]>
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 | |
1d3ba0bf | 9 | * or https://opensource.org/licenses/CDDL-1.0. |
34dc7c2f BB |
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. |
93e28d66 | 23 | * Copyright (c) 2011, 2019 by Delphix. All rights reserved. |
adfe9d93 | 24 | * Copyright 2015 Nexenta Systems, Inc. All rights reserved. |
0c66c32d | 25 | * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. |
3c67d83a | 26 | * Copyright 2013 Saso Kiselkov. All rights reserved. |
0ea05c64 | 27 | * Copyright (c) 2017 Datto Inc. |
cc99f275 | 28 | * Copyright (c) 2017, Intel Corporation. |
e3570464 | 29 | * Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>. All rights reserved. |
4647353c | 30 | * Copyright (c) 2023, Klara Inc. |
34dc7c2f BB |
31 | */ |
32 | ||
34dc7c2f | 33 | #include <sys/zfs_context.h> |
985c33b1 | 34 | #include <sys/zfs_chksum.h> |
34dc7c2f BB |
35 | #include <sys/spa_impl.h> |
36 | #include <sys/zio.h> | |
37 | #include <sys/zio_checksum.h> | |
38 | #include <sys/zio_compress.h> | |
39 | #include <sys/dmu.h> | |
40 | #include <sys/dmu_tx.h> | |
41 | #include <sys/zap.h> | |
42 | #include <sys/zil.h> | |
43 | #include <sys/vdev_impl.h> | |
619f0976 | 44 | #include <sys/vdev_initialize.h> |
1b939560 | 45 | #include <sys/vdev_trim.h> |
bc25c932 | 46 | #include <sys/vdev_file.h> |
ab9f4b0b | 47 | #include <sys/vdev_raidz.h> |
34dc7c2f BB |
48 | #include <sys/metaslab.h> |
49 | #include <sys/uberblock_impl.h> | |
50 | #include <sys/txg.h> | |
51 | #include <sys/avl.h> | |
52 | #include <sys/unique.h> | |
53 | #include <sys/dsl_pool.h> | |
54 | #include <sys/dsl_dir.h> | |
55 | #include <sys/dsl_prop.h> | |
26685276 | 56 | #include <sys/fm/util.h> |
428870ff | 57 | #include <sys/dsl_scan.h> |
34dc7c2f BB |
58 | #include <sys/fs/zfs.h> |
59 | #include <sys/metaslab_impl.h> | |
b128c09f | 60 | #include <sys/arc.h> |
67a1b037 | 61 | #include <sys/brt.h> |
428870ff | 62 | #include <sys/ddt.h> |
1421c891 | 63 | #include <sys/kstat.h> |
34dc7c2f | 64 | #include "zfs_prop.h" |
ca577779 | 65 | #include <sys/btree.h> |
3c67d83a | 66 | #include <sys/zfeature.h> |
bced7e3a | 67 | #include <sys/qat.h> |
10b3c7f5 | 68 | #include <sys/zstd/zstd.h> |
34dc7c2f BB |
69 | |
70 | /* | |
71 | * SPA locking | |
72 | * | |
93e28d66 | 73 | * There are three basic locks for managing spa_t structures: |
34dc7c2f BB |
74 | * |
75 | * spa_namespace_lock (global mutex) | |
76 | * | |
77 | * This lock must be acquired to do any of the following: | |
78 | * | |
79 | * - Lookup a spa_t by name | |
80 | * - Add or remove a spa_t from the namespace | |
81 | * - Increase spa_refcount from non-zero | |
82 | * - Check if spa_refcount is zero | |
83 | * - Rename a spa_t | |
84 | * - add/remove/attach/detach devices | |
85 | * - Held for the duration of create/destroy/import/export | |
86 | * | |
87 | * It does not need to handle recursion. A create or destroy may | |
88 | * reference objects (files or zvols) in other pools, but by | |
89 | * definition they must have an existing reference, and will never need | |
90 | * to lookup a spa_t by name. | |
91 | * | |
c13060e4 | 92 | * spa_refcount (per-spa zfs_refcount_t protected by mutex) |
34dc7c2f BB |
93 | * |
94 | * This reference count keep track of any active users of the spa_t. The | |
95 | * spa_t cannot be destroyed or freed while this is non-zero. Internally, | |
96 | * the refcount is never really 'zero' - opening a pool implicitly keeps | |
b128c09f | 97 | * some references in the DMU. Internally we check against spa_minref, but |
34dc7c2f BB |
98 | * present the image of a zero/non-zero value to consumers. |
99 | * | |
b128c09f | 100 | * spa_config_lock[] (per-spa array of rwlocks) |
34dc7c2f BB |
101 | * |
102 | * This protects the spa_t from config changes, and must be held in | |
103 | * the following circumstances: | |
104 | * | |
105 | * - RW_READER to perform I/O to the spa | |
106 | * - RW_WRITER to change the vdev config | |
107 | * | |
34dc7c2f BB |
108 | * The locking order is fairly straightforward: |
109 | * | |
110 | * spa_namespace_lock -> spa_refcount | |
111 | * | |
112 | * The namespace lock must be acquired to increase the refcount from 0 | |
113 | * or to check if it is zero. | |
114 | * | |
b128c09f | 115 | * spa_refcount -> spa_config_lock[] |
34dc7c2f BB |
116 | * |
117 | * There must be at least one valid reference on the spa_t to acquire | |
118 | * the config lock. | |
119 | * | |
b128c09f | 120 | * spa_namespace_lock -> spa_config_lock[] |
34dc7c2f BB |
121 | * |
122 | * The namespace lock must always be taken before the config lock. | |
123 | * | |
124 | * | |
b128c09f | 125 | * The spa_namespace_lock can be acquired directly and is globally visible. |
34dc7c2f | 126 | * |
b128c09f BB |
127 | * The namespace is manipulated using the following functions, all of which |
128 | * require the spa_namespace_lock to be held. | |
34dc7c2f BB |
129 | * |
130 | * spa_lookup() Lookup a spa_t by name. | |
131 | * | |
132 | * spa_add() Create a new spa_t in the namespace. | |
133 | * | |
134 | * spa_remove() Remove a spa_t from the namespace. This also | |
135 | * frees up any memory associated with the spa_t. | |
136 | * | |
137 | * spa_next() Returns the next spa_t in the system, or the | |
138 | * first if NULL is passed. | |
139 | * | |
140 | * spa_evict_all() Shutdown and remove all spa_t structures in | |
141 | * the system. | |
142 | * | |
143 | * spa_guid_exists() Determine whether a pool/device guid exists. | |
144 | * | |
145 | * The spa_refcount is manipulated using the following functions: | |
146 | * | |
147 | * spa_open_ref() Adds a reference to the given spa_t. Must be | |
148 | * called with spa_namespace_lock held if the | |
149 | * refcount is currently zero. | |
150 | * | |
151 | * spa_close() Remove a reference from the spa_t. This will | |
152 | * not free the spa_t or remove it from the | |
153 | * namespace. No locking is required. | |
154 | * | |
155 | * spa_refcount_zero() Returns true if the refcount is currently | |
156 | * zero. Must be called with spa_namespace_lock | |
157 | * held. | |
158 | * | |
b128c09f BB |
159 | * The spa_config_lock[] is an array of rwlocks, ordered as follows: |
160 | * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV. | |
161 | * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}(). | |
162 | * | |
163 | * To read the configuration, it suffices to hold one of these locks as reader. | |
164 | * To modify the configuration, you must hold all locks as writer. To modify | |
165 | * vdev state without altering the vdev tree's topology (e.g. online/offline), | |
166 | * you must hold SCL_STATE and SCL_ZIO as writer. | |
167 | * | |
168 | * We use these distinct config locks to avoid recursive lock entry. | |
169 | * For example, spa_sync() (which holds SCL_CONFIG as reader) induces | |
170 | * block allocations (SCL_ALLOC), which may require reading space maps | |
171 | * from disk (dmu_read() -> zio_read() -> SCL_ZIO). | |
172 | * | |
173 | * The spa config locks cannot be normal rwlocks because we need the | |
174 | * ability to hand off ownership. For example, SCL_ZIO is acquired | |
175 | * by the issuing thread and later released by an interrupt thread. | |
176 | * They do, however, obey the usual write-wanted semantics to prevent | |
177 | * writer (i.e. system administrator) starvation. | |
178 | * | |
179 | * The lock acquisition rules are as follows: | |
180 | * | |
181 | * SCL_CONFIG | |
182 | * Protects changes to the vdev tree topology, such as vdev | |
183 | * add/remove/attach/detach. Protects the dirty config list | |
184 | * (spa_config_dirty_list) and the set of spares and l2arc devices. | |
185 | * | |
186 | * SCL_STATE | |
187 | * Protects changes to pool state and vdev state, such as vdev | |
188 | * online/offline/fault/degrade/clear. Protects the dirty state list | |
189 | * (spa_state_dirty_list) and global pool state (spa_state). | |
190 | * | |
191 | * SCL_ALLOC | |
192 | * Protects changes to metaslab groups and classes. | |
193 | * Held as reader by metaslab_alloc() and metaslab_claim(). | |
194 | * | |
195 | * SCL_ZIO | |
196 | * Held by bp-level zios (those which have no io_vd upon entry) | |
197 | * to prevent changes to the vdev tree. The bp-level zio implicitly | |
198 | * protects all of its vdev child zios, which do not hold SCL_ZIO. | |
199 | * | |
200 | * SCL_FREE | |
201 | * Protects changes to metaslab groups and classes. | |
202 | * Held as reader by metaslab_free(). SCL_FREE is distinct from | |
203 | * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free | |
204 | * blocks in zio_done() while another i/o that holds either | |
205 | * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete. | |
206 | * | |
207 | * SCL_VDEV | |
208 | * Held as reader to prevent changes to the vdev tree during trivial | |
428870ff | 209 | * inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the |
b128c09f BB |
210 | * other locks, and lower than all of them, to ensure that it's safe |
211 | * to acquire regardless of caller context. | |
212 | * | |
213 | * In addition, the following rules apply: | |
214 | * | |
215 | * (a) spa_props_lock protects pool properties, spa_config and spa_config_list. | |
216 | * The lock ordering is SCL_CONFIG > spa_props_lock. | |
217 | * | |
218 | * (b) I/O operations on leaf vdevs. For any zio operation that takes | |
219 | * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(), | |
220 | * or zio_write_phys() -- the caller must ensure that the config cannot | |
221 | * cannot change in the interim, and that the vdev cannot be reopened. | |
222 | * SCL_STATE as reader suffices for both. | |
34dc7c2f BB |
223 | * |
224 | * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). | |
225 | * | |
226 | * spa_vdev_enter() Acquire the namespace lock and the config lock | |
227 | * for writing. | |
228 | * | |
229 | * spa_vdev_exit() Release the config lock, wait for all I/O | |
230 | * to complete, sync the updated configs to the | |
231 | * cache, and release the namespace lock. | |
232 | * | |
b128c09f BB |
233 | * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit(). |
234 | * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual | |
235 | * locking is, always, based on spa_namespace_lock and spa_config_lock[]. | |
34dc7c2f BB |
236 | */ |
237 | ||
238 | static avl_tree_t spa_namespace_avl; | |
239 | kmutex_t spa_namespace_lock; | |
240 | static kcondvar_t spa_namespace_cv; | |
18168da7 | 241 | static const int spa_max_replication_override = SPA_DVAS_PER_BP; |
34dc7c2f BB |
242 | |
243 | static kmutex_t spa_spare_lock; | |
244 | static avl_tree_t spa_spare_avl; | |
245 | static kmutex_t spa_l2cache_lock; | |
246 | static avl_tree_t spa_l2cache_avl; | |
247 | ||
da92d5cb | 248 | spa_mode_t spa_mode_global = SPA_MODE_UNINIT; |
34dc7c2f | 249 | |
0b39b9f9 | 250 | #ifdef ZFS_DEBUG |
a1d477c2 MA |
251 | /* |
252 | * Everything except dprintf, set_error, spa, and indirect_remap is on | |
253 | * by default in debug builds. | |
254 | */ | |
255 | int zfs_flags = ~(ZFS_DEBUG_DPRINTF | ZFS_DEBUG_SET_ERROR | | |
964c2d69 | 256 | ZFS_DEBUG_INDIRECT_REMAP); |
0b39b9f9 PS |
257 | #else |
258 | int zfs_flags = 0; | |
259 | #endif | |
260 | ||
261 | /* | |
262 | * zfs_recover can be set to nonzero to attempt to recover from | |
263 | * otherwise-fatal errors, typically caused by on-disk corruption. When | |
264 | * set, calls to zfs_panic_recover() will turn into warning messages. | |
265 | * This should only be used as a last resort, as it typically results | |
266 | * in leaked space, or worse. | |
267 | */ | |
268 | int zfs_recover = B_FALSE; | |
269 | ||
270 | /* | |
271 | * If destroy encounters an EIO while reading metadata (e.g. indirect | |
272 | * blocks), space referenced by the missing metadata can not be freed. | |
273 | * Normally this causes the background destroy to become "stalled", as | |
274 | * it is unable to make forward progress. While in this stalled state, | |
275 | * all remaining space to free from the error-encountering filesystem is | |
276 | * "temporarily leaked". Set this flag to cause it to ignore the EIO, | |
277 | * permanently leak the space from indirect blocks that can not be read, | |
278 | * and continue to free everything else that it can. | |
279 | * | |
280 | * The default, "stalling" behavior is useful if the storage partially | |
281 | * fails (i.e. some but not all i/os fail), and then later recovers. In | |
282 | * this case, we will be able to continue pool operations while it is | |
283 | * partially failed, and when it recovers, we can continue to free the | |
284 | * space, with no leaks. However, note that this case is actually | |
285 | * fairly rare. | |
286 | * | |
287 | * Typically pools either (a) fail completely (but perhaps temporarily, | |
288 | * e.g. a top-level vdev going offline), or (b) have localized, | |
289 | * permanent errors (e.g. disk returns the wrong data due to bit flip or | |
290 | * firmware bug). In case (a), this setting does not matter because the | |
291 | * pool will be suspended and the sync thread will not be able to make | |
292 | * forward progress regardless. In case (b), because the error is | |
293 | * permanent, the best we can do is leak the minimum amount of space, | |
294 | * which is what setting this flag will do. Therefore, it is reasonable | |
295 | * for this flag to normally be set, but we chose the more conservative | |
296 | * approach of not setting it, so that there is no possibility of | |
297 | * leaking space in the "partial temporary" failure case. | |
298 | */ | |
299 | int zfs_free_leak_on_eio = B_FALSE; | |
300 | ||
cc92e9d0 | 301 | /* |
e8b96c60 MA |
302 | * Expiration time in milliseconds. This value has two meanings. First it is |
303 | * used to determine when the spa_deadman() logic should fire. By default the | |
8fb1ede1 | 304 | * spa_deadman() will fire if spa_sync() has not completed in 600 seconds. |
e8b96c60 MA |
305 | * Secondly, the value determines if an I/O is considered "hung". Any I/O that |
306 | * has not completed in zfs_deadman_synctime_ms is considered "hung" resulting | |
8fb1ede1 | 307 | * in one of three behaviors controlled by zfs_deadman_failmode. |
cc92e9d0 | 308 | */ |
ab8d9c17 | 309 | uint64_t zfs_deadman_synctime_ms = 600000UL; /* 10 min. */ |
8fb1ede1 BB |
310 | |
311 | /* | |
312 | * This value controls the maximum amount of time zio_wait() will block for an | |
313 | * outstanding IO. By default this is 300 seconds at which point the "hung" | |
314 | * behavior will be applied as described for zfs_deadman_synctime_ms. | |
315 | */ | |
ab8d9c17 | 316 | uint64_t zfs_deadman_ziotime_ms = 300000UL; /* 5 min. */ |
cc92e9d0 | 317 | |
b81a3ddc TC |
318 | /* |
319 | * Check time in milliseconds. This defines the frequency at which we check | |
320 | * for hung I/O. | |
321 | */ | |
ab8d9c17 | 322 | uint64_t zfs_deadman_checktime_ms = 60000UL; /* 1 min. */ |
b81a3ddc | 323 | |
cc92e9d0 GW |
324 | /* |
325 | * By default the deadman is enabled. | |
326 | */ | |
18168da7 | 327 | int zfs_deadman_enabled = B_TRUE; |
cc92e9d0 | 328 | |
8fb1ede1 BB |
329 | /* |
330 | * Controls the behavior of the deadman when it detects a "hung" I/O. | |
331 | * Valid values are zfs_deadman_failmode=<wait|continue|panic>. | |
332 | * | |
333 | * wait - Wait for the "hung" I/O (default) | |
334 | * continue - Attempt to recover from a "hung" I/O | |
335 | * panic - Panic the system | |
336 | */ | |
18168da7 | 337 | const char *zfs_deadman_failmode = "wait"; |
8fb1ede1 | 338 | |
e8b96c60 MA |
339 | /* |
340 | * The worst case is single-sector max-parity RAID-Z blocks, in which | |
341 | * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1) | |
342 | * times the size; so just assume that. Add to this the fact that | |
343 | * we can have up to 3 DVAs per bp, and one more factor of 2 because | |
344 | * the block may be dittoed with up to 3 DVAs by ddt_sync(). All together, | |
345 | * the worst case is: | |
346 | * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2 == 24 | |
347 | */ | |
fdc2d303 | 348 | uint_t spa_asize_inflation = 24; |
e8b96c60 | 349 | |
3d45fdd6 MA |
350 | /* |
351 | * Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space in | |
f01eaed4 PS |
352 | * the pool to be consumed (bounded by spa_max_slop). This ensures that we |
353 | * don't run the pool completely out of space, due to unaccounted changes (e.g. | |
354 | * to the MOS). It also limits the worst-case time to allocate space. If we | |
355 | * have less than this amount of free space, most ZPL operations (e.g. write, | |
356 | * create) will return ENOSPC. The ZIL metaslabs (spa_embedded_log_class) are | |
357 | * also part of this 3.2% of space which can't be consumed by normal writes; | |
358 | * the slop space "proper" (spa_get_slop_space()) is decreased by the embedded | |
359 | * log space. | |
3d45fdd6 MA |
360 | * |
361 | * Certain operations (e.g. file removal, most administrative actions) can | |
362 | * use half the slop space. They will only return ENOSPC if less than half | |
363 | * the slop space is free. Typically, once the pool has less than the slop | |
364 | * space free, the user will use these operations to free up space in the pool. | |
365 | * These are the operations that call dsl_pool_adjustedsize() with the netfree | |
366 | * argument set to TRUE. | |
367 | * | |
d2734cce SD |
368 | * Operations that are almost guaranteed to free up space in the absence of |
369 | * a pool checkpoint can use up to three quarters of the slop space | |
370 | * (e.g zfs destroy). | |
371 | * | |
3d45fdd6 MA |
372 | * A very restricted set of operations are always permitted, regardless of |
373 | * the amount of free space. These are the operations that call | |
d2734cce SD |
374 | * dsl_sync_task(ZFS_SPACE_CHECK_NONE). If these operations result in a net |
375 | * increase in the amount of space used, it is possible to run the pool | |
376 | * completely out of space, causing it to be permanently read-only. | |
3d45fdd6 | 377 | * |
d7958b4c MA |
378 | * Note that on very small pools, the slop space will be larger than |
379 | * 3.2%, in an effort to have it be at least spa_min_slop (128MB), | |
380 | * but we never allow it to be more than half the pool size. | |
381 | * | |
f01eaed4 PS |
382 | * Further, on very large pools, the slop space will be smaller than |
383 | * 3.2%, to avoid reserving much more space than we actually need; bounded | |
384 | * by spa_max_slop (128GB). | |
385 | * | |
3d45fdd6 MA |
386 | * See also the comments in zfs_space_check_t. |
387 | */ | |
fdc2d303 | 388 | uint_t spa_slop_shift = 5; |
18168da7 AZ |
389 | static const uint64_t spa_min_slop = 128ULL * 1024 * 1024; |
390 | static const uint64_t spa_max_slop = 128ULL * 1024 * 1024 * 1024; | |
3bd4df38 EN |
391 | |
392 | /* | |
393 | * Number of allocators to use, per spa instance | |
394 | */ | |
395 | static int spa_num_allocators = 4; | |
492f64e9 | 396 | |
95f71c01 EN |
397 | /* |
398 | * Spa active allocator. | |
399 | * Valid values are zfs_active_allocator=<dynamic|cursor|new-dynamic>. | |
400 | */ | |
401 | const char *zfs_active_allocator = "dynamic"; | |
3d45fdd6 | 402 | |
4a0ee12a PZ |
403 | void |
404 | spa_load_failed(spa_t *spa, const char *fmt, ...) | |
405 | { | |
406 | va_list adx; | |
407 | char buf[256]; | |
408 | ||
409 | va_start(adx, fmt); | |
410 | (void) vsnprintf(buf, sizeof (buf), fmt, adx); | |
411 | va_end(adx); | |
412 | ||
6cb8e530 PZ |
413 | zfs_dbgmsg("spa_load(%s, config %s): FAILED: %s", spa->spa_name, |
414 | spa->spa_trust_config ? "trusted" : "untrusted", buf); | |
4a0ee12a PZ |
415 | } |
416 | ||
4a0ee12a PZ |
417 | void |
418 | spa_load_note(spa_t *spa, const char *fmt, ...) | |
419 | { | |
420 | va_list adx; | |
421 | char buf[256]; | |
422 | ||
423 | va_start(adx, fmt); | |
424 | (void) vsnprintf(buf, sizeof (buf), fmt, adx); | |
425 | va_end(adx); | |
426 | ||
6cb8e530 PZ |
427 | zfs_dbgmsg("spa_load(%s, config %s): %s", spa->spa_name, |
428 | spa->spa_trust_config ? "trusted" : "untrusted", buf); | |
687e4d7f DB |
429 | |
430 | spa_import_progress_set_notes_nolog(spa, "%s", buf); | |
4a0ee12a PZ |
431 | } |
432 | ||
cc99f275 DB |
433 | /* |
434 | * By default dedup and user data indirects land in the special class | |
435 | */ | |
18168da7 AZ |
436 | static int zfs_ddt_data_is_special = B_TRUE; |
437 | static int zfs_user_indirect_is_special = B_TRUE; | |
cc99f275 DB |
438 | |
439 | /* | |
440 | * The percentage of special class final space reserved for metadata only. | |
441 | * Once we allocate 100 - zfs_special_class_metadata_reserve_pct we only | |
442 | * let metadata into the class. | |
443 | */ | |
fdc2d303 | 444 | static uint_t zfs_special_class_metadata_reserve_pct = 25; |
cc99f275 | 445 | |
34dc7c2f BB |
446 | /* |
447 | * ========================================================================== | |
448 | * SPA config locking | |
449 | * ========================================================================== | |
450 | */ | |
451 | static void | |
b128c09f BB |
452 | spa_config_lock_init(spa_t *spa) |
453 | { | |
1c27024e | 454 | for (int i = 0; i < SCL_LOCKS; i++) { |
b128c09f BB |
455 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
456 | mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL); | |
457 | cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL); | |
b128c09f BB |
458 | scl->scl_writer = NULL; |
459 | scl->scl_write_wanted = 0; | |
42afb12d | 460 | scl->scl_count = 0; |
b128c09f | 461 | } |
34dc7c2f BB |
462 | } |
463 | ||
464 | static void | |
b128c09f BB |
465 | spa_config_lock_destroy(spa_t *spa) |
466 | { | |
1c27024e | 467 | for (int i = 0; i < SCL_LOCKS; i++) { |
b128c09f BB |
468 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
469 | mutex_destroy(&scl->scl_lock); | |
470 | cv_destroy(&scl->scl_cv); | |
b128c09f BB |
471 | ASSERT(scl->scl_writer == NULL); |
472 | ASSERT(scl->scl_write_wanted == 0); | |
42afb12d | 473 | ASSERT(scl->scl_count == 0); |
b128c09f BB |
474 | } |
475 | } | |
476 | ||
477 | int | |
a926aab9 | 478 | spa_config_tryenter(spa_t *spa, int locks, const void *tag, krw_t rw) |
34dc7c2f | 479 | { |
1c27024e | 480 | for (int i = 0; i < SCL_LOCKS; i++) { |
b128c09f BB |
481 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
482 | if (!(locks & (1 << i))) | |
483 | continue; | |
484 | mutex_enter(&scl->scl_lock); | |
485 | if (rw == RW_READER) { | |
486 | if (scl->scl_writer || scl->scl_write_wanted) { | |
487 | mutex_exit(&scl->scl_lock); | |
adfe9d93 SK |
488 | spa_config_exit(spa, locks & ((1 << i) - 1), |
489 | tag); | |
b128c09f BB |
490 | return (0); |
491 | } | |
492 | } else { | |
493 | ASSERT(scl->scl_writer != curthread); | |
42afb12d | 494 | if (scl->scl_count != 0) { |
b128c09f | 495 | mutex_exit(&scl->scl_lock); |
adfe9d93 SK |
496 | spa_config_exit(spa, locks & ((1 << i) - 1), |
497 | tag); | |
b128c09f BB |
498 | return (0); |
499 | } | |
500 | scl->scl_writer = curthread; | |
501 | } | |
42afb12d | 502 | scl->scl_count++; |
b128c09f BB |
503 | mutex_exit(&scl->scl_lock); |
504 | } | |
505 | return (1); | |
34dc7c2f BB |
506 | } |
507 | ||
71d191ef HW |
508 | static void |
509 | spa_config_enter_impl(spa_t *spa, int locks, const void *tag, krw_t rw, | |
510 | int mmp_flag) | |
34dc7c2f | 511 | { |
14e4e3cb | 512 | (void) tag; |
45d1cae3 BB |
513 | int wlocks_held = 0; |
514 | ||
13fe0198 MA |
515 | ASSERT3U(SCL_LOCKS, <, sizeof (wlocks_held) * NBBY); |
516 | ||
1c27024e | 517 | for (int i = 0; i < SCL_LOCKS; i++) { |
b128c09f | 518 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
45d1cae3 BB |
519 | if (scl->scl_writer == curthread) |
520 | wlocks_held |= (1 << i); | |
b128c09f BB |
521 | if (!(locks & (1 << i))) |
522 | continue; | |
523 | mutex_enter(&scl->scl_lock); | |
524 | if (rw == RW_READER) { | |
71d191ef HW |
525 | while (scl->scl_writer || |
526 | (!mmp_flag && scl->scl_write_wanted)) { | |
b128c09f BB |
527 | cv_wait(&scl->scl_cv, &scl->scl_lock); |
528 | } | |
529 | } else { | |
530 | ASSERT(scl->scl_writer != curthread); | |
42afb12d | 531 | while (scl->scl_count != 0) { |
b128c09f BB |
532 | scl->scl_write_wanted++; |
533 | cv_wait(&scl->scl_cv, &scl->scl_lock); | |
534 | scl->scl_write_wanted--; | |
535 | } | |
536 | scl->scl_writer = curthread; | |
537 | } | |
42afb12d | 538 | scl->scl_count++; |
b128c09f | 539 | mutex_exit(&scl->scl_lock); |
34dc7c2f | 540 | } |
a1d477c2 | 541 | ASSERT3U(wlocks_held, <=, locks); |
34dc7c2f BB |
542 | } |
543 | ||
71d191ef HW |
544 | void |
545 | spa_config_enter(spa_t *spa, int locks, const void *tag, krw_t rw) | |
546 | { | |
547 | spa_config_enter_impl(spa, locks, tag, rw, 0); | |
548 | } | |
549 | ||
550 | /* | |
551 | * The spa_config_enter_mmp() allows the mmp thread to cut in front of | |
552 | * outstanding write lock requests. This is needed since the mmp updates are | |
553 | * time sensitive and failure to service them promptly will result in a | |
554 | * suspended pool. This pool suspension has been seen in practice when there is | |
555 | * a single disk in a pool that is responding slowly and presumably about to | |
556 | * fail. | |
557 | */ | |
558 | ||
559 | void | |
560 | spa_config_enter_mmp(spa_t *spa, int locks, const void *tag, krw_t rw) | |
561 | { | |
562 | spa_config_enter_impl(spa, locks, tag, rw, 1); | |
563 | } | |
564 | ||
34dc7c2f | 565 | void |
dc04a8c7 | 566 | spa_config_exit(spa_t *spa, int locks, const void *tag) |
34dc7c2f | 567 | { |
14e4e3cb | 568 | (void) tag; |
1c27024e | 569 | for (int i = SCL_LOCKS - 1; i >= 0; i--) { |
b128c09f BB |
570 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
571 | if (!(locks & (1 << i))) | |
572 | continue; | |
573 | mutex_enter(&scl->scl_lock); | |
42afb12d AM |
574 | ASSERT(scl->scl_count > 0); |
575 | if (--scl->scl_count == 0) { | |
b128c09f BB |
576 | ASSERT(scl->scl_writer == NULL || |
577 | scl->scl_writer == curthread); | |
578 | scl->scl_writer = NULL; /* OK in either case */ | |
579 | cv_broadcast(&scl->scl_cv); | |
580 | } | |
581 | mutex_exit(&scl->scl_lock); | |
34dc7c2f | 582 | } |
34dc7c2f BB |
583 | } |
584 | ||
b128c09f BB |
585 | int |
586 | spa_config_held(spa_t *spa, int locks, krw_t rw) | |
34dc7c2f | 587 | { |
1c27024e | 588 | int locks_held = 0; |
34dc7c2f | 589 | |
1c27024e | 590 | for (int i = 0; i < SCL_LOCKS; i++) { |
b128c09f BB |
591 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
592 | if (!(locks & (1 << i))) | |
593 | continue; | |
42afb12d | 594 | if ((rw == RW_READER && scl->scl_count != 0) || |
b128c09f BB |
595 | (rw == RW_WRITER && scl->scl_writer == curthread)) |
596 | locks_held |= 1 << i; | |
597 | } | |
598 | ||
599 | return (locks_held); | |
34dc7c2f BB |
600 | } |
601 | ||
602 | /* | |
603 | * ========================================================================== | |
604 | * SPA namespace functions | |
605 | * ========================================================================== | |
606 | */ | |
607 | ||
608 | /* | |
609 | * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. | |
610 | * Returns NULL if no matching spa_t is found. | |
611 | */ | |
612 | spa_t * | |
613 | spa_lookup(const char *name) | |
614 | { | |
b128c09f BB |
615 | static spa_t search; /* spa_t is large; don't allocate on stack */ |
616 | spa_t *spa; | |
34dc7c2f | 617 | avl_index_t where; |
34dc7c2f BB |
618 | char *cp; |
619 | ||
620 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
621 | ||
13fe0198 MA |
622 | (void) strlcpy(search.spa_name, name, sizeof (search.spa_name)); |
623 | ||
34dc7c2f BB |
624 | /* |
625 | * If it's a full dataset name, figure out the pool name and | |
626 | * just use that. | |
627 | */ | |
da536844 | 628 | cp = strpbrk(search.spa_name, "/@#"); |
13fe0198 | 629 | if (cp != NULL) |
34dc7c2f | 630 | *cp = '\0'; |
34dc7c2f | 631 | |
34dc7c2f BB |
632 | spa = avl_find(&spa_namespace_avl, &search, &where); |
633 | ||
34dc7c2f BB |
634 | return (spa); |
635 | } | |
636 | ||
cc92e9d0 GW |
637 | /* |
638 | * Fires when spa_sync has not completed within zfs_deadman_synctime_ms. | |
639 | * If the zfs_deadman_enabled flag is set then it inspects all vdev queues | |
640 | * looking for potentially hung I/Os. | |
641 | */ | |
642 | void | |
643 | spa_deadman(void *arg) | |
644 | { | |
645 | spa_t *spa = arg; | |
646 | ||
b81a3ddc TC |
647 | /* Disable the deadman if the pool is suspended. */ |
648 | if (spa_suspended(spa)) | |
649 | return; | |
650 | ||
cc92e9d0 GW |
651 | zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu", |
652 | (gethrtime() - spa->spa_sync_starttime) / NANOSEC, | |
8e739b2c | 653 | (u_longlong_t)++spa->spa_deadman_calls); |
cc92e9d0 | 654 | if (zfs_deadman_enabled) |
8fb1ede1 | 655 | vdev_deadman(spa->spa_root_vdev, FTAG); |
cc92e9d0 | 656 | |
57ddcda1 | 657 | spa->spa_deadman_tqid = taskq_dispatch_delay(system_delay_taskq, |
f764edf0 | 658 | spa_deadman, spa, TQ_SLEEP, ddi_get_lbolt() + |
b81a3ddc | 659 | MSEC_TO_TICK(zfs_deadman_checktime_ms)); |
cc92e9d0 GW |
660 | } |
661 | ||
65c7cc49 | 662 | static int |
93e28d66 SD |
663 | spa_log_sm_sort_by_txg(const void *va, const void *vb) |
664 | { | |
665 | const spa_log_sm_t *a = va; | |
666 | const spa_log_sm_t *b = vb; | |
667 | ||
ca577779 | 668 | return (TREE_CMP(a->sls_txg, b->sls_txg)); |
93e28d66 SD |
669 | } |
670 | ||
34dc7c2f BB |
671 | /* |
672 | * Create an uninitialized spa_t with the given name. Requires | |
673 | * spa_namespace_lock. The caller must ensure that the spa_t doesn't already | |
674 | * exist by calling spa_lookup() first. | |
675 | */ | |
676 | spa_t * | |
428870ff | 677 | spa_add(const char *name, nvlist_t *config, const char *altroot) |
34dc7c2f BB |
678 | { |
679 | spa_t *spa; | |
b128c09f | 680 | spa_config_dirent_t *dp; |
34dc7c2f BB |
681 | |
682 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
683 | ||
79c76d5b | 684 | spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); |
34dc7c2f | 685 | |
34dc7c2f | 686 | mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 687 | mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); |
428870ff | 688 | mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); |
0c66c32d | 689 | mutex_init(&spa->spa_evicting_os_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 690 | mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); |
428870ff | 691 | mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 692 | mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); |
3c67d83a | 693 | mutex_init(&spa->spa_cksum_tmpls_lock, NULL, MUTEX_DEFAULT, NULL); |
428870ff BB |
694 | mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); |
695 | mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL); | |
696 | mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL); | |
4eb30c68 | 697 | mutex_init(&spa->spa_feat_stats_lock, NULL, MUTEX_DEFAULT, NULL); |
93e28d66 | 698 | mutex_init(&spa->spa_flushed_ms_lock, NULL, MUTEX_DEFAULT, NULL); |
e60e158e | 699 | mutex_init(&spa->spa_activities_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f BB |
700 | |
701 | cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); | |
0c66c32d | 702 | cv_init(&spa->spa_evicting_os_cv, NULL, CV_DEFAULT, NULL); |
428870ff | 703 | cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL); |
34dc7c2f | 704 | cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); |
b128c09f | 705 | cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL); |
e60e158e JG |
706 | cv_init(&spa->spa_activities_cv, NULL, CV_DEFAULT, NULL); |
707 | cv_init(&spa->spa_waiters_cv, NULL, CV_DEFAULT, NULL); | |
34dc7c2f | 708 | |
1c27024e | 709 | for (int t = 0; t < TXG_SIZE; t++) |
428870ff BB |
710 | bplist_create(&spa->spa_free_bplist[t]); |
711 | ||
b128c09f | 712 | (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name)); |
34dc7c2f BB |
713 | spa->spa_state = POOL_STATE_UNINITIALIZED; |
714 | spa->spa_freeze_txg = UINT64_MAX; | |
715 | spa->spa_final_txg = UINT64_MAX; | |
428870ff BB |
716 | spa->spa_load_max_txg = UINT64_MAX; |
717 | spa->spa_proc = &p0; | |
718 | spa->spa_proc_state = SPA_PROC_NONE; | |
6cb8e530 | 719 | spa->spa_trust_config = B_TRUE; |
25f06d67 | 720 | spa->spa_hostid = zone_get_hostid(NULL); |
34dc7c2f | 721 | |
e8b96c60 | 722 | spa->spa_deadman_synctime = MSEC2NSEC(zfs_deadman_synctime_ms); |
8fb1ede1 BB |
723 | spa->spa_deadman_ziotime = MSEC2NSEC(zfs_deadman_ziotime_ms); |
724 | spa_set_deadman_failmode(spa, zfs_deadman_failmode); | |
95f71c01 | 725 | spa_set_allocator(spa, zfs_active_allocator); |
cc92e9d0 | 726 | |
424fd7c3 | 727 | zfs_refcount_create(&spa->spa_refcount); |
b128c09f | 728 | spa_config_lock_init(spa); |
1421c891 | 729 | spa_stats_init(spa); |
34dc7c2f BB |
730 | |
731 | avl_add(&spa_namespace_avl, spa); | |
732 | ||
34dc7c2f BB |
733 | /* |
734 | * Set the alternate root, if there is one. | |
735 | */ | |
0336f3d0 | 736 | if (altroot) |
34dc7c2f | 737 | spa->spa_root = spa_strdup(altroot); |
34dc7c2f | 738 | |
3bd4df38 EN |
739 | /* Do not allow more allocators than CPUs. */ |
740 | spa->spa_alloc_count = MIN(MAX(spa_num_allocators, 1), boot_ncpus); | |
741 | ||
1b50749c AM |
742 | spa->spa_allocs = kmem_zalloc(spa->spa_alloc_count * |
743 | sizeof (spa_alloc_t), KM_SLEEP); | |
492f64e9 | 744 | for (int i = 0; i < spa->spa_alloc_count; i++) { |
1b50749c AM |
745 | mutex_init(&spa->spa_allocs[i].spaa_lock, NULL, MUTEX_DEFAULT, |
746 | NULL); | |
747 | avl_create(&spa->spa_allocs[i].spaa_tree, zio_bookmark_compare, | |
8469b5aa | 748 | sizeof (zio_t), offsetof(zio_t, io_queue_node.a)); |
492f64e9 | 749 | } |
3bd4df38 | 750 | |
93e28d66 SD |
751 | avl_create(&spa->spa_metaslabs_by_flushed, metaslab_sort_by_flushed, |
752 | sizeof (metaslab_t), offsetof(metaslab_t, ms_spa_txg_node)); | |
753 | avl_create(&spa->spa_sm_logs_by_txg, spa_log_sm_sort_by_txg, | |
754 | sizeof (spa_log_sm_t), offsetof(spa_log_sm_t, sls_node)); | |
755 | list_create(&spa->spa_log_summary, sizeof (log_summary_entry_t), | |
756 | offsetof(log_summary_entry_t, lse_node)); | |
3dfb57a3 | 757 | |
b128c09f BB |
758 | /* |
759 | * Every pool starts with the default cachefile | |
760 | */ | |
761 | list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t), | |
762 | offsetof(spa_config_dirent_t, scd_link)); | |
763 | ||
79c76d5b | 764 | dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP); |
428870ff | 765 | dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path); |
b128c09f BB |
766 | list_insert_head(&spa->spa_config_list, dp); |
767 | ||
572e2857 | 768 | VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME, |
79c76d5b | 769 | KM_SLEEP) == 0); |
572e2857 | 770 | |
9ae529ec CS |
771 | if (config != NULL) { |
772 | nvlist_t *features; | |
773 | ||
774 | if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ, | |
775 | &features) == 0) { | |
776 | VERIFY(nvlist_dup(features, &spa->spa_label_features, | |
777 | 0) == 0); | |
778 | } | |
779 | ||
428870ff | 780 | VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); |
9ae529ec CS |
781 | } |
782 | ||
783 | if (spa->spa_label_features == NULL) { | |
784 | VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME, | |
79c76d5b | 785 | KM_SLEEP) == 0); |
9ae529ec | 786 | } |
428870ff | 787 | |
c3520e7f MA |
788 | spa->spa_min_ashift = INT_MAX; |
789 | spa->spa_max_ashift = 0; | |
b2255edc | 790 | spa->spa_min_alloc = INT_MAX; |
d9bb583c | 791 | spa->spa_gcd_alloc = INT_MAX; |
c3520e7f | 792 | |
e8a20144 GN |
793 | /* Reset cached value */ |
794 | spa->spa_dedup_dspace = ~0ULL; | |
795 | ||
b0bc7a84 MG |
796 | /* |
797 | * As a pool is being created, treat all features as disabled by | |
798 | * setting SPA_FEATURE_DISABLED for all entries in the feature | |
799 | * refcount cache. | |
800 | */ | |
1c27024e | 801 | for (int i = 0; i < SPA_FEATURES; i++) { |
b0bc7a84 MG |
802 | spa->spa_feat_refcount_cache[i] = SPA_FEATURE_DISABLED; |
803 | } | |
804 | ||
3d31aad8 OF |
805 | list_create(&spa->spa_leaf_list, sizeof (vdev_t), |
806 | offsetof(vdev_t, vdev_leaf_node)); | |
807 | ||
34dc7c2f BB |
808 | return (spa); |
809 | } | |
810 | ||
811 | /* | |
812 | * Removes a spa_t from the namespace, freeing up any memory used. Requires | |
813 | * spa_namespace_lock. This is called only after the spa_t has been closed and | |
814 | * deactivated. | |
815 | */ | |
816 | void | |
817 | spa_remove(spa_t *spa) | |
818 | { | |
b128c09f BB |
819 | spa_config_dirent_t *dp; |
820 | ||
34dc7c2f | 821 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); |
93e28d66 | 822 | ASSERT(spa_state(spa) == POOL_STATE_UNINITIALIZED); |
424fd7c3 | 823 | ASSERT3U(zfs_refcount_count(&spa->spa_refcount), ==, 0); |
e60e158e | 824 | ASSERT0(spa->spa_waiters); |
34dc7c2f | 825 | |
428870ff BB |
826 | nvlist_free(spa->spa_config_splitting); |
827 | ||
34dc7c2f BB |
828 | avl_remove(&spa_namespace_avl, spa); |
829 | cv_broadcast(&spa_namespace_cv); | |
830 | ||
0336f3d0 | 831 | if (spa->spa_root) |
34dc7c2f | 832 | spa_strfree(spa->spa_root); |
34dc7c2f | 833 | |
b3ad3f48 | 834 | while ((dp = list_remove_head(&spa->spa_config_list)) != NULL) { |
b128c09f BB |
835 | if (dp->scd_path != NULL) |
836 | spa_strfree(dp->scd_path); | |
837 | kmem_free(dp, sizeof (spa_config_dirent_t)); | |
838 | } | |
34dc7c2f | 839 | |
492f64e9 | 840 | for (int i = 0; i < spa->spa_alloc_count; i++) { |
1b50749c AM |
841 | avl_destroy(&spa->spa_allocs[i].spaa_tree); |
842 | mutex_destroy(&spa->spa_allocs[i].spaa_lock); | |
492f64e9 | 843 | } |
1b50749c AM |
844 | kmem_free(spa->spa_allocs, spa->spa_alloc_count * |
845 | sizeof (spa_alloc_t)); | |
492f64e9 | 846 | |
93e28d66 SD |
847 | avl_destroy(&spa->spa_metaslabs_by_flushed); |
848 | avl_destroy(&spa->spa_sm_logs_by_txg); | |
849 | list_destroy(&spa->spa_log_summary); | |
b128c09f | 850 | list_destroy(&spa->spa_config_list); |
3d31aad8 | 851 | list_destroy(&spa->spa_leaf_list); |
34dc7c2f | 852 | |
9ae529ec | 853 | nvlist_free(spa->spa_label_features); |
572e2857 | 854 | nvlist_free(spa->spa_load_info); |
417104bd | 855 | nvlist_free(spa->spa_feat_stats); |
34dc7c2f BB |
856 | spa_config_set(spa, NULL); |
857 | ||
424fd7c3 | 858 | zfs_refcount_destroy(&spa->spa_refcount); |
34dc7c2f | 859 | |
1421c891 | 860 | spa_stats_destroy(spa); |
b128c09f | 861 | spa_config_lock_destroy(spa); |
34dc7c2f | 862 | |
1c27024e | 863 | for (int t = 0; t < TXG_SIZE; t++) |
428870ff BB |
864 | bplist_destroy(&spa->spa_free_bplist[t]); |
865 | ||
3c67d83a TH |
866 | zio_checksum_templates_free(spa); |
867 | ||
34dc7c2f | 868 | cv_destroy(&spa->spa_async_cv); |
0c66c32d | 869 | cv_destroy(&spa->spa_evicting_os_cv); |
428870ff | 870 | cv_destroy(&spa->spa_proc_cv); |
34dc7c2f | 871 | cv_destroy(&spa->spa_scrub_io_cv); |
b128c09f | 872 | cv_destroy(&spa->spa_suspend_cv); |
e60e158e JG |
873 | cv_destroy(&spa->spa_activities_cv); |
874 | cv_destroy(&spa->spa_waiters_cv); | |
34dc7c2f | 875 | |
93e28d66 | 876 | mutex_destroy(&spa->spa_flushed_ms_lock); |
34dc7c2f | 877 | mutex_destroy(&spa->spa_async_lock); |
34dc7c2f | 878 | mutex_destroy(&spa->spa_errlist_lock); |
428870ff | 879 | mutex_destroy(&spa->spa_errlog_lock); |
0c66c32d | 880 | mutex_destroy(&spa->spa_evicting_os_lock); |
34dc7c2f | 881 | mutex_destroy(&spa->spa_history_lock); |
428870ff | 882 | mutex_destroy(&spa->spa_proc_lock); |
34dc7c2f | 883 | mutex_destroy(&spa->spa_props_lock); |
3c67d83a | 884 | mutex_destroy(&spa->spa_cksum_tmpls_lock); |
428870ff | 885 | mutex_destroy(&spa->spa_scrub_lock); |
b128c09f | 886 | mutex_destroy(&spa->spa_suspend_lock); |
428870ff | 887 | mutex_destroy(&spa->spa_vdev_top_lock); |
4eb30c68 | 888 | mutex_destroy(&spa->spa_feat_stats_lock); |
e60e158e | 889 | mutex_destroy(&spa->spa_activities_lock); |
34dc7c2f BB |
890 | |
891 | kmem_free(spa, sizeof (spa_t)); | |
892 | } | |
893 | ||
894 | /* | |
895 | * Given a pool, return the next pool in the namespace, or NULL if there is | |
896 | * none. If 'prev' is NULL, return the first pool. | |
897 | */ | |
898 | spa_t * | |
899 | spa_next(spa_t *prev) | |
900 | { | |
901 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
902 | ||
903 | if (prev) | |
904 | return (AVL_NEXT(&spa_namespace_avl, prev)); | |
905 | else | |
906 | return (avl_first(&spa_namespace_avl)); | |
907 | } | |
908 | ||
909 | /* | |
910 | * ========================================================================== | |
911 | * SPA refcount functions | |
912 | * ========================================================================== | |
913 | */ | |
914 | ||
915 | /* | |
916 | * Add a reference to the given spa_t. Must have at least one reference, or | |
917 | * have the namespace lock held. | |
918 | */ | |
919 | void | |
a926aab9 | 920 | spa_open_ref(spa_t *spa, const void *tag) |
34dc7c2f | 921 | { |
424fd7c3 | 922 | ASSERT(zfs_refcount_count(&spa->spa_refcount) >= spa->spa_minref || |
34dc7c2f | 923 | MUTEX_HELD(&spa_namespace_lock)); |
c13060e4 | 924 | (void) zfs_refcount_add(&spa->spa_refcount, tag); |
34dc7c2f BB |
925 | } |
926 | ||
927 | /* | |
928 | * Remove a reference to the given spa_t. Must have at least one reference, or | |
929 | * have the namespace lock held. | |
930 | */ | |
931 | void | |
a926aab9 | 932 | spa_close(spa_t *spa, const void *tag) |
34dc7c2f | 933 | { |
424fd7c3 | 934 | ASSERT(zfs_refcount_count(&spa->spa_refcount) > spa->spa_minref || |
34dc7c2f | 935 | MUTEX_HELD(&spa_namespace_lock)); |
424fd7c3 | 936 | (void) zfs_refcount_remove(&spa->spa_refcount, tag); |
34dc7c2f BB |
937 | } |
938 | ||
0c66c32d JG |
939 | /* |
940 | * Remove a reference to the given spa_t held by a dsl dir that is | |
941 | * being asynchronously released. Async releases occur from a taskq | |
942 | * performing eviction of dsl datasets and dirs. The namespace lock | |
943 | * isn't held and the hold by the object being evicted may contribute to | |
944 | * spa_minref (e.g. dataset or directory released during pool export), | |
945 | * so the asserts in spa_close() do not apply. | |
946 | */ | |
947 | void | |
a926aab9 | 948 | spa_async_close(spa_t *spa, const void *tag) |
0c66c32d | 949 | { |
424fd7c3 | 950 | (void) zfs_refcount_remove(&spa->spa_refcount, tag); |
0c66c32d JG |
951 | } |
952 | ||
34dc7c2f BB |
953 | /* |
954 | * Check to see if the spa refcount is zero. Must be called with | |
b128c09f | 955 | * spa_namespace_lock held. We really compare against spa_minref, which is the |
34dc7c2f BB |
956 | * number of references acquired when opening a pool |
957 | */ | |
958 | boolean_t | |
959 | spa_refcount_zero(spa_t *spa) | |
960 | { | |
961 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
962 | ||
424fd7c3 | 963 | return (zfs_refcount_count(&spa->spa_refcount) == spa->spa_minref); |
34dc7c2f BB |
964 | } |
965 | ||
966 | /* | |
967 | * ========================================================================== | |
968 | * SPA spare and l2cache tracking | |
969 | * ========================================================================== | |
970 | */ | |
971 | ||
972 | /* | |
973 | * Hot spares and cache devices are tracked using the same code below, | |
974 | * for 'auxiliary' devices. | |
975 | */ | |
976 | ||
977 | typedef struct spa_aux { | |
978 | uint64_t aux_guid; | |
979 | uint64_t aux_pool; | |
980 | avl_node_t aux_avl; | |
981 | int aux_count; | |
982 | } spa_aux_t; | |
983 | ||
ee36c709 | 984 | static inline int |
34dc7c2f BB |
985 | spa_aux_compare(const void *a, const void *b) |
986 | { | |
ee36c709 GN |
987 | const spa_aux_t *sa = (const spa_aux_t *)a; |
988 | const spa_aux_t *sb = (const spa_aux_t *)b; | |
34dc7c2f | 989 | |
ca577779 | 990 | return (TREE_CMP(sa->aux_guid, sb->aux_guid)); |
34dc7c2f BB |
991 | } |
992 | ||
65c7cc49 | 993 | static void |
34dc7c2f BB |
994 | spa_aux_add(vdev_t *vd, avl_tree_t *avl) |
995 | { | |
996 | avl_index_t where; | |
997 | spa_aux_t search; | |
998 | spa_aux_t *aux; | |
999 | ||
1000 | search.aux_guid = vd->vdev_guid; | |
1001 | if ((aux = avl_find(avl, &search, &where)) != NULL) { | |
1002 | aux->aux_count++; | |
1003 | } else { | |
79c76d5b | 1004 | aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP); |
34dc7c2f BB |
1005 | aux->aux_guid = vd->vdev_guid; |
1006 | aux->aux_count = 1; | |
1007 | avl_insert(avl, aux, where); | |
1008 | } | |
1009 | } | |
1010 | ||
65c7cc49 | 1011 | static void |
34dc7c2f BB |
1012 | spa_aux_remove(vdev_t *vd, avl_tree_t *avl) |
1013 | { | |
1014 | spa_aux_t search; | |
1015 | spa_aux_t *aux; | |
1016 | avl_index_t where; | |
1017 | ||
1018 | search.aux_guid = vd->vdev_guid; | |
1019 | aux = avl_find(avl, &search, &where); | |
1020 | ||
1021 | ASSERT(aux != NULL); | |
1022 | ||
1023 | if (--aux->aux_count == 0) { | |
1024 | avl_remove(avl, aux); | |
1025 | kmem_free(aux, sizeof (spa_aux_t)); | |
1026 | } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { | |
1027 | aux->aux_pool = 0ULL; | |
1028 | } | |
1029 | } | |
1030 | ||
65c7cc49 | 1031 | static boolean_t |
b128c09f | 1032 | spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl) |
34dc7c2f BB |
1033 | { |
1034 | spa_aux_t search, *found; | |
34dc7c2f BB |
1035 | |
1036 | search.aux_guid = guid; | |
b128c09f | 1037 | found = avl_find(avl, &search, NULL); |
34dc7c2f BB |
1038 | |
1039 | if (pool) { | |
1040 | if (found) | |
1041 | *pool = found->aux_pool; | |
1042 | else | |
1043 | *pool = 0ULL; | |
1044 | } | |
1045 | ||
b128c09f BB |
1046 | if (refcnt) { |
1047 | if (found) | |
1048 | *refcnt = found->aux_count; | |
1049 | else | |
1050 | *refcnt = 0; | |
1051 | } | |
1052 | ||
34dc7c2f BB |
1053 | return (found != NULL); |
1054 | } | |
1055 | ||
65c7cc49 | 1056 | static void |
34dc7c2f BB |
1057 | spa_aux_activate(vdev_t *vd, avl_tree_t *avl) |
1058 | { | |
1059 | spa_aux_t search, *found; | |
1060 | avl_index_t where; | |
1061 | ||
1062 | search.aux_guid = vd->vdev_guid; | |
1063 | found = avl_find(avl, &search, &where); | |
1064 | ASSERT(found != NULL); | |
1065 | ASSERT(found->aux_pool == 0ULL); | |
1066 | ||
1067 | found->aux_pool = spa_guid(vd->vdev_spa); | |
1068 | } | |
1069 | ||
1070 | /* | |
1071 | * Spares are tracked globally due to the following constraints: | |
1072 | * | |
aa755b35 MA |
1073 | * - A spare may be part of multiple pools. |
1074 | * - A spare may be added to a pool even if it's actively in use within | |
34dc7c2f | 1075 | * another pool. |
aa755b35 | 1076 | * - A spare in use in any pool can only be the source of a replacement if |
34dc7c2f BB |
1077 | * the target is a spare in the same pool. |
1078 | * | |
1079 | * We keep track of all spares on the system through the use of a reference | |
1080 | * counted AVL tree. When a vdev is added as a spare, or used as a replacement | |
1081 | * spare, then we bump the reference count in the AVL tree. In addition, we set | |
1082 | * the 'vdev_isspare' member to indicate that the device is a spare (active or | |
1083 | * inactive). When a spare is made active (used to replace a device in the | |
1084 | * pool), we also keep track of which pool its been made a part of. | |
1085 | * | |
1086 | * The 'spa_spare_lock' protects the AVL tree. These functions are normally | |
1087 | * called under the spa_namespace lock as part of vdev reconfiguration. The | |
1088 | * separate spare lock exists for the status query path, which does not need to | |
1089 | * be completely consistent with respect to other vdev configuration changes. | |
1090 | */ | |
1091 | ||
1092 | static int | |
1093 | spa_spare_compare(const void *a, const void *b) | |
1094 | { | |
1095 | return (spa_aux_compare(a, b)); | |
1096 | } | |
1097 | ||
1098 | void | |
1099 | spa_spare_add(vdev_t *vd) | |
1100 | { | |
1101 | mutex_enter(&spa_spare_lock); | |
1102 | ASSERT(!vd->vdev_isspare); | |
1103 | spa_aux_add(vd, &spa_spare_avl); | |
1104 | vd->vdev_isspare = B_TRUE; | |
1105 | mutex_exit(&spa_spare_lock); | |
1106 | } | |
1107 | ||
1108 | void | |
1109 | spa_spare_remove(vdev_t *vd) | |
1110 | { | |
1111 | mutex_enter(&spa_spare_lock); | |
1112 | ASSERT(vd->vdev_isspare); | |
1113 | spa_aux_remove(vd, &spa_spare_avl); | |
1114 | vd->vdev_isspare = B_FALSE; | |
1115 | mutex_exit(&spa_spare_lock); | |
1116 | } | |
1117 | ||
1118 | boolean_t | |
b128c09f | 1119 | spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt) |
34dc7c2f BB |
1120 | { |
1121 | boolean_t found; | |
1122 | ||
1123 | mutex_enter(&spa_spare_lock); | |
b128c09f | 1124 | found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl); |
34dc7c2f BB |
1125 | mutex_exit(&spa_spare_lock); |
1126 | ||
1127 | return (found); | |
1128 | } | |
1129 | ||
1130 | void | |
1131 | spa_spare_activate(vdev_t *vd) | |
1132 | { | |
1133 | mutex_enter(&spa_spare_lock); | |
1134 | ASSERT(vd->vdev_isspare); | |
1135 | spa_aux_activate(vd, &spa_spare_avl); | |
1136 | mutex_exit(&spa_spare_lock); | |
1137 | } | |
1138 | ||
1139 | /* | |
1140 | * Level 2 ARC devices are tracked globally for the same reasons as spares. | |
1141 | * Cache devices currently only support one pool per cache device, and so | |
1142 | * for these devices the aux reference count is currently unused beyond 1. | |
1143 | */ | |
1144 | ||
1145 | static int | |
1146 | spa_l2cache_compare(const void *a, const void *b) | |
1147 | { | |
1148 | return (spa_aux_compare(a, b)); | |
1149 | } | |
1150 | ||
1151 | void | |
1152 | spa_l2cache_add(vdev_t *vd) | |
1153 | { | |
1154 | mutex_enter(&spa_l2cache_lock); | |
1155 | ASSERT(!vd->vdev_isl2cache); | |
1156 | spa_aux_add(vd, &spa_l2cache_avl); | |
1157 | vd->vdev_isl2cache = B_TRUE; | |
1158 | mutex_exit(&spa_l2cache_lock); | |
1159 | } | |
1160 | ||
1161 | void | |
1162 | spa_l2cache_remove(vdev_t *vd) | |
1163 | { | |
1164 | mutex_enter(&spa_l2cache_lock); | |
1165 | ASSERT(vd->vdev_isl2cache); | |
1166 | spa_aux_remove(vd, &spa_l2cache_avl); | |
1167 | vd->vdev_isl2cache = B_FALSE; | |
1168 | mutex_exit(&spa_l2cache_lock); | |
1169 | } | |
1170 | ||
1171 | boolean_t | |
1172 | spa_l2cache_exists(uint64_t guid, uint64_t *pool) | |
1173 | { | |
1174 | boolean_t found; | |
1175 | ||
1176 | mutex_enter(&spa_l2cache_lock); | |
b128c09f | 1177 | found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl); |
34dc7c2f BB |
1178 | mutex_exit(&spa_l2cache_lock); |
1179 | ||
1180 | return (found); | |
1181 | } | |
1182 | ||
1183 | void | |
1184 | spa_l2cache_activate(vdev_t *vd) | |
1185 | { | |
1186 | mutex_enter(&spa_l2cache_lock); | |
1187 | ASSERT(vd->vdev_isl2cache); | |
1188 | spa_aux_activate(vd, &spa_l2cache_avl); | |
1189 | mutex_exit(&spa_l2cache_lock); | |
1190 | } | |
1191 | ||
34dc7c2f BB |
1192 | /* |
1193 | * ========================================================================== | |
1194 | * SPA vdev locking | |
1195 | * ========================================================================== | |
1196 | */ | |
1197 | ||
1198 | /* | |
1199 | * Lock the given spa_t for the purpose of adding or removing a vdev. | |
1200 | * Grabs the global spa_namespace_lock plus the spa config lock for writing. | |
1201 | * It returns the next transaction group for the spa_t. | |
1202 | */ | |
1203 | uint64_t | |
1204 | spa_vdev_enter(spa_t *spa) | |
1205 | { | |
428870ff | 1206 | mutex_enter(&spa->spa_vdev_top_lock); |
34dc7c2f | 1207 | mutex_enter(&spa_namespace_lock); |
1b939560 BB |
1208 | |
1209 | vdev_autotrim_stop_all(spa); | |
1210 | ||
428870ff BB |
1211 | return (spa_vdev_config_enter(spa)); |
1212 | } | |
1213 | ||
9a49d3f3 BB |
1214 | /* |
1215 | * The same as spa_vdev_enter() above but additionally takes the guid of | |
1216 | * the vdev being detached. When there is a rebuild in process it will be | |
1217 | * suspended while the vdev tree is modified then resumed by spa_vdev_exit(). | |
1218 | * The rebuild is canceled if only a single child remains after the detach. | |
1219 | */ | |
1220 | uint64_t | |
1221 | spa_vdev_detach_enter(spa_t *spa, uint64_t guid) | |
1222 | { | |
1223 | mutex_enter(&spa->spa_vdev_top_lock); | |
1224 | mutex_enter(&spa_namespace_lock); | |
1225 | ||
1226 | vdev_autotrim_stop_all(spa); | |
1227 | ||
1228 | if (guid != 0) { | |
1229 | vdev_t *vd = spa_lookup_by_guid(spa, guid, B_FALSE); | |
1230 | if (vd) { | |
1231 | vdev_rebuild_stop_wait(vd->vdev_top); | |
1232 | } | |
1233 | } | |
1234 | ||
1235 | return (spa_vdev_config_enter(spa)); | |
1236 | } | |
1237 | ||
428870ff BB |
1238 | /* |
1239 | * Internal implementation for spa_vdev_enter(). Used when a vdev | |
1240 | * operation requires multiple syncs (i.e. removing a device) while | |
1241 | * keeping the spa_namespace_lock held. | |
1242 | */ | |
1243 | uint64_t | |
1244 | spa_vdev_config_enter(spa_t *spa) | |
1245 | { | |
1246 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
34dc7c2f | 1247 | |
b128c09f | 1248 | spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); |
34dc7c2f BB |
1249 | |
1250 | return (spa_last_synced_txg(spa) + 1); | |
1251 | } | |
1252 | ||
1253 | /* | |
428870ff BB |
1254 | * Used in combination with spa_vdev_config_enter() to allow the syncing |
1255 | * of multiple transactions without releasing the spa_namespace_lock. | |
34dc7c2f | 1256 | */ |
428870ff | 1257 | void |
dd66857d AZ |
1258 | spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, |
1259 | const char *tag) | |
34dc7c2f | 1260 | { |
1c27024e DB |
1261 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); |
1262 | ||
34dc7c2f BB |
1263 | int config_changed = B_FALSE; |
1264 | ||
1265 | ASSERT(txg > spa_last_synced_txg(spa)); | |
1266 | ||
b128c09f BB |
1267 | spa->spa_pending_vdev = NULL; |
1268 | ||
34dc7c2f BB |
1269 | /* |
1270 | * Reassess the DTLs. | |
1271 | */ | |
9a49d3f3 | 1272 | vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE, B_FALSE); |
34dc7c2f | 1273 | |
b128c09f | 1274 | if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) { |
34dc7c2f | 1275 | config_changed = B_TRUE; |
428870ff | 1276 | spa->spa_config_generation++; |
34dc7c2f BB |
1277 | } |
1278 | ||
428870ff BB |
1279 | /* |
1280 | * Verify the metaslab classes. | |
1281 | */ | |
1282 | ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0); | |
1283 | ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0); | |
aa755b35 | 1284 | ASSERT(metaslab_class_validate(spa_embedded_log_class(spa)) == 0); |
cc99f275 DB |
1285 | ASSERT(metaslab_class_validate(spa_special_class(spa)) == 0); |
1286 | ASSERT(metaslab_class_validate(spa_dedup_class(spa)) == 0); | |
428870ff | 1287 | |
b128c09f | 1288 | spa_config_exit(spa, SCL_ALL, spa); |
34dc7c2f | 1289 | |
428870ff BB |
1290 | /* |
1291 | * Panic the system if the specified tag requires it. This | |
1292 | * is useful for ensuring that configurations are updated | |
1293 | * transactionally. | |
1294 | */ | |
1295 | if (zio_injection_enabled) | |
1296 | zio_handle_panic_injection(spa, tag, 0); | |
1297 | ||
34dc7c2f BB |
1298 | /* |
1299 | * Note: this txg_wait_synced() is important because it ensures | |
1300 | * that there won't be more than one config change per txg. | |
1301 | * This allows us to use the txg as the generation number. | |
1302 | */ | |
1303 | if (error == 0) | |
1304 | txg_wait_synced(spa->spa_dsl_pool, txg); | |
1305 | ||
1306 | if (vd != NULL) { | |
93cf2076 | 1307 | ASSERT(!vd->vdev_detached || vd->vdev_dtl_sm == NULL); |
619f0976 GW |
1308 | if (vd->vdev_ops->vdev_op_leaf) { |
1309 | mutex_enter(&vd->vdev_initialize_lock); | |
c10d37dd GW |
1310 | vdev_initialize_stop(vd, VDEV_INITIALIZE_CANCELED, |
1311 | NULL); | |
619f0976 | 1312 | mutex_exit(&vd->vdev_initialize_lock); |
1b939560 BB |
1313 | |
1314 | mutex_enter(&vd->vdev_trim_lock); | |
1315 | vdev_trim_stop(vd, VDEV_TRIM_CANCELED, NULL); | |
1316 | mutex_exit(&vd->vdev_trim_lock); | |
619f0976 GW |
1317 | } |
1318 | ||
1b939560 BB |
1319 | /* |
1320 | * The vdev may be both a leaf and top-level device. | |
1321 | */ | |
1322 | vdev_autotrim_stop_wait(vd); | |
1323 | ||
75a089ed | 1324 | spa_config_enter(spa, SCL_STATE_ALL, spa, RW_WRITER); |
34dc7c2f | 1325 | vdev_free(vd); |
75a089ed | 1326 | spa_config_exit(spa, SCL_STATE_ALL, spa); |
34dc7c2f BB |
1327 | } |
1328 | ||
1329 | /* | |
1330 | * If the config changed, update the config cache. | |
1331 | */ | |
1332 | if (config_changed) | |
3a74f488 | 1333 | spa_write_cachefile(spa, B_FALSE, B_TRUE, B_TRUE); |
428870ff | 1334 | } |
34dc7c2f | 1335 | |
428870ff BB |
1336 | /* |
1337 | * Unlock the spa_t after adding or removing a vdev. Besides undoing the | |
1338 | * locking of spa_vdev_enter(), we also want make sure the transactions have | |
1339 | * synced to disk, and then update the global configuration cache with the new | |
1340 | * information. | |
1341 | */ | |
1342 | int | |
1343 | spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) | |
1344 | { | |
1b939560 | 1345 | vdev_autotrim_restart(spa); |
9a49d3f3 | 1346 | vdev_rebuild_restart(spa); |
1b939560 | 1347 | |
428870ff | 1348 | spa_vdev_config_exit(spa, vd, txg, error, FTAG); |
34dc7c2f | 1349 | mutex_exit(&spa_namespace_lock); |
428870ff | 1350 | mutex_exit(&spa->spa_vdev_top_lock); |
34dc7c2f BB |
1351 | |
1352 | return (error); | |
1353 | } | |
1354 | ||
b128c09f BB |
1355 | /* |
1356 | * Lock the given spa_t for the purpose of changing vdev state. | |
1357 | */ | |
1358 | void | |
428870ff | 1359 | spa_vdev_state_enter(spa_t *spa, int oplocks) |
b128c09f | 1360 | { |
428870ff BB |
1361 | int locks = SCL_STATE_ALL | oplocks; |
1362 | ||
1363 | /* | |
1364 | * Root pools may need to read of the underlying devfs filesystem | |
1365 | * when opening up a vdev. Unfortunately if we're holding the | |
1366 | * SCL_ZIO lock it will result in a deadlock when we try to issue | |
1367 | * the read from the root filesystem. Instead we "prefetch" | |
1368 | * the associated vnodes that we need prior to opening the | |
1369 | * underlying devices and cache them so that we can prevent | |
1370 | * any I/O when we are doing the actual open. | |
1371 | */ | |
1372 | if (spa_is_root(spa)) { | |
1373 | int low = locks & ~(SCL_ZIO - 1); | |
1374 | int high = locks & ~low; | |
1375 | ||
1376 | spa_config_enter(spa, high, spa, RW_WRITER); | |
1377 | vdev_hold(spa->spa_root_vdev); | |
1378 | spa_config_enter(spa, low, spa, RW_WRITER); | |
1379 | } else { | |
1380 | spa_config_enter(spa, locks, spa, RW_WRITER); | |
1381 | } | |
1382 | spa->spa_vdev_locks = locks; | |
b128c09f BB |
1383 | } |
1384 | ||
1385 | int | |
1386 | spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error) | |
1387 | { | |
428870ff | 1388 | boolean_t config_changed = B_FALSE; |
4a283c7f TH |
1389 | vdev_t *vdev_top; |
1390 | ||
1391 | if (vd == NULL || vd == spa->spa_root_vdev) { | |
1392 | vdev_top = spa->spa_root_vdev; | |
1393 | } else { | |
1394 | vdev_top = vd->vdev_top; | |
1395 | } | |
428870ff BB |
1396 | |
1397 | if (vd != NULL || error == 0) | |
9a49d3f3 | 1398 | vdev_dtl_reassess(vdev_top, 0, 0, B_FALSE, B_FALSE); |
428870ff BB |
1399 | |
1400 | if (vd != NULL) { | |
4a283c7f TH |
1401 | if (vd != spa->spa_root_vdev) |
1402 | vdev_state_dirty(vdev_top); | |
1403 | ||
428870ff BB |
1404 | config_changed = B_TRUE; |
1405 | spa->spa_config_generation++; | |
1406 | } | |
b128c09f | 1407 | |
428870ff BB |
1408 | if (spa_is_root(spa)) |
1409 | vdev_rele(spa->spa_root_vdev); | |
1410 | ||
1411 | ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL); | |
1412 | spa_config_exit(spa, spa->spa_vdev_locks, spa); | |
b128c09f | 1413 | |
fb5f0bc8 BB |
1414 | /* |
1415 | * If anything changed, wait for it to sync. This ensures that, | |
76d04993 | 1416 | * from the system administrator's perspective, zpool(8) commands |
fb5f0bc8 BB |
1417 | * are synchronous. This is important for things like zpool offline: |
1418 | * when the command completes, you expect no further I/O from ZFS. | |
1419 | */ | |
1420 | if (vd != NULL) | |
1421 | txg_wait_synced(spa->spa_dsl_pool, 0); | |
1422 | ||
428870ff BB |
1423 | /* |
1424 | * If the config changed, update the config cache. | |
1425 | */ | |
1426 | if (config_changed) { | |
1427 | mutex_enter(&spa_namespace_lock); | |
55c12724 | 1428 | spa_write_cachefile(spa, B_FALSE, B_TRUE, B_FALSE); |
428870ff BB |
1429 | mutex_exit(&spa_namespace_lock); |
1430 | } | |
1431 | ||
b128c09f BB |
1432 | return (error); |
1433 | } | |
1434 | ||
34dc7c2f BB |
1435 | /* |
1436 | * ========================================================================== | |
1437 | * Miscellaneous functions | |
1438 | * ========================================================================== | |
1439 | */ | |
1440 | ||
9ae529ec | 1441 | void |
b0bc7a84 | 1442 | spa_activate_mos_feature(spa_t *spa, const char *feature, dmu_tx_t *tx) |
9ae529ec | 1443 | { |
fa86b5db MA |
1444 | if (!nvlist_exists(spa->spa_label_features, feature)) { |
1445 | fnvlist_add_boolean(spa->spa_label_features, feature); | |
b0bc7a84 MG |
1446 | /* |
1447 | * When we are creating the pool (tx_txg==TXG_INITIAL), we can't | |
1448 | * dirty the vdev config because lock SCL_CONFIG is not held. | |
1449 | * Thankfully, in this case we don't need to dirty the config | |
1450 | * because it will be written out anyway when we finish | |
1451 | * creating the pool. | |
1452 | */ | |
1453 | if (tx->tx_txg != TXG_INITIAL) | |
1454 | vdev_config_dirty(spa->spa_root_vdev); | |
fa86b5db | 1455 | } |
9ae529ec CS |
1456 | } |
1457 | ||
1458 | void | |
1459 | spa_deactivate_mos_feature(spa_t *spa, const char *feature) | |
1460 | { | |
fa86b5db MA |
1461 | if (nvlist_remove_all(spa->spa_label_features, feature) == 0) |
1462 | vdev_config_dirty(spa->spa_root_vdev); | |
9ae529ec CS |
1463 | } |
1464 | ||
34dc7c2f | 1465 | /* |
572e2857 BB |
1466 | * Return the spa_t associated with given pool_guid, if it exists. If |
1467 | * device_guid is non-zero, determine whether the pool exists *and* contains | |
1468 | * a device with the specified device_guid. | |
34dc7c2f | 1469 | */ |
572e2857 BB |
1470 | spa_t * |
1471 | spa_by_guid(uint64_t pool_guid, uint64_t device_guid) | |
34dc7c2f BB |
1472 | { |
1473 | spa_t *spa; | |
1474 | avl_tree_t *t = &spa_namespace_avl; | |
1475 | ||
1476 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
1477 | ||
1478 | for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { | |
1479 | if (spa->spa_state == POOL_STATE_UNINITIALIZED) | |
1480 | continue; | |
1481 | if (spa->spa_root_vdev == NULL) | |
1482 | continue; | |
1483 | if (spa_guid(spa) == pool_guid) { | |
1484 | if (device_guid == 0) | |
1485 | break; | |
1486 | ||
1487 | if (vdev_lookup_by_guid(spa->spa_root_vdev, | |
1488 | device_guid) != NULL) | |
1489 | break; | |
1490 | ||
1491 | /* | |
1492 | * Check any devices we may be in the process of adding. | |
1493 | */ | |
1494 | if (spa->spa_pending_vdev) { | |
1495 | if (vdev_lookup_by_guid(spa->spa_pending_vdev, | |
1496 | device_guid) != NULL) | |
1497 | break; | |
1498 | } | |
1499 | } | |
1500 | } | |
1501 | ||
572e2857 BB |
1502 | return (spa); |
1503 | } | |
1504 | ||
1505 | /* | |
1506 | * Determine whether a pool with the given pool_guid exists. | |
1507 | */ | |
1508 | boolean_t | |
1509 | spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) | |
1510 | { | |
1511 | return (spa_by_guid(pool_guid, device_guid) != NULL); | |
34dc7c2f BB |
1512 | } |
1513 | ||
1514 | char * | |
1515 | spa_strdup(const char *s) | |
1516 | { | |
1517 | size_t len; | |
1518 | char *new; | |
1519 | ||
1520 | len = strlen(s); | |
79c76d5b | 1521 | new = kmem_alloc(len + 1, KM_SLEEP); |
861166b0 | 1522 | memcpy(new, s, len + 1); |
34dc7c2f BB |
1523 | |
1524 | return (new); | |
1525 | } | |
1526 | ||
1527 | void | |
1528 | spa_strfree(char *s) | |
1529 | { | |
1530 | kmem_free(s, strlen(s) + 1); | |
1531 | } | |
1532 | ||
428870ff BB |
1533 | uint64_t |
1534 | spa_generate_guid(spa_t *spa) | |
34dc7c2f | 1535 | { |
29274c9f | 1536 | uint64_t guid; |
34dc7c2f | 1537 | |
428870ff | 1538 | if (spa != NULL) { |
29274c9f AM |
1539 | do { |
1540 | (void) random_get_pseudo_bytes((void *)&guid, | |
1541 | sizeof (guid)); | |
1542 | } while (guid == 0 || spa_guid_exists(spa_guid(spa), guid)); | |
428870ff | 1543 | } else { |
29274c9f AM |
1544 | do { |
1545 | (void) random_get_pseudo_bytes((void *)&guid, | |
1546 | sizeof (guid)); | |
1547 | } while (guid == 0 || spa_guid_exists(guid, 0)); | |
34dc7c2f BB |
1548 | } |
1549 | ||
428870ff BB |
1550 | return (guid); |
1551 | } | |
1552 | ||
1553 | void | |
b0bc7a84 | 1554 | snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp) |
428870ff | 1555 | { |
9ae529ec | 1556 | char type[256]; |
a926aab9 AZ |
1557 | const char *checksum = NULL; |
1558 | const char *compress = NULL; | |
34dc7c2f | 1559 | |
428870ff | 1560 | if (bp != NULL) { |
9ae529ec CS |
1561 | if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) { |
1562 | dmu_object_byteswap_t bswap = | |
1563 | DMU_OT_BYTESWAP(BP_GET_TYPE(bp)); | |
1564 | (void) snprintf(type, sizeof (type), "bswap %s %s", | |
1565 | DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ? | |
1566 | "metadata" : "data", | |
1567 | dmu_ot_byteswap[bswap].ob_name); | |
1568 | } else { | |
1569 | (void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name, | |
1570 | sizeof (type)); | |
1571 | } | |
9b67f605 MA |
1572 | if (!BP_IS_EMBEDDED(bp)) { |
1573 | checksum = | |
1574 | zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name; | |
1575 | } | |
428870ff | 1576 | compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name; |
34dc7c2f BB |
1577 | } |
1578 | ||
97143b9d | 1579 | SNPRINTF_BLKPTR(kmem_scnprintf, ' ', buf, buflen, bp, type, checksum, |
5c27ec10 | 1580 | compress); |
34dc7c2f BB |
1581 | } |
1582 | ||
1583 | void | |
1584 | spa_freeze(spa_t *spa) | |
1585 | { | |
1586 | uint64_t freeze_txg = 0; | |
1587 | ||
b128c09f | 1588 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f BB |
1589 | if (spa->spa_freeze_txg == UINT64_MAX) { |
1590 | freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; | |
1591 | spa->spa_freeze_txg = freeze_txg; | |
1592 | } | |
b128c09f | 1593 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
1594 | if (freeze_txg != 0) |
1595 | txg_wait_synced(spa_get_dsl(spa), freeze_txg); | |
1596 | } | |
1597 | ||
0b39b9f9 PS |
1598 | void |
1599 | zfs_panic_recover(const char *fmt, ...) | |
1600 | { | |
1601 | va_list adx; | |
1602 | ||
1603 | va_start(adx, fmt); | |
1604 | vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); | |
1605 | va_end(adx); | |
1606 | } | |
1607 | ||
428870ff BB |
1608 | /* |
1609 | * This is a stripped-down version of strtoull, suitable only for converting | |
d3cc8b15 | 1610 | * lowercase hexadecimal numbers that don't overflow. |
428870ff BB |
1611 | */ |
1612 | uint64_t | |
e19572e4 | 1613 | zfs_strtonum(const char *str, char **nptr) |
428870ff BB |
1614 | { |
1615 | uint64_t val = 0; | |
1616 | char c; | |
1617 | int digit; | |
1618 | ||
1619 | while ((c = *str) != '\0') { | |
1620 | if (c >= '0' && c <= '9') | |
1621 | digit = c - '0'; | |
1622 | else if (c >= 'a' && c <= 'f') | |
1623 | digit = 10 + c - 'a'; | |
1624 | else | |
1625 | break; | |
1626 | ||
1627 | val *= 16; | |
1628 | val += digit; | |
1629 | ||
1630 | str++; | |
1631 | } | |
1632 | ||
1633 | if (nptr) | |
1634 | *nptr = (char *)str; | |
1635 | ||
1636 | return (val); | |
1637 | } | |
1638 | ||
cc99f275 DB |
1639 | void |
1640 | spa_activate_allocation_classes(spa_t *spa, dmu_tx_t *tx) | |
1641 | { | |
1642 | /* | |
1643 | * We bump the feature refcount for each special vdev added to the pool | |
1644 | */ | |
1645 | ASSERT(spa_feature_is_enabled(spa, SPA_FEATURE_ALLOCATION_CLASSES)); | |
1646 | spa_feature_incr(spa, SPA_FEATURE_ALLOCATION_CLASSES, tx); | |
1647 | } | |
1648 | ||
34dc7c2f BB |
1649 | /* |
1650 | * ========================================================================== | |
1651 | * Accessor functions | |
1652 | * ========================================================================== | |
1653 | */ | |
1654 | ||
b128c09f BB |
1655 | boolean_t |
1656 | spa_shutting_down(spa_t *spa) | |
34dc7c2f | 1657 | { |
b128c09f | 1658 | return (spa->spa_async_suspended); |
34dc7c2f BB |
1659 | } |
1660 | ||
1661 | dsl_pool_t * | |
1662 | spa_get_dsl(spa_t *spa) | |
1663 | { | |
1664 | return (spa->spa_dsl_pool); | |
1665 | } | |
1666 | ||
9ae529ec CS |
1667 | boolean_t |
1668 | spa_is_initializing(spa_t *spa) | |
1669 | { | |
1670 | return (spa->spa_is_initializing); | |
1671 | } | |
1672 | ||
a1d477c2 MA |
1673 | boolean_t |
1674 | spa_indirect_vdevs_loaded(spa_t *spa) | |
1675 | { | |
1676 | return (spa->spa_indirect_vdevs_loaded); | |
1677 | } | |
1678 | ||
34dc7c2f BB |
1679 | blkptr_t * |
1680 | spa_get_rootblkptr(spa_t *spa) | |
1681 | { | |
1682 | return (&spa->spa_ubsync.ub_rootbp); | |
1683 | } | |
1684 | ||
1685 | void | |
1686 | spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) | |
1687 | { | |
1688 | spa->spa_uberblock.ub_rootbp = *bp; | |
1689 | } | |
1690 | ||
1691 | void | |
1692 | spa_altroot(spa_t *spa, char *buf, size_t buflen) | |
1693 | { | |
1694 | if (spa->spa_root == NULL) | |
1695 | buf[0] = '\0'; | |
1696 | else | |
7584fbe8 | 1697 | (void) strlcpy(buf, spa->spa_root, buflen); |
34dc7c2f BB |
1698 | } |
1699 | ||
fdc2d303 | 1700 | uint32_t |
34dc7c2f BB |
1701 | spa_sync_pass(spa_t *spa) |
1702 | { | |
1703 | return (spa->spa_sync_pass); | |
1704 | } | |
1705 | ||
1706 | char * | |
1707 | spa_name(spa_t *spa) | |
1708 | { | |
34dc7c2f BB |
1709 | return (spa->spa_name); |
1710 | } | |
1711 | ||
1712 | uint64_t | |
1713 | spa_guid(spa_t *spa) | |
1714 | { | |
3bc7e0fb GW |
1715 | dsl_pool_t *dp = spa_get_dsl(spa); |
1716 | uint64_t guid; | |
1717 | ||
34dc7c2f BB |
1718 | /* |
1719 | * If we fail to parse the config during spa_load(), we can go through | |
1720 | * the error path (which posts an ereport) and end up here with no root | |
3541dc6d | 1721 | * vdev. We stash the original pool guid in 'spa_config_guid' to handle |
34dc7c2f BB |
1722 | * this case. |
1723 | */ | |
3bc7e0fb GW |
1724 | if (spa->spa_root_vdev == NULL) |
1725 | return (spa->spa_config_guid); | |
1726 | ||
1727 | guid = spa->spa_last_synced_guid != 0 ? | |
1728 | spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid; | |
1729 | ||
1730 | /* | |
1731 | * Return the most recently synced out guid unless we're | |
1732 | * in syncing context. | |
1733 | */ | |
1734 | if (dp && dsl_pool_sync_context(dp)) | |
34dc7c2f BB |
1735 | return (spa->spa_root_vdev->vdev_guid); |
1736 | else | |
3bc7e0fb | 1737 | return (guid); |
3541dc6d GA |
1738 | } |
1739 | ||
1740 | uint64_t | |
1741 | spa_load_guid(spa_t *spa) | |
1742 | { | |
1743 | /* | |
1744 | * This is a GUID that exists solely as a reference for the | |
1745 | * purposes of the arc. It is generated at load time, and | |
1746 | * is never written to persistent storage. | |
1747 | */ | |
1748 | return (spa->spa_load_guid); | |
34dc7c2f BB |
1749 | } |
1750 | ||
1751 | uint64_t | |
1752 | spa_last_synced_txg(spa_t *spa) | |
1753 | { | |
1754 | return (spa->spa_ubsync.ub_txg); | |
1755 | } | |
1756 | ||
1757 | uint64_t | |
1758 | spa_first_txg(spa_t *spa) | |
1759 | { | |
1760 | return (spa->spa_first_txg); | |
1761 | } | |
1762 | ||
428870ff BB |
1763 | uint64_t |
1764 | spa_syncing_txg(spa_t *spa) | |
1765 | { | |
1766 | return (spa->spa_syncing_txg); | |
1767 | } | |
1768 | ||
3b7f360c GW |
1769 | /* |
1770 | * Return the last txg where data can be dirtied. The final txgs | |
1771 | * will be used to just clear out any deferred frees that remain. | |
1772 | */ | |
1773 | uint64_t | |
1774 | spa_final_dirty_txg(spa_t *spa) | |
1775 | { | |
1776 | return (spa->spa_final_txg - TXG_DEFER_SIZE); | |
1777 | } | |
1778 | ||
b128c09f | 1779 | pool_state_t |
34dc7c2f BB |
1780 | spa_state(spa_t *spa) |
1781 | { | |
1782 | return (spa->spa_state); | |
1783 | } | |
1784 | ||
428870ff BB |
1785 | spa_load_state_t |
1786 | spa_load_state(spa_t *spa) | |
34dc7c2f | 1787 | { |
428870ff | 1788 | return (spa->spa_load_state); |
34dc7c2f BB |
1789 | } |
1790 | ||
34dc7c2f | 1791 | uint64_t |
428870ff | 1792 | spa_freeze_txg(spa_t *spa) |
34dc7c2f | 1793 | { |
428870ff | 1794 | return (spa->spa_freeze_txg); |
34dc7c2f BB |
1795 | } |
1796 | ||
047187c1 | 1797 | /* |
1798 | * Return the inflated asize for a logical write in bytes. This is used by the | |
1799 | * DMU to calculate the space a logical write will require on disk. | |
1800 | * If lsize is smaller than the largest physical block size allocatable on this | |
1801 | * pool we use its value instead, since the write will end up using the whole | |
1802 | * block anyway. | |
1803 | */ | |
34dc7c2f | 1804 | uint64_t |
3ec3bc21 | 1805 | spa_get_worst_case_asize(spa_t *spa, uint64_t lsize) |
34dc7c2f | 1806 | { |
047187c1 | 1807 | if (lsize == 0) |
1808 | return (0); /* No inflation needed */ | |
1809 | return (MAX(lsize, 1 << spa->spa_max_ashift) * spa_asize_inflation); | |
34dc7c2f BB |
1810 | } |
1811 | ||
3d45fdd6 | 1812 | /* |
aa755b35 MA |
1813 | * Return the amount of slop space in bytes. It is typically 1/32 of the pool |
1814 | * (3.2%), minus the embedded log space. On very small pools, it may be | |
f01eaed4 PS |
1815 | * slightly larger than this. On very large pools, it will be capped to |
1816 | * the value of spa_max_slop. The embedded log space is not included in | |
aa755b35 MA |
1817 | * spa_dspace. By subtracting it, the usable space (per "zfs list") is a |
1818 | * constant 97% of the total space, regardless of metaslab size (assuming the | |
1819 | * default spa_slop_shift=5 and a non-tiny pool). | |
1820 | * | |
1821 | * See the comment above spa_slop_shift for more details. | |
3d45fdd6 MA |
1822 | */ |
1823 | uint64_t | |
4ea3f864 GM |
1824 | spa_get_slop_space(spa_t *spa) |
1825 | { | |
1325434b RE |
1826 | uint64_t space = 0; |
1827 | uint64_t slop = 0; | |
1828 | ||
1829 | /* | |
1830 | * Make sure spa_dedup_dspace has been set. | |
1831 | */ | |
1832 | if (spa->spa_dedup_dspace == ~0ULL) | |
1833 | spa_update_dspace(spa); | |
1834 | ||
1835 | /* | |
1836 | * spa_get_dspace() includes the space only logically "used" by | |
1837 | * deduplicated data, so since it's not useful to reserve more | |
1838 | * space with more deduplicated data, we subtract that out here. | |
1839 | */ | |
6cc93ccd B |
1840 | space = |
1841 | spa_get_dspace(spa) - spa->spa_dedup_dspace - brt_get_dspace(spa); | |
1325434b | 1842 | slop = MIN(space >> spa_slop_shift, spa_max_slop); |
aa755b35 MA |
1843 | |
1844 | /* | |
1845 | * Subtract the embedded log space, but no more than half the (3.2%) | |
1846 | * unusable space. Note, the "no more than half" is only relevant if | |
1847 | * zfs_embedded_slog_min_ms >> spa_slop_shift < 2, which is not true by | |
1848 | * default. | |
1849 | */ | |
1850 | uint64_t embedded_log = | |
1851 | metaslab_class_get_dspace(spa_embedded_log_class(spa)); | |
1852 | slop -= MIN(embedded_log, slop >> 1); | |
1853 | ||
1854 | /* | |
1855 | * Slop space should be at least spa_min_slop, but no more than half | |
1856 | * the entire pool. | |
1857 | */ | |
1858 | slop = MAX(slop, MIN(space >> 1, spa_min_slop)); | |
1859 | return (slop); | |
3d45fdd6 MA |
1860 | } |
1861 | ||
34dc7c2f BB |
1862 | uint64_t |
1863 | spa_get_dspace(spa_t *spa) | |
1864 | { | |
428870ff | 1865 | return (spa->spa_dspace); |
34dc7c2f BB |
1866 | } |
1867 | ||
d2734cce SD |
1868 | uint64_t |
1869 | spa_get_checkpoint_space(spa_t *spa) | |
1870 | { | |
1871 | return (spa->spa_checkpoint_info.sci_dspace); | |
1872 | } | |
1873 | ||
428870ff BB |
1874 | void |
1875 | spa_update_dspace(spa_t *spa) | |
34dc7c2f | 1876 | { |
428870ff | 1877 | spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) + |
67a1b037 | 1878 | ddt_get_dedup_dspace(spa) + brt_get_dspace(spa); |
2a673e76 | 1879 | if (spa->spa_nonallocating_dspace > 0) { |
a1d477c2 | 1880 | /* |
2a673e76 AJ |
1881 | * Subtract the space provided by all non-allocating vdevs that |
1882 | * contribute to dspace. If a file is overwritten, its old | |
1883 | * blocks are freed and new blocks are allocated. If there are | |
1884 | * no snapshots of the file, the available space should remain | |
1885 | * the same. The old blocks could be freed from the | |
1886 | * non-allocating vdev, but the new blocks must be allocated on | |
1887 | * other (allocating) vdevs. By reserving the entire size of | |
1888 | * the non-allocating vdevs (including allocated space), we | |
1889 | * ensure that there will be enough space on the allocating | |
1890 | * vdevs for this file overwrite to succeed. | |
a1d477c2 MA |
1891 | * |
1892 | * Note that the DMU/DSL doesn't actually know or care | |
1893 | * how much space is allocated (it does its own tracking | |
1894 | * of how much space has been logically used). So it | |
1895 | * doesn't matter that the data we are moving may be | |
2a673e76 | 1896 | * allocated twice (on the old device and the new device). |
a1d477c2 | 1897 | */ |
2a673e76 AJ |
1898 | ASSERT3U(spa->spa_dspace, >=, spa->spa_nonallocating_dspace); |
1899 | spa->spa_dspace -= spa->spa_nonallocating_dspace; | |
a1d477c2 | 1900 | } |
34dc7c2f BB |
1901 | } |
1902 | ||
1903 | /* | |
1904 | * Return the failure mode that has been set to this pool. The default | |
1905 | * behavior will be to block all I/Os when a complete failure occurs. | |
1906 | */ | |
8fb1ede1 | 1907 | uint64_t |
34dc7c2f BB |
1908 | spa_get_failmode(spa_t *spa) |
1909 | { | |
1910 | return (spa->spa_failmode); | |
1911 | } | |
1912 | ||
b128c09f BB |
1913 | boolean_t |
1914 | spa_suspended(spa_t *spa) | |
1915 | { | |
cec3a0a1 | 1916 | return (spa->spa_suspended != ZIO_SUSPEND_NONE); |
b128c09f BB |
1917 | } |
1918 | ||
34dc7c2f BB |
1919 | uint64_t |
1920 | spa_version(spa_t *spa) | |
1921 | { | |
1922 | return (spa->spa_ubsync.ub_version); | |
1923 | } | |
1924 | ||
428870ff BB |
1925 | boolean_t |
1926 | spa_deflate(spa_t *spa) | |
1927 | { | |
1928 | return (spa->spa_deflate); | |
1929 | } | |
1930 | ||
1931 | metaslab_class_t * | |
1932 | spa_normal_class(spa_t *spa) | |
1933 | { | |
1934 | return (spa->spa_normal_class); | |
1935 | } | |
1936 | ||
1937 | metaslab_class_t * | |
1938 | spa_log_class(spa_t *spa) | |
1939 | { | |
1940 | return (spa->spa_log_class); | |
1941 | } | |
1942 | ||
aa755b35 MA |
1943 | metaslab_class_t * |
1944 | spa_embedded_log_class(spa_t *spa) | |
1945 | { | |
1946 | return (spa->spa_embedded_log_class); | |
1947 | } | |
1948 | ||
cc99f275 DB |
1949 | metaslab_class_t * |
1950 | spa_special_class(spa_t *spa) | |
1951 | { | |
1952 | return (spa->spa_special_class); | |
1953 | } | |
1954 | ||
1955 | metaslab_class_t * | |
1956 | spa_dedup_class(spa_t *spa) | |
1957 | { | |
1958 | return (spa->spa_dedup_class); | |
1959 | } | |
1960 | ||
1961 | /* | |
1962 | * Locate an appropriate allocation class | |
1963 | */ | |
1964 | metaslab_class_t * | |
1965 | spa_preferred_class(spa_t *spa, uint64_t size, dmu_object_type_t objtype, | |
1966 | uint_t level, uint_t special_smallblk) | |
1967 | { | |
aa755b35 MA |
1968 | /* |
1969 | * ZIL allocations determine their class in zio_alloc_zil(). | |
1970 | */ | |
1971 | ASSERT(objtype != DMU_OT_INTENT_LOG); | |
cc99f275 DB |
1972 | |
1973 | boolean_t has_special_class = spa->spa_special_class->mc_groups != 0; | |
1974 | ||
1975 | if (DMU_OT_IS_DDT(objtype)) { | |
1976 | if (spa->spa_dedup_class->mc_groups != 0) | |
1977 | return (spa_dedup_class(spa)); | |
1978 | else if (has_special_class && zfs_ddt_data_is_special) | |
1979 | return (spa_special_class(spa)); | |
1980 | else | |
1981 | return (spa_normal_class(spa)); | |
1982 | } | |
1983 | ||
1984 | /* Indirect blocks for user data can land in special if allowed */ | |
1985 | if (level > 0 && (DMU_OT_IS_FILE(objtype) || objtype == DMU_OT_ZVOL)) { | |
1986 | if (has_special_class && zfs_user_indirect_is_special) | |
1987 | return (spa_special_class(spa)); | |
1988 | else | |
1989 | return (spa_normal_class(spa)); | |
1990 | } | |
1991 | ||
1992 | if (DMU_OT_IS_METADATA(objtype) || level > 0) { | |
1993 | if (has_special_class) | |
1994 | return (spa_special_class(spa)); | |
1995 | else | |
1996 | return (spa_normal_class(spa)); | |
1997 | } | |
1998 | ||
1999 | /* | |
2000 | * Allow small file blocks in special class in some cases (like | |
2001 | * for the dRAID vdev feature). But always leave a reserve of | |
2002 | * zfs_special_class_metadata_reserve_pct exclusively for metadata. | |
2003 | */ | |
2004 | if (DMU_OT_IS_FILE(objtype) && | |
44170969 | 2005 | has_special_class && size <= special_smallblk) { |
cc99f275 DB |
2006 | metaslab_class_t *special = spa_special_class(spa); |
2007 | uint64_t alloc = metaslab_class_get_alloc(special); | |
2008 | uint64_t space = metaslab_class_get_space(special); | |
2009 | uint64_t limit = | |
2010 | (space * (100 - zfs_special_class_metadata_reserve_pct)) | |
2011 | / 100; | |
2012 | ||
2013 | if (alloc < limit) | |
2014 | return (special); | |
2015 | } | |
2016 | ||
2017 | return (spa_normal_class(spa)); | |
2018 | } | |
2019 | ||
0c66c32d JG |
2020 | void |
2021 | spa_evicting_os_register(spa_t *spa, objset_t *os) | |
2022 | { | |
2023 | mutex_enter(&spa->spa_evicting_os_lock); | |
2024 | list_insert_head(&spa->spa_evicting_os_list, os); | |
2025 | mutex_exit(&spa->spa_evicting_os_lock); | |
2026 | } | |
2027 | ||
2028 | void | |
2029 | spa_evicting_os_deregister(spa_t *spa, objset_t *os) | |
2030 | { | |
2031 | mutex_enter(&spa->spa_evicting_os_lock); | |
2032 | list_remove(&spa->spa_evicting_os_list, os); | |
2033 | cv_broadcast(&spa->spa_evicting_os_cv); | |
2034 | mutex_exit(&spa->spa_evicting_os_lock); | |
2035 | } | |
2036 | ||
2037 | void | |
2038 | spa_evicting_os_wait(spa_t *spa) | |
2039 | { | |
2040 | mutex_enter(&spa->spa_evicting_os_lock); | |
2041 | while (!list_is_empty(&spa->spa_evicting_os_list)) | |
2042 | cv_wait(&spa->spa_evicting_os_cv, &spa->spa_evicting_os_lock); | |
2043 | mutex_exit(&spa->spa_evicting_os_lock); | |
2044 | ||
2045 | dmu_buf_user_evict_wait(); | |
2046 | } | |
2047 | ||
34dc7c2f BB |
2048 | int |
2049 | spa_max_replication(spa_t *spa) | |
2050 | { | |
2051 | /* | |
2052 | * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to | |
2053 | * handle BPs with more than one DVA allocated. Set our max | |
2054 | * replication level accordingly. | |
2055 | */ | |
2056 | if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) | |
2057 | return (1); | |
2058 | return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); | |
2059 | } | |
2060 | ||
428870ff BB |
2061 | int |
2062 | spa_prev_software_version(spa_t *spa) | |
2063 | { | |
2064 | return (spa->spa_prev_software_version); | |
2065 | } | |
2066 | ||
cc92e9d0 GW |
2067 | uint64_t |
2068 | spa_deadman_synctime(spa_t *spa) | |
2069 | { | |
2070 | return (spa->spa_deadman_synctime); | |
2071 | } | |
2072 | ||
1b939560 BB |
2073 | spa_autotrim_t |
2074 | spa_get_autotrim(spa_t *spa) | |
2075 | { | |
2076 | return (spa->spa_autotrim); | |
2077 | } | |
2078 | ||
8fb1ede1 BB |
2079 | uint64_t |
2080 | spa_deadman_ziotime(spa_t *spa) | |
2081 | { | |
2082 | return (spa->spa_deadman_ziotime); | |
2083 | } | |
2084 | ||
2085 | uint64_t | |
2086 | spa_get_deadman_failmode(spa_t *spa) | |
2087 | { | |
2088 | return (spa->spa_deadman_failmode); | |
2089 | } | |
2090 | ||
2091 | void | |
2092 | spa_set_deadman_failmode(spa_t *spa, const char *failmode) | |
2093 | { | |
2094 | if (strcmp(failmode, "wait") == 0) | |
2095 | spa->spa_deadman_failmode = ZIO_FAILURE_MODE_WAIT; | |
2096 | else if (strcmp(failmode, "continue") == 0) | |
2097 | spa->spa_deadman_failmode = ZIO_FAILURE_MODE_CONTINUE; | |
2098 | else if (strcmp(failmode, "panic") == 0) | |
2099 | spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC; | |
2100 | else | |
2101 | spa->spa_deadman_failmode = ZIO_FAILURE_MODE_WAIT; | |
2102 | } | |
2103 | ||
57940b43 RM |
2104 | void |
2105 | spa_set_deadman_ziotime(hrtime_t ns) | |
2106 | { | |
2107 | spa_t *spa = NULL; | |
2108 | ||
2109 | if (spa_mode_global != SPA_MODE_UNINIT) { | |
2110 | mutex_enter(&spa_namespace_lock); | |
2111 | while ((spa = spa_next(spa)) != NULL) | |
2112 | spa->spa_deadman_ziotime = ns; | |
2113 | mutex_exit(&spa_namespace_lock); | |
2114 | } | |
2115 | } | |
2116 | ||
2117 | void | |
2118 | spa_set_deadman_synctime(hrtime_t ns) | |
2119 | { | |
2120 | spa_t *spa = NULL; | |
2121 | ||
2122 | if (spa_mode_global != SPA_MODE_UNINIT) { | |
2123 | mutex_enter(&spa_namespace_lock); | |
2124 | while ((spa = spa_next(spa)) != NULL) | |
2125 | spa->spa_deadman_synctime = ns; | |
2126 | mutex_exit(&spa_namespace_lock); | |
2127 | } | |
2128 | } | |
2129 | ||
34dc7c2f | 2130 | uint64_t |
428870ff | 2131 | dva_get_dsize_sync(spa_t *spa, const dva_t *dva) |
34dc7c2f | 2132 | { |
428870ff BB |
2133 | uint64_t asize = DVA_GET_ASIZE(dva); |
2134 | uint64_t dsize = asize; | |
34dc7c2f | 2135 | |
428870ff | 2136 | ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); |
34dc7c2f | 2137 | |
428870ff BB |
2138 | if (asize != 0 && spa->spa_deflate) { |
2139 | vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva)); | |
2c33b912 BB |
2140 | if (vd != NULL) |
2141 | dsize = (asize >> SPA_MINBLOCKSHIFT) * | |
2142 | vd->vdev_deflate_ratio; | |
34dc7c2f | 2143 | } |
428870ff BB |
2144 | |
2145 | return (dsize); | |
2146 | } | |
2147 | ||
2148 | uint64_t | |
2149 | bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp) | |
2150 | { | |
2151 | uint64_t dsize = 0; | |
2152 | ||
1c27024e | 2153 | for (int d = 0; d < BP_GET_NDVAS(bp); d++) |
428870ff BB |
2154 | dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); |
2155 | ||
2156 | return (dsize); | |
2157 | } | |
2158 | ||
2159 | uint64_t | |
2160 | bp_get_dsize(spa_t *spa, const blkptr_t *bp) | |
2161 | { | |
2162 | uint64_t dsize = 0; | |
2163 | ||
2164 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2165 | ||
1c27024e | 2166 | for (int d = 0; d < BP_GET_NDVAS(bp); d++) |
428870ff BB |
2167 | dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); |
2168 | ||
b128c09f | 2169 | spa_config_exit(spa, SCL_VDEV, FTAG); |
428870ff BB |
2170 | |
2171 | return (dsize); | |
34dc7c2f BB |
2172 | } |
2173 | ||
dae3e9ea DB |
2174 | uint64_t |
2175 | spa_dirty_data(spa_t *spa) | |
2176 | { | |
2177 | return (spa->spa_dsl_pool->dp_dirty_total); | |
2178 | } | |
2179 | ||
ca95f70d OF |
2180 | /* |
2181 | * ========================================================================== | |
2182 | * SPA Import Progress Routines | |
2183 | * ========================================================================== | |
2184 | */ | |
2185 | ||
2186 | typedef struct spa_import_progress { | |
2187 | uint64_t pool_guid; /* unique id for updates */ | |
2188 | char *pool_name; | |
2189 | spa_load_state_t spa_load_state; | |
687e4d7f | 2190 | char *spa_load_notes; |
ca95f70d OF |
2191 | uint64_t mmp_sec_remaining; /* MMP activity check */ |
2192 | uint64_t spa_load_max_txg; /* rewind txg */ | |
2193 | procfs_list_node_t smh_node; | |
2194 | } spa_import_progress_t; | |
2195 | ||
2196 | spa_history_list_t *spa_import_progress_list = NULL; | |
2197 | ||
2198 | static int | |
2199 | spa_import_progress_show_header(struct seq_file *f) | |
2200 | { | |
687e4d7f | 2201 | seq_printf(f, "%-20s %-14s %-14s %-12s %-16s %s\n", "pool_guid", |
ca95f70d | 2202 | "load_state", "multihost_secs", "max_txg", |
687e4d7f | 2203 | "pool_name", "notes"); |
ca95f70d OF |
2204 | return (0); |
2205 | } | |
2206 | ||
2207 | static int | |
2208 | spa_import_progress_show(struct seq_file *f, void *data) | |
2209 | { | |
2210 | spa_import_progress_t *sip = (spa_import_progress_t *)data; | |
2211 | ||
687e4d7f | 2212 | seq_printf(f, "%-20llu %-14llu %-14llu %-12llu %-16s %s\n", |
ca95f70d OF |
2213 | (u_longlong_t)sip->pool_guid, (u_longlong_t)sip->spa_load_state, |
2214 | (u_longlong_t)sip->mmp_sec_remaining, | |
2215 | (u_longlong_t)sip->spa_load_max_txg, | |
687e4d7f DB |
2216 | (sip->pool_name ? sip->pool_name : "-"), |
2217 | (sip->spa_load_notes ? sip->spa_load_notes : "-")); | |
ca95f70d OF |
2218 | |
2219 | return (0); | |
2220 | } | |
2221 | ||
2222 | /* Remove oldest elements from list until there are no more than 'size' left */ | |
2223 | static void | |
2224 | spa_import_progress_truncate(spa_history_list_t *shl, unsigned int size) | |
2225 | { | |
2226 | spa_import_progress_t *sip; | |
2227 | while (shl->size > size) { | |
2228 | sip = list_remove_head(&shl->procfs_list.pl_list); | |
2229 | if (sip->pool_name) | |
2230 | spa_strfree(sip->pool_name); | |
687e4d7f DB |
2231 | if (sip->spa_load_notes) |
2232 | kmem_strfree(sip->spa_load_notes); | |
ca95f70d OF |
2233 | kmem_free(sip, sizeof (spa_import_progress_t)); |
2234 | shl->size--; | |
2235 | } | |
2236 | ||
2237 | IMPLY(size == 0, list_is_empty(&shl->procfs_list.pl_list)); | |
2238 | } | |
2239 | ||
2240 | static void | |
2241 | spa_import_progress_init(void) | |
2242 | { | |
2243 | spa_import_progress_list = kmem_zalloc(sizeof (spa_history_list_t), | |
2244 | KM_SLEEP); | |
2245 | ||
2246 | spa_import_progress_list->size = 0; | |
2247 | ||
2248 | spa_import_progress_list->procfs_list.pl_private = | |
2249 | spa_import_progress_list; | |
2250 | ||
2251 | procfs_list_install("zfs", | |
7b8363d7 | 2252 | NULL, |
ca95f70d OF |
2253 | "import_progress", |
2254 | 0644, | |
2255 | &spa_import_progress_list->procfs_list, | |
2256 | spa_import_progress_show, | |
2257 | spa_import_progress_show_header, | |
2258 | NULL, | |
2259 | offsetof(spa_import_progress_t, smh_node)); | |
2260 | } | |
2261 | ||
2262 | static void | |
2263 | spa_import_progress_destroy(void) | |
2264 | { | |
2265 | spa_history_list_t *shl = spa_import_progress_list; | |
2266 | procfs_list_uninstall(&shl->procfs_list); | |
2267 | spa_import_progress_truncate(shl, 0); | |
ca95f70d | 2268 | procfs_list_destroy(&shl->procfs_list); |
75c09c50 | 2269 | kmem_free(shl, sizeof (spa_history_list_t)); |
ca95f70d OF |
2270 | } |
2271 | ||
2272 | int | |
2273 | spa_import_progress_set_state(uint64_t pool_guid, | |
2274 | spa_load_state_t load_state) | |
2275 | { | |
2276 | spa_history_list_t *shl = spa_import_progress_list; | |
2277 | spa_import_progress_t *sip; | |
2278 | int error = ENOENT; | |
2279 | ||
2280 | if (shl->size == 0) | |
2281 | return (0); | |
2282 | ||
2283 | mutex_enter(&shl->procfs_list.pl_lock); | |
2284 | for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL; | |
2285 | sip = list_prev(&shl->procfs_list.pl_list, sip)) { | |
2286 | if (sip->pool_guid == pool_guid) { | |
2287 | sip->spa_load_state = load_state; | |
687e4d7f DB |
2288 | if (sip->spa_load_notes != NULL) { |
2289 | kmem_strfree(sip->spa_load_notes); | |
2290 | sip->spa_load_notes = NULL; | |
2291 | } | |
ca95f70d OF |
2292 | error = 0; |
2293 | break; | |
2294 | } | |
2295 | } | |
2296 | mutex_exit(&shl->procfs_list.pl_lock); | |
2297 | ||
2298 | return (error); | |
2299 | } | |
2300 | ||
687e4d7f DB |
2301 | static void |
2302 | spa_import_progress_set_notes_impl(spa_t *spa, boolean_t log_dbgmsg, | |
2303 | const char *fmt, va_list adx) | |
2304 | { | |
2305 | spa_history_list_t *shl = spa_import_progress_list; | |
2306 | spa_import_progress_t *sip; | |
2307 | uint64_t pool_guid = spa_guid(spa); | |
2308 | ||
2309 | if (shl->size == 0) | |
2310 | return; | |
2311 | ||
2312 | char *notes = kmem_vasprintf(fmt, adx); | |
2313 | ||
2314 | mutex_enter(&shl->procfs_list.pl_lock); | |
2315 | for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL; | |
2316 | sip = list_prev(&shl->procfs_list.pl_list, sip)) { | |
2317 | if (sip->pool_guid == pool_guid) { | |
2318 | if (sip->spa_load_notes != NULL) { | |
2319 | kmem_strfree(sip->spa_load_notes); | |
2320 | sip->spa_load_notes = NULL; | |
2321 | } | |
2322 | sip->spa_load_notes = notes; | |
2323 | if (log_dbgmsg) | |
2324 | zfs_dbgmsg("'%s' %s", sip->pool_name, notes); | |
2325 | notes = NULL; | |
2326 | break; | |
2327 | } | |
2328 | } | |
2329 | mutex_exit(&shl->procfs_list.pl_lock); | |
2330 | if (notes != NULL) | |
2331 | kmem_strfree(notes); | |
2332 | } | |
2333 | ||
2334 | void | |
2335 | spa_import_progress_set_notes(spa_t *spa, const char *fmt, ...) | |
2336 | { | |
2337 | va_list adx; | |
2338 | ||
2339 | va_start(adx, fmt); | |
2340 | spa_import_progress_set_notes_impl(spa, B_TRUE, fmt, adx); | |
2341 | va_end(adx); | |
2342 | } | |
2343 | ||
2344 | void | |
2345 | spa_import_progress_set_notes_nolog(spa_t *spa, const char *fmt, ...) | |
2346 | { | |
2347 | va_list adx; | |
2348 | ||
2349 | va_start(adx, fmt); | |
2350 | spa_import_progress_set_notes_impl(spa, B_FALSE, fmt, adx); | |
2351 | va_end(adx); | |
2352 | } | |
2353 | ||
ca95f70d OF |
2354 | int |
2355 | spa_import_progress_set_max_txg(uint64_t pool_guid, uint64_t load_max_txg) | |
2356 | { | |
2357 | spa_history_list_t *shl = spa_import_progress_list; | |
2358 | spa_import_progress_t *sip; | |
2359 | int error = ENOENT; | |
2360 | ||
2361 | if (shl->size == 0) | |
2362 | return (0); | |
2363 | ||
2364 | mutex_enter(&shl->procfs_list.pl_lock); | |
2365 | for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL; | |
2366 | sip = list_prev(&shl->procfs_list.pl_list, sip)) { | |
2367 | if (sip->pool_guid == pool_guid) { | |
2368 | sip->spa_load_max_txg = load_max_txg; | |
2369 | error = 0; | |
2370 | break; | |
2371 | } | |
2372 | } | |
2373 | mutex_exit(&shl->procfs_list.pl_lock); | |
2374 | ||
2375 | return (error); | |
2376 | } | |
2377 | ||
2378 | int | |
2379 | spa_import_progress_set_mmp_check(uint64_t pool_guid, | |
2380 | uint64_t mmp_sec_remaining) | |
2381 | { | |
2382 | spa_history_list_t *shl = spa_import_progress_list; | |
2383 | spa_import_progress_t *sip; | |
2384 | int error = ENOENT; | |
2385 | ||
2386 | if (shl->size == 0) | |
2387 | return (0); | |
2388 | ||
2389 | mutex_enter(&shl->procfs_list.pl_lock); | |
2390 | for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL; | |
2391 | sip = list_prev(&shl->procfs_list.pl_list, sip)) { | |
2392 | if (sip->pool_guid == pool_guid) { | |
2393 | sip->mmp_sec_remaining = mmp_sec_remaining; | |
2394 | error = 0; | |
2395 | break; | |
2396 | } | |
2397 | } | |
2398 | mutex_exit(&shl->procfs_list.pl_lock); | |
2399 | ||
2400 | return (error); | |
2401 | } | |
2402 | ||
2403 | /* | |
2404 | * A new import is in progress, add an entry. | |
2405 | */ | |
2406 | void | |
2407 | spa_import_progress_add(spa_t *spa) | |
2408 | { | |
2409 | spa_history_list_t *shl = spa_import_progress_list; | |
2410 | spa_import_progress_t *sip; | |
d1807f16 | 2411 | const char *poolname = NULL; |
ca95f70d OF |
2412 | |
2413 | sip = kmem_zalloc(sizeof (spa_import_progress_t), KM_SLEEP); | |
2414 | sip->pool_guid = spa_guid(spa); | |
2415 | ||
2416 | (void) nvlist_lookup_string(spa->spa_config, ZPOOL_CONFIG_POOL_NAME, | |
2417 | &poolname); | |
2418 | if (poolname == NULL) | |
2419 | poolname = spa_name(spa); | |
2420 | sip->pool_name = spa_strdup(poolname); | |
2421 | sip->spa_load_state = spa_load_state(spa); | |
687e4d7f | 2422 | sip->spa_load_notes = NULL; |
ca95f70d OF |
2423 | |
2424 | mutex_enter(&shl->procfs_list.pl_lock); | |
2425 | procfs_list_add(&shl->procfs_list, sip); | |
2426 | shl->size++; | |
2427 | mutex_exit(&shl->procfs_list.pl_lock); | |
2428 | } | |
2429 | ||
2430 | void | |
2431 | spa_import_progress_remove(uint64_t pool_guid) | |
2432 | { | |
2433 | spa_history_list_t *shl = spa_import_progress_list; | |
2434 | spa_import_progress_t *sip; | |
2435 | ||
2436 | mutex_enter(&shl->procfs_list.pl_lock); | |
2437 | for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL; | |
2438 | sip = list_prev(&shl->procfs_list.pl_list, sip)) { | |
2439 | if (sip->pool_guid == pool_guid) { | |
2440 | if (sip->pool_name) | |
2441 | spa_strfree(sip->pool_name); | |
687e4d7f DB |
2442 | if (sip->spa_load_notes) |
2443 | spa_strfree(sip->spa_load_notes); | |
ca95f70d OF |
2444 | list_remove(&shl->procfs_list.pl_list, sip); |
2445 | shl->size--; | |
2446 | kmem_free(sip, sizeof (spa_import_progress_t)); | |
2447 | break; | |
2448 | } | |
2449 | } | |
2450 | mutex_exit(&shl->procfs_list.pl_lock); | |
2451 | } | |
2452 | ||
34dc7c2f BB |
2453 | /* |
2454 | * ========================================================================== | |
2455 | * Initialization and Termination | |
2456 | * ========================================================================== | |
2457 | */ | |
2458 | ||
2459 | static int | |
2460 | spa_name_compare(const void *a1, const void *a2) | |
2461 | { | |
2462 | const spa_t *s1 = a1; | |
2463 | const spa_t *s2 = a2; | |
2464 | int s; | |
2465 | ||
2466 | s = strcmp(s1->spa_name, s2->spa_name); | |
ee36c709 | 2467 | |
ca577779 | 2468 | return (TREE_ISIGN(s)); |
34dc7c2f BB |
2469 | } |
2470 | ||
34dc7c2f | 2471 | void |
0bc8fd78 | 2472 | spa_boot_init(void) |
34dc7c2f BB |
2473 | { |
2474 | spa_config_load(); | |
2475 | } | |
2476 | ||
2477 | void | |
da92d5cb | 2478 | spa_init(spa_mode_t mode) |
34dc7c2f BB |
2479 | { |
2480 | mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); | |
2481 | mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); | |
2482 | mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); | |
2483 | cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); | |
2484 | ||
2485 | avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), | |
2486 | offsetof(spa_t, spa_avl)); | |
2487 | ||
2488 | avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), | |
2489 | offsetof(spa_aux_t, aux_avl)); | |
2490 | ||
2491 | avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), | |
2492 | offsetof(spa_aux_t, aux_avl)); | |
2493 | ||
fb5f0bc8 | 2494 | spa_mode_global = mode; |
34dc7c2f | 2495 | |
498877ba | 2496 | #ifndef _KERNEL |
da92d5cb | 2497 | if (spa_mode_global != SPA_MODE_READ && dprintf_find_string("watch")) { |
498877ba MA |
2498 | struct sigaction sa; |
2499 | ||
2500 | sa.sa_flags = SA_SIGINFO; | |
2501 | sigemptyset(&sa.sa_mask); | |
2502 | sa.sa_sigaction = arc_buf_sigsegv; | |
2503 | ||
2504 | if (sigaction(SIGSEGV, &sa, NULL) == -1) { | |
2505 | perror("could not enable watchpoints: " | |
2506 | "sigaction(SIGSEGV, ...) = "); | |
2507 | } else { | |
2508 | arc_watch = B_TRUE; | |
2509 | } | |
2510 | } | |
2511 | #endif | |
2512 | ||
26685276 | 2513 | fm_init(); |
424fd7c3 | 2514 | zfs_refcount_init(); |
34dc7c2f | 2515 | unique_init(); |
ca577779 PD |
2516 | zfs_btree_init(); |
2517 | metaslab_stat_init(); | |
67a1b037 | 2518 | brt_init(); |
ecf3d9b8 | 2519 | ddt_init(); |
34dc7c2f BB |
2520 | zio_init(); |
2521 | dmu_init(); | |
2522 | zil_init(); | |
551905dd | 2523 | vdev_mirror_stat_init(); |
ab9f4b0b | 2524 | vdev_raidz_math_init(); |
da8f51e1 | 2525 | vdev_file_init(); |
34dc7c2f | 2526 | zfs_prop_init(); |
985c33b1 | 2527 | chksum_init(); |
34dc7c2f | 2528 | zpool_prop_init(); |
9ae529ec | 2529 | zpool_feature_init(); |
34dc7c2f | 2530 | spa_config_load(); |
2a673e76 | 2531 | vdev_prop_init(); |
b128c09f | 2532 | l2arc_start(); |
d4a72f23 | 2533 | scan_init(); |
6a9d6359 | 2534 | qat_init(); |
ca95f70d | 2535 | spa_import_progress_init(); |
34dc7c2f BB |
2536 | } |
2537 | ||
2538 | void | |
2539 | spa_fini(void) | |
2540 | { | |
b128c09f BB |
2541 | l2arc_stop(); |
2542 | ||
34dc7c2f BB |
2543 | spa_evict_all(); |
2544 | ||
da8f51e1 | 2545 | vdev_file_fini(); |
551905dd | 2546 | vdev_mirror_stat_fini(); |
ab9f4b0b | 2547 | vdev_raidz_math_fini(); |
985c33b1 | 2548 | chksum_fini(); |
34dc7c2f BB |
2549 | zil_fini(); |
2550 | dmu_fini(); | |
2551 | zio_fini(); | |
ecf3d9b8 | 2552 | ddt_fini(); |
67a1b037 | 2553 | brt_fini(); |
ca577779 PD |
2554 | metaslab_stat_fini(); |
2555 | zfs_btree_fini(); | |
34dc7c2f | 2556 | unique_fini(); |
424fd7c3 | 2557 | zfs_refcount_fini(); |
26685276 | 2558 | fm_fini(); |
d4a72f23 | 2559 | scan_fini(); |
6a9d6359 | 2560 | qat_fini(); |
ca95f70d | 2561 | spa_import_progress_destroy(); |
34dc7c2f BB |
2562 | |
2563 | avl_destroy(&spa_namespace_avl); | |
2564 | avl_destroy(&spa_spare_avl); | |
2565 | avl_destroy(&spa_l2cache_avl); | |
2566 | ||
2567 | cv_destroy(&spa_namespace_cv); | |
2568 | mutex_destroy(&spa_namespace_lock); | |
2569 | mutex_destroy(&spa_spare_lock); | |
2570 | mutex_destroy(&spa_l2cache_lock); | |
2571 | } | |
2572 | ||
2573 | /* | |
aa755b35 | 2574 | * Return whether this pool has a dedicated slog device. No locking needed. |
34dc7c2f | 2575 | * It's not a problem if the wrong answer is returned as it's only for |
aa755b35 | 2576 | * performance and not correctness. |
34dc7c2f BB |
2577 | */ |
2578 | boolean_t | |
2579 | spa_has_slogs(spa_t *spa) | |
2580 | { | |
f8020c93 | 2581 | return (spa->spa_log_class->mc_groups != 0); |
34dc7c2f | 2582 | } |
b128c09f | 2583 | |
428870ff BB |
2584 | spa_log_state_t |
2585 | spa_get_log_state(spa_t *spa) | |
2586 | { | |
2587 | return (spa->spa_log_state); | |
2588 | } | |
2589 | ||
2590 | void | |
2591 | spa_set_log_state(spa_t *spa, spa_log_state_t state) | |
2592 | { | |
2593 | spa->spa_log_state = state; | |
2594 | } | |
2595 | ||
b128c09f BB |
2596 | boolean_t |
2597 | spa_is_root(spa_t *spa) | |
2598 | { | |
2599 | return (spa->spa_is_root); | |
2600 | } | |
fb5f0bc8 BB |
2601 | |
2602 | boolean_t | |
2603 | spa_writeable(spa_t *spa) | |
2604 | { | |
da92d5cb | 2605 | return (!!(spa->spa_mode & SPA_MODE_WRITE) && spa->spa_trust_config); |
fb5f0bc8 BB |
2606 | } |
2607 | ||
acbad6ff AR |
2608 | /* |
2609 | * Returns true if there is a pending sync task in any of the current | |
2610 | * syncing txg, the current quiescing txg, or the current open txg. | |
2611 | */ | |
2612 | boolean_t | |
2613 | spa_has_pending_synctask(spa_t *spa) | |
2614 | { | |
d2734cce SD |
2615 | return (!txg_all_lists_empty(&spa->spa_dsl_pool->dp_sync_tasks) || |
2616 | !txg_all_lists_empty(&spa->spa_dsl_pool->dp_early_sync_tasks)); | |
acbad6ff AR |
2617 | } |
2618 | ||
da92d5cb | 2619 | spa_mode_t |
fb5f0bc8 BB |
2620 | spa_mode(spa_t *spa) |
2621 | { | |
2622 | return (spa->spa_mode); | |
2623 | } | |
428870ff BB |
2624 | |
2625 | uint64_t | |
2626 | spa_bootfs(spa_t *spa) | |
2627 | { | |
2628 | return (spa->spa_bootfs); | |
2629 | } | |
2630 | ||
2631 | uint64_t | |
2632 | spa_delegation(spa_t *spa) | |
2633 | { | |
2634 | return (spa->spa_delegation); | |
2635 | } | |
2636 | ||
2637 | objset_t * | |
2638 | spa_meta_objset(spa_t *spa) | |
2639 | { | |
2640 | return (spa->spa_meta_objset); | |
2641 | } | |
2642 | ||
2643 | enum zio_checksum | |
2644 | spa_dedup_checksum(spa_t *spa) | |
2645 | { | |
2646 | return (spa->spa_dedup_checksum); | |
2647 | } | |
2648 | ||
2649 | /* | |
2650 | * Reset pool scan stat per scan pass (or reboot). | |
2651 | */ | |
2652 | void | |
2653 | spa_scan_stat_init(spa_t *spa) | |
2654 | { | |
2655 | /* data not stored on disk */ | |
2656 | spa->spa_scan_pass_start = gethrestime_sec(); | |
0ea05c64 AP |
2657 | if (dsl_scan_is_paused_scrub(spa->spa_dsl_pool->dp_scan)) |
2658 | spa->spa_scan_pass_scrub_pause = spa->spa_scan_pass_start; | |
2659 | else | |
2660 | spa->spa_scan_pass_scrub_pause = 0; | |
482eeef8 GA |
2661 | |
2662 | if (dsl_errorscrub_is_paused(spa->spa_dsl_pool->dp_scan)) | |
2663 | spa->spa_scan_pass_errorscrub_pause = spa->spa_scan_pass_start; | |
2664 | else | |
2665 | spa->spa_scan_pass_errorscrub_pause = 0; | |
2666 | ||
0ea05c64 | 2667 | spa->spa_scan_pass_scrub_spent_paused = 0; |
428870ff | 2668 | spa->spa_scan_pass_exam = 0; |
d4a72f23 | 2669 | spa->spa_scan_pass_issued = 0; |
482eeef8 GA |
2670 | |
2671 | // error scrub stats | |
2672 | spa->spa_scan_pass_errorscrub_spent_paused = 0; | |
428870ff BB |
2673 | } |
2674 | ||
2675 | /* | |
2676 | * Get scan stats for zpool status reports | |
2677 | */ | |
2678 | int | |
2679 | spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps) | |
2680 | { | |
2681 | dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL; | |
2682 | ||
482eeef8 GA |
2683 | if (scn == NULL || (scn->scn_phys.scn_func == POOL_SCAN_NONE && |
2684 | scn->errorscrub_phys.dep_func == POOL_SCAN_NONE)) | |
2e528b49 | 2685 | return (SET_ERROR(ENOENT)); |
482eeef8 | 2686 | |
861166b0 | 2687 | memset(ps, 0, sizeof (pool_scan_stat_t)); |
428870ff BB |
2688 | |
2689 | /* data stored on disk */ | |
2690 | ps->pss_func = scn->scn_phys.scn_func; | |
d4a72f23 | 2691 | ps->pss_state = scn->scn_phys.scn_state; |
428870ff BB |
2692 | ps->pss_start_time = scn->scn_phys.scn_start_time; |
2693 | ps->pss_end_time = scn->scn_phys.scn_end_time; | |
2694 | ps->pss_to_examine = scn->scn_phys.scn_to_examine; | |
d4677269 | 2695 | ps->pss_examined = scn->scn_phys.scn_examined; |
fa7b2390 | 2696 | ps->pss_skipped = scn->scn_phys.scn_skipped; |
428870ff BB |
2697 | ps->pss_processed = scn->scn_phys.scn_processed; |
2698 | ps->pss_errors = scn->scn_phys.scn_errors; | |
428870ff BB |
2699 | |
2700 | /* data not stored on disk */ | |
428870ff | 2701 | ps->pss_pass_exam = spa->spa_scan_pass_exam; |
d4677269 | 2702 | ps->pss_pass_start = spa->spa_scan_pass_start; |
0ea05c64 AP |
2703 | ps->pss_pass_scrub_pause = spa->spa_scan_pass_scrub_pause; |
2704 | ps->pss_pass_scrub_spent_paused = spa->spa_scan_pass_scrub_spent_paused; | |
d4677269 TC |
2705 | ps->pss_pass_issued = spa->spa_scan_pass_issued; |
2706 | ps->pss_issued = | |
2707 | scn->scn_issued_before_pass + spa->spa_scan_pass_issued; | |
428870ff | 2708 | |
482eeef8 GA |
2709 | /* error scrub data stored on disk */ |
2710 | ps->pss_error_scrub_func = scn->errorscrub_phys.dep_func; | |
2711 | ps->pss_error_scrub_state = scn->errorscrub_phys.dep_state; | |
2712 | ps->pss_error_scrub_start = scn->errorscrub_phys.dep_start_time; | |
2713 | ps->pss_error_scrub_end = scn->errorscrub_phys.dep_end_time; | |
2714 | ps->pss_error_scrub_examined = scn->errorscrub_phys.dep_examined; | |
2715 | ps->pss_error_scrub_to_be_examined = | |
2716 | scn->errorscrub_phys.dep_to_examine; | |
2717 | ||
2718 | /* error scrub data not stored on disk */ | |
2719 | ps->pss_pass_error_scrub_pause = spa->spa_scan_pass_errorscrub_pause; | |
2720 | ||
428870ff BB |
2721 | return (0); |
2722 | } | |
c28b2279 | 2723 | |
f1512ee6 MA |
2724 | int |
2725 | spa_maxblocksize(spa_t *spa) | |
2726 | { | |
2727 | if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) | |
2728 | return (SPA_MAXBLOCKSIZE); | |
2729 | else | |
2730 | return (SPA_OLD_MAXBLOCKSIZE); | |
2731 | } | |
2732 | ||
a1d477c2 MA |
2733 | |
2734 | /* | |
2735 | * Returns the txg that the last device removal completed. No indirect mappings | |
2736 | * have been added since this txg. | |
2737 | */ | |
2738 | uint64_t | |
2739 | spa_get_last_removal_txg(spa_t *spa) | |
2740 | { | |
2741 | uint64_t vdevid; | |
2742 | uint64_t ret = -1ULL; | |
2743 | ||
2744 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2745 | /* | |
2746 | * sr_prev_indirect_vdev is only modified while holding all the | |
2747 | * config locks, so it is sufficient to hold SCL_VDEV as reader when | |
2748 | * examining it. | |
2749 | */ | |
2750 | vdevid = spa->spa_removing_phys.sr_prev_indirect_vdev; | |
2751 | ||
2752 | while (vdevid != -1ULL) { | |
2753 | vdev_t *vd = vdev_lookup_top(spa, vdevid); | |
2754 | vdev_indirect_births_t *vib = vd->vdev_indirect_births; | |
2755 | ||
2756 | ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops); | |
2757 | ||
2758 | /* | |
2759 | * If the removal did not remap any data, we don't care. | |
2760 | */ | |
2761 | if (vdev_indirect_births_count(vib) != 0) { | |
2762 | ret = vdev_indirect_births_last_entry_txg(vib); | |
2763 | break; | |
2764 | } | |
2765 | ||
2766 | vdevid = vd->vdev_indirect_config.vic_prev_indirect_vdev; | |
2767 | } | |
2768 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2769 | ||
2770 | IMPLY(ret != -1ULL, | |
2771 | spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL)); | |
2772 | ||
2773 | return (ret); | |
2774 | } | |
2775 | ||
50c957f7 NB |
2776 | int |
2777 | spa_maxdnodesize(spa_t *spa) | |
2778 | { | |
2779 | if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE)) | |
2780 | return (DNODE_MAX_SIZE); | |
2781 | else | |
2782 | return (DNODE_MIN_SIZE); | |
2783 | } | |
2784 | ||
379ca9cf OF |
2785 | boolean_t |
2786 | spa_multihost(spa_t *spa) | |
2787 | { | |
2788 | return (spa->spa_multihost ? B_TRUE : B_FALSE); | |
2789 | } | |
2790 | ||
25f06d67 BB |
2791 | uint32_t |
2792 | spa_get_hostid(spa_t *spa) | |
379ca9cf | 2793 | { |
25f06d67 | 2794 | return (spa->spa_hostid); |
379ca9cf OF |
2795 | } |
2796 | ||
6cb8e530 PZ |
2797 | boolean_t |
2798 | spa_trust_config(spa_t *spa) | |
2799 | { | |
2800 | return (spa->spa_trust_config); | |
2801 | } | |
2802 | ||
2803 | uint64_t | |
2804 | spa_missing_tvds_allowed(spa_t *spa) | |
2805 | { | |
2806 | return (spa->spa_missing_tvds_allowed); | |
2807 | } | |
2808 | ||
93e28d66 SD |
2809 | space_map_t * |
2810 | spa_syncing_log_sm(spa_t *spa) | |
2811 | { | |
2812 | return (spa->spa_syncing_log_sm); | |
2813 | } | |
2814 | ||
6cb8e530 PZ |
2815 | void |
2816 | spa_set_missing_tvds(spa_t *spa, uint64_t missing) | |
2817 | { | |
2818 | spa->spa_missing_tvds = missing; | |
2819 | } | |
2820 | ||
f0ed6c74 TH |
2821 | /* |
2822 | * Return the pool state string ("ONLINE", "DEGRADED", "SUSPENDED", etc). | |
2823 | */ | |
2824 | const char * | |
2825 | spa_state_to_name(spa_t *spa) | |
2826 | { | |
78fac8d9 RE |
2827 | ASSERT3P(spa, !=, NULL); |
2828 | ||
2829 | /* | |
2830 | * it is possible for the spa to exist, without root vdev | |
2831 | * as the spa transitions during import/export | |
2832 | */ | |
2833 | vdev_t *rvd = spa->spa_root_vdev; | |
2834 | if (rvd == NULL) { | |
2835 | return ("TRANSITIONING"); | |
2836 | } | |
2837 | vdev_state_t state = rvd->vdev_state; | |
2838 | vdev_aux_t aux = rvd->vdev_stat.vs_aux; | |
f0ed6c74 | 2839 | |
4647353c | 2840 | if (spa_suspended(spa)) |
f0ed6c74 TH |
2841 | return ("SUSPENDED"); |
2842 | ||
2843 | switch (state) { | |
2844 | case VDEV_STATE_CLOSED: | |
2845 | case VDEV_STATE_OFFLINE: | |
2846 | return ("OFFLINE"); | |
2847 | case VDEV_STATE_REMOVED: | |
2848 | return ("REMOVED"); | |
2849 | case VDEV_STATE_CANT_OPEN: | |
2850 | if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG) | |
2851 | return ("FAULTED"); | |
2852 | else if (aux == VDEV_AUX_SPLIT_POOL) | |
2853 | return ("SPLIT"); | |
2854 | else | |
2855 | return ("UNAVAIL"); | |
2856 | case VDEV_STATE_FAULTED: | |
2857 | return ("FAULTED"); | |
2858 | case VDEV_STATE_DEGRADED: | |
2859 | return ("DEGRADED"); | |
2860 | case VDEV_STATE_HEALTHY: | |
2861 | return ("ONLINE"); | |
2862 | default: | |
2863 | break; | |
2864 | } | |
2865 | ||
2866 | return ("UNKNOWN"); | |
2867 | } | |
2868 | ||
d2734cce SD |
2869 | boolean_t |
2870 | spa_top_vdevs_spacemap_addressable(spa_t *spa) | |
2871 | { | |
2872 | vdev_t *rvd = spa->spa_root_vdev; | |
2873 | for (uint64_t c = 0; c < rvd->vdev_children; c++) { | |
2874 | if (!vdev_is_spacemap_addressable(rvd->vdev_child[c])) | |
2875 | return (B_FALSE); | |
2876 | } | |
2877 | return (B_TRUE); | |
2878 | } | |
2879 | ||
2880 | boolean_t | |
2881 | spa_has_checkpoint(spa_t *spa) | |
2882 | { | |
2883 | return (spa->spa_checkpoint_txg != 0); | |
2884 | } | |
2885 | ||
2886 | boolean_t | |
2887 | spa_importing_readonly_checkpoint(spa_t *spa) | |
2888 | { | |
2889 | return ((spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT) && | |
da92d5cb | 2890 | spa->spa_mode == SPA_MODE_READ); |
d2734cce SD |
2891 | } |
2892 | ||
2893 | uint64_t | |
2894 | spa_min_claim_txg(spa_t *spa) | |
2895 | { | |
2896 | uint64_t checkpoint_txg = spa->spa_uberblock.ub_checkpoint_txg; | |
2897 | ||
2898 | if (checkpoint_txg != 0) | |
2899 | return (checkpoint_txg + 1); | |
2900 | ||
2901 | return (spa->spa_first_txg); | |
2902 | } | |
2903 | ||
2904 | /* | |
2905 | * If there is a checkpoint, async destroys may consume more space from | |
2906 | * the pool instead of freeing it. In an attempt to save the pool from | |
2907 | * getting suspended when it is about to run out of space, we stop | |
2908 | * processing async destroys. | |
2909 | */ | |
2910 | boolean_t | |
2911 | spa_suspend_async_destroy(spa_t *spa) | |
2912 | { | |
2913 | dsl_pool_t *dp = spa_get_dsl(spa); | |
2914 | ||
2915 | uint64_t unreserved = dsl_pool_unreserved_space(dp, | |
2916 | ZFS_SPACE_CHECK_EXTRA_RESERVED); | |
2917 | uint64_t used = dsl_dir_phys(dp->dp_root_dir)->dd_used_bytes; | |
2918 | uint64_t avail = (unreserved > used) ? (unreserved - used) : 0; | |
2919 | ||
2920 | if (spa_has_checkpoint(spa) && avail == 0) | |
2921 | return (B_TRUE); | |
2922 | ||
2923 | return (B_FALSE); | |
2924 | } | |
2925 | ||
93ce2b4c | 2926 | #if defined(_KERNEL) |
8fb1ede1 | 2927 | |
e64e84ec MM |
2928 | int |
2929 | param_set_deadman_failmode_common(const char *val) | |
8fb1ede1 BB |
2930 | { |
2931 | spa_t *spa = NULL; | |
2932 | char *p; | |
2933 | ||
2934 | if (val == NULL) | |
e64e84ec | 2935 | return (SET_ERROR(EINVAL)); |
8fb1ede1 BB |
2936 | |
2937 | if ((p = strchr(val, '\n')) != NULL) | |
2938 | *p = '\0'; | |
2939 | ||
2940 | if (strcmp(val, "wait") != 0 && strcmp(val, "continue") != 0 && | |
2941 | strcmp(val, "panic")) | |
e64e84ec | 2942 | return (SET_ERROR(EINVAL)); |
8fb1ede1 | 2943 | |
da92d5cb | 2944 | if (spa_mode_global != SPA_MODE_UNINIT) { |
d1043e2f TC |
2945 | mutex_enter(&spa_namespace_lock); |
2946 | while ((spa = spa_next(spa)) != NULL) | |
2947 | spa_set_deadman_failmode(spa, val); | |
2948 | mutex_exit(&spa_namespace_lock); | |
2949 | } | |
8fb1ede1 | 2950 | |
e64e84ec | 2951 | return (0); |
8fb1ede1 | 2952 | } |
03fdcb9a MM |
2953 | #endif |
2954 | ||
c28b2279 BB |
2955 | /* Namespace manipulation */ |
2956 | EXPORT_SYMBOL(spa_lookup); | |
2957 | EXPORT_SYMBOL(spa_add); | |
2958 | EXPORT_SYMBOL(spa_remove); | |
2959 | EXPORT_SYMBOL(spa_next); | |
2960 | ||
2961 | /* Refcount functions */ | |
2962 | EXPORT_SYMBOL(spa_open_ref); | |
2963 | EXPORT_SYMBOL(spa_close); | |
2964 | EXPORT_SYMBOL(spa_refcount_zero); | |
2965 | ||
2966 | /* Pool configuration lock */ | |
2967 | EXPORT_SYMBOL(spa_config_tryenter); | |
2968 | EXPORT_SYMBOL(spa_config_enter); | |
2969 | EXPORT_SYMBOL(spa_config_exit); | |
2970 | EXPORT_SYMBOL(spa_config_held); | |
2971 | ||
2972 | /* Pool vdev add/remove lock */ | |
2973 | EXPORT_SYMBOL(spa_vdev_enter); | |
2974 | EXPORT_SYMBOL(spa_vdev_exit); | |
2975 | ||
2976 | /* Pool vdev state change lock */ | |
2977 | EXPORT_SYMBOL(spa_vdev_state_enter); | |
2978 | EXPORT_SYMBOL(spa_vdev_state_exit); | |
2979 | ||
2980 | /* Accessor functions */ | |
2981 | EXPORT_SYMBOL(spa_shutting_down); | |
2982 | EXPORT_SYMBOL(spa_get_dsl); | |
2983 | EXPORT_SYMBOL(spa_get_rootblkptr); | |
2984 | EXPORT_SYMBOL(spa_set_rootblkptr); | |
2985 | EXPORT_SYMBOL(spa_altroot); | |
2986 | EXPORT_SYMBOL(spa_sync_pass); | |
2987 | EXPORT_SYMBOL(spa_name); | |
2988 | EXPORT_SYMBOL(spa_guid); | |
2989 | EXPORT_SYMBOL(spa_last_synced_txg); | |
2990 | EXPORT_SYMBOL(spa_first_txg); | |
2991 | EXPORT_SYMBOL(spa_syncing_txg); | |
2992 | EXPORT_SYMBOL(spa_version); | |
2993 | EXPORT_SYMBOL(spa_state); | |
2994 | EXPORT_SYMBOL(spa_load_state); | |
2995 | EXPORT_SYMBOL(spa_freeze_txg); | |
c28b2279 BB |
2996 | EXPORT_SYMBOL(spa_get_dspace); |
2997 | EXPORT_SYMBOL(spa_update_dspace); | |
2998 | EXPORT_SYMBOL(spa_deflate); | |
2999 | EXPORT_SYMBOL(spa_normal_class); | |
3000 | EXPORT_SYMBOL(spa_log_class); | |
cc99f275 DB |
3001 | EXPORT_SYMBOL(spa_special_class); |
3002 | EXPORT_SYMBOL(spa_preferred_class); | |
c28b2279 BB |
3003 | EXPORT_SYMBOL(spa_max_replication); |
3004 | EXPORT_SYMBOL(spa_prev_software_version); | |
3005 | EXPORT_SYMBOL(spa_get_failmode); | |
3006 | EXPORT_SYMBOL(spa_suspended); | |
3007 | EXPORT_SYMBOL(spa_bootfs); | |
3008 | EXPORT_SYMBOL(spa_delegation); | |
3009 | EXPORT_SYMBOL(spa_meta_objset); | |
f1512ee6 | 3010 | EXPORT_SYMBOL(spa_maxblocksize); |
50c957f7 | 3011 | EXPORT_SYMBOL(spa_maxdnodesize); |
c28b2279 BB |
3012 | |
3013 | /* Miscellaneous support routines */ | |
c28b2279 BB |
3014 | EXPORT_SYMBOL(spa_guid_exists); |
3015 | EXPORT_SYMBOL(spa_strdup); | |
3016 | EXPORT_SYMBOL(spa_strfree); | |
c28b2279 | 3017 | EXPORT_SYMBOL(spa_generate_guid); |
b0bc7a84 | 3018 | EXPORT_SYMBOL(snprintf_blkptr); |
c28b2279 BB |
3019 | EXPORT_SYMBOL(spa_freeze); |
3020 | EXPORT_SYMBOL(spa_upgrade); | |
3021 | EXPORT_SYMBOL(spa_evict_all); | |
3022 | EXPORT_SYMBOL(spa_lookup_by_guid); | |
3023 | EXPORT_SYMBOL(spa_has_spare); | |
3024 | EXPORT_SYMBOL(dva_get_dsize_sync); | |
3025 | EXPORT_SYMBOL(bp_get_dsize_sync); | |
3026 | EXPORT_SYMBOL(bp_get_dsize); | |
3027 | EXPORT_SYMBOL(spa_has_slogs); | |
3028 | EXPORT_SYMBOL(spa_is_root); | |
3029 | EXPORT_SYMBOL(spa_writeable); | |
3030 | EXPORT_SYMBOL(spa_mode); | |
c28b2279 | 3031 | EXPORT_SYMBOL(spa_namespace_lock); |
6cb8e530 PZ |
3032 | EXPORT_SYMBOL(spa_trust_config); |
3033 | EXPORT_SYMBOL(spa_missing_tvds_allowed); | |
3034 | EXPORT_SYMBOL(spa_set_missing_tvds); | |
f0ed6c74 | 3035 | EXPORT_SYMBOL(spa_state_to_name); |
d2734cce SD |
3036 | EXPORT_SYMBOL(spa_importing_readonly_checkpoint); |
3037 | EXPORT_SYMBOL(spa_min_claim_txg); | |
3038 | EXPORT_SYMBOL(spa_suspend_async_destroy); | |
3039 | EXPORT_SYMBOL(spa_has_checkpoint); | |
3040 | EXPORT_SYMBOL(spa_top_vdevs_spacemap_addressable); | |
cc92e9d0 | 3041 | |
03fdcb9a MM |
3042 | ZFS_MODULE_PARAM(zfs, zfs_, flags, UINT, ZMOD_RW, |
3043 | "Set additional debugging flags"); | |
0b39b9f9 | 3044 | |
03fdcb9a MM |
3045 | ZFS_MODULE_PARAM(zfs, zfs_, recover, INT, ZMOD_RW, |
3046 | "Set to attempt to recover from fatal errors"); | |
0b39b9f9 | 3047 | |
03fdcb9a | 3048 | ZFS_MODULE_PARAM(zfs, zfs_, free_leak_on_eio, INT, ZMOD_RW, |
0b39b9f9 PS |
3049 | "Set to ignore IO errors during free and permanently leak the space"); |
3050 | ||
ab8d9c17 | 3051 | ZFS_MODULE_PARAM(zfs_deadman, zfs_deadman_, checktime_ms, U64, ZMOD_RW, |
03fdcb9a MM |
3052 | "Dead I/O check interval in milliseconds"); |
3053 | ||
35aa9dc6 | 3054 | ZFS_MODULE_PARAM(zfs_deadman, zfs_deadman_, enabled, INT, ZMOD_RW, |
03fdcb9a MM |
3055 | "Enable deadman timer"); |
3056 | ||
fdc2d303 | 3057 | ZFS_MODULE_PARAM(zfs_spa, spa_, asize_inflation, UINT, ZMOD_RW, |
03fdcb9a MM |
3058 | "SPA size estimate multiplication factor"); |
3059 | ||
3060 | ZFS_MODULE_PARAM(zfs, zfs_, ddt_data_is_special, INT, ZMOD_RW, | |
3061 | "Place DDT data into the special class"); | |
3062 | ||
3063 | ZFS_MODULE_PARAM(zfs, zfs_, user_indirect_is_special, INT, ZMOD_RW, | |
3064 | "Place user data indirect blocks into the special class"); | |
3065 | ||
03fdcb9a | 3066 | /* BEGIN CSTYLED */ |
e64e84ec MM |
3067 | ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, failmode, |
3068 | param_set_deadman_failmode, param_get_charp, ZMOD_RW, | |
3069 | "Failmode for deadman timer"); | |
3070 | ||
2a3aa5a1 | 3071 | ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, synctime_ms, |
ab8d9c17 | 3072 | param_set_deadman_synctime, spl_param_get_u64, ZMOD_RW, |
2a3aa5a1 MM |
3073 | "Pool sync expiration time in milliseconds"); |
3074 | ||
3075 | ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, ziotime_ms, | |
ab8d9c17 | 3076 | param_set_deadman_ziotime, spl_param_get_u64, ZMOD_RW, |
2a3aa5a1 MM |
3077 | "IO expiration time in milliseconds"); |
3078 | ||
fdc2d303 | 3079 | ZFS_MODULE_PARAM(zfs, zfs_, special_class_metadata_reserve_pct, UINT, ZMOD_RW, |
1f02ecc5 D |
3080 | "Small file blocks in special vdevs depends on this much " |
3081 | "free space available"); | |
02730c33 | 3082 | /* END CSTYLED */ |
2a3aa5a1 | 3083 | |
7e3df9db | 3084 | ZFS_MODULE_PARAM_CALL(zfs_spa, spa_, slop_shift, param_set_slop_shift, |
fdc2d303 | 3085 | param_get_uint, ZMOD_RW, "Reserved free space in pool"); |
3bd4df38 EN |
3086 | |
3087 | ZFS_MODULE_PARAM(zfs, spa_, num_allocators, INT, ZMOD_RW, | |
3088 | "Number of allocators per spa, capped by ncpus"); |