]>
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. |
34dc7c2f BB |
25 | */ |
26 | ||
428870ff BB |
27 | /* Portions Copyright 2010 Robert Milkowski */ |
28 | ||
34dc7c2f BB |
29 | #include <sys/cred.h> |
30 | #include <sys/zfs_context.h> | |
31 | #include <sys/dmu_objset.h> | |
32 | #include <sys/dsl_dir.h> | |
33 | #include <sys/dsl_dataset.h> | |
34 | #include <sys/dsl_prop.h> | |
35 | #include <sys/dsl_pool.h> | |
36 | #include <sys/dsl_synctask.h> | |
37 | #include <sys/dsl_deleg.h> | |
38 | #include <sys/dnode.h> | |
39 | #include <sys/dbuf.h> | |
40 | #include <sys/zvol.h> | |
41 | #include <sys/dmu_tx.h> | |
34dc7c2f BB |
42 | #include <sys/zap.h> |
43 | #include <sys/zil.h> | |
44 | #include <sys/dmu_impl.h> | |
45 | #include <sys/zfs_ioctl.h> | |
428870ff | 46 | #include <sys/sa.h> |
572e2857 | 47 | #include <sys/zfs_onexit.h> |
13fe0198 | 48 | #include <sys/dsl_destroy.h> |
572e2857 BB |
49 | |
50 | /* | |
51 | * Needed to close a window in dnode_move() that allows the objset to be freed | |
52 | * before it can be safely accessed. | |
53 | */ | |
54 | krwlock_t os_lock; | |
55 | ||
56 | void | |
57 | dmu_objset_init(void) | |
58 | { | |
59 | rw_init(&os_lock, NULL, RW_DEFAULT, NULL); | |
60 | } | |
61 | ||
62 | void | |
63 | dmu_objset_fini(void) | |
64 | { | |
65 | rw_destroy(&os_lock); | |
66 | } | |
34dc7c2f BB |
67 | |
68 | spa_t * | |
69 | dmu_objset_spa(objset_t *os) | |
70 | { | |
428870ff | 71 | return (os->os_spa); |
34dc7c2f BB |
72 | } |
73 | ||
74 | zilog_t * | |
75 | dmu_objset_zil(objset_t *os) | |
76 | { | |
428870ff | 77 | return (os->os_zil); |
34dc7c2f BB |
78 | } |
79 | ||
80 | dsl_pool_t * | |
81 | dmu_objset_pool(objset_t *os) | |
82 | { | |
83 | dsl_dataset_t *ds; | |
84 | ||
428870ff | 85 | if ((ds = os->os_dsl_dataset) != NULL && ds->ds_dir) |
34dc7c2f BB |
86 | return (ds->ds_dir->dd_pool); |
87 | else | |
428870ff | 88 | return (spa_get_dsl(os->os_spa)); |
34dc7c2f BB |
89 | } |
90 | ||
91 | dsl_dataset_t * | |
92 | dmu_objset_ds(objset_t *os) | |
93 | { | |
428870ff | 94 | return (os->os_dsl_dataset); |
34dc7c2f BB |
95 | } |
96 | ||
97 | dmu_objset_type_t | |
98 | dmu_objset_type(objset_t *os) | |
99 | { | |
428870ff | 100 | return (os->os_phys->os_type); |
34dc7c2f BB |
101 | } |
102 | ||
103 | void | |
104 | dmu_objset_name(objset_t *os, char *buf) | |
105 | { | |
428870ff | 106 | dsl_dataset_name(os->os_dsl_dataset, buf); |
34dc7c2f BB |
107 | } |
108 | ||
109 | uint64_t | |
110 | dmu_objset_id(objset_t *os) | |
111 | { | |
428870ff | 112 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
113 | |
114 | return (ds ? ds->ds_object : 0); | |
115 | } | |
116 | ||
faf0f58c | 117 | zfs_sync_type_t |
428870ff BB |
118 | dmu_objset_syncprop(objset_t *os) |
119 | { | |
120 | return (os->os_sync); | |
121 | } | |
122 | ||
faf0f58c | 123 | zfs_logbias_op_t |
428870ff BB |
124 | dmu_objset_logbias(objset_t *os) |
125 | { | |
126 | return (os->os_logbias); | |
127 | } | |
128 | ||
34dc7c2f BB |
129 | static void |
130 | checksum_changed_cb(void *arg, uint64_t newval) | |
131 | { | |
428870ff | 132 | objset_t *os = arg; |
34dc7c2f BB |
133 | |
134 | /* | |
135 | * Inheritance should have been done by now. | |
136 | */ | |
137 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
138 | ||
428870ff | 139 | os->os_checksum = zio_checksum_select(newval, ZIO_CHECKSUM_ON_VALUE); |
34dc7c2f BB |
140 | } |
141 | ||
142 | static void | |
143 | compression_changed_cb(void *arg, uint64_t newval) | |
144 | { | |
428870ff | 145 | objset_t *os = arg; |
34dc7c2f BB |
146 | |
147 | /* | |
148 | * Inheritance and range checking should have been done by now. | |
149 | */ | |
150 | ASSERT(newval != ZIO_COMPRESS_INHERIT); | |
151 | ||
428870ff | 152 | os->os_compress = zio_compress_select(newval, ZIO_COMPRESS_ON_VALUE); |
34dc7c2f BB |
153 | } |
154 | ||
155 | static void | |
156 | copies_changed_cb(void *arg, uint64_t newval) | |
157 | { | |
428870ff | 158 | objset_t *os = arg; |
34dc7c2f BB |
159 | |
160 | /* | |
161 | * Inheritance and range checking should have been done by now. | |
162 | */ | |
163 | ASSERT(newval > 0); | |
428870ff | 164 | ASSERT(newval <= spa_max_replication(os->os_spa)); |
34dc7c2f | 165 | |
428870ff BB |
166 | os->os_copies = newval; |
167 | } | |
168 | ||
169 | static void | |
170 | dedup_changed_cb(void *arg, uint64_t newval) | |
171 | { | |
172 | objset_t *os = arg; | |
173 | spa_t *spa = os->os_spa; | |
174 | enum zio_checksum checksum; | |
175 | ||
176 | /* | |
177 | * Inheritance should have been done by now. | |
178 | */ | |
179 | ASSERT(newval != ZIO_CHECKSUM_INHERIT); | |
180 | ||
181 | checksum = zio_checksum_dedup_select(spa, newval, ZIO_CHECKSUM_OFF); | |
182 | ||
183 | os->os_dedup_checksum = checksum & ZIO_CHECKSUM_MASK; | |
184 | os->os_dedup_verify = !!(checksum & ZIO_CHECKSUM_VERIFY); | |
34dc7c2f BB |
185 | } |
186 | ||
b128c09f BB |
187 | static void |
188 | primary_cache_changed_cb(void *arg, uint64_t newval) | |
189 | { | |
428870ff | 190 | objset_t *os = arg; |
b128c09f BB |
191 | |
192 | /* | |
193 | * Inheritance and range checking should have been done by now. | |
194 | */ | |
195 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
196 | newval == ZFS_CACHE_METADATA); | |
197 | ||
428870ff | 198 | os->os_primary_cache = newval; |
b128c09f BB |
199 | } |
200 | ||
201 | static void | |
202 | secondary_cache_changed_cb(void *arg, uint64_t newval) | |
203 | { | |
428870ff | 204 | objset_t *os = arg; |
b128c09f BB |
205 | |
206 | /* | |
207 | * Inheritance and range checking should have been done by now. | |
208 | */ | |
209 | ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || | |
210 | newval == ZFS_CACHE_METADATA); | |
211 | ||
428870ff BB |
212 | os->os_secondary_cache = newval; |
213 | } | |
214 | ||
215 | static void | |
216 | sync_changed_cb(void *arg, uint64_t newval) | |
217 | { | |
218 | objset_t *os = arg; | |
219 | ||
220 | /* | |
221 | * Inheritance and range checking should have been done by now. | |
222 | */ | |
223 | ASSERT(newval == ZFS_SYNC_STANDARD || newval == ZFS_SYNC_ALWAYS || | |
224 | newval == ZFS_SYNC_DISABLED); | |
225 | ||
226 | os->os_sync = newval; | |
227 | if (os->os_zil) | |
228 | zil_set_sync(os->os_zil, newval); | |
229 | } | |
230 | ||
faf0f58c MA |
231 | static void |
232 | redundant_metadata_changed_cb(void *arg, uint64_t newval) | |
233 | { | |
234 | objset_t *os = arg; | |
235 | ||
236 | /* | |
237 | * Inheritance and range checking should have been done by now. | |
238 | */ | |
239 | ASSERT(newval == ZFS_REDUNDANT_METADATA_ALL || | |
240 | newval == ZFS_REDUNDANT_METADATA_MOST); | |
241 | ||
242 | os->os_redundant_metadata = newval; | |
243 | } | |
244 | ||
428870ff BB |
245 | static void |
246 | logbias_changed_cb(void *arg, uint64_t newval) | |
247 | { | |
248 | objset_t *os = arg; | |
249 | ||
250 | ASSERT(newval == ZFS_LOGBIAS_LATENCY || | |
251 | newval == ZFS_LOGBIAS_THROUGHPUT); | |
252 | os->os_logbias = newval; | |
253 | if (os->os_zil) | |
254 | zil_set_logbias(os->os_zil, newval); | |
b128c09f BB |
255 | } |
256 | ||
34dc7c2f BB |
257 | void |
258 | dmu_objset_byteswap(void *buf, size_t size) | |
259 | { | |
260 | objset_phys_t *osp = buf; | |
261 | ||
9babb374 | 262 | ASSERT(size == OBJSET_OLD_PHYS_SIZE || size == sizeof (objset_phys_t)); |
34dc7c2f BB |
263 | dnode_byteswap(&osp->os_meta_dnode); |
264 | byteswap_uint64_array(&osp->os_zil_header, sizeof (zil_header_t)); | |
265 | osp->os_type = BSWAP_64(osp->os_type); | |
9babb374 BB |
266 | osp->os_flags = BSWAP_64(osp->os_flags); |
267 | if (size == sizeof (objset_phys_t)) { | |
268 | dnode_byteswap(&osp->os_userused_dnode); | |
269 | dnode_byteswap(&osp->os_groupused_dnode); | |
270 | } | |
34dc7c2f BB |
271 | } |
272 | ||
273 | int | |
274 | dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, | |
428870ff | 275 | objset_t **osp) |
34dc7c2f | 276 | { |
428870ff | 277 | objset_t *os; |
b128c09f | 278 | int i, err; |
34dc7c2f BB |
279 | |
280 | ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock)); | |
281 | ||
b8d06fca | 282 | os = kmem_zalloc(sizeof (objset_t), KM_PUSHPAGE); |
428870ff BB |
283 | os->os_dsl_dataset = ds; |
284 | os->os_spa = spa; | |
285 | os->os_rootbp = bp; | |
286 | if (!BP_IS_HOLE(os->os_rootbp)) { | |
34dc7c2f BB |
287 | uint32_t aflags = ARC_WAIT; |
288 | zbookmark_t zb; | |
428870ff BB |
289 | SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET, |
290 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
291 | ||
292 | if (DMU_OS_IS_L2CACHEABLE(os)) | |
b128c09f | 293 | aflags |= ARC_L2CACHE; |
3a17a7a9 SK |
294 | if (DMU_OS_IS_L2COMPRESSIBLE(os)) |
295 | aflags |= ARC_L2COMPRESS; | |
34dc7c2f | 296 | |
428870ff | 297 | dprintf_bp(os->os_rootbp, "reading %s", ""); |
294f6806 | 298 | err = arc_read(NULL, spa, os->os_rootbp, |
428870ff | 299 | arc_getbuf_func, &os->os_phys_buf, |
34dc7c2f | 300 | ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL, &aflags, &zb); |
13fe0198 | 301 | if (err != 0) { |
428870ff | 302 | kmem_free(os, sizeof (objset_t)); |
b128c09f BB |
303 | /* convert checksum errors into IO errors */ |
304 | if (err == ECKSUM) | |
2e528b49 | 305 | err = SET_ERROR(EIO); |
34dc7c2f BB |
306 | return (err); |
307 | } | |
9babb374 BB |
308 | |
309 | /* Increase the blocksize if we are permitted. */ | |
310 | if (spa_version(spa) >= SPA_VERSION_USERSPACE && | |
428870ff | 311 | arc_buf_size(os->os_phys_buf) < sizeof (objset_phys_t)) { |
9babb374 | 312 | arc_buf_t *buf = arc_buf_alloc(spa, |
428870ff | 313 | sizeof (objset_phys_t), &os->os_phys_buf, |
9babb374 BB |
314 | ARC_BUFC_METADATA); |
315 | bzero(buf->b_data, sizeof (objset_phys_t)); | |
428870ff BB |
316 | bcopy(os->os_phys_buf->b_data, buf->b_data, |
317 | arc_buf_size(os->os_phys_buf)); | |
318 | (void) arc_buf_remove_ref(os->os_phys_buf, | |
319 | &os->os_phys_buf); | |
320 | os->os_phys_buf = buf; | |
9babb374 BB |
321 | } |
322 | ||
428870ff BB |
323 | os->os_phys = os->os_phys_buf->b_data; |
324 | os->os_flags = os->os_phys->os_flags; | |
34dc7c2f | 325 | } else { |
9babb374 BB |
326 | int size = spa_version(spa) >= SPA_VERSION_USERSPACE ? |
327 | sizeof (objset_phys_t) : OBJSET_OLD_PHYS_SIZE; | |
428870ff BB |
328 | os->os_phys_buf = arc_buf_alloc(spa, size, |
329 | &os->os_phys_buf, ARC_BUFC_METADATA); | |
330 | os->os_phys = os->os_phys_buf->b_data; | |
331 | bzero(os->os_phys, size); | |
34dc7c2f BB |
332 | } |
333 | ||
334 | /* | |
335 | * Note: the changed_cb will be called once before the register | |
336 | * func returns, thus changing the checksum/compression from the | |
b128c09f BB |
337 | * default (fletcher2/off). Snapshots don't need to know about |
338 | * checksum/compression/copies. | |
34dc7c2f | 339 | */ |
9b67f605 | 340 | if (ds != NULL) { |
13fe0198 MA |
341 | err = dsl_prop_register(ds, |
342 | zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE), | |
428870ff | 343 | primary_cache_changed_cb, os); |
13fe0198 MA |
344 | if (err == 0) { |
345 | err = dsl_prop_register(ds, | |
346 | zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE), | |
428870ff | 347 | secondary_cache_changed_cb, os); |
13fe0198 | 348 | } |
b128c09f | 349 | if (!dsl_dataset_is_snapshot(ds)) { |
13fe0198 MA |
350 | if (err == 0) { |
351 | err = dsl_prop_register(ds, | |
352 | zfs_prop_to_name(ZFS_PROP_CHECKSUM), | |
428870ff | 353 | checksum_changed_cb, os); |
13fe0198 MA |
354 | } |
355 | if (err == 0) { | |
356 | err = dsl_prop_register(ds, | |
357 | zfs_prop_to_name(ZFS_PROP_COMPRESSION), | |
428870ff | 358 | compression_changed_cb, os); |
13fe0198 MA |
359 | } |
360 | if (err == 0) { | |
361 | err = dsl_prop_register(ds, | |
362 | zfs_prop_to_name(ZFS_PROP_COPIES), | |
428870ff | 363 | copies_changed_cb, os); |
13fe0198 MA |
364 | } |
365 | if (err == 0) { | |
366 | err = dsl_prop_register(ds, | |
367 | zfs_prop_to_name(ZFS_PROP_DEDUP), | |
428870ff | 368 | dedup_changed_cb, os); |
13fe0198 MA |
369 | } |
370 | if (err == 0) { | |
371 | err = dsl_prop_register(ds, | |
372 | zfs_prop_to_name(ZFS_PROP_LOGBIAS), | |
428870ff | 373 | logbias_changed_cb, os); |
13fe0198 MA |
374 | } |
375 | if (err == 0) { | |
376 | err = dsl_prop_register(ds, | |
377 | zfs_prop_to_name(ZFS_PROP_SYNC), | |
428870ff | 378 | sync_changed_cb, os); |
13fe0198 | 379 | } |
faf0f58c MA |
380 | if (err == 0) { |
381 | err = dsl_prop_register(ds, | |
382 | zfs_prop_to_name( | |
383 | ZFS_PROP_REDUNDANT_METADATA), | |
384 | redundant_metadata_changed_cb, os); | |
385 | } | |
b128c09f | 386 | } |
13fe0198 | 387 | if (err != 0) { |
428870ff | 388 | VERIFY(arc_buf_remove_ref(os->os_phys_buf, |
13fe0198 | 389 | &os->os_phys_buf)); |
428870ff | 390 | kmem_free(os, sizeof (objset_t)); |
34dc7c2f BB |
391 | return (err); |
392 | } | |
9b67f605 | 393 | } else { |
34dc7c2f | 394 | /* It's the meta-objset. */ |
428870ff BB |
395 | os->os_checksum = ZIO_CHECKSUM_FLETCHER_4; |
396 | os->os_compress = ZIO_COMPRESS_LZJB; | |
397 | os->os_copies = spa_max_replication(spa); | |
398 | os->os_dedup_checksum = ZIO_CHECKSUM_OFF; | |
faf0f58c MA |
399 | os->os_dedup_verify = B_FALSE; |
400 | os->os_logbias = ZFS_LOGBIAS_LATENCY; | |
401 | os->os_sync = ZFS_SYNC_STANDARD; | |
428870ff BB |
402 | os->os_primary_cache = ZFS_CACHE_ALL; |
403 | os->os_secondary_cache = ZFS_CACHE_ALL; | |
34dc7c2f BB |
404 | } |
405 | ||
572e2857 BB |
406 | if (ds == NULL || !dsl_dataset_is_snapshot(ds)) |
407 | os->os_zil_header = os->os_phys->os_zil_header; | |
428870ff | 408 | os->os_zil = zil_alloc(os, &os->os_zil_header); |
34dc7c2f BB |
409 | |
410 | for (i = 0; i < TXG_SIZE; i++) { | |
428870ff | 411 | list_create(&os->os_dirty_dnodes[i], sizeof (dnode_t), |
34dc7c2f | 412 | offsetof(dnode_t, dn_dirty_link[i])); |
428870ff | 413 | list_create(&os->os_free_dnodes[i], sizeof (dnode_t), |
34dc7c2f BB |
414 | offsetof(dnode_t, dn_dirty_link[i])); |
415 | } | |
428870ff | 416 | list_create(&os->os_dnodes, sizeof (dnode_t), |
34dc7c2f | 417 | offsetof(dnode_t, dn_link)); |
428870ff | 418 | list_create(&os->os_downgraded_dbufs, sizeof (dmu_buf_impl_t), |
34dc7c2f BB |
419 | offsetof(dmu_buf_impl_t, db_link)); |
420 | ||
428870ff BB |
421 | mutex_init(&os->os_lock, NULL, MUTEX_DEFAULT, NULL); |
422 | mutex_init(&os->os_obj_lock, NULL, MUTEX_DEFAULT, NULL); | |
423 | mutex_init(&os->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL); | |
424 | ||
572e2857 BB |
425 | DMU_META_DNODE(os) = dnode_special_open(os, |
426 | &os->os_phys->os_meta_dnode, DMU_META_DNODE_OBJECT, | |
427 | &os->os_meta_dnode); | |
428870ff | 428 | if (arc_buf_size(os->os_phys_buf) >= sizeof (objset_phys_t)) { |
572e2857 BB |
429 | DMU_USERUSED_DNODE(os) = dnode_special_open(os, |
430 | &os->os_phys->os_userused_dnode, DMU_USERUSED_OBJECT, | |
431 | &os->os_userused_dnode); | |
432 | DMU_GROUPUSED_DNODE(os) = dnode_special_open(os, | |
433 | &os->os_phys->os_groupused_dnode, DMU_GROUPUSED_OBJECT, | |
434 | &os->os_groupused_dnode); | |
9babb374 | 435 | } |
34dc7c2f | 436 | |
428870ff | 437 | *osp = os; |
34dc7c2f BB |
438 | return (0); |
439 | } | |
440 | ||
428870ff BB |
441 | int |
442 | dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp) | |
34dc7c2f | 443 | { |
428870ff | 444 | int err = 0; |
34dc7c2f BB |
445 | |
446 | mutex_enter(&ds->ds_opening_lock); | |
9b67f605 MA |
447 | if (ds->ds_objset == NULL) { |
448 | objset_t *os; | |
34dc7c2f | 449 | err = dmu_objset_open_impl(dsl_dataset_get_spa(ds), |
9b67f605 MA |
450 | ds, dsl_dataset_get_blkptr(ds), &os); |
451 | ||
452 | if (err == 0) { | |
453 | mutex_enter(&ds->ds_lock); | |
454 | ASSERT(ds->ds_objset == NULL); | |
455 | ds->ds_objset = os; | |
456 | mutex_exit(&ds->ds_lock); | |
457 | } | |
34dc7c2f | 458 | } |
9b67f605 | 459 | *osp = ds->ds_objset; |
34dc7c2f | 460 | mutex_exit(&ds->ds_opening_lock); |
428870ff | 461 | return (err); |
34dc7c2f BB |
462 | } |
463 | ||
13fe0198 MA |
464 | /* |
465 | * Holds the pool while the objset is held. Therefore only one objset | |
466 | * can be held at a time. | |
467 | */ | |
34dc7c2f | 468 | int |
428870ff | 469 | dmu_objset_hold(const char *name, void *tag, objset_t **osp) |
34dc7c2f | 470 | { |
13fe0198 | 471 | dsl_pool_t *dp; |
428870ff | 472 | dsl_dataset_t *ds; |
34dc7c2f BB |
473 | int err; |
474 | ||
13fe0198 MA |
475 | err = dsl_pool_hold(name, tag, &dp); |
476 | if (err != 0) | |
477 | return (err); | |
478 | err = dsl_dataset_hold(dp, name, tag, &ds); | |
479 | if (err != 0) { | |
480 | dsl_pool_rele(dp, tag); | |
428870ff | 481 | return (err); |
13fe0198 | 482 | } |
428870ff BB |
483 | |
484 | err = dmu_objset_from_ds(ds, osp); | |
13fe0198 | 485 | if (err != 0) { |
428870ff | 486 | dsl_dataset_rele(ds, tag); |
13fe0198 MA |
487 | dsl_pool_rele(dp, tag); |
488 | } | |
428870ff | 489 | |
34dc7c2f BB |
490 | return (err); |
491 | } | |
492 | ||
13fe0198 MA |
493 | /* |
494 | * dsl_pool must not be held when this is called. | |
495 | * Upon successful return, there will be a longhold on the dataset, | |
496 | * and the dsl_pool will not be held. | |
497 | */ | |
34dc7c2f | 498 | int |
428870ff BB |
499 | dmu_objset_own(const char *name, dmu_objset_type_t type, |
500 | boolean_t readonly, void *tag, objset_t **osp) | |
34dc7c2f | 501 | { |
13fe0198 | 502 | dsl_pool_t *dp; |
34dc7c2f BB |
503 | dsl_dataset_t *ds; |
504 | int err; | |
505 | ||
13fe0198 MA |
506 | err = dsl_pool_hold(name, FTAG, &dp); |
507 | if (err != 0) | |
508 | return (err); | |
509 | err = dsl_dataset_own(dp, name, tag, &ds); | |
510 | if (err != 0) { | |
511 | dsl_pool_rele(dp, FTAG); | |
34dc7c2f | 512 | return (err); |
13fe0198 | 513 | } |
34dc7c2f | 514 | |
428870ff | 515 | err = dmu_objset_from_ds(ds, osp); |
13fe0198 MA |
516 | dsl_pool_rele(dp, FTAG); |
517 | if (err != 0) { | |
428870ff BB |
518 | dsl_dataset_disown(ds, tag); |
519 | } else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) { | |
13fe0198 | 520 | dsl_dataset_disown(ds, tag); |
2e528b49 | 521 | return (SET_ERROR(EINVAL)); |
428870ff | 522 | } else if (!readonly && dsl_dataset_is_snapshot(ds)) { |
13fe0198 | 523 | dsl_dataset_disown(ds, tag); |
2e528b49 | 524 | return (SET_ERROR(EROFS)); |
34dc7c2f BB |
525 | } |
526 | return (err); | |
527 | } | |
528 | ||
529 | void | |
428870ff | 530 | dmu_objset_rele(objset_t *os, void *tag) |
34dc7c2f | 531 | { |
13fe0198 | 532 | dsl_pool_t *dp = dmu_objset_pool(os); |
428870ff | 533 | dsl_dataset_rele(os->os_dsl_dataset, tag); |
13fe0198 | 534 | dsl_pool_rele(dp, tag); |
428870ff | 535 | } |
b128c09f | 536 | |
831baf06 KW |
537 | /* |
538 | * When we are called, os MUST refer to an objset associated with a dataset | |
539 | * that is owned by 'tag'; that is, is held and long held by 'tag' and ds_owner | |
540 | * == tag. We will then release and reacquire ownership of the dataset while | |
541 | * holding the pool config_rwlock to avoid intervening namespace or ownership | |
542 | * changes may occur. | |
543 | * | |
544 | * This exists solely to accommodate zfs_ioc_userspace_upgrade()'s desire to | |
545 | * release the hold on its dataset and acquire a new one on the dataset of the | |
546 | * same name so that it can be partially torn down and reconstructed. | |
547 | */ | |
548 | void | |
549 | dmu_objset_refresh_ownership(objset_t *os, void *tag) | |
550 | { | |
551 | dsl_pool_t *dp; | |
552 | dsl_dataset_t *ds, *newds; | |
553 | char name[MAXNAMELEN]; | |
554 | ||
555 | ds = os->os_dsl_dataset; | |
556 | VERIFY3P(ds, !=, NULL); | |
557 | VERIFY3P(ds->ds_owner, ==, tag); | |
558 | VERIFY(dsl_dataset_long_held(ds)); | |
559 | ||
560 | dsl_dataset_name(ds, name); | |
561 | dp = dmu_objset_pool(os); | |
562 | dsl_pool_config_enter(dp, FTAG); | |
563 | dmu_objset_disown(os, tag); | |
564 | VERIFY0(dsl_dataset_own(dp, name, tag, &newds)); | |
565 | VERIFY3P(newds, ==, os->os_dsl_dataset); | |
566 | dsl_pool_config_exit(dp, FTAG); | |
567 | } | |
568 | ||
428870ff BB |
569 | void |
570 | dmu_objset_disown(objset_t *os, void *tag) | |
571 | { | |
572 | dsl_dataset_disown(os->os_dsl_dataset, tag); | |
34dc7c2f BB |
573 | } |
574 | ||
13fe0198 | 575 | void |
34dc7c2f BB |
576 | dmu_objset_evict_dbufs(objset_t *os) |
577 | { | |
34dc7c2f BB |
578 | dnode_t *dn; |
579 | ||
428870ff | 580 | mutex_enter(&os->os_lock); |
34dc7c2f BB |
581 | |
582 | /* process the mdn last, since the other dnodes have holds on it */ | |
572e2857 BB |
583 | list_remove(&os->os_dnodes, DMU_META_DNODE(os)); |
584 | list_insert_tail(&os->os_dnodes, DMU_META_DNODE(os)); | |
34dc7c2f BB |
585 | |
586 | /* | |
587 | * Find the first dnode with holds. We have to do this dance | |
588 | * because dnode_add_ref() only works if you already have a | |
589 | * hold. If there are no holds then it has no dbufs so OK to | |
590 | * skip. | |
591 | */ | |
428870ff | 592 | for (dn = list_head(&os->os_dnodes); |
34dc7c2f | 593 | dn && !dnode_add_ref(dn, FTAG); |
428870ff | 594 | dn = list_next(&os->os_dnodes, dn)) |
34dc7c2f BB |
595 | continue; |
596 | ||
597 | while (dn) { | |
598 | dnode_t *next_dn = dn; | |
599 | ||
600 | do { | |
428870ff | 601 | next_dn = list_next(&os->os_dnodes, next_dn); |
34dc7c2f BB |
602 | } while (next_dn && !dnode_add_ref(next_dn, FTAG)); |
603 | ||
428870ff | 604 | mutex_exit(&os->os_lock); |
34dc7c2f BB |
605 | dnode_evict_dbufs(dn); |
606 | dnode_rele(dn, FTAG); | |
428870ff | 607 | mutex_enter(&os->os_lock); |
34dc7c2f BB |
608 | dn = next_dn; |
609 | } | |
428870ff | 610 | mutex_exit(&os->os_lock); |
34dc7c2f BB |
611 | } |
612 | ||
613 | void | |
428870ff | 614 | dmu_objset_evict(objset_t *os) |
34dc7c2f | 615 | { |
d6320ddb | 616 | int t; |
34dc7c2f | 617 | |
6f1ffb06 MA |
618 | dsl_dataset_t *ds = os->os_dsl_dataset; |
619 | ||
d6320ddb | 620 | for (t = 0; t < TXG_SIZE; t++) |
428870ff | 621 | ASSERT(!dmu_objset_is_dirty(os, t)); |
34dc7c2f | 622 | |
b128c09f BB |
623 | if (ds) { |
624 | if (!dsl_dataset_is_snapshot(ds)) { | |
13fe0198 MA |
625 | VERIFY0(dsl_prop_unregister(ds, |
626 | zfs_prop_to_name(ZFS_PROP_CHECKSUM), | |
428870ff | 627 | checksum_changed_cb, os)); |
13fe0198 MA |
628 | VERIFY0(dsl_prop_unregister(ds, |
629 | zfs_prop_to_name(ZFS_PROP_COMPRESSION), | |
428870ff | 630 | compression_changed_cb, os)); |
13fe0198 MA |
631 | VERIFY0(dsl_prop_unregister(ds, |
632 | zfs_prop_to_name(ZFS_PROP_COPIES), | |
428870ff | 633 | copies_changed_cb, os)); |
13fe0198 MA |
634 | VERIFY0(dsl_prop_unregister(ds, |
635 | zfs_prop_to_name(ZFS_PROP_DEDUP), | |
428870ff | 636 | dedup_changed_cb, os)); |
13fe0198 MA |
637 | VERIFY0(dsl_prop_unregister(ds, |
638 | zfs_prop_to_name(ZFS_PROP_LOGBIAS), | |
428870ff | 639 | logbias_changed_cb, os)); |
13fe0198 MA |
640 | VERIFY0(dsl_prop_unregister(ds, |
641 | zfs_prop_to_name(ZFS_PROP_SYNC), | |
428870ff | 642 | sync_changed_cb, os)); |
faf0f58c MA |
643 | VERIFY0(dsl_prop_unregister(ds, |
644 | zfs_prop_to_name(ZFS_PROP_REDUNDANT_METADATA), | |
645 | redundant_metadata_changed_cb, os)); | |
b128c09f | 646 | } |
13fe0198 MA |
647 | VERIFY0(dsl_prop_unregister(ds, |
648 | zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE), | |
428870ff | 649 | primary_cache_changed_cb, os)); |
13fe0198 MA |
650 | VERIFY0(dsl_prop_unregister(ds, |
651 | zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE), | |
428870ff | 652 | secondary_cache_changed_cb, os)); |
34dc7c2f BB |
653 | } |
654 | ||
428870ff BB |
655 | if (os->os_sa) |
656 | sa_tear_down(os); | |
657 | ||
13fe0198 | 658 | dmu_objset_evict_dbufs(os); |
34dc7c2f | 659 | |
572e2857 BB |
660 | dnode_special_close(&os->os_meta_dnode); |
661 | if (DMU_USERUSED_DNODE(os)) { | |
662 | dnode_special_close(&os->os_userused_dnode); | |
663 | dnode_special_close(&os->os_groupused_dnode); | |
9babb374 | 664 | } |
428870ff BB |
665 | zil_free(os->os_zil); |
666 | ||
667 | ASSERT3P(list_head(&os->os_dnodes), ==, NULL); | |
34dc7c2f | 668 | |
13fe0198 | 669 | VERIFY(arc_buf_remove_ref(os->os_phys_buf, &os->os_phys_buf)); |
572e2857 BB |
670 | |
671 | /* | |
672 | * This is a barrier to prevent the objset from going away in | |
673 | * dnode_move() until we can safely ensure that the objset is still in | |
674 | * use. We consider the objset valid before the barrier and invalid | |
675 | * after the barrier. | |
676 | */ | |
677 | rw_enter(&os_lock, RW_READER); | |
678 | rw_exit(&os_lock); | |
679 | ||
428870ff BB |
680 | mutex_destroy(&os->os_lock); |
681 | mutex_destroy(&os->os_obj_lock); | |
682 | mutex_destroy(&os->os_user_ptr_lock); | |
683 | kmem_free(os, sizeof (objset_t)); | |
684 | } | |
9babb374 | 685 | |
428870ff BB |
686 | timestruc_t |
687 | dmu_objset_snap_cmtime(objset_t *os) | |
688 | { | |
689 | return (dsl_dir_snap_cmtime(os->os_dsl_dataset->ds_dir)); | |
34dc7c2f BB |
690 | } |
691 | ||
692 | /* called from dsl for meta-objset */ | |
428870ff | 693 | objset_t * |
34dc7c2f BB |
694 | dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, |
695 | dmu_objset_type_t type, dmu_tx_t *tx) | |
696 | { | |
428870ff | 697 | objset_t *os; |
34dc7c2f BB |
698 | dnode_t *mdn; |
699 | ||
700 | ASSERT(dmu_tx_is_syncing(tx)); | |
13fe0198 | 701 | |
572e2857 | 702 | if (ds != NULL) |
13fe0198 | 703 | VERIFY0(dmu_objset_from_ds(ds, &os)); |
572e2857 | 704 | else |
13fe0198 | 705 | VERIFY0(dmu_objset_open_impl(spa, NULL, bp, &os)); |
572e2857 BB |
706 | |
707 | mdn = DMU_META_DNODE(os); | |
34dc7c2f BB |
708 | |
709 | dnode_allocate(mdn, DMU_OT_DNODE, 1 << DNODE_BLOCK_SHIFT, | |
710 | DN_MAX_INDBLKSHIFT, DMU_OT_NONE, 0, tx); | |
711 | ||
712 | /* | |
713 | * We don't want to have to increase the meta-dnode's nlevels | |
714 | * later, because then we could do it in quescing context while | |
715 | * we are also accessing it in open context. | |
716 | * | |
717 | * This precaution is not necessary for the MOS (ds == NULL), | |
718 | * because the MOS is only updated in syncing context. | |
719 | * This is most fortunate: the MOS is the only objset that | |
720 | * needs to be synced multiple times as spa_sync() iterates | |
721 | * to convergence, so minimizing its dn_nlevels matters. | |
722 | */ | |
723 | if (ds != NULL) { | |
724 | int levels = 1; | |
725 | ||
726 | /* | |
727 | * Determine the number of levels necessary for the meta-dnode | |
728 | * to contain DN_MAX_OBJECT dnodes. | |
729 | */ | |
730 | while ((uint64_t)mdn->dn_nblkptr << (mdn->dn_datablkshift + | |
731 | (levels - 1) * (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) < | |
732 | DN_MAX_OBJECT * sizeof (dnode_phys_t)) | |
733 | levels++; | |
734 | ||
735 | mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] = | |
736 | mdn->dn_nlevels = levels; | |
737 | } | |
738 | ||
739 | ASSERT(type != DMU_OST_NONE); | |
740 | ASSERT(type != DMU_OST_ANY); | |
741 | ASSERT(type < DMU_OST_NUMTYPES); | |
428870ff BB |
742 | os->os_phys->os_type = type; |
743 | if (dmu_objset_userused_enabled(os)) { | |
744 | os->os_phys->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; | |
745 | os->os_flags = os->os_phys->os_flags; | |
9babb374 | 746 | } |
34dc7c2f BB |
747 | |
748 | dsl_dataset_dirty(ds, tx); | |
749 | ||
428870ff | 750 | return (os); |
34dc7c2f BB |
751 | } |
752 | ||
13fe0198 MA |
753 | typedef struct dmu_objset_create_arg { |
754 | const char *doca_name; | |
755 | cred_t *doca_cred; | |
756 | void (*doca_userfunc)(objset_t *os, void *arg, | |
757 | cred_t *cr, dmu_tx_t *tx); | |
758 | void *doca_userarg; | |
759 | dmu_objset_type_t doca_type; | |
760 | uint64_t doca_flags; | |
761 | } dmu_objset_create_arg_t; | |
34dc7c2f BB |
762 | |
763 | /*ARGSUSED*/ | |
764 | static int | |
13fe0198 | 765 | dmu_objset_create_check(void *arg, dmu_tx_t *tx) |
34dc7c2f | 766 | { |
13fe0198 MA |
767 | dmu_objset_create_arg_t *doca = arg; |
768 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
769 | dsl_dir_t *pdd; | |
770 | const char *tail; | |
771 | int error; | |
34dc7c2f | 772 | |
13fe0198 | 773 | if (strchr(doca->doca_name, '@') != NULL) |
2e528b49 | 774 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 775 | |
13fe0198 MA |
776 | error = dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail); |
777 | if (error != 0) | |
778 | return (error); | |
779 | if (tail == NULL) { | |
780 | dsl_dir_rele(pdd, FTAG); | |
2e528b49 | 781 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 782 | } |
13fe0198 | 783 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f BB |
784 | |
785 | return (0); | |
786 | } | |
787 | ||
788 | static void | |
13fe0198 | 789 | dmu_objset_create_sync(void *arg, dmu_tx_t *tx) |
34dc7c2f | 790 | { |
13fe0198 MA |
791 | dmu_objset_create_arg_t *doca = arg; |
792 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
793 | dsl_dir_t *pdd; | |
794 | const char *tail; | |
6f1ffb06 | 795 | dsl_dataset_t *ds; |
13fe0198 | 796 | uint64_t obj; |
6f1ffb06 | 797 | blkptr_t *bp; |
13fe0198 | 798 | objset_t *os; |
34dc7c2f | 799 | |
13fe0198 | 800 | VERIFY0(dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail)); |
34dc7c2f | 801 | |
13fe0198 MA |
802 | obj = dsl_dataset_create_sync(pdd, tail, NULL, doca->doca_flags, |
803 | doca->doca_cred, tx); | |
34dc7c2f | 804 | |
13fe0198 | 805 | VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); |
6f1ffb06 | 806 | bp = dsl_dataset_get_blkptr(ds); |
13fe0198 MA |
807 | os = dmu_objset_create_impl(pdd->dd_pool->dp_spa, |
808 | ds, bp, doca->doca_type, tx); | |
34dc7c2f | 809 | |
13fe0198 MA |
810 | if (doca->doca_userfunc != NULL) { |
811 | doca->doca_userfunc(os, doca->doca_userarg, | |
812 | doca->doca_cred, tx); | |
34dc7c2f BB |
813 | } |
814 | ||
13fe0198 | 815 | spa_history_log_internal_ds(ds, "create", tx, ""); |
6f1ffb06 | 816 | dsl_dataset_rele(ds, FTAG); |
13fe0198 | 817 | dsl_dir_rele(pdd, FTAG); |
34dc7c2f BB |
818 | } |
819 | ||
820 | int | |
428870ff | 821 | dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, |
34dc7c2f BB |
822 | void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg) |
823 | { | |
13fe0198 | 824 | dmu_objset_create_arg_t doca; |
34dc7c2f | 825 | |
13fe0198 MA |
826 | doca.doca_name = name; |
827 | doca.doca_cred = CRED(); | |
828 | doca.doca_flags = flags; | |
829 | doca.doca_userfunc = func; | |
830 | doca.doca_userarg = arg; | |
831 | doca.doca_type = type; | |
34dc7c2f | 832 | |
13fe0198 MA |
833 | return (dsl_sync_task(name, |
834 | dmu_objset_create_check, dmu_objset_create_sync, &doca, 5)); | |
34dc7c2f BB |
835 | } |
836 | ||
13fe0198 MA |
837 | typedef struct dmu_objset_clone_arg { |
838 | const char *doca_clone; | |
839 | const char *doca_origin; | |
840 | cred_t *doca_cred; | |
841 | } dmu_objset_clone_arg_t; | |
842 | ||
843 | /*ARGSUSED*/ | |
844 | static int | |
845 | dmu_objset_clone_check(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 846 | { |
13fe0198 | 847 | dmu_objset_clone_arg_t *doca = arg; |
428870ff BB |
848 | dsl_dir_t *pdd; |
849 | const char *tail; | |
13fe0198 MA |
850 | int error; |
851 | dsl_dataset_t *origin; | |
852 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
34dc7c2f | 853 | |
13fe0198 | 854 | if (strchr(doca->doca_clone, '@') != NULL) |
2e528b49 | 855 | return (SET_ERROR(EINVAL)); |
13fe0198 MA |
856 | |
857 | error = dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail); | |
858 | if (error != 0) | |
859 | return (error); | |
428870ff | 860 | if (tail == NULL) { |
13fe0198 | 861 | dsl_dir_rele(pdd, FTAG); |
2e528b49 | 862 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 863 | } |
13fe0198 MA |
864 | /* You can't clone across pools. */ |
865 | if (pdd->dd_pool != dp) { | |
866 | dsl_dir_rele(pdd, FTAG); | |
2e528b49 | 867 | return (SET_ERROR(EXDEV)); |
428870ff | 868 | } |
13fe0198 | 869 | dsl_dir_rele(pdd, FTAG); |
428870ff | 870 | |
13fe0198 MA |
871 | error = dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin); |
872 | if (error != 0) | |
572e2857 BB |
873 | return (error); |
874 | ||
13fe0198 MA |
875 | /* You can't clone across pools. */ |
876 | if (origin->ds_dir->dd_pool != dp) { | |
877 | dsl_dataset_rele(origin, FTAG); | |
2e528b49 | 878 | return (SET_ERROR(EXDEV)); |
572e2857 | 879 | } |
9babb374 | 880 | |
13fe0198 MA |
881 | /* You can only clone snapshots, not the head datasets. */ |
882 | if (!dsl_dataset_is_snapshot(origin)) { | |
883 | dsl_dataset_rele(origin, FTAG); | |
2e528b49 | 884 | return (SET_ERROR(EINVAL)); |
428870ff | 885 | } |
13fe0198 | 886 | dsl_dataset_rele(origin, FTAG); |
572e2857 | 887 | |
13fe0198 | 888 | return (0); |
9babb374 | 889 | } |
34dc7c2f | 890 | |
13fe0198 MA |
891 | static void |
892 | dmu_objset_clone_sync(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 893 | { |
13fe0198 MA |
894 | dmu_objset_clone_arg_t *doca = arg; |
895 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
896 | dsl_dir_t *pdd; | |
897 | const char *tail; | |
898 | dsl_dataset_t *origin, *ds; | |
899 | uint64_t obj; | |
900 | char namebuf[MAXNAMELEN]; | |
6f1ffb06 | 901 | |
13fe0198 MA |
902 | VERIFY0(dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail)); |
903 | VERIFY0(dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin)); | |
6f1ffb06 | 904 | |
13fe0198 MA |
905 | obj = dsl_dataset_create_sync(pdd, tail, origin, 0, |
906 | doca->doca_cred, tx); | |
34dc7c2f | 907 | |
13fe0198 MA |
908 | VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); |
909 | dsl_dataset_name(origin, namebuf); | |
910 | spa_history_log_internal_ds(ds, "clone", tx, | |
911 | "origin=%s (%llu)", namebuf, origin->ds_object); | |
912 | dsl_dataset_rele(ds, FTAG); | |
913 | dsl_dataset_rele(origin, FTAG); | |
914 | dsl_dir_rele(pdd, FTAG); | |
34dc7c2f BB |
915 | } |
916 | ||
917 | int | |
13fe0198 | 918 | dmu_objset_clone(const char *clone, const char *origin) |
34dc7c2f | 919 | { |
13fe0198 | 920 | dmu_objset_clone_arg_t doca; |
34dc7c2f | 921 | |
13fe0198 MA |
922 | doca.doca_clone = clone; |
923 | doca.doca_origin = origin; | |
924 | doca.doca_cred = CRED(); | |
572e2857 | 925 | |
13fe0198 MA |
926 | return (dsl_sync_task(clone, |
927 | dmu_objset_clone_check, dmu_objset_clone_sync, &doca, 5)); | |
6f1ffb06 MA |
928 | } |
929 | ||
930 | int | |
931 | dmu_objset_snapshot_one(const char *fsname, const char *snapname) | |
932 | { | |
933 | int err; | |
934 | char *longsnap = kmem_asprintf("%s@%s", fsname, snapname); | |
935 | nvlist_t *snaps = fnvlist_alloc(); | |
936 | ||
937 | fnvlist_add_boolean(snaps, longsnap); | |
6f1ffb06 | 938 | strfree(longsnap); |
13fe0198 MA |
939 | err = dsl_dataset_snapshot(snaps, NULL, NULL); |
940 | fnvlist_free(snaps); | |
6f1ffb06 MA |
941 | return (err); |
942 | } | |
943 | ||
34dc7c2f | 944 | static void |
9babb374 | 945 | dmu_objset_sync_dnodes(list_t *list, list_t *newlist, dmu_tx_t *tx) |
34dc7c2f BB |
946 | { |
947 | dnode_t *dn; | |
948 | ||
c65aa5b2 | 949 | while ((dn = list_head(list))) { |
34dc7c2f BB |
950 | ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT); |
951 | ASSERT(dn->dn_dbuf->db_data_pending); | |
952 | /* | |
9babb374 BB |
953 | * Initialize dn_zio outside dnode_sync() because the |
954 | * meta-dnode needs to set it ouside dnode_sync(). | |
34dc7c2f BB |
955 | */ |
956 | dn->dn_zio = dn->dn_dbuf->db_data_pending->dr_zio; | |
957 | ASSERT(dn->dn_zio); | |
958 | ||
959 | ASSERT3U(dn->dn_nlevels, <=, DN_MAX_LEVELS); | |
960 | list_remove(list, dn); | |
9babb374 BB |
961 | |
962 | if (newlist) { | |
963 | (void) dnode_add_ref(dn, newlist); | |
964 | list_insert_tail(newlist, dn); | |
965 | } | |
966 | ||
34dc7c2f BB |
967 | dnode_sync(dn, tx); |
968 | } | |
969 | } | |
970 | ||
971 | /* ARGSUSED */ | |
972 | static void | |
428870ff | 973 | dmu_objset_write_ready(zio_t *zio, arc_buf_t *abuf, void *arg) |
34dc7c2f | 974 | { |
d6320ddb BB |
975 | int i; |
976 | ||
b128c09f | 977 | blkptr_t *bp = zio->io_bp; |
428870ff | 978 | objset_t *os = arg; |
34dc7c2f | 979 | dnode_phys_t *dnp = &os->os_phys->os_meta_dnode; |
34dc7c2f | 980 | |
9b67f605 | 981 | ASSERT(!BP_IS_EMBEDDED(bp)); |
13fe0198 MA |
982 | ASSERT3P(bp, ==, os->os_rootbp); |
983 | ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_OBJSET); | |
984 | ASSERT0(BP_GET_LEVEL(bp)); | |
34dc7c2f BB |
985 | |
986 | /* | |
9babb374 BB |
987 | * Update rootbp fill count: it should be the number of objects |
988 | * allocated in the object set (not counting the "special" | |
989 | * objects that are stored in the objset_phys_t -- the meta | |
990 | * dnode and user/group accounting objects). | |
34dc7c2f | 991 | */ |
9babb374 | 992 | bp->blk_fill = 0; |
d6320ddb | 993 | for (i = 0; i < dnp->dn_nblkptr; i++) |
9b67f605 | 994 | bp->blk_fill += BP_GET_FILL(&dnp->dn_blkptr[i]); |
428870ff BB |
995 | } |
996 | ||
997 | /* ARGSUSED */ | |
998 | static void | |
999 | dmu_objset_write_done(zio_t *zio, arc_buf_t *abuf, void *arg) | |
1000 | { | |
1001 | blkptr_t *bp = zio->io_bp; | |
1002 | blkptr_t *bp_orig = &zio->io_bp_orig; | |
1003 | objset_t *os = arg; | |
34dc7c2f | 1004 | |
b128c09f | 1005 | if (zio->io_flags & ZIO_FLAG_IO_REWRITE) { |
428870ff | 1006 | ASSERT(BP_EQUAL(bp, bp_orig)); |
b128c09f | 1007 | } else { |
428870ff BB |
1008 | dsl_dataset_t *ds = os->os_dsl_dataset; |
1009 | dmu_tx_t *tx = os->os_synctx; | |
1010 | ||
1011 | (void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE); | |
1012 | dsl_dataset_block_born(ds, bp, tx); | |
34dc7c2f BB |
1013 | } |
1014 | } | |
1015 | ||
34dc7c2f BB |
1016 | /* called from dsl */ |
1017 | void | |
428870ff | 1018 | dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx) |
34dc7c2f BB |
1019 | { |
1020 | int txgoff; | |
1021 | zbookmark_t zb; | |
428870ff | 1022 | zio_prop_t zp; |
34dc7c2f BB |
1023 | zio_t *zio; |
1024 | list_t *list; | |
9babb374 | 1025 | list_t *newlist = NULL; |
34dc7c2f BB |
1026 | dbuf_dirty_record_t *dr; |
1027 | ||
1028 | dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg); | |
1029 | ||
1030 | ASSERT(dmu_tx_is_syncing(tx)); | |
1031 | /* XXX the write_done callback should really give us the tx... */ | |
1032 | os->os_synctx = tx; | |
1033 | ||
1034 | if (os->os_dsl_dataset == NULL) { | |
1035 | /* | |
1036 | * This is the MOS. If we have upgraded, | |
1037 | * spa_max_replication() could change, so reset | |
1038 | * os_copies here. | |
1039 | */ | |
1040 | os->os_copies = spa_max_replication(os->os_spa); | |
1041 | } | |
1042 | ||
1043 | /* | |
1044 | * Create the root block IO | |
1045 | */ | |
428870ff BB |
1046 | SET_BOOKMARK(&zb, os->os_dsl_dataset ? |
1047 | os->os_dsl_dataset->ds_object : DMU_META_OBJSET, | |
1048 | ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); | |
294f6806 | 1049 | arc_release(os->os_phys_buf, &os->os_phys_buf); |
b128c09f | 1050 | |
428870ff | 1051 | dmu_write_policy(os, NULL, 0, 0, &zp); |
9babb374 | 1052 | |
428870ff | 1053 | zio = arc_write(pio, os->os_spa, tx->tx_txg, |
3a17a7a9 SK |
1054 | os->os_rootbp, os->os_phys_buf, DMU_OS_IS_L2CACHEABLE(os), |
1055 | DMU_OS_IS_L2COMPRESSIBLE(os), &zp, dmu_objset_write_ready, | |
e8b96c60 | 1056 | NULL, dmu_objset_write_done, os, ZIO_PRIORITY_ASYNC_WRITE, |
3a17a7a9 | 1057 | ZIO_FLAG_MUSTSUCCEED, &zb); |
34dc7c2f BB |
1058 | |
1059 | /* | |
9babb374 | 1060 | * Sync special dnodes - the parent IO for the sync is the root block |
34dc7c2f | 1061 | */ |
572e2857 BB |
1062 | DMU_META_DNODE(os)->dn_zio = zio; |
1063 | dnode_sync(DMU_META_DNODE(os), tx); | |
34dc7c2f | 1064 | |
9babb374 BB |
1065 | os->os_phys->os_flags = os->os_flags; |
1066 | ||
572e2857 BB |
1067 | if (DMU_USERUSED_DNODE(os) && |
1068 | DMU_USERUSED_DNODE(os)->dn_type != DMU_OT_NONE) { | |
1069 | DMU_USERUSED_DNODE(os)->dn_zio = zio; | |
1070 | dnode_sync(DMU_USERUSED_DNODE(os), tx); | |
1071 | DMU_GROUPUSED_DNODE(os)->dn_zio = zio; | |
1072 | dnode_sync(DMU_GROUPUSED_DNODE(os), tx); | |
9babb374 BB |
1073 | } |
1074 | ||
34dc7c2f BB |
1075 | txgoff = tx->tx_txg & TXG_MASK; |
1076 | ||
9babb374 BB |
1077 | if (dmu_objset_userused_enabled(os)) { |
1078 | newlist = &os->os_synced_dnodes; | |
1079 | /* | |
1080 | * We must create the list here because it uses the | |
1081 | * dn_dirty_link[] of this txg. | |
1082 | */ | |
1083 | list_create(newlist, sizeof (dnode_t), | |
1084 | offsetof(dnode_t, dn_dirty_link[txgoff])); | |
1085 | } | |
1086 | ||
1087 | dmu_objset_sync_dnodes(&os->os_free_dnodes[txgoff], newlist, tx); | |
1088 | dmu_objset_sync_dnodes(&os->os_dirty_dnodes[txgoff], newlist, tx); | |
34dc7c2f | 1089 | |
572e2857 | 1090 | list = &DMU_META_DNODE(os)->dn_dirty_records[txgoff]; |
6f1ffb06 | 1091 | while ((dr = list_head(list))) { |
13fe0198 | 1092 | ASSERT0(dr->dr_dbuf->db_level); |
34dc7c2f BB |
1093 | list_remove(list, dr); |
1094 | if (dr->dr_zio) | |
1095 | zio_nowait(dr->dr_zio); | |
1096 | } | |
1097 | /* | |
1098 | * Free intent log blocks up to this tx. | |
1099 | */ | |
1100 | zil_sync(os->os_zil, tx); | |
b128c09f | 1101 | os->os_phys->os_zil_header = os->os_zil_header; |
34dc7c2f BB |
1102 | zio_nowait(zio); |
1103 | } | |
1104 | ||
428870ff BB |
1105 | boolean_t |
1106 | dmu_objset_is_dirty(objset_t *os, uint64_t txg) | |
1107 | { | |
1108 | return (!list_is_empty(&os->os_dirty_dnodes[txg & TXG_MASK]) || | |
1109 | !list_is_empty(&os->os_free_dnodes[txg & TXG_MASK])); | |
1110 | } | |
1111 | ||
572e2857 | 1112 | static objset_used_cb_t *used_cbs[DMU_OST_NUMTYPES]; |
9babb374 BB |
1113 | |
1114 | void | |
1115 | dmu_objset_register_type(dmu_objset_type_t ost, objset_used_cb_t *cb) | |
1116 | { | |
1117 | used_cbs[ost] = cb; | |
1118 | } | |
1119 | ||
1120 | boolean_t | |
428870ff | 1121 | dmu_objset_userused_enabled(objset_t *os) |
9babb374 BB |
1122 | { |
1123 | return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE && | |
572e2857 BB |
1124 | used_cbs[os->os_phys->os_type] != NULL && |
1125 | DMU_USERUSED_DNODE(os) != NULL); | |
9babb374 BB |
1126 | } |
1127 | ||
428870ff BB |
1128 | static void |
1129 | do_userquota_update(objset_t *os, uint64_t used, uint64_t flags, | |
1130 | uint64_t user, uint64_t group, boolean_t subtract, dmu_tx_t *tx) | |
1131 | { | |
1132 | if ((flags & DNODE_FLAG_USERUSED_ACCOUNTED)) { | |
1133 | int64_t delta = DNODE_SIZE + used; | |
1134 | if (subtract) | |
1135 | delta = -delta; | |
1136 | VERIFY3U(0, ==, zap_increment_int(os, DMU_USERUSED_OBJECT, | |
1137 | user, delta, tx)); | |
1138 | VERIFY3U(0, ==, zap_increment_int(os, DMU_GROUPUSED_OBJECT, | |
1139 | group, delta, tx)); | |
1140 | } | |
1141 | } | |
1142 | ||
9babb374 | 1143 | void |
428870ff | 1144 | dmu_objset_do_userquota_updates(objset_t *os, dmu_tx_t *tx) |
9babb374 BB |
1145 | { |
1146 | dnode_t *dn; | |
1147 | list_t *list = &os->os_synced_dnodes; | |
9babb374 BB |
1148 | |
1149 | ASSERT(list_head(list) == NULL || dmu_objset_userused_enabled(os)); | |
1150 | ||
6f1ffb06 | 1151 | while ((dn = list_head(list))) { |
572e2857 | 1152 | int flags; |
9babb374 | 1153 | ASSERT(!DMU_OBJECT_IS_SPECIAL(dn->dn_object)); |
9babb374 BB |
1154 | ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE || |
1155 | dn->dn_phys->dn_flags & | |
1156 | DNODE_FLAG_USERUSED_ACCOUNTED); | |
1157 | ||
1158 | /* Allocate the user/groupused objects if necessary. */ | |
572e2857 | 1159 | if (DMU_USERUSED_DNODE(os)->dn_type == DMU_OT_NONE) { |
428870ff | 1160 | VERIFY(0 == zap_create_claim(os, |
9babb374 BB |
1161 | DMU_USERUSED_OBJECT, |
1162 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
428870ff | 1163 | VERIFY(0 == zap_create_claim(os, |
9babb374 BB |
1164 | DMU_GROUPUSED_OBJECT, |
1165 | DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); | |
1166 | } | |
1167 | ||
1168 | /* | |
428870ff BB |
1169 | * We intentionally modify the zap object even if the |
1170 | * net delta is zero. Otherwise | |
1171 | * the block of the zap obj could be shared between | |
1172 | * datasets but need to be different between them after | |
1173 | * a bprewrite. | |
9babb374 | 1174 | */ |
9babb374 | 1175 | |
572e2857 BB |
1176 | flags = dn->dn_id_flags; |
1177 | ASSERT(flags); | |
1178 | if (flags & DN_ID_OLD_EXIST) { | |
428870ff BB |
1179 | do_userquota_update(os, dn->dn_oldused, dn->dn_oldflags, |
1180 | dn->dn_olduid, dn->dn_oldgid, B_TRUE, tx); | |
1181 | } | |
572e2857 | 1182 | if (flags & DN_ID_NEW_EXIST) { |
428870ff BB |
1183 | do_userquota_update(os, DN_USED_BYTES(dn->dn_phys), |
1184 | dn->dn_phys->dn_flags, dn->dn_newuid, | |
1185 | dn->dn_newgid, B_FALSE, tx); | |
1186 | } | |
1187 | ||
572e2857 | 1188 | mutex_enter(&dn->dn_mtx); |
428870ff BB |
1189 | dn->dn_oldused = 0; |
1190 | dn->dn_oldflags = 0; | |
1191 | if (dn->dn_id_flags & DN_ID_NEW_EXIST) { | |
1192 | dn->dn_olduid = dn->dn_newuid; | |
1193 | dn->dn_oldgid = dn->dn_newgid; | |
1194 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
1195 | if (dn->dn_bonuslen == 0) | |
1196 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1197 | else | |
1198 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1199 | } | |
1200 | dn->dn_id_flags &= ~(DN_ID_NEW_EXIST); | |
9babb374 BB |
1201 | mutex_exit(&dn->dn_mtx); |
1202 | ||
1203 | list_remove(list, dn); | |
1204 | dnode_rele(dn, list); | |
1205 | } | |
1206 | } | |
1207 | ||
428870ff BB |
1208 | /* |
1209 | * Returns a pointer to data to find uid/gid from | |
1210 | * | |
1211 | * If a dirty record for transaction group that is syncing can't | |
1212 | * be found then NULL is returned. In the NULL case it is assumed | |
1213 | * the uid/gid aren't changing. | |
1214 | */ | |
1215 | static void * | |
1216 | dmu_objset_userquota_find_data(dmu_buf_impl_t *db, dmu_tx_t *tx) | |
1217 | { | |
1218 | dbuf_dirty_record_t *dr, **drp; | |
1219 | void *data; | |
1220 | ||
1221 | if (db->db_dirtycnt == 0) | |
1222 | return (db->db.db_data); /* Nothing is changing */ | |
1223 | ||
1224 | for (drp = &db->db_last_dirty; (dr = *drp) != NULL; drp = &dr->dr_next) | |
1225 | if (dr->dr_txg == tx->tx_txg) | |
1226 | break; | |
1227 | ||
572e2857 | 1228 | if (dr == NULL) { |
428870ff | 1229 | data = NULL; |
572e2857 BB |
1230 | } else { |
1231 | dnode_t *dn; | |
1232 | ||
1233 | DB_DNODE_ENTER(dr->dr_dbuf); | |
1234 | dn = DB_DNODE(dr->dr_dbuf); | |
1235 | ||
1236 | if (dn->dn_bonuslen == 0 && | |
1237 | dr->dr_dbuf->db_blkid == DMU_SPILL_BLKID) | |
1238 | data = dr->dt.dl.dr_data->b_data; | |
1239 | else | |
1240 | data = dr->dt.dl.dr_data; | |
1241 | ||
1242 | DB_DNODE_EXIT(dr->dr_dbuf); | |
1243 | } | |
1244 | ||
428870ff BB |
1245 | return (data); |
1246 | } | |
1247 | ||
1248 | void | |
1249 | dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) | |
1250 | { | |
1251 | objset_t *os = dn->dn_objset; | |
1252 | void *data = NULL; | |
1253 | dmu_buf_impl_t *db = NULL; | |
a117a6d6 GW |
1254 | uint64_t *user = NULL; |
1255 | uint64_t *group = NULL; | |
428870ff BB |
1256 | int flags = dn->dn_id_flags; |
1257 | int error; | |
1258 | boolean_t have_spill = B_FALSE; | |
1259 | ||
1260 | if (!dmu_objset_userused_enabled(dn->dn_objset)) | |
1261 | return; | |
1262 | ||
1263 | if (before && (flags & (DN_ID_CHKED_BONUS|DN_ID_OLD_EXIST| | |
1264 | DN_ID_CHKED_SPILL))) | |
1265 | return; | |
1266 | ||
1267 | if (before && dn->dn_bonuslen != 0) | |
1268 | data = DN_BONUS(dn->dn_phys); | |
1269 | else if (!before && dn->dn_bonuslen != 0) { | |
1270 | if (dn->dn_bonus) { | |
1271 | db = dn->dn_bonus; | |
1272 | mutex_enter(&db->db_mtx); | |
1273 | data = dmu_objset_userquota_find_data(db, tx); | |
1274 | } else { | |
1275 | data = DN_BONUS(dn->dn_phys); | |
1276 | } | |
1277 | } else if (dn->dn_bonuslen == 0 && dn->dn_bonustype == DMU_OT_SA) { | |
1278 | int rf = 0; | |
1279 | ||
1280 | if (RW_WRITE_HELD(&dn->dn_struct_rwlock)) | |
1281 | rf |= DB_RF_HAVESTRUCT; | |
572e2857 BB |
1282 | error = dmu_spill_hold_by_dnode(dn, |
1283 | rf | DB_RF_MUST_SUCCEED, | |
428870ff BB |
1284 | FTAG, (dmu_buf_t **)&db); |
1285 | ASSERT(error == 0); | |
1286 | mutex_enter(&db->db_mtx); | |
1287 | data = (before) ? db->db.db_data : | |
1288 | dmu_objset_userquota_find_data(db, tx); | |
1289 | have_spill = B_TRUE; | |
1290 | } else { | |
1291 | mutex_enter(&dn->dn_mtx); | |
1292 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1293 | mutex_exit(&dn->dn_mtx); | |
1294 | return; | |
1295 | } | |
1296 | ||
1297 | if (before) { | |
1298 | ASSERT(data); | |
1299 | user = &dn->dn_olduid; | |
1300 | group = &dn->dn_oldgid; | |
1301 | } else if (data) { | |
1302 | user = &dn->dn_newuid; | |
1303 | group = &dn->dn_newgid; | |
1304 | } | |
1305 | ||
1306 | /* | |
1307 | * Must always call the callback in case the object | |
1308 | * type has changed and that type isn't an object type to track | |
1309 | */ | |
1310 | error = used_cbs[os->os_phys->os_type](dn->dn_bonustype, data, | |
1311 | user, group); | |
1312 | ||
1313 | /* | |
1314 | * Preserve existing uid/gid when the callback can't determine | |
1315 | * what the new uid/gid are and the callback returned EEXIST. | |
1316 | * The EEXIST error tells us to just use the existing uid/gid. | |
1317 | * If we don't know what the old values are then just assign | |
1318 | * them to 0, since that is a new file being created. | |
1319 | */ | |
1320 | if (!before && data == NULL && error == EEXIST) { | |
1321 | if (flags & DN_ID_OLD_EXIST) { | |
1322 | dn->dn_newuid = dn->dn_olduid; | |
1323 | dn->dn_newgid = dn->dn_oldgid; | |
1324 | } else { | |
1325 | dn->dn_newuid = 0; | |
1326 | dn->dn_newgid = 0; | |
1327 | } | |
1328 | error = 0; | |
1329 | } | |
1330 | ||
1331 | if (db) | |
1332 | mutex_exit(&db->db_mtx); | |
1333 | ||
1334 | mutex_enter(&dn->dn_mtx); | |
1335 | if (error == 0 && before) | |
1336 | dn->dn_id_flags |= DN_ID_OLD_EXIST; | |
1337 | if (error == 0 && !before) | |
1338 | dn->dn_id_flags |= DN_ID_NEW_EXIST; | |
1339 | ||
1340 | if (have_spill) { | |
1341 | dn->dn_id_flags |= DN_ID_CHKED_SPILL; | |
1342 | } else { | |
1343 | dn->dn_id_flags |= DN_ID_CHKED_BONUS; | |
1344 | } | |
1345 | mutex_exit(&dn->dn_mtx); | |
1346 | if (have_spill) | |
1347 | dmu_buf_rele((dmu_buf_t *)db, FTAG); | |
1348 | } | |
1349 | ||
9babb374 BB |
1350 | boolean_t |
1351 | dmu_objset_userspace_present(objset_t *os) | |
1352 | { | |
428870ff | 1353 | return (os->os_phys->os_flags & |
9babb374 BB |
1354 | OBJSET_FLAG_USERACCOUNTING_COMPLETE); |
1355 | } | |
1356 | ||
1357 | int | |
1358 | dmu_objset_userspace_upgrade(objset_t *os) | |
1359 | { | |
1360 | uint64_t obj; | |
1361 | int err = 0; | |
1362 | ||
1363 | if (dmu_objset_userspace_present(os)) | |
1364 | return (0); | |
428870ff | 1365 | if (!dmu_objset_userused_enabled(os)) |
2e528b49 | 1366 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 1367 | if (dmu_objset_is_snapshot(os)) |
2e528b49 | 1368 | return (SET_ERROR(EINVAL)); |
9babb374 BB |
1369 | |
1370 | /* | |
1371 | * We simply need to mark every object dirty, so that it will be | |
1372 | * synced out and now accounted. If this is called | |
1373 | * concurrently, or if we already did some work before crashing, | |
1374 | * that's fine, since we track each object's accounted state | |
1375 | * independently. | |
1376 | */ | |
1377 | ||
1378 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
45d1cae3 | 1379 | dmu_tx_t *tx; |
9babb374 BB |
1380 | dmu_buf_t *db; |
1381 | int objerr; | |
1382 | ||
1383 | if (issig(JUSTLOOKING) && issig(FORREAL)) | |
2e528b49 | 1384 | return (SET_ERROR(EINTR)); |
9babb374 BB |
1385 | |
1386 | objerr = dmu_bonus_hold(os, obj, FTAG, &db); | |
13fe0198 | 1387 | if (objerr != 0) |
9babb374 | 1388 | continue; |
45d1cae3 | 1389 | tx = dmu_tx_create(os); |
9babb374 BB |
1390 | dmu_tx_hold_bonus(tx, obj); |
1391 | objerr = dmu_tx_assign(tx, TXG_WAIT); | |
13fe0198 | 1392 | if (objerr != 0) { |
9babb374 BB |
1393 | dmu_tx_abort(tx); |
1394 | continue; | |
1395 | } | |
1396 | dmu_buf_will_dirty(db, tx); | |
1397 | dmu_buf_rele(db, FTAG); | |
1398 | dmu_tx_commit(tx); | |
1399 | } | |
1400 | ||
428870ff | 1401 | os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; |
9babb374 BB |
1402 | txg_wait_synced(dmu_objset_pool(os), 0); |
1403 | return (0); | |
1404 | } | |
1405 | ||
34dc7c2f BB |
1406 | void |
1407 | dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, | |
1408 | uint64_t *usedobjsp, uint64_t *availobjsp) | |
1409 | { | |
428870ff | 1410 | dsl_dataset_space(os->os_dsl_dataset, refdbytesp, availbytesp, |
34dc7c2f BB |
1411 | usedobjsp, availobjsp); |
1412 | } | |
1413 | ||
1414 | uint64_t | |
1415 | dmu_objset_fsid_guid(objset_t *os) | |
1416 | { | |
428870ff | 1417 | return (dsl_dataset_fsid_guid(os->os_dsl_dataset)); |
34dc7c2f BB |
1418 | } |
1419 | ||
1420 | void | |
1421 | dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat) | |
1422 | { | |
428870ff BB |
1423 | stat->dds_type = os->os_phys->os_type; |
1424 | if (os->os_dsl_dataset) | |
1425 | dsl_dataset_fast_stat(os->os_dsl_dataset, stat); | |
34dc7c2f BB |
1426 | } |
1427 | ||
1428 | void | |
1429 | dmu_objset_stats(objset_t *os, nvlist_t *nv) | |
1430 | { | |
428870ff BB |
1431 | ASSERT(os->os_dsl_dataset || |
1432 | os->os_phys->os_type == DMU_OST_META); | |
34dc7c2f | 1433 | |
428870ff BB |
1434 | if (os->os_dsl_dataset != NULL) |
1435 | dsl_dataset_stats(os->os_dsl_dataset, nv); | |
34dc7c2f BB |
1436 | |
1437 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_TYPE, | |
428870ff | 1438 | os->os_phys->os_type); |
9babb374 BB |
1439 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERACCOUNTING, |
1440 | dmu_objset_userspace_present(os)); | |
34dc7c2f BB |
1441 | } |
1442 | ||
1443 | int | |
1444 | dmu_objset_is_snapshot(objset_t *os) | |
1445 | { | |
428870ff BB |
1446 | if (os->os_dsl_dataset != NULL) |
1447 | return (dsl_dataset_is_snapshot(os->os_dsl_dataset)); | |
34dc7c2f BB |
1448 | else |
1449 | return (B_FALSE); | |
1450 | } | |
1451 | ||
1452 | int | |
1453 | dmu_snapshot_realname(objset_t *os, char *name, char *real, int maxlen, | |
1454 | boolean_t *conflict) | |
1455 | { | |
428870ff | 1456 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
1457 | uint64_t ignored; |
1458 | ||
1459 | if (ds->ds_phys->ds_snapnames_zapobj == 0) | |
2e528b49 | 1460 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1461 | |
1462 | return (zap_lookup_norm(ds->ds_dir->dd_pool->dp_meta_objset, | |
1463 | ds->ds_phys->ds_snapnames_zapobj, name, 8, 1, &ignored, MT_FIRST, | |
1464 | real, maxlen, conflict)); | |
1465 | } | |
1466 | ||
1467 | int | |
1468 | dmu_snapshot_list_next(objset_t *os, int namelen, char *name, | |
1469 | uint64_t *idp, uint64_t *offp, boolean_t *case_conflict) | |
1470 | { | |
428870ff | 1471 | dsl_dataset_t *ds = os->os_dsl_dataset; |
34dc7c2f BB |
1472 | zap_cursor_t cursor; |
1473 | zap_attribute_t attr; | |
1474 | ||
13fe0198 MA |
1475 | ASSERT(dsl_pool_config_held(dmu_objset_pool(os))); |
1476 | ||
34dc7c2f | 1477 | if (ds->ds_phys->ds_snapnames_zapobj == 0) |
2e528b49 | 1478 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1479 | |
1480 | zap_cursor_init_serialized(&cursor, | |
1481 | ds->ds_dir->dd_pool->dp_meta_objset, | |
1482 | ds->ds_phys->ds_snapnames_zapobj, *offp); | |
1483 | ||
1484 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
1485 | zap_cursor_fini(&cursor); | |
2e528b49 | 1486 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1487 | } |
1488 | ||
1489 | if (strlen(attr.za_name) + 1 > namelen) { | |
1490 | zap_cursor_fini(&cursor); | |
2e528b49 | 1491 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
1492 | } |
1493 | ||
1494 | (void) strcpy(name, attr.za_name); | |
1495 | if (idp) | |
1496 | *idp = attr.za_first_integer; | |
1497 | if (case_conflict) | |
1498 | *case_conflict = attr.za_normalization_conflict; | |
1499 | zap_cursor_advance(&cursor); | |
1500 | *offp = zap_cursor_serialize(&cursor); | |
1501 | zap_cursor_fini(&cursor); | |
1502 | ||
1503 | return (0); | |
1504 | } | |
1505 | ||
ebe7e575 | 1506 | int |
6772fb67 | 1507 | dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *value) |
ebe7e575 | 1508 | { |
d1d7e268 | 1509 | return (dsl_dataset_snap_lookup(os->os_dsl_dataset, name, value)); |
ebe7e575 BB |
1510 | } |
1511 | ||
34dc7c2f BB |
1512 | int |
1513 | dmu_dir_list_next(objset_t *os, int namelen, char *name, | |
1514 | uint64_t *idp, uint64_t *offp) | |
1515 | { | |
428870ff | 1516 | dsl_dir_t *dd = os->os_dsl_dataset->ds_dir; |
34dc7c2f BB |
1517 | zap_cursor_t cursor; |
1518 | zap_attribute_t attr; | |
1519 | ||
1520 | /* there is no next dir on a snapshot! */ | |
428870ff | 1521 | if (os->os_dsl_dataset->ds_object != |
34dc7c2f | 1522 | dd->dd_phys->dd_head_dataset_obj) |
2e528b49 | 1523 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1524 | |
1525 | zap_cursor_init_serialized(&cursor, | |
1526 | dd->dd_pool->dp_meta_objset, | |
1527 | dd->dd_phys->dd_child_dir_zapobj, *offp); | |
1528 | ||
1529 | if (zap_cursor_retrieve(&cursor, &attr) != 0) { | |
1530 | zap_cursor_fini(&cursor); | |
2e528b49 | 1531 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1532 | } |
1533 | ||
1534 | if (strlen(attr.za_name) + 1 > namelen) { | |
1535 | zap_cursor_fini(&cursor); | |
2e528b49 | 1536 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
1537 | } |
1538 | ||
1539 | (void) strcpy(name, attr.za_name); | |
1540 | if (idp) | |
1541 | *idp = attr.za_first_integer; | |
1542 | zap_cursor_advance(&cursor); | |
1543 | *offp = zap_cursor_serialize(&cursor); | |
1544 | zap_cursor_fini(&cursor); | |
1545 | ||
1546 | return (0); | |
1547 | } | |
1548 | ||
1549 | /* | |
13fe0198 | 1550 | * Find objsets under and including ddobj, call func(ds) on each. |
34dc7c2f BB |
1551 | */ |
1552 | int | |
13fe0198 MA |
1553 | dmu_objset_find_dp(dsl_pool_t *dp, uint64_t ddobj, |
1554 | int func(dsl_pool_t *, dsl_dataset_t *, void *), void *arg, int flags) | |
b128c09f | 1555 | { |
13fe0198 MA |
1556 | dsl_dir_t *dd; |
1557 | dsl_dataset_t *ds; | |
1558 | zap_cursor_t zc; | |
1559 | zap_attribute_t *attr; | |
1560 | uint64_t thisobj; | |
1561 | int err; | |
1562 | ||
1563 | ASSERT(dsl_pool_config_held(dp)); | |
1564 | ||
1565 | err = dsl_dir_hold_obj(dp, ddobj, NULL, FTAG, &dd); | |
1566 | if (err != 0) | |
1567 | return (err); | |
1568 | ||
1569 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ | |
1570 | if (dd->dd_myname[0] == '$') { | |
1571 | dsl_dir_rele(dd, FTAG); | |
1572 | return (0); | |
1573 | } | |
1574 | ||
1575 | thisobj = dd->dd_phys->dd_head_dataset_obj; | |
4cf652e5 | 1576 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_PUSHPAGE); |
13fe0198 MA |
1577 | |
1578 | /* | |
1579 | * Iterate over all children. | |
1580 | */ | |
1581 | if (flags & DS_FIND_CHILDREN) { | |
1582 | for (zap_cursor_init(&zc, dp->dp_meta_objset, | |
1583 | dd->dd_phys->dd_child_dir_zapobj); | |
1584 | zap_cursor_retrieve(&zc, attr) == 0; | |
1585 | (void) zap_cursor_advance(&zc)) { | |
1586 | ASSERT3U(attr->za_integer_length, ==, | |
1587 | sizeof (uint64_t)); | |
1588 | ASSERT3U(attr->za_num_integers, ==, 1); | |
1589 | ||
1590 | err = dmu_objset_find_dp(dp, attr->za_first_integer, | |
1591 | func, arg, flags); | |
1592 | if (err != 0) | |
1593 | break; | |
1594 | } | |
1595 | zap_cursor_fini(&zc); | |
1596 | ||
1597 | if (err != 0) { | |
1598 | dsl_dir_rele(dd, FTAG); | |
1599 | kmem_free(attr, sizeof (zap_attribute_t)); | |
1600 | return (err); | |
1601 | } | |
1602 | } | |
1603 | ||
1604 | /* | |
1605 | * Iterate over all snapshots. | |
1606 | */ | |
1607 | if (flags & DS_FIND_SNAPSHOTS) { | |
1608 | dsl_dataset_t *ds; | |
1609 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1610 | ||
1611 | if (err == 0) { | |
1612 | uint64_t snapobj = ds->ds_phys->ds_snapnames_zapobj; | |
1613 | dsl_dataset_rele(ds, FTAG); | |
1614 | ||
1615 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
1616 | zap_cursor_retrieve(&zc, attr) == 0; | |
1617 | (void) zap_cursor_advance(&zc)) { | |
1618 | ASSERT3U(attr->za_integer_length, ==, | |
1619 | sizeof (uint64_t)); | |
1620 | ASSERT3U(attr->za_num_integers, ==, 1); | |
1621 | ||
1622 | err = dsl_dataset_hold_obj(dp, | |
1623 | attr->za_first_integer, FTAG, &ds); | |
1624 | if (err != 0) | |
1625 | break; | |
1626 | err = func(dp, ds, arg); | |
1627 | dsl_dataset_rele(ds, FTAG); | |
1628 | if (err != 0) | |
1629 | break; | |
1630 | } | |
1631 | zap_cursor_fini(&zc); | |
1632 | } | |
1633 | } | |
1634 | ||
1635 | dsl_dir_rele(dd, FTAG); | |
1636 | kmem_free(attr, sizeof (zap_attribute_t)); | |
1637 | ||
1638 | if (err != 0) | |
1639 | return (err); | |
1640 | ||
1641 | /* | |
1642 | * Apply to self. | |
1643 | */ | |
1644 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); | |
1645 | if (err != 0) | |
1646 | return (err); | |
1647 | err = func(dp, ds, arg); | |
1648 | dsl_dataset_rele(ds, FTAG); | |
1649 | return (err); | |
b128c09f BB |
1650 | } |
1651 | ||
1652 | /* | |
13fe0198 MA |
1653 | * Find all objsets under name, and for each, call 'func(child_name, arg)'. |
1654 | * The dp_config_rwlock must not be held when this is called, and it | |
1655 | * will not be held when the callback is called. | |
1656 | * Therefore this function should only be used when the pool is not changing | |
1657 | * (e.g. in syncing context), or the callback can deal with the possible races. | |
b128c09f | 1658 | */ |
13fe0198 MA |
1659 | static int |
1660 | dmu_objset_find_impl(spa_t *spa, const char *name, | |
1661 | int func(const char *, void *), void *arg, int flags) | |
34dc7c2f BB |
1662 | { |
1663 | dsl_dir_t *dd; | |
13fe0198 | 1664 | dsl_pool_t *dp = spa_get_dsl(spa); |
b128c09f | 1665 | dsl_dataset_t *ds; |
34dc7c2f BB |
1666 | zap_cursor_t zc; |
1667 | zap_attribute_t *attr; | |
1668 | char *child; | |
b128c09f BB |
1669 | uint64_t thisobj; |
1670 | int err; | |
34dc7c2f | 1671 | |
13fe0198 MA |
1672 | dsl_pool_config_enter(dp, FTAG); |
1673 | ||
1674 | err = dsl_dir_hold(dp, name, FTAG, &dd, NULL); | |
1675 | if (err != 0) { | |
1676 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f | 1677 | return (err); |
13fe0198 | 1678 | } |
34dc7c2f | 1679 | |
b128c09f BB |
1680 | /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ |
1681 | if (dd->dd_myname[0] == '$') { | |
13fe0198 MA |
1682 | dsl_dir_rele(dd, FTAG); |
1683 | dsl_pool_config_exit(dp, FTAG); | |
b128c09f BB |
1684 | return (0); |
1685 | } | |
1686 | ||
1687 | thisobj = dd->dd_phys->dd_head_dataset_obj; | |
ed2e1576 | 1688 | attr = kmem_alloc(sizeof (zap_attribute_t), KM_PUSHPAGE); |
34dc7c2f BB |
1689 | |
1690 | /* | |
1691 | * Iterate over all children. | |
1692 | */ | |
1693 | if (flags & DS_FIND_CHILDREN) { | |
b128c09f | 1694 | for (zap_cursor_init(&zc, dp->dp_meta_objset, |
34dc7c2f BB |
1695 | dd->dd_phys->dd_child_dir_zapobj); |
1696 | zap_cursor_retrieve(&zc, attr) == 0; | |
1697 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 MA |
1698 | ASSERT3U(attr->za_integer_length, ==, |
1699 | sizeof (uint64_t)); | |
1700 | ASSERT3U(attr->za_num_integers, ==, 1); | |
34dc7c2f | 1701 | |
428870ff | 1702 | child = kmem_asprintf("%s/%s", name, attr->za_name); |
13fe0198 MA |
1703 | dsl_pool_config_exit(dp, FTAG); |
1704 | err = dmu_objset_find_impl(spa, child, | |
1705 | func, arg, flags); | |
1706 | dsl_pool_config_enter(dp, FTAG); | |
428870ff | 1707 | strfree(child); |
13fe0198 | 1708 | if (err != 0) |
34dc7c2f BB |
1709 | break; |
1710 | } | |
1711 | zap_cursor_fini(&zc); | |
1712 | ||
13fe0198 MA |
1713 | if (err != 0) { |
1714 | dsl_dir_rele(dd, FTAG); | |
1715 | dsl_pool_config_exit(dp, FTAG); | |
34dc7c2f BB |
1716 | kmem_free(attr, sizeof (zap_attribute_t)); |
1717 | return (err); | |
1718 | } | |
1719 | } | |
1720 | ||
1721 | /* | |
1722 | * Iterate over all snapshots. | |
1723 | */ | |
b128c09f | 1724 | if (flags & DS_FIND_SNAPSHOTS) { |
b128c09f | 1725 | err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); |
b128c09f BB |
1726 | |
1727 | if (err == 0) { | |
1728 | uint64_t snapobj = ds->ds_phys->ds_snapnames_zapobj; | |
1729 | dsl_dataset_rele(ds, FTAG); | |
1730 | ||
1731 | for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); | |
1732 | zap_cursor_retrieve(&zc, attr) == 0; | |
1733 | (void) zap_cursor_advance(&zc)) { | |
13fe0198 | 1734 | ASSERT3U(attr->za_integer_length, ==, |
b128c09f | 1735 | sizeof (uint64_t)); |
13fe0198 | 1736 | ASSERT3U(attr->za_num_integers, ==, 1); |
b128c09f | 1737 | |
428870ff BB |
1738 | child = kmem_asprintf("%s@%s", |
1739 | name, attr->za_name); | |
13fe0198 MA |
1740 | dsl_pool_config_exit(dp, FTAG); |
1741 | err = func(child, arg); | |
1742 | dsl_pool_config_enter(dp, FTAG); | |
428870ff | 1743 | strfree(child); |
13fe0198 | 1744 | if (err != 0) |
b128c09f BB |
1745 | break; |
1746 | } | |
1747 | zap_cursor_fini(&zc); | |
34dc7c2f | 1748 | } |
34dc7c2f BB |
1749 | } |
1750 | ||
13fe0198 | 1751 | dsl_dir_rele(dd, FTAG); |
34dc7c2f | 1752 | kmem_free(attr, sizeof (zap_attribute_t)); |
13fe0198 | 1753 | dsl_pool_config_exit(dp, FTAG); |
34dc7c2f | 1754 | |
13fe0198 | 1755 | if (err != 0) |
34dc7c2f BB |
1756 | return (err); |
1757 | ||
13fe0198 MA |
1758 | /* Apply to self. */ |
1759 | return (func(name, arg)); | |
34dc7c2f BB |
1760 | } |
1761 | ||
13fe0198 MA |
1762 | /* |
1763 | * See comment above dmu_objset_find_impl(). | |
1764 | */ | |
d164b209 | 1765 | int |
13fe0198 MA |
1766 | dmu_objset_find(char *name, int func(const char *, void *), void *arg, |
1767 | int flags) | |
d164b209 | 1768 | { |
13fe0198 MA |
1769 | spa_t *spa; |
1770 | int error; | |
d164b209 | 1771 | |
13fe0198 MA |
1772 | error = spa_open(name, &spa, FTAG); |
1773 | if (error != 0) | |
1774 | return (error); | |
1775 | error = dmu_objset_find_impl(spa, name, func, arg, flags); | |
1776 | spa_close(spa, FTAG); | |
1777 | return (error); | |
d164b209 BB |
1778 | } |
1779 | ||
34dc7c2f BB |
1780 | void |
1781 | dmu_objset_set_user(objset_t *os, void *user_ptr) | |
1782 | { | |
428870ff BB |
1783 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
1784 | os->os_user_ptr = user_ptr; | |
34dc7c2f BB |
1785 | } |
1786 | ||
1787 | void * | |
1788 | dmu_objset_get_user(objset_t *os) | |
1789 | { | |
428870ff BB |
1790 | ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); |
1791 | return (os->os_user_ptr); | |
34dc7c2f | 1792 | } |
c28b2279 | 1793 | |
13fe0198 MA |
1794 | /* |
1795 | * Determine name of filesystem, given name of snapshot. | |
1796 | * buf must be at least MAXNAMELEN bytes | |
1797 | */ | |
1798 | int | |
1799 | dmu_fsname(const char *snapname, char *buf) | |
1800 | { | |
1801 | char *atp = strchr(snapname, '@'); | |
1802 | if (atp == NULL) | |
2e528b49 | 1803 | return (SET_ERROR(EINVAL)); |
13fe0198 | 1804 | if (atp - snapname >= MAXNAMELEN) |
2e528b49 | 1805 | return (SET_ERROR(ENAMETOOLONG)); |
13fe0198 MA |
1806 | (void) strlcpy(buf, snapname, atp - snapname + 1); |
1807 | return (0); | |
1808 | } | |
1809 | ||
c28b2279 | 1810 | #if defined(_KERNEL) && defined(HAVE_SPL) |
f0fd83be | 1811 | EXPORT_SYMBOL(dmu_objset_zil); |
c28b2279 | 1812 | EXPORT_SYMBOL(dmu_objset_pool); |
f0fd83be BB |
1813 | EXPORT_SYMBOL(dmu_objset_ds); |
1814 | EXPORT_SYMBOL(dmu_objset_type); | |
c28b2279 BB |
1815 | EXPORT_SYMBOL(dmu_objset_name); |
1816 | EXPORT_SYMBOL(dmu_objset_hold); | |
1817 | EXPORT_SYMBOL(dmu_objset_own); | |
1818 | EXPORT_SYMBOL(dmu_objset_rele); | |
1819 | EXPORT_SYMBOL(dmu_objset_disown); | |
1820 | EXPORT_SYMBOL(dmu_objset_from_ds); | |
1821 | EXPORT_SYMBOL(dmu_objset_create); | |
1822 | EXPORT_SYMBOL(dmu_objset_clone); | |
c28b2279 BB |
1823 | EXPORT_SYMBOL(dmu_objset_stats); |
1824 | EXPORT_SYMBOL(dmu_objset_fast_stat); | |
54a179e7 | 1825 | EXPORT_SYMBOL(dmu_objset_spa); |
c28b2279 BB |
1826 | EXPORT_SYMBOL(dmu_objset_space); |
1827 | EXPORT_SYMBOL(dmu_objset_fsid_guid); | |
1828 | EXPORT_SYMBOL(dmu_objset_find); | |
c28b2279 BB |
1829 | EXPORT_SYMBOL(dmu_objset_byteswap); |
1830 | EXPORT_SYMBOL(dmu_objset_evict_dbufs); | |
1831 | EXPORT_SYMBOL(dmu_objset_snap_cmtime); | |
1832 | ||
1833 | EXPORT_SYMBOL(dmu_objset_sync); | |
1834 | EXPORT_SYMBOL(dmu_objset_is_dirty); | |
1835 | EXPORT_SYMBOL(dmu_objset_create_impl); | |
1836 | EXPORT_SYMBOL(dmu_objset_open_impl); | |
1837 | EXPORT_SYMBOL(dmu_objset_evict); | |
1838 | EXPORT_SYMBOL(dmu_objset_register_type); | |
1839 | EXPORT_SYMBOL(dmu_objset_do_userquota_updates); | |
1840 | EXPORT_SYMBOL(dmu_objset_userquota_get_ids); | |
1841 | EXPORT_SYMBOL(dmu_objset_userused_enabled); | |
1842 | EXPORT_SYMBOL(dmu_objset_userspace_upgrade); | |
1843 | EXPORT_SYMBOL(dmu_objset_userspace_present); | |
1844 | #endif |