Print this page
3741 zfs needs better comments
Submitted by: Will Andrews <willa@spectralogic.com>
Submitted by: Justin Gibbs <justing@spectralogic.com>
Submitted by: Alan Somers <alans@spectralogic.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.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.
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 47 #include <sys/fs/zfs.h>
48 48
49 49 #ifdef __cplusplus
50 50 extern "C" {
51 51 #endif
52 52
53 53 struct uio;
54 54 struct xuio;
55 55 struct page;
56 56 struct vnode;
57 57 struct spa;
58 58 struct zilog;
59 59 struct zio;
60 60 struct blkptr;
61 61 struct zap_cursor;
62 62 struct dsl_dataset;
63 63 struct dsl_pool;
64 64 struct dnode;
65 65 struct drr_begin;
66 66 struct drr_end;
67 67 struct zbookmark;
68 68 struct spa;
69 69 struct nvlist;
70 70 struct arc_buf;
71 71 struct zio_prop;
72 72 struct sa_handle;
73 73
74 74 typedef struct objset objset_t;
75 75 typedef struct dmu_tx dmu_tx_t;
76 76 typedef struct dsl_dir dsl_dir_t;
77 77
78 78 typedef enum dmu_object_byteswap {
79 79 DMU_BSWAP_UINT8,
80 80 DMU_BSWAP_UINT16,
81 81 DMU_BSWAP_UINT32,
82 82 DMU_BSWAP_UINT64,
83 83 DMU_BSWAP_ZAP,
84 84 DMU_BSWAP_DNODE,
85 85 DMU_BSWAP_OBJSET,
86 86 DMU_BSWAP_ZNODE,
87 87 DMU_BSWAP_OLDACL,
88 88 DMU_BSWAP_ACL,
89 89 /*
90 90 * Allocating a new byteswap type number makes the on-disk format
91 91 * incompatible with any other format that uses the same number.
92 92 *
93 93 * Data can usually be structured to work with one of the
94 94 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
95 95 */
96 96 DMU_BSWAP_NUMFUNCS
97 97 } dmu_object_byteswap_t;
98 98
99 99 #define DMU_OT_NEWTYPE 0x80
100 100 #define DMU_OT_METADATA 0x40
101 101 #define DMU_OT_BYTESWAP_MASK 0x3f
102 102
103 103 /*
104 104 * Defines a uint8_t object type. Object types specify if the data
105 105 * in the object is metadata (boolean) and how to byteswap the data
106 106 * (dmu_object_byteswap_t).
107 107 */
108 108 #define DMU_OT(byteswap, metadata) \
109 109 (DMU_OT_NEWTYPE | \
110 110 ((metadata) ? DMU_OT_METADATA : 0) | \
111 111 ((byteswap) & DMU_OT_BYTESWAP_MASK))
112 112
113 113 #define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \
114 114 ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \
115 115 (ot) < DMU_OT_NUMTYPES)
116 116
117 117 #define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \
118 118 ((ot) & DMU_OT_METADATA) : \
119 119 dmu_ot[(ot)].ot_metadata)
120 120
121 121 #define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \
122 122 ((ot) & DMU_OT_BYTESWAP_MASK) : \
123 123 dmu_ot[(ot)].ot_byteswap)
124 124
125 125 typedef enum dmu_object_type {
126 126 DMU_OT_NONE,
127 127 /* general: */
128 128 DMU_OT_OBJECT_DIRECTORY, /* ZAP */
129 129 DMU_OT_OBJECT_ARRAY, /* UINT64 */
130 130 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */
131 131 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */
132 132 DMU_OT_BPOBJ, /* UINT64 */
133 133 DMU_OT_BPOBJ_HDR, /* UINT64 */
134 134 /* spa: */
135 135 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */
136 136 DMU_OT_SPACE_MAP, /* UINT64 */
137 137 /* zil: */
138 138 DMU_OT_INTENT_LOG, /* UINT64 */
139 139 /* dmu: */
140 140 DMU_OT_DNODE, /* DNODE */
141 141 DMU_OT_OBJSET, /* OBJSET */
142 142 /* dsl: */
143 143 DMU_OT_DSL_DIR, /* UINT64 */
144 144 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */
145 145 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */
146 146 DMU_OT_DSL_PROPS, /* ZAP */
147 147 DMU_OT_DSL_DATASET, /* UINT64 */
148 148 /* zpl: */
149 149 DMU_OT_ZNODE, /* ZNODE */
150 150 DMU_OT_OLDACL, /* Old ACL */
151 151 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */
152 152 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */
153 153 DMU_OT_MASTER_NODE, /* ZAP */
154 154 DMU_OT_UNLINKED_SET, /* ZAP */
155 155 /* zvol: */
156 156 DMU_OT_ZVOL, /* UINT8 */
157 157 DMU_OT_ZVOL_PROP, /* ZAP */
158 158 /* other; for testing only! */
159 159 DMU_OT_PLAIN_OTHER, /* UINT8 */
160 160 DMU_OT_UINT64_OTHER, /* UINT64 */
161 161 DMU_OT_ZAP_OTHER, /* ZAP */
162 162 /* new object types: */
163 163 DMU_OT_ERROR_LOG, /* ZAP */
164 164 DMU_OT_SPA_HISTORY, /* UINT8 */
165 165 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */
166 166 DMU_OT_POOL_PROPS, /* ZAP */
167 167 DMU_OT_DSL_PERMS, /* ZAP */
168 168 DMU_OT_ACL, /* ACL */
169 169 DMU_OT_SYSACL, /* SYSACL */
170 170 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */
171 171 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */
172 172 DMU_OT_NEXT_CLONES, /* ZAP */
173 173 DMU_OT_SCAN_QUEUE, /* ZAP */
174 174 DMU_OT_USERGROUP_USED, /* ZAP */
175 175 DMU_OT_USERGROUP_QUOTA, /* ZAP */
176 176 DMU_OT_USERREFS, /* ZAP */
177 177 DMU_OT_DDT_ZAP, /* ZAP */
178 178 DMU_OT_DDT_STATS, /* ZAP */
179 179 DMU_OT_SA, /* System attr */
180 180 DMU_OT_SA_MASTER_NODE, /* ZAP */
181 181 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */
182 182 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */
183 183 DMU_OT_SCAN_XLATE, /* ZAP */
184 184 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */
185 185 DMU_OT_DEADLIST, /* ZAP */
186 186 DMU_OT_DEADLIST_HDR, /* UINT64 */
187 187 DMU_OT_DSL_CLONES, /* ZAP */
188 188 DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */
189 189 /*
190 190 * Do not allocate new object types here. Doing so makes the on-disk
191 191 * format incompatible with any other format that uses the same object
192 192 * type number.
193 193 *
194 194 * When creating an object which does not have one of the above types
195 195 * use the DMU_OTN_* type with the correct byteswap and metadata
196 196 * values.
197 197 *
198 198 * The DMU_OTN_* types do not have entries in the dmu_ot table,
199 199 * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead
200 200 * of indexing into dmu_ot directly (this works for both DMU_OT_* types
201 201 * and DMU_OTN_* types).
202 202 */
203 203 DMU_OT_NUMTYPES,
204 204
205 205 /*
206 206 * Names for valid types declared with DMU_OT().
207 207 */
208 208 DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE),
209 209 DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE),
210 210 DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE),
211 211 DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE),
212 212 DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE),
213 213 DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE),
214 214 DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE),
215 215 DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE),
216 216 DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE),
217 217 DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE),
218 218 } dmu_object_type_t;
219 219
220 220 typedef enum txg_how {
221 221 TXG_WAIT = 1,
222 222 TXG_NOWAIT,
223 223 } txg_how_t;
224 224
225 225 void byteswap_uint64_array(void *buf, size_t size);
226 226 void byteswap_uint32_array(void *buf, size_t size);
227 227 void byteswap_uint16_array(void *buf, size_t size);
228 228 void byteswap_uint8_array(void *buf, size_t size);
229 229 void zap_byteswap(void *buf, size_t size);
230 230 void zfs_oldacl_byteswap(void *buf, size_t size);
231 231 void zfs_acl_byteswap(void *buf, size_t size);
232 232 void zfs_znode_byteswap(void *buf, size_t size);
233 233
234 234 #define DS_FIND_SNAPSHOTS (1<<0)
235 235 #define DS_FIND_CHILDREN (1<<1)
236 236
237 237 /*
238 238 * The maximum number of bytes that can be accessed as part of one
239 239 * operation, including metadata.
240 240 */
241 241 #define DMU_MAX_ACCESS (10<<20) /* 10MB */
242 242 #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
243 243
244 244 #define DMU_USERUSED_OBJECT (-1ULL)
245 245 #define DMU_GROUPUSED_OBJECT (-2ULL)
246 246 #define DMU_DEADLIST_OBJECT (-3ULL)
247 247
248 248 /*
249 249 * artificial blkids for bonus buffer and spill blocks
250 250 */
251 251 #define DMU_BONUS_BLKID (-1ULL)
252 252 #define DMU_SPILL_BLKID (-2ULL)
253 253 /*
254 254 * Public routines to create, destroy, open, and close objsets.
255 255 */
256 256 int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
257 257 int dmu_objset_own(const char *name, dmu_objset_type_t type,
258 258 boolean_t readonly, void *tag, objset_t **osp);
259 259 void dmu_objset_rele(objset_t *os, void *tag);
260 260 void dmu_objset_disown(objset_t *os, void *tag);
261 261 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
262 262
263 263 void dmu_objset_evict_dbufs(objset_t *os);
264 264 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
265 265 void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
266 266 int dmu_objset_clone(const char *name, const char *origin);
267 267 int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer,
268 268 struct nvlist *errlist);
269 269 int dmu_objset_snapshot_one(const char *fsname, const char *snapname);
270 270 int dmu_objset_snapshot_tmp(const char *, const char *, int);
271 271 int dmu_objset_find(char *name, int func(const char *, void *), void *arg,
272 272 int flags);
273 273 void dmu_objset_byteswap(void *buf, size_t size);
274 274 int dsl_dataset_rename_snapshot(const char *fsname,
275 275 const char *oldsnapname, const char *newsnapname, boolean_t recursive);
276 276
277 277 typedef struct dmu_buf {
278 278 uint64_t db_object; /* object that this buffer is part of */
279 279 uint64_t db_offset; /* byte offset in this object */
280 280 uint64_t db_size; /* size of buffer in bytes */
281 281 void *db_data; /* data in buffer */
282 282 } dmu_buf_t;
283 283
284 284 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
285 285
286 286 /*
287 287 * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
288 288 */
289 289 #define DMU_POOL_DIRECTORY_OBJECT 1
290 290 #define DMU_POOL_CONFIG "config"
291 291 #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write"
292 292 #define DMU_POOL_FEATURES_FOR_READ "features_for_read"
293 293 #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions"
294 294 #define DMU_POOL_ROOT_DATASET "root_dataset"
295 295 #define DMU_POOL_SYNC_BPOBJ "sync_bplist"
296 296 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub"
297 297 #define DMU_POOL_ERRLOG_LAST "errlog_last"
298 298 #define DMU_POOL_SPARES "spares"
299 299 #define DMU_POOL_DEFLATE "deflate"
300 300 #define DMU_POOL_HISTORY "history"
301 301 #define DMU_POOL_PROPS "pool_props"
302 302 #define DMU_POOL_L2CACHE "l2cache"
303 303 #define DMU_POOL_TMP_USERREFS "tmp_userrefs"
304 304 #define DMU_POOL_DDT "DDT-%s-%s-%s"
305 305 #define DMU_POOL_DDT_STATS "DDT-statistics"
306 306 #define DMU_POOL_CREATION_VERSION "creation_version"
307 307 #define DMU_POOL_SCAN "scan"
308 308 #define DMU_POOL_FREE_BPOBJ "free_bpobj"
309 309 #define DMU_POOL_BPTREE_OBJ "bptree_obj"
310 310 #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj"
311 311
312 312 /*
313 313 * Allocate an object from this objset. The range of object numbers
314 314 * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode.
315 315 *
316 316 * The transaction must be assigned to a txg. The newly allocated
317 317 * object will be "held" in the transaction (ie. you can modify the
318 318 * newly allocated object in this transaction).
319 319 *
320 320 * dmu_object_alloc() chooses an object and returns it in *objectp.
321 321 *
322 322 * dmu_object_claim() allocates a specific object number. If that
323 323 * number is already allocated, it fails and returns EEXIST.
324 324 *
325 325 * Return 0 on success, or ENOSPC or EEXIST as specified above.
326 326 */
327 327 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
328 328 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
329 329 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
330 330 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
331 331 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
332 332 int blocksize, dmu_object_type_t bonustype, int bonuslen);
333 333
334 334 /*
335 335 * Free an object from this objset.
336 336 *
337 337 * The object's data will be freed as well (ie. you don't need to call
338 338 * dmu_free(object, 0, -1, tx)).
339 339 *
340 340 * The object need not be held in the transaction.
341 341 *
342 342 * If there are any holds on this object's buffers (via dmu_buf_hold()),
343 343 * or tx holds on the object (via dmu_tx_hold_object()), you can not
344 344 * free it; it fails and returns EBUSY.
345 345 *
346 346 * If the object is not allocated, it fails and returns ENOENT.
347 347 *
348 348 * Return 0 on success, or EBUSY or ENOENT as specified above.
349 349 */
350 350 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
351 351
352 352 /*
353 353 * Find the next allocated or free object.
354 354 *
355 355 * The objectp parameter is in-out. It will be updated to be the next
356 356 * object which is allocated. Ignore objects which have not been
357 357 * modified since txg.
358 358 *
359 359 * XXX Can only be called on a objset with no dirty data.
360 360 *
361 361 * Returns 0 on success, or ENOENT if there are no more objects.
362 362 */
363 363 int dmu_object_next(objset_t *os, uint64_t *objectp,
364 364 boolean_t hole, uint64_t txg);
365 365
366 366 /*
367 367 * Set the data blocksize for an object.
368 368 *
369 369 * The object cannot have any blocks allcated beyond the first. If
370 370 * the first block is allocated already, the new size must be greater
371 371 * than the current block size. If these conditions are not met,
372 372 * ENOTSUP will be returned.
373 373 *
374 374 * Returns 0 on success, or EBUSY if there are any holds on the object
375 375 * contents, or ENOTSUP as described above.
376 376 */
377 377 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
378 378 int ibs, dmu_tx_t *tx);
379 379
380 380 /*
381 381 * Set the checksum property on a dnode. The new checksum algorithm will
382 382 * apply to all newly written blocks; existing blocks will not be affected.
383 383 */
384 384 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
385 385 dmu_tx_t *tx);
386 386
387 387 /*
388 388 * Set the compress property on a dnode. The new compression algorithm will
389 389 * apply to all newly written blocks; existing blocks will not be affected.
390 390 */
391 391 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
392 392 dmu_tx_t *tx);
393 393
394 394 /*
395 395 * Decide how to write a block: checksum, compression, number of copies, etc.
396 396 */
397 397 #define WP_NOFILL 0x1
398 398 #define WP_DMU_SYNC 0x2
399 399 #define WP_SPILL 0x4
400 400
401 401 void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp,
↓ open down ↓ |
401 lines elided |
↑ open up ↑ |
402 402 struct zio_prop *zp);
403 403 /*
404 404 * The bonus data is accessed more or less like a regular buffer.
405 405 * You must dmu_bonus_hold() to get the buffer, which will give you a
406 406 * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
407 407 * data. As with any normal buffer, you must call dmu_buf_read() to
408 408 * read db_data, dmu_buf_will_dirty() before modifying it, and the
409 409 * object must be held in an assigned transaction before calling
410 410 * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus
411 411 * buffer as well. You must release your hold with dmu_buf_rele().
412 + *
413 + * Returns ENOENT, EIO, or 0.
412 414 */
413 415 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
414 416 int dmu_bonus_max(void);
415 417 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
416 418 int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
417 419 dmu_object_type_t dmu_get_bonustype(dmu_buf_t *);
418 420 int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
419 421
420 422 /*
421 423 * Special spill buffer support used by "SA" framework
422 424 */
423 425
424 426 int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
425 427 int dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags,
426 428 void *tag, dmu_buf_t **dbp);
427 429 int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
428 430
429 431 /*
430 432 * Obtain the DMU buffer from the specified object which contains the
431 433 * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so
432 434 * that it will remain in memory. You must release the hold with
433 435 * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your
434 436 * hold. You must have a hold on any dmu_buf_t* you pass to the DMU.
435 437 *
436 438 * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
437 439 * on the returned buffer before reading or writing the buffer's
438 440 * db_data. The comments for those routines describe what particular
439 441 * operations are valid after calling them.
440 442 *
441 443 * The object number must be a valid, allocated object number.
442 444 */
443 445 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
444 446 void *tag, dmu_buf_t **, int flags);
445 447 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
446 448 void dmu_buf_rele(dmu_buf_t *db, void *tag);
447 449 uint64_t dmu_buf_refcount(dmu_buf_t *db);
448 450
449 451 /*
450 452 * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
451 453 * range of an object. A pointer to an array of dmu_buf_t*'s is
452 454 * returned (in *dbpp).
453 455 *
454 456 * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
455 457 * frees the array. The hold on the array of buffers MUST be released
456 458 * with dmu_buf_rele_array. You can NOT release the hold on each buffer
457 459 * individually with dmu_buf_rele.
458 460 */
459 461 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
460 462 uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
461 463 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
462 464
463 465 /*
464 466 * Returns NULL on success, or the existing user ptr if it's already
465 467 * been set.
466 468 *
467 469 * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
468 470 *
469 471 * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
470 472 * will be set to db->db_data when you are allowed to access it. Note
471 473 * that db->db_data (the pointer) can change when you do dmu_buf_read(),
472 474 * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
473 475 * *user_data_ptr_ptr will be set to the new value when it changes.
474 476 *
475 477 * If non-NULL, pageout func will be called when this buffer is being
476 478 * excised from the cache, so that you can clean up the data structure
477 479 * pointed to by user_ptr.
478 480 *
479 481 * dmu_evict_user() will call the pageout func for all buffers in a
480 482 * objset with a given pageout func.
481 483 */
482 484 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
483 485 dmu_buf_evict_func_t *pageout_func);
484 486 /*
485 487 * set_user_ie is the same as set_user, but request immediate eviction
486 488 * when hold count goes to zero.
487 489 */
488 490 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
489 491 void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
490 492 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
491 493 void *user_ptr, void *user_data_ptr_ptr,
492 494 dmu_buf_evict_func_t *pageout_func);
493 495 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
494 496
495 497 /*
496 498 * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
497 499 */
498 500 void *dmu_buf_get_user(dmu_buf_t *db);
499 501
500 502 /*
501 503 * Returns the blkptr associated with this dbuf, or NULL if not set.
502 504 */
503 505 struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db);
504 506
505 507 /*
506 508 * Indicate that you are going to modify the buffer's data (db_data).
507 509 *
508 510 * The transaction (tx) must be assigned to a txg (ie. you've called
509 511 * dmu_tx_assign()). The buffer's object must be held in the tx
510 512 * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
511 513 */
512 514 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
513 515
514 516 /*
515 517 * Tells if the given dbuf is freeable.
516 518 */
517 519 boolean_t dmu_buf_freeable(dmu_buf_t *);
518 520
519 521 /*
520 522 * You must create a transaction, then hold the objects which you will
521 523 * (or might) modify as part of this transaction. Then you must assign
522 524 * the transaction to a transaction group. Once the transaction has
523 525 * been assigned, you can modify buffers which belong to held objects as
524 526 * part of this transaction. You can't modify buffers before the
525 527 * transaction has been assigned; you can't modify buffers which don't
526 528 * belong to objects which this transaction holds; you can't hold
527 529 * objects once the transaction has been assigned. You may hold an
528 530 * object which you are going to free (with dmu_object_free()), but you
529 531 * don't have to.
530 532 *
531 533 * You can abort the transaction before it has been assigned.
532 534 *
533 535 * Note that you may hold buffers (with dmu_buf_hold) at any time,
534 536 * regardless of transaction state.
535 537 */
536 538
537 539 #define DMU_NEW_OBJECT (-1ULL)
538 540 #define DMU_OBJECT_END (-1ULL)
539 541
540 542 dmu_tx_t *dmu_tx_create(objset_t *os);
541 543 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
542 544 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
543 545 uint64_t len);
544 546 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
545 547 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
546 548 void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
547 549 void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
548 550 void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
549 551 void dmu_tx_abort(dmu_tx_t *tx);
550 552 int dmu_tx_assign(dmu_tx_t *tx, enum txg_how txg_how);
551 553 void dmu_tx_wait(dmu_tx_t *tx);
552 554 void dmu_tx_commit(dmu_tx_t *tx);
553 555
554 556 /*
555 557 * To register a commit callback, dmu_tx_callback_register() must be called.
556 558 *
557 559 * dcb_data is a pointer to caller private data that is passed on as a
558 560 * callback parameter. The caller is responsible for properly allocating and
559 561 * freeing it.
560 562 *
561 563 * When registering a callback, the transaction must be already created, but
562 564 * it cannot be committed or aborted. It can be assigned to a txg or not.
563 565 *
564 566 * The callback will be called after the transaction has been safely written
565 567 * to stable storage and will also be called if the dmu_tx is aborted.
566 568 * If there is any error which prevents the transaction from being committed to
567 569 * disk, the callback will be called with a value of error != 0.
568 570 */
569 571 typedef void dmu_tx_callback_func_t(void *dcb_data, int error);
570 572
571 573 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
572 574 void *dcb_data);
573 575
574 576 /*
575 577 * Free up the data blocks for a defined range of a file. If size is
576 578 * -1, the range from offset to end-of-file is freed.
577 579 */
578 580 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
579 581 uint64_t size, dmu_tx_t *tx);
580 582 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
581 583 uint64_t size);
582 584 int dmu_free_object(objset_t *os, uint64_t object);
583 585
584 586 /*
585 587 * Convenience functions.
586 588 *
587 589 * Canfail routines will return 0 on success, or an errno if there is a
588 590 * nonrecoverable I/O error.
589 591 */
590 592 #define DMU_READ_PREFETCH 0 /* prefetch */
591 593 #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */
592 594 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
593 595 void *buf, uint32_t flags);
594 596 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
595 597 const void *buf, dmu_tx_t *tx);
596 598 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
597 599 dmu_tx_t *tx);
598 600 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
599 601 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
600 602 dmu_tx_t *tx);
601 603 int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
602 604 dmu_tx_t *tx);
603 605 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
604 606 uint64_t size, struct page *pp, dmu_tx_t *tx);
605 607 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
606 608 void dmu_return_arcbuf(struct arc_buf *buf);
607 609 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
608 610 dmu_tx_t *tx);
609 611 int dmu_xuio_init(struct xuio *uio, int niov);
610 612 void dmu_xuio_fini(struct xuio *uio);
611 613 int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
612 614 size_t n);
613 615 int dmu_xuio_cnt(struct xuio *uio);
614 616 struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
615 617 void dmu_xuio_clear(struct xuio *uio, int i);
616 618 void xuio_stat_wbuf_copied();
617 619 void xuio_stat_wbuf_nocopy();
618 620
619 621 extern int zfs_prefetch_disable;
620 622
621 623 /*
622 624 * Asynchronously try to read in the data.
623 625 */
624 626 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
625 627 uint64_t len);
626 628
627 629 typedef struct dmu_object_info {
628 630 /* All sizes are in bytes unless otherwise indicated. */
629 631 uint32_t doi_data_block_size;
630 632 uint32_t doi_metadata_block_size;
631 633 dmu_object_type_t doi_type;
632 634 dmu_object_type_t doi_bonus_type;
633 635 uint64_t doi_bonus_size;
634 636 uint8_t doi_indirection; /* 2 = dnode->indirect->data */
635 637 uint8_t doi_checksum;
636 638 uint8_t doi_compress;
637 639 uint8_t doi_pad[5];
638 640 uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */
639 641 uint64_t doi_max_offset;
640 642 uint64_t doi_fill_count; /* number of non-empty blocks */
641 643 } dmu_object_info_t;
642 644
643 645 typedef void arc_byteswap_func_t(void *buf, size_t size);
644 646
645 647 typedef struct dmu_object_type_info {
646 648 dmu_object_byteswap_t ot_byteswap;
647 649 boolean_t ot_metadata;
648 650 char *ot_name;
649 651 } dmu_object_type_info_t;
650 652
651 653 typedef struct dmu_object_byteswap_info {
652 654 arc_byteswap_func_t *ob_func;
653 655 char *ob_name;
654 656 } dmu_object_byteswap_info_t;
655 657
656 658 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
↓ open down ↓ |
235 lines elided |
↑ open up ↑ |
657 659 extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS];
658 660
659 661 /*
660 662 * Get information on a DMU object.
661 663 *
662 664 * Return 0 on success or ENOENT if object is not allocated.
663 665 *
664 666 * If doi is NULL, just indicates whether the object exists.
665 667 */
666 668 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
669 +/* Like dmu_object_info, but faster if you have a held dnode in hand. */
667 670 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
671 +/* Like dmu_object_info, but faster if you have a held dbuf in hand. */
668 672 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
673 +/*
674 + * Like dmu_object_info_from_db, but faster still when you only care about
675 + * the size. This is specifically optimized for zfs_getattr().
676 + */
669 677 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
670 678 u_longlong_t *nblk512);
671 679
672 680 typedef struct dmu_objset_stats {
673 681 uint64_t dds_num_clones; /* number of clones of this */
674 682 uint64_t dds_creation_txg;
675 683 uint64_t dds_guid;
676 684 dmu_objset_type_t dds_type;
677 685 uint8_t dds_is_snapshot;
678 686 uint8_t dds_inconsistent;
679 687 char dds_origin[MAXNAMELEN];
680 688 } dmu_objset_stats_t;
681 689
682 690 /*
683 691 * Get stats on a dataset.
684 692 */
685 693 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
686 694
687 695 /*
688 696 * Add entries to the nvlist for all the objset's properties. See
689 697 * zfs_prop_table[] and zfs(1m) for details on the properties.
690 698 */
691 699 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
692 700
693 701 /*
694 702 * Get the space usage statistics for statvfs().
695 703 *
696 704 * refdbytes is the amount of space "referenced" by this objset.
697 705 * availbytes is the amount of space available to this objset, taking
698 706 * into account quotas & reservations, assuming that no other objsets
699 707 * use the space first. These values correspond to the 'referenced' and
700 708 * 'available' properties, described in the zfs(1m) manpage.
701 709 *
702 710 * usedobjs and availobjs are the number of objects currently allocated,
703 711 * and available.
704 712 */
705 713 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
706 714 uint64_t *usedobjsp, uint64_t *availobjsp);
707 715
708 716 /*
709 717 * The fsid_guid is a 56-bit ID that can change to avoid collisions.
710 718 * (Contrast with the ds_guid which is a 64-bit ID that will never
711 719 * change, so there is a small probability that it will collide.)
712 720 */
713 721 uint64_t dmu_objset_fsid_guid(objset_t *os);
714 722
715 723 /*
716 724 * Get the [cm]time for an objset's snapshot dir
717 725 */
718 726 timestruc_t dmu_objset_snap_cmtime(objset_t *os);
719 727
720 728 int dmu_objset_is_snapshot(objset_t *os);
721 729
722 730 extern struct spa *dmu_objset_spa(objset_t *os);
723 731 extern struct zilog *dmu_objset_zil(objset_t *os);
724 732 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
725 733 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
726 734 extern void dmu_objset_name(objset_t *os, char *buf);
727 735 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
728 736 extern uint64_t dmu_objset_id(objset_t *os);
729 737 extern uint64_t dmu_objset_syncprop(objset_t *os);
730 738 extern uint64_t dmu_objset_logbias(objset_t *os);
731 739 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
732 740 uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
733 741 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
734 742 int maxlen, boolean_t *conflict);
735 743 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
736 744 uint64_t *idp, uint64_t *offp);
737 745
738 746 typedef int objset_used_cb_t(dmu_object_type_t bonustype,
739 747 void *bonus, uint64_t *userp, uint64_t *groupp);
740 748 extern void dmu_objset_register_type(dmu_objset_type_t ost,
741 749 objset_used_cb_t *cb);
742 750 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
743 751 extern void *dmu_objset_get_user(objset_t *os);
744 752
745 753 /*
746 754 * Return the txg number for the given assigned transaction.
747 755 */
748 756 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
749 757
750 758 /*
751 759 * Synchronous write.
752 760 * If a parent zio is provided this function initiates a write on the
753 761 * provided buffer as a child of the parent zio.
754 762 * In the absence of a parent zio, the write is completed synchronously.
755 763 * At write completion, blk is filled with the bp of the written block.
756 764 * Note that while the data covered by this function will be on stable
757 765 * storage when the write completes this new data does not become a
758 766 * permanent part of the file until the associated transaction commits.
759 767 */
760 768
761 769 /*
762 770 * {zfs,zvol,ztest}_get_done() args
763 771 */
764 772 typedef struct zgd {
765 773 struct zilog *zgd_zilog;
766 774 struct blkptr *zgd_bp;
767 775 dmu_buf_t *zgd_db;
768 776 struct rl *zgd_rl;
769 777 void *zgd_private;
770 778 } zgd_t;
771 779
772 780 typedef void dmu_sync_cb_t(zgd_t *arg, int error);
773 781 int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
774 782
775 783 /*
776 784 * Find the next hole or data block in file starting at *off
777 785 * Return found offset in *off. Return ESRCH for end of file.
778 786 */
779 787 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
780 788 uint64_t *off);
781 789
782 790 /*
783 791 * Initial setup and final teardown.
784 792 */
785 793 extern void dmu_init(void);
786 794 extern void dmu_fini(void);
787 795
788 796 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
789 797 uint64_t object, uint64_t offset, int len);
790 798 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
791 799 dmu_traverse_cb_t cb, void *arg);
792 800
793 801 int dmu_diff(const char *tosnap_name, const char *fromsnap_name,
794 802 struct vnode *vp, offset_t *offp);
795 803
796 804 /* CRC64 table */
797 805 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */
798 806 extern uint64_t zfs_crc64_table[256];
799 807
800 808 #ifdef __cplusplus
801 809 }
802 810 #endif
803 811
804 812 #endif /* _SYS_DMU_H */
↓ open down ↓ |
126 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX