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