]>
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 | /* | |
428870ff | 22 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
2e528b49 | 23 | * Copyright (c) 2013 by Delphix. All rights reserved. |
3541dc6d | 24 | * Copyright 2011 Nexenta Systems, Inc. All rights reserved. |
34dc7c2f BB |
25 | */ |
26 | ||
34dc7c2f BB |
27 | #include <sys/zfs_context.h> |
28 | #include <sys/spa_impl.h> | |
29 | #include <sys/zio.h> | |
30 | #include <sys/zio_checksum.h> | |
31 | #include <sys/zio_compress.h> | |
32 | #include <sys/dmu.h> | |
33 | #include <sys/dmu_tx.h> | |
34 | #include <sys/zap.h> | |
35 | #include <sys/zil.h> | |
36 | #include <sys/vdev_impl.h> | |
37 | #include <sys/metaslab.h> | |
38 | #include <sys/uberblock_impl.h> | |
39 | #include <sys/txg.h> | |
40 | #include <sys/avl.h> | |
41 | #include <sys/unique.h> | |
42 | #include <sys/dsl_pool.h> | |
43 | #include <sys/dsl_dir.h> | |
44 | #include <sys/dsl_prop.h> | |
26685276 | 45 | #include <sys/fm/util.h> |
428870ff | 46 | #include <sys/dsl_scan.h> |
34dc7c2f BB |
47 | #include <sys/fs/zfs.h> |
48 | #include <sys/metaslab_impl.h> | |
b128c09f | 49 | #include <sys/arc.h> |
428870ff | 50 | #include <sys/ddt.h> |
1421c891 | 51 | #include <sys/kstat.h> |
34dc7c2f | 52 | #include "zfs_prop.h" |
9ae529ec | 53 | #include "zfeature_common.h" |
34dc7c2f BB |
54 | |
55 | /* | |
56 | * SPA locking | |
57 | * | |
58 | * There are four basic locks for managing spa_t structures: | |
59 | * | |
60 | * spa_namespace_lock (global mutex) | |
61 | * | |
62 | * This lock must be acquired to do any of the following: | |
63 | * | |
64 | * - Lookup a spa_t by name | |
65 | * - Add or remove a spa_t from the namespace | |
66 | * - Increase spa_refcount from non-zero | |
67 | * - Check if spa_refcount is zero | |
68 | * - Rename a spa_t | |
69 | * - add/remove/attach/detach devices | |
70 | * - Held for the duration of create/destroy/import/export | |
71 | * | |
72 | * It does not need to handle recursion. A create or destroy may | |
73 | * reference objects (files or zvols) in other pools, but by | |
74 | * definition they must have an existing reference, and will never need | |
75 | * to lookup a spa_t by name. | |
76 | * | |
77 | * spa_refcount (per-spa refcount_t protected by mutex) | |
78 | * | |
79 | * This reference count keep track of any active users of the spa_t. The | |
80 | * spa_t cannot be destroyed or freed while this is non-zero. Internally, | |
81 | * the refcount is never really 'zero' - opening a pool implicitly keeps | |
b128c09f | 82 | * some references in the DMU. Internally we check against spa_minref, but |
34dc7c2f BB |
83 | * present the image of a zero/non-zero value to consumers. |
84 | * | |
b128c09f | 85 | * spa_config_lock[] (per-spa array of rwlocks) |
34dc7c2f BB |
86 | * |
87 | * This protects the spa_t from config changes, and must be held in | |
88 | * the following circumstances: | |
89 | * | |
90 | * - RW_READER to perform I/O to the spa | |
91 | * - RW_WRITER to change the vdev config | |
92 | * | |
34dc7c2f BB |
93 | * The locking order is fairly straightforward: |
94 | * | |
95 | * spa_namespace_lock -> spa_refcount | |
96 | * | |
97 | * The namespace lock must be acquired to increase the refcount from 0 | |
98 | * or to check if it is zero. | |
99 | * | |
b128c09f | 100 | * spa_refcount -> spa_config_lock[] |
34dc7c2f BB |
101 | * |
102 | * There must be at least one valid reference on the spa_t to acquire | |
103 | * the config lock. | |
104 | * | |
b128c09f | 105 | * spa_namespace_lock -> spa_config_lock[] |
34dc7c2f BB |
106 | * |
107 | * The namespace lock must always be taken before the config lock. | |
108 | * | |
109 | * | |
b128c09f | 110 | * The spa_namespace_lock can be acquired directly and is globally visible. |
34dc7c2f | 111 | * |
b128c09f BB |
112 | * The namespace is manipulated using the following functions, all of which |
113 | * require the spa_namespace_lock to be held. | |
34dc7c2f BB |
114 | * |
115 | * spa_lookup() Lookup a spa_t by name. | |
116 | * | |
117 | * spa_add() Create a new spa_t in the namespace. | |
118 | * | |
119 | * spa_remove() Remove a spa_t from the namespace. This also | |
120 | * frees up any memory associated with the spa_t. | |
121 | * | |
122 | * spa_next() Returns the next spa_t in the system, or the | |
123 | * first if NULL is passed. | |
124 | * | |
125 | * spa_evict_all() Shutdown and remove all spa_t structures in | |
126 | * the system. | |
127 | * | |
128 | * spa_guid_exists() Determine whether a pool/device guid exists. | |
129 | * | |
130 | * The spa_refcount is manipulated using the following functions: | |
131 | * | |
132 | * spa_open_ref() Adds a reference to the given spa_t. Must be | |
133 | * called with spa_namespace_lock held if the | |
134 | * refcount is currently zero. | |
135 | * | |
136 | * spa_close() Remove a reference from the spa_t. This will | |
137 | * not free the spa_t or remove it from the | |
138 | * namespace. No locking is required. | |
139 | * | |
140 | * spa_refcount_zero() Returns true if the refcount is currently | |
141 | * zero. Must be called with spa_namespace_lock | |
142 | * held. | |
143 | * | |
b128c09f BB |
144 | * The spa_config_lock[] is an array of rwlocks, ordered as follows: |
145 | * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV. | |
146 | * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}(). | |
147 | * | |
148 | * To read the configuration, it suffices to hold one of these locks as reader. | |
149 | * To modify the configuration, you must hold all locks as writer. To modify | |
150 | * vdev state without altering the vdev tree's topology (e.g. online/offline), | |
151 | * you must hold SCL_STATE and SCL_ZIO as writer. | |
152 | * | |
153 | * We use these distinct config locks to avoid recursive lock entry. | |
154 | * For example, spa_sync() (which holds SCL_CONFIG as reader) induces | |
155 | * block allocations (SCL_ALLOC), which may require reading space maps | |
156 | * from disk (dmu_read() -> zio_read() -> SCL_ZIO). | |
157 | * | |
158 | * The spa config locks cannot be normal rwlocks because we need the | |
159 | * ability to hand off ownership. For example, SCL_ZIO is acquired | |
160 | * by the issuing thread and later released by an interrupt thread. | |
161 | * They do, however, obey the usual write-wanted semantics to prevent | |
162 | * writer (i.e. system administrator) starvation. | |
163 | * | |
164 | * The lock acquisition rules are as follows: | |
165 | * | |
166 | * SCL_CONFIG | |
167 | * Protects changes to the vdev tree topology, such as vdev | |
168 | * add/remove/attach/detach. Protects the dirty config list | |
169 | * (spa_config_dirty_list) and the set of spares and l2arc devices. | |
170 | * | |
171 | * SCL_STATE | |
172 | * Protects changes to pool state and vdev state, such as vdev | |
173 | * online/offline/fault/degrade/clear. Protects the dirty state list | |
174 | * (spa_state_dirty_list) and global pool state (spa_state). | |
175 | * | |
176 | * SCL_ALLOC | |
177 | * Protects changes to metaslab groups and classes. | |
178 | * Held as reader by metaslab_alloc() and metaslab_claim(). | |
179 | * | |
180 | * SCL_ZIO | |
181 | * Held by bp-level zios (those which have no io_vd upon entry) | |
182 | * to prevent changes to the vdev tree. The bp-level zio implicitly | |
183 | * protects all of its vdev child zios, which do not hold SCL_ZIO. | |
184 | * | |
185 | * SCL_FREE | |
186 | * Protects changes to metaslab groups and classes. | |
187 | * Held as reader by metaslab_free(). SCL_FREE is distinct from | |
188 | * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free | |
189 | * blocks in zio_done() while another i/o that holds either | |
190 | * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete. | |
191 | * | |
192 | * SCL_VDEV | |
193 | * Held as reader to prevent changes to the vdev tree during trivial | |
428870ff | 194 | * inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the |
b128c09f BB |
195 | * other locks, and lower than all of them, to ensure that it's safe |
196 | * to acquire regardless of caller context. | |
197 | * | |
198 | * In addition, the following rules apply: | |
199 | * | |
200 | * (a) spa_props_lock protects pool properties, spa_config and spa_config_list. | |
201 | * The lock ordering is SCL_CONFIG > spa_props_lock. | |
202 | * | |
203 | * (b) I/O operations on leaf vdevs. For any zio operation that takes | |
204 | * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(), | |
205 | * or zio_write_phys() -- the caller must ensure that the config cannot | |
206 | * cannot change in the interim, and that the vdev cannot be reopened. | |
207 | * SCL_STATE as reader suffices for both. | |
34dc7c2f BB |
208 | * |
209 | * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). | |
210 | * | |
211 | * spa_vdev_enter() Acquire the namespace lock and the config lock | |
212 | * for writing. | |
213 | * | |
214 | * spa_vdev_exit() Release the config lock, wait for all I/O | |
215 | * to complete, sync the updated configs to the | |
216 | * cache, and release the namespace lock. | |
217 | * | |
b128c09f BB |
218 | * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit(). |
219 | * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual | |
220 | * locking is, always, based on spa_namespace_lock and spa_config_lock[]. | |
221 | * | |
9ae529ec | 222 | * spa_rename() is also implemented within this file since it requires |
b128c09f | 223 | * manipulation of the namespace. |
34dc7c2f BB |
224 | */ |
225 | ||
226 | static avl_tree_t spa_namespace_avl; | |
227 | kmutex_t spa_namespace_lock; | |
228 | static kcondvar_t spa_namespace_cv; | |
229 | static int spa_active_count; | |
230 | int spa_max_replication_override = SPA_DVAS_PER_BP; | |
231 | ||
232 | static kmutex_t spa_spare_lock; | |
233 | static avl_tree_t spa_spare_avl; | |
234 | static kmutex_t spa_l2cache_lock; | |
235 | static avl_tree_t spa_l2cache_avl; | |
236 | ||
237 | kmem_cache_t *spa_buffer_pool; | |
fb5f0bc8 | 238 | int spa_mode_global; |
34dc7c2f | 239 | |
cc92e9d0 | 240 | /* |
e8b96c60 MA |
241 | * Expiration time in milliseconds. This value has two meanings. First it is |
242 | * used to determine when the spa_deadman() logic should fire. By default the | |
243 | * spa_deadman() will fire if spa_sync() has not completed in 1000 seconds. | |
244 | * Secondly, the value determines if an I/O is considered "hung". Any I/O that | |
245 | * has not completed in zfs_deadman_synctime_ms is considered "hung" resulting | |
246 | * in a system panic. | |
cc92e9d0 | 247 | */ |
e8b96c60 | 248 | unsigned long zfs_deadman_synctime_ms = 1000000ULL; |
cc92e9d0 GW |
249 | |
250 | /* | |
251 | * By default the deadman is enabled. | |
252 | */ | |
253 | int zfs_deadman_enabled = 1; | |
254 | ||
e8b96c60 MA |
255 | /* |
256 | * The worst case is single-sector max-parity RAID-Z blocks, in which | |
257 | * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1) | |
258 | * times the size; so just assume that. Add to this the fact that | |
259 | * we can have up to 3 DVAs per bp, and one more factor of 2 because | |
260 | * the block may be dittoed with up to 3 DVAs by ddt_sync(). All together, | |
261 | * the worst case is: | |
262 | * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2 == 24 | |
263 | */ | |
264 | int spa_asize_inflation = 24; | |
265 | ||
34dc7c2f BB |
266 | /* |
267 | * ========================================================================== | |
268 | * SPA config locking | |
269 | * ========================================================================== | |
270 | */ | |
271 | static void | |
b128c09f BB |
272 | spa_config_lock_init(spa_t *spa) |
273 | { | |
d6320ddb BB |
274 | int i; |
275 | ||
276 | for (i = 0; i < SCL_LOCKS; i++) { | |
b128c09f BB |
277 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
278 | mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL); | |
279 | cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL); | |
13fe0198 | 280 | refcount_create_untracked(&scl->scl_count); |
b128c09f BB |
281 | scl->scl_writer = NULL; |
282 | scl->scl_write_wanted = 0; | |
283 | } | |
34dc7c2f BB |
284 | } |
285 | ||
286 | static void | |
b128c09f BB |
287 | spa_config_lock_destroy(spa_t *spa) |
288 | { | |
d6320ddb BB |
289 | int i; |
290 | ||
291 | for (i = 0; i < SCL_LOCKS; i++) { | |
b128c09f BB |
292 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
293 | mutex_destroy(&scl->scl_lock); | |
294 | cv_destroy(&scl->scl_cv); | |
295 | refcount_destroy(&scl->scl_count); | |
296 | ASSERT(scl->scl_writer == NULL); | |
297 | ASSERT(scl->scl_write_wanted == 0); | |
298 | } | |
299 | } | |
300 | ||
301 | int | |
302 | spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw) | |
34dc7c2f | 303 | { |
d6320ddb BB |
304 | int i; |
305 | ||
306 | for (i = 0; i < SCL_LOCKS; i++) { | |
b128c09f BB |
307 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
308 | if (!(locks & (1 << i))) | |
309 | continue; | |
310 | mutex_enter(&scl->scl_lock); | |
311 | if (rw == RW_READER) { | |
312 | if (scl->scl_writer || scl->scl_write_wanted) { | |
313 | mutex_exit(&scl->scl_lock); | |
314 | spa_config_exit(spa, locks ^ (1 << i), tag); | |
315 | return (0); | |
316 | } | |
317 | } else { | |
318 | ASSERT(scl->scl_writer != curthread); | |
319 | if (!refcount_is_zero(&scl->scl_count)) { | |
320 | mutex_exit(&scl->scl_lock); | |
321 | spa_config_exit(spa, locks ^ (1 << i), tag); | |
322 | return (0); | |
323 | } | |
324 | scl->scl_writer = curthread; | |
325 | } | |
326 | (void) refcount_add(&scl->scl_count, tag); | |
327 | mutex_exit(&scl->scl_lock); | |
328 | } | |
329 | return (1); | |
34dc7c2f BB |
330 | } |
331 | ||
332 | void | |
b128c09f | 333 | spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw) |
34dc7c2f | 334 | { |
45d1cae3 | 335 | int wlocks_held = 0; |
d6320ddb | 336 | int i; |
45d1cae3 | 337 | |
13fe0198 MA |
338 | ASSERT3U(SCL_LOCKS, <, sizeof (wlocks_held) * NBBY); |
339 | ||
d6320ddb | 340 | for (i = 0; i < SCL_LOCKS; i++) { |
b128c09f | 341 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
45d1cae3 BB |
342 | if (scl->scl_writer == curthread) |
343 | wlocks_held |= (1 << i); | |
b128c09f BB |
344 | if (!(locks & (1 << i))) |
345 | continue; | |
346 | mutex_enter(&scl->scl_lock); | |
347 | if (rw == RW_READER) { | |
348 | while (scl->scl_writer || scl->scl_write_wanted) { | |
349 | cv_wait(&scl->scl_cv, &scl->scl_lock); | |
350 | } | |
351 | } else { | |
352 | ASSERT(scl->scl_writer != curthread); | |
353 | while (!refcount_is_zero(&scl->scl_count)) { | |
354 | scl->scl_write_wanted++; | |
355 | cv_wait(&scl->scl_cv, &scl->scl_lock); | |
356 | scl->scl_write_wanted--; | |
357 | } | |
358 | scl->scl_writer = curthread; | |
359 | } | |
360 | (void) refcount_add(&scl->scl_count, tag); | |
361 | mutex_exit(&scl->scl_lock); | |
34dc7c2f | 362 | } |
45d1cae3 | 363 | ASSERT(wlocks_held <= locks); |
34dc7c2f BB |
364 | } |
365 | ||
366 | void | |
b128c09f | 367 | spa_config_exit(spa_t *spa, int locks, void *tag) |
34dc7c2f | 368 | { |
d6320ddb BB |
369 | int i; |
370 | ||
371 | for (i = SCL_LOCKS - 1; i >= 0; i--) { | |
b128c09f BB |
372 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
373 | if (!(locks & (1 << i))) | |
374 | continue; | |
375 | mutex_enter(&scl->scl_lock); | |
376 | ASSERT(!refcount_is_zero(&scl->scl_count)); | |
377 | if (refcount_remove(&scl->scl_count, tag) == 0) { | |
378 | ASSERT(scl->scl_writer == NULL || | |
379 | scl->scl_writer == curthread); | |
380 | scl->scl_writer = NULL; /* OK in either case */ | |
381 | cv_broadcast(&scl->scl_cv); | |
382 | } | |
383 | mutex_exit(&scl->scl_lock); | |
34dc7c2f | 384 | } |
34dc7c2f BB |
385 | } |
386 | ||
b128c09f BB |
387 | int |
388 | spa_config_held(spa_t *spa, int locks, krw_t rw) | |
34dc7c2f | 389 | { |
d6320ddb | 390 | int i, locks_held = 0; |
34dc7c2f | 391 | |
d6320ddb | 392 | for (i = 0; i < SCL_LOCKS; i++) { |
b128c09f BB |
393 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
394 | if (!(locks & (1 << i))) | |
395 | continue; | |
396 | if ((rw == RW_READER && !refcount_is_zero(&scl->scl_count)) || | |
397 | (rw == RW_WRITER && scl->scl_writer == curthread)) | |
398 | locks_held |= 1 << i; | |
399 | } | |
400 | ||
401 | return (locks_held); | |
34dc7c2f BB |
402 | } |
403 | ||
404 | /* | |
405 | * ========================================================================== | |
406 | * SPA namespace functions | |
407 | * ========================================================================== | |
408 | */ | |
409 | ||
410 | /* | |
411 | * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. | |
412 | * Returns NULL if no matching spa_t is found. | |
413 | */ | |
414 | spa_t * | |
415 | spa_lookup(const char *name) | |
416 | { | |
b128c09f BB |
417 | static spa_t search; /* spa_t is large; don't allocate on stack */ |
418 | spa_t *spa; | |
34dc7c2f | 419 | avl_index_t where; |
34dc7c2f BB |
420 | char *cp; |
421 | ||
422 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
423 | ||
13fe0198 MA |
424 | (void) strlcpy(search.spa_name, name, sizeof (search.spa_name)); |
425 | ||
34dc7c2f BB |
426 | /* |
427 | * If it's a full dataset name, figure out the pool name and | |
428 | * just use that. | |
429 | */ | |
13fe0198 MA |
430 | cp = strpbrk(search.spa_name, "/@"); |
431 | if (cp != NULL) | |
34dc7c2f | 432 | *cp = '\0'; |
34dc7c2f | 433 | |
34dc7c2f BB |
434 | spa = avl_find(&spa_namespace_avl, &search, &where); |
435 | ||
34dc7c2f BB |
436 | return (spa); |
437 | } | |
438 | ||
cc92e9d0 GW |
439 | /* |
440 | * Fires when spa_sync has not completed within zfs_deadman_synctime_ms. | |
441 | * If the zfs_deadman_enabled flag is set then it inspects all vdev queues | |
442 | * looking for potentially hung I/Os. | |
443 | */ | |
444 | void | |
445 | spa_deadman(void *arg) | |
446 | { | |
447 | spa_t *spa = arg; | |
448 | ||
449 | zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu", | |
450 | (gethrtime() - spa->spa_sync_starttime) / NANOSEC, | |
451 | ++spa->spa_deadman_calls); | |
452 | if (zfs_deadman_enabled) | |
453 | vdev_deadman(spa->spa_root_vdev); | |
454 | ||
455 | spa->spa_deadman_tqid = taskq_dispatch_delay(system_taskq, | |
cbfa294d | 456 | spa_deadman, spa, TQ_PUSHPAGE, ddi_get_lbolt() + |
cc92e9d0 GW |
457 | NSEC_TO_TICK(spa->spa_deadman_synctime)); |
458 | } | |
459 | ||
34dc7c2f BB |
460 | /* |
461 | * Create an uninitialized spa_t with the given name. Requires | |
462 | * spa_namespace_lock. The caller must ensure that the spa_t doesn't already | |
463 | * exist by calling spa_lookup() first. | |
464 | */ | |
465 | spa_t * | |
428870ff | 466 | spa_add(const char *name, nvlist_t *config, const char *altroot) |
34dc7c2f BB |
467 | { |
468 | spa_t *spa; | |
b128c09f | 469 | spa_config_dirent_t *dp; |
d6320ddb | 470 | int t; |
34dc7c2f BB |
471 | |
472 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
473 | ||
b8d06fca | 474 | spa = kmem_zalloc(sizeof (spa_t), KM_PUSHPAGE | KM_NODEBUG); |
34dc7c2f | 475 | |
34dc7c2f | 476 | mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 477 | mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); |
428870ff | 478 | mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 479 | mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); |
428870ff | 480 | mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 481 | mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); |
428870ff BB |
482 | mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); |
483 | mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL); | |
484 | mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL); | |
34dc7c2f BB |
485 | |
486 | cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); | |
428870ff | 487 | cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL); |
34dc7c2f | 488 | cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); |
b128c09f | 489 | cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL); |
34dc7c2f | 490 | |
d6320ddb | 491 | for (t = 0; t < TXG_SIZE; t++) |
428870ff BB |
492 | bplist_create(&spa->spa_free_bplist[t]); |
493 | ||
b128c09f | 494 | (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name)); |
34dc7c2f BB |
495 | spa->spa_state = POOL_STATE_UNINITIALIZED; |
496 | spa->spa_freeze_txg = UINT64_MAX; | |
497 | spa->spa_final_txg = UINT64_MAX; | |
428870ff BB |
498 | spa->spa_load_max_txg = UINT64_MAX; |
499 | spa->spa_proc = &p0; | |
500 | spa->spa_proc_state = SPA_PROC_NONE; | |
34dc7c2f | 501 | |
e8b96c60 | 502 | spa->spa_deadman_synctime = MSEC2NSEC(zfs_deadman_synctime_ms); |
cc92e9d0 | 503 | |
34dc7c2f | 504 | refcount_create(&spa->spa_refcount); |
b128c09f | 505 | spa_config_lock_init(spa); |
1421c891 | 506 | spa_stats_init(spa); |
34dc7c2f BB |
507 | |
508 | avl_add(&spa_namespace_avl, spa); | |
509 | ||
34dc7c2f BB |
510 | /* |
511 | * Set the alternate root, if there is one. | |
512 | */ | |
513 | if (altroot) { | |
514 | spa->spa_root = spa_strdup(altroot); | |
515 | spa_active_count++; | |
516 | } | |
517 | ||
b128c09f BB |
518 | /* |
519 | * Every pool starts with the default cachefile | |
520 | */ | |
521 | list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t), | |
522 | offsetof(spa_config_dirent_t, scd_link)); | |
523 | ||
b8d06fca | 524 | dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_PUSHPAGE); |
428870ff | 525 | dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path); |
b128c09f BB |
526 | list_insert_head(&spa->spa_config_list, dp); |
527 | ||
572e2857 | 528 | VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME, |
b8d06fca | 529 | KM_PUSHPAGE) == 0); |
572e2857 | 530 | |
9ae529ec CS |
531 | if (config != NULL) { |
532 | nvlist_t *features; | |
533 | ||
534 | if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ, | |
535 | &features) == 0) { | |
536 | VERIFY(nvlist_dup(features, &spa->spa_label_features, | |
537 | 0) == 0); | |
538 | } | |
539 | ||
428870ff | 540 | VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); |
9ae529ec CS |
541 | } |
542 | ||
543 | if (spa->spa_label_features == NULL) { | |
544 | VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME, | |
545 | KM_SLEEP) == 0); | |
546 | } | |
428870ff | 547 | |
13fe0198 MA |
548 | spa->spa_debug = ((zfs_flags & ZFS_DEBUG_SPA) != 0); |
549 | ||
34dc7c2f BB |
550 | return (spa); |
551 | } | |
552 | ||
553 | /* | |
554 | * Removes a spa_t from the namespace, freeing up any memory used. Requires | |
555 | * spa_namespace_lock. This is called only after the spa_t has been closed and | |
556 | * deactivated. | |
557 | */ | |
558 | void | |
559 | spa_remove(spa_t *spa) | |
560 | { | |
b128c09f | 561 | spa_config_dirent_t *dp; |
d6320ddb | 562 | int t; |
b128c09f | 563 | |
34dc7c2f BB |
564 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); |
565 | ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); | |
34dc7c2f | 566 | |
428870ff BB |
567 | nvlist_free(spa->spa_config_splitting); |
568 | ||
34dc7c2f BB |
569 | avl_remove(&spa_namespace_avl, spa); |
570 | cv_broadcast(&spa_namespace_cv); | |
571 | ||
572 | if (spa->spa_root) { | |
573 | spa_strfree(spa->spa_root); | |
574 | spa_active_count--; | |
575 | } | |
576 | ||
b128c09f BB |
577 | while ((dp = list_head(&spa->spa_config_list)) != NULL) { |
578 | list_remove(&spa->spa_config_list, dp); | |
579 | if (dp->scd_path != NULL) | |
580 | spa_strfree(dp->scd_path); | |
581 | kmem_free(dp, sizeof (spa_config_dirent_t)); | |
582 | } | |
34dc7c2f | 583 | |
b128c09f | 584 | list_destroy(&spa->spa_config_list); |
34dc7c2f | 585 | |
9ae529ec | 586 | nvlist_free(spa->spa_label_features); |
572e2857 | 587 | nvlist_free(spa->spa_load_info); |
34dc7c2f BB |
588 | spa_config_set(spa, NULL); |
589 | ||
590 | refcount_destroy(&spa->spa_refcount); | |
591 | ||
1421c891 | 592 | spa_stats_destroy(spa); |
b128c09f | 593 | spa_config_lock_destroy(spa); |
34dc7c2f | 594 | |
d6320ddb | 595 | for (t = 0; t < TXG_SIZE; t++) |
428870ff BB |
596 | bplist_destroy(&spa->spa_free_bplist[t]); |
597 | ||
34dc7c2f | 598 | cv_destroy(&spa->spa_async_cv); |
428870ff | 599 | cv_destroy(&spa->spa_proc_cv); |
34dc7c2f | 600 | cv_destroy(&spa->spa_scrub_io_cv); |
b128c09f | 601 | cv_destroy(&spa->spa_suspend_cv); |
34dc7c2f | 602 | |
34dc7c2f | 603 | mutex_destroy(&spa->spa_async_lock); |
34dc7c2f | 604 | mutex_destroy(&spa->spa_errlist_lock); |
428870ff | 605 | mutex_destroy(&spa->spa_errlog_lock); |
34dc7c2f | 606 | mutex_destroy(&spa->spa_history_lock); |
428870ff | 607 | mutex_destroy(&spa->spa_proc_lock); |
34dc7c2f | 608 | mutex_destroy(&spa->spa_props_lock); |
428870ff | 609 | mutex_destroy(&spa->spa_scrub_lock); |
b128c09f | 610 | mutex_destroy(&spa->spa_suspend_lock); |
428870ff | 611 | mutex_destroy(&spa->spa_vdev_top_lock); |
34dc7c2f BB |
612 | |
613 | kmem_free(spa, sizeof (spa_t)); | |
614 | } | |
615 | ||
616 | /* | |
617 | * Given a pool, return the next pool in the namespace, or NULL if there is | |
618 | * none. If 'prev' is NULL, return the first pool. | |
619 | */ | |
620 | spa_t * | |
621 | spa_next(spa_t *prev) | |
622 | { | |
623 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
624 | ||
625 | if (prev) | |
626 | return (AVL_NEXT(&spa_namespace_avl, prev)); | |
627 | else | |
628 | return (avl_first(&spa_namespace_avl)); | |
629 | } | |
630 | ||
631 | /* | |
632 | * ========================================================================== | |
633 | * SPA refcount functions | |
634 | * ========================================================================== | |
635 | */ | |
636 | ||
637 | /* | |
638 | * Add a reference to the given spa_t. Must have at least one reference, or | |
639 | * have the namespace lock held. | |
640 | */ | |
641 | void | |
642 | spa_open_ref(spa_t *spa, void *tag) | |
643 | { | |
b128c09f | 644 | ASSERT(refcount_count(&spa->spa_refcount) >= spa->spa_minref || |
34dc7c2f | 645 | MUTEX_HELD(&spa_namespace_lock)); |
34dc7c2f BB |
646 | (void) refcount_add(&spa->spa_refcount, tag); |
647 | } | |
648 | ||
649 | /* | |
650 | * Remove a reference to the given spa_t. Must have at least one reference, or | |
651 | * have the namespace lock held. | |
652 | */ | |
653 | void | |
654 | spa_close(spa_t *spa, void *tag) | |
655 | { | |
b128c09f | 656 | ASSERT(refcount_count(&spa->spa_refcount) > spa->spa_minref || |
34dc7c2f | 657 | MUTEX_HELD(&spa_namespace_lock)); |
34dc7c2f BB |
658 | (void) refcount_remove(&spa->spa_refcount, tag); |
659 | } | |
660 | ||
661 | /* | |
662 | * Check to see if the spa refcount is zero. Must be called with | |
b128c09f | 663 | * spa_namespace_lock held. We really compare against spa_minref, which is the |
34dc7c2f BB |
664 | * number of references acquired when opening a pool |
665 | */ | |
666 | boolean_t | |
667 | spa_refcount_zero(spa_t *spa) | |
668 | { | |
669 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
670 | ||
b128c09f | 671 | return (refcount_count(&spa->spa_refcount) == spa->spa_minref); |
34dc7c2f BB |
672 | } |
673 | ||
674 | /* | |
675 | * ========================================================================== | |
676 | * SPA spare and l2cache tracking | |
677 | * ========================================================================== | |
678 | */ | |
679 | ||
680 | /* | |
681 | * Hot spares and cache devices are tracked using the same code below, | |
682 | * for 'auxiliary' devices. | |
683 | */ | |
684 | ||
685 | typedef struct spa_aux { | |
686 | uint64_t aux_guid; | |
687 | uint64_t aux_pool; | |
688 | avl_node_t aux_avl; | |
689 | int aux_count; | |
690 | } spa_aux_t; | |
691 | ||
692 | static int | |
693 | spa_aux_compare(const void *a, const void *b) | |
694 | { | |
695 | const spa_aux_t *sa = a; | |
696 | const spa_aux_t *sb = b; | |
697 | ||
698 | if (sa->aux_guid < sb->aux_guid) | |
699 | return (-1); | |
700 | else if (sa->aux_guid > sb->aux_guid) | |
701 | return (1); | |
702 | else | |
703 | return (0); | |
704 | } | |
705 | ||
706 | void | |
707 | spa_aux_add(vdev_t *vd, avl_tree_t *avl) | |
708 | { | |
709 | avl_index_t where; | |
710 | spa_aux_t search; | |
711 | spa_aux_t *aux; | |
712 | ||
713 | search.aux_guid = vd->vdev_guid; | |
714 | if ((aux = avl_find(avl, &search, &where)) != NULL) { | |
715 | aux->aux_count++; | |
716 | } else { | |
b8d06fca | 717 | aux = kmem_zalloc(sizeof (spa_aux_t), KM_PUSHPAGE); |
34dc7c2f BB |
718 | aux->aux_guid = vd->vdev_guid; |
719 | aux->aux_count = 1; | |
720 | avl_insert(avl, aux, where); | |
721 | } | |
722 | } | |
723 | ||
724 | void | |
725 | spa_aux_remove(vdev_t *vd, avl_tree_t *avl) | |
726 | { | |
727 | spa_aux_t search; | |
728 | spa_aux_t *aux; | |
729 | avl_index_t where; | |
730 | ||
731 | search.aux_guid = vd->vdev_guid; | |
732 | aux = avl_find(avl, &search, &where); | |
733 | ||
734 | ASSERT(aux != NULL); | |
735 | ||
736 | if (--aux->aux_count == 0) { | |
737 | avl_remove(avl, aux); | |
738 | kmem_free(aux, sizeof (spa_aux_t)); | |
739 | } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { | |
740 | aux->aux_pool = 0ULL; | |
741 | } | |
742 | } | |
743 | ||
744 | boolean_t | |
b128c09f | 745 | spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl) |
34dc7c2f BB |
746 | { |
747 | spa_aux_t search, *found; | |
34dc7c2f BB |
748 | |
749 | search.aux_guid = guid; | |
b128c09f | 750 | found = avl_find(avl, &search, NULL); |
34dc7c2f BB |
751 | |
752 | if (pool) { | |
753 | if (found) | |
754 | *pool = found->aux_pool; | |
755 | else | |
756 | *pool = 0ULL; | |
757 | } | |
758 | ||
b128c09f BB |
759 | if (refcnt) { |
760 | if (found) | |
761 | *refcnt = found->aux_count; | |
762 | else | |
763 | *refcnt = 0; | |
764 | } | |
765 | ||
34dc7c2f BB |
766 | return (found != NULL); |
767 | } | |
768 | ||
769 | void | |
770 | spa_aux_activate(vdev_t *vd, avl_tree_t *avl) | |
771 | { | |
772 | spa_aux_t search, *found; | |
773 | avl_index_t where; | |
774 | ||
775 | search.aux_guid = vd->vdev_guid; | |
776 | found = avl_find(avl, &search, &where); | |
777 | ASSERT(found != NULL); | |
778 | ASSERT(found->aux_pool == 0ULL); | |
779 | ||
780 | found->aux_pool = spa_guid(vd->vdev_spa); | |
781 | } | |
782 | ||
783 | /* | |
784 | * Spares are tracked globally due to the following constraints: | |
785 | * | |
786 | * - A spare may be part of multiple pools. | |
787 | * - A spare may be added to a pool even if it's actively in use within | |
788 | * another pool. | |
789 | * - A spare in use in any pool can only be the source of a replacement if | |
790 | * the target is a spare in the same pool. | |
791 | * | |
792 | * We keep track of all spares on the system through the use of a reference | |
793 | * counted AVL tree. When a vdev is added as a spare, or used as a replacement | |
794 | * spare, then we bump the reference count in the AVL tree. In addition, we set | |
795 | * the 'vdev_isspare' member to indicate that the device is a spare (active or | |
796 | * inactive). When a spare is made active (used to replace a device in the | |
797 | * pool), we also keep track of which pool its been made a part of. | |
798 | * | |
799 | * The 'spa_spare_lock' protects the AVL tree. These functions are normally | |
800 | * called under the spa_namespace lock as part of vdev reconfiguration. The | |
801 | * separate spare lock exists for the status query path, which does not need to | |
802 | * be completely consistent with respect to other vdev configuration changes. | |
803 | */ | |
804 | ||
805 | static int | |
806 | spa_spare_compare(const void *a, const void *b) | |
807 | { | |
808 | return (spa_aux_compare(a, b)); | |
809 | } | |
810 | ||
811 | void | |
812 | spa_spare_add(vdev_t *vd) | |
813 | { | |
814 | mutex_enter(&spa_spare_lock); | |
815 | ASSERT(!vd->vdev_isspare); | |
816 | spa_aux_add(vd, &spa_spare_avl); | |
817 | vd->vdev_isspare = B_TRUE; | |
818 | mutex_exit(&spa_spare_lock); | |
819 | } | |
820 | ||
821 | void | |
822 | spa_spare_remove(vdev_t *vd) | |
823 | { | |
824 | mutex_enter(&spa_spare_lock); | |
825 | ASSERT(vd->vdev_isspare); | |
826 | spa_aux_remove(vd, &spa_spare_avl); | |
827 | vd->vdev_isspare = B_FALSE; | |
828 | mutex_exit(&spa_spare_lock); | |
829 | } | |
830 | ||
831 | boolean_t | |
b128c09f | 832 | spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt) |
34dc7c2f BB |
833 | { |
834 | boolean_t found; | |
835 | ||
836 | mutex_enter(&spa_spare_lock); | |
b128c09f | 837 | found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl); |
34dc7c2f BB |
838 | mutex_exit(&spa_spare_lock); |
839 | ||
840 | return (found); | |
841 | } | |
842 | ||
843 | void | |
844 | spa_spare_activate(vdev_t *vd) | |
845 | { | |
846 | mutex_enter(&spa_spare_lock); | |
847 | ASSERT(vd->vdev_isspare); | |
848 | spa_aux_activate(vd, &spa_spare_avl); | |
849 | mutex_exit(&spa_spare_lock); | |
850 | } | |
851 | ||
852 | /* | |
853 | * Level 2 ARC devices are tracked globally for the same reasons as spares. | |
854 | * Cache devices currently only support one pool per cache device, and so | |
855 | * for these devices the aux reference count is currently unused beyond 1. | |
856 | */ | |
857 | ||
858 | static int | |
859 | spa_l2cache_compare(const void *a, const void *b) | |
860 | { | |
861 | return (spa_aux_compare(a, b)); | |
862 | } | |
863 | ||
864 | void | |
865 | spa_l2cache_add(vdev_t *vd) | |
866 | { | |
867 | mutex_enter(&spa_l2cache_lock); | |
868 | ASSERT(!vd->vdev_isl2cache); | |
869 | spa_aux_add(vd, &spa_l2cache_avl); | |
870 | vd->vdev_isl2cache = B_TRUE; | |
871 | mutex_exit(&spa_l2cache_lock); | |
872 | } | |
873 | ||
874 | void | |
875 | spa_l2cache_remove(vdev_t *vd) | |
876 | { | |
877 | mutex_enter(&spa_l2cache_lock); | |
878 | ASSERT(vd->vdev_isl2cache); | |
879 | spa_aux_remove(vd, &spa_l2cache_avl); | |
880 | vd->vdev_isl2cache = B_FALSE; | |
881 | mutex_exit(&spa_l2cache_lock); | |
882 | } | |
883 | ||
884 | boolean_t | |
885 | spa_l2cache_exists(uint64_t guid, uint64_t *pool) | |
886 | { | |
887 | boolean_t found; | |
888 | ||
889 | mutex_enter(&spa_l2cache_lock); | |
b128c09f | 890 | found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl); |
34dc7c2f BB |
891 | mutex_exit(&spa_l2cache_lock); |
892 | ||
893 | return (found); | |
894 | } | |
895 | ||
896 | void | |
897 | spa_l2cache_activate(vdev_t *vd) | |
898 | { | |
899 | mutex_enter(&spa_l2cache_lock); | |
900 | ASSERT(vd->vdev_isl2cache); | |
901 | spa_aux_activate(vd, &spa_l2cache_avl); | |
902 | mutex_exit(&spa_l2cache_lock); | |
903 | } | |
904 | ||
34dc7c2f BB |
905 | /* |
906 | * ========================================================================== | |
907 | * SPA vdev locking | |
908 | * ========================================================================== | |
909 | */ | |
910 | ||
911 | /* | |
912 | * Lock the given spa_t for the purpose of adding or removing a vdev. | |
913 | * Grabs the global spa_namespace_lock plus the spa config lock for writing. | |
914 | * It returns the next transaction group for the spa_t. | |
915 | */ | |
916 | uint64_t | |
917 | spa_vdev_enter(spa_t *spa) | |
918 | { | |
428870ff | 919 | mutex_enter(&spa->spa_vdev_top_lock); |
34dc7c2f | 920 | mutex_enter(&spa_namespace_lock); |
428870ff BB |
921 | return (spa_vdev_config_enter(spa)); |
922 | } | |
923 | ||
924 | /* | |
925 | * Internal implementation for spa_vdev_enter(). Used when a vdev | |
926 | * operation requires multiple syncs (i.e. removing a device) while | |
927 | * keeping the spa_namespace_lock held. | |
928 | */ | |
929 | uint64_t | |
930 | spa_vdev_config_enter(spa_t *spa) | |
931 | { | |
932 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
34dc7c2f | 933 | |
b128c09f | 934 | spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); |
34dc7c2f BB |
935 | |
936 | return (spa_last_synced_txg(spa) + 1); | |
937 | } | |
938 | ||
939 | /* | |
428870ff BB |
940 | * Used in combination with spa_vdev_config_enter() to allow the syncing |
941 | * of multiple transactions without releasing the spa_namespace_lock. | |
34dc7c2f | 942 | */ |
428870ff BB |
943 | void |
944 | spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag) | |
34dc7c2f BB |
945 | { |
946 | int config_changed = B_FALSE; | |
947 | ||
d6320ddb | 948 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); |
34dc7c2f BB |
949 | ASSERT(txg > spa_last_synced_txg(spa)); |
950 | ||
b128c09f BB |
951 | spa->spa_pending_vdev = NULL; |
952 | ||
34dc7c2f BB |
953 | /* |
954 | * Reassess the DTLs. | |
955 | */ | |
956 | vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); | |
957 | ||
b128c09f | 958 | if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) { |
34dc7c2f | 959 | config_changed = B_TRUE; |
428870ff | 960 | spa->spa_config_generation++; |
34dc7c2f BB |
961 | } |
962 | ||
428870ff BB |
963 | /* |
964 | * Verify the metaslab classes. | |
965 | */ | |
966 | ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0); | |
967 | ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0); | |
968 | ||
b128c09f | 969 | spa_config_exit(spa, SCL_ALL, spa); |
34dc7c2f | 970 | |
428870ff BB |
971 | /* |
972 | * Panic the system if the specified tag requires it. This | |
973 | * is useful for ensuring that configurations are updated | |
974 | * transactionally. | |
975 | */ | |
976 | if (zio_injection_enabled) | |
977 | zio_handle_panic_injection(spa, tag, 0); | |
978 | ||
34dc7c2f BB |
979 | /* |
980 | * Note: this txg_wait_synced() is important because it ensures | |
981 | * that there won't be more than one config change per txg. | |
982 | * This allows us to use the txg as the generation number. | |
983 | */ | |
984 | if (error == 0) | |
985 | txg_wait_synced(spa->spa_dsl_pool, txg); | |
986 | ||
987 | if (vd != NULL) { | |
fb5f0bc8 BB |
988 | ASSERT(!vd->vdev_detached || vd->vdev_dtl_smo.smo_object == 0); |
989 | spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); | |
34dc7c2f | 990 | vdev_free(vd); |
fb5f0bc8 | 991 | spa_config_exit(spa, SCL_ALL, spa); |
34dc7c2f BB |
992 | } |
993 | ||
994 | /* | |
995 | * If the config changed, update the config cache. | |
996 | */ | |
997 | if (config_changed) | |
b128c09f | 998 | spa_config_sync(spa, B_FALSE, B_TRUE); |
428870ff | 999 | } |
34dc7c2f | 1000 | |
428870ff BB |
1001 | /* |
1002 | * Unlock the spa_t after adding or removing a vdev. Besides undoing the | |
1003 | * locking of spa_vdev_enter(), we also want make sure the transactions have | |
1004 | * synced to disk, and then update the global configuration cache with the new | |
1005 | * information. | |
1006 | */ | |
1007 | int | |
1008 | spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) | |
1009 | { | |
1010 | spa_vdev_config_exit(spa, vd, txg, error, FTAG); | |
34dc7c2f | 1011 | mutex_exit(&spa_namespace_lock); |
428870ff | 1012 | mutex_exit(&spa->spa_vdev_top_lock); |
34dc7c2f BB |
1013 | |
1014 | return (error); | |
1015 | } | |
1016 | ||
b128c09f BB |
1017 | /* |
1018 | * Lock the given spa_t for the purpose of changing vdev state. | |
1019 | */ | |
1020 | void | |
428870ff | 1021 | spa_vdev_state_enter(spa_t *spa, int oplocks) |
b128c09f | 1022 | { |
428870ff BB |
1023 | int locks = SCL_STATE_ALL | oplocks; |
1024 | ||
1025 | /* | |
1026 | * Root pools may need to read of the underlying devfs filesystem | |
1027 | * when opening up a vdev. Unfortunately if we're holding the | |
1028 | * SCL_ZIO lock it will result in a deadlock when we try to issue | |
1029 | * the read from the root filesystem. Instead we "prefetch" | |
1030 | * the associated vnodes that we need prior to opening the | |
1031 | * underlying devices and cache them so that we can prevent | |
1032 | * any I/O when we are doing the actual open. | |
1033 | */ | |
1034 | if (spa_is_root(spa)) { | |
1035 | int low = locks & ~(SCL_ZIO - 1); | |
1036 | int high = locks & ~low; | |
1037 | ||
1038 | spa_config_enter(spa, high, spa, RW_WRITER); | |
1039 | vdev_hold(spa->spa_root_vdev); | |
1040 | spa_config_enter(spa, low, spa, RW_WRITER); | |
1041 | } else { | |
1042 | spa_config_enter(spa, locks, spa, RW_WRITER); | |
1043 | } | |
1044 | spa->spa_vdev_locks = locks; | |
b128c09f BB |
1045 | } |
1046 | ||
1047 | int | |
1048 | spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error) | |
1049 | { | |
428870ff BB |
1050 | boolean_t config_changed = B_FALSE; |
1051 | ||
1052 | if (vd != NULL || error == 0) | |
1053 | vdev_dtl_reassess(vd ? vd->vdev_top : spa->spa_root_vdev, | |
1054 | 0, 0, B_FALSE); | |
1055 | ||
1056 | if (vd != NULL) { | |
b128c09f | 1057 | vdev_state_dirty(vd->vdev_top); |
428870ff BB |
1058 | config_changed = B_TRUE; |
1059 | spa->spa_config_generation++; | |
1060 | } | |
b128c09f | 1061 | |
428870ff BB |
1062 | if (spa_is_root(spa)) |
1063 | vdev_rele(spa->spa_root_vdev); | |
1064 | ||
1065 | ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL); | |
1066 | spa_config_exit(spa, spa->spa_vdev_locks, spa); | |
b128c09f | 1067 | |
fb5f0bc8 BB |
1068 | /* |
1069 | * If anything changed, wait for it to sync. This ensures that, | |
1070 | * from the system administrator's perspective, zpool(1M) commands | |
1071 | * are synchronous. This is important for things like zpool offline: | |
1072 | * when the command completes, you expect no further I/O from ZFS. | |
1073 | */ | |
1074 | if (vd != NULL) | |
1075 | txg_wait_synced(spa->spa_dsl_pool, 0); | |
1076 | ||
428870ff BB |
1077 | /* |
1078 | * If the config changed, update the config cache. | |
1079 | */ | |
1080 | if (config_changed) { | |
1081 | mutex_enter(&spa_namespace_lock); | |
1082 | spa_config_sync(spa, B_FALSE, B_TRUE); | |
1083 | mutex_exit(&spa_namespace_lock); | |
1084 | } | |
1085 | ||
b128c09f BB |
1086 | return (error); |
1087 | } | |
1088 | ||
34dc7c2f BB |
1089 | /* |
1090 | * ========================================================================== | |
1091 | * Miscellaneous functions | |
1092 | * ========================================================================== | |
1093 | */ | |
1094 | ||
9ae529ec CS |
1095 | void |
1096 | spa_activate_mos_feature(spa_t *spa, const char *feature) | |
1097 | { | |
1098 | (void) nvlist_add_boolean(spa->spa_label_features, feature); | |
1099 | vdev_config_dirty(spa->spa_root_vdev); | |
1100 | } | |
1101 | ||
1102 | void | |
1103 | spa_deactivate_mos_feature(spa_t *spa, const char *feature) | |
1104 | { | |
1105 | (void) nvlist_remove_all(spa->spa_label_features, feature); | |
1106 | vdev_config_dirty(spa->spa_root_vdev); | |
1107 | } | |
1108 | ||
34dc7c2f BB |
1109 | /* |
1110 | * Rename a spa_t. | |
1111 | */ | |
1112 | int | |
1113 | spa_rename(const char *name, const char *newname) | |
1114 | { | |
1115 | spa_t *spa; | |
1116 | int err; | |
1117 | ||
1118 | /* | |
1119 | * Lookup the spa_t and grab the config lock for writing. We need to | |
1120 | * actually open the pool so that we can sync out the necessary labels. | |
1121 | * It's OK to call spa_open() with the namespace lock held because we | |
1122 | * allow recursive calls for other reasons. | |
1123 | */ | |
1124 | mutex_enter(&spa_namespace_lock); | |
1125 | if ((err = spa_open(name, &spa, FTAG)) != 0) { | |
1126 | mutex_exit(&spa_namespace_lock); | |
1127 | return (err); | |
1128 | } | |
1129 | ||
b128c09f | 1130 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f BB |
1131 | |
1132 | avl_remove(&spa_namespace_avl, spa); | |
b128c09f | 1133 | (void) strlcpy(spa->spa_name, newname, sizeof (spa->spa_name)); |
34dc7c2f BB |
1134 | avl_add(&spa_namespace_avl, spa); |
1135 | ||
1136 | /* | |
1137 | * Sync all labels to disk with the new names by marking the root vdev | |
1138 | * dirty and waiting for it to sync. It will pick up the new pool name | |
1139 | * during the sync. | |
1140 | */ | |
1141 | vdev_config_dirty(spa->spa_root_vdev); | |
1142 | ||
b128c09f | 1143 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
1144 | |
1145 | txg_wait_synced(spa->spa_dsl_pool, 0); | |
1146 | ||
1147 | /* | |
1148 | * Sync the updated config cache. | |
1149 | */ | |
b128c09f | 1150 | spa_config_sync(spa, B_FALSE, B_TRUE); |
34dc7c2f BB |
1151 | |
1152 | spa_close(spa, FTAG); | |
1153 | ||
1154 | mutex_exit(&spa_namespace_lock); | |
1155 | ||
1156 | return (0); | |
1157 | } | |
1158 | ||
34dc7c2f | 1159 | /* |
572e2857 BB |
1160 | * Return the spa_t associated with given pool_guid, if it exists. If |
1161 | * device_guid is non-zero, determine whether the pool exists *and* contains | |
1162 | * a device with the specified device_guid. | |
34dc7c2f | 1163 | */ |
572e2857 BB |
1164 | spa_t * |
1165 | spa_by_guid(uint64_t pool_guid, uint64_t device_guid) | |
34dc7c2f BB |
1166 | { |
1167 | spa_t *spa; | |
1168 | avl_tree_t *t = &spa_namespace_avl; | |
1169 | ||
1170 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
1171 | ||
1172 | for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { | |
1173 | if (spa->spa_state == POOL_STATE_UNINITIALIZED) | |
1174 | continue; | |
1175 | if (spa->spa_root_vdev == NULL) | |
1176 | continue; | |
1177 | if (spa_guid(spa) == pool_guid) { | |
1178 | if (device_guid == 0) | |
1179 | break; | |
1180 | ||
1181 | if (vdev_lookup_by_guid(spa->spa_root_vdev, | |
1182 | device_guid) != NULL) | |
1183 | break; | |
1184 | ||
1185 | /* | |
1186 | * Check any devices we may be in the process of adding. | |
1187 | */ | |
1188 | if (spa->spa_pending_vdev) { | |
1189 | if (vdev_lookup_by_guid(spa->spa_pending_vdev, | |
1190 | device_guid) != NULL) | |
1191 | break; | |
1192 | } | |
1193 | } | |
1194 | } | |
1195 | ||
572e2857 BB |
1196 | return (spa); |
1197 | } | |
1198 | ||
1199 | /* | |
1200 | * Determine whether a pool with the given pool_guid exists. | |
1201 | */ | |
1202 | boolean_t | |
1203 | spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) | |
1204 | { | |
1205 | return (spa_by_guid(pool_guid, device_guid) != NULL); | |
34dc7c2f BB |
1206 | } |
1207 | ||
1208 | char * | |
1209 | spa_strdup(const char *s) | |
1210 | { | |
1211 | size_t len; | |
1212 | char *new; | |
1213 | ||
1214 | len = strlen(s); | |
b8d06fca | 1215 | new = kmem_alloc(len + 1, KM_PUSHPAGE); |
34dc7c2f BB |
1216 | bcopy(s, new, len); |
1217 | new[len] = '\0'; | |
1218 | ||
1219 | return (new); | |
1220 | } | |
1221 | ||
1222 | void | |
1223 | spa_strfree(char *s) | |
1224 | { | |
1225 | kmem_free(s, strlen(s) + 1); | |
1226 | } | |
1227 | ||
1228 | uint64_t | |
1229 | spa_get_random(uint64_t range) | |
1230 | { | |
1231 | uint64_t r; | |
1232 | ||
1233 | ASSERT(range != 0); | |
1234 | ||
1235 | (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); | |
1236 | ||
1237 | return (r % range); | |
1238 | } | |
1239 | ||
428870ff BB |
1240 | uint64_t |
1241 | spa_generate_guid(spa_t *spa) | |
34dc7c2f | 1242 | { |
428870ff | 1243 | uint64_t guid = spa_get_random(-1ULL); |
34dc7c2f | 1244 | |
428870ff BB |
1245 | if (spa != NULL) { |
1246 | while (guid == 0 || spa_guid_exists(spa_guid(spa), guid)) | |
1247 | guid = spa_get_random(-1ULL); | |
1248 | } else { | |
1249 | while (guid == 0 || spa_guid_exists(guid, 0)) | |
1250 | guid = spa_get_random(-1ULL); | |
34dc7c2f BB |
1251 | } |
1252 | ||
428870ff BB |
1253 | return (guid); |
1254 | } | |
1255 | ||
1256 | void | |
1257 | sprintf_blkptr(char *buf, const blkptr_t *bp) | |
1258 | { | |
9ae529ec | 1259 | char type[256]; |
428870ff BB |
1260 | char *checksum = NULL; |
1261 | char *compress = NULL; | |
34dc7c2f | 1262 | |
428870ff | 1263 | if (bp != NULL) { |
9ae529ec CS |
1264 | if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) { |
1265 | dmu_object_byteswap_t bswap = | |
1266 | DMU_OT_BYTESWAP(BP_GET_TYPE(bp)); | |
1267 | (void) snprintf(type, sizeof (type), "bswap %s %s", | |
1268 | DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ? | |
1269 | "metadata" : "data", | |
1270 | dmu_ot_byteswap[bswap].ob_name); | |
1271 | } else { | |
1272 | (void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name, | |
1273 | sizeof (type)); | |
1274 | } | |
428870ff BB |
1275 | checksum = zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name; |
1276 | compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name; | |
34dc7c2f BB |
1277 | } |
1278 | ||
428870ff | 1279 | SPRINTF_BLKPTR(snprintf, ' ', buf, bp, type, checksum, compress); |
34dc7c2f BB |
1280 | } |
1281 | ||
1282 | void | |
1283 | spa_freeze(spa_t *spa) | |
1284 | { | |
1285 | uint64_t freeze_txg = 0; | |
1286 | ||
b128c09f | 1287 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f BB |
1288 | if (spa->spa_freeze_txg == UINT64_MAX) { |
1289 | freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; | |
1290 | spa->spa_freeze_txg = freeze_txg; | |
1291 | } | |
b128c09f | 1292 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
1293 | if (freeze_txg != 0) |
1294 | txg_wait_synced(spa_get_dsl(spa), freeze_txg); | |
1295 | } | |
1296 | ||
428870ff BB |
1297 | /* |
1298 | * This is a stripped-down version of strtoull, suitable only for converting | |
d3cc8b15 | 1299 | * lowercase hexadecimal numbers that don't overflow. |
428870ff BB |
1300 | */ |
1301 | uint64_t | |
1302 | strtonum(const char *str, char **nptr) | |
1303 | { | |
1304 | uint64_t val = 0; | |
1305 | char c; | |
1306 | int digit; | |
1307 | ||
1308 | while ((c = *str) != '\0') { | |
1309 | if (c >= '0' && c <= '9') | |
1310 | digit = c - '0'; | |
1311 | else if (c >= 'a' && c <= 'f') | |
1312 | digit = 10 + c - 'a'; | |
1313 | else | |
1314 | break; | |
1315 | ||
1316 | val *= 16; | |
1317 | val += digit; | |
1318 | ||
1319 | str++; | |
1320 | } | |
1321 | ||
1322 | if (nptr) | |
1323 | *nptr = (char *)str; | |
1324 | ||
1325 | return (val); | |
1326 | } | |
1327 | ||
34dc7c2f BB |
1328 | /* |
1329 | * ========================================================================== | |
1330 | * Accessor functions | |
1331 | * ========================================================================== | |
1332 | */ | |
1333 | ||
b128c09f BB |
1334 | boolean_t |
1335 | spa_shutting_down(spa_t *spa) | |
34dc7c2f | 1336 | { |
b128c09f | 1337 | return (spa->spa_async_suspended); |
34dc7c2f BB |
1338 | } |
1339 | ||
1340 | dsl_pool_t * | |
1341 | spa_get_dsl(spa_t *spa) | |
1342 | { | |
1343 | return (spa->spa_dsl_pool); | |
1344 | } | |
1345 | ||
9ae529ec CS |
1346 | boolean_t |
1347 | spa_is_initializing(spa_t *spa) | |
1348 | { | |
1349 | return (spa->spa_is_initializing); | |
1350 | } | |
1351 | ||
34dc7c2f BB |
1352 | blkptr_t * |
1353 | spa_get_rootblkptr(spa_t *spa) | |
1354 | { | |
1355 | return (&spa->spa_ubsync.ub_rootbp); | |
1356 | } | |
1357 | ||
1358 | void | |
1359 | spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) | |
1360 | { | |
1361 | spa->spa_uberblock.ub_rootbp = *bp; | |
1362 | } | |
1363 | ||
1364 | void | |
1365 | spa_altroot(spa_t *spa, char *buf, size_t buflen) | |
1366 | { | |
1367 | if (spa->spa_root == NULL) | |
1368 | buf[0] = '\0'; | |
1369 | else | |
1370 | (void) strncpy(buf, spa->spa_root, buflen); | |
1371 | } | |
1372 | ||
1373 | int | |
1374 | spa_sync_pass(spa_t *spa) | |
1375 | { | |
1376 | return (spa->spa_sync_pass); | |
1377 | } | |
1378 | ||
1379 | char * | |
1380 | spa_name(spa_t *spa) | |
1381 | { | |
34dc7c2f BB |
1382 | return (spa->spa_name); |
1383 | } | |
1384 | ||
1385 | uint64_t | |
1386 | spa_guid(spa_t *spa) | |
1387 | { | |
3bc7e0fb GW |
1388 | dsl_pool_t *dp = spa_get_dsl(spa); |
1389 | uint64_t guid; | |
1390 | ||
34dc7c2f BB |
1391 | /* |
1392 | * If we fail to parse the config during spa_load(), we can go through | |
1393 | * the error path (which posts an ereport) and end up here with no root | |
3541dc6d | 1394 | * vdev. We stash the original pool guid in 'spa_config_guid' to handle |
34dc7c2f BB |
1395 | * this case. |
1396 | */ | |
3bc7e0fb GW |
1397 | if (spa->spa_root_vdev == NULL) |
1398 | return (spa->spa_config_guid); | |
1399 | ||
1400 | guid = spa->spa_last_synced_guid != 0 ? | |
1401 | spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid; | |
1402 | ||
1403 | /* | |
1404 | * Return the most recently synced out guid unless we're | |
1405 | * in syncing context. | |
1406 | */ | |
1407 | if (dp && dsl_pool_sync_context(dp)) | |
34dc7c2f BB |
1408 | return (spa->spa_root_vdev->vdev_guid); |
1409 | else | |
3bc7e0fb | 1410 | return (guid); |
3541dc6d GA |
1411 | } |
1412 | ||
1413 | uint64_t | |
1414 | spa_load_guid(spa_t *spa) | |
1415 | { | |
1416 | /* | |
1417 | * This is a GUID that exists solely as a reference for the | |
1418 | * purposes of the arc. It is generated at load time, and | |
1419 | * is never written to persistent storage. | |
1420 | */ | |
1421 | return (spa->spa_load_guid); | |
34dc7c2f BB |
1422 | } |
1423 | ||
1424 | uint64_t | |
1425 | spa_last_synced_txg(spa_t *spa) | |
1426 | { | |
1427 | return (spa->spa_ubsync.ub_txg); | |
1428 | } | |
1429 | ||
1430 | uint64_t | |
1431 | spa_first_txg(spa_t *spa) | |
1432 | { | |
1433 | return (spa->spa_first_txg); | |
1434 | } | |
1435 | ||
428870ff BB |
1436 | uint64_t |
1437 | spa_syncing_txg(spa_t *spa) | |
1438 | { | |
1439 | return (spa->spa_syncing_txg); | |
1440 | } | |
1441 | ||
b128c09f | 1442 | pool_state_t |
34dc7c2f BB |
1443 | spa_state(spa_t *spa) |
1444 | { | |
1445 | return (spa->spa_state); | |
1446 | } | |
1447 | ||
428870ff BB |
1448 | spa_load_state_t |
1449 | spa_load_state(spa_t *spa) | |
34dc7c2f | 1450 | { |
428870ff | 1451 | return (spa->spa_load_state); |
34dc7c2f BB |
1452 | } |
1453 | ||
34dc7c2f | 1454 | uint64_t |
428870ff | 1455 | spa_freeze_txg(spa_t *spa) |
34dc7c2f | 1456 | { |
428870ff | 1457 | return (spa->spa_freeze_txg); |
34dc7c2f BB |
1458 | } |
1459 | ||
428870ff | 1460 | /* ARGSUSED */ |
34dc7c2f | 1461 | uint64_t |
428870ff | 1462 | spa_get_asize(spa_t *spa, uint64_t lsize) |
34dc7c2f | 1463 | { |
e8b96c60 | 1464 | return (lsize * spa_asize_inflation); |
34dc7c2f BB |
1465 | } |
1466 | ||
34dc7c2f BB |
1467 | uint64_t |
1468 | spa_get_dspace(spa_t *spa) | |
1469 | { | |
428870ff | 1470 | return (spa->spa_dspace); |
34dc7c2f BB |
1471 | } |
1472 | ||
428870ff BB |
1473 | void |
1474 | spa_update_dspace(spa_t *spa) | |
34dc7c2f | 1475 | { |
428870ff BB |
1476 | spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) + |
1477 | ddt_get_dedup_dspace(spa); | |
34dc7c2f BB |
1478 | } |
1479 | ||
1480 | /* | |
1481 | * Return the failure mode that has been set to this pool. The default | |
1482 | * behavior will be to block all I/Os when a complete failure occurs. | |
1483 | */ | |
1484 | uint8_t | |
1485 | spa_get_failmode(spa_t *spa) | |
1486 | { | |
1487 | return (spa->spa_failmode); | |
1488 | } | |
1489 | ||
b128c09f BB |
1490 | boolean_t |
1491 | spa_suspended(spa_t *spa) | |
1492 | { | |
1493 | return (spa->spa_suspended); | |
1494 | } | |
1495 | ||
34dc7c2f BB |
1496 | uint64_t |
1497 | spa_version(spa_t *spa) | |
1498 | { | |
1499 | return (spa->spa_ubsync.ub_version); | |
1500 | } | |
1501 | ||
428870ff BB |
1502 | boolean_t |
1503 | spa_deflate(spa_t *spa) | |
1504 | { | |
1505 | return (spa->spa_deflate); | |
1506 | } | |
1507 | ||
1508 | metaslab_class_t * | |
1509 | spa_normal_class(spa_t *spa) | |
1510 | { | |
1511 | return (spa->spa_normal_class); | |
1512 | } | |
1513 | ||
1514 | metaslab_class_t * | |
1515 | spa_log_class(spa_t *spa) | |
1516 | { | |
1517 | return (spa->spa_log_class); | |
1518 | } | |
1519 | ||
34dc7c2f BB |
1520 | int |
1521 | spa_max_replication(spa_t *spa) | |
1522 | { | |
1523 | /* | |
1524 | * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to | |
1525 | * handle BPs with more than one DVA allocated. Set our max | |
1526 | * replication level accordingly. | |
1527 | */ | |
1528 | if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) | |
1529 | return (1); | |
1530 | return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); | |
1531 | } | |
1532 | ||
428870ff BB |
1533 | int |
1534 | spa_prev_software_version(spa_t *spa) | |
1535 | { | |
1536 | return (spa->spa_prev_software_version); | |
1537 | } | |
1538 | ||
cc92e9d0 GW |
1539 | uint64_t |
1540 | spa_deadman_synctime(spa_t *spa) | |
1541 | { | |
1542 | return (spa->spa_deadman_synctime); | |
1543 | } | |
1544 | ||
34dc7c2f | 1545 | uint64_t |
428870ff | 1546 | dva_get_dsize_sync(spa_t *spa, const dva_t *dva) |
34dc7c2f | 1547 | { |
428870ff BB |
1548 | uint64_t asize = DVA_GET_ASIZE(dva); |
1549 | uint64_t dsize = asize; | |
34dc7c2f | 1550 | |
428870ff | 1551 | ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); |
34dc7c2f | 1552 | |
428870ff BB |
1553 | if (asize != 0 && spa->spa_deflate) { |
1554 | vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva)); | |
1555 | dsize = (asize >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio; | |
34dc7c2f | 1556 | } |
428870ff BB |
1557 | |
1558 | return (dsize); | |
1559 | } | |
1560 | ||
1561 | uint64_t | |
1562 | bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp) | |
1563 | { | |
1564 | uint64_t dsize = 0; | |
d6320ddb | 1565 | int d; |
428870ff | 1566 | |
d6320ddb | 1567 | for (d = 0; d < SPA_DVAS_PER_BP; d++) |
428870ff BB |
1568 | dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); |
1569 | ||
1570 | return (dsize); | |
1571 | } | |
1572 | ||
1573 | uint64_t | |
1574 | bp_get_dsize(spa_t *spa, const blkptr_t *bp) | |
1575 | { | |
1576 | uint64_t dsize = 0; | |
d6320ddb | 1577 | int d; |
428870ff BB |
1578 | |
1579 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
1580 | ||
d6320ddb | 1581 | for (d = 0; d < SPA_DVAS_PER_BP; d++) |
428870ff BB |
1582 | dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); |
1583 | ||
b128c09f | 1584 | spa_config_exit(spa, SCL_VDEV, FTAG); |
428870ff BB |
1585 | |
1586 | return (dsize); | |
34dc7c2f BB |
1587 | } |
1588 | ||
1589 | /* | |
1590 | * ========================================================================== | |
1591 | * Initialization and Termination | |
1592 | * ========================================================================== | |
1593 | */ | |
1594 | ||
1595 | static int | |
1596 | spa_name_compare(const void *a1, const void *a2) | |
1597 | { | |
1598 | const spa_t *s1 = a1; | |
1599 | const spa_t *s2 = a2; | |
1600 | int s; | |
1601 | ||
1602 | s = strcmp(s1->spa_name, s2->spa_name); | |
1603 | if (s > 0) | |
1604 | return (1); | |
1605 | if (s < 0) | |
1606 | return (-1); | |
1607 | return (0); | |
1608 | } | |
1609 | ||
34dc7c2f | 1610 | void |
0bc8fd78 | 1611 | spa_boot_init(void) |
34dc7c2f BB |
1612 | { |
1613 | spa_config_load(); | |
1614 | } | |
1615 | ||
1616 | void | |
1617 | spa_init(int mode) | |
1618 | { | |
1619 | mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); | |
1620 | mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); | |
1621 | mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); | |
1622 | cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); | |
1623 | ||
1624 | avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), | |
1625 | offsetof(spa_t, spa_avl)); | |
1626 | ||
1627 | avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), | |
1628 | offsetof(spa_aux_t, aux_avl)); | |
1629 | ||
1630 | avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), | |
1631 | offsetof(spa_aux_t, aux_avl)); | |
1632 | ||
fb5f0bc8 | 1633 | spa_mode_global = mode; |
34dc7c2f | 1634 | |
498877ba MA |
1635 | #ifndef _KERNEL |
1636 | if (spa_mode_global != FREAD && dprintf_find_string("watch")) { | |
1637 | struct sigaction sa; | |
1638 | ||
1639 | sa.sa_flags = SA_SIGINFO; | |
1640 | sigemptyset(&sa.sa_mask); | |
1641 | sa.sa_sigaction = arc_buf_sigsegv; | |
1642 | ||
1643 | if (sigaction(SIGSEGV, &sa, NULL) == -1) { | |
1644 | perror("could not enable watchpoints: " | |
1645 | "sigaction(SIGSEGV, ...) = "); | |
1646 | } else { | |
1647 | arc_watch = B_TRUE; | |
1648 | } | |
1649 | } | |
1650 | #endif | |
1651 | ||
26685276 | 1652 | fm_init(); |
34dc7c2f BB |
1653 | refcount_init(); |
1654 | unique_init(); | |
55d85d5a | 1655 | space_map_init(); |
34dc7c2f BB |
1656 | zio_init(); |
1657 | dmu_init(); | |
1658 | zil_init(); | |
1659 | vdev_cache_stat_init(); | |
1660 | zfs_prop_init(); | |
1661 | zpool_prop_init(); | |
9ae529ec | 1662 | zpool_feature_init(); |
34dc7c2f | 1663 | spa_config_load(); |
b128c09f | 1664 | l2arc_start(); |
34dc7c2f BB |
1665 | } |
1666 | ||
1667 | void | |
1668 | spa_fini(void) | |
1669 | { | |
b128c09f BB |
1670 | l2arc_stop(); |
1671 | ||
34dc7c2f BB |
1672 | spa_evict_all(); |
1673 | ||
1674 | vdev_cache_stat_fini(); | |
1675 | zil_fini(); | |
1676 | dmu_fini(); | |
1677 | zio_fini(); | |
55d85d5a | 1678 | space_map_fini(); |
34dc7c2f BB |
1679 | unique_fini(); |
1680 | refcount_fini(); | |
26685276 | 1681 | fm_fini(); |
34dc7c2f BB |
1682 | |
1683 | avl_destroy(&spa_namespace_avl); | |
1684 | avl_destroy(&spa_spare_avl); | |
1685 | avl_destroy(&spa_l2cache_avl); | |
1686 | ||
1687 | cv_destroy(&spa_namespace_cv); | |
1688 | mutex_destroy(&spa_namespace_lock); | |
1689 | mutex_destroy(&spa_spare_lock); | |
1690 | mutex_destroy(&spa_l2cache_lock); | |
1691 | } | |
1692 | ||
1693 | /* | |
1694 | * Return whether this pool has slogs. No locking needed. | |
1695 | * It's not a problem if the wrong answer is returned as it's only for | |
1696 | * performance and not correctness | |
1697 | */ | |
1698 | boolean_t | |
1699 | spa_has_slogs(spa_t *spa) | |
1700 | { | |
1701 | return (spa->spa_log_class->mc_rotor != NULL); | |
1702 | } | |
b128c09f | 1703 | |
428870ff BB |
1704 | spa_log_state_t |
1705 | spa_get_log_state(spa_t *spa) | |
1706 | { | |
1707 | return (spa->spa_log_state); | |
1708 | } | |
1709 | ||
1710 | void | |
1711 | spa_set_log_state(spa_t *spa, spa_log_state_t state) | |
1712 | { | |
1713 | spa->spa_log_state = state; | |
1714 | } | |
1715 | ||
b128c09f BB |
1716 | boolean_t |
1717 | spa_is_root(spa_t *spa) | |
1718 | { | |
1719 | return (spa->spa_is_root); | |
1720 | } | |
fb5f0bc8 BB |
1721 | |
1722 | boolean_t | |
1723 | spa_writeable(spa_t *spa) | |
1724 | { | |
1725 | return (!!(spa->spa_mode & FWRITE)); | |
1726 | } | |
1727 | ||
1728 | int | |
1729 | spa_mode(spa_t *spa) | |
1730 | { | |
1731 | return (spa->spa_mode); | |
1732 | } | |
428870ff BB |
1733 | |
1734 | uint64_t | |
1735 | spa_bootfs(spa_t *spa) | |
1736 | { | |
1737 | return (spa->spa_bootfs); | |
1738 | } | |
1739 | ||
1740 | uint64_t | |
1741 | spa_delegation(spa_t *spa) | |
1742 | { | |
1743 | return (spa->spa_delegation); | |
1744 | } | |
1745 | ||
1746 | objset_t * | |
1747 | spa_meta_objset(spa_t *spa) | |
1748 | { | |
1749 | return (spa->spa_meta_objset); | |
1750 | } | |
1751 | ||
1752 | enum zio_checksum | |
1753 | spa_dedup_checksum(spa_t *spa) | |
1754 | { | |
1755 | return (spa->spa_dedup_checksum); | |
1756 | } | |
1757 | ||
1758 | /* | |
1759 | * Reset pool scan stat per scan pass (or reboot). | |
1760 | */ | |
1761 | void | |
1762 | spa_scan_stat_init(spa_t *spa) | |
1763 | { | |
1764 | /* data not stored on disk */ | |
1765 | spa->spa_scan_pass_start = gethrestime_sec(); | |
1766 | spa->spa_scan_pass_exam = 0; | |
1767 | vdev_scan_stat_init(spa->spa_root_vdev); | |
1768 | } | |
1769 | ||
1770 | /* | |
1771 | * Get scan stats for zpool status reports | |
1772 | */ | |
1773 | int | |
1774 | spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps) | |
1775 | { | |
1776 | dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL; | |
1777 | ||
1778 | if (scn == NULL || scn->scn_phys.scn_func == POOL_SCAN_NONE) | |
2e528b49 | 1779 | return (SET_ERROR(ENOENT)); |
428870ff BB |
1780 | bzero(ps, sizeof (pool_scan_stat_t)); |
1781 | ||
1782 | /* data stored on disk */ | |
1783 | ps->pss_func = scn->scn_phys.scn_func; | |
1784 | ps->pss_start_time = scn->scn_phys.scn_start_time; | |
1785 | ps->pss_end_time = scn->scn_phys.scn_end_time; | |
1786 | ps->pss_to_examine = scn->scn_phys.scn_to_examine; | |
1787 | ps->pss_examined = scn->scn_phys.scn_examined; | |
1788 | ps->pss_to_process = scn->scn_phys.scn_to_process; | |
1789 | ps->pss_processed = scn->scn_phys.scn_processed; | |
1790 | ps->pss_errors = scn->scn_phys.scn_errors; | |
1791 | ps->pss_state = scn->scn_phys.scn_state; | |
1792 | ||
1793 | /* data not stored on disk */ | |
1794 | ps->pss_pass_start = spa->spa_scan_pass_start; | |
1795 | ps->pss_pass_exam = spa->spa_scan_pass_exam; | |
1796 | ||
1797 | return (0); | |
1798 | } | |
c28b2279 | 1799 | |
6d974228 GW |
1800 | boolean_t |
1801 | spa_debug_enabled(spa_t *spa) | |
1802 | { | |
1803 | return (spa->spa_debug); | |
1804 | } | |
1805 | ||
c28b2279 BB |
1806 | #if defined(_KERNEL) && defined(HAVE_SPL) |
1807 | /* Namespace manipulation */ | |
1808 | EXPORT_SYMBOL(spa_lookup); | |
1809 | EXPORT_SYMBOL(spa_add); | |
1810 | EXPORT_SYMBOL(spa_remove); | |
1811 | EXPORT_SYMBOL(spa_next); | |
1812 | ||
1813 | /* Refcount functions */ | |
1814 | EXPORT_SYMBOL(spa_open_ref); | |
1815 | EXPORT_SYMBOL(spa_close); | |
1816 | EXPORT_SYMBOL(spa_refcount_zero); | |
1817 | ||
1818 | /* Pool configuration lock */ | |
1819 | EXPORT_SYMBOL(spa_config_tryenter); | |
1820 | EXPORT_SYMBOL(spa_config_enter); | |
1821 | EXPORT_SYMBOL(spa_config_exit); | |
1822 | EXPORT_SYMBOL(spa_config_held); | |
1823 | ||
1824 | /* Pool vdev add/remove lock */ | |
1825 | EXPORT_SYMBOL(spa_vdev_enter); | |
1826 | EXPORT_SYMBOL(spa_vdev_exit); | |
1827 | ||
1828 | /* Pool vdev state change lock */ | |
1829 | EXPORT_SYMBOL(spa_vdev_state_enter); | |
1830 | EXPORT_SYMBOL(spa_vdev_state_exit); | |
1831 | ||
1832 | /* Accessor functions */ | |
1833 | EXPORT_SYMBOL(spa_shutting_down); | |
1834 | EXPORT_SYMBOL(spa_get_dsl); | |
1835 | EXPORT_SYMBOL(spa_get_rootblkptr); | |
1836 | EXPORT_SYMBOL(spa_set_rootblkptr); | |
1837 | EXPORT_SYMBOL(spa_altroot); | |
1838 | EXPORT_SYMBOL(spa_sync_pass); | |
1839 | EXPORT_SYMBOL(spa_name); | |
1840 | EXPORT_SYMBOL(spa_guid); | |
1841 | EXPORT_SYMBOL(spa_last_synced_txg); | |
1842 | EXPORT_SYMBOL(spa_first_txg); | |
1843 | EXPORT_SYMBOL(spa_syncing_txg); | |
1844 | EXPORT_SYMBOL(spa_version); | |
1845 | EXPORT_SYMBOL(spa_state); | |
1846 | EXPORT_SYMBOL(spa_load_state); | |
1847 | EXPORT_SYMBOL(spa_freeze_txg); | |
1848 | EXPORT_SYMBOL(spa_get_asize); | |
1849 | EXPORT_SYMBOL(spa_get_dspace); | |
1850 | EXPORT_SYMBOL(spa_update_dspace); | |
1851 | EXPORT_SYMBOL(spa_deflate); | |
1852 | EXPORT_SYMBOL(spa_normal_class); | |
1853 | EXPORT_SYMBOL(spa_log_class); | |
1854 | EXPORT_SYMBOL(spa_max_replication); | |
1855 | EXPORT_SYMBOL(spa_prev_software_version); | |
1856 | EXPORT_SYMBOL(spa_get_failmode); | |
1857 | EXPORT_SYMBOL(spa_suspended); | |
1858 | EXPORT_SYMBOL(spa_bootfs); | |
1859 | EXPORT_SYMBOL(spa_delegation); | |
1860 | EXPORT_SYMBOL(spa_meta_objset); | |
1861 | ||
1862 | /* Miscellaneous support routines */ | |
1863 | EXPORT_SYMBOL(spa_rename); | |
1864 | EXPORT_SYMBOL(spa_guid_exists); | |
1865 | EXPORT_SYMBOL(spa_strdup); | |
1866 | EXPORT_SYMBOL(spa_strfree); | |
1867 | EXPORT_SYMBOL(spa_get_random); | |
1868 | EXPORT_SYMBOL(spa_generate_guid); | |
1869 | EXPORT_SYMBOL(sprintf_blkptr); | |
1870 | EXPORT_SYMBOL(spa_freeze); | |
1871 | EXPORT_SYMBOL(spa_upgrade); | |
1872 | EXPORT_SYMBOL(spa_evict_all); | |
1873 | EXPORT_SYMBOL(spa_lookup_by_guid); | |
1874 | EXPORT_SYMBOL(spa_has_spare); | |
1875 | EXPORT_SYMBOL(dva_get_dsize_sync); | |
1876 | EXPORT_SYMBOL(bp_get_dsize_sync); | |
1877 | EXPORT_SYMBOL(bp_get_dsize); | |
1878 | EXPORT_SYMBOL(spa_has_slogs); | |
1879 | EXPORT_SYMBOL(spa_is_root); | |
1880 | EXPORT_SYMBOL(spa_writeable); | |
1881 | EXPORT_SYMBOL(spa_mode); | |
1882 | ||
1883 | EXPORT_SYMBOL(spa_namespace_lock); | |
cc92e9d0 | 1884 | |
e8b96c60 | 1885 | module_param(zfs_deadman_synctime_ms, ulong, 0644); |
d1d7e268 | 1886 | MODULE_PARM_DESC(zfs_deadman_synctime_ms, "Expiration time in milliseconds"); |
cc92e9d0 GW |
1887 | |
1888 | module_param(zfs_deadman_enabled, int, 0644); | |
1889 | MODULE_PARM_DESC(zfs_deadman_enabled, "Enable deadman timer"); | |
e8b96c60 MA |
1890 | |
1891 | module_param(spa_asize_inflation, int, 0644); | |
1892 | MODULE_PARM_DESC(spa_asize_inflation, | |
d1d7e268 | 1893 | "SPA size estimate multiplication factor"); |
c28b2279 | 1894 | #endif |