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