]>
Commit | Line | Data |
---|---|---|
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 | |
1d3ba0bf | 9 | * or https://opensource.org/licenses/CDDL-1.0. |
34dc7c2f BB |
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. |
196bee4c | 23 | * Copyright (c) 2011, 2020 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. |
cc99f275 | 29 | * Copyright (c) 2017, Intel Corporation. |
5405be03 | 30 | * Copyright (c) 2022 Hewlett Packard Enterprise Development LP. |
34dc7c2f BB |
31 | */ |
32 | ||
428870ff BB |
33 | /* Portions Copyright 2010 Robert Milkowski */ |
34 | ||
34dc7c2f BB |
35 | #ifndef _SYS_DMU_H |
36 | #define _SYS_DMU_H | |
37 | ||
34dc7c2f BB |
38 | /* |
39 | * This file describes the interface that the DMU provides for its | |
40 | * consumers. | |
41 | * | |
42 | * The DMU also interacts with the SPA. That interface is described in | |
43 | * dmu_spa.h. | |
44 | */ | |
45 | ||
0c66c32d | 46 | #include <sys/zfs_context.h> |
34dc7c2f | 47 | #include <sys/inttypes.h> |
34dc7c2f | 48 | #include <sys/cred.h> |
6f1ffb06 | 49 | #include <sys/fs/zfs.h> |
2aa34383 | 50 | #include <sys/zio_compress.h> |
fcff0f35 | 51 | #include <sys/zio_priority.h> |
59e6e7ca | 52 | #include <sys/uio.h> |
da92d5cb | 53 | #include <sys/zfs_file.h> |
34dc7c2f BB |
54 | |
55 | #ifdef __cplusplus | |
56 | extern "C" { | |
57 | #endif | |
58 | ||
34dc7c2f BB |
59 | struct page; |
60 | struct vnode; | |
61 | struct spa; | |
62 | struct zilog; | |
63 | struct zio; | |
64 | struct blkptr; | |
65 | struct zap_cursor; | |
66 | struct dsl_dataset; | |
67 | struct dsl_pool; | |
68 | struct dnode; | |
69 | struct drr_begin; | |
70 | struct drr_end; | |
5dbd68a3 | 71 | struct zbookmark_phys; |
34dc7c2f BB |
72 | struct spa; |
73 | struct nvlist; | |
9babb374 | 74 | struct arc_buf; |
428870ff BB |
75 | struct zio_prop; |
76 | struct sa_handle; | |
b5256303 | 77 | struct dsl_crypto_params; |
5d43cc9a | 78 | struct locked_range; |
34dc7c2f BB |
79 | |
80 | typedef struct objset objset_t; | |
81 | typedef struct dmu_tx dmu_tx_t; | |
82 | typedef struct dsl_dir dsl_dir_t; | |
2bce8049 | 83 | typedef struct dnode dnode_t; |
34dc7c2f | 84 | |
9ae529ec CS |
85 | typedef enum dmu_object_byteswap { |
86 | DMU_BSWAP_UINT8, | |
87 | DMU_BSWAP_UINT16, | |
88 | DMU_BSWAP_UINT32, | |
89 | DMU_BSWAP_UINT64, | |
90 | DMU_BSWAP_ZAP, | |
91 | DMU_BSWAP_DNODE, | |
92 | DMU_BSWAP_OBJSET, | |
93 | DMU_BSWAP_ZNODE, | |
94 | DMU_BSWAP_OLDACL, | |
95 | DMU_BSWAP_ACL, | |
96 | /* | |
97 | * Allocating a new byteswap type number makes the on-disk format | |
98 | * incompatible with any other format that uses the same number. | |
99 | * | |
100 | * Data can usually be structured to work with one of the | |
101 | * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. | |
102 | */ | |
103 | DMU_BSWAP_NUMFUNCS | |
104 | } dmu_object_byteswap_t; | |
105 | ||
106 | #define DMU_OT_NEWTYPE 0x80 | |
107 | #define DMU_OT_METADATA 0x40 | |
b5256303 TC |
108 | #define DMU_OT_ENCRYPTED 0x20 |
109 | #define DMU_OT_BYTESWAP_MASK 0x1f | |
9ae529ec CS |
110 | |
111 | /* | |
112 | * Defines a uint8_t object type. Object types specify if the data | |
113 | * in the object is metadata (boolean) and how to byteswap the data | |
2e5dc449 MA |
114 | * (dmu_object_byteswap_t). All of the types created by this method |
115 | * are cached in the dbuf metadata cache. | |
9ae529ec | 116 | */ |
b5256303 | 117 | #define DMU_OT(byteswap, metadata, encrypted) \ |
9ae529ec CS |
118 | (DMU_OT_NEWTYPE | \ |
119 | ((metadata) ? DMU_OT_METADATA : 0) | \ | |
b5256303 | 120 | ((encrypted) ? DMU_OT_ENCRYPTED : 0) | \ |
9ae529ec CS |
121 | ((byteswap) & DMU_OT_BYTESWAP_MASK)) |
122 | ||
123 | #define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \ | |
124 | ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \ | |
125 | (ot) < DMU_OT_NUMTYPES) | |
126 | ||
2e5dc449 MA |
127 | #define DMU_OT_IS_METADATA_CACHED(ot) (((ot) & DMU_OT_NEWTYPE) ? \ |
128 | B_TRUE : dmu_ot[(ot)].ot_dbuf_metadata_cache) | |
129 | ||
5c27ec10 MA |
130 | /* |
131 | * MDB doesn't have dmu_ot; it defines these macros itself. | |
132 | */ | |
133 | #ifndef ZFS_MDB | |
134 | #define DMU_OT_IS_METADATA_IMPL(ot) (dmu_ot[ot].ot_metadata) | |
135 | #define DMU_OT_IS_ENCRYPTED_IMPL(ot) (dmu_ot[ot].ot_encrypt) | |
136 | #define DMU_OT_BYTESWAP_IMPL(ot) (dmu_ot[ot].ot_byteswap) | |
137 | #endif | |
138 | ||
9ae529ec | 139 | #define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \ |
57240735 | 140 | (((ot) & DMU_OT_METADATA) != 0) : \ |
5c27ec10 | 141 | DMU_OT_IS_METADATA_IMPL(ot)) |
9ae529ec | 142 | |
cc99f275 DB |
143 | #define DMU_OT_IS_DDT(ot) \ |
144 | ((ot) == DMU_OT_DDT_ZAP) | |
145 | ||
5405be03 A |
146 | #define DMU_OT_IS_CRITICAL(ot) \ |
147 | (DMU_OT_IS_METADATA(ot) && \ | |
148 | (ot) != DMU_OT_DNODE && \ | |
149 | (ot) != DMU_OT_DIRECTORY_CONTENTS && \ | |
150 | (ot) != DMU_OT_SA) | |
151 | ||
cc99f275 DB |
152 | /* Note: ztest uses DMU_OT_UINT64_OTHER as a proxy for file blocks */ |
153 | #define DMU_OT_IS_FILE(ot) \ | |
154 | ((ot) == DMU_OT_PLAIN_FILE_CONTENTS || (ot) == DMU_OT_UINT64_OTHER) | |
155 | ||
b5256303 | 156 | #define DMU_OT_IS_ENCRYPTED(ot) (((ot) & DMU_OT_NEWTYPE) ? \ |
57240735 | 157 | (((ot) & DMU_OT_ENCRYPTED) != 0) : \ |
5c27ec10 | 158 | DMU_OT_IS_ENCRYPTED_IMPL(ot)) |
b5256303 | 159 | |
9b67f605 MA |
160 | /* |
161 | * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't | |
162 | * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill | |
163 | * is repurposed for embedded BPs. | |
164 | */ | |
165 | #define DMU_OT_HAS_FILL(ot) \ | |
166 | ((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET) | |
167 | ||
9ae529ec CS |
168 | #define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \ |
169 | ((ot) & DMU_OT_BYTESWAP_MASK) : \ | |
5c27ec10 | 170 | DMU_OT_BYTESWAP_IMPL(ot)) |
9ae529ec | 171 | |
34dc7c2f BB |
172 | typedef enum dmu_object_type { |
173 | DMU_OT_NONE, | |
174 | /* general: */ | |
175 | DMU_OT_OBJECT_DIRECTORY, /* ZAP */ | |
176 | DMU_OT_OBJECT_ARRAY, /* UINT64 */ | |
177 | DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ | |
178 | DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ | |
428870ff BB |
179 | DMU_OT_BPOBJ, /* UINT64 */ |
180 | DMU_OT_BPOBJ_HDR, /* UINT64 */ | |
34dc7c2f BB |
181 | /* spa: */ |
182 | DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ | |
183 | DMU_OT_SPACE_MAP, /* UINT64 */ | |
184 | /* zil: */ | |
185 | DMU_OT_INTENT_LOG, /* UINT64 */ | |
186 | /* dmu: */ | |
187 | DMU_OT_DNODE, /* DNODE */ | |
188 | DMU_OT_OBJSET, /* OBJSET */ | |
189 | /* dsl: */ | |
190 | DMU_OT_DSL_DIR, /* UINT64 */ | |
191 | DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ | |
192 | DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ | |
193 | DMU_OT_DSL_PROPS, /* ZAP */ | |
194 | DMU_OT_DSL_DATASET, /* UINT64 */ | |
195 | /* zpl: */ | |
196 | DMU_OT_ZNODE, /* ZNODE */ | |
197 | DMU_OT_OLDACL, /* Old ACL */ | |
198 | DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ | |
199 | DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ | |
200 | DMU_OT_MASTER_NODE, /* ZAP */ | |
201 | DMU_OT_UNLINKED_SET, /* ZAP */ | |
202 | /* zvol: */ | |
203 | DMU_OT_ZVOL, /* UINT8 */ | |
204 | DMU_OT_ZVOL_PROP, /* ZAP */ | |
205 | /* other; for testing only! */ | |
206 | DMU_OT_PLAIN_OTHER, /* UINT8 */ | |
207 | DMU_OT_UINT64_OTHER, /* UINT64 */ | |
208 | DMU_OT_ZAP_OTHER, /* ZAP */ | |
209 | /* new object types: */ | |
210 | DMU_OT_ERROR_LOG, /* ZAP */ | |
211 | DMU_OT_SPA_HISTORY, /* UINT8 */ | |
212 | DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ | |
213 | DMU_OT_POOL_PROPS, /* ZAP */ | |
214 | DMU_OT_DSL_PERMS, /* ZAP */ | |
215 | DMU_OT_ACL, /* ACL */ | |
216 | DMU_OT_SYSACL, /* SYSACL */ | |
217 | DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ | |
218 | DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ | |
b128c09f | 219 | DMU_OT_NEXT_CLONES, /* ZAP */ |
428870ff | 220 | DMU_OT_SCAN_QUEUE, /* ZAP */ |
9babb374 BB |
221 | DMU_OT_USERGROUP_USED, /* ZAP */ |
222 | DMU_OT_USERGROUP_QUOTA, /* ZAP */ | |
45d1cae3 | 223 | DMU_OT_USERREFS, /* ZAP */ |
428870ff BB |
224 | DMU_OT_DDT_ZAP, /* ZAP */ |
225 | DMU_OT_DDT_STATS, /* ZAP */ | |
226 | DMU_OT_SA, /* System attr */ | |
227 | DMU_OT_SA_MASTER_NODE, /* ZAP */ | |
228 | DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ | |
229 | DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ | |
230 | DMU_OT_SCAN_XLATE, /* ZAP */ | |
231 | DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ | |
232 | DMU_OT_DEADLIST, /* ZAP */ | |
233 | DMU_OT_DEADLIST_HDR, /* UINT64 */ | |
234 | DMU_OT_DSL_CLONES, /* ZAP */ | |
235 | DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */ | |
9ae529ec CS |
236 | /* |
237 | * Do not allocate new object types here. Doing so makes the on-disk | |
238 | * format incompatible with any other format that uses the same object | |
239 | * type number. | |
240 | * | |
241 | * When creating an object which does not have one of the above types | |
242 | * use the DMU_OTN_* type with the correct byteswap and metadata | |
243 | * values. | |
244 | * | |
245 | * The DMU_OTN_* types do not have entries in the dmu_ot table, | |
cc99f275 | 246 | * use the DMU_OT_IS_METADATA() and DMU_OT_BYTESWAP() macros instead |
9ae529ec CS |
247 | * of indexing into dmu_ot directly (this works for both DMU_OT_* types |
248 | * and DMU_OTN_* types). | |
249 | */ | |
250 | DMU_OT_NUMTYPES, | |
251 | ||
252 | /* | |
253 | * Names for valid types declared with DMU_OT(). | |
254 | */ | |
b5256303 TC |
255 | DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_FALSE), |
256 | DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_FALSE), | |
257 | DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_FALSE), | |
258 | DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_FALSE), | |
259 | DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_FALSE), | |
260 | DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_FALSE), | |
261 | DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_FALSE), | |
262 | DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_FALSE), | |
263 | DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_FALSE), | |
264 | DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_FALSE), | |
265 | ||
266 | DMU_OTN_UINT8_ENC_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_TRUE), | |
267 | DMU_OTN_UINT8_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_TRUE), | |
268 | DMU_OTN_UINT16_ENC_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_TRUE), | |
269 | DMU_OTN_UINT16_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_TRUE), | |
270 | DMU_OTN_UINT32_ENC_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_TRUE), | |
271 | DMU_OTN_UINT32_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_TRUE), | |
272 | DMU_OTN_UINT64_ENC_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_TRUE), | |
273 | DMU_OTN_UINT64_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_TRUE), | |
274 | DMU_OTN_ZAP_ENC_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_TRUE), | |
275 | DMU_OTN_ZAP_ENC_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_TRUE), | |
34dc7c2f BB |
276 | } dmu_object_type_t; |
277 | ||
0735ecb3 PS |
278 | /* |
279 | * These flags are intended to be used to specify the "txg_how" | |
280 | * parameter when calling the dmu_tx_assign() function. See the comment | |
281 | * above dmu_tx_assign() for more details on the meaning of these flags. | |
282 | */ | |
283 | #define TXG_NOWAIT (0ULL) | |
284 | #define TXG_WAIT (1ULL<<0) | |
285 | #define TXG_NOTHROTTLE (1ULL<<1) | |
13fe0198 | 286 | |
34dc7c2f BB |
287 | void byteswap_uint64_array(void *buf, size_t size); |
288 | void byteswap_uint32_array(void *buf, size_t size); | |
289 | void byteswap_uint16_array(void *buf, size_t size); | |
290 | void byteswap_uint8_array(void *buf, size_t size); | |
291 | void zap_byteswap(void *buf, size_t size); | |
292 | void zfs_oldacl_byteswap(void *buf, size_t size); | |
293 | void zfs_acl_byteswap(void *buf, size_t size); | |
294 | void zfs_znode_byteswap(void *buf, size_t size); | |
295 | ||
34dc7c2f BB |
296 | #define DS_FIND_SNAPSHOTS (1<<0) |
297 | #define DS_FIND_CHILDREN (1<<1) | |
9c43027b | 298 | #define DS_FIND_SERIALIZE (1<<2) |
34dc7c2f BB |
299 | |
300 | /* | |
301 | * The maximum number of bytes that can be accessed as part of one | |
302 | * operation, including metadata. | |
303 | */ | |
f1512ee6 | 304 | #define DMU_MAX_ACCESS (64 * 1024 * 1024) /* 64MB */ |
b128c09f | 305 | #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */ |
34dc7c2f | 306 | |
9babb374 BB |
307 | #define DMU_USERUSED_OBJECT (-1ULL) |
308 | #define DMU_GROUPUSED_OBJECT (-2ULL) | |
9c5167d1 | 309 | #define DMU_PROJECTUSED_OBJECT (-3ULL) |
9babb374 | 310 | |
1de321e6 | 311 | /* |
9c5167d1 | 312 | * Zap prefix for object accounting in DMU_{USER,GROUP,PROJECT}USED_OBJECT. |
1de321e6 JX |
313 | */ |
314 | #define DMU_OBJACCT_PREFIX "obj-" | |
315 | #define DMU_OBJACCT_PREFIX_LEN 4 | |
316 | ||
428870ff BB |
317 | /* |
318 | * artificial blkids for bonus buffer and spill blocks | |
319 | */ | |
320 | #define DMU_BONUS_BLKID (-1ULL) | |
321 | #define DMU_SPILL_BLKID (-2ULL) | |
b5256303 | 322 | |
34dc7c2f BB |
323 | /* |
324 | * Public routines to create, destroy, open, and close objsets. | |
325 | */ | |
b5256303 TC |
326 | typedef void dmu_objset_create_sync_func_t(objset_t *os, void *arg, |
327 | cred_t *cr, dmu_tx_t *tx); | |
328 | ||
a926aab9 | 329 | int dmu_objset_hold(const char *name, const void *tag, objset_t **osp); |
428870ff | 330 | int dmu_objset_own(const char *name, dmu_objset_type_t type, |
a926aab9 AZ |
331 | boolean_t readonly, boolean_t key_required, const void *tag, |
332 | objset_t **osp); | |
333 | void dmu_objset_rele(objset_t *os, const void *tag); | |
334 | void dmu_objset_disown(objset_t *os, boolean_t key_required, const void *tag); | |
428870ff BB |
335 | int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp); |
336 | ||
13fe0198 | 337 | void dmu_objset_evict_dbufs(objset_t *os); |
428870ff | 338 | int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, |
b5256303 TC |
339 | struct dsl_crypto_params *dcp, dmu_objset_create_sync_func_t func, |
340 | void *arg); | |
13fe0198 MA |
341 | int dmu_objset_clone(const char *name, const char *origin); |
342 | int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer, | |
6f1ffb06 | 343 | struct nvlist *errlist); |
6f1ffb06 | 344 | int dmu_objset_snapshot_one(const char *fsname, const char *snapname); |
5df7e9d8 | 345 | int dmu_objset_find(const char *name, int func(const char *, void *), void *arg, |
34dc7c2f BB |
346 | int flags); |
347 | void dmu_objset_byteswap(void *buf, size_t size); | |
13fe0198 MA |
348 | int dsl_dataset_rename_snapshot(const char *fsname, |
349 | const char *oldsnapname, const char *newsnapname, boolean_t recursive); | |
34dc7c2f BB |
350 | |
351 | typedef struct dmu_buf { | |
352 | uint64_t db_object; /* object that this buffer is part of */ | |
353 | uint64_t db_offset; /* byte offset in this object */ | |
354 | uint64_t db_size; /* size of buffer in bytes */ | |
355 | void *db_data; /* data in buffer */ | |
356 | } dmu_buf_t; | |
357 | ||
34dc7c2f BB |
358 | /* |
359 | * The names of zap entries in the DIRECTORY_OBJECT of the MOS. | |
360 | */ | |
361 | #define DMU_POOL_DIRECTORY_OBJECT 1 | |
362 | #define DMU_POOL_CONFIG "config" | |
9ae529ec CS |
363 | #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write" |
364 | #define DMU_POOL_FEATURES_FOR_READ "features_for_read" | |
365 | #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions" | |
b0bc7a84 | 366 | #define DMU_POOL_FEATURE_ENABLED_TXG "feature_enabled_txg" |
34dc7c2f | 367 | #define DMU_POOL_ROOT_DATASET "root_dataset" |
428870ff | 368 | #define DMU_POOL_SYNC_BPOBJ "sync_bplist" |
34dc7c2f BB |
369 | #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" |
370 | #define DMU_POOL_ERRLOG_LAST "errlog_last" | |
371 | #define DMU_POOL_SPARES "spares" | |
372 | #define DMU_POOL_DEFLATE "deflate" | |
373 | #define DMU_POOL_HISTORY "history" | |
374 | #define DMU_POOL_PROPS "pool_props" | |
375 | #define DMU_POOL_L2CACHE "l2cache" | |
428870ff BB |
376 | #define DMU_POOL_TMP_USERREFS "tmp_userrefs" |
377 | #define DMU_POOL_DDT "DDT-%s-%s-%s" | |
378 | #define DMU_POOL_DDT_STATS "DDT-statistics" | |
379 | #define DMU_POOL_CREATION_VERSION "creation_version" | |
380 | #define DMU_POOL_SCAN "scan" | |
381 | #define DMU_POOL_FREE_BPOBJ "free_bpobj" | |
9ae529ec | 382 | #define DMU_POOL_BPTREE_OBJ "bptree_obj" |
753c3839 | 383 | #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj" |
3c67d83a | 384 | #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" |
e0ab3ab5 | 385 | #define DMU_POOL_VDEV_ZAP_MAP "com.delphix:vdev_zap_map" |
a1d477c2 MA |
386 | #define DMU_POOL_REMOVING "com.delphix:removing" |
387 | #define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj" | |
388 | #define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect" | |
d2734cce | 389 | #define DMU_POOL_ZPOOL_CHECKPOINT "com.delphix:zpool_checkpoint" |
93e28d66 | 390 | #define DMU_POOL_LOG_SPACEMAP_ZAP "com.delphix:log_spacemap_zap" |
37f03da8 | 391 | #define DMU_POOL_DELETED_CLONES "com.delphix:deleted_clones" |
b128c09f | 392 | |
34dc7c2f BB |
393 | /* |
394 | * Allocate an object from this objset. The range of object numbers | |
395 | * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode. | |
396 | * | |
397 | * The transaction must be assigned to a txg. The newly allocated | |
398 | * object will be "held" in the transaction (ie. you can modify the | |
399 | * newly allocated object in this transaction). | |
400 | * | |
401 | * dmu_object_alloc() chooses an object and returns it in *objectp. | |
402 | * | |
403 | * dmu_object_claim() allocates a specific object number. If that | |
404 | * number is already allocated, it fails and returns EEXIST. | |
405 | * | |
406 | * Return 0 on success, or ENOSPC or EEXIST as specified above. | |
407 | */ | |
408 | uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot, | |
409 | int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); | |
3a549dc7 MA |
410 | uint64_t dmu_object_alloc_ibs(objset_t *os, dmu_object_type_t ot, int blocksize, |
411 | int indirect_blockshift, | |
412 | dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx); | |
50c957f7 NB |
413 | uint64_t dmu_object_alloc_dnsize(objset_t *os, dmu_object_type_t ot, |
414 | int blocksize, dmu_object_type_t bonus_type, int bonus_len, | |
415 | int dnodesize, dmu_tx_t *tx); | |
6955b401 BB |
416 | uint64_t dmu_object_alloc_hold(objset_t *os, dmu_object_type_t ot, |
417 | int blocksize, int indirect_blockshift, dmu_object_type_t bonustype, | |
a926aab9 | 418 | int bonuslen, int dnodesize, dnode_t **allocated_dnode, const void *tag, |
6955b401 | 419 | dmu_tx_t *tx); |
34dc7c2f BB |
420 | int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot, |
421 | int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); | |
50c957f7 NB |
422 | int dmu_object_claim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot, |
423 | int blocksize, dmu_object_type_t bonus_type, int bonus_len, | |
424 | int dnodesize, dmu_tx_t *tx); | |
34dc7c2f | 425 | int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot, |
6c59307a | 426 | int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp); |
50c957f7 NB |
427 | int dmu_object_reclaim_dnsize(objset_t *os, uint64_t object, |
428 | dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype, | |
caf9dd20 BB |
429 | int bonuslen, int dnodesize, boolean_t keep_spill, dmu_tx_t *tx); |
430 | int dmu_object_rm_spill(objset_t *os, uint64_t object, dmu_tx_t *tx); | |
34dc7c2f BB |
431 | |
432 | /* | |
433 | * Free an object from this objset. | |
434 | * | |
435 | * The object's data will be freed as well (ie. you don't need to call | |
436 | * dmu_free(object, 0, -1, tx)). | |
437 | * | |
438 | * The object need not be held in the transaction. | |
439 | * | |
440 | * If there are any holds on this object's buffers (via dmu_buf_hold()), | |
441 | * or tx holds on the object (via dmu_tx_hold_object()), you can not | |
442 | * free it; it fails and returns EBUSY. | |
443 | * | |
444 | * If the object is not allocated, it fails and returns ENOENT. | |
445 | * | |
446 | * Return 0 on success, or EBUSY or ENOENT as specified above. | |
447 | */ | |
448 | int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx); | |
449 | ||
450 | /* | |
451 | * Find the next allocated or free object. | |
452 | * | |
453 | * The objectp parameter is in-out. It will be updated to be the next | |
454 | * object which is allocated. Ignore objects which have not been | |
455 | * modified since txg. | |
456 | * | |
457 | * XXX Can only be called on a objset with no dirty data. | |
458 | * | |
459 | * Returns 0 on success, or ENOENT if there are no more objects. | |
460 | */ | |
461 | int dmu_object_next(objset_t *os, uint64_t *objectp, | |
462 | boolean_t hole, uint64_t txg); | |
463 | ||
b5256303 TC |
464 | /* |
465 | * Set the number of levels on a dnode. nlevels must be greater than the | |
466 | * current number of levels or an EINVAL will be returned. | |
467 | */ | |
468 | int dmu_object_set_nlevels(objset_t *os, uint64_t object, int nlevels, | |
469 | dmu_tx_t *tx); | |
470 | ||
34dc7c2f BB |
471 | /* |
472 | * Set the data blocksize for an object. | |
473 | * | |
cf7c5a03 | 474 | * The object cannot have any blocks allocated beyond the first. If |
34dc7c2f BB |
475 | * the first block is allocated already, the new size must be greater |
476 | * than the current block size. If these conditions are not met, | |
477 | * ENOTSUP will be returned. | |
478 | * | |
479 | * Returns 0 on success, or EBUSY if there are any holds on the object | |
480 | * contents, or ENOTSUP as described above. | |
481 | */ | |
482 | int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, | |
483 | int ibs, dmu_tx_t *tx); | |
484 | ||
ae76f45c TC |
485 | /* |
486 | * Manually set the maxblkid on a dnode. This will adjust nlevels accordingly | |
369aa501 TC |
487 | * to accommodate the change. When calling this function, the caller must |
488 | * ensure that the object's nlevels can sufficiently support the new maxblkid. | |
ae76f45c TC |
489 | */ |
490 | int dmu_object_set_maxblkid(objset_t *os, uint64_t object, uint64_t maxblkid, | |
491 | dmu_tx_t *tx); | |
492 | ||
34dc7c2f BB |
493 | /* |
494 | * Set the checksum property on a dnode. The new checksum algorithm will | |
495 | * apply to all newly written blocks; existing blocks will not be affected. | |
496 | */ | |
497 | void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, | |
498 | dmu_tx_t *tx); | |
499 | ||
500 | /* | |
501 | * Set the compress property on a dnode. The new compression algorithm will | |
502 | * apply to all newly written blocks; existing blocks will not be affected. | |
503 | */ | |
504 | void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, | |
505 | dmu_tx_t *tx); | |
506 | ||
9b840763 | 507 | void dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset, |
9b67f605 MA |
508 | void *data, uint8_t etype, uint8_t comp, int uncompressed_size, |
509 | int compressed_size, int byteorder, dmu_tx_t *tx); | |
30af21b0 PD |
510 | void dmu_redact(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, |
511 | dmu_tx_t *tx); | |
9b67f605 | 512 | |
34dc7c2f | 513 | /* |
428870ff | 514 | * Decide how to write a block: checksum, compression, number of copies, etc. |
34dc7c2f | 515 | */ |
428870ff BB |
516 | #define WP_NOFILL 0x1 |
517 | #define WP_DMU_SYNC 0x2 | |
518 | #define WP_SPILL 0x4 | |
519 | ||
82644107 MA |
520 | void dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, |
521 | struct zio_prop *zp); | |
b5256303 | 522 | |
34dc7c2f BB |
523 | /* |
524 | * The bonus data is accessed more or less like a regular buffer. | |
525 | * You must dmu_bonus_hold() to get the buffer, which will give you a | |
526 | * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus | |
a1d477c2 MA |
527 | * data. As with any normal buffer, you must call dmu_buf_will_dirty() |
528 | * before modifying it, and the | |
34dc7c2f BB |
529 | * object must be held in an assigned transaction before calling |
530 | * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus | |
4e95cc99 | 531 | * buffer as well. You must release what you hold with dmu_buf_rele(). |
e49f1e20 WA |
532 | * |
533 | * Returns ENOENT, EIO, or 0. | |
34dc7c2f | 534 | */ |
a926aab9 AZ |
535 | int dmu_bonus_hold(objset_t *os, uint64_t object, const void *tag, |
536 | dmu_buf_t **dbp); | |
537 | int dmu_bonus_hold_by_dnode(dnode_t *dn, const void *tag, dmu_buf_t **dbp, | |
6955b401 | 538 | uint32_t flags); |
34dc7c2f BB |
539 | int dmu_bonus_max(void); |
540 | int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *); | |
428870ff | 541 | int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *); |
572e2857 | 542 | dmu_object_type_t dmu_get_bonustype(dmu_buf_t *); |
428870ff BB |
543 | int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *); |
544 | ||
545 | /* | |
546 | * Special spill buffer support used by "SA" framework | |
547 | */ | |
548 | ||
a926aab9 | 549 | int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, uint32_t flags, const void *tag, |
e7504d7a | 550 | dmu_buf_t **dbp); |
2bce8049 | 551 | int dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags, |
a926aab9 AZ |
552 | const void *tag, dmu_buf_t **dbp); |
553 | int dmu_spill_hold_existing(dmu_buf_t *bonus, const void *tag, dmu_buf_t **dbp); | |
34dc7c2f BB |
554 | |
555 | /* | |
556 | * Obtain the DMU buffer from the specified object which contains the | |
557 | * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so | |
558 | * that it will remain in memory. You must release the hold with | |
4e95cc99 JE |
559 | * dmu_buf_rele(). You must not access the dmu_buf_t after releasing |
560 | * what you hold. You must have a hold on any dmu_buf_t* you pass to the DMU. | |
34dc7c2f BB |
561 | * |
562 | * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill | |
563 | * on the returned buffer before reading or writing the buffer's | |
564 | * db_data. The comments for those routines describe what particular | |
565 | * operations are valid after calling them. | |
566 | * | |
567 | * The object number must be a valid, allocated object number. | |
568 | */ | |
569 | int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, | |
a926aab9 | 570 | const void *tag, dmu_buf_t **, int flags); |
afbc6179 | 571 | int dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset, |
a926aab9 AZ |
572 | uint64_t length, int read, const void *tag, int *numbufsp, |
573 | dmu_buf_t ***dbpp); | |
2bce8049 | 574 | int dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset, |
a926aab9 | 575 | const void *tag, dmu_buf_t **dbp, int flags); |
af1698f5 | 576 | int dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset, |
a926aab9 | 577 | uint64_t length, boolean_t read, const void *tag, int *numbufsp, |
af1698f5 | 578 | dmu_buf_t ***dbpp, uint32_t flags); |
6ebebace JG |
579 | /* |
580 | * Add a reference to a dmu buffer that has already been held via | |
581 | * dmu_buf_hold() in the current context. | |
582 | */ | |
dd66857d | 583 | void dmu_buf_add_ref(dmu_buf_t *db, const void *tag); |
6ebebace JG |
584 | |
585 | /* | |
586 | * Attempt to add a reference to a dmu buffer that is in an unknown state, | |
587 | * using a pointer that may have been invalidated by eviction processing. | |
588 | * The request will succeed if the passed in dbuf still represents the | |
589 | * same os/object/blkid, is ineligible for eviction, and has at least | |
590 | * one hold by a user other than the syncer. | |
591 | */ | |
592 | boolean_t dmu_buf_try_add_ref(dmu_buf_t *, objset_t *os, uint64_t object, | |
a926aab9 | 593 | uint64_t blkid, const void *tag); |
6ebebace | 594 | |
a926aab9 | 595 | void dmu_buf_rele(dmu_buf_t *db, const void *tag); |
34dc7c2f | 596 | uint64_t dmu_buf_refcount(dmu_buf_t *db); |
cd32e5db | 597 | uint64_t dmu_buf_user_refcount(dmu_buf_t *db); |
34dc7c2f BB |
598 | |
599 | /* | |
600 | * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a | |
601 | * range of an object. A pointer to an array of dmu_buf_t*'s is | |
602 | * returned (in *dbpp). | |
603 | * | |
604 | * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and | |
605 | * frees the array. The hold on the array of buffers MUST be released | |
606 | * with dmu_buf_rele_array. You can NOT release the hold on each buffer | |
607 | * individually with dmu_buf_rele. | |
608 | */ | |
609 | int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset, | |
a926aab9 | 610 | uint64_t length, boolean_t read, const void *tag, |
7f60329a | 611 | int *numbufsp, dmu_buf_t ***dbpp); |
a926aab9 | 612 | void dmu_buf_rele_array(dmu_buf_t **, int numbufs, const void *tag); |
34dc7c2f | 613 | |
0c66c32d JG |
614 | typedef void dmu_buf_evict_func_t(void *user_ptr); |
615 | ||
34dc7c2f | 616 | /* |
0c66c32d JG |
617 | * A DMU buffer user object may be associated with a dbuf for the |
618 | * duration of its lifetime. This allows the user of a dbuf (client) | |
619 | * to attach private data to a dbuf (e.g. in-core only data such as a | |
620 | * dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified | |
621 | * when that dbuf has been evicted. Clients typically respond to the | |
622 | * eviction notification by freeing their private data, thus ensuring | |
623 | * the same lifetime for both dbuf and private data. | |
34dc7c2f | 624 | * |
0c66c32d JG |
625 | * The mapping from a dmu_buf_user_t to any client private data is the |
626 | * client's responsibility. All current consumers of the API with private | |
627 | * data embed a dmu_buf_user_t as the first member of the structure for | |
628 | * their private data. This allows conversions between the two types | |
629 | * with a simple cast. Since the DMU buf user API never needs access | |
630 | * to the private data, other strategies can be employed if necessary | |
631 | * or convenient for the client (e.g. using container_of() to do the | |
632 | * conversion for private data that cannot have the dmu_buf_user_t as | |
633 | * its first member). | |
34dc7c2f | 634 | * |
0c66c32d JG |
635 | * Eviction callbacks are executed without the dbuf mutex held or any |
636 | * other type of mechanism to guarantee that the dbuf is still available. | |
637 | * For this reason, users must assume the dbuf has already been freed | |
638 | * and not reference the dbuf from the callback context. | |
34dc7c2f | 639 | * |
0c66c32d JG |
640 | * Users requesting "immediate eviction" are notified as soon as the dbuf |
641 | * is only referenced by dirty records (dirties == holds). Otherwise the | |
642 | * notification occurs after eviction processing for the dbuf begins. | |
34dc7c2f | 643 | */ |
0c66c32d JG |
644 | typedef struct dmu_buf_user { |
645 | /* | |
646 | * Asynchronous user eviction callback state. | |
647 | */ | |
648 | taskq_ent_t dbu_tqent; | |
649 | ||
39efbde7 GM |
650 | /* |
651 | * This instance's eviction function pointers. | |
652 | * | |
653 | * dbu_evict_func_sync is called synchronously and then | |
654 | * dbu_evict_func_async is executed asynchronously on a taskq. | |
655 | */ | |
656 | dmu_buf_evict_func_t *dbu_evict_func_sync; | |
657 | dmu_buf_evict_func_t *dbu_evict_func_async; | |
0c66c32d JG |
658 | #ifdef ZFS_DEBUG |
659 | /* | |
660 | * Pointer to user's dbuf pointer. NULL for clients that do | |
661 | * not associate a dbuf with their user data. | |
662 | * | |
663 | * The dbuf pointer is cleared upon eviction so as to catch | |
664 | * use-after-evict bugs in clients. | |
665 | */ | |
666 | dmu_buf_t **dbu_clear_on_evict_dbufp; | |
667 | #endif | |
668 | } dmu_buf_user_t; | |
669 | ||
34dc7c2f | 670 | /* |
0c66c32d JG |
671 | * Initialize the given dmu_buf_user_t instance with the eviction function |
672 | * evict_func, to be called when the user is evicted. | |
673 | * | |
674 | * NOTE: This function should only be called once on a given dmu_buf_user_t. | |
675 | * To allow enforcement of this, dbu must already be zeroed on entry. | |
34dc7c2f | 676 | */ |
0c66c32d | 677 | static inline void |
39efbde7 | 678 | dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func_sync, |
bfafe178 KE |
679 | dmu_buf_evict_func_t *evict_func_async, |
680 | dmu_buf_t **clear_on_evict_dbufp __maybe_unused) | |
0c66c32d | 681 | { |
39efbde7 GM |
682 | ASSERT(dbu->dbu_evict_func_sync == NULL); |
683 | ASSERT(dbu->dbu_evict_func_async == NULL); | |
684 | ||
685 | /* must have at least one evict func */ | |
686 | IMPLY(evict_func_sync == NULL, evict_func_async != NULL); | |
687 | dbu->dbu_evict_func_sync = evict_func_sync; | |
688 | dbu->dbu_evict_func_async = evict_func_async; | |
f467b05a | 689 | taskq_init_ent(&dbu->dbu_tqent); |
0c66c32d JG |
690 | #ifdef ZFS_DEBUG |
691 | dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp; | |
692 | #endif | |
693 | } | |
34dc7c2f BB |
694 | |
695 | /* | |
0c66c32d JG |
696 | * Attach user data to a dbuf and mark it for normal (when the dbuf's |
697 | * data is cleared or its reference count goes to zero) eviction processing. | |
698 | * | |
699 | * Returns NULL on success, or the existing user if another user currently | |
700 | * owns the buffer. | |
701 | */ | |
702 | void *dmu_buf_set_user(dmu_buf_t *db, dmu_buf_user_t *user); | |
703 | ||
704 | /* | |
705 | * Attach user data to a dbuf and mark it for immediate (its dirty and | |
706 | * reference counts are equal) eviction processing. | |
707 | * | |
708 | * Returns NULL on success, or the existing user if another user currently | |
709 | * owns the buffer. | |
710 | */ | |
711 | void *dmu_buf_set_user_ie(dmu_buf_t *db, dmu_buf_user_t *user); | |
712 | ||
713 | /* | |
714 | * Replace the current user of a dbuf. | |
715 | * | |
716 | * If given the current user of a dbuf, replaces the dbuf's user with | |
717 | * "new_user" and returns the user data pointer that was replaced. | |
718 | * Otherwise returns the current, and unmodified, dbuf user pointer. | |
719 | */ | |
720 | void *dmu_buf_replace_user(dmu_buf_t *db, | |
721 | dmu_buf_user_t *old_user, dmu_buf_user_t *new_user); | |
722 | ||
723 | /* | |
724 | * Remove the specified user data for a DMU buffer. | |
725 | * | |
726 | * Returns the user that was removed on success, or the current user if | |
727 | * another user currently owns the buffer. | |
728 | */ | |
729 | void *dmu_buf_remove_user(dmu_buf_t *db, dmu_buf_user_t *user); | |
730 | ||
731 | /* | |
732 | * Returns the user data (dmu_buf_user_t *) associated with this dbuf. | |
34dc7c2f BB |
733 | */ |
734 | void *dmu_buf_get_user(dmu_buf_t *db); | |
735 | ||
8bea9815 | 736 | objset_t *dmu_buf_get_objset(dmu_buf_t *db); |
2bce8049 MA |
737 | dnode_t *dmu_buf_dnode_enter(dmu_buf_t *db); |
738 | void dmu_buf_dnode_exit(dmu_buf_t *db); | |
8bea9815 | 739 | |
0c66c32d JG |
740 | /* Block until any in-progress dmu buf user evictions complete. */ |
741 | void dmu_buf_user_evict_wait(void); | |
742 | ||
03c6040b GW |
743 | /* |
744 | * Returns the blkptr associated with this dbuf, or NULL if not set. | |
745 | */ | |
746 | struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db); | |
747 | ||
34dc7c2f BB |
748 | /* |
749 | * Indicate that you are going to modify the buffer's data (db_data). | |
750 | * | |
751 | * The transaction (tx) must be assigned to a txg (ie. you've called | |
752 | * dmu_tx_assign()). The buffer's object must be held in the tx | |
753 | * (ie. you've called dmu_tx_hold_object(tx, db->db_object)). | |
754 | */ | |
755 | void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx); | |
a73e8fdb | 756 | boolean_t dmu_buf_is_dirty(dmu_buf_t *db, dmu_tx_t *tx); |
0c03d21a MA |
757 | void dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder, |
758 | const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_tx_t *tx); | |
34dc7c2f BB |
759 | |
760 | /* | |
761 | * You must create a transaction, then hold the objects which you will | |
762 | * (or might) modify as part of this transaction. Then you must assign | |
763 | * the transaction to a transaction group. Once the transaction has | |
764 | * been assigned, you can modify buffers which belong to held objects as | |
765 | * part of this transaction. You can't modify buffers before the | |
766 | * transaction has been assigned; you can't modify buffers which don't | |
767 | * belong to objects which this transaction holds; you can't hold | |
768 | * objects once the transaction has been assigned. You may hold an | |
769 | * object which you are going to free (with dmu_object_free()), but you | |
770 | * don't have to. | |
771 | * | |
772 | * You can abort the transaction before it has been assigned. | |
773 | * | |
774 | * Note that you may hold buffers (with dmu_buf_hold) at any time, | |
775 | * regardless of transaction state. | |
776 | */ | |
777 | ||
778 | #define DMU_NEW_OBJECT (-1ULL) | |
779 | #define DMU_OBJECT_END (-1ULL) | |
780 | ||
781 | dmu_tx_t *dmu_tx_create(objset_t *os); | |
782 | void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len); | |
0eef1bde | 783 | void dmu_tx_hold_write_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, |
784 | int len); | |
903c3613 BB |
785 | void dmu_tx_hold_append(dmu_tx_t *tx, uint64_t object, uint64_t off, int len); |
786 | void dmu_tx_hold_append_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, | |
787 | int len); | |
67a1b037 PJD |
788 | void dmu_tx_hold_clone_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, |
789 | int len); | |
34dc7c2f BB |
790 | void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, |
791 | uint64_t len); | |
0eef1bde | 792 | void dmu_tx_hold_free_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, |
793 | uint64_t len); | |
9babb374 | 794 | void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name); |
0eef1bde | 795 | void dmu_tx_hold_zap_by_dnode(dmu_tx_t *tx, dnode_t *dn, int add, |
796 | const char *name); | |
34dc7c2f | 797 | void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object); |
0eef1bde | 798 | void dmu_tx_hold_bonus_by_dnode(dmu_tx_t *tx, dnode_t *dn); |
428870ff BB |
799 | void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object); |
800 | void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow); | |
801 | void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size); | |
34dc7c2f | 802 | void dmu_tx_abort(dmu_tx_t *tx); |
0735ecb3 | 803 | int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how); |
34dc7c2f BB |
804 | void dmu_tx_wait(dmu_tx_t *tx); |
805 | void dmu_tx_commit(dmu_tx_t *tx); | |
19d55079 | 806 | void dmu_tx_mark_netfree(dmu_tx_t *tx); |
34dc7c2f | 807 | |
428870ff BB |
808 | /* |
809 | * To register a commit callback, dmu_tx_callback_register() must be called. | |
810 | * | |
811 | * dcb_data is a pointer to caller private data that is passed on as a | |
812 | * callback parameter. The caller is responsible for properly allocating and | |
813 | * freeing it. | |
814 | * | |
815 | * When registering a callback, the transaction must be already created, but | |
816 | * it cannot be committed or aborted. It can be assigned to a txg or not. | |
817 | * | |
818 | * The callback will be called after the transaction has been safely written | |
819 | * to stable storage and will also be called if the dmu_tx is aborted. | |
820 | * If there is any error which prevents the transaction from being committed to | |
821 | * disk, the callback will be called with a value of error != 0. | |
823d48bf | 822 | * |
823 | * When multiple callbacks are registered to the transaction, the callbacks | |
824 | * will be called in reverse order to let Lustre, the only user of commit | |
825 | * callback currently, take the fast path of its commit callback handling. | |
428870ff BB |
826 | */ |
827 | typedef void dmu_tx_callback_func_t(void *dcb_data, int error); | |
828 | ||
829 | void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func, | |
830 | void *dcb_data); | |
823d48bf | 831 | void dmu_tx_do_callbacks(list_t *cb_list, int error); |
428870ff | 832 | |
34dc7c2f BB |
833 | /* |
834 | * Free up the data blocks for a defined range of a file. If size is | |
9ae529ec | 835 | * -1, the range from offset to end-of-file is freed. |
34dc7c2f BB |
836 | */ |
837 | int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, | |
440a3eb9 | 838 | uint64_t size, dmu_tx_t *tx); |
b128c09f | 839 | int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, |
440a3eb9 | 840 | uint64_t size); |
b663a23d | 841 | int dmu_free_long_object(objset_t *os, uint64_t object); |
34dc7c2f BB |
842 | |
843 | /* | |
844 | * Convenience functions. | |
845 | * | |
846 | * Canfail routines will return 0 on success, or an errno if there is a | |
847 | * nonrecoverable I/O error. | |
848 | */ | |
9babb374 BB |
849 | #define DMU_READ_PREFETCH 0 /* prefetch */ |
850 | #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */ | |
b5256303 | 851 | #define DMU_READ_NO_DECRYPT 2 /* don't decrypt */ |
34dc7c2f | 852 | int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, |
9babb374 | 853 | void *buf, uint32_t flags); |
0eef1bde | 854 | int dmu_read_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, void *buf, |
855 | uint32_t flags); | |
34dc7c2f BB |
856 | void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, |
857 | const void *buf, dmu_tx_t *tx); | |
0eef1bde | 858 | void dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, |
859 | const void *buf, dmu_tx_t *tx); | |
b128c09f BB |
860 | void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, |
861 | dmu_tx_t *tx); | |
60101509 | 862 | #ifdef _KERNEL |
d0cd9a5c BA |
863 | int dmu_read_uio(objset_t *os, uint64_t object, zfs_uio_t *uio, uint64_t size); |
864 | int dmu_read_uio_dbuf(dmu_buf_t *zdb, zfs_uio_t *uio, uint64_t size); | |
865 | int dmu_read_uio_dnode(dnode_t *dn, zfs_uio_t *uio, uint64_t size); | |
866 | int dmu_write_uio(objset_t *os, uint64_t object, zfs_uio_t *uio, uint64_t size, | |
872e8d26 | 867 | dmu_tx_t *tx); |
d0cd9a5c | 868 | int dmu_write_uio_dbuf(dmu_buf_t *zdb, zfs_uio_t *uio, uint64_t size, |
872e8d26 | 869 | dmu_tx_t *tx); |
d0cd9a5c | 870 | int dmu_write_uio_dnode(dnode_t *dn, zfs_uio_t *uio, uint64_t size, |
5228cf01 | 871 | dmu_tx_t *tx); |
872e8d26 | 872 | #endif |
9babb374 BB |
873 | struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size); |
874 | void dmu_return_arcbuf(struct arc_buf *buf); | |
305781da | 875 | int dmu_assign_arcbuf_by_dnode(dnode_t *dn, uint64_t offset, |
440a3eb9 | 876 | struct arc_buf *buf, dmu_tx_t *tx); |
305781da | 877 | int dmu_assign_arcbuf_by_dbuf(dmu_buf_t *handle, uint64_t offset, |
440a3eb9 TC |
878 | struct arc_buf *buf, dmu_tx_t *tx); |
879 | #define dmu_assign_arcbuf dmu_assign_arcbuf_by_dbuf | |
fdc2d303 | 880 | extern uint_t zfs_max_recordsize; |
34dc7c2f BB |
881 | |
882 | /* | |
883 | * Asynchronously try to read in the data. | |
884 | */ | |
fcff0f35 PD |
885 | void dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset, |
886 | uint64_t len, enum zio_priority pri); | |
34dc7c2f BB |
887 | |
888 | typedef struct dmu_object_info { | |
428870ff | 889 | /* All sizes are in bytes unless otherwise indicated. */ |
34dc7c2f BB |
890 | uint32_t doi_data_block_size; |
891 | uint32_t doi_metadata_block_size; | |
34dc7c2f BB |
892 | dmu_object_type_t doi_type; |
893 | dmu_object_type_t doi_bonus_type; | |
428870ff | 894 | uint64_t doi_bonus_size; |
34dc7c2f BB |
895 | uint8_t doi_indirection; /* 2 = dnode->indirect->data */ |
896 | uint8_t doi_checksum; | |
897 | uint8_t doi_compress; | |
6c59307a MA |
898 | uint8_t doi_nblkptr; |
899 | uint8_t doi_pad[4]; | |
50c957f7 | 900 | uint64_t doi_dnodesize; |
428870ff BB |
901 | uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */ |
902 | uint64_t doi_max_offset; | |
903 | uint64_t doi_fill_count; /* number of non-empty blocks */ | |
34dc7c2f BB |
904 | } dmu_object_info_t; |
905 | ||
b01615d5 | 906 | typedef void (*const arc_byteswap_func_t)(void *buf, size_t size); |
34dc7c2f BB |
907 | |
908 | typedef struct dmu_object_type_info { | |
9ae529ec | 909 | dmu_object_byteswap_t ot_byteswap; |
34dc7c2f | 910 | boolean_t ot_metadata; |
2e5dc449 | 911 | boolean_t ot_dbuf_metadata_cache; |
b5256303 | 912 | boolean_t ot_encrypt; |
a926aab9 | 913 | const char *ot_name; |
34dc7c2f BB |
914 | } dmu_object_type_info_t; |
915 | ||
b01615d5 RY |
916 | typedef const struct dmu_object_byteswap_info { |
917 | arc_byteswap_func_t ob_func; | |
a926aab9 | 918 | const char *ob_name; |
9ae529ec CS |
919 | } dmu_object_byteswap_info_t; |
920 | ||
34dc7c2f | 921 | extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES]; |
a2d5643f | 922 | extern dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS]; |
34dc7c2f BB |
923 | |
924 | /* | |
925 | * Get information on a DMU object. | |
926 | * | |
927 | * Return 0 on success or ENOENT if object is not allocated. | |
928 | * | |
929 | * If doi is NULL, just indicates whether the object exists. | |
930 | */ | |
931 | int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi); | |
e0b0ca98 | 932 | void __dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi); |
e49f1e20 | 933 | /* Like dmu_object_info, but faster if you have a held dnode in hand. */ |
2bce8049 | 934 | void dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi); |
e49f1e20 | 935 | /* Like dmu_object_info, but faster if you have a held dbuf in hand. */ |
34dc7c2f | 936 | void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi); |
e49f1e20 WA |
937 | /* |
938 | * Like dmu_object_info_from_db, but faster still when you only care about | |
327000ce | 939 | * the size. |
e49f1e20 | 940 | */ |
34dc7c2f BB |
941 | void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, |
942 | u_longlong_t *nblk512); | |
943 | ||
50c957f7 NB |
944 | void dmu_object_dnsize_from_db(dmu_buf_t *db, int *dnsize); |
945 | ||
34dc7c2f BB |
946 | typedef struct dmu_objset_stats { |
947 | uint64_t dds_num_clones; /* number of clones of this */ | |
948 | uint64_t dds_creation_txg; | |
949 | uint64_t dds_guid; | |
950 | dmu_objset_type_t dds_type; | |
951 | uint8_t dds_is_snapshot; | |
952 | uint8_t dds_inconsistent; | |
30af21b0 | 953 | uint8_t dds_redacted; |
eca7b760 | 954 | char dds_origin[ZFS_MAX_DATASET_NAME_LEN]; |
34dc7c2f BB |
955 | } dmu_objset_stats_t; |
956 | ||
957 | /* | |
958 | * Get stats on a dataset. | |
959 | */ | |
960 | void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat); | |
961 | ||
962 | /* | |
963 | * Add entries to the nvlist for all the objset's properties. See | |
964 | * zfs_prop_table[] and zfs(1m) for details on the properties. | |
965 | */ | |
966 | void dmu_objset_stats(objset_t *os, struct nvlist *nv); | |
967 | ||
968 | /* | |
969 | * Get the space usage statistics for statvfs(). | |
970 | * | |
971 | * refdbytes is the amount of space "referenced" by this objset. | |
972 | * availbytes is the amount of space available to this objset, taking | |
973 | * into account quotas & reservations, assuming that no other objsets | |
974 | * use the space first. These values correspond to the 'referenced' and | |
975 | * 'available' properties, described in the zfs(1m) manpage. | |
976 | * | |
977 | * usedobjs and availobjs are the number of objects currently allocated, | |
978 | * and available. | |
979 | */ | |
980 | void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, | |
981 | uint64_t *usedobjsp, uint64_t *availobjsp); | |
982 | ||
983 | /* | |
984 | * The fsid_guid is a 56-bit ID that can change to avoid collisions. | |
985 | * (Contrast with the ds_guid which is a 64-bit ID that will never | |
986 | * change, so there is a small probability that it will collide.) | |
987 | */ | |
988 | uint64_t dmu_objset_fsid_guid(objset_t *os); | |
989 | ||
428870ff BB |
990 | /* |
991 | * Get the [cm]time for an objset's snapshot dir | |
992 | */ | |
6413c95f | 993 | inode_timespec_t dmu_objset_snap_cmtime(objset_t *os); |
428870ff | 994 | |
34dc7c2f BB |
995 | int dmu_objset_is_snapshot(objset_t *os); |
996 | ||
997 | extern struct spa *dmu_objset_spa(objset_t *os); | |
998 | extern struct zilog *dmu_objset_zil(objset_t *os); | |
999 | extern struct dsl_pool *dmu_objset_pool(objset_t *os); | |
1000 | extern struct dsl_dataset *dmu_objset_ds(objset_t *os); | |
1001 | extern void dmu_objset_name(objset_t *os, char *buf); | |
1002 | extern dmu_objset_type_t dmu_objset_type(objset_t *os); | |
1003 | extern uint64_t dmu_objset_id(objset_t *os); | |
50c957f7 | 1004 | extern uint64_t dmu_objset_dnodesize(objset_t *os); |
faf0f58c MA |
1005 | extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os); |
1006 | extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os); | |
37f03da8 | 1007 | extern int dmu_objset_blksize(objset_t *os); |
34dc7c2f BB |
1008 | extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name, |
1009 | uint64_t *id, uint64_t *offp, boolean_t *case_conflict); | |
6772fb67 | 1010 | extern int dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *val); |
4d55ea81 | 1011 | extern int dmu_snapshot_realname(objset_t *os, const char *name, char *real, |
34dc7c2f BB |
1012 | int maxlen, boolean_t *conflict); |
1013 | extern int dmu_dir_list_next(objset_t *os, int namelen, char *name, | |
1014 | uint64_t *idp, uint64_t *offp); | |
9babb374 | 1015 | |
7bcb7f08 MA |
1016 | typedef struct zfs_file_info { |
1017 | uint64_t zfi_user; | |
1018 | uint64_t zfi_group; | |
1019 | uint64_t zfi_project; | |
1020 | uint64_t zfi_generation; | |
1021 | } zfs_file_info_t; | |
1022 | ||
1023 | typedef int file_info_cb_t(dmu_object_type_t bonustype, const void *data, | |
1024 | struct zfs_file_info *zoi); | |
9babb374 | 1025 | extern void dmu_objset_register_type(dmu_objset_type_t ost, |
7bcb7f08 | 1026 | file_info_cb_t *cb); |
34dc7c2f BB |
1027 | extern void dmu_objset_set_user(objset_t *os, void *user_ptr); |
1028 | extern void *dmu_objset_get_user(objset_t *os); | |
1029 | ||
1030 | /* | |
1031 | * Return the txg number for the given assigned transaction. | |
1032 | */ | |
1033 | uint64_t dmu_tx_get_txg(dmu_tx_t *tx); | |
1034 | ||
1035 | /* | |
1036 | * Synchronous write. | |
1037 | * If a parent zio is provided this function initiates a write on the | |
1038 | * provided buffer as a child of the parent zio. | |
1039 | * In the absence of a parent zio, the write is completed synchronously. | |
1040 | * At write completion, blk is filled with the bp of the written block. | |
1041 | * Note that while the data covered by this function will be on stable | |
1042 | * storage when the write completes this new data does not become a | |
1043 | * permanent part of the file until the associated transaction commits. | |
1044 | */ | |
428870ff BB |
1045 | |
1046 | /* | |
1047 | * {zfs,zvol,ztest}_get_done() args | |
1048 | */ | |
1049 | typedef struct zgd { | |
1ce23dca | 1050 | struct lwb *zgd_lwb; |
428870ff BB |
1051 | struct blkptr *zgd_bp; |
1052 | dmu_buf_t *zgd_db; | |
bd4dde8e | 1053 | struct zfs_locked_range *zgd_lr; |
428870ff BB |
1054 | void *zgd_private; |
1055 | } zgd_t; | |
1056 | ||
1057 | typedef void dmu_sync_cb_t(zgd_t *arg, int error); | |
1058 | int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd); | |
34dc7c2f BB |
1059 | |
1060 | /* | |
1061 | * Find the next hole or data block in file starting at *off | |
1062 | * Return found offset in *off. Return ESRCH for end of file. | |
1063 | */ | |
1064 | int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, | |
1065 | uint64_t *off); | |
1066 | ||
67a1b037 | 1067 | int dmu_read_l0_bps(objset_t *os, uint64_t object, uint64_t offset, |
d0d91f18 | 1068 | uint64_t length, struct blkptr *bps, size_t *nbpsp); |
bd8c6bd6 | 1069 | int dmu_brt_clone(objset_t *os, uint64_t object, uint64_t offset, |
67a1b037 PJD |
1070 | uint64_t length, dmu_tx_t *tx, const struct blkptr *bps, size_t nbps, |
1071 | boolean_t replay); | |
1072 | ||
34dc7c2f BB |
1073 | /* |
1074 | * Initial setup and final teardown. | |
1075 | */ | |
1076 | extern void dmu_init(void); | |
1077 | extern void dmu_fini(void); | |
1078 | ||
1079 | typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp, | |
1080 | uint64_t object, uint64_t offset, int len); | |
1081 | void dmu_traverse_objset(objset_t *os, uint64_t txg_start, | |
1082 | dmu_traverse_cb_t cb, void *arg); | |
1083 | ||
13fe0198 | 1084 | int dmu_diff(const char *tosnap_name, const char *fromsnap_name, |
da92d5cb | 1085 | zfs_file_t *fp, offset_t *offp); |
572e2857 | 1086 | |
34dc7c2f BB |
1087 | /* CRC64 table */ |
1088 | #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ | |
1089 | extern uint64_t zfs_crc64_table[256]; | |
1090 | ||
fdc2d303 | 1091 | extern uint_t dmu_prefetch_max; |
600a02b8 | 1092 | |
34dc7c2f BB |
1093 | #ifdef __cplusplus |
1094 | } | |
1095 | #endif | |
1096 | ||
1097 | #endif /* _SYS_DMU_H */ |