]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
1 | /* |
2 | * CDDL HEADER START | |
3 | * | |
4 | * The contents of this file are subject to the terms of the | |
5 | * Common Development and Distribution License (the "License"). | |
6 | * You may not use this file except in compliance with the License. | |
7 | * | |
8 | * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE | |
9 | * or http://www.opensolaris.org/os/licensing. | |
10 | * See the License for the specific language governing permissions | |
11 | * and limitations under the License. | |
12 | * | |
13 | * When distributing Covered Code, include this CDDL HEADER in each | |
14 | * file and include the License file at usr/src/OPENSOLARIS.LICENSE. | |
15 | * If applicable, add the following below this CDDL HEADER, with the | |
16 | * fields enclosed by brackets "[]" replaced with your own identifying | |
17 | * information: Portions Copyright [yyyy] [name of copyright owner] | |
18 | * | |
19 | * CDDL HEADER END | |
20 | */ | |
21 | /* | |
428870ff | 22 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
bc77ba73 | 23 | * Copyright (c) 2012, 2016 by Delphix. All rights reserved. |
3a17a7a9 | 24 | * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. |
788eb90c | 25 | * Copyright (c) 2013, Joyent, Inc. All rights reserved. |
0c66c32d | 26 | * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. |
1cddb8c9 | 27 | * Copyright (c) 2015 Nexenta Systems, Inc. All rights reserved. |
9c43027b | 28 | * Copyright (c) 2015, STRATO AG, Inc. All rights reserved. |
a0bd735a | 29 | * Copyright (c) 2016 Actifio, Inc. All rights reserved. |
34dc7c2f BB |
30 | */ |
31 | ||
428870ff BB |
32 | /* Portions Copyright 2010 Robert Milkowski */ |
33 | ||
34dc7c2f BB |
34 | #include <sys/cred.h> |
35 | #include <sys/zfs_context.h> | |
36 | #include <sys/dmu_objset.h> | |
37 | #include <sys/dsl_dir.h> | |
38 | #include <sys/dsl_dataset.h> | |
39 | #include <sys/dsl_prop.h> | |
40 | #include <sys/dsl_pool.h> | |
41 | #include <sys/dsl_synctask.h> | |
42 | #include <sys/dsl_deleg.h> | |
43 | #include <sys/dnode.h> | |
44 | #include <sys/dbuf.h> | |
45 | #include <sys/zvol.h> | |
46 | #include <sys/dmu_tx.h> | |
34dc7c2f BB |
47 | #include <sys/zap.h> |
48 | #include <sys/zil.h> | |
49 | #include <sys/dmu_impl.h> | |
50 | #include <sys/zfs_ioctl.h> | |
428870ff | 51 | #include <sys/sa.h> |
572e2857 | 52 | #include <sys/zfs_onexit.h> |
13fe0198 | 53 | #include <sys/dsl_destroy.h> |
9c43027b | 54 | #include <sys/vdev.h> |
f74b821a | 55 | #include <sys/policy.h> |
572e2857 BB |
56 | |
57 | /* | |
58 | * Needed to close a window in dnode_move() that allows the objset to be freed | |
59 | * before it can be safely accessed. | |
60 | */ | |
61 | krwlock_t os_lock; | |
62 | ||
9c43027b AJ |
63 | /* |
64 | * Tunable to overwrite the maximum number of threads for the parallization | |
65 | * of dmu_objset_find_dp, needed to speed up the import of pools with many | |
66 | * datasets. | |
67 | * Default is 4 times the number of leaf vdevs. | |
68 | */ | |
69 | int dmu_find_threads = 0; | |
70 | ||
68cbd56e NB |
71 | /* |
72 | * Backfill lower metadnode objects after this many have been freed. | |
73 | * Backfilling negatively impacts object creation rates, so only do it | |
74 | * if there are enough holes to fill. | |
75 | */ | |
76 | int dmu_rescan_dnode_threshold = 1 << DN_MAX_INDBLKSHIFT; | |
77 | ||
9c43027b AJ |
78 | static void dmu_objset_find_dp_cb(void *arg); |
79 | ||
572e2857 BB |
80 | void |
81 | dmu_objset_init(void) | |
82 | { | |
83 | rw_init(&os_lock, NULL, RW_DEFAULT, NULL); | |
84 | } | |
85 | ||
86 | void | |
87 | dmu_objset_fini(void) | |
88 | { | |
89 | rw_destroy(&os_lock); | |
90 | } | |
34dc7c2f BB |
91 | |
92 | spa_t * | |
93 | dmu_objset_spa(objset_t *os) | |
94 | { | |
428870ff | 95 | return (os->os_spa); |
34dc7c2f BB |
96 | } |
97 | ||
98 | zilog_t * | |
99 | dmu_objset_zil(objset_t *os) | |
100 | { | |
428870ff | 101 | return (os->os_zil); |
34dc7c2f BB |
102 | } |
103 | ||
104 | dsl_pool_t * | |
105 | dmu_objset_pool(objset_t *os) | |
106 | { | |
107 | dsl_dataset_t *ds; | |
108 | ||
428870ff | 109 | if ((ds = os->os_dsl_dataset) != NULL && ds->ds_dir) |
34dc7c2f BB |
110 | return (ds->ds_dir->dd_pool); |
111 | else | |
428870ff | 112 | return (spa_get_dsl(os->os_spa)); |
34dc7c2f BB |
113 | } |
114 | ||
115 | dsl_dataset_t * | |
116 | dmu_objset_ds(objset_t *os) | |
117 | { | |
428870ff | 118 | return (os->os_dsl_dataset); |
34dc7c2f BB |
119 | } |
120 | ||
121 | dmu_objset_type_t | |
122 | dmu_objset_type(objset_t *os) | |
123 | { | |
428870ff | 124 | return (os->os_phys->os_type); |
34dc7c2f BB |
125 | } |
126 | ||
127 | void | |
128 | dmu_objset_name(objset_t *os, char *buf) | |
129 | { | |
428870ff | 130 | dsl_dataset_name(os->os_dsl_dataset, buf); |
34dc7c2f BB |
131 | } |
132 | ||
133 | uint64_t | |
134 | dmu_objset_id(objset_t *os) | |
135 | { | |
428870ff | 136 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
137 | |
138 | return (ds ? ds->ds_object : 0); | |
139 | } | |
140 | ||
50c957f7 NB |
141 | uint64_t |
142 | dmu_objset_dnodesize(objset_t *os) | |
143 | { | |
144 | return (os->os_dnodesize); | |
145 | } | |
146 | ||
faf0f58c | 147 | zfs_sync_type_t |
428870ff BB |
148 | dmu_objset_syncprop(objset_t *os) |
149 | { | |
150 | return (os->os_sync); | |
151 | } | |
152 | ||
faf0f58c | 153 | zfs_logbias_op_t |
428870ff BB |
154 | dmu_objset_logbias(objset_t *os) |
155 | { | |
156 | return (os->os_logbias); | |
157 | } | |
158 | ||
34dc7c2f BB |
159 | static void |
160 | checksum_changed_cb(void *arg, uint64_t newval) | |
161 | { | |
428870ff | 162 | objset_t *os = arg; |
34dc7c2f BB |
163 | |
164 | /* | |
165 | * Inheritance should have been done by now. | |
166 | */ | |
167 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
168 | ||
428870ff | 169 | os->os_checksum = zio_checksum_select(newval, ZIO_CHECKSUM_ON_VALUE); |
34dc7c2f BB |
170 | } |
171 | ||
172 | static void | |
173 | compression_changed_cb(void *arg, uint64_t newval) | |
174 | { | |
428870ff | 175 | objset_t *os = arg; |
34dc7c2f BB |
176 | |
177 | /* | |
178 | * Inheritance and range checking should have been done by now. | |
179 | */ | |
180 | ASSERT(newval != ZIO_COMPRESS_INHERIT); | |
181 | ||
99197f03 JG |
182 | os->os_compress = zio_compress_select(os->os_spa, newval, |
183 | ZIO_COMPRESS_ON); | |
34dc7c2f BB |
184 | } |
185 | ||
186 | static void | |
187 | copies_changed_cb(void *arg, uint64_t newval) | |
188 | { | |
428870ff | 189 | objset_t *os = arg; |
34dc7c2f BB |
190 | |
191 | /* | |
192 | * Inheritance and range checking should have been done by now. | |
193 | */ | |
194 | ASSERT(newval > 0); | |
428870ff | 195 | ASSERT(newval <= spa_max_replication(os->os_spa)); |
34dc7c2f | 196 | |
428870ff BB |
197 | os->os_copies = newval; |
198 | } | |
199 | ||
200 | static void | |
201 | dedup_changed_cb(void *arg, uint64_t newval) | |
202 | { | |
203 | objset_t *os = arg; | |
204 | spa_t *spa = os->os_spa; | |
205 | enum zio_checksum checksum; | |
206 | ||
207 | /* | |
208 | * Inheritance should have been done by now. | |
209 | */ | |
210 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
211 | ||
212 | checksum = zio_checksum_dedup_select(spa, newval, ZIO_CHECKSUM_OFF); | |
213 | ||
214 | os->os_dedup_checksum = checksum & ZIO_CHECKSUM_MASK; | |
215 | os->os_dedup_verify = !!(checksum & ZIO_CHECKSUM_VERIFY); | |
34dc7c2f BB |
216 | } |
217 | ||
b128c09f BB |
218 | static void |
219 | primary_cache_changed_cb(void *arg, uint64_t newval) | |
220 | { | |
428870ff | 221 | objset_t *os = arg; |
b128c09f BB |
222 | |
223 | /* | |
224 | * Inheritance and range checking should have been done by now. | |
225 | */ | |
226 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
227 | newval == ZFS_CACHE_METADATA); | |
228 | ||
428870ff | 229 | os->os_primary_cache = newval; |
b128c09f BB |
230 | } |
231 | ||
232 | static void | |
233 | secondary_cache_changed_cb(void *arg, uint64_t newval) | |
234 | { | |
428870ff | 235 | objset_t *os = arg; |
b128c09f BB |
236 | |
237 | /* | |
238 | * Inheritance and range checking should have been done by now. | |
239 | */ | |
240 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
241 | newval == ZFS_CACHE_METADATA); | |
242 | ||
428870ff BB |
243 | os->os_secondary_cache = newval; |
244 | } | |
245 | ||
246 | static void | |
247 | sync_changed_cb(void *arg, uint64_t newval) | |
248 | { | |
249 | objset_t *os = arg; | |
250 | ||
251 | /* | |
252 | * Inheritance and range checking should have been done by now. | |
253 | */ | |
254 | ASSERT(newval == ZFS_SYNC_STANDARD || newval == ZFS_SYNC_ALWAYS || | |
255 | newval == ZFS_SYNC_DISABLED); | |
256 | ||
257 | os->os_sync = newval; | |
258 | if (os->os_zil) | |
259 | zil_set_sync(os->os_zil, newval); | |
260 | } | |
261 | ||
faf0f58c MA |
262 | static void |
263 | redundant_metadata_changed_cb(void *arg, uint64_t newval) | |
264 | { | |
265 | objset_t *os = arg; | |
266 | ||
267 | /* | |
268 | * Inheritance and range checking should have been done by now. | |
269 | */ | |
270 | ASSERT(newval == ZFS_REDUNDANT_METADATA_ALL || | |
271 | newval == ZFS_REDUNDANT_METADATA_MOST); | |
272 | ||
273 | os->os_redundant_metadata = newval; | |
274 | } | |
275 | ||
50c957f7 NB |
276 | static void |
277 | dnodesize_changed_cb(void *arg, uint64_t newval) | |
278 | { | |
279 | objset_t *os = arg; | |
280 | ||
281 | switch (newval) { | |
282 | case ZFS_DNSIZE_LEGACY: | |
283 | os->os_dnodesize = DNODE_MIN_SIZE; | |
284 | break; | |
285 | case ZFS_DNSIZE_AUTO: | |
286 | /* | |
287 | * Choose a dnode size that will work well for most | |
288 | * workloads if the user specified "auto". Future code | |
289 | * improvements could dynamically select a dnode size | |
290 | * based on observed workload patterns. | |
291 | */ | |
292 | os->os_dnodesize = DNODE_MIN_SIZE * 2; | |
293 | break; | |
294 | case ZFS_DNSIZE_1K: | |
295 | case ZFS_DNSIZE_2K: | |
296 | case ZFS_DNSIZE_4K: | |
297 | case ZFS_DNSIZE_8K: | |
298 | case ZFS_DNSIZE_16K: | |
299 | os->os_dnodesize = newval; | |
300 | break; | |
301 | } | |
302 | } | |
303 | ||
428870ff BB |
304 | static void |
305 | logbias_changed_cb(void *arg, uint64_t newval) | |
306 | { | |
307 | objset_t *os = arg; | |
308 | ||
309 | ASSERT(newval == ZFS_LOGBIAS_LATENCY || | |
310 | newval == ZFS_LOGBIAS_THROUGHPUT); | |
311 | os->os_logbias = newval; | |
312 | if (os->os_zil) | |
313 | zil_set_logbias(os->os_zil, newval); | |
b128c09f BB |
314 | } |
315 | ||
f1512ee6 MA |
316 | static void |
317 | recordsize_changed_cb(void *arg, uint64_t newval) | |
318 | { | |
319 | objset_t *os = arg; | |
320 | ||
321 | os->os_recordsize = newval; | |
322 | } | |
323 | ||
34dc7c2f BB |
324 | void |
325 | dmu_objset_byteswap(void *buf, size_t size) | |
326 | { | |
327 | objset_phys_t *osp = buf; | |
328 | ||
9babb374 | 329 | ASSERT(size == OBJSET_OLD_PHYS_SIZE || size == sizeof (objset_phys_t)); |
34dc7c2f BB |
330 | dnode_byteswap(&osp->os_meta_dnode); |
331 | byteswap_uint64_array(&osp->os_zil_header, sizeof (zil_header_t)); | |
332 | osp->os_type = BSWAP_64(osp->os_type); | |
9babb374 BB |
333 | osp->os_flags = BSWAP_64(osp->os_flags); |
334 | if (size == sizeof (objset_phys_t)) { | |
335 | dnode_byteswap(&osp->os_userused_dnode); | |
336 | dnode_byteswap(&osp->os_groupused_dnode); | |
337 | } | |
34dc7c2f BB |
338 | } |
339 | ||
340 | int | |
341 | dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, | |
428870ff | 342 | objset_t **osp) |
34dc7c2f | 343 | { |
428870ff | 344 | objset_t *os; |
b128c09f | 345 | int i, err; |
34dc7c2f BB |
346 | |
347 | ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock)); | |
348 | ||
79c76d5b | 349 | os = kmem_zalloc(sizeof (objset_t), KM_SLEEP); |
428870ff BB |
350 | os->os_dsl_dataset = ds; |
351 | os->os_spa = spa; | |
352 | os->os_rootbp = bp; | |
353 | if (!BP_IS_HOLE(os->os_rootbp)) { | |
2a432414 | 354 | arc_flags_t aflags = ARC_FLAG_WAIT; |
5dbd68a3 | 355 | zbookmark_phys_t zb; |
428870ff BB |
356 | SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET, |
357 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
358 | ||
359 | if (DMU_OS_IS_L2CACHEABLE(os)) | |
2a432414 | 360 | aflags |= ARC_FLAG_L2CACHE; |
3a17a7a9 | 361 | if (DMU_OS_IS_L2COMPRESSIBLE(os)) |
2a432414 | 362 | aflags |= ARC_FLAG_L2COMPRESS; |
34dc7c2f | 363 | |
428870ff | 364 | dprintf_bp(os->os_rootbp, "reading %s", ""); |
294f6806 | 365 | err = arc_read(NULL, spa, os->os_rootbp, |
428870ff | 366 | arc_getbuf_func, &os->os_phys_buf, |
34dc7c2f | 367 | ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL, &aflags, &zb); |
13fe0198 | 368 | if (err != 0) { |
428870ff | 369 | kmem_free(os, sizeof (objset_t)); |
b128c09f BB |
370 | /* convert checksum errors into IO errors */ |
371 | if (err == ECKSUM) | |
2e528b49 | 372 | err = SET_ERROR(EIO); |
34dc7c2f BB |
373 | return (err); |
374 | } | |
9babb374 BB |
375 | |
376 | /* Increase the blocksize if we are permitted. */ | |
377 | if (spa_version(spa) >= SPA_VERSION_USERSPACE && | |
428870ff | 378 | arc_buf_size(os->os_phys_buf) < sizeof (objset_phys_t)) { |
9babb374 | 379 | arc_buf_t *buf = arc_buf_alloc(spa, |
428870ff | 380 | sizeof (objset_phys_t), &os->os_phys_buf, |
9babb374 BB |
381 | ARC_BUFC_METADATA); |
382 | bzero(buf->b_data, sizeof (objset_phys_t)); | |
428870ff BB |
383 | bcopy(os->os_phys_buf->b_data, buf->b_data, |
384 | arc_buf_size(os->os_phys_buf)); | |
385 | (void) arc_buf_remove_ref(os->os_phys_buf, | |
386 | &os->os_phys_buf); | |
387 | os->os_phys_buf = buf; | |
9babb374 BB |
388 | } |
389 | ||
428870ff BB |
390 | os->os_phys = os->os_phys_buf->b_data; |
391 | os->os_flags = os->os_phys->os_flags; | |
34dc7c2f | 392 | } else { |
9babb374 BB |
393 | int size = spa_version(spa) >= SPA_VERSION_USERSPACE ? |
394 | sizeof (objset_phys_t) : OBJSET_OLD_PHYS_SIZE; | |
428870ff BB |
395 | os->os_phys_buf = arc_buf_alloc(spa, size, |
396 | &os->os_phys_buf, ARC_BUFC_METADATA); | |
397 | os->os_phys = os->os_phys_buf->b_data; | |
398 | bzero(os->os_phys, size); | |
34dc7c2f BB |
399 | } |
400 | ||
401 | /* | |
402 | * Note: the changed_cb will be called once before the register | |
403 | * func returns, thus changing the checksum/compression from the | |
b128c09f BB |
404 | * default (fletcher2/off). Snapshots don't need to know about |
405 | * checksum/compression/copies. | |
34dc7c2f | 406 | */ |
9b67f605 | 407 | if (ds != NULL) { |
13fe0198 MA |
408 | err = dsl_prop_register(ds, |
409 | zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE), | |
428870ff | 410 | primary_cache_changed_cb, os); |
13fe0198 MA |
411 | if (err == 0) { |
412 | err = dsl_prop_register(ds, | |
413 | zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE), | |
428870ff | 414 | secondary_cache_changed_cb, os); |
13fe0198 | 415 | } |
0c66c32d | 416 | if (!ds->ds_is_snapshot) { |
13fe0198 MA |
417 | if (err == 0) { |
418 | err = dsl_prop_register(ds, | |
419 | zfs_prop_to_name(ZFS_PROP_CHECKSUM), | |
428870ff | 420 | checksum_changed_cb, os); |
13fe0198 MA |
421 | } |
422 | if (err == 0) { | |
423 | err = dsl_prop_register(ds, | |
424 | zfs_prop_to_name(ZFS_PROP_COMPRESSION), | |
428870ff | 425 | compression_changed_cb, os); |
13fe0198 MA |
426 | } |
427 | if (err == 0) { | |
428 | err = dsl_prop_register(ds, | |
429 | zfs_prop_to_name(ZFS_PROP_COPIES), | |
428870ff | 430 | copies_changed_cb, os); |
13fe0198 MA |
431 | } |
432 | if (err == 0) { | |
433 | err = dsl_prop_register(ds, | |
434 | zfs_prop_to_name(ZFS_PROP_DEDUP), | |
428870ff | 435 | dedup_changed_cb, os); |
13fe0198 MA |
436 | } |
437 | if (err == 0) { | |
438 | err = dsl_prop_register(ds, | |
439 | zfs_prop_to_name(ZFS_PROP_LOGBIAS), | |
428870ff | 440 | logbias_changed_cb, os); |
13fe0198 MA |
441 | } |
442 | if (err == 0) { | |
443 | err = dsl_prop_register(ds, | |
444 | zfs_prop_to_name(ZFS_PROP_SYNC), | |
428870ff | 445 | sync_changed_cb, os); |
13fe0198 | 446 | } |
faf0f58c MA |
447 | if (err == 0) { |
448 | err = dsl_prop_register(ds, | |
449 | zfs_prop_to_name( | |
450 | ZFS_PROP_REDUNDANT_METADATA), | |
451 | redundant_metadata_changed_cb, os); | |
452 | } | |
f1512ee6 MA |
453 | if (err == 0) { |
454 | err = dsl_prop_register(ds, | |
455 | zfs_prop_to_name(ZFS_PROP_RECORDSIZE), | |
456 | recordsize_changed_cb, os); | |
457 | } | |
50c957f7 NB |
458 | if (err == 0) { |
459 | err = dsl_prop_register(ds, | |
460 | zfs_prop_to_name(ZFS_PROP_DNODESIZE), | |
461 | dnodesize_changed_cb, os); | |
462 | } | |
b128c09f | 463 | } |
13fe0198 | 464 | if (err != 0) { |
428870ff | 465 | VERIFY(arc_buf_remove_ref(os->os_phys_buf, |
13fe0198 | 466 | &os->os_phys_buf)); |
428870ff | 467 | kmem_free(os, sizeof (objset_t)); |
34dc7c2f BB |
468 | return (err); |
469 | } | |
9b67f605 | 470 | } else { |
34dc7c2f | 471 | /* It's the meta-objset. */ |
428870ff | 472 | os->os_checksum = ZIO_CHECKSUM_FLETCHER_4; |
99197f03 | 473 | os->os_compress = ZIO_COMPRESS_ON; |
428870ff BB |
474 | os->os_copies = spa_max_replication(spa); |
475 | os->os_dedup_checksum = ZIO_CHECKSUM_OFF; | |
faf0f58c MA |
476 | os->os_dedup_verify = B_FALSE; |
477 | os->os_logbias = ZFS_LOGBIAS_LATENCY; | |
478 | os->os_sync = ZFS_SYNC_STANDARD; | |
428870ff BB |
479 | os->os_primary_cache = ZFS_CACHE_ALL; |
480 | os->os_secondary_cache = ZFS_CACHE_ALL; | |
50c957f7 | 481 | os->os_dnodesize = DNODE_MIN_SIZE; |
34dc7c2f BB |
482 | } |
483 | ||
0c66c32d | 484 | if (ds == NULL || !ds->ds_is_snapshot) |
572e2857 | 485 | os->os_zil_header = os->os_phys->os_zil_header; |
428870ff | 486 | os->os_zil = zil_alloc(os, &os->os_zil_header); |
34dc7c2f BB |
487 | |
488 | for (i = 0; i < TXG_SIZE; i++) { | |
428870ff | 489 | list_create(&os->os_dirty_dnodes[i], sizeof (dnode_t), |
34dc7c2f | 490 | offsetof(dnode_t, dn_dirty_link[i])); |
428870ff | 491 | list_create(&os->os_free_dnodes[i], sizeof (dnode_t), |
34dc7c2f BB |
492 | offsetof(dnode_t, dn_dirty_link[i])); |
493 | } | |
428870ff | 494 | list_create(&os->os_dnodes, sizeof (dnode_t), |
34dc7c2f | 495 | offsetof(dnode_t, dn_link)); |
428870ff | 496 | list_create(&os->os_downgraded_dbufs, sizeof (dmu_buf_impl_t), |
34dc7c2f BB |
497 | offsetof(dmu_buf_impl_t, db_link)); |
498 | ||
0c66c32d JG |
499 | list_link_init(&os->os_evicting_node); |
500 | ||
428870ff BB |
501 | mutex_init(&os->os_lock, NULL, MUTEX_DEFAULT, NULL); |
502 | mutex_init(&os->os_obj_lock, NULL, MUTEX_DEFAULT, NULL); | |
503 | mutex_init(&os->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL); | |
504 | ||
0c66c32d JG |
505 | dnode_special_open(os, &os->os_phys->os_meta_dnode, |
506 | DMU_META_DNODE_OBJECT, &os->os_meta_dnode); | |
428870ff | 507 | if (arc_buf_size(os->os_phys_buf) >= sizeof (objset_phys_t)) { |
0c66c32d JG |
508 | dnode_special_open(os, &os->os_phys->os_userused_dnode, |
509 | DMU_USERUSED_OBJECT, &os->os_userused_dnode); | |
510 | dnode_special_open(os, &os->os_phys->os_groupused_dnode, | |
511 | DMU_GROUPUSED_OBJECT, &os->os_groupused_dnode); | |
9babb374 | 512 | } |
34dc7c2f | 513 | |
428870ff | 514 | *osp = os; |
34dc7c2f BB |
515 | return (0); |
516 | } | |
517 | ||
428870ff BB |
518 | int |
519 | dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp) | |
34dc7c2f | 520 | { |
428870ff | 521 | int err = 0; |
34dc7c2f BB |
522 | |
523 | mutex_enter(&ds->ds_opening_lock); | |
9b67f605 MA |
524 | if (ds->ds_objset == NULL) { |
525 | objset_t *os; | |
34dc7c2f | 526 | err = dmu_objset_open_impl(dsl_dataset_get_spa(ds), |
9b67f605 MA |
527 | ds, dsl_dataset_get_blkptr(ds), &os); |
528 | ||
529 | if (err == 0) { | |
530 | mutex_enter(&ds->ds_lock); | |
531 | ASSERT(ds->ds_objset == NULL); | |
532 | ds->ds_objset = os; | |
533 | mutex_exit(&ds->ds_lock); | |
534 | } | |
34dc7c2f | 535 | } |
9b67f605 | 536 | *osp = ds->ds_objset; |
34dc7c2f | 537 | mutex_exit(&ds->ds_opening_lock); |
428870ff | 538 | return (err); |
34dc7c2f BB |
539 | } |
540 | ||
13fe0198 MA |
541 | /* |
542 | * Holds the pool while the objset is held. Therefore only one objset | |
543 | * can be held at a time. | |
544 | */ | |
34dc7c2f | 545 | int |
428870ff | 546 | dmu_objset_hold(const char *name, void *tag, objset_t **osp) |
34dc7c2f | 547 | { |
13fe0198 | 548 | dsl_pool_t *dp; |
428870ff | 549 | dsl_dataset_t *ds; |
34dc7c2f BB |
550 | int err; |
551 | ||
13fe0198 MA |
552 | err = dsl_pool_hold(name, tag, &dp); |
553 | if (err != 0) | |
554 | return (err); | |
555 | err = dsl_dataset_hold(dp, name, tag, &ds); | |
556 | if (err != 0) { | |
557 | dsl_pool_rele(dp, tag); | |
428870ff | 558 | return (err); |
13fe0198 | 559 | } |
428870ff BB |
560 | |
561 | err = dmu_objset_from_ds(ds, osp); | |
13fe0198 | 562 | if (err != 0) { |
428870ff | 563 | dsl_dataset_rele(ds, tag); |
13fe0198 MA |
564 | dsl_pool_rele(dp, tag); |
565 | } | |
428870ff | 566 | |
34dc7c2f BB |
567 | return (err); |
568 | } | |
569 | ||
9c43027b AJ |
570 | static int |
571 | dmu_objset_own_impl(dsl_dataset_t *ds, dmu_objset_type_t type, | |
572 | boolean_t readonly, void *tag, objset_t **osp) | |
573 | { | |
574 | int err; | |
575 | ||
576 | err = dmu_objset_from_ds(ds, osp); | |
577 | if (err != 0) { | |
578 | dsl_dataset_disown(ds, tag); | |
579 | } else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) { | |
580 | dsl_dataset_disown(ds, tag); | |
581 | return (SET_ERROR(EINVAL)); | |
582 | } else if (!readonly && dsl_dataset_is_snapshot(ds)) { | |
583 | dsl_dataset_disown(ds, tag); | |
584 | return (SET_ERROR(EROFS)); | |
585 | } | |
586 | return (err); | |
587 | } | |
588 | ||
13fe0198 MA |
589 | /* |
590 | * dsl_pool must not be held when this is called. | |
591 | * Upon successful return, there will be a longhold on the dataset, | |
592 | * and the dsl_pool will not be held. | |
593 | */ | |
34dc7c2f | 594 | int |
428870ff BB |
595 | dmu_objset_own(const char *name, dmu_objset_type_t type, |
596 | boolean_t readonly, void *tag, objset_t **osp) | |
34dc7c2f | 597 | { |
13fe0198 | 598 | dsl_pool_t *dp; |
34dc7c2f BB |
599 | dsl_dataset_t *ds; |
600 | int err; | |
601 | ||
13fe0198 MA |
602 | err = dsl_pool_hold(name, FTAG, &dp); |
603 | if (err != 0) | |
604 | return (err); | |
605 | err = dsl_dataset_own(dp, name, tag, &ds); | |
606 | if (err != 0) { | |
607 | dsl_pool_rele(dp, FTAG); | |
34dc7c2f | 608 | return (err); |
13fe0198 | 609 | } |
9c43027b | 610 | err = dmu_objset_own_impl(ds, type, readonly, tag, osp); |
13fe0198 | 611 | dsl_pool_rele(dp, FTAG); |
9c43027b | 612 | |
34dc7c2f BB |
613 | return (err); |
614 | } | |
615 | ||
9c43027b AJ |
616 | int |
617 | dmu_objset_own_obj(dsl_pool_t *dp, uint64_t obj, dmu_objset_type_t type, | |
618 | boolean_t readonly, void *tag, objset_t **osp) | |
619 | { | |
620 | dsl_dataset_t *ds; | |
621 | int err; | |
622 | ||
623 | err = dsl_dataset_own_obj(dp, obj, tag, &ds); | |
624 | if (err != 0) | |
625 | return (err); | |
626 | ||
627 | return (dmu_objset_own_impl(ds, type, readonly, tag, osp)); | |
628 | } | |
629 | ||
34dc7c2f | 630 | void |
428870ff | 631 | dmu_objset_rele(objset_t *os, void *tag) |
34dc7c2f | 632 | { |
13fe0198 | 633 | dsl_pool_t *dp = dmu_objset_pool(os); |
428870ff | 634 | dsl_dataset_rele(os->os_dsl_dataset, tag); |
13fe0198 | 635 | dsl_pool_rele(dp, tag); |
428870ff | 636 | } |
b128c09f | 637 | |
831baf06 KW |
638 | /* |
639 | * When we are called, os MUST refer to an objset associated with a dataset | |
640 | * that is owned by 'tag'; that is, is held and long held by 'tag' and ds_owner | |
641 | * == tag. We will then release and reacquire ownership of the dataset while | |
642 | * holding the pool config_rwlock to avoid intervening namespace or ownership | |
643 | * changes may occur. | |
644 | * | |
645 | * This exists solely to accommodate zfs_ioc_userspace_upgrade()'s desire to | |
646 | * release the hold on its dataset and acquire a new one on the dataset of the | |
647 | * same name so that it can be partially torn down and reconstructed. | |
648 | */ | |
649 | void | |
650 | dmu_objset_refresh_ownership(objset_t *os, void *tag) | |
651 | { | |
652 | dsl_pool_t *dp; | |
653 | dsl_dataset_t *ds, *newds; | |
654 | char name[MAXNAMELEN]; | |
655 | ||
656 | ds = os->os_dsl_dataset; | |
657 | VERIFY3P(ds, !=, NULL); | |
658 | VERIFY3P(ds->ds_owner, ==, tag); | |
659 | VERIFY(dsl_dataset_long_held(ds)); | |
660 | ||
661 | dsl_dataset_name(ds, name); | |
662 | dp = dmu_objset_pool(os); | |
663 | dsl_pool_config_enter(dp, FTAG); | |
664 | dmu_objset_disown(os, tag); | |
665 | VERIFY0(dsl_dataset_own(dp, name, tag, &newds)); | |
666 | VERIFY3P(newds, ==, os->os_dsl_dataset); | |
667 | dsl_pool_config_exit(dp, FTAG); | |
668 | } | |
669 | ||
428870ff BB |
670 | void |
671 | dmu_objset_disown(objset_t *os, void *tag) | |
672 | { | |
673 | dsl_dataset_disown(os->os_dsl_dataset, tag); | |
34dc7c2f BB |
674 | } |
675 | ||
13fe0198 | 676 | void |
34dc7c2f BB |
677 | dmu_objset_evict_dbufs(objset_t *os) |
678 | { | |
0c66c32d | 679 | dnode_t *dn_marker; |
34dc7c2f BB |
680 | dnode_t *dn; |
681 | ||
0c66c32d | 682 | dn_marker = kmem_alloc(sizeof (dnode_t), KM_SLEEP); |
34dc7c2f | 683 | |
0c66c32d JG |
684 | mutex_enter(&os->os_lock); |
685 | dn = list_head(&os->os_dnodes); | |
686 | while (dn != NULL) { | |
687 | /* | |
688 | * Skip dnodes without holds. We have to do this dance | |
689 | * because dnode_add_ref() only works if there is already a | |
690 | * hold. If the dnode has no holds, then it has no dbufs. | |
691 | */ | |
692 | if (dnode_add_ref(dn, FTAG)) { | |
693 | list_insert_after(&os->os_dnodes, dn, dn_marker); | |
694 | mutex_exit(&os->os_lock); | |
34dc7c2f | 695 | |
0c66c32d JG |
696 | dnode_evict_dbufs(dn); |
697 | dnode_rele(dn, FTAG); | |
34dc7c2f | 698 | |
0c66c32d JG |
699 | mutex_enter(&os->os_lock); |
700 | dn = list_next(&os->os_dnodes, dn_marker); | |
701 | list_remove(&os->os_dnodes, dn_marker); | |
702 | } else { | |
703 | dn = list_next(&os->os_dnodes, dn); | |
704 | } | |
705 | } | |
706 | mutex_exit(&os->os_lock); | |
34dc7c2f | 707 | |
0c66c32d | 708 | kmem_free(dn_marker, sizeof (dnode_t)); |
34dc7c2f | 709 | |
0c66c32d JG |
710 | if (DMU_USERUSED_DNODE(os) != NULL) { |
711 | dnode_evict_dbufs(DMU_GROUPUSED_DNODE(os)); | |
712 | dnode_evict_dbufs(DMU_USERUSED_DNODE(os)); | |
34dc7c2f | 713 | } |
0c66c32d | 714 | dnode_evict_dbufs(DMU_META_DNODE(os)); |
34dc7c2f BB |
715 | } |
716 | ||
0c66c32d JG |
717 | /* |
718 | * Objset eviction processing is split into into two pieces. | |
719 | * The first marks the objset as evicting, evicts any dbufs that | |
720 | * have a refcount of zero, and then queues up the objset for the | |
721 | * second phase of eviction. Once os->os_dnodes has been cleared by | |
722 | * dnode_buf_pageout()->dnode_destroy(), the second phase is executed. | |
723 | * The second phase closes the special dnodes, dequeues the objset from | |
724 | * the list of those undergoing eviction, and finally frees the objset. | |
725 | * | |
726 | * NOTE: Due to asynchronous eviction processing (invocation of | |
727 | * dnode_buf_pageout()), it is possible for the meta dnode for the | |
728 | * objset to have no holds even though os->os_dnodes is not empty. | |
729 | */ | |
34dc7c2f | 730 | void |
428870ff | 731 | dmu_objset_evict(objset_t *os) |
34dc7c2f | 732 | { |
d6320ddb | 733 | int t; |
34dc7c2f | 734 | |
6f1ffb06 MA |
735 | dsl_dataset_t *ds = os->os_dsl_dataset; |
736 | ||
d6320ddb | 737 | for (t = 0; t < TXG_SIZE; t++) |
428870ff | 738 | ASSERT(!dmu_objset_is_dirty(os, t)); |
34dc7c2f | 739 | |
0eb21616 JG |
740 | if (ds) |
741 | dsl_prop_unregister_all(ds, os); | |
34dc7c2f | 742 | |
428870ff BB |
743 | if (os->os_sa) |
744 | sa_tear_down(os); | |
745 | ||
13fe0198 | 746 | dmu_objset_evict_dbufs(os); |
34dc7c2f | 747 | |
0c66c32d JG |
748 | mutex_enter(&os->os_lock); |
749 | spa_evicting_os_register(os->os_spa, os); | |
750 | if (list_is_empty(&os->os_dnodes)) { | |
751 | mutex_exit(&os->os_lock); | |
752 | dmu_objset_evict_done(os); | |
753 | } else { | |
754 | mutex_exit(&os->os_lock); | |
755 | } | |
756 | } | |
757 | ||
758 | void | |
759 | dmu_objset_evict_done(objset_t *os) | |
760 | { | |
761 | ASSERT3P(list_head(&os->os_dnodes), ==, NULL); | |
762 | ||
572e2857 BB |
763 | dnode_special_close(&os->os_meta_dnode); |
764 | if (DMU_USERUSED_DNODE(os)) { | |
765 | dnode_special_close(&os->os_userused_dnode); | |
766 | dnode_special_close(&os->os_groupused_dnode); | |
9babb374 | 767 | } |
428870ff BB |
768 | zil_free(os->os_zil); |
769 | ||
13fe0198 | 770 | VERIFY(arc_buf_remove_ref(os->os_phys_buf, &os->os_phys_buf)); |
572e2857 BB |
771 | |
772 | /* | |
773 | * This is a barrier to prevent the objset from going away in | |
774 | * dnode_move() until we can safely ensure that the objset is still in | |
775 | * use. We consider the objset valid before the barrier and invalid | |
776 | * after the barrier. | |
777 | */ | |
778 | rw_enter(&os_lock, RW_READER); | |
779 | rw_exit(&os_lock); | |
780 | ||
428870ff BB |
781 | mutex_destroy(&os->os_lock); |
782 | mutex_destroy(&os->os_obj_lock); | |
783 | mutex_destroy(&os->os_user_ptr_lock); | |
0c66c32d | 784 | spa_evicting_os_deregister(os->os_spa, os); |
428870ff BB |
785 | kmem_free(os, sizeof (objset_t)); |
786 | } | |
9babb374 | 787 | |
428870ff BB |
788 | timestruc_t |
789 | dmu_objset_snap_cmtime(objset_t *os) | |
790 | { | |
791 | return (dsl_dir_snap_cmtime(os->os_dsl_dataset->ds_dir)); | |
34dc7c2f BB |
792 | } |
793 | ||
794 | /* called from dsl for meta-objset */ | |
428870ff | 795 | objset_t * |
34dc7c2f BB |
796 | dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, |
797 | dmu_objset_type_t type, dmu_tx_t *tx) | |
798 | { | |
428870ff | 799 | objset_t *os; |
34dc7c2f BB |
800 | dnode_t *mdn; |
801 | ||
802 | ASSERT(dmu_tx_is_syncing(tx)); | |
13fe0198 | 803 | |
572e2857 | 804 | if (ds != NULL) |
13fe0198 | 805 | VERIFY0(dmu_objset_from_ds(ds, &os)); |
572e2857 | 806 | else |
13fe0198 | 807 | VERIFY0(dmu_objset_open_impl(spa, NULL, bp, &os)); |
572e2857 BB |
808 | |
809 | mdn = DMU_META_DNODE(os); | |
34dc7c2f | 810 | |
50c957f7 NB |
811 | dnode_allocate(mdn, DMU_OT_DNODE, DNODE_BLOCK_SIZE, DN_MAX_INDBLKSHIFT, |
812 | DMU_OT_NONE, 0, DNODE_MIN_SLOTS, tx); | |
34dc7c2f BB |
813 | |
814 | /* | |
815 | * We don't want to have to increase the meta-dnode's nlevels | |
816 | * later, because then we could do it in quescing context while | |
817 | * we are also accessing it in open context. | |
818 | * | |
819 | * This precaution is not necessary for the MOS (ds == NULL), | |
820 | * because the MOS is only updated in syncing context. | |
821 | * This is most fortunate: the MOS is the only objset that | |
822 | * needs to be synced multiple times as spa_sync() iterates | |
823 | * to convergence, so minimizing its dn_nlevels matters. | |
824 | */ | |
825 | if (ds != NULL) { | |
826 | int levels = 1; | |
827 | ||
828 | /* | |
829 | * Determine the number of levels necessary for the meta-dnode | |
830 | * to contain DN_MAX_OBJECT dnodes. | |
831 | */ | |
832 | while ((uint64_t)mdn->dn_nblkptr << (mdn->dn_datablkshift + | |
833 | (levels - 1) * (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) < | |
834 | DN_MAX_OBJECT * sizeof (dnode_phys_t)) | |
835 | levels++; | |
836 | ||
837 | mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] = | |
838 | mdn->dn_nlevels = levels; | |
839 | } | |
840 | ||
841 | ASSERT(type != DMU_OST_NONE); | |
842 | ASSERT(type != DMU_OST_ANY); | |
843 | ASSERT(type < DMU_OST_NUMTYPES); | |
428870ff BB |
844 | os->os_phys->os_type = type; |
845 | if (dmu_objset_userused_enabled(os)) { | |
846 | os->os_phys->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; | |
847 | os->os_flags = os->os_phys->os_flags; | |
9babb374 | 848 | } |
34dc7c2f BB |
849 | |
850 | dsl_dataset_dirty(ds, tx); | |
851 | ||
428870ff | 852 | return (os); |
34dc7c2f BB |
853 | } |
854 | ||
13fe0198 MA |
855 | typedef struct dmu_objset_create_arg { |
856 | const char *doca_name; | |
857 | cred_t *doca_cred; | |
858 | void (*doca_userfunc)(objset_t *os, void *arg, | |
859 | cred_t *cr, dmu_tx_t *tx); | |
860 | void *doca_userarg; | |
861 | dmu_objset_type_t doca_type; | |
862 | uint64_t doca_flags; | |
863 | } dmu_objset_create_arg_t; | |
34dc7c2f BB |
864 | |
865 | /*ARGSUSED*/ | |
866 | static int | |
13fe0198 | 867 | dmu_objset_create_check(void *arg, dmu_tx_t *tx) |
34dc7c2f | 868 | { |
13fe0198 MA |
869 | dmu_objset_create_arg_t *doca = arg; |
870 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
871 | dsl_dir_t *pdd; | |
872 | const char *tail; | |
873 | int error; | |
34dc7c2f | 874 | |
13fe0198 | 875 | if (strchr(doca->doca_name, '@') != NULL) |
2e528b49 | 876 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 877 | |
13fe0198 MA |
878 | error = dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail); |
879 | if (error != 0) | |
880 | return (error); | |
881 | if (tail == NULL) { | |
882 | dsl_dir_rele(pdd, FTAG); | |
2e528b49 | 883 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 884 | } |
788eb90c JJ |
885 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
886 | doca->doca_cred); | |
13fe0198 | 887 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f | 888 | |
788eb90c | 889 | return (error); |
34dc7c2f BB |
890 | } |
891 | ||
892 | static void | |
13fe0198 | 893 | dmu_objset_create_sync(void *arg, dmu_tx_t *tx) |
34dc7c2f | 894 | { |
13fe0198 MA |
895 | dmu_objset_create_arg_t *doca = arg; |
896 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
897 | dsl_dir_t *pdd; | |
898 | const char *tail; | |
6f1ffb06 | 899 | dsl_dataset_t *ds; |
13fe0198 | 900 | uint64_t obj; |
6f1ffb06 | 901 | blkptr_t *bp; |
13fe0198 | 902 | objset_t *os; |
34dc7c2f | 903 | |
13fe0198 | 904 | VERIFY0(dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail)); |
34dc7c2f | 905 | |
13fe0198 MA |
906 | obj = dsl_dataset_create_sync(pdd, tail, NULL, doca->doca_flags, |
907 | doca->doca_cred, tx); | |
34dc7c2f | 908 | |
13fe0198 | 909 | VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); |
6f1ffb06 | 910 | bp = dsl_dataset_get_blkptr(ds); |
13fe0198 MA |
911 | os = dmu_objset_create_impl(pdd->dd_pool->dp_spa, |
912 | ds, bp, doca->doca_type, tx); | |
34dc7c2f | 913 | |
13fe0198 MA |
914 | if (doca->doca_userfunc != NULL) { |
915 | doca->doca_userfunc(os, doca->doca_userarg, | |
916 | doca->doca_cred, tx); | |
34dc7c2f BB |
917 | } |
918 | ||
13fe0198 | 919 | spa_history_log_internal_ds(ds, "create", tx, ""); |
a0bd735a BP |
920 | zvol_create_minors(dp->dp_spa, doca->doca_name, B_TRUE); |
921 | ||
6f1ffb06 | 922 | dsl_dataset_rele(ds, FTAG); |
13fe0198 | 923 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f BB |
924 | } |
925 | ||
926 | int | |
428870ff | 927 | dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, |
34dc7c2f BB |
928 | void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg) |
929 | { | |
13fe0198 | 930 | dmu_objset_create_arg_t doca; |
34dc7c2f | 931 | |
13fe0198 MA |
932 | doca.doca_name = name; |
933 | doca.doca_cred = CRED(); | |
934 | doca.doca_flags = flags; | |
935 | doca.doca_userfunc = func; | |
936 | doca.doca_userarg = arg; | |
937 | doca.doca_type = type; | |
34dc7c2f | 938 | |
13fe0198 | 939 | return (dsl_sync_task(name, |
3d45fdd6 MA |
940 | dmu_objset_create_check, dmu_objset_create_sync, &doca, |
941 | 5, ZFS_SPACE_CHECK_NORMAL)); | |
34dc7c2f BB |
942 | } |
943 | ||
13fe0198 MA |
944 | typedef struct dmu_objset_clone_arg { |
945 | const char *doca_clone; | |
946 | const char *doca_origin; | |
947 | cred_t *doca_cred; | |
948 | } dmu_objset_clone_arg_t; | |
949 | ||
950 | /*ARGSUSED*/ | |
951 | static int | |
952 | dmu_objset_clone_check(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 953 | { |
13fe0198 | 954 | dmu_objset_clone_arg_t *doca = arg; |
428870ff BB |
955 | dsl_dir_t *pdd; |
956 | const char *tail; | |
13fe0198 MA |
957 | int error; |
958 | dsl_dataset_t *origin; | |
959 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
34dc7c2f | 960 | |
13fe0198 | 961 | if (strchr(doca->doca_clone, '@') != NULL) |
2e528b49 | 962 | return (SET_ERROR(EINVAL)); |
13fe0198 MA |
963 | |
964 | error = dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail); | |
965 | if (error != 0) | |
966 | return (error); | |
428870ff | 967 | if (tail == NULL) { |
13fe0198 | 968 | dsl_dir_rele(pdd, FTAG); |
2e528b49 | 969 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 970 | } |
1cddb8c9 | 971 | |
788eb90c JJ |
972 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
973 | doca->doca_cred); | |
974 | if (error != 0) { | |
975 | dsl_dir_rele(pdd, FTAG); | |
976 | return (SET_ERROR(EDQUOT)); | |
977 | } | |
13fe0198 | 978 | dsl_dir_rele(pdd, FTAG); |
428870ff | 979 | |
13fe0198 MA |
980 | error = dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin); |
981 | if (error != 0) | |
572e2857 BB |
982 | return (error); |
983 | ||
13fe0198 | 984 | /* You can only clone snapshots, not the head datasets. */ |
0c66c32d | 985 | if (!origin->ds_is_snapshot) { |
13fe0198 | 986 | dsl_dataset_rele(origin, FTAG); |
2e528b49 | 987 | return (SET_ERROR(EINVAL)); |
428870ff | 988 | } |
13fe0198 | 989 | dsl_dataset_rele(origin, FTAG); |
572e2857 | 990 | |
13fe0198 | 991 | return (0); |
9babb374 | 992 | } |
34dc7c2f | 993 | |
13fe0198 MA |
994 | static void |
995 | dmu_objset_clone_sync(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 996 | { |
13fe0198 MA |
997 | dmu_objset_clone_arg_t *doca = arg; |
998 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
999 | dsl_dir_t *pdd; | |
1000 | const char *tail; | |
1001 | dsl_dataset_t *origin, *ds; | |
1002 | uint64_t obj; | |
1003 | char namebuf[MAXNAMELEN]; | |
6f1ffb06 | 1004 | |
13fe0198 MA |
1005 | VERIFY0(dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail)); |
1006 | VERIFY0(dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin)); | |
6f1ffb06 | 1007 | |
13fe0198 MA |
1008 | obj = dsl_dataset_create_sync(pdd, tail, origin, 0, |
1009 | doca->doca_cred, tx); | |
34dc7c2f | 1010 | |
13fe0198 MA |
1011 | VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); |
1012 | dsl_dataset_name(origin, namebuf); | |
1013 | spa_history_log_internal_ds(ds, "clone", tx, | |
1014 | "origin=%s (%llu)", namebuf, origin->ds_object); | |
a0bd735a | 1015 | zvol_create_minors(dp->dp_spa, doca->doca_clone, B_TRUE); |
13fe0198 MA |
1016 | dsl_dataset_rele(ds, FTAG); |
1017 | dsl_dataset_rele(origin, FTAG); | |
1018 | dsl_dir_rele(pdd, FTAG); | |
34dc7c2f BB |
1019 | } |
1020 | ||
1021 | int | |
13fe0198 | 1022 | dmu_objset_clone(const char *clone, const char *origin) |
34dc7c2f | 1023 | { |
13fe0198 | 1024 | dmu_objset_clone_arg_t doca; |
34dc7c2f | 1025 | |
13fe0198 MA |
1026 | doca.doca_clone = clone; |
1027 | doca.doca_origin = origin; | |
1028 | doca.doca_cred = CRED(); | |
572e2857 | 1029 | |
13fe0198 | 1030 | return (dsl_sync_task(clone, |
3d45fdd6 MA |
1031 | dmu_objset_clone_check, dmu_objset_clone_sync, &doca, |
1032 | 5, ZFS_SPACE_CHECK_NORMAL)); | |
6f1ffb06 MA |
1033 | } |
1034 | ||
1035 | int | |
1036 | dmu_objset_snapshot_one(const char *fsname, const char *snapname) | |
1037 | { | |
1038 | int err; | |
1039 | char *longsnap = kmem_asprintf("%s@%s", fsname, snapname); | |
1040 | nvlist_t *snaps = fnvlist_alloc(); | |
1041 | ||
1042 | fnvlist_add_boolean(snaps, longsnap); | |
6f1ffb06 | 1043 | strfree(longsnap); |
13fe0198 MA |
1044 | err = dsl_dataset_snapshot(snaps, NULL, NULL); |
1045 | fnvlist_free(snaps); | |
6f1ffb06 MA |
1046 | return (err); |
1047 | } | |
1048 | ||
34dc7c2f | 1049 | static void |
9babb374 | 1050 | dmu_objset_sync_dnodes(list_t *list, list_t *newlist, dmu_tx_t *tx) |
34dc7c2f BB |
1051 | { |
1052 | dnode_t *dn; | |
1053 | ||
c65aa5b2 | 1054 | while ((dn = list_head(list))) { |
34dc7c2f BB |
1055 | ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT); |
1056 | ASSERT(dn->dn_dbuf->db_data_pending); | |
1057 | /* | |
9babb374 BB |
1058 | * Initialize dn_zio outside dnode_sync() because the |
1059 | * meta-dnode needs to set it ouside dnode_sync(). | |
34dc7c2f BB |
1060 | */ |
1061 | dn->dn_zio = dn->dn_dbuf->db_data_pending->dr_zio; | |
1062 | ASSERT(dn->dn_zio); | |
1063 | ||
1064 | ASSERT3U(dn->dn_nlevels, <=, DN_MAX_LEVELS); | |
1065 | list_remove(list, dn); | |
9babb374 BB |
1066 | |
1067 | if (newlist) { | |
1068 | (void) dnode_add_ref(dn, newlist); | |
1069 | list_insert_tail(newlist, dn); | |
1070 | } | |
1071 | ||
34dc7c2f BB |
1072 | dnode_sync(dn, tx); |
1073 | } | |
1074 | } | |
1075 | ||
1076 | /* ARGSUSED */ | |
1077 | static void | |
428870ff | 1078 | dmu_objset_write_ready(zio_t *zio, arc_buf_t *abuf, void *arg) |
34dc7c2f | 1079 | { |
d6320ddb BB |
1080 | int i; |
1081 | ||
b128c09f | 1082 | blkptr_t *bp = zio->io_bp; |
428870ff | 1083 | objset_t *os = arg; |
34dc7c2f | 1084 | dnode_phys_t *dnp = &os->os_phys->os_meta_dnode; |
34dc7c2f | 1085 | |
9b67f605 | 1086 | ASSERT(!BP_IS_EMBEDDED(bp)); |
13fe0198 MA |
1087 | ASSERT3P(bp, ==, os->os_rootbp); |
1088 | ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_OBJSET); | |
1089 | ASSERT0(BP_GET_LEVEL(bp)); | |
34dc7c2f BB |
1090 | |
1091 | /* | |
9babb374 BB |
1092 | * Update rootbp fill count: it should be the number of objects |
1093 | * allocated in the object set (not counting the "special" | |
1094 | * objects that are stored in the objset_phys_t -- the meta | |
1095 | * dnode and user/group accounting objects). | |
34dc7c2f | 1096 | */ |
9babb374 | 1097 | bp->blk_fill = 0; |
d6320ddb | 1098 | for (i = 0; i < dnp->dn_nblkptr; i++) |
9b67f605 | 1099 | bp->blk_fill += BP_GET_FILL(&dnp->dn_blkptr[i]); |
428870ff BB |
1100 | } |
1101 | ||
1102 | /* ARGSUSED */ | |
1103 | static void | |
1104 | dmu_objset_write_done(zio_t *zio, arc_buf_t *abuf, void *arg) | |
1105 | { | |
1106 | blkptr_t *bp = zio->io_bp; | |
1107 | blkptr_t *bp_orig = &zio->io_bp_orig; | |
1108 | objset_t *os = arg; | |
34dc7c2f | 1109 | |
b128c09f | 1110 | if (zio->io_flags & ZIO_FLAG_IO_REWRITE) { |
428870ff | 1111 | ASSERT(BP_EQUAL(bp, bp_orig)); |
b128c09f | 1112 | } else { |
428870ff BB |
1113 | dsl_dataset_t *ds = os->os_dsl_dataset; |
1114 | dmu_tx_t *tx = os->os_synctx; | |
1115 | ||
1116 | (void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE); | |
1117 | dsl_dataset_block_born(ds, bp, tx); | |
34dc7c2f BB |
1118 | } |
1119 | } | |
1120 | ||
34dc7c2f BB |
1121 | /* called from dsl */ |
1122 | void | |
428870ff | 1123 | dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx) |
34dc7c2f BB |
1124 | { |
1125 | int txgoff; | |
5dbd68a3 | 1126 | zbookmark_phys_t zb; |
428870ff | 1127 | zio_prop_t zp; |
34dc7c2f BB |
1128 | zio_t *zio; |
1129 | list_t *list; | |
9babb374 | 1130 | list_t *newlist = NULL; |
34dc7c2f BB |
1131 | dbuf_dirty_record_t *dr; |
1132 | ||
1133 | dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg); | |
1134 | ||
1135 | ASSERT(dmu_tx_is_syncing(tx)); | |
1136 | /* XXX the write_done callback should really give us the tx... */ | |
1137 | os->os_synctx = tx; | |
1138 | ||
1139 | if (os->os_dsl_dataset == NULL) { | |
1140 | /* | |
1141 | * This is the MOS. If we have upgraded, | |
1142 | * spa_max_replication() could change, so reset | |
1143 | * os_copies here. | |
1144 | */ | |
1145 | os->os_copies = spa_max_replication(os->os_spa); | |
1146 | } | |
1147 | ||
1148 | /* | |
1149 | * Create the root block IO | |
1150 | */ | |
428870ff BB |
1151 | SET_BOOKMARK(&zb, os->os_dsl_dataset ? |
1152 | os->os_dsl_dataset->ds_object : DMU_META_OBJSET, | |
1153 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
294f6806 | 1154 | arc_release(os->os_phys_buf, &os->os_phys_buf); |
b128c09f | 1155 | |
428870ff | 1156 | dmu_write_policy(os, NULL, 0, 0, &zp); |
9babb374 | 1157 | |
428870ff | 1158 | zio = arc_write(pio, os->os_spa, tx->tx_txg, |
3a17a7a9 | 1159 | os->os_rootbp, os->os_phys_buf, DMU_OS_IS_L2CACHEABLE(os), |
bc77ba73 PD |
1160 | DMU_OS_IS_L2COMPRESSIBLE(os), |
1161 | &zp, dmu_objset_write_ready, NULL, NULL, dmu_objset_write_done, | |
1162 | os, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb); | |
34dc7c2f BB |
1163 | |
1164 | /* | |
9babb374 | 1165 | * Sync special dnodes - the parent IO for the sync is the root block |
34dc7c2f | 1166 | */ |
572e2857 BB |
1167 | DMU_META_DNODE(os)->dn_zio = zio; |
1168 | dnode_sync(DMU_META_DNODE(os), tx); | |
34dc7c2f | 1169 | |
9babb374 BB |
1170 | os->os_phys->os_flags = os->os_flags; |
1171 | ||
572e2857 BB |
1172 | if (DMU_USERUSED_DNODE(os) && |
1173 | DMU_USERUSED_DNODE(os)->dn_type != DMU_OT_NONE) { | |
1174 | DMU_USERUSED_DNODE(os)->dn_zio = zio; | |
1175 | dnode_sync(DMU_USERUSED_DNODE(os), tx); | |
1176 | DMU_GROUPUSED_DNODE(os)->dn_zio = zio; | |
1177 | dnode_sync(DMU_GROUPUSED_DNODE(os), tx); | |
9babb374 BB |
1178 | } |
1179 | ||
34dc7c2f BB |
1180 | txgoff = tx->tx_txg & TXG_MASK; |
1181 | ||
9babb374 BB |
1182 | if (dmu_objset_userused_enabled(os)) { |
1183 | newlist = &os->os_synced_dnodes; | |
1184 | /* | |
1185 | * We must create the list here because it uses the | |
1186 | * dn_dirty_link[] of this txg. | |
1187 | */ | |
1188 | list_create(newlist, sizeof (dnode_t), | |
1189 | offsetof(dnode_t, dn_dirty_link[txgoff])); | |
1190 | } | |
1191 | ||
1192 | dmu_objset_sync_dnodes(&os->os_free_dnodes[txgoff], newlist, tx); | |
1193 | dmu_objset_sync_dnodes(&os->os_dirty_dnodes[txgoff], newlist, tx); | |
34dc7c2f | 1194 | |
572e2857 | 1195 | list = &DMU_META_DNODE(os)->dn_dirty_records[txgoff]; |
6f1ffb06 | 1196 | while ((dr = list_head(list))) { |
13fe0198 | 1197 | ASSERT0(dr->dr_dbuf->db_level); |
34dc7c2f BB |
1198 | list_remove(list, dr); |
1199 | if (dr->dr_zio) | |
1200 | zio_nowait(dr->dr_zio); | |
1201 | } | |
68cbd56e NB |
1202 | |
1203 | /* Enable dnode backfill if enough objects have been freed. */ | |
1204 | if (os->os_freed_dnodes >= dmu_rescan_dnode_threshold) { | |
1205 | os->os_rescan_dnodes = B_TRUE; | |
1206 | os->os_freed_dnodes = 0; | |
1207 | } | |
1208 | ||
34dc7c2f BB |
1209 | /* |
1210 | * Free intent log blocks up to this tx. | |
1211 | */ | |
1212 | zil_sync(os->os_zil, tx); | |
b128c09f | 1213 | os->os_phys->os_zil_header = os->os_zil_header; |
34dc7c2f BB |
1214 | zio_nowait(zio); |
1215 | } | |
1216 | ||
428870ff BB |
1217 | boolean_t |
1218 | dmu_objset_is_dirty(objset_t *os, uint64_t txg) | |
1219 | { | |
1220 | return (!list_is_empty(&os->os_dirty_dnodes[txg & TXG_MASK]) || | |
1221 | !list_is_empty(&os->os_free_dnodes[txg & TXG_MASK])); | |
1222 | } | |
1223 | ||
572e2857 | 1224 | static objset_used_cb_t *used_cbs[DMU_OST_NUMTYPES]; |
9babb374 BB |
1225 | |
1226 | void | |
1227 | dmu_objset_register_type(dmu_objset_type_t ost, objset_used_cb_t *cb) | |
1228 | { | |
1229 | used_cbs[ost] = cb; | |
1230 | } | |
1231 | ||
1232 | boolean_t | |
428870ff | 1233 | dmu_objset_userused_enabled(objset_t *os) |
9babb374 BB |
1234 | { |
1235 | return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE && | |
572e2857 BB |
1236 | used_cbs[os->os_phys->os_type] != NULL && |
1237 | DMU_USERUSED_DNODE(os) != NULL); | |
9babb374 BB |
1238 | } |
1239 | ||
428870ff BB |
1240 | static void |
1241 | do_userquota_update(objset_t *os, uint64_t used, uint64_t flags, | |
1242 | uint64_t user, uint64_t group, boolean_t subtract, dmu_tx_t *tx) | |
1243 | { | |
1244 | if ((flags & DNODE_FLAG_USERUSED_ACCOUNTED)) { | |
50c957f7 | 1245 | int64_t delta = DNODE_MIN_SIZE + used; |
428870ff BB |
1246 | if (subtract) |
1247 | delta = -delta; | |
1248 | VERIFY3U(0, ==, zap_increment_int(os, DMU_USERUSED_OBJECT, | |
1249 | user, delta, tx)); | |
1250 | VERIFY3U(0, ==, zap_increment_int(os, DMU_GROUPUSED_OBJECT, | |
1251 | group, delta, tx)); | |
1252 | } | |
1253 | } | |
1254 | ||
9babb374 | 1255 | void |
428870ff | 1256 | dmu_objset_do_userquota_updates(objset_t *os, dmu_tx_t *tx) |
9babb374 BB |
1257 | { |
1258 | dnode_t *dn; | |
1259 | list_t *list = &os->os_synced_dnodes; | |
9babb374 BB |
1260 | |
1261 | ASSERT(list_head(list) == NULL || dmu_objset_userused_enabled(os)); | |
1262 | ||
6f1ffb06 | 1263 | while ((dn = list_head(list))) { |
572e2857 | 1264 | int flags; |
9babb374 | 1265 | ASSERT(!DMU_OBJECT_IS_SPECIAL(dn->dn_object)); |
9babb374 BB |
1266 | ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE || |
1267 | dn->dn_phys->dn_flags & | |
1268 | DNODE_FLAG_USERUSED_ACCOUNTED); | |
1269 | ||
1270 | /* Allocate the user/groupused objects if necessary. */ | |
572e2857 | 1271 | if (DMU_USERUSED_DNODE(os)->dn_type == DMU_OT_NONE) { |
428870ff | 1272 | VERIFY(0 == zap_create_claim(os, |
9babb374 BB |
1273 | DMU_USERUSED_OBJECT, |
1274 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
428870ff | 1275 | VERIFY(0 == zap_create_claim(os, |
9babb374 BB |
1276 | DMU_GROUPUSED_OBJECT, |
1277 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
1278 | } | |
1279 | ||
1280 | /* | |
428870ff BB |
1281 | * We intentionally modify the zap object even if the |
1282 | * net delta is zero. Otherwise | |
1283 | * the block of the zap obj could be shared between | |
1284 | * datasets but need to be different between them after | |
1285 | * a bprewrite. | |
9babb374 | 1286 | */ |
9babb374 | 1287 | |
572e2857 BB |
1288 | flags = dn->dn_id_flags; |
1289 | ASSERT(flags); | |
1290 | if (flags & DN_ID_OLD_EXIST) { | |
428870ff BB |
1291 | do_userquota_update(os, dn->dn_oldused, dn->dn_oldflags, |
1292 | dn->dn_olduid, dn->dn_oldgid, B_TRUE, tx); | |
1293 | } | |
572e2857 | 1294 | if (flags & DN_ID_NEW_EXIST) { |
428870ff BB |
1295 | do_userquota_update(os, DN_USED_BYTES(dn->dn_phys), |
1296 | dn->dn_phys->dn_flags, dn->dn_newuid, | |
1297 | dn->dn_newgid, B_FALSE, tx); | |
1298 | } | |
1299 | ||
572e2857 | 1300 | mutex_enter(&dn->dn_mtx); |
428870ff BB |
1301 | dn->dn_oldused = 0; |
1302 | dn->dn_oldflags = 0; | |
1303 | if (dn->dn_id_flags & DN_ID_NEW_EXIST) { | |
1304 | dn->dn_olduid = dn->dn_newuid; | |
1305 | dn->dn_oldgid = dn->dn_newgid; | |
1306 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
1307 | if (dn->dn_bonuslen == 0) | |
1308 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1309 | else | |
1310 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1311 | } | |
1312 | dn->dn_id_flags &= ~(DN_ID_NEW_EXIST); | |
9babb374 BB |
1313 | mutex_exit(&dn->dn_mtx); |
1314 | ||
1315 | list_remove(list, dn); | |
1316 | dnode_rele(dn, list); | |
1317 | } | |
1318 | } | |
1319 | ||
428870ff BB |
1320 | /* |
1321 | * Returns a pointer to data to find uid/gid from | |
1322 | * | |
1323 | * If a dirty record for transaction group that is syncing can't | |
1324 | * be found then NULL is returned. In the NULL case it is assumed | |
1325 | * the uid/gid aren't changing. | |
1326 | */ | |
1327 | static void * | |
1328 | dmu_objset_userquota_find_data(dmu_buf_impl_t *db, dmu_tx_t *tx) | |
1329 | { | |
1330 | dbuf_dirty_record_t *dr, **drp; | |
1331 | void *data; | |
1332 | ||
1333 | if (db->db_dirtycnt == 0) | |
1334 | return (db->db.db_data); /* Nothing is changing */ | |
1335 | ||
1336 | for (drp = &db->db_last_dirty; (dr = *drp) != NULL; drp = &dr->dr_next) | |
1337 | if (dr->dr_txg == tx->tx_txg) | |
1338 | break; | |
1339 | ||
572e2857 | 1340 | if (dr == NULL) { |
428870ff | 1341 | data = NULL; |
572e2857 BB |
1342 | } else { |
1343 | dnode_t *dn; | |
1344 | ||
1345 | DB_DNODE_ENTER(dr->dr_dbuf); | |
1346 | dn = DB_DNODE(dr->dr_dbuf); | |
1347 | ||
1348 | if (dn->dn_bonuslen == 0 && | |
1349 | dr->dr_dbuf->db_blkid == DMU_SPILL_BLKID) | |
1350 | data = dr->dt.dl.dr_data->b_data; | |
1351 | else | |
1352 | data = dr->dt.dl.dr_data; | |
1353 | ||
1354 | DB_DNODE_EXIT(dr->dr_dbuf); | |
1355 | } | |
1356 | ||
428870ff BB |
1357 | return (data); |
1358 | } | |
1359 | ||
1360 | void | |
1361 | dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) | |
1362 | { | |
1363 | objset_t *os = dn->dn_objset; | |
1364 | void *data = NULL; | |
1365 | dmu_buf_impl_t *db = NULL; | |
a117a6d6 GW |
1366 | uint64_t *user = NULL; |
1367 | uint64_t *group = NULL; | |
428870ff BB |
1368 | int flags = dn->dn_id_flags; |
1369 | int error; | |
1370 | boolean_t have_spill = B_FALSE; | |
1371 | ||
1372 | if (!dmu_objset_userused_enabled(dn->dn_objset)) | |
1373 | return; | |
1374 | ||
1375 | if (before && (flags & (DN_ID_CHKED_BONUS|DN_ID_OLD_EXIST| | |
1376 | DN_ID_CHKED_SPILL))) | |
1377 | return; | |
1378 | ||
1379 | if (before && dn->dn_bonuslen != 0) | |
1380 | data = DN_BONUS(dn->dn_phys); | |
1381 | else if (!before && dn->dn_bonuslen != 0) { | |
a3000f93 BB |
1382 | if (dn->dn_bonus) { |
1383 | db = dn->dn_bonus; | |
428870ff BB |
1384 | mutex_enter(&db->db_mtx); |
1385 | data = dmu_objset_userquota_find_data(db, tx); | |
1386 | } else { | |
1387 | data = DN_BONUS(dn->dn_phys); | |
1388 | } | |
1389 | } else if (dn->dn_bonuslen == 0 && dn->dn_bonustype == DMU_OT_SA) { | |
1390 | int rf = 0; | |
1391 | ||
1392 | if (RW_WRITE_HELD(&dn->dn_struct_rwlock)) | |
1393 | rf |= DB_RF_HAVESTRUCT; | |
572e2857 BB |
1394 | error = dmu_spill_hold_by_dnode(dn, |
1395 | rf | DB_RF_MUST_SUCCEED, | |
428870ff BB |
1396 | FTAG, (dmu_buf_t **)&db); |
1397 | ASSERT(error == 0); | |
1398 | mutex_enter(&db->db_mtx); | |
1399 | data = (before) ? db->db.db_data : | |
1400 | dmu_objset_userquota_find_data(db, tx); | |
1401 | have_spill = B_TRUE; | |
1402 | } else { | |
1403 | mutex_enter(&dn->dn_mtx); | |
1404 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1405 | mutex_exit(&dn->dn_mtx); | |
1406 | return; | |
1407 | } | |
1408 | ||
1409 | if (before) { | |
1410 | ASSERT(data); | |
1411 | user = &dn->dn_olduid; | |
1412 | group = &dn->dn_oldgid; | |
1413 | } else if (data) { | |
1414 | user = &dn->dn_newuid; | |
1415 | group = &dn->dn_newgid; | |
1416 | } | |
1417 | ||
1418 | /* | |
1419 | * Must always call the callback in case the object | |
1420 | * type has changed and that type isn't an object type to track | |
1421 | */ | |
1422 | error = used_cbs[os->os_phys->os_type](dn->dn_bonustype, data, | |
1423 | user, group); | |
1424 | ||
1425 | /* | |
1426 | * Preserve existing uid/gid when the callback can't determine | |
1427 | * what the new uid/gid are and the callback returned EEXIST. | |
1428 | * The EEXIST error tells us to just use the existing uid/gid. | |
1429 | * If we don't know what the old values are then just assign | |
1430 | * them to 0, since that is a new file being created. | |
1431 | */ | |
1432 | if (!before && data == NULL && error == EEXIST) { | |
1433 | if (flags & DN_ID_OLD_EXIST) { | |
1434 | dn->dn_newuid = dn->dn_olduid; | |
1435 | dn->dn_newgid = dn->dn_oldgid; | |
1436 | } else { | |
1437 | dn->dn_newuid = 0; | |
1438 | dn->dn_newgid = 0; | |
1439 | } | |
1440 | error = 0; | |
1441 | } | |
1442 | ||
1443 | if (db) | |
1444 | mutex_exit(&db->db_mtx); | |
1445 | ||
1446 | mutex_enter(&dn->dn_mtx); | |
1447 | if (error == 0 && before) | |
1448 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
1449 | if (error == 0 && !before) | |
1450 | dn->dn_id_flags |= DN_ID_NEW_EXIST; | |
1451 | ||
1452 | if (have_spill) { | |
1453 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1454 | } else { | |
1455 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1456 | } | |
1457 | mutex_exit(&dn->dn_mtx); | |
a3000f93 | 1458 | if (have_spill) |
428870ff BB |
1459 | dmu_buf_rele((dmu_buf_t *)db, FTAG); |
1460 | } | |
1461 | ||
9babb374 BB |
1462 | boolean_t |
1463 | dmu_objset_userspace_present(objset_t *os) | |
1464 | { | |
428870ff | 1465 | return (os->os_phys->os_flags & |
9babb374 BB |
1466 | OBJSET_FLAG_USERACCOUNTING_COMPLETE); |
1467 | } | |
1468 | ||
1469 | int | |
1470 | dmu_objset_userspace_upgrade(objset_t *os) | |
1471 | { | |
1472 | uint64_t obj; | |
1473 | int err = 0; | |
1474 | ||
1475 | if (dmu_objset_userspace_present(os)) | |
1476 | return (0); | |
428870ff | 1477 | if (!dmu_objset_userused_enabled(os)) |
2e528b49 | 1478 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 1479 | if (dmu_objset_is_snapshot(os)) |
2e528b49 | 1480 | return (SET_ERROR(EINVAL)); |
9babb374 BB |
1481 | |
1482 | /* | |
1483 | * We simply need to mark every object dirty, so that it will be | |
1484 | * synced out and now accounted. If this is called | |
1485 | * concurrently, or if we already did some work before crashing, | |
1486 | * that's fine, since we track each object's accounted state | |
1487 | * independently. | |
1488 | */ | |
1489 | ||
1490 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
45d1cae3 | 1491 | dmu_tx_t *tx; |
9babb374 BB |
1492 | dmu_buf_t *db; |
1493 | int objerr; | |
1494 | ||
1495 | if (issig(JUSTLOOKING) && issig(FORREAL)) | |
2e528b49 | 1496 | return (SET_ERROR(EINTR)); |
9babb374 BB |
1497 | |
1498 | objerr = dmu_bonus_hold(os, obj, FTAG, &db); | |
13fe0198 | 1499 | if (objerr != 0) |
9babb374 | 1500 | continue; |
45d1cae3 | 1501 | tx = dmu_tx_create(os); |
9babb374 BB |
1502 | dmu_tx_hold_bonus(tx, obj); |
1503 | objerr = dmu_tx_assign(tx, TXG_WAIT); | |
13fe0198 | 1504 | if (objerr != 0) { |
9babb374 BB |
1505 | dmu_tx_abort(tx); |
1506 | continue; | |
1507 | } | |
1508 | dmu_buf_will_dirty(db, tx); | |
1509 | dmu_buf_rele(db, FTAG); | |
1510 | dmu_tx_commit(tx); | |
1511 | } | |
1512 | ||
428870ff | 1513 | os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; |
9babb374 BB |
1514 | txg_wait_synced(dmu_objset_pool(os), 0); |
1515 | return (0); | |
1516 | } | |
1517 | ||
34dc7c2f BB |
1518 | void |
1519 | dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, | |
1520 | uint64_t *usedobjsp, uint64_t *availobjsp) | |
1521 | { | |
428870ff | 1522 | dsl_dataset_space(os->os_dsl_dataset, refdbytesp, availbytesp, |
34dc7c2f BB |
1523 | usedobjsp, availobjsp); |
1524 | } | |
1525 | ||
1526 | uint64_t | |
1527 | dmu_objset_fsid_guid(objset_t *os) | |
1528 | { | |
428870ff | 1529 | return (dsl_dataset_fsid_guid(os->os_dsl_dataset)); |
34dc7c2f BB |
1530 | } |
1531 | ||
1532 | void | |
1533 | dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat) | |
1534 | { | |
428870ff BB |
1535 | stat->dds_type = os->os_phys->os_type; |
1536 | if (os->os_dsl_dataset) | |
1537 | dsl_dataset_fast_stat(os->os_dsl_dataset, stat); | |
34dc7c2f BB |
1538 | } |
1539 | ||
1540 | void | |
1541 | dmu_objset_stats(objset_t *os, nvlist_t *nv) | |
1542 | { | |
428870ff BB |
1543 | ASSERT(os->os_dsl_dataset || |
1544 | os->os_phys->os_type == DMU_OST_META); | |
34dc7c2f | 1545 | |
428870ff BB |
1546 | if (os->os_dsl_dataset != NULL) |
1547 | dsl_dataset_stats(os->os_dsl_dataset, nv); | |
34dc7c2f BB |
1548 | |
1549 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_TYPE, | |
428870ff | 1550 | os->os_phys->os_type); |
9babb374 BB |
1551 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERACCOUNTING, |
1552 | dmu_objset_userspace_present(os)); | |
34dc7c2f BB |
1553 | } |
1554 | ||
1555 | int | |
1556 | dmu_objset_is_snapshot(objset_t *os) | |
1557 | { | |
428870ff | 1558 | if (os->os_dsl_dataset != NULL) |
0c66c32d | 1559 | return (os->os_dsl_dataset->ds_is_snapshot); |
34dc7c2f BB |
1560 | else |
1561 | return (B_FALSE); | |
1562 | } | |
1563 | ||
1564 | int | |
1565 | dmu_snapshot_realname(objset_t *os, char *name, char *real, int maxlen, | |
1566 | boolean_t *conflict) | |
1567 | { | |
428870ff | 1568 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
1569 | uint64_t ignored; |
1570 | ||
d683ddbb | 1571 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 1572 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1573 | |
1574 | return (zap_lookup_norm(ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb JG |
1575 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, name, 8, 1, &ignored, |
1576 | MT_FIRST, real, maxlen, conflict)); | |
34dc7c2f BB |
1577 | } |
1578 | ||
1579 | int | |
1580 | dmu_snapshot_list_next(objset_t *os, int namelen, char *name, | |
1581 | uint64_t *idp, uint64_t *offp, boolean_t *case_conflict) | |
1582 | { | |
428870ff | 1583 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
1584 | zap_cursor_t cursor; |
1585 | zap_attribute_t attr; | |
1586 | ||
13fe0198 MA |
1587 | ASSERT(dsl_pool_config_held(dmu_objset_pool(os))); |
1588 | ||
d683ddbb | 1589 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 1590 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1591 | |
1592 | zap_cursor_init_serialized(&cursor, | |
1593 | ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb | 1594 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, *offp); |
34dc7c2f BB |
1595 | |
1596 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
1597 | zap_cursor_fini(&cursor); | |
2e528b49 | 1598 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1599 | } |
1600 | ||
1601 | if (strlen(attr.za_name) + 1 > namelen) { | |
1602 | zap_cursor_fini(&cursor); | |
2e528b49 | 1603 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
1604 | } |
1605 | ||
1606 | (void) strcpy(name, attr.za_name); | |
1607 | if (idp) | |
1608 | *idp = attr.za_first_integer; | |
1609 | if (case_conflict) | |
1610 | *case_conflict = attr.za_normalization_conflict; | |
1611 | zap_cursor_advance(&cursor); | |
1612 | *offp = zap_cursor_serialize(&cursor); | |
1613 | zap_cursor_fini(&cursor); | |
1614 | ||
1615 | return (0); | |
1616 | } | |
1617 | ||
ebe7e575 | 1618 | int |
6772fb67 | 1619 | dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *value) |
ebe7e575 | 1620 | { |
d1d7e268 | 1621 | return (dsl_dataset_snap_lookup(os->os_dsl_dataset, name, value)); |
ebe7e575 BB |
1622 | } |
1623 | ||
34dc7c2f BB |
1624 | int |
1625 | dmu_dir_list_next(objset_t *os, int namelen, char *name, | |
1626 | uint64_t *idp, uint64_t *offp) | |
1627 | { | |
428870ff | 1628 | dsl_dir_t *dd = os->os_dsl_dataset->ds_dir; |
34dc7c2f BB |
1629 | zap_cursor_t cursor; |
1630 | zap_attribute_t attr; | |
1631 | ||
1632 | /* there is no next dir on a snapshot! */ | |
428870ff | 1633 | if (os->os_dsl_dataset->ds_object != |
d683ddbb | 1634 | dsl_dir_phys(dd)->dd_head_dataset_obj) |
2e528b49 | 1635 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1636 | |
1637 | zap_cursor_init_serialized(&cursor, | |
1638 | dd->dd_pool->dp_meta_objset, | |
d683ddbb | 1639 | dsl_dir_phys(dd)->dd_child_dir_zapobj, *offp); |
34dc7c2f BB |
1640 | |
1641 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
1642 | zap_cursor_fini(&cursor); | |
2e528b49 | 1643 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1644 | } |
1645 | ||
1646 | if (strlen(attr.za_name) + 1 > namelen) { | |
1647 | zap_cursor_fini(&cursor); | |
2e528b49 | 1648 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
1649 | } |
1650 | ||
1651 | (void) strcpy(name, attr.za_name); | |
1652 | if (idp) | |
1653 | *idp = attr.za_first_integer; | |
1654 | zap_cursor_advance(&cursor); | |
1655 | *offp = zap_cursor_serialize(&cursor); | |
1656 | zap_cursor_fini(&cursor); | |
1657 | ||
1658 | return (0); | |
1659 | } | |
1660 | ||
9c43027b AJ |
1661 | typedef struct dmu_objset_find_ctx { |
1662 | taskq_t *dc_tq; | |
1663 | dsl_pool_t *dc_dp; | |
1664 | uint64_t dc_ddobj; | |
1665 | int (*dc_func)(dsl_pool_t *, dsl_dataset_t *, void *); | |
1666 | void *dc_arg; | |
1667 | int dc_flags; | |
1668 | kmutex_t *dc_error_lock; | |
1669 | int *dc_error; | |
1670 | } dmu_objset_find_ctx_t; | |
1671 | ||
1672 | static void | |
1673 | dmu_objset_find_dp_impl(dmu_objset_find_ctx_t *dcp) | |
b128c09f | 1674 | { |
9c43027b AJ |
1675 | dsl_pool_t *dp = dcp->dc_dp; |
1676 | dmu_objset_find_ctx_t *child_dcp; | |
13fe0198 MA |
1677 | dsl_dir_t *dd; |
1678 | dsl_dataset_t *ds; | |
1679 | zap_cursor_t zc; | |
1680 | zap_attribute_t *attr; | |
1681 | uint64_t thisobj; | |
9c43027b | 1682 | int err = 0; |
13fe0198 | 1683 | |
9c43027b AJ |
1684 | /* don't process if there already was an error */ |
1685 | if (*dcp->dc_error != 0) | |
1686 | goto out; | |
13fe0198 | 1687 | |
9c43027b | 1688 | err = dsl_dir_hold_obj(dp, dcp->dc_ddobj, NULL, FTAG, &dd); |
13fe0198 | 1689 | if (err != 0) |
9c43027b | 1690 | goto out; |
13fe0198 MA |
1691 | |
1692 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ | |
1693 | if (dd->dd_myname[0] == '$') { | |
1694 | dsl_dir_rele(dd, FTAG); | |
9c43027b | 1695 | goto out; |
13fe0198 MA |
1696 | } |
1697 | ||
d683ddbb | 1698 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 1699 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
13fe0198 MA |
1700 | |
1701 | /* | |
1702 | * Iterate over all children. | |
1703 | */ | |
9c43027b | 1704 | if (dcp->dc_flags & DS_FIND_CHILDREN) { |
13fe0198 | 1705 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
d683ddbb | 1706 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
13fe0198 MA |
1707 | zap_cursor_retrieve(&zc, attr) == 0; |
1708 | (void) zap_cursor_advance(&zc)) { | |
1709 | ASSERT3U(attr->za_integer_length, ==, | |
1710 | sizeof (uint64_t)); | |
1711 | ASSERT3U(attr->za_num_integers, ==, 1); | |
1712 | ||
9c43027b AJ |
1713 | child_dcp = kmem_alloc(sizeof (*child_dcp), KM_SLEEP); |
1714 | *child_dcp = *dcp; | |
1715 | child_dcp->dc_ddobj = attr->za_first_integer; | |
1716 | if (dcp->dc_tq != NULL) | |
1717 | (void) taskq_dispatch(dcp->dc_tq, | |
1718 | dmu_objset_find_dp_cb, child_dcp, TQ_SLEEP); | |
1719 | else | |
1720 | dmu_objset_find_dp_impl(child_dcp); | |
13fe0198 MA |
1721 | } |
1722 | zap_cursor_fini(&zc); | |
13fe0198 MA |
1723 | } |
1724 | ||
1725 | /* | |
1726 | * Iterate over all snapshots. | |
1727 | */ | |
9c43027b | 1728 | if (dcp->dc_flags & DS_FIND_SNAPSHOTS) { |
13fe0198 MA |
1729 | dsl_dataset_t *ds; |
1730 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1731 | ||
1732 | if (err == 0) { | |
d683ddbb JG |
1733 | uint64_t snapobj; |
1734 | ||
1735 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
13fe0198 MA |
1736 | dsl_dataset_rele(ds, FTAG); |
1737 | ||
1738 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
1739 | zap_cursor_retrieve(&zc, attr) == 0; | |
1740 | (void) zap_cursor_advance(&zc)) { | |
1741 | ASSERT3U(attr->za_integer_length, ==, | |
1742 | sizeof (uint64_t)); | |
1743 | ASSERT3U(attr->za_num_integers, ==, 1); | |
1744 | ||
1745 | err = dsl_dataset_hold_obj(dp, | |
1746 | attr->za_first_integer, FTAG, &ds); | |
1747 | if (err != 0) | |
1748 | break; | |
9c43027b | 1749 | err = dcp->dc_func(dp, ds, dcp->dc_arg); |
13fe0198 MA |
1750 | dsl_dataset_rele(ds, FTAG); |
1751 | if (err != 0) | |
1752 | break; | |
1753 | } | |
1754 | zap_cursor_fini(&zc); | |
1755 | } | |
1756 | } | |
1757 | ||
1758 | dsl_dir_rele(dd, FTAG); | |
1759 | kmem_free(attr, sizeof (zap_attribute_t)); | |
1760 | ||
1761 | if (err != 0) | |
9c43027b | 1762 | goto out; |
13fe0198 MA |
1763 | |
1764 | /* | |
1765 | * Apply to self. | |
1766 | */ | |
1767 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1768 | if (err != 0) | |
9c43027b AJ |
1769 | goto out; |
1770 | err = dcp->dc_func(dp, ds, dcp->dc_arg); | |
13fe0198 | 1771 | dsl_dataset_rele(ds, FTAG); |
9c43027b AJ |
1772 | |
1773 | out: | |
1774 | if (err != 0) { | |
1775 | mutex_enter(dcp->dc_error_lock); | |
1776 | /* only keep first error */ | |
1777 | if (*dcp->dc_error == 0) | |
1778 | *dcp->dc_error = err; | |
1779 | mutex_exit(dcp->dc_error_lock); | |
1780 | } | |
1781 | ||
1782 | kmem_free(dcp, sizeof (*dcp)); | |
1783 | } | |
1784 | ||
1785 | static void | |
1786 | dmu_objset_find_dp_cb(void *arg) | |
1787 | { | |
1788 | dmu_objset_find_ctx_t *dcp = arg; | |
1789 | dsl_pool_t *dp = dcp->dc_dp; | |
1790 | ||
5e8cd5d1 AJ |
1791 | /* |
1792 | * We need to get a pool_config_lock here, as there are several | |
1793 | * asssert(pool_config_held) down the stack. Getting a lock via | |
1794 | * dsl_pool_config_enter is risky, as it might be stalled by a | |
1795 | * pending writer. This would deadlock, as the write lock can | |
1796 | * only be granted when our parent thread gives up the lock. | |
1797 | * The _prio interface gives us priority over a pending writer. | |
1798 | */ | |
1799 | dsl_pool_config_enter_prio(dp, FTAG); | |
9c43027b AJ |
1800 | |
1801 | dmu_objset_find_dp_impl(dcp); | |
1802 | ||
1803 | dsl_pool_config_exit(dp, FTAG); | |
1804 | } | |
1805 | ||
1806 | /* | |
1807 | * Find objsets under and including ddobj, call func(ds) on each. | |
1808 | * The order for the enumeration is completely undefined. | |
1809 | * func is called with dsl_pool_config held. | |
1810 | */ | |
1811 | int | |
1812 | dmu_objset_find_dp(dsl_pool_t *dp, uint64_t ddobj, | |
1813 | int func(dsl_pool_t *, dsl_dataset_t *, void *), void *arg, int flags) | |
1814 | { | |
1815 | int error = 0; | |
1816 | taskq_t *tq = NULL; | |
1817 | int ntasks; | |
1818 | dmu_objset_find_ctx_t *dcp; | |
1819 | kmutex_t err_lock; | |
1820 | ||
1821 | mutex_init(&err_lock, NULL, MUTEX_DEFAULT, NULL); | |
1822 | dcp = kmem_alloc(sizeof (*dcp), KM_SLEEP); | |
1823 | dcp->dc_tq = NULL; | |
1824 | dcp->dc_dp = dp; | |
1825 | dcp->dc_ddobj = ddobj; | |
1826 | dcp->dc_func = func; | |
1827 | dcp->dc_arg = arg; | |
1828 | dcp->dc_flags = flags; | |
1829 | dcp->dc_error_lock = &err_lock; | |
1830 | dcp->dc_error = &error; | |
1831 | ||
1832 | if ((flags & DS_FIND_SERIALIZE) || dsl_pool_config_held_writer(dp)) { | |
1833 | /* | |
1834 | * In case a write lock is held we can't make use of | |
1835 | * parallelism, as down the stack of the worker threads | |
1836 | * the lock is asserted via dsl_pool_config_held. | |
1837 | * In case of a read lock this is solved by getting a read | |
1838 | * lock in each worker thread, which isn't possible in case | |
1839 | * of a writer lock. So we fall back to the synchronous path | |
1840 | * here. | |
1841 | * In the future it might be possible to get some magic into | |
1842 | * dsl_pool_config_held in a way that it returns true for | |
1843 | * the worker threads so that a single lock held from this | |
1844 | * thread suffices. For now, stay single threaded. | |
1845 | */ | |
1846 | dmu_objset_find_dp_impl(dcp); | |
e5676636 | 1847 | mutex_destroy(&err_lock); |
9c43027b AJ |
1848 | |
1849 | return (error); | |
1850 | } | |
1851 | ||
1852 | ntasks = dmu_find_threads; | |
1853 | if (ntasks == 0) | |
1854 | ntasks = vdev_count_leaves(dp->dp_spa) * 4; | |
1229323d | 1855 | tq = taskq_create("dmu_objset_find", ntasks, maxclsyspri, ntasks, |
9c43027b AJ |
1856 | INT_MAX, 0); |
1857 | if (tq == NULL) { | |
1858 | kmem_free(dcp, sizeof (*dcp)); | |
e5676636 BB |
1859 | mutex_destroy(&err_lock); |
1860 | ||
9c43027b AJ |
1861 | return (SET_ERROR(ENOMEM)); |
1862 | } | |
1863 | dcp->dc_tq = tq; | |
1864 | ||
1865 | /* dcp will be freed by task */ | |
1866 | (void) taskq_dispatch(tq, dmu_objset_find_dp_cb, dcp, TQ_SLEEP); | |
1867 | ||
1868 | /* | |
1869 | * PORTING: this code relies on the property of taskq_wait to wait | |
1870 | * until no more tasks are queued and no more tasks are active. As | |
1871 | * we always queue new tasks from within other tasks, task_wait | |
1872 | * reliably waits for the full recursion to finish, even though we | |
1873 | * enqueue new tasks after taskq_wait has been called. | |
1874 | * On platforms other than illumos, taskq_wait may not have this | |
1875 | * property. | |
1876 | */ | |
1877 | taskq_wait(tq); | |
1878 | taskq_destroy(tq); | |
1879 | mutex_destroy(&err_lock); | |
1880 | ||
1881 | return (error); | |
b128c09f BB |
1882 | } |
1883 | ||
1884 | /* | |
13fe0198 MA |
1885 | * Find all objsets under name, and for each, call 'func(child_name, arg)'. |
1886 | * The dp_config_rwlock must not be held when this is called, and it | |
1887 | * will not be held when the callback is called. | |
1888 | * Therefore this function should only be used when the pool is not changing | |
1889 | * (e.g. in syncing context), or the callback can deal with the possible races. | |
b128c09f | 1890 | */ |
13fe0198 MA |
1891 | static int |
1892 | dmu_objset_find_impl(spa_t *spa, const char *name, | |
1893 | int func(const char *, void *), void *arg, int flags) | |
34dc7c2f BB |
1894 | { |
1895 | dsl_dir_t *dd; | |
13fe0198 | 1896 | dsl_pool_t *dp = spa_get_dsl(spa); |
b128c09f | 1897 | dsl_dataset_t *ds; |
34dc7c2f BB |
1898 | zap_cursor_t zc; |
1899 | zap_attribute_t *attr; | |
1900 | char *child; | |
b128c09f BB |
1901 | uint64_t thisobj; |
1902 | int err; | |
34dc7c2f | 1903 | |
13fe0198 MA |
1904 | dsl_pool_config_enter(dp, FTAG); |
1905 | ||
1906 | err = dsl_dir_hold(dp, name, FTAG, &dd, NULL); | |
1907 | if (err != 0) { | |
1908 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f | 1909 | return (err); |
13fe0198 | 1910 | } |
34dc7c2f | 1911 | |
b128c09f BB |
1912 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ |
1913 | if (dd->dd_myname[0] == '$') { | |
13fe0198 MA |
1914 | dsl_dir_rele(dd, FTAG); |
1915 | dsl_pool_config_exit(dp, FTAG); | |
b128c09f BB |
1916 | return (0); |
1917 | } | |
1918 | ||
d683ddbb | 1919 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 1920 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
34dc7c2f BB |
1921 | |
1922 | /* | |
1923 | * Iterate over all children. | |
1924 | */ | |
1925 | if (flags & DS_FIND_CHILDREN) { | |
b128c09f | 1926 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
d683ddbb | 1927 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
34dc7c2f BB |
1928 | zap_cursor_retrieve(&zc, attr) == 0; |
1929 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 MA |
1930 | ASSERT3U(attr->za_integer_length, ==, |
1931 | sizeof (uint64_t)); | |
1932 | ASSERT3U(attr->za_num_integers, ==, 1); | |
34dc7c2f | 1933 | |
428870ff | 1934 | child = kmem_asprintf("%s/%s", name, attr->za_name); |
13fe0198 MA |
1935 | dsl_pool_config_exit(dp, FTAG); |
1936 | err = dmu_objset_find_impl(spa, child, | |
1937 | func, arg, flags); | |
1938 | dsl_pool_config_enter(dp, FTAG); | |
428870ff | 1939 | strfree(child); |
13fe0198 | 1940 | if (err != 0) |
34dc7c2f BB |
1941 | break; |
1942 | } | |
1943 | zap_cursor_fini(&zc); | |
1944 | ||
13fe0198 MA |
1945 | if (err != 0) { |
1946 | dsl_dir_rele(dd, FTAG); | |
1947 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f BB |
1948 | kmem_free(attr, sizeof (zap_attribute_t)); |
1949 | return (err); | |
1950 | } | |
1951 | } | |
1952 | ||
1953 | /* | |
1954 | * Iterate over all snapshots. | |
1955 | */ | |
b128c09f | 1956 | if (flags & DS_FIND_SNAPSHOTS) { |
b128c09f | 1957 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); |
b128c09f BB |
1958 | |
1959 | if (err == 0) { | |
d683ddbb JG |
1960 | uint64_t snapobj; |
1961 | ||
1962 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
b128c09f BB |
1963 | dsl_dataset_rele(ds, FTAG); |
1964 | ||
1965 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
1966 | zap_cursor_retrieve(&zc, attr) == 0; | |
1967 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 | 1968 | ASSERT3U(attr->za_integer_length, ==, |
b128c09f | 1969 | sizeof (uint64_t)); |
13fe0198 | 1970 | ASSERT3U(attr->za_num_integers, ==, 1); |
b128c09f | 1971 | |
428870ff BB |
1972 | child = kmem_asprintf("%s@%s", |
1973 | name, attr->za_name); | |
13fe0198 MA |
1974 | dsl_pool_config_exit(dp, FTAG); |
1975 | err = func(child, arg); | |
1976 | dsl_pool_config_enter(dp, FTAG); | |
428870ff | 1977 | strfree(child); |
13fe0198 | 1978 | if (err != 0) |
b128c09f BB |
1979 | break; |
1980 | } | |
1981 | zap_cursor_fini(&zc); | |
34dc7c2f | 1982 | } |
34dc7c2f BB |
1983 | } |
1984 | ||
13fe0198 | 1985 | dsl_dir_rele(dd, FTAG); |
34dc7c2f | 1986 | kmem_free(attr, sizeof (zap_attribute_t)); |
13fe0198 | 1987 | dsl_pool_config_exit(dp, FTAG); |
34dc7c2f | 1988 | |
13fe0198 | 1989 | if (err != 0) |
34dc7c2f BB |
1990 | return (err); |
1991 | ||
13fe0198 MA |
1992 | /* Apply to self. */ |
1993 | return (func(name, arg)); | |
34dc7c2f BB |
1994 | } |
1995 | ||
13fe0198 MA |
1996 | /* |
1997 | * See comment above dmu_objset_find_impl(). | |
1998 | */ | |
d164b209 | 1999 | int |
13fe0198 MA |
2000 | dmu_objset_find(char *name, int func(const char *, void *), void *arg, |
2001 | int flags) | |
d164b209 | 2002 | { |
13fe0198 MA |
2003 | spa_t *spa; |
2004 | int error; | |
d164b209 | 2005 | |
13fe0198 MA |
2006 | error = spa_open(name, &spa, FTAG); |
2007 | if (error != 0) | |
2008 | return (error); | |
2009 | error = dmu_objset_find_impl(spa, name, func, arg, flags); | |
2010 | spa_close(spa, FTAG); | |
2011 | return (error); | |
d164b209 BB |
2012 | } |
2013 | ||
34dc7c2f BB |
2014 | void |
2015 | dmu_objset_set_user(objset_t *os, void *user_ptr) | |
2016 | { | |
428870ff BB |
2017 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
2018 | os->os_user_ptr = user_ptr; | |
34dc7c2f BB |
2019 | } |
2020 | ||
2021 | void * | |
2022 | dmu_objset_get_user(objset_t *os) | |
2023 | { | |
428870ff BB |
2024 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
2025 | return (os->os_user_ptr); | |
34dc7c2f | 2026 | } |
c28b2279 | 2027 | |
13fe0198 MA |
2028 | /* |
2029 | * Determine name of filesystem, given name of snapshot. | |
2030 | * buf must be at least MAXNAMELEN bytes | |
2031 | */ | |
2032 | int | |
2033 | dmu_fsname(const char *snapname, char *buf) | |
2034 | { | |
2035 | char *atp = strchr(snapname, '@'); | |
2036 | if (atp == NULL) | |
2e528b49 | 2037 | return (SET_ERROR(EINVAL)); |
13fe0198 | 2038 | if (atp - snapname >= MAXNAMELEN) |
2e528b49 | 2039 | return (SET_ERROR(ENAMETOOLONG)); |
13fe0198 MA |
2040 | (void) strlcpy(buf, snapname, atp - snapname + 1); |
2041 | return (0); | |
2042 | } | |
2043 | ||
c28b2279 | 2044 | #if defined(_KERNEL) && defined(HAVE_SPL) |
f0fd83be | 2045 | EXPORT_SYMBOL(dmu_objset_zil); |
c28b2279 | 2046 | EXPORT_SYMBOL(dmu_objset_pool); |
f0fd83be BB |
2047 | EXPORT_SYMBOL(dmu_objset_ds); |
2048 | EXPORT_SYMBOL(dmu_objset_type); | |
c28b2279 BB |
2049 | EXPORT_SYMBOL(dmu_objset_name); |
2050 | EXPORT_SYMBOL(dmu_objset_hold); | |
2051 | EXPORT_SYMBOL(dmu_objset_own); | |
2052 | EXPORT_SYMBOL(dmu_objset_rele); | |
2053 | EXPORT_SYMBOL(dmu_objset_disown); | |
2054 | EXPORT_SYMBOL(dmu_objset_from_ds); | |
2055 | EXPORT_SYMBOL(dmu_objset_create); | |
2056 | EXPORT_SYMBOL(dmu_objset_clone); | |
c28b2279 BB |
2057 | EXPORT_SYMBOL(dmu_objset_stats); |
2058 | EXPORT_SYMBOL(dmu_objset_fast_stat); | |
54a179e7 | 2059 | EXPORT_SYMBOL(dmu_objset_spa); |
c28b2279 BB |
2060 | EXPORT_SYMBOL(dmu_objset_space); |
2061 | EXPORT_SYMBOL(dmu_objset_fsid_guid); | |
2062 | EXPORT_SYMBOL(dmu_objset_find); | |
c28b2279 BB |
2063 | EXPORT_SYMBOL(dmu_objset_byteswap); |
2064 | EXPORT_SYMBOL(dmu_objset_evict_dbufs); | |
2065 | EXPORT_SYMBOL(dmu_objset_snap_cmtime); | |
50c957f7 | 2066 | EXPORT_SYMBOL(dmu_objset_dnodesize); |
c28b2279 BB |
2067 | |
2068 | EXPORT_SYMBOL(dmu_objset_sync); | |
2069 | EXPORT_SYMBOL(dmu_objset_is_dirty); | |
2070 | EXPORT_SYMBOL(dmu_objset_create_impl); | |
2071 | EXPORT_SYMBOL(dmu_objset_open_impl); | |
2072 | EXPORT_SYMBOL(dmu_objset_evict); | |
2073 | EXPORT_SYMBOL(dmu_objset_register_type); | |
2074 | EXPORT_SYMBOL(dmu_objset_do_userquota_updates); | |
2075 | EXPORT_SYMBOL(dmu_objset_userquota_get_ids); | |
2076 | EXPORT_SYMBOL(dmu_objset_userused_enabled); | |
2077 | EXPORT_SYMBOL(dmu_objset_userspace_upgrade); | |
2078 | EXPORT_SYMBOL(dmu_objset_userspace_present); | |
2079 | #endif |