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 https://opensource.org/licenses/CDDL-1.0. 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, const 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, const void *tag, 325 objset_t **osp); 326 void dmu_objset_rele(objset_t *os, const void *tag); 327 void dmu_objset_disown(objset_t *os, boolean_t key_required, const void *tag); 328 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp); 329 330 void dmu_objset_evict_dbufs(objset_t *os); 331 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, 332 struct dsl_crypto_params *dcp, dmu_objset_create_sync_func_t func, 333 void *arg); 334 int dmu_objset_clone(const char *name, const char *origin); 335 int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer, 336 struct nvlist *errlist); 337 int dmu_objset_snapshot_one(const char *fsname, const char *snapname); 338 int dmu_objset_find(const char *name, int func(const char *, void *), void *arg, 339 int flags); 340 void dmu_objset_byteswap(void *buf, size_t size); 341 int dsl_dataset_rename_snapshot(const char *fsname, 342 const char *oldsnapname, const char *newsnapname, boolean_t recursive); 343 344 typedef struct dmu_buf { 345 uint64_t db_object; /* object that this buffer is part of */ 346 uint64_t db_offset; /* byte offset in this object */ 347 uint64_t db_size; /* size of buffer in bytes */ 348 void *db_data; /* data in buffer */ 349 } dmu_buf_t; 350 351 /* 352 * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 353 */ 354 #define DMU_POOL_DIRECTORY_OBJECT 1 355 #define DMU_POOL_CONFIG "config" 356 #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write" 357 #define DMU_POOL_FEATURES_FOR_READ "features_for_read" 358 #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions" 359 #define DMU_POOL_FEATURE_ENABLED_TXG "feature_enabled_txg" 360 #define DMU_POOL_ROOT_DATASET "root_dataset" 361 #define DMU_POOL_SYNC_BPOBJ "sync_bplist" 362 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 363 #define DMU_POOL_ERRLOG_LAST "errlog_last" 364 #define DMU_POOL_SPARES "spares" 365 #define DMU_POOL_DEFLATE "deflate" 366 #define DMU_POOL_HISTORY "history" 367 #define DMU_POOL_PROPS "pool_props" 368 #define DMU_POOL_L2CACHE "l2cache" 369 #define DMU_POOL_TMP_USERREFS "tmp_userrefs" 370 #define DMU_POOL_DDT "DDT-%s-%s-%s" 371 #define DMU_POOL_DDT_STATS "DDT-statistics" 372 #define DMU_POOL_CREATION_VERSION "creation_version" 373 #define DMU_POOL_SCAN "scan" 374 #define DMU_POOL_FREE_BPOBJ "free_bpobj" 375 #define DMU_POOL_BPTREE_OBJ "bptree_obj" 376 #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj" 377 #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" 378 #define DMU_POOL_VDEV_ZAP_MAP "com.delphix:vdev_zap_map" 379 #define DMU_POOL_REMOVING "com.delphix:removing" 380 #define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj" 381 #define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect" 382 #define DMU_POOL_ZPOOL_CHECKPOINT "com.delphix:zpool_checkpoint" 383 #define DMU_POOL_LOG_SPACEMAP_ZAP "com.delphix:log_spacemap_zap" 384 #define DMU_POOL_DELETED_CLONES "com.delphix:deleted_clones" 385 386 /* 387 * Allocate an object from this objset. The range of object numbers 388 * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode. 389 * 390 * The transaction must be assigned to a txg. The newly allocated 391 * object will be "held" in the transaction (ie. you can modify the 392 * newly allocated object in this transaction). 393 * 394 * dmu_object_alloc() chooses an object and returns it in *objectp. 395 * 396 * dmu_object_claim() allocates a specific object number. If that 397 * number is already allocated, it fails and returns EEXIST. 398 * 399 * Return 0 on success, or ENOSPC or EEXIST as specified above. 400 */ 401 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot, 402 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 403 uint64_t dmu_object_alloc_ibs(objset_t *os, dmu_object_type_t ot, int blocksize, 404 int indirect_blockshift, 405 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx); 406 uint64_t dmu_object_alloc_dnsize(objset_t *os, dmu_object_type_t ot, 407 int blocksize, dmu_object_type_t bonus_type, int bonus_len, 408 int dnodesize, dmu_tx_t *tx); 409 uint64_t dmu_object_alloc_hold(objset_t *os, dmu_object_type_t ot, 410 int blocksize, int indirect_blockshift, dmu_object_type_t bonustype, 411 int bonuslen, int dnodesize, dnode_t **allocated_dnode, const void *tag, 412 dmu_tx_t *tx); 413 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot, 414 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 415 int dmu_object_claim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot, 416 int blocksize, dmu_object_type_t bonus_type, int bonus_len, 417 int dnodesize, dmu_tx_t *tx); 418 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot, 419 int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp); 420 int dmu_object_reclaim_dnsize(objset_t *os, uint64_t object, 421 dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype, 422 int bonuslen, int dnodesize, boolean_t keep_spill, dmu_tx_t *tx); 423 int dmu_object_rm_spill(objset_t *os, uint64_t object, dmu_tx_t *tx); 424 425 /* 426 * Free an object from this objset. 427 * 428 * The object's data will be freed as well (ie. you don't need to call 429 * dmu_free(object, 0, -1, tx)). 430 * 431 * The object need not be held in the transaction. 432 * 433 * If there are any holds on this object's buffers (via dmu_buf_hold()), 434 * or tx holds on the object (via dmu_tx_hold_object()), you can not 435 * free it; it fails and returns EBUSY. 436 * 437 * If the object is not allocated, it fails and returns ENOENT. 438 * 439 * Return 0 on success, or EBUSY or ENOENT as specified above. 440 */ 441 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx); 442 443 /* 444 * Find the next allocated or free object. 445 * 446 * The objectp parameter is in-out. It will be updated to be the next 447 * object which is allocated. Ignore objects which have not been 448 * modified since txg. 449 * 450 * XXX Can only be called on a objset with no dirty data. 451 * 452 * Returns 0 on success, or ENOENT if there are no more objects. 453 */ 454 int dmu_object_next(objset_t *os, uint64_t *objectp, 455 boolean_t hole, uint64_t txg); 456 457 /* 458 * Set the number of levels on a dnode. nlevels must be greater than the 459 * current number of levels or an EINVAL will be returned. 460 */ 461 int dmu_object_set_nlevels(objset_t *os, uint64_t object, int nlevels, 462 dmu_tx_t *tx); 463 464 /* 465 * Set the data blocksize for an object. 466 * 467 * The object cannot have any blocks allocated beyond the first. If 468 * the first block is allocated already, the new size must be greater 469 * than the current block size. If these conditions are not met, 470 * ENOTSUP will be returned. 471 * 472 * Returns 0 on success, or EBUSY if there are any holds on the object 473 * contents, or ENOTSUP as described above. 474 */ 475 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, 476 int ibs, dmu_tx_t *tx); 477 478 /* 479 * Manually set the maxblkid on a dnode. This will adjust nlevels accordingly 480 * to accommodate the change. When calling this function, the caller must 481 * ensure that the object's nlevels can sufficiently support the new maxblkid. 482 */ 483 int dmu_object_set_maxblkid(objset_t *os, uint64_t object, uint64_t maxblkid, 484 dmu_tx_t *tx); 485 486 /* 487 * Set the checksum property on a dnode. The new checksum algorithm will 488 * apply to all newly written blocks; existing blocks will not be affected. 489 */ 490 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, 491 dmu_tx_t *tx); 492 493 /* 494 * Set the compress property on a dnode. The new compression algorithm will 495 * apply to all newly written blocks; existing blocks will not be affected. 496 */ 497 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, 498 dmu_tx_t *tx); 499 500 void dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset, 501 void *data, uint8_t etype, uint8_t comp, int uncompressed_size, 502 int compressed_size, int byteorder, dmu_tx_t *tx); 503 void dmu_redact(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 504 dmu_tx_t *tx); 505 506 /* 507 * Decide how to write a block: checksum, compression, number of copies, etc. 508 */ 509 #define WP_NOFILL 0x1 510 #define WP_DMU_SYNC 0x2 511 #define WP_SPILL 0x4 512 513 void dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, 514 struct zio_prop *zp); 515 516 /* 517 * The bonus data is accessed more or less like a regular buffer. 518 * You must dmu_bonus_hold() to get the buffer, which will give you a 519 * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus 520 * data. As with any normal buffer, you must call dmu_buf_will_dirty() 521 * before modifying it, and the 522 * object must be held in an assigned transaction before calling 523 * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus 524 * buffer as well. You must release what you hold with dmu_buf_rele(). 525 * 526 * Returns ENOENT, EIO, or 0. 527 */ 528 int dmu_bonus_hold(objset_t *os, uint64_t object, const void *tag, 529 dmu_buf_t **dbp); 530 int dmu_bonus_hold_by_dnode(dnode_t *dn, const void *tag, dmu_buf_t **dbp, 531 uint32_t flags); 532 int dmu_bonus_max(void); 533 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *); 534 int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *); 535 dmu_object_type_t dmu_get_bonustype(dmu_buf_t *); 536 int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *); 537 538 /* 539 * Special spill buffer support used by "SA" framework 540 */ 541 542 int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, uint32_t flags, const void *tag, 543 dmu_buf_t **dbp); 544 int dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags, 545 const void *tag, dmu_buf_t **dbp); 546 int dmu_spill_hold_existing(dmu_buf_t *bonus, const void *tag, dmu_buf_t **dbp); 547 548 /* 549 * Obtain the DMU buffer from the specified object which contains the 550 * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so 551 * that it will remain in memory. You must release the hold with 552 * dmu_buf_rele(). You must not access the dmu_buf_t after releasing 553 * what you hold. You must have a hold on any dmu_buf_t* you pass to the DMU. 554 * 555 * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill 556 * on the returned buffer before reading or writing the buffer's 557 * db_data. The comments for those routines describe what particular 558 * operations are valid after calling them. 559 * 560 * The object number must be a valid, allocated object number. 561 */ 562 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, 563 const void *tag, dmu_buf_t **, int flags); 564 int dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset, 565 uint64_t length, int read, const void *tag, int *numbufsp, 566 dmu_buf_t ***dbpp); 567 int dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset, 568 const void *tag, dmu_buf_t **dbp, int flags); 569 int dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset, 570 uint64_t length, boolean_t read, const void *tag, int *numbufsp, 571 dmu_buf_t ***dbpp, uint32_t flags); 572 /* 573 * Add a reference to a dmu buffer that has already been held via 574 * dmu_buf_hold() in the current context. 575 */ 576 void dmu_buf_add_ref(dmu_buf_t *db, const void *tag); 577 578 /* 579 * Attempt to add a reference to a dmu buffer that is in an unknown state, 580 * using a pointer that may have been invalidated by eviction processing. 581 * The request will succeed if the passed in dbuf still represents the 582 * same os/object/blkid, is ineligible for eviction, and has at least 583 * one hold by a user other than the syncer. 584 */ 585 boolean_t dmu_buf_try_add_ref(dmu_buf_t *, objset_t *os, uint64_t object, 586 uint64_t blkid, const void *tag); 587 588 void dmu_buf_rele(dmu_buf_t *db, const void *tag); 589 uint64_t dmu_buf_refcount(dmu_buf_t *db); 590 uint64_t dmu_buf_user_refcount(dmu_buf_t *db); 591 592 /* 593 * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a 594 * range of an object. A pointer to an array of dmu_buf_t*'s is 595 * returned (in *dbpp). 596 * 597 * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and 598 * frees the array. The hold on the array of buffers MUST be released 599 * with dmu_buf_rele_array. You can NOT release the hold on each buffer 600 * individually with dmu_buf_rele. 601 */ 602 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset, 603 uint64_t length, boolean_t read, const void *tag, 604 int *numbufsp, dmu_buf_t ***dbpp); 605 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, const void *tag); 606 607 typedef void dmu_buf_evict_func_t(void *user_ptr); 608 609 /* 610 * A DMU buffer user object may be associated with a dbuf for the 611 * duration of its lifetime. This allows the user of a dbuf (client) 612 * to attach private data to a dbuf (e.g. in-core only data such as a 613 * dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified 614 * when that dbuf has been evicted. Clients typically respond to the 615 * eviction notification by freeing their private data, thus ensuring 616 * the same lifetime for both dbuf and private data. 617 * 618 * The mapping from a dmu_buf_user_t to any client private data is the 619 * client's responsibility. All current consumers of the API with private 620 * data embed a dmu_buf_user_t as the first member of the structure for 621 * their private data. This allows conversions between the two types 622 * with a simple cast. Since the DMU buf user API never needs access 623 * to the private data, other strategies can be employed if necessary 624 * or convenient for the client (e.g. using container_of() to do the 625 * conversion for private data that cannot have the dmu_buf_user_t as 626 * its first member). 627 * 628 * Eviction callbacks are executed without the dbuf mutex held or any 629 * other type of mechanism to guarantee that the dbuf is still available. 630 * For this reason, users must assume the dbuf has already been freed 631 * and not reference the dbuf from the callback context. 632 * 633 * Users requesting "immediate eviction" are notified as soon as the dbuf 634 * is only referenced by dirty records (dirties == holds). Otherwise the 635 * notification occurs after eviction processing for the dbuf begins. 636 */ 637 typedef struct dmu_buf_user { 638 /* 639 * Asynchronous user eviction callback state. 640 */ 641 taskq_ent_t dbu_tqent; 642 643 /* 644 * This instance's eviction function pointers. 645 * 646 * dbu_evict_func_sync is called synchronously and then 647 * dbu_evict_func_async is executed asynchronously on a taskq. 648 */ 649 dmu_buf_evict_func_t *dbu_evict_func_sync; 650 dmu_buf_evict_func_t *dbu_evict_func_async; 651 #ifdef ZFS_DEBUG 652 /* 653 * Pointer to user's dbuf pointer. NULL for clients that do 654 * not associate a dbuf with their user data. 655 * 656 * The dbuf pointer is cleared upon eviction so as to catch 657 * use-after-evict bugs in clients. 658 */ 659 dmu_buf_t **dbu_clear_on_evict_dbufp; 660 #endif 661 } dmu_buf_user_t; 662 663 /* 664 * Initialize the given dmu_buf_user_t instance with the eviction function 665 * evict_func, to be called when the user is evicted. 666 * 667 * NOTE: This function should only be called once on a given dmu_buf_user_t. 668 * To allow enforcement of this, dbu must already be zeroed on entry. 669 */ 670 static inline void 671 dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func_sync, 672 dmu_buf_evict_func_t *evict_func_async, 673 dmu_buf_t **clear_on_evict_dbufp __maybe_unused) 674 { 675 ASSERT(dbu->dbu_evict_func_sync == NULL); 676 ASSERT(dbu->dbu_evict_func_async == NULL); 677 678 /* must have at least one evict func */ 679 IMPLY(evict_func_sync == NULL, evict_func_async != NULL); 680 dbu->dbu_evict_func_sync = evict_func_sync; 681 dbu->dbu_evict_func_async = evict_func_async; 682 taskq_init_ent(&dbu->dbu_tqent); 683 #ifdef ZFS_DEBUG 684 dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp; 685 #endif 686 } 687 688 /* 689 * Attach user data to a dbuf and mark it for normal (when the dbuf's 690 * data is cleared or its reference count goes to zero) eviction processing. 691 * 692 * Returns NULL on success, or the existing user if another user currently 693 * owns the buffer. 694 */ 695 void *dmu_buf_set_user(dmu_buf_t *db, dmu_buf_user_t *user); 696 697 /* 698 * Attach user data to a dbuf and mark it for immediate (its dirty and 699 * reference counts are equal) eviction processing. 700 * 701 * Returns NULL on success, or the existing user if another user currently 702 * owns the buffer. 703 */ 704 void *dmu_buf_set_user_ie(dmu_buf_t *db, dmu_buf_user_t *user); 705 706 /* 707 * Replace the current user of a dbuf. 708 * 709 * If given the current user of a dbuf, replaces the dbuf's user with 710 * "new_user" and returns the user data pointer that was replaced. 711 * Otherwise returns the current, and unmodified, dbuf user pointer. 712 */ 713 void *dmu_buf_replace_user(dmu_buf_t *db, 714 dmu_buf_user_t *old_user, dmu_buf_user_t *new_user); 715 716 /* 717 * Remove the specified user data for a DMU buffer. 718 * 719 * Returns the user that was removed on success, or the current user if 720 * another user currently owns the buffer. 721 */ 722 void *dmu_buf_remove_user(dmu_buf_t *db, dmu_buf_user_t *user); 723 724 /* 725 * Returns the user data (dmu_buf_user_t *) associated with this dbuf. 726 */ 727 void *dmu_buf_get_user(dmu_buf_t *db); 728 729 objset_t *dmu_buf_get_objset(dmu_buf_t *db); 730 dnode_t *dmu_buf_dnode_enter(dmu_buf_t *db); 731 void dmu_buf_dnode_exit(dmu_buf_t *db); 732 733 /* Block until any in-progress dmu buf user evictions complete. */ 734 void dmu_buf_user_evict_wait(void); 735 736 /* 737 * Returns the blkptr associated with this dbuf, or NULL if not set. 738 */ 739 struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db); 740 741 /* 742 * Indicate that you are going to modify the buffer's data (db_data). 743 * 744 * The transaction (tx) must be assigned to a txg (ie. you've called 745 * dmu_tx_assign()). The buffer's object must be held in the tx 746 * (ie. you've called dmu_tx_hold_object(tx, db->db_object)). 747 */ 748 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx); 749 boolean_t dmu_buf_is_dirty(dmu_buf_t *db, dmu_tx_t *tx); 750 void dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder, 751 const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_tx_t *tx); 752 753 /* 754 * You must create a transaction, then hold the objects which you will 755 * (or might) modify as part of this transaction. Then you must assign 756 * the transaction to a transaction group. Once the transaction has 757 * been assigned, you can modify buffers which belong to held objects as 758 * part of this transaction. You can't modify buffers before the 759 * transaction has been assigned; you can't modify buffers which don't 760 * belong to objects which this transaction holds; you can't hold 761 * objects once the transaction has been assigned. You may hold an 762 * object which you are going to free (with dmu_object_free()), but you 763 * don't have to. 764 * 765 * You can abort the transaction before it has been assigned. 766 * 767 * Note that you may hold buffers (with dmu_buf_hold) at any time, 768 * regardless of transaction state. 769 */ 770 771 #define DMU_NEW_OBJECT (-1ULL) 772 #define DMU_OBJECT_END (-1ULL) 773 774 dmu_tx_t *dmu_tx_create(objset_t *os); 775 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len); 776 void dmu_tx_hold_write_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, 777 int len); 778 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, 779 uint64_t len); 780 void dmu_tx_hold_free_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, 781 uint64_t len); 782 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name); 783 void dmu_tx_hold_zap_by_dnode(dmu_tx_t *tx, dnode_t *dn, int add, 784 const char *name); 785 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object); 786 void dmu_tx_hold_bonus_by_dnode(dmu_tx_t *tx, dnode_t *dn); 787 void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object); 788 void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow); 789 void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size); 790 void dmu_tx_abort(dmu_tx_t *tx); 791 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how); 792 void dmu_tx_wait(dmu_tx_t *tx); 793 void dmu_tx_commit(dmu_tx_t *tx); 794 void dmu_tx_mark_netfree(dmu_tx_t *tx); 795 796 /* 797 * To register a commit callback, dmu_tx_callback_register() must be called. 798 * 799 * dcb_data is a pointer to caller private data that is passed on as a 800 * callback parameter. The caller is responsible for properly allocating and 801 * freeing it. 802 * 803 * When registering a callback, the transaction must be already created, but 804 * it cannot be committed or aborted. It can be assigned to a txg or not. 805 * 806 * The callback will be called after the transaction has been safely written 807 * to stable storage and will also be called if the dmu_tx is aborted. 808 * If there is any error which prevents the transaction from being committed to 809 * disk, the callback will be called with a value of error != 0. 810 * 811 * When multiple callbacks are registered to the transaction, the callbacks 812 * will be called in reverse order to let Lustre, the only user of commit 813 * callback currently, take the fast path of its commit callback handling. 814 */ 815 typedef void dmu_tx_callback_func_t(void *dcb_data, int error); 816 817 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func, 818 void *dcb_data); 819 void dmu_tx_do_callbacks(list_t *cb_list, int error); 820 821 /* 822 * Free up the data blocks for a defined range of a file. If size is 823 * -1, the range from offset to end-of-file is freed. 824 */ 825 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, 826 uint64_t size, dmu_tx_t *tx); 827 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, 828 uint64_t size); 829 int dmu_free_long_object(objset_t *os, uint64_t object); 830 831 /* 832 * Convenience functions. 833 * 834 * Canfail routines will return 0 on success, or an errno if there is a 835 * nonrecoverable I/O error. 836 */ 837 #define DMU_READ_PREFETCH 0 /* prefetch */ 838 #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */ 839 #define DMU_READ_NO_DECRYPT 2 /* don't decrypt */ 840 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 841 void *buf, uint32_t flags); 842 int dmu_read_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, void *buf, 843 uint32_t flags); 844 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 845 const void *buf, dmu_tx_t *tx); 846 void dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, 847 const void *buf, dmu_tx_t *tx); 848 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 849 dmu_tx_t *tx); 850 #ifdef _KERNEL 851 int dmu_read_uio(objset_t *os, uint64_t object, zfs_uio_t *uio, uint64_t size); 852 int dmu_read_uio_dbuf(dmu_buf_t *zdb, zfs_uio_t *uio, uint64_t size); 853 int dmu_read_uio_dnode(dnode_t *dn, zfs_uio_t *uio, uint64_t size); 854 int dmu_write_uio(objset_t *os, uint64_t object, zfs_uio_t *uio, uint64_t size, 855 dmu_tx_t *tx); 856 int dmu_write_uio_dbuf(dmu_buf_t *zdb, zfs_uio_t *uio, uint64_t size, 857 dmu_tx_t *tx); 858 int dmu_write_uio_dnode(dnode_t *dn, zfs_uio_t *uio, uint64_t size, 859 dmu_tx_t *tx); 860 #endif 861 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size); 862 void dmu_return_arcbuf(struct arc_buf *buf); 863 int dmu_assign_arcbuf_by_dnode(dnode_t *dn, uint64_t offset, 864 struct arc_buf *buf, dmu_tx_t *tx); 865 int dmu_assign_arcbuf_by_dbuf(dmu_buf_t *handle, uint64_t offset, 866 struct arc_buf *buf, dmu_tx_t *tx); 867 #define dmu_assign_arcbuf dmu_assign_arcbuf_by_dbuf 868 extern int zfs_max_recordsize; 869 870 /* 871 * Asynchronously try to read in the data. 872 */ 873 void dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset, 874 uint64_t len, enum zio_priority pri); 875 876 typedef struct dmu_object_info { 877 /* All sizes are in bytes unless otherwise indicated. */ 878 uint32_t doi_data_block_size; 879 uint32_t doi_metadata_block_size; 880 dmu_object_type_t doi_type; 881 dmu_object_type_t doi_bonus_type; 882 uint64_t doi_bonus_size; 883 uint8_t doi_indirection; /* 2 = dnode->indirect->data */ 884 uint8_t doi_checksum; 885 uint8_t doi_compress; 886 uint8_t doi_nblkptr; 887 uint8_t doi_pad[4]; 888 uint64_t doi_dnodesize; 889 uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */ 890 uint64_t doi_max_offset; 891 uint64_t doi_fill_count; /* number of non-empty blocks */ 892 } dmu_object_info_t; 893 894 typedef void (*const arc_byteswap_func_t)(void *buf, size_t size); 895 896 typedef struct dmu_object_type_info { 897 dmu_object_byteswap_t ot_byteswap; 898 boolean_t ot_metadata; 899 boolean_t ot_dbuf_metadata_cache; 900 boolean_t ot_encrypt; 901 const char *ot_name; 902 } dmu_object_type_info_t; 903 904 typedef const struct dmu_object_byteswap_info { 905 arc_byteswap_func_t ob_func; 906 const char *ob_name; 907 } dmu_object_byteswap_info_t; 908 909 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES]; 910 extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS]; 911 912 /* 913 * Get information on a DMU object. 914 * 915 * Return 0 on success or ENOENT if object is not allocated. 916 * 917 * If doi is NULL, just indicates whether the object exists. 918 */ 919 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi); 920 void __dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi); 921 /* Like dmu_object_info, but faster if you have a held dnode in hand. */ 922 void dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi); 923 /* Like dmu_object_info, but faster if you have a held dbuf in hand. */ 924 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi); 925 /* 926 * Like dmu_object_info_from_db, but faster still when you only care about 927 * the size. 928 */ 929 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, 930 u_longlong_t *nblk512); 931 932 void dmu_object_dnsize_from_db(dmu_buf_t *db, int *dnsize); 933 934 typedef struct dmu_objset_stats { 935 uint64_t dds_num_clones; /* number of clones of this */ 936 uint64_t dds_creation_txg; 937 uint64_t dds_guid; 938 dmu_objset_type_t dds_type; 939 uint8_t dds_is_snapshot; 940 uint8_t dds_inconsistent; 941 uint8_t dds_redacted; 942 char dds_origin[ZFS_MAX_DATASET_NAME_LEN]; 943 } dmu_objset_stats_t; 944 945 /* 946 * Get stats on a dataset. 947 */ 948 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat); 949 950 /* 951 * Add entries to the nvlist for all the objset's properties. See 952 * zfs_prop_table[] and zfs(1m) for details on the properties. 953 */ 954 void dmu_objset_stats(objset_t *os, struct nvlist *nv); 955 956 /* 957 * Get the space usage statistics for statvfs(). 958 * 959 * refdbytes is the amount of space "referenced" by this objset. 960 * availbytes is the amount of space available to this objset, taking 961 * into account quotas & reservations, assuming that no other objsets 962 * use the space first. These values correspond to the 'referenced' and 963 * 'available' properties, described in the zfs(1m) manpage. 964 * 965 * usedobjs and availobjs are the number of objects currently allocated, 966 * and available. 967 */ 968 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, 969 uint64_t *usedobjsp, uint64_t *availobjsp); 970 971 /* 972 * The fsid_guid is a 56-bit ID that can change to avoid collisions. 973 * (Contrast with the ds_guid which is a 64-bit ID that will never 974 * change, so there is a small probability that it will collide.) 975 */ 976 uint64_t dmu_objset_fsid_guid(objset_t *os); 977 978 /* 979 * Get the [cm]time for an objset's snapshot dir 980 */ 981 inode_timespec_t dmu_objset_snap_cmtime(objset_t *os); 982 983 int dmu_objset_is_snapshot(objset_t *os); 984 985 extern struct spa *dmu_objset_spa(objset_t *os); 986 extern struct zilog *dmu_objset_zil(objset_t *os); 987 extern struct dsl_pool *dmu_objset_pool(objset_t *os); 988 extern struct dsl_dataset *dmu_objset_ds(objset_t *os); 989 extern void dmu_objset_name(objset_t *os, char *buf); 990 extern dmu_objset_type_t dmu_objset_type(objset_t *os); 991 extern uint64_t dmu_objset_id(objset_t *os); 992 extern uint64_t dmu_objset_dnodesize(objset_t *os); 993 extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os); 994 extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os); 995 extern int dmu_objset_blksize(objset_t *os); 996 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name, 997 uint64_t *id, uint64_t *offp, boolean_t *case_conflict); 998 extern int dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *val); 999 extern int dmu_snapshot_realname(objset_t *os, const char *name, char *real, 1000 int maxlen, boolean_t *conflict); 1001 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name, 1002 uint64_t *idp, uint64_t *offp); 1003 1004 typedef struct zfs_file_info { 1005 uint64_t zfi_user; 1006 uint64_t zfi_group; 1007 uint64_t zfi_project; 1008 uint64_t zfi_generation; 1009 } zfs_file_info_t; 1010 1011 typedef int file_info_cb_t(dmu_object_type_t bonustype, const void *data, 1012 struct zfs_file_info *zoi); 1013 extern void dmu_objset_register_type(dmu_objset_type_t ost, 1014 file_info_cb_t *cb); 1015 extern void dmu_objset_set_user(objset_t *os, void *user_ptr); 1016 extern void *dmu_objset_get_user(objset_t *os); 1017 1018 /* 1019 * Return the txg number for the given assigned transaction. 1020 */ 1021 uint64_t dmu_tx_get_txg(dmu_tx_t *tx); 1022 1023 /* 1024 * Synchronous write. 1025 * If a parent zio is provided this function initiates a write on the 1026 * provided buffer as a child of the parent zio. 1027 * In the absence of a parent zio, the write is completed synchronously. 1028 * At write completion, blk is filled with the bp of the written block. 1029 * Note that while the data covered by this function will be on stable 1030 * storage when the write completes this new data does not become a 1031 * permanent part of the file until the associated transaction commits. 1032 */ 1033 1034 /* 1035 * {zfs,zvol,ztest}_get_done() args 1036 */ 1037 typedef struct zgd { 1038 struct lwb *zgd_lwb; 1039 struct blkptr *zgd_bp; 1040 dmu_buf_t *zgd_db; 1041 struct zfs_locked_range *zgd_lr; 1042 void *zgd_private; 1043 } zgd_t; 1044 1045 typedef void dmu_sync_cb_t(zgd_t *arg, int error); 1046 int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd); 1047 1048 /* 1049 * Find the next hole or data block in file starting at *off 1050 * Return found offset in *off. Return ESRCH for end of file. 1051 */ 1052 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, 1053 uint64_t *off); 1054 1055 /* 1056 * Initial setup and final teardown. 1057 */ 1058 extern void dmu_init(void); 1059 extern void dmu_fini(void); 1060 1061 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp, 1062 uint64_t object, uint64_t offset, int len); 1063 void dmu_traverse_objset(objset_t *os, uint64_t txg_start, 1064 dmu_traverse_cb_t cb, void *arg); 1065 1066 int dmu_diff(const char *tosnap_name, const char *fromsnap_name, 1067 zfs_file_t *fp, offset_t *offp); 1068 1069 /* CRC64 table */ 1070 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ 1071 extern uint64_t zfs_crc64_table[256]; 1072 1073 extern int dmu_prefetch_max; 1074 1075 #ifdef __cplusplus 1076 } 1077 #endif 1078 1079 #endif /* _SYS_DMU_H */ 1080