]>
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. |
9ae529ec | 23 | * Copyright (c) 2012 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> |
34dc7c2f | 51 | #include "zfs_prop.h" |
9ae529ec | 52 | #include "zfeature_common.h" |
34dc7c2f BB |
53 | |
54 | /* | |
55 | * SPA locking | |
56 | * | |
57 | * There are four basic locks for managing spa_t structures: | |
58 | * | |
59 | * spa_namespace_lock (global mutex) | |
60 | * | |
61 | * This lock must be acquired to do any of the following: | |
62 | * | |
63 | * - Lookup a spa_t by name | |
64 | * - Add or remove a spa_t from the namespace | |
65 | * - Increase spa_refcount from non-zero | |
66 | * - Check if spa_refcount is zero | |
67 | * - Rename a spa_t | |
68 | * - add/remove/attach/detach devices | |
69 | * - Held for the duration of create/destroy/import/export | |
70 | * | |
71 | * It does not need to handle recursion. A create or destroy may | |
72 | * reference objects (files or zvols) in other pools, but by | |
73 | * definition they must have an existing reference, and will never need | |
74 | * to lookup a spa_t by name. | |
75 | * | |
76 | * spa_refcount (per-spa refcount_t protected by mutex) | |
77 | * | |
78 | * This reference count keep track of any active users of the spa_t. The | |
79 | * spa_t cannot be destroyed or freed while this is non-zero. Internally, | |
80 | * the refcount is never really 'zero' - opening a pool implicitly keeps | |
b128c09f | 81 | * some references in the DMU. Internally we check against spa_minref, but |
34dc7c2f BB |
82 | * present the image of a zero/non-zero value to consumers. |
83 | * | |
b128c09f | 84 | * spa_config_lock[] (per-spa array of rwlocks) |
34dc7c2f BB |
85 | * |
86 | * This protects the spa_t from config changes, and must be held in | |
87 | * the following circumstances: | |
88 | * | |
89 | * - RW_READER to perform I/O to the spa | |
90 | * - RW_WRITER to change the vdev config | |
91 | * | |
34dc7c2f BB |
92 | * The locking order is fairly straightforward: |
93 | * | |
94 | * spa_namespace_lock -> spa_refcount | |
95 | * | |
96 | * The namespace lock must be acquired to increase the refcount from 0 | |
97 | * or to check if it is zero. | |
98 | * | |
b128c09f | 99 | * spa_refcount -> spa_config_lock[] |
34dc7c2f BB |
100 | * |
101 | * There must be at least one valid reference on the spa_t to acquire | |
102 | * the config lock. | |
103 | * | |
b128c09f | 104 | * spa_namespace_lock -> spa_config_lock[] |
34dc7c2f BB |
105 | * |
106 | * The namespace lock must always be taken before the config lock. | |
107 | * | |
108 | * | |
b128c09f | 109 | * The spa_namespace_lock can be acquired directly and is globally visible. |
34dc7c2f | 110 | * |
b128c09f BB |
111 | * The namespace is manipulated using the following functions, all of which |
112 | * require the spa_namespace_lock to be held. | |
34dc7c2f BB |
113 | * |
114 | * spa_lookup() Lookup a spa_t by name. | |
115 | * | |
116 | * spa_add() Create a new spa_t in the namespace. | |
117 | * | |
118 | * spa_remove() Remove a spa_t from the namespace. This also | |
119 | * frees up any memory associated with the spa_t. | |
120 | * | |
121 | * spa_next() Returns the next spa_t in the system, or the | |
122 | * first if NULL is passed. | |
123 | * | |
124 | * spa_evict_all() Shutdown and remove all spa_t structures in | |
125 | * the system. | |
126 | * | |
127 | * spa_guid_exists() Determine whether a pool/device guid exists. | |
128 | * | |
129 | * The spa_refcount is manipulated using the following functions: | |
130 | * | |
131 | * spa_open_ref() Adds a reference to the given spa_t. Must be | |
132 | * called with spa_namespace_lock held if the | |
133 | * refcount is currently zero. | |
134 | * | |
135 | * spa_close() Remove a reference from the spa_t. This will | |
136 | * not free the spa_t or remove it from the | |
137 | * namespace. No locking is required. | |
138 | * | |
139 | * spa_refcount_zero() Returns true if the refcount is currently | |
140 | * zero. Must be called with spa_namespace_lock | |
141 | * held. | |
142 | * | |
b128c09f BB |
143 | * The spa_config_lock[] is an array of rwlocks, ordered as follows: |
144 | * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV. | |
145 | * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}(). | |
146 | * | |
147 | * To read the configuration, it suffices to hold one of these locks as reader. | |
148 | * To modify the configuration, you must hold all locks as writer. To modify | |
149 | * vdev state without altering the vdev tree's topology (e.g. online/offline), | |
150 | * you must hold SCL_STATE and SCL_ZIO as writer. | |
151 | * | |
152 | * We use these distinct config locks to avoid recursive lock entry. | |
153 | * For example, spa_sync() (which holds SCL_CONFIG as reader) induces | |
154 | * block allocations (SCL_ALLOC), which may require reading space maps | |
155 | * from disk (dmu_read() -> zio_read() -> SCL_ZIO). | |
156 | * | |
157 | * The spa config locks cannot be normal rwlocks because we need the | |
158 | * ability to hand off ownership. For example, SCL_ZIO is acquired | |
159 | * by the issuing thread and later released by an interrupt thread. | |
160 | * They do, however, obey the usual write-wanted semantics to prevent | |
161 | * writer (i.e. system administrator) starvation. | |
162 | * | |
163 | * The lock acquisition rules are as follows: | |
164 | * | |
165 | * SCL_CONFIG | |
166 | * Protects changes to the vdev tree topology, such as vdev | |
167 | * add/remove/attach/detach. Protects the dirty config list | |
168 | * (spa_config_dirty_list) and the set of spares and l2arc devices. | |
169 | * | |
170 | * SCL_STATE | |
171 | * Protects changes to pool state and vdev state, such as vdev | |
172 | * online/offline/fault/degrade/clear. Protects the dirty state list | |
173 | * (spa_state_dirty_list) and global pool state (spa_state). | |
174 | * | |
175 | * SCL_ALLOC | |
176 | * Protects changes to metaslab groups and classes. | |
177 | * Held as reader by metaslab_alloc() and metaslab_claim(). | |
178 | * | |
179 | * SCL_ZIO | |
180 | * Held by bp-level zios (those which have no io_vd upon entry) | |
181 | * to prevent changes to the vdev tree. The bp-level zio implicitly | |
182 | * protects all of its vdev child zios, which do not hold SCL_ZIO. | |
183 | * | |
184 | * SCL_FREE | |
185 | * Protects changes to metaslab groups and classes. | |
186 | * Held as reader by metaslab_free(). SCL_FREE is distinct from | |
187 | * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free | |
188 | * blocks in zio_done() while another i/o that holds either | |
189 | * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete. | |
190 | * | |
191 | * SCL_VDEV | |
192 | * Held as reader to prevent changes to the vdev tree during trivial | |
428870ff | 193 | * inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the |
b128c09f BB |
194 | * other locks, and lower than all of them, to ensure that it's safe |
195 | * to acquire regardless of caller context. | |
196 | * | |
197 | * In addition, the following rules apply: | |
198 | * | |
199 | * (a) spa_props_lock protects pool properties, spa_config and spa_config_list. | |
200 | * The lock ordering is SCL_CONFIG > spa_props_lock. | |
201 | * | |
202 | * (b) I/O operations on leaf vdevs. For any zio operation that takes | |
203 | * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(), | |
204 | * or zio_write_phys() -- the caller must ensure that the config cannot | |
205 | * cannot change in the interim, and that the vdev cannot be reopened. | |
206 | * SCL_STATE as reader suffices for both. | |
34dc7c2f BB |
207 | * |
208 | * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). | |
209 | * | |
210 | * spa_vdev_enter() Acquire the namespace lock and the config lock | |
211 | * for writing. | |
212 | * | |
213 | * spa_vdev_exit() Release the config lock, wait for all I/O | |
214 | * to complete, sync the updated configs to the | |
215 | * cache, and release the namespace lock. | |
216 | * | |
b128c09f BB |
217 | * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit(). |
218 | * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual | |
219 | * locking is, always, based on spa_namespace_lock and spa_config_lock[]. | |
220 | * | |
9ae529ec | 221 | * spa_rename() is also implemented within this file since it requires |
b128c09f | 222 | * manipulation of the namespace. |
34dc7c2f BB |
223 | */ |
224 | ||
225 | static avl_tree_t spa_namespace_avl; | |
226 | kmutex_t spa_namespace_lock; | |
227 | static kcondvar_t spa_namespace_cv; | |
228 | static int spa_active_count; | |
229 | int spa_max_replication_override = SPA_DVAS_PER_BP; | |
230 | ||
231 | static kmutex_t spa_spare_lock; | |
232 | static avl_tree_t spa_spare_avl; | |
233 | static kmutex_t spa_l2cache_lock; | |
234 | static avl_tree_t spa_l2cache_avl; | |
235 | ||
236 | kmem_cache_t *spa_buffer_pool; | |
fb5f0bc8 | 237 | int spa_mode_global; |
34dc7c2f | 238 | |
34dc7c2f BB |
239 | /* |
240 | * ========================================================================== | |
241 | * SPA config locking | |
242 | * ========================================================================== | |
243 | */ | |
244 | static void | |
b128c09f BB |
245 | spa_config_lock_init(spa_t *spa) |
246 | { | |
d6320ddb BB |
247 | int i; |
248 | ||
249 | for (i = 0; i < SCL_LOCKS; i++) { | |
b128c09f BB |
250 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
251 | mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL); | |
252 | cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL); | |
253 | refcount_create(&scl->scl_count); | |
254 | scl->scl_writer = NULL; | |
255 | scl->scl_write_wanted = 0; | |
256 | } | |
34dc7c2f BB |
257 | } |
258 | ||
259 | static void | |
b128c09f BB |
260 | spa_config_lock_destroy(spa_t *spa) |
261 | { | |
d6320ddb BB |
262 | int i; |
263 | ||
264 | for (i = 0; i < SCL_LOCKS; i++) { | |
b128c09f BB |
265 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
266 | mutex_destroy(&scl->scl_lock); | |
267 | cv_destroy(&scl->scl_cv); | |
268 | refcount_destroy(&scl->scl_count); | |
269 | ASSERT(scl->scl_writer == NULL); | |
270 | ASSERT(scl->scl_write_wanted == 0); | |
271 | } | |
272 | } | |
273 | ||
274 | int | |
275 | spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw) | |
34dc7c2f | 276 | { |
d6320ddb BB |
277 | int i; |
278 | ||
279 | for (i = 0; i < SCL_LOCKS; i++) { | |
b128c09f BB |
280 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
281 | if (!(locks & (1 << i))) | |
282 | continue; | |
283 | mutex_enter(&scl->scl_lock); | |
284 | if (rw == RW_READER) { | |
285 | if (scl->scl_writer || scl->scl_write_wanted) { | |
286 | mutex_exit(&scl->scl_lock); | |
287 | spa_config_exit(spa, locks ^ (1 << i), tag); | |
288 | return (0); | |
289 | } | |
290 | } else { | |
291 | ASSERT(scl->scl_writer != curthread); | |
292 | if (!refcount_is_zero(&scl->scl_count)) { | |
293 | mutex_exit(&scl->scl_lock); | |
294 | spa_config_exit(spa, locks ^ (1 << i), tag); | |
295 | return (0); | |
296 | } | |
297 | scl->scl_writer = curthread; | |
298 | } | |
299 | (void) refcount_add(&scl->scl_count, tag); | |
300 | mutex_exit(&scl->scl_lock); | |
301 | } | |
302 | return (1); | |
34dc7c2f BB |
303 | } |
304 | ||
305 | void | |
b128c09f | 306 | spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw) |
34dc7c2f | 307 | { |
45d1cae3 | 308 | int wlocks_held = 0; |
d6320ddb | 309 | int i; |
45d1cae3 | 310 | |
d6320ddb | 311 | for (i = 0; i < SCL_LOCKS; i++) { |
b128c09f | 312 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
45d1cae3 BB |
313 | if (scl->scl_writer == curthread) |
314 | wlocks_held |= (1 << i); | |
b128c09f BB |
315 | if (!(locks & (1 << i))) |
316 | continue; | |
317 | mutex_enter(&scl->scl_lock); | |
318 | if (rw == RW_READER) { | |
319 | while (scl->scl_writer || scl->scl_write_wanted) { | |
320 | cv_wait(&scl->scl_cv, &scl->scl_lock); | |
321 | } | |
322 | } else { | |
323 | ASSERT(scl->scl_writer != curthread); | |
324 | while (!refcount_is_zero(&scl->scl_count)) { | |
325 | scl->scl_write_wanted++; | |
326 | cv_wait(&scl->scl_cv, &scl->scl_lock); | |
327 | scl->scl_write_wanted--; | |
328 | } | |
329 | scl->scl_writer = curthread; | |
330 | } | |
331 | (void) refcount_add(&scl->scl_count, tag); | |
332 | mutex_exit(&scl->scl_lock); | |
34dc7c2f | 333 | } |
45d1cae3 | 334 | ASSERT(wlocks_held <= locks); |
34dc7c2f BB |
335 | } |
336 | ||
337 | void | |
b128c09f | 338 | spa_config_exit(spa_t *spa, int locks, void *tag) |
34dc7c2f | 339 | { |
d6320ddb BB |
340 | int i; |
341 | ||
342 | for (i = SCL_LOCKS - 1; i >= 0; i--) { | |
b128c09f BB |
343 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
344 | if (!(locks & (1 << i))) | |
345 | continue; | |
346 | mutex_enter(&scl->scl_lock); | |
347 | ASSERT(!refcount_is_zero(&scl->scl_count)); | |
348 | if (refcount_remove(&scl->scl_count, tag) == 0) { | |
349 | ASSERT(scl->scl_writer == NULL || | |
350 | scl->scl_writer == curthread); | |
351 | scl->scl_writer = NULL; /* OK in either case */ | |
352 | cv_broadcast(&scl->scl_cv); | |
353 | } | |
354 | mutex_exit(&scl->scl_lock); | |
34dc7c2f | 355 | } |
34dc7c2f BB |
356 | } |
357 | ||
b128c09f BB |
358 | int |
359 | spa_config_held(spa_t *spa, int locks, krw_t rw) | |
34dc7c2f | 360 | { |
d6320ddb | 361 | int i, locks_held = 0; |
34dc7c2f | 362 | |
d6320ddb | 363 | for (i = 0; i < SCL_LOCKS; i++) { |
b128c09f BB |
364 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
365 | if (!(locks & (1 << i))) | |
366 | continue; | |
367 | if ((rw == RW_READER && !refcount_is_zero(&scl->scl_count)) || | |
368 | (rw == RW_WRITER && scl->scl_writer == curthread)) | |
369 | locks_held |= 1 << i; | |
370 | } | |
371 | ||
372 | return (locks_held); | |
34dc7c2f BB |
373 | } |
374 | ||
375 | /* | |
376 | * ========================================================================== | |
377 | * SPA namespace functions | |
378 | * ========================================================================== | |
379 | */ | |
380 | ||
381 | /* | |
382 | * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. | |
383 | * Returns NULL if no matching spa_t is found. | |
384 | */ | |
385 | spa_t * | |
386 | spa_lookup(const char *name) | |
387 | { | |
b128c09f BB |
388 | static spa_t search; /* spa_t is large; don't allocate on stack */ |
389 | spa_t *spa; | |
34dc7c2f | 390 | avl_index_t where; |
d4ed6673 | 391 | char c = 0; |
34dc7c2f BB |
392 | char *cp; |
393 | ||
394 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
395 | ||
396 | /* | |
397 | * If it's a full dataset name, figure out the pool name and | |
398 | * just use that. | |
399 | */ | |
400 | cp = strpbrk(name, "/@"); | |
401 | if (cp) { | |
402 | c = *cp; | |
403 | *cp = '\0'; | |
404 | } | |
405 | ||
b128c09f | 406 | (void) strlcpy(search.spa_name, name, sizeof (search.spa_name)); |
34dc7c2f BB |
407 | spa = avl_find(&spa_namespace_avl, &search, &where); |
408 | ||
409 | if (cp) | |
410 | *cp = c; | |
411 | ||
412 | return (spa); | |
413 | } | |
414 | ||
415 | /* | |
416 | * Create an uninitialized spa_t with the given name. Requires | |
417 | * spa_namespace_lock. The caller must ensure that the spa_t doesn't already | |
418 | * exist by calling spa_lookup() first. | |
419 | */ | |
420 | spa_t * | |
428870ff | 421 | spa_add(const char *name, nvlist_t *config, const char *altroot) |
34dc7c2f BB |
422 | { |
423 | spa_t *spa; | |
b128c09f | 424 | spa_config_dirent_t *dp; |
d6320ddb | 425 | int t; |
34dc7c2f BB |
426 | |
427 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
428 | ||
b8d06fca | 429 | spa = kmem_zalloc(sizeof (spa_t), KM_PUSHPAGE | KM_NODEBUG); |
34dc7c2f | 430 | |
34dc7c2f | 431 | mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 432 | mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); |
428870ff | 433 | mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 434 | mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); |
428870ff | 435 | mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 436 | mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); |
428870ff BB |
437 | mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); |
438 | mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL); | |
439 | mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL); | |
34dc7c2f BB |
440 | |
441 | cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); | |
428870ff | 442 | cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL); |
34dc7c2f | 443 | cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); |
b128c09f | 444 | cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL); |
34dc7c2f | 445 | |
d6320ddb | 446 | for (t = 0; t < TXG_SIZE; t++) |
428870ff BB |
447 | bplist_create(&spa->spa_free_bplist[t]); |
448 | ||
b128c09f | 449 | (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name)); |
34dc7c2f BB |
450 | spa->spa_state = POOL_STATE_UNINITIALIZED; |
451 | spa->spa_freeze_txg = UINT64_MAX; | |
452 | spa->spa_final_txg = UINT64_MAX; | |
428870ff BB |
453 | spa->spa_load_max_txg = UINT64_MAX; |
454 | spa->spa_proc = &p0; | |
455 | spa->spa_proc_state = SPA_PROC_NONE; | |
34dc7c2f BB |
456 | |
457 | refcount_create(&spa->spa_refcount); | |
b128c09f | 458 | spa_config_lock_init(spa); |
34dc7c2f BB |
459 | |
460 | avl_add(&spa_namespace_avl, spa); | |
461 | ||
34dc7c2f BB |
462 | /* |
463 | * Set the alternate root, if there is one. | |
464 | */ | |
465 | if (altroot) { | |
466 | spa->spa_root = spa_strdup(altroot); | |
467 | spa_active_count++; | |
468 | } | |
469 | ||
b128c09f BB |
470 | /* |
471 | * Every pool starts with the default cachefile | |
472 | */ | |
473 | list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t), | |
474 | offsetof(spa_config_dirent_t, scd_link)); | |
475 | ||
b8d06fca | 476 | dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_PUSHPAGE); |
428870ff | 477 | dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path); |
b128c09f BB |
478 | list_insert_head(&spa->spa_config_list, dp); |
479 | ||
572e2857 | 480 | VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME, |
b8d06fca | 481 | KM_PUSHPAGE) == 0); |
572e2857 | 482 | |
9ae529ec CS |
483 | if (config != NULL) { |
484 | nvlist_t *features; | |
485 | ||
486 | if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ, | |
487 | &features) == 0) { | |
488 | VERIFY(nvlist_dup(features, &spa->spa_label_features, | |
489 | 0) == 0); | |
490 | } | |
491 | ||
428870ff | 492 | VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); |
9ae529ec CS |
493 | } |
494 | ||
495 | if (spa->spa_label_features == NULL) { | |
496 | VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME, | |
497 | KM_SLEEP) == 0); | |
498 | } | |
428870ff | 499 | |
34dc7c2f BB |
500 | return (spa); |
501 | } | |
502 | ||
503 | /* | |
504 | * Removes a spa_t from the namespace, freeing up any memory used. Requires | |
505 | * spa_namespace_lock. This is called only after the spa_t has been closed and | |
506 | * deactivated. | |
507 | */ | |
508 | void | |
509 | spa_remove(spa_t *spa) | |
510 | { | |
b128c09f | 511 | spa_config_dirent_t *dp; |
d6320ddb | 512 | int t; |
b128c09f | 513 | |
34dc7c2f BB |
514 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); |
515 | ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); | |
34dc7c2f | 516 | |
428870ff BB |
517 | nvlist_free(spa->spa_config_splitting); |
518 | ||
34dc7c2f BB |
519 | avl_remove(&spa_namespace_avl, spa); |
520 | cv_broadcast(&spa_namespace_cv); | |
521 | ||
522 | if (spa->spa_root) { | |
523 | spa_strfree(spa->spa_root); | |
524 | spa_active_count--; | |
525 | } | |
526 | ||
b128c09f BB |
527 | while ((dp = list_head(&spa->spa_config_list)) != NULL) { |
528 | list_remove(&spa->spa_config_list, dp); | |
529 | if (dp->scd_path != NULL) | |
530 | spa_strfree(dp->scd_path); | |
531 | kmem_free(dp, sizeof (spa_config_dirent_t)); | |
532 | } | |
34dc7c2f | 533 | |
b128c09f | 534 | list_destroy(&spa->spa_config_list); |
34dc7c2f | 535 | |
9ae529ec | 536 | nvlist_free(spa->spa_label_features); |
572e2857 | 537 | nvlist_free(spa->spa_load_info); |
34dc7c2f BB |
538 | spa_config_set(spa, NULL); |
539 | ||
540 | refcount_destroy(&spa->spa_refcount); | |
541 | ||
b128c09f | 542 | spa_config_lock_destroy(spa); |
34dc7c2f | 543 | |
d6320ddb | 544 | for (t = 0; t < TXG_SIZE; t++) |
428870ff BB |
545 | bplist_destroy(&spa->spa_free_bplist[t]); |
546 | ||
34dc7c2f | 547 | cv_destroy(&spa->spa_async_cv); |
428870ff | 548 | cv_destroy(&spa->spa_proc_cv); |
34dc7c2f | 549 | cv_destroy(&spa->spa_scrub_io_cv); |
b128c09f | 550 | cv_destroy(&spa->spa_suspend_cv); |
34dc7c2f | 551 | |
34dc7c2f | 552 | mutex_destroy(&spa->spa_async_lock); |
34dc7c2f | 553 | mutex_destroy(&spa->spa_errlist_lock); |
428870ff | 554 | mutex_destroy(&spa->spa_errlog_lock); |
34dc7c2f | 555 | mutex_destroy(&spa->spa_history_lock); |
428870ff | 556 | mutex_destroy(&spa->spa_proc_lock); |
34dc7c2f | 557 | mutex_destroy(&spa->spa_props_lock); |
428870ff | 558 | mutex_destroy(&spa->spa_scrub_lock); |
b128c09f | 559 | mutex_destroy(&spa->spa_suspend_lock); |
428870ff | 560 | mutex_destroy(&spa->spa_vdev_top_lock); |
34dc7c2f BB |
561 | |
562 | kmem_free(spa, sizeof (spa_t)); | |
563 | } | |
564 | ||
565 | /* | |
566 | * Given a pool, return the next pool in the namespace, or NULL if there is | |
567 | * none. If 'prev' is NULL, return the first pool. | |
568 | */ | |
569 | spa_t * | |
570 | spa_next(spa_t *prev) | |
571 | { | |
572 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
573 | ||
574 | if (prev) | |
575 | return (AVL_NEXT(&spa_namespace_avl, prev)); | |
576 | else | |
577 | return (avl_first(&spa_namespace_avl)); | |
578 | } | |
579 | ||
580 | /* | |
581 | * ========================================================================== | |
582 | * SPA refcount functions | |
583 | * ========================================================================== | |
584 | */ | |
585 | ||
586 | /* | |
587 | * Add a reference to the given spa_t. Must have at least one reference, or | |
588 | * have the namespace lock held. | |
589 | */ | |
590 | void | |
591 | spa_open_ref(spa_t *spa, void *tag) | |
592 | { | |
b128c09f | 593 | ASSERT(refcount_count(&spa->spa_refcount) >= spa->spa_minref || |
34dc7c2f | 594 | MUTEX_HELD(&spa_namespace_lock)); |
34dc7c2f BB |
595 | (void) refcount_add(&spa->spa_refcount, tag); |
596 | } | |
597 | ||
598 | /* | |
599 | * Remove a reference to the given spa_t. Must have at least one reference, or | |
600 | * have the namespace lock held. | |
601 | */ | |
602 | void | |
603 | spa_close(spa_t *spa, void *tag) | |
604 | { | |
b128c09f | 605 | ASSERT(refcount_count(&spa->spa_refcount) > spa->spa_minref || |
34dc7c2f | 606 | MUTEX_HELD(&spa_namespace_lock)); |
34dc7c2f BB |
607 | (void) refcount_remove(&spa->spa_refcount, tag); |
608 | } | |
609 | ||
610 | /* | |
611 | * Check to see if the spa refcount is zero. Must be called with | |
b128c09f | 612 | * spa_namespace_lock held. We really compare against spa_minref, which is the |
34dc7c2f BB |
613 | * number of references acquired when opening a pool |
614 | */ | |
615 | boolean_t | |
616 | spa_refcount_zero(spa_t *spa) | |
617 | { | |
618 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
619 | ||
b128c09f | 620 | return (refcount_count(&spa->spa_refcount) == spa->spa_minref); |
34dc7c2f BB |
621 | } |
622 | ||
623 | /* | |
624 | * ========================================================================== | |
625 | * SPA spare and l2cache tracking | |
626 | * ========================================================================== | |
627 | */ | |
628 | ||
629 | /* | |
630 | * Hot spares and cache devices are tracked using the same code below, | |
631 | * for 'auxiliary' devices. | |
632 | */ | |
633 | ||
634 | typedef struct spa_aux { | |
635 | uint64_t aux_guid; | |
636 | uint64_t aux_pool; | |
637 | avl_node_t aux_avl; | |
638 | int aux_count; | |
639 | } spa_aux_t; | |
640 | ||
641 | static int | |
642 | spa_aux_compare(const void *a, const void *b) | |
643 | { | |
644 | const spa_aux_t *sa = a; | |
645 | const spa_aux_t *sb = b; | |
646 | ||
647 | if (sa->aux_guid < sb->aux_guid) | |
648 | return (-1); | |
649 | else if (sa->aux_guid > sb->aux_guid) | |
650 | return (1); | |
651 | else | |
652 | return (0); | |
653 | } | |
654 | ||
655 | void | |
656 | spa_aux_add(vdev_t *vd, avl_tree_t *avl) | |
657 | { | |
658 | avl_index_t where; | |
659 | spa_aux_t search; | |
660 | spa_aux_t *aux; | |
661 | ||
662 | search.aux_guid = vd->vdev_guid; | |
663 | if ((aux = avl_find(avl, &search, &where)) != NULL) { | |
664 | aux->aux_count++; | |
665 | } else { | |
b8d06fca | 666 | aux = kmem_zalloc(sizeof (spa_aux_t), KM_PUSHPAGE); |
34dc7c2f BB |
667 | aux->aux_guid = vd->vdev_guid; |
668 | aux->aux_count = 1; | |
669 | avl_insert(avl, aux, where); | |
670 | } | |
671 | } | |
672 | ||
673 | void | |
674 | spa_aux_remove(vdev_t *vd, avl_tree_t *avl) | |
675 | { | |
676 | spa_aux_t search; | |
677 | spa_aux_t *aux; | |
678 | avl_index_t where; | |
679 | ||
680 | search.aux_guid = vd->vdev_guid; | |
681 | aux = avl_find(avl, &search, &where); | |
682 | ||
683 | ASSERT(aux != NULL); | |
684 | ||
685 | if (--aux->aux_count == 0) { | |
686 | avl_remove(avl, aux); | |
687 | kmem_free(aux, sizeof (spa_aux_t)); | |
688 | } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { | |
689 | aux->aux_pool = 0ULL; | |
690 | } | |
691 | } | |
692 | ||
693 | boolean_t | |
b128c09f | 694 | spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl) |
34dc7c2f BB |
695 | { |
696 | spa_aux_t search, *found; | |
34dc7c2f BB |
697 | |
698 | search.aux_guid = guid; | |
b128c09f | 699 | found = avl_find(avl, &search, NULL); |
34dc7c2f BB |
700 | |
701 | if (pool) { | |
702 | if (found) | |
703 | *pool = found->aux_pool; | |
704 | else | |
705 | *pool = 0ULL; | |
706 | } | |
707 | ||
b128c09f BB |
708 | if (refcnt) { |
709 | if (found) | |
710 | *refcnt = found->aux_count; | |
711 | else | |
712 | *refcnt = 0; | |
713 | } | |
714 | ||
34dc7c2f BB |
715 | return (found != NULL); |
716 | } | |
717 | ||
718 | void | |
719 | spa_aux_activate(vdev_t *vd, avl_tree_t *avl) | |
720 | { | |
721 | spa_aux_t search, *found; | |
722 | avl_index_t where; | |
723 | ||
724 | search.aux_guid = vd->vdev_guid; | |
725 | found = avl_find(avl, &search, &where); | |
726 | ASSERT(found != NULL); | |
727 | ASSERT(found->aux_pool == 0ULL); | |
728 | ||
729 | found->aux_pool = spa_guid(vd->vdev_spa); | |
730 | } | |
731 | ||
732 | /* | |
733 | * Spares are tracked globally due to the following constraints: | |
734 | * | |
735 | * - A spare may be part of multiple pools. | |
736 | * - A spare may be added to a pool even if it's actively in use within | |
737 | * another pool. | |
738 | * - A spare in use in any pool can only be the source of a replacement if | |
739 | * the target is a spare in the same pool. | |
740 | * | |
741 | * We keep track of all spares on the system through the use of a reference | |
742 | * counted AVL tree. When a vdev is added as a spare, or used as a replacement | |
743 | * spare, then we bump the reference count in the AVL tree. In addition, we set | |
744 | * the 'vdev_isspare' member to indicate that the device is a spare (active or | |
745 | * inactive). When a spare is made active (used to replace a device in the | |
746 | * pool), we also keep track of which pool its been made a part of. | |
747 | * | |
748 | * The 'spa_spare_lock' protects the AVL tree. These functions are normally | |
749 | * called under the spa_namespace lock as part of vdev reconfiguration. The | |
750 | * separate spare lock exists for the status query path, which does not need to | |
751 | * be completely consistent with respect to other vdev configuration changes. | |
752 | */ | |
753 | ||
754 | static int | |
755 | spa_spare_compare(const void *a, const void *b) | |
756 | { | |
757 | return (spa_aux_compare(a, b)); | |
758 | } | |
759 | ||
760 | void | |
761 | spa_spare_add(vdev_t *vd) | |
762 | { | |
763 | mutex_enter(&spa_spare_lock); | |
764 | ASSERT(!vd->vdev_isspare); | |
765 | spa_aux_add(vd, &spa_spare_avl); | |
766 | vd->vdev_isspare = B_TRUE; | |
767 | mutex_exit(&spa_spare_lock); | |
768 | } | |
769 | ||
770 | void | |
771 | spa_spare_remove(vdev_t *vd) | |
772 | { | |
773 | mutex_enter(&spa_spare_lock); | |
774 | ASSERT(vd->vdev_isspare); | |
775 | spa_aux_remove(vd, &spa_spare_avl); | |
776 | vd->vdev_isspare = B_FALSE; | |
777 | mutex_exit(&spa_spare_lock); | |
778 | } | |
779 | ||
780 | boolean_t | |
b128c09f | 781 | spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt) |
34dc7c2f BB |
782 | { |
783 | boolean_t found; | |
784 | ||
785 | mutex_enter(&spa_spare_lock); | |
b128c09f | 786 | found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl); |
34dc7c2f BB |
787 | mutex_exit(&spa_spare_lock); |
788 | ||
789 | return (found); | |
790 | } | |
791 | ||
792 | void | |
793 | spa_spare_activate(vdev_t *vd) | |
794 | { | |
795 | mutex_enter(&spa_spare_lock); | |
796 | ASSERT(vd->vdev_isspare); | |
797 | spa_aux_activate(vd, &spa_spare_avl); | |
798 | mutex_exit(&spa_spare_lock); | |
799 | } | |
800 | ||
801 | /* | |
802 | * Level 2 ARC devices are tracked globally for the same reasons as spares. | |
803 | * Cache devices currently only support one pool per cache device, and so | |
804 | * for these devices the aux reference count is currently unused beyond 1. | |
805 | */ | |
806 | ||
807 | static int | |
808 | spa_l2cache_compare(const void *a, const void *b) | |
809 | { | |
810 | return (spa_aux_compare(a, b)); | |
811 | } | |
812 | ||
813 | void | |
814 | spa_l2cache_add(vdev_t *vd) | |
815 | { | |
816 | mutex_enter(&spa_l2cache_lock); | |
817 | ASSERT(!vd->vdev_isl2cache); | |
818 | spa_aux_add(vd, &spa_l2cache_avl); | |
819 | vd->vdev_isl2cache = B_TRUE; | |
820 | mutex_exit(&spa_l2cache_lock); | |
821 | } | |
822 | ||
823 | void | |
824 | spa_l2cache_remove(vdev_t *vd) | |
825 | { | |
826 | mutex_enter(&spa_l2cache_lock); | |
827 | ASSERT(vd->vdev_isl2cache); | |
828 | spa_aux_remove(vd, &spa_l2cache_avl); | |
829 | vd->vdev_isl2cache = B_FALSE; | |
830 | mutex_exit(&spa_l2cache_lock); | |
831 | } | |
832 | ||
833 | boolean_t | |
834 | spa_l2cache_exists(uint64_t guid, uint64_t *pool) | |
835 | { | |
836 | boolean_t found; | |
837 | ||
838 | mutex_enter(&spa_l2cache_lock); | |
b128c09f | 839 | found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl); |
34dc7c2f BB |
840 | mutex_exit(&spa_l2cache_lock); |
841 | ||
842 | return (found); | |
843 | } | |
844 | ||
845 | void | |
846 | spa_l2cache_activate(vdev_t *vd) | |
847 | { | |
848 | mutex_enter(&spa_l2cache_lock); | |
849 | ASSERT(vd->vdev_isl2cache); | |
850 | spa_aux_activate(vd, &spa_l2cache_avl); | |
851 | mutex_exit(&spa_l2cache_lock); | |
852 | } | |
853 | ||
34dc7c2f BB |
854 | /* |
855 | * ========================================================================== | |
856 | * SPA vdev locking | |
857 | * ========================================================================== | |
858 | */ | |
859 | ||
860 | /* | |
861 | * Lock the given spa_t for the purpose of adding or removing a vdev. | |
862 | * Grabs the global spa_namespace_lock plus the spa config lock for writing. | |
863 | * It returns the next transaction group for the spa_t. | |
864 | */ | |
865 | uint64_t | |
866 | spa_vdev_enter(spa_t *spa) | |
867 | { | |
428870ff | 868 | mutex_enter(&spa->spa_vdev_top_lock); |
34dc7c2f | 869 | mutex_enter(&spa_namespace_lock); |
428870ff BB |
870 | return (spa_vdev_config_enter(spa)); |
871 | } | |
872 | ||
873 | /* | |
874 | * Internal implementation for spa_vdev_enter(). Used when a vdev | |
875 | * operation requires multiple syncs (i.e. removing a device) while | |
876 | * keeping the spa_namespace_lock held. | |
877 | */ | |
878 | uint64_t | |
879 | spa_vdev_config_enter(spa_t *spa) | |
880 | { | |
881 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
34dc7c2f | 882 | |
b128c09f | 883 | spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); |
34dc7c2f BB |
884 | |
885 | return (spa_last_synced_txg(spa) + 1); | |
886 | } | |
887 | ||
888 | /* | |
428870ff BB |
889 | * Used in combination with spa_vdev_config_enter() to allow the syncing |
890 | * of multiple transactions without releasing the spa_namespace_lock. | |
34dc7c2f | 891 | */ |
428870ff BB |
892 | void |
893 | spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag) | |
34dc7c2f BB |
894 | { |
895 | int config_changed = B_FALSE; | |
896 | ||
d6320ddb | 897 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); |
34dc7c2f BB |
898 | ASSERT(txg > spa_last_synced_txg(spa)); |
899 | ||
b128c09f BB |
900 | spa->spa_pending_vdev = NULL; |
901 | ||
34dc7c2f BB |
902 | /* |
903 | * Reassess the DTLs. | |
904 | */ | |
905 | vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); | |
906 | ||
b128c09f | 907 | if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) { |
34dc7c2f | 908 | config_changed = B_TRUE; |
428870ff | 909 | spa->spa_config_generation++; |
34dc7c2f BB |
910 | } |
911 | ||
428870ff BB |
912 | /* |
913 | * Verify the metaslab classes. | |
914 | */ | |
915 | ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0); | |
916 | ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0); | |
917 | ||
b128c09f | 918 | spa_config_exit(spa, SCL_ALL, spa); |
34dc7c2f | 919 | |
428870ff BB |
920 | /* |
921 | * Panic the system if the specified tag requires it. This | |
922 | * is useful for ensuring that configurations are updated | |
923 | * transactionally. | |
924 | */ | |
925 | if (zio_injection_enabled) | |
926 | zio_handle_panic_injection(spa, tag, 0); | |
927 | ||
34dc7c2f BB |
928 | /* |
929 | * Note: this txg_wait_synced() is important because it ensures | |
930 | * that there won't be more than one config change per txg. | |
931 | * This allows us to use the txg as the generation number. | |
932 | */ | |
933 | if (error == 0) | |
934 | txg_wait_synced(spa->spa_dsl_pool, txg); | |
935 | ||
936 | if (vd != NULL) { | |
fb5f0bc8 BB |
937 | ASSERT(!vd->vdev_detached || vd->vdev_dtl_smo.smo_object == 0); |
938 | spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); | |
34dc7c2f | 939 | vdev_free(vd); |
fb5f0bc8 | 940 | spa_config_exit(spa, SCL_ALL, spa); |
34dc7c2f BB |
941 | } |
942 | ||
943 | /* | |
944 | * If the config changed, update the config cache. | |
945 | */ | |
946 | if (config_changed) | |
b128c09f | 947 | spa_config_sync(spa, B_FALSE, B_TRUE); |
428870ff | 948 | } |
34dc7c2f | 949 | |
428870ff BB |
950 | /* |
951 | * Unlock the spa_t after adding or removing a vdev. Besides undoing the | |
952 | * locking of spa_vdev_enter(), we also want make sure the transactions have | |
953 | * synced to disk, and then update the global configuration cache with the new | |
954 | * information. | |
955 | */ | |
956 | int | |
957 | spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) | |
958 | { | |
959 | spa_vdev_config_exit(spa, vd, txg, error, FTAG); | |
34dc7c2f | 960 | mutex_exit(&spa_namespace_lock); |
428870ff | 961 | mutex_exit(&spa->spa_vdev_top_lock); |
34dc7c2f BB |
962 | |
963 | return (error); | |
964 | } | |
965 | ||
b128c09f BB |
966 | /* |
967 | * Lock the given spa_t for the purpose of changing vdev state. | |
968 | */ | |
969 | void | |
428870ff | 970 | spa_vdev_state_enter(spa_t *spa, int oplocks) |
b128c09f | 971 | { |
428870ff BB |
972 | int locks = SCL_STATE_ALL | oplocks; |
973 | ||
974 | /* | |
975 | * Root pools may need to read of the underlying devfs filesystem | |
976 | * when opening up a vdev. Unfortunately if we're holding the | |
977 | * SCL_ZIO lock it will result in a deadlock when we try to issue | |
978 | * the read from the root filesystem. Instead we "prefetch" | |
979 | * the associated vnodes that we need prior to opening the | |
980 | * underlying devices and cache them so that we can prevent | |
981 | * any I/O when we are doing the actual open. | |
982 | */ | |
983 | if (spa_is_root(spa)) { | |
984 | int low = locks & ~(SCL_ZIO - 1); | |
985 | int high = locks & ~low; | |
986 | ||
987 | spa_config_enter(spa, high, spa, RW_WRITER); | |
988 | vdev_hold(spa->spa_root_vdev); | |
989 | spa_config_enter(spa, low, spa, RW_WRITER); | |
990 | } else { | |
991 | spa_config_enter(spa, locks, spa, RW_WRITER); | |
992 | } | |
993 | spa->spa_vdev_locks = locks; | |
b128c09f BB |
994 | } |
995 | ||
996 | int | |
997 | spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error) | |
998 | { | |
428870ff BB |
999 | boolean_t config_changed = B_FALSE; |
1000 | ||
1001 | if (vd != NULL || error == 0) | |
1002 | vdev_dtl_reassess(vd ? vd->vdev_top : spa->spa_root_vdev, | |
1003 | 0, 0, B_FALSE); | |
1004 | ||
1005 | if (vd != NULL) { | |
b128c09f | 1006 | vdev_state_dirty(vd->vdev_top); |
428870ff BB |
1007 | config_changed = B_TRUE; |
1008 | spa->spa_config_generation++; | |
1009 | } | |
b128c09f | 1010 | |
428870ff BB |
1011 | if (spa_is_root(spa)) |
1012 | vdev_rele(spa->spa_root_vdev); | |
1013 | ||
1014 | ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL); | |
1015 | spa_config_exit(spa, spa->spa_vdev_locks, spa); | |
b128c09f | 1016 | |
fb5f0bc8 BB |
1017 | /* |
1018 | * If anything changed, wait for it to sync. This ensures that, | |
1019 | * from the system administrator's perspective, zpool(1M) commands | |
1020 | * are synchronous. This is important for things like zpool offline: | |
1021 | * when the command completes, you expect no further I/O from ZFS. | |
1022 | */ | |
1023 | if (vd != NULL) | |
1024 | txg_wait_synced(spa->spa_dsl_pool, 0); | |
1025 | ||
428870ff BB |
1026 | /* |
1027 | * If the config changed, update the config cache. | |
1028 | */ | |
1029 | if (config_changed) { | |
1030 | mutex_enter(&spa_namespace_lock); | |
1031 | spa_config_sync(spa, B_FALSE, B_TRUE); | |
1032 | mutex_exit(&spa_namespace_lock); | |
1033 | } | |
1034 | ||
b128c09f BB |
1035 | return (error); |
1036 | } | |
1037 | ||
34dc7c2f BB |
1038 | /* |
1039 | * ========================================================================== | |
1040 | * Miscellaneous functions | |
1041 | * ========================================================================== | |
1042 | */ | |
1043 | ||
9ae529ec CS |
1044 | void |
1045 | spa_activate_mos_feature(spa_t *spa, const char *feature) | |
1046 | { | |
1047 | (void) nvlist_add_boolean(spa->spa_label_features, feature); | |
1048 | vdev_config_dirty(spa->spa_root_vdev); | |
1049 | } | |
1050 | ||
1051 | void | |
1052 | spa_deactivate_mos_feature(spa_t *spa, const char *feature) | |
1053 | { | |
1054 | (void) nvlist_remove_all(spa->spa_label_features, feature); | |
1055 | vdev_config_dirty(spa->spa_root_vdev); | |
1056 | } | |
1057 | ||
34dc7c2f BB |
1058 | /* |
1059 | * Rename a spa_t. | |
1060 | */ | |
1061 | int | |
1062 | spa_rename(const char *name, const char *newname) | |
1063 | { | |
1064 | spa_t *spa; | |
1065 | int err; | |
1066 | ||
1067 | /* | |
1068 | * Lookup the spa_t and grab the config lock for writing. We need to | |
1069 | * actually open the pool so that we can sync out the necessary labels. | |
1070 | * It's OK to call spa_open() with the namespace lock held because we | |
1071 | * allow recursive calls for other reasons. | |
1072 | */ | |
1073 | mutex_enter(&spa_namespace_lock); | |
1074 | if ((err = spa_open(name, &spa, FTAG)) != 0) { | |
1075 | mutex_exit(&spa_namespace_lock); | |
1076 | return (err); | |
1077 | } | |
1078 | ||
b128c09f | 1079 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f BB |
1080 | |
1081 | avl_remove(&spa_namespace_avl, spa); | |
b128c09f | 1082 | (void) strlcpy(spa->spa_name, newname, sizeof (spa->spa_name)); |
34dc7c2f BB |
1083 | avl_add(&spa_namespace_avl, spa); |
1084 | ||
1085 | /* | |
1086 | * Sync all labels to disk with the new names by marking the root vdev | |
1087 | * dirty and waiting for it to sync. It will pick up the new pool name | |
1088 | * during the sync. | |
1089 | */ | |
1090 | vdev_config_dirty(spa->spa_root_vdev); | |
1091 | ||
b128c09f | 1092 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
1093 | |
1094 | txg_wait_synced(spa->spa_dsl_pool, 0); | |
1095 | ||
1096 | /* | |
1097 | * Sync the updated config cache. | |
1098 | */ | |
b128c09f | 1099 | spa_config_sync(spa, B_FALSE, B_TRUE); |
34dc7c2f BB |
1100 | |
1101 | spa_close(spa, FTAG); | |
1102 | ||
1103 | mutex_exit(&spa_namespace_lock); | |
1104 | ||
1105 | return (0); | |
1106 | } | |
1107 | ||
34dc7c2f | 1108 | /* |
572e2857 BB |
1109 | * Return the spa_t associated with given pool_guid, if it exists. If |
1110 | * device_guid is non-zero, determine whether the pool exists *and* contains | |
1111 | * a device with the specified device_guid. | |
34dc7c2f | 1112 | */ |
572e2857 BB |
1113 | spa_t * |
1114 | spa_by_guid(uint64_t pool_guid, uint64_t device_guid) | |
34dc7c2f BB |
1115 | { |
1116 | spa_t *spa; | |
1117 | avl_tree_t *t = &spa_namespace_avl; | |
1118 | ||
1119 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
1120 | ||
1121 | for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { | |
1122 | if (spa->spa_state == POOL_STATE_UNINITIALIZED) | |
1123 | continue; | |
1124 | if (spa->spa_root_vdev == NULL) | |
1125 | continue; | |
1126 | if (spa_guid(spa) == pool_guid) { | |
1127 | if (device_guid == 0) | |
1128 | break; | |
1129 | ||
1130 | if (vdev_lookup_by_guid(spa->spa_root_vdev, | |
1131 | device_guid) != NULL) | |
1132 | break; | |
1133 | ||
1134 | /* | |
1135 | * Check any devices we may be in the process of adding. | |
1136 | */ | |
1137 | if (spa->spa_pending_vdev) { | |
1138 | if (vdev_lookup_by_guid(spa->spa_pending_vdev, | |
1139 | device_guid) != NULL) | |
1140 | break; | |
1141 | } | |
1142 | } | |
1143 | } | |
1144 | ||
572e2857 BB |
1145 | return (spa); |
1146 | } | |
1147 | ||
1148 | /* | |
1149 | * Determine whether a pool with the given pool_guid exists. | |
1150 | */ | |
1151 | boolean_t | |
1152 | spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) | |
1153 | { | |
1154 | return (spa_by_guid(pool_guid, device_guid) != NULL); | |
34dc7c2f BB |
1155 | } |
1156 | ||
1157 | char * | |
1158 | spa_strdup(const char *s) | |
1159 | { | |
1160 | size_t len; | |
1161 | char *new; | |
1162 | ||
1163 | len = strlen(s); | |
b8d06fca | 1164 | new = kmem_alloc(len + 1, KM_PUSHPAGE); |
34dc7c2f BB |
1165 | bcopy(s, new, len); |
1166 | new[len] = '\0'; | |
1167 | ||
1168 | return (new); | |
1169 | } | |
1170 | ||
1171 | void | |
1172 | spa_strfree(char *s) | |
1173 | { | |
1174 | kmem_free(s, strlen(s) + 1); | |
1175 | } | |
1176 | ||
1177 | uint64_t | |
1178 | spa_get_random(uint64_t range) | |
1179 | { | |
1180 | uint64_t r; | |
1181 | ||
1182 | ASSERT(range != 0); | |
1183 | ||
1184 | (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); | |
1185 | ||
1186 | return (r % range); | |
1187 | } | |
1188 | ||
428870ff BB |
1189 | uint64_t |
1190 | spa_generate_guid(spa_t *spa) | |
34dc7c2f | 1191 | { |
428870ff | 1192 | uint64_t guid = spa_get_random(-1ULL); |
34dc7c2f | 1193 | |
428870ff BB |
1194 | if (spa != NULL) { |
1195 | while (guid == 0 || spa_guid_exists(spa_guid(spa), guid)) | |
1196 | guid = spa_get_random(-1ULL); | |
1197 | } else { | |
1198 | while (guid == 0 || spa_guid_exists(guid, 0)) | |
1199 | guid = spa_get_random(-1ULL); | |
34dc7c2f BB |
1200 | } |
1201 | ||
428870ff BB |
1202 | return (guid); |
1203 | } | |
1204 | ||
1205 | void | |
1206 | sprintf_blkptr(char *buf, const blkptr_t *bp) | |
1207 | { | |
9ae529ec | 1208 | char type[256]; |
428870ff BB |
1209 | char *checksum = NULL; |
1210 | char *compress = NULL; | |
34dc7c2f | 1211 | |
428870ff | 1212 | if (bp != NULL) { |
9ae529ec CS |
1213 | if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) { |
1214 | dmu_object_byteswap_t bswap = | |
1215 | DMU_OT_BYTESWAP(BP_GET_TYPE(bp)); | |
1216 | (void) snprintf(type, sizeof (type), "bswap %s %s", | |
1217 | DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ? | |
1218 | "metadata" : "data", | |
1219 | dmu_ot_byteswap[bswap].ob_name); | |
1220 | } else { | |
1221 | (void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name, | |
1222 | sizeof (type)); | |
1223 | } | |
428870ff BB |
1224 | checksum = zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name; |
1225 | compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name; | |
34dc7c2f BB |
1226 | } |
1227 | ||
428870ff | 1228 | SPRINTF_BLKPTR(snprintf, ' ', buf, bp, type, checksum, compress); |
34dc7c2f BB |
1229 | } |
1230 | ||
1231 | void | |
1232 | spa_freeze(spa_t *spa) | |
1233 | { | |
1234 | uint64_t freeze_txg = 0; | |
1235 | ||
b128c09f | 1236 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f BB |
1237 | if (spa->spa_freeze_txg == UINT64_MAX) { |
1238 | freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; | |
1239 | spa->spa_freeze_txg = freeze_txg; | |
1240 | } | |
b128c09f | 1241 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
1242 | if (freeze_txg != 0) |
1243 | txg_wait_synced(spa_get_dsl(spa), freeze_txg); | |
1244 | } | |
1245 | ||
428870ff BB |
1246 | /* |
1247 | * This is a stripped-down version of strtoull, suitable only for converting | |
1248 | * lowercase hexidecimal numbers that don't overflow. | |
1249 | */ | |
1250 | uint64_t | |
1251 | strtonum(const char *str, char **nptr) | |
1252 | { | |
1253 | uint64_t val = 0; | |
1254 | char c; | |
1255 | int digit; | |
1256 | ||
1257 | while ((c = *str) != '\0') { | |
1258 | if (c >= '0' && c <= '9') | |
1259 | digit = c - '0'; | |
1260 | else if (c >= 'a' && c <= 'f') | |
1261 | digit = 10 + c - 'a'; | |
1262 | else | |
1263 | break; | |
1264 | ||
1265 | val *= 16; | |
1266 | val += digit; | |
1267 | ||
1268 | str++; | |
1269 | } | |
1270 | ||
1271 | if (nptr) | |
1272 | *nptr = (char *)str; | |
1273 | ||
1274 | return (val); | |
1275 | } | |
1276 | ||
34dc7c2f BB |
1277 | /* |
1278 | * ========================================================================== | |
1279 | * Accessor functions | |
1280 | * ========================================================================== | |
1281 | */ | |
1282 | ||
b128c09f BB |
1283 | boolean_t |
1284 | spa_shutting_down(spa_t *spa) | |
34dc7c2f | 1285 | { |
b128c09f | 1286 | return (spa->spa_async_suspended); |
34dc7c2f BB |
1287 | } |
1288 | ||
1289 | dsl_pool_t * | |
1290 | spa_get_dsl(spa_t *spa) | |
1291 | { | |
1292 | return (spa->spa_dsl_pool); | |
1293 | } | |
1294 | ||
9ae529ec CS |
1295 | boolean_t |
1296 | spa_is_initializing(spa_t *spa) | |
1297 | { | |
1298 | return (spa->spa_is_initializing); | |
1299 | } | |
1300 | ||
34dc7c2f BB |
1301 | blkptr_t * |
1302 | spa_get_rootblkptr(spa_t *spa) | |
1303 | { | |
1304 | return (&spa->spa_ubsync.ub_rootbp); | |
1305 | } | |
1306 | ||
1307 | void | |
1308 | spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) | |
1309 | { | |
1310 | spa->spa_uberblock.ub_rootbp = *bp; | |
1311 | } | |
1312 | ||
1313 | void | |
1314 | spa_altroot(spa_t *spa, char *buf, size_t buflen) | |
1315 | { | |
1316 | if (spa->spa_root == NULL) | |
1317 | buf[0] = '\0'; | |
1318 | else | |
1319 | (void) strncpy(buf, spa->spa_root, buflen); | |
1320 | } | |
1321 | ||
1322 | int | |
1323 | spa_sync_pass(spa_t *spa) | |
1324 | { | |
1325 | return (spa->spa_sync_pass); | |
1326 | } | |
1327 | ||
1328 | char * | |
1329 | spa_name(spa_t *spa) | |
1330 | { | |
34dc7c2f BB |
1331 | return (spa->spa_name); |
1332 | } | |
1333 | ||
1334 | uint64_t | |
1335 | spa_guid(spa_t *spa) | |
1336 | { | |
3bc7e0fb GW |
1337 | dsl_pool_t *dp = spa_get_dsl(spa); |
1338 | uint64_t guid; | |
1339 | ||
34dc7c2f BB |
1340 | /* |
1341 | * If we fail to parse the config during spa_load(), we can go through | |
1342 | * the error path (which posts an ereport) and end up here with no root | |
3541dc6d | 1343 | * vdev. We stash the original pool guid in 'spa_config_guid' to handle |
34dc7c2f BB |
1344 | * this case. |
1345 | */ | |
3bc7e0fb GW |
1346 | if (spa->spa_root_vdev == NULL) |
1347 | return (spa->spa_config_guid); | |
1348 | ||
1349 | guid = spa->spa_last_synced_guid != 0 ? | |
1350 | spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid; | |
1351 | ||
1352 | /* | |
1353 | * Return the most recently synced out guid unless we're | |
1354 | * in syncing context. | |
1355 | */ | |
1356 | if (dp && dsl_pool_sync_context(dp)) | |
34dc7c2f BB |
1357 | return (spa->spa_root_vdev->vdev_guid); |
1358 | else | |
3bc7e0fb | 1359 | return (guid); |
3541dc6d GA |
1360 | } |
1361 | ||
1362 | uint64_t | |
1363 | spa_load_guid(spa_t *spa) | |
1364 | { | |
1365 | /* | |
1366 | * This is a GUID that exists solely as a reference for the | |
1367 | * purposes of the arc. It is generated at load time, and | |
1368 | * is never written to persistent storage. | |
1369 | */ | |
1370 | return (spa->spa_load_guid); | |
34dc7c2f BB |
1371 | } |
1372 | ||
1373 | uint64_t | |
1374 | spa_last_synced_txg(spa_t *spa) | |
1375 | { | |
1376 | return (spa->spa_ubsync.ub_txg); | |
1377 | } | |
1378 | ||
1379 | uint64_t | |
1380 | spa_first_txg(spa_t *spa) | |
1381 | { | |
1382 | return (spa->spa_first_txg); | |
1383 | } | |
1384 | ||
428870ff BB |
1385 | uint64_t |
1386 | spa_syncing_txg(spa_t *spa) | |
1387 | { | |
1388 | return (spa->spa_syncing_txg); | |
1389 | } | |
1390 | ||
b128c09f | 1391 | pool_state_t |
34dc7c2f BB |
1392 | spa_state(spa_t *spa) |
1393 | { | |
1394 | return (spa->spa_state); | |
1395 | } | |
1396 | ||
428870ff BB |
1397 | spa_load_state_t |
1398 | spa_load_state(spa_t *spa) | |
34dc7c2f | 1399 | { |
428870ff | 1400 | return (spa->spa_load_state); |
34dc7c2f BB |
1401 | } |
1402 | ||
34dc7c2f | 1403 | uint64_t |
428870ff | 1404 | spa_freeze_txg(spa_t *spa) |
34dc7c2f | 1405 | { |
428870ff | 1406 | return (spa->spa_freeze_txg); |
34dc7c2f BB |
1407 | } |
1408 | ||
428870ff | 1409 | /* ARGSUSED */ |
34dc7c2f | 1410 | uint64_t |
428870ff | 1411 | spa_get_asize(spa_t *spa, uint64_t lsize) |
34dc7c2f | 1412 | { |
428870ff BB |
1413 | /* |
1414 | * The worst case is single-sector max-parity RAID-Z blocks, in which | |
1415 | * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1) | |
1416 | * times the size; so just assume that. Add to this the fact that | |
1417 | * we can have up to 3 DVAs per bp, and one more factor of 2 because | |
1418 | * the block may be dittoed with up to 3 DVAs by ddt_sync(). | |
1419 | */ | |
1420 | return (lsize * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2); | |
34dc7c2f BB |
1421 | } |
1422 | ||
34dc7c2f BB |
1423 | uint64_t |
1424 | spa_get_dspace(spa_t *spa) | |
1425 | { | |
428870ff | 1426 | return (spa->spa_dspace); |
34dc7c2f BB |
1427 | } |
1428 | ||
428870ff BB |
1429 | void |
1430 | spa_update_dspace(spa_t *spa) | |
34dc7c2f | 1431 | { |
428870ff BB |
1432 | spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) + |
1433 | ddt_get_dedup_dspace(spa); | |
34dc7c2f BB |
1434 | } |
1435 | ||
1436 | /* | |
1437 | * Return the failure mode that has been set to this pool. The default | |
1438 | * behavior will be to block all I/Os when a complete failure occurs. | |
1439 | */ | |
1440 | uint8_t | |
1441 | spa_get_failmode(spa_t *spa) | |
1442 | { | |
1443 | return (spa->spa_failmode); | |
1444 | } | |
1445 | ||
b128c09f BB |
1446 | boolean_t |
1447 | spa_suspended(spa_t *spa) | |
1448 | { | |
1449 | return (spa->spa_suspended); | |
1450 | } | |
1451 | ||
34dc7c2f BB |
1452 | uint64_t |
1453 | spa_version(spa_t *spa) | |
1454 | { | |
1455 | return (spa->spa_ubsync.ub_version); | |
1456 | } | |
1457 | ||
428870ff BB |
1458 | boolean_t |
1459 | spa_deflate(spa_t *spa) | |
1460 | { | |
1461 | return (spa->spa_deflate); | |
1462 | } | |
1463 | ||
1464 | metaslab_class_t * | |
1465 | spa_normal_class(spa_t *spa) | |
1466 | { | |
1467 | return (spa->spa_normal_class); | |
1468 | } | |
1469 | ||
1470 | metaslab_class_t * | |
1471 | spa_log_class(spa_t *spa) | |
1472 | { | |
1473 | return (spa->spa_log_class); | |
1474 | } | |
1475 | ||
34dc7c2f BB |
1476 | int |
1477 | spa_max_replication(spa_t *spa) | |
1478 | { | |
1479 | /* | |
1480 | * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to | |
1481 | * handle BPs with more than one DVA allocated. Set our max | |
1482 | * replication level accordingly. | |
1483 | */ | |
1484 | if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) | |
1485 | return (1); | |
1486 | return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); | |
1487 | } | |
1488 | ||
428870ff BB |
1489 | int |
1490 | spa_prev_software_version(spa_t *spa) | |
1491 | { | |
1492 | return (spa->spa_prev_software_version); | |
1493 | } | |
1494 | ||
34dc7c2f | 1495 | uint64_t |
428870ff | 1496 | dva_get_dsize_sync(spa_t *spa, const dva_t *dva) |
34dc7c2f | 1497 | { |
428870ff BB |
1498 | uint64_t asize = DVA_GET_ASIZE(dva); |
1499 | uint64_t dsize = asize; | |
34dc7c2f | 1500 | |
428870ff | 1501 | ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); |
34dc7c2f | 1502 | |
428870ff BB |
1503 | if (asize != 0 && spa->spa_deflate) { |
1504 | vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva)); | |
1505 | dsize = (asize >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio; | |
34dc7c2f | 1506 | } |
428870ff BB |
1507 | |
1508 | return (dsize); | |
1509 | } | |
1510 | ||
1511 | uint64_t | |
1512 | bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp) | |
1513 | { | |
1514 | uint64_t dsize = 0; | |
d6320ddb | 1515 | int d; |
428870ff | 1516 | |
d6320ddb | 1517 | for (d = 0; d < SPA_DVAS_PER_BP; d++) |
428870ff BB |
1518 | dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); |
1519 | ||
1520 | return (dsize); | |
1521 | } | |
1522 | ||
1523 | uint64_t | |
1524 | bp_get_dsize(spa_t *spa, const blkptr_t *bp) | |
1525 | { | |
1526 | uint64_t dsize = 0; | |
d6320ddb | 1527 | int d; |
428870ff BB |
1528 | |
1529 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
1530 | ||
d6320ddb | 1531 | for (d = 0; d < SPA_DVAS_PER_BP; d++) |
428870ff BB |
1532 | dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); |
1533 | ||
b128c09f | 1534 | spa_config_exit(spa, SCL_VDEV, FTAG); |
428870ff BB |
1535 | |
1536 | return (dsize); | |
34dc7c2f BB |
1537 | } |
1538 | ||
1539 | /* | |
1540 | * ========================================================================== | |
1541 | * Initialization and Termination | |
1542 | * ========================================================================== | |
1543 | */ | |
1544 | ||
1545 | static int | |
1546 | spa_name_compare(const void *a1, const void *a2) | |
1547 | { | |
1548 | const spa_t *s1 = a1; | |
1549 | const spa_t *s2 = a2; | |
1550 | int s; | |
1551 | ||
1552 | s = strcmp(s1->spa_name, s2->spa_name); | |
1553 | if (s > 0) | |
1554 | return (1); | |
1555 | if (s < 0) | |
1556 | return (-1); | |
1557 | return (0); | |
1558 | } | |
1559 | ||
34dc7c2f | 1560 | void |
0bc8fd78 | 1561 | spa_boot_init(void) |
34dc7c2f BB |
1562 | { |
1563 | spa_config_load(); | |
1564 | } | |
1565 | ||
1566 | void | |
1567 | spa_init(int mode) | |
1568 | { | |
1569 | mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); | |
1570 | mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); | |
1571 | mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); | |
1572 | cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); | |
1573 | ||
1574 | avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), | |
1575 | offsetof(spa_t, spa_avl)); | |
1576 | ||
1577 | avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), | |
1578 | offsetof(spa_aux_t, aux_avl)); | |
1579 | ||
1580 | avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), | |
1581 | offsetof(spa_aux_t, aux_avl)); | |
1582 | ||
fb5f0bc8 | 1583 | spa_mode_global = mode; |
34dc7c2f | 1584 | |
26685276 | 1585 | fm_init(); |
34dc7c2f BB |
1586 | refcount_init(); |
1587 | unique_init(); | |
1588 | zio_init(); | |
1589 | dmu_init(); | |
1590 | zil_init(); | |
1591 | vdev_cache_stat_init(); | |
1592 | zfs_prop_init(); | |
1593 | zpool_prop_init(); | |
9ae529ec | 1594 | zpool_feature_init(); |
34dc7c2f | 1595 | spa_config_load(); |
b128c09f | 1596 | l2arc_start(); |
34dc7c2f BB |
1597 | } |
1598 | ||
1599 | void | |
1600 | spa_fini(void) | |
1601 | { | |
b128c09f BB |
1602 | l2arc_stop(); |
1603 | ||
34dc7c2f BB |
1604 | spa_evict_all(); |
1605 | ||
1606 | vdev_cache_stat_fini(); | |
1607 | zil_fini(); | |
1608 | dmu_fini(); | |
1609 | zio_fini(); | |
1610 | unique_fini(); | |
1611 | refcount_fini(); | |
26685276 | 1612 | fm_fini(); |
34dc7c2f BB |
1613 | |
1614 | avl_destroy(&spa_namespace_avl); | |
1615 | avl_destroy(&spa_spare_avl); | |
1616 | avl_destroy(&spa_l2cache_avl); | |
1617 | ||
1618 | cv_destroy(&spa_namespace_cv); | |
1619 | mutex_destroy(&spa_namespace_lock); | |
1620 | mutex_destroy(&spa_spare_lock); | |
1621 | mutex_destroy(&spa_l2cache_lock); | |
1622 | } | |
1623 | ||
1624 | /* | |
1625 | * Return whether this pool has slogs. No locking needed. | |
1626 | * It's not a problem if the wrong answer is returned as it's only for | |
1627 | * performance and not correctness | |
1628 | */ | |
1629 | boolean_t | |
1630 | spa_has_slogs(spa_t *spa) | |
1631 | { | |
1632 | return (spa->spa_log_class->mc_rotor != NULL); | |
1633 | } | |
b128c09f | 1634 | |
428870ff BB |
1635 | spa_log_state_t |
1636 | spa_get_log_state(spa_t *spa) | |
1637 | { | |
1638 | return (spa->spa_log_state); | |
1639 | } | |
1640 | ||
1641 | void | |
1642 | spa_set_log_state(spa_t *spa, spa_log_state_t state) | |
1643 | { | |
1644 | spa->spa_log_state = state; | |
1645 | } | |
1646 | ||
b128c09f BB |
1647 | boolean_t |
1648 | spa_is_root(spa_t *spa) | |
1649 | { | |
1650 | return (spa->spa_is_root); | |
1651 | } | |
fb5f0bc8 BB |
1652 | |
1653 | boolean_t | |
1654 | spa_writeable(spa_t *spa) | |
1655 | { | |
1656 | return (!!(spa->spa_mode & FWRITE)); | |
1657 | } | |
1658 | ||
1659 | int | |
1660 | spa_mode(spa_t *spa) | |
1661 | { | |
1662 | return (spa->spa_mode); | |
1663 | } | |
428870ff BB |
1664 | |
1665 | uint64_t | |
1666 | spa_bootfs(spa_t *spa) | |
1667 | { | |
1668 | return (spa->spa_bootfs); | |
1669 | } | |
1670 | ||
1671 | uint64_t | |
1672 | spa_delegation(spa_t *spa) | |
1673 | { | |
1674 | return (spa->spa_delegation); | |
1675 | } | |
1676 | ||
1677 | objset_t * | |
1678 | spa_meta_objset(spa_t *spa) | |
1679 | { | |
1680 | return (spa->spa_meta_objset); | |
1681 | } | |
1682 | ||
1683 | enum zio_checksum | |
1684 | spa_dedup_checksum(spa_t *spa) | |
1685 | { | |
1686 | return (spa->spa_dedup_checksum); | |
1687 | } | |
1688 | ||
1689 | /* | |
1690 | * Reset pool scan stat per scan pass (or reboot). | |
1691 | */ | |
1692 | void | |
1693 | spa_scan_stat_init(spa_t *spa) | |
1694 | { | |
1695 | /* data not stored on disk */ | |
1696 | spa->spa_scan_pass_start = gethrestime_sec(); | |
1697 | spa->spa_scan_pass_exam = 0; | |
1698 | vdev_scan_stat_init(spa->spa_root_vdev); | |
1699 | } | |
1700 | ||
1701 | /* | |
1702 | * Get scan stats for zpool status reports | |
1703 | */ | |
1704 | int | |
1705 | spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps) | |
1706 | { | |
1707 | dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL; | |
1708 | ||
1709 | if (scn == NULL || scn->scn_phys.scn_func == POOL_SCAN_NONE) | |
1710 | return (ENOENT); | |
1711 | bzero(ps, sizeof (pool_scan_stat_t)); | |
1712 | ||
1713 | /* data stored on disk */ | |
1714 | ps->pss_func = scn->scn_phys.scn_func; | |
1715 | ps->pss_start_time = scn->scn_phys.scn_start_time; | |
1716 | ps->pss_end_time = scn->scn_phys.scn_end_time; | |
1717 | ps->pss_to_examine = scn->scn_phys.scn_to_examine; | |
1718 | ps->pss_examined = scn->scn_phys.scn_examined; | |
1719 | ps->pss_to_process = scn->scn_phys.scn_to_process; | |
1720 | ps->pss_processed = scn->scn_phys.scn_processed; | |
1721 | ps->pss_errors = scn->scn_phys.scn_errors; | |
1722 | ps->pss_state = scn->scn_phys.scn_state; | |
1723 | ||
1724 | /* data not stored on disk */ | |
1725 | ps->pss_pass_start = spa->spa_scan_pass_start; | |
1726 | ps->pss_pass_exam = spa->spa_scan_pass_exam; | |
1727 | ||
1728 | return (0); | |
1729 | } | |
c28b2279 | 1730 | |
6d974228 GW |
1731 | boolean_t |
1732 | spa_debug_enabled(spa_t *spa) | |
1733 | { | |
1734 | return (spa->spa_debug); | |
1735 | } | |
1736 | ||
c28b2279 BB |
1737 | #if defined(_KERNEL) && defined(HAVE_SPL) |
1738 | /* Namespace manipulation */ | |
1739 | EXPORT_SYMBOL(spa_lookup); | |
1740 | EXPORT_SYMBOL(spa_add); | |
1741 | EXPORT_SYMBOL(spa_remove); | |
1742 | EXPORT_SYMBOL(spa_next); | |
1743 | ||
1744 | /* Refcount functions */ | |
1745 | EXPORT_SYMBOL(spa_open_ref); | |
1746 | EXPORT_SYMBOL(spa_close); | |
1747 | EXPORT_SYMBOL(spa_refcount_zero); | |
1748 | ||
1749 | /* Pool configuration lock */ | |
1750 | EXPORT_SYMBOL(spa_config_tryenter); | |
1751 | EXPORT_SYMBOL(spa_config_enter); | |
1752 | EXPORT_SYMBOL(spa_config_exit); | |
1753 | EXPORT_SYMBOL(spa_config_held); | |
1754 | ||
1755 | /* Pool vdev add/remove lock */ | |
1756 | EXPORT_SYMBOL(spa_vdev_enter); | |
1757 | EXPORT_SYMBOL(spa_vdev_exit); | |
1758 | ||
1759 | /* Pool vdev state change lock */ | |
1760 | EXPORT_SYMBOL(spa_vdev_state_enter); | |
1761 | EXPORT_SYMBOL(spa_vdev_state_exit); | |
1762 | ||
1763 | /* Accessor functions */ | |
1764 | EXPORT_SYMBOL(spa_shutting_down); | |
1765 | EXPORT_SYMBOL(spa_get_dsl); | |
1766 | EXPORT_SYMBOL(spa_get_rootblkptr); | |
1767 | EXPORT_SYMBOL(spa_set_rootblkptr); | |
1768 | EXPORT_SYMBOL(spa_altroot); | |
1769 | EXPORT_SYMBOL(spa_sync_pass); | |
1770 | EXPORT_SYMBOL(spa_name); | |
1771 | EXPORT_SYMBOL(spa_guid); | |
1772 | EXPORT_SYMBOL(spa_last_synced_txg); | |
1773 | EXPORT_SYMBOL(spa_first_txg); | |
1774 | EXPORT_SYMBOL(spa_syncing_txg); | |
1775 | EXPORT_SYMBOL(spa_version); | |
1776 | EXPORT_SYMBOL(spa_state); | |
1777 | EXPORT_SYMBOL(spa_load_state); | |
1778 | EXPORT_SYMBOL(spa_freeze_txg); | |
1779 | EXPORT_SYMBOL(spa_get_asize); | |
1780 | EXPORT_SYMBOL(spa_get_dspace); | |
1781 | EXPORT_SYMBOL(spa_update_dspace); | |
1782 | EXPORT_SYMBOL(spa_deflate); | |
1783 | EXPORT_SYMBOL(spa_normal_class); | |
1784 | EXPORT_SYMBOL(spa_log_class); | |
1785 | EXPORT_SYMBOL(spa_max_replication); | |
1786 | EXPORT_SYMBOL(spa_prev_software_version); | |
1787 | EXPORT_SYMBOL(spa_get_failmode); | |
1788 | EXPORT_SYMBOL(spa_suspended); | |
1789 | EXPORT_SYMBOL(spa_bootfs); | |
1790 | EXPORT_SYMBOL(spa_delegation); | |
1791 | EXPORT_SYMBOL(spa_meta_objset); | |
1792 | ||
1793 | /* Miscellaneous support routines */ | |
1794 | EXPORT_SYMBOL(spa_rename); | |
1795 | EXPORT_SYMBOL(spa_guid_exists); | |
1796 | EXPORT_SYMBOL(spa_strdup); | |
1797 | EXPORT_SYMBOL(spa_strfree); | |
1798 | EXPORT_SYMBOL(spa_get_random); | |
1799 | EXPORT_SYMBOL(spa_generate_guid); | |
1800 | EXPORT_SYMBOL(sprintf_blkptr); | |
1801 | EXPORT_SYMBOL(spa_freeze); | |
1802 | EXPORT_SYMBOL(spa_upgrade); | |
1803 | EXPORT_SYMBOL(spa_evict_all); | |
1804 | EXPORT_SYMBOL(spa_lookup_by_guid); | |
1805 | EXPORT_SYMBOL(spa_has_spare); | |
1806 | EXPORT_SYMBOL(dva_get_dsize_sync); | |
1807 | EXPORT_SYMBOL(bp_get_dsize_sync); | |
1808 | EXPORT_SYMBOL(bp_get_dsize); | |
1809 | EXPORT_SYMBOL(spa_has_slogs); | |
1810 | EXPORT_SYMBOL(spa_is_root); | |
1811 | EXPORT_SYMBOL(spa_writeable); | |
1812 | EXPORT_SYMBOL(spa_mode); | |
1813 | ||
1814 | EXPORT_SYMBOL(spa_namespace_lock); | |
1815 | #endif |