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 /* 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 * Copyright (c) 2012, 2018 by Delphix. All rights reserved. 25 * Copyright 2017 Nexenta Systems, Inc. 26 */ 27 28 #ifndef _SYS_ZAP_H 29 #define _SYS_ZAP_H 30 31 /* 32 * ZAP - ZFS Attribute Processor 33 * 34 * The ZAP is a module which sits on top of the DMU (Data Management 35 * Unit) and implements a higher-level storage primitive using DMU 36 * objects. Its primary consumer is the ZPL (ZFS Posix Layer). 37 * 38 * A "zapobj" is a DMU object which the ZAP uses to stores attributes. 39 * Users should use only zap routines to access a zapobj - they should 40 * not access the DMU object directly using DMU routines. 41 * 42 * The attributes stored in a zapobj are name-value pairs. The name is 43 * a zero-terminated string of up to ZAP_MAXNAMELEN bytes (including 44 * terminating NULL). The value is an array of integers, which may be 45 * 1, 2, 4, or 8 bytes long. The total space used by the array (number 46 * of integers * integer length) can be up to ZAP_MAXVALUELEN bytes. 47 * Note that an 8-byte integer value can be used to store the location 48 * (object number) of another dmu object (which may be itself a zapobj). 49 * Note that you can use a zero-length attribute to store a single bit 50 * of information - the attribute is present or not. 51 * 52 * The ZAP routines are thread-safe. However, you must observe the 53 * DMU's restriction that a transaction may not be operated on 54 * concurrently. 55 * 56 * Any of the routines that return an int may return an I/O error (EIO 57 * or ECHECKSUM). 58 * 59 * 60 * Implementation / Performance Notes: 61 * 62 * The ZAP is intended to operate most efficiently on attributes with 63 * short (49 bytes or less) names and single 8-byte values, for which 64 * the microzap will be used. The ZAP should be efficient enough so 65 * that the user does not need to cache these attributes. 66 * 67 * The ZAP's locking scheme makes its routines thread-safe. Operations 68 * on different zapobjs will be processed concurrently. Operations on 69 * the same zapobj which only read data will be processed concurrently. 70 * Operations on the same zapobj which modify data will be processed 71 * concurrently when there are many attributes in the zapobj (because 72 * the ZAP uses per-block locking - more than 128 * (number of cpus) 73 * small attributes will suffice). 74 */ 75 76 /* 77 * We're using zero-terminated byte strings (ie. ASCII or UTF-8 C 78 * strings) for the names of attributes, rather than a byte string 79 * bounded by an explicit length. If some day we want to support names 80 * in character sets which have embedded zeros (eg. UTF-16, UTF-32), 81 * we'll have to add routines for using length-bounded strings. 82 */ 83 84 #include <sys/dmu.h> 85 86 #ifdef __cplusplus 87 extern "C" { 88 #endif 89 90 /* 91 * Specifies matching criteria for ZAP lookups. 92 * MT_NORMALIZE Use ZAP normalization flags, which can include both 93 * unicode normalization and case-insensitivity. 94 * MT_MATCH_CASE Do case-sensitive lookups even if MT_NORMALIZE is 95 * specified and ZAP normalization flags include 96 * U8_TEXTPREP_TOUPPER. 97 */ 98 typedef enum matchtype { 99 MT_NORMALIZE = 1 << 0, 100 MT_MATCH_CASE = 1 << 1, 101 } matchtype_t; 102 103 typedef enum zap_flags { 104 /* Use 64-bit hash value (serialized cursors will always use 64-bits) */ 105 ZAP_FLAG_HASH64 = 1 << 0, 106 /* Key is binary, not string (zap_add_uint64() can be used) */ 107 ZAP_FLAG_UINT64_KEY = 1 << 1, 108 /* 109 * First word of key (which must be an array of uint64) is 110 * already randomly distributed. 111 */ 112 ZAP_FLAG_PRE_HASHED_KEY = 1 << 2, 113 } zap_flags_t; 114 115 /* 116 * Create a new zapobj with no attributes and return its object number. 117 */ 118 uint64_t zap_create(objset_t *ds, dmu_object_type_t ot, 119 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx); 120 uint64_t zap_create_dnsize(objset_t *ds, dmu_object_type_t ot, 121 dmu_object_type_t bonustype, int bonuslen, int dnodesize, dmu_tx_t *tx); 122 uint64_t zap_create_norm(objset_t *ds, int normflags, dmu_object_type_t ot, 123 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx); 124 uint64_t zap_create_norm_dnsize(objset_t *ds, int normflags, 125 dmu_object_type_t ot, dmu_object_type_t bonustype, int bonuslen, 126 int dnodesize, dmu_tx_t *tx); 127 uint64_t zap_create_flags(objset_t *os, int normflags, zap_flags_t flags, 128 dmu_object_type_t ot, int leaf_blockshift, int indirect_blockshift, 129 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx); 130 uint64_t zap_create_flags_dnsize(objset_t *os, int normflags, 131 zap_flags_t flags, dmu_object_type_t ot, int leaf_blockshift, 132 int indirect_blockshift, dmu_object_type_t bonustype, int bonuslen, 133 int dnodesize, dmu_tx_t *tx); 134 uint64_t zap_create_hold(objset_t *os, int normflags, zap_flags_t flags, 135 dmu_object_type_t ot, int leaf_blockshift, int indirect_blockshift, 136 dmu_object_type_t bonustype, int bonuslen, int dnodesize, 137 dnode_t **allocated_dnode, void *tag, dmu_tx_t *tx); 138 139 uint64_t zap_create_link(objset_t *os, dmu_object_type_t ot, 140 uint64_t parent_obj, const char *name, dmu_tx_t *tx); 141 uint64_t zap_create_link_dnsize(objset_t *os, dmu_object_type_t ot, 142 uint64_t parent_obj, const char *name, int dnodesize, dmu_tx_t *tx); 143 144 /* 145 * Initialize an already-allocated object. 146 */ 147 void mzap_create_impl(dnode_t *dn, int normflags, zap_flags_t flags, 148 dmu_tx_t *tx); 149 150 /* 151 * Create a new zapobj with no attributes from the given (unallocated) 152 * object number. 153 */ 154 int zap_create_claim(objset_t *ds, uint64_t obj, dmu_object_type_t ot, 155 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx); 156 int zap_create_claim_dnsize(objset_t *ds, uint64_t obj, dmu_object_type_t ot, 157 dmu_object_type_t bonustype, int bonuslen, int dnodesize, dmu_tx_t *tx); 158 int zap_create_claim_norm(objset_t *ds, uint64_t obj, 159 int normflags, dmu_object_type_t ot, 160 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx); 161 int zap_create_claim_norm_dnsize(objset_t *ds, uint64_t obj, 162 int normflags, dmu_object_type_t ot, 163 dmu_object_type_t bonustype, int bonuslen, int dnodesize, dmu_tx_t *tx); 164 165 /* 166 * The zapobj passed in must be a valid ZAP object for all of the 167 * following routines. 168 */ 169 170 /* 171 * Destroy this zapobj and all its attributes. 172 * 173 * Frees the object number using dmu_object_free. 174 */ 175 int zap_destroy(objset_t *ds, uint64_t zapobj, dmu_tx_t *tx); 176 177 /* 178 * Manipulate attributes. 179 * 180 * 'integer_size' is in bytes, and must be 1, 2, 4, or 8. 181 */ 182 183 /* 184 * Retrieve the contents of the attribute with the given name. 185 * 186 * If the requested attribute does not exist, the call will fail and 187 * return ENOENT. 188 * 189 * If 'integer_size' is smaller than the attribute's integer size, the 190 * call will fail and return EINVAL. 191 * 192 * If 'integer_size' is equal to or larger than the attribute's integer 193 * size, the call will succeed and return 0. 194 * 195 * When converting to a larger integer size, the integers will be treated as 196 * unsigned (ie. no sign-extension will be performed). 197 * 198 * 'num_integers' is the length (in integers) of 'buf'. 199 * 200 * If the attribute is longer than the buffer, as many integers as will 201 * fit will be transferred to 'buf'. If the entire attribute was not 202 * transferred, the call will return EOVERFLOW. 203 */ 204 int zap_lookup(objset_t *ds, uint64_t zapobj, const char *name, 205 uint64_t integer_size, uint64_t num_integers, void *buf); 206 207 /* 208 * If rn_len is nonzero, realname will be set to the name of the found 209 * entry (which may be different from the requested name if matchtype is 210 * not MT_EXACT). 211 * 212 * If normalization_conflictp is not NULL, it will be set if there is 213 * another name with the same case/unicode normalized form. 214 */ 215 int zap_lookup_norm(objset_t *ds, uint64_t zapobj, const char *name, 216 uint64_t integer_size, uint64_t num_integers, void *buf, 217 matchtype_t mt, char *realname, int rn_len, 218 boolean_t *normalization_conflictp); 219 int zap_lookup_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key, 220 int key_numints, uint64_t integer_size, uint64_t num_integers, void *buf); 221 int zap_contains(objset_t *ds, uint64_t zapobj, const char *name); 222 int zap_prefetch(objset_t *os, uint64_t zapobj, const char *name); 223 int zap_prefetch_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key, 224 int key_numints); 225 226 int zap_lookup_by_dnode(dnode_t *dn, const char *name, 227 uint64_t integer_size, uint64_t num_integers, void *buf); 228 int zap_lookup_norm_by_dnode(dnode_t *dn, const char *name, 229 uint64_t integer_size, uint64_t num_integers, void *buf, 230 matchtype_t mt, char *realname, int rn_len, 231 boolean_t *ncp); 232 233 int zap_count_write_by_dnode(dnode_t *dn, const char *name, 234 int add, zfs_refcount_t *towrite, zfs_refcount_t *tooverwrite); 235 236 /* 237 * Create an attribute with the given name and value. 238 * 239 * If an attribute with the given name already exists, the call will 240 * fail and return EEXIST. 241 */ 242 int zap_add(objset_t *ds, uint64_t zapobj, const char *key, 243 int integer_size, uint64_t num_integers, 244 const void *val, dmu_tx_t *tx); 245 int zap_add_by_dnode(dnode_t *dn, const char *key, 246 int integer_size, uint64_t num_integers, 247 const void *val, dmu_tx_t *tx); 248 int zap_add_uint64(objset_t *ds, uint64_t zapobj, const uint64_t *key, 249 int key_numints, int integer_size, uint64_t num_integers, 250 const void *val, dmu_tx_t *tx); 251 252 /* 253 * Set the attribute with the given name to the given value. If an 254 * attribute with the given name does not exist, it will be created. If 255 * an attribute with the given name already exists, the previous value 256 * will be overwritten. The integer_size may be different from the 257 * existing attribute's integer size, in which case the attribute's 258 * integer size will be updated to the new value. 259 */ 260 int zap_update(objset_t *ds, uint64_t zapobj, const char *name, 261 int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx); 262 int zap_update_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key, 263 int key_numints, 264 int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx); 265 266 /* 267 * Get the length (in integers) and the integer size of the specified 268 * attribute. 269 * 270 * If the requested attribute does not exist, the call will fail and 271 * return ENOENT. 272 */ 273 int zap_length(objset_t *ds, uint64_t zapobj, const char *name, 274 uint64_t *integer_size, uint64_t *num_integers); 275 int zap_length_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key, 276 int key_numints, uint64_t *integer_size, uint64_t *num_integers); 277 278 /* 279 * Remove the specified attribute. 280 * 281 * If the specified attribute does not exist, the call will fail and 282 * return ENOENT. 283 */ 284 int zap_remove(objset_t *ds, uint64_t zapobj, const char *name, dmu_tx_t *tx); 285 int zap_remove_norm(objset_t *ds, uint64_t zapobj, const char *name, 286 matchtype_t mt, dmu_tx_t *tx); 287 int zap_remove_by_dnode(dnode_t *dn, const char *name, dmu_tx_t *tx); 288 int zap_remove_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key, 289 int key_numints, dmu_tx_t *tx); 290 291 /* 292 * Returns (in *count) the number of attributes in the specified zap 293 * object. 294 */ 295 int zap_count(objset_t *ds, uint64_t zapobj, uint64_t *count); 296 297 /* 298 * Returns (in name) the name of the entry whose (value & mask) 299 * (za_first_integer) is value, or ENOENT if not found. The string 300 * pointed to by name must be at least 256 bytes long. If mask==0, the 301 * match must be exact (ie, same as mask=-1ULL). 302 */ 303 int zap_value_search(objset_t *os, uint64_t zapobj, 304 uint64_t value, uint64_t mask, char *name); 305 306 /* 307 * Transfer all the entries from fromobj into intoobj. Only works on 308 * int_size=8 num_integers=1 values. Fails if there are any duplicated 309 * entries. 310 */ 311 int zap_join(objset_t *os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t *tx); 312 313 /* Same as zap_join, but set the values to 'value'. */ 314 int zap_join_key(objset_t *os, uint64_t fromobj, uint64_t intoobj, 315 uint64_t value, dmu_tx_t *tx); 316 317 /* Same as zap_join, but add together any duplicated entries. */ 318 int zap_join_increment(objset_t *os, uint64_t fromobj, uint64_t intoobj, 319 dmu_tx_t *tx); 320 321 /* 322 * Manipulate entries where the name + value are the "same" (the name is 323 * a stringified version of the value). 324 */ 325 int zap_add_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx); 326 int zap_remove_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx); 327 int zap_lookup_int(objset_t *os, uint64_t obj, uint64_t value); 328 int zap_increment_int(objset_t *os, uint64_t obj, uint64_t key, int64_t delta, 329 dmu_tx_t *tx); 330 331 /* Here the key is an int and the value is a different int. */ 332 int zap_add_int_key(objset_t *os, uint64_t obj, 333 uint64_t key, uint64_t value, dmu_tx_t *tx); 334 int zap_update_int_key(objset_t *os, uint64_t obj, 335 uint64_t key, uint64_t value, dmu_tx_t *tx); 336 int zap_lookup_int_key(objset_t *os, uint64_t obj, 337 uint64_t key, uint64_t *valuep); 338 339 int zap_increment(objset_t *os, uint64_t obj, const char *name, int64_t delta, 340 dmu_tx_t *tx); 341 342 struct zap; 343 struct zap_leaf; 344 typedef struct zap_cursor { 345 /* This structure is opaque! */ 346 objset_t *zc_objset; 347 struct zap *zc_zap; 348 struct zap_leaf *zc_leaf; 349 uint64_t zc_zapobj; 350 uint64_t zc_serialized; 351 uint64_t zc_hash; 352 uint32_t zc_cd; 353 boolean_t zc_prefetch; 354 } zap_cursor_t; 355 356 typedef struct { 357 int za_integer_length; 358 /* 359 * za_normalization_conflict will be set if there are additional 360 * entries with this normalized form (eg, "foo" and "Foo"). 361 */ 362 boolean_t za_normalization_conflict; 363 uint64_t za_num_integers; 364 uint64_t za_first_integer; /* no sign extension for <8byte ints */ 365 char za_name[ZAP_MAXNAMELEN]; 366 } zap_attribute_t; 367 368 /* 369 * The interface for listing all the attributes of a zapobj can be 370 * thought of as cursor moving down a list of the attributes one by 371 * one. The cookie returned by the zap_cursor_serialize routine is 372 * persistent across system calls (and across reboot, even). 373 */ 374 375 /* 376 * Initialize a zap cursor, pointing to the "first" attribute of the 377 * zapobj. You must _fini the cursor when you are done with it. 378 */ 379 void zap_cursor_init(zap_cursor_t *zc, objset_t *os, uint64_t zapobj); 380 void zap_cursor_init_noprefetch(zap_cursor_t *zc, objset_t *os, 381 uint64_t zapobj); 382 void zap_cursor_fini(zap_cursor_t *zc); 383 384 /* 385 * Get the attribute currently pointed to by the cursor. Returns 386 * ENOENT if at the end of the attributes. 387 */ 388 int zap_cursor_retrieve(zap_cursor_t *zc, zap_attribute_t *za); 389 390 /* 391 * Advance the cursor to the next attribute. 392 */ 393 void zap_cursor_advance(zap_cursor_t *zc); 394 395 /* 396 * Get a persistent cookie pointing to the current position of the zap 397 * cursor. The low 4 bits in the cookie are always zero, and thus can 398 * be used as to differentiate a serialized cookie from a different type 399 * of value. The cookie will be less than 2^32 as long as there are 400 * fewer than 2^22 (4.2 million) entries in the zap object. 401 */ 402 uint64_t zap_cursor_serialize(zap_cursor_t *zc); 403 404 /* 405 * Initialize a zap cursor pointing to the position recorded by 406 * zap_cursor_serialize (in the "serialized" argument). You can also 407 * use a "serialized" argument of 0 to start at the beginning of the 408 * zapobj (ie. zap_cursor_init_serialized(..., 0) is equivalent to 409 * zap_cursor_init(...).) 410 */ 411 void zap_cursor_init_serialized(zap_cursor_t *zc, objset_t *ds, 412 uint64_t zapobj, uint64_t serialized); 413 414 415 #define ZAP_HISTOGRAM_SIZE 10 416 417 typedef struct zap_stats { 418 /* 419 * Size of the pointer table (in number of entries). 420 * This is always a power of 2, or zero if it's a microzap. 421 * In general, it should be considerably greater than zs_num_leafs. 422 */ 423 uint64_t zs_ptrtbl_len; 424 425 uint64_t zs_blocksize; /* size of zap blocks */ 426 427 /* 428 * The number of blocks used. Note that some blocks may be 429 * wasted because old ptrtbl's and large name/value blocks are 430 * not reused. (Although their space is reclaimed, we don't 431 * reuse those offsets in the object.) 432 */ 433 uint64_t zs_num_blocks; 434 435 /* 436 * Pointer table values from zap_ptrtbl in the zap_phys_t 437 */ 438 uint64_t zs_ptrtbl_nextblk; /* next (larger) copy start block */ 439 uint64_t zs_ptrtbl_blks_copied; /* number source blocks copied */ 440 uint64_t zs_ptrtbl_zt_blk; /* starting block number */ 441 uint64_t zs_ptrtbl_zt_numblks; /* number of blocks */ 442 uint64_t zs_ptrtbl_zt_shift; /* bits to index it */ 443 444 /* 445 * Values of the other members of the zap_phys_t 446 */ 447 uint64_t zs_block_type; /* ZBT_HEADER */ 448 uint64_t zs_magic; /* ZAP_MAGIC */ 449 uint64_t zs_num_leafs; /* The number of leaf blocks */ 450 uint64_t zs_num_entries; /* The number of zap entries */ 451 uint64_t zs_salt; /* salt to stir into hash function */ 452 453 /* 454 * Histograms. For all histograms, the last index 455 * (ZAP_HISTOGRAM_SIZE-1) includes any values which are greater 456 * than what can be represented. For example 457 * zs_leafs_with_n5_entries[ZAP_HISTOGRAM_SIZE-1] is the number 458 * of leafs with more than 45 entries. 459 */ 460 461 /* 462 * zs_leafs_with_n_pointers[n] is the number of leafs with 463 * 2^n pointers to it. 464 */ 465 uint64_t zs_leafs_with_2n_pointers[ZAP_HISTOGRAM_SIZE]; 466 467 /* 468 * zs_leafs_with_n_entries[n] is the number of leafs with 469 * [n*5, (n+1)*5) entries. In the current implementation, there 470 * can be at most 55 entries in any block, but there may be 471 * fewer if the name or value is large, or the block is not 472 * completely full. 473 */ 474 uint64_t zs_blocks_with_n5_entries[ZAP_HISTOGRAM_SIZE]; 475 476 /* 477 * zs_leafs_n_tenths_full[n] is the number of leafs whose 478 * fullness is in the range [n/10, (n+1)/10). 479 */ 480 uint64_t zs_blocks_n_tenths_full[ZAP_HISTOGRAM_SIZE]; 481 482 /* 483 * zs_entries_using_n_chunks[n] is the number of entries which 484 * consume n 24-byte chunks. (Note, large names/values only use 485 * one chunk, but contribute to zs_num_blocks_large.) 486 */ 487 uint64_t zs_entries_using_n_chunks[ZAP_HISTOGRAM_SIZE]; 488 489 /* 490 * zs_buckets_with_n_entries[n] is the number of buckets (each 491 * leaf has 64 buckets) with n entries. 492 * zs_buckets_with_n_entries[1] should be very close to 493 * zs_num_entries. 494 */ 495 uint64_t zs_buckets_with_n_entries[ZAP_HISTOGRAM_SIZE]; 496 } zap_stats_t; 497 498 /* 499 * Get statistics about a ZAP object. Note: you need to be aware of the 500 * internal implementation of the ZAP to correctly interpret some of the 501 * statistics. This interface shouldn't be relied on unless you really 502 * know what you're doing. 503 */ 504 int zap_get_stats(objset_t *ds, uint64_t zapobj, zap_stats_t *zs); 505 506 #ifdef __cplusplus 507 } 508 #endif 509 510 #endif /* _SYS_ZAP_H */ 511