]>
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. |
34dc7c2f BB |
26 | */ |
27 | ||
428870ff BB |
28 | /* Portions Copyright 2010 Robert Milkowski */ |
29 | ||
34dc7c2f BB |
30 | #include <sys/cred.h> |
31 | #include <sys/zfs_context.h> | |
32 | #include <sys/dmu_objset.h> | |
33 | #include <sys/dsl_dir.h> | |
34 | #include <sys/dsl_dataset.h> | |
35 | #include <sys/dsl_prop.h> | |
36 | #include <sys/dsl_pool.h> | |
37 | #include <sys/dsl_synctask.h> | |
38 | #include <sys/dsl_deleg.h> | |
39 | #include <sys/dnode.h> | |
40 | #include <sys/dbuf.h> | |
41 | #include <sys/zvol.h> | |
42 | #include <sys/dmu_tx.h> | |
34dc7c2f BB |
43 | #include <sys/zap.h> |
44 | #include <sys/zil.h> | |
45 | #include <sys/dmu_impl.h> | |
46 | #include <sys/zfs_ioctl.h> | |
428870ff | 47 | #include <sys/sa.h> |
572e2857 | 48 | #include <sys/zfs_onexit.h> |
13fe0198 | 49 | #include <sys/dsl_destroy.h> |
572e2857 BB |
50 | |
51 | /* | |
52 | * Needed to close a window in dnode_move() that allows the objset to be freed | |
53 | * before it can be safely accessed. | |
54 | */ | |
55 | krwlock_t os_lock; | |
56 | ||
57 | void | |
58 | dmu_objset_init(void) | |
59 | { | |
60 | rw_init(&os_lock, NULL, RW_DEFAULT, NULL); | |
61 | } | |
62 | ||
63 | void | |
64 | dmu_objset_fini(void) | |
65 | { | |
66 | rw_destroy(&os_lock); | |
67 | } | |
34dc7c2f BB |
68 | |
69 | spa_t * | |
70 | dmu_objset_spa(objset_t *os) | |
71 | { | |
428870ff | 72 | return (os->os_spa); |
34dc7c2f BB |
73 | } |
74 | ||
75 | zilog_t * | |
76 | dmu_objset_zil(objset_t *os) | |
77 | { | |
428870ff | 78 | return (os->os_zil); |
34dc7c2f BB |
79 | } |
80 | ||
81 | dsl_pool_t * | |
82 | dmu_objset_pool(objset_t *os) | |
83 | { | |
84 | dsl_dataset_t *ds; | |
85 | ||
428870ff | 86 | if ((ds = os->os_dsl_dataset) != NULL && ds->ds_dir) |
34dc7c2f BB |
87 | return (ds->ds_dir->dd_pool); |
88 | else | |
428870ff | 89 | return (spa_get_dsl(os->os_spa)); |
34dc7c2f BB |
90 | } |
91 | ||
92 | dsl_dataset_t * | |
93 | dmu_objset_ds(objset_t *os) | |
94 | { | |
428870ff | 95 | return (os->os_dsl_dataset); |
34dc7c2f BB |
96 | } |
97 | ||
98 | dmu_objset_type_t | |
99 | dmu_objset_type(objset_t *os) | |
100 | { | |
428870ff | 101 | return (os->os_phys->os_type); |
34dc7c2f BB |
102 | } |
103 | ||
104 | void | |
105 | dmu_objset_name(objset_t *os, char *buf) | |
106 | { | |
428870ff | 107 | dsl_dataset_name(os->os_dsl_dataset, buf); |
34dc7c2f BB |
108 | } |
109 | ||
110 | uint64_t | |
111 | dmu_objset_id(objset_t *os) | |
112 | { | |
428870ff | 113 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
114 | |
115 | return (ds ? ds->ds_object : 0); | |
116 | } | |
117 | ||
faf0f58c | 118 | zfs_sync_type_t |
428870ff BB |
119 | dmu_objset_syncprop(objset_t *os) |
120 | { | |
121 | return (os->os_sync); | |
122 | } | |
123 | ||
faf0f58c | 124 | zfs_logbias_op_t |
428870ff BB |
125 | dmu_objset_logbias(objset_t *os) |
126 | { | |
127 | return (os->os_logbias); | |
128 | } | |
129 | ||
34dc7c2f BB |
130 | static void |
131 | checksum_changed_cb(void *arg, uint64_t newval) | |
132 | { | |
428870ff | 133 | objset_t *os = arg; |
34dc7c2f BB |
134 | |
135 | /* | |
136 | * Inheritance should have been done by now. | |
137 | */ | |
138 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
139 | ||
428870ff | 140 | os->os_checksum = zio_checksum_select(newval, ZIO_CHECKSUM_ON_VALUE); |
34dc7c2f BB |
141 | } |
142 | ||
143 | static void | |
144 | compression_changed_cb(void *arg, uint64_t newval) | |
145 | { | |
428870ff | 146 | objset_t *os = arg; |
34dc7c2f BB |
147 | |
148 | /* | |
149 | * Inheritance and range checking should have been done by now. | |
150 | */ | |
151 | ASSERT(newval != ZIO_COMPRESS_INHERIT); | |
152 | ||
428870ff | 153 | os->os_compress = zio_compress_select(newval, ZIO_COMPRESS_ON_VALUE); |
34dc7c2f BB |
154 | } |
155 | ||
156 | static void | |
157 | copies_changed_cb(void *arg, uint64_t newval) | |
158 | { | |
428870ff | 159 | objset_t *os = arg; |
34dc7c2f BB |
160 | |
161 | /* | |
162 | * Inheritance and range checking should have been done by now. | |
163 | */ | |
164 | ASSERT(newval > 0); | |
428870ff | 165 | ASSERT(newval <= spa_max_replication(os->os_spa)); |
34dc7c2f | 166 | |
428870ff BB |
167 | os->os_copies = newval; |
168 | } | |
169 | ||
170 | static void | |
171 | dedup_changed_cb(void *arg, uint64_t newval) | |
172 | { | |
173 | objset_t *os = arg; | |
174 | spa_t *spa = os->os_spa; | |
175 | enum zio_checksum checksum; | |
176 | ||
177 | /* | |
178 | * Inheritance should have been done by now. | |
179 | */ | |
180 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
181 | ||
182 | checksum = zio_checksum_dedup_select(spa, newval, ZIO_CHECKSUM_OFF); | |
183 | ||
184 | os->os_dedup_checksum = checksum & ZIO_CHECKSUM_MASK; | |
185 | os->os_dedup_verify = !!(checksum & ZIO_CHECKSUM_VERIFY); | |
34dc7c2f BB |
186 | } |
187 | ||
b128c09f BB |
188 | static void |
189 | primary_cache_changed_cb(void *arg, uint64_t newval) | |
190 | { | |
428870ff | 191 | objset_t *os = arg; |
b128c09f BB |
192 | |
193 | /* | |
194 | * Inheritance and range checking should have been done by now. | |
195 | */ | |
196 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
197 | newval == ZFS_CACHE_METADATA); | |
198 | ||
428870ff | 199 | os->os_primary_cache = newval; |
b128c09f BB |
200 | } |
201 | ||
202 | static void | |
203 | secondary_cache_changed_cb(void *arg, uint64_t newval) | |
204 | { | |
428870ff | 205 | objset_t *os = arg; |
b128c09f BB |
206 | |
207 | /* | |
208 | * Inheritance and range checking should have been done by now. | |
209 | */ | |
210 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
211 | newval == ZFS_CACHE_METADATA); | |
212 | ||
428870ff BB |
213 | os->os_secondary_cache = newval; |
214 | } | |
215 | ||
216 | static void | |
217 | sync_changed_cb(void *arg, uint64_t newval) | |
218 | { | |
219 | objset_t *os = arg; | |
220 | ||
221 | /* | |
222 | * Inheritance and range checking should have been done by now. | |
223 | */ | |
224 | ASSERT(newval == ZFS_SYNC_STANDARD || newval == ZFS_SYNC_ALWAYS || | |
225 | newval == ZFS_SYNC_DISABLED); | |
226 | ||
227 | os->os_sync = newval; | |
228 | if (os->os_zil) | |
229 | zil_set_sync(os->os_zil, newval); | |
230 | } | |
231 | ||
faf0f58c MA |
232 | static void |
233 | redundant_metadata_changed_cb(void *arg, uint64_t newval) | |
234 | { | |
235 | objset_t *os = arg; | |
236 | ||
237 | /* | |
238 | * Inheritance and range checking should have been done by now. | |
239 | */ | |
240 | ASSERT(newval == ZFS_REDUNDANT_METADATA_ALL || | |
241 | newval == ZFS_REDUNDANT_METADATA_MOST); | |
242 | ||
243 | os->os_redundant_metadata = newval; | |
244 | } | |
245 | ||
428870ff BB |
246 | static void |
247 | logbias_changed_cb(void *arg, uint64_t newval) | |
248 | { | |
249 | objset_t *os = arg; | |
250 | ||
251 | ASSERT(newval == ZFS_LOGBIAS_LATENCY || | |
252 | newval == ZFS_LOGBIAS_THROUGHPUT); | |
253 | os->os_logbias = newval; | |
254 | if (os->os_zil) | |
255 | zil_set_logbias(os->os_zil, newval); | |
b128c09f BB |
256 | } |
257 | ||
34dc7c2f BB |
258 | void |
259 | dmu_objset_byteswap(void *buf, size_t size) | |
260 | { | |
261 | objset_phys_t *osp = buf; | |
262 | ||
9babb374 | 263 | ASSERT(size == OBJSET_OLD_PHYS_SIZE || size == sizeof (objset_phys_t)); |
34dc7c2f BB |
264 | dnode_byteswap(&osp->os_meta_dnode); |
265 | byteswap_uint64_array(&osp->os_zil_header, sizeof (zil_header_t)); | |
266 | osp->os_type = BSWAP_64(osp->os_type); | |
9babb374 BB |
267 | osp->os_flags = BSWAP_64(osp->os_flags); |
268 | if (size == sizeof (objset_phys_t)) { | |
269 | dnode_byteswap(&osp->os_userused_dnode); | |
270 | dnode_byteswap(&osp->os_groupused_dnode); | |
271 | } | |
34dc7c2f BB |
272 | } |
273 | ||
274 | int | |
275 | dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, | |
428870ff | 276 | objset_t **osp) |
34dc7c2f | 277 | { |
428870ff | 278 | objset_t *os; |
b128c09f | 279 | int i, err; |
34dc7c2f BB |
280 | |
281 | ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock)); | |
282 | ||
79c76d5b | 283 | os = kmem_zalloc(sizeof (objset_t), KM_SLEEP); |
428870ff BB |
284 | os->os_dsl_dataset = ds; |
285 | os->os_spa = spa; | |
286 | os->os_rootbp = bp; | |
287 | if (!BP_IS_HOLE(os->os_rootbp)) { | |
34dc7c2f | 288 | uint32_t aflags = ARC_WAIT; |
5dbd68a3 | 289 | zbookmark_phys_t zb; |
428870ff BB |
290 | SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET, |
291 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
292 | ||
293 | if (DMU_OS_IS_L2CACHEABLE(os)) | |
b128c09f | 294 | aflags |= ARC_L2CACHE; |
3a17a7a9 SK |
295 | if (DMU_OS_IS_L2COMPRESSIBLE(os)) |
296 | aflags |= ARC_L2COMPRESS; | |
34dc7c2f | 297 | |
428870ff | 298 | dprintf_bp(os->os_rootbp, "reading %s", ""); |
294f6806 | 299 | err = arc_read(NULL, spa, os->os_rootbp, |
428870ff | 300 | arc_getbuf_func, &os->os_phys_buf, |
34dc7c2f | 301 | ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL, &aflags, &zb); |
13fe0198 | 302 | if (err != 0) { |
428870ff | 303 | kmem_free(os, sizeof (objset_t)); |
b128c09f BB |
304 | /* convert checksum errors into IO errors */ |
305 | if (err == ECKSUM) | |
2e528b49 | 306 | err = SET_ERROR(EIO); |
34dc7c2f BB |
307 | return (err); |
308 | } | |
9babb374 BB |
309 | |
310 | /* Increase the blocksize if we are permitted. */ | |
311 | if (spa_version(spa) >= SPA_VERSION_USERSPACE && | |
428870ff | 312 | arc_buf_size(os->os_phys_buf) < sizeof (objset_phys_t)) { |
9babb374 | 313 | arc_buf_t *buf = arc_buf_alloc(spa, |
428870ff | 314 | sizeof (objset_phys_t), &os->os_phys_buf, |
9babb374 BB |
315 | ARC_BUFC_METADATA); |
316 | bzero(buf->b_data, sizeof (objset_phys_t)); | |
428870ff BB |
317 | bcopy(os->os_phys_buf->b_data, buf->b_data, |
318 | arc_buf_size(os->os_phys_buf)); | |
319 | (void) arc_buf_remove_ref(os->os_phys_buf, | |
320 | &os->os_phys_buf); | |
321 | os->os_phys_buf = buf; | |
9babb374 BB |
322 | } |
323 | ||
428870ff BB |
324 | os->os_phys = os->os_phys_buf->b_data; |
325 | os->os_flags = os->os_phys->os_flags; | |
34dc7c2f | 326 | } else { |
9babb374 BB |
327 | int size = spa_version(spa) >= SPA_VERSION_USERSPACE ? |
328 | sizeof (objset_phys_t) : OBJSET_OLD_PHYS_SIZE; | |
428870ff BB |
329 | os->os_phys_buf = arc_buf_alloc(spa, size, |
330 | &os->os_phys_buf, ARC_BUFC_METADATA); | |
331 | os->os_phys = os->os_phys_buf->b_data; | |
332 | bzero(os->os_phys, size); | |
34dc7c2f BB |
333 | } |
334 | ||
335 | /* | |
336 | * Note: the changed_cb will be called once before the register | |
337 | * func returns, thus changing the checksum/compression from the | |
b128c09f BB |
338 | * default (fletcher2/off). Snapshots don't need to know about |
339 | * checksum/compression/copies. | |
34dc7c2f | 340 | */ |
9b67f605 | 341 | if (ds != NULL) { |
13fe0198 MA |
342 | err = dsl_prop_register(ds, |
343 | zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE), | |
428870ff | 344 | primary_cache_changed_cb, os); |
13fe0198 MA |
345 | if (err == 0) { |
346 | err = dsl_prop_register(ds, | |
347 | zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE), | |
428870ff | 348 | secondary_cache_changed_cb, os); |
13fe0198 | 349 | } |
b128c09f | 350 | if (!dsl_dataset_is_snapshot(ds)) { |
13fe0198 MA |
351 | if (err == 0) { |
352 | err = dsl_prop_register(ds, | |
353 | zfs_prop_to_name(ZFS_PROP_CHECKSUM), | |
428870ff | 354 | checksum_changed_cb, os); |
13fe0198 MA |
355 | } |
356 | if (err == 0) { | |
357 | err = dsl_prop_register(ds, | |
358 | zfs_prop_to_name(ZFS_PROP_COMPRESSION), | |
428870ff | 359 | compression_changed_cb, os); |
13fe0198 MA |
360 | } |
361 | if (err == 0) { | |
362 | err = dsl_prop_register(ds, | |
363 | zfs_prop_to_name(ZFS_PROP_COPIES), | |
428870ff | 364 | copies_changed_cb, os); |
13fe0198 MA |
365 | } |
366 | if (err == 0) { | |
367 | err = dsl_prop_register(ds, | |
368 | zfs_prop_to_name(ZFS_PROP_DEDUP), | |
428870ff | 369 | dedup_changed_cb, os); |
13fe0198 MA |
370 | } |
371 | if (err == 0) { | |
372 | err = dsl_prop_register(ds, | |
373 | zfs_prop_to_name(ZFS_PROP_LOGBIAS), | |
428870ff | 374 | logbias_changed_cb, os); |
13fe0198 MA |
375 | } |
376 | if (err == 0) { | |
377 | err = dsl_prop_register(ds, | |
378 | zfs_prop_to_name(ZFS_PROP_SYNC), | |
428870ff | 379 | sync_changed_cb, os); |
13fe0198 | 380 | } |
faf0f58c MA |
381 | if (err == 0) { |
382 | err = dsl_prop_register(ds, | |
383 | zfs_prop_to_name( | |
384 | ZFS_PROP_REDUNDANT_METADATA), | |
385 | redundant_metadata_changed_cb, os); | |
386 | } | |
b128c09f | 387 | } |
13fe0198 | 388 | if (err != 0) { |
428870ff | 389 | VERIFY(arc_buf_remove_ref(os->os_phys_buf, |
13fe0198 | 390 | &os->os_phys_buf)); |
428870ff | 391 | kmem_free(os, sizeof (objset_t)); |
34dc7c2f BB |
392 | return (err); |
393 | } | |
9b67f605 | 394 | } else { |
34dc7c2f | 395 | /* It's the meta-objset. */ |
428870ff BB |
396 | os->os_checksum = ZIO_CHECKSUM_FLETCHER_4; |
397 | os->os_compress = ZIO_COMPRESS_LZJB; | |
398 | os->os_copies = spa_max_replication(spa); | |
399 | os->os_dedup_checksum = ZIO_CHECKSUM_OFF; | |
faf0f58c MA |
400 | os->os_dedup_verify = B_FALSE; |
401 | os->os_logbias = ZFS_LOGBIAS_LATENCY; | |
402 | os->os_sync = ZFS_SYNC_STANDARD; | |
428870ff BB |
403 | os->os_primary_cache = ZFS_CACHE_ALL; |
404 | os->os_secondary_cache = ZFS_CACHE_ALL; | |
34dc7c2f BB |
405 | } |
406 | ||
572e2857 BB |
407 | if (ds == NULL || !dsl_dataset_is_snapshot(ds)) |
408 | os->os_zil_header = os->os_phys->os_zil_header; | |
428870ff | 409 | os->os_zil = zil_alloc(os, &os->os_zil_header); |
34dc7c2f BB |
410 | |
411 | for (i = 0; i < TXG_SIZE; i++) { | |
428870ff | 412 | list_create(&os->os_dirty_dnodes[i], sizeof (dnode_t), |
34dc7c2f | 413 | offsetof(dnode_t, dn_dirty_link[i])); |
428870ff | 414 | list_create(&os->os_free_dnodes[i], sizeof (dnode_t), |
34dc7c2f BB |
415 | offsetof(dnode_t, dn_dirty_link[i])); |
416 | } | |
428870ff | 417 | list_create(&os->os_dnodes, sizeof (dnode_t), |
34dc7c2f | 418 | offsetof(dnode_t, dn_link)); |
428870ff | 419 | list_create(&os->os_downgraded_dbufs, sizeof (dmu_buf_impl_t), |
34dc7c2f BB |
420 | offsetof(dmu_buf_impl_t, db_link)); |
421 | ||
428870ff BB |
422 | mutex_init(&os->os_lock, NULL, MUTEX_DEFAULT, NULL); |
423 | mutex_init(&os->os_obj_lock, NULL, MUTEX_DEFAULT, NULL); | |
424 | mutex_init(&os->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL); | |
425 | ||
572e2857 BB |
426 | DMU_META_DNODE(os) = dnode_special_open(os, |
427 | &os->os_phys->os_meta_dnode, DMU_META_DNODE_OBJECT, | |
428 | &os->os_meta_dnode); | |
428870ff | 429 | if (arc_buf_size(os->os_phys_buf) >= sizeof (objset_phys_t)) { |
572e2857 BB |
430 | DMU_USERUSED_DNODE(os) = dnode_special_open(os, |
431 | &os->os_phys->os_userused_dnode, DMU_USERUSED_OBJECT, | |
432 | &os->os_userused_dnode); | |
433 | DMU_GROUPUSED_DNODE(os) = dnode_special_open(os, | |
434 | &os->os_phys->os_groupused_dnode, DMU_GROUPUSED_OBJECT, | |
435 | &os->os_groupused_dnode); | |
9babb374 | 436 | } |
34dc7c2f | 437 | |
428870ff | 438 | *osp = os; |
34dc7c2f BB |
439 | return (0); |
440 | } | |
441 | ||
428870ff BB |
442 | int |
443 | dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp) | |
34dc7c2f | 444 | { |
428870ff | 445 | int err = 0; |
34dc7c2f BB |
446 | |
447 | mutex_enter(&ds->ds_opening_lock); | |
9b67f605 MA |
448 | if (ds->ds_objset == NULL) { |
449 | objset_t *os; | |
34dc7c2f | 450 | err = dmu_objset_open_impl(dsl_dataset_get_spa(ds), |
9b67f605 MA |
451 | ds, dsl_dataset_get_blkptr(ds), &os); |
452 | ||
453 | if (err == 0) { | |
454 | mutex_enter(&ds->ds_lock); | |
455 | ASSERT(ds->ds_objset == NULL); | |
456 | ds->ds_objset = os; | |
457 | mutex_exit(&ds->ds_lock); | |
458 | } | |
34dc7c2f | 459 | } |
9b67f605 | 460 | *osp = ds->ds_objset; |
34dc7c2f | 461 | mutex_exit(&ds->ds_opening_lock); |
428870ff | 462 | return (err); |
34dc7c2f BB |
463 | } |
464 | ||
13fe0198 MA |
465 | /* |
466 | * Holds the pool while the objset is held. Therefore only one objset | |
467 | * can be held at a time. | |
468 | */ | |
34dc7c2f | 469 | int |
428870ff | 470 | dmu_objset_hold(const char *name, void *tag, objset_t **osp) |
34dc7c2f | 471 | { |
13fe0198 | 472 | dsl_pool_t *dp; |
428870ff | 473 | dsl_dataset_t *ds; |
34dc7c2f BB |
474 | int err; |
475 | ||
13fe0198 MA |
476 | err = dsl_pool_hold(name, tag, &dp); |
477 | if (err != 0) | |
478 | return (err); | |
479 | err = dsl_dataset_hold(dp, name, tag, &ds); | |
480 | if (err != 0) { | |
481 | dsl_pool_rele(dp, tag); | |
428870ff | 482 | return (err); |
13fe0198 | 483 | } |
428870ff BB |
484 | |
485 | err = dmu_objset_from_ds(ds, osp); | |
13fe0198 | 486 | if (err != 0) { |
428870ff | 487 | dsl_dataset_rele(ds, tag); |
13fe0198 MA |
488 | dsl_pool_rele(dp, tag); |
489 | } | |
428870ff | 490 | |
34dc7c2f BB |
491 | return (err); |
492 | } | |
493 | ||
13fe0198 MA |
494 | /* |
495 | * dsl_pool must not be held when this is called. | |
496 | * Upon successful return, there will be a longhold on the dataset, | |
497 | * and the dsl_pool will not be held. | |
498 | */ | |
34dc7c2f | 499 | int |
428870ff BB |
500 | dmu_objset_own(const char *name, dmu_objset_type_t type, |
501 | boolean_t readonly, void *tag, objset_t **osp) | |
34dc7c2f | 502 | { |
13fe0198 | 503 | dsl_pool_t *dp; |
34dc7c2f BB |
504 | dsl_dataset_t *ds; |
505 | int err; | |
506 | ||
13fe0198 MA |
507 | err = dsl_pool_hold(name, FTAG, &dp); |
508 | if (err != 0) | |
509 | return (err); | |
510 | err = dsl_dataset_own(dp, name, tag, &ds); | |
511 | if (err != 0) { | |
512 | dsl_pool_rele(dp, FTAG); | |
34dc7c2f | 513 | return (err); |
13fe0198 | 514 | } |
34dc7c2f | 515 | |
428870ff | 516 | err = dmu_objset_from_ds(ds, osp); |
13fe0198 MA |
517 | dsl_pool_rele(dp, FTAG); |
518 | if (err != 0) { | |
428870ff BB |
519 | dsl_dataset_disown(ds, tag); |
520 | } else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) { | |
13fe0198 | 521 | dsl_dataset_disown(ds, tag); |
2e528b49 | 522 | return (SET_ERROR(EINVAL)); |
428870ff | 523 | } else if (!readonly && dsl_dataset_is_snapshot(ds)) { |
13fe0198 | 524 | dsl_dataset_disown(ds, tag); |
2e528b49 | 525 | return (SET_ERROR(EROFS)); |
34dc7c2f BB |
526 | } |
527 | return (err); | |
528 | } | |
529 | ||
530 | void | |
428870ff | 531 | dmu_objset_rele(objset_t *os, void *tag) |
34dc7c2f | 532 | { |
13fe0198 | 533 | dsl_pool_t *dp = dmu_objset_pool(os); |
428870ff | 534 | dsl_dataset_rele(os->os_dsl_dataset, tag); |
13fe0198 | 535 | dsl_pool_rele(dp, tag); |
428870ff | 536 | } |
b128c09f | 537 | |
831baf06 KW |
538 | /* |
539 | * When we are called, os MUST refer to an objset associated with a dataset | |
540 | * that is owned by 'tag'; that is, is held and long held by 'tag' and ds_owner | |
541 | * == tag. We will then release and reacquire ownership of the dataset while | |
542 | * holding the pool config_rwlock to avoid intervening namespace or ownership | |
543 | * changes may occur. | |
544 | * | |
545 | * This exists solely to accommodate zfs_ioc_userspace_upgrade()'s desire to | |
546 | * release the hold on its dataset and acquire a new one on the dataset of the | |
547 | * same name so that it can be partially torn down and reconstructed. | |
548 | */ | |
549 | void | |
550 | dmu_objset_refresh_ownership(objset_t *os, void *tag) | |
551 | { | |
552 | dsl_pool_t *dp; | |
553 | dsl_dataset_t *ds, *newds; | |
554 | char name[MAXNAMELEN]; | |
555 | ||
556 | ds = os->os_dsl_dataset; | |
557 | VERIFY3P(ds, !=, NULL); | |
558 | VERIFY3P(ds->ds_owner, ==, tag); | |
559 | VERIFY(dsl_dataset_long_held(ds)); | |
560 | ||
561 | dsl_dataset_name(ds, name); | |
562 | dp = dmu_objset_pool(os); | |
563 | dsl_pool_config_enter(dp, FTAG); | |
564 | dmu_objset_disown(os, tag); | |
565 | VERIFY0(dsl_dataset_own(dp, name, tag, &newds)); | |
566 | VERIFY3P(newds, ==, os->os_dsl_dataset); | |
567 | dsl_pool_config_exit(dp, FTAG); | |
568 | } | |
569 | ||
428870ff BB |
570 | void |
571 | dmu_objset_disown(objset_t *os, void *tag) | |
572 | { | |
573 | dsl_dataset_disown(os->os_dsl_dataset, tag); | |
34dc7c2f BB |
574 | } |
575 | ||
13fe0198 | 576 | void |
34dc7c2f BB |
577 | dmu_objset_evict_dbufs(objset_t *os) |
578 | { | |
34dc7c2f BB |
579 | dnode_t *dn; |
580 | ||
428870ff | 581 | mutex_enter(&os->os_lock); |
34dc7c2f BB |
582 | |
583 | /* process the mdn last, since the other dnodes have holds on it */ | |
572e2857 BB |
584 | list_remove(&os->os_dnodes, DMU_META_DNODE(os)); |
585 | list_insert_tail(&os->os_dnodes, DMU_META_DNODE(os)); | |
34dc7c2f BB |
586 | |
587 | /* | |
588 | * Find the first dnode with holds. We have to do this dance | |
589 | * because dnode_add_ref() only works if you already have a | |
590 | * hold. If there are no holds then it has no dbufs so OK to | |
591 | * skip. | |
592 | */ | |
428870ff | 593 | for (dn = list_head(&os->os_dnodes); |
34dc7c2f | 594 | dn && !dnode_add_ref(dn, FTAG); |
428870ff | 595 | dn = list_next(&os->os_dnodes, dn)) |
34dc7c2f BB |
596 | continue; |
597 | ||
598 | while (dn) { | |
599 | dnode_t *next_dn = dn; | |
600 | ||
601 | do { | |
428870ff | 602 | next_dn = list_next(&os->os_dnodes, next_dn); |
34dc7c2f BB |
603 | } while (next_dn && !dnode_add_ref(next_dn, FTAG)); |
604 | ||
428870ff | 605 | mutex_exit(&os->os_lock); |
34dc7c2f BB |
606 | dnode_evict_dbufs(dn); |
607 | dnode_rele(dn, FTAG); | |
428870ff | 608 | mutex_enter(&os->os_lock); |
34dc7c2f BB |
609 | dn = next_dn; |
610 | } | |
428870ff | 611 | mutex_exit(&os->os_lock); |
34dc7c2f BB |
612 | } |
613 | ||
614 | void | |
428870ff | 615 | dmu_objset_evict(objset_t *os) |
34dc7c2f | 616 | { |
d6320ddb | 617 | int t; |
34dc7c2f | 618 | |
6f1ffb06 MA |
619 | dsl_dataset_t *ds = os->os_dsl_dataset; |
620 | ||
d6320ddb | 621 | for (t = 0; t < TXG_SIZE; t++) |
428870ff | 622 | ASSERT(!dmu_objset_is_dirty(os, t)); |
34dc7c2f | 623 | |
b128c09f BB |
624 | if (ds) { |
625 | if (!dsl_dataset_is_snapshot(ds)) { | |
13fe0198 MA |
626 | VERIFY0(dsl_prop_unregister(ds, |
627 | zfs_prop_to_name(ZFS_PROP_CHECKSUM), | |
428870ff | 628 | checksum_changed_cb, os)); |
13fe0198 MA |
629 | VERIFY0(dsl_prop_unregister(ds, |
630 | zfs_prop_to_name(ZFS_PROP_COMPRESSION), | |
428870ff | 631 | compression_changed_cb, os)); |
13fe0198 MA |
632 | VERIFY0(dsl_prop_unregister(ds, |
633 | zfs_prop_to_name(ZFS_PROP_COPIES), | |
428870ff | 634 | copies_changed_cb, os)); |
13fe0198 MA |
635 | VERIFY0(dsl_prop_unregister(ds, |
636 | zfs_prop_to_name(ZFS_PROP_DEDUP), | |
428870ff | 637 | dedup_changed_cb, os)); |
13fe0198 MA |
638 | VERIFY0(dsl_prop_unregister(ds, |
639 | zfs_prop_to_name(ZFS_PROP_LOGBIAS), | |
428870ff | 640 | logbias_changed_cb, os)); |
13fe0198 MA |
641 | VERIFY0(dsl_prop_unregister(ds, |
642 | zfs_prop_to_name(ZFS_PROP_SYNC), | |
428870ff | 643 | sync_changed_cb, os)); |
faf0f58c MA |
644 | VERIFY0(dsl_prop_unregister(ds, |
645 | zfs_prop_to_name(ZFS_PROP_REDUNDANT_METADATA), | |
646 | redundant_metadata_changed_cb, os)); | |
b128c09f | 647 | } |
13fe0198 MA |
648 | VERIFY0(dsl_prop_unregister(ds, |
649 | zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE), | |
428870ff | 650 | primary_cache_changed_cb, os)); |
13fe0198 MA |
651 | VERIFY0(dsl_prop_unregister(ds, |
652 | zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE), | |
428870ff | 653 | secondary_cache_changed_cb, os)); |
34dc7c2f BB |
654 | } |
655 | ||
428870ff BB |
656 | if (os->os_sa) |
657 | sa_tear_down(os); | |
658 | ||
13fe0198 | 659 | dmu_objset_evict_dbufs(os); |
34dc7c2f | 660 | |
572e2857 BB |
661 | dnode_special_close(&os->os_meta_dnode); |
662 | if (DMU_USERUSED_DNODE(os)) { | |
663 | dnode_special_close(&os->os_userused_dnode); | |
664 | dnode_special_close(&os->os_groupused_dnode); | |
9babb374 | 665 | } |
428870ff BB |
666 | zil_free(os->os_zil); |
667 | ||
668 | ASSERT3P(list_head(&os->os_dnodes), ==, NULL); | |
34dc7c2f | 669 | |
13fe0198 | 670 | VERIFY(arc_buf_remove_ref(os->os_phys_buf, &os->os_phys_buf)); |
572e2857 BB |
671 | |
672 | /* | |
673 | * This is a barrier to prevent the objset from going away in | |
674 | * dnode_move() until we can safely ensure that the objset is still in | |
675 | * use. We consider the objset valid before the barrier and invalid | |
676 | * after the barrier. | |
677 | */ | |
678 | rw_enter(&os_lock, RW_READER); | |
679 | rw_exit(&os_lock); | |
680 | ||
428870ff BB |
681 | mutex_destroy(&os->os_lock); |
682 | mutex_destroy(&os->os_obj_lock); | |
683 | mutex_destroy(&os->os_user_ptr_lock); | |
684 | kmem_free(os, sizeof (objset_t)); | |
685 | } | |
9babb374 | 686 | |
428870ff BB |
687 | timestruc_t |
688 | dmu_objset_snap_cmtime(objset_t *os) | |
689 | { | |
690 | return (dsl_dir_snap_cmtime(os->os_dsl_dataset->ds_dir)); | |
34dc7c2f BB |
691 | } |
692 | ||
693 | /* called from dsl for meta-objset */ | |
428870ff | 694 | objset_t * |
34dc7c2f BB |
695 | dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, |
696 | dmu_objset_type_t type, dmu_tx_t *tx) | |
697 | { | |
428870ff | 698 | objset_t *os; |
34dc7c2f BB |
699 | dnode_t *mdn; |
700 | ||
701 | ASSERT(dmu_tx_is_syncing(tx)); | |
13fe0198 | 702 | |
572e2857 | 703 | if (ds != NULL) |
13fe0198 | 704 | VERIFY0(dmu_objset_from_ds(ds, &os)); |
572e2857 | 705 | else |
13fe0198 | 706 | VERIFY0(dmu_objset_open_impl(spa, NULL, bp, &os)); |
572e2857 BB |
707 | |
708 | mdn = DMU_META_DNODE(os); | |
34dc7c2f BB |
709 | |
710 | dnode_allocate(mdn, DMU_OT_DNODE, 1 << DNODE_BLOCK_SHIFT, | |
711 | DN_MAX_INDBLKSHIFT, DMU_OT_NONE, 0, tx); | |
712 | ||
713 | /* | |
714 | * We don't want to have to increase the meta-dnode's nlevels | |
715 | * later, because then we could do it in quescing context while | |
716 | * we are also accessing it in open context. | |
717 | * | |
718 | * This precaution is not necessary for the MOS (ds == NULL), | |
719 | * because the MOS is only updated in syncing context. | |
720 | * This is most fortunate: the MOS is the only objset that | |
721 | * needs to be synced multiple times as spa_sync() iterates | |
722 | * to convergence, so minimizing its dn_nlevels matters. | |
723 | */ | |
724 | if (ds != NULL) { | |
725 | int levels = 1; | |
726 | ||
727 | /* | |
728 | * Determine the number of levels necessary for the meta-dnode | |
729 | * to contain DN_MAX_OBJECT dnodes. | |
730 | */ | |
731 | while ((uint64_t)mdn->dn_nblkptr << (mdn->dn_datablkshift + | |
732 | (levels - 1) * (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) < | |
733 | DN_MAX_OBJECT * sizeof (dnode_phys_t)) | |
734 | levels++; | |
735 | ||
736 | mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] = | |
737 | mdn->dn_nlevels = levels; | |
738 | } | |
739 | ||
740 | ASSERT(type != DMU_OST_NONE); | |
741 | ASSERT(type != DMU_OST_ANY); | |
742 | ASSERT(type < DMU_OST_NUMTYPES); | |
428870ff BB |
743 | os->os_phys->os_type = type; |
744 | if (dmu_objset_userused_enabled(os)) { | |
745 | os->os_phys->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; | |
746 | os->os_flags = os->os_phys->os_flags; | |
9babb374 | 747 | } |
34dc7c2f BB |
748 | |
749 | dsl_dataset_dirty(ds, tx); | |
750 | ||
428870ff | 751 | return (os); |
34dc7c2f BB |
752 | } |
753 | ||
13fe0198 MA |
754 | typedef struct dmu_objset_create_arg { |
755 | const char *doca_name; | |
756 | cred_t *doca_cred; | |
757 | void (*doca_userfunc)(objset_t *os, void *arg, | |
758 | cred_t *cr, dmu_tx_t *tx); | |
759 | void *doca_userarg; | |
760 | dmu_objset_type_t doca_type; | |
761 | uint64_t doca_flags; | |
762 | } dmu_objset_create_arg_t; | |
34dc7c2f BB |
763 | |
764 | /*ARGSUSED*/ | |
765 | static int | |
13fe0198 | 766 | dmu_objset_create_check(void *arg, dmu_tx_t *tx) |
34dc7c2f | 767 | { |
13fe0198 MA |
768 | dmu_objset_create_arg_t *doca = arg; |
769 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
770 | dsl_dir_t *pdd; | |
771 | const char *tail; | |
772 | int error; | |
34dc7c2f | 773 | |
13fe0198 | 774 | if (strchr(doca->doca_name, '@') != NULL) |
2e528b49 | 775 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 776 | |
13fe0198 MA |
777 | error = dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail); |
778 | if (error != 0) | |
779 | return (error); | |
780 | if (tail == NULL) { | |
781 | dsl_dir_rele(pdd, FTAG); | |
2e528b49 | 782 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 783 | } |
788eb90c JJ |
784 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
785 | doca->doca_cred); | |
13fe0198 | 786 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f | 787 | |
788eb90c | 788 | return (error); |
34dc7c2f BB |
789 | } |
790 | ||
791 | static void | |
13fe0198 | 792 | dmu_objset_create_sync(void *arg, dmu_tx_t *tx) |
34dc7c2f | 793 | { |
13fe0198 MA |
794 | dmu_objset_create_arg_t *doca = arg; |
795 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
796 | dsl_dir_t *pdd; | |
797 | const char *tail; | |
6f1ffb06 | 798 | dsl_dataset_t *ds; |
13fe0198 | 799 | uint64_t obj; |
6f1ffb06 | 800 | blkptr_t *bp; |
13fe0198 | 801 | objset_t *os; |
34dc7c2f | 802 | |
13fe0198 | 803 | VERIFY0(dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail)); |
34dc7c2f | 804 | |
13fe0198 MA |
805 | obj = dsl_dataset_create_sync(pdd, tail, NULL, doca->doca_flags, |
806 | doca->doca_cred, tx); | |
34dc7c2f | 807 | |
13fe0198 | 808 | VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); |
6f1ffb06 | 809 | bp = dsl_dataset_get_blkptr(ds); |
13fe0198 MA |
810 | os = dmu_objset_create_impl(pdd->dd_pool->dp_spa, |
811 | ds, bp, doca->doca_type, tx); | |
34dc7c2f | 812 | |
13fe0198 MA |
813 | if (doca->doca_userfunc != NULL) { |
814 | doca->doca_userfunc(os, doca->doca_userarg, | |
815 | doca->doca_cred, tx); | |
34dc7c2f BB |
816 | } |
817 | ||
13fe0198 | 818 | spa_history_log_internal_ds(ds, "create", tx, ""); |
6f1ffb06 | 819 | dsl_dataset_rele(ds, FTAG); |
13fe0198 | 820 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f BB |
821 | } |
822 | ||
823 | int | |
428870ff | 824 | dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, |
34dc7c2f BB |
825 | void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg) |
826 | { | |
13fe0198 | 827 | dmu_objset_create_arg_t doca; |
34dc7c2f | 828 | |
13fe0198 MA |
829 | doca.doca_name = name; |
830 | doca.doca_cred = CRED(); | |
831 | doca.doca_flags = flags; | |
832 | doca.doca_userfunc = func; | |
833 | doca.doca_userarg = arg; | |
834 | doca.doca_type = type; | |
34dc7c2f | 835 | |
13fe0198 MA |
836 | return (dsl_sync_task(name, |
837 | dmu_objset_create_check, dmu_objset_create_sync, &doca, 5)); | |
34dc7c2f BB |
838 | } |
839 | ||
13fe0198 MA |
840 | typedef struct dmu_objset_clone_arg { |
841 | const char *doca_clone; | |
842 | const char *doca_origin; | |
843 | cred_t *doca_cred; | |
844 | } dmu_objset_clone_arg_t; | |
845 | ||
846 | /*ARGSUSED*/ | |
847 | static int | |
848 | dmu_objset_clone_check(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 849 | { |
13fe0198 | 850 | dmu_objset_clone_arg_t *doca = arg; |
428870ff BB |
851 | dsl_dir_t *pdd; |
852 | const char *tail; | |
13fe0198 MA |
853 | int error; |
854 | dsl_dataset_t *origin; | |
855 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
34dc7c2f | 856 | |
13fe0198 | 857 | if (strchr(doca->doca_clone, '@') != NULL) |
2e528b49 | 858 | return (SET_ERROR(EINVAL)); |
13fe0198 MA |
859 | |
860 | error = dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail); | |
861 | if (error != 0) | |
862 | return (error); | |
428870ff | 863 | if (tail == NULL) { |
13fe0198 | 864 | dsl_dir_rele(pdd, FTAG); |
2e528b49 | 865 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 866 | } |
13fe0198 MA |
867 | /* You can't clone across pools. */ |
868 | if (pdd->dd_pool != dp) { | |
869 | dsl_dir_rele(pdd, FTAG); | |
2e528b49 | 870 | return (SET_ERROR(EXDEV)); |
428870ff | 871 | } |
788eb90c JJ |
872 | error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, |
873 | doca->doca_cred); | |
874 | if (error != 0) { | |
875 | dsl_dir_rele(pdd, FTAG); | |
876 | return (SET_ERROR(EDQUOT)); | |
877 | } | |
13fe0198 | 878 | dsl_dir_rele(pdd, FTAG); |
428870ff | 879 | |
13fe0198 MA |
880 | error = dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin); |
881 | if (error != 0) | |
572e2857 BB |
882 | return (error); |
883 | ||
13fe0198 MA |
884 | /* You can't clone across pools. */ |
885 | if (origin->ds_dir->dd_pool != dp) { | |
886 | dsl_dataset_rele(origin, FTAG); | |
2e528b49 | 887 | return (SET_ERROR(EXDEV)); |
572e2857 | 888 | } |
9babb374 | 889 | |
13fe0198 MA |
890 | /* You can only clone snapshots, not the head datasets. */ |
891 | if (!dsl_dataset_is_snapshot(origin)) { | |
892 | dsl_dataset_rele(origin, FTAG); | |
2e528b49 | 893 | return (SET_ERROR(EINVAL)); |
428870ff | 894 | } |
13fe0198 | 895 | dsl_dataset_rele(origin, FTAG); |
572e2857 | 896 | |
13fe0198 | 897 | return (0); |
9babb374 | 898 | } |
34dc7c2f | 899 | |
13fe0198 MA |
900 | static void |
901 | dmu_objset_clone_sync(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 902 | { |
13fe0198 MA |
903 | dmu_objset_clone_arg_t *doca = arg; |
904 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
905 | dsl_dir_t *pdd; | |
906 | const char *tail; | |
907 | dsl_dataset_t *origin, *ds; | |
908 | uint64_t obj; | |
909 | char namebuf[MAXNAMELEN]; | |
6f1ffb06 | 910 | |
13fe0198 MA |
911 | VERIFY0(dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail)); |
912 | VERIFY0(dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin)); | |
6f1ffb06 | 913 | |
13fe0198 MA |
914 | obj = dsl_dataset_create_sync(pdd, tail, origin, 0, |
915 | doca->doca_cred, tx); | |
34dc7c2f | 916 | |
13fe0198 MA |
917 | VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); |
918 | dsl_dataset_name(origin, namebuf); | |
919 | spa_history_log_internal_ds(ds, "clone", tx, | |
920 | "origin=%s (%llu)", namebuf, origin->ds_object); | |
921 | dsl_dataset_rele(ds, FTAG); | |
922 | dsl_dataset_rele(origin, FTAG); | |
923 | dsl_dir_rele(pdd, FTAG); | |
34dc7c2f BB |
924 | } |
925 | ||
926 | int | |
13fe0198 | 927 | dmu_objset_clone(const char *clone, const char *origin) |
34dc7c2f | 928 | { |
13fe0198 | 929 | dmu_objset_clone_arg_t doca; |
34dc7c2f | 930 | |
13fe0198 MA |
931 | doca.doca_clone = clone; |
932 | doca.doca_origin = origin; | |
933 | doca.doca_cred = CRED(); | |
572e2857 | 934 | |
13fe0198 MA |
935 | return (dsl_sync_task(clone, |
936 | dmu_objset_clone_check, dmu_objset_clone_sync, &doca, 5)); | |
6f1ffb06 MA |
937 | } |
938 | ||
939 | int | |
940 | dmu_objset_snapshot_one(const char *fsname, const char *snapname) | |
941 | { | |
942 | int err; | |
943 | char *longsnap = kmem_asprintf("%s@%s", fsname, snapname); | |
944 | nvlist_t *snaps = fnvlist_alloc(); | |
945 | ||
946 | fnvlist_add_boolean(snaps, longsnap); | |
6f1ffb06 | 947 | strfree(longsnap); |
13fe0198 MA |
948 | err = dsl_dataset_snapshot(snaps, NULL, NULL); |
949 | fnvlist_free(snaps); | |
6f1ffb06 MA |
950 | return (err); |
951 | } | |
952 | ||
34dc7c2f | 953 | static void |
9babb374 | 954 | dmu_objset_sync_dnodes(list_t *list, list_t *newlist, dmu_tx_t *tx) |
34dc7c2f BB |
955 | { |
956 | dnode_t *dn; | |
957 | ||
c65aa5b2 | 958 | while ((dn = list_head(list))) { |
34dc7c2f BB |
959 | ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT); |
960 | ASSERT(dn->dn_dbuf->db_data_pending); | |
961 | /* | |
9babb374 BB |
962 | * Initialize dn_zio outside dnode_sync() because the |
963 | * meta-dnode needs to set it ouside dnode_sync(). | |
34dc7c2f BB |
964 | */ |
965 | dn->dn_zio = dn->dn_dbuf->db_data_pending->dr_zio; | |
966 | ASSERT(dn->dn_zio); | |
967 | ||
968 | ASSERT3U(dn->dn_nlevels, <=, DN_MAX_LEVELS); | |
969 | list_remove(list, dn); | |
9babb374 BB |
970 | |
971 | if (newlist) { | |
972 | (void) dnode_add_ref(dn, newlist); | |
973 | list_insert_tail(newlist, dn); | |
974 | } | |
975 | ||
34dc7c2f BB |
976 | dnode_sync(dn, tx); |
977 | } | |
978 | } | |
979 | ||
980 | /* ARGSUSED */ | |
981 | static void | |
428870ff | 982 | dmu_objset_write_ready(zio_t *zio, arc_buf_t *abuf, void *arg) |
34dc7c2f | 983 | { |
d6320ddb BB |
984 | int i; |
985 | ||
b128c09f | 986 | blkptr_t *bp = zio->io_bp; |
428870ff | 987 | objset_t *os = arg; |
34dc7c2f | 988 | dnode_phys_t *dnp = &os->os_phys->os_meta_dnode; |
34dc7c2f | 989 | |
9b67f605 | 990 | ASSERT(!BP_IS_EMBEDDED(bp)); |
13fe0198 MA |
991 | ASSERT3P(bp, ==, os->os_rootbp); |
992 | ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_OBJSET); | |
993 | ASSERT0(BP_GET_LEVEL(bp)); | |
34dc7c2f BB |
994 | |
995 | /* | |
9babb374 BB |
996 | * Update rootbp fill count: it should be the number of objects |
997 | * allocated in the object set (not counting the "special" | |
998 | * objects that are stored in the objset_phys_t -- the meta | |
999 | * dnode and user/group accounting objects). | |
34dc7c2f | 1000 | */ |
9babb374 | 1001 | bp->blk_fill = 0; |
d6320ddb | 1002 | for (i = 0; i < dnp->dn_nblkptr; i++) |
9b67f605 | 1003 | bp->blk_fill += BP_GET_FILL(&dnp->dn_blkptr[i]); |
428870ff BB |
1004 | } |
1005 | ||
1006 | /* ARGSUSED */ | |
1007 | static void | |
1008 | dmu_objset_write_done(zio_t *zio, arc_buf_t *abuf, void *arg) | |
1009 | { | |
1010 | blkptr_t *bp = zio->io_bp; | |
1011 | blkptr_t *bp_orig = &zio->io_bp_orig; | |
1012 | objset_t *os = arg; | |
34dc7c2f | 1013 | |
b128c09f | 1014 | if (zio->io_flags & ZIO_FLAG_IO_REWRITE) { |
428870ff | 1015 | ASSERT(BP_EQUAL(bp, bp_orig)); |
b128c09f | 1016 | } else { |
428870ff BB |
1017 | dsl_dataset_t *ds = os->os_dsl_dataset; |
1018 | dmu_tx_t *tx = os->os_synctx; | |
1019 | ||
1020 | (void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE); | |
1021 | dsl_dataset_block_born(ds, bp, tx); | |
34dc7c2f BB |
1022 | } |
1023 | } | |
1024 | ||
34dc7c2f BB |
1025 | /* called from dsl */ |
1026 | void | |
428870ff | 1027 | dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx) |
34dc7c2f BB |
1028 | { |
1029 | int txgoff; | |
5dbd68a3 | 1030 | zbookmark_phys_t zb; |
428870ff | 1031 | zio_prop_t zp; |
34dc7c2f BB |
1032 | zio_t *zio; |
1033 | list_t *list; | |
9babb374 | 1034 | list_t *newlist = NULL; |
34dc7c2f BB |
1035 | dbuf_dirty_record_t *dr; |
1036 | ||
1037 | dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg); | |
1038 | ||
1039 | ASSERT(dmu_tx_is_syncing(tx)); | |
1040 | /* XXX the write_done callback should really give us the tx... */ | |
1041 | os->os_synctx = tx; | |
1042 | ||
1043 | if (os->os_dsl_dataset == NULL) { | |
1044 | /* | |
1045 | * This is the MOS. If we have upgraded, | |
1046 | * spa_max_replication() could change, so reset | |
1047 | * os_copies here. | |
1048 | */ | |
1049 | os->os_copies = spa_max_replication(os->os_spa); | |
1050 | } | |
1051 | ||
1052 | /* | |
1053 | * Create the root block IO | |
1054 | */ | |
428870ff BB |
1055 | SET_BOOKMARK(&zb, os->os_dsl_dataset ? |
1056 | os->os_dsl_dataset->ds_object : DMU_META_OBJSET, | |
1057 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
294f6806 | 1058 | arc_release(os->os_phys_buf, &os->os_phys_buf); |
b128c09f | 1059 | |
428870ff | 1060 | dmu_write_policy(os, NULL, 0, 0, &zp); |
9babb374 | 1061 | |
428870ff | 1062 | zio = arc_write(pio, os->os_spa, tx->tx_txg, |
3a17a7a9 SK |
1063 | os->os_rootbp, os->os_phys_buf, DMU_OS_IS_L2CACHEABLE(os), |
1064 | DMU_OS_IS_L2COMPRESSIBLE(os), &zp, dmu_objset_write_ready, | |
e8b96c60 | 1065 | NULL, dmu_objset_write_done, os, ZIO_PRIORITY_ASYNC_WRITE, |
3a17a7a9 | 1066 | ZIO_FLAG_MUSTSUCCEED, &zb); |
34dc7c2f BB |
1067 | |
1068 | /* | |
9babb374 | 1069 | * Sync special dnodes - the parent IO for the sync is the root block |
34dc7c2f | 1070 | */ |
572e2857 BB |
1071 | DMU_META_DNODE(os)->dn_zio = zio; |
1072 | dnode_sync(DMU_META_DNODE(os), tx); | |
34dc7c2f | 1073 | |
9babb374 BB |
1074 | os->os_phys->os_flags = os->os_flags; |
1075 | ||
572e2857 BB |
1076 | if (DMU_USERUSED_DNODE(os) && |
1077 | DMU_USERUSED_DNODE(os)->dn_type != DMU_OT_NONE) { | |
1078 | DMU_USERUSED_DNODE(os)->dn_zio = zio; | |
1079 | dnode_sync(DMU_USERUSED_DNODE(os), tx); | |
1080 | DMU_GROUPUSED_DNODE(os)->dn_zio = zio; | |
1081 | dnode_sync(DMU_GROUPUSED_DNODE(os), tx); | |
9babb374 BB |
1082 | } |
1083 | ||
34dc7c2f BB |
1084 | txgoff = tx->tx_txg & TXG_MASK; |
1085 | ||
9babb374 BB |
1086 | if (dmu_objset_userused_enabled(os)) { |
1087 | newlist = &os->os_synced_dnodes; | |
1088 | /* | |
1089 | * We must create the list here because it uses the | |
1090 | * dn_dirty_link[] of this txg. | |
1091 | */ | |
1092 | list_create(newlist, sizeof (dnode_t), | |
1093 | offsetof(dnode_t, dn_dirty_link[txgoff])); | |
1094 | } | |
1095 | ||
1096 | dmu_objset_sync_dnodes(&os->os_free_dnodes[txgoff], newlist, tx); | |
1097 | dmu_objset_sync_dnodes(&os->os_dirty_dnodes[txgoff], newlist, tx); | |
34dc7c2f | 1098 | |
572e2857 | 1099 | list = &DMU_META_DNODE(os)->dn_dirty_records[txgoff]; |
6f1ffb06 | 1100 | while ((dr = list_head(list))) { |
13fe0198 | 1101 | ASSERT0(dr->dr_dbuf->db_level); |
34dc7c2f BB |
1102 | list_remove(list, dr); |
1103 | if (dr->dr_zio) | |
1104 | zio_nowait(dr->dr_zio); | |
1105 | } | |
1106 | /* | |
1107 | * Free intent log blocks up to this tx. | |
1108 | */ | |
1109 | zil_sync(os->os_zil, tx); | |
b128c09f | 1110 | os->os_phys->os_zil_header = os->os_zil_header; |
34dc7c2f BB |
1111 | zio_nowait(zio); |
1112 | } | |
1113 | ||
428870ff BB |
1114 | boolean_t |
1115 | dmu_objset_is_dirty(objset_t *os, uint64_t txg) | |
1116 | { | |
1117 | return (!list_is_empty(&os->os_dirty_dnodes[txg & TXG_MASK]) || | |
1118 | !list_is_empty(&os->os_free_dnodes[txg & TXG_MASK])); | |
1119 | } | |
1120 | ||
572e2857 | 1121 | static objset_used_cb_t *used_cbs[DMU_OST_NUMTYPES]; |
9babb374 BB |
1122 | |
1123 | void | |
1124 | dmu_objset_register_type(dmu_objset_type_t ost, objset_used_cb_t *cb) | |
1125 | { | |
1126 | used_cbs[ost] = cb; | |
1127 | } | |
1128 | ||
1129 | boolean_t | |
428870ff | 1130 | dmu_objset_userused_enabled(objset_t *os) |
9babb374 BB |
1131 | { |
1132 | return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE && | |
572e2857 BB |
1133 | used_cbs[os->os_phys->os_type] != NULL && |
1134 | DMU_USERUSED_DNODE(os) != NULL); | |
9babb374 BB |
1135 | } |
1136 | ||
428870ff BB |
1137 | static void |
1138 | do_userquota_update(objset_t *os, uint64_t used, uint64_t flags, | |
1139 | uint64_t user, uint64_t group, boolean_t subtract, dmu_tx_t *tx) | |
1140 | { | |
1141 | if ((flags & DNODE_FLAG_USERUSED_ACCOUNTED)) { | |
1142 | int64_t delta = DNODE_SIZE + used; | |
1143 | if (subtract) | |
1144 | delta = -delta; | |
1145 | VERIFY3U(0, ==, zap_increment_int(os, DMU_USERUSED_OBJECT, | |
1146 | user, delta, tx)); | |
1147 | VERIFY3U(0, ==, zap_increment_int(os, DMU_GROUPUSED_OBJECT, | |
1148 | group, delta, tx)); | |
1149 | } | |
1150 | } | |
1151 | ||
9babb374 | 1152 | void |
428870ff | 1153 | dmu_objset_do_userquota_updates(objset_t *os, dmu_tx_t *tx) |
9babb374 BB |
1154 | { |
1155 | dnode_t *dn; | |
1156 | list_t *list = &os->os_synced_dnodes; | |
9babb374 BB |
1157 | |
1158 | ASSERT(list_head(list) == NULL || dmu_objset_userused_enabled(os)); | |
1159 | ||
6f1ffb06 | 1160 | while ((dn = list_head(list))) { |
572e2857 | 1161 | int flags; |
9babb374 | 1162 | ASSERT(!DMU_OBJECT_IS_SPECIAL(dn->dn_object)); |
9babb374 BB |
1163 | ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE || |
1164 | dn->dn_phys->dn_flags & | |
1165 | DNODE_FLAG_USERUSED_ACCOUNTED); | |
1166 | ||
1167 | /* Allocate the user/groupused objects if necessary. */ | |
572e2857 | 1168 | if (DMU_USERUSED_DNODE(os)->dn_type == DMU_OT_NONE) { |
428870ff | 1169 | VERIFY(0 == zap_create_claim(os, |
9babb374 BB |
1170 | DMU_USERUSED_OBJECT, |
1171 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
428870ff | 1172 | VERIFY(0 == zap_create_claim(os, |
9babb374 BB |
1173 | DMU_GROUPUSED_OBJECT, |
1174 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
1175 | } | |
1176 | ||
1177 | /* | |
428870ff BB |
1178 | * We intentionally modify the zap object even if the |
1179 | * net delta is zero. Otherwise | |
1180 | * the block of the zap obj could be shared between | |
1181 | * datasets but need to be different between them after | |
1182 | * a bprewrite. | |
9babb374 | 1183 | */ |
9babb374 | 1184 | |
572e2857 BB |
1185 | flags = dn->dn_id_flags; |
1186 | ASSERT(flags); | |
1187 | if (flags & DN_ID_OLD_EXIST) { | |
428870ff BB |
1188 | do_userquota_update(os, dn->dn_oldused, dn->dn_oldflags, |
1189 | dn->dn_olduid, dn->dn_oldgid, B_TRUE, tx); | |
1190 | } | |
572e2857 | 1191 | if (flags & DN_ID_NEW_EXIST) { |
428870ff BB |
1192 | do_userquota_update(os, DN_USED_BYTES(dn->dn_phys), |
1193 | dn->dn_phys->dn_flags, dn->dn_newuid, | |
1194 | dn->dn_newgid, B_FALSE, tx); | |
1195 | } | |
1196 | ||
572e2857 | 1197 | mutex_enter(&dn->dn_mtx); |
428870ff BB |
1198 | dn->dn_oldused = 0; |
1199 | dn->dn_oldflags = 0; | |
1200 | if (dn->dn_id_flags & DN_ID_NEW_EXIST) { | |
1201 | dn->dn_olduid = dn->dn_newuid; | |
1202 | dn->dn_oldgid = dn->dn_newgid; | |
1203 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
1204 | if (dn->dn_bonuslen == 0) | |
1205 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1206 | else | |
1207 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1208 | } | |
1209 | dn->dn_id_flags &= ~(DN_ID_NEW_EXIST); | |
9babb374 BB |
1210 | mutex_exit(&dn->dn_mtx); |
1211 | ||
1212 | list_remove(list, dn); | |
1213 | dnode_rele(dn, list); | |
1214 | } | |
1215 | } | |
1216 | ||
428870ff BB |
1217 | /* |
1218 | * Returns a pointer to data to find uid/gid from | |
1219 | * | |
1220 | * If a dirty record for transaction group that is syncing can't | |
1221 | * be found then NULL is returned. In the NULL case it is assumed | |
1222 | * the uid/gid aren't changing. | |
1223 | */ | |
1224 | static void * | |
1225 | dmu_objset_userquota_find_data(dmu_buf_impl_t *db, dmu_tx_t *tx) | |
1226 | { | |
1227 | dbuf_dirty_record_t *dr, **drp; | |
1228 | void *data; | |
1229 | ||
1230 | if (db->db_dirtycnt == 0) | |
1231 | return (db->db.db_data); /* Nothing is changing */ | |
1232 | ||
1233 | for (drp = &db->db_last_dirty; (dr = *drp) != NULL; drp = &dr->dr_next) | |
1234 | if (dr->dr_txg == tx->tx_txg) | |
1235 | break; | |
1236 | ||
572e2857 | 1237 | if (dr == NULL) { |
428870ff | 1238 | data = NULL; |
572e2857 BB |
1239 | } else { |
1240 | dnode_t *dn; | |
1241 | ||
1242 | DB_DNODE_ENTER(dr->dr_dbuf); | |
1243 | dn = DB_DNODE(dr->dr_dbuf); | |
1244 | ||
1245 | if (dn->dn_bonuslen == 0 && | |
1246 | dr->dr_dbuf->db_blkid == DMU_SPILL_BLKID) | |
1247 | data = dr->dt.dl.dr_data->b_data; | |
1248 | else | |
1249 | data = dr->dt.dl.dr_data; | |
1250 | ||
1251 | DB_DNODE_EXIT(dr->dr_dbuf); | |
1252 | } | |
1253 | ||
428870ff BB |
1254 | return (data); |
1255 | } | |
1256 | ||
1257 | void | |
1258 | dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) | |
1259 | { | |
1260 | objset_t *os = dn->dn_objset; | |
1261 | void *data = NULL; | |
1262 | dmu_buf_impl_t *db = NULL; | |
a117a6d6 GW |
1263 | uint64_t *user = NULL; |
1264 | uint64_t *group = NULL; | |
428870ff BB |
1265 | int flags = dn->dn_id_flags; |
1266 | int error; | |
1267 | boolean_t have_spill = B_FALSE; | |
1268 | ||
1269 | if (!dmu_objset_userused_enabled(dn->dn_objset)) | |
1270 | return; | |
1271 | ||
1272 | if (before && (flags & (DN_ID_CHKED_BONUS|DN_ID_OLD_EXIST| | |
1273 | DN_ID_CHKED_SPILL))) | |
1274 | return; | |
1275 | ||
1276 | if (before && dn->dn_bonuslen != 0) | |
1277 | data = DN_BONUS(dn->dn_phys); | |
1278 | else if (!before && dn->dn_bonuslen != 0) { | |
1279 | if (dn->dn_bonus) { | |
1280 | db = dn->dn_bonus; | |
1281 | mutex_enter(&db->db_mtx); | |
1282 | data = dmu_objset_userquota_find_data(db, tx); | |
1283 | } else { | |
1284 | data = DN_BONUS(dn->dn_phys); | |
1285 | } | |
1286 | } else if (dn->dn_bonuslen == 0 && dn->dn_bonustype == DMU_OT_SA) { | |
1287 | int rf = 0; | |
1288 | ||
1289 | if (RW_WRITE_HELD(&dn->dn_struct_rwlock)) | |
1290 | rf |= DB_RF_HAVESTRUCT; | |
572e2857 BB |
1291 | error = dmu_spill_hold_by_dnode(dn, |
1292 | rf | DB_RF_MUST_SUCCEED, | |
428870ff BB |
1293 | FTAG, (dmu_buf_t **)&db); |
1294 | ASSERT(error == 0); | |
1295 | mutex_enter(&db->db_mtx); | |
1296 | data = (before) ? db->db.db_data : | |
1297 | dmu_objset_userquota_find_data(db, tx); | |
1298 | have_spill = B_TRUE; | |
1299 | } else { | |
1300 | mutex_enter(&dn->dn_mtx); | |
1301 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1302 | mutex_exit(&dn->dn_mtx); | |
1303 | return; | |
1304 | } | |
1305 | ||
1306 | if (before) { | |
1307 | ASSERT(data); | |
1308 | user = &dn->dn_olduid; | |
1309 | group = &dn->dn_oldgid; | |
1310 | } else if (data) { | |
1311 | user = &dn->dn_newuid; | |
1312 | group = &dn->dn_newgid; | |
1313 | } | |
1314 | ||
1315 | /* | |
1316 | * Must always call the callback in case the object | |
1317 | * type has changed and that type isn't an object type to track | |
1318 | */ | |
1319 | error = used_cbs[os->os_phys->os_type](dn->dn_bonustype, data, | |
1320 | user, group); | |
1321 | ||
1322 | /* | |
1323 | * Preserve existing uid/gid when the callback can't determine | |
1324 | * what the new uid/gid are and the callback returned EEXIST. | |
1325 | * The EEXIST error tells us to just use the existing uid/gid. | |
1326 | * If we don't know what the old values are then just assign | |
1327 | * them to 0, since that is a new file being created. | |
1328 | */ | |
1329 | if (!before && data == NULL && error == EEXIST) { | |
1330 | if (flags & DN_ID_OLD_EXIST) { | |
1331 | dn->dn_newuid = dn->dn_olduid; | |
1332 | dn->dn_newgid = dn->dn_oldgid; | |
1333 | } else { | |
1334 | dn->dn_newuid = 0; | |
1335 | dn->dn_newgid = 0; | |
1336 | } | |
1337 | error = 0; | |
1338 | } | |
1339 | ||
1340 | if (db) | |
1341 | mutex_exit(&db->db_mtx); | |
1342 | ||
1343 | mutex_enter(&dn->dn_mtx); | |
1344 | if (error == 0 && before) | |
1345 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
1346 | if (error == 0 && !before) | |
1347 | dn->dn_id_flags |= DN_ID_NEW_EXIST; | |
1348 | ||
1349 | if (have_spill) { | |
1350 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1351 | } else { | |
1352 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1353 | } | |
1354 | mutex_exit(&dn->dn_mtx); | |
1355 | if (have_spill) | |
1356 | dmu_buf_rele((dmu_buf_t *)db, FTAG); | |
1357 | } | |
1358 | ||
9babb374 BB |
1359 | boolean_t |
1360 | dmu_objset_userspace_present(objset_t *os) | |
1361 | { | |
428870ff | 1362 | return (os->os_phys->os_flags & |
9babb374 BB |
1363 | OBJSET_FLAG_USERACCOUNTING_COMPLETE); |
1364 | } | |
1365 | ||
1366 | int | |
1367 | dmu_objset_userspace_upgrade(objset_t *os) | |
1368 | { | |
1369 | uint64_t obj; | |
1370 | int err = 0; | |
1371 | ||
1372 | if (dmu_objset_userspace_present(os)) | |
1373 | return (0); | |
428870ff | 1374 | if (!dmu_objset_userused_enabled(os)) |
2e528b49 | 1375 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 1376 | if (dmu_objset_is_snapshot(os)) |
2e528b49 | 1377 | return (SET_ERROR(EINVAL)); |
9babb374 BB |
1378 | |
1379 | /* | |
1380 | * We simply need to mark every object dirty, so that it will be | |
1381 | * synced out and now accounted. If this is called | |
1382 | * concurrently, or if we already did some work before crashing, | |
1383 | * that's fine, since we track each object's accounted state | |
1384 | * independently. | |
1385 | */ | |
1386 | ||
1387 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
45d1cae3 | 1388 | dmu_tx_t *tx; |
9babb374 BB |
1389 | dmu_buf_t *db; |
1390 | int objerr; | |
1391 | ||
1392 | if (issig(JUSTLOOKING) && issig(FORREAL)) | |
2e528b49 | 1393 | return (SET_ERROR(EINTR)); |
9babb374 BB |
1394 | |
1395 | objerr = dmu_bonus_hold(os, obj, FTAG, &db); | |
13fe0198 | 1396 | if (objerr != 0) |
9babb374 | 1397 | continue; |
45d1cae3 | 1398 | tx = dmu_tx_create(os); |
9babb374 BB |
1399 | dmu_tx_hold_bonus(tx, obj); |
1400 | objerr = dmu_tx_assign(tx, TXG_WAIT); | |
13fe0198 | 1401 | if (objerr != 0) { |
9babb374 BB |
1402 | dmu_tx_abort(tx); |
1403 | continue; | |
1404 | } | |
1405 | dmu_buf_will_dirty(db, tx); | |
1406 | dmu_buf_rele(db, FTAG); | |
1407 | dmu_tx_commit(tx); | |
1408 | } | |
1409 | ||
428870ff | 1410 | os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; |
9babb374 BB |
1411 | txg_wait_synced(dmu_objset_pool(os), 0); |
1412 | return (0); | |
1413 | } | |
1414 | ||
34dc7c2f BB |
1415 | void |
1416 | dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, | |
1417 | uint64_t *usedobjsp, uint64_t *availobjsp) | |
1418 | { | |
428870ff | 1419 | dsl_dataset_space(os->os_dsl_dataset, refdbytesp, availbytesp, |
34dc7c2f BB |
1420 | usedobjsp, availobjsp); |
1421 | } | |
1422 | ||
1423 | uint64_t | |
1424 | dmu_objset_fsid_guid(objset_t *os) | |
1425 | { | |
428870ff | 1426 | return (dsl_dataset_fsid_guid(os->os_dsl_dataset)); |
34dc7c2f BB |
1427 | } |
1428 | ||
1429 | void | |
1430 | dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat) | |
1431 | { | |
428870ff BB |
1432 | stat->dds_type = os->os_phys->os_type; |
1433 | if (os->os_dsl_dataset) | |
1434 | dsl_dataset_fast_stat(os->os_dsl_dataset, stat); | |
34dc7c2f BB |
1435 | } |
1436 | ||
1437 | void | |
1438 | dmu_objset_stats(objset_t *os, nvlist_t *nv) | |
1439 | { | |
428870ff BB |
1440 | ASSERT(os->os_dsl_dataset || |
1441 | os->os_phys->os_type == DMU_OST_META); | |
34dc7c2f | 1442 | |
428870ff BB |
1443 | if (os->os_dsl_dataset != NULL) |
1444 | dsl_dataset_stats(os->os_dsl_dataset, nv); | |
34dc7c2f BB |
1445 | |
1446 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_TYPE, | |
428870ff | 1447 | os->os_phys->os_type); |
9babb374 BB |
1448 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERACCOUNTING, |
1449 | dmu_objset_userspace_present(os)); | |
34dc7c2f BB |
1450 | } |
1451 | ||
1452 | int | |
1453 | dmu_objset_is_snapshot(objset_t *os) | |
1454 | { | |
428870ff BB |
1455 | if (os->os_dsl_dataset != NULL) |
1456 | return (dsl_dataset_is_snapshot(os->os_dsl_dataset)); | |
34dc7c2f BB |
1457 | else |
1458 | return (B_FALSE); | |
1459 | } | |
1460 | ||
1461 | int | |
1462 | dmu_snapshot_realname(objset_t *os, char *name, char *real, int maxlen, | |
1463 | boolean_t *conflict) | |
1464 | { | |
428870ff | 1465 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
1466 | uint64_t ignored; |
1467 | ||
d683ddbb | 1468 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 1469 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1470 | |
1471 | return (zap_lookup_norm(ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb JG |
1472 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, name, 8, 1, &ignored, |
1473 | MT_FIRST, real, maxlen, conflict)); | |
34dc7c2f BB |
1474 | } |
1475 | ||
1476 | int | |
1477 | dmu_snapshot_list_next(objset_t *os, int namelen, char *name, | |
1478 | uint64_t *idp, uint64_t *offp, boolean_t *case_conflict) | |
1479 | { | |
428870ff | 1480 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
1481 | zap_cursor_t cursor; |
1482 | zap_attribute_t attr; | |
1483 | ||
13fe0198 MA |
1484 | ASSERT(dsl_pool_config_held(dmu_objset_pool(os))); |
1485 | ||
d683ddbb | 1486 | if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) |
2e528b49 | 1487 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1488 | |
1489 | zap_cursor_init_serialized(&cursor, | |
1490 | ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb | 1491 | dsl_dataset_phys(ds)->ds_snapnames_zapobj, *offp); |
34dc7c2f BB |
1492 | |
1493 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
1494 | zap_cursor_fini(&cursor); | |
2e528b49 | 1495 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1496 | } |
1497 | ||
1498 | if (strlen(attr.za_name) + 1 > namelen) { | |
1499 | zap_cursor_fini(&cursor); | |
2e528b49 | 1500 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
1501 | } |
1502 | ||
1503 | (void) strcpy(name, attr.za_name); | |
1504 | if (idp) | |
1505 | *idp = attr.za_first_integer; | |
1506 | if (case_conflict) | |
1507 | *case_conflict = attr.za_normalization_conflict; | |
1508 | zap_cursor_advance(&cursor); | |
1509 | *offp = zap_cursor_serialize(&cursor); | |
1510 | zap_cursor_fini(&cursor); | |
1511 | ||
1512 | return (0); | |
1513 | } | |
1514 | ||
ebe7e575 | 1515 | int |
6772fb67 | 1516 | dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *value) |
ebe7e575 | 1517 | { |
d1d7e268 | 1518 | return (dsl_dataset_snap_lookup(os->os_dsl_dataset, name, value)); |
ebe7e575 BB |
1519 | } |
1520 | ||
34dc7c2f BB |
1521 | int |
1522 | dmu_dir_list_next(objset_t *os, int namelen, char *name, | |
1523 | uint64_t *idp, uint64_t *offp) | |
1524 | { | |
428870ff | 1525 | dsl_dir_t *dd = os->os_dsl_dataset->ds_dir; |
34dc7c2f BB |
1526 | zap_cursor_t cursor; |
1527 | zap_attribute_t attr; | |
1528 | ||
1529 | /* there is no next dir on a snapshot! */ | |
428870ff | 1530 | if (os->os_dsl_dataset->ds_object != |
d683ddbb | 1531 | dsl_dir_phys(dd)->dd_head_dataset_obj) |
2e528b49 | 1532 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1533 | |
1534 | zap_cursor_init_serialized(&cursor, | |
1535 | dd->dd_pool->dp_meta_objset, | |
d683ddbb | 1536 | dsl_dir_phys(dd)->dd_child_dir_zapobj, *offp); |
34dc7c2f BB |
1537 | |
1538 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
1539 | zap_cursor_fini(&cursor); | |
2e528b49 | 1540 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1541 | } |
1542 | ||
1543 | if (strlen(attr.za_name) + 1 > namelen) { | |
1544 | zap_cursor_fini(&cursor); | |
2e528b49 | 1545 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
1546 | } |
1547 | ||
1548 | (void) strcpy(name, attr.za_name); | |
1549 | if (idp) | |
1550 | *idp = attr.za_first_integer; | |
1551 | zap_cursor_advance(&cursor); | |
1552 | *offp = zap_cursor_serialize(&cursor); | |
1553 | zap_cursor_fini(&cursor); | |
1554 | ||
1555 | return (0); | |
1556 | } | |
1557 | ||
1558 | /* | |
13fe0198 | 1559 | * Find objsets under and including ddobj, call func(ds) on each. |
34dc7c2f BB |
1560 | */ |
1561 | int | |
13fe0198 MA |
1562 | dmu_objset_find_dp(dsl_pool_t *dp, uint64_t ddobj, |
1563 | int func(dsl_pool_t *, dsl_dataset_t *, void *), void *arg, int flags) | |
b128c09f | 1564 | { |
13fe0198 MA |
1565 | dsl_dir_t *dd; |
1566 | dsl_dataset_t *ds; | |
1567 | zap_cursor_t zc; | |
1568 | zap_attribute_t *attr; | |
1569 | uint64_t thisobj; | |
1570 | int err; | |
1571 | ||
1572 | ASSERT(dsl_pool_config_held(dp)); | |
1573 | ||
1574 | err = dsl_dir_hold_obj(dp, ddobj, NULL, FTAG, &dd); | |
1575 | if (err != 0) | |
1576 | return (err); | |
1577 | ||
1578 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ | |
1579 | if (dd->dd_myname[0] == '$') { | |
1580 | dsl_dir_rele(dd, FTAG); | |
1581 | return (0); | |
1582 | } | |
1583 | ||
d683ddbb | 1584 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 1585 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
13fe0198 MA |
1586 | |
1587 | /* | |
1588 | * Iterate over all children. | |
1589 | */ | |
1590 | if (flags & DS_FIND_CHILDREN) { | |
1591 | for (zap_cursor_init(&zc, dp->dp_meta_objset, | |
d683ddbb | 1592 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
13fe0198 MA |
1593 | zap_cursor_retrieve(&zc, attr) == 0; |
1594 | (void) zap_cursor_advance(&zc)) { | |
1595 | ASSERT3U(attr->za_integer_length, ==, | |
1596 | sizeof (uint64_t)); | |
1597 | ASSERT3U(attr->za_num_integers, ==, 1); | |
1598 | ||
1599 | err = dmu_objset_find_dp(dp, attr->za_first_integer, | |
1600 | func, arg, flags); | |
1601 | if (err != 0) | |
1602 | break; | |
1603 | } | |
1604 | zap_cursor_fini(&zc); | |
1605 | ||
1606 | if (err != 0) { | |
1607 | dsl_dir_rele(dd, FTAG); | |
1608 | kmem_free(attr, sizeof (zap_attribute_t)); | |
1609 | return (err); | |
1610 | } | |
1611 | } | |
1612 | ||
1613 | /* | |
1614 | * Iterate over all snapshots. | |
1615 | */ | |
1616 | if (flags & DS_FIND_SNAPSHOTS) { | |
1617 | dsl_dataset_t *ds; | |
1618 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1619 | ||
1620 | if (err == 0) { | |
d683ddbb JG |
1621 | uint64_t snapobj; |
1622 | ||
1623 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
13fe0198 MA |
1624 | dsl_dataset_rele(ds, FTAG); |
1625 | ||
1626 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
1627 | zap_cursor_retrieve(&zc, attr) == 0; | |
1628 | (void) zap_cursor_advance(&zc)) { | |
1629 | ASSERT3U(attr->za_integer_length, ==, | |
1630 | sizeof (uint64_t)); | |
1631 | ASSERT3U(attr->za_num_integers, ==, 1); | |
1632 | ||
1633 | err = dsl_dataset_hold_obj(dp, | |
1634 | attr->za_first_integer, FTAG, &ds); | |
1635 | if (err != 0) | |
1636 | break; | |
1637 | err = func(dp, ds, arg); | |
1638 | dsl_dataset_rele(ds, FTAG); | |
1639 | if (err != 0) | |
1640 | break; | |
1641 | } | |
1642 | zap_cursor_fini(&zc); | |
1643 | } | |
1644 | } | |
1645 | ||
1646 | dsl_dir_rele(dd, FTAG); | |
1647 | kmem_free(attr, sizeof (zap_attribute_t)); | |
1648 | ||
1649 | if (err != 0) | |
1650 | return (err); | |
1651 | ||
1652 | /* | |
1653 | * Apply to self. | |
1654 | */ | |
1655 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1656 | if (err != 0) | |
1657 | return (err); | |
1658 | err = func(dp, ds, arg); | |
1659 | dsl_dataset_rele(ds, FTAG); | |
1660 | return (err); | |
b128c09f BB |
1661 | } |
1662 | ||
1663 | /* | |
13fe0198 MA |
1664 | * Find all objsets under name, and for each, call 'func(child_name, arg)'. |
1665 | * The dp_config_rwlock must not be held when this is called, and it | |
1666 | * will not be held when the callback is called. | |
1667 | * Therefore this function should only be used when the pool is not changing | |
1668 | * (e.g. in syncing context), or the callback can deal with the possible races. | |
b128c09f | 1669 | */ |
13fe0198 MA |
1670 | static int |
1671 | dmu_objset_find_impl(spa_t *spa, const char *name, | |
1672 | int func(const char *, void *), void *arg, int flags) | |
34dc7c2f BB |
1673 | { |
1674 | dsl_dir_t *dd; | |
13fe0198 | 1675 | dsl_pool_t *dp = spa_get_dsl(spa); |
b128c09f | 1676 | dsl_dataset_t *ds; |
34dc7c2f BB |
1677 | zap_cursor_t zc; |
1678 | zap_attribute_t *attr; | |
1679 | char *child; | |
b128c09f BB |
1680 | uint64_t thisobj; |
1681 | int err; | |
34dc7c2f | 1682 | |
13fe0198 MA |
1683 | dsl_pool_config_enter(dp, FTAG); |
1684 | ||
1685 | err = dsl_dir_hold(dp, name, FTAG, &dd, NULL); | |
1686 | if (err != 0) { | |
1687 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f | 1688 | return (err); |
13fe0198 | 1689 | } |
34dc7c2f | 1690 | |
b128c09f BB |
1691 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ |
1692 | if (dd->dd_myname[0] == '$') { | |
13fe0198 MA |
1693 | dsl_dir_rele(dd, FTAG); |
1694 | dsl_pool_config_exit(dp, FTAG); | |
b128c09f BB |
1695 | return (0); |
1696 | } | |
1697 | ||
d683ddbb | 1698 | thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
79c76d5b | 1699 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); |
34dc7c2f BB |
1700 | |
1701 | /* | |
1702 | * Iterate over all children. | |
1703 | */ | |
1704 | if (flags & DS_FIND_CHILDREN) { | |
b128c09f | 1705 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
d683ddbb | 1706 | dsl_dir_phys(dd)->dd_child_dir_zapobj); |
34dc7c2f BB |
1707 | zap_cursor_retrieve(&zc, attr) == 0; |
1708 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 MA |
1709 | ASSERT3U(attr->za_integer_length, ==, |
1710 | sizeof (uint64_t)); | |
1711 | ASSERT3U(attr->za_num_integers, ==, 1); | |
34dc7c2f | 1712 | |
428870ff | 1713 | child = kmem_asprintf("%s/%s", name, attr->za_name); |
13fe0198 MA |
1714 | dsl_pool_config_exit(dp, FTAG); |
1715 | err = dmu_objset_find_impl(spa, child, | |
1716 | func, arg, flags); | |
1717 | dsl_pool_config_enter(dp, FTAG); | |
428870ff | 1718 | strfree(child); |
13fe0198 | 1719 | if (err != 0) |
34dc7c2f BB |
1720 | break; |
1721 | } | |
1722 | zap_cursor_fini(&zc); | |
1723 | ||
13fe0198 MA |
1724 | if (err != 0) { |
1725 | dsl_dir_rele(dd, FTAG); | |
1726 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f BB |
1727 | kmem_free(attr, sizeof (zap_attribute_t)); |
1728 | return (err); | |
1729 | } | |
1730 | } | |
1731 | ||
1732 | /* | |
1733 | * Iterate over all snapshots. | |
1734 | */ | |
b128c09f | 1735 | if (flags & DS_FIND_SNAPSHOTS) { |
b128c09f | 1736 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); |
b128c09f BB |
1737 | |
1738 | if (err == 0) { | |
d683ddbb JG |
1739 | uint64_t snapobj; |
1740 | ||
1741 | snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; | |
b128c09f BB |
1742 | dsl_dataset_rele(ds, FTAG); |
1743 | ||
1744 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
1745 | zap_cursor_retrieve(&zc, attr) == 0; | |
1746 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 | 1747 | ASSERT3U(attr->za_integer_length, ==, |
b128c09f | 1748 | sizeof (uint64_t)); |
13fe0198 | 1749 | ASSERT3U(attr->za_num_integers, ==, 1); |
b128c09f | 1750 | |
428870ff BB |
1751 | child = kmem_asprintf("%s@%s", |
1752 | name, attr->za_name); | |
13fe0198 MA |
1753 | dsl_pool_config_exit(dp, FTAG); |
1754 | err = func(child, arg); | |
1755 | dsl_pool_config_enter(dp, FTAG); | |
428870ff | 1756 | strfree(child); |
13fe0198 | 1757 | if (err != 0) |
b128c09f BB |
1758 | break; |
1759 | } | |
1760 | zap_cursor_fini(&zc); | |
34dc7c2f | 1761 | } |
34dc7c2f BB |
1762 | } |
1763 | ||
13fe0198 | 1764 | dsl_dir_rele(dd, FTAG); |
34dc7c2f | 1765 | kmem_free(attr, sizeof (zap_attribute_t)); |
13fe0198 | 1766 | dsl_pool_config_exit(dp, FTAG); |
34dc7c2f | 1767 | |
13fe0198 | 1768 | if (err != 0) |
34dc7c2f BB |
1769 | return (err); |
1770 | ||
13fe0198 MA |
1771 | /* Apply to self. */ |
1772 | return (func(name, arg)); | |
34dc7c2f BB |
1773 | } |
1774 | ||
13fe0198 MA |
1775 | /* |
1776 | * See comment above dmu_objset_find_impl(). | |
1777 | */ | |
d164b209 | 1778 | int |
13fe0198 MA |
1779 | dmu_objset_find(char *name, int func(const char *, void *), void *arg, |
1780 | int flags) | |
d164b209 | 1781 | { |
13fe0198 MA |
1782 | spa_t *spa; |
1783 | int error; | |
d164b209 | 1784 | |
13fe0198 MA |
1785 | error = spa_open(name, &spa, FTAG); |
1786 | if (error != 0) | |
1787 | return (error); | |
1788 | error = dmu_objset_find_impl(spa, name, func, arg, flags); | |
1789 | spa_close(spa, FTAG); | |
1790 | return (error); | |
d164b209 BB |
1791 | } |
1792 | ||
34dc7c2f BB |
1793 | void |
1794 | dmu_objset_set_user(objset_t *os, void *user_ptr) | |
1795 | { | |
428870ff BB |
1796 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
1797 | os->os_user_ptr = user_ptr; | |
34dc7c2f BB |
1798 | } |
1799 | ||
1800 | void * | |
1801 | dmu_objset_get_user(objset_t *os) | |
1802 | { | |
428870ff BB |
1803 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
1804 | return (os->os_user_ptr); | |
34dc7c2f | 1805 | } |
c28b2279 | 1806 | |
13fe0198 MA |
1807 | /* |
1808 | * Determine name of filesystem, given name of snapshot. | |
1809 | * buf must be at least MAXNAMELEN bytes | |
1810 | */ | |
1811 | int | |
1812 | dmu_fsname(const char *snapname, char *buf) | |
1813 | { | |
1814 | char *atp = strchr(snapname, '@'); | |
1815 | if (atp == NULL) | |
2e528b49 | 1816 | return (SET_ERROR(EINVAL)); |
13fe0198 | 1817 | if (atp - snapname >= MAXNAMELEN) |
2e528b49 | 1818 | return (SET_ERROR(ENAMETOOLONG)); |
13fe0198 MA |
1819 | (void) strlcpy(buf, snapname, atp - snapname + 1); |
1820 | return (0); | |
1821 | } | |
1822 | ||
c28b2279 | 1823 | #if defined(_KERNEL) && defined(HAVE_SPL) |
f0fd83be | 1824 | EXPORT_SYMBOL(dmu_objset_zil); |
c28b2279 | 1825 | EXPORT_SYMBOL(dmu_objset_pool); |
f0fd83be BB |
1826 | EXPORT_SYMBOL(dmu_objset_ds); |
1827 | EXPORT_SYMBOL(dmu_objset_type); | |
c28b2279 BB |
1828 | EXPORT_SYMBOL(dmu_objset_name); |
1829 | EXPORT_SYMBOL(dmu_objset_hold); | |
1830 | EXPORT_SYMBOL(dmu_objset_own); | |
1831 | EXPORT_SYMBOL(dmu_objset_rele); | |
1832 | EXPORT_SYMBOL(dmu_objset_disown); | |
1833 | EXPORT_SYMBOL(dmu_objset_from_ds); | |
1834 | EXPORT_SYMBOL(dmu_objset_create); | |
1835 | EXPORT_SYMBOL(dmu_objset_clone); | |
c28b2279 BB |
1836 | EXPORT_SYMBOL(dmu_objset_stats); |
1837 | EXPORT_SYMBOL(dmu_objset_fast_stat); | |
54a179e7 | 1838 | EXPORT_SYMBOL(dmu_objset_spa); |
c28b2279 BB |
1839 | EXPORT_SYMBOL(dmu_objset_space); |
1840 | EXPORT_SYMBOL(dmu_objset_fsid_guid); | |
1841 | EXPORT_SYMBOL(dmu_objset_find); | |
c28b2279 BB |
1842 | EXPORT_SYMBOL(dmu_objset_byteswap); |
1843 | EXPORT_SYMBOL(dmu_objset_evict_dbufs); | |
1844 | EXPORT_SYMBOL(dmu_objset_snap_cmtime); | |
1845 | ||
1846 | EXPORT_SYMBOL(dmu_objset_sync); | |
1847 | EXPORT_SYMBOL(dmu_objset_is_dirty); | |
1848 | EXPORT_SYMBOL(dmu_objset_create_impl); | |
1849 | EXPORT_SYMBOL(dmu_objset_open_impl); | |
1850 | EXPORT_SYMBOL(dmu_objset_evict); | |
1851 | EXPORT_SYMBOL(dmu_objset_register_type); | |
1852 | EXPORT_SYMBOL(dmu_objset_do_userquota_updates); | |
1853 | EXPORT_SYMBOL(dmu_objset_userquota_get_ids); | |
1854 | EXPORT_SYMBOL(dmu_objset_userused_enabled); | |
1855 | EXPORT_SYMBOL(dmu_objset_userspace_upgrade); | |
1856 | EXPORT_SYMBOL(dmu_objset_userspace_present); | |
1857 | #endif |