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 /*
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011, 2016 by Delphix. All rights reserved.
24 * Copyright (c) 2013 Steven Hartland. All rights reserved.
25 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
26 * Copyright (c) 2014 Integros [integros.com]
27 * Copyright 2016 Nexenta Systems, Inc. All rights reserved.
28 */
29
30 #include <sys/dsl_pool.h>
31 #include <sys/dsl_dataset.h>
32 #include <sys/dsl_prop.h>
33 #include <sys/dsl_dir.h>
34 #include <sys/dsl_synctask.h>
35 #include <sys/dsl_scan.h>
36 #include <sys/dnode.h>
37 #include <sys/dmu_tx.h>
38 #include <sys/dmu_objset.h>
39 #include <sys/arc.h>
40 #include <sys/zap.h>
41 #include <sys/zio.h>
42 #include <sys/zfs_context.h>
43 #include <sys/fs/zfs.h>
44 #include <sys/zfs_znode.h>
45 #include <sys/spa_impl.h>
46 #include <sys/dsl_deadlist.h>
47 #include <sys/bptree.h>
48 #include <sys/zfeature.h>
49 #include <sys/zil_impl.h>
50 #include <sys/dsl_userhold.h>
51
52 #if defined(__FreeBSD__) && defined(_KERNEL)
53 #include <sys/types.h>
54 #include <sys/sysctl.h>
55 #endif
56
57 /*
58 * ZFS Write Throttle
59 * ------------------
60 *
61 * ZFS must limit the rate of incoming writes to the rate at which it is able
62 * to sync data modifications to the backend storage. Throttling by too much
63 * creates an artificial limit; throttling by too little can only be sustained
64 * for short periods and would lead to highly lumpy performance. On a per-pool
65 * basis, ZFS tracks the amount of modified (dirty) data. As operations change
66 * data, the amount of dirty data increases; as ZFS syncs out data, the amount
67 * of dirty data decreases. When the amount of dirty data exceeds a
68 * predetermined threshold further modifications are blocked until the amount
69 * of dirty data decreases (as data is synced out).
70 *
71 * The limit on dirty data is tunable, and should be adjusted according to
72 * both the IO capacity and available memory of the system. The larger the
73 * window, the more ZFS is able to aggregate and amortize metadata (and data)
74 * changes. However, memory is a limited resource, and allowing for more dirty
75 * data comes at the cost of keeping other useful data in memory (for example
76 * ZFS data cached by the ARC).
77 *
78 * Implementation
79 *
80 * As buffers are modified dsl_pool_willuse_space() increments both the per-
81 * txg (dp_dirty_pertxg[]) and poolwide (dp_dirty_total) accounting of
82 * dirty space used; dsl_pool_dirty_space() decrements those values as data
83 * is synced out from dsl_pool_sync(). While only the poolwide value is
84 * relevant, the per-txg value is useful for debugging. The tunable
85 * zfs_dirty_data_max determines the dirty space limit. Once that value is
86 * exceeded, new writes are halted until space frees up.
87 *
88 * The zfs_dirty_data_sync tunable dictates the threshold at which we
89 * ensure that there is a txg syncing (see the comment in txg.c for a full
90 * description of transaction group stages).
91 *
92 * The IO scheduler uses both the dirty space limit and current amount of
93 * dirty data as inputs. Those values affect the number of concurrent IOs ZFS
94 * issues. See the comment in vdev_queue.c for details of the IO scheduler.
95 *
96 * The delay is also calculated based on the amount of dirty data. See the
97 * comment above dmu_tx_delay() for details.
98 */
99
100 /*
101 * zfs_dirty_data_max will be set to zfs_dirty_data_max_percent% of all memory,
102 * capped at zfs_dirty_data_max_max. It can also be overridden in /etc/system.
103 */
104 uint64_t zfs_dirty_data_max;
105 uint64_t zfs_dirty_data_max_max = 4ULL * 1024 * 1024 * 1024;
106 int zfs_dirty_data_max_percent = 10;
107
108 /*
109 * If there is at least this much dirty data, push out a txg.
110 */
111 uint64_t zfs_dirty_data_sync = 64 * 1024 * 1024;
112
113 /*
114 * Once there is this amount of dirty data, the dmu_tx_delay() will kick in
115 * and delay each transaction.
116 * This value should be >= zfs_vdev_async_write_active_max_dirty_percent.
117 */
118 int zfs_delay_min_dirty_percent = 60;
119
120 /*
121 * This controls how quickly the delay approaches infinity.
122 * Larger values cause it to delay more for a given amount of dirty data.
123 * Therefore larger values will cause there to be less dirty data for a
124 * given throughput.
125 *
126 * For the smoothest delay, this value should be about 1 billion divided
127 * by the maximum number of operations per second. This will smoothly
128 * handle between 10x and 1/10th this number.
129 *
130 * Note: zfs_delay_scale * zfs_dirty_data_max must be < 2^64, due to the
131 * multiply in dmu_tx_delay().
132 */
133 uint64_t zfs_delay_scale = 1000 * 1000 * 1000 / 2000;
134
135
136 #if defined(__FreeBSD__) && defined(_KERNEL)
137
138 extern int zfs_vdev_async_write_active_max_dirty_percent;
139
140 SYSCTL_DECL(_vfs_zfs);
141
142 SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, dirty_data_max, CTLFLAG_RWTUN,
143 &zfs_dirty_data_max, 0,
144 "The maximum amount of dirty data in bytes after which new writes are "
145 "halted until space becomes available");
146
147 SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, dirty_data_max_max, CTLFLAG_RDTUN,
148 &zfs_dirty_data_max_max, 0,
149 "The absolute cap on dirty_data_max when auto calculating");
150
151 static int sysctl_zfs_dirty_data_max_percent(SYSCTL_HANDLER_ARGS);
152 SYSCTL_PROC(_vfs_zfs, OID_AUTO, dirty_data_max_percent,
153 CTLTYPE_INT | CTLFLAG_MPSAFE | CTLFLAG_RWTUN, 0, sizeof(int),
154 sysctl_zfs_dirty_data_max_percent, "I",
155 "The percent of physical memory used to auto calculate dirty_data_max");
156
157 SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, dirty_data_sync, CTLFLAG_RWTUN,
158 &zfs_dirty_data_sync, 0,
159 "Force a txg if the number of dirty buffer bytes exceed this value");
160
161 static int sysctl_zfs_delay_min_dirty_percent(SYSCTL_HANDLER_ARGS);
162 /* No zfs_delay_min_dirty_percent tunable due to limit requirements */
163 SYSCTL_PROC(_vfs_zfs, OID_AUTO, delay_min_dirty_percent,
164 CTLTYPE_INT | CTLFLAG_MPSAFE | CTLFLAG_RW, 0, sizeof(int),
165 sysctl_zfs_delay_min_dirty_percent, "I",
166 "The limit of outstanding dirty data before transations are delayed");
167
168 static int sysctl_zfs_delay_scale(SYSCTL_HANDLER_ARGS);
169 /* No zfs_delay_scale tunable due to limit requirements */
170 SYSCTL_PROC(_vfs_zfs, OID_AUTO, delay_scale,
171 CTLTYPE_U64 | CTLFLAG_MPSAFE | CTLFLAG_RW, 0, sizeof(uint64_t),
172 sysctl_zfs_delay_scale, "QU",
173 "Controls how quickly the delay approaches infinity");
174
175 static int
sysctl_zfs_dirty_data_max_percent(SYSCTL_HANDLER_ARGS)176 sysctl_zfs_dirty_data_max_percent(SYSCTL_HANDLER_ARGS)
177 {
178 int val, err;
179
180 val = zfs_dirty_data_max_percent;
181 err = sysctl_handle_int(oidp, &val, 0, req);
182 if (err != 0 || req->newptr == NULL)
183 return (err);
184
185 if (val < 0 || val > 100)
186 return (EINVAL);
187
188 zfs_dirty_data_max_percent = val;
189
190 return (0);
191 }
192
193 static int
sysctl_zfs_delay_min_dirty_percent(SYSCTL_HANDLER_ARGS)194 sysctl_zfs_delay_min_dirty_percent(SYSCTL_HANDLER_ARGS)
195 {
196 int val, err;
197
198 val = zfs_delay_min_dirty_percent;
199 err = sysctl_handle_int(oidp, &val, 0, req);
200 if (err != 0 || req->newptr == NULL)
201 return (err);
202
203 if (val < zfs_vdev_async_write_active_max_dirty_percent)
204 return (EINVAL);
205
206 zfs_delay_min_dirty_percent = val;
207
208 return (0);
209 }
210
211 static int
sysctl_zfs_delay_scale(SYSCTL_HANDLER_ARGS)212 sysctl_zfs_delay_scale(SYSCTL_HANDLER_ARGS)
213 {
214 uint64_t val;
215 int err;
216
217 val = zfs_delay_scale;
218 err = sysctl_handle_64(oidp, &val, 0, req);
219 if (err != 0 || req->newptr == NULL)
220 return (err);
221
222 if (val > UINT64_MAX / zfs_dirty_data_max)
223 return (EINVAL);
224
225 zfs_delay_scale = val;
226
227 return (0);
228 }
229 #endif
230
231 hrtime_t zfs_throttle_delay = MSEC2NSEC(10);
232 hrtime_t zfs_throttle_resolution = MSEC2NSEC(10);
233
234 int
dsl_pool_open_special_dir(dsl_pool_t * dp,const char * name,dsl_dir_t ** ddp)235 dsl_pool_open_special_dir(dsl_pool_t *dp, const char *name, dsl_dir_t **ddp)
236 {
237 uint64_t obj;
238 int err;
239
240 err = zap_lookup(dp->dp_meta_objset,
241 dsl_dir_phys(dp->dp_root_dir)->dd_child_dir_zapobj,
242 name, sizeof (obj), 1, &obj);
243 if (err)
244 return (err);
245
246 return (dsl_dir_hold_obj(dp, obj, name, dp, ddp));
247 }
248
249 static dsl_pool_t *
dsl_pool_open_impl(spa_t * spa,uint64_t txg)250 dsl_pool_open_impl(spa_t *spa, uint64_t txg)
251 {
252 dsl_pool_t *dp;
253 blkptr_t *bp = spa_get_rootblkptr(spa);
254
255 dp = kmem_zalloc(sizeof (dsl_pool_t), KM_SLEEP);
256 dp->dp_spa = spa;
257 dp->dp_meta_rootbp = *bp;
258 rrw_init(&dp->dp_config_rwlock, B_TRUE);
259 txg_init(dp, txg);
260
261 txg_list_create(&dp->dp_dirty_datasets,
262 offsetof(dsl_dataset_t, ds_dirty_link));
263 txg_list_create(&dp->dp_dirty_zilogs,
264 offsetof(zilog_t, zl_dirty_link));
265 txg_list_create(&dp->dp_dirty_dirs,
266 offsetof(dsl_dir_t, dd_dirty_link));
267 txg_list_create(&dp->dp_sync_tasks,
268 offsetof(dsl_sync_task_t, dst_node));
269
270 mutex_init(&dp->dp_lock, NULL, MUTEX_DEFAULT, NULL);
271 cv_init(&dp->dp_spaceavail_cv, NULL, CV_DEFAULT, NULL);
272
273 dp->dp_vnrele_taskq = taskq_create("zfs_vn_rele_taskq", 1, minclsyspri,
274 1, 4, 0);
275
276 return (dp);
277 }
278
279 int
dsl_pool_init(spa_t * spa,uint64_t txg,dsl_pool_t ** dpp)280 dsl_pool_init(spa_t *spa, uint64_t txg, dsl_pool_t **dpp)
281 {
282 int err;
283 dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);
284
285 err = dmu_objset_open_impl(spa, NULL, &dp->dp_meta_rootbp,
286 &dp->dp_meta_objset);
287 if (err != 0)
288 dsl_pool_close(dp);
289 else
290 *dpp = dp;
291
292 return (err);
293 }
294
295 int
dsl_pool_open(dsl_pool_t * dp)296 dsl_pool_open(dsl_pool_t *dp)
297 {
298 int err;
299 dsl_dir_t *dd;
300 dsl_dataset_t *ds;
301 uint64_t obj;
302
303 rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG);
304 err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
305 DMU_POOL_ROOT_DATASET, sizeof (uint64_t), 1,
306 &dp->dp_root_dir_obj);
307 if (err)
308 goto out;
309
310 err = dsl_dir_hold_obj(dp, dp->dp_root_dir_obj,
311 NULL, dp, &dp->dp_root_dir);
312 if (err)
313 goto out;
314
315 err = dsl_pool_open_special_dir(dp, MOS_DIR_NAME, &dp->dp_mos_dir);
316 if (err)
317 goto out;
318
319 if (spa_version(dp->dp_spa) >= SPA_VERSION_ORIGIN) {
320 err = dsl_pool_open_special_dir(dp, ORIGIN_DIR_NAME, &dd);
321 if (err)
322 goto out;
323 err = dsl_dataset_hold_obj(dp,
324 dsl_dir_phys(dd)->dd_head_dataset_obj, FTAG, &ds);
325 if (err == 0) {
326 err = dsl_dataset_hold_obj(dp,
327 dsl_dataset_phys(ds)->ds_prev_snap_obj, dp,
328 &dp->dp_origin_snap);
329 dsl_dataset_rele(ds, FTAG);
330 }
331 dsl_dir_rele(dd, dp);
332 if (err)
333 goto out;
334 }
335
336 if (spa_version(dp->dp_spa) >= SPA_VERSION_DEADLISTS) {
337 err = dsl_pool_open_special_dir(dp, FREE_DIR_NAME,
338 &dp->dp_free_dir);
339 if (err)
340 goto out;
341
342 err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
343 DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj);
344 if (err)
345 goto out;
346 VERIFY0(bpobj_open(&dp->dp_free_bpobj,
347 dp->dp_meta_objset, obj));
348 }
349
350 /*
351 * Note: errors ignored, because the leak dir will not exist if we
352 * have not encountered a leak yet.
353 */
354 (void) dsl_pool_open_special_dir(dp, LEAK_DIR_NAME,
355 &dp->dp_leak_dir);
356
357 if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_ASYNC_DESTROY)) {
358 err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
359 DMU_POOL_BPTREE_OBJ, sizeof (uint64_t), 1,
360 &dp->dp_bptree_obj);
361 if (err != 0)
362 goto out;
363 }
364
365 if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_EMPTY_BPOBJ)) {
366 err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
367 DMU_POOL_EMPTY_BPOBJ, sizeof (uint64_t), 1,
368 &dp->dp_empty_bpobj);
369 if (err != 0)
370 goto out;
371 }
372
373 err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
374 DMU_POOL_TMP_USERREFS, sizeof (uint64_t), 1,
375 &dp->dp_tmp_userrefs_obj);
376 if (err == ENOENT)
377 err = 0;
378 if (err)
379 goto out;
380
381 err = dsl_scan_init(dp, dp->dp_tx.tx_open_txg);
382
383 out:
384 rrw_exit(&dp->dp_config_rwlock, FTAG);
385 return (err);
386 }
387
388 void
dsl_pool_close(dsl_pool_t * dp)389 dsl_pool_close(dsl_pool_t *dp)
390 {
391 /*
392 * Drop our references from dsl_pool_open().
393 *
394 * Since we held the origin_snap from "syncing" context (which
395 * includes pool-opening context), it actually only got a "ref"
396 * and not a hold, so just drop that here.
397 */
398 if (dp->dp_origin_snap)
399 dsl_dataset_rele(dp->dp_origin_snap, dp);
400 if (dp->dp_mos_dir)
401 dsl_dir_rele(dp->dp_mos_dir, dp);
402 if (dp->dp_free_dir)
403 dsl_dir_rele(dp->dp_free_dir, dp);
404 if (dp->dp_leak_dir)
405 dsl_dir_rele(dp->dp_leak_dir, dp);
406 if (dp->dp_root_dir)
407 dsl_dir_rele(dp->dp_root_dir, dp);
408
409 bpobj_close(&dp->dp_free_bpobj);
410
411 /* undo the dmu_objset_open_impl(mos) from dsl_pool_open() */
412 if (dp->dp_meta_objset)
413 dmu_objset_evict(dp->dp_meta_objset);
414
415 txg_list_destroy(&dp->dp_dirty_datasets);
416 txg_list_destroy(&dp->dp_dirty_zilogs);
417 txg_list_destroy(&dp->dp_sync_tasks);
418 txg_list_destroy(&dp->dp_dirty_dirs);
419
420 /*
421 * We can't set retry to TRUE since we're explicitly specifying
422 * a spa to flush. This is good enough; any missed buffers for
423 * this spa won't cause trouble, and they'll eventually fall
424 * out of the ARC just like any other unused buffer.
425 */
426 arc_flush(dp->dp_spa, FALSE);
427
428 txg_fini(dp);
429 dsl_scan_fini(dp);
430 dmu_buf_user_evict_wait();
431
432 rrw_destroy(&dp->dp_config_rwlock);
433 mutex_destroy(&dp->dp_lock);
434 cv_destroy(&dp->dp_spaceavail_cv);
435 taskq_destroy(dp->dp_vnrele_taskq);
436 if (dp->dp_blkstats)
437 kmem_free(dp->dp_blkstats, sizeof (zfs_all_blkstats_t));
438 kmem_free(dp, sizeof (dsl_pool_t));
439 }
440
441 dsl_pool_t *
dsl_pool_create(spa_t * spa,nvlist_t * zplprops,uint64_t txg)442 dsl_pool_create(spa_t *spa, nvlist_t *zplprops, uint64_t txg)
443 {
444 int err;
445 dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);
446 dmu_tx_t *tx = dmu_tx_create_assigned(dp, txg);
447 objset_t *os;
448 dsl_dataset_t *ds;
449 uint64_t obj;
450
451 rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG);
452
453 /* create and open the MOS (meta-objset) */
454 dp->dp_meta_objset = dmu_objset_create_impl(spa,
455 NULL, &dp->dp_meta_rootbp, DMU_OST_META, tx);
456
457 /* create the pool directory */
458 err = zap_create_claim(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
459 DMU_OT_OBJECT_DIRECTORY, DMU_OT_NONE, 0, tx);
460 ASSERT0(err);
461
462 /* Initialize scan structures */
463 VERIFY0(dsl_scan_init(dp, txg));
464
465 /* create and open the root dir */
466 dp->dp_root_dir_obj = dsl_dir_create_sync(dp, NULL, NULL, tx);
467 VERIFY0(dsl_dir_hold_obj(dp, dp->dp_root_dir_obj,
468 NULL, dp, &dp->dp_root_dir));
469
470 /* create and open the meta-objset dir */
471 (void) dsl_dir_create_sync(dp, dp->dp_root_dir, MOS_DIR_NAME, tx);
472 VERIFY0(dsl_pool_open_special_dir(dp,
473 MOS_DIR_NAME, &dp->dp_mos_dir));
474
475 if (spa_version(spa) >= SPA_VERSION_DEADLISTS) {
476 /* create and open the free dir */
477 (void) dsl_dir_create_sync(dp, dp->dp_root_dir,
478 FREE_DIR_NAME, tx);
479 VERIFY0(dsl_pool_open_special_dir(dp,
480 FREE_DIR_NAME, &dp->dp_free_dir));
481
482 /* create and open the free_bplist */
483 obj = bpobj_alloc(dp->dp_meta_objset, SPA_OLD_MAXBLOCKSIZE, tx);
484 VERIFY(zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
485 DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj, tx) == 0);
486 VERIFY0(bpobj_open(&dp->dp_free_bpobj,
487 dp->dp_meta_objset, obj));
488 }
489
490 if (spa_version(spa) >= SPA_VERSION_DSL_SCRUB)
491 dsl_pool_create_origin(dp, tx);
492
493 /* create the root dataset */
494 obj = dsl_dataset_create_sync_dd(dp->dp_root_dir, NULL, 0, tx);
495
496 /* create the root objset */
497 VERIFY0(dsl_dataset_hold_obj(dp, obj, FTAG, &ds));
498 rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
499 os = dmu_objset_create_impl(dp->dp_spa, ds,
500 dsl_dataset_get_blkptr(ds), DMU_OST_ZFS, tx);
501 rrw_exit(&ds->ds_bp_rwlock, FTAG);
502 #ifdef _KERNEL
503 zfs_create_fs(os, kcred, zplprops, tx);
504 #endif
505 dsl_dataset_rele(ds, FTAG);
506
507 dmu_tx_commit(tx);
508
509 rrw_exit(&dp->dp_config_rwlock, FTAG);
510
511 return (dp);
512 }
513
514 /*
515 * Account for the meta-objset space in its placeholder dsl_dir.
516 */
517 void
dsl_pool_mos_diduse_space(dsl_pool_t * dp,int64_t used,int64_t comp,int64_t uncomp)518 dsl_pool_mos_diduse_space(dsl_pool_t *dp,
519 int64_t used, int64_t comp, int64_t uncomp)
520 {
521 ASSERT3U(comp, ==, uncomp); /* it's all metadata */
522 mutex_enter(&dp->dp_lock);
523 dp->dp_mos_used_delta += used;
524 dp->dp_mos_compressed_delta += comp;
525 dp->dp_mos_uncompressed_delta += uncomp;
526 mutex_exit(&dp->dp_lock);
527 }
528
529 static void
dsl_pool_sync_mos(dsl_pool_t * dp,dmu_tx_t * tx)530 dsl_pool_sync_mos(dsl_pool_t *dp, dmu_tx_t *tx)
531 {
532 zio_t *zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
533 dmu_objset_sync(dp->dp_meta_objset, zio, tx);
534 VERIFY0(zio_wait(zio));
535 dprintf_bp(&dp->dp_meta_rootbp, "meta objset rootbp is %s", "");
536 spa_set_rootblkptr(dp->dp_spa, &dp->dp_meta_rootbp);
537 }
538
539 static void
dsl_pool_dirty_delta(dsl_pool_t * dp,int64_t delta)540 dsl_pool_dirty_delta(dsl_pool_t *dp, int64_t delta)
541 {
542 ASSERT(MUTEX_HELD(&dp->dp_lock));
543
544 if (delta < 0)
545 ASSERT3U(-delta, <=, dp->dp_dirty_total);
546
547 dp->dp_dirty_total += delta;
548
549 /*
550 * Note: we signal even when increasing dp_dirty_total.
551 * This ensures forward progress -- each thread wakes the next waiter.
552 */
553 if (dp->dp_dirty_total <= zfs_dirty_data_max)
554 cv_signal(&dp->dp_spaceavail_cv);
555 }
556
557 void
dsl_pool_sync(dsl_pool_t * dp,uint64_t txg)558 dsl_pool_sync(dsl_pool_t *dp, uint64_t txg)
559 {
560 zio_t *zio;
561 dmu_tx_t *tx;
562 dsl_dir_t *dd;
563 dsl_dataset_t *ds;
564 objset_t *mos = dp->dp_meta_objset;
565 list_t synced_datasets;
566
567 list_create(&synced_datasets, sizeof (dsl_dataset_t),
568 offsetof(dsl_dataset_t, ds_synced_link));
569
570 tx = dmu_tx_create_assigned(dp, txg);
571
572 /*
573 * Write out all dirty blocks of dirty datasets.
574 */
575 zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
576 while ((ds = txg_list_remove(&dp->dp_dirty_datasets, txg)) != NULL) {
577 /*
578 * We must not sync any non-MOS datasets twice, because
579 * we may have taken a snapshot of them. However, we
580 * may sync newly-created datasets on pass 2.
581 */
582 ASSERT(!list_link_active(&ds->ds_synced_link));
583 list_insert_tail(&synced_datasets, ds);
584 dsl_dataset_sync(ds, zio, tx);
585 }
586 VERIFY0(zio_wait(zio));
587
588 /*
589 * We have written all of the accounted dirty data, so our
590 * dp_space_towrite should now be zero. However, some seldom-used
591 * code paths do not adhere to this (e.g. dbuf_undirty(), also
592 * rounding error in dbuf_write_physdone).
593 * Shore up the accounting of any dirtied space now.
594 */
595 dsl_pool_undirty_space(dp, dp->dp_dirty_pertxg[txg & TXG_MASK], txg);
596
597 /*
598 * Update the long range free counter after
599 * we're done syncing user data
600 */
601 mutex_enter(&dp->dp_lock);
602 ASSERT(spa_sync_pass(dp->dp_spa) == 1 ||
603 dp->dp_long_free_dirty_pertxg[txg & TXG_MASK] == 0);
604 dp->dp_long_free_dirty_pertxg[txg & TXG_MASK] = 0;
605 mutex_exit(&dp->dp_lock);
606
607 /*
608 * After the data blocks have been written (ensured by the zio_wait()
609 * above), update the user/group space accounting.
610 */
611 for (ds = list_head(&synced_datasets); ds != NULL;
612 ds = list_next(&synced_datasets, ds)) {
613 dmu_objset_do_userquota_updates(ds->ds_objset, tx);
614 }
615
616 /*
617 * Sync the datasets again to push out the changes due to
618 * userspace updates. This must be done before we process the
619 * sync tasks, so that any snapshots will have the correct
620 * user accounting information (and we won't get confused
621 * about which blocks are part of the snapshot).
622 */
623 zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
624 while ((ds = txg_list_remove(&dp->dp_dirty_datasets, txg)) != NULL) {
625 ASSERT(list_link_active(&ds->ds_synced_link));
626 dmu_buf_rele(ds->ds_dbuf, ds);
627 dsl_dataset_sync(ds, zio, tx);
628 }
629 VERIFY0(zio_wait(zio));
630
631 /*
632 * Now that the datasets have been completely synced, we can
633 * clean up our in-memory structures accumulated while syncing:
634 *
635 * - move dead blocks from the pending deadlist to the on-disk deadlist
636 * - release hold from dsl_dataset_dirty()
637 */
638 while ((ds = list_remove_head(&synced_datasets)) != NULL) {
639 dsl_dataset_sync_done(ds, tx);
640 }
641 while ((dd = txg_list_remove(&dp->dp_dirty_dirs, txg)) != NULL) {
642 dsl_dir_sync(dd, tx);
643 }
644
645 /*
646 * The MOS's space is accounted for in the pool/$MOS
647 * (dp_mos_dir). We can't modify the mos while we're syncing
648 * it, so we remember the deltas and apply them here.
649 */
650 if (dp->dp_mos_used_delta != 0 || dp->dp_mos_compressed_delta != 0 ||
651 dp->dp_mos_uncompressed_delta != 0) {
652 dsl_dir_diduse_space(dp->dp_mos_dir, DD_USED_HEAD,
653 dp->dp_mos_used_delta,
654 dp->dp_mos_compressed_delta,
655 dp->dp_mos_uncompressed_delta, tx);
656 dp->dp_mos_used_delta = 0;
657 dp->dp_mos_compressed_delta = 0;
658 dp->dp_mos_uncompressed_delta = 0;
659 }
660
661 if (list_head(&mos->os_dirty_dnodes[txg & TXG_MASK]) != NULL ||
662 list_head(&mos->os_free_dnodes[txg & TXG_MASK]) != NULL) {
663 dsl_pool_sync_mos(dp, tx);
664 }
665
666 /*
667 * If we modify a dataset in the same txg that we want to destroy it,
668 * its dsl_dir's dd_dbuf will be dirty, and thus have a hold on it.
669 * dsl_dir_destroy_check() will fail if there are unexpected holds.
670 * Therefore, we want to sync the MOS (thus syncing the dd_dbuf
671 * and clearing the hold on it) before we process the sync_tasks.
672 * The MOS data dirtied by the sync_tasks will be synced on the next
673 * pass.
674 */
675 if (!txg_list_empty(&dp->dp_sync_tasks, txg)) {
676 dsl_sync_task_t *dst;
677 /*
678 * No more sync tasks should have been added while we
679 * were syncing.
680 */
681 ASSERT3U(spa_sync_pass(dp->dp_spa), ==, 1);
682 while ((dst = txg_list_remove(&dp->dp_sync_tasks, txg)) != NULL)
683 dsl_sync_task_sync(dst, tx);
684 }
685
686 dmu_tx_commit(tx);
687
688 DTRACE_PROBE2(dsl_pool_sync__done, dsl_pool_t *dp, dp, uint64_t, txg);
689 }
690
691 void
dsl_pool_sync_done(dsl_pool_t * dp,uint64_t txg)692 dsl_pool_sync_done(dsl_pool_t *dp, uint64_t txg)
693 {
694 zilog_t *zilog;
695
696 while (zilog = txg_list_head(&dp->dp_dirty_zilogs, txg)) {
697 dsl_dataset_t *ds = dmu_objset_ds(zilog->zl_os);
698 /*
699 * We don't remove the zilog from the dp_dirty_zilogs
700 * list until after we've cleaned it. This ensures that
701 * callers of zilog_is_dirty() receive an accurate
702 * answer when they are racing with the spa sync thread.
703 */
704 zil_clean(zilog, txg);
705 (void) txg_list_remove_this(&dp->dp_dirty_zilogs, zilog, txg);
706 ASSERT(!dmu_objset_is_dirty(zilog->zl_os, txg));
707 dmu_buf_rele(ds->ds_dbuf, zilog);
708 }
709 ASSERT(!dmu_objset_is_dirty(dp->dp_meta_objset, txg));
710 }
711
712 /*
713 * TRUE if the current thread is the tx_sync_thread or if we
714 * are being called from SPA context during pool initialization.
715 */
716 int
dsl_pool_sync_context(dsl_pool_t * dp)717 dsl_pool_sync_context(dsl_pool_t *dp)
718 {
719 return (curthread == dp->dp_tx.tx_sync_thread ||
720 spa_is_initializing(dp->dp_spa));
721 }
722
723 uint64_t
dsl_pool_adjustedsize(dsl_pool_t * dp,boolean_t netfree)724 dsl_pool_adjustedsize(dsl_pool_t *dp, boolean_t netfree)
725 {
726 uint64_t space, resv;
727
728 /*
729 * If we're trying to assess whether it's OK to do a free,
730 * cut the reservation in half to allow forward progress
731 * (e.g. make it possible to rm(1) files from a full pool).
732 */
733 space = spa_get_dspace(dp->dp_spa);
734 resv = spa_get_slop_space(dp->dp_spa);
735 if (netfree)
736 resv >>= 1;
737
738 return (space - resv);
739 }
740
741 boolean_t
dsl_pool_need_dirty_delay(dsl_pool_t * dp)742 dsl_pool_need_dirty_delay(dsl_pool_t *dp)
743 {
744 uint64_t delay_min_bytes =
745 zfs_dirty_data_max * zfs_delay_min_dirty_percent / 100;
746 boolean_t rv;
747
748 mutex_enter(&dp->dp_lock);
749 if (dp->dp_dirty_total > zfs_dirty_data_sync)
750 txg_kick(dp);
751 rv = (dp->dp_dirty_total > delay_min_bytes);
752 mutex_exit(&dp->dp_lock);
753 return (rv);
754 }
755
756 void
dsl_pool_dirty_space(dsl_pool_t * dp,int64_t space,dmu_tx_t * tx)757 dsl_pool_dirty_space(dsl_pool_t *dp, int64_t space, dmu_tx_t *tx)
758 {
759 if (space > 0) {
760 mutex_enter(&dp->dp_lock);
761 dp->dp_dirty_pertxg[tx->tx_txg & TXG_MASK] += space;
762 dsl_pool_dirty_delta(dp, space);
763 mutex_exit(&dp->dp_lock);
764 }
765 }
766
767 void
dsl_pool_undirty_space(dsl_pool_t * dp,int64_t space,uint64_t txg)768 dsl_pool_undirty_space(dsl_pool_t *dp, int64_t space, uint64_t txg)
769 {
770 ASSERT3S(space, >=, 0);
771 if (space == 0)
772 return;
773 mutex_enter(&dp->dp_lock);
774 if (dp->dp_dirty_pertxg[txg & TXG_MASK] < space) {
775 /* XXX writing something we didn't dirty? */
776 space = dp->dp_dirty_pertxg[txg & TXG_MASK];
777 }
778 ASSERT3U(dp->dp_dirty_pertxg[txg & TXG_MASK], >=, space);
779 dp->dp_dirty_pertxg[txg & TXG_MASK] -= space;
780 ASSERT3U(dp->dp_dirty_total, >=, space);
781 dsl_pool_dirty_delta(dp, -space);
782 mutex_exit(&dp->dp_lock);
783 }
784
785 /* ARGSUSED */
786 static int
upgrade_clones_cb(dsl_pool_t * dp,dsl_dataset_t * hds,void * arg)787 upgrade_clones_cb(dsl_pool_t *dp, dsl_dataset_t *hds, void *arg)
788 {
789 dmu_tx_t *tx = arg;
790 dsl_dataset_t *ds, *prev = NULL;
791 int err;
792
793 err = dsl_dataset_hold_obj(dp, hds->ds_object, FTAG, &ds);
794 if (err)
795 return (err);
796
797 while (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) {
798 err = dsl_dataset_hold_obj(dp,
799 dsl_dataset_phys(ds)->ds_prev_snap_obj, FTAG, &prev);
800 if (err) {
801 dsl_dataset_rele(ds, FTAG);
802 return (err);
803 }
804
805 if (dsl_dataset_phys(prev)->ds_next_snap_obj != ds->ds_object)
806 break;
807 dsl_dataset_rele(ds, FTAG);
808 ds = prev;
809 prev = NULL;
810 }
811
812 if (prev == NULL) {
813 prev = dp->dp_origin_snap;
814
815 /*
816 * The $ORIGIN can't have any data, or the accounting
817 * will be wrong.
818 */
819 rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
820 ASSERT0(dsl_dataset_phys(prev)->ds_bp.blk_birth);
821 rrw_exit(&ds->ds_bp_rwlock, FTAG);
822
823 /* The origin doesn't get attached to itself */
824 if (ds->ds_object == prev->ds_object) {
825 dsl_dataset_rele(ds, FTAG);
826 return (0);
827 }
828
829 dmu_buf_will_dirty(ds->ds_dbuf, tx);
830 dsl_dataset_phys(ds)->ds_prev_snap_obj = prev->ds_object;
831 dsl_dataset_phys(ds)->ds_prev_snap_txg =
832 dsl_dataset_phys(prev)->ds_creation_txg;
833
834 dmu_buf_will_dirty(ds->ds_dir->dd_dbuf, tx);
835 dsl_dir_phys(ds->ds_dir)->dd_origin_obj = prev->ds_object;
836
837 dmu_buf_will_dirty(prev->ds_dbuf, tx);
838 dsl_dataset_phys(prev)->ds_num_children++;
839
840 if (dsl_dataset_phys(ds)->ds_next_snap_obj == 0) {
841 ASSERT(ds->ds_prev == NULL);
842 VERIFY0(dsl_dataset_hold_obj(dp,
843 dsl_dataset_phys(ds)->ds_prev_snap_obj,
844 ds, &ds->ds_prev));
845 }
846 }
847
848 ASSERT3U(dsl_dir_phys(ds->ds_dir)->dd_origin_obj, ==, prev->ds_object);
849 ASSERT3U(dsl_dataset_phys(ds)->ds_prev_snap_obj, ==, prev->ds_object);
850
851 if (dsl_dataset_phys(prev)->ds_next_clones_obj == 0) {
852 dmu_buf_will_dirty(prev->ds_dbuf, tx);
853 dsl_dataset_phys(prev)->ds_next_clones_obj =
854 zap_create(dp->dp_meta_objset,
855 DMU_OT_NEXT_CLONES, DMU_OT_NONE, 0, tx);
856 }
857 VERIFY0(zap_add_int(dp->dp_meta_objset,
858 dsl_dataset_phys(prev)->ds_next_clones_obj, ds->ds_object, tx));
859
860 dsl_dataset_rele(ds, FTAG);
861 if (prev != dp->dp_origin_snap)
862 dsl_dataset_rele(prev, FTAG);
863 return (0);
864 }
865
866 void
dsl_pool_upgrade_clones(dsl_pool_t * dp,dmu_tx_t * tx)867 dsl_pool_upgrade_clones(dsl_pool_t *dp, dmu_tx_t *tx)
868 {
869 ASSERT(dmu_tx_is_syncing(tx));
870 ASSERT(dp->dp_origin_snap != NULL);
871
872 VERIFY0(dmu_objset_find_dp(dp, dp->dp_root_dir_obj, upgrade_clones_cb,
873 tx, DS_FIND_CHILDREN | DS_FIND_SERIALIZE));
874 }
875
876 /* ARGSUSED */
877 static int
upgrade_dir_clones_cb(dsl_pool_t * dp,dsl_dataset_t * ds,void * arg)878 upgrade_dir_clones_cb(dsl_pool_t *dp, dsl_dataset_t *ds, void *arg)
879 {
880 dmu_tx_t *tx = arg;
881 objset_t *mos = dp->dp_meta_objset;
882
883 if (dsl_dir_phys(ds->ds_dir)->dd_origin_obj != 0) {
884 dsl_dataset_t *origin;
885
886 VERIFY0(dsl_dataset_hold_obj(dp,
887 dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &origin));
888
889 if (dsl_dir_phys(origin->ds_dir)->dd_clones == 0) {
890 dmu_buf_will_dirty(origin->ds_dir->dd_dbuf, tx);
891 dsl_dir_phys(origin->ds_dir)->dd_clones =
892 zap_create(mos, DMU_OT_DSL_CLONES, DMU_OT_NONE,
893 0, tx);
894 }
895
896 VERIFY0(zap_add_int(dp->dp_meta_objset,
897 dsl_dir_phys(origin->ds_dir)->dd_clones,
898 ds->ds_object, tx));
899
900 dsl_dataset_rele(origin, FTAG);
901 }
902 return (0);
903 }
904
905 void
dsl_pool_upgrade_dir_clones(dsl_pool_t * dp,dmu_tx_t * tx)906 dsl_pool_upgrade_dir_clones(dsl_pool_t *dp, dmu_tx_t *tx)
907 {
908 ASSERT(dmu_tx_is_syncing(tx));
909 uint64_t obj;
910
911 (void) dsl_dir_create_sync(dp, dp->dp_root_dir, FREE_DIR_NAME, tx);
912 VERIFY0(dsl_pool_open_special_dir(dp,
913 FREE_DIR_NAME, &dp->dp_free_dir));
914
915 /*
916 * We can't use bpobj_alloc(), because spa_version() still
917 * returns the old version, and we need a new-version bpobj with
918 * subobj support. So call dmu_object_alloc() directly.
919 */
920 obj = dmu_object_alloc(dp->dp_meta_objset, DMU_OT_BPOBJ,
921 SPA_OLD_MAXBLOCKSIZE, DMU_OT_BPOBJ_HDR, sizeof (bpobj_phys_t), tx);
922 VERIFY0(zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
923 DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj, tx));
924 VERIFY0(bpobj_open(&dp->dp_free_bpobj, dp->dp_meta_objset, obj));
925
926 VERIFY0(dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
927 upgrade_dir_clones_cb, tx, DS_FIND_CHILDREN | DS_FIND_SERIALIZE));
928 }
929
930 void
dsl_pool_create_origin(dsl_pool_t * dp,dmu_tx_t * tx)931 dsl_pool_create_origin(dsl_pool_t *dp, dmu_tx_t *tx)
932 {
933 uint64_t dsobj;
934 dsl_dataset_t *ds;
935
936 ASSERT(dmu_tx_is_syncing(tx));
937 ASSERT(dp->dp_origin_snap == NULL);
938 ASSERT(rrw_held(&dp->dp_config_rwlock, RW_WRITER));
939
940 /* create the origin dir, ds, & snap-ds */
941 dsobj = dsl_dataset_create_sync(dp->dp_root_dir, ORIGIN_DIR_NAME,
942 NULL, 0, kcred, tx);
943 VERIFY0(dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));
944 dsl_dataset_snapshot_sync_impl(ds, ORIGIN_DIR_NAME, tx);
945 VERIFY0(dsl_dataset_hold_obj(dp, dsl_dataset_phys(ds)->ds_prev_snap_obj,
946 dp, &dp->dp_origin_snap));
947 dsl_dataset_rele(ds, FTAG);
948 }
949
950 taskq_t *
dsl_pool_vnrele_taskq(dsl_pool_t * dp)951 dsl_pool_vnrele_taskq(dsl_pool_t *dp)
952 {
953 return (dp->dp_vnrele_taskq);
954 }
955
956 /*
957 * Walk through the pool-wide zap object of temporary snapshot user holds
958 * and release them.
959 */
960 void
dsl_pool_clean_tmp_userrefs(dsl_pool_t * dp)961 dsl_pool_clean_tmp_userrefs(dsl_pool_t *dp)
962 {
963 zap_attribute_t za;
964 zap_cursor_t zc;
965 objset_t *mos = dp->dp_meta_objset;
966 uint64_t zapobj = dp->dp_tmp_userrefs_obj;
967 nvlist_t *holds;
968
969 if (zapobj == 0)
970 return;
971 ASSERT(spa_version(dp->dp_spa) >= SPA_VERSION_USERREFS);
972
973 holds = fnvlist_alloc();
974
975 for (zap_cursor_init(&zc, mos, zapobj);
976 zap_cursor_retrieve(&zc, &za) == 0;
977 zap_cursor_advance(&zc)) {
978 char *htag;
979 nvlist_t *tags;
980
981 htag = strchr(za.za_name, '-');
982 *htag = '\0';
983 ++htag;
984 if (nvlist_lookup_nvlist(holds, za.za_name, &tags) != 0) {
985 tags = fnvlist_alloc();
986 fnvlist_add_boolean(tags, htag);
987 fnvlist_add_nvlist(holds, za.za_name, tags);
988 fnvlist_free(tags);
989 } else {
990 fnvlist_add_boolean(tags, htag);
991 }
992 }
993 dsl_dataset_user_release_tmp(dp, holds);
994 fnvlist_free(holds);
995 zap_cursor_fini(&zc);
996 }
997
998 /*
999 * Create the pool-wide zap object for storing temporary snapshot holds.
1000 */
1001 void
dsl_pool_user_hold_create_obj(dsl_pool_t * dp,dmu_tx_t * tx)1002 dsl_pool_user_hold_create_obj(dsl_pool_t *dp, dmu_tx_t *tx)
1003 {
1004 objset_t *mos = dp->dp_meta_objset;
1005
1006 ASSERT(dp->dp_tmp_userrefs_obj == 0);
1007 ASSERT(dmu_tx_is_syncing(tx));
1008
1009 dp->dp_tmp_userrefs_obj = zap_create_link(mos, DMU_OT_USERREFS,
1010 DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_TMP_USERREFS, tx);
1011 }
1012
1013 static int
dsl_pool_user_hold_rele_impl(dsl_pool_t * dp,uint64_t dsobj,const char * tag,uint64_t now,dmu_tx_t * tx,boolean_t holding)1014 dsl_pool_user_hold_rele_impl(dsl_pool_t *dp, uint64_t dsobj,
1015 const char *tag, uint64_t now, dmu_tx_t *tx, boolean_t holding)
1016 {
1017 objset_t *mos = dp->dp_meta_objset;
1018 uint64_t zapobj = dp->dp_tmp_userrefs_obj;
1019 char *name;
1020 int error;
1021
1022 ASSERT(spa_version(dp->dp_spa) >= SPA_VERSION_USERREFS);
1023 ASSERT(dmu_tx_is_syncing(tx));
1024
1025 /*
1026 * If the pool was created prior to SPA_VERSION_USERREFS, the
1027 * zap object for temporary holds might not exist yet.
1028 */
1029 if (zapobj == 0) {
1030 if (holding) {
1031 dsl_pool_user_hold_create_obj(dp, tx);
1032 zapobj = dp->dp_tmp_userrefs_obj;
1033 } else {
1034 return (SET_ERROR(ENOENT));
1035 }
1036 }
1037
1038 name = kmem_asprintf("%llx-%s", (u_longlong_t)dsobj, tag);
1039 if (holding)
1040 error = zap_add(mos, zapobj, name, 8, 1, &now, tx);
1041 else
1042 error = zap_remove(mos, zapobj, name, tx);
1043 strfree(name);
1044
1045 return (error);
1046 }
1047
1048 /*
1049 * Add a temporary hold for the given dataset object and tag.
1050 */
1051 int
dsl_pool_user_hold(dsl_pool_t * dp,uint64_t dsobj,const char * tag,uint64_t now,dmu_tx_t * tx)1052 dsl_pool_user_hold(dsl_pool_t *dp, uint64_t dsobj, const char *tag,
1053 uint64_t now, dmu_tx_t *tx)
1054 {
1055 return (dsl_pool_user_hold_rele_impl(dp, dsobj, tag, now, tx, B_TRUE));
1056 }
1057
1058 /*
1059 * Release a temporary hold for the given dataset object and tag.
1060 */
1061 int
dsl_pool_user_release(dsl_pool_t * dp,uint64_t dsobj,const char * tag,dmu_tx_t * tx)1062 dsl_pool_user_release(dsl_pool_t *dp, uint64_t dsobj, const char *tag,
1063 dmu_tx_t *tx)
1064 {
1065 return (dsl_pool_user_hold_rele_impl(dp, dsobj, tag, 0,
1066 tx, B_FALSE));
1067 }
1068
1069 /*
1070 * DSL Pool Configuration Lock
1071 *
1072 * The dp_config_rwlock protects against changes to DSL state (e.g. dataset
1073 * creation / destruction / rename / property setting). It must be held for
1074 * read to hold a dataset or dsl_dir. I.e. you must call
1075 * dsl_pool_config_enter() or dsl_pool_hold() before calling
1076 * dsl_{dataset,dir}_hold{_obj}. In most circumstances, the dp_config_rwlock
1077 * must be held continuously until all datasets and dsl_dirs are released.
1078 *
1079 * The only exception to this rule is that if a "long hold" is placed on
1080 * a dataset, then the dp_config_rwlock may be dropped while the dataset
1081 * is still held. The long hold will prevent the dataset from being
1082 * destroyed -- the destroy will fail with EBUSY. A long hold can be
1083 * obtained by calling dsl_dataset_long_hold(), or by "owning" a dataset
1084 * (by calling dsl_{dataset,objset}_{try}own{_obj}).
1085 *
1086 * Legitimate long-holders (including owners) should be long-running, cancelable
1087 * tasks that should cause "zfs destroy" to fail. This includes DMU
1088 * consumers (i.e. a ZPL filesystem being mounted or ZVOL being open),
1089 * "zfs send", and "zfs diff". There are several other long-holders whose
1090 * uses are suboptimal (e.g. "zfs promote", and zil_suspend()).
1091 *
1092 * The usual formula for long-holding would be:
1093 * dsl_pool_hold()
1094 * dsl_dataset_hold()
1095 * ... perform checks ...
1096 * dsl_dataset_long_hold()
1097 * dsl_pool_rele()
1098 * ... perform long-running task ...
1099 * dsl_dataset_long_rele()
1100 * dsl_dataset_rele()
1101 *
1102 * Note that when the long hold is released, the dataset is still held but
1103 * the pool is not held. The dataset may change arbitrarily during this time
1104 * (e.g. it could be destroyed). Therefore you shouldn't do anything to the
1105 * dataset except release it.
1106 *
1107 * User-initiated operations (e.g. ioctls, zfs_ioc_*()) are either read-only
1108 * or modifying operations.
1109 *
1110 * Modifying operations should generally use dsl_sync_task(). The synctask
1111 * infrastructure enforces proper locking strategy with respect to the
1112 * dp_config_rwlock. See the comment above dsl_sync_task() for details.
1113 *
1114 * Read-only operations will manually hold the pool, then the dataset, obtain
1115 * information from the dataset, then release the pool and dataset.
1116 * dmu_objset_{hold,rele}() are convenience routines that also do the pool
1117 * hold/rele.
1118 */
1119
1120 int
dsl_pool_hold(const char * name,void * tag,dsl_pool_t ** dp)1121 dsl_pool_hold(const char *name, void *tag, dsl_pool_t **dp)
1122 {
1123 spa_t *spa;
1124 int error;
1125
1126 error = spa_open(name, &spa, tag);
1127 if (error == 0) {
1128 *dp = spa_get_dsl(spa);
1129 dsl_pool_config_enter(*dp, tag);
1130 }
1131 return (error);
1132 }
1133
1134 void
dsl_pool_rele(dsl_pool_t * dp,void * tag)1135 dsl_pool_rele(dsl_pool_t *dp, void *tag)
1136 {
1137 dsl_pool_config_exit(dp, tag);
1138 spa_close(dp->dp_spa, tag);
1139 }
1140
1141 void
dsl_pool_config_enter(dsl_pool_t * dp,void * tag)1142 dsl_pool_config_enter(dsl_pool_t *dp, void *tag)
1143 {
1144 /*
1145 * We use a "reentrant" reader-writer lock, but not reentrantly.
1146 *
1147 * The rrwlock can (with the track_all flag) track all reading threads,
1148 * which is very useful for debugging which code path failed to release
1149 * the lock, and for verifying that the *current* thread does hold
1150 * the lock.
1151 *
1152 * (Unlike a rwlock, which knows that N threads hold it for
1153 * read, but not *which* threads, so rw_held(RW_READER) returns TRUE
1154 * if any thread holds it for read, even if this thread doesn't).
1155 */
1156 ASSERT(!rrw_held(&dp->dp_config_rwlock, RW_READER));
1157 rrw_enter(&dp->dp_config_rwlock, RW_READER, tag);
1158 }
1159
1160 void
dsl_pool_config_enter_prio(dsl_pool_t * dp,void * tag)1161 dsl_pool_config_enter_prio(dsl_pool_t *dp, void *tag)
1162 {
1163 ASSERT(!rrw_held(&dp->dp_config_rwlock, RW_READER));
1164 rrw_enter_read_prio(&dp->dp_config_rwlock, tag);
1165 }
1166
1167 void
dsl_pool_config_exit(dsl_pool_t * dp,void * tag)1168 dsl_pool_config_exit(dsl_pool_t *dp, void *tag)
1169 {
1170 rrw_exit(&dp->dp_config_rwlock, tag);
1171 }
1172
1173 boolean_t
dsl_pool_config_held(dsl_pool_t * dp)1174 dsl_pool_config_held(dsl_pool_t *dp)
1175 {
1176 return (RRW_LOCK_HELD(&dp->dp_config_rwlock));
1177 }
1178
1179 boolean_t
dsl_pool_config_held_writer(dsl_pool_t * dp)1180 dsl_pool_config_held_writer(dsl_pool_t *dp)
1181 {
1182 return (RRW_WRITE_HELD(&dp->dp_config_rwlock));
1183 }
1184