Print this page
3742 zfs comments need cleaner, more consistent style
Submitted by: Will Andrews <willa@spectralogic.com>
Submitted by: Alan Somers <alans@spectralogic.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Eric Schrock <eric.schrock@delphix.com>
| Split |
Close |
| Expand all |
| Collapse all |
--- old/usr/src/uts/common/fs/zfs/spa_misc.c
+++ new/usr/src/uts/common/fs/zfs/spa_misc.c
1 1 /*
2 2 * CDDL HEADER START
3 3 *
4 4 * The contents of this file are subject to the terms of the
5 5 * Common Development and Distribution License (the "License").
6 6 * You may not use this file except in compliance with the License.
7 7 *
8 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 9 * or http://www.opensolaris.org/os/licensing.
10 10 * See the License for the specific language governing permissions
11 11 * and limitations under the License.
12 12 *
13 13 * When distributing Covered Code, include this CDDL HEADER in each
14 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 15 * If applicable, add the following below this CDDL HEADER, with the
16 16 * fields enclosed by brackets "[]" replaced with your own identifying
17 17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 18 *
19 19 * CDDL HEADER END
20 20 */
21 21 /*
22 22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 23 * Copyright (c) 2013 by Delphix. All rights reserved.
24 24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 25 */
26 26
27 27 #include <sys/zfs_context.h>
28 28 #include <sys/spa_impl.h>
29 29 #include <sys/spa_boot.h>
30 30 #include <sys/zio.h>
31 31 #include <sys/zio_checksum.h>
32 32 #include <sys/zio_compress.h>
33 33 #include <sys/dmu.h>
34 34 #include <sys/dmu_tx.h>
35 35 #include <sys/zap.h>
36 36 #include <sys/zil.h>
37 37 #include <sys/vdev_impl.h>
38 38 #include <sys/metaslab.h>
39 39 #include <sys/uberblock_impl.h>
40 40 #include <sys/txg.h>
41 41 #include <sys/avl.h>
42 42 #include <sys/unique.h>
43 43 #include <sys/dsl_pool.h>
44 44 #include <sys/dsl_dir.h>
45 45 #include <sys/dsl_prop.h>
46 46 #include <sys/dsl_scan.h>
47 47 #include <sys/fs/zfs.h>
48 48 #include <sys/metaslab_impl.h>
49 49 #include <sys/arc.h>
50 50 #include <sys/ddt.h>
51 51 #include "zfs_prop.h"
52 52 #include "zfeature_common.h"
53 53
54 54 /*
55 55 * SPA locking
56 56 *
57 57 * There are four basic locks for managing spa_t structures:
58 58 *
59 59 * spa_namespace_lock (global mutex)
60 60 *
61 61 * This lock must be acquired to do any of the following:
62 62 *
63 63 * - Lookup a spa_t by name
64 64 * - Add or remove a spa_t from the namespace
65 65 * - Increase spa_refcount from non-zero
66 66 * - Check if spa_refcount is zero
67 67 * - Rename a spa_t
68 68 * - add/remove/attach/detach devices
69 69 * - Held for the duration of create/destroy/import/export
70 70 *
71 71 * It does not need to handle recursion. A create or destroy may
72 72 * reference objects (files or zvols) in other pools, but by
73 73 * definition they must have an existing reference, and will never need
74 74 * to lookup a spa_t by name.
75 75 *
76 76 * spa_refcount (per-spa refcount_t protected by mutex)
77 77 *
78 78 * This reference count keep track of any active users of the spa_t. The
79 79 * spa_t cannot be destroyed or freed while this is non-zero. Internally,
80 80 * the refcount is never really 'zero' - opening a pool implicitly keeps
81 81 * some references in the DMU. Internally we check against spa_minref, but
82 82 * present the image of a zero/non-zero value to consumers.
83 83 *
84 84 * spa_config_lock[] (per-spa array of rwlocks)
85 85 *
86 86 * This protects the spa_t from config changes, and must be held in
87 87 * the following circumstances:
88 88 *
89 89 * - RW_READER to perform I/O to the spa
90 90 * - RW_WRITER to change the vdev config
91 91 *
92 92 * The locking order is fairly straightforward:
93 93 *
94 94 * spa_namespace_lock -> spa_refcount
95 95 *
96 96 * The namespace lock must be acquired to increase the refcount from 0
97 97 * or to check if it is zero.
98 98 *
99 99 * spa_refcount -> spa_config_lock[]
100 100 *
101 101 * There must be at least one valid reference on the spa_t to acquire
102 102 * the config lock.
103 103 *
104 104 * spa_namespace_lock -> spa_config_lock[]
105 105 *
106 106 * The namespace lock must always be taken before the config lock.
107 107 *
108 108 *
109 109 * The spa_namespace_lock can be acquired directly and is globally visible.
110 110 *
111 111 * The namespace is manipulated using the following functions, all of which
112 112 * require the spa_namespace_lock to be held.
113 113 *
114 114 * spa_lookup() Lookup a spa_t by name.
115 115 *
116 116 * spa_add() Create a new spa_t in the namespace.
117 117 *
118 118 * spa_remove() Remove a spa_t from the namespace. This also
119 119 * frees up any memory associated with the spa_t.
120 120 *
121 121 * spa_next() Returns the next spa_t in the system, or the
122 122 * first if NULL is passed.
123 123 *
124 124 * spa_evict_all() Shutdown and remove all spa_t structures in
125 125 * the system.
126 126 *
127 127 * spa_guid_exists() Determine whether a pool/device guid exists.
128 128 *
129 129 * The spa_refcount is manipulated using the following functions:
130 130 *
131 131 * spa_open_ref() Adds a reference to the given spa_t. Must be
132 132 * called with spa_namespace_lock held if the
133 133 * refcount is currently zero.
134 134 *
135 135 * spa_close() Remove a reference from the spa_t. This will
136 136 * not free the spa_t or remove it from the
137 137 * namespace. No locking is required.
138 138 *
139 139 * spa_refcount_zero() Returns true if the refcount is currently
140 140 * zero. Must be called with spa_namespace_lock
141 141 * held.
142 142 *
143 143 * The spa_config_lock[] is an array of rwlocks, ordered as follows:
144 144 * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV.
145 145 * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}().
146 146 *
147 147 * To read the configuration, it suffices to hold one of these locks as reader.
148 148 * To modify the configuration, you must hold all locks as writer. To modify
149 149 * vdev state without altering the vdev tree's topology (e.g. online/offline),
150 150 * you must hold SCL_STATE and SCL_ZIO as writer.
151 151 *
152 152 * We use these distinct config locks to avoid recursive lock entry.
153 153 * For example, spa_sync() (which holds SCL_CONFIG as reader) induces
154 154 * block allocations (SCL_ALLOC), which may require reading space maps
155 155 * from disk (dmu_read() -> zio_read() -> SCL_ZIO).
156 156 *
157 157 * The spa config locks cannot be normal rwlocks because we need the
158 158 * ability to hand off ownership. For example, SCL_ZIO is acquired
159 159 * by the issuing thread and later released by an interrupt thread.
160 160 * They do, however, obey the usual write-wanted semantics to prevent
161 161 * writer (i.e. system administrator) starvation.
162 162 *
163 163 * The lock acquisition rules are as follows:
164 164 *
165 165 * SCL_CONFIG
166 166 * Protects changes to the vdev tree topology, such as vdev
167 167 * add/remove/attach/detach. Protects the dirty config list
168 168 * (spa_config_dirty_list) and the set of spares and l2arc devices.
169 169 *
170 170 * SCL_STATE
171 171 * Protects changes to pool state and vdev state, such as vdev
172 172 * online/offline/fault/degrade/clear. Protects the dirty state list
173 173 * (spa_state_dirty_list) and global pool state (spa_state).
174 174 *
175 175 * SCL_ALLOC
176 176 * Protects changes to metaslab groups and classes.
177 177 * Held as reader by metaslab_alloc() and metaslab_claim().
178 178 *
179 179 * SCL_ZIO
180 180 * Held by bp-level zios (those which have no io_vd upon entry)
181 181 * to prevent changes to the vdev tree. The bp-level zio implicitly
182 182 * protects all of its vdev child zios, which do not hold SCL_ZIO.
183 183 *
184 184 * SCL_FREE
185 185 * Protects changes to metaslab groups and classes.
186 186 * Held as reader by metaslab_free(). SCL_FREE is distinct from
187 187 * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free
188 188 * blocks in zio_done() while another i/o that holds either
189 189 * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete.
190 190 *
191 191 * SCL_VDEV
192 192 * Held as reader to prevent changes to the vdev tree during trivial
193 193 * inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the
194 194 * other locks, and lower than all of them, to ensure that it's safe
195 195 * to acquire regardless of caller context.
196 196 *
197 197 * In addition, the following rules apply:
198 198 *
199 199 * (a) spa_props_lock protects pool properties, spa_config and spa_config_list.
200 200 * The lock ordering is SCL_CONFIG > spa_props_lock.
201 201 *
202 202 * (b) I/O operations on leaf vdevs. For any zio operation that takes
203 203 * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(),
204 204 * or zio_write_phys() -- the caller must ensure that the config cannot
205 205 * cannot change in the interim, and that the vdev cannot be reopened.
206 206 * SCL_STATE as reader suffices for both.
207 207 *
208 208 * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit().
209 209 *
210 210 * spa_vdev_enter() Acquire the namespace lock and the config lock
211 211 * for writing.
212 212 *
213 213 * spa_vdev_exit() Release the config lock, wait for all I/O
214 214 * to complete, sync the updated configs to the
215 215 * cache, and release the namespace lock.
216 216 *
217 217 * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit().
218 218 * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual
219 219 * locking is, always, based on spa_namespace_lock and spa_config_lock[].
220 220 *
221 221 * spa_rename() is also implemented within this file since it requires
222 222 * manipulation of the namespace.
223 223 */
224 224
225 225 static avl_tree_t spa_namespace_avl;
226 226 kmutex_t spa_namespace_lock;
227 227 static kcondvar_t spa_namespace_cv;
228 228 static int spa_active_count;
229 229 int spa_max_replication_override = SPA_DVAS_PER_BP;
230 230
231 231 static kmutex_t spa_spare_lock;
232 232 static avl_tree_t spa_spare_avl;
233 233 static kmutex_t spa_l2cache_lock;
234 234 static avl_tree_t spa_l2cache_avl;
235 235
236 236 kmem_cache_t *spa_buffer_pool;
237 237 int spa_mode_global;
238 238
239 239 #ifdef ZFS_DEBUG
240 240 /* Everything except dprintf and spa is on by default in debug builds */
241 241 int zfs_flags = ~(ZFS_DEBUG_DPRINTF | ZFS_DEBUG_SPA);
242 242 #else
243 243 int zfs_flags = 0;
244 244 #endif
245 245
246 246 /*
247 247 * zfs_recover can be set to nonzero to attempt to recover from
248 248 * otherwise-fatal errors, typically caused by on-disk corruption. When
249 249 * set, calls to zfs_panic_recover() will turn into warning messages.
250 250 */
251 251 int zfs_recover = 0;
252 252
253 253 extern int zfs_txg_synctime_ms;
254 254
255 255 /*
256 256 * Expiration time in units of zfs_txg_synctime_ms. This value has two
257 257 * meanings. First it is used to determine when the spa_deadman logic
258 258 * should fire. By default the spa_deadman will fire if spa_sync has
259 259 * not completed in 1000 * zfs_txg_synctime_ms (i.e. 1000 seconds).
260 260 * Secondly, the value determines if an I/O is considered "hung".
261 261 * Any I/O that has not completed in zfs_deadman_synctime is considered
262 262 * "hung" resulting in a system panic.
263 263 */
264 264 uint64_t zfs_deadman_synctime = 1000ULL;
265 265
266 266 /*
267 267 * Override the zfs deadman behavior via /etc/system. By default the
268 268 * deadman is enabled except on VMware and sparc deployments.
269 269 */
270 270 int zfs_deadman_enabled = -1;
271 271
272 272
273 273 /*
274 274 * ==========================================================================
275 275 * SPA config locking
276 276 * ==========================================================================
277 277 */
278 278 static void
279 279 spa_config_lock_init(spa_t *spa)
280 280 {
281 281 for (int i = 0; i < SCL_LOCKS; i++) {
282 282 spa_config_lock_t *scl = &spa->spa_config_lock[i];
283 283 mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL);
284 284 cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL);
285 285 refcount_create_untracked(&scl->scl_count);
286 286 scl->scl_writer = NULL;
287 287 scl->scl_write_wanted = 0;
288 288 }
289 289 }
290 290
291 291 static void
292 292 spa_config_lock_destroy(spa_t *spa)
293 293 {
294 294 for (int i = 0; i < SCL_LOCKS; i++) {
295 295 spa_config_lock_t *scl = &spa->spa_config_lock[i];
296 296 mutex_destroy(&scl->scl_lock);
297 297 cv_destroy(&scl->scl_cv);
298 298 refcount_destroy(&scl->scl_count);
299 299 ASSERT(scl->scl_writer == NULL);
300 300 ASSERT(scl->scl_write_wanted == 0);
301 301 }
302 302 }
303 303
304 304 int
305 305 spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw)
306 306 {
307 307 for (int i = 0; i < SCL_LOCKS; i++) {
308 308 spa_config_lock_t *scl = &spa->spa_config_lock[i];
309 309 if (!(locks & (1 << i)))
310 310 continue;
311 311 mutex_enter(&scl->scl_lock);
312 312 if (rw == RW_READER) {
313 313 if (scl->scl_writer || scl->scl_write_wanted) {
314 314 mutex_exit(&scl->scl_lock);
315 315 spa_config_exit(spa, locks ^ (1 << i), tag);
316 316 return (0);
317 317 }
318 318 } else {
319 319 ASSERT(scl->scl_writer != curthread);
320 320 if (!refcount_is_zero(&scl->scl_count)) {
321 321 mutex_exit(&scl->scl_lock);
322 322 spa_config_exit(spa, locks ^ (1 << i), tag);
323 323 return (0);
324 324 }
325 325 scl->scl_writer = curthread;
326 326 }
327 327 (void) refcount_add(&scl->scl_count, tag);
328 328 mutex_exit(&scl->scl_lock);
329 329 }
330 330 return (1);
331 331 }
332 332
333 333 void
334 334 spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw)
335 335 {
336 336 int wlocks_held = 0;
337 337
338 338 ASSERT3U(SCL_LOCKS, <, sizeof (wlocks_held) * NBBY);
339 339
340 340 for (int i = 0; i < SCL_LOCKS; i++) {
341 341 spa_config_lock_t *scl = &spa->spa_config_lock[i];
342 342 if (scl->scl_writer == curthread)
343 343 wlocks_held |= (1 << i);
344 344 if (!(locks & (1 << i)))
345 345 continue;
346 346 mutex_enter(&scl->scl_lock);
347 347 if (rw == RW_READER) {
348 348 while (scl->scl_writer || scl->scl_write_wanted) {
349 349 cv_wait(&scl->scl_cv, &scl->scl_lock);
350 350 }
351 351 } else {
352 352 ASSERT(scl->scl_writer != curthread);
353 353 while (!refcount_is_zero(&scl->scl_count)) {
354 354 scl->scl_write_wanted++;
355 355 cv_wait(&scl->scl_cv, &scl->scl_lock);
356 356 scl->scl_write_wanted--;
357 357 }
358 358 scl->scl_writer = curthread;
359 359 }
360 360 (void) refcount_add(&scl->scl_count, tag);
361 361 mutex_exit(&scl->scl_lock);
362 362 }
363 363 ASSERT(wlocks_held <= locks);
364 364 }
365 365
366 366 void
367 367 spa_config_exit(spa_t *spa, int locks, void *tag)
368 368 {
369 369 for (int i = SCL_LOCKS - 1; i >= 0; i--) {
370 370 spa_config_lock_t *scl = &spa->spa_config_lock[i];
371 371 if (!(locks & (1 << i)))
372 372 continue;
373 373 mutex_enter(&scl->scl_lock);
374 374 ASSERT(!refcount_is_zero(&scl->scl_count));
375 375 if (refcount_remove(&scl->scl_count, tag) == 0) {
376 376 ASSERT(scl->scl_writer == NULL ||
377 377 scl->scl_writer == curthread);
378 378 scl->scl_writer = NULL; /* OK in either case */
379 379 cv_broadcast(&scl->scl_cv);
380 380 }
381 381 mutex_exit(&scl->scl_lock);
382 382 }
383 383 }
384 384
385 385 int
386 386 spa_config_held(spa_t *spa, int locks, krw_t rw)
387 387 {
388 388 int locks_held = 0;
389 389
390 390 for (int i = 0; i < SCL_LOCKS; i++) {
391 391 spa_config_lock_t *scl = &spa->spa_config_lock[i];
392 392 if (!(locks & (1 << i)))
393 393 continue;
394 394 if ((rw == RW_READER && !refcount_is_zero(&scl->scl_count)) ||
395 395 (rw == RW_WRITER && scl->scl_writer == curthread))
396 396 locks_held |= 1 << i;
397 397 }
398 398
399 399 return (locks_held);
400 400 }
401 401
402 402 /*
403 403 * ==========================================================================
404 404 * SPA namespace functions
405 405 * ==========================================================================
406 406 */
407 407
408 408 /*
409 409 * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held.
410 410 * Returns NULL if no matching spa_t is found.
411 411 */
412 412 spa_t *
413 413 spa_lookup(const char *name)
414 414 {
415 415 static spa_t search; /* spa_t is large; don't allocate on stack */
416 416 spa_t *spa;
417 417 avl_index_t where;
418 418 char *cp;
419 419
420 420 ASSERT(MUTEX_HELD(&spa_namespace_lock));
421 421
422 422 (void) strlcpy(search.spa_name, name, sizeof (search.spa_name));
423 423
424 424 /*
425 425 * If it's a full dataset name, figure out the pool name and
426 426 * just use that.
427 427 */
428 428 cp = strpbrk(search.spa_name, "/@");
429 429 if (cp != NULL)
430 430 *cp = '\0';
431 431
432 432 spa = avl_find(&spa_namespace_avl, &search, &where);
433 433
434 434 return (spa);
435 435 }
436 436
437 437 /*
438 438 * Fires when spa_sync has not completed within zfs_deadman_synctime_ms.
439 439 * If the zfs_deadman_enabled flag is set then it inspects all vdev queues
440 440 * looking for potentially hung I/Os.
441 441 */
442 442 void
443 443 spa_deadman(void *arg)
444 444 {
445 445 spa_t *spa = arg;
446 446
447 447 zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu",
448 448 (gethrtime() - spa->spa_sync_starttime) / NANOSEC,
449 449 ++spa->spa_deadman_calls);
450 450 if (zfs_deadman_enabled)
451 451 vdev_deadman(spa->spa_root_vdev);
452 452 }
453 453
454 454 /*
455 455 * Create an uninitialized spa_t with the given name. Requires
456 456 * spa_namespace_lock. The caller must ensure that the spa_t doesn't already
457 457 * exist by calling spa_lookup() first.
458 458 */
459 459 spa_t *
460 460 spa_add(const char *name, nvlist_t *config, const char *altroot)
461 461 {
462 462 spa_t *spa;
463 463 spa_config_dirent_t *dp;
464 464 cyc_handler_t hdlr;
465 465 cyc_time_t when;
466 466
467 467 ASSERT(MUTEX_HELD(&spa_namespace_lock));
468 468
469 469 spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP);
470 470
471 471 mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL);
472 472 mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL);
473 473 mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL);
474 474 mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL);
475 475 mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL);
476 476 mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL);
477 477 mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL);
478 478 mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL);
479 479 mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL);
480 480 mutex_init(&spa->spa_iokstat_lock, NULL, MUTEX_DEFAULT, NULL);
481 481
482 482 cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL);
483 483 cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL);
484 484 cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL);
485 485 cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL);
486 486
487 487 for (int t = 0; t < TXG_SIZE; t++)
488 488 bplist_create(&spa->spa_free_bplist[t]);
489 489
490 490 (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name));
491 491 spa->spa_state = POOL_STATE_UNINITIALIZED;
492 492 spa->spa_freeze_txg = UINT64_MAX;
493 493 spa->spa_final_txg = UINT64_MAX;
494 494 spa->spa_load_max_txg = UINT64_MAX;
495 495 spa->spa_proc = &p0;
496 496 spa->spa_proc_state = SPA_PROC_NONE;
497 497
498 498 hdlr.cyh_func = spa_deadman;
499 499 hdlr.cyh_arg = spa;
500 500 hdlr.cyh_level = CY_LOW_LEVEL;
501 501
502 502 spa->spa_deadman_synctime = MSEC2NSEC(zfs_deadman_synctime *
503 503 zfs_txg_synctime_ms);
504 504
505 505 /*
506 506 * This determines how often we need to check for hung I/Os after
507 507 * the cyclic has already fired. Since checking for hung I/Os is
508 508 * an expensive operation we don't want to check too frequently.
509 509 * Instead wait for 5 synctimes before checking again.
510 510 */
511 511 when.cyt_interval = MSEC2NSEC(5 * zfs_txg_synctime_ms);
512 512 when.cyt_when = CY_INFINITY;
513 513 mutex_enter(&cpu_lock);
514 514 spa->spa_deadman_cycid = cyclic_add(&hdlr, &when);
515 515 mutex_exit(&cpu_lock);
516 516
517 517 refcount_create(&spa->spa_refcount);
518 518 spa_config_lock_init(spa);
519 519
520 520 avl_add(&spa_namespace_avl, spa);
521 521
522 522 /*
523 523 * Set the alternate root, if there is one.
524 524 */
525 525 if (altroot) {
526 526 spa->spa_root = spa_strdup(altroot);
527 527 spa_active_count++;
528 528 }
529 529
530 530 /*
531 531 * Every pool starts with the default cachefile
532 532 */
533 533 list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t),
534 534 offsetof(spa_config_dirent_t, scd_link));
535 535
536 536 dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP);
537 537 dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path);
538 538 list_insert_head(&spa->spa_config_list, dp);
539 539
540 540 VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME,
541 541 KM_SLEEP) == 0);
542 542
543 543 if (config != NULL) {
544 544 nvlist_t *features;
545 545
546 546 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ,
547 547 &features) == 0) {
548 548 VERIFY(nvlist_dup(features, &spa->spa_label_features,
549 549 0) == 0);
550 550 }
551 551
552 552 VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0);
553 553 }
554 554
555 555 if (spa->spa_label_features == NULL) {
556 556 VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME,
557 557 KM_SLEEP) == 0);
558 558 }
559 559
560 560 spa->spa_iokstat = kstat_create("zfs", 0, name,
561 561 "disk", KSTAT_TYPE_IO, 1, 0);
562 562 if (spa->spa_iokstat) {
563 563 spa->spa_iokstat->ks_lock = &spa->spa_iokstat_lock;
564 564 kstat_install(spa->spa_iokstat);
565 565 }
566 566
567 567 spa->spa_debug = ((zfs_flags & ZFS_DEBUG_SPA) != 0);
568 568
569 569 return (spa);
570 570 }
571 571
572 572 /*
573 573 * Removes a spa_t from the namespace, freeing up any memory used. Requires
574 574 * spa_namespace_lock. This is called only after the spa_t has been closed and
575 575 * deactivated.
576 576 */
577 577 void
578 578 spa_remove(spa_t *spa)
579 579 {
580 580 spa_config_dirent_t *dp;
581 581
582 582 ASSERT(MUTEX_HELD(&spa_namespace_lock));
583 583 ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED);
584 584
585 585 nvlist_free(spa->spa_config_splitting);
586 586
587 587 avl_remove(&spa_namespace_avl, spa);
588 588 cv_broadcast(&spa_namespace_cv);
589 589
590 590 if (spa->spa_root) {
591 591 spa_strfree(spa->spa_root);
592 592 spa_active_count--;
593 593 }
594 594
595 595 while ((dp = list_head(&spa->spa_config_list)) != NULL) {
596 596 list_remove(&spa->spa_config_list, dp);
597 597 if (dp->scd_path != NULL)
598 598 spa_strfree(dp->scd_path);
599 599 kmem_free(dp, sizeof (spa_config_dirent_t));
600 600 }
601 601
602 602 list_destroy(&spa->spa_config_list);
603 603
604 604 nvlist_free(spa->spa_label_features);
605 605 nvlist_free(spa->spa_load_info);
606 606 spa_config_set(spa, NULL);
607 607
608 608 mutex_enter(&cpu_lock);
609 609 if (spa->spa_deadman_cycid != CYCLIC_NONE)
610 610 cyclic_remove(spa->spa_deadman_cycid);
611 611 mutex_exit(&cpu_lock);
612 612 spa->spa_deadman_cycid = CYCLIC_NONE;
613 613
614 614 refcount_destroy(&spa->spa_refcount);
615 615
616 616 spa_config_lock_destroy(spa);
617 617
618 618 kstat_delete(spa->spa_iokstat);
619 619 spa->spa_iokstat = NULL;
620 620
621 621 for (int t = 0; t < TXG_SIZE; t++)
622 622 bplist_destroy(&spa->spa_free_bplist[t]);
623 623
624 624 cv_destroy(&spa->spa_async_cv);
625 625 cv_destroy(&spa->spa_proc_cv);
626 626 cv_destroy(&spa->spa_scrub_io_cv);
627 627 cv_destroy(&spa->spa_suspend_cv);
628 628
629 629 mutex_destroy(&spa->spa_async_lock);
630 630 mutex_destroy(&spa->spa_errlist_lock);
631 631 mutex_destroy(&spa->spa_errlog_lock);
632 632 mutex_destroy(&spa->spa_history_lock);
633 633 mutex_destroy(&spa->spa_proc_lock);
634 634 mutex_destroy(&spa->spa_props_lock);
635 635 mutex_destroy(&spa->spa_scrub_lock);
636 636 mutex_destroy(&spa->spa_suspend_lock);
637 637 mutex_destroy(&spa->spa_vdev_top_lock);
638 638 mutex_destroy(&spa->spa_iokstat_lock);
639 639
640 640 kmem_free(spa, sizeof (spa_t));
641 641 }
642 642
643 643 /*
644 644 * Given a pool, return the next pool in the namespace, or NULL if there is
645 645 * none. If 'prev' is NULL, return the first pool.
646 646 */
647 647 spa_t *
648 648 spa_next(spa_t *prev)
649 649 {
650 650 ASSERT(MUTEX_HELD(&spa_namespace_lock));
651 651
652 652 if (prev)
653 653 return (AVL_NEXT(&spa_namespace_avl, prev));
654 654 else
655 655 return (avl_first(&spa_namespace_avl));
656 656 }
657 657
658 658 /*
659 659 * ==========================================================================
660 660 * SPA refcount functions
661 661 * ==========================================================================
662 662 */
663 663
664 664 /*
665 665 * Add a reference to the given spa_t. Must have at least one reference, or
666 666 * have the namespace lock held.
667 667 */
668 668 void
669 669 spa_open_ref(spa_t *spa, void *tag)
670 670 {
671 671 ASSERT(refcount_count(&spa->spa_refcount) >= spa->spa_minref ||
672 672 MUTEX_HELD(&spa_namespace_lock));
673 673 (void) refcount_add(&spa->spa_refcount, tag);
674 674 }
675 675
676 676 /*
677 677 * Remove a reference to the given spa_t. Must have at least one reference, or
678 678 * have the namespace lock held.
679 679 */
680 680 void
681 681 spa_close(spa_t *spa, void *tag)
682 682 {
683 683 ASSERT(refcount_count(&spa->spa_refcount) > spa->spa_minref ||
684 684 MUTEX_HELD(&spa_namespace_lock));
685 685 (void) refcount_remove(&spa->spa_refcount, tag);
686 686 }
687 687
688 688 /*
689 689 * Check to see if the spa refcount is zero. Must be called with
690 690 * spa_namespace_lock held. We really compare against spa_minref, which is the
691 691 * number of references acquired when opening a pool
692 692 */
693 693 boolean_t
694 694 spa_refcount_zero(spa_t *spa)
695 695 {
696 696 ASSERT(MUTEX_HELD(&spa_namespace_lock));
697 697
698 698 return (refcount_count(&spa->spa_refcount) == spa->spa_minref);
699 699 }
700 700
701 701 /*
702 702 * ==========================================================================
703 703 * SPA spare and l2cache tracking
704 704 * ==========================================================================
705 705 */
706 706
707 707 /*
708 708 * Hot spares and cache devices are tracked using the same code below,
709 709 * for 'auxiliary' devices.
710 710 */
711 711
712 712 typedef struct spa_aux {
713 713 uint64_t aux_guid;
714 714 uint64_t aux_pool;
715 715 avl_node_t aux_avl;
716 716 int aux_count;
717 717 } spa_aux_t;
718 718
719 719 static int
720 720 spa_aux_compare(const void *a, const void *b)
721 721 {
722 722 const spa_aux_t *sa = a;
723 723 const spa_aux_t *sb = b;
724 724
725 725 if (sa->aux_guid < sb->aux_guid)
726 726 return (-1);
727 727 else if (sa->aux_guid > sb->aux_guid)
728 728 return (1);
729 729 else
730 730 return (0);
731 731 }
732 732
733 733 void
734 734 spa_aux_add(vdev_t *vd, avl_tree_t *avl)
735 735 {
736 736 avl_index_t where;
737 737 spa_aux_t search;
738 738 spa_aux_t *aux;
739 739
740 740 search.aux_guid = vd->vdev_guid;
741 741 if ((aux = avl_find(avl, &search, &where)) != NULL) {
742 742 aux->aux_count++;
743 743 } else {
744 744 aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP);
745 745 aux->aux_guid = vd->vdev_guid;
746 746 aux->aux_count = 1;
747 747 avl_insert(avl, aux, where);
748 748 }
749 749 }
750 750
751 751 void
752 752 spa_aux_remove(vdev_t *vd, avl_tree_t *avl)
753 753 {
754 754 spa_aux_t search;
755 755 spa_aux_t *aux;
756 756 avl_index_t where;
757 757
758 758 search.aux_guid = vd->vdev_guid;
759 759 aux = avl_find(avl, &search, &where);
760 760
761 761 ASSERT(aux != NULL);
762 762
763 763 if (--aux->aux_count == 0) {
764 764 avl_remove(avl, aux);
765 765 kmem_free(aux, sizeof (spa_aux_t));
766 766 } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) {
767 767 aux->aux_pool = 0ULL;
768 768 }
769 769 }
770 770
771 771 boolean_t
772 772 spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl)
773 773 {
774 774 spa_aux_t search, *found;
775 775
776 776 search.aux_guid = guid;
777 777 found = avl_find(avl, &search, NULL);
778 778
779 779 if (pool) {
780 780 if (found)
781 781 *pool = found->aux_pool;
782 782 else
783 783 *pool = 0ULL;
784 784 }
785 785
786 786 if (refcnt) {
787 787 if (found)
788 788 *refcnt = found->aux_count;
789 789 else
790 790 *refcnt = 0;
791 791 }
792 792
793 793 return (found != NULL);
794 794 }
795 795
796 796 void
797 797 spa_aux_activate(vdev_t *vd, avl_tree_t *avl)
798 798 {
799 799 spa_aux_t search, *found;
800 800 avl_index_t where;
801 801
802 802 search.aux_guid = vd->vdev_guid;
803 803 found = avl_find(avl, &search, &where);
804 804 ASSERT(found != NULL);
805 805 ASSERT(found->aux_pool == 0ULL);
806 806
807 807 found->aux_pool = spa_guid(vd->vdev_spa);
808 808 }
809 809
810 810 /*
811 811 * Spares are tracked globally due to the following constraints:
812 812 *
813 813 * - A spare may be part of multiple pools.
814 814 * - A spare may be added to a pool even if it's actively in use within
815 815 * another pool.
816 816 * - A spare in use in any pool can only be the source of a replacement if
817 817 * the target is a spare in the same pool.
818 818 *
819 819 * We keep track of all spares on the system through the use of a reference
820 820 * counted AVL tree. When a vdev is added as a spare, or used as a replacement
821 821 * spare, then we bump the reference count in the AVL tree. In addition, we set
822 822 * the 'vdev_isspare' member to indicate that the device is a spare (active or
823 823 * inactive). When a spare is made active (used to replace a device in the
824 824 * pool), we also keep track of which pool its been made a part of.
825 825 *
826 826 * The 'spa_spare_lock' protects the AVL tree. These functions are normally
827 827 * called under the spa_namespace lock as part of vdev reconfiguration. The
828 828 * separate spare lock exists for the status query path, which does not need to
829 829 * be completely consistent with respect to other vdev configuration changes.
830 830 */
831 831
832 832 static int
833 833 spa_spare_compare(const void *a, const void *b)
834 834 {
835 835 return (spa_aux_compare(a, b));
836 836 }
837 837
838 838 void
839 839 spa_spare_add(vdev_t *vd)
840 840 {
841 841 mutex_enter(&spa_spare_lock);
842 842 ASSERT(!vd->vdev_isspare);
843 843 spa_aux_add(vd, &spa_spare_avl);
844 844 vd->vdev_isspare = B_TRUE;
845 845 mutex_exit(&spa_spare_lock);
846 846 }
847 847
848 848 void
849 849 spa_spare_remove(vdev_t *vd)
850 850 {
851 851 mutex_enter(&spa_spare_lock);
852 852 ASSERT(vd->vdev_isspare);
853 853 spa_aux_remove(vd, &spa_spare_avl);
854 854 vd->vdev_isspare = B_FALSE;
855 855 mutex_exit(&spa_spare_lock);
856 856 }
857 857
858 858 boolean_t
859 859 spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt)
860 860 {
861 861 boolean_t found;
862 862
863 863 mutex_enter(&spa_spare_lock);
864 864 found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl);
865 865 mutex_exit(&spa_spare_lock);
866 866
867 867 return (found);
868 868 }
869 869
870 870 void
871 871 spa_spare_activate(vdev_t *vd)
872 872 {
873 873 mutex_enter(&spa_spare_lock);
874 874 ASSERT(vd->vdev_isspare);
875 875 spa_aux_activate(vd, &spa_spare_avl);
876 876 mutex_exit(&spa_spare_lock);
877 877 }
878 878
879 879 /*
880 880 * Level 2 ARC devices are tracked globally for the same reasons as spares.
881 881 * Cache devices currently only support one pool per cache device, and so
882 882 * for these devices the aux reference count is currently unused beyond 1.
883 883 */
884 884
885 885 static int
886 886 spa_l2cache_compare(const void *a, const void *b)
887 887 {
888 888 return (spa_aux_compare(a, b));
889 889 }
890 890
891 891 void
892 892 spa_l2cache_add(vdev_t *vd)
893 893 {
894 894 mutex_enter(&spa_l2cache_lock);
895 895 ASSERT(!vd->vdev_isl2cache);
896 896 spa_aux_add(vd, &spa_l2cache_avl);
897 897 vd->vdev_isl2cache = B_TRUE;
898 898 mutex_exit(&spa_l2cache_lock);
899 899 }
900 900
901 901 void
902 902 spa_l2cache_remove(vdev_t *vd)
903 903 {
904 904 mutex_enter(&spa_l2cache_lock);
905 905 ASSERT(vd->vdev_isl2cache);
906 906 spa_aux_remove(vd, &spa_l2cache_avl);
907 907 vd->vdev_isl2cache = B_FALSE;
908 908 mutex_exit(&spa_l2cache_lock);
909 909 }
910 910
911 911 boolean_t
912 912 spa_l2cache_exists(uint64_t guid, uint64_t *pool)
913 913 {
914 914 boolean_t found;
915 915
916 916 mutex_enter(&spa_l2cache_lock);
917 917 found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl);
918 918 mutex_exit(&spa_l2cache_lock);
919 919
920 920 return (found);
921 921 }
922 922
923 923 void
924 924 spa_l2cache_activate(vdev_t *vd)
925 925 {
926 926 mutex_enter(&spa_l2cache_lock);
927 927 ASSERT(vd->vdev_isl2cache);
928 928 spa_aux_activate(vd, &spa_l2cache_avl);
929 929 mutex_exit(&spa_l2cache_lock);
930 930 }
931 931
932 932 /*
933 933 * ==========================================================================
934 934 * SPA vdev locking
935 935 * ==========================================================================
936 936 */
937 937
938 938 /*
939 939 * Lock the given spa_t for the purpose of adding or removing a vdev.
940 940 * Grabs the global spa_namespace_lock plus the spa config lock for writing.
941 941 * It returns the next transaction group for the spa_t.
942 942 */
943 943 uint64_t
944 944 spa_vdev_enter(spa_t *spa)
945 945 {
946 946 mutex_enter(&spa->spa_vdev_top_lock);
947 947 mutex_enter(&spa_namespace_lock);
948 948 return (spa_vdev_config_enter(spa));
949 949 }
950 950
951 951 /*
952 952 * Internal implementation for spa_vdev_enter(). Used when a vdev
953 953 * operation requires multiple syncs (i.e. removing a device) while
954 954 * keeping the spa_namespace_lock held.
955 955 */
956 956 uint64_t
957 957 spa_vdev_config_enter(spa_t *spa)
958 958 {
959 959 ASSERT(MUTEX_HELD(&spa_namespace_lock));
960 960
961 961 spa_config_enter(spa, SCL_ALL, spa, RW_WRITER);
962 962
963 963 return (spa_last_synced_txg(spa) + 1);
964 964 }
965 965
966 966 /*
967 967 * Used in combination with spa_vdev_config_enter() to allow the syncing
968 968 * of multiple transactions without releasing the spa_namespace_lock.
969 969 */
970 970 void
971 971 spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag)
972 972 {
973 973 ASSERT(MUTEX_HELD(&spa_namespace_lock));
974 974
975 975 int config_changed = B_FALSE;
976 976
977 977 ASSERT(txg > spa_last_synced_txg(spa));
978 978
979 979 spa->spa_pending_vdev = NULL;
980 980
981 981 /*
982 982 * Reassess the DTLs.
983 983 */
984 984 vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE);
985 985
986 986 if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) {
987 987 config_changed = B_TRUE;
988 988 spa->spa_config_generation++;
989 989 }
990 990
991 991 /*
992 992 * Verify the metaslab classes.
993 993 */
994 994 ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0);
995 995 ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0);
996 996
997 997 spa_config_exit(spa, SCL_ALL, spa);
998 998
999 999 /*
1000 1000 * Panic the system if the specified tag requires it. This
1001 1001 * is useful for ensuring that configurations are updated
1002 1002 * transactionally.
1003 1003 */
1004 1004 if (zio_injection_enabled)
1005 1005 zio_handle_panic_injection(spa, tag, 0);
1006 1006
1007 1007 /*
1008 1008 * Note: this txg_wait_synced() is important because it ensures
1009 1009 * that there won't be more than one config change per txg.
1010 1010 * This allows us to use the txg as the generation number.
1011 1011 */
1012 1012 if (error == 0)
1013 1013 txg_wait_synced(spa->spa_dsl_pool, txg);
1014 1014
1015 1015 if (vd != NULL) {
1016 1016 ASSERT(!vd->vdev_detached || vd->vdev_dtl_smo.smo_object == 0);
1017 1017 spa_config_enter(spa, SCL_ALL, spa, RW_WRITER);
1018 1018 vdev_free(vd);
1019 1019 spa_config_exit(spa, SCL_ALL, spa);
1020 1020 }
1021 1021
1022 1022 /*
1023 1023 * If the config changed, update the config cache.
1024 1024 */
1025 1025 if (config_changed)
1026 1026 spa_config_sync(spa, B_FALSE, B_TRUE);
1027 1027 }
1028 1028
1029 1029 /*
1030 1030 * Unlock the spa_t after adding or removing a vdev. Besides undoing the
1031 1031 * locking of spa_vdev_enter(), we also want make sure the transactions have
1032 1032 * synced to disk, and then update the global configuration cache with the new
1033 1033 * information.
1034 1034 */
1035 1035 int
1036 1036 spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error)
1037 1037 {
1038 1038 spa_vdev_config_exit(spa, vd, txg, error, FTAG);
1039 1039 mutex_exit(&spa_namespace_lock);
1040 1040 mutex_exit(&spa->spa_vdev_top_lock);
1041 1041
1042 1042 return (error);
1043 1043 }
1044 1044
1045 1045 /*
1046 1046 * Lock the given spa_t for the purpose of changing vdev state.
1047 1047 */
1048 1048 void
1049 1049 spa_vdev_state_enter(spa_t *spa, int oplocks)
1050 1050 {
1051 1051 int locks = SCL_STATE_ALL | oplocks;
1052 1052
1053 1053 /*
1054 1054 * Root pools may need to read of the underlying devfs filesystem
1055 1055 * when opening up a vdev. Unfortunately if we're holding the
1056 1056 * SCL_ZIO lock it will result in a deadlock when we try to issue
1057 1057 * the read from the root filesystem. Instead we "prefetch"
1058 1058 * the associated vnodes that we need prior to opening the
1059 1059 * underlying devices and cache them so that we can prevent
1060 1060 * any I/O when we are doing the actual open.
1061 1061 */
1062 1062 if (spa_is_root(spa)) {
1063 1063 int low = locks & ~(SCL_ZIO - 1);
1064 1064 int high = locks & ~low;
1065 1065
1066 1066 spa_config_enter(spa, high, spa, RW_WRITER);
1067 1067 vdev_hold(spa->spa_root_vdev);
1068 1068 spa_config_enter(spa, low, spa, RW_WRITER);
1069 1069 } else {
1070 1070 spa_config_enter(spa, locks, spa, RW_WRITER);
1071 1071 }
1072 1072 spa->spa_vdev_locks = locks;
1073 1073 }
1074 1074
1075 1075 int
1076 1076 spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error)
1077 1077 {
1078 1078 boolean_t config_changed = B_FALSE;
1079 1079
1080 1080 if (vd != NULL || error == 0)
1081 1081 vdev_dtl_reassess(vd ? vd->vdev_top : spa->spa_root_vdev,
1082 1082 0, 0, B_FALSE);
1083 1083
1084 1084 if (vd != NULL) {
1085 1085 vdev_state_dirty(vd->vdev_top);
1086 1086 config_changed = B_TRUE;
1087 1087 spa->spa_config_generation++;
1088 1088 }
1089 1089
1090 1090 if (spa_is_root(spa))
1091 1091 vdev_rele(spa->spa_root_vdev);
1092 1092
1093 1093 ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL);
1094 1094 spa_config_exit(spa, spa->spa_vdev_locks, spa);
1095 1095
1096 1096 /*
1097 1097 * If anything changed, wait for it to sync. This ensures that,
1098 1098 * from the system administrator's perspective, zpool(1M) commands
1099 1099 * are synchronous. This is important for things like zpool offline:
1100 1100 * when the command completes, you expect no further I/O from ZFS.
1101 1101 */
1102 1102 if (vd != NULL)
1103 1103 txg_wait_synced(spa->spa_dsl_pool, 0);
1104 1104
1105 1105 /*
1106 1106 * If the config changed, update the config cache.
1107 1107 */
1108 1108 if (config_changed) {
1109 1109 mutex_enter(&spa_namespace_lock);
1110 1110 spa_config_sync(spa, B_FALSE, B_TRUE);
1111 1111 mutex_exit(&spa_namespace_lock);
1112 1112 }
1113 1113
1114 1114 return (error);
1115 1115 }
1116 1116
1117 1117 /*
1118 1118 * ==========================================================================
1119 1119 * Miscellaneous functions
1120 1120 * ==========================================================================
1121 1121 */
1122 1122
1123 1123 void
1124 1124 spa_activate_mos_feature(spa_t *spa, const char *feature)
1125 1125 {
1126 1126 (void) nvlist_add_boolean(spa->spa_label_features, feature);
1127 1127 vdev_config_dirty(spa->spa_root_vdev);
1128 1128 }
1129 1129
1130 1130 void
1131 1131 spa_deactivate_mos_feature(spa_t *spa, const char *feature)
1132 1132 {
1133 1133 (void) nvlist_remove_all(spa->spa_label_features, feature);
1134 1134 vdev_config_dirty(spa->spa_root_vdev);
1135 1135 }
1136 1136
1137 1137 /*
1138 1138 * Rename a spa_t.
1139 1139 */
1140 1140 int
1141 1141 spa_rename(const char *name, const char *newname)
1142 1142 {
1143 1143 spa_t *spa;
1144 1144 int err;
1145 1145
1146 1146 /*
1147 1147 * Lookup the spa_t and grab the config lock for writing. We need to
1148 1148 * actually open the pool so that we can sync out the necessary labels.
1149 1149 * It's OK to call spa_open() with the namespace lock held because we
1150 1150 * allow recursive calls for other reasons.
1151 1151 */
1152 1152 mutex_enter(&spa_namespace_lock);
1153 1153 if ((err = spa_open(name, &spa, FTAG)) != 0) {
1154 1154 mutex_exit(&spa_namespace_lock);
1155 1155 return (err);
1156 1156 }
1157 1157
1158 1158 spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
1159 1159
1160 1160 avl_remove(&spa_namespace_avl, spa);
1161 1161 (void) strlcpy(spa->spa_name, newname, sizeof (spa->spa_name));
1162 1162 avl_add(&spa_namespace_avl, spa);
1163 1163
1164 1164 /*
1165 1165 * Sync all labels to disk with the new names by marking the root vdev
1166 1166 * dirty and waiting for it to sync. It will pick up the new pool name
1167 1167 * during the sync.
1168 1168 */
1169 1169 vdev_config_dirty(spa->spa_root_vdev);
1170 1170
1171 1171 spa_config_exit(spa, SCL_ALL, FTAG);
1172 1172
1173 1173 txg_wait_synced(spa->spa_dsl_pool, 0);
1174 1174
1175 1175 /*
1176 1176 * Sync the updated config cache.
1177 1177 */
1178 1178 spa_config_sync(spa, B_FALSE, B_TRUE);
1179 1179
1180 1180 spa_close(spa, FTAG);
1181 1181
1182 1182 mutex_exit(&spa_namespace_lock);
1183 1183
1184 1184 return (0);
1185 1185 }
1186 1186
1187 1187 /*
1188 1188 * Return the spa_t associated with given pool_guid, if it exists. If
1189 1189 * device_guid is non-zero, determine whether the pool exists *and* contains
1190 1190 * a device with the specified device_guid.
1191 1191 */
1192 1192 spa_t *
1193 1193 spa_by_guid(uint64_t pool_guid, uint64_t device_guid)
1194 1194 {
1195 1195 spa_t *spa;
1196 1196 avl_tree_t *t = &spa_namespace_avl;
1197 1197
1198 1198 ASSERT(MUTEX_HELD(&spa_namespace_lock));
1199 1199
1200 1200 for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) {
1201 1201 if (spa->spa_state == POOL_STATE_UNINITIALIZED)
1202 1202 continue;
1203 1203 if (spa->spa_root_vdev == NULL)
1204 1204 continue;
1205 1205 if (spa_guid(spa) == pool_guid) {
1206 1206 if (device_guid == 0)
1207 1207 break;
1208 1208
1209 1209 if (vdev_lookup_by_guid(spa->spa_root_vdev,
1210 1210 device_guid) != NULL)
1211 1211 break;
1212 1212
1213 1213 /*
1214 1214 * Check any devices we may be in the process of adding.
1215 1215 */
1216 1216 if (spa->spa_pending_vdev) {
1217 1217 if (vdev_lookup_by_guid(spa->spa_pending_vdev,
1218 1218 device_guid) != NULL)
1219 1219 break;
1220 1220 }
1221 1221 }
1222 1222 }
1223 1223
1224 1224 return (spa);
1225 1225 }
1226 1226
1227 1227 /*
1228 1228 * Determine whether a pool with the given pool_guid exists.
1229 1229 */
1230 1230 boolean_t
1231 1231 spa_guid_exists(uint64_t pool_guid, uint64_t device_guid)
1232 1232 {
1233 1233 return (spa_by_guid(pool_guid, device_guid) != NULL);
1234 1234 }
1235 1235
1236 1236 char *
1237 1237 spa_strdup(const char *s)
1238 1238 {
1239 1239 size_t len;
1240 1240 char *new;
1241 1241
1242 1242 len = strlen(s);
1243 1243 new = kmem_alloc(len + 1, KM_SLEEP);
1244 1244 bcopy(s, new, len);
1245 1245 new[len] = '\0';
1246 1246
1247 1247 return (new);
1248 1248 }
1249 1249
1250 1250 void
1251 1251 spa_strfree(char *s)
1252 1252 {
1253 1253 kmem_free(s, strlen(s) + 1);
1254 1254 }
1255 1255
1256 1256 uint64_t
1257 1257 spa_get_random(uint64_t range)
1258 1258 {
1259 1259 uint64_t r;
1260 1260
1261 1261 ASSERT(range != 0);
1262 1262
1263 1263 (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t));
1264 1264
1265 1265 return (r % range);
1266 1266 }
1267 1267
1268 1268 uint64_t
1269 1269 spa_generate_guid(spa_t *spa)
1270 1270 {
1271 1271 uint64_t guid = spa_get_random(-1ULL);
1272 1272
1273 1273 if (spa != NULL) {
1274 1274 while (guid == 0 || spa_guid_exists(spa_guid(spa), guid))
1275 1275 guid = spa_get_random(-1ULL);
1276 1276 } else {
1277 1277 while (guid == 0 || spa_guid_exists(guid, 0))
1278 1278 guid = spa_get_random(-1ULL);
1279 1279 }
1280 1280
1281 1281 return (guid);
1282 1282 }
1283 1283
1284 1284 void
1285 1285 sprintf_blkptr(char *buf, const blkptr_t *bp)
1286 1286 {
1287 1287 char type[256];
1288 1288 char *checksum = NULL;
1289 1289 char *compress = NULL;
1290 1290
1291 1291 if (bp != NULL) {
1292 1292 if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) {
1293 1293 dmu_object_byteswap_t bswap =
1294 1294 DMU_OT_BYTESWAP(BP_GET_TYPE(bp));
1295 1295 (void) snprintf(type, sizeof (type), "bswap %s %s",
1296 1296 DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ?
1297 1297 "metadata" : "data",
1298 1298 dmu_ot_byteswap[bswap].ob_name);
1299 1299 } else {
1300 1300 (void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name,
1301 1301 sizeof (type));
1302 1302 }
1303 1303 checksum = zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name;
1304 1304 compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name;
1305 1305 }
1306 1306
1307 1307 SPRINTF_BLKPTR(snprintf, ' ', buf, bp, type, checksum, compress);
1308 1308 }
1309 1309
1310 1310 void
1311 1311 spa_freeze(spa_t *spa)
1312 1312 {
1313 1313 uint64_t freeze_txg = 0;
1314 1314
1315 1315 spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
1316 1316 if (spa->spa_freeze_txg == UINT64_MAX) {
1317 1317 freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE;
1318 1318 spa->spa_freeze_txg = freeze_txg;
1319 1319 }
1320 1320 spa_config_exit(spa, SCL_ALL, FTAG);
1321 1321 if (freeze_txg != 0)
1322 1322 txg_wait_synced(spa_get_dsl(spa), freeze_txg);
1323 1323 }
1324 1324
1325 1325 void
1326 1326 zfs_panic_recover(const char *fmt, ...)
|
↓ open down ↓ |
1326 lines elided |
↑ open up ↑ |
1327 1327 {
1328 1328 va_list adx;
1329 1329
1330 1330 va_start(adx, fmt);
1331 1331 vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx);
1332 1332 va_end(adx);
1333 1333 }
1334 1334
1335 1335 /*
1336 1336 * This is a stripped-down version of strtoull, suitable only for converting
1337 - * lowercase hexidecimal numbers that don't overflow.
1337 + * lowercase hexadecimal numbers that don't overflow.
1338 1338 */
1339 1339 uint64_t
1340 1340 strtonum(const char *str, char **nptr)
1341 1341 {
1342 1342 uint64_t val = 0;
1343 1343 char c;
1344 1344 int digit;
1345 1345
1346 1346 while ((c = *str) != '\0') {
1347 1347 if (c >= '0' && c <= '9')
1348 1348 digit = c - '0';
1349 1349 else if (c >= 'a' && c <= 'f')
1350 1350 digit = 10 + c - 'a';
1351 1351 else
1352 1352 break;
1353 1353
1354 1354 val *= 16;
1355 1355 val += digit;
1356 1356
1357 1357 str++;
1358 1358 }
1359 1359
1360 1360 if (nptr)
1361 1361 *nptr = (char *)str;
1362 1362
1363 1363 return (val);
1364 1364 }
1365 1365
1366 1366 /*
1367 1367 * ==========================================================================
1368 1368 * Accessor functions
1369 1369 * ==========================================================================
1370 1370 */
1371 1371
1372 1372 boolean_t
1373 1373 spa_shutting_down(spa_t *spa)
1374 1374 {
1375 1375 return (spa->spa_async_suspended);
1376 1376 }
1377 1377
1378 1378 dsl_pool_t *
1379 1379 spa_get_dsl(spa_t *spa)
1380 1380 {
1381 1381 return (spa->spa_dsl_pool);
1382 1382 }
1383 1383
1384 1384 boolean_t
1385 1385 spa_is_initializing(spa_t *spa)
1386 1386 {
1387 1387 return (spa->spa_is_initializing);
1388 1388 }
1389 1389
1390 1390 blkptr_t *
1391 1391 spa_get_rootblkptr(spa_t *spa)
1392 1392 {
1393 1393 return (&spa->spa_ubsync.ub_rootbp);
1394 1394 }
1395 1395
1396 1396 void
1397 1397 spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp)
1398 1398 {
1399 1399 spa->spa_uberblock.ub_rootbp = *bp;
1400 1400 }
1401 1401
1402 1402 void
1403 1403 spa_altroot(spa_t *spa, char *buf, size_t buflen)
1404 1404 {
1405 1405 if (spa->spa_root == NULL)
1406 1406 buf[0] = '\0';
1407 1407 else
1408 1408 (void) strncpy(buf, spa->spa_root, buflen);
1409 1409 }
1410 1410
1411 1411 int
1412 1412 spa_sync_pass(spa_t *spa)
1413 1413 {
1414 1414 return (spa->spa_sync_pass);
1415 1415 }
1416 1416
1417 1417 char *
1418 1418 spa_name(spa_t *spa)
1419 1419 {
1420 1420 return (spa->spa_name);
1421 1421 }
1422 1422
1423 1423 uint64_t
1424 1424 spa_guid(spa_t *spa)
1425 1425 {
1426 1426 dsl_pool_t *dp = spa_get_dsl(spa);
1427 1427 uint64_t guid;
1428 1428
1429 1429 /*
1430 1430 * If we fail to parse the config during spa_load(), we can go through
1431 1431 * the error path (which posts an ereport) and end up here with no root
1432 1432 * vdev. We stash the original pool guid in 'spa_config_guid' to handle
1433 1433 * this case.
1434 1434 */
1435 1435 if (spa->spa_root_vdev == NULL)
1436 1436 return (spa->spa_config_guid);
1437 1437
1438 1438 guid = spa->spa_last_synced_guid != 0 ?
1439 1439 spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid;
1440 1440
1441 1441 /*
1442 1442 * Return the most recently synced out guid unless we're
1443 1443 * in syncing context.
1444 1444 */
1445 1445 if (dp && dsl_pool_sync_context(dp))
1446 1446 return (spa->spa_root_vdev->vdev_guid);
1447 1447 else
1448 1448 return (guid);
1449 1449 }
1450 1450
1451 1451 uint64_t
1452 1452 spa_load_guid(spa_t *spa)
1453 1453 {
1454 1454 /*
1455 1455 * This is a GUID that exists solely as a reference for the
1456 1456 * purposes of the arc. It is generated at load time, and
1457 1457 * is never written to persistent storage.
1458 1458 */
1459 1459 return (spa->spa_load_guid);
1460 1460 }
1461 1461
1462 1462 uint64_t
1463 1463 spa_last_synced_txg(spa_t *spa)
1464 1464 {
1465 1465 return (spa->spa_ubsync.ub_txg);
1466 1466 }
1467 1467
1468 1468 uint64_t
1469 1469 spa_first_txg(spa_t *spa)
1470 1470 {
1471 1471 return (spa->spa_first_txg);
1472 1472 }
1473 1473
1474 1474 uint64_t
1475 1475 spa_syncing_txg(spa_t *spa)
1476 1476 {
1477 1477 return (spa->spa_syncing_txg);
1478 1478 }
1479 1479
1480 1480 pool_state_t
1481 1481 spa_state(spa_t *spa)
1482 1482 {
1483 1483 return (spa->spa_state);
1484 1484 }
1485 1485
1486 1486 spa_load_state_t
1487 1487 spa_load_state(spa_t *spa)
1488 1488 {
1489 1489 return (spa->spa_load_state);
1490 1490 }
1491 1491
1492 1492 uint64_t
1493 1493 spa_freeze_txg(spa_t *spa)
1494 1494 {
1495 1495 return (spa->spa_freeze_txg);
1496 1496 }
1497 1497
1498 1498 /* ARGSUSED */
1499 1499 uint64_t
1500 1500 spa_get_asize(spa_t *spa, uint64_t lsize)
1501 1501 {
1502 1502 /*
1503 1503 * The worst case is single-sector max-parity RAID-Z blocks, in which
1504 1504 * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1)
1505 1505 * times the size; so just assume that. Add to this the fact that
1506 1506 * we can have up to 3 DVAs per bp, and one more factor of 2 because
1507 1507 * the block may be dittoed with up to 3 DVAs by ddt_sync().
1508 1508 */
1509 1509 return (lsize * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2);
1510 1510 }
1511 1511
1512 1512 uint64_t
1513 1513 spa_get_dspace(spa_t *spa)
1514 1514 {
1515 1515 return (spa->spa_dspace);
1516 1516 }
1517 1517
1518 1518 void
1519 1519 spa_update_dspace(spa_t *spa)
1520 1520 {
1521 1521 spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) +
1522 1522 ddt_get_dedup_dspace(spa);
1523 1523 }
1524 1524
1525 1525 /*
1526 1526 * Return the failure mode that has been set to this pool. The default
1527 1527 * behavior will be to block all I/Os when a complete failure occurs.
1528 1528 */
1529 1529 uint8_t
1530 1530 spa_get_failmode(spa_t *spa)
1531 1531 {
1532 1532 return (spa->spa_failmode);
1533 1533 }
1534 1534
1535 1535 boolean_t
1536 1536 spa_suspended(spa_t *spa)
1537 1537 {
1538 1538 return (spa->spa_suspended);
1539 1539 }
1540 1540
1541 1541 uint64_t
1542 1542 spa_version(spa_t *spa)
1543 1543 {
1544 1544 return (spa->spa_ubsync.ub_version);
1545 1545 }
1546 1546
1547 1547 boolean_t
1548 1548 spa_deflate(spa_t *spa)
1549 1549 {
1550 1550 return (spa->spa_deflate);
1551 1551 }
1552 1552
1553 1553 metaslab_class_t *
1554 1554 spa_normal_class(spa_t *spa)
1555 1555 {
1556 1556 return (spa->spa_normal_class);
1557 1557 }
1558 1558
1559 1559 metaslab_class_t *
1560 1560 spa_log_class(spa_t *spa)
1561 1561 {
1562 1562 return (spa->spa_log_class);
1563 1563 }
1564 1564
1565 1565 int
1566 1566 spa_max_replication(spa_t *spa)
1567 1567 {
1568 1568 /*
1569 1569 * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to
1570 1570 * handle BPs with more than one DVA allocated. Set our max
1571 1571 * replication level accordingly.
1572 1572 */
1573 1573 if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS)
1574 1574 return (1);
1575 1575 return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override));
1576 1576 }
1577 1577
1578 1578 int
1579 1579 spa_prev_software_version(spa_t *spa)
1580 1580 {
1581 1581 return (spa->spa_prev_software_version);
1582 1582 }
1583 1583
1584 1584 uint64_t
1585 1585 spa_deadman_synctime(spa_t *spa)
1586 1586 {
1587 1587 return (spa->spa_deadman_synctime);
1588 1588 }
1589 1589
1590 1590 uint64_t
1591 1591 dva_get_dsize_sync(spa_t *spa, const dva_t *dva)
1592 1592 {
1593 1593 uint64_t asize = DVA_GET_ASIZE(dva);
1594 1594 uint64_t dsize = asize;
1595 1595
1596 1596 ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);
1597 1597
1598 1598 if (asize != 0 && spa->spa_deflate) {
1599 1599 vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva));
1600 1600 dsize = (asize >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio;
1601 1601 }
1602 1602
1603 1603 return (dsize);
1604 1604 }
1605 1605
1606 1606 uint64_t
1607 1607 bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp)
1608 1608 {
1609 1609 uint64_t dsize = 0;
1610 1610
1611 1611 for (int d = 0; d < SPA_DVAS_PER_BP; d++)
1612 1612 dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]);
1613 1613
1614 1614 return (dsize);
1615 1615 }
1616 1616
1617 1617 uint64_t
1618 1618 bp_get_dsize(spa_t *spa, const blkptr_t *bp)
1619 1619 {
1620 1620 uint64_t dsize = 0;
1621 1621
1622 1622 spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
1623 1623
1624 1624 for (int d = 0; d < SPA_DVAS_PER_BP; d++)
1625 1625 dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]);
1626 1626
1627 1627 spa_config_exit(spa, SCL_VDEV, FTAG);
1628 1628
1629 1629 return (dsize);
1630 1630 }
1631 1631
1632 1632 /*
1633 1633 * ==========================================================================
1634 1634 * Initialization and Termination
1635 1635 * ==========================================================================
1636 1636 */
1637 1637
1638 1638 static int
1639 1639 spa_name_compare(const void *a1, const void *a2)
1640 1640 {
1641 1641 const spa_t *s1 = a1;
1642 1642 const spa_t *s2 = a2;
1643 1643 int s;
1644 1644
1645 1645 s = strcmp(s1->spa_name, s2->spa_name);
1646 1646 if (s > 0)
1647 1647 return (1);
1648 1648 if (s < 0)
1649 1649 return (-1);
1650 1650 return (0);
1651 1651 }
1652 1652
1653 1653 int
1654 1654 spa_busy(void)
1655 1655 {
1656 1656 return (spa_active_count);
1657 1657 }
1658 1658
1659 1659 void
1660 1660 spa_boot_init()
1661 1661 {
1662 1662 spa_config_load();
1663 1663 }
1664 1664
1665 1665 void
1666 1666 spa_init(int mode)
1667 1667 {
1668 1668 mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL);
1669 1669 mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL);
1670 1670 mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL);
1671 1671 cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL);
1672 1672
1673 1673 avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t),
1674 1674 offsetof(spa_t, spa_avl));
1675 1675
1676 1676 avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t),
1677 1677 offsetof(spa_aux_t, aux_avl));
1678 1678
1679 1679 avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t),
1680 1680 offsetof(spa_aux_t, aux_avl));
1681 1681
1682 1682 spa_mode_global = mode;
1683 1683
1684 1684 #ifdef _KERNEL
1685 1685 spa_arch_init();
1686 1686 #else
1687 1687 if (spa_mode_global != FREAD && dprintf_find_string("watch")) {
1688 1688 arc_procfd = open("/proc/self/ctl", O_WRONLY);
1689 1689 if (arc_procfd == -1) {
1690 1690 perror("could not enable watchpoints: "
1691 1691 "opening /proc/self/ctl failed: ");
1692 1692 } else {
1693 1693 arc_watch = B_TRUE;
1694 1694 }
1695 1695 }
1696 1696 #endif
1697 1697
1698 1698 refcount_init();
1699 1699 unique_init();
1700 1700 space_map_init();
1701 1701 zio_init();
1702 1702 dmu_init();
1703 1703 zil_init();
1704 1704 vdev_cache_stat_init();
1705 1705 zfs_prop_init();
1706 1706 zpool_prop_init();
1707 1707 zpool_feature_init();
1708 1708 spa_config_load();
1709 1709 l2arc_start();
1710 1710 }
1711 1711
1712 1712 void
1713 1713 spa_fini(void)
1714 1714 {
1715 1715 l2arc_stop();
1716 1716
1717 1717 spa_evict_all();
1718 1718
1719 1719 vdev_cache_stat_fini();
1720 1720 zil_fini();
1721 1721 dmu_fini();
1722 1722 zio_fini();
1723 1723 space_map_fini();
1724 1724 unique_fini();
1725 1725 refcount_fini();
1726 1726
1727 1727 avl_destroy(&spa_namespace_avl);
1728 1728 avl_destroy(&spa_spare_avl);
1729 1729 avl_destroy(&spa_l2cache_avl);
1730 1730
1731 1731 cv_destroy(&spa_namespace_cv);
1732 1732 mutex_destroy(&spa_namespace_lock);
1733 1733 mutex_destroy(&spa_spare_lock);
1734 1734 mutex_destroy(&spa_l2cache_lock);
1735 1735 }
1736 1736
1737 1737 /*
1738 1738 * Return whether this pool has slogs. No locking needed.
1739 1739 * It's not a problem if the wrong answer is returned as it's only for
1740 1740 * performance and not correctness
1741 1741 */
1742 1742 boolean_t
1743 1743 spa_has_slogs(spa_t *spa)
1744 1744 {
1745 1745 return (spa->spa_log_class->mc_rotor != NULL);
1746 1746 }
1747 1747
1748 1748 spa_log_state_t
1749 1749 spa_get_log_state(spa_t *spa)
1750 1750 {
1751 1751 return (spa->spa_log_state);
1752 1752 }
1753 1753
1754 1754 void
1755 1755 spa_set_log_state(spa_t *spa, spa_log_state_t state)
1756 1756 {
1757 1757 spa->spa_log_state = state;
1758 1758 }
1759 1759
1760 1760 boolean_t
1761 1761 spa_is_root(spa_t *spa)
1762 1762 {
1763 1763 return (spa->spa_is_root);
1764 1764 }
1765 1765
1766 1766 boolean_t
1767 1767 spa_writeable(spa_t *spa)
1768 1768 {
1769 1769 return (!!(spa->spa_mode & FWRITE));
1770 1770 }
1771 1771
1772 1772 int
1773 1773 spa_mode(spa_t *spa)
1774 1774 {
1775 1775 return (spa->spa_mode);
1776 1776 }
1777 1777
1778 1778 uint64_t
1779 1779 spa_bootfs(spa_t *spa)
1780 1780 {
1781 1781 return (spa->spa_bootfs);
1782 1782 }
1783 1783
1784 1784 uint64_t
1785 1785 spa_delegation(spa_t *spa)
1786 1786 {
1787 1787 return (spa->spa_delegation);
1788 1788 }
1789 1789
1790 1790 objset_t *
1791 1791 spa_meta_objset(spa_t *spa)
1792 1792 {
1793 1793 return (spa->spa_meta_objset);
1794 1794 }
1795 1795
1796 1796 enum zio_checksum
1797 1797 spa_dedup_checksum(spa_t *spa)
1798 1798 {
1799 1799 return (spa->spa_dedup_checksum);
1800 1800 }
1801 1801
1802 1802 /*
1803 1803 * Reset pool scan stat per scan pass (or reboot).
1804 1804 */
1805 1805 void
1806 1806 spa_scan_stat_init(spa_t *spa)
1807 1807 {
1808 1808 /* data not stored on disk */
1809 1809 spa->spa_scan_pass_start = gethrestime_sec();
1810 1810 spa->spa_scan_pass_exam = 0;
1811 1811 vdev_scan_stat_init(spa->spa_root_vdev);
1812 1812 }
1813 1813
1814 1814 /*
1815 1815 * Get scan stats for zpool status reports
1816 1816 */
1817 1817 int
1818 1818 spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps)
1819 1819 {
1820 1820 dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL;
1821 1821
1822 1822 if (scn == NULL || scn->scn_phys.scn_func == POOL_SCAN_NONE)
1823 1823 return (SET_ERROR(ENOENT));
1824 1824 bzero(ps, sizeof (pool_scan_stat_t));
1825 1825
1826 1826 /* data stored on disk */
1827 1827 ps->pss_func = scn->scn_phys.scn_func;
1828 1828 ps->pss_start_time = scn->scn_phys.scn_start_time;
1829 1829 ps->pss_end_time = scn->scn_phys.scn_end_time;
1830 1830 ps->pss_to_examine = scn->scn_phys.scn_to_examine;
1831 1831 ps->pss_examined = scn->scn_phys.scn_examined;
1832 1832 ps->pss_to_process = scn->scn_phys.scn_to_process;
1833 1833 ps->pss_processed = scn->scn_phys.scn_processed;
1834 1834 ps->pss_errors = scn->scn_phys.scn_errors;
1835 1835 ps->pss_state = scn->scn_phys.scn_state;
1836 1836
1837 1837 /* data not stored on disk */
1838 1838 ps->pss_pass_start = spa->spa_scan_pass_start;
1839 1839 ps->pss_pass_exam = spa->spa_scan_pass_exam;
1840 1840
1841 1841 return (0);
1842 1842 }
1843 1843
1844 1844 boolean_t
1845 1845 spa_debug_enabled(spa_t *spa)
1846 1846 {
1847 1847 return (spa->spa_debug);
1848 1848 }
|
↓ open down ↓ |
501 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX