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 2010 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #ifndef _SYS_DMU_H 27 #define _SYS_DMU_H 28 29 /* 30 * This file describes the interface that the DMU provides for its 31 * consumers. 32 * 33 * The DMU also interacts with the SPA. That interface is described in 34 * dmu_spa.h. 35 */ 36 37 #include <sys/inttypes.h> 38 #include <sys/types.h> 39 #include <sys/param.h> 40 #include <sys/cred.h> 41 #include <sys/time.h> 42 43 #ifdef __cplusplus 44 extern "C" { 45 #endif 46 47 struct uio; 48 struct xuio; 49 struct page; 50 struct vnode; 51 struct spa; 52 struct zilog; 53 struct zio; 54 struct blkptr; 55 struct zap_cursor; 56 struct dsl_dataset; 57 struct dsl_pool; 58 struct dnode; 59 struct drr_begin; 60 struct drr_end; 61 struct zbookmark; 62 struct spa; 63 struct nvlist; 64 struct arc_buf; 65 struct zio_prop; 66 struct sa_handle; 67 68 typedef struct objset objset_t; 69 typedef struct dmu_tx dmu_tx_t; 70 typedef struct dsl_dir dsl_dir_t; 71 72 typedef enum dmu_object_type { 73 DMU_OT_NONE, 74 /* general: */ 75 DMU_OT_OBJECT_DIRECTORY, /* ZAP */ 76 DMU_OT_OBJECT_ARRAY, /* UINT64 */ 77 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ 78 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ 79 DMU_OT_BPLIST, /* UINT64 */ 80 DMU_OT_BPLIST_HDR, /* UINT64 */ 81 /* spa: */ 82 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ 83 DMU_OT_SPACE_MAP, /* UINT64 */ 84 /* zil: */ 85 DMU_OT_INTENT_LOG, /* UINT64 */ 86 /* dmu: */ 87 DMU_OT_DNODE, /* DNODE */ 88 DMU_OT_OBJSET, /* OBJSET */ 89 /* dsl: */ 90 DMU_OT_DSL_DIR, /* UINT64 */ 91 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ 92 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ 93 DMU_OT_DSL_PROPS, /* ZAP */ 94 DMU_OT_DSL_DATASET, /* UINT64 */ 95 /* zpl: */ 96 DMU_OT_ZNODE, /* ZNODE */ 97 DMU_OT_OLDACL, /* Old ACL */ 98 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ 99 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ 100 DMU_OT_MASTER_NODE, /* ZAP */ 101 DMU_OT_UNLINKED_SET, /* ZAP */ 102 /* zvol: */ 103 DMU_OT_ZVOL, /* UINT8 */ 104 DMU_OT_ZVOL_PROP, /* ZAP */ 105 /* other; for testing only! */ 106 DMU_OT_PLAIN_OTHER, /* UINT8 */ 107 DMU_OT_UINT64_OTHER, /* UINT64 */ 108 DMU_OT_ZAP_OTHER, /* ZAP */ 109 /* new object types: */ 110 DMU_OT_ERROR_LOG, /* ZAP */ 111 DMU_OT_SPA_HISTORY, /* UINT8 */ 112 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ 113 DMU_OT_POOL_PROPS, /* ZAP */ 114 DMU_OT_DSL_PERMS, /* ZAP */ 115 DMU_OT_ACL, /* ACL */ 116 DMU_OT_SYSACL, /* SYSACL */ 117 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ 118 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ 119 DMU_OT_NEXT_CLONES, /* ZAP */ 120 DMU_OT_SCRUB_QUEUE, /* ZAP */ 121 DMU_OT_USERGROUP_USED, /* ZAP */ 122 DMU_OT_USERGROUP_QUOTA, /* ZAP */ 123 DMU_OT_USERREFS, /* ZAP */ 124 DMU_OT_DDT_ZAP, /* ZAP */ 125 DMU_OT_DDT_STATS, /* ZAP */ 126 DMU_OT_SA, /* System attr */ 127 DMU_OT_SA_MASTER_NODE, /* ZAP */ 128 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ 129 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ 130 DMU_OT_NUMTYPES 131 } dmu_object_type_t; 132 133 typedef enum dmu_objset_type { 134 DMU_OST_NONE, 135 DMU_OST_META, 136 DMU_OST_ZFS, 137 DMU_OST_ZVOL, 138 DMU_OST_OTHER, /* For testing only! */ 139 DMU_OST_ANY, /* Be careful! */ 140 DMU_OST_NUMTYPES 141 } dmu_objset_type_t; 142 143 void byteswap_uint64_array(void *buf, size_t size); 144 void byteswap_uint32_array(void *buf, size_t size); 145 void byteswap_uint16_array(void *buf, size_t size); 146 void byteswap_uint8_array(void *buf, size_t size); 147 void zap_byteswap(void *buf, size_t size); 148 void zfs_oldacl_byteswap(void *buf, size_t size); 149 void zfs_acl_byteswap(void *buf, size_t size); 150 void zfs_znode_byteswap(void *buf, size_t size); 151 152 #define DS_FIND_SNAPSHOTS (1<<0) 153 #define DS_FIND_CHILDREN (1<<1) 154 155 /* 156 * The maximum number of bytes that can be accessed as part of one 157 * operation, including metadata. 158 */ 159 #define DMU_MAX_ACCESS (10<<20) /* 10MB */ 160 #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */ 161 162 #define DMU_USERUSED_OBJECT (-1ULL) 163 #define DMU_GROUPUSED_OBJECT (-2ULL) 164 #define DMU_DEADLIST_OBJECT (-3ULL) 165 166 /* 167 * artificial blkids for bonus buffer and spill blocks 168 */ 169 #define DMU_BONUS_BLKID (-1ULL) 170 #define DMU_SPILL_BLKID (-2ULL) 171 /* 172 * Public routines to create, destroy, open, and close objsets. 173 */ 174 int dmu_objset_hold(const char *name, void *tag, objset_t **osp); 175 int dmu_objset_own(const char *name, dmu_objset_type_t type, 176 boolean_t readonly, void *tag, objset_t **osp); 177 void dmu_objset_rele(objset_t *os, void *tag); 178 void dmu_objset_disown(objset_t *os, void *tag); 179 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp); 180 181 int dmu_objset_evict_dbufs(objset_t *os); 182 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, 183 void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg); 184 int dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin, 185 uint64_t flags); 186 int dmu_objset_destroy(const char *name, boolean_t defer); 187 int dmu_snapshots_destroy(char *fsname, char *snapname, boolean_t defer); 188 int dmu_objset_snapshot(char *fsname, char *snapname, struct nvlist *props, 189 boolean_t recursive); 190 int dmu_objset_rename(const char *name, const char *newname, 191 boolean_t recursive); 192 int dmu_objset_find(char *name, int func(const char *, void *), void *arg, 193 int flags); 194 void dmu_objset_byteswap(void *buf, size_t size); 195 196 typedef struct dmu_buf { 197 uint64_t db_object; /* object that this buffer is part of */ 198 uint64_t db_offset; /* byte offset in this object */ 199 uint64_t db_size; /* size of buffer in bytes */ 200 void *db_data; /* data in buffer */ 201 } dmu_buf_t; 202 203 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr); 204 205 /* 206 * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 207 */ 208 #define DMU_POOL_DIRECTORY_OBJECT 1 209 #define DMU_POOL_CONFIG "config" 210 #define DMU_POOL_ROOT_DATASET "root_dataset" 211 #define DMU_POOL_SYNC_BPLIST "sync_bplist" 212 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 213 #define DMU_POOL_ERRLOG_LAST "errlog_last" 214 #define DMU_POOL_SPARES "spares" 215 #define DMU_POOL_DEFLATE "deflate" 216 #define DMU_POOL_HISTORY "history" 217 #define DMU_POOL_PROPS "pool_props" 218 #define DMU_POOL_L2CACHE "l2cache" 219 #define DMU_POOL_TMP_USERREFS "tmp_userrefs" 220 #define DMU_POOL_DDT "DDT-%s-%s-%s" 221 #define DMU_POOL_DDT_STATS "DDT-statistics" 222 223 /* 4x8 zbookmark_t */ 224 #define DMU_POOL_SCRUB_BOOKMARK "scrub_bookmark" 225 /* 4x8 ddt_bookmark_t */ 226 #define DMU_POOL_SCRUB_DDT_BOOKMARK "scrub_ddt_bookmark" 227 /* 1x8 max_class */ 228 #define DMU_POOL_SCRUB_DDT_CLASS_MAX "scrub_ddt_class_max" 229 /* 1x8 zap obj DMU_OT_SCRUB_QUEUE */ 230 #define DMU_POOL_SCRUB_QUEUE "scrub_queue" 231 /* 1x8 txg */ 232 #define DMU_POOL_SCRUB_MIN_TXG "scrub_min_txg" 233 /* 1x8 txg */ 234 #define DMU_POOL_SCRUB_MAX_TXG "scrub_max_txg" 235 /* 1x4 enum scrub_func */ 236 #define DMU_POOL_SCRUB_FUNC "scrub_func" 237 /* 1x8 count */ 238 #define DMU_POOL_SCRUB_ERRORS "scrub_errors" 239 240 /* 241 * Allocate an object from this objset. The range of object numbers 242 * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode. 243 * 244 * The transaction must be assigned to a txg. The newly allocated 245 * object will be "held" in the transaction (ie. you can modify the 246 * newly allocated object in this transaction). 247 * 248 * dmu_object_alloc() chooses an object and returns it in *objectp. 249 * 250 * dmu_object_claim() allocates a specific object number. If that 251 * number is already allocated, it fails and returns EEXIST. 252 * 253 * Return 0 on success, or ENOSPC or EEXIST as specified above. 254 */ 255 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot, 256 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 257 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot, 258 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 259 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot, 260 int blocksize, dmu_object_type_t bonustype, int bonuslen); 261 262 /* 263 * Free an object from this objset. 264 * 265 * The object's data will be freed as well (ie. you don't need to call 266 * dmu_free(object, 0, -1, tx)). 267 * 268 * The object need not be held in the transaction. 269 * 270 * If there are any holds on this object's buffers (via dmu_buf_hold()), 271 * or tx holds on the object (via dmu_tx_hold_object()), you can not 272 * free it; it fails and returns EBUSY. 273 * 274 * If the object is not allocated, it fails and returns ENOENT. 275 * 276 * Return 0 on success, or EBUSY or ENOENT as specified above. 277 */ 278 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx); 279 280 /* 281 * Find the next allocated or free object. 282 * 283 * The objectp parameter is in-out. It will be updated to be the next 284 * object which is allocated. Ignore objects which have not been 285 * modified since txg. 286 * 287 * XXX Can only be called on a objset with no dirty data. 288 * 289 * Returns 0 on success, or ENOENT if there are no more objects. 290 */ 291 int dmu_object_next(objset_t *os, uint64_t *objectp, 292 boolean_t hole, uint64_t txg); 293 294 /* 295 * Set the data blocksize for an object. 296 * 297 * The object cannot have any blocks allcated beyond the first. If 298 * the first block is allocated already, the new size must be greater 299 * than the current block size. If these conditions are not met, 300 * ENOTSUP will be returned. 301 * 302 * Returns 0 on success, or EBUSY if there are any holds on the object 303 * contents, or ENOTSUP as described above. 304 */ 305 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, 306 int ibs, dmu_tx_t *tx); 307 308 /* 309 * Set the checksum property on a dnode. The new checksum algorithm will 310 * apply to all newly written blocks; existing blocks will not be affected. 311 */ 312 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, 313 dmu_tx_t *tx); 314 315 /* 316 * Set the compress property on a dnode. The new compression algorithm will 317 * apply to all newly written blocks; existing blocks will not be affected. 318 */ 319 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, 320 dmu_tx_t *tx); 321 322 /* 323 * Decide how to write a block: checksum, compression, number of copies, etc. 324 */ 325 #define WP_NOFILL 0x1 326 #define WP_DMU_SYNC 0x2 327 #define WP_SPILL 0x4 328 329 void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp, 330 struct zio_prop *zp); 331 /* 332 * The bonus data is accessed more or less like a regular buffer. 333 * You must dmu_bonus_hold() to get the buffer, which will give you a 334 * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus 335 * data. As with any normal buffer, you must call dmu_buf_read() to 336 * read db_data, dmu_buf_will_dirty() before modifying it, and the 337 * object must be held in an assigned transaction before calling 338 * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus 339 * buffer as well. You must release your hold with dmu_buf_rele(). 340 */ 341 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **); 342 int dmu_bonus_max(void); 343 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *); 344 int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *); 345 int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *); 346 347 /* 348 * Special spill buffer support used by "SA" framework 349 */ 350 351 int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); 352 int dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags, 353 void *tag, dmu_buf_t **dbp); 354 int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); 355 356 /* 357 * Obtain the DMU buffer from the specified object which contains the 358 * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so 359 * that it will remain in memory. You must release the hold with 360 * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your 361 * hold. You must have a hold on any dmu_buf_t* you pass to the DMU. 362 * 363 * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill 364 * on the returned buffer before reading or writing the buffer's 365 * db_data. The comments for those routines describe what particular 366 * operations are valid after calling them. 367 * 368 * The object number must be a valid, allocated object number. 369 */ 370 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, 371 void *tag, dmu_buf_t **); 372 void dmu_buf_add_ref(dmu_buf_t *db, void* tag); 373 void dmu_buf_rele(dmu_buf_t *db, void *tag); 374 uint64_t dmu_buf_refcount(dmu_buf_t *db); 375 376 /* 377 * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a 378 * range of an object. A pointer to an array of dmu_buf_t*'s is 379 * returned (in *dbpp). 380 * 381 * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and 382 * frees the array. The hold on the array of buffers MUST be released 383 * with dmu_buf_rele_array. You can NOT release the hold on each buffer 384 * individually with dmu_buf_rele. 385 */ 386 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset, 387 uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp); 388 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag); 389 390 /* 391 * Returns NULL on success, or the existing user ptr if it's already 392 * been set. 393 * 394 * user_ptr is for use by the user and can be obtained via dmu_buf_get_user(). 395 * 396 * user_data_ptr_ptr should be NULL, or a pointer to a pointer which 397 * will be set to db->db_data when you are allowed to access it. Note 398 * that db->db_data (the pointer) can change when you do dmu_buf_read(), 399 * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill(). 400 * *user_data_ptr_ptr will be set to the new value when it changes. 401 * 402 * If non-NULL, pageout func will be called when this buffer is being 403 * excised from the cache, so that you can clean up the data structure 404 * pointed to by user_ptr. 405 * 406 * dmu_evict_user() will call the pageout func for all buffers in a 407 * objset with a given pageout func. 408 */ 409 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr, 410 dmu_buf_evict_func_t *pageout_func); 411 /* 412 * set_user_ie is the same as set_user, but request immediate eviction 413 * when hold count goes to zero. 414 */ 415 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr, 416 void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func); 417 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr, 418 void *user_ptr, void *user_data_ptr_ptr, 419 dmu_buf_evict_func_t *pageout_func); 420 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func); 421 422 /* 423 * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set. 424 */ 425 void *dmu_buf_get_user(dmu_buf_t *db); 426 427 /* 428 * Indicate that you are going to modify the buffer's data (db_data). 429 * 430 * The transaction (tx) must be assigned to a txg (ie. you've called 431 * dmu_tx_assign()). The buffer's object must be held in the tx 432 * (ie. you've called dmu_tx_hold_object(tx, db->db_object)). 433 */ 434 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx); 435 436 /* 437 * Tells if the given dbuf is freeable. 438 */ 439 boolean_t dmu_buf_freeable(dmu_buf_t *); 440 441 /* 442 * You must create a transaction, then hold the objects which you will 443 * (or might) modify as part of this transaction. Then you must assign 444 * the transaction to a transaction group. Once the transaction has 445 * been assigned, you can modify buffers which belong to held objects as 446 * part of this transaction. You can't modify buffers before the 447 * transaction has been assigned; you can't modify buffers which don't 448 * belong to objects which this transaction holds; you can't hold 449 * objects once the transaction has been assigned. You may hold an 450 * object which you are going to free (with dmu_object_free()), but you 451 * don't have to. 452 * 453 * You can abort the transaction before it has been assigned. 454 * 455 * Note that you may hold buffers (with dmu_buf_hold) at any time, 456 * regardless of transaction state. 457 */ 458 459 #define DMU_NEW_OBJECT (-1ULL) 460 #define DMU_OBJECT_END (-1ULL) 461 462 dmu_tx_t *dmu_tx_create(objset_t *os); 463 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len); 464 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, 465 uint64_t len); 466 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name); 467 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object); 468 void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object); 469 void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow); 470 void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size); 471 void dmu_tx_abort(dmu_tx_t *tx); 472 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how); 473 void dmu_tx_wait(dmu_tx_t *tx); 474 void dmu_tx_commit(dmu_tx_t *tx); 475 476 /* 477 * To register a commit callback, dmu_tx_callback_register() must be called. 478 * 479 * dcb_data is a pointer to caller private data that is passed on as a 480 * callback parameter. The caller is responsible for properly allocating and 481 * freeing it. 482 * 483 * When registering a callback, the transaction must be already created, but 484 * it cannot be committed or aborted. It can be assigned to a txg or not. 485 * 486 * The callback will be called after the transaction has been safely written 487 * to stable storage and will also be called if the dmu_tx is aborted. 488 * If there is any error which prevents the transaction from being committed to 489 * disk, the callback will be called with a value of error != 0. 490 */ 491 typedef void dmu_tx_callback_func_t(void *dcb_data, int error); 492 493 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func, 494 void *dcb_data); 495 496 /* 497 * Free up the data blocks for a defined range of a file. If size is 498 * zero, the range from offset to end-of-file is freed. 499 */ 500 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, 501 uint64_t size, dmu_tx_t *tx); 502 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, 503 uint64_t size); 504 int dmu_free_object(objset_t *os, uint64_t object); 505 506 /* 507 * Convenience functions. 508 * 509 * Canfail routines will return 0 on success, or an errno if there is a 510 * nonrecoverable I/O error. 511 */ 512 #define DMU_READ_PREFETCH 0 /* prefetch */ 513 #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */ 514 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 515 void *buf, uint32_t flags); 516 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 517 const void *buf, dmu_tx_t *tx); 518 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 519 dmu_tx_t *tx); 520 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size); 521 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size, 522 dmu_tx_t *tx); 523 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, 524 uint64_t size, struct page *pp, dmu_tx_t *tx); 525 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size); 526 void dmu_return_arcbuf(struct arc_buf *buf); 527 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf, 528 dmu_tx_t *tx); 529 int dmu_xuio_init(struct xuio *uio, int niov); 530 void dmu_xuio_fini(struct xuio *uio); 531 int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off, 532 size_t n); 533 int dmu_xuio_cnt(struct xuio *uio); 534 struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i); 535 void dmu_xuio_clear(struct xuio *uio, int i); 536 void xuio_stat_wbuf_copied(); 537 void xuio_stat_wbuf_nocopy(); 538 539 extern int zfs_prefetch_disable; 540 541 /* 542 * Asynchronously try to read in the data. 543 */ 544 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset, 545 uint64_t len); 546 547 typedef struct dmu_object_info { 548 /* All sizes are in bytes unless otherwise indicated. */ 549 uint32_t doi_data_block_size; 550 uint32_t doi_metadata_block_size; 551 dmu_object_type_t doi_type; 552 dmu_object_type_t doi_bonus_type; 553 uint64_t doi_bonus_size; 554 uint8_t doi_indirection; /* 2 = dnode->indirect->data */ 555 uint8_t doi_checksum; 556 uint8_t doi_compress; 557 uint8_t doi_pad[5]; 558 uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */ 559 uint64_t doi_max_offset; 560 uint64_t doi_fill_count; /* number of non-empty blocks */ 561 } dmu_object_info_t; 562 563 typedef void arc_byteswap_func_t(void *buf, size_t size); 564 565 typedef struct dmu_object_type_info { 566 arc_byteswap_func_t *ot_byteswap; 567 boolean_t ot_metadata; 568 char *ot_name; 569 } dmu_object_type_info_t; 570 571 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES]; 572 573 /* 574 * Get information on a DMU object. 575 * 576 * Return 0 on success or ENOENT if object is not allocated. 577 * 578 * If doi is NULL, just indicates whether the object exists. 579 */ 580 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi); 581 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi); 582 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi); 583 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, 584 u_longlong_t *nblk512); 585 586 typedef struct dmu_objset_stats { 587 uint64_t dds_num_clones; /* number of clones of this */ 588 uint64_t dds_creation_txg; 589 uint64_t dds_guid; 590 dmu_objset_type_t dds_type; 591 uint8_t dds_is_snapshot; 592 uint8_t dds_inconsistent; 593 char dds_origin[MAXNAMELEN]; 594 } dmu_objset_stats_t; 595 596 /* 597 * Get stats on a dataset. 598 */ 599 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat); 600 601 /* 602 * Add entries to the nvlist for all the objset's properties. See 603 * zfs_prop_table[] and zfs(1m) for details on the properties. 604 */ 605 void dmu_objset_stats(objset_t *os, struct nvlist *nv); 606 607 /* 608 * Get the space usage statistics for statvfs(). 609 * 610 * refdbytes is the amount of space "referenced" by this objset. 611 * availbytes is the amount of space available to this objset, taking 612 * into account quotas & reservations, assuming that no other objsets 613 * use the space first. These values correspond to the 'referenced' and 614 * 'available' properties, described in the zfs(1m) manpage. 615 * 616 * usedobjs and availobjs are the number of objects currently allocated, 617 * and available. 618 */ 619 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, 620 uint64_t *usedobjsp, uint64_t *availobjsp); 621 622 /* 623 * The fsid_guid is a 56-bit ID that can change to avoid collisions. 624 * (Contrast with the ds_guid which is a 64-bit ID that will never 625 * change, so there is a small probability that it will collide.) 626 */ 627 uint64_t dmu_objset_fsid_guid(objset_t *os); 628 629 /* 630 * Get the [cm]time for an objset's snapshot dir 631 */ 632 timestruc_t dmu_objset_snap_cmtime(objset_t *os); 633 634 int dmu_objset_is_snapshot(objset_t *os); 635 636 extern struct spa *dmu_objset_spa(objset_t *os); 637 extern struct zilog *dmu_objset_zil(objset_t *os); 638 extern struct dsl_pool *dmu_objset_pool(objset_t *os); 639 extern struct dsl_dataset *dmu_objset_ds(objset_t *os); 640 extern void dmu_objset_name(objset_t *os, char *buf); 641 extern dmu_objset_type_t dmu_objset_type(objset_t *os); 642 extern uint64_t dmu_objset_id(objset_t *os); 643 extern uint64_t dmu_objset_logbias(objset_t *os); 644 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name, 645 uint64_t *id, uint64_t *offp, boolean_t *case_conflict); 646 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real, 647 int maxlen, boolean_t *conflict); 648 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name, 649 uint64_t *idp, uint64_t *offp); 650 651 typedef int objset_used_cb_t(dmu_object_type_t bonustype, 652 void *bonus, uint64_t *userp, uint64_t *groupp); 653 extern void dmu_objset_register_type(dmu_objset_type_t ost, 654 objset_used_cb_t *cb); 655 extern void dmu_objset_set_user(objset_t *os, void *user_ptr); 656 extern void *dmu_objset_get_user(objset_t *os); 657 658 /* 659 * Return the txg number for the given assigned transaction. 660 */ 661 uint64_t dmu_tx_get_txg(dmu_tx_t *tx); 662 663 /* 664 * Synchronous write. 665 * If a parent zio is provided this function initiates a write on the 666 * provided buffer as a child of the parent zio. 667 * In the absence of a parent zio, the write is completed synchronously. 668 * At write completion, blk is filled with the bp of the written block. 669 * Note that while the data covered by this function will be on stable 670 * storage when the write completes this new data does not become a 671 * permanent part of the file until the associated transaction commits. 672 */ 673 674 /* 675 * {zfs,zvol,ztest}_get_done() args 676 */ 677 typedef struct zgd { 678 struct zilog *zgd_zilog; 679 struct blkptr *zgd_bp; 680 dmu_buf_t *zgd_db; 681 struct rl *zgd_rl; 682 void *zgd_private; 683 } zgd_t; 684 685 typedef void dmu_sync_cb_t(zgd_t *arg, int error); 686 int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd); 687 688 /* 689 * Find the next hole or data block in file starting at *off 690 * Return found offset in *off. Return ESRCH for end of file. 691 */ 692 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, 693 uint64_t *off); 694 695 /* 696 * Initial setup and final teardown. 697 */ 698 extern void dmu_init(void); 699 extern void dmu_fini(void); 700 701 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp, 702 uint64_t object, uint64_t offset, int len); 703 void dmu_traverse_objset(objset_t *os, uint64_t txg_start, 704 dmu_traverse_cb_t cb, void *arg); 705 706 int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin, 707 struct vnode *vp, offset_t *off); 708 709 typedef struct dmu_recv_cookie { 710 /* 711 * This structure is opaque! 712 * 713 * If logical and real are different, we are recving the stream 714 * into the "real" temporary clone, and then switching it with 715 * the "logical" target. 716 */ 717 struct dsl_dataset *drc_logical_ds; 718 struct dsl_dataset *drc_real_ds; 719 struct drr_begin *drc_drrb; 720 char *drc_tosnap; 721 char *drc_top_ds; 722 boolean_t drc_newfs; 723 boolean_t drc_force; 724 } dmu_recv_cookie_t; 725 726 int dmu_recv_begin(char *tofs, char *tosnap, char *topds, struct drr_begin *, 727 boolean_t force, objset_t *origin, dmu_recv_cookie_t *); 728 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp); 729 int dmu_recv_end(dmu_recv_cookie_t *drc); 730 731 /* CRC64 table */ 732 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ 733 extern uint64_t zfs_crc64_table[256]; 734 735 #ifdef __cplusplus 736 } 737 #endif 738 739 #endif /* _SYS_DMU_H */ 740