]> git.proxmox.com Git - mirror_zfs.git/blame - include/sys/dmu.h
OpenZFS 7614, 9064 - zfs device evacuation/removal
[mirror_zfs.git] / include / sys / dmu.h
CommitLineData
34dc7c2f
BB
1/*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21/*
428870ff 22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
82644107 23 * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
632a242e 24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
37abac6d 25 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
6c59307a 26 * Copyright 2014 HybridCluster. All rights reserved.
0c66c32d 27 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
3c67d83a 28 * Copyright 2013 Saso Kiselkov. All rights reserved.
34dc7c2f
BB
29 */
30
428870ff
BB
31/* Portions Copyright 2010 Robert Milkowski */
32
34dc7c2f
BB
33#ifndef _SYS_DMU_H
34#define _SYS_DMU_H
35
34dc7c2f
BB
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
0c66c32d 44#include <sys/zfs_context.h>
34dc7c2f 45#include <sys/inttypes.h>
34dc7c2f 46#include <sys/cred.h>
6f1ffb06 47#include <sys/fs/zfs.h>
2aa34383 48#include <sys/zio_compress.h>
fcff0f35 49#include <sys/zio_priority.h>
59e6e7ca 50#include <sys/uio.h>
34dc7c2f
BB
51
52#ifdef __cplusplus
53extern "C" {
54#endif
55
34dc7c2f
BB
56struct page;
57struct vnode;
58struct spa;
59struct zilog;
60struct zio;
61struct blkptr;
62struct zap_cursor;
63struct dsl_dataset;
64struct dsl_pool;
65struct dnode;
66struct drr_begin;
67struct drr_end;
5dbd68a3 68struct zbookmark_phys;
34dc7c2f
BB
69struct spa;
70struct nvlist;
9babb374 71struct arc_buf;
428870ff
BB
72struct zio_prop;
73struct sa_handle;
b5256303 74struct dsl_crypto_params;
34dc7c2f
BB
75
76typedef struct objset objset_t;
77typedef struct dmu_tx dmu_tx_t;
78typedef struct dsl_dir dsl_dir_t;
2bce8049 79typedef struct dnode dnode_t;
34dc7c2f 80
9ae529ec
CS
81typedef 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
b5256303
TC
104#define DMU_OT_ENCRYPTED 0x20
105#define DMU_OT_BYTESWAP_MASK 0x1f
9ae529ec
CS
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).
111 */
b5256303 112#define DMU_OT(byteswap, metadata, encrypted) \
9ae529ec
CS
113 (DMU_OT_NEWTYPE | \
114 ((metadata) ? DMU_OT_METADATA : 0) | \
b5256303 115 ((encrypted) ? DMU_OT_ENCRYPTED : 0) | \
9ae529ec
CS
116 ((byteswap) & DMU_OT_BYTESWAP_MASK))
117
118#define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \
119 ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \
120 (ot) < DMU_OT_NUMTYPES)
121
5c27ec10
MA
122/*
123 * MDB doesn't have dmu_ot; it defines these macros itself.
124 */
125#ifndef ZFS_MDB
126#define DMU_OT_IS_METADATA_IMPL(ot) (dmu_ot[ot].ot_metadata)
127#define DMU_OT_IS_ENCRYPTED_IMPL(ot) (dmu_ot[ot].ot_encrypt)
128#define DMU_OT_BYTESWAP_IMPL(ot) (dmu_ot[ot].ot_byteswap)
129#endif
130
9ae529ec
CS
131#define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \
132 ((ot) & DMU_OT_METADATA) : \
5c27ec10 133 DMU_OT_IS_METADATA_IMPL(ot))
9ae529ec 134
b5256303
TC
135#define DMU_OT_IS_ENCRYPTED(ot) (((ot) & DMU_OT_NEWTYPE) ? \
136 ((ot) & DMU_OT_ENCRYPTED) : \
5c27ec10 137 DMU_OT_IS_ENCRYPTED_IMPL(ot))
b5256303 138
9b67f605
MA
139/*
140 * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't
141 * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill
142 * is repurposed for embedded BPs.
143 */
144#define DMU_OT_HAS_FILL(ot) \
145 ((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET)
146
9ae529ec
CS
147#define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \
148 ((ot) & DMU_OT_BYTESWAP_MASK) : \
5c27ec10 149 DMU_OT_BYTESWAP_IMPL(ot))
9ae529ec 150
34dc7c2f
BB
151typedef enum dmu_object_type {
152 DMU_OT_NONE,
153 /* general: */
154 DMU_OT_OBJECT_DIRECTORY, /* ZAP */
155 DMU_OT_OBJECT_ARRAY, /* UINT64 */
156 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */
157 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */
428870ff
BB
158 DMU_OT_BPOBJ, /* UINT64 */
159 DMU_OT_BPOBJ_HDR, /* UINT64 */
34dc7c2f
BB
160 /* spa: */
161 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */
162 DMU_OT_SPACE_MAP, /* UINT64 */
163 /* zil: */
164 DMU_OT_INTENT_LOG, /* UINT64 */
165 /* dmu: */
166 DMU_OT_DNODE, /* DNODE */
167 DMU_OT_OBJSET, /* OBJSET */
168 /* dsl: */
169 DMU_OT_DSL_DIR, /* UINT64 */
170 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */
171 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */
172 DMU_OT_DSL_PROPS, /* ZAP */
173 DMU_OT_DSL_DATASET, /* UINT64 */
174 /* zpl: */
175 DMU_OT_ZNODE, /* ZNODE */
176 DMU_OT_OLDACL, /* Old ACL */
177 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */
178 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */
179 DMU_OT_MASTER_NODE, /* ZAP */
180 DMU_OT_UNLINKED_SET, /* ZAP */
181 /* zvol: */
182 DMU_OT_ZVOL, /* UINT8 */
183 DMU_OT_ZVOL_PROP, /* ZAP */
184 /* other; for testing only! */
185 DMU_OT_PLAIN_OTHER, /* UINT8 */
186 DMU_OT_UINT64_OTHER, /* UINT64 */
187 DMU_OT_ZAP_OTHER, /* ZAP */
188 /* new object types: */
189 DMU_OT_ERROR_LOG, /* ZAP */
190 DMU_OT_SPA_HISTORY, /* UINT8 */
191 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */
192 DMU_OT_POOL_PROPS, /* ZAP */
193 DMU_OT_DSL_PERMS, /* ZAP */
194 DMU_OT_ACL, /* ACL */
195 DMU_OT_SYSACL, /* SYSACL */
196 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */
197 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */
b128c09f 198 DMU_OT_NEXT_CLONES, /* ZAP */
428870ff 199 DMU_OT_SCAN_QUEUE, /* ZAP */
9babb374
BB
200 DMU_OT_USERGROUP_USED, /* ZAP */
201 DMU_OT_USERGROUP_QUOTA, /* ZAP */
45d1cae3 202 DMU_OT_USERREFS, /* ZAP */
428870ff
BB
203 DMU_OT_DDT_ZAP, /* ZAP */
204 DMU_OT_DDT_STATS, /* ZAP */
205 DMU_OT_SA, /* System attr */
206 DMU_OT_SA_MASTER_NODE, /* ZAP */
207 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */
208 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */
209 DMU_OT_SCAN_XLATE, /* ZAP */
210 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */
211 DMU_OT_DEADLIST, /* ZAP */
212 DMU_OT_DEADLIST_HDR, /* UINT64 */
213 DMU_OT_DSL_CLONES, /* ZAP */
214 DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */
9ae529ec
CS
215 /*
216 * Do not allocate new object types here. Doing so makes the on-disk
217 * format incompatible with any other format that uses the same object
218 * type number.
219 *
220 * When creating an object which does not have one of the above types
221 * use the DMU_OTN_* type with the correct byteswap and metadata
222 * values.
223 *
224 * The DMU_OTN_* types do not have entries in the dmu_ot table,
225 * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead
226 * of indexing into dmu_ot directly (this works for both DMU_OT_* types
227 * and DMU_OTN_* types).
228 */
229 DMU_OT_NUMTYPES,
230
231 /*
232 * Names for valid types declared with DMU_OT().
233 */
b5256303
TC
234 DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_FALSE),
235 DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_FALSE),
236 DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_FALSE),
237 DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_FALSE),
238 DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_FALSE),
239 DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_FALSE),
240 DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_FALSE),
241 DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_FALSE),
242 DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_FALSE),
243 DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_FALSE),
244
245 DMU_OTN_UINT8_ENC_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_TRUE),
246 DMU_OTN_UINT8_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_TRUE),
247 DMU_OTN_UINT16_ENC_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_TRUE),
248 DMU_OTN_UINT16_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_TRUE),
249 DMU_OTN_UINT32_ENC_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_TRUE),
250 DMU_OTN_UINT32_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_TRUE),
251 DMU_OTN_UINT64_ENC_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_TRUE),
252 DMU_OTN_UINT64_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_TRUE),
253 DMU_OTN_ZAP_ENC_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_TRUE),
254 DMU_OTN_ZAP_ENC_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_TRUE),
34dc7c2f
BB
255} dmu_object_type_t;
256
0735ecb3
PS
257/*
258 * These flags are intended to be used to specify the "txg_how"
259 * parameter when calling the dmu_tx_assign() function. See the comment
260 * above dmu_tx_assign() for more details on the meaning of these flags.
261 */
262#define TXG_NOWAIT (0ULL)
263#define TXG_WAIT (1ULL<<0)
264#define TXG_NOTHROTTLE (1ULL<<1)
13fe0198 265
34dc7c2f
BB
266void byteswap_uint64_array(void *buf, size_t size);
267void byteswap_uint32_array(void *buf, size_t size);
268void byteswap_uint16_array(void *buf, size_t size);
269void byteswap_uint8_array(void *buf, size_t size);
270void zap_byteswap(void *buf, size_t size);
271void zfs_oldacl_byteswap(void *buf, size_t size);
272void zfs_acl_byteswap(void *buf, size_t size);
273void zfs_znode_byteswap(void *buf, size_t size);
274
34dc7c2f
BB
275#define DS_FIND_SNAPSHOTS (1<<0)
276#define DS_FIND_CHILDREN (1<<1)
9c43027b 277#define DS_FIND_SERIALIZE (1<<2)
34dc7c2f
BB
278
279/*
280 * The maximum number of bytes that can be accessed as part of one
281 * operation, including metadata.
282 */
f1512ee6 283#define DMU_MAX_ACCESS (64 * 1024 * 1024) /* 64MB */
b128c09f 284#define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
34dc7c2f 285
9babb374
BB
286#define DMU_USERUSED_OBJECT (-1ULL)
287#define DMU_GROUPUSED_OBJECT (-2ULL)
9c5167d1 288#define DMU_PROJECTUSED_OBJECT (-3ULL)
9babb374 289
1de321e6 290/*
9c5167d1 291 * Zap prefix for object accounting in DMU_{USER,GROUP,PROJECT}USED_OBJECT.
1de321e6
JX
292 */
293#define DMU_OBJACCT_PREFIX "obj-"
294#define DMU_OBJACCT_PREFIX_LEN 4
295
428870ff
BB
296/*
297 * artificial blkids for bonus buffer and spill blocks
298 */
299#define DMU_BONUS_BLKID (-1ULL)
300#define DMU_SPILL_BLKID (-2ULL)
b5256303 301
34dc7c2f
BB
302/*
303 * Public routines to create, destroy, open, and close objsets.
304 */
b5256303
TC
305typedef void dmu_objset_create_sync_func_t(objset_t *os, void *arg,
306 cred_t *cr, dmu_tx_t *tx);
307
428870ff
BB
308int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
309int dmu_objset_own(const char *name, dmu_objset_type_t type,
b5256303 310 boolean_t readonly, boolean_t key_required, void *tag, objset_t **osp);
428870ff 311void dmu_objset_rele(objset_t *os, void *tag);
b5256303 312void dmu_objset_disown(objset_t *os, boolean_t key_required, void *tag);
428870ff
BB
313int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
314
13fe0198 315void dmu_objset_evict_dbufs(objset_t *os);
428870ff 316int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
b5256303
TC
317 struct dsl_crypto_params *dcp, dmu_objset_create_sync_func_t func,
318 void *arg);
13fe0198
MA
319int dmu_objset_clone(const char *name, const char *origin);
320int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer,
6f1ffb06 321 struct nvlist *errlist);
6f1ffb06
MA
322int dmu_objset_snapshot_one(const char *fsname, const char *snapname);
323int dmu_objset_snapshot_tmp(const char *, const char *, int);
428870ff 324int dmu_objset_find(char *name, int func(const char *, void *), void *arg,
34dc7c2f
BB
325 int flags);
326void dmu_objset_byteswap(void *buf, size_t size);
13fe0198
MA
327int dsl_dataset_rename_snapshot(const char *fsname,
328 const char *oldsnapname, const char *newsnapname, boolean_t recursive);
a1d477c2 329int dmu_objset_remap_indirects(const char *fsname);
34dc7c2f
BB
330
331typedef struct dmu_buf {
332 uint64_t db_object; /* object that this buffer is part of */
333 uint64_t db_offset; /* byte offset in this object */
334 uint64_t db_size; /* size of buffer in bytes */
335 void *db_data; /* data in buffer */
336} dmu_buf_t;
337
34dc7c2f
BB
338/*
339 * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
340 */
341#define DMU_POOL_DIRECTORY_OBJECT 1
342#define DMU_POOL_CONFIG "config"
9ae529ec
CS
343#define DMU_POOL_FEATURES_FOR_WRITE "features_for_write"
344#define DMU_POOL_FEATURES_FOR_READ "features_for_read"
345#define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions"
b0bc7a84 346#define DMU_POOL_FEATURE_ENABLED_TXG "feature_enabled_txg"
34dc7c2f 347#define DMU_POOL_ROOT_DATASET "root_dataset"
428870ff 348#define DMU_POOL_SYNC_BPOBJ "sync_bplist"
34dc7c2f
BB
349#define DMU_POOL_ERRLOG_SCRUB "errlog_scrub"
350#define DMU_POOL_ERRLOG_LAST "errlog_last"
351#define DMU_POOL_SPARES "spares"
352#define DMU_POOL_DEFLATE "deflate"
353#define DMU_POOL_HISTORY "history"
354#define DMU_POOL_PROPS "pool_props"
355#define DMU_POOL_L2CACHE "l2cache"
428870ff
BB
356#define DMU_POOL_TMP_USERREFS "tmp_userrefs"
357#define DMU_POOL_DDT "DDT-%s-%s-%s"
358#define DMU_POOL_DDT_STATS "DDT-statistics"
359#define DMU_POOL_CREATION_VERSION "creation_version"
360#define DMU_POOL_SCAN "scan"
361#define DMU_POOL_FREE_BPOBJ "free_bpobj"
9ae529ec 362#define DMU_POOL_BPTREE_OBJ "bptree_obj"
753c3839 363#define DMU_POOL_EMPTY_BPOBJ "empty_bpobj"
3c67d83a 364#define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt"
e0ab3ab5 365#define DMU_POOL_VDEV_ZAP_MAP "com.delphix:vdev_zap_map"
a1d477c2
MA
366#define DMU_POOL_REMOVING "com.delphix:removing"
367#define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj"
368#define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect"
b128c09f 369
34dc7c2f
BB
370/*
371 * Allocate an object from this objset. The range of object numbers
372 * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode.
373 *
374 * The transaction must be assigned to a txg. The newly allocated
375 * object will be "held" in the transaction (ie. you can modify the
376 * newly allocated object in this transaction).
377 *
378 * dmu_object_alloc() chooses an object and returns it in *objectp.
379 *
380 * dmu_object_claim() allocates a specific object number. If that
381 * number is already allocated, it fails and returns EEXIST.
382 *
383 * Return 0 on success, or ENOSPC or EEXIST as specified above.
384 */
385uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
386 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
50c957f7
NB
387uint64_t dmu_object_alloc_dnsize(objset_t *os, dmu_object_type_t ot,
388 int blocksize, dmu_object_type_t bonus_type, int bonus_len,
389 int dnodesize, dmu_tx_t *tx);
34dc7c2f
BB
390int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
391 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
50c957f7
NB
392int dmu_object_claim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot,
393 int blocksize, dmu_object_type_t bonus_type, int bonus_len,
394 int dnodesize, dmu_tx_t *tx);
34dc7c2f 395int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
6c59307a 396 int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp);
50c957f7
NB
397int dmu_object_reclaim_dnsize(objset_t *os, uint64_t object,
398 dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype,
399 int bonuslen, int dnodesize, dmu_tx_t *txp);
34dc7c2f
BB
400
401/*
402 * Free an object from this objset.
403 *
404 * The object's data will be freed as well (ie. you don't need to call
405 * dmu_free(object, 0, -1, tx)).
406 *
407 * The object need not be held in the transaction.
408 *
409 * If there are any holds on this object's buffers (via dmu_buf_hold()),
410 * or tx holds on the object (via dmu_tx_hold_object()), you can not
411 * free it; it fails and returns EBUSY.
412 *
413 * If the object is not allocated, it fails and returns ENOENT.
414 *
415 * Return 0 on success, or EBUSY or ENOENT as specified above.
416 */
417int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
418
419/*
420 * Find the next allocated or free object.
421 *
422 * The objectp parameter is in-out. It will be updated to be the next
423 * object which is allocated. Ignore objects which have not been
424 * modified since txg.
425 *
426 * XXX Can only be called on a objset with no dirty data.
427 *
428 * Returns 0 on success, or ENOENT if there are no more objects.
429 */
430int dmu_object_next(objset_t *os, uint64_t *objectp,
431 boolean_t hole, uint64_t txg);
432
b5256303
TC
433/*
434 * Set the number of levels on a dnode. nlevels must be greater than the
435 * current number of levels or an EINVAL will be returned.
436 */
437int dmu_object_set_nlevels(objset_t *os, uint64_t object, int nlevels,
438 dmu_tx_t *tx);
439
34dc7c2f
BB
440/*
441 * Set the data blocksize for an object.
442 *
443 * The object cannot have any blocks allcated beyond the first. If
444 * the first block is allocated already, the new size must be greater
445 * than the current block size. If these conditions are not met,
446 * ENOTSUP will be returned.
447 *
448 * Returns 0 on success, or EBUSY if there are any holds on the object
449 * contents, or ENOTSUP as described above.
450 */
451int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
452 int ibs, dmu_tx_t *tx);
453
ae76f45c
TC
454/*
455 * Manually set the maxblkid on a dnode. This will adjust nlevels accordingly
456 * to accommodate the change.
457 */
458int dmu_object_set_maxblkid(objset_t *os, uint64_t object, uint64_t maxblkid,
459 dmu_tx_t *tx);
460
34dc7c2f
BB
461/*
462 * Set the checksum property on a dnode. The new checksum algorithm will
463 * apply to all newly written blocks; existing blocks will not be affected.
464 */
465void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
466 dmu_tx_t *tx);
467
468/*
469 * Set the compress property on a dnode. The new compression algorithm will
470 * apply to all newly written blocks; existing blocks will not be affected.
471 */
472void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
473 dmu_tx_t *tx);
474
9b840763
TC
475int dmu_object_dirty_raw(objset_t *os, uint64_t object, dmu_tx_t *tx);
476
a1d477c2
MA
477int dmu_object_remap_indirects(objset_t *os, uint64_t object, uint64_t txg);
478
9b840763 479void dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset,
9b67f605
MA
480 void *data, uint8_t etype, uint8_t comp, int uncompressed_size,
481 int compressed_size, int byteorder, dmu_tx_t *tx);
482
34dc7c2f 483/*
428870ff 484 * Decide how to write a block: checksum, compression, number of copies, etc.
34dc7c2f 485 */
428870ff
BB
486#define WP_NOFILL 0x1
487#define WP_DMU_SYNC 0x2
488#define WP_SPILL 0x4
489
82644107
MA
490void dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp,
491 struct zio_prop *zp);
b5256303 492
34dc7c2f
BB
493/*
494 * The bonus data is accessed more or less like a regular buffer.
495 * You must dmu_bonus_hold() to get the buffer, which will give you a
496 * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
a1d477c2
MA
497 * data. As with any normal buffer, you must call dmu_buf_will_dirty()
498 * before modifying it, and the
34dc7c2f
BB
499 * object must be held in an assigned transaction before calling
500 * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus
4e95cc99 501 * buffer as well. You must release what you hold with dmu_buf_rele().
e49f1e20
WA
502 *
503 * Returns ENOENT, EIO, or 0.
34dc7c2f 504 */
b5256303
TC
505int dmu_bonus_hold_impl(objset_t *os, uint64_t object, void *tag,
506 uint32_t flags, dmu_buf_t **dbp);
34dc7c2f
BB
507int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
508int dmu_bonus_max(void);
509int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
428870ff 510int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
572e2857 511dmu_object_type_t dmu_get_bonustype(dmu_buf_t *);
428870ff
BB
512int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
513
514/*
515 * Special spill buffer support used by "SA" framework
516 */
517
518int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
2bce8049 519int dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags,
428870ff
BB
520 void *tag, dmu_buf_t **dbp);
521int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
34dc7c2f
BB
522
523/*
524 * Obtain the DMU buffer from the specified object which contains the
525 * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so
526 * that it will remain in memory. You must release the hold with
4e95cc99
JE
527 * dmu_buf_rele(). You must not access the dmu_buf_t after releasing
528 * what you hold. You must have a hold on any dmu_buf_t* you pass to the DMU.
34dc7c2f
BB
529 *
530 * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
531 * on the returned buffer before reading or writing the buffer's
532 * db_data. The comments for those routines describe what particular
533 * operations are valid after calling them.
534 *
535 * The object number must be a valid, allocated object number.
536 */
537int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
428870ff 538 void *tag, dmu_buf_t **, int flags);
2bce8049
MA
539int dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset,
540 void *tag, dmu_buf_t **dbp, int flags);
6ebebace
JG
541
542/*
543 * Add a reference to a dmu buffer that has already been held via
544 * dmu_buf_hold() in the current context.
545 */
34dc7c2f 546void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
6ebebace
JG
547
548/*
549 * Attempt to add a reference to a dmu buffer that is in an unknown state,
550 * using a pointer that may have been invalidated by eviction processing.
551 * The request will succeed if the passed in dbuf still represents the
552 * same os/object/blkid, is ineligible for eviction, and has at least
553 * one hold by a user other than the syncer.
554 */
555boolean_t dmu_buf_try_add_ref(dmu_buf_t *, objset_t *os, uint64_t object,
556 uint64_t blkid, void *tag);
557
34dc7c2f
BB
558void dmu_buf_rele(dmu_buf_t *db, void *tag);
559uint64_t dmu_buf_refcount(dmu_buf_t *db);
560
561/*
562 * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
563 * range of an object. A pointer to an array of dmu_buf_t*'s is
564 * returned (in *dbpp).
565 *
566 * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
567 * frees the array. The hold on the array of buffers MUST be released
568 * with dmu_buf_rele_array. You can NOT release the hold on each buffer
569 * individually with dmu_buf_rele.
570 */
571int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
7f60329a
MA
572 uint64_t length, boolean_t read, void *tag,
573 int *numbufsp, dmu_buf_t ***dbpp);
34dc7c2f
BB
574void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
575
0c66c32d
JG
576typedef void dmu_buf_evict_func_t(void *user_ptr);
577
34dc7c2f 578/*
0c66c32d
JG
579 * A DMU buffer user object may be associated with a dbuf for the
580 * duration of its lifetime. This allows the user of a dbuf (client)
581 * to attach private data to a dbuf (e.g. in-core only data such as a
582 * dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified
583 * when that dbuf has been evicted. Clients typically respond to the
584 * eviction notification by freeing their private data, thus ensuring
585 * the same lifetime for both dbuf and private data.
34dc7c2f 586 *
0c66c32d
JG
587 * The mapping from a dmu_buf_user_t to any client private data is the
588 * client's responsibility. All current consumers of the API with private
589 * data embed a dmu_buf_user_t as the first member of the structure for
590 * their private data. This allows conversions between the two types
591 * with a simple cast. Since the DMU buf user API never needs access
592 * to the private data, other strategies can be employed if necessary
593 * or convenient for the client (e.g. using container_of() to do the
594 * conversion for private data that cannot have the dmu_buf_user_t as
595 * its first member).
34dc7c2f 596 *
0c66c32d
JG
597 * Eviction callbacks are executed without the dbuf mutex held or any
598 * other type of mechanism to guarantee that the dbuf is still available.
599 * For this reason, users must assume the dbuf has already been freed
600 * and not reference the dbuf from the callback context.
34dc7c2f 601 *
0c66c32d
JG
602 * Users requesting "immediate eviction" are notified as soon as the dbuf
603 * is only referenced by dirty records (dirties == holds). Otherwise the
604 * notification occurs after eviction processing for the dbuf begins.
34dc7c2f 605 */
0c66c32d
JG
606typedef struct dmu_buf_user {
607 /*
608 * Asynchronous user eviction callback state.
609 */
610 taskq_ent_t dbu_tqent;
611
39efbde7
GM
612 /*
613 * This instance's eviction function pointers.
614 *
615 * dbu_evict_func_sync is called synchronously and then
616 * dbu_evict_func_async is executed asynchronously on a taskq.
617 */
618 dmu_buf_evict_func_t *dbu_evict_func_sync;
619 dmu_buf_evict_func_t *dbu_evict_func_async;
0c66c32d
JG
620#ifdef ZFS_DEBUG
621 /*
622 * Pointer to user's dbuf pointer. NULL for clients that do
623 * not associate a dbuf with their user data.
624 *
625 * The dbuf pointer is cleared upon eviction so as to catch
626 * use-after-evict bugs in clients.
627 */
628 dmu_buf_t **dbu_clear_on_evict_dbufp;
629#endif
630} dmu_buf_user_t;
631
34dc7c2f 632/*
0c66c32d
JG
633 * Initialize the given dmu_buf_user_t instance with the eviction function
634 * evict_func, to be called when the user is evicted.
635 *
636 * NOTE: This function should only be called once on a given dmu_buf_user_t.
637 * To allow enforcement of this, dbu must already be zeroed on entry.
34dc7c2f 638 */
c4434877 639/*ARGSUSED*/
0c66c32d 640static inline void
39efbde7
GM
641dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func_sync,
642 dmu_buf_evict_func_t *evict_func_async, dmu_buf_t **clear_on_evict_dbufp)
0c66c32d 643{
39efbde7
GM
644 ASSERT(dbu->dbu_evict_func_sync == NULL);
645 ASSERT(dbu->dbu_evict_func_async == NULL);
646
647 /* must have at least one evict func */
648 IMPLY(evict_func_sync == NULL, evict_func_async != NULL);
649 dbu->dbu_evict_func_sync = evict_func_sync;
650 dbu->dbu_evict_func_async = evict_func_async;
f467b05a 651 taskq_init_ent(&dbu->dbu_tqent);
0c66c32d
JG
652#ifdef ZFS_DEBUG
653 dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp;
654#endif
655}
34dc7c2f
BB
656
657/*
0c66c32d
JG
658 * Attach user data to a dbuf and mark it for normal (when the dbuf's
659 * data is cleared or its reference count goes to zero) eviction processing.
660 *
661 * Returns NULL on success, or the existing user if another user currently
662 * owns the buffer.
663 */
664void *dmu_buf_set_user(dmu_buf_t *db, dmu_buf_user_t *user);
665
666/*
667 * Attach user data to a dbuf and mark it for immediate (its dirty and
668 * reference counts are equal) eviction processing.
669 *
670 * Returns NULL on success, or the existing user if another user currently
671 * owns the buffer.
672 */
673void *dmu_buf_set_user_ie(dmu_buf_t *db, dmu_buf_user_t *user);
674
675/*
676 * Replace the current user of a dbuf.
677 *
678 * If given the current user of a dbuf, replaces the dbuf's user with
679 * "new_user" and returns the user data pointer that was replaced.
680 * Otherwise returns the current, and unmodified, dbuf user pointer.
681 */
682void *dmu_buf_replace_user(dmu_buf_t *db,
683 dmu_buf_user_t *old_user, dmu_buf_user_t *new_user);
684
685/*
686 * Remove the specified user data for a DMU buffer.
687 *
688 * Returns the user that was removed on success, or the current user if
689 * another user currently owns the buffer.
690 */
691void *dmu_buf_remove_user(dmu_buf_t *db, dmu_buf_user_t *user);
692
693/*
694 * Returns the user data (dmu_buf_user_t *) associated with this dbuf.
34dc7c2f
BB
695 */
696void *dmu_buf_get_user(dmu_buf_t *db);
697
8bea9815 698objset_t *dmu_buf_get_objset(dmu_buf_t *db);
2bce8049
MA
699dnode_t *dmu_buf_dnode_enter(dmu_buf_t *db);
700void dmu_buf_dnode_exit(dmu_buf_t *db);
8bea9815 701
0c66c32d
JG
702/* Block until any in-progress dmu buf user evictions complete. */
703void dmu_buf_user_evict_wait(void);
704
03c6040b
GW
705/*
706 * Returns the blkptr associated with this dbuf, or NULL if not set.
707 */
708struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db);
709
34dc7c2f
BB
710/*
711 * Indicate that you are going to modify the buffer's data (db_data).
712 *
713 * The transaction (tx) must be assigned to a txg (ie. you've called
714 * dmu_tx_assign()). The buffer's object must be held in the tx
715 * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
716 */
717void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
b5256303 718void dmu_buf_will_change_crypt_params(dmu_buf_t *db, dmu_tx_t *tx);
34dc7c2f
BB
719
720/*
721 * You must create a transaction, then hold the objects which you will
722 * (or might) modify as part of this transaction. Then you must assign
723 * the transaction to a transaction group. Once the transaction has
724 * been assigned, you can modify buffers which belong to held objects as
725 * part of this transaction. You can't modify buffers before the
726 * transaction has been assigned; you can't modify buffers which don't
727 * belong to objects which this transaction holds; you can't hold
728 * objects once the transaction has been assigned. You may hold an
729 * object which you are going to free (with dmu_object_free()), but you
730 * don't have to.
731 *
732 * You can abort the transaction before it has been assigned.
733 *
734 * Note that you may hold buffers (with dmu_buf_hold) at any time,
735 * regardless of transaction state.
736 */
737
738#define DMU_NEW_OBJECT (-1ULL)
739#define DMU_OBJECT_END (-1ULL)
740
741dmu_tx_t *dmu_tx_create(objset_t *os);
742void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
0eef1bde 743void dmu_tx_hold_write_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off,
744 int len);
34dc7c2f
BB
745void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
746 uint64_t len);
0eef1bde 747void dmu_tx_hold_free_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off,
748 uint64_t len);
a1d477c2 749void dmu_tx_hold_remap_l1indirect(dmu_tx_t *tx, uint64_t object);
9babb374 750void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
0eef1bde 751void dmu_tx_hold_zap_by_dnode(dmu_tx_t *tx, dnode_t *dn, int add,
752 const char *name);
34dc7c2f 753void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
0eef1bde 754void dmu_tx_hold_bonus_by_dnode(dmu_tx_t *tx, dnode_t *dn);
428870ff
BB
755void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
756void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
757void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
34dc7c2f 758void dmu_tx_abort(dmu_tx_t *tx);
0735ecb3 759int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
34dc7c2f
BB
760void dmu_tx_wait(dmu_tx_t *tx);
761void dmu_tx_commit(dmu_tx_t *tx);
19d55079 762void dmu_tx_mark_netfree(dmu_tx_t *tx);
34dc7c2f 763
428870ff
BB
764/*
765 * To register a commit callback, dmu_tx_callback_register() must be called.
766 *
767 * dcb_data is a pointer to caller private data that is passed on as a
768 * callback parameter. The caller is responsible for properly allocating and
769 * freeing it.
770 *
771 * When registering a callback, the transaction must be already created, but
772 * it cannot be committed or aborted. It can be assigned to a txg or not.
773 *
774 * The callback will be called after the transaction has been safely written
775 * to stable storage and will also be called if the dmu_tx is aborted.
776 * If there is any error which prevents the transaction from being committed to
777 * disk, the callback will be called with a value of error != 0.
823d48bf 778 *
779 * When multiple callbacks are registered to the transaction, the callbacks
780 * will be called in reverse order to let Lustre, the only user of commit
781 * callback currently, take the fast path of its commit callback handling.
428870ff
BB
782 */
783typedef void dmu_tx_callback_func_t(void *dcb_data, int error);
784
785void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
786 void *dcb_data);
823d48bf 787void dmu_tx_do_callbacks(list_t *cb_list, int error);
428870ff 788
34dc7c2f
BB
789/*
790 * Free up the data blocks for a defined range of a file. If size is
9ae529ec 791 * -1, the range from offset to end-of-file is freed.
34dc7c2f
BB
792 */
793int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
440a3eb9 794 uint64_t size, dmu_tx_t *tx);
b128c09f 795int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
440a3eb9
TC
796 uint64_t size);
797int dmu_free_long_range_raw(objset_t *os, uint64_t object, uint64_t offset,
798 uint64_t size);
b663a23d 799int dmu_free_long_object(objset_t *os, uint64_t object);
440a3eb9 800int dmu_free_long_object_raw(objset_t *os, uint64_t object);
34dc7c2f
BB
801
802/*
803 * Convenience functions.
804 *
805 * Canfail routines will return 0 on success, or an errno if there is a
806 * nonrecoverable I/O error.
807 */
9babb374
BB
808#define DMU_READ_PREFETCH 0 /* prefetch */
809#define DMU_READ_NO_PREFETCH 1 /* don't prefetch */
b5256303 810#define DMU_READ_NO_DECRYPT 2 /* don't decrypt */
34dc7c2f 811int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
9babb374 812 void *buf, uint32_t flags);
0eef1bde 813int dmu_read_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, void *buf,
814 uint32_t flags);
34dc7c2f
BB
815void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
816 const void *buf, dmu_tx_t *tx);
0eef1bde 817void dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size,
818 const void *buf, dmu_tx_t *tx);
b128c09f
BB
819void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
820 dmu_tx_t *tx);
60101509 821#ifdef _KERNEL
61e90960 822#include <linux/blkdev_compat.h>
872e8d26 823int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
804e0504 824int dmu_read_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size);
5228cf01 825int dmu_read_uio_dnode(dnode_t *dn, struct uio *uio, uint64_t size);
872e8d26
BB
826int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
827 dmu_tx_t *tx);
828int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
829 dmu_tx_t *tx);
5228cf01
RY
830int dmu_write_uio_dnode(dnode_t *dn, struct uio *uio, uint64_t size,
831 dmu_tx_t *tx);
872e8d26 832#endif
9babb374
BB
833struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
834void dmu_return_arcbuf(struct arc_buf *buf);
440a3eb9
TC
835void dmu_assign_arcbuf_by_dnode(dnode_t *dn, uint64_t offset,
836 struct arc_buf *buf, dmu_tx_t *tx);
837void dmu_assign_arcbuf_by_dbuf(dmu_buf_t *handle, uint64_t offset,
838 struct arc_buf *buf, dmu_tx_t *tx);
839#define dmu_assign_arcbuf dmu_assign_arcbuf_by_dbuf
095495e0
TC
840int dmu_convert_mdn_block_to_raw(objset_t *os, uint64_t firstobj,
841 boolean_t byteorder, const uint8_t *salt, const uint8_t *iv,
842 const uint8_t *mac, dmu_tx_t *tx);
b5256303
TC
843void dmu_copy_from_buf(objset_t *os, uint64_t object, uint64_t offset,
844 dmu_buf_t *handle, dmu_tx_t *tx);
5a6765cf 845#ifdef HAVE_UIO_ZEROCOPY
428870ff
BB
846int dmu_xuio_init(struct xuio *uio, int niov);
847void dmu_xuio_fini(struct xuio *uio);
848int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
849 size_t n);
850int dmu_xuio_cnt(struct xuio *uio);
851struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
852void dmu_xuio_clear(struct xuio *uio, int i);
5a6765cf 853#endif /* HAVE_UIO_ZEROCOPY */
0bc8fd78
BB
854void xuio_stat_wbuf_copied(void);
855void xuio_stat_wbuf_nocopy(void);
34dc7c2f
BB
856
857extern int zfs_prefetch_disable;
f1512ee6 858extern int zfs_max_recordsize;
34dc7c2f
BB
859
860/*
861 * Asynchronously try to read in the data.
862 */
fcff0f35
PD
863void dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset,
864 uint64_t len, enum zio_priority pri);
34dc7c2f
BB
865
866typedef struct dmu_object_info {
428870ff 867 /* All sizes are in bytes unless otherwise indicated. */
34dc7c2f
BB
868 uint32_t doi_data_block_size;
869 uint32_t doi_metadata_block_size;
34dc7c2f
BB
870 dmu_object_type_t doi_type;
871 dmu_object_type_t doi_bonus_type;
428870ff 872 uint64_t doi_bonus_size;
34dc7c2f
BB
873 uint8_t doi_indirection; /* 2 = dnode->indirect->data */
874 uint8_t doi_checksum;
875 uint8_t doi_compress;
6c59307a
MA
876 uint8_t doi_nblkptr;
877 uint8_t doi_pad[4];
50c957f7 878 uint64_t doi_dnodesize;
428870ff
BB
879 uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */
880 uint64_t doi_max_offset;
881 uint64_t doi_fill_count; /* number of non-empty blocks */
34dc7c2f
BB
882} dmu_object_info_t;
883
b01615d5 884typedef void (*const arc_byteswap_func_t)(void *buf, size_t size);
34dc7c2f
BB
885
886typedef struct dmu_object_type_info {
9ae529ec 887 dmu_object_byteswap_t ot_byteswap;
34dc7c2f 888 boolean_t ot_metadata;
b5256303 889 boolean_t ot_encrypt;
34dc7c2f
BB
890 char *ot_name;
891} dmu_object_type_info_t;
892
b01615d5
RY
893typedef const struct dmu_object_byteswap_info {
894 arc_byteswap_func_t ob_func;
9ae529ec
CS
895 char *ob_name;
896} dmu_object_byteswap_info_t;
897
34dc7c2f 898extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
9ae529ec 899extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS];
34dc7c2f
BB
900
901/*
902 * Get information on a DMU object.
903 *
904 * Return 0 on success or ENOENT if object is not allocated.
905 *
906 * If doi is NULL, just indicates whether the object exists.
907 */
908int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
e0b0ca98 909void __dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
e49f1e20 910/* Like dmu_object_info, but faster if you have a held dnode in hand. */
2bce8049 911void dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi);
e49f1e20 912/* Like dmu_object_info, but faster if you have a held dbuf in hand. */
34dc7c2f 913void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
e49f1e20
WA
914/*
915 * Like dmu_object_info_from_db, but faster still when you only care about
916 * the size. This is specifically optimized for zfs_getattr().
917 */
34dc7c2f
BB
918void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
919 u_longlong_t *nblk512);
920
50c957f7
NB
921void dmu_object_dnsize_from_db(dmu_buf_t *db, int *dnsize);
922
34dc7c2f
BB
923typedef struct dmu_objset_stats {
924 uint64_t dds_num_clones; /* number of clones of this */
925 uint64_t dds_creation_txg;
926 uint64_t dds_guid;
927 dmu_objset_type_t dds_type;
928 uint8_t dds_is_snapshot;
929 uint8_t dds_inconsistent;
eca7b760 930 char dds_origin[ZFS_MAX_DATASET_NAME_LEN];
34dc7c2f
BB
931} dmu_objset_stats_t;
932
933/*
934 * Get stats on a dataset.
935 */
936void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
937
938/*
939 * Add entries to the nvlist for all the objset's properties. See
940 * zfs_prop_table[] and zfs(1m) for details on the properties.
941 */
942void dmu_objset_stats(objset_t *os, struct nvlist *nv);
943
944/*
945 * Get the space usage statistics for statvfs().
946 *
947 * refdbytes is the amount of space "referenced" by this objset.
948 * availbytes is the amount of space available to this objset, taking
949 * into account quotas & reservations, assuming that no other objsets
950 * use the space first. These values correspond to the 'referenced' and
951 * 'available' properties, described in the zfs(1m) manpage.
952 *
953 * usedobjs and availobjs are the number of objects currently allocated,
954 * and available.
955 */
956void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
957 uint64_t *usedobjsp, uint64_t *availobjsp);
958
959/*
960 * The fsid_guid is a 56-bit ID that can change to avoid collisions.
961 * (Contrast with the ds_guid which is a 64-bit ID that will never
962 * change, so there is a small probability that it will collide.)
963 */
964uint64_t dmu_objset_fsid_guid(objset_t *os);
965
428870ff
BB
966/*
967 * Get the [cm]time for an objset's snapshot dir
968 */
969timestruc_t dmu_objset_snap_cmtime(objset_t *os);
970
34dc7c2f
BB
971int dmu_objset_is_snapshot(objset_t *os);
972
973extern struct spa *dmu_objset_spa(objset_t *os);
974extern struct zilog *dmu_objset_zil(objset_t *os);
975extern struct dsl_pool *dmu_objset_pool(objset_t *os);
976extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
977extern void dmu_objset_name(objset_t *os, char *buf);
978extern dmu_objset_type_t dmu_objset_type(objset_t *os);
979extern uint64_t dmu_objset_id(objset_t *os);
50c957f7 980extern uint64_t dmu_objset_dnodesize(objset_t *os);
faf0f58c
MA
981extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os);
982extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os);
34dc7c2f
BB
983extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
984 uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
6772fb67 985extern int dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *val);
34dc7c2f
BB
986extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
987 int maxlen, boolean_t *conflict);
988extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
989 uint64_t *idp, uint64_t *offp);
9babb374 990
428870ff 991typedef int objset_used_cb_t(dmu_object_type_t bonustype,
9c5167d1 992 void *bonus, uint64_t *userp, uint64_t *groupp, uint64_t *projectp);
9babb374
BB
993extern void dmu_objset_register_type(dmu_objset_type_t ost,
994 objset_used_cb_t *cb);
34dc7c2f
BB
995extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
996extern void *dmu_objset_get_user(objset_t *os);
997
998/*
999 * Return the txg number for the given assigned transaction.
1000 */
1001uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
1002
1003/*
1004 * Synchronous write.
1005 * If a parent zio is provided this function initiates a write on the
1006 * provided buffer as a child of the parent zio.
1007 * In the absence of a parent zio, the write is completed synchronously.
1008 * At write completion, blk is filled with the bp of the written block.
1009 * Note that while the data covered by this function will be on stable
1010 * storage when the write completes this new data does not become a
1011 * permanent part of the file until the associated transaction commits.
1012 */
428870ff
BB
1013
1014/*
1015 * {zfs,zvol,ztest}_get_done() args
1016 */
1017typedef struct zgd {
1ce23dca 1018 struct lwb *zgd_lwb;
428870ff
BB
1019 struct blkptr *zgd_bp;
1020 dmu_buf_t *zgd_db;
1021 struct rl *zgd_rl;
1022 void *zgd_private;
1023} zgd_t;
1024
1025typedef void dmu_sync_cb_t(zgd_t *arg, int error);
1026int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
34dc7c2f
BB
1027
1028/*
1029 * Find the next hole or data block in file starting at *off
1030 * Return found offset in *off. Return ESRCH for end of file.
1031 */
1032int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
1033 uint64_t *off);
1034
1035/*
1036 * Initial setup and final teardown.
1037 */
1038extern void dmu_init(void);
1039extern void dmu_fini(void);
1040
1041typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
1042 uint64_t object, uint64_t offset, int len);
1043void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
1044 dmu_traverse_cb_t cb, void *arg);
1045
13fe0198
MA
1046int dmu_diff(const char *tosnap_name, const char *fromsnap_name,
1047 struct vnode *vp, offset_t *offp);
572e2857 1048
34dc7c2f
BB
1049/* CRC64 table */
1050#define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */
1051extern uint64_t zfs_crc64_table[256];
1052
1053#ifdef __cplusplus
1054}
1055#endif
1056
1057#endif /* _SYS_DMU_H */