1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23 * Copyright (c) 2011, 2020 by Delphix. All rights reserved. 24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 25 * Copyright (c) 2012, Joyent, Inc. All rights reserved. 26 * Copyright 2014 HybridCluster. All rights reserved. 27 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. 28 * Copyright 2013 Saso Kiselkov. All rights reserved. 29 * Copyright (c) 2017, Intel Corporation. 30 */ 31 32 /* Portions Copyright 2010 Robert Milkowski */ 33 34 #ifndef _SYS_DMU_H 35 #define _SYS_DMU_H 36 37 /* 38 * This file describes the interface that the DMU provides for its 39 * consumers. 40 * 41 * The DMU also interacts with the SPA. That interface is described in 42 * dmu_spa.h. 43 */ 44 45 #include <sys/zfs_context.h> 46 #include <sys/inttypes.h> 47 #include <sys/cred.h> 48 #include <sys/fs/zfs.h> 49 #include <sys/zio_compress.h> 50 #include <sys/zio_priority.h> 51 #include <sys/uio.h> 52 #include <sys/zfs_file.h> 53 54 #ifdef __cplusplus 55 extern "C" { 56 #endif 57 58 struct page; 59 struct vnode; 60 struct spa; 61 struct zilog; 62 struct zio; 63 struct blkptr; 64 struct zap_cursor; 65 struct dsl_dataset; 66 struct dsl_pool; 67 struct dnode; 68 struct drr_begin; 69 struct drr_end; 70 struct zbookmark_phys; 71 struct spa; 72 struct nvlist; 73 struct arc_buf; 74 struct zio_prop; 75 struct sa_handle; 76 struct dsl_crypto_params; 77 struct locked_range; 78 79 typedef struct objset objset_t; 80 typedef struct dmu_tx dmu_tx_t; 81 typedef struct dsl_dir dsl_dir_t; 82 typedef struct dnode dnode_t; 83 84 typedef enum dmu_object_byteswap { 85 DMU_BSWAP_UINT8, 86 DMU_BSWAP_UINT16, 87 DMU_BSWAP_UINT32, 88 DMU_BSWAP_UINT64, 89 DMU_BSWAP_ZAP, 90 DMU_BSWAP_DNODE, 91 DMU_BSWAP_OBJSET, 92 DMU_BSWAP_ZNODE, 93 DMU_BSWAP_OLDACL, 94 DMU_BSWAP_ACL, 95 /* 96 * Allocating a new byteswap type number makes the on-disk format 97 * incompatible with any other format that uses the same number. 98 * 99 * Data can usually be structured to work with one of the 100 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. 101 */ 102 DMU_BSWAP_NUMFUNCS 103 } dmu_object_byteswap_t; 104 105 #define DMU_OT_NEWTYPE 0x80 106 #define DMU_OT_METADATA 0x40 107 #define DMU_OT_ENCRYPTED 0x20 108 #define DMU_OT_BYTESWAP_MASK 0x1f 109 110 /* 111 * Defines a uint8_t object type. Object types specify if the data 112 * in the object is metadata (boolean) and how to byteswap the data 113 * (dmu_object_byteswap_t). All of the types created by this method 114 * are cached in the dbuf metadata cache. 115 */ 116 #define DMU_OT(byteswap, metadata, encrypted) \ 117 (DMU_OT_NEWTYPE | \ 118 ((metadata) ? DMU_OT_METADATA : 0) | \ 119 ((encrypted) ? DMU_OT_ENCRYPTED : 0) | \ 120 ((byteswap) & DMU_OT_BYTESWAP_MASK)) 121 122 #define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 123 ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \ 124 (ot) < DMU_OT_NUMTYPES) 125 126 #define DMU_OT_IS_METADATA_CACHED(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 127 B_TRUE : dmu_ot[(ot)].ot_dbuf_metadata_cache) 128 129 /* 130 * MDB doesn't have dmu_ot; it defines these macros itself. 131 */ 132 #ifndef ZFS_MDB 133 #define DMU_OT_IS_METADATA_IMPL(ot) (dmu_ot[ot].ot_metadata) 134 #define DMU_OT_IS_ENCRYPTED_IMPL(ot) (dmu_ot[ot].ot_encrypt) 135 #define DMU_OT_BYTESWAP_IMPL(ot) (dmu_ot[ot].ot_byteswap) 136 #endif 137 138 #define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 139 ((ot) & DMU_OT_METADATA) : \ 140 DMU_OT_IS_METADATA_IMPL(ot)) 141 142 #define DMU_OT_IS_DDT(ot) \ 143 ((ot) == DMU_OT_DDT_ZAP) 144 145 /* Note: ztest uses DMU_OT_UINT64_OTHER as a proxy for file blocks */ 146 #define DMU_OT_IS_FILE(ot) \ 147 ((ot) == DMU_OT_PLAIN_FILE_CONTENTS || (ot) == DMU_OT_UINT64_OTHER) 148 149 #define DMU_OT_IS_ENCRYPTED(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 150 ((ot) & DMU_OT_ENCRYPTED) : \ 151 DMU_OT_IS_ENCRYPTED_IMPL(ot)) 152 153 /* 154 * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't 155 * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill 156 * is repurposed for embedded BPs. 157 */ 158 #define DMU_OT_HAS_FILL(ot) \ 159 ((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET) 160 161 #define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 162 ((ot) & DMU_OT_BYTESWAP_MASK) : \ 163 DMU_OT_BYTESWAP_IMPL(ot)) 164 165 typedef enum dmu_object_type { 166 DMU_OT_NONE, 167 /* general: */ 168 DMU_OT_OBJECT_DIRECTORY, /* ZAP */ 169 DMU_OT_OBJECT_ARRAY, /* UINT64 */ 170 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ 171 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ 172 DMU_OT_BPOBJ, /* UINT64 */ 173 DMU_OT_BPOBJ_HDR, /* UINT64 */ 174 /* spa: */ 175 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ 176 DMU_OT_SPACE_MAP, /* UINT64 */ 177 /* zil: */ 178 DMU_OT_INTENT_LOG, /* UINT64 */ 179 /* dmu: */ 180 DMU_OT_DNODE, /* DNODE */ 181 DMU_OT_OBJSET, /* OBJSET */ 182 /* dsl: */ 183 DMU_OT_DSL_DIR, /* UINT64 */ 184 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ 185 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ 186 DMU_OT_DSL_PROPS, /* ZAP */ 187 DMU_OT_DSL_DATASET, /* UINT64 */ 188 /* zpl: */ 189 DMU_OT_ZNODE, /* ZNODE */ 190 DMU_OT_OLDACL, /* Old ACL */ 191 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ 192 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ 193 DMU_OT_MASTER_NODE, /* ZAP */ 194 DMU_OT_UNLINKED_SET, /* ZAP */ 195 /* zvol: */ 196 DMU_OT_ZVOL, /* UINT8 */ 197 DMU_OT_ZVOL_PROP, /* ZAP */ 198 /* other; for testing only! */ 199 DMU_OT_PLAIN_OTHER, /* UINT8 */ 200 DMU_OT_UINT64_OTHER, /* UINT64 */ 201 DMU_OT_ZAP_OTHER, /* ZAP */ 202 /* new object types: */ 203 DMU_OT_ERROR_LOG, /* ZAP */ 204 DMU_OT_SPA_HISTORY, /* UINT8 */ 205 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ 206 DMU_OT_POOL_PROPS, /* ZAP */ 207 DMU_OT_DSL_PERMS, /* ZAP */ 208 DMU_OT_ACL, /* ACL */ 209 DMU_OT_SYSACL, /* SYSACL */ 210 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ 211 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ 212 DMU_OT_NEXT_CLONES, /* ZAP */ 213 DMU_OT_SCAN_QUEUE, /* ZAP */ 214 DMU_OT_USERGROUP_USED, /* ZAP */ 215 DMU_OT_USERGROUP_QUOTA, /* ZAP */ 216 DMU_OT_USERREFS, /* ZAP */ 217 DMU_OT_DDT_ZAP, /* ZAP */ 218 DMU_OT_DDT_STATS, /* ZAP */ 219 DMU_OT_SA, /* System attr */ 220 DMU_OT_SA_MASTER_NODE, /* ZAP */ 221 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ 222 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ 223 DMU_OT_SCAN_XLATE, /* ZAP */ 224 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ 225 DMU_OT_DEADLIST, /* ZAP */ 226 DMU_OT_DEADLIST_HDR, /* UINT64 */ 227 DMU_OT_DSL_CLONES, /* ZAP */ 228 DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */ 229 /* 230 * Do not allocate new object types here. Doing so makes the on-disk 231 * format incompatible with any other format that uses the same object 232 * type number. 233 * 234 * When creating an object which does not have one of the above types 235 * use the DMU_OTN_* type with the correct byteswap and metadata 236 * values. 237 * 238 * The DMU_OTN_* types do not have entries in the dmu_ot table, 239 * use the DMU_OT_IS_METADATA() and DMU_OT_BYTESWAP() macros instead 240 * of indexing into dmu_ot directly (this works for both DMU_OT_* types 241 * and DMU_OTN_* types). 242 */ 243 DMU_OT_NUMTYPES, 244 245 /* 246 * Names for valid types declared with DMU_OT(). 247 */ 248 DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_FALSE), 249 DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_FALSE), 250 DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_FALSE), 251 DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_FALSE), 252 DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_FALSE), 253 DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_FALSE), 254 DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_FALSE), 255 DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_FALSE), 256 DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_FALSE), 257 DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_FALSE), 258 259 DMU_OTN_UINT8_ENC_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_TRUE), 260 DMU_OTN_UINT8_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_TRUE), 261 DMU_OTN_UINT16_ENC_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_TRUE), 262 DMU_OTN_UINT16_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_TRUE), 263 DMU_OTN_UINT32_ENC_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_TRUE), 264 DMU_OTN_UINT32_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_TRUE), 265 DMU_OTN_UINT64_ENC_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_TRUE), 266 DMU_OTN_UINT64_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_TRUE), 267 DMU_OTN_ZAP_ENC_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_TRUE), 268 DMU_OTN_ZAP_ENC_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_TRUE), 269 } dmu_object_type_t; 270 271 /* 272 * These flags are intended to be used to specify the "txg_how" 273 * parameter when calling the dmu_tx_assign() function. See the comment 274 * above dmu_tx_assign() for more details on the meaning of these flags. 275 */ 276 #define TXG_NOWAIT (0ULL) 277 #define TXG_WAIT (1ULL<<0) 278 #define TXG_NOTHROTTLE (1ULL<<1) 279 280 void byteswap_uint64_array(void *buf, size_t size); 281 void byteswap_uint32_array(void *buf, size_t size); 282 void byteswap_uint16_array(void *buf, size_t size); 283 void byteswap_uint8_array(void *buf, size_t size); 284 void zap_byteswap(void *buf, size_t size); 285 void zfs_oldacl_byteswap(void *buf, size_t size); 286 void zfs_acl_byteswap(void *buf, size_t size); 287 void zfs_znode_byteswap(void *buf, size_t size); 288 289 #define DS_FIND_SNAPSHOTS (1<<0) 290 #define DS_FIND_CHILDREN (1<<1) 291 #define DS_FIND_SERIALIZE (1<<2) 292 293 /* 294 * The maximum number of bytes that can be accessed as part of one 295 * operation, including metadata. 296 */ 297 #define DMU_MAX_ACCESS (64 * 1024 * 1024) /* 64MB */ 298 #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */ 299 300 #define DMU_USERUSED_OBJECT (-1ULL) 301 #define DMU_GROUPUSED_OBJECT (-2ULL) 302 #define DMU_PROJECTUSED_OBJECT (-3ULL) 303 304 /* 305 * Zap prefix for object accounting in DMU_{USER,GROUP,PROJECT}USED_OBJECT. 306 */ 307 #define DMU_OBJACCT_PREFIX "obj-" 308 #define DMU_OBJACCT_PREFIX_LEN 4 309 310 /* 311 * artificial blkids for bonus buffer and spill blocks 312 */ 313 #define DMU_BONUS_BLKID (-1ULL) 314 #define DMU_SPILL_BLKID (-2ULL) 315 316 /* 317 * Public routines to create, destroy, open, and close objsets. 318 */ 319 typedef void dmu_objset_create_sync_func_t(objset_t *os, void *arg, 320 cred_t *cr, dmu_tx_t *tx); 321 322 int dmu_objset_hold(const char *name, void *tag, objset_t **osp); 323 int dmu_objset_own(const char *name, dmu_objset_type_t type, 324 boolean_t readonly, boolean_t key_required, void *tag, objset_t **osp); 325 void dmu_objset_rele(objset_t *os, void *tag); 326 void dmu_objset_disown(objset_t *os, boolean_t key_required, void *tag); 327 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp); 328 329 void dmu_objset_evict_dbufs(objset_t *os); 330 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, 331 struct dsl_crypto_params *dcp, dmu_objset_create_sync_func_t func, 332 void *arg); 333 int dmu_objset_clone(const char *name, const char *origin); 334 int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer, 335 struct nvlist *errlist); 336 int dmu_objset_snapshot_one(const char *fsname, const char *snapname); 337 int dmu_objset_find(const char *name, int func(const char *, void *), void *arg, 338 int flags); 339 void dmu_objset_byteswap(void *buf, size_t size); 340 int dsl_dataset_rename_snapshot(const char *fsname, 341 const char *oldsnapname, const char *newsnapname, boolean_t recursive); 342 343 typedef struct dmu_buf { 344 uint64_t db_object; /* object that this buffer is part of */ 345 uint64_t db_offset; /* byte offset in this object */ 346 uint64_t db_size; /* size of buffer in bytes */ 347 void *db_data; /* data in buffer */ 348 } dmu_buf_t; 349 350 /* 351 * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 352 */ 353 #define DMU_POOL_DIRECTORY_OBJECT 1 354 #define DMU_POOL_CONFIG "config" 355 #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write" 356 #define DMU_POOL_FEATURES_FOR_READ "features_for_read" 357 #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions" 358 #define DMU_POOL_FEATURE_ENABLED_TXG "feature_enabled_txg" 359 #define DMU_POOL_ROOT_DATASET "root_dataset" 360 #define DMU_POOL_SYNC_BPOBJ "sync_bplist" 361 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 362 #define DMU_POOL_ERRLOG_LAST "errlog_last" 363 #define DMU_POOL_SPARES "spares" 364 #define DMU_POOL_DEFLATE "deflate" 365 #define DMU_POOL_HISTORY "history" 366 #define DMU_POOL_PROPS "pool_props" 367 #define DMU_POOL_L2CACHE "l2cache" 368 #define DMU_POOL_TMP_USERREFS "tmp_userrefs" 369 #define DMU_POOL_DDT "DDT-%s-%s-%s" 370 #define DMU_POOL_DDT_STATS "DDT-statistics" 371 #define DMU_POOL_CREATION_VERSION "creation_version" 372 #define DMU_POOL_SCAN "scan" 373 #define DMU_POOL_FREE_BPOBJ "free_bpobj" 374 #define DMU_POOL_BPTREE_OBJ "bptree_obj" 375 #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj" 376 #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" 377 #define DMU_POOL_VDEV_ZAP_MAP "com.delphix:vdev_zap_map" 378 #define DMU_POOL_REMOVING "com.delphix:removing" 379 #define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj" 380 #define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect" 381 #define DMU_POOL_ZPOOL_CHECKPOINT "com.delphix:zpool_checkpoint" 382 #define DMU_POOL_LOG_SPACEMAP_ZAP "com.delphix:log_spacemap_zap" 383 #define DMU_POOL_DELETED_CLONES "com.delphix:deleted_clones" 384 385 /* 386 * Allocate an object from this objset. The range of object numbers 387 * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode. 388 * 389 * The transaction must be assigned to a txg. The newly allocated 390 * object will be "held" in the transaction (ie. you can modify the 391 * newly allocated object in this transaction). 392 * 393 * dmu_object_alloc() chooses an object and returns it in *objectp. 394 * 395 * dmu_object_claim() allocates a specific object number. If that 396 * number is already allocated, it fails and returns EEXIST. 397 * 398 * Return 0 on success, or ENOSPC or EEXIST as specified above. 399 */ 400 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot, 401 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 402 uint64_t dmu_object_alloc_ibs(objset_t *os, dmu_object_type_t ot, int blocksize, 403 int indirect_blockshift, 404 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx); 405 uint64_t dmu_object_alloc_dnsize(objset_t *os, dmu_object_type_t ot, 406 int blocksize, dmu_object_type_t bonus_type, int bonus_len, 407 int dnodesize, dmu_tx_t *tx); 408 uint64_t dmu_object_alloc_hold(objset_t *os, dmu_object_type_t ot, 409 int blocksize, int indirect_blockshift, dmu_object_type_t bonustype, 410 int bonuslen, int dnodesize, dnode_t **allocated_dnode, void *tag, 411 dmu_tx_t *tx); 412 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot, 413 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 414 int dmu_object_claim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot, 415 int blocksize, dmu_object_type_t bonus_type, int bonus_len, 416 int dnodesize, dmu_tx_t *tx); 417 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot, 418 int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp); 419 int dmu_object_reclaim_dnsize(objset_t *os, uint64_t object, 420 dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype, 421 int bonuslen, int dnodesize, boolean_t keep_spill, dmu_tx_t *tx); 422 int dmu_object_rm_spill(objset_t *os, uint64_t object, dmu_tx_t *tx); 423 424 /* 425 * Free an object from this objset. 426 * 427 * The object's data will be freed as well (ie. you don't need to call 428 * dmu_free(object, 0, -1, tx)). 429 * 430 * The object need not be held in the transaction. 431 * 432 * If there are any holds on this object's buffers (via dmu_buf_hold()), 433 * or tx holds on the object (via dmu_tx_hold_object()), you can not 434 * free it; it fails and returns EBUSY. 435 * 436 * If the object is not allocated, it fails and returns ENOENT. 437 * 438 * Return 0 on success, or EBUSY or ENOENT as specified above. 439 */ 440 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx); 441 442 /* 443 * Find the next allocated or free object. 444 * 445 * The objectp parameter is in-out. It will be updated to be the next 446 * object which is allocated. Ignore objects which have not been 447 * modified since txg. 448 * 449 * XXX Can only be called on a objset with no dirty data. 450 * 451 * Returns 0 on success, or ENOENT if there are no more objects. 452 */ 453 int dmu_object_next(objset_t *os, uint64_t *objectp, 454 boolean_t hole, uint64_t txg); 455 456 /* 457 * Set the number of levels on a dnode. nlevels must be greater than the 458 * current number of levels or an EINVAL will be returned. 459 */ 460 int dmu_object_set_nlevels(objset_t *os, uint64_t object, int nlevels, 461 dmu_tx_t *tx); 462 463 /* 464 * Set the data blocksize for an object. 465 * 466 * The object cannot have any blocks allocated beyond the first. If 467 * the first block is allocated already, the new size must be greater 468 * than the current block size. If these conditions are not met, 469 * ENOTSUP will be returned. 470 * 471 * Returns 0 on success, or EBUSY if there are any holds on the object 472 * contents, or ENOTSUP as described above. 473 */ 474 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, 475 int ibs, dmu_tx_t *tx); 476 477 /* 478 * Manually set the maxblkid on a dnode. This will adjust nlevels accordingly 479 * to accommodate the change. When calling this function, the caller must 480 * ensure that the object's nlevels can sufficiently support the new maxblkid. 481 */ 482 int dmu_object_set_maxblkid(objset_t *os, uint64_t object, uint64_t maxblkid, 483 dmu_tx_t *tx); 484 485 /* 486 * Set the checksum property on a dnode. The new checksum algorithm will 487 * apply to all newly written blocks; existing blocks will not be affected. 488 */ 489 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, 490 dmu_tx_t *tx); 491 492 /* 493 * Set the compress property on a dnode. The new compression algorithm will 494 * apply to all newly written blocks; existing blocks will not be affected. 495 */ 496 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, 497 dmu_tx_t *tx); 498 499 void dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset, 500 void *data, uint8_t etype, uint8_t comp, int uncompressed_size, 501 int compressed_size, int byteorder, dmu_tx_t *tx); 502 void dmu_redact(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 503 dmu_tx_t *tx); 504 505 /* 506 * Decide how to write a block: checksum, compression, number of copies, etc. 507 */ 508 #define WP_NOFILL 0x1 509 #define WP_DMU_SYNC 0x2 510 #define WP_SPILL 0x4 511 512 void dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, 513 struct zio_prop *zp); 514 515 /* 516 * The bonus data is accessed more or less like a regular buffer. 517 * You must dmu_bonus_hold() to get the buffer, which will give you a 518 * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus 519 * data. As with any normal buffer, you must call dmu_buf_will_dirty() 520 * before modifying it, and the 521 * object must be held in an assigned transaction before calling 522 * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus 523 * buffer as well. You must release what you hold with dmu_buf_rele(). 524 * 525 * Returns ENOENT, EIO, or 0. 526 */ 527 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **dbp); 528 int dmu_bonus_hold_by_dnode(dnode_t *dn, void *tag, dmu_buf_t **dbp, 529 uint32_t flags); 530 int dmu_bonus_max(void); 531 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *); 532 int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *); 533 dmu_object_type_t dmu_get_bonustype(dmu_buf_t *); 534 int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *); 535 536 /* 537 * Special spill buffer support used by "SA" framework 538 */ 539 540 int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, uint32_t flags, void *tag, 541 dmu_buf_t **dbp); 542 int dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags, 543 void *tag, dmu_buf_t **dbp); 544 int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); 545 546 /* 547 * Obtain the DMU buffer from the specified object which contains the 548 * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so 549 * that it will remain in memory. You must release the hold with 550 * dmu_buf_rele(). You must not access the dmu_buf_t after releasing 551 * what you hold. You must have a hold on any dmu_buf_t* you pass to the DMU. 552 * 553 * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill 554 * on the returned buffer before reading or writing the buffer's 555 * db_data. The comments for those routines describe what particular 556 * operations are valid after calling them. 557 * 558 * The object number must be a valid, allocated object number. 559 */ 560 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, 561 void *tag, dmu_buf_t **, int flags); 562 int dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset, 563 void *tag, dmu_buf_t **dbp, int flags); 564 int dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset, 565 uint64_t length, boolean_t read, void *tag, int *numbufsp, 566 dmu_buf_t ***dbpp, uint32_t flags); 567 /* 568 * Add a reference to a dmu buffer that has already been held via 569 * dmu_buf_hold() in the current context. 570 */ 571 void dmu_buf_add_ref(dmu_buf_t *db, void* tag); 572 573 /* 574 * Attempt to add a reference to a dmu buffer that is in an unknown state, 575 * using a pointer that may have been invalidated by eviction processing. 576 * The request will succeed if the passed in dbuf still represents the 577 * same os/object/blkid, is ineligible for eviction, and has at least 578 * one hold by a user other than the syncer. 579 */ 580 boolean_t dmu_buf_try_add_ref(dmu_buf_t *, objset_t *os, uint64_t object, 581 uint64_t blkid, void *tag); 582 583 void dmu_buf_rele(dmu_buf_t *db, void *tag); 584 uint64_t dmu_buf_refcount(dmu_buf_t *db); 585 uint64_t dmu_buf_user_refcount(dmu_buf_t *db); 586 587 /* 588 * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a 589 * range of an object. A pointer to an array of dmu_buf_t*'s is 590 * returned (in *dbpp). 591 * 592 * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and 593 * frees the array. The hold on the array of buffers MUST be released 594 * with dmu_buf_rele_array. You can NOT release the hold on each buffer 595 * individually with dmu_buf_rele. 596 */ 597 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset, 598 uint64_t length, boolean_t read, void *tag, 599 int *numbufsp, dmu_buf_t ***dbpp); 600 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag); 601 602 typedef void dmu_buf_evict_func_t(void *user_ptr); 603 604 /* 605 * A DMU buffer user object may be associated with a dbuf for the 606 * duration of its lifetime. This allows the user of a dbuf (client) 607 * to attach private data to a dbuf (e.g. in-core only data such as a 608 * dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified 609 * when that dbuf has been evicted. Clients typically respond to the 610 * eviction notification by freeing their private data, thus ensuring 611 * the same lifetime for both dbuf and private data. 612 * 613 * The mapping from a dmu_buf_user_t to any client private data is the 614 * client's responsibility. All current consumers of the API with private 615 * data embed a dmu_buf_user_t as the first member of the structure for 616 * their private data. This allows conversions between the two types 617 * with a simple cast. Since the DMU buf user API never needs access 618 * to the private data, other strategies can be employed if necessary 619 * or convenient for the client (e.g. using container_of() to do the 620 * conversion for private data that cannot have the dmu_buf_user_t as 621 * its first member). 622 * 623 * Eviction callbacks are executed without the dbuf mutex held or any 624 * other type of mechanism to guarantee that the dbuf is still available. 625 * For this reason, users must assume the dbuf has already been freed 626 * and not reference the dbuf from the callback context. 627 * 628 * Users requesting "immediate eviction" are notified as soon as the dbuf 629 * is only referenced by dirty records (dirties == holds). Otherwise the 630 * notification occurs after eviction processing for the dbuf begins. 631 */ 632 typedef struct dmu_buf_user { 633 /* 634 * Asynchronous user eviction callback state. 635 */ 636 taskq_ent_t dbu_tqent; 637 638 /* 639 * This instance's eviction function pointers. 640 * 641 * dbu_evict_func_sync is called synchronously and then 642 * dbu_evict_func_async is executed asynchronously on a taskq. 643 */ 644 dmu_buf_evict_func_t *dbu_evict_func_sync; 645 dmu_buf_evict_func_t *dbu_evict_func_async; 646 #ifdef ZFS_DEBUG 647 /* 648 * Pointer to user's dbuf pointer. NULL for clients that do 649 * not associate a dbuf with their user data. 650 * 651 * The dbuf pointer is cleared upon eviction so as to catch 652 * use-after-evict bugs in clients. 653 */ 654 dmu_buf_t **dbu_clear_on_evict_dbufp; 655 #endif 656 } dmu_buf_user_t; 657 658 /* 659 * Initialize the given dmu_buf_user_t instance with the eviction function 660 * evict_func, to be called when the user is evicted. 661 * 662 * NOTE: This function should only be called once on a given dmu_buf_user_t. 663 * To allow enforcement of this, dbu must already be zeroed on entry. 664 */ 665 /*ARGSUSED*/ 666 static inline void 667 dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func_sync, 668 dmu_buf_evict_func_t *evict_func_async, 669 dmu_buf_t **clear_on_evict_dbufp __maybe_unused) 670 { 671 ASSERT(dbu->dbu_evict_func_sync == NULL); 672 ASSERT(dbu->dbu_evict_func_async == NULL); 673 674 /* must have at least one evict func */ 675 IMPLY(evict_func_sync == NULL, evict_func_async != NULL); 676 dbu->dbu_evict_func_sync = evict_func_sync; 677 dbu->dbu_evict_func_async = evict_func_async; 678 taskq_init_ent(&dbu->dbu_tqent); 679 #ifdef ZFS_DEBUG 680 dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp; 681 #endif 682 } 683 684 /* 685 * Attach user data to a dbuf and mark it for normal (when the dbuf's 686 * data is cleared or its reference count goes to zero) eviction processing. 687 * 688 * Returns NULL on success, or the existing user if another user currently 689 * owns the buffer. 690 */ 691 void *dmu_buf_set_user(dmu_buf_t *db, dmu_buf_user_t *user); 692 693 /* 694 * Attach user data to a dbuf and mark it for immediate (its dirty and 695 * reference counts are equal) eviction processing. 696 * 697 * Returns NULL on success, or the existing user if another user currently 698 * owns the buffer. 699 */ 700 void *dmu_buf_set_user_ie(dmu_buf_t *db, dmu_buf_user_t *user); 701 702 /* 703 * Replace the current user of a dbuf. 704 * 705 * If given the current user of a dbuf, replaces the dbuf's user with 706 * "new_user" and returns the user data pointer that was replaced. 707 * Otherwise returns the current, and unmodified, dbuf user pointer. 708 */ 709 void *dmu_buf_replace_user(dmu_buf_t *db, 710 dmu_buf_user_t *old_user, dmu_buf_user_t *new_user); 711 712 /* 713 * Remove the specified user data for a DMU buffer. 714 * 715 * Returns the user that was removed on success, or the current user if 716 * another user currently owns the buffer. 717 */ 718 void *dmu_buf_remove_user(dmu_buf_t *db, dmu_buf_user_t *user); 719 720 /* 721 * Returns the user data (dmu_buf_user_t *) associated with this dbuf. 722 */ 723 void *dmu_buf_get_user(dmu_buf_t *db); 724 725 objset_t *dmu_buf_get_objset(dmu_buf_t *db); 726 dnode_t *dmu_buf_dnode_enter(dmu_buf_t *db); 727 void dmu_buf_dnode_exit(dmu_buf_t *db); 728 729 /* Block until any in-progress dmu buf user evictions complete. */ 730 void dmu_buf_user_evict_wait(void); 731 732 /* 733 * Returns the blkptr associated with this dbuf, or NULL if not set. 734 */ 735 struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db); 736 737 /* 738 * Indicate that you are going to modify the buffer's data (db_data). 739 * 740 * The transaction (tx) must be assigned to a txg (ie. you've called 741 * dmu_tx_assign()). The buffer's object must be held in the tx 742 * (ie. you've called dmu_tx_hold_object(tx, db->db_object)). 743 */ 744 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx); 745 boolean_t dmu_buf_is_dirty(dmu_buf_t *db, dmu_tx_t *tx); 746 void dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder, 747 const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_tx_t *tx); 748 749 /* 750 * You must create a transaction, then hold the objects which you will 751 * (or might) modify as part of this transaction. Then you must assign 752 * the transaction to a transaction group. Once the transaction has 753 * been assigned, you can modify buffers which belong to held objects as 754 * part of this transaction. You can't modify buffers before the 755 * transaction has been assigned; you can't modify buffers which don't 756 * belong to objects which this transaction holds; you can't hold 757 * objects once the transaction has been assigned. You may hold an 758 * object which you are going to free (with dmu_object_free()), but you 759 * don't have to. 760 * 761 * You can abort the transaction before it has been assigned. 762 * 763 * Note that you may hold buffers (with dmu_buf_hold) at any time, 764 * regardless of transaction state. 765 */ 766 767 #define DMU_NEW_OBJECT (-1ULL) 768 #define DMU_OBJECT_END (-1ULL) 769 770 dmu_tx_t *dmu_tx_create(objset_t *os); 771 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len); 772 void dmu_tx_hold_write_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, 773 int len); 774 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, 775 uint64_t len); 776 void dmu_tx_hold_free_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, 777 uint64_t len); 778 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name); 779 void dmu_tx_hold_zap_by_dnode(dmu_tx_t *tx, dnode_t *dn, int add, 780 const char *name); 781 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object); 782 void dmu_tx_hold_bonus_by_dnode(dmu_tx_t *tx, dnode_t *dn); 783 void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object); 784 void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow); 785 void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size); 786 void dmu_tx_abort(dmu_tx_t *tx); 787 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how); 788 void dmu_tx_wait(dmu_tx_t *tx); 789 void dmu_tx_commit(dmu_tx_t *tx); 790 void dmu_tx_mark_netfree(dmu_tx_t *tx); 791 792 /* 793 * To register a commit callback, dmu_tx_callback_register() must be called. 794 * 795 * dcb_data is a pointer to caller private data that is passed on as a 796 * callback parameter. The caller is responsible for properly allocating and 797 * freeing it. 798 * 799 * When registering a callback, the transaction must be already created, but 800 * it cannot be committed or aborted. It can be assigned to a txg or not. 801 * 802 * The callback will be called after the transaction has been safely written 803 * to stable storage and will also be called if the dmu_tx is aborted. 804 * If there is any error which prevents the transaction from being committed to 805 * disk, the callback will be called with a value of error != 0. 806 * 807 * When multiple callbacks are registered to the transaction, the callbacks 808 * will be called in reverse order to let Lustre, the only user of commit 809 * callback currently, take the fast path of its commit callback handling. 810 */ 811 typedef void dmu_tx_callback_func_t(void *dcb_data, int error); 812 813 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func, 814 void *dcb_data); 815 void dmu_tx_do_callbacks(list_t *cb_list, int error); 816 817 /* 818 * Free up the data blocks for a defined range of a file. If size is 819 * -1, the range from offset to end-of-file is freed. 820 */ 821 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, 822 uint64_t size, dmu_tx_t *tx); 823 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, 824 uint64_t size); 825 int dmu_free_long_object(objset_t *os, uint64_t object); 826 827 /* 828 * Convenience functions. 829 * 830 * Canfail routines will return 0 on success, or an errno if there is a 831 * nonrecoverable I/O error. 832 */ 833 #define DMU_READ_PREFETCH 0 /* prefetch */ 834 #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */ 835 #define DMU_READ_NO_DECRYPT 2 /* don't decrypt */ 836 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 837 void *buf, uint32_t flags); 838 int dmu_read_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, void *buf, 839 uint32_t flags); 840 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 841 const void *buf, dmu_tx_t *tx); 842 void dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, 843 const void *buf, dmu_tx_t *tx); 844 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 845 dmu_tx_t *tx); 846 #ifdef _KERNEL 847 int dmu_read_uio(objset_t *os, uint64_t object, zfs_uio_t *uio, uint64_t size); 848 int dmu_read_uio_dbuf(dmu_buf_t *zdb, zfs_uio_t *uio, uint64_t size); 849 int dmu_read_uio_dnode(dnode_t *dn, zfs_uio_t *uio, uint64_t size); 850 int dmu_write_uio(objset_t *os, uint64_t object, zfs_uio_t *uio, uint64_t size, 851 dmu_tx_t *tx); 852 int dmu_write_uio_dbuf(dmu_buf_t *zdb, zfs_uio_t *uio, uint64_t size, 853 dmu_tx_t *tx); 854 int dmu_write_uio_dnode(dnode_t *dn, zfs_uio_t *uio, uint64_t size, 855 dmu_tx_t *tx); 856 #endif 857 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size); 858 void dmu_return_arcbuf(struct arc_buf *buf); 859 int dmu_assign_arcbuf_by_dnode(dnode_t *dn, uint64_t offset, 860 struct arc_buf *buf, dmu_tx_t *tx); 861 int dmu_assign_arcbuf_by_dbuf(dmu_buf_t *handle, uint64_t offset, 862 struct arc_buf *buf, dmu_tx_t *tx); 863 #define dmu_assign_arcbuf dmu_assign_arcbuf_by_dbuf 864 extern int zfs_prefetch_disable; 865 extern int zfs_max_recordsize; 866 867 /* 868 * Asynchronously try to read in the data. 869 */ 870 void dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset, 871 uint64_t len, enum zio_priority pri); 872 873 typedef struct dmu_object_info { 874 /* All sizes are in bytes unless otherwise indicated. */ 875 uint32_t doi_data_block_size; 876 uint32_t doi_metadata_block_size; 877 dmu_object_type_t doi_type; 878 dmu_object_type_t doi_bonus_type; 879 uint64_t doi_bonus_size; 880 uint8_t doi_indirection; /* 2 = dnode->indirect->data */ 881 uint8_t doi_checksum; 882 uint8_t doi_compress; 883 uint8_t doi_nblkptr; 884 uint8_t doi_pad[4]; 885 uint64_t doi_dnodesize; 886 uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */ 887 uint64_t doi_max_offset; 888 uint64_t doi_fill_count; /* number of non-empty blocks */ 889 } dmu_object_info_t; 890 891 typedef void (*const arc_byteswap_func_t)(void *buf, size_t size); 892 893 typedef struct dmu_object_type_info { 894 dmu_object_byteswap_t ot_byteswap; 895 boolean_t ot_metadata; 896 boolean_t ot_dbuf_metadata_cache; 897 boolean_t ot_encrypt; 898 char *ot_name; 899 } dmu_object_type_info_t; 900 901 typedef const struct dmu_object_byteswap_info { 902 arc_byteswap_func_t ob_func; 903 char *ob_name; 904 } dmu_object_byteswap_info_t; 905 906 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES]; 907 extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS]; 908 909 /* 910 * Get information on a DMU object. 911 * 912 * Return 0 on success or ENOENT if object is not allocated. 913 * 914 * If doi is NULL, just indicates whether the object exists. 915 */ 916 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi); 917 void __dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi); 918 /* Like dmu_object_info, but faster if you have a held dnode in hand. */ 919 void dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi); 920 /* Like dmu_object_info, but faster if you have a held dbuf in hand. */ 921 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi); 922 /* 923 * Like dmu_object_info_from_db, but faster still when you only care about 924 * the size. 925 */ 926 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, 927 u_longlong_t *nblk512); 928 929 void dmu_object_dnsize_from_db(dmu_buf_t *db, int *dnsize); 930 931 typedef struct dmu_objset_stats { 932 uint64_t dds_num_clones; /* number of clones of this */ 933 uint64_t dds_creation_txg; 934 uint64_t dds_guid; 935 dmu_objset_type_t dds_type; 936 uint8_t dds_is_snapshot; 937 uint8_t dds_inconsistent; 938 uint8_t dds_redacted; 939 char dds_origin[ZFS_MAX_DATASET_NAME_LEN]; 940 } dmu_objset_stats_t; 941 942 /* 943 * Get stats on a dataset. 944 */ 945 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat); 946 947 /* 948 * Add entries to the nvlist for all the objset's properties. See 949 * zfs_prop_table[] and zfs(1m) for details on the properties. 950 */ 951 void dmu_objset_stats(objset_t *os, struct nvlist *nv); 952 953 /* 954 * Get the space usage statistics for statvfs(). 955 * 956 * refdbytes is the amount of space "referenced" by this objset. 957 * availbytes is the amount of space available to this objset, taking 958 * into account quotas & reservations, assuming that no other objsets 959 * use the space first. These values correspond to the 'referenced' and 960 * 'available' properties, described in the zfs(1m) manpage. 961 * 962 * usedobjs and availobjs are the number of objects currently allocated, 963 * and available. 964 */ 965 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, 966 uint64_t *usedobjsp, uint64_t *availobjsp); 967 968 /* 969 * The fsid_guid is a 56-bit ID that can change to avoid collisions. 970 * (Contrast with the ds_guid which is a 64-bit ID that will never 971 * change, so there is a small probability that it will collide.) 972 */ 973 uint64_t dmu_objset_fsid_guid(objset_t *os); 974 975 /* 976 * Get the [cm]time for an objset's snapshot dir 977 */ 978 inode_timespec_t dmu_objset_snap_cmtime(objset_t *os); 979 980 int dmu_objset_is_snapshot(objset_t *os); 981 982 extern struct spa *dmu_objset_spa(objset_t *os); 983 extern struct zilog *dmu_objset_zil(objset_t *os); 984 extern struct dsl_pool *dmu_objset_pool(objset_t *os); 985 extern struct dsl_dataset *dmu_objset_ds(objset_t *os); 986 extern void dmu_objset_name(objset_t *os, char *buf); 987 extern dmu_objset_type_t dmu_objset_type(objset_t *os); 988 extern uint64_t dmu_objset_id(objset_t *os); 989 extern uint64_t dmu_objset_dnodesize(objset_t *os); 990 extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os); 991 extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os); 992 extern int dmu_objset_blksize(objset_t *os); 993 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name, 994 uint64_t *id, uint64_t *offp, boolean_t *case_conflict); 995 extern int dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *val); 996 extern int dmu_snapshot_realname(objset_t *os, const char *name, char *real, 997 int maxlen, boolean_t *conflict); 998 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name, 999 uint64_t *idp, uint64_t *offp); 1000 1001 typedef struct zfs_file_info { 1002 uint64_t zfi_user; 1003 uint64_t zfi_group; 1004 uint64_t zfi_project; 1005 uint64_t zfi_generation; 1006 } zfs_file_info_t; 1007 1008 typedef int file_info_cb_t(dmu_object_type_t bonustype, const void *data, 1009 struct zfs_file_info *zoi); 1010 extern void dmu_objset_register_type(dmu_objset_type_t ost, 1011 file_info_cb_t *cb); 1012 extern void dmu_objset_set_user(objset_t *os, void *user_ptr); 1013 extern void *dmu_objset_get_user(objset_t *os); 1014 1015 /* 1016 * Return the txg number for the given assigned transaction. 1017 */ 1018 uint64_t dmu_tx_get_txg(dmu_tx_t *tx); 1019 1020 /* 1021 * Synchronous write. 1022 * If a parent zio is provided this function initiates a write on the 1023 * provided buffer as a child of the parent zio. 1024 * In the absence of a parent zio, the write is completed synchronously. 1025 * At write completion, blk is filled with the bp of the written block. 1026 * Note that while the data covered by this function will be on stable 1027 * storage when the write completes this new data does not become a 1028 * permanent part of the file until the associated transaction commits. 1029 */ 1030 1031 /* 1032 * {zfs,zvol,ztest}_get_done() args 1033 */ 1034 typedef struct zgd { 1035 struct lwb *zgd_lwb; 1036 struct blkptr *zgd_bp; 1037 dmu_buf_t *zgd_db; 1038 struct zfs_locked_range *zgd_lr; 1039 void *zgd_private; 1040 } zgd_t; 1041 1042 typedef void dmu_sync_cb_t(zgd_t *arg, int error); 1043 int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd); 1044 1045 /* 1046 * Find the next hole or data block in file starting at *off 1047 * Return found offset in *off. Return ESRCH for end of file. 1048 */ 1049 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, 1050 uint64_t *off); 1051 1052 /* 1053 * Initial setup and final teardown. 1054 */ 1055 extern void dmu_init(void); 1056 extern void dmu_fini(void); 1057 1058 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp, 1059 uint64_t object, uint64_t offset, int len); 1060 void dmu_traverse_objset(objset_t *os, uint64_t txg_start, 1061 dmu_traverse_cb_t cb, void *arg); 1062 1063 int dmu_diff(const char *tosnap_name, const char *fromsnap_name, 1064 zfs_file_t *fp, offset_t *offp); 1065 1066 /* CRC64 table */ 1067 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ 1068 extern uint64_t zfs_crc64_table[256]; 1069 1070 #ifdef __cplusplus 1071 } 1072 #endif 1073 1074 #endif /* _SYS_DMU_H */ 1075