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 * Copyright (c) 2011, 2019 by Delphix. All rights reserved. 24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 25 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. 26 * Copyright 2013 Saso Kiselkov. All rights reserved. 27 * Copyright (c) 2014 Integros [integros.com] 28 * Copyright 2017 Joyent, Inc. 29 * Copyright (c) 2017 Datto Inc. 30 * Copyright (c) 2017, Intel Corporation. 31 */ 32 33 #ifndef _SYS_SPA_H 34 #define _SYS_SPA_H 35 36 #include <sys/avl.h> 37 #include <sys/zfs_context.h> 38 #include <sys/nvpair.h> 39 #include <sys/sysevent.h> 40 #include <sys/sysmacros.h> 41 #include <sys/types.h> 42 #include <sys/fs/zfs.h> 43 #include <sys/dmu.h> 44 #include <sys/space_map.h> 45 46 #ifdef __cplusplus 47 extern "C" { 48 #endif 49 50 /* 51 * Forward references that lots of things need. 52 */ 53 typedef struct spa spa_t; 54 typedef struct vdev vdev_t; 55 typedef struct metaslab metaslab_t; 56 typedef struct metaslab_group metaslab_group_t; 57 typedef struct metaslab_class metaslab_class_t; 58 typedef struct zio zio_t; 59 typedef struct zilog zilog_t; 60 typedef struct spa_aux_vdev spa_aux_vdev_t; 61 typedef struct ddt ddt_t; 62 typedef struct ddt_entry ddt_entry_t; 63 struct dsl_pool; 64 struct dsl_dataset; 65 struct dsl_crypto_params; 66 67 /* 68 * General-purpose 32-bit and 64-bit bitfield encodings. 69 */ 70 #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) 71 #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) 72 #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) 73 #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) 74 75 #define BF32_GET(x, low, len) BF32_DECODE(x, low, len) 76 #define BF64_GET(x, low, len) BF64_DECODE(x, low, len) 77 78 #define BF32_SET(x, low, len, val) do { \ 79 ASSERT3U(val, <, 1U << (len)); \ 80 ASSERT3U(low + len, <=, 32); \ 81 (x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \ 82 _NOTE(CONSTCOND) } while (0) 83 84 #define BF64_SET(x, low, len, val) do { \ 85 ASSERT3U(val, <, 1ULL << (len)); \ 86 ASSERT3U(low + len, <=, 64); \ 87 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \ 88 _NOTE(CONSTCOND) } while (0) 89 90 #define BF32_GET_SB(x, low, len, shift, bias) \ 91 ((BF32_GET(x, low, len) + (bias)) << (shift)) 92 #define BF64_GET_SB(x, low, len, shift, bias) \ 93 ((BF64_GET(x, low, len) + (bias)) << (shift)) 94 95 #define BF32_SET_SB(x, low, len, shift, bias, val) do { \ 96 ASSERT(IS_P2ALIGNED(val, 1U << shift)); \ 97 ASSERT3S((val) >> (shift), >=, bias); \ 98 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \ 99 _NOTE(CONSTCOND) } while (0) 100 #define BF64_SET_SB(x, low, len, shift, bias, val) do { \ 101 ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \ 102 ASSERT3S((val) >> (shift), >=, bias); \ 103 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \ 104 _NOTE(CONSTCOND) } while (0) 105 106 /* 107 * We currently support block sizes from 512 bytes to 16MB. 108 * The benefits of larger blocks, and thus larger IO, need to be weighed 109 * against the cost of COWing a giant block to modify one byte, and the 110 * large latency of reading or writing a large block. 111 * 112 * Note that although blocks up to 16MB are supported, the recordsize 113 * property can not be set larger than zfs_max_recordsize (default 1MB). 114 * See the comment near zfs_max_recordsize in dsl_dataset.c for details. 115 * 116 * Note that although the LSIZE field of the blkptr_t can store sizes up 117 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to 118 * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB. 119 */ 120 #define SPA_MINBLOCKSHIFT 9 121 #define SPA_OLD_MAXBLOCKSHIFT 17 122 #define SPA_MAXBLOCKSHIFT 24 123 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 124 #define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT) 125 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 126 127 /* 128 * Alignment Shift (ashift) is an immutable, internal top-level vdev property 129 * which can only be set at vdev creation time. Physical writes are always done 130 * according to it, which makes 2^ashift the smallest possible IO on a vdev. 131 * 132 * We currently allow values ranging from 512 bytes (2^9 = 512) to 64 KiB 133 * (2^16 = 65,536). 134 */ 135 #define ASHIFT_MIN 9 136 #define ASHIFT_MAX 16 137 138 /* 139 * Size of block to hold the configuration data (a packed nvlist) 140 */ 141 #define SPA_CONFIG_BLOCKSIZE (1ULL << 14) 142 143 /* 144 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. 145 * The ASIZE encoding should be at least 64 times larger (6 more bits) 146 * to support up to 4-way RAID-Z mirror mode with worst-case gang block 147 * overhead, three DVAs per bp, plus one more bit in case we do anything 148 * else that expands the ASIZE. 149 */ 150 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ 151 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ 152 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ 153 154 #define SPA_COMPRESSBITS 7 155 #define SPA_VDEVBITS 24 156 157 /* 158 * All SPA data is represented by 128-bit data virtual addresses (DVAs). 159 * The members of the dva_t should be considered opaque outside the SPA. 160 */ 161 typedef struct dva { 162 uint64_t dva_word[2]; 163 } dva_t; 164 165 /* 166 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. 167 */ 168 typedef struct zio_cksum { 169 uint64_t zc_word[4]; 170 } zio_cksum_t; 171 172 /* 173 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept 174 * secret and is suitable for use in MAC algorithms as the key. 175 */ 176 typedef struct zio_cksum_salt { 177 uint8_t zcs_bytes[32]; 178 } zio_cksum_salt_t; 179 180 /* 181 * Each block is described by its DVAs, time of birth, checksum, etc. 182 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 183 * 184 * 64 56 48 40 32 24 16 8 0 185 * +-------+-------+-------+-------+-------+-------+-------+-------+ 186 * 0 | pad | vdev1 | GRID | ASIZE | 187 * +-------+-------+-------+-------+-------+-------+-------+-------+ 188 * 1 |G| offset1 | 189 * +-------+-------+-------+-------+-------+-------+-------+-------+ 190 * 2 | pad | vdev2 | GRID | ASIZE | 191 * +-------+-------+-------+-------+-------+-------+-------+-------+ 192 * 3 |G| offset2 | 193 * +-------+-------+-------+-------+-------+-------+-------+-------+ 194 * 4 | pad | vdev3 | GRID | ASIZE | 195 * +-------+-------+-------+-------+-------+-------+-------+-------+ 196 * 5 |G| offset3 | 197 * +-------+-------+-------+-------+-------+-------+-------+-------+ 198 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 199 * +-------+-------+-------+-------+-------+-------+-------+-------+ 200 * 7 | padding | 201 * +-------+-------+-------+-------+-------+-------+-------+-------+ 202 * 8 | padding | 203 * +-------+-------+-------+-------+-------+-------+-------+-------+ 204 * 9 | physical birth txg | 205 * +-------+-------+-------+-------+-------+-------+-------+-------+ 206 * a | logical birth txg | 207 * +-------+-------+-------+-------+-------+-------+-------+-------+ 208 * b | fill count | 209 * +-------+-------+-------+-------+-------+-------+-------+-------+ 210 * c | checksum[0] | 211 * +-------+-------+-------+-------+-------+-------+-------+-------+ 212 * d | checksum[1] | 213 * +-------+-------+-------+-------+-------+-------+-------+-------+ 214 * e | checksum[2] | 215 * +-------+-------+-------+-------+-------+-------+-------+-------+ 216 * f | checksum[3] | 217 * +-------+-------+-------+-------+-------+-------+-------+-------+ 218 * 219 * Legend: 220 * 221 * vdev virtual device ID 222 * offset offset into virtual device 223 * LSIZE logical size 224 * PSIZE physical size (after compression) 225 * ASIZE allocated size (including RAID-Z parity and gang block headers) 226 * GRID RAID-Z layout information (reserved for future use) 227 * cksum checksum function 228 * comp compression function 229 * G gang block indicator 230 * B byteorder (endianness) 231 * D dedup 232 * X encryption 233 * E blkptr_t contains embedded data (see below) 234 * lvl level of indirection 235 * type DMU object type 236 * phys birth txg when dva[0] was written; zero if same as logical birth txg 237 * note that typically all the dva's would be written in this 238 * txg, but they could be different if they were moved by 239 * device removal. 240 * log. birth transaction group in which the block was logically born 241 * fill count number of non-zero blocks under this bp 242 * checksum[4] 256-bit checksum of the data this bp describes 243 */ 244 245 /* 246 * The blkptr_t's of encrypted blocks also need to store the encryption 247 * parameters so that the block can be decrypted. This layout is as follows: 248 * 249 * 64 56 48 40 32 24 16 8 0 250 * +-------+-------+-------+-------+-------+-------+-------+-------+ 251 * 0 | vdev1 | GRID | ASIZE | 252 * +-------+-------+-------+-------+-------+-------+-------+-------+ 253 * 1 |G| offset1 | 254 * +-------+-------+-------+-------+-------+-------+-------+-------+ 255 * 2 | vdev2 | GRID | ASIZE | 256 * +-------+-------+-------+-------+-------+-------+-------+-------+ 257 * 3 |G| offset2 | 258 * +-------+-------+-------+-------+-------+-------+-------+-------+ 259 * 4 | salt | 260 * +-------+-------+-------+-------+-------+-------+-------+-------+ 261 * 5 | IV1 | 262 * +-------+-------+-------+-------+-------+-------+-------+-------+ 263 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 264 * +-------+-------+-------+-------+-------+-------+-------+-------+ 265 * 7 | padding | 266 * +-------+-------+-------+-------+-------+-------+-------+-------+ 267 * 8 | padding | 268 * +-------+-------+-------+-------+-------+-------+-------+-------+ 269 * 9 | physical birth txg | 270 * +-------+-------+-------+-------+-------+-------+-------+-------+ 271 * a | logical birth txg | 272 * +-------+-------+-------+-------+-------+-------+-------+-------+ 273 * b | IV2 | fill count | 274 * +-------+-------+-------+-------+-------+-------+-------+-------+ 275 * c | checksum[0] | 276 * +-------+-------+-------+-------+-------+-------+-------+-------+ 277 * d | checksum[1] | 278 * +-------+-------+-------+-------+-------+-------+-------+-------+ 279 * e | MAC[0] | 280 * +-------+-------+-------+-------+-------+-------+-------+-------+ 281 * f | MAC[1] | 282 * +-------+-------+-------+-------+-------+-------+-------+-------+ 283 * 284 * Legend: 285 * 286 * salt Salt for generating encryption keys 287 * IV1 First 64 bits of encryption IV 288 * X Block requires encryption handling (set to 1) 289 * E blkptr_t contains embedded data (set to 0, see below) 290 * fill count number of non-zero blocks under this bp (truncated to 32 bits) 291 * IV2 Last 32 bits of encryption IV 292 * checksum[2] 128-bit checksum of the data this bp describes 293 * MAC[2] 128-bit message authentication code for this data 294 * 295 * The X bit being set indicates that this block is one of 3 types. If this is 296 * a level 0 block with an encrypted object type, the block is encrypted 297 * (see BP_IS_ENCRYPTED()). If this is a level 0 block with an unencrypted 298 * object type, this block is authenticated with an HMAC (see 299 * BP_IS_AUTHENTICATED()). Otherwise (if level > 0), this bp will use the MAC 300 * words to store a checksum-of-MACs from the level below (see 301 * BP_HAS_INDIRECT_MAC_CKSUM()). For convenience in the code, BP_IS_PROTECTED() 302 * refers to both encrypted and authenticated blocks and BP_USES_CRYPT() 303 * refers to any of these 3 kinds of blocks. 304 * 305 * The additional encryption parameters are the salt, IV, and MAC which are 306 * explained in greater detail in the block comment at the top of zio_crypt.c. 307 * The MAC occupies half of the checksum space since it serves a very similar 308 * purpose: to prevent data corruption on disk. The only functional difference 309 * is that the checksum is used to detect on-disk corruption whether or not the 310 * encryption key is loaded and the MAC provides additional protection against 311 * malicious disk tampering. We use the 3rd DVA to store the salt and first 312 * 64 bits of the IV. As a result encrypted blocks can only have 2 copies 313 * maximum instead of the normal 3. The last 32 bits of the IV are stored in 314 * the upper bits of what is usually the fill count. Note that only blocks at 315 * level 0 or -2 are ever encrypted, which allows us to guarantee that these 316 * 32 bits are not trampled over by other code (see zio_crypt.c for details). 317 * The salt and IV are not used for authenticated bps or bps with an indirect 318 * MAC checksum, so these blocks can utilize all 3 DVAs and the full 64 bits 319 * for the fill count. 320 */ 321 322 /* 323 * "Embedded" blkptr_t's don't actually point to a block, instead they 324 * have a data payload embedded in the blkptr_t itself. See the comment 325 * in blkptr.c for more details. 326 * 327 * The blkptr_t is laid out as follows: 328 * 329 * 64 56 48 40 32 24 16 8 0 330 * +-------+-------+-------+-------+-------+-------+-------+-------+ 331 * 0 | payload | 332 * 1 | payload | 333 * 2 | payload | 334 * 3 | payload | 335 * 4 | payload | 336 * 5 | payload | 337 * +-------+-------+-------+-------+-------+-------+-------+-------+ 338 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 339 * +-------+-------+-------+-------+-------+-------+-------+-------+ 340 * 7 | payload | 341 * 8 | payload | 342 * 9 | payload | 343 * +-------+-------+-------+-------+-------+-------+-------+-------+ 344 * a | logical birth txg | 345 * +-------+-------+-------+-------+-------+-------+-------+-------+ 346 * b | payload | 347 * c | payload | 348 * d | payload | 349 * e | payload | 350 * f | payload | 351 * +-------+-------+-------+-------+-------+-------+-------+-------+ 352 * 353 * Legend: 354 * 355 * payload contains the embedded data 356 * B (byteorder) byteorder (endianness) 357 * D (dedup) padding (set to zero) 358 * X encryption (set to zero; see above) 359 * E (embedded) set to one 360 * lvl indirection level 361 * type DMU object type 362 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 363 * comp compression function of payload 364 * PSIZE size of payload after compression, in bytes 365 * LSIZE logical size of payload, in bytes 366 * note that 25 bits is enough to store the largest 367 * "normal" BP's LSIZE (2^16 * 2^9) in bytes 368 * log. birth transaction group in which the block was logically born 369 * 370 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 371 * bp's they are stored in units of SPA_MINBLOCKSHIFT. 372 * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 373 * The B, D, X, lvl, type, and comp fields are stored the same as with normal 374 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 375 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 376 * other macros, as they assert that they are only used on BP's of the correct 377 * "embedded-ness". Encrypted blkptr_t's cannot be embedded because they use 378 * the payload space for encryption parameters (see the comment above on 379 * how encryption parameters are stored). 380 */ 381 382 #define BPE_GET_ETYPE(bp) \ 383 (ASSERT(BP_IS_EMBEDDED(bp)), \ 384 BF64_GET((bp)->blk_prop, 40, 8)) 385 #define BPE_SET_ETYPE(bp, t) do { \ 386 ASSERT(BP_IS_EMBEDDED(bp)); \ 387 BF64_SET((bp)->blk_prop, 40, 8, t); \ 388 _NOTE(CONSTCOND) } while (0) 389 390 #define BPE_GET_LSIZE(bp) \ 391 (ASSERT(BP_IS_EMBEDDED(bp)), \ 392 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 393 #define BPE_SET_LSIZE(bp, x) do { \ 394 ASSERT(BP_IS_EMBEDDED(bp)); \ 395 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 396 _NOTE(CONSTCOND) } while (0) 397 398 #define BPE_GET_PSIZE(bp) \ 399 (ASSERT(BP_IS_EMBEDDED(bp)), \ 400 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 401 #define BPE_SET_PSIZE(bp, x) do { \ 402 ASSERT(BP_IS_EMBEDDED(bp)); \ 403 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 404 _NOTE(CONSTCOND) } while (0) 405 406 typedef enum bp_embedded_type { 407 BP_EMBEDDED_TYPE_DATA, 408 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ 409 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED 410 } bp_embedded_type_t; 411 412 #define BPE_NUM_WORDS 14 413 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 414 #define BPE_IS_PAYLOADWORD(bp, wp) \ 415 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 416 417 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 418 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 419 #define SPA_SYNC_MIN_VDEVS 3 /* min vdevs to update during sync */ 420 421 /* 422 * A block is a hole when it has either 1) never been written to, or 423 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads 424 * without physically allocating disk space. Holes are represented in the 425 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is 426 * done through the BP_IS_HOLE macro. For holes, the logical size, level, 427 * DMU object type, and birth times are all also stored for holes that 428 * were written to at some point (i.e. were punched after having been filled). 429 */ 430 typedef struct blkptr { 431 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 432 uint64_t blk_prop; /* size, compression, type, etc */ 433 uint64_t blk_pad[2]; /* Extra space for the future */ 434 uint64_t blk_phys_birth; /* txg when block was allocated */ 435 uint64_t blk_birth; /* transaction group at birth */ 436 uint64_t blk_fill; /* fill count */ 437 zio_cksum_t blk_cksum; /* 256-bit checksum */ 438 } blkptr_t; 439 440 /* 441 * Macros to get and set fields in a bp or DVA. 442 */ 443 #define DVA_GET_ASIZE(dva) \ 444 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 445 #define DVA_SET_ASIZE(dva, x) \ 446 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 447 SPA_MINBLOCKSHIFT, 0, x) 448 449 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 450 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 451 452 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, SPA_VDEVBITS) 453 #define DVA_SET_VDEV(dva, x) \ 454 BF64_SET((dva)->dva_word[0], 32, SPA_VDEVBITS, x) 455 456 #define DVA_GET_OFFSET(dva) \ 457 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 458 #define DVA_SET_OFFSET(dva, x) \ 459 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 460 461 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 462 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 463 464 #define BP_GET_LSIZE(bp) \ 465 (BP_IS_EMBEDDED(bp) ? \ 466 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 467 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 468 #define BP_SET_LSIZE(bp, x) do { \ 469 ASSERT(!BP_IS_EMBEDDED(bp)); \ 470 BF64_SET_SB((bp)->blk_prop, \ 471 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 472 _NOTE(CONSTCOND) } while (0) 473 474 #define BP_GET_PSIZE(bp) \ 475 (BP_IS_EMBEDDED(bp) ? 0 : \ 476 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 477 #define BP_SET_PSIZE(bp, x) do { \ 478 ASSERT(!BP_IS_EMBEDDED(bp)); \ 479 BF64_SET_SB((bp)->blk_prop, \ 480 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 481 _NOTE(CONSTCOND) } while (0) 482 483 #define BP_GET_COMPRESS(bp) \ 484 BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS) 485 #define BP_SET_COMPRESS(bp, x) \ 486 BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x) 487 488 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 489 #define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x) 490 491 #define BP_GET_CHECKSUM(bp) \ 492 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \ 493 BF64_GET((bp)->blk_prop, 40, 8)) 494 #define BP_SET_CHECKSUM(bp, x) do { \ 495 ASSERT(!BP_IS_EMBEDDED(bp)); \ 496 BF64_SET((bp)->blk_prop, 40, 8, x); \ 497 _NOTE(CONSTCOND) } while (0) 498 499 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 500 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 501 502 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 503 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 504 505 /* encrypted, authenticated, and MAC cksum bps use the same bit */ 506 #define BP_USES_CRYPT(bp) BF64_GET((bp)->blk_prop, 61, 1) 507 #define BP_SET_CRYPT(bp, x) BF64_SET((bp)->blk_prop, 61, 1, x) 508 509 #define BP_IS_ENCRYPTED(bp) \ 510 (BP_USES_CRYPT(bp) && \ 511 BP_GET_LEVEL(bp) == 0 && \ 512 DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp))) 513 514 #define BP_IS_AUTHENTICATED(bp) \ 515 (BP_USES_CRYPT(bp) && \ 516 BP_GET_LEVEL(bp) == 0 && \ 517 !DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp))) 518 519 #define BP_HAS_INDIRECT_MAC_CKSUM(bp) \ 520 (BP_USES_CRYPT(bp) && BP_GET_LEVEL(bp) > 0) 521 522 #define BP_IS_PROTECTED(bp) \ 523 (BP_IS_ENCRYPTED(bp) || BP_IS_AUTHENTICATED(bp)) 524 525 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 526 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 527 528 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 529 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 530 531 #define BP_PHYSICAL_BIRTH(bp) \ 532 (BP_IS_EMBEDDED(bp) ? 0 : \ 533 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 534 535 #define BP_SET_BIRTH(bp, logical, physical) \ 536 { \ 537 ASSERT(!BP_IS_EMBEDDED(bp)); \ 538 (bp)->blk_birth = (logical); \ 539 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \ 540 } 541 542 #define BP_GET_FILL(bp) \ 543 ((BP_IS_ENCRYPTED(bp)) ? BF64_GET((bp)->blk_fill, 0, 32) : \ 544 ((BP_IS_EMBEDDED(bp)) ? 1 : (bp)->blk_fill)) 545 546 #define BP_SET_FILL(bp, fill) \ 547 { \ 548 if (BP_IS_ENCRYPTED(bp)) \ 549 BF64_SET((bp)->blk_fill, 0, 32, fill); \ 550 else \ 551 (bp)->blk_fill = fill; \ 552 } 553 554 #define BP_GET_IV2(bp) \ 555 (ASSERT(BP_IS_ENCRYPTED(bp)), \ 556 BF64_GET((bp)->blk_fill, 32, 32)) 557 #define BP_SET_IV2(bp, iv2) \ 558 { \ 559 ASSERT(BP_IS_ENCRYPTED(bp)); \ 560 BF64_SET((bp)->blk_fill, 32, 32, iv2); \ 561 } 562 563 #define BP_IS_METADATA(bp) \ 564 (BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) 565 566 #define BP_GET_ASIZE(bp) \ 567 (BP_IS_EMBEDDED(bp) ? 0 : \ 568 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 569 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 570 (DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp))) 571 572 #define BP_GET_UCSIZE(bp) \ 573 (BP_IS_METADATA(bp) ? BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)) 574 575 #define BP_GET_NDVAS(bp) \ 576 (BP_IS_EMBEDDED(bp) ? 0 : \ 577 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 578 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 579 (!!DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp))) 580 581 #define BP_COUNT_GANG(bp) \ 582 (BP_IS_EMBEDDED(bp) ? 0 : \ 583 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \ 584 DVA_GET_GANG(&(bp)->blk_dva[1]) + \ 585 (DVA_GET_GANG(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))) 586 587 #define DVA_EQUAL(dva1, dva2) \ 588 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 589 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 590 591 #define BP_EQUAL(bp1, bp2) \ 592 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \ 593 (bp1)->blk_birth == (bp2)->blk_birth && \ 594 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \ 595 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \ 596 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2])) 597 598 #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 599 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 600 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 601 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 602 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 603 604 #define ZIO_CHECKSUM_MAC_EQUAL(zc1, zc2) \ 605 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 606 ((zc1).zc_word[1] - (zc2).zc_word[1]))) 607 608 #define ZIO_CHECKSUM_IS_ZERO(zc) \ 609 (0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \ 610 (zc)->zc_word[2] | (zc)->zc_word[3])) 611 612 #define ZIO_CHECKSUM_BSWAP(zcp) \ 613 { \ 614 (zcp)->zc_word[0] = BSWAP_64((zcp)->zc_word[0]); \ 615 (zcp)->zc_word[1] = BSWAP_64((zcp)->zc_word[1]); \ 616 (zcp)->zc_word[2] = BSWAP_64((zcp)->zc_word[2]); \ 617 (zcp)->zc_word[3] = BSWAP_64((zcp)->zc_word[3]); \ 618 } 619 620 621 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 622 623 #define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ 624 { \ 625 (zcp)->zc_word[0] = w0; \ 626 (zcp)->zc_word[1] = w1; \ 627 (zcp)->zc_word[2] = w2; \ 628 (zcp)->zc_word[3] = w3; \ 629 } 630 631 #define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0]) 632 #define BP_IS_GANG(bp) \ 633 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp))) 634 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 635 (dva)->dva_word[1] == 0ULL) 636 #define BP_IS_HOLE(bp) \ 637 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp))) 638 639 /* BP_IS_RAIDZ(bp) assumes no block compression */ 640 #define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \ 641 BP_GET_PSIZE(bp)) 642 643 #define BP_ZERO(bp) \ 644 { \ 645 (bp)->blk_dva[0].dva_word[0] = 0; \ 646 (bp)->blk_dva[0].dva_word[1] = 0; \ 647 (bp)->blk_dva[1].dva_word[0] = 0; \ 648 (bp)->blk_dva[1].dva_word[1] = 0; \ 649 (bp)->blk_dva[2].dva_word[0] = 0; \ 650 (bp)->blk_dva[2].dva_word[1] = 0; \ 651 (bp)->blk_prop = 0; \ 652 (bp)->blk_pad[0] = 0; \ 653 (bp)->blk_pad[1] = 0; \ 654 (bp)->blk_phys_birth = 0; \ 655 (bp)->blk_birth = 0; \ 656 (bp)->blk_fill = 0; \ 657 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 658 } 659 660 #ifdef _BIG_ENDIAN 661 #define ZFS_HOST_BYTEORDER (0ULL) 662 #else 663 #define ZFS_HOST_BYTEORDER (1ULL) 664 #endif 665 666 #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) 667 668 #define BP_SPRINTF_LEN 400 669 670 /* 671 * This macro allows code sharing between zfs, libzpool, and mdb. 672 * 'func' is either snprintf() or mdb_snprintf(). 673 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line. 674 */ 675 #define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \ 676 { \ 677 static const char *copyname[] = \ 678 { "zero", "single", "double", "triple" }; \ 679 int len = 0; \ 680 int copies = 0; \ 681 const char *crypt_type; \ 682 if (bp != NULL) { \ 683 if (BP_IS_ENCRYPTED(bp)) { \ 684 crypt_type = "encrypted"; \ 685 } else if (BP_IS_AUTHENTICATED(bp)) { \ 686 crypt_type = "authenticated"; \ 687 } else if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) { \ 688 crypt_type = "indirect-MAC"; \ 689 } else { \ 690 crypt_type = "unencrypted"; \ 691 } \ 692 } \ 693 if (bp == NULL) { \ 694 len += func(buf + len, size - len, "<NULL>"); \ 695 } else if (BP_IS_HOLE(bp)) { \ 696 len += func(buf + len, size - len, \ 697 "HOLE [L%llu %s] " \ 698 "size=%llxL birth=%lluL", \ 699 (u_longlong_t)BP_GET_LEVEL(bp), \ 700 type, \ 701 (u_longlong_t)BP_GET_LSIZE(bp), \ 702 (u_longlong_t)bp->blk_birth); \ 703 } else if (BP_IS_EMBEDDED(bp)) { \ 704 len = func(buf + len, size - len, \ 705 "EMBEDDED [L%llu %s] et=%u %s " \ 706 "size=%llxL/%llxP birth=%lluL", \ 707 (u_longlong_t)BP_GET_LEVEL(bp), \ 708 type, \ 709 (int)BPE_GET_ETYPE(bp), \ 710 compress, \ 711 (u_longlong_t)BPE_GET_LSIZE(bp), \ 712 (u_longlong_t)BPE_GET_PSIZE(bp), \ 713 (u_longlong_t)bp->blk_birth); \ 714 } else { \ 715 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \ 716 const dva_t *dva = &bp->blk_dva[d]; \ 717 if (DVA_IS_VALID(dva)) \ 718 copies++; \ 719 len += func(buf + len, size - len, \ 720 "DVA[%d]=<%llu:%llx:%llx>%c", d, \ 721 (u_longlong_t)DVA_GET_VDEV(dva), \ 722 (u_longlong_t)DVA_GET_OFFSET(dva), \ 723 (u_longlong_t)DVA_GET_ASIZE(dva), \ 724 ws); \ 725 } \ 726 if (BP_IS_ENCRYPTED(bp)) { \ 727 len += func(buf + len, size - len, \ 728 "salt=%llx iv=%llx:%llx%c", \ 729 (u_longlong_t)bp->blk_dva[2].dva_word[0], \ 730 (u_longlong_t)bp->blk_dva[2].dva_word[1], \ 731 (u_longlong_t)BP_GET_IV2(bp), \ 732 ws); \ 733 } \ 734 if (BP_IS_GANG(bp) && \ 735 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \ 736 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \ 737 copies--; \ 738 len += func(buf + len, size - len, \ 739 "[L%llu %s] %s %s %s %s %s %s %s%c" \ 740 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \ 741 "cksum=%llx:%llx:%llx:%llx", \ 742 (u_longlong_t)BP_GET_LEVEL(bp), \ 743 type, \ 744 checksum, \ 745 compress, \ 746 crypt_type, \ 747 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \ 748 BP_IS_GANG(bp) ? "gang" : "contiguous", \ 749 BP_GET_DEDUP(bp) ? "dedup" : "unique", \ 750 copyname[copies], \ 751 ws, \ 752 (u_longlong_t)BP_GET_LSIZE(bp), \ 753 (u_longlong_t)BP_GET_PSIZE(bp), \ 754 (u_longlong_t)bp->blk_birth, \ 755 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \ 756 (u_longlong_t)BP_GET_FILL(bp), \ 757 ws, \ 758 (u_longlong_t)bp->blk_cksum.zc_word[0], \ 759 (u_longlong_t)bp->blk_cksum.zc_word[1], \ 760 (u_longlong_t)bp->blk_cksum.zc_word[2], \ 761 (u_longlong_t)bp->blk_cksum.zc_word[3]); \ 762 } \ 763 ASSERT(len < size); \ 764 } 765 766 #define BP_GET_BUFC_TYPE(bp) \ 767 (BP_IS_METADATA(bp) ? ARC_BUFC_METADATA : ARC_BUFC_DATA) 768 769 typedef enum spa_import_type { 770 SPA_IMPORT_EXISTING, 771 SPA_IMPORT_ASSEMBLE 772 } spa_import_type_t; 773 774 /* 775 * Send TRIM commands in-line during normal pool operation while deleting. 776 * OFF: no 777 * ON: yes 778 */ 779 typedef enum { 780 SPA_AUTOTRIM_OFF = 0, /* default */ 781 SPA_AUTOTRIM_ON 782 } spa_autotrim_t; 783 784 /* 785 * Reason TRIM command was issued, used internally for accounting purposes. 786 */ 787 typedef enum trim_type { 788 TRIM_TYPE_MANUAL = 0, 789 TRIM_TYPE_AUTO = 1, 790 } trim_type_t; 791 792 /* state manipulation functions */ 793 extern int spa_open(const char *pool, spa_t **, void *tag); 794 extern int spa_open_rewind(const char *pool, spa_t **, void *tag, 795 nvlist_t *policy, nvlist_t **config); 796 extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot, 797 size_t buflen); 798 extern int spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props, 799 nvlist_t *zplprops, struct dsl_crypto_params *dcp); 800 extern int spa_import_rootpool(char *devpath, char *devid); 801 extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props, 802 uint64_t flags); 803 extern nvlist_t *spa_tryimport(nvlist_t *tryconfig); 804 extern int spa_destroy(char *pool); 805 extern int spa_checkpoint(const char *pool); 806 extern int spa_checkpoint_discard(const char *pool); 807 extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force, 808 boolean_t hardforce); 809 extern int spa_reset(char *pool); 810 extern void spa_async_request(spa_t *spa, int flag); 811 extern void spa_async_unrequest(spa_t *spa, int flag); 812 extern void spa_async_suspend(spa_t *spa); 813 extern void spa_async_resume(spa_t *spa); 814 extern spa_t *spa_inject_addref(char *pool); 815 extern void spa_inject_delref(spa_t *spa); 816 extern void spa_scan_stat_init(spa_t *spa); 817 extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps); 818 819 #define SPA_ASYNC_CONFIG_UPDATE 0x01 820 #define SPA_ASYNC_REMOVE 0x02 821 #define SPA_ASYNC_PROBE 0x04 822 #define SPA_ASYNC_RESILVER_DONE 0x08 823 #define SPA_ASYNC_RESILVER 0x10 824 #define SPA_ASYNC_AUTOEXPAND 0x20 825 #define SPA_ASYNC_REMOVE_DONE 0x40 826 #define SPA_ASYNC_REMOVE_STOP 0x80 827 #define SPA_ASYNC_INITIALIZE_RESTART 0x100 828 #define SPA_ASYNC_TRIM_RESTART 0x200 829 #define SPA_ASYNC_AUTOTRIM_RESTART 0x400 830 831 /* 832 * Controls the behavior of spa_vdev_remove(). 833 */ 834 #define SPA_REMOVE_UNSPARE 0x01 835 #define SPA_REMOVE_DONE 0x02 836 837 /* device manipulation */ 838 extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot); 839 extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, 840 int replacing); 841 extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, 842 int replace_done); 843 extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare); 844 extern boolean_t spa_vdev_remove_active(spa_t *spa); 845 extern int spa_vdev_initialize(spa_t *spa, nvlist_t *nv, uint64_t cmd_type, 846 nvlist_t *vdev_errlist); 847 extern int spa_vdev_trim(spa_t *spa, nvlist_t *nv, uint64_t cmd_type, 848 uint64_t rate, boolean_t partial, boolean_t secure, nvlist_t *vdev_errlist); 849 extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath); 850 extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru); 851 extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config, 852 nvlist_t *props, boolean_t exp); 853 854 /* spare state (which is global across all pools) */ 855 extern void spa_spare_add(vdev_t *vd); 856 extern void spa_spare_remove(vdev_t *vd); 857 extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt); 858 extern void spa_spare_activate(vdev_t *vd); 859 860 /* L2ARC state (which is global across all pools) */ 861 extern void spa_l2cache_add(vdev_t *vd); 862 extern void spa_l2cache_remove(vdev_t *vd); 863 extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool); 864 extern void spa_l2cache_activate(vdev_t *vd); 865 extern void spa_l2cache_drop(spa_t *spa); 866 867 /* scanning */ 868 extern int spa_scan(spa_t *spa, pool_scan_func_t func); 869 extern int spa_scan_stop(spa_t *spa); 870 extern int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t flag); 871 872 /* spa syncing */ 873 extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */ 874 extern void spa_sync_allpools(void); 875 876 /* spa namespace global mutex */ 877 extern kmutex_t spa_namespace_lock; 878 879 /* 880 * SPA configuration functions in spa_config.c 881 */ 882 883 #define SPA_CONFIG_UPDATE_POOL 0 884 #define SPA_CONFIG_UPDATE_VDEVS 1 885 886 extern void spa_write_cachefile(spa_t *, boolean_t, boolean_t); 887 extern void spa_config_load(void); 888 extern nvlist_t *spa_all_configs(uint64_t *); 889 extern void spa_config_set(spa_t *spa, nvlist_t *config); 890 extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, 891 int getstats); 892 extern void spa_config_update(spa_t *spa, int what); 893 894 /* 895 * Miscellaneous SPA routines in spa_misc.c 896 */ 897 898 /* Namespace manipulation */ 899 extern spa_t *spa_lookup(const char *name); 900 extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot); 901 extern void spa_remove(spa_t *spa); 902 extern spa_t *spa_next(spa_t *prev); 903 904 /* Refcount functions */ 905 extern void spa_open_ref(spa_t *spa, void *tag); 906 extern void spa_close(spa_t *spa, void *tag); 907 extern void spa_async_close(spa_t *spa, void *tag); 908 extern boolean_t spa_refcount_zero(spa_t *spa); 909 910 #define SCL_NONE 0x00 911 #define SCL_CONFIG 0x01 912 #define SCL_STATE 0x02 913 #define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */ 914 #define SCL_ALLOC 0x08 915 #define SCL_ZIO 0x10 916 #define SCL_FREE 0x20 917 #define SCL_VDEV 0x40 918 #define SCL_LOCKS 7 919 #define SCL_ALL ((1 << SCL_LOCKS) - 1) 920 #define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO) 921 922 /* Assorted pool IO kstats */ 923 typedef struct spa_iostats { 924 kstat_named_t trim_extents_written; 925 kstat_named_t trim_bytes_written; 926 kstat_named_t trim_extents_skipped; 927 kstat_named_t trim_bytes_skipped; 928 kstat_named_t trim_extents_failed; 929 kstat_named_t trim_bytes_failed; 930 kstat_named_t autotrim_extents_written; 931 kstat_named_t autotrim_bytes_written; 932 kstat_named_t autotrim_extents_skipped; 933 kstat_named_t autotrim_bytes_skipped; 934 kstat_named_t autotrim_extents_failed; 935 kstat_named_t autotrim_bytes_failed; 936 } spa_iostats_t; 937 938 /* Pool configuration locks */ 939 extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw); 940 extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw); 941 extern void spa_config_exit(spa_t *spa, int locks, void *tag); 942 extern int spa_config_held(spa_t *spa, int locks, krw_t rw); 943 944 /* Pool vdev add/remove lock */ 945 extern uint64_t spa_vdev_enter(spa_t *spa); 946 extern uint64_t spa_vdev_config_enter(spa_t *spa); 947 extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, 948 int error, char *tag); 949 extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error); 950 951 /* Pool vdev state change lock */ 952 extern void spa_vdev_state_enter(spa_t *spa, int oplock); 953 extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error); 954 955 /* Log state */ 956 typedef enum spa_log_state { 957 SPA_LOG_UNKNOWN = 0, /* unknown log state */ 958 SPA_LOG_MISSING, /* missing log(s) */ 959 SPA_LOG_CLEAR, /* clear the log(s) */ 960 SPA_LOG_GOOD, /* log(s) are good */ 961 } spa_log_state_t; 962 963 extern spa_log_state_t spa_get_log_state(spa_t *spa); 964 extern void spa_set_log_state(spa_t *spa, spa_log_state_t state); 965 extern int spa_reset_logs(spa_t *spa); 966 967 /* Log claim callback */ 968 extern void spa_claim_notify(zio_t *zio); 969 970 /* Accessor functions */ 971 extern boolean_t spa_shutting_down(spa_t *spa); 972 extern struct dsl_pool *spa_get_dsl(spa_t *spa); 973 extern boolean_t spa_is_initializing(spa_t *spa); 974 extern boolean_t spa_indirect_vdevs_loaded(spa_t *spa); 975 extern blkptr_t *spa_get_rootblkptr(spa_t *spa); 976 extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp); 977 extern void spa_altroot(spa_t *, char *, size_t); 978 extern int spa_sync_pass(spa_t *spa); 979 extern char *spa_name(spa_t *spa); 980 extern uint64_t spa_guid(spa_t *spa); 981 extern uint64_t spa_load_guid(spa_t *spa); 982 extern uint64_t spa_last_synced_txg(spa_t *spa); 983 extern uint64_t spa_first_txg(spa_t *spa); 984 extern uint64_t spa_syncing_txg(spa_t *spa); 985 extern uint64_t spa_final_dirty_txg(spa_t *spa); 986 extern uint64_t spa_version(spa_t *spa); 987 extern pool_state_t spa_state(spa_t *spa); 988 extern spa_load_state_t spa_load_state(spa_t *spa); 989 extern uint64_t spa_freeze_txg(spa_t *spa); 990 extern uint64_t spa_get_worst_case_asize(spa_t *spa, uint64_t lsize); 991 extern uint64_t spa_get_dspace(spa_t *spa); 992 extern uint64_t spa_get_checkpoint_space(spa_t *spa); 993 extern uint64_t spa_get_slop_space(spa_t *spa); 994 extern void spa_update_dspace(spa_t *spa); 995 extern uint64_t spa_version(spa_t *spa); 996 extern boolean_t spa_deflate(spa_t *spa); 997 extern metaslab_class_t *spa_normal_class(spa_t *spa); 998 extern metaslab_class_t *spa_log_class(spa_t *spa); 999 extern metaslab_class_t *spa_special_class(spa_t *spa); 1000 extern metaslab_class_t *spa_dedup_class(spa_t *spa); 1001 extern metaslab_class_t *spa_preferred_class(spa_t *spa, uint64_t size, 1002 dmu_object_type_t objtype, uint_t level, uint_t special_smallblk); 1003 1004 extern void spa_evicting_os_register(spa_t *, objset_t *os); 1005 extern void spa_evicting_os_deregister(spa_t *, objset_t *os); 1006 extern void spa_evicting_os_wait(spa_t *spa); 1007 extern int spa_max_replication(spa_t *spa); 1008 extern int spa_prev_software_version(spa_t *spa); 1009 extern int spa_busy(void); 1010 extern uint8_t spa_get_failmode(spa_t *spa); 1011 extern boolean_t spa_suspended(spa_t *spa); 1012 extern uint64_t spa_bootfs(spa_t *spa); 1013 extern uint64_t spa_delegation(spa_t *spa); 1014 extern objset_t *spa_meta_objset(spa_t *spa); 1015 extern space_map_t *spa_syncing_log_sm(spa_t *spa); 1016 extern uint64_t spa_deadman_synctime(spa_t *spa); 1017 extern uint64_t spa_dirty_data(spa_t *spa); 1018 extern spa_autotrim_t spa_get_autotrim(spa_t *spa); 1019 1020 /* Miscellaneous support routines */ 1021 extern void spa_load_failed(spa_t *spa, const char *fmt, ...); 1022 extern void spa_load_note(spa_t *spa, const char *fmt, ...); 1023 extern void spa_activate_mos_feature(spa_t *spa, const char *feature, 1024 dmu_tx_t *tx); 1025 extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature); 1026 extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid); 1027 extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid); 1028 extern char *spa_strdup(const char *); 1029 extern void spa_strfree(char *); 1030 extern uint64_t spa_get_random(uint64_t range); 1031 extern uint64_t spa_generate_guid(spa_t *spa); 1032 extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp); 1033 extern void spa_freeze(spa_t *spa); 1034 extern int spa_change_guid(spa_t *spa); 1035 extern void spa_upgrade(spa_t *spa, uint64_t version); 1036 extern void spa_evict_all(void); 1037 extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid, 1038 boolean_t l2cache); 1039 extern boolean_t spa_has_spare(spa_t *, uint64_t guid); 1040 extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva); 1041 extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp); 1042 extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp); 1043 extern boolean_t spa_has_slogs(spa_t *spa); 1044 extern boolean_t spa_is_root(spa_t *spa); 1045 extern boolean_t spa_writeable(spa_t *spa); 1046 extern boolean_t spa_has_pending_synctask(spa_t *spa); 1047 extern int spa_maxblocksize(spa_t *spa); 1048 extern int spa_maxdnodesize(spa_t *spa); 1049 extern boolean_t spa_multihost(spa_t *spa); 1050 extern unsigned long spa_get_hostid(void); 1051 extern boolean_t spa_has_checkpoint(spa_t *spa); 1052 extern boolean_t spa_importing_readonly_checkpoint(spa_t *spa); 1053 extern boolean_t spa_suspend_async_destroy(spa_t *spa); 1054 extern uint64_t spa_min_claim_txg(spa_t *spa); 1055 extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp); 1056 extern boolean_t zfs_dva_valid(spa_t *spa, const dva_t *dva, 1057 const blkptr_t *bp); 1058 typedef void (*spa_remap_cb_t)(uint64_t vdev, uint64_t offset, uint64_t size, 1059 void *arg); 1060 extern boolean_t spa_remap_blkptr(spa_t *spa, blkptr_t *bp, 1061 spa_remap_cb_t callback, void *arg); 1062 extern uint64_t spa_get_last_removal_txg(spa_t *spa); 1063 extern boolean_t spa_trust_config(spa_t *spa); 1064 extern uint64_t spa_missing_tvds_allowed(spa_t *spa); 1065 extern void spa_set_missing_tvds(spa_t *spa, uint64_t missing); 1066 extern boolean_t spa_top_vdevs_spacemap_addressable(spa_t *spa); 1067 extern uint64_t spa_total_metaslabs(spa_t *spa); 1068 extern void spa_activate_allocation_classes(spa_t *, dmu_tx_t *); 1069 1070 extern int spa_mode(spa_t *spa); 1071 extern uint64_t zfs_strtonum(const char *str, char **nptr); 1072 1073 extern char *spa_his_ievent_table[]; 1074 1075 extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx); 1076 extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read, 1077 char *his_buf); 1078 extern int spa_history_log(spa_t *spa, const char *his_buf); 1079 extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl); 1080 extern void spa_history_log_version(spa_t *spa, const char *operation); 1081 extern void spa_history_log_internal(spa_t *spa, const char *operation, 1082 dmu_tx_t *tx, const char *fmt, ...); 1083 extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op, 1084 dmu_tx_t *tx, const char *fmt, ...); 1085 extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation, 1086 dmu_tx_t *tx, const char *fmt, ...); 1087 1088 /* error handling */ 1089 struct zbookmark_phys; 1090 extern void spa_log_error(spa_t *spa, const struct zbookmark_phys *zb); 1091 extern void zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd, 1092 const struct zbookmark_phys *zb, struct zio *zio, uint64_t stateoroffset, 1093 uint64_t length); 1094 extern void zfs_post_remove(spa_t *spa, vdev_t *vd); 1095 extern void zfs_post_state_change(spa_t *spa, vdev_t *vd); 1096 extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd); 1097 extern uint64_t spa_get_errlog_size(spa_t *spa); 1098 extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count); 1099 extern void spa_errlog_rotate(spa_t *spa); 1100 extern void spa_errlog_drain(spa_t *spa); 1101 extern void spa_errlog_sync(spa_t *spa, uint64_t txg); 1102 extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub); 1103 1104 /* vdev cache */ 1105 extern void vdev_cache_stat_init(void); 1106 extern void vdev_cache_stat_fini(void); 1107 1108 /* vdev mirror */ 1109 extern void vdev_mirror_stat_init(void); 1110 extern void vdev_mirror_stat_fini(void); 1111 1112 /* Initialization and termination */ 1113 extern void spa_init(int flags); 1114 extern void spa_fini(void); 1115 extern void spa_boot_init(void); 1116 1117 /* properties */ 1118 extern int spa_prop_set(spa_t *spa, nvlist_t *nvp); 1119 extern int spa_prop_get(spa_t *spa, nvlist_t **nvp); 1120 extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx); 1121 extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t); 1122 1123 /* asynchronous event notification */ 1124 extern void spa_event_notify(spa_t *spa, vdev_t *vdev, nvlist_t *hist_nvl, 1125 const char *name); 1126 extern sysevent_t *spa_event_create(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl, 1127 const char *name); 1128 extern void spa_event_post(sysevent_t *ev); 1129 extern void spa_event_discard(sysevent_t *ev); 1130 1131 #ifdef ZFS_DEBUG 1132 #define dprintf_bp(bp, fmt, ...) do { \ 1133 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 1134 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \ 1135 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \ 1136 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \ 1137 kmem_free(__blkbuf, BP_SPRINTF_LEN); \ 1138 } \ 1139 _NOTE(CONSTCOND) } while (0) 1140 #else 1141 #define dprintf_bp(bp, fmt, ...) 1142 #endif 1143 1144 extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */ 1145 1146 #ifdef __cplusplus 1147 } 1148 #endif 1149 1150 #endif /* _SYS_SPA_H */ 1151