]>
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 | |
1d3ba0bf | 9 | * or https://opensource.org/licenses/CDDL-1.0. |
34dc7c2f BB |
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. |
ba67d821 | 24 | * Copyright (c) 2012, 2020 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. |
d8d418ff | 32 | * Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved. |
10b3c7f5 MN |
33 | * Copyright (c) 2019, Klara Inc. |
34 | * Copyright (c) 2019, Allan Jude | |
5405be03 | 35 | * Copyright (c) 2022 Hewlett Packard Enterprise Development LP. |
34dc7c2f BB |
36 | */ |
37 | ||
428870ff BB |
38 | /* Portions Copyright 2010 Robert Milkowski */ |
39 | ||
34dc7c2f BB |
40 | #include <sys/cred.h> |
41 | #include <sys/zfs_context.h> | |
42 | #include <sys/dmu_objset.h> | |
43 | #include <sys/dsl_dir.h> | |
44 | #include <sys/dsl_dataset.h> | |
45 | #include <sys/dsl_prop.h> | |
46 | #include <sys/dsl_pool.h> | |
47 | #include <sys/dsl_synctask.h> | |
48 | #include <sys/dsl_deleg.h> | |
49 | #include <sys/dnode.h> | |
50 | #include <sys/dbuf.h> | |
51 | #include <sys/zvol.h> | |
52 | #include <sys/dmu_tx.h> | |
34dc7c2f BB |
53 | #include <sys/zap.h> |
54 | #include <sys/zil.h> | |
55 | #include <sys/dmu_impl.h> | |
56 | #include <sys/zfs_ioctl.h> | |
428870ff | 57 | #include <sys/sa.h> |
572e2857 | 58 | #include <sys/zfs_onexit.h> |
13fe0198 | 59 | #include <sys/dsl_destroy.h> |
9c43027b | 60 | #include <sys/vdev.h> |
a1d477c2 | 61 | #include <sys/zfeature.h> |
f74b821a | 62 | #include <sys/policy.h> |
1de321e6 | 63 | #include <sys/spa_impl.h> |
03916905 | 64 | #include <sys/dmu_recv.h> |
9c5167d1 | 65 | #include <sys/zfs_project.h> |
a7ed98d8 | 66 | #include "zfs_namecheck.h" |
c9d62d13 GA |
67 | #include <sys/vdev_impl.h> |
68 | #include <sys/arc.h> | |
572e2857 BB |
69 | |
70 | /* | |
71 | * Needed to close a window in dnode_move() that allows the objset to be freed | |
72 | * before it can be safely accessed. | |
73 | */ | |
74 | krwlock_t os_lock; | |
75 | ||
9c43027b | 76 | /* |
4e33ba4c | 77 | * Tunable to overwrite the maximum number of threads for the parallelization |
9c43027b AJ |
78 | * of dmu_objset_find_dp, needed to speed up the import of pools with many |
79 | * datasets. | |
80 | * Default is 4 times the number of leaf vdevs. | |
81 | */ | |
18168da7 | 82 | static const int dmu_find_threads = 0; |
9c43027b | 83 | |
68cbd56e NB |
84 | /* |
85 | * Backfill lower metadnode objects after this many have been freed. | |
86 | * Backfilling negatively impacts object creation rates, so only do it | |
87 | * if there are enough holes to fill. | |
88 | */ | |
18168da7 | 89 | static const int dmu_rescan_dnode_threshold = 1 << DN_MAX_INDBLKSHIFT; |
68cbd56e | 90 | |
18168da7 | 91 | static const char *upgrade_tag = "upgrade_tag"; |
c0daec32 | 92 | |
9c43027b AJ |
93 | static void dmu_objset_find_dp_cb(void *arg); |
94 | ||
1de321e6 JX |
95 | static void dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb); |
96 | static void dmu_objset_upgrade_stop(objset_t *os); | |
97 | ||
572e2857 BB |
98 | void |
99 | dmu_objset_init(void) | |
100 | { | |
101 | rw_init(&os_lock, NULL, RW_DEFAULT, NULL); | |
102 | } | |
103 | ||
104 | void | |
105 | dmu_objset_fini(void) | |
106 | { | |
107 | rw_destroy(&os_lock); | |
108 | } | |
34dc7c2f BB |
109 | |
110 | spa_t * | |
111 | dmu_objset_spa(objset_t *os) | |
112 | { | |
428870ff | 113 | return (os->os_spa); |
34dc7c2f BB |
114 | } |
115 | ||
116 | zilog_t * | |
117 | dmu_objset_zil(objset_t *os) | |
118 | { | |
428870ff | 119 | return (os->os_zil); |
34dc7c2f BB |
120 | } |
121 | ||
122 | dsl_pool_t * | |
123 | dmu_objset_pool(objset_t *os) | |
124 | { | |
125 | dsl_dataset_t *ds; | |
126 | ||
428870ff | 127 | if ((ds = os->os_dsl_dataset) != NULL && ds->ds_dir) |
34dc7c2f BB |
128 | return (ds->ds_dir->dd_pool); |
129 | else | |
428870ff | 130 | return (spa_get_dsl(os->os_spa)); |
34dc7c2f BB |
131 | } |
132 | ||
133 | dsl_dataset_t * | |
134 | dmu_objset_ds(objset_t *os) | |
135 | { | |
428870ff | 136 | return (os->os_dsl_dataset); |
34dc7c2f BB |
137 | } |
138 | ||
139 | dmu_objset_type_t | |
140 | dmu_objset_type(objset_t *os) | |
141 | { | |
428870ff | 142 | return (os->os_phys->os_type); |
34dc7c2f BB |
143 | } |
144 | ||
145 | void | |
146 | dmu_objset_name(objset_t *os, char *buf) | |
147 | { | |
428870ff | 148 | dsl_dataset_name(os->os_dsl_dataset, buf); |
34dc7c2f BB |
149 | } |
150 | ||
151 | uint64_t | |
152 | dmu_objset_id(objset_t *os) | |
153 | { | |
428870ff | 154 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
155 | |
156 | return (ds ? ds->ds_object : 0); | |
157 | } | |
158 | ||
50c957f7 NB |
159 | uint64_t |
160 | dmu_objset_dnodesize(objset_t *os) | |
161 | { | |
162 | return (os->os_dnodesize); | |
163 | } | |
164 | ||
faf0f58c | 165 | zfs_sync_type_t |
428870ff BB |
166 | dmu_objset_syncprop(objset_t *os) |
167 | { | |
168 | return (os->os_sync); | |
169 | } | |
170 | ||
faf0f58c | 171 | zfs_logbias_op_t |
428870ff BB |
172 | dmu_objset_logbias(objset_t *os) |
173 | { | |
174 | return (os->os_logbias); | |
175 | } | |
176 | ||
34dc7c2f BB |
177 | static void |
178 | checksum_changed_cb(void *arg, uint64_t newval) | |
179 | { | |
428870ff | 180 | objset_t *os = arg; |
34dc7c2f BB |
181 | |
182 | /* | |
183 | * Inheritance should have been done by now. | |
184 | */ | |
185 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
186 | ||
428870ff | 187 | os->os_checksum = zio_checksum_select(newval, ZIO_CHECKSUM_ON_VALUE); |
34dc7c2f BB |
188 | } |
189 | ||
190 | static void | |
191 | compression_changed_cb(void *arg, uint64_t newval) | |
192 | { | |
428870ff | 193 | objset_t *os = arg; |
34dc7c2f BB |
194 | |
195 | /* | |
196 | * Inheritance and range checking should have been done by now. | |
197 | */ | |
198 | ASSERT(newval != ZIO_COMPRESS_INHERIT); | |
199 | ||
10b3c7f5 MN |
200 | os->os_compress = zio_compress_select(os->os_spa, |
201 | ZIO_COMPRESS_ALGO(newval), ZIO_COMPRESS_ON); | |
202 | os->os_complevel = zio_complevel_select(os->os_spa, os->os_compress, | |
203 | ZIO_COMPRESS_LEVEL(newval), ZIO_COMPLEVEL_DEFAULT); | |
34dc7c2f BB |
204 | } |
205 | ||
206 | static void | |
207 | copies_changed_cb(void *arg, uint64_t newval) | |
208 | { | |
428870ff | 209 | objset_t *os = arg; |
34dc7c2f BB |
210 | |
211 | /* | |
212 | * Inheritance and range checking should have been done by now. | |
213 | */ | |
214 | ASSERT(newval > 0); | |
428870ff | 215 | ASSERT(newval <= spa_max_replication(os->os_spa)); |
34dc7c2f | 216 | |
428870ff BB |
217 | os->os_copies = newval; |
218 | } | |
219 | ||
220 | static void | |
221 | dedup_changed_cb(void *arg, uint64_t newval) | |
222 | { | |
223 | objset_t *os = arg; | |
224 | spa_t *spa = os->os_spa; | |
225 | enum zio_checksum checksum; | |
226 | ||
227 | /* | |
228 | * Inheritance should have been done by now. | |
229 | */ | |
230 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
231 | ||
232 | checksum = zio_checksum_dedup_select(spa, newval, ZIO_CHECKSUM_OFF); | |
233 | ||
234 | os->os_dedup_checksum = checksum & ZIO_CHECKSUM_MASK; | |
235 | os->os_dedup_verify = !!(checksum & ZIO_CHECKSUM_VERIFY); | |
34dc7c2f BB |
236 | } |
237 | ||
b128c09f BB |
238 | static void |
239 | primary_cache_changed_cb(void *arg, uint64_t newval) | |
240 | { | |
428870ff | 241 | objset_t *os = arg; |
b128c09f BB |
242 | |
243 | /* | |
244 | * Inheritance and range checking should have been done by now. | |
245 | */ | |
246 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
247 | newval == ZFS_CACHE_METADATA); | |
248 | ||
428870ff | 249 | os->os_primary_cache = newval; |
b128c09f BB |
250 | } |
251 | ||
252 | static void | |
253 | secondary_cache_changed_cb(void *arg, uint64_t newval) | |
254 | { | |
428870ff | 255 | objset_t *os = arg; |
b128c09f BB |
256 | |
257 | /* | |
258 | * Inheritance and range checking should have been done by now. | |
259 | */ | |
260 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
261 | newval == ZFS_CACHE_METADATA); | |
262 | ||
428870ff BB |
263 | os->os_secondary_cache = newval; |
264 | } | |
265 | ||
07345ac2 BB |
266 | static void |
267 | prefetch_changed_cb(void *arg, uint64_t newval) | |
268 | { | |
269 | objset_t *os = arg; | |
270 | ||
271 | /* | |
272 | * Inheritance should have been done by now. | |
273 | */ | |
274 | ASSERT(newval == ZFS_PREFETCH_ALL || newval == ZFS_PREFETCH_NONE || | |
275 | newval == ZFS_PREFETCH_METADATA); | |
276 | os->os_prefetch = newval; | |
277 | } | |
278 | ||
428870ff BB |
279 | static void |
280 | sync_changed_cb(void *arg, uint64_t newval) | |
281 | { | |
282 | objset_t *os = arg; | |
283 | ||
284 | /* | |
285 | * Inheritance and range checking should have been done by now. | |
286 | */ | |
287 | ASSERT(newval == ZFS_SYNC_STANDARD || newval == ZFS_SYNC_ALWAYS || | |
288 | newval == ZFS_SYNC_DISABLED); | |
289 | ||
290 | os->os_sync = newval; | |
291 | if (os->os_zil) | |
292 | zil_set_sync(os->os_zil, newval); | |
293 | } | |
294 | ||
faf0f58c MA |
295 | static void |
296 | redundant_metadata_changed_cb(void *arg, uint64_t newval) | |
297 | { | |
298 | objset_t *os = arg; | |
299 | ||
300 | /* | |
301 | * Inheritance and range checking should have been done by now. | |
302 | */ | |
303 | ASSERT(newval == ZFS_REDUNDANT_METADATA_ALL || | |
5405be03 A |
304 | newval == ZFS_REDUNDANT_METADATA_MOST || |
305 | newval == ZFS_REDUNDANT_METADATA_SOME || | |
306 | newval == ZFS_REDUNDANT_METADATA_NONE); | |
faf0f58c MA |
307 | |
308 | os->os_redundant_metadata = newval; | |
309 | } | |
310 | ||
50c957f7 NB |
311 | static void |
312 | dnodesize_changed_cb(void *arg, uint64_t newval) | |
313 | { | |
314 | objset_t *os = arg; | |
315 | ||
316 | switch (newval) { | |
317 | case ZFS_DNSIZE_LEGACY: | |
318 | os->os_dnodesize = DNODE_MIN_SIZE; | |
319 | break; | |
320 | case ZFS_DNSIZE_AUTO: | |
321 | /* | |
322 | * Choose a dnode size that will work well for most | |
323 | * workloads if the user specified "auto". Future code | |
324 | * improvements could dynamically select a dnode size | |
325 | * based on observed workload patterns. | |
326 | */ | |
327 | os->os_dnodesize = DNODE_MIN_SIZE * 2; | |
328 | break; | |
329 | case ZFS_DNSIZE_1K: | |
330 | case ZFS_DNSIZE_2K: | |
331 | case ZFS_DNSIZE_4K: | |
332 | case ZFS_DNSIZE_8K: | |
333 | case ZFS_DNSIZE_16K: | |
334 | os->os_dnodesize = newval; | |
335 | break; | |
336 | } | |
337 | } | |
338 | ||
cc99f275 DB |
339 | static void |
340 | smallblk_changed_cb(void *arg, uint64_t newval) | |
341 | { | |
342 | objset_t *os = arg; | |
343 | ||
344 | /* | |
345 | * Inheritance and range checking should have been done by now. | |
346 | */ | |
5aa69a57 | 347 | ASSERT(newval <= SPA_MAXBLOCKSIZE); |
cc99f275 DB |
348 | ASSERT(ISP2(newval)); |
349 | ||
350 | os->os_zpl_special_smallblock = newval; | |
351 | } | |
352 | ||
428870ff BB |
353 | static void |
354 | logbias_changed_cb(void *arg, uint64_t newval) | |
355 | { | |
356 | objset_t *os = arg; | |
357 | ||
358 | ASSERT(newval == ZFS_LOGBIAS_LATENCY || | |
359 | newval == ZFS_LOGBIAS_THROUGHPUT); | |
360 | os->os_logbias = newval; | |
361 | if (os->os_zil) | |
362 | zil_set_logbias(os->os_zil, newval); | |
b128c09f BB |
363 | } |
364 | ||
f1512ee6 MA |
365 | static void |
366 | recordsize_changed_cb(void *arg, uint64_t newval) | |
367 | { | |
368 | objset_t *os = arg; | |
369 | ||
370 | os->os_recordsize = newval; | |
371 | } | |
372 | ||
34dc7c2f BB |
373 | void |
374 | dmu_objset_byteswap(void *buf, size_t size) | |
375 | { | |
376 | objset_phys_t *osp = buf; | |
377 | ||
9c5167d1 NF |
378 | ASSERT(size == OBJSET_PHYS_SIZE_V1 || size == OBJSET_PHYS_SIZE_V2 || |
379 | size == sizeof (objset_phys_t)); | |
34dc7c2f BB |
380 | dnode_byteswap(&osp->os_meta_dnode); |
381 | byteswap_uint64_array(&osp->os_zil_header, sizeof (zil_header_t)); | |
382 | osp->os_type = BSWAP_64(osp->os_type); | |
9babb374 | 383 | osp->os_flags = BSWAP_64(osp->os_flags); |
9c5167d1 | 384 | if (size >= OBJSET_PHYS_SIZE_V2) { |
9babb374 BB |
385 | dnode_byteswap(&osp->os_userused_dnode); |
386 | dnode_byteswap(&osp->os_groupused_dnode); | |
9c5167d1 NF |
387 | if (size >= sizeof (objset_phys_t)) |
388 | dnode_byteswap(&osp->os_projectused_dnode); | |
9babb374 | 389 | } |
34dc7c2f BB |
390 | } |
391 | ||
64fc7762 MA |
392 | /* |
393 | * The hash is a CRC-based hash of the objset_t pointer and the object number. | |
394 | */ | |
395 | static uint64_t | |
396 | dnode_hash(const objset_t *os, uint64_t obj) | |
397 | { | |
398 | uintptr_t osv = (uintptr_t)os; | |
399 | uint64_t crc = -1ULL; | |
400 | ||
401 | ASSERT(zfs_crc64_table[128] == ZFS_CRC64_POLY); | |
402 | /* | |
8fd3a5d0 AM |
403 | * The lower 11 bits of the pointer don't have much entropy, because |
404 | * the objset_t is more than 1KB long and so likely aligned to 2KB. | |
64fc7762 | 405 | */ |
8fd3a5d0 | 406 | crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (osv >> 11)) & 0xFF]; |
64fc7762 MA |
407 | crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 0)) & 0xFF]; |
408 | crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 8)) & 0xFF]; | |
409 | crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 16)) & 0xFF]; | |
410 | ||
411 | crc ^= (osv>>14) ^ (obj>>24); | |
412 | ||
413 | return (crc); | |
414 | } | |
415 | ||
65c7cc49 | 416 | static unsigned int |
64fc7762 MA |
417 | dnode_multilist_index_func(multilist_t *ml, void *obj) |
418 | { | |
419 | dnode_t *dn = obj; | |
5b7053a9 AM |
420 | |
421 | /* | |
422 | * The low order bits of the hash value are thought to be | |
423 | * distributed evenly. Otherwise, in the case that the multilist | |
424 | * has a power of two number of sublists, each sublists' usage | |
425 | * would not be evenly distributed. In this context full 64bit | |
426 | * division would be a waste of time, so limit it to 32 bits. | |
427 | */ | |
428 | return ((unsigned int)dnode_hash(dn->dn_objset, dn->dn_object) % | |
64fc7762 MA |
429 | multilist_get_num_sublists(ml)); |
430 | } | |
431 | ||
c9d62d13 GA |
432 | static inline boolean_t |
433 | dmu_os_is_l2cacheable(objset_t *os) | |
434 | { | |
ca9e32d3 GA |
435 | if (os->os_secondary_cache == ZFS_CACHE_ALL || |
436 | os->os_secondary_cache == ZFS_CACHE_METADATA) { | |
437 | if (l2arc_exclude_special == 0) | |
438 | return (B_TRUE); | |
439 | ||
440 | blkptr_t *bp = os->os_rootbp; | |
441 | if (bp == NULL || BP_IS_HOLE(bp)) | |
442 | return (B_FALSE); | |
c9d62d13 GA |
443 | uint64_t vdev = DVA_GET_VDEV(bp->blk_dva); |
444 | vdev_t *rvd = os->os_spa->spa_root_vdev; | |
ca9e32d3 | 445 | vdev_t *vd = NULL; |
c9d62d13 GA |
446 | |
447 | if (vdev < rvd->vdev_children) | |
448 | vd = rvd->vdev_child[vdev]; | |
449 | ||
ca9e32d3 GA |
450 | if (vd == NULL) |
451 | return (B_TRUE); | |
c9d62d13 | 452 | |
ca9e32d3 GA |
453 | if (vd->vdev_alloc_bias != VDEV_BIAS_SPECIAL && |
454 | vd->vdev_alloc_bias != VDEV_BIAS_DEDUP) | |
455 | return (B_TRUE); | |
c9d62d13 | 456 | } |
c9d62d13 GA |
457 | return (B_FALSE); |
458 | } | |
459 | ||
a1d477c2 MA |
460 | /* |
461 | * Instantiates the objset_t in-memory structure corresponding to the | |
462 | * objset_phys_t that's pointed to by the specified blkptr_t. | |
463 | */ | |
34dc7c2f BB |
464 | int |
465 | dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, | |
428870ff | 466 | objset_t **osp) |
34dc7c2f | 467 | { |
428870ff | 468 | objset_t *os; |
b128c09f | 469 | int i, err; |
34dc7c2f BB |
470 | |
471 | ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock)); | |
30af21b0 | 472 | ASSERT(!BP_IS_REDACTED(bp)); |
34dc7c2f | 473 | |
0fdd6106 MA |
474 | /* |
475 | * We need the pool config lock to get properties. | |
476 | */ | |
477 | ASSERT(ds == NULL || dsl_pool_config_held(ds->ds_dir->dd_pool)); | |
478 | ||
a1d477c2 MA |
479 | /* |
480 | * The $ORIGIN dataset (if it exists) doesn't have an associated | |
481 | * objset, so there's no reason to open it. The $ORIGIN dataset | |
482 | * will not exist on pools older than SPA_VERSION_ORIGIN. | |
483 | */ | |
484 | if (ds != NULL && spa_get_dsl(spa) != NULL && | |
485 | spa_get_dsl(spa)->dp_origin_snap != NULL) { | |
486 | ASSERT3P(ds->ds_dir, !=, | |
487 | spa_get_dsl(spa)->dp_origin_snap->ds_dir); | |
488 | } | |
489 | ||
79c76d5b | 490 | os = kmem_zalloc(sizeof (objset_t), KM_SLEEP); |
428870ff BB |
491 | os->os_dsl_dataset = ds; |
492 | os->os_spa = spa; | |
493 | os->os_rootbp = bp; | |
494 | if (!BP_IS_HOLE(os->os_rootbp)) { | |
2a432414 | 495 | arc_flags_t aflags = ARC_FLAG_WAIT; |
5dbd68a3 | 496 | zbookmark_phys_t zb; |
9c5167d1 | 497 | int size; |
4938d01d | 498 | zio_flag_t zio_flags = ZIO_FLAG_CANFAIL; |
428870ff BB |
499 | SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET, |
500 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
501 | ||
c9d62d13 | 502 | if (dmu_os_is_l2cacheable(os)) |
2a432414 | 503 | aflags |= ARC_FLAG_L2CACHE; |
34dc7c2f | 504 | |
b5256303 TC |
505 | if (ds != NULL && ds->ds_dir->dd_crypto_obj != 0) { |
506 | ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF); | |
507 | ASSERT(BP_IS_AUTHENTICATED(bp)); | |
508 | zio_flags |= ZIO_FLAG_RAW; | |
509 | } | |
510 | ||
428870ff | 511 | dprintf_bp(os->os_rootbp, "reading %s", ""); |
294f6806 | 512 | err = arc_read(NULL, spa, os->os_rootbp, |
428870ff | 513 | arc_getbuf_func, &os->os_phys_buf, |
b5256303 | 514 | ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zb); |
13fe0198 | 515 | if (err != 0) { |
428870ff | 516 | kmem_free(os, sizeof (objset_t)); |
b128c09f BB |
517 | /* convert checksum errors into IO errors */ |
518 | if (err == ECKSUM) | |
2e528b49 | 519 | err = SET_ERROR(EIO); |
34dc7c2f BB |
520 | return (err); |
521 | } | |
9babb374 | 522 | |
9c5167d1 NF |
523 | if (spa_version(spa) < SPA_VERSION_USERSPACE) |
524 | size = OBJSET_PHYS_SIZE_V1; | |
525 | else if (!spa_feature_is_enabled(spa, | |
526 | SPA_FEATURE_PROJECT_QUOTA)) | |
527 | size = OBJSET_PHYS_SIZE_V2; | |
528 | else | |
529 | size = sizeof (objset_phys_t); | |
530 | ||
9babb374 | 531 | /* Increase the blocksize if we are permitted. */ |
9c5167d1 | 532 | if (arc_buf_size(os->os_phys_buf) < size) { |
2aa34383 | 533 | arc_buf_t *buf = arc_alloc_buf(spa, &os->os_phys_buf, |
9c5167d1 | 534 | ARC_BUFC_METADATA, size); |
861166b0 AZ |
535 | memset(buf->b_data, 0, size); |
536 | memcpy(buf->b_data, os->os_phys_buf->b_data, | |
428870ff | 537 | arc_buf_size(os->os_phys_buf)); |
d3c2ae1c | 538 | arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); |
428870ff | 539 | os->os_phys_buf = buf; |
9babb374 BB |
540 | } |
541 | ||
428870ff BB |
542 | os->os_phys = os->os_phys_buf->b_data; |
543 | os->os_flags = os->os_phys->os_flags; | |
34dc7c2f | 544 | } else { |
9babb374 | 545 | int size = spa_version(spa) >= SPA_VERSION_USERSPACE ? |
9c5167d1 | 546 | sizeof (objset_phys_t) : OBJSET_PHYS_SIZE_V1; |
2aa34383 DK |
547 | os->os_phys_buf = arc_alloc_buf(spa, &os->os_phys_buf, |
548 | ARC_BUFC_METADATA, size); | |
428870ff | 549 | os->os_phys = os->os_phys_buf->b_data; |
861166b0 | 550 | memset(os->os_phys, 0, size); |
34dc7c2f | 551 | } |
2e5dc449 MA |
552 | /* |
553 | * These properties will be filled in by the logic in zfs_get_zplprop() | |
554 | * when they are queried for the first time. | |
555 | */ | |
556 | os->os_version = OBJSET_PROP_UNINITIALIZED; | |
557 | os->os_normalization = OBJSET_PROP_UNINITIALIZED; | |
558 | os->os_utf8only = OBJSET_PROP_UNINITIALIZED; | |
559 | os->os_casesensitivity = OBJSET_PROP_UNINITIALIZED; | |
34dc7c2f BB |
560 | |
561 | /* | |
562 | * Note: the changed_cb will be called once before the register | |
563 | * func returns, thus changing the checksum/compression from the | |
b128c09f BB |
564 | * default (fletcher2/off). Snapshots don't need to know about |
565 | * checksum/compression/copies. | |
34dc7c2f | 566 | */ |
9b67f605 | 567 | if (ds != NULL) { |
b5256303 TC |
568 | os->os_encrypted = (ds->ds_dir->dd_crypto_obj != 0); |
569 | ||
13fe0198 MA |
570 | err = dsl_prop_register(ds, |
571 | zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE), | |
428870ff | 572 | primary_cache_changed_cb, os); |
13fe0198 MA |
573 | if (err == 0) { |
574 | err = dsl_prop_register(ds, | |
575 | zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE), | |
428870ff | 576 | secondary_cache_changed_cb, os); |
13fe0198 | 577 | } |
07345ac2 BB |
578 | if (err == 0) { |
579 | err = dsl_prop_register(ds, | |
580 | zfs_prop_to_name(ZFS_PROP_PREFETCH), | |
581 | prefetch_changed_cb, os); | |
582 | } | |
0c66c32d | 583 | if (!ds->ds_is_snapshot) { |
13fe0198 MA |
584 | if (err == 0) { |
585 | err = dsl_prop_register(ds, | |
586 | zfs_prop_to_name(ZFS_PROP_CHECKSUM), | |
428870ff | 587 | checksum_changed_cb, os); |
13fe0198 MA |
588 | } |
589 | if (err == 0) { | |
590 | err = dsl_prop_register(ds, | |
591 | zfs_prop_to_name(ZFS_PROP_COMPRESSION), | |
428870ff | 592 | compression_changed_cb, os); |
13fe0198 MA |
593 | } |
594 | if (err == 0) { | |
595 | err = dsl_prop_register(ds, | |
596 | zfs_prop_to_name(ZFS_PROP_COPIES), | |
428870ff | 597 | copies_changed_cb, os); |
13fe0198 MA |
598 | } |
599 | if (err == 0) { | |
600 | err = dsl_prop_register(ds, | |
601 | zfs_prop_to_name(ZFS_PROP_DEDUP), | |
428870ff | 602 | dedup_changed_cb, os); |
13fe0198 MA |
603 | } |
604 | if (err == 0) { | |
605 | err = dsl_prop_register(ds, | |
606 | zfs_prop_to_name(ZFS_PROP_LOGBIAS), | |
428870ff | 607 | logbias_changed_cb, os); |
13fe0198 MA |
608 | } |
609 | if (err == 0) { | |
610 | err = dsl_prop_register(ds, | |
611 | zfs_prop_to_name(ZFS_PROP_SYNC), | |
428870ff | 612 | sync_changed_cb, os); |
13fe0198 | 613 | } |
faf0f58c MA |
614 | if (err == 0) { |
615 | err = dsl_prop_register(ds, | |
616 | zfs_prop_to_name( | |
617 | ZFS_PROP_REDUNDANT_METADATA), | |
618 | redundant_metadata_changed_cb, os); | |
619 | } | |
f1512ee6 MA |
620 | if (err == 0) { |
621 | err = dsl_prop_register(ds, | |
622 | zfs_prop_to_name(ZFS_PROP_RECORDSIZE), | |
623 | recordsize_changed_cb, os); | |
624 | } | |
50c957f7 NB |
625 | if (err == 0) { |
626 | err = dsl_prop_register(ds, | |
627 | zfs_prop_to_name(ZFS_PROP_DNODESIZE), | |
628 | dnodesize_changed_cb, os); | |
629 | } | |
cc99f275 DB |
630 | if (err == 0) { |
631 | err = dsl_prop_register(ds, | |
632 | zfs_prop_to_name( | |
633 | ZFS_PROP_SPECIAL_SMALL_BLOCKS), | |
634 | smallblk_changed_cb, os); | |
635 | } | |
b128c09f | 636 | } |
13fe0198 | 637 | if (err != 0) { |
d3c2ae1c | 638 | arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); |
428870ff | 639 | kmem_free(os, sizeof (objset_t)); |
34dc7c2f BB |
640 | return (err); |
641 | } | |
9b67f605 | 642 | } else { |
34dc7c2f | 643 | /* It's the meta-objset. */ |
428870ff | 644 | os->os_checksum = ZIO_CHECKSUM_FLETCHER_4; |
99197f03 | 645 | os->os_compress = ZIO_COMPRESS_ON; |
10b3c7f5 | 646 | os->os_complevel = ZIO_COMPLEVEL_DEFAULT; |
b5256303 | 647 | os->os_encrypted = B_FALSE; |
428870ff BB |
648 | os->os_copies = spa_max_replication(spa); |
649 | os->os_dedup_checksum = ZIO_CHECKSUM_OFF; | |
faf0f58c MA |
650 | os->os_dedup_verify = B_FALSE; |
651 | os->os_logbias = ZFS_LOGBIAS_LATENCY; | |
652 | os->os_sync = ZFS_SYNC_STANDARD; | |
428870ff BB |
653 | os->os_primary_cache = ZFS_CACHE_ALL; |
654 | os->os_secondary_cache = ZFS_CACHE_ALL; | |
50c957f7 | 655 | os->os_dnodesize = DNODE_MIN_SIZE; |
07345ac2 | 656 | os->os_prefetch = ZFS_PREFETCH_ALL; |
34dc7c2f BB |
657 | } |
658 | ||
0c66c32d | 659 | if (ds == NULL || !ds->ds_is_snapshot) |
572e2857 | 660 | os->os_zil_header = os->os_phys->os_zil_header; |
428870ff | 661 | os->os_zil = zil_alloc(os, &os->os_zil_header); |
34dc7c2f BB |
662 | |
663 | for (i = 0; i < TXG_SIZE; i++) { | |
ffdf019c | 664 | multilist_create(&os->os_dirty_dnodes[i], sizeof (dnode_t), |
64fc7762 MA |
665 | offsetof(dnode_t, dn_dirty_link[i]), |
666 | dnode_multilist_index_func); | |
34dc7c2f | 667 | } |
428870ff | 668 | list_create(&os->os_dnodes, sizeof (dnode_t), |
34dc7c2f | 669 | offsetof(dnode_t, dn_link)); |
428870ff | 670 | list_create(&os->os_downgraded_dbufs, sizeof (dmu_buf_impl_t), |
34dc7c2f BB |
671 | offsetof(dmu_buf_impl_t, db_link)); |
672 | ||
0c66c32d JG |
673 | list_link_init(&os->os_evicting_node); |
674 | ||
428870ff | 675 | mutex_init(&os->os_lock, NULL, MUTEX_DEFAULT, NULL); |
64fc7762 | 676 | mutex_init(&os->os_userused_lock, NULL, MUTEX_DEFAULT, NULL); |
428870ff BB |
677 | mutex_init(&os->os_obj_lock, NULL, MUTEX_DEFAULT, NULL); |
678 | mutex_init(&os->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL); | |
dbeb8796 MA |
679 | os->os_obj_next_percpu_len = boot_ncpus; |
680 | os->os_obj_next_percpu = kmem_zalloc(os->os_obj_next_percpu_len * | |
681 | sizeof (os->os_obj_next_percpu[0]), KM_SLEEP); | |
428870ff | 682 | |
0c66c32d JG |
683 | dnode_special_open(os, &os->os_phys->os_meta_dnode, |
684 | DMU_META_DNODE_OBJECT, &os->os_meta_dnode); | |
9c5167d1 | 685 | if (OBJSET_BUF_HAS_USERUSED(os->os_phys_buf)) { |
0c66c32d JG |
686 | dnode_special_open(os, &os->os_phys->os_userused_dnode, |
687 | DMU_USERUSED_OBJECT, &os->os_userused_dnode); | |
688 | dnode_special_open(os, &os->os_phys->os_groupused_dnode, | |
689 | DMU_GROUPUSED_OBJECT, &os->os_groupused_dnode); | |
9c5167d1 NF |
690 | if (OBJSET_BUF_HAS_PROJECTUSED(os->os_phys_buf)) |
691 | dnode_special_open(os, | |
692 | &os->os_phys->os_projectused_dnode, | |
693 | DMU_PROJECTUSED_OBJECT, &os->os_projectused_dnode); | |
9babb374 | 694 | } |
34dc7c2f | 695 | |
1de321e6 JX |
696 | mutex_init(&os->os_upgrade_lock, NULL, MUTEX_DEFAULT, NULL); |
697 | ||
428870ff | 698 | *osp = os; |
34dc7c2f BB |
699 | return (0); |
700 | } | |
701 | ||
428870ff BB |
702 | int |
703 | dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp) | |
34dc7c2f | 704 | { |
428870ff | 705 | int err = 0; |
34dc7c2f | 706 | |
47dfff3b | 707 | /* |
0fdd6106 MA |
708 | * We need the pool_config lock to manipulate the dsl_dataset_t. |
709 | * Even if the dataset is long-held, we need the pool_config lock | |
710 | * to open the objset, as it needs to get properties. | |
47dfff3b | 711 | */ |
0fdd6106 | 712 | ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool)); |
47dfff3b | 713 | |
34dc7c2f | 714 | mutex_enter(&ds->ds_opening_lock); |
9b67f605 MA |
715 | if (ds->ds_objset == NULL) { |
716 | objset_t *os; | |
cc9bb3e5 | 717 | rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); |
34dc7c2f | 718 | err = dmu_objset_open_impl(dsl_dataset_get_spa(ds), |
9b67f605 | 719 | ds, dsl_dataset_get_blkptr(ds), &os); |
cc9bb3e5 | 720 | rrw_exit(&ds->ds_bp_rwlock, FTAG); |
9b67f605 MA |
721 | |
722 | if (err == 0) { | |
723 | mutex_enter(&ds->ds_lock); | |
724 | ASSERT(ds->ds_objset == NULL); | |
725 | ds->ds_objset = os; | |
726 | mutex_exit(&ds->ds_lock); | |
727 | } | |
34dc7c2f | 728 | } |
9b67f605 | 729 | *osp = ds->ds_objset; |
34dc7c2f | 730 | mutex_exit(&ds->ds_opening_lock); |
428870ff | 731 | return (err); |
34dc7c2f BB |
732 | } |
733 | ||
13fe0198 MA |
734 | /* |
735 | * Holds the pool while the objset is held. Therefore only one objset | |
736 | * can be held at a time. | |
737 | */ | |
34dc7c2f | 738 | int |
a926aab9 | 739 | dmu_objset_hold_flags(const char *name, boolean_t decrypt, const void *tag, |
b5256303 | 740 | objset_t **osp) |
34dc7c2f | 741 | { |
13fe0198 | 742 | dsl_pool_t *dp; |
428870ff | 743 | dsl_dataset_t *ds; |
34dc7c2f | 744 | int err; |
40ab927a | 745 | ds_hold_flags_t flags; |
34dc7c2f | 746 | |
40ab927a | 747 | flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE; |
13fe0198 MA |
748 | err = dsl_pool_hold(name, tag, &dp); |
749 | if (err != 0) | |
750 | return (err); | |
b5256303 | 751 | err = dsl_dataset_hold_flags(dp, name, flags, tag, &ds); |
13fe0198 MA |
752 | if (err != 0) { |
753 | dsl_pool_rele(dp, tag); | |
428870ff | 754 | return (err); |
13fe0198 | 755 | } |
428870ff BB |
756 | |
757 | err = dmu_objset_from_ds(ds, osp); | |
13fe0198 | 758 | if (err != 0) { |
428870ff | 759 | dsl_dataset_rele(ds, tag); |
13fe0198 MA |
760 | dsl_pool_rele(dp, tag); |
761 | } | |
428870ff | 762 | |
34dc7c2f BB |
763 | return (err); |
764 | } | |
765 | ||
b5256303 | 766 | int |
a926aab9 | 767 | dmu_objset_hold(const char *name, const void *tag, objset_t **osp) |
b5256303 TC |
768 | { |
769 | return (dmu_objset_hold_flags(name, B_FALSE, tag, osp)); | |
770 | } | |
771 | ||
9c43027b AJ |
772 | static int |
773 | dmu_objset_own_impl(dsl_dataset_t *ds, dmu_objset_type_t type, | |
a926aab9 | 774 | boolean_t readonly, boolean_t decrypt, const void *tag, objset_t **osp) |
9c43027b | 775 | { |
14e4e3cb | 776 | (void) tag; |
9c43027b | 777 | |
14e4e3cb | 778 | int err = dmu_objset_from_ds(ds, osp); |
9c43027b | 779 | if (err != 0) { |
b5256303 | 780 | return (err); |
9c43027b | 781 | } else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) { |
9c43027b AJ |
782 | return (SET_ERROR(EINVAL)); |
783 | } else if (!readonly && dsl_dataset_is_snapshot(ds)) { | |
9c43027b | 784 | return (SET_ERROR(EROFS)); |
ae76f45c TC |
785 | } else if (!readonly && decrypt && |
786 | dsl_dir_incompatible_encryption_version(ds->ds_dir)) { | |
787 | return (SET_ERROR(EROFS)); | |
9c43027b | 788 | } |
b5256303 TC |
789 | |
790 | /* if we are decrypting, we can now check MACs in os->os_phys_buf */ | |
791 | if (decrypt && arc_is_unauthenticated((*osp)->os_phys_buf)) { | |
a2c2ed1b TC |
792 | zbookmark_phys_t zb; |
793 | ||
794 | SET_BOOKMARK(&zb, ds->ds_object, ZB_ROOT_OBJECT, | |
795 | ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
b5256303 | 796 | err = arc_untransform((*osp)->os_phys_buf, (*osp)->os_spa, |
a2c2ed1b | 797 | &zb, B_FALSE); |
b5256303 TC |
798 | if (err != 0) |
799 | return (err); | |
800 | ||
801 | ASSERT0(arc_is_unauthenticated((*osp)->os_phys_buf)); | |
802 | } | |
803 | ||
804 | return (0); | |
9c43027b AJ |
805 | } |
806 | ||
13fe0198 MA |
807 | /* |
808 | * dsl_pool must not be held when this is called. | |
809 | * Upon successful return, there will be a longhold on the dataset, | |
810 | * and the dsl_pool will not be held. | |
811 | */ | |
34dc7c2f | 812 | int |
428870ff | 813 | dmu_objset_own(const char *name, dmu_objset_type_t type, |
a926aab9 | 814 | boolean_t readonly, boolean_t decrypt, const void *tag, objset_t **osp) |
34dc7c2f | 815 | { |
13fe0198 | 816 | dsl_pool_t *dp; |
34dc7c2f BB |
817 | dsl_dataset_t *ds; |
818 | int err; | |
40ab927a | 819 | ds_hold_flags_t flags; |
34dc7c2f | 820 | |
40ab927a | 821 | flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE; |
13fe0198 MA |
822 | err = dsl_pool_hold(name, FTAG, &dp); |
823 | if (err != 0) | |
824 | return (err); | |
b5256303 | 825 | err = dsl_dataset_own(dp, name, flags, tag, &ds); |
13fe0198 MA |
826 | if (err != 0) { |
827 | dsl_pool_rele(dp, FTAG); | |
34dc7c2f | 828 | return (err); |
13fe0198 | 829 | } |
b5256303 TC |
830 | err = dmu_objset_own_impl(ds, type, readonly, decrypt, tag, osp); |
831 | if (err != 0) { | |
832 | dsl_dataset_disown(ds, flags, tag); | |
833 | dsl_pool_rele(dp, FTAG); | |
834 | return (err); | |
835 | } | |
836 | ||
163a8c28 TC |
837 | /* |
838 | * User accounting requires the dataset to be decrypted and rw. | |
839 | * We also don't begin user accounting during claiming to help | |
840 | * speed up pool import times and to keep this txg reserved | |
841 | * completely for recovery work. | |
842 | */ | |
4072f465 | 843 | if (!readonly && !dp->dp_spa->spa_claiming && |
844 | (ds->ds_dir->dd_crypto_obj == 0 || decrypt)) { | |
845 | if (dmu_objset_userobjspace_upgradable(*osp) || | |
846 | dmu_objset_projectquota_upgradable(*osp)) { | |
847 | dmu_objset_id_quota_upgrade(*osp); | |
848 | } else if (dmu_objset_userused_enabled(*osp)) { | |
849 | dmu_objset_userspace_upgrade(*osp); | |
850 | } | |
851 | } | |
1de321e6 | 852 | |
c0daec32 | 853 | dsl_pool_rele(dp, FTAG); |
b5256303 | 854 | return (0); |
34dc7c2f BB |
855 | } |
856 | ||
9c43027b AJ |
857 | int |
858 | dmu_objset_own_obj(dsl_pool_t *dp, uint64_t obj, dmu_objset_type_t type, | |
a926aab9 | 859 | boolean_t readonly, boolean_t decrypt, const void *tag, objset_t **osp) |
9c43027b AJ |
860 | { |
861 | dsl_dataset_t *ds; | |
862 | int err; | |
40ab927a | 863 | ds_hold_flags_t flags; |
9c43027b | 864 | |
40ab927a | 865 | flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE; |
b5256303 | 866 | err = dsl_dataset_own_obj(dp, obj, flags, tag, &ds); |
9c43027b AJ |
867 | if (err != 0) |
868 | return (err); | |
869 | ||
b5256303 TC |
870 | err = dmu_objset_own_impl(ds, type, readonly, decrypt, tag, osp); |
871 | if (err != 0) { | |
872 | dsl_dataset_disown(ds, flags, tag); | |
873 | return (err); | |
874 | } | |
875 | ||
876 | return (0); | |
9c43027b AJ |
877 | } |
878 | ||
34dc7c2f | 879 | void |
a926aab9 | 880 | dmu_objset_rele_flags(objset_t *os, boolean_t decrypt, const void *tag) |
34dc7c2f | 881 | { |
40ab927a | 882 | ds_hold_flags_t flags; |
13fe0198 | 883 | dsl_pool_t *dp = dmu_objset_pool(os); |
40ab927a TS |
884 | |
885 | flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE; | |
b5256303 | 886 | dsl_dataset_rele_flags(os->os_dsl_dataset, flags, tag); |
13fe0198 | 887 | dsl_pool_rele(dp, tag); |
428870ff | 888 | } |
b128c09f | 889 | |
b5256303 | 890 | void |
a926aab9 | 891 | dmu_objset_rele(objset_t *os, const void *tag) |
b5256303 TC |
892 | { |
893 | dmu_objset_rele_flags(os, B_FALSE, tag); | |
894 | } | |
895 | ||
831baf06 KW |
896 | /* |
897 | * When we are called, os MUST refer to an objset associated with a dataset | |
898 | * that is owned by 'tag'; that is, is held and long held by 'tag' and ds_owner | |
899 | * == tag. We will then release and reacquire ownership of the dataset while | |
900 | * holding the pool config_rwlock to avoid intervening namespace or ownership | |
901 | * changes may occur. | |
902 | * | |
903 | * This exists solely to accommodate zfs_ioc_userspace_upgrade()'s desire to | |
904 | * release the hold on its dataset and acquire a new one on the dataset of the | |
905 | * same name so that it can be partially torn down and reconstructed. | |
906 | */ | |
907 | void | |
5e00213e | 908 | dmu_objset_refresh_ownership(dsl_dataset_t *ds, dsl_dataset_t **newds, |
a926aab9 | 909 | boolean_t decrypt, const void *tag) |
831baf06 KW |
910 | { |
911 | dsl_pool_t *dp; | |
eca7b760 | 912 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
40ab927a | 913 | ds_hold_flags_t flags; |
831baf06 | 914 | |
40ab927a | 915 | flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE; |
831baf06 KW |
916 | VERIFY3P(ds, !=, NULL); |
917 | VERIFY3P(ds->ds_owner, ==, tag); | |
918 | VERIFY(dsl_dataset_long_held(ds)); | |
919 | ||
920 | dsl_dataset_name(ds, name); | |
5e00213e | 921 | dp = ds->ds_dir->dd_pool; |
831baf06 | 922 | dsl_pool_config_enter(dp, FTAG); |
40ab927a TS |
923 | dsl_dataset_disown(ds, flags, tag); |
924 | VERIFY0(dsl_dataset_own(dp, name, flags, tag, newds)); | |
831baf06 KW |
925 | dsl_pool_config_exit(dp, FTAG); |
926 | } | |
927 | ||
428870ff | 928 | void |
a926aab9 | 929 | dmu_objset_disown(objset_t *os, boolean_t decrypt, const void *tag) |
428870ff | 930 | { |
40ab927a TS |
931 | ds_hold_flags_t flags; |
932 | ||
933 | flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE; | |
1de321e6 JX |
934 | /* |
935 | * Stop upgrading thread | |
936 | */ | |
937 | dmu_objset_upgrade_stop(os); | |
40ab927a | 938 | dsl_dataset_disown(os->os_dsl_dataset, flags, tag); |
34dc7c2f BB |
939 | } |
940 | ||
13fe0198 | 941 | void |
34dc7c2f BB |
942 | dmu_objset_evict_dbufs(objset_t *os) |
943 | { | |
0c66c32d | 944 | dnode_t *dn_marker; |
34dc7c2f BB |
945 | dnode_t *dn; |
946 | ||
0c66c32d | 947 | dn_marker = kmem_alloc(sizeof (dnode_t), KM_SLEEP); |
34dc7c2f | 948 | |
0c66c32d JG |
949 | mutex_enter(&os->os_lock); |
950 | dn = list_head(&os->os_dnodes); | |
951 | while (dn != NULL) { | |
952 | /* | |
953 | * Skip dnodes without holds. We have to do this dance | |
954 | * because dnode_add_ref() only works if there is already a | |
955 | * hold. If the dnode has no holds, then it has no dbufs. | |
956 | */ | |
957 | if (dnode_add_ref(dn, FTAG)) { | |
958 | list_insert_after(&os->os_dnodes, dn, dn_marker); | |
959 | mutex_exit(&os->os_lock); | |
34dc7c2f | 960 | |
0c66c32d JG |
961 | dnode_evict_dbufs(dn); |
962 | dnode_rele(dn, FTAG); | |
34dc7c2f | 963 | |
0c66c32d JG |
964 | mutex_enter(&os->os_lock); |
965 | dn = list_next(&os->os_dnodes, dn_marker); | |
966 | list_remove(&os->os_dnodes, dn_marker); | |
967 | } else { | |
968 | dn = list_next(&os->os_dnodes, dn); | |
969 | } | |
970 | } | |
971 | mutex_exit(&os->os_lock); | |
34dc7c2f | 972 | |
0c66c32d | 973 | kmem_free(dn_marker, sizeof (dnode_t)); |
34dc7c2f | 974 | |
0c66c32d | 975 | if (DMU_USERUSED_DNODE(os) != NULL) { |
9c5167d1 NF |
976 | if (DMU_PROJECTUSED_DNODE(os) != NULL) |
977 | dnode_evict_dbufs(DMU_PROJECTUSED_DNODE(os)); | |
0c66c32d JG |
978 | dnode_evict_dbufs(DMU_GROUPUSED_DNODE(os)); |
979 | dnode_evict_dbufs(DMU_USERUSED_DNODE(os)); | |
34dc7c2f | 980 | } |
0c66c32d | 981 | dnode_evict_dbufs(DMU_META_DNODE(os)); |
34dc7c2f BB |
982 | } |
983 | ||
0c66c32d JG |
984 | /* |
985 | * Objset eviction processing is split into into two pieces. | |
986 | * The first marks the objset as evicting, evicts any dbufs that | |
987 | * have a refcount of zero, and then queues up the objset for the | |
988 | * second phase of eviction. Once os->os_dnodes has been cleared by | |
989 | * dnode_buf_pageout()->dnode_destroy(), the second phase is executed. | |
990 | * The second phase closes the special dnodes, dequeues the objset from | |
991 | * the list of those undergoing eviction, and finally frees the objset. | |
992 | * | |
993 | * NOTE: Due to asynchronous eviction processing (invocation of | |
994 | * dnode_buf_pageout()), it is possible for the meta dnode for the | |
995 | * objset to have no holds even though os->os_dnodes is not empty. | |
996 | */ | |
34dc7c2f | 997 | void |
428870ff | 998 | dmu_objset_evict(objset_t *os) |
34dc7c2f | 999 | { |
6f1ffb06 MA |
1000 | dsl_dataset_t *ds = os->os_dsl_dataset; |
1001 | ||
1c27024e | 1002 | for (int t = 0; t < TXG_SIZE; t++) |
428870ff | 1003 | ASSERT(!dmu_objset_is_dirty(os, t)); |
34dc7c2f | 1004 | |
0eb21616 JG |
1005 | if (ds) |
1006 | dsl_prop_unregister_all(ds, os); | |
34dc7c2f | 1007 | |
428870ff BB |
1008 | if (os->os_sa) |
1009 | sa_tear_down(os); | |
1010 | ||
13fe0198 | 1011 | dmu_objset_evict_dbufs(os); |
34dc7c2f | 1012 | |
0c66c32d JG |
1013 | mutex_enter(&os->os_lock); |
1014 | spa_evicting_os_register(os->os_spa, os); | |
1015 | if (list_is_empty(&os->os_dnodes)) { | |
1016 | mutex_exit(&os->os_lock); | |
1017 | dmu_objset_evict_done(os); | |
1018 | } else { | |
1019 | mutex_exit(&os->os_lock); | |
1020 | } | |
b5256303 TC |
1021 | |
1022 | ||
0c66c32d JG |
1023 | } |
1024 | ||
1025 | void | |
1026 | dmu_objset_evict_done(objset_t *os) | |
1027 | { | |
1028 | ASSERT3P(list_head(&os->os_dnodes), ==, NULL); | |
1029 | ||
572e2857 BB |
1030 | dnode_special_close(&os->os_meta_dnode); |
1031 | if (DMU_USERUSED_DNODE(os)) { | |
9c5167d1 NF |
1032 | if (DMU_PROJECTUSED_DNODE(os)) |
1033 | dnode_special_close(&os->os_projectused_dnode); | |
572e2857 BB |
1034 | dnode_special_close(&os->os_userused_dnode); |
1035 | dnode_special_close(&os->os_groupused_dnode); | |
9babb374 | 1036 | } |
428870ff BB |
1037 | zil_free(os->os_zil); |
1038 | ||
d3c2ae1c | 1039 | arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); |
572e2857 BB |
1040 | |
1041 | /* | |
1042 | * This is a barrier to prevent the objset from going away in | |
1043 | * dnode_move() until we can safely ensure that the objset is still in | |
1044 | * use. We consider the objset valid before the barrier and invalid | |
1045 | * after the barrier. | |
1046 | */ | |
1047 | rw_enter(&os_lock, RW_READER); | |
1048 | rw_exit(&os_lock); | |
1049 | ||
dbeb8796 MA |
1050 | kmem_free(os->os_obj_next_percpu, |
1051 | os->os_obj_next_percpu_len * sizeof (os->os_obj_next_percpu[0])); | |
1052 | ||
428870ff | 1053 | mutex_destroy(&os->os_lock); |
64fc7762 | 1054 | mutex_destroy(&os->os_userused_lock); |
428870ff BB |
1055 | mutex_destroy(&os->os_obj_lock); |
1056 | mutex_destroy(&os->os_user_ptr_lock); | |
c17486b2 | 1057 | mutex_destroy(&os->os_upgrade_lock); |
ffdf019c AM |
1058 | for (int i = 0; i < TXG_SIZE; i++) |
1059 | multilist_destroy(&os->os_dirty_dnodes[i]); | |
0c66c32d | 1060 | spa_evicting_os_deregister(os->os_spa, os); |
428870ff BB |
1061 | kmem_free(os, sizeof (objset_t)); |
1062 | } | |
9babb374 | 1063 | |
6413c95f | 1064 | inode_timespec_t |
428870ff BB |
1065 | dmu_objset_snap_cmtime(objset_t *os) |
1066 | { | |
1067 | return (dsl_dir_snap_cmtime(os->os_dsl_dataset->ds_dir)); | |
34dc7c2f BB |
1068 | } |
1069 | ||
428870ff | 1070 | objset_t * |
b5256303 TC |
1071 | dmu_objset_create_impl_dnstats(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, |
1072 | dmu_objset_type_t type, int levels, int blksz, int ibs, dmu_tx_t *tx) | |
34dc7c2f | 1073 | { |
428870ff | 1074 | objset_t *os; |
34dc7c2f BB |
1075 | dnode_t *mdn; |
1076 | ||
1077 | ASSERT(dmu_tx_is_syncing(tx)); | |
13fe0198 | 1078 | |
b5256303 TC |
1079 | if (blksz == 0) |
1080 | blksz = DNODE_BLOCK_SIZE; | |
4807c0ba | 1081 | if (ibs == 0) |
b5256303 TC |
1082 | ibs = DN_MAX_INDBLKSHIFT; |
1083 | ||
572e2857 | 1084 | if (ds != NULL) |
13fe0198 | 1085 | VERIFY0(dmu_objset_from_ds(ds, &os)); |
572e2857 | 1086 | else |
13fe0198 | 1087 | VERIFY0(dmu_objset_open_impl(spa, NULL, bp, &os)); |
572e2857 BB |
1088 | |
1089 | mdn = DMU_META_DNODE(os); | |
34dc7c2f | 1090 | |
b5256303 TC |
1091 | dnode_allocate(mdn, DMU_OT_DNODE, blksz, ibs, DMU_OT_NONE, 0, |
1092 | DNODE_MIN_SLOTS, tx); | |
34dc7c2f BB |
1093 | |
1094 | /* | |
1095 | * We don't want to have to increase the meta-dnode's nlevels | |
e1cfd73f | 1096 | * later, because then we could do it in quiescing context while |
34dc7c2f BB |
1097 | * we are also accessing it in open context. |
1098 | * | |
1099 | * This precaution is not necessary for the MOS (ds == NULL), | |
1100 | * because the MOS is only updated in syncing context. | |
1101 | * This is most fortunate: the MOS is the only objset that | |
1102 | * needs to be synced multiple times as spa_sync() iterates | |
1103 | * to convergence, so minimizing its dn_nlevels matters. | |
1104 | */ | |
1105 | if (ds != NULL) { | |
b5256303 TC |
1106 | if (levels == 0) { |
1107 | levels = 1; | |
1108 | ||
1109 | /* | |
1110 | * Determine the number of levels necessary for the | |
1111 | * meta-dnode to contain DN_MAX_OBJECT dnodes. Note | |
1112 | * that in order to ensure that we do not overflow | |
1113 | * 64 bits, there has to be a nlevels that gives us a | |
1114 | * number of blocks > DN_MAX_OBJECT but < 2^64. | |
1115 | * Therefore, (mdn->dn_indblkshift - SPA_BLKPTRSHIFT) | |
1116 | * (10) must be less than (64 - log2(DN_MAX_OBJECT)) | |
1117 | * (16). | |
1118 | */ | |
1119 | while ((uint64_t)mdn->dn_nblkptr << | |
1120 | (mdn->dn_datablkshift - DNODE_SHIFT + (levels - 1) * | |
1121 | (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) < | |
1122 | DN_MAX_OBJECT) | |
1123 | levels++; | |
1124 | } | |
34dc7c2f BB |
1125 | |
1126 | mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] = | |
1127 | mdn->dn_nlevels = levels; | |
1128 | } | |
1129 | ||
1130 | ASSERT(type != DMU_OST_NONE); | |
1131 | ASSERT(type != DMU_OST_ANY); | |
1132 | ASSERT(type < DMU_OST_NUMTYPES); | |
428870ff | 1133 | os->os_phys->os_type = type; |
b5256303 TC |
1134 | |
1135 | /* | |
1136 | * Enable user accounting if it is enabled and this is not an | |
1137 | * encrypted receive. | |
1138 | */ | |
1139 | if (dmu_objset_userused_enabled(os) && | |
1140 | (!os->os_encrypted || !dmu_objset_is_receiving(os))) { | |
428870ff | 1141 | os->os_phys->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; |
1de321e6 | 1142 | if (dmu_objset_userobjused_enabled(os)) { |
ae7e7006 | 1143 | ASSERT3P(ds, !=, NULL); |
d52d80b7 PD |
1144 | ds->ds_feature_activation[ |
1145 | SPA_FEATURE_USEROBJ_ACCOUNTING] = (void *)B_TRUE; | |
1de321e6 JX |
1146 | os->os_phys->os_flags |= |
1147 | OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE; | |
1148 | } | |
9c5167d1 | 1149 | if (dmu_objset_projectquota_enabled(os)) { |
08641d90 | 1150 | ASSERT3P(ds, !=, NULL); |
d52d80b7 PD |
1151 | ds->ds_feature_activation[ |
1152 | SPA_FEATURE_PROJECT_QUOTA] = (void *)B_TRUE; | |
9c5167d1 NF |
1153 | os->os_phys->os_flags |= |
1154 | OBJSET_FLAG_PROJECTQUOTA_COMPLETE; | |
1155 | } | |
428870ff | 1156 | os->os_flags = os->os_phys->os_flags; |
9babb374 | 1157 | } |
34dc7c2f BB |
1158 | |
1159 | dsl_dataset_dirty(ds, tx); | |
1160 | ||
428870ff | 1161 | return (os); |
34dc7c2f BB |
1162 | } |
1163 | ||
b5256303 TC |
1164 | /* called from dsl for meta-objset */ |
1165 | objset_t * | |
1166 | dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, | |
1167 | dmu_objset_type_t type, dmu_tx_t *tx) | |
1168 | { | |
1169 | return (dmu_objset_create_impl_dnstats(spa, ds, bp, type, 0, 0, 0, tx)); | |
1170 | } | |
1171 | ||
13fe0198 MA |
1172 | typedef struct dmu_objset_create_arg { |
1173 | const char *doca_name; | |
1174 | cred_t *doca_cred; | |
e59a377a | 1175 | proc_t *doca_proc; |
13fe0198 MA |
1176 | void (*doca_userfunc)(objset_t *os, void *arg, |
1177 | cred_t *cr, dmu_tx_t *tx); | |
1178 | void *doca_userarg; | |
1179 | dmu_objset_type_t doca_type; | |
1180 | uint64_t doca_flags; | |
b5256303 | 1181 | dsl_crypto_params_t *doca_dcp; |
13fe0198 | 1182 | } dmu_objset_create_arg_t; |
34dc7c2f | 1183 | |
34dc7c2f | 1184 | static int |
13fe0198 | 1185 | dmu_objset_create_check(void *arg, dmu_tx_t *tx) |
34dc7c2f | 1186 | { |
13fe0198 MA |
1187 | dmu_objset_create_arg_t *doca = arg; |
1188 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
1189 | dsl_dir_t *pdd; | |
d8d418ff | 1190 | dsl_dataset_t *parentds; |
1191 | objset_t *parentos; | |
13fe0198 MA |
1192 | const char *tail; |
1193 | int error; | |
34dc7c2f | 1194 | |
13fe0198 | 1195 | if (strchr(doca->doca_name, '@') != NULL) |
2e528b49 | 1196 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1197 | |
eca7b760 IK |
1198 | if (strlen(doca->doca_name) >= ZFS_MAX_DATASET_NAME_LEN) |
1199 | return (SET_ERROR(ENAMETOOLONG)); | |
1200 | ||
a7ed98d8 SD |
1201 | if (dataset_nestcheck(doca->doca_name) != 0) |
1202 | return (SET_ERROR(ENAMETOOLONG)); | |
1203 | ||
13fe0198 MA |
1204 | error = dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail); |
1205 | if (error != 0) | |
1206 | return (error); | |
1207 | if (tail == NULL) { | |
1208 | dsl_dir_rele(pdd, FTAG); | |
2e528b49 | 1209 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 1210 | } |
b5256303 | 1211 | |
1fff937a | 1212 | error = dmu_objset_create_crypt_check(pdd, doca->doca_dcp, NULL); |
b5256303 TC |
1213 | if (error != 0) { |
1214 | dsl_dir_rele(pdd, FTAG); | |
1215 | return (error); | |
1216 | } | |
1217 | ||
788eb90c | 1218 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
e59a377a | 1219 | doca->doca_cred, doca->doca_proc); |
d8d418ff | 1220 | if (error != 0) { |
1221 | dsl_dir_rele(pdd, FTAG); | |
1222 | return (error); | |
1223 | } | |
b5256303 | 1224 | |
d8d418ff | 1225 | /* can't create below anything but filesystems (eg. no ZVOLs) */ |
1226 | error = dsl_dataset_hold_obj(pdd->dd_pool, | |
1227 | dsl_dir_phys(pdd)->dd_head_dataset_obj, FTAG, &parentds); | |
1228 | if (error != 0) { | |
1229 | dsl_dir_rele(pdd, FTAG); | |
1230 | return (error); | |
1231 | } | |
1232 | error = dmu_objset_from_ds(parentds, &parentos); | |
1233 | if (error != 0) { | |
1234 | dsl_dataset_rele(parentds, FTAG); | |
1235 | dsl_dir_rele(pdd, FTAG); | |
1236 | return (error); | |
1237 | } | |
1238 | if (dmu_objset_type(parentos) != DMU_OST_ZFS) { | |
1239 | dsl_dataset_rele(parentds, FTAG); | |
1240 | dsl_dir_rele(pdd, FTAG); | |
1241 | return (SET_ERROR(ZFS_ERR_WRONG_PARENT)); | |
1242 | } | |
1243 | dsl_dataset_rele(parentds, FTAG); | |
13fe0198 | 1244 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f | 1245 | |
788eb90c | 1246 | return (error); |
34dc7c2f BB |
1247 | } |
1248 | ||
1249 | static void | |
13fe0198 | 1250 | dmu_objset_create_sync(void *arg, dmu_tx_t *tx) |
34dc7c2f | 1251 | { |
13fe0198 MA |
1252 | dmu_objset_create_arg_t *doca = arg; |
1253 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
52ce99dd | 1254 | spa_t *spa = dp->dp_spa; |
13fe0198 MA |
1255 | dsl_dir_t *pdd; |
1256 | const char *tail; | |
6f1ffb06 | 1257 | dsl_dataset_t *ds; |
13fe0198 | 1258 | uint64_t obj; |
6f1ffb06 | 1259 | blkptr_t *bp; |
13fe0198 | 1260 | objset_t *os; |
b5256303 | 1261 | zio_t *rzio; |
34dc7c2f | 1262 | |
13fe0198 | 1263 | VERIFY0(dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail)); |
34dc7c2f | 1264 | |
13fe0198 | 1265 | obj = dsl_dataset_create_sync(pdd, tail, NULL, doca->doca_flags, |
b5256303 | 1266 | doca->doca_cred, doca->doca_dcp, tx); |
34dc7c2f | 1267 | |
b5256303 TC |
1268 | VERIFY0(dsl_dataset_hold_obj_flags(pdd->dd_pool, obj, |
1269 | DS_HOLD_FLAG_DECRYPT, FTAG, &ds)); | |
cc9bb3e5 | 1270 | rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); |
6f1ffb06 | 1271 | bp = dsl_dataset_get_blkptr(ds); |
52ce99dd | 1272 | os = dmu_objset_create_impl(spa, ds, bp, doca->doca_type, tx); |
cc9bb3e5 | 1273 | rrw_exit(&ds->ds_bp_rwlock, FTAG); |
34dc7c2f | 1274 | |
13fe0198 MA |
1275 | if (doca->doca_userfunc != NULL) { |
1276 | doca->doca_userfunc(os, doca->doca_userarg, | |
1277 | doca->doca_cred, tx); | |
34dc7c2f BB |
1278 | } |
1279 | ||
b5256303 | 1280 | /* |
4807c0ba | 1281 | * The doca_userfunc() may write out some data that needs to be |
b5256303 TC |
1282 | * encrypted if the dataset is encrypted (specifically the root |
1283 | * directory). This data must be written out before the encryption | |
1284 | * key mapping is removed by dsl_dataset_rele_flags(). Force the | |
1285 | * I/O to occur immediately by invoking the relevant sections of | |
1286 | * dsl_pool_sync(). | |
1287 | */ | |
1288 | if (os->os_encrypted) { | |
1289 | dsl_dataset_t *tmpds = NULL; | |
1290 | boolean_t need_sync_done = B_FALSE; | |
1291 | ||
4807c0ba TC |
1292 | mutex_enter(&ds->ds_lock); |
1293 | ds->ds_owner = FTAG; | |
1294 | mutex_exit(&ds->ds_lock); | |
1295 | ||
52ce99dd | 1296 | rzio = zio_root(spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED); |
4807c0ba TC |
1297 | tmpds = txg_list_remove_this(&dp->dp_dirty_datasets, ds, |
1298 | tx->tx_txg); | |
b5256303 | 1299 | if (tmpds != NULL) { |
b5256303 TC |
1300 | dsl_dataset_sync(ds, rzio, tx); |
1301 | need_sync_done = B_TRUE; | |
1302 | } | |
1303 | VERIFY0(zio_wait(rzio)); | |
1304 | ||
ba67d821 | 1305 | dmu_objset_sync_done(os, tx); |
b5256303 | 1306 | taskq_wait(dp->dp_sync_taskq); |
52ce99dd TC |
1307 | if (txg_list_member(&dp->dp_dirty_datasets, ds, tx->tx_txg)) { |
1308 | ASSERT3P(ds->ds_key_mapping, !=, NULL); | |
1309 | key_mapping_rele(spa, ds->ds_key_mapping, ds); | |
1310 | } | |
b5256303 | 1311 | |
52ce99dd | 1312 | rzio = zio_root(spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED); |
4807c0ba TC |
1313 | tmpds = txg_list_remove_this(&dp->dp_dirty_datasets, ds, |
1314 | tx->tx_txg); | |
b5256303 | 1315 | if (tmpds != NULL) { |
b5256303 TC |
1316 | dmu_buf_rele(ds->ds_dbuf, ds); |
1317 | dsl_dataset_sync(ds, rzio, tx); | |
1318 | } | |
1319 | VERIFY0(zio_wait(rzio)); | |
1320 | ||
52ce99dd TC |
1321 | if (need_sync_done) { |
1322 | ASSERT3P(ds->ds_key_mapping, !=, NULL); | |
1323 | key_mapping_rele(spa, ds->ds_key_mapping, ds); | |
b5256303 | 1324 | dsl_dataset_sync_done(ds, tx); |
d816bc5e | 1325 | dmu_buf_rele(ds->ds_dbuf, ds); |
52ce99dd | 1326 | } |
4807c0ba TC |
1327 | |
1328 | mutex_enter(&ds->ds_lock); | |
1329 | ds->ds_owner = NULL; | |
1330 | mutex_exit(&ds->ds_lock); | |
b5256303 TC |
1331 | } |
1332 | ||
74756182 | 1333 | spa_history_log_internal_ds(ds, "create", tx, " "); |
a0bd735a | 1334 | |
b5256303 | 1335 | dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT, FTAG); |
13fe0198 | 1336 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f BB |
1337 | } |
1338 | ||
1339 | int | |
428870ff | 1340 | dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, |
b5256303 | 1341 | dsl_crypto_params_t *dcp, dmu_objset_create_sync_func_t func, void *arg) |
34dc7c2f | 1342 | { |
13fe0198 | 1343 | dmu_objset_create_arg_t doca; |
b5256303 | 1344 | dsl_crypto_params_t tmp_dcp = { 0 }; |
34dc7c2f | 1345 | |
13fe0198 MA |
1346 | doca.doca_name = name; |
1347 | doca.doca_cred = CRED(); | |
e59a377a | 1348 | doca.doca_proc = curproc; |
13fe0198 MA |
1349 | doca.doca_flags = flags; |
1350 | doca.doca_userfunc = func; | |
1351 | doca.doca_userarg = arg; | |
1352 | doca.doca_type = type; | |
34dc7c2f | 1353 | |
b5256303 TC |
1354 | /* |
1355 | * Some callers (mostly for testing) do not provide a dcp on their | |
1356 | * own but various code inside the sync task will require it to be | |
1357 | * allocated. Rather than adding NULL checks throughout this code | |
1358 | * or adding dummy dcp's to all of the callers we simply create a | |
1359 | * dummy one here and use that. This zero dcp will have the same | |
85ce3f4f | 1360 | * effect as asking for inheritance of all encryption params. |
b5256303 TC |
1361 | */ |
1362 | doca.doca_dcp = (dcp != NULL) ? dcp : &tmp_dcp; | |
1363 | ||
ec213971 | 1364 | int rv = dsl_sync_task(name, |
3d45fdd6 | 1365 | dmu_objset_create_check, dmu_objset_create_sync, &doca, |
ec213971 MA |
1366 | 6, ZFS_SPACE_CHECK_NORMAL); |
1367 | ||
1368 | if (rv == 0) | |
1369 | zvol_create_minor(name); | |
1370 | return (rv); | |
34dc7c2f BB |
1371 | } |
1372 | ||
13fe0198 MA |
1373 | typedef struct dmu_objset_clone_arg { |
1374 | const char *doca_clone; | |
1375 | const char *doca_origin; | |
1376 | cred_t *doca_cred; | |
e59a377a | 1377 | proc_t *doca_proc; |
13fe0198 MA |
1378 | } dmu_objset_clone_arg_t; |
1379 | ||
13fe0198 MA |
1380 | static int |
1381 | dmu_objset_clone_check(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 1382 | { |
13fe0198 | 1383 | dmu_objset_clone_arg_t *doca = arg; |
428870ff BB |
1384 | dsl_dir_t *pdd; |
1385 | const char *tail; | |
13fe0198 MA |
1386 | int error; |
1387 | dsl_dataset_t *origin; | |
1388 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
34dc7c2f | 1389 | |
13fe0198 | 1390 | if (strchr(doca->doca_clone, '@') != NULL) |
2e528b49 | 1391 | return (SET_ERROR(EINVAL)); |
13fe0198 | 1392 | |
eca7b760 IK |
1393 | if (strlen(doca->doca_clone) >= ZFS_MAX_DATASET_NAME_LEN) |
1394 | return (SET_ERROR(ENAMETOOLONG)); | |
1395 | ||
13fe0198 MA |
1396 | error = dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail); |
1397 | if (error != 0) | |
1398 | return (error); | |
428870ff | 1399 | if (tail == NULL) { |
13fe0198 | 1400 | dsl_dir_rele(pdd, FTAG); |
2e528b49 | 1401 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 1402 | } |
1cddb8c9 | 1403 | |
788eb90c | 1404 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
e59a377a | 1405 | doca->doca_cred, doca->doca_proc); |
788eb90c JJ |
1406 | if (error != 0) { |
1407 | dsl_dir_rele(pdd, FTAG); | |
1408 | return (SET_ERROR(EDQUOT)); | |
1409 | } | |
428870ff | 1410 | |
13fe0198 | 1411 | error = dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin); |
b5256303 TC |
1412 | if (error != 0) { |
1413 | dsl_dir_rele(pdd, FTAG); | |
572e2857 | 1414 | return (error); |
b5256303 | 1415 | } |
572e2857 | 1416 | |
13fe0198 | 1417 | /* You can only clone snapshots, not the head datasets. */ |
0c66c32d | 1418 | if (!origin->ds_is_snapshot) { |
13fe0198 | 1419 | dsl_dataset_rele(origin, FTAG); |
b5256303 | 1420 | dsl_dir_rele(pdd, FTAG); |
2e528b49 | 1421 | return (SET_ERROR(EINVAL)); |
428870ff | 1422 | } |
b5256303 | 1423 | |
13fe0198 | 1424 | dsl_dataset_rele(origin, FTAG); |
b5256303 | 1425 | dsl_dir_rele(pdd, FTAG); |
572e2857 | 1426 | |
13fe0198 | 1427 | return (0); |
9babb374 | 1428 | } |
34dc7c2f | 1429 | |
13fe0198 MA |
1430 | static void |
1431 | dmu_objset_clone_sync(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 1432 | { |
13fe0198 MA |
1433 | dmu_objset_clone_arg_t *doca = arg; |
1434 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
1435 | dsl_dir_t *pdd; | |
1436 | const char *tail; | |
1437 | dsl_dataset_t *origin, *ds; | |
1438 | uint64_t obj; | |
eca7b760 | 1439 | char namebuf[ZFS_MAX_DATASET_NAME_LEN]; |
6f1ffb06 | 1440 | |
13fe0198 MA |
1441 | VERIFY0(dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail)); |
1442 | VERIFY0(dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin)); | |
6f1ffb06 | 1443 | |
13fe0198 | 1444 | obj = dsl_dataset_create_sync(pdd, tail, origin, 0, |
b5256303 | 1445 | doca->doca_cred, NULL, tx); |
34dc7c2f | 1446 | |
13fe0198 MA |
1447 | VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); |
1448 | dsl_dataset_name(origin, namebuf); | |
1449 | spa_history_log_internal_ds(ds, "clone", tx, | |
74756182 | 1450 | "origin=%s (%llu)", namebuf, (u_longlong_t)origin->ds_object); |
13fe0198 MA |
1451 | dsl_dataset_rele(ds, FTAG); |
1452 | dsl_dataset_rele(origin, FTAG); | |
1453 | dsl_dir_rele(pdd, FTAG); | |
34dc7c2f BB |
1454 | } |
1455 | ||
1456 | int | |
13fe0198 | 1457 | dmu_objset_clone(const char *clone, const char *origin) |
34dc7c2f | 1458 | { |
13fe0198 | 1459 | dmu_objset_clone_arg_t doca; |
34dc7c2f | 1460 | |
13fe0198 MA |
1461 | doca.doca_clone = clone; |
1462 | doca.doca_origin = origin; | |
1463 | doca.doca_cred = CRED(); | |
e59a377a | 1464 | doca.doca_proc = curproc; |
572e2857 | 1465 | |
ec213971 | 1466 | int rv = dsl_sync_task(clone, |
3d45fdd6 | 1467 | dmu_objset_clone_check, dmu_objset_clone_sync, &doca, |
ec213971 MA |
1468 | 6, ZFS_SPACE_CHECK_NORMAL); |
1469 | ||
1470 | if (rv == 0) | |
1471 | zvol_create_minor(clone); | |
1472 | ||
1473 | return (rv); | |
6f1ffb06 MA |
1474 | } |
1475 | ||
1476 | int | |
1477 | dmu_objset_snapshot_one(const char *fsname, const char *snapname) | |
1478 | { | |
1479 | int err; | |
1480 | char *longsnap = kmem_asprintf("%s@%s", fsname, snapname); | |
1481 | nvlist_t *snaps = fnvlist_alloc(); | |
1482 | ||
1483 | fnvlist_add_boolean(snaps, longsnap); | |
e4f5fa12 | 1484 | kmem_strfree(longsnap); |
13fe0198 MA |
1485 | err = dsl_dataset_snapshot(snaps, NULL, NULL); |
1486 | fnvlist_free(snaps); | |
6f1ffb06 MA |
1487 | return (err); |
1488 | } | |
1489 | ||
1de321e6 JX |
1490 | static void |
1491 | dmu_objset_upgrade_task_cb(void *data) | |
1492 | { | |
1493 | objset_t *os = data; | |
1494 | ||
1495 | mutex_enter(&os->os_upgrade_lock); | |
1496 | os->os_upgrade_status = EINTR; | |
1497 | if (!os->os_upgrade_exit) { | |
39372fa2 AF |
1498 | int status; |
1499 | ||
1de321e6 JX |
1500 | mutex_exit(&os->os_upgrade_lock); |
1501 | ||
39372fa2 AF |
1502 | status = os->os_upgrade_cb(os); |
1503 | ||
1de321e6 | 1504 | mutex_enter(&os->os_upgrade_lock); |
39372fa2 AF |
1505 | |
1506 | os->os_upgrade_status = status; | |
1de321e6 JX |
1507 | } |
1508 | os->os_upgrade_exit = B_TRUE; | |
1509 | os->os_upgrade_id = 0; | |
1510 | mutex_exit(&os->os_upgrade_lock); | |
c0daec32 | 1511 | dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag); |
1de321e6 JX |
1512 | } |
1513 | ||
1514 | static void | |
1515 | dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb) | |
1516 | { | |
1517 | if (os->os_upgrade_id != 0) | |
1518 | return; | |
1519 | ||
c0daec32 AB |
1520 | ASSERT(dsl_pool_config_held(dmu_objset_pool(os))); |
1521 | dsl_dataset_long_hold(dmu_objset_ds(os), upgrade_tag); | |
1522 | ||
1de321e6 JX |
1523 | mutex_enter(&os->os_upgrade_lock); |
1524 | if (os->os_upgrade_id == 0 && os->os_upgrade_status == 0) { | |
1525 | os->os_upgrade_exit = B_FALSE; | |
1526 | os->os_upgrade_cb = cb; | |
1527 | os->os_upgrade_id = taskq_dispatch( | |
1528 | os->os_spa->spa_upgrade_taskq, | |
1529 | dmu_objset_upgrade_task_cb, os, TQ_SLEEP); | |
c0daec32 AB |
1530 | if (os->os_upgrade_id == TASKQID_INVALID) { |
1531 | dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag); | |
1de321e6 | 1532 | os->os_upgrade_status = ENOMEM; |
c0daec32 | 1533 | } |
39372fa2 AF |
1534 | } else { |
1535 | dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag); | |
1de321e6 JX |
1536 | } |
1537 | mutex_exit(&os->os_upgrade_lock); | |
1538 | } | |
1539 | ||
1540 | static void | |
1541 | dmu_objset_upgrade_stop(objset_t *os) | |
1542 | { | |
1543 | mutex_enter(&os->os_upgrade_lock); | |
1544 | os->os_upgrade_exit = B_TRUE; | |
1545 | if (os->os_upgrade_id != 0) { | |
1546 | taskqid_t id = os->os_upgrade_id; | |
1547 | ||
1548 | os->os_upgrade_id = 0; | |
1549 | mutex_exit(&os->os_upgrade_lock); | |
1550 | ||
c0daec32 AB |
1551 | if ((taskq_cancel_id(os->os_spa->spa_upgrade_taskq, id)) == 0) { |
1552 | dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag); | |
1553 | } | |
4807c0ba | 1554 | txg_wait_synced(os->os_spa->spa_dsl_pool, 0); |
1de321e6 JX |
1555 | } else { |
1556 | mutex_exit(&os->os_upgrade_lock); | |
1557 | } | |
1558 | } | |
1559 | ||
34dc7c2f | 1560 | static void |
64fc7762 | 1561 | dmu_objset_sync_dnodes(multilist_sublist_t *list, dmu_tx_t *tx) |
34dc7c2f BB |
1562 | { |
1563 | dnode_t *dn; | |
1564 | ||
64fc7762 | 1565 | while ((dn = multilist_sublist_head(list)) != NULL) { |
34dc7c2f BB |
1566 | ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT); |
1567 | ASSERT(dn->dn_dbuf->db_data_pending); | |
1568 | /* | |
9babb374 | 1569 | * Initialize dn_zio outside dnode_sync() because the |
93e28d66 | 1570 | * meta-dnode needs to set it outside dnode_sync(). |
34dc7c2f BB |
1571 | */ |
1572 | dn->dn_zio = dn->dn_dbuf->db_data_pending->dr_zio; | |
1573 | ASSERT(dn->dn_zio); | |
1574 | ||
1575 | ASSERT3U(dn->dn_nlevels, <=, DN_MAX_LEVELS); | |
64fc7762 | 1576 | multilist_sublist_remove(list, dn); |
9babb374 | 1577 | |
edc1e713 | 1578 | /* |
ba67d821 MA |
1579 | * See the comment above dnode_rele_task() for an explanation |
1580 | * of why this dnode hold is always needed (even when not | |
1581 | * doing user accounting). | |
edc1e713 | 1582 | */ |
ffdf019c | 1583 | multilist_t *newlist = &dn->dn_objset->os_synced_dnodes; |
ba67d821 MA |
1584 | (void) dnode_add_ref(dn, newlist); |
1585 | multilist_insert(newlist, dn); | |
9babb374 | 1586 | |
34dc7c2f BB |
1587 | dnode_sync(dn, tx); |
1588 | } | |
1589 | } | |
1590 | ||
34dc7c2f | 1591 | static void |
428870ff | 1592 | dmu_objset_write_ready(zio_t *zio, arc_buf_t *abuf, void *arg) |
34dc7c2f | 1593 | { |
14e4e3cb | 1594 | (void) abuf; |
b128c09f | 1595 | blkptr_t *bp = zio->io_bp; |
428870ff | 1596 | objset_t *os = arg; |
34dc7c2f | 1597 | dnode_phys_t *dnp = &os->os_phys->os_meta_dnode; |
b5256303 | 1598 | uint64_t fill = 0; |
34dc7c2f | 1599 | |
9b67f605 | 1600 | ASSERT(!BP_IS_EMBEDDED(bp)); |
13fe0198 MA |
1601 | ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_OBJSET); |
1602 | ASSERT0(BP_GET_LEVEL(bp)); | |
34dc7c2f BB |
1603 | |
1604 | /* | |
9babb374 BB |
1605 | * Update rootbp fill count: it should be the number of objects |
1606 | * allocated in the object set (not counting the "special" | |
1607 | * objects that are stored in the objset_phys_t -- the meta | |
9c5167d1 | 1608 | * dnode and user/group/project accounting objects). |
34dc7c2f | 1609 | */ |
1c27024e | 1610 | for (int i = 0; i < dnp->dn_nblkptr; i++) |
b5256303 TC |
1611 | fill += BP_GET_FILL(&dnp->dn_blkptr[i]); |
1612 | ||
1613 | BP_SET_FILL(bp, fill); | |
1614 | ||
cc9bb3e5 GM |
1615 | if (os->os_dsl_dataset != NULL) |
1616 | rrw_enter(&os->os_dsl_dataset->ds_bp_rwlock, RW_WRITER, FTAG); | |
1617 | *os->os_rootbp = *bp; | |
1618 | if (os->os_dsl_dataset != NULL) | |
1619 | rrw_exit(&os->os_dsl_dataset->ds_bp_rwlock, FTAG); | |
428870ff BB |
1620 | } |
1621 | ||
428870ff BB |
1622 | static void |
1623 | dmu_objset_write_done(zio_t *zio, arc_buf_t *abuf, void *arg) | |
1624 | { | |
14e4e3cb | 1625 | (void) abuf; |
428870ff BB |
1626 | blkptr_t *bp = zio->io_bp; |
1627 | blkptr_t *bp_orig = &zio->io_bp_orig; | |
1628 | objset_t *os = arg; | |
34dc7c2f | 1629 | |
b128c09f | 1630 | if (zio->io_flags & ZIO_FLAG_IO_REWRITE) { |
428870ff | 1631 | ASSERT(BP_EQUAL(bp, bp_orig)); |
b128c09f | 1632 | } else { |
428870ff BB |
1633 | dsl_dataset_t *ds = os->os_dsl_dataset; |
1634 | dmu_tx_t *tx = os->os_synctx; | |
1635 | ||
1636 | (void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE); | |
1637 | dsl_dataset_block_born(ds, bp, tx); | |
34dc7c2f | 1638 | } |
cc9bb3e5 | 1639 | kmem_free(bp, sizeof (*bp)); |
34dc7c2f BB |
1640 | } |
1641 | ||
3bd4df38 EN |
1642 | typedef struct sync_objset_arg { |
1643 | zio_t *soa_zio; | |
1644 | objset_t *soa_os; | |
1645 | dmu_tx_t *soa_tx; | |
1646 | kmutex_t soa_mutex; | |
1647 | int soa_count; | |
1648 | taskq_ent_t soa_tq_ent; | |
1649 | } sync_objset_arg_t; | |
1650 | ||
64fc7762 | 1651 | typedef struct sync_dnodes_arg { |
3bd4df38 EN |
1652 | multilist_t *sda_list; |
1653 | int sda_sublist_idx; | |
1654 | multilist_t *sda_newlist; | |
1655 | sync_objset_arg_t *sda_soa; | |
64fc7762 MA |
1656 | } sync_dnodes_arg_t; |
1657 | ||
3bd4df38 EN |
1658 | static void sync_meta_dnode_task(void *arg); |
1659 | ||
64fc7762 MA |
1660 | static void |
1661 | sync_dnodes_task(void *arg) | |
1662 | { | |
1663 | sync_dnodes_arg_t *sda = arg; | |
3bd4df38 EN |
1664 | sync_objset_arg_t *soa = sda->sda_soa; |
1665 | objset_t *os = soa->soa_os; | |
64fc7762 | 1666 | |
645b8330 | 1667 | uint_t allocator = spa_acq_allocator(os->os_spa); |
64fc7762 | 1668 | multilist_sublist_t *ms = |
997f85b4 | 1669 | multilist_sublist_lock_idx(sda->sda_list, sda->sda_sublist_idx); |
64fc7762 | 1670 | |
3bd4df38 | 1671 | dmu_objset_sync_dnodes(ms, soa->soa_tx); |
64fc7762 MA |
1672 | |
1673 | multilist_sublist_unlock(ms); | |
645b8330 | 1674 | spa_rel_allocator(os->os_spa, allocator); |
64fc7762 MA |
1675 | |
1676 | kmem_free(sda, sizeof (*sda)); | |
3bd4df38 EN |
1677 | |
1678 | mutex_enter(&soa->soa_mutex); | |
1679 | ASSERT(soa->soa_count != 0); | |
1680 | if (--soa->soa_count != 0) { | |
1681 | mutex_exit(&soa->soa_mutex); | |
1682 | return; | |
1683 | } | |
1684 | mutex_exit(&soa->soa_mutex); | |
1685 | ||
1686 | taskq_dispatch_ent(dmu_objset_pool(os)->dp_sync_taskq, | |
1687 | sync_meta_dnode_task, soa, TQ_FRONT, &soa->soa_tq_ent); | |
64fc7762 MA |
1688 | } |
1689 | ||
3bd4df38 EN |
1690 | /* |
1691 | * Issue the zio_nowait() for all dirty record zios on the meta dnode, | |
1692 | * then trigger the callback for the zil_sync. This runs once for each | |
1693 | * objset, only after any/all sublists in the objset have been synced. | |
1694 | */ | |
1695 | static void | |
1696 | sync_meta_dnode_task(void *arg) | |
1697 | { | |
1698 | sync_objset_arg_t *soa = arg; | |
1699 | objset_t *os = soa->soa_os; | |
1700 | dmu_tx_t *tx = soa->soa_tx; | |
1701 | int txgoff = tx->tx_txg & TXG_MASK; | |
1702 | dbuf_dirty_record_t *dr; | |
1703 | ||
1704 | ASSERT0(soa->soa_count); | |
1705 | ||
1706 | list_t *list = &DMU_META_DNODE(os)->dn_dirty_records[txgoff]; | |
1707 | while ((dr = list_remove_head(list)) != NULL) { | |
1708 | ASSERT0(dr->dr_dbuf->db_level); | |
1709 | zio_nowait(dr->dr_zio); | |
1710 | } | |
1711 | ||
1712 | /* Enable dnode backfill if enough objects have been freed. */ | |
1713 | if (os->os_freed_dnodes >= dmu_rescan_dnode_threshold) { | |
1714 | os->os_rescan_dnodes = B_TRUE; | |
1715 | os->os_freed_dnodes = 0; | |
1716 | } | |
1717 | ||
1718 | /* | |
1719 | * Free intent log blocks up to this tx. | |
1720 | */ | |
1721 | zil_sync(os->os_zil, tx); | |
1722 | os->os_phys->os_zil_header = os->os_zil_header; | |
1723 | zio_nowait(soa->soa_zio); | |
1724 | ||
1725 | mutex_destroy(&soa->soa_mutex); | |
1726 | kmem_free(soa, sizeof (*soa)); | |
1727 | } | |
64fc7762 | 1728 | |
34dc7c2f BB |
1729 | /* called from dsl */ |
1730 | void | |
428870ff | 1731 | dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx) |
34dc7c2f BB |
1732 | { |
1733 | int txgoff; | |
5dbd68a3 | 1734 | zbookmark_phys_t zb; |
428870ff | 1735 | zio_prop_t zp; |
34dc7c2f | 1736 | zio_t *zio; |
fc754677 AM |
1737 | int num_sublists; |
1738 | multilist_t *ml; | |
cc9bb3e5 GM |
1739 | blkptr_t *blkptr_copy = kmem_alloc(sizeof (*os->os_rootbp), KM_SLEEP); |
1740 | *blkptr_copy = *os->os_rootbp; | |
34dc7c2f | 1741 | |
8e739b2c | 1742 | dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", (u_longlong_t)tx->tx_txg); |
34dc7c2f BB |
1743 | |
1744 | ASSERT(dmu_tx_is_syncing(tx)); | |
1745 | /* XXX the write_done callback should really give us the tx... */ | |
1746 | os->os_synctx = tx; | |
1747 | ||
1748 | if (os->os_dsl_dataset == NULL) { | |
1749 | /* | |
1750 | * This is the MOS. If we have upgraded, | |
1751 | * spa_max_replication() could change, so reset | |
1752 | * os_copies here. | |
1753 | */ | |
1754 | os->os_copies = spa_max_replication(os->os_spa); | |
1755 | } | |
1756 | ||
1757 | /* | |
1758 | * Create the root block IO | |
1759 | */ | |
428870ff BB |
1760 | SET_BOOKMARK(&zb, os->os_dsl_dataset ? |
1761 | os->os_dsl_dataset->ds_object : DMU_META_OBJSET, | |
1762 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
294f6806 | 1763 | arc_release(os->os_phys_buf, &os->os_phys_buf); |
b128c09f | 1764 | |
82644107 | 1765 | dmu_write_policy(os, NULL, 0, 0, &zp); |
9babb374 | 1766 | |
b5256303 | 1767 | /* |
0c03d21a MA |
1768 | * If we are either claiming the ZIL or doing a raw receive, write |
1769 | * out the os_phys_buf raw. Neither of these actions will effect the | |
1770 | * MAC at this point. | |
b5256303 | 1771 | */ |
0c03d21a MA |
1772 | if (os->os_raw_receive || |
1773 | os->os_next_write_raw[tx->tx_txg & TXG_MASK]) { | |
b5256303 | 1774 | ASSERT(os->os_encrypted); |
b5256303 TC |
1775 | arc_convert_to_raw(os->os_phys_buf, |
1776 | os->os_dsl_dataset->ds_object, ZFS_HOST_BYTEORDER, | |
1777 | DMU_OT_OBJSET, NULL, NULL, NULL); | |
1778 | } | |
1779 | ||
428870ff | 1780 | zio = arc_write(pio, os->os_spa, tx->tx_txg, |
ed2f7ba0 | 1781 | blkptr_copy, os->os_phys_buf, B_FALSE, dmu_os_is_l2cacheable(os), |
ccec7fbe | 1782 | &zp, dmu_objset_write_ready, NULL, dmu_objset_write_done, |
bc77ba73 | 1783 | os, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb); |
34dc7c2f BB |
1784 | |
1785 | /* | |
9babb374 | 1786 | * Sync special dnodes - the parent IO for the sync is the root block |
34dc7c2f | 1787 | */ |
572e2857 BB |
1788 | DMU_META_DNODE(os)->dn_zio = zio; |
1789 | dnode_sync(DMU_META_DNODE(os), tx); | |
34dc7c2f | 1790 | |
9babb374 BB |
1791 | os->os_phys->os_flags = os->os_flags; |
1792 | ||
572e2857 BB |
1793 | if (DMU_USERUSED_DNODE(os) && |
1794 | DMU_USERUSED_DNODE(os)->dn_type != DMU_OT_NONE) { | |
1795 | DMU_USERUSED_DNODE(os)->dn_zio = zio; | |
1796 | dnode_sync(DMU_USERUSED_DNODE(os), tx); | |
1797 | DMU_GROUPUSED_DNODE(os)->dn_zio = zio; | |
1798 | dnode_sync(DMU_GROUPUSED_DNODE(os), tx); | |
9babb374 BB |
1799 | } |
1800 | ||
9c5167d1 NF |
1801 | if (DMU_PROJECTUSED_DNODE(os) && |
1802 | DMU_PROJECTUSED_DNODE(os)->dn_type != DMU_OT_NONE) { | |
1803 | DMU_PROJECTUSED_DNODE(os)->dn_zio = zio; | |
1804 | dnode_sync(DMU_PROJECTUSED_DNODE(os), tx); | |
1805 | } | |
1806 | ||
34dc7c2f BB |
1807 | txgoff = tx->tx_txg & TXG_MASK; |
1808 | ||
ba67d821 MA |
1809 | /* |
1810 | * We must create the list here because it uses the | |
1811 | * dn_dirty_link[] of this txg. But it may already | |
1812 | * exist because we call dsl_dataset_sync() twice per txg. | |
1813 | */ | |
ffdf019c AM |
1814 | if (os->os_synced_dnodes.ml_sublists == NULL) { |
1815 | multilist_create(&os->os_synced_dnodes, sizeof (dnode_t), | |
ba67d821 MA |
1816 | offsetof(dnode_t, dn_dirty_link[txgoff]), |
1817 | dnode_multilist_index_func); | |
1818 | } else { | |
ffdf019c | 1819 | ASSERT3U(os->os_synced_dnodes.ml_offset, ==, |
ba67d821 | 1820 | offsetof(dnode_t, dn_dirty_link[txgoff])); |
9babb374 BB |
1821 | } |
1822 | ||
3bd4df38 EN |
1823 | /* |
1824 | * zio_nowait(zio) is done after any/all sublist and meta dnode | |
1825 | * zios have been nowaited, and the zil_sync() has been performed. | |
1826 | * The soa is freed at the end of sync_meta_dnode_task. | |
1827 | */ | |
1828 | sync_objset_arg_t *soa = kmem_alloc(sizeof (*soa), KM_SLEEP); | |
1829 | soa->soa_zio = zio; | |
1830 | soa->soa_os = os; | |
1831 | soa->soa_tx = tx; | |
1832 | taskq_init_ent(&soa->soa_tq_ent); | |
1833 | mutex_init(&soa->soa_mutex, NULL, MUTEX_DEFAULT, NULL); | |
1834 | ||
ffdf019c | 1835 | ml = &os->os_dirty_dnodes[txgoff]; |
3bd4df38 EN |
1836 | soa->soa_count = num_sublists = multilist_get_num_sublists(ml); |
1837 | ||
fc754677 AM |
1838 | for (int i = 0; i < num_sublists; i++) { |
1839 | if (multilist_sublist_is_empty_idx(ml, i)) | |
3bd4df38 | 1840 | soa->soa_count--; |
64fc7762 | 1841 | } |
34dc7c2f | 1842 | |
3bd4df38 EN |
1843 | if (soa->soa_count == 0) { |
1844 | taskq_dispatch_ent(dmu_objset_pool(os)->dp_sync_taskq, | |
1845 | sync_meta_dnode_task, soa, TQ_FRONT, &soa->soa_tq_ent); | |
1846 | } else { | |
1847 | /* | |
1848 | * Sync sublists in parallel. The last to finish | |
1849 | * (i.e., when soa->soa_count reaches zero) must | |
1850 | * dispatch sync_meta_dnode_task. | |
1851 | */ | |
1852 | for (int i = 0; i < num_sublists; i++) { | |
1853 | if (multilist_sublist_is_empty_idx(ml, i)) | |
1854 | continue; | |
1855 | sync_dnodes_arg_t *sda = | |
1856 | kmem_alloc(sizeof (*sda), KM_SLEEP); | |
1857 | sda->sda_list = ml; | |
1858 | sda->sda_sublist_idx = i; | |
1859 | sda->sda_soa = soa; | |
1860 | (void) taskq_dispatch( | |
1861 | dmu_objset_pool(os)->dp_sync_taskq, | |
1862 | sync_dnodes_task, sda, 0); | |
1863 | /* sync_dnodes_task frees sda */ | |
1864 | } | |
68cbd56e | 1865 | } |
34dc7c2f BB |
1866 | } |
1867 | ||
428870ff BB |
1868 | boolean_t |
1869 | dmu_objset_is_dirty(objset_t *os, uint64_t txg) | |
1870 | { | |
ffdf019c | 1871 | return (!multilist_is_empty(&os->os_dirty_dnodes[txg & TXG_MASK])); |
428870ff BB |
1872 | } |
1873 | ||
7bcb7f08 | 1874 | static file_info_cb_t *file_cbs[DMU_OST_NUMTYPES]; |
9babb374 BB |
1875 | |
1876 | void | |
7bcb7f08 | 1877 | dmu_objset_register_type(dmu_objset_type_t ost, file_info_cb_t *cb) |
9babb374 | 1878 | { |
7bcb7f08 MA |
1879 | file_cbs[ost] = cb; |
1880 | } | |
1881 | ||
1882 | int | |
1883 | dmu_get_file_info(objset_t *os, dmu_object_type_t bonustype, const void *data, | |
1884 | zfs_file_info_t *zfi) | |
1885 | { | |
1886 | file_info_cb_t *cb = file_cbs[os->os_phys->os_type]; | |
1887 | if (cb == NULL) | |
1888 | return (EINVAL); | |
1889 | return (cb(bonustype, data, zfi)); | |
9babb374 BB |
1890 | } |
1891 | ||
1892 | boolean_t | |
428870ff | 1893 | dmu_objset_userused_enabled(objset_t *os) |
9babb374 BB |
1894 | { |
1895 | return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE && | |
7bcb7f08 | 1896 | file_cbs[os->os_phys->os_type] != NULL && |
572e2857 | 1897 | DMU_USERUSED_DNODE(os) != NULL); |
9babb374 BB |
1898 | } |
1899 | ||
1de321e6 JX |
1900 | boolean_t |
1901 | dmu_objset_userobjused_enabled(objset_t *os) | |
1902 | { | |
1903 | return (dmu_objset_userused_enabled(os) && | |
1904 | spa_feature_is_enabled(os->os_spa, SPA_FEATURE_USEROBJ_ACCOUNTING)); | |
1905 | } | |
1906 | ||
9c5167d1 NF |
1907 | boolean_t |
1908 | dmu_objset_projectquota_enabled(objset_t *os) | |
1909 | { | |
7bcb7f08 | 1910 | return (file_cbs[os->os_phys->os_type] != NULL && |
9c5167d1 NF |
1911 | DMU_PROJECTUSED_DNODE(os) != NULL && |
1912 | spa_feature_is_enabled(os->os_spa, SPA_FEATURE_PROJECT_QUOTA)); | |
1913 | } | |
1914 | ||
9b7a83cb JX |
1915 | typedef struct userquota_node { |
1916 | /* must be in the first filed, see userquota_update_cache() */ | |
1917 | char uqn_id[20 + DMU_OBJACCT_PREFIX_LEN]; | |
1918 | int64_t uqn_delta; | |
1919 | avl_node_t uqn_node; | |
1920 | } userquota_node_t; | |
1921 | ||
1922 | typedef struct userquota_cache { | |
1923 | avl_tree_t uqc_user_deltas; | |
1924 | avl_tree_t uqc_group_deltas; | |
9c5167d1 | 1925 | avl_tree_t uqc_project_deltas; |
9b7a83cb JX |
1926 | } userquota_cache_t; |
1927 | ||
1928 | static int | |
1929 | userquota_compare(const void *l, const void *r) | |
1930 | { | |
1931 | const userquota_node_t *luqn = l; | |
1932 | const userquota_node_t *ruqn = r; | |
e4ffa98d | 1933 | int rv; |
9b7a83cb JX |
1934 | |
1935 | /* | |
1936 | * NB: can only access uqn_id because userquota_update_cache() doesn't | |
1937 | * pass in an entire userquota_node_t. | |
1938 | */ | |
e4ffa98d BB |
1939 | rv = strcmp(luqn->uqn_id, ruqn->uqn_id); |
1940 | ||
ca577779 | 1941 | return (TREE_ISIGN(rv)); |
9b7a83cb JX |
1942 | } |
1943 | ||
1944 | static void | |
1945 | do_userquota_cacheflush(objset_t *os, userquota_cache_t *cache, dmu_tx_t *tx) | |
1946 | { | |
1947 | void *cookie; | |
1948 | userquota_node_t *uqn; | |
1949 | ||
1950 | ASSERT(dmu_tx_is_syncing(tx)); | |
1951 | ||
1952 | cookie = NULL; | |
1953 | while ((uqn = avl_destroy_nodes(&cache->uqc_user_deltas, | |
1954 | &cookie)) != NULL) { | |
64fc7762 MA |
1955 | /* |
1956 | * os_userused_lock protects against concurrent calls to | |
1957 | * zap_increment_int(). It's needed because zap_increment_int() | |
1958 | * is not thread-safe (i.e. not atomic). | |
1959 | */ | |
1960 | mutex_enter(&os->os_userused_lock); | |
9b7a83cb JX |
1961 | VERIFY0(zap_increment(os, DMU_USERUSED_OBJECT, |
1962 | uqn->uqn_id, uqn->uqn_delta, tx)); | |
64fc7762 | 1963 | mutex_exit(&os->os_userused_lock); |
9b7a83cb JX |
1964 | kmem_free(uqn, sizeof (*uqn)); |
1965 | } | |
1966 | avl_destroy(&cache->uqc_user_deltas); | |
1967 | ||
1968 | cookie = NULL; | |
1969 | while ((uqn = avl_destroy_nodes(&cache->uqc_group_deltas, | |
1970 | &cookie)) != NULL) { | |
64fc7762 | 1971 | mutex_enter(&os->os_userused_lock); |
9b7a83cb JX |
1972 | VERIFY0(zap_increment(os, DMU_GROUPUSED_OBJECT, |
1973 | uqn->uqn_id, uqn->uqn_delta, tx)); | |
64fc7762 | 1974 | mutex_exit(&os->os_userused_lock); |
9b7a83cb JX |
1975 | kmem_free(uqn, sizeof (*uqn)); |
1976 | } | |
1977 | avl_destroy(&cache->uqc_group_deltas); | |
9c5167d1 NF |
1978 | |
1979 | if (dmu_objset_projectquota_enabled(os)) { | |
1980 | cookie = NULL; | |
1981 | while ((uqn = avl_destroy_nodes(&cache->uqc_project_deltas, | |
1982 | &cookie)) != NULL) { | |
1983 | mutex_enter(&os->os_userused_lock); | |
1984 | VERIFY0(zap_increment(os, DMU_PROJECTUSED_OBJECT, | |
1985 | uqn->uqn_id, uqn->uqn_delta, tx)); | |
1986 | mutex_exit(&os->os_userused_lock); | |
1987 | kmem_free(uqn, sizeof (*uqn)); | |
1988 | } | |
1989 | avl_destroy(&cache->uqc_project_deltas); | |
1990 | } | |
9b7a83cb JX |
1991 | } |
1992 | ||
1993 | static void | |
1994 | userquota_update_cache(avl_tree_t *avl, const char *id, int64_t delta) | |
1995 | { | |
1996 | userquota_node_t *uqn; | |
1997 | avl_index_t idx; | |
1998 | ||
1999 | ASSERT(strlen(id) < sizeof (uqn->uqn_id)); | |
2000 | /* | |
2001 | * Use id directly for searching because uqn_id is the first field of | |
2002 | * userquota_node_t and fields after uqn_id won't be accessed in | |
2003 | * avl_find(). | |
2004 | */ | |
2005 | uqn = avl_find(avl, (const void *)id, &idx); | |
2006 | if (uqn == NULL) { | |
2007 | uqn = kmem_zalloc(sizeof (*uqn), KM_SLEEP); | |
1f51b525 | 2008 | strlcpy(uqn->uqn_id, id, sizeof (uqn->uqn_id)); |
9b7a83cb JX |
2009 | avl_insert(avl, uqn, idx); |
2010 | } | |
2011 | uqn->uqn_delta += delta; | |
2012 | } | |
2013 | ||
428870ff | 2014 | static void |
9c5167d1 NF |
2015 | do_userquota_update(objset_t *os, userquota_cache_t *cache, uint64_t used, |
2016 | uint64_t flags, uint64_t user, uint64_t group, uint64_t project, | |
2017 | boolean_t subtract) | |
428870ff | 2018 | { |
9c5167d1 | 2019 | if (flags & DNODE_FLAG_USERUSED_ACCOUNTED) { |
50c957f7 | 2020 | int64_t delta = DNODE_MIN_SIZE + used; |
9b7a83cb JX |
2021 | char name[20]; |
2022 | ||
428870ff BB |
2023 | if (subtract) |
2024 | delta = -delta; | |
9b7a83cb | 2025 | |
c9e319fa | 2026 | (void) snprintf(name, sizeof (name), "%llx", (longlong_t)user); |
9b7a83cb JX |
2027 | userquota_update_cache(&cache->uqc_user_deltas, name, delta); |
2028 | ||
c9e319fa | 2029 | (void) snprintf(name, sizeof (name), "%llx", (longlong_t)group); |
9b7a83cb | 2030 | userquota_update_cache(&cache->uqc_group_deltas, name, delta); |
9c5167d1 NF |
2031 | |
2032 | if (dmu_objset_projectquota_enabled(os)) { | |
c9e319fa JL |
2033 | (void) snprintf(name, sizeof (name), "%llx", |
2034 | (longlong_t)project); | |
9c5167d1 NF |
2035 | userquota_update_cache(&cache->uqc_project_deltas, |
2036 | name, delta); | |
2037 | } | |
428870ff BB |
2038 | } |
2039 | } | |
2040 | ||
1de321e6 | 2041 | static void |
9c5167d1 NF |
2042 | do_userobjquota_update(objset_t *os, userquota_cache_t *cache, uint64_t flags, |
2043 | uint64_t user, uint64_t group, uint64_t project, boolean_t subtract) | |
1de321e6 JX |
2044 | { |
2045 | if (flags & DNODE_FLAG_USEROBJUSED_ACCOUNTED) { | |
2046 | char name[20 + DMU_OBJACCT_PREFIX_LEN]; | |
9b7a83cb | 2047 | int delta = subtract ? -1 : 1; |
1de321e6 JX |
2048 | |
2049 | (void) snprintf(name, sizeof (name), DMU_OBJACCT_PREFIX "%llx", | |
2050 | (longlong_t)user); | |
9b7a83cb | 2051 | userquota_update_cache(&cache->uqc_user_deltas, name, delta); |
1de321e6 JX |
2052 | |
2053 | (void) snprintf(name, sizeof (name), DMU_OBJACCT_PREFIX "%llx", | |
2054 | (longlong_t)group); | |
9b7a83cb | 2055 | userquota_update_cache(&cache->uqc_group_deltas, name, delta); |
9c5167d1 NF |
2056 | |
2057 | if (dmu_objset_projectquota_enabled(os)) { | |
2058 | (void) snprintf(name, sizeof (name), | |
2059 | DMU_OBJACCT_PREFIX "%llx", (longlong_t)project); | |
2060 | userquota_update_cache(&cache->uqc_project_deltas, | |
2061 | name, delta); | |
2062 | } | |
1de321e6 JX |
2063 | } |
2064 | } | |
2065 | ||
64fc7762 MA |
2066 | typedef struct userquota_updates_arg { |
2067 | objset_t *uua_os; | |
2068 | int uua_sublist_idx; | |
2069 | dmu_tx_t *uua_tx; | |
2070 | } userquota_updates_arg_t; | |
2071 | ||
2072 | static void | |
2073 | userquota_updates_task(void *arg) | |
9babb374 | 2074 | { |
64fc7762 MA |
2075 | userquota_updates_arg_t *uua = arg; |
2076 | objset_t *os = uua->uua_os; | |
2077 | dmu_tx_t *tx = uua->uua_tx; | |
9babb374 | 2078 | dnode_t *dn; |
9b7a83cb | 2079 | userquota_cache_t cache = { { 0 } }; |
9babb374 | 2080 | |
997f85b4 AM |
2081 | multilist_sublist_t *list = multilist_sublist_lock_idx( |
2082 | &os->os_synced_dnodes, uua->uua_sublist_idx); | |
9babb374 | 2083 | |
64fc7762 MA |
2084 | ASSERT(multilist_sublist_head(list) == NULL || |
2085 | dmu_objset_userused_enabled(os)); | |
9b7a83cb JX |
2086 | avl_create(&cache.uqc_user_deltas, userquota_compare, |
2087 | sizeof (userquota_node_t), offsetof(userquota_node_t, uqn_node)); | |
2088 | avl_create(&cache.uqc_group_deltas, userquota_compare, | |
2089 | sizeof (userquota_node_t), offsetof(userquota_node_t, uqn_node)); | |
9c5167d1 NF |
2090 | if (dmu_objset_projectquota_enabled(os)) |
2091 | avl_create(&cache.uqc_project_deltas, userquota_compare, | |
2092 | sizeof (userquota_node_t), offsetof(userquota_node_t, | |
2093 | uqn_node)); | |
9b7a83cb | 2094 | |
64fc7762 | 2095 | while ((dn = multilist_sublist_head(list)) != NULL) { |
572e2857 | 2096 | int flags; |
9babb374 | 2097 | ASSERT(!DMU_OBJECT_IS_SPECIAL(dn->dn_object)); |
9babb374 BB |
2098 | ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE || |
2099 | dn->dn_phys->dn_flags & | |
2100 | DNODE_FLAG_USERUSED_ACCOUNTED); | |
2101 | ||
572e2857 BB |
2102 | flags = dn->dn_id_flags; |
2103 | ASSERT(flags); | |
2104 | if (flags & DN_ID_OLD_EXIST) { | |
9c5167d1 NF |
2105 | do_userquota_update(os, &cache, dn->dn_oldused, |
2106 | dn->dn_oldflags, dn->dn_olduid, dn->dn_oldgid, | |
2107 | dn->dn_oldprojid, B_TRUE); | |
2108 | do_userobjquota_update(os, &cache, dn->dn_oldflags, | |
2109 | dn->dn_olduid, dn->dn_oldgid, | |
2110 | dn->dn_oldprojid, B_TRUE); | |
428870ff | 2111 | } |
572e2857 | 2112 | if (flags & DN_ID_NEW_EXIST) { |
9c5167d1 | 2113 | do_userquota_update(os, &cache, |
9b7a83cb | 2114 | DN_USED_BYTES(dn->dn_phys), dn->dn_phys->dn_flags, |
9c5167d1 NF |
2115 | dn->dn_newuid, dn->dn_newgid, |
2116 | dn->dn_newprojid, B_FALSE); | |
2117 | do_userobjquota_update(os, &cache, | |
2118 | dn->dn_phys->dn_flags, dn->dn_newuid, dn->dn_newgid, | |
2119 | dn->dn_newprojid, B_FALSE); | |
428870ff BB |
2120 | } |
2121 | ||
572e2857 | 2122 | mutex_enter(&dn->dn_mtx); |
428870ff BB |
2123 | dn->dn_oldused = 0; |
2124 | dn->dn_oldflags = 0; | |
2125 | if (dn->dn_id_flags & DN_ID_NEW_EXIST) { | |
2126 | dn->dn_olduid = dn->dn_newuid; | |
2127 | dn->dn_oldgid = dn->dn_newgid; | |
9c5167d1 | 2128 | dn->dn_oldprojid = dn->dn_newprojid; |
428870ff BB |
2129 | dn->dn_id_flags |= DN_ID_OLD_EXIST; |
2130 | if (dn->dn_bonuslen == 0) | |
2131 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
2132 | else | |
2133 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
2134 | } | |
2135 | dn->dn_id_flags &= ~(DN_ID_NEW_EXIST); | |
9babb374 BB |
2136 | mutex_exit(&dn->dn_mtx); |
2137 | ||
64fc7762 | 2138 | multilist_sublist_remove(list, dn); |
ffdf019c | 2139 | dnode_rele(dn, &os->os_synced_dnodes); |
9babb374 | 2140 | } |
9b7a83cb | 2141 | do_userquota_cacheflush(os, &cache, tx); |
64fc7762 MA |
2142 | multilist_sublist_unlock(list); |
2143 | kmem_free(uua, sizeof (*uua)); | |
2144 | } | |
2145 | ||
ba67d821 MA |
2146 | /* |
2147 | * Release dnode holds from dmu_objset_sync_dnodes(). When the dnode is being | |
2148 | * synced (i.e. we have issued the zio's for blocks in the dnode), it can't be | |
2149 | * evicted because the block containing the dnode can't be evicted until it is | |
2150 | * written out. However, this hold is necessary to prevent the dnode_t from | |
2151 | * being moved (via dnode_move()) while it's still referenced by | |
2152 | * dbuf_dirty_record_t:dr_dnode. And dr_dnode is needed for | |
2153 | * dirty_lightweight_leaf-type dirty records. | |
2154 | * | |
2155 | * If we are doing user-object accounting, the dnode_rele() happens from | |
2156 | * userquota_updates_task() instead. | |
2157 | */ | |
2158 | static void | |
2159 | dnode_rele_task(void *arg) | |
64fc7762 | 2160 | { |
ba67d821 MA |
2161 | userquota_updates_arg_t *uua = arg; |
2162 | objset_t *os = uua->uua_os; | |
2163 | ||
997f85b4 AM |
2164 | multilist_sublist_t *list = multilist_sublist_lock_idx( |
2165 | &os->os_synced_dnodes, uua->uua_sublist_idx); | |
fc754677 | 2166 | |
ba67d821 MA |
2167 | dnode_t *dn; |
2168 | while ((dn = multilist_sublist_head(list)) != NULL) { | |
2169 | multilist_sublist_remove(list, dn); | |
ffdf019c | 2170 | dnode_rele(dn, &os->os_synced_dnodes); |
ba67d821 MA |
2171 | } |
2172 | multilist_sublist_unlock(list); | |
2173 | kmem_free(uua, sizeof (*uua)); | |
2174 | } | |
2175 | ||
2176 | /* | |
2177 | * Return TRUE if userquota updates are needed. | |
2178 | */ | |
2179 | static boolean_t | |
2180 | dmu_objset_do_userquota_updates_prep(objset_t *os, dmu_tx_t *tx) | |
2181 | { | |
64fc7762 | 2182 | if (!dmu_objset_userused_enabled(os)) |
ba67d821 | 2183 | return (B_FALSE); |
64fc7762 | 2184 | |
163a8c28 TC |
2185 | /* |
2186 | * If this is a raw receive just return and handle accounting | |
2187 | * later when we have the keys loaded. We also don't do user | |
2188 | * accounting during claiming since the datasets are not owned | |
2189 | * for the duration of claiming and this txg should only be | |
2190 | * used for recovery. | |
2191 | */ | |
b5256303 | 2192 | if (os->os_encrypted && dmu_objset_is_receiving(os)) |
ba67d821 | 2193 | return (B_FALSE); |
b5256303 | 2194 | |
163a8c28 | 2195 | if (tx->tx_txg <= os->os_spa->spa_claim_max_txg) |
ba67d821 | 2196 | return (B_FALSE); |
163a8c28 | 2197 | |
9c5167d1 | 2198 | /* Allocate the user/group/project used objects if necessary. */ |
64fc7762 MA |
2199 | if (DMU_USERUSED_DNODE(os)->dn_type == DMU_OT_NONE) { |
2200 | VERIFY0(zap_create_claim(os, | |
2201 | DMU_USERUSED_OBJECT, | |
2202 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
2203 | VERIFY0(zap_create_claim(os, | |
2204 | DMU_GROUPUSED_OBJECT, | |
2205 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
2206 | } | |
2207 | ||
9c5167d1 NF |
2208 | if (dmu_objset_projectquota_enabled(os) && |
2209 | DMU_PROJECTUSED_DNODE(os)->dn_type == DMU_OT_NONE) { | |
2210 | VERIFY0(zap_create_claim(os, DMU_PROJECTUSED_OBJECT, | |
2211 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
2212 | } | |
ba67d821 MA |
2213 | return (B_TRUE); |
2214 | } | |
9c5167d1 | 2215 | |
ba67d821 MA |
2216 | /* |
2217 | * Dispatch taskq tasks to dp_sync_taskq to update the user accounting, and | |
2218 | * also release the holds on the dnodes from dmu_objset_sync_dnodes(). | |
2219 | * The caller must taskq_wait(dp_sync_taskq). | |
2220 | */ | |
2221 | void | |
2222 | dmu_objset_sync_done(objset_t *os, dmu_tx_t *tx) | |
2223 | { | |
2224 | boolean_t need_userquota = dmu_objset_do_userquota_updates_prep(os, tx); | |
2225 | ||
ffdf019c | 2226 | int num_sublists = multilist_get_num_sublists(&os->os_synced_dnodes); |
fc754677 | 2227 | for (int i = 0; i < num_sublists; i++) { |
64fc7762 MA |
2228 | userquota_updates_arg_t *uua = |
2229 | kmem_alloc(sizeof (*uua), KM_SLEEP); | |
2230 | uua->uua_os = os; | |
2231 | uua->uua_sublist_idx = i; | |
2232 | uua->uua_tx = tx; | |
ba67d821 MA |
2233 | |
2234 | /* | |
2235 | * If we don't need to update userquotas, use | |
2236 | * dnode_rele_task() to call dnode_rele() | |
2237 | */ | |
64fc7762 | 2238 | (void) taskq_dispatch(dmu_objset_pool(os)->dp_sync_taskq, |
ba67d821 MA |
2239 | need_userquota ? userquota_updates_task : dnode_rele_task, |
2240 | uua, 0); | |
64fc7762 MA |
2241 | /* callback frees uua */ |
2242 | } | |
9babb374 BB |
2243 | } |
2244 | ||
ba67d821 | 2245 | |
428870ff BB |
2246 | /* |
2247 | * Returns a pointer to data to find uid/gid from | |
2248 | * | |
2249 | * If a dirty record for transaction group that is syncing can't | |
2250 | * be found then NULL is returned. In the NULL case it is assumed | |
2251 | * the uid/gid aren't changing. | |
2252 | */ | |
2253 | static void * | |
2254 | dmu_objset_userquota_find_data(dmu_buf_impl_t *db, dmu_tx_t *tx) | |
2255 | { | |
cccbed9f | 2256 | dbuf_dirty_record_t *dr; |
428870ff BB |
2257 | void *data; |
2258 | ||
2259 | if (db->db_dirtycnt == 0) | |
2260 | return (db->db.db_data); /* Nothing is changing */ | |
2261 | ||
cccbed9f | 2262 | dr = dbuf_find_dirty_eq(db, tx->tx_txg); |
428870ff | 2263 | |
572e2857 | 2264 | if (dr == NULL) { |
428870ff | 2265 | data = NULL; |
572e2857 | 2266 | } else { |
ba67d821 | 2267 | if (dr->dr_dnode->dn_bonuslen == 0 && |
572e2857 BB |
2268 | dr->dr_dbuf->db_blkid == DMU_SPILL_BLKID) |
2269 | data = dr->dt.dl.dr_data->b_data; | |
2270 | else | |
2271 | data = dr->dt.dl.dr_data; | |
572e2857 BB |
2272 | } |
2273 | ||
428870ff BB |
2274 | return (data); |
2275 | } | |
2276 | ||
2277 | void | |
2278 | dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) | |
2279 | { | |
2280 | objset_t *os = dn->dn_objset; | |
2281 | void *data = NULL; | |
2282 | dmu_buf_impl_t *db = NULL; | |
428870ff BB |
2283 | int flags = dn->dn_id_flags; |
2284 | int error; | |
2285 | boolean_t have_spill = B_FALSE; | |
2286 | ||
2287 | if (!dmu_objset_userused_enabled(dn->dn_objset)) | |
2288 | return; | |
2289 | ||
b5256303 TC |
2290 | /* |
2291 | * Raw receives introduce a problem with user accounting. Raw | |
2292 | * receives cannot update the user accounting info because the | |
2293 | * user ids and the sizes are encrypted. To guarantee that we | |
2294 | * never end up with bad user accounting, we simply disable it | |
2295 | * during raw receives. We also disable this for normal receives | |
2296 | * so that an incremental raw receive may be done on top of an | |
2297 | * existing non-raw receive. | |
2298 | */ | |
2299 | if (os->os_encrypted && dmu_objset_is_receiving(os)) | |
2300 | return; | |
2301 | ||
428870ff BB |
2302 | if (before && (flags & (DN_ID_CHKED_BONUS|DN_ID_OLD_EXIST| |
2303 | DN_ID_CHKED_SPILL))) | |
2304 | return; | |
2305 | ||
2306 | if (before && dn->dn_bonuslen != 0) | |
2307 | data = DN_BONUS(dn->dn_phys); | |
2308 | else if (!before && dn->dn_bonuslen != 0) { | |
a3000f93 BB |
2309 | if (dn->dn_bonus) { |
2310 | db = dn->dn_bonus; | |
428870ff BB |
2311 | mutex_enter(&db->db_mtx); |
2312 | data = dmu_objset_userquota_find_data(db, tx); | |
2313 | } else { | |
2314 | data = DN_BONUS(dn->dn_phys); | |
2315 | } | |
2316 | } else if (dn->dn_bonuslen == 0 && dn->dn_bonustype == DMU_OT_SA) { | |
2317 | int rf = 0; | |
2318 | ||
2319 | if (RW_WRITE_HELD(&dn->dn_struct_rwlock)) | |
2320 | rf |= DB_RF_HAVESTRUCT; | |
572e2857 BB |
2321 | error = dmu_spill_hold_by_dnode(dn, |
2322 | rf | DB_RF_MUST_SUCCEED, | |
428870ff BB |
2323 | FTAG, (dmu_buf_t **)&db); |
2324 | ASSERT(error == 0); | |
2325 | mutex_enter(&db->db_mtx); | |
2326 | data = (before) ? db->db.db_data : | |
2327 | dmu_objset_userquota_find_data(db, tx); | |
2328 | have_spill = B_TRUE; | |
2329 | } else { | |
2330 | mutex_enter(&dn->dn_mtx); | |
2331 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
2332 | mutex_exit(&dn->dn_mtx); | |
2333 | return; | |
2334 | } | |
2335 | ||
428870ff BB |
2336 | /* |
2337 | * Must always call the callback in case the object | |
2338 | * type has changed and that type isn't an object type to track | |
2339 | */ | |
7bcb7f08 MA |
2340 | zfs_file_info_t zfi; |
2341 | error = file_cbs[os->os_phys->os_type](dn->dn_bonustype, data, &zfi); | |
2342 | ||
2343 | if (before) { | |
2344 | ASSERT(data); | |
2345 | dn->dn_olduid = zfi.zfi_user; | |
2346 | dn->dn_oldgid = zfi.zfi_group; | |
2347 | dn->dn_oldprojid = zfi.zfi_project; | |
2348 | } else if (data) { | |
2349 | dn->dn_newuid = zfi.zfi_user; | |
2350 | dn->dn_newgid = zfi.zfi_group; | |
2351 | dn->dn_newprojid = zfi.zfi_project; | |
2352 | } | |
428870ff BB |
2353 | |
2354 | /* | |
2355 | * Preserve existing uid/gid when the callback can't determine | |
2356 | * what the new uid/gid are and the callback returned EEXIST. | |
2357 | * The EEXIST error tells us to just use the existing uid/gid. | |
2358 | * If we don't know what the old values are then just assign | |
2359 | * them to 0, since that is a new file being created. | |
2360 | */ | |
2361 | if (!before && data == NULL && error == EEXIST) { | |
2362 | if (flags & DN_ID_OLD_EXIST) { | |
2363 | dn->dn_newuid = dn->dn_olduid; | |
2364 | dn->dn_newgid = dn->dn_oldgid; | |
2705ebf0 | 2365 | dn->dn_newprojid = dn->dn_oldprojid; |
428870ff BB |
2366 | } else { |
2367 | dn->dn_newuid = 0; | |
2368 | dn->dn_newgid = 0; | |
9c5167d1 | 2369 | dn->dn_newprojid = ZFS_DEFAULT_PROJID; |
428870ff BB |
2370 | } |
2371 | error = 0; | |
2372 | } | |
2373 | ||
2374 | if (db) | |
2375 | mutex_exit(&db->db_mtx); | |
2376 | ||
2377 | mutex_enter(&dn->dn_mtx); | |
2378 | if (error == 0 && before) | |
2379 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
2380 | if (error == 0 && !before) | |
2381 | dn->dn_id_flags |= DN_ID_NEW_EXIST; | |
2382 | ||
2383 | if (have_spill) { | |
2384 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
2385 | } else { | |
2386 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
2387 | } | |
2388 | mutex_exit(&dn->dn_mtx); | |
a3000f93 | 2389 | if (have_spill) |
428870ff BB |
2390 | dmu_buf_rele((dmu_buf_t *)db, FTAG); |
2391 | } | |
2392 | ||
9babb374 BB |
2393 | boolean_t |
2394 | dmu_objset_userspace_present(objset_t *os) | |
2395 | { | |
428870ff | 2396 | return (os->os_phys->os_flags & |
9babb374 BB |
2397 | OBJSET_FLAG_USERACCOUNTING_COMPLETE); |
2398 | } | |
2399 | ||
1de321e6 JX |
2400 | boolean_t |
2401 | dmu_objset_userobjspace_present(objset_t *os) | |
2402 | { | |
2403 | return (os->os_phys->os_flags & | |
2404 | OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE); | |
2405 | } | |
2406 | ||
9c5167d1 NF |
2407 | boolean_t |
2408 | dmu_objset_projectquota_present(objset_t *os) | |
2409 | { | |
2410 | return (os->os_phys->os_flags & | |
2411 | OBJSET_FLAG_PROJECTQUOTA_COMPLETE); | |
2412 | } | |
2413 | ||
1de321e6 JX |
2414 | static int |
2415 | dmu_objset_space_upgrade(objset_t *os) | |
9babb374 BB |
2416 | { |
2417 | uint64_t obj; | |
2418 | int err = 0; | |
2419 | ||
9babb374 BB |
2420 | /* |
2421 | * We simply need to mark every object dirty, so that it will be | |
2422 | * synced out and now accounted. If this is called | |
2423 | * concurrently, or if we already did some work before crashing, | |
2424 | * that's fine, since we track each object's accounted state | |
2425 | * independently. | |
2426 | */ | |
2427 | ||
2428 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
45d1cae3 | 2429 | dmu_tx_t *tx; |
9babb374 BB |
2430 | dmu_buf_t *db; |
2431 | int objerr; | |
2432 | ||
1de321e6 JX |
2433 | mutex_enter(&os->os_upgrade_lock); |
2434 | if (os->os_upgrade_exit) | |
2435 | err = SET_ERROR(EINTR); | |
2436 | mutex_exit(&os->os_upgrade_lock); | |
2437 | if (err != 0) | |
2438 | return (err); | |
2439 | ||
9babb374 | 2440 | if (issig(JUSTLOOKING) && issig(FORREAL)) |
2e528b49 | 2441 | return (SET_ERROR(EINTR)); |
9babb374 BB |
2442 | |
2443 | objerr = dmu_bonus_hold(os, obj, FTAG, &db); | |
13fe0198 | 2444 | if (objerr != 0) |
9babb374 | 2445 | continue; |
45d1cae3 | 2446 | tx = dmu_tx_create(os); |
9babb374 BB |
2447 | dmu_tx_hold_bonus(tx, obj); |
2448 | objerr = dmu_tx_assign(tx, TXG_WAIT); | |
13fe0198 | 2449 | if (objerr != 0) { |
d22323e8 | 2450 | dmu_buf_rele(db, FTAG); |
9babb374 BB |
2451 | dmu_tx_abort(tx); |
2452 | continue; | |
2453 | } | |
2454 | dmu_buf_will_dirty(db, tx); | |
2455 | dmu_buf_rele(db, FTAG); | |
2456 | dmu_tx_commit(tx); | |
2457 | } | |
1de321e6 JX |
2458 | return (0); |
2459 | } | |
2460 | ||
4072f465 | 2461 | static int |
2462 | dmu_objset_userspace_upgrade_cb(objset_t *os) | |
1de321e6 JX |
2463 | { |
2464 | int err = 0; | |
2465 | ||
2466 | if (dmu_objset_userspace_present(os)) | |
2467 | return (0); | |
2468 | if (dmu_objset_is_snapshot(os)) | |
2469 | return (SET_ERROR(EINVAL)); | |
2470 | if (!dmu_objset_userused_enabled(os)) | |
2471 | return (SET_ERROR(ENOTSUP)); | |
2472 | ||
2473 | err = dmu_objset_space_upgrade(os); | |
2474 | if (err) | |
2475 | return (err); | |
9babb374 | 2476 | |
428870ff | 2477 | os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; |
9babb374 BB |
2478 | txg_wait_synced(dmu_objset_pool(os), 0); |
2479 | return (0); | |
2480 | } | |
2481 | ||
4072f465 | 2482 | void |
2483 | dmu_objset_userspace_upgrade(objset_t *os) | |
2484 | { | |
2485 | dmu_objset_upgrade(os, dmu_objset_userspace_upgrade_cb); | |
2486 | } | |
2487 | ||
1de321e6 | 2488 | static int |
9c5167d1 | 2489 | dmu_objset_id_quota_upgrade_cb(objset_t *os) |
1de321e6 JX |
2490 | { |
2491 | int err = 0; | |
2492 | ||
9c5167d1 NF |
2493 | if (dmu_objset_userobjspace_present(os) && |
2494 | dmu_objset_projectquota_present(os)) | |
1de321e6 JX |
2495 | return (0); |
2496 | if (dmu_objset_is_snapshot(os)) | |
2497 | return (SET_ERROR(EINVAL)); | |
4072f465 | 2498 | if (!dmu_objset_userused_enabled(os)) |
1de321e6 | 2499 | return (SET_ERROR(ENOTSUP)); |
9c5167d1 NF |
2500 | if (!dmu_objset_projectquota_enabled(os) && |
2501 | dmu_objset_userobjspace_present(os)) | |
2502 | return (SET_ERROR(ENOTSUP)); | |
1de321e6 | 2503 | |
0f32b1f7 GA |
2504 | if (dmu_objset_userobjused_enabled(os)) |
2505 | dmu_objset_ds(os)->ds_feature_activation[ | |
2506 | SPA_FEATURE_USEROBJ_ACCOUNTING] = (void *)B_TRUE; | |
2507 | if (dmu_objset_projectquota_enabled(os)) | |
2508 | dmu_objset_ds(os)->ds_feature_activation[ | |
2509 | SPA_FEATURE_PROJECT_QUOTA] = (void *)B_TRUE; | |
2510 | ||
1de321e6 JX |
2511 | err = dmu_objset_space_upgrade(os); |
2512 | if (err) | |
2513 | return (err); | |
2514 | ||
4072f465 | 2515 | os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; |
2516 | if (dmu_objset_userobjused_enabled(os)) | |
2517 | os->os_flags |= OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE; | |
9c5167d1 NF |
2518 | if (dmu_objset_projectquota_enabled(os)) |
2519 | os->os_flags |= OBJSET_FLAG_PROJECTQUOTA_COMPLETE; | |
2520 | ||
1de321e6 JX |
2521 | txg_wait_synced(dmu_objset_pool(os), 0); |
2522 | return (0); | |
2523 | } | |
2524 | ||
2525 | void | |
9c5167d1 | 2526 | dmu_objset_id_quota_upgrade(objset_t *os) |
1de321e6 | 2527 | { |
9c5167d1 | 2528 | dmu_objset_upgrade(os, dmu_objset_id_quota_upgrade_cb); |
1de321e6 JX |
2529 | } |
2530 | ||
126ae9f4 JX |
2531 | boolean_t |
2532 | dmu_objset_userobjspace_upgradable(objset_t *os) | |
2533 | { | |
2534 | return (dmu_objset_type(os) == DMU_OST_ZFS && | |
2535 | !dmu_objset_is_snapshot(os) && | |
2536 | dmu_objset_userobjused_enabled(os) && | |
bb1be77a | 2537 | !dmu_objset_userobjspace_present(os) && |
2538 | spa_writeable(dmu_objset_spa(os))); | |
126ae9f4 JX |
2539 | } |
2540 | ||
9c5167d1 NF |
2541 | boolean_t |
2542 | dmu_objset_projectquota_upgradable(objset_t *os) | |
2543 | { | |
2544 | return (dmu_objset_type(os) == DMU_OST_ZFS && | |
2545 | !dmu_objset_is_snapshot(os) && | |
2546 | dmu_objset_projectquota_enabled(os) && | |
bb1be77a | 2547 | !dmu_objset_projectquota_present(os) && |
2548 | spa_writeable(dmu_objset_spa(os))); | |
9c5167d1 NF |
2549 | } |
2550 | ||
34dc7c2f BB |
2551 | void |
2552 | dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, | |
2553 | uint64_t *usedobjsp, uint64_t *availobjsp) | |
2554 | { | |
428870ff | 2555 | dsl_dataset_space(os->os_dsl_dataset, refdbytesp, availbytesp, |
34dc7c2f BB |
2556 | usedobjsp, availobjsp); |
2557 | } | |
2558 | ||
2559 | uint64_t | |
2560 | dmu_objset_fsid_guid(objset_t *os) | |
2561 | { | |
428870ff | 2562 | return (dsl_dataset_fsid_guid(os->os_dsl_dataset)); |
34dc7c2f BB |
2563 | } |
2564 | ||
2565 | void | |
2566 | dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat) | |
2567 | { | |
428870ff BB |
2568 | stat->dds_type = os->os_phys->os_type; |
2569 | if (os->os_dsl_dataset) | |
2570 | dsl_dataset_fast_stat(os->os_dsl_dataset, stat); | |
34dc7c2f BB |
2571 | } |
2572 | ||
2573 | void | |
2574 | dmu_objset_stats(objset_t *os, nvlist_t *nv) | |
2575 | { | |
428870ff BB |
2576 | ASSERT(os->os_dsl_dataset || |
2577 | os->os_phys->os_type == DMU_OST_META); | |
34dc7c2f | 2578 | |
428870ff BB |
2579 | if (os->os_dsl_dataset != NULL) |
2580 | dsl_dataset_stats(os->os_dsl_dataset, nv); | |
34dc7c2f BB |
2581 | |
2582 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_TYPE, | |
428870ff | 2583 | os->os_phys->os_type); |
9babb374 BB |
2584 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERACCOUNTING, |
2585 | dmu_objset_userspace_present(os)); | |
34dc7c2f BB |
2586 | } |
2587 | ||
2588 | int | |
2589 | dmu_objset_is_snapshot(objset_t *os) | |
2590 | { | |
428870ff | 2591 | if (os->os_dsl_dataset != NULL) |
0c66c32d | 2592 | return (os->os_dsl_dataset->ds_is_snapshot); |
34dc7c2f BB |
2593 | else |
2594 | return (B_FALSE); | |
2595 | } | |
2596 | ||
2597 | int | |
4d55ea81 | 2598 | dmu_snapshot_realname(objset_t *os, const char *name, char *real, int maxlen, |
34dc7c2f BB |
2599 | boolean_t *conflict) |
2600 | { | |
428870ff | 2601 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
2602 | uint64_t ignored; |
2603 | ||
d683ddbb | 2604 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 2605 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2606 | |
2607 | return (zap_lookup_norm(ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb | 2608 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, name, 8, 1, &ignored, |
9b7b9cd3 | 2609 | MT_NORMALIZE, real, maxlen, conflict)); |
34dc7c2f BB |
2610 | } |
2611 | ||
2612 | int | |
2613 | dmu_snapshot_list_next(objset_t *os, int namelen, char *name, | |
2614 | uint64_t *idp, uint64_t *offp, boolean_t *case_conflict) | |
2615 | { | |
428870ff | 2616 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
2617 | zap_cursor_t cursor; |
2618 | zap_attribute_t attr; | |
2619 | ||
13fe0198 MA |
2620 | ASSERT(dsl_pool_config_held(dmu_objset_pool(os))); |
2621 | ||
d683ddbb | 2622 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 2623 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2624 | |
2625 | zap_cursor_init_serialized(&cursor, | |
2626 | ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb | 2627 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, *offp); |
34dc7c2f BB |
2628 | |
2629 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
2630 | zap_cursor_fini(&cursor); | |
2e528b49 | 2631 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2632 | } |
2633 | ||
2634 | if (strlen(attr.za_name) + 1 > namelen) { | |
2635 | zap_cursor_fini(&cursor); | |
2e528b49 | 2636 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
2637 | } |
2638 | ||
c9e319fa | 2639 | (void) strlcpy(name, attr.za_name, namelen); |
34dc7c2f BB |
2640 | if (idp) |
2641 | *idp = attr.za_first_integer; | |
2642 | if (case_conflict) | |
2643 | *case_conflict = attr.za_normalization_conflict; | |
2644 | zap_cursor_advance(&cursor); | |
2645 | *offp = zap_cursor_serialize(&cursor); | |
2646 | zap_cursor_fini(&cursor); | |
2647 | ||
2648 | return (0); | |
2649 | } | |
2650 | ||
ebe7e575 | 2651 | int |
6772fb67 | 2652 | dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *value) |
ebe7e575 | 2653 | { |
d1d7e268 | 2654 | return (dsl_dataset_snap_lookup(os->os_dsl_dataset, name, value)); |
ebe7e575 BB |
2655 | } |
2656 | ||
34dc7c2f BB |
2657 | int |
2658 | dmu_dir_list_next(objset_t *os, int namelen, char *name, | |
2659 | uint64_t *idp, uint64_t *offp) | |
2660 | { | |
428870ff | 2661 | dsl_dir_t *dd = os->os_dsl_dataset->ds_dir; |
34dc7c2f BB |
2662 | zap_cursor_t cursor; |
2663 | zap_attribute_t attr; | |
2664 | ||
2665 | /* there is no next dir on a snapshot! */ | |
428870ff | 2666 | if (os->os_dsl_dataset->ds_object != |
d683ddbb | 2667 | dsl_dir_phys(dd)->dd_head_dataset_obj) |
2e528b49 | 2668 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2669 | |
2670 | zap_cursor_init_serialized(&cursor, | |
2671 | dd->dd_pool->dp_meta_objset, | |
d683ddbb | 2672 | dsl_dir_phys(dd)->dd_child_dir_zapobj, *offp); |
34dc7c2f BB |
2673 | |
2674 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
2675 | zap_cursor_fini(&cursor); | |
2e528b49 | 2676 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2677 | } |
2678 | ||
2679 | if (strlen(attr.za_name) + 1 > namelen) { | |
2680 | zap_cursor_fini(&cursor); | |
2e528b49 | 2681 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
2682 | } |
2683 | ||
c9e319fa | 2684 | (void) strlcpy(name, attr.za_name, namelen); |
34dc7c2f BB |
2685 | if (idp) |
2686 | *idp = attr.za_first_integer; | |
2687 | zap_cursor_advance(&cursor); | |
2688 | *offp = zap_cursor_serialize(&cursor); | |
2689 | zap_cursor_fini(&cursor); | |
2690 | ||
2691 | return (0); | |
2692 | } | |
2693 | ||
9c43027b AJ |
2694 | typedef struct dmu_objset_find_ctx { |
2695 | taskq_t *dc_tq; | |
2696 | dsl_pool_t *dc_dp; | |
2697 | uint64_t dc_ddobj; | |
1149ba64 | 2698 | char *dc_ddname; /* last component of ddobj's name */ |
9c43027b AJ |
2699 | int (*dc_func)(dsl_pool_t *, dsl_dataset_t *, void *); |
2700 | void *dc_arg; | |
2701 | int dc_flags; | |
2702 | kmutex_t *dc_error_lock; | |
2703 | int *dc_error; | |
2704 | } dmu_objset_find_ctx_t; | |
2705 | ||
2706 | static void | |
2707 | dmu_objset_find_dp_impl(dmu_objset_find_ctx_t *dcp) | |
b128c09f | 2708 | { |
9c43027b | 2709 | dsl_pool_t *dp = dcp->dc_dp; |
13fe0198 MA |
2710 | dsl_dir_t *dd; |
2711 | dsl_dataset_t *ds; | |
2712 | zap_cursor_t zc; | |
2713 | zap_attribute_t *attr; | |
2714 | uint64_t thisobj; | |
9c43027b | 2715 | int err = 0; |
13fe0198 | 2716 | |
9c43027b AJ |
2717 | /* don't process if there already was an error */ |
2718 | if (*dcp->dc_error != 0) | |
2719 | goto out; | |
13fe0198 | 2720 | |
1149ba64 GM |
2721 | /* |
2722 | * Note: passing the name (dc_ddname) here is optional, but it | |
2723 | * improves performance because we don't need to call | |
2724 | * zap_value_search() to determine the name. | |
2725 | */ | |
2726 | err = dsl_dir_hold_obj(dp, dcp->dc_ddobj, dcp->dc_ddname, FTAG, &dd); | |
13fe0198 | 2727 | if (err != 0) |
9c43027b | 2728 | goto out; |
13fe0198 MA |
2729 | |
2730 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ | |
2731 | if (dd->dd_myname[0] == '$') { | |
2732 | dsl_dir_rele(dd, FTAG); | |
9c43027b | 2733 | goto out; |
13fe0198 MA |
2734 | } |
2735 | ||
d683ddbb | 2736 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 2737 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
13fe0198 MA |
2738 | |
2739 | /* | |
2740 | * Iterate over all children. | |
2741 | */ | |
9c43027b | 2742 | if (dcp->dc_flags & DS_FIND_CHILDREN) { |
13fe0198 | 2743 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
d683ddbb | 2744 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
13fe0198 MA |
2745 | zap_cursor_retrieve(&zc, attr) == 0; |
2746 | (void) zap_cursor_advance(&zc)) { | |
2747 | ASSERT3U(attr->za_integer_length, ==, | |
2748 | sizeof (uint64_t)); | |
2749 | ASSERT3U(attr->za_num_integers, ==, 1); | |
2750 | ||
1c27024e | 2751 | dmu_objset_find_ctx_t *child_dcp = |
1149ba64 | 2752 | kmem_alloc(sizeof (*child_dcp), KM_SLEEP); |
9c43027b AJ |
2753 | *child_dcp = *dcp; |
2754 | child_dcp->dc_ddobj = attr->za_first_integer; | |
1149ba64 | 2755 | child_dcp->dc_ddname = spa_strdup(attr->za_name); |
9c43027b AJ |
2756 | if (dcp->dc_tq != NULL) |
2757 | (void) taskq_dispatch(dcp->dc_tq, | |
2758 | dmu_objset_find_dp_cb, child_dcp, TQ_SLEEP); | |
2759 | else | |
2760 | dmu_objset_find_dp_impl(child_dcp); | |
13fe0198 MA |
2761 | } |
2762 | zap_cursor_fini(&zc); | |
13fe0198 MA |
2763 | } |
2764 | ||
2765 | /* | |
2766 | * Iterate over all snapshots. | |
2767 | */ | |
9c43027b | 2768 | if (dcp->dc_flags & DS_FIND_SNAPSHOTS) { |
13fe0198 MA |
2769 | dsl_dataset_t *ds; |
2770 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
2771 | ||
2772 | if (err == 0) { | |
d683ddbb JG |
2773 | uint64_t snapobj; |
2774 | ||
2775 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
13fe0198 MA |
2776 | dsl_dataset_rele(ds, FTAG); |
2777 | ||
2778 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
2779 | zap_cursor_retrieve(&zc, attr) == 0; | |
2780 | (void) zap_cursor_advance(&zc)) { | |
2781 | ASSERT3U(attr->za_integer_length, ==, | |
2782 | sizeof (uint64_t)); | |
2783 | ASSERT3U(attr->za_num_integers, ==, 1); | |
2784 | ||
2785 | err = dsl_dataset_hold_obj(dp, | |
2786 | attr->za_first_integer, FTAG, &ds); | |
2787 | if (err != 0) | |
2788 | break; | |
9c43027b | 2789 | err = dcp->dc_func(dp, ds, dcp->dc_arg); |
13fe0198 MA |
2790 | dsl_dataset_rele(ds, FTAG); |
2791 | if (err != 0) | |
2792 | break; | |
2793 | } | |
2794 | zap_cursor_fini(&zc); | |
2795 | } | |
2796 | } | |
2797 | ||
13fe0198 MA |
2798 | kmem_free(attr, sizeof (zap_attribute_t)); |
2799 | ||
1149ba64 GM |
2800 | if (err != 0) { |
2801 | dsl_dir_rele(dd, FTAG); | |
9c43027b | 2802 | goto out; |
1149ba64 | 2803 | } |
13fe0198 MA |
2804 | |
2805 | /* | |
2806 | * Apply to self. | |
2807 | */ | |
2808 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1149ba64 GM |
2809 | |
2810 | /* | |
2811 | * Note: we hold the dir while calling dsl_dataset_hold_obj() so | |
2812 | * that the dir will remain cached, and we won't have to re-instantiate | |
2813 | * it (which could be expensive due to finding its name via | |
2814 | * zap_value_search()). | |
2815 | */ | |
2816 | dsl_dir_rele(dd, FTAG); | |
13fe0198 | 2817 | if (err != 0) |
9c43027b AJ |
2818 | goto out; |
2819 | err = dcp->dc_func(dp, ds, dcp->dc_arg); | |
13fe0198 | 2820 | dsl_dataset_rele(ds, FTAG); |
9c43027b AJ |
2821 | |
2822 | out: | |
2823 | if (err != 0) { | |
2824 | mutex_enter(dcp->dc_error_lock); | |
2825 | /* only keep first error */ | |
2826 | if (*dcp->dc_error == 0) | |
2827 | *dcp->dc_error = err; | |
2828 | mutex_exit(dcp->dc_error_lock); | |
2829 | } | |
2830 | ||
1149ba64 GM |
2831 | if (dcp->dc_ddname != NULL) |
2832 | spa_strfree(dcp->dc_ddname); | |
9c43027b AJ |
2833 | kmem_free(dcp, sizeof (*dcp)); |
2834 | } | |
2835 | ||
2836 | static void | |
2837 | dmu_objset_find_dp_cb(void *arg) | |
2838 | { | |
2839 | dmu_objset_find_ctx_t *dcp = arg; | |
2840 | dsl_pool_t *dp = dcp->dc_dp; | |
2841 | ||
5e8cd5d1 AJ |
2842 | /* |
2843 | * We need to get a pool_config_lock here, as there are several | |
e1cfd73f | 2844 | * assert(pool_config_held) down the stack. Getting a lock via |
5e8cd5d1 AJ |
2845 | * dsl_pool_config_enter is risky, as it might be stalled by a |
2846 | * pending writer. This would deadlock, as the write lock can | |
2847 | * only be granted when our parent thread gives up the lock. | |
2848 | * The _prio interface gives us priority over a pending writer. | |
2849 | */ | |
2850 | dsl_pool_config_enter_prio(dp, FTAG); | |
9c43027b AJ |
2851 | |
2852 | dmu_objset_find_dp_impl(dcp); | |
2853 | ||
2854 | dsl_pool_config_exit(dp, FTAG); | |
2855 | } | |
2856 | ||
2857 | /* | |
2858 | * Find objsets under and including ddobj, call func(ds) on each. | |
2859 | * The order for the enumeration is completely undefined. | |
2860 | * func is called with dsl_pool_config held. | |
2861 | */ | |
2862 | int | |
2863 | dmu_objset_find_dp(dsl_pool_t *dp, uint64_t ddobj, | |
2864 | int func(dsl_pool_t *, dsl_dataset_t *, void *), void *arg, int flags) | |
2865 | { | |
2866 | int error = 0; | |
2867 | taskq_t *tq = NULL; | |
2868 | int ntasks; | |
2869 | dmu_objset_find_ctx_t *dcp; | |
2870 | kmutex_t err_lock; | |
2871 | ||
2872 | mutex_init(&err_lock, NULL, MUTEX_DEFAULT, NULL); | |
2873 | dcp = kmem_alloc(sizeof (*dcp), KM_SLEEP); | |
2874 | dcp->dc_tq = NULL; | |
2875 | dcp->dc_dp = dp; | |
2876 | dcp->dc_ddobj = ddobj; | |
1149ba64 | 2877 | dcp->dc_ddname = NULL; |
9c43027b AJ |
2878 | dcp->dc_func = func; |
2879 | dcp->dc_arg = arg; | |
2880 | dcp->dc_flags = flags; | |
2881 | dcp->dc_error_lock = &err_lock; | |
2882 | dcp->dc_error = &error; | |
2883 | ||
2884 | if ((flags & DS_FIND_SERIALIZE) || dsl_pool_config_held_writer(dp)) { | |
2885 | /* | |
2886 | * In case a write lock is held we can't make use of | |
2887 | * parallelism, as down the stack of the worker threads | |
2888 | * the lock is asserted via dsl_pool_config_held. | |
2889 | * In case of a read lock this is solved by getting a read | |
2890 | * lock in each worker thread, which isn't possible in case | |
2891 | * of a writer lock. So we fall back to the synchronous path | |
2892 | * here. | |
2893 | * In the future it might be possible to get some magic into | |
2894 | * dsl_pool_config_held in a way that it returns true for | |
2895 | * the worker threads so that a single lock held from this | |
2896 | * thread suffices. For now, stay single threaded. | |
2897 | */ | |
2898 | dmu_objset_find_dp_impl(dcp); | |
e5676636 | 2899 | mutex_destroy(&err_lock); |
9c43027b AJ |
2900 | |
2901 | return (error); | |
2902 | } | |
2903 | ||
2904 | ntasks = dmu_find_threads; | |
2905 | if (ntasks == 0) | |
2906 | ntasks = vdev_count_leaves(dp->dp_spa) * 4; | |
1229323d | 2907 | tq = taskq_create("dmu_objset_find", ntasks, maxclsyspri, ntasks, |
9c43027b AJ |
2908 | INT_MAX, 0); |
2909 | if (tq == NULL) { | |
2910 | kmem_free(dcp, sizeof (*dcp)); | |
e5676636 BB |
2911 | mutex_destroy(&err_lock); |
2912 | ||
9c43027b AJ |
2913 | return (SET_ERROR(ENOMEM)); |
2914 | } | |
2915 | dcp->dc_tq = tq; | |
2916 | ||
2917 | /* dcp will be freed by task */ | |
2918 | (void) taskq_dispatch(tq, dmu_objset_find_dp_cb, dcp, TQ_SLEEP); | |
2919 | ||
2920 | /* | |
2921 | * PORTING: this code relies on the property of taskq_wait to wait | |
2922 | * until no more tasks are queued and no more tasks are active. As | |
2923 | * we always queue new tasks from within other tasks, task_wait | |
2924 | * reliably waits for the full recursion to finish, even though we | |
2925 | * enqueue new tasks after taskq_wait has been called. | |
2926 | * On platforms other than illumos, taskq_wait may not have this | |
2927 | * property. | |
2928 | */ | |
2929 | taskq_wait(tq); | |
2930 | taskq_destroy(tq); | |
2931 | mutex_destroy(&err_lock); | |
2932 | ||
2933 | return (error); | |
b128c09f BB |
2934 | } |
2935 | ||
2936 | /* | |
13fe0198 MA |
2937 | * Find all objsets under name, and for each, call 'func(child_name, arg)'. |
2938 | * The dp_config_rwlock must not be held when this is called, and it | |
2939 | * will not be held when the callback is called. | |
2940 | * Therefore this function should only be used when the pool is not changing | |
2941 | * (e.g. in syncing context), or the callback can deal with the possible races. | |
b128c09f | 2942 | */ |
13fe0198 MA |
2943 | static int |
2944 | dmu_objset_find_impl(spa_t *spa, const char *name, | |
2945 | int func(const char *, void *), void *arg, int flags) | |
34dc7c2f BB |
2946 | { |
2947 | dsl_dir_t *dd; | |
13fe0198 | 2948 | dsl_pool_t *dp = spa_get_dsl(spa); |
b128c09f | 2949 | dsl_dataset_t *ds; |
34dc7c2f BB |
2950 | zap_cursor_t zc; |
2951 | zap_attribute_t *attr; | |
2952 | char *child; | |
b128c09f BB |
2953 | uint64_t thisobj; |
2954 | int err; | |
34dc7c2f | 2955 | |
13fe0198 MA |
2956 | dsl_pool_config_enter(dp, FTAG); |
2957 | ||
2958 | err = dsl_dir_hold(dp, name, FTAG, &dd, NULL); | |
2959 | if (err != 0) { | |
2960 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f | 2961 | return (err); |
13fe0198 | 2962 | } |
34dc7c2f | 2963 | |
b128c09f BB |
2964 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ |
2965 | if (dd->dd_myname[0] == '$') { | |
13fe0198 MA |
2966 | dsl_dir_rele(dd, FTAG); |
2967 | dsl_pool_config_exit(dp, FTAG); | |
b128c09f BB |
2968 | return (0); |
2969 | } | |
2970 | ||
d683ddbb | 2971 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 2972 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
34dc7c2f BB |
2973 | |
2974 | /* | |
2975 | * Iterate over all children. | |
2976 | */ | |
2977 | if (flags & DS_FIND_CHILDREN) { | |
b128c09f | 2978 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
d683ddbb | 2979 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
34dc7c2f BB |
2980 | zap_cursor_retrieve(&zc, attr) == 0; |
2981 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 MA |
2982 | ASSERT3U(attr->za_integer_length, ==, |
2983 | sizeof (uint64_t)); | |
2984 | ASSERT3U(attr->za_num_integers, ==, 1); | |
34dc7c2f | 2985 | |
428870ff | 2986 | child = kmem_asprintf("%s/%s", name, attr->za_name); |
13fe0198 MA |
2987 | dsl_pool_config_exit(dp, FTAG); |
2988 | err = dmu_objset_find_impl(spa, child, | |
2989 | func, arg, flags); | |
2990 | dsl_pool_config_enter(dp, FTAG); | |
e4f5fa12 | 2991 | kmem_strfree(child); |
13fe0198 | 2992 | if (err != 0) |
34dc7c2f BB |
2993 | break; |
2994 | } | |
2995 | zap_cursor_fini(&zc); | |
2996 | ||
13fe0198 MA |
2997 | if (err != 0) { |
2998 | dsl_dir_rele(dd, FTAG); | |
2999 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f BB |
3000 | kmem_free(attr, sizeof (zap_attribute_t)); |
3001 | return (err); | |
3002 | } | |
3003 | } | |
3004 | ||
3005 | /* | |
3006 | * Iterate over all snapshots. | |
3007 | */ | |
b128c09f | 3008 | if (flags & DS_FIND_SNAPSHOTS) { |
b128c09f | 3009 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); |
b128c09f BB |
3010 | |
3011 | if (err == 0) { | |
d683ddbb JG |
3012 | uint64_t snapobj; |
3013 | ||
3014 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
b128c09f BB |
3015 | dsl_dataset_rele(ds, FTAG); |
3016 | ||
3017 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
3018 | zap_cursor_retrieve(&zc, attr) == 0; | |
3019 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 | 3020 | ASSERT3U(attr->za_integer_length, ==, |
b128c09f | 3021 | sizeof (uint64_t)); |
13fe0198 | 3022 | ASSERT3U(attr->za_num_integers, ==, 1); |
b128c09f | 3023 | |
428870ff BB |
3024 | child = kmem_asprintf("%s@%s", |
3025 | name, attr->za_name); | |
13fe0198 MA |
3026 | dsl_pool_config_exit(dp, FTAG); |
3027 | err = func(child, arg); | |
3028 | dsl_pool_config_enter(dp, FTAG); | |
e4f5fa12 | 3029 | kmem_strfree(child); |
13fe0198 | 3030 | if (err != 0) |
b128c09f BB |
3031 | break; |
3032 | } | |
3033 | zap_cursor_fini(&zc); | |
34dc7c2f | 3034 | } |
34dc7c2f BB |
3035 | } |
3036 | ||
13fe0198 | 3037 | dsl_dir_rele(dd, FTAG); |
34dc7c2f | 3038 | kmem_free(attr, sizeof (zap_attribute_t)); |
13fe0198 | 3039 | dsl_pool_config_exit(dp, FTAG); |
34dc7c2f | 3040 | |
13fe0198 | 3041 | if (err != 0) |
34dc7c2f BB |
3042 | return (err); |
3043 | ||
13fe0198 MA |
3044 | /* Apply to self. */ |
3045 | return (func(name, arg)); | |
34dc7c2f BB |
3046 | } |
3047 | ||
13fe0198 MA |
3048 | /* |
3049 | * See comment above dmu_objset_find_impl(). | |
3050 | */ | |
d164b209 | 3051 | int |
5df7e9d8 | 3052 | dmu_objset_find(const char *name, int func(const char *, void *), void *arg, |
13fe0198 | 3053 | int flags) |
d164b209 | 3054 | { |
13fe0198 MA |
3055 | spa_t *spa; |
3056 | int error; | |
d164b209 | 3057 | |
13fe0198 MA |
3058 | error = spa_open(name, &spa, FTAG); |
3059 | if (error != 0) | |
3060 | return (error); | |
3061 | error = dmu_objset_find_impl(spa, name, func, arg, flags); | |
3062 | spa_close(spa, FTAG); | |
3063 | return (error); | |
d164b209 BB |
3064 | } |
3065 | ||
ae76f45c TC |
3066 | boolean_t |
3067 | dmu_objset_incompatible_encryption_version(objset_t *os) | |
3068 | { | |
3069 | return (dsl_dir_incompatible_encryption_version( | |
3070 | os->os_dsl_dataset->ds_dir)); | |
3071 | } | |
3072 | ||
34dc7c2f BB |
3073 | void |
3074 | dmu_objset_set_user(objset_t *os, void *user_ptr) | |
3075 | { | |
428870ff BB |
3076 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
3077 | os->os_user_ptr = user_ptr; | |
34dc7c2f BB |
3078 | } |
3079 | ||
3080 | void * | |
3081 | dmu_objset_get_user(objset_t *os) | |
3082 | { | |
428870ff BB |
3083 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
3084 | return (os->os_user_ptr); | |
34dc7c2f | 3085 | } |
c28b2279 | 3086 | |
13fe0198 MA |
3087 | /* |
3088 | * Determine name of filesystem, given name of snapshot. | |
eca7b760 | 3089 | * buf must be at least ZFS_MAX_DATASET_NAME_LEN bytes |
13fe0198 MA |
3090 | */ |
3091 | int | |
3092 | dmu_fsname(const char *snapname, char *buf) | |
3093 | { | |
3094 | char *atp = strchr(snapname, '@'); | |
3095 | if (atp == NULL) | |
2e528b49 | 3096 | return (SET_ERROR(EINVAL)); |
eca7b760 | 3097 | if (atp - snapname >= ZFS_MAX_DATASET_NAME_LEN) |
2e528b49 | 3098 | return (SET_ERROR(ENAMETOOLONG)); |
13fe0198 MA |
3099 | (void) strlcpy(buf, snapname, atp - snapname + 1); |
3100 | return (0); | |
3101 | } | |
3102 | ||
3ec3bc21 | 3103 | /* |
0f8ff49e SD |
3104 | * Call when we think we're going to write/free space in open context |
3105 | * to track the amount of dirty data in the open txg, which is also the | |
3106 | * amount of memory that can not be evicted until this txg syncs. | |
3107 | * | |
3108 | * Note that there are two conditions where this can be called from | |
3109 | * syncing context: | |
3110 | * | |
3111 | * [1] When we just created the dataset, in which case we go on with | |
3112 | * updating any accounting of dirty data as usual. | |
3113 | * [2] When we are dirtying MOS data, in which case we only update the | |
3114 | * pool's accounting of dirty data. | |
3ec3bc21 BB |
3115 | */ |
3116 | void | |
3117 | dmu_objset_willuse_space(objset_t *os, int64_t space, dmu_tx_t *tx) | |
3118 | { | |
3119 | dsl_dataset_t *ds = os->os_dsl_dataset; | |
3120 | int64_t aspace = spa_get_worst_case_asize(os->os_spa, space); | |
3121 | ||
3122 | if (ds != NULL) { | |
3123 | dsl_dir_willuse_space(ds->ds_dir, aspace, tx); | |
3ec3bc21 | 3124 | } |
0f8ff49e SD |
3125 | |
3126 | dsl_pool_dirty_space(dmu_tx_pool(tx), space, tx); | |
3ec3bc21 BB |
3127 | } |
3128 | ||
93ce2b4c | 3129 | #if defined(_KERNEL) |
f0fd83be | 3130 | EXPORT_SYMBOL(dmu_objset_zil); |
c28b2279 | 3131 | EXPORT_SYMBOL(dmu_objset_pool); |
f0fd83be BB |
3132 | EXPORT_SYMBOL(dmu_objset_ds); |
3133 | EXPORT_SYMBOL(dmu_objset_type); | |
c28b2279 BB |
3134 | EXPORT_SYMBOL(dmu_objset_name); |
3135 | EXPORT_SYMBOL(dmu_objset_hold); | |
b5256303 | 3136 | EXPORT_SYMBOL(dmu_objset_hold_flags); |
c28b2279 BB |
3137 | EXPORT_SYMBOL(dmu_objset_own); |
3138 | EXPORT_SYMBOL(dmu_objset_rele); | |
b5256303 | 3139 | EXPORT_SYMBOL(dmu_objset_rele_flags); |
c28b2279 BB |
3140 | EXPORT_SYMBOL(dmu_objset_disown); |
3141 | EXPORT_SYMBOL(dmu_objset_from_ds); | |
3142 | EXPORT_SYMBOL(dmu_objset_create); | |
3143 | EXPORT_SYMBOL(dmu_objset_clone); | |
c28b2279 BB |
3144 | EXPORT_SYMBOL(dmu_objset_stats); |
3145 | EXPORT_SYMBOL(dmu_objset_fast_stat); | |
54a179e7 | 3146 | EXPORT_SYMBOL(dmu_objset_spa); |
c28b2279 BB |
3147 | EXPORT_SYMBOL(dmu_objset_space); |
3148 | EXPORT_SYMBOL(dmu_objset_fsid_guid); | |
3149 | EXPORT_SYMBOL(dmu_objset_find); | |
c28b2279 BB |
3150 | EXPORT_SYMBOL(dmu_objset_byteswap); |
3151 | EXPORT_SYMBOL(dmu_objset_evict_dbufs); | |
3152 | EXPORT_SYMBOL(dmu_objset_snap_cmtime); | |
50c957f7 | 3153 | EXPORT_SYMBOL(dmu_objset_dnodesize); |
c28b2279 BB |
3154 | |
3155 | EXPORT_SYMBOL(dmu_objset_sync); | |
3156 | EXPORT_SYMBOL(dmu_objset_is_dirty); | |
b5256303 | 3157 | EXPORT_SYMBOL(dmu_objset_create_impl_dnstats); |
c28b2279 BB |
3158 | EXPORT_SYMBOL(dmu_objset_create_impl); |
3159 | EXPORT_SYMBOL(dmu_objset_open_impl); | |
3160 | EXPORT_SYMBOL(dmu_objset_evict); | |
3161 | EXPORT_SYMBOL(dmu_objset_register_type); | |
ba67d821 | 3162 | EXPORT_SYMBOL(dmu_objset_sync_done); |
c28b2279 BB |
3163 | EXPORT_SYMBOL(dmu_objset_userquota_get_ids); |
3164 | EXPORT_SYMBOL(dmu_objset_userused_enabled); | |
3165 | EXPORT_SYMBOL(dmu_objset_userspace_upgrade); | |
3166 | EXPORT_SYMBOL(dmu_objset_userspace_present); | |
1de321e6 | 3167 | EXPORT_SYMBOL(dmu_objset_userobjused_enabled); |
126ae9f4 | 3168 | EXPORT_SYMBOL(dmu_objset_userobjspace_upgradable); |
1de321e6 | 3169 | EXPORT_SYMBOL(dmu_objset_userobjspace_present); |
9c5167d1 NF |
3170 | EXPORT_SYMBOL(dmu_objset_projectquota_enabled); |
3171 | EXPORT_SYMBOL(dmu_objset_projectquota_present); | |
3172 | EXPORT_SYMBOL(dmu_objset_projectquota_upgradable); | |
3173 | EXPORT_SYMBOL(dmu_objset_id_quota_upgrade); | |
c28b2279 | 3174 | #endif |