]> git.proxmox.com Git - mirror_zfs.git/blob - include/sys/dmu.h
67f4be1c2438814018eedb2f08795481649424b2
[mirror_zfs.git] / include / sys / dmu.h
1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21 /*
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
26 * Copyright 2014 HybridCluster. All rights reserved.
27 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
28 * Copyright 2013 Saso Kiselkov. All rights reserved.
29 */
30
31 /* Portions Copyright 2010 Robert Milkowski */
32
33 #ifndef _SYS_DMU_H
34 #define _SYS_DMU_H
35
36 /*
37 * This file describes the interface that the DMU provides for its
38 * consumers.
39 *
40 * The DMU also interacts with the SPA. That interface is described in
41 * dmu_spa.h.
42 */
43
44 #include <sys/zfs_context.h>
45 #include <sys/inttypes.h>
46 #include <sys/cred.h>
47 #include <sys/fs/zfs.h>
48 #include <sys/zio_compress.h>
49 #include <sys/zio_priority.h>
50 #include <sys/uio.h>
51
52 #ifdef __cplusplus
53 extern "C" {
54 #endif
55
56 struct page;
57 struct vnode;
58 struct spa;
59 struct zilog;
60 struct zio;
61 struct blkptr;
62 struct zap_cursor;
63 struct dsl_dataset;
64 struct dsl_pool;
65 struct dnode;
66 struct drr_begin;
67 struct drr_end;
68 struct zbookmark_phys;
69 struct spa;
70 struct nvlist;
71 struct arc_buf;
72 struct zio_prop;
73 struct sa_handle;
74 struct dsl_crypto_params;
75
76 typedef struct objset objset_t;
77 typedef struct dmu_tx dmu_tx_t;
78 typedef struct dsl_dir dsl_dir_t;
79 typedef struct dnode dnode_t;
80
81 typedef enum dmu_object_byteswap {
82 DMU_BSWAP_UINT8,
83 DMU_BSWAP_UINT16,
84 DMU_BSWAP_UINT32,
85 DMU_BSWAP_UINT64,
86 DMU_BSWAP_ZAP,
87 DMU_BSWAP_DNODE,
88 DMU_BSWAP_OBJSET,
89 DMU_BSWAP_ZNODE,
90 DMU_BSWAP_OLDACL,
91 DMU_BSWAP_ACL,
92 /*
93 * Allocating a new byteswap type number makes the on-disk format
94 * incompatible with any other format that uses the same number.
95 *
96 * Data can usually be structured to work with one of the
97 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
98 */
99 DMU_BSWAP_NUMFUNCS
100 } dmu_object_byteswap_t;
101
102 #define DMU_OT_NEWTYPE 0x80
103 #define DMU_OT_METADATA 0x40
104 #define DMU_OT_ENCRYPTED 0x20
105 #define DMU_OT_BYTESWAP_MASK 0x1f
106
107 /*
108 * Defines a uint8_t object type. Object types specify if the data
109 * in the object is metadata (boolean) and how to byteswap the data
110 * (dmu_object_byteswap_t). All of the types created by this method
111 * are cached in the dbuf metadata cache.
112 */
113 #define DMU_OT(byteswap, metadata, encrypted) \
114 (DMU_OT_NEWTYPE | \
115 ((metadata) ? DMU_OT_METADATA : 0) | \
116 ((encrypted) ? DMU_OT_ENCRYPTED : 0) | \
117 ((byteswap) & DMU_OT_BYTESWAP_MASK))
118
119 #define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \
120 ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \
121 (ot) < DMU_OT_NUMTYPES)
122
123 #define DMU_OT_IS_METADATA_CACHED(ot) (((ot) & DMU_OT_NEWTYPE) ? \
124 B_TRUE : dmu_ot[(ot)].ot_dbuf_metadata_cache)
125
126 /*
127 * MDB doesn't have dmu_ot; it defines these macros itself.
128 */
129 #ifndef ZFS_MDB
130 #define DMU_OT_IS_METADATA_IMPL(ot) (dmu_ot[ot].ot_metadata)
131 #define DMU_OT_IS_ENCRYPTED_IMPL(ot) (dmu_ot[ot].ot_encrypt)
132 #define DMU_OT_BYTESWAP_IMPL(ot) (dmu_ot[ot].ot_byteswap)
133 #endif
134
135 #define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \
136 ((ot) & DMU_OT_METADATA) : \
137 DMU_OT_IS_METADATA_IMPL(ot))
138
139 #define DMU_OT_IS_ENCRYPTED(ot) (((ot) & DMU_OT_NEWTYPE) ? \
140 ((ot) & DMU_OT_ENCRYPTED) : \
141 DMU_OT_IS_ENCRYPTED_IMPL(ot))
142
143 /*
144 * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't
145 * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill
146 * is repurposed for embedded BPs.
147 */
148 #define DMU_OT_HAS_FILL(ot) \
149 ((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET)
150
151 #define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \
152 ((ot) & DMU_OT_BYTESWAP_MASK) : \
153 DMU_OT_BYTESWAP_IMPL(ot))
154
155 typedef enum dmu_object_type {
156 DMU_OT_NONE,
157 /* general: */
158 DMU_OT_OBJECT_DIRECTORY, /* ZAP */
159 DMU_OT_OBJECT_ARRAY, /* UINT64 */
160 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */
161 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */
162 DMU_OT_BPOBJ, /* UINT64 */
163 DMU_OT_BPOBJ_HDR, /* UINT64 */
164 /* spa: */
165 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */
166 DMU_OT_SPACE_MAP, /* UINT64 */
167 /* zil: */
168 DMU_OT_INTENT_LOG, /* UINT64 */
169 /* dmu: */
170 DMU_OT_DNODE, /* DNODE */
171 DMU_OT_OBJSET, /* OBJSET */
172 /* dsl: */
173 DMU_OT_DSL_DIR, /* UINT64 */
174 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */
175 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */
176 DMU_OT_DSL_PROPS, /* ZAP */
177 DMU_OT_DSL_DATASET, /* UINT64 */
178 /* zpl: */
179 DMU_OT_ZNODE, /* ZNODE */
180 DMU_OT_OLDACL, /* Old ACL */
181 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */
182 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */
183 DMU_OT_MASTER_NODE, /* ZAP */
184 DMU_OT_UNLINKED_SET, /* ZAP */
185 /* zvol: */
186 DMU_OT_ZVOL, /* UINT8 */
187 DMU_OT_ZVOL_PROP, /* ZAP */
188 /* other; for testing only! */
189 DMU_OT_PLAIN_OTHER, /* UINT8 */
190 DMU_OT_UINT64_OTHER, /* UINT64 */
191 DMU_OT_ZAP_OTHER, /* ZAP */
192 /* new object types: */
193 DMU_OT_ERROR_LOG, /* ZAP */
194 DMU_OT_SPA_HISTORY, /* UINT8 */
195 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */
196 DMU_OT_POOL_PROPS, /* ZAP */
197 DMU_OT_DSL_PERMS, /* ZAP */
198 DMU_OT_ACL, /* ACL */
199 DMU_OT_SYSACL, /* SYSACL */
200 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */
201 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */
202 DMU_OT_NEXT_CLONES, /* ZAP */
203 DMU_OT_SCAN_QUEUE, /* ZAP */
204 DMU_OT_USERGROUP_USED, /* ZAP */
205 DMU_OT_USERGROUP_QUOTA, /* ZAP */
206 DMU_OT_USERREFS, /* ZAP */
207 DMU_OT_DDT_ZAP, /* ZAP */
208 DMU_OT_DDT_STATS, /* ZAP */
209 DMU_OT_SA, /* System attr */
210 DMU_OT_SA_MASTER_NODE, /* ZAP */
211 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */
212 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */
213 DMU_OT_SCAN_XLATE, /* ZAP */
214 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */
215 DMU_OT_DEADLIST, /* ZAP */
216 DMU_OT_DEADLIST_HDR, /* UINT64 */
217 DMU_OT_DSL_CLONES, /* ZAP */
218 DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */
219 /*
220 * Do not allocate new object types here. Doing so makes the on-disk
221 * format incompatible with any other format that uses the same object
222 * type number.
223 *
224 * When creating an object which does not have one of the above types
225 * use the DMU_OTN_* type with the correct byteswap and metadata
226 * values.
227 *
228 * The DMU_OTN_* types do not have entries in the dmu_ot table,
229 * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead
230 * of indexing into dmu_ot directly (this works for both DMU_OT_* types
231 * and DMU_OTN_* types).
232 */
233 DMU_OT_NUMTYPES,
234
235 /*
236 * Names for valid types declared with DMU_OT().
237 */
238 DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_FALSE),
239 DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_FALSE),
240 DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_FALSE),
241 DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_FALSE),
242 DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_FALSE),
243 DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_FALSE),
244 DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_FALSE),
245 DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_FALSE),
246 DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_FALSE),
247 DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_FALSE),
248
249 DMU_OTN_UINT8_ENC_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_TRUE),
250 DMU_OTN_UINT8_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_TRUE),
251 DMU_OTN_UINT16_ENC_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_TRUE),
252 DMU_OTN_UINT16_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_TRUE),
253 DMU_OTN_UINT32_ENC_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_TRUE),
254 DMU_OTN_UINT32_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_TRUE),
255 DMU_OTN_UINT64_ENC_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_TRUE),
256 DMU_OTN_UINT64_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_TRUE),
257 DMU_OTN_ZAP_ENC_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_TRUE),
258 DMU_OTN_ZAP_ENC_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_TRUE),
259 } dmu_object_type_t;
260
261 /*
262 * These flags are intended to be used to specify the "txg_how"
263 * parameter when calling the dmu_tx_assign() function. See the comment
264 * above dmu_tx_assign() for more details on the meaning of these flags.
265 */
266 #define TXG_NOWAIT (0ULL)
267 #define TXG_WAIT (1ULL<<0)
268 #define TXG_NOTHROTTLE (1ULL<<1)
269
270 void byteswap_uint64_array(void *buf, size_t size);
271 void byteswap_uint32_array(void *buf, size_t size);
272 void byteswap_uint16_array(void *buf, size_t size);
273 void byteswap_uint8_array(void *buf, size_t size);
274 void zap_byteswap(void *buf, size_t size);
275 void zfs_oldacl_byteswap(void *buf, size_t size);
276 void zfs_acl_byteswap(void *buf, size_t size);
277 void zfs_znode_byteswap(void *buf, size_t size);
278
279 #define DS_FIND_SNAPSHOTS (1<<0)
280 #define DS_FIND_CHILDREN (1<<1)
281 #define DS_FIND_SERIALIZE (1<<2)
282
283 /*
284 * The maximum number of bytes that can be accessed as part of one
285 * operation, including metadata.
286 */
287 #define DMU_MAX_ACCESS (64 * 1024 * 1024) /* 64MB */
288 #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
289
290 #define DMU_USERUSED_OBJECT (-1ULL)
291 #define DMU_GROUPUSED_OBJECT (-2ULL)
292 #define DMU_PROJECTUSED_OBJECT (-3ULL)
293
294 /*
295 * Zap prefix for object accounting in DMU_{USER,GROUP,PROJECT}USED_OBJECT.
296 */
297 #define DMU_OBJACCT_PREFIX "obj-"
298 #define DMU_OBJACCT_PREFIX_LEN 4
299
300 /*
301 * artificial blkids for bonus buffer and spill blocks
302 */
303 #define DMU_BONUS_BLKID (-1ULL)
304 #define DMU_SPILL_BLKID (-2ULL)
305
306 /*
307 * Public routines to create, destroy, open, and close objsets.
308 */
309 typedef void dmu_objset_create_sync_func_t(objset_t *os, void *arg,
310 cred_t *cr, dmu_tx_t *tx);
311
312 int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
313 int dmu_objset_own(const char *name, dmu_objset_type_t type,
314 boolean_t readonly, boolean_t key_required, void *tag, objset_t **osp);
315 void dmu_objset_rele(objset_t *os, void *tag);
316 void dmu_objset_disown(objset_t *os, boolean_t key_required, void *tag);
317 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
318
319 void dmu_objset_evict_dbufs(objset_t *os);
320 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
321 struct dsl_crypto_params *dcp, dmu_objset_create_sync_func_t func,
322 void *arg);
323 int dmu_objset_clone(const char *name, const char *origin);
324 int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer,
325 struct nvlist *errlist);
326 int dmu_objset_snapshot_one(const char *fsname, const char *snapname);
327 int dmu_objset_snapshot_tmp(const char *, const char *, int);
328 int dmu_objset_find(char *name, int func(const char *, void *), void *arg,
329 int flags);
330 void dmu_objset_byteswap(void *buf, size_t size);
331 int dsl_dataset_rename_snapshot(const char *fsname,
332 const char *oldsnapname, const char *newsnapname, boolean_t recursive);
333 int dmu_objset_remap_indirects(const char *fsname);
334
335 typedef struct dmu_buf {
336 uint64_t db_object; /* object that this buffer is part of */
337 uint64_t db_offset; /* byte offset in this object */
338 uint64_t db_size; /* size of buffer in bytes */
339 void *db_data; /* data in buffer */
340 } dmu_buf_t;
341
342 /*
343 * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
344 */
345 #define DMU_POOL_DIRECTORY_OBJECT 1
346 #define DMU_POOL_CONFIG "config"
347 #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write"
348 #define DMU_POOL_FEATURES_FOR_READ "features_for_read"
349 #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions"
350 #define DMU_POOL_FEATURE_ENABLED_TXG "feature_enabled_txg"
351 #define DMU_POOL_ROOT_DATASET "root_dataset"
352 #define DMU_POOL_SYNC_BPOBJ "sync_bplist"
353 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub"
354 #define DMU_POOL_ERRLOG_LAST "errlog_last"
355 #define DMU_POOL_SPARES "spares"
356 #define DMU_POOL_DEFLATE "deflate"
357 #define DMU_POOL_HISTORY "history"
358 #define DMU_POOL_PROPS "pool_props"
359 #define DMU_POOL_L2CACHE "l2cache"
360 #define DMU_POOL_TMP_USERREFS "tmp_userrefs"
361 #define DMU_POOL_DDT "DDT-%s-%s-%s"
362 #define DMU_POOL_DDT_STATS "DDT-statistics"
363 #define DMU_POOL_CREATION_VERSION "creation_version"
364 #define DMU_POOL_SCAN "scan"
365 #define DMU_POOL_FREE_BPOBJ "free_bpobj"
366 #define DMU_POOL_BPTREE_OBJ "bptree_obj"
367 #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj"
368 #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt"
369 #define DMU_POOL_VDEV_ZAP_MAP "com.delphix:vdev_zap_map"
370 #define DMU_POOL_REMOVING "com.delphix:removing"
371 #define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj"
372 #define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect"
373 #define DMU_POOL_ZPOOL_CHECKPOINT "com.delphix:zpool_checkpoint"
374
375 /*
376 * Allocate an object from this objset. The range of object numbers
377 * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode.
378 *
379 * The transaction must be assigned to a txg. The newly allocated
380 * object will be "held" in the transaction (ie. you can modify the
381 * newly allocated object in this transaction).
382 *
383 * dmu_object_alloc() chooses an object and returns it in *objectp.
384 *
385 * dmu_object_claim() allocates a specific object number. If that
386 * number is already allocated, it fails and returns EEXIST.
387 *
388 * Return 0 on success, or ENOSPC or EEXIST as specified above.
389 */
390 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
391 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
392 uint64_t dmu_object_alloc_ibs(objset_t *os, dmu_object_type_t ot, int blocksize,
393 int indirect_blockshift,
394 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
395 uint64_t dmu_object_alloc_dnsize(objset_t *os, dmu_object_type_t ot,
396 int blocksize, dmu_object_type_t bonus_type, int bonus_len,
397 int dnodesize, dmu_tx_t *tx);
398 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
399 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
400 int dmu_object_claim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot,
401 int blocksize, dmu_object_type_t bonus_type, int bonus_len,
402 int dnodesize, dmu_tx_t *tx);
403 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
404 int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp);
405 int dmu_object_reclaim_dnsize(objset_t *os, uint64_t object,
406 dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype,
407 int bonuslen, int dnodesize, dmu_tx_t *txp);
408
409 /*
410 * Free an object from this objset.
411 *
412 * The object's data will be freed as well (ie. you don't need to call
413 * dmu_free(object, 0, -1, tx)).
414 *
415 * The object need not be held in the transaction.
416 *
417 * If there are any holds on this object's buffers (via dmu_buf_hold()),
418 * or tx holds on the object (via dmu_tx_hold_object()), you can not
419 * free it; it fails and returns EBUSY.
420 *
421 * If the object is not allocated, it fails and returns ENOENT.
422 *
423 * Return 0 on success, or EBUSY or ENOENT as specified above.
424 */
425 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
426
427 /*
428 * Find the next allocated or free object.
429 *
430 * The objectp parameter is in-out. It will be updated to be the next
431 * object which is allocated. Ignore objects which have not been
432 * modified since txg.
433 *
434 * XXX Can only be called on a objset with no dirty data.
435 *
436 * Returns 0 on success, or ENOENT if there are no more objects.
437 */
438 int dmu_object_next(objset_t *os, uint64_t *objectp,
439 boolean_t hole, uint64_t txg);
440
441 /*
442 * Set the number of levels on a dnode. nlevels must be greater than the
443 * current number of levels or an EINVAL will be returned.
444 */
445 int dmu_object_set_nlevels(objset_t *os, uint64_t object, int nlevels,
446 dmu_tx_t *tx);
447
448 /*
449 * Set the data blocksize for an object.
450 *
451 * The object cannot have any blocks allcated beyond the first. If
452 * the first block is allocated already, the new size must be greater
453 * than the current block size. If these conditions are not met,
454 * ENOTSUP will be returned.
455 *
456 * Returns 0 on success, or EBUSY if there are any holds on the object
457 * contents, or ENOTSUP as described above.
458 */
459 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
460 int ibs, dmu_tx_t *tx);
461
462 /*
463 * Manually set the maxblkid on a dnode. This will adjust nlevels accordingly
464 * to accommodate the change.
465 */
466 int dmu_object_set_maxblkid(objset_t *os, uint64_t object, uint64_t maxblkid,
467 dmu_tx_t *tx);
468
469 /*
470 * Set the checksum property on a dnode. The new checksum algorithm will
471 * apply to all newly written blocks; existing blocks will not be affected.
472 */
473 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
474 dmu_tx_t *tx);
475
476 /*
477 * Set the compress property on a dnode. The new compression algorithm will
478 * apply to all newly written blocks; existing blocks will not be affected.
479 */
480 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
481 dmu_tx_t *tx);
482
483
484 int dmu_object_remap_indirects(objset_t *os, uint64_t object, uint64_t txg);
485
486 void dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset,
487 void *data, uint8_t etype, uint8_t comp, int uncompressed_size,
488 int compressed_size, int byteorder, dmu_tx_t *tx);
489
490 /*
491 * Decide how to write a block: checksum, compression, number of copies, etc.
492 */
493 #define WP_NOFILL 0x1
494 #define WP_DMU_SYNC 0x2
495 #define WP_SPILL 0x4
496
497 void dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp,
498 struct zio_prop *zp);
499
500 /*
501 * The bonus data is accessed more or less like a regular buffer.
502 * You must dmu_bonus_hold() to get the buffer, which will give you a
503 * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
504 * data. As with any normal buffer, you must call dmu_buf_will_dirty()
505 * before modifying it, and the
506 * object must be held in an assigned transaction before calling
507 * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus
508 * buffer as well. You must release what you hold with dmu_buf_rele().
509 *
510 * Returns ENOENT, EIO, or 0.
511 */
512 int dmu_bonus_hold_impl(objset_t *os, uint64_t object, void *tag,
513 uint32_t flags, dmu_buf_t **dbp);
514 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
515 int dmu_bonus_max(void);
516 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
517 int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
518 dmu_object_type_t dmu_get_bonustype(dmu_buf_t *);
519 int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
520
521 /*
522 * Special spill buffer support used by "SA" framework
523 */
524
525 int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, uint32_t flags, void *tag,
526 dmu_buf_t **dbp);
527 int dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags,
528 void *tag, dmu_buf_t **dbp);
529 int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
530
531 /*
532 * Obtain the DMU buffer from the specified object which contains the
533 * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so
534 * that it will remain in memory. You must release the hold with
535 * dmu_buf_rele(). You must not access the dmu_buf_t after releasing
536 * what you hold. You must have a hold on any dmu_buf_t* you pass to the DMU.
537 *
538 * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
539 * on the returned buffer before reading or writing the buffer's
540 * db_data. The comments for those routines describe what particular
541 * operations are valid after calling them.
542 *
543 * The object number must be a valid, allocated object number.
544 */
545 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
546 void *tag, dmu_buf_t **, int flags);
547 int dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset,
548 void *tag, dmu_buf_t **dbp, int flags);
549
550 /*
551 * Add a reference to a dmu buffer that has already been held via
552 * dmu_buf_hold() in the current context.
553 */
554 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
555
556 /*
557 * Attempt to add a reference to a dmu buffer that is in an unknown state,
558 * using a pointer that may have been invalidated by eviction processing.
559 * The request will succeed if the passed in dbuf still represents the
560 * same os/object/blkid, is ineligible for eviction, and has at least
561 * one hold by a user other than the syncer.
562 */
563 boolean_t dmu_buf_try_add_ref(dmu_buf_t *, objset_t *os, uint64_t object,
564 uint64_t blkid, void *tag);
565
566 void dmu_buf_rele(dmu_buf_t *db, void *tag);
567 uint64_t dmu_buf_refcount(dmu_buf_t *db);
568 uint64_t dmu_buf_user_refcount(dmu_buf_t *db);
569
570 /*
571 * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
572 * range of an object. A pointer to an array of dmu_buf_t*'s is
573 * returned (in *dbpp).
574 *
575 * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
576 * frees the array. The hold on the array of buffers MUST be released
577 * with dmu_buf_rele_array. You can NOT release the hold on each buffer
578 * individually with dmu_buf_rele.
579 */
580 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
581 uint64_t length, boolean_t read, void *tag,
582 int *numbufsp, dmu_buf_t ***dbpp);
583 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
584
585 typedef void dmu_buf_evict_func_t(void *user_ptr);
586
587 /*
588 * A DMU buffer user object may be associated with a dbuf for the
589 * duration of its lifetime. This allows the user of a dbuf (client)
590 * to attach private data to a dbuf (e.g. in-core only data such as a
591 * dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified
592 * when that dbuf has been evicted. Clients typically respond to the
593 * eviction notification by freeing their private data, thus ensuring
594 * the same lifetime for both dbuf and private data.
595 *
596 * The mapping from a dmu_buf_user_t to any client private data is the
597 * client's responsibility. All current consumers of the API with private
598 * data embed a dmu_buf_user_t as the first member of the structure for
599 * their private data. This allows conversions between the two types
600 * with a simple cast. Since the DMU buf user API never needs access
601 * to the private data, other strategies can be employed if necessary
602 * or convenient for the client (e.g. using container_of() to do the
603 * conversion for private data that cannot have the dmu_buf_user_t as
604 * its first member).
605 *
606 * Eviction callbacks are executed without the dbuf mutex held or any
607 * other type of mechanism to guarantee that the dbuf is still available.
608 * For this reason, users must assume the dbuf has already been freed
609 * and not reference the dbuf from the callback context.
610 *
611 * Users requesting "immediate eviction" are notified as soon as the dbuf
612 * is only referenced by dirty records (dirties == holds). Otherwise the
613 * notification occurs after eviction processing for the dbuf begins.
614 */
615 typedef struct dmu_buf_user {
616 /*
617 * Asynchronous user eviction callback state.
618 */
619 taskq_ent_t dbu_tqent;
620
621 /*
622 * This instance's eviction function pointers.
623 *
624 * dbu_evict_func_sync is called synchronously and then
625 * dbu_evict_func_async is executed asynchronously on a taskq.
626 */
627 dmu_buf_evict_func_t *dbu_evict_func_sync;
628 dmu_buf_evict_func_t *dbu_evict_func_async;
629 #ifdef ZFS_DEBUG
630 /*
631 * Pointer to user's dbuf pointer. NULL for clients that do
632 * not associate a dbuf with their user data.
633 *
634 * The dbuf pointer is cleared upon eviction so as to catch
635 * use-after-evict bugs in clients.
636 */
637 dmu_buf_t **dbu_clear_on_evict_dbufp;
638 #endif
639 } dmu_buf_user_t;
640
641 /*
642 * Initialize the given dmu_buf_user_t instance with the eviction function
643 * evict_func, to be called when the user is evicted.
644 *
645 * NOTE: This function should only be called once on a given dmu_buf_user_t.
646 * To allow enforcement of this, dbu must already be zeroed on entry.
647 */
648 /*ARGSUSED*/
649 static inline void
650 dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func_sync,
651 dmu_buf_evict_func_t *evict_func_async, dmu_buf_t **clear_on_evict_dbufp)
652 {
653 ASSERT(dbu->dbu_evict_func_sync == NULL);
654 ASSERT(dbu->dbu_evict_func_async == NULL);
655
656 /* must have at least one evict func */
657 IMPLY(evict_func_sync == NULL, evict_func_async != NULL);
658 dbu->dbu_evict_func_sync = evict_func_sync;
659 dbu->dbu_evict_func_async = evict_func_async;
660 taskq_init_ent(&dbu->dbu_tqent);
661 #ifdef ZFS_DEBUG
662 dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp;
663 #endif
664 }
665
666 /*
667 * Attach user data to a dbuf and mark it for normal (when the dbuf's
668 * data is cleared or its reference count goes to zero) eviction processing.
669 *
670 * Returns NULL on success, or the existing user if another user currently
671 * owns the buffer.
672 */
673 void *dmu_buf_set_user(dmu_buf_t *db, dmu_buf_user_t *user);
674
675 /*
676 * Attach user data to a dbuf and mark it for immediate (its dirty and
677 * reference counts are equal) eviction processing.
678 *
679 * Returns NULL on success, or the existing user if another user currently
680 * owns the buffer.
681 */
682 void *dmu_buf_set_user_ie(dmu_buf_t *db, dmu_buf_user_t *user);
683
684 /*
685 * Replace the current user of a dbuf.
686 *
687 * If given the current user of a dbuf, replaces the dbuf's user with
688 * "new_user" and returns the user data pointer that was replaced.
689 * Otherwise returns the current, and unmodified, dbuf user pointer.
690 */
691 void *dmu_buf_replace_user(dmu_buf_t *db,
692 dmu_buf_user_t *old_user, dmu_buf_user_t *new_user);
693
694 /*
695 * Remove the specified user data for a DMU buffer.
696 *
697 * Returns the user that was removed on success, or the current user if
698 * another user currently owns the buffer.
699 */
700 void *dmu_buf_remove_user(dmu_buf_t *db, dmu_buf_user_t *user);
701
702 /*
703 * Returns the user data (dmu_buf_user_t *) associated with this dbuf.
704 */
705 void *dmu_buf_get_user(dmu_buf_t *db);
706
707 objset_t *dmu_buf_get_objset(dmu_buf_t *db);
708 dnode_t *dmu_buf_dnode_enter(dmu_buf_t *db);
709 void dmu_buf_dnode_exit(dmu_buf_t *db);
710
711 /* Block until any in-progress dmu buf user evictions complete. */
712 void dmu_buf_user_evict_wait(void);
713
714 /*
715 * Returns the blkptr associated with this dbuf, or NULL if not set.
716 */
717 struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db);
718
719 /*
720 * Indicate that you are going to modify the buffer's data (db_data).
721 *
722 * The transaction (tx) must be assigned to a txg (ie. you've called
723 * dmu_tx_assign()). The buffer's object must be held in the tx
724 * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
725 */
726 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
727 void dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder,
728 const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_tx_t *tx);
729
730 /*
731 * You must create a transaction, then hold the objects which you will
732 * (or might) modify as part of this transaction. Then you must assign
733 * the transaction to a transaction group. Once the transaction has
734 * been assigned, you can modify buffers which belong to held objects as
735 * part of this transaction. You can't modify buffers before the
736 * transaction has been assigned; you can't modify buffers which don't
737 * belong to objects which this transaction holds; you can't hold
738 * objects once the transaction has been assigned. You may hold an
739 * object which you are going to free (with dmu_object_free()), but you
740 * don't have to.
741 *
742 * You can abort the transaction before it has been assigned.
743 *
744 * Note that you may hold buffers (with dmu_buf_hold) at any time,
745 * regardless of transaction state.
746 */
747
748 #define DMU_NEW_OBJECT (-1ULL)
749 #define DMU_OBJECT_END (-1ULL)
750
751 dmu_tx_t *dmu_tx_create(objset_t *os);
752 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
753 void dmu_tx_hold_write_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off,
754 int len);
755 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
756 uint64_t len);
757 void dmu_tx_hold_free_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off,
758 uint64_t len);
759 void dmu_tx_hold_remap_l1indirect(dmu_tx_t *tx, uint64_t object);
760 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
761 void dmu_tx_hold_zap_by_dnode(dmu_tx_t *tx, dnode_t *dn, int add,
762 const char *name);
763 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
764 void dmu_tx_hold_bonus_by_dnode(dmu_tx_t *tx, dnode_t *dn);
765 void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
766 void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
767 void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
768 void dmu_tx_abort(dmu_tx_t *tx);
769 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
770 void dmu_tx_wait(dmu_tx_t *tx);
771 void dmu_tx_commit(dmu_tx_t *tx);
772 void dmu_tx_mark_netfree(dmu_tx_t *tx);
773
774 /*
775 * To register a commit callback, dmu_tx_callback_register() must be called.
776 *
777 * dcb_data is a pointer to caller private data that is passed on as a
778 * callback parameter. The caller is responsible for properly allocating and
779 * freeing it.
780 *
781 * When registering a callback, the transaction must be already created, but
782 * it cannot be committed or aborted. It can be assigned to a txg or not.
783 *
784 * The callback will be called after the transaction has been safely written
785 * to stable storage and will also be called if the dmu_tx is aborted.
786 * If there is any error which prevents the transaction from being committed to
787 * disk, the callback will be called with a value of error != 0.
788 *
789 * When multiple callbacks are registered to the transaction, the callbacks
790 * will be called in reverse order to let Lustre, the only user of commit
791 * callback currently, take the fast path of its commit callback handling.
792 */
793 typedef void dmu_tx_callback_func_t(void *dcb_data, int error);
794
795 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
796 void *dcb_data);
797 void dmu_tx_do_callbacks(list_t *cb_list, int error);
798
799 /*
800 * Free up the data blocks for a defined range of a file. If size is
801 * -1, the range from offset to end-of-file is freed.
802 */
803 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
804 uint64_t size, dmu_tx_t *tx);
805 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
806 uint64_t size);
807 int dmu_free_long_object(objset_t *os, uint64_t object);
808
809 /*
810 * Convenience functions.
811 *
812 * Canfail routines will return 0 on success, or an errno if there is a
813 * nonrecoverable I/O error.
814 */
815 #define DMU_READ_PREFETCH 0 /* prefetch */
816 #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */
817 #define DMU_READ_NO_DECRYPT 2 /* don't decrypt */
818 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
819 void *buf, uint32_t flags);
820 int dmu_read_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, void *buf,
821 uint32_t flags);
822 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
823 const void *buf, dmu_tx_t *tx);
824 void dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size,
825 const void *buf, dmu_tx_t *tx);
826 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
827 dmu_tx_t *tx);
828 #ifdef _KERNEL
829 #include <linux/blkdev_compat.h>
830 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
831 int dmu_read_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size);
832 int dmu_read_uio_dnode(dnode_t *dn, struct uio *uio, uint64_t size);
833 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
834 dmu_tx_t *tx);
835 int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
836 dmu_tx_t *tx);
837 int dmu_write_uio_dnode(dnode_t *dn, struct uio *uio, uint64_t size,
838 dmu_tx_t *tx);
839 #endif
840 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
841 void dmu_return_arcbuf(struct arc_buf *buf);
842 void dmu_assign_arcbuf_by_dnode(dnode_t *dn, uint64_t offset,
843 struct arc_buf *buf, dmu_tx_t *tx);
844 void dmu_assign_arcbuf_by_dbuf(dmu_buf_t *handle, uint64_t offset,
845 struct arc_buf *buf, dmu_tx_t *tx);
846 #define dmu_assign_arcbuf dmu_assign_arcbuf_by_dbuf
847 void dmu_copy_from_buf(objset_t *os, uint64_t object, uint64_t offset,
848 dmu_buf_t *handle, dmu_tx_t *tx);
849 #ifdef HAVE_UIO_ZEROCOPY
850 int dmu_xuio_init(struct xuio *uio, int niov);
851 void dmu_xuio_fini(struct xuio *uio);
852 int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
853 size_t n);
854 int dmu_xuio_cnt(struct xuio *uio);
855 struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
856 void dmu_xuio_clear(struct xuio *uio, int i);
857 #endif /* HAVE_UIO_ZEROCOPY */
858 void xuio_stat_wbuf_copied(void);
859 void xuio_stat_wbuf_nocopy(void);
860
861 extern int zfs_prefetch_disable;
862 extern int zfs_max_recordsize;
863
864 /*
865 * Asynchronously try to read in the data.
866 */
867 void dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset,
868 uint64_t len, enum zio_priority pri);
869
870 typedef struct dmu_object_info {
871 /* All sizes are in bytes unless otherwise indicated. */
872 uint32_t doi_data_block_size;
873 uint32_t doi_metadata_block_size;
874 dmu_object_type_t doi_type;
875 dmu_object_type_t doi_bonus_type;
876 uint64_t doi_bonus_size;
877 uint8_t doi_indirection; /* 2 = dnode->indirect->data */
878 uint8_t doi_checksum;
879 uint8_t doi_compress;
880 uint8_t doi_nblkptr;
881 uint8_t doi_pad[4];
882 uint64_t doi_dnodesize;
883 uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */
884 uint64_t doi_max_offset;
885 uint64_t doi_fill_count; /* number of non-empty blocks */
886 } dmu_object_info_t;
887
888 typedef void (*const arc_byteswap_func_t)(void *buf, size_t size);
889
890 typedef struct dmu_object_type_info {
891 dmu_object_byteswap_t ot_byteswap;
892 boolean_t ot_metadata;
893 boolean_t ot_dbuf_metadata_cache;
894 boolean_t ot_encrypt;
895 char *ot_name;
896 } dmu_object_type_info_t;
897
898 typedef const struct dmu_object_byteswap_info {
899 arc_byteswap_func_t ob_func;
900 char *ob_name;
901 } dmu_object_byteswap_info_t;
902
903 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
904 extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS];
905
906 /*
907 * Get information on a DMU object.
908 *
909 * Return 0 on success or ENOENT if object is not allocated.
910 *
911 * If doi is NULL, just indicates whether the object exists.
912 */
913 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
914 void __dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
915 /* Like dmu_object_info, but faster if you have a held dnode in hand. */
916 void dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi);
917 /* Like dmu_object_info, but faster if you have a held dbuf in hand. */
918 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
919 /*
920 * Like dmu_object_info_from_db, but faster still when you only care about
921 * the size. This is specifically optimized for zfs_getattr().
922 */
923 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
924 u_longlong_t *nblk512);
925
926 void dmu_object_dnsize_from_db(dmu_buf_t *db, int *dnsize);
927
928 typedef struct dmu_objset_stats {
929 uint64_t dds_num_clones; /* number of clones of this */
930 uint64_t dds_creation_txg;
931 uint64_t dds_guid;
932 dmu_objset_type_t dds_type;
933 uint8_t dds_is_snapshot;
934 uint8_t dds_inconsistent;
935 char dds_origin[ZFS_MAX_DATASET_NAME_LEN];
936 } dmu_objset_stats_t;
937
938 /*
939 * Get stats on a dataset.
940 */
941 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
942
943 /*
944 * Add entries to the nvlist for all the objset's properties. See
945 * zfs_prop_table[] and zfs(1m) for details on the properties.
946 */
947 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
948
949 /*
950 * Get the space usage statistics for statvfs().
951 *
952 * refdbytes is the amount of space "referenced" by this objset.
953 * availbytes is the amount of space available to this objset, taking
954 * into account quotas & reservations, assuming that no other objsets
955 * use the space first. These values correspond to the 'referenced' and
956 * 'available' properties, described in the zfs(1m) manpage.
957 *
958 * usedobjs and availobjs are the number of objects currently allocated,
959 * and available.
960 */
961 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
962 uint64_t *usedobjsp, uint64_t *availobjsp);
963
964 /*
965 * The fsid_guid is a 56-bit ID that can change to avoid collisions.
966 * (Contrast with the ds_guid which is a 64-bit ID that will never
967 * change, so there is a small probability that it will collide.)
968 */
969 uint64_t dmu_objset_fsid_guid(objset_t *os);
970
971 /*
972 * Get the [cm]time for an objset's snapshot dir
973 */
974 inode_timespec_t dmu_objset_snap_cmtime(objset_t *os);
975
976 int dmu_objset_is_snapshot(objset_t *os);
977
978 extern struct spa *dmu_objset_spa(objset_t *os);
979 extern struct zilog *dmu_objset_zil(objset_t *os);
980 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
981 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
982 extern void dmu_objset_name(objset_t *os, char *buf);
983 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
984 extern uint64_t dmu_objset_id(objset_t *os);
985 extern uint64_t dmu_objset_dnodesize(objset_t *os);
986 extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os);
987 extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os);
988 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
989 uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
990 extern int dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *val);
991 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
992 int maxlen, boolean_t *conflict);
993 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
994 uint64_t *idp, uint64_t *offp);
995
996 typedef int objset_used_cb_t(dmu_object_type_t bonustype,
997 void *bonus, uint64_t *userp, uint64_t *groupp, uint64_t *projectp);
998 extern void dmu_objset_register_type(dmu_objset_type_t ost,
999 objset_used_cb_t *cb);
1000 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
1001 extern void *dmu_objset_get_user(objset_t *os);
1002
1003 /*
1004 * Return the txg number for the given assigned transaction.
1005 */
1006 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
1007
1008 /*
1009 * Synchronous write.
1010 * If a parent zio is provided this function initiates a write on the
1011 * provided buffer as a child of the parent zio.
1012 * In the absence of a parent zio, the write is completed synchronously.
1013 * At write completion, blk is filled with the bp of the written block.
1014 * Note that while the data covered by this function will be on stable
1015 * storage when the write completes this new data does not become a
1016 * permanent part of the file until the associated transaction commits.
1017 */
1018
1019 /*
1020 * {zfs,zvol,ztest}_get_done() args
1021 */
1022 typedef struct zgd {
1023 struct lwb *zgd_lwb;
1024 struct blkptr *zgd_bp;
1025 dmu_buf_t *zgd_db;
1026 struct rl *zgd_rl;
1027 void *zgd_private;
1028 } zgd_t;
1029
1030 typedef void dmu_sync_cb_t(zgd_t *arg, int error);
1031 int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
1032
1033 /*
1034 * Find the next hole or data block in file starting at *off
1035 * Return found offset in *off. Return ESRCH for end of file.
1036 */
1037 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
1038 uint64_t *off);
1039
1040 /*
1041 * Initial setup and final teardown.
1042 */
1043 extern void dmu_init(void);
1044 extern void dmu_fini(void);
1045
1046 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
1047 uint64_t object, uint64_t offset, int len);
1048 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
1049 dmu_traverse_cb_t cb, void *arg);
1050
1051 int dmu_diff(const char *tosnap_name, const char *fromsnap_name,
1052 struct vnode *vp, offset_t *offp);
1053
1054 /* CRC64 table */
1055 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */
1056 extern uint64_t zfs_crc64_table[256];
1057
1058 #ifdef __cplusplus
1059 }
1060 #endif
1061
1062 #endif /* _SYS_DMU_H */