]>
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 | */ | |
9b7b9cd3 | 21 | |
34dc7c2f | 22 | /* |
428870ff | 23 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
64fc7762 | 24 | * Copyright (c) 2012, 2017 by Delphix. All rights reserved. |
3a17a7a9 | 25 | * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. |
788eb90c | 26 | * Copyright (c) 2013, Joyent, Inc. All rights reserved. |
0c66c32d | 27 | * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. |
9c43027b | 28 | * Copyright (c) 2015, STRATO AG, Inc. All rights reserved. |
a0bd735a | 29 | * Copyright (c) 2016 Actifio, Inc. All rights reserved. |
9b7b9cd3 | 30 | * Copyright 2017 Nexenta Systems, Inc. |
c0daec32 | 31 | * Copyright (c) 2017 Open-E, Inc. All Rights Reserved. |
34dc7c2f BB |
32 | */ |
33 | ||
428870ff BB |
34 | /* Portions Copyright 2010 Robert Milkowski */ |
35 | ||
1de321e6 | 36 | #include <sys/zfeature.h> |
34dc7c2f BB |
37 | #include <sys/cred.h> |
38 | #include <sys/zfs_context.h> | |
39 | #include <sys/dmu_objset.h> | |
40 | #include <sys/dsl_dir.h> | |
41 | #include <sys/dsl_dataset.h> | |
42 | #include <sys/dsl_prop.h> | |
43 | #include <sys/dsl_pool.h> | |
44 | #include <sys/dsl_synctask.h> | |
45 | #include <sys/dsl_deleg.h> | |
46 | #include <sys/dnode.h> | |
47 | #include <sys/dbuf.h> | |
48 | #include <sys/zvol.h> | |
49 | #include <sys/dmu_tx.h> | |
34dc7c2f BB |
50 | #include <sys/zap.h> |
51 | #include <sys/zil.h> | |
52 | #include <sys/dmu_impl.h> | |
53 | #include <sys/zfs_ioctl.h> | |
428870ff | 54 | #include <sys/sa.h> |
572e2857 | 55 | #include <sys/zfs_onexit.h> |
13fe0198 | 56 | #include <sys/dsl_destroy.h> |
9c43027b | 57 | #include <sys/vdev.h> |
f74b821a | 58 | #include <sys/policy.h> |
1de321e6 | 59 | #include <sys/spa_impl.h> |
b5256303 | 60 | #include <sys/dmu_send.h> |
9c5167d1 | 61 | #include <sys/zfs_project.h> |
572e2857 BB |
62 | |
63 | /* | |
64 | * Needed to close a window in dnode_move() that allows the objset to be freed | |
65 | * before it can be safely accessed. | |
66 | */ | |
67 | krwlock_t os_lock; | |
68 | ||
9c43027b | 69 | /* |
4e33ba4c | 70 | * Tunable to overwrite the maximum number of threads for the parallelization |
9c43027b AJ |
71 | * of dmu_objset_find_dp, needed to speed up the import of pools with many |
72 | * datasets. | |
73 | * Default is 4 times the number of leaf vdevs. | |
74 | */ | |
75 | int dmu_find_threads = 0; | |
76 | ||
68cbd56e NB |
77 | /* |
78 | * Backfill lower metadnode objects after this many have been freed. | |
79 | * Backfilling negatively impacts object creation rates, so only do it | |
80 | * if there are enough holes to fill. | |
81 | */ | |
82 | int dmu_rescan_dnode_threshold = 1 << DN_MAX_INDBLKSHIFT; | |
83 | ||
c0daec32 AB |
84 | static char *upgrade_tag = "upgrade_tag"; |
85 | ||
9c43027b AJ |
86 | static void dmu_objset_find_dp_cb(void *arg); |
87 | ||
1de321e6 JX |
88 | static void dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb); |
89 | static void dmu_objset_upgrade_stop(objset_t *os); | |
90 | ||
572e2857 BB |
91 | void |
92 | dmu_objset_init(void) | |
93 | { | |
94 | rw_init(&os_lock, NULL, RW_DEFAULT, NULL); | |
95 | } | |
96 | ||
97 | void | |
98 | dmu_objset_fini(void) | |
99 | { | |
100 | rw_destroy(&os_lock); | |
101 | } | |
34dc7c2f BB |
102 | |
103 | spa_t * | |
104 | dmu_objset_spa(objset_t *os) | |
105 | { | |
428870ff | 106 | return (os->os_spa); |
34dc7c2f BB |
107 | } |
108 | ||
109 | zilog_t * | |
110 | dmu_objset_zil(objset_t *os) | |
111 | { | |
428870ff | 112 | return (os->os_zil); |
34dc7c2f BB |
113 | } |
114 | ||
115 | dsl_pool_t * | |
116 | dmu_objset_pool(objset_t *os) | |
117 | { | |
118 | dsl_dataset_t *ds; | |
119 | ||
428870ff | 120 | if ((ds = os->os_dsl_dataset) != NULL && ds->ds_dir) |
34dc7c2f BB |
121 | return (ds->ds_dir->dd_pool); |
122 | else | |
428870ff | 123 | return (spa_get_dsl(os->os_spa)); |
34dc7c2f BB |
124 | } |
125 | ||
126 | dsl_dataset_t * | |
127 | dmu_objset_ds(objset_t *os) | |
128 | { | |
428870ff | 129 | return (os->os_dsl_dataset); |
34dc7c2f BB |
130 | } |
131 | ||
132 | dmu_objset_type_t | |
133 | dmu_objset_type(objset_t *os) | |
134 | { | |
428870ff | 135 | return (os->os_phys->os_type); |
34dc7c2f BB |
136 | } |
137 | ||
138 | void | |
139 | dmu_objset_name(objset_t *os, char *buf) | |
140 | { | |
428870ff | 141 | dsl_dataset_name(os->os_dsl_dataset, buf); |
34dc7c2f BB |
142 | } |
143 | ||
144 | uint64_t | |
145 | dmu_objset_id(objset_t *os) | |
146 | { | |
428870ff | 147 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
148 | |
149 | return (ds ? ds->ds_object : 0); | |
150 | } | |
151 | ||
50c957f7 NB |
152 | uint64_t |
153 | dmu_objset_dnodesize(objset_t *os) | |
154 | { | |
155 | return (os->os_dnodesize); | |
156 | } | |
157 | ||
faf0f58c | 158 | zfs_sync_type_t |
428870ff BB |
159 | dmu_objset_syncprop(objset_t *os) |
160 | { | |
161 | return (os->os_sync); | |
162 | } | |
163 | ||
faf0f58c | 164 | zfs_logbias_op_t |
428870ff BB |
165 | dmu_objset_logbias(objset_t *os) |
166 | { | |
167 | return (os->os_logbias); | |
168 | } | |
169 | ||
34dc7c2f BB |
170 | static void |
171 | checksum_changed_cb(void *arg, uint64_t newval) | |
172 | { | |
428870ff | 173 | objset_t *os = arg; |
34dc7c2f BB |
174 | |
175 | /* | |
176 | * Inheritance should have been done by now. | |
177 | */ | |
178 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
179 | ||
428870ff | 180 | os->os_checksum = zio_checksum_select(newval, ZIO_CHECKSUM_ON_VALUE); |
34dc7c2f BB |
181 | } |
182 | ||
183 | static void | |
184 | compression_changed_cb(void *arg, uint64_t newval) | |
185 | { | |
428870ff | 186 | objset_t *os = arg; |
34dc7c2f BB |
187 | |
188 | /* | |
189 | * Inheritance and range checking should have been done by now. | |
190 | */ | |
191 | ASSERT(newval != ZIO_COMPRESS_INHERIT); | |
192 | ||
99197f03 JG |
193 | os->os_compress = zio_compress_select(os->os_spa, newval, |
194 | ZIO_COMPRESS_ON); | |
34dc7c2f BB |
195 | } |
196 | ||
197 | static void | |
198 | copies_changed_cb(void *arg, uint64_t newval) | |
199 | { | |
428870ff | 200 | objset_t *os = arg; |
34dc7c2f BB |
201 | |
202 | /* | |
203 | * Inheritance and range checking should have been done by now. | |
204 | */ | |
205 | ASSERT(newval > 0); | |
428870ff | 206 | ASSERT(newval <= spa_max_replication(os->os_spa)); |
34dc7c2f | 207 | |
428870ff BB |
208 | os->os_copies = newval; |
209 | } | |
210 | ||
211 | static void | |
212 | dedup_changed_cb(void *arg, uint64_t newval) | |
213 | { | |
214 | objset_t *os = arg; | |
215 | spa_t *spa = os->os_spa; | |
216 | enum zio_checksum checksum; | |
217 | ||
218 | /* | |
219 | * Inheritance should have been done by now. | |
220 | */ | |
221 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
222 | ||
223 | checksum = zio_checksum_dedup_select(spa, newval, ZIO_CHECKSUM_OFF); | |
224 | ||
225 | os->os_dedup_checksum = checksum & ZIO_CHECKSUM_MASK; | |
226 | os->os_dedup_verify = !!(checksum & ZIO_CHECKSUM_VERIFY); | |
34dc7c2f BB |
227 | } |
228 | ||
b128c09f BB |
229 | static void |
230 | primary_cache_changed_cb(void *arg, uint64_t newval) | |
231 | { | |
428870ff | 232 | objset_t *os = arg; |
b128c09f BB |
233 | |
234 | /* | |
235 | * Inheritance and range checking should have been done by now. | |
236 | */ | |
237 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
238 | newval == ZFS_CACHE_METADATA); | |
239 | ||
428870ff | 240 | os->os_primary_cache = newval; |
b128c09f BB |
241 | } |
242 | ||
243 | static void | |
244 | secondary_cache_changed_cb(void *arg, uint64_t newval) | |
245 | { | |
428870ff | 246 | objset_t *os = arg; |
b128c09f BB |
247 | |
248 | /* | |
249 | * Inheritance and range checking should have been done by now. | |
250 | */ | |
251 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
252 | newval == ZFS_CACHE_METADATA); | |
253 | ||
428870ff BB |
254 | os->os_secondary_cache = newval; |
255 | } | |
256 | ||
257 | static void | |
258 | sync_changed_cb(void *arg, uint64_t newval) | |
259 | { | |
260 | objset_t *os = arg; | |
261 | ||
262 | /* | |
263 | * Inheritance and range checking should have been done by now. | |
264 | */ | |
265 | ASSERT(newval == ZFS_SYNC_STANDARD || newval == ZFS_SYNC_ALWAYS || | |
266 | newval == ZFS_SYNC_DISABLED); | |
267 | ||
268 | os->os_sync = newval; | |
269 | if (os->os_zil) | |
270 | zil_set_sync(os->os_zil, newval); | |
271 | } | |
272 | ||
faf0f58c MA |
273 | static void |
274 | redundant_metadata_changed_cb(void *arg, uint64_t newval) | |
275 | { | |
276 | objset_t *os = arg; | |
277 | ||
278 | /* | |
279 | * Inheritance and range checking should have been done by now. | |
280 | */ | |
281 | ASSERT(newval == ZFS_REDUNDANT_METADATA_ALL || | |
282 | newval == ZFS_REDUNDANT_METADATA_MOST); | |
283 | ||
284 | os->os_redundant_metadata = newval; | |
285 | } | |
286 | ||
50c957f7 NB |
287 | static void |
288 | dnodesize_changed_cb(void *arg, uint64_t newval) | |
289 | { | |
290 | objset_t *os = arg; | |
291 | ||
292 | switch (newval) { | |
293 | case ZFS_DNSIZE_LEGACY: | |
294 | os->os_dnodesize = DNODE_MIN_SIZE; | |
295 | break; | |
296 | case ZFS_DNSIZE_AUTO: | |
297 | /* | |
298 | * Choose a dnode size that will work well for most | |
299 | * workloads if the user specified "auto". Future code | |
300 | * improvements could dynamically select a dnode size | |
301 | * based on observed workload patterns. | |
302 | */ | |
303 | os->os_dnodesize = DNODE_MIN_SIZE * 2; | |
304 | break; | |
305 | case ZFS_DNSIZE_1K: | |
306 | case ZFS_DNSIZE_2K: | |
307 | case ZFS_DNSIZE_4K: | |
308 | case ZFS_DNSIZE_8K: | |
309 | case ZFS_DNSIZE_16K: | |
310 | os->os_dnodesize = newval; | |
311 | break; | |
312 | } | |
313 | } | |
314 | ||
428870ff BB |
315 | static void |
316 | logbias_changed_cb(void *arg, uint64_t newval) | |
317 | { | |
318 | objset_t *os = arg; | |
319 | ||
320 | ASSERT(newval == ZFS_LOGBIAS_LATENCY || | |
321 | newval == ZFS_LOGBIAS_THROUGHPUT); | |
322 | os->os_logbias = newval; | |
323 | if (os->os_zil) | |
324 | zil_set_logbias(os->os_zil, newval); | |
b128c09f BB |
325 | } |
326 | ||
f1512ee6 MA |
327 | static void |
328 | recordsize_changed_cb(void *arg, uint64_t newval) | |
329 | { | |
330 | objset_t *os = arg; | |
331 | ||
332 | os->os_recordsize = newval; | |
333 | } | |
334 | ||
34dc7c2f BB |
335 | void |
336 | dmu_objset_byteswap(void *buf, size_t size) | |
337 | { | |
338 | objset_phys_t *osp = buf; | |
339 | ||
9c5167d1 NF |
340 | ASSERT(size == OBJSET_PHYS_SIZE_V1 || size == OBJSET_PHYS_SIZE_V2 || |
341 | size == sizeof (objset_phys_t)); | |
34dc7c2f BB |
342 | dnode_byteswap(&osp->os_meta_dnode); |
343 | byteswap_uint64_array(&osp->os_zil_header, sizeof (zil_header_t)); | |
344 | osp->os_type = BSWAP_64(osp->os_type); | |
9babb374 | 345 | osp->os_flags = BSWAP_64(osp->os_flags); |
9c5167d1 | 346 | if (size >= OBJSET_PHYS_SIZE_V2) { |
9babb374 BB |
347 | dnode_byteswap(&osp->os_userused_dnode); |
348 | dnode_byteswap(&osp->os_groupused_dnode); | |
9c5167d1 NF |
349 | if (size >= sizeof (objset_phys_t)) |
350 | dnode_byteswap(&osp->os_projectused_dnode); | |
9babb374 | 351 | } |
34dc7c2f BB |
352 | } |
353 | ||
64fc7762 MA |
354 | /* |
355 | * The hash is a CRC-based hash of the objset_t pointer and the object number. | |
356 | */ | |
357 | static uint64_t | |
358 | dnode_hash(const objset_t *os, uint64_t obj) | |
359 | { | |
360 | uintptr_t osv = (uintptr_t)os; | |
361 | uint64_t crc = -1ULL; | |
362 | ||
363 | ASSERT(zfs_crc64_table[128] == ZFS_CRC64_POLY); | |
364 | /* | |
365 | * The low 6 bits of the pointer don't have much entropy, because | |
366 | * the objset_t is larger than 2^6 bytes long. | |
367 | */ | |
368 | crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (osv >> 6)) & 0xFF]; | |
369 | crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 0)) & 0xFF]; | |
370 | crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 8)) & 0xFF]; | |
371 | crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 16)) & 0xFF]; | |
372 | ||
373 | crc ^= (osv>>14) ^ (obj>>24); | |
374 | ||
375 | return (crc); | |
376 | } | |
377 | ||
378 | unsigned int | |
379 | dnode_multilist_index_func(multilist_t *ml, void *obj) | |
380 | { | |
381 | dnode_t *dn = obj; | |
382 | return (dnode_hash(dn->dn_objset, dn->dn_object) % | |
383 | multilist_get_num_sublists(ml)); | |
384 | } | |
385 | ||
34dc7c2f BB |
386 | int |
387 | dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, | |
428870ff | 388 | objset_t **osp) |
34dc7c2f | 389 | { |
428870ff | 390 | objset_t *os; |
b128c09f | 391 | int i, err; |
34dc7c2f BB |
392 | |
393 | ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock)); | |
394 | ||
79c76d5b | 395 | os = kmem_zalloc(sizeof (objset_t), KM_SLEEP); |
428870ff BB |
396 | os->os_dsl_dataset = ds; |
397 | os->os_spa = spa; | |
398 | os->os_rootbp = bp; | |
399 | if (!BP_IS_HOLE(os->os_rootbp)) { | |
2a432414 | 400 | arc_flags_t aflags = ARC_FLAG_WAIT; |
5dbd68a3 | 401 | zbookmark_phys_t zb; |
9c5167d1 | 402 | int size; |
b5256303 | 403 | enum zio_flag zio_flags = ZIO_FLAG_CANFAIL; |
428870ff BB |
404 | SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET, |
405 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
406 | ||
407 | if (DMU_OS_IS_L2CACHEABLE(os)) | |
2a432414 | 408 | aflags |= ARC_FLAG_L2CACHE; |
34dc7c2f | 409 | |
b5256303 TC |
410 | if (ds != NULL && ds->ds_dir->dd_crypto_obj != 0) { |
411 | ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF); | |
412 | ASSERT(BP_IS_AUTHENTICATED(bp)); | |
413 | zio_flags |= ZIO_FLAG_RAW; | |
414 | } | |
415 | ||
428870ff | 416 | dprintf_bp(os->os_rootbp, "reading %s", ""); |
294f6806 | 417 | err = arc_read(NULL, spa, os->os_rootbp, |
428870ff | 418 | arc_getbuf_func, &os->os_phys_buf, |
b5256303 | 419 | ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zb); |
13fe0198 | 420 | if (err != 0) { |
428870ff | 421 | kmem_free(os, sizeof (objset_t)); |
b128c09f BB |
422 | /* convert checksum errors into IO errors */ |
423 | if (err == ECKSUM) | |
2e528b49 | 424 | err = SET_ERROR(EIO); |
34dc7c2f BB |
425 | return (err); |
426 | } | |
9babb374 | 427 | |
9c5167d1 NF |
428 | if (spa_version(spa) < SPA_VERSION_USERSPACE) |
429 | size = OBJSET_PHYS_SIZE_V1; | |
430 | else if (!spa_feature_is_enabled(spa, | |
431 | SPA_FEATURE_PROJECT_QUOTA)) | |
432 | size = OBJSET_PHYS_SIZE_V2; | |
433 | else | |
434 | size = sizeof (objset_phys_t); | |
435 | ||
9babb374 | 436 | /* Increase the blocksize if we are permitted. */ |
9c5167d1 | 437 | if (arc_buf_size(os->os_phys_buf) < size) { |
2aa34383 | 438 | arc_buf_t *buf = arc_alloc_buf(spa, &os->os_phys_buf, |
9c5167d1 NF |
439 | ARC_BUFC_METADATA, size); |
440 | bzero(buf->b_data, size); | |
428870ff BB |
441 | bcopy(os->os_phys_buf->b_data, buf->b_data, |
442 | arc_buf_size(os->os_phys_buf)); | |
d3c2ae1c | 443 | arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); |
428870ff | 444 | os->os_phys_buf = buf; |
9babb374 BB |
445 | } |
446 | ||
428870ff BB |
447 | os->os_phys = os->os_phys_buf->b_data; |
448 | os->os_flags = os->os_phys->os_flags; | |
34dc7c2f | 449 | } else { |
9babb374 | 450 | int size = spa_version(spa) >= SPA_VERSION_USERSPACE ? |
9c5167d1 | 451 | sizeof (objset_phys_t) : OBJSET_PHYS_SIZE_V1; |
2aa34383 DK |
452 | os->os_phys_buf = arc_alloc_buf(spa, &os->os_phys_buf, |
453 | ARC_BUFC_METADATA, size); | |
428870ff BB |
454 | os->os_phys = os->os_phys_buf->b_data; |
455 | bzero(os->os_phys, size); | |
34dc7c2f BB |
456 | } |
457 | ||
458 | /* | |
459 | * Note: the changed_cb will be called once before the register | |
460 | * func returns, thus changing the checksum/compression from the | |
b128c09f BB |
461 | * default (fletcher2/off). Snapshots don't need to know about |
462 | * checksum/compression/copies. | |
34dc7c2f | 463 | */ |
9b67f605 | 464 | if (ds != NULL) { |
47dfff3b MA |
465 | boolean_t needlock = B_FALSE; |
466 | ||
b5256303 TC |
467 | os->os_encrypted = (ds->ds_dir->dd_crypto_obj != 0); |
468 | ||
47dfff3b MA |
469 | /* |
470 | * Note: it's valid to open the objset if the dataset is | |
471 | * long-held, in which case the pool_config lock will not | |
472 | * be held. | |
473 | */ | |
474 | if (!dsl_pool_config_held(dmu_objset_pool(os))) { | |
475 | needlock = B_TRUE; | |
476 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); | |
477 | } | |
b5256303 | 478 | |
13fe0198 MA |
479 | err = dsl_prop_register(ds, |
480 | zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE), | |
428870ff | 481 | primary_cache_changed_cb, os); |
13fe0198 MA |
482 | if (err == 0) { |
483 | err = dsl_prop_register(ds, | |
484 | zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE), | |
428870ff | 485 | secondary_cache_changed_cb, os); |
13fe0198 | 486 | } |
0c66c32d | 487 | if (!ds->ds_is_snapshot) { |
13fe0198 MA |
488 | if (err == 0) { |
489 | err = dsl_prop_register(ds, | |
490 | zfs_prop_to_name(ZFS_PROP_CHECKSUM), | |
428870ff | 491 | checksum_changed_cb, os); |
13fe0198 MA |
492 | } |
493 | if (err == 0) { | |
494 | err = dsl_prop_register(ds, | |
495 | zfs_prop_to_name(ZFS_PROP_COMPRESSION), | |
428870ff | 496 | compression_changed_cb, os); |
13fe0198 MA |
497 | } |
498 | if (err == 0) { | |
499 | err = dsl_prop_register(ds, | |
500 | zfs_prop_to_name(ZFS_PROP_COPIES), | |
428870ff | 501 | copies_changed_cb, os); |
13fe0198 MA |
502 | } |
503 | if (err == 0) { | |
504 | err = dsl_prop_register(ds, | |
505 | zfs_prop_to_name(ZFS_PROP_DEDUP), | |
428870ff | 506 | dedup_changed_cb, os); |
13fe0198 MA |
507 | } |
508 | if (err == 0) { | |
509 | err = dsl_prop_register(ds, | |
510 | zfs_prop_to_name(ZFS_PROP_LOGBIAS), | |
428870ff | 511 | logbias_changed_cb, os); |
13fe0198 MA |
512 | } |
513 | if (err == 0) { | |
514 | err = dsl_prop_register(ds, | |
515 | zfs_prop_to_name(ZFS_PROP_SYNC), | |
428870ff | 516 | sync_changed_cb, os); |
13fe0198 | 517 | } |
faf0f58c MA |
518 | if (err == 0) { |
519 | err = dsl_prop_register(ds, | |
520 | zfs_prop_to_name( | |
521 | ZFS_PROP_REDUNDANT_METADATA), | |
522 | redundant_metadata_changed_cb, os); | |
523 | } | |
f1512ee6 MA |
524 | if (err == 0) { |
525 | err = dsl_prop_register(ds, | |
526 | zfs_prop_to_name(ZFS_PROP_RECORDSIZE), | |
527 | recordsize_changed_cb, os); | |
528 | } | |
50c957f7 NB |
529 | if (err == 0) { |
530 | err = dsl_prop_register(ds, | |
531 | zfs_prop_to_name(ZFS_PROP_DNODESIZE), | |
532 | dnodesize_changed_cb, os); | |
533 | } | |
b128c09f | 534 | } |
47dfff3b MA |
535 | if (needlock) |
536 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
13fe0198 | 537 | if (err != 0) { |
d3c2ae1c | 538 | arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); |
428870ff | 539 | kmem_free(os, sizeof (objset_t)); |
34dc7c2f BB |
540 | return (err); |
541 | } | |
9b67f605 | 542 | } else { |
34dc7c2f | 543 | /* It's the meta-objset. */ |
428870ff | 544 | os->os_checksum = ZIO_CHECKSUM_FLETCHER_4; |
99197f03 | 545 | os->os_compress = ZIO_COMPRESS_ON; |
b5256303 | 546 | os->os_encrypted = B_FALSE; |
428870ff BB |
547 | os->os_copies = spa_max_replication(spa); |
548 | os->os_dedup_checksum = ZIO_CHECKSUM_OFF; | |
faf0f58c MA |
549 | os->os_dedup_verify = B_FALSE; |
550 | os->os_logbias = ZFS_LOGBIAS_LATENCY; | |
551 | os->os_sync = ZFS_SYNC_STANDARD; | |
428870ff BB |
552 | os->os_primary_cache = ZFS_CACHE_ALL; |
553 | os->os_secondary_cache = ZFS_CACHE_ALL; | |
50c957f7 | 554 | os->os_dnodesize = DNODE_MIN_SIZE; |
34dc7c2f BB |
555 | } |
556 | ||
0c66c32d | 557 | if (ds == NULL || !ds->ds_is_snapshot) |
572e2857 | 558 | os->os_zil_header = os->os_phys->os_zil_header; |
428870ff | 559 | os->os_zil = zil_alloc(os, &os->os_zil_header); |
34dc7c2f BB |
560 | |
561 | for (i = 0; i < TXG_SIZE; i++) { | |
64fc7762 MA |
562 | os->os_dirty_dnodes[i] = multilist_create(sizeof (dnode_t), |
563 | offsetof(dnode_t, dn_dirty_link[i]), | |
564 | dnode_multilist_index_func); | |
34dc7c2f | 565 | } |
428870ff | 566 | list_create(&os->os_dnodes, sizeof (dnode_t), |
34dc7c2f | 567 | offsetof(dnode_t, dn_link)); |
428870ff | 568 | list_create(&os->os_downgraded_dbufs, sizeof (dmu_buf_impl_t), |
34dc7c2f BB |
569 | offsetof(dmu_buf_impl_t, db_link)); |
570 | ||
0c66c32d JG |
571 | list_link_init(&os->os_evicting_node); |
572 | ||
428870ff | 573 | mutex_init(&os->os_lock, NULL, MUTEX_DEFAULT, NULL); |
64fc7762 | 574 | mutex_init(&os->os_userused_lock, NULL, MUTEX_DEFAULT, NULL); |
428870ff BB |
575 | mutex_init(&os->os_obj_lock, NULL, MUTEX_DEFAULT, NULL); |
576 | mutex_init(&os->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL); | |
dbeb8796 MA |
577 | os->os_obj_next_percpu_len = boot_ncpus; |
578 | os->os_obj_next_percpu = kmem_zalloc(os->os_obj_next_percpu_len * | |
579 | sizeof (os->os_obj_next_percpu[0]), KM_SLEEP); | |
428870ff | 580 | |
0c66c32d JG |
581 | dnode_special_open(os, &os->os_phys->os_meta_dnode, |
582 | DMU_META_DNODE_OBJECT, &os->os_meta_dnode); | |
9c5167d1 | 583 | if (OBJSET_BUF_HAS_USERUSED(os->os_phys_buf)) { |
0c66c32d JG |
584 | dnode_special_open(os, &os->os_phys->os_userused_dnode, |
585 | DMU_USERUSED_OBJECT, &os->os_userused_dnode); | |
586 | dnode_special_open(os, &os->os_phys->os_groupused_dnode, | |
587 | DMU_GROUPUSED_OBJECT, &os->os_groupused_dnode); | |
9c5167d1 NF |
588 | if (OBJSET_BUF_HAS_PROJECTUSED(os->os_phys_buf)) |
589 | dnode_special_open(os, | |
590 | &os->os_phys->os_projectused_dnode, | |
591 | DMU_PROJECTUSED_OBJECT, &os->os_projectused_dnode); | |
9babb374 | 592 | } |
34dc7c2f | 593 | |
1de321e6 JX |
594 | mutex_init(&os->os_upgrade_lock, NULL, MUTEX_DEFAULT, NULL); |
595 | ||
428870ff | 596 | *osp = os; |
34dc7c2f BB |
597 | return (0); |
598 | } | |
599 | ||
428870ff BB |
600 | int |
601 | dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp) | |
34dc7c2f | 602 | { |
428870ff | 603 | int err = 0; |
34dc7c2f | 604 | |
47dfff3b MA |
605 | /* |
606 | * We shouldn't be doing anything with dsl_dataset_t's unless the | |
607 | * pool_config lock is held, or the dataset is long-held. | |
608 | */ | |
609 | ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool) || | |
610 | dsl_dataset_long_held(ds)); | |
611 | ||
34dc7c2f | 612 | mutex_enter(&ds->ds_opening_lock); |
9b67f605 MA |
613 | if (ds->ds_objset == NULL) { |
614 | objset_t *os; | |
cc9bb3e5 | 615 | rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); |
34dc7c2f | 616 | err = dmu_objset_open_impl(dsl_dataset_get_spa(ds), |
9b67f605 | 617 | ds, dsl_dataset_get_blkptr(ds), &os); |
cc9bb3e5 | 618 | rrw_exit(&ds->ds_bp_rwlock, FTAG); |
9b67f605 MA |
619 | |
620 | if (err == 0) { | |
621 | mutex_enter(&ds->ds_lock); | |
622 | ASSERT(ds->ds_objset == NULL); | |
623 | ds->ds_objset = os; | |
624 | mutex_exit(&ds->ds_lock); | |
625 | } | |
34dc7c2f | 626 | } |
9b67f605 | 627 | *osp = ds->ds_objset; |
34dc7c2f | 628 | mutex_exit(&ds->ds_opening_lock); |
428870ff | 629 | return (err); |
34dc7c2f BB |
630 | } |
631 | ||
13fe0198 MA |
632 | /* |
633 | * Holds the pool while the objset is held. Therefore only one objset | |
634 | * can be held at a time. | |
635 | */ | |
34dc7c2f | 636 | int |
b5256303 TC |
637 | dmu_objset_hold_flags(const char *name, boolean_t decrypt, void *tag, |
638 | objset_t **osp) | |
34dc7c2f | 639 | { |
13fe0198 | 640 | dsl_pool_t *dp; |
428870ff | 641 | dsl_dataset_t *ds; |
34dc7c2f | 642 | int err; |
b5256303 | 643 | ds_hold_flags_t flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0; |
34dc7c2f | 644 | |
13fe0198 MA |
645 | err = dsl_pool_hold(name, tag, &dp); |
646 | if (err != 0) | |
647 | return (err); | |
b5256303 | 648 | err = dsl_dataset_hold_flags(dp, name, flags, tag, &ds); |
13fe0198 MA |
649 | if (err != 0) { |
650 | dsl_pool_rele(dp, tag); | |
428870ff | 651 | return (err); |
13fe0198 | 652 | } |
428870ff BB |
653 | |
654 | err = dmu_objset_from_ds(ds, osp); | |
13fe0198 | 655 | if (err != 0) { |
428870ff | 656 | dsl_dataset_rele(ds, tag); |
13fe0198 MA |
657 | dsl_pool_rele(dp, tag); |
658 | } | |
428870ff | 659 | |
34dc7c2f BB |
660 | return (err); |
661 | } | |
662 | ||
b5256303 TC |
663 | int |
664 | dmu_objset_hold(const char *name, void *tag, objset_t **osp) | |
665 | { | |
666 | return (dmu_objset_hold_flags(name, B_FALSE, tag, osp)); | |
667 | } | |
668 | ||
9c43027b AJ |
669 | static int |
670 | dmu_objset_own_impl(dsl_dataset_t *ds, dmu_objset_type_t type, | |
b5256303 | 671 | boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp) |
9c43027b AJ |
672 | { |
673 | int err; | |
674 | ||
675 | err = dmu_objset_from_ds(ds, osp); | |
676 | if (err != 0) { | |
b5256303 | 677 | return (err); |
9c43027b | 678 | } else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) { |
9c43027b AJ |
679 | return (SET_ERROR(EINVAL)); |
680 | } else if (!readonly && dsl_dataset_is_snapshot(ds)) { | |
9c43027b | 681 | return (SET_ERROR(EROFS)); |
ae76f45c TC |
682 | } else if (!readonly && decrypt && |
683 | dsl_dir_incompatible_encryption_version(ds->ds_dir)) { | |
684 | return (SET_ERROR(EROFS)); | |
9c43027b | 685 | } |
b5256303 TC |
686 | |
687 | /* if we are decrypting, we can now check MACs in os->os_phys_buf */ | |
688 | if (decrypt && arc_is_unauthenticated((*osp)->os_phys_buf)) { | |
a2c2ed1b TC |
689 | zbookmark_phys_t zb; |
690 | ||
691 | SET_BOOKMARK(&zb, ds->ds_object, ZB_ROOT_OBJECT, | |
692 | ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
b5256303 | 693 | err = arc_untransform((*osp)->os_phys_buf, (*osp)->os_spa, |
a2c2ed1b | 694 | &zb, B_FALSE); |
b5256303 TC |
695 | if (err != 0) |
696 | return (err); | |
697 | ||
698 | ASSERT0(arc_is_unauthenticated((*osp)->os_phys_buf)); | |
699 | } | |
700 | ||
701 | return (0); | |
9c43027b AJ |
702 | } |
703 | ||
13fe0198 MA |
704 | /* |
705 | * dsl_pool must not be held when this is called. | |
706 | * Upon successful return, there will be a longhold on the dataset, | |
707 | * and the dsl_pool will not be held. | |
708 | */ | |
34dc7c2f | 709 | int |
428870ff | 710 | dmu_objset_own(const char *name, dmu_objset_type_t type, |
b5256303 | 711 | boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp) |
34dc7c2f | 712 | { |
13fe0198 | 713 | dsl_pool_t *dp; |
34dc7c2f BB |
714 | dsl_dataset_t *ds; |
715 | int err; | |
b5256303 | 716 | ds_hold_flags_t flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0; |
34dc7c2f | 717 | |
13fe0198 MA |
718 | err = dsl_pool_hold(name, FTAG, &dp); |
719 | if (err != 0) | |
720 | return (err); | |
b5256303 | 721 | err = dsl_dataset_own(dp, name, flags, tag, &ds); |
13fe0198 MA |
722 | if (err != 0) { |
723 | dsl_pool_rele(dp, FTAG); | |
34dc7c2f | 724 | return (err); |
13fe0198 | 725 | } |
b5256303 TC |
726 | err = dmu_objset_own_impl(ds, type, readonly, decrypt, tag, osp); |
727 | if (err != 0) { | |
728 | dsl_dataset_disown(ds, flags, tag); | |
729 | dsl_pool_rele(dp, FTAG); | |
730 | return (err); | |
731 | } | |
732 | ||
163a8c28 TC |
733 | /* |
734 | * User accounting requires the dataset to be decrypted and rw. | |
735 | * We also don't begin user accounting during claiming to help | |
736 | * speed up pool import times and to keep this txg reserved | |
737 | * completely for recovery work. | |
738 | */ | |
9c5167d1 NF |
739 | if ((dmu_objset_userobjspace_upgradable(*osp) || |
740 | dmu_objset_projectquota_upgradable(*osp)) && | |
163a8c28 | 741 | !readonly && !dp->dp_spa->spa_claiming && |
4807c0ba | 742 | (ds->ds_dir->dd_crypto_obj == 0 || decrypt)) |
9c5167d1 | 743 | dmu_objset_id_quota_upgrade(*osp); |
1de321e6 | 744 | |
c0daec32 | 745 | dsl_pool_rele(dp, FTAG); |
b5256303 | 746 | return (0); |
34dc7c2f BB |
747 | } |
748 | ||
9c43027b AJ |
749 | int |
750 | dmu_objset_own_obj(dsl_pool_t *dp, uint64_t obj, dmu_objset_type_t type, | |
b5256303 | 751 | boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp) |
9c43027b AJ |
752 | { |
753 | dsl_dataset_t *ds; | |
754 | int err; | |
b5256303 | 755 | ds_hold_flags_t flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0; |
9c43027b | 756 | |
b5256303 | 757 | err = dsl_dataset_own_obj(dp, obj, flags, tag, &ds); |
9c43027b AJ |
758 | if (err != 0) |
759 | return (err); | |
760 | ||
b5256303 TC |
761 | err = dmu_objset_own_impl(ds, type, readonly, decrypt, tag, osp); |
762 | if (err != 0) { | |
763 | dsl_dataset_disown(ds, flags, tag); | |
764 | return (err); | |
765 | } | |
766 | ||
767 | return (0); | |
9c43027b AJ |
768 | } |
769 | ||
34dc7c2f | 770 | void |
b5256303 | 771 | dmu_objset_rele_flags(objset_t *os, boolean_t decrypt, void *tag) |
34dc7c2f | 772 | { |
b5256303 TC |
773 | ds_hold_flags_t flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0; |
774 | ||
13fe0198 | 775 | dsl_pool_t *dp = dmu_objset_pool(os); |
b5256303 | 776 | dsl_dataset_rele_flags(os->os_dsl_dataset, flags, tag); |
13fe0198 | 777 | dsl_pool_rele(dp, tag); |
428870ff | 778 | } |
b128c09f | 779 | |
b5256303 TC |
780 | void |
781 | dmu_objset_rele(objset_t *os, void *tag) | |
782 | { | |
783 | dmu_objset_rele_flags(os, B_FALSE, tag); | |
784 | } | |
785 | ||
831baf06 KW |
786 | /* |
787 | * When we are called, os MUST refer to an objset associated with a dataset | |
788 | * that is owned by 'tag'; that is, is held and long held by 'tag' and ds_owner | |
789 | * == tag. We will then release and reacquire ownership of the dataset while | |
790 | * holding the pool config_rwlock to avoid intervening namespace or ownership | |
791 | * changes may occur. | |
792 | * | |
793 | * This exists solely to accommodate zfs_ioc_userspace_upgrade()'s desire to | |
794 | * release the hold on its dataset and acquire a new one on the dataset of the | |
795 | * same name so that it can be partially torn down and reconstructed. | |
796 | */ | |
797 | void | |
5e00213e AG |
798 | dmu_objset_refresh_ownership(dsl_dataset_t *ds, dsl_dataset_t **newds, |
799 | boolean_t decrypt, void *tag) | |
831baf06 KW |
800 | { |
801 | dsl_pool_t *dp; | |
eca7b760 | 802 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
831baf06 | 803 | |
831baf06 KW |
804 | VERIFY3P(ds, !=, NULL); |
805 | VERIFY3P(ds->ds_owner, ==, tag); | |
806 | VERIFY(dsl_dataset_long_held(ds)); | |
807 | ||
808 | dsl_dataset_name(ds, name); | |
5e00213e | 809 | dp = ds->ds_dir->dd_pool; |
831baf06 | 810 | dsl_pool_config_enter(dp, FTAG); |
5e00213e | 811 | dsl_dataset_disown(ds, decrypt, tag); |
b5256303 | 812 | VERIFY0(dsl_dataset_own(dp, name, |
5e00213e | 813 | (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0, tag, newds)); |
831baf06 KW |
814 | dsl_pool_config_exit(dp, FTAG); |
815 | } | |
816 | ||
428870ff | 817 | void |
b5256303 | 818 | dmu_objset_disown(objset_t *os, boolean_t decrypt, void *tag) |
428870ff | 819 | { |
1de321e6 JX |
820 | /* |
821 | * Stop upgrading thread | |
822 | */ | |
823 | dmu_objset_upgrade_stop(os); | |
b5256303 TC |
824 | dsl_dataset_disown(os->os_dsl_dataset, |
825 | (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0, tag); | |
34dc7c2f BB |
826 | } |
827 | ||
13fe0198 | 828 | void |
34dc7c2f BB |
829 | dmu_objset_evict_dbufs(objset_t *os) |
830 | { | |
0c66c32d | 831 | dnode_t *dn_marker; |
34dc7c2f BB |
832 | dnode_t *dn; |
833 | ||
0c66c32d | 834 | dn_marker = kmem_alloc(sizeof (dnode_t), KM_SLEEP); |
34dc7c2f | 835 | |
0c66c32d JG |
836 | mutex_enter(&os->os_lock); |
837 | dn = list_head(&os->os_dnodes); | |
838 | while (dn != NULL) { | |
839 | /* | |
840 | * Skip dnodes without holds. We have to do this dance | |
841 | * because dnode_add_ref() only works if there is already a | |
842 | * hold. If the dnode has no holds, then it has no dbufs. | |
843 | */ | |
844 | if (dnode_add_ref(dn, FTAG)) { | |
845 | list_insert_after(&os->os_dnodes, dn, dn_marker); | |
846 | mutex_exit(&os->os_lock); | |
34dc7c2f | 847 | |
0c66c32d JG |
848 | dnode_evict_dbufs(dn); |
849 | dnode_rele(dn, FTAG); | |
34dc7c2f | 850 | |
0c66c32d JG |
851 | mutex_enter(&os->os_lock); |
852 | dn = list_next(&os->os_dnodes, dn_marker); | |
853 | list_remove(&os->os_dnodes, dn_marker); | |
854 | } else { | |
855 | dn = list_next(&os->os_dnodes, dn); | |
856 | } | |
857 | } | |
858 | mutex_exit(&os->os_lock); | |
34dc7c2f | 859 | |
0c66c32d | 860 | kmem_free(dn_marker, sizeof (dnode_t)); |
34dc7c2f | 861 | |
0c66c32d | 862 | if (DMU_USERUSED_DNODE(os) != NULL) { |
9c5167d1 NF |
863 | if (DMU_PROJECTUSED_DNODE(os) != NULL) |
864 | dnode_evict_dbufs(DMU_PROJECTUSED_DNODE(os)); | |
0c66c32d JG |
865 | dnode_evict_dbufs(DMU_GROUPUSED_DNODE(os)); |
866 | dnode_evict_dbufs(DMU_USERUSED_DNODE(os)); | |
34dc7c2f | 867 | } |
0c66c32d | 868 | dnode_evict_dbufs(DMU_META_DNODE(os)); |
34dc7c2f BB |
869 | } |
870 | ||
0c66c32d JG |
871 | /* |
872 | * Objset eviction processing is split into into two pieces. | |
873 | * The first marks the objset as evicting, evicts any dbufs that | |
874 | * have a refcount of zero, and then queues up the objset for the | |
875 | * second phase of eviction. Once os->os_dnodes has been cleared by | |
876 | * dnode_buf_pageout()->dnode_destroy(), the second phase is executed. | |
877 | * The second phase closes the special dnodes, dequeues the objset from | |
878 | * the list of those undergoing eviction, and finally frees the objset. | |
879 | * | |
880 | * NOTE: Due to asynchronous eviction processing (invocation of | |
881 | * dnode_buf_pageout()), it is possible for the meta dnode for the | |
882 | * objset to have no holds even though os->os_dnodes is not empty. | |
883 | */ | |
34dc7c2f | 884 | void |
428870ff | 885 | dmu_objset_evict(objset_t *os) |
34dc7c2f | 886 | { |
6f1ffb06 MA |
887 | dsl_dataset_t *ds = os->os_dsl_dataset; |
888 | ||
1c27024e | 889 | for (int t = 0; t < TXG_SIZE; t++) |
428870ff | 890 | ASSERT(!dmu_objset_is_dirty(os, t)); |
34dc7c2f | 891 | |
0eb21616 JG |
892 | if (ds) |
893 | dsl_prop_unregister_all(ds, os); | |
34dc7c2f | 894 | |
428870ff BB |
895 | if (os->os_sa) |
896 | sa_tear_down(os); | |
897 | ||
13fe0198 | 898 | dmu_objset_evict_dbufs(os); |
34dc7c2f | 899 | |
0c66c32d JG |
900 | mutex_enter(&os->os_lock); |
901 | spa_evicting_os_register(os->os_spa, os); | |
902 | if (list_is_empty(&os->os_dnodes)) { | |
903 | mutex_exit(&os->os_lock); | |
904 | dmu_objset_evict_done(os); | |
905 | } else { | |
906 | mutex_exit(&os->os_lock); | |
907 | } | |
b5256303 TC |
908 | |
909 | ||
0c66c32d JG |
910 | } |
911 | ||
912 | void | |
913 | dmu_objset_evict_done(objset_t *os) | |
914 | { | |
915 | ASSERT3P(list_head(&os->os_dnodes), ==, NULL); | |
916 | ||
572e2857 BB |
917 | dnode_special_close(&os->os_meta_dnode); |
918 | if (DMU_USERUSED_DNODE(os)) { | |
9c5167d1 NF |
919 | if (DMU_PROJECTUSED_DNODE(os)) |
920 | dnode_special_close(&os->os_projectused_dnode); | |
572e2857 BB |
921 | dnode_special_close(&os->os_userused_dnode); |
922 | dnode_special_close(&os->os_groupused_dnode); | |
9babb374 | 923 | } |
428870ff BB |
924 | zil_free(os->os_zil); |
925 | ||
d3c2ae1c | 926 | arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); |
572e2857 BB |
927 | |
928 | /* | |
929 | * This is a barrier to prevent the objset from going away in | |
930 | * dnode_move() until we can safely ensure that the objset is still in | |
931 | * use. We consider the objset valid before the barrier and invalid | |
932 | * after the barrier. | |
933 | */ | |
934 | rw_enter(&os_lock, RW_READER); | |
935 | rw_exit(&os_lock); | |
936 | ||
dbeb8796 MA |
937 | kmem_free(os->os_obj_next_percpu, |
938 | os->os_obj_next_percpu_len * sizeof (os->os_obj_next_percpu[0])); | |
939 | ||
428870ff | 940 | mutex_destroy(&os->os_lock); |
64fc7762 | 941 | mutex_destroy(&os->os_userused_lock); |
428870ff BB |
942 | mutex_destroy(&os->os_obj_lock); |
943 | mutex_destroy(&os->os_user_ptr_lock); | |
c17486b2 | 944 | mutex_destroy(&os->os_upgrade_lock); |
64fc7762 MA |
945 | for (int i = 0; i < TXG_SIZE; i++) { |
946 | multilist_destroy(os->os_dirty_dnodes[i]); | |
947 | } | |
0c66c32d | 948 | spa_evicting_os_deregister(os->os_spa, os); |
428870ff BB |
949 | kmem_free(os, sizeof (objset_t)); |
950 | } | |
9babb374 | 951 | |
428870ff BB |
952 | timestruc_t |
953 | dmu_objset_snap_cmtime(objset_t *os) | |
954 | { | |
955 | return (dsl_dir_snap_cmtime(os->os_dsl_dataset->ds_dir)); | |
34dc7c2f BB |
956 | } |
957 | ||
428870ff | 958 | objset_t * |
b5256303 TC |
959 | dmu_objset_create_impl_dnstats(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, |
960 | dmu_objset_type_t type, int levels, int blksz, int ibs, dmu_tx_t *tx) | |
34dc7c2f | 961 | { |
428870ff | 962 | objset_t *os; |
34dc7c2f BB |
963 | dnode_t *mdn; |
964 | ||
965 | ASSERT(dmu_tx_is_syncing(tx)); | |
13fe0198 | 966 | |
b5256303 TC |
967 | if (blksz == 0) |
968 | blksz = DNODE_BLOCK_SIZE; | |
4807c0ba | 969 | if (ibs == 0) |
b5256303 TC |
970 | ibs = DN_MAX_INDBLKSHIFT; |
971 | ||
572e2857 | 972 | if (ds != NULL) |
13fe0198 | 973 | VERIFY0(dmu_objset_from_ds(ds, &os)); |
572e2857 | 974 | else |
13fe0198 | 975 | VERIFY0(dmu_objset_open_impl(spa, NULL, bp, &os)); |
572e2857 BB |
976 | |
977 | mdn = DMU_META_DNODE(os); | |
34dc7c2f | 978 | |
b5256303 TC |
979 | dnode_allocate(mdn, DMU_OT_DNODE, blksz, ibs, DMU_OT_NONE, 0, |
980 | DNODE_MIN_SLOTS, tx); | |
34dc7c2f BB |
981 | |
982 | /* | |
983 | * We don't want to have to increase the meta-dnode's nlevels | |
984 | * later, because then we could do it in quescing context while | |
985 | * we are also accessing it in open context. | |
986 | * | |
987 | * This precaution is not necessary for the MOS (ds == NULL), | |
988 | * because the MOS is only updated in syncing context. | |
989 | * This is most fortunate: the MOS is the only objset that | |
990 | * needs to be synced multiple times as spa_sync() iterates | |
991 | * to convergence, so minimizing its dn_nlevels matters. | |
992 | */ | |
993 | if (ds != NULL) { | |
b5256303 TC |
994 | if (levels == 0) { |
995 | levels = 1; | |
996 | ||
997 | /* | |
998 | * Determine the number of levels necessary for the | |
999 | * meta-dnode to contain DN_MAX_OBJECT dnodes. Note | |
1000 | * that in order to ensure that we do not overflow | |
1001 | * 64 bits, there has to be a nlevels that gives us a | |
1002 | * number of blocks > DN_MAX_OBJECT but < 2^64. | |
1003 | * Therefore, (mdn->dn_indblkshift - SPA_BLKPTRSHIFT) | |
1004 | * (10) must be less than (64 - log2(DN_MAX_OBJECT)) | |
1005 | * (16). | |
1006 | */ | |
1007 | while ((uint64_t)mdn->dn_nblkptr << | |
1008 | (mdn->dn_datablkshift - DNODE_SHIFT + (levels - 1) * | |
1009 | (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) < | |
1010 | DN_MAX_OBJECT) | |
1011 | levels++; | |
1012 | } | |
34dc7c2f BB |
1013 | |
1014 | mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] = | |
1015 | mdn->dn_nlevels = levels; | |
1016 | } | |
1017 | ||
1018 | ASSERT(type != DMU_OST_NONE); | |
1019 | ASSERT(type != DMU_OST_ANY); | |
1020 | ASSERT(type < DMU_OST_NUMTYPES); | |
428870ff | 1021 | os->os_phys->os_type = type; |
b5256303 TC |
1022 | |
1023 | /* | |
1024 | * Enable user accounting if it is enabled and this is not an | |
1025 | * encrypted receive. | |
1026 | */ | |
1027 | if (dmu_objset_userused_enabled(os) && | |
1028 | (!os->os_encrypted || !dmu_objset_is_receiving(os))) { | |
428870ff | 1029 | os->os_phys->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; |
1de321e6 JX |
1030 | if (dmu_objset_userobjused_enabled(os)) { |
1031 | ds->ds_feature_activation_needed[ | |
1032 | SPA_FEATURE_USEROBJ_ACCOUNTING] = B_TRUE; | |
1033 | os->os_phys->os_flags |= | |
1034 | OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE; | |
1035 | } | |
9c5167d1 NF |
1036 | if (dmu_objset_projectquota_enabled(os)) { |
1037 | ds->ds_feature_activation_needed[ | |
1038 | SPA_FEATURE_PROJECT_QUOTA] = B_TRUE; | |
1039 | os->os_phys->os_flags |= | |
1040 | OBJSET_FLAG_PROJECTQUOTA_COMPLETE; | |
1041 | } | |
428870ff | 1042 | os->os_flags = os->os_phys->os_flags; |
9babb374 | 1043 | } |
34dc7c2f BB |
1044 | |
1045 | dsl_dataset_dirty(ds, tx); | |
1046 | ||
428870ff | 1047 | return (os); |
34dc7c2f BB |
1048 | } |
1049 | ||
b5256303 TC |
1050 | /* called from dsl for meta-objset */ |
1051 | objset_t * | |
1052 | dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, | |
1053 | dmu_objset_type_t type, dmu_tx_t *tx) | |
1054 | { | |
1055 | return (dmu_objset_create_impl_dnstats(spa, ds, bp, type, 0, 0, 0, tx)); | |
1056 | } | |
1057 | ||
13fe0198 MA |
1058 | typedef struct dmu_objset_create_arg { |
1059 | const char *doca_name; | |
1060 | cred_t *doca_cred; | |
1061 | void (*doca_userfunc)(objset_t *os, void *arg, | |
1062 | cred_t *cr, dmu_tx_t *tx); | |
1063 | void *doca_userarg; | |
1064 | dmu_objset_type_t doca_type; | |
1065 | uint64_t doca_flags; | |
b5256303 | 1066 | dsl_crypto_params_t *doca_dcp; |
13fe0198 | 1067 | } dmu_objset_create_arg_t; |
34dc7c2f BB |
1068 | |
1069 | /*ARGSUSED*/ | |
1070 | static int | |
13fe0198 | 1071 | dmu_objset_create_check(void *arg, dmu_tx_t *tx) |
34dc7c2f | 1072 | { |
13fe0198 MA |
1073 | dmu_objset_create_arg_t *doca = arg; |
1074 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
1075 | dsl_dir_t *pdd; | |
1076 | const char *tail; | |
1077 | int error; | |
34dc7c2f | 1078 | |
13fe0198 | 1079 | if (strchr(doca->doca_name, '@') != NULL) |
2e528b49 | 1080 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1081 | |
eca7b760 IK |
1082 | if (strlen(doca->doca_name) >= ZFS_MAX_DATASET_NAME_LEN) |
1083 | return (SET_ERROR(ENAMETOOLONG)); | |
1084 | ||
13fe0198 MA |
1085 | error = dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail); |
1086 | if (error != 0) | |
1087 | return (error); | |
1088 | if (tail == NULL) { | |
1089 | dsl_dir_rele(pdd, FTAG); | |
2e528b49 | 1090 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 1091 | } |
b5256303 TC |
1092 | |
1093 | error = dmu_objset_create_crypt_check(pdd, doca->doca_dcp); | |
1094 | if (error != 0) { | |
1095 | dsl_dir_rele(pdd, FTAG); | |
1096 | return (error); | |
1097 | } | |
1098 | ||
788eb90c JJ |
1099 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
1100 | doca->doca_cred); | |
b5256303 | 1101 | |
13fe0198 | 1102 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f | 1103 | |
788eb90c | 1104 | return (error); |
34dc7c2f BB |
1105 | } |
1106 | ||
1107 | static void | |
13fe0198 | 1108 | dmu_objset_create_sync(void *arg, dmu_tx_t *tx) |
34dc7c2f | 1109 | { |
13fe0198 MA |
1110 | dmu_objset_create_arg_t *doca = arg; |
1111 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
1112 | dsl_dir_t *pdd; | |
1113 | const char *tail; | |
6f1ffb06 | 1114 | dsl_dataset_t *ds; |
13fe0198 | 1115 | uint64_t obj; |
6f1ffb06 | 1116 | blkptr_t *bp; |
13fe0198 | 1117 | objset_t *os; |
b5256303 | 1118 | zio_t *rzio; |
34dc7c2f | 1119 | |
13fe0198 | 1120 | VERIFY0(dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail)); |
34dc7c2f | 1121 | |
13fe0198 | 1122 | obj = dsl_dataset_create_sync(pdd, tail, NULL, doca->doca_flags, |
b5256303 | 1123 | doca->doca_cred, doca->doca_dcp, tx); |
34dc7c2f | 1124 | |
b5256303 TC |
1125 | VERIFY0(dsl_dataset_hold_obj_flags(pdd->dd_pool, obj, |
1126 | DS_HOLD_FLAG_DECRYPT, FTAG, &ds)); | |
cc9bb3e5 | 1127 | rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); |
6f1ffb06 | 1128 | bp = dsl_dataset_get_blkptr(ds); |
13fe0198 MA |
1129 | os = dmu_objset_create_impl(pdd->dd_pool->dp_spa, |
1130 | ds, bp, doca->doca_type, tx); | |
cc9bb3e5 | 1131 | rrw_exit(&ds->ds_bp_rwlock, FTAG); |
34dc7c2f | 1132 | |
13fe0198 MA |
1133 | if (doca->doca_userfunc != NULL) { |
1134 | doca->doca_userfunc(os, doca->doca_userarg, | |
1135 | doca->doca_cred, tx); | |
34dc7c2f BB |
1136 | } |
1137 | ||
b5256303 | 1138 | /* |
4807c0ba | 1139 | * The doca_userfunc() may write out some data that needs to be |
b5256303 TC |
1140 | * encrypted if the dataset is encrypted (specifically the root |
1141 | * directory). This data must be written out before the encryption | |
1142 | * key mapping is removed by dsl_dataset_rele_flags(). Force the | |
1143 | * I/O to occur immediately by invoking the relevant sections of | |
1144 | * dsl_pool_sync(). | |
1145 | */ | |
1146 | if (os->os_encrypted) { | |
1147 | dsl_dataset_t *tmpds = NULL; | |
1148 | boolean_t need_sync_done = B_FALSE; | |
1149 | ||
4807c0ba TC |
1150 | mutex_enter(&ds->ds_lock); |
1151 | ds->ds_owner = FTAG; | |
1152 | mutex_exit(&ds->ds_lock); | |
1153 | ||
b5256303 | 1154 | rzio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED); |
4807c0ba TC |
1155 | tmpds = txg_list_remove_this(&dp->dp_dirty_datasets, ds, |
1156 | tx->tx_txg); | |
b5256303 | 1157 | if (tmpds != NULL) { |
b5256303 TC |
1158 | dsl_dataset_sync(ds, rzio, tx); |
1159 | need_sync_done = B_TRUE; | |
1160 | } | |
1161 | VERIFY0(zio_wait(rzio)); | |
1162 | ||
1163 | dmu_objset_do_userquota_updates(os, tx); | |
1164 | taskq_wait(dp->dp_sync_taskq); | |
1165 | ||
1166 | rzio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED); | |
4807c0ba TC |
1167 | tmpds = txg_list_remove_this(&dp->dp_dirty_datasets, ds, |
1168 | tx->tx_txg); | |
b5256303 | 1169 | if (tmpds != NULL) { |
b5256303 TC |
1170 | dmu_buf_rele(ds->ds_dbuf, ds); |
1171 | dsl_dataset_sync(ds, rzio, tx); | |
1172 | } | |
1173 | VERIFY0(zio_wait(rzio)); | |
1174 | ||
1175 | if (need_sync_done) | |
1176 | dsl_dataset_sync_done(ds, tx); | |
4807c0ba TC |
1177 | |
1178 | mutex_enter(&ds->ds_lock); | |
1179 | ds->ds_owner = NULL; | |
1180 | mutex_exit(&ds->ds_lock); | |
b5256303 TC |
1181 | } |
1182 | ||
13fe0198 | 1183 | spa_history_log_internal_ds(ds, "create", tx, ""); |
a0bd735a BP |
1184 | zvol_create_minors(dp->dp_spa, doca->doca_name, B_TRUE); |
1185 | ||
b5256303 | 1186 | dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT, FTAG); |
13fe0198 | 1187 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f BB |
1188 | } |
1189 | ||
1190 | int | |
428870ff | 1191 | dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, |
b5256303 | 1192 | dsl_crypto_params_t *dcp, dmu_objset_create_sync_func_t func, void *arg) |
34dc7c2f | 1193 | { |
13fe0198 | 1194 | dmu_objset_create_arg_t doca; |
b5256303 | 1195 | dsl_crypto_params_t tmp_dcp = { 0 }; |
34dc7c2f | 1196 | |
13fe0198 MA |
1197 | doca.doca_name = name; |
1198 | doca.doca_cred = CRED(); | |
1199 | doca.doca_flags = flags; | |
1200 | doca.doca_userfunc = func; | |
1201 | doca.doca_userarg = arg; | |
1202 | doca.doca_type = type; | |
34dc7c2f | 1203 | |
b5256303 TC |
1204 | /* |
1205 | * Some callers (mostly for testing) do not provide a dcp on their | |
1206 | * own but various code inside the sync task will require it to be | |
1207 | * allocated. Rather than adding NULL checks throughout this code | |
1208 | * or adding dummy dcp's to all of the callers we simply create a | |
1209 | * dummy one here and use that. This zero dcp will have the same | |
1210 | * effect as asking for inheritence of all encryption params. | |
1211 | */ | |
1212 | doca.doca_dcp = (dcp != NULL) ? dcp : &tmp_dcp; | |
1213 | ||
13fe0198 | 1214 | return (dsl_sync_task(name, |
3d45fdd6 | 1215 | dmu_objset_create_check, dmu_objset_create_sync, &doca, |
b5256303 | 1216 | 6, ZFS_SPACE_CHECK_NORMAL)); |
34dc7c2f BB |
1217 | } |
1218 | ||
13fe0198 MA |
1219 | typedef struct dmu_objset_clone_arg { |
1220 | const char *doca_clone; | |
1221 | const char *doca_origin; | |
1222 | cred_t *doca_cred; | |
1223 | } dmu_objset_clone_arg_t; | |
1224 | ||
1225 | /*ARGSUSED*/ | |
1226 | static int | |
1227 | dmu_objset_clone_check(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 1228 | { |
13fe0198 | 1229 | dmu_objset_clone_arg_t *doca = arg; |
428870ff BB |
1230 | dsl_dir_t *pdd; |
1231 | const char *tail; | |
13fe0198 MA |
1232 | int error; |
1233 | dsl_dataset_t *origin; | |
1234 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
34dc7c2f | 1235 | |
13fe0198 | 1236 | if (strchr(doca->doca_clone, '@') != NULL) |
2e528b49 | 1237 | return (SET_ERROR(EINVAL)); |
13fe0198 | 1238 | |
eca7b760 IK |
1239 | if (strlen(doca->doca_clone) >= ZFS_MAX_DATASET_NAME_LEN) |
1240 | return (SET_ERROR(ENAMETOOLONG)); | |
1241 | ||
13fe0198 MA |
1242 | error = dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail); |
1243 | if (error != 0) | |
1244 | return (error); | |
428870ff | 1245 | if (tail == NULL) { |
13fe0198 | 1246 | dsl_dir_rele(pdd, FTAG); |
2e528b49 | 1247 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 1248 | } |
1cddb8c9 | 1249 | |
788eb90c JJ |
1250 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
1251 | doca->doca_cred); | |
1252 | if (error != 0) { | |
1253 | dsl_dir_rele(pdd, FTAG); | |
1254 | return (SET_ERROR(EDQUOT)); | |
1255 | } | |
428870ff | 1256 | |
13fe0198 | 1257 | error = dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin); |
b5256303 TC |
1258 | if (error != 0) { |
1259 | dsl_dir_rele(pdd, FTAG); | |
572e2857 | 1260 | return (error); |
b5256303 | 1261 | } |
572e2857 | 1262 | |
13fe0198 | 1263 | /* You can only clone snapshots, not the head datasets. */ |
0c66c32d | 1264 | if (!origin->ds_is_snapshot) { |
13fe0198 | 1265 | dsl_dataset_rele(origin, FTAG); |
b5256303 | 1266 | dsl_dir_rele(pdd, FTAG); |
2e528b49 | 1267 | return (SET_ERROR(EINVAL)); |
428870ff | 1268 | } |
b5256303 TC |
1269 | |
1270 | error = dmu_objset_clone_crypt_check(pdd, origin->ds_dir); | |
1271 | if (error != 0) { | |
1272 | dsl_dataset_rele(origin, FTAG); | |
1273 | dsl_dir_rele(pdd, FTAG); | |
1274 | return (error); | |
1275 | } | |
1276 | ||
13fe0198 | 1277 | dsl_dataset_rele(origin, FTAG); |
b5256303 | 1278 | dsl_dir_rele(pdd, FTAG); |
572e2857 | 1279 | |
13fe0198 | 1280 | return (0); |
9babb374 | 1281 | } |
34dc7c2f | 1282 | |
13fe0198 MA |
1283 | static void |
1284 | dmu_objset_clone_sync(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 1285 | { |
13fe0198 MA |
1286 | dmu_objset_clone_arg_t *doca = arg; |
1287 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
1288 | dsl_dir_t *pdd; | |
1289 | const char *tail; | |
1290 | dsl_dataset_t *origin, *ds; | |
1291 | uint64_t obj; | |
eca7b760 | 1292 | char namebuf[ZFS_MAX_DATASET_NAME_LEN]; |
6f1ffb06 | 1293 | |
13fe0198 MA |
1294 | VERIFY0(dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail)); |
1295 | VERIFY0(dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin)); | |
6f1ffb06 | 1296 | |
13fe0198 | 1297 | obj = dsl_dataset_create_sync(pdd, tail, origin, 0, |
b5256303 | 1298 | doca->doca_cred, NULL, tx); |
34dc7c2f | 1299 | |
13fe0198 MA |
1300 | VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); |
1301 | dsl_dataset_name(origin, namebuf); | |
1302 | spa_history_log_internal_ds(ds, "clone", tx, | |
1303 | "origin=%s (%llu)", namebuf, origin->ds_object); | |
a0bd735a | 1304 | zvol_create_minors(dp->dp_spa, doca->doca_clone, B_TRUE); |
13fe0198 MA |
1305 | dsl_dataset_rele(ds, FTAG); |
1306 | dsl_dataset_rele(origin, FTAG); | |
1307 | dsl_dir_rele(pdd, FTAG); | |
34dc7c2f BB |
1308 | } |
1309 | ||
1310 | int | |
13fe0198 | 1311 | dmu_objset_clone(const char *clone, const char *origin) |
34dc7c2f | 1312 | { |
13fe0198 | 1313 | dmu_objset_clone_arg_t doca; |
34dc7c2f | 1314 | |
13fe0198 MA |
1315 | doca.doca_clone = clone; |
1316 | doca.doca_origin = origin; | |
1317 | doca.doca_cred = CRED(); | |
572e2857 | 1318 | |
13fe0198 | 1319 | return (dsl_sync_task(clone, |
3d45fdd6 | 1320 | dmu_objset_clone_check, dmu_objset_clone_sync, &doca, |
b5256303 | 1321 | 6, ZFS_SPACE_CHECK_NORMAL)); |
6f1ffb06 MA |
1322 | } |
1323 | ||
1324 | int | |
1325 | dmu_objset_snapshot_one(const char *fsname, const char *snapname) | |
1326 | { | |
1327 | int err; | |
1328 | char *longsnap = kmem_asprintf("%s@%s", fsname, snapname); | |
1329 | nvlist_t *snaps = fnvlist_alloc(); | |
1330 | ||
1331 | fnvlist_add_boolean(snaps, longsnap); | |
6f1ffb06 | 1332 | strfree(longsnap); |
13fe0198 MA |
1333 | err = dsl_dataset_snapshot(snaps, NULL, NULL); |
1334 | fnvlist_free(snaps); | |
6f1ffb06 MA |
1335 | return (err); |
1336 | } | |
1337 | ||
1de321e6 JX |
1338 | static void |
1339 | dmu_objset_upgrade_task_cb(void *data) | |
1340 | { | |
1341 | objset_t *os = data; | |
1342 | ||
1343 | mutex_enter(&os->os_upgrade_lock); | |
1344 | os->os_upgrade_status = EINTR; | |
1345 | if (!os->os_upgrade_exit) { | |
1346 | mutex_exit(&os->os_upgrade_lock); | |
1347 | ||
1348 | os->os_upgrade_status = os->os_upgrade_cb(os); | |
1349 | mutex_enter(&os->os_upgrade_lock); | |
1350 | } | |
1351 | os->os_upgrade_exit = B_TRUE; | |
1352 | os->os_upgrade_id = 0; | |
1353 | mutex_exit(&os->os_upgrade_lock); | |
c0daec32 | 1354 | dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag); |
1de321e6 JX |
1355 | } |
1356 | ||
1357 | static void | |
1358 | dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb) | |
1359 | { | |
1360 | if (os->os_upgrade_id != 0) | |
1361 | return; | |
1362 | ||
c0daec32 AB |
1363 | ASSERT(dsl_pool_config_held(dmu_objset_pool(os))); |
1364 | dsl_dataset_long_hold(dmu_objset_ds(os), upgrade_tag); | |
1365 | ||
1de321e6 JX |
1366 | mutex_enter(&os->os_upgrade_lock); |
1367 | if (os->os_upgrade_id == 0 && os->os_upgrade_status == 0) { | |
1368 | os->os_upgrade_exit = B_FALSE; | |
1369 | os->os_upgrade_cb = cb; | |
1370 | os->os_upgrade_id = taskq_dispatch( | |
1371 | os->os_spa->spa_upgrade_taskq, | |
1372 | dmu_objset_upgrade_task_cb, os, TQ_SLEEP); | |
c0daec32 AB |
1373 | if (os->os_upgrade_id == TASKQID_INVALID) { |
1374 | dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag); | |
1de321e6 | 1375 | os->os_upgrade_status = ENOMEM; |
c0daec32 | 1376 | } |
1de321e6 JX |
1377 | } |
1378 | mutex_exit(&os->os_upgrade_lock); | |
1379 | } | |
1380 | ||
1381 | static void | |
1382 | dmu_objset_upgrade_stop(objset_t *os) | |
1383 | { | |
1384 | mutex_enter(&os->os_upgrade_lock); | |
1385 | os->os_upgrade_exit = B_TRUE; | |
1386 | if (os->os_upgrade_id != 0) { | |
1387 | taskqid_t id = os->os_upgrade_id; | |
1388 | ||
1389 | os->os_upgrade_id = 0; | |
1390 | mutex_exit(&os->os_upgrade_lock); | |
1391 | ||
c0daec32 AB |
1392 | if ((taskq_cancel_id(os->os_spa->spa_upgrade_taskq, id)) == 0) { |
1393 | dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag); | |
1394 | } | |
4807c0ba | 1395 | txg_wait_synced(os->os_spa->spa_dsl_pool, 0); |
1de321e6 JX |
1396 | } else { |
1397 | mutex_exit(&os->os_upgrade_lock); | |
1398 | } | |
1399 | } | |
1400 | ||
34dc7c2f | 1401 | static void |
64fc7762 | 1402 | dmu_objset_sync_dnodes(multilist_sublist_t *list, dmu_tx_t *tx) |
34dc7c2f BB |
1403 | { |
1404 | dnode_t *dn; | |
1405 | ||
64fc7762 | 1406 | while ((dn = multilist_sublist_head(list)) != NULL) { |
34dc7c2f BB |
1407 | ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT); |
1408 | ASSERT(dn->dn_dbuf->db_data_pending); | |
1409 | /* | |
9babb374 BB |
1410 | * Initialize dn_zio outside dnode_sync() because the |
1411 | * meta-dnode needs to set it ouside dnode_sync(). | |
34dc7c2f BB |
1412 | */ |
1413 | dn->dn_zio = dn->dn_dbuf->db_data_pending->dr_zio; | |
1414 | ASSERT(dn->dn_zio); | |
1415 | ||
1416 | ASSERT3U(dn->dn_nlevels, <=, DN_MAX_LEVELS); | |
64fc7762 | 1417 | multilist_sublist_remove(list, dn); |
9babb374 | 1418 | |
edc1e713 TC |
1419 | /* |
1420 | * If we are not doing useraccounting (os_synced_dnodes == NULL) | |
1421 | * we are done with this dnode for this txg. Unset dn_dirty_txg | |
1422 | * if later txgs aren't dirtying it so that future holders do | |
1423 | * not get a stale value. Otherwise, we will do this in | |
1424 | * userquota_updates_task() when processing has completely | |
1425 | * finished for this txg. | |
1426 | */ | |
64fc7762 MA |
1427 | multilist_t *newlist = dn->dn_objset->os_synced_dnodes; |
1428 | if (newlist != NULL) { | |
9babb374 | 1429 | (void) dnode_add_ref(dn, newlist); |
64fc7762 | 1430 | multilist_insert(newlist, dn); |
edc1e713 TC |
1431 | } else { |
1432 | mutex_enter(&dn->dn_mtx); | |
1433 | if (dn->dn_dirty_txg == tx->tx_txg) | |
1434 | dn->dn_dirty_txg = 0; | |
1435 | mutex_exit(&dn->dn_mtx); | |
9babb374 BB |
1436 | } |
1437 | ||
34dc7c2f BB |
1438 | dnode_sync(dn, tx); |
1439 | } | |
1440 | } | |
1441 | ||
1442 | /* ARGSUSED */ | |
1443 | static void | |
428870ff | 1444 | dmu_objset_write_ready(zio_t *zio, arc_buf_t *abuf, void *arg) |
34dc7c2f | 1445 | { |
b128c09f | 1446 | blkptr_t *bp = zio->io_bp; |
428870ff | 1447 | objset_t *os = arg; |
34dc7c2f | 1448 | dnode_phys_t *dnp = &os->os_phys->os_meta_dnode; |
b5256303 | 1449 | uint64_t fill = 0; |
34dc7c2f | 1450 | |
9b67f605 | 1451 | ASSERT(!BP_IS_EMBEDDED(bp)); |
13fe0198 MA |
1452 | ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_OBJSET); |
1453 | ASSERT0(BP_GET_LEVEL(bp)); | |
34dc7c2f BB |
1454 | |
1455 | /* | |
9babb374 BB |
1456 | * Update rootbp fill count: it should be the number of objects |
1457 | * allocated in the object set (not counting the "special" | |
1458 | * objects that are stored in the objset_phys_t -- the meta | |
9c5167d1 | 1459 | * dnode and user/group/project accounting objects). |
34dc7c2f | 1460 | */ |
1c27024e | 1461 | for (int i = 0; i < dnp->dn_nblkptr; i++) |
b5256303 TC |
1462 | fill += BP_GET_FILL(&dnp->dn_blkptr[i]); |
1463 | ||
1464 | BP_SET_FILL(bp, fill); | |
1465 | ||
cc9bb3e5 GM |
1466 | if (os->os_dsl_dataset != NULL) |
1467 | rrw_enter(&os->os_dsl_dataset->ds_bp_rwlock, RW_WRITER, FTAG); | |
1468 | *os->os_rootbp = *bp; | |
1469 | if (os->os_dsl_dataset != NULL) | |
1470 | rrw_exit(&os->os_dsl_dataset->ds_bp_rwlock, FTAG); | |
428870ff BB |
1471 | } |
1472 | ||
1473 | /* ARGSUSED */ | |
1474 | static void | |
1475 | dmu_objset_write_done(zio_t *zio, arc_buf_t *abuf, void *arg) | |
1476 | { | |
1477 | blkptr_t *bp = zio->io_bp; | |
1478 | blkptr_t *bp_orig = &zio->io_bp_orig; | |
1479 | objset_t *os = arg; | |
34dc7c2f | 1480 | |
b128c09f | 1481 | if (zio->io_flags & ZIO_FLAG_IO_REWRITE) { |
428870ff | 1482 | ASSERT(BP_EQUAL(bp, bp_orig)); |
b128c09f | 1483 | } else { |
428870ff BB |
1484 | dsl_dataset_t *ds = os->os_dsl_dataset; |
1485 | dmu_tx_t *tx = os->os_synctx; | |
1486 | ||
1487 | (void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE); | |
1488 | dsl_dataset_block_born(ds, bp, tx); | |
34dc7c2f | 1489 | } |
cc9bb3e5 | 1490 | kmem_free(bp, sizeof (*bp)); |
34dc7c2f BB |
1491 | } |
1492 | ||
64fc7762 MA |
1493 | typedef struct sync_dnodes_arg { |
1494 | multilist_t *sda_list; | |
1495 | int sda_sublist_idx; | |
1496 | multilist_t *sda_newlist; | |
1497 | dmu_tx_t *sda_tx; | |
1498 | } sync_dnodes_arg_t; | |
1499 | ||
1500 | static void | |
1501 | sync_dnodes_task(void *arg) | |
1502 | { | |
1503 | sync_dnodes_arg_t *sda = arg; | |
1504 | ||
1505 | multilist_sublist_t *ms = | |
1506 | multilist_sublist_lock(sda->sda_list, sda->sda_sublist_idx); | |
1507 | ||
1508 | dmu_objset_sync_dnodes(ms, sda->sda_tx); | |
1509 | ||
1510 | multilist_sublist_unlock(ms); | |
1511 | ||
1512 | kmem_free(sda, sizeof (*sda)); | |
1513 | } | |
1514 | ||
1515 | ||
34dc7c2f BB |
1516 | /* called from dsl */ |
1517 | void | |
428870ff | 1518 | dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx) |
34dc7c2f BB |
1519 | { |
1520 | int txgoff; | |
5dbd68a3 | 1521 | zbookmark_phys_t zb; |
428870ff | 1522 | zio_prop_t zp; |
34dc7c2f BB |
1523 | zio_t *zio; |
1524 | list_t *list; | |
1525 | dbuf_dirty_record_t *dr; | |
cc9bb3e5 GM |
1526 | blkptr_t *blkptr_copy = kmem_alloc(sizeof (*os->os_rootbp), KM_SLEEP); |
1527 | *blkptr_copy = *os->os_rootbp; | |
34dc7c2f BB |
1528 | |
1529 | dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg); | |
1530 | ||
1531 | ASSERT(dmu_tx_is_syncing(tx)); | |
1532 | /* XXX the write_done callback should really give us the tx... */ | |
1533 | os->os_synctx = tx; | |
1534 | ||
1535 | if (os->os_dsl_dataset == NULL) { | |
1536 | /* | |
1537 | * This is the MOS. If we have upgraded, | |
1538 | * spa_max_replication() could change, so reset | |
1539 | * os_copies here. | |
1540 | */ | |
1541 | os->os_copies = spa_max_replication(os->os_spa); | |
1542 | } | |
1543 | ||
1544 | /* | |
1545 | * Create the root block IO | |
1546 | */ | |
428870ff BB |
1547 | SET_BOOKMARK(&zb, os->os_dsl_dataset ? |
1548 | os->os_dsl_dataset->ds_object : DMU_META_OBJSET, | |
1549 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
294f6806 | 1550 | arc_release(os->os_phys_buf, &os->os_phys_buf); |
b128c09f | 1551 | |
82644107 | 1552 | dmu_write_policy(os, NULL, 0, 0, &zp); |
9babb374 | 1553 | |
b5256303 TC |
1554 | /* |
1555 | * If we are either claiming the ZIL or doing a raw receive write out | |
1556 | * the os_phys_buf raw. Neither of these actions will effect the MAC | |
1557 | * at this point. | |
1558 | */ | |
1b66810b | 1559 | if (os->os_next_write_raw[tx->tx_txg & TXG_MASK]) { |
b5256303 | 1560 | ASSERT(os->os_encrypted); |
1b66810b | 1561 | os->os_next_write_raw[tx->tx_txg & TXG_MASK] = B_FALSE; |
b5256303 TC |
1562 | arc_convert_to_raw(os->os_phys_buf, |
1563 | os->os_dsl_dataset->ds_object, ZFS_HOST_BYTEORDER, | |
1564 | DMU_OT_OBJSET, NULL, NULL, NULL); | |
1565 | } | |
1566 | ||
428870ff | 1567 | zio = arc_write(pio, os->os_spa, tx->tx_txg, |
cc9bb3e5 | 1568 | blkptr_copy, os->os_phys_buf, DMU_OS_IS_L2CACHEABLE(os), |
bc77ba73 PD |
1569 | &zp, dmu_objset_write_ready, NULL, NULL, dmu_objset_write_done, |
1570 | os, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb); | |
34dc7c2f BB |
1571 | |
1572 | /* | |
9babb374 | 1573 | * Sync special dnodes - the parent IO for the sync is the root block |
34dc7c2f | 1574 | */ |
572e2857 BB |
1575 | DMU_META_DNODE(os)->dn_zio = zio; |
1576 | dnode_sync(DMU_META_DNODE(os), tx); | |
34dc7c2f | 1577 | |
9babb374 BB |
1578 | os->os_phys->os_flags = os->os_flags; |
1579 | ||
572e2857 BB |
1580 | if (DMU_USERUSED_DNODE(os) && |
1581 | DMU_USERUSED_DNODE(os)->dn_type != DMU_OT_NONE) { | |
1582 | DMU_USERUSED_DNODE(os)->dn_zio = zio; | |
1583 | dnode_sync(DMU_USERUSED_DNODE(os), tx); | |
1584 | DMU_GROUPUSED_DNODE(os)->dn_zio = zio; | |
1585 | dnode_sync(DMU_GROUPUSED_DNODE(os), tx); | |
9babb374 BB |
1586 | } |
1587 | ||
9c5167d1 NF |
1588 | if (DMU_PROJECTUSED_DNODE(os) && |
1589 | DMU_PROJECTUSED_DNODE(os)->dn_type != DMU_OT_NONE) { | |
1590 | DMU_PROJECTUSED_DNODE(os)->dn_zio = zio; | |
1591 | dnode_sync(DMU_PROJECTUSED_DNODE(os), tx); | |
1592 | } | |
1593 | ||
34dc7c2f BB |
1594 | txgoff = tx->tx_txg & TXG_MASK; |
1595 | ||
b5256303 TC |
1596 | if (dmu_objset_userused_enabled(os) && |
1597 | (!os->os_encrypted || !dmu_objset_is_receiving(os))) { | |
9babb374 BB |
1598 | /* |
1599 | * We must create the list here because it uses the | |
64fc7762 MA |
1600 | * dn_dirty_link[] of this txg. But it may already |
1601 | * exist because we call dsl_dataset_sync() twice per txg. | |
9babb374 | 1602 | */ |
64fc7762 MA |
1603 | if (os->os_synced_dnodes == NULL) { |
1604 | os->os_synced_dnodes = | |
1605 | multilist_create(sizeof (dnode_t), | |
1606 | offsetof(dnode_t, dn_dirty_link[txgoff]), | |
1607 | dnode_multilist_index_func); | |
1608 | } else { | |
1609 | ASSERT3U(os->os_synced_dnodes->ml_offset, ==, | |
1610 | offsetof(dnode_t, dn_dirty_link[txgoff])); | |
1611 | } | |
9babb374 BB |
1612 | } |
1613 | ||
64fc7762 MA |
1614 | for (int i = 0; |
1615 | i < multilist_get_num_sublists(os->os_dirty_dnodes[txgoff]); i++) { | |
1616 | sync_dnodes_arg_t *sda = kmem_alloc(sizeof (*sda), KM_SLEEP); | |
1617 | sda->sda_list = os->os_dirty_dnodes[txgoff]; | |
1618 | sda->sda_sublist_idx = i; | |
1619 | sda->sda_tx = tx; | |
1620 | (void) taskq_dispatch(dmu_objset_pool(os)->dp_sync_taskq, | |
1621 | sync_dnodes_task, sda, 0); | |
1622 | /* callback frees sda */ | |
1623 | } | |
1624 | taskq_wait(dmu_objset_pool(os)->dp_sync_taskq); | |
34dc7c2f | 1625 | |
572e2857 | 1626 | list = &DMU_META_DNODE(os)->dn_dirty_records[txgoff]; |
64fc7762 | 1627 | while ((dr = list_head(list)) != NULL) { |
13fe0198 | 1628 | ASSERT0(dr->dr_dbuf->db_level); |
34dc7c2f BB |
1629 | list_remove(list, dr); |
1630 | if (dr->dr_zio) | |
1631 | zio_nowait(dr->dr_zio); | |
1632 | } | |
68cbd56e NB |
1633 | |
1634 | /* Enable dnode backfill if enough objects have been freed. */ | |
1635 | if (os->os_freed_dnodes >= dmu_rescan_dnode_threshold) { | |
1636 | os->os_rescan_dnodes = B_TRUE; | |
1637 | os->os_freed_dnodes = 0; | |
1638 | } | |
1639 | ||
34dc7c2f BB |
1640 | /* |
1641 | * Free intent log blocks up to this tx. | |
1642 | */ | |
1643 | zil_sync(os->os_zil, tx); | |
b128c09f | 1644 | os->os_phys->os_zil_header = os->os_zil_header; |
34dc7c2f BB |
1645 | zio_nowait(zio); |
1646 | } | |
1647 | ||
428870ff BB |
1648 | boolean_t |
1649 | dmu_objset_is_dirty(objset_t *os, uint64_t txg) | |
1650 | { | |
64fc7762 | 1651 | return (!multilist_is_empty(os->os_dirty_dnodes[txg & TXG_MASK])); |
428870ff BB |
1652 | } |
1653 | ||
572e2857 | 1654 | static objset_used_cb_t *used_cbs[DMU_OST_NUMTYPES]; |
9babb374 BB |
1655 | |
1656 | void | |
1657 | dmu_objset_register_type(dmu_objset_type_t ost, objset_used_cb_t *cb) | |
1658 | { | |
1659 | used_cbs[ost] = cb; | |
1660 | } | |
1661 | ||
1662 | boolean_t | |
428870ff | 1663 | dmu_objset_userused_enabled(objset_t *os) |
9babb374 BB |
1664 | { |
1665 | return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE && | |
572e2857 BB |
1666 | used_cbs[os->os_phys->os_type] != NULL && |
1667 | DMU_USERUSED_DNODE(os) != NULL); | |
9babb374 BB |
1668 | } |
1669 | ||
1de321e6 JX |
1670 | boolean_t |
1671 | dmu_objset_userobjused_enabled(objset_t *os) | |
1672 | { | |
1673 | return (dmu_objset_userused_enabled(os) && | |
1674 | spa_feature_is_enabled(os->os_spa, SPA_FEATURE_USEROBJ_ACCOUNTING)); | |
1675 | } | |
1676 | ||
9c5167d1 NF |
1677 | boolean_t |
1678 | dmu_objset_projectquota_enabled(objset_t *os) | |
1679 | { | |
1680 | return (used_cbs[os->os_phys->os_type] != NULL && | |
1681 | DMU_PROJECTUSED_DNODE(os) != NULL && | |
1682 | spa_feature_is_enabled(os->os_spa, SPA_FEATURE_PROJECT_QUOTA)); | |
1683 | } | |
1684 | ||
9b7a83cb JX |
1685 | typedef struct userquota_node { |
1686 | /* must be in the first filed, see userquota_update_cache() */ | |
1687 | char uqn_id[20 + DMU_OBJACCT_PREFIX_LEN]; | |
1688 | int64_t uqn_delta; | |
1689 | avl_node_t uqn_node; | |
1690 | } userquota_node_t; | |
1691 | ||
1692 | typedef struct userquota_cache { | |
1693 | avl_tree_t uqc_user_deltas; | |
1694 | avl_tree_t uqc_group_deltas; | |
9c5167d1 | 1695 | avl_tree_t uqc_project_deltas; |
9b7a83cb JX |
1696 | } userquota_cache_t; |
1697 | ||
1698 | static int | |
1699 | userquota_compare(const void *l, const void *r) | |
1700 | { | |
1701 | const userquota_node_t *luqn = l; | |
1702 | const userquota_node_t *ruqn = r; | |
e4ffa98d | 1703 | int rv; |
9b7a83cb JX |
1704 | |
1705 | /* | |
1706 | * NB: can only access uqn_id because userquota_update_cache() doesn't | |
1707 | * pass in an entire userquota_node_t. | |
1708 | */ | |
e4ffa98d BB |
1709 | rv = strcmp(luqn->uqn_id, ruqn->uqn_id); |
1710 | ||
1711 | return (AVL_ISIGN(rv)); | |
9b7a83cb JX |
1712 | } |
1713 | ||
1714 | static void | |
1715 | do_userquota_cacheflush(objset_t *os, userquota_cache_t *cache, dmu_tx_t *tx) | |
1716 | { | |
1717 | void *cookie; | |
1718 | userquota_node_t *uqn; | |
1719 | ||
1720 | ASSERT(dmu_tx_is_syncing(tx)); | |
1721 | ||
1722 | cookie = NULL; | |
1723 | while ((uqn = avl_destroy_nodes(&cache->uqc_user_deltas, | |
1724 | &cookie)) != NULL) { | |
64fc7762 MA |
1725 | /* |
1726 | * os_userused_lock protects against concurrent calls to | |
1727 | * zap_increment_int(). It's needed because zap_increment_int() | |
1728 | * is not thread-safe (i.e. not atomic). | |
1729 | */ | |
1730 | mutex_enter(&os->os_userused_lock); | |
9b7a83cb JX |
1731 | VERIFY0(zap_increment(os, DMU_USERUSED_OBJECT, |
1732 | uqn->uqn_id, uqn->uqn_delta, tx)); | |
64fc7762 | 1733 | mutex_exit(&os->os_userused_lock); |
9b7a83cb JX |
1734 | kmem_free(uqn, sizeof (*uqn)); |
1735 | } | |
1736 | avl_destroy(&cache->uqc_user_deltas); | |
1737 | ||
1738 | cookie = NULL; | |
1739 | while ((uqn = avl_destroy_nodes(&cache->uqc_group_deltas, | |
1740 | &cookie)) != NULL) { | |
64fc7762 | 1741 | mutex_enter(&os->os_userused_lock); |
9b7a83cb JX |
1742 | VERIFY0(zap_increment(os, DMU_GROUPUSED_OBJECT, |
1743 | uqn->uqn_id, uqn->uqn_delta, tx)); | |
64fc7762 | 1744 | mutex_exit(&os->os_userused_lock); |
9b7a83cb JX |
1745 | kmem_free(uqn, sizeof (*uqn)); |
1746 | } | |
1747 | avl_destroy(&cache->uqc_group_deltas); | |
9c5167d1 NF |
1748 | |
1749 | if (dmu_objset_projectquota_enabled(os)) { | |
1750 | cookie = NULL; | |
1751 | while ((uqn = avl_destroy_nodes(&cache->uqc_project_deltas, | |
1752 | &cookie)) != NULL) { | |
1753 | mutex_enter(&os->os_userused_lock); | |
1754 | VERIFY0(zap_increment(os, DMU_PROJECTUSED_OBJECT, | |
1755 | uqn->uqn_id, uqn->uqn_delta, tx)); | |
1756 | mutex_exit(&os->os_userused_lock); | |
1757 | kmem_free(uqn, sizeof (*uqn)); | |
1758 | } | |
1759 | avl_destroy(&cache->uqc_project_deltas); | |
1760 | } | |
9b7a83cb JX |
1761 | } |
1762 | ||
1763 | static void | |
1764 | userquota_update_cache(avl_tree_t *avl, const char *id, int64_t delta) | |
1765 | { | |
1766 | userquota_node_t *uqn; | |
1767 | avl_index_t idx; | |
1768 | ||
1769 | ASSERT(strlen(id) < sizeof (uqn->uqn_id)); | |
1770 | /* | |
1771 | * Use id directly for searching because uqn_id is the first field of | |
1772 | * userquota_node_t and fields after uqn_id won't be accessed in | |
1773 | * avl_find(). | |
1774 | */ | |
1775 | uqn = avl_find(avl, (const void *)id, &idx); | |
1776 | if (uqn == NULL) { | |
1777 | uqn = kmem_zalloc(sizeof (*uqn), KM_SLEEP); | |
1f51b525 | 1778 | strlcpy(uqn->uqn_id, id, sizeof (uqn->uqn_id)); |
9b7a83cb JX |
1779 | avl_insert(avl, uqn, idx); |
1780 | } | |
1781 | uqn->uqn_delta += delta; | |
1782 | } | |
1783 | ||
428870ff | 1784 | static void |
9c5167d1 NF |
1785 | do_userquota_update(objset_t *os, userquota_cache_t *cache, uint64_t used, |
1786 | uint64_t flags, uint64_t user, uint64_t group, uint64_t project, | |
1787 | boolean_t subtract) | |
428870ff | 1788 | { |
9c5167d1 | 1789 | if (flags & DNODE_FLAG_USERUSED_ACCOUNTED) { |
50c957f7 | 1790 | int64_t delta = DNODE_MIN_SIZE + used; |
9b7a83cb JX |
1791 | char name[20]; |
1792 | ||
428870ff BB |
1793 | if (subtract) |
1794 | delta = -delta; | |
9b7a83cb JX |
1795 | |
1796 | (void) sprintf(name, "%llx", (longlong_t)user); | |
1797 | userquota_update_cache(&cache->uqc_user_deltas, name, delta); | |
1798 | ||
1799 | (void) sprintf(name, "%llx", (longlong_t)group); | |
1800 | userquota_update_cache(&cache->uqc_group_deltas, name, delta); | |
9c5167d1 NF |
1801 | |
1802 | if (dmu_objset_projectquota_enabled(os)) { | |
1803 | (void) sprintf(name, "%llx", (longlong_t)project); | |
1804 | userquota_update_cache(&cache->uqc_project_deltas, | |
1805 | name, delta); | |
1806 | } | |
428870ff BB |
1807 | } |
1808 | } | |
1809 | ||
1de321e6 | 1810 | static void |
9c5167d1 NF |
1811 | do_userobjquota_update(objset_t *os, userquota_cache_t *cache, uint64_t flags, |
1812 | uint64_t user, uint64_t group, uint64_t project, boolean_t subtract) | |
1de321e6 JX |
1813 | { |
1814 | if (flags & DNODE_FLAG_USEROBJUSED_ACCOUNTED) { | |
1815 | char name[20 + DMU_OBJACCT_PREFIX_LEN]; | |
9b7a83cb | 1816 | int delta = subtract ? -1 : 1; |
1de321e6 JX |
1817 | |
1818 | (void) snprintf(name, sizeof (name), DMU_OBJACCT_PREFIX "%llx", | |
1819 | (longlong_t)user); | |
9b7a83cb | 1820 | userquota_update_cache(&cache->uqc_user_deltas, name, delta); |
1de321e6 JX |
1821 | |
1822 | (void) snprintf(name, sizeof (name), DMU_OBJACCT_PREFIX "%llx", | |
1823 | (longlong_t)group); | |
9b7a83cb | 1824 | userquota_update_cache(&cache->uqc_group_deltas, name, delta); |
9c5167d1 NF |
1825 | |
1826 | if (dmu_objset_projectquota_enabled(os)) { | |
1827 | (void) snprintf(name, sizeof (name), | |
1828 | DMU_OBJACCT_PREFIX "%llx", (longlong_t)project); | |
1829 | userquota_update_cache(&cache->uqc_project_deltas, | |
1830 | name, delta); | |
1831 | } | |
1de321e6 JX |
1832 | } |
1833 | } | |
1834 | ||
64fc7762 MA |
1835 | typedef struct userquota_updates_arg { |
1836 | objset_t *uua_os; | |
1837 | int uua_sublist_idx; | |
1838 | dmu_tx_t *uua_tx; | |
1839 | } userquota_updates_arg_t; | |
1840 | ||
1841 | static void | |
1842 | userquota_updates_task(void *arg) | |
9babb374 | 1843 | { |
64fc7762 MA |
1844 | userquota_updates_arg_t *uua = arg; |
1845 | objset_t *os = uua->uua_os; | |
1846 | dmu_tx_t *tx = uua->uua_tx; | |
9babb374 | 1847 | dnode_t *dn; |
9b7a83cb | 1848 | userquota_cache_t cache = { { 0 } }; |
9babb374 | 1849 | |
64fc7762 MA |
1850 | multilist_sublist_t *list = |
1851 | multilist_sublist_lock(os->os_synced_dnodes, uua->uua_sublist_idx); | |
9babb374 | 1852 | |
64fc7762 MA |
1853 | ASSERT(multilist_sublist_head(list) == NULL || |
1854 | dmu_objset_userused_enabled(os)); | |
9b7a83cb JX |
1855 | avl_create(&cache.uqc_user_deltas, userquota_compare, |
1856 | sizeof (userquota_node_t), offsetof(userquota_node_t, uqn_node)); | |
1857 | avl_create(&cache.uqc_group_deltas, userquota_compare, | |
1858 | sizeof (userquota_node_t), offsetof(userquota_node_t, uqn_node)); | |
9c5167d1 NF |
1859 | if (dmu_objset_projectquota_enabled(os)) |
1860 | avl_create(&cache.uqc_project_deltas, userquota_compare, | |
1861 | sizeof (userquota_node_t), offsetof(userquota_node_t, | |
1862 | uqn_node)); | |
9b7a83cb | 1863 | |
64fc7762 | 1864 | while ((dn = multilist_sublist_head(list)) != NULL) { |
572e2857 | 1865 | int flags; |
9babb374 | 1866 | ASSERT(!DMU_OBJECT_IS_SPECIAL(dn->dn_object)); |
9babb374 BB |
1867 | ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE || |
1868 | dn->dn_phys->dn_flags & | |
1869 | DNODE_FLAG_USERUSED_ACCOUNTED); | |
1870 | ||
572e2857 BB |
1871 | flags = dn->dn_id_flags; |
1872 | ASSERT(flags); | |
1873 | if (flags & DN_ID_OLD_EXIST) { | |
9c5167d1 NF |
1874 | do_userquota_update(os, &cache, dn->dn_oldused, |
1875 | dn->dn_oldflags, dn->dn_olduid, dn->dn_oldgid, | |
1876 | dn->dn_oldprojid, B_TRUE); | |
1877 | do_userobjquota_update(os, &cache, dn->dn_oldflags, | |
1878 | dn->dn_olduid, dn->dn_oldgid, | |
1879 | dn->dn_oldprojid, B_TRUE); | |
428870ff | 1880 | } |
572e2857 | 1881 | if (flags & DN_ID_NEW_EXIST) { |
9c5167d1 | 1882 | do_userquota_update(os, &cache, |
9b7a83cb | 1883 | DN_USED_BYTES(dn->dn_phys), dn->dn_phys->dn_flags, |
9c5167d1 NF |
1884 | dn->dn_newuid, dn->dn_newgid, |
1885 | dn->dn_newprojid, B_FALSE); | |
1886 | do_userobjquota_update(os, &cache, | |
1887 | dn->dn_phys->dn_flags, dn->dn_newuid, dn->dn_newgid, | |
1888 | dn->dn_newprojid, B_FALSE); | |
428870ff BB |
1889 | } |
1890 | ||
572e2857 | 1891 | mutex_enter(&dn->dn_mtx); |
428870ff BB |
1892 | dn->dn_oldused = 0; |
1893 | dn->dn_oldflags = 0; | |
1894 | if (dn->dn_id_flags & DN_ID_NEW_EXIST) { | |
1895 | dn->dn_olduid = dn->dn_newuid; | |
1896 | dn->dn_oldgid = dn->dn_newgid; | |
9c5167d1 | 1897 | dn->dn_oldprojid = dn->dn_newprojid; |
428870ff BB |
1898 | dn->dn_id_flags |= DN_ID_OLD_EXIST; |
1899 | if (dn->dn_bonuslen == 0) | |
1900 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1901 | else | |
1902 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1903 | } | |
1904 | dn->dn_id_flags &= ~(DN_ID_NEW_EXIST); | |
edc1e713 TC |
1905 | if (dn->dn_dirty_txg == spa_syncing_txg(os->os_spa)) |
1906 | dn->dn_dirty_txg = 0; | |
9babb374 BB |
1907 | mutex_exit(&dn->dn_mtx); |
1908 | ||
64fc7762 MA |
1909 | multilist_sublist_remove(list, dn); |
1910 | dnode_rele(dn, os->os_synced_dnodes); | |
9babb374 | 1911 | } |
9b7a83cb | 1912 | do_userquota_cacheflush(os, &cache, tx); |
64fc7762 MA |
1913 | multilist_sublist_unlock(list); |
1914 | kmem_free(uua, sizeof (*uua)); | |
1915 | } | |
1916 | ||
1917 | void | |
1918 | dmu_objset_do_userquota_updates(objset_t *os, dmu_tx_t *tx) | |
1919 | { | |
1920 | if (!dmu_objset_userused_enabled(os)) | |
1921 | return; | |
1922 | ||
163a8c28 TC |
1923 | /* |
1924 | * If this is a raw receive just return and handle accounting | |
1925 | * later when we have the keys loaded. We also don't do user | |
1926 | * accounting during claiming since the datasets are not owned | |
1927 | * for the duration of claiming and this txg should only be | |
1928 | * used for recovery. | |
1929 | */ | |
b5256303 TC |
1930 | if (os->os_encrypted && dmu_objset_is_receiving(os)) |
1931 | return; | |
1932 | ||
163a8c28 TC |
1933 | if (tx->tx_txg <= os->os_spa->spa_claim_max_txg) |
1934 | return; | |
1935 | ||
9c5167d1 | 1936 | /* Allocate the user/group/project used objects if necessary. */ |
64fc7762 MA |
1937 | if (DMU_USERUSED_DNODE(os)->dn_type == DMU_OT_NONE) { |
1938 | VERIFY0(zap_create_claim(os, | |
1939 | DMU_USERUSED_OBJECT, | |
1940 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
1941 | VERIFY0(zap_create_claim(os, | |
1942 | DMU_GROUPUSED_OBJECT, | |
1943 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
1944 | } | |
1945 | ||
9c5167d1 NF |
1946 | if (dmu_objset_projectquota_enabled(os) && |
1947 | DMU_PROJECTUSED_DNODE(os)->dn_type == DMU_OT_NONE) { | |
1948 | VERIFY0(zap_create_claim(os, DMU_PROJECTUSED_OBJECT, | |
1949 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
1950 | } | |
1951 | ||
64fc7762 MA |
1952 | for (int i = 0; |
1953 | i < multilist_get_num_sublists(os->os_synced_dnodes); i++) { | |
1954 | userquota_updates_arg_t *uua = | |
1955 | kmem_alloc(sizeof (*uua), KM_SLEEP); | |
1956 | uua->uua_os = os; | |
1957 | uua->uua_sublist_idx = i; | |
1958 | uua->uua_tx = tx; | |
1959 | /* note: caller does taskq_wait() */ | |
1960 | (void) taskq_dispatch(dmu_objset_pool(os)->dp_sync_taskq, | |
1961 | userquota_updates_task, uua, 0); | |
1962 | /* callback frees uua */ | |
1963 | } | |
9babb374 BB |
1964 | } |
1965 | ||
428870ff BB |
1966 | /* |
1967 | * Returns a pointer to data to find uid/gid from | |
1968 | * | |
1969 | * If a dirty record for transaction group that is syncing can't | |
1970 | * be found then NULL is returned. In the NULL case it is assumed | |
1971 | * the uid/gid aren't changing. | |
1972 | */ | |
1973 | static void * | |
1974 | dmu_objset_userquota_find_data(dmu_buf_impl_t *db, dmu_tx_t *tx) | |
1975 | { | |
1976 | dbuf_dirty_record_t *dr, **drp; | |
1977 | void *data; | |
1978 | ||
1979 | if (db->db_dirtycnt == 0) | |
1980 | return (db->db.db_data); /* Nothing is changing */ | |
1981 | ||
1982 | for (drp = &db->db_last_dirty; (dr = *drp) != NULL; drp = &dr->dr_next) | |
1983 | if (dr->dr_txg == tx->tx_txg) | |
1984 | break; | |
1985 | ||
572e2857 | 1986 | if (dr == NULL) { |
428870ff | 1987 | data = NULL; |
572e2857 BB |
1988 | } else { |
1989 | dnode_t *dn; | |
1990 | ||
1991 | DB_DNODE_ENTER(dr->dr_dbuf); | |
1992 | dn = DB_DNODE(dr->dr_dbuf); | |
1993 | ||
1994 | if (dn->dn_bonuslen == 0 && | |
1995 | dr->dr_dbuf->db_blkid == DMU_SPILL_BLKID) | |
1996 | data = dr->dt.dl.dr_data->b_data; | |
1997 | else | |
1998 | data = dr->dt.dl.dr_data; | |
1999 | ||
2000 | DB_DNODE_EXIT(dr->dr_dbuf); | |
2001 | } | |
2002 | ||
428870ff BB |
2003 | return (data); |
2004 | } | |
2005 | ||
2006 | void | |
2007 | dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) | |
2008 | { | |
2009 | objset_t *os = dn->dn_objset; | |
2010 | void *data = NULL; | |
2011 | dmu_buf_impl_t *db = NULL; | |
a117a6d6 GW |
2012 | uint64_t *user = NULL; |
2013 | uint64_t *group = NULL; | |
9c5167d1 | 2014 | uint64_t *project = NULL; |
428870ff BB |
2015 | int flags = dn->dn_id_flags; |
2016 | int error; | |
2017 | boolean_t have_spill = B_FALSE; | |
2018 | ||
2019 | if (!dmu_objset_userused_enabled(dn->dn_objset)) | |
2020 | return; | |
2021 | ||
b5256303 TC |
2022 | /* |
2023 | * Raw receives introduce a problem with user accounting. Raw | |
2024 | * receives cannot update the user accounting info because the | |
2025 | * user ids and the sizes are encrypted. To guarantee that we | |
2026 | * never end up with bad user accounting, we simply disable it | |
2027 | * during raw receives. We also disable this for normal receives | |
2028 | * so that an incremental raw receive may be done on top of an | |
2029 | * existing non-raw receive. | |
2030 | */ | |
2031 | if (os->os_encrypted && dmu_objset_is_receiving(os)) | |
2032 | return; | |
2033 | ||
428870ff BB |
2034 | if (before && (flags & (DN_ID_CHKED_BONUS|DN_ID_OLD_EXIST| |
2035 | DN_ID_CHKED_SPILL))) | |
2036 | return; | |
2037 | ||
2038 | if (before && dn->dn_bonuslen != 0) | |
2039 | data = DN_BONUS(dn->dn_phys); | |
2040 | else if (!before && dn->dn_bonuslen != 0) { | |
a3000f93 BB |
2041 | if (dn->dn_bonus) { |
2042 | db = dn->dn_bonus; | |
428870ff BB |
2043 | mutex_enter(&db->db_mtx); |
2044 | data = dmu_objset_userquota_find_data(db, tx); | |
2045 | } else { | |
2046 | data = DN_BONUS(dn->dn_phys); | |
2047 | } | |
2048 | } else if (dn->dn_bonuslen == 0 && dn->dn_bonustype == DMU_OT_SA) { | |
2049 | int rf = 0; | |
2050 | ||
2051 | if (RW_WRITE_HELD(&dn->dn_struct_rwlock)) | |
2052 | rf |= DB_RF_HAVESTRUCT; | |
572e2857 BB |
2053 | error = dmu_spill_hold_by_dnode(dn, |
2054 | rf | DB_RF_MUST_SUCCEED, | |
428870ff BB |
2055 | FTAG, (dmu_buf_t **)&db); |
2056 | ASSERT(error == 0); | |
2057 | mutex_enter(&db->db_mtx); | |
2058 | data = (before) ? db->db.db_data : | |
2059 | dmu_objset_userquota_find_data(db, tx); | |
2060 | have_spill = B_TRUE; | |
2061 | } else { | |
2062 | mutex_enter(&dn->dn_mtx); | |
2063 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
2064 | mutex_exit(&dn->dn_mtx); | |
2065 | return; | |
2066 | } | |
2067 | ||
2068 | if (before) { | |
2069 | ASSERT(data); | |
2070 | user = &dn->dn_olduid; | |
2071 | group = &dn->dn_oldgid; | |
9c5167d1 | 2072 | project = &dn->dn_oldprojid; |
428870ff BB |
2073 | } else if (data) { |
2074 | user = &dn->dn_newuid; | |
2075 | group = &dn->dn_newgid; | |
9c5167d1 | 2076 | project = &dn->dn_newprojid; |
428870ff BB |
2077 | } |
2078 | ||
2079 | /* | |
2080 | * Must always call the callback in case the object | |
2081 | * type has changed and that type isn't an object type to track | |
2082 | */ | |
2083 | error = used_cbs[os->os_phys->os_type](dn->dn_bonustype, data, | |
9c5167d1 | 2084 | user, group, project); |
428870ff BB |
2085 | |
2086 | /* | |
2087 | * Preserve existing uid/gid when the callback can't determine | |
2088 | * what the new uid/gid are and the callback returned EEXIST. | |
2089 | * The EEXIST error tells us to just use the existing uid/gid. | |
2090 | * If we don't know what the old values are then just assign | |
2091 | * them to 0, since that is a new file being created. | |
2092 | */ | |
2093 | if (!before && data == NULL && error == EEXIST) { | |
2094 | if (flags & DN_ID_OLD_EXIST) { | |
2095 | dn->dn_newuid = dn->dn_olduid; | |
2096 | dn->dn_newgid = dn->dn_oldgid; | |
2705ebf0 | 2097 | dn->dn_newprojid = dn->dn_oldprojid; |
428870ff BB |
2098 | } else { |
2099 | dn->dn_newuid = 0; | |
2100 | dn->dn_newgid = 0; | |
9c5167d1 | 2101 | dn->dn_newprojid = ZFS_DEFAULT_PROJID; |
428870ff BB |
2102 | } |
2103 | error = 0; | |
2104 | } | |
2105 | ||
2106 | if (db) | |
2107 | mutex_exit(&db->db_mtx); | |
2108 | ||
2109 | mutex_enter(&dn->dn_mtx); | |
2110 | if (error == 0 && before) | |
2111 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
2112 | if (error == 0 && !before) | |
2113 | dn->dn_id_flags |= DN_ID_NEW_EXIST; | |
2114 | ||
2115 | if (have_spill) { | |
2116 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
2117 | } else { | |
2118 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
2119 | } | |
2120 | mutex_exit(&dn->dn_mtx); | |
a3000f93 | 2121 | if (have_spill) |
428870ff BB |
2122 | dmu_buf_rele((dmu_buf_t *)db, FTAG); |
2123 | } | |
2124 | ||
9babb374 BB |
2125 | boolean_t |
2126 | dmu_objset_userspace_present(objset_t *os) | |
2127 | { | |
428870ff | 2128 | return (os->os_phys->os_flags & |
9babb374 BB |
2129 | OBJSET_FLAG_USERACCOUNTING_COMPLETE); |
2130 | } | |
2131 | ||
1de321e6 JX |
2132 | boolean_t |
2133 | dmu_objset_userobjspace_present(objset_t *os) | |
2134 | { | |
2135 | return (os->os_phys->os_flags & | |
2136 | OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE); | |
2137 | } | |
2138 | ||
9c5167d1 NF |
2139 | boolean_t |
2140 | dmu_objset_projectquota_present(objset_t *os) | |
2141 | { | |
2142 | return (os->os_phys->os_flags & | |
2143 | OBJSET_FLAG_PROJECTQUOTA_COMPLETE); | |
2144 | } | |
2145 | ||
1de321e6 JX |
2146 | static int |
2147 | dmu_objset_space_upgrade(objset_t *os) | |
9babb374 BB |
2148 | { |
2149 | uint64_t obj; | |
2150 | int err = 0; | |
2151 | ||
9babb374 BB |
2152 | /* |
2153 | * We simply need to mark every object dirty, so that it will be | |
2154 | * synced out and now accounted. If this is called | |
2155 | * concurrently, or if we already did some work before crashing, | |
2156 | * that's fine, since we track each object's accounted state | |
2157 | * independently. | |
2158 | */ | |
2159 | ||
2160 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
45d1cae3 | 2161 | dmu_tx_t *tx; |
9babb374 BB |
2162 | dmu_buf_t *db; |
2163 | int objerr; | |
2164 | ||
1de321e6 JX |
2165 | mutex_enter(&os->os_upgrade_lock); |
2166 | if (os->os_upgrade_exit) | |
2167 | err = SET_ERROR(EINTR); | |
2168 | mutex_exit(&os->os_upgrade_lock); | |
2169 | if (err != 0) | |
2170 | return (err); | |
2171 | ||
9babb374 | 2172 | if (issig(JUSTLOOKING) && issig(FORREAL)) |
2e528b49 | 2173 | return (SET_ERROR(EINTR)); |
9babb374 BB |
2174 | |
2175 | objerr = dmu_bonus_hold(os, obj, FTAG, &db); | |
13fe0198 | 2176 | if (objerr != 0) |
9babb374 | 2177 | continue; |
45d1cae3 | 2178 | tx = dmu_tx_create(os); |
9babb374 BB |
2179 | dmu_tx_hold_bonus(tx, obj); |
2180 | objerr = dmu_tx_assign(tx, TXG_WAIT); | |
13fe0198 | 2181 | if (objerr != 0) { |
d22323e8 | 2182 | dmu_buf_rele(db, FTAG); |
9babb374 BB |
2183 | dmu_tx_abort(tx); |
2184 | continue; | |
2185 | } | |
2186 | dmu_buf_will_dirty(db, tx); | |
2187 | dmu_buf_rele(db, FTAG); | |
2188 | dmu_tx_commit(tx); | |
2189 | } | |
1de321e6 JX |
2190 | return (0); |
2191 | } | |
2192 | ||
2193 | int | |
2194 | dmu_objset_userspace_upgrade(objset_t *os) | |
2195 | { | |
2196 | int err = 0; | |
2197 | ||
2198 | if (dmu_objset_userspace_present(os)) | |
2199 | return (0); | |
2200 | if (dmu_objset_is_snapshot(os)) | |
2201 | return (SET_ERROR(EINVAL)); | |
2202 | if (!dmu_objset_userused_enabled(os)) | |
2203 | return (SET_ERROR(ENOTSUP)); | |
2204 | ||
2205 | err = dmu_objset_space_upgrade(os); | |
2206 | if (err) | |
2207 | return (err); | |
9babb374 | 2208 | |
428870ff | 2209 | os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; |
9babb374 BB |
2210 | txg_wait_synced(dmu_objset_pool(os), 0); |
2211 | return (0); | |
2212 | } | |
2213 | ||
1de321e6 | 2214 | static int |
9c5167d1 | 2215 | dmu_objset_id_quota_upgrade_cb(objset_t *os) |
1de321e6 JX |
2216 | { |
2217 | int err = 0; | |
2218 | ||
9c5167d1 NF |
2219 | if (dmu_objset_userobjspace_present(os) && |
2220 | dmu_objset_projectquota_present(os)) | |
1de321e6 JX |
2221 | return (0); |
2222 | if (dmu_objset_is_snapshot(os)) | |
2223 | return (SET_ERROR(EINVAL)); | |
2224 | if (!dmu_objset_userobjused_enabled(os)) | |
2225 | return (SET_ERROR(ENOTSUP)); | |
9c5167d1 NF |
2226 | if (!dmu_objset_projectquota_enabled(os) && |
2227 | dmu_objset_userobjspace_present(os)) | |
2228 | return (SET_ERROR(ENOTSUP)); | |
1de321e6 JX |
2229 | |
2230 | dmu_objset_ds(os)->ds_feature_activation_needed[ | |
2231 | SPA_FEATURE_USEROBJ_ACCOUNTING] = B_TRUE; | |
9c5167d1 NF |
2232 | if (dmu_objset_projectquota_enabled(os)) |
2233 | dmu_objset_ds(os)->ds_feature_activation_needed[ | |
2234 | SPA_FEATURE_PROJECT_QUOTA] = B_TRUE; | |
1de321e6 JX |
2235 | |
2236 | err = dmu_objset_space_upgrade(os); | |
2237 | if (err) | |
2238 | return (err); | |
2239 | ||
2240 | os->os_flags |= OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE; | |
9c5167d1 NF |
2241 | if (dmu_objset_projectquota_enabled(os)) |
2242 | os->os_flags |= OBJSET_FLAG_PROJECTQUOTA_COMPLETE; | |
2243 | ||
1de321e6 JX |
2244 | txg_wait_synced(dmu_objset_pool(os), 0); |
2245 | return (0); | |
2246 | } | |
2247 | ||
2248 | void | |
9c5167d1 | 2249 | dmu_objset_id_quota_upgrade(objset_t *os) |
1de321e6 | 2250 | { |
9c5167d1 | 2251 | dmu_objset_upgrade(os, dmu_objset_id_quota_upgrade_cb); |
1de321e6 JX |
2252 | } |
2253 | ||
126ae9f4 JX |
2254 | boolean_t |
2255 | dmu_objset_userobjspace_upgradable(objset_t *os) | |
2256 | { | |
2257 | return (dmu_objset_type(os) == DMU_OST_ZFS && | |
2258 | !dmu_objset_is_snapshot(os) && | |
2259 | dmu_objset_userobjused_enabled(os) && | |
2260 | !dmu_objset_userobjspace_present(os)); | |
2261 | } | |
2262 | ||
9c5167d1 NF |
2263 | boolean_t |
2264 | dmu_objset_projectquota_upgradable(objset_t *os) | |
2265 | { | |
2266 | return (dmu_objset_type(os) == DMU_OST_ZFS && | |
2267 | !dmu_objset_is_snapshot(os) && | |
2268 | dmu_objset_projectquota_enabled(os) && | |
2269 | !dmu_objset_projectquota_present(os)); | |
2270 | } | |
2271 | ||
34dc7c2f BB |
2272 | void |
2273 | dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, | |
2274 | uint64_t *usedobjsp, uint64_t *availobjsp) | |
2275 | { | |
428870ff | 2276 | dsl_dataset_space(os->os_dsl_dataset, refdbytesp, availbytesp, |
34dc7c2f BB |
2277 | usedobjsp, availobjsp); |
2278 | } | |
2279 | ||
2280 | uint64_t | |
2281 | dmu_objset_fsid_guid(objset_t *os) | |
2282 | { | |
428870ff | 2283 | return (dsl_dataset_fsid_guid(os->os_dsl_dataset)); |
34dc7c2f BB |
2284 | } |
2285 | ||
2286 | void | |
2287 | dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat) | |
2288 | { | |
428870ff BB |
2289 | stat->dds_type = os->os_phys->os_type; |
2290 | if (os->os_dsl_dataset) | |
2291 | dsl_dataset_fast_stat(os->os_dsl_dataset, stat); | |
34dc7c2f BB |
2292 | } |
2293 | ||
2294 | void | |
2295 | dmu_objset_stats(objset_t *os, nvlist_t *nv) | |
2296 | { | |
428870ff BB |
2297 | ASSERT(os->os_dsl_dataset || |
2298 | os->os_phys->os_type == DMU_OST_META); | |
34dc7c2f | 2299 | |
428870ff BB |
2300 | if (os->os_dsl_dataset != NULL) |
2301 | dsl_dataset_stats(os->os_dsl_dataset, nv); | |
34dc7c2f BB |
2302 | |
2303 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_TYPE, | |
428870ff | 2304 | os->os_phys->os_type); |
9babb374 BB |
2305 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERACCOUNTING, |
2306 | dmu_objset_userspace_present(os)); | |
34dc7c2f BB |
2307 | } |
2308 | ||
2309 | int | |
2310 | dmu_objset_is_snapshot(objset_t *os) | |
2311 | { | |
428870ff | 2312 | if (os->os_dsl_dataset != NULL) |
0c66c32d | 2313 | return (os->os_dsl_dataset->ds_is_snapshot); |
34dc7c2f BB |
2314 | else |
2315 | return (B_FALSE); | |
2316 | } | |
2317 | ||
2318 | int | |
2319 | dmu_snapshot_realname(objset_t *os, char *name, char *real, int maxlen, | |
2320 | boolean_t *conflict) | |
2321 | { | |
428870ff | 2322 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
2323 | uint64_t ignored; |
2324 | ||
d683ddbb | 2325 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 2326 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2327 | |
2328 | return (zap_lookup_norm(ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb | 2329 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, name, 8, 1, &ignored, |
9b7b9cd3 | 2330 | MT_NORMALIZE, real, maxlen, conflict)); |
34dc7c2f BB |
2331 | } |
2332 | ||
2333 | int | |
2334 | dmu_snapshot_list_next(objset_t *os, int namelen, char *name, | |
2335 | uint64_t *idp, uint64_t *offp, boolean_t *case_conflict) | |
2336 | { | |
428870ff | 2337 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
2338 | zap_cursor_t cursor; |
2339 | zap_attribute_t attr; | |
2340 | ||
13fe0198 MA |
2341 | ASSERT(dsl_pool_config_held(dmu_objset_pool(os))); |
2342 | ||
d683ddbb | 2343 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 2344 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2345 | |
2346 | zap_cursor_init_serialized(&cursor, | |
2347 | ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb | 2348 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, *offp); |
34dc7c2f BB |
2349 | |
2350 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
2351 | zap_cursor_fini(&cursor); | |
2e528b49 | 2352 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2353 | } |
2354 | ||
2355 | if (strlen(attr.za_name) + 1 > namelen) { | |
2356 | zap_cursor_fini(&cursor); | |
2e528b49 | 2357 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
2358 | } |
2359 | ||
2360 | (void) strcpy(name, attr.za_name); | |
2361 | if (idp) | |
2362 | *idp = attr.za_first_integer; | |
2363 | if (case_conflict) | |
2364 | *case_conflict = attr.za_normalization_conflict; | |
2365 | zap_cursor_advance(&cursor); | |
2366 | *offp = zap_cursor_serialize(&cursor); | |
2367 | zap_cursor_fini(&cursor); | |
2368 | ||
2369 | return (0); | |
2370 | } | |
2371 | ||
ebe7e575 | 2372 | int |
6772fb67 | 2373 | dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *value) |
ebe7e575 | 2374 | { |
d1d7e268 | 2375 | return (dsl_dataset_snap_lookup(os->os_dsl_dataset, name, value)); |
ebe7e575 BB |
2376 | } |
2377 | ||
34dc7c2f BB |
2378 | int |
2379 | dmu_dir_list_next(objset_t *os, int namelen, char *name, | |
2380 | uint64_t *idp, uint64_t *offp) | |
2381 | { | |
428870ff | 2382 | dsl_dir_t *dd = os->os_dsl_dataset->ds_dir; |
34dc7c2f BB |
2383 | zap_cursor_t cursor; |
2384 | zap_attribute_t attr; | |
2385 | ||
2386 | /* there is no next dir on a snapshot! */ | |
428870ff | 2387 | if (os->os_dsl_dataset->ds_object != |
d683ddbb | 2388 | dsl_dir_phys(dd)->dd_head_dataset_obj) |
2e528b49 | 2389 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2390 | |
2391 | zap_cursor_init_serialized(&cursor, | |
2392 | dd->dd_pool->dp_meta_objset, | |
d683ddbb | 2393 | dsl_dir_phys(dd)->dd_child_dir_zapobj, *offp); |
34dc7c2f BB |
2394 | |
2395 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
2396 | zap_cursor_fini(&cursor); | |
2e528b49 | 2397 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2398 | } |
2399 | ||
2400 | if (strlen(attr.za_name) + 1 > namelen) { | |
2401 | zap_cursor_fini(&cursor); | |
2e528b49 | 2402 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
2403 | } |
2404 | ||
2405 | (void) strcpy(name, attr.za_name); | |
2406 | if (idp) | |
2407 | *idp = attr.za_first_integer; | |
2408 | zap_cursor_advance(&cursor); | |
2409 | *offp = zap_cursor_serialize(&cursor); | |
2410 | zap_cursor_fini(&cursor); | |
2411 | ||
2412 | return (0); | |
2413 | } | |
2414 | ||
9c43027b AJ |
2415 | typedef struct dmu_objset_find_ctx { |
2416 | taskq_t *dc_tq; | |
2417 | dsl_pool_t *dc_dp; | |
2418 | uint64_t dc_ddobj; | |
1149ba64 | 2419 | char *dc_ddname; /* last component of ddobj's name */ |
9c43027b AJ |
2420 | int (*dc_func)(dsl_pool_t *, dsl_dataset_t *, void *); |
2421 | void *dc_arg; | |
2422 | int dc_flags; | |
2423 | kmutex_t *dc_error_lock; | |
2424 | int *dc_error; | |
2425 | } dmu_objset_find_ctx_t; | |
2426 | ||
2427 | static void | |
2428 | dmu_objset_find_dp_impl(dmu_objset_find_ctx_t *dcp) | |
b128c09f | 2429 | { |
9c43027b | 2430 | dsl_pool_t *dp = dcp->dc_dp; |
13fe0198 MA |
2431 | dsl_dir_t *dd; |
2432 | dsl_dataset_t *ds; | |
2433 | zap_cursor_t zc; | |
2434 | zap_attribute_t *attr; | |
2435 | uint64_t thisobj; | |
9c43027b | 2436 | int err = 0; |
13fe0198 | 2437 | |
9c43027b AJ |
2438 | /* don't process if there already was an error */ |
2439 | if (*dcp->dc_error != 0) | |
2440 | goto out; | |
13fe0198 | 2441 | |
1149ba64 GM |
2442 | /* |
2443 | * Note: passing the name (dc_ddname) here is optional, but it | |
2444 | * improves performance because we don't need to call | |
2445 | * zap_value_search() to determine the name. | |
2446 | */ | |
2447 | err = dsl_dir_hold_obj(dp, dcp->dc_ddobj, dcp->dc_ddname, FTAG, &dd); | |
13fe0198 | 2448 | if (err != 0) |
9c43027b | 2449 | goto out; |
13fe0198 MA |
2450 | |
2451 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ | |
2452 | if (dd->dd_myname[0] == '$') { | |
2453 | dsl_dir_rele(dd, FTAG); | |
9c43027b | 2454 | goto out; |
13fe0198 MA |
2455 | } |
2456 | ||
d683ddbb | 2457 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 2458 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
13fe0198 MA |
2459 | |
2460 | /* | |
2461 | * Iterate over all children. | |
2462 | */ | |
9c43027b | 2463 | if (dcp->dc_flags & DS_FIND_CHILDREN) { |
13fe0198 | 2464 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
d683ddbb | 2465 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
13fe0198 MA |
2466 | zap_cursor_retrieve(&zc, attr) == 0; |
2467 | (void) zap_cursor_advance(&zc)) { | |
2468 | ASSERT3U(attr->za_integer_length, ==, | |
2469 | sizeof (uint64_t)); | |
2470 | ASSERT3U(attr->za_num_integers, ==, 1); | |
2471 | ||
1c27024e | 2472 | dmu_objset_find_ctx_t *child_dcp = |
1149ba64 | 2473 | kmem_alloc(sizeof (*child_dcp), KM_SLEEP); |
9c43027b AJ |
2474 | *child_dcp = *dcp; |
2475 | child_dcp->dc_ddobj = attr->za_first_integer; | |
1149ba64 | 2476 | child_dcp->dc_ddname = spa_strdup(attr->za_name); |
9c43027b AJ |
2477 | if (dcp->dc_tq != NULL) |
2478 | (void) taskq_dispatch(dcp->dc_tq, | |
2479 | dmu_objset_find_dp_cb, child_dcp, TQ_SLEEP); | |
2480 | else | |
2481 | dmu_objset_find_dp_impl(child_dcp); | |
13fe0198 MA |
2482 | } |
2483 | zap_cursor_fini(&zc); | |
13fe0198 MA |
2484 | } |
2485 | ||
2486 | /* | |
2487 | * Iterate over all snapshots. | |
2488 | */ | |
9c43027b | 2489 | if (dcp->dc_flags & DS_FIND_SNAPSHOTS) { |
13fe0198 MA |
2490 | dsl_dataset_t *ds; |
2491 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
2492 | ||
2493 | if (err == 0) { | |
d683ddbb JG |
2494 | uint64_t snapobj; |
2495 | ||
2496 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
13fe0198 MA |
2497 | dsl_dataset_rele(ds, FTAG); |
2498 | ||
2499 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
2500 | zap_cursor_retrieve(&zc, attr) == 0; | |
2501 | (void) zap_cursor_advance(&zc)) { | |
2502 | ASSERT3U(attr->za_integer_length, ==, | |
2503 | sizeof (uint64_t)); | |
2504 | ASSERT3U(attr->za_num_integers, ==, 1); | |
2505 | ||
2506 | err = dsl_dataset_hold_obj(dp, | |
2507 | attr->za_first_integer, FTAG, &ds); | |
2508 | if (err != 0) | |
2509 | break; | |
9c43027b | 2510 | err = dcp->dc_func(dp, ds, dcp->dc_arg); |
13fe0198 MA |
2511 | dsl_dataset_rele(ds, FTAG); |
2512 | if (err != 0) | |
2513 | break; | |
2514 | } | |
2515 | zap_cursor_fini(&zc); | |
2516 | } | |
2517 | } | |
2518 | ||
13fe0198 MA |
2519 | kmem_free(attr, sizeof (zap_attribute_t)); |
2520 | ||
1149ba64 GM |
2521 | if (err != 0) { |
2522 | dsl_dir_rele(dd, FTAG); | |
9c43027b | 2523 | goto out; |
1149ba64 | 2524 | } |
13fe0198 MA |
2525 | |
2526 | /* | |
2527 | * Apply to self. | |
2528 | */ | |
2529 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1149ba64 GM |
2530 | |
2531 | /* | |
2532 | * Note: we hold the dir while calling dsl_dataset_hold_obj() so | |
2533 | * that the dir will remain cached, and we won't have to re-instantiate | |
2534 | * it (which could be expensive due to finding its name via | |
2535 | * zap_value_search()). | |
2536 | */ | |
2537 | dsl_dir_rele(dd, FTAG); | |
13fe0198 | 2538 | if (err != 0) |
9c43027b AJ |
2539 | goto out; |
2540 | err = dcp->dc_func(dp, ds, dcp->dc_arg); | |
13fe0198 | 2541 | dsl_dataset_rele(ds, FTAG); |
9c43027b AJ |
2542 | |
2543 | out: | |
2544 | if (err != 0) { | |
2545 | mutex_enter(dcp->dc_error_lock); | |
2546 | /* only keep first error */ | |
2547 | if (*dcp->dc_error == 0) | |
2548 | *dcp->dc_error = err; | |
2549 | mutex_exit(dcp->dc_error_lock); | |
2550 | } | |
2551 | ||
1149ba64 GM |
2552 | if (dcp->dc_ddname != NULL) |
2553 | spa_strfree(dcp->dc_ddname); | |
9c43027b AJ |
2554 | kmem_free(dcp, sizeof (*dcp)); |
2555 | } | |
2556 | ||
2557 | static void | |
2558 | dmu_objset_find_dp_cb(void *arg) | |
2559 | { | |
2560 | dmu_objset_find_ctx_t *dcp = arg; | |
2561 | dsl_pool_t *dp = dcp->dc_dp; | |
2562 | ||
5e8cd5d1 AJ |
2563 | /* |
2564 | * We need to get a pool_config_lock here, as there are several | |
2565 | * asssert(pool_config_held) down the stack. Getting a lock via | |
2566 | * dsl_pool_config_enter is risky, as it might be stalled by a | |
2567 | * pending writer. This would deadlock, as the write lock can | |
2568 | * only be granted when our parent thread gives up the lock. | |
2569 | * The _prio interface gives us priority over a pending writer. | |
2570 | */ | |
2571 | dsl_pool_config_enter_prio(dp, FTAG); | |
9c43027b AJ |
2572 | |
2573 | dmu_objset_find_dp_impl(dcp); | |
2574 | ||
2575 | dsl_pool_config_exit(dp, FTAG); | |
2576 | } | |
2577 | ||
2578 | /* | |
2579 | * Find objsets under and including ddobj, call func(ds) on each. | |
2580 | * The order for the enumeration is completely undefined. | |
2581 | * func is called with dsl_pool_config held. | |
2582 | */ | |
2583 | int | |
2584 | dmu_objset_find_dp(dsl_pool_t *dp, uint64_t ddobj, | |
2585 | int func(dsl_pool_t *, dsl_dataset_t *, void *), void *arg, int flags) | |
2586 | { | |
2587 | int error = 0; | |
2588 | taskq_t *tq = NULL; | |
2589 | int ntasks; | |
2590 | dmu_objset_find_ctx_t *dcp; | |
2591 | kmutex_t err_lock; | |
2592 | ||
2593 | mutex_init(&err_lock, NULL, MUTEX_DEFAULT, NULL); | |
2594 | dcp = kmem_alloc(sizeof (*dcp), KM_SLEEP); | |
2595 | dcp->dc_tq = NULL; | |
2596 | dcp->dc_dp = dp; | |
2597 | dcp->dc_ddobj = ddobj; | |
1149ba64 | 2598 | dcp->dc_ddname = NULL; |
9c43027b AJ |
2599 | dcp->dc_func = func; |
2600 | dcp->dc_arg = arg; | |
2601 | dcp->dc_flags = flags; | |
2602 | dcp->dc_error_lock = &err_lock; | |
2603 | dcp->dc_error = &error; | |
2604 | ||
2605 | if ((flags & DS_FIND_SERIALIZE) || dsl_pool_config_held_writer(dp)) { | |
2606 | /* | |
2607 | * In case a write lock is held we can't make use of | |
2608 | * parallelism, as down the stack of the worker threads | |
2609 | * the lock is asserted via dsl_pool_config_held. | |
2610 | * In case of a read lock this is solved by getting a read | |
2611 | * lock in each worker thread, which isn't possible in case | |
2612 | * of a writer lock. So we fall back to the synchronous path | |
2613 | * here. | |
2614 | * In the future it might be possible to get some magic into | |
2615 | * dsl_pool_config_held in a way that it returns true for | |
2616 | * the worker threads so that a single lock held from this | |
2617 | * thread suffices. For now, stay single threaded. | |
2618 | */ | |
2619 | dmu_objset_find_dp_impl(dcp); | |
e5676636 | 2620 | mutex_destroy(&err_lock); |
9c43027b AJ |
2621 | |
2622 | return (error); | |
2623 | } | |
2624 | ||
2625 | ntasks = dmu_find_threads; | |
2626 | if (ntasks == 0) | |
2627 | ntasks = vdev_count_leaves(dp->dp_spa) * 4; | |
1229323d | 2628 | tq = taskq_create("dmu_objset_find", ntasks, maxclsyspri, ntasks, |
9c43027b AJ |
2629 | INT_MAX, 0); |
2630 | if (tq == NULL) { | |
2631 | kmem_free(dcp, sizeof (*dcp)); | |
e5676636 BB |
2632 | mutex_destroy(&err_lock); |
2633 | ||
9c43027b AJ |
2634 | return (SET_ERROR(ENOMEM)); |
2635 | } | |
2636 | dcp->dc_tq = tq; | |
2637 | ||
2638 | /* dcp will be freed by task */ | |
2639 | (void) taskq_dispatch(tq, dmu_objset_find_dp_cb, dcp, TQ_SLEEP); | |
2640 | ||
2641 | /* | |
2642 | * PORTING: this code relies on the property of taskq_wait to wait | |
2643 | * until no more tasks are queued and no more tasks are active. As | |
2644 | * we always queue new tasks from within other tasks, task_wait | |
2645 | * reliably waits for the full recursion to finish, even though we | |
2646 | * enqueue new tasks after taskq_wait has been called. | |
2647 | * On platforms other than illumos, taskq_wait may not have this | |
2648 | * property. | |
2649 | */ | |
2650 | taskq_wait(tq); | |
2651 | taskq_destroy(tq); | |
2652 | mutex_destroy(&err_lock); | |
2653 | ||
2654 | return (error); | |
b128c09f BB |
2655 | } |
2656 | ||
2657 | /* | |
13fe0198 MA |
2658 | * Find all objsets under name, and for each, call 'func(child_name, arg)'. |
2659 | * The dp_config_rwlock must not be held when this is called, and it | |
2660 | * will not be held when the callback is called. | |
2661 | * Therefore this function should only be used when the pool is not changing | |
2662 | * (e.g. in syncing context), or the callback can deal with the possible races. | |
b128c09f | 2663 | */ |
13fe0198 MA |
2664 | static int |
2665 | dmu_objset_find_impl(spa_t *spa, const char *name, | |
2666 | int func(const char *, void *), void *arg, int flags) | |
34dc7c2f BB |
2667 | { |
2668 | dsl_dir_t *dd; | |
13fe0198 | 2669 | dsl_pool_t *dp = spa_get_dsl(spa); |
b128c09f | 2670 | dsl_dataset_t *ds; |
34dc7c2f BB |
2671 | zap_cursor_t zc; |
2672 | zap_attribute_t *attr; | |
2673 | char *child; | |
b128c09f BB |
2674 | uint64_t thisobj; |
2675 | int err; | |
34dc7c2f | 2676 | |
13fe0198 MA |
2677 | dsl_pool_config_enter(dp, FTAG); |
2678 | ||
2679 | err = dsl_dir_hold(dp, name, FTAG, &dd, NULL); | |
2680 | if (err != 0) { | |
2681 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f | 2682 | return (err); |
13fe0198 | 2683 | } |
34dc7c2f | 2684 | |
b128c09f BB |
2685 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ |
2686 | if (dd->dd_myname[0] == '$') { | |
13fe0198 MA |
2687 | dsl_dir_rele(dd, FTAG); |
2688 | dsl_pool_config_exit(dp, FTAG); | |
b128c09f BB |
2689 | return (0); |
2690 | } | |
2691 | ||
d683ddbb | 2692 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 2693 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
34dc7c2f BB |
2694 | |
2695 | /* | |
2696 | * Iterate over all children. | |
2697 | */ | |
2698 | if (flags & DS_FIND_CHILDREN) { | |
b128c09f | 2699 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
d683ddbb | 2700 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
34dc7c2f BB |
2701 | zap_cursor_retrieve(&zc, attr) == 0; |
2702 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 MA |
2703 | ASSERT3U(attr->za_integer_length, ==, |
2704 | sizeof (uint64_t)); | |
2705 | ASSERT3U(attr->za_num_integers, ==, 1); | |
34dc7c2f | 2706 | |
428870ff | 2707 | child = kmem_asprintf("%s/%s", name, attr->za_name); |
13fe0198 MA |
2708 | dsl_pool_config_exit(dp, FTAG); |
2709 | err = dmu_objset_find_impl(spa, child, | |
2710 | func, arg, flags); | |
2711 | dsl_pool_config_enter(dp, FTAG); | |
428870ff | 2712 | strfree(child); |
13fe0198 | 2713 | if (err != 0) |
34dc7c2f BB |
2714 | break; |
2715 | } | |
2716 | zap_cursor_fini(&zc); | |
2717 | ||
13fe0198 MA |
2718 | if (err != 0) { |
2719 | dsl_dir_rele(dd, FTAG); | |
2720 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f BB |
2721 | kmem_free(attr, sizeof (zap_attribute_t)); |
2722 | return (err); | |
2723 | } | |
2724 | } | |
2725 | ||
2726 | /* | |
2727 | * Iterate over all snapshots. | |
2728 | */ | |
b128c09f | 2729 | if (flags & DS_FIND_SNAPSHOTS) { |
b128c09f | 2730 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); |
b128c09f BB |
2731 | |
2732 | if (err == 0) { | |
d683ddbb JG |
2733 | uint64_t snapobj; |
2734 | ||
2735 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
b128c09f BB |
2736 | dsl_dataset_rele(ds, FTAG); |
2737 | ||
2738 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
2739 | zap_cursor_retrieve(&zc, attr) == 0; | |
2740 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 | 2741 | ASSERT3U(attr->za_integer_length, ==, |
b128c09f | 2742 | sizeof (uint64_t)); |
13fe0198 | 2743 | ASSERT3U(attr->za_num_integers, ==, 1); |
b128c09f | 2744 | |
428870ff BB |
2745 | child = kmem_asprintf("%s@%s", |
2746 | name, attr->za_name); | |
13fe0198 MA |
2747 | dsl_pool_config_exit(dp, FTAG); |
2748 | err = func(child, arg); | |
2749 | dsl_pool_config_enter(dp, FTAG); | |
428870ff | 2750 | strfree(child); |
13fe0198 | 2751 | if (err != 0) |
b128c09f BB |
2752 | break; |
2753 | } | |
2754 | zap_cursor_fini(&zc); | |
34dc7c2f | 2755 | } |
34dc7c2f BB |
2756 | } |
2757 | ||
13fe0198 | 2758 | dsl_dir_rele(dd, FTAG); |
34dc7c2f | 2759 | kmem_free(attr, sizeof (zap_attribute_t)); |
13fe0198 | 2760 | dsl_pool_config_exit(dp, FTAG); |
34dc7c2f | 2761 | |
13fe0198 | 2762 | if (err != 0) |
34dc7c2f BB |
2763 | return (err); |
2764 | ||
13fe0198 MA |
2765 | /* Apply to self. */ |
2766 | return (func(name, arg)); | |
34dc7c2f BB |
2767 | } |
2768 | ||
13fe0198 MA |
2769 | /* |
2770 | * See comment above dmu_objset_find_impl(). | |
2771 | */ | |
d164b209 | 2772 | int |
13fe0198 MA |
2773 | dmu_objset_find(char *name, int func(const char *, void *), void *arg, |
2774 | int flags) | |
d164b209 | 2775 | { |
13fe0198 MA |
2776 | spa_t *spa; |
2777 | int error; | |
d164b209 | 2778 | |
13fe0198 MA |
2779 | error = spa_open(name, &spa, FTAG); |
2780 | if (error != 0) | |
2781 | return (error); | |
2782 | error = dmu_objset_find_impl(spa, name, func, arg, flags); | |
2783 | spa_close(spa, FTAG); | |
2784 | return (error); | |
d164b209 BB |
2785 | } |
2786 | ||
ae76f45c TC |
2787 | boolean_t |
2788 | dmu_objset_incompatible_encryption_version(objset_t *os) | |
2789 | { | |
2790 | return (dsl_dir_incompatible_encryption_version( | |
2791 | os->os_dsl_dataset->ds_dir)); | |
2792 | } | |
2793 | ||
34dc7c2f BB |
2794 | void |
2795 | dmu_objset_set_user(objset_t *os, void *user_ptr) | |
2796 | { | |
428870ff BB |
2797 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
2798 | os->os_user_ptr = user_ptr; | |
34dc7c2f BB |
2799 | } |
2800 | ||
2801 | void * | |
2802 | dmu_objset_get_user(objset_t *os) | |
2803 | { | |
428870ff BB |
2804 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
2805 | return (os->os_user_ptr); | |
34dc7c2f | 2806 | } |
c28b2279 | 2807 | |
13fe0198 MA |
2808 | /* |
2809 | * Determine name of filesystem, given name of snapshot. | |
eca7b760 | 2810 | * buf must be at least ZFS_MAX_DATASET_NAME_LEN bytes |
13fe0198 MA |
2811 | */ |
2812 | int | |
2813 | dmu_fsname(const char *snapname, char *buf) | |
2814 | { | |
2815 | char *atp = strchr(snapname, '@'); | |
2816 | if (atp == NULL) | |
2e528b49 | 2817 | return (SET_ERROR(EINVAL)); |
eca7b760 | 2818 | if (atp - snapname >= ZFS_MAX_DATASET_NAME_LEN) |
2e528b49 | 2819 | return (SET_ERROR(ENAMETOOLONG)); |
13fe0198 MA |
2820 | (void) strlcpy(buf, snapname, atp - snapname + 1); |
2821 | return (0); | |
2822 | } | |
2823 | ||
3ec3bc21 BB |
2824 | /* |
2825 | * Call when we think we're going to write/free space in open context to track | |
2826 | * the amount of dirty data in the open txg, which is also the amount | |
2827 | * of memory that can not be evicted until this txg syncs. | |
2828 | */ | |
2829 | void | |
2830 | dmu_objset_willuse_space(objset_t *os, int64_t space, dmu_tx_t *tx) | |
2831 | { | |
2832 | dsl_dataset_t *ds = os->os_dsl_dataset; | |
2833 | int64_t aspace = spa_get_worst_case_asize(os->os_spa, space); | |
2834 | ||
2835 | if (ds != NULL) { | |
2836 | dsl_dir_willuse_space(ds->ds_dir, aspace, tx); | |
2837 | dsl_pool_dirty_space(dmu_tx_pool(tx), space, tx); | |
2838 | } | |
2839 | } | |
2840 | ||
c28b2279 | 2841 | #if defined(_KERNEL) && defined(HAVE_SPL) |
f0fd83be | 2842 | EXPORT_SYMBOL(dmu_objset_zil); |
c28b2279 | 2843 | EXPORT_SYMBOL(dmu_objset_pool); |
f0fd83be BB |
2844 | EXPORT_SYMBOL(dmu_objset_ds); |
2845 | EXPORT_SYMBOL(dmu_objset_type); | |
c28b2279 BB |
2846 | EXPORT_SYMBOL(dmu_objset_name); |
2847 | EXPORT_SYMBOL(dmu_objset_hold); | |
b5256303 | 2848 | EXPORT_SYMBOL(dmu_objset_hold_flags); |
c28b2279 BB |
2849 | EXPORT_SYMBOL(dmu_objset_own); |
2850 | EXPORT_SYMBOL(dmu_objset_rele); | |
b5256303 | 2851 | EXPORT_SYMBOL(dmu_objset_rele_flags); |
c28b2279 BB |
2852 | EXPORT_SYMBOL(dmu_objset_disown); |
2853 | EXPORT_SYMBOL(dmu_objset_from_ds); | |
2854 | EXPORT_SYMBOL(dmu_objset_create); | |
2855 | EXPORT_SYMBOL(dmu_objset_clone); | |
c28b2279 BB |
2856 | EXPORT_SYMBOL(dmu_objset_stats); |
2857 | EXPORT_SYMBOL(dmu_objset_fast_stat); | |
54a179e7 | 2858 | EXPORT_SYMBOL(dmu_objset_spa); |
c28b2279 BB |
2859 | EXPORT_SYMBOL(dmu_objset_space); |
2860 | EXPORT_SYMBOL(dmu_objset_fsid_guid); | |
2861 | EXPORT_SYMBOL(dmu_objset_find); | |
c28b2279 BB |
2862 | EXPORT_SYMBOL(dmu_objset_byteswap); |
2863 | EXPORT_SYMBOL(dmu_objset_evict_dbufs); | |
2864 | EXPORT_SYMBOL(dmu_objset_snap_cmtime); | |
50c957f7 | 2865 | EXPORT_SYMBOL(dmu_objset_dnodesize); |
c28b2279 BB |
2866 | |
2867 | EXPORT_SYMBOL(dmu_objset_sync); | |
2868 | EXPORT_SYMBOL(dmu_objset_is_dirty); | |
b5256303 | 2869 | EXPORT_SYMBOL(dmu_objset_create_impl_dnstats); |
c28b2279 BB |
2870 | EXPORT_SYMBOL(dmu_objset_create_impl); |
2871 | EXPORT_SYMBOL(dmu_objset_open_impl); | |
2872 | EXPORT_SYMBOL(dmu_objset_evict); | |
2873 | EXPORT_SYMBOL(dmu_objset_register_type); | |
2874 | EXPORT_SYMBOL(dmu_objset_do_userquota_updates); | |
2875 | EXPORT_SYMBOL(dmu_objset_userquota_get_ids); | |
2876 | EXPORT_SYMBOL(dmu_objset_userused_enabled); | |
2877 | EXPORT_SYMBOL(dmu_objset_userspace_upgrade); | |
2878 | EXPORT_SYMBOL(dmu_objset_userspace_present); | |
1de321e6 | 2879 | EXPORT_SYMBOL(dmu_objset_userobjused_enabled); |
126ae9f4 | 2880 | EXPORT_SYMBOL(dmu_objset_userobjspace_upgradable); |
1de321e6 | 2881 | EXPORT_SYMBOL(dmu_objset_userobjspace_present); |
9c5167d1 NF |
2882 | EXPORT_SYMBOL(dmu_objset_projectquota_enabled); |
2883 | EXPORT_SYMBOL(dmu_objset_projectquota_present); | |
2884 | EXPORT_SYMBOL(dmu_objset_projectquota_upgradable); | |
2885 | EXPORT_SYMBOL(dmu_objset_id_quota_upgrade); | |
c28b2279 | 2886 | #endif |