]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
1 | /* |
2 | * CDDL HEADER START | |
3 | * | |
4 | * The contents of this file are subject to the terms of the | |
5 | * Common Development and Distribution License (the "License"). | |
6 | * You may not use this file except in compliance with the License. | |
7 | * | |
8 | * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE | |
9 | * or http://www.opensolaris.org/os/licensing. | |
10 | * See the License for the specific language governing permissions | |
11 | * and limitations under the License. | |
12 | * | |
13 | * When distributing Covered Code, include this CDDL HEADER in each | |
14 | * file and include the License file at usr/src/OPENSOLARIS.LICENSE. | |
15 | * If applicable, add the following below this CDDL HEADER, with the | |
16 | * fields enclosed by brackets "[]" replaced with your own identifying | |
17 | * information: Portions Copyright [yyyy] [name of copyright owner] | |
18 | * | |
19 | * CDDL HEADER END | |
20 | */ | |
21 | /* | |
428870ff | 22 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
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 BB |
32 | #include <sys/zfs_context.h> |
33 | #include <sys/spa_impl.h> | |
34 | #include <sys/zio.h> | |
35 | #include <sys/zio_checksum.h> | |
36 | #include <sys/zio_compress.h> | |
37 | #include <sys/dmu.h> | |
38 | #include <sys/dmu_tx.h> | |
39 | #include <sys/zap.h> | |
40 | #include <sys/zil.h> | |
41 | #include <sys/vdev_impl.h> | |
619f0976 | 42 | #include <sys/vdev_initialize.h> |
1b939560 | 43 | #include <sys/vdev_trim.h> |
bc25c932 | 44 | #include <sys/vdev_file.h> |
ab9f4b0b | 45 | #include <sys/vdev_raidz.h> |
34dc7c2f BB |
46 | #include <sys/metaslab.h> |
47 | #include <sys/uberblock_impl.h> | |
48 | #include <sys/txg.h> | |
49 | #include <sys/avl.h> | |
50 | #include <sys/unique.h> | |
51 | #include <sys/dsl_pool.h> | |
52 | #include <sys/dsl_dir.h> | |
53 | #include <sys/dsl_prop.h> | |
26685276 | 54 | #include <sys/fm/util.h> |
428870ff | 55 | #include <sys/dsl_scan.h> |
34dc7c2f BB |
56 | #include <sys/fs/zfs.h> |
57 | #include <sys/metaslab_impl.h> | |
b128c09f | 58 | #include <sys/arc.h> |
428870ff | 59 | #include <sys/ddt.h> |
1421c891 | 60 | #include <sys/kstat.h> |
34dc7c2f | 61 | #include "zfs_prop.h" |
ca577779 | 62 | #include <sys/btree.h> |
3c67d83a | 63 | #include <sys/zfeature.h> |
bced7e3a | 64 | #include <sys/qat.h> |
10b3c7f5 | 65 | #include <sys/zstd/zstd.h> |
34dc7c2f BB |
66 | |
67 | /* | |
68 | * SPA locking | |
69 | * | |
93e28d66 | 70 | * There are three basic locks for managing spa_t structures: |
34dc7c2f BB |
71 | * |
72 | * spa_namespace_lock (global mutex) | |
73 | * | |
74 | * This lock must be acquired to do any of the following: | |
75 | * | |
76 | * - Lookup a spa_t by name | |
77 | * - Add or remove a spa_t from the namespace | |
78 | * - Increase spa_refcount from non-zero | |
79 | * - Check if spa_refcount is zero | |
80 | * - Rename a spa_t | |
81 | * - add/remove/attach/detach devices | |
82 | * - Held for the duration of create/destroy/import/export | |
83 | * | |
84 | * It does not need to handle recursion. A create or destroy may | |
85 | * reference objects (files or zvols) in other pools, but by | |
86 | * definition they must have an existing reference, and will never need | |
87 | * to lookup a spa_t by name. | |
88 | * | |
c13060e4 | 89 | * spa_refcount (per-spa zfs_refcount_t protected by mutex) |
34dc7c2f BB |
90 | * |
91 | * This reference count keep track of any active users of the spa_t. The | |
92 | * spa_t cannot be destroyed or freed while this is non-zero. Internally, | |
93 | * the refcount is never really 'zero' - opening a pool implicitly keeps | |
b128c09f | 94 | * some references in the DMU. Internally we check against spa_minref, but |
34dc7c2f BB |
95 | * present the image of a zero/non-zero value to consumers. |
96 | * | |
b128c09f | 97 | * spa_config_lock[] (per-spa array of rwlocks) |
34dc7c2f BB |
98 | * |
99 | * This protects the spa_t from config changes, and must be held in | |
100 | * the following circumstances: | |
101 | * | |
102 | * - RW_READER to perform I/O to the spa | |
103 | * - RW_WRITER to change the vdev config | |
104 | * | |
34dc7c2f BB |
105 | * The locking order is fairly straightforward: |
106 | * | |
107 | * spa_namespace_lock -> spa_refcount | |
108 | * | |
109 | * The namespace lock must be acquired to increase the refcount from 0 | |
110 | * or to check if it is zero. | |
111 | * | |
b128c09f | 112 | * spa_refcount -> spa_config_lock[] |
34dc7c2f BB |
113 | * |
114 | * There must be at least one valid reference on the spa_t to acquire | |
115 | * the config lock. | |
116 | * | |
b128c09f | 117 | * spa_namespace_lock -> spa_config_lock[] |
34dc7c2f BB |
118 | * |
119 | * The namespace lock must always be taken before the config lock. | |
120 | * | |
121 | * | |
b128c09f | 122 | * The spa_namespace_lock can be acquired directly and is globally visible. |
34dc7c2f | 123 | * |
b128c09f BB |
124 | * The namespace is manipulated using the following functions, all of which |
125 | * require the spa_namespace_lock to be held. | |
34dc7c2f BB |
126 | * |
127 | * spa_lookup() Lookup a spa_t by name. | |
128 | * | |
129 | * spa_add() Create a new spa_t in the namespace. | |
130 | * | |
131 | * spa_remove() Remove a spa_t from the namespace. This also | |
132 | * frees up any memory associated with the spa_t. | |
133 | * | |
134 | * spa_next() Returns the next spa_t in the system, or the | |
135 | * first if NULL is passed. | |
136 | * | |
137 | * spa_evict_all() Shutdown and remove all spa_t structures in | |
138 | * the system. | |
139 | * | |
140 | * spa_guid_exists() Determine whether a pool/device guid exists. | |
141 | * | |
142 | * The spa_refcount is manipulated using the following functions: | |
143 | * | |
144 | * spa_open_ref() Adds a reference to the given spa_t. Must be | |
145 | * called with spa_namespace_lock held if the | |
146 | * refcount is currently zero. | |
147 | * | |
148 | * spa_close() Remove a reference from the spa_t. This will | |
149 | * not free the spa_t or remove it from the | |
150 | * namespace. No locking is required. | |
151 | * | |
152 | * spa_refcount_zero() Returns true if the refcount is currently | |
153 | * zero. Must be called with spa_namespace_lock | |
154 | * held. | |
155 | * | |
b128c09f BB |
156 | * The spa_config_lock[] is an array of rwlocks, ordered as follows: |
157 | * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV. | |
158 | * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}(). | |
159 | * | |
160 | * To read the configuration, it suffices to hold one of these locks as reader. | |
161 | * To modify the configuration, you must hold all locks as writer. To modify | |
162 | * vdev state without altering the vdev tree's topology (e.g. online/offline), | |
163 | * you must hold SCL_STATE and SCL_ZIO as writer. | |
164 | * | |
165 | * We use these distinct config locks to avoid recursive lock entry. | |
166 | * For example, spa_sync() (which holds SCL_CONFIG as reader) induces | |
167 | * block allocations (SCL_ALLOC), which may require reading space maps | |
168 | * from disk (dmu_read() -> zio_read() -> SCL_ZIO). | |
169 | * | |
170 | * The spa config locks cannot be normal rwlocks because we need the | |
171 | * ability to hand off ownership. For example, SCL_ZIO is acquired | |
172 | * by the issuing thread and later released by an interrupt thread. | |
173 | * They do, however, obey the usual write-wanted semantics to prevent | |
174 | * writer (i.e. system administrator) starvation. | |
175 | * | |
176 | * The lock acquisition rules are as follows: | |
177 | * | |
178 | * SCL_CONFIG | |
179 | * Protects changes to the vdev tree topology, such as vdev | |
180 | * add/remove/attach/detach. Protects the dirty config list | |
181 | * (spa_config_dirty_list) and the set of spares and l2arc devices. | |
182 | * | |
183 | * SCL_STATE | |
184 | * Protects changes to pool state and vdev state, such as vdev | |
185 | * online/offline/fault/degrade/clear. Protects the dirty state list | |
186 | * (spa_state_dirty_list) and global pool state (spa_state). | |
187 | * | |
188 | * SCL_ALLOC | |
189 | * Protects changes to metaslab groups and classes. | |
190 | * Held as reader by metaslab_alloc() and metaslab_claim(). | |
191 | * | |
192 | * SCL_ZIO | |
193 | * Held by bp-level zios (those which have no io_vd upon entry) | |
194 | * to prevent changes to the vdev tree. The bp-level zio implicitly | |
195 | * protects all of its vdev child zios, which do not hold SCL_ZIO. | |
196 | * | |
197 | * SCL_FREE | |
198 | * Protects changes to metaslab groups and classes. | |
199 | * Held as reader by metaslab_free(). SCL_FREE is distinct from | |
200 | * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free | |
201 | * blocks in zio_done() while another i/o that holds either | |
202 | * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete. | |
203 | * | |
204 | * SCL_VDEV | |
205 | * Held as reader to prevent changes to the vdev tree during trivial | |
428870ff | 206 | * inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the |
b128c09f BB |
207 | * other locks, and lower than all of them, to ensure that it's safe |
208 | * to acquire regardless of caller context. | |
209 | * | |
210 | * In addition, the following rules apply: | |
211 | * | |
212 | * (a) spa_props_lock protects pool properties, spa_config and spa_config_list. | |
213 | * The lock ordering is SCL_CONFIG > spa_props_lock. | |
214 | * | |
215 | * (b) I/O operations on leaf vdevs. For any zio operation that takes | |
216 | * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(), | |
217 | * or zio_write_phys() -- the caller must ensure that the config cannot | |
218 | * cannot change in the interim, and that the vdev cannot be reopened. | |
219 | * SCL_STATE as reader suffices for both. | |
34dc7c2f BB |
220 | * |
221 | * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). | |
222 | * | |
223 | * spa_vdev_enter() Acquire the namespace lock and the config lock | |
224 | * for writing. | |
225 | * | |
226 | * spa_vdev_exit() Release the config lock, wait for all I/O | |
227 | * to complete, sync the updated configs to the | |
228 | * cache, and release the namespace lock. | |
229 | * | |
b128c09f BB |
230 | * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit(). |
231 | * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual | |
232 | * locking is, always, based on spa_namespace_lock and spa_config_lock[]. | |
34dc7c2f BB |
233 | */ |
234 | ||
235 | static avl_tree_t spa_namespace_avl; | |
236 | kmutex_t spa_namespace_lock; | |
237 | static kcondvar_t spa_namespace_cv; | |
18168da7 | 238 | static const int spa_max_replication_override = SPA_DVAS_PER_BP; |
34dc7c2f BB |
239 | |
240 | static kmutex_t spa_spare_lock; | |
241 | static avl_tree_t spa_spare_avl; | |
242 | static kmutex_t spa_l2cache_lock; | |
243 | static avl_tree_t spa_l2cache_avl; | |
244 | ||
da92d5cb | 245 | spa_mode_t spa_mode_global = SPA_MODE_UNINIT; |
34dc7c2f | 246 | |
0b39b9f9 | 247 | #ifdef ZFS_DEBUG |
a1d477c2 MA |
248 | /* |
249 | * Everything except dprintf, set_error, spa, and indirect_remap is on | |
250 | * by default in debug builds. | |
251 | */ | |
252 | int zfs_flags = ~(ZFS_DEBUG_DPRINTF | ZFS_DEBUG_SET_ERROR | | |
964c2d69 | 253 | ZFS_DEBUG_INDIRECT_REMAP); |
0b39b9f9 PS |
254 | #else |
255 | int zfs_flags = 0; | |
256 | #endif | |
257 | ||
258 | /* | |
259 | * zfs_recover can be set to nonzero to attempt to recover from | |
260 | * otherwise-fatal errors, typically caused by on-disk corruption. When | |
261 | * set, calls to zfs_panic_recover() will turn into warning messages. | |
262 | * This should only be used as a last resort, as it typically results | |
263 | * in leaked space, or worse. | |
264 | */ | |
265 | int zfs_recover = B_FALSE; | |
266 | ||
267 | /* | |
268 | * If destroy encounters an EIO while reading metadata (e.g. indirect | |
269 | * blocks), space referenced by the missing metadata can not be freed. | |
270 | * Normally this causes the background destroy to become "stalled", as | |
271 | * it is unable to make forward progress. While in this stalled state, | |
272 | * all remaining space to free from the error-encountering filesystem is | |
273 | * "temporarily leaked". Set this flag to cause it to ignore the EIO, | |
274 | * permanently leak the space from indirect blocks that can not be read, | |
275 | * and continue to free everything else that it can. | |
276 | * | |
277 | * The default, "stalling" behavior is useful if the storage partially | |
278 | * fails (i.e. some but not all i/os fail), and then later recovers. In | |
279 | * this case, we will be able to continue pool operations while it is | |
280 | * partially failed, and when it recovers, we can continue to free the | |
281 | * space, with no leaks. However, note that this case is actually | |
282 | * fairly rare. | |
283 | * | |
284 | * Typically pools either (a) fail completely (but perhaps temporarily, | |
285 | * e.g. a top-level vdev going offline), or (b) have localized, | |
286 | * permanent errors (e.g. disk returns the wrong data due to bit flip or | |
287 | * firmware bug). In case (a), this setting does not matter because the | |
288 | * pool will be suspended and the sync thread will not be able to make | |
289 | * forward progress regardless. In case (b), because the error is | |
290 | * permanent, the best we can do is leak the minimum amount of space, | |
291 | * which is what setting this flag will do. Therefore, it is reasonable | |
292 | * for this flag to normally be set, but we chose the more conservative | |
293 | * approach of not setting it, so that there is no possibility of | |
294 | * leaking space in the "partial temporary" failure case. | |
295 | */ | |
296 | int zfs_free_leak_on_eio = B_FALSE; | |
297 | ||
cc92e9d0 | 298 | /* |
e8b96c60 MA |
299 | * Expiration time in milliseconds. This value has two meanings. First it is |
300 | * used to determine when the spa_deadman() logic should fire. By default the | |
8fb1ede1 | 301 | * spa_deadman() will fire if spa_sync() has not completed in 600 seconds. |
e8b96c60 MA |
302 | * Secondly, the value determines if an I/O is considered "hung". Any I/O that |
303 | * has not completed in zfs_deadman_synctime_ms is considered "hung" resulting | |
8fb1ede1 | 304 | * in one of three behaviors controlled by zfs_deadman_failmode. |
cc92e9d0 | 305 | */ |
18168da7 | 306 | unsigned long zfs_deadman_synctime_ms = 600000UL; /* 10 min. */ |
8fb1ede1 BB |
307 | |
308 | /* | |
309 | * This value controls the maximum amount of time zio_wait() will block for an | |
310 | * outstanding IO. By default this is 300 seconds at which point the "hung" | |
311 | * behavior will be applied as described for zfs_deadman_synctime_ms. | |
312 | */ | |
18168da7 | 313 | unsigned long zfs_deadman_ziotime_ms = 300000UL; /* 5 min. */ |
cc92e9d0 | 314 | |
b81a3ddc TC |
315 | /* |
316 | * Check time in milliseconds. This defines the frequency at which we check | |
317 | * for hung I/O. | |
318 | */ | |
18168da7 | 319 | unsigned long zfs_deadman_checktime_ms = 60000UL; /* 1 min. */ |
b81a3ddc | 320 | |
cc92e9d0 GW |
321 | /* |
322 | * By default the deadman is enabled. | |
323 | */ | |
18168da7 | 324 | int zfs_deadman_enabled = B_TRUE; |
cc92e9d0 | 325 | |
8fb1ede1 BB |
326 | /* |
327 | * Controls the behavior of the deadman when it detects a "hung" I/O. | |
328 | * Valid values are zfs_deadman_failmode=<wait|continue|panic>. | |
329 | * | |
330 | * wait - Wait for the "hung" I/O (default) | |
331 | * continue - Attempt to recover from a "hung" I/O | |
332 | * panic - Panic the system | |
333 | */ | |
18168da7 | 334 | const char *zfs_deadman_failmode = "wait"; |
8fb1ede1 | 335 | |
e8b96c60 MA |
336 | /* |
337 | * The worst case is single-sector max-parity RAID-Z blocks, in which | |
338 | * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1) | |
339 | * times the size; so just assume that. Add to this the fact that | |
340 | * we can have up to 3 DVAs per bp, and one more factor of 2 because | |
341 | * the block may be dittoed with up to 3 DVAs by ddt_sync(). All together, | |
342 | * the worst case is: | |
343 | * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2 == 24 | |
344 | */ | |
345 | int spa_asize_inflation = 24; | |
346 | ||
3d45fdd6 MA |
347 | /* |
348 | * Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space in | |
f01eaed4 PS |
349 | * the pool to be consumed (bounded by spa_max_slop). This ensures that we |
350 | * don't run the pool completely out of space, due to unaccounted changes (e.g. | |
351 | * to the MOS). It also limits the worst-case time to allocate space. If we | |
352 | * have less than this amount of free space, most ZPL operations (e.g. write, | |
353 | * create) will return ENOSPC. The ZIL metaslabs (spa_embedded_log_class) are | |
354 | * also part of this 3.2% of space which can't be consumed by normal writes; | |
355 | * the slop space "proper" (spa_get_slop_space()) is decreased by the embedded | |
356 | * log space. | |
3d45fdd6 MA |
357 | * |
358 | * Certain operations (e.g. file removal, most administrative actions) can | |
359 | * use half the slop space. They will only return ENOSPC if less than half | |
360 | * the slop space is free. Typically, once the pool has less than the slop | |
361 | * space free, the user will use these operations to free up space in the pool. | |
362 | * These are the operations that call dsl_pool_adjustedsize() with the netfree | |
363 | * argument set to TRUE. | |
364 | * | |
d2734cce SD |
365 | * Operations that are almost guaranteed to free up space in the absence of |
366 | * a pool checkpoint can use up to three quarters of the slop space | |
367 | * (e.g zfs destroy). | |
368 | * | |
3d45fdd6 MA |
369 | * A very restricted set of operations are always permitted, regardless of |
370 | * the amount of free space. These are the operations that call | |
d2734cce SD |
371 | * dsl_sync_task(ZFS_SPACE_CHECK_NONE). If these operations result in a net |
372 | * increase in the amount of space used, it is possible to run the pool | |
373 | * completely out of space, causing it to be permanently read-only. | |
3d45fdd6 | 374 | * |
d7958b4c MA |
375 | * Note that on very small pools, the slop space will be larger than |
376 | * 3.2%, in an effort to have it be at least spa_min_slop (128MB), | |
377 | * but we never allow it to be more than half the pool size. | |
378 | * | |
f01eaed4 PS |
379 | * Further, on very large pools, the slop space will be smaller than |
380 | * 3.2%, to avoid reserving much more space than we actually need; bounded | |
381 | * by spa_max_slop (128GB). | |
382 | * | |
3d45fdd6 MA |
383 | * See also the comments in zfs_space_check_t. |
384 | */ | |
385 | int spa_slop_shift = 5; | |
18168da7 AZ |
386 | static const uint64_t spa_min_slop = 128ULL * 1024 * 1024; |
387 | static const uint64_t spa_max_slop = 128ULL * 1024 * 1024 * 1024; | |
388 | static const int spa_allocators = 4; | |
492f64e9 | 389 | |
3d45fdd6 | 390 | |
4a0ee12a PZ |
391 | void |
392 | spa_load_failed(spa_t *spa, const char *fmt, ...) | |
393 | { | |
394 | va_list adx; | |
395 | char buf[256]; | |
396 | ||
397 | va_start(adx, fmt); | |
398 | (void) vsnprintf(buf, sizeof (buf), fmt, adx); | |
399 | va_end(adx); | |
400 | ||
6cb8e530 PZ |
401 | zfs_dbgmsg("spa_load(%s, config %s): FAILED: %s", spa->spa_name, |
402 | spa->spa_trust_config ? "trusted" : "untrusted", buf); | |
4a0ee12a PZ |
403 | } |
404 | ||
4a0ee12a PZ |
405 | void |
406 | spa_load_note(spa_t *spa, const char *fmt, ...) | |
407 | { | |
408 | va_list adx; | |
409 | char buf[256]; | |
410 | ||
411 | va_start(adx, fmt); | |
412 | (void) vsnprintf(buf, sizeof (buf), fmt, adx); | |
413 | va_end(adx); | |
414 | ||
6cb8e530 PZ |
415 | zfs_dbgmsg("spa_load(%s, config %s): %s", spa->spa_name, |
416 | spa->spa_trust_config ? "trusted" : "untrusted", buf); | |
4a0ee12a PZ |
417 | } |
418 | ||
cc99f275 DB |
419 | /* |
420 | * By default dedup and user data indirects land in the special class | |
421 | */ | |
18168da7 AZ |
422 | static int zfs_ddt_data_is_special = B_TRUE; |
423 | static int zfs_user_indirect_is_special = B_TRUE; | |
cc99f275 DB |
424 | |
425 | /* | |
426 | * The percentage of special class final space reserved for metadata only. | |
427 | * Once we allocate 100 - zfs_special_class_metadata_reserve_pct we only | |
428 | * let metadata into the class. | |
429 | */ | |
18168da7 | 430 | static int zfs_special_class_metadata_reserve_pct = 25; |
cc99f275 | 431 | |
34dc7c2f BB |
432 | /* |
433 | * ========================================================================== | |
434 | * SPA config locking | |
435 | * ========================================================================== | |
436 | */ | |
437 | static void | |
b128c09f BB |
438 | spa_config_lock_init(spa_t *spa) |
439 | { | |
1c27024e | 440 | for (int i = 0; i < SCL_LOCKS; i++) { |
b128c09f BB |
441 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
442 | mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL); | |
443 | cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL); | |
b128c09f BB |
444 | scl->scl_writer = NULL; |
445 | scl->scl_write_wanted = 0; | |
42afb12d | 446 | scl->scl_count = 0; |
b128c09f | 447 | } |
34dc7c2f BB |
448 | } |
449 | ||
450 | static void | |
b128c09f BB |
451 | spa_config_lock_destroy(spa_t *spa) |
452 | { | |
1c27024e | 453 | for (int i = 0; i < SCL_LOCKS; i++) { |
b128c09f BB |
454 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
455 | mutex_destroy(&scl->scl_lock); | |
456 | cv_destroy(&scl->scl_cv); | |
b128c09f BB |
457 | ASSERT(scl->scl_writer == NULL); |
458 | ASSERT(scl->scl_write_wanted == 0); | |
42afb12d | 459 | ASSERT(scl->scl_count == 0); |
b128c09f BB |
460 | } |
461 | } | |
462 | ||
463 | int | |
464 | spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw) | |
34dc7c2f | 465 | { |
1c27024e | 466 | for (int i = 0; i < SCL_LOCKS; i++) { |
b128c09f BB |
467 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
468 | if (!(locks & (1 << i))) | |
469 | continue; | |
470 | mutex_enter(&scl->scl_lock); | |
471 | if (rw == RW_READER) { | |
472 | if (scl->scl_writer || scl->scl_write_wanted) { | |
473 | mutex_exit(&scl->scl_lock); | |
adfe9d93 SK |
474 | spa_config_exit(spa, locks & ((1 << i) - 1), |
475 | tag); | |
b128c09f BB |
476 | return (0); |
477 | } | |
478 | } else { | |
479 | ASSERT(scl->scl_writer != curthread); | |
42afb12d | 480 | if (scl->scl_count != 0) { |
b128c09f | 481 | mutex_exit(&scl->scl_lock); |
adfe9d93 SK |
482 | spa_config_exit(spa, locks & ((1 << i) - 1), |
483 | tag); | |
b128c09f BB |
484 | return (0); |
485 | } | |
486 | scl->scl_writer = curthread; | |
487 | } | |
42afb12d | 488 | scl->scl_count++; |
b128c09f BB |
489 | mutex_exit(&scl->scl_lock); |
490 | } | |
491 | return (1); | |
34dc7c2f BB |
492 | } |
493 | ||
494 | void | |
dc04a8c7 | 495 | spa_config_enter(spa_t *spa, int locks, const void *tag, krw_t rw) |
34dc7c2f | 496 | { |
14e4e3cb | 497 | (void) tag; |
45d1cae3 BB |
498 | int wlocks_held = 0; |
499 | ||
13fe0198 MA |
500 | ASSERT3U(SCL_LOCKS, <, sizeof (wlocks_held) * NBBY); |
501 | ||
1c27024e | 502 | for (int i = 0; i < SCL_LOCKS; i++) { |
b128c09f | 503 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
45d1cae3 BB |
504 | if (scl->scl_writer == curthread) |
505 | wlocks_held |= (1 << i); | |
b128c09f BB |
506 | if (!(locks & (1 << i))) |
507 | continue; | |
508 | mutex_enter(&scl->scl_lock); | |
509 | if (rw == RW_READER) { | |
510 | while (scl->scl_writer || scl->scl_write_wanted) { | |
511 | cv_wait(&scl->scl_cv, &scl->scl_lock); | |
512 | } | |
513 | } else { | |
514 | ASSERT(scl->scl_writer != curthread); | |
42afb12d | 515 | while (scl->scl_count != 0) { |
b128c09f BB |
516 | scl->scl_write_wanted++; |
517 | cv_wait(&scl->scl_cv, &scl->scl_lock); | |
518 | scl->scl_write_wanted--; | |
519 | } | |
520 | scl->scl_writer = curthread; | |
521 | } | |
42afb12d | 522 | scl->scl_count++; |
b128c09f | 523 | mutex_exit(&scl->scl_lock); |
34dc7c2f | 524 | } |
a1d477c2 | 525 | ASSERT3U(wlocks_held, <=, locks); |
34dc7c2f BB |
526 | } |
527 | ||
528 | void | |
dc04a8c7 | 529 | spa_config_exit(spa_t *spa, int locks, const void *tag) |
34dc7c2f | 530 | { |
14e4e3cb | 531 | (void) tag; |
1c27024e | 532 | for (int i = SCL_LOCKS - 1; i >= 0; i--) { |
b128c09f BB |
533 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
534 | if (!(locks & (1 << i))) | |
535 | continue; | |
536 | mutex_enter(&scl->scl_lock); | |
42afb12d AM |
537 | ASSERT(scl->scl_count > 0); |
538 | if (--scl->scl_count == 0) { | |
b128c09f BB |
539 | ASSERT(scl->scl_writer == NULL || |
540 | scl->scl_writer == curthread); | |
541 | scl->scl_writer = NULL; /* OK in either case */ | |
542 | cv_broadcast(&scl->scl_cv); | |
543 | } | |
544 | mutex_exit(&scl->scl_lock); | |
34dc7c2f | 545 | } |
34dc7c2f BB |
546 | } |
547 | ||
b128c09f BB |
548 | int |
549 | spa_config_held(spa_t *spa, int locks, krw_t rw) | |
34dc7c2f | 550 | { |
1c27024e | 551 | int locks_held = 0; |
34dc7c2f | 552 | |
1c27024e | 553 | for (int i = 0; i < SCL_LOCKS; i++) { |
b128c09f BB |
554 | spa_config_lock_t *scl = &spa->spa_config_lock[i]; |
555 | if (!(locks & (1 << i))) | |
556 | continue; | |
42afb12d | 557 | if ((rw == RW_READER && scl->scl_count != 0) || |
b128c09f BB |
558 | (rw == RW_WRITER && scl->scl_writer == curthread)) |
559 | locks_held |= 1 << i; | |
560 | } | |
561 | ||
562 | return (locks_held); | |
34dc7c2f BB |
563 | } |
564 | ||
565 | /* | |
566 | * ========================================================================== | |
567 | * SPA namespace functions | |
568 | * ========================================================================== | |
569 | */ | |
570 | ||
571 | /* | |
572 | * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. | |
573 | * Returns NULL if no matching spa_t is found. | |
574 | */ | |
575 | spa_t * | |
576 | spa_lookup(const char *name) | |
577 | { | |
b128c09f BB |
578 | static spa_t search; /* spa_t is large; don't allocate on stack */ |
579 | spa_t *spa; | |
34dc7c2f | 580 | avl_index_t where; |
34dc7c2f BB |
581 | char *cp; |
582 | ||
583 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
584 | ||
13fe0198 MA |
585 | (void) strlcpy(search.spa_name, name, sizeof (search.spa_name)); |
586 | ||
34dc7c2f BB |
587 | /* |
588 | * If it's a full dataset name, figure out the pool name and | |
589 | * just use that. | |
590 | */ | |
da536844 | 591 | cp = strpbrk(search.spa_name, "/@#"); |
13fe0198 | 592 | if (cp != NULL) |
34dc7c2f | 593 | *cp = '\0'; |
34dc7c2f | 594 | |
34dc7c2f BB |
595 | spa = avl_find(&spa_namespace_avl, &search, &where); |
596 | ||
34dc7c2f BB |
597 | return (spa); |
598 | } | |
599 | ||
cc92e9d0 GW |
600 | /* |
601 | * Fires when spa_sync has not completed within zfs_deadman_synctime_ms. | |
602 | * If the zfs_deadman_enabled flag is set then it inspects all vdev queues | |
603 | * looking for potentially hung I/Os. | |
604 | */ | |
605 | void | |
606 | spa_deadman(void *arg) | |
607 | { | |
608 | spa_t *spa = arg; | |
609 | ||
b81a3ddc TC |
610 | /* Disable the deadman if the pool is suspended. */ |
611 | if (spa_suspended(spa)) | |
612 | return; | |
613 | ||
cc92e9d0 GW |
614 | zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu", |
615 | (gethrtime() - spa->spa_sync_starttime) / NANOSEC, | |
8e739b2c | 616 | (u_longlong_t)++spa->spa_deadman_calls); |
cc92e9d0 | 617 | if (zfs_deadman_enabled) |
8fb1ede1 | 618 | vdev_deadman(spa->spa_root_vdev, FTAG); |
cc92e9d0 | 619 | |
57ddcda1 | 620 | spa->spa_deadman_tqid = taskq_dispatch_delay(system_delay_taskq, |
f764edf0 | 621 | spa_deadman, spa, TQ_SLEEP, ddi_get_lbolt() + |
b81a3ddc | 622 | MSEC_TO_TICK(zfs_deadman_checktime_ms)); |
cc92e9d0 GW |
623 | } |
624 | ||
65c7cc49 | 625 | static int |
93e28d66 SD |
626 | spa_log_sm_sort_by_txg(const void *va, const void *vb) |
627 | { | |
628 | const spa_log_sm_t *a = va; | |
629 | const spa_log_sm_t *b = vb; | |
630 | ||
ca577779 | 631 | return (TREE_CMP(a->sls_txg, b->sls_txg)); |
93e28d66 SD |
632 | } |
633 | ||
34dc7c2f BB |
634 | /* |
635 | * Create an uninitialized spa_t with the given name. Requires | |
636 | * spa_namespace_lock. The caller must ensure that the spa_t doesn't already | |
637 | * exist by calling spa_lookup() first. | |
638 | */ | |
639 | spa_t * | |
428870ff | 640 | spa_add(const char *name, nvlist_t *config, const char *altroot) |
34dc7c2f BB |
641 | { |
642 | spa_t *spa; | |
b128c09f | 643 | spa_config_dirent_t *dp; |
34dc7c2f BB |
644 | |
645 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
646 | ||
79c76d5b | 647 | spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); |
34dc7c2f | 648 | |
34dc7c2f | 649 | mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 650 | mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); |
428870ff | 651 | mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); |
0c66c32d | 652 | mutex_init(&spa->spa_evicting_os_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 653 | mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); |
428870ff | 654 | mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 655 | mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); |
3c67d83a | 656 | mutex_init(&spa->spa_cksum_tmpls_lock, NULL, MUTEX_DEFAULT, NULL); |
428870ff BB |
657 | mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); |
658 | mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL); | |
659 | mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL); | |
4eb30c68 | 660 | mutex_init(&spa->spa_feat_stats_lock, NULL, MUTEX_DEFAULT, NULL); |
93e28d66 | 661 | mutex_init(&spa->spa_flushed_ms_lock, NULL, MUTEX_DEFAULT, NULL); |
e60e158e | 662 | mutex_init(&spa->spa_activities_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f BB |
663 | |
664 | cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); | |
0c66c32d | 665 | cv_init(&spa->spa_evicting_os_cv, NULL, CV_DEFAULT, NULL); |
428870ff | 666 | cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL); |
34dc7c2f | 667 | cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); |
b128c09f | 668 | cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL); |
e60e158e JG |
669 | cv_init(&spa->spa_activities_cv, NULL, CV_DEFAULT, NULL); |
670 | cv_init(&spa->spa_waiters_cv, NULL, CV_DEFAULT, NULL); | |
34dc7c2f | 671 | |
1c27024e | 672 | for (int t = 0; t < TXG_SIZE; t++) |
428870ff BB |
673 | bplist_create(&spa->spa_free_bplist[t]); |
674 | ||
b128c09f | 675 | (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name)); |
34dc7c2f BB |
676 | spa->spa_state = POOL_STATE_UNINITIALIZED; |
677 | spa->spa_freeze_txg = UINT64_MAX; | |
678 | spa->spa_final_txg = UINT64_MAX; | |
428870ff BB |
679 | spa->spa_load_max_txg = UINT64_MAX; |
680 | spa->spa_proc = &p0; | |
681 | spa->spa_proc_state = SPA_PROC_NONE; | |
6cb8e530 | 682 | spa->spa_trust_config = B_TRUE; |
25f06d67 | 683 | spa->spa_hostid = zone_get_hostid(NULL); |
34dc7c2f | 684 | |
e8b96c60 | 685 | spa->spa_deadman_synctime = MSEC2NSEC(zfs_deadman_synctime_ms); |
8fb1ede1 BB |
686 | spa->spa_deadman_ziotime = MSEC2NSEC(zfs_deadman_ziotime_ms); |
687 | spa_set_deadman_failmode(spa, zfs_deadman_failmode); | |
cc92e9d0 | 688 | |
424fd7c3 | 689 | zfs_refcount_create(&spa->spa_refcount); |
b128c09f | 690 | spa_config_lock_init(spa); |
1421c891 | 691 | spa_stats_init(spa); |
34dc7c2f BB |
692 | |
693 | avl_add(&spa_namespace_avl, spa); | |
694 | ||
34dc7c2f BB |
695 | /* |
696 | * Set the alternate root, if there is one. | |
697 | */ | |
0336f3d0 | 698 | if (altroot) |
34dc7c2f | 699 | spa->spa_root = spa_strdup(altroot); |
34dc7c2f | 700 | |
492f64e9 | 701 | spa->spa_alloc_count = spa_allocators; |
1b50749c AM |
702 | spa->spa_allocs = kmem_zalloc(spa->spa_alloc_count * |
703 | sizeof (spa_alloc_t), KM_SLEEP); | |
492f64e9 | 704 | for (int i = 0; i < spa->spa_alloc_count; i++) { |
1b50749c AM |
705 | mutex_init(&spa->spa_allocs[i].spaa_lock, NULL, MUTEX_DEFAULT, |
706 | NULL); | |
707 | avl_create(&spa->spa_allocs[i].spaa_tree, zio_bookmark_compare, | |
492f64e9 PD |
708 | sizeof (zio_t), offsetof(zio_t, io_alloc_node)); |
709 | } | |
93e28d66 SD |
710 | avl_create(&spa->spa_metaslabs_by_flushed, metaslab_sort_by_flushed, |
711 | sizeof (metaslab_t), offsetof(metaslab_t, ms_spa_txg_node)); | |
712 | avl_create(&spa->spa_sm_logs_by_txg, spa_log_sm_sort_by_txg, | |
713 | sizeof (spa_log_sm_t), offsetof(spa_log_sm_t, sls_node)); | |
714 | list_create(&spa->spa_log_summary, sizeof (log_summary_entry_t), | |
715 | offsetof(log_summary_entry_t, lse_node)); | |
3dfb57a3 | 716 | |
b128c09f BB |
717 | /* |
718 | * Every pool starts with the default cachefile | |
719 | */ | |
720 | list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t), | |
721 | offsetof(spa_config_dirent_t, scd_link)); | |
722 | ||
79c76d5b | 723 | dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP); |
428870ff | 724 | dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path); |
b128c09f BB |
725 | list_insert_head(&spa->spa_config_list, dp); |
726 | ||
572e2857 | 727 | VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME, |
79c76d5b | 728 | KM_SLEEP) == 0); |
572e2857 | 729 | |
9ae529ec CS |
730 | if (config != NULL) { |
731 | nvlist_t *features; | |
732 | ||
733 | if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ, | |
734 | &features) == 0) { | |
735 | VERIFY(nvlist_dup(features, &spa->spa_label_features, | |
736 | 0) == 0); | |
737 | } | |
738 | ||
428870ff | 739 | VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); |
9ae529ec CS |
740 | } |
741 | ||
742 | if (spa->spa_label_features == NULL) { | |
743 | VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME, | |
79c76d5b | 744 | KM_SLEEP) == 0); |
9ae529ec | 745 | } |
428870ff | 746 | |
c3520e7f MA |
747 | spa->spa_min_ashift = INT_MAX; |
748 | spa->spa_max_ashift = 0; | |
b2255edc | 749 | spa->spa_min_alloc = INT_MAX; |
c3520e7f | 750 | |
e8a20144 GN |
751 | /* Reset cached value */ |
752 | spa->spa_dedup_dspace = ~0ULL; | |
753 | ||
b0bc7a84 MG |
754 | /* |
755 | * As a pool is being created, treat all features as disabled by | |
756 | * setting SPA_FEATURE_DISABLED for all entries in the feature | |
757 | * refcount cache. | |
758 | */ | |
1c27024e | 759 | for (int i = 0; i < SPA_FEATURES; i++) { |
b0bc7a84 MG |
760 | spa->spa_feat_refcount_cache[i] = SPA_FEATURE_DISABLED; |
761 | } | |
762 | ||
3d31aad8 OF |
763 | list_create(&spa->spa_leaf_list, sizeof (vdev_t), |
764 | offsetof(vdev_t, vdev_leaf_node)); | |
765 | ||
34dc7c2f BB |
766 | return (spa); |
767 | } | |
768 | ||
769 | /* | |
770 | * Removes a spa_t from the namespace, freeing up any memory used. Requires | |
771 | * spa_namespace_lock. This is called only after the spa_t has been closed and | |
772 | * deactivated. | |
773 | */ | |
774 | void | |
775 | spa_remove(spa_t *spa) | |
776 | { | |
b128c09f BB |
777 | spa_config_dirent_t *dp; |
778 | ||
34dc7c2f | 779 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); |
93e28d66 | 780 | ASSERT(spa_state(spa) == POOL_STATE_UNINITIALIZED); |
424fd7c3 | 781 | ASSERT3U(zfs_refcount_count(&spa->spa_refcount), ==, 0); |
e60e158e | 782 | ASSERT0(spa->spa_waiters); |
34dc7c2f | 783 | |
428870ff BB |
784 | nvlist_free(spa->spa_config_splitting); |
785 | ||
34dc7c2f BB |
786 | avl_remove(&spa_namespace_avl, spa); |
787 | cv_broadcast(&spa_namespace_cv); | |
788 | ||
0336f3d0 | 789 | if (spa->spa_root) |
34dc7c2f | 790 | spa_strfree(spa->spa_root); |
34dc7c2f | 791 | |
b128c09f BB |
792 | while ((dp = list_head(&spa->spa_config_list)) != NULL) { |
793 | list_remove(&spa->spa_config_list, dp); | |
794 | if (dp->scd_path != NULL) | |
795 | spa_strfree(dp->scd_path); | |
796 | kmem_free(dp, sizeof (spa_config_dirent_t)); | |
797 | } | |
34dc7c2f | 798 | |
492f64e9 | 799 | for (int i = 0; i < spa->spa_alloc_count; i++) { |
1b50749c AM |
800 | avl_destroy(&spa->spa_allocs[i].spaa_tree); |
801 | mutex_destroy(&spa->spa_allocs[i].spaa_lock); | |
492f64e9 | 802 | } |
1b50749c AM |
803 | kmem_free(spa->spa_allocs, spa->spa_alloc_count * |
804 | sizeof (spa_alloc_t)); | |
492f64e9 | 805 | |
93e28d66 SD |
806 | avl_destroy(&spa->spa_metaslabs_by_flushed); |
807 | avl_destroy(&spa->spa_sm_logs_by_txg); | |
808 | list_destroy(&spa->spa_log_summary); | |
b128c09f | 809 | list_destroy(&spa->spa_config_list); |
3d31aad8 | 810 | list_destroy(&spa->spa_leaf_list); |
34dc7c2f | 811 | |
9ae529ec | 812 | nvlist_free(spa->spa_label_features); |
572e2857 | 813 | nvlist_free(spa->spa_load_info); |
417104bd | 814 | nvlist_free(spa->spa_feat_stats); |
34dc7c2f BB |
815 | spa_config_set(spa, NULL); |
816 | ||
424fd7c3 | 817 | zfs_refcount_destroy(&spa->spa_refcount); |
34dc7c2f | 818 | |
1421c891 | 819 | spa_stats_destroy(spa); |
b128c09f | 820 | spa_config_lock_destroy(spa); |
34dc7c2f | 821 | |
1c27024e | 822 | for (int t = 0; t < TXG_SIZE; t++) |
428870ff BB |
823 | bplist_destroy(&spa->spa_free_bplist[t]); |
824 | ||
3c67d83a TH |
825 | zio_checksum_templates_free(spa); |
826 | ||
34dc7c2f | 827 | cv_destroy(&spa->spa_async_cv); |
0c66c32d | 828 | cv_destroy(&spa->spa_evicting_os_cv); |
428870ff | 829 | cv_destroy(&spa->spa_proc_cv); |
34dc7c2f | 830 | cv_destroy(&spa->spa_scrub_io_cv); |
b128c09f | 831 | cv_destroy(&spa->spa_suspend_cv); |
e60e158e JG |
832 | cv_destroy(&spa->spa_activities_cv); |
833 | cv_destroy(&spa->spa_waiters_cv); | |
34dc7c2f | 834 | |
93e28d66 | 835 | mutex_destroy(&spa->spa_flushed_ms_lock); |
34dc7c2f | 836 | mutex_destroy(&spa->spa_async_lock); |
34dc7c2f | 837 | mutex_destroy(&spa->spa_errlist_lock); |
428870ff | 838 | mutex_destroy(&spa->spa_errlog_lock); |
0c66c32d | 839 | mutex_destroy(&spa->spa_evicting_os_lock); |
34dc7c2f | 840 | mutex_destroy(&spa->spa_history_lock); |
428870ff | 841 | mutex_destroy(&spa->spa_proc_lock); |
34dc7c2f | 842 | mutex_destroy(&spa->spa_props_lock); |
3c67d83a | 843 | mutex_destroy(&spa->spa_cksum_tmpls_lock); |
428870ff | 844 | mutex_destroy(&spa->spa_scrub_lock); |
b128c09f | 845 | mutex_destroy(&spa->spa_suspend_lock); |
428870ff | 846 | mutex_destroy(&spa->spa_vdev_top_lock); |
4eb30c68 | 847 | mutex_destroy(&spa->spa_feat_stats_lock); |
e60e158e | 848 | mutex_destroy(&spa->spa_activities_lock); |
34dc7c2f BB |
849 | |
850 | kmem_free(spa, sizeof (spa_t)); | |
851 | } | |
852 | ||
853 | /* | |
854 | * Given a pool, return the next pool in the namespace, or NULL if there is | |
855 | * none. If 'prev' is NULL, return the first pool. | |
856 | */ | |
857 | spa_t * | |
858 | spa_next(spa_t *prev) | |
859 | { | |
860 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
861 | ||
862 | if (prev) | |
863 | return (AVL_NEXT(&spa_namespace_avl, prev)); | |
864 | else | |
865 | return (avl_first(&spa_namespace_avl)); | |
866 | } | |
867 | ||
868 | /* | |
869 | * ========================================================================== | |
870 | * SPA refcount functions | |
871 | * ========================================================================== | |
872 | */ | |
873 | ||
874 | /* | |
875 | * Add a reference to the given spa_t. Must have at least one reference, or | |
876 | * have the namespace lock held. | |
877 | */ | |
878 | void | |
879 | spa_open_ref(spa_t *spa, void *tag) | |
880 | { | |
424fd7c3 | 881 | ASSERT(zfs_refcount_count(&spa->spa_refcount) >= spa->spa_minref || |
34dc7c2f | 882 | MUTEX_HELD(&spa_namespace_lock)); |
c13060e4 | 883 | (void) zfs_refcount_add(&spa->spa_refcount, tag); |
34dc7c2f BB |
884 | } |
885 | ||
886 | /* | |
887 | * Remove a reference to the given spa_t. Must have at least one reference, or | |
888 | * have the namespace lock held. | |
889 | */ | |
890 | void | |
891 | spa_close(spa_t *spa, void *tag) | |
892 | { | |
424fd7c3 | 893 | ASSERT(zfs_refcount_count(&spa->spa_refcount) > spa->spa_minref || |
34dc7c2f | 894 | MUTEX_HELD(&spa_namespace_lock)); |
424fd7c3 | 895 | (void) zfs_refcount_remove(&spa->spa_refcount, tag); |
34dc7c2f BB |
896 | } |
897 | ||
0c66c32d JG |
898 | /* |
899 | * Remove a reference to the given spa_t held by a dsl dir that is | |
900 | * being asynchronously released. Async releases occur from a taskq | |
901 | * performing eviction of dsl datasets and dirs. The namespace lock | |
902 | * isn't held and the hold by the object being evicted may contribute to | |
903 | * spa_minref (e.g. dataset or directory released during pool export), | |
904 | * so the asserts in spa_close() do not apply. | |
905 | */ | |
906 | void | |
907 | spa_async_close(spa_t *spa, void *tag) | |
908 | { | |
424fd7c3 | 909 | (void) zfs_refcount_remove(&spa->spa_refcount, tag); |
0c66c32d JG |
910 | } |
911 | ||
34dc7c2f BB |
912 | /* |
913 | * Check to see if the spa refcount is zero. Must be called with | |
b128c09f | 914 | * spa_namespace_lock held. We really compare against spa_minref, which is the |
34dc7c2f BB |
915 | * number of references acquired when opening a pool |
916 | */ | |
917 | boolean_t | |
918 | spa_refcount_zero(spa_t *spa) | |
919 | { | |
920 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
921 | ||
424fd7c3 | 922 | return (zfs_refcount_count(&spa->spa_refcount) == spa->spa_minref); |
34dc7c2f BB |
923 | } |
924 | ||
925 | /* | |
926 | * ========================================================================== | |
927 | * SPA spare and l2cache tracking | |
928 | * ========================================================================== | |
929 | */ | |
930 | ||
931 | /* | |
932 | * Hot spares and cache devices are tracked using the same code below, | |
933 | * for 'auxiliary' devices. | |
934 | */ | |
935 | ||
936 | typedef struct spa_aux { | |
937 | uint64_t aux_guid; | |
938 | uint64_t aux_pool; | |
939 | avl_node_t aux_avl; | |
940 | int aux_count; | |
941 | } spa_aux_t; | |
942 | ||
ee36c709 | 943 | static inline int |
34dc7c2f BB |
944 | spa_aux_compare(const void *a, const void *b) |
945 | { | |
ee36c709 GN |
946 | const spa_aux_t *sa = (const spa_aux_t *)a; |
947 | const spa_aux_t *sb = (const spa_aux_t *)b; | |
34dc7c2f | 948 | |
ca577779 | 949 | return (TREE_CMP(sa->aux_guid, sb->aux_guid)); |
34dc7c2f BB |
950 | } |
951 | ||
65c7cc49 | 952 | static void |
34dc7c2f BB |
953 | spa_aux_add(vdev_t *vd, avl_tree_t *avl) |
954 | { | |
955 | avl_index_t where; | |
956 | spa_aux_t search; | |
957 | spa_aux_t *aux; | |
958 | ||
959 | search.aux_guid = vd->vdev_guid; | |
960 | if ((aux = avl_find(avl, &search, &where)) != NULL) { | |
961 | aux->aux_count++; | |
962 | } else { | |
79c76d5b | 963 | aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP); |
34dc7c2f BB |
964 | aux->aux_guid = vd->vdev_guid; |
965 | aux->aux_count = 1; | |
966 | avl_insert(avl, aux, where); | |
967 | } | |
968 | } | |
969 | ||
65c7cc49 | 970 | static void |
34dc7c2f BB |
971 | spa_aux_remove(vdev_t *vd, avl_tree_t *avl) |
972 | { | |
973 | spa_aux_t search; | |
974 | spa_aux_t *aux; | |
975 | avl_index_t where; | |
976 | ||
977 | search.aux_guid = vd->vdev_guid; | |
978 | aux = avl_find(avl, &search, &where); | |
979 | ||
980 | ASSERT(aux != NULL); | |
981 | ||
982 | if (--aux->aux_count == 0) { | |
983 | avl_remove(avl, aux); | |
984 | kmem_free(aux, sizeof (spa_aux_t)); | |
985 | } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { | |
986 | aux->aux_pool = 0ULL; | |
987 | } | |
988 | } | |
989 | ||
65c7cc49 | 990 | static boolean_t |
b128c09f | 991 | spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl) |
34dc7c2f BB |
992 | { |
993 | spa_aux_t search, *found; | |
34dc7c2f BB |
994 | |
995 | search.aux_guid = guid; | |
b128c09f | 996 | found = avl_find(avl, &search, NULL); |
34dc7c2f BB |
997 | |
998 | if (pool) { | |
999 | if (found) | |
1000 | *pool = found->aux_pool; | |
1001 | else | |
1002 | *pool = 0ULL; | |
1003 | } | |
1004 | ||
b128c09f BB |
1005 | if (refcnt) { |
1006 | if (found) | |
1007 | *refcnt = found->aux_count; | |
1008 | else | |
1009 | *refcnt = 0; | |
1010 | } | |
1011 | ||
34dc7c2f BB |
1012 | return (found != NULL); |
1013 | } | |
1014 | ||
65c7cc49 | 1015 | static void |
34dc7c2f BB |
1016 | spa_aux_activate(vdev_t *vd, avl_tree_t *avl) |
1017 | { | |
1018 | spa_aux_t search, *found; | |
1019 | avl_index_t where; | |
1020 | ||
1021 | search.aux_guid = vd->vdev_guid; | |
1022 | found = avl_find(avl, &search, &where); | |
1023 | ASSERT(found != NULL); | |
1024 | ASSERT(found->aux_pool == 0ULL); | |
1025 | ||
1026 | found->aux_pool = spa_guid(vd->vdev_spa); | |
1027 | } | |
1028 | ||
1029 | /* | |
1030 | * Spares are tracked globally due to the following constraints: | |
1031 | * | |
aa755b35 MA |
1032 | * - A spare may be part of multiple pools. |
1033 | * - A spare may be added to a pool even if it's actively in use within | |
34dc7c2f | 1034 | * another pool. |
aa755b35 | 1035 | * - A spare in use in any pool can only be the source of a replacement if |
34dc7c2f BB |
1036 | * the target is a spare in the same pool. |
1037 | * | |
1038 | * We keep track of all spares on the system through the use of a reference | |
1039 | * counted AVL tree. When a vdev is added as a spare, or used as a replacement | |
1040 | * spare, then we bump the reference count in the AVL tree. In addition, we set | |
1041 | * the 'vdev_isspare' member to indicate that the device is a spare (active or | |
1042 | * inactive). When a spare is made active (used to replace a device in the | |
1043 | * pool), we also keep track of which pool its been made a part of. | |
1044 | * | |
1045 | * The 'spa_spare_lock' protects the AVL tree. These functions are normally | |
1046 | * called under the spa_namespace lock as part of vdev reconfiguration. The | |
1047 | * separate spare lock exists for the status query path, which does not need to | |
1048 | * be completely consistent with respect to other vdev configuration changes. | |
1049 | */ | |
1050 | ||
1051 | static int | |
1052 | spa_spare_compare(const void *a, const void *b) | |
1053 | { | |
1054 | return (spa_aux_compare(a, b)); | |
1055 | } | |
1056 | ||
1057 | void | |
1058 | spa_spare_add(vdev_t *vd) | |
1059 | { | |
1060 | mutex_enter(&spa_spare_lock); | |
1061 | ASSERT(!vd->vdev_isspare); | |
1062 | spa_aux_add(vd, &spa_spare_avl); | |
1063 | vd->vdev_isspare = B_TRUE; | |
1064 | mutex_exit(&spa_spare_lock); | |
1065 | } | |
1066 | ||
1067 | void | |
1068 | spa_spare_remove(vdev_t *vd) | |
1069 | { | |
1070 | mutex_enter(&spa_spare_lock); | |
1071 | ASSERT(vd->vdev_isspare); | |
1072 | spa_aux_remove(vd, &spa_spare_avl); | |
1073 | vd->vdev_isspare = B_FALSE; | |
1074 | mutex_exit(&spa_spare_lock); | |
1075 | } | |
1076 | ||
1077 | boolean_t | |
b128c09f | 1078 | spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt) |
34dc7c2f BB |
1079 | { |
1080 | boolean_t found; | |
1081 | ||
1082 | mutex_enter(&spa_spare_lock); | |
b128c09f | 1083 | found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl); |
34dc7c2f BB |
1084 | mutex_exit(&spa_spare_lock); |
1085 | ||
1086 | return (found); | |
1087 | } | |
1088 | ||
1089 | void | |
1090 | spa_spare_activate(vdev_t *vd) | |
1091 | { | |
1092 | mutex_enter(&spa_spare_lock); | |
1093 | ASSERT(vd->vdev_isspare); | |
1094 | spa_aux_activate(vd, &spa_spare_avl); | |
1095 | mutex_exit(&spa_spare_lock); | |
1096 | } | |
1097 | ||
1098 | /* | |
1099 | * Level 2 ARC devices are tracked globally for the same reasons as spares. | |
1100 | * Cache devices currently only support one pool per cache device, and so | |
1101 | * for these devices the aux reference count is currently unused beyond 1. | |
1102 | */ | |
1103 | ||
1104 | static int | |
1105 | spa_l2cache_compare(const void *a, const void *b) | |
1106 | { | |
1107 | return (spa_aux_compare(a, b)); | |
1108 | } | |
1109 | ||
1110 | void | |
1111 | spa_l2cache_add(vdev_t *vd) | |
1112 | { | |
1113 | mutex_enter(&spa_l2cache_lock); | |
1114 | ASSERT(!vd->vdev_isl2cache); | |
1115 | spa_aux_add(vd, &spa_l2cache_avl); | |
1116 | vd->vdev_isl2cache = B_TRUE; | |
1117 | mutex_exit(&spa_l2cache_lock); | |
1118 | } | |
1119 | ||
1120 | void | |
1121 | spa_l2cache_remove(vdev_t *vd) | |
1122 | { | |
1123 | mutex_enter(&spa_l2cache_lock); | |
1124 | ASSERT(vd->vdev_isl2cache); | |
1125 | spa_aux_remove(vd, &spa_l2cache_avl); | |
1126 | vd->vdev_isl2cache = B_FALSE; | |
1127 | mutex_exit(&spa_l2cache_lock); | |
1128 | } | |
1129 | ||
1130 | boolean_t | |
1131 | spa_l2cache_exists(uint64_t guid, uint64_t *pool) | |
1132 | { | |
1133 | boolean_t found; | |
1134 | ||
1135 | mutex_enter(&spa_l2cache_lock); | |
b128c09f | 1136 | found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl); |
34dc7c2f BB |
1137 | mutex_exit(&spa_l2cache_lock); |
1138 | ||
1139 | return (found); | |
1140 | } | |
1141 | ||
1142 | void | |
1143 | spa_l2cache_activate(vdev_t *vd) | |
1144 | { | |
1145 | mutex_enter(&spa_l2cache_lock); | |
1146 | ASSERT(vd->vdev_isl2cache); | |
1147 | spa_aux_activate(vd, &spa_l2cache_avl); | |
1148 | mutex_exit(&spa_l2cache_lock); | |
1149 | } | |
1150 | ||
34dc7c2f BB |
1151 | /* |
1152 | * ========================================================================== | |
1153 | * SPA vdev locking | |
1154 | * ========================================================================== | |
1155 | */ | |
1156 | ||
1157 | /* | |
1158 | * Lock the given spa_t for the purpose of adding or removing a vdev. | |
1159 | * Grabs the global spa_namespace_lock plus the spa config lock for writing. | |
1160 | * It returns the next transaction group for the spa_t. | |
1161 | */ | |
1162 | uint64_t | |
1163 | spa_vdev_enter(spa_t *spa) | |
1164 | { | |
428870ff | 1165 | mutex_enter(&spa->spa_vdev_top_lock); |
34dc7c2f | 1166 | mutex_enter(&spa_namespace_lock); |
1b939560 BB |
1167 | |
1168 | vdev_autotrim_stop_all(spa); | |
1169 | ||
428870ff BB |
1170 | return (spa_vdev_config_enter(spa)); |
1171 | } | |
1172 | ||
9a49d3f3 BB |
1173 | /* |
1174 | * The same as spa_vdev_enter() above but additionally takes the guid of | |
1175 | * the vdev being detached. When there is a rebuild in process it will be | |
1176 | * suspended while the vdev tree is modified then resumed by spa_vdev_exit(). | |
1177 | * The rebuild is canceled if only a single child remains after the detach. | |
1178 | */ | |
1179 | uint64_t | |
1180 | spa_vdev_detach_enter(spa_t *spa, uint64_t guid) | |
1181 | { | |
1182 | mutex_enter(&spa->spa_vdev_top_lock); | |
1183 | mutex_enter(&spa_namespace_lock); | |
1184 | ||
1185 | vdev_autotrim_stop_all(spa); | |
1186 | ||
1187 | if (guid != 0) { | |
1188 | vdev_t *vd = spa_lookup_by_guid(spa, guid, B_FALSE); | |
1189 | if (vd) { | |
1190 | vdev_rebuild_stop_wait(vd->vdev_top); | |
1191 | } | |
1192 | } | |
1193 | ||
1194 | return (spa_vdev_config_enter(spa)); | |
1195 | } | |
1196 | ||
428870ff BB |
1197 | /* |
1198 | * Internal implementation for spa_vdev_enter(). Used when a vdev | |
1199 | * operation requires multiple syncs (i.e. removing a device) while | |
1200 | * keeping the spa_namespace_lock held. | |
1201 | */ | |
1202 | uint64_t | |
1203 | spa_vdev_config_enter(spa_t *spa) | |
1204 | { | |
1205 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
34dc7c2f | 1206 | |
b128c09f | 1207 | spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); |
34dc7c2f BB |
1208 | |
1209 | return (spa_last_synced_txg(spa) + 1); | |
1210 | } | |
1211 | ||
1212 | /* | |
428870ff BB |
1213 | * Used in combination with spa_vdev_config_enter() to allow the syncing |
1214 | * of multiple transactions without releasing the spa_namespace_lock. | |
34dc7c2f | 1215 | */ |
428870ff BB |
1216 | void |
1217 | spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag) | |
34dc7c2f | 1218 | { |
1c27024e DB |
1219 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); |
1220 | ||
34dc7c2f BB |
1221 | int config_changed = B_FALSE; |
1222 | ||
1223 | ASSERT(txg > spa_last_synced_txg(spa)); | |
1224 | ||
b128c09f BB |
1225 | spa->spa_pending_vdev = NULL; |
1226 | ||
34dc7c2f BB |
1227 | /* |
1228 | * Reassess the DTLs. | |
1229 | */ | |
9a49d3f3 | 1230 | vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE, B_FALSE); |
34dc7c2f | 1231 | |
b128c09f | 1232 | if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) { |
34dc7c2f | 1233 | config_changed = B_TRUE; |
428870ff | 1234 | spa->spa_config_generation++; |
34dc7c2f BB |
1235 | } |
1236 | ||
428870ff BB |
1237 | /* |
1238 | * Verify the metaslab classes. | |
1239 | */ | |
1240 | ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0); | |
1241 | ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0); | |
aa755b35 | 1242 | ASSERT(metaslab_class_validate(spa_embedded_log_class(spa)) == 0); |
cc99f275 DB |
1243 | ASSERT(metaslab_class_validate(spa_special_class(spa)) == 0); |
1244 | ASSERT(metaslab_class_validate(spa_dedup_class(spa)) == 0); | |
428870ff | 1245 | |
b128c09f | 1246 | spa_config_exit(spa, SCL_ALL, spa); |
34dc7c2f | 1247 | |
428870ff BB |
1248 | /* |
1249 | * Panic the system if the specified tag requires it. This | |
1250 | * is useful for ensuring that configurations are updated | |
1251 | * transactionally. | |
1252 | */ | |
1253 | if (zio_injection_enabled) | |
1254 | zio_handle_panic_injection(spa, tag, 0); | |
1255 | ||
34dc7c2f BB |
1256 | /* |
1257 | * Note: this txg_wait_synced() is important because it ensures | |
1258 | * that there won't be more than one config change per txg. | |
1259 | * This allows us to use the txg as the generation number. | |
1260 | */ | |
1261 | if (error == 0) | |
1262 | txg_wait_synced(spa->spa_dsl_pool, txg); | |
1263 | ||
1264 | if (vd != NULL) { | |
93cf2076 | 1265 | ASSERT(!vd->vdev_detached || vd->vdev_dtl_sm == NULL); |
619f0976 GW |
1266 | if (vd->vdev_ops->vdev_op_leaf) { |
1267 | mutex_enter(&vd->vdev_initialize_lock); | |
c10d37dd GW |
1268 | vdev_initialize_stop(vd, VDEV_INITIALIZE_CANCELED, |
1269 | NULL); | |
619f0976 | 1270 | mutex_exit(&vd->vdev_initialize_lock); |
1b939560 BB |
1271 | |
1272 | mutex_enter(&vd->vdev_trim_lock); | |
1273 | vdev_trim_stop(vd, VDEV_TRIM_CANCELED, NULL); | |
1274 | mutex_exit(&vd->vdev_trim_lock); | |
619f0976 GW |
1275 | } |
1276 | ||
1b939560 BB |
1277 | /* |
1278 | * The vdev may be both a leaf and top-level device. | |
1279 | */ | |
1280 | vdev_autotrim_stop_wait(vd); | |
1281 | ||
75a089ed | 1282 | spa_config_enter(spa, SCL_STATE_ALL, spa, RW_WRITER); |
34dc7c2f | 1283 | vdev_free(vd); |
75a089ed | 1284 | spa_config_exit(spa, SCL_STATE_ALL, spa); |
34dc7c2f BB |
1285 | } |
1286 | ||
1287 | /* | |
1288 | * If the config changed, update the config cache. | |
1289 | */ | |
1290 | if (config_changed) | |
a1d477c2 | 1291 | spa_write_cachefile(spa, B_FALSE, B_TRUE); |
428870ff | 1292 | } |
34dc7c2f | 1293 | |
428870ff BB |
1294 | /* |
1295 | * Unlock the spa_t after adding or removing a vdev. Besides undoing the | |
1296 | * locking of spa_vdev_enter(), we also want make sure the transactions have | |
1297 | * synced to disk, and then update the global configuration cache with the new | |
1298 | * information. | |
1299 | */ | |
1300 | int | |
1301 | spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) | |
1302 | { | |
1b939560 | 1303 | vdev_autotrim_restart(spa); |
9a49d3f3 | 1304 | vdev_rebuild_restart(spa); |
1b939560 | 1305 | |
428870ff | 1306 | spa_vdev_config_exit(spa, vd, txg, error, FTAG); |
34dc7c2f | 1307 | mutex_exit(&spa_namespace_lock); |
428870ff | 1308 | mutex_exit(&spa->spa_vdev_top_lock); |
34dc7c2f BB |
1309 | |
1310 | return (error); | |
1311 | } | |
1312 | ||
b128c09f BB |
1313 | /* |
1314 | * Lock the given spa_t for the purpose of changing vdev state. | |
1315 | */ | |
1316 | void | |
428870ff | 1317 | spa_vdev_state_enter(spa_t *spa, int oplocks) |
b128c09f | 1318 | { |
428870ff BB |
1319 | int locks = SCL_STATE_ALL | oplocks; |
1320 | ||
1321 | /* | |
1322 | * Root pools may need to read of the underlying devfs filesystem | |
1323 | * when opening up a vdev. Unfortunately if we're holding the | |
1324 | * SCL_ZIO lock it will result in a deadlock when we try to issue | |
1325 | * the read from the root filesystem. Instead we "prefetch" | |
1326 | * the associated vnodes that we need prior to opening the | |
1327 | * underlying devices and cache them so that we can prevent | |
1328 | * any I/O when we are doing the actual open. | |
1329 | */ | |
1330 | if (spa_is_root(spa)) { | |
1331 | int low = locks & ~(SCL_ZIO - 1); | |
1332 | int high = locks & ~low; | |
1333 | ||
1334 | spa_config_enter(spa, high, spa, RW_WRITER); | |
1335 | vdev_hold(spa->spa_root_vdev); | |
1336 | spa_config_enter(spa, low, spa, RW_WRITER); | |
1337 | } else { | |
1338 | spa_config_enter(spa, locks, spa, RW_WRITER); | |
1339 | } | |
1340 | spa->spa_vdev_locks = locks; | |
b128c09f BB |
1341 | } |
1342 | ||
1343 | int | |
1344 | spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error) | |
1345 | { | |
428870ff | 1346 | boolean_t config_changed = B_FALSE; |
4a283c7f TH |
1347 | vdev_t *vdev_top; |
1348 | ||
1349 | if (vd == NULL || vd == spa->spa_root_vdev) { | |
1350 | vdev_top = spa->spa_root_vdev; | |
1351 | } else { | |
1352 | vdev_top = vd->vdev_top; | |
1353 | } | |
428870ff BB |
1354 | |
1355 | if (vd != NULL || error == 0) | |
9a49d3f3 | 1356 | vdev_dtl_reassess(vdev_top, 0, 0, B_FALSE, B_FALSE); |
428870ff BB |
1357 | |
1358 | if (vd != NULL) { | |
4a283c7f TH |
1359 | if (vd != spa->spa_root_vdev) |
1360 | vdev_state_dirty(vdev_top); | |
1361 | ||
428870ff BB |
1362 | config_changed = B_TRUE; |
1363 | spa->spa_config_generation++; | |
1364 | } | |
b128c09f | 1365 | |
428870ff BB |
1366 | if (spa_is_root(spa)) |
1367 | vdev_rele(spa->spa_root_vdev); | |
1368 | ||
1369 | ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL); | |
1370 | spa_config_exit(spa, spa->spa_vdev_locks, spa); | |
b128c09f | 1371 | |
fb5f0bc8 BB |
1372 | /* |
1373 | * If anything changed, wait for it to sync. This ensures that, | |
76d04993 | 1374 | * from the system administrator's perspective, zpool(8) commands |
fb5f0bc8 BB |
1375 | * are synchronous. This is important for things like zpool offline: |
1376 | * when the command completes, you expect no further I/O from ZFS. | |
1377 | */ | |
1378 | if (vd != NULL) | |
1379 | txg_wait_synced(spa->spa_dsl_pool, 0); | |
1380 | ||
428870ff BB |
1381 | /* |
1382 | * If the config changed, update the config cache. | |
1383 | */ | |
1384 | if (config_changed) { | |
1385 | mutex_enter(&spa_namespace_lock); | |
a1d477c2 | 1386 | spa_write_cachefile(spa, B_FALSE, B_TRUE); |
428870ff BB |
1387 | mutex_exit(&spa_namespace_lock); |
1388 | } | |
1389 | ||
b128c09f BB |
1390 | return (error); |
1391 | } | |
1392 | ||
34dc7c2f BB |
1393 | /* |
1394 | * ========================================================================== | |
1395 | * Miscellaneous functions | |
1396 | * ========================================================================== | |
1397 | */ | |
1398 | ||
9ae529ec | 1399 | void |
b0bc7a84 | 1400 | spa_activate_mos_feature(spa_t *spa, const char *feature, dmu_tx_t *tx) |
9ae529ec | 1401 | { |
fa86b5db MA |
1402 | if (!nvlist_exists(spa->spa_label_features, feature)) { |
1403 | fnvlist_add_boolean(spa->spa_label_features, feature); | |
b0bc7a84 MG |
1404 | /* |
1405 | * When we are creating the pool (tx_txg==TXG_INITIAL), we can't | |
1406 | * dirty the vdev config because lock SCL_CONFIG is not held. | |
1407 | * Thankfully, in this case we don't need to dirty the config | |
1408 | * because it will be written out anyway when we finish | |
1409 | * creating the pool. | |
1410 | */ | |
1411 | if (tx->tx_txg != TXG_INITIAL) | |
1412 | vdev_config_dirty(spa->spa_root_vdev); | |
fa86b5db | 1413 | } |
9ae529ec CS |
1414 | } |
1415 | ||
1416 | void | |
1417 | spa_deactivate_mos_feature(spa_t *spa, const char *feature) | |
1418 | { | |
fa86b5db MA |
1419 | if (nvlist_remove_all(spa->spa_label_features, feature) == 0) |
1420 | vdev_config_dirty(spa->spa_root_vdev); | |
9ae529ec CS |
1421 | } |
1422 | ||
34dc7c2f | 1423 | /* |
572e2857 BB |
1424 | * Return the spa_t associated with given pool_guid, if it exists. If |
1425 | * device_guid is non-zero, determine whether the pool exists *and* contains | |
1426 | * a device with the specified device_guid. | |
34dc7c2f | 1427 | */ |
572e2857 BB |
1428 | spa_t * |
1429 | spa_by_guid(uint64_t pool_guid, uint64_t device_guid) | |
34dc7c2f BB |
1430 | { |
1431 | spa_t *spa; | |
1432 | avl_tree_t *t = &spa_namespace_avl; | |
1433 | ||
1434 | ASSERT(MUTEX_HELD(&spa_namespace_lock)); | |
1435 | ||
1436 | for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { | |
1437 | if (spa->spa_state == POOL_STATE_UNINITIALIZED) | |
1438 | continue; | |
1439 | if (spa->spa_root_vdev == NULL) | |
1440 | continue; | |
1441 | if (spa_guid(spa) == pool_guid) { | |
1442 | if (device_guid == 0) | |
1443 | break; | |
1444 | ||
1445 | if (vdev_lookup_by_guid(spa->spa_root_vdev, | |
1446 | device_guid) != NULL) | |
1447 | break; | |
1448 | ||
1449 | /* | |
1450 | * Check any devices we may be in the process of adding. | |
1451 | */ | |
1452 | if (spa->spa_pending_vdev) { | |
1453 | if (vdev_lookup_by_guid(spa->spa_pending_vdev, | |
1454 | device_guid) != NULL) | |
1455 | break; | |
1456 | } | |
1457 | } | |
1458 | } | |
1459 | ||
572e2857 BB |
1460 | return (spa); |
1461 | } | |
1462 | ||
1463 | /* | |
1464 | * Determine whether a pool with the given pool_guid exists. | |
1465 | */ | |
1466 | boolean_t | |
1467 | spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) | |
1468 | { | |
1469 | return (spa_by_guid(pool_guid, device_guid) != NULL); | |
34dc7c2f BB |
1470 | } |
1471 | ||
1472 | char * | |
1473 | spa_strdup(const char *s) | |
1474 | { | |
1475 | size_t len; | |
1476 | char *new; | |
1477 | ||
1478 | len = strlen(s); | |
79c76d5b | 1479 | new = kmem_alloc(len + 1, KM_SLEEP); |
34dc7c2f BB |
1480 | bcopy(s, new, len); |
1481 | new[len] = '\0'; | |
1482 | ||
1483 | return (new); | |
1484 | } | |
1485 | ||
1486 | void | |
1487 | spa_strfree(char *s) | |
1488 | { | |
1489 | kmem_free(s, strlen(s) + 1); | |
1490 | } | |
1491 | ||
428870ff BB |
1492 | uint64_t |
1493 | spa_generate_guid(spa_t *spa) | |
34dc7c2f | 1494 | { |
29274c9f | 1495 | uint64_t guid; |
34dc7c2f | 1496 | |
428870ff | 1497 | if (spa != NULL) { |
29274c9f AM |
1498 | do { |
1499 | (void) random_get_pseudo_bytes((void *)&guid, | |
1500 | sizeof (guid)); | |
1501 | } while (guid == 0 || spa_guid_exists(spa_guid(spa), guid)); | |
428870ff | 1502 | } else { |
29274c9f AM |
1503 | do { |
1504 | (void) random_get_pseudo_bytes((void *)&guid, | |
1505 | sizeof (guid)); | |
1506 | } while (guid == 0 || spa_guid_exists(guid, 0)); | |
34dc7c2f BB |
1507 | } |
1508 | ||
428870ff BB |
1509 | return (guid); |
1510 | } | |
1511 | ||
1512 | void | |
b0bc7a84 | 1513 | snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp) |
428870ff | 1514 | { |
9ae529ec | 1515 | char type[256]; |
428870ff BB |
1516 | char *checksum = NULL; |
1517 | char *compress = NULL; | |
34dc7c2f | 1518 | |
428870ff | 1519 | if (bp != NULL) { |
9ae529ec CS |
1520 | if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) { |
1521 | dmu_object_byteswap_t bswap = | |
1522 | DMU_OT_BYTESWAP(BP_GET_TYPE(bp)); | |
1523 | (void) snprintf(type, sizeof (type), "bswap %s %s", | |
1524 | DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ? | |
1525 | "metadata" : "data", | |
1526 | dmu_ot_byteswap[bswap].ob_name); | |
1527 | } else { | |
1528 | (void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name, | |
1529 | sizeof (type)); | |
1530 | } | |
9b67f605 MA |
1531 | if (!BP_IS_EMBEDDED(bp)) { |
1532 | checksum = | |
1533 | zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name; | |
1534 | } | |
428870ff | 1535 | compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name; |
34dc7c2f BB |
1536 | } |
1537 | ||
b0bc7a84 | 1538 | SNPRINTF_BLKPTR(snprintf, ' ', buf, buflen, bp, type, checksum, |
5c27ec10 | 1539 | compress); |
34dc7c2f BB |
1540 | } |
1541 | ||
1542 | void | |
1543 | spa_freeze(spa_t *spa) | |
1544 | { | |
1545 | uint64_t freeze_txg = 0; | |
1546 | ||
b128c09f | 1547 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
34dc7c2f BB |
1548 | if (spa->spa_freeze_txg == UINT64_MAX) { |
1549 | freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; | |
1550 | spa->spa_freeze_txg = freeze_txg; | |
1551 | } | |
b128c09f | 1552 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
1553 | if (freeze_txg != 0) |
1554 | txg_wait_synced(spa_get_dsl(spa), freeze_txg); | |
1555 | } | |
1556 | ||
0b39b9f9 PS |
1557 | void |
1558 | zfs_panic_recover(const char *fmt, ...) | |
1559 | { | |
1560 | va_list adx; | |
1561 | ||
1562 | va_start(adx, fmt); | |
1563 | vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); | |
1564 | va_end(adx); | |
1565 | } | |
1566 | ||
428870ff BB |
1567 | /* |
1568 | * This is a stripped-down version of strtoull, suitable only for converting | |
d3cc8b15 | 1569 | * lowercase hexadecimal numbers that don't overflow. |
428870ff BB |
1570 | */ |
1571 | uint64_t | |
e19572e4 | 1572 | zfs_strtonum(const char *str, char **nptr) |
428870ff BB |
1573 | { |
1574 | uint64_t val = 0; | |
1575 | char c; | |
1576 | int digit; | |
1577 | ||
1578 | while ((c = *str) != '\0') { | |
1579 | if (c >= '0' && c <= '9') | |
1580 | digit = c - '0'; | |
1581 | else if (c >= 'a' && c <= 'f') | |
1582 | digit = 10 + c - 'a'; | |
1583 | else | |
1584 | break; | |
1585 | ||
1586 | val *= 16; | |
1587 | val += digit; | |
1588 | ||
1589 | str++; | |
1590 | } | |
1591 | ||
1592 | if (nptr) | |
1593 | *nptr = (char *)str; | |
1594 | ||
1595 | return (val); | |
1596 | } | |
1597 | ||
cc99f275 DB |
1598 | void |
1599 | spa_activate_allocation_classes(spa_t *spa, dmu_tx_t *tx) | |
1600 | { | |
1601 | /* | |
1602 | * We bump the feature refcount for each special vdev added to the pool | |
1603 | */ | |
1604 | ASSERT(spa_feature_is_enabled(spa, SPA_FEATURE_ALLOCATION_CLASSES)); | |
1605 | spa_feature_incr(spa, SPA_FEATURE_ALLOCATION_CLASSES, tx); | |
1606 | } | |
1607 | ||
34dc7c2f BB |
1608 | /* |
1609 | * ========================================================================== | |
1610 | * Accessor functions | |
1611 | * ========================================================================== | |
1612 | */ | |
1613 | ||
b128c09f BB |
1614 | boolean_t |
1615 | spa_shutting_down(spa_t *spa) | |
34dc7c2f | 1616 | { |
b128c09f | 1617 | return (spa->spa_async_suspended); |
34dc7c2f BB |
1618 | } |
1619 | ||
1620 | dsl_pool_t * | |
1621 | spa_get_dsl(spa_t *spa) | |
1622 | { | |
1623 | return (spa->spa_dsl_pool); | |
1624 | } | |
1625 | ||
9ae529ec CS |
1626 | boolean_t |
1627 | spa_is_initializing(spa_t *spa) | |
1628 | { | |
1629 | return (spa->spa_is_initializing); | |
1630 | } | |
1631 | ||
a1d477c2 MA |
1632 | boolean_t |
1633 | spa_indirect_vdevs_loaded(spa_t *spa) | |
1634 | { | |
1635 | return (spa->spa_indirect_vdevs_loaded); | |
1636 | } | |
1637 | ||
34dc7c2f BB |
1638 | blkptr_t * |
1639 | spa_get_rootblkptr(spa_t *spa) | |
1640 | { | |
1641 | return (&spa->spa_ubsync.ub_rootbp); | |
1642 | } | |
1643 | ||
1644 | void | |
1645 | spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) | |
1646 | { | |
1647 | spa->spa_uberblock.ub_rootbp = *bp; | |
1648 | } | |
1649 | ||
1650 | void | |
1651 | spa_altroot(spa_t *spa, char *buf, size_t buflen) | |
1652 | { | |
1653 | if (spa->spa_root == NULL) | |
1654 | buf[0] = '\0'; | |
1655 | else | |
1656 | (void) strncpy(buf, spa->spa_root, buflen); | |
1657 | } | |
1658 | ||
1659 | int | |
1660 | spa_sync_pass(spa_t *spa) | |
1661 | { | |
1662 | return (spa->spa_sync_pass); | |
1663 | } | |
1664 | ||
1665 | char * | |
1666 | spa_name(spa_t *spa) | |
1667 | { | |
34dc7c2f BB |
1668 | return (spa->spa_name); |
1669 | } | |
1670 | ||
1671 | uint64_t | |
1672 | spa_guid(spa_t *spa) | |
1673 | { | |
3bc7e0fb GW |
1674 | dsl_pool_t *dp = spa_get_dsl(spa); |
1675 | uint64_t guid; | |
1676 | ||
34dc7c2f BB |
1677 | /* |
1678 | * If we fail to parse the config during spa_load(), we can go through | |
1679 | * the error path (which posts an ereport) and end up here with no root | |
3541dc6d | 1680 | * vdev. We stash the original pool guid in 'spa_config_guid' to handle |
34dc7c2f BB |
1681 | * this case. |
1682 | */ | |
3bc7e0fb GW |
1683 | if (spa->spa_root_vdev == NULL) |
1684 | return (spa->spa_config_guid); | |
1685 | ||
1686 | guid = spa->spa_last_synced_guid != 0 ? | |
1687 | spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid; | |
1688 | ||
1689 | /* | |
1690 | * Return the most recently synced out guid unless we're | |
1691 | * in syncing context. | |
1692 | */ | |
1693 | if (dp && dsl_pool_sync_context(dp)) | |
34dc7c2f BB |
1694 | return (spa->spa_root_vdev->vdev_guid); |
1695 | else | |
3bc7e0fb | 1696 | return (guid); |
3541dc6d GA |
1697 | } |
1698 | ||
1699 | uint64_t | |
1700 | spa_load_guid(spa_t *spa) | |
1701 | { | |
1702 | /* | |
1703 | * This is a GUID that exists solely as a reference for the | |
1704 | * purposes of the arc. It is generated at load time, and | |
1705 | * is never written to persistent storage. | |
1706 | */ | |
1707 | return (spa->spa_load_guid); | |
34dc7c2f BB |
1708 | } |
1709 | ||
1710 | uint64_t | |
1711 | spa_last_synced_txg(spa_t *spa) | |
1712 | { | |
1713 | return (spa->spa_ubsync.ub_txg); | |
1714 | } | |
1715 | ||
1716 | uint64_t | |
1717 | spa_first_txg(spa_t *spa) | |
1718 | { | |
1719 | return (spa->spa_first_txg); | |
1720 | } | |
1721 | ||
428870ff BB |
1722 | uint64_t |
1723 | spa_syncing_txg(spa_t *spa) | |
1724 | { | |
1725 | return (spa->spa_syncing_txg); | |
1726 | } | |
1727 | ||
3b7f360c GW |
1728 | /* |
1729 | * Return the last txg where data can be dirtied. The final txgs | |
1730 | * will be used to just clear out any deferred frees that remain. | |
1731 | */ | |
1732 | uint64_t | |
1733 | spa_final_dirty_txg(spa_t *spa) | |
1734 | { | |
1735 | return (spa->spa_final_txg - TXG_DEFER_SIZE); | |
1736 | } | |
1737 | ||
b128c09f | 1738 | pool_state_t |
34dc7c2f BB |
1739 | spa_state(spa_t *spa) |
1740 | { | |
1741 | return (spa->spa_state); | |
1742 | } | |
1743 | ||
428870ff BB |
1744 | spa_load_state_t |
1745 | spa_load_state(spa_t *spa) | |
34dc7c2f | 1746 | { |
428870ff | 1747 | return (spa->spa_load_state); |
34dc7c2f BB |
1748 | } |
1749 | ||
34dc7c2f | 1750 | uint64_t |
428870ff | 1751 | spa_freeze_txg(spa_t *spa) |
34dc7c2f | 1752 | { |
428870ff | 1753 | return (spa->spa_freeze_txg); |
34dc7c2f BB |
1754 | } |
1755 | ||
047187c1 | 1756 | /* |
1757 | * Return the inflated asize for a logical write in bytes. This is used by the | |
1758 | * DMU to calculate the space a logical write will require on disk. | |
1759 | * If lsize is smaller than the largest physical block size allocatable on this | |
1760 | * pool we use its value instead, since the write will end up using the whole | |
1761 | * block anyway. | |
1762 | */ | |
34dc7c2f | 1763 | uint64_t |
3ec3bc21 | 1764 | spa_get_worst_case_asize(spa_t *spa, uint64_t lsize) |
34dc7c2f | 1765 | { |
047187c1 | 1766 | if (lsize == 0) |
1767 | return (0); /* No inflation needed */ | |
1768 | return (MAX(lsize, 1 << spa->spa_max_ashift) * spa_asize_inflation); | |
34dc7c2f BB |
1769 | } |
1770 | ||
3d45fdd6 | 1771 | /* |
aa755b35 MA |
1772 | * Return the amount of slop space in bytes. It is typically 1/32 of the pool |
1773 | * (3.2%), minus the embedded log space. On very small pools, it may be | |
f01eaed4 PS |
1774 | * slightly larger than this. On very large pools, it will be capped to |
1775 | * the value of spa_max_slop. The embedded log space is not included in | |
aa755b35 MA |
1776 | * spa_dspace. By subtracting it, the usable space (per "zfs list") is a |
1777 | * constant 97% of the total space, regardless of metaslab size (assuming the | |
1778 | * default spa_slop_shift=5 and a non-tiny pool). | |
1779 | * | |
1780 | * See the comment above spa_slop_shift for more details. | |
3d45fdd6 MA |
1781 | */ |
1782 | uint64_t | |
4ea3f864 GM |
1783 | spa_get_slop_space(spa_t *spa) |
1784 | { | |
1325434b RE |
1785 | uint64_t space = 0; |
1786 | uint64_t slop = 0; | |
1787 | ||
1788 | /* | |
1789 | * Make sure spa_dedup_dspace has been set. | |
1790 | */ | |
1791 | if (spa->spa_dedup_dspace == ~0ULL) | |
1792 | spa_update_dspace(spa); | |
1793 | ||
1794 | /* | |
1795 | * spa_get_dspace() includes the space only logically "used" by | |
1796 | * deduplicated data, so since it's not useful to reserve more | |
1797 | * space with more deduplicated data, we subtract that out here. | |
1798 | */ | |
1799 | space = spa_get_dspace(spa) - spa->spa_dedup_dspace; | |
1800 | slop = MIN(space >> spa_slop_shift, spa_max_slop); | |
aa755b35 MA |
1801 | |
1802 | /* | |
1803 | * Subtract the embedded log space, but no more than half the (3.2%) | |
1804 | * unusable space. Note, the "no more than half" is only relevant if | |
1805 | * zfs_embedded_slog_min_ms >> spa_slop_shift < 2, which is not true by | |
1806 | * default. | |
1807 | */ | |
1808 | uint64_t embedded_log = | |
1809 | metaslab_class_get_dspace(spa_embedded_log_class(spa)); | |
1810 | slop -= MIN(embedded_log, slop >> 1); | |
1811 | ||
1812 | /* | |
1813 | * Slop space should be at least spa_min_slop, but no more than half | |
1814 | * the entire pool. | |
1815 | */ | |
1816 | slop = MAX(slop, MIN(space >> 1, spa_min_slop)); | |
1817 | return (slop); | |
3d45fdd6 MA |
1818 | } |
1819 | ||
34dc7c2f BB |
1820 | uint64_t |
1821 | spa_get_dspace(spa_t *spa) | |
1822 | { | |
428870ff | 1823 | return (spa->spa_dspace); |
34dc7c2f BB |
1824 | } |
1825 | ||
d2734cce SD |
1826 | uint64_t |
1827 | spa_get_checkpoint_space(spa_t *spa) | |
1828 | { | |
1829 | return (spa->spa_checkpoint_info.sci_dspace); | |
1830 | } | |
1831 | ||
428870ff BB |
1832 | void |
1833 | spa_update_dspace(spa_t *spa) | |
34dc7c2f | 1834 | { |
428870ff BB |
1835 | spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) + |
1836 | ddt_get_dedup_dspace(spa); | |
2a673e76 | 1837 | if (spa->spa_nonallocating_dspace > 0) { |
a1d477c2 | 1838 | /* |
2a673e76 AJ |
1839 | * Subtract the space provided by all non-allocating vdevs that |
1840 | * contribute to dspace. If a file is overwritten, its old | |
1841 | * blocks are freed and new blocks are allocated. If there are | |
1842 | * no snapshots of the file, the available space should remain | |
1843 | * the same. The old blocks could be freed from the | |
1844 | * non-allocating vdev, but the new blocks must be allocated on | |
1845 | * other (allocating) vdevs. By reserving the entire size of | |
1846 | * the non-allocating vdevs (including allocated space), we | |
1847 | * ensure that there will be enough space on the allocating | |
1848 | * vdevs for this file overwrite to succeed. | |
a1d477c2 MA |
1849 | * |
1850 | * Note that the DMU/DSL doesn't actually know or care | |
1851 | * how much space is allocated (it does its own tracking | |
1852 | * of how much space has been logically used). So it | |
1853 | * doesn't matter that the data we are moving may be | |
2a673e76 | 1854 | * allocated twice (on the old device and the new device). |
a1d477c2 | 1855 | */ |
2a673e76 AJ |
1856 | ASSERT3U(spa->spa_dspace, >=, spa->spa_nonallocating_dspace); |
1857 | spa->spa_dspace -= spa->spa_nonallocating_dspace; | |
a1d477c2 | 1858 | } |
34dc7c2f BB |
1859 | } |
1860 | ||
1861 | /* | |
1862 | * Return the failure mode that has been set to this pool. The default | |
1863 | * behavior will be to block all I/Os when a complete failure occurs. | |
1864 | */ | |
8fb1ede1 | 1865 | uint64_t |
34dc7c2f BB |
1866 | spa_get_failmode(spa_t *spa) |
1867 | { | |
1868 | return (spa->spa_failmode); | |
1869 | } | |
1870 | ||
b128c09f BB |
1871 | boolean_t |
1872 | spa_suspended(spa_t *spa) | |
1873 | { | |
cec3a0a1 | 1874 | return (spa->spa_suspended != ZIO_SUSPEND_NONE); |
b128c09f BB |
1875 | } |
1876 | ||
34dc7c2f BB |
1877 | uint64_t |
1878 | spa_version(spa_t *spa) | |
1879 | { | |
1880 | return (spa->spa_ubsync.ub_version); | |
1881 | } | |
1882 | ||
428870ff BB |
1883 | boolean_t |
1884 | spa_deflate(spa_t *spa) | |
1885 | { | |
1886 | return (spa->spa_deflate); | |
1887 | } | |
1888 | ||
1889 | metaslab_class_t * | |
1890 | spa_normal_class(spa_t *spa) | |
1891 | { | |
1892 | return (spa->spa_normal_class); | |
1893 | } | |
1894 | ||
1895 | metaslab_class_t * | |
1896 | spa_log_class(spa_t *spa) | |
1897 | { | |
1898 | return (spa->spa_log_class); | |
1899 | } | |
1900 | ||
aa755b35 MA |
1901 | metaslab_class_t * |
1902 | spa_embedded_log_class(spa_t *spa) | |
1903 | { | |
1904 | return (spa->spa_embedded_log_class); | |
1905 | } | |
1906 | ||
cc99f275 DB |
1907 | metaslab_class_t * |
1908 | spa_special_class(spa_t *spa) | |
1909 | { | |
1910 | return (spa->spa_special_class); | |
1911 | } | |
1912 | ||
1913 | metaslab_class_t * | |
1914 | spa_dedup_class(spa_t *spa) | |
1915 | { | |
1916 | return (spa->spa_dedup_class); | |
1917 | } | |
1918 | ||
1919 | /* | |
1920 | * Locate an appropriate allocation class | |
1921 | */ | |
1922 | metaslab_class_t * | |
1923 | spa_preferred_class(spa_t *spa, uint64_t size, dmu_object_type_t objtype, | |
1924 | uint_t level, uint_t special_smallblk) | |
1925 | { | |
aa755b35 MA |
1926 | /* |
1927 | * ZIL allocations determine their class in zio_alloc_zil(). | |
1928 | */ | |
1929 | ASSERT(objtype != DMU_OT_INTENT_LOG); | |
cc99f275 DB |
1930 | |
1931 | boolean_t has_special_class = spa->spa_special_class->mc_groups != 0; | |
1932 | ||
1933 | if (DMU_OT_IS_DDT(objtype)) { | |
1934 | if (spa->spa_dedup_class->mc_groups != 0) | |
1935 | return (spa_dedup_class(spa)); | |
1936 | else if (has_special_class && zfs_ddt_data_is_special) | |
1937 | return (spa_special_class(spa)); | |
1938 | else | |
1939 | return (spa_normal_class(spa)); | |
1940 | } | |
1941 | ||
1942 | /* Indirect blocks for user data can land in special if allowed */ | |
1943 | if (level > 0 && (DMU_OT_IS_FILE(objtype) || objtype == DMU_OT_ZVOL)) { | |
1944 | if (has_special_class && zfs_user_indirect_is_special) | |
1945 | return (spa_special_class(spa)); | |
1946 | else | |
1947 | return (spa_normal_class(spa)); | |
1948 | } | |
1949 | ||
1950 | if (DMU_OT_IS_METADATA(objtype) || level > 0) { | |
1951 | if (has_special_class) | |
1952 | return (spa_special_class(spa)); | |
1953 | else | |
1954 | return (spa_normal_class(spa)); | |
1955 | } | |
1956 | ||
1957 | /* | |
1958 | * Allow small file blocks in special class in some cases (like | |
1959 | * for the dRAID vdev feature). But always leave a reserve of | |
1960 | * zfs_special_class_metadata_reserve_pct exclusively for metadata. | |
1961 | */ | |
1962 | if (DMU_OT_IS_FILE(objtype) && | |
44170969 | 1963 | has_special_class && size <= special_smallblk) { |
cc99f275 DB |
1964 | metaslab_class_t *special = spa_special_class(spa); |
1965 | uint64_t alloc = metaslab_class_get_alloc(special); | |
1966 | uint64_t space = metaslab_class_get_space(special); | |
1967 | uint64_t limit = | |
1968 | (space * (100 - zfs_special_class_metadata_reserve_pct)) | |
1969 | / 100; | |
1970 | ||
1971 | if (alloc < limit) | |
1972 | return (special); | |
1973 | } | |
1974 | ||
1975 | return (spa_normal_class(spa)); | |
1976 | } | |
1977 | ||
0c66c32d JG |
1978 | void |
1979 | spa_evicting_os_register(spa_t *spa, objset_t *os) | |
1980 | { | |
1981 | mutex_enter(&spa->spa_evicting_os_lock); | |
1982 | list_insert_head(&spa->spa_evicting_os_list, os); | |
1983 | mutex_exit(&spa->spa_evicting_os_lock); | |
1984 | } | |
1985 | ||
1986 | void | |
1987 | spa_evicting_os_deregister(spa_t *spa, objset_t *os) | |
1988 | { | |
1989 | mutex_enter(&spa->spa_evicting_os_lock); | |
1990 | list_remove(&spa->spa_evicting_os_list, os); | |
1991 | cv_broadcast(&spa->spa_evicting_os_cv); | |
1992 | mutex_exit(&spa->spa_evicting_os_lock); | |
1993 | } | |
1994 | ||
1995 | void | |
1996 | spa_evicting_os_wait(spa_t *spa) | |
1997 | { | |
1998 | mutex_enter(&spa->spa_evicting_os_lock); | |
1999 | while (!list_is_empty(&spa->spa_evicting_os_list)) | |
2000 | cv_wait(&spa->spa_evicting_os_cv, &spa->spa_evicting_os_lock); | |
2001 | mutex_exit(&spa->spa_evicting_os_lock); | |
2002 | ||
2003 | dmu_buf_user_evict_wait(); | |
2004 | } | |
2005 | ||
34dc7c2f BB |
2006 | int |
2007 | spa_max_replication(spa_t *spa) | |
2008 | { | |
2009 | /* | |
2010 | * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to | |
2011 | * handle BPs with more than one DVA allocated. Set our max | |
2012 | * replication level accordingly. | |
2013 | */ | |
2014 | if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) | |
2015 | return (1); | |
2016 | return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); | |
2017 | } | |
2018 | ||
428870ff BB |
2019 | int |
2020 | spa_prev_software_version(spa_t *spa) | |
2021 | { | |
2022 | return (spa->spa_prev_software_version); | |
2023 | } | |
2024 | ||
cc92e9d0 GW |
2025 | uint64_t |
2026 | spa_deadman_synctime(spa_t *spa) | |
2027 | { | |
2028 | return (spa->spa_deadman_synctime); | |
2029 | } | |
2030 | ||
1b939560 BB |
2031 | spa_autotrim_t |
2032 | spa_get_autotrim(spa_t *spa) | |
2033 | { | |
2034 | return (spa->spa_autotrim); | |
2035 | } | |
2036 | ||
8fb1ede1 BB |
2037 | uint64_t |
2038 | spa_deadman_ziotime(spa_t *spa) | |
2039 | { | |
2040 | return (spa->spa_deadman_ziotime); | |
2041 | } | |
2042 | ||
2043 | uint64_t | |
2044 | spa_get_deadman_failmode(spa_t *spa) | |
2045 | { | |
2046 | return (spa->spa_deadman_failmode); | |
2047 | } | |
2048 | ||
2049 | void | |
2050 | spa_set_deadman_failmode(spa_t *spa, const char *failmode) | |
2051 | { | |
2052 | if (strcmp(failmode, "wait") == 0) | |
2053 | spa->spa_deadman_failmode = ZIO_FAILURE_MODE_WAIT; | |
2054 | else if (strcmp(failmode, "continue") == 0) | |
2055 | spa->spa_deadman_failmode = ZIO_FAILURE_MODE_CONTINUE; | |
2056 | else if (strcmp(failmode, "panic") == 0) | |
2057 | spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC; | |
2058 | else | |
2059 | spa->spa_deadman_failmode = ZIO_FAILURE_MODE_WAIT; | |
2060 | } | |
2061 | ||
57940b43 RM |
2062 | void |
2063 | spa_set_deadman_ziotime(hrtime_t ns) | |
2064 | { | |
2065 | spa_t *spa = NULL; | |
2066 | ||
2067 | if (spa_mode_global != SPA_MODE_UNINIT) { | |
2068 | mutex_enter(&spa_namespace_lock); | |
2069 | while ((spa = spa_next(spa)) != NULL) | |
2070 | spa->spa_deadman_ziotime = ns; | |
2071 | mutex_exit(&spa_namespace_lock); | |
2072 | } | |
2073 | } | |
2074 | ||
2075 | void | |
2076 | spa_set_deadman_synctime(hrtime_t ns) | |
2077 | { | |
2078 | spa_t *spa = NULL; | |
2079 | ||
2080 | if (spa_mode_global != SPA_MODE_UNINIT) { | |
2081 | mutex_enter(&spa_namespace_lock); | |
2082 | while ((spa = spa_next(spa)) != NULL) | |
2083 | spa->spa_deadman_synctime = ns; | |
2084 | mutex_exit(&spa_namespace_lock); | |
2085 | } | |
2086 | } | |
2087 | ||
34dc7c2f | 2088 | uint64_t |
428870ff | 2089 | dva_get_dsize_sync(spa_t *spa, const dva_t *dva) |
34dc7c2f | 2090 | { |
428870ff BB |
2091 | uint64_t asize = DVA_GET_ASIZE(dva); |
2092 | uint64_t dsize = asize; | |
34dc7c2f | 2093 | |
428870ff | 2094 | ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); |
34dc7c2f | 2095 | |
428870ff BB |
2096 | if (asize != 0 && spa->spa_deflate) { |
2097 | vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva)); | |
2c33b912 BB |
2098 | if (vd != NULL) |
2099 | dsize = (asize >> SPA_MINBLOCKSHIFT) * | |
2100 | vd->vdev_deflate_ratio; | |
34dc7c2f | 2101 | } |
428870ff BB |
2102 | |
2103 | return (dsize); | |
2104 | } | |
2105 | ||
2106 | uint64_t | |
2107 | bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp) | |
2108 | { | |
2109 | uint64_t dsize = 0; | |
2110 | ||
1c27024e | 2111 | for (int d = 0; d < BP_GET_NDVAS(bp); d++) |
428870ff BB |
2112 | dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); |
2113 | ||
2114 | return (dsize); | |
2115 | } | |
2116 | ||
2117 | uint64_t | |
2118 | bp_get_dsize(spa_t *spa, const blkptr_t *bp) | |
2119 | { | |
2120 | uint64_t dsize = 0; | |
2121 | ||
2122 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2123 | ||
1c27024e | 2124 | for (int d = 0; d < BP_GET_NDVAS(bp); d++) |
428870ff BB |
2125 | dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); |
2126 | ||
b128c09f | 2127 | spa_config_exit(spa, SCL_VDEV, FTAG); |
428870ff BB |
2128 | |
2129 | return (dsize); | |
34dc7c2f BB |
2130 | } |
2131 | ||
dae3e9ea DB |
2132 | uint64_t |
2133 | spa_dirty_data(spa_t *spa) | |
2134 | { | |
2135 | return (spa->spa_dsl_pool->dp_dirty_total); | |
2136 | } | |
2137 | ||
ca95f70d OF |
2138 | /* |
2139 | * ========================================================================== | |
2140 | * SPA Import Progress Routines | |
2141 | * ========================================================================== | |
2142 | */ | |
2143 | ||
2144 | typedef struct spa_import_progress { | |
2145 | uint64_t pool_guid; /* unique id for updates */ | |
2146 | char *pool_name; | |
2147 | spa_load_state_t spa_load_state; | |
2148 | uint64_t mmp_sec_remaining; /* MMP activity check */ | |
2149 | uint64_t spa_load_max_txg; /* rewind txg */ | |
2150 | procfs_list_node_t smh_node; | |
2151 | } spa_import_progress_t; | |
2152 | ||
2153 | spa_history_list_t *spa_import_progress_list = NULL; | |
2154 | ||
2155 | static int | |
2156 | spa_import_progress_show_header(struct seq_file *f) | |
2157 | { | |
2158 | seq_printf(f, "%-20s %-14s %-14s %-12s %s\n", "pool_guid", | |
2159 | "load_state", "multihost_secs", "max_txg", | |
2160 | "pool_name"); | |
2161 | return (0); | |
2162 | } | |
2163 | ||
2164 | static int | |
2165 | spa_import_progress_show(struct seq_file *f, void *data) | |
2166 | { | |
2167 | spa_import_progress_t *sip = (spa_import_progress_t *)data; | |
2168 | ||
2169 | seq_printf(f, "%-20llu %-14llu %-14llu %-12llu %s\n", | |
2170 | (u_longlong_t)sip->pool_guid, (u_longlong_t)sip->spa_load_state, | |
2171 | (u_longlong_t)sip->mmp_sec_remaining, | |
2172 | (u_longlong_t)sip->spa_load_max_txg, | |
2173 | (sip->pool_name ? sip->pool_name : "-")); | |
2174 | ||
2175 | return (0); | |
2176 | } | |
2177 | ||
2178 | /* Remove oldest elements from list until there are no more than 'size' left */ | |
2179 | static void | |
2180 | spa_import_progress_truncate(spa_history_list_t *shl, unsigned int size) | |
2181 | { | |
2182 | spa_import_progress_t *sip; | |
2183 | while (shl->size > size) { | |
2184 | sip = list_remove_head(&shl->procfs_list.pl_list); | |
2185 | if (sip->pool_name) | |
2186 | spa_strfree(sip->pool_name); | |
2187 | kmem_free(sip, sizeof (spa_import_progress_t)); | |
2188 | shl->size--; | |
2189 | } | |
2190 | ||
2191 | IMPLY(size == 0, list_is_empty(&shl->procfs_list.pl_list)); | |
2192 | } | |
2193 | ||
2194 | static void | |
2195 | spa_import_progress_init(void) | |
2196 | { | |
2197 | spa_import_progress_list = kmem_zalloc(sizeof (spa_history_list_t), | |
2198 | KM_SLEEP); | |
2199 | ||
2200 | spa_import_progress_list->size = 0; | |
2201 | ||
2202 | spa_import_progress_list->procfs_list.pl_private = | |
2203 | spa_import_progress_list; | |
2204 | ||
2205 | procfs_list_install("zfs", | |
7b8363d7 | 2206 | NULL, |
ca95f70d OF |
2207 | "import_progress", |
2208 | 0644, | |
2209 | &spa_import_progress_list->procfs_list, | |
2210 | spa_import_progress_show, | |
2211 | spa_import_progress_show_header, | |
2212 | NULL, | |
2213 | offsetof(spa_import_progress_t, smh_node)); | |
2214 | } | |
2215 | ||
2216 | static void | |
2217 | spa_import_progress_destroy(void) | |
2218 | { | |
2219 | spa_history_list_t *shl = spa_import_progress_list; | |
2220 | procfs_list_uninstall(&shl->procfs_list); | |
2221 | spa_import_progress_truncate(shl, 0); | |
ca95f70d | 2222 | procfs_list_destroy(&shl->procfs_list); |
75c09c50 | 2223 | kmem_free(shl, sizeof (spa_history_list_t)); |
ca95f70d OF |
2224 | } |
2225 | ||
2226 | int | |
2227 | spa_import_progress_set_state(uint64_t pool_guid, | |
2228 | spa_load_state_t load_state) | |
2229 | { | |
2230 | spa_history_list_t *shl = spa_import_progress_list; | |
2231 | spa_import_progress_t *sip; | |
2232 | int error = ENOENT; | |
2233 | ||
2234 | if (shl->size == 0) | |
2235 | return (0); | |
2236 | ||
2237 | mutex_enter(&shl->procfs_list.pl_lock); | |
2238 | for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL; | |
2239 | sip = list_prev(&shl->procfs_list.pl_list, sip)) { | |
2240 | if (sip->pool_guid == pool_guid) { | |
2241 | sip->spa_load_state = load_state; | |
2242 | error = 0; | |
2243 | break; | |
2244 | } | |
2245 | } | |
2246 | mutex_exit(&shl->procfs_list.pl_lock); | |
2247 | ||
2248 | return (error); | |
2249 | } | |
2250 | ||
2251 | int | |
2252 | spa_import_progress_set_max_txg(uint64_t pool_guid, uint64_t load_max_txg) | |
2253 | { | |
2254 | spa_history_list_t *shl = spa_import_progress_list; | |
2255 | spa_import_progress_t *sip; | |
2256 | int error = ENOENT; | |
2257 | ||
2258 | if (shl->size == 0) | |
2259 | return (0); | |
2260 | ||
2261 | mutex_enter(&shl->procfs_list.pl_lock); | |
2262 | for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL; | |
2263 | sip = list_prev(&shl->procfs_list.pl_list, sip)) { | |
2264 | if (sip->pool_guid == pool_guid) { | |
2265 | sip->spa_load_max_txg = load_max_txg; | |
2266 | error = 0; | |
2267 | break; | |
2268 | } | |
2269 | } | |
2270 | mutex_exit(&shl->procfs_list.pl_lock); | |
2271 | ||
2272 | return (error); | |
2273 | } | |
2274 | ||
2275 | int | |
2276 | spa_import_progress_set_mmp_check(uint64_t pool_guid, | |
2277 | uint64_t mmp_sec_remaining) | |
2278 | { | |
2279 | spa_history_list_t *shl = spa_import_progress_list; | |
2280 | spa_import_progress_t *sip; | |
2281 | int error = ENOENT; | |
2282 | ||
2283 | if (shl->size == 0) | |
2284 | return (0); | |
2285 | ||
2286 | mutex_enter(&shl->procfs_list.pl_lock); | |
2287 | for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL; | |
2288 | sip = list_prev(&shl->procfs_list.pl_list, sip)) { | |
2289 | if (sip->pool_guid == pool_guid) { | |
2290 | sip->mmp_sec_remaining = mmp_sec_remaining; | |
2291 | error = 0; | |
2292 | break; | |
2293 | } | |
2294 | } | |
2295 | mutex_exit(&shl->procfs_list.pl_lock); | |
2296 | ||
2297 | return (error); | |
2298 | } | |
2299 | ||
2300 | /* | |
2301 | * A new import is in progress, add an entry. | |
2302 | */ | |
2303 | void | |
2304 | spa_import_progress_add(spa_t *spa) | |
2305 | { | |
2306 | spa_history_list_t *shl = spa_import_progress_list; | |
2307 | spa_import_progress_t *sip; | |
2308 | char *poolname = NULL; | |
2309 | ||
2310 | sip = kmem_zalloc(sizeof (spa_import_progress_t), KM_SLEEP); | |
2311 | sip->pool_guid = spa_guid(spa); | |
2312 | ||
2313 | (void) nvlist_lookup_string(spa->spa_config, ZPOOL_CONFIG_POOL_NAME, | |
2314 | &poolname); | |
2315 | if (poolname == NULL) | |
2316 | poolname = spa_name(spa); | |
2317 | sip->pool_name = spa_strdup(poolname); | |
2318 | sip->spa_load_state = spa_load_state(spa); | |
2319 | ||
2320 | mutex_enter(&shl->procfs_list.pl_lock); | |
2321 | procfs_list_add(&shl->procfs_list, sip); | |
2322 | shl->size++; | |
2323 | mutex_exit(&shl->procfs_list.pl_lock); | |
2324 | } | |
2325 | ||
2326 | void | |
2327 | spa_import_progress_remove(uint64_t pool_guid) | |
2328 | { | |
2329 | spa_history_list_t *shl = spa_import_progress_list; | |
2330 | spa_import_progress_t *sip; | |
2331 | ||
2332 | mutex_enter(&shl->procfs_list.pl_lock); | |
2333 | for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL; | |
2334 | sip = list_prev(&shl->procfs_list.pl_list, sip)) { | |
2335 | if (sip->pool_guid == pool_guid) { | |
2336 | if (sip->pool_name) | |
2337 | spa_strfree(sip->pool_name); | |
2338 | list_remove(&shl->procfs_list.pl_list, sip); | |
2339 | shl->size--; | |
2340 | kmem_free(sip, sizeof (spa_import_progress_t)); | |
2341 | break; | |
2342 | } | |
2343 | } | |
2344 | mutex_exit(&shl->procfs_list.pl_lock); | |
2345 | } | |
2346 | ||
34dc7c2f BB |
2347 | /* |
2348 | * ========================================================================== | |
2349 | * Initialization and Termination | |
2350 | * ========================================================================== | |
2351 | */ | |
2352 | ||
2353 | static int | |
2354 | spa_name_compare(const void *a1, const void *a2) | |
2355 | { | |
2356 | const spa_t *s1 = a1; | |
2357 | const spa_t *s2 = a2; | |
2358 | int s; | |
2359 | ||
2360 | s = strcmp(s1->spa_name, s2->spa_name); | |
ee36c709 | 2361 | |
ca577779 | 2362 | return (TREE_ISIGN(s)); |
34dc7c2f BB |
2363 | } |
2364 | ||
34dc7c2f | 2365 | void |
0bc8fd78 | 2366 | spa_boot_init(void) |
34dc7c2f BB |
2367 | { |
2368 | spa_config_load(); | |
2369 | } | |
2370 | ||
2371 | void | |
da92d5cb | 2372 | spa_init(spa_mode_t mode) |
34dc7c2f BB |
2373 | { |
2374 | mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); | |
2375 | mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); | |
2376 | mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); | |
2377 | cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); | |
2378 | ||
2379 | avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), | |
2380 | offsetof(spa_t, spa_avl)); | |
2381 | ||
2382 | avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), | |
2383 | offsetof(spa_aux_t, aux_avl)); | |
2384 | ||
2385 | avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), | |
2386 | offsetof(spa_aux_t, aux_avl)); | |
2387 | ||
fb5f0bc8 | 2388 | spa_mode_global = mode; |
34dc7c2f | 2389 | |
498877ba | 2390 | #ifndef _KERNEL |
da92d5cb | 2391 | if (spa_mode_global != SPA_MODE_READ && dprintf_find_string("watch")) { |
498877ba MA |
2392 | struct sigaction sa; |
2393 | ||
2394 | sa.sa_flags = SA_SIGINFO; | |
2395 | sigemptyset(&sa.sa_mask); | |
2396 | sa.sa_sigaction = arc_buf_sigsegv; | |
2397 | ||
2398 | if (sigaction(SIGSEGV, &sa, NULL) == -1) { | |
2399 | perror("could not enable watchpoints: " | |
2400 | "sigaction(SIGSEGV, ...) = "); | |
2401 | } else { | |
2402 | arc_watch = B_TRUE; | |
2403 | } | |
2404 | } | |
2405 | #endif | |
2406 | ||
26685276 | 2407 | fm_init(); |
424fd7c3 | 2408 | zfs_refcount_init(); |
34dc7c2f | 2409 | unique_init(); |
ca577779 PD |
2410 | zfs_btree_init(); |
2411 | metaslab_stat_init(); | |
ecf3d9b8 | 2412 | ddt_init(); |
34dc7c2f BB |
2413 | zio_init(); |
2414 | dmu_init(); | |
2415 | zil_init(); | |
2416 | vdev_cache_stat_init(); | |
551905dd | 2417 | vdev_mirror_stat_init(); |
ab9f4b0b | 2418 | vdev_raidz_math_init(); |
da8f51e1 | 2419 | vdev_file_init(); |
34dc7c2f BB |
2420 | zfs_prop_init(); |
2421 | zpool_prop_init(); | |
9ae529ec | 2422 | zpool_feature_init(); |
34dc7c2f | 2423 | spa_config_load(); |
2a673e76 | 2424 | vdev_prop_init(); |
b128c09f | 2425 | l2arc_start(); |
d4a72f23 | 2426 | scan_init(); |
6a9d6359 | 2427 | qat_init(); |
ca95f70d | 2428 | spa_import_progress_init(); |
34dc7c2f BB |
2429 | } |
2430 | ||
2431 | void | |
2432 | spa_fini(void) | |
2433 | { | |
b128c09f BB |
2434 | l2arc_stop(); |
2435 | ||
34dc7c2f BB |
2436 | spa_evict_all(); |
2437 | ||
da8f51e1 | 2438 | vdev_file_fini(); |
34dc7c2f | 2439 | vdev_cache_stat_fini(); |
551905dd | 2440 | vdev_mirror_stat_fini(); |
ab9f4b0b | 2441 | vdev_raidz_math_fini(); |
34dc7c2f BB |
2442 | zil_fini(); |
2443 | dmu_fini(); | |
2444 | zio_fini(); | |
ecf3d9b8 | 2445 | ddt_fini(); |
ca577779 PD |
2446 | metaslab_stat_fini(); |
2447 | zfs_btree_fini(); | |
34dc7c2f | 2448 | unique_fini(); |
424fd7c3 | 2449 | zfs_refcount_fini(); |
26685276 | 2450 | fm_fini(); |
d4a72f23 | 2451 | scan_fini(); |
6a9d6359 | 2452 | qat_fini(); |
ca95f70d | 2453 | spa_import_progress_destroy(); |
34dc7c2f BB |
2454 | |
2455 | avl_destroy(&spa_namespace_avl); | |
2456 | avl_destroy(&spa_spare_avl); | |
2457 | avl_destroy(&spa_l2cache_avl); | |
2458 | ||
2459 | cv_destroy(&spa_namespace_cv); | |
2460 | mutex_destroy(&spa_namespace_lock); | |
2461 | mutex_destroy(&spa_spare_lock); | |
2462 | mutex_destroy(&spa_l2cache_lock); | |
2463 | } | |
2464 | ||
2465 | /* | |
aa755b35 | 2466 | * Return whether this pool has a dedicated slog device. No locking needed. |
34dc7c2f | 2467 | * It's not a problem if the wrong answer is returned as it's only for |
aa755b35 | 2468 | * performance and not correctness. |
34dc7c2f BB |
2469 | */ |
2470 | boolean_t | |
2471 | spa_has_slogs(spa_t *spa) | |
2472 | { | |
f8020c93 | 2473 | return (spa->spa_log_class->mc_groups != 0); |
34dc7c2f | 2474 | } |
b128c09f | 2475 | |
428870ff BB |
2476 | spa_log_state_t |
2477 | spa_get_log_state(spa_t *spa) | |
2478 | { | |
2479 | return (spa->spa_log_state); | |
2480 | } | |
2481 | ||
2482 | void | |
2483 | spa_set_log_state(spa_t *spa, spa_log_state_t state) | |
2484 | { | |
2485 | spa->spa_log_state = state; | |
2486 | } | |
2487 | ||
b128c09f BB |
2488 | boolean_t |
2489 | spa_is_root(spa_t *spa) | |
2490 | { | |
2491 | return (spa->spa_is_root); | |
2492 | } | |
fb5f0bc8 BB |
2493 | |
2494 | boolean_t | |
2495 | spa_writeable(spa_t *spa) | |
2496 | { | |
da92d5cb | 2497 | return (!!(spa->spa_mode & SPA_MODE_WRITE) && spa->spa_trust_config); |
fb5f0bc8 BB |
2498 | } |
2499 | ||
acbad6ff AR |
2500 | /* |
2501 | * Returns true if there is a pending sync task in any of the current | |
2502 | * syncing txg, the current quiescing txg, or the current open txg. | |
2503 | */ | |
2504 | boolean_t | |
2505 | spa_has_pending_synctask(spa_t *spa) | |
2506 | { | |
d2734cce SD |
2507 | return (!txg_all_lists_empty(&spa->spa_dsl_pool->dp_sync_tasks) || |
2508 | !txg_all_lists_empty(&spa->spa_dsl_pool->dp_early_sync_tasks)); | |
acbad6ff AR |
2509 | } |
2510 | ||
da92d5cb | 2511 | spa_mode_t |
fb5f0bc8 BB |
2512 | spa_mode(spa_t *spa) |
2513 | { | |
2514 | return (spa->spa_mode); | |
2515 | } | |
428870ff BB |
2516 | |
2517 | uint64_t | |
2518 | spa_bootfs(spa_t *spa) | |
2519 | { | |
2520 | return (spa->spa_bootfs); | |
2521 | } | |
2522 | ||
2523 | uint64_t | |
2524 | spa_delegation(spa_t *spa) | |
2525 | { | |
2526 | return (spa->spa_delegation); | |
2527 | } | |
2528 | ||
2529 | objset_t * | |
2530 | spa_meta_objset(spa_t *spa) | |
2531 | { | |
2532 | return (spa->spa_meta_objset); | |
2533 | } | |
2534 | ||
2535 | enum zio_checksum | |
2536 | spa_dedup_checksum(spa_t *spa) | |
2537 | { | |
2538 | return (spa->spa_dedup_checksum); | |
2539 | } | |
2540 | ||
2541 | /* | |
2542 | * Reset pool scan stat per scan pass (or reboot). | |
2543 | */ | |
2544 | void | |
2545 | spa_scan_stat_init(spa_t *spa) | |
2546 | { | |
2547 | /* data not stored on disk */ | |
2548 | spa->spa_scan_pass_start = gethrestime_sec(); | |
0ea05c64 AP |
2549 | if (dsl_scan_is_paused_scrub(spa->spa_dsl_pool->dp_scan)) |
2550 | spa->spa_scan_pass_scrub_pause = spa->spa_scan_pass_start; | |
2551 | else | |
2552 | spa->spa_scan_pass_scrub_pause = 0; | |
2553 | spa->spa_scan_pass_scrub_spent_paused = 0; | |
428870ff | 2554 | spa->spa_scan_pass_exam = 0; |
d4a72f23 | 2555 | spa->spa_scan_pass_issued = 0; |
428870ff BB |
2556 | vdev_scan_stat_init(spa->spa_root_vdev); |
2557 | } | |
2558 | ||
2559 | /* | |
2560 | * Get scan stats for zpool status reports | |
2561 | */ | |
2562 | int | |
2563 | spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps) | |
2564 | { | |
2565 | dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL; | |
2566 | ||
2567 | if (scn == NULL || scn->scn_phys.scn_func == POOL_SCAN_NONE) | |
2e528b49 | 2568 | return (SET_ERROR(ENOENT)); |
428870ff BB |
2569 | bzero(ps, sizeof (pool_scan_stat_t)); |
2570 | ||
2571 | /* data stored on disk */ | |
2572 | ps->pss_func = scn->scn_phys.scn_func; | |
d4a72f23 | 2573 | ps->pss_state = scn->scn_phys.scn_state; |
428870ff BB |
2574 | ps->pss_start_time = scn->scn_phys.scn_start_time; |
2575 | ps->pss_end_time = scn->scn_phys.scn_end_time; | |
2576 | ps->pss_to_examine = scn->scn_phys.scn_to_examine; | |
d4677269 | 2577 | ps->pss_examined = scn->scn_phys.scn_examined; |
428870ff BB |
2578 | ps->pss_to_process = scn->scn_phys.scn_to_process; |
2579 | ps->pss_processed = scn->scn_phys.scn_processed; | |
2580 | ps->pss_errors = scn->scn_phys.scn_errors; | |
428870ff BB |
2581 | |
2582 | /* data not stored on disk */ | |
428870ff | 2583 | ps->pss_pass_exam = spa->spa_scan_pass_exam; |
d4677269 | 2584 | ps->pss_pass_start = spa->spa_scan_pass_start; |
0ea05c64 AP |
2585 | ps->pss_pass_scrub_pause = spa->spa_scan_pass_scrub_pause; |
2586 | ps->pss_pass_scrub_spent_paused = spa->spa_scan_pass_scrub_spent_paused; | |
d4677269 TC |
2587 | ps->pss_pass_issued = spa->spa_scan_pass_issued; |
2588 | ps->pss_issued = | |
2589 | scn->scn_issued_before_pass + spa->spa_scan_pass_issued; | |
428870ff BB |
2590 | |
2591 | return (0); | |
2592 | } | |
c28b2279 | 2593 | |
f1512ee6 MA |
2594 | int |
2595 | spa_maxblocksize(spa_t *spa) | |
2596 | { | |
2597 | if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) | |
2598 | return (SPA_MAXBLOCKSIZE); | |
2599 | else | |
2600 | return (SPA_OLD_MAXBLOCKSIZE); | |
2601 | } | |
2602 | ||
a1d477c2 MA |
2603 | |
2604 | /* | |
2605 | * Returns the txg that the last device removal completed. No indirect mappings | |
2606 | * have been added since this txg. | |
2607 | */ | |
2608 | uint64_t | |
2609 | spa_get_last_removal_txg(spa_t *spa) | |
2610 | { | |
2611 | uint64_t vdevid; | |
2612 | uint64_t ret = -1ULL; | |
2613 | ||
2614 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2615 | /* | |
2616 | * sr_prev_indirect_vdev is only modified while holding all the | |
2617 | * config locks, so it is sufficient to hold SCL_VDEV as reader when | |
2618 | * examining it. | |
2619 | */ | |
2620 | vdevid = spa->spa_removing_phys.sr_prev_indirect_vdev; | |
2621 | ||
2622 | while (vdevid != -1ULL) { | |
2623 | vdev_t *vd = vdev_lookup_top(spa, vdevid); | |
2624 | vdev_indirect_births_t *vib = vd->vdev_indirect_births; | |
2625 | ||
2626 | ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops); | |
2627 | ||
2628 | /* | |
2629 | * If the removal did not remap any data, we don't care. | |
2630 | */ | |
2631 | if (vdev_indirect_births_count(vib) != 0) { | |
2632 | ret = vdev_indirect_births_last_entry_txg(vib); | |
2633 | break; | |
2634 | } | |
2635 | ||
2636 | vdevid = vd->vdev_indirect_config.vic_prev_indirect_vdev; | |
2637 | } | |
2638 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2639 | ||
2640 | IMPLY(ret != -1ULL, | |
2641 | spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL)); | |
2642 | ||
2643 | return (ret); | |
2644 | } | |
2645 | ||
50c957f7 NB |
2646 | int |
2647 | spa_maxdnodesize(spa_t *spa) | |
2648 | { | |
2649 | if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE)) | |
2650 | return (DNODE_MAX_SIZE); | |
2651 | else | |
2652 | return (DNODE_MIN_SIZE); | |
2653 | } | |
2654 | ||
379ca9cf OF |
2655 | boolean_t |
2656 | spa_multihost(spa_t *spa) | |
2657 | { | |
2658 | return (spa->spa_multihost ? B_TRUE : B_FALSE); | |
2659 | } | |
2660 | ||
25f06d67 BB |
2661 | uint32_t |
2662 | spa_get_hostid(spa_t *spa) | |
379ca9cf | 2663 | { |
25f06d67 | 2664 | return (spa->spa_hostid); |
379ca9cf OF |
2665 | } |
2666 | ||
6cb8e530 PZ |
2667 | boolean_t |
2668 | spa_trust_config(spa_t *spa) | |
2669 | { | |
2670 | return (spa->spa_trust_config); | |
2671 | } | |
2672 | ||
2673 | uint64_t | |
2674 | spa_missing_tvds_allowed(spa_t *spa) | |
2675 | { | |
2676 | return (spa->spa_missing_tvds_allowed); | |
2677 | } | |
2678 | ||
93e28d66 SD |
2679 | space_map_t * |
2680 | spa_syncing_log_sm(spa_t *spa) | |
2681 | { | |
2682 | return (spa->spa_syncing_log_sm); | |
2683 | } | |
2684 | ||
6cb8e530 PZ |
2685 | void |
2686 | spa_set_missing_tvds(spa_t *spa, uint64_t missing) | |
2687 | { | |
2688 | spa->spa_missing_tvds = missing; | |
2689 | } | |
2690 | ||
f0ed6c74 TH |
2691 | /* |
2692 | * Return the pool state string ("ONLINE", "DEGRADED", "SUSPENDED", etc). | |
2693 | */ | |
2694 | const char * | |
2695 | spa_state_to_name(spa_t *spa) | |
2696 | { | |
78fac8d9 RE |
2697 | ASSERT3P(spa, !=, NULL); |
2698 | ||
2699 | /* | |
2700 | * it is possible for the spa to exist, without root vdev | |
2701 | * as the spa transitions during import/export | |
2702 | */ | |
2703 | vdev_t *rvd = spa->spa_root_vdev; | |
2704 | if (rvd == NULL) { | |
2705 | return ("TRANSITIONING"); | |
2706 | } | |
2707 | vdev_state_t state = rvd->vdev_state; | |
2708 | vdev_aux_t aux = rvd->vdev_stat.vs_aux; | |
f0ed6c74 TH |
2709 | |
2710 | if (spa_suspended(spa) && | |
2711 | (spa_get_failmode(spa) != ZIO_FAILURE_MODE_CONTINUE)) | |
2712 | return ("SUSPENDED"); | |
2713 | ||
2714 | switch (state) { | |
2715 | case VDEV_STATE_CLOSED: | |
2716 | case VDEV_STATE_OFFLINE: | |
2717 | return ("OFFLINE"); | |
2718 | case VDEV_STATE_REMOVED: | |
2719 | return ("REMOVED"); | |
2720 | case VDEV_STATE_CANT_OPEN: | |
2721 | if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG) | |
2722 | return ("FAULTED"); | |
2723 | else if (aux == VDEV_AUX_SPLIT_POOL) | |
2724 | return ("SPLIT"); | |
2725 | else | |
2726 | return ("UNAVAIL"); | |
2727 | case VDEV_STATE_FAULTED: | |
2728 | return ("FAULTED"); | |
2729 | case VDEV_STATE_DEGRADED: | |
2730 | return ("DEGRADED"); | |
2731 | case VDEV_STATE_HEALTHY: | |
2732 | return ("ONLINE"); | |
2733 | default: | |
2734 | break; | |
2735 | } | |
2736 | ||
2737 | return ("UNKNOWN"); | |
2738 | } | |
2739 | ||
d2734cce SD |
2740 | boolean_t |
2741 | spa_top_vdevs_spacemap_addressable(spa_t *spa) | |
2742 | { | |
2743 | vdev_t *rvd = spa->spa_root_vdev; | |
2744 | for (uint64_t c = 0; c < rvd->vdev_children; c++) { | |
2745 | if (!vdev_is_spacemap_addressable(rvd->vdev_child[c])) | |
2746 | return (B_FALSE); | |
2747 | } | |
2748 | return (B_TRUE); | |
2749 | } | |
2750 | ||
2751 | boolean_t | |
2752 | spa_has_checkpoint(spa_t *spa) | |
2753 | { | |
2754 | return (spa->spa_checkpoint_txg != 0); | |
2755 | } | |
2756 | ||
2757 | boolean_t | |
2758 | spa_importing_readonly_checkpoint(spa_t *spa) | |
2759 | { | |
2760 | return ((spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT) && | |
da92d5cb | 2761 | spa->spa_mode == SPA_MODE_READ); |
d2734cce SD |
2762 | } |
2763 | ||
2764 | uint64_t | |
2765 | spa_min_claim_txg(spa_t *spa) | |
2766 | { | |
2767 | uint64_t checkpoint_txg = spa->spa_uberblock.ub_checkpoint_txg; | |
2768 | ||
2769 | if (checkpoint_txg != 0) | |
2770 | return (checkpoint_txg + 1); | |
2771 | ||
2772 | return (spa->spa_first_txg); | |
2773 | } | |
2774 | ||
2775 | /* | |
2776 | * If there is a checkpoint, async destroys may consume more space from | |
2777 | * the pool instead of freeing it. In an attempt to save the pool from | |
2778 | * getting suspended when it is about to run out of space, we stop | |
2779 | * processing async destroys. | |
2780 | */ | |
2781 | boolean_t | |
2782 | spa_suspend_async_destroy(spa_t *spa) | |
2783 | { | |
2784 | dsl_pool_t *dp = spa_get_dsl(spa); | |
2785 | ||
2786 | uint64_t unreserved = dsl_pool_unreserved_space(dp, | |
2787 | ZFS_SPACE_CHECK_EXTRA_RESERVED); | |
2788 | uint64_t used = dsl_dir_phys(dp->dp_root_dir)->dd_used_bytes; | |
2789 | uint64_t avail = (unreserved > used) ? (unreserved - used) : 0; | |
2790 | ||
2791 | if (spa_has_checkpoint(spa) && avail == 0) | |
2792 | return (B_TRUE); | |
2793 | ||
2794 | return (B_FALSE); | |
2795 | } | |
2796 | ||
93ce2b4c | 2797 | #if defined(_KERNEL) |
8fb1ede1 | 2798 | |
e64e84ec MM |
2799 | int |
2800 | param_set_deadman_failmode_common(const char *val) | |
8fb1ede1 BB |
2801 | { |
2802 | spa_t *spa = NULL; | |
2803 | char *p; | |
2804 | ||
2805 | if (val == NULL) | |
e64e84ec | 2806 | return (SET_ERROR(EINVAL)); |
8fb1ede1 BB |
2807 | |
2808 | if ((p = strchr(val, '\n')) != NULL) | |
2809 | *p = '\0'; | |
2810 | ||
2811 | if (strcmp(val, "wait") != 0 && strcmp(val, "continue") != 0 && | |
2812 | strcmp(val, "panic")) | |
e64e84ec | 2813 | return (SET_ERROR(EINVAL)); |
8fb1ede1 | 2814 | |
da92d5cb | 2815 | if (spa_mode_global != SPA_MODE_UNINIT) { |
d1043e2f TC |
2816 | mutex_enter(&spa_namespace_lock); |
2817 | while ((spa = spa_next(spa)) != NULL) | |
2818 | spa_set_deadman_failmode(spa, val); | |
2819 | mutex_exit(&spa_namespace_lock); | |
2820 | } | |
8fb1ede1 | 2821 | |
e64e84ec | 2822 | return (0); |
8fb1ede1 | 2823 | } |
03fdcb9a MM |
2824 | #endif |
2825 | ||
c28b2279 BB |
2826 | /* Namespace manipulation */ |
2827 | EXPORT_SYMBOL(spa_lookup); | |
2828 | EXPORT_SYMBOL(spa_add); | |
2829 | EXPORT_SYMBOL(spa_remove); | |
2830 | EXPORT_SYMBOL(spa_next); | |
2831 | ||
2832 | /* Refcount functions */ | |
2833 | EXPORT_SYMBOL(spa_open_ref); | |
2834 | EXPORT_SYMBOL(spa_close); | |
2835 | EXPORT_SYMBOL(spa_refcount_zero); | |
2836 | ||
2837 | /* Pool configuration lock */ | |
2838 | EXPORT_SYMBOL(spa_config_tryenter); | |
2839 | EXPORT_SYMBOL(spa_config_enter); | |
2840 | EXPORT_SYMBOL(spa_config_exit); | |
2841 | EXPORT_SYMBOL(spa_config_held); | |
2842 | ||
2843 | /* Pool vdev add/remove lock */ | |
2844 | EXPORT_SYMBOL(spa_vdev_enter); | |
2845 | EXPORT_SYMBOL(spa_vdev_exit); | |
2846 | ||
2847 | /* Pool vdev state change lock */ | |
2848 | EXPORT_SYMBOL(spa_vdev_state_enter); | |
2849 | EXPORT_SYMBOL(spa_vdev_state_exit); | |
2850 | ||
2851 | /* Accessor functions */ | |
2852 | EXPORT_SYMBOL(spa_shutting_down); | |
2853 | EXPORT_SYMBOL(spa_get_dsl); | |
2854 | EXPORT_SYMBOL(spa_get_rootblkptr); | |
2855 | EXPORT_SYMBOL(spa_set_rootblkptr); | |
2856 | EXPORT_SYMBOL(spa_altroot); | |
2857 | EXPORT_SYMBOL(spa_sync_pass); | |
2858 | EXPORT_SYMBOL(spa_name); | |
2859 | EXPORT_SYMBOL(spa_guid); | |
2860 | EXPORT_SYMBOL(spa_last_synced_txg); | |
2861 | EXPORT_SYMBOL(spa_first_txg); | |
2862 | EXPORT_SYMBOL(spa_syncing_txg); | |
2863 | EXPORT_SYMBOL(spa_version); | |
2864 | EXPORT_SYMBOL(spa_state); | |
2865 | EXPORT_SYMBOL(spa_load_state); | |
2866 | EXPORT_SYMBOL(spa_freeze_txg); | |
c28b2279 BB |
2867 | EXPORT_SYMBOL(spa_get_dspace); |
2868 | EXPORT_SYMBOL(spa_update_dspace); | |
2869 | EXPORT_SYMBOL(spa_deflate); | |
2870 | EXPORT_SYMBOL(spa_normal_class); | |
2871 | EXPORT_SYMBOL(spa_log_class); | |
cc99f275 DB |
2872 | EXPORT_SYMBOL(spa_special_class); |
2873 | EXPORT_SYMBOL(spa_preferred_class); | |
c28b2279 BB |
2874 | EXPORT_SYMBOL(spa_max_replication); |
2875 | EXPORT_SYMBOL(spa_prev_software_version); | |
2876 | EXPORT_SYMBOL(spa_get_failmode); | |
2877 | EXPORT_SYMBOL(spa_suspended); | |
2878 | EXPORT_SYMBOL(spa_bootfs); | |
2879 | EXPORT_SYMBOL(spa_delegation); | |
2880 | EXPORT_SYMBOL(spa_meta_objset); | |
f1512ee6 | 2881 | EXPORT_SYMBOL(spa_maxblocksize); |
50c957f7 | 2882 | EXPORT_SYMBOL(spa_maxdnodesize); |
c28b2279 BB |
2883 | |
2884 | /* Miscellaneous support routines */ | |
c28b2279 BB |
2885 | EXPORT_SYMBOL(spa_guid_exists); |
2886 | EXPORT_SYMBOL(spa_strdup); | |
2887 | EXPORT_SYMBOL(spa_strfree); | |
c28b2279 | 2888 | EXPORT_SYMBOL(spa_generate_guid); |
b0bc7a84 | 2889 | EXPORT_SYMBOL(snprintf_blkptr); |
c28b2279 BB |
2890 | EXPORT_SYMBOL(spa_freeze); |
2891 | EXPORT_SYMBOL(spa_upgrade); | |
2892 | EXPORT_SYMBOL(spa_evict_all); | |
2893 | EXPORT_SYMBOL(spa_lookup_by_guid); | |
2894 | EXPORT_SYMBOL(spa_has_spare); | |
2895 | EXPORT_SYMBOL(dva_get_dsize_sync); | |
2896 | EXPORT_SYMBOL(bp_get_dsize_sync); | |
2897 | EXPORT_SYMBOL(bp_get_dsize); | |
2898 | EXPORT_SYMBOL(spa_has_slogs); | |
2899 | EXPORT_SYMBOL(spa_is_root); | |
2900 | EXPORT_SYMBOL(spa_writeable); | |
2901 | EXPORT_SYMBOL(spa_mode); | |
c28b2279 | 2902 | EXPORT_SYMBOL(spa_namespace_lock); |
6cb8e530 PZ |
2903 | EXPORT_SYMBOL(spa_trust_config); |
2904 | EXPORT_SYMBOL(spa_missing_tvds_allowed); | |
2905 | EXPORT_SYMBOL(spa_set_missing_tvds); | |
f0ed6c74 | 2906 | EXPORT_SYMBOL(spa_state_to_name); |
d2734cce SD |
2907 | EXPORT_SYMBOL(spa_importing_readonly_checkpoint); |
2908 | EXPORT_SYMBOL(spa_min_claim_txg); | |
2909 | EXPORT_SYMBOL(spa_suspend_async_destroy); | |
2910 | EXPORT_SYMBOL(spa_has_checkpoint); | |
2911 | EXPORT_SYMBOL(spa_top_vdevs_spacemap_addressable); | |
cc92e9d0 | 2912 | |
03fdcb9a MM |
2913 | ZFS_MODULE_PARAM(zfs, zfs_, flags, UINT, ZMOD_RW, |
2914 | "Set additional debugging flags"); | |
0b39b9f9 | 2915 | |
03fdcb9a MM |
2916 | ZFS_MODULE_PARAM(zfs, zfs_, recover, INT, ZMOD_RW, |
2917 | "Set to attempt to recover from fatal errors"); | |
0b39b9f9 | 2918 | |
03fdcb9a | 2919 | ZFS_MODULE_PARAM(zfs, zfs_, free_leak_on_eio, INT, ZMOD_RW, |
0b39b9f9 PS |
2920 | "Set to ignore IO errors during free and permanently leak the space"); |
2921 | ||
35aa9dc6 | 2922 | ZFS_MODULE_PARAM(zfs_deadman, zfs_deadman_, checktime_ms, ULONG, ZMOD_RW, |
03fdcb9a MM |
2923 | "Dead I/O check interval in milliseconds"); |
2924 | ||
35aa9dc6 | 2925 | ZFS_MODULE_PARAM(zfs_deadman, zfs_deadman_, enabled, INT, ZMOD_RW, |
03fdcb9a MM |
2926 | "Enable deadman timer"); |
2927 | ||
2928 | ZFS_MODULE_PARAM(zfs_spa, spa_, asize_inflation, INT, ZMOD_RW, | |
2929 | "SPA size estimate multiplication factor"); | |
2930 | ||
2931 | ZFS_MODULE_PARAM(zfs, zfs_, ddt_data_is_special, INT, ZMOD_RW, | |
2932 | "Place DDT data into the special class"); | |
2933 | ||
2934 | ZFS_MODULE_PARAM(zfs, zfs_, user_indirect_is_special, INT, ZMOD_RW, | |
2935 | "Place user data indirect blocks into the special class"); | |
2936 | ||
03fdcb9a | 2937 | /* BEGIN CSTYLED */ |
e64e84ec MM |
2938 | ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, failmode, |
2939 | param_set_deadman_failmode, param_get_charp, ZMOD_RW, | |
2940 | "Failmode for deadman timer"); | |
2941 | ||
2a3aa5a1 MM |
2942 | ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, synctime_ms, |
2943 | param_set_deadman_synctime, param_get_ulong, ZMOD_RW, | |
2944 | "Pool sync expiration time in milliseconds"); | |
2945 | ||
2946 | ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, ziotime_ms, | |
2947 | param_set_deadman_ziotime, param_get_ulong, ZMOD_RW, | |
2948 | "IO expiration time in milliseconds"); | |
2949 | ||
03fdcb9a | 2950 | ZFS_MODULE_PARAM(zfs, zfs_, special_class_metadata_reserve_pct, INT, ZMOD_RW, |
1f02ecc5 D |
2951 | "Small file blocks in special vdevs depends on this much " |
2952 | "free space available"); | |
02730c33 | 2953 | /* END CSTYLED */ |
2a3aa5a1 | 2954 | |
7e3df9db | 2955 | ZFS_MODULE_PARAM_CALL(zfs_spa, spa_, slop_shift, param_set_slop_shift, |
2a3aa5a1 | 2956 | param_get_int, ZMOD_RW, "Reserved free space in pool"); |