]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
1 | /* |
2 | * CDDL HEADER START | |
3 | * | |
4 | * The contents of this file are subject to the terms of the | |
5 | * Common Development and Distribution License (the "License"). | |
6 | * You may not use this file except in compliance with the License. | |
7 | * | |
8 | * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE | |
9 | * or http://www.opensolaris.org/os/licensing. | |
10 | * See the License for the specific language governing permissions | |
11 | * and limitations under the License. | |
12 | * | |
13 | * When distributing Covered Code, include this CDDL HEADER in each | |
14 | * file and include the License file at usr/src/OPENSOLARIS.LICENSE. | |
15 | * If applicable, add the following below this CDDL HEADER, with the | |
16 | * fields enclosed by brackets "[]" replaced with your own identifying | |
17 | * information: Portions Copyright [yyyy] [name of copyright owner] | |
18 | * | |
19 | * CDDL HEADER END | |
20 | */ | |
9b7b9cd3 | 21 | |
34dc7c2f | 22 | /* |
428870ff | 23 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
d52d80b7 | 24 | * Copyright (c) 2012, 2018 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. |
34dc7c2f BB |
33 | */ |
34 | ||
428870ff BB |
35 | /* Portions Copyright 2010 Robert Milkowski */ |
36 | ||
1de321e6 | 37 | #include <sys/zfeature.h> |
34dc7c2f BB |
38 | #include <sys/cred.h> |
39 | #include <sys/zfs_context.h> | |
40 | #include <sys/dmu_objset.h> | |
41 | #include <sys/dsl_dir.h> | |
42 | #include <sys/dsl_dataset.h> | |
43 | #include <sys/dsl_prop.h> | |
44 | #include <sys/dsl_pool.h> | |
45 | #include <sys/dsl_synctask.h> | |
46 | #include <sys/dsl_deleg.h> | |
47 | #include <sys/dnode.h> | |
48 | #include <sys/dbuf.h> | |
49 | #include <sys/zvol.h> | |
50 | #include <sys/dmu_tx.h> | |
34dc7c2f BB |
51 | #include <sys/zap.h> |
52 | #include <sys/zil.h> | |
53 | #include <sys/dmu_impl.h> | |
54 | #include <sys/zfs_ioctl.h> | |
428870ff | 55 | #include <sys/sa.h> |
572e2857 | 56 | #include <sys/zfs_onexit.h> |
13fe0198 | 57 | #include <sys/dsl_destroy.h> |
9c43027b | 58 | #include <sys/vdev.h> |
a1d477c2 | 59 | #include <sys/zfeature.h> |
f74b821a | 60 | #include <sys/policy.h> |
1de321e6 | 61 | #include <sys/spa_impl.h> |
03916905 | 62 | #include <sys/dmu_recv.h> |
9c5167d1 | 63 | #include <sys/zfs_project.h> |
a7ed98d8 | 64 | #include "zfs_namecheck.h" |
572e2857 BB |
65 | |
66 | /* | |
67 | * Needed to close a window in dnode_move() that allows the objset to be freed | |
68 | * before it can be safely accessed. | |
69 | */ | |
70 | krwlock_t os_lock; | |
71 | ||
9c43027b | 72 | /* |
4e33ba4c | 73 | * Tunable to overwrite the maximum number of threads for the parallelization |
9c43027b AJ |
74 | * of dmu_objset_find_dp, needed to speed up the import of pools with many |
75 | * datasets. | |
76 | * Default is 4 times the number of leaf vdevs. | |
77 | */ | |
78 | int dmu_find_threads = 0; | |
79 | ||
68cbd56e NB |
80 | /* |
81 | * Backfill lower metadnode objects after this many have been freed. | |
82 | * Backfilling negatively impacts object creation rates, so only do it | |
83 | * if there are enough holes to fill. | |
84 | */ | |
85 | int dmu_rescan_dnode_threshold = 1 << DN_MAX_INDBLKSHIFT; | |
86 | ||
c0daec32 AB |
87 | static char *upgrade_tag = "upgrade_tag"; |
88 | ||
9c43027b AJ |
89 | static void dmu_objset_find_dp_cb(void *arg); |
90 | ||
1de321e6 JX |
91 | static void dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb); |
92 | static void dmu_objset_upgrade_stop(objset_t *os); | |
93 | ||
572e2857 BB |
94 | void |
95 | dmu_objset_init(void) | |
96 | { | |
97 | rw_init(&os_lock, NULL, RW_DEFAULT, NULL); | |
98 | } | |
99 | ||
100 | void | |
101 | dmu_objset_fini(void) | |
102 | { | |
103 | rw_destroy(&os_lock); | |
104 | } | |
34dc7c2f BB |
105 | |
106 | spa_t * | |
107 | dmu_objset_spa(objset_t *os) | |
108 | { | |
428870ff | 109 | return (os->os_spa); |
34dc7c2f BB |
110 | } |
111 | ||
112 | zilog_t * | |
113 | dmu_objset_zil(objset_t *os) | |
114 | { | |
428870ff | 115 | return (os->os_zil); |
34dc7c2f BB |
116 | } |
117 | ||
118 | dsl_pool_t * | |
119 | dmu_objset_pool(objset_t *os) | |
120 | { | |
121 | dsl_dataset_t *ds; | |
122 | ||
428870ff | 123 | if ((ds = os->os_dsl_dataset) != NULL && ds->ds_dir) |
34dc7c2f BB |
124 | return (ds->ds_dir->dd_pool); |
125 | else | |
428870ff | 126 | return (spa_get_dsl(os->os_spa)); |
34dc7c2f BB |
127 | } |
128 | ||
129 | dsl_dataset_t * | |
130 | dmu_objset_ds(objset_t *os) | |
131 | { | |
428870ff | 132 | return (os->os_dsl_dataset); |
34dc7c2f BB |
133 | } |
134 | ||
135 | dmu_objset_type_t | |
136 | dmu_objset_type(objset_t *os) | |
137 | { | |
428870ff | 138 | return (os->os_phys->os_type); |
34dc7c2f BB |
139 | } |
140 | ||
141 | void | |
142 | dmu_objset_name(objset_t *os, char *buf) | |
143 | { | |
428870ff | 144 | dsl_dataset_name(os->os_dsl_dataset, buf); |
34dc7c2f BB |
145 | } |
146 | ||
147 | uint64_t | |
148 | dmu_objset_id(objset_t *os) | |
149 | { | |
428870ff | 150 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
151 | |
152 | return (ds ? ds->ds_object : 0); | |
153 | } | |
154 | ||
50c957f7 NB |
155 | uint64_t |
156 | dmu_objset_dnodesize(objset_t *os) | |
157 | { | |
158 | return (os->os_dnodesize); | |
159 | } | |
160 | ||
faf0f58c | 161 | zfs_sync_type_t |
428870ff BB |
162 | dmu_objset_syncprop(objset_t *os) |
163 | { | |
164 | return (os->os_sync); | |
165 | } | |
166 | ||
faf0f58c | 167 | zfs_logbias_op_t |
428870ff BB |
168 | dmu_objset_logbias(objset_t *os) |
169 | { | |
170 | return (os->os_logbias); | |
171 | } | |
172 | ||
34dc7c2f BB |
173 | static void |
174 | checksum_changed_cb(void *arg, uint64_t newval) | |
175 | { | |
428870ff | 176 | objset_t *os = arg; |
34dc7c2f BB |
177 | |
178 | /* | |
179 | * Inheritance should have been done by now. | |
180 | */ | |
181 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
182 | ||
428870ff | 183 | os->os_checksum = zio_checksum_select(newval, ZIO_CHECKSUM_ON_VALUE); |
34dc7c2f BB |
184 | } |
185 | ||
186 | static void | |
187 | compression_changed_cb(void *arg, uint64_t newval) | |
188 | { | |
428870ff | 189 | objset_t *os = arg; |
34dc7c2f BB |
190 | |
191 | /* | |
192 | * Inheritance and range checking should have been done by now. | |
193 | */ | |
194 | ASSERT(newval != ZIO_COMPRESS_INHERIT); | |
195 | ||
99197f03 JG |
196 | os->os_compress = zio_compress_select(os->os_spa, newval, |
197 | ZIO_COMPRESS_ON); | |
34dc7c2f BB |
198 | } |
199 | ||
200 | static void | |
201 | copies_changed_cb(void *arg, uint64_t newval) | |
202 | { | |
428870ff | 203 | objset_t *os = arg; |
34dc7c2f BB |
204 | |
205 | /* | |
206 | * Inheritance and range checking should have been done by now. | |
207 | */ | |
208 | ASSERT(newval > 0); | |
428870ff | 209 | ASSERT(newval <= spa_max_replication(os->os_spa)); |
34dc7c2f | 210 | |
428870ff BB |
211 | os->os_copies = newval; |
212 | } | |
213 | ||
214 | static void | |
215 | dedup_changed_cb(void *arg, uint64_t newval) | |
216 | { | |
217 | objset_t *os = arg; | |
218 | spa_t *spa = os->os_spa; | |
219 | enum zio_checksum checksum; | |
220 | ||
221 | /* | |
222 | * Inheritance should have been done by now. | |
223 | */ | |
224 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
225 | ||
226 | checksum = zio_checksum_dedup_select(spa, newval, ZIO_CHECKSUM_OFF); | |
227 | ||
228 | os->os_dedup_checksum = checksum & ZIO_CHECKSUM_MASK; | |
229 | os->os_dedup_verify = !!(checksum & ZIO_CHECKSUM_VERIFY); | |
34dc7c2f BB |
230 | } |
231 | ||
b128c09f BB |
232 | static void |
233 | primary_cache_changed_cb(void *arg, uint64_t newval) | |
234 | { | |
428870ff | 235 | objset_t *os = arg; |
b128c09f BB |
236 | |
237 | /* | |
238 | * Inheritance and range checking should have been done by now. | |
239 | */ | |
240 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
241 | newval == ZFS_CACHE_METADATA); | |
242 | ||
428870ff | 243 | os->os_primary_cache = newval; |
b128c09f BB |
244 | } |
245 | ||
246 | static void | |
247 | secondary_cache_changed_cb(void *arg, uint64_t newval) | |
248 | { | |
428870ff | 249 | objset_t *os = arg; |
b128c09f BB |
250 | |
251 | /* | |
252 | * Inheritance and range checking should have been done by now. | |
253 | */ | |
254 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
255 | newval == ZFS_CACHE_METADATA); | |
256 | ||
428870ff BB |
257 | os->os_secondary_cache = newval; |
258 | } | |
259 | ||
260 | static void | |
261 | sync_changed_cb(void *arg, uint64_t newval) | |
262 | { | |
263 | objset_t *os = arg; | |
264 | ||
265 | /* | |
266 | * Inheritance and range checking should have been done by now. | |
267 | */ | |
268 | ASSERT(newval == ZFS_SYNC_STANDARD || newval == ZFS_SYNC_ALWAYS || | |
269 | newval == ZFS_SYNC_DISABLED); | |
270 | ||
271 | os->os_sync = newval; | |
272 | if (os->os_zil) | |
273 | zil_set_sync(os->os_zil, newval); | |
274 | } | |
275 | ||
faf0f58c MA |
276 | static void |
277 | redundant_metadata_changed_cb(void *arg, uint64_t newval) | |
278 | { | |
279 | objset_t *os = arg; | |
280 | ||
281 | /* | |
282 | * Inheritance and range checking should have been done by now. | |
283 | */ | |
284 | ASSERT(newval == ZFS_REDUNDANT_METADATA_ALL || | |
285 | newval == ZFS_REDUNDANT_METADATA_MOST); | |
286 | ||
287 | os->os_redundant_metadata = newval; | |
288 | } | |
289 | ||
50c957f7 NB |
290 | static void |
291 | dnodesize_changed_cb(void *arg, uint64_t newval) | |
292 | { | |
293 | objset_t *os = arg; | |
294 | ||
295 | switch (newval) { | |
296 | case ZFS_DNSIZE_LEGACY: | |
297 | os->os_dnodesize = DNODE_MIN_SIZE; | |
298 | break; | |
299 | case ZFS_DNSIZE_AUTO: | |
300 | /* | |
301 | * Choose a dnode size that will work well for most | |
302 | * workloads if the user specified "auto". Future code | |
303 | * improvements could dynamically select a dnode size | |
304 | * based on observed workload patterns. | |
305 | */ | |
306 | os->os_dnodesize = DNODE_MIN_SIZE * 2; | |
307 | break; | |
308 | case ZFS_DNSIZE_1K: | |
309 | case ZFS_DNSIZE_2K: | |
310 | case ZFS_DNSIZE_4K: | |
311 | case ZFS_DNSIZE_8K: | |
312 | case ZFS_DNSIZE_16K: | |
313 | os->os_dnodesize = newval; | |
314 | break; | |
315 | } | |
316 | } | |
317 | ||
cc99f275 DB |
318 | static void |
319 | smallblk_changed_cb(void *arg, uint64_t newval) | |
320 | { | |
321 | objset_t *os = arg; | |
322 | ||
323 | /* | |
324 | * Inheritance and range checking should have been done by now. | |
325 | */ | |
326 | ASSERT(newval <= SPA_OLD_MAXBLOCKSIZE); | |
327 | ASSERT(ISP2(newval)); | |
328 | ||
329 | os->os_zpl_special_smallblock = newval; | |
330 | } | |
331 | ||
428870ff BB |
332 | static void |
333 | logbias_changed_cb(void *arg, uint64_t newval) | |
334 | { | |
335 | objset_t *os = arg; | |
336 | ||
337 | ASSERT(newval == ZFS_LOGBIAS_LATENCY || | |
338 | newval == ZFS_LOGBIAS_THROUGHPUT); | |
339 | os->os_logbias = newval; | |
340 | if (os->os_zil) | |
341 | zil_set_logbias(os->os_zil, newval); | |
b128c09f BB |
342 | } |
343 | ||
f1512ee6 MA |
344 | static void |
345 | recordsize_changed_cb(void *arg, uint64_t newval) | |
346 | { | |
347 | objset_t *os = arg; | |
348 | ||
349 | os->os_recordsize = newval; | |
350 | } | |
351 | ||
34dc7c2f BB |
352 | void |
353 | dmu_objset_byteswap(void *buf, size_t size) | |
354 | { | |
355 | objset_phys_t *osp = buf; | |
356 | ||
9c5167d1 NF |
357 | ASSERT(size == OBJSET_PHYS_SIZE_V1 || size == OBJSET_PHYS_SIZE_V2 || |
358 | size == sizeof (objset_phys_t)); | |
34dc7c2f BB |
359 | dnode_byteswap(&osp->os_meta_dnode); |
360 | byteswap_uint64_array(&osp->os_zil_header, sizeof (zil_header_t)); | |
361 | osp->os_type = BSWAP_64(osp->os_type); | |
9babb374 | 362 | osp->os_flags = BSWAP_64(osp->os_flags); |
9c5167d1 | 363 | if (size >= OBJSET_PHYS_SIZE_V2) { |
9babb374 BB |
364 | dnode_byteswap(&osp->os_userused_dnode); |
365 | dnode_byteswap(&osp->os_groupused_dnode); | |
9c5167d1 NF |
366 | if (size >= sizeof (objset_phys_t)) |
367 | dnode_byteswap(&osp->os_projectused_dnode); | |
9babb374 | 368 | } |
34dc7c2f BB |
369 | } |
370 | ||
64fc7762 MA |
371 | /* |
372 | * The hash is a CRC-based hash of the objset_t pointer and the object number. | |
373 | */ | |
374 | static uint64_t | |
375 | dnode_hash(const objset_t *os, uint64_t obj) | |
376 | { | |
377 | uintptr_t osv = (uintptr_t)os; | |
378 | uint64_t crc = -1ULL; | |
379 | ||
380 | ASSERT(zfs_crc64_table[128] == ZFS_CRC64_POLY); | |
381 | /* | |
382 | * The low 6 bits of the pointer don't have much entropy, because | |
383 | * the objset_t is larger than 2^6 bytes long. | |
384 | */ | |
385 | crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (osv >> 6)) & 0xFF]; | |
386 | crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 0)) & 0xFF]; | |
387 | crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 8)) & 0xFF]; | |
388 | crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 16)) & 0xFF]; | |
389 | ||
390 | crc ^= (osv>>14) ^ (obj>>24); | |
391 | ||
392 | return (crc); | |
393 | } | |
394 | ||
395 | unsigned int | |
396 | dnode_multilist_index_func(multilist_t *ml, void *obj) | |
397 | { | |
398 | dnode_t *dn = obj; | |
399 | return (dnode_hash(dn->dn_objset, dn->dn_object) % | |
400 | multilist_get_num_sublists(ml)); | |
401 | } | |
402 | ||
a1d477c2 MA |
403 | /* |
404 | * Instantiates the objset_t in-memory structure corresponding to the | |
405 | * objset_phys_t that's pointed to by the specified blkptr_t. | |
406 | */ | |
34dc7c2f BB |
407 | int |
408 | dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, | |
428870ff | 409 | objset_t **osp) |
34dc7c2f | 410 | { |
428870ff | 411 | objset_t *os; |
b128c09f | 412 | int i, err; |
34dc7c2f BB |
413 | |
414 | ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock)); | |
30af21b0 | 415 | ASSERT(!BP_IS_REDACTED(bp)); |
34dc7c2f | 416 | |
0fdd6106 MA |
417 | /* |
418 | * We need the pool config lock to get properties. | |
419 | */ | |
420 | ASSERT(ds == NULL || dsl_pool_config_held(ds->ds_dir->dd_pool)); | |
421 | ||
a1d477c2 MA |
422 | /* |
423 | * The $ORIGIN dataset (if it exists) doesn't have an associated | |
424 | * objset, so there's no reason to open it. The $ORIGIN dataset | |
425 | * will not exist on pools older than SPA_VERSION_ORIGIN. | |
426 | */ | |
427 | if (ds != NULL && spa_get_dsl(spa) != NULL && | |
428 | spa_get_dsl(spa)->dp_origin_snap != NULL) { | |
429 | ASSERT3P(ds->ds_dir, !=, | |
430 | spa_get_dsl(spa)->dp_origin_snap->ds_dir); | |
431 | } | |
432 | ||
79c76d5b | 433 | os = kmem_zalloc(sizeof (objset_t), KM_SLEEP); |
428870ff BB |
434 | os->os_dsl_dataset = ds; |
435 | os->os_spa = spa; | |
436 | os->os_rootbp = bp; | |
437 | if (!BP_IS_HOLE(os->os_rootbp)) { | |
2a432414 | 438 | arc_flags_t aflags = ARC_FLAG_WAIT; |
5dbd68a3 | 439 | zbookmark_phys_t zb; |
9c5167d1 | 440 | int size; |
b5256303 | 441 | enum zio_flag zio_flags = ZIO_FLAG_CANFAIL; |
428870ff BB |
442 | SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET, |
443 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
444 | ||
445 | if (DMU_OS_IS_L2CACHEABLE(os)) | |
2a432414 | 446 | aflags |= ARC_FLAG_L2CACHE; |
34dc7c2f | 447 | |
b5256303 TC |
448 | if (ds != NULL && ds->ds_dir->dd_crypto_obj != 0) { |
449 | ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF); | |
450 | ASSERT(BP_IS_AUTHENTICATED(bp)); | |
451 | zio_flags |= ZIO_FLAG_RAW; | |
452 | } | |
453 | ||
428870ff | 454 | dprintf_bp(os->os_rootbp, "reading %s", ""); |
294f6806 | 455 | err = arc_read(NULL, spa, os->os_rootbp, |
428870ff | 456 | arc_getbuf_func, &os->os_phys_buf, |
b5256303 | 457 | ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zb); |
13fe0198 | 458 | if (err != 0) { |
428870ff | 459 | kmem_free(os, sizeof (objset_t)); |
b128c09f BB |
460 | /* convert checksum errors into IO errors */ |
461 | if (err == ECKSUM) | |
2e528b49 | 462 | err = SET_ERROR(EIO); |
34dc7c2f BB |
463 | return (err); |
464 | } | |
9babb374 | 465 | |
9c5167d1 NF |
466 | if (spa_version(spa) < SPA_VERSION_USERSPACE) |
467 | size = OBJSET_PHYS_SIZE_V1; | |
468 | else if (!spa_feature_is_enabled(spa, | |
469 | SPA_FEATURE_PROJECT_QUOTA)) | |
470 | size = OBJSET_PHYS_SIZE_V2; | |
471 | else | |
472 | size = sizeof (objset_phys_t); | |
473 | ||
9babb374 | 474 | /* Increase the blocksize if we are permitted. */ |
9c5167d1 | 475 | if (arc_buf_size(os->os_phys_buf) < size) { |
2aa34383 | 476 | arc_buf_t *buf = arc_alloc_buf(spa, &os->os_phys_buf, |
9c5167d1 NF |
477 | ARC_BUFC_METADATA, size); |
478 | bzero(buf->b_data, size); | |
428870ff BB |
479 | bcopy(os->os_phys_buf->b_data, buf->b_data, |
480 | arc_buf_size(os->os_phys_buf)); | |
d3c2ae1c | 481 | arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); |
428870ff | 482 | os->os_phys_buf = buf; |
9babb374 BB |
483 | } |
484 | ||
428870ff BB |
485 | os->os_phys = os->os_phys_buf->b_data; |
486 | os->os_flags = os->os_phys->os_flags; | |
34dc7c2f | 487 | } else { |
9babb374 | 488 | int size = spa_version(spa) >= SPA_VERSION_USERSPACE ? |
9c5167d1 | 489 | sizeof (objset_phys_t) : OBJSET_PHYS_SIZE_V1; |
2aa34383 DK |
490 | os->os_phys_buf = arc_alloc_buf(spa, &os->os_phys_buf, |
491 | ARC_BUFC_METADATA, size); | |
428870ff BB |
492 | os->os_phys = os->os_phys_buf->b_data; |
493 | bzero(os->os_phys, size); | |
34dc7c2f | 494 | } |
2e5dc449 MA |
495 | /* |
496 | * These properties will be filled in by the logic in zfs_get_zplprop() | |
497 | * when they are queried for the first time. | |
498 | */ | |
499 | os->os_version = OBJSET_PROP_UNINITIALIZED; | |
500 | os->os_normalization = OBJSET_PROP_UNINITIALIZED; | |
501 | os->os_utf8only = OBJSET_PROP_UNINITIALIZED; | |
502 | os->os_casesensitivity = OBJSET_PROP_UNINITIALIZED; | |
34dc7c2f BB |
503 | |
504 | /* | |
505 | * Note: the changed_cb will be called once before the register | |
506 | * func returns, thus changing the checksum/compression from the | |
b128c09f BB |
507 | * default (fletcher2/off). Snapshots don't need to know about |
508 | * checksum/compression/copies. | |
34dc7c2f | 509 | */ |
9b67f605 | 510 | if (ds != NULL) { |
b5256303 TC |
511 | os->os_encrypted = (ds->ds_dir->dd_crypto_obj != 0); |
512 | ||
13fe0198 MA |
513 | err = dsl_prop_register(ds, |
514 | zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE), | |
428870ff | 515 | primary_cache_changed_cb, os); |
13fe0198 MA |
516 | if (err == 0) { |
517 | err = dsl_prop_register(ds, | |
518 | zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE), | |
428870ff | 519 | secondary_cache_changed_cb, os); |
13fe0198 | 520 | } |
0c66c32d | 521 | if (!ds->ds_is_snapshot) { |
13fe0198 MA |
522 | if (err == 0) { |
523 | err = dsl_prop_register(ds, | |
524 | zfs_prop_to_name(ZFS_PROP_CHECKSUM), | |
428870ff | 525 | checksum_changed_cb, os); |
13fe0198 MA |
526 | } |
527 | if (err == 0) { | |
528 | err = dsl_prop_register(ds, | |
529 | zfs_prop_to_name(ZFS_PROP_COMPRESSION), | |
428870ff | 530 | compression_changed_cb, os); |
13fe0198 MA |
531 | } |
532 | if (err == 0) { | |
533 | err = dsl_prop_register(ds, | |
534 | zfs_prop_to_name(ZFS_PROP_COPIES), | |
428870ff | 535 | copies_changed_cb, os); |
13fe0198 MA |
536 | } |
537 | if (err == 0) { | |
538 | err = dsl_prop_register(ds, | |
539 | zfs_prop_to_name(ZFS_PROP_DEDUP), | |
428870ff | 540 | dedup_changed_cb, os); |
13fe0198 MA |
541 | } |
542 | if (err == 0) { | |
543 | err = dsl_prop_register(ds, | |
544 | zfs_prop_to_name(ZFS_PROP_LOGBIAS), | |
428870ff | 545 | logbias_changed_cb, os); |
13fe0198 MA |
546 | } |
547 | if (err == 0) { | |
548 | err = dsl_prop_register(ds, | |
549 | zfs_prop_to_name(ZFS_PROP_SYNC), | |
428870ff | 550 | sync_changed_cb, os); |
13fe0198 | 551 | } |
faf0f58c MA |
552 | if (err == 0) { |
553 | err = dsl_prop_register(ds, | |
554 | zfs_prop_to_name( | |
555 | ZFS_PROP_REDUNDANT_METADATA), | |
556 | redundant_metadata_changed_cb, os); | |
557 | } | |
f1512ee6 MA |
558 | if (err == 0) { |
559 | err = dsl_prop_register(ds, | |
560 | zfs_prop_to_name(ZFS_PROP_RECORDSIZE), | |
561 | recordsize_changed_cb, os); | |
562 | } | |
50c957f7 NB |
563 | if (err == 0) { |
564 | err = dsl_prop_register(ds, | |
565 | zfs_prop_to_name(ZFS_PROP_DNODESIZE), | |
566 | dnodesize_changed_cb, os); | |
567 | } | |
cc99f275 DB |
568 | if (err == 0) { |
569 | err = dsl_prop_register(ds, | |
570 | zfs_prop_to_name( | |
571 | ZFS_PROP_SPECIAL_SMALL_BLOCKS), | |
572 | smallblk_changed_cb, os); | |
573 | } | |
b128c09f | 574 | } |
13fe0198 | 575 | if (err != 0) { |
d3c2ae1c | 576 | arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); |
428870ff | 577 | kmem_free(os, sizeof (objset_t)); |
34dc7c2f BB |
578 | return (err); |
579 | } | |
9b67f605 | 580 | } else { |
34dc7c2f | 581 | /* It's the meta-objset. */ |
428870ff | 582 | os->os_checksum = ZIO_CHECKSUM_FLETCHER_4; |
99197f03 | 583 | os->os_compress = ZIO_COMPRESS_ON; |
b5256303 | 584 | os->os_encrypted = B_FALSE; |
428870ff BB |
585 | os->os_copies = spa_max_replication(spa); |
586 | os->os_dedup_checksum = ZIO_CHECKSUM_OFF; | |
faf0f58c MA |
587 | os->os_dedup_verify = B_FALSE; |
588 | os->os_logbias = ZFS_LOGBIAS_LATENCY; | |
589 | os->os_sync = ZFS_SYNC_STANDARD; | |
428870ff BB |
590 | os->os_primary_cache = ZFS_CACHE_ALL; |
591 | os->os_secondary_cache = ZFS_CACHE_ALL; | |
50c957f7 | 592 | os->os_dnodesize = DNODE_MIN_SIZE; |
34dc7c2f BB |
593 | } |
594 | ||
0c66c32d | 595 | if (ds == NULL || !ds->ds_is_snapshot) |
572e2857 | 596 | os->os_zil_header = os->os_phys->os_zil_header; |
428870ff | 597 | os->os_zil = zil_alloc(os, &os->os_zil_header); |
34dc7c2f BB |
598 | |
599 | for (i = 0; i < TXG_SIZE; i++) { | |
64fc7762 MA |
600 | os->os_dirty_dnodes[i] = multilist_create(sizeof (dnode_t), |
601 | offsetof(dnode_t, dn_dirty_link[i]), | |
602 | dnode_multilist_index_func); | |
34dc7c2f | 603 | } |
428870ff | 604 | list_create(&os->os_dnodes, sizeof (dnode_t), |
34dc7c2f | 605 | offsetof(dnode_t, dn_link)); |
428870ff | 606 | list_create(&os->os_downgraded_dbufs, sizeof (dmu_buf_impl_t), |
34dc7c2f BB |
607 | offsetof(dmu_buf_impl_t, db_link)); |
608 | ||
0c66c32d JG |
609 | list_link_init(&os->os_evicting_node); |
610 | ||
428870ff | 611 | mutex_init(&os->os_lock, NULL, MUTEX_DEFAULT, NULL); |
64fc7762 | 612 | mutex_init(&os->os_userused_lock, NULL, MUTEX_DEFAULT, NULL); |
428870ff BB |
613 | mutex_init(&os->os_obj_lock, NULL, MUTEX_DEFAULT, NULL); |
614 | mutex_init(&os->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL); | |
dbeb8796 MA |
615 | os->os_obj_next_percpu_len = boot_ncpus; |
616 | os->os_obj_next_percpu = kmem_zalloc(os->os_obj_next_percpu_len * | |
617 | sizeof (os->os_obj_next_percpu[0]), KM_SLEEP); | |
428870ff | 618 | |
0c66c32d JG |
619 | dnode_special_open(os, &os->os_phys->os_meta_dnode, |
620 | DMU_META_DNODE_OBJECT, &os->os_meta_dnode); | |
9c5167d1 | 621 | if (OBJSET_BUF_HAS_USERUSED(os->os_phys_buf)) { |
0c66c32d JG |
622 | dnode_special_open(os, &os->os_phys->os_userused_dnode, |
623 | DMU_USERUSED_OBJECT, &os->os_userused_dnode); | |
624 | dnode_special_open(os, &os->os_phys->os_groupused_dnode, | |
625 | DMU_GROUPUSED_OBJECT, &os->os_groupused_dnode); | |
9c5167d1 NF |
626 | if (OBJSET_BUF_HAS_PROJECTUSED(os->os_phys_buf)) |
627 | dnode_special_open(os, | |
628 | &os->os_phys->os_projectused_dnode, | |
629 | DMU_PROJECTUSED_OBJECT, &os->os_projectused_dnode); | |
9babb374 | 630 | } |
34dc7c2f | 631 | |
1de321e6 JX |
632 | mutex_init(&os->os_upgrade_lock, NULL, MUTEX_DEFAULT, NULL); |
633 | ||
428870ff | 634 | *osp = os; |
34dc7c2f BB |
635 | return (0); |
636 | } | |
637 | ||
428870ff BB |
638 | int |
639 | dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp) | |
34dc7c2f | 640 | { |
428870ff | 641 | int err = 0; |
34dc7c2f | 642 | |
47dfff3b | 643 | /* |
0fdd6106 MA |
644 | * We need the pool_config lock to manipulate the dsl_dataset_t. |
645 | * Even if the dataset is long-held, we need the pool_config lock | |
646 | * to open the objset, as it needs to get properties. | |
47dfff3b | 647 | */ |
0fdd6106 | 648 | ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool)); |
47dfff3b | 649 | |
34dc7c2f | 650 | mutex_enter(&ds->ds_opening_lock); |
9b67f605 MA |
651 | if (ds->ds_objset == NULL) { |
652 | objset_t *os; | |
cc9bb3e5 | 653 | rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); |
34dc7c2f | 654 | err = dmu_objset_open_impl(dsl_dataset_get_spa(ds), |
9b67f605 | 655 | ds, dsl_dataset_get_blkptr(ds), &os); |
cc9bb3e5 | 656 | rrw_exit(&ds->ds_bp_rwlock, FTAG); |
9b67f605 MA |
657 | |
658 | if (err == 0) { | |
659 | mutex_enter(&ds->ds_lock); | |
660 | ASSERT(ds->ds_objset == NULL); | |
661 | ds->ds_objset = os; | |
662 | mutex_exit(&ds->ds_lock); | |
663 | } | |
34dc7c2f | 664 | } |
9b67f605 | 665 | *osp = ds->ds_objset; |
34dc7c2f | 666 | mutex_exit(&ds->ds_opening_lock); |
428870ff | 667 | return (err); |
34dc7c2f BB |
668 | } |
669 | ||
13fe0198 MA |
670 | /* |
671 | * Holds the pool while the objset is held. Therefore only one objset | |
672 | * can be held at a time. | |
673 | */ | |
34dc7c2f | 674 | int |
b5256303 TC |
675 | dmu_objset_hold_flags(const char *name, boolean_t decrypt, void *tag, |
676 | objset_t **osp) | |
34dc7c2f | 677 | { |
13fe0198 | 678 | dsl_pool_t *dp; |
428870ff | 679 | dsl_dataset_t *ds; |
34dc7c2f | 680 | int err; |
b5256303 | 681 | ds_hold_flags_t flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0; |
34dc7c2f | 682 | |
13fe0198 MA |
683 | err = dsl_pool_hold(name, tag, &dp); |
684 | if (err != 0) | |
685 | return (err); | |
b5256303 | 686 | err = dsl_dataset_hold_flags(dp, name, flags, tag, &ds); |
13fe0198 MA |
687 | if (err != 0) { |
688 | dsl_pool_rele(dp, tag); | |
428870ff | 689 | return (err); |
13fe0198 | 690 | } |
428870ff BB |
691 | |
692 | err = dmu_objset_from_ds(ds, osp); | |
13fe0198 | 693 | if (err != 0) { |
428870ff | 694 | dsl_dataset_rele(ds, tag); |
13fe0198 MA |
695 | dsl_pool_rele(dp, tag); |
696 | } | |
428870ff | 697 | |
34dc7c2f BB |
698 | return (err); |
699 | } | |
700 | ||
b5256303 TC |
701 | int |
702 | dmu_objset_hold(const char *name, void *tag, objset_t **osp) | |
703 | { | |
704 | return (dmu_objset_hold_flags(name, B_FALSE, tag, osp)); | |
705 | } | |
706 | ||
9c43027b AJ |
707 | static int |
708 | dmu_objset_own_impl(dsl_dataset_t *ds, dmu_objset_type_t type, | |
b5256303 | 709 | boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp) |
9c43027b AJ |
710 | { |
711 | int err; | |
712 | ||
713 | err = dmu_objset_from_ds(ds, osp); | |
714 | if (err != 0) { | |
b5256303 | 715 | return (err); |
9c43027b | 716 | } else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) { |
9c43027b AJ |
717 | return (SET_ERROR(EINVAL)); |
718 | } else if (!readonly && dsl_dataset_is_snapshot(ds)) { | |
9c43027b | 719 | return (SET_ERROR(EROFS)); |
ae76f45c TC |
720 | } else if (!readonly && decrypt && |
721 | dsl_dir_incompatible_encryption_version(ds->ds_dir)) { | |
722 | return (SET_ERROR(EROFS)); | |
9c43027b | 723 | } |
b5256303 TC |
724 | |
725 | /* if we are decrypting, we can now check MACs in os->os_phys_buf */ | |
726 | if (decrypt && arc_is_unauthenticated((*osp)->os_phys_buf)) { | |
a2c2ed1b TC |
727 | zbookmark_phys_t zb; |
728 | ||
729 | SET_BOOKMARK(&zb, ds->ds_object, ZB_ROOT_OBJECT, | |
730 | ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
b5256303 | 731 | err = arc_untransform((*osp)->os_phys_buf, (*osp)->os_spa, |
a2c2ed1b | 732 | &zb, B_FALSE); |
b5256303 TC |
733 | if (err != 0) |
734 | return (err); | |
735 | ||
736 | ASSERT0(arc_is_unauthenticated((*osp)->os_phys_buf)); | |
737 | } | |
738 | ||
739 | return (0); | |
9c43027b AJ |
740 | } |
741 | ||
13fe0198 MA |
742 | /* |
743 | * dsl_pool must not be held when this is called. | |
744 | * Upon successful return, there will be a longhold on the dataset, | |
745 | * and the dsl_pool will not be held. | |
746 | */ | |
34dc7c2f | 747 | int |
428870ff | 748 | dmu_objset_own(const char *name, dmu_objset_type_t type, |
b5256303 | 749 | boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp) |
34dc7c2f | 750 | { |
13fe0198 | 751 | dsl_pool_t *dp; |
34dc7c2f BB |
752 | dsl_dataset_t *ds; |
753 | int err; | |
b5256303 | 754 | ds_hold_flags_t flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0; |
34dc7c2f | 755 | |
13fe0198 MA |
756 | err = dsl_pool_hold(name, FTAG, &dp); |
757 | if (err != 0) | |
758 | return (err); | |
b5256303 | 759 | err = dsl_dataset_own(dp, name, flags, tag, &ds); |
13fe0198 MA |
760 | if (err != 0) { |
761 | dsl_pool_rele(dp, FTAG); | |
34dc7c2f | 762 | return (err); |
13fe0198 | 763 | } |
b5256303 TC |
764 | err = dmu_objset_own_impl(ds, type, readonly, decrypt, tag, osp); |
765 | if (err != 0) { | |
766 | dsl_dataset_disown(ds, flags, tag); | |
767 | dsl_pool_rele(dp, FTAG); | |
768 | return (err); | |
769 | } | |
770 | ||
163a8c28 TC |
771 | /* |
772 | * User accounting requires the dataset to be decrypted and rw. | |
773 | * We also don't begin user accounting during claiming to help | |
774 | * speed up pool import times and to keep this txg reserved | |
775 | * completely for recovery work. | |
776 | */ | |
9c5167d1 NF |
777 | if ((dmu_objset_userobjspace_upgradable(*osp) || |
778 | dmu_objset_projectquota_upgradable(*osp)) && | |
163a8c28 | 779 | !readonly && !dp->dp_spa->spa_claiming && |
4807c0ba | 780 | (ds->ds_dir->dd_crypto_obj == 0 || decrypt)) |
9c5167d1 | 781 | dmu_objset_id_quota_upgrade(*osp); |
1de321e6 | 782 | |
c0daec32 | 783 | dsl_pool_rele(dp, FTAG); |
b5256303 | 784 | return (0); |
34dc7c2f BB |
785 | } |
786 | ||
9c43027b AJ |
787 | int |
788 | dmu_objset_own_obj(dsl_pool_t *dp, uint64_t obj, dmu_objset_type_t type, | |
b5256303 | 789 | boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp) |
9c43027b AJ |
790 | { |
791 | dsl_dataset_t *ds; | |
792 | int err; | |
b5256303 | 793 | ds_hold_flags_t flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0; |
9c43027b | 794 | |
b5256303 | 795 | err = dsl_dataset_own_obj(dp, obj, flags, tag, &ds); |
9c43027b AJ |
796 | if (err != 0) |
797 | return (err); | |
798 | ||
b5256303 TC |
799 | err = dmu_objset_own_impl(ds, type, readonly, decrypt, tag, osp); |
800 | if (err != 0) { | |
801 | dsl_dataset_disown(ds, flags, tag); | |
802 | return (err); | |
803 | } | |
804 | ||
805 | return (0); | |
9c43027b AJ |
806 | } |
807 | ||
34dc7c2f | 808 | void |
b5256303 | 809 | dmu_objset_rele_flags(objset_t *os, boolean_t decrypt, void *tag) |
34dc7c2f | 810 | { |
b5256303 TC |
811 | ds_hold_flags_t flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0; |
812 | ||
13fe0198 | 813 | dsl_pool_t *dp = dmu_objset_pool(os); |
b5256303 | 814 | dsl_dataset_rele_flags(os->os_dsl_dataset, flags, tag); |
13fe0198 | 815 | dsl_pool_rele(dp, tag); |
428870ff | 816 | } |
b128c09f | 817 | |
b5256303 TC |
818 | void |
819 | dmu_objset_rele(objset_t *os, void *tag) | |
820 | { | |
821 | dmu_objset_rele_flags(os, B_FALSE, tag); | |
822 | } | |
823 | ||
831baf06 KW |
824 | /* |
825 | * When we are called, os MUST refer to an objset associated with a dataset | |
826 | * that is owned by 'tag'; that is, is held and long held by 'tag' and ds_owner | |
827 | * == tag. We will then release and reacquire ownership of the dataset while | |
828 | * holding the pool config_rwlock to avoid intervening namespace or ownership | |
829 | * changes may occur. | |
830 | * | |
831 | * This exists solely to accommodate zfs_ioc_userspace_upgrade()'s desire to | |
832 | * release the hold on its dataset and acquire a new one on the dataset of the | |
833 | * same name so that it can be partially torn down and reconstructed. | |
834 | */ | |
835 | void | |
5e00213e AG |
836 | dmu_objset_refresh_ownership(dsl_dataset_t *ds, dsl_dataset_t **newds, |
837 | boolean_t decrypt, void *tag) | |
831baf06 KW |
838 | { |
839 | dsl_pool_t *dp; | |
eca7b760 | 840 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
831baf06 | 841 | |
831baf06 KW |
842 | VERIFY3P(ds, !=, NULL); |
843 | VERIFY3P(ds->ds_owner, ==, tag); | |
844 | VERIFY(dsl_dataset_long_held(ds)); | |
845 | ||
846 | dsl_dataset_name(ds, name); | |
5e00213e | 847 | dp = ds->ds_dir->dd_pool; |
831baf06 | 848 | dsl_pool_config_enter(dp, FTAG); |
5e00213e | 849 | dsl_dataset_disown(ds, decrypt, tag); |
b5256303 | 850 | VERIFY0(dsl_dataset_own(dp, name, |
5e00213e | 851 | (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0, tag, newds)); |
831baf06 KW |
852 | dsl_pool_config_exit(dp, FTAG); |
853 | } | |
854 | ||
428870ff | 855 | void |
b5256303 | 856 | dmu_objset_disown(objset_t *os, boolean_t decrypt, void *tag) |
428870ff | 857 | { |
1de321e6 JX |
858 | /* |
859 | * Stop upgrading thread | |
860 | */ | |
861 | dmu_objset_upgrade_stop(os); | |
b5256303 TC |
862 | dsl_dataset_disown(os->os_dsl_dataset, |
863 | (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0, tag); | |
34dc7c2f BB |
864 | } |
865 | ||
13fe0198 | 866 | void |
34dc7c2f BB |
867 | dmu_objset_evict_dbufs(objset_t *os) |
868 | { | |
0c66c32d | 869 | dnode_t *dn_marker; |
34dc7c2f BB |
870 | dnode_t *dn; |
871 | ||
0c66c32d | 872 | dn_marker = kmem_alloc(sizeof (dnode_t), KM_SLEEP); |
34dc7c2f | 873 | |
0c66c32d JG |
874 | mutex_enter(&os->os_lock); |
875 | dn = list_head(&os->os_dnodes); | |
876 | while (dn != NULL) { | |
877 | /* | |
878 | * Skip dnodes without holds. We have to do this dance | |
879 | * because dnode_add_ref() only works if there is already a | |
880 | * hold. If the dnode has no holds, then it has no dbufs. | |
881 | */ | |
882 | if (dnode_add_ref(dn, FTAG)) { | |
883 | list_insert_after(&os->os_dnodes, dn, dn_marker); | |
884 | mutex_exit(&os->os_lock); | |
34dc7c2f | 885 | |
0c66c32d JG |
886 | dnode_evict_dbufs(dn); |
887 | dnode_rele(dn, FTAG); | |
34dc7c2f | 888 | |
0c66c32d JG |
889 | mutex_enter(&os->os_lock); |
890 | dn = list_next(&os->os_dnodes, dn_marker); | |
891 | list_remove(&os->os_dnodes, dn_marker); | |
892 | } else { | |
893 | dn = list_next(&os->os_dnodes, dn); | |
894 | } | |
895 | } | |
896 | mutex_exit(&os->os_lock); | |
34dc7c2f | 897 | |
0c66c32d | 898 | kmem_free(dn_marker, sizeof (dnode_t)); |
34dc7c2f | 899 | |
0c66c32d | 900 | if (DMU_USERUSED_DNODE(os) != NULL) { |
9c5167d1 NF |
901 | if (DMU_PROJECTUSED_DNODE(os) != NULL) |
902 | dnode_evict_dbufs(DMU_PROJECTUSED_DNODE(os)); | |
0c66c32d JG |
903 | dnode_evict_dbufs(DMU_GROUPUSED_DNODE(os)); |
904 | dnode_evict_dbufs(DMU_USERUSED_DNODE(os)); | |
34dc7c2f | 905 | } |
0c66c32d | 906 | dnode_evict_dbufs(DMU_META_DNODE(os)); |
34dc7c2f BB |
907 | } |
908 | ||
0c66c32d JG |
909 | /* |
910 | * Objset eviction processing is split into into two pieces. | |
911 | * The first marks the objset as evicting, evicts any dbufs that | |
912 | * have a refcount of zero, and then queues up the objset for the | |
913 | * second phase of eviction. Once os->os_dnodes has been cleared by | |
914 | * dnode_buf_pageout()->dnode_destroy(), the second phase is executed. | |
915 | * The second phase closes the special dnodes, dequeues the objset from | |
916 | * the list of those undergoing eviction, and finally frees the objset. | |
917 | * | |
918 | * NOTE: Due to asynchronous eviction processing (invocation of | |
919 | * dnode_buf_pageout()), it is possible for the meta dnode for the | |
920 | * objset to have no holds even though os->os_dnodes is not empty. | |
921 | */ | |
34dc7c2f | 922 | void |
428870ff | 923 | dmu_objset_evict(objset_t *os) |
34dc7c2f | 924 | { |
6f1ffb06 MA |
925 | dsl_dataset_t *ds = os->os_dsl_dataset; |
926 | ||
1c27024e | 927 | for (int t = 0; t < TXG_SIZE; t++) |
428870ff | 928 | ASSERT(!dmu_objset_is_dirty(os, t)); |
34dc7c2f | 929 | |
0eb21616 JG |
930 | if (ds) |
931 | dsl_prop_unregister_all(ds, os); | |
34dc7c2f | 932 | |
428870ff BB |
933 | if (os->os_sa) |
934 | sa_tear_down(os); | |
935 | ||
13fe0198 | 936 | dmu_objset_evict_dbufs(os); |
34dc7c2f | 937 | |
0c66c32d JG |
938 | mutex_enter(&os->os_lock); |
939 | spa_evicting_os_register(os->os_spa, os); | |
940 | if (list_is_empty(&os->os_dnodes)) { | |
941 | mutex_exit(&os->os_lock); | |
942 | dmu_objset_evict_done(os); | |
943 | } else { | |
944 | mutex_exit(&os->os_lock); | |
945 | } | |
b5256303 TC |
946 | |
947 | ||
0c66c32d JG |
948 | } |
949 | ||
950 | void | |
951 | dmu_objset_evict_done(objset_t *os) | |
952 | { | |
953 | ASSERT3P(list_head(&os->os_dnodes), ==, NULL); | |
954 | ||
572e2857 BB |
955 | dnode_special_close(&os->os_meta_dnode); |
956 | if (DMU_USERUSED_DNODE(os)) { | |
9c5167d1 NF |
957 | if (DMU_PROJECTUSED_DNODE(os)) |
958 | dnode_special_close(&os->os_projectused_dnode); | |
572e2857 BB |
959 | dnode_special_close(&os->os_userused_dnode); |
960 | dnode_special_close(&os->os_groupused_dnode); | |
9babb374 | 961 | } |
428870ff BB |
962 | zil_free(os->os_zil); |
963 | ||
d3c2ae1c | 964 | arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); |
572e2857 BB |
965 | |
966 | /* | |
967 | * This is a barrier to prevent the objset from going away in | |
968 | * dnode_move() until we can safely ensure that the objset is still in | |
969 | * use. We consider the objset valid before the barrier and invalid | |
970 | * after the barrier. | |
971 | */ | |
972 | rw_enter(&os_lock, RW_READER); | |
973 | rw_exit(&os_lock); | |
974 | ||
dbeb8796 MA |
975 | kmem_free(os->os_obj_next_percpu, |
976 | os->os_obj_next_percpu_len * sizeof (os->os_obj_next_percpu[0])); | |
977 | ||
428870ff | 978 | mutex_destroy(&os->os_lock); |
64fc7762 | 979 | mutex_destroy(&os->os_userused_lock); |
428870ff BB |
980 | mutex_destroy(&os->os_obj_lock); |
981 | mutex_destroy(&os->os_user_ptr_lock); | |
c17486b2 | 982 | mutex_destroy(&os->os_upgrade_lock); |
64fc7762 MA |
983 | for (int i = 0; i < TXG_SIZE; i++) { |
984 | multilist_destroy(os->os_dirty_dnodes[i]); | |
985 | } | |
0c66c32d | 986 | spa_evicting_os_deregister(os->os_spa, os); |
428870ff BB |
987 | kmem_free(os, sizeof (objset_t)); |
988 | } | |
9babb374 | 989 | |
6413c95f | 990 | inode_timespec_t |
428870ff BB |
991 | dmu_objset_snap_cmtime(objset_t *os) |
992 | { | |
993 | return (dsl_dir_snap_cmtime(os->os_dsl_dataset->ds_dir)); | |
34dc7c2f BB |
994 | } |
995 | ||
428870ff | 996 | objset_t * |
b5256303 TC |
997 | dmu_objset_create_impl_dnstats(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, |
998 | dmu_objset_type_t type, int levels, int blksz, int ibs, dmu_tx_t *tx) | |
34dc7c2f | 999 | { |
428870ff | 1000 | objset_t *os; |
34dc7c2f BB |
1001 | dnode_t *mdn; |
1002 | ||
1003 | ASSERT(dmu_tx_is_syncing(tx)); | |
13fe0198 | 1004 | |
b5256303 TC |
1005 | if (blksz == 0) |
1006 | blksz = DNODE_BLOCK_SIZE; | |
4807c0ba | 1007 | if (ibs == 0) |
b5256303 TC |
1008 | ibs = DN_MAX_INDBLKSHIFT; |
1009 | ||
572e2857 | 1010 | if (ds != NULL) |
13fe0198 | 1011 | VERIFY0(dmu_objset_from_ds(ds, &os)); |
572e2857 | 1012 | else |
13fe0198 | 1013 | VERIFY0(dmu_objset_open_impl(spa, NULL, bp, &os)); |
572e2857 BB |
1014 | |
1015 | mdn = DMU_META_DNODE(os); | |
34dc7c2f | 1016 | |
b5256303 TC |
1017 | dnode_allocate(mdn, DMU_OT_DNODE, blksz, ibs, DMU_OT_NONE, 0, |
1018 | DNODE_MIN_SLOTS, tx); | |
34dc7c2f BB |
1019 | |
1020 | /* | |
1021 | * We don't want to have to increase the meta-dnode's nlevels | |
e1cfd73f | 1022 | * later, because then we could do it in quiescing context while |
34dc7c2f BB |
1023 | * we are also accessing it in open context. |
1024 | * | |
1025 | * This precaution is not necessary for the MOS (ds == NULL), | |
1026 | * because the MOS is only updated in syncing context. | |
1027 | * This is most fortunate: the MOS is the only objset that | |
1028 | * needs to be synced multiple times as spa_sync() iterates | |
1029 | * to convergence, so minimizing its dn_nlevels matters. | |
1030 | */ | |
1031 | if (ds != NULL) { | |
b5256303 TC |
1032 | if (levels == 0) { |
1033 | levels = 1; | |
1034 | ||
1035 | /* | |
1036 | * Determine the number of levels necessary for the | |
1037 | * meta-dnode to contain DN_MAX_OBJECT dnodes. Note | |
1038 | * that in order to ensure that we do not overflow | |
1039 | * 64 bits, there has to be a nlevels that gives us a | |
1040 | * number of blocks > DN_MAX_OBJECT but < 2^64. | |
1041 | * Therefore, (mdn->dn_indblkshift - SPA_BLKPTRSHIFT) | |
1042 | * (10) must be less than (64 - log2(DN_MAX_OBJECT)) | |
1043 | * (16). | |
1044 | */ | |
1045 | while ((uint64_t)mdn->dn_nblkptr << | |
1046 | (mdn->dn_datablkshift - DNODE_SHIFT + (levels - 1) * | |
1047 | (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) < | |
1048 | DN_MAX_OBJECT) | |
1049 | levels++; | |
1050 | } | |
34dc7c2f BB |
1051 | |
1052 | mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] = | |
1053 | mdn->dn_nlevels = levels; | |
1054 | } | |
1055 | ||
1056 | ASSERT(type != DMU_OST_NONE); | |
1057 | ASSERT(type != DMU_OST_ANY); | |
1058 | ASSERT(type < DMU_OST_NUMTYPES); | |
428870ff | 1059 | os->os_phys->os_type = type; |
b5256303 TC |
1060 | |
1061 | /* | |
1062 | * Enable user accounting if it is enabled and this is not an | |
1063 | * encrypted receive. | |
1064 | */ | |
1065 | if (dmu_objset_userused_enabled(os) && | |
1066 | (!os->os_encrypted || !dmu_objset_is_receiving(os))) { | |
428870ff | 1067 | os->os_phys->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; |
1de321e6 | 1068 | if (dmu_objset_userobjused_enabled(os)) { |
d52d80b7 PD |
1069 | ds->ds_feature_activation[ |
1070 | SPA_FEATURE_USEROBJ_ACCOUNTING] = (void *)B_TRUE; | |
1de321e6 JX |
1071 | os->os_phys->os_flags |= |
1072 | OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE; | |
1073 | } | |
9c5167d1 | 1074 | if (dmu_objset_projectquota_enabled(os)) { |
d52d80b7 PD |
1075 | ds->ds_feature_activation[ |
1076 | SPA_FEATURE_PROJECT_QUOTA] = (void *)B_TRUE; | |
9c5167d1 NF |
1077 | os->os_phys->os_flags |= |
1078 | OBJSET_FLAG_PROJECTQUOTA_COMPLETE; | |
1079 | } | |
428870ff | 1080 | os->os_flags = os->os_phys->os_flags; |
9babb374 | 1081 | } |
34dc7c2f BB |
1082 | |
1083 | dsl_dataset_dirty(ds, tx); | |
1084 | ||
428870ff | 1085 | return (os); |
34dc7c2f BB |
1086 | } |
1087 | ||
b5256303 TC |
1088 | /* called from dsl for meta-objset */ |
1089 | objset_t * | |
1090 | dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, | |
1091 | dmu_objset_type_t type, dmu_tx_t *tx) | |
1092 | { | |
1093 | return (dmu_objset_create_impl_dnstats(spa, ds, bp, type, 0, 0, 0, tx)); | |
1094 | } | |
1095 | ||
13fe0198 MA |
1096 | typedef struct dmu_objset_create_arg { |
1097 | const char *doca_name; | |
1098 | cred_t *doca_cred; | |
1099 | void (*doca_userfunc)(objset_t *os, void *arg, | |
1100 | cred_t *cr, dmu_tx_t *tx); | |
1101 | void *doca_userarg; | |
1102 | dmu_objset_type_t doca_type; | |
1103 | uint64_t doca_flags; | |
b5256303 | 1104 | dsl_crypto_params_t *doca_dcp; |
13fe0198 | 1105 | } dmu_objset_create_arg_t; |
34dc7c2f BB |
1106 | |
1107 | /*ARGSUSED*/ | |
1108 | static int | |
13fe0198 | 1109 | dmu_objset_create_check(void *arg, dmu_tx_t *tx) |
34dc7c2f | 1110 | { |
13fe0198 MA |
1111 | dmu_objset_create_arg_t *doca = arg; |
1112 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
1113 | dsl_dir_t *pdd; | |
d8d418ff | 1114 | dsl_dataset_t *parentds; |
1115 | objset_t *parentos; | |
13fe0198 MA |
1116 | const char *tail; |
1117 | int error; | |
34dc7c2f | 1118 | |
13fe0198 | 1119 | if (strchr(doca->doca_name, '@') != NULL) |
2e528b49 | 1120 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1121 | |
eca7b760 IK |
1122 | if (strlen(doca->doca_name) >= ZFS_MAX_DATASET_NAME_LEN) |
1123 | return (SET_ERROR(ENAMETOOLONG)); | |
1124 | ||
a7ed98d8 SD |
1125 | if (dataset_nestcheck(doca->doca_name) != 0) |
1126 | return (SET_ERROR(ENAMETOOLONG)); | |
1127 | ||
13fe0198 MA |
1128 | error = dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail); |
1129 | if (error != 0) | |
1130 | return (error); | |
1131 | if (tail == NULL) { | |
1132 | dsl_dir_rele(pdd, FTAG); | |
2e528b49 | 1133 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 1134 | } |
b5256303 | 1135 | |
1fff937a | 1136 | error = dmu_objset_create_crypt_check(pdd, doca->doca_dcp, NULL); |
b5256303 TC |
1137 | if (error != 0) { |
1138 | dsl_dir_rele(pdd, FTAG); | |
1139 | return (error); | |
1140 | } | |
1141 | ||
788eb90c JJ |
1142 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
1143 | doca->doca_cred); | |
d8d418ff | 1144 | if (error != 0) { |
1145 | dsl_dir_rele(pdd, FTAG); | |
1146 | return (error); | |
1147 | } | |
b5256303 | 1148 | |
d8d418ff | 1149 | /* can't create below anything but filesystems (eg. no ZVOLs) */ |
1150 | error = dsl_dataset_hold_obj(pdd->dd_pool, | |
1151 | dsl_dir_phys(pdd)->dd_head_dataset_obj, FTAG, &parentds); | |
1152 | if (error != 0) { | |
1153 | dsl_dir_rele(pdd, FTAG); | |
1154 | return (error); | |
1155 | } | |
1156 | error = dmu_objset_from_ds(parentds, &parentos); | |
1157 | if (error != 0) { | |
1158 | dsl_dataset_rele(parentds, FTAG); | |
1159 | dsl_dir_rele(pdd, FTAG); | |
1160 | return (error); | |
1161 | } | |
1162 | if (dmu_objset_type(parentos) != DMU_OST_ZFS) { | |
1163 | dsl_dataset_rele(parentds, FTAG); | |
1164 | dsl_dir_rele(pdd, FTAG); | |
1165 | return (SET_ERROR(ZFS_ERR_WRONG_PARENT)); | |
1166 | } | |
1167 | dsl_dataset_rele(parentds, FTAG); | |
13fe0198 | 1168 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f | 1169 | |
788eb90c | 1170 | return (error); |
34dc7c2f BB |
1171 | } |
1172 | ||
1173 | static void | |
13fe0198 | 1174 | dmu_objset_create_sync(void *arg, dmu_tx_t *tx) |
34dc7c2f | 1175 | { |
13fe0198 MA |
1176 | dmu_objset_create_arg_t *doca = arg; |
1177 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
52ce99dd | 1178 | spa_t *spa = dp->dp_spa; |
13fe0198 MA |
1179 | dsl_dir_t *pdd; |
1180 | const char *tail; | |
6f1ffb06 | 1181 | dsl_dataset_t *ds; |
13fe0198 | 1182 | uint64_t obj; |
6f1ffb06 | 1183 | blkptr_t *bp; |
13fe0198 | 1184 | objset_t *os; |
b5256303 | 1185 | zio_t *rzio; |
34dc7c2f | 1186 | |
13fe0198 | 1187 | VERIFY0(dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail)); |
34dc7c2f | 1188 | |
13fe0198 | 1189 | obj = dsl_dataset_create_sync(pdd, tail, NULL, doca->doca_flags, |
b5256303 | 1190 | doca->doca_cred, doca->doca_dcp, tx); |
34dc7c2f | 1191 | |
b5256303 TC |
1192 | VERIFY0(dsl_dataset_hold_obj_flags(pdd->dd_pool, obj, |
1193 | DS_HOLD_FLAG_DECRYPT, FTAG, &ds)); | |
cc9bb3e5 | 1194 | rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); |
6f1ffb06 | 1195 | bp = dsl_dataset_get_blkptr(ds); |
52ce99dd | 1196 | os = dmu_objset_create_impl(spa, ds, bp, doca->doca_type, tx); |
cc9bb3e5 | 1197 | rrw_exit(&ds->ds_bp_rwlock, FTAG); |
34dc7c2f | 1198 | |
13fe0198 MA |
1199 | if (doca->doca_userfunc != NULL) { |
1200 | doca->doca_userfunc(os, doca->doca_userarg, | |
1201 | doca->doca_cred, tx); | |
34dc7c2f BB |
1202 | } |
1203 | ||
b5256303 | 1204 | /* |
4807c0ba | 1205 | * The doca_userfunc() may write out some data that needs to be |
b5256303 TC |
1206 | * encrypted if the dataset is encrypted (specifically the root |
1207 | * directory). This data must be written out before the encryption | |
1208 | * key mapping is removed by dsl_dataset_rele_flags(). Force the | |
1209 | * I/O to occur immediately by invoking the relevant sections of | |
1210 | * dsl_pool_sync(). | |
1211 | */ | |
1212 | if (os->os_encrypted) { | |
1213 | dsl_dataset_t *tmpds = NULL; | |
1214 | boolean_t need_sync_done = B_FALSE; | |
1215 | ||
4807c0ba TC |
1216 | mutex_enter(&ds->ds_lock); |
1217 | ds->ds_owner = FTAG; | |
1218 | mutex_exit(&ds->ds_lock); | |
1219 | ||
52ce99dd | 1220 | rzio = zio_root(spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED); |
4807c0ba TC |
1221 | tmpds = txg_list_remove_this(&dp->dp_dirty_datasets, ds, |
1222 | tx->tx_txg); | |
b5256303 | 1223 | if (tmpds != NULL) { |
b5256303 TC |
1224 | dsl_dataset_sync(ds, rzio, tx); |
1225 | need_sync_done = B_TRUE; | |
1226 | } | |
1227 | VERIFY0(zio_wait(rzio)); | |
1228 | ||
1229 | dmu_objset_do_userquota_updates(os, tx); | |
1230 | taskq_wait(dp->dp_sync_taskq); | |
52ce99dd TC |
1231 | if (txg_list_member(&dp->dp_dirty_datasets, ds, tx->tx_txg)) { |
1232 | ASSERT3P(ds->ds_key_mapping, !=, NULL); | |
1233 | key_mapping_rele(spa, ds->ds_key_mapping, ds); | |
1234 | } | |
b5256303 | 1235 | |
52ce99dd | 1236 | rzio = zio_root(spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED); |
4807c0ba TC |
1237 | tmpds = txg_list_remove_this(&dp->dp_dirty_datasets, ds, |
1238 | tx->tx_txg); | |
b5256303 | 1239 | if (tmpds != NULL) { |
b5256303 TC |
1240 | dmu_buf_rele(ds->ds_dbuf, ds); |
1241 | dsl_dataset_sync(ds, rzio, tx); | |
1242 | } | |
1243 | VERIFY0(zio_wait(rzio)); | |
1244 | ||
52ce99dd TC |
1245 | if (need_sync_done) { |
1246 | ASSERT3P(ds->ds_key_mapping, !=, NULL); | |
1247 | key_mapping_rele(spa, ds->ds_key_mapping, ds); | |
b5256303 | 1248 | dsl_dataset_sync_done(ds, tx); |
52ce99dd | 1249 | } |
4807c0ba TC |
1250 | |
1251 | mutex_enter(&ds->ds_lock); | |
1252 | ds->ds_owner = NULL; | |
1253 | mutex_exit(&ds->ds_lock); | |
b5256303 TC |
1254 | } |
1255 | ||
74756182 | 1256 | spa_history_log_internal_ds(ds, "create", tx, " "); |
a0bd735a | 1257 | |
b5256303 | 1258 | dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT, FTAG); |
13fe0198 | 1259 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f BB |
1260 | } |
1261 | ||
1262 | int | |
428870ff | 1263 | dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, |
b5256303 | 1264 | dsl_crypto_params_t *dcp, dmu_objset_create_sync_func_t func, void *arg) |
34dc7c2f | 1265 | { |
13fe0198 | 1266 | dmu_objset_create_arg_t doca; |
b5256303 | 1267 | dsl_crypto_params_t tmp_dcp = { 0 }; |
34dc7c2f | 1268 | |
13fe0198 MA |
1269 | doca.doca_name = name; |
1270 | doca.doca_cred = CRED(); | |
1271 | doca.doca_flags = flags; | |
1272 | doca.doca_userfunc = func; | |
1273 | doca.doca_userarg = arg; | |
1274 | doca.doca_type = type; | |
34dc7c2f | 1275 | |
b5256303 TC |
1276 | /* |
1277 | * Some callers (mostly for testing) do not provide a dcp on their | |
1278 | * own but various code inside the sync task will require it to be | |
1279 | * allocated. Rather than adding NULL checks throughout this code | |
1280 | * or adding dummy dcp's to all of the callers we simply create a | |
1281 | * dummy one here and use that. This zero dcp will have the same | |
85ce3f4f | 1282 | * effect as asking for inheritance of all encryption params. |
b5256303 TC |
1283 | */ |
1284 | doca.doca_dcp = (dcp != NULL) ? dcp : &tmp_dcp; | |
1285 | ||
ec213971 | 1286 | int rv = dsl_sync_task(name, |
3d45fdd6 | 1287 | dmu_objset_create_check, dmu_objset_create_sync, &doca, |
ec213971 MA |
1288 | 6, ZFS_SPACE_CHECK_NORMAL); |
1289 | ||
1290 | if (rv == 0) | |
1291 | zvol_create_minor(name); | |
1292 | return (rv); | |
34dc7c2f BB |
1293 | } |
1294 | ||
13fe0198 MA |
1295 | typedef struct dmu_objset_clone_arg { |
1296 | const char *doca_clone; | |
1297 | const char *doca_origin; | |
1298 | cred_t *doca_cred; | |
1299 | } dmu_objset_clone_arg_t; | |
1300 | ||
1301 | /*ARGSUSED*/ | |
1302 | static int | |
1303 | dmu_objset_clone_check(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 1304 | { |
13fe0198 | 1305 | dmu_objset_clone_arg_t *doca = arg; |
428870ff BB |
1306 | dsl_dir_t *pdd; |
1307 | const char *tail; | |
13fe0198 MA |
1308 | int error; |
1309 | dsl_dataset_t *origin; | |
1310 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
34dc7c2f | 1311 | |
13fe0198 | 1312 | if (strchr(doca->doca_clone, '@') != NULL) |
2e528b49 | 1313 | return (SET_ERROR(EINVAL)); |
13fe0198 | 1314 | |
eca7b760 IK |
1315 | if (strlen(doca->doca_clone) >= ZFS_MAX_DATASET_NAME_LEN) |
1316 | return (SET_ERROR(ENAMETOOLONG)); | |
1317 | ||
13fe0198 MA |
1318 | error = dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail); |
1319 | if (error != 0) | |
1320 | return (error); | |
428870ff | 1321 | if (tail == NULL) { |
13fe0198 | 1322 | dsl_dir_rele(pdd, FTAG); |
2e528b49 | 1323 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 1324 | } |
1cddb8c9 | 1325 | |
788eb90c JJ |
1326 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
1327 | doca->doca_cred); | |
1328 | if (error != 0) { | |
1329 | dsl_dir_rele(pdd, FTAG); | |
1330 | return (SET_ERROR(EDQUOT)); | |
1331 | } | |
428870ff | 1332 | |
13fe0198 | 1333 | error = dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin); |
b5256303 TC |
1334 | if (error != 0) { |
1335 | dsl_dir_rele(pdd, FTAG); | |
572e2857 | 1336 | return (error); |
b5256303 | 1337 | } |
572e2857 | 1338 | |
13fe0198 | 1339 | /* You can only clone snapshots, not the head datasets. */ |
0c66c32d | 1340 | if (!origin->ds_is_snapshot) { |
13fe0198 | 1341 | dsl_dataset_rele(origin, FTAG); |
b5256303 | 1342 | dsl_dir_rele(pdd, FTAG); |
2e528b49 | 1343 | return (SET_ERROR(EINVAL)); |
428870ff | 1344 | } |
b5256303 | 1345 | |
13fe0198 | 1346 | dsl_dataset_rele(origin, FTAG); |
b5256303 | 1347 | dsl_dir_rele(pdd, FTAG); |
572e2857 | 1348 | |
13fe0198 | 1349 | return (0); |
9babb374 | 1350 | } |
34dc7c2f | 1351 | |
13fe0198 MA |
1352 | static void |
1353 | dmu_objset_clone_sync(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 1354 | { |
13fe0198 MA |
1355 | dmu_objset_clone_arg_t *doca = arg; |
1356 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
1357 | dsl_dir_t *pdd; | |
1358 | const char *tail; | |
1359 | dsl_dataset_t *origin, *ds; | |
1360 | uint64_t obj; | |
eca7b760 | 1361 | char namebuf[ZFS_MAX_DATASET_NAME_LEN]; |
6f1ffb06 | 1362 | |
13fe0198 MA |
1363 | VERIFY0(dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail)); |
1364 | VERIFY0(dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin)); | |
6f1ffb06 | 1365 | |
13fe0198 | 1366 | obj = dsl_dataset_create_sync(pdd, tail, origin, 0, |
b5256303 | 1367 | doca->doca_cred, NULL, tx); |
34dc7c2f | 1368 | |
13fe0198 MA |
1369 | VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); |
1370 | dsl_dataset_name(origin, namebuf); | |
1371 | spa_history_log_internal_ds(ds, "clone", tx, | |
74756182 | 1372 | "origin=%s (%llu)", namebuf, (u_longlong_t)origin->ds_object); |
13fe0198 MA |
1373 | dsl_dataset_rele(ds, FTAG); |
1374 | dsl_dataset_rele(origin, FTAG); | |
1375 | dsl_dir_rele(pdd, FTAG); | |
34dc7c2f BB |
1376 | } |
1377 | ||
1378 | int | |
13fe0198 | 1379 | dmu_objset_clone(const char *clone, const char *origin) |
34dc7c2f | 1380 | { |
13fe0198 | 1381 | dmu_objset_clone_arg_t doca; |
34dc7c2f | 1382 | |
13fe0198 MA |
1383 | doca.doca_clone = clone; |
1384 | doca.doca_origin = origin; | |
1385 | doca.doca_cred = CRED(); | |
572e2857 | 1386 | |
ec213971 | 1387 | int rv = dsl_sync_task(clone, |
3d45fdd6 | 1388 | dmu_objset_clone_check, dmu_objset_clone_sync, &doca, |
ec213971 MA |
1389 | 6, ZFS_SPACE_CHECK_NORMAL); |
1390 | ||
1391 | if (rv == 0) | |
1392 | zvol_create_minor(clone); | |
1393 | ||
1394 | return (rv); | |
6f1ffb06 MA |
1395 | } |
1396 | ||
1397 | int | |
1398 | dmu_objset_snapshot_one(const char *fsname, const char *snapname) | |
1399 | { | |
1400 | int err; | |
1401 | char *longsnap = kmem_asprintf("%s@%s", fsname, snapname); | |
1402 | nvlist_t *snaps = fnvlist_alloc(); | |
1403 | ||
1404 | fnvlist_add_boolean(snaps, longsnap); | |
e4f5fa12 | 1405 | kmem_strfree(longsnap); |
13fe0198 MA |
1406 | err = dsl_dataset_snapshot(snaps, NULL, NULL); |
1407 | fnvlist_free(snaps); | |
6f1ffb06 MA |
1408 | return (err); |
1409 | } | |
1410 | ||
1de321e6 JX |
1411 | static void |
1412 | dmu_objset_upgrade_task_cb(void *data) | |
1413 | { | |
1414 | objset_t *os = data; | |
1415 | ||
1416 | mutex_enter(&os->os_upgrade_lock); | |
1417 | os->os_upgrade_status = EINTR; | |
1418 | if (!os->os_upgrade_exit) { | |
1419 | mutex_exit(&os->os_upgrade_lock); | |
1420 | ||
1421 | os->os_upgrade_status = os->os_upgrade_cb(os); | |
1422 | mutex_enter(&os->os_upgrade_lock); | |
1423 | } | |
1424 | os->os_upgrade_exit = B_TRUE; | |
1425 | os->os_upgrade_id = 0; | |
1426 | mutex_exit(&os->os_upgrade_lock); | |
c0daec32 | 1427 | dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag); |
1de321e6 JX |
1428 | } |
1429 | ||
1430 | static void | |
1431 | dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb) | |
1432 | { | |
1433 | if (os->os_upgrade_id != 0) | |
1434 | return; | |
1435 | ||
c0daec32 AB |
1436 | ASSERT(dsl_pool_config_held(dmu_objset_pool(os))); |
1437 | dsl_dataset_long_hold(dmu_objset_ds(os), upgrade_tag); | |
1438 | ||
1de321e6 JX |
1439 | mutex_enter(&os->os_upgrade_lock); |
1440 | if (os->os_upgrade_id == 0 && os->os_upgrade_status == 0) { | |
1441 | os->os_upgrade_exit = B_FALSE; | |
1442 | os->os_upgrade_cb = cb; | |
1443 | os->os_upgrade_id = taskq_dispatch( | |
1444 | os->os_spa->spa_upgrade_taskq, | |
1445 | dmu_objset_upgrade_task_cb, os, TQ_SLEEP); | |
c0daec32 AB |
1446 | if (os->os_upgrade_id == TASKQID_INVALID) { |
1447 | dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag); | |
1de321e6 | 1448 | os->os_upgrade_status = ENOMEM; |
c0daec32 | 1449 | } |
1de321e6 JX |
1450 | } |
1451 | mutex_exit(&os->os_upgrade_lock); | |
1452 | } | |
1453 | ||
1454 | static void | |
1455 | dmu_objset_upgrade_stop(objset_t *os) | |
1456 | { | |
1457 | mutex_enter(&os->os_upgrade_lock); | |
1458 | os->os_upgrade_exit = B_TRUE; | |
1459 | if (os->os_upgrade_id != 0) { | |
1460 | taskqid_t id = os->os_upgrade_id; | |
1461 | ||
1462 | os->os_upgrade_id = 0; | |
1463 | mutex_exit(&os->os_upgrade_lock); | |
1464 | ||
c0daec32 AB |
1465 | if ((taskq_cancel_id(os->os_spa->spa_upgrade_taskq, id)) == 0) { |
1466 | dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag); | |
1467 | } | |
4807c0ba | 1468 | txg_wait_synced(os->os_spa->spa_dsl_pool, 0); |
1de321e6 JX |
1469 | } else { |
1470 | mutex_exit(&os->os_upgrade_lock); | |
1471 | } | |
1472 | } | |
1473 | ||
34dc7c2f | 1474 | static void |
64fc7762 | 1475 | dmu_objset_sync_dnodes(multilist_sublist_t *list, dmu_tx_t *tx) |
34dc7c2f BB |
1476 | { |
1477 | dnode_t *dn; | |
1478 | ||
64fc7762 | 1479 | while ((dn = multilist_sublist_head(list)) != NULL) { |
34dc7c2f BB |
1480 | ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT); |
1481 | ASSERT(dn->dn_dbuf->db_data_pending); | |
1482 | /* | |
9babb374 | 1483 | * Initialize dn_zio outside dnode_sync() because the |
93e28d66 | 1484 | * meta-dnode needs to set it outside dnode_sync(). |
34dc7c2f BB |
1485 | */ |
1486 | dn->dn_zio = dn->dn_dbuf->db_data_pending->dr_zio; | |
1487 | ASSERT(dn->dn_zio); | |
1488 | ||
1489 | ASSERT3U(dn->dn_nlevels, <=, DN_MAX_LEVELS); | |
64fc7762 | 1490 | multilist_sublist_remove(list, dn); |
9babb374 | 1491 | |
edc1e713 TC |
1492 | /* |
1493 | * If we are not doing useraccounting (os_synced_dnodes == NULL) | |
1494 | * we are done with this dnode for this txg. Unset dn_dirty_txg | |
1495 | * if later txgs aren't dirtying it so that future holders do | |
1496 | * not get a stale value. Otherwise, we will do this in | |
1497 | * userquota_updates_task() when processing has completely | |
1498 | * finished for this txg. | |
1499 | */ | |
64fc7762 MA |
1500 | multilist_t *newlist = dn->dn_objset->os_synced_dnodes; |
1501 | if (newlist != NULL) { | |
9babb374 | 1502 | (void) dnode_add_ref(dn, newlist); |
64fc7762 | 1503 | multilist_insert(newlist, dn); |
edc1e713 TC |
1504 | } else { |
1505 | mutex_enter(&dn->dn_mtx); | |
1506 | if (dn->dn_dirty_txg == tx->tx_txg) | |
1507 | dn->dn_dirty_txg = 0; | |
1508 | mutex_exit(&dn->dn_mtx); | |
9babb374 BB |
1509 | } |
1510 | ||
34dc7c2f BB |
1511 | dnode_sync(dn, tx); |
1512 | } | |
1513 | } | |
1514 | ||
1515 | /* ARGSUSED */ | |
1516 | static void | |
428870ff | 1517 | dmu_objset_write_ready(zio_t *zio, arc_buf_t *abuf, void *arg) |
34dc7c2f | 1518 | { |
b128c09f | 1519 | blkptr_t *bp = zio->io_bp; |
428870ff | 1520 | objset_t *os = arg; |
34dc7c2f | 1521 | dnode_phys_t *dnp = &os->os_phys->os_meta_dnode; |
b5256303 | 1522 | uint64_t fill = 0; |
34dc7c2f | 1523 | |
9b67f605 | 1524 | ASSERT(!BP_IS_EMBEDDED(bp)); |
13fe0198 MA |
1525 | ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_OBJSET); |
1526 | ASSERT0(BP_GET_LEVEL(bp)); | |
34dc7c2f BB |
1527 | |
1528 | /* | |
9babb374 BB |
1529 | * Update rootbp fill count: it should be the number of objects |
1530 | * allocated in the object set (not counting the "special" | |
1531 | * objects that are stored in the objset_phys_t -- the meta | |
9c5167d1 | 1532 | * dnode and user/group/project accounting objects). |
34dc7c2f | 1533 | */ |
1c27024e | 1534 | for (int i = 0; i < dnp->dn_nblkptr; i++) |
b5256303 TC |
1535 | fill += BP_GET_FILL(&dnp->dn_blkptr[i]); |
1536 | ||
1537 | BP_SET_FILL(bp, fill); | |
1538 | ||
cc9bb3e5 GM |
1539 | if (os->os_dsl_dataset != NULL) |
1540 | rrw_enter(&os->os_dsl_dataset->ds_bp_rwlock, RW_WRITER, FTAG); | |
1541 | *os->os_rootbp = *bp; | |
1542 | if (os->os_dsl_dataset != NULL) | |
1543 | rrw_exit(&os->os_dsl_dataset->ds_bp_rwlock, FTAG); | |
428870ff BB |
1544 | } |
1545 | ||
1546 | /* ARGSUSED */ | |
1547 | static void | |
1548 | dmu_objset_write_done(zio_t *zio, arc_buf_t *abuf, void *arg) | |
1549 | { | |
1550 | blkptr_t *bp = zio->io_bp; | |
1551 | blkptr_t *bp_orig = &zio->io_bp_orig; | |
1552 | objset_t *os = arg; | |
34dc7c2f | 1553 | |
b128c09f | 1554 | if (zio->io_flags & ZIO_FLAG_IO_REWRITE) { |
428870ff | 1555 | ASSERT(BP_EQUAL(bp, bp_orig)); |
b128c09f | 1556 | } else { |
428870ff BB |
1557 | dsl_dataset_t *ds = os->os_dsl_dataset; |
1558 | dmu_tx_t *tx = os->os_synctx; | |
1559 | ||
1560 | (void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE); | |
1561 | dsl_dataset_block_born(ds, bp, tx); | |
34dc7c2f | 1562 | } |
cc9bb3e5 | 1563 | kmem_free(bp, sizeof (*bp)); |
34dc7c2f BB |
1564 | } |
1565 | ||
64fc7762 MA |
1566 | typedef struct sync_dnodes_arg { |
1567 | multilist_t *sda_list; | |
1568 | int sda_sublist_idx; | |
1569 | multilist_t *sda_newlist; | |
1570 | dmu_tx_t *sda_tx; | |
1571 | } sync_dnodes_arg_t; | |
1572 | ||
1573 | static void | |
1574 | sync_dnodes_task(void *arg) | |
1575 | { | |
1576 | sync_dnodes_arg_t *sda = arg; | |
1577 | ||
1578 | multilist_sublist_t *ms = | |
1579 | multilist_sublist_lock(sda->sda_list, sda->sda_sublist_idx); | |
1580 | ||
1581 | dmu_objset_sync_dnodes(ms, sda->sda_tx); | |
1582 | ||
1583 | multilist_sublist_unlock(ms); | |
1584 | ||
1585 | kmem_free(sda, sizeof (*sda)); | |
1586 | } | |
1587 | ||
1588 | ||
34dc7c2f BB |
1589 | /* called from dsl */ |
1590 | void | |
428870ff | 1591 | dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx) |
34dc7c2f BB |
1592 | { |
1593 | int txgoff; | |
5dbd68a3 | 1594 | zbookmark_phys_t zb; |
428870ff | 1595 | zio_prop_t zp; |
34dc7c2f BB |
1596 | zio_t *zio; |
1597 | list_t *list; | |
1598 | dbuf_dirty_record_t *dr; | |
fc754677 AM |
1599 | int num_sublists; |
1600 | multilist_t *ml; | |
cc9bb3e5 GM |
1601 | blkptr_t *blkptr_copy = kmem_alloc(sizeof (*os->os_rootbp), KM_SLEEP); |
1602 | *blkptr_copy = *os->os_rootbp; | |
34dc7c2f BB |
1603 | |
1604 | dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg); | |
1605 | ||
1606 | ASSERT(dmu_tx_is_syncing(tx)); | |
1607 | /* XXX the write_done callback should really give us the tx... */ | |
1608 | os->os_synctx = tx; | |
1609 | ||
1610 | if (os->os_dsl_dataset == NULL) { | |
1611 | /* | |
1612 | * This is the MOS. If we have upgraded, | |
1613 | * spa_max_replication() could change, so reset | |
1614 | * os_copies here. | |
1615 | */ | |
1616 | os->os_copies = spa_max_replication(os->os_spa); | |
1617 | } | |
1618 | ||
1619 | /* | |
1620 | * Create the root block IO | |
1621 | */ | |
428870ff BB |
1622 | SET_BOOKMARK(&zb, os->os_dsl_dataset ? |
1623 | os->os_dsl_dataset->ds_object : DMU_META_OBJSET, | |
1624 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
294f6806 | 1625 | arc_release(os->os_phys_buf, &os->os_phys_buf); |
b128c09f | 1626 | |
82644107 | 1627 | dmu_write_policy(os, NULL, 0, 0, &zp); |
9babb374 | 1628 | |
b5256303 | 1629 | /* |
0c03d21a MA |
1630 | * If we are either claiming the ZIL or doing a raw receive, write |
1631 | * out the os_phys_buf raw. Neither of these actions will effect the | |
1632 | * MAC at this point. | |
b5256303 | 1633 | */ |
0c03d21a MA |
1634 | if (os->os_raw_receive || |
1635 | os->os_next_write_raw[tx->tx_txg & TXG_MASK]) { | |
b5256303 | 1636 | ASSERT(os->os_encrypted); |
b5256303 TC |
1637 | arc_convert_to_raw(os->os_phys_buf, |
1638 | os->os_dsl_dataset->ds_object, ZFS_HOST_BYTEORDER, | |
1639 | DMU_OT_OBJSET, NULL, NULL, NULL); | |
1640 | } | |
1641 | ||
428870ff | 1642 | zio = arc_write(pio, os->os_spa, tx->tx_txg, |
cc9bb3e5 | 1643 | blkptr_copy, os->os_phys_buf, DMU_OS_IS_L2CACHEABLE(os), |
bc77ba73 PD |
1644 | &zp, dmu_objset_write_ready, NULL, NULL, dmu_objset_write_done, |
1645 | os, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb); | |
34dc7c2f BB |
1646 | |
1647 | /* | |
9babb374 | 1648 | * Sync special dnodes - the parent IO for the sync is the root block |
34dc7c2f | 1649 | */ |
572e2857 BB |
1650 | DMU_META_DNODE(os)->dn_zio = zio; |
1651 | dnode_sync(DMU_META_DNODE(os), tx); | |
34dc7c2f | 1652 | |
9babb374 BB |
1653 | os->os_phys->os_flags = os->os_flags; |
1654 | ||
572e2857 BB |
1655 | if (DMU_USERUSED_DNODE(os) && |
1656 | DMU_USERUSED_DNODE(os)->dn_type != DMU_OT_NONE) { | |
1657 | DMU_USERUSED_DNODE(os)->dn_zio = zio; | |
1658 | dnode_sync(DMU_USERUSED_DNODE(os), tx); | |
1659 | DMU_GROUPUSED_DNODE(os)->dn_zio = zio; | |
1660 | dnode_sync(DMU_GROUPUSED_DNODE(os), tx); | |
9babb374 BB |
1661 | } |
1662 | ||
9c5167d1 NF |
1663 | if (DMU_PROJECTUSED_DNODE(os) && |
1664 | DMU_PROJECTUSED_DNODE(os)->dn_type != DMU_OT_NONE) { | |
1665 | DMU_PROJECTUSED_DNODE(os)->dn_zio = zio; | |
1666 | dnode_sync(DMU_PROJECTUSED_DNODE(os), tx); | |
1667 | } | |
1668 | ||
34dc7c2f BB |
1669 | txgoff = tx->tx_txg & TXG_MASK; |
1670 | ||
b5256303 TC |
1671 | if (dmu_objset_userused_enabled(os) && |
1672 | (!os->os_encrypted || !dmu_objset_is_receiving(os))) { | |
9babb374 BB |
1673 | /* |
1674 | * We must create the list here because it uses the | |
64fc7762 MA |
1675 | * dn_dirty_link[] of this txg. But it may already |
1676 | * exist because we call dsl_dataset_sync() twice per txg. | |
9babb374 | 1677 | */ |
64fc7762 MA |
1678 | if (os->os_synced_dnodes == NULL) { |
1679 | os->os_synced_dnodes = | |
1680 | multilist_create(sizeof (dnode_t), | |
1681 | offsetof(dnode_t, dn_dirty_link[txgoff]), | |
1682 | dnode_multilist_index_func); | |
1683 | } else { | |
1684 | ASSERT3U(os->os_synced_dnodes->ml_offset, ==, | |
1685 | offsetof(dnode_t, dn_dirty_link[txgoff])); | |
1686 | } | |
9babb374 BB |
1687 | } |
1688 | ||
fc754677 AM |
1689 | ml = os->os_dirty_dnodes[txgoff]; |
1690 | num_sublists = multilist_get_num_sublists(ml); | |
1691 | for (int i = 0; i < num_sublists; i++) { | |
1692 | if (multilist_sublist_is_empty_idx(ml, i)) | |
1693 | continue; | |
64fc7762 | 1694 | sync_dnodes_arg_t *sda = kmem_alloc(sizeof (*sda), KM_SLEEP); |
fc754677 | 1695 | sda->sda_list = ml; |
64fc7762 MA |
1696 | sda->sda_sublist_idx = i; |
1697 | sda->sda_tx = tx; | |
1698 | (void) taskq_dispatch(dmu_objset_pool(os)->dp_sync_taskq, | |
1699 | sync_dnodes_task, sda, 0); | |
1700 | /* callback frees sda */ | |
1701 | } | |
1702 | taskq_wait(dmu_objset_pool(os)->dp_sync_taskq); | |
34dc7c2f | 1703 | |
572e2857 | 1704 | list = &DMU_META_DNODE(os)->dn_dirty_records[txgoff]; |
64fc7762 | 1705 | while ((dr = list_head(list)) != NULL) { |
13fe0198 | 1706 | ASSERT0(dr->dr_dbuf->db_level); |
34dc7c2f | 1707 | list_remove(list, dr); |
9cdf7b1f | 1708 | zio_nowait(dr->dr_zio); |
34dc7c2f | 1709 | } |
68cbd56e NB |
1710 | |
1711 | /* Enable dnode backfill if enough objects have been freed. */ | |
1712 | if (os->os_freed_dnodes >= dmu_rescan_dnode_threshold) { | |
1713 | os->os_rescan_dnodes = B_TRUE; | |
1714 | os->os_freed_dnodes = 0; | |
1715 | } | |
1716 | ||
34dc7c2f BB |
1717 | /* |
1718 | * Free intent log blocks up to this tx. | |
1719 | */ | |
1720 | zil_sync(os->os_zil, tx); | |
b128c09f | 1721 | os->os_phys->os_zil_header = os->os_zil_header; |
34dc7c2f BB |
1722 | zio_nowait(zio); |
1723 | } | |
1724 | ||
428870ff BB |
1725 | boolean_t |
1726 | dmu_objset_is_dirty(objset_t *os, uint64_t txg) | |
1727 | { | |
64fc7762 | 1728 | return (!multilist_is_empty(os->os_dirty_dnodes[txg & TXG_MASK])); |
428870ff BB |
1729 | } |
1730 | ||
572e2857 | 1731 | static objset_used_cb_t *used_cbs[DMU_OST_NUMTYPES]; |
9babb374 BB |
1732 | |
1733 | void | |
1734 | dmu_objset_register_type(dmu_objset_type_t ost, objset_used_cb_t *cb) | |
1735 | { | |
1736 | used_cbs[ost] = cb; | |
1737 | } | |
1738 | ||
1739 | boolean_t | |
428870ff | 1740 | dmu_objset_userused_enabled(objset_t *os) |
9babb374 BB |
1741 | { |
1742 | return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE && | |
572e2857 BB |
1743 | used_cbs[os->os_phys->os_type] != NULL && |
1744 | DMU_USERUSED_DNODE(os) != NULL); | |
9babb374 BB |
1745 | } |
1746 | ||
1de321e6 JX |
1747 | boolean_t |
1748 | dmu_objset_userobjused_enabled(objset_t *os) | |
1749 | { | |
1750 | return (dmu_objset_userused_enabled(os) && | |
1751 | spa_feature_is_enabled(os->os_spa, SPA_FEATURE_USEROBJ_ACCOUNTING)); | |
1752 | } | |
1753 | ||
9c5167d1 NF |
1754 | boolean_t |
1755 | dmu_objset_projectquota_enabled(objset_t *os) | |
1756 | { | |
1757 | return (used_cbs[os->os_phys->os_type] != NULL && | |
1758 | DMU_PROJECTUSED_DNODE(os) != NULL && | |
1759 | spa_feature_is_enabled(os->os_spa, SPA_FEATURE_PROJECT_QUOTA)); | |
1760 | } | |
1761 | ||
9b7a83cb JX |
1762 | typedef struct userquota_node { |
1763 | /* must be in the first filed, see userquota_update_cache() */ | |
1764 | char uqn_id[20 + DMU_OBJACCT_PREFIX_LEN]; | |
1765 | int64_t uqn_delta; | |
1766 | avl_node_t uqn_node; | |
1767 | } userquota_node_t; | |
1768 | ||
1769 | typedef struct userquota_cache { | |
1770 | avl_tree_t uqc_user_deltas; | |
1771 | avl_tree_t uqc_group_deltas; | |
9c5167d1 | 1772 | avl_tree_t uqc_project_deltas; |
9b7a83cb JX |
1773 | } userquota_cache_t; |
1774 | ||
1775 | static int | |
1776 | userquota_compare(const void *l, const void *r) | |
1777 | { | |
1778 | const userquota_node_t *luqn = l; | |
1779 | const userquota_node_t *ruqn = r; | |
e4ffa98d | 1780 | int rv; |
9b7a83cb JX |
1781 | |
1782 | /* | |
1783 | * NB: can only access uqn_id because userquota_update_cache() doesn't | |
1784 | * pass in an entire userquota_node_t. | |
1785 | */ | |
e4ffa98d BB |
1786 | rv = strcmp(luqn->uqn_id, ruqn->uqn_id); |
1787 | ||
ca577779 | 1788 | return (TREE_ISIGN(rv)); |
9b7a83cb JX |
1789 | } |
1790 | ||
1791 | static void | |
1792 | do_userquota_cacheflush(objset_t *os, userquota_cache_t *cache, dmu_tx_t *tx) | |
1793 | { | |
1794 | void *cookie; | |
1795 | userquota_node_t *uqn; | |
1796 | ||
1797 | ASSERT(dmu_tx_is_syncing(tx)); | |
1798 | ||
1799 | cookie = NULL; | |
1800 | while ((uqn = avl_destroy_nodes(&cache->uqc_user_deltas, | |
1801 | &cookie)) != NULL) { | |
64fc7762 MA |
1802 | /* |
1803 | * os_userused_lock protects against concurrent calls to | |
1804 | * zap_increment_int(). It's needed because zap_increment_int() | |
1805 | * is not thread-safe (i.e. not atomic). | |
1806 | */ | |
1807 | mutex_enter(&os->os_userused_lock); | |
9b7a83cb JX |
1808 | VERIFY0(zap_increment(os, DMU_USERUSED_OBJECT, |
1809 | uqn->uqn_id, uqn->uqn_delta, tx)); | |
64fc7762 | 1810 | mutex_exit(&os->os_userused_lock); |
9b7a83cb JX |
1811 | kmem_free(uqn, sizeof (*uqn)); |
1812 | } | |
1813 | avl_destroy(&cache->uqc_user_deltas); | |
1814 | ||
1815 | cookie = NULL; | |
1816 | while ((uqn = avl_destroy_nodes(&cache->uqc_group_deltas, | |
1817 | &cookie)) != NULL) { | |
64fc7762 | 1818 | mutex_enter(&os->os_userused_lock); |
9b7a83cb JX |
1819 | VERIFY0(zap_increment(os, DMU_GROUPUSED_OBJECT, |
1820 | uqn->uqn_id, uqn->uqn_delta, tx)); | |
64fc7762 | 1821 | mutex_exit(&os->os_userused_lock); |
9b7a83cb JX |
1822 | kmem_free(uqn, sizeof (*uqn)); |
1823 | } | |
1824 | avl_destroy(&cache->uqc_group_deltas); | |
9c5167d1 NF |
1825 | |
1826 | if (dmu_objset_projectquota_enabled(os)) { | |
1827 | cookie = NULL; | |
1828 | while ((uqn = avl_destroy_nodes(&cache->uqc_project_deltas, | |
1829 | &cookie)) != NULL) { | |
1830 | mutex_enter(&os->os_userused_lock); | |
1831 | VERIFY0(zap_increment(os, DMU_PROJECTUSED_OBJECT, | |
1832 | uqn->uqn_id, uqn->uqn_delta, tx)); | |
1833 | mutex_exit(&os->os_userused_lock); | |
1834 | kmem_free(uqn, sizeof (*uqn)); | |
1835 | } | |
1836 | avl_destroy(&cache->uqc_project_deltas); | |
1837 | } | |
9b7a83cb JX |
1838 | } |
1839 | ||
1840 | static void | |
1841 | userquota_update_cache(avl_tree_t *avl, const char *id, int64_t delta) | |
1842 | { | |
1843 | userquota_node_t *uqn; | |
1844 | avl_index_t idx; | |
1845 | ||
1846 | ASSERT(strlen(id) < sizeof (uqn->uqn_id)); | |
1847 | /* | |
1848 | * Use id directly for searching because uqn_id is the first field of | |
1849 | * userquota_node_t and fields after uqn_id won't be accessed in | |
1850 | * avl_find(). | |
1851 | */ | |
1852 | uqn = avl_find(avl, (const void *)id, &idx); | |
1853 | if (uqn == NULL) { | |
1854 | uqn = kmem_zalloc(sizeof (*uqn), KM_SLEEP); | |
1f51b525 | 1855 | strlcpy(uqn->uqn_id, id, sizeof (uqn->uqn_id)); |
9b7a83cb JX |
1856 | avl_insert(avl, uqn, idx); |
1857 | } | |
1858 | uqn->uqn_delta += delta; | |
1859 | } | |
1860 | ||
428870ff | 1861 | static void |
9c5167d1 NF |
1862 | do_userquota_update(objset_t *os, userquota_cache_t *cache, uint64_t used, |
1863 | uint64_t flags, uint64_t user, uint64_t group, uint64_t project, | |
1864 | boolean_t subtract) | |
428870ff | 1865 | { |
9c5167d1 | 1866 | if (flags & DNODE_FLAG_USERUSED_ACCOUNTED) { |
50c957f7 | 1867 | int64_t delta = DNODE_MIN_SIZE + used; |
9b7a83cb JX |
1868 | char name[20]; |
1869 | ||
428870ff BB |
1870 | if (subtract) |
1871 | delta = -delta; | |
9b7a83cb JX |
1872 | |
1873 | (void) sprintf(name, "%llx", (longlong_t)user); | |
1874 | userquota_update_cache(&cache->uqc_user_deltas, name, delta); | |
1875 | ||
1876 | (void) sprintf(name, "%llx", (longlong_t)group); | |
1877 | userquota_update_cache(&cache->uqc_group_deltas, name, delta); | |
9c5167d1 NF |
1878 | |
1879 | if (dmu_objset_projectquota_enabled(os)) { | |
1880 | (void) sprintf(name, "%llx", (longlong_t)project); | |
1881 | userquota_update_cache(&cache->uqc_project_deltas, | |
1882 | name, delta); | |
1883 | } | |
428870ff BB |
1884 | } |
1885 | } | |
1886 | ||
1de321e6 | 1887 | static void |
9c5167d1 NF |
1888 | do_userobjquota_update(objset_t *os, userquota_cache_t *cache, uint64_t flags, |
1889 | uint64_t user, uint64_t group, uint64_t project, boolean_t subtract) | |
1de321e6 JX |
1890 | { |
1891 | if (flags & DNODE_FLAG_USEROBJUSED_ACCOUNTED) { | |
1892 | char name[20 + DMU_OBJACCT_PREFIX_LEN]; | |
9b7a83cb | 1893 | int delta = subtract ? -1 : 1; |
1de321e6 JX |
1894 | |
1895 | (void) snprintf(name, sizeof (name), DMU_OBJACCT_PREFIX "%llx", | |
1896 | (longlong_t)user); | |
9b7a83cb | 1897 | userquota_update_cache(&cache->uqc_user_deltas, name, delta); |
1de321e6 JX |
1898 | |
1899 | (void) snprintf(name, sizeof (name), DMU_OBJACCT_PREFIX "%llx", | |
1900 | (longlong_t)group); | |
9b7a83cb | 1901 | userquota_update_cache(&cache->uqc_group_deltas, name, delta); |
9c5167d1 NF |
1902 | |
1903 | if (dmu_objset_projectquota_enabled(os)) { | |
1904 | (void) snprintf(name, sizeof (name), | |
1905 | DMU_OBJACCT_PREFIX "%llx", (longlong_t)project); | |
1906 | userquota_update_cache(&cache->uqc_project_deltas, | |
1907 | name, delta); | |
1908 | } | |
1de321e6 JX |
1909 | } |
1910 | } | |
1911 | ||
64fc7762 MA |
1912 | typedef struct userquota_updates_arg { |
1913 | objset_t *uua_os; | |
1914 | int uua_sublist_idx; | |
1915 | dmu_tx_t *uua_tx; | |
1916 | } userquota_updates_arg_t; | |
1917 | ||
1918 | static void | |
1919 | userquota_updates_task(void *arg) | |
9babb374 | 1920 | { |
64fc7762 MA |
1921 | userquota_updates_arg_t *uua = arg; |
1922 | objset_t *os = uua->uua_os; | |
1923 | dmu_tx_t *tx = uua->uua_tx; | |
9babb374 | 1924 | dnode_t *dn; |
9b7a83cb | 1925 | userquota_cache_t cache = { { 0 } }; |
9babb374 | 1926 | |
64fc7762 MA |
1927 | multilist_sublist_t *list = |
1928 | multilist_sublist_lock(os->os_synced_dnodes, uua->uua_sublist_idx); | |
9babb374 | 1929 | |
64fc7762 MA |
1930 | ASSERT(multilist_sublist_head(list) == NULL || |
1931 | dmu_objset_userused_enabled(os)); | |
9b7a83cb JX |
1932 | avl_create(&cache.uqc_user_deltas, userquota_compare, |
1933 | sizeof (userquota_node_t), offsetof(userquota_node_t, uqn_node)); | |
1934 | avl_create(&cache.uqc_group_deltas, userquota_compare, | |
1935 | sizeof (userquota_node_t), offsetof(userquota_node_t, uqn_node)); | |
9c5167d1 NF |
1936 | if (dmu_objset_projectquota_enabled(os)) |
1937 | avl_create(&cache.uqc_project_deltas, userquota_compare, | |
1938 | sizeof (userquota_node_t), offsetof(userquota_node_t, | |
1939 | uqn_node)); | |
9b7a83cb | 1940 | |
64fc7762 | 1941 | while ((dn = multilist_sublist_head(list)) != NULL) { |
572e2857 | 1942 | int flags; |
9babb374 | 1943 | ASSERT(!DMU_OBJECT_IS_SPECIAL(dn->dn_object)); |
9babb374 BB |
1944 | ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE || |
1945 | dn->dn_phys->dn_flags & | |
1946 | DNODE_FLAG_USERUSED_ACCOUNTED); | |
1947 | ||
572e2857 BB |
1948 | flags = dn->dn_id_flags; |
1949 | ASSERT(flags); | |
1950 | if (flags & DN_ID_OLD_EXIST) { | |
9c5167d1 NF |
1951 | do_userquota_update(os, &cache, dn->dn_oldused, |
1952 | dn->dn_oldflags, dn->dn_olduid, dn->dn_oldgid, | |
1953 | dn->dn_oldprojid, B_TRUE); | |
1954 | do_userobjquota_update(os, &cache, dn->dn_oldflags, | |
1955 | dn->dn_olduid, dn->dn_oldgid, | |
1956 | dn->dn_oldprojid, B_TRUE); | |
428870ff | 1957 | } |
572e2857 | 1958 | if (flags & DN_ID_NEW_EXIST) { |
9c5167d1 | 1959 | do_userquota_update(os, &cache, |
9b7a83cb | 1960 | DN_USED_BYTES(dn->dn_phys), dn->dn_phys->dn_flags, |
9c5167d1 NF |
1961 | dn->dn_newuid, dn->dn_newgid, |
1962 | dn->dn_newprojid, B_FALSE); | |
1963 | do_userobjquota_update(os, &cache, | |
1964 | dn->dn_phys->dn_flags, dn->dn_newuid, dn->dn_newgid, | |
1965 | dn->dn_newprojid, B_FALSE); | |
428870ff BB |
1966 | } |
1967 | ||
572e2857 | 1968 | mutex_enter(&dn->dn_mtx); |
428870ff BB |
1969 | dn->dn_oldused = 0; |
1970 | dn->dn_oldflags = 0; | |
1971 | if (dn->dn_id_flags & DN_ID_NEW_EXIST) { | |
1972 | dn->dn_olduid = dn->dn_newuid; | |
1973 | dn->dn_oldgid = dn->dn_newgid; | |
9c5167d1 | 1974 | dn->dn_oldprojid = dn->dn_newprojid; |
428870ff BB |
1975 | dn->dn_id_flags |= DN_ID_OLD_EXIST; |
1976 | if (dn->dn_bonuslen == 0) | |
1977 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1978 | else | |
1979 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1980 | } | |
1981 | dn->dn_id_flags &= ~(DN_ID_NEW_EXIST); | |
edc1e713 TC |
1982 | if (dn->dn_dirty_txg == spa_syncing_txg(os->os_spa)) |
1983 | dn->dn_dirty_txg = 0; | |
9babb374 BB |
1984 | mutex_exit(&dn->dn_mtx); |
1985 | ||
64fc7762 MA |
1986 | multilist_sublist_remove(list, dn); |
1987 | dnode_rele(dn, os->os_synced_dnodes); | |
9babb374 | 1988 | } |
9b7a83cb | 1989 | do_userquota_cacheflush(os, &cache, tx); |
64fc7762 MA |
1990 | multilist_sublist_unlock(list); |
1991 | kmem_free(uua, sizeof (*uua)); | |
1992 | } | |
1993 | ||
1994 | void | |
1995 | dmu_objset_do_userquota_updates(objset_t *os, dmu_tx_t *tx) | |
1996 | { | |
fc754677 AM |
1997 | int num_sublists; |
1998 | ||
64fc7762 MA |
1999 | if (!dmu_objset_userused_enabled(os)) |
2000 | return; | |
2001 | ||
163a8c28 TC |
2002 | /* |
2003 | * If this is a raw receive just return and handle accounting | |
2004 | * later when we have the keys loaded. We also don't do user | |
2005 | * accounting during claiming since the datasets are not owned | |
2006 | * for the duration of claiming and this txg should only be | |
2007 | * used for recovery. | |
2008 | */ | |
b5256303 TC |
2009 | if (os->os_encrypted && dmu_objset_is_receiving(os)) |
2010 | return; | |
2011 | ||
163a8c28 TC |
2012 | if (tx->tx_txg <= os->os_spa->spa_claim_max_txg) |
2013 | return; | |
2014 | ||
9c5167d1 | 2015 | /* Allocate the user/group/project used objects if necessary. */ |
64fc7762 MA |
2016 | if (DMU_USERUSED_DNODE(os)->dn_type == DMU_OT_NONE) { |
2017 | VERIFY0(zap_create_claim(os, | |
2018 | DMU_USERUSED_OBJECT, | |
2019 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
2020 | VERIFY0(zap_create_claim(os, | |
2021 | DMU_GROUPUSED_OBJECT, | |
2022 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
2023 | } | |
2024 | ||
9c5167d1 NF |
2025 | if (dmu_objset_projectquota_enabled(os) && |
2026 | DMU_PROJECTUSED_DNODE(os)->dn_type == DMU_OT_NONE) { | |
2027 | VERIFY0(zap_create_claim(os, DMU_PROJECTUSED_OBJECT, | |
2028 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
2029 | } | |
2030 | ||
fc754677 AM |
2031 | num_sublists = multilist_get_num_sublists(os->os_synced_dnodes); |
2032 | for (int i = 0; i < num_sublists; i++) { | |
2033 | if (multilist_sublist_is_empty_idx(os->os_synced_dnodes, i)) | |
2034 | continue; | |
64fc7762 MA |
2035 | userquota_updates_arg_t *uua = |
2036 | kmem_alloc(sizeof (*uua), KM_SLEEP); | |
2037 | uua->uua_os = os; | |
2038 | uua->uua_sublist_idx = i; | |
2039 | uua->uua_tx = tx; | |
2040 | /* note: caller does taskq_wait() */ | |
2041 | (void) taskq_dispatch(dmu_objset_pool(os)->dp_sync_taskq, | |
2042 | userquota_updates_task, uua, 0); | |
2043 | /* callback frees uua */ | |
2044 | } | |
9babb374 BB |
2045 | } |
2046 | ||
428870ff BB |
2047 | /* |
2048 | * Returns a pointer to data to find uid/gid from | |
2049 | * | |
2050 | * If a dirty record for transaction group that is syncing can't | |
2051 | * be found then NULL is returned. In the NULL case it is assumed | |
2052 | * the uid/gid aren't changing. | |
2053 | */ | |
2054 | static void * | |
2055 | dmu_objset_userquota_find_data(dmu_buf_impl_t *db, dmu_tx_t *tx) | |
2056 | { | |
cccbed9f | 2057 | dbuf_dirty_record_t *dr; |
428870ff BB |
2058 | void *data; |
2059 | ||
2060 | if (db->db_dirtycnt == 0) | |
2061 | return (db->db.db_data); /* Nothing is changing */ | |
2062 | ||
cccbed9f | 2063 | dr = dbuf_find_dirty_eq(db, tx->tx_txg); |
428870ff | 2064 | |
572e2857 | 2065 | if (dr == NULL) { |
428870ff | 2066 | data = NULL; |
572e2857 BB |
2067 | } else { |
2068 | dnode_t *dn; | |
2069 | ||
2070 | DB_DNODE_ENTER(dr->dr_dbuf); | |
2071 | dn = DB_DNODE(dr->dr_dbuf); | |
2072 | ||
2073 | if (dn->dn_bonuslen == 0 && | |
2074 | dr->dr_dbuf->db_blkid == DMU_SPILL_BLKID) | |
2075 | data = dr->dt.dl.dr_data->b_data; | |
2076 | else | |
2077 | data = dr->dt.dl.dr_data; | |
2078 | ||
2079 | DB_DNODE_EXIT(dr->dr_dbuf); | |
2080 | } | |
2081 | ||
428870ff BB |
2082 | return (data); |
2083 | } | |
2084 | ||
2085 | void | |
2086 | dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) | |
2087 | { | |
2088 | objset_t *os = dn->dn_objset; | |
2089 | void *data = NULL; | |
2090 | dmu_buf_impl_t *db = NULL; | |
a117a6d6 GW |
2091 | uint64_t *user = NULL; |
2092 | uint64_t *group = NULL; | |
9c5167d1 | 2093 | uint64_t *project = NULL; |
428870ff BB |
2094 | int flags = dn->dn_id_flags; |
2095 | int error; | |
2096 | boolean_t have_spill = B_FALSE; | |
2097 | ||
2098 | if (!dmu_objset_userused_enabled(dn->dn_objset)) | |
2099 | return; | |
2100 | ||
b5256303 TC |
2101 | /* |
2102 | * Raw receives introduce a problem with user accounting. Raw | |
2103 | * receives cannot update the user accounting info because the | |
2104 | * user ids and the sizes are encrypted. To guarantee that we | |
2105 | * never end up with bad user accounting, we simply disable it | |
2106 | * during raw receives. We also disable this for normal receives | |
2107 | * so that an incremental raw receive may be done on top of an | |
2108 | * existing non-raw receive. | |
2109 | */ | |
2110 | if (os->os_encrypted && dmu_objset_is_receiving(os)) | |
2111 | return; | |
2112 | ||
428870ff BB |
2113 | if (before && (flags & (DN_ID_CHKED_BONUS|DN_ID_OLD_EXIST| |
2114 | DN_ID_CHKED_SPILL))) | |
2115 | return; | |
2116 | ||
2117 | if (before && dn->dn_bonuslen != 0) | |
2118 | data = DN_BONUS(dn->dn_phys); | |
2119 | else if (!before && dn->dn_bonuslen != 0) { | |
a3000f93 BB |
2120 | if (dn->dn_bonus) { |
2121 | db = dn->dn_bonus; | |
428870ff BB |
2122 | mutex_enter(&db->db_mtx); |
2123 | data = dmu_objset_userquota_find_data(db, tx); | |
2124 | } else { | |
2125 | data = DN_BONUS(dn->dn_phys); | |
2126 | } | |
2127 | } else if (dn->dn_bonuslen == 0 && dn->dn_bonustype == DMU_OT_SA) { | |
2128 | int rf = 0; | |
2129 | ||
2130 | if (RW_WRITE_HELD(&dn->dn_struct_rwlock)) | |
2131 | rf |= DB_RF_HAVESTRUCT; | |
572e2857 BB |
2132 | error = dmu_spill_hold_by_dnode(dn, |
2133 | rf | DB_RF_MUST_SUCCEED, | |
428870ff BB |
2134 | FTAG, (dmu_buf_t **)&db); |
2135 | ASSERT(error == 0); | |
2136 | mutex_enter(&db->db_mtx); | |
2137 | data = (before) ? db->db.db_data : | |
2138 | dmu_objset_userquota_find_data(db, tx); | |
2139 | have_spill = B_TRUE; | |
2140 | } else { | |
2141 | mutex_enter(&dn->dn_mtx); | |
2142 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
2143 | mutex_exit(&dn->dn_mtx); | |
2144 | return; | |
2145 | } | |
2146 | ||
2147 | if (before) { | |
2148 | ASSERT(data); | |
2149 | user = &dn->dn_olduid; | |
2150 | group = &dn->dn_oldgid; | |
9c5167d1 | 2151 | project = &dn->dn_oldprojid; |
428870ff BB |
2152 | } else if (data) { |
2153 | user = &dn->dn_newuid; | |
2154 | group = &dn->dn_newgid; | |
9c5167d1 | 2155 | project = &dn->dn_newprojid; |
428870ff BB |
2156 | } |
2157 | ||
2158 | /* | |
2159 | * Must always call the callback in case the object | |
2160 | * type has changed and that type isn't an object type to track | |
2161 | */ | |
2162 | error = used_cbs[os->os_phys->os_type](dn->dn_bonustype, data, | |
9c5167d1 | 2163 | user, group, project); |
428870ff BB |
2164 | |
2165 | /* | |
2166 | * Preserve existing uid/gid when the callback can't determine | |
2167 | * what the new uid/gid are and the callback returned EEXIST. | |
2168 | * The EEXIST error tells us to just use the existing uid/gid. | |
2169 | * If we don't know what the old values are then just assign | |
2170 | * them to 0, since that is a new file being created. | |
2171 | */ | |
2172 | if (!before && data == NULL && error == EEXIST) { | |
2173 | if (flags & DN_ID_OLD_EXIST) { | |
2174 | dn->dn_newuid = dn->dn_olduid; | |
2175 | dn->dn_newgid = dn->dn_oldgid; | |
2705ebf0 | 2176 | dn->dn_newprojid = dn->dn_oldprojid; |
428870ff BB |
2177 | } else { |
2178 | dn->dn_newuid = 0; | |
2179 | dn->dn_newgid = 0; | |
9c5167d1 | 2180 | dn->dn_newprojid = ZFS_DEFAULT_PROJID; |
428870ff BB |
2181 | } |
2182 | error = 0; | |
2183 | } | |
2184 | ||
2185 | if (db) | |
2186 | mutex_exit(&db->db_mtx); | |
2187 | ||
2188 | mutex_enter(&dn->dn_mtx); | |
2189 | if (error == 0 && before) | |
2190 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
2191 | if (error == 0 && !before) | |
2192 | dn->dn_id_flags |= DN_ID_NEW_EXIST; | |
2193 | ||
2194 | if (have_spill) { | |
2195 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
2196 | } else { | |
2197 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
2198 | } | |
2199 | mutex_exit(&dn->dn_mtx); | |
a3000f93 | 2200 | if (have_spill) |
428870ff BB |
2201 | dmu_buf_rele((dmu_buf_t *)db, FTAG); |
2202 | } | |
2203 | ||
9babb374 BB |
2204 | boolean_t |
2205 | dmu_objset_userspace_present(objset_t *os) | |
2206 | { | |
428870ff | 2207 | return (os->os_phys->os_flags & |
9babb374 BB |
2208 | OBJSET_FLAG_USERACCOUNTING_COMPLETE); |
2209 | } | |
2210 | ||
1de321e6 JX |
2211 | boolean_t |
2212 | dmu_objset_userobjspace_present(objset_t *os) | |
2213 | { | |
2214 | return (os->os_phys->os_flags & | |
2215 | OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE); | |
2216 | } | |
2217 | ||
9c5167d1 NF |
2218 | boolean_t |
2219 | dmu_objset_projectquota_present(objset_t *os) | |
2220 | { | |
2221 | return (os->os_phys->os_flags & | |
2222 | OBJSET_FLAG_PROJECTQUOTA_COMPLETE); | |
2223 | } | |
2224 | ||
1de321e6 JX |
2225 | static int |
2226 | dmu_objset_space_upgrade(objset_t *os) | |
9babb374 BB |
2227 | { |
2228 | uint64_t obj; | |
2229 | int err = 0; | |
2230 | ||
9babb374 BB |
2231 | /* |
2232 | * We simply need to mark every object dirty, so that it will be | |
2233 | * synced out and now accounted. If this is called | |
2234 | * concurrently, or if we already did some work before crashing, | |
2235 | * that's fine, since we track each object's accounted state | |
2236 | * independently. | |
2237 | */ | |
2238 | ||
2239 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
45d1cae3 | 2240 | dmu_tx_t *tx; |
9babb374 BB |
2241 | dmu_buf_t *db; |
2242 | int objerr; | |
2243 | ||
1de321e6 JX |
2244 | mutex_enter(&os->os_upgrade_lock); |
2245 | if (os->os_upgrade_exit) | |
2246 | err = SET_ERROR(EINTR); | |
2247 | mutex_exit(&os->os_upgrade_lock); | |
2248 | if (err != 0) | |
2249 | return (err); | |
2250 | ||
9babb374 | 2251 | if (issig(JUSTLOOKING) && issig(FORREAL)) |
2e528b49 | 2252 | return (SET_ERROR(EINTR)); |
9babb374 BB |
2253 | |
2254 | objerr = dmu_bonus_hold(os, obj, FTAG, &db); | |
13fe0198 | 2255 | if (objerr != 0) |
9babb374 | 2256 | continue; |
45d1cae3 | 2257 | tx = dmu_tx_create(os); |
9babb374 BB |
2258 | dmu_tx_hold_bonus(tx, obj); |
2259 | objerr = dmu_tx_assign(tx, TXG_WAIT); | |
13fe0198 | 2260 | if (objerr != 0) { |
d22323e8 | 2261 | dmu_buf_rele(db, FTAG); |
9babb374 BB |
2262 | dmu_tx_abort(tx); |
2263 | continue; | |
2264 | } | |
2265 | dmu_buf_will_dirty(db, tx); | |
2266 | dmu_buf_rele(db, FTAG); | |
2267 | dmu_tx_commit(tx); | |
2268 | } | |
1de321e6 JX |
2269 | return (0); |
2270 | } | |
2271 | ||
2272 | int | |
2273 | dmu_objset_userspace_upgrade(objset_t *os) | |
2274 | { | |
2275 | int err = 0; | |
2276 | ||
2277 | if (dmu_objset_userspace_present(os)) | |
2278 | return (0); | |
2279 | if (dmu_objset_is_snapshot(os)) | |
2280 | return (SET_ERROR(EINVAL)); | |
2281 | if (!dmu_objset_userused_enabled(os)) | |
2282 | return (SET_ERROR(ENOTSUP)); | |
2283 | ||
2284 | err = dmu_objset_space_upgrade(os); | |
2285 | if (err) | |
2286 | return (err); | |
9babb374 | 2287 | |
428870ff | 2288 | os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; |
9babb374 BB |
2289 | txg_wait_synced(dmu_objset_pool(os), 0); |
2290 | return (0); | |
2291 | } | |
2292 | ||
1de321e6 | 2293 | static int |
9c5167d1 | 2294 | dmu_objset_id_quota_upgrade_cb(objset_t *os) |
1de321e6 JX |
2295 | { |
2296 | int err = 0; | |
2297 | ||
9c5167d1 NF |
2298 | if (dmu_objset_userobjspace_present(os) && |
2299 | dmu_objset_projectquota_present(os)) | |
1de321e6 JX |
2300 | return (0); |
2301 | if (dmu_objset_is_snapshot(os)) | |
2302 | return (SET_ERROR(EINVAL)); | |
2303 | if (!dmu_objset_userobjused_enabled(os)) | |
2304 | return (SET_ERROR(ENOTSUP)); | |
9c5167d1 NF |
2305 | if (!dmu_objset_projectquota_enabled(os) && |
2306 | dmu_objset_userobjspace_present(os)) | |
2307 | return (SET_ERROR(ENOTSUP)); | |
1de321e6 | 2308 | |
d52d80b7 PD |
2309 | dmu_objset_ds(os)->ds_feature_activation[ |
2310 | SPA_FEATURE_USEROBJ_ACCOUNTING] = (void *)B_TRUE; | |
9c5167d1 | 2311 | if (dmu_objset_projectquota_enabled(os)) |
d52d80b7 PD |
2312 | dmu_objset_ds(os)->ds_feature_activation[ |
2313 | SPA_FEATURE_PROJECT_QUOTA] = (void *)B_TRUE; | |
1de321e6 JX |
2314 | |
2315 | err = dmu_objset_space_upgrade(os); | |
2316 | if (err) | |
2317 | return (err); | |
2318 | ||
2319 | os->os_flags |= OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE; | |
9c5167d1 NF |
2320 | if (dmu_objset_projectquota_enabled(os)) |
2321 | os->os_flags |= OBJSET_FLAG_PROJECTQUOTA_COMPLETE; | |
2322 | ||
1de321e6 JX |
2323 | txg_wait_synced(dmu_objset_pool(os), 0); |
2324 | return (0); | |
2325 | } | |
2326 | ||
2327 | void | |
9c5167d1 | 2328 | dmu_objset_id_quota_upgrade(objset_t *os) |
1de321e6 | 2329 | { |
9c5167d1 | 2330 | dmu_objset_upgrade(os, dmu_objset_id_quota_upgrade_cb); |
1de321e6 JX |
2331 | } |
2332 | ||
126ae9f4 JX |
2333 | boolean_t |
2334 | dmu_objset_userobjspace_upgradable(objset_t *os) | |
2335 | { | |
2336 | return (dmu_objset_type(os) == DMU_OST_ZFS && | |
2337 | !dmu_objset_is_snapshot(os) && | |
2338 | dmu_objset_userobjused_enabled(os) && | |
bb1be77a | 2339 | !dmu_objset_userobjspace_present(os) && |
2340 | spa_writeable(dmu_objset_spa(os))); | |
126ae9f4 JX |
2341 | } |
2342 | ||
9c5167d1 NF |
2343 | boolean_t |
2344 | dmu_objset_projectquota_upgradable(objset_t *os) | |
2345 | { | |
2346 | return (dmu_objset_type(os) == DMU_OST_ZFS && | |
2347 | !dmu_objset_is_snapshot(os) && | |
2348 | dmu_objset_projectquota_enabled(os) && | |
bb1be77a | 2349 | !dmu_objset_projectquota_present(os) && |
2350 | spa_writeable(dmu_objset_spa(os))); | |
9c5167d1 NF |
2351 | } |
2352 | ||
34dc7c2f BB |
2353 | void |
2354 | dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, | |
2355 | uint64_t *usedobjsp, uint64_t *availobjsp) | |
2356 | { | |
428870ff | 2357 | dsl_dataset_space(os->os_dsl_dataset, refdbytesp, availbytesp, |
34dc7c2f BB |
2358 | usedobjsp, availobjsp); |
2359 | } | |
2360 | ||
2361 | uint64_t | |
2362 | dmu_objset_fsid_guid(objset_t *os) | |
2363 | { | |
428870ff | 2364 | return (dsl_dataset_fsid_guid(os->os_dsl_dataset)); |
34dc7c2f BB |
2365 | } |
2366 | ||
2367 | void | |
2368 | dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat) | |
2369 | { | |
428870ff BB |
2370 | stat->dds_type = os->os_phys->os_type; |
2371 | if (os->os_dsl_dataset) | |
2372 | dsl_dataset_fast_stat(os->os_dsl_dataset, stat); | |
34dc7c2f BB |
2373 | } |
2374 | ||
2375 | void | |
2376 | dmu_objset_stats(objset_t *os, nvlist_t *nv) | |
2377 | { | |
428870ff BB |
2378 | ASSERT(os->os_dsl_dataset || |
2379 | os->os_phys->os_type == DMU_OST_META); | |
34dc7c2f | 2380 | |
428870ff BB |
2381 | if (os->os_dsl_dataset != NULL) |
2382 | dsl_dataset_stats(os->os_dsl_dataset, nv); | |
34dc7c2f BB |
2383 | |
2384 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_TYPE, | |
428870ff | 2385 | os->os_phys->os_type); |
9babb374 BB |
2386 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERACCOUNTING, |
2387 | dmu_objset_userspace_present(os)); | |
34dc7c2f BB |
2388 | } |
2389 | ||
2390 | int | |
2391 | dmu_objset_is_snapshot(objset_t *os) | |
2392 | { | |
428870ff | 2393 | if (os->os_dsl_dataset != NULL) |
0c66c32d | 2394 | return (os->os_dsl_dataset->ds_is_snapshot); |
34dc7c2f BB |
2395 | else |
2396 | return (B_FALSE); | |
2397 | } | |
2398 | ||
2399 | int | |
2400 | dmu_snapshot_realname(objset_t *os, char *name, char *real, int maxlen, | |
2401 | boolean_t *conflict) | |
2402 | { | |
428870ff | 2403 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
2404 | uint64_t ignored; |
2405 | ||
d683ddbb | 2406 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 2407 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2408 | |
2409 | return (zap_lookup_norm(ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb | 2410 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, name, 8, 1, &ignored, |
9b7b9cd3 | 2411 | MT_NORMALIZE, real, maxlen, conflict)); |
34dc7c2f BB |
2412 | } |
2413 | ||
2414 | int | |
2415 | dmu_snapshot_list_next(objset_t *os, int namelen, char *name, | |
2416 | uint64_t *idp, uint64_t *offp, boolean_t *case_conflict) | |
2417 | { | |
428870ff | 2418 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
2419 | zap_cursor_t cursor; |
2420 | zap_attribute_t attr; | |
2421 | ||
13fe0198 MA |
2422 | ASSERT(dsl_pool_config_held(dmu_objset_pool(os))); |
2423 | ||
d683ddbb | 2424 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 2425 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2426 | |
2427 | zap_cursor_init_serialized(&cursor, | |
2428 | ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb | 2429 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, *offp); |
34dc7c2f BB |
2430 | |
2431 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
2432 | zap_cursor_fini(&cursor); | |
2e528b49 | 2433 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2434 | } |
2435 | ||
2436 | if (strlen(attr.za_name) + 1 > namelen) { | |
2437 | zap_cursor_fini(&cursor); | |
2e528b49 | 2438 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
2439 | } |
2440 | ||
2441 | (void) strcpy(name, attr.za_name); | |
2442 | if (idp) | |
2443 | *idp = attr.za_first_integer; | |
2444 | if (case_conflict) | |
2445 | *case_conflict = attr.za_normalization_conflict; | |
2446 | zap_cursor_advance(&cursor); | |
2447 | *offp = zap_cursor_serialize(&cursor); | |
2448 | zap_cursor_fini(&cursor); | |
2449 | ||
2450 | return (0); | |
2451 | } | |
2452 | ||
ebe7e575 | 2453 | int |
6772fb67 | 2454 | dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *value) |
ebe7e575 | 2455 | { |
d1d7e268 | 2456 | return (dsl_dataset_snap_lookup(os->os_dsl_dataset, name, value)); |
ebe7e575 BB |
2457 | } |
2458 | ||
34dc7c2f BB |
2459 | int |
2460 | dmu_dir_list_next(objset_t *os, int namelen, char *name, | |
2461 | uint64_t *idp, uint64_t *offp) | |
2462 | { | |
428870ff | 2463 | dsl_dir_t *dd = os->os_dsl_dataset->ds_dir; |
34dc7c2f BB |
2464 | zap_cursor_t cursor; |
2465 | zap_attribute_t attr; | |
2466 | ||
2467 | /* there is no next dir on a snapshot! */ | |
428870ff | 2468 | if (os->os_dsl_dataset->ds_object != |
d683ddbb | 2469 | dsl_dir_phys(dd)->dd_head_dataset_obj) |
2e528b49 | 2470 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2471 | |
2472 | zap_cursor_init_serialized(&cursor, | |
2473 | dd->dd_pool->dp_meta_objset, | |
d683ddbb | 2474 | dsl_dir_phys(dd)->dd_child_dir_zapobj, *offp); |
34dc7c2f BB |
2475 | |
2476 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
2477 | zap_cursor_fini(&cursor); | |
2e528b49 | 2478 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2479 | } |
2480 | ||
2481 | if (strlen(attr.za_name) + 1 > namelen) { | |
2482 | zap_cursor_fini(&cursor); | |
2e528b49 | 2483 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
2484 | } |
2485 | ||
2486 | (void) strcpy(name, attr.za_name); | |
2487 | if (idp) | |
2488 | *idp = attr.za_first_integer; | |
2489 | zap_cursor_advance(&cursor); | |
2490 | *offp = zap_cursor_serialize(&cursor); | |
2491 | zap_cursor_fini(&cursor); | |
2492 | ||
2493 | return (0); | |
2494 | } | |
2495 | ||
9c43027b AJ |
2496 | typedef struct dmu_objset_find_ctx { |
2497 | taskq_t *dc_tq; | |
2498 | dsl_pool_t *dc_dp; | |
2499 | uint64_t dc_ddobj; | |
1149ba64 | 2500 | char *dc_ddname; /* last component of ddobj's name */ |
9c43027b AJ |
2501 | int (*dc_func)(dsl_pool_t *, dsl_dataset_t *, void *); |
2502 | void *dc_arg; | |
2503 | int dc_flags; | |
2504 | kmutex_t *dc_error_lock; | |
2505 | int *dc_error; | |
2506 | } dmu_objset_find_ctx_t; | |
2507 | ||
2508 | static void | |
2509 | dmu_objset_find_dp_impl(dmu_objset_find_ctx_t *dcp) | |
b128c09f | 2510 | { |
9c43027b | 2511 | dsl_pool_t *dp = dcp->dc_dp; |
13fe0198 MA |
2512 | dsl_dir_t *dd; |
2513 | dsl_dataset_t *ds; | |
2514 | zap_cursor_t zc; | |
2515 | zap_attribute_t *attr; | |
2516 | uint64_t thisobj; | |
9c43027b | 2517 | int err = 0; |
13fe0198 | 2518 | |
9c43027b AJ |
2519 | /* don't process if there already was an error */ |
2520 | if (*dcp->dc_error != 0) | |
2521 | goto out; | |
13fe0198 | 2522 | |
1149ba64 GM |
2523 | /* |
2524 | * Note: passing the name (dc_ddname) here is optional, but it | |
2525 | * improves performance because we don't need to call | |
2526 | * zap_value_search() to determine the name. | |
2527 | */ | |
2528 | err = dsl_dir_hold_obj(dp, dcp->dc_ddobj, dcp->dc_ddname, FTAG, &dd); | |
13fe0198 | 2529 | if (err != 0) |
9c43027b | 2530 | goto out; |
13fe0198 MA |
2531 | |
2532 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ | |
2533 | if (dd->dd_myname[0] == '$') { | |
2534 | dsl_dir_rele(dd, FTAG); | |
9c43027b | 2535 | goto out; |
13fe0198 MA |
2536 | } |
2537 | ||
d683ddbb | 2538 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 2539 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
13fe0198 MA |
2540 | |
2541 | /* | |
2542 | * Iterate over all children. | |
2543 | */ | |
9c43027b | 2544 | if (dcp->dc_flags & DS_FIND_CHILDREN) { |
13fe0198 | 2545 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
d683ddbb | 2546 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
13fe0198 MA |
2547 | zap_cursor_retrieve(&zc, attr) == 0; |
2548 | (void) zap_cursor_advance(&zc)) { | |
2549 | ASSERT3U(attr->za_integer_length, ==, | |
2550 | sizeof (uint64_t)); | |
2551 | ASSERT3U(attr->za_num_integers, ==, 1); | |
2552 | ||
1c27024e | 2553 | dmu_objset_find_ctx_t *child_dcp = |
1149ba64 | 2554 | kmem_alloc(sizeof (*child_dcp), KM_SLEEP); |
9c43027b AJ |
2555 | *child_dcp = *dcp; |
2556 | child_dcp->dc_ddobj = attr->za_first_integer; | |
1149ba64 | 2557 | child_dcp->dc_ddname = spa_strdup(attr->za_name); |
9c43027b AJ |
2558 | if (dcp->dc_tq != NULL) |
2559 | (void) taskq_dispatch(dcp->dc_tq, | |
2560 | dmu_objset_find_dp_cb, child_dcp, TQ_SLEEP); | |
2561 | else | |
2562 | dmu_objset_find_dp_impl(child_dcp); | |
13fe0198 MA |
2563 | } |
2564 | zap_cursor_fini(&zc); | |
13fe0198 MA |
2565 | } |
2566 | ||
2567 | /* | |
2568 | * Iterate over all snapshots. | |
2569 | */ | |
9c43027b | 2570 | if (dcp->dc_flags & DS_FIND_SNAPSHOTS) { |
13fe0198 MA |
2571 | dsl_dataset_t *ds; |
2572 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
2573 | ||
2574 | if (err == 0) { | |
d683ddbb JG |
2575 | uint64_t snapobj; |
2576 | ||
2577 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
13fe0198 MA |
2578 | dsl_dataset_rele(ds, FTAG); |
2579 | ||
2580 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
2581 | zap_cursor_retrieve(&zc, attr) == 0; | |
2582 | (void) zap_cursor_advance(&zc)) { | |
2583 | ASSERT3U(attr->za_integer_length, ==, | |
2584 | sizeof (uint64_t)); | |
2585 | ASSERT3U(attr->za_num_integers, ==, 1); | |
2586 | ||
2587 | err = dsl_dataset_hold_obj(dp, | |
2588 | attr->za_first_integer, FTAG, &ds); | |
2589 | if (err != 0) | |
2590 | break; | |
9c43027b | 2591 | err = dcp->dc_func(dp, ds, dcp->dc_arg); |
13fe0198 MA |
2592 | dsl_dataset_rele(ds, FTAG); |
2593 | if (err != 0) | |
2594 | break; | |
2595 | } | |
2596 | zap_cursor_fini(&zc); | |
2597 | } | |
2598 | } | |
2599 | ||
13fe0198 MA |
2600 | kmem_free(attr, sizeof (zap_attribute_t)); |
2601 | ||
1149ba64 GM |
2602 | if (err != 0) { |
2603 | dsl_dir_rele(dd, FTAG); | |
9c43027b | 2604 | goto out; |
1149ba64 | 2605 | } |
13fe0198 MA |
2606 | |
2607 | /* | |
2608 | * Apply to self. | |
2609 | */ | |
2610 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1149ba64 GM |
2611 | |
2612 | /* | |
2613 | * Note: we hold the dir while calling dsl_dataset_hold_obj() so | |
2614 | * that the dir will remain cached, and we won't have to re-instantiate | |
2615 | * it (which could be expensive due to finding its name via | |
2616 | * zap_value_search()). | |
2617 | */ | |
2618 | dsl_dir_rele(dd, FTAG); | |
13fe0198 | 2619 | if (err != 0) |
9c43027b AJ |
2620 | goto out; |
2621 | err = dcp->dc_func(dp, ds, dcp->dc_arg); | |
13fe0198 | 2622 | dsl_dataset_rele(ds, FTAG); |
9c43027b AJ |
2623 | |
2624 | out: | |
2625 | if (err != 0) { | |
2626 | mutex_enter(dcp->dc_error_lock); | |
2627 | /* only keep first error */ | |
2628 | if (*dcp->dc_error == 0) | |
2629 | *dcp->dc_error = err; | |
2630 | mutex_exit(dcp->dc_error_lock); | |
2631 | } | |
2632 | ||
1149ba64 GM |
2633 | if (dcp->dc_ddname != NULL) |
2634 | spa_strfree(dcp->dc_ddname); | |
9c43027b AJ |
2635 | kmem_free(dcp, sizeof (*dcp)); |
2636 | } | |
2637 | ||
2638 | static void | |
2639 | dmu_objset_find_dp_cb(void *arg) | |
2640 | { | |
2641 | dmu_objset_find_ctx_t *dcp = arg; | |
2642 | dsl_pool_t *dp = dcp->dc_dp; | |
2643 | ||
5e8cd5d1 AJ |
2644 | /* |
2645 | * We need to get a pool_config_lock here, as there are several | |
e1cfd73f | 2646 | * assert(pool_config_held) down the stack. Getting a lock via |
5e8cd5d1 AJ |
2647 | * dsl_pool_config_enter is risky, as it might be stalled by a |
2648 | * pending writer. This would deadlock, as the write lock can | |
2649 | * only be granted when our parent thread gives up the lock. | |
2650 | * The _prio interface gives us priority over a pending writer. | |
2651 | */ | |
2652 | dsl_pool_config_enter_prio(dp, FTAG); | |
9c43027b AJ |
2653 | |
2654 | dmu_objset_find_dp_impl(dcp); | |
2655 | ||
2656 | dsl_pool_config_exit(dp, FTAG); | |
2657 | } | |
2658 | ||
2659 | /* | |
2660 | * Find objsets under and including ddobj, call func(ds) on each. | |
2661 | * The order for the enumeration is completely undefined. | |
2662 | * func is called with dsl_pool_config held. | |
2663 | */ | |
2664 | int | |
2665 | dmu_objset_find_dp(dsl_pool_t *dp, uint64_t ddobj, | |
2666 | int func(dsl_pool_t *, dsl_dataset_t *, void *), void *arg, int flags) | |
2667 | { | |
2668 | int error = 0; | |
2669 | taskq_t *tq = NULL; | |
2670 | int ntasks; | |
2671 | dmu_objset_find_ctx_t *dcp; | |
2672 | kmutex_t err_lock; | |
2673 | ||
2674 | mutex_init(&err_lock, NULL, MUTEX_DEFAULT, NULL); | |
2675 | dcp = kmem_alloc(sizeof (*dcp), KM_SLEEP); | |
2676 | dcp->dc_tq = NULL; | |
2677 | dcp->dc_dp = dp; | |
2678 | dcp->dc_ddobj = ddobj; | |
1149ba64 | 2679 | dcp->dc_ddname = NULL; |
9c43027b AJ |
2680 | dcp->dc_func = func; |
2681 | dcp->dc_arg = arg; | |
2682 | dcp->dc_flags = flags; | |
2683 | dcp->dc_error_lock = &err_lock; | |
2684 | dcp->dc_error = &error; | |
2685 | ||
2686 | if ((flags & DS_FIND_SERIALIZE) || dsl_pool_config_held_writer(dp)) { | |
2687 | /* | |
2688 | * In case a write lock is held we can't make use of | |
2689 | * parallelism, as down the stack of the worker threads | |
2690 | * the lock is asserted via dsl_pool_config_held. | |
2691 | * In case of a read lock this is solved by getting a read | |
2692 | * lock in each worker thread, which isn't possible in case | |
2693 | * of a writer lock. So we fall back to the synchronous path | |
2694 | * here. | |
2695 | * In the future it might be possible to get some magic into | |
2696 | * dsl_pool_config_held in a way that it returns true for | |
2697 | * the worker threads so that a single lock held from this | |
2698 | * thread suffices. For now, stay single threaded. | |
2699 | */ | |
2700 | dmu_objset_find_dp_impl(dcp); | |
e5676636 | 2701 | mutex_destroy(&err_lock); |
9c43027b AJ |
2702 | |
2703 | return (error); | |
2704 | } | |
2705 | ||
2706 | ntasks = dmu_find_threads; | |
2707 | if (ntasks == 0) | |
2708 | ntasks = vdev_count_leaves(dp->dp_spa) * 4; | |
1229323d | 2709 | tq = taskq_create("dmu_objset_find", ntasks, maxclsyspri, ntasks, |
9c43027b AJ |
2710 | INT_MAX, 0); |
2711 | if (tq == NULL) { | |
2712 | kmem_free(dcp, sizeof (*dcp)); | |
e5676636 BB |
2713 | mutex_destroy(&err_lock); |
2714 | ||
9c43027b AJ |
2715 | return (SET_ERROR(ENOMEM)); |
2716 | } | |
2717 | dcp->dc_tq = tq; | |
2718 | ||
2719 | /* dcp will be freed by task */ | |
2720 | (void) taskq_dispatch(tq, dmu_objset_find_dp_cb, dcp, TQ_SLEEP); | |
2721 | ||
2722 | /* | |
2723 | * PORTING: this code relies on the property of taskq_wait to wait | |
2724 | * until no more tasks are queued and no more tasks are active. As | |
2725 | * we always queue new tasks from within other tasks, task_wait | |
2726 | * reliably waits for the full recursion to finish, even though we | |
2727 | * enqueue new tasks after taskq_wait has been called. | |
2728 | * On platforms other than illumos, taskq_wait may not have this | |
2729 | * property. | |
2730 | */ | |
2731 | taskq_wait(tq); | |
2732 | taskq_destroy(tq); | |
2733 | mutex_destroy(&err_lock); | |
2734 | ||
2735 | return (error); | |
b128c09f BB |
2736 | } |
2737 | ||
2738 | /* | |
13fe0198 MA |
2739 | * Find all objsets under name, and for each, call 'func(child_name, arg)'. |
2740 | * The dp_config_rwlock must not be held when this is called, and it | |
2741 | * will not be held when the callback is called. | |
2742 | * Therefore this function should only be used when the pool is not changing | |
2743 | * (e.g. in syncing context), or the callback can deal with the possible races. | |
b128c09f | 2744 | */ |
13fe0198 MA |
2745 | static int |
2746 | dmu_objset_find_impl(spa_t *spa, const char *name, | |
2747 | int func(const char *, void *), void *arg, int flags) | |
34dc7c2f BB |
2748 | { |
2749 | dsl_dir_t *dd; | |
13fe0198 | 2750 | dsl_pool_t *dp = spa_get_dsl(spa); |
b128c09f | 2751 | dsl_dataset_t *ds; |
34dc7c2f BB |
2752 | zap_cursor_t zc; |
2753 | zap_attribute_t *attr; | |
2754 | char *child; | |
b128c09f BB |
2755 | uint64_t thisobj; |
2756 | int err; | |
34dc7c2f | 2757 | |
13fe0198 MA |
2758 | dsl_pool_config_enter(dp, FTAG); |
2759 | ||
2760 | err = dsl_dir_hold(dp, name, FTAG, &dd, NULL); | |
2761 | if (err != 0) { | |
2762 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f | 2763 | return (err); |
13fe0198 | 2764 | } |
34dc7c2f | 2765 | |
b128c09f BB |
2766 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ |
2767 | if (dd->dd_myname[0] == '$') { | |
13fe0198 MA |
2768 | dsl_dir_rele(dd, FTAG); |
2769 | dsl_pool_config_exit(dp, FTAG); | |
b128c09f BB |
2770 | return (0); |
2771 | } | |
2772 | ||
d683ddbb | 2773 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 2774 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
34dc7c2f BB |
2775 | |
2776 | /* | |
2777 | * Iterate over all children. | |
2778 | */ | |
2779 | if (flags & DS_FIND_CHILDREN) { | |
b128c09f | 2780 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
d683ddbb | 2781 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
34dc7c2f BB |
2782 | zap_cursor_retrieve(&zc, attr) == 0; |
2783 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 MA |
2784 | ASSERT3U(attr->za_integer_length, ==, |
2785 | sizeof (uint64_t)); | |
2786 | ASSERT3U(attr->za_num_integers, ==, 1); | |
34dc7c2f | 2787 | |
428870ff | 2788 | child = kmem_asprintf("%s/%s", name, attr->za_name); |
13fe0198 MA |
2789 | dsl_pool_config_exit(dp, FTAG); |
2790 | err = dmu_objset_find_impl(spa, child, | |
2791 | func, arg, flags); | |
2792 | dsl_pool_config_enter(dp, FTAG); | |
e4f5fa12 | 2793 | kmem_strfree(child); |
13fe0198 | 2794 | if (err != 0) |
34dc7c2f BB |
2795 | break; |
2796 | } | |
2797 | zap_cursor_fini(&zc); | |
2798 | ||
13fe0198 MA |
2799 | if (err != 0) { |
2800 | dsl_dir_rele(dd, FTAG); | |
2801 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f BB |
2802 | kmem_free(attr, sizeof (zap_attribute_t)); |
2803 | return (err); | |
2804 | } | |
2805 | } | |
2806 | ||
2807 | /* | |
2808 | * Iterate over all snapshots. | |
2809 | */ | |
b128c09f | 2810 | if (flags & DS_FIND_SNAPSHOTS) { |
b128c09f | 2811 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); |
b128c09f BB |
2812 | |
2813 | if (err == 0) { | |
d683ddbb JG |
2814 | uint64_t snapobj; |
2815 | ||
2816 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
b128c09f BB |
2817 | dsl_dataset_rele(ds, FTAG); |
2818 | ||
2819 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
2820 | zap_cursor_retrieve(&zc, attr) == 0; | |
2821 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 | 2822 | ASSERT3U(attr->za_integer_length, ==, |
b128c09f | 2823 | sizeof (uint64_t)); |
13fe0198 | 2824 | ASSERT3U(attr->za_num_integers, ==, 1); |
b128c09f | 2825 | |
428870ff BB |
2826 | child = kmem_asprintf("%s@%s", |
2827 | name, attr->za_name); | |
13fe0198 MA |
2828 | dsl_pool_config_exit(dp, FTAG); |
2829 | err = func(child, arg); | |
2830 | dsl_pool_config_enter(dp, FTAG); | |
e4f5fa12 | 2831 | kmem_strfree(child); |
13fe0198 | 2832 | if (err != 0) |
b128c09f BB |
2833 | break; |
2834 | } | |
2835 | zap_cursor_fini(&zc); | |
34dc7c2f | 2836 | } |
34dc7c2f BB |
2837 | } |
2838 | ||
13fe0198 | 2839 | dsl_dir_rele(dd, FTAG); |
34dc7c2f | 2840 | kmem_free(attr, sizeof (zap_attribute_t)); |
13fe0198 | 2841 | dsl_pool_config_exit(dp, FTAG); |
34dc7c2f | 2842 | |
13fe0198 | 2843 | if (err != 0) |
34dc7c2f BB |
2844 | return (err); |
2845 | ||
13fe0198 MA |
2846 | /* Apply to self. */ |
2847 | return (func(name, arg)); | |
34dc7c2f BB |
2848 | } |
2849 | ||
13fe0198 MA |
2850 | /* |
2851 | * See comment above dmu_objset_find_impl(). | |
2852 | */ | |
d164b209 | 2853 | int |
5df7e9d8 | 2854 | dmu_objset_find(const char *name, int func(const char *, void *), void *arg, |
13fe0198 | 2855 | int flags) |
d164b209 | 2856 | { |
13fe0198 MA |
2857 | spa_t *spa; |
2858 | int error; | |
d164b209 | 2859 | |
13fe0198 MA |
2860 | error = spa_open(name, &spa, FTAG); |
2861 | if (error != 0) | |
2862 | return (error); | |
2863 | error = dmu_objset_find_impl(spa, name, func, arg, flags); | |
2864 | spa_close(spa, FTAG); | |
2865 | return (error); | |
d164b209 BB |
2866 | } |
2867 | ||
ae76f45c TC |
2868 | boolean_t |
2869 | dmu_objset_incompatible_encryption_version(objset_t *os) | |
2870 | { | |
2871 | return (dsl_dir_incompatible_encryption_version( | |
2872 | os->os_dsl_dataset->ds_dir)); | |
2873 | } | |
2874 | ||
34dc7c2f BB |
2875 | void |
2876 | dmu_objset_set_user(objset_t *os, void *user_ptr) | |
2877 | { | |
428870ff BB |
2878 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
2879 | os->os_user_ptr = user_ptr; | |
34dc7c2f BB |
2880 | } |
2881 | ||
2882 | void * | |
2883 | dmu_objset_get_user(objset_t *os) | |
2884 | { | |
428870ff BB |
2885 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
2886 | return (os->os_user_ptr); | |
34dc7c2f | 2887 | } |
c28b2279 | 2888 | |
13fe0198 MA |
2889 | /* |
2890 | * Determine name of filesystem, given name of snapshot. | |
eca7b760 | 2891 | * buf must be at least ZFS_MAX_DATASET_NAME_LEN bytes |
13fe0198 MA |
2892 | */ |
2893 | int | |
2894 | dmu_fsname(const char *snapname, char *buf) | |
2895 | { | |
2896 | char *atp = strchr(snapname, '@'); | |
2897 | if (atp == NULL) | |
2e528b49 | 2898 | return (SET_ERROR(EINVAL)); |
eca7b760 | 2899 | if (atp - snapname >= ZFS_MAX_DATASET_NAME_LEN) |
2e528b49 | 2900 | return (SET_ERROR(ENAMETOOLONG)); |
13fe0198 MA |
2901 | (void) strlcpy(buf, snapname, atp - snapname + 1); |
2902 | return (0); | |
2903 | } | |
2904 | ||
3ec3bc21 | 2905 | /* |
0f8ff49e SD |
2906 | * Call when we think we're going to write/free space in open context |
2907 | * to track the amount of dirty data in the open txg, which is also the | |
2908 | * amount of memory that can not be evicted until this txg syncs. | |
2909 | * | |
2910 | * Note that there are two conditions where this can be called from | |
2911 | * syncing context: | |
2912 | * | |
2913 | * [1] When we just created the dataset, in which case we go on with | |
2914 | * updating any accounting of dirty data as usual. | |
2915 | * [2] When we are dirtying MOS data, in which case we only update the | |
2916 | * pool's accounting of dirty data. | |
3ec3bc21 BB |
2917 | */ |
2918 | void | |
2919 | dmu_objset_willuse_space(objset_t *os, int64_t space, dmu_tx_t *tx) | |
2920 | { | |
2921 | dsl_dataset_t *ds = os->os_dsl_dataset; | |
2922 | int64_t aspace = spa_get_worst_case_asize(os->os_spa, space); | |
2923 | ||
2924 | if (ds != NULL) { | |
2925 | dsl_dir_willuse_space(ds->ds_dir, aspace, tx); | |
3ec3bc21 | 2926 | } |
0f8ff49e SD |
2927 | |
2928 | dsl_pool_dirty_space(dmu_tx_pool(tx), space, tx); | |
3ec3bc21 BB |
2929 | } |
2930 | ||
93ce2b4c | 2931 | #if defined(_KERNEL) |
f0fd83be | 2932 | EXPORT_SYMBOL(dmu_objset_zil); |
c28b2279 | 2933 | EXPORT_SYMBOL(dmu_objset_pool); |
f0fd83be BB |
2934 | EXPORT_SYMBOL(dmu_objset_ds); |
2935 | EXPORT_SYMBOL(dmu_objset_type); | |
c28b2279 BB |
2936 | EXPORT_SYMBOL(dmu_objset_name); |
2937 | EXPORT_SYMBOL(dmu_objset_hold); | |
b5256303 | 2938 | EXPORT_SYMBOL(dmu_objset_hold_flags); |
c28b2279 BB |
2939 | EXPORT_SYMBOL(dmu_objset_own); |
2940 | EXPORT_SYMBOL(dmu_objset_rele); | |
b5256303 | 2941 | EXPORT_SYMBOL(dmu_objset_rele_flags); |
c28b2279 BB |
2942 | EXPORT_SYMBOL(dmu_objset_disown); |
2943 | EXPORT_SYMBOL(dmu_objset_from_ds); | |
2944 | EXPORT_SYMBOL(dmu_objset_create); | |
2945 | EXPORT_SYMBOL(dmu_objset_clone); | |
c28b2279 BB |
2946 | EXPORT_SYMBOL(dmu_objset_stats); |
2947 | EXPORT_SYMBOL(dmu_objset_fast_stat); | |
54a179e7 | 2948 | EXPORT_SYMBOL(dmu_objset_spa); |
c28b2279 BB |
2949 | EXPORT_SYMBOL(dmu_objset_space); |
2950 | EXPORT_SYMBOL(dmu_objset_fsid_guid); | |
2951 | EXPORT_SYMBOL(dmu_objset_find); | |
c28b2279 BB |
2952 | EXPORT_SYMBOL(dmu_objset_byteswap); |
2953 | EXPORT_SYMBOL(dmu_objset_evict_dbufs); | |
2954 | EXPORT_SYMBOL(dmu_objset_snap_cmtime); | |
50c957f7 | 2955 | EXPORT_SYMBOL(dmu_objset_dnodesize); |
c28b2279 BB |
2956 | |
2957 | EXPORT_SYMBOL(dmu_objset_sync); | |
2958 | EXPORT_SYMBOL(dmu_objset_is_dirty); | |
b5256303 | 2959 | EXPORT_SYMBOL(dmu_objset_create_impl_dnstats); |
c28b2279 BB |
2960 | EXPORT_SYMBOL(dmu_objset_create_impl); |
2961 | EXPORT_SYMBOL(dmu_objset_open_impl); | |
2962 | EXPORT_SYMBOL(dmu_objset_evict); | |
2963 | EXPORT_SYMBOL(dmu_objset_register_type); | |
2964 | EXPORT_SYMBOL(dmu_objset_do_userquota_updates); | |
2965 | EXPORT_SYMBOL(dmu_objset_userquota_get_ids); | |
2966 | EXPORT_SYMBOL(dmu_objset_userused_enabled); | |
2967 | EXPORT_SYMBOL(dmu_objset_userspace_upgrade); | |
2968 | EXPORT_SYMBOL(dmu_objset_userspace_present); | |
1de321e6 | 2969 | EXPORT_SYMBOL(dmu_objset_userobjused_enabled); |
126ae9f4 | 2970 | EXPORT_SYMBOL(dmu_objset_userobjspace_upgradable); |
1de321e6 | 2971 | EXPORT_SYMBOL(dmu_objset_userobjspace_present); |
9c5167d1 NF |
2972 | EXPORT_SYMBOL(dmu_objset_projectquota_enabled); |
2973 | EXPORT_SYMBOL(dmu_objset_projectquota_present); | |
2974 | EXPORT_SYMBOL(dmu_objset_projectquota_upgradable); | |
2975 | EXPORT_SYMBOL(dmu_objset_id_quota_upgrade); | |
c28b2279 | 2976 | #endif |