]>
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. |
bc77ba73 | 23 | * Copyright (c) 2012, 2016 by Delphix. All rights reserved. |
3a17a7a9 | 24 | * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. |
788eb90c | 25 | * Copyright (c) 2013, Joyent, Inc. All rights reserved. |
0c66c32d | 26 | * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. |
1cddb8c9 | 27 | * Copyright (c) 2015 Nexenta Systems, Inc. All rights reserved. |
9c43027b | 28 | * Copyright (c) 2015, STRATO AG, Inc. All rights reserved. |
a0bd735a | 29 | * Copyright (c) 2016 Actifio, Inc. All rights reserved. |
34dc7c2f BB |
30 | */ |
31 | ||
428870ff BB |
32 | /* Portions Copyright 2010 Robert Milkowski */ |
33 | ||
34dc7c2f BB |
34 | #include <sys/cred.h> |
35 | #include <sys/zfs_context.h> | |
36 | #include <sys/dmu_objset.h> | |
37 | #include <sys/dsl_dir.h> | |
38 | #include <sys/dsl_dataset.h> | |
39 | #include <sys/dsl_prop.h> | |
40 | #include <sys/dsl_pool.h> | |
41 | #include <sys/dsl_synctask.h> | |
42 | #include <sys/dsl_deleg.h> | |
43 | #include <sys/dnode.h> | |
44 | #include <sys/dbuf.h> | |
45 | #include <sys/zvol.h> | |
46 | #include <sys/dmu_tx.h> | |
34dc7c2f BB |
47 | #include <sys/zap.h> |
48 | #include <sys/zil.h> | |
49 | #include <sys/dmu_impl.h> | |
50 | #include <sys/zfs_ioctl.h> | |
428870ff | 51 | #include <sys/sa.h> |
572e2857 | 52 | #include <sys/zfs_onexit.h> |
13fe0198 | 53 | #include <sys/dsl_destroy.h> |
9c43027b | 54 | #include <sys/vdev.h> |
f74b821a | 55 | #include <sys/policy.h> |
572e2857 BB |
56 | |
57 | /* | |
58 | * Needed to close a window in dnode_move() that allows the objset to be freed | |
59 | * before it can be safely accessed. | |
60 | */ | |
61 | krwlock_t os_lock; | |
62 | ||
9c43027b AJ |
63 | /* |
64 | * Tunable to overwrite the maximum number of threads for the parallization | |
65 | * of dmu_objset_find_dp, needed to speed up the import of pools with many | |
66 | * datasets. | |
67 | * Default is 4 times the number of leaf vdevs. | |
68 | */ | |
69 | int dmu_find_threads = 0; | |
70 | ||
68cbd56e NB |
71 | /* |
72 | * Backfill lower metadnode objects after this many have been freed. | |
73 | * Backfilling negatively impacts object creation rates, so only do it | |
74 | * if there are enough holes to fill. | |
75 | */ | |
76 | int dmu_rescan_dnode_threshold = 1 << DN_MAX_INDBLKSHIFT; | |
77 | ||
9c43027b AJ |
78 | static void dmu_objset_find_dp_cb(void *arg); |
79 | ||
572e2857 BB |
80 | void |
81 | dmu_objset_init(void) | |
82 | { | |
83 | rw_init(&os_lock, NULL, RW_DEFAULT, NULL); | |
84 | } | |
85 | ||
86 | void | |
87 | dmu_objset_fini(void) | |
88 | { | |
89 | rw_destroy(&os_lock); | |
90 | } | |
34dc7c2f BB |
91 | |
92 | spa_t * | |
93 | dmu_objset_spa(objset_t *os) | |
94 | { | |
428870ff | 95 | return (os->os_spa); |
34dc7c2f BB |
96 | } |
97 | ||
98 | zilog_t * | |
99 | dmu_objset_zil(objset_t *os) | |
100 | { | |
428870ff | 101 | return (os->os_zil); |
34dc7c2f BB |
102 | } |
103 | ||
104 | dsl_pool_t * | |
105 | dmu_objset_pool(objset_t *os) | |
106 | { | |
107 | dsl_dataset_t *ds; | |
108 | ||
428870ff | 109 | if ((ds = os->os_dsl_dataset) != NULL && ds->ds_dir) |
34dc7c2f BB |
110 | return (ds->ds_dir->dd_pool); |
111 | else | |
428870ff | 112 | return (spa_get_dsl(os->os_spa)); |
34dc7c2f BB |
113 | } |
114 | ||
115 | dsl_dataset_t * | |
116 | dmu_objset_ds(objset_t *os) | |
117 | { | |
428870ff | 118 | return (os->os_dsl_dataset); |
34dc7c2f BB |
119 | } |
120 | ||
121 | dmu_objset_type_t | |
122 | dmu_objset_type(objset_t *os) | |
123 | { | |
428870ff | 124 | return (os->os_phys->os_type); |
34dc7c2f BB |
125 | } |
126 | ||
127 | void | |
128 | dmu_objset_name(objset_t *os, char *buf) | |
129 | { | |
428870ff | 130 | dsl_dataset_name(os->os_dsl_dataset, buf); |
34dc7c2f BB |
131 | } |
132 | ||
133 | uint64_t | |
134 | dmu_objset_id(objset_t *os) | |
135 | { | |
428870ff | 136 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
137 | |
138 | return (ds ? ds->ds_object : 0); | |
139 | } | |
140 | ||
50c957f7 NB |
141 | uint64_t |
142 | dmu_objset_dnodesize(objset_t *os) | |
143 | { | |
144 | return (os->os_dnodesize); | |
145 | } | |
146 | ||
faf0f58c | 147 | zfs_sync_type_t |
428870ff BB |
148 | dmu_objset_syncprop(objset_t *os) |
149 | { | |
150 | return (os->os_sync); | |
151 | } | |
152 | ||
faf0f58c | 153 | zfs_logbias_op_t |
428870ff BB |
154 | dmu_objset_logbias(objset_t *os) |
155 | { | |
156 | return (os->os_logbias); | |
157 | } | |
158 | ||
34dc7c2f BB |
159 | static void |
160 | checksum_changed_cb(void *arg, uint64_t newval) | |
161 | { | |
428870ff | 162 | objset_t *os = arg; |
34dc7c2f BB |
163 | |
164 | /* | |
165 | * Inheritance should have been done by now. | |
166 | */ | |
167 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
168 | ||
428870ff | 169 | os->os_checksum = zio_checksum_select(newval, ZIO_CHECKSUM_ON_VALUE); |
34dc7c2f BB |
170 | } |
171 | ||
172 | static void | |
173 | compression_changed_cb(void *arg, uint64_t newval) | |
174 | { | |
428870ff | 175 | objset_t *os = arg; |
34dc7c2f BB |
176 | |
177 | /* | |
178 | * Inheritance and range checking should have been done by now. | |
179 | */ | |
180 | ASSERT(newval != ZIO_COMPRESS_INHERIT); | |
181 | ||
99197f03 JG |
182 | os->os_compress = zio_compress_select(os->os_spa, newval, |
183 | ZIO_COMPRESS_ON); | |
34dc7c2f BB |
184 | } |
185 | ||
186 | static void | |
187 | copies_changed_cb(void *arg, uint64_t newval) | |
188 | { | |
428870ff | 189 | objset_t *os = arg; |
34dc7c2f BB |
190 | |
191 | /* | |
192 | * Inheritance and range checking should have been done by now. | |
193 | */ | |
194 | ASSERT(newval > 0); | |
428870ff | 195 | ASSERT(newval <= spa_max_replication(os->os_spa)); |
34dc7c2f | 196 | |
428870ff BB |
197 | os->os_copies = newval; |
198 | } | |
199 | ||
200 | static void | |
201 | dedup_changed_cb(void *arg, uint64_t newval) | |
202 | { | |
203 | objset_t *os = arg; | |
204 | spa_t *spa = os->os_spa; | |
205 | enum zio_checksum checksum; | |
206 | ||
207 | /* | |
208 | * Inheritance should have been done by now. | |
209 | */ | |
210 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
211 | ||
212 | checksum = zio_checksum_dedup_select(spa, newval, ZIO_CHECKSUM_OFF); | |
213 | ||
214 | os->os_dedup_checksum = checksum & ZIO_CHECKSUM_MASK; | |
215 | os->os_dedup_verify = !!(checksum & ZIO_CHECKSUM_VERIFY); | |
34dc7c2f BB |
216 | } |
217 | ||
b128c09f BB |
218 | static void |
219 | primary_cache_changed_cb(void *arg, uint64_t newval) | |
220 | { | |
428870ff | 221 | objset_t *os = arg; |
b128c09f BB |
222 | |
223 | /* | |
224 | * Inheritance and range checking should have been done by now. | |
225 | */ | |
226 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
227 | newval == ZFS_CACHE_METADATA); | |
228 | ||
428870ff | 229 | os->os_primary_cache = newval; |
b128c09f BB |
230 | } |
231 | ||
232 | static void | |
233 | secondary_cache_changed_cb(void *arg, uint64_t newval) | |
234 | { | |
428870ff | 235 | objset_t *os = arg; |
b128c09f BB |
236 | |
237 | /* | |
238 | * Inheritance and range checking should have been done by now. | |
239 | */ | |
240 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
241 | newval == ZFS_CACHE_METADATA); | |
242 | ||
428870ff BB |
243 | os->os_secondary_cache = newval; |
244 | } | |
245 | ||
246 | static void | |
247 | sync_changed_cb(void *arg, uint64_t newval) | |
248 | { | |
249 | objset_t *os = arg; | |
250 | ||
251 | /* | |
252 | * Inheritance and range checking should have been done by now. | |
253 | */ | |
254 | ASSERT(newval == ZFS_SYNC_STANDARD || newval == ZFS_SYNC_ALWAYS || | |
255 | newval == ZFS_SYNC_DISABLED); | |
256 | ||
257 | os->os_sync = newval; | |
258 | if (os->os_zil) | |
259 | zil_set_sync(os->os_zil, newval); | |
260 | } | |
261 | ||
faf0f58c MA |
262 | static void |
263 | redundant_metadata_changed_cb(void *arg, uint64_t newval) | |
264 | { | |
265 | objset_t *os = arg; | |
266 | ||
267 | /* | |
268 | * Inheritance and range checking should have been done by now. | |
269 | */ | |
270 | ASSERT(newval == ZFS_REDUNDANT_METADATA_ALL || | |
271 | newval == ZFS_REDUNDANT_METADATA_MOST); | |
272 | ||
273 | os->os_redundant_metadata = newval; | |
274 | } | |
275 | ||
50c957f7 NB |
276 | static void |
277 | dnodesize_changed_cb(void *arg, uint64_t newval) | |
278 | { | |
279 | objset_t *os = arg; | |
280 | ||
281 | switch (newval) { | |
282 | case ZFS_DNSIZE_LEGACY: | |
283 | os->os_dnodesize = DNODE_MIN_SIZE; | |
284 | break; | |
285 | case ZFS_DNSIZE_AUTO: | |
286 | /* | |
287 | * Choose a dnode size that will work well for most | |
288 | * workloads if the user specified "auto". Future code | |
289 | * improvements could dynamically select a dnode size | |
290 | * based on observed workload patterns. | |
291 | */ | |
292 | os->os_dnodesize = DNODE_MIN_SIZE * 2; | |
293 | break; | |
294 | case ZFS_DNSIZE_1K: | |
295 | case ZFS_DNSIZE_2K: | |
296 | case ZFS_DNSIZE_4K: | |
297 | case ZFS_DNSIZE_8K: | |
298 | case ZFS_DNSIZE_16K: | |
299 | os->os_dnodesize = newval; | |
300 | break; | |
301 | } | |
302 | } | |
303 | ||
428870ff BB |
304 | static void |
305 | logbias_changed_cb(void *arg, uint64_t newval) | |
306 | { | |
307 | objset_t *os = arg; | |
308 | ||
309 | ASSERT(newval == ZFS_LOGBIAS_LATENCY || | |
310 | newval == ZFS_LOGBIAS_THROUGHPUT); | |
311 | os->os_logbias = newval; | |
312 | if (os->os_zil) | |
313 | zil_set_logbias(os->os_zil, newval); | |
b128c09f BB |
314 | } |
315 | ||
f1512ee6 MA |
316 | static void |
317 | recordsize_changed_cb(void *arg, uint64_t newval) | |
318 | { | |
319 | objset_t *os = arg; | |
320 | ||
321 | os->os_recordsize = newval; | |
322 | } | |
323 | ||
34dc7c2f BB |
324 | void |
325 | dmu_objset_byteswap(void *buf, size_t size) | |
326 | { | |
327 | objset_phys_t *osp = buf; | |
328 | ||
9babb374 | 329 | ASSERT(size == OBJSET_OLD_PHYS_SIZE || size == sizeof (objset_phys_t)); |
34dc7c2f BB |
330 | dnode_byteswap(&osp->os_meta_dnode); |
331 | byteswap_uint64_array(&osp->os_zil_header, sizeof (zil_header_t)); | |
332 | osp->os_type = BSWAP_64(osp->os_type); | |
9babb374 BB |
333 | osp->os_flags = BSWAP_64(osp->os_flags); |
334 | if (size == sizeof (objset_phys_t)) { | |
335 | dnode_byteswap(&osp->os_userused_dnode); | |
336 | dnode_byteswap(&osp->os_groupused_dnode); | |
337 | } | |
34dc7c2f BB |
338 | } |
339 | ||
340 | int | |
341 | dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, | |
428870ff | 342 | objset_t **osp) |
34dc7c2f | 343 | { |
428870ff | 344 | objset_t *os; |
b128c09f | 345 | int i, err; |
34dc7c2f BB |
346 | |
347 | ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock)); | |
348 | ||
79c76d5b | 349 | os = kmem_zalloc(sizeof (objset_t), KM_SLEEP); |
428870ff BB |
350 | os->os_dsl_dataset = ds; |
351 | os->os_spa = spa; | |
352 | os->os_rootbp = bp; | |
353 | if (!BP_IS_HOLE(os->os_rootbp)) { | |
2a432414 | 354 | arc_flags_t aflags = ARC_FLAG_WAIT; |
5dbd68a3 | 355 | zbookmark_phys_t zb; |
428870ff BB |
356 | SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET, |
357 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
358 | ||
359 | if (DMU_OS_IS_L2CACHEABLE(os)) | |
2a432414 | 360 | aflags |= ARC_FLAG_L2CACHE; |
34dc7c2f | 361 | |
428870ff | 362 | dprintf_bp(os->os_rootbp, "reading %s", ""); |
294f6806 | 363 | err = arc_read(NULL, spa, os->os_rootbp, |
428870ff | 364 | arc_getbuf_func, &os->os_phys_buf, |
34dc7c2f | 365 | ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL, &aflags, &zb); |
13fe0198 | 366 | if (err != 0) { |
428870ff | 367 | kmem_free(os, sizeof (objset_t)); |
b128c09f BB |
368 | /* convert checksum errors into IO errors */ |
369 | if (err == ECKSUM) | |
2e528b49 | 370 | err = SET_ERROR(EIO); |
34dc7c2f BB |
371 | return (err); |
372 | } | |
9babb374 BB |
373 | |
374 | /* Increase the blocksize if we are permitted. */ | |
375 | if (spa_version(spa) >= SPA_VERSION_USERSPACE && | |
428870ff | 376 | arc_buf_size(os->os_phys_buf) < sizeof (objset_phys_t)) { |
2aa34383 DK |
377 | arc_buf_t *buf = arc_alloc_buf(spa, &os->os_phys_buf, |
378 | ARC_BUFC_METADATA, sizeof (objset_phys_t)); | |
9babb374 | 379 | bzero(buf->b_data, sizeof (objset_phys_t)); |
428870ff BB |
380 | bcopy(os->os_phys_buf->b_data, buf->b_data, |
381 | arc_buf_size(os->os_phys_buf)); | |
d3c2ae1c | 382 | arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); |
428870ff | 383 | os->os_phys_buf = buf; |
9babb374 BB |
384 | } |
385 | ||
428870ff BB |
386 | os->os_phys = os->os_phys_buf->b_data; |
387 | os->os_flags = os->os_phys->os_flags; | |
34dc7c2f | 388 | } else { |
9babb374 BB |
389 | int size = spa_version(spa) >= SPA_VERSION_USERSPACE ? |
390 | sizeof (objset_phys_t) : OBJSET_OLD_PHYS_SIZE; | |
2aa34383 DK |
391 | os->os_phys_buf = arc_alloc_buf(spa, &os->os_phys_buf, |
392 | ARC_BUFC_METADATA, size); | |
428870ff BB |
393 | os->os_phys = os->os_phys_buf->b_data; |
394 | bzero(os->os_phys, size); | |
34dc7c2f BB |
395 | } |
396 | ||
397 | /* | |
398 | * Note: the changed_cb will be called once before the register | |
399 | * func returns, thus changing the checksum/compression from the | |
b128c09f BB |
400 | * default (fletcher2/off). Snapshots don't need to know about |
401 | * checksum/compression/copies. | |
34dc7c2f | 402 | */ |
9b67f605 | 403 | if (ds != NULL) { |
47dfff3b MA |
404 | boolean_t needlock = B_FALSE; |
405 | ||
406 | /* | |
407 | * Note: it's valid to open the objset if the dataset is | |
408 | * long-held, in which case the pool_config lock will not | |
409 | * be held. | |
410 | */ | |
411 | if (!dsl_pool_config_held(dmu_objset_pool(os))) { | |
412 | needlock = B_TRUE; | |
413 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); | |
414 | } | |
13fe0198 MA |
415 | err = dsl_prop_register(ds, |
416 | zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE), | |
428870ff | 417 | primary_cache_changed_cb, os); |
13fe0198 MA |
418 | if (err == 0) { |
419 | err = dsl_prop_register(ds, | |
420 | zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE), | |
428870ff | 421 | secondary_cache_changed_cb, os); |
13fe0198 | 422 | } |
0c66c32d | 423 | if (!ds->ds_is_snapshot) { |
13fe0198 MA |
424 | if (err == 0) { |
425 | err = dsl_prop_register(ds, | |
426 | zfs_prop_to_name(ZFS_PROP_CHECKSUM), | |
428870ff | 427 | checksum_changed_cb, os); |
13fe0198 MA |
428 | } |
429 | if (err == 0) { | |
430 | err = dsl_prop_register(ds, | |
431 | zfs_prop_to_name(ZFS_PROP_COMPRESSION), | |
428870ff | 432 | compression_changed_cb, os); |
13fe0198 MA |
433 | } |
434 | if (err == 0) { | |
435 | err = dsl_prop_register(ds, | |
436 | zfs_prop_to_name(ZFS_PROP_COPIES), | |
428870ff | 437 | copies_changed_cb, os); |
13fe0198 MA |
438 | } |
439 | if (err == 0) { | |
440 | err = dsl_prop_register(ds, | |
441 | zfs_prop_to_name(ZFS_PROP_DEDUP), | |
428870ff | 442 | dedup_changed_cb, os); |
13fe0198 MA |
443 | } |
444 | if (err == 0) { | |
445 | err = dsl_prop_register(ds, | |
446 | zfs_prop_to_name(ZFS_PROP_LOGBIAS), | |
428870ff | 447 | logbias_changed_cb, os); |
13fe0198 MA |
448 | } |
449 | if (err == 0) { | |
450 | err = dsl_prop_register(ds, | |
451 | zfs_prop_to_name(ZFS_PROP_SYNC), | |
428870ff | 452 | sync_changed_cb, os); |
13fe0198 | 453 | } |
faf0f58c MA |
454 | if (err == 0) { |
455 | err = dsl_prop_register(ds, | |
456 | zfs_prop_to_name( | |
457 | ZFS_PROP_REDUNDANT_METADATA), | |
458 | redundant_metadata_changed_cb, os); | |
459 | } | |
f1512ee6 MA |
460 | if (err == 0) { |
461 | err = dsl_prop_register(ds, | |
462 | zfs_prop_to_name(ZFS_PROP_RECORDSIZE), | |
463 | recordsize_changed_cb, os); | |
464 | } | |
50c957f7 NB |
465 | if (err == 0) { |
466 | err = dsl_prop_register(ds, | |
467 | zfs_prop_to_name(ZFS_PROP_DNODESIZE), | |
468 | dnodesize_changed_cb, os); | |
469 | } | |
b128c09f | 470 | } |
47dfff3b MA |
471 | if (needlock) |
472 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
13fe0198 | 473 | if (err != 0) { |
d3c2ae1c | 474 | arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); |
428870ff | 475 | kmem_free(os, sizeof (objset_t)); |
34dc7c2f BB |
476 | return (err); |
477 | } | |
9b67f605 | 478 | } else { |
34dc7c2f | 479 | /* It's the meta-objset. */ |
428870ff | 480 | os->os_checksum = ZIO_CHECKSUM_FLETCHER_4; |
99197f03 | 481 | os->os_compress = ZIO_COMPRESS_ON; |
428870ff BB |
482 | os->os_copies = spa_max_replication(spa); |
483 | os->os_dedup_checksum = ZIO_CHECKSUM_OFF; | |
faf0f58c MA |
484 | os->os_dedup_verify = B_FALSE; |
485 | os->os_logbias = ZFS_LOGBIAS_LATENCY; | |
486 | os->os_sync = ZFS_SYNC_STANDARD; | |
428870ff BB |
487 | os->os_primary_cache = ZFS_CACHE_ALL; |
488 | os->os_secondary_cache = ZFS_CACHE_ALL; | |
50c957f7 | 489 | os->os_dnodesize = DNODE_MIN_SIZE; |
34dc7c2f BB |
490 | } |
491 | ||
0c66c32d | 492 | if (ds == NULL || !ds->ds_is_snapshot) |
572e2857 | 493 | os->os_zil_header = os->os_phys->os_zil_header; |
428870ff | 494 | os->os_zil = zil_alloc(os, &os->os_zil_header); |
34dc7c2f BB |
495 | |
496 | for (i = 0; i < TXG_SIZE; i++) { | |
428870ff | 497 | list_create(&os->os_dirty_dnodes[i], sizeof (dnode_t), |
34dc7c2f | 498 | offsetof(dnode_t, dn_dirty_link[i])); |
428870ff | 499 | list_create(&os->os_free_dnodes[i], sizeof (dnode_t), |
34dc7c2f BB |
500 | offsetof(dnode_t, dn_dirty_link[i])); |
501 | } | |
428870ff | 502 | list_create(&os->os_dnodes, sizeof (dnode_t), |
34dc7c2f | 503 | offsetof(dnode_t, dn_link)); |
428870ff | 504 | list_create(&os->os_downgraded_dbufs, sizeof (dmu_buf_impl_t), |
34dc7c2f BB |
505 | offsetof(dmu_buf_impl_t, db_link)); |
506 | ||
0c66c32d JG |
507 | list_link_init(&os->os_evicting_node); |
508 | ||
428870ff BB |
509 | mutex_init(&os->os_lock, NULL, MUTEX_DEFAULT, NULL); |
510 | mutex_init(&os->os_obj_lock, NULL, MUTEX_DEFAULT, NULL); | |
511 | mutex_init(&os->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL); | |
512 | ||
0c66c32d JG |
513 | dnode_special_open(os, &os->os_phys->os_meta_dnode, |
514 | DMU_META_DNODE_OBJECT, &os->os_meta_dnode); | |
428870ff | 515 | if (arc_buf_size(os->os_phys_buf) >= sizeof (objset_phys_t)) { |
0c66c32d JG |
516 | dnode_special_open(os, &os->os_phys->os_userused_dnode, |
517 | DMU_USERUSED_OBJECT, &os->os_userused_dnode); | |
518 | dnode_special_open(os, &os->os_phys->os_groupused_dnode, | |
519 | DMU_GROUPUSED_OBJECT, &os->os_groupused_dnode); | |
9babb374 | 520 | } |
34dc7c2f | 521 | |
428870ff | 522 | *osp = os; |
34dc7c2f BB |
523 | return (0); |
524 | } | |
525 | ||
428870ff BB |
526 | int |
527 | dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp) | |
34dc7c2f | 528 | { |
428870ff | 529 | int err = 0; |
34dc7c2f | 530 | |
47dfff3b MA |
531 | /* |
532 | * We shouldn't be doing anything with dsl_dataset_t's unless the | |
533 | * pool_config lock is held, or the dataset is long-held. | |
534 | */ | |
535 | ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool) || | |
536 | dsl_dataset_long_held(ds)); | |
537 | ||
34dc7c2f | 538 | mutex_enter(&ds->ds_opening_lock); |
9b67f605 MA |
539 | if (ds->ds_objset == NULL) { |
540 | objset_t *os; | |
34dc7c2f | 541 | err = dmu_objset_open_impl(dsl_dataset_get_spa(ds), |
9b67f605 MA |
542 | ds, dsl_dataset_get_blkptr(ds), &os); |
543 | ||
544 | if (err == 0) { | |
545 | mutex_enter(&ds->ds_lock); | |
546 | ASSERT(ds->ds_objset == NULL); | |
547 | ds->ds_objset = os; | |
548 | mutex_exit(&ds->ds_lock); | |
549 | } | |
34dc7c2f | 550 | } |
9b67f605 | 551 | *osp = ds->ds_objset; |
34dc7c2f | 552 | mutex_exit(&ds->ds_opening_lock); |
428870ff | 553 | return (err); |
34dc7c2f BB |
554 | } |
555 | ||
13fe0198 MA |
556 | /* |
557 | * Holds the pool while the objset is held. Therefore only one objset | |
558 | * can be held at a time. | |
559 | */ | |
34dc7c2f | 560 | int |
428870ff | 561 | dmu_objset_hold(const char *name, void *tag, objset_t **osp) |
34dc7c2f | 562 | { |
13fe0198 | 563 | dsl_pool_t *dp; |
428870ff | 564 | dsl_dataset_t *ds; |
34dc7c2f BB |
565 | int err; |
566 | ||
13fe0198 MA |
567 | err = dsl_pool_hold(name, tag, &dp); |
568 | if (err != 0) | |
569 | return (err); | |
570 | err = dsl_dataset_hold(dp, name, tag, &ds); | |
571 | if (err != 0) { | |
572 | dsl_pool_rele(dp, tag); | |
428870ff | 573 | return (err); |
13fe0198 | 574 | } |
428870ff BB |
575 | |
576 | err = dmu_objset_from_ds(ds, osp); | |
13fe0198 | 577 | if (err != 0) { |
428870ff | 578 | dsl_dataset_rele(ds, tag); |
13fe0198 MA |
579 | dsl_pool_rele(dp, tag); |
580 | } | |
428870ff | 581 | |
34dc7c2f BB |
582 | return (err); |
583 | } | |
584 | ||
9c43027b AJ |
585 | static int |
586 | dmu_objset_own_impl(dsl_dataset_t *ds, dmu_objset_type_t type, | |
587 | boolean_t readonly, void *tag, objset_t **osp) | |
588 | { | |
589 | int err; | |
590 | ||
591 | err = dmu_objset_from_ds(ds, osp); | |
592 | if (err != 0) { | |
593 | dsl_dataset_disown(ds, tag); | |
594 | } else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) { | |
595 | dsl_dataset_disown(ds, tag); | |
596 | return (SET_ERROR(EINVAL)); | |
597 | } else if (!readonly && dsl_dataset_is_snapshot(ds)) { | |
598 | dsl_dataset_disown(ds, tag); | |
599 | return (SET_ERROR(EROFS)); | |
600 | } | |
601 | return (err); | |
602 | } | |
603 | ||
13fe0198 MA |
604 | /* |
605 | * dsl_pool must not be held when this is called. | |
606 | * Upon successful return, there will be a longhold on the dataset, | |
607 | * and the dsl_pool will not be held. | |
608 | */ | |
34dc7c2f | 609 | int |
428870ff BB |
610 | dmu_objset_own(const char *name, dmu_objset_type_t type, |
611 | boolean_t readonly, void *tag, objset_t **osp) | |
34dc7c2f | 612 | { |
13fe0198 | 613 | dsl_pool_t *dp; |
34dc7c2f BB |
614 | dsl_dataset_t *ds; |
615 | int err; | |
616 | ||
13fe0198 MA |
617 | err = dsl_pool_hold(name, FTAG, &dp); |
618 | if (err != 0) | |
619 | return (err); | |
620 | err = dsl_dataset_own(dp, name, tag, &ds); | |
621 | if (err != 0) { | |
622 | dsl_pool_rele(dp, FTAG); | |
34dc7c2f | 623 | return (err); |
13fe0198 | 624 | } |
9c43027b | 625 | err = dmu_objset_own_impl(ds, type, readonly, tag, osp); |
13fe0198 | 626 | dsl_pool_rele(dp, FTAG); |
9c43027b | 627 | |
34dc7c2f BB |
628 | return (err); |
629 | } | |
630 | ||
9c43027b AJ |
631 | int |
632 | dmu_objset_own_obj(dsl_pool_t *dp, uint64_t obj, dmu_objset_type_t type, | |
633 | boolean_t readonly, void *tag, objset_t **osp) | |
634 | { | |
635 | dsl_dataset_t *ds; | |
636 | int err; | |
637 | ||
638 | err = dsl_dataset_own_obj(dp, obj, tag, &ds); | |
639 | if (err != 0) | |
640 | return (err); | |
641 | ||
642 | return (dmu_objset_own_impl(ds, type, readonly, tag, osp)); | |
643 | } | |
644 | ||
34dc7c2f | 645 | void |
428870ff | 646 | dmu_objset_rele(objset_t *os, void *tag) |
34dc7c2f | 647 | { |
13fe0198 | 648 | dsl_pool_t *dp = dmu_objset_pool(os); |
428870ff | 649 | dsl_dataset_rele(os->os_dsl_dataset, tag); |
13fe0198 | 650 | dsl_pool_rele(dp, tag); |
428870ff | 651 | } |
b128c09f | 652 | |
831baf06 KW |
653 | /* |
654 | * When we are called, os MUST refer to an objset associated with a dataset | |
655 | * that is owned by 'tag'; that is, is held and long held by 'tag' and ds_owner | |
656 | * == tag. We will then release and reacquire ownership of the dataset while | |
657 | * holding the pool config_rwlock to avoid intervening namespace or ownership | |
658 | * changes may occur. | |
659 | * | |
660 | * This exists solely to accommodate zfs_ioc_userspace_upgrade()'s desire to | |
661 | * release the hold on its dataset and acquire a new one on the dataset of the | |
662 | * same name so that it can be partially torn down and reconstructed. | |
663 | */ | |
664 | void | |
665 | dmu_objset_refresh_ownership(objset_t *os, void *tag) | |
666 | { | |
667 | dsl_pool_t *dp; | |
668 | dsl_dataset_t *ds, *newds; | |
eca7b760 | 669 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
831baf06 KW |
670 | |
671 | ds = os->os_dsl_dataset; | |
672 | VERIFY3P(ds, !=, NULL); | |
673 | VERIFY3P(ds->ds_owner, ==, tag); | |
674 | VERIFY(dsl_dataset_long_held(ds)); | |
675 | ||
676 | dsl_dataset_name(ds, name); | |
677 | dp = dmu_objset_pool(os); | |
678 | dsl_pool_config_enter(dp, FTAG); | |
679 | dmu_objset_disown(os, tag); | |
680 | VERIFY0(dsl_dataset_own(dp, name, tag, &newds)); | |
681 | VERIFY3P(newds, ==, os->os_dsl_dataset); | |
682 | dsl_pool_config_exit(dp, FTAG); | |
683 | } | |
684 | ||
428870ff BB |
685 | void |
686 | dmu_objset_disown(objset_t *os, void *tag) | |
687 | { | |
688 | dsl_dataset_disown(os->os_dsl_dataset, tag); | |
34dc7c2f BB |
689 | } |
690 | ||
13fe0198 | 691 | void |
34dc7c2f BB |
692 | dmu_objset_evict_dbufs(objset_t *os) |
693 | { | |
0c66c32d | 694 | dnode_t *dn_marker; |
34dc7c2f BB |
695 | dnode_t *dn; |
696 | ||
0c66c32d | 697 | dn_marker = kmem_alloc(sizeof (dnode_t), KM_SLEEP); |
34dc7c2f | 698 | |
0c66c32d JG |
699 | mutex_enter(&os->os_lock); |
700 | dn = list_head(&os->os_dnodes); | |
701 | while (dn != NULL) { | |
702 | /* | |
703 | * Skip dnodes without holds. We have to do this dance | |
704 | * because dnode_add_ref() only works if there is already a | |
705 | * hold. If the dnode has no holds, then it has no dbufs. | |
706 | */ | |
707 | if (dnode_add_ref(dn, FTAG)) { | |
708 | list_insert_after(&os->os_dnodes, dn, dn_marker); | |
709 | mutex_exit(&os->os_lock); | |
34dc7c2f | 710 | |
0c66c32d JG |
711 | dnode_evict_dbufs(dn); |
712 | dnode_rele(dn, FTAG); | |
34dc7c2f | 713 | |
0c66c32d JG |
714 | mutex_enter(&os->os_lock); |
715 | dn = list_next(&os->os_dnodes, dn_marker); | |
716 | list_remove(&os->os_dnodes, dn_marker); | |
717 | } else { | |
718 | dn = list_next(&os->os_dnodes, dn); | |
719 | } | |
720 | } | |
721 | mutex_exit(&os->os_lock); | |
34dc7c2f | 722 | |
0c66c32d | 723 | kmem_free(dn_marker, sizeof (dnode_t)); |
34dc7c2f | 724 | |
0c66c32d JG |
725 | if (DMU_USERUSED_DNODE(os) != NULL) { |
726 | dnode_evict_dbufs(DMU_GROUPUSED_DNODE(os)); | |
727 | dnode_evict_dbufs(DMU_USERUSED_DNODE(os)); | |
34dc7c2f | 728 | } |
0c66c32d | 729 | dnode_evict_dbufs(DMU_META_DNODE(os)); |
34dc7c2f BB |
730 | } |
731 | ||
0c66c32d JG |
732 | /* |
733 | * Objset eviction processing is split into into two pieces. | |
734 | * The first marks the objset as evicting, evicts any dbufs that | |
735 | * have a refcount of zero, and then queues up the objset for the | |
736 | * second phase of eviction. Once os->os_dnodes has been cleared by | |
737 | * dnode_buf_pageout()->dnode_destroy(), the second phase is executed. | |
738 | * The second phase closes the special dnodes, dequeues the objset from | |
739 | * the list of those undergoing eviction, and finally frees the objset. | |
740 | * | |
741 | * NOTE: Due to asynchronous eviction processing (invocation of | |
742 | * dnode_buf_pageout()), it is possible for the meta dnode for the | |
743 | * objset to have no holds even though os->os_dnodes is not empty. | |
744 | */ | |
34dc7c2f | 745 | void |
428870ff | 746 | dmu_objset_evict(objset_t *os) |
34dc7c2f | 747 | { |
d6320ddb | 748 | int t; |
34dc7c2f | 749 | |
6f1ffb06 MA |
750 | dsl_dataset_t *ds = os->os_dsl_dataset; |
751 | ||
d6320ddb | 752 | for (t = 0; t < TXG_SIZE; t++) |
428870ff | 753 | ASSERT(!dmu_objset_is_dirty(os, t)); |
34dc7c2f | 754 | |
0eb21616 JG |
755 | if (ds) |
756 | dsl_prop_unregister_all(ds, os); | |
34dc7c2f | 757 | |
428870ff BB |
758 | if (os->os_sa) |
759 | sa_tear_down(os); | |
760 | ||
13fe0198 | 761 | dmu_objset_evict_dbufs(os); |
34dc7c2f | 762 | |
0c66c32d JG |
763 | mutex_enter(&os->os_lock); |
764 | spa_evicting_os_register(os->os_spa, os); | |
765 | if (list_is_empty(&os->os_dnodes)) { | |
766 | mutex_exit(&os->os_lock); | |
767 | dmu_objset_evict_done(os); | |
768 | } else { | |
769 | mutex_exit(&os->os_lock); | |
770 | } | |
771 | } | |
772 | ||
773 | void | |
774 | dmu_objset_evict_done(objset_t *os) | |
775 | { | |
776 | ASSERT3P(list_head(&os->os_dnodes), ==, NULL); | |
777 | ||
572e2857 BB |
778 | dnode_special_close(&os->os_meta_dnode); |
779 | if (DMU_USERUSED_DNODE(os)) { | |
780 | dnode_special_close(&os->os_userused_dnode); | |
781 | dnode_special_close(&os->os_groupused_dnode); | |
9babb374 | 782 | } |
428870ff BB |
783 | zil_free(os->os_zil); |
784 | ||
d3c2ae1c | 785 | arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); |
572e2857 BB |
786 | |
787 | /* | |
788 | * This is a barrier to prevent the objset from going away in | |
789 | * dnode_move() until we can safely ensure that the objset is still in | |
790 | * use. We consider the objset valid before the barrier and invalid | |
791 | * after the barrier. | |
792 | */ | |
793 | rw_enter(&os_lock, RW_READER); | |
794 | rw_exit(&os_lock); | |
795 | ||
428870ff BB |
796 | mutex_destroy(&os->os_lock); |
797 | mutex_destroy(&os->os_obj_lock); | |
798 | mutex_destroy(&os->os_user_ptr_lock); | |
0c66c32d | 799 | spa_evicting_os_deregister(os->os_spa, os); |
428870ff BB |
800 | kmem_free(os, sizeof (objset_t)); |
801 | } | |
9babb374 | 802 | |
428870ff BB |
803 | timestruc_t |
804 | dmu_objset_snap_cmtime(objset_t *os) | |
805 | { | |
806 | return (dsl_dir_snap_cmtime(os->os_dsl_dataset->ds_dir)); | |
34dc7c2f BB |
807 | } |
808 | ||
809 | /* called from dsl for meta-objset */ | |
428870ff | 810 | objset_t * |
34dc7c2f BB |
811 | dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, |
812 | dmu_objset_type_t type, dmu_tx_t *tx) | |
813 | { | |
428870ff | 814 | objset_t *os; |
34dc7c2f BB |
815 | dnode_t *mdn; |
816 | ||
817 | ASSERT(dmu_tx_is_syncing(tx)); | |
13fe0198 | 818 | |
572e2857 | 819 | if (ds != NULL) |
13fe0198 | 820 | VERIFY0(dmu_objset_from_ds(ds, &os)); |
572e2857 | 821 | else |
13fe0198 | 822 | VERIFY0(dmu_objset_open_impl(spa, NULL, bp, &os)); |
572e2857 BB |
823 | |
824 | mdn = DMU_META_DNODE(os); | |
34dc7c2f | 825 | |
50c957f7 NB |
826 | dnode_allocate(mdn, DMU_OT_DNODE, DNODE_BLOCK_SIZE, DN_MAX_INDBLKSHIFT, |
827 | DMU_OT_NONE, 0, DNODE_MIN_SLOTS, tx); | |
34dc7c2f BB |
828 | |
829 | /* | |
830 | * We don't want to have to increase the meta-dnode's nlevels | |
831 | * later, because then we could do it in quescing context while | |
832 | * we are also accessing it in open context. | |
833 | * | |
834 | * This precaution is not necessary for the MOS (ds == NULL), | |
835 | * because the MOS is only updated in syncing context. | |
836 | * This is most fortunate: the MOS is the only objset that | |
837 | * needs to be synced multiple times as spa_sync() iterates | |
838 | * to convergence, so minimizing its dn_nlevels matters. | |
839 | */ | |
840 | if (ds != NULL) { | |
841 | int levels = 1; | |
842 | ||
843 | /* | |
844 | * Determine the number of levels necessary for the meta-dnode | |
845 | * to contain DN_MAX_OBJECT dnodes. | |
846 | */ | |
847 | while ((uint64_t)mdn->dn_nblkptr << (mdn->dn_datablkshift + | |
848 | (levels - 1) * (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) < | |
849 | DN_MAX_OBJECT * sizeof (dnode_phys_t)) | |
850 | levels++; | |
851 | ||
852 | mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] = | |
853 | mdn->dn_nlevels = levels; | |
854 | } | |
855 | ||
856 | ASSERT(type != DMU_OST_NONE); | |
857 | ASSERT(type != DMU_OST_ANY); | |
858 | ASSERT(type < DMU_OST_NUMTYPES); | |
428870ff BB |
859 | os->os_phys->os_type = type; |
860 | if (dmu_objset_userused_enabled(os)) { | |
861 | os->os_phys->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; | |
862 | os->os_flags = os->os_phys->os_flags; | |
9babb374 | 863 | } |
34dc7c2f BB |
864 | |
865 | dsl_dataset_dirty(ds, tx); | |
866 | ||
428870ff | 867 | return (os); |
34dc7c2f BB |
868 | } |
869 | ||
13fe0198 MA |
870 | typedef struct dmu_objset_create_arg { |
871 | const char *doca_name; | |
872 | cred_t *doca_cred; | |
873 | void (*doca_userfunc)(objset_t *os, void *arg, | |
874 | cred_t *cr, dmu_tx_t *tx); | |
875 | void *doca_userarg; | |
876 | dmu_objset_type_t doca_type; | |
877 | uint64_t doca_flags; | |
878 | } dmu_objset_create_arg_t; | |
34dc7c2f BB |
879 | |
880 | /*ARGSUSED*/ | |
881 | static int | |
13fe0198 | 882 | dmu_objset_create_check(void *arg, dmu_tx_t *tx) |
34dc7c2f | 883 | { |
13fe0198 MA |
884 | dmu_objset_create_arg_t *doca = arg; |
885 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
886 | dsl_dir_t *pdd; | |
887 | const char *tail; | |
888 | int error; | |
34dc7c2f | 889 | |
13fe0198 | 890 | if (strchr(doca->doca_name, '@') != NULL) |
2e528b49 | 891 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 892 | |
eca7b760 IK |
893 | if (strlen(doca->doca_name) >= ZFS_MAX_DATASET_NAME_LEN) |
894 | return (SET_ERROR(ENAMETOOLONG)); | |
895 | ||
13fe0198 MA |
896 | error = dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail); |
897 | if (error != 0) | |
898 | return (error); | |
899 | if (tail == NULL) { | |
900 | dsl_dir_rele(pdd, FTAG); | |
2e528b49 | 901 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 902 | } |
788eb90c JJ |
903 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
904 | doca->doca_cred); | |
13fe0198 | 905 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f | 906 | |
788eb90c | 907 | return (error); |
34dc7c2f BB |
908 | } |
909 | ||
910 | static void | |
13fe0198 | 911 | dmu_objset_create_sync(void *arg, dmu_tx_t *tx) |
34dc7c2f | 912 | { |
13fe0198 MA |
913 | dmu_objset_create_arg_t *doca = arg; |
914 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
915 | dsl_dir_t *pdd; | |
916 | const char *tail; | |
6f1ffb06 | 917 | dsl_dataset_t *ds; |
13fe0198 | 918 | uint64_t obj; |
6f1ffb06 | 919 | blkptr_t *bp; |
13fe0198 | 920 | objset_t *os; |
34dc7c2f | 921 | |
13fe0198 | 922 | VERIFY0(dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail)); |
34dc7c2f | 923 | |
13fe0198 MA |
924 | obj = dsl_dataset_create_sync(pdd, tail, NULL, doca->doca_flags, |
925 | doca->doca_cred, tx); | |
34dc7c2f | 926 | |
13fe0198 | 927 | VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); |
6f1ffb06 | 928 | bp = dsl_dataset_get_blkptr(ds); |
13fe0198 MA |
929 | os = dmu_objset_create_impl(pdd->dd_pool->dp_spa, |
930 | ds, bp, doca->doca_type, tx); | |
34dc7c2f | 931 | |
13fe0198 MA |
932 | if (doca->doca_userfunc != NULL) { |
933 | doca->doca_userfunc(os, doca->doca_userarg, | |
934 | doca->doca_cred, tx); | |
34dc7c2f BB |
935 | } |
936 | ||
13fe0198 | 937 | spa_history_log_internal_ds(ds, "create", tx, ""); |
a0bd735a BP |
938 | zvol_create_minors(dp->dp_spa, doca->doca_name, B_TRUE); |
939 | ||
6f1ffb06 | 940 | dsl_dataset_rele(ds, FTAG); |
13fe0198 | 941 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f BB |
942 | } |
943 | ||
944 | int | |
428870ff | 945 | dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, |
34dc7c2f BB |
946 | void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg) |
947 | { | |
13fe0198 | 948 | dmu_objset_create_arg_t doca; |
34dc7c2f | 949 | |
13fe0198 MA |
950 | doca.doca_name = name; |
951 | doca.doca_cred = CRED(); | |
952 | doca.doca_flags = flags; | |
953 | doca.doca_userfunc = func; | |
954 | doca.doca_userarg = arg; | |
955 | doca.doca_type = type; | |
34dc7c2f | 956 | |
13fe0198 | 957 | return (dsl_sync_task(name, |
3d45fdd6 MA |
958 | dmu_objset_create_check, dmu_objset_create_sync, &doca, |
959 | 5, ZFS_SPACE_CHECK_NORMAL)); | |
34dc7c2f BB |
960 | } |
961 | ||
13fe0198 MA |
962 | typedef struct dmu_objset_clone_arg { |
963 | const char *doca_clone; | |
964 | const char *doca_origin; | |
965 | cred_t *doca_cred; | |
966 | } dmu_objset_clone_arg_t; | |
967 | ||
968 | /*ARGSUSED*/ | |
969 | static int | |
970 | dmu_objset_clone_check(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 971 | { |
13fe0198 | 972 | dmu_objset_clone_arg_t *doca = arg; |
428870ff BB |
973 | dsl_dir_t *pdd; |
974 | const char *tail; | |
13fe0198 MA |
975 | int error; |
976 | dsl_dataset_t *origin; | |
977 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
34dc7c2f | 978 | |
13fe0198 | 979 | if (strchr(doca->doca_clone, '@') != NULL) |
2e528b49 | 980 | return (SET_ERROR(EINVAL)); |
13fe0198 | 981 | |
eca7b760 IK |
982 | if (strlen(doca->doca_clone) >= ZFS_MAX_DATASET_NAME_LEN) |
983 | return (SET_ERROR(ENAMETOOLONG)); | |
984 | ||
13fe0198 MA |
985 | error = dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail); |
986 | if (error != 0) | |
987 | return (error); | |
428870ff | 988 | if (tail == NULL) { |
13fe0198 | 989 | dsl_dir_rele(pdd, FTAG); |
2e528b49 | 990 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 991 | } |
1cddb8c9 | 992 | |
788eb90c JJ |
993 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
994 | doca->doca_cred); | |
995 | if (error != 0) { | |
996 | dsl_dir_rele(pdd, FTAG); | |
997 | return (SET_ERROR(EDQUOT)); | |
998 | } | |
13fe0198 | 999 | dsl_dir_rele(pdd, FTAG); |
428870ff | 1000 | |
13fe0198 MA |
1001 | error = dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin); |
1002 | if (error != 0) | |
572e2857 BB |
1003 | return (error); |
1004 | ||
13fe0198 | 1005 | /* You can only clone snapshots, not the head datasets. */ |
0c66c32d | 1006 | if (!origin->ds_is_snapshot) { |
13fe0198 | 1007 | dsl_dataset_rele(origin, FTAG); |
2e528b49 | 1008 | return (SET_ERROR(EINVAL)); |
428870ff | 1009 | } |
13fe0198 | 1010 | dsl_dataset_rele(origin, FTAG); |
572e2857 | 1011 | |
13fe0198 | 1012 | return (0); |
9babb374 | 1013 | } |
34dc7c2f | 1014 | |
13fe0198 MA |
1015 | static void |
1016 | dmu_objset_clone_sync(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 1017 | { |
13fe0198 MA |
1018 | dmu_objset_clone_arg_t *doca = arg; |
1019 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
1020 | dsl_dir_t *pdd; | |
1021 | const char *tail; | |
1022 | dsl_dataset_t *origin, *ds; | |
1023 | uint64_t obj; | |
eca7b760 | 1024 | char namebuf[ZFS_MAX_DATASET_NAME_LEN]; |
6f1ffb06 | 1025 | |
13fe0198 MA |
1026 | VERIFY0(dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail)); |
1027 | VERIFY0(dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin)); | |
6f1ffb06 | 1028 | |
13fe0198 MA |
1029 | obj = dsl_dataset_create_sync(pdd, tail, origin, 0, |
1030 | doca->doca_cred, tx); | |
34dc7c2f | 1031 | |
13fe0198 MA |
1032 | VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); |
1033 | dsl_dataset_name(origin, namebuf); | |
1034 | spa_history_log_internal_ds(ds, "clone", tx, | |
1035 | "origin=%s (%llu)", namebuf, origin->ds_object); | |
a0bd735a | 1036 | zvol_create_minors(dp->dp_spa, doca->doca_clone, B_TRUE); |
13fe0198 MA |
1037 | dsl_dataset_rele(ds, FTAG); |
1038 | dsl_dataset_rele(origin, FTAG); | |
1039 | dsl_dir_rele(pdd, FTAG); | |
34dc7c2f BB |
1040 | } |
1041 | ||
1042 | int | |
13fe0198 | 1043 | dmu_objset_clone(const char *clone, const char *origin) |
34dc7c2f | 1044 | { |
13fe0198 | 1045 | dmu_objset_clone_arg_t doca; |
34dc7c2f | 1046 | |
13fe0198 MA |
1047 | doca.doca_clone = clone; |
1048 | doca.doca_origin = origin; | |
1049 | doca.doca_cred = CRED(); | |
572e2857 | 1050 | |
13fe0198 | 1051 | return (dsl_sync_task(clone, |
3d45fdd6 MA |
1052 | dmu_objset_clone_check, dmu_objset_clone_sync, &doca, |
1053 | 5, ZFS_SPACE_CHECK_NORMAL)); | |
6f1ffb06 MA |
1054 | } |
1055 | ||
1056 | int | |
1057 | dmu_objset_snapshot_one(const char *fsname, const char *snapname) | |
1058 | { | |
1059 | int err; | |
1060 | char *longsnap = kmem_asprintf("%s@%s", fsname, snapname); | |
1061 | nvlist_t *snaps = fnvlist_alloc(); | |
1062 | ||
1063 | fnvlist_add_boolean(snaps, longsnap); | |
6f1ffb06 | 1064 | strfree(longsnap); |
13fe0198 MA |
1065 | err = dsl_dataset_snapshot(snaps, NULL, NULL); |
1066 | fnvlist_free(snaps); | |
6f1ffb06 MA |
1067 | return (err); |
1068 | } | |
1069 | ||
34dc7c2f | 1070 | static void |
9babb374 | 1071 | dmu_objset_sync_dnodes(list_t *list, list_t *newlist, dmu_tx_t *tx) |
34dc7c2f BB |
1072 | { |
1073 | dnode_t *dn; | |
1074 | ||
c65aa5b2 | 1075 | while ((dn = list_head(list))) { |
34dc7c2f BB |
1076 | ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT); |
1077 | ASSERT(dn->dn_dbuf->db_data_pending); | |
1078 | /* | |
9babb374 BB |
1079 | * Initialize dn_zio outside dnode_sync() because the |
1080 | * meta-dnode needs to set it ouside dnode_sync(). | |
34dc7c2f BB |
1081 | */ |
1082 | dn->dn_zio = dn->dn_dbuf->db_data_pending->dr_zio; | |
1083 | ASSERT(dn->dn_zio); | |
1084 | ||
1085 | ASSERT3U(dn->dn_nlevels, <=, DN_MAX_LEVELS); | |
1086 | list_remove(list, dn); | |
9babb374 BB |
1087 | |
1088 | if (newlist) { | |
1089 | (void) dnode_add_ref(dn, newlist); | |
1090 | list_insert_tail(newlist, dn); | |
1091 | } | |
1092 | ||
34dc7c2f BB |
1093 | dnode_sync(dn, tx); |
1094 | } | |
1095 | } | |
1096 | ||
1097 | /* ARGSUSED */ | |
1098 | static void | |
428870ff | 1099 | dmu_objset_write_ready(zio_t *zio, arc_buf_t *abuf, void *arg) |
34dc7c2f | 1100 | { |
d6320ddb BB |
1101 | int i; |
1102 | ||
b128c09f | 1103 | blkptr_t *bp = zio->io_bp; |
428870ff | 1104 | objset_t *os = arg; |
34dc7c2f | 1105 | dnode_phys_t *dnp = &os->os_phys->os_meta_dnode; |
34dc7c2f | 1106 | |
9b67f605 | 1107 | ASSERT(!BP_IS_EMBEDDED(bp)); |
13fe0198 MA |
1108 | ASSERT3P(bp, ==, os->os_rootbp); |
1109 | ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_OBJSET); | |
1110 | ASSERT0(BP_GET_LEVEL(bp)); | |
34dc7c2f BB |
1111 | |
1112 | /* | |
9babb374 BB |
1113 | * Update rootbp fill count: it should be the number of objects |
1114 | * allocated in the object set (not counting the "special" | |
1115 | * objects that are stored in the objset_phys_t -- the meta | |
1116 | * dnode and user/group accounting objects). | |
34dc7c2f | 1117 | */ |
9babb374 | 1118 | bp->blk_fill = 0; |
d6320ddb | 1119 | for (i = 0; i < dnp->dn_nblkptr; i++) |
9b67f605 | 1120 | bp->blk_fill += BP_GET_FILL(&dnp->dn_blkptr[i]); |
428870ff BB |
1121 | } |
1122 | ||
1123 | /* ARGSUSED */ | |
1124 | static void | |
1125 | dmu_objset_write_done(zio_t *zio, arc_buf_t *abuf, void *arg) | |
1126 | { | |
1127 | blkptr_t *bp = zio->io_bp; | |
1128 | blkptr_t *bp_orig = &zio->io_bp_orig; | |
1129 | objset_t *os = arg; | |
34dc7c2f | 1130 | |
b128c09f | 1131 | if (zio->io_flags & ZIO_FLAG_IO_REWRITE) { |
428870ff | 1132 | ASSERT(BP_EQUAL(bp, bp_orig)); |
b128c09f | 1133 | } else { |
428870ff BB |
1134 | dsl_dataset_t *ds = os->os_dsl_dataset; |
1135 | dmu_tx_t *tx = os->os_synctx; | |
1136 | ||
1137 | (void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE); | |
1138 | dsl_dataset_block_born(ds, bp, tx); | |
34dc7c2f BB |
1139 | } |
1140 | } | |
1141 | ||
34dc7c2f BB |
1142 | /* called from dsl */ |
1143 | void | |
428870ff | 1144 | dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx) |
34dc7c2f BB |
1145 | { |
1146 | int txgoff; | |
5dbd68a3 | 1147 | zbookmark_phys_t zb; |
428870ff | 1148 | zio_prop_t zp; |
34dc7c2f BB |
1149 | zio_t *zio; |
1150 | list_t *list; | |
9babb374 | 1151 | list_t *newlist = NULL; |
34dc7c2f BB |
1152 | dbuf_dirty_record_t *dr; |
1153 | ||
1154 | dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg); | |
1155 | ||
1156 | ASSERT(dmu_tx_is_syncing(tx)); | |
1157 | /* XXX the write_done callback should really give us the tx... */ | |
1158 | os->os_synctx = tx; | |
1159 | ||
1160 | if (os->os_dsl_dataset == NULL) { | |
1161 | /* | |
1162 | * This is the MOS. If we have upgraded, | |
1163 | * spa_max_replication() could change, so reset | |
1164 | * os_copies here. | |
1165 | */ | |
1166 | os->os_copies = spa_max_replication(os->os_spa); | |
1167 | } | |
1168 | ||
1169 | /* | |
1170 | * Create the root block IO | |
1171 | */ | |
428870ff BB |
1172 | SET_BOOKMARK(&zb, os->os_dsl_dataset ? |
1173 | os->os_dsl_dataset->ds_object : DMU_META_OBJSET, | |
1174 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
294f6806 | 1175 | arc_release(os->os_phys_buf, &os->os_phys_buf); |
b128c09f | 1176 | |
2aa34383 | 1177 | dmu_write_policy(os, NULL, 0, 0, ZIO_COMPRESS_INHERIT, &zp); |
9babb374 | 1178 | |
428870ff | 1179 | zio = arc_write(pio, os->os_spa, tx->tx_txg, |
3a17a7a9 | 1180 | os->os_rootbp, os->os_phys_buf, DMU_OS_IS_L2CACHEABLE(os), |
bc77ba73 PD |
1181 | &zp, dmu_objset_write_ready, NULL, NULL, dmu_objset_write_done, |
1182 | os, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb); | |
34dc7c2f BB |
1183 | |
1184 | /* | |
9babb374 | 1185 | * Sync special dnodes - the parent IO for the sync is the root block |
34dc7c2f | 1186 | */ |
572e2857 BB |
1187 | DMU_META_DNODE(os)->dn_zio = zio; |
1188 | dnode_sync(DMU_META_DNODE(os), tx); | |
34dc7c2f | 1189 | |
9babb374 BB |
1190 | os->os_phys->os_flags = os->os_flags; |
1191 | ||
572e2857 BB |
1192 | if (DMU_USERUSED_DNODE(os) && |
1193 | DMU_USERUSED_DNODE(os)->dn_type != DMU_OT_NONE) { | |
1194 | DMU_USERUSED_DNODE(os)->dn_zio = zio; | |
1195 | dnode_sync(DMU_USERUSED_DNODE(os), tx); | |
1196 | DMU_GROUPUSED_DNODE(os)->dn_zio = zio; | |
1197 | dnode_sync(DMU_GROUPUSED_DNODE(os), tx); | |
9babb374 BB |
1198 | } |
1199 | ||
34dc7c2f BB |
1200 | txgoff = tx->tx_txg & TXG_MASK; |
1201 | ||
9babb374 BB |
1202 | if (dmu_objset_userused_enabled(os)) { |
1203 | newlist = &os->os_synced_dnodes; | |
1204 | /* | |
1205 | * We must create the list here because it uses the | |
1206 | * dn_dirty_link[] of this txg. | |
1207 | */ | |
1208 | list_create(newlist, sizeof (dnode_t), | |
1209 | offsetof(dnode_t, dn_dirty_link[txgoff])); | |
1210 | } | |
1211 | ||
1212 | dmu_objset_sync_dnodes(&os->os_free_dnodes[txgoff], newlist, tx); | |
1213 | dmu_objset_sync_dnodes(&os->os_dirty_dnodes[txgoff], newlist, tx); | |
34dc7c2f | 1214 | |
572e2857 | 1215 | list = &DMU_META_DNODE(os)->dn_dirty_records[txgoff]; |
6f1ffb06 | 1216 | while ((dr = list_head(list))) { |
13fe0198 | 1217 | ASSERT0(dr->dr_dbuf->db_level); |
34dc7c2f BB |
1218 | list_remove(list, dr); |
1219 | if (dr->dr_zio) | |
1220 | zio_nowait(dr->dr_zio); | |
1221 | } | |
68cbd56e NB |
1222 | |
1223 | /* Enable dnode backfill if enough objects have been freed. */ | |
1224 | if (os->os_freed_dnodes >= dmu_rescan_dnode_threshold) { | |
1225 | os->os_rescan_dnodes = B_TRUE; | |
1226 | os->os_freed_dnodes = 0; | |
1227 | } | |
1228 | ||
34dc7c2f BB |
1229 | /* |
1230 | * Free intent log blocks up to this tx. | |
1231 | */ | |
1232 | zil_sync(os->os_zil, tx); | |
b128c09f | 1233 | os->os_phys->os_zil_header = os->os_zil_header; |
34dc7c2f BB |
1234 | zio_nowait(zio); |
1235 | } | |
1236 | ||
428870ff BB |
1237 | boolean_t |
1238 | dmu_objset_is_dirty(objset_t *os, uint64_t txg) | |
1239 | { | |
1240 | return (!list_is_empty(&os->os_dirty_dnodes[txg & TXG_MASK]) || | |
1241 | !list_is_empty(&os->os_free_dnodes[txg & TXG_MASK])); | |
1242 | } | |
1243 | ||
572e2857 | 1244 | static objset_used_cb_t *used_cbs[DMU_OST_NUMTYPES]; |
9babb374 BB |
1245 | |
1246 | void | |
1247 | dmu_objset_register_type(dmu_objset_type_t ost, objset_used_cb_t *cb) | |
1248 | { | |
1249 | used_cbs[ost] = cb; | |
1250 | } | |
1251 | ||
1252 | boolean_t | |
428870ff | 1253 | dmu_objset_userused_enabled(objset_t *os) |
9babb374 BB |
1254 | { |
1255 | return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE && | |
572e2857 BB |
1256 | used_cbs[os->os_phys->os_type] != NULL && |
1257 | DMU_USERUSED_DNODE(os) != NULL); | |
9babb374 BB |
1258 | } |
1259 | ||
428870ff BB |
1260 | static void |
1261 | do_userquota_update(objset_t *os, uint64_t used, uint64_t flags, | |
1262 | uint64_t user, uint64_t group, boolean_t subtract, dmu_tx_t *tx) | |
1263 | { | |
1264 | if ((flags & DNODE_FLAG_USERUSED_ACCOUNTED)) { | |
50c957f7 | 1265 | int64_t delta = DNODE_MIN_SIZE + used; |
428870ff BB |
1266 | if (subtract) |
1267 | delta = -delta; | |
1268 | VERIFY3U(0, ==, zap_increment_int(os, DMU_USERUSED_OBJECT, | |
1269 | user, delta, tx)); | |
1270 | VERIFY3U(0, ==, zap_increment_int(os, DMU_GROUPUSED_OBJECT, | |
1271 | group, delta, tx)); | |
1272 | } | |
1273 | } | |
1274 | ||
9babb374 | 1275 | void |
428870ff | 1276 | dmu_objset_do_userquota_updates(objset_t *os, dmu_tx_t *tx) |
9babb374 BB |
1277 | { |
1278 | dnode_t *dn; | |
1279 | list_t *list = &os->os_synced_dnodes; | |
9babb374 BB |
1280 | |
1281 | ASSERT(list_head(list) == NULL || dmu_objset_userused_enabled(os)); | |
1282 | ||
6f1ffb06 | 1283 | while ((dn = list_head(list))) { |
572e2857 | 1284 | int flags; |
9babb374 | 1285 | ASSERT(!DMU_OBJECT_IS_SPECIAL(dn->dn_object)); |
9babb374 BB |
1286 | ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE || |
1287 | dn->dn_phys->dn_flags & | |
1288 | DNODE_FLAG_USERUSED_ACCOUNTED); | |
1289 | ||
1290 | /* Allocate the user/groupused objects if necessary. */ | |
572e2857 | 1291 | if (DMU_USERUSED_DNODE(os)->dn_type == DMU_OT_NONE) { |
428870ff | 1292 | VERIFY(0 == zap_create_claim(os, |
9babb374 BB |
1293 | DMU_USERUSED_OBJECT, |
1294 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
428870ff | 1295 | VERIFY(0 == zap_create_claim(os, |
9babb374 BB |
1296 | DMU_GROUPUSED_OBJECT, |
1297 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
1298 | } | |
1299 | ||
1300 | /* | |
428870ff BB |
1301 | * We intentionally modify the zap object even if the |
1302 | * net delta is zero. Otherwise | |
1303 | * the block of the zap obj could be shared between | |
1304 | * datasets but need to be different between them after | |
1305 | * a bprewrite. | |
9babb374 | 1306 | */ |
9babb374 | 1307 | |
572e2857 BB |
1308 | flags = dn->dn_id_flags; |
1309 | ASSERT(flags); | |
1310 | if (flags & DN_ID_OLD_EXIST) { | |
428870ff BB |
1311 | do_userquota_update(os, dn->dn_oldused, dn->dn_oldflags, |
1312 | dn->dn_olduid, dn->dn_oldgid, B_TRUE, tx); | |
1313 | } | |
572e2857 | 1314 | if (flags & DN_ID_NEW_EXIST) { |
428870ff BB |
1315 | do_userquota_update(os, DN_USED_BYTES(dn->dn_phys), |
1316 | dn->dn_phys->dn_flags, dn->dn_newuid, | |
1317 | dn->dn_newgid, B_FALSE, tx); | |
1318 | } | |
1319 | ||
572e2857 | 1320 | mutex_enter(&dn->dn_mtx); |
428870ff BB |
1321 | dn->dn_oldused = 0; |
1322 | dn->dn_oldflags = 0; | |
1323 | if (dn->dn_id_flags & DN_ID_NEW_EXIST) { | |
1324 | dn->dn_olduid = dn->dn_newuid; | |
1325 | dn->dn_oldgid = dn->dn_newgid; | |
1326 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
1327 | if (dn->dn_bonuslen == 0) | |
1328 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1329 | else | |
1330 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1331 | } | |
1332 | dn->dn_id_flags &= ~(DN_ID_NEW_EXIST); | |
9babb374 BB |
1333 | mutex_exit(&dn->dn_mtx); |
1334 | ||
1335 | list_remove(list, dn); | |
1336 | dnode_rele(dn, list); | |
1337 | } | |
1338 | } | |
1339 | ||
428870ff BB |
1340 | /* |
1341 | * Returns a pointer to data to find uid/gid from | |
1342 | * | |
1343 | * If a dirty record for transaction group that is syncing can't | |
1344 | * be found then NULL is returned. In the NULL case it is assumed | |
1345 | * the uid/gid aren't changing. | |
1346 | */ | |
1347 | static void * | |
1348 | dmu_objset_userquota_find_data(dmu_buf_impl_t *db, dmu_tx_t *tx) | |
1349 | { | |
1350 | dbuf_dirty_record_t *dr, **drp; | |
1351 | void *data; | |
1352 | ||
1353 | if (db->db_dirtycnt == 0) | |
1354 | return (db->db.db_data); /* Nothing is changing */ | |
1355 | ||
1356 | for (drp = &db->db_last_dirty; (dr = *drp) != NULL; drp = &dr->dr_next) | |
1357 | if (dr->dr_txg == tx->tx_txg) | |
1358 | break; | |
1359 | ||
572e2857 | 1360 | if (dr == NULL) { |
428870ff | 1361 | data = NULL; |
572e2857 BB |
1362 | } else { |
1363 | dnode_t *dn; | |
1364 | ||
1365 | DB_DNODE_ENTER(dr->dr_dbuf); | |
1366 | dn = DB_DNODE(dr->dr_dbuf); | |
1367 | ||
1368 | if (dn->dn_bonuslen == 0 && | |
1369 | dr->dr_dbuf->db_blkid == DMU_SPILL_BLKID) | |
1370 | data = dr->dt.dl.dr_data->b_data; | |
1371 | else | |
1372 | data = dr->dt.dl.dr_data; | |
1373 | ||
1374 | DB_DNODE_EXIT(dr->dr_dbuf); | |
1375 | } | |
1376 | ||
428870ff BB |
1377 | return (data); |
1378 | } | |
1379 | ||
1380 | void | |
1381 | dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) | |
1382 | { | |
1383 | objset_t *os = dn->dn_objset; | |
1384 | void *data = NULL; | |
1385 | dmu_buf_impl_t *db = NULL; | |
a117a6d6 GW |
1386 | uint64_t *user = NULL; |
1387 | uint64_t *group = NULL; | |
428870ff BB |
1388 | int flags = dn->dn_id_flags; |
1389 | int error; | |
1390 | boolean_t have_spill = B_FALSE; | |
1391 | ||
1392 | if (!dmu_objset_userused_enabled(dn->dn_objset)) | |
1393 | return; | |
1394 | ||
1395 | if (before && (flags & (DN_ID_CHKED_BONUS|DN_ID_OLD_EXIST| | |
1396 | DN_ID_CHKED_SPILL))) | |
1397 | return; | |
1398 | ||
1399 | if (before && dn->dn_bonuslen != 0) | |
1400 | data = DN_BONUS(dn->dn_phys); | |
1401 | else if (!before && dn->dn_bonuslen != 0) { | |
a3000f93 BB |
1402 | if (dn->dn_bonus) { |
1403 | db = dn->dn_bonus; | |
428870ff BB |
1404 | mutex_enter(&db->db_mtx); |
1405 | data = dmu_objset_userquota_find_data(db, tx); | |
1406 | } else { | |
1407 | data = DN_BONUS(dn->dn_phys); | |
1408 | } | |
1409 | } else if (dn->dn_bonuslen == 0 && dn->dn_bonustype == DMU_OT_SA) { | |
1410 | int rf = 0; | |
1411 | ||
1412 | if (RW_WRITE_HELD(&dn->dn_struct_rwlock)) | |
1413 | rf |= DB_RF_HAVESTRUCT; | |
572e2857 BB |
1414 | error = dmu_spill_hold_by_dnode(dn, |
1415 | rf | DB_RF_MUST_SUCCEED, | |
428870ff BB |
1416 | FTAG, (dmu_buf_t **)&db); |
1417 | ASSERT(error == 0); | |
1418 | mutex_enter(&db->db_mtx); | |
1419 | data = (before) ? db->db.db_data : | |
1420 | dmu_objset_userquota_find_data(db, tx); | |
1421 | have_spill = B_TRUE; | |
1422 | } else { | |
1423 | mutex_enter(&dn->dn_mtx); | |
1424 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1425 | mutex_exit(&dn->dn_mtx); | |
1426 | return; | |
1427 | } | |
1428 | ||
1429 | if (before) { | |
1430 | ASSERT(data); | |
1431 | user = &dn->dn_olduid; | |
1432 | group = &dn->dn_oldgid; | |
1433 | } else if (data) { | |
1434 | user = &dn->dn_newuid; | |
1435 | group = &dn->dn_newgid; | |
1436 | } | |
1437 | ||
1438 | /* | |
1439 | * Must always call the callback in case the object | |
1440 | * type has changed and that type isn't an object type to track | |
1441 | */ | |
1442 | error = used_cbs[os->os_phys->os_type](dn->dn_bonustype, data, | |
1443 | user, group); | |
1444 | ||
1445 | /* | |
1446 | * Preserve existing uid/gid when the callback can't determine | |
1447 | * what the new uid/gid are and the callback returned EEXIST. | |
1448 | * The EEXIST error tells us to just use the existing uid/gid. | |
1449 | * If we don't know what the old values are then just assign | |
1450 | * them to 0, since that is a new file being created. | |
1451 | */ | |
1452 | if (!before && data == NULL && error == EEXIST) { | |
1453 | if (flags & DN_ID_OLD_EXIST) { | |
1454 | dn->dn_newuid = dn->dn_olduid; | |
1455 | dn->dn_newgid = dn->dn_oldgid; | |
1456 | } else { | |
1457 | dn->dn_newuid = 0; | |
1458 | dn->dn_newgid = 0; | |
1459 | } | |
1460 | error = 0; | |
1461 | } | |
1462 | ||
1463 | if (db) | |
1464 | mutex_exit(&db->db_mtx); | |
1465 | ||
1466 | mutex_enter(&dn->dn_mtx); | |
1467 | if (error == 0 && before) | |
1468 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
1469 | if (error == 0 && !before) | |
1470 | dn->dn_id_flags |= DN_ID_NEW_EXIST; | |
1471 | ||
1472 | if (have_spill) { | |
1473 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1474 | } else { | |
1475 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1476 | } | |
1477 | mutex_exit(&dn->dn_mtx); | |
a3000f93 | 1478 | if (have_spill) |
428870ff BB |
1479 | dmu_buf_rele((dmu_buf_t *)db, FTAG); |
1480 | } | |
1481 | ||
9babb374 BB |
1482 | boolean_t |
1483 | dmu_objset_userspace_present(objset_t *os) | |
1484 | { | |
428870ff | 1485 | return (os->os_phys->os_flags & |
9babb374 BB |
1486 | OBJSET_FLAG_USERACCOUNTING_COMPLETE); |
1487 | } | |
1488 | ||
1489 | int | |
1490 | dmu_objset_userspace_upgrade(objset_t *os) | |
1491 | { | |
1492 | uint64_t obj; | |
1493 | int err = 0; | |
1494 | ||
1495 | if (dmu_objset_userspace_present(os)) | |
1496 | return (0); | |
428870ff | 1497 | if (!dmu_objset_userused_enabled(os)) |
2e528b49 | 1498 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 1499 | if (dmu_objset_is_snapshot(os)) |
2e528b49 | 1500 | return (SET_ERROR(EINVAL)); |
9babb374 BB |
1501 | |
1502 | /* | |
1503 | * We simply need to mark every object dirty, so that it will be | |
1504 | * synced out and now accounted. If this is called | |
1505 | * concurrently, or if we already did some work before crashing, | |
1506 | * that's fine, since we track each object's accounted state | |
1507 | * independently. | |
1508 | */ | |
1509 | ||
1510 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
45d1cae3 | 1511 | dmu_tx_t *tx; |
9babb374 BB |
1512 | dmu_buf_t *db; |
1513 | int objerr; | |
1514 | ||
1515 | if (issig(JUSTLOOKING) && issig(FORREAL)) | |
2e528b49 | 1516 | return (SET_ERROR(EINTR)); |
9babb374 BB |
1517 | |
1518 | objerr = dmu_bonus_hold(os, obj, FTAG, &db); | |
13fe0198 | 1519 | if (objerr != 0) |
9babb374 | 1520 | continue; |
45d1cae3 | 1521 | tx = dmu_tx_create(os); |
9babb374 BB |
1522 | dmu_tx_hold_bonus(tx, obj); |
1523 | objerr = dmu_tx_assign(tx, TXG_WAIT); | |
13fe0198 | 1524 | if (objerr != 0) { |
9babb374 BB |
1525 | dmu_tx_abort(tx); |
1526 | continue; | |
1527 | } | |
1528 | dmu_buf_will_dirty(db, tx); | |
1529 | dmu_buf_rele(db, FTAG); | |
1530 | dmu_tx_commit(tx); | |
1531 | } | |
1532 | ||
428870ff | 1533 | os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; |
9babb374 BB |
1534 | txg_wait_synced(dmu_objset_pool(os), 0); |
1535 | return (0); | |
1536 | } | |
1537 | ||
34dc7c2f BB |
1538 | void |
1539 | dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, | |
1540 | uint64_t *usedobjsp, uint64_t *availobjsp) | |
1541 | { | |
428870ff | 1542 | dsl_dataset_space(os->os_dsl_dataset, refdbytesp, availbytesp, |
34dc7c2f BB |
1543 | usedobjsp, availobjsp); |
1544 | } | |
1545 | ||
1546 | uint64_t | |
1547 | dmu_objset_fsid_guid(objset_t *os) | |
1548 | { | |
428870ff | 1549 | return (dsl_dataset_fsid_guid(os->os_dsl_dataset)); |
34dc7c2f BB |
1550 | } |
1551 | ||
1552 | void | |
1553 | dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat) | |
1554 | { | |
428870ff BB |
1555 | stat->dds_type = os->os_phys->os_type; |
1556 | if (os->os_dsl_dataset) | |
1557 | dsl_dataset_fast_stat(os->os_dsl_dataset, stat); | |
34dc7c2f BB |
1558 | } |
1559 | ||
1560 | void | |
1561 | dmu_objset_stats(objset_t *os, nvlist_t *nv) | |
1562 | { | |
428870ff BB |
1563 | ASSERT(os->os_dsl_dataset || |
1564 | os->os_phys->os_type == DMU_OST_META); | |
34dc7c2f | 1565 | |
428870ff BB |
1566 | if (os->os_dsl_dataset != NULL) |
1567 | dsl_dataset_stats(os->os_dsl_dataset, nv); | |
34dc7c2f BB |
1568 | |
1569 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_TYPE, | |
428870ff | 1570 | os->os_phys->os_type); |
9babb374 BB |
1571 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERACCOUNTING, |
1572 | dmu_objset_userspace_present(os)); | |
34dc7c2f BB |
1573 | } |
1574 | ||
1575 | int | |
1576 | dmu_objset_is_snapshot(objset_t *os) | |
1577 | { | |
428870ff | 1578 | if (os->os_dsl_dataset != NULL) |
0c66c32d | 1579 | return (os->os_dsl_dataset->ds_is_snapshot); |
34dc7c2f BB |
1580 | else |
1581 | return (B_FALSE); | |
1582 | } | |
1583 | ||
1584 | int | |
1585 | dmu_snapshot_realname(objset_t *os, char *name, char *real, int maxlen, | |
1586 | boolean_t *conflict) | |
1587 | { | |
428870ff | 1588 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
1589 | uint64_t ignored; |
1590 | ||
d683ddbb | 1591 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 1592 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1593 | |
1594 | return (zap_lookup_norm(ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb JG |
1595 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, name, 8, 1, &ignored, |
1596 | MT_FIRST, real, maxlen, conflict)); | |
34dc7c2f BB |
1597 | } |
1598 | ||
1599 | int | |
1600 | dmu_snapshot_list_next(objset_t *os, int namelen, char *name, | |
1601 | uint64_t *idp, uint64_t *offp, boolean_t *case_conflict) | |
1602 | { | |
428870ff | 1603 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
1604 | zap_cursor_t cursor; |
1605 | zap_attribute_t attr; | |
1606 | ||
13fe0198 MA |
1607 | ASSERT(dsl_pool_config_held(dmu_objset_pool(os))); |
1608 | ||
d683ddbb | 1609 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 1610 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1611 | |
1612 | zap_cursor_init_serialized(&cursor, | |
1613 | ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb | 1614 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, *offp); |
34dc7c2f BB |
1615 | |
1616 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
1617 | zap_cursor_fini(&cursor); | |
2e528b49 | 1618 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1619 | } |
1620 | ||
1621 | if (strlen(attr.za_name) + 1 > namelen) { | |
1622 | zap_cursor_fini(&cursor); | |
2e528b49 | 1623 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
1624 | } |
1625 | ||
1626 | (void) strcpy(name, attr.za_name); | |
1627 | if (idp) | |
1628 | *idp = attr.za_first_integer; | |
1629 | if (case_conflict) | |
1630 | *case_conflict = attr.za_normalization_conflict; | |
1631 | zap_cursor_advance(&cursor); | |
1632 | *offp = zap_cursor_serialize(&cursor); | |
1633 | zap_cursor_fini(&cursor); | |
1634 | ||
1635 | return (0); | |
1636 | } | |
1637 | ||
ebe7e575 | 1638 | int |
6772fb67 | 1639 | dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *value) |
ebe7e575 | 1640 | { |
d1d7e268 | 1641 | return (dsl_dataset_snap_lookup(os->os_dsl_dataset, name, value)); |
ebe7e575 BB |
1642 | } |
1643 | ||
34dc7c2f BB |
1644 | int |
1645 | dmu_dir_list_next(objset_t *os, int namelen, char *name, | |
1646 | uint64_t *idp, uint64_t *offp) | |
1647 | { | |
428870ff | 1648 | dsl_dir_t *dd = os->os_dsl_dataset->ds_dir; |
34dc7c2f BB |
1649 | zap_cursor_t cursor; |
1650 | zap_attribute_t attr; | |
1651 | ||
1652 | /* there is no next dir on a snapshot! */ | |
428870ff | 1653 | if (os->os_dsl_dataset->ds_object != |
d683ddbb | 1654 | dsl_dir_phys(dd)->dd_head_dataset_obj) |
2e528b49 | 1655 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1656 | |
1657 | zap_cursor_init_serialized(&cursor, | |
1658 | dd->dd_pool->dp_meta_objset, | |
d683ddbb | 1659 | dsl_dir_phys(dd)->dd_child_dir_zapobj, *offp); |
34dc7c2f BB |
1660 | |
1661 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
1662 | zap_cursor_fini(&cursor); | |
2e528b49 | 1663 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1664 | } |
1665 | ||
1666 | if (strlen(attr.za_name) + 1 > namelen) { | |
1667 | zap_cursor_fini(&cursor); | |
2e528b49 | 1668 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
1669 | } |
1670 | ||
1671 | (void) strcpy(name, attr.za_name); | |
1672 | if (idp) | |
1673 | *idp = attr.za_first_integer; | |
1674 | zap_cursor_advance(&cursor); | |
1675 | *offp = zap_cursor_serialize(&cursor); | |
1676 | zap_cursor_fini(&cursor); | |
1677 | ||
1678 | return (0); | |
1679 | } | |
1680 | ||
9c43027b AJ |
1681 | typedef struct dmu_objset_find_ctx { |
1682 | taskq_t *dc_tq; | |
1683 | dsl_pool_t *dc_dp; | |
1684 | uint64_t dc_ddobj; | |
1685 | int (*dc_func)(dsl_pool_t *, dsl_dataset_t *, void *); | |
1686 | void *dc_arg; | |
1687 | int dc_flags; | |
1688 | kmutex_t *dc_error_lock; | |
1689 | int *dc_error; | |
1690 | } dmu_objset_find_ctx_t; | |
1691 | ||
1692 | static void | |
1693 | dmu_objset_find_dp_impl(dmu_objset_find_ctx_t *dcp) | |
b128c09f | 1694 | { |
9c43027b AJ |
1695 | dsl_pool_t *dp = dcp->dc_dp; |
1696 | dmu_objset_find_ctx_t *child_dcp; | |
13fe0198 MA |
1697 | dsl_dir_t *dd; |
1698 | dsl_dataset_t *ds; | |
1699 | zap_cursor_t zc; | |
1700 | zap_attribute_t *attr; | |
1701 | uint64_t thisobj; | |
9c43027b | 1702 | int err = 0; |
13fe0198 | 1703 | |
9c43027b AJ |
1704 | /* don't process if there already was an error */ |
1705 | if (*dcp->dc_error != 0) | |
1706 | goto out; | |
13fe0198 | 1707 | |
9c43027b | 1708 | err = dsl_dir_hold_obj(dp, dcp->dc_ddobj, NULL, FTAG, &dd); |
13fe0198 | 1709 | if (err != 0) |
9c43027b | 1710 | goto out; |
13fe0198 MA |
1711 | |
1712 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ | |
1713 | if (dd->dd_myname[0] == '$') { | |
1714 | dsl_dir_rele(dd, FTAG); | |
9c43027b | 1715 | goto out; |
13fe0198 MA |
1716 | } |
1717 | ||
d683ddbb | 1718 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 1719 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
13fe0198 MA |
1720 | |
1721 | /* | |
1722 | * Iterate over all children. | |
1723 | */ | |
9c43027b | 1724 | if (dcp->dc_flags & DS_FIND_CHILDREN) { |
13fe0198 | 1725 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
d683ddbb | 1726 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
13fe0198 MA |
1727 | zap_cursor_retrieve(&zc, attr) == 0; |
1728 | (void) zap_cursor_advance(&zc)) { | |
1729 | ASSERT3U(attr->za_integer_length, ==, | |
1730 | sizeof (uint64_t)); | |
1731 | ASSERT3U(attr->za_num_integers, ==, 1); | |
1732 | ||
9c43027b AJ |
1733 | child_dcp = kmem_alloc(sizeof (*child_dcp), KM_SLEEP); |
1734 | *child_dcp = *dcp; | |
1735 | child_dcp->dc_ddobj = attr->za_first_integer; | |
1736 | if (dcp->dc_tq != NULL) | |
1737 | (void) taskq_dispatch(dcp->dc_tq, | |
1738 | dmu_objset_find_dp_cb, child_dcp, TQ_SLEEP); | |
1739 | else | |
1740 | dmu_objset_find_dp_impl(child_dcp); | |
13fe0198 MA |
1741 | } |
1742 | zap_cursor_fini(&zc); | |
13fe0198 MA |
1743 | } |
1744 | ||
1745 | /* | |
1746 | * Iterate over all snapshots. | |
1747 | */ | |
9c43027b | 1748 | if (dcp->dc_flags & DS_FIND_SNAPSHOTS) { |
13fe0198 MA |
1749 | dsl_dataset_t *ds; |
1750 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1751 | ||
1752 | if (err == 0) { | |
d683ddbb JG |
1753 | uint64_t snapobj; |
1754 | ||
1755 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
13fe0198 MA |
1756 | dsl_dataset_rele(ds, FTAG); |
1757 | ||
1758 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
1759 | zap_cursor_retrieve(&zc, attr) == 0; | |
1760 | (void) zap_cursor_advance(&zc)) { | |
1761 | ASSERT3U(attr->za_integer_length, ==, | |
1762 | sizeof (uint64_t)); | |
1763 | ASSERT3U(attr->za_num_integers, ==, 1); | |
1764 | ||
1765 | err = dsl_dataset_hold_obj(dp, | |
1766 | attr->za_first_integer, FTAG, &ds); | |
1767 | if (err != 0) | |
1768 | break; | |
9c43027b | 1769 | err = dcp->dc_func(dp, ds, dcp->dc_arg); |
13fe0198 MA |
1770 | dsl_dataset_rele(ds, FTAG); |
1771 | if (err != 0) | |
1772 | break; | |
1773 | } | |
1774 | zap_cursor_fini(&zc); | |
1775 | } | |
1776 | } | |
1777 | ||
1778 | dsl_dir_rele(dd, FTAG); | |
1779 | kmem_free(attr, sizeof (zap_attribute_t)); | |
1780 | ||
1781 | if (err != 0) | |
9c43027b | 1782 | goto out; |
13fe0198 MA |
1783 | |
1784 | /* | |
1785 | * Apply to self. | |
1786 | */ | |
1787 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1788 | if (err != 0) | |
9c43027b AJ |
1789 | goto out; |
1790 | err = dcp->dc_func(dp, ds, dcp->dc_arg); | |
13fe0198 | 1791 | dsl_dataset_rele(ds, FTAG); |
9c43027b AJ |
1792 | |
1793 | out: | |
1794 | if (err != 0) { | |
1795 | mutex_enter(dcp->dc_error_lock); | |
1796 | /* only keep first error */ | |
1797 | if (*dcp->dc_error == 0) | |
1798 | *dcp->dc_error = err; | |
1799 | mutex_exit(dcp->dc_error_lock); | |
1800 | } | |
1801 | ||
1802 | kmem_free(dcp, sizeof (*dcp)); | |
1803 | } | |
1804 | ||
1805 | static void | |
1806 | dmu_objset_find_dp_cb(void *arg) | |
1807 | { | |
1808 | dmu_objset_find_ctx_t *dcp = arg; | |
1809 | dsl_pool_t *dp = dcp->dc_dp; | |
1810 | ||
5e8cd5d1 AJ |
1811 | /* |
1812 | * We need to get a pool_config_lock here, as there are several | |
1813 | * asssert(pool_config_held) down the stack. Getting a lock via | |
1814 | * dsl_pool_config_enter is risky, as it might be stalled by a | |
1815 | * pending writer. This would deadlock, as the write lock can | |
1816 | * only be granted when our parent thread gives up the lock. | |
1817 | * The _prio interface gives us priority over a pending writer. | |
1818 | */ | |
1819 | dsl_pool_config_enter_prio(dp, FTAG); | |
9c43027b AJ |
1820 | |
1821 | dmu_objset_find_dp_impl(dcp); | |
1822 | ||
1823 | dsl_pool_config_exit(dp, FTAG); | |
1824 | } | |
1825 | ||
1826 | /* | |
1827 | * Find objsets under and including ddobj, call func(ds) on each. | |
1828 | * The order for the enumeration is completely undefined. | |
1829 | * func is called with dsl_pool_config held. | |
1830 | */ | |
1831 | int | |
1832 | dmu_objset_find_dp(dsl_pool_t *dp, uint64_t ddobj, | |
1833 | int func(dsl_pool_t *, dsl_dataset_t *, void *), void *arg, int flags) | |
1834 | { | |
1835 | int error = 0; | |
1836 | taskq_t *tq = NULL; | |
1837 | int ntasks; | |
1838 | dmu_objset_find_ctx_t *dcp; | |
1839 | kmutex_t err_lock; | |
1840 | ||
1841 | mutex_init(&err_lock, NULL, MUTEX_DEFAULT, NULL); | |
1842 | dcp = kmem_alloc(sizeof (*dcp), KM_SLEEP); | |
1843 | dcp->dc_tq = NULL; | |
1844 | dcp->dc_dp = dp; | |
1845 | dcp->dc_ddobj = ddobj; | |
1846 | dcp->dc_func = func; | |
1847 | dcp->dc_arg = arg; | |
1848 | dcp->dc_flags = flags; | |
1849 | dcp->dc_error_lock = &err_lock; | |
1850 | dcp->dc_error = &error; | |
1851 | ||
1852 | if ((flags & DS_FIND_SERIALIZE) || dsl_pool_config_held_writer(dp)) { | |
1853 | /* | |
1854 | * In case a write lock is held we can't make use of | |
1855 | * parallelism, as down the stack of the worker threads | |
1856 | * the lock is asserted via dsl_pool_config_held. | |
1857 | * In case of a read lock this is solved by getting a read | |
1858 | * lock in each worker thread, which isn't possible in case | |
1859 | * of a writer lock. So we fall back to the synchronous path | |
1860 | * here. | |
1861 | * In the future it might be possible to get some magic into | |
1862 | * dsl_pool_config_held in a way that it returns true for | |
1863 | * the worker threads so that a single lock held from this | |
1864 | * thread suffices. For now, stay single threaded. | |
1865 | */ | |
1866 | dmu_objset_find_dp_impl(dcp); | |
e5676636 | 1867 | mutex_destroy(&err_lock); |
9c43027b AJ |
1868 | |
1869 | return (error); | |
1870 | } | |
1871 | ||
1872 | ntasks = dmu_find_threads; | |
1873 | if (ntasks == 0) | |
1874 | ntasks = vdev_count_leaves(dp->dp_spa) * 4; | |
1229323d | 1875 | tq = taskq_create("dmu_objset_find", ntasks, maxclsyspri, ntasks, |
9c43027b AJ |
1876 | INT_MAX, 0); |
1877 | if (tq == NULL) { | |
1878 | kmem_free(dcp, sizeof (*dcp)); | |
e5676636 BB |
1879 | mutex_destroy(&err_lock); |
1880 | ||
9c43027b AJ |
1881 | return (SET_ERROR(ENOMEM)); |
1882 | } | |
1883 | dcp->dc_tq = tq; | |
1884 | ||
1885 | /* dcp will be freed by task */ | |
1886 | (void) taskq_dispatch(tq, dmu_objset_find_dp_cb, dcp, TQ_SLEEP); | |
1887 | ||
1888 | /* | |
1889 | * PORTING: this code relies on the property of taskq_wait to wait | |
1890 | * until no more tasks are queued and no more tasks are active. As | |
1891 | * we always queue new tasks from within other tasks, task_wait | |
1892 | * reliably waits for the full recursion to finish, even though we | |
1893 | * enqueue new tasks after taskq_wait has been called. | |
1894 | * On platforms other than illumos, taskq_wait may not have this | |
1895 | * property. | |
1896 | */ | |
1897 | taskq_wait(tq); | |
1898 | taskq_destroy(tq); | |
1899 | mutex_destroy(&err_lock); | |
1900 | ||
1901 | return (error); | |
b128c09f BB |
1902 | } |
1903 | ||
1904 | /* | |
13fe0198 MA |
1905 | * Find all objsets under name, and for each, call 'func(child_name, arg)'. |
1906 | * The dp_config_rwlock must not be held when this is called, and it | |
1907 | * will not be held when the callback is called. | |
1908 | * Therefore this function should only be used when the pool is not changing | |
1909 | * (e.g. in syncing context), or the callback can deal with the possible races. | |
b128c09f | 1910 | */ |
13fe0198 MA |
1911 | static int |
1912 | dmu_objset_find_impl(spa_t *spa, const char *name, | |
1913 | int func(const char *, void *), void *arg, int flags) | |
34dc7c2f BB |
1914 | { |
1915 | dsl_dir_t *dd; | |
13fe0198 | 1916 | dsl_pool_t *dp = spa_get_dsl(spa); |
b128c09f | 1917 | dsl_dataset_t *ds; |
34dc7c2f BB |
1918 | zap_cursor_t zc; |
1919 | zap_attribute_t *attr; | |
1920 | char *child; | |
b128c09f BB |
1921 | uint64_t thisobj; |
1922 | int err; | |
34dc7c2f | 1923 | |
13fe0198 MA |
1924 | dsl_pool_config_enter(dp, FTAG); |
1925 | ||
1926 | err = dsl_dir_hold(dp, name, FTAG, &dd, NULL); | |
1927 | if (err != 0) { | |
1928 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f | 1929 | return (err); |
13fe0198 | 1930 | } |
34dc7c2f | 1931 | |
b128c09f BB |
1932 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ |
1933 | if (dd->dd_myname[0] == '$') { | |
13fe0198 MA |
1934 | dsl_dir_rele(dd, FTAG); |
1935 | dsl_pool_config_exit(dp, FTAG); | |
b128c09f BB |
1936 | return (0); |
1937 | } | |
1938 | ||
d683ddbb | 1939 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 1940 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
34dc7c2f BB |
1941 | |
1942 | /* | |
1943 | * Iterate over all children. | |
1944 | */ | |
1945 | if (flags & DS_FIND_CHILDREN) { | |
b128c09f | 1946 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
d683ddbb | 1947 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
34dc7c2f BB |
1948 | zap_cursor_retrieve(&zc, attr) == 0; |
1949 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 MA |
1950 | ASSERT3U(attr->za_integer_length, ==, |
1951 | sizeof (uint64_t)); | |
1952 | ASSERT3U(attr->za_num_integers, ==, 1); | |
34dc7c2f | 1953 | |
428870ff | 1954 | child = kmem_asprintf("%s/%s", name, attr->za_name); |
13fe0198 MA |
1955 | dsl_pool_config_exit(dp, FTAG); |
1956 | err = dmu_objset_find_impl(spa, child, | |
1957 | func, arg, flags); | |
1958 | dsl_pool_config_enter(dp, FTAG); | |
428870ff | 1959 | strfree(child); |
13fe0198 | 1960 | if (err != 0) |
34dc7c2f BB |
1961 | break; |
1962 | } | |
1963 | zap_cursor_fini(&zc); | |
1964 | ||
13fe0198 MA |
1965 | if (err != 0) { |
1966 | dsl_dir_rele(dd, FTAG); | |
1967 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f BB |
1968 | kmem_free(attr, sizeof (zap_attribute_t)); |
1969 | return (err); | |
1970 | } | |
1971 | } | |
1972 | ||
1973 | /* | |
1974 | * Iterate over all snapshots. | |
1975 | */ | |
b128c09f | 1976 | if (flags & DS_FIND_SNAPSHOTS) { |
b128c09f | 1977 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); |
b128c09f BB |
1978 | |
1979 | if (err == 0) { | |
d683ddbb JG |
1980 | uint64_t snapobj; |
1981 | ||
1982 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
b128c09f BB |
1983 | dsl_dataset_rele(ds, FTAG); |
1984 | ||
1985 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
1986 | zap_cursor_retrieve(&zc, attr) == 0; | |
1987 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 | 1988 | ASSERT3U(attr->za_integer_length, ==, |
b128c09f | 1989 | sizeof (uint64_t)); |
13fe0198 | 1990 | ASSERT3U(attr->za_num_integers, ==, 1); |
b128c09f | 1991 | |
428870ff BB |
1992 | child = kmem_asprintf("%s@%s", |
1993 | name, attr->za_name); | |
13fe0198 MA |
1994 | dsl_pool_config_exit(dp, FTAG); |
1995 | err = func(child, arg); | |
1996 | dsl_pool_config_enter(dp, FTAG); | |
428870ff | 1997 | strfree(child); |
13fe0198 | 1998 | if (err != 0) |
b128c09f BB |
1999 | break; |
2000 | } | |
2001 | zap_cursor_fini(&zc); | |
34dc7c2f | 2002 | } |
34dc7c2f BB |
2003 | } |
2004 | ||
13fe0198 | 2005 | dsl_dir_rele(dd, FTAG); |
34dc7c2f | 2006 | kmem_free(attr, sizeof (zap_attribute_t)); |
13fe0198 | 2007 | dsl_pool_config_exit(dp, FTAG); |
34dc7c2f | 2008 | |
13fe0198 | 2009 | if (err != 0) |
34dc7c2f BB |
2010 | return (err); |
2011 | ||
13fe0198 MA |
2012 | /* Apply to self. */ |
2013 | return (func(name, arg)); | |
34dc7c2f BB |
2014 | } |
2015 | ||
13fe0198 MA |
2016 | /* |
2017 | * See comment above dmu_objset_find_impl(). | |
2018 | */ | |
d164b209 | 2019 | int |
13fe0198 MA |
2020 | dmu_objset_find(char *name, int func(const char *, void *), void *arg, |
2021 | int flags) | |
d164b209 | 2022 | { |
13fe0198 MA |
2023 | spa_t *spa; |
2024 | int error; | |
d164b209 | 2025 | |
13fe0198 MA |
2026 | error = spa_open(name, &spa, FTAG); |
2027 | if (error != 0) | |
2028 | return (error); | |
2029 | error = dmu_objset_find_impl(spa, name, func, arg, flags); | |
2030 | spa_close(spa, FTAG); | |
2031 | return (error); | |
d164b209 BB |
2032 | } |
2033 | ||
34dc7c2f BB |
2034 | void |
2035 | dmu_objset_set_user(objset_t *os, void *user_ptr) | |
2036 | { | |
428870ff BB |
2037 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
2038 | os->os_user_ptr = user_ptr; | |
34dc7c2f BB |
2039 | } |
2040 | ||
2041 | void * | |
2042 | dmu_objset_get_user(objset_t *os) | |
2043 | { | |
428870ff BB |
2044 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
2045 | return (os->os_user_ptr); | |
34dc7c2f | 2046 | } |
c28b2279 | 2047 | |
13fe0198 MA |
2048 | /* |
2049 | * Determine name of filesystem, given name of snapshot. | |
eca7b760 | 2050 | * buf must be at least ZFS_MAX_DATASET_NAME_LEN bytes |
13fe0198 MA |
2051 | */ |
2052 | int | |
2053 | dmu_fsname(const char *snapname, char *buf) | |
2054 | { | |
2055 | char *atp = strchr(snapname, '@'); | |
2056 | if (atp == NULL) | |
2e528b49 | 2057 | return (SET_ERROR(EINVAL)); |
eca7b760 | 2058 | if (atp - snapname >= ZFS_MAX_DATASET_NAME_LEN) |
2e528b49 | 2059 | return (SET_ERROR(ENAMETOOLONG)); |
13fe0198 MA |
2060 | (void) strlcpy(buf, snapname, atp - snapname + 1); |
2061 | return (0); | |
2062 | } | |
2063 | ||
c28b2279 | 2064 | #if defined(_KERNEL) && defined(HAVE_SPL) |
f0fd83be | 2065 | EXPORT_SYMBOL(dmu_objset_zil); |
c28b2279 | 2066 | EXPORT_SYMBOL(dmu_objset_pool); |
f0fd83be BB |
2067 | EXPORT_SYMBOL(dmu_objset_ds); |
2068 | EXPORT_SYMBOL(dmu_objset_type); | |
c28b2279 BB |
2069 | EXPORT_SYMBOL(dmu_objset_name); |
2070 | EXPORT_SYMBOL(dmu_objset_hold); | |
2071 | EXPORT_SYMBOL(dmu_objset_own); | |
2072 | EXPORT_SYMBOL(dmu_objset_rele); | |
2073 | EXPORT_SYMBOL(dmu_objset_disown); | |
2074 | EXPORT_SYMBOL(dmu_objset_from_ds); | |
2075 | EXPORT_SYMBOL(dmu_objset_create); | |
2076 | EXPORT_SYMBOL(dmu_objset_clone); | |
c28b2279 BB |
2077 | EXPORT_SYMBOL(dmu_objset_stats); |
2078 | EXPORT_SYMBOL(dmu_objset_fast_stat); | |
54a179e7 | 2079 | EXPORT_SYMBOL(dmu_objset_spa); |
c28b2279 BB |
2080 | EXPORT_SYMBOL(dmu_objset_space); |
2081 | EXPORT_SYMBOL(dmu_objset_fsid_guid); | |
2082 | EXPORT_SYMBOL(dmu_objset_find); | |
c28b2279 BB |
2083 | EXPORT_SYMBOL(dmu_objset_byteswap); |
2084 | EXPORT_SYMBOL(dmu_objset_evict_dbufs); | |
2085 | EXPORT_SYMBOL(dmu_objset_snap_cmtime); | |
50c957f7 | 2086 | EXPORT_SYMBOL(dmu_objset_dnodesize); |
c28b2279 BB |
2087 | |
2088 | EXPORT_SYMBOL(dmu_objset_sync); | |
2089 | EXPORT_SYMBOL(dmu_objset_is_dirty); | |
2090 | EXPORT_SYMBOL(dmu_objset_create_impl); | |
2091 | EXPORT_SYMBOL(dmu_objset_open_impl); | |
2092 | EXPORT_SYMBOL(dmu_objset_evict); | |
2093 | EXPORT_SYMBOL(dmu_objset_register_type); | |
2094 | EXPORT_SYMBOL(dmu_objset_do_userquota_updates); | |
2095 | EXPORT_SYMBOL(dmu_objset_userquota_get_ids); | |
2096 | EXPORT_SYMBOL(dmu_objset_userused_enabled); | |
2097 | EXPORT_SYMBOL(dmu_objset_userspace_upgrade); | |
2098 | EXPORT_SYMBOL(dmu_objset_userspace_present); | |
2099 | #endif |