]>
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)) { |
d3c2ae1c | 377 | arc_buf_t *buf = arc_alloc_buf(spa, |
428870ff | 378 | sizeof (objset_phys_t), &os->os_phys_buf, |
9babb374 BB |
379 | ARC_BUFC_METADATA); |
380 | bzero(buf->b_data, sizeof (objset_phys_t)); | |
428870ff BB |
381 | bcopy(os->os_phys_buf->b_data, buf->b_data, |
382 | arc_buf_size(os->os_phys_buf)); | |
d3c2ae1c | 383 | arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); |
428870ff | 384 | os->os_phys_buf = buf; |
9babb374 BB |
385 | } |
386 | ||
428870ff BB |
387 | os->os_phys = os->os_phys_buf->b_data; |
388 | os->os_flags = os->os_phys->os_flags; | |
34dc7c2f | 389 | } else { |
9babb374 BB |
390 | int size = spa_version(spa) >= SPA_VERSION_USERSPACE ? |
391 | sizeof (objset_phys_t) : OBJSET_OLD_PHYS_SIZE; | |
d3c2ae1c | 392 | os->os_phys_buf = arc_alloc_buf(spa, size, |
428870ff BB |
393 | &os->os_phys_buf, ARC_BUFC_METADATA); |
394 | os->os_phys = os->os_phys_buf->b_data; | |
395 | bzero(os->os_phys, size); | |
34dc7c2f BB |
396 | } |
397 | ||
398 | /* | |
399 | * Note: the changed_cb will be called once before the register | |
400 | * func returns, thus changing the checksum/compression from the | |
b128c09f BB |
401 | * default (fletcher2/off). Snapshots don't need to know about |
402 | * checksum/compression/copies. | |
34dc7c2f | 403 | */ |
9b67f605 | 404 | if (ds != NULL) { |
47dfff3b MA |
405 | boolean_t needlock = B_FALSE; |
406 | ||
407 | /* | |
408 | * Note: it's valid to open the objset if the dataset is | |
409 | * long-held, in which case the pool_config lock will not | |
410 | * be held. | |
411 | */ | |
412 | if (!dsl_pool_config_held(dmu_objset_pool(os))) { | |
413 | needlock = B_TRUE; | |
414 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); | |
415 | } | |
13fe0198 MA |
416 | err = dsl_prop_register(ds, |
417 | zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE), | |
428870ff | 418 | primary_cache_changed_cb, os); |
13fe0198 MA |
419 | if (err == 0) { |
420 | err = dsl_prop_register(ds, | |
421 | zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE), | |
428870ff | 422 | secondary_cache_changed_cb, os); |
13fe0198 | 423 | } |
0c66c32d | 424 | if (!ds->ds_is_snapshot) { |
13fe0198 MA |
425 | if (err == 0) { |
426 | err = dsl_prop_register(ds, | |
427 | zfs_prop_to_name(ZFS_PROP_CHECKSUM), | |
428870ff | 428 | checksum_changed_cb, os); |
13fe0198 MA |
429 | } |
430 | if (err == 0) { | |
431 | err = dsl_prop_register(ds, | |
432 | zfs_prop_to_name(ZFS_PROP_COMPRESSION), | |
428870ff | 433 | compression_changed_cb, os); |
13fe0198 MA |
434 | } |
435 | if (err == 0) { | |
436 | err = dsl_prop_register(ds, | |
437 | zfs_prop_to_name(ZFS_PROP_COPIES), | |
428870ff | 438 | copies_changed_cb, os); |
13fe0198 MA |
439 | } |
440 | if (err == 0) { | |
441 | err = dsl_prop_register(ds, | |
442 | zfs_prop_to_name(ZFS_PROP_DEDUP), | |
428870ff | 443 | dedup_changed_cb, os); |
13fe0198 MA |
444 | } |
445 | if (err == 0) { | |
446 | err = dsl_prop_register(ds, | |
447 | zfs_prop_to_name(ZFS_PROP_LOGBIAS), | |
428870ff | 448 | logbias_changed_cb, os); |
13fe0198 MA |
449 | } |
450 | if (err == 0) { | |
451 | err = dsl_prop_register(ds, | |
452 | zfs_prop_to_name(ZFS_PROP_SYNC), | |
428870ff | 453 | sync_changed_cb, os); |
13fe0198 | 454 | } |
faf0f58c MA |
455 | if (err == 0) { |
456 | err = dsl_prop_register(ds, | |
457 | zfs_prop_to_name( | |
458 | ZFS_PROP_REDUNDANT_METADATA), | |
459 | redundant_metadata_changed_cb, os); | |
460 | } | |
f1512ee6 MA |
461 | if (err == 0) { |
462 | err = dsl_prop_register(ds, | |
463 | zfs_prop_to_name(ZFS_PROP_RECORDSIZE), | |
464 | recordsize_changed_cb, os); | |
465 | } | |
50c957f7 NB |
466 | if (err == 0) { |
467 | err = dsl_prop_register(ds, | |
468 | zfs_prop_to_name(ZFS_PROP_DNODESIZE), | |
469 | dnodesize_changed_cb, os); | |
470 | } | |
b128c09f | 471 | } |
47dfff3b MA |
472 | if (needlock) |
473 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
13fe0198 | 474 | if (err != 0) { |
d3c2ae1c | 475 | arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); |
428870ff | 476 | kmem_free(os, sizeof (objset_t)); |
34dc7c2f BB |
477 | return (err); |
478 | } | |
9b67f605 | 479 | } else { |
34dc7c2f | 480 | /* It's the meta-objset. */ |
428870ff | 481 | os->os_checksum = ZIO_CHECKSUM_FLETCHER_4; |
99197f03 | 482 | os->os_compress = ZIO_COMPRESS_ON; |
428870ff BB |
483 | os->os_copies = spa_max_replication(spa); |
484 | os->os_dedup_checksum = ZIO_CHECKSUM_OFF; | |
faf0f58c MA |
485 | os->os_dedup_verify = B_FALSE; |
486 | os->os_logbias = ZFS_LOGBIAS_LATENCY; | |
487 | os->os_sync = ZFS_SYNC_STANDARD; | |
428870ff BB |
488 | os->os_primary_cache = ZFS_CACHE_ALL; |
489 | os->os_secondary_cache = ZFS_CACHE_ALL; | |
50c957f7 | 490 | os->os_dnodesize = DNODE_MIN_SIZE; |
34dc7c2f BB |
491 | } |
492 | ||
0c66c32d | 493 | if (ds == NULL || !ds->ds_is_snapshot) |
572e2857 | 494 | os->os_zil_header = os->os_phys->os_zil_header; |
428870ff | 495 | os->os_zil = zil_alloc(os, &os->os_zil_header); |
34dc7c2f BB |
496 | |
497 | for (i = 0; i < TXG_SIZE; i++) { | |
428870ff | 498 | list_create(&os->os_dirty_dnodes[i], sizeof (dnode_t), |
34dc7c2f | 499 | offsetof(dnode_t, dn_dirty_link[i])); |
428870ff | 500 | list_create(&os->os_free_dnodes[i], sizeof (dnode_t), |
34dc7c2f BB |
501 | offsetof(dnode_t, dn_dirty_link[i])); |
502 | } | |
428870ff | 503 | list_create(&os->os_dnodes, sizeof (dnode_t), |
34dc7c2f | 504 | offsetof(dnode_t, dn_link)); |
428870ff | 505 | list_create(&os->os_downgraded_dbufs, sizeof (dmu_buf_impl_t), |
34dc7c2f BB |
506 | offsetof(dmu_buf_impl_t, db_link)); |
507 | ||
0c66c32d JG |
508 | list_link_init(&os->os_evicting_node); |
509 | ||
428870ff BB |
510 | mutex_init(&os->os_lock, NULL, MUTEX_DEFAULT, NULL); |
511 | mutex_init(&os->os_obj_lock, NULL, MUTEX_DEFAULT, NULL); | |
512 | mutex_init(&os->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL); | |
513 | ||
0c66c32d JG |
514 | dnode_special_open(os, &os->os_phys->os_meta_dnode, |
515 | DMU_META_DNODE_OBJECT, &os->os_meta_dnode); | |
428870ff | 516 | if (arc_buf_size(os->os_phys_buf) >= sizeof (objset_phys_t)) { |
0c66c32d JG |
517 | dnode_special_open(os, &os->os_phys->os_userused_dnode, |
518 | DMU_USERUSED_OBJECT, &os->os_userused_dnode); | |
519 | dnode_special_open(os, &os->os_phys->os_groupused_dnode, | |
520 | DMU_GROUPUSED_OBJECT, &os->os_groupused_dnode); | |
9babb374 | 521 | } |
34dc7c2f | 522 | |
428870ff | 523 | *osp = os; |
34dc7c2f BB |
524 | return (0); |
525 | } | |
526 | ||
428870ff BB |
527 | int |
528 | dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp) | |
34dc7c2f | 529 | { |
428870ff | 530 | int err = 0; |
34dc7c2f | 531 | |
47dfff3b MA |
532 | /* |
533 | * We shouldn't be doing anything with dsl_dataset_t's unless the | |
534 | * pool_config lock is held, or the dataset is long-held. | |
535 | */ | |
536 | ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool) || | |
537 | dsl_dataset_long_held(ds)); | |
538 | ||
34dc7c2f | 539 | mutex_enter(&ds->ds_opening_lock); |
9b67f605 MA |
540 | if (ds->ds_objset == NULL) { |
541 | objset_t *os; | |
34dc7c2f | 542 | err = dmu_objset_open_impl(dsl_dataset_get_spa(ds), |
9b67f605 MA |
543 | ds, dsl_dataset_get_blkptr(ds), &os); |
544 | ||
545 | if (err == 0) { | |
546 | mutex_enter(&ds->ds_lock); | |
547 | ASSERT(ds->ds_objset == NULL); | |
548 | ds->ds_objset = os; | |
549 | mutex_exit(&ds->ds_lock); | |
550 | } | |
34dc7c2f | 551 | } |
9b67f605 | 552 | *osp = ds->ds_objset; |
34dc7c2f | 553 | mutex_exit(&ds->ds_opening_lock); |
428870ff | 554 | return (err); |
34dc7c2f BB |
555 | } |
556 | ||
13fe0198 MA |
557 | /* |
558 | * Holds the pool while the objset is held. Therefore only one objset | |
559 | * can be held at a time. | |
560 | */ | |
34dc7c2f | 561 | int |
428870ff | 562 | dmu_objset_hold(const char *name, void *tag, objset_t **osp) |
34dc7c2f | 563 | { |
13fe0198 | 564 | dsl_pool_t *dp; |
428870ff | 565 | dsl_dataset_t *ds; |
34dc7c2f BB |
566 | int err; |
567 | ||
13fe0198 MA |
568 | err = dsl_pool_hold(name, tag, &dp); |
569 | if (err != 0) | |
570 | return (err); | |
571 | err = dsl_dataset_hold(dp, name, tag, &ds); | |
572 | if (err != 0) { | |
573 | dsl_pool_rele(dp, tag); | |
428870ff | 574 | return (err); |
13fe0198 | 575 | } |
428870ff BB |
576 | |
577 | err = dmu_objset_from_ds(ds, osp); | |
13fe0198 | 578 | if (err != 0) { |
428870ff | 579 | dsl_dataset_rele(ds, tag); |
13fe0198 MA |
580 | dsl_pool_rele(dp, tag); |
581 | } | |
428870ff | 582 | |
34dc7c2f BB |
583 | return (err); |
584 | } | |
585 | ||
9c43027b AJ |
586 | static int |
587 | dmu_objset_own_impl(dsl_dataset_t *ds, dmu_objset_type_t type, | |
588 | boolean_t readonly, void *tag, objset_t **osp) | |
589 | { | |
590 | int err; | |
591 | ||
592 | err = dmu_objset_from_ds(ds, osp); | |
593 | if (err != 0) { | |
594 | dsl_dataset_disown(ds, tag); | |
595 | } else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) { | |
596 | dsl_dataset_disown(ds, tag); | |
597 | return (SET_ERROR(EINVAL)); | |
598 | } else if (!readonly && dsl_dataset_is_snapshot(ds)) { | |
599 | dsl_dataset_disown(ds, tag); | |
600 | return (SET_ERROR(EROFS)); | |
601 | } | |
602 | return (err); | |
603 | } | |
604 | ||
13fe0198 MA |
605 | /* |
606 | * dsl_pool must not be held when this is called. | |
607 | * Upon successful return, there will be a longhold on the dataset, | |
608 | * and the dsl_pool will not be held. | |
609 | */ | |
34dc7c2f | 610 | int |
428870ff BB |
611 | dmu_objset_own(const char *name, dmu_objset_type_t type, |
612 | boolean_t readonly, void *tag, objset_t **osp) | |
34dc7c2f | 613 | { |
13fe0198 | 614 | dsl_pool_t *dp; |
34dc7c2f BB |
615 | dsl_dataset_t *ds; |
616 | int err; | |
617 | ||
13fe0198 MA |
618 | err = dsl_pool_hold(name, FTAG, &dp); |
619 | if (err != 0) | |
620 | return (err); | |
621 | err = dsl_dataset_own(dp, name, tag, &ds); | |
622 | if (err != 0) { | |
623 | dsl_pool_rele(dp, FTAG); | |
34dc7c2f | 624 | return (err); |
13fe0198 | 625 | } |
9c43027b | 626 | err = dmu_objset_own_impl(ds, type, readonly, tag, osp); |
13fe0198 | 627 | dsl_pool_rele(dp, FTAG); |
9c43027b | 628 | |
34dc7c2f BB |
629 | return (err); |
630 | } | |
631 | ||
9c43027b AJ |
632 | int |
633 | dmu_objset_own_obj(dsl_pool_t *dp, uint64_t obj, dmu_objset_type_t type, | |
634 | boolean_t readonly, void *tag, objset_t **osp) | |
635 | { | |
636 | dsl_dataset_t *ds; | |
637 | int err; | |
638 | ||
639 | err = dsl_dataset_own_obj(dp, obj, tag, &ds); | |
640 | if (err != 0) | |
641 | return (err); | |
642 | ||
643 | return (dmu_objset_own_impl(ds, type, readonly, tag, osp)); | |
644 | } | |
645 | ||
34dc7c2f | 646 | void |
428870ff | 647 | dmu_objset_rele(objset_t *os, void *tag) |
34dc7c2f | 648 | { |
13fe0198 | 649 | dsl_pool_t *dp = dmu_objset_pool(os); |
428870ff | 650 | dsl_dataset_rele(os->os_dsl_dataset, tag); |
13fe0198 | 651 | dsl_pool_rele(dp, tag); |
428870ff | 652 | } |
b128c09f | 653 | |
831baf06 KW |
654 | /* |
655 | * When we are called, os MUST refer to an objset associated with a dataset | |
656 | * that is owned by 'tag'; that is, is held and long held by 'tag' and ds_owner | |
657 | * == tag. We will then release and reacquire ownership of the dataset while | |
658 | * holding the pool config_rwlock to avoid intervening namespace or ownership | |
659 | * changes may occur. | |
660 | * | |
661 | * This exists solely to accommodate zfs_ioc_userspace_upgrade()'s desire to | |
662 | * release the hold on its dataset and acquire a new one on the dataset of the | |
663 | * same name so that it can be partially torn down and reconstructed. | |
664 | */ | |
665 | void | |
666 | dmu_objset_refresh_ownership(objset_t *os, void *tag) | |
667 | { | |
668 | dsl_pool_t *dp; | |
669 | dsl_dataset_t *ds, *newds; | |
eca7b760 | 670 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
831baf06 KW |
671 | |
672 | ds = os->os_dsl_dataset; | |
673 | VERIFY3P(ds, !=, NULL); | |
674 | VERIFY3P(ds->ds_owner, ==, tag); | |
675 | VERIFY(dsl_dataset_long_held(ds)); | |
676 | ||
677 | dsl_dataset_name(ds, name); | |
678 | dp = dmu_objset_pool(os); | |
679 | dsl_pool_config_enter(dp, FTAG); | |
680 | dmu_objset_disown(os, tag); | |
681 | VERIFY0(dsl_dataset_own(dp, name, tag, &newds)); | |
682 | VERIFY3P(newds, ==, os->os_dsl_dataset); | |
683 | dsl_pool_config_exit(dp, FTAG); | |
684 | } | |
685 | ||
428870ff BB |
686 | void |
687 | dmu_objset_disown(objset_t *os, void *tag) | |
688 | { | |
689 | dsl_dataset_disown(os->os_dsl_dataset, tag); | |
34dc7c2f BB |
690 | } |
691 | ||
13fe0198 | 692 | void |
34dc7c2f BB |
693 | dmu_objset_evict_dbufs(objset_t *os) |
694 | { | |
0c66c32d | 695 | dnode_t *dn_marker; |
34dc7c2f BB |
696 | dnode_t *dn; |
697 | ||
0c66c32d | 698 | dn_marker = kmem_alloc(sizeof (dnode_t), KM_SLEEP); |
34dc7c2f | 699 | |
0c66c32d JG |
700 | mutex_enter(&os->os_lock); |
701 | dn = list_head(&os->os_dnodes); | |
702 | while (dn != NULL) { | |
703 | /* | |
704 | * Skip dnodes without holds. We have to do this dance | |
705 | * because dnode_add_ref() only works if there is already a | |
706 | * hold. If the dnode has no holds, then it has no dbufs. | |
707 | */ | |
708 | if (dnode_add_ref(dn, FTAG)) { | |
709 | list_insert_after(&os->os_dnodes, dn, dn_marker); | |
710 | mutex_exit(&os->os_lock); | |
34dc7c2f | 711 | |
0c66c32d JG |
712 | dnode_evict_dbufs(dn); |
713 | dnode_rele(dn, FTAG); | |
34dc7c2f | 714 | |
0c66c32d JG |
715 | mutex_enter(&os->os_lock); |
716 | dn = list_next(&os->os_dnodes, dn_marker); | |
717 | list_remove(&os->os_dnodes, dn_marker); | |
718 | } else { | |
719 | dn = list_next(&os->os_dnodes, dn); | |
720 | } | |
721 | } | |
722 | mutex_exit(&os->os_lock); | |
34dc7c2f | 723 | |
0c66c32d | 724 | kmem_free(dn_marker, sizeof (dnode_t)); |
34dc7c2f | 725 | |
0c66c32d JG |
726 | if (DMU_USERUSED_DNODE(os) != NULL) { |
727 | dnode_evict_dbufs(DMU_GROUPUSED_DNODE(os)); | |
728 | dnode_evict_dbufs(DMU_USERUSED_DNODE(os)); | |
34dc7c2f | 729 | } |
0c66c32d | 730 | dnode_evict_dbufs(DMU_META_DNODE(os)); |
34dc7c2f BB |
731 | } |
732 | ||
0c66c32d JG |
733 | /* |
734 | * Objset eviction processing is split into into two pieces. | |
735 | * The first marks the objset as evicting, evicts any dbufs that | |
736 | * have a refcount of zero, and then queues up the objset for the | |
737 | * second phase of eviction. Once os->os_dnodes has been cleared by | |
738 | * dnode_buf_pageout()->dnode_destroy(), the second phase is executed. | |
739 | * The second phase closes the special dnodes, dequeues the objset from | |
740 | * the list of those undergoing eviction, and finally frees the objset. | |
741 | * | |
742 | * NOTE: Due to asynchronous eviction processing (invocation of | |
743 | * dnode_buf_pageout()), it is possible for the meta dnode for the | |
744 | * objset to have no holds even though os->os_dnodes is not empty. | |
745 | */ | |
34dc7c2f | 746 | void |
428870ff | 747 | dmu_objset_evict(objset_t *os) |
34dc7c2f | 748 | { |
d6320ddb | 749 | int t; |
34dc7c2f | 750 | |
6f1ffb06 MA |
751 | dsl_dataset_t *ds = os->os_dsl_dataset; |
752 | ||
d6320ddb | 753 | for (t = 0; t < TXG_SIZE; t++) |
428870ff | 754 | ASSERT(!dmu_objset_is_dirty(os, t)); |
34dc7c2f | 755 | |
0eb21616 JG |
756 | if (ds) |
757 | dsl_prop_unregister_all(ds, os); | |
34dc7c2f | 758 | |
428870ff BB |
759 | if (os->os_sa) |
760 | sa_tear_down(os); | |
761 | ||
13fe0198 | 762 | dmu_objset_evict_dbufs(os); |
34dc7c2f | 763 | |
0c66c32d JG |
764 | mutex_enter(&os->os_lock); |
765 | spa_evicting_os_register(os->os_spa, os); | |
766 | if (list_is_empty(&os->os_dnodes)) { | |
767 | mutex_exit(&os->os_lock); | |
768 | dmu_objset_evict_done(os); | |
769 | } else { | |
770 | mutex_exit(&os->os_lock); | |
771 | } | |
772 | } | |
773 | ||
774 | void | |
775 | dmu_objset_evict_done(objset_t *os) | |
776 | { | |
777 | ASSERT3P(list_head(&os->os_dnodes), ==, NULL); | |
778 | ||
572e2857 BB |
779 | dnode_special_close(&os->os_meta_dnode); |
780 | if (DMU_USERUSED_DNODE(os)) { | |
781 | dnode_special_close(&os->os_userused_dnode); | |
782 | dnode_special_close(&os->os_groupused_dnode); | |
9babb374 | 783 | } |
428870ff BB |
784 | zil_free(os->os_zil); |
785 | ||
d3c2ae1c | 786 | arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); |
572e2857 BB |
787 | |
788 | /* | |
789 | * This is a barrier to prevent the objset from going away in | |
790 | * dnode_move() until we can safely ensure that the objset is still in | |
791 | * use. We consider the objset valid before the barrier and invalid | |
792 | * after the barrier. | |
793 | */ | |
794 | rw_enter(&os_lock, RW_READER); | |
795 | rw_exit(&os_lock); | |
796 | ||
428870ff BB |
797 | mutex_destroy(&os->os_lock); |
798 | mutex_destroy(&os->os_obj_lock); | |
799 | mutex_destroy(&os->os_user_ptr_lock); | |
0c66c32d | 800 | spa_evicting_os_deregister(os->os_spa, os); |
428870ff BB |
801 | kmem_free(os, sizeof (objset_t)); |
802 | } | |
9babb374 | 803 | |
428870ff BB |
804 | timestruc_t |
805 | dmu_objset_snap_cmtime(objset_t *os) | |
806 | { | |
807 | return (dsl_dir_snap_cmtime(os->os_dsl_dataset->ds_dir)); | |
34dc7c2f BB |
808 | } |
809 | ||
810 | /* called from dsl for meta-objset */ | |
428870ff | 811 | objset_t * |
34dc7c2f BB |
812 | dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, |
813 | dmu_objset_type_t type, dmu_tx_t *tx) | |
814 | { | |
428870ff | 815 | objset_t *os; |
34dc7c2f BB |
816 | dnode_t *mdn; |
817 | ||
818 | ASSERT(dmu_tx_is_syncing(tx)); | |
13fe0198 | 819 | |
572e2857 | 820 | if (ds != NULL) |
13fe0198 | 821 | VERIFY0(dmu_objset_from_ds(ds, &os)); |
572e2857 | 822 | else |
13fe0198 | 823 | VERIFY0(dmu_objset_open_impl(spa, NULL, bp, &os)); |
572e2857 BB |
824 | |
825 | mdn = DMU_META_DNODE(os); | |
34dc7c2f | 826 | |
50c957f7 NB |
827 | dnode_allocate(mdn, DMU_OT_DNODE, DNODE_BLOCK_SIZE, DN_MAX_INDBLKSHIFT, |
828 | DMU_OT_NONE, 0, DNODE_MIN_SLOTS, tx); | |
34dc7c2f BB |
829 | |
830 | /* | |
831 | * We don't want to have to increase the meta-dnode's nlevels | |
832 | * later, because then we could do it in quescing context while | |
833 | * we are also accessing it in open context. | |
834 | * | |
835 | * This precaution is not necessary for the MOS (ds == NULL), | |
836 | * because the MOS is only updated in syncing context. | |
837 | * This is most fortunate: the MOS is the only objset that | |
838 | * needs to be synced multiple times as spa_sync() iterates | |
839 | * to convergence, so minimizing its dn_nlevels matters. | |
840 | */ | |
841 | if (ds != NULL) { | |
842 | int levels = 1; | |
843 | ||
844 | /* | |
845 | * Determine the number of levels necessary for the meta-dnode | |
846 | * to contain DN_MAX_OBJECT dnodes. | |
847 | */ | |
848 | while ((uint64_t)mdn->dn_nblkptr << (mdn->dn_datablkshift + | |
849 | (levels - 1) * (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) < | |
850 | DN_MAX_OBJECT * sizeof (dnode_phys_t)) | |
851 | levels++; | |
852 | ||
853 | mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] = | |
854 | mdn->dn_nlevels = levels; | |
855 | } | |
856 | ||
857 | ASSERT(type != DMU_OST_NONE); | |
858 | ASSERT(type != DMU_OST_ANY); | |
859 | ASSERT(type < DMU_OST_NUMTYPES); | |
428870ff BB |
860 | os->os_phys->os_type = type; |
861 | if (dmu_objset_userused_enabled(os)) { | |
862 | os->os_phys->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; | |
863 | os->os_flags = os->os_phys->os_flags; | |
9babb374 | 864 | } |
34dc7c2f BB |
865 | |
866 | dsl_dataset_dirty(ds, tx); | |
867 | ||
428870ff | 868 | return (os); |
34dc7c2f BB |
869 | } |
870 | ||
13fe0198 MA |
871 | typedef struct dmu_objset_create_arg { |
872 | const char *doca_name; | |
873 | cred_t *doca_cred; | |
874 | void (*doca_userfunc)(objset_t *os, void *arg, | |
875 | cred_t *cr, dmu_tx_t *tx); | |
876 | void *doca_userarg; | |
877 | dmu_objset_type_t doca_type; | |
878 | uint64_t doca_flags; | |
879 | } dmu_objset_create_arg_t; | |
34dc7c2f BB |
880 | |
881 | /*ARGSUSED*/ | |
882 | static int | |
13fe0198 | 883 | dmu_objset_create_check(void *arg, dmu_tx_t *tx) |
34dc7c2f | 884 | { |
13fe0198 MA |
885 | dmu_objset_create_arg_t *doca = arg; |
886 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
887 | dsl_dir_t *pdd; | |
888 | const char *tail; | |
889 | int error; | |
34dc7c2f | 890 | |
13fe0198 | 891 | if (strchr(doca->doca_name, '@') != NULL) |
2e528b49 | 892 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 893 | |
eca7b760 IK |
894 | if (strlen(doca->doca_name) >= ZFS_MAX_DATASET_NAME_LEN) |
895 | return (SET_ERROR(ENAMETOOLONG)); | |
896 | ||
13fe0198 MA |
897 | error = dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail); |
898 | if (error != 0) | |
899 | return (error); | |
900 | if (tail == NULL) { | |
901 | dsl_dir_rele(pdd, FTAG); | |
2e528b49 | 902 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 903 | } |
788eb90c JJ |
904 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
905 | doca->doca_cred); | |
13fe0198 | 906 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f | 907 | |
788eb90c | 908 | return (error); |
34dc7c2f BB |
909 | } |
910 | ||
911 | static void | |
13fe0198 | 912 | dmu_objset_create_sync(void *arg, dmu_tx_t *tx) |
34dc7c2f | 913 | { |
13fe0198 MA |
914 | dmu_objset_create_arg_t *doca = arg; |
915 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
916 | dsl_dir_t *pdd; | |
917 | const char *tail; | |
6f1ffb06 | 918 | dsl_dataset_t *ds; |
13fe0198 | 919 | uint64_t obj; |
6f1ffb06 | 920 | blkptr_t *bp; |
13fe0198 | 921 | objset_t *os; |
34dc7c2f | 922 | |
13fe0198 | 923 | VERIFY0(dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail)); |
34dc7c2f | 924 | |
13fe0198 MA |
925 | obj = dsl_dataset_create_sync(pdd, tail, NULL, doca->doca_flags, |
926 | doca->doca_cred, tx); | |
34dc7c2f | 927 | |
13fe0198 | 928 | VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); |
6f1ffb06 | 929 | bp = dsl_dataset_get_blkptr(ds); |
13fe0198 MA |
930 | os = dmu_objset_create_impl(pdd->dd_pool->dp_spa, |
931 | ds, bp, doca->doca_type, tx); | |
34dc7c2f | 932 | |
13fe0198 MA |
933 | if (doca->doca_userfunc != NULL) { |
934 | doca->doca_userfunc(os, doca->doca_userarg, | |
935 | doca->doca_cred, tx); | |
34dc7c2f BB |
936 | } |
937 | ||
13fe0198 | 938 | spa_history_log_internal_ds(ds, "create", tx, ""); |
a0bd735a BP |
939 | zvol_create_minors(dp->dp_spa, doca->doca_name, B_TRUE); |
940 | ||
6f1ffb06 | 941 | dsl_dataset_rele(ds, FTAG); |
13fe0198 | 942 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f BB |
943 | } |
944 | ||
945 | int | |
428870ff | 946 | dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, |
34dc7c2f BB |
947 | void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg) |
948 | { | |
13fe0198 | 949 | dmu_objset_create_arg_t doca; |
34dc7c2f | 950 | |
13fe0198 MA |
951 | doca.doca_name = name; |
952 | doca.doca_cred = CRED(); | |
953 | doca.doca_flags = flags; | |
954 | doca.doca_userfunc = func; | |
955 | doca.doca_userarg = arg; | |
956 | doca.doca_type = type; | |
34dc7c2f | 957 | |
13fe0198 | 958 | return (dsl_sync_task(name, |
3d45fdd6 MA |
959 | dmu_objset_create_check, dmu_objset_create_sync, &doca, |
960 | 5, ZFS_SPACE_CHECK_NORMAL)); | |
34dc7c2f BB |
961 | } |
962 | ||
13fe0198 MA |
963 | typedef struct dmu_objset_clone_arg { |
964 | const char *doca_clone; | |
965 | const char *doca_origin; | |
966 | cred_t *doca_cred; | |
967 | } dmu_objset_clone_arg_t; | |
968 | ||
969 | /*ARGSUSED*/ | |
970 | static int | |
971 | dmu_objset_clone_check(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 972 | { |
13fe0198 | 973 | dmu_objset_clone_arg_t *doca = arg; |
428870ff BB |
974 | dsl_dir_t *pdd; |
975 | const char *tail; | |
13fe0198 MA |
976 | int error; |
977 | dsl_dataset_t *origin; | |
978 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
34dc7c2f | 979 | |
13fe0198 | 980 | if (strchr(doca->doca_clone, '@') != NULL) |
2e528b49 | 981 | return (SET_ERROR(EINVAL)); |
13fe0198 | 982 | |
eca7b760 IK |
983 | if (strlen(doca->doca_clone) >= ZFS_MAX_DATASET_NAME_LEN) |
984 | return (SET_ERROR(ENAMETOOLONG)); | |
985 | ||
13fe0198 MA |
986 | error = dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail); |
987 | if (error != 0) | |
988 | return (error); | |
428870ff | 989 | if (tail == NULL) { |
13fe0198 | 990 | dsl_dir_rele(pdd, FTAG); |
2e528b49 | 991 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 992 | } |
1cddb8c9 | 993 | |
788eb90c JJ |
994 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
995 | doca->doca_cred); | |
996 | if (error != 0) { | |
997 | dsl_dir_rele(pdd, FTAG); | |
998 | return (SET_ERROR(EDQUOT)); | |
999 | } | |
13fe0198 | 1000 | dsl_dir_rele(pdd, FTAG); |
428870ff | 1001 | |
13fe0198 MA |
1002 | error = dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin); |
1003 | if (error != 0) | |
572e2857 BB |
1004 | return (error); |
1005 | ||
13fe0198 | 1006 | /* You can only clone snapshots, not the head datasets. */ |
0c66c32d | 1007 | if (!origin->ds_is_snapshot) { |
13fe0198 | 1008 | dsl_dataset_rele(origin, FTAG); |
2e528b49 | 1009 | return (SET_ERROR(EINVAL)); |
428870ff | 1010 | } |
13fe0198 | 1011 | dsl_dataset_rele(origin, FTAG); |
572e2857 | 1012 | |
13fe0198 | 1013 | return (0); |
9babb374 | 1014 | } |
34dc7c2f | 1015 | |
13fe0198 MA |
1016 | static void |
1017 | dmu_objset_clone_sync(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 1018 | { |
13fe0198 MA |
1019 | dmu_objset_clone_arg_t *doca = arg; |
1020 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
1021 | dsl_dir_t *pdd; | |
1022 | const char *tail; | |
1023 | dsl_dataset_t *origin, *ds; | |
1024 | uint64_t obj; | |
eca7b760 | 1025 | char namebuf[ZFS_MAX_DATASET_NAME_LEN]; |
6f1ffb06 | 1026 | |
13fe0198 MA |
1027 | VERIFY0(dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail)); |
1028 | VERIFY0(dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin)); | |
6f1ffb06 | 1029 | |
13fe0198 MA |
1030 | obj = dsl_dataset_create_sync(pdd, tail, origin, 0, |
1031 | doca->doca_cred, tx); | |
34dc7c2f | 1032 | |
13fe0198 MA |
1033 | VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); |
1034 | dsl_dataset_name(origin, namebuf); | |
1035 | spa_history_log_internal_ds(ds, "clone", tx, | |
1036 | "origin=%s (%llu)", namebuf, origin->ds_object); | |
a0bd735a | 1037 | zvol_create_minors(dp->dp_spa, doca->doca_clone, B_TRUE); |
13fe0198 MA |
1038 | dsl_dataset_rele(ds, FTAG); |
1039 | dsl_dataset_rele(origin, FTAG); | |
1040 | dsl_dir_rele(pdd, FTAG); | |
34dc7c2f BB |
1041 | } |
1042 | ||
1043 | int | |
13fe0198 | 1044 | dmu_objset_clone(const char *clone, const char *origin) |
34dc7c2f | 1045 | { |
13fe0198 | 1046 | dmu_objset_clone_arg_t doca; |
34dc7c2f | 1047 | |
13fe0198 MA |
1048 | doca.doca_clone = clone; |
1049 | doca.doca_origin = origin; | |
1050 | doca.doca_cred = CRED(); | |
572e2857 | 1051 | |
13fe0198 | 1052 | return (dsl_sync_task(clone, |
3d45fdd6 MA |
1053 | dmu_objset_clone_check, dmu_objset_clone_sync, &doca, |
1054 | 5, ZFS_SPACE_CHECK_NORMAL)); | |
6f1ffb06 MA |
1055 | } |
1056 | ||
1057 | int | |
1058 | dmu_objset_snapshot_one(const char *fsname, const char *snapname) | |
1059 | { | |
1060 | int err; | |
1061 | char *longsnap = kmem_asprintf("%s@%s", fsname, snapname); | |
1062 | nvlist_t *snaps = fnvlist_alloc(); | |
1063 | ||
1064 | fnvlist_add_boolean(snaps, longsnap); | |
6f1ffb06 | 1065 | strfree(longsnap); |
13fe0198 MA |
1066 | err = dsl_dataset_snapshot(snaps, NULL, NULL); |
1067 | fnvlist_free(snaps); | |
6f1ffb06 MA |
1068 | return (err); |
1069 | } | |
1070 | ||
34dc7c2f | 1071 | static void |
9babb374 | 1072 | dmu_objset_sync_dnodes(list_t *list, list_t *newlist, dmu_tx_t *tx) |
34dc7c2f BB |
1073 | { |
1074 | dnode_t *dn; | |
1075 | ||
c65aa5b2 | 1076 | while ((dn = list_head(list))) { |
34dc7c2f BB |
1077 | ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT); |
1078 | ASSERT(dn->dn_dbuf->db_data_pending); | |
1079 | /* | |
9babb374 BB |
1080 | * Initialize dn_zio outside dnode_sync() because the |
1081 | * meta-dnode needs to set it ouside dnode_sync(). | |
34dc7c2f BB |
1082 | */ |
1083 | dn->dn_zio = dn->dn_dbuf->db_data_pending->dr_zio; | |
1084 | ASSERT(dn->dn_zio); | |
1085 | ||
1086 | ASSERT3U(dn->dn_nlevels, <=, DN_MAX_LEVELS); | |
1087 | list_remove(list, dn); | |
9babb374 BB |
1088 | |
1089 | if (newlist) { | |
1090 | (void) dnode_add_ref(dn, newlist); | |
1091 | list_insert_tail(newlist, dn); | |
1092 | } | |
1093 | ||
34dc7c2f BB |
1094 | dnode_sync(dn, tx); |
1095 | } | |
1096 | } | |
1097 | ||
1098 | /* ARGSUSED */ | |
1099 | static void | |
428870ff | 1100 | dmu_objset_write_ready(zio_t *zio, arc_buf_t *abuf, void *arg) |
34dc7c2f | 1101 | { |
d6320ddb BB |
1102 | int i; |
1103 | ||
b128c09f | 1104 | blkptr_t *bp = zio->io_bp; |
428870ff | 1105 | objset_t *os = arg; |
34dc7c2f | 1106 | dnode_phys_t *dnp = &os->os_phys->os_meta_dnode; |
34dc7c2f | 1107 | |
9b67f605 | 1108 | ASSERT(!BP_IS_EMBEDDED(bp)); |
13fe0198 MA |
1109 | ASSERT3P(bp, ==, os->os_rootbp); |
1110 | ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_OBJSET); | |
1111 | ASSERT0(BP_GET_LEVEL(bp)); | |
34dc7c2f BB |
1112 | |
1113 | /* | |
9babb374 BB |
1114 | * Update rootbp fill count: it should be the number of objects |
1115 | * allocated in the object set (not counting the "special" | |
1116 | * objects that are stored in the objset_phys_t -- the meta | |
1117 | * dnode and user/group accounting objects). | |
34dc7c2f | 1118 | */ |
9babb374 | 1119 | bp->blk_fill = 0; |
d6320ddb | 1120 | for (i = 0; i < dnp->dn_nblkptr; i++) |
9b67f605 | 1121 | bp->blk_fill += BP_GET_FILL(&dnp->dn_blkptr[i]); |
428870ff BB |
1122 | } |
1123 | ||
1124 | /* ARGSUSED */ | |
1125 | static void | |
1126 | dmu_objset_write_done(zio_t *zio, arc_buf_t *abuf, void *arg) | |
1127 | { | |
1128 | blkptr_t *bp = zio->io_bp; | |
1129 | blkptr_t *bp_orig = &zio->io_bp_orig; | |
1130 | objset_t *os = arg; | |
34dc7c2f | 1131 | |
b128c09f | 1132 | if (zio->io_flags & ZIO_FLAG_IO_REWRITE) { |
428870ff | 1133 | ASSERT(BP_EQUAL(bp, bp_orig)); |
b128c09f | 1134 | } else { |
428870ff BB |
1135 | dsl_dataset_t *ds = os->os_dsl_dataset; |
1136 | dmu_tx_t *tx = os->os_synctx; | |
1137 | ||
1138 | (void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE); | |
1139 | dsl_dataset_block_born(ds, bp, tx); | |
34dc7c2f BB |
1140 | } |
1141 | } | |
1142 | ||
34dc7c2f BB |
1143 | /* called from dsl */ |
1144 | void | |
428870ff | 1145 | dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx) |
34dc7c2f BB |
1146 | { |
1147 | int txgoff; | |
5dbd68a3 | 1148 | zbookmark_phys_t zb; |
428870ff | 1149 | zio_prop_t zp; |
34dc7c2f BB |
1150 | zio_t *zio; |
1151 | list_t *list; | |
9babb374 | 1152 | list_t *newlist = NULL; |
34dc7c2f BB |
1153 | dbuf_dirty_record_t *dr; |
1154 | ||
1155 | dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg); | |
1156 | ||
1157 | ASSERT(dmu_tx_is_syncing(tx)); | |
1158 | /* XXX the write_done callback should really give us the tx... */ | |
1159 | os->os_synctx = tx; | |
1160 | ||
1161 | if (os->os_dsl_dataset == NULL) { | |
1162 | /* | |
1163 | * This is the MOS. If we have upgraded, | |
1164 | * spa_max_replication() could change, so reset | |
1165 | * os_copies here. | |
1166 | */ | |
1167 | os->os_copies = spa_max_replication(os->os_spa); | |
1168 | } | |
1169 | ||
1170 | /* | |
1171 | * Create the root block IO | |
1172 | */ | |
428870ff BB |
1173 | SET_BOOKMARK(&zb, os->os_dsl_dataset ? |
1174 | os->os_dsl_dataset->ds_object : DMU_META_OBJSET, | |
1175 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
294f6806 | 1176 | arc_release(os->os_phys_buf, &os->os_phys_buf); |
b128c09f | 1177 | |
428870ff | 1178 | dmu_write_policy(os, NULL, 0, 0, &zp); |
9babb374 | 1179 | |
428870ff | 1180 | zio = arc_write(pio, os->os_spa, tx->tx_txg, |
3a17a7a9 | 1181 | os->os_rootbp, os->os_phys_buf, DMU_OS_IS_L2CACHEABLE(os), |
bc77ba73 PD |
1182 | &zp, dmu_objset_write_ready, NULL, NULL, dmu_objset_write_done, |
1183 | os, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb); | |
34dc7c2f BB |
1184 | |
1185 | /* | |
9babb374 | 1186 | * Sync special dnodes - the parent IO for the sync is the root block |
34dc7c2f | 1187 | */ |
572e2857 BB |
1188 | DMU_META_DNODE(os)->dn_zio = zio; |
1189 | dnode_sync(DMU_META_DNODE(os), tx); | |
34dc7c2f | 1190 | |
9babb374 BB |
1191 | os->os_phys->os_flags = os->os_flags; |
1192 | ||
572e2857 BB |
1193 | if (DMU_USERUSED_DNODE(os) && |
1194 | DMU_USERUSED_DNODE(os)->dn_type != DMU_OT_NONE) { | |
1195 | DMU_USERUSED_DNODE(os)->dn_zio = zio; | |
1196 | dnode_sync(DMU_USERUSED_DNODE(os), tx); | |
1197 | DMU_GROUPUSED_DNODE(os)->dn_zio = zio; | |
1198 | dnode_sync(DMU_GROUPUSED_DNODE(os), tx); | |
9babb374 BB |
1199 | } |
1200 | ||
34dc7c2f BB |
1201 | txgoff = tx->tx_txg & TXG_MASK; |
1202 | ||
9babb374 BB |
1203 | if (dmu_objset_userused_enabled(os)) { |
1204 | newlist = &os->os_synced_dnodes; | |
1205 | /* | |
1206 | * We must create the list here because it uses the | |
1207 | * dn_dirty_link[] of this txg. | |
1208 | */ | |
1209 | list_create(newlist, sizeof (dnode_t), | |
1210 | offsetof(dnode_t, dn_dirty_link[txgoff])); | |
1211 | } | |
1212 | ||
1213 | dmu_objset_sync_dnodes(&os->os_free_dnodes[txgoff], newlist, tx); | |
1214 | dmu_objset_sync_dnodes(&os->os_dirty_dnodes[txgoff], newlist, tx); | |
34dc7c2f | 1215 | |
572e2857 | 1216 | list = &DMU_META_DNODE(os)->dn_dirty_records[txgoff]; |
6f1ffb06 | 1217 | while ((dr = list_head(list))) { |
13fe0198 | 1218 | ASSERT0(dr->dr_dbuf->db_level); |
34dc7c2f BB |
1219 | list_remove(list, dr); |
1220 | if (dr->dr_zio) | |
1221 | zio_nowait(dr->dr_zio); | |
1222 | } | |
68cbd56e NB |
1223 | |
1224 | /* Enable dnode backfill if enough objects have been freed. */ | |
1225 | if (os->os_freed_dnodes >= dmu_rescan_dnode_threshold) { | |
1226 | os->os_rescan_dnodes = B_TRUE; | |
1227 | os->os_freed_dnodes = 0; | |
1228 | } | |
1229 | ||
34dc7c2f BB |
1230 | /* |
1231 | * Free intent log blocks up to this tx. | |
1232 | */ | |
1233 | zil_sync(os->os_zil, tx); | |
b128c09f | 1234 | os->os_phys->os_zil_header = os->os_zil_header; |
34dc7c2f BB |
1235 | zio_nowait(zio); |
1236 | } | |
1237 | ||
428870ff BB |
1238 | boolean_t |
1239 | dmu_objset_is_dirty(objset_t *os, uint64_t txg) | |
1240 | { | |
1241 | return (!list_is_empty(&os->os_dirty_dnodes[txg & TXG_MASK]) || | |
1242 | !list_is_empty(&os->os_free_dnodes[txg & TXG_MASK])); | |
1243 | } | |
1244 | ||
572e2857 | 1245 | static objset_used_cb_t *used_cbs[DMU_OST_NUMTYPES]; |
9babb374 BB |
1246 | |
1247 | void | |
1248 | dmu_objset_register_type(dmu_objset_type_t ost, objset_used_cb_t *cb) | |
1249 | { | |
1250 | used_cbs[ost] = cb; | |
1251 | } | |
1252 | ||
1253 | boolean_t | |
428870ff | 1254 | dmu_objset_userused_enabled(objset_t *os) |
9babb374 BB |
1255 | { |
1256 | return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE && | |
572e2857 BB |
1257 | used_cbs[os->os_phys->os_type] != NULL && |
1258 | DMU_USERUSED_DNODE(os) != NULL); | |
9babb374 BB |
1259 | } |
1260 | ||
428870ff BB |
1261 | static void |
1262 | do_userquota_update(objset_t *os, uint64_t used, uint64_t flags, | |
1263 | uint64_t user, uint64_t group, boolean_t subtract, dmu_tx_t *tx) | |
1264 | { | |
1265 | if ((flags & DNODE_FLAG_USERUSED_ACCOUNTED)) { | |
50c957f7 | 1266 | int64_t delta = DNODE_MIN_SIZE + used; |
428870ff BB |
1267 | if (subtract) |
1268 | delta = -delta; | |
1269 | VERIFY3U(0, ==, zap_increment_int(os, DMU_USERUSED_OBJECT, | |
1270 | user, delta, tx)); | |
1271 | VERIFY3U(0, ==, zap_increment_int(os, DMU_GROUPUSED_OBJECT, | |
1272 | group, delta, tx)); | |
1273 | } | |
1274 | } | |
1275 | ||
9babb374 | 1276 | void |
428870ff | 1277 | dmu_objset_do_userquota_updates(objset_t *os, dmu_tx_t *tx) |
9babb374 BB |
1278 | { |
1279 | dnode_t *dn; | |
1280 | list_t *list = &os->os_synced_dnodes; | |
9babb374 BB |
1281 | |
1282 | ASSERT(list_head(list) == NULL || dmu_objset_userused_enabled(os)); | |
1283 | ||
6f1ffb06 | 1284 | while ((dn = list_head(list))) { |
572e2857 | 1285 | int flags; |
9babb374 | 1286 | ASSERT(!DMU_OBJECT_IS_SPECIAL(dn->dn_object)); |
9babb374 BB |
1287 | ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE || |
1288 | dn->dn_phys->dn_flags & | |
1289 | DNODE_FLAG_USERUSED_ACCOUNTED); | |
1290 | ||
1291 | /* Allocate the user/groupused objects if necessary. */ | |
572e2857 | 1292 | if (DMU_USERUSED_DNODE(os)->dn_type == DMU_OT_NONE) { |
428870ff | 1293 | VERIFY(0 == zap_create_claim(os, |
9babb374 BB |
1294 | DMU_USERUSED_OBJECT, |
1295 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
428870ff | 1296 | VERIFY(0 == zap_create_claim(os, |
9babb374 BB |
1297 | DMU_GROUPUSED_OBJECT, |
1298 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
1299 | } | |
1300 | ||
1301 | /* | |
428870ff BB |
1302 | * We intentionally modify the zap object even if the |
1303 | * net delta is zero. Otherwise | |
1304 | * the block of the zap obj could be shared between | |
1305 | * datasets but need to be different between them after | |
1306 | * a bprewrite. | |
9babb374 | 1307 | */ |
9babb374 | 1308 | |
572e2857 BB |
1309 | flags = dn->dn_id_flags; |
1310 | ASSERT(flags); | |
1311 | if (flags & DN_ID_OLD_EXIST) { | |
428870ff BB |
1312 | do_userquota_update(os, dn->dn_oldused, dn->dn_oldflags, |
1313 | dn->dn_olduid, dn->dn_oldgid, B_TRUE, tx); | |
1314 | } | |
572e2857 | 1315 | if (flags & DN_ID_NEW_EXIST) { |
428870ff BB |
1316 | do_userquota_update(os, DN_USED_BYTES(dn->dn_phys), |
1317 | dn->dn_phys->dn_flags, dn->dn_newuid, | |
1318 | dn->dn_newgid, B_FALSE, tx); | |
1319 | } | |
1320 | ||
572e2857 | 1321 | mutex_enter(&dn->dn_mtx); |
428870ff BB |
1322 | dn->dn_oldused = 0; |
1323 | dn->dn_oldflags = 0; | |
1324 | if (dn->dn_id_flags & DN_ID_NEW_EXIST) { | |
1325 | dn->dn_olduid = dn->dn_newuid; | |
1326 | dn->dn_oldgid = dn->dn_newgid; | |
1327 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
1328 | if (dn->dn_bonuslen == 0) | |
1329 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1330 | else | |
1331 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1332 | } | |
1333 | dn->dn_id_flags &= ~(DN_ID_NEW_EXIST); | |
9babb374 BB |
1334 | mutex_exit(&dn->dn_mtx); |
1335 | ||
1336 | list_remove(list, dn); | |
1337 | dnode_rele(dn, list); | |
1338 | } | |
1339 | } | |
1340 | ||
428870ff BB |
1341 | /* |
1342 | * Returns a pointer to data to find uid/gid from | |
1343 | * | |
1344 | * If a dirty record for transaction group that is syncing can't | |
1345 | * be found then NULL is returned. In the NULL case it is assumed | |
1346 | * the uid/gid aren't changing. | |
1347 | */ | |
1348 | static void * | |
1349 | dmu_objset_userquota_find_data(dmu_buf_impl_t *db, dmu_tx_t *tx) | |
1350 | { | |
1351 | dbuf_dirty_record_t *dr, **drp; | |
1352 | void *data; | |
1353 | ||
1354 | if (db->db_dirtycnt == 0) | |
1355 | return (db->db.db_data); /* Nothing is changing */ | |
1356 | ||
1357 | for (drp = &db->db_last_dirty; (dr = *drp) != NULL; drp = &dr->dr_next) | |
1358 | if (dr->dr_txg == tx->tx_txg) | |
1359 | break; | |
1360 | ||
572e2857 | 1361 | if (dr == NULL) { |
428870ff | 1362 | data = NULL; |
572e2857 BB |
1363 | } else { |
1364 | dnode_t *dn; | |
1365 | ||
1366 | DB_DNODE_ENTER(dr->dr_dbuf); | |
1367 | dn = DB_DNODE(dr->dr_dbuf); | |
1368 | ||
1369 | if (dn->dn_bonuslen == 0 && | |
1370 | dr->dr_dbuf->db_blkid == DMU_SPILL_BLKID) | |
1371 | data = dr->dt.dl.dr_data->b_data; | |
1372 | else | |
1373 | data = dr->dt.dl.dr_data; | |
1374 | ||
1375 | DB_DNODE_EXIT(dr->dr_dbuf); | |
1376 | } | |
1377 | ||
428870ff BB |
1378 | return (data); |
1379 | } | |
1380 | ||
1381 | void | |
1382 | dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) | |
1383 | { | |
1384 | objset_t *os = dn->dn_objset; | |
1385 | void *data = NULL; | |
1386 | dmu_buf_impl_t *db = NULL; | |
a117a6d6 GW |
1387 | uint64_t *user = NULL; |
1388 | uint64_t *group = NULL; | |
428870ff BB |
1389 | int flags = dn->dn_id_flags; |
1390 | int error; | |
1391 | boolean_t have_spill = B_FALSE; | |
1392 | ||
1393 | if (!dmu_objset_userused_enabled(dn->dn_objset)) | |
1394 | return; | |
1395 | ||
1396 | if (before && (flags & (DN_ID_CHKED_BONUS|DN_ID_OLD_EXIST| | |
1397 | DN_ID_CHKED_SPILL))) | |
1398 | return; | |
1399 | ||
1400 | if (before && dn->dn_bonuslen != 0) | |
1401 | data = DN_BONUS(dn->dn_phys); | |
1402 | else if (!before && dn->dn_bonuslen != 0) { | |
a3000f93 BB |
1403 | if (dn->dn_bonus) { |
1404 | db = dn->dn_bonus; | |
428870ff BB |
1405 | mutex_enter(&db->db_mtx); |
1406 | data = dmu_objset_userquota_find_data(db, tx); | |
1407 | } else { | |
1408 | data = DN_BONUS(dn->dn_phys); | |
1409 | } | |
1410 | } else if (dn->dn_bonuslen == 0 && dn->dn_bonustype == DMU_OT_SA) { | |
1411 | int rf = 0; | |
1412 | ||
1413 | if (RW_WRITE_HELD(&dn->dn_struct_rwlock)) | |
1414 | rf |= DB_RF_HAVESTRUCT; | |
572e2857 BB |
1415 | error = dmu_spill_hold_by_dnode(dn, |
1416 | rf | DB_RF_MUST_SUCCEED, | |
428870ff BB |
1417 | FTAG, (dmu_buf_t **)&db); |
1418 | ASSERT(error == 0); | |
1419 | mutex_enter(&db->db_mtx); | |
1420 | data = (before) ? db->db.db_data : | |
1421 | dmu_objset_userquota_find_data(db, tx); | |
1422 | have_spill = B_TRUE; | |
1423 | } else { | |
1424 | mutex_enter(&dn->dn_mtx); | |
1425 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1426 | mutex_exit(&dn->dn_mtx); | |
1427 | return; | |
1428 | } | |
1429 | ||
1430 | if (before) { | |
1431 | ASSERT(data); | |
1432 | user = &dn->dn_olduid; | |
1433 | group = &dn->dn_oldgid; | |
1434 | } else if (data) { | |
1435 | user = &dn->dn_newuid; | |
1436 | group = &dn->dn_newgid; | |
1437 | } | |
1438 | ||
1439 | /* | |
1440 | * Must always call the callback in case the object | |
1441 | * type has changed and that type isn't an object type to track | |
1442 | */ | |
1443 | error = used_cbs[os->os_phys->os_type](dn->dn_bonustype, data, | |
1444 | user, group); | |
1445 | ||
1446 | /* | |
1447 | * Preserve existing uid/gid when the callback can't determine | |
1448 | * what the new uid/gid are and the callback returned EEXIST. | |
1449 | * The EEXIST error tells us to just use the existing uid/gid. | |
1450 | * If we don't know what the old values are then just assign | |
1451 | * them to 0, since that is a new file being created. | |
1452 | */ | |
1453 | if (!before && data == NULL && error == EEXIST) { | |
1454 | if (flags & DN_ID_OLD_EXIST) { | |
1455 | dn->dn_newuid = dn->dn_olduid; | |
1456 | dn->dn_newgid = dn->dn_oldgid; | |
1457 | } else { | |
1458 | dn->dn_newuid = 0; | |
1459 | dn->dn_newgid = 0; | |
1460 | } | |
1461 | error = 0; | |
1462 | } | |
1463 | ||
1464 | if (db) | |
1465 | mutex_exit(&db->db_mtx); | |
1466 | ||
1467 | mutex_enter(&dn->dn_mtx); | |
1468 | if (error == 0 && before) | |
1469 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
1470 | if (error == 0 && !before) | |
1471 | dn->dn_id_flags |= DN_ID_NEW_EXIST; | |
1472 | ||
1473 | if (have_spill) { | |
1474 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1475 | } else { | |
1476 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1477 | } | |
1478 | mutex_exit(&dn->dn_mtx); | |
a3000f93 | 1479 | if (have_spill) |
428870ff BB |
1480 | dmu_buf_rele((dmu_buf_t *)db, FTAG); |
1481 | } | |
1482 | ||
9babb374 BB |
1483 | boolean_t |
1484 | dmu_objset_userspace_present(objset_t *os) | |
1485 | { | |
428870ff | 1486 | return (os->os_phys->os_flags & |
9babb374 BB |
1487 | OBJSET_FLAG_USERACCOUNTING_COMPLETE); |
1488 | } | |
1489 | ||
1490 | int | |
1491 | dmu_objset_userspace_upgrade(objset_t *os) | |
1492 | { | |
1493 | uint64_t obj; | |
1494 | int err = 0; | |
1495 | ||
1496 | if (dmu_objset_userspace_present(os)) | |
1497 | return (0); | |
428870ff | 1498 | if (!dmu_objset_userused_enabled(os)) |
2e528b49 | 1499 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 1500 | if (dmu_objset_is_snapshot(os)) |
2e528b49 | 1501 | return (SET_ERROR(EINVAL)); |
9babb374 BB |
1502 | |
1503 | /* | |
1504 | * We simply need to mark every object dirty, so that it will be | |
1505 | * synced out and now accounted. If this is called | |
1506 | * concurrently, or if we already did some work before crashing, | |
1507 | * that's fine, since we track each object's accounted state | |
1508 | * independently. | |
1509 | */ | |
1510 | ||
1511 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
45d1cae3 | 1512 | dmu_tx_t *tx; |
9babb374 BB |
1513 | dmu_buf_t *db; |
1514 | int objerr; | |
1515 | ||
1516 | if (issig(JUSTLOOKING) && issig(FORREAL)) | |
2e528b49 | 1517 | return (SET_ERROR(EINTR)); |
9babb374 BB |
1518 | |
1519 | objerr = dmu_bonus_hold(os, obj, FTAG, &db); | |
13fe0198 | 1520 | if (objerr != 0) |
9babb374 | 1521 | continue; |
45d1cae3 | 1522 | tx = dmu_tx_create(os); |
9babb374 BB |
1523 | dmu_tx_hold_bonus(tx, obj); |
1524 | objerr = dmu_tx_assign(tx, TXG_WAIT); | |
13fe0198 | 1525 | if (objerr != 0) { |
9babb374 BB |
1526 | dmu_tx_abort(tx); |
1527 | continue; | |
1528 | } | |
1529 | dmu_buf_will_dirty(db, tx); | |
1530 | dmu_buf_rele(db, FTAG); | |
1531 | dmu_tx_commit(tx); | |
1532 | } | |
1533 | ||
428870ff | 1534 | os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; |
9babb374 BB |
1535 | txg_wait_synced(dmu_objset_pool(os), 0); |
1536 | return (0); | |
1537 | } | |
1538 | ||
34dc7c2f BB |
1539 | void |
1540 | dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, | |
1541 | uint64_t *usedobjsp, uint64_t *availobjsp) | |
1542 | { | |
428870ff | 1543 | dsl_dataset_space(os->os_dsl_dataset, refdbytesp, availbytesp, |
34dc7c2f BB |
1544 | usedobjsp, availobjsp); |
1545 | } | |
1546 | ||
1547 | uint64_t | |
1548 | dmu_objset_fsid_guid(objset_t *os) | |
1549 | { | |
428870ff | 1550 | return (dsl_dataset_fsid_guid(os->os_dsl_dataset)); |
34dc7c2f BB |
1551 | } |
1552 | ||
1553 | void | |
1554 | dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat) | |
1555 | { | |
428870ff BB |
1556 | stat->dds_type = os->os_phys->os_type; |
1557 | if (os->os_dsl_dataset) | |
1558 | dsl_dataset_fast_stat(os->os_dsl_dataset, stat); | |
34dc7c2f BB |
1559 | } |
1560 | ||
1561 | void | |
1562 | dmu_objset_stats(objset_t *os, nvlist_t *nv) | |
1563 | { | |
428870ff BB |
1564 | ASSERT(os->os_dsl_dataset || |
1565 | os->os_phys->os_type == DMU_OST_META); | |
34dc7c2f | 1566 | |
428870ff BB |
1567 | if (os->os_dsl_dataset != NULL) |
1568 | dsl_dataset_stats(os->os_dsl_dataset, nv); | |
34dc7c2f BB |
1569 | |
1570 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_TYPE, | |
428870ff | 1571 | os->os_phys->os_type); |
9babb374 BB |
1572 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERACCOUNTING, |
1573 | dmu_objset_userspace_present(os)); | |
34dc7c2f BB |
1574 | } |
1575 | ||
1576 | int | |
1577 | dmu_objset_is_snapshot(objset_t *os) | |
1578 | { | |
428870ff | 1579 | if (os->os_dsl_dataset != NULL) |
0c66c32d | 1580 | return (os->os_dsl_dataset->ds_is_snapshot); |
34dc7c2f BB |
1581 | else |
1582 | return (B_FALSE); | |
1583 | } | |
1584 | ||
1585 | int | |
1586 | dmu_snapshot_realname(objset_t *os, char *name, char *real, int maxlen, | |
1587 | boolean_t *conflict) | |
1588 | { | |
428870ff | 1589 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
1590 | uint64_t ignored; |
1591 | ||
d683ddbb | 1592 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 1593 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1594 | |
1595 | return (zap_lookup_norm(ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb JG |
1596 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, name, 8, 1, &ignored, |
1597 | MT_FIRST, real, maxlen, conflict)); | |
34dc7c2f BB |
1598 | } |
1599 | ||
1600 | int | |
1601 | dmu_snapshot_list_next(objset_t *os, int namelen, char *name, | |
1602 | uint64_t *idp, uint64_t *offp, boolean_t *case_conflict) | |
1603 | { | |
428870ff | 1604 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
1605 | zap_cursor_t cursor; |
1606 | zap_attribute_t attr; | |
1607 | ||
13fe0198 MA |
1608 | ASSERT(dsl_pool_config_held(dmu_objset_pool(os))); |
1609 | ||
d683ddbb | 1610 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 1611 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1612 | |
1613 | zap_cursor_init_serialized(&cursor, | |
1614 | ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb | 1615 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, *offp); |
34dc7c2f BB |
1616 | |
1617 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
1618 | zap_cursor_fini(&cursor); | |
2e528b49 | 1619 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1620 | } |
1621 | ||
1622 | if (strlen(attr.za_name) + 1 > namelen) { | |
1623 | zap_cursor_fini(&cursor); | |
2e528b49 | 1624 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
1625 | } |
1626 | ||
1627 | (void) strcpy(name, attr.za_name); | |
1628 | if (idp) | |
1629 | *idp = attr.za_first_integer; | |
1630 | if (case_conflict) | |
1631 | *case_conflict = attr.za_normalization_conflict; | |
1632 | zap_cursor_advance(&cursor); | |
1633 | *offp = zap_cursor_serialize(&cursor); | |
1634 | zap_cursor_fini(&cursor); | |
1635 | ||
1636 | return (0); | |
1637 | } | |
1638 | ||
ebe7e575 | 1639 | int |
6772fb67 | 1640 | dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *value) |
ebe7e575 | 1641 | { |
d1d7e268 | 1642 | return (dsl_dataset_snap_lookup(os->os_dsl_dataset, name, value)); |
ebe7e575 BB |
1643 | } |
1644 | ||
34dc7c2f BB |
1645 | int |
1646 | dmu_dir_list_next(objset_t *os, int namelen, char *name, | |
1647 | uint64_t *idp, uint64_t *offp) | |
1648 | { | |
428870ff | 1649 | dsl_dir_t *dd = os->os_dsl_dataset->ds_dir; |
34dc7c2f BB |
1650 | zap_cursor_t cursor; |
1651 | zap_attribute_t attr; | |
1652 | ||
1653 | /* there is no next dir on a snapshot! */ | |
428870ff | 1654 | if (os->os_dsl_dataset->ds_object != |
d683ddbb | 1655 | dsl_dir_phys(dd)->dd_head_dataset_obj) |
2e528b49 | 1656 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1657 | |
1658 | zap_cursor_init_serialized(&cursor, | |
1659 | dd->dd_pool->dp_meta_objset, | |
d683ddbb | 1660 | dsl_dir_phys(dd)->dd_child_dir_zapobj, *offp); |
34dc7c2f BB |
1661 | |
1662 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
1663 | zap_cursor_fini(&cursor); | |
2e528b49 | 1664 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1665 | } |
1666 | ||
1667 | if (strlen(attr.za_name) + 1 > namelen) { | |
1668 | zap_cursor_fini(&cursor); | |
2e528b49 | 1669 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
1670 | } |
1671 | ||
1672 | (void) strcpy(name, attr.za_name); | |
1673 | if (idp) | |
1674 | *idp = attr.za_first_integer; | |
1675 | zap_cursor_advance(&cursor); | |
1676 | *offp = zap_cursor_serialize(&cursor); | |
1677 | zap_cursor_fini(&cursor); | |
1678 | ||
1679 | return (0); | |
1680 | } | |
1681 | ||
9c43027b AJ |
1682 | typedef struct dmu_objset_find_ctx { |
1683 | taskq_t *dc_tq; | |
1684 | dsl_pool_t *dc_dp; | |
1685 | uint64_t dc_ddobj; | |
1686 | int (*dc_func)(dsl_pool_t *, dsl_dataset_t *, void *); | |
1687 | void *dc_arg; | |
1688 | int dc_flags; | |
1689 | kmutex_t *dc_error_lock; | |
1690 | int *dc_error; | |
1691 | } dmu_objset_find_ctx_t; | |
1692 | ||
1693 | static void | |
1694 | dmu_objset_find_dp_impl(dmu_objset_find_ctx_t *dcp) | |
b128c09f | 1695 | { |
9c43027b AJ |
1696 | dsl_pool_t *dp = dcp->dc_dp; |
1697 | dmu_objset_find_ctx_t *child_dcp; | |
13fe0198 MA |
1698 | dsl_dir_t *dd; |
1699 | dsl_dataset_t *ds; | |
1700 | zap_cursor_t zc; | |
1701 | zap_attribute_t *attr; | |
1702 | uint64_t thisobj; | |
9c43027b | 1703 | int err = 0; |
13fe0198 | 1704 | |
9c43027b AJ |
1705 | /* don't process if there already was an error */ |
1706 | if (*dcp->dc_error != 0) | |
1707 | goto out; | |
13fe0198 | 1708 | |
9c43027b | 1709 | err = dsl_dir_hold_obj(dp, dcp->dc_ddobj, NULL, FTAG, &dd); |
13fe0198 | 1710 | if (err != 0) |
9c43027b | 1711 | goto out; |
13fe0198 MA |
1712 | |
1713 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ | |
1714 | if (dd->dd_myname[0] == '$') { | |
1715 | dsl_dir_rele(dd, FTAG); | |
9c43027b | 1716 | goto out; |
13fe0198 MA |
1717 | } |
1718 | ||
d683ddbb | 1719 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 1720 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
13fe0198 MA |
1721 | |
1722 | /* | |
1723 | * Iterate over all children. | |
1724 | */ | |
9c43027b | 1725 | if (dcp->dc_flags & DS_FIND_CHILDREN) { |
13fe0198 | 1726 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
d683ddbb | 1727 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
13fe0198 MA |
1728 | zap_cursor_retrieve(&zc, attr) == 0; |
1729 | (void) zap_cursor_advance(&zc)) { | |
1730 | ASSERT3U(attr->za_integer_length, ==, | |
1731 | sizeof (uint64_t)); | |
1732 | ASSERT3U(attr->za_num_integers, ==, 1); | |
1733 | ||
9c43027b AJ |
1734 | child_dcp = kmem_alloc(sizeof (*child_dcp), KM_SLEEP); |
1735 | *child_dcp = *dcp; | |
1736 | child_dcp->dc_ddobj = attr->za_first_integer; | |
1737 | if (dcp->dc_tq != NULL) | |
1738 | (void) taskq_dispatch(dcp->dc_tq, | |
1739 | dmu_objset_find_dp_cb, child_dcp, TQ_SLEEP); | |
1740 | else | |
1741 | dmu_objset_find_dp_impl(child_dcp); | |
13fe0198 MA |
1742 | } |
1743 | zap_cursor_fini(&zc); | |
13fe0198 MA |
1744 | } |
1745 | ||
1746 | /* | |
1747 | * Iterate over all snapshots. | |
1748 | */ | |
9c43027b | 1749 | if (dcp->dc_flags & DS_FIND_SNAPSHOTS) { |
13fe0198 MA |
1750 | dsl_dataset_t *ds; |
1751 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1752 | ||
1753 | if (err == 0) { | |
d683ddbb JG |
1754 | uint64_t snapobj; |
1755 | ||
1756 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
13fe0198 MA |
1757 | dsl_dataset_rele(ds, FTAG); |
1758 | ||
1759 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
1760 | zap_cursor_retrieve(&zc, attr) == 0; | |
1761 | (void) zap_cursor_advance(&zc)) { | |
1762 | ASSERT3U(attr->za_integer_length, ==, | |
1763 | sizeof (uint64_t)); | |
1764 | ASSERT3U(attr->za_num_integers, ==, 1); | |
1765 | ||
1766 | err = dsl_dataset_hold_obj(dp, | |
1767 | attr->za_first_integer, FTAG, &ds); | |
1768 | if (err != 0) | |
1769 | break; | |
9c43027b | 1770 | err = dcp->dc_func(dp, ds, dcp->dc_arg); |
13fe0198 MA |
1771 | dsl_dataset_rele(ds, FTAG); |
1772 | if (err != 0) | |
1773 | break; | |
1774 | } | |
1775 | zap_cursor_fini(&zc); | |
1776 | } | |
1777 | } | |
1778 | ||
1779 | dsl_dir_rele(dd, FTAG); | |
1780 | kmem_free(attr, sizeof (zap_attribute_t)); | |
1781 | ||
1782 | if (err != 0) | |
9c43027b | 1783 | goto out; |
13fe0198 MA |
1784 | |
1785 | /* | |
1786 | * Apply to self. | |
1787 | */ | |
1788 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1789 | if (err != 0) | |
9c43027b AJ |
1790 | goto out; |
1791 | err = dcp->dc_func(dp, ds, dcp->dc_arg); | |
13fe0198 | 1792 | dsl_dataset_rele(ds, FTAG); |
9c43027b AJ |
1793 | |
1794 | out: | |
1795 | if (err != 0) { | |
1796 | mutex_enter(dcp->dc_error_lock); | |
1797 | /* only keep first error */ | |
1798 | if (*dcp->dc_error == 0) | |
1799 | *dcp->dc_error = err; | |
1800 | mutex_exit(dcp->dc_error_lock); | |
1801 | } | |
1802 | ||
1803 | kmem_free(dcp, sizeof (*dcp)); | |
1804 | } | |
1805 | ||
1806 | static void | |
1807 | dmu_objset_find_dp_cb(void *arg) | |
1808 | { | |
1809 | dmu_objset_find_ctx_t *dcp = arg; | |
1810 | dsl_pool_t *dp = dcp->dc_dp; | |
1811 | ||
5e8cd5d1 AJ |
1812 | /* |
1813 | * We need to get a pool_config_lock here, as there are several | |
1814 | * asssert(pool_config_held) down the stack. Getting a lock via | |
1815 | * dsl_pool_config_enter is risky, as it might be stalled by a | |
1816 | * pending writer. This would deadlock, as the write lock can | |
1817 | * only be granted when our parent thread gives up the lock. | |
1818 | * The _prio interface gives us priority over a pending writer. | |
1819 | */ | |
1820 | dsl_pool_config_enter_prio(dp, FTAG); | |
9c43027b AJ |
1821 | |
1822 | dmu_objset_find_dp_impl(dcp); | |
1823 | ||
1824 | dsl_pool_config_exit(dp, FTAG); | |
1825 | } | |
1826 | ||
1827 | /* | |
1828 | * Find objsets under and including ddobj, call func(ds) on each. | |
1829 | * The order for the enumeration is completely undefined. | |
1830 | * func is called with dsl_pool_config held. | |
1831 | */ | |
1832 | int | |
1833 | dmu_objset_find_dp(dsl_pool_t *dp, uint64_t ddobj, | |
1834 | int func(dsl_pool_t *, dsl_dataset_t *, void *), void *arg, int flags) | |
1835 | { | |
1836 | int error = 0; | |
1837 | taskq_t *tq = NULL; | |
1838 | int ntasks; | |
1839 | dmu_objset_find_ctx_t *dcp; | |
1840 | kmutex_t err_lock; | |
1841 | ||
1842 | mutex_init(&err_lock, NULL, MUTEX_DEFAULT, NULL); | |
1843 | dcp = kmem_alloc(sizeof (*dcp), KM_SLEEP); | |
1844 | dcp->dc_tq = NULL; | |
1845 | dcp->dc_dp = dp; | |
1846 | dcp->dc_ddobj = ddobj; | |
1847 | dcp->dc_func = func; | |
1848 | dcp->dc_arg = arg; | |
1849 | dcp->dc_flags = flags; | |
1850 | dcp->dc_error_lock = &err_lock; | |
1851 | dcp->dc_error = &error; | |
1852 | ||
1853 | if ((flags & DS_FIND_SERIALIZE) || dsl_pool_config_held_writer(dp)) { | |
1854 | /* | |
1855 | * In case a write lock is held we can't make use of | |
1856 | * parallelism, as down the stack of the worker threads | |
1857 | * the lock is asserted via dsl_pool_config_held. | |
1858 | * In case of a read lock this is solved by getting a read | |
1859 | * lock in each worker thread, which isn't possible in case | |
1860 | * of a writer lock. So we fall back to the synchronous path | |
1861 | * here. | |
1862 | * In the future it might be possible to get some magic into | |
1863 | * dsl_pool_config_held in a way that it returns true for | |
1864 | * the worker threads so that a single lock held from this | |
1865 | * thread suffices. For now, stay single threaded. | |
1866 | */ | |
1867 | dmu_objset_find_dp_impl(dcp); | |
e5676636 | 1868 | mutex_destroy(&err_lock); |
9c43027b AJ |
1869 | |
1870 | return (error); | |
1871 | } | |
1872 | ||
1873 | ntasks = dmu_find_threads; | |
1874 | if (ntasks == 0) | |
1875 | ntasks = vdev_count_leaves(dp->dp_spa) * 4; | |
1229323d | 1876 | tq = taskq_create("dmu_objset_find", ntasks, maxclsyspri, ntasks, |
9c43027b AJ |
1877 | INT_MAX, 0); |
1878 | if (tq == NULL) { | |
1879 | kmem_free(dcp, sizeof (*dcp)); | |
e5676636 BB |
1880 | mutex_destroy(&err_lock); |
1881 | ||
9c43027b AJ |
1882 | return (SET_ERROR(ENOMEM)); |
1883 | } | |
1884 | dcp->dc_tq = tq; | |
1885 | ||
1886 | /* dcp will be freed by task */ | |
1887 | (void) taskq_dispatch(tq, dmu_objset_find_dp_cb, dcp, TQ_SLEEP); | |
1888 | ||
1889 | /* | |
1890 | * PORTING: this code relies on the property of taskq_wait to wait | |
1891 | * until no more tasks are queued and no more tasks are active. As | |
1892 | * we always queue new tasks from within other tasks, task_wait | |
1893 | * reliably waits for the full recursion to finish, even though we | |
1894 | * enqueue new tasks after taskq_wait has been called. | |
1895 | * On platforms other than illumos, taskq_wait may not have this | |
1896 | * property. | |
1897 | */ | |
1898 | taskq_wait(tq); | |
1899 | taskq_destroy(tq); | |
1900 | mutex_destroy(&err_lock); | |
1901 | ||
1902 | return (error); | |
b128c09f BB |
1903 | } |
1904 | ||
1905 | /* | |
13fe0198 MA |
1906 | * Find all objsets under name, and for each, call 'func(child_name, arg)'. |
1907 | * The dp_config_rwlock must not be held when this is called, and it | |
1908 | * will not be held when the callback is called. | |
1909 | * Therefore this function should only be used when the pool is not changing | |
1910 | * (e.g. in syncing context), or the callback can deal with the possible races. | |
b128c09f | 1911 | */ |
13fe0198 MA |
1912 | static int |
1913 | dmu_objset_find_impl(spa_t *spa, const char *name, | |
1914 | int func(const char *, void *), void *arg, int flags) | |
34dc7c2f BB |
1915 | { |
1916 | dsl_dir_t *dd; | |
13fe0198 | 1917 | dsl_pool_t *dp = spa_get_dsl(spa); |
b128c09f | 1918 | dsl_dataset_t *ds; |
34dc7c2f BB |
1919 | zap_cursor_t zc; |
1920 | zap_attribute_t *attr; | |
1921 | char *child; | |
b128c09f BB |
1922 | uint64_t thisobj; |
1923 | int err; | |
34dc7c2f | 1924 | |
13fe0198 MA |
1925 | dsl_pool_config_enter(dp, FTAG); |
1926 | ||
1927 | err = dsl_dir_hold(dp, name, FTAG, &dd, NULL); | |
1928 | if (err != 0) { | |
1929 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f | 1930 | return (err); |
13fe0198 | 1931 | } |
34dc7c2f | 1932 | |
b128c09f BB |
1933 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ |
1934 | if (dd->dd_myname[0] == '$') { | |
13fe0198 MA |
1935 | dsl_dir_rele(dd, FTAG); |
1936 | dsl_pool_config_exit(dp, FTAG); | |
b128c09f BB |
1937 | return (0); |
1938 | } | |
1939 | ||
d683ddbb | 1940 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 1941 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
34dc7c2f BB |
1942 | |
1943 | /* | |
1944 | * Iterate over all children. | |
1945 | */ | |
1946 | if (flags & DS_FIND_CHILDREN) { | |
b128c09f | 1947 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
d683ddbb | 1948 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
34dc7c2f BB |
1949 | zap_cursor_retrieve(&zc, attr) == 0; |
1950 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 MA |
1951 | ASSERT3U(attr->za_integer_length, ==, |
1952 | sizeof (uint64_t)); | |
1953 | ASSERT3U(attr->za_num_integers, ==, 1); | |
34dc7c2f | 1954 | |
428870ff | 1955 | child = kmem_asprintf("%s/%s", name, attr->za_name); |
13fe0198 MA |
1956 | dsl_pool_config_exit(dp, FTAG); |
1957 | err = dmu_objset_find_impl(spa, child, | |
1958 | func, arg, flags); | |
1959 | dsl_pool_config_enter(dp, FTAG); | |
428870ff | 1960 | strfree(child); |
13fe0198 | 1961 | if (err != 0) |
34dc7c2f BB |
1962 | break; |
1963 | } | |
1964 | zap_cursor_fini(&zc); | |
1965 | ||
13fe0198 MA |
1966 | if (err != 0) { |
1967 | dsl_dir_rele(dd, FTAG); | |
1968 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f BB |
1969 | kmem_free(attr, sizeof (zap_attribute_t)); |
1970 | return (err); | |
1971 | } | |
1972 | } | |
1973 | ||
1974 | /* | |
1975 | * Iterate over all snapshots. | |
1976 | */ | |
b128c09f | 1977 | if (flags & DS_FIND_SNAPSHOTS) { |
b128c09f | 1978 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); |
b128c09f BB |
1979 | |
1980 | if (err == 0) { | |
d683ddbb JG |
1981 | uint64_t snapobj; |
1982 | ||
1983 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
b128c09f BB |
1984 | dsl_dataset_rele(ds, FTAG); |
1985 | ||
1986 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
1987 | zap_cursor_retrieve(&zc, attr) == 0; | |
1988 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 | 1989 | ASSERT3U(attr->za_integer_length, ==, |
b128c09f | 1990 | sizeof (uint64_t)); |
13fe0198 | 1991 | ASSERT3U(attr->za_num_integers, ==, 1); |
b128c09f | 1992 | |
428870ff BB |
1993 | child = kmem_asprintf("%s@%s", |
1994 | name, attr->za_name); | |
13fe0198 MA |
1995 | dsl_pool_config_exit(dp, FTAG); |
1996 | err = func(child, arg); | |
1997 | dsl_pool_config_enter(dp, FTAG); | |
428870ff | 1998 | strfree(child); |
13fe0198 | 1999 | if (err != 0) |
b128c09f BB |
2000 | break; |
2001 | } | |
2002 | zap_cursor_fini(&zc); | |
34dc7c2f | 2003 | } |
34dc7c2f BB |
2004 | } |
2005 | ||
13fe0198 | 2006 | dsl_dir_rele(dd, FTAG); |
34dc7c2f | 2007 | kmem_free(attr, sizeof (zap_attribute_t)); |
13fe0198 | 2008 | dsl_pool_config_exit(dp, FTAG); |
34dc7c2f | 2009 | |
13fe0198 | 2010 | if (err != 0) |
34dc7c2f BB |
2011 | return (err); |
2012 | ||
13fe0198 MA |
2013 | /* Apply to self. */ |
2014 | return (func(name, arg)); | |
34dc7c2f BB |
2015 | } |
2016 | ||
13fe0198 MA |
2017 | /* |
2018 | * See comment above dmu_objset_find_impl(). | |
2019 | */ | |
d164b209 | 2020 | int |
13fe0198 MA |
2021 | dmu_objset_find(char *name, int func(const char *, void *), void *arg, |
2022 | int flags) | |
d164b209 | 2023 | { |
13fe0198 MA |
2024 | spa_t *spa; |
2025 | int error; | |
d164b209 | 2026 | |
13fe0198 MA |
2027 | error = spa_open(name, &spa, FTAG); |
2028 | if (error != 0) | |
2029 | return (error); | |
2030 | error = dmu_objset_find_impl(spa, name, func, arg, flags); | |
2031 | spa_close(spa, FTAG); | |
2032 | return (error); | |
d164b209 BB |
2033 | } |
2034 | ||
34dc7c2f BB |
2035 | void |
2036 | dmu_objset_set_user(objset_t *os, void *user_ptr) | |
2037 | { | |
428870ff BB |
2038 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
2039 | os->os_user_ptr = user_ptr; | |
34dc7c2f BB |
2040 | } |
2041 | ||
2042 | void * | |
2043 | dmu_objset_get_user(objset_t *os) | |
2044 | { | |
428870ff BB |
2045 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
2046 | return (os->os_user_ptr); | |
34dc7c2f | 2047 | } |
c28b2279 | 2048 | |
13fe0198 MA |
2049 | /* |
2050 | * Determine name of filesystem, given name of snapshot. | |
eca7b760 | 2051 | * buf must be at least ZFS_MAX_DATASET_NAME_LEN bytes |
13fe0198 MA |
2052 | */ |
2053 | int | |
2054 | dmu_fsname(const char *snapname, char *buf) | |
2055 | { | |
2056 | char *atp = strchr(snapname, '@'); | |
2057 | if (atp == NULL) | |
2e528b49 | 2058 | return (SET_ERROR(EINVAL)); |
eca7b760 | 2059 | if (atp - snapname >= ZFS_MAX_DATASET_NAME_LEN) |
2e528b49 | 2060 | return (SET_ERROR(ENAMETOOLONG)); |
13fe0198 MA |
2061 | (void) strlcpy(buf, snapname, atp - snapname + 1); |
2062 | return (0); | |
2063 | } | |
2064 | ||
c28b2279 | 2065 | #if defined(_KERNEL) && defined(HAVE_SPL) |
f0fd83be | 2066 | EXPORT_SYMBOL(dmu_objset_zil); |
c28b2279 | 2067 | EXPORT_SYMBOL(dmu_objset_pool); |
f0fd83be BB |
2068 | EXPORT_SYMBOL(dmu_objset_ds); |
2069 | EXPORT_SYMBOL(dmu_objset_type); | |
c28b2279 BB |
2070 | EXPORT_SYMBOL(dmu_objset_name); |
2071 | EXPORT_SYMBOL(dmu_objset_hold); | |
2072 | EXPORT_SYMBOL(dmu_objset_own); | |
2073 | EXPORT_SYMBOL(dmu_objset_rele); | |
2074 | EXPORT_SYMBOL(dmu_objset_disown); | |
2075 | EXPORT_SYMBOL(dmu_objset_from_ds); | |
2076 | EXPORT_SYMBOL(dmu_objset_create); | |
2077 | EXPORT_SYMBOL(dmu_objset_clone); | |
c28b2279 BB |
2078 | EXPORT_SYMBOL(dmu_objset_stats); |
2079 | EXPORT_SYMBOL(dmu_objset_fast_stat); | |
54a179e7 | 2080 | EXPORT_SYMBOL(dmu_objset_spa); |
c28b2279 BB |
2081 | EXPORT_SYMBOL(dmu_objset_space); |
2082 | EXPORT_SYMBOL(dmu_objset_fsid_guid); | |
2083 | EXPORT_SYMBOL(dmu_objset_find); | |
c28b2279 BB |
2084 | EXPORT_SYMBOL(dmu_objset_byteswap); |
2085 | EXPORT_SYMBOL(dmu_objset_evict_dbufs); | |
2086 | EXPORT_SYMBOL(dmu_objset_snap_cmtime); | |
50c957f7 | 2087 | EXPORT_SYMBOL(dmu_objset_dnodesize); |
c28b2279 BB |
2088 | |
2089 | EXPORT_SYMBOL(dmu_objset_sync); | |
2090 | EXPORT_SYMBOL(dmu_objset_is_dirty); | |
2091 | EXPORT_SYMBOL(dmu_objset_create_impl); | |
2092 | EXPORT_SYMBOL(dmu_objset_open_impl); | |
2093 | EXPORT_SYMBOL(dmu_objset_evict); | |
2094 | EXPORT_SYMBOL(dmu_objset_register_type); | |
2095 | EXPORT_SYMBOL(dmu_objset_do_userquota_updates); | |
2096 | EXPORT_SYMBOL(dmu_objset_userquota_get_ids); | |
2097 | EXPORT_SYMBOL(dmu_objset_userused_enabled); | |
2098 | EXPORT_SYMBOL(dmu_objset_userspace_upgrade); | |
2099 | EXPORT_SYMBOL(dmu_objset_userspace_present); | |
2100 | #endif |