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 */ 24 25 /* Portions Copyright 2010 Robert Milkowski */ 26 27 #ifndef _SYS_DMU_H 28 #define _SYS_DMU_H 29 30 /* 31 * This file describes the interface that the DMU provides for its 32 * consumers. 33 * 34 * The DMU also interacts with the SPA. That interface is described in 35 * dmu_spa.h. 36 */ 37 38 #include <sys/inttypes.h> 39 #include <sys/types.h> 40 #include <sys/param.h> 41 #include <sys/cred.h> 42 #include <sys/time.h> 43 44 #ifdef __cplusplus 45 extern "C" { 46 #endif 47 48 struct uio; 49 struct xuio; 50 struct page; 51 struct vnode; 52 struct spa; 53 struct zilog; 54 struct zio; 55 struct blkptr; 56 struct zap_cursor; 57 struct dsl_dataset; 58 struct dsl_pool; 59 struct dnode; 60 struct drr_begin; 61 struct drr_end; 62 struct zbookmark; 63 struct spa; 64 struct nvlist; 65 struct arc_buf; 66 struct zio_prop; 67 struct sa_handle; 68 69 typedef struct objset objset_t; 70 typedef struct dmu_tx dmu_tx_t; 71 typedef struct dsl_dir dsl_dir_t; 72 73 typedef enum dmu_object_type { 74 DMU_OT_NONE, 75 /* general: */ 76 DMU_OT_OBJECT_DIRECTORY, /* ZAP */ 77 DMU_OT_OBJECT_ARRAY, /* UINT64 */ 78 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ 79 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ 80 DMU_OT_BPLIST, /* UINT64 */ 81 DMU_OT_BPLIST_HDR, /* UINT64 */ 82 /* spa: */ 83 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ 84 DMU_OT_SPACE_MAP, /* UINT64 */ 85 /* zil: */ 86 DMU_OT_INTENT_LOG, /* UINT64 */ 87 /* dmu: */ 88 DMU_OT_DNODE, /* DNODE */ 89 DMU_OT_OBJSET, /* OBJSET */ 90 /* dsl: */ 91 DMU_OT_DSL_DIR, /* UINT64 */ 92 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ 93 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ 94 DMU_OT_DSL_PROPS, /* ZAP */ 95 DMU_OT_DSL_DATASET, /* UINT64 */ 96 /* zpl: */ 97 DMU_OT_ZNODE, /* ZNODE */ 98 DMU_OT_OLDACL, /* Old ACL */ 99 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ 100 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ 101 DMU_OT_MASTER_NODE, /* ZAP */ 102 DMU_OT_UNLINKED_SET, /* ZAP */ 103 /* zvol: */ 104 DMU_OT_ZVOL, /* UINT8 */ 105 DMU_OT_ZVOL_PROP, /* ZAP */ 106 /* other; for testing only! */ 107 DMU_OT_PLAIN_OTHER, /* UINT8 */ 108 DMU_OT_UINT64_OTHER, /* UINT64 */ 109 DMU_OT_ZAP_OTHER, /* ZAP */ 110 /* new object types: */ 111 DMU_OT_ERROR_LOG, /* ZAP */ 112 DMU_OT_SPA_HISTORY, /* UINT8 */ 113 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ 114 DMU_OT_POOL_PROPS, /* ZAP */ 115 DMU_OT_DSL_PERMS, /* ZAP */ 116 DMU_OT_ACL, /* ACL */ 117 DMU_OT_SYSACL, /* SYSACL */ 118 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ 119 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ 120 DMU_OT_NEXT_CLONES, /* ZAP */ 121 DMU_OT_SCRUB_QUEUE, /* ZAP */ 122 DMU_OT_USERGROUP_USED, /* ZAP */ 123 DMU_OT_USERGROUP_QUOTA, /* ZAP */ 124 DMU_OT_USERREFS, /* ZAP */ 125 DMU_OT_DDT_ZAP, /* ZAP */ 126 DMU_OT_DDT_STATS, /* ZAP */ 127 DMU_OT_SA, /* System attr */ 128 DMU_OT_SA_MASTER_NODE, /* ZAP */ 129 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ 130 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ 131 DMU_OT_NUMTYPES 132 } dmu_object_type_t; 133 134 typedef enum dmu_objset_type { 135 DMU_OST_NONE, 136 DMU_OST_META, 137 DMU_OST_ZFS, 138 DMU_OST_ZVOL, 139 DMU_OST_OTHER, /* For testing only! */ 140 DMU_OST_ANY, /* Be careful! */ 141 DMU_OST_NUMTYPES 142 } dmu_objset_type_t; 143 144 void byteswap_uint64_array(void *buf, size_t size); 145 void byteswap_uint32_array(void *buf, size_t size); 146 void byteswap_uint16_array(void *buf, size_t size); 147 void byteswap_uint8_array(void *buf, size_t size); 148 void zap_byteswap(void *buf, size_t size); 149 void zfs_oldacl_byteswap(void *buf, size_t size); 150 void zfs_acl_byteswap(void *buf, size_t size); 151 void zfs_znode_byteswap(void *buf, size_t size); 152 153 #define DS_FIND_SNAPSHOTS (1<<0) 154 #define DS_FIND_CHILDREN (1<<1) 155 156 /* 157 * The maximum number of bytes that can be accessed as part of one 158 * operation, including metadata. 159 */ 160 #define DMU_MAX_ACCESS (10<<20) /* 10MB */ 161 #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */ 162 163 #define DMU_USERUSED_OBJECT (-1ULL) 164 #define DMU_GROUPUSED_OBJECT (-2ULL) 165 #define DMU_DEADLIST_OBJECT (-3ULL) 166 167 /* 168 * artificial blkids for bonus buffer and spill blocks 169 */ 170 #define DMU_BONUS_BLKID (-1ULL) 171 #define DMU_SPILL_BLKID (-2ULL) 172 /* 173 * Public routines to create, destroy, open, and close objsets. 174 */ 175 int dmu_objset_hold(const char *name, void *tag, objset_t **osp); 176 int dmu_objset_own(const char *name, dmu_objset_type_t type, 177 boolean_t readonly, void *tag, objset_t **osp); 178 void dmu_objset_rele(objset_t *os, void *tag); 179 void dmu_objset_disown(objset_t *os, void *tag); 180 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp); 181 182 int dmu_objset_evict_dbufs(objset_t *os); 183 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, 184 void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg); 185 int dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin, 186 uint64_t flags); 187 int dmu_objset_destroy(const char *name, boolean_t defer); 188 int dmu_snapshots_destroy(char *fsname, char *snapname, boolean_t defer); 189 int dmu_objset_snapshot(char *fsname, char *snapname, struct nvlist *props, 190 boolean_t recursive); 191 int dmu_objset_rename(const char *name, const char *newname, 192 boolean_t recursive); 193 int dmu_objset_find(char *name, int func(const char *, void *), void *arg, 194 int flags); 195 void dmu_objset_byteswap(void *buf, size_t size); 196 197 typedef struct dmu_buf { 198 uint64_t db_object; /* object that this buffer is part of */ 199 uint64_t db_offset; /* byte offset in this object */ 200 uint64_t db_size; /* size of buffer in bytes */ 201 void *db_data; /* data in buffer */ 202 } dmu_buf_t; 203 204 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr); 205 206 /* 207 * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 208 */ 209 #define DMU_POOL_DIRECTORY_OBJECT 1 210 #define DMU_POOL_CONFIG "config" 211 #define DMU_POOL_ROOT_DATASET "root_dataset" 212 #define DMU_POOL_SYNC_BPLIST "sync_bplist" 213 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 214 #define DMU_POOL_ERRLOG_LAST "errlog_last" 215 #define DMU_POOL_SPARES "spares" 216 #define DMU_POOL_DEFLATE "deflate" 217 #define DMU_POOL_HISTORY "history" 218 #define DMU_POOL_PROPS "pool_props" 219 #define DMU_POOL_L2CACHE "l2cache" 220 #define DMU_POOL_TMP_USERREFS "tmp_userrefs" 221 #define DMU_POOL_DDT "DDT-%s-%s-%s" 222 #define DMU_POOL_DDT_STATS "DDT-statistics" 223 224 /* 4x8 zbookmark_t */ 225 #define DMU_POOL_SCRUB_BOOKMARK "scrub_bookmark" 226 /* 4x8 ddt_bookmark_t */ 227 #define DMU_POOL_SCRUB_DDT_BOOKMARK "scrub_ddt_bookmark" 228 /* 1x8 max_class */ 229 #define DMU_POOL_SCRUB_DDT_CLASS_MAX "scrub_ddt_class_max" 230 /* 1x8 zap obj DMU_OT_SCRUB_QUEUE */ 231 #define DMU_POOL_SCRUB_QUEUE "scrub_queue" 232 /* 1x8 txg */ 233 #define DMU_POOL_SCRUB_MIN_TXG "scrub_min_txg" 234 /* 1x8 txg */ 235 #define DMU_POOL_SCRUB_MAX_TXG "scrub_max_txg" 236 /* 1x4 enum scrub_func */ 237 #define DMU_POOL_SCRUB_FUNC "scrub_func" 238 /* 1x8 count */ 239 #define DMU_POOL_SCRUB_ERRORS "scrub_errors" 240 241 /* 242 * Allocate an object from this objset. The range of object numbers 243 * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode. 244 * 245 * The transaction must be assigned to a txg. The newly allocated 246 * object will be "held" in the transaction (ie. you can modify the 247 * newly allocated object in this transaction). 248 * 249 * dmu_object_alloc() chooses an object and returns it in *objectp. 250 * 251 * dmu_object_claim() allocates a specific object number. If that 252 * number is already allocated, it fails and returns EEXIST. 253 * 254 * Return 0 on success, or ENOSPC or EEXIST as specified above. 255 */ 256 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot, 257 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 258 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot, 259 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 260 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot, 261 int blocksize, dmu_object_type_t bonustype, int bonuslen); 262 263 /* 264 * Free an object from this objset. 265 * 266 * The object's data will be freed as well (ie. you don't need to call 267 * dmu_free(object, 0, -1, tx)). 268 * 269 * The object need not be held in the transaction. 270 * 271 * If there are any holds on this object's buffers (via dmu_buf_hold()), 272 * or tx holds on the object (via dmu_tx_hold_object()), you can not 273 * free it; it fails and returns EBUSY. 274 * 275 * If the object is not allocated, it fails and returns ENOENT. 276 * 277 * Return 0 on success, or EBUSY or ENOENT as specified above. 278 */ 279 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx); 280 281 /* 282 * Find the next allocated or free object. 283 * 284 * The objectp parameter is in-out. It will be updated to be the next 285 * object which is allocated. Ignore objects which have not been 286 * modified since txg. 287 * 288 * XXX Can only be called on a objset with no dirty data. 289 * 290 * Returns 0 on success, or ENOENT if there are no more objects. 291 */ 292 int dmu_object_next(objset_t *os, uint64_t *objectp, 293 boolean_t hole, uint64_t txg); 294 295 /* 296 * Set the data blocksize for an object. 297 * 298 * The object cannot have any blocks allcated beyond the first. If 299 * the first block is allocated already, the new size must be greater 300 * than the current block size. If these conditions are not met, 301 * ENOTSUP will be returned. 302 * 303 * Returns 0 on success, or EBUSY if there are any holds on the object 304 * contents, or ENOTSUP as described above. 305 */ 306 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, 307 int ibs, dmu_tx_t *tx); 308 309 /* 310 * Set the checksum property on a dnode. The new checksum algorithm will 311 * apply to all newly written blocks; existing blocks will not be affected. 312 */ 313 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, 314 dmu_tx_t *tx); 315 316 /* 317 * Set the compress property on a dnode. The new compression algorithm will 318 * apply to all newly written blocks; existing blocks will not be affected. 319 */ 320 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, 321 dmu_tx_t *tx); 322 323 /* 324 * Decide how to write a block: checksum, compression, number of copies, etc. 325 */ 326 #define WP_NOFILL 0x1 327 #define WP_DMU_SYNC 0x2 328 #define WP_SPILL 0x4 329 330 void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp, 331 struct zio_prop *zp); 332 /* 333 * The bonus data is accessed more or less like a regular buffer. 334 * You must dmu_bonus_hold() to get the buffer, which will give you a 335 * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus 336 * data. As with any normal buffer, you must call dmu_buf_read() to 337 * read db_data, dmu_buf_will_dirty() before modifying it, and the 338 * object must be held in an assigned transaction before calling 339 * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus 340 * buffer as well. You must release your hold with dmu_buf_rele(). 341 */ 342 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **); 343 int dmu_bonus_max(void); 344 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *); 345 int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *); 346 int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *); 347 348 /* 349 * Special spill buffer support used by "SA" framework 350 */ 351 352 int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); 353 int dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags, 354 void *tag, dmu_buf_t **dbp); 355 int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); 356 357 /* 358 * Obtain the DMU buffer from the specified object which contains the 359 * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so 360 * that it will remain in memory. You must release the hold with 361 * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your 362 * hold. You must have a hold on any dmu_buf_t* you pass to the DMU. 363 * 364 * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill 365 * on the returned buffer before reading or writing the buffer's 366 * db_data. The comments for those routines describe what particular 367 * operations are valid after calling them. 368 * 369 * The object number must be a valid, allocated object number. 370 */ 371 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, 372 void *tag, dmu_buf_t **, int flags); 373 void dmu_buf_add_ref(dmu_buf_t *db, void* tag); 374 void dmu_buf_rele(dmu_buf_t *db, void *tag); 375 uint64_t dmu_buf_refcount(dmu_buf_t *db); 376 377 /* 378 * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a 379 * range of an object. A pointer to an array of dmu_buf_t*'s is 380 * returned (in *dbpp). 381 * 382 * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and 383 * frees the array. The hold on the array of buffers MUST be released 384 * with dmu_buf_rele_array. You can NOT release the hold on each buffer 385 * individually with dmu_buf_rele. 386 */ 387 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset, 388 uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp); 389 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag); 390 391 /* 392 * Returns NULL on success, or the existing user ptr if it's already 393 * been set. 394 * 395 * user_ptr is for use by the user and can be obtained via dmu_buf_get_user(). 396 * 397 * user_data_ptr_ptr should be NULL, or a pointer to a pointer which 398 * will be set to db->db_data when you are allowed to access it. Note 399 * that db->db_data (the pointer) can change when you do dmu_buf_read(), 400 * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill(). 401 * *user_data_ptr_ptr will be set to the new value when it changes. 402 * 403 * If non-NULL, pageout func will be called when this buffer is being 404 * excised from the cache, so that you can clean up the data structure 405 * pointed to by user_ptr. 406 * 407 * dmu_evict_user() will call the pageout func for all buffers in a 408 * objset with a given pageout func. 409 */ 410 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr, 411 dmu_buf_evict_func_t *pageout_func); 412 /* 413 * set_user_ie is the same as set_user, but request immediate eviction 414 * when hold count goes to zero. 415 */ 416 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr, 417 void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func); 418 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr, 419 void *user_ptr, void *user_data_ptr_ptr, 420 dmu_buf_evict_func_t *pageout_func); 421 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func); 422 423 /* 424 * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set. 425 */ 426 void *dmu_buf_get_user(dmu_buf_t *db); 427 428 /* 429 * Indicate that you are going to modify the buffer's data (db_data). 430 * 431 * The transaction (tx) must be assigned to a txg (ie. you've called 432 * dmu_tx_assign()). The buffer's object must be held in the tx 433 * (ie. you've called dmu_tx_hold_object(tx, db->db_object)). 434 */ 435 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx); 436 437 /* 438 * Tells if the given dbuf is freeable. 439 */ 440 boolean_t dmu_buf_freeable(dmu_buf_t *); 441 442 /* 443 * You must create a transaction, then hold the objects which you will 444 * (or might) modify as part of this transaction. Then you must assign 445 * the transaction to a transaction group. Once the transaction has 446 * been assigned, you can modify buffers which belong to held objects as 447 * part of this transaction. You can't modify buffers before the 448 * transaction has been assigned; you can't modify buffers which don't 449 * belong to objects which this transaction holds; you can't hold 450 * objects once the transaction has been assigned. You may hold an 451 * object which you are going to free (with dmu_object_free()), but you 452 * don't have to. 453 * 454 * You can abort the transaction before it has been assigned. 455 * 456 * Note that you may hold buffers (with dmu_buf_hold) at any time, 457 * regardless of transaction state. 458 */ 459 460 #define DMU_NEW_OBJECT (-1ULL) 461 #define DMU_OBJECT_END (-1ULL) 462 463 dmu_tx_t *dmu_tx_create(objset_t *os); 464 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len); 465 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, 466 uint64_t len); 467 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name); 468 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object); 469 void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object); 470 void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow); 471 void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size); 472 void dmu_tx_abort(dmu_tx_t *tx); 473 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how); 474 void dmu_tx_wait(dmu_tx_t *tx); 475 void dmu_tx_commit(dmu_tx_t *tx); 476 477 /* 478 * To register a commit callback, dmu_tx_callback_register() must be called. 479 * 480 * dcb_data is a pointer to caller private data that is passed on as a 481 * callback parameter. The caller is responsible for properly allocating and 482 * freeing it. 483 * 484 * When registering a callback, the transaction must be already created, but 485 * it cannot be committed or aborted. It can be assigned to a txg or not. 486 * 487 * The callback will be called after the transaction has been safely written 488 * to stable storage and will also be called if the dmu_tx is aborted. 489 * If there is any error which prevents the transaction from being committed to 490 * disk, the callback will be called with a value of error != 0. 491 */ 492 typedef void dmu_tx_callback_func_t(void *dcb_data, int error); 493 494 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func, 495 void *dcb_data); 496 497 /* 498 * Free up the data blocks for a defined range of a file. If size is 499 * zero, the range from offset to end-of-file is freed. 500 */ 501 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, 502 uint64_t size, dmu_tx_t *tx); 503 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, 504 uint64_t size); 505 int dmu_free_object(objset_t *os, uint64_t object); 506 507 /* 508 * Convenience functions. 509 * 510 * Canfail routines will return 0 on success, or an errno if there is a 511 * nonrecoverable I/O error. 512 */ 513 #define DMU_READ_PREFETCH 0 /* prefetch */ 514 #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */ 515 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 516 void *buf, uint32_t flags); 517 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 518 const void *buf, dmu_tx_t *tx); 519 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 520 dmu_tx_t *tx); 521 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size); 522 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size, 523 dmu_tx_t *tx); 524 int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size, 525 dmu_tx_t *tx); 526 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, 527 uint64_t size, struct page *pp, dmu_tx_t *tx); 528 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size); 529 void dmu_return_arcbuf(struct arc_buf *buf); 530 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf, 531 dmu_tx_t *tx); 532 int dmu_xuio_init(struct xuio *uio, int niov); 533 void dmu_xuio_fini(struct xuio *uio); 534 int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off, 535 size_t n); 536 int dmu_xuio_cnt(struct xuio *uio); 537 struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i); 538 void dmu_xuio_clear(struct xuio *uio, int i); 539 void xuio_stat_wbuf_copied(); 540 void xuio_stat_wbuf_nocopy(); 541 542 extern int zfs_prefetch_disable; 543 544 /* 545 * Asynchronously try to read in the data. 546 */ 547 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset, 548 uint64_t len); 549 550 typedef struct dmu_object_info { 551 /* All sizes are in bytes unless otherwise indicated. */ 552 uint32_t doi_data_block_size; 553 uint32_t doi_metadata_block_size; 554 dmu_object_type_t doi_type; 555 dmu_object_type_t doi_bonus_type; 556 uint64_t doi_bonus_size; 557 uint8_t doi_indirection; /* 2 = dnode->indirect->data */ 558 uint8_t doi_checksum; 559 uint8_t doi_compress; 560 uint8_t doi_pad[5]; 561 uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */ 562 uint64_t doi_max_offset; 563 uint64_t doi_fill_count; /* number of non-empty blocks */ 564 } dmu_object_info_t; 565 566 typedef void arc_byteswap_func_t(void *buf, size_t size); 567 568 typedef struct dmu_object_type_info { 569 arc_byteswap_func_t *ot_byteswap; 570 boolean_t ot_metadata; 571 char *ot_name; 572 } dmu_object_type_info_t; 573 574 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES]; 575 576 /* 577 * Get information on a DMU object. 578 * 579 * Return 0 on success or ENOENT if object is not allocated. 580 * 581 * If doi is NULL, just indicates whether the object exists. 582 */ 583 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi); 584 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi); 585 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi); 586 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, 587 u_longlong_t *nblk512); 588 589 typedef struct dmu_objset_stats { 590 uint64_t dds_num_clones; /* number of clones of this */ 591 uint64_t dds_creation_txg; 592 uint64_t dds_guid; 593 dmu_objset_type_t dds_type; 594 uint8_t dds_is_snapshot; 595 uint8_t dds_inconsistent; 596 char dds_origin[MAXNAMELEN]; 597 } dmu_objset_stats_t; 598 599 /* 600 * Get stats on a dataset. 601 */ 602 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat); 603 604 /* 605 * Add entries to the nvlist for all the objset's properties. See 606 * zfs_prop_table[] and zfs(1m) for details on the properties. 607 */ 608 void dmu_objset_stats(objset_t *os, struct nvlist *nv); 609 610 /* 611 * Get the space usage statistics for statvfs(). 612 * 613 * refdbytes is the amount of space "referenced" by this objset. 614 * availbytes is the amount of space available to this objset, taking 615 * into account quotas & reservations, assuming that no other objsets 616 * use the space first. These values correspond to the 'referenced' and 617 * 'available' properties, described in the zfs(1m) manpage. 618 * 619 * usedobjs and availobjs are the number of objects currently allocated, 620 * and available. 621 */ 622 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, 623 uint64_t *usedobjsp, uint64_t *availobjsp); 624 625 /* 626 * The fsid_guid is a 56-bit ID that can change to avoid collisions. 627 * (Contrast with the ds_guid which is a 64-bit ID that will never 628 * change, so there is a small probability that it will collide.) 629 */ 630 uint64_t dmu_objset_fsid_guid(objset_t *os); 631 632 /* 633 * Get the [cm]time for an objset's snapshot dir 634 */ 635 timestruc_t dmu_objset_snap_cmtime(objset_t *os); 636 637 int dmu_objset_is_snapshot(objset_t *os); 638 639 extern struct spa *dmu_objset_spa(objset_t *os); 640 extern struct zilog *dmu_objset_zil(objset_t *os); 641 extern struct dsl_pool *dmu_objset_pool(objset_t *os); 642 extern struct dsl_dataset *dmu_objset_ds(objset_t *os); 643 extern void dmu_objset_name(objset_t *os, char *buf); 644 extern dmu_objset_type_t dmu_objset_type(objset_t *os); 645 extern uint64_t dmu_objset_id(objset_t *os); 646 extern uint64_t dmu_objset_syncprop(objset_t *os); 647 extern uint64_t dmu_objset_logbias(objset_t *os); 648 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name, 649 uint64_t *id, uint64_t *offp, boolean_t *case_conflict); 650 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real, 651 int maxlen, boolean_t *conflict); 652 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name, 653 uint64_t *idp, uint64_t *offp); 654 655 typedef int objset_used_cb_t(dmu_object_type_t bonustype, 656 void *bonus, uint64_t *userp, uint64_t *groupp); 657 extern void dmu_objset_register_type(dmu_objset_type_t ost, 658 objset_used_cb_t *cb); 659 extern void dmu_objset_set_user(objset_t *os, void *user_ptr); 660 extern void *dmu_objset_get_user(objset_t *os); 661 662 /* 663 * Return the txg number for the given assigned transaction. 664 */ 665 uint64_t dmu_tx_get_txg(dmu_tx_t *tx); 666 667 /* 668 * Synchronous write. 669 * If a parent zio is provided this function initiates a write on the 670 * provided buffer as a child of the parent zio. 671 * In the absence of a parent zio, the write is completed synchronously. 672 * At write completion, blk is filled with the bp of the written block. 673 * Note that while the data covered by this function will be on stable 674 * storage when the write completes this new data does not become a 675 * permanent part of the file until the associated transaction commits. 676 */ 677 678 /* 679 * {zfs,zvol,ztest}_get_done() args 680 */ 681 typedef struct zgd { 682 struct zilog *zgd_zilog; 683 struct blkptr *zgd_bp; 684 dmu_buf_t *zgd_db; 685 struct rl *zgd_rl; 686 void *zgd_private; 687 } zgd_t; 688 689 typedef void dmu_sync_cb_t(zgd_t *arg, int error); 690 int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd); 691 692 /* 693 * Find the next hole or data block in file starting at *off 694 * Return found offset in *off. Return ESRCH for end of file. 695 */ 696 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, 697 uint64_t *off); 698 699 /* 700 * Initial setup and final teardown. 701 */ 702 extern void dmu_init(void); 703 extern void dmu_fini(void); 704 705 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp, 706 uint64_t object, uint64_t offset, int len); 707 void dmu_traverse_objset(objset_t *os, uint64_t txg_start, 708 dmu_traverse_cb_t cb, void *arg); 709 710 int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin, 711 struct vnode *vp, offset_t *off); 712 713 typedef struct dmu_recv_cookie { 714 /* 715 * This structure is opaque! 716 * 717 * If logical and real are different, we are recving the stream 718 * into the "real" temporary clone, and then switching it with 719 * the "logical" target. 720 */ 721 struct dsl_dataset *drc_logical_ds; 722 struct dsl_dataset *drc_real_ds; 723 struct drr_begin *drc_drrb; 724 char *drc_tosnap; 725 char *drc_top_ds; 726 boolean_t drc_newfs; 727 boolean_t drc_force; 728 } dmu_recv_cookie_t; 729 730 int dmu_recv_begin(char *tofs, char *tosnap, char *topds, struct drr_begin *, 731 boolean_t force, objset_t *origin, dmu_recv_cookie_t *); 732 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp); 733 int dmu_recv_end(dmu_recv_cookie_t *drc); 734 735 /* CRC64 table */ 736 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ 737 extern uint64_t zfs_crc64_table[256]; 738 739 #ifdef __cplusplus 740 } 741 #endif 742 743 #endif /* _SYS_DMU_H */ 744