Print this page
2882 implement libzfs_core
2883 changing "canmount" property to "on" should not always remount dataset
2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Chris Siden <christopher.siden@delphix.com>
Reviewed by: Garrett D'Amore <garrett@damore.org>
Reviewed by: Bill Pijewski <wdp@joyent.com>
Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com>
Split |
Close |
Expand all |
Collapse all |
--- old/usr/src/uts/common/fs/zfs/sys/dmu.h
+++ new/usr/src/uts/common/fs/zfs/sys/dmu.h
1 1 /*
2 2 * CDDL HEADER START
3 3 *
4 4 * The contents of this file are subject to the terms of the
5 5 * Common Development and Distribution License (the "License").
6 6 * You may not use this file except in compliance with the License.
7 7 *
8 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 9 * or http://www.opensolaris.org/os/licensing.
10 10 * See the License for the specific language governing permissions
11 11 * and limitations under the License.
12 12 *
13 13 * When distributing Covered Code, include this CDDL HEADER in each
14 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 15 * If applicable, add the following below this CDDL HEADER, with the
16 16 * fields enclosed by brackets "[]" replaced with your own identifying
17 17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 18 *
19 19 * CDDL HEADER END
20 20 */
21 21
22 22 /*
23 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 24 * Copyright (c) 2012 by Delphix. All rights reserved.
25 25 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
26 26 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
27 27 */
28 28
29 29 /* Portions Copyright 2010 Robert Milkowski */
30 30
31 31 #ifndef _SYS_DMU_H
32 32 #define _SYS_DMU_H
33 33
34 34 /*
35 35 * This file describes the interface that the DMU provides for its
36 36 * consumers.
↓ open down ↓ |
36 lines elided |
↑ open up ↑ |
37 37 *
38 38 * The DMU also interacts with the SPA. That interface is described in
39 39 * dmu_spa.h.
40 40 */
41 41
42 42 #include <sys/inttypes.h>
43 43 #include <sys/types.h>
44 44 #include <sys/param.h>
45 45 #include <sys/cred.h>
46 46 #include <sys/time.h>
47 +#include <sys/fs/zfs.h>
47 48
48 49 #ifdef __cplusplus
49 50 extern "C" {
50 51 #endif
51 52
52 53 struct uio;
53 54 struct xuio;
54 55 struct page;
55 56 struct vnode;
56 57 struct spa;
57 58 struct zilog;
58 59 struct zio;
59 60 struct blkptr;
60 61 struct zap_cursor;
61 62 struct dsl_dataset;
62 63 struct dsl_pool;
63 64 struct dnode;
64 65 struct drr_begin;
65 66 struct drr_end;
66 67 struct zbookmark;
67 68 struct spa;
68 69 struct nvlist;
69 70 struct arc_buf;
70 71 struct zio_prop;
71 72 struct sa_handle;
72 73
73 74 typedef struct objset objset_t;
74 75 typedef struct dmu_tx dmu_tx_t;
75 76 typedef struct dsl_dir dsl_dir_t;
76 77
77 78 typedef enum dmu_object_byteswap {
78 79 DMU_BSWAP_UINT8,
79 80 DMU_BSWAP_UINT16,
80 81 DMU_BSWAP_UINT32,
81 82 DMU_BSWAP_UINT64,
82 83 DMU_BSWAP_ZAP,
83 84 DMU_BSWAP_DNODE,
84 85 DMU_BSWAP_OBJSET,
85 86 DMU_BSWAP_ZNODE,
86 87 DMU_BSWAP_OLDACL,
87 88 DMU_BSWAP_ACL,
88 89 /*
89 90 * Allocating a new byteswap type number makes the on-disk format
90 91 * incompatible with any other format that uses the same number.
91 92 *
92 93 * Data can usually be structured to work with one of the
93 94 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
94 95 */
95 96 DMU_BSWAP_NUMFUNCS
96 97 } dmu_object_byteswap_t;
97 98
98 99 #define DMU_OT_NEWTYPE 0x80
99 100 #define DMU_OT_METADATA 0x40
100 101 #define DMU_OT_BYTESWAP_MASK 0x3f
101 102
102 103 /*
103 104 * Defines a uint8_t object type. Object types specify if the data
104 105 * in the object is metadata (boolean) and how to byteswap the data
105 106 * (dmu_object_byteswap_t).
106 107 */
107 108 #define DMU_OT(byteswap, metadata) \
108 109 (DMU_OT_NEWTYPE | \
109 110 ((metadata) ? DMU_OT_METADATA : 0) | \
110 111 ((byteswap) & DMU_OT_BYTESWAP_MASK))
111 112
112 113 #define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \
113 114 ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \
114 115 (ot) < DMU_OT_NUMTYPES)
115 116
116 117 #define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \
117 118 ((ot) & DMU_OT_METADATA) : \
118 119 dmu_ot[(ot)].ot_metadata)
119 120
120 121 #define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \
121 122 ((ot) & DMU_OT_BYTESWAP_MASK) : \
122 123 dmu_ot[(ot)].ot_byteswap)
123 124
124 125 typedef enum dmu_object_type {
125 126 DMU_OT_NONE,
126 127 /* general: */
127 128 DMU_OT_OBJECT_DIRECTORY, /* ZAP */
128 129 DMU_OT_OBJECT_ARRAY, /* UINT64 */
129 130 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */
130 131 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */
131 132 DMU_OT_BPOBJ, /* UINT64 */
132 133 DMU_OT_BPOBJ_HDR, /* UINT64 */
133 134 /* spa: */
134 135 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */
135 136 DMU_OT_SPACE_MAP, /* UINT64 */
136 137 /* zil: */
137 138 DMU_OT_INTENT_LOG, /* UINT64 */
138 139 /* dmu: */
139 140 DMU_OT_DNODE, /* DNODE */
140 141 DMU_OT_OBJSET, /* OBJSET */
141 142 /* dsl: */
142 143 DMU_OT_DSL_DIR, /* UINT64 */
143 144 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */
144 145 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */
145 146 DMU_OT_DSL_PROPS, /* ZAP */
146 147 DMU_OT_DSL_DATASET, /* UINT64 */
147 148 /* zpl: */
148 149 DMU_OT_ZNODE, /* ZNODE */
149 150 DMU_OT_OLDACL, /* Old ACL */
150 151 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */
151 152 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */
152 153 DMU_OT_MASTER_NODE, /* ZAP */
153 154 DMU_OT_UNLINKED_SET, /* ZAP */
154 155 /* zvol: */
155 156 DMU_OT_ZVOL, /* UINT8 */
156 157 DMU_OT_ZVOL_PROP, /* ZAP */
157 158 /* other; for testing only! */
158 159 DMU_OT_PLAIN_OTHER, /* UINT8 */
159 160 DMU_OT_UINT64_OTHER, /* UINT64 */
160 161 DMU_OT_ZAP_OTHER, /* ZAP */
161 162 /* new object types: */
162 163 DMU_OT_ERROR_LOG, /* ZAP */
163 164 DMU_OT_SPA_HISTORY, /* UINT8 */
164 165 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */
165 166 DMU_OT_POOL_PROPS, /* ZAP */
166 167 DMU_OT_DSL_PERMS, /* ZAP */
167 168 DMU_OT_ACL, /* ACL */
168 169 DMU_OT_SYSACL, /* SYSACL */
169 170 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */
170 171 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */
171 172 DMU_OT_NEXT_CLONES, /* ZAP */
172 173 DMU_OT_SCAN_QUEUE, /* ZAP */
173 174 DMU_OT_USERGROUP_USED, /* ZAP */
174 175 DMU_OT_USERGROUP_QUOTA, /* ZAP */
175 176 DMU_OT_USERREFS, /* ZAP */
176 177 DMU_OT_DDT_ZAP, /* ZAP */
177 178 DMU_OT_DDT_STATS, /* ZAP */
178 179 DMU_OT_SA, /* System attr */
179 180 DMU_OT_SA_MASTER_NODE, /* ZAP */
180 181 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */
181 182 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */
182 183 DMU_OT_SCAN_XLATE, /* ZAP */
183 184 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */
184 185 DMU_OT_DEADLIST, /* ZAP */
185 186 DMU_OT_DEADLIST_HDR, /* UINT64 */
186 187 DMU_OT_DSL_CLONES, /* ZAP */
187 188 DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */
188 189 /*
189 190 * Do not allocate new object types here. Doing so makes the on-disk
190 191 * format incompatible with any other format that uses the same object
191 192 * type number.
192 193 *
193 194 * When creating an object which does not have one of the above types
194 195 * use the DMU_OTN_* type with the correct byteswap and metadata
195 196 * values.
196 197 *
197 198 * The DMU_OTN_* types do not have entries in the dmu_ot table,
198 199 * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead
199 200 * of indexing into dmu_ot directly (this works for both DMU_OT_* types
200 201 * and DMU_OTN_* types).
201 202 */
202 203 DMU_OT_NUMTYPES,
203 204
204 205 /*
205 206 * Names for valid types declared with DMU_OT().
206 207 */
207 208 DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE),
208 209 DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE),
↓ open down ↓ |
152 lines elided |
↑ open up ↑ |
209 210 DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE),
210 211 DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE),
211 212 DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE),
212 213 DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE),
213 214 DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE),
214 215 DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE),
215 216 DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE),
216 217 DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE),
217 218 } dmu_object_type_t;
218 219
219 -typedef enum dmu_objset_type {
220 - DMU_OST_NONE,
221 - DMU_OST_META,
222 - DMU_OST_ZFS,
223 - DMU_OST_ZVOL,
224 - DMU_OST_OTHER, /* For testing only! */
225 - DMU_OST_ANY, /* Be careful! */
226 - DMU_OST_NUMTYPES
227 -} dmu_objset_type_t;
228 -
229 220 void byteswap_uint64_array(void *buf, size_t size);
230 221 void byteswap_uint32_array(void *buf, size_t size);
231 222 void byteswap_uint16_array(void *buf, size_t size);
232 223 void byteswap_uint8_array(void *buf, size_t size);
233 224 void zap_byteswap(void *buf, size_t size);
234 225 void zfs_oldacl_byteswap(void *buf, size_t size);
235 226 void zfs_acl_byteswap(void *buf, size_t size);
236 227 void zfs_znode_byteswap(void *buf, size_t size);
237 228
238 229 #define DS_FIND_SNAPSHOTS (1<<0)
239 230 #define DS_FIND_CHILDREN (1<<1)
240 231
241 232 /*
242 233 * The maximum number of bytes that can be accessed as part of one
243 234 * operation, including metadata.
244 235 */
245 236 #define DMU_MAX_ACCESS (10<<20) /* 10MB */
246 237 #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
247 238
248 239 #define DMU_USERUSED_OBJECT (-1ULL)
249 240 #define DMU_GROUPUSED_OBJECT (-2ULL)
250 241 #define DMU_DEADLIST_OBJECT (-3ULL)
251 242
252 243 /*
253 244 * artificial blkids for bonus buffer and spill blocks
254 245 */
255 246 #define DMU_BONUS_BLKID (-1ULL)
256 247 #define DMU_SPILL_BLKID (-2ULL)
257 248 /*
258 249 * Public routines to create, destroy, open, and close objsets.
259 250 */
260 251 int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
261 252 int dmu_objset_own(const char *name, dmu_objset_type_t type,
262 253 boolean_t readonly, void *tag, objset_t **osp);
↓ open down ↓ |
24 lines elided |
↑ open up ↑ |
263 254 void dmu_objset_rele(objset_t *os, void *tag);
264 255 void dmu_objset_disown(objset_t *os, void *tag);
265 256 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
266 257
267 258 int dmu_objset_evict_dbufs(objset_t *os);
268 259 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
269 260 void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
270 261 int dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin,
271 262 uint64_t flags);
272 263 int dmu_objset_destroy(const char *name, boolean_t defer);
273 -int dmu_snapshots_destroy_nvl(struct nvlist *snaps, boolean_t defer, char *);
274 -int dmu_objset_snapshot(char *fsname, char *snapname, char *tag,
275 - struct nvlist *props, boolean_t recursive, boolean_t temporary, int fd);
264 +int dmu_snapshots_destroy_nvl(struct nvlist *snaps, boolean_t defer,
265 + struct nvlist *errlist);
266 +int dmu_objset_snapshot(struct nvlist *snaps, struct nvlist *, struct nvlist *);
267 +int dmu_objset_snapshot_one(const char *fsname, const char *snapname);
268 +int dmu_objset_snapshot_tmp(const char *, const char *, int);
276 269 int dmu_objset_rename(const char *name, const char *newname,
277 270 boolean_t recursive);
278 271 int dmu_objset_find(char *name, int func(const char *, void *), void *arg,
279 272 int flags);
280 273 void dmu_objset_byteswap(void *buf, size_t size);
281 274
282 275 typedef struct dmu_buf {
283 276 uint64_t db_object; /* object that this buffer is part of */
284 277 uint64_t db_offset; /* byte offset in this object */
285 278 uint64_t db_size; /* size of buffer in bytes */
286 279 void *db_data; /* data in buffer */
287 280 } dmu_buf_t;
288 281
289 282 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
290 283
291 284 /*
292 285 * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
293 286 */
294 287 #define DMU_POOL_DIRECTORY_OBJECT 1
295 288 #define DMU_POOL_CONFIG "config"
296 289 #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write"
297 290 #define DMU_POOL_FEATURES_FOR_READ "features_for_read"
298 291 #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions"
299 292 #define DMU_POOL_ROOT_DATASET "root_dataset"
300 293 #define DMU_POOL_SYNC_BPOBJ "sync_bplist"
301 294 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub"
302 295 #define DMU_POOL_ERRLOG_LAST "errlog_last"
303 296 #define DMU_POOL_SPARES "spares"
304 297 #define DMU_POOL_DEFLATE "deflate"
305 298 #define DMU_POOL_HISTORY "history"
306 299 #define DMU_POOL_PROPS "pool_props"
307 300 #define DMU_POOL_L2CACHE "l2cache"
308 301 #define DMU_POOL_TMP_USERREFS "tmp_userrefs"
309 302 #define DMU_POOL_DDT "DDT-%s-%s-%s"
310 303 #define DMU_POOL_DDT_STATS "DDT-statistics"
311 304 #define DMU_POOL_CREATION_VERSION "creation_version"
312 305 #define DMU_POOL_SCAN "scan"
313 306 #define DMU_POOL_FREE_BPOBJ "free_bpobj"
314 307 #define DMU_POOL_BPTREE_OBJ "bptree_obj"
315 308
316 309 /*
317 310 * Allocate an object from this objset. The range of object numbers
318 311 * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode.
319 312 *
320 313 * The transaction must be assigned to a txg. The newly allocated
321 314 * object will be "held" in the transaction (ie. you can modify the
322 315 * newly allocated object in this transaction).
323 316 *
324 317 * dmu_object_alloc() chooses an object and returns it in *objectp.
325 318 *
326 319 * dmu_object_claim() allocates a specific object number. If that
327 320 * number is already allocated, it fails and returns EEXIST.
328 321 *
329 322 * Return 0 on success, or ENOSPC or EEXIST as specified above.
330 323 */
331 324 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
332 325 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
333 326 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
334 327 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
335 328 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
336 329 int blocksize, dmu_object_type_t bonustype, int bonuslen);
337 330
338 331 /*
339 332 * Free an object from this objset.
340 333 *
341 334 * The object's data will be freed as well (ie. you don't need to call
342 335 * dmu_free(object, 0, -1, tx)).
343 336 *
344 337 * The object need not be held in the transaction.
345 338 *
346 339 * If there are any holds on this object's buffers (via dmu_buf_hold()),
347 340 * or tx holds on the object (via dmu_tx_hold_object()), you can not
348 341 * free it; it fails and returns EBUSY.
349 342 *
350 343 * If the object is not allocated, it fails and returns ENOENT.
351 344 *
352 345 * Return 0 on success, or EBUSY or ENOENT as specified above.
353 346 */
354 347 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
355 348
356 349 /*
357 350 * Find the next allocated or free object.
358 351 *
359 352 * The objectp parameter is in-out. It will be updated to be the next
360 353 * object which is allocated. Ignore objects which have not been
361 354 * modified since txg.
362 355 *
363 356 * XXX Can only be called on a objset with no dirty data.
364 357 *
365 358 * Returns 0 on success, or ENOENT if there are no more objects.
366 359 */
367 360 int dmu_object_next(objset_t *os, uint64_t *objectp,
368 361 boolean_t hole, uint64_t txg);
369 362
370 363 /*
371 364 * Set the data blocksize for an object.
372 365 *
373 366 * The object cannot have any blocks allcated beyond the first. If
374 367 * the first block is allocated already, the new size must be greater
375 368 * than the current block size. If these conditions are not met,
376 369 * ENOTSUP will be returned.
377 370 *
378 371 * Returns 0 on success, or EBUSY if there are any holds on the object
379 372 * contents, or ENOTSUP as described above.
380 373 */
381 374 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
382 375 int ibs, dmu_tx_t *tx);
383 376
384 377 /*
385 378 * Set the checksum property on a dnode. The new checksum algorithm will
386 379 * apply to all newly written blocks; existing blocks will not be affected.
387 380 */
388 381 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
389 382 dmu_tx_t *tx);
390 383
391 384 /*
392 385 * Set the compress property on a dnode. The new compression algorithm will
393 386 * apply to all newly written blocks; existing blocks will not be affected.
394 387 */
395 388 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
396 389 dmu_tx_t *tx);
397 390
398 391 /*
399 392 * Decide how to write a block: checksum, compression, number of copies, etc.
400 393 */
401 394 #define WP_NOFILL 0x1
402 395 #define WP_DMU_SYNC 0x2
403 396 #define WP_SPILL 0x4
404 397
405 398 void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp,
406 399 struct zio_prop *zp);
407 400 /*
408 401 * The bonus data is accessed more or less like a regular buffer.
409 402 * You must dmu_bonus_hold() to get the buffer, which will give you a
410 403 * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
411 404 * data. As with any normal buffer, you must call dmu_buf_read() to
412 405 * read db_data, dmu_buf_will_dirty() before modifying it, and the
413 406 * object must be held in an assigned transaction before calling
414 407 * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus
415 408 * buffer as well. You must release your hold with dmu_buf_rele().
416 409 */
417 410 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
418 411 int dmu_bonus_max(void);
419 412 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
420 413 int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
421 414 dmu_object_type_t dmu_get_bonustype(dmu_buf_t *);
422 415 int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
423 416
424 417 /*
425 418 * Special spill buffer support used by "SA" framework
426 419 */
427 420
428 421 int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
429 422 int dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags,
430 423 void *tag, dmu_buf_t **dbp);
431 424 int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
432 425
433 426 /*
434 427 * Obtain the DMU buffer from the specified object which contains the
435 428 * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so
436 429 * that it will remain in memory. You must release the hold with
437 430 * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your
438 431 * hold. You must have a hold on any dmu_buf_t* you pass to the DMU.
439 432 *
440 433 * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
441 434 * on the returned buffer before reading or writing the buffer's
442 435 * db_data. The comments for those routines describe what particular
443 436 * operations are valid after calling them.
444 437 *
445 438 * The object number must be a valid, allocated object number.
446 439 */
447 440 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
448 441 void *tag, dmu_buf_t **, int flags);
449 442 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
450 443 void dmu_buf_rele(dmu_buf_t *db, void *tag);
451 444 uint64_t dmu_buf_refcount(dmu_buf_t *db);
452 445
453 446 /*
454 447 * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
455 448 * range of an object. A pointer to an array of dmu_buf_t*'s is
456 449 * returned (in *dbpp).
457 450 *
458 451 * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
459 452 * frees the array. The hold on the array of buffers MUST be released
460 453 * with dmu_buf_rele_array. You can NOT release the hold on each buffer
461 454 * individually with dmu_buf_rele.
462 455 */
463 456 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
464 457 uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
465 458 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
466 459
467 460 /*
468 461 * Returns NULL on success, or the existing user ptr if it's already
469 462 * been set.
470 463 *
471 464 * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
472 465 *
473 466 * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
474 467 * will be set to db->db_data when you are allowed to access it. Note
475 468 * that db->db_data (the pointer) can change when you do dmu_buf_read(),
476 469 * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
477 470 * *user_data_ptr_ptr will be set to the new value when it changes.
478 471 *
479 472 * If non-NULL, pageout func will be called when this buffer is being
480 473 * excised from the cache, so that you can clean up the data structure
481 474 * pointed to by user_ptr.
482 475 *
483 476 * dmu_evict_user() will call the pageout func for all buffers in a
484 477 * objset with a given pageout func.
485 478 */
486 479 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
487 480 dmu_buf_evict_func_t *pageout_func);
488 481 /*
489 482 * set_user_ie is the same as set_user, but request immediate eviction
490 483 * when hold count goes to zero.
491 484 */
492 485 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
493 486 void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
494 487 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
495 488 void *user_ptr, void *user_data_ptr_ptr,
496 489 dmu_buf_evict_func_t *pageout_func);
497 490 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
498 491
499 492 /*
500 493 * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
501 494 */
502 495 void *dmu_buf_get_user(dmu_buf_t *db);
503 496
504 497 /*
505 498 * Indicate that you are going to modify the buffer's data (db_data).
506 499 *
507 500 * The transaction (tx) must be assigned to a txg (ie. you've called
508 501 * dmu_tx_assign()). The buffer's object must be held in the tx
509 502 * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
510 503 */
511 504 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
512 505
513 506 /*
514 507 * Tells if the given dbuf is freeable.
515 508 */
516 509 boolean_t dmu_buf_freeable(dmu_buf_t *);
517 510
518 511 /*
519 512 * You must create a transaction, then hold the objects which you will
520 513 * (or might) modify as part of this transaction. Then you must assign
521 514 * the transaction to a transaction group. Once the transaction has
522 515 * been assigned, you can modify buffers which belong to held objects as
523 516 * part of this transaction. You can't modify buffers before the
524 517 * transaction has been assigned; you can't modify buffers which don't
525 518 * belong to objects which this transaction holds; you can't hold
526 519 * objects once the transaction has been assigned. You may hold an
527 520 * object which you are going to free (with dmu_object_free()), but you
528 521 * don't have to.
529 522 *
530 523 * You can abort the transaction before it has been assigned.
531 524 *
532 525 * Note that you may hold buffers (with dmu_buf_hold) at any time,
533 526 * regardless of transaction state.
534 527 */
535 528
536 529 #define DMU_NEW_OBJECT (-1ULL)
537 530 #define DMU_OBJECT_END (-1ULL)
538 531
539 532 dmu_tx_t *dmu_tx_create(objset_t *os);
540 533 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
541 534 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
542 535 uint64_t len);
543 536 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
544 537 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
545 538 void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
546 539 void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
547 540 void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
548 541 void dmu_tx_abort(dmu_tx_t *tx);
549 542 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
550 543 void dmu_tx_wait(dmu_tx_t *tx);
551 544 void dmu_tx_commit(dmu_tx_t *tx);
552 545
553 546 /*
554 547 * To register a commit callback, dmu_tx_callback_register() must be called.
555 548 *
556 549 * dcb_data is a pointer to caller private data that is passed on as a
557 550 * callback parameter. The caller is responsible for properly allocating and
558 551 * freeing it.
559 552 *
560 553 * When registering a callback, the transaction must be already created, but
561 554 * it cannot be committed or aborted. It can be assigned to a txg or not.
562 555 *
563 556 * The callback will be called after the transaction has been safely written
564 557 * to stable storage and will also be called if the dmu_tx is aborted.
565 558 * If there is any error which prevents the transaction from being committed to
566 559 * disk, the callback will be called with a value of error != 0.
567 560 */
568 561 typedef void dmu_tx_callback_func_t(void *dcb_data, int error);
569 562
570 563 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
571 564 void *dcb_data);
572 565
573 566 /*
574 567 * Free up the data blocks for a defined range of a file. If size is
575 568 * -1, the range from offset to end-of-file is freed.
576 569 */
577 570 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
578 571 uint64_t size, dmu_tx_t *tx);
579 572 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
580 573 uint64_t size);
581 574 int dmu_free_object(objset_t *os, uint64_t object);
582 575
583 576 /*
584 577 * Convenience functions.
585 578 *
586 579 * Canfail routines will return 0 on success, or an errno if there is a
587 580 * nonrecoverable I/O error.
588 581 */
589 582 #define DMU_READ_PREFETCH 0 /* prefetch */
590 583 #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */
591 584 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
592 585 void *buf, uint32_t flags);
593 586 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
594 587 const void *buf, dmu_tx_t *tx);
595 588 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
596 589 dmu_tx_t *tx);
597 590 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
598 591 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
599 592 dmu_tx_t *tx);
600 593 int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
601 594 dmu_tx_t *tx);
602 595 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
603 596 uint64_t size, struct page *pp, dmu_tx_t *tx);
604 597 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
605 598 void dmu_return_arcbuf(struct arc_buf *buf);
606 599 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
607 600 dmu_tx_t *tx);
608 601 int dmu_xuio_init(struct xuio *uio, int niov);
609 602 void dmu_xuio_fini(struct xuio *uio);
610 603 int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
611 604 size_t n);
612 605 int dmu_xuio_cnt(struct xuio *uio);
613 606 struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
614 607 void dmu_xuio_clear(struct xuio *uio, int i);
615 608 void xuio_stat_wbuf_copied();
616 609 void xuio_stat_wbuf_nocopy();
617 610
618 611 extern int zfs_prefetch_disable;
619 612
620 613 /*
621 614 * Asynchronously try to read in the data.
622 615 */
623 616 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
624 617 uint64_t len);
625 618
626 619 typedef struct dmu_object_info {
627 620 /* All sizes are in bytes unless otherwise indicated. */
628 621 uint32_t doi_data_block_size;
629 622 uint32_t doi_metadata_block_size;
630 623 dmu_object_type_t doi_type;
631 624 dmu_object_type_t doi_bonus_type;
632 625 uint64_t doi_bonus_size;
633 626 uint8_t doi_indirection; /* 2 = dnode->indirect->data */
634 627 uint8_t doi_checksum;
635 628 uint8_t doi_compress;
636 629 uint8_t doi_pad[5];
637 630 uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */
638 631 uint64_t doi_max_offset;
639 632 uint64_t doi_fill_count; /* number of non-empty blocks */
640 633 } dmu_object_info_t;
641 634
642 635 typedef void arc_byteswap_func_t(void *buf, size_t size);
643 636
644 637 typedef struct dmu_object_type_info {
645 638 dmu_object_byteswap_t ot_byteswap;
646 639 boolean_t ot_metadata;
647 640 char *ot_name;
648 641 } dmu_object_type_info_t;
649 642
650 643 typedef struct dmu_object_byteswap_info {
651 644 arc_byteswap_func_t *ob_func;
652 645 char *ob_name;
653 646 } dmu_object_byteswap_info_t;
654 647
655 648 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
656 649 extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS];
657 650
658 651 /*
659 652 * Get information on a DMU object.
660 653 *
661 654 * Return 0 on success or ENOENT if object is not allocated.
662 655 *
663 656 * If doi is NULL, just indicates whether the object exists.
664 657 */
665 658 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
666 659 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
667 660 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
668 661 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
669 662 u_longlong_t *nblk512);
670 663
671 664 typedef struct dmu_objset_stats {
672 665 uint64_t dds_num_clones; /* number of clones of this */
673 666 uint64_t dds_creation_txg;
674 667 uint64_t dds_guid;
675 668 dmu_objset_type_t dds_type;
676 669 uint8_t dds_is_snapshot;
677 670 uint8_t dds_inconsistent;
678 671 char dds_origin[MAXNAMELEN];
679 672 } dmu_objset_stats_t;
680 673
681 674 /*
682 675 * Get stats on a dataset.
683 676 */
684 677 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
685 678
686 679 /*
687 680 * Add entries to the nvlist for all the objset's properties. See
688 681 * zfs_prop_table[] and zfs(1m) for details on the properties.
689 682 */
690 683 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
691 684
692 685 /*
693 686 * Get the space usage statistics for statvfs().
694 687 *
695 688 * refdbytes is the amount of space "referenced" by this objset.
696 689 * availbytes is the amount of space available to this objset, taking
697 690 * into account quotas & reservations, assuming that no other objsets
698 691 * use the space first. These values correspond to the 'referenced' and
699 692 * 'available' properties, described in the zfs(1m) manpage.
700 693 *
701 694 * usedobjs and availobjs are the number of objects currently allocated,
702 695 * and available.
703 696 */
704 697 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
705 698 uint64_t *usedobjsp, uint64_t *availobjsp);
706 699
707 700 /*
708 701 * The fsid_guid is a 56-bit ID that can change to avoid collisions.
709 702 * (Contrast with the ds_guid which is a 64-bit ID that will never
710 703 * change, so there is a small probability that it will collide.)
711 704 */
712 705 uint64_t dmu_objset_fsid_guid(objset_t *os);
713 706
714 707 /*
715 708 * Get the [cm]time for an objset's snapshot dir
716 709 */
717 710 timestruc_t dmu_objset_snap_cmtime(objset_t *os);
718 711
719 712 int dmu_objset_is_snapshot(objset_t *os);
720 713
721 714 extern struct spa *dmu_objset_spa(objset_t *os);
722 715 extern struct zilog *dmu_objset_zil(objset_t *os);
723 716 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
724 717 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
725 718 extern void dmu_objset_name(objset_t *os, char *buf);
726 719 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
727 720 extern uint64_t dmu_objset_id(objset_t *os);
728 721 extern uint64_t dmu_objset_syncprop(objset_t *os);
729 722 extern uint64_t dmu_objset_logbias(objset_t *os);
730 723 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
731 724 uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
732 725 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
733 726 int maxlen, boolean_t *conflict);
734 727 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
735 728 uint64_t *idp, uint64_t *offp);
736 729
737 730 typedef int objset_used_cb_t(dmu_object_type_t bonustype,
738 731 void *bonus, uint64_t *userp, uint64_t *groupp);
739 732 extern void dmu_objset_register_type(dmu_objset_type_t ost,
740 733 objset_used_cb_t *cb);
741 734 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
742 735 extern void *dmu_objset_get_user(objset_t *os);
743 736
744 737 /*
745 738 * Return the txg number for the given assigned transaction.
746 739 */
747 740 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
748 741
749 742 /*
750 743 * Synchronous write.
751 744 * If a parent zio is provided this function initiates a write on the
752 745 * provided buffer as a child of the parent zio.
753 746 * In the absence of a parent zio, the write is completed synchronously.
754 747 * At write completion, blk is filled with the bp of the written block.
755 748 * Note that while the data covered by this function will be on stable
756 749 * storage when the write completes this new data does not become a
757 750 * permanent part of the file until the associated transaction commits.
758 751 */
759 752
760 753 /*
761 754 * {zfs,zvol,ztest}_get_done() args
762 755 */
763 756 typedef struct zgd {
764 757 struct zilog *zgd_zilog;
765 758 struct blkptr *zgd_bp;
766 759 dmu_buf_t *zgd_db;
767 760 struct rl *zgd_rl;
768 761 void *zgd_private;
769 762 } zgd_t;
770 763
771 764 typedef void dmu_sync_cb_t(zgd_t *arg, int error);
772 765 int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
773 766
774 767 /*
775 768 * Find the next hole or data block in file starting at *off
776 769 * Return found offset in *off. Return ESRCH for end of file.
777 770 */
778 771 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
779 772 uint64_t *off);
780 773
781 774 /*
↓ open down ↓ |
496 lines elided |
↑ open up ↑ |
782 775 * Initial setup and final teardown.
783 776 */
784 777 extern void dmu_init(void);
785 778 extern void dmu_fini(void);
786 779
787 780 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
788 781 uint64_t object, uint64_t offset, int len);
789 782 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
790 783 dmu_traverse_cb_t cb, void *arg);
791 784
792 -int dmu_send(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
785 +int dmu_send(objset_t *tosnap, objset_t *fromsnap,
793 786 int outfd, struct vnode *vp, offset_t *off);
794 -int dmu_send_estimate(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorign,
795 - uint64_t *sizep);
787 +int dmu_send_estimate(objset_t *tosnap, objset_t *fromsnap, uint64_t *sizep);
796 788
797 789 typedef struct dmu_recv_cookie {
798 790 /*
799 791 * This structure is opaque!
800 792 *
801 793 * If logical and real are different, we are recving the stream
802 794 * into the "real" temporary clone, and then switching it with
803 795 * the "logical" target.
804 796 */
805 797 struct dsl_dataset *drc_logical_ds;
806 798 struct dsl_dataset *drc_real_ds;
807 799 struct drr_begin *drc_drrb;
808 800 char *drc_tosnap;
809 801 char *drc_top_ds;
810 802 boolean_t drc_newfs;
811 803 boolean_t drc_force;
812 804 struct avl_tree *drc_guid_to_ds_map;
813 805 } dmu_recv_cookie_t;
814 806
815 807 int dmu_recv_begin(char *tofs, char *tosnap, char *topds, struct drr_begin *,
816 808 boolean_t force, objset_t *origin, dmu_recv_cookie_t *);
817 809 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp,
818 810 int cleanup_fd, uint64_t *action_handlep);
819 811 int dmu_recv_end(dmu_recv_cookie_t *drc);
820 812
821 813 int dmu_diff(objset_t *tosnap, objset_t *fromsnap, struct vnode *vp,
822 814 offset_t *off);
823 815
824 816 /* CRC64 table */
825 817 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */
826 818 extern uint64_t zfs_crc64_table[256];
827 819
828 820 #ifdef __cplusplus
829 821 }
830 822 #endif
831 823
832 824 #endif /* _SYS_DMU_H */
↓ open down ↓ |
27 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX