]>
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. |
d96eb2b1 | 24 | * Copyright 2011 Nexenta Systems, Inc. All rights reserved. |
e0ab3ab5 | 25 | * Copyright (c) 2011, 2015 by Delphix. All rights reserved. |
12fa0466 | 26 | * Copyright 2017 Joyent, Inc. |
34dc7c2f BB |
27 | */ |
28 | ||
34dc7c2f | 29 | #include <sys/spa.h> |
e6cfd633 | 30 | #include <sys/fm/fs/zfs.h> |
34dc7c2f BB |
31 | #include <sys/spa_impl.h> |
32 | #include <sys/nvpair.h> | |
33 | #include <sys/uio.h> | |
34 | #include <sys/fs/zfs.h> | |
35 | #include <sys/vdev_impl.h> | |
36 | #include <sys/zfs_ioctl.h> | |
34dc7c2f BB |
37 | #include <sys/systeminfo.h> |
38 | #include <sys/sunddi.h> | |
9ae529ec | 39 | #include <sys/zfeature.h> |
34dc7c2f BB |
40 | #ifdef _KERNEL |
41 | #include <sys/kobj.h> | |
d164b209 | 42 | #include <sys/zone.h> |
34dc7c2f BB |
43 | #endif |
44 | ||
45 | /* | |
46 | * Pool configuration repository. | |
47 | * | |
48 | * Pool configuration is stored as a packed nvlist on the filesystem. By | |
49 | * default, all pools are stored in /etc/zfs/zpool.cache and loaded on boot | |
50 | * (when the ZFS module is loaded). Pools can also have the 'cachefile' | |
51 | * property set that allows them to be stored in an alternate location until | |
52 | * the control of external software. | |
53 | * | |
54 | * For each cache file, we have a single nvlist which holds all the | |
55 | * configuration information. When the module loads, we read this information | |
56 | * from /etc/zfs/zpool.cache and populate the SPA namespace. This namespace is | |
57 | * maintained independently in spa.c. Whenever the namespace is modified, or | |
58 | * the configuration of a pool is changed, we call spa_config_sync(), which | |
59 | * walks through all the active pools and writes the configuration to disk. | |
60 | */ | |
61 | ||
62 | static uint64_t spa_config_generation = 1; | |
63 | ||
64 | /* | |
65 | * This can be overridden in userland to preserve an alternate namespace for | |
66 | * userland pools when doing testing. | |
67 | */ | |
c28b2279 | 68 | char *spa_config_path = ZPOOL_CACHE; |
1611bb7b | 69 | int zfs_autoimport_disable = 1; |
34dc7c2f BB |
70 | |
71 | /* | |
72 | * Called when the module is first loaded, this routine loads the configuration | |
73 | * file into the SPA namespace. It does not actually open or load the pools; it | |
74 | * only populates the namespace. | |
75 | */ | |
76 | void | |
77 | spa_config_load(void) | |
78 | { | |
79 | void *buf = NULL; | |
80 | nvlist_t *nvlist, *child; | |
81 | nvpair_t *nvpair; | |
b128c09f | 82 | char *pathname; |
34dc7c2f BB |
83 | struct _buf *file; |
84 | uint64_t fsize; | |
85 | ||
1611bb7b | 86 | #ifdef _KERNEL |
31455ab1 BB |
87 | if (zfs_autoimport_disable) |
88 | return; | |
1611bb7b | 89 | #endif |
31455ab1 | 90 | |
34dc7c2f BB |
91 | /* |
92 | * Open the configuration file. | |
93 | */ | |
79c76d5b | 94 | pathname = kmem_alloc(MAXPATHLEN, KM_SLEEP); |
b128c09f BB |
95 | |
96 | (void) snprintf(pathname, MAXPATHLEN, "%s%s", | |
97 | (rootdir != NULL) ? "./" : "", spa_config_path); | |
34dc7c2f BB |
98 | |
99 | file = kobj_open_file(pathname); | |
b128c09f BB |
100 | |
101 | kmem_free(pathname, MAXPATHLEN); | |
102 | ||
34dc7c2f BB |
103 | if (file == (struct _buf *)-1) |
104 | return; | |
105 | ||
106 | if (kobj_get_filesize(file, &fsize) != 0) | |
107 | goto out; | |
108 | ||
79c76d5b | 109 | buf = kmem_alloc(fsize, KM_SLEEP); |
34dc7c2f BB |
110 | |
111 | /* | |
112 | * Read the nvlist from the file. | |
113 | */ | |
114 | if (kobj_read_file(file, buf, fsize, 0) < 0) | |
115 | goto out; | |
116 | ||
117 | /* | |
118 | * Unpack the nvlist. | |
119 | */ | |
79c76d5b | 120 | if (nvlist_unpack(buf, fsize, &nvlist, KM_SLEEP) != 0) |
34dc7c2f BB |
121 | goto out; |
122 | ||
123 | /* | |
124 | * Iterate over all elements in the nvlist, creating a new spa_t for | |
125 | * each one with the specified configuration. | |
126 | */ | |
127 | mutex_enter(&spa_namespace_lock); | |
128 | nvpair = NULL; | |
129 | while ((nvpair = nvlist_next_nvpair(nvlist, nvpair)) != NULL) { | |
34dc7c2f BB |
130 | if (nvpair_type(nvpair) != DATA_TYPE_NVLIST) |
131 | continue; | |
132 | ||
e0ab3ab5 | 133 | child = fnvpair_value_nvlist(nvpair); |
34dc7c2f BB |
134 | |
135 | if (spa_lookup(nvpair_name(nvpair)) != NULL) | |
136 | continue; | |
428870ff | 137 | (void) spa_add(nvpair_name(nvpair), child, NULL); |
34dc7c2f BB |
138 | } |
139 | mutex_exit(&spa_namespace_lock); | |
140 | ||
141 | nvlist_free(nvlist); | |
142 | ||
143 | out: | |
144 | if (buf != NULL) | |
145 | kmem_free(buf, fsize); | |
146 | ||
147 | kobj_close_file(file); | |
148 | } | |
149 | ||
bbf1ad67 BB |
150 | static int |
151 | spa_config_remove(spa_config_dirent_t *dp) | |
152 | { | |
153 | #if defined(__linux__) && defined(_KERNEL) | |
154 | int error, flags = FWRITE | FTRUNC; | |
155 | uio_seg_t seg = UIO_SYSSPACE; | |
156 | vnode_t *vp; | |
157 | ||
158 | error = vn_open(dp->scd_path, seg, flags, 0644, &vp, 0, 0); | |
159 | if (error == 0) { | |
160 | (void) VOP_FSYNC(vp, FSYNC, kcred, NULL); | |
161 | (void) VOP_CLOSE(vp, 0, 1, 0, kcred, NULL); | |
162 | } | |
163 | ||
164 | return (error); | |
165 | #else | |
166 | return (vn_remove(dp->scd_path, UIO_SYSSPACE, RMFILE)); | |
167 | #endif | |
168 | } | |
169 | ||
e6cfd633 | 170 | static int |
b128c09f | 171 | spa_config_write(spa_config_dirent_t *dp, nvlist_t *nvl) |
34dc7c2f | 172 | { |
34dc7c2f BB |
173 | size_t buflen; |
174 | char *buf; | |
175 | vnode_t *vp; | |
176 | int oflags = FWRITE | FTRUNC | FCREAT | FOFFMAX; | |
b128c09f | 177 | char *temp; |
e6cfd633 | 178 | int err; |
b128c09f BB |
179 | |
180 | /* | |
181 | * If the nvlist is empty (NULL), then remove the old cachefile. | |
182 | */ | |
183 | if (nvl == NULL) { | |
bbf1ad67 | 184 | err = spa_config_remove(dp); |
d9549cba AB |
185 | if (err == ENOENT) |
186 | err = 0; | |
bbf1ad67 | 187 | |
e6cfd633 | 188 | return (err); |
b128c09f | 189 | } |
34dc7c2f BB |
190 | |
191 | /* | |
192 | * Pack the configuration into a buffer. | |
193 | */ | |
e0ab3ab5 | 194 | buf = fnvlist_pack(nvl, &buflen); |
79c76d5b | 195 | temp = kmem_zalloc(MAXPATHLEN, KM_SLEEP); |
34dc7c2f | 196 | |
151f84e2 | 197 | #if defined(__linux__) && defined(_KERNEL) |
efc412b6 BB |
198 | /* |
199 | * Write the configuration to disk. Due to the complexity involved | |
bbf1ad67 BB |
200 | * in performing a rename and remove from within the kernel the file |
201 | * is instead truncated and overwritten in place. This way we always | |
202 | * have a consistent view of the data or a zero length file. | |
efc412b6 | 203 | */ |
e6cfd633 WA |
204 | err = vn_open(dp->scd_path, UIO_SYSSPACE, oflags, 0644, &vp, 0, 0); |
205 | if (err == 0) { | |
206 | err = vn_rdwr(UIO_WRITE, vp, buf, buflen, 0, | |
efc412b6 | 207 | UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, NULL); |
e6cfd633 WA |
208 | if (err == 0) |
209 | err = VOP_FSYNC(vp, FSYNC, kcred, NULL); | |
efc412b6 BB |
210 | |
211 | (void) VOP_CLOSE(vp, oflags, 1, 0, kcred, NULL); | |
e6cfd633 | 212 | if (err) |
bbf1ad67 | 213 | (void) spa_config_remove(dp); |
efc412b6 BB |
214 | } |
215 | #else | |
34dc7c2f BB |
216 | /* |
217 | * Write the configuration to disk. We need to do the traditional | |
218 | * 'write to temporary file, sync, move over original' to make sure we | |
219 | * always have a consistent view of the data. | |
220 | */ | |
b128c09f | 221 | (void) snprintf(temp, MAXPATHLEN, "%s.tmp", dp->scd_path); |
34dc7c2f | 222 | |
e6cfd633 WA |
223 | err = vn_open(temp, UIO_SYSSPACE, oflags, 0644, &vp, CRCREAT, 0); |
224 | if (err == 0) { | |
225 | err = vn_rdwr(UIO_WRITE, vp, buf, buflen, 0, UIO_SYSSPACE, | |
226 | 0, RLIM64_INFINITY, kcred, NULL); | |
227 | if (err == 0) | |
228 | err = VOP_FSYNC(vp, FSYNC, kcred, NULL); | |
229 | if (err == 0) | |
230 | err = vn_rename(temp, dp->scd_path, UIO_SYSSPACE); | |
b128c09f | 231 | (void) VOP_CLOSE(vp, oflags, 1, 0, kcred, NULL); |
34dc7c2f BB |
232 | } |
233 | ||
b128c09f | 234 | (void) vn_remove(temp, UIO_SYSSPACE, RMFILE); |
efc412b6 | 235 | #endif |
34dc7c2f | 236 | |
e0ab3ab5 | 237 | fnvlist_pack_free(buf, buflen); |
b128c09f | 238 | kmem_free(temp, MAXPATHLEN); |
e6cfd633 | 239 | return (err); |
34dc7c2f BB |
240 | } |
241 | ||
242 | /* | |
b128c09f | 243 | * Synchronize pool configuration to disk. This must be called with the |
5d1f7fb6 GW |
244 | * namespace lock held. Synchronizing the pool cache is typically done after |
245 | * the configuration has been synced to the MOS. This exposes a window where | |
246 | * the MOS config will have been updated but the cache file has not. If | |
247 | * the system were to crash at that instant then the cached config may not | |
4e33ba4c | 248 | * contain the correct information to open the pool and an explicit import |
5d1f7fb6 | 249 | * would be required. |
34dc7c2f BB |
250 | */ |
251 | void | |
b128c09f | 252 | spa_config_sync(spa_t *target, boolean_t removing, boolean_t postsysevent) |
34dc7c2f | 253 | { |
b128c09f BB |
254 | spa_config_dirent_t *dp, *tdp; |
255 | nvlist_t *nvl; | |
26b42f3f | 256 | char *pool_name; |
e6cfd633 WA |
257 | boolean_t ccw_failure; |
258 | int error = 0; | |
34dc7c2f BB |
259 | |
260 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
261 | ||
45d1cae3 | 262 | if (rootdir == NULL || !(spa_mode_global & FWRITE)) |
fb5f0bc8 BB |
263 | return; |
264 | ||
b128c09f BB |
265 | /* |
266 | * Iterate over all cachefiles for the pool, past or present. When the | |
267 | * cachefile is changed, the new one is pushed onto this list, allowing | |
268 | * us to update previous cachefiles that no longer contain this pool. | |
269 | */ | |
e6cfd633 | 270 | ccw_failure = B_FALSE; |
b128c09f BB |
271 | for (dp = list_head(&target->spa_config_list); dp != NULL; |
272 | dp = list_next(&target->spa_config_list, dp)) { | |
273 | spa_t *spa = NULL; | |
274 | if (dp->scd_path == NULL) | |
275 | continue; | |
276 | ||
277 | /* | |
278 | * Iterate over all pools, adding any matching pools to 'nvl'. | |
279 | */ | |
280 | nvl = NULL; | |
281 | while ((spa = spa_next(spa)) != NULL) { | |
c61f97f4 GW |
282 | /* |
283 | * Skip over our own pool if we're about to remove | |
284 | * ourselves from the spa namespace or any pool that | |
285 | * is readonly. Since we cannot guarantee that a | |
286 | * readonly pool would successfully import upon reboot, | |
287 | * we don't allow them to be written to the cache file. | |
288 | */ | |
289 | if ((spa == target && removing) || | |
290 | !spa_writeable(spa)) | |
b128c09f BB |
291 | continue; |
292 | ||
293 | mutex_enter(&spa->spa_props_lock); | |
294 | tdp = list_head(&spa->spa_config_list); | |
295 | if (spa->spa_config == NULL || | |
ccc92611 | 296 | tdp == NULL || |
b128c09f BB |
297 | tdp->scd_path == NULL || |
298 | strcmp(tdp->scd_path, dp->scd_path) != 0) { | |
299 | mutex_exit(&spa->spa_props_lock); | |
300 | continue; | |
301 | } | |
302 | ||
303 | if (nvl == NULL) | |
e0ab3ab5 | 304 | nvl = fnvlist_alloc(); |
b128c09f | 305 | |
e0ab3ab5 JS |
306 | if (spa->spa_import_flags & ZFS_IMPORT_TEMP_NAME) |
307 | pool_name = fnvlist_lookup_string( | |
308 | spa->spa_config, ZPOOL_CONFIG_POOL_NAME); | |
309 | else | |
26b42f3f RY |
310 | pool_name = spa_name(spa); |
311 | ||
e0ab3ab5 | 312 | fnvlist_add_nvlist(nvl, pool_name, spa->spa_config); |
b128c09f BB |
313 | mutex_exit(&spa->spa_props_lock); |
314 | } | |
315 | ||
e6cfd633 WA |
316 | error = spa_config_write(dp, nvl); |
317 | if (error != 0) | |
318 | ccw_failure = B_TRUE; | |
b128c09f BB |
319 | nvlist_free(nvl); |
320 | } | |
34dc7c2f | 321 | |
e6cfd633 WA |
322 | if (ccw_failure) { |
323 | /* | |
324 | * Keep trying so that configuration data is | |
325 | * written if/when any temporary filesystem | |
326 | * resource issues are resolved. | |
327 | */ | |
328 | if (target->spa_ccw_fail_time == 0) { | |
329 | zfs_ereport_post(FM_EREPORT_ZFS_CONFIG_CACHE_WRITE, | |
b5256303 | 330 | target, NULL, NULL, NULL, 0, 0); |
e6cfd633 WA |
331 | } |
332 | target->spa_ccw_fail_time = gethrtime(); | |
333 | spa_async_request(target, SPA_ASYNC_CONFIG_UPDATE); | |
334 | } else { | |
335 | /* | |
336 | * Do not rate limit future attempts to update | |
337 | * the config cache. | |
338 | */ | |
339 | target->spa_ccw_fail_time = 0; | |
340 | } | |
341 | ||
34dc7c2f | 342 | /* |
b128c09f | 343 | * Remove any config entries older than the current one. |
34dc7c2f | 344 | */ |
b128c09f BB |
345 | dp = list_head(&target->spa_config_list); |
346 | while ((tdp = list_next(&target->spa_config_list, dp)) != NULL) { | |
347 | list_remove(&target->spa_config_list, tdp); | |
348 | if (tdp->scd_path != NULL) | |
349 | spa_strfree(tdp->scd_path); | |
350 | kmem_free(tdp, sizeof (spa_config_dirent_t)); | |
34dc7c2f BB |
351 | } |
352 | ||
353 | spa_config_generation++; | |
b128c09f BB |
354 | |
355 | if (postsysevent) | |
12fa0466 | 356 | spa_event_notify(target, NULL, NULL, ESC_ZFS_CONFIG_SYNC); |
34dc7c2f BB |
357 | } |
358 | ||
359 | /* | |
360 | * Sigh. Inside a local zone, we don't have access to /etc/zfs/zpool.cache, | |
361 | * and we don't want to allow the local zone to see all the pools anyway. | |
362 | * So we have to invent the ZFS_IOC_CONFIG ioctl to grab the configuration | |
363 | * information for all pool visible within the zone. | |
364 | */ | |
365 | nvlist_t * | |
366 | spa_all_configs(uint64_t *generation) | |
367 | { | |
368 | nvlist_t *pools; | |
b128c09f | 369 | spa_t *spa = NULL; |
34dc7c2f BB |
370 | |
371 | if (*generation == spa_config_generation) | |
372 | return (NULL); | |
373 | ||
e0ab3ab5 | 374 | pools = fnvlist_alloc(); |
34dc7c2f | 375 | |
34dc7c2f BB |
376 | mutex_enter(&spa_namespace_lock); |
377 | while ((spa = spa_next(spa)) != NULL) { | |
378 | if (INGLOBALZONE(curproc) || | |
379 | zone_dataset_visible(spa_name(spa), NULL)) { | |
b128c09f | 380 | mutex_enter(&spa->spa_props_lock); |
e0ab3ab5 JS |
381 | fnvlist_add_nvlist(pools, spa_name(spa), |
382 | spa->spa_config); | |
b128c09f | 383 | mutex_exit(&spa->spa_props_lock); |
34dc7c2f BB |
384 | } |
385 | } | |
34dc7c2f | 386 | *generation = spa_config_generation; |
b128c09f | 387 | mutex_exit(&spa_namespace_lock); |
34dc7c2f BB |
388 | |
389 | return (pools); | |
390 | } | |
391 | ||
392 | void | |
393 | spa_config_set(spa_t *spa, nvlist_t *config) | |
394 | { | |
b128c09f | 395 | mutex_enter(&spa->spa_props_lock); |
8a5fc748 | 396 | nvlist_free(spa->spa_config); |
34dc7c2f | 397 | spa->spa_config = config; |
b128c09f | 398 | mutex_exit(&spa->spa_props_lock); |
34dc7c2f BB |
399 | } |
400 | ||
401 | /* | |
402 | * Generate the pool's configuration based on the current in-core state. | |
d3cc8b15 | 403 | * |
34dc7c2f BB |
404 | * We infer whether to generate a complete config or just one top-level config |
405 | * based on whether vd is the root vdev. | |
406 | */ | |
407 | nvlist_t * | |
408 | spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats) | |
409 | { | |
410 | nvlist_t *config, *nvroot; | |
411 | vdev_t *rvd = spa->spa_root_vdev; | |
412 | unsigned long hostid = 0; | |
b128c09f | 413 | boolean_t locked = B_FALSE; |
428870ff | 414 | uint64_t split_guid; |
26b42f3f | 415 | char *pool_name; |
34dc7c2f | 416 | |
b128c09f | 417 | if (vd == NULL) { |
34dc7c2f | 418 | vd = rvd; |
b128c09f BB |
419 | locked = B_TRUE; |
420 | spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER); | |
421 | } | |
422 | ||
423 | ASSERT(spa_config_held(spa, SCL_CONFIG | SCL_STATE, RW_READER) == | |
424 | (SCL_CONFIG | SCL_STATE)); | |
34dc7c2f BB |
425 | |
426 | /* | |
427 | * If txg is -1, report the current value of spa->spa_config_txg. | |
428 | */ | |
429 | if (txg == -1ULL) | |
430 | txg = spa->spa_config_txg; | |
431 | ||
26b42f3f RY |
432 | /* |
433 | * Originally, users had to handle spa namespace collisions by either | |
434 | * exporting the already imported pool or by specifying a new name for | |
435 | * the pool with a conflicting name. In the case of root pools from | |
436 | * virtual guests, neither approach to collision resolution is | |
437 | * reasonable. This is addressed by extending the new name syntax with | |
438 | * an option to specify that the new name is temporary. When specified, | |
439 | * ZFS_IMPORT_TEMP_NAME will be set in spa->spa_import_flags to tell us | |
440 | * to use the previous name, which we do below. | |
441 | */ | |
442 | if (spa->spa_import_flags & ZFS_IMPORT_TEMP_NAME) { | |
443 | VERIFY0(nvlist_lookup_string(spa->spa_config, | |
02730c33 | 444 | ZPOOL_CONFIG_POOL_NAME, &pool_name)); |
26b42f3f RY |
445 | } else |
446 | pool_name = spa_name(spa); | |
447 | ||
e0ab3ab5 | 448 | config = fnvlist_alloc(); |
d96eb2b1 | 449 | |
e0ab3ab5 | 450 | fnvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, spa_version(spa)); |
3500a145 | 451 | fnvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, pool_name); |
e0ab3ab5 JS |
452 | fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, spa_state(spa)); |
453 | fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG, txg); | |
454 | fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID, spa_guid(spa)); | |
455 | fnvlist_add_uint64(config, ZPOOL_CONFIG_ERRATA, spa->spa_errata); | |
456 | if (spa->spa_comment != NULL) | |
457 | fnvlist_add_string(config, ZPOOL_CONFIG_COMMENT, | |
458 | spa->spa_comment); | |
d96eb2b1 | 459 | |
379ca9cf | 460 | hostid = spa_get_hostid(); |
e0ab3ab5 JS |
461 | if (hostid != 0) |
462 | fnvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID, hostid); | |
463 | fnvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME, utsname()->nodename); | |
34dc7c2f | 464 | |
1c27024e | 465 | int config_gen_flags = 0; |
34dc7c2f | 466 | if (vd != rvd) { |
e0ab3ab5 JS |
467 | fnvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID, |
468 | vd->vdev_top->vdev_guid); | |
469 | fnvlist_add_uint64(config, ZPOOL_CONFIG_GUID, | |
470 | vd->vdev_guid); | |
34dc7c2f | 471 | if (vd->vdev_isspare) |
e0ab3ab5 JS |
472 | fnvlist_add_uint64(config, |
473 | ZPOOL_CONFIG_IS_SPARE, 1ULL); | |
34dc7c2f | 474 | if (vd->vdev_islog) |
e0ab3ab5 JS |
475 | fnvlist_add_uint64(config, |
476 | ZPOOL_CONFIG_IS_LOG, 1ULL); | |
34dc7c2f | 477 | vd = vd->vdev_top; /* label contains top config */ |
428870ff BB |
478 | } else { |
479 | /* | |
480 | * Only add the (potentially large) split information | |
481 | * in the mos config, and not in the vdev labels | |
482 | */ | |
483 | if (spa->spa_config_splitting != NULL) | |
e0ab3ab5 JS |
484 | fnvlist_add_nvlist(config, ZPOOL_CONFIG_SPLIT, |
485 | spa->spa_config_splitting); | |
486 | ||
487 | fnvlist_add_boolean(config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS); | |
488 | ||
489 | config_gen_flags |= VDEV_CONFIG_MOS; | |
34dc7c2f BB |
490 | } |
491 | ||
428870ff BB |
492 | /* |
493 | * Add the top-level config. We even add this on pools which | |
572e2857 | 494 | * don't support holes in the namespace. |
428870ff BB |
495 | */ |
496 | vdev_top_config_generate(spa, config); | |
497 | ||
498 | /* | |
499 | * If we're splitting, record the original pool's guid. | |
500 | */ | |
501 | if (spa->spa_config_splitting != NULL && | |
502 | nvlist_lookup_uint64(spa->spa_config_splitting, | |
503 | ZPOOL_CONFIG_SPLIT_GUID, &split_guid) == 0) { | |
e0ab3ab5 | 504 | fnvlist_add_uint64(config, ZPOOL_CONFIG_SPLIT_GUID, split_guid); |
428870ff BB |
505 | } |
506 | ||
e0ab3ab5 JS |
507 | nvroot = vdev_config_generate(spa, vd, getstats, config_gen_flags); |
508 | fnvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot); | |
34dc7c2f BB |
509 | nvlist_free(nvroot); |
510 | ||
9ae529ec CS |
511 | /* |
512 | * Store what's necessary for reading the MOS in the label. | |
513 | */ | |
e0ab3ab5 JS |
514 | fnvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ, |
515 | spa->spa_label_features); | |
9ae529ec | 516 | |
428870ff BB |
517 | if (getstats && spa_load_state(spa) == SPA_LOAD_NONE) { |
518 | ddt_histogram_t *ddh; | |
519 | ddt_stat_t *dds; | |
520 | ddt_object_t *ddo; | |
521 | ||
79c76d5b | 522 | ddh = kmem_zalloc(sizeof (ddt_histogram_t), KM_SLEEP); |
428870ff | 523 | ddt_get_dedup_histogram(spa, ddh); |
e0ab3ab5 | 524 | fnvlist_add_uint64_array(config, |
428870ff | 525 | ZPOOL_CONFIG_DDT_HISTOGRAM, |
e0ab3ab5 | 526 | (uint64_t *)ddh, sizeof (*ddh) / sizeof (uint64_t)); |
428870ff BB |
527 | kmem_free(ddh, sizeof (ddt_histogram_t)); |
528 | ||
79c76d5b | 529 | ddo = kmem_zalloc(sizeof (ddt_object_t), KM_SLEEP); |
428870ff | 530 | ddt_get_dedup_object_stats(spa, ddo); |
e0ab3ab5 | 531 | fnvlist_add_uint64_array(config, |
428870ff | 532 | ZPOOL_CONFIG_DDT_OBJ_STATS, |
e0ab3ab5 | 533 | (uint64_t *)ddo, sizeof (*ddo) / sizeof (uint64_t)); |
428870ff BB |
534 | kmem_free(ddo, sizeof (ddt_object_t)); |
535 | ||
79c76d5b | 536 | dds = kmem_zalloc(sizeof (ddt_stat_t), KM_SLEEP); |
428870ff | 537 | ddt_get_dedup_stats(spa, dds); |
e0ab3ab5 | 538 | fnvlist_add_uint64_array(config, |
428870ff | 539 | ZPOOL_CONFIG_DDT_STATS, |
e0ab3ab5 | 540 | (uint64_t *)dds, sizeof (*dds) / sizeof (uint64_t)); |
428870ff BB |
541 | kmem_free(dds, sizeof (ddt_stat_t)); |
542 | } | |
543 | ||
b128c09f BB |
544 | if (locked) |
545 | spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); | |
546 | ||
34dc7c2f BB |
547 | return (config); |
548 | } | |
549 | ||
34dc7c2f BB |
550 | /* |
551 | * Update all disk labels, generate a fresh config based on the current | |
552 | * in-core state, and sync the global config cache (do not sync the config | |
553 | * cache if this is a booting rootpool). | |
554 | */ | |
555 | void | |
45d1cae3 | 556 | spa_config_update(spa_t *spa, int what) |
34dc7c2f BB |
557 | { |
558 | vdev_t *rvd = spa->spa_root_vdev; | |
559 | uint64_t txg; | |
560 | int c; | |
561 | ||
562 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
563 | ||
b128c09f | 564 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f BB |
565 | txg = spa_last_synced_txg(spa) + 1; |
566 | if (what == SPA_CONFIG_UPDATE_POOL) { | |
567 | vdev_config_dirty(rvd); | |
568 | } else { | |
569 | /* | |
570 | * If we have top-level vdevs that were added but have | |
571 | * not yet been prepared for allocation, do that now. | |
572 | * (It's safe now because the config cache is up to date, | |
573 | * so it will be able to translate the new DVAs.) | |
574 | * See comments in spa_vdev_add() for full details. | |
575 | */ | |
576 | for (c = 0; c < rvd->vdev_children; c++) { | |
577 | vdev_t *tvd = rvd->vdev_child[c]; | |
9babb374 BB |
578 | if (tvd->vdev_ms_array == 0) |
579 | vdev_metaslab_set_size(tvd); | |
580 | vdev_expand(tvd, txg); | |
34dc7c2f BB |
581 | } |
582 | } | |
b128c09f | 583 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
584 | |
585 | /* | |
586 | * Wait for the mosconfig to be regenerated and synced. | |
587 | */ | |
588 | txg_wait_synced(spa->spa_dsl_pool, txg); | |
589 | ||
590 | /* | |
591 | * Update the global config cache to reflect the new mosconfig. | |
592 | */ | |
45d1cae3 | 593 | if (!spa->spa_is_root) |
b128c09f | 594 | spa_config_sync(spa, B_FALSE, what != SPA_CONFIG_UPDATE_POOL); |
34dc7c2f BB |
595 | |
596 | if (what == SPA_CONFIG_UPDATE_POOL) | |
45d1cae3 | 597 | spa_config_update(spa, SPA_CONFIG_UPDATE_VDEVS); |
34dc7c2f | 598 | } |
c28b2279 BB |
599 | |
600 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
601 | EXPORT_SYMBOL(spa_config_sync); | |
602 | EXPORT_SYMBOL(spa_config_load); | |
603 | EXPORT_SYMBOL(spa_all_configs); | |
604 | EXPORT_SYMBOL(spa_config_set); | |
605 | EXPORT_SYMBOL(spa_config_generate); | |
606 | EXPORT_SYMBOL(spa_config_update); | |
607 | ||
608 | module_param(spa_config_path, charp, 0444); | |
609 | MODULE_PARM_DESC(spa_config_path, "SPA config file (/etc/zfs/zpool.cache)"); | |
31455ab1 BB |
610 | |
611 | module_param(zfs_autoimport_disable, int, 0644); | |
612 | MODULE_PARM_DESC(zfs_autoimport_disable, "Disable pool import at module load"); | |
613 | ||
c28b2279 | 614 | #endif |