]>
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 | */ | |
21 | ||
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. |
7011fb60 | 25 | * Copyright 2013 Nexenta Systems, Inc. All rights reserved. |
a38718a6 | 26 | */ |
34dc7c2f | 27 | |
34dc7c2f | 28 | /* |
e49f1e20 WA |
29 | * SPA: Storage Pool Allocator |
30 | * | |
34dc7c2f BB |
31 | * This file contains all the routines used when modifying on-disk SPA state. |
32 | * This includes opening, importing, destroying, exporting a pool, and syncing a | |
33 | * pool. | |
34 | */ | |
35 | ||
36 | #include <sys/zfs_context.h> | |
37 | #include <sys/fm/fs/zfs.h> | |
38 | #include <sys/spa_impl.h> | |
39 | #include <sys/zio.h> | |
40 | #include <sys/zio_checksum.h> | |
34dc7c2f BB |
41 | #include <sys/dmu.h> |
42 | #include <sys/dmu_tx.h> | |
43 | #include <sys/zap.h> | |
44 | #include <sys/zil.h> | |
428870ff | 45 | #include <sys/ddt.h> |
34dc7c2f | 46 | #include <sys/vdev_impl.h> |
c28b2279 | 47 | #include <sys/vdev_disk.h> |
34dc7c2f | 48 | #include <sys/metaslab.h> |
428870ff | 49 | #include <sys/metaslab_impl.h> |
34dc7c2f BB |
50 | #include <sys/uberblock_impl.h> |
51 | #include <sys/txg.h> | |
52 | #include <sys/avl.h> | |
53 | #include <sys/dmu_traverse.h> | |
54 | #include <sys/dmu_objset.h> | |
55 | #include <sys/unique.h> | |
56 | #include <sys/dsl_pool.h> | |
57 | #include <sys/dsl_dataset.h> | |
58 | #include <sys/dsl_dir.h> | |
59 | #include <sys/dsl_prop.h> | |
60 | #include <sys/dsl_synctask.h> | |
61 | #include <sys/fs/zfs.h> | |
62 | #include <sys/arc.h> | |
63 | #include <sys/callb.h> | |
64 | #include <sys/systeminfo.h> | |
34dc7c2f | 65 | #include <sys/spa_boot.h> |
9babb374 | 66 | #include <sys/zfs_ioctl.h> |
428870ff | 67 | #include <sys/dsl_scan.h> |
9ae529ec | 68 | #include <sys/zfeature.h> |
13fe0198 | 69 | #include <sys/dsl_destroy.h> |
526af785 | 70 | #include <sys/zvol.h> |
34dc7c2f | 71 | |
d164b209 | 72 | #ifdef _KERNEL |
428870ff BB |
73 | #include <sys/bootprops.h> |
74 | #include <sys/callb.h> | |
75 | #include <sys/cpupart.h> | |
76 | #include <sys/pool.h> | |
77 | #include <sys/sysdc.h> | |
d164b209 BB |
78 | #include <sys/zone.h> |
79 | #endif /* _KERNEL */ | |
80 | ||
34dc7c2f BB |
81 | #include "zfs_prop.h" |
82 | #include "zfs_comutil.h" | |
83 | ||
428870ff | 84 | typedef enum zti_modes { |
7ef5e54e AL |
85 | ZTI_MODE_FIXED, /* value is # of threads (min 1) */ |
86 | ZTI_MODE_ONLINE_PERCENT, /* value is % of online CPUs */ | |
87 | ZTI_MODE_BATCH, /* cpu-intensive; value is ignored */ | |
88 | ZTI_MODE_NULL, /* don't create a taskq */ | |
89 | ZTI_NMODES | |
428870ff | 90 | } zti_modes_t; |
34dc7c2f | 91 | |
7ef5e54e AL |
92 | #define ZTI_P(n, q) { ZTI_MODE_FIXED, (n), (q) } |
93 | #define ZTI_PCT(n) { ZTI_MODE_ONLINE_PERCENT, (n), 1 } | |
94 | #define ZTI_BATCH { ZTI_MODE_BATCH, 0, 1 } | |
95 | #define ZTI_NULL { ZTI_MODE_NULL, 0, 0 } | |
9babb374 | 96 | |
7ef5e54e AL |
97 | #define ZTI_N(n) ZTI_P(n, 1) |
98 | #define ZTI_ONE ZTI_N(1) | |
9babb374 BB |
99 | |
100 | typedef struct zio_taskq_info { | |
7ef5e54e | 101 | zti_modes_t zti_mode; |
428870ff | 102 | uint_t zti_value; |
7ef5e54e | 103 | uint_t zti_count; |
9babb374 BB |
104 | } zio_taskq_info_t; |
105 | ||
106 | static const char *const zio_taskq_types[ZIO_TASKQ_TYPES] = { | |
451041db | 107 | "iss", "iss_h", "int", "int_h" |
9babb374 BB |
108 | }; |
109 | ||
428870ff | 110 | /* |
7ef5e54e AL |
111 | * This table defines the taskq settings for each ZFS I/O type. When |
112 | * initializing a pool, we use this table to create an appropriately sized | |
113 | * taskq. Some operations are low volume and therefore have a small, static | |
114 | * number of threads assigned to their taskqs using the ZTI_N(#) or ZTI_ONE | |
115 | * macros. Other operations process a large amount of data; the ZTI_BATCH | |
116 | * macro causes us to create a taskq oriented for throughput. Some operations | |
117 | * are so high frequency and short-lived that the taskq itself can become a a | |
118 | * point of lock contention. The ZTI_P(#, #) macro indicates that we need an | |
119 | * additional degree of parallelism specified by the number of threads per- | |
120 | * taskq and the number of taskqs; when dispatching an event in this case, the | |
121 | * particular taskq is chosen at random. | |
122 | * | |
123 | * The different taskq priorities are to handle the different contexts (issue | |
124 | * and interrupt) and then to reserve threads for ZIO_PRIORITY_NOW I/Os that | |
125 | * need to be handled with minimum delay. | |
428870ff BB |
126 | */ |
127 | const zio_taskq_info_t zio_taskqs[ZIO_TYPES][ZIO_TASKQ_TYPES] = { | |
128 | /* ISSUE ISSUE_HIGH INTR INTR_HIGH */ | |
7ef5e54e AL |
129 | { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* NULL */ |
130 | { ZTI_N(8), ZTI_NULL, ZTI_BATCH, ZTI_NULL }, /* READ */ | |
131 | { ZTI_BATCH, ZTI_N(5), ZTI_N(16), ZTI_N(5) }, /* WRITE */ | |
132 | { ZTI_P(4, 8), ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* FREE */ | |
133 | { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* CLAIM */ | |
134 | { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* IOCTL */ | |
9babb374 BB |
135 | }; |
136 | ||
13fe0198 MA |
137 | static void spa_sync_version(void *arg, dmu_tx_t *tx); |
138 | static void spa_sync_props(void *arg, dmu_tx_t *tx); | |
b128c09f | 139 | static boolean_t spa_has_active_shared_spare(spa_t *spa); |
bf701a83 | 140 | static inline int spa_load_impl(spa_t *spa, uint64_t, nvlist_t *config, |
428870ff BB |
141 | spa_load_state_t state, spa_import_type_t type, boolean_t mosconfig, |
142 | char **ereport); | |
572e2857 | 143 | static void spa_vdev_resilver_done(spa_t *spa); |
428870ff BB |
144 | |
145 | uint_t zio_taskq_batch_pct = 100; /* 1 thread per cpu in pset */ | |
146 | id_t zio_taskq_psrset_bind = PS_NONE; | |
147 | boolean_t zio_taskq_sysdc = B_TRUE; /* use SDC scheduling class */ | |
148 | uint_t zio_taskq_basedc = 80; /* base duty cycle */ | |
149 | ||
150 | boolean_t spa_create_process = B_TRUE; /* no process ==> no sysdc */ | |
151 | ||
152 | /* | |
153 | * This (illegal) pool name is used when temporarily importing a spa_t in order | |
154 | * to get the vdev stats associated with the imported devices. | |
155 | */ | |
156 | #define TRYIMPORT_NAME "$import" | |
34dc7c2f BB |
157 | |
158 | /* | |
159 | * ========================================================================== | |
160 | * SPA properties routines | |
161 | * ========================================================================== | |
162 | */ | |
163 | ||
164 | /* | |
165 | * Add a (source=src, propname=propval) list to an nvlist. | |
166 | */ | |
167 | static void | |
168 | spa_prop_add_list(nvlist_t *nvl, zpool_prop_t prop, char *strval, | |
169 | uint64_t intval, zprop_source_t src) | |
170 | { | |
171 | const char *propname = zpool_prop_to_name(prop); | |
172 | nvlist_t *propval; | |
173 | ||
b8d06fca | 174 | VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_PUSHPAGE) == 0); |
34dc7c2f BB |
175 | VERIFY(nvlist_add_uint64(propval, ZPROP_SOURCE, src) == 0); |
176 | ||
177 | if (strval != NULL) | |
178 | VERIFY(nvlist_add_string(propval, ZPROP_VALUE, strval) == 0); | |
179 | else | |
180 | VERIFY(nvlist_add_uint64(propval, ZPROP_VALUE, intval) == 0); | |
181 | ||
182 | VERIFY(nvlist_add_nvlist(nvl, propname, propval) == 0); | |
183 | nvlist_free(propval); | |
184 | } | |
185 | ||
186 | /* | |
187 | * Get property values from the spa configuration. | |
188 | */ | |
189 | static void | |
190 | spa_prop_get_config(spa_t *spa, nvlist_t **nvp) | |
191 | { | |
1bd201e7 | 192 | vdev_t *rvd = spa->spa_root_vdev; |
9ae529ec | 193 | dsl_pool_t *pool = spa->spa_dsl_pool; |
d164b209 | 194 | uint64_t size; |
428870ff | 195 | uint64_t alloc; |
1bd201e7 | 196 | uint64_t space; |
34dc7c2f BB |
197 | uint64_t cap, version; |
198 | zprop_source_t src = ZPROP_SRC_NONE; | |
b128c09f | 199 | spa_config_dirent_t *dp; |
1bd201e7 | 200 | int c; |
b128c09f BB |
201 | |
202 | ASSERT(MUTEX_HELD(&spa->spa_props_lock)); | |
34dc7c2f | 203 | |
1bd201e7 | 204 | if (rvd != NULL) { |
428870ff BB |
205 | alloc = metaslab_class_get_alloc(spa_normal_class(spa)); |
206 | size = metaslab_class_get_space(spa_normal_class(spa)); | |
d164b209 BB |
207 | spa_prop_add_list(*nvp, ZPOOL_PROP_NAME, spa_name(spa), 0, src); |
208 | spa_prop_add_list(*nvp, ZPOOL_PROP_SIZE, NULL, size, src); | |
428870ff BB |
209 | spa_prop_add_list(*nvp, ZPOOL_PROP_ALLOCATED, NULL, alloc, src); |
210 | spa_prop_add_list(*nvp, ZPOOL_PROP_FREE, NULL, | |
211 | size - alloc, src); | |
1bd201e7 CS |
212 | |
213 | space = 0; | |
214 | for (c = 0; c < rvd->vdev_children; c++) { | |
215 | vdev_t *tvd = rvd->vdev_child[c]; | |
216 | space += tvd->vdev_max_asize - tvd->vdev_asize; | |
217 | } | |
218 | spa_prop_add_list(*nvp, ZPOOL_PROP_EXPANDSZ, NULL, space, | |
219 | src); | |
220 | ||
572e2857 BB |
221 | spa_prop_add_list(*nvp, ZPOOL_PROP_READONLY, NULL, |
222 | (spa_mode(spa) == FREAD), src); | |
d164b209 | 223 | |
428870ff | 224 | cap = (size == 0) ? 0 : (alloc * 100 / size); |
d164b209 BB |
225 | spa_prop_add_list(*nvp, ZPOOL_PROP_CAPACITY, NULL, cap, src); |
226 | ||
428870ff BB |
227 | spa_prop_add_list(*nvp, ZPOOL_PROP_DEDUPRATIO, NULL, |
228 | ddt_get_pool_dedup_ratio(spa), src); | |
229 | ||
d164b209 | 230 | spa_prop_add_list(*nvp, ZPOOL_PROP_HEALTH, NULL, |
1bd201e7 | 231 | rvd->vdev_state, src); |
d164b209 BB |
232 | |
233 | version = spa_version(spa); | |
234 | if (version == zpool_prop_default_numeric(ZPOOL_PROP_VERSION)) | |
235 | src = ZPROP_SRC_DEFAULT; | |
236 | else | |
237 | src = ZPROP_SRC_LOCAL; | |
238 | spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL, version, src); | |
239 | } | |
34dc7c2f | 240 | |
9ae529ec CS |
241 | if (pool != NULL) { |
242 | dsl_dir_t *freedir = pool->dp_free_dir; | |
243 | ||
244 | /* | |
245 | * The $FREE directory was introduced in SPA_VERSION_DEADLISTS, | |
246 | * when opening pools before this version freedir will be NULL. | |
247 | */ | |
248 | if (freedir != NULL) { | |
249 | spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING, NULL, | |
250 | freedir->dd_phys->dd_used_bytes, src); | |
251 | } else { | |
252 | spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING, | |
253 | NULL, 0, src); | |
254 | } | |
255 | } | |
256 | ||
34dc7c2f | 257 | spa_prop_add_list(*nvp, ZPOOL_PROP_GUID, NULL, spa_guid(spa), src); |
34dc7c2f | 258 | |
d96eb2b1 DM |
259 | if (spa->spa_comment != NULL) { |
260 | spa_prop_add_list(*nvp, ZPOOL_PROP_COMMENT, spa->spa_comment, | |
261 | 0, ZPROP_SRC_LOCAL); | |
262 | } | |
263 | ||
34dc7c2f BB |
264 | if (spa->spa_root != NULL) |
265 | spa_prop_add_list(*nvp, ZPOOL_PROP_ALTROOT, spa->spa_root, | |
266 | 0, ZPROP_SRC_LOCAL); | |
267 | ||
b128c09f BB |
268 | if ((dp = list_head(&spa->spa_config_list)) != NULL) { |
269 | if (dp->scd_path == NULL) { | |
34dc7c2f | 270 | spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE, |
b128c09f BB |
271 | "none", 0, ZPROP_SRC_LOCAL); |
272 | } else if (strcmp(dp->scd_path, spa_config_path) != 0) { | |
34dc7c2f | 273 | spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE, |
b128c09f | 274 | dp->scd_path, 0, ZPROP_SRC_LOCAL); |
34dc7c2f BB |
275 | } |
276 | } | |
277 | } | |
278 | ||
279 | /* | |
280 | * Get zpool property values. | |
281 | */ | |
282 | int | |
283 | spa_prop_get(spa_t *spa, nvlist_t **nvp) | |
284 | { | |
428870ff | 285 | objset_t *mos = spa->spa_meta_objset; |
34dc7c2f BB |
286 | zap_cursor_t zc; |
287 | zap_attribute_t za; | |
34dc7c2f BB |
288 | int err; |
289 | ||
b8d06fca | 290 | err = nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_PUSHPAGE); |
c28b2279 BB |
291 | if (err) |
292 | return err; | |
34dc7c2f | 293 | |
b128c09f BB |
294 | mutex_enter(&spa->spa_props_lock); |
295 | ||
34dc7c2f BB |
296 | /* |
297 | * Get properties from the spa config. | |
298 | */ | |
299 | spa_prop_get_config(spa, nvp); | |
300 | ||
34dc7c2f | 301 | /* If no pool property object, no more prop to get. */ |
428870ff | 302 | if (mos == NULL || spa->spa_pool_props_object == 0) { |
34dc7c2f | 303 | mutex_exit(&spa->spa_props_lock); |
c28b2279 | 304 | goto out; |
34dc7c2f BB |
305 | } |
306 | ||
307 | /* | |
308 | * Get properties from the MOS pool property object. | |
309 | */ | |
310 | for (zap_cursor_init(&zc, mos, spa->spa_pool_props_object); | |
311 | (err = zap_cursor_retrieve(&zc, &za)) == 0; | |
312 | zap_cursor_advance(&zc)) { | |
313 | uint64_t intval = 0; | |
314 | char *strval = NULL; | |
315 | zprop_source_t src = ZPROP_SRC_DEFAULT; | |
316 | zpool_prop_t prop; | |
317 | ||
318 | if ((prop = zpool_name_to_prop(za.za_name)) == ZPROP_INVAL) | |
319 | continue; | |
320 | ||
321 | switch (za.za_integer_length) { | |
322 | case 8: | |
323 | /* integer property */ | |
324 | if (za.za_first_integer != | |
325 | zpool_prop_default_numeric(prop)) | |
326 | src = ZPROP_SRC_LOCAL; | |
327 | ||
328 | if (prop == ZPOOL_PROP_BOOTFS) { | |
329 | dsl_pool_t *dp; | |
330 | dsl_dataset_t *ds = NULL; | |
331 | ||
332 | dp = spa_get_dsl(spa); | |
13fe0198 | 333 | dsl_pool_config_enter(dp, FTAG); |
c65aa5b2 BB |
334 | if ((err = dsl_dataset_hold_obj(dp, |
335 | za.za_first_integer, FTAG, &ds))) { | |
13fe0198 | 336 | dsl_pool_config_exit(dp, FTAG); |
34dc7c2f BB |
337 | break; |
338 | } | |
339 | ||
340 | strval = kmem_alloc( | |
341 | MAXNAMELEN + strlen(MOS_DIR_NAME) + 1, | |
b8d06fca | 342 | KM_PUSHPAGE); |
34dc7c2f | 343 | dsl_dataset_name(ds, strval); |
b128c09f | 344 | dsl_dataset_rele(ds, FTAG); |
13fe0198 | 345 | dsl_pool_config_exit(dp, FTAG); |
34dc7c2f BB |
346 | } else { |
347 | strval = NULL; | |
348 | intval = za.za_first_integer; | |
349 | } | |
350 | ||
351 | spa_prop_add_list(*nvp, prop, strval, intval, src); | |
352 | ||
353 | if (strval != NULL) | |
354 | kmem_free(strval, | |
355 | MAXNAMELEN + strlen(MOS_DIR_NAME) + 1); | |
356 | ||
357 | break; | |
358 | ||
359 | case 1: | |
360 | /* string property */ | |
b8d06fca | 361 | strval = kmem_alloc(za.za_num_integers, KM_PUSHPAGE); |
34dc7c2f BB |
362 | err = zap_lookup(mos, spa->spa_pool_props_object, |
363 | za.za_name, 1, za.za_num_integers, strval); | |
364 | if (err) { | |
365 | kmem_free(strval, za.za_num_integers); | |
366 | break; | |
367 | } | |
368 | spa_prop_add_list(*nvp, prop, strval, 0, src); | |
369 | kmem_free(strval, za.za_num_integers); | |
370 | break; | |
371 | ||
372 | default: | |
373 | break; | |
374 | } | |
375 | } | |
376 | zap_cursor_fini(&zc); | |
377 | mutex_exit(&spa->spa_props_lock); | |
378 | out: | |
379 | if (err && err != ENOENT) { | |
380 | nvlist_free(*nvp); | |
381 | *nvp = NULL; | |
382 | return (err); | |
383 | } | |
384 | ||
385 | return (0); | |
386 | } | |
387 | ||
388 | /* | |
389 | * Validate the given pool properties nvlist and modify the list | |
390 | * for the property values to be set. | |
391 | */ | |
392 | static int | |
393 | spa_prop_validate(spa_t *spa, nvlist_t *props) | |
394 | { | |
395 | nvpair_t *elem; | |
396 | int error = 0, reset_bootfs = 0; | |
d4ed6673 | 397 | uint64_t objnum = 0; |
9ae529ec | 398 | boolean_t has_feature = B_FALSE; |
34dc7c2f BB |
399 | |
400 | elem = NULL; | |
401 | while ((elem = nvlist_next_nvpair(props, elem)) != NULL) { | |
34dc7c2f | 402 | uint64_t intval; |
9ae529ec CS |
403 | char *strval, *slash, *check, *fname; |
404 | const char *propname = nvpair_name(elem); | |
405 | zpool_prop_t prop = zpool_name_to_prop(propname); | |
406 | ||
407 | switch ((int)prop) { | |
408 | case ZPROP_INVAL: | |
409 | if (!zpool_prop_feature(propname)) { | |
2e528b49 | 410 | error = SET_ERROR(EINVAL); |
9ae529ec CS |
411 | break; |
412 | } | |
413 | ||
414 | /* | |
415 | * Sanitize the input. | |
416 | */ | |
417 | if (nvpair_type(elem) != DATA_TYPE_UINT64) { | |
2e528b49 | 418 | error = SET_ERROR(EINVAL); |
9ae529ec CS |
419 | break; |
420 | } | |
421 | ||
422 | if (nvpair_value_uint64(elem, &intval) != 0) { | |
2e528b49 | 423 | error = SET_ERROR(EINVAL); |
9ae529ec CS |
424 | break; |
425 | } | |
34dc7c2f | 426 | |
9ae529ec | 427 | if (intval != 0) { |
2e528b49 | 428 | error = SET_ERROR(EINVAL); |
9ae529ec CS |
429 | break; |
430 | } | |
34dc7c2f | 431 | |
9ae529ec CS |
432 | fname = strchr(propname, '@') + 1; |
433 | if (zfeature_lookup_name(fname, NULL) != 0) { | |
2e528b49 | 434 | error = SET_ERROR(EINVAL); |
9ae529ec CS |
435 | break; |
436 | } | |
437 | ||
438 | has_feature = B_TRUE; | |
439 | break; | |
34dc7c2f | 440 | |
34dc7c2f BB |
441 | case ZPOOL_PROP_VERSION: |
442 | error = nvpair_value_uint64(elem, &intval); | |
443 | if (!error && | |
9ae529ec CS |
444 | (intval < spa_version(spa) || |
445 | intval > SPA_VERSION_BEFORE_FEATURES || | |
446 | has_feature)) | |
2e528b49 | 447 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
448 | break; |
449 | ||
450 | case ZPOOL_PROP_DELEGATION: | |
451 | case ZPOOL_PROP_AUTOREPLACE: | |
b128c09f | 452 | case ZPOOL_PROP_LISTSNAPS: |
9babb374 | 453 | case ZPOOL_PROP_AUTOEXPAND: |
34dc7c2f BB |
454 | error = nvpair_value_uint64(elem, &intval); |
455 | if (!error && intval > 1) | |
2e528b49 | 456 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
457 | break; |
458 | ||
459 | case ZPOOL_PROP_BOOTFS: | |
9babb374 BB |
460 | /* |
461 | * If the pool version is less than SPA_VERSION_BOOTFS, | |
462 | * or the pool is still being created (version == 0), | |
463 | * the bootfs property cannot be set. | |
464 | */ | |
34dc7c2f | 465 | if (spa_version(spa) < SPA_VERSION_BOOTFS) { |
2e528b49 | 466 | error = SET_ERROR(ENOTSUP); |
34dc7c2f BB |
467 | break; |
468 | } | |
469 | ||
470 | /* | |
b128c09f | 471 | * Make sure the vdev config is bootable |
34dc7c2f | 472 | */ |
b128c09f | 473 | if (!vdev_is_bootable(spa->spa_root_vdev)) { |
2e528b49 | 474 | error = SET_ERROR(ENOTSUP); |
34dc7c2f BB |
475 | break; |
476 | } | |
477 | ||
478 | reset_bootfs = 1; | |
479 | ||
480 | error = nvpair_value_string(elem, &strval); | |
481 | ||
482 | if (!error) { | |
9ae529ec | 483 | objset_t *os; |
b128c09f BB |
484 | uint64_t compress; |
485 | ||
34dc7c2f BB |
486 | if (strval == NULL || strval[0] == '\0') { |
487 | objnum = zpool_prop_default_numeric( | |
488 | ZPOOL_PROP_BOOTFS); | |
489 | break; | |
490 | } | |
491 | ||
c65aa5b2 | 492 | if ((error = dmu_objset_hold(strval,FTAG,&os))) |
34dc7c2f | 493 | break; |
b128c09f | 494 | |
428870ff BB |
495 | /* Must be ZPL and not gzip compressed. */ |
496 | ||
497 | if (dmu_objset_type(os) != DMU_OST_ZFS) { | |
2e528b49 | 498 | error = SET_ERROR(ENOTSUP); |
13fe0198 MA |
499 | } else if ((error = |
500 | dsl_prop_get_int_ds(dmu_objset_ds(os), | |
b128c09f | 501 | zfs_prop_to_name(ZFS_PROP_COMPRESSION), |
13fe0198 | 502 | &compress)) == 0 && |
b128c09f | 503 | !BOOTFS_COMPRESS_VALID(compress)) { |
2e528b49 | 504 | error = SET_ERROR(ENOTSUP); |
b128c09f BB |
505 | } else { |
506 | objnum = dmu_objset_id(os); | |
507 | } | |
428870ff | 508 | dmu_objset_rele(os, FTAG); |
34dc7c2f BB |
509 | } |
510 | break; | |
b128c09f | 511 | |
34dc7c2f BB |
512 | case ZPOOL_PROP_FAILUREMODE: |
513 | error = nvpair_value_uint64(elem, &intval); | |
514 | if (!error && (intval < ZIO_FAILURE_MODE_WAIT || | |
515 | intval > ZIO_FAILURE_MODE_PANIC)) | |
2e528b49 | 516 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
517 | |
518 | /* | |
519 | * This is a special case which only occurs when | |
520 | * the pool has completely failed. This allows | |
521 | * the user to change the in-core failmode property | |
522 | * without syncing it out to disk (I/Os might | |
523 | * currently be blocked). We do this by returning | |
524 | * EIO to the caller (spa_prop_set) to trick it | |
525 | * into thinking we encountered a property validation | |
526 | * error. | |
527 | */ | |
b128c09f | 528 | if (!error && spa_suspended(spa)) { |
34dc7c2f | 529 | spa->spa_failmode = intval; |
2e528b49 | 530 | error = SET_ERROR(EIO); |
34dc7c2f BB |
531 | } |
532 | break; | |
533 | ||
534 | case ZPOOL_PROP_CACHEFILE: | |
535 | if ((error = nvpair_value_string(elem, &strval)) != 0) | |
536 | break; | |
537 | ||
538 | if (strval[0] == '\0') | |
539 | break; | |
540 | ||
541 | if (strcmp(strval, "none") == 0) | |
542 | break; | |
543 | ||
544 | if (strval[0] != '/') { | |
2e528b49 | 545 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
546 | break; |
547 | } | |
548 | ||
549 | slash = strrchr(strval, '/'); | |
550 | ASSERT(slash != NULL); | |
551 | ||
552 | if (slash[1] == '\0' || strcmp(slash, "/.") == 0 || | |
553 | strcmp(slash, "/..") == 0) | |
2e528b49 | 554 | error = SET_ERROR(EINVAL); |
34dc7c2f | 555 | break; |
428870ff | 556 | |
d96eb2b1 DM |
557 | case ZPOOL_PROP_COMMENT: |
558 | if ((error = nvpair_value_string(elem, &strval)) != 0) | |
559 | break; | |
560 | for (check = strval; *check != '\0'; check++) { | |
561 | if (!isprint(*check)) { | |
2e528b49 | 562 | error = SET_ERROR(EINVAL); |
d96eb2b1 DM |
563 | break; |
564 | } | |
565 | check++; | |
566 | } | |
567 | if (strlen(strval) > ZPROP_MAX_COMMENT) | |
2e528b49 | 568 | error = SET_ERROR(E2BIG); |
d96eb2b1 DM |
569 | break; |
570 | ||
428870ff BB |
571 | case ZPOOL_PROP_DEDUPDITTO: |
572 | if (spa_version(spa) < SPA_VERSION_DEDUP) | |
2e528b49 | 573 | error = SET_ERROR(ENOTSUP); |
428870ff BB |
574 | else |
575 | error = nvpair_value_uint64(elem, &intval); | |
576 | if (error == 0 && | |
577 | intval != 0 && intval < ZIO_DEDUPDITTO_MIN) | |
2e528b49 | 578 | error = SET_ERROR(EINVAL); |
428870ff | 579 | break; |
e75c13c3 BB |
580 | |
581 | default: | |
582 | break; | |
34dc7c2f BB |
583 | } |
584 | ||
585 | if (error) | |
586 | break; | |
587 | } | |
588 | ||
589 | if (!error && reset_bootfs) { | |
590 | error = nvlist_remove(props, | |
591 | zpool_prop_to_name(ZPOOL_PROP_BOOTFS), DATA_TYPE_STRING); | |
592 | ||
593 | if (!error) { | |
594 | error = nvlist_add_uint64(props, | |
595 | zpool_prop_to_name(ZPOOL_PROP_BOOTFS), objnum); | |
596 | } | |
597 | } | |
598 | ||
599 | return (error); | |
600 | } | |
601 | ||
d164b209 BB |
602 | void |
603 | spa_configfile_set(spa_t *spa, nvlist_t *nvp, boolean_t need_sync) | |
604 | { | |
605 | char *cachefile; | |
606 | spa_config_dirent_t *dp; | |
607 | ||
608 | if (nvlist_lookup_string(nvp, zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), | |
609 | &cachefile) != 0) | |
610 | return; | |
611 | ||
612 | dp = kmem_alloc(sizeof (spa_config_dirent_t), | |
b8d06fca | 613 | KM_PUSHPAGE); |
d164b209 BB |
614 | |
615 | if (cachefile[0] == '\0') | |
616 | dp->scd_path = spa_strdup(spa_config_path); | |
617 | else if (strcmp(cachefile, "none") == 0) | |
618 | dp->scd_path = NULL; | |
619 | else | |
620 | dp->scd_path = spa_strdup(cachefile); | |
621 | ||
622 | list_insert_head(&spa->spa_config_list, dp); | |
623 | if (need_sync) | |
624 | spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE); | |
625 | } | |
626 | ||
34dc7c2f BB |
627 | int |
628 | spa_prop_set(spa_t *spa, nvlist_t *nvp) | |
629 | { | |
630 | int error; | |
9ae529ec | 631 | nvpair_t *elem = NULL; |
d164b209 | 632 | boolean_t need_sync = B_FALSE; |
34dc7c2f BB |
633 | |
634 | if ((error = spa_prop_validate(spa, nvp)) != 0) | |
635 | return (error); | |
636 | ||
d164b209 | 637 | while ((elem = nvlist_next_nvpair(nvp, elem)) != NULL) { |
9ae529ec | 638 | zpool_prop_t prop = zpool_name_to_prop(nvpair_name(elem)); |
d164b209 | 639 | |
572e2857 BB |
640 | if (prop == ZPOOL_PROP_CACHEFILE || |
641 | prop == ZPOOL_PROP_ALTROOT || | |
642 | prop == ZPOOL_PROP_READONLY) | |
d164b209 BB |
643 | continue; |
644 | ||
9ae529ec CS |
645 | if (prop == ZPOOL_PROP_VERSION || prop == ZPROP_INVAL) { |
646 | uint64_t ver; | |
647 | ||
648 | if (prop == ZPOOL_PROP_VERSION) { | |
649 | VERIFY(nvpair_value_uint64(elem, &ver) == 0); | |
650 | } else { | |
651 | ASSERT(zpool_prop_feature(nvpair_name(elem))); | |
652 | ver = SPA_VERSION_FEATURES; | |
653 | need_sync = B_TRUE; | |
654 | } | |
655 | ||
656 | /* Save time if the version is already set. */ | |
657 | if (ver == spa_version(spa)) | |
658 | continue; | |
659 | ||
660 | /* | |
661 | * In addition to the pool directory object, we might | |
662 | * create the pool properties object, the features for | |
663 | * read object, the features for write object, or the | |
664 | * feature descriptions object. | |
665 | */ | |
13fe0198 MA |
666 | error = dsl_sync_task(spa->spa_name, NULL, |
667 | spa_sync_version, &ver, 6); | |
9ae529ec CS |
668 | if (error) |
669 | return (error); | |
670 | continue; | |
671 | } | |
672 | ||
d164b209 BB |
673 | need_sync = B_TRUE; |
674 | break; | |
675 | } | |
676 | ||
9ae529ec | 677 | if (need_sync) { |
13fe0198 MA |
678 | return (dsl_sync_task(spa->spa_name, NULL, spa_sync_props, |
679 | nvp, 6)); | |
9ae529ec CS |
680 | } |
681 | ||
682 | return (0); | |
34dc7c2f BB |
683 | } |
684 | ||
685 | /* | |
686 | * If the bootfs property value is dsobj, clear it. | |
687 | */ | |
688 | void | |
689 | spa_prop_clear_bootfs(spa_t *spa, uint64_t dsobj, dmu_tx_t *tx) | |
690 | { | |
691 | if (spa->spa_bootfs == dsobj && spa->spa_pool_props_object != 0) { | |
692 | VERIFY(zap_remove(spa->spa_meta_objset, | |
693 | spa->spa_pool_props_object, | |
694 | zpool_prop_to_name(ZPOOL_PROP_BOOTFS), tx) == 0); | |
695 | spa->spa_bootfs = 0; | |
696 | } | |
697 | } | |
698 | ||
3bc7e0fb GW |
699 | /*ARGSUSED*/ |
700 | static int | |
13fe0198 | 701 | spa_change_guid_check(void *arg, dmu_tx_t *tx) |
3bc7e0fb | 702 | { |
13fe0198 | 703 | spa_t *spa = dmu_tx_pool(tx)->dp_spa; |
3bc7e0fb GW |
704 | vdev_t *rvd = spa->spa_root_vdev; |
705 | uint64_t vdev_state; | |
13fe0198 | 706 | ASSERTV(uint64_t *newguid = arg); |
3bc7e0fb GW |
707 | |
708 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); | |
709 | vdev_state = rvd->vdev_state; | |
710 | spa_config_exit(spa, SCL_STATE, FTAG); | |
711 | ||
712 | if (vdev_state != VDEV_STATE_HEALTHY) | |
2e528b49 | 713 | return (SET_ERROR(ENXIO)); |
3bc7e0fb GW |
714 | |
715 | ASSERT3U(spa_guid(spa), !=, *newguid); | |
716 | ||
717 | return (0); | |
718 | } | |
719 | ||
720 | static void | |
13fe0198 | 721 | spa_change_guid_sync(void *arg, dmu_tx_t *tx) |
3bc7e0fb | 722 | { |
13fe0198 MA |
723 | uint64_t *newguid = arg; |
724 | spa_t *spa = dmu_tx_pool(tx)->dp_spa; | |
3bc7e0fb GW |
725 | uint64_t oldguid; |
726 | vdev_t *rvd = spa->spa_root_vdev; | |
727 | ||
728 | oldguid = spa_guid(spa); | |
729 | ||
730 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); | |
731 | rvd->vdev_guid = *newguid; | |
732 | rvd->vdev_guid_sum += (*newguid - oldguid); | |
733 | vdev_config_dirty(rvd); | |
734 | spa_config_exit(spa, SCL_STATE, FTAG); | |
735 | ||
6f1ffb06 MA |
736 | spa_history_log_internal(spa, "guid change", tx, "old=%llu new=%llu", |
737 | oldguid, *newguid); | |
3bc7e0fb GW |
738 | } |
739 | ||
3541dc6d GA |
740 | /* |
741 | * Change the GUID for the pool. This is done so that we can later | |
742 | * re-import a pool built from a clone of our own vdevs. We will modify | |
743 | * the root vdev's guid, our own pool guid, and then mark all of our | |
744 | * vdevs dirty. Note that we must make sure that all our vdevs are | |
745 | * online when we do this, or else any vdevs that weren't present | |
746 | * would be orphaned from our pool. We are also going to issue a | |
747 | * sysevent to update any watchers. | |
748 | */ | |
749 | int | |
750 | spa_change_guid(spa_t *spa) | |
751 | { | |
3bc7e0fb GW |
752 | int error; |
753 | uint64_t guid; | |
3541dc6d | 754 | |
621dd7bb | 755 | mutex_enter(&spa->spa_vdev_top_lock); |
3bc7e0fb GW |
756 | mutex_enter(&spa_namespace_lock); |
757 | guid = spa_generate_guid(NULL); | |
3541dc6d | 758 | |
13fe0198 MA |
759 | error = dsl_sync_task(spa->spa_name, spa_change_guid_check, |
760 | spa_change_guid_sync, &guid, 5); | |
3541dc6d | 761 | |
3bc7e0fb GW |
762 | if (error == 0) { |
763 | spa_config_sync(spa, B_FALSE, B_TRUE); | |
764 | spa_event_notify(spa, NULL, FM_EREPORT_ZFS_POOL_REGUID); | |
765 | } | |
3541dc6d | 766 | |
3bc7e0fb | 767 | mutex_exit(&spa_namespace_lock); |
621dd7bb | 768 | mutex_exit(&spa->spa_vdev_top_lock); |
3541dc6d | 769 | |
3bc7e0fb | 770 | return (error); |
3541dc6d GA |
771 | } |
772 | ||
34dc7c2f BB |
773 | /* |
774 | * ========================================================================== | |
775 | * SPA state manipulation (open/create/destroy/import/export) | |
776 | * ========================================================================== | |
777 | */ | |
778 | ||
779 | static int | |
780 | spa_error_entry_compare(const void *a, const void *b) | |
781 | { | |
782 | spa_error_entry_t *sa = (spa_error_entry_t *)a; | |
783 | spa_error_entry_t *sb = (spa_error_entry_t *)b; | |
784 | int ret; | |
785 | ||
786 | ret = bcmp(&sa->se_bookmark, &sb->se_bookmark, | |
787 | sizeof (zbookmark_t)); | |
788 | ||
789 | if (ret < 0) | |
790 | return (-1); | |
791 | else if (ret > 0) | |
792 | return (1); | |
793 | else | |
794 | return (0); | |
795 | } | |
796 | ||
797 | /* | |
798 | * Utility function which retrieves copies of the current logs and | |
799 | * re-initializes them in the process. | |
800 | */ | |
801 | void | |
802 | spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub) | |
803 | { | |
804 | ASSERT(MUTEX_HELD(&spa->spa_errlist_lock)); | |
805 | ||
806 | bcopy(&spa->spa_errlist_last, last, sizeof (avl_tree_t)); | |
807 | bcopy(&spa->spa_errlist_scrub, scrub, sizeof (avl_tree_t)); | |
808 | ||
809 | avl_create(&spa->spa_errlist_scrub, | |
810 | spa_error_entry_compare, sizeof (spa_error_entry_t), | |
811 | offsetof(spa_error_entry_t, se_avl)); | |
812 | avl_create(&spa->spa_errlist_last, | |
813 | spa_error_entry_compare, sizeof (spa_error_entry_t), | |
814 | offsetof(spa_error_entry_t, se_avl)); | |
815 | } | |
816 | ||
7ef5e54e AL |
817 | static void |
818 | spa_taskqs_init(spa_t *spa, zio_type_t t, zio_taskq_type_t q) | |
34dc7c2f | 819 | { |
7ef5e54e AL |
820 | const zio_taskq_info_t *ztip = &zio_taskqs[t][q]; |
821 | enum zti_modes mode = ztip->zti_mode; | |
822 | uint_t value = ztip->zti_value; | |
823 | uint_t count = ztip->zti_count; | |
824 | spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q]; | |
825 | char name[32]; | |
826 | uint_t i, flags = 0; | |
428870ff | 827 | boolean_t batch = B_FALSE; |
34dc7c2f | 828 | |
7ef5e54e AL |
829 | if (mode == ZTI_MODE_NULL) { |
830 | tqs->stqs_count = 0; | |
831 | tqs->stqs_taskq = NULL; | |
832 | return; | |
833 | } | |
428870ff | 834 | |
7ef5e54e | 835 | ASSERT3U(count, >, 0); |
428870ff | 836 | |
7ef5e54e AL |
837 | tqs->stqs_count = count; |
838 | tqs->stqs_taskq = kmem_alloc(count * sizeof (taskq_t *), KM_SLEEP); | |
428870ff | 839 | |
7ef5e54e AL |
840 | for (i = 0; i < count; i++) { |
841 | taskq_t *tq; | |
428870ff | 842 | |
7ef5e54e AL |
843 | switch (mode) { |
844 | case ZTI_MODE_FIXED: | |
845 | ASSERT3U(value, >=, 1); | |
846 | value = MAX(value, 1); | |
847 | break; | |
848 | ||
849 | case ZTI_MODE_BATCH: | |
850 | batch = B_TRUE; | |
851 | flags |= TASKQ_THREADS_CPU_PCT; | |
852 | value = zio_taskq_batch_pct; | |
853 | break; | |
854 | ||
855 | case ZTI_MODE_ONLINE_PERCENT: | |
856 | flags |= TASKQ_THREADS_CPU_PCT; | |
857 | break; | |
858 | ||
859 | default: | |
860 | panic("unrecognized mode for %s_%s taskq (%u:%u) in " | |
861 | "spa_activate()", | |
862 | zio_type_name[t], zio_taskq_types[q], mode, value); | |
863 | break; | |
864 | } | |
865 | ||
866 | if (count > 1) { | |
867 | (void) snprintf(name, sizeof (name), "%s_%s_%u", | |
868 | zio_type_name[t], zio_taskq_types[q], i); | |
869 | } else { | |
870 | (void) snprintf(name, sizeof (name), "%s_%s", | |
871 | zio_type_name[t], zio_taskq_types[q]); | |
872 | } | |
873 | ||
874 | if (zio_taskq_sysdc && spa->spa_proc != &p0) { | |
875 | if (batch) | |
876 | flags |= TASKQ_DC_BATCH; | |
877 | ||
878 | tq = taskq_create_sysdc(name, value, 50, INT_MAX, | |
879 | spa->spa_proc, zio_taskq_basedc, flags); | |
880 | } else { | |
881 | tq = taskq_create_proc(name, value, maxclsyspri, 50, | |
882 | INT_MAX, spa->spa_proc, flags); | |
883 | } | |
884 | ||
885 | tqs->stqs_taskq[i] = tq; | |
886 | } | |
887 | } | |
888 | ||
889 | static void | |
890 | spa_taskqs_fini(spa_t *spa, zio_type_t t, zio_taskq_type_t q) | |
891 | { | |
892 | spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q]; | |
893 | uint_t i; | |
894 | ||
895 | if (tqs->stqs_taskq == NULL) { | |
896 | ASSERT3U(tqs->stqs_count, ==, 0); | |
897 | return; | |
898 | } | |
899 | ||
900 | for (i = 0; i < tqs->stqs_count; i++) { | |
901 | ASSERT3P(tqs->stqs_taskq[i], !=, NULL); | |
902 | taskq_destroy(tqs->stqs_taskq[i]); | |
428870ff | 903 | } |
34dc7c2f | 904 | |
7ef5e54e AL |
905 | kmem_free(tqs->stqs_taskq, tqs->stqs_count * sizeof (taskq_t *)); |
906 | tqs->stqs_taskq = NULL; | |
907 | } | |
34dc7c2f | 908 | |
7ef5e54e AL |
909 | /* |
910 | * Dispatch a task to the appropriate taskq for the ZFS I/O type and priority. | |
911 | * Note that a type may have multiple discrete taskqs to avoid lock contention | |
912 | * on the taskq itself. In that case we choose which taskq at random by using | |
913 | * the low bits of gethrtime(). | |
914 | */ | |
915 | void | |
916 | spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q, | |
917 | task_func_t *func, void *arg, uint_t flags, taskq_ent_t *ent) | |
918 | { | |
919 | spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q]; | |
920 | taskq_t *tq; | |
921 | ||
922 | ASSERT3P(tqs->stqs_taskq, !=, NULL); | |
923 | ASSERT3U(tqs->stqs_count, !=, 0); | |
924 | ||
925 | if (tqs->stqs_count == 1) { | |
926 | tq = tqs->stqs_taskq[0]; | |
927 | } else { | |
c12936b1 | 928 | tq = tqs->stqs_taskq[((uint64_t)gethrtime()) % tqs->stqs_count]; |
428870ff | 929 | } |
7ef5e54e AL |
930 | |
931 | taskq_dispatch_ent(tq, func, arg, flags, ent); | |
428870ff BB |
932 | } |
933 | ||
044baf00 BB |
934 | /* |
935 | * Same as spa_taskq_dispatch_ent() but block on the task until completion. | |
936 | */ | |
937 | void | |
938 | spa_taskq_dispatch_sync(spa_t *spa, zio_type_t t, zio_taskq_type_t q, | |
939 | task_func_t *func, void *arg, uint_t flags) | |
940 | { | |
941 | spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q]; | |
942 | taskq_t *tq; | |
943 | taskqid_t id; | |
944 | ||
945 | ASSERT3P(tqs->stqs_taskq, !=, NULL); | |
946 | ASSERT3U(tqs->stqs_count, !=, 0); | |
947 | ||
948 | if (tqs->stqs_count == 1) { | |
949 | tq = tqs->stqs_taskq[0]; | |
950 | } else { | |
c12936b1 | 951 | tq = tqs->stqs_taskq[((uint64_t)gethrtime()) % tqs->stqs_count]; |
044baf00 BB |
952 | } |
953 | ||
954 | id = taskq_dispatch(tq, func, arg, flags); | |
955 | if (id) | |
956 | taskq_wait_id(tq, id); | |
957 | } | |
958 | ||
428870ff BB |
959 | static void |
960 | spa_create_zio_taskqs(spa_t *spa) | |
961 | { | |
d6320ddb BB |
962 | int t, q; |
963 | ||
964 | for (t = 0; t < ZIO_TYPES; t++) { | |
965 | for (q = 0; q < ZIO_TASKQ_TYPES; q++) { | |
7ef5e54e | 966 | spa_taskqs_init(spa, t, q); |
428870ff BB |
967 | } |
968 | } | |
969 | } | |
9babb374 | 970 | |
7b89a549 | 971 | #if defined(_KERNEL) && defined(HAVE_SPA_THREAD) |
428870ff BB |
972 | static void |
973 | spa_thread(void *arg) | |
974 | { | |
975 | callb_cpr_t cprinfo; | |
9babb374 | 976 | |
428870ff BB |
977 | spa_t *spa = arg; |
978 | user_t *pu = PTOU(curproc); | |
9babb374 | 979 | |
428870ff BB |
980 | CALLB_CPR_INIT(&cprinfo, &spa->spa_proc_lock, callb_generic_cpr, |
981 | spa->spa_name); | |
9babb374 | 982 | |
428870ff BB |
983 | ASSERT(curproc != &p0); |
984 | (void) snprintf(pu->u_psargs, sizeof (pu->u_psargs), | |
985 | "zpool-%s", spa->spa_name); | |
986 | (void) strlcpy(pu->u_comm, pu->u_psargs, sizeof (pu->u_comm)); | |
987 | ||
988 | /* bind this thread to the requested psrset */ | |
989 | if (zio_taskq_psrset_bind != PS_NONE) { | |
990 | pool_lock(); | |
991 | mutex_enter(&cpu_lock); | |
992 | mutex_enter(&pidlock); | |
993 | mutex_enter(&curproc->p_lock); | |
994 | ||
995 | if (cpupart_bind_thread(curthread, zio_taskq_psrset_bind, | |
996 | 0, NULL, NULL) == 0) { | |
997 | curthread->t_bind_pset = zio_taskq_psrset_bind; | |
998 | } else { | |
999 | cmn_err(CE_WARN, | |
1000 | "Couldn't bind process for zfs pool \"%s\" to " | |
1001 | "pset %d\n", spa->spa_name, zio_taskq_psrset_bind); | |
1002 | } | |
1003 | ||
1004 | mutex_exit(&curproc->p_lock); | |
1005 | mutex_exit(&pidlock); | |
1006 | mutex_exit(&cpu_lock); | |
1007 | pool_unlock(); | |
1008 | } | |
1009 | ||
1010 | if (zio_taskq_sysdc) { | |
1011 | sysdc_thread_enter(curthread, 100, 0); | |
1012 | } | |
1013 | ||
1014 | spa->spa_proc = curproc; | |
1015 | spa->spa_did = curthread->t_did; | |
1016 | ||
1017 | spa_create_zio_taskqs(spa); | |
1018 | ||
1019 | mutex_enter(&spa->spa_proc_lock); | |
1020 | ASSERT(spa->spa_proc_state == SPA_PROC_CREATED); | |
1021 | ||
1022 | spa->spa_proc_state = SPA_PROC_ACTIVE; | |
1023 | cv_broadcast(&spa->spa_proc_cv); | |
1024 | ||
1025 | CALLB_CPR_SAFE_BEGIN(&cprinfo); | |
1026 | while (spa->spa_proc_state == SPA_PROC_ACTIVE) | |
1027 | cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock); | |
1028 | CALLB_CPR_SAFE_END(&cprinfo, &spa->spa_proc_lock); | |
1029 | ||
1030 | ASSERT(spa->spa_proc_state == SPA_PROC_DEACTIVATE); | |
1031 | spa->spa_proc_state = SPA_PROC_GONE; | |
1032 | spa->spa_proc = &p0; | |
1033 | cv_broadcast(&spa->spa_proc_cv); | |
1034 | CALLB_CPR_EXIT(&cprinfo); /* drops spa_proc_lock */ | |
1035 | ||
1036 | mutex_enter(&curproc->p_lock); | |
1037 | lwp_exit(); | |
1038 | } | |
1039 | #endif | |
1040 | ||
1041 | /* | |
1042 | * Activate an uninitialized pool. | |
1043 | */ | |
1044 | static void | |
1045 | spa_activate(spa_t *spa, int mode) | |
1046 | { | |
1047 | ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); | |
1048 | ||
1049 | spa->spa_state = POOL_STATE_ACTIVE; | |
1050 | spa->spa_mode = mode; | |
1051 | ||
1052 | spa->spa_normal_class = metaslab_class_create(spa, zfs_metaslab_ops); | |
1053 | spa->spa_log_class = metaslab_class_create(spa, zfs_metaslab_ops); | |
1054 | ||
1055 | /* Try to create a covering process */ | |
1056 | mutex_enter(&spa->spa_proc_lock); | |
1057 | ASSERT(spa->spa_proc_state == SPA_PROC_NONE); | |
1058 | ASSERT(spa->spa_proc == &p0); | |
1059 | spa->spa_did = 0; | |
1060 | ||
7b89a549 | 1061 | #ifdef HAVE_SPA_THREAD |
428870ff BB |
1062 | /* Only create a process if we're going to be around a while. */ |
1063 | if (spa_create_process && strcmp(spa->spa_name, TRYIMPORT_NAME) != 0) { | |
1064 | if (newproc(spa_thread, (caddr_t)spa, syscid, maxclsyspri, | |
1065 | NULL, 0) == 0) { | |
1066 | spa->spa_proc_state = SPA_PROC_CREATED; | |
1067 | while (spa->spa_proc_state == SPA_PROC_CREATED) { | |
1068 | cv_wait(&spa->spa_proc_cv, | |
1069 | &spa->spa_proc_lock); | |
9babb374 | 1070 | } |
428870ff BB |
1071 | ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE); |
1072 | ASSERT(spa->spa_proc != &p0); | |
1073 | ASSERT(spa->spa_did != 0); | |
1074 | } else { | |
1075 | #ifdef _KERNEL | |
1076 | cmn_err(CE_WARN, | |
1077 | "Couldn't create process for zfs pool \"%s\"\n", | |
1078 | spa->spa_name); | |
1079 | #endif | |
b128c09f | 1080 | } |
34dc7c2f | 1081 | } |
7b89a549 | 1082 | #endif /* HAVE_SPA_THREAD */ |
428870ff BB |
1083 | mutex_exit(&spa->spa_proc_lock); |
1084 | ||
1085 | /* If we didn't create a process, we need to create our taskqs. */ | |
1086 | if (spa->spa_proc == &p0) { | |
1087 | spa_create_zio_taskqs(spa); | |
1088 | } | |
34dc7c2f | 1089 | |
b128c09f BB |
1090 | list_create(&spa->spa_config_dirty_list, sizeof (vdev_t), |
1091 | offsetof(vdev_t, vdev_config_dirty_node)); | |
1092 | list_create(&spa->spa_state_dirty_list, sizeof (vdev_t), | |
1093 | offsetof(vdev_t, vdev_state_dirty_node)); | |
34dc7c2f BB |
1094 | |
1095 | txg_list_create(&spa->spa_vdev_txg_list, | |
1096 | offsetof(struct vdev, vdev_txg_node)); | |
1097 | ||
1098 | avl_create(&spa->spa_errlist_scrub, | |
1099 | spa_error_entry_compare, sizeof (spa_error_entry_t), | |
1100 | offsetof(spa_error_entry_t, se_avl)); | |
1101 | avl_create(&spa->spa_errlist_last, | |
1102 | spa_error_entry_compare, sizeof (spa_error_entry_t), | |
1103 | offsetof(spa_error_entry_t, se_avl)); | |
1104 | } | |
1105 | ||
1106 | /* | |
1107 | * Opposite of spa_activate(). | |
1108 | */ | |
1109 | static void | |
1110 | spa_deactivate(spa_t *spa) | |
1111 | { | |
d6320ddb BB |
1112 | int t, q; |
1113 | ||
34dc7c2f BB |
1114 | ASSERT(spa->spa_sync_on == B_FALSE); |
1115 | ASSERT(spa->spa_dsl_pool == NULL); | |
1116 | ASSERT(spa->spa_root_vdev == NULL); | |
9babb374 | 1117 | ASSERT(spa->spa_async_zio_root == NULL); |
34dc7c2f BB |
1118 | ASSERT(spa->spa_state != POOL_STATE_UNINITIALIZED); |
1119 | ||
1120 | txg_list_destroy(&spa->spa_vdev_txg_list); | |
1121 | ||
b128c09f BB |
1122 | list_destroy(&spa->spa_config_dirty_list); |
1123 | list_destroy(&spa->spa_state_dirty_list); | |
34dc7c2f | 1124 | |
cc92e9d0 GW |
1125 | taskq_cancel_id(system_taskq, spa->spa_deadman_tqid); |
1126 | ||
d6320ddb BB |
1127 | for (t = 0; t < ZIO_TYPES; t++) { |
1128 | for (q = 0; q < ZIO_TASKQ_TYPES; q++) { | |
7ef5e54e | 1129 | spa_taskqs_fini(spa, t, q); |
b128c09f | 1130 | } |
34dc7c2f BB |
1131 | } |
1132 | ||
1133 | metaslab_class_destroy(spa->spa_normal_class); | |
1134 | spa->spa_normal_class = NULL; | |
1135 | ||
1136 | metaslab_class_destroy(spa->spa_log_class); | |
1137 | spa->spa_log_class = NULL; | |
1138 | ||
1139 | /* | |
1140 | * If this was part of an import or the open otherwise failed, we may | |
1141 | * still have errors left in the queues. Empty them just in case. | |
1142 | */ | |
1143 | spa_errlog_drain(spa); | |
1144 | ||
1145 | avl_destroy(&spa->spa_errlist_scrub); | |
1146 | avl_destroy(&spa->spa_errlist_last); | |
1147 | ||
1148 | spa->spa_state = POOL_STATE_UNINITIALIZED; | |
428870ff BB |
1149 | |
1150 | mutex_enter(&spa->spa_proc_lock); | |
1151 | if (spa->spa_proc_state != SPA_PROC_NONE) { | |
1152 | ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE); | |
1153 | spa->spa_proc_state = SPA_PROC_DEACTIVATE; | |
1154 | cv_broadcast(&spa->spa_proc_cv); | |
1155 | while (spa->spa_proc_state == SPA_PROC_DEACTIVATE) { | |
1156 | ASSERT(spa->spa_proc != &p0); | |
1157 | cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock); | |
1158 | } | |
1159 | ASSERT(spa->spa_proc_state == SPA_PROC_GONE); | |
1160 | spa->spa_proc_state = SPA_PROC_NONE; | |
1161 | } | |
1162 | ASSERT(spa->spa_proc == &p0); | |
1163 | mutex_exit(&spa->spa_proc_lock); | |
1164 | ||
1165 | /* | |
1166 | * We want to make sure spa_thread() has actually exited the ZFS | |
1167 | * module, so that the module can't be unloaded out from underneath | |
1168 | * it. | |
1169 | */ | |
1170 | if (spa->spa_did != 0) { | |
1171 | thread_join(spa->spa_did); | |
1172 | spa->spa_did = 0; | |
1173 | } | |
34dc7c2f BB |
1174 | } |
1175 | ||
1176 | /* | |
1177 | * Verify a pool configuration, and construct the vdev tree appropriately. This | |
1178 | * will create all the necessary vdevs in the appropriate layout, with each vdev | |
1179 | * in the CLOSED state. This will prep the pool before open/creation/import. | |
1180 | * All vdev validation is done by the vdev_alloc() routine. | |
1181 | */ | |
1182 | static int | |
1183 | spa_config_parse(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, | |
1184 | uint_t id, int atype) | |
1185 | { | |
1186 | nvlist_t **child; | |
9babb374 | 1187 | uint_t children; |
34dc7c2f | 1188 | int error; |
d6320ddb | 1189 | int c; |
34dc7c2f BB |
1190 | |
1191 | if ((error = vdev_alloc(spa, vdp, nv, parent, id, atype)) != 0) | |
1192 | return (error); | |
1193 | ||
1194 | if ((*vdp)->vdev_ops->vdev_op_leaf) | |
1195 | return (0); | |
1196 | ||
b128c09f BB |
1197 | error = nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, |
1198 | &child, &children); | |
1199 | ||
1200 | if (error == ENOENT) | |
1201 | return (0); | |
1202 | ||
1203 | if (error) { | |
34dc7c2f BB |
1204 | vdev_free(*vdp); |
1205 | *vdp = NULL; | |
2e528b49 | 1206 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1207 | } |
1208 | ||
d6320ddb | 1209 | for (c = 0; c < children; c++) { |
34dc7c2f BB |
1210 | vdev_t *vd; |
1211 | if ((error = spa_config_parse(spa, &vd, child[c], *vdp, c, | |
1212 | atype)) != 0) { | |
1213 | vdev_free(*vdp); | |
1214 | *vdp = NULL; | |
1215 | return (error); | |
1216 | } | |
1217 | } | |
1218 | ||
1219 | ASSERT(*vdp != NULL); | |
1220 | ||
1221 | return (0); | |
1222 | } | |
1223 | ||
1224 | /* | |
1225 | * Opposite of spa_load(). | |
1226 | */ | |
1227 | static void | |
1228 | spa_unload(spa_t *spa) | |
1229 | { | |
1230 | int i; | |
1231 | ||
b128c09f BB |
1232 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); |
1233 | ||
34dc7c2f BB |
1234 | /* |
1235 | * Stop async tasks. | |
1236 | */ | |
1237 | spa_async_suspend(spa); | |
1238 | ||
1239 | /* | |
1240 | * Stop syncing. | |
1241 | */ | |
1242 | if (spa->spa_sync_on) { | |
1243 | txg_sync_stop(spa->spa_dsl_pool); | |
1244 | spa->spa_sync_on = B_FALSE; | |
1245 | } | |
1246 | ||
1247 | /* | |
b128c09f | 1248 | * Wait for any outstanding async I/O to complete. |
34dc7c2f | 1249 | */ |
9babb374 BB |
1250 | if (spa->spa_async_zio_root != NULL) { |
1251 | (void) zio_wait(spa->spa_async_zio_root); | |
1252 | spa->spa_async_zio_root = NULL; | |
1253 | } | |
34dc7c2f | 1254 | |
428870ff BB |
1255 | bpobj_close(&spa->spa_deferred_bpobj); |
1256 | ||
34dc7c2f BB |
1257 | /* |
1258 | * Close the dsl pool. | |
1259 | */ | |
1260 | if (spa->spa_dsl_pool) { | |
1261 | dsl_pool_close(spa->spa_dsl_pool); | |
1262 | spa->spa_dsl_pool = NULL; | |
428870ff | 1263 | spa->spa_meta_objset = NULL; |
34dc7c2f BB |
1264 | } |
1265 | ||
428870ff BB |
1266 | ddt_unload(spa); |
1267 | ||
fb5f0bc8 BB |
1268 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
1269 | ||
1270 | /* | |
1271 | * Drop and purge level 2 cache | |
1272 | */ | |
1273 | spa_l2cache_drop(spa); | |
1274 | ||
34dc7c2f BB |
1275 | /* |
1276 | * Close all vdevs. | |
1277 | */ | |
1278 | if (spa->spa_root_vdev) | |
1279 | vdev_free(spa->spa_root_vdev); | |
1280 | ASSERT(spa->spa_root_vdev == NULL); | |
1281 | ||
1282 | for (i = 0; i < spa->spa_spares.sav_count; i++) | |
1283 | vdev_free(spa->spa_spares.sav_vdevs[i]); | |
1284 | if (spa->spa_spares.sav_vdevs) { | |
1285 | kmem_free(spa->spa_spares.sav_vdevs, | |
1286 | spa->spa_spares.sav_count * sizeof (void *)); | |
1287 | spa->spa_spares.sav_vdevs = NULL; | |
1288 | } | |
1289 | if (spa->spa_spares.sav_config) { | |
1290 | nvlist_free(spa->spa_spares.sav_config); | |
1291 | spa->spa_spares.sav_config = NULL; | |
1292 | } | |
b128c09f | 1293 | spa->spa_spares.sav_count = 0; |
34dc7c2f | 1294 | |
5ffb9d1d GW |
1295 | for (i = 0; i < spa->spa_l2cache.sav_count; i++) { |
1296 | vdev_clear_stats(spa->spa_l2cache.sav_vdevs[i]); | |
34dc7c2f | 1297 | vdev_free(spa->spa_l2cache.sav_vdevs[i]); |
5ffb9d1d | 1298 | } |
34dc7c2f BB |
1299 | if (spa->spa_l2cache.sav_vdevs) { |
1300 | kmem_free(spa->spa_l2cache.sav_vdevs, | |
1301 | spa->spa_l2cache.sav_count * sizeof (void *)); | |
1302 | spa->spa_l2cache.sav_vdevs = NULL; | |
1303 | } | |
1304 | if (spa->spa_l2cache.sav_config) { | |
1305 | nvlist_free(spa->spa_l2cache.sav_config); | |
1306 | spa->spa_l2cache.sav_config = NULL; | |
1307 | } | |
b128c09f | 1308 | spa->spa_l2cache.sav_count = 0; |
34dc7c2f BB |
1309 | |
1310 | spa->spa_async_suspended = 0; | |
fb5f0bc8 | 1311 | |
d96eb2b1 DM |
1312 | if (spa->spa_comment != NULL) { |
1313 | spa_strfree(spa->spa_comment); | |
1314 | spa->spa_comment = NULL; | |
1315 | } | |
1316 | ||
fb5f0bc8 | 1317 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
1318 | } |
1319 | ||
1320 | /* | |
1321 | * Load (or re-load) the current list of vdevs describing the active spares for | |
1322 | * this pool. When this is called, we have some form of basic information in | |
1323 | * 'spa_spares.sav_config'. We parse this into vdevs, try to open them, and | |
1324 | * then re-generate a more complete list including status information. | |
1325 | */ | |
1326 | static void | |
1327 | spa_load_spares(spa_t *spa) | |
1328 | { | |
1329 | nvlist_t **spares; | |
1330 | uint_t nspares; | |
1331 | int i; | |
1332 | vdev_t *vd, *tvd; | |
1333 | ||
b128c09f BB |
1334 | ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); |
1335 | ||
34dc7c2f BB |
1336 | /* |
1337 | * First, close and free any existing spare vdevs. | |
1338 | */ | |
1339 | for (i = 0; i < spa->spa_spares.sav_count; i++) { | |
1340 | vd = spa->spa_spares.sav_vdevs[i]; | |
1341 | ||
1342 | /* Undo the call to spa_activate() below */ | |
b128c09f BB |
1343 | if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid, |
1344 | B_FALSE)) != NULL && tvd->vdev_isspare) | |
34dc7c2f BB |
1345 | spa_spare_remove(tvd); |
1346 | vdev_close(vd); | |
1347 | vdev_free(vd); | |
1348 | } | |
1349 | ||
1350 | if (spa->spa_spares.sav_vdevs) | |
1351 | kmem_free(spa->spa_spares.sav_vdevs, | |
1352 | spa->spa_spares.sav_count * sizeof (void *)); | |
1353 | ||
1354 | if (spa->spa_spares.sav_config == NULL) | |
1355 | nspares = 0; | |
1356 | else | |
1357 | VERIFY(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config, | |
1358 | ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0); | |
1359 | ||
1360 | spa->spa_spares.sav_count = (int)nspares; | |
1361 | spa->spa_spares.sav_vdevs = NULL; | |
1362 | ||
1363 | if (nspares == 0) | |
1364 | return; | |
1365 | ||
1366 | /* | |
1367 | * Construct the array of vdevs, opening them to get status in the | |
1368 | * process. For each spare, there is potentially two different vdev_t | |
1369 | * structures associated with it: one in the list of spares (used only | |
1370 | * for basic validation purposes) and one in the active vdev | |
1371 | * configuration (if it's spared in). During this phase we open and | |
1372 | * validate each vdev on the spare list. If the vdev also exists in the | |
1373 | * active configuration, then we also mark this vdev as an active spare. | |
1374 | */ | |
1375 | spa->spa_spares.sav_vdevs = kmem_alloc(nspares * sizeof (void *), | |
b8d06fca | 1376 | KM_PUSHPAGE); |
34dc7c2f BB |
1377 | for (i = 0; i < spa->spa_spares.sav_count; i++) { |
1378 | VERIFY(spa_config_parse(spa, &vd, spares[i], NULL, 0, | |
1379 | VDEV_ALLOC_SPARE) == 0); | |
1380 | ASSERT(vd != NULL); | |
1381 | ||
1382 | spa->spa_spares.sav_vdevs[i] = vd; | |
1383 | ||
b128c09f BB |
1384 | if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid, |
1385 | B_FALSE)) != NULL) { | |
34dc7c2f BB |
1386 | if (!tvd->vdev_isspare) |
1387 | spa_spare_add(tvd); | |
1388 | ||
1389 | /* | |
1390 | * We only mark the spare active if we were successfully | |
1391 | * able to load the vdev. Otherwise, importing a pool | |
1392 | * with a bad active spare would result in strange | |
1393 | * behavior, because multiple pool would think the spare | |
1394 | * is actively in use. | |
1395 | * | |
1396 | * There is a vulnerability here to an equally bizarre | |
1397 | * circumstance, where a dead active spare is later | |
1398 | * brought back to life (onlined or otherwise). Given | |
1399 | * the rarity of this scenario, and the extra complexity | |
1400 | * it adds, we ignore the possibility. | |
1401 | */ | |
1402 | if (!vdev_is_dead(tvd)) | |
1403 | spa_spare_activate(tvd); | |
1404 | } | |
1405 | ||
b128c09f | 1406 | vd->vdev_top = vd; |
9babb374 | 1407 | vd->vdev_aux = &spa->spa_spares; |
b128c09f | 1408 | |
34dc7c2f BB |
1409 | if (vdev_open(vd) != 0) |
1410 | continue; | |
1411 | ||
34dc7c2f BB |
1412 | if (vdev_validate_aux(vd) == 0) |
1413 | spa_spare_add(vd); | |
1414 | } | |
1415 | ||
1416 | /* | |
1417 | * Recompute the stashed list of spares, with status information | |
1418 | * this time. | |
1419 | */ | |
1420 | VERIFY(nvlist_remove(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, | |
1421 | DATA_TYPE_NVLIST_ARRAY) == 0); | |
1422 | ||
1423 | spares = kmem_alloc(spa->spa_spares.sav_count * sizeof (void *), | |
b8d06fca | 1424 | KM_PUSHPAGE); |
34dc7c2f BB |
1425 | for (i = 0; i < spa->spa_spares.sav_count; i++) |
1426 | spares[i] = vdev_config_generate(spa, | |
428870ff | 1427 | spa->spa_spares.sav_vdevs[i], B_TRUE, VDEV_CONFIG_SPARE); |
34dc7c2f BB |
1428 | VERIFY(nvlist_add_nvlist_array(spa->spa_spares.sav_config, |
1429 | ZPOOL_CONFIG_SPARES, spares, spa->spa_spares.sav_count) == 0); | |
1430 | for (i = 0; i < spa->spa_spares.sav_count; i++) | |
1431 | nvlist_free(spares[i]); | |
1432 | kmem_free(spares, spa->spa_spares.sav_count * sizeof (void *)); | |
1433 | } | |
1434 | ||
1435 | /* | |
1436 | * Load (or re-load) the current list of vdevs describing the active l2cache for | |
1437 | * this pool. When this is called, we have some form of basic information in | |
1438 | * 'spa_l2cache.sav_config'. We parse this into vdevs, try to open them, and | |
1439 | * then re-generate a more complete list including status information. | |
1440 | * Devices which are already active have their details maintained, and are | |
1441 | * not re-opened. | |
1442 | */ | |
1443 | static void | |
1444 | spa_load_l2cache(spa_t *spa) | |
1445 | { | |
1446 | nvlist_t **l2cache; | |
1447 | uint_t nl2cache; | |
1448 | int i, j, oldnvdevs; | |
9babb374 | 1449 | uint64_t guid; |
a117a6d6 | 1450 | vdev_t *vd, **oldvdevs, **newvdevs; |
34dc7c2f BB |
1451 | spa_aux_vdev_t *sav = &spa->spa_l2cache; |
1452 | ||
b128c09f BB |
1453 | ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); |
1454 | ||
34dc7c2f BB |
1455 | if (sav->sav_config != NULL) { |
1456 | VERIFY(nvlist_lookup_nvlist_array(sav->sav_config, | |
1457 | ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0); | |
b8d06fca | 1458 | newvdevs = kmem_alloc(nl2cache * sizeof (void *), KM_PUSHPAGE); |
34dc7c2f BB |
1459 | } else { |
1460 | nl2cache = 0; | |
a117a6d6 | 1461 | newvdevs = NULL; |
34dc7c2f BB |
1462 | } |
1463 | ||
1464 | oldvdevs = sav->sav_vdevs; | |
1465 | oldnvdevs = sav->sav_count; | |
1466 | sav->sav_vdevs = NULL; | |
1467 | sav->sav_count = 0; | |
1468 | ||
1469 | /* | |
1470 | * Process new nvlist of vdevs. | |
1471 | */ | |
1472 | for (i = 0; i < nl2cache; i++) { | |
1473 | VERIFY(nvlist_lookup_uint64(l2cache[i], ZPOOL_CONFIG_GUID, | |
1474 | &guid) == 0); | |
1475 | ||
1476 | newvdevs[i] = NULL; | |
1477 | for (j = 0; j < oldnvdevs; j++) { | |
1478 | vd = oldvdevs[j]; | |
1479 | if (vd != NULL && guid == vd->vdev_guid) { | |
1480 | /* | |
1481 | * Retain previous vdev for add/remove ops. | |
1482 | */ | |
1483 | newvdevs[i] = vd; | |
1484 | oldvdevs[j] = NULL; | |
1485 | break; | |
1486 | } | |
1487 | } | |
1488 | ||
1489 | if (newvdevs[i] == NULL) { | |
1490 | /* | |
1491 | * Create new vdev | |
1492 | */ | |
1493 | VERIFY(spa_config_parse(spa, &vd, l2cache[i], NULL, 0, | |
1494 | VDEV_ALLOC_L2CACHE) == 0); | |
1495 | ASSERT(vd != NULL); | |
1496 | newvdevs[i] = vd; | |
1497 | ||
1498 | /* | |
1499 | * Commit this vdev as an l2cache device, | |
1500 | * even if it fails to open. | |
1501 | */ | |
1502 | spa_l2cache_add(vd); | |
1503 | ||
b128c09f BB |
1504 | vd->vdev_top = vd; |
1505 | vd->vdev_aux = sav; | |
1506 | ||
1507 | spa_l2cache_activate(vd); | |
1508 | ||
34dc7c2f BB |
1509 | if (vdev_open(vd) != 0) |
1510 | continue; | |
1511 | ||
34dc7c2f BB |
1512 | (void) vdev_validate_aux(vd); |
1513 | ||
9babb374 BB |
1514 | if (!vdev_is_dead(vd)) |
1515 | l2arc_add_vdev(spa, vd); | |
34dc7c2f BB |
1516 | } |
1517 | } | |
1518 | ||
1519 | /* | |
1520 | * Purge vdevs that were dropped | |
1521 | */ | |
1522 | for (i = 0; i < oldnvdevs; i++) { | |
1523 | uint64_t pool; | |
1524 | ||
1525 | vd = oldvdevs[i]; | |
1526 | if (vd != NULL) { | |
5ffb9d1d GW |
1527 | ASSERT(vd->vdev_isl2cache); |
1528 | ||
fb5f0bc8 BB |
1529 | if (spa_l2cache_exists(vd->vdev_guid, &pool) && |
1530 | pool != 0ULL && l2arc_vdev_present(vd)) | |
34dc7c2f | 1531 | l2arc_remove_vdev(vd); |
5ffb9d1d GW |
1532 | vdev_clear_stats(vd); |
1533 | vdev_free(vd); | |
34dc7c2f BB |
1534 | } |
1535 | } | |
1536 | ||
1537 | if (oldvdevs) | |
1538 | kmem_free(oldvdevs, oldnvdevs * sizeof (void *)); | |
1539 | ||
1540 | if (sav->sav_config == NULL) | |
1541 | goto out; | |
1542 | ||
1543 | sav->sav_vdevs = newvdevs; | |
1544 | sav->sav_count = (int)nl2cache; | |
1545 | ||
1546 | /* | |
1547 | * Recompute the stashed list of l2cache devices, with status | |
1548 | * information this time. | |
1549 | */ | |
1550 | VERIFY(nvlist_remove(sav->sav_config, ZPOOL_CONFIG_L2CACHE, | |
1551 | DATA_TYPE_NVLIST_ARRAY) == 0); | |
1552 | ||
b8d06fca | 1553 | l2cache = kmem_alloc(sav->sav_count * sizeof (void *), KM_PUSHPAGE); |
34dc7c2f BB |
1554 | for (i = 0; i < sav->sav_count; i++) |
1555 | l2cache[i] = vdev_config_generate(spa, | |
428870ff | 1556 | sav->sav_vdevs[i], B_TRUE, VDEV_CONFIG_L2CACHE); |
34dc7c2f BB |
1557 | VERIFY(nvlist_add_nvlist_array(sav->sav_config, |
1558 | ZPOOL_CONFIG_L2CACHE, l2cache, sav->sav_count) == 0); | |
1559 | out: | |
1560 | for (i = 0; i < sav->sav_count; i++) | |
1561 | nvlist_free(l2cache[i]); | |
1562 | if (sav->sav_count) | |
1563 | kmem_free(l2cache, sav->sav_count * sizeof (void *)); | |
1564 | } | |
1565 | ||
1566 | static int | |
1567 | load_nvlist(spa_t *spa, uint64_t obj, nvlist_t **value) | |
1568 | { | |
1569 | dmu_buf_t *db; | |
1570 | char *packed = NULL; | |
1571 | size_t nvsize = 0; | |
1572 | int error; | |
1573 | *value = NULL; | |
1574 | ||
c3275b56 BB |
1575 | error = dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db); |
1576 | if (error) | |
1577 | return (error); | |
1578 | ||
34dc7c2f BB |
1579 | nvsize = *(uint64_t *)db->db_data; |
1580 | dmu_buf_rele(db, FTAG); | |
1581 | ||
b8d06fca | 1582 | packed = kmem_alloc(nvsize, KM_PUSHPAGE | KM_NODEBUG); |
9babb374 BB |
1583 | error = dmu_read(spa->spa_meta_objset, obj, 0, nvsize, packed, |
1584 | DMU_READ_PREFETCH); | |
34dc7c2f BB |
1585 | if (error == 0) |
1586 | error = nvlist_unpack(packed, nvsize, value, 0); | |
1587 | kmem_free(packed, nvsize); | |
1588 | ||
1589 | return (error); | |
1590 | } | |
1591 | ||
1592 | /* | |
1593 | * Checks to see if the given vdev could not be opened, in which case we post a | |
1594 | * sysevent to notify the autoreplace code that the device has been removed. | |
1595 | */ | |
1596 | static void | |
1597 | spa_check_removed(vdev_t *vd) | |
1598 | { | |
d6320ddb BB |
1599 | int c; |
1600 | ||
1601 | for (c = 0; c < vd->vdev_children; c++) | |
34dc7c2f BB |
1602 | spa_check_removed(vd->vdev_child[c]); |
1603 | ||
7011fb60 YP |
1604 | if (vd->vdev_ops->vdev_op_leaf && vdev_is_dead(vd) && |
1605 | !vd->vdev_ishole) { | |
26685276 BB |
1606 | zfs_ereport_post(FM_EREPORT_RESOURCE_AUTOREPLACE, |
1607 | vd->vdev_spa, vd, NULL, 0, 0); | |
1608 | spa_event_notify(vd->vdev_spa, vd, FM_EREPORT_ZFS_DEVICE_CHECK); | |
34dc7c2f BB |
1609 | } |
1610 | } | |
1611 | ||
9babb374 | 1612 | /* |
572e2857 | 1613 | * Validate the current config against the MOS config |
9babb374 | 1614 | */ |
572e2857 BB |
1615 | static boolean_t |
1616 | spa_config_valid(spa_t *spa, nvlist_t *config) | |
9babb374 | 1617 | { |
572e2857 BB |
1618 | vdev_t *mrvd, *rvd = spa->spa_root_vdev; |
1619 | nvlist_t *nv; | |
d6320ddb | 1620 | int c, i; |
572e2857 BB |
1621 | |
1622 | VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nv) == 0); | |
1623 | ||
1624 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); | |
1625 | VERIFY(spa_config_parse(spa, &mrvd, nv, NULL, 0, VDEV_ALLOC_LOAD) == 0); | |
1626 | ||
1627 | ASSERT3U(rvd->vdev_children, ==, mrvd->vdev_children); | |
9babb374 | 1628 | |
428870ff | 1629 | /* |
572e2857 BB |
1630 | * If we're doing a normal import, then build up any additional |
1631 | * diagnostic information about missing devices in this config. | |
1632 | * We'll pass this up to the user for further processing. | |
428870ff | 1633 | */ |
572e2857 BB |
1634 | if (!(spa->spa_import_flags & ZFS_IMPORT_MISSING_LOG)) { |
1635 | nvlist_t **child, *nv; | |
1636 | uint64_t idx = 0; | |
1637 | ||
1638 | child = kmem_alloc(rvd->vdev_children * sizeof (nvlist_t **), | |
b8d06fca RY |
1639 | KM_PUSHPAGE); |
1640 | VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_PUSHPAGE) == 0); | |
572e2857 | 1641 | |
d6320ddb | 1642 | for (c = 0; c < rvd->vdev_children; c++) { |
572e2857 BB |
1643 | vdev_t *tvd = rvd->vdev_child[c]; |
1644 | vdev_t *mtvd = mrvd->vdev_child[c]; | |
1645 | ||
1646 | if (tvd->vdev_ops == &vdev_missing_ops && | |
1647 | mtvd->vdev_ops != &vdev_missing_ops && | |
1648 | mtvd->vdev_islog) | |
1649 | child[idx++] = vdev_config_generate(spa, mtvd, | |
1650 | B_FALSE, 0); | |
1651 | } | |
9babb374 | 1652 | |
572e2857 BB |
1653 | if (idx) { |
1654 | VERIFY(nvlist_add_nvlist_array(nv, | |
1655 | ZPOOL_CONFIG_CHILDREN, child, idx) == 0); | |
1656 | VERIFY(nvlist_add_nvlist(spa->spa_load_info, | |
1657 | ZPOOL_CONFIG_MISSING_DEVICES, nv) == 0); | |
1658 | ||
d6320ddb | 1659 | for (i = 0; i < idx; i++) |
572e2857 BB |
1660 | nvlist_free(child[i]); |
1661 | } | |
1662 | nvlist_free(nv); | |
1663 | kmem_free(child, rvd->vdev_children * sizeof (char **)); | |
1664 | } | |
1665 | ||
1666 | /* | |
1667 | * Compare the root vdev tree with the information we have | |
1668 | * from the MOS config (mrvd). Check each top-level vdev | |
1669 | * with the corresponding MOS config top-level (mtvd). | |
1670 | */ | |
d6320ddb | 1671 | for (c = 0; c < rvd->vdev_children; c++) { |
572e2857 BB |
1672 | vdev_t *tvd = rvd->vdev_child[c]; |
1673 | vdev_t *mtvd = mrvd->vdev_child[c]; | |
1674 | ||
1675 | /* | |
1676 | * Resolve any "missing" vdevs in the current configuration. | |
1677 | * If we find that the MOS config has more accurate information | |
1678 | * about the top-level vdev then use that vdev instead. | |
1679 | */ | |
1680 | if (tvd->vdev_ops == &vdev_missing_ops && | |
1681 | mtvd->vdev_ops != &vdev_missing_ops) { | |
1682 | ||
1683 | if (!(spa->spa_import_flags & ZFS_IMPORT_MISSING_LOG)) | |
1684 | continue; | |
1685 | ||
1686 | /* | |
1687 | * Device specific actions. | |
1688 | */ | |
1689 | if (mtvd->vdev_islog) { | |
1690 | spa_set_log_state(spa, SPA_LOG_CLEAR); | |
1691 | } else { | |
1692 | /* | |
1693 | * XXX - once we have 'readonly' pool | |
1694 | * support we should be able to handle | |
1695 | * missing data devices by transitioning | |
1696 | * the pool to readonly. | |
1697 | */ | |
1698 | continue; | |
1699 | } | |
1700 | ||
1701 | /* | |
1702 | * Swap the missing vdev with the data we were | |
1703 | * able to obtain from the MOS config. | |
1704 | */ | |
1705 | vdev_remove_child(rvd, tvd); | |
1706 | vdev_remove_child(mrvd, mtvd); | |
1707 | ||
1708 | vdev_add_child(rvd, mtvd); | |
1709 | vdev_add_child(mrvd, tvd); | |
1710 | ||
1711 | spa_config_exit(spa, SCL_ALL, FTAG); | |
1712 | vdev_load(mtvd); | |
1713 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); | |
1714 | ||
1715 | vdev_reopen(rvd); | |
1716 | } else if (mtvd->vdev_islog) { | |
1717 | /* | |
1718 | * Load the slog device's state from the MOS config | |
1719 | * since it's possible that the label does not | |
1720 | * contain the most up-to-date information. | |
1721 | */ | |
1722 | vdev_load_log_state(tvd, mtvd); | |
1723 | vdev_reopen(tvd); | |
1724 | } | |
9babb374 | 1725 | } |
572e2857 | 1726 | vdev_free(mrvd); |
428870ff | 1727 | spa_config_exit(spa, SCL_ALL, FTAG); |
572e2857 BB |
1728 | |
1729 | /* | |
1730 | * Ensure we were able to validate the config. | |
1731 | */ | |
1732 | return (rvd->vdev_guid_sum == spa->spa_uberblock.ub_guid_sum); | |
9babb374 BB |
1733 | } |
1734 | ||
b128c09f BB |
1735 | /* |
1736 | * Check for missing log devices | |
1737 | */ | |
13fe0198 | 1738 | static boolean_t |
b128c09f BB |
1739 | spa_check_logs(spa_t *spa) |
1740 | { | |
13fe0198 MA |
1741 | boolean_t rv = B_FALSE; |
1742 | ||
b128c09f | 1743 | switch (spa->spa_log_state) { |
e75c13c3 BB |
1744 | default: |
1745 | break; | |
b128c09f BB |
1746 | case SPA_LOG_MISSING: |
1747 | /* need to recheck in case slog has been restored */ | |
1748 | case SPA_LOG_UNKNOWN: | |
13fe0198 MA |
1749 | rv = (dmu_objset_find(spa->spa_name, zil_check_log_chain, |
1750 | NULL, DS_FIND_CHILDREN) != 0); | |
1751 | if (rv) | |
428870ff | 1752 | spa_set_log_state(spa, SPA_LOG_MISSING); |
b128c09f | 1753 | break; |
b128c09f | 1754 | } |
13fe0198 | 1755 | return (rv); |
b128c09f BB |
1756 | } |
1757 | ||
428870ff BB |
1758 | static boolean_t |
1759 | spa_passivate_log(spa_t *spa) | |
34dc7c2f | 1760 | { |
428870ff BB |
1761 | vdev_t *rvd = spa->spa_root_vdev; |
1762 | boolean_t slog_found = B_FALSE; | |
d6320ddb | 1763 | int c; |
b128c09f | 1764 | |
428870ff | 1765 | ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER)); |
fb5f0bc8 | 1766 | |
428870ff BB |
1767 | if (!spa_has_slogs(spa)) |
1768 | return (B_FALSE); | |
34dc7c2f | 1769 | |
d6320ddb | 1770 | for (c = 0; c < rvd->vdev_children; c++) { |
428870ff BB |
1771 | vdev_t *tvd = rvd->vdev_child[c]; |
1772 | metaslab_group_t *mg = tvd->vdev_mg; | |
34dc7c2f | 1773 | |
428870ff BB |
1774 | if (tvd->vdev_islog) { |
1775 | metaslab_group_passivate(mg); | |
1776 | slog_found = B_TRUE; | |
1777 | } | |
34dc7c2f BB |
1778 | } |
1779 | ||
428870ff BB |
1780 | return (slog_found); |
1781 | } | |
34dc7c2f | 1782 | |
428870ff BB |
1783 | static void |
1784 | spa_activate_log(spa_t *spa) | |
1785 | { | |
1786 | vdev_t *rvd = spa->spa_root_vdev; | |
d6320ddb | 1787 | int c; |
34dc7c2f | 1788 | |
428870ff BB |
1789 | ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER)); |
1790 | ||
d6320ddb | 1791 | for (c = 0; c < rvd->vdev_children; c++) { |
428870ff BB |
1792 | vdev_t *tvd = rvd->vdev_child[c]; |
1793 | metaslab_group_t *mg = tvd->vdev_mg; | |
1794 | ||
1795 | if (tvd->vdev_islog) | |
1796 | metaslab_group_activate(mg); | |
34dc7c2f | 1797 | } |
428870ff | 1798 | } |
34dc7c2f | 1799 | |
428870ff BB |
1800 | int |
1801 | spa_offline_log(spa_t *spa) | |
1802 | { | |
13fe0198 | 1803 | int error; |
9babb374 | 1804 | |
13fe0198 MA |
1805 | error = dmu_objset_find(spa_name(spa), zil_vdev_offline, |
1806 | NULL, DS_FIND_CHILDREN); | |
1807 | if (error == 0) { | |
428870ff BB |
1808 | /* |
1809 | * We successfully offlined the log device, sync out the | |
1810 | * current txg so that the "stubby" block can be removed | |
1811 | * by zil_sync(). | |
1812 | */ | |
1813 | txg_wait_synced(spa->spa_dsl_pool, 0); | |
1814 | } | |
1815 | return (error); | |
1816 | } | |
34dc7c2f | 1817 | |
428870ff BB |
1818 | static void |
1819 | spa_aux_check_removed(spa_aux_vdev_t *sav) | |
1820 | { | |
d6320ddb BB |
1821 | int i; |
1822 | ||
1823 | for (i = 0; i < sav->sav_count; i++) | |
428870ff BB |
1824 | spa_check_removed(sav->sav_vdevs[i]); |
1825 | } | |
34dc7c2f | 1826 | |
428870ff BB |
1827 | void |
1828 | spa_claim_notify(zio_t *zio) | |
1829 | { | |
1830 | spa_t *spa = zio->io_spa; | |
34dc7c2f | 1831 | |
428870ff BB |
1832 | if (zio->io_error) |
1833 | return; | |
34dc7c2f | 1834 | |
428870ff BB |
1835 | mutex_enter(&spa->spa_props_lock); /* any mutex will do */ |
1836 | if (spa->spa_claim_max_txg < zio->io_bp->blk_birth) | |
1837 | spa->spa_claim_max_txg = zio->io_bp->blk_birth; | |
1838 | mutex_exit(&spa->spa_props_lock); | |
1839 | } | |
34dc7c2f | 1840 | |
428870ff BB |
1841 | typedef struct spa_load_error { |
1842 | uint64_t sle_meta_count; | |
1843 | uint64_t sle_data_count; | |
1844 | } spa_load_error_t; | |
34dc7c2f | 1845 | |
428870ff BB |
1846 | static void |
1847 | spa_load_verify_done(zio_t *zio) | |
1848 | { | |
1849 | blkptr_t *bp = zio->io_bp; | |
1850 | spa_load_error_t *sle = zio->io_private; | |
1851 | dmu_object_type_t type = BP_GET_TYPE(bp); | |
1852 | int error = zio->io_error; | |
34dc7c2f | 1853 | |
428870ff | 1854 | if (error) { |
9ae529ec | 1855 | if ((BP_GET_LEVEL(bp) != 0 || DMU_OT_IS_METADATA(type)) && |
428870ff BB |
1856 | type != DMU_OT_INTENT_LOG) |
1857 | atomic_add_64(&sle->sle_meta_count, 1); | |
1858 | else | |
1859 | atomic_add_64(&sle->sle_data_count, 1); | |
34dc7c2f | 1860 | } |
428870ff BB |
1861 | zio_data_buf_free(zio->io_data, zio->io_size); |
1862 | } | |
34dc7c2f | 1863 | |
428870ff BB |
1864 | /*ARGSUSED*/ |
1865 | static int | |
1866 | spa_load_verify_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp, | |
294f6806 | 1867 | const zbookmark_t *zb, const dnode_phys_t *dnp, void *arg) |
428870ff BB |
1868 | { |
1869 | if (bp != NULL) { | |
1870 | zio_t *rio = arg; | |
1871 | size_t size = BP_GET_PSIZE(bp); | |
1872 | void *data = zio_data_buf_alloc(size); | |
34dc7c2f | 1873 | |
428870ff BB |
1874 | zio_nowait(zio_read(rio, spa, bp, data, size, |
1875 | spa_load_verify_done, rio->io_private, ZIO_PRIORITY_SCRUB, | |
1876 | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_CANFAIL | | |
1877 | ZIO_FLAG_SCRUB | ZIO_FLAG_RAW, zb)); | |
34dc7c2f | 1878 | } |
428870ff BB |
1879 | return (0); |
1880 | } | |
34dc7c2f | 1881 | |
428870ff BB |
1882 | static int |
1883 | spa_load_verify(spa_t *spa) | |
1884 | { | |
1885 | zio_t *rio; | |
1886 | spa_load_error_t sle = { 0 }; | |
1887 | zpool_rewind_policy_t policy; | |
1888 | boolean_t verify_ok = B_FALSE; | |
1889 | int error; | |
34dc7c2f | 1890 | |
428870ff | 1891 | zpool_get_rewind_policy(spa->spa_config, &policy); |
34dc7c2f | 1892 | |
428870ff BB |
1893 | if (policy.zrp_request & ZPOOL_NEVER_REWIND) |
1894 | return (0); | |
34dc7c2f | 1895 | |
428870ff BB |
1896 | rio = zio_root(spa, NULL, &sle, |
1897 | ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE); | |
34dc7c2f | 1898 | |
428870ff BB |
1899 | error = traverse_pool(spa, spa->spa_verify_min_txg, |
1900 | TRAVERSE_PRE | TRAVERSE_PREFETCH, spa_load_verify_cb, rio); | |
1901 | ||
1902 | (void) zio_wait(rio); | |
1903 | ||
1904 | spa->spa_load_meta_errors = sle.sle_meta_count; | |
1905 | spa->spa_load_data_errors = sle.sle_data_count; | |
1906 | ||
1907 | if (!error && sle.sle_meta_count <= policy.zrp_maxmeta && | |
1908 | sle.sle_data_count <= policy.zrp_maxdata) { | |
572e2857 BB |
1909 | int64_t loss = 0; |
1910 | ||
428870ff BB |
1911 | verify_ok = B_TRUE; |
1912 | spa->spa_load_txg = spa->spa_uberblock.ub_txg; | |
1913 | spa->spa_load_txg_ts = spa->spa_uberblock.ub_timestamp; | |
572e2857 BB |
1914 | |
1915 | loss = spa->spa_last_ubsync_txg_ts - spa->spa_load_txg_ts; | |
1916 | VERIFY(nvlist_add_uint64(spa->spa_load_info, | |
1917 | ZPOOL_CONFIG_LOAD_TIME, spa->spa_load_txg_ts) == 0); | |
1918 | VERIFY(nvlist_add_int64(spa->spa_load_info, | |
1919 | ZPOOL_CONFIG_REWIND_TIME, loss) == 0); | |
1920 | VERIFY(nvlist_add_uint64(spa->spa_load_info, | |
1921 | ZPOOL_CONFIG_LOAD_DATA_ERRORS, sle.sle_data_count) == 0); | |
428870ff BB |
1922 | } else { |
1923 | spa->spa_load_max_txg = spa->spa_uberblock.ub_txg; | |
1924 | } | |
1925 | ||
1926 | if (error) { | |
1927 | if (error != ENXIO && error != EIO) | |
2e528b49 | 1928 | error = SET_ERROR(EIO); |
428870ff BB |
1929 | return (error); |
1930 | } | |
1931 | ||
1932 | return (verify_ok ? 0 : EIO); | |
1933 | } | |
1934 | ||
1935 | /* | |
1936 | * Find a value in the pool props object. | |
1937 | */ | |
1938 | static void | |
1939 | spa_prop_find(spa_t *spa, zpool_prop_t prop, uint64_t *val) | |
1940 | { | |
1941 | (void) zap_lookup(spa->spa_meta_objset, spa->spa_pool_props_object, | |
1942 | zpool_prop_to_name(prop), sizeof (uint64_t), 1, val); | |
1943 | } | |
1944 | ||
1945 | /* | |
1946 | * Find a value in the pool directory object. | |
1947 | */ | |
1948 | static int | |
1949 | spa_dir_prop(spa_t *spa, const char *name, uint64_t *val) | |
1950 | { | |
1951 | return (zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, | |
1952 | name, sizeof (uint64_t), 1, val)); | |
1953 | } | |
1954 | ||
1955 | static int | |
1956 | spa_vdev_err(vdev_t *vdev, vdev_aux_t aux, int err) | |
1957 | { | |
1958 | vdev_set_state(vdev, B_TRUE, VDEV_STATE_CANT_OPEN, aux); | |
1959 | return (err); | |
1960 | } | |
1961 | ||
1962 | /* | |
1963 | * Fix up config after a partly-completed split. This is done with the | |
1964 | * ZPOOL_CONFIG_SPLIT nvlist. Both the splitting pool and the split-off | |
1965 | * pool have that entry in their config, but only the splitting one contains | |
1966 | * a list of all the guids of the vdevs that are being split off. | |
1967 | * | |
1968 | * This function determines what to do with that list: either rejoin | |
1969 | * all the disks to the pool, or complete the splitting process. To attempt | |
1970 | * the rejoin, each disk that is offlined is marked online again, and | |
1971 | * we do a reopen() call. If the vdev label for every disk that was | |
1972 | * marked online indicates it was successfully split off (VDEV_AUX_SPLIT_POOL) | |
1973 | * then we call vdev_split() on each disk, and complete the split. | |
1974 | * | |
1975 | * Otherwise we leave the config alone, with all the vdevs in place in | |
1976 | * the original pool. | |
1977 | */ | |
1978 | static void | |
1979 | spa_try_repair(spa_t *spa, nvlist_t *config) | |
1980 | { | |
1981 | uint_t extracted; | |
1982 | uint64_t *glist; | |
1983 | uint_t i, gcount; | |
1984 | nvlist_t *nvl; | |
1985 | vdev_t **vd; | |
1986 | boolean_t attempt_reopen; | |
1987 | ||
1988 | if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT, &nvl) != 0) | |
1989 | return; | |
1990 | ||
1991 | /* check that the config is complete */ | |
1992 | if (nvlist_lookup_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST, | |
1993 | &glist, &gcount) != 0) | |
1994 | return; | |
1995 | ||
b8d06fca | 1996 | vd = kmem_zalloc(gcount * sizeof (vdev_t *), KM_PUSHPAGE); |
428870ff BB |
1997 | |
1998 | /* attempt to online all the vdevs & validate */ | |
1999 | attempt_reopen = B_TRUE; | |
2000 | for (i = 0; i < gcount; i++) { | |
2001 | if (glist[i] == 0) /* vdev is hole */ | |
2002 | continue; | |
2003 | ||
2004 | vd[i] = spa_lookup_by_guid(spa, glist[i], B_FALSE); | |
2005 | if (vd[i] == NULL) { | |
2006 | /* | |
2007 | * Don't bother attempting to reopen the disks; | |
2008 | * just do the split. | |
2009 | */ | |
2010 | attempt_reopen = B_FALSE; | |
2011 | } else { | |
2012 | /* attempt to re-online it */ | |
2013 | vd[i]->vdev_offline = B_FALSE; | |
2014 | } | |
2015 | } | |
2016 | ||
2017 | if (attempt_reopen) { | |
2018 | vdev_reopen(spa->spa_root_vdev); | |
2019 | ||
2020 | /* check each device to see what state it's in */ | |
2021 | for (extracted = 0, i = 0; i < gcount; i++) { | |
2022 | if (vd[i] != NULL && | |
2023 | vd[i]->vdev_stat.vs_aux != VDEV_AUX_SPLIT_POOL) | |
2024 | break; | |
2025 | ++extracted; | |
2026 | } | |
2027 | } | |
2028 | ||
2029 | /* | |
2030 | * If every disk has been moved to the new pool, or if we never | |
2031 | * even attempted to look at them, then we split them off for | |
2032 | * good. | |
2033 | */ | |
2034 | if (!attempt_reopen || gcount == extracted) { | |
2035 | for (i = 0; i < gcount; i++) | |
2036 | if (vd[i] != NULL) | |
2037 | vdev_split(vd[i]); | |
2038 | vdev_reopen(spa->spa_root_vdev); | |
2039 | } | |
2040 | ||
2041 | kmem_free(vd, gcount * sizeof (vdev_t *)); | |
2042 | } | |
2043 | ||
2044 | static int | |
2045 | spa_load(spa_t *spa, spa_load_state_t state, spa_import_type_t type, | |
2046 | boolean_t mosconfig) | |
2047 | { | |
2048 | nvlist_t *config = spa->spa_config; | |
2049 | char *ereport = FM_EREPORT_ZFS_POOL; | |
d96eb2b1 | 2050 | char *comment; |
428870ff BB |
2051 | int error; |
2052 | uint64_t pool_guid; | |
2053 | nvlist_t *nvl; | |
2054 | ||
2055 | if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pool_guid)) | |
2e528b49 | 2056 | return (SET_ERROR(EINVAL)); |
428870ff | 2057 | |
d96eb2b1 DM |
2058 | ASSERT(spa->spa_comment == NULL); |
2059 | if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0) | |
2060 | spa->spa_comment = spa_strdup(comment); | |
2061 | ||
428870ff BB |
2062 | /* |
2063 | * Versioning wasn't explicitly added to the label until later, so if | |
2064 | * it's not present treat it as the initial version. | |
2065 | */ | |
2066 | if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, | |
2067 | &spa->spa_ubsync.ub_version) != 0) | |
2068 | spa->spa_ubsync.ub_version = SPA_VERSION_INITIAL; | |
2069 | ||
2070 | (void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, | |
2071 | &spa->spa_config_txg); | |
2072 | ||
2073 | if ((state == SPA_LOAD_IMPORT || state == SPA_LOAD_TRYIMPORT) && | |
2074 | spa_guid_exists(pool_guid, 0)) { | |
2e528b49 | 2075 | error = SET_ERROR(EEXIST); |
428870ff | 2076 | } else { |
3541dc6d | 2077 | spa->spa_config_guid = pool_guid; |
428870ff BB |
2078 | |
2079 | if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT, | |
2080 | &nvl) == 0) { | |
2081 | VERIFY(nvlist_dup(nvl, &spa->spa_config_splitting, | |
b8d06fca | 2082 | KM_PUSHPAGE) == 0); |
428870ff BB |
2083 | } |
2084 | ||
9ae529ec CS |
2085 | nvlist_free(spa->spa_load_info); |
2086 | spa->spa_load_info = fnvlist_alloc(); | |
2087 | ||
572e2857 | 2088 | gethrestime(&spa->spa_loaded_ts); |
428870ff BB |
2089 | error = spa_load_impl(spa, pool_guid, config, state, type, |
2090 | mosconfig, &ereport); | |
2091 | } | |
2092 | ||
2093 | spa->spa_minref = refcount_count(&spa->spa_refcount); | |
572e2857 BB |
2094 | if (error) { |
2095 | if (error != EEXIST) { | |
2096 | spa->spa_loaded_ts.tv_sec = 0; | |
2097 | spa->spa_loaded_ts.tv_nsec = 0; | |
2098 | } | |
2099 | if (error != EBADF) { | |
2100 | zfs_ereport_post(ereport, spa, NULL, NULL, 0, 0); | |
2101 | } | |
2102 | } | |
428870ff BB |
2103 | spa->spa_load_state = error ? SPA_LOAD_ERROR : SPA_LOAD_NONE; |
2104 | spa->spa_ena = 0; | |
2105 | ||
2106 | return (error); | |
2107 | } | |
2108 | ||
2109 | /* | |
2110 | * Load an existing storage pool, using the pool's builtin spa_config as a | |
2111 | * source of configuration information. | |
2112 | */ | |
bf701a83 BB |
2113 | __attribute__((always_inline)) |
2114 | static inline int | |
428870ff BB |
2115 | spa_load_impl(spa_t *spa, uint64_t pool_guid, nvlist_t *config, |
2116 | spa_load_state_t state, spa_import_type_t type, boolean_t mosconfig, | |
2117 | char **ereport) | |
2118 | { | |
2119 | int error = 0; | |
2120 | nvlist_t *nvroot = NULL; | |
9ae529ec | 2121 | nvlist_t *label; |
428870ff BB |
2122 | vdev_t *rvd; |
2123 | uberblock_t *ub = &spa->spa_uberblock; | |
572e2857 | 2124 | uint64_t children, config_cache_txg = spa->spa_config_txg; |
428870ff BB |
2125 | int orig_mode = spa->spa_mode; |
2126 | int parse; | |
2127 | uint64_t obj; | |
9ae529ec | 2128 | boolean_t missing_feat_write = B_FALSE; |
428870ff BB |
2129 | |
2130 | /* | |
2131 | * If this is an untrusted config, access the pool in read-only mode. | |
2132 | * This prevents things like resilvering recently removed devices. | |
2133 | */ | |
2134 | if (!mosconfig) | |
2135 | spa->spa_mode = FREAD; | |
2136 | ||
2137 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
2138 | ||
2139 | spa->spa_load_state = state; | |
2140 | ||
2141 | if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot)) | |
2e528b49 | 2142 | return (SET_ERROR(EINVAL)); |
428870ff BB |
2143 | |
2144 | parse = (type == SPA_IMPORT_EXISTING ? | |
2145 | VDEV_ALLOC_LOAD : VDEV_ALLOC_SPLIT); | |
2146 | ||
2147 | /* | |
2148 | * Create "The Godfather" zio to hold all async IOs | |
2149 | */ | |
2150 | spa->spa_async_zio_root = zio_root(spa, NULL, NULL, | |
2151 | ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_GODFATHER); | |
2152 | ||
2153 | /* | |
2154 | * Parse the configuration into a vdev tree. We explicitly set the | |
2155 | * value that will be returned by spa_version() since parsing the | |
2156 | * configuration requires knowing the version number. | |
2157 | */ | |
2158 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); | |
2159 | error = spa_config_parse(spa, &rvd, nvroot, NULL, 0, parse); | |
2160 | spa_config_exit(spa, SCL_ALL, FTAG); | |
2161 | ||
2162 | if (error != 0) | |
2163 | return (error); | |
2164 | ||
2165 | ASSERT(spa->spa_root_vdev == rvd); | |
2166 | ||
2167 | if (type != SPA_IMPORT_ASSEMBLE) { | |
2168 | ASSERT(spa_guid(spa) == pool_guid); | |
2169 | } | |
2170 | ||
2171 | /* | |
2172 | * Try to open all vdevs, loading each label in the process. | |
2173 | */ | |
2174 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); | |
2175 | error = vdev_open(rvd); | |
2176 | spa_config_exit(spa, SCL_ALL, FTAG); | |
2177 | if (error != 0) | |
2178 | return (error); | |
2179 | ||
2180 | /* | |
2181 | * We need to validate the vdev labels against the configuration that | |
2182 | * we have in hand, which is dependent on the setting of mosconfig. If | |
2183 | * mosconfig is true then we're validating the vdev labels based on | |
2184 | * that config. Otherwise, we're validating against the cached config | |
2185 | * (zpool.cache) that was read when we loaded the zfs module, and then | |
2186 | * later we will recursively call spa_load() and validate against | |
2187 | * the vdev config. | |
2188 | * | |
2189 | * If we're assembling a new pool that's been split off from an | |
2190 | * existing pool, the labels haven't yet been updated so we skip | |
2191 | * validation for now. | |
2192 | */ | |
2193 | if (type != SPA_IMPORT_ASSEMBLE) { | |
2194 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); | |
c7f2d69d | 2195 | error = vdev_validate(rvd, mosconfig); |
428870ff BB |
2196 | spa_config_exit(spa, SCL_ALL, FTAG); |
2197 | ||
2198 | if (error != 0) | |
2199 | return (error); | |
2200 | ||
2201 | if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) | |
2e528b49 | 2202 | return (SET_ERROR(ENXIO)); |
428870ff BB |
2203 | } |
2204 | ||
2205 | /* | |
2206 | * Find the best uberblock. | |
2207 | */ | |
9ae529ec | 2208 | vdev_uberblock_load(rvd, ub, &label); |
428870ff BB |
2209 | |
2210 | /* | |
2211 | * If we weren't able to find a single valid uberblock, return failure. | |
2212 | */ | |
9ae529ec CS |
2213 | if (ub->ub_txg == 0) { |
2214 | nvlist_free(label); | |
428870ff | 2215 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, ENXIO)); |
9ae529ec | 2216 | } |
428870ff BB |
2217 | |
2218 | /* | |
9ae529ec | 2219 | * If the pool has an unsupported version we can't open it. |
428870ff | 2220 | */ |
9ae529ec CS |
2221 | if (!SPA_VERSION_IS_SUPPORTED(ub->ub_version)) { |
2222 | nvlist_free(label); | |
428870ff | 2223 | return (spa_vdev_err(rvd, VDEV_AUX_VERSION_NEWER, ENOTSUP)); |
9ae529ec CS |
2224 | } |
2225 | ||
2226 | if (ub->ub_version >= SPA_VERSION_FEATURES) { | |
2227 | nvlist_t *features; | |
2228 | ||
2229 | /* | |
2230 | * If we weren't able to find what's necessary for reading the | |
2231 | * MOS in the label, return failure. | |
2232 | */ | |
2233 | if (label == NULL || nvlist_lookup_nvlist(label, | |
2234 | ZPOOL_CONFIG_FEATURES_FOR_READ, &features) != 0) { | |
2235 | nvlist_free(label); | |
2236 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, | |
2237 | ENXIO)); | |
2238 | } | |
2239 | ||
2240 | /* | |
2241 | * Update our in-core representation with the definitive values | |
2242 | * from the label. | |
2243 | */ | |
2244 | nvlist_free(spa->spa_label_features); | |
2245 | VERIFY(nvlist_dup(features, &spa->spa_label_features, 0) == 0); | |
2246 | } | |
2247 | ||
2248 | nvlist_free(label); | |
2249 | ||
2250 | /* | |
2251 | * Look through entries in the label nvlist's features_for_read. If | |
2252 | * there is a feature listed there which we don't understand then we | |
2253 | * cannot open a pool. | |
2254 | */ | |
2255 | if (ub->ub_version >= SPA_VERSION_FEATURES) { | |
2256 | nvlist_t *unsup_feat; | |
2257 | nvpair_t *nvp; | |
2258 | ||
2259 | VERIFY(nvlist_alloc(&unsup_feat, NV_UNIQUE_NAME, KM_SLEEP) == | |
2260 | 0); | |
2261 | ||
2262 | for (nvp = nvlist_next_nvpair(spa->spa_label_features, NULL); | |
2263 | nvp != NULL; | |
2264 | nvp = nvlist_next_nvpair(spa->spa_label_features, nvp)) { | |
2265 | if (!zfeature_is_supported(nvpair_name(nvp))) { | |
2266 | VERIFY(nvlist_add_string(unsup_feat, | |
2267 | nvpair_name(nvp), "") == 0); | |
2268 | } | |
2269 | } | |
2270 | ||
2271 | if (!nvlist_empty(unsup_feat)) { | |
2272 | VERIFY(nvlist_add_nvlist(spa->spa_load_info, | |
2273 | ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat) == 0); | |
2274 | nvlist_free(unsup_feat); | |
2275 | return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT, | |
2276 | ENOTSUP)); | |
2277 | } | |
2278 | ||
2279 | nvlist_free(unsup_feat); | |
2280 | } | |
428870ff BB |
2281 | |
2282 | /* | |
2283 | * If the vdev guid sum doesn't match the uberblock, we have an | |
572e2857 BB |
2284 | * incomplete configuration. We first check to see if the pool |
2285 | * is aware of the complete config (i.e ZPOOL_CONFIG_VDEV_CHILDREN). | |
2286 | * If it is, defer the vdev_guid_sum check till later so we | |
2287 | * can handle missing vdevs. | |
428870ff | 2288 | */ |
572e2857 BB |
2289 | if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_VDEV_CHILDREN, |
2290 | &children) != 0 && mosconfig && type != SPA_IMPORT_ASSEMBLE && | |
428870ff BB |
2291 | rvd->vdev_guid_sum != ub->ub_guid_sum) |
2292 | return (spa_vdev_err(rvd, VDEV_AUX_BAD_GUID_SUM, ENXIO)); | |
2293 | ||
2294 | if (type != SPA_IMPORT_ASSEMBLE && spa->spa_config_splitting) { | |
2295 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); | |
2296 | spa_try_repair(spa, config); | |
2297 | spa_config_exit(spa, SCL_ALL, FTAG); | |
2298 | nvlist_free(spa->spa_config_splitting); | |
2299 | spa->spa_config_splitting = NULL; | |
2300 | } | |
2301 | ||
2302 | /* | |
2303 | * Initialize internal SPA structures. | |
2304 | */ | |
2305 | spa->spa_state = POOL_STATE_ACTIVE; | |
2306 | spa->spa_ubsync = spa->spa_uberblock; | |
2307 | spa->spa_verify_min_txg = spa->spa_extreme_rewind ? | |
2308 | TXG_INITIAL - 1 : spa_last_synced_txg(spa) - TXG_DEFER_SIZE - 1; | |
2309 | spa->spa_first_txg = spa->spa_last_ubsync_txg ? | |
2310 | spa->spa_last_ubsync_txg : spa_last_synced_txg(spa) + 1; | |
2311 | spa->spa_claim_max_txg = spa->spa_first_txg; | |
2312 | spa->spa_prev_software_version = ub->ub_software_version; | |
2313 | ||
9ae529ec | 2314 | error = dsl_pool_init(spa, spa->spa_first_txg, &spa->spa_dsl_pool); |
428870ff BB |
2315 | if (error) |
2316 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
2317 | spa->spa_meta_objset = spa->spa_dsl_pool->dp_meta_objset; | |
2318 | ||
2319 | if (spa_dir_prop(spa, DMU_POOL_CONFIG, &spa->spa_config_object) != 0) | |
2320 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
2321 | ||
9ae529ec CS |
2322 | if (spa_version(spa) >= SPA_VERSION_FEATURES) { |
2323 | boolean_t missing_feat_read = B_FALSE; | |
b9b24bb4 | 2324 | nvlist_t *unsup_feat, *enabled_feat; |
9ae529ec CS |
2325 | |
2326 | if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_READ, | |
2327 | &spa->spa_feat_for_read_obj) != 0) { | |
2328 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
2329 | } | |
2330 | ||
2331 | if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_WRITE, | |
2332 | &spa->spa_feat_for_write_obj) != 0) { | |
2333 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
2334 | } | |
2335 | ||
2336 | if (spa_dir_prop(spa, DMU_POOL_FEATURE_DESCRIPTIONS, | |
2337 | &spa->spa_feat_desc_obj) != 0) { | |
2338 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
2339 | } | |
2340 | ||
b9b24bb4 CS |
2341 | enabled_feat = fnvlist_alloc(); |
2342 | unsup_feat = fnvlist_alloc(); | |
9ae529ec CS |
2343 | |
2344 | if (!feature_is_supported(spa->spa_meta_objset, | |
2345 | spa->spa_feat_for_read_obj, spa->spa_feat_desc_obj, | |
b9b24bb4 | 2346 | unsup_feat, enabled_feat)) |
9ae529ec CS |
2347 | missing_feat_read = B_TRUE; |
2348 | ||
2349 | if (spa_writeable(spa) || state == SPA_LOAD_TRYIMPORT) { | |
2350 | if (!feature_is_supported(spa->spa_meta_objset, | |
2351 | spa->spa_feat_for_write_obj, spa->spa_feat_desc_obj, | |
b9b24bb4 | 2352 | unsup_feat, enabled_feat)) { |
9ae529ec | 2353 | missing_feat_write = B_TRUE; |
b9b24bb4 | 2354 | } |
9ae529ec CS |
2355 | } |
2356 | ||
b9b24bb4 CS |
2357 | fnvlist_add_nvlist(spa->spa_load_info, |
2358 | ZPOOL_CONFIG_ENABLED_FEAT, enabled_feat); | |
2359 | ||
9ae529ec | 2360 | if (!nvlist_empty(unsup_feat)) { |
b9b24bb4 CS |
2361 | fnvlist_add_nvlist(spa->spa_load_info, |
2362 | ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat); | |
9ae529ec CS |
2363 | } |
2364 | ||
b9b24bb4 CS |
2365 | fnvlist_free(enabled_feat); |
2366 | fnvlist_free(unsup_feat); | |
9ae529ec CS |
2367 | |
2368 | if (!missing_feat_read) { | |
2369 | fnvlist_add_boolean(spa->spa_load_info, | |
2370 | ZPOOL_CONFIG_CAN_RDONLY); | |
2371 | } | |
2372 | ||
2373 | /* | |
2374 | * If the state is SPA_LOAD_TRYIMPORT, our objective is | |
2375 | * twofold: to determine whether the pool is available for | |
2376 | * import in read-write mode and (if it is not) whether the | |
2377 | * pool is available for import in read-only mode. If the pool | |
2378 | * is available for import in read-write mode, it is displayed | |
2379 | * as available in userland; if it is not available for import | |
2380 | * in read-only mode, it is displayed as unavailable in | |
2381 | * userland. If the pool is available for import in read-only | |
2382 | * mode but not read-write mode, it is displayed as unavailable | |
2383 | * in userland with a special note that the pool is actually | |
2384 | * available for open in read-only mode. | |
2385 | * | |
2386 | * As a result, if the state is SPA_LOAD_TRYIMPORT and we are | |
2387 | * missing a feature for write, we must first determine whether | |
2388 | * the pool can be opened read-only before returning to | |
2389 | * userland in order to know whether to display the | |
2390 | * abovementioned note. | |
2391 | */ | |
2392 | if (missing_feat_read || (missing_feat_write && | |
2393 | spa_writeable(spa))) { | |
2394 | return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT, | |
2395 | ENOTSUP)); | |
2396 | } | |
2397 | } | |
2398 | ||
2399 | spa->spa_is_initializing = B_TRUE; | |
2400 | error = dsl_pool_open(spa->spa_dsl_pool); | |
2401 | spa->spa_is_initializing = B_FALSE; | |
2402 | if (error != 0) | |
2403 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
2404 | ||
428870ff BB |
2405 | if (!mosconfig) { |
2406 | uint64_t hostid; | |
2407 | nvlist_t *policy = NULL, *nvconfig; | |
2408 | ||
2409 | if (load_nvlist(spa, spa->spa_config_object, &nvconfig) != 0) | |
2410 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
2411 | ||
2412 | if (!spa_is_root(spa) && nvlist_lookup_uint64(nvconfig, | |
b128c09f | 2413 | ZPOOL_CONFIG_HOSTID, &hostid) == 0) { |
34dc7c2f BB |
2414 | char *hostname; |
2415 | unsigned long myhostid = 0; | |
2416 | ||
428870ff | 2417 | VERIFY(nvlist_lookup_string(nvconfig, |
34dc7c2f BB |
2418 | ZPOOL_CONFIG_HOSTNAME, &hostname) == 0); |
2419 | ||
d164b209 BB |
2420 | #ifdef _KERNEL |
2421 | myhostid = zone_get_hostid(NULL); | |
2422 | #else /* _KERNEL */ | |
2423 | /* | |
2424 | * We're emulating the system's hostid in userland, so | |
2425 | * we can't use zone_get_hostid(). | |
2426 | */ | |
34dc7c2f | 2427 | (void) ddi_strtoul(hw_serial, NULL, 10, &myhostid); |
d164b209 | 2428 | #endif /* _KERNEL */ |
34dc7c2f | 2429 | if (hostid != 0 && myhostid != 0 && |
d164b209 | 2430 | hostid != myhostid) { |
428870ff | 2431 | nvlist_free(nvconfig); |
34dc7c2f BB |
2432 | cmn_err(CE_WARN, "pool '%s' could not be " |
2433 | "loaded as it was last accessed by " | |
b128c09f | 2434 | "another system (host: %s hostid: 0x%lx). " |
3cee2262 | 2435 | "See: http://zfsonlinux.org/msg/ZFS-8000-EY", |
b128c09f | 2436 | spa_name(spa), hostname, |
34dc7c2f | 2437 | (unsigned long)hostid); |
2e528b49 | 2438 | return (SET_ERROR(EBADF)); |
34dc7c2f BB |
2439 | } |
2440 | } | |
428870ff BB |
2441 | if (nvlist_lookup_nvlist(spa->spa_config, |
2442 | ZPOOL_REWIND_POLICY, &policy) == 0) | |
2443 | VERIFY(nvlist_add_nvlist(nvconfig, | |
2444 | ZPOOL_REWIND_POLICY, policy) == 0); | |
34dc7c2f | 2445 | |
428870ff | 2446 | spa_config_set(spa, nvconfig); |
34dc7c2f BB |
2447 | spa_unload(spa); |
2448 | spa_deactivate(spa); | |
fb5f0bc8 | 2449 | spa_activate(spa, orig_mode); |
34dc7c2f | 2450 | |
428870ff | 2451 | return (spa_load(spa, state, SPA_IMPORT_EXISTING, B_TRUE)); |
34dc7c2f BB |
2452 | } |
2453 | ||
428870ff BB |
2454 | if (spa_dir_prop(spa, DMU_POOL_SYNC_BPOBJ, &obj) != 0) |
2455 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
2456 | error = bpobj_open(&spa->spa_deferred_bpobj, spa->spa_meta_objset, obj); | |
2457 | if (error != 0) | |
2458 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
34dc7c2f BB |
2459 | |
2460 | /* | |
2461 | * Load the bit that tells us to use the new accounting function | |
2462 | * (raid-z deflation). If we have an older pool, this will not | |
2463 | * be present. | |
2464 | */ | |
428870ff BB |
2465 | error = spa_dir_prop(spa, DMU_POOL_DEFLATE, &spa->spa_deflate); |
2466 | if (error != 0 && error != ENOENT) | |
2467 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
2468 | ||
2469 | error = spa_dir_prop(spa, DMU_POOL_CREATION_VERSION, | |
2470 | &spa->spa_creation_version); | |
2471 | if (error != 0 && error != ENOENT) | |
2472 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
34dc7c2f BB |
2473 | |
2474 | /* | |
2475 | * Load the persistent error log. If we have an older pool, this will | |
2476 | * not be present. | |
2477 | */ | |
428870ff BB |
2478 | error = spa_dir_prop(spa, DMU_POOL_ERRLOG_LAST, &spa->spa_errlog_last); |
2479 | if (error != 0 && error != ENOENT) | |
2480 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
34dc7c2f | 2481 | |
428870ff BB |
2482 | error = spa_dir_prop(spa, DMU_POOL_ERRLOG_SCRUB, |
2483 | &spa->spa_errlog_scrub); | |
2484 | if (error != 0 && error != ENOENT) | |
2485 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
34dc7c2f BB |
2486 | |
2487 | /* | |
2488 | * Load the history object. If we have an older pool, this | |
2489 | * will not be present. | |
2490 | */ | |
428870ff BB |
2491 | error = spa_dir_prop(spa, DMU_POOL_HISTORY, &spa->spa_history); |
2492 | if (error != 0 && error != ENOENT) | |
2493 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
2494 | ||
2495 | /* | |
2496 | * If we're assembling the pool from the split-off vdevs of | |
2497 | * an existing pool, we don't want to attach the spares & cache | |
2498 | * devices. | |
2499 | */ | |
34dc7c2f BB |
2500 | |
2501 | /* | |
2502 | * Load any hot spares for this pool. | |
2503 | */ | |
428870ff BB |
2504 | error = spa_dir_prop(spa, DMU_POOL_SPARES, &spa->spa_spares.sav_object); |
2505 | if (error != 0 && error != ENOENT) | |
2506 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
2507 | if (error == 0 && type != SPA_IMPORT_ASSEMBLE) { | |
34dc7c2f BB |
2508 | ASSERT(spa_version(spa) >= SPA_VERSION_SPARES); |
2509 | if (load_nvlist(spa, spa->spa_spares.sav_object, | |
428870ff BB |
2510 | &spa->spa_spares.sav_config) != 0) |
2511 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
34dc7c2f | 2512 | |
b128c09f | 2513 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f | 2514 | spa_load_spares(spa); |
b128c09f | 2515 | spa_config_exit(spa, SCL_ALL, FTAG); |
428870ff BB |
2516 | } else if (error == 0) { |
2517 | spa->spa_spares.sav_sync = B_TRUE; | |
34dc7c2f BB |
2518 | } |
2519 | ||
2520 | /* | |
2521 | * Load any level 2 ARC devices for this pool. | |
2522 | */ | |
428870ff | 2523 | error = spa_dir_prop(spa, DMU_POOL_L2CACHE, |
34dc7c2f | 2524 | &spa->spa_l2cache.sav_object); |
428870ff BB |
2525 | if (error != 0 && error != ENOENT) |
2526 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
2527 | if (error == 0 && type != SPA_IMPORT_ASSEMBLE) { | |
34dc7c2f BB |
2528 | ASSERT(spa_version(spa) >= SPA_VERSION_L2CACHE); |
2529 | if (load_nvlist(spa, spa->spa_l2cache.sav_object, | |
428870ff BB |
2530 | &spa->spa_l2cache.sav_config) != 0) |
2531 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
34dc7c2f | 2532 | |
b128c09f | 2533 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f | 2534 | spa_load_l2cache(spa); |
b128c09f | 2535 | spa_config_exit(spa, SCL_ALL, FTAG); |
428870ff BB |
2536 | } else if (error == 0) { |
2537 | spa->spa_l2cache.sav_sync = B_TRUE; | |
b128c09f BB |
2538 | } |
2539 | ||
34dc7c2f BB |
2540 | spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION); |
2541 | ||
428870ff BB |
2542 | error = spa_dir_prop(spa, DMU_POOL_PROPS, &spa->spa_pool_props_object); |
2543 | if (error && error != ENOENT) | |
2544 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
34dc7c2f BB |
2545 | |
2546 | if (error == 0) { | |
428870ff BB |
2547 | uint64_t autoreplace; |
2548 | ||
2549 | spa_prop_find(spa, ZPOOL_PROP_BOOTFS, &spa->spa_bootfs); | |
2550 | spa_prop_find(spa, ZPOOL_PROP_AUTOREPLACE, &autoreplace); | |
2551 | spa_prop_find(spa, ZPOOL_PROP_DELEGATION, &spa->spa_delegation); | |
2552 | spa_prop_find(spa, ZPOOL_PROP_FAILUREMODE, &spa->spa_failmode); | |
2553 | spa_prop_find(spa, ZPOOL_PROP_AUTOEXPAND, &spa->spa_autoexpand); | |
2554 | spa_prop_find(spa, ZPOOL_PROP_DEDUPDITTO, | |
2555 | &spa->spa_dedup_ditto); | |
2556 | ||
2557 | spa->spa_autoreplace = (autoreplace != 0); | |
34dc7c2f BB |
2558 | } |
2559 | ||
2560 | /* | |
2561 | * If the 'autoreplace' property is set, then post a resource notifying | |
2562 | * the ZFS DE that it should not issue any faults for unopenable | |
2563 | * devices. We also iterate over the vdevs, and post a sysevent for any | |
2564 | * unopenable vdevs so that the normal autoreplace handler can take | |
2565 | * over. | |
2566 | */ | |
428870ff | 2567 | if (spa->spa_autoreplace && state != SPA_LOAD_TRYIMPORT) { |
34dc7c2f | 2568 | spa_check_removed(spa->spa_root_vdev); |
428870ff BB |
2569 | /* |
2570 | * For the import case, this is done in spa_import(), because | |
2571 | * at this point we're using the spare definitions from | |
2572 | * the MOS config, not necessarily from the userland config. | |
2573 | */ | |
2574 | if (state != SPA_LOAD_IMPORT) { | |
2575 | spa_aux_check_removed(&spa->spa_spares); | |
2576 | spa_aux_check_removed(&spa->spa_l2cache); | |
2577 | } | |
2578 | } | |
34dc7c2f BB |
2579 | |
2580 | /* | |
2581 | * Load the vdev state for all toplevel vdevs. | |
2582 | */ | |
2583 | vdev_load(rvd); | |
2584 | ||
2585 | /* | |
2586 | * Propagate the leaf DTLs we just loaded all the way up the tree. | |
2587 | */ | |
b128c09f | 2588 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f | 2589 | vdev_dtl_reassess(rvd, 0, 0, B_FALSE); |
b128c09f | 2590 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f | 2591 | |
428870ff BB |
2592 | /* |
2593 | * Load the DDTs (dedup tables). | |
2594 | */ | |
2595 | error = ddt_load(spa); | |
2596 | if (error != 0) | |
2597 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
2598 | ||
2599 | spa_update_dspace(spa); | |
2600 | ||
428870ff | 2601 | /* |
572e2857 BB |
2602 | * Validate the config, using the MOS config to fill in any |
2603 | * information which might be missing. If we fail to validate | |
2604 | * the config then declare the pool unfit for use. If we're | |
2605 | * assembling a pool from a split, the log is not transferred | |
2606 | * over. | |
428870ff BB |
2607 | */ |
2608 | if (type != SPA_IMPORT_ASSEMBLE) { | |
2609 | nvlist_t *nvconfig; | |
2610 | ||
2611 | if (load_nvlist(spa, spa->spa_config_object, &nvconfig) != 0) | |
2612 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); | |
2613 | ||
572e2857 BB |
2614 | if (!spa_config_valid(spa, nvconfig)) { |
2615 | nvlist_free(nvconfig); | |
2616 | return (spa_vdev_err(rvd, VDEV_AUX_BAD_GUID_SUM, | |
2617 | ENXIO)); | |
2618 | } | |
428870ff BB |
2619 | nvlist_free(nvconfig); |
2620 | ||
572e2857 | 2621 | /* |
9ae529ec | 2622 | * Now that we've validated the config, check the state of the |
572e2857 BB |
2623 | * root vdev. If it can't be opened, it indicates one or |
2624 | * more toplevel vdevs are faulted. | |
2625 | */ | |
2626 | if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) | |
2e528b49 | 2627 | return (SET_ERROR(ENXIO)); |
572e2857 | 2628 | |
428870ff BB |
2629 | if (spa_check_logs(spa)) { |
2630 | *ereport = FM_EREPORT_ZFS_LOG_REPLAY; | |
2631 | return (spa_vdev_err(rvd, VDEV_AUX_BAD_LOG, ENXIO)); | |
2632 | } | |
2633 | } | |
2634 | ||
9ae529ec CS |
2635 | if (missing_feat_write) { |
2636 | ASSERT(state == SPA_LOAD_TRYIMPORT); | |
2637 | ||
2638 | /* | |
2639 | * At this point, we know that we can open the pool in | |
2640 | * read-only mode but not read-write mode. We now have enough | |
2641 | * information and can return to userland. | |
2642 | */ | |
2643 | return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT, ENOTSUP)); | |
2644 | } | |
2645 | ||
572e2857 BB |
2646 | /* |
2647 | * We've successfully opened the pool, verify that we're ready | |
2648 | * to start pushing transactions. | |
2649 | */ | |
2650 | if (state != SPA_LOAD_TRYIMPORT) { | |
c65aa5b2 | 2651 | if ((error = spa_load_verify(spa))) |
572e2857 BB |
2652 | return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, |
2653 | error)); | |
2654 | } | |
2655 | ||
428870ff BB |
2656 | if (spa_writeable(spa) && (state == SPA_LOAD_RECOVER || |
2657 | spa->spa_load_max_txg == UINT64_MAX)) { | |
34dc7c2f BB |
2658 | dmu_tx_t *tx; |
2659 | int need_update = B_FALSE; | |
d6320ddb | 2660 | int c; |
fb5f0bc8 BB |
2661 | |
2662 | ASSERT(state != SPA_LOAD_TRYIMPORT); | |
34dc7c2f BB |
2663 | |
2664 | /* | |
2665 | * Claim log blocks that haven't been committed yet. | |
2666 | * This must all happen in a single txg. | |
428870ff BB |
2667 | * Note: spa_claim_max_txg is updated by spa_claim_notify(), |
2668 | * invoked from zil_claim_log_block()'s i/o done callback. | |
2669 | * Price of rollback is that we abandon the log. | |
34dc7c2f | 2670 | */ |
428870ff BB |
2671 | spa->spa_claiming = B_TRUE; |
2672 | ||
34dc7c2f BB |
2673 | tx = dmu_tx_create_assigned(spa_get_dsl(spa), |
2674 | spa_first_txg(spa)); | |
b128c09f | 2675 | (void) dmu_objset_find(spa_name(spa), |
34dc7c2f BB |
2676 | zil_claim, tx, DS_FIND_CHILDREN); |
2677 | dmu_tx_commit(tx); | |
2678 | ||
428870ff BB |
2679 | spa->spa_claiming = B_FALSE; |
2680 | ||
2681 | spa_set_log_state(spa, SPA_LOG_GOOD); | |
34dc7c2f BB |
2682 | spa->spa_sync_on = B_TRUE; |
2683 | txg_sync_start(spa->spa_dsl_pool); | |
2684 | ||
2685 | /* | |
428870ff BB |
2686 | * Wait for all claims to sync. We sync up to the highest |
2687 | * claimed log block birth time so that claimed log blocks | |
2688 | * don't appear to be from the future. spa_claim_max_txg | |
2689 | * will have been set for us by either zil_check_log_chain() | |
2690 | * (invoked from spa_check_logs()) or zil_claim() above. | |
34dc7c2f | 2691 | */ |
428870ff | 2692 | txg_wait_synced(spa->spa_dsl_pool, spa->spa_claim_max_txg); |
34dc7c2f BB |
2693 | |
2694 | /* | |
2695 | * If the config cache is stale, or we have uninitialized | |
2696 | * metaslabs (see spa_vdev_add()), then update the config. | |
45d1cae3 | 2697 | * |
572e2857 | 2698 | * If this is a verbatim import, trust the current |
45d1cae3 | 2699 | * in-core spa_config and update the disk labels. |
34dc7c2f BB |
2700 | */ |
2701 | if (config_cache_txg != spa->spa_config_txg || | |
572e2857 BB |
2702 | state == SPA_LOAD_IMPORT || |
2703 | state == SPA_LOAD_RECOVER || | |
2704 | (spa->spa_import_flags & ZFS_IMPORT_VERBATIM)) | |
34dc7c2f BB |
2705 | need_update = B_TRUE; |
2706 | ||
d6320ddb | 2707 | for (c = 0; c < rvd->vdev_children; c++) |
34dc7c2f BB |
2708 | if (rvd->vdev_child[c]->vdev_ms_array == 0) |
2709 | need_update = B_TRUE; | |
2710 | ||
2711 | /* | |
2712 | * Update the config cache asychronously in case we're the | |
2713 | * root pool, in which case the config cache isn't writable yet. | |
2714 | */ | |
2715 | if (need_update) | |
2716 | spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE); | |
fb5f0bc8 BB |
2717 | |
2718 | /* | |
2719 | * Check all DTLs to see if anything needs resilvering. | |
2720 | */ | |
428870ff BB |
2721 | if (!dsl_scan_resilvering(spa->spa_dsl_pool) && |
2722 | vdev_resilver_needed(rvd, NULL, NULL)) | |
fb5f0bc8 | 2723 | spa_async_request(spa, SPA_ASYNC_RESILVER); |
428870ff | 2724 | |
6f1ffb06 MA |
2725 | /* |
2726 | * Log the fact that we booted up (so that we can detect if | |
2727 | * we rebooted in the middle of an operation). | |
2728 | */ | |
2729 | spa_history_log_version(spa, "open"); | |
2730 | ||
428870ff BB |
2731 | /* |
2732 | * Delete any inconsistent datasets. | |
2733 | */ | |
2734 | (void) dmu_objset_find(spa_name(spa), | |
2735 | dsl_destroy_inconsistent, NULL, DS_FIND_CHILDREN); | |
2736 | ||
2737 | /* | |
2738 | * Clean up any stale temporary dataset userrefs. | |
2739 | */ | |
2740 | dsl_pool_clean_tmp_userrefs(spa->spa_dsl_pool); | |
34dc7c2f BB |
2741 | } |
2742 | ||
428870ff BB |
2743 | return (0); |
2744 | } | |
34dc7c2f | 2745 | |
428870ff BB |
2746 | static int |
2747 | spa_load_retry(spa_t *spa, spa_load_state_t state, int mosconfig) | |
2748 | { | |
572e2857 BB |
2749 | int mode = spa->spa_mode; |
2750 | ||
428870ff BB |
2751 | spa_unload(spa); |
2752 | spa_deactivate(spa); | |
2753 | ||
2754 | spa->spa_load_max_txg--; | |
2755 | ||
572e2857 | 2756 | spa_activate(spa, mode); |
428870ff BB |
2757 | spa_async_suspend(spa); |
2758 | ||
2759 | return (spa_load(spa, state, SPA_IMPORT_EXISTING, mosconfig)); | |
2760 | } | |
2761 | ||
9ae529ec CS |
2762 | /* |
2763 | * If spa_load() fails this function will try loading prior txg's. If | |
2764 | * 'state' is SPA_LOAD_RECOVER and one of these loads succeeds the pool | |
2765 | * will be rewound to that txg. If 'state' is not SPA_LOAD_RECOVER this | |
2766 | * function will not rewind the pool and will return the same error as | |
2767 | * spa_load(). | |
2768 | */ | |
428870ff BB |
2769 | static int |
2770 | spa_load_best(spa_t *spa, spa_load_state_t state, int mosconfig, | |
2771 | uint64_t max_request, int rewind_flags) | |
2772 | { | |
9ae529ec | 2773 | nvlist_t *loadinfo = NULL; |
428870ff BB |
2774 | nvlist_t *config = NULL; |
2775 | int load_error, rewind_error; | |
2776 | uint64_t safe_rewind_txg; | |
2777 | uint64_t min_txg; | |
2778 | ||
2779 | if (spa->spa_load_txg && state == SPA_LOAD_RECOVER) { | |
2780 | spa->spa_load_max_txg = spa->spa_load_txg; | |
2781 | spa_set_log_state(spa, SPA_LOG_CLEAR); | |
2782 | } else { | |
2783 | spa->spa_load_max_txg = max_request; | |
2784 | } | |
2785 | ||
2786 | load_error = rewind_error = spa_load(spa, state, SPA_IMPORT_EXISTING, | |
2787 | mosconfig); | |
2788 | if (load_error == 0) | |
2789 | return (0); | |
2790 | ||
2791 | if (spa->spa_root_vdev != NULL) | |
2792 | config = spa_config_generate(spa, NULL, -1ULL, B_TRUE); | |
2793 | ||
2794 | spa->spa_last_ubsync_txg = spa->spa_uberblock.ub_txg; | |
2795 | spa->spa_last_ubsync_txg_ts = spa->spa_uberblock.ub_timestamp; | |
2796 | ||
2797 | if (rewind_flags & ZPOOL_NEVER_REWIND) { | |
2798 | nvlist_free(config); | |
2799 | return (load_error); | |
2800 | } | |
2801 | ||
9ae529ec CS |
2802 | if (state == SPA_LOAD_RECOVER) { |
2803 | /* Price of rolling back is discarding txgs, including log */ | |
428870ff | 2804 | spa_set_log_state(spa, SPA_LOG_CLEAR); |
9ae529ec CS |
2805 | } else { |
2806 | /* | |
2807 | * If we aren't rolling back save the load info from our first | |
2808 | * import attempt so that we can restore it after attempting | |
2809 | * to rewind. | |
2810 | */ | |
2811 | loadinfo = spa->spa_load_info; | |
2812 | spa->spa_load_info = fnvlist_alloc(); | |
2813 | } | |
428870ff BB |
2814 | |
2815 | spa->spa_load_max_txg = spa->spa_last_ubsync_txg; | |
2816 | safe_rewind_txg = spa->spa_last_ubsync_txg - TXG_DEFER_SIZE; | |
2817 | min_txg = (rewind_flags & ZPOOL_EXTREME_REWIND) ? | |
2818 | TXG_INITIAL : safe_rewind_txg; | |
2819 | ||
2820 | /* | |
2821 | * Continue as long as we're finding errors, we're still within | |
2822 | * the acceptable rewind range, and we're still finding uberblocks | |
2823 | */ | |
2824 | while (rewind_error && spa->spa_uberblock.ub_txg >= min_txg && | |
2825 | spa->spa_uberblock.ub_txg <= spa->spa_load_max_txg) { | |
2826 | if (spa->spa_load_max_txg < safe_rewind_txg) | |
2827 | spa->spa_extreme_rewind = B_TRUE; | |
2828 | rewind_error = spa_load_retry(spa, state, mosconfig); | |
2829 | } | |
2830 | ||
428870ff BB |
2831 | spa->spa_extreme_rewind = B_FALSE; |
2832 | spa->spa_load_max_txg = UINT64_MAX; | |
2833 | ||
2834 | if (config && (rewind_error || state != SPA_LOAD_RECOVER)) | |
2835 | spa_config_set(spa, config); | |
2836 | ||
9ae529ec CS |
2837 | if (state == SPA_LOAD_RECOVER) { |
2838 | ASSERT3P(loadinfo, ==, NULL); | |
2839 | return (rewind_error); | |
2840 | } else { | |
2841 | /* Store the rewind info as part of the initial load info */ | |
2842 | fnvlist_add_nvlist(loadinfo, ZPOOL_CONFIG_REWIND_INFO, | |
2843 | spa->spa_load_info); | |
2844 | ||
2845 | /* Restore the initial load info */ | |
2846 | fnvlist_free(spa->spa_load_info); | |
2847 | spa->spa_load_info = loadinfo; | |
2848 | ||
2849 | return (load_error); | |
2850 | } | |
34dc7c2f BB |
2851 | } |
2852 | ||
2853 | /* | |
2854 | * Pool Open/Import | |
2855 | * | |
2856 | * The import case is identical to an open except that the configuration is sent | |
2857 | * down from userland, instead of grabbed from the configuration cache. For the | |
2858 | * case of an open, the pool configuration will exist in the | |
2859 | * POOL_STATE_UNINITIALIZED state. | |
2860 | * | |
2861 | * The stats information (gen/count/ustats) is used to gather vdev statistics at | |
2862 | * the same time open the pool, without having to keep around the spa_t in some | |
2863 | * ambiguous state. | |
2864 | */ | |
2865 | static int | |
428870ff BB |
2866 | spa_open_common(const char *pool, spa_t **spapp, void *tag, nvlist_t *nvpolicy, |
2867 | nvlist_t **config) | |
34dc7c2f BB |
2868 | { |
2869 | spa_t *spa; | |
572e2857 | 2870 | spa_load_state_t state = SPA_LOAD_OPEN; |
34dc7c2f | 2871 | int error; |
34dc7c2f | 2872 | int locked = B_FALSE; |
526af785 | 2873 | int firstopen = B_FALSE; |
34dc7c2f BB |
2874 | |
2875 | *spapp = NULL; | |
2876 | ||
2877 | /* | |
2878 | * As disgusting as this is, we need to support recursive calls to this | |
2879 | * function because dsl_dir_open() is called during spa_load(), and ends | |
2880 | * up calling spa_open() again. The real fix is to figure out how to | |
2881 | * avoid dsl_dir_open() calling this in the first place. | |
2882 | */ | |
2883 | if (mutex_owner(&spa_namespace_lock) != curthread) { | |
2884 | mutex_enter(&spa_namespace_lock); | |
2885 | locked = B_TRUE; | |
2886 | } | |
2887 | ||
2888 | if ((spa = spa_lookup(pool)) == NULL) { | |
2889 | if (locked) | |
2890 | mutex_exit(&spa_namespace_lock); | |
2e528b49 | 2891 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 2892 | } |
428870ff | 2893 | |
34dc7c2f | 2894 | if (spa->spa_state == POOL_STATE_UNINITIALIZED) { |
428870ff BB |
2895 | zpool_rewind_policy_t policy; |
2896 | ||
526af785 PJD |
2897 | firstopen = B_TRUE; |
2898 | ||
428870ff BB |
2899 | zpool_get_rewind_policy(nvpolicy ? nvpolicy : spa->spa_config, |
2900 | &policy); | |
2901 | if (policy.zrp_request & ZPOOL_DO_REWIND) | |
2902 | state = SPA_LOAD_RECOVER; | |
34dc7c2f | 2903 | |
fb5f0bc8 | 2904 | spa_activate(spa, spa_mode_global); |
34dc7c2f | 2905 | |
428870ff BB |
2906 | if (state != SPA_LOAD_RECOVER) |
2907 | spa->spa_last_ubsync_txg = spa->spa_load_txg = 0; | |
2908 | ||
2909 | error = spa_load_best(spa, state, B_FALSE, policy.zrp_txg, | |
2910 | policy.zrp_request); | |
34dc7c2f BB |
2911 | |
2912 | if (error == EBADF) { | |
2913 | /* | |
2914 | * If vdev_validate() returns failure (indicated by | |
2915 | * EBADF), it indicates that one of the vdevs indicates | |
2916 | * that the pool has been exported or destroyed. If | |
2917 | * this is the case, the config cache is out of sync and | |
2918 | * we should remove the pool from the namespace. | |
2919 | */ | |
34dc7c2f BB |
2920 | spa_unload(spa); |
2921 | spa_deactivate(spa); | |
b128c09f | 2922 | spa_config_sync(spa, B_TRUE, B_TRUE); |
34dc7c2f | 2923 | spa_remove(spa); |
34dc7c2f BB |
2924 | if (locked) |
2925 | mutex_exit(&spa_namespace_lock); | |
2e528b49 | 2926 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2927 | } |
2928 | ||
2929 | if (error) { | |
2930 | /* | |
2931 | * We can't open the pool, but we still have useful | |
2932 | * information: the state of each vdev after the | |
2933 | * attempted vdev_open(). Return this to the user. | |
2934 | */ | |
572e2857 | 2935 | if (config != NULL && spa->spa_config) { |
428870ff | 2936 | VERIFY(nvlist_dup(spa->spa_config, config, |
b8d06fca | 2937 | KM_PUSHPAGE) == 0); |
572e2857 BB |
2938 | VERIFY(nvlist_add_nvlist(*config, |
2939 | ZPOOL_CONFIG_LOAD_INFO, | |
2940 | spa->spa_load_info) == 0); | |
2941 | } | |
34dc7c2f BB |
2942 | spa_unload(spa); |
2943 | spa_deactivate(spa); | |
428870ff | 2944 | spa->spa_last_open_failed = error; |
34dc7c2f BB |
2945 | if (locked) |
2946 | mutex_exit(&spa_namespace_lock); | |
2947 | *spapp = NULL; | |
2948 | return (error); | |
34dc7c2f | 2949 | } |
34dc7c2f BB |
2950 | } |
2951 | ||
2952 | spa_open_ref(spa, tag); | |
2953 | ||
b128c09f | 2954 | if (config != NULL) |
34dc7c2f | 2955 | *config = spa_config_generate(spa, NULL, -1ULL, B_TRUE); |
34dc7c2f | 2956 | |
572e2857 BB |
2957 | /* |
2958 | * If we've recovered the pool, pass back any information we | |
2959 | * gathered while doing the load. | |
2960 | */ | |
2961 | if (state == SPA_LOAD_RECOVER) { | |
2962 | VERIFY(nvlist_add_nvlist(*config, ZPOOL_CONFIG_LOAD_INFO, | |
2963 | spa->spa_load_info) == 0); | |
2964 | } | |
2965 | ||
428870ff BB |
2966 | if (locked) { |
2967 | spa->spa_last_open_failed = 0; | |
2968 | spa->spa_last_ubsync_txg = 0; | |
2969 | spa->spa_load_txg = 0; | |
2970 | mutex_exit(&spa_namespace_lock); | |
2971 | } | |
2972 | ||
526af785 PJD |
2973 | #ifdef _KERNEL |
2974 | if (firstopen) | |
2975 | zvol_create_minors(spa->spa_name); | |
2976 | #endif | |
2977 | ||
428870ff BB |
2978 | *spapp = spa; |
2979 | ||
34dc7c2f BB |
2980 | return (0); |
2981 | } | |
2982 | ||
428870ff BB |
2983 | int |
2984 | spa_open_rewind(const char *name, spa_t **spapp, void *tag, nvlist_t *policy, | |
2985 | nvlist_t **config) | |
2986 | { | |
2987 | return (spa_open_common(name, spapp, tag, policy, config)); | |
2988 | } | |
2989 | ||
34dc7c2f BB |
2990 | int |
2991 | spa_open(const char *name, spa_t **spapp, void *tag) | |
2992 | { | |
428870ff | 2993 | return (spa_open_common(name, spapp, tag, NULL, NULL)); |
34dc7c2f BB |
2994 | } |
2995 | ||
2996 | /* | |
2997 | * Lookup the given spa_t, incrementing the inject count in the process, | |
2998 | * preventing it from being exported or destroyed. | |
2999 | */ | |
3000 | spa_t * | |
3001 | spa_inject_addref(char *name) | |
3002 | { | |
3003 | spa_t *spa; | |
3004 | ||
3005 | mutex_enter(&spa_namespace_lock); | |
3006 | if ((spa = spa_lookup(name)) == NULL) { | |
3007 | mutex_exit(&spa_namespace_lock); | |
3008 | return (NULL); | |
3009 | } | |
3010 | spa->spa_inject_ref++; | |
3011 | mutex_exit(&spa_namespace_lock); | |
3012 | ||
3013 | return (spa); | |
3014 | } | |
3015 | ||
3016 | void | |
3017 | spa_inject_delref(spa_t *spa) | |
3018 | { | |
3019 | mutex_enter(&spa_namespace_lock); | |
3020 | spa->spa_inject_ref--; | |
3021 | mutex_exit(&spa_namespace_lock); | |
3022 | } | |
3023 | ||
3024 | /* | |
3025 | * Add spares device information to the nvlist. | |
3026 | */ | |
3027 | static void | |
3028 | spa_add_spares(spa_t *spa, nvlist_t *config) | |
3029 | { | |
3030 | nvlist_t **spares; | |
3031 | uint_t i, nspares; | |
3032 | nvlist_t *nvroot; | |
3033 | uint64_t guid; | |
3034 | vdev_stat_t *vs; | |
3035 | uint_t vsc; | |
3036 | uint64_t pool; | |
3037 | ||
9babb374 BB |
3038 | ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER)); |
3039 | ||
34dc7c2f BB |
3040 | if (spa->spa_spares.sav_count == 0) |
3041 | return; | |
3042 | ||
3043 | VERIFY(nvlist_lookup_nvlist(config, | |
3044 | ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); | |
3045 | VERIFY(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config, | |
3046 | ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0); | |
3047 | if (nspares != 0) { | |
3048 | VERIFY(nvlist_add_nvlist_array(nvroot, | |
3049 | ZPOOL_CONFIG_SPARES, spares, nspares) == 0); | |
3050 | VERIFY(nvlist_lookup_nvlist_array(nvroot, | |
3051 | ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0); | |
3052 | ||
3053 | /* | |
3054 | * Go through and find any spares which have since been | |
3055 | * repurposed as an active spare. If this is the case, update | |
3056 | * their status appropriately. | |
3057 | */ | |
3058 | for (i = 0; i < nspares; i++) { | |
3059 | VERIFY(nvlist_lookup_uint64(spares[i], | |
3060 | ZPOOL_CONFIG_GUID, &guid) == 0); | |
b128c09f BB |
3061 | if (spa_spare_exists(guid, &pool, NULL) && |
3062 | pool != 0ULL) { | |
34dc7c2f | 3063 | VERIFY(nvlist_lookup_uint64_array( |
428870ff | 3064 | spares[i], ZPOOL_CONFIG_VDEV_STATS, |
34dc7c2f BB |
3065 | (uint64_t **)&vs, &vsc) == 0); |
3066 | vs->vs_state = VDEV_STATE_CANT_OPEN; | |
3067 | vs->vs_aux = VDEV_AUX_SPARED; | |
3068 | } | |
3069 | } | |
3070 | } | |
3071 | } | |
3072 | ||
3073 | /* | |
3074 | * Add l2cache device information to the nvlist, including vdev stats. | |
3075 | */ | |
3076 | static void | |
3077 | spa_add_l2cache(spa_t *spa, nvlist_t *config) | |
3078 | { | |
3079 | nvlist_t **l2cache; | |
3080 | uint_t i, j, nl2cache; | |
3081 | nvlist_t *nvroot; | |
3082 | uint64_t guid; | |
3083 | vdev_t *vd; | |
3084 | vdev_stat_t *vs; | |
3085 | uint_t vsc; | |
3086 | ||
9babb374 BB |
3087 | ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER)); |
3088 | ||
34dc7c2f BB |
3089 | if (spa->spa_l2cache.sav_count == 0) |
3090 | return; | |
3091 | ||
34dc7c2f BB |
3092 | VERIFY(nvlist_lookup_nvlist(config, |
3093 | ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); | |
3094 | VERIFY(nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config, | |
3095 | ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0); | |
3096 | if (nl2cache != 0) { | |
3097 | VERIFY(nvlist_add_nvlist_array(nvroot, | |
3098 | ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache) == 0); | |
3099 | VERIFY(nvlist_lookup_nvlist_array(nvroot, | |
3100 | ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0); | |
3101 | ||
3102 | /* | |
3103 | * Update level 2 cache device stats. | |
3104 | */ | |
3105 | ||
3106 | for (i = 0; i < nl2cache; i++) { | |
3107 | VERIFY(nvlist_lookup_uint64(l2cache[i], | |
3108 | ZPOOL_CONFIG_GUID, &guid) == 0); | |
3109 | ||
3110 | vd = NULL; | |
3111 | for (j = 0; j < spa->spa_l2cache.sav_count; j++) { | |
3112 | if (guid == | |
3113 | spa->spa_l2cache.sav_vdevs[j]->vdev_guid) { | |
3114 | vd = spa->spa_l2cache.sav_vdevs[j]; | |
3115 | break; | |
3116 | } | |
3117 | } | |
3118 | ASSERT(vd != NULL); | |
3119 | ||
3120 | VERIFY(nvlist_lookup_uint64_array(l2cache[i], | |
428870ff BB |
3121 | ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc) |
3122 | == 0); | |
34dc7c2f BB |
3123 | vdev_get_stats(vd, vs); |
3124 | } | |
3125 | } | |
34dc7c2f BB |
3126 | } |
3127 | ||
9ae529ec CS |
3128 | static void |
3129 | spa_add_feature_stats(spa_t *spa, nvlist_t *config) | |
3130 | { | |
3131 | nvlist_t *features; | |
3132 | zap_cursor_t zc; | |
3133 | zap_attribute_t za; | |
3134 | ||
3135 | ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER)); | |
3136 | VERIFY(nvlist_alloc(&features, NV_UNIQUE_NAME, KM_SLEEP) == 0); | |
3137 | ||
3138 | if (spa->spa_feat_for_read_obj != 0) { | |
3139 | for (zap_cursor_init(&zc, spa->spa_meta_objset, | |
3140 | spa->spa_feat_for_read_obj); | |
3141 | zap_cursor_retrieve(&zc, &za) == 0; | |
3142 | zap_cursor_advance(&zc)) { | |
3143 | ASSERT(za.za_integer_length == sizeof (uint64_t) && | |
3144 | za.za_num_integers == 1); | |
3145 | VERIFY3U(0, ==, nvlist_add_uint64(features, za.za_name, | |
3146 | za.za_first_integer)); | |
3147 | } | |
3148 | zap_cursor_fini(&zc); | |
3149 | } | |
3150 | ||
3151 | if (spa->spa_feat_for_write_obj != 0) { | |
3152 | for (zap_cursor_init(&zc, spa->spa_meta_objset, | |
3153 | spa->spa_feat_for_write_obj); | |
3154 | zap_cursor_retrieve(&zc, &za) == 0; | |
3155 | zap_cursor_advance(&zc)) { | |
3156 | ASSERT(za.za_integer_length == sizeof (uint64_t) && | |
3157 | za.za_num_integers == 1); | |
3158 | VERIFY3U(0, ==, nvlist_add_uint64(features, za.za_name, | |
3159 | za.za_first_integer)); | |
3160 | } | |
3161 | zap_cursor_fini(&zc); | |
3162 | } | |
3163 | ||
3164 | VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURE_STATS, | |
3165 | features) == 0); | |
3166 | nvlist_free(features); | |
3167 | } | |
3168 | ||
34dc7c2f | 3169 | int |
9ae529ec CS |
3170 | spa_get_stats(const char *name, nvlist_t **config, |
3171 | char *altroot, size_t buflen) | |
34dc7c2f BB |
3172 | { |
3173 | int error; | |
3174 | spa_t *spa; | |
3175 | ||
3176 | *config = NULL; | |
428870ff | 3177 | error = spa_open_common(name, &spa, FTAG, NULL, config); |
34dc7c2f | 3178 | |
9babb374 BB |
3179 | if (spa != NULL) { |
3180 | /* | |
3181 | * This still leaves a window of inconsistency where the spares | |
3182 | * or l2cache devices could change and the config would be | |
3183 | * self-inconsistent. | |
3184 | */ | |
3185 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); | |
34dc7c2f | 3186 | |
9babb374 | 3187 | if (*config != NULL) { |
572e2857 BB |
3188 | uint64_t loadtimes[2]; |
3189 | ||
3190 | loadtimes[0] = spa->spa_loaded_ts.tv_sec; | |
3191 | loadtimes[1] = spa->spa_loaded_ts.tv_nsec; | |
3192 | VERIFY(nvlist_add_uint64_array(*config, | |
3193 | ZPOOL_CONFIG_LOADED_TIME, loadtimes, 2) == 0); | |
3194 | ||
b128c09f | 3195 | VERIFY(nvlist_add_uint64(*config, |
9babb374 BB |
3196 | ZPOOL_CONFIG_ERRCOUNT, |
3197 | spa_get_errlog_size(spa)) == 0); | |
3198 | ||
3199 | if (spa_suspended(spa)) | |
3200 | VERIFY(nvlist_add_uint64(*config, | |
3201 | ZPOOL_CONFIG_SUSPENDED, | |
3202 | spa->spa_failmode) == 0); | |
b128c09f | 3203 | |
9babb374 BB |
3204 | spa_add_spares(spa, *config); |
3205 | spa_add_l2cache(spa, *config); | |
9ae529ec | 3206 | spa_add_feature_stats(spa, *config); |
9babb374 | 3207 | } |
34dc7c2f BB |
3208 | } |
3209 | ||
3210 | /* | |
3211 | * We want to get the alternate root even for faulted pools, so we cheat | |
3212 | * and call spa_lookup() directly. | |
3213 | */ | |
3214 | if (altroot) { | |
3215 | if (spa == NULL) { | |
3216 | mutex_enter(&spa_namespace_lock); | |
3217 | spa = spa_lookup(name); | |
3218 | if (spa) | |
3219 | spa_altroot(spa, altroot, buflen); | |
3220 | else | |
3221 | altroot[0] = '\0'; | |
3222 | spa = NULL; | |
3223 | mutex_exit(&spa_namespace_lock); | |
3224 | } else { | |
3225 | spa_altroot(spa, altroot, buflen); | |
3226 | } | |
3227 | } | |
3228 | ||
9babb374 BB |
3229 | if (spa != NULL) { |
3230 | spa_config_exit(spa, SCL_CONFIG, FTAG); | |
34dc7c2f | 3231 | spa_close(spa, FTAG); |
9babb374 | 3232 | } |
34dc7c2f BB |
3233 | |
3234 | return (error); | |
3235 | } | |
3236 | ||
3237 | /* | |
3238 | * Validate that the auxiliary device array is well formed. We must have an | |
3239 | * array of nvlists, each which describes a valid leaf vdev. If this is an | |
3240 | * import (mode is VDEV_ALLOC_SPARE), then we allow corrupted spares to be | |
3241 | * specified, as long as they are well-formed. | |
3242 | */ | |
3243 | static int | |
3244 | spa_validate_aux_devs(spa_t *spa, nvlist_t *nvroot, uint64_t crtxg, int mode, | |
3245 | spa_aux_vdev_t *sav, const char *config, uint64_t version, | |
3246 | vdev_labeltype_t label) | |
3247 | { | |
3248 | nvlist_t **dev; | |
3249 | uint_t i, ndev; | |
3250 | vdev_t *vd; | |
3251 | int error; | |
3252 | ||
b128c09f BB |
3253 | ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); |
3254 | ||
34dc7c2f BB |
3255 | /* |
3256 | * It's acceptable to have no devs specified. | |
3257 | */ | |
3258 | if (nvlist_lookup_nvlist_array(nvroot, config, &dev, &ndev) != 0) | |
3259 | return (0); | |
3260 | ||
3261 | if (ndev == 0) | |
2e528b49 | 3262 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3263 | |
3264 | /* | |
3265 | * Make sure the pool is formatted with a version that supports this | |
3266 | * device type. | |
3267 | */ | |
3268 | if (spa_version(spa) < version) | |
2e528b49 | 3269 | return (SET_ERROR(ENOTSUP)); |
34dc7c2f BB |
3270 | |
3271 | /* | |
3272 | * Set the pending device list so we correctly handle device in-use | |
3273 | * checking. | |
3274 | */ | |
3275 | sav->sav_pending = dev; | |
3276 | sav->sav_npending = ndev; | |
3277 | ||
3278 | for (i = 0; i < ndev; i++) { | |
3279 | if ((error = spa_config_parse(spa, &vd, dev[i], NULL, 0, | |
3280 | mode)) != 0) | |
3281 | goto out; | |
3282 | ||
3283 | if (!vd->vdev_ops->vdev_op_leaf) { | |
3284 | vdev_free(vd); | |
2e528b49 | 3285 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
3286 | goto out; |
3287 | } | |
3288 | ||
3289 | /* | |
b128c09f BB |
3290 | * The L2ARC currently only supports disk devices in |
3291 | * kernel context. For user-level testing, we allow it. | |
34dc7c2f | 3292 | */ |
b128c09f | 3293 | #ifdef _KERNEL |
34dc7c2f BB |
3294 | if ((strcmp(config, ZPOOL_CONFIG_L2CACHE) == 0) && |
3295 | strcmp(vd->vdev_ops->vdev_op_type, VDEV_TYPE_DISK) != 0) { | |
2e528b49 | 3296 | error = SET_ERROR(ENOTBLK); |
5ffb9d1d | 3297 | vdev_free(vd); |
34dc7c2f BB |
3298 | goto out; |
3299 | } | |
b128c09f | 3300 | #endif |
34dc7c2f BB |
3301 | vd->vdev_top = vd; |
3302 | ||
3303 | if ((error = vdev_open(vd)) == 0 && | |
3304 | (error = vdev_label_init(vd, crtxg, label)) == 0) { | |
3305 | VERIFY(nvlist_add_uint64(dev[i], ZPOOL_CONFIG_GUID, | |
3306 | vd->vdev_guid) == 0); | |
3307 | } | |
3308 | ||
3309 | vdev_free(vd); | |
3310 | ||
3311 | if (error && | |
3312 | (mode != VDEV_ALLOC_SPARE && mode != VDEV_ALLOC_L2CACHE)) | |
3313 | goto out; | |
3314 | else | |
3315 | error = 0; | |
3316 | } | |
3317 | ||
3318 | out: | |
3319 | sav->sav_pending = NULL; | |
3320 | sav->sav_npending = 0; | |
3321 | return (error); | |
3322 | } | |
3323 | ||
3324 | static int | |
3325 | spa_validate_aux(spa_t *spa, nvlist_t *nvroot, uint64_t crtxg, int mode) | |
3326 | { | |
3327 | int error; | |
3328 | ||
b128c09f BB |
3329 | ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); |
3330 | ||
34dc7c2f BB |
3331 | if ((error = spa_validate_aux_devs(spa, nvroot, crtxg, mode, |
3332 | &spa->spa_spares, ZPOOL_CONFIG_SPARES, SPA_VERSION_SPARES, | |
3333 | VDEV_LABEL_SPARE)) != 0) { | |
3334 | return (error); | |
3335 | } | |
3336 | ||
3337 | return (spa_validate_aux_devs(spa, nvroot, crtxg, mode, | |
3338 | &spa->spa_l2cache, ZPOOL_CONFIG_L2CACHE, SPA_VERSION_L2CACHE, | |
3339 | VDEV_LABEL_L2CACHE)); | |
3340 | } | |
3341 | ||
3342 | static void | |
3343 | spa_set_aux_vdevs(spa_aux_vdev_t *sav, nvlist_t **devs, int ndevs, | |
3344 | const char *config) | |
3345 | { | |
3346 | int i; | |
3347 | ||
3348 | if (sav->sav_config != NULL) { | |
3349 | nvlist_t **olddevs; | |
3350 | uint_t oldndevs; | |
3351 | nvlist_t **newdevs; | |
3352 | ||
3353 | /* | |
3354 | * Generate new dev list by concatentating with the | |
3355 | * current dev list. | |
3356 | */ | |
3357 | VERIFY(nvlist_lookup_nvlist_array(sav->sav_config, config, | |
3358 | &olddevs, &oldndevs) == 0); | |
3359 | ||
3360 | newdevs = kmem_alloc(sizeof (void *) * | |
b8d06fca | 3361 | (ndevs + oldndevs), KM_PUSHPAGE); |
34dc7c2f BB |
3362 | for (i = 0; i < oldndevs; i++) |
3363 | VERIFY(nvlist_dup(olddevs[i], &newdevs[i], | |
b8d06fca | 3364 | KM_PUSHPAGE) == 0); |
34dc7c2f BB |
3365 | for (i = 0; i < ndevs; i++) |
3366 | VERIFY(nvlist_dup(devs[i], &newdevs[i + oldndevs], | |
b8d06fca | 3367 | KM_PUSHPAGE) == 0); |
34dc7c2f BB |
3368 | |
3369 | VERIFY(nvlist_remove(sav->sav_config, config, | |
3370 | DATA_TYPE_NVLIST_ARRAY) == 0); | |
3371 | ||
3372 | VERIFY(nvlist_add_nvlist_array(sav->sav_config, | |
3373 | config, newdevs, ndevs + oldndevs) == 0); | |
3374 | for (i = 0; i < oldndevs + ndevs; i++) | |
3375 | nvlist_free(newdevs[i]); | |
3376 | kmem_free(newdevs, (oldndevs + ndevs) * sizeof (void *)); | |
3377 | } else { | |
3378 | /* | |
3379 | * Generate a new dev list. | |
3380 | */ | |
3381 | VERIFY(nvlist_alloc(&sav->sav_config, NV_UNIQUE_NAME, | |
b8d06fca | 3382 | KM_PUSHPAGE) == 0); |
34dc7c2f BB |
3383 | VERIFY(nvlist_add_nvlist_array(sav->sav_config, config, |
3384 | devs, ndevs) == 0); | |
3385 | } | |
3386 | } | |
3387 | ||
3388 | /* | |
3389 | * Stop and drop level 2 ARC devices | |
3390 | */ | |
3391 | void | |
3392 | spa_l2cache_drop(spa_t *spa) | |
3393 | { | |
3394 | vdev_t *vd; | |
3395 | int i; | |
3396 | spa_aux_vdev_t *sav = &spa->spa_l2cache; | |
3397 | ||
3398 | for (i = 0; i < sav->sav_count; i++) { | |
3399 | uint64_t pool; | |
3400 | ||
3401 | vd = sav->sav_vdevs[i]; | |
3402 | ASSERT(vd != NULL); | |
3403 | ||
fb5f0bc8 BB |
3404 | if (spa_l2cache_exists(vd->vdev_guid, &pool) && |
3405 | pool != 0ULL && l2arc_vdev_present(vd)) | |
34dc7c2f | 3406 | l2arc_remove_vdev(vd); |
34dc7c2f BB |
3407 | } |
3408 | } | |
3409 | ||
3410 | /* | |
3411 | * Pool Creation | |
3412 | */ | |
3413 | int | |
3414 | spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props, | |
6f1ffb06 | 3415 | nvlist_t *zplprops) |
34dc7c2f BB |
3416 | { |
3417 | spa_t *spa; | |
3418 | char *altroot = NULL; | |
3419 | vdev_t *rvd; | |
3420 | dsl_pool_t *dp; | |
3421 | dmu_tx_t *tx; | |
9babb374 | 3422 | int error = 0; |
34dc7c2f BB |
3423 | uint64_t txg = TXG_INITIAL; |
3424 | nvlist_t **spares, **l2cache; | |
3425 | uint_t nspares, nl2cache; | |
428870ff | 3426 | uint64_t version, obj; |
9ae529ec CS |
3427 | boolean_t has_features; |
3428 | nvpair_t *elem; | |
d6320ddb | 3429 | int c; |
34dc7c2f BB |
3430 | |
3431 | /* | |
3432 | * If this pool already exists, return failure. | |
3433 | */ | |
3434 | mutex_enter(&spa_namespace_lock); | |
3435 | if (spa_lookup(pool) != NULL) { | |
3436 | mutex_exit(&spa_namespace_lock); | |
2e528b49 | 3437 | return (SET_ERROR(EEXIST)); |
34dc7c2f BB |
3438 | } |
3439 | ||
3440 | /* | |
3441 | * Allocate a new spa_t structure. | |
3442 | */ | |
3443 | (void) nvlist_lookup_string(props, | |
3444 | zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot); | |
428870ff | 3445 | spa = spa_add(pool, NULL, altroot); |
fb5f0bc8 | 3446 | spa_activate(spa, spa_mode_global); |
34dc7c2f | 3447 | |
34dc7c2f | 3448 | if (props && (error = spa_prop_validate(spa, props))) { |
34dc7c2f BB |
3449 | spa_deactivate(spa); |
3450 | spa_remove(spa); | |
b128c09f | 3451 | mutex_exit(&spa_namespace_lock); |
34dc7c2f BB |
3452 | return (error); |
3453 | } | |
3454 | ||
9ae529ec CS |
3455 | has_features = B_FALSE; |
3456 | for (elem = nvlist_next_nvpair(props, NULL); | |
3457 | elem != NULL; elem = nvlist_next_nvpair(props, elem)) { | |
3458 | if (zpool_prop_feature(nvpair_name(elem))) | |
3459 | has_features = B_TRUE; | |
3460 | } | |
3461 | ||
3462 | if (has_features || nvlist_lookup_uint64(props, | |
3463 | zpool_prop_to_name(ZPOOL_PROP_VERSION), &version) != 0) { | |
34dc7c2f | 3464 | version = SPA_VERSION; |
9ae529ec CS |
3465 | } |
3466 | ASSERT(SPA_VERSION_IS_SUPPORTED(version)); | |
428870ff BB |
3467 | |
3468 | spa->spa_first_txg = txg; | |
3469 | spa->spa_uberblock.ub_txg = txg - 1; | |
34dc7c2f BB |
3470 | spa->spa_uberblock.ub_version = version; |
3471 | spa->spa_ubsync = spa->spa_uberblock; | |
3472 | ||
9babb374 BB |
3473 | /* |
3474 | * Create "The Godfather" zio to hold all async IOs | |
3475 | */ | |
3476 | spa->spa_async_zio_root = zio_root(spa, NULL, NULL, | |
3477 | ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_GODFATHER); | |
3478 | ||
34dc7c2f BB |
3479 | /* |
3480 | * Create the root vdev. | |
3481 | */ | |
b128c09f | 3482 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f BB |
3483 | |
3484 | error = spa_config_parse(spa, &rvd, nvroot, NULL, 0, VDEV_ALLOC_ADD); | |
3485 | ||
3486 | ASSERT(error != 0 || rvd != NULL); | |
3487 | ASSERT(error != 0 || spa->spa_root_vdev == rvd); | |
3488 | ||
3489 | if (error == 0 && !zfs_allocatable_devs(nvroot)) | |
2e528b49 | 3490 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
3491 | |
3492 | if (error == 0 && | |
3493 | (error = vdev_create(rvd, txg, B_FALSE)) == 0 && | |
3494 | (error = spa_validate_aux(spa, nvroot, txg, | |
3495 | VDEV_ALLOC_ADD)) == 0) { | |
d6320ddb | 3496 | for (c = 0; c < rvd->vdev_children; c++) { |
9babb374 BB |
3497 | vdev_metaslab_set_size(rvd->vdev_child[c]); |
3498 | vdev_expand(rvd->vdev_child[c], txg); | |
3499 | } | |
34dc7c2f BB |
3500 | } |
3501 | ||
b128c09f | 3502 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
3503 | |
3504 | if (error != 0) { | |
3505 | spa_unload(spa); | |
3506 | spa_deactivate(spa); | |
3507 | spa_remove(spa); | |
3508 | mutex_exit(&spa_namespace_lock); | |
3509 | return (error); | |
3510 | } | |
3511 | ||
3512 | /* | |
3513 | * Get the list of spares, if specified. | |
3514 | */ | |
3515 | if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, | |
3516 | &spares, &nspares) == 0) { | |
3517 | VERIFY(nvlist_alloc(&spa->spa_spares.sav_config, NV_UNIQUE_NAME, | |
b8d06fca | 3518 | KM_PUSHPAGE) == 0); |
34dc7c2f BB |
3519 | VERIFY(nvlist_add_nvlist_array(spa->spa_spares.sav_config, |
3520 | ZPOOL_CONFIG_SPARES, spares, nspares) == 0); | |
b128c09f | 3521 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f | 3522 | spa_load_spares(spa); |
b128c09f | 3523 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
3524 | spa->spa_spares.sav_sync = B_TRUE; |
3525 | } | |
3526 | ||
3527 | /* | |
3528 | * Get the list of level 2 cache devices, if specified. | |
3529 | */ | |
3530 | if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, | |
3531 | &l2cache, &nl2cache) == 0) { | |
3532 | VERIFY(nvlist_alloc(&spa->spa_l2cache.sav_config, | |
b8d06fca | 3533 | NV_UNIQUE_NAME, KM_PUSHPAGE) == 0); |
34dc7c2f BB |
3534 | VERIFY(nvlist_add_nvlist_array(spa->spa_l2cache.sav_config, |
3535 | ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache) == 0); | |
b128c09f | 3536 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f | 3537 | spa_load_l2cache(spa); |
b128c09f | 3538 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
3539 | spa->spa_l2cache.sav_sync = B_TRUE; |
3540 | } | |
3541 | ||
9ae529ec | 3542 | spa->spa_is_initializing = B_TRUE; |
b128c09f | 3543 | spa->spa_dsl_pool = dp = dsl_pool_create(spa, zplprops, txg); |
34dc7c2f | 3544 | spa->spa_meta_objset = dp->dp_meta_objset; |
9ae529ec | 3545 | spa->spa_is_initializing = B_FALSE; |
34dc7c2f | 3546 | |
428870ff BB |
3547 | /* |
3548 | * Create DDTs (dedup tables). | |
3549 | */ | |
3550 | ddt_create(spa); | |
3551 | ||
3552 | spa_update_dspace(spa); | |
3553 | ||
34dc7c2f BB |
3554 | tx = dmu_tx_create_assigned(dp, txg); |
3555 | ||
3556 | /* | |
3557 | * Create the pool config object. | |
3558 | */ | |
3559 | spa->spa_config_object = dmu_object_alloc(spa->spa_meta_objset, | |
b128c09f | 3560 | DMU_OT_PACKED_NVLIST, SPA_CONFIG_BLOCKSIZE, |
34dc7c2f BB |
3561 | DMU_OT_PACKED_NVLIST_SIZE, sizeof (uint64_t), tx); |
3562 | ||
3563 | if (zap_add(spa->spa_meta_objset, | |
3564 | DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CONFIG, | |
3565 | sizeof (uint64_t), 1, &spa->spa_config_object, tx) != 0) { | |
3566 | cmn_err(CE_PANIC, "failed to add pool config"); | |
3567 | } | |
3568 | ||
9ae529ec CS |
3569 | if (spa_version(spa) >= SPA_VERSION_FEATURES) |
3570 | spa_feature_create_zap_objects(spa, tx); | |
3571 | ||
428870ff BB |
3572 | if (zap_add(spa->spa_meta_objset, |
3573 | DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CREATION_VERSION, | |
3574 | sizeof (uint64_t), 1, &version, tx) != 0) { | |
3575 | cmn_err(CE_PANIC, "failed to add pool version"); | |
3576 | } | |
3577 | ||
34dc7c2f BB |
3578 | /* Newly created pools with the right version are always deflated. */ |
3579 | if (version >= SPA_VERSION_RAIDZ_DEFLATE) { | |
3580 | spa->spa_deflate = TRUE; | |
3581 | if (zap_add(spa->spa_meta_objset, | |
3582 | DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE, | |
3583 | sizeof (uint64_t), 1, &spa->spa_deflate, tx) != 0) { | |
3584 | cmn_err(CE_PANIC, "failed to add deflate"); | |
3585 | } | |
3586 | } | |
3587 | ||
3588 | /* | |
428870ff | 3589 | * Create the deferred-free bpobj. Turn off compression |
34dc7c2f BB |
3590 | * because sync-to-convergence takes longer if the blocksize |
3591 | * keeps changing. | |
3592 | */ | |
428870ff BB |
3593 | obj = bpobj_alloc(spa->spa_meta_objset, 1 << 14, tx); |
3594 | dmu_object_set_compress(spa->spa_meta_objset, obj, | |
34dc7c2f | 3595 | ZIO_COMPRESS_OFF, tx); |
34dc7c2f | 3596 | if (zap_add(spa->spa_meta_objset, |
428870ff BB |
3597 | DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SYNC_BPOBJ, |
3598 | sizeof (uint64_t), 1, &obj, tx) != 0) { | |
3599 | cmn_err(CE_PANIC, "failed to add bpobj"); | |
34dc7c2f | 3600 | } |
428870ff BB |
3601 | VERIFY3U(0, ==, bpobj_open(&spa->spa_deferred_bpobj, |
3602 | spa->spa_meta_objset, obj)); | |
34dc7c2f BB |
3603 | |
3604 | /* | |
3605 | * Create the pool's history object. | |
3606 | */ | |
3607 | if (version >= SPA_VERSION_ZPOOL_HISTORY) | |
3608 | spa_history_create_obj(spa, tx); | |
3609 | ||
3610 | /* | |
3611 | * Set pool properties. | |
3612 | */ | |
3613 | spa->spa_bootfs = zpool_prop_default_numeric(ZPOOL_PROP_BOOTFS); | |
3614 | spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION); | |
3615 | spa->spa_failmode = zpool_prop_default_numeric(ZPOOL_PROP_FAILUREMODE); | |
9babb374 | 3616 | spa->spa_autoexpand = zpool_prop_default_numeric(ZPOOL_PROP_AUTOEXPAND); |
428870ff | 3617 | |
d164b209 BB |
3618 | if (props != NULL) { |
3619 | spa_configfile_set(spa, props, B_FALSE); | |
13fe0198 | 3620 | spa_sync_props(props, tx); |
d164b209 | 3621 | } |
34dc7c2f BB |
3622 | |
3623 | dmu_tx_commit(tx); | |
3624 | ||
3625 | spa->spa_sync_on = B_TRUE; | |
3626 | txg_sync_start(spa->spa_dsl_pool); | |
3627 | ||
3628 | /* | |
3629 | * We explicitly wait for the first transaction to complete so that our | |
3630 | * bean counters are appropriately updated. | |
3631 | */ | |
3632 | txg_wait_synced(spa->spa_dsl_pool, txg); | |
3633 | ||
b128c09f | 3634 | spa_config_sync(spa, B_FALSE, B_TRUE); |
34dc7c2f | 3635 | |
6f1ffb06 | 3636 | spa_history_log_version(spa, "create"); |
34dc7c2f | 3637 | |
b128c09f BB |
3638 | spa->spa_minref = refcount_count(&spa->spa_refcount); |
3639 | ||
d164b209 BB |
3640 | mutex_exit(&spa_namespace_lock); |
3641 | ||
34dc7c2f BB |
3642 | return (0); |
3643 | } | |
3644 | ||
9babb374 | 3645 | #ifdef _KERNEL |
34dc7c2f | 3646 | /* |
9babb374 BB |
3647 | * Get the root pool information from the root disk, then import the root pool |
3648 | * during the system boot up time. | |
34dc7c2f | 3649 | */ |
9babb374 BB |
3650 | extern int vdev_disk_read_rootlabel(char *, char *, nvlist_t **); |
3651 | ||
3652 | static nvlist_t * | |
3653 | spa_generate_rootconf(char *devpath, char *devid, uint64_t *guid) | |
3654 | { | |
3655 | nvlist_t *config; | |
3656 | nvlist_t *nvtop, *nvroot; | |
3657 | uint64_t pgid; | |
3658 | ||
3659 | if (vdev_disk_read_rootlabel(devpath, devid, &config) != 0) | |
3660 | return (NULL); | |
3661 | ||
3662 | /* | |
3663 | * Add this top-level vdev to the child array. | |
3664 | */ | |
3665 | VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, | |
3666 | &nvtop) == 0); | |
3667 | VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, | |
3668 | &pgid) == 0); | |
3669 | VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, guid) == 0); | |
3670 | ||
3671 | /* | |
3672 | * Put this pool's top-level vdevs into a root vdev. | |
3673 | */ | |
b8d06fca | 3674 | VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_PUSHPAGE) == 0); |
9babb374 BB |
3675 | VERIFY(nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE, |
3676 | VDEV_TYPE_ROOT) == 0); | |
3677 | VERIFY(nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) == 0); | |
3678 | VERIFY(nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, pgid) == 0); | |
3679 | VERIFY(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, | |
3680 | &nvtop, 1) == 0); | |
3681 | ||
3682 | /* | |
3683 | * Replace the existing vdev_tree with the new root vdev in | |
3684 | * this pool's configuration (remove the old, add the new). | |
3685 | */ | |
3686 | VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0); | |
3687 | nvlist_free(nvroot); | |
3688 | return (config); | |
3689 | } | |
3690 | ||
3691 | /* | |
3692 | * Walk the vdev tree and see if we can find a device with "better" | |
3693 | * configuration. A configuration is "better" if the label on that | |
3694 | * device has a more recent txg. | |
3695 | */ | |
3696 | static void | |
3697 | spa_alt_rootvdev(vdev_t *vd, vdev_t **avd, uint64_t *txg) | |
3698 | { | |
d6320ddb BB |
3699 | int c; |
3700 | ||
3701 | for (c = 0; c < vd->vdev_children; c++) | |
9babb374 BB |
3702 | spa_alt_rootvdev(vd->vdev_child[c], avd, txg); |
3703 | ||
3704 | if (vd->vdev_ops->vdev_op_leaf) { | |
3705 | nvlist_t *label; | |
3706 | uint64_t label_txg; | |
3707 | ||
3708 | if (vdev_disk_read_rootlabel(vd->vdev_physpath, vd->vdev_devid, | |
3709 | &label) != 0) | |
3710 | return; | |
3711 | ||
3712 | VERIFY(nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_TXG, | |
3713 | &label_txg) == 0); | |
3714 | ||
3715 | /* | |
3716 | * Do we have a better boot device? | |
3717 | */ | |
3718 | if (label_txg > *txg) { | |
3719 | *txg = label_txg; | |
3720 | *avd = vd; | |
3721 | } | |
3722 | nvlist_free(label); | |
3723 | } | |
3724 | } | |
3725 | ||
3726 | /* | |
3727 | * Import a root pool. | |
3728 | * | |
3729 | * For x86. devpath_list will consist of devid and/or physpath name of | |
3730 | * the vdev (e.g. "id1,sd@SSEAGATE..." or "/pci@1f,0/ide@d/disk@0,0:a"). | |
3731 | * The GRUB "findroot" command will return the vdev we should boot. | |
3732 | * | |
3733 | * For Sparc, devpath_list consists the physpath name of the booting device | |
3734 | * no matter the rootpool is a single device pool or a mirrored pool. | |
3735 | * e.g. | |
3736 | * "/pci@1f,0/ide@d/disk@0,0:a" | |
3737 | */ | |
3738 | int | |
3739 | spa_import_rootpool(char *devpath, char *devid) | |
3740 | { | |
3741 | spa_t *spa; | |
3742 | vdev_t *rvd, *bvd, *avd = NULL; | |
3743 | nvlist_t *config, *nvtop; | |
3744 | uint64_t guid, txg; | |
3745 | char *pname; | |
3746 | int error; | |
3747 | ||
3748 | /* | |
3749 | * Read the label from the boot device and generate a configuration. | |
3750 | */ | |
428870ff BB |
3751 | config = spa_generate_rootconf(devpath, devid, &guid); |
3752 | #if defined(_OBP) && defined(_KERNEL) | |
3753 | if (config == NULL) { | |
3754 | if (strstr(devpath, "/iscsi/ssd") != NULL) { | |
3755 | /* iscsi boot */ | |
3756 | get_iscsi_bootpath_phy(devpath); | |
3757 | config = spa_generate_rootconf(devpath, devid, &guid); | |
3758 | } | |
3759 | } | |
3760 | #endif | |
3761 | if (config == NULL) { | |
9ae529ec | 3762 | cmn_err(CE_NOTE, "Cannot read the pool label from '%s'", |
9babb374 | 3763 | devpath); |
2e528b49 | 3764 | return (SET_ERROR(EIO)); |
9babb374 BB |
3765 | } |
3766 | ||
3767 | VERIFY(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, | |
3768 | &pname) == 0); | |
3769 | VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, &txg) == 0); | |
3770 | ||
3771 | mutex_enter(&spa_namespace_lock); | |
3772 | if ((spa = spa_lookup(pname)) != NULL) { | |
3773 | /* | |
3774 | * Remove the existing root pool from the namespace so that we | |
3775 | * can replace it with the correct config we just read in. | |
3776 | */ | |
3777 | spa_remove(spa); | |
3778 | } | |
3779 | ||
428870ff | 3780 | spa = spa_add(pname, config, NULL); |
9babb374 | 3781 | spa->spa_is_root = B_TRUE; |
572e2857 | 3782 | spa->spa_import_flags = ZFS_IMPORT_VERBATIM; |
9babb374 BB |
3783 | |
3784 | /* | |
3785 | * Build up a vdev tree based on the boot device's label config. | |
3786 | */ | |
3787 | VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, | |
3788 | &nvtop) == 0); | |
3789 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); | |
3790 | error = spa_config_parse(spa, &rvd, nvtop, NULL, 0, | |
3791 | VDEV_ALLOC_ROOTPOOL); | |
3792 | spa_config_exit(spa, SCL_ALL, FTAG); | |
3793 | if (error) { | |
3794 | mutex_exit(&spa_namespace_lock); | |
3795 | nvlist_free(config); | |
3796 | cmn_err(CE_NOTE, "Can not parse the config for pool '%s'", | |
3797 | pname); | |
3798 | return (error); | |
3799 | } | |
3800 | ||
3801 | /* | |
3802 | * Get the boot vdev. | |
3803 | */ | |
3804 | if ((bvd = vdev_lookup_by_guid(rvd, guid)) == NULL) { | |
3805 | cmn_err(CE_NOTE, "Can not find the boot vdev for guid %llu", | |
3806 | (u_longlong_t)guid); | |
2e528b49 | 3807 | error = SET_ERROR(ENOENT); |
9babb374 BB |
3808 | goto out; |
3809 | } | |
3810 | ||
3811 | /* | |
3812 | * Determine if there is a better boot device. | |
3813 | */ | |
3814 | avd = bvd; | |
3815 | spa_alt_rootvdev(rvd, &avd, &txg); | |
3816 | if (avd != bvd) { | |
3817 | cmn_err(CE_NOTE, "The boot device is 'degraded'. Please " | |
3818 | "try booting from '%s'", avd->vdev_path); | |
2e528b49 | 3819 | error = SET_ERROR(EINVAL); |
9babb374 BB |
3820 | goto out; |
3821 | } | |
3822 | ||
3823 | /* | |
3824 | * If the boot device is part of a spare vdev then ensure that | |
3825 | * we're booting off the active spare. | |
3826 | */ | |
3827 | if (bvd->vdev_parent->vdev_ops == &vdev_spare_ops && | |
3828 | !bvd->vdev_isspare) { | |
3829 | cmn_err(CE_NOTE, "The boot device is currently spared. Please " | |
3830 | "try booting from '%s'", | |
572e2857 BB |
3831 | bvd->vdev_parent-> |
3832 | vdev_child[bvd->vdev_parent->vdev_children - 1]->vdev_path); | |
2e528b49 | 3833 | error = SET_ERROR(EINVAL); |
9babb374 BB |
3834 | goto out; |
3835 | } | |
3836 | ||
9babb374 BB |
3837 | error = 0; |
3838 | out: | |
3839 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); | |
3840 | vdev_free(rvd); | |
3841 | spa_config_exit(spa, SCL_ALL, FTAG); | |
3842 | mutex_exit(&spa_namespace_lock); | |
3843 | ||
3844 | nvlist_free(config); | |
3845 | return (error); | |
3846 | } | |
3847 | ||
3848 | #endif | |
3849 | ||
9babb374 BB |
3850 | /* |
3851 | * Import a non-root pool into the system. | |
3852 | */ | |
3853 | int | |
13fe0198 | 3854 | spa_import(char *pool, nvlist_t *config, nvlist_t *props, uint64_t flags) |
34dc7c2f BB |
3855 | { |
3856 | spa_t *spa; | |
3857 | char *altroot = NULL; | |
428870ff BB |
3858 | spa_load_state_t state = SPA_LOAD_IMPORT; |
3859 | zpool_rewind_policy_t policy; | |
572e2857 BB |
3860 | uint64_t mode = spa_mode_global; |
3861 | uint64_t readonly = B_FALSE; | |
9babb374 | 3862 | int error; |
34dc7c2f BB |
3863 | nvlist_t *nvroot; |
3864 | nvlist_t **spares, **l2cache; | |
3865 | uint_t nspares, nl2cache; | |
34dc7c2f BB |
3866 | |
3867 | /* | |
3868 | * If a pool with this name exists, return failure. | |
3869 | */ | |
3870 | mutex_enter(&spa_namespace_lock); | |
428870ff | 3871 | if (spa_lookup(pool) != NULL) { |
9babb374 | 3872 | mutex_exit(&spa_namespace_lock); |
2e528b49 | 3873 | return (SET_ERROR(EEXIST)); |
34dc7c2f BB |
3874 | } |
3875 | ||
3876 | /* | |
3877 | * Create and initialize the spa structure. | |
3878 | */ | |
3879 | (void) nvlist_lookup_string(props, | |
3880 | zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot); | |
572e2857 BB |
3881 | (void) nvlist_lookup_uint64(props, |
3882 | zpool_prop_to_name(ZPOOL_PROP_READONLY), &readonly); | |
3883 | if (readonly) | |
3884 | mode = FREAD; | |
428870ff | 3885 | spa = spa_add(pool, config, altroot); |
572e2857 BB |
3886 | spa->spa_import_flags = flags; |
3887 | ||
3888 | /* | |
3889 | * Verbatim import - Take a pool and insert it into the namespace | |
3890 | * as if it had been loaded at boot. | |
3891 | */ | |
3892 | if (spa->spa_import_flags & ZFS_IMPORT_VERBATIM) { | |
3893 | if (props != NULL) | |
3894 | spa_configfile_set(spa, props, B_FALSE); | |
3895 | ||
3896 | spa_config_sync(spa, B_FALSE, B_TRUE); | |
3897 | ||
3898 | mutex_exit(&spa_namespace_lock); | |
6f1ffb06 | 3899 | spa_history_log_version(spa, "import"); |
572e2857 BB |
3900 | |
3901 | return (0); | |
3902 | } | |
3903 | ||
3904 | spa_activate(spa, mode); | |
34dc7c2f | 3905 | |
9babb374 BB |
3906 | /* |
3907 | * Don't start async tasks until we know everything is healthy. | |
3908 | */ | |
3909 | spa_async_suspend(spa); | |
b128c09f | 3910 | |
572e2857 BB |
3911 | zpool_get_rewind_policy(config, &policy); |
3912 | if (policy.zrp_request & ZPOOL_DO_REWIND) | |
3913 | state = SPA_LOAD_RECOVER; | |
3914 | ||
34dc7c2f | 3915 | /* |
9babb374 BB |
3916 | * Pass off the heavy lifting to spa_load(). Pass TRUE for mosconfig |
3917 | * because the user-supplied config is actually the one to trust when | |
b128c09f | 3918 | * doing an import. |
34dc7c2f | 3919 | */ |
428870ff BB |
3920 | if (state != SPA_LOAD_RECOVER) |
3921 | spa->spa_last_ubsync_txg = spa->spa_load_txg = 0; | |
572e2857 | 3922 | |
428870ff BB |
3923 | error = spa_load_best(spa, state, B_TRUE, policy.zrp_txg, |
3924 | policy.zrp_request); | |
3925 | ||
3926 | /* | |
572e2857 BB |
3927 | * Propagate anything learned while loading the pool and pass it |
3928 | * back to caller (i.e. rewind info, missing devices, etc). | |
428870ff | 3929 | */ |
572e2857 BB |
3930 | VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, |
3931 | spa->spa_load_info) == 0); | |
34dc7c2f | 3932 | |
b128c09f | 3933 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f | 3934 | /* |
9babb374 BB |
3935 | * Toss any existing sparelist, as it doesn't have any validity |
3936 | * anymore, and conflicts with spa_has_spare(). | |
34dc7c2f | 3937 | */ |
9babb374 | 3938 | if (spa->spa_spares.sav_config) { |
34dc7c2f BB |
3939 | nvlist_free(spa->spa_spares.sav_config); |
3940 | spa->spa_spares.sav_config = NULL; | |
3941 | spa_load_spares(spa); | |
3942 | } | |
9babb374 | 3943 | if (spa->spa_l2cache.sav_config) { |
34dc7c2f BB |
3944 | nvlist_free(spa->spa_l2cache.sav_config); |
3945 | spa->spa_l2cache.sav_config = NULL; | |
3946 | spa_load_l2cache(spa); | |
3947 | } | |
3948 | ||
3949 | VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, | |
3950 | &nvroot) == 0); | |
3951 | if (error == 0) | |
9babb374 BB |
3952 | error = spa_validate_aux(spa, nvroot, -1ULL, |
3953 | VDEV_ALLOC_SPARE); | |
34dc7c2f BB |
3954 | if (error == 0) |
3955 | error = spa_validate_aux(spa, nvroot, -1ULL, | |
3956 | VDEV_ALLOC_L2CACHE); | |
b128c09f | 3957 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f | 3958 | |
d164b209 BB |
3959 | if (props != NULL) |
3960 | spa_configfile_set(spa, props, B_FALSE); | |
3961 | ||
fb5f0bc8 BB |
3962 | if (error != 0 || (props && spa_writeable(spa) && |
3963 | (error = spa_prop_set(spa, props)))) { | |
9babb374 BB |
3964 | spa_unload(spa); |
3965 | spa_deactivate(spa); | |
3966 | spa_remove(spa); | |
34dc7c2f BB |
3967 | mutex_exit(&spa_namespace_lock); |
3968 | return (error); | |
3969 | } | |
3970 | ||
572e2857 BB |
3971 | spa_async_resume(spa); |
3972 | ||
34dc7c2f BB |
3973 | /* |
3974 | * Override any spares and level 2 cache devices as specified by | |
3975 | * the user, as these may have correct device names/devids, etc. | |
3976 | */ | |
3977 | if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, | |
3978 | &spares, &nspares) == 0) { | |
3979 | if (spa->spa_spares.sav_config) | |
3980 | VERIFY(nvlist_remove(spa->spa_spares.sav_config, | |
3981 | ZPOOL_CONFIG_SPARES, DATA_TYPE_NVLIST_ARRAY) == 0); | |
3982 | else | |
3983 | VERIFY(nvlist_alloc(&spa->spa_spares.sav_config, | |
b8d06fca | 3984 | NV_UNIQUE_NAME, KM_PUSHPAGE) == 0); |
34dc7c2f BB |
3985 | VERIFY(nvlist_add_nvlist_array(spa->spa_spares.sav_config, |
3986 | ZPOOL_CONFIG_SPARES, spares, nspares) == 0); | |
b128c09f | 3987 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f | 3988 | spa_load_spares(spa); |
b128c09f | 3989 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
3990 | spa->spa_spares.sav_sync = B_TRUE; |
3991 | } | |
3992 | if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, | |
3993 | &l2cache, &nl2cache) == 0) { | |
3994 | if (spa->spa_l2cache.sav_config) | |
3995 | VERIFY(nvlist_remove(spa->spa_l2cache.sav_config, | |
3996 | ZPOOL_CONFIG_L2CACHE, DATA_TYPE_NVLIST_ARRAY) == 0); | |
3997 | else | |
3998 | VERIFY(nvlist_alloc(&spa->spa_l2cache.sav_config, | |
b8d06fca | 3999 | NV_UNIQUE_NAME, KM_PUSHPAGE) == 0); |
34dc7c2f BB |
4000 | VERIFY(nvlist_add_nvlist_array(spa->spa_l2cache.sav_config, |
4001 | ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache) == 0); | |
b128c09f | 4002 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f | 4003 | spa_load_l2cache(spa); |
b128c09f | 4004 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
4005 | spa->spa_l2cache.sav_sync = B_TRUE; |
4006 | } | |
4007 | ||
428870ff BB |
4008 | /* |
4009 | * Check for any removed devices. | |
4010 | */ | |
4011 | if (spa->spa_autoreplace) { | |
4012 | spa_aux_check_removed(&spa->spa_spares); | |
4013 | spa_aux_check_removed(&spa->spa_l2cache); | |
4014 | } | |
4015 | ||
fb5f0bc8 | 4016 | if (spa_writeable(spa)) { |
b128c09f BB |
4017 | /* |
4018 | * Update the config cache to include the newly-imported pool. | |
4019 | */ | |
45d1cae3 | 4020 | spa_config_update(spa, SPA_CONFIG_UPDATE_POOL); |
b128c09f | 4021 | } |
34dc7c2f | 4022 | |
34dc7c2f | 4023 | /* |
9babb374 BB |
4024 | * It's possible that the pool was expanded while it was exported. |
4025 | * We kick off an async task to handle this for us. | |
34dc7c2f | 4026 | */ |
9babb374 | 4027 | spa_async_request(spa, SPA_ASYNC_AUTOEXPAND); |
b128c09f | 4028 | |
9babb374 | 4029 | mutex_exit(&spa_namespace_lock); |
6f1ffb06 | 4030 | spa_history_log_version(spa, "import"); |
b128c09f | 4031 | |
526af785 PJD |
4032 | #ifdef _KERNEL |
4033 | zvol_create_minors(pool); | |
4034 | #endif | |
4035 | ||
b128c09f BB |
4036 | return (0); |
4037 | } | |
4038 | ||
34dc7c2f BB |
4039 | nvlist_t * |
4040 | spa_tryimport(nvlist_t *tryconfig) | |
4041 | { | |
4042 | nvlist_t *config = NULL; | |
4043 | char *poolname; | |
4044 | spa_t *spa; | |
4045 | uint64_t state; | |
d164b209 | 4046 | int error; |
34dc7c2f BB |
4047 | |
4048 | if (nvlist_lookup_string(tryconfig, ZPOOL_CONFIG_POOL_NAME, &poolname)) | |
4049 | return (NULL); | |
4050 | ||
4051 | if (nvlist_lookup_uint64(tryconfig, ZPOOL_CONFIG_POOL_STATE, &state)) | |
4052 | return (NULL); | |
4053 | ||
4054 | /* | |
4055 | * Create and initialize the spa structure. | |
4056 | */ | |
4057 | mutex_enter(&spa_namespace_lock); | |
428870ff | 4058 | spa = spa_add(TRYIMPORT_NAME, tryconfig, NULL); |
fb5f0bc8 | 4059 | spa_activate(spa, FREAD); |
34dc7c2f BB |
4060 | |
4061 | /* | |
4062 | * Pass off the heavy lifting to spa_load(). | |
4063 | * Pass TRUE for mosconfig because the user-supplied config | |
4064 | * is actually the one to trust when doing an import. | |
4065 | */ | |
428870ff | 4066 | error = spa_load(spa, SPA_LOAD_TRYIMPORT, SPA_IMPORT_EXISTING, B_TRUE); |
34dc7c2f BB |
4067 | |
4068 | /* | |
4069 | * If 'tryconfig' was at least parsable, return the current config. | |
4070 | */ | |
4071 | if (spa->spa_root_vdev != NULL) { | |
34dc7c2f | 4072 | config = spa_config_generate(spa, NULL, -1ULL, B_TRUE); |
34dc7c2f BB |
4073 | VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, |
4074 | poolname) == 0); | |
4075 | VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, | |
4076 | state) == 0); | |
4077 | VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TIMESTAMP, | |
4078 | spa->spa_uberblock.ub_timestamp) == 0); | |
9ae529ec CS |
4079 | VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, |
4080 | spa->spa_load_info) == 0); | |
34dc7c2f BB |
4081 | |
4082 | /* | |
4083 | * If the bootfs property exists on this pool then we | |
4084 | * copy it out so that external consumers can tell which | |
4085 | * pools are bootable. | |
4086 | */ | |
d164b209 | 4087 | if ((!error || error == EEXIST) && spa->spa_bootfs) { |
b8d06fca | 4088 | char *tmpname = kmem_alloc(MAXPATHLEN, KM_PUSHPAGE); |
34dc7c2f BB |
4089 | |
4090 | /* | |
4091 | * We have to play games with the name since the | |
4092 | * pool was opened as TRYIMPORT_NAME. | |
4093 | */ | |
b128c09f | 4094 | if (dsl_dsobj_to_dsname(spa_name(spa), |
34dc7c2f BB |
4095 | spa->spa_bootfs, tmpname) == 0) { |
4096 | char *cp; | |
b8d06fca | 4097 | char *dsname = kmem_alloc(MAXPATHLEN, KM_PUSHPAGE); |
34dc7c2f BB |
4098 | |
4099 | cp = strchr(tmpname, '/'); | |
4100 | if (cp == NULL) { | |
4101 | (void) strlcpy(dsname, tmpname, | |
4102 | MAXPATHLEN); | |
4103 | } else { | |
4104 | (void) snprintf(dsname, MAXPATHLEN, | |
4105 | "%s/%s", poolname, ++cp); | |
4106 | } | |
4107 | VERIFY(nvlist_add_string(config, | |
4108 | ZPOOL_CONFIG_BOOTFS, dsname) == 0); | |
4109 | kmem_free(dsname, MAXPATHLEN); | |
4110 | } | |
4111 | kmem_free(tmpname, MAXPATHLEN); | |
4112 | } | |
4113 | ||
4114 | /* | |
4115 | * Add the list of hot spares and level 2 cache devices. | |
4116 | */ | |
9babb374 | 4117 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); |
34dc7c2f BB |
4118 | spa_add_spares(spa, config); |
4119 | spa_add_l2cache(spa, config); | |
9babb374 | 4120 | spa_config_exit(spa, SCL_CONFIG, FTAG); |
34dc7c2f BB |
4121 | } |
4122 | ||
4123 | spa_unload(spa); | |
4124 | spa_deactivate(spa); | |
4125 | spa_remove(spa); | |
4126 | mutex_exit(&spa_namespace_lock); | |
4127 | ||
4128 | return (config); | |
4129 | } | |
4130 | ||
4131 | /* | |
4132 | * Pool export/destroy | |
4133 | * | |
4134 | * The act of destroying or exporting a pool is very simple. We make sure there | |
4135 | * is no more pending I/O and any references to the pool are gone. Then, we | |
4136 | * update the pool state and sync all the labels to disk, removing the | |
fb5f0bc8 BB |
4137 | * configuration from the cache afterwards. If the 'hardforce' flag is set, then |
4138 | * we don't sync the labels or remove the configuration cache. | |
34dc7c2f BB |
4139 | */ |
4140 | static int | |
b128c09f | 4141 | spa_export_common(char *pool, int new_state, nvlist_t **oldconfig, |
fb5f0bc8 | 4142 | boolean_t force, boolean_t hardforce) |
34dc7c2f BB |
4143 | { |
4144 | spa_t *spa; | |
4145 | ||
4146 | if (oldconfig) | |
4147 | *oldconfig = NULL; | |
4148 | ||
fb5f0bc8 | 4149 | if (!(spa_mode_global & FWRITE)) |
2e528b49 | 4150 | return (SET_ERROR(EROFS)); |
34dc7c2f BB |
4151 | |
4152 | mutex_enter(&spa_namespace_lock); | |
4153 | if ((spa = spa_lookup(pool)) == NULL) { | |
4154 | mutex_exit(&spa_namespace_lock); | |
2e528b49 | 4155 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
4156 | } |
4157 | ||
4158 | /* | |
4159 | * Put a hold on the pool, drop the namespace lock, stop async tasks, | |
4160 | * reacquire the namespace lock, and see if we can export. | |
4161 | */ | |
4162 | spa_open_ref(spa, FTAG); | |
4163 | mutex_exit(&spa_namespace_lock); | |
4164 | spa_async_suspend(spa); | |
4165 | mutex_enter(&spa_namespace_lock); | |
4166 | spa_close(spa, FTAG); | |
4167 | ||
4168 | /* | |
4169 | * The pool will be in core if it's openable, | |
4170 | * in which case we can modify its state. | |
4171 | */ | |
4172 | if (spa->spa_state != POOL_STATE_UNINITIALIZED && spa->spa_sync_on) { | |
4173 | /* | |
4174 | * Objsets may be open only because they're dirty, so we | |
4175 | * have to force it to sync before checking spa_refcnt. | |
4176 | */ | |
34dc7c2f BB |
4177 | txg_wait_synced(spa->spa_dsl_pool, 0); |
4178 | ||
4179 | /* | |
4180 | * A pool cannot be exported or destroyed if there are active | |
4181 | * references. If we are resetting a pool, allow references by | |
4182 | * fault injection handlers. | |
4183 | */ | |
4184 | if (!spa_refcount_zero(spa) || | |
4185 | (spa->spa_inject_ref != 0 && | |
4186 | new_state != POOL_STATE_UNINITIALIZED)) { | |
34dc7c2f BB |
4187 | spa_async_resume(spa); |
4188 | mutex_exit(&spa_namespace_lock); | |
2e528b49 | 4189 | return (SET_ERROR(EBUSY)); |
34dc7c2f BB |
4190 | } |
4191 | ||
b128c09f BB |
4192 | /* |
4193 | * A pool cannot be exported if it has an active shared spare. | |
4194 | * This is to prevent other pools stealing the active spare | |
4195 | * from an exported pool. At user's own will, such pool can | |
4196 | * be forcedly exported. | |
4197 | */ | |
4198 | if (!force && new_state == POOL_STATE_EXPORTED && | |
4199 | spa_has_active_shared_spare(spa)) { | |
4200 | spa_async_resume(spa); | |
4201 | mutex_exit(&spa_namespace_lock); | |
2e528b49 | 4202 | return (SET_ERROR(EXDEV)); |
b128c09f | 4203 | } |
34dc7c2f BB |
4204 | |
4205 | /* | |
4206 | * We want this to be reflected on every label, | |
4207 | * so mark them all dirty. spa_unload() will do the | |
4208 | * final sync that pushes these changes out. | |
4209 | */ | |
fb5f0bc8 | 4210 | if (new_state != POOL_STATE_UNINITIALIZED && !hardforce) { |
b128c09f | 4211 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f | 4212 | spa->spa_state = new_state; |
428870ff BB |
4213 | spa->spa_final_txg = spa_last_synced_txg(spa) + |
4214 | TXG_DEFER_SIZE + 1; | |
34dc7c2f | 4215 | vdev_config_dirty(spa->spa_root_vdev); |
b128c09f | 4216 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
4217 | } |
4218 | } | |
4219 | ||
26685276 | 4220 | spa_event_notify(spa, NULL, FM_EREPORT_ZFS_POOL_DESTROY); |
34dc7c2f BB |
4221 | |
4222 | if (spa->spa_state != POOL_STATE_UNINITIALIZED) { | |
4223 | spa_unload(spa); | |
4224 | spa_deactivate(spa); | |
4225 | } | |
4226 | ||
4227 | if (oldconfig && spa->spa_config) | |
4228 | VERIFY(nvlist_dup(spa->spa_config, oldconfig, 0) == 0); | |
4229 | ||
4230 | if (new_state != POOL_STATE_UNINITIALIZED) { | |
fb5f0bc8 BB |
4231 | if (!hardforce) |
4232 | spa_config_sync(spa, B_TRUE, B_TRUE); | |
34dc7c2f | 4233 | spa_remove(spa); |
34dc7c2f BB |
4234 | } |
4235 | mutex_exit(&spa_namespace_lock); | |
4236 | ||
4237 | return (0); | |
4238 | } | |
4239 | ||
4240 | /* | |
4241 | * Destroy a storage pool. | |
4242 | */ | |
4243 | int | |
4244 | spa_destroy(char *pool) | |
4245 | { | |
fb5f0bc8 BB |
4246 | return (spa_export_common(pool, POOL_STATE_DESTROYED, NULL, |
4247 | B_FALSE, B_FALSE)); | |
34dc7c2f BB |
4248 | } |
4249 | ||
4250 | /* | |
4251 | * Export a storage pool. | |
4252 | */ | |
4253 | int | |
fb5f0bc8 BB |
4254 | spa_export(char *pool, nvlist_t **oldconfig, boolean_t force, |
4255 | boolean_t hardforce) | |
34dc7c2f | 4256 | { |
fb5f0bc8 BB |
4257 | return (spa_export_common(pool, POOL_STATE_EXPORTED, oldconfig, |
4258 | force, hardforce)); | |
34dc7c2f BB |
4259 | } |
4260 | ||
4261 | /* | |
4262 | * Similar to spa_export(), this unloads the spa_t without actually removing it | |
4263 | * from the namespace in any way. | |
4264 | */ | |
4265 | int | |
4266 | spa_reset(char *pool) | |
4267 | { | |
b128c09f | 4268 | return (spa_export_common(pool, POOL_STATE_UNINITIALIZED, NULL, |
fb5f0bc8 | 4269 | B_FALSE, B_FALSE)); |
34dc7c2f BB |
4270 | } |
4271 | ||
34dc7c2f BB |
4272 | /* |
4273 | * ========================================================================== | |
4274 | * Device manipulation | |
4275 | * ========================================================================== | |
4276 | */ | |
4277 | ||
4278 | /* | |
4279 | * Add a device to a storage pool. | |
4280 | */ | |
4281 | int | |
4282 | spa_vdev_add(spa_t *spa, nvlist_t *nvroot) | |
4283 | { | |
428870ff | 4284 | uint64_t txg, id; |
fb5f0bc8 | 4285 | int error; |
34dc7c2f BB |
4286 | vdev_t *rvd = spa->spa_root_vdev; |
4287 | vdev_t *vd, *tvd; | |
4288 | nvlist_t **spares, **l2cache; | |
4289 | uint_t nspares, nl2cache; | |
d6320ddb | 4290 | int c; |
34dc7c2f | 4291 | |
572e2857 BB |
4292 | ASSERT(spa_writeable(spa)); |
4293 | ||
34dc7c2f BB |
4294 | txg = spa_vdev_enter(spa); |
4295 | ||
4296 | if ((error = spa_config_parse(spa, &vd, nvroot, NULL, 0, | |
4297 | VDEV_ALLOC_ADD)) != 0) | |
4298 | return (spa_vdev_exit(spa, NULL, txg, error)); | |
4299 | ||
b128c09f | 4300 | spa->spa_pending_vdev = vd; /* spa_vdev_exit() will clear this */ |
34dc7c2f BB |
4301 | |
4302 | if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares, | |
4303 | &nspares) != 0) | |
4304 | nspares = 0; | |
4305 | ||
4306 | if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2cache, | |
4307 | &nl2cache) != 0) | |
4308 | nl2cache = 0; | |
4309 | ||
b128c09f | 4310 | if (vd->vdev_children == 0 && nspares == 0 && nl2cache == 0) |
34dc7c2f | 4311 | return (spa_vdev_exit(spa, vd, txg, EINVAL)); |
34dc7c2f | 4312 | |
b128c09f BB |
4313 | if (vd->vdev_children != 0 && |
4314 | (error = vdev_create(vd, txg, B_FALSE)) != 0) | |
4315 | return (spa_vdev_exit(spa, vd, txg, error)); | |
34dc7c2f BB |
4316 | |
4317 | /* | |
4318 | * We must validate the spares and l2cache devices after checking the | |
4319 | * children. Otherwise, vdev_inuse() will blindly overwrite the spare. | |
4320 | */ | |
b128c09f | 4321 | if ((error = spa_validate_aux(spa, nvroot, txg, VDEV_ALLOC_ADD)) != 0) |
34dc7c2f | 4322 | return (spa_vdev_exit(spa, vd, txg, error)); |
34dc7c2f BB |
4323 | |
4324 | /* | |
4325 | * Transfer each new top-level vdev from vd to rvd. | |
4326 | */ | |
d6320ddb | 4327 | for (c = 0; c < vd->vdev_children; c++) { |
428870ff BB |
4328 | |
4329 | /* | |
4330 | * Set the vdev id to the first hole, if one exists. | |
4331 | */ | |
4332 | for (id = 0; id < rvd->vdev_children; id++) { | |
4333 | if (rvd->vdev_child[id]->vdev_ishole) { | |
4334 | vdev_free(rvd->vdev_child[id]); | |
4335 | break; | |
4336 | } | |
4337 | } | |
34dc7c2f BB |
4338 | tvd = vd->vdev_child[c]; |
4339 | vdev_remove_child(vd, tvd); | |
428870ff | 4340 | tvd->vdev_id = id; |
34dc7c2f BB |
4341 | vdev_add_child(rvd, tvd); |
4342 | vdev_config_dirty(tvd); | |
4343 | } | |
4344 | ||
4345 | if (nspares != 0) { | |
4346 | spa_set_aux_vdevs(&spa->spa_spares, spares, nspares, | |
4347 | ZPOOL_CONFIG_SPARES); | |
4348 | spa_load_spares(spa); | |
4349 | spa->spa_spares.sav_sync = B_TRUE; | |
4350 | } | |
4351 | ||
4352 | if (nl2cache != 0) { | |
4353 | spa_set_aux_vdevs(&spa->spa_l2cache, l2cache, nl2cache, | |
4354 | ZPOOL_CONFIG_L2CACHE); | |
4355 | spa_load_l2cache(spa); | |
4356 | spa->spa_l2cache.sav_sync = B_TRUE; | |
4357 | } | |
4358 | ||
4359 | /* | |
4360 | * We have to be careful when adding new vdevs to an existing pool. | |
4361 | * If other threads start allocating from these vdevs before we | |
4362 | * sync the config cache, and we lose power, then upon reboot we may | |
4363 | * fail to open the pool because there are DVAs that the config cache | |
4364 | * can't translate. Therefore, we first add the vdevs without | |
4365 | * initializing metaslabs; sync the config cache (via spa_vdev_exit()); | |
4366 | * and then let spa_config_update() initialize the new metaslabs. | |
4367 | * | |
4368 | * spa_load() checks for added-but-not-initialized vdevs, so that | |
4369 | * if we lose power at any point in this sequence, the remaining | |
4370 | * steps will be completed the next time we load the pool. | |
4371 | */ | |
4372 | (void) spa_vdev_exit(spa, vd, txg, 0); | |
4373 | ||
4374 | mutex_enter(&spa_namespace_lock); | |
4375 | spa_config_update(spa, SPA_CONFIG_UPDATE_POOL); | |
4376 | mutex_exit(&spa_namespace_lock); | |
4377 | ||
4378 | return (0); | |
4379 | } | |
4380 | ||
4381 | /* | |
4382 | * Attach a device to a mirror. The arguments are the path to any device | |
4383 | * in the mirror, and the nvroot for the new device. If the path specifies | |
4384 | * a device that is not mirrored, we automatically insert the mirror vdev. | |
4385 | * | |
4386 | * If 'replacing' is specified, the new device is intended to replace the | |
4387 | * existing device; in this case the two devices are made into their own | |
4388 | * mirror using the 'replacing' vdev, which is functionally identical to | |
4389 | * the mirror vdev (it actually reuses all the same ops) but has a few | |
4390 | * extra rules: you can't attach to it after it's been created, and upon | |
4391 | * completion of resilvering, the first disk (the one being replaced) | |
4392 | * is automatically detached. | |
4393 | */ | |
4394 | int | |
4395 | spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing) | |
4396 | { | |
428870ff | 4397 | uint64_t txg, dtl_max_txg; |
34dc7c2f BB |
4398 | vdev_t *oldvd, *newvd, *newrootvd, *pvd, *tvd; |
4399 | vdev_ops_t *pvops; | |
b128c09f BB |
4400 | char *oldvdpath, *newvdpath; |
4401 | int newvd_isspare; | |
4402 | int error; | |
2e528b49 | 4403 | ASSERTV(vdev_t *rvd = spa->spa_root_vdev); |
34dc7c2f | 4404 | |
572e2857 BB |
4405 | ASSERT(spa_writeable(spa)); |
4406 | ||
34dc7c2f BB |
4407 | txg = spa_vdev_enter(spa); |
4408 | ||
b128c09f | 4409 | oldvd = spa_lookup_by_guid(spa, guid, B_FALSE); |
34dc7c2f BB |
4410 | |
4411 | if (oldvd == NULL) | |
4412 | return (spa_vdev_exit(spa, NULL, txg, ENODEV)); | |
4413 | ||
4414 | if (!oldvd->vdev_ops->vdev_op_leaf) | |
4415 | return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); | |
4416 | ||
4417 | pvd = oldvd->vdev_parent; | |
4418 | ||
4419 | if ((error = spa_config_parse(spa, &newrootvd, nvroot, NULL, 0, | |
5ffb9d1d | 4420 | VDEV_ALLOC_ATTACH)) != 0) |
34dc7c2f BB |
4421 | return (spa_vdev_exit(spa, NULL, txg, EINVAL)); |
4422 | ||
4423 | if (newrootvd->vdev_children != 1) | |
4424 | return (spa_vdev_exit(spa, newrootvd, txg, EINVAL)); | |
4425 | ||
4426 | newvd = newrootvd->vdev_child[0]; | |
4427 | ||
4428 | if (!newvd->vdev_ops->vdev_op_leaf) | |
4429 | return (spa_vdev_exit(spa, newrootvd, txg, EINVAL)); | |
4430 | ||
4431 | if ((error = vdev_create(newrootvd, txg, replacing)) != 0) | |
4432 | return (spa_vdev_exit(spa, newrootvd, txg, error)); | |
4433 | ||
4434 | /* | |
4435 | * Spares can't replace logs | |
4436 | */ | |
b128c09f | 4437 | if (oldvd->vdev_top->vdev_islog && newvd->vdev_isspare) |
34dc7c2f BB |
4438 | return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); |
4439 | ||
4440 | if (!replacing) { | |
4441 | /* | |
4442 | * For attach, the only allowable parent is a mirror or the root | |
4443 | * vdev. | |
4444 | */ | |
4445 | if (pvd->vdev_ops != &vdev_mirror_ops && | |
4446 | pvd->vdev_ops != &vdev_root_ops) | |
4447 | return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); | |
4448 | ||
4449 | pvops = &vdev_mirror_ops; | |
4450 | } else { | |
4451 | /* | |
4452 | * Active hot spares can only be replaced by inactive hot | |
4453 | * spares. | |
4454 | */ | |
4455 | if (pvd->vdev_ops == &vdev_spare_ops && | |
572e2857 | 4456 | oldvd->vdev_isspare && |
34dc7c2f BB |
4457 | !spa_has_spare(spa, newvd->vdev_guid)) |
4458 | return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); | |
4459 | ||
4460 | /* | |
4461 | * If the source is a hot spare, and the parent isn't already a | |
4462 | * spare, then we want to create a new hot spare. Otherwise, we | |
4463 | * want to create a replacing vdev. The user is not allowed to | |
4464 | * attach to a spared vdev child unless the 'isspare' state is | |
4465 | * the same (spare replaces spare, non-spare replaces | |
4466 | * non-spare). | |
4467 | */ | |
572e2857 BB |
4468 | if (pvd->vdev_ops == &vdev_replacing_ops && |
4469 | spa_version(spa) < SPA_VERSION_MULTI_REPLACE) { | |
34dc7c2f | 4470 | return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); |
572e2857 BB |
4471 | } else if (pvd->vdev_ops == &vdev_spare_ops && |
4472 | newvd->vdev_isspare != oldvd->vdev_isspare) { | |
34dc7c2f | 4473 | return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); |
572e2857 BB |
4474 | } |
4475 | ||
4476 | if (newvd->vdev_isspare) | |
34dc7c2f BB |
4477 | pvops = &vdev_spare_ops; |
4478 | else | |
4479 | pvops = &vdev_replacing_ops; | |
4480 | } | |
4481 | ||
4482 | /* | |
9babb374 | 4483 | * Make sure the new device is big enough. |
34dc7c2f | 4484 | */ |
9babb374 | 4485 | if (newvd->vdev_asize < vdev_get_min_asize(oldvd)) |
34dc7c2f BB |
4486 | return (spa_vdev_exit(spa, newrootvd, txg, EOVERFLOW)); |
4487 | ||
4488 | /* | |
4489 | * The new device cannot have a higher alignment requirement | |
4490 | * than the top-level vdev. | |
4491 | */ | |
4492 | if (newvd->vdev_ashift > oldvd->vdev_top->vdev_ashift) | |
4493 | return (spa_vdev_exit(spa, newrootvd, txg, EDOM)); | |
4494 | ||
4495 | /* | |
4496 | * If this is an in-place replacement, update oldvd's path and devid | |
4497 | * to make it distinguishable from newvd, and unopenable from now on. | |
4498 | */ | |
4499 | if (strcmp(oldvd->vdev_path, newvd->vdev_path) == 0) { | |
4500 | spa_strfree(oldvd->vdev_path); | |
4501 | oldvd->vdev_path = kmem_alloc(strlen(newvd->vdev_path) + 5, | |
b8d06fca | 4502 | KM_PUSHPAGE); |
34dc7c2f BB |
4503 | (void) sprintf(oldvd->vdev_path, "%s/%s", |
4504 | newvd->vdev_path, "old"); | |
4505 | if (oldvd->vdev_devid != NULL) { | |
4506 | spa_strfree(oldvd->vdev_devid); | |
4507 | oldvd->vdev_devid = NULL; | |
4508 | } | |
4509 | } | |
4510 | ||
572e2857 BB |
4511 | /* mark the device being resilvered */ |
4512 | newvd->vdev_resilvering = B_TRUE; | |
4513 | ||
34dc7c2f BB |
4514 | /* |
4515 | * If the parent is not a mirror, or if we're replacing, insert the new | |
4516 | * mirror/replacing/spare vdev above oldvd. | |
4517 | */ | |
4518 | if (pvd->vdev_ops != pvops) | |
4519 | pvd = vdev_add_parent(oldvd, pvops); | |
4520 | ||
4521 | ASSERT(pvd->vdev_top->vdev_parent == rvd); | |
4522 | ASSERT(pvd->vdev_ops == pvops); | |
4523 | ASSERT(oldvd->vdev_parent == pvd); | |
4524 | ||
4525 | /* | |
4526 | * Extract the new device from its root and add it to pvd. | |
4527 | */ | |
4528 | vdev_remove_child(newrootvd, newvd); | |
4529 | newvd->vdev_id = pvd->vdev_children; | |
428870ff | 4530 | newvd->vdev_crtxg = oldvd->vdev_crtxg; |
34dc7c2f BB |
4531 | vdev_add_child(pvd, newvd); |
4532 | ||
34dc7c2f BB |
4533 | tvd = newvd->vdev_top; |
4534 | ASSERT(pvd->vdev_top == tvd); | |
4535 | ASSERT(tvd->vdev_parent == rvd); | |
4536 | ||
4537 | vdev_config_dirty(tvd); | |
4538 | ||
4539 | /* | |
428870ff BB |
4540 | * Set newvd's DTL to [TXG_INITIAL, dtl_max_txg) so that we account |
4541 | * for any dmu_sync-ed blocks. It will propagate upward when | |
4542 | * spa_vdev_exit() calls vdev_dtl_reassess(). | |
34dc7c2f | 4543 | */ |
428870ff | 4544 | dtl_max_txg = txg + TXG_CONCURRENT_STATES; |
34dc7c2f | 4545 | |
428870ff BB |
4546 | vdev_dtl_dirty(newvd, DTL_MISSING, TXG_INITIAL, |
4547 | dtl_max_txg - TXG_INITIAL); | |
34dc7c2f | 4548 | |
9babb374 | 4549 | if (newvd->vdev_isspare) { |
34dc7c2f | 4550 | spa_spare_activate(newvd); |
26685276 | 4551 | spa_event_notify(spa, newvd, FM_EREPORT_ZFS_DEVICE_SPARE); |
9babb374 BB |
4552 | } |
4553 | ||
b128c09f BB |
4554 | oldvdpath = spa_strdup(oldvd->vdev_path); |
4555 | newvdpath = spa_strdup(newvd->vdev_path); | |
4556 | newvd_isspare = newvd->vdev_isspare; | |
34dc7c2f BB |
4557 | |
4558 | /* | |
4559 | * Mark newvd's DTL dirty in this txg. | |
4560 | */ | |
4561 | vdev_dirty(tvd, VDD_DTL, newvd, txg); | |
4562 | ||
428870ff BB |
4563 | /* |
4564 | * Restart the resilver | |
4565 | */ | |
4566 | dsl_resilver_restart(spa->spa_dsl_pool, dtl_max_txg); | |
4567 | ||
4568 | /* | |
4569 | * Commit the config | |
4570 | */ | |
4571 | (void) spa_vdev_exit(spa, newrootvd, dtl_max_txg, 0); | |
34dc7c2f | 4572 | |
6f1ffb06 | 4573 | spa_history_log_internal(spa, "vdev attach", NULL, |
428870ff | 4574 | "%s vdev=%s %s vdev=%s", |
45d1cae3 BB |
4575 | replacing && newvd_isspare ? "spare in" : |
4576 | replacing ? "replace" : "attach", newvdpath, | |
4577 | replacing ? "for" : "to", oldvdpath); | |
b128c09f BB |
4578 | |
4579 | spa_strfree(oldvdpath); | |
4580 | spa_strfree(newvdpath); | |
4581 | ||
572e2857 | 4582 | if (spa->spa_bootfs) |
26685276 | 4583 | spa_event_notify(spa, newvd, FM_EREPORT_ZFS_BOOTFS_VDEV_ATTACH); |
572e2857 | 4584 | |
34dc7c2f BB |
4585 | return (0); |
4586 | } | |
4587 | ||
4588 | /* | |
4589 | * Detach a device from a mirror or replacing vdev. | |
d3cc8b15 | 4590 | * |
34dc7c2f BB |
4591 | * If 'replace_done' is specified, only detach if the parent |
4592 | * is a replacing vdev. | |
4593 | */ | |
4594 | int | |
fb5f0bc8 | 4595 | spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, int replace_done) |
34dc7c2f BB |
4596 | { |
4597 | uint64_t txg; | |
fb5f0bc8 | 4598 | int error; |
34dc7c2f BB |
4599 | vdev_t *vd, *pvd, *cvd, *tvd; |
4600 | boolean_t unspare = B_FALSE; | |
d4ed6673 | 4601 | uint64_t unspare_guid = 0; |
428870ff | 4602 | char *vdpath; |
d6320ddb | 4603 | int c, t; |
2e528b49 | 4604 | ASSERTV(vdev_t *rvd = spa->spa_root_vdev); |
572e2857 BB |
4605 | ASSERT(spa_writeable(spa)); |
4606 | ||
34dc7c2f BB |
4607 | txg = spa_vdev_enter(spa); |
4608 | ||
b128c09f | 4609 | vd = spa_lookup_by_guid(spa, guid, B_FALSE); |
34dc7c2f BB |
4610 | |
4611 | if (vd == NULL) | |
4612 | return (spa_vdev_exit(spa, NULL, txg, ENODEV)); | |
4613 | ||
4614 | if (!vd->vdev_ops->vdev_op_leaf) | |
4615 | return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); | |
4616 | ||
4617 | pvd = vd->vdev_parent; | |
4618 | ||
fb5f0bc8 BB |
4619 | /* |
4620 | * If the parent/child relationship is not as expected, don't do it. | |
4621 | * Consider M(A,R(B,C)) -- that is, a mirror of A with a replacing | |
4622 | * vdev that's replacing B with C. The user's intent in replacing | |
4623 | * is to go from M(A,B) to M(A,C). If the user decides to cancel | |
4624 | * the replace by detaching C, the expected behavior is to end up | |
4625 | * M(A,B). But suppose that right after deciding to detach C, | |
4626 | * the replacement of B completes. We would have M(A,C), and then | |
4627 | * ask to detach C, which would leave us with just A -- not what | |
4628 | * the user wanted. To prevent this, we make sure that the | |
4629 | * parent/child relationship hasn't changed -- in this example, | |
4630 | * that C's parent is still the replacing vdev R. | |
4631 | */ | |
4632 | if (pvd->vdev_guid != pguid && pguid != 0) | |
4633 | return (spa_vdev_exit(spa, NULL, txg, EBUSY)); | |
4634 | ||
34dc7c2f | 4635 | /* |
572e2857 | 4636 | * Only 'replacing' or 'spare' vdevs can be replaced. |
34dc7c2f | 4637 | */ |
572e2857 BB |
4638 | if (replace_done && pvd->vdev_ops != &vdev_replacing_ops && |
4639 | pvd->vdev_ops != &vdev_spare_ops) | |
4640 | return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); | |
34dc7c2f BB |
4641 | |
4642 | ASSERT(pvd->vdev_ops != &vdev_spare_ops || | |
4643 | spa_version(spa) >= SPA_VERSION_SPARES); | |
4644 | ||
4645 | /* | |
4646 | * Only mirror, replacing, and spare vdevs support detach. | |
4647 | */ | |
4648 | if (pvd->vdev_ops != &vdev_replacing_ops && | |
4649 | pvd->vdev_ops != &vdev_mirror_ops && | |
4650 | pvd->vdev_ops != &vdev_spare_ops) | |
4651 | return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); | |
4652 | ||
4653 | /* | |
fb5f0bc8 BB |
4654 | * If this device has the only valid copy of some data, |
4655 | * we cannot safely detach it. | |
34dc7c2f | 4656 | */ |
fb5f0bc8 | 4657 | if (vdev_dtl_required(vd)) |
34dc7c2f BB |
4658 | return (spa_vdev_exit(spa, NULL, txg, EBUSY)); |
4659 | ||
fb5f0bc8 | 4660 | ASSERT(pvd->vdev_children >= 2); |
34dc7c2f | 4661 | |
b128c09f BB |
4662 | /* |
4663 | * If we are detaching the second disk from a replacing vdev, then | |
4664 | * check to see if we changed the original vdev's path to have "/old" | |
4665 | * at the end in spa_vdev_attach(). If so, undo that change now. | |
4666 | */ | |
572e2857 BB |
4667 | if (pvd->vdev_ops == &vdev_replacing_ops && vd->vdev_id > 0 && |
4668 | vd->vdev_path != NULL) { | |
4669 | size_t len = strlen(vd->vdev_path); | |
4670 | ||
d6320ddb | 4671 | for (c = 0; c < pvd->vdev_children; c++) { |
572e2857 BB |
4672 | cvd = pvd->vdev_child[c]; |
4673 | ||
4674 | if (cvd == vd || cvd->vdev_path == NULL) | |
4675 | continue; | |
4676 | ||
4677 | if (strncmp(cvd->vdev_path, vd->vdev_path, len) == 0 && | |
4678 | strcmp(cvd->vdev_path + len, "/old") == 0) { | |
4679 | spa_strfree(cvd->vdev_path); | |
4680 | cvd->vdev_path = spa_strdup(vd->vdev_path); | |
4681 | break; | |
4682 | } | |
b128c09f BB |
4683 | } |
4684 | } | |
4685 | ||
34dc7c2f BB |
4686 | /* |
4687 | * If we are detaching the original disk from a spare, then it implies | |
4688 | * that the spare should become a real disk, and be removed from the | |
4689 | * active spare list for the pool. | |
4690 | */ | |
4691 | if (pvd->vdev_ops == &vdev_spare_ops && | |
572e2857 BB |
4692 | vd->vdev_id == 0 && |
4693 | pvd->vdev_child[pvd->vdev_children - 1]->vdev_isspare) | |
34dc7c2f BB |
4694 | unspare = B_TRUE; |
4695 | ||
4696 | /* | |
4697 | * Erase the disk labels so the disk can be used for other things. | |
4698 | * This must be done after all other error cases are handled, | |
4699 | * but before we disembowel vd (so we can still do I/O to it). | |
4700 | * But if we can't do it, don't treat the error as fatal -- | |
4701 | * it may be that the unwritability of the disk is the reason | |
4702 | * it's being detached! | |
4703 | */ | |
4704 | error = vdev_label_init(vd, 0, VDEV_LABEL_REMOVE); | |
4705 | ||
4706 | /* | |
4707 | * Remove vd from its parent and compact the parent's children. | |
4708 | */ | |
4709 | vdev_remove_child(pvd, vd); | |
4710 | vdev_compact_children(pvd); | |
4711 | ||
4712 | /* | |
4713 | * Remember one of the remaining children so we can get tvd below. | |
4714 | */ | |
572e2857 | 4715 | cvd = pvd->vdev_child[pvd->vdev_children - 1]; |
34dc7c2f BB |
4716 | |
4717 | /* | |
4718 | * If we need to remove the remaining child from the list of hot spares, | |
fb5f0bc8 BB |
4719 | * do it now, marking the vdev as no longer a spare in the process. |
4720 | * We must do this before vdev_remove_parent(), because that can | |
4721 | * change the GUID if it creates a new toplevel GUID. For a similar | |
4722 | * reason, we must remove the spare now, in the same txg as the detach; | |
4723 | * otherwise someone could attach a new sibling, change the GUID, and | |
4724 | * the subsequent attempt to spa_vdev_remove(unspare_guid) would fail. | |
34dc7c2f BB |
4725 | */ |
4726 | if (unspare) { | |
4727 | ASSERT(cvd->vdev_isspare); | |
4728 | spa_spare_remove(cvd); | |
4729 | unspare_guid = cvd->vdev_guid; | |
fb5f0bc8 | 4730 | (void) spa_vdev_remove(spa, unspare_guid, B_TRUE); |
572e2857 | 4731 | cvd->vdev_unspare = B_TRUE; |
34dc7c2f BB |
4732 | } |
4733 | ||
428870ff BB |
4734 | /* |
4735 | * If the parent mirror/replacing vdev only has one child, | |
4736 | * the parent is no longer needed. Remove it from the tree. | |
4737 | */ | |
572e2857 BB |
4738 | if (pvd->vdev_children == 1) { |
4739 | if (pvd->vdev_ops == &vdev_spare_ops) | |
4740 | cvd->vdev_unspare = B_FALSE; | |
428870ff | 4741 | vdev_remove_parent(cvd); |
572e2857 BB |
4742 | } |
4743 | ||
428870ff BB |
4744 | |
4745 | /* | |
4746 | * We don't set tvd until now because the parent we just removed | |
4747 | * may have been the previous top-level vdev. | |
4748 | */ | |
4749 | tvd = cvd->vdev_top; | |
4750 | ASSERT(tvd->vdev_parent == rvd); | |
4751 | ||
4752 | /* | |
4753 | * Reevaluate the parent vdev state. | |
4754 | */ | |
4755 | vdev_propagate_state(cvd); | |
4756 | ||
4757 | /* | |
4758 | * If the 'autoexpand' property is set on the pool then automatically | |
4759 | * try to expand the size of the pool. For example if the device we | |
4760 | * just detached was smaller than the others, it may be possible to | |
4761 | * add metaslabs (i.e. grow the pool). We need to reopen the vdev | |
4762 | * first so that we can obtain the updated sizes of the leaf vdevs. | |
4763 | */ | |
4764 | if (spa->spa_autoexpand) { | |
4765 | vdev_reopen(tvd); | |
4766 | vdev_expand(tvd, txg); | |
4767 | } | |
4768 | ||
4769 | vdev_config_dirty(tvd); | |
4770 | ||
4771 | /* | |
4772 | * Mark vd's DTL as dirty in this txg. vdev_dtl_sync() will see that | |
4773 | * vd->vdev_detached is set and free vd's DTL object in syncing context. | |
4774 | * But first make sure we're not on any *other* txg's DTL list, to | |
4775 | * prevent vd from being accessed after it's freed. | |
4776 | */ | |
4777 | vdpath = spa_strdup(vd->vdev_path); | |
d6320ddb | 4778 | for (t = 0; t < TXG_SIZE; t++) |
428870ff BB |
4779 | (void) txg_list_remove_this(&tvd->vdev_dtl_list, vd, t); |
4780 | vd->vdev_detached = B_TRUE; | |
4781 | vdev_dirty(tvd, VDD_DTL, vd, txg); | |
4782 | ||
26685276 | 4783 | spa_event_notify(spa, vd, FM_EREPORT_ZFS_DEVICE_REMOVE); |
428870ff | 4784 | |
572e2857 BB |
4785 | /* hang on to the spa before we release the lock */ |
4786 | spa_open_ref(spa, FTAG); | |
4787 | ||
428870ff BB |
4788 | error = spa_vdev_exit(spa, vd, txg, 0); |
4789 | ||
6f1ffb06 | 4790 | spa_history_log_internal(spa, "detach", NULL, |
428870ff BB |
4791 | "vdev=%s", vdpath); |
4792 | spa_strfree(vdpath); | |
4793 | ||
4794 | /* | |
4795 | * If this was the removal of the original device in a hot spare vdev, | |
4796 | * then we want to go through and remove the device from the hot spare | |
4797 | * list of every other pool. | |
4798 | */ | |
4799 | if (unspare) { | |
572e2857 BB |
4800 | spa_t *altspa = NULL; |
4801 | ||
428870ff | 4802 | mutex_enter(&spa_namespace_lock); |
572e2857 BB |
4803 | while ((altspa = spa_next(altspa)) != NULL) { |
4804 | if (altspa->spa_state != POOL_STATE_ACTIVE || | |
4805 | altspa == spa) | |
428870ff | 4806 | continue; |
572e2857 BB |
4807 | |
4808 | spa_open_ref(altspa, FTAG); | |
428870ff | 4809 | mutex_exit(&spa_namespace_lock); |
572e2857 | 4810 | (void) spa_vdev_remove(altspa, unspare_guid, B_TRUE); |
428870ff | 4811 | mutex_enter(&spa_namespace_lock); |
572e2857 | 4812 | spa_close(altspa, FTAG); |
428870ff BB |
4813 | } |
4814 | mutex_exit(&spa_namespace_lock); | |
572e2857 BB |
4815 | |
4816 | /* search the rest of the vdevs for spares to remove */ | |
4817 | spa_vdev_resilver_done(spa); | |
428870ff BB |
4818 | } |
4819 | ||
572e2857 BB |
4820 | /* all done with the spa; OK to release */ |
4821 | mutex_enter(&spa_namespace_lock); | |
4822 | spa_close(spa, FTAG); | |
4823 | mutex_exit(&spa_namespace_lock); | |
4824 | ||
428870ff BB |
4825 | return (error); |
4826 | } | |
4827 | ||
4828 | /* | |
4829 | * Split a set of devices from their mirrors, and create a new pool from them. | |
4830 | */ | |
4831 | int | |
4832 | spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config, | |
4833 | nvlist_t *props, boolean_t exp) | |
4834 | { | |
4835 | int error = 0; | |
4836 | uint64_t txg, *glist; | |
4837 | spa_t *newspa; | |
4838 | uint_t c, children, lastlog; | |
4839 | nvlist_t **child, *nvl, *tmp; | |
4840 | dmu_tx_t *tx; | |
4841 | char *altroot = NULL; | |
4842 | vdev_t *rvd, **vml = NULL; /* vdev modify list */ | |
4843 | boolean_t activate_slog; | |
4844 | ||
572e2857 | 4845 | ASSERT(spa_writeable(spa)); |
428870ff BB |
4846 | |
4847 | txg = spa_vdev_enter(spa); | |
4848 | ||
4849 | /* clear the log and flush everything up to now */ | |
4850 | activate_slog = spa_passivate_log(spa); | |
4851 | (void) spa_vdev_config_exit(spa, NULL, txg, 0, FTAG); | |
4852 | error = spa_offline_log(spa); | |
4853 | txg = spa_vdev_config_enter(spa); | |
4854 | ||
4855 | if (activate_slog) | |
4856 | spa_activate_log(spa); | |
4857 | ||
4858 | if (error != 0) | |
4859 | return (spa_vdev_exit(spa, NULL, txg, error)); | |
4860 | ||
4861 | /* check new spa name before going any further */ | |
4862 | if (spa_lookup(newname) != NULL) | |
4863 | return (spa_vdev_exit(spa, NULL, txg, EEXIST)); | |
4864 | ||
4865 | /* | |
4866 | * scan through all the children to ensure they're all mirrors | |
4867 | */ | |
4868 | if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvl) != 0 || | |
4869 | nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_CHILDREN, &child, | |
4870 | &children) != 0) | |
4871 | return (spa_vdev_exit(spa, NULL, txg, EINVAL)); | |
4872 | ||
4873 | /* first, check to ensure we've got the right child count */ | |
4874 | rvd = spa->spa_root_vdev; | |
4875 | lastlog = 0; | |
4876 | for (c = 0; c < rvd->vdev_children; c++) { | |
4877 | vdev_t *vd = rvd->vdev_child[c]; | |
4878 | ||
4879 | /* don't count the holes & logs as children */ | |
4880 | if (vd->vdev_islog || vd->vdev_ishole) { | |
4881 | if (lastlog == 0) | |
4882 | lastlog = c; | |
4883 | continue; | |
4884 | } | |
4885 | ||
4886 | lastlog = 0; | |
4887 | } | |
4888 | if (children != (lastlog != 0 ? lastlog : rvd->vdev_children)) | |
4889 | return (spa_vdev_exit(spa, NULL, txg, EINVAL)); | |
4890 | ||
4891 | /* next, ensure no spare or cache devices are part of the split */ | |
4892 | if (nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_SPARES, &tmp) == 0 || | |
4893 | nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_L2CACHE, &tmp) == 0) | |
4894 | return (spa_vdev_exit(spa, NULL, txg, EINVAL)); | |
4895 | ||
b8d06fca RY |
4896 | vml = kmem_zalloc(children * sizeof (vdev_t *), KM_PUSHPAGE); |
4897 | glist = kmem_zalloc(children * sizeof (uint64_t), KM_PUSHPAGE); | |
428870ff BB |
4898 | |
4899 | /* then, loop over each vdev and validate it */ | |
4900 | for (c = 0; c < children; c++) { | |
4901 | uint64_t is_hole = 0; | |
4902 | ||
4903 | (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE, | |
4904 | &is_hole); | |
4905 | ||
4906 | if (is_hole != 0) { | |
4907 | if (spa->spa_root_vdev->vdev_child[c]->vdev_ishole || | |
4908 | spa->spa_root_vdev->vdev_child[c]->vdev_islog) { | |
4909 | continue; | |
4910 | } else { | |
2e528b49 | 4911 | error = SET_ERROR(EINVAL); |
428870ff BB |
4912 | break; |
4913 | } | |
4914 | } | |
4915 | ||
4916 | /* which disk is going to be split? */ | |
4917 | if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_GUID, | |
4918 | &glist[c]) != 0) { | |
2e528b49 | 4919 | error = SET_ERROR(EINVAL); |
428870ff BB |
4920 | break; |
4921 | } | |
4922 | ||
4923 | /* look it up in the spa */ | |
4924 | vml[c] = spa_lookup_by_guid(spa, glist[c], B_FALSE); | |
4925 | if (vml[c] == NULL) { | |
2e528b49 | 4926 | error = SET_ERROR(ENODEV); |
428870ff BB |
4927 | break; |
4928 | } | |
4929 | ||
4930 | /* make sure there's nothing stopping the split */ | |
4931 | if (vml[c]->vdev_parent->vdev_ops != &vdev_mirror_ops || | |
4932 | vml[c]->vdev_islog || | |
4933 | vml[c]->vdev_ishole || | |
4934 | vml[c]->vdev_isspare || | |
4935 | vml[c]->vdev_isl2cache || | |
4936 | !vdev_writeable(vml[c]) || | |
4937 | vml[c]->vdev_children != 0 || | |
4938 | vml[c]->vdev_state != VDEV_STATE_HEALTHY || | |
4939 | c != spa->spa_root_vdev->vdev_child[c]->vdev_id) { | |
2e528b49 | 4940 | error = SET_ERROR(EINVAL); |
428870ff BB |
4941 | break; |
4942 | } | |
4943 | ||
4944 | if (vdev_dtl_required(vml[c])) { | |
2e528b49 | 4945 | error = SET_ERROR(EBUSY); |
428870ff BB |
4946 | break; |
4947 | } | |
4948 | ||
4949 | /* we need certain info from the top level */ | |
4950 | VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_ARRAY, | |
4951 | vml[c]->vdev_top->vdev_ms_array) == 0); | |
4952 | VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_SHIFT, | |
4953 | vml[c]->vdev_top->vdev_ms_shift) == 0); | |
4954 | VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_ASIZE, | |
4955 | vml[c]->vdev_top->vdev_asize) == 0); | |
4956 | VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_ASHIFT, | |
4957 | vml[c]->vdev_top->vdev_ashift) == 0); | |
4958 | } | |
4959 | ||
4960 | if (error != 0) { | |
4961 | kmem_free(vml, children * sizeof (vdev_t *)); | |
4962 | kmem_free(glist, children * sizeof (uint64_t)); | |
4963 | return (spa_vdev_exit(spa, NULL, txg, error)); | |
4964 | } | |
4965 | ||
4966 | /* stop writers from using the disks */ | |
4967 | for (c = 0; c < children; c++) { | |
4968 | if (vml[c] != NULL) | |
4969 | vml[c]->vdev_offline = B_TRUE; | |
4970 | } | |
4971 | vdev_reopen(spa->spa_root_vdev); | |
34dc7c2f BB |
4972 | |
4973 | /* | |
428870ff BB |
4974 | * Temporarily record the splitting vdevs in the spa config. This |
4975 | * will disappear once the config is regenerated. | |
34dc7c2f | 4976 | */ |
b8d06fca | 4977 | VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME, KM_PUSHPAGE) == 0); |
428870ff BB |
4978 | VERIFY(nvlist_add_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST, |
4979 | glist, children) == 0); | |
4980 | kmem_free(glist, children * sizeof (uint64_t)); | |
34dc7c2f | 4981 | |
428870ff BB |
4982 | mutex_enter(&spa->spa_props_lock); |
4983 | VERIFY(nvlist_add_nvlist(spa->spa_config, ZPOOL_CONFIG_SPLIT, | |
4984 | nvl) == 0); | |
4985 | mutex_exit(&spa->spa_props_lock); | |
4986 | spa->spa_config_splitting = nvl; | |
4987 | vdev_config_dirty(spa->spa_root_vdev); | |
4988 | ||
4989 | /* configure and create the new pool */ | |
4990 | VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, newname) == 0); | |
4991 | VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, | |
4992 | exp ? POOL_STATE_EXPORTED : POOL_STATE_ACTIVE) == 0); | |
4993 | VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, | |
4994 | spa_version(spa)) == 0); | |
4995 | VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG, | |
4996 | spa->spa_config_txg) == 0); | |
4997 | VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID, | |
4998 | spa_generate_guid(NULL)) == 0); | |
4999 | (void) nvlist_lookup_string(props, | |
5000 | zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot); | |
34dc7c2f | 5001 | |
428870ff BB |
5002 | /* add the new pool to the namespace */ |
5003 | newspa = spa_add(newname, config, altroot); | |
5004 | newspa->spa_config_txg = spa->spa_config_txg; | |
5005 | spa_set_log_state(newspa, SPA_LOG_CLEAR); | |
5006 | ||
5007 | /* release the spa config lock, retaining the namespace lock */ | |
5008 | spa_vdev_config_exit(spa, NULL, txg, 0, FTAG); | |
5009 | ||
5010 | if (zio_injection_enabled) | |
5011 | zio_handle_panic_injection(spa, FTAG, 1); | |
5012 | ||
5013 | spa_activate(newspa, spa_mode_global); | |
5014 | spa_async_suspend(newspa); | |
5015 | ||
5016 | /* create the new pool from the disks of the original pool */ | |
5017 | error = spa_load(newspa, SPA_LOAD_IMPORT, SPA_IMPORT_ASSEMBLE, B_TRUE); | |
5018 | if (error) | |
5019 | goto out; | |
5020 | ||
5021 | /* if that worked, generate a real config for the new pool */ | |
5022 | if (newspa->spa_root_vdev != NULL) { | |
5023 | VERIFY(nvlist_alloc(&newspa->spa_config_splitting, | |
b8d06fca | 5024 | NV_UNIQUE_NAME, KM_PUSHPAGE) == 0); |
428870ff BB |
5025 | VERIFY(nvlist_add_uint64(newspa->spa_config_splitting, |
5026 | ZPOOL_CONFIG_SPLIT_GUID, spa_guid(spa)) == 0); | |
5027 | spa_config_set(newspa, spa_config_generate(newspa, NULL, -1ULL, | |
5028 | B_TRUE)); | |
9babb374 | 5029 | } |
34dc7c2f | 5030 | |
428870ff BB |
5031 | /* set the props */ |
5032 | if (props != NULL) { | |
5033 | spa_configfile_set(newspa, props, B_FALSE); | |
5034 | error = spa_prop_set(newspa, props); | |
5035 | if (error) | |
5036 | goto out; | |
5037 | } | |
34dc7c2f | 5038 | |
428870ff BB |
5039 | /* flush everything */ |
5040 | txg = spa_vdev_config_enter(newspa); | |
5041 | vdev_config_dirty(newspa->spa_root_vdev); | |
5042 | (void) spa_vdev_config_exit(newspa, NULL, txg, 0, FTAG); | |
34dc7c2f | 5043 | |
428870ff BB |
5044 | if (zio_injection_enabled) |
5045 | zio_handle_panic_injection(spa, FTAG, 2); | |
34dc7c2f | 5046 | |
428870ff | 5047 | spa_async_resume(newspa); |
34dc7c2f | 5048 | |
428870ff BB |
5049 | /* finally, update the original pool's config */ |
5050 | txg = spa_vdev_config_enter(spa); | |
5051 | tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); | |
5052 | error = dmu_tx_assign(tx, TXG_WAIT); | |
5053 | if (error != 0) | |
5054 | dmu_tx_abort(tx); | |
5055 | for (c = 0; c < children; c++) { | |
5056 | if (vml[c] != NULL) { | |
5057 | vdev_split(vml[c]); | |
5058 | if (error == 0) | |
6f1ffb06 MA |
5059 | spa_history_log_internal(spa, "detach", tx, |
5060 | "vdev=%s", vml[c]->vdev_path); | |
428870ff | 5061 | vdev_free(vml[c]); |
34dc7c2f | 5062 | } |
34dc7c2f | 5063 | } |
428870ff BB |
5064 | vdev_config_dirty(spa->spa_root_vdev); |
5065 | spa->spa_config_splitting = NULL; | |
5066 | nvlist_free(nvl); | |
5067 | if (error == 0) | |
5068 | dmu_tx_commit(tx); | |
5069 | (void) spa_vdev_exit(spa, NULL, txg, 0); | |
5070 | ||
5071 | if (zio_injection_enabled) | |
5072 | zio_handle_panic_injection(spa, FTAG, 3); | |
5073 | ||
5074 | /* split is complete; log a history record */ | |
6f1ffb06 MA |
5075 | spa_history_log_internal(newspa, "split", NULL, |
5076 | "from pool %s", spa_name(spa)); | |
428870ff BB |
5077 | |
5078 | kmem_free(vml, children * sizeof (vdev_t *)); | |
5079 | ||
5080 | /* if we're not going to mount the filesystems in userland, export */ | |
5081 | if (exp) | |
5082 | error = spa_export_common(newname, POOL_STATE_EXPORTED, NULL, | |
5083 | B_FALSE, B_FALSE); | |
5084 | ||
5085 | return (error); | |
5086 | ||
5087 | out: | |
5088 | spa_unload(newspa); | |
5089 | spa_deactivate(newspa); | |
5090 | spa_remove(newspa); | |
5091 | ||
5092 | txg = spa_vdev_config_enter(spa); | |
5093 | ||
5094 | /* re-online all offlined disks */ | |
5095 | for (c = 0; c < children; c++) { | |
5096 | if (vml[c] != NULL) | |
5097 | vml[c]->vdev_offline = B_FALSE; | |
5098 | } | |
5099 | vdev_reopen(spa->spa_root_vdev); | |
5100 | ||
5101 | nvlist_free(spa->spa_config_splitting); | |
5102 | spa->spa_config_splitting = NULL; | |
5103 | (void) spa_vdev_exit(spa, NULL, txg, error); | |
34dc7c2f | 5104 | |
428870ff | 5105 | kmem_free(vml, children * sizeof (vdev_t *)); |
34dc7c2f BB |
5106 | return (error); |
5107 | } | |
5108 | ||
b128c09f BB |
5109 | static nvlist_t * |
5110 | spa_nvlist_lookup_by_guid(nvlist_t **nvpp, int count, uint64_t target_guid) | |
34dc7c2f | 5111 | { |
d6320ddb BB |
5112 | int i; |
5113 | ||
5114 | for (i = 0; i < count; i++) { | |
b128c09f | 5115 | uint64_t guid; |
34dc7c2f | 5116 | |
b128c09f BB |
5117 | VERIFY(nvlist_lookup_uint64(nvpp[i], ZPOOL_CONFIG_GUID, |
5118 | &guid) == 0); | |
34dc7c2f | 5119 | |
b128c09f BB |
5120 | if (guid == target_guid) |
5121 | return (nvpp[i]); | |
34dc7c2f BB |
5122 | } |
5123 | ||
b128c09f | 5124 | return (NULL); |
34dc7c2f BB |
5125 | } |
5126 | ||
b128c09f BB |
5127 | static void |
5128 | spa_vdev_remove_aux(nvlist_t *config, char *name, nvlist_t **dev, int count, | |
5129 | nvlist_t *dev_to_remove) | |
34dc7c2f | 5130 | { |
b128c09f | 5131 | nvlist_t **newdev = NULL; |
d6320ddb | 5132 | int i, j; |
34dc7c2f | 5133 | |
b128c09f | 5134 | if (count > 1) |
b8d06fca | 5135 | newdev = kmem_alloc((count - 1) * sizeof (void *), KM_PUSHPAGE); |
34dc7c2f | 5136 | |
d6320ddb | 5137 | for (i = 0, j = 0; i < count; i++) { |
b128c09f BB |
5138 | if (dev[i] == dev_to_remove) |
5139 | continue; | |
b8d06fca | 5140 | VERIFY(nvlist_dup(dev[i], &newdev[j++], KM_PUSHPAGE) == 0); |
34dc7c2f BB |
5141 | } |
5142 | ||
b128c09f BB |
5143 | VERIFY(nvlist_remove(config, name, DATA_TYPE_NVLIST_ARRAY) == 0); |
5144 | VERIFY(nvlist_add_nvlist_array(config, name, newdev, count - 1) == 0); | |
34dc7c2f | 5145 | |
d6320ddb | 5146 | for (i = 0; i < count - 1; i++) |
b128c09f | 5147 | nvlist_free(newdev[i]); |
34dc7c2f | 5148 | |
b128c09f BB |
5149 | if (count > 1) |
5150 | kmem_free(newdev, (count - 1) * sizeof (void *)); | |
34dc7c2f BB |
5151 | } |
5152 | ||
428870ff BB |
5153 | /* |
5154 | * Evacuate the device. | |
5155 | */ | |
5156 | static int | |
5157 | spa_vdev_remove_evacuate(spa_t *spa, vdev_t *vd) | |
5158 | { | |
5159 | uint64_t txg; | |
5160 | int error = 0; | |
5161 | ||
5162 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
5163 | ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); | |
5164 | ASSERT(vd == vd->vdev_top); | |
5165 | ||
5166 | /* | |
5167 | * Evacuate the device. We don't hold the config lock as writer | |
5168 | * since we need to do I/O but we do keep the | |
5169 | * spa_namespace_lock held. Once this completes the device | |
5170 | * should no longer have any blocks allocated on it. | |
5171 | */ | |
5172 | if (vd->vdev_islog) { | |
5173 | if (vd->vdev_stat.vs_alloc != 0) | |
5174 | error = spa_offline_log(spa); | |
5175 | } else { | |
2e528b49 | 5176 | error = SET_ERROR(ENOTSUP); |
428870ff BB |
5177 | } |
5178 | ||
5179 | if (error) | |
5180 | return (error); | |
5181 | ||
5182 | /* | |
5183 | * The evacuation succeeded. Remove any remaining MOS metadata | |
5184 | * associated with this vdev, and wait for these changes to sync. | |
5185 | */ | |
c99c9001 | 5186 | ASSERT0(vd->vdev_stat.vs_alloc); |
428870ff BB |
5187 | txg = spa_vdev_config_enter(spa); |
5188 | vd->vdev_removing = B_TRUE; | |
5189 | vdev_dirty(vd, 0, NULL, txg); | |
5190 | vdev_config_dirty(vd); | |
5191 | spa_vdev_config_exit(spa, NULL, txg, 0, FTAG); | |
5192 | ||
5193 | return (0); | |
5194 | } | |
5195 | ||
5196 | /* | |
5197 | * Complete the removal by cleaning up the namespace. | |
5198 | */ | |
5199 | static void | |
5200 | spa_vdev_remove_from_namespace(spa_t *spa, vdev_t *vd) | |
5201 | { | |
5202 | vdev_t *rvd = spa->spa_root_vdev; | |
5203 | uint64_t id = vd->vdev_id; | |
5204 | boolean_t last_vdev = (id == (rvd->vdev_children - 1)); | |
5205 | ||
5206 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
5207 | ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); | |
5208 | ASSERT(vd == vd->vdev_top); | |
5209 | ||
5210 | /* | |
5211 | * Only remove any devices which are empty. | |
5212 | */ | |
5213 | if (vd->vdev_stat.vs_alloc != 0) | |
5214 | return; | |
5215 | ||
5216 | (void) vdev_label_init(vd, 0, VDEV_LABEL_REMOVE); | |
5217 | ||
5218 | if (list_link_active(&vd->vdev_state_dirty_node)) | |
5219 | vdev_state_clean(vd); | |
5220 | if (list_link_active(&vd->vdev_config_dirty_node)) | |
5221 | vdev_config_clean(vd); | |
5222 | ||
5223 | vdev_free(vd); | |
5224 | ||
5225 | if (last_vdev) { | |
5226 | vdev_compact_children(rvd); | |
5227 | } else { | |
5228 | vd = vdev_alloc_common(spa, id, 0, &vdev_hole_ops); | |
5229 | vdev_add_child(rvd, vd); | |
5230 | } | |
5231 | vdev_config_dirty(rvd); | |
5232 | ||
5233 | /* | |
5234 | * Reassess the health of our root vdev. | |
5235 | */ | |
5236 | vdev_reopen(rvd); | |
5237 | } | |
5238 | ||
5239 | /* | |
5240 | * Remove a device from the pool - | |
5241 | * | |
5242 | * Removing a device from the vdev namespace requires several steps | |
5243 | * and can take a significant amount of time. As a result we use | |
5244 | * the spa_vdev_config_[enter/exit] functions which allow us to | |
5245 | * grab and release the spa_config_lock while still holding the namespace | |
5246 | * lock. During each step the configuration is synced out. | |
d3cc8b15 WA |
5247 | * |
5248 | * Currently, this supports removing only hot spares, slogs, and level 2 ARC | |
5249 | * devices. | |
34dc7c2f BB |
5250 | */ |
5251 | int | |
5252 | spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare) | |
5253 | { | |
5254 | vdev_t *vd; | |
428870ff | 5255 | metaslab_group_t *mg; |
b128c09f | 5256 | nvlist_t **spares, **l2cache, *nv; |
fb5f0bc8 | 5257 | uint64_t txg = 0; |
428870ff | 5258 | uint_t nspares, nl2cache; |
34dc7c2f | 5259 | int error = 0; |
fb5f0bc8 | 5260 | boolean_t locked = MUTEX_HELD(&spa_namespace_lock); |
34dc7c2f | 5261 | |
572e2857 BB |
5262 | ASSERT(spa_writeable(spa)); |
5263 | ||
fb5f0bc8 BB |
5264 | if (!locked) |
5265 | txg = spa_vdev_enter(spa); | |
34dc7c2f | 5266 | |
b128c09f | 5267 | vd = spa_lookup_by_guid(spa, guid, B_FALSE); |
34dc7c2f BB |
5268 | |
5269 | if (spa->spa_spares.sav_vdevs != NULL && | |
34dc7c2f | 5270 | nvlist_lookup_nvlist_array(spa->spa_spares.sav_config, |
b128c09f BB |
5271 | ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0 && |
5272 | (nv = spa_nvlist_lookup_by_guid(spares, nspares, guid)) != NULL) { | |
5273 | /* | |
5274 | * Only remove the hot spare if it's not currently in use | |
5275 | * in this pool. | |
5276 | */ | |
5277 | if (vd == NULL || unspare) { | |
5278 | spa_vdev_remove_aux(spa->spa_spares.sav_config, | |
5279 | ZPOOL_CONFIG_SPARES, spares, nspares, nv); | |
5280 | spa_load_spares(spa); | |
5281 | spa->spa_spares.sav_sync = B_TRUE; | |
5282 | } else { | |
2e528b49 | 5283 | error = SET_ERROR(EBUSY); |
b128c09f BB |
5284 | } |
5285 | } else if (spa->spa_l2cache.sav_vdevs != NULL && | |
34dc7c2f | 5286 | nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config, |
b128c09f BB |
5287 | ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0 && |
5288 | (nv = spa_nvlist_lookup_by_guid(l2cache, nl2cache, guid)) != NULL) { | |
5289 | /* | |
5290 | * Cache devices can always be removed. | |
5291 | */ | |
5292 | spa_vdev_remove_aux(spa->spa_l2cache.sav_config, | |
5293 | ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache, nv); | |
34dc7c2f BB |
5294 | spa_load_l2cache(spa); |
5295 | spa->spa_l2cache.sav_sync = B_TRUE; | |
428870ff BB |
5296 | } else if (vd != NULL && vd->vdev_islog) { |
5297 | ASSERT(!locked); | |
5298 | ASSERT(vd == vd->vdev_top); | |
5299 | ||
5300 | /* | |
5301 | * XXX - Once we have bp-rewrite this should | |
5302 | * become the common case. | |
5303 | */ | |
5304 | ||
5305 | mg = vd->vdev_mg; | |
5306 | ||
5307 | /* | |
5308 | * Stop allocating from this vdev. | |
5309 | */ | |
5310 | metaslab_group_passivate(mg); | |
5311 | ||
5312 | /* | |
5313 | * Wait for the youngest allocations and frees to sync, | |
5314 | * and then wait for the deferral of those frees to finish. | |
5315 | */ | |
5316 | spa_vdev_config_exit(spa, NULL, | |
5317 | txg + TXG_CONCURRENT_STATES + TXG_DEFER_SIZE, 0, FTAG); | |
5318 | ||
5319 | /* | |
5320 | * Attempt to evacuate the vdev. | |
5321 | */ | |
5322 | error = spa_vdev_remove_evacuate(spa, vd); | |
5323 | ||
5324 | txg = spa_vdev_config_enter(spa); | |
5325 | ||
5326 | /* | |
5327 | * If we couldn't evacuate the vdev, unwind. | |
5328 | */ | |
5329 | if (error) { | |
5330 | metaslab_group_activate(mg); | |
5331 | return (spa_vdev_exit(spa, NULL, txg, error)); | |
5332 | } | |
5333 | ||
5334 | /* | |
5335 | * Clean up the vdev namespace. | |
5336 | */ | |
5337 | spa_vdev_remove_from_namespace(spa, vd); | |
5338 | ||
b128c09f BB |
5339 | } else if (vd != NULL) { |
5340 | /* | |
5341 | * Normal vdevs cannot be removed (yet). | |
5342 | */ | |
2e528b49 | 5343 | error = SET_ERROR(ENOTSUP); |
b128c09f BB |
5344 | } else { |
5345 | /* | |
5346 | * There is no vdev of any kind with the specified guid. | |
5347 | */ | |
2e528b49 | 5348 | error = SET_ERROR(ENOENT); |
34dc7c2f BB |
5349 | } |
5350 | ||
fb5f0bc8 BB |
5351 | if (!locked) |
5352 | return (spa_vdev_exit(spa, NULL, txg, error)); | |
5353 | ||
5354 | return (error); | |
34dc7c2f BB |
5355 | } |
5356 | ||
5357 | /* | |
5358 | * Find any device that's done replacing, or a vdev marked 'unspare' that's | |
d3cc8b15 | 5359 | * currently spared, so we can detach it. |
34dc7c2f BB |
5360 | */ |
5361 | static vdev_t * | |
5362 | spa_vdev_resilver_done_hunt(vdev_t *vd) | |
5363 | { | |
5364 | vdev_t *newvd, *oldvd; | |
d6320ddb | 5365 | int c; |
34dc7c2f | 5366 | |
d6320ddb | 5367 | for (c = 0; c < vd->vdev_children; c++) { |
34dc7c2f BB |
5368 | oldvd = spa_vdev_resilver_done_hunt(vd->vdev_child[c]); |
5369 | if (oldvd != NULL) | |
5370 | return (oldvd); | |
5371 | } | |
5372 | ||
621dd7bb GW |
5373 | if (vd->vdev_resilvering && vdev_dtl_empty(vd, DTL_MISSING) && |
5374 | vdev_dtl_empty(vd, DTL_OUTAGE)) { | |
5375 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
5376 | vd->vdev_resilvering = B_FALSE; | |
5377 | vdev_config_dirty(vd->vdev_top); | |
5378 | } | |
5379 | ||
34dc7c2f | 5380 | /* |
572e2857 BB |
5381 | * Check for a completed replacement. We always consider the first |
5382 | * vdev in the list to be the oldest vdev, and the last one to be | |
5383 | * the newest (see spa_vdev_attach() for how that works). In | |
5384 | * the case where the newest vdev is faulted, we will not automatically | |
5385 | * remove it after a resilver completes. This is OK as it will require | |
5386 | * user intervention to determine which disk the admin wishes to keep. | |
34dc7c2f | 5387 | */ |
572e2857 BB |
5388 | if (vd->vdev_ops == &vdev_replacing_ops) { |
5389 | ASSERT(vd->vdev_children > 1); | |
5390 | ||
5391 | newvd = vd->vdev_child[vd->vdev_children - 1]; | |
34dc7c2f | 5392 | oldvd = vd->vdev_child[0]; |
34dc7c2f | 5393 | |
fb5f0bc8 | 5394 | if (vdev_dtl_empty(newvd, DTL_MISSING) && |
428870ff | 5395 | vdev_dtl_empty(newvd, DTL_OUTAGE) && |
fb5f0bc8 | 5396 | !vdev_dtl_required(oldvd)) |
34dc7c2f | 5397 | return (oldvd); |
34dc7c2f BB |
5398 | } |
5399 | ||
5400 | /* | |
5401 | * Check for a completed resilver with the 'unspare' flag set. | |
5402 | */ | |
572e2857 BB |
5403 | if (vd->vdev_ops == &vdev_spare_ops) { |
5404 | vdev_t *first = vd->vdev_child[0]; | |
5405 | vdev_t *last = vd->vdev_child[vd->vdev_children - 1]; | |
5406 | ||
5407 | if (last->vdev_unspare) { | |
5408 | oldvd = first; | |
5409 | newvd = last; | |
5410 | } else if (first->vdev_unspare) { | |
5411 | oldvd = last; | |
5412 | newvd = first; | |
5413 | } else { | |
5414 | oldvd = NULL; | |
5415 | } | |
34dc7c2f | 5416 | |
572e2857 | 5417 | if (oldvd != NULL && |
fb5f0bc8 | 5418 | vdev_dtl_empty(newvd, DTL_MISSING) && |
428870ff | 5419 | vdev_dtl_empty(newvd, DTL_OUTAGE) && |
572e2857 | 5420 | !vdev_dtl_required(oldvd)) |
34dc7c2f | 5421 | return (oldvd); |
572e2857 BB |
5422 | |
5423 | /* | |
5424 | * If there are more than two spares attached to a disk, | |
5425 | * and those spares are not required, then we want to | |
5426 | * attempt to free them up now so that they can be used | |
5427 | * by other pools. Once we're back down to a single | |
5428 | * disk+spare, we stop removing them. | |
5429 | */ | |
5430 | if (vd->vdev_children > 2) { | |
5431 | newvd = vd->vdev_child[1]; | |
5432 | ||
5433 | if (newvd->vdev_isspare && last->vdev_isspare && | |
5434 | vdev_dtl_empty(last, DTL_MISSING) && | |
5435 | vdev_dtl_empty(last, DTL_OUTAGE) && | |
5436 | !vdev_dtl_required(newvd)) | |
5437 | return (newvd); | |
34dc7c2f | 5438 | } |
34dc7c2f BB |
5439 | } |
5440 | ||
5441 | return (NULL); | |
5442 | } | |
5443 | ||
5444 | static void | |
5445 | spa_vdev_resilver_done(spa_t *spa) | |
5446 | { | |
fb5f0bc8 BB |
5447 | vdev_t *vd, *pvd, *ppvd; |
5448 | uint64_t guid, sguid, pguid, ppguid; | |
34dc7c2f | 5449 | |
fb5f0bc8 | 5450 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f BB |
5451 | |
5452 | while ((vd = spa_vdev_resilver_done_hunt(spa->spa_root_vdev)) != NULL) { | |
fb5f0bc8 BB |
5453 | pvd = vd->vdev_parent; |
5454 | ppvd = pvd->vdev_parent; | |
34dc7c2f | 5455 | guid = vd->vdev_guid; |
fb5f0bc8 BB |
5456 | pguid = pvd->vdev_guid; |
5457 | ppguid = ppvd->vdev_guid; | |
5458 | sguid = 0; | |
34dc7c2f BB |
5459 | /* |
5460 | * If we have just finished replacing a hot spared device, then | |
5461 | * we need to detach the parent's first child (the original hot | |
5462 | * spare) as well. | |
5463 | */ | |
572e2857 BB |
5464 | if (ppvd->vdev_ops == &vdev_spare_ops && pvd->vdev_id == 0 && |
5465 | ppvd->vdev_children == 2) { | |
34dc7c2f | 5466 | ASSERT(pvd->vdev_ops == &vdev_replacing_ops); |
fb5f0bc8 | 5467 | sguid = ppvd->vdev_child[1]->vdev_guid; |
34dc7c2f | 5468 | } |
fb5f0bc8 BB |
5469 | spa_config_exit(spa, SCL_ALL, FTAG); |
5470 | if (spa_vdev_detach(spa, guid, pguid, B_TRUE) != 0) | |
34dc7c2f | 5471 | return; |
fb5f0bc8 | 5472 | if (sguid && spa_vdev_detach(spa, sguid, ppguid, B_TRUE) != 0) |
34dc7c2f | 5473 | return; |
fb5f0bc8 | 5474 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f BB |
5475 | } |
5476 | ||
fb5f0bc8 | 5477 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
5478 | } |
5479 | ||
5480 | /* | |
428870ff | 5481 | * Update the stored path or FRU for this vdev. |
34dc7c2f BB |
5482 | */ |
5483 | int | |
9babb374 BB |
5484 | spa_vdev_set_common(spa_t *spa, uint64_t guid, const char *value, |
5485 | boolean_t ispath) | |
34dc7c2f | 5486 | { |
b128c09f | 5487 | vdev_t *vd; |
428870ff | 5488 | boolean_t sync = B_FALSE; |
34dc7c2f | 5489 | |
572e2857 BB |
5490 | ASSERT(spa_writeable(spa)); |
5491 | ||
428870ff | 5492 | spa_vdev_state_enter(spa, SCL_ALL); |
34dc7c2f | 5493 | |
9babb374 | 5494 | if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL) |
428870ff | 5495 | return (spa_vdev_state_exit(spa, NULL, ENOENT)); |
34dc7c2f BB |
5496 | |
5497 | if (!vd->vdev_ops->vdev_op_leaf) | |
428870ff | 5498 | return (spa_vdev_state_exit(spa, NULL, ENOTSUP)); |
34dc7c2f | 5499 | |
9babb374 | 5500 | if (ispath) { |
428870ff BB |
5501 | if (strcmp(value, vd->vdev_path) != 0) { |
5502 | spa_strfree(vd->vdev_path); | |
5503 | vd->vdev_path = spa_strdup(value); | |
5504 | sync = B_TRUE; | |
5505 | } | |
9babb374 | 5506 | } else { |
428870ff BB |
5507 | if (vd->vdev_fru == NULL) { |
5508 | vd->vdev_fru = spa_strdup(value); | |
5509 | sync = B_TRUE; | |
5510 | } else if (strcmp(value, vd->vdev_fru) != 0) { | |
9babb374 | 5511 | spa_strfree(vd->vdev_fru); |
428870ff BB |
5512 | vd->vdev_fru = spa_strdup(value); |
5513 | sync = B_TRUE; | |
5514 | } | |
9babb374 | 5515 | } |
34dc7c2f | 5516 | |
428870ff | 5517 | return (spa_vdev_state_exit(spa, sync ? vd : NULL, 0)); |
34dc7c2f BB |
5518 | } |
5519 | ||
9babb374 BB |
5520 | int |
5521 | spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath) | |
5522 | { | |
5523 | return (spa_vdev_set_common(spa, guid, newpath, B_TRUE)); | |
5524 | } | |
5525 | ||
5526 | int | |
5527 | spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru) | |
5528 | { | |
5529 | return (spa_vdev_set_common(spa, guid, newfru, B_FALSE)); | |
5530 | } | |
5531 | ||
34dc7c2f BB |
5532 | /* |
5533 | * ========================================================================== | |
428870ff | 5534 | * SPA Scanning |
34dc7c2f BB |
5535 | * ========================================================================== |
5536 | */ | |
5537 | ||
34dc7c2f | 5538 | int |
428870ff BB |
5539 | spa_scan_stop(spa_t *spa) |
5540 | { | |
5541 | ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); | |
5542 | if (dsl_scan_resilvering(spa->spa_dsl_pool)) | |
2e528b49 | 5543 | return (SET_ERROR(EBUSY)); |
428870ff BB |
5544 | return (dsl_scan_cancel(spa->spa_dsl_pool)); |
5545 | } | |
5546 | ||
5547 | int | |
5548 | spa_scan(spa_t *spa, pool_scan_func_t func) | |
34dc7c2f | 5549 | { |
b128c09f | 5550 | ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); |
34dc7c2f | 5551 | |
428870ff | 5552 | if (func >= POOL_SCAN_FUNCS || func == POOL_SCAN_NONE) |
2e528b49 | 5553 | return (SET_ERROR(ENOTSUP)); |
34dc7c2f | 5554 | |
34dc7c2f | 5555 | /* |
b128c09f BB |
5556 | * If a resilver was requested, but there is no DTL on a |
5557 | * writeable leaf device, we have nothing to do. | |
34dc7c2f | 5558 | */ |
428870ff | 5559 | if (func == POOL_SCAN_RESILVER && |
b128c09f BB |
5560 | !vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL)) { |
5561 | spa_async_request(spa, SPA_ASYNC_RESILVER_DONE); | |
34dc7c2f BB |
5562 | return (0); |
5563 | } | |
5564 | ||
428870ff | 5565 | return (dsl_scan(spa->spa_dsl_pool, func)); |
34dc7c2f BB |
5566 | } |
5567 | ||
5568 | /* | |
5569 | * ========================================================================== | |
5570 | * SPA async task processing | |
5571 | * ========================================================================== | |
5572 | */ | |
5573 | ||
5574 | static void | |
5575 | spa_async_remove(spa_t *spa, vdev_t *vd) | |
5576 | { | |
d6320ddb BB |
5577 | int c; |
5578 | ||
b128c09f | 5579 | if (vd->vdev_remove_wanted) { |
428870ff BB |
5580 | vd->vdev_remove_wanted = B_FALSE; |
5581 | vd->vdev_delayed_close = B_FALSE; | |
b128c09f | 5582 | vdev_set_state(vd, B_FALSE, VDEV_STATE_REMOVED, VDEV_AUX_NONE); |
428870ff BB |
5583 | |
5584 | /* | |
5585 | * We want to clear the stats, but we don't want to do a full | |
5586 | * vdev_clear() as that will cause us to throw away | |
5587 | * degraded/faulted state as well as attempt to reopen the | |
5588 | * device, all of which is a waste. | |
5589 | */ | |
5590 | vd->vdev_stat.vs_read_errors = 0; | |
5591 | vd->vdev_stat.vs_write_errors = 0; | |
5592 | vd->vdev_stat.vs_checksum_errors = 0; | |
5593 | ||
b128c09f BB |
5594 | vdev_state_dirty(vd->vdev_top); |
5595 | } | |
34dc7c2f | 5596 | |
d6320ddb | 5597 | for (c = 0; c < vd->vdev_children; c++) |
b128c09f BB |
5598 | spa_async_remove(spa, vd->vdev_child[c]); |
5599 | } | |
5600 | ||
5601 | static void | |
5602 | spa_async_probe(spa_t *spa, vdev_t *vd) | |
5603 | { | |
d6320ddb BB |
5604 | int c; |
5605 | ||
b128c09f | 5606 | if (vd->vdev_probe_wanted) { |
428870ff | 5607 | vd->vdev_probe_wanted = B_FALSE; |
b128c09f | 5608 | vdev_reopen(vd); /* vdev_open() does the actual probe */ |
34dc7c2f | 5609 | } |
b128c09f | 5610 | |
d6320ddb | 5611 | for (c = 0; c < vd->vdev_children; c++) |
b128c09f | 5612 | spa_async_probe(spa, vd->vdev_child[c]); |
34dc7c2f BB |
5613 | } |
5614 | ||
9babb374 BB |
5615 | static void |
5616 | spa_async_autoexpand(spa_t *spa, vdev_t *vd) | |
5617 | { | |
d6320ddb | 5618 | int c; |
9babb374 BB |
5619 | |
5620 | if (!spa->spa_autoexpand) | |
5621 | return; | |
5622 | ||
d6320ddb | 5623 | for (c = 0; c < vd->vdev_children; c++) { |
9babb374 BB |
5624 | vdev_t *cvd = vd->vdev_child[c]; |
5625 | spa_async_autoexpand(spa, cvd); | |
5626 | } | |
5627 | ||
5628 | if (!vd->vdev_ops->vdev_op_leaf || vd->vdev_physpath == NULL) | |
5629 | return; | |
5630 | ||
26685276 | 5631 | spa_event_notify(vd->vdev_spa, vd, FM_EREPORT_ZFS_DEVICE_AUTOEXPAND); |
9babb374 BB |
5632 | } |
5633 | ||
34dc7c2f BB |
5634 | static void |
5635 | spa_async_thread(spa_t *spa) | |
5636 | { | |
d6320ddb | 5637 | int tasks, i; |
34dc7c2f BB |
5638 | |
5639 | ASSERT(spa->spa_sync_on); | |
5640 | ||
5641 | mutex_enter(&spa->spa_async_lock); | |
5642 | tasks = spa->spa_async_tasks; | |
5643 | spa->spa_async_tasks = 0; | |
5644 | mutex_exit(&spa->spa_async_lock); | |
5645 | ||
5646 | /* | |
5647 | * See if the config needs to be updated. | |
5648 | */ | |
5649 | if (tasks & SPA_ASYNC_CONFIG_UPDATE) { | |
428870ff | 5650 | uint64_t old_space, new_space; |
9babb374 | 5651 | |
34dc7c2f | 5652 | mutex_enter(&spa_namespace_lock); |
428870ff | 5653 | old_space = metaslab_class_get_space(spa_normal_class(spa)); |
34dc7c2f | 5654 | spa_config_update(spa, SPA_CONFIG_UPDATE_POOL); |
428870ff | 5655 | new_space = metaslab_class_get_space(spa_normal_class(spa)); |
34dc7c2f | 5656 | mutex_exit(&spa_namespace_lock); |
9babb374 BB |
5657 | |
5658 | /* | |
5659 | * If the pool grew as a result of the config update, | |
5660 | * then log an internal history event. | |
5661 | */ | |
428870ff | 5662 | if (new_space != old_space) { |
6f1ffb06 | 5663 | spa_history_log_internal(spa, "vdev online", NULL, |
45d1cae3 | 5664 | "pool '%s' size: %llu(+%llu)", |
428870ff | 5665 | spa_name(spa), new_space, new_space - old_space); |
9babb374 | 5666 | } |
34dc7c2f BB |
5667 | } |
5668 | ||
5669 | /* | |
5670 | * See if any devices need to be marked REMOVED. | |
34dc7c2f | 5671 | */ |
b128c09f | 5672 | if (tasks & SPA_ASYNC_REMOVE) { |
428870ff | 5673 | spa_vdev_state_enter(spa, SCL_NONE); |
34dc7c2f | 5674 | spa_async_remove(spa, spa->spa_root_vdev); |
d6320ddb | 5675 | for (i = 0; i < spa->spa_l2cache.sav_count; i++) |
b128c09f | 5676 | spa_async_remove(spa, spa->spa_l2cache.sav_vdevs[i]); |
d6320ddb | 5677 | for (i = 0; i < spa->spa_spares.sav_count; i++) |
b128c09f BB |
5678 | spa_async_remove(spa, spa->spa_spares.sav_vdevs[i]); |
5679 | (void) spa_vdev_state_exit(spa, NULL, 0); | |
34dc7c2f BB |
5680 | } |
5681 | ||
9babb374 BB |
5682 | if ((tasks & SPA_ASYNC_AUTOEXPAND) && !spa_suspended(spa)) { |
5683 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); | |
5684 | spa_async_autoexpand(spa, spa->spa_root_vdev); | |
5685 | spa_config_exit(spa, SCL_CONFIG, FTAG); | |
5686 | } | |
5687 | ||
34dc7c2f | 5688 | /* |
b128c09f | 5689 | * See if any devices need to be probed. |
34dc7c2f | 5690 | */ |
b128c09f | 5691 | if (tasks & SPA_ASYNC_PROBE) { |
428870ff | 5692 | spa_vdev_state_enter(spa, SCL_NONE); |
b128c09f BB |
5693 | spa_async_probe(spa, spa->spa_root_vdev); |
5694 | (void) spa_vdev_state_exit(spa, NULL, 0); | |
5695 | } | |
34dc7c2f BB |
5696 | |
5697 | /* | |
b128c09f | 5698 | * If any devices are done replacing, detach them. |
34dc7c2f | 5699 | */ |
b128c09f BB |
5700 | if (tasks & SPA_ASYNC_RESILVER_DONE) |
5701 | spa_vdev_resilver_done(spa); | |
34dc7c2f BB |
5702 | |
5703 | /* | |
5704 | * Kick off a resilver. | |
5705 | */ | |
b128c09f | 5706 | if (tasks & SPA_ASYNC_RESILVER) |
428870ff | 5707 | dsl_resilver_restart(spa->spa_dsl_pool, 0); |
34dc7c2f BB |
5708 | |
5709 | /* | |
5710 | * Let the world know that we're done. | |
5711 | */ | |
5712 | mutex_enter(&spa->spa_async_lock); | |
5713 | spa->spa_async_thread = NULL; | |
5714 | cv_broadcast(&spa->spa_async_cv); | |
5715 | mutex_exit(&spa->spa_async_lock); | |
5716 | thread_exit(); | |
5717 | } | |
5718 | ||
5719 | void | |
5720 | spa_async_suspend(spa_t *spa) | |
5721 | { | |
5722 | mutex_enter(&spa->spa_async_lock); | |
5723 | spa->spa_async_suspended++; | |
5724 | while (spa->spa_async_thread != NULL) | |
5725 | cv_wait(&spa->spa_async_cv, &spa->spa_async_lock); | |
5726 | mutex_exit(&spa->spa_async_lock); | |
5727 | } | |
5728 | ||
5729 | void | |
5730 | spa_async_resume(spa_t *spa) | |
5731 | { | |
5732 | mutex_enter(&spa->spa_async_lock); | |
5733 | ASSERT(spa->spa_async_suspended != 0); | |
5734 | spa->spa_async_suspended--; | |
5735 | mutex_exit(&spa->spa_async_lock); | |
5736 | } | |
5737 | ||
5738 | static void | |
5739 | spa_async_dispatch(spa_t *spa) | |
5740 | { | |
5741 | mutex_enter(&spa->spa_async_lock); | |
5742 | if (spa->spa_async_tasks && !spa->spa_async_suspended && | |
5743 | spa->spa_async_thread == NULL && | |
5744 | rootdir != NULL && !vn_is_readonly(rootdir)) | |
5745 | spa->spa_async_thread = thread_create(NULL, 0, | |
5746 | spa_async_thread, spa, 0, &p0, TS_RUN, maxclsyspri); | |
5747 | mutex_exit(&spa->spa_async_lock); | |
5748 | } | |
5749 | ||
5750 | void | |
5751 | spa_async_request(spa_t *spa, int task) | |
5752 | { | |
428870ff | 5753 | zfs_dbgmsg("spa=%s async request task=%u", spa->spa_name, task); |
34dc7c2f BB |
5754 | mutex_enter(&spa->spa_async_lock); |
5755 | spa->spa_async_tasks |= task; | |
5756 | mutex_exit(&spa->spa_async_lock); | |
5757 | } | |
5758 | ||
5759 | /* | |
5760 | * ========================================================================== | |
5761 | * SPA syncing routines | |
5762 | * ========================================================================== | |
5763 | */ | |
5764 | ||
428870ff BB |
5765 | static int |
5766 | bpobj_enqueue_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx) | |
34dc7c2f | 5767 | { |
428870ff BB |
5768 | bpobj_t *bpo = arg; |
5769 | bpobj_enqueue(bpo, bp, tx); | |
5770 | return (0); | |
5771 | } | |
34dc7c2f | 5772 | |
428870ff BB |
5773 | static int |
5774 | spa_free_sync_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx) | |
5775 | { | |
5776 | zio_t *zio = arg; | |
34dc7c2f | 5777 | |
428870ff BB |
5778 | zio_nowait(zio_free_sync(zio, zio->io_spa, dmu_tx_get_txg(tx), bp, |
5779 | zio->io_flags)); | |
5780 | return (0); | |
34dc7c2f BB |
5781 | } |
5782 | ||
5783 | static void | |
5784 | spa_sync_nvlist(spa_t *spa, uint64_t obj, nvlist_t *nv, dmu_tx_t *tx) | |
5785 | { | |
5786 | char *packed = NULL; | |
b128c09f | 5787 | size_t bufsize; |
34dc7c2f BB |
5788 | size_t nvsize = 0; |
5789 | dmu_buf_t *db; | |
5790 | ||
5791 | VERIFY(nvlist_size(nv, &nvsize, NV_ENCODE_XDR) == 0); | |
5792 | ||
b128c09f BB |
5793 | /* |
5794 | * Write full (SPA_CONFIG_BLOCKSIZE) blocks of configuration | |
5795 | * information. This avoids the dbuf_will_dirty() path and | |
5796 | * saves us a pre-read to get data we don't actually care about. | |
5797 | */ | |
9ae529ec | 5798 | bufsize = P2ROUNDUP((uint64_t)nvsize, SPA_CONFIG_BLOCKSIZE); |
b8d06fca | 5799 | packed = vmem_alloc(bufsize, KM_PUSHPAGE); |
34dc7c2f BB |
5800 | |
5801 | VERIFY(nvlist_pack(nv, &packed, &nvsize, NV_ENCODE_XDR, | |
b8d06fca | 5802 | KM_PUSHPAGE) == 0); |
b128c09f | 5803 | bzero(packed + nvsize, bufsize - nvsize); |
34dc7c2f | 5804 | |
b128c09f | 5805 | dmu_write(spa->spa_meta_objset, obj, 0, bufsize, packed, tx); |
34dc7c2f | 5806 | |
00b46022 | 5807 | vmem_free(packed, bufsize); |
34dc7c2f BB |
5808 | |
5809 | VERIFY(0 == dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db)); | |
5810 | dmu_buf_will_dirty(db, tx); | |
5811 | *(uint64_t *)db->db_data = nvsize; | |
5812 | dmu_buf_rele(db, FTAG); | |
5813 | } | |
5814 | ||
5815 | static void | |
5816 | spa_sync_aux_dev(spa_t *spa, spa_aux_vdev_t *sav, dmu_tx_t *tx, | |
5817 | const char *config, const char *entry) | |
5818 | { | |
5819 | nvlist_t *nvroot; | |
5820 | nvlist_t **list; | |
5821 | int i; | |
5822 | ||
5823 | if (!sav->sav_sync) | |
5824 | return; | |
5825 | ||
5826 | /* | |
5827 | * Update the MOS nvlist describing the list of available devices. | |
5828 | * spa_validate_aux() will have already made sure this nvlist is | |
5829 | * valid and the vdevs are labeled appropriately. | |
5830 | */ | |
5831 | if (sav->sav_object == 0) { | |
5832 | sav->sav_object = dmu_object_alloc(spa->spa_meta_objset, | |
5833 | DMU_OT_PACKED_NVLIST, 1 << 14, DMU_OT_PACKED_NVLIST_SIZE, | |
5834 | sizeof (uint64_t), tx); | |
5835 | VERIFY(zap_update(spa->spa_meta_objset, | |
5836 | DMU_POOL_DIRECTORY_OBJECT, entry, sizeof (uint64_t), 1, | |
5837 | &sav->sav_object, tx) == 0); | |
5838 | } | |
5839 | ||
b8d06fca | 5840 | VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_PUSHPAGE) == 0); |
34dc7c2f BB |
5841 | if (sav->sav_count == 0) { |
5842 | VERIFY(nvlist_add_nvlist_array(nvroot, config, NULL, 0) == 0); | |
5843 | } else { | |
b8d06fca | 5844 | list = kmem_alloc(sav->sav_count * sizeof (void *), KM_PUSHPAGE); |
34dc7c2f BB |
5845 | for (i = 0; i < sav->sav_count; i++) |
5846 | list[i] = vdev_config_generate(spa, sav->sav_vdevs[i], | |
428870ff | 5847 | B_FALSE, VDEV_CONFIG_L2CACHE); |
34dc7c2f BB |
5848 | VERIFY(nvlist_add_nvlist_array(nvroot, config, list, |
5849 | sav->sav_count) == 0); | |
5850 | for (i = 0; i < sav->sav_count; i++) | |
5851 | nvlist_free(list[i]); | |
5852 | kmem_free(list, sav->sav_count * sizeof (void *)); | |
5853 | } | |
5854 | ||
5855 | spa_sync_nvlist(spa, sav->sav_object, nvroot, tx); | |
5856 | nvlist_free(nvroot); | |
5857 | ||
5858 | sav->sav_sync = B_FALSE; | |
5859 | } | |
5860 | ||
5861 | static void | |
5862 | spa_sync_config_object(spa_t *spa, dmu_tx_t *tx) | |
5863 | { | |
5864 | nvlist_t *config; | |
5865 | ||
b128c09f | 5866 | if (list_is_empty(&spa->spa_config_dirty_list)) |
34dc7c2f BB |
5867 | return; |
5868 | ||
b128c09f BB |
5869 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); |
5870 | ||
5871 | config = spa_config_generate(spa, spa->spa_root_vdev, | |
5872 | dmu_tx_get_txg(tx), B_FALSE); | |
5873 | ||
ea0b2538 GW |
5874 | /* |
5875 | * If we're upgrading the spa version then make sure that | |
5876 | * the config object gets updated with the correct version. | |
5877 | */ | |
5878 | if (spa->spa_ubsync.ub_version < spa->spa_uberblock.ub_version) | |
5879 | fnvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, | |
5880 | spa->spa_uberblock.ub_version); | |
5881 | ||
b128c09f | 5882 | spa_config_exit(spa, SCL_STATE, FTAG); |
34dc7c2f BB |
5883 | |
5884 | if (spa->spa_config_syncing) | |
5885 | nvlist_free(spa->spa_config_syncing); | |
5886 | spa->spa_config_syncing = config; | |
5887 | ||
5888 | spa_sync_nvlist(spa, spa->spa_config_object, config, tx); | |
5889 | } | |
5890 | ||
9ae529ec | 5891 | static void |
13fe0198 | 5892 | spa_sync_version(void *arg, dmu_tx_t *tx) |
9ae529ec | 5893 | { |
13fe0198 MA |
5894 | uint64_t *versionp = arg; |
5895 | uint64_t version = *versionp; | |
5896 | spa_t *spa = dmu_tx_pool(tx)->dp_spa; | |
9ae529ec CS |
5897 | |
5898 | /* | |
5899 | * Setting the version is special cased when first creating the pool. | |
5900 | */ | |
5901 | ASSERT(tx->tx_txg != TXG_INITIAL); | |
5902 | ||
8dca0a9a | 5903 | ASSERT(SPA_VERSION_IS_SUPPORTED(version)); |
9ae529ec CS |
5904 | ASSERT(version >= spa_version(spa)); |
5905 | ||
5906 | spa->spa_uberblock.ub_version = version; | |
5907 | vdev_config_dirty(spa->spa_root_vdev); | |
6f1ffb06 | 5908 | spa_history_log_internal(spa, "set", tx, "version=%lld", version); |
9ae529ec CS |
5909 | } |
5910 | ||
34dc7c2f BB |
5911 | /* |
5912 | * Set zpool properties. | |
5913 | */ | |
5914 | static void | |
13fe0198 | 5915 | spa_sync_props(void *arg, dmu_tx_t *tx) |
34dc7c2f | 5916 | { |
13fe0198 MA |
5917 | nvlist_t *nvp = arg; |
5918 | spa_t *spa = dmu_tx_pool(tx)->dp_spa; | |
34dc7c2f | 5919 | objset_t *mos = spa->spa_meta_objset; |
9ae529ec | 5920 | nvpair_t *elem = NULL; |
b128c09f BB |
5921 | |
5922 | mutex_enter(&spa->spa_props_lock); | |
34dc7c2f | 5923 | |
34dc7c2f | 5924 | while ((elem = nvlist_next_nvpair(nvp, elem))) { |
9ae529ec CS |
5925 | uint64_t intval; |
5926 | char *strval, *fname; | |
5927 | zpool_prop_t prop; | |
5928 | const char *propname; | |
5929 | zprop_type_t proptype; | |
5930 | zfeature_info_t *feature; | |
5931 | ||
5932 | prop = zpool_name_to_prop(nvpair_name(elem)); | |
5933 | switch ((int)prop) { | |
5934 | case ZPROP_INVAL: | |
5935 | /* | |
5936 | * We checked this earlier in spa_prop_validate(). | |
5937 | */ | |
5938 | ASSERT(zpool_prop_feature(nvpair_name(elem))); | |
5939 | ||
5940 | fname = strchr(nvpair_name(elem), '@') + 1; | |
5941 | VERIFY3U(0, ==, zfeature_lookup_name(fname, &feature)); | |
5942 | ||
5943 | spa_feature_enable(spa, feature, tx); | |
6f1ffb06 MA |
5944 | spa_history_log_internal(spa, "set", tx, |
5945 | "%s=enabled", nvpair_name(elem)); | |
9ae529ec CS |
5946 | break; |
5947 | ||
34dc7c2f | 5948 | case ZPOOL_PROP_VERSION: |
9ae529ec | 5949 | VERIFY(nvpair_value_uint64(elem, &intval) == 0); |
34dc7c2f | 5950 | /* |
9ae529ec CS |
5951 | * The version is synced seperatly before other |
5952 | * properties and should be correct by now. | |
34dc7c2f | 5953 | */ |
9ae529ec | 5954 | ASSERT3U(spa_version(spa), >=, intval); |
34dc7c2f BB |
5955 | break; |
5956 | ||
5957 | case ZPOOL_PROP_ALTROOT: | |
5958 | /* | |
5959 | * 'altroot' is a non-persistent property. It should | |
5960 | * have been set temporarily at creation or import time. | |
5961 | */ | |
5962 | ASSERT(spa->spa_root != NULL); | |
5963 | break; | |
5964 | ||
572e2857 | 5965 | case ZPOOL_PROP_READONLY: |
34dc7c2f BB |
5966 | case ZPOOL_PROP_CACHEFILE: |
5967 | /* | |
572e2857 BB |
5968 | * 'readonly' and 'cachefile' are also non-persisitent |
5969 | * properties. | |
34dc7c2f | 5970 | */ |
34dc7c2f | 5971 | break; |
d96eb2b1 DM |
5972 | case ZPOOL_PROP_COMMENT: |
5973 | VERIFY(nvpair_value_string(elem, &strval) == 0); | |
5974 | if (spa->spa_comment != NULL) | |
5975 | spa_strfree(spa->spa_comment); | |
5976 | spa->spa_comment = spa_strdup(strval); | |
5977 | /* | |
5978 | * We need to dirty the configuration on all the vdevs | |
5979 | * so that their labels get updated. It's unnecessary | |
5980 | * to do this for pool creation since the vdev's | |
5981 | * configuratoin has already been dirtied. | |
5982 | */ | |
5983 | if (tx->tx_txg != TXG_INITIAL) | |
5984 | vdev_config_dirty(spa->spa_root_vdev); | |
6f1ffb06 MA |
5985 | spa_history_log_internal(spa, "set", tx, |
5986 | "%s=%s", nvpair_name(elem), strval); | |
d96eb2b1 | 5987 | break; |
34dc7c2f BB |
5988 | default: |
5989 | /* | |
5990 | * Set pool property values in the poolprops mos object. | |
5991 | */ | |
34dc7c2f | 5992 | if (spa->spa_pool_props_object == 0) { |
9ae529ec CS |
5993 | spa->spa_pool_props_object = |
5994 | zap_create_link(mos, DMU_OT_POOL_PROPS, | |
34dc7c2f | 5995 | DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_PROPS, |
9ae529ec | 5996 | tx); |
34dc7c2f | 5997 | } |
34dc7c2f BB |
5998 | |
5999 | /* normalize the property name */ | |
6000 | propname = zpool_prop_to_name(prop); | |
6001 | proptype = zpool_prop_get_type(prop); | |
6002 | ||
6003 | if (nvpair_type(elem) == DATA_TYPE_STRING) { | |
6004 | ASSERT(proptype == PROP_TYPE_STRING); | |
6005 | VERIFY(nvpair_value_string(elem, &strval) == 0); | |
6006 | VERIFY(zap_update(mos, | |
6007 | spa->spa_pool_props_object, propname, | |
6008 | 1, strlen(strval) + 1, strval, tx) == 0); | |
6f1ffb06 MA |
6009 | spa_history_log_internal(spa, "set", tx, |
6010 | "%s=%s", nvpair_name(elem), strval); | |
34dc7c2f BB |
6011 | } else if (nvpair_type(elem) == DATA_TYPE_UINT64) { |
6012 | VERIFY(nvpair_value_uint64(elem, &intval) == 0); | |
6013 | ||
6014 | if (proptype == PROP_TYPE_INDEX) { | |
6015 | const char *unused; | |
6016 | VERIFY(zpool_prop_index_to_string( | |
6017 | prop, intval, &unused) == 0); | |
6018 | } | |
6019 | VERIFY(zap_update(mos, | |
6020 | spa->spa_pool_props_object, propname, | |
6021 | 8, 1, &intval, tx) == 0); | |
6f1ffb06 MA |
6022 | spa_history_log_internal(spa, "set", tx, |
6023 | "%s=%lld", nvpair_name(elem), intval); | |
34dc7c2f BB |
6024 | } else { |
6025 | ASSERT(0); /* not allowed */ | |
6026 | } | |
6027 | ||
6028 | switch (prop) { | |
6029 | case ZPOOL_PROP_DELEGATION: | |
6030 | spa->spa_delegation = intval; | |
6031 | break; | |
6032 | case ZPOOL_PROP_BOOTFS: | |
6033 | spa->spa_bootfs = intval; | |
6034 | break; | |
6035 | case ZPOOL_PROP_FAILUREMODE: | |
6036 | spa->spa_failmode = intval; | |
6037 | break; | |
9babb374 BB |
6038 | case ZPOOL_PROP_AUTOEXPAND: |
6039 | spa->spa_autoexpand = intval; | |
428870ff BB |
6040 | if (tx->tx_txg != TXG_INITIAL) |
6041 | spa_async_request(spa, | |
6042 | SPA_ASYNC_AUTOEXPAND); | |
6043 | break; | |
6044 | case ZPOOL_PROP_DEDUPDITTO: | |
6045 | spa->spa_dedup_ditto = intval; | |
9babb374 | 6046 | break; |
34dc7c2f BB |
6047 | default: |
6048 | break; | |
6049 | } | |
6050 | } | |
6051 | ||
34dc7c2f | 6052 | } |
b128c09f BB |
6053 | |
6054 | mutex_exit(&spa->spa_props_lock); | |
34dc7c2f BB |
6055 | } |
6056 | ||
428870ff BB |
6057 | /* |
6058 | * Perform one-time upgrade on-disk changes. spa_version() does not | |
6059 | * reflect the new version this txg, so there must be no changes this | |
6060 | * txg to anything that the upgrade code depends on after it executes. | |
6061 | * Therefore this must be called after dsl_pool_sync() does the sync | |
6062 | * tasks. | |
6063 | */ | |
6064 | static void | |
6065 | spa_sync_upgrades(spa_t *spa, dmu_tx_t *tx) | |
6066 | { | |
6067 | dsl_pool_t *dp = spa->spa_dsl_pool; | |
6068 | ||
6069 | ASSERT(spa->spa_sync_pass == 1); | |
6070 | ||
13fe0198 MA |
6071 | rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG); |
6072 | ||
428870ff BB |
6073 | if (spa->spa_ubsync.ub_version < SPA_VERSION_ORIGIN && |
6074 | spa->spa_uberblock.ub_version >= SPA_VERSION_ORIGIN) { | |
6075 | dsl_pool_create_origin(dp, tx); | |
6076 | ||
6077 | /* Keeping the origin open increases spa_minref */ | |
6078 | spa->spa_minref += 3; | |
6079 | } | |
6080 | ||
6081 | if (spa->spa_ubsync.ub_version < SPA_VERSION_NEXT_CLONES && | |
6082 | spa->spa_uberblock.ub_version >= SPA_VERSION_NEXT_CLONES) { | |
6083 | dsl_pool_upgrade_clones(dp, tx); | |
6084 | } | |
6085 | ||
6086 | if (spa->spa_ubsync.ub_version < SPA_VERSION_DIR_CLONES && | |
6087 | spa->spa_uberblock.ub_version >= SPA_VERSION_DIR_CLONES) { | |
6088 | dsl_pool_upgrade_dir_clones(dp, tx); | |
6089 | ||
6090 | /* Keeping the freedir open increases spa_minref */ | |
6091 | spa->spa_minref += 3; | |
6092 | } | |
9ae529ec CS |
6093 | |
6094 | if (spa->spa_ubsync.ub_version < SPA_VERSION_FEATURES && | |
6095 | spa->spa_uberblock.ub_version >= SPA_VERSION_FEATURES) { | |
6096 | spa_feature_create_zap_objects(spa, tx); | |
6097 | } | |
13fe0198 | 6098 | rrw_exit(&dp->dp_config_rwlock, FTAG); |
428870ff BB |
6099 | } |
6100 | ||
34dc7c2f BB |
6101 | /* |
6102 | * Sync the specified transaction group. New blocks may be dirtied as | |
6103 | * part of the process, so we iterate until it converges. | |
6104 | */ | |
6105 | void | |
6106 | spa_sync(spa_t *spa, uint64_t txg) | |
6107 | { | |
6108 | dsl_pool_t *dp = spa->spa_dsl_pool; | |
6109 | objset_t *mos = spa->spa_meta_objset; | |
428870ff BB |
6110 | bpobj_t *defer_bpo = &spa->spa_deferred_bpobj; |
6111 | bplist_t *free_bpl = &spa->spa_free_bplist[txg & TXG_MASK]; | |
34dc7c2f BB |
6112 | vdev_t *rvd = spa->spa_root_vdev; |
6113 | vdev_t *vd; | |
34dc7c2f | 6114 | dmu_tx_t *tx; |
b128c09f | 6115 | int error; |
d6320ddb | 6116 | int c; |
34dc7c2f | 6117 | |
572e2857 BB |
6118 | VERIFY(spa_writeable(spa)); |
6119 | ||
34dc7c2f BB |
6120 | /* |
6121 | * Lock out configuration changes. | |
6122 | */ | |
b128c09f | 6123 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); |
34dc7c2f BB |
6124 | |
6125 | spa->spa_syncing_txg = txg; | |
6126 | spa->spa_sync_pass = 0; | |
6127 | ||
b128c09f BB |
6128 | /* |
6129 | * If there are any pending vdev state changes, convert them | |
6130 | * into config changes that go out with this transaction group. | |
6131 | */ | |
6132 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); | |
fb5f0bc8 BB |
6133 | while (list_head(&spa->spa_state_dirty_list) != NULL) { |
6134 | /* | |
6135 | * We need the write lock here because, for aux vdevs, | |
6136 | * calling vdev_config_dirty() modifies sav_config. | |
6137 | * This is ugly and will become unnecessary when we | |
6138 | * eliminate the aux vdev wart by integrating all vdevs | |
6139 | * into the root vdev tree. | |
6140 | */ | |
6141 | spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); | |
6142 | spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_WRITER); | |
6143 | while ((vd = list_head(&spa->spa_state_dirty_list)) != NULL) { | |
6144 | vdev_state_clean(vd); | |
6145 | vdev_config_dirty(vd); | |
6146 | } | |
6147 | spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); | |
6148 | spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER); | |
b128c09f BB |
6149 | } |
6150 | spa_config_exit(spa, SCL_STATE, FTAG); | |
6151 | ||
34dc7c2f BB |
6152 | tx = dmu_tx_create_assigned(dp, txg); |
6153 | ||
cc92e9d0 GW |
6154 | spa->spa_sync_starttime = gethrtime(); |
6155 | taskq_cancel_id(system_taskq, spa->spa_deadman_tqid); | |
6156 | spa->spa_deadman_tqid = taskq_dispatch_delay(system_taskq, | |
cbfa294d | 6157 | spa_deadman, spa, TQ_PUSHPAGE, ddi_get_lbolt() + |
cc92e9d0 GW |
6158 | NSEC_TO_TICK(spa->spa_deadman_synctime)); |
6159 | ||
34dc7c2f BB |
6160 | /* |
6161 | * If we are upgrading to SPA_VERSION_RAIDZ_DEFLATE this txg, | |
6162 | * set spa_deflate if we have no raid-z vdevs. | |
6163 | */ | |
6164 | if (spa->spa_ubsync.ub_version < SPA_VERSION_RAIDZ_DEFLATE && | |
6165 | spa->spa_uberblock.ub_version >= SPA_VERSION_RAIDZ_DEFLATE) { | |
6166 | int i; | |
6167 | ||
6168 | for (i = 0; i < rvd->vdev_children; i++) { | |
6169 | vd = rvd->vdev_child[i]; | |
6170 | if (vd->vdev_deflate_ratio != SPA_MINBLOCKSIZE) | |
6171 | break; | |
6172 | } | |
6173 | if (i == rvd->vdev_children) { | |
6174 | spa->spa_deflate = TRUE; | |
6175 | VERIFY(0 == zap_add(spa->spa_meta_objset, | |
6176 | DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE, | |
6177 | sizeof (uint64_t), 1, &spa->spa_deflate, tx)); | |
6178 | } | |
6179 | } | |
6180 | ||
6181 | /* | |
428870ff BB |
6182 | * If anything has changed in this txg, or if someone is waiting |
6183 | * for this txg to sync (eg, spa_vdev_remove()), push the | |
6184 | * deferred frees from the previous txg. If not, leave them | |
6185 | * alone so that we don't generate work on an otherwise idle | |
6186 | * system. | |
34dc7c2f BB |
6187 | */ |
6188 | if (!txg_list_empty(&dp->dp_dirty_datasets, txg) || | |
6189 | !txg_list_empty(&dp->dp_dirty_dirs, txg) || | |
428870ff BB |
6190 | !txg_list_empty(&dp->dp_sync_tasks, txg) || |
6191 | ((dsl_scan_active(dp->dp_scan) || | |
6192 | txg_sync_waiting(dp)) && !spa_shutting_down(spa))) { | |
6193 | zio_t *zio = zio_root(spa, NULL, NULL, 0); | |
6194 | VERIFY3U(bpobj_iterate(defer_bpo, | |
6195 | spa_free_sync_cb, zio, tx), ==, 0); | |
c99c9001 | 6196 | VERIFY0(zio_wait(zio)); |
428870ff | 6197 | } |
34dc7c2f BB |
6198 | |
6199 | /* | |
6200 | * Iterate to convergence. | |
6201 | */ | |
6202 | do { | |
428870ff | 6203 | int pass = ++spa->spa_sync_pass; |
34dc7c2f BB |
6204 | |
6205 | spa_sync_config_object(spa, tx); | |
6206 | spa_sync_aux_dev(spa, &spa->spa_spares, tx, | |
6207 | ZPOOL_CONFIG_SPARES, DMU_POOL_SPARES); | |
6208 | spa_sync_aux_dev(spa, &spa->spa_l2cache, tx, | |
6209 | ZPOOL_CONFIG_L2CACHE, DMU_POOL_L2CACHE); | |
6210 | spa_errlog_sync(spa, txg); | |
6211 | dsl_pool_sync(dp, txg); | |
6212 | ||
55d85d5a | 6213 | if (pass < zfs_sync_pass_deferred_free) { |
428870ff BB |
6214 | zio_t *zio = zio_root(spa, NULL, NULL, 0); |
6215 | bplist_iterate(free_bpl, spa_free_sync_cb, | |
6216 | zio, tx); | |
6217 | VERIFY(zio_wait(zio) == 0); | |
6218 | } else { | |
6219 | bplist_iterate(free_bpl, bpobj_enqueue_cb, | |
6220 | defer_bpo, tx); | |
34dc7c2f BB |
6221 | } |
6222 | ||
428870ff BB |
6223 | ddt_sync(spa, txg); |
6224 | dsl_scan_sync(dp, tx); | |
34dc7c2f | 6225 | |
c65aa5b2 | 6226 | while ((vd = txg_list_remove(&spa->spa_vdev_txg_list, txg))) |
428870ff BB |
6227 | vdev_sync(vd, txg); |
6228 | ||
6229 | if (pass == 1) | |
6230 | spa_sync_upgrades(spa, tx); | |
34dc7c2f | 6231 | |
428870ff | 6232 | } while (dmu_objset_is_dirty(mos, txg)); |
34dc7c2f BB |
6233 | |
6234 | /* | |
6235 | * Rewrite the vdev configuration (which includes the uberblock) | |
6236 | * to commit the transaction group. | |
6237 | * | |
6238 | * If there are no dirty vdevs, we sync the uberblock to a few | |
6239 | * random top-level vdevs that are known to be visible in the | |
b128c09f BB |
6240 | * config cache (see spa_vdev_add() for a complete description). |
6241 | * If there *are* dirty vdevs, sync the uberblock to all vdevs. | |
34dc7c2f | 6242 | */ |
b128c09f BB |
6243 | for (;;) { |
6244 | /* | |
6245 | * We hold SCL_STATE to prevent vdev open/close/etc. | |
6246 | * while we're attempting to write the vdev labels. | |
6247 | */ | |
6248 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); | |
6249 | ||
6250 | if (list_is_empty(&spa->spa_config_dirty_list)) { | |
6251 | vdev_t *svd[SPA_DVAS_PER_BP]; | |
6252 | int svdcount = 0; | |
6253 | int children = rvd->vdev_children; | |
6254 | int c0 = spa_get_random(children); | |
b128c09f | 6255 | |
d6320ddb | 6256 | for (c = 0; c < children; c++) { |
b128c09f BB |
6257 | vd = rvd->vdev_child[(c0 + c) % children]; |
6258 | if (vd->vdev_ms_array == 0 || vd->vdev_islog) | |
6259 | continue; | |
6260 | svd[svdcount++] = vd; | |
6261 | if (svdcount == SPA_DVAS_PER_BP) | |
6262 | break; | |
6263 | } | |
9babb374 BB |
6264 | error = vdev_config_sync(svd, svdcount, txg, B_FALSE); |
6265 | if (error != 0) | |
6266 | error = vdev_config_sync(svd, svdcount, txg, | |
6267 | B_TRUE); | |
b128c09f BB |
6268 | } else { |
6269 | error = vdev_config_sync(rvd->vdev_child, | |
9babb374 BB |
6270 | rvd->vdev_children, txg, B_FALSE); |
6271 | if (error != 0) | |
6272 | error = vdev_config_sync(rvd->vdev_child, | |
6273 | rvd->vdev_children, txg, B_TRUE); | |
34dc7c2f | 6274 | } |
34dc7c2f | 6275 | |
3bc7e0fb GW |
6276 | if (error == 0) |
6277 | spa->spa_last_synced_guid = rvd->vdev_guid; | |
6278 | ||
b128c09f BB |
6279 | spa_config_exit(spa, SCL_STATE, FTAG); |
6280 | ||
6281 | if (error == 0) | |
6282 | break; | |
6283 | zio_suspend(spa, NULL); | |
6284 | zio_resume_wait(spa); | |
6285 | } | |
34dc7c2f BB |
6286 | dmu_tx_commit(tx); |
6287 | ||
cc92e9d0 GW |
6288 | taskq_cancel_id(system_taskq, spa->spa_deadman_tqid); |
6289 | spa->spa_deadman_tqid = 0; | |
6290 | ||
34dc7c2f BB |
6291 | /* |
6292 | * Clear the dirty config list. | |
6293 | */ | |
b128c09f | 6294 | while ((vd = list_head(&spa->spa_config_dirty_list)) != NULL) |
34dc7c2f BB |
6295 | vdev_config_clean(vd); |
6296 | ||
6297 | /* | |
6298 | * Now that the new config has synced transactionally, | |
6299 | * let it become visible to the config cache. | |
6300 | */ | |
6301 | if (spa->spa_config_syncing != NULL) { | |
6302 | spa_config_set(spa, spa->spa_config_syncing); | |
6303 | spa->spa_config_txg = txg; | |
6304 | spa->spa_config_syncing = NULL; | |
6305 | } | |
6306 | ||
34dc7c2f | 6307 | spa->spa_ubsync = spa->spa_uberblock; |
34dc7c2f | 6308 | |
428870ff | 6309 | dsl_pool_sync_done(dp, txg); |
34dc7c2f BB |
6310 | |
6311 | /* | |
6312 | * Update usable space statistics. | |
6313 | */ | |
c65aa5b2 | 6314 | while ((vd = txg_list_remove(&spa->spa_vdev_txg_list, TXG_CLEAN(txg)))) |
34dc7c2f BB |
6315 | vdev_sync_done(vd, txg); |
6316 | ||
428870ff BB |
6317 | spa_update_dspace(spa); |
6318 | ||
34dc7c2f BB |
6319 | /* |
6320 | * It had better be the case that we didn't dirty anything | |
6321 | * since vdev_config_sync(). | |
6322 | */ | |
6323 | ASSERT(txg_list_empty(&dp->dp_dirty_datasets, txg)); | |
6324 | ASSERT(txg_list_empty(&dp->dp_dirty_dirs, txg)); | |
6325 | ASSERT(txg_list_empty(&spa->spa_vdev_txg_list, txg)); | |
428870ff BB |
6326 | |
6327 | spa->spa_sync_pass = 0; | |
34dc7c2f | 6328 | |
b128c09f | 6329 | spa_config_exit(spa, SCL_CONFIG, FTAG); |
34dc7c2f | 6330 | |
428870ff BB |
6331 | spa_handle_ignored_writes(spa); |
6332 | ||
34dc7c2f BB |
6333 | /* |
6334 | * If any async tasks have been requested, kick them off. | |
6335 | */ | |
6336 | spa_async_dispatch(spa); | |
6337 | } | |
6338 | ||
6339 | /* | |
6340 | * Sync all pools. We don't want to hold the namespace lock across these | |
6341 | * operations, so we take a reference on the spa_t and drop the lock during the | |
6342 | * sync. | |
6343 | */ | |
6344 | void | |
6345 | spa_sync_allpools(void) | |
6346 | { | |
6347 | spa_t *spa = NULL; | |
6348 | mutex_enter(&spa_namespace_lock); | |
6349 | while ((spa = spa_next(spa)) != NULL) { | |
572e2857 BB |
6350 | if (spa_state(spa) != POOL_STATE_ACTIVE || |
6351 | !spa_writeable(spa) || spa_suspended(spa)) | |
34dc7c2f BB |
6352 | continue; |
6353 | spa_open_ref(spa, FTAG); | |
6354 | mutex_exit(&spa_namespace_lock); | |
6355 | txg_wait_synced(spa_get_dsl(spa), 0); | |
6356 | mutex_enter(&spa_namespace_lock); | |
6357 | spa_close(spa, FTAG); | |
6358 | } | |
6359 | mutex_exit(&spa_namespace_lock); | |
6360 | } | |
6361 | ||
6362 | /* | |
6363 | * ========================================================================== | |
6364 | * Miscellaneous routines | |
6365 | * ========================================================================== | |
6366 | */ | |
6367 | ||
6368 | /* | |
6369 | * Remove all pools in the system. | |
6370 | */ | |
6371 | void | |
6372 | spa_evict_all(void) | |
6373 | { | |
6374 | spa_t *spa; | |
6375 | ||
6376 | /* | |
6377 | * Remove all cached state. All pools should be closed now, | |
6378 | * so every spa in the AVL tree should be unreferenced. | |
6379 | */ | |
6380 | mutex_enter(&spa_namespace_lock); | |
6381 | while ((spa = spa_next(NULL)) != NULL) { | |
6382 | /* | |
6383 | * Stop async tasks. The async thread may need to detach | |
6384 | * a device that's been replaced, which requires grabbing | |
6385 | * spa_namespace_lock, so we must drop it here. | |
6386 | */ | |
6387 | spa_open_ref(spa, FTAG); | |
6388 | mutex_exit(&spa_namespace_lock); | |
6389 | spa_async_suspend(spa); | |
6390 | mutex_enter(&spa_namespace_lock); | |
34dc7c2f BB |
6391 | spa_close(spa, FTAG); |
6392 | ||
6393 | if (spa->spa_state != POOL_STATE_UNINITIALIZED) { | |
6394 | spa_unload(spa); | |
6395 | spa_deactivate(spa); | |
6396 | } | |
6397 | spa_remove(spa); | |
6398 | } | |
6399 | mutex_exit(&spa_namespace_lock); | |
6400 | } | |
6401 | ||
6402 | vdev_t * | |
9babb374 | 6403 | spa_lookup_by_guid(spa_t *spa, uint64_t guid, boolean_t aux) |
34dc7c2f | 6404 | { |
b128c09f BB |
6405 | vdev_t *vd; |
6406 | int i; | |
6407 | ||
6408 | if ((vd = vdev_lookup_by_guid(spa->spa_root_vdev, guid)) != NULL) | |
6409 | return (vd); | |
6410 | ||
9babb374 | 6411 | if (aux) { |
b128c09f BB |
6412 | for (i = 0; i < spa->spa_l2cache.sav_count; i++) { |
6413 | vd = spa->spa_l2cache.sav_vdevs[i]; | |
9babb374 BB |
6414 | if (vd->vdev_guid == guid) |
6415 | return (vd); | |
6416 | } | |
6417 | ||
6418 | for (i = 0; i < spa->spa_spares.sav_count; i++) { | |
6419 | vd = spa->spa_spares.sav_vdevs[i]; | |
b128c09f BB |
6420 | if (vd->vdev_guid == guid) |
6421 | return (vd); | |
6422 | } | |
6423 | } | |
6424 | ||
6425 | return (NULL); | |
34dc7c2f BB |
6426 | } |
6427 | ||
6428 | void | |
6429 | spa_upgrade(spa_t *spa, uint64_t version) | |
6430 | { | |
572e2857 BB |
6431 | ASSERT(spa_writeable(spa)); |
6432 | ||
b128c09f | 6433 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f BB |
6434 | |
6435 | /* | |
6436 | * This should only be called for a non-faulted pool, and since a | |
6437 | * future version would result in an unopenable pool, this shouldn't be | |
6438 | * possible. | |
6439 | */ | |
8dca0a9a | 6440 | ASSERT(SPA_VERSION_IS_SUPPORTED(spa->spa_uberblock.ub_version)); |
34dc7c2f BB |
6441 | ASSERT(version >= spa->spa_uberblock.ub_version); |
6442 | ||
6443 | spa->spa_uberblock.ub_version = version; | |
6444 | vdev_config_dirty(spa->spa_root_vdev); | |
6445 | ||
b128c09f | 6446 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
6447 | |
6448 | txg_wait_synced(spa_get_dsl(spa), 0); | |
6449 | } | |
6450 | ||
6451 | boolean_t | |
6452 | spa_has_spare(spa_t *spa, uint64_t guid) | |
6453 | { | |
6454 | int i; | |
6455 | uint64_t spareguid; | |
6456 | spa_aux_vdev_t *sav = &spa->spa_spares; | |
6457 | ||
6458 | for (i = 0; i < sav->sav_count; i++) | |
6459 | if (sav->sav_vdevs[i]->vdev_guid == guid) | |
6460 | return (B_TRUE); | |
6461 | ||
6462 | for (i = 0; i < sav->sav_npending; i++) { | |
6463 | if (nvlist_lookup_uint64(sav->sav_pending[i], ZPOOL_CONFIG_GUID, | |
6464 | &spareguid) == 0 && spareguid == guid) | |
6465 | return (B_TRUE); | |
6466 | } | |
6467 | ||
6468 | return (B_FALSE); | |
6469 | } | |
6470 | ||
b128c09f BB |
6471 | /* |
6472 | * Check if a pool has an active shared spare device. | |
6473 | * Note: reference count of an active spare is 2, as a spare and as a replace | |
6474 | */ | |
6475 | static boolean_t | |
6476 | spa_has_active_shared_spare(spa_t *spa) | |
6477 | { | |
6478 | int i, refcnt; | |
6479 | uint64_t pool; | |
6480 | spa_aux_vdev_t *sav = &spa->spa_spares; | |
6481 | ||
6482 | for (i = 0; i < sav->sav_count; i++) { | |
6483 | if (spa_spare_exists(sav->sav_vdevs[i]->vdev_guid, &pool, | |
6484 | &refcnt) && pool != 0ULL && pool == spa_guid(spa) && | |
6485 | refcnt > 2) | |
6486 | return (B_TRUE); | |
6487 | } | |
6488 | ||
6489 | return (B_FALSE); | |
6490 | } | |
6491 | ||
34dc7c2f | 6492 | /* |
26685276 | 6493 | * Post a FM_EREPORT_ZFS_* event from sys/fm/fs/zfs.h. The payload will be |
34dc7c2f BB |
6494 | * filled in from the spa and (optionally) the vdev. This doesn't do anything |
6495 | * in the userland libzpool, as we don't want consumers to misinterpret ztest | |
6496 | * or zdb as real changes. | |
6497 | */ | |
6498 | void | |
6499 | spa_event_notify(spa_t *spa, vdev_t *vd, const char *name) | |
6500 | { | |
6501 | #ifdef _KERNEL | |
26685276 | 6502 | zfs_ereport_post(name, spa, vd, NULL, 0, 0); |
34dc7c2f BB |
6503 | #endif |
6504 | } | |
c28b2279 BB |
6505 | |
6506 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
6507 | /* state manipulation functions */ | |
6508 | EXPORT_SYMBOL(spa_open); | |
6509 | EXPORT_SYMBOL(spa_open_rewind); | |
6510 | EXPORT_SYMBOL(spa_get_stats); | |
6511 | EXPORT_SYMBOL(spa_create); | |
6512 | EXPORT_SYMBOL(spa_import_rootpool); | |
6513 | EXPORT_SYMBOL(spa_import); | |
6514 | EXPORT_SYMBOL(spa_tryimport); | |
6515 | EXPORT_SYMBOL(spa_destroy); | |
6516 | EXPORT_SYMBOL(spa_export); | |
6517 | EXPORT_SYMBOL(spa_reset); | |
6518 | EXPORT_SYMBOL(spa_async_request); | |
6519 | EXPORT_SYMBOL(spa_async_suspend); | |
6520 | EXPORT_SYMBOL(spa_async_resume); | |
6521 | EXPORT_SYMBOL(spa_inject_addref); | |
6522 | EXPORT_SYMBOL(spa_inject_delref); | |
6523 | EXPORT_SYMBOL(spa_scan_stat_init); | |
6524 | EXPORT_SYMBOL(spa_scan_get_stats); | |
6525 | ||
6526 | /* device maniion */ | |
6527 | EXPORT_SYMBOL(spa_vdev_add); | |
6528 | EXPORT_SYMBOL(spa_vdev_attach); | |
6529 | EXPORT_SYMBOL(spa_vdev_detach); | |
6530 | EXPORT_SYMBOL(spa_vdev_remove); | |
6531 | EXPORT_SYMBOL(spa_vdev_setpath); | |
6532 | EXPORT_SYMBOL(spa_vdev_setfru); | |
6533 | EXPORT_SYMBOL(spa_vdev_split_mirror); | |
6534 | ||
6535 | /* spare statech is global across all pools) */ | |
6536 | EXPORT_SYMBOL(spa_spare_add); | |
6537 | EXPORT_SYMBOL(spa_spare_remove); | |
6538 | EXPORT_SYMBOL(spa_spare_exists); | |
6539 | EXPORT_SYMBOL(spa_spare_activate); | |
6540 | ||
6541 | /* L2ARC statech is global across all pools) */ | |
6542 | EXPORT_SYMBOL(spa_l2cache_add); | |
6543 | EXPORT_SYMBOL(spa_l2cache_remove); | |
6544 | EXPORT_SYMBOL(spa_l2cache_exists); | |
6545 | EXPORT_SYMBOL(spa_l2cache_activate); | |
6546 | EXPORT_SYMBOL(spa_l2cache_drop); | |
6547 | ||
6548 | /* scanning */ | |
6549 | EXPORT_SYMBOL(spa_scan); | |
6550 | EXPORT_SYMBOL(spa_scan_stop); | |
6551 | ||
6552 | /* spa syncing */ | |
6553 | EXPORT_SYMBOL(spa_sync); /* only for DMU use */ | |
6554 | EXPORT_SYMBOL(spa_sync_allpools); | |
6555 | ||
6556 | /* properties */ | |
6557 | EXPORT_SYMBOL(spa_prop_set); | |
6558 | EXPORT_SYMBOL(spa_prop_get); | |
6559 | EXPORT_SYMBOL(spa_prop_clear_bootfs); | |
6560 | ||
6561 | /* asynchronous event notification */ | |
6562 | EXPORT_SYMBOL(spa_event_notify); | |
6563 | #endif |