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 * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 26 */ 27 28 /* Portions Copyright 2010 Robert Milkowski */ 29 30 #ifndef _SYS_DMU_H 31 #define _SYS_DMU_H 32 33 /* 34 * This file describes the interface that the DMU provides for its 35 * consumers. 36 * 37 * The DMU also interacts with the SPA. That interface is described in 38 * dmu_spa.h. 39 */ 40 41 #include <sys/inttypes.h> 42 #include <sys/types.h> 43 #include <sys/param.h> 44 #include <sys/cred.h> 45 #include <sys/time.h> 46 47 #ifdef __cplusplus 48 extern "C" { 49 #endif 50 51 struct uio; 52 struct xuio; 53 struct page; 54 struct vnode; 55 struct spa; 56 struct zilog; 57 struct zio; 58 struct blkptr; 59 struct zap_cursor; 60 struct dsl_dataset; 61 struct dsl_pool; 62 struct dnode; 63 struct drr_begin; 64 struct drr_end; 65 struct zbookmark; 66 struct spa; 67 struct nvlist; 68 struct arc_buf; 69 struct zio_prop; 70 struct sa_handle; 71 72 typedef struct objset objset_t; 73 typedef struct dmu_tx dmu_tx_t; 74 typedef struct dsl_dir dsl_dir_t; 75 76 typedef enum dmu_object_type { 77 DMU_OT_NONE, 78 /* general: */ 79 DMU_OT_OBJECT_DIRECTORY, /* ZAP */ 80 DMU_OT_OBJECT_ARRAY, /* UINT64 */ 81 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ 82 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ 83 DMU_OT_BPOBJ, /* UINT64 */ 84 DMU_OT_BPOBJ_HDR, /* UINT64 */ 85 /* spa: */ 86 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ 87 DMU_OT_SPACE_MAP, /* UINT64 */ 88 /* zil: */ 89 DMU_OT_INTENT_LOG, /* UINT64 */ 90 /* dmu: */ 91 DMU_OT_DNODE, /* DNODE */ 92 DMU_OT_OBJSET, /* OBJSET */ 93 /* dsl: */ 94 DMU_OT_DSL_DIR, /* UINT64 */ 95 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ 96 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ 97 DMU_OT_DSL_PROPS, /* ZAP */ 98 DMU_OT_DSL_DATASET, /* UINT64 */ 99 /* zpl: */ 100 DMU_OT_ZNODE, /* ZNODE */ 101 DMU_OT_OLDACL, /* Old ACL */ 102 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ 103 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ 104 DMU_OT_MASTER_NODE, /* ZAP */ 105 DMU_OT_UNLINKED_SET, /* ZAP */ 106 /* zvol: */ 107 DMU_OT_ZVOL, /* UINT8 */ 108 DMU_OT_ZVOL_PROP, /* ZAP */ 109 /* other; for testing only! */ 110 DMU_OT_PLAIN_OTHER, /* UINT8 */ 111 DMU_OT_UINT64_OTHER, /* UINT64 */ 112 DMU_OT_ZAP_OTHER, /* ZAP */ 113 /* new object types: */ 114 DMU_OT_ERROR_LOG, /* ZAP */ 115 DMU_OT_SPA_HISTORY, /* UINT8 */ 116 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ 117 DMU_OT_POOL_PROPS, /* ZAP */ 118 DMU_OT_DSL_PERMS, /* ZAP */ 119 DMU_OT_ACL, /* ACL */ 120 DMU_OT_SYSACL, /* SYSACL */ 121 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ 122 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ 123 DMU_OT_NEXT_CLONES, /* ZAP */ 124 DMU_OT_SCAN_QUEUE, /* ZAP */ 125 DMU_OT_USERGROUP_USED, /* ZAP */ 126 DMU_OT_USERGROUP_QUOTA, /* ZAP */ 127 DMU_OT_USERREFS, /* ZAP */ 128 DMU_OT_DDT_ZAP, /* ZAP */ 129 DMU_OT_DDT_STATS, /* ZAP */ 130 DMU_OT_SA, /* System attr */ 131 DMU_OT_SA_MASTER_NODE, /* ZAP */ 132 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ 133 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ 134 DMU_OT_SCAN_XLATE, /* ZAP */ 135 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ 136 DMU_OT_DEADLIST, /* ZAP */ 137 DMU_OT_DEADLIST_HDR, /* UINT64 */ 138 DMU_OT_DSL_CLONES, /* ZAP */ 139 DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */ 140 DMU_OT_NUMTYPES 141 } dmu_object_type_t; 142 143 typedef enum dmu_objset_type { 144 DMU_OST_NONE, 145 DMU_OST_META, 146 DMU_OST_ZFS, 147 DMU_OST_ZVOL, 148 DMU_OST_OTHER, /* For testing only! */ 149 DMU_OST_ANY, /* Be careful! */ 150 DMU_OST_NUMTYPES 151 } dmu_objset_type_t; 152 153 void byteswap_uint64_array(void *buf, size_t size); 154 void byteswap_uint32_array(void *buf, size_t size); 155 void byteswap_uint16_array(void *buf, size_t size); 156 void byteswap_uint8_array(void *buf, size_t size); 157 void zap_byteswap(void *buf, size_t size); 158 void zfs_oldacl_byteswap(void *buf, size_t size); 159 void zfs_acl_byteswap(void *buf, size_t size); 160 void zfs_znode_byteswap(void *buf, size_t size); 161 162 #define DS_FIND_SNAPSHOTS (1<<0) 163 #define DS_FIND_CHILDREN (1<<1) 164 165 /* 166 * The maximum number of bytes that can be accessed as part of one 167 * operation, including metadata. 168 */ 169 #define DMU_MAX_ACCESS (10<<20) /* 10MB */ 170 #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */ 171 172 #define DMU_USERUSED_OBJECT (-1ULL) 173 #define DMU_GROUPUSED_OBJECT (-2ULL) 174 #define DMU_DEADLIST_OBJECT (-3ULL) 175 176 /* 177 * artificial blkids for bonus buffer and spill blocks 178 */ 179 #define DMU_BONUS_BLKID (-1ULL) 180 #define DMU_SPILL_BLKID (-2ULL) 181 /* 182 * Public routines to create, destroy, open, and close objsets. 183 */ 184 int dmu_objset_hold(const char *name, void *tag, objset_t **osp); 185 int dmu_objset_own(const char *name, dmu_objset_type_t type, 186 boolean_t readonly, void *tag, objset_t **osp); 187 void dmu_objset_rele(objset_t *os, void *tag); 188 void dmu_objset_disown(objset_t *os, void *tag); 189 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp); 190 191 int dmu_objset_evict_dbufs(objset_t *os); 192 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, 193 void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg); 194 int dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin, 195 uint64_t flags); 196 int dmu_objset_destroy(const char *name, boolean_t defer); 197 int dmu_snapshots_destroy(char *fsname, char *snapname, boolean_t defer); 198 int dmu_objset_snapshot(char *fsname, char *snapname, char *tag, 199 struct nvlist *props, boolean_t recursive, boolean_t temporary, int fd); 200 int dmu_objset_rename(const char *name, const char *newname, 201 boolean_t recursive); 202 int dmu_objset_find(char *name, int func(const char *, void *), void *arg, 203 int flags); 204 void dmu_objset_byteswap(void *buf, size_t size); 205 206 typedef struct dmu_buf { 207 uint64_t db_object; /* object that this buffer is part of */ 208 uint64_t db_offset; /* byte offset in this object */ 209 uint64_t db_size; /* size of buffer in bytes */ 210 void *db_data; /* data in buffer */ 211 } dmu_buf_t; 212 213 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr); 214 215 /* 216 * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 217 */ 218 #define DMU_POOL_DIRECTORY_OBJECT 1 219 #define DMU_POOL_CONFIG "config" 220 #define DMU_POOL_ROOT_DATASET "root_dataset" 221 #define DMU_POOL_SYNC_BPOBJ "sync_bplist" 222 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 223 #define DMU_POOL_ERRLOG_LAST "errlog_last" 224 #define DMU_POOL_SPARES "spares" 225 #define DMU_POOL_DEFLATE "deflate" 226 #define DMU_POOL_HISTORY "history" 227 #define DMU_POOL_PROPS "pool_props" 228 #define DMU_POOL_L2CACHE "l2cache" 229 #define DMU_POOL_TMP_USERREFS "tmp_userrefs" 230 #define DMU_POOL_DDT "DDT-%s-%s-%s" 231 #define DMU_POOL_DDT_STATS "DDT-statistics" 232 #define DMU_POOL_CREATION_VERSION "creation_version" 233 #define DMU_POOL_SCAN "scan" 234 #define DMU_POOL_FREE_BPOBJ "free_bpobj" 235 236 /* 237 * Allocate an object from this objset. The range of object numbers 238 * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode. 239 * 240 * The transaction must be assigned to a txg. The newly allocated 241 * object will be "held" in the transaction (ie. you can modify the 242 * newly allocated object in this transaction). 243 * 244 * dmu_object_alloc() chooses an object and returns it in *objectp. 245 * 246 * dmu_object_claim() allocates a specific object number. If that 247 * number is already allocated, it fails and returns EEXIST. 248 * 249 * Return 0 on success, or ENOSPC or EEXIST as specified above. 250 */ 251 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot, 252 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 253 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot, 254 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 255 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot, 256 int blocksize, dmu_object_type_t bonustype, int bonuslen); 257 258 /* 259 * Free an object from this objset. 260 * 261 * The object's data will be freed as well (ie. you don't need to call 262 * dmu_free(object, 0, -1, tx)). 263 * 264 * The object need not be held in the transaction. 265 * 266 * If there are any holds on this object's buffers (via dmu_buf_hold()), 267 * or tx holds on the object (via dmu_tx_hold_object()), you can not 268 * free it; it fails and returns EBUSY. 269 * 270 * If the object is not allocated, it fails and returns ENOENT. 271 * 272 * Return 0 on success, or EBUSY or ENOENT as specified above. 273 */ 274 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx); 275 276 /* 277 * Find the next allocated or free object. 278 * 279 * The objectp parameter is in-out. It will be updated to be the next 280 * object which is allocated. Ignore objects which have not been 281 * modified since txg. 282 * 283 * XXX Can only be called on a objset with no dirty data. 284 * 285 * Returns 0 on success, or ENOENT if there are no more objects. 286 */ 287 int dmu_object_next(objset_t *os, uint64_t *objectp, 288 boolean_t hole, uint64_t txg); 289 290 /* 291 * Set the data blocksize for an object. 292 * 293 * The object cannot have any blocks allcated beyond the first. If 294 * the first block is allocated already, the new size must be greater 295 * than the current block size. If these conditions are not met, 296 * ENOTSUP will be returned. 297 * 298 * Returns 0 on success, or EBUSY if there are any holds on the object 299 * contents, or ENOTSUP as described above. 300 */ 301 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, 302 int ibs, dmu_tx_t *tx); 303 304 /* 305 * Set the checksum property on a dnode. The new checksum algorithm will 306 * apply to all newly written blocks; existing blocks will not be affected. 307 */ 308 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, 309 dmu_tx_t *tx); 310 311 /* 312 * Set the compress property on a dnode. The new compression algorithm will 313 * apply to all newly written blocks; existing blocks will not be affected. 314 */ 315 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, 316 dmu_tx_t *tx); 317 318 /* 319 * Decide how to write a block: checksum, compression, number of copies, etc. 320 */ 321 #define WP_NOFILL 0x1 322 #define WP_DMU_SYNC 0x2 323 #define WP_SPILL 0x4 324 325 void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp, 326 struct zio_prop *zp); 327 /* 328 * The bonus data is accessed more or less like a regular buffer. 329 * You must dmu_bonus_hold() to get the buffer, which will give you a 330 * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus 331 * data. As with any normal buffer, you must call dmu_buf_read() to 332 * read db_data, dmu_buf_will_dirty() before modifying it, and the 333 * object must be held in an assigned transaction before calling 334 * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus 335 * buffer as well. You must release your hold with dmu_buf_rele(). 336 */ 337 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **); 338 int dmu_bonus_max(void); 339 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *); 340 int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *); 341 dmu_object_type_t dmu_get_bonustype(dmu_buf_t *); 342 int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *); 343 344 /* 345 * Special spill buffer support used by "SA" framework 346 */ 347 348 int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); 349 int dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags, 350 void *tag, dmu_buf_t **dbp); 351 int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); 352 353 /* 354 * Obtain the DMU buffer from the specified object which contains the 355 * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so 356 * that it will remain in memory. You must release the hold with 357 * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your 358 * hold. You must have a hold on any dmu_buf_t* you pass to the DMU. 359 * 360 * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill 361 * on the returned buffer before reading or writing the buffer's 362 * db_data. The comments for those routines describe what particular 363 * operations are valid after calling them. 364 * 365 * The object number must be a valid, allocated object number. 366 */ 367 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, 368 void *tag, dmu_buf_t **, int flags); 369 void dmu_buf_add_ref(dmu_buf_t *db, void* tag); 370 void dmu_buf_rele(dmu_buf_t *db, void *tag); 371 uint64_t dmu_buf_refcount(dmu_buf_t *db); 372 373 /* 374 * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a 375 * range of an object. A pointer to an array of dmu_buf_t*'s is 376 * returned (in *dbpp). 377 * 378 * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and 379 * frees the array. The hold on the array of buffers MUST be released 380 * with dmu_buf_rele_array. You can NOT release the hold on each buffer 381 * individually with dmu_buf_rele. 382 */ 383 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset, 384 uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp); 385 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag); 386 387 /* 388 * Returns NULL on success, or the existing user ptr if it's already 389 * been set. 390 * 391 * user_ptr is for use by the user and can be obtained via dmu_buf_get_user(). 392 * 393 * user_data_ptr_ptr should be NULL, or a pointer to a pointer which 394 * will be set to db->db_data when you are allowed to access it. Note 395 * that db->db_data (the pointer) can change when you do dmu_buf_read(), 396 * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill(). 397 * *user_data_ptr_ptr will be set to the new value when it changes. 398 * 399 * If non-NULL, pageout func will be called when this buffer is being 400 * excised from the cache, so that you can clean up the data structure 401 * pointed to by user_ptr. 402 * 403 * dmu_evict_user() will call the pageout func for all buffers in a 404 * objset with a given pageout func. 405 */ 406 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr, 407 dmu_buf_evict_func_t *pageout_func); 408 /* 409 * set_user_ie is the same as set_user, but request immediate eviction 410 * when hold count goes to zero. 411 */ 412 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr, 413 void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func); 414 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr, 415 void *user_ptr, void *user_data_ptr_ptr, 416 dmu_buf_evict_func_t *pageout_func); 417 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func); 418 419 /* 420 * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set. 421 */ 422 void *dmu_buf_get_user(dmu_buf_t *db); 423 424 /* 425 * Indicate that you are going to modify the buffer's data (db_data). 426 * 427 * The transaction (tx) must be assigned to a txg (ie. you've called 428 * dmu_tx_assign()). The buffer's object must be held in the tx 429 * (ie. you've called dmu_tx_hold_object(tx, db->db_object)). 430 */ 431 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx); 432 433 /* 434 * Tells if the given dbuf is freeable. 435 */ 436 boolean_t dmu_buf_freeable(dmu_buf_t *); 437 438 /* 439 * You must create a transaction, then hold the objects which you will 440 * (or might) modify as part of this transaction. Then you must assign 441 * the transaction to a transaction group. Once the transaction has 442 * been assigned, you can modify buffers which belong to held objects as 443 * part of this transaction. You can't modify buffers before the 444 * transaction has been assigned; you can't modify buffers which don't 445 * belong to objects which this transaction holds; you can't hold 446 * objects once the transaction has been assigned. You may hold an 447 * object which you are going to free (with dmu_object_free()), but you 448 * don't have to. 449 * 450 * You can abort the transaction before it has been assigned. 451 * 452 * Note that you may hold buffers (with dmu_buf_hold) at any time, 453 * regardless of transaction state. 454 */ 455 456 #define DMU_NEW_OBJECT (-1ULL) 457 #define DMU_OBJECT_END (-1ULL) 458 459 dmu_tx_t *dmu_tx_create(objset_t *os); 460 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len); 461 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, 462 uint64_t len); 463 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name); 464 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object); 465 void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object); 466 void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow); 467 void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size); 468 void dmu_tx_abort(dmu_tx_t *tx); 469 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how); 470 void dmu_tx_wait(dmu_tx_t *tx); 471 void dmu_tx_commit(dmu_tx_t *tx); 472 473 /* 474 * To register a commit callback, dmu_tx_callback_register() must be called. 475 * 476 * dcb_data is a pointer to caller private data that is passed on as a 477 * callback parameter. The caller is responsible for properly allocating and 478 * freeing it. 479 * 480 * When registering a callback, the transaction must be already created, but 481 * it cannot be committed or aborted. It can be assigned to a txg or not. 482 * 483 * The callback will be called after the transaction has been safely written 484 * to stable storage and will also be called if the dmu_tx is aborted. 485 * If there is any error which prevents the transaction from being committed to 486 * disk, the callback will be called with a value of error != 0. 487 */ 488 typedef void dmu_tx_callback_func_t(void *dcb_data, int error); 489 490 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func, 491 void *dcb_data); 492 493 /* 494 * Free up the data blocks for a defined range of a file. If size is 495 * zero, the range from offset to end-of-file is freed. 496 */ 497 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, 498 uint64_t size, dmu_tx_t *tx); 499 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, 500 uint64_t size); 501 int dmu_free_object(objset_t *os, uint64_t object); 502 503 /* 504 * Convenience functions. 505 * 506 * Canfail routines will return 0 on success, or an errno if there is a 507 * nonrecoverable I/O error. 508 */ 509 #define DMU_READ_PREFETCH 0 /* prefetch */ 510 #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */ 511 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 512 void *buf, uint32_t flags); 513 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 514 const void *buf, dmu_tx_t *tx); 515 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 516 dmu_tx_t *tx); 517 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size); 518 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size, 519 dmu_tx_t *tx); 520 int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size, 521 dmu_tx_t *tx); 522 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, 523 uint64_t size, struct page *pp, dmu_tx_t *tx); 524 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size); 525 void dmu_return_arcbuf(struct arc_buf *buf); 526 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf, 527 dmu_tx_t *tx); 528 int dmu_xuio_init(struct xuio *uio, int niov); 529 void dmu_xuio_fini(struct xuio *uio); 530 int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off, 531 size_t n); 532 int dmu_xuio_cnt(struct xuio *uio); 533 struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i); 534 void dmu_xuio_clear(struct xuio *uio, int i); 535 void xuio_stat_wbuf_copied(); 536 void xuio_stat_wbuf_nocopy(); 537 538 extern int zfs_prefetch_disable; 539 540 /* 541 * Asynchronously try to read in the data. 542 */ 543 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset, 544 uint64_t len); 545 546 typedef struct dmu_object_info { 547 /* All sizes are in bytes unless otherwise indicated. */ 548 uint32_t doi_data_block_size; 549 uint32_t doi_metadata_block_size; 550 dmu_object_type_t doi_type; 551 dmu_object_type_t doi_bonus_type; 552 uint64_t doi_bonus_size; 553 uint8_t doi_indirection; /* 2 = dnode->indirect->data */ 554 uint8_t doi_checksum; 555 uint8_t doi_compress; 556 uint8_t doi_pad[5]; 557 uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */ 558 uint64_t doi_max_offset; 559 uint64_t doi_fill_count; /* number of non-empty blocks */ 560 } dmu_object_info_t; 561 562 typedef void arc_byteswap_func_t(void *buf, size_t size); 563 564 typedef struct dmu_object_type_info { 565 arc_byteswap_func_t *ot_byteswap; 566 boolean_t ot_metadata; 567 char *ot_name; 568 } dmu_object_type_info_t; 569 570 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES]; 571 572 /* 573 * Get information on a DMU object. 574 * 575 * Return 0 on success or ENOENT if object is not allocated. 576 * 577 * If doi is NULL, just indicates whether the object exists. 578 */ 579 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi); 580 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi); 581 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi); 582 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, 583 u_longlong_t *nblk512); 584 585 typedef struct dmu_objset_stats { 586 uint64_t dds_num_clones; /* number of clones of this */ 587 uint64_t dds_creation_txg; 588 uint64_t dds_guid; 589 dmu_objset_type_t dds_type; 590 uint8_t dds_is_snapshot; 591 uint8_t dds_inconsistent; 592 char dds_origin[MAXNAMELEN]; 593 } dmu_objset_stats_t; 594 595 /* 596 * Get stats on a dataset. 597 */ 598 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat); 599 600 /* 601 * Add entries to the nvlist for all the objset's properties. See 602 * zfs_prop_table[] and zfs(1m) for details on the properties. 603 */ 604 void dmu_objset_stats(objset_t *os, struct nvlist *nv); 605 606 /* 607 * Get the space usage statistics for statvfs(). 608 * 609 * refdbytes is the amount of space "referenced" by this objset. 610 * availbytes is the amount of space available to this objset, taking 611 * into account quotas & reservations, assuming that no other objsets 612 * use the space first. These values correspond to the 'referenced' and 613 * 'available' properties, described in the zfs(1m) manpage. 614 * 615 * usedobjs and availobjs are the number of objects currently allocated, 616 * and available. 617 */ 618 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, 619 uint64_t *usedobjsp, uint64_t *availobjsp); 620 621 /* 622 * The fsid_guid is a 56-bit ID that can change to avoid collisions. 623 * (Contrast with the ds_guid which is a 64-bit ID that will never 624 * change, so there is a small probability that it will collide.) 625 */ 626 uint64_t dmu_objset_fsid_guid(objset_t *os); 627 628 /* 629 * Get the [cm]time for an objset's snapshot dir 630 */ 631 timestruc_t dmu_objset_snap_cmtime(objset_t *os); 632 633 int dmu_objset_is_snapshot(objset_t *os); 634 635 extern struct spa *dmu_objset_spa(objset_t *os); 636 extern struct zilog *dmu_objset_zil(objset_t *os); 637 extern struct dsl_pool *dmu_objset_pool(objset_t *os); 638 extern struct dsl_dataset *dmu_objset_ds(objset_t *os); 639 extern void dmu_objset_name(objset_t *os, char *buf); 640 extern dmu_objset_type_t dmu_objset_type(objset_t *os); 641 extern uint64_t dmu_objset_id(objset_t *os); 642 extern uint64_t dmu_objset_syncprop(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 struct avl_tree *drc_guid_to_ds_map; 725 } dmu_recv_cookie_t; 726 727 int dmu_recv_begin(char *tofs, char *tosnap, char *topds, struct drr_begin *, 728 boolean_t force, objset_t *origin, dmu_recv_cookie_t *); 729 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp, 730 int cleanup_fd, uint64_t *action_handlep); 731 int dmu_recv_end(dmu_recv_cookie_t *drc); 732 733 int dmu_diff(objset_t *tosnap, objset_t *fromsnap, struct vnode *vp, 734 offset_t *off); 735 736 /* CRC64 table */ 737 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ 738 extern uint64_t zfs_crc64_table[256]; 739 740 #ifdef __cplusplus 741 } 742 #endif 743 744 #endif /* _SYS_DMU_H */ 745