]>
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. |
9ae529ec | 23 | * Copyright (c) 2012 by Delphix. All rights reserved. |
37abac6d | 24 | * Copyright (c) 2012, Joyent, Inc. All rights reserved. |
34dc7c2f BB |
25 | */ |
26 | ||
428870ff BB |
27 | /* Portions Copyright 2010 Robert Milkowski */ |
28 | ||
34dc7c2f BB |
29 | #ifndef _SYS_DMU_H |
30 | #define _SYS_DMU_H | |
31 | ||
34dc7c2f BB |
32 | /* |
33 | * This file describes the interface that the DMU provides for its | |
34 | * consumers. | |
35 | * | |
36 | * The DMU also interacts with the SPA. That interface is described in | |
37 | * dmu_spa.h. | |
38 | */ | |
39 | ||
40 | #include <sys/inttypes.h> | |
41 | #include <sys/types.h> | |
42 | #include <sys/param.h> | |
43 | #include <sys/cred.h> | |
428870ff | 44 | #include <sys/time.h> |
59e6e7ca | 45 | #include <sys/uio.h> |
34dc7c2f BB |
46 | |
47 | #ifdef __cplusplus | |
48 | extern "C" { | |
49 | #endif | |
50 | ||
34dc7c2f BB |
51 | struct page; |
52 | struct vnode; | |
53 | struct spa; | |
54 | struct zilog; | |
55 | struct zio; | |
56 | struct blkptr; | |
57 | struct zap_cursor; | |
58 | struct dsl_dataset; | |
59 | struct dsl_pool; | |
60 | struct dnode; | |
61 | struct drr_begin; | |
62 | struct drr_end; | |
63 | struct zbookmark; | |
64 | struct spa; | |
65 | struct nvlist; | |
9babb374 | 66 | struct arc_buf; |
428870ff BB |
67 | struct zio_prop; |
68 | struct sa_handle; | |
34dc7c2f BB |
69 | |
70 | typedef struct objset objset_t; | |
71 | typedef struct dmu_tx dmu_tx_t; | |
72 | typedef struct dsl_dir dsl_dir_t; | |
73 | ||
9ae529ec CS |
74 | typedef enum dmu_object_byteswap { |
75 | DMU_BSWAP_UINT8, | |
76 | DMU_BSWAP_UINT16, | |
77 | DMU_BSWAP_UINT32, | |
78 | DMU_BSWAP_UINT64, | |
79 | DMU_BSWAP_ZAP, | |
80 | DMU_BSWAP_DNODE, | |
81 | DMU_BSWAP_OBJSET, | |
82 | DMU_BSWAP_ZNODE, | |
83 | DMU_BSWAP_OLDACL, | |
84 | DMU_BSWAP_ACL, | |
85 | /* | |
86 | * Allocating a new byteswap type number makes the on-disk format | |
87 | * incompatible with any other format that uses the same number. | |
88 | * | |
89 | * Data can usually be structured to work with one of the | |
90 | * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. | |
91 | */ | |
92 | DMU_BSWAP_NUMFUNCS | |
93 | } dmu_object_byteswap_t; | |
94 | ||
95 | #define DMU_OT_NEWTYPE 0x80 | |
96 | #define DMU_OT_METADATA 0x40 | |
97 | #define DMU_OT_BYTESWAP_MASK 0x3f | |
98 | ||
99 | /* | |
100 | * Defines a uint8_t object type. Object types specify if the data | |
101 | * in the object is metadata (boolean) and how to byteswap the data | |
102 | * (dmu_object_byteswap_t). | |
103 | */ | |
104 | #define DMU_OT(byteswap, metadata) \ | |
105 | (DMU_OT_NEWTYPE | \ | |
106 | ((metadata) ? DMU_OT_METADATA : 0) | \ | |
107 | ((byteswap) & DMU_OT_BYTESWAP_MASK)) | |
108 | ||
109 | #define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \ | |
110 | ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \ | |
111 | (ot) < DMU_OT_NUMTYPES) | |
112 | ||
113 | #define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \ | |
114 | ((ot) & DMU_OT_METADATA) : \ | |
115 | dmu_ot[(int)(ot)].ot_metadata) | |
116 | ||
117 | #define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \ | |
118 | ((ot) & DMU_OT_BYTESWAP_MASK) : \ | |
119 | dmu_ot[(int)(ot)].ot_byteswap) | |
120 | ||
34dc7c2f BB |
121 | typedef enum dmu_object_type { |
122 | DMU_OT_NONE, | |
123 | /* general: */ | |
124 | DMU_OT_OBJECT_DIRECTORY, /* ZAP */ | |
125 | DMU_OT_OBJECT_ARRAY, /* UINT64 */ | |
126 | DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ | |
127 | DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ | |
428870ff BB |
128 | DMU_OT_BPOBJ, /* UINT64 */ |
129 | DMU_OT_BPOBJ_HDR, /* UINT64 */ | |
34dc7c2f BB |
130 | /* spa: */ |
131 | DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ | |
132 | DMU_OT_SPACE_MAP, /* UINT64 */ | |
133 | /* zil: */ | |
134 | DMU_OT_INTENT_LOG, /* UINT64 */ | |
135 | /* dmu: */ | |
136 | DMU_OT_DNODE, /* DNODE */ | |
137 | DMU_OT_OBJSET, /* OBJSET */ | |
138 | /* dsl: */ | |
139 | DMU_OT_DSL_DIR, /* UINT64 */ | |
140 | DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ | |
141 | DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ | |
142 | DMU_OT_DSL_PROPS, /* ZAP */ | |
143 | DMU_OT_DSL_DATASET, /* UINT64 */ | |
144 | /* zpl: */ | |
145 | DMU_OT_ZNODE, /* ZNODE */ | |
146 | DMU_OT_OLDACL, /* Old ACL */ | |
147 | DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ | |
148 | DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ | |
149 | DMU_OT_MASTER_NODE, /* ZAP */ | |
150 | DMU_OT_UNLINKED_SET, /* ZAP */ | |
151 | /* zvol: */ | |
152 | DMU_OT_ZVOL, /* UINT8 */ | |
153 | DMU_OT_ZVOL_PROP, /* ZAP */ | |
154 | /* other; for testing only! */ | |
155 | DMU_OT_PLAIN_OTHER, /* UINT8 */ | |
156 | DMU_OT_UINT64_OTHER, /* UINT64 */ | |
157 | DMU_OT_ZAP_OTHER, /* ZAP */ | |
158 | /* new object types: */ | |
159 | DMU_OT_ERROR_LOG, /* ZAP */ | |
160 | DMU_OT_SPA_HISTORY, /* UINT8 */ | |
161 | DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ | |
162 | DMU_OT_POOL_PROPS, /* ZAP */ | |
163 | DMU_OT_DSL_PERMS, /* ZAP */ | |
164 | DMU_OT_ACL, /* ACL */ | |
165 | DMU_OT_SYSACL, /* SYSACL */ | |
166 | DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ | |
167 | DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ | |
b128c09f | 168 | DMU_OT_NEXT_CLONES, /* ZAP */ |
428870ff | 169 | DMU_OT_SCAN_QUEUE, /* ZAP */ |
9babb374 BB |
170 | DMU_OT_USERGROUP_USED, /* ZAP */ |
171 | DMU_OT_USERGROUP_QUOTA, /* ZAP */ | |
45d1cae3 | 172 | DMU_OT_USERREFS, /* ZAP */ |
428870ff BB |
173 | DMU_OT_DDT_ZAP, /* ZAP */ |
174 | DMU_OT_DDT_STATS, /* ZAP */ | |
175 | DMU_OT_SA, /* System attr */ | |
176 | DMU_OT_SA_MASTER_NODE, /* ZAP */ | |
177 | DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ | |
178 | DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ | |
179 | DMU_OT_SCAN_XLATE, /* ZAP */ | |
180 | DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ | |
181 | DMU_OT_DEADLIST, /* ZAP */ | |
182 | DMU_OT_DEADLIST_HDR, /* UINT64 */ | |
183 | DMU_OT_DSL_CLONES, /* ZAP */ | |
184 | DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */ | |
9ae529ec CS |
185 | /* |
186 | * Do not allocate new object types here. Doing so makes the on-disk | |
187 | * format incompatible with any other format that uses the same object | |
188 | * type number. | |
189 | * | |
190 | * When creating an object which does not have one of the above types | |
191 | * use the DMU_OTN_* type with the correct byteswap and metadata | |
192 | * values. | |
193 | * | |
194 | * The DMU_OTN_* types do not have entries in the dmu_ot table, | |
195 | * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead | |
196 | * of indexing into dmu_ot directly (this works for both DMU_OT_* types | |
197 | * and DMU_OTN_* types). | |
198 | */ | |
199 | DMU_OT_NUMTYPES, | |
200 | ||
201 | /* | |
202 | * Names for valid types declared with DMU_OT(). | |
203 | */ | |
204 | DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE), | |
205 | DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE), | |
206 | DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE), | |
207 | DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE), | |
208 | DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE), | |
209 | DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE), | |
210 | DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE), | |
211 | DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE), | |
212 | DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE), | |
213 | DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE), | |
34dc7c2f BB |
214 | } dmu_object_type_t; |
215 | ||
216 | typedef enum dmu_objset_type { | |
217 | DMU_OST_NONE, | |
218 | DMU_OST_META, | |
219 | DMU_OST_ZFS, | |
220 | DMU_OST_ZVOL, | |
221 | DMU_OST_OTHER, /* For testing only! */ | |
222 | DMU_OST_ANY, /* Be careful! */ | |
223 | DMU_OST_NUMTYPES | |
224 | } dmu_objset_type_t; | |
225 | ||
226 | void byteswap_uint64_array(void *buf, size_t size); | |
227 | void byteswap_uint32_array(void *buf, size_t size); | |
228 | void byteswap_uint16_array(void *buf, size_t size); | |
229 | void byteswap_uint8_array(void *buf, size_t size); | |
230 | void zap_byteswap(void *buf, size_t size); | |
231 | void zfs_oldacl_byteswap(void *buf, size_t size); | |
232 | void zfs_acl_byteswap(void *buf, size_t size); | |
233 | void zfs_znode_byteswap(void *buf, size_t size); | |
234 | ||
34dc7c2f BB |
235 | #define DS_FIND_SNAPSHOTS (1<<0) |
236 | #define DS_FIND_CHILDREN (1<<1) | |
237 | ||
238 | /* | |
239 | * The maximum number of bytes that can be accessed as part of one | |
240 | * operation, including metadata. | |
241 | */ | |
242 | #define DMU_MAX_ACCESS (10<<20) /* 10MB */ | |
b128c09f | 243 | #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */ |
34dc7c2f | 244 | |
9babb374 BB |
245 | #define DMU_USERUSED_OBJECT (-1ULL) |
246 | #define DMU_GROUPUSED_OBJECT (-2ULL) | |
428870ff | 247 | #define DMU_DEADLIST_OBJECT (-3ULL) |
9babb374 | 248 | |
428870ff BB |
249 | /* |
250 | * artificial blkids for bonus buffer and spill blocks | |
251 | */ | |
252 | #define DMU_BONUS_BLKID (-1ULL) | |
253 | #define DMU_SPILL_BLKID (-2ULL) | |
34dc7c2f BB |
254 | /* |
255 | * Public routines to create, destroy, open, and close objsets. | |
256 | */ | |
428870ff BB |
257 | int dmu_objset_hold(const char *name, void *tag, objset_t **osp); |
258 | int dmu_objset_own(const char *name, dmu_objset_type_t type, | |
259 | boolean_t readonly, void *tag, objset_t **osp); | |
260 | void dmu_objset_rele(objset_t *os, void *tag); | |
261 | void dmu_objset_disown(objset_t *os, void *tag); | |
262 | int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp); | |
263 | ||
34dc7c2f | 264 | int dmu_objset_evict_dbufs(objset_t *os); |
428870ff | 265 | int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, |
34dc7c2f | 266 | void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg); |
428870ff BB |
267 | int dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin, |
268 | uint64_t flags); | |
45d1cae3 | 269 | int dmu_objset_destroy(const char *name, boolean_t defer); |
330d06f9 | 270 | int dmu_snapshots_destroy_nvl(struct nvlist *snaps, boolean_t defer, char *); |
572e2857 BB |
271 | int dmu_objset_snapshot(char *fsname, char *snapname, char *tag, |
272 | struct nvlist *props, boolean_t recursive, boolean_t temporary, int fd); | |
34dc7c2f BB |
273 | int dmu_objset_rename(const char *name, const char *newname, |
274 | boolean_t recursive); | |
428870ff | 275 | int dmu_objset_find(char *name, int func(const char *, void *), void *arg, |
34dc7c2f BB |
276 | int flags); |
277 | void dmu_objset_byteswap(void *buf, size_t size); | |
278 | ||
279 | typedef struct dmu_buf { | |
280 | uint64_t db_object; /* object that this buffer is part of */ | |
281 | uint64_t db_offset; /* byte offset in this object */ | |
282 | uint64_t db_size; /* size of buffer in bytes */ | |
283 | void *db_data; /* data in buffer */ | |
284 | } dmu_buf_t; | |
285 | ||
286 | typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr); | |
287 | ||
288 | /* | |
289 | * The names of zap entries in the DIRECTORY_OBJECT of the MOS. | |
290 | */ | |
291 | #define DMU_POOL_DIRECTORY_OBJECT 1 | |
292 | #define DMU_POOL_CONFIG "config" | |
9ae529ec CS |
293 | #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write" |
294 | #define DMU_POOL_FEATURES_FOR_READ "features_for_read" | |
295 | #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions" | |
34dc7c2f | 296 | #define DMU_POOL_ROOT_DATASET "root_dataset" |
428870ff | 297 | #define DMU_POOL_SYNC_BPOBJ "sync_bplist" |
34dc7c2f BB |
298 | #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" |
299 | #define DMU_POOL_ERRLOG_LAST "errlog_last" | |
300 | #define DMU_POOL_SPARES "spares" | |
301 | #define DMU_POOL_DEFLATE "deflate" | |
302 | #define DMU_POOL_HISTORY "history" | |
303 | #define DMU_POOL_PROPS "pool_props" | |
304 | #define DMU_POOL_L2CACHE "l2cache" | |
428870ff BB |
305 | #define DMU_POOL_TMP_USERREFS "tmp_userrefs" |
306 | #define DMU_POOL_DDT "DDT-%s-%s-%s" | |
307 | #define DMU_POOL_DDT_STATS "DDT-statistics" | |
308 | #define DMU_POOL_CREATION_VERSION "creation_version" | |
309 | #define DMU_POOL_SCAN "scan" | |
310 | #define DMU_POOL_FREE_BPOBJ "free_bpobj" | |
9ae529ec | 311 | #define DMU_POOL_BPTREE_OBJ "bptree_obj" |
753c3839 | 312 | #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj" |
b128c09f | 313 | |
34dc7c2f BB |
314 | /* |
315 | * Allocate an object from this objset. The range of object numbers | |
316 | * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode. | |
317 | * | |
318 | * The transaction must be assigned to a txg. The newly allocated | |
319 | * object will be "held" in the transaction (ie. you can modify the | |
320 | * newly allocated object in this transaction). | |
321 | * | |
322 | * dmu_object_alloc() chooses an object and returns it in *objectp. | |
323 | * | |
324 | * dmu_object_claim() allocates a specific object number. If that | |
325 | * number is already allocated, it fails and returns EEXIST. | |
326 | * | |
327 | * Return 0 on success, or ENOSPC or EEXIST as specified above. | |
328 | */ | |
329 | uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot, | |
330 | int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); | |
331 | int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot, | |
332 | int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); | |
333 | int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot, | |
9babb374 | 334 | int blocksize, dmu_object_type_t bonustype, int bonuslen); |
34dc7c2f BB |
335 | |
336 | /* | |
337 | * Free an object from this objset. | |
338 | * | |
339 | * The object's data will be freed as well (ie. you don't need to call | |
340 | * dmu_free(object, 0, -1, tx)). | |
341 | * | |
342 | * The object need not be held in the transaction. | |
343 | * | |
344 | * If there are any holds on this object's buffers (via dmu_buf_hold()), | |
345 | * or tx holds on the object (via dmu_tx_hold_object()), you can not | |
346 | * free it; it fails and returns EBUSY. | |
347 | * | |
348 | * If the object is not allocated, it fails and returns ENOENT. | |
349 | * | |
350 | * Return 0 on success, or EBUSY or ENOENT as specified above. | |
351 | */ | |
352 | int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx); | |
353 | ||
354 | /* | |
355 | * Find the next allocated or free object. | |
356 | * | |
357 | * The objectp parameter is in-out. It will be updated to be the next | |
358 | * object which is allocated. Ignore objects which have not been | |
359 | * modified since txg. | |
360 | * | |
361 | * XXX Can only be called on a objset with no dirty data. | |
362 | * | |
363 | * Returns 0 on success, or ENOENT if there are no more objects. | |
364 | */ | |
365 | int dmu_object_next(objset_t *os, uint64_t *objectp, | |
366 | boolean_t hole, uint64_t txg); | |
367 | ||
368 | /* | |
369 | * Set the data blocksize for an object. | |
370 | * | |
371 | * The object cannot have any blocks allcated beyond the first. If | |
372 | * the first block is allocated already, the new size must be greater | |
373 | * than the current block size. If these conditions are not met, | |
374 | * ENOTSUP will be returned. | |
375 | * | |
376 | * Returns 0 on success, or EBUSY if there are any holds on the object | |
377 | * contents, or ENOTSUP as described above. | |
378 | */ | |
379 | int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, | |
380 | int ibs, dmu_tx_t *tx); | |
381 | ||
382 | /* | |
383 | * Set the checksum property on a dnode. The new checksum algorithm will | |
384 | * apply to all newly written blocks; existing blocks will not be affected. | |
385 | */ | |
386 | void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, | |
387 | dmu_tx_t *tx); | |
388 | ||
389 | /* | |
390 | * Set the compress property on a dnode. The new compression algorithm will | |
391 | * apply to all newly written blocks; existing blocks will not be affected. | |
392 | */ | |
393 | void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, | |
394 | dmu_tx_t *tx); | |
395 | ||
396 | /* | |
428870ff | 397 | * Decide how to write a block: checksum, compression, number of copies, etc. |
34dc7c2f | 398 | */ |
428870ff BB |
399 | #define WP_NOFILL 0x1 |
400 | #define WP_DMU_SYNC 0x2 | |
401 | #define WP_SPILL 0x4 | |
402 | ||
403 | void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp, | |
404 | struct zio_prop *zp); | |
34dc7c2f BB |
405 | /* |
406 | * The bonus data is accessed more or less like a regular buffer. | |
407 | * You must dmu_bonus_hold() to get the buffer, which will give you a | |
408 | * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus | |
409 | * data. As with any normal buffer, you must call dmu_buf_read() to | |
410 | * read db_data, dmu_buf_will_dirty() before modifying it, and the | |
411 | * object must be held in an assigned transaction before calling | |
412 | * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus | |
413 | * buffer as well. You must release your hold with dmu_buf_rele(). | |
414 | */ | |
415 | int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **); | |
416 | int dmu_bonus_max(void); | |
417 | int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *); | |
428870ff | 418 | int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *); |
572e2857 | 419 | dmu_object_type_t dmu_get_bonustype(dmu_buf_t *); |
428870ff BB |
420 | int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *); |
421 | ||
422 | /* | |
423 | * Special spill buffer support used by "SA" framework | |
424 | */ | |
425 | ||
426 | int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); | |
427 | int dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags, | |
428 | void *tag, dmu_buf_t **dbp); | |
429 | int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); | |
34dc7c2f BB |
430 | |
431 | /* | |
432 | * Obtain the DMU buffer from the specified object which contains the | |
433 | * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so | |
434 | * that it will remain in memory. You must release the hold with | |
435 | * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your | |
436 | * hold. You must have a hold on any dmu_buf_t* you pass to the DMU. | |
437 | * | |
438 | * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill | |
439 | * on the returned buffer before reading or writing the buffer's | |
440 | * db_data. The comments for those routines describe what particular | |
441 | * operations are valid after calling them. | |
442 | * | |
443 | * The object number must be a valid, allocated object number. | |
444 | */ | |
445 | int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, | |
428870ff | 446 | void *tag, dmu_buf_t **, int flags); |
34dc7c2f BB |
447 | void dmu_buf_add_ref(dmu_buf_t *db, void* tag); |
448 | void dmu_buf_rele(dmu_buf_t *db, void *tag); | |
449 | uint64_t dmu_buf_refcount(dmu_buf_t *db); | |
450 | ||
451 | /* | |
452 | * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a | |
453 | * range of an object. A pointer to an array of dmu_buf_t*'s is | |
454 | * returned (in *dbpp). | |
455 | * | |
456 | * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and | |
457 | * frees the array. The hold on the array of buffers MUST be released | |
458 | * with dmu_buf_rele_array. You can NOT release the hold on each buffer | |
459 | * individually with dmu_buf_rele. | |
460 | */ | |
461 | int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset, | |
462 | uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp); | |
463 | void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag); | |
464 | ||
465 | /* | |
466 | * Returns NULL on success, or the existing user ptr if it's already | |
467 | * been set. | |
468 | * | |
469 | * user_ptr is for use by the user and can be obtained via dmu_buf_get_user(). | |
470 | * | |
471 | * user_data_ptr_ptr should be NULL, or a pointer to a pointer which | |
472 | * will be set to db->db_data when you are allowed to access it. Note | |
473 | * that db->db_data (the pointer) can change when you do dmu_buf_read(), | |
474 | * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill(). | |
475 | * *user_data_ptr_ptr will be set to the new value when it changes. | |
476 | * | |
477 | * If non-NULL, pageout func will be called when this buffer is being | |
478 | * excised from the cache, so that you can clean up the data structure | |
479 | * pointed to by user_ptr. | |
480 | * | |
481 | * dmu_evict_user() will call the pageout func for all buffers in a | |
482 | * objset with a given pageout func. | |
483 | */ | |
484 | void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr, | |
485 | dmu_buf_evict_func_t *pageout_func); | |
486 | /* | |
487 | * set_user_ie is the same as set_user, but request immediate eviction | |
488 | * when hold count goes to zero. | |
489 | */ | |
490 | void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr, | |
491 | void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func); | |
492 | void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr, | |
493 | void *user_ptr, void *user_data_ptr_ptr, | |
494 | dmu_buf_evict_func_t *pageout_func); | |
495 | void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func); | |
496 | ||
497 | /* | |
498 | * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set. | |
499 | */ | |
500 | void *dmu_buf_get_user(dmu_buf_t *db); | |
501 | ||
502 | /* | |
503 | * Indicate that you are going to modify the buffer's data (db_data). | |
504 | * | |
505 | * The transaction (tx) must be assigned to a txg (ie. you've called | |
506 | * dmu_tx_assign()). The buffer's object must be held in the tx | |
507 | * (ie. you've called dmu_tx_hold_object(tx, db->db_object)). | |
508 | */ | |
509 | void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx); | |
510 | ||
9babb374 BB |
511 | /* |
512 | * Tells if the given dbuf is freeable. | |
513 | */ | |
514 | boolean_t dmu_buf_freeable(dmu_buf_t *); | |
515 | ||
34dc7c2f BB |
516 | /* |
517 | * You must create a transaction, then hold the objects which you will | |
518 | * (or might) modify as part of this transaction. Then you must assign | |
519 | * the transaction to a transaction group. Once the transaction has | |
520 | * been assigned, you can modify buffers which belong to held objects as | |
521 | * part of this transaction. You can't modify buffers before the | |
522 | * transaction has been assigned; you can't modify buffers which don't | |
523 | * belong to objects which this transaction holds; you can't hold | |
524 | * objects once the transaction has been assigned. You may hold an | |
525 | * object which you are going to free (with dmu_object_free()), but you | |
526 | * don't have to. | |
527 | * | |
528 | * You can abort the transaction before it has been assigned. | |
529 | * | |
530 | * Note that you may hold buffers (with dmu_buf_hold) at any time, | |
531 | * regardless of transaction state. | |
532 | */ | |
533 | ||
534 | #define DMU_NEW_OBJECT (-1ULL) | |
535 | #define DMU_OBJECT_END (-1ULL) | |
536 | ||
537 | dmu_tx_t *dmu_tx_create(objset_t *os); | |
538 | void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len); | |
539 | void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, | |
540 | uint64_t len); | |
9babb374 | 541 | void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name); |
34dc7c2f | 542 | void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object); |
428870ff BB |
543 | void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object); |
544 | void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow); | |
545 | void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size); | |
34dc7c2f BB |
546 | void dmu_tx_abort(dmu_tx_t *tx); |
547 | int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how); | |
548 | void dmu_tx_wait(dmu_tx_t *tx); | |
549 | void dmu_tx_commit(dmu_tx_t *tx); | |
550 | ||
428870ff BB |
551 | /* |
552 | * To register a commit callback, dmu_tx_callback_register() must be called. | |
553 | * | |
554 | * dcb_data is a pointer to caller private data that is passed on as a | |
555 | * callback parameter. The caller is responsible for properly allocating and | |
556 | * freeing it. | |
557 | * | |
558 | * When registering a callback, the transaction must be already created, but | |
559 | * it cannot be committed or aborted. It can be assigned to a txg or not. | |
560 | * | |
561 | * The callback will be called after the transaction has been safely written | |
562 | * to stable storage and will also be called if the dmu_tx is aborted. | |
563 | * If there is any error which prevents the transaction from being committed to | |
564 | * disk, the callback will be called with a value of error != 0. | |
565 | */ | |
566 | typedef void dmu_tx_callback_func_t(void *dcb_data, int error); | |
567 | ||
568 | void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func, | |
569 | void *dcb_data); | |
570 | ||
34dc7c2f BB |
571 | /* |
572 | * Free up the data blocks for a defined range of a file. If size is | |
9ae529ec | 573 | * -1, the range from offset to end-of-file is freed. |
34dc7c2f BB |
574 | */ |
575 | int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, | |
576 | uint64_t size, dmu_tx_t *tx); | |
b128c09f BB |
577 | int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, |
578 | uint64_t size); | |
579 | int dmu_free_object(objset_t *os, uint64_t object); | |
34dc7c2f BB |
580 | |
581 | /* | |
582 | * Convenience functions. | |
583 | * | |
584 | * Canfail routines will return 0 on success, or an errno if there is a | |
585 | * nonrecoverable I/O error. | |
586 | */ | |
9babb374 BB |
587 | #define DMU_READ_PREFETCH 0 /* prefetch */ |
588 | #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */ | |
34dc7c2f | 589 | int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, |
9babb374 | 590 | void *buf, uint32_t flags); |
34dc7c2f BB |
591 | void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, |
592 | const void *buf, dmu_tx_t *tx); | |
b128c09f BB |
593 | void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, |
594 | dmu_tx_t *tx); | |
60101509 | 595 | #ifdef _KERNEL |
61e90960 | 596 | #include <linux/blkdev_compat.h> |
60101509 | 597 | int dmu_read_req(objset_t *os, uint64_t object, struct request *req); |
872e8d26 BB |
598 | int dmu_write_req(objset_t *os, uint64_t object, struct request *req, |
599 | dmu_tx_t *tx); | |
600 | int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size); | |
601 | int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size, | |
602 | dmu_tx_t *tx); | |
603 | int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size, | |
604 | dmu_tx_t *tx); | |
872e8d26 | 605 | #endif |
9babb374 BB |
606 | struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size); |
607 | void dmu_return_arcbuf(struct arc_buf *buf); | |
608 | void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf, | |
609 | dmu_tx_t *tx); | |
428870ff BB |
610 | int dmu_xuio_init(struct xuio *uio, int niov); |
611 | void dmu_xuio_fini(struct xuio *uio); | |
612 | int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off, | |
613 | size_t n); | |
614 | int dmu_xuio_cnt(struct xuio *uio); | |
615 | struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i); | |
616 | void dmu_xuio_clear(struct xuio *uio, int i); | |
0bc8fd78 BB |
617 | void xuio_stat_wbuf_copied(void); |
618 | void xuio_stat_wbuf_nocopy(void); | |
34dc7c2f BB |
619 | |
620 | extern int zfs_prefetch_disable; | |
621 | ||
622 | /* | |
623 | * Asynchronously try to read in the data. | |
624 | */ | |
625 | void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset, | |
626 | uint64_t len); | |
627 | ||
628 | typedef struct dmu_object_info { | |
428870ff | 629 | /* All sizes are in bytes unless otherwise indicated. */ |
34dc7c2f BB |
630 | uint32_t doi_data_block_size; |
631 | uint32_t doi_metadata_block_size; | |
34dc7c2f BB |
632 | dmu_object_type_t doi_type; |
633 | dmu_object_type_t doi_bonus_type; | |
428870ff | 634 | uint64_t doi_bonus_size; |
34dc7c2f BB |
635 | uint8_t doi_indirection; /* 2 = dnode->indirect->data */ |
636 | uint8_t doi_checksum; | |
637 | uint8_t doi_compress; | |
638 | uint8_t doi_pad[5]; | |
428870ff BB |
639 | uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */ |
640 | uint64_t doi_max_offset; | |
641 | uint64_t doi_fill_count; /* number of non-empty blocks */ | |
34dc7c2f BB |
642 | } dmu_object_info_t; |
643 | ||
644 | typedef void arc_byteswap_func_t(void *buf, size_t size); | |
645 | ||
646 | typedef struct dmu_object_type_info { | |
9ae529ec | 647 | dmu_object_byteswap_t ot_byteswap; |
34dc7c2f BB |
648 | boolean_t ot_metadata; |
649 | char *ot_name; | |
650 | } dmu_object_type_info_t; | |
651 | ||
9ae529ec CS |
652 | typedef struct dmu_object_byteswap_info { |
653 | arc_byteswap_func_t *ob_func; | |
654 | char *ob_name; | |
655 | } dmu_object_byteswap_info_t; | |
656 | ||
34dc7c2f | 657 | extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES]; |
9ae529ec | 658 | extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS]; |
34dc7c2f BB |
659 | |
660 | /* | |
661 | * Get information on a DMU object. | |
662 | * | |
663 | * Return 0 on success or ENOENT if object is not allocated. | |
664 | * | |
665 | * If doi is NULL, just indicates whether the object exists. | |
666 | */ | |
667 | int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi); | |
668 | void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi); | |
669 | void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi); | |
670 | void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, | |
671 | u_longlong_t *nblk512); | |
672 | ||
673 | typedef struct dmu_objset_stats { | |
674 | uint64_t dds_num_clones; /* number of clones of this */ | |
675 | uint64_t dds_creation_txg; | |
676 | uint64_t dds_guid; | |
677 | dmu_objset_type_t dds_type; | |
678 | uint8_t dds_is_snapshot; | |
679 | uint8_t dds_inconsistent; | |
680 | char dds_origin[MAXNAMELEN]; | |
681 | } dmu_objset_stats_t; | |
682 | ||
683 | /* | |
684 | * Get stats on a dataset. | |
685 | */ | |
686 | void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat); | |
687 | ||
688 | /* | |
689 | * Add entries to the nvlist for all the objset's properties. See | |
690 | * zfs_prop_table[] and zfs(1m) for details on the properties. | |
691 | */ | |
692 | void dmu_objset_stats(objset_t *os, struct nvlist *nv); | |
693 | ||
694 | /* | |
695 | * Get the space usage statistics for statvfs(). | |
696 | * | |
697 | * refdbytes is the amount of space "referenced" by this objset. | |
698 | * availbytes is the amount of space available to this objset, taking | |
699 | * into account quotas & reservations, assuming that no other objsets | |
700 | * use the space first. These values correspond to the 'referenced' and | |
701 | * 'available' properties, described in the zfs(1m) manpage. | |
702 | * | |
703 | * usedobjs and availobjs are the number of objects currently allocated, | |
704 | * and available. | |
705 | */ | |
706 | void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, | |
707 | uint64_t *usedobjsp, uint64_t *availobjsp); | |
708 | ||
709 | /* | |
710 | * The fsid_guid is a 56-bit ID that can change to avoid collisions. | |
711 | * (Contrast with the ds_guid which is a 64-bit ID that will never | |
712 | * change, so there is a small probability that it will collide.) | |
713 | */ | |
714 | uint64_t dmu_objset_fsid_guid(objset_t *os); | |
715 | ||
428870ff BB |
716 | /* |
717 | * Get the [cm]time for an objset's snapshot dir | |
718 | */ | |
719 | timestruc_t dmu_objset_snap_cmtime(objset_t *os); | |
720 | ||
34dc7c2f BB |
721 | int dmu_objset_is_snapshot(objset_t *os); |
722 | ||
723 | extern struct spa *dmu_objset_spa(objset_t *os); | |
724 | extern struct zilog *dmu_objset_zil(objset_t *os); | |
725 | extern struct dsl_pool *dmu_objset_pool(objset_t *os); | |
726 | extern struct dsl_dataset *dmu_objset_ds(objset_t *os); | |
727 | extern void dmu_objset_name(objset_t *os, char *buf); | |
728 | extern dmu_objset_type_t dmu_objset_type(objset_t *os); | |
729 | extern uint64_t dmu_objset_id(objset_t *os); | |
428870ff BB |
730 | extern uint64_t dmu_objset_syncprop(objset_t *os); |
731 | extern uint64_t dmu_objset_logbias(objset_t *os); | |
34dc7c2f BB |
732 | extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name, |
733 | uint64_t *id, uint64_t *offp, boolean_t *case_conflict); | |
6772fb67 | 734 | extern int dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *val); |
34dc7c2f BB |
735 | extern int dmu_snapshot_realname(objset_t *os, char *name, char *real, |
736 | int maxlen, boolean_t *conflict); | |
737 | extern int dmu_dir_list_next(objset_t *os, int namelen, char *name, | |
738 | uint64_t *idp, uint64_t *offp); | |
9babb374 | 739 | |
428870ff BB |
740 | typedef int objset_used_cb_t(dmu_object_type_t bonustype, |
741 | void *bonus, uint64_t *userp, uint64_t *groupp); | |
9babb374 BB |
742 | extern void dmu_objset_register_type(dmu_objset_type_t ost, |
743 | objset_used_cb_t *cb); | |
34dc7c2f BB |
744 | extern void dmu_objset_set_user(objset_t *os, void *user_ptr); |
745 | extern void *dmu_objset_get_user(objset_t *os); | |
746 | ||
747 | /* | |
748 | * Return the txg number for the given assigned transaction. | |
749 | */ | |
750 | uint64_t dmu_tx_get_txg(dmu_tx_t *tx); | |
751 | ||
752 | /* | |
753 | * Synchronous write. | |
754 | * If a parent zio is provided this function initiates a write on the | |
755 | * provided buffer as a child of the parent zio. | |
756 | * In the absence of a parent zio, the write is completed synchronously. | |
757 | * At write completion, blk is filled with the bp of the written block. | |
758 | * Note that while the data covered by this function will be on stable | |
759 | * storage when the write completes this new data does not become a | |
760 | * permanent part of the file until the associated transaction commits. | |
761 | */ | |
428870ff BB |
762 | |
763 | /* | |
764 | * {zfs,zvol,ztest}_get_done() args | |
765 | */ | |
766 | typedef struct zgd { | |
767 | struct zilog *zgd_zilog; | |
768 | struct blkptr *zgd_bp; | |
769 | dmu_buf_t *zgd_db; | |
770 | struct rl *zgd_rl; | |
771 | void *zgd_private; | |
772 | } zgd_t; | |
773 | ||
774 | typedef void dmu_sync_cb_t(zgd_t *arg, int error); | |
775 | int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd); | |
34dc7c2f BB |
776 | |
777 | /* | |
778 | * Find the next hole or data block in file starting at *off | |
779 | * Return found offset in *off. Return ESRCH for end of file. | |
780 | */ | |
781 | int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, | |
782 | uint64_t *off); | |
783 | ||
784 | /* | |
785 | * Initial setup and final teardown. | |
786 | */ | |
787 | extern void dmu_init(void); | |
788 | extern void dmu_fini(void); | |
789 | ||
790 | typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp, | |
791 | uint64_t object, uint64_t offset, int len); | |
792 | void dmu_traverse_objset(objset_t *os, uint64_t txg_start, | |
793 | dmu_traverse_cb_t cb, void *arg); | |
794 | ||
37abac6d BP |
795 | int dmu_send(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin, |
796 | int outfd, struct vnode *vp, offset_t *off); | |
330d06f9 MA |
797 | int dmu_send_estimate(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorign, |
798 | uint64_t *sizep); | |
34dc7c2f BB |
799 | |
800 | typedef struct dmu_recv_cookie { | |
801 | /* | |
802 | * This structure is opaque! | |
803 | * | |
804 | * If logical and real are different, we are recving the stream | |
805 | * into the "real" temporary clone, and then switching it with | |
806 | * the "logical" target. | |
807 | */ | |
808 | struct dsl_dataset *drc_logical_ds; | |
809 | struct dsl_dataset *drc_real_ds; | |
810 | struct drr_begin *drc_drrb; | |
811 | char *drc_tosnap; | |
428870ff | 812 | char *drc_top_ds; |
34dc7c2f BB |
813 | boolean_t drc_newfs; |
814 | boolean_t drc_force; | |
8d35c149 | 815 | struct avl_tree *drc_guid_to_ds_map; |
34dc7c2f BB |
816 | } dmu_recv_cookie_t; |
817 | ||
428870ff | 818 | int dmu_recv_begin(char *tofs, char *tosnap, char *topds, struct drr_begin *, |
45d1cae3 | 819 | boolean_t force, objset_t *origin, dmu_recv_cookie_t *); |
572e2857 BB |
820 | int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp, |
821 | int cleanup_fd, uint64_t *action_handlep); | |
34dc7c2f | 822 | int dmu_recv_end(dmu_recv_cookie_t *drc); |
34dc7c2f | 823 | |
572e2857 BB |
824 | int dmu_diff(objset_t *tosnap, objset_t *fromsnap, struct vnode *vp, |
825 | offset_t *off); | |
826 | ||
34dc7c2f BB |
827 | /* CRC64 table */ |
828 | #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ | |
829 | extern uint64_t zfs_crc64_table[256]; | |
830 | ||
831 | #ifdef __cplusplus | |
832 | } | |
833 | #endif | |
834 | ||
835 | #endif /* _SYS_DMU_H */ |