]>
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. |
faf0f58c | 23 | * Copyright (c) 2012, 2014 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. |
34dc7c2f BB |
29 | */ |
30 | ||
428870ff BB |
31 | /* Portions Copyright 2010 Robert Milkowski */ |
32 | ||
34dc7c2f BB |
33 | #include <sys/cred.h> |
34 | #include <sys/zfs_context.h> | |
35 | #include <sys/dmu_objset.h> | |
36 | #include <sys/dsl_dir.h> | |
37 | #include <sys/dsl_dataset.h> | |
38 | #include <sys/dsl_prop.h> | |
39 | #include <sys/dsl_pool.h> | |
40 | #include <sys/dsl_synctask.h> | |
41 | #include <sys/dsl_deleg.h> | |
42 | #include <sys/dnode.h> | |
43 | #include <sys/dbuf.h> | |
44 | #include <sys/zvol.h> | |
45 | #include <sys/dmu_tx.h> | |
34dc7c2f BB |
46 | #include <sys/zap.h> |
47 | #include <sys/zil.h> | |
48 | #include <sys/dmu_impl.h> | |
49 | #include <sys/zfs_ioctl.h> | |
428870ff | 50 | #include <sys/sa.h> |
572e2857 | 51 | #include <sys/zfs_onexit.h> |
13fe0198 | 52 | #include <sys/dsl_destroy.h> |
9c43027b | 53 | #include <sys/vdev.h> |
572e2857 BB |
54 | |
55 | /* | |
56 | * Needed to close a window in dnode_move() that allows the objset to be freed | |
57 | * before it can be safely accessed. | |
58 | */ | |
59 | krwlock_t os_lock; | |
60 | ||
9c43027b AJ |
61 | /* |
62 | * Tunable to overwrite the maximum number of threads for the parallization | |
63 | * of dmu_objset_find_dp, needed to speed up the import of pools with many | |
64 | * datasets. | |
65 | * Default is 4 times the number of leaf vdevs. | |
66 | */ | |
67 | int dmu_find_threads = 0; | |
68 | ||
69 | static void dmu_objset_find_dp_cb(void *arg); | |
70 | ||
572e2857 BB |
71 | void |
72 | dmu_objset_init(void) | |
73 | { | |
74 | rw_init(&os_lock, NULL, RW_DEFAULT, NULL); | |
75 | } | |
76 | ||
77 | void | |
78 | dmu_objset_fini(void) | |
79 | { | |
80 | rw_destroy(&os_lock); | |
81 | } | |
34dc7c2f BB |
82 | |
83 | spa_t * | |
84 | dmu_objset_spa(objset_t *os) | |
85 | { | |
428870ff | 86 | return (os->os_spa); |
34dc7c2f BB |
87 | } |
88 | ||
89 | zilog_t * | |
90 | dmu_objset_zil(objset_t *os) | |
91 | { | |
428870ff | 92 | return (os->os_zil); |
34dc7c2f BB |
93 | } |
94 | ||
95 | dsl_pool_t * | |
96 | dmu_objset_pool(objset_t *os) | |
97 | { | |
98 | dsl_dataset_t *ds; | |
99 | ||
428870ff | 100 | if ((ds = os->os_dsl_dataset) != NULL && ds->ds_dir) |
34dc7c2f BB |
101 | return (ds->ds_dir->dd_pool); |
102 | else | |
428870ff | 103 | return (spa_get_dsl(os->os_spa)); |
34dc7c2f BB |
104 | } |
105 | ||
106 | dsl_dataset_t * | |
107 | dmu_objset_ds(objset_t *os) | |
108 | { | |
428870ff | 109 | return (os->os_dsl_dataset); |
34dc7c2f BB |
110 | } |
111 | ||
112 | dmu_objset_type_t | |
113 | dmu_objset_type(objset_t *os) | |
114 | { | |
428870ff | 115 | return (os->os_phys->os_type); |
34dc7c2f BB |
116 | } |
117 | ||
118 | void | |
119 | dmu_objset_name(objset_t *os, char *buf) | |
120 | { | |
428870ff | 121 | dsl_dataset_name(os->os_dsl_dataset, buf); |
34dc7c2f BB |
122 | } |
123 | ||
124 | uint64_t | |
125 | dmu_objset_id(objset_t *os) | |
126 | { | |
428870ff | 127 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
128 | |
129 | return (ds ? ds->ds_object : 0); | |
130 | } | |
131 | ||
faf0f58c | 132 | zfs_sync_type_t |
428870ff BB |
133 | dmu_objset_syncprop(objset_t *os) |
134 | { | |
135 | return (os->os_sync); | |
136 | } | |
137 | ||
faf0f58c | 138 | zfs_logbias_op_t |
428870ff BB |
139 | dmu_objset_logbias(objset_t *os) |
140 | { | |
141 | return (os->os_logbias); | |
142 | } | |
143 | ||
34dc7c2f BB |
144 | static void |
145 | checksum_changed_cb(void *arg, uint64_t newval) | |
146 | { | |
428870ff | 147 | objset_t *os = arg; |
34dc7c2f BB |
148 | |
149 | /* | |
150 | * Inheritance should have been done by now. | |
151 | */ | |
152 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
153 | ||
428870ff | 154 | os->os_checksum = zio_checksum_select(newval, ZIO_CHECKSUM_ON_VALUE); |
34dc7c2f BB |
155 | } |
156 | ||
157 | static void | |
158 | compression_changed_cb(void *arg, uint64_t newval) | |
159 | { | |
428870ff | 160 | objset_t *os = arg; |
34dc7c2f BB |
161 | |
162 | /* | |
163 | * Inheritance and range checking should have been done by now. | |
164 | */ | |
165 | ASSERT(newval != ZIO_COMPRESS_INHERIT); | |
166 | ||
99197f03 JG |
167 | os->os_compress = zio_compress_select(os->os_spa, newval, |
168 | ZIO_COMPRESS_ON); | |
34dc7c2f BB |
169 | } |
170 | ||
171 | static void | |
172 | copies_changed_cb(void *arg, uint64_t newval) | |
173 | { | |
428870ff | 174 | objset_t *os = arg; |
34dc7c2f BB |
175 | |
176 | /* | |
177 | * Inheritance and range checking should have been done by now. | |
178 | */ | |
179 | ASSERT(newval > 0); | |
428870ff | 180 | ASSERT(newval <= spa_max_replication(os->os_spa)); |
34dc7c2f | 181 | |
428870ff BB |
182 | os->os_copies = newval; |
183 | } | |
184 | ||
185 | static void | |
186 | dedup_changed_cb(void *arg, uint64_t newval) | |
187 | { | |
188 | objset_t *os = arg; | |
189 | spa_t *spa = os->os_spa; | |
190 | enum zio_checksum checksum; | |
191 | ||
192 | /* | |
193 | * Inheritance should have been done by now. | |
194 | */ | |
195 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
196 | ||
197 | checksum = zio_checksum_dedup_select(spa, newval, ZIO_CHECKSUM_OFF); | |
198 | ||
199 | os->os_dedup_checksum = checksum & ZIO_CHECKSUM_MASK; | |
200 | os->os_dedup_verify = !!(checksum & ZIO_CHECKSUM_VERIFY); | |
34dc7c2f BB |
201 | } |
202 | ||
b128c09f BB |
203 | static void |
204 | primary_cache_changed_cb(void *arg, uint64_t newval) | |
205 | { | |
428870ff | 206 | objset_t *os = arg; |
b128c09f BB |
207 | |
208 | /* | |
209 | * Inheritance and range checking should have been done by now. | |
210 | */ | |
211 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
212 | newval == ZFS_CACHE_METADATA); | |
213 | ||
428870ff | 214 | os->os_primary_cache = newval; |
b128c09f BB |
215 | } |
216 | ||
217 | static void | |
218 | secondary_cache_changed_cb(void *arg, uint64_t newval) | |
219 | { | |
428870ff | 220 | objset_t *os = arg; |
b128c09f BB |
221 | |
222 | /* | |
223 | * Inheritance and range checking should have been done by now. | |
224 | */ | |
225 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
226 | newval == ZFS_CACHE_METADATA); | |
227 | ||
428870ff BB |
228 | os->os_secondary_cache = newval; |
229 | } | |
230 | ||
231 | static void | |
232 | sync_changed_cb(void *arg, uint64_t newval) | |
233 | { | |
234 | objset_t *os = arg; | |
235 | ||
236 | /* | |
237 | * Inheritance and range checking should have been done by now. | |
238 | */ | |
239 | ASSERT(newval == ZFS_SYNC_STANDARD || newval == ZFS_SYNC_ALWAYS || | |
240 | newval == ZFS_SYNC_DISABLED); | |
241 | ||
242 | os->os_sync = newval; | |
243 | if (os->os_zil) | |
244 | zil_set_sync(os->os_zil, newval); | |
245 | } | |
246 | ||
faf0f58c MA |
247 | static void |
248 | redundant_metadata_changed_cb(void *arg, uint64_t newval) | |
249 | { | |
250 | objset_t *os = arg; | |
251 | ||
252 | /* | |
253 | * Inheritance and range checking should have been done by now. | |
254 | */ | |
255 | ASSERT(newval == ZFS_REDUNDANT_METADATA_ALL || | |
256 | newval == ZFS_REDUNDANT_METADATA_MOST); | |
257 | ||
258 | os->os_redundant_metadata = newval; | |
259 | } | |
260 | ||
428870ff BB |
261 | static void |
262 | logbias_changed_cb(void *arg, uint64_t newval) | |
263 | { | |
264 | objset_t *os = arg; | |
265 | ||
266 | ASSERT(newval == ZFS_LOGBIAS_LATENCY || | |
267 | newval == ZFS_LOGBIAS_THROUGHPUT); | |
268 | os->os_logbias = newval; | |
269 | if (os->os_zil) | |
270 | zil_set_logbias(os->os_zil, newval); | |
b128c09f BB |
271 | } |
272 | ||
f1512ee6 MA |
273 | static void |
274 | recordsize_changed_cb(void *arg, uint64_t newval) | |
275 | { | |
276 | objset_t *os = arg; | |
277 | ||
278 | os->os_recordsize = newval; | |
279 | } | |
280 | ||
34dc7c2f BB |
281 | void |
282 | dmu_objset_byteswap(void *buf, size_t size) | |
283 | { | |
284 | objset_phys_t *osp = buf; | |
285 | ||
9babb374 | 286 | ASSERT(size == OBJSET_OLD_PHYS_SIZE || size == sizeof (objset_phys_t)); |
34dc7c2f BB |
287 | dnode_byteswap(&osp->os_meta_dnode); |
288 | byteswap_uint64_array(&osp->os_zil_header, sizeof (zil_header_t)); | |
289 | osp->os_type = BSWAP_64(osp->os_type); | |
9babb374 BB |
290 | osp->os_flags = BSWAP_64(osp->os_flags); |
291 | if (size == sizeof (objset_phys_t)) { | |
292 | dnode_byteswap(&osp->os_userused_dnode); | |
293 | dnode_byteswap(&osp->os_groupused_dnode); | |
294 | } | |
34dc7c2f BB |
295 | } |
296 | ||
297 | int | |
298 | dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, | |
428870ff | 299 | objset_t **osp) |
34dc7c2f | 300 | { |
428870ff | 301 | objset_t *os; |
b128c09f | 302 | int i, err; |
34dc7c2f BB |
303 | |
304 | ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock)); | |
305 | ||
79c76d5b | 306 | os = kmem_zalloc(sizeof (objset_t), KM_SLEEP); |
428870ff BB |
307 | os->os_dsl_dataset = ds; |
308 | os->os_spa = spa; | |
309 | os->os_rootbp = bp; | |
310 | if (!BP_IS_HOLE(os->os_rootbp)) { | |
2a432414 | 311 | arc_flags_t aflags = ARC_FLAG_WAIT; |
5dbd68a3 | 312 | zbookmark_phys_t zb; |
428870ff BB |
313 | SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET, |
314 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
315 | ||
316 | if (DMU_OS_IS_L2CACHEABLE(os)) | |
2a432414 | 317 | aflags |= ARC_FLAG_L2CACHE; |
3a17a7a9 | 318 | if (DMU_OS_IS_L2COMPRESSIBLE(os)) |
2a432414 | 319 | aflags |= ARC_FLAG_L2COMPRESS; |
34dc7c2f | 320 | |
428870ff | 321 | dprintf_bp(os->os_rootbp, "reading %s", ""); |
294f6806 | 322 | err = arc_read(NULL, spa, os->os_rootbp, |
428870ff | 323 | arc_getbuf_func, &os->os_phys_buf, |
34dc7c2f | 324 | ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL, &aflags, &zb); |
13fe0198 | 325 | if (err != 0) { |
428870ff | 326 | kmem_free(os, sizeof (objset_t)); |
b128c09f BB |
327 | /* convert checksum errors into IO errors */ |
328 | if (err == ECKSUM) | |
2e528b49 | 329 | err = SET_ERROR(EIO); |
34dc7c2f BB |
330 | return (err); |
331 | } | |
9babb374 BB |
332 | |
333 | /* Increase the blocksize if we are permitted. */ | |
334 | if (spa_version(spa) >= SPA_VERSION_USERSPACE && | |
428870ff | 335 | arc_buf_size(os->os_phys_buf) < sizeof (objset_phys_t)) { |
9babb374 | 336 | arc_buf_t *buf = arc_buf_alloc(spa, |
428870ff | 337 | sizeof (objset_phys_t), &os->os_phys_buf, |
9babb374 BB |
338 | ARC_BUFC_METADATA); |
339 | bzero(buf->b_data, sizeof (objset_phys_t)); | |
428870ff BB |
340 | bcopy(os->os_phys_buf->b_data, buf->b_data, |
341 | arc_buf_size(os->os_phys_buf)); | |
342 | (void) arc_buf_remove_ref(os->os_phys_buf, | |
343 | &os->os_phys_buf); | |
344 | os->os_phys_buf = buf; | |
9babb374 BB |
345 | } |
346 | ||
428870ff BB |
347 | os->os_phys = os->os_phys_buf->b_data; |
348 | os->os_flags = os->os_phys->os_flags; | |
34dc7c2f | 349 | } else { |
9babb374 BB |
350 | int size = spa_version(spa) >= SPA_VERSION_USERSPACE ? |
351 | sizeof (objset_phys_t) : OBJSET_OLD_PHYS_SIZE; | |
428870ff BB |
352 | os->os_phys_buf = arc_buf_alloc(spa, size, |
353 | &os->os_phys_buf, ARC_BUFC_METADATA); | |
354 | os->os_phys = os->os_phys_buf->b_data; | |
355 | bzero(os->os_phys, size); | |
34dc7c2f BB |
356 | } |
357 | ||
358 | /* | |
359 | * Note: the changed_cb will be called once before the register | |
360 | * func returns, thus changing the checksum/compression from the | |
b128c09f BB |
361 | * default (fletcher2/off). Snapshots don't need to know about |
362 | * checksum/compression/copies. | |
34dc7c2f | 363 | */ |
9b67f605 | 364 | if (ds != NULL) { |
13fe0198 MA |
365 | err = dsl_prop_register(ds, |
366 | zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE), | |
428870ff | 367 | primary_cache_changed_cb, os); |
13fe0198 MA |
368 | if (err == 0) { |
369 | err = dsl_prop_register(ds, | |
370 | zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE), | |
428870ff | 371 | secondary_cache_changed_cb, os); |
13fe0198 | 372 | } |
0c66c32d | 373 | if (!ds->ds_is_snapshot) { |
13fe0198 MA |
374 | if (err == 0) { |
375 | err = dsl_prop_register(ds, | |
376 | zfs_prop_to_name(ZFS_PROP_CHECKSUM), | |
428870ff | 377 | checksum_changed_cb, os); |
13fe0198 MA |
378 | } |
379 | if (err == 0) { | |
380 | err = dsl_prop_register(ds, | |
381 | zfs_prop_to_name(ZFS_PROP_COMPRESSION), | |
428870ff | 382 | compression_changed_cb, os); |
13fe0198 MA |
383 | } |
384 | if (err == 0) { | |
385 | err = dsl_prop_register(ds, | |
386 | zfs_prop_to_name(ZFS_PROP_COPIES), | |
428870ff | 387 | copies_changed_cb, os); |
13fe0198 MA |
388 | } |
389 | if (err == 0) { | |
390 | err = dsl_prop_register(ds, | |
391 | zfs_prop_to_name(ZFS_PROP_DEDUP), | |
428870ff | 392 | dedup_changed_cb, os); |
13fe0198 MA |
393 | } |
394 | if (err == 0) { | |
395 | err = dsl_prop_register(ds, | |
396 | zfs_prop_to_name(ZFS_PROP_LOGBIAS), | |
428870ff | 397 | logbias_changed_cb, os); |
13fe0198 MA |
398 | } |
399 | if (err == 0) { | |
400 | err = dsl_prop_register(ds, | |
401 | zfs_prop_to_name(ZFS_PROP_SYNC), | |
428870ff | 402 | sync_changed_cb, os); |
13fe0198 | 403 | } |
faf0f58c MA |
404 | if (err == 0) { |
405 | err = dsl_prop_register(ds, | |
406 | zfs_prop_to_name( | |
407 | ZFS_PROP_REDUNDANT_METADATA), | |
408 | redundant_metadata_changed_cb, os); | |
409 | } | |
f1512ee6 MA |
410 | if (err == 0) { |
411 | err = dsl_prop_register(ds, | |
412 | zfs_prop_to_name(ZFS_PROP_RECORDSIZE), | |
413 | recordsize_changed_cb, os); | |
414 | } | |
b128c09f | 415 | } |
13fe0198 | 416 | if (err != 0) { |
428870ff | 417 | VERIFY(arc_buf_remove_ref(os->os_phys_buf, |
13fe0198 | 418 | &os->os_phys_buf)); |
428870ff | 419 | kmem_free(os, sizeof (objset_t)); |
34dc7c2f BB |
420 | return (err); |
421 | } | |
9b67f605 | 422 | } else { |
34dc7c2f | 423 | /* It's the meta-objset. */ |
428870ff | 424 | os->os_checksum = ZIO_CHECKSUM_FLETCHER_4; |
99197f03 | 425 | os->os_compress = ZIO_COMPRESS_ON; |
428870ff BB |
426 | os->os_copies = spa_max_replication(spa); |
427 | os->os_dedup_checksum = ZIO_CHECKSUM_OFF; | |
faf0f58c MA |
428 | os->os_dedup_verify = B_FALSE; |
429 | os->os_logbias = ZFS_LOGBIAS_LATENCY; | |
430 | os->os_sync = ZFS_SYNC_STANDARD; | |
428870ff BB |
431 | os->os_primary_cache = ZFS_CACHE_ALL; |
432 | os->os_secondary_cache = ZFS_CACHE_ALL; | |
34dc7c2f BB |
433 | } |
434 | ||
0c66c32d | 435 | if (ds == NULL || !ds->ds_is_snapshot) |
572e2857 | 436 | os->os_zil_header = os->os_phys->os_zil_header; |
428870ff | 437 | os->os_zil = zil_alloc(os, &os->os_zil_header); |
34dc7c2f BB |
438 | |
439 | for (i = 0; i < TXG_SIZE; i++) { | |
428870ff | 440 | list_create(&os->os_dirty_dnodes[i], sizeof (dnode_t), |
34dc7c2f | 441 | offsetof(dnode_t, dn_dirty_link[i])); |
428870ff | 442 | list_create(&os->os_free_dnodes[i], sizeof (dnode_t), |
34dc7c2f BB |
443 | offsetof(dnode_t, dn_dirty_link[i])); |
444 | } | |
428870ff | 445 | list_create(&os->os_dnodes, sizeof (dnode_t), |
34dc7c2f | 446 | offsetof(dnode_t, dn_link)); |
428870ff | 447 | list_create(&os->os_downgraded_dbufs, sizeof (dmu_buf_impl_t), |
34dc7c2f BB |
448 | offsetof(dmu_buf_impl_t, db_link)); |
449 | ||
0c66c32d JG |
450 | list_link_init(&os->os_evicting_node); |
451 | ||
428870ff BB |
452 | mutex_init(&os->os_lock, NULL, MUTEX_DEFAULT, NULL); |
453 | mutex_init(&os->os_obj_lock, NULL, MUTEX_DEFAULT, NULL); | |
454 | mutex_init(&os->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL); | |
455 | ||
0c66c32d JG |
456 | dnode_special_open(os, &os->os_phys->os_meta_dnode, |
457 | DMU_META_DNODE_OBJECT, &os->os_meta_dnode); | |
428870ff | 458 | if (arc_buf_size(os->os_phys_buf) >= sizeof (objset_phys_t)) { |
0c66c32d JG |
459 | dnode_special_open(os, &os->os_phys->os_userused_dnode, |
460 | DMU_USERUSED_OBJECT, &os->os_userused_dnode); | |
461 | dnode_special_open(os, &os->os_phys->os_groupused_dnode, | |
462 | DMU_GROUPUSED_OBJECT, &os->os_groupused_dnode); | |
9babb374 | 463 | } |
34dc7c2f | 464 | |
428870ff | 465 | *osp = os; |
34dc7c2f BB |
466 | return (0); |
467 | } | |
468 | ||
428870ff BB |
469 | int |
470 | dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp) | |
34dc7c2f | 471 | { |
428870ff | 472 | int err = 0; |
34dc7c2f BB |
473 | |
474 | mutex_enter(&ds->ds_opening_lock); | |
9b67f605 MA |
475 | if (ds->ds_objset == NULL) { |
476 | objset_t *os; | |
34dc7c2f | 477 | err = dmu_objset_open_impl(dsl_dataset_get_spa(ds), |
9b67f605 MA |
478 | ds, dsl_dataset_get_blkptr(ds), &os); |
479 | ||
480 | if (err == 0) { | |
481 | mutex_enter(&ds->ds_lock); | |
482 | ASSERT(ds->ds_objset == NULL); | |
483 | ds->ds_objset = os; | |
484 | mutex_exit(&ds->ds_lock); | |
485 | } | |
34dc7c2f | 486 | } |
9b67f605 | 487 | *osp = ds->ds_objset; |
34dc7c2f | 488 | mutex_exit(&ds->ds_opening_lock); |
428870ff | 489 | return (err); |
34dc7c2f BB |
490 | } |
491 | ||
13fe0198 MA |
492 | /* |
493 | * Holds the pool while the objset is held. Therefore only one objset | |
494 | * can be held at a time. | |
495 | */ | |
34dc7c2f | 496 | int |
428870ff | 497 | dmu_objset_hold(const char *name, void *tag, objset_t **osp) |
34dc7c2f | 498 | { |
13fe0198 | 499 | dsl_pool_t *dp; |
428870ff | 500 | dsl_dataset_t *ds; |
34dc7c2f BB |
501 | int err; |
502 | ||
13fe0198 MA |
503 | err = dsl_pool_hold(name, tag, &dp); |
504 | if (err != 0) | |
505 | return (err); | |
506 | err = dsl_dataset_hold(dp, name, tag, &ds); | |
507 | if (err != 0) { | |
508 | dsl_pool_rele(dp, tag); | |
428870ff | 509 | return (err); |
13fe0198 | 510 | } |
428870ff BB |
511 | |
512 | err = dmu_objset_from_ds(ds, osp); | |
13fe0198 | 513 | if (err != 0) { |
428870ff | 514 | dsl_dataset_rele(ds, tag); |
13fe0198 MA |
515 | dsl_pool_rele(dp, tag); |
516 | } | |
428870ff | 517 | |
34dc7c2f BB |
518 | return (err); |
519 | } | |
520 | ||
9c43027b AJ |
521 | static int |
522 | dmu_objset_own_impl(dsl_dataset_t *ds, dmu_objset_type_t type, | |
523 | boolean_t readonly, void *tag, objset_t **osp) | |
524 | { | |
525 | int err; | |
526 | ||
527 | err = dmu_objset_from_ds(ds, osp); | |
528 | if (err != 0) { | |
529 | dsl_dataset_disown(ds, tag); | |
530 | } else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) { | |
531 | dsl_dataset_disown(ds, tag); | |
532 | return (SET_ERROR(EINVAL)); | |
533 | } else if (!readonly && dsl_dataset_is_snapshot(ds)) { | |
534 | dsl_dataset_disown(ds, tag); | |
535 | return (SET_ERROR(EROFS)); | |
536 | } | |
537 | return (err); | |
538 | } | |
539 | ||
13fe0198 MA |
540 | /* |
541 | * dsl_pool must not be held when this is called. | |
542 | * Upon successful return, there will be a longhold on the dataset, | |
543 | * and the dsl_pool will not be held. | |
544 | */ | |
34dc7c2f | 545 | int |
428870ff BB |
546 | dmu_objset_own(const char *name, dmu_objset_type_t type, |
547 | boolean_t readonly, void *tag, objset_t **osp) | |
34dc7c2f | 548 | { |
13fe0198 | 549 | dsl_pool_t *dp; |
34dc7c2f BB |
550 | dsl_dataset_t *ds; |
551 | int err; | |
552 | ||
13fe0198 MA |
553 | err = dsl_pool_hold(name, FTAG, &dp); |
554 | if (err != 0) | |
555 | return (err); | |
556 | err = dsl_dataset_own(dp, name, tag, &ds); | |
557 | if (err != 0) { | |
558 | dsl_pool_rele(dp, FTAG); | |
34dc7c2f | 559 | return (err); |
13fe0198 | 560 | } |
9c43027b | 561 | err = dmu_objset_own_impl(ds, type, readonly, tag, osp); |
13fe0198 | 562 | dsl_pool_rele(dp, FTAG); |
9c43027b | 563 | |
34dc7c2f BB |
564 | return (err); |
565 | } | |
566 | ||
9c43027b AJ |
567 | int |
568 | dmu_objset_own_obj(dsl_pool_t *dp, uint64_t obj, dmu_objset_type_t type, | |
569 | boolean_t readonly, void *tag, objset_t **osp) | |
570 | { | |
571 | dsl_dataset_t *ds; | |
572 | int err; | |
573 | ||
574 | err = dsl_dataset_own_obj(dp, obj, tag, &ds); | |
575 | if (err != 0) | |
576 | return (err); | |
577 | ||
578 | return (dmu_objset_own_impl(ds, type, readonly, tag, osp)); | |
579 | } | |
580 | ||
34dc7c2f | 581 | void |
428870ff | 582 | dmu_objset_rele(objset_t *os, void *tag) |
34dc7c2f | 583 | { |
13fe0198 | 584 | dsl_pool_t *dp = dmu_objset_pool(os); |
428870ff | 585 | dsl_dataset_rele(os->os_dsl_dataset, tag); |
13fe0198 | 586 | dsl_pool_rele(dp, tag); |
428870ff | 587 | } |
b128c09f | 588 | |
831baf06 KW |
589 | /* |
590 | * When we are called, os MUST refer to an objset associated with a dataset | |
591 | * that is owned by 'tag'; that is, is held and long held by 'tag' and ds_owner | |
592 | * == tag. We will then release and reacquire ownership of the dataset while | |
593 | * holding the pool config_rwlock to avoid intervening namespace or ownership | |
594 | * changes may occur. | |
595 | * | |
596 | * This exists solely to accommodate zfs_ioc_userspace_upgrade()'s desire to | |
597 | * release the hold on its dataset and acquire a new one on the dataset of the | |
598 | * same name so that it can be partially torn down and reconstructed. | |
599 | */ | |
600 | void | |
601 | dmu_objset_refresh_ownership(objset_t *os, void *tag) | |
602 | { | |
603 | dsl_pool_t *dp; | |
604 | dsl_dataset_t *ds, *newds; | |
605 | char name[MAXNAMELEN]; | |
606 | ||
607 | ds = os->os_dsl_dataset; | |
608 | VERIFY3P(ds, !=, NULL); | |
609 | VERIFY3P(ds->ds_owner, ==, tag); | |
610 | VERIFY(dsl_dataset_long_held(ds)); | |
611 | ||
612 | dsl_dataset_name(ds, name); | |
613 | dp = dmu_objset_pool(os); | |
614 | dsl_pool_config_enter(dp, FTAG); | |
615 | dmu_objset_disown(os, tag); | |
616 | VERIFY0(dsl_dataset_own(dp, name, tag, &newds)); | |
617 | VERIFY3P(newds, ==, os->os_dsl_dataset); | |
618 | dsl_pool_config_exit(dp, FTAG); | |
619 | } | |
620 | ||
428870ff BB |
621 | void |
622 | dmu_objset_disown(objset_t *os, void *tag) | |
623 | { | |
624 | dsl_dataset_disown(os->os_dsl_dataset, tag); | |
34dc7c2f BB |
625 | } |
626 | ||
13fe0198 | 627 | void |
34dc7c2f BB |
628 | dmu_objset_evict_dbufs(objset_t *os) |
629 | { | |
0c66c32d | 630 | dnode_t *dn_marker; |
34dc7c2f BB |
631 | dnode_t *dn; |
632 | ||
0c66c32d | 633 | dn_marker = kmem_alloc(sizeof (dnode_t), KM_SLEEP); |
34dc7c2f | 634 | |
0c66c32d JG |
635 | mutex_enter(&os->os_lock); |
636 | dn = list_head(&os->os_dnodes); | |
637 | while (dn != NULL) { | |
638 | /* | |
639 | * Skip dnodes without holds. We have to do this dance | |
640 | * because dnode_add_ref() only works if there is already a | |
641 | * hold. If the dnode has no holds, then it has no dbufs. | |
642 | */ | |
643 | if (dnode_add_ref(dn, FTAG)) { | |
644 | list_insert_after(&os->os_dnodes, dn, dn_marker); | |
645 | mutex_exit(&os->os_lock); | |
34dc7c2f | 646 | |
0c66c32d JG |
647 | dnode_evict_dbufs(dn); |
648 | dnode_rele(dn, FTAG); | |
34dc7c2f | 649 | |
0c66c32d JG |
650 | mutex_enter(&os->os_lock); |
651 | dn = list_next(&os->os_dnodes, dn_marker); | |
652 | list_remove(&os->os_dnodes, dn_marker); | |
653 | } else { | |
654 | dn = list_next(&os->os_dnodes, dn); | |
655 | } | |
656 | } | |
657 | mutex_exit(&os->os_lock); | |
34dc7c2f | 658 | |
0c66c32d | 659 | kmem_free(dn_marker, sizeof (dnode_t)); |
34dc7c2f | 660 | |
0c66c32d JG |
661 | if (DMU_USERUSED_DNODE(os) != NULL) { |
662 | dnode_evict_dbufs(DMU_GROUPUSED_DNODE(os)); | |
663 | dnode_evict_dbufs(DMU_USERUSED_DNODE(os)); | |
34dc7c2f | 664 | } |
0c66c32d | 665 | dnode_evict_dbufs(DMU_META_DNODE(os)); |
34dc7c2f BB |
666 | } |
667 | ||
0c66c32d JG |
668 | /* |
669 | * Objset eviction processing is split into into two pieces. | |
670 | * The first marks the objset as evicting, evicts any dbufs that | |
671 | * have a refcount of zero, and then queues up the objset for the | |
672 | * second phase of eviction. Once os->os_dnodes has been cleared by | |
673 | * dnode_buf_pageout()->dnode_destroy(), the second phase is executed. | |
674 | * The second phase closes the special dnodes, dequeues the objset from | |
675 | * the list of those undergoing eviction, and finally frees the objset. | |
676 | * | |
677 | * NOTE: Due to asynchronous eviction processing (invocation of | |
678 | * dnode_buf_pageout()), it is possible for the meta dnode for the | |
679 | * objset to have no holds even though os->os_dnodes is not empty. | |
680 | */ | |
34dc7c2f | 681 | void |
428870ff | 682 | dmu_objset_evict(objset_t *os) |
34dc7c2f | 683 | { |
d6320ddb | 684 | int t; |
34dc7c2f | 685 | |
6f1ffb06 MA |
686 | dsl_dataset_t *ds = os->os_dsl_dataset; |
687 | ||
d6320ddb | 688 | for (t = 0; t < TXG_SIZE; t++) |
428870ff | 689 | ASSERT(!dmu_objset_is_dirty(os, t)); |
34dc7c2f | 690 | |
0eb21616 JG |
691 | if (ds) |
692 | dsl_prop_unregister_all(ds, os); | |
34dc7c2f | 693 | |
428870ff BB |
694 | if (os->os_sa) |
695 | sa_tear_down(os); | |
696 | ||
13fe0198 | 697 | dmu_objset_evict_dbufs(os); |
34dc7c2f | 698 | |
0c66c32d JG |
699 | mutex_enter(&os->os_lock); |
700 | spa_evicting_os_register(os->os_spa, os); | |
701 | if (list_is_empty(&os->os_dnodes)) { | |
702 | mutex_exit(&os->os_lock); | |
703 | dmu_objset_evict_done(os); | |
704 | } else { | |
705 | mutex_exit(&os->os_lock); | |
706 | } | |
707 | } | |
708 | ||
709 | void | |
710 | dmu_objset_evict_done(objset_t *os) | |
711 | { | |
712 | ASSERT3P(list_head(&os->os_dnodes), ==, NULL); | |
713 | ||
572e2857 BB |
714 | dnode_special_close(&os->os_meta_dnode); |
715 | if (DMU_USERUSED_DNODE(os)) { | |
716 | dnode_special_close(&os->os_userused_dnode); | |
717 | dnode_special_close(&os->os_groupused_dnode); | |
9babb374 | 718 | } |
428870ff BB |
719 | zil_free(os->os_zil); |
720 | ||
13fe0198 | 721 | VERIFY(arc_buf_remove_ref(os->os_phys_buf, &os->os_phys_buf)); |
572e2857 BB |
722 | |
723 | /* | |
724 | * This is a barrier to prevent the objset from going away in | |
725 | * dnode_move() until we can safely ensure that the objset is still in | |
726 | * use. We consider the objset valid before the barrier and invalid | |
727 | * after the barrier. | |
728 | */ | |
729 | rw_enter(&os_lock, RW_READER); | |
730 | rw_exit(&os_lock); | |
731 | ||
428870ff BB |
732 | mutex_destroy(&os->os_lock); |
733 | mutex_destroy(&os->os_obj_lock); | |
734 | mutex_destroy(&os->os_user_ptr_lock); | |
0c66c32d | 735 | spa_evicting_os_deregister(os->os_spa, os); |
428870ff BB |
736 | kmem_free(os, sizeof (objset_t)); |
737 | } | |
9babb374 | 738 | |
428870ff BB |
739 | timestruc_t |
740 | dmu_objset_snap_cmtime(objset_t *os) | |
741 | { | |
742 | return (dsl_dir_snap_cmtime(os->os_dsl_dataset->ds_dir)); | |
34dc7c2f BB |
743 | } |
744 | ||
745 | /* called from dsl for meta-objset */ | |
428870ff | 746 | objset_t * |
34dc7c2f BB |
747 | dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, |
748 | dmu_objset_type_t type, dmu_tx_t *tx) | |
749 | { | |
428870ff | 750 | objset_t *os; |
34dc7c2f BB |
751 | dnode_t *mdn; |
752 | ||
753 | ASSERT(dmu_tx_is_syncing(tx)); | |
13fe0198 | 754 | |
572e2857 | 755 | if (ds != NULL) |
13fe0198 | 756 | VERIFY0(dmu_objset_from_ds(ds, &os)); |
572e2857 | 757 | else |
13fe0198 | 758 | VERIFY0(dmu_objset_open_impl(spa, NULL, bp, &os)); |
572e2857 BB |
759 | |
760 | mdn = DMU_META_DNODE(os); | |
34dc7c2f BB |
761 | |
762 | dnode_allocate(mdn, DMU_OT_DNODE, 1 << DNODE_BLOCK_SHIFT, | |
763 | DN_MAX_INDBLKSHIFT, DMU_OT_NONE, 0, tx); | |
764 | ||
765 | /* | |
766 | * We don't want to have to increase the meta-dnode's nlevels | |
767 | * later, because then we could do it in quescing context while | |
768 | * we are also accessing it in open context. | |
769 | * | |
770 | * This precaution is not necessary for the MOS (ds == NULL), | |
771 | * because the MOS is only updated in syncing context. | |
772 | * This is most fortunate: the MOS is the only objset that | |
773 | * needs to be synced multiple times as spa_sync() iterates | |
774 | * to convergence, so minimizing its dn_nlevels matters. | |
775 | */ | |
776 | if (ds != NULL) { | |
777 | int levels = 1; | |
778 | ||
779 | /* | |
780 | * Determine the number of levels necessary for the meta-dnode | |
781 | * to contain DN_MAX_OBJECT dnodes. | |
782 | */ | |
783 | while ((uint64_t)mdn->dn_nblkptr << (mdn->dn_datablkshift + | |
784 | (levels - 1) * (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) < | |
785 | DN_MAX_OBJECT * sizeof (dnode_phys_t)) | |
786 | levels++; | |
787 | ||
788 | mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] = | |
789 | mdn->dn_nlevels = levels; | |
790 | } | |
791 | ||
792 | ASSERT(type != DMU_OST_NONE); | |
793 | ASSERT(type != DMU_OST_ANY); | |
794 | ASSERT(type < DMU_OST_NUMTYPES); | |
428870ff BB |
795 | os->os_phys->os_type = type; |
796 | if (dmu_objset_userused_enabled(os)) { | |
797 | os->os_phys->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; | |
798 | os->os_flags = os->os_phys->os_flags; | |
9babb374 | 799 | } |
34dc7c2f BB |
800 | |
801 | dsl_dataset_dirty(ds, tx); | |
802 | ||
428870ff | 803 | return (os); |
34dc7c2f BB |
804 | } |
805 | ||
13fe0198 MA |
806 | typedef struct dmu_objset_create_arg { |
807 | const char *doca_name; | |
808 | cred_t *doca_cred; | |
809 | void (*doca_userfunc)(objset_t *os, void *arg, | |
810 | cred_t *cr, dmu_tx_t *tx); | |
811 | void *doca_userarg; | |
812 | dmu_objset_type_t doca_type; | |
813 | uint64_t doca_flags; | |
814 | } dmu_objset_create_arg_t; | |
34dc7c2f BB |
815 | |
816 | /*ARGSUSED*/ | |
817 | static int | |
13fe0198 | 818 | dmu_objset_create_check(void *arg, dmu_tx_t *tx) |
34dc7c2f | 819 | { |
13fe0198 MA |
820 | dmu_objset_create_arg_t *doca = arg; |
821 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
822 | dsl_dir_t *pdd; | |
823 | const char *tail; | |
824 | int error; | |
34dc7c2f | 825 | |
13fe0198 | 826 | if (strchr(doca->doca_name, '@') != NULL) |
2e528b49 | 827 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 828 | |
13fe0198 MA |
829 | error = dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail); |
830 | if (error != 0) | |
831 | return (error); | |
832 | if (tail == NULL) { | |
833 | dsl_dir_rele(pdd, FTAG); | |
2e528b49 | 834 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 835 | } |
788eb90c JJ |
836 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
837 | doca->doca_cred); | |
13fe0198 | 838 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f | 839 | |
788eb90c | 840 | return (error); |
34dc7c2f BB |
841 | } |
842 | ||
843 | static void | |
13fe0198 | 844 | dmu_objset_create_sync(void *arg, dmu_tx_t *tx) |
34dc7c2f | 845 | { |
13fe0198 MA |
846 | dmu_objset_create_arg_t *doca = arg; |
847 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
848 | dsl_dir_t *pdd; | |
849 | const char *tail; | |
6f1ffb06 | 850 | dsl_dataset_t *ds; |
13fe0198 | 851 | uint64_t obj; |
6f1ffb06 | 852 | blkptr_t *bp; |
13fe0198 | 853 | objset_t *os; |
34dc7c2f | 854 | |
13fe0198 | 855 | VERIFY0(dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail)); |
34dc7c2f | 856 | |
13fe0198 MA |
857 | obj = dsl_dataset_create_sync(pdd, tail, NULL, doca->doca_flags, |
858 | doca->doca_cred, tx); | |
34dc7c2f | 859 | |
13fe0198 | 860 | VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); |
6f1ffb06 | 861 | bp = dsl_dataset_get_blkptr(ds); |
13fe0198 MA |
862 | os = dmu_objset_create_impl(pdd->dd_pool->dp_spa, |
863 | ds, bp, doca->doca_type, tx); | |
34dc7c2f | 864 | |
13fe0198 MA |
865 | if (doca->doca_userfunc != NULL) { |
866 | doca->doca_userfunc(os, doca->doca_userarg, | |
867 | doca->doca_cred, tx); | |
34dc7c2f BB |
868 | } |
869 | ||
13fe0198 | 870 | spa_history_log_internal_ds(ds, "create", tx, ""); |
6f1ffb06 | 871 | dsl_dataset_rele(ds, FTAG); |
13fe0198 | 872 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f BB |
873 | } |
874 | ||
875 | int | |
428870ff | 876 | dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, |
34dc7c2f BB |
877 | void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg) |
878 | { | |
13fe0198 | 879 | dmu_objset_create_arg_t doca; |
34dc7c2f | 880 | |
13fe0198 MA |
881 | doca.doca_name = name; |
882 | doca.doca_cred = CRED(); | |
883 | doca.doca_flags = flags; | |
884 | doca.doca_userfunc = func; | |
885 | doca.doca_userarg = arg; | |
886 | doca.doca_type = type; | |
34dc7c2f | 887 | |
13fe0198 | 888 | return (dsl_sync_task(name, |
3d45fdd6 MA |
889 | dmu_objset_create_check, dmu_objset_create_sync, &doca, |
890 | 5, ZFS_SPACE_CHECK_NORMAL)); | |
34dc7c2f BB |
891 | } |
892 | ||
13fe0198 MA |
893 | typedef struct dmu_objset_clone_arg { |
894 | const char *doca_clone; | |
895 | const char *doca_origin; | |
896 | cred_t *doca_cred; | |
897 | } dmu_objset_clone_arg_t; | |
898 | ||
899 | /*ARGSUSED*/ | |
900 | static int | |
901 | dmu_objset_clone_check(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 902 | { |
13fe0198 | 903 | dmu_objset_clone_arg_t *doca = arg; |
428870ff BB |
904 | dsl_dir_t *pdd; |
905 | const char *tail; | |
13fe0198 MA |
906 | int error; |
907 | dsl_dataset_t *origin; | |
908 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
34dc7c2f | 909 | |
13fe0198 | 910 | if (strchr(doca->doca_clone, '@') != NULL) |
2e528b49 | 911 | return (SET_ERROR(EINVAL)); |
13fe0198 MA |
912 | |
913 | error = dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail); | |
914 | if (error != 0) | |
915 | return (error); | |
428870ff | 916 | if (tail == NULL) { |
13fe0198 | 917 | dsl_dir_rele(pdd, FTAG); |
2e528b49 | 918 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 919 | } |
1cddb8c9 | 920 | |
788eb90c JJ |
921 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
922 | doca->doca_cred); | |
923 | if (error != 0) { | |
924 | dsl_dir_rele(pdd, FTAG); | |
925 | return (SET_ERROR(EDQUOT)); | |
926 | } | |
13fe0198 | 927 | dsl_dir_rele(pdd, FTAG); |
428870ff | 928 | |
13fe0198 MA |
929 | error = dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin); |
930 | if (error != 0) | |
572e2857 BB |
931 | return (error); |
932 | ||
13fe0198 | 933 | /* You can only clone snapshots, not the head datasets. */ |
0c66c32d | 934 | if (!origin->ds_is_snapshot) { |
13fe0198 | 935 | dsl_dataset_rele(origin, FTAG); |
2e528b49 | 936 | return (SET_ERROR(EINVAL)); |
428870ff | 937 | } |
13fe0198 | 938 | dsl_dataset_rele(origin, FTAG); |
572e2857 | 939 | |
13fe0198 | 940 | return (0); |
9babb374 | 941 | } |
34dc7c2f | 942 | |
13fe0198 MA |
943 | static void |
944 | dmu_objset_clone_sync(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 945 | { |
13fe0198 MA |
946 | dmu_objset_clone_arg_t *doca = arg; |
947 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
948 | dsl_dir_t *pdd; | |
949 | const char *tail; | |
950 | dsl_dataset_t *origin, *ds; | |
951 | uint64_t obj; | |
952 | char namebuf[MAXNAMELEN]; | |
6f1ffb06 | 953 | |
13fe0198 MA |
954 | VERIFY0(dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail)); |
955 | VERIFY0(dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin)); | |
6f1ffb06 | 956 | |
13fe0198 MA |
957 | obj = dsl_dataset_create_sync(pdd, tail, origin, 0, |
958 | doca->doca_cred, tx); | |
34dc7c2f | 959 | |
13fe0198 MA |
960 | VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); |
961 | dsl_dataset_name(origin, namebuf); | |
962 | spa_history_log_internal_ds(ds, "clone", tx, | |
963 | "origin=%s (%llu)", namebuf, origin->ds_object); | |
964 | dsl_dataset_rele(ds, FTAG); | |
965 | dsl_dataset_rele(origin, FTAG); | |
966 | dsl_dir_rele(pdd, FTAG); | |
34dc7c2f BB |
967 | } |
968 | ||
969 | int | |
13fe0198 | 970 | dmu_objset_clone(const char *clone, const char *origin) |
34dc7c2f | 971 | { |
13fe0198 | 972 | dmu_objset_clone_arg_t doca; |
34dc7c2f | 973 | |
13fe0198 MA |
974 | doca.doca_clone = clone; |
975 | doca.doca_origin = origin; | |
976 | doca.doca_cred = CRED(); | |
572e2857 | 977 | |
13fe0198 | 978 | return (dsl_sync_task(clone, |
3d45fdd6 MA |
979 | dmu_objset_clone_check, dmu_objset_clone_sync, &doca, |
980 | 5, ZFS_SPACE_CHECK_NORMAL)); | |
6f1ffb06 MA |
981 | } |
982 | ||
983 | int | |
984 | dmu_objset_snapshot_one(const char *fsname, const char *snapname) | |
985 | { | |
986 | int err; | |
987 | char *longsnap = kmem_asprintf("%s@%s", fsname, snapname); | |
988 | nvlist_t *snaps = fnvlist_alloc(); | |
989 | ||
990 | fnvlist_add_boolean(snaps, longsnap); | |
6f1ffb06 | 991 | strfree(longsnap); |
13fe0198 MA |
992 | err = dsl_dataset_snapshot(snaps, NULL, NULL); |
993 | fnvlist_free(snaps); | |
6f1ffb06 MA |
994 | return (err); |
995 | } | |
996 | ||
34dc7c2f | 997 | static void |
9babb374 | 998 | dmu_objset_sync_dnodes(list_t *list, list_t *newlist, dmu_tx_t *tx) |
34dc7c2f BB |
999 | { |
1000 | dnode_t *dn; | |
1001 | ||
c65aa5b2 | 1002 | while ((dn = list_head(list))) { |
34dc7c2f BB |
1003 | ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT); |
1004 | ASSERT(dn->dn_dbuf->db_data_pending); | |
1005 | /* | |
9babb374 BB |
1006 | * Initialize dn_zio outside dnode_sync() because the |
1007 | * meta-dnode needs to set it ouside dnode_sync(). | |
34dc7c2f BB |
1008 | */ |
1009 | dn->dn_zio = dn->dn_dbuf->db_data_pending->dr_zio; | |
1010 | ASSERT(dn->dn_zio); | |
1011 | ||
1012 | ASSERT3U(dn->dn_nlevels, <=, DN_MAX_LEVELS); | |
1013 | list_remove(list, dn); | |
9babb374 BB |
1014 | |
1015 | if (newlist) { | |
1016 | (void) dnode_add_ref(dn, newlist); | |
1017 | list_insert_tail(newlist, dn); | |
1018 | } | |
1019 | ||
34dc7c2f BB |
1020 | dnode_sync(dn, tx); |
1021 | } | |
1022 | } | |
1023 | ||
1024 | /* ARGSUSED */ | |
1025 | static void | |
428870ff | 1026 | dmu_objset_write_ready(zio_t *zio, arc_buf_t *abuf, void *arg) |
34dc7c2f | 1027 | { |
d6320ddb BB |
1028 | int i; |
1029 | ||
b128c09f | 1030 | blkptr_t *bp = zio->io_bp; |
428870ff | 1031 | objset_t *os = arg; |
34dc7c2f | 1032 | dnode_phys_t *dnp = &os->os_phys->os_meta_dnode; |
34dc7c2f | 1033 | |
9b67f605 | 1034 | ASSERT(!BP_IS_EMBEDDED(bp)); |
13fe0198 MA |
1035 | ASSERT3P(bp, ==, os->os_rootbp); |
1036 | ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_OBJSET); | |
1037 | ASSERT0(BP_GET_LEVEL(bp)); | |
34dc7c2f BB |
1038 | |
1039 | /* | |
9babb374 BB |
1040 | * Update rootbp fill count: it should be the number of objects |
1041 | * allocated in the object set (not counting the "special" | |
1042 | * objects that are stored in the objset_phys_t -- the meta | |
1043 | * dnode and user/group accounting objects). | |
34dc7c2f | 1044 | */ |
9babb374 | 1045 | bp->blk_fill = 0; |
d6320ddb | 1046 | for (i = 0; i < dnp->dn_nblkptr; i++) |
9b67f605 | 1047 | bp->blk_fill += BP_GET_FILL(&dnp->dn_blkptr[i]); |
428870ff BB |
1048 | } |
1049 | ||
1050 | /* ARGSUSED */ | |
1051 | static void | |
1052 | dmu_objset_write_done(zio_t *zio, arc_buf_t *abuf, void *arg) | |
1053 | { | |
1054 | blkptr_t *bp = zio->io_bp; | |
1055 | blkptr_t *bp_orig = &zio->io_bp_orig; | |
1056 | objset_t *os = arg; | |
34dc7c2f | 1057 | |
b128c09f | 1058 | if (zio->io_flags & ZIO_FLAG_IO_REWRITE) { |
428870ff | 1059 | ASSERT(BP_EQUAL(bp, bp_orig)); |
b128c09f | 1060 | } else { |
428870ff BB |
1061 | dsl_dataset_t *ds = os->os_dsl_dataset; |
1062 | dmu_tx_t *tx = os->os_synctx; | |
1063 | ||
1064 | (void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE); | |
1065 | dsl_dataset_block_born(ds, bp, tx); | |
34dc7c2f BB |
1066 | } |
1067 | } | |
1068 | ||
34dc7c2f BB |
1069 | /* called from dsl */ |
1070 | void | |
428870ff | 1071 | dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx) |
34dc7c2f BB |
1072 | { |
1073 | int txgoff; | |
5dbd68a3 | 1074 | zbookmark_phys_t zb; |
428870ff | 1075 | zio_prop_t zp; |
34dc7c2f BB |
1076 | zio_t *zio; |
1077 | list_t *list; | |
9babb374 | 1078 | list_t *newlist = NULL; |
34dc7c2f BB |
1079 | dbuf_dirty_record_t *dr; |
1080 | ||
1081 | dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg); | |
1082 | ||
1083 | ASSERT(dmu_tx_is_syncing(tx)); | |
1084 | /* XXX the write_done callback should really give us the tx... */ | |
1085 | os->os_synctx = tx; | |
1086 | ||
1087 | if (os->os_dsl_dataset == NULL) { | |
1088 | /* | |
1089 | * This is the MOS. If we have upgraded, | |
1090 | * spa_max_replication() could change, so reset | |
1091 | * os_copies here. | |
1092 | */ | |
1093 | os->os_copies = spa_max_replication(os->os_spa); | |
1094 | } | |
1095 | ||
1096 | /* | |
1097 | * Create the root block IO | |
1098 | */ | |
428870ff BB |
1099 | SET_BOOKMARK(&zb, os->os_dsl_dataset ? |
1100 | os->os_dsl_dataset->ds_object : DMU_META_OBJSET, | |
1101 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
294f6806 | 1102 | arc_release(os->os_phys_buf, &os->os_phys_buf); |
b128c09f | 1103 | |
428870ff | 1104 | dmu_write_policy(os, NULL, 0, 0, &zp); |
9babb374 | 1105 | |
428870ff | 1106 | zio = arc_write(pio, os->os_spa, tx->tx_txg, |
3a17a7a9 SK |
1107 | os->os_rootbp, os->os_phys_buf, DMU_OS_IS_L2CACHEABLE(os), |
1108 | DMU_OS_IS_L2COMPRESSIBLE(os), &zp, dmu_objset_write_ready, | |
e8b96c60 | 1109 | NULL, dmu_objset_write_done, os, ZIO_PRIORITY_ASYNC_WRITE, |
3a17a7a9 | 1110 | ZIO_FLAG_MUSTSUCCEED, &zb); |
34dc7c2f BB |
1111 | |
1112 | /* | |
9babb374 | 1113 | * Sync special dnodes - the parent IO for the sync is the root block |
34dc7c2f | 1114 | */ |
572e2857 BB |
1115 | DMU_META_DNODE(os)->dn_zio = zio; |
1116 | dnode_sync(DMU_META_DNODE(os), tx); | |
34dc7c2f | 1117 | |
9babb374 BB |
1118 | os->os_phys->os_flags = os->os_flags; |
1119 | ||
572e2857 BB |
1120 | if (DMU_USERUSED_DNODE(os) && |
1121 | DMU_USERUSED_DNODE(os)->dn_type != DMU_OT_NONE) { | |
1122 | DMU_USERUSED_DNODE(os)->dn_zio = zio; | |
1123 | dnode_sync(DMU_USERUSED_DNODE(os), tx); | |
1124 | DMU_GROUPUSED_DNODE(os)->dn_zio = zio; | |
1125 | dnode_sync(DMU_GROUPUSED_DNODE(os), tx); | |
9babb374 BB |
1126 | } |
1127 | ||
34dc7c2f BB |
1128 | txgoff = tx->tx_txg & TXG_MASK; |
1129 | ||
9babb374 BB |
1130 | if (dmu_objset_userused_enabled(os)) { |
1131 | newlist = &os->os_synced_dnodes; | |
1132 | /* | |
1133 | * We must create the list here because it uses the | |
1134 | * dn_dirty_link[] of this txg. | |
1135 | */ | |
1136 | list_create(newlist, sizeof (dnode_t), | |
1137 | offsetof(dnode_t, dn_dirty_link[txgoff])); | |
1138 | } | |
1139 | ||
1140 | dmu_objset_sync_dnodes(&os->os_free_dnodes[txgoff], newlist, tx); | |
1141 | dmu_objset_sync_dnodes(&os->os_dirty_dnodes[txgoff], newlist, tx); | |
34dc7c2f | 1142 | |
572e2857 | 1143 | list = &DMU_META_DNODE(os)->dn_dirty_records[txgoff]; |
6f1ffb06 | 1144 | while ((dr = list_head(list))) { |
13fe0198 | 1145 | ASSERT0(dr->dr_dbuf->db_level); |
34dc7c2f BB |
1146 | list_remove(list, dr); |
1147 | if (dr->dr_zio) | |
1148 | zio_nowait(dr->dr_zio); | |
1149 | } | |
1150 | /* | |
1151 | * Free intent log blocks up to this tx. | |
1152 | */ | |
1153 | zil_sync(os->os_zil, tx); | |
b128c09f | 1154 | os->os_phys->os_zil_header = os->os_zil_header; |
34dc7c2f BB |
1155 | zio_nowait(zio); |
1156 | } | |
1157 | ||
428870ff BB |
1158 | boolean_t |
1159 | dmu_objset_is_dirty(objset_t *os, uint64_t txg) | |
1160 | { | |
1161 | return (!list_is_empty(&os->os_dirty_dnodes[txg & TXG_MASK]) || | |
1162 | !list_is_empty(&os->os_free_dnodes[txg & TXG_MASK])); | |
1163 | } | |
1164 | ||
572e2857 | 1165 | static objset_used_cb_t *used_cbs[DMU_OST_NUMTYPES]; |
9babb374 BB |
1166 | |
1167 | void | |
1168 | dmu_objset_register_type(dmu_objset_type_t ost, objset_used_cb_t *cb) | |
1169 | { | |
1170 | used_cbs[ost] = cb; | |
1171 | } | |
1172 | ||
1173 | boolean_t | |
428870ff | 1174 | dmu_objset_userused_enabled(objset_t *os) |
9babb374 BB |
1175 | { |
1176 | return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE && | |
572e2857 BB |
1177 | used_cbs[os->os_phys->os_type] != NULL && |
1178 | DMU_USERUSED_DNODE(os) != NULL); | |
9babb374 BB |
1179 | } |
1180 | ||
428870ff BB |
1181 | static void |
1182 | do_userquota_update(objset_t *os, uint64_t used, uint64_t flags, | |
1183 | uint64_t user, uint64_t group, boolean_t subtract, dmu_tx_t *tx) | |
1184 | { | |
1185 | if ((flags & DNODE_FLAG_USERUSED_ACCOUNTED)) { | |
1186 | int64_t delta = DNODE_SIZE + used; | |
1187 | if (subtract) | |
1188 | delta = -delta; | |
1189 | VERIFY3U(0, ==, zap_increment_int(os, DMU_USERUSED_OBJECT, | |
1190 | user, delta, tx)); | |
1191 | VERIFY3U(0, ==, zap_increment_int(os, DMU_GROUPUSED_OBJECT, | |
1192 | group, delta, tx)); | |
1193 | } | |
1194 | } | |
1195 | ||
9babb374 | 1196 | void |
428870ff | 1197 | dmu_objset_do_userquota_updates(objset_t *os, dmu_tx_t *tx) |
9babb374 BB |
1198 | { |
1199 | dnode_t *dn; | |
1200 | list_t *list = &os->os_synced_dnodes; | |
9babb374 BB |
1201 | |
1202 | ASSERT(list_head(list) == NULL || dmu_objset_userused_enabled(os)); | |
1203 | ||
6f1ffb06 | 1204 | while ((dn = list_head(list))) { |
572e2857 | 1205 | int flags; |
9babb374 | 1206 | ASSERT(!DMU_OBJECT_IS_SPECIAL(dn->dn_object)); |
9babb374 BB |
1207 | ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE || |
1208 | dn->dn_phys->dn_flags & | |
1209 | DNODE_FLAG_USERUSED_ACCOUNTED); | |
1210 | ||
1211 | /* Allocate the user/groupused objects if necessary. */ | |
572e2857 | 1212 | if (DMU_USERUSED_DNODE(os)->dn_type == DMU_OT_NONE) { |
428870ff | 1213 | VERIFY(0 == zap_create_claim(os, |
9babb374 BB |
1214 | DMU_USERUSED_OBJECT, |
1215 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
428870ff | 1216 | VERIFY(0 == zap_create_claim(os, |
9babb374 BB |
1217 | DMU_GROUPUSED_OBJECT, |
1218 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
1219 | } | |
1220 | ||
1221 | /* | |
428870ff BB |
1222 | * We intentionally modify the zap object even if the |
1223 | * net delta is zero. Otherwise | |
1224 | * the block of the zap obj could be shared between | |
1225 | * datasets but need to be different between them after | |
1226 | * a bprewrite. | |
9babb374 | 1227 | */ |
9babb374 | 1228 | |
572e2857 BB |
1229 | flags = dn->dn_id_flags; |
1230 | ASSERT(flags); | |
1231 | if (flags & DN_ID_OLD_EXIST) { | |
428870ff BB |
1232 | do_userquota_update(os, dn->dn_oldused, dn->dn_oldflags, |
1233 | dn->dn_olduid, dn->dn_oldgid, B_TRUE, tx); | |
1234 | } | |
572e2857 | 1235 | if (flags & DN_ID_NEW_EXIST) { |
428870ff BB |
1236 | do_userquota_update(os, DN_USED_BYTES(dn->dn_phys), |
1237 | dn->dn_phys->dn_flags, dn->dn_newuid, | |
1238 | dn->dn_newgid, B_FALSE, tx); | |
1239 | } | |
1240 | ||
572e2857 | 1241 | mutex_enter(&dn->dn_mtx); |
428870ff BB |
1242 | dn->dn_oldused = 0; |
1243 | dn->dn_oldflags = 0; | |
1244 | if (dn->dn_id_flags & DN_ID_NEW_EXIST) { | |
1245 | dn->dn_olduid = dn->dn_newuid; | |
1246 | dn->dn_oldgid = dn->dn_newgid; | |
1247 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
1248 | if (dn->dn_bonuslen == 0) | |
1249 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1250 | else | |
1251 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1252 | } | |
1253 | dn->dn_id_flags &= ~(DN_ID_NEW_EXIST); | |
9babb374 BB |
1254 | mutex_exit(&dn->dn_mtx); |
1255 | ||
1256 | list_remove(list, dn); | |
1257 | dnode_rele(dn, list); | |
1258 | } | |
1259 | } | |
1260 | ||
428870ff BB |
1261 | /* |
1262 | * Returns a pointer to data to find uid/gid from | |
1263 | * | |
1264 | * If a dirty record for transaction group that is syncing can't | |
1265 | * be found then NULL is returned. In the NULL case it is assumed | |
1266 | * the uid/gid aren't changing. | |
1267 | */ | |
1268 | static void * | |
1269 | dmu_objset_userquota_find_data(dmu_buf_impl_t *db, dmu_tx_t *tx) | |
1270 | { | |
1271 | dbuf_dirty_record_t *dr, **drp; | |
1272 | void *data; | |
1273 | ||
1274 | if (db->db_dirtycnt == 0) | |
1275 | return (db->db.db_data); /* Nothing is changing */ | |
1276 | ||
1277 | for (drp = &db->db_last_dirty; (dr = *drp) != NULL; drp = &dr->dr_next) | |
1278 | if (dr->dr_txg == tx->tx_txg) | |
1279 | break; | |
1280 | ||
572e2857 | 1281 | if (dr == NULL) { |
428870ff | 1282 | data = NULL; |
572e2857 BB |
1283 | } else { |
1284 | dnode_t *dn; | |
1285 | ||
1286 | DB_DNODE_ENTER(dr->dr_dbuf); | |
1287 | dn = DB_DNODE(dr->dr_dbuf); | |
1288 | ||
1289 | if (dn->dn_bonuslen == 0 && | |
1290 | dr->dr_dbuf->db_blkid == DMU_SPILL_BLKID) | |
1291 | data = dr->dt.dl.dr_data->b_data; | |
1292 | else | |
1293 | data = dr->dt.dl.dr_data; | |
1294 | ||
1295 | DB_DNODE_EXIT(dr->dr_dbuf); | |
1296 | } | |
1297 | ||
428870ff BB |
1298 | return (data); |
1299 | } | |
1300 | ||
1301 | void | |
1302 | dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) | |
1303 | { | |
1304 | objset_t *os = dn->dn_objset; | |
1305 | void *data = NULL; | |
1306 | dmu_buf_impl_t *db = NULL; | |
a117a6d6 GW |
1307 | uint64_t *user = NULL; |
1308 | uint64_t *group = NULL; | |
428870ff BB |
1309 | int flags = dn->dn_id_flags; |
1310 | int error; | |
1311 | boolean_t have_spill = B_FALSE; | |
1312 | ||
1313 | if (!dmu_objset_userused_enabled(dn->dn_objset)) | |
1314 | return; | |
1315 | ||
1316 | if (before && (flags & (DN_ID_CHKED_BONUS|DN_ID_OLD_EXIST| | |
1317 | DN_ID_CHKED_SPILL))) | |
1318 | return; | |
1319 | ||
1320 | if (before && dn->dn_bonuslen != 0) | |
1321 | data = DN_BONUS(dn->dn_phys); | |
1322 | else if (!before && dn->dn_bonuslen != 0) { | |
a3000f93 BB |
1323 | if (dn->dn_bonus) { |
1324 | db = dn->dn_bonus; | |
428870ff BB |
1325 | mutex_enter(&db->db_mtx); |
1326 | data = dmu_objset_userquota_find_data(db, tx); | |
1327 | } else { | |
1328 | data = DN_BONUS(dn->dn_phys); | |
1329 | } | |
1330 | } else if (dn->dn_bonuslen == 0 && dn->dn_bonustype == DMU_OT_SA) { | |
1331 | int rf = 0; | |
1332 | ||
1333 | if (RW_WRITE_HELD(&dn->dn_struct_rwlock)) | |
1334 | rf |= DB_RF_HAVESTRUCT; | |
572e2857 BB |
1335 | error = dmu_spill_hold_by_dnode(dn, |
1336 | rf | DB_RF_MUST_SUCCEED, | |
428870ff BB |
1337 | FTAG, (dmu_buf_t **)&db); |
1338 | ASSERT(error == 0); | |
1339 | mutex_enter(&db->db_mtx); | |
1340 | data = (before) ? db->db.db_data : | |
1341 | dmu_objset_userquota_find_data(db, tx); | |
1342 | have_spill = B_TRUE; | |
1343 | } else { | |
1344 | mutex_enter(&dn->dn_mtx); | |
1345 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1346 | mutex_exit(&dn->dn_mtx); | |
1347 | return; | |
1348 | } | |
1349 | ||
1350 | if (before) { | |
1351 | ASSERT(data); | |
1352 | user = &dn->dn_olduid; | |
1353 | group = &dn->dn_oldgid; | |
1354 | } else if (data) { | |
1355 | user = &dn->dn_newuid; | |
1356 | group = &dn->dn_newgid; | |
1357 | } | |
1358 | ||
1359 | /* | |
1360 | * Must always call the callback in case the object | |
1361 | * type has changed and that type isn't an object type to track | |
1362 | */ | |
1363 | error = used_cbs[os->os_phys->os_type](dn->dn_bonustype, data, | |
1364 | user, group); | |
1365 | ||
1366 | /* | |
1367 | * Preserve existing uid/gid when the callback can't determine | |
1368 | * what the new uid/gid are and the callback returned EEXIST. | |
1369 | * The EEXIST error tells us to just use the existing uid/gid. | |
1370 | * If we don't know what the old values are then just assign | |
1371 | * them to 0, since that is a new file being created. | |
1372 | */ | |
1373 | if (!before && data == NULL && error == EEXIST) { | |
1374 | if (flags & DN_ID_OLD_EXIST) { | |
1375 | dn->dn_newuid = dn->dn_olduid; | |
1376 | dn->dn_newgid = dn->dn_oldgid; | |
1377 | } else { | |
1378 | dn->dn_newuid = 0; | |
1379 | dn->dn_newgid = 0; | |
1380 | } | |
1381 | error = 0; | |
1382 | } | |
1383 | ||
1384 | if (db) | |
1385 | mutex_exit(&db->db_mtx); | |
1386 | ||
1387 | mutex_enter(&dn->dn_mtx); | |
1388 | if (error == 0 && before) | |
1389 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
1390 | if (error == 0 && !before) | |
1391 | dn->dn_id_flags |= DN_ID_NEW_EXIST; | |
1392 | ||
1393 | if (have_spill) { | |
1394 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1395 | } else { | |
1396 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1397 | } | |
1398 | mutex_exit(&dn->dn_mtx); | |
a3000f93 | 1399 | if (have_spill) |
428870ff BB |
1400 | dmu_buf_rele((dmu_buf_t *)db, FTAG); |
1401 | } | |
1402 | ||
9babb374 BB |
1403 | boolean_t |
1404 | dmu_objset_userspace_present(objset_t *os) | |
1405 | { | |
428870ff | 1406 | return (os->os_phys->os_flags & |
9babb374 BB |
1407 | OBJSET_FLAG_USERACCOUNTING_COMPLETE); |
1408 | } | |
1409 | ||
1410 | int | |
1411 | dmu_objset_userspace_upgrade(objset_t *os) | |
1412 | { | |
1413 | uint64_t obj; | |
1414 | int err = 0; | |
1415 | ||
1416 | if (dmu_objset_userspace_present(os)) | |
1417 | return (0); | |
428870ff | 1418 | if (!dmu_objset_userused_enabled(os)) |
2e528b49 | 1419 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 1420 | if (dmu_objset_is_snapshot(os)) |
2e528b49 | 1421 | return (SET_ERROR(EINVAL)); |
9babb374 BB |
1422 | |
1423 | /* | |
1424 | * We simply need to mark every object dirty, so that it will be | |
1425 | * synced out and now accounted. If this is called | |
1426 | * concurrently, or if we already did some work before crashing, | |
1427 | * that's fine, since we track each object's accounted state | |
1428 | * independently. | |
1429 | */ | |
1430 | ||
1431 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
45d1cae3 | 1432 | dmu_tx_t *tx; |
9babb374 BB |
1433 | dmu_buf_t *db; |
1434 | int objerr; | |
1435 | ||
1436 | if (issig(JUSTLOOKING) && issig(FORREAL)) | |
2e528b49 | 1437 | return (SET_ERROR(EINTR)); |
9babb374 BB |
1438 | |
1439 | objerr = dmu_bonus_hold(os, obj, FTAG, &db); | |
13fe0198 | 1440 | if (objerr != 0) |
9babb374 | 1441 | continue; |
45d1cae3 | 1442 | tx = dmu_tx_create(os); |
9babb374 BB |
1443 | dmu_tx_hold_bonus(tx, obj); |
1444 | objerr = dmu_tx_assign(tx, TXG_WAIT); | |
13fe0198 | 1445 | if (objerr != 0) { |
9babb374 BB |
1446 | dmu_tx_abort(tx); |
1447 | continue; | |
1448 | } | |
1449 | dmu_buf_will_dirty(db, tx); | |
1450 | dmu_buf_rele(db, FTAG); | |
1451 | dmu_tx_commit(tx); | |
1452 | } | |
1453 | ||
428870ff | 1454 | os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; |
9babb374 BB |
1455 | txg_wait_synced(dmu_objset_pool(os), 0); |
1456 | return (0); | |
1457 | } | |
1458 | ||
34dc7c2f BB |
1459 | void |
1460 | dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, | |
1461 | uint64_t *usedobjsp, uint64_t *availobjsp) | |
1462 | { | |
428870ff | 1463 | dsl_dataset_space(os->os_dsl_dataset, refdbytesp, availbytesp, |
34dc7c2f BB |
1464 | usedobjsp, availobjsp); |
1465 | } | |
1466 | ||
1467 | uint64_t | |
1468 | dmu_objset_fsid_guid(objset_t *os) | |
1469 | { | |
428870ff | 1470 | return (dsl_dataset_fsid_guid(os->os_dsl_dataset)); |
34dc7c2f BB |
1471 | } |
1472 | ||
1473 | void | |
1474 | dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat) | |
1475 | { | |
428870ff BB |
1476 | stat->dds_type = os->os_phys->os_type; |
1477 | if (os->os_dsl_dataset) | |
1478 | dsl_dataset_fast_stat(os->os_dsl_dataset, stat); | |
34dc7c2f BB |
1479 | } |
1480 | ||
1481 | void | |
1482 | dmu_objset_stats(objset_t *os, nvlist_t *nv) | |
1483 | { | |
428870ff BB |
1484 | ASSERT(os->os_dsl_dataset || |
1485 | os->os_phys->os_type == DMU_OST_META); | |
34dc7c2f | 1486 | |
428870ff BB |
1487 | if (os->os_dsl_dataset != NULL) |
1488 | dsl_dataset_stats(os->os_dsl_dataset, nv); | |
34dc7c2f BB |
1489 | |
1490 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_TYPE, | |
428870ff | 1491 | os->os_phys->os_type); |
9babb374 BB |
1492 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERACCOUNTING, |
1493 | dmu_objset_userspace_present(os)); | |
34dc7c2f BB |
1494 | } |
1495 | ||
1496 | int | |
1497 | dmu_objset_is_snapshot(objset_t *os) | |
1498 | { | |
428870ff | 1499 | if (os->os_dsl_dataset != NULL) |
0c66c32d | 1500 | return (os->os_dsl_dataset->ds_is_snapshot); |
34dc7c2f BB |
1501 | else |
1502 | return (B_FALSE); | |
1503 | } | |
1504 | ||
1505 | int | |
1506 | dmu_snapshot_realname(objset_t *os, char *name, char *real, int maxlen, | |
1507 | boolean_t *conflict) | |
1508 | { | |
428870ff | 1509 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
1510 | uint64_t ignored; |
1511 | ||
d683ddbb | 1512 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 1513 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1514 | |
1515 | return (zap_lookup_norm(ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb JG |
1516 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, name, 8, 1, &ignored, |
1517 | MT_FIRST, real, maxlen, conflict)); | |
34dc7c2f BB |
1518 | } |
1519 | ||
1520 | int | |
1521 | dmu_snapshot_list_next(objset_t *os, int namelen, char *name, | |
1522 | uint64_t *idp, uint64_t *offp, boolean_t *case_conflict) | |
1523 | { | |
428870ff | 1524 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
1525 | zap_cursor_t cursor; |
1526 | zap_attribute_t attr; | |
1527 | ||
13fe0198 MA |
1528 | ASSERT(dsl_pool_config_held(dmu_objset_pool(os))); |
1529 | ||
d683ddbb | 1530 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 1531 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1532 | |
1533 | zap_cursor_init_serialized(&cursor, | |
1534 | ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb | 1535 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, *offp); |
34dc7c2f BB |
1536 | |
1537 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
1538 | zap_cursor_fini(&cursor); | |
2e528b49 | 1539 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1540 | } |
1541 | ||
1542 | if (strlen(attr.za_name) + 1 > namelen) { | |
1543 | zap_cursor_fini(&cursor); | |
2e528b49 | 1544 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
1545 | } |
1546 | ||
1547 | (void) strcpy(name, attr.za_name); | |
1548 | if (idp) | |
1549 | *idp = attr.za_first_integer; | |
1550 | if (case_conflict) | |
1551 | *case_conflict = attr.za_normalization_conflict; | |
1552 | zap_cursor_advance(&cursor); | |
1553 | *offp = zap_cursor_serialize(&cursor); | |
1554 | zap_cursor_fini(&cursor); | |
1555 | ||
1556 | return (0); | |
1557 | } | |
1558 | ||
ebe7e575 | 1559 | int |
6772fb67 | 1560 | dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *value) |
ebe7e575 | 1561 | { |
d1d7e268 | 1562 | return (dsl_dataset_snap_lookup(os->os_dsl_dataset, name, value)); |
ebe7e575 BB |
1563 | } |
1564 | ||
34dc7c2f BB |
1565 | int |
1566 | dmu_dir_list_next(objset_t *os, int namelen, char *name, | |
1567 | uint64_t *idp, uint64_t *offp) | |
1568 | { | |
428870ff | 1569 | dsl_dir_t *dd = os->os_dsl_dataset->ds_dir; |
34dc7c2f BB |
1570 | zap_cursor_t cursor; |
1571 | zap_attribute_t attr; | |
1572 | ||
1573 | /* there is no next dir on a snapshot! */ | |
428870ff | 1574 | if (os->os_dsl_dataset->ds_object != |
d683ddbb | 1575 | dsl_dir_phys(dd)->dd_head_dataset_obj) |
2e528b49 | 1576 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1577 | |
1578 | zap_cursor_init_serialized(&cursor, | |
1579 | dd->dd_pool->dp_meta_objset, | |
d683ddbb | 1580 | dsl_dir_phys(dd)->dd_child_dir_zapobj, *offp); |
34dc7c2f BB |
1581 | |
1582 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
1583 | zap_cursor_fini(&cursor); | |
2e528b49 | 1584 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1585 | } |
1586 | ||
1587 | if (strlen(attr.za_name) + 1 > namelen) { | |
1588 | zap_cursor_fini(&cursor); | |
2e528b49 | 1589 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
1590 | } |
1591 | ||
1592 | (void) strcpy(name, attr.za_name); | |
1593 | if (idp) | |
1594 | *idp = attr.za_first_integer; | |
1595 | zap_cursor_advance(&cursor); | |
1596 | *offp = zap_cursor_serialize(&cursor); | |
1597 | zap_cursor_fini(&cursor); | |
1598 | ||
1599 | return (0); | |
1600 | } | |
1601 | ||
9c43027b AJ |
1602 | typedef struct dmu_objset_find_ctx { |
1603 | taskq_t *dc_tq; | |
1604 | dsl_pool_t *dc_dp; | |
1605 | uint64_t dc_ddobj; | |
1606 | int (*dc_func)(dsl_pool_t *, dsl_dataset_t *, void *); | |
1607 | void *dc_arg; | |
1608 | int dc_flags; | |
1609 | kmutex_t *dc_error_lock; | |
1610 | int *dc_error; | |
1611 | } dmu_objset_find_ctx_t; | |
1612 | ||
1613 | static void | |
1614 | dmu_objset_find_dp_impl(dmu_objset_find_ctx_t *dcp) | |
b128c09f | 1615 | { |
9c43027b AJ |
1616 | dsl_pool_t *dp = dcp->dc_dp; |
1617 | dmu_objset_find_ctx_t *child_dcp; | |
13fe0198 MA |
1618 | dsl_dir_t *dd; |
1619 | dsl_dataset_t *ds; | |
1620 | zap_cursor_t zc; | |
1621 | zap_attribute_t *attr; | |
1622 | uint64_t thisobj; | |
9c43027b | 1623 | int err = 0; |
13fe0198 | 1624 | |
9c43027b AJ |
1625 | /* don't process if there already was an error */ |
1626 | if (*dcp->dc_error != 0) | |
1627 | goto out; | |
13fe0198 | 1628 | |
9c43027b | 1629 | err = dsl_dir_hold_obj(dp, dcp->dc_ddobj, NULL, FTAG, &dd); |
13fe0198 | 1630 | if (err != 0) |
9c43027b | 1631 | goto out; |
13fe0198 MA |
1632 | |
1633 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ | |
1634 | if (dd->dd_myname[0] == '$') { | |
1635 | dsl_dir_rele(dd, FTAG); | |
9c43027b | 1636 | goto out; |
13fe0198 MA |
1637 | } |
1638 | ||
d683ddbb | 1639 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 1640 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
13fe0198 MA |
1641 | |
1642 | /* | |
1643 | * Iterate over all children. | |
1644 | */ | |
9c43027b | 1645 | if (dcp->dc_flags & DS_FIND_CHILDREN) { |
13fe0198 | 1646 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
d683ddbb | 1647 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
13fe0198 MA |
1648 | zap_cursor_retrieve(&zc, attr) == 0; |
1649 | (void) zap_cursor_advance(&zc)) { | |
1650 | ASSERT3U(attr->za_integer_length, ==, | |
1651 | sizeof (uint64_t)); | |
1652 | ASSERT3U(attr->za_num_integers, ==, 1); | |
1653 | ||
9c43027b AJ |
1654 | child_dcp = kmem_alloc(sizeof (*child_dcp), KM_SLEEP); |
1655 | *child_dcp = *dcp; | |
1656 | child_dcp->dc_ddobj = attr->za_first_integer; | |
1657 | if (dcp->dc_tq != NULL) | |
1658 | (void) taskq_dispatch(dcp->dc_tq, | |
1659 | dmu_objset_find_dp_cb, child_dcp, TQ_SLEEP); | |
1660 | else | |
1661 | dmu_objset_find_dp_impl(child_dcp); | |
13fe0198 MA |
1662 | } |
1663 | zap_cursor_fini(&zc); | |
13fe0198 MA |
1664 | } |
1665 | ||
1666 | /* | |
1667 | * Iterate over all snapshots. | |
1668 | */ | |
9c43027b | 1669 | if (dcp->dc_flags & DS_FIND_SNAPSHOTS) { |
13fe0198 MA |
1670 | dsl_dataset_t *ds; |
1671 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1672 | ||
1673 | if (err == 0) { | |
d683ddbb JG |
1674 | uint64_t snapobj; |
1675 | ||
1676 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
13fe0198 MA |
1677 | dsl_dataset_rele(ds, FTAG); |
1678 | ||
1679 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
1680 | zap_cursor_retrieve(&zc, attr) == 0; | |
1681 | (void) zap_cursor_advance(&zc)) { | |
1682 | ASSERT3U(attr->za_integer_length, ==, | |
1683 | sizeof (uint64_t)); | |
1684 | ASSERT3U(attr->za_num_integers, ==, 1); | |
1685 | ||
1686 | err = dsl_dataset_hold_obj(dp, | |
1687 | attr->za_first_integer, FTAG, &ds); | |
1688 | if (err != 0) | |
1689 | break; | |
9c43027b | 1690 | err = dcp->dc_func(dp, ds, dcp->dc_arg); |
13fe0198 MA |
1691 | dsl_dataset_rele(ds, FTAG); |
1692 | if (err != 0) | |
1693 | break; | |
1694 | } | |
1695 | zap_cursor_fini(&zc); | |
1696 | } | |
1697 | } | |
1698 | ||
1699 | dsl_dir_rele(dd, FTAG); | |
1700 | kmem_free(attr, sizeof (zap_attribute_t)); | |
1701 | ||
1702 | if (err != 0) | |
9c43027b | 1703 | goto out; |
13fe0198 MA |
1704 | |
1705 | /* | |
1706 | * Apply to self. | |
1707 | */ | |
1708 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1709 | if (err != 0) | |
9c43027b AJ |
1710 | goto out; |
1711 | err = dcp->dc_func(dp, ds, dcp->dc_arg); | |
13fe0198 | 1712 | dsl_dataset_rele(ds, FTAG); |
9c43027b AJ |
1713 | |
1714 | out: | |
1715 | if (err != 0) { | |
1716 | mutex_enter(dcp->dc_error_lock); | |
1717 | /* only keep first error */ | |
1718 | if (*dcp->dc_error == 0) | |
1719 | *dcp->dc_error = err; | |
1720 | mutex_exit(dcp->dc_error_lock); | |
1721 | } | |
1722 | ||
1723 | kmem_free(dcp, sizeof (*dcp)); | |
1724 | } | |
1725 | ||
1726 | static void | |
1727 | dmu_objset_find_dp_cb(void *arg) | |
1728 | { | |
1729 | dmu_objset_find_ctx_t *dcp = arg; | |
1730 | dsl_pool_t *dp = dcp->dc_dp; | |
1731 | ||
5e8cd5d1 AJ |
1732 | /* |
1733 | * We need to get a pool_config_lock here, as there are several | |
1734 | * asssert(pool_config_held) down the stack. Getting a lock via | |
1735 | * dsl_pool_config_enter is risky, as it might be stalled by a | |
1736 | * pending writer. This would deadlock, as the write lock can | |
1737 | * only be granted when our parent thread gives up the lock. | |
1738 | * The _prio interface gives us priority over a pending writer. | |
1739 | */ | |
1740 | dsl_pool_config_enter_prio(dp, FTAG); | |
9c43027b AJ |
1741 | |
1742 | dmu_objset_find_dp_impl(dcp); | |
1743 | ||
1744 | dsl_pool_config_exit(dp, FTAG); | |
1745 | } | |
1746 | ||
1747 | /* | |
1748 | * Find objsets under and including ddobj, call func(ds) on each. | |
1749 | * The order for the enumeration is completely undefined. | |
1750 | * func is called with dsl_pool_config held. | |
1751 | */ | |
1752 | int | |
1753 | dmu_objset_find_dp(dsl_pool_t *dp, uint64_t ddobj, | |
1754 | int func(dsl_pool_t *, dsl_dataset_t *, void *), void *arg, int flags) | |
1755 | { | |
1756 | int error = 0; | |
1757 | taskq_t *tq = NULL; | |
1758 | int ntasks; | |
1759 | dmu_objset_find_ctx_t *dcp; | |
1760 | kmutex_t err_lock; | |
1761 | ||
1762 | mutex_init(&err_lock, NULL, MUTEX_DEFAULT, NULL); | |
1763 | dcp = kmem_alloc(sizeof (*dcp), KM_SLEEP); | |
1764 | dcp->dc_tq = NULL; | |
1765 | dcp->dc_dp = dp; | |
1766 | dcp->dc_ddobj = ddobj; | |
1767 | dcp->dc_func = func; | |
1768 | dcp->dc_arg = arg; | |
1769 | dcp->dc_flags = flags; | |
1770 | dcp->dc_error_lock = &err_lock; | |
1771 | dcp->dc_error = &error; | |
1772 | ||
1773 | if ((flags & DS_FIND_SERIALIZE) || dsl_pool_config_held_writer(dp)) { | |
1774 | /* | |
1775 | * In case a write lock is held we can't make use of | |
1776 | * parallelism, as down the stack of the worker threads | |
1777 | * the lock is asserted via dsl_pool_config_held. | |
1778 | * In case of a read lock this is solved by getting a read | |
1779 | * lock in each worker thread, which isn't possible in case | |
1780 | * of a writer lock. So we fall back to the synchronous path | |
1781 | * here. | |
1782 | * In the future it might be possible to get some magic into | |
1783 | * dsl_pool_config_held in a way that it returns true for | |
1784 | * the worker threads so that a single lock held from this | |
1785 | * thread suffices. For now, stay single threaded. | |
1786 | */ | |
1787 | dmu_objset_find_dp_impl(dcp); | |
e5676636 | 1788 | mutex_destroy(&err_lock); |
9c43027b AJ |
1789 | |
1790 | return (error); | |
1791 | } | |
1792 | ||
1793 | ntasks = dmu_find_threads; | |
1794 | if (ntasks == 0) | |
1795 | ntasks = vdev_count_leaves(dp->dp_spa) * 4; | |
1229323d | 1796 | tq = taskq_create("dmu_objset_find", ntasks, maxclsyspri, ntasks, |
9c43027b AJ |
1797 | INT_MAX, 0); |
1798 | if (tq == NULL) { | |
1799 | kmem_free(dcp, sizeof (*dcp)); | |
e5676636 BB |
1800 | mutex_destroy(&err_lock); |
1801 | ||
9c43027b AJ |
1802 | return (SET_ERROR(ENOMEM)); |
1803 | } | |
1804 | dcp->dc_tq = tq; | |
1805 | ||
1806 | /* dcp will be freed by task */ | |
1807 | (void) taskq_dispatch(tq, dmu_objset_find_dp_cb, dcp, TQ_SLEEP); | |
1808 | ||
1809 | /* | |
1810 | * PORTING: this code relies on the property of taskq_wait to wait | |
1811 | * until no more tasks are queued and no more tasks are active. As | |
1812 | * we always queue new tasks from within other tasks, task_wait | |
1813 | * reliably waits for the full recursion to finish, even though we | |
1814 | * enqueue new tasks after taskq_wait has been called. | |
1815 | * On platforms other than illumos, taskq_wait may not have this | |
1816 | * property. | |
1817 | */ | |
1818 | taskq_wait(tq); | |
1819 | taskq_destroy(tq); | |
1820 | mutex_destroy(&err_lock); | |
1821 | ||
1822 | return (error); | |
b128c09f BB |
1823 | } |
1824 | ||
1825 | /* | |
13fe0198 MA |
1826 | * Find all objsets under name, and for each, call 'func(child_name, arg)'. |
1827 | * The dp_config_rwlock must not be held when this is called, and it | |
1828 | * will not be held when the callback is called. | |
1829 | * Therefore this function should only be used when the pool is not changing | |
1830 | * (e.g. in syncing context), or the callback can deal with the possible races. | |
b128c09f | 1831 | */ |
13fe0198 MA |
1832 | static int |
1833 | dmu_objset_find_impl(spa_t *spa, const char *name, | |
1834 | int func(const char *, void *), void *arg, int flags) | |
34dc7c2f BB |
1835 | { |
1836 | dsl_dir_t *dd; | |
13fe0198 | 1837 | dsl_pool_t *dp = spa_get_dsl(spa); |
b128c09f | 1838 | dsl_dataset_t *ds; |
34dc7c2f BB |
1839 | zap_cursor_t zc; |
1840 | zap_attribute_t *attr; | |
1841 | char *child; | |
b128c09f BB |
1842 | uint64_t thisobj; |
1843 | int err; | |
34dc7c2f | 1844 | |
13fe0198 MA |
1845 | dsl_pool_config_enter(dp, FTAG); |
1846 | ||
1847 | err = dsl_dir_hold(dp, name, FTAG, &dd, NULL); | |
1848 | if (err != 0) { | |
1849 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f | 1850 | return (err); |
13fe0198 | 1851 | } |
34dc7c2f | 1852 | |
b128c09f BB |
1853 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ |
1854 | if (dd->dd_myname[0] == '$') { | |
13fe0198 MA |
1855 | dsl_dir_rele(dd, FTAG); |
1856 | dsl_pool_config_exit(dp, FTAG); | |
b128c09f BB |
1857 | return (0); |
1858 | } | |
1859 | ||
d683ddbb | 1860 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 1861 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
34dc7c2f BB |
1862 | |
1863 | /* | |
1864 | * Iterate over all children. | |
1865 | */ | |
1866 | if (flags & DS_FIND_CHILDREN) { | |
b128c09f | 1867 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
d683ddbb | 1868 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
34dc7c2f BB |
1869 | zap_cursor_retrieve(&zc, attr) == 0; |
1870 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 MA |
1871 | ASSERT3U(attr->za_integer_length, ==, |
1872 | sizeof (uint64_t)); | |
1873 | ASSERT3U(attr->za_num_integers, ==, 1); | |
34dc7c2f | 1874 | |
428870ff | 1875 | child = kmem_asprintf("%s/%s", name, attr->za_name); |
13fe0198 MA |
1876 | dsl_pool_config_exit(dp, FTAG); |
1877 | err = dmu_objset_find_impl(spa, child, | |
1878 | func, arg, flags); | |
1879 | dsl_pool_config_enter(dp, FTAG); | |
428870ff | 1880 | strfree(child); |
13fe0198 | 1881 | if (err != 0) |
34dc7c2f BB |
1882 | break; |
1883 | } | |
1884 | zap_cursor_fini(&zc); | |
1885 | ||
13fe0198 MA |
1886 | if (err != 0) { |
1887 | dsl_dir_rele(dd, FTAG); | |
1888 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f BB |
1889 | kmem_free(attr, sizeof (zap_attribute_t)); |
1890 | return (err); | |
1891 | } | |
1892 | } | |
1893 | ||
1894 | /* | |
1895 | * Iterate over all snapshots. | |
1896 | */ | |
b128c09f | 1897 | if (flags & DS_FIND_SNAPSHOTS) { |
b128c09f | 1898 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); |
b128c09f BB |
1899 | |
1900 | if (err == 0) { | |
d683ddbb JG |
1901 | uint64_t snapobj; |
1902 | ||
1903 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
b128c09f BB |
1904 | dsl_dataset_rele(ds, FTAG); |
1905 | ||
1906 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
1907 | zap_cursor_retrieve(&zc, attr) == 0; | |
1908 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 | 1909 | ASSERT3U(attr->za_integer_length, ==, |
b128c09f | 1910 | sizeof (uint64_t)); |
13fe0198 | 1911 | ASSERT3U(attr->za_num_integers, ==, 1); |
b128c09f | 1912 | |
428870ff BB |
1913 | child = kmem_asprintf("%s@%s", |
1914 | name, attr->za_name); | |
13fe0198 MA |
1915 | dsl_pool_config_exit(dp, FTAG); |
1916 | err = func(child, arg); | |
1917 | dsl_pool_config_enter(dp, FTAG); | |
428870ff | 1918 | strfree(child); |
13fe0198 | 1919 | if (err != 0) |
b128c09f BB |
1920 | break; |
1921 | } | |
1922 | zap_cursor_fini(&zc); | |
34dc7c2f | 1923 | } |
34dc7c2f BB |
1924 | } |
1925 | ||
13fe0198 | 1926 | dsl_dir_rele(dd, FTAG); |
34dc7c2f | 1927 | kmem_free(attr, sizeof (zap_attribute_t)); |
13fe0198 | 1928 | dsl_pool_config_exit(dp, FTAG); |
34dc7c2f | 1929 | |
13fe0198 | 1930 | if (err != 0) |
34dc7c2f BB |
1931 | return (err); |
1932 | ||
13fe0198 MA |
1933 | /* Apply to self. */ |
1934 | return (func(name, arg)); | |
34dc7c2f BB |
1935 | } |
1936 | ||
13fe0198 MA |
1937 | /* |
1938 | * See comment above dmu_objset_find_impl(). | |
1939 | */ | |
d164b209 | 1940 | int |
13fe0198 MA |
1941 | dmu_objset_find(char *name, int func(const char *, void *), void *arg, |
1942 | int flags) | |
d164b209 | 1943 | { |
13fe0198 MA |
1944 | spa_t *spa; |
1945 | int error; | |
d164b209 | 1946 | |
13fe0198 MA |
1947 | error = spa_open(name, &spa, FTAG); |
1948 | if (error != 0) | |
1949 | return (error); | |
1950 | error = dmu_objset_find_impl(spa, name, func, arg, flags); | |
1951 | spa_close(spa, FTAG); | |
1952 | return (error); | |
d164b209 BB |
1953 | } |
1954 | ||
34dc7c2f BB |
1955 | void |
1956 | dmu_objset_set_user(objset_t *os, void *user_ptr) | |
1957 | { | |
428870ff BB |
1958 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
1959 | os->os_user_ptr = user_ptr; | |
34dc7c2f BB |
1960 | } |
1961 | ||
1962 | void * | |
1963 | dmu_objset_get_user(objset_t *os) | |
1964 | { | |
428870ff BB |
1965 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
1966 | return (os->os_user_ptr); | |
34dc7c2f | 1967 | } |
c28b2279 | 1968 | |
13fe0198 MA |
1969 | /* |
1970 | * Determine name of filesystem, given name of snapshot. | |
1971 | * buf must be at least MAXNAMELEN bytes | |
1972 | */ | |
1973 | int | |
1974 | dmu_fsname(const char *snapname, char *buf) | |
1975 | { | |
1976 | char *atp = strchr(snapname, '@'); | |
1977 | if (atp == NULL) | |
2e528b49 | 1978 | return (SET_ERROR(EINVAL)); |
13fe0198 | 1979 | if (atp - snapname >= MAXNAMELEN) |
2e528b49 | 1980 | return (SET_ERROR(ENAMETOOLONG)); |
13fe0198 MA |
1981 | (void) strlcpy(buf, snapname, atp - snapname + 1); |
1982 | return (0); | |
1983 | } | |
1984 | ||
c28b2279 | 1985 | #if defined(_KERNEL) && defined(HAVE_SPL) |
f0fd83be | 1986 | EXPORT_SYMBOL(dmu_objset_zil); |
c28b2279 | 1987 | EXPORT_SYMBOL(dmu_objset_pool); |
f0fd83be BB |
1988 | EXPORT_SYMBOL(dmu_objset_ds); |
1989 | EXPORT_SYMBOL(dmu_objset_type); | |
c28b2279 BB |
1990 | EXPORT_SYMBOL(dmu_objset_name); |
1991 | EXPORT_SYMBOL(dmu_objset_hold); | |
1992 | EXPORT_SYMBOL(dmu_objset_own); | |
1993 | EXPORT_SYMBOL(dmu_objset_rele); | |
1994 | EXPORT_SYMBOL(dmu_objset_disown); | |
1995 | EXPORT_SYMBOL(dmu_objset_from_ds); | |
1996 | EXPORT_SYMBOL(dmu_objset_create); | |
1997 | EXPORT_SYMBOL(dmu_objset_clone); | |
c28b2279 BB |
1998 | EXPORT_SYMBOL(dmu_objset_stats); |
1999 | EXPORT_SYMBOL(dmu_objset_fast_stat); | |
54a179e7 | 2000 | EXPORT_SYMBOL(dmu_objset_spa); |
c28b2279 BB |
2001 | EXPORT_SYMBOL(dmu_objset_space); |
2002 | EXPORT_SYMBOL(dmu_objset_fsid_guid); | |
2003 | EXPORT_SYMBOL(dmu_objset_find); | |
c28b2279 BB |
2004 | EXPORT_SYMBOL(dmu_objset_byteswap); |
2005 | EXPORT_SYMBOL(dmu_objset_evict_dbufs); | |
2006 | EXPORT_SYMBOL(dmu_objset_snap_cmtime); | |
2007 | ||
2008 | EXPORT_SYMBOL(dmu_objset_sync); | |
2009 | EXPORT_SYMBOL(dmu_objset_is_dirty); | |
2010 | EXPORT_SYMBOL(dmu_objset_create_impl); | |
2011 | EXPORT_SYMBOL(dmu_objset_open_impl); | |
2012 | EXPORT_SYMBOL(dmu_objset_evict); | |
2013 | EXPORT_SYMBOL(dmu_objset_register_type); | |
2014 | EXPORT_SYMBOL(dmu_objset_do_userquota_updates); | |
2015 | EXPORT_SYMBOL(dmu_objset_userquota_get_ids); | |
2016 | EXPORT_SYMBOL(dmu_objset_userused_enabled); | |
2017 | EXPORT_SYMBOL(dmu_objset_userspace_upgrade); | |
2018 | EXPORT_SYMBOL(dmu_objset_userspace_present); | |
2019 | #endif |