Print this page
2619 asynchronous destruction of ZFS file systems
2747 SPA versioning with zfs feature flags
Reviewed by: Matt Ahrens <mahrens@delphix.com>
Reviewed by: George Wilson <gwilson@delphix.com>
Reviewed by: Richard Lowe <richlowe@richlowe.net>
Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com>
Approved by: Dan McDonald <danmcd@nexenta.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
|
↓ open down ↓ |
10 lines elided |
↑ open up ↑ |
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 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 - * Copyright (c) 2011 by Delphix. All rights reserved.
24 - */
25 -/*
24 + * Copyright (c) 2012 by Delphix. All rights reserved.
26 25 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
27 26 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
28 27 */
29 28
30 29 /* Portions Copyright 2010 Robert Milkowski */
31 30
32 31 #ifndef _SYS_DMU_H
33 32 #define _SYS_DMU_H
34 33
35 34 /*
36 35 * This file describes the interface that the DMU provides for its
37 36 * consumers.
38 37 *
39 38 * The DMU also interacts with the SPA. That interface is described in
40 39 * dmu_spa.h.
41 40 */
42 41
43 42 #include <sys/inttypes.h>
44 43 #include <sys/types.h>
45 44 #include <sys/param.h>
46 45 #include <sys/cred.h>
47 46 #include <sys/time.h>
48 47
49 48 #ifdef __cplusplus
50 49 extern "C" {
51 50 #endif
52 51
53 52 struct uio;
54 53 struct xuio;
55 54 struct page;
56 55 struct vnode;
57 56 struct spa;
58 57 struct zilog;
59 58 struct zio;
60 59 struct blkptr;
61 60 struct zap_cursor;
62 61 struct dsl_dataset;
63 62 struct dsl_pool;
64 63 struct dnode;
65 64 struct drr_begin;
66 65 struct drr_end;
67 66 struct zbookmark;
|
↓ open down ↓ |
32 lines elided |
↑ open up ↑ |
68 67 struct spa;
69 68 struct nvlist;
70 69 struct arc_buf;
71 70 struct zio_prop;
72 71 struct sa_handle;
73 72
74 73 typedef struct objset objset_t;
75 74 typedef struct dmu_tx dmu_tx_t;
76 75 typedef struct dsl_dir dsl_dir_t;
77 76
77 +typedef enum dmu_object_byteswap {
78 + DMU_BSWAP_UINT8,
79 + DMU_BSWAP_UINT16,
80 + DMU_BSWAP_UINT32,
81 + DMU_BSWAP_UINT64,
82 + DMU_BSWAP_ZAP,
83 + DMU_BSWAP_DNODE,
84 + DMU_BSWAP_OBJSET,
85 + DMU_BSWAP_ZNODE,
86 + DMU_BSWAP_OLDACL,
87 + DMU_BSWAP_ACL,
88 + /*
89 + * Allocating a new byteswap type number makes the on-disk format
90 + * incompatible with any other format that uses the same number.
91 + *
92 + * Data can usually be structured to work with one of the
93 + * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
94 + */
95 + DMU_BSWAP_NUMFUNCS
96 +} dmu_object_byteswap_t;
97 +
98 +#define DMU_OT_NEWTYPE 0x80
99 +#define DMU_OT_METADATA 0x40
100 +#define DMU_OT_BYTESWAP_MASK 0x3f
101 +
102 +/*
103 + * Defines a uint8_t object type. Object types specify if the data
104 + * in the object is metadata (boolean) and how to byteswap the data
105 + * (dmu_object_byteswap_t).
106 + */
107 +#define DMU_OT(byteswap, metadata) \
108 + (DMU_OT_NEWTYPE | \
109 + ((metadata) ? DMU_OT_METADATA : 0) | \
110 + ((byteswap) & DMU_OT_BYTESWAP_MASK))
111 +
112 +#define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \
113 + ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \
114 + (ot) < DMU_OT_NUMTYPES)
115 +
116 +#define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \
117 + ((ot) & DMU_OT_METADATA) : \
118 + dmu_ot[(ot)].ot_metadata)
119 +
120 +#define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \
121 + ((ot) & DMU_OT_BYTESWAP_MASK) : \
122 + dmu_ot[(ot)].ot_byteswap)
123 +
78 124 typedef enum dmu_object_type {
79 125 DMU_OT_NONE,
80 126 /* general: */
81 127 DMU_OT_OBJECT_DIRECTORY, /* ZAP */
82 128 DMU_OT_OBJECT_ARRAY, /* UINT64 */
83 129 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */
84 130 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */
85 131 DMU_OT_BPOBJ, /* UINT64 */
86 132 DMU_OT_BPOBJ_HDR, /* UINT64 */
87 133 /* spa: */
88 134 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */
89 135 DMU_OT_SPACE_MAP, /* UINT64 */
90 136 /* zil: */
91 137 DMU_OT_INTENT_LOG, /* UINT64 */
92 138 /* dmu: */
93 139 DMU_OT_DNODE, /* DNODE */
94 140 DMU_OT_OBJSET, /* OBJSET */
95 141 /* dsl: */
96 142 DMU_OT_DSL_DIR, /* UINT64 */
97 143 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */
98 144 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */
99 145 DMU_OT_DSL_PROPS, /* ZAP */
100 146 DMU_OT_DSL_DATASET, /* UINT64 */
101 147 /* zpl: */
102 148 DMU_OT_ZNODE, /* ZNODE */
103 149 DMU_OT_OLDACL, /* Old ACL */
104 150 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */
105 151 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */
106 152 DMU_OT_MASTER_NODE, /* ZAP */
107 153 DMU_OT_UNLINKED_SET, /* ZAP */
108 154 /* zvol: */
109 155 DMU_OT_ZVOL, /* UINT8 */
110 156 DMU_OT_ZVOL_PROP, /* ZAP */
111 157 /* other; for testing only! */
112 158 DMU_OT_PLAIN_OTHER, /* UINT8 */
113 159 DMU_OT_UINT64_OTHER, /* UINT64 */
114 160 DMU_OT_ZAP_OTHER, /* ZAP */
115 161 /* new object types: */
116 162 DMU_OT_ERROR_LOG, /* ZAP */
117 163 DMU_OT_SPA_HISTORY, /* UINT8 */
118 164 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */
119 165 DMU_OT_POOL_PROPS, /* ZAP */
120 166 DMU_OT_DSL_PERMS, /* ZAP */
121 167 DMU_OT_ACL, /* ACL */
122 168 DMU_OT_SYSACL, /* SYSACL */
123 169 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */
124 170 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */
125 171 DMU_OT_NEXT_CLONES, /* ZAP */
126 172 DMU_OT_SCAN_QUEUE, /* ZAP */
127 173 DMU_OT_USERGROUP_USED, /* ZAP */
128 174 DMU_OT_USERGROUP_QUOTA, /* ZAP */
129 175 DMU_OT_USERREFS, /* ZAP */
130 176 DMU_OT_DDT_ZAP, /* ZAP */
131 177 DMU_OT_DDT_STATS, /* ZAP */
|
↓ open down ↓ |
44 lines elided |
↑ open up ↑ |
132 178 DMU_OT_SA, /* System attr */
133 179 DMU_OT_SA_MASTER_NODE, /* ZAP */
134 180 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */
135 181 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */
136 182 DMU_OT_SCAN_XLATE, /* ZAP */
137 183 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */
138 184 DMU_OT_DEADLIST, /* ZAP */
139 185 DMU_OT_DEADLIST_HDR, /* UINT64 */
140 186 DMU_OT_DSL_CLONES, /* ZAP */
141 187 DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */
142 - DMU_OT_NUMTYPES
188 + /*
189 + * Do not allocate new object types here. Doing so makes the on-disk
190 + * format incompatible with any other format that uses the same object
191 + * type number.
192 + *
193 + * When creating an object which does not have one of the above types
194 + * use the DMU_OTN_* type with the correct byteswap and metadata
195 + * values.
196 + *
197 + * The DMU_OTN_* types do not have entries in the dmu_ot table,
198 + * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead
199 + * of indexing into dmu_ot directly (this works for both DMU_OT_* types
200 + * and DMU_OTN_* types).
201 + */
202 + DMU_OT_NUMTYPES,
203 +
204 + /*
205 + * Names for valid types declared with DMU_OT().
206 + */
207 + DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE),
208 + DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE),
209 + DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE),
210 + DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE),
211 + DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE),
212 + DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE),
213 + DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE),
214 + DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE),
215 + DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE),
216 + DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE),
143 217 } dmu_object_type_t;
144 218
145 219 typedef enum dmu_objset_type {
146 220 DMU_OST_NONE,
147 221 DMU_OST_META,
148 222 DMU_OST_ZFS,
149 223 DMU_OST_ZVOL,
150 224 DMU_OST_OTHER, /* For testing only! */
151 225 DMU_OST_ANY, /* Be careful! */
152 226 DMU_OST_NUMTYPES
153 227 } dmu_objset_type_t;
154 228
155 229 void byteswap_uint64_array(void *buf, size_t size);
156 230 void byteswap_uint32_array(void *buf, size_t size);
157 231 void byteswap_uint16_array(void *buf, size_t size);
158 232 void byteswap_uint8_array(void *buf, size_t size);
159 233 void zap_byteswap(void *buf, size_t size);
160 234 void zfs_oldacl_byteswap(void *buf, size_t size);
161 235 void zfs_acl_byteswap(void *buf, size_t size);
162 236 void zfs_znode_byteswap(void *buf, size_t size);
163 237
164 238 #define DS_FIND_SNAPSHOTS (1<<0)
165 239 #define DS_FIND_CHILDREN (1<<1)
166 240
167 241 /*
168 242 * The maximum number of bytes that can be accessed as part of one
169 243 * operation, including metadata.
170 244 */
171 245 #define DMU_MAX_ACCESS (10<<20) /* 10MB */
172 246 #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
173 247
174 248 #define DMU_USERUSED_OBJECT (-1ULL)
175 249 #define DMU_GROUPUSED_OBJECT (-2ULL)
176 250 #define DMU_DEADLIST_OBJECT (-3ULL)
177 251
178 252 /*
179 253 * artificial blkids for bonus buffer and spill blocks
180 254 */
181 255 #define DMU_BONUS_BLKID (-1ULL)
182 256 #define DMU_SPILL_BLKID (-2ULL)
183 257 /*
184 258 * Public routines to create, destroy, open, and close objsets.
185 259 */
186 260 int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
187 261 int dmu_objset_own(const char *name, dmu_objset_type_t type,
188 262 boolean_t readonly, void *tag, objset_t **osp);
189 263 void dmu_objset_rele(objset_t *os, void *tag);
190 264 void dmu_objset_disown(objset_t *os, void *tag);
191 265 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
192 266
193 267 int dmu_objset_evict_dbufs(objset_t *os);
194 268 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
195 269 void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
196 270 int dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin,
197 271 uint64_t flags);
198 272 int dmu_objset_destroy(const char *name, boolean_t defer);
199 273 int dmu_snapshots_destroy_nvl(struct nvlist *snaps, boolean_t defer, char *);
200 274 int dmu_objset_snapshot(char *fsname, char *snapname, char *tag,
201 275 struct nvlist *props, boolean_t recursive, boolean_t temporary, int fd);
202 276 int dmu_objset_rename(const char *name, const char *newname,
203 277 boolean_t recursive);
204 278 int dmu_objset_find(char *name, int func(const char *, void *), void *arg,
205 279 int flags);
206 280 void dmu_objset_byteswap(void *buf, size_t size);
207 281
208 282 typedef struct dmu_buf {
209 283 uint64_t db_object; /* object that this buffer is part of */
210 284 uint64_t db_offset; /* byte offset in this object */
211 285 uint64_t db_size; /* size of buffer in bytes */
|
↓ open down ↓ |
59 lines elided |
↑ open up ↑ |
212 286 void *db_data; /* data in buffer */
213 287 } dmu_buf_t;
214 288
215 289 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
216 290
217 291 /*
218 292 * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
219 293 */
220 294 #define DMU_POOL_DIRECTORY_OBJECT 1
221 295 #define DMU_POOL_CONFIG "config"
296 +#define DMU_POOL_FEATURES_FOR_WRITE "features_for_write"
297 +#define DMU_POOL_FEATURES_FOR_READ "features_for_read"
298 +#define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions"
222 299 #define DMU_POOL_ROOT_DATASET "root_dataset"
223 300 #define DMU_POOL_SYNC_BPOBJ "sync_bplist"
224 301 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub"
225 302 #define DMU_POOL_ERRLOG_LAST "errlog_last"
226 303 #define DMU_POOL_SPARES "spares"
227 304 #define DMU_POOL_DEFLATE "deflate"
228 305 #define DMU_POOL_HISTORY "history"
229 306 #define DMU_POOL_PROPS "pool_props"
230 307 #define DMU_POOL_L2CACHE "l2cache"
231 308 #define DMU_POOL_TMP_USERREFS "tmp_userrefs"
232 309 #define DMU_POOL_DDT "DDT-%s-%s-%s"
233 310 #define DMU_POOL_DDT_STATS "DDT-statistics"
234 311 #define DMU_POOL_CREATION_VERSION "creation_version"
235 312 #define DMU_POOL_SCAN "scan"
236 313 #define DMU_POOL_FREE_BPOBJ "free_bpobj"
314 +#define DMU_POOL_BPTREE_OBJ "bptree_obj"
237 315
238 316 /*
239 317 * Allocate an object from this objset. The range of object numbers
240 318 * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode.
241 319 *
242 320 * The transaction must be assigned to a txg. The newly allocated
243 321 * object will be "held" in the transaction (ie. you can modify the
244 322 * newly allocated object in this transaction).
245 323 *
246 324 * dmu_object_alloc() chooses an object and returns it in *objectp.
247 325 *
248 326 * dmu_object_claim() allocates a specific object number. If that
249 327 * number is already allocated, it fails and returns EEXIST.
250 328 *
251 329 * Return 0 on success, or ENOSPC or EEXIST as specified above.
252 330 */
253 331 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
254 332 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
255 333 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
256 334 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
257 335 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
258 336 int blocksize, dmu_object_type_t bonustype, int bonuslen);
259 337
260 338 /*
261 339 * Free an object from this objset.
262 340 *
263 341 * The object's data will be freed as well (ie. you don't need to call
264 342 * dmu_free(object, 0, -1, tx)).
265 343 *
266 344 * The object need not be held in the transaction.
267 345 *
268 346 * If there are any holds on this object's buffers (via dmu_buf_hold()),
269 347 * or tx holds on the object (via dmu_tx_hold_object()), you can not
270 348 * free it; it fails and returns EBUSY.
271 349 *
272 350 * If the object is not allocated, it fails and returns ENOENT.
273 351 *
274 352 * Return 0 on success, or EBUSY or ENOENT as specified above.
275 353 */
276 354 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
277 355
278 356 /*
279 357 * Find the next allocated or free object.
280 358 *
281 359 * The objectp parameter is in-out. It will be updated to be the next
282 360 * object which is allocated. Ignore objects which have not been
283 361 * modified since txg.
284 362 *
285 363 * XXX Can only be called on a objset with no dirty data.
286 364 *
287 365 * Returns 0 on success, or ENOENT if there are no more objects.
288 366 */
289 367 int dmu_object_next(objset_t *os, uint64_t *objectp,
290 368 boolean_t hole, uint64_t txg);
291 369
292 370 /*
293 371 * Set the data blocksize for an object.
294 372 *
295 373 * The object cannot have any blocks allcated beyond the first. If
296 374 * the first block is allocated already, the new size must be greater
297 375 * than the current block size. If these conditions are not met,
298 376 * ENOTSUP will be returned.
299 377 *
300 378 * Returns 0 on success, or EBUSY if there are any holds on the object
301 379 * contents, or ENOTSUP as described above.
302 380 */
303 381 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
304 382 int ibs, dmu_tx_t *tx);
305 383
306 384 /*
307 385 * Set the checksum property on a dnode. The new checksum algorithm will
308 386 * apply to all newly written blocks; existing blocks will not be affected.
309 387 */
310 388 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
311 389 dmu_tx_t *tx);
312 390
313 391 /*
314 392 * Set the compress property on a dnode. The new compression algorithm will
315 393 * apply to all newly written blocks; existing blocks will not be affected.
316 394 */
317 395 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
318 396 dmu_tx_t *tx);
319 397
320 398 /*
321 399 * Decide how to write a block: checksum, compression, number of copies, etc.
322 400 */
323 401 #define WP_NOFILL 0x1
324 402 #define WP_DMU_SYNC 0x2
325 403 #define WP_SPILL 0x4
326 404
327 405 void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp,
328 406 struct zio_prop *zp);
329 407 /*
330 408 * The bonus data is accessed more or less like a regular buffer.
331 409 * You must dmu_bonus_hold() to get the buffer, which will give you a
332 410 * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
333 411 * data. As with any normal buffer, you must call dmu_buf_read() to
334 412 * read db_data, dmu_buf_will_dirty() before modifying it, and the
335 413 * object must be held in an assigned transaction before calling
336 414 * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus
337 415 * buffer as well. You must release your hold with dmu_buf_rele().
338 416 */
339 417 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
340 418 int dmu_bonus_max(void);
341 419 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
342 420 int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
343 421 dmu_object_type_t dmu_get_bonustype(dmu_buf_t *);
344 422 int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
345 423
346 424 /*
347 425 * Special spill buffer support used by "SA" framework
348 426 */
349 427
350 428 int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
351 429 int dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags,
352 430 void *tag, dmu_buf_t **dbp);
353 431 int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
354 432
355 433 /*
356 434 * Obtain the DMU buffer from the specified object which contains the
357 435 * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so
358 436 * that it will remain in memory. You must release the hold with
359 437 * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your
360 438 * hold. You must have a hold on any dmu_buf_t* you pass to the DMU.
361 439 *
362 440 * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
363 441 * on the returned buffer before reading or writing the buffer's
364 442 * db_data. The comments for those routines describe what particular
365 443 * operations are valid after calling them.
366 444 *
367 445 * The object number must be a valid, allocated object number.
368 446 */
369 447 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
370 448 void *tag, dmu_buf_t **, int flags);
371 449 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
372 450 void dmu_buf_rele(dmu_buf_t *db, void *tag);
373 451 uint64_t dmu_buf_refcount(dmu_buf_t *db);
374 452
375 453 /*
376 454 * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
377 455 * range of an object. A pointer to an array of dmu_buf_t*'s is
378 456 * returned (in *dbpp).
379 457 *
380 458 * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
381 459 * frees the array. The hold on the array of buffers MUST be released
382 460 * with dmu_buf_rele_array. You can NOT release the hold on each buffer
383 461 * individually with dmu_buf_rele.
384 462 */
385 463 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
386 464 uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
387 465 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
388 466
389 467 /*
390 468 * Returns NULL on success, or the existing user ptr if it's already
391 469 * been set.
392 470 *
393 471 * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
394 472 *
395 473 * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
396 474 * will be set to db->db_data when you are allowed to access it. Note
397 475 * that db->db_data (the pointer) can change when you do dmu_buf_read(),
398 476 * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
399 477 * *user_data_ptr_ptr will be set to the new value when it changes.
400 478 *
401 479 * If non-NULL, pageout func will be called when this buffer is being
402 480 * excised from the cache, so that you can clean up the data structure
403 481 * pointed to by user_ptr.
404 482 *
405 483 * dmu_evict_user() will call the pageout func for all buffers in a
406 484 * objset with a given pageout func.
407 485 */
408 486 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
409 487 dmu_buf_evict_func_t *pageout_func);
410 488 /*
411 489 * set_user_ie is the same as set_user, but request immediate eviction
412 490 * when hold count goes to zero.
413 491 */
414 492 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
415 493 void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
416 494 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
417 495 void *user_ptr, void *user_data_ptr_ptr,
418 496 dmu_buf_evict_func_t *pageout_func);
419 497 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
420 498
421 499 /*
422 500 * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
423 501 */
424 502 void *dmu_buf_get_user(dmu_buf_t *db);
425 503
426 504 /*
427 505 * Indicate that you are going to modify the buffer's data (db_data).
428 506 *
429 507 * The transaction (tx) must be assigned to a txg (ie. you've called
430 508 * dmu_tx_assign()). The buffer's object must be held in the tx
431 509 * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
432 510 */
433 511 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
434 512
435 513 /*
436 514 * Tells if the given dbuf is freeable.
437 515 */
438 516 boolean_t dmu_buf_freeable(dmu_buf_t *);
439 517
440 518 /*
441 519 * You must create a transaction, then hold the objects which you will
442 520 * (or might) modify as part of this transaction. Then you must assign
443 521 * the transaction to a transaction group. Once the transaction has
444 522 * been assigned, you can modify buffers which belong to held objects as
445 523 * part of this transaction. You can't modify buffers before the
446 524 * transaction has been assigned; you can't modify buffers which don't
447 525 * belong to objects which this transaction holds; you can't hold
448 526 * objects once the transaction has been assigned. You may hold an
449 527 * object which you are going to free (with dmu_object_free()), but you
450 528 * don't have to.
451 529 *
452 530 * You can abort the transaction before it has been assigned.
453 531 *
454 532 * Note that you may hold buffers (with dmu_buf_hold) at any time,
455 533 * regardless of transaction state.
456 534 */
457 535
458 536 #define DMU_NEW_OBJECT (-1ULL)
459 537 #define DMU_OBJECT_END (-1ULL)
460 538
461 539 dmu_tx_t *dmu_tx_create(objset_t *os);
462 540 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
463 541 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
464 542 uint64_t len);
465 543 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
466 544 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
467 545 void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
468 546 void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
469 547 void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
470 548 void dmu_tx_abort(dmu_tx_t *tx);
471 549 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
472 550 void dmu_tx_wait(dmu_tx_t *tx);
473 551 void dmu_tx_commit(dmu_tx_t *tx);
474 552
475 553 /*
476 554 * To register a commit callback, dmu_tx_callback_register() must be called.
477 555 *
478 556 * dcb_data is a pointer to caller private data that is passed on as a
479 557 * callback parameter. The caller is responsible for properly allocating and
480 558 * freeing it.
481 559 *
482 560 * When registering a callback, the transaction must be already created, but
483 561 * it cannot be committed or aborted. It can be assigned to a txg or not.
484 562 *
485 563 * The callback will be called after the transaction has been safely written
486 564 * to stable storage and will also be called if the dmu_tx is aborted.
|
↓ open down ↓ |
240 lines elided |
↑ open up ↑ |
487 565 * If there is any error which prevents the transaction from being committed to
488 566 * disk, the callback will be called with a value of error != 0.
489 567 */
490 568 typedef void dmu_tx_callback_func_t(void *dcb_data, int error);
491 569
492 570 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
493 571 void *dcb_data);
494 572
495 573 /*
496 574 * Free up the data blocks for a defined range of a file. If size is
497 - * zero, the range from offset to end-of-file is freed.
575 + * -1, the range from offset to end-of-file is freed.
498 576 */
499 577 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
500 578 uint64_t size, dmu_tx_t *tx);
501 579 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
502 580 uint64_t size);
503 581 int dmu_free_object(objset_t *os, uint64_t object);
504 582
505 583 /*
506 584 * Convenience functions.
507 585 *
508 586 * Canfail routines will return 0 on success, or an errno if there is a
509 587 * nonrecoverable I/O error.
510 588 */
511 589 #define DMU_READ_PREFETCH 0 /* prefetch */
512 590 #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */
513 591 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
514 592 void *buf, uint32_t flags);
515 593 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
516 594 const void *buf, dmu_tx_t *tx);
517 595 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
518 596 dmu_tx_t *tx);
519 597 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
520 598 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
521 599 dmu_tx_t *tx);
522 600 int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
523 601 dmu_tx_t *tx);
524 602 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
525 603 uint64_t size, struct page *pp, dmu_tx_t *tx);
526 604 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
527 605 void dmu_return_arcbuf(struct arc_buf *buf);
528 606 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
529 607 dmu_tx_t *tx);
530 608 int dmu_xuio_init(struct xuio *uio, int niov);
531 609 void dmu_xuio_fini(struct xuio *uio);
532 610 int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
533 611 size_t n);
534 612 int dmu_xuio_cnt(struct xuio *uio);
535 613 struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
536 614 void dmu_xuio_clear(struct xuio *uio, int i);
537 615 void xuio_stat_wbuf_copied();
538 616 void xuio_stat_wbuf_nocopy();
539 617
540 618 extern int zfs_prefetch_disable;
541 619
542 620 /*
543 621 * Asynchronously try to read in the data.
544 622 */
545 623 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
546 624 uint64_t len);
547 625
548 626 typedef struct dmu_object_info {
549 627 /* All sizes are in bytes unless otherwise indicated. */
550 628 uint32_t doi_data_block_size;
551 629 uint32_t doi_metadata_block_size;
552 630 dmu_object_type_t doi_type;
553 631 dmu_object_type_t doi_bonus_type;
554 632 uint64_t doi_bonus_size;
555 633 uint8_t doi_indirection; /* 2 = dnode->indirect->data */
556 634 uint8_t doi_checksum;
|
↓ open down ↓ |
49 lines elided |
↑ open up ↑ |
557 635 uint8_t doi_compress;
558 636 uint8_t doi_pad[5];
559 637 uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */
560 638 uint64_t doi_max_offset;
561 639 uint64_t doi_fill_count; /* number of non-empty blocks */
562 640 } dmu_object_info_t;
563 641
564 642 typedef void arc_byteswap_func_t(void *buf, size_t size);
565 643
566 644 typedef struct dmu_object_type_info {
567 - arc_byteswap_func_t *ot_byteswap;
645 + dmu_object_byteswap_t ot_byteswap;
568 646 boolean_t ot_metadata;
569 647 char *ot_name;
570 648 } dmu_object_type_info_t;
571 649
650 +typedef struct dmu_object_byteswap_info {
651 + arc_byteswap_func_t *ob_func;
652 + char *ob_name;
653 +} dmu_object_byteswap_info_t;
654 +
572 655 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
656 +extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS];
573 657
574 658 /*
575 659 * Get information on a DMU object.
576 660 *
577 661 * Return 0 on success or ENOENT if object is not allocated.
578 662 *
579 663 * If doi is NULL, just indicates whether the object exists.
580 664 */
581 665 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
582 666 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
583 667 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
584 668 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
585 669 u_longlong_t *nblk512);
586 670
587 671 typedef struct dmu_objset_stats {
588 672 uint64_t dds_num_clones; /* number of clones of this */
589 673 uint64_t dds_creation_txg;
590 674 uint64_t dds_guid;
591 675 dmu_objset_type_t dds_type;
592 676 uint8_t dds_is_snapshot;
593 677 uint8_t dds_inconsistent;
594 678 char dds_origin[MAXNAMELEN];
595 679 } dmu_objset_stats_t;
596 680
597 681 /*
598 682 * Get stats on a dataset.
599 683 */
600 684 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
601 685
602 686 /*
603 687 * Add entries to the nvlist for all the objset's properties. See
604 688 * zfs_prop_table[] and zfs(1m) for details on the properties.
605 689 */
606 690 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
607 691
608 692 /*
609 693 * Get the space usage statistics for statvfs().
610 694 *
611 695 * refdbytes is the amount of space "referenced" by this objset.
612 696 * availbytes is the amount of space available to this objset, taking
613 697 * into account quotas & reservations, assuming that no other objsets
614 698 * use the space first. These values correspond to the 'referenced' and
615 699 * 'available' properties, described in the zfs(1m) manpage.
616 700 *
617 701 * usedobjs and availobjs are the number of objects currently allocated,
618 702 * and available.
619 703 */
620 704 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
621 705 uint64_t *usedobjsp, uint64_t *availobjsp);
622 706
623 707 /*
624 708 * The fsid_guid is a 56-bit ID that can change to avoid collisions.
625 709 * (Contrast with the ds_guid which is a 64-bit ID that will never
626 710 * change, so there is a small probability that it will collide.)
627 711 */
628 712 uint64_t dmu_objset_fsid_guid(objset_t *os);
629 713
630 714 /*
631 715 * Get the [cm]time for an objset's snapshot dir
632 716 */
633 717 timestruc_t dmu_objset_snap_cmtime(objset_t *os);
634 718
635 719 int dmu_objset_is_snapshot(objset_t *os);
636 720
637 721 extern struct spa *dmu_objset_spa(objset_t *os);
638 722 extern struct zilog *dmu_objset_zil(objset_t *os);
639 723 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
640 724 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
641 725 extern void dmu_objset_name(objset_t *os, char *buf);
642 726 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
643 727 extern uint64_t dmu_objset_id(objset_t *os);
644 728 extern uint64_t dmu_objset_syncprop(objset_t *os);
645 729 extern uint64_t dmu_objset_logbias(objset_t *os);
646 730 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
647 731 uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
648 732 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
649 733 int maxlen, boolean_t *conflict);
650 734 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
651 735 uint64_t *idp, uint64_t *offp);
652 736
653 737 typedef int objset_used_cb_t(dmu_object_type_t bonustype,
654 738 void *bonus, uint64_t *userp, uint64_t *groupp);
655 739 extern void dmu_objset_register_type(dmu_objset_type_t ost,
656 740 objset_used_cb_t *cb);
657 741 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
658 742 extern void *dmu_objset_get_user(objset_t *os);
659 743
660 744 /*
661 745 * Return the txg number for the given assigned transaction.
662 746 */
663 747 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
664 748
665 749 /*
666 750 * Synchronous write.
667 751 * If a parent zio is provided this function initiates a write on the
668 752 * provided buffer as a child of the parent zio.
669 753 * In the absence of a parent zio, the write is completed synchronously.
670 754 * At write completion, blk is filled with the bp of the written block.
671 755 * Note that while the data covered by this function will be on stable
672 756 * storage when the write completes this new data does not become a
673 757 * permanent part of the file until the associated transaction commits.
674 758 */
675 759
676 760 /*
677 761 * {zfs,zvol,ztest}_get_done() args
678 762 */
679 763 typedef struct zgd {
680 764 struct zilog *zgd_zilog;
681 765 struct blkptr *zgd_bp;
682 766 dmu_buf_t *zgd_db;
683 767 struct rl *zgd_rl;
684 768 void *zgd_private;
685 769 } zgd_t;
686 770
687 771 typedef void dmu_sync_cb_t(zgd_t *arg, int error);
688 772 int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
689 773
690 774 /*
691 775 * Find the next hole or data block in file starting at *off
692 776 * Return found offset in *off. Return ESRCH for end of file.
693 777 */
694 778 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
695 779 uint64_t *off);
696 780
697 781 /*
698 782 * Initial setup and final teardown.
699 783 */
700 784 extern void dmu_init(void);
701 785 extern void dmu_fini(void);
702 786
703 787 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
704 788 uint64_t object, uint64_t offset, int len);
705 789 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
706 790 dmu_traverse_cb_t cb, void *arg);
707 791
708 792 int dmu_send(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
709 793 int outfd, struct vnode *vp, offset_t *off);
710 794 int dmu_send_estimate(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorign,
711 795 uint64_t *sizep);
712 796
713 797 typedef struct dmu_recv_cookie {
714 798 /*
715 799 * This structure is opaque!
716 800 *
717 801 * If logical and real are different, we are recving the stream
718 802 * into the "real" temporary clone, and then switching it with
719 803 * the "logical" target.
720 804 */
721 805 struct dsl_dataset *drc_logical_ds;
722 806 struct dsl_dataset *drc_real_ds;
723 807 struct drr_begin *drc_drrb;
724 808 char *drc_tosnap;
725 809 char *drc_top_ds;
726 810 boolean_t drc_newfs;
727 811 boolean_t drc_force;
728 812 struct avl_tree *drc_guid_to_ds_map;
729 813 } dmu_recv_cookie_t;
730 814
731 815 int dmu_recv_begin(char *tofs, char *tosnap, char *topds, struct drr_begin *,
732 816 boolean_t force, objset_t *origin, dmu_recv_cookie_t *);
733 817 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp,
734 818 int cleanup_fd, uint64_t *action_handlep);
735 819 int dmu_recv_end(dmu_recv_cookie_t *drc);
736 820
737 821 int dmu_diff(objset_t *tosnap, objset_t *fromsnap, struct vnode *vp,
738 822 offset_t *off);
739 823
740 824 /* CRC64 table */
741 825 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */
742 826 extern uint64_t zfs_crc64_table[256];
743 827
744 828 #ifdef __cplusplus
745 829 }
746 830 #endif
747 831
748 832 #endif /* _SYS_DMU_H */
|
↓ open down ↓ |
166 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX