]>
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. |
3a17a7a9 | 23 | * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. |
34dc7c2f BB |
24 | */ |
25 | ||
428870ff BB |
26 | /* Portions Copyright 2010 Robert Milkowski */ |
27 | ||
34dc7c2f BB |
28 | #include <sys/cred.h> |
29 | #include <sys/zfs_context.h> | |
30 | #include <sys/dmu_objset.h> | |
31 | #include <sys/dsl_dir.h> | |
32 | #include <sys/dsl_dataset.h> | |
33 | #include <sys/dsl_prop.h> | |
34 | #include <sys/dsl_pool.h> | |
35 | #include <sys/dsl_synctask.h> | |
36 | #include <sys/dsl_deleg.h> | |
37 | #include <sys/dnode.h> | |
38 | #include <sys/dbuf.h> | |
39 | #include <sys/zvol.h> | |
40 | #include <sys/dmu_tx.h> | |
34dc7c2f BB |
41 | #include <sys/zap.h> |
42 | #include <sys/zil.h> | |
43 | #include <sys/dmu_impl.h> | |
44 | #include <sys/zfs_ioctl.h> | |
428870ff | 45 | #include <sys/sa.h> |
572e2857 BB |
46 | #include <sys/zfs_onexit.h> |
47 | ||
48 | /* | |
49 | * Needed to close a window in dnode_move() that allows the objset to be freed | |
50 | * before it can be safely accessed. | |
51 | */ | |
52 | krwlock_t os_lock; | |
53 | ||
54 | void | |
55 | dmu_objset_init(void) | |
56 | { | |
57 | rw_init(&os_lock, NULL, RW_DEFAULT, NULL); | |
58 | } | |
59 | ||
60 | void | |
61 | dmu_objset_fini(void) | |
62 | { | |
63 | rw_destroy(&os_lock); | |
64 | } | |
34dc7c2f BB |
65 | |
66 | spa_t * | |
67 | dmu_objset_spa(objset_t *os) | |
68 | { | |
428870ff | 69 | return (os->os_spa); |
34dc7c2f BB |
70 | } |
71 | ||
72 | zilog_t * | |
73 | dmu_objset_zil(objset_t *os) | |
74 | { | |
428870ff | 75 | return (os->os_zil); |
34dc7c2f BB |
76 | } |
77 | ||
78 | dsl_pool_t * | |
79 | dmu_objset_pool(objset_t *os) | |
80 | { | |
81 | dsl_dataset_t *ds; | |
82 | ||
428870ff | 83 | if ((ds = os->os_dsl_dataset) != NULL && ds->ds_dir) |
34dc7c2f BB |
84 | return (ds->ds_dir->dd_pool); |
85 | else | |
428870ff | 86 | return (spa_get_dsl(os->os_spa)); |
34dc7c2f BB |
87 | } |
88 | ||
89 | dsl_dataset_t * | |
90 | dmu_objset_ds(objset_t *os) | |
91 | { | |
428870ff | 92 | return (os->os_dsl_dataset); |
34dc7c2f BB |
93 | } |
94 | ||
95 | dmu_objset_type_t | |
96 | dmu_objset_type(objset_t *os) | |
97 | { | |
428870ff | 98 | return (os->os_phys->os_type); |
34dc7c2f BB |
99 | } |
100 | ||
101 | void | |
102 | dmu_objset_name(objset_t *os, char *buf) | |
103 | { | |
428870ff | 104 | dsl_dataset_name(os->os_dsl_dataset, buf); |
34dc7c2f BB |
105 | } |
106 | ||
107 | uint64_t | |
108 | dmu_objset_id(objset_t *os) | |
109 | { | |
428870ff | 110 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
111 | |
112 | return (ds ? ds->ds_object : 0); | |
113 | } | |
114 | ||
428870ff BB |
115 | uint64_t |
116 | dmu_objset_syncprop(objset_t *os) | |
117 | { | |
118 | return (os->os_sync); | |
119 | } | |
120 | ||
121 | uint64_t | |
122 | dmu_objset_logbias(objset_t *os) | |
123 | { | |
124 | return (os->os_logbias); | |
125 | } | |
126 | ||
34dc7c2f BB |
127 | static void |
128 | checksum_changed_cb(void *arg, uint64_t newval) | |
129 | { | |
428870ff | 130 | objset_t *os = arg; |
34dc7c2f BB |
131 | |
132 | /* | |
133 | * Inheritance should have been done by now. | |
134 | */ | |
135 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
136 | ||
428870ff | 137 | os->os_checksum = zio_checksum_select(newval, ZIO_CHECKSUM_ON_VALUE); |
34dc7c2f BB |
138 | } |
139 | ||
140 | static void | |
141 | compression_changed_cb(void *arg, uint64_t newval) | |
142 | { | |
428870ff | 143 | objset_t *os = arg; |
34dc7c2f BB |
144 | |
145 | /* | |
146 | * Inheritance and range checking should have been done by now. | |
147 | */ | |
148 | ASSERT(newval != ZIO_COMPRESS_INHERIT); | |
149 | ||
428870ff | 150 | os->os_compress = zio_compress_select(newval, ZIO_COMPRESS_ON_VALUE); |
34dc7c2f BB |
151 | } |
152 | ||
153 | static void | |
154 | copies_changed_cb(void *arg, uint64_t newval) | |
155 | { | |
428870ff | 156 | objset_t *os = arg; |
34dc7c2f BB |
157 | |
158 | /* | |
159 | * Inheritance and range checking should have been done by now. | |
160 | */ | |
161 | ASSERT(newval > 0); | |
428870ff | 162 | ASSERT(newval <= spa_max_replication(os->os_spa)); |
34dc7c2f | 163 | |
428870ff BB |
164 | os->os_copies = newval; |
165 | } | |
166 | ||
167 | static void | |
168 | dedup_changed_cb(void *arg, uint64_t newval) | |
169 | { | |
170 | objset_t *os = arg; | |
171 | spa_t *spa = os->os_spa; | |
172 | enum zio_checksum checksum; | |
173 | ||
174 | /* | |
175 | * Inheritance should have been done by now. | |
176 | */ | |
177 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
178 | ||
179 | checksum = zio_checksum_dedup_select(spa, newval, ZIO_CHECKSUM_OFF); | |
180 | ||
181 | os->os_dedup_checksum = checksum & ZIO_CHECKSUM_MASK; | |
182 | os->os_dedup_verify = !!(checksum & ZIO_CHECKSUM_VERIFY); | |
34dc7c2f BB |
183 | } |
184 | ||
b128c09f BB |
185 | static void |
186 | primary_cache_changed_cb(void *arg, uint64_t newval) | |
187 | { | |
428870ff | 188 | objset_t *os = arg; |
b128c09f BB |
189 | |
190 | /* | |
191 | * Inheritance and range checking should have been done by now. | |
192 | */ | |
193 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
194 | newval == ZFS_CACHE_METADATA); | |
195 | ||
428870ff | 196 | os->os_primary_cache = newval; |
b128c09f BB |
197 | } |
198 | ||
199 | static void | |
200 | secondary_cache_changed_cb(void *arg, uint64_t newval) | |
201 | { | |
428870ff | 202 | objset_t *os = arg; |
b128c09f BB |
203 | |
204 | /* | |
205 | * Inheritance and range checking should have been done by now. | |
206 | */ | |
207 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
208 | newval == ZFS_CACHE_METADATA); | |
209 | ||
428870ff BB |
210 | os->os_secondary_cache = newval; |
211 | } | |
212 | ||
213 | static void | |
214 | sync_changed_cb(void *arg, uint64_t newval) | |
215 | { | |
216 | objset_t *os = arg; | |
217 | ||
218 | /* | |
219 | * Inheritance and range checking should have been done by now. | |
220 | */ | |
221 | ASSERT(newval == ZFS_SYNC_STANDARD || newval == ZFS_SYNC_ALWAYS || | |
222 | newval == ZFS_SYNC_DISABLED); | |
223 | ||
224 | os->os_sync = newval; | |
225 | if (os->os_zil) | |
226 | zil_set_sync(os->os_zil, newval); | |
227 | } | |
228 | ||
229 | static void | |
230 | logbias_changed_cb(void *arg, uint64_t newval) | |
231 | { | |
232 | objset_t *os = arg; | |
233 | ||
234 | ASSERT(newval == ZFS_LOGBIAS_LATENCY || | |
235 | newval == ZFS_LOGBIAS_THROUGHPUT); | |
236 | os->os_logbias = newval; | |
237 | if (os->os_zil) | |
238 | zil_set_logbias(os->os_zil, newval); | |
b128c09f BB |
239 | } |
240 | ||
34dc7c2f BB |
241 | void |
242 | dmu_objset_byteswap(void *buf, size_t size) | |
243 | { | |
244 | objset_phys_t *osp = buf; | |
245 | ||
9babb374 | 246 | ASSERT(size == OBJSET_OLD_PHYS_SIZE || size == sizeof (objset_phys_t)); |
34dc7c2f BB |
247 | dnode_byteswap(&osp->os_meta_dnode); |
248 | byteswap_uint64_array(&osp->os_zil_header, sizeof (zil_header_t)); | |
249 | osp->os_type = BSWAP_64(osp->os_type); | |
9babb374 BB |
250 | osp->os_flags = BSWAP_64(osp->os_flags); |
251 | if (size == sizeof (objset_phys_t)) { | |
252 | dnode_byteswap(&osp->os_userused_dnode); | |
253 | dnode_byteswap(&osp->os_groupused_dnode); | |
254 | } | |
34dc7c2f BB |
255 | } |
256 | ||
257 | int | |
258 | dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, | |
428870ff | 259 | objset_t **osp) |
34dc7c2f | 260 | { |
428870ff | 261 | objset_t *os; |
b128c09f | 262 | int i, err; |
34dc7c2f BB |
263 | |
264 | ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock)); | |
265 | ||
b8d06fca | 266 | os = kmem_zalloc(sizeof (objset_t), KM_PUSHPAGE); |
428870ff BB |
267 | os->os_dsl_dataset = ds; |
268 | os->os_spa = spa; | |
269 | os->os_rootbp = bp; | |
270 | if (!BP_IS_HOLE(os->os_rootbp)) { | |
34dc7c2f BB |
271 | uint32_t aflags = ARC_WAIT; |
272 | zbookmark_t zb; | |
428870ff BB |
273 | SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET, |
274 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
275 | ||
276 | if (DMU_OS_IS_L2CACHEABLE(os)) | |
b128c09f | 277 | aflags |= ARC_L2CACHE; |
3a17a7a9 SK |
278 | if (DMU_OS_IS_L2COMPRESSIBLE(os)) |
279 | aflags |= ARC_L2COMPRESS; | |
34dc7c2f | 280 | |
428870ff | 281 | dprintf_bp(os->os_rootbp, "reading %s", ""); |
294f6806 | 282 | err = arc_read(NULL, spa, os->os_rootbp, |
428870ff | 283 | arc_getbuf_func, &os->os_phys_buf, |
34dc7c2f BB |
284 | ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL, &aflags, &zb); |
285 | if (err) { | |
428870ff | 286 | kmem_free(os, sizeof (objset_t)); |
b128c09f BB |
287 | /* convert checksum errors into IO errors */ |
288 | if (err == ECKSUM) | |
289 | err = EIO; | |
34dc7c2f BB |
290 | return (err); |
291 | } | |
9babb374 BB |
292 | |
293 | /* Increase the blocksize if we are permitted. */ | |
294 | if (spa_version(spa) >= SPA_VERSION_USERSPACE && | |
428870ff | 295 | arc_buf_size(os->os_phys_buf) < sizeof (objset_phys_t)) { |
9babb374 | 296 | arc_buf_t *buf = arc_buf_alloc(spa, |
428870ff | 297 | sizeof (objset_phys_t), &os->os_phys_buf, |
9babb374 BB |
298 | ARC_BUFC_METADATA); |
299 | bzero(buf->b_data, sizeof (objset_phys_t)); | |
428870ff BB |
300 | bcopy(os->os_phys_buf->b_data, buf->b_data, |
301 | arc_buf_size(os->os_phys_buf)); | |
302 | (void) arc_buf_remove_ref(os->os_phys_buf, | |
303 | &os->os_phys_buf); | |
304 | os->os_phys_buf = buf; | |
9babb374 BB |
305 | } |
306 | ||
428870ff BB |
307 | os->os_phys = os->os_phys_buf->b_data; |
308 | os->os_flags = os->os_phys->os_flags; | |
34dc7c2f | 309 | } else { |
9babb374 BB |
310 | int size = spa_version(spa) >= SPA_VERSION_USERSPACE ? |
311 | sizeof (objset_phys_t) : OBJSET_OLD_PHYS_SIZE; | |
428870ff BB |
312 | os->os_phys_buf = arc_buf_alloc(spa, size, |
313 | &os->os_phys_buf, ARC_BUFC_METADATA); | |
314 | os->os_phys = os->os_phys_buf->b_data; | |
315 | bzero(os->os_phys, size); | |
34dc7c2f BB |
316 | } |
317 | ||
318 | /* | |
319 | * Note: the changed_cb will be called once before the register | |
320 | * func returns, thus changing the checksum/compression from the | |
b128c09f BB |
321 | * default (fletcher2/off). Snapshots don't need to know about |
322 | * checksum/compression/copies. | |
34dc7c2f | 323 | */ |
b128c09f BB |
324 | if (ds) { |
325 | err = dsl_prop_register(ds, "primarycache", | |
428870ff | 326 | primary_cache_changed_cb, os); |
34dc7c2f | 327 | if (err == 0) |
b128c09f | 328 | err = dsl_prop_register(ds, "secondarycache", |
428870ff | 329 | secondary_cache_changed_cb, os); |
b128c09f BB |
330 | if (!dsl_dataset_is_snapshot(ds)) { |
331 | if (err == 0) | |
332 | err = dsl_prop_register(ds, "checksum", | |
428870ff | 333 | checksum_changed_cb, os); |
b128c09f BB |
334 | if (err == 0) |
335 | err = dsl_prop_register(ds, "compression", | |
428870ff | 336 | compression_changed_cb, os); |
b128c09f BB |
337 | if (err == 0) |
338 | err = dsl_prop_register(ds, "copies", | |
428870ff BB |
339 | copies_changed_cb, os); |
340 | if (err == 0) | |
341 | err = dsl_prop_register(ds, "dedup", | |
342 | dedup_changed_cb, os); | |
343 | if (err == 0) | |
344 | err = dsl_prop_register(ds, "logbias", | |
345 | logbias_changed_cb, os); | |
346 | if (err == 0) | |
347 | err = dsl_prop_register(ds, "sync", | |
348 | sync_changed_cb, os); | |
b128c09f | 349 | } |
34dc7c2f | 350 | if (err) { |
428870ff BB |
351 | VERIFY(arc_buf_remove_ref(os->os_phys_buf, |
352 | &os->os_phys_buf) == 1); | |
353 | kmem_free(os, sizeof (objset_t)); | |
34dc7c2f BB |
354 | return (err); |
355 | } | |
356 | } else if (ds == NULL) { | |
357 | /* It's the meta-objset. */ | |
428870ff BB |
358 | os->os_checksum = ZIO_CHECKSUM_FLETCHER_4; |
359 | os->os_compress = ZIO_COMPRESS_LZJB; | |
360 | os->os_copies = spa_max_replication(spa); | |
361 | os->os_dedup_checksum = ZIO_CHECKSUM_OFF; | |
362 | os->os_dedup_verify = 0; | |
363 | os->os_logbias = 0; | |
364 | os->os_sync = 0; | |
365 | os->os_primary_cache = ZFS_CACHE_ALL; | |
366 | os->os_secondary_cache = ZFS_CACHE_ALL; | |
34dc7c2f BB |
367 | } |
368 | ||
572e2857 BB |
369 | if (ds == NULL || !dsl_dataset_is_snapshot(ds)) |
370 | os->os_zil_header = os->os_phys->os_zil_header; | |
428870ff | 371 | os->os_zil = zil_alloc(os, &os->os_zil_header); |
34dc7c2f BB |
372 | |
373 | for (i = 0; i < TXG_SIZE; i++) { | |
428870ff | 374 | list_create(&os->os_dirty_dnodes[i], sizeof (dnode_t), |
34dc7c2f | 375 | offsetof(dnode_t, dn_dirty_link[i])); |
428870ff | 376 | list_create(&os->os_free_dnodes[i], sizeof (dnode_t), |
34dc7c2f BB |
377 | offsetof(dnode_t, dn_dirty_link[i])); |
378 | } | |
428870ff | 379 | list_create(&os->os_dnodes, sizeof (dnode_t), |
34dc7c2f | 380 | offsetof(dnode_t, dn_link)); |
428870ff | 381 | list_create(&os->os_downgraded_dbufs, sizeof (dmu_buf_impl_t), |
34dc7c2f BB |
382 | offsetof(dmu_buf_impl_t, db_link)); |
383 | ||
428870ff BB |
384 | mutex_init(&os->os_lock, NULL, MUTEX_DEFAULT, NULL); |
385 | mutex_init(&os->os_obj_lock, NULL, MUTEX_DEFAULT, NULL); | |
386 | mutex_init(&os->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL); | |
387 | ||
572e2857 BB |
388 | DMU_META_DNODE(os) = dnode_special_open(os, |
389 | &os->os_phys->os_meta_dnode, DMU_META_DNODE_OBJECT, | |
390 | &os->os_meta_dnode); | |
428870ff | 391 | if (arc_buf_size(os->os_phys_buf) >= sizeof (objset_phys_t)) { |
572e2857 BB |
392 | DMU_USERUSED_DNODE(os) = dnode_special_open(os, |
393 | &os->os_phys->os_userused_dnode, DMU_USERUSED_OBJECT, | |
394 | &os->os_userused_dnode); | |
395 | DMU_GROUPUSED_DNODE(os) = dnode_special_open(os, | |
396 | &os->os_phys->os_groupused_dnode, DMU_GROUPUSED_OBJECT, | |
397 | &os->os_groupused_dnode); | |
9babb374 | 398 | } |
34dc7c2f BB |
399 | |
400 | /* | |
401 | * We should be the only thread trying to do this because we | |
402 | * have ds_opening_lock | |
403 | */ | |
404 | if (ds) { | |
428870ff BB |
405 | mutex_enter(&ds->ds_lock); |
406 | ASSERT(ds->ds_objset == NULL); | |
407 | ds->ds_objset = os; | |
408 | mutex_exit(&ds->ds_lock); | |
34dc7c2f BB |
409 | } |
410 | ||
428870ff | 411 | *osp = os; |
34dc7c2f BB |
412 | return (0); |
413 | } | |
414 | ||
428870ff BB |
415 | int |
416 | dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp) | |
34dc7c2f | 417 | { |
428870ff | 418 | int err = 0; |
34dc7c2f BB |
419 | |
420 | mutex_enter(&ds->ds_opening_lock); | |
428870ff BB |
421 | *osp = ds->ds_objset; |
422 | if (*osp == NULL) { | |
34dc7c2f | 423 | err = dmu_objset_open_impl(dsl_dataset_get_spa(ds), |
572e2857 | 424 | ds, dsl_dataset_get_blkptr(ds), osp); |
34dc7c2f BB |
425 | } |
426 | mutex_exit(&ds->ds_opening_lock); | |
428870ff | 427 | return (err); |
34dc7c2f BB |
428 | } |
429 | ||
428870ff | 430 | /* called from zpl */ |
34dc7c2f | 431 | int |
428870ff | 432 | dmu_objset_hold(const char *name, void *tag, objset_t **osp) |
34dc7c2f | 433 | { |
428870ff | 434 | dsl_dataset_t *ds; |
34dc7c2f BB |
435 | int err; |
436 | ||
428870ff | 437 | err = dsl_dataset_hold(name, tag, &ds); |
34dc7c2f | 438 | if (err) |
428870ff BB |
439 | return (err); |
440 | ||
441 | err = dmu_objset_from_ds(ds, osp); | |
442 | if (err) | |
443 | dsl_dataset_rele(ds, tag); | |
444 | ||
34dc7c2f BB |
445 | return (err); |
446 | } | |
447 | ||
448 | /* called from zpl */ | |
449 | int | |
428870ff BB |
450 | dmu_objset_own(const char *name, dmu_objset_type_t type, |
451 | boolean_t readonly, void *tag, objset_t **osp) | |
34dc7c2f | 452 | { |
34dc7c2f BB |
453 | dsl_dataset_t *ds; |
454 | int err; | |
455 | ||
428870ff BB |
456 | err = dsl_dataset_own(name, B_FALSE, tag, &ds); |
457 | if (err) | |
34dc7c2f | 458 | return (err); |
34dc7c2f | 459 | |
428870ff | 460 | err = dmu_objset_from_ds(ds, osp); |
34dc7c2f | 461 | if (err) { |
428870ff BB |
462 | dsl_dataset_disown(ds, tag); |
463 | } else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) { | |
464 | dmu_objset_disown(*osp, tag); | |
465 | return (EINVAL); | |
466 | } else if (!readonly && dsl_dataset_is_snapshot(ds)) { | |
467 | dmu_objset_disown(*osp, tag); | |
468 | return (EROFS); | |
34dc7c2f BB |
469 | } |
470 | return (err); | |
471 | } | |
472 | ||
473 | void | |
428870ff | 474 | dmu_objset_rele(objset_t *os, void *tag) |
34dc7c2f | 475 | { |
428870ff BB |
476 | dsl_dataset_rele(os->os_dsl_dataset, tag); |
477 | } | |
b128c09f | 478 | |
428870ff BB |
479 | void |
480 | dmu_objset_disown(objset_t *os, void *tag) | |
481 | { | |
482 | dsl_dataset_disown(os->os_dsl_dataset, tag); | |
34dc7c2f BB |
483 | } |
484 | ||
485 | int | |
486 | dmu_objset_evict_dbufs(objset_t *os) | |
487 | { | |
34dc7c2f BB |
488 | dnode_t *dn; |
489 | ||
428870ff | 490 | mutex_enter(&os->os_lock); |
34dc7c2f BB |
491 | |
492 | /* process the mdn last, since the other dnodes have holds on it */ | |
572e2857 BB |
493 | list_remove(&os->os_dnodes, DMU_META_DNODE(os)); |
494 | list_insert_tail(&os->os_dnodes, DMU_META_DNODE(os)); | |
34dc7c2f BB |
495 | |
496 | /* | |
497 | * Find the first dnode with holds. We have to do this dance | |
498 | * because dnode_add_ref() only works if you already have a | |
499 | * hold. If there are no holds then it has no dbufs so OK to | |
500 | * skip. | |
501 | */ | |
428870ff | 502 | for (dn = list_head(&os->os_dnodes); |
34dc7c2f | 503 | dn && !dnode_add_ref(dn, FTAG); |
428870ff | 504 | dn = list_next(&os->os_dnodes, dn)) |
34dc7c2f BB |
505 | continue; |
506 | ||
507 | while (dn) { | |
508 | dnode_t *next_dn = dn; | |
509 | ||
510 | do { | |
428870ff | 511 | next_dn = list_next(&os->os_dnodes, next_dn); |
34dc7c2f BB |
512 | } while (next_dn && !dnode_add_ref(next_dn, FTAG)); |
513 | ||
428870ff | 514 | mutex_exit(&os->os_lock); |
34dc7c2f BB |
515 | dnode_evict_dbufs(dn); |
516 | dnode_rele(dn, FTAG); | |
428870ff | 517 | mutex_enter(&os->os_lock); |
34dc7c2f BB |
518 | dn = next_dn; |
519 | } | |
572e2857 | 520 | dn = list_head(&os->os_dnodes); |
428870ff | 521 | mutex_exit(&os->os_lock); |
572e2857 | 522 | return (dn != DMU_META_DNODE(os)); |
34dc7c2f BB |
523 | } |
524 | ||
525 | void | |
428870ff | 526 | dmu_objset_evict(objset_t *os) |
34dc7c2f | 527 | { |
428870ff | 528 | dsl_dataset_t *ds = os->os_dsl_dataset; |
d6320ddb | 529 | int t; |
34dc7c2f | 530 | |
d6320ddb | 531 | for (t = 0; t < TXG_SIZE; t++) |
428870ff | 532 | ASSERT(!dmu_objset_is_dirty(os, t)); |
34dc7c2f | 533 | |
b128c09f BB |
534 | if (ds) { |
535 | if (!dsl_dataset_is_snapshot(ds)) { | |
536 | VERIFY(0 == dsl_prop_unregister(ds, "checksum", | |
428870ff | 537 | checksum_changed_cb, os)); |
b128c09f | 538 | VERIFY(0 == dsl_prop_unregister(ds, "compression", |
428870ff | 539 | compression_changed_cb, os)); |
b128c09f | 540 | VERIFY(0 == dsl_prop_unregister(ds, "copies", |
428870ff BB |
541 | copies_changed_cb, os)); |
542 | VERIFY(0 == dsl_prop_unregister(ds, "dedup", | |
543 | dedup_changed_cb, os)); | |
544 | VERIFY(0 == dsl_prop_unregister(ds, "logbias", | |
545 | logbias_changed_cb, os)); | |
546 | VERIFY(0 == dsl_prop_unregister(ds, "sync", | |
547 | sync_changed_cb, os)); | |
b128c09f BB |
548 | } |
549 | VERIFY(0 == dsl_prop_unregister(ds, "primarycache", | |
428870ff | 550 | primary_cache_changed_cb, os)); |
b128c09f | 551 | VERIFY(0 == dsl_prop_unregister(ds, "secondarycache", |
428870ff | 552 | secondary_cache_changed_cb, os)); |
34dc7c2f BB |
553 | } |
554 | ||
428870ff BB |
555 | if (os->os_sa) |
556 | sa_tear_down(os); | |
557 | ||
34dc7c2f BB |
558 | /* |
559 | * We should need only a single pass over the dnode list, since | |
560 | * nothing can be added to the list at this point. | |
561 | */ | |
428870ff | 562 | (void) dmu_objset_evict_dbufs(os); |
34dc7c2f | 563 | |
572e2857 BB |
564 | dnode_special_close(&os->os_meta_dnode); |
565 | if (DMU_USERUSED_DNODE(os)) { | |
566 | dnode_special_close(&os->os_userused_dnode); | |
567 | dnode_special_close(&os->os_groupused_dnode); | |
9babb374 | 568 | } |
428870ff BB |
569 | zil_free(os->os_zil); |
570 | ||
571 | ASSERT3P(list_head(&os->os_dnodes), ==, NULL); | |
34dc7c2f | 572 | |
428870ff | 573 | VERIFY(arc_buf_remove_ref(os->os_phys_buf, &os->os_phys_buf) == 1); |
572e2857 BB |
574 | |
575 | /* | |
576 | * This is a barrier to prevent the objset from going away in | |
577 | * dnode_move() until we can safely ensure that the objset is still in | |
578 | * use. We consider the objset valid before the barrier and invalid | |
579 | * after the barrier. | |
580 | */ | |
581 | rw_enter(&os_lock, RW_READER); | |
582 | rw_exit(&os_lock); | |
583 | ||
428870ff BB |
584 | mutex_destroy(&os->os_lock); |
585 | mutex_destroy(&os->os_obj_lock); | |
586 | mutex_destroy(&os->os_user_ptr_lock); | |
587 | kmem_free(os, sizeof (objset_t)); | |
588 | } | |
9babb374 | 589 | |
428870ff BB |
590 | timestruc_t |
591 | dmu_objset_snap_cmtime(objset_t *os) | |
592 | { | |
593 | return (dsl_dir_snap_cmtime(os->os_dsl_dataset->ds_dir)); | |
34dc7c2f BB |
594 | } |
595 | ||
596 | /* called from dsl for meta-objset */ | |
428870ff | 597 | objset_t * |
34dc7c2f BB |
598 | dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, |
599 | dmu_objset_type_t type, dmu_tx_t *tx) | |
600 | { | |
428870ff | 601 | objset_t *os; |
34dc7c2f BB |
602 | dnode_t *mdn; |
603 | ||
604 | ASSERT(dmu_tx_is_syncing(tx)); | |
572e2857 BB |
605 | if (ds != NULL) |
606 | VERIFY(0 == dmu_objset_from_ds(ds, &os)); | |
607 | else | |
608 | VERIFY(0 == dmu_objset_open_impl(spa, NULL, bp, &os)); | |
609 | ||
610 | mdn = DMU_META_DNODE(os); | |
34dc7c2f BB |
611 | |
612 | dnode_allocate(mdn, DMU_OT_DNODE, 1 << DNODE_BLOCK_SHIFT, | |
613 | DN_MAX_INDBLKSHIFT, DMU_OT_NONE, 0, tx); | |
614 | ||
615 | /* | |
616 | * We don't want to have to increase the meta-dnode's nlevels | |
617 | * later, because then we could do it in quescing context while | |
618 | * we are also accessing it in open context. | |
619 | * | |
620 | * This precaution is not necessary for the MOS (ds == NULL), | |
621 | * because the MOS is only updated in syncing context. | |
622 | * This is most fortunate: the MOS is the only objset that | |
623 | * needs to be synced multiple times as spa_sync() iterates | |
624 | * to convergence, so minimizing its dn_nlevels matters. | |
625 | */ | |
626 | if (ds != NULL) { | |
627 | int levels = 1; | |
628 | ||
629 | /* | |
630 | * Determine the number of levels necessary for the meta-dnode | |
631 | * to contain DN_MAX_OBJECT dnodes. | |
632 | */ | |
633 | while ((uint64_t)mdn->dn_nblkptr << (mdn->dn_datablkshift + | |
634 | (levels - 1) * (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) < | |
635 | DN_MAX_OBJECT * sizeof (dnode_phys_t)) | |
636 | levels++; | |
637 | ||
638 | mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] = | |
639 | mdn->dn_nlevels = levels; | |
640 | } | |
641 | ||
642 | ASSERT(type != DMU_OST_NONE); | |
643 | ASSERT(type != DMU_OST_ANY); | |
644 | ASSERT(type < DMU_OST_NUMTYPES); | |
428870ff BB |
645 | os->os_phys->os_type = type; |
646 | if (dmu_objset_userused_enabled(os)) { | |
647 | os->os_phys->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; | |
648 | os->os_flags = os->os_phys->os_flags; | |
9babb374 | 649 | } |
34dc7c2f BB |
650 | |
651 | dsl_dataset_dirty(ds, tx); | |
652 | ||
428870ff | 653 | return (os); |
34dc7c2f BB |
654 | } |
655 | ||
656 | struct oscarg { | |
657 | void (*userfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); | |
658 | void *userarg; | |
428870ff | 659 | dsl_dataset_t *clone_origin; |
34dc7c2f BB |
660 | const char *lastname; |
661 | dmu_objset_type_t type; | |
662 | uint64_t flags; | |
428870ff | 663 | cred_t *cr; |
34dc7c2f BB |
664 | }; |
665 | ||
666 | /*ARGSUSED*/ | |
667 | static int | |
668 | dmu_objset_create_check(void *arg1, void *arg2, dmu_tx_t *tx) | |
669 | { | |
670 | dsl_dir_t *dd = arg1; | |
671 | struct oscarg *oa = arg2; | |
672 | objset_t *mos = dd->dd_pool->dp_meta_objset; | |
673 | int err; | |
674 | uint64_t ddobj; | |
675 | ||
676 | err = zap_lookup(mos, dd->dd_phys->dd_child_dir_zapobj, | |
677 | oa->lastname, sizeof (uint64_t), 1, &ddobj); | |
678 | if (err != ENOENT) | |
679 | return (err ? err : EEXIST); | |
680 | ||
428870ff BB |
681 | if (oa->clone_origin != NULL) { |
682 | /* You can't clone across pools. */ | |
683 | if (oa->clone_origin->ds_dir->dd_pool != dd->dd_pool) | |
34dc7c2f BB |
684 | return (EXDEV); |
685 | ||
428870ff BB |
686 | /* You can only clone snapshots, not the head datasets. */ |
687 | if (!dsl_dataset_is_snapshot(oa->clone_origin)) | |
34dc7c2f BB |
688 | return (EINVAL); |
689 | } | |
690 | ||
691 | return (0); | |
692 | } | |
693 | ||
694 | static void | |
428870ff | 695 | dmu_objset_create_sync(void *arg1, void *arg2, dmu_tx_t *tx) |
34dc7c2f BB |
696 | { |
697 | dsl_dir_t *dd = arg1; | |
572e2857 | 698 | spa_t *spa = dd->dd_pool->dp_spa; |
34dc7c2f | 699 | struct oscarg *oa = arg2; |
572e2857 | 700 | uint64_t obj; |
34dc7c2f BB |
701 | |
702 | ASSERT(dmu_tx_is_syncing(tx)); | |
703 | ||
572e2857 | 704 | obj = dsl_dataset_create_sync(dd, oa->lastname, |
428870ff | 705 | oa->clone_origin, oa->flags, oa->cr, tx); |
34dc7c2f | 706 | |
428870ff | 707 | if (oa->clone_origin == NULL) { |
572e2857 | 708 | dsl_pool_t *dp = dd->dd_pool; |
428870ff BB |
709 | dsl_dataset_t *ds; |
710 | blkptr_t *bp; | |
711 | objset_t *os; | |
34dc7c2f | 712 | |
572e2857 | 713 | VERIFY3U(0, ==, dsl_dataset_hold_obj(dp, obj, FTAG, &ds)); |
428870ff BB |
714 | bp = dsl_dataset_get_blkptr(ds); |
715 | ASSERT(BP_IS_HOLE(bp)); | |
716 | ||
572e2857 | 717 | os = dmu_objset_create_impl(spa, ds, bp, oa->type, tx); |
34dc7c2f BB |
718 | |
719 | if (oa->userfunc) | |
428870ff BB |
720 | oa->userfunc(os, oa->userarg, oa->cr, tx); |
721 | dsl_dataset_rele(ds, FTAG); | |
34dc7c2f BB |
722 | } |
723 | ||
572e2857 | 724 | spa_history_log_internal(LOG_DS_CREATE, spa, tx, "dataset = %llu", obj); |
34dc7c2f BB |
725 | } |
726 | ||
727 | int | |
428870ff | 728 | dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, |
34dc7c2f BB |
729 | void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg) |
730 | { | |
731 | dsl_dir_t *pdd; | |
732 | const char *tail; | |
733 | int err = 0; | |
734 | struct oscarg oa = { 0 }; | |
735 | ||
736 | ASSERT(strchr(name, '@') == NULL); | |
737 | err = dsl_dir_open(name, FTAG, &pdd, &tail); | |
738 | if (err) | |
739 | return (err); | |
740 | if (tail == NULL) { | |
741 | dsl_dir_close(pdd, FTAG); | |
742 | return (EEXIST); | |
743 | } | |
744 | ||
34dc7c2f BB |
745 | oa.userfunc = func; |
746 | oa.userarg = arg; | |
747 | oa.lastname = tail; | |
748 | oa.type = type; | |
749 | oa.flags = flags; | |
428870ff | 750 | oa.cr = CRED(); |
34dc7c2f | 751 | |
34dc7c2f BB |
752 | err = dsl_sync_task_do(pdd->dd_pool, dmu_objset_create_check, |
753 | dmu_objset_create_sync, pdd, &oa, 5); | |
754 | dsl_dir_close(pdd, FTAG); | |
755 | return (err); | |
756 | } | |
757 | ||
758 | int | |
428870ff | 759 | dmu_objset_clone(const char *name, dsl_dataset_t *clone_origin, uint64_t flags) |
34dc7c2f | 760 | { |
428870ff BB |
761 | dsl_dir_t *pdd; |
762 | const char *tail; | |
763 | int err = 0; | |
764 | struct oscarg oa = { 0 }; | |
34dc7c2f | 765 | |
428870ff BB |
766 | ASSERT(strchr(name, '@') == NULL); |
767 | err = dsl_dir_open(name, FTAG, &pdd, &tail); | |
768 | if (err) | |
769 | return (err); | |
770 | if (tail == NULL) { | |
771 | dsl_dir_close(pdd, FTAG); | |
772 | return (EEXIST); | |
34dc7c2f BB |
773 | } |
774 | ||
428870ff BB |
775 | oa.lastname = tail; |
776 | oa.clone_origin = clone_origin; | |
777 | oa.flags = flags; | |
778 | oa.cr = CRED(); | |
779 | ||
780 | err = dsl_sync_task_do(pdd->dd_pool, dmu_objset_create_check, | |
781 | dmu_objset_create_sync, pdd, &oa, 5); | |
782 | dsl_dir_close(pdd, FTAG); | |
783 | return (err); | |
34dc7c2f BB |
784 | } |
785 | ||
34dc7c2f | 786 | int |
428870ff | 787 | dmu_objset_destroy(const char *name, boolean_t defer) |
34dc7c2f | 788 | { |
34dc7c2f | 789 | dsl_dataset_t *ds; |
428870ff | 790 | int error; |
34dc7c2f | 791 | |
428870ff BB |
792 | error = dsl_dataset_own(name, B_TRUE, FTAG, &ds); |
793 | if (error == 0) { | |
428870ff BB |
794 | error = dsl_dataset_destroy(ds, FTAG, defer); |
795 | /* dsl_dataset_destroy() closes the ds. */ | |
796 | } | |
797 | ||
798 | return (error); | |
34dc7c2f BB |
799 | } |
800 | ||
801 | struct snaparg { | |
802 | dsl_sync_task_group_t *dstg; | |
803 | char *snapname; | |
572e2857 | 804 | char *htag; |
34dc7c2f | 805 | char failed[MAXPATHLEN]; |
428870ff | 806 | boolean_t recursive; |
572e2857 BB |
807 | boolean_t needsuspend; |
808 | boolean_t temporary; | |
9babb374 | 809 | nvlist_t *props; |
572e2857 BB |
810 | struct dsl_ds_holdarg *ha; /* only needed in the temporary case */ |
811 | dsl_dataset_t *newds; | |
34dc7c2f BB |
812 | }; |
813 | ||
9babb374 BB |
814 | static int |
815 | snapshot_check(void *arg1, void *arg2, dmu_tx_t *tx) | |
816 | { | |
817 | objset_t *os = arg1; | |
818 | struct snaparg *sn = arg2; | |
572e2857 | 819 | int error; |
9babb374 BB |
820 | |
821 | /* The props have already been checked by zfs_check_userprops(). */ | |
822 | ||
572e2857 BB |
823 | error = dsl_dataset_snapshot_check(os->os_dsl_dataset, |
824 | sn->snapname, tx); | |
825 | if (error) | |
826 | return (error); | |
827 | ||
828 | if (sn->temporary) { | |
829 | /* | |
830 | * Ideally we would just call | |
831 | * dsl_dataset_user_hold_check() and | |
832 | * dsl_dataset_destroy_check() here. However the | |
833 | * dataset we want to hold and destroy is the snapshot | |
834 | * that we just confirmed we can create, but it won't | |
835 | * exist until after these checks are run. Do any | |
836 | * checks we can here and if more checks are added to | |
837 | * those routines in the future, similar checks may be | |
838 | * necessary here. | |
839 | */ | |
840 | if (spa_version(os->os_spa) < SPA_VERSION_USERREFS) | |
841 | return (ENOTSUP); | |
842 | /* | |
843 | * Not checking number of tags because the tag will be | |
844 | * unique, as it will be the only tag. | |
845 | */ | |
846 | if (strlen(sn->htag) + MAX_TAG_PREFIX_LEN >= MAXNAMELEN) | |
847 | return (E2BIG); | |
848 | ||
beb99944 | 849 | sn->ha = kmem_alloc(sizeof(struct dsl_ds_holdarg), KM_PUSHPAGE); |
572e2857 BB |
850 | sn->ha->temphold = B_TRUE; |
851 | sn->ha->htag = sn->htag; | |
852 | } | |
853 | return (error); | |
9babb374 BB |
854 | } |
855 | ||
856 | static void | |
428870ff | 857 | snapshot_sync(void *arg1, void *arg2, dmu_tx_t *tx) |
9babb374 BB |
858 | { |
859 | objset_t *os = arg1; | |
428870ff | 860 | dsl_dataset_t *ds = os->os_dsl_dataset; |
9babb374 BB |
861 | struct snaparg *sn = arg2; |
862 | ||
428870ff | 863 | dsl_dataset_snapshot_sync(ds, sn->snapname, tx); |
9babb374 | 864 | |
428870ff BB |
865 | if (sn->props) { |
866 | dsl_props_arg_t pa; | |
867 | pa.pa_props = sn->props; | |
868 | pa.pa_source = ZPROP_SRC_LOCAL; | |
869 | dsl_props_set_sync(ds->ds_prev, &pa, tx); | |
870 | } | |
572e2857 BB |
871 | |
872 | if (sn->temporary) { | |
873 | struct dsl_ds_destroyarg da; | |
874 | ||
875 | dsl_dataset_user_hold_sync(ds->ds_prev, sn->ha, tx); | |
876 | kmem_free(sn->ha, sizeof (struct dsl_ds_holdarg)); | |
877 | sn->ha = NULL; | |
878 | sn->newds = ds->ds_prev; | |
879 | ||
880 | da.ds = ds->ds_prev; | |
881 | da.defer = B_TRUE; | |
882 | dsl_dataset_destroy_sync(&da, FTAG, tx); | |
883 | } | |
9babb374 | 884 | } |
34dc7c2f BB |
885 | |
886 | static int | |
428870ff | 887 | dmu_objset_snapshot_one(const char *name, void *arg) |
34dc7c2f BB |
888 | { |
889 | struct snaparg *sn = arg; | |
890 | objset_t *os; | |
34dc7c2f | 891 | int err; |
428870ff BB |
892 | char *cp; |
893 | ||
894 | /* | |
895 | * If the objset starts with a '%', then ignore it unless it was | |
896 | * explicitly named (ie, not recursive). These hidden datasets | |
897 | * are always inconsistent, and by not opening them here, we can | |
898 | * avoid a race with dsl_dir_destroy_check(). | |
899 | */ | |
900 | cp = strrchr(name, '/'); | |
901 | if (cp && cp[1] == '%' && sn->recursive) | |
902 | return (0); | |
34dc7c2f BB |
903 | |
904 | (void) strcpy(sn->failed, name); | |
905 | ||
906 | /* | |
428870ff BB |
907 | * Check permissions if we are doing a recursive snapshot. The |
908 | * permission checks for the starting dataset have already been | |
909 | * performed in zfs_secpolicy_snapshot() | |
34dc7c2f | 910 | */ |
428870ff | 911 | if (sn->recursive && (err = zfs_secpolicy_snapshot_perms(name, CRED()))) |
34dc7c2f BB |
912 | return (err); |
913 | ||
428870ff | 914 | err = dmu_objset_hold(name, sn, &os); |
34dc7c2f BB |
915 | if (err != 0) |
916 | return (err); | |
917 | ||
428870ff BB |
918 | /* |
919 | * If the objset is in an inconsistent state (eg, in the process | |
920 | * of being destroyed), don't snapshot it. As with %hidden | |
921 | * datasets, we return EBUSY if this name was explicitly | |
922 | * requested (ie, not recursive), and otherwise ignore it. | |
923 | */ | |
924 | if (os->os_dsl_dataset->ds_phys->ds_flags & DS_FLAG_INCONSISTENT) { | |
925 | dmu_objset_rele(os, sn); | |
926 | return (sn->recursive ? 0 : EBUSY); | |
34dc7c2f BB |
927 | } |
928 | ||
572e2857 BB |
929 | if (sn->needsuspend) { |
930 | err = zil_suspend(dmu_objset_zil(os)); | |
931 | if (err) { | |
932 | dmu_objset_rele(os, sn); | |
933 | return (err); | |
934 | } | |
34dc7c2f | 935 | } |
572e2857 BB |
936 | dsl_sync_task_create(sn->dstg, snapshot_check, snapshot_sync, |
937 | os, sn, 3); | |
34dc7c2f | 938 | |
572e2857 | 939 | return (0); |
34dc7c2f BB |
940 | } |
941 | ||
942 | int | |
572e2857 BB |
943 | dmu_objset_snapshot(char *fsname, char *snapname, char *tag, |
944 | nvlist_t *props, boolean_t recursive, boolean_t temporary, int cleanup_fd) | |
34dc7c2f BB |
945 | { |
946 | dsl_sync_task_t *dst; | |
a8ac8e71 | 947 | struct snaparg *sn; |
34dc7c2f | 948 | spa_t *spa; |
572e2857 | 949 | minor_t minor; |
34dc7c2f BB |
950 | int err; |
951 | ||
a8ac8e71 BB |
952 | sn = kmem_alloc(sizeof (struct snaparg), KM_SLEEP); |
953 | (void) strcpy(sn->failed, fsname); | |
34dc7c2f BB |
954 | |
955 | err = spa_open(fsname, &spa, FTAG); | |
a8ac8e71 BB |
956 | if (err) { |
957 | kmem_free(sn, sizeof (struct snaparg)); | |
34dc7c2f | 958 | return (err); |
a8ac8e71 | 959 | } |
34dc7c2f | 960 | |
572e2857 BB |
961 | if (temporary) { |
962 | if (cleanup_fd < 0) { | |
963 | spa_close(spa, FTAG); | |
964 | return (EINVAL); | |
965 | } | |
966 | if ((err = zfs_onexit_fd_hold(cleanup_fd, &minor)) != 0) { | |
967 | spa_close(spa, FTAG); | |
968 | return (err); | |
969 | } | |
970 | } | |
971 | ||
a8ac8e71 BB |
972 | sn->dstg = dsl_sync_task_group_create(spa_get_dsl(spa)); |
973 | sn->snapname = snapname; | |
974 | sn->htag = tag; | |
975 | sn->props = props; | |
976 | sn->recursive = recursive; | |
977 | sn->needsuspend = (spa_version(spa) < SPA_VERSION_FAST_SNAP); | |
978 | sn->temporary = temporary; | |
979 | sn->ha = NULL; | |
980 | sn->newds = NULL; | |
34dc7c2f BB |
981 | |
982 | if (recursive) { | |
34dc7c2f | 983 | err = dmu_objset_find(fsname, |
a8ac8e71 | 984 | dmu_objset_snapshot_one, sn, DS_FIND_CHILDREN); |
34dc7c2f | 985 | } else { |
a8ac8e71 | 986 | err = dmu_objset_snapshot_one(fsname, sn); |
34dc7c2f BB |
987 | } |
988 | ||
9babb374 | 989 | if (err == 0) |
a8ac8e71 | 990 | err = dsl_sync_task_group_wait(sn->dstg); |
34dc7c2f | 991 | |
a8ac8e71 BB |
992 | for (dst = list_head(&sn->dstg->dstg_tasks); dst; |
993 | dst = list_next(&sn->dstg->dstg_tasks, dst)) { | |
9babb374 | 994 | objset_t *os = dst->dst_arg1; |
428870ff | 995 | dsl_dataset_t *ds = os->os_dsl_dataset; |
572e2857 | 996 | if (dst->dst_err) { |
a8ac8e71 | 997 | dsl_dataset_name(ds, sn->failed); |
572e2857 | 998 | } else if (temporary) { |
a8ac8e71 | 999 | dsl_register_onexit_hold_cleanup(sn->newds, tag, minor); |
572e2857 | 1000 | } |
a8ac8e71 | 1001 | if (sn->needsuspend) |
572e2857 | 1002 | zil_resume(dmu_objset_zil(os)); |
a8ac8e71 | 1003 | dmu_objset_rele(os, sn); |
34dc7c2f BB |
1004 | } |
1005 | ||
34dc7c2f | 1006 | if (err) |
a8ac8e71 | 1007 | (void) strcpy(fsname, sn->failed); |
572e2857 BB |
1008 | if (temporary) |
1009 | zfs_onexit_fd_rele(cleanup_fd); | |
a8ac8e71 | 1010 | dsl_sync_task_group_destroy(sn->dstg); |
34dc7c2f | 1011 | spa_close(spa, FTAG); |
a8ac8e71 | 1012 | kmem_free(sn, sizeof (struct snaparg)); |
34dc7c2f BB |
1013 | return (err); |
1014 | } | |
1015 | ||
1016 | static void | |
9babb374 | 1017 | dmu_objset_sync_dnodes(list_t *list, list_t *newlist, dmu_tx_t *tx) |
34dc7c2f BB |
1018 | { |
1019 | dnode_t *dn; | |
1020 | ||
c65aa5b2 | 1021 | while ((dn = list_head(list))) { |
34dc7c2f BB |
1022 | ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT); |
1023 | ASSERT(dn->dn_dbuf->db_data_pending); | |
1024 | /* | |
9babb374 BB |
1025 | * Initialize dn_zio outside dnode_sync() because the |
1026 | * meta-dnode needs to set it ouside dnode_sync(). | |
34dc7c2f BB |
1027 | */ |
1028 | dn->dn_zio = dn->dn_dbuf->db_data_pending->dr_zio; | |
1029 | ASSERT(dn->dn_zio); | |
1030 | ||
1031 | ASSERT3U(dn->dn_nlevels, <=, DN_MAX_LEVELS); | |
1032 | list_remove(list, dn); | |
9babb374 BB |
1033 | |
1034 | if (newlist) { | |
1035 | (void) dnode_add_ref(dn, newlist); | |
1036 | list_insert_tail(newlist, dn); | |
1037 | } | |
1038 | ||
34dc7c2f BB |
1039 | dnode_sync(dn, tx); |
1040 | } | |
1041 | } | |
1042 | ||
1043 | /* ARGSUSED */ | |
1044 | static void | |
428870ff | 1045 | dmu_objset_write_ready(zio_t *zio, arc_buf_t *abuf, void *arg) |
34dc7c2f | 1046 | { |
d6320ddb BB |
1047 | int i; |
1048 | ||
b128c09f | 1049 | blkptr_t *bp = zio->io_bp; |
428870ff | 1050 | objset_t *os = arg; |
34dc7c2f | 1051 | dnode_phys_t *dnp = &os->os_phys->os_meta_dnode; |
34dc7c2f | 1052 | |
b128c09f BB |
1053 | ASSERT(bp == os->os_rootbp); |
1054 | ASSERT(BP_GET_TYPE(bp) == DMU_OT_OBJSET); | |
1055 | ASSERT(BP_GET_LEVEL(bp) == 0); | |
34dc7c2f BB |
1056 | |
1057 | /* | |
9babb374 BB |
1058 | * Update rootbp fill count: it should be the number of objects |
1059 | * allocated in the object set (not counting the "special" | |
1060 | * objects that are stored in the objset_phys_t -- the meta | |
1061 | * dnode and user/group accounting objects). | |
34dc7c2f | 1062 | */ |
9babb374 | 1063 | bp->blk_fill = 0; |
d6320ddb | 1064 | for (i = 0; i < dnp->dn_nblkptr; i++) |
34dc7c2f | 1065 | bp->blk_fill += dnp->dn_blkptr[i].blk_fill; |
428870ff BB |
1066 | } |
1067 | ||
1068 | /* ARGSUSED */ | |
1069 | static void | |
1070 | dmu_objset_write_done(zio_t *zio, arc_buf_t *abuf, void *arg) | |
1071 | { | |
1072 | blkptr_t *bp = zio->io_bp; | |
1073 | blkptr_t *bp_orig = &zio->io_bp_orig; | |
1074 | objset_t *os = arg; | |
34dc7c2f | 1075 | |
b128c09f | 1076 | if (zio->io_flags & ZIO_FLAG_IO_REWRITE) { |
428870ff | 1077 | ASSERT(BP_EQUAL(bp, bp_orig)); |
b128c09f | 1078 | } else { |
428870ff BB |
1079 | dsl_dataset_t *ds = os->os_dsl_dataset; |
1080 | dmu_tx_t *tx = os->os_synctx; | |
1081 | ||
1082 | (void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE); | |
1083 | dsl_dataset_block_born(ds, bp, tx); | |
34dc7c2f BB |
1084 | } |
1085 | } | |
1086 | ||
34dc7c2f BB |
1087 | /* called from dsl */ |
1088 | void | |
428870ff | 1089 | dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx) |
34dc7c2f BB |
1090 | { |
1091 | int txgoff; | |
1092 | zbookmark_t zb; | |
428870ff | 1093 | zio_prop_t zp; |
34dc7c2f BB |
1094 | zio_t *zio; |
1095 | list_t *list; | |
9babb374 | 1096 | list_t *newlist = NULL; |
34dc7c2f BB |
1097 | dbuf_dirty_record_t *dr; |
1098 | ||
1099 | dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg); | |
1100 | ||
1101 | ASSERT(dmu_tx_is_syncing(tx)); | |
1102 | /* XXX the write_done callback should really give us the tx... */ | |
1103 | os->os_synctx = tx; | |
1104 | ||
1105 | if (os->os_dsl_dataset == NULL) { | |
1106 | /* | |
1107 | * This is the MOS. If we have upgraded, | |
1108 | * spa_max_replication() could change, so reset | |
1109 | * os_copies here. | |
1110 | */ | |
1111 | os->os_copies = spa_max_replication(os->os_spa); | |
1112 | } | |
1113 | ||
1114 | /* | |
1115 | * Create the root block IO | |
1116 | */ | |
428870ff BB |
1117 | SET_BOOKMARK(&zb, os->os_dsl_dataset ? |
1118 | os->os_dsl_dataset->ds_object : DMU_META_OBJSET, | |
1119 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
294f6806 | 1120 | arc_release(os->os_phys_buf, &os->os_phys_buf); |
b128c09f | 1121 | |
428870ff | 1122 | dmu_write_policy(os, NULL, 0, 0, &zp); |
9babb374 | 1123 | |
428870ff | 1124 | zio = arc_write(pio, os->os_spa, tx->tx_txg, |
3a17a7a9 SK |
1125 | os->os_rootbp, os->os_phys_buf, DMU_OS_IS_L2CACHEABLE(os), |
1126 | DMU_OS_IS_L2COMPRESSIBLE(os), &zp, dmu_objset_write_ready, | |
1127 | dmu_objset_write_done, os, ZIO_PRIORITY_ASYNC_WRITE, | |
1128 | ZIO_FLAG_MUSTSUCCEED, &zb); | |
34dc7c2f BB |
1129 | |
1130 | /* | |
9babb374 | 1131 | * Sync special dnodes - the parent IO for the sync is the root block |
34dc7c2f | 1132 | */ |
572e2857 BB |
1133 | DMU_META_DNODE(os)->dn_zio = zio; |
1134 | dnode_sync(DMU_META_DNODE(os), tx); | |
34dc7c2f | 1135 | |
9babb374 BB |
1136 | os->os_phys->os_flags = os->os_flags; |
1137 | ||
572e2857 BB |
1138 | if (DMU_USERUSED_DNODE(os) && |
1139 | DMU_USERUSED_DNODE(os)->dn_type != DMU_OT_NONE) { | |
1140 | DMU_USERUSED_DNODE(os)->dn_zio = zio; | |
1141 | dnode_sync(DMU_USERUSED_DNODE(os), tx); | |
1142 | DMU_GROUPUSED_DNODE(os)->dn_zio = zio; | |
1143 | dnode_sync(DMU_GROUPUSED_DNODE(os), tx); | |
9babb374 BB |
1144 | } |
1145 | ||
34dc7c2f BB |
1146 | txgoff = tx->tx_txg & TXG_MASK; |
1147 | ||
9babb374 BB |
1148 | if (dmu_objset_userused_enabled(os)) { |
1149 | newlist = &os->os_synced_dnodes; | |
1150 | /* | |
1151 | * We must create the list here because it uses the | |
1152 | * dn_dirty_link[] of this txg. | |
1153 | */ | |
1154 | list_create(newlist, sizeof (dnode_t), | |
1155 | offsetof(dnode_t, dn_dirty_link[txgoff])); | |
1156 | } | |
1157 | ||
1158 | dmu_objset_sync_dnodes(&os->os_free_dnodes[txgoff], newlist, tx); | |
1159 | dmu_objset_sync_dnodes(&os->os_dirty_dnodes[txgoff], newlist, tx); | |
34dc7c2f | 1160 | |
572e2857 | 1161 | list = &DMU_META_DNODE(os)->dn_dirty_records[txgoff]; |
c65aa5b2 | 1162 | while ((dr = list_head(list)) != NULL) { |
34dc7c2f BB |
1163 | ASSERT(dr->dr_dbuf->db_level == 0); |
1164 | list_remove(list, dr); | |
1165 | if (dr->dr_zio) | |
1166 | zio_nowait(dr->dr_zio); | |
1167 | } | |
1168 | /* | |
1169 | * Free intent log blocks up to this tx. | |
1170 | */ | |
1171 | zil_sync(os->os_zil, tx); | |
b128c09f | 1172 | os->os_phys->os_zil_header = os->os_zil_header; |
34dc7c2f BB |
1173 | zio_nowait(zio); |
1174 | } | |
1175 | ||
428870ff BB |
1176 | boolean_t |
1177 | dmu_objset_is_dirty(objset_t *os, uint64_t txg) | |
1178 | { | |
1179 | return (!list_is_empty(&os->os_dirty_dnodes[txg & TXG_MASK]) || | |
1180 | !list_is_empty(&os->os_free_dnodes[txg & TXG_MASK])); | |
1181 | } | |
1182 | ||
572e2857 | 1183 | static objset_used_cb_t *used_cbs[DMU_OST_NUMTYPES]; |
9babb374 BB |
1184 | |
1185 | void | |
1186 | dmu_objset_register_type(dmu_objset_type_t ost, objset_used_cb_t *cb) | |
1187 | { | |
1188 | used_cbs[ost] = cb; | |
1189 | } | |
1190 | ||
1191 | boolean_t | |
428870ff | 1192 | dmu_objset_userused_enabled(objset_t *os) |
9babb374 BB |
1193 | { |
1194 | return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE && | |
572e2857 BB |
1195 | used_cbs[os->os_phys->os_type] != NULL && |
1196 | DMU_USERUSED_DNODE(os) != NULL); | |
9babb374 BB |
1197 | } |
1198 | ||
428870ff BB |
1199 | static void |
1200 | do_userquota_update(objset_t *os, uint64_t used, uint64_t flags, | |
1201 | uint64_t user, uint64_t group, boolean_t subtract, dmu_tx_t *tx) | |
1202 | { | |
1203 | if ((flags & DNODE_FLAG_USERUSED_ACCOUNTED)) { | |
1204 | int64_t delta = DNODE_SIZE + used; | |
1205 | if (subtract) | |
1206 | delta = -delta; | |
1207 | VERIFY3U(0, ==, zap_increment_int(os, DMU_USERUSED_OBJECT, | |
1208 | user, delta, tx)); | |
1209 | VERIFY3U(0, ==, zap_increment_int(os, DMU_GROUPUSED_OBJECT, | |
1210 | group, delta, tx)); | |
1211 | } | |
1212 | } | |
1213 | ||
9babb374 | 1214 | void |
428870ff | 1215 | dmu_objset_do_userquota_updates(objset_t *os, dmu_tx_t *tx) |
9babb374 BB |
1216 | { |
1217 | dnode_t *dn; | |
1218 | list_t *list = &os->os_synced_dnodes; | |
9babb374 BB |
1219 | |
1220 | ASSERT(list_head(list) == NULL || dmu_objset_userused_enabled(os)); | |
1221 | ||
c65aa5b2 | 1222 | while ((dn = list_head(list)) != NULL) { |
572e2857 | 1223 | int flags; |
9babb374 | 1224 | ASSERT(!DMU_OBJECT_IS_SPECIAL(dn->dn_object)); |
9babb374 BB |
1225 | ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE || |
1226 | dn->dn_phys->dn_flags & | |
1227 | DNODE_FLAG_USERUSED_ACCOUNTED); | |
1228 | ||
1229 | /* Allocate the user/groupused objects if necessary. */ | |
572e2857 | 1230 | if (DMU_USERUSED_DNODE(os)->dn_type == DMU_OT_NONE) { |
428870ff | 1231 | VERIFY(0 == zap_create_claim(os, |
9babb374 BB |
1232 | DMU_USERUSED_OBJECT, |
1233 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
428870ff | 1234 | VERIFY(0 == zap_create_claim(os, |
9babb374 BB |
1235 | DMU_GROUPUSED_OBJECT, |
1236 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
1237 | } | |
1238 | ||
1239 | /* | |
428870ff BB |
1240 | * We intentionally modify the zap object even if the |
1241 | * net delta is zero. Otherwise | |
1242 | * the block of the zap obj could be shared between | |
1243 | * datasets but need to be different between them after | |
1244 | * a bprewrite. | |
9babb374 | 1245 | */ |
9babb374 | 1246 | |
572e2857 BB |
1247 | flags = dn->dn_id_flags; |
1248 | ASSERT(flags); | |
1249 | if (flags & DN_ID_OLD_EXIST) { | |
428870ff BB |
1250 | do_userquota_update(os, dn->dn_oldused, dn->dn_oldflags, |
1251 | dn->dn_olduid, dn->dn_oldgid, B_TRUE, tx); | |
1252 | } | |
572e2857 | 1253 | if (flags & DN_ID_NEW_EXIST) { |
428870ff BB |
1254 | do_userquota_update(os, DN_USED_BYTES(dn->dn_phys), |
1255 | dn->dn_phys->dn_flags, dn->dn_newuid, | |
1256 | dn->dn_newgid, B_FALSE, tx); | |
1257 | } | |
1258 | ||
572e2857 | 1259 | mutex_enter(&dn->dn_mtx); |
428870ff BB |
1260 | dn->dn_oldused = 0; |
1261 | dn->dn_oldflags = 0; | |
1262 | if (dn->dn_id_flags & DN_ID_NEW_EXIST) { | |
1263 | dn->dn_olduid = dn->dn_newuid; | |
1264 | dn->dn_oldgid = dn->dn_newgid; | |
1265 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
1266 | if (dn->dn_bonuslen == 0) | |
1267 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1268 | else | |
1269 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1270 | } | |
1271 | dn->dn_id_flags &= ~(DN_ID_NEW_EXIST); | |
9babb374 BB |
1272 | mutex_exit(&dn->dn_mtx); |
1273 | ||
1274 | list_remove(list, dn); | |
1275 | dnode_rele(dn, list); | |
1276 | } | |
1277 | } | |
1278 | ||
428870ff BB |
1279 | /* |
1280 | * Returns a pointer to data to find uid/gid from | |
1281 | * | |
1282 | * If a dirty record for transaction group that is syncing can't | |
1283 | * be found then NULL is returned. In the NULL case it is assumed | |
1284 | * the uid/gid aren't changing. | |
1285 | */ | |
1286 | static void * | |
1287 | dmu_objset_userquota_find_data(dmu_buf_impl_t *db, dmu_tx_t *tx) | |
1288 | { | |
1289 | dbuf_dirty_record_t *dr, **drp; | |
1290 | void *data; | |
1291 | ||
1292 | if (db->db_dirtycnt == 0) | |
1293 | return (db->db.db_data); /* Nothing is changing */ | |
1294 | ||
1295 | for (drp = &db->db_last_dirty; (dr = *drp) != NULL; drp = &dr->dr_next) | |
1296 | if (dr->dr_txg == tx->tx_txg) | |
1297 | break; | |
1298 | ||
572e2857 | 1299 | if (dr == NULL) { |
428870ff | 1300 | data = NULL; |
572e2857 BB |
1301 | } else { |
1302 | dnode_t *dn; | |
1303 | ||
1304 | DB_DNODE_ENTER(dr->dr_dbuf); | |
1305 | dn = DB_DNODE(dr->dr_dbuf); | |
1306 | ||
1307 | if (dn->dn_bonuslen == 0 && | |
1308 | dr->dr_dbuf->db_blkid == DMU_SPILL_BLKID) | |
1309 | data = dr->dt.dl.dr_data->b_data; | |
1310 | else | |
1311 | data = dr->dt.dl.dr_data; | |
1312 | ||
1313 | DB_DNODE_EXIT(dr->dr_dbuf); | |
1314 | } | |
1315 | ||
428870ff BB |
1316 | return (data); |
1317 | } | |
1318 | ||
1319 | void | |
1320 | dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) | |
1321 | { | |
1322 | objset_t *os = dn->dn_objset; | |
1323 | void *data = NULL; | |
1324 | dmu_buf_impl_t *db = NULL; | |
d4ed6673 | 1325 | uint64_t *user = NULL, *group = NULL; |
428870ff BB |
1326 | int flags = dn->dn_id_flags; |
1327 | int error; | |
1328 | boolean_t have_spill = B_FALSE; | |
1329 | ||
1330 | if (!dmu_objset_userused_enabled(dn->dn_objset)) | |
1331 | return; | |
1332 | ||
1333 | if (before && (flags & (DN_ID_CHKED_BONUS|DN_ID_OLD_EXIST| | |
1334 | DN_ID_CHKED_SPILL))) | |
1335 | return; | |
1336 | ||
1337 | if (before && dn->dn_bonuslen != 0) | |
1338 | data = DN_BONUS(dn->dn_phys); | |
1339 | else if (!before && dn->dn_bonuslen != 0) { | |
1340 | if (dn->dn_bonus) { | |
1341 | db = dn->dn_bonus; | |
1342 | mutex_enter(&db->db_mtx); | |
1343 | data = dmu_objset_userquota_find_data(db, tx); | |
1344 | } else { | |
1345 | data = DN_BONUS(dn->dn_phys); | |
1346 | } | |
1347 | } else if (dn->dn_bonuslen == 0 && dn->dn_bonustype == DMU_OT_SA) { | |
1348 | int rf = 0; | |
1349 | ||
1350 | if (RW_WRITE_HELD(&dn->dn_struct_rwlock)) | |
1351 | rf |= DB_RF_HAVESTRUCT; | |
572e2857 BB |
1352 | error = dmu_spill_hold_by_dnode(dn, |
1353 | rf | DB_RF_MUST_SUCCEED, | |
428870ff BB |
1354 | FTAG, (dmu_buf_t **)&db); |
1355 | ASSERT(error == 0); | |
1356 | mutex_enter(&db->db_mtx); | |
1357 | data = (before) ? db->db.db_data : | |
1358 | dmu_objset_userquota_find_data(db, tx); | |
1359 | have_spill = B_TRUE; | |
1360 | } else { | |
1361 | mutex_enter(&dn->dn_mtx); | |
1362 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1363 | mutex_exit(&dn->dn_mtx); | |
1364 | return; | |
1365 | } | |
1366 | ||
1367 | if (before) { | |
1368 | ASSERT(data); | |
1369 | user = &dn->dn_olduid; | |
1370 | group = &dn->dn_oldgid; | |
1371 | } else if (data) { | |
1372 | user = &dn->dn_newuid; | |
1373 | group = &dn->dn_newgid; | |
1374 | } | |
1375 | ||
1376 | /* | |
1377 | * Must always call the callback in case the object | |
1378 | * type has changed and that type isn't an object type to track | |
1379 | */ | |
1380 | error = used_cbs[os->os_phys->os_type](dn->dn_bonustype, data, | |
1381 | user, group); | |
1382 | ||
1383 | /* | |
1384 | * Preserve existing uid/gid when the callback can't determine | |
1385 | * what the new uid/gid are and the callback returned EEXIST. | |
1386 | * The EEXIST error tells us to just use the existing uid/gid. | |
1387 | * If we don't know what the old values are then just assign | |
1388 | * them to 0, since that is a new file being created. | |
1389 | */ | |
1390 | if (!before && data == NULL && error == EEXIST) { | |
1391 | if (flags & DN_ID_OLD_EXIST) { | |
1392 | dn->dn_newuid = dn->dn_olduid; | |
1393 | dn->dn_newgid = dn->dn_oldgid; | |
1394 | } else { | |
1395 | dn->dn_newuid = 0; | |
1396 | dn->dn_newgid = 0; | |
1397 | } | |
1398 | error = 0; | |
1399 | } | |
1400 | ||
1401 | if (db) | |
1402 | mutex_exit(&db->db_mtx); | |
1403 | ||
1404 | mutex_enter(&dn->dn_mtx); | |
1405 | if (error == 0 && before) | |
1406 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
1407 | if (error == 0 && !before) | |
1408 | dn->dn_id_flags |= DN_ID_NEW_EXIST; | |
1409 | ||
1410 | if (have_spill) { | |
1411 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1412 | } else { | |
1413 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1414 | } | |
1415 | mutex_exit(&dn->dn_mtx); | |
1416 | if (have_spill) | |
1417 | dmu_buf_rele((dmu_buf_t *)db, FTAG); | |
1418 | } | |
1419 | ||
9babb374 BB |
1420 | boolean_t |
1421 | dmu_objset_userspace_present(objset_t *os) | |
1422 | { | |
428870ff | 1423 | return (os->os_phys->os_flags & |
9babb374 BB |
1424 | OBJSET_FLAG_USERACCOUNTING_COMPLETE); |
1425 | } | |
1426 | ||
1427 | int | |
1428 | dmu_objset_userspace_upgrade(objset_t *os) | |
1429 | { | |
1430 | uint64_t obj; | |
1431 | int err = 0; | |
1432 | ||
1433 | if (dmu_objset_userspace_present(os)) | |
1434 | return (0); | |
428870ff | 1435 | if (!dmu_objset_userused_enabled(os)) |
9babb374 BB |
1436 | return (ENOTSUP); |
1437 | if (dmu_objset_is_snapshot(os)) | |
1438 | return (EINVAL); | |
1439 | ||
1440 | /* | |
1441 | * We simply need to mark every object dirty, so that it will be | |
1442 | * synced out and now accounted. If this is called | |
1443 | * concurrently, or if we already did some work before crashing, | |
1444 | * that's fine, since we track each object's accounted state | |
1445 | * independently. | |
1446 | */ | |
1447 | ||
1448 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
45d1cae3 | 1449 | dmu_tx_t *tx; |
9babb374 BB |
1450 | dmu_buf_t *db; |
1451 | int objerr; | |
1452 | ||
1453 | if (issig(JUSTLOOKING) && issig(FORREAL)) | |
1454 | return (EINTR); | |
1455 | ||
1456 | objerr = dmu_bonus_hold(os, obj, FTAG, &db); | |
1457 | if (objerr) | |
1458 | continue; | |
45d1cae3 | 1459 | tx = dmu_tx_create(os); |
9babb374 BB |
1460 | dmu_tx_hold_bonus(tx, obj); |
1461 | objerr = dmu_tx_assign(tx, TXG_WAIT); | |
1462 | if (objerr) { | |
1463 | dmu_tx_abort(tx); | |
1464 | continue; | |
1465 | } | |
1466 | dmu_buf_will_dirty(db, tx); | |
1467 | dmu_buf_rele(db, FTAG); | |
1468 | dmu_tx_commit(tx); | |
1469 | } | |
1470 | ||
428870ff | 1471 | os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; |
9babb374 BB |
1472 | txg_wait_synced(dmu_objset_pool(os), 0); |
1473 | return (0); | |
1474 | } | |
1475 | ||
34dc7c2f BB |
1476 | void |
1477 | dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, | |
1478 | uint64_t *usedobjsp, uint64_t *availobjsp) | |
1479 | { | |
428870ff | 1480 | dsl_dataset_space(os->os_dsl_dataset, refdbytesp, availbytesp, |
34dc7c2f BB |
1481 | usedobjsp, availobjsp); |
1482 | } | |
1483 | ||
1484 | uint64_t | |
1485 | dmu_objset_fsid_guid(objset_t *os) | |
1486 | { | |
428870ff | 1487 | return (dsl_dataset_fsid_guid(os->os_dsl_dataset)); |
34dc7c2f BB |
1488 | } |
1489 | ||
1490 | void | |
1491 | dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat) | |
1492 | { | |
428870ff BB |
1493 | stat->dds_type = os->os_phys->os_type; |
1494 | if (os->os_dsl_dataset) | |
1495 | dsl_dataset_fast_stat(os->os_dsl_dataset, stat); | |
34dc7c2f BB |
1496 | } |
1497 | ||
1498 | void | |
1499 | dmu_objset_stats(objset_t *os, nvlist_t *nv) | |
1500 | { | |
428870ff BB |
1501 | ASSERT(os->os_dsl_dataset || |
1502 | os->os_phys->os_type == DMU_OST_META); | |
34dc7c2f | 1503 | |
428870ff BB |
1504 | if (os->os_dsl_dataset != NULL) |
1505 | dsl_dataset_stats(os->os_dsl_dataset, nv); | |
34dc7c2f BB |
1506 | |
1507 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_TYPE, | |
428870ff | 1508 | os->os_phys->os_type); |
9babb374 BB |
1509 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERACCOUNTING, |
1510 | dmu_objset_userspace_present(os)); | |
34dc7c2f BB |
1511 | } |
1512 | ||
1513 | int | |
1514 | dmu_objset_is_snapshot(objset_t *os) | |
1515 | { | |
428870ff BB |
1516 | if (os->os_dsl_dataset != NULL) |
1517 | return (dsl_dataset_is_snapshot(os->os_dsl_dataset)); | |
34dc7c2f BB |
1518 | else |
1519 | return (B_FALSE); | |
1520 | } | |
1521 | ||
1522 | int | |
1523 | dmu_snapshot_realname(objset_t *os, char *name, char *real, int maxlen, | |
1524 | boolean_t *conflict) | |
1525 | { | |
428870ff | 1526 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
1527 | uint64_t ignored; |
1528 | ||
1529 | if (ds->ds_phys->ds_snapnames_zapobj == 0) | |
1530 | return (ENOENT); | |
1531 | ||
1532 | return (zap_lookup_norm(ds->ds_dir->dd_pool->dp_meta_objset, | |
1533 | ds->ds_phys->ds_snapnames_zapobj, name, 8, 1, &ignored, MT_FIRST, | |
1534 | real, maxlen, conflict)); | |
1535 | } | |
1536 | ||
1537 | int | |
1538 | dmu_snapshot_list_next(objset_t *os, int namelen, char *name, | |
1539 | uint64_t *idp, uint64_t *offp, boolean_t *case_conflict) | |
1540 | { | |
428870ff | 1541 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
1542 | zap_cursor_t cursor; |
1543 | zap_attribute_t attr; | |
1544 | ||
1545 | if (ds->ds_phys->ds_snapnames_zapobj == 0) | |
1546 | return (ENOENT); | |
1547 | ||
1548 | zap_cursor_init_serialized(&cursor, | |
1549 | ds->ds_dir->dd_pool->dp_meta_objset, | |
1550 | ds->ds_phys->ds_snapnames_zapobj, *offp); | |
1551 | ||
1552 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
1553 | zap_cursor_fini(&cursor); | |
1554 | return (ENOENT); | |
1555 | } | |
1556 | ||
1557 | if (strlen(attr.za_name) + 1 > namelen) { | |
1558 | zap_cursor_fini(&cursor); | |
1559 | return (ENAMETOOLONG); | |
1560 | } | |
1561 | ||
1562 | (void) strcpy(name, attr.za_name); | |
1563 | if (idp) | |
1564 | *idp = attr.za_first_integer; | |
1565 | if (case_conflict) | |
1566 | *case_conflict = attr.za_normalization_conflict; | |
1567 | zap_cursor_advance(&cursor); | |
1568 | *offp = zap_cursor_serialize(&cursor); | |
1569 | zap_cursor_fini(&cursor); | |
1570 | ||
1571 | return (0); | |
1572 | } | |
1573 | ||
ebe7e575 | 1574 | int |
6772fb67 | 1575 | dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *value) |
ebe7e575 | 1576 | { |
6772fb67 | 1577 | return dsl_dataset_snap_lookup(os->os_dsl_dataset, name, value); |
ebe7e575 BB |
1578 | } |
1579 | ||
34dc7c2f BB |
1580 | int |
1581 | dmu_dir_list_next(objset_t *os, int namelen, char *name, | |
1582 | uint64_t *idp, uint64_t *offp) | |
1583 | { | |
428870ff | 1584 | dsl_dir_t *dd = os->os_dsl_dataset->ds_dir; |
34dc7c2f BB |
1585 | zap_cursor_t cursor; |
1586 | zap_attribute_t attr; | |
1587 | ||
1588 | /* there is no next dir on a snapshot! */ | |
428870ff | 1589 | if (os->os_dsl_dataset->ds_object != |
34dc7c2f BB |
1590 | dd->dd_phys->dd_head_dataset_obj) |
1591 | return (ENOENT); | |
1592 | ||
1593 | zap_cursor_init_serialized(&cursor, | |
1594 | dd->dd_pool->dp_meta_objset, | |
1595 | dd->dd_phys->dd_child_dir_zapobj, *offp); | |
1596 | ||
1597 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
1598 | zap_cursor_fini(&cursor); | |
1599 | return (ENOENT); | |
1600 | } | |
1601 | ||
1602 | if (strlen(attr.za_name) + 1 > namelen) { | |
1603 | zap_cursor_fini(&cursor); | |
1604 | return (ENAMETOOLONG); | |
1605 | } | |
1606 | ||
1607 | (void) strcpy(name, attr.za_name); | |
1608 | if (idp) | |
1609 | *idp = attr.za_first_integer; | |
1610 | zap_cursor_advance(&cursor); | |
1611 | *offp = zap_cursor_serialize(&cursor); | |
1612 | zap_cursor_fini(&cursor); | |
1613 | ||
1614 | return (0); | |
1615 | } | |
1616 | ||
b128c09f | 1617 | struct findarg { |
428870ff | 1618 | int (*func)(const char *, void *); |
b128c09f BB |
1619 | void *arg; |
1620 | }; | |
1621 | ||
1622 | /* ARGSUSED */ | |
1623 | static int | |
1624 | findfunc(spa_t *spa, uint64_t dsobj, const char *dsname, void *arg) | |
1625 | { | |
1626 | struct findarg *fa = arg; | |
428870ff | 1627 | return (fa->func(dsname, fa->arg)); |
b128c09f BB |
1628 | } |
1629 | ||
34dc7c2f BB |
1630 | /* |
1631 | * Find all objsets under name, and for each, call 'func(child_name, arg)'. | |
b128c09f | 1632 | * Perhaps change all callers to use dmu_objset_find_spa()? |
34dc7c2f BB |
1633 | */ |
1634 | int | |
428870ff BB |
1635 | dmu_objset_find(char *name, int func(const char *, void *), void *arg, |
1636 | int flags) | |
b128c09f BB |
1637 | { |
1638 | struct findarg fa; | |
1639 | fa.func = func; | |
1640 | fa.arg = arg; | |
1641 | return (dmu_objset_find_spa(NULL, name, findfunc, &fa, flags)); | |
1642 | } | |
1643 | ||
1644 | /* | |
1645 | * Find all objsets under name, call func on each | |
1646 | */ | |
1647 | int | |
1648 | dmu_objset_find_spa(spa_t *spa, const char *name, | |
1649 | int func(spa_t *, uint64_t, const char *, void *), void *arg, int flags) | |
34dc7c2f BB |
1650 | { |
1651 | dsl_dir_t *dd; | |
b128c09f BB |
1652 | dsl_pool_t *dp; |
1653 | dsl_dataset_t *ds; | |
34dc7c2f BB |
1654 | zap_cursor_t zc; |
1655 | zap_attribute_t *attr; | |
1656 | char *child; | |
b128c09f BB |
1657 | uint64_t thisobj; |
1658 | int err; | |
34dc7c2f | 1659 | |
b128c09f BB |
1660 | if (name == NULL) |
1661 | name = spa_name(spa); | |
1662 | err = dsl_dir_open_spa(spa, name, FTAG, &dd, NULL); | |
34dc7c2f BB |
1663 | if (err) |
1664 | return (err); | |
1665 | ||
b128c09f BB |
1666 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ |
1667 | if (dd->dd_myname[0] == '$') { | |
1668 | dsl_dir_close(dd, FTAG); | |
1669 | return (0); | |
1670 | } | |
1671 | ||
1672 | thisobj = dd->dd_phys->dd_head_dataset_obj; | |
ed2e1576 | 1673 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_PUSHPAGE); |
b128c09f | 1674 | dp = dd->dd_pool; |
34dc7c2f BB |
1675 | |
1676 | /* | |
1677 | * Iterate over all children. | |
1678 | */ | |
1679 | if (flags & DS_FIND_CHILDREN) { | |
b128c09f | 1680 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
34dc7c2f BB |
1681 | dd->dd_phys->dd_child_dir_zapobj); |
1682 | zap_cursor_retrieve(&zc, attr) == 0; | |
1683 | (void) zap_cursor_advance(&zc)) { | |
1684 | ASSERT(attr->za_integer_length == sizeof (uint64_t)); | |
1685 | ASSERT(attr->za_num_integers == 1); | |
1686 | ||
428870ff | 1687 | child = kmem_asprintf("%s/%s", name, attr->za_name); |
b128c09f | 1688 | err = dmu_objset_find_spa(spa, child, func, arg, flags); |
428870ff | 1689 | strfree(child); |
34dc7c2f BB |
1690 | if (err) |
1691 | break; | |
1692 | } | |
1693 | zap_cursor_fini(&zc); | |
1694 | ||
1695 | if (err) { | |
1696 | dsl_dir_close(dd, FTAG); | |
1697 | kmem_free(attr, sizeof (zap_attribute_t)); | |
1698 | return (err); | |
1699 | } | |
1700 | } | |
1701 | ||
1702 | /* | |
1703 | * Iterate over all snapshots. | |
1704 | */ | |
b128c09f BB |
1705 | if (flags & DS_FIND_SNAPSHOTS) { |
1706 | if (!dsl_pool_sync_context(dp)) | |
1707 | rw_enter(&dp->dp_config_rwlock, RW_READER); | |
1708 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1709 | if (!dsl_pool_sync_context(dp)) | |
1710 | rw_exit(&dp->dp_config_rwlock); | |
1711 | ||
1712 | if (err == 0) { | |
1713 | uint64_t snapobj = ds->ds_phys->ds_snapnames_zapobj; | |
1714 | dsl_dataset_rele(ds, FTAG); | |
1715 | ||
1716 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
1717 | zap_cursor_retrieve(&zc, attr) == 0; | |
1718 | (void) zap_cursor_advance(&zc)) { | |
1719 | ASSERT(attr->za_integer_length == | |
1720 | sizeof (uint64_t)); | |
1721 | ASSERT(attr->za_num_integers == 1); | |
1722 | ||
428870ff BB |
1723 | child = kmem_asprintf("%s@%s", |
1724 | name, attr->za_name); | |
b128c09f BB |
1725 | err = func(spa, attr->za_first_integer, |
1726 | child, arg); | |
428870ff | 1727 | strfree(child); |
b128c09f BB |
1728 | if (err) |
1729 | break; | |
1730 | } | |
1731 | zap_cursor_fini(&zc); | |
34dc7c2f | 1732 | } |
34dc7c2f BB |
1733 | } |
1734 | ||
1735 | dsl_dir_close(dd, FTAG); | |
1736 | kmem_free(attr, sizeof (zap_attribute_t)); | |
1737 | ||
1738 | if (err) | |
1739 | return (err); | |
1740 | ||
1741 | /* | |
1742 | * Apply to self if appropriate. | |
1743 | */ | |
b128c09f | 1744 | err = func(spa, thisobj, name, arg); |
34dc7c2f BB |
1745 | return (err); |
1746 | } | |
1747 | ||
d164b209 BB |
1748 | /* ARGSUSED */ |
1749 | int | |
428870ff | 1750 | dmu_objset_prefetch(const char *name, void *arg) |
d164b209 BB |
1751 | { |
1752 | dsl_dataset_t *ds; | |
1753 | ||
1754 | if (dsl_dataset_hold(name, FTAG, &ds)) | |
1755 | return (0); | |
1756 | ||
1757 | if (!BP_IS_HOLE(&ds->ds_phys->ds_bp)) { | |
1758 | mutex_enter(&ds->ds_opening_lock); | |
428870ff | 1759 | if (ds->ds_objset == NULL) { |
d164b209 BB |
1760 | uint32_t aflags = ARC_NOWAIT | ARC_PREFETCH; |
1761 | zbookmark_t zb; | |
1762 | ||
428870ff BB |
1763 | SET_BOOKMARK(&zb, ds->ds_object, ZB_ROOT_OBJECT, |
1764 | ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
d164b209 | 1765 | |
294f6806 | 1766 | (void) arc_read(NULL, dsl_dataset_get_spa(ds), |
d164b209 BB |
1767 | &ds->ds_phys->ds_bp, NULL, NULL, |
1768 | ZIO_PRIORITY_ASYNC_READ, | |
1769 | ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, | |
1770 | &aflags, &zb); | |
1771 | } | |
1772 | mutex_exit(&ds->ds_opening_lock); | |
1773 | } | |
1774 | ||
1775 | dsl_dataset_rele(ds, FTAG); | |
1776 | return (0); | |
1777 | } | |
1778 | ||
34dc7c2f BB |
1779 | void |
1780 | dmu_objset_set_user(objset_t *os, void *user_ptr) | |
1781 | { | |
428870ff BB |
1782 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
1783 | os->os_user_ptr = user_ptr; | |
34dc7c2f BB |
1784 | } |
1785 | ||
1786 | void * | |
1787 | dmu_objset_get_user(objset_t *os) | |
1788 | { | |
428870ff BB |
1789 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
1790 | return (os->os_user_ptr); | |
34dc7c2f | 1791 | } |
c28b2279 BB |
1792 | |
1793 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
f0fd83be | 1794 | EXPORT_SYMBOL(dmu_objset_zil); |
c28b2279 | 1795 | EXPORT_SYMBOL(dmu_objset_pool); |
f0fd83be BB |
1796 | EXPORT_SYMBOL(dmu_objset_ds); |
1797 | EXPORT_SYMBOL(dmu_objset_type); | |
c28b2279 BB |
1798 | EXPORT_SYMBOL(dmu_objset_name); |
1799 | EXPORT_SYMBOL(dmu_objset_hold); | |
1800 | EXPORT_SYMBOL(dmu_objset_own); | |
1801 | EXPORT_SYMBOL(dmu_objset_rele); | |
1802 | EXPORT_SYMBOL(dmu_objset_disown); | |
1803 | EXPORT_SYMBOL(dmu_objset_from_ds); | |
1804 | EXPORT_SYMBOL(dmu_objset_create); | |
1805 | EXPORT_SYMBOL(dmu_objset_clone); | |
1806 | EXPORT_SYMBOL(dmu_objset_destroy); | |
1807 | EXPORT_SYMBOL(dmu_objset_snapshot); | |
1808 | EXPORT_SYMBOL(dmu_objset_stats); | |
1809 | EXPORT_SYMBOL(dmu_objset_fast_stat); | |
54a179e7 | 1810 | EXPORT_SYMBOL(dmu_objset_spa); |
c28b2279 BB |
1811 | EXPORT_SYMBOL(dmu_objset_space); |
1812 | EXPORT_SYMBOL(dmu_objset_fsid_guid); | |
1813 | EXPORT_SYMBOL(dmu_objset_find); | |
1814 | EXPORT_SYMBOL(dmu_objset_find_spa); | |
1815 | EXPORT_SYMBOL(dmu_objset_prefetch); | |
1816 | EXPORT_SYMBOL(dmu_objset_byteswap); | |
1817 | EXPORT_SYMBOL(dmu_objset_evict_dbufs); | |
1818 | EXPORT_SYMBOL(dmu_objset_snap_cmtime); | |
1819 | ||
1820 | EXPORT_SYMBOL(dmu_objset_sync); | |
1821 | EXPORT_SYMBOL(dmu_objset_is_dirty); | |
1822 | EXPORT_SYMBOL(dmu_objset_create_impl); | |
1823 | EXPORT_SYMBOL(dmu_objset_open_impl); | |
1824 | EXPORT_SYMBOL(dmu_objset_evict); | |
1825 | EXPORT_SYMBOL(dmu_objset_register_type); | |
1826 | EXPORT_SYMBOL(dmu_objset_do_userquota_updates); | |
1827 | EXPORT_SYMBOL(dmu_objset_userquota_get_ids); | |
1828 | EXPORT_SYMBOL(dmu_objset_userused_enabled); | |
1829 | EXPORT_SYMBOL(dmu_objset_userspace_upgrade); | |
1830 | EXPORT_SYMBOL(dmu_objset_userspace_present); | |
1831 | #endif |