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 2013 Saso Kiselkov. All rights reserved. 24 * Copyright (c) 2011, 2014 by Delphix. All rights reserved. 25 * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 26 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. 27 * Copyright 2013 Saso Kiselkov. All rights reserved. 28 * Copyright (c) 2014 Integros [integros.com] 29 */ 30 31 #ifndef _SYS_SPA_H 32 #define _SYS_SPA_H 33 34 #include <sys/avl.h> 35 #include <sys/zfs_context.h> 36 #include <sys/nvpair.h> 37 #include <sys/sysmacros.h> 38 #include <sys/types.h> 39 #include <sys/fs/zfs.h> 40 41 #ifdef __cplusplus 42 extern "C" { 43 #endif 44 45 /* 46 * Forward references that lots of things need. 47 */ 48 typedef struct spa spa_t; 49 typedef struct vdev vdev_t; 50 typedef struct metaslab metaslab_t; 51 typedef struct metaslab_group metaslab_group_t; 52 typedef struct metaslab_class metaslab_class_t; 53 typedef struct zio zio_t; 54 typedef struct zilog zilog_t; 55 typedef struct spa_aux_vdev spa_aux_vdev_t; 56 typedef struct ddt ddt_t; 57 typedef struct ddt_entry ddt_entry_t; 58 struct dsl_pool; 59 struct dsl_dataset; 60 61 /* 62 * General-purpose 32-bit and 64-bit bitfield encodings. 63 */ 64 #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) 65 #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) 66 #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) 67 #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) 68 69 #define BF32_GET(x, low, len) BF32_DECODE(x, low, len) 70 #define BF64_GET(x, low, len) BF64_DECODE(x, low, len) 71 72 #define BF32_SET(x, low, len, val) do { \ 73 ASSERT3U(val, <, 1U << (len)); \ 74 ASSERT3U(low + len, <=, 32); \ 75 (x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \ 76 _NOTE(CONSTCOND) } while (0) 77 78 #define BF64_SET(x, low, len, val) do { \ 79 ASSERT3U(val, <, 1ULL << (len)); \ 80 ASSERT3U(low + len, <=, 64); \ 81 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \ 82 _NOTE(CONSTCOND) } while (0) 83 84 #define BF32_GET_SB(x, low, len, shift, bias) \ 85 ((BF32_GET(x, low, len) + (bias)) << (shift)) 86 #define BF64_GET_SB(x, low, len, shift, bias) \ 87 ((BF64_GET(x, low, len) + (bias)) << (shift)) 88 89 #define BF32_SET_SB(x, low, len, shift, bias, val) do { \ 90 ASSERT(IS_P2ALIGNED(val, 1U << shift)); \ 91 ASSERT3S((val) >> (shift), >=, bias); \ 92 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \ 93 _NOTE(CONSTCOND) } while (0) 94 #define BF64_SET_SB(x, low, len, shift, bias, val) do { \ 95 ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \ 96 ASSERT3S((val) >> (shift), >=, bias); \ 97 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \ 98 _NOTE(CONSTCOND) } while (0) 99 100 /* 101 * We currently support block sizes from 512 bytes to 16MB. 102 * The benefits of larger blocks, and thus larger IO, need to be weighed 103 * against the cost of COWing a giant block to modify one byte, and the 104 * large latency of reading or writing a large block. 105 * 106 * Note that although blocks up to 16MB are supported, the recordsize 107 * property can not be set larger than zfs_max_recordsize (default 1MB). 108 * See the comment near zfs_max_recordsize in dsl_dataset.c for details. 109 * 110 * Note that although the LSIZE field of the blkptr_t can store sizes up 111 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to 112 * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB. 113 */ 114 #define SPA_MINBLOCKSHIFT 9 115 #define SPA_OLD_MAXBLOCKSHIFT 17 116 #define SPA_MAXBLOCKSHIFT 24 117 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 118 #define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT) 119 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 120 121 /* 122 * Size of block to hold the configuration data (a packed nvlist) 123 */ 124 #define SPA_CONFIG_BLOCKSIZE (1ULL << 14) 125 126 /* 127 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. 128 * The ASIZE encoding should be at least 64 times larger (6 more bits) 129 * to support up to 4-way RAID-Z mirror mode with worst-case gang block 130 * overhead, three DVAs per bp, plus one more bit in case we do anything 131 * else that expands the ASIZE. 132 */ 133 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ 134 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ 135 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ 136 137 /* 138 * All SPA data is represented by 128-bit data virtual addresses (DVAs). 139 * The members of the dva_t should be considered opaque outside the SPA. 140 */ 141 typedef struct dva { 142 uint64_t dva_word[2]; 143 } dva_t; 144 145 /* 146 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. 147 */ 148 typedef struct zio_cksum { 149 uint64_t zc_word[4]; 150 } zio_cksum_t; 151 152 /* 153 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept 154 * secret and is suitable for use in MAC algorithms as the key. 155 */ 156 typedef struct zio_cksum_salt { 157 uint8_t zcs_bytes[32]; 158 } zio_cksum_salt_t; 159 160 /* 161 * Each block is described by its DVAs, time of birth, checksum, etc. 162 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 163 * 164 * 64 56 48 40 32 24 16 8 0 165 * +-------+-------+-------+-------+-------+-------+-------+-------+ 166 * 0 | vdev1 | GRID | ASIZE | 167 * +-------+-------+-------+-------+-------+-------+-------+-------+ 168 * 1 |G| offset1 | 169 * +-------+-------+-------+-------+-------+-------+-------+-------+ 170 * 2 | vdev2 | GRID | ASIZE | 171 * +-------+-------+-------+-------+-------+-------+-------+-------+ 172 * 3 |G| offset2 | 173 * +-------+-------+-------+-------+-------+-------+-------+-------+ 174 * 4 | vdev3 | GRID | ASIZE | 175 * +-------+-------+-------+-------+-------+-------+-------+-------+ 176 * 5 |G| offset3 | 177 * +-------+-------+-------+-------+-------+-------+-------+-------+ 178 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 179 * +-------+-------+-------+-------+-------+-------+-------+-------+ 180 * 7 | padding | 181 * +-------+-------+-------+-------+-------+-------+-------+-------+ 182 * 8 | padding | 183 * +-------+-------+-------+-------+-------+-------+-------+-------+ 184 * 9 | physical birth txg | 185 * +-------+-------+-------+-------+-------+-------+-------+-------+ 186 * a | logical birth txg | 187 * +-------+-------+-------+-------+-------+-------+-------+-------+ 188 * b | fill count | 189 * +-------+-------+-------+-------+-------+-------+-------+-------+ 190 * c | checksum[0] | 191 * +-------+-------+-------+-------+-------+-------+-------+-------+ 192 * d | checksum[1] | 193 * +-------+-------+-------+-------+-------+-------+-------+-------+ 194 * e | checksum[2] | 195 * +-------+-------+-------+-------+-------+-------+-------+-------+ 196 * f | checksum[3] | 197 * +-------+-------+-------+-------+-------+-------+-------+-------+ 198 * 199 * Legend: 200 * 201 * vdev virtual device ID 202 * offset offset into virtual device 203 * LSIZE logical size 204 * PSIZE physical size (after compression) 205 * ASIZE allocated size (including RAID-Z parity and gang block headers) 206 * GRID RAID-Z layout information (reserved for future use) 207 * cksum checksum function 208 * comp compression function 209 * G gang block indicator 210 * B byteorder (endianness) 211 * D dedup 212 * X encryption (on version 30, which is not supported) 213 * E blkptr_t contains embedded data (see below) 214 * lvl level of indirection 215 * type DMU object type 216 * phys birth txg of block allocation; zero if same as logical birth txg 217 * log. birth transaction group in which the block was logically born 218 * fill count number of non-zero blocks under this bp 219 * checksum[4] 256-bit checksum of the data this bp describes 220 */ 221 222 /* 223 * "Embedded" blkptr_t's don't actually point to a block, instead they 224 * have a data payload embedded in the blkptr_t itself. See the comment 225 * in blkptr.c for more details. 226 * 227 * The blkptr_t is laid out as follows: 228 * 229 * 64 56 48 40 32 24 16 8 0 230 * +-------+-------+-------+-------+-------+-------+-------+-------+ 231 * 0 | payload | 232 * 1 | payload | 233 * 2 | payload | 234 * 3 | payload | 235 * 4 | payload | 236 * 5 | payload | 237 * +-------+-------+-------+-------+-------+-------+-------+-------+ 238 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 239 * +-------+-------+-------+-------+-------+-------+-------+-------+ 240 * 7 | payload | 241 * 8 | payload | 242 * 9 | payload | 243 * +-------+-------+-------+-------+-------+-------+-------+-------+ 244 * a | logical birth txg | 245 * +-------+-------+-------+-------+-------+-------+-------+-------+ 246 * b | payload | 247 * c | payload | 248 * d | payload | 249 * e | payload | 250 * f | payload | 251 * +-------+-------+-------+-------+-------+-------+-------+-------+ 252 * 253 * Legend: 254 * 255 * payload contains the embedded data 256 * B (byteorder) byteorder (endianness) 257 * D (dedup) padding (set to zero) 258 * X encryption (set to zero; see above) 259 * E (embedded) set to one 260 * lvl indirection level 261 * type DMU object type 262 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 263 * comp compression function of payload 264 * PSIZE size of payload after compression, in bytes 265 * LSIZE logical size of payload, in bytes 266 * note that 25 bits is enough to store the largest 267 * "normal" BP's LSIZE (2^16 * 2^9) in bytes 268 * log. birth transaction group in which the block was logically born 269 * 270 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 271 * bp's they are stored in units of SPA_MINBLOCKSHIFT. 272 * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 273 * The B, D, X, lvl, type, and comp fields are stored the same as with normal 274 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 275 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 276 * other macros, as they assert that they are only used on BP's of the correct 277 * "embedded-ness". 278 */ 279 280 #define BPE_GET_ETYPE(bp) \ 281 (ASSERT(BP_IS_EMBEDDED(bp)), \ 282 BF64_GET((bp)->blk_prop, 40, 8)) 283 #define BPE_SET_ETYPE(bp, t) do { \ 284 ASSERT(BP_IS_EMBEDDED(bp)); \ 285 BF64_SET((bp)->blk_prop, 40, 8, t); \ 286 _NOTE(CONSTCOND) } while (0) 287 288 #define BPE_GET_LSIZE(bp) \ 289 (ASSERT(BP_IS_EMBEDDED(bp)), \ 290 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 291 #define BPE_SET_LSIZE(bp, x) do { \ 292 ASSERT(BP_IS_EMBEDDED(bp)); \ 293 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 294 _NOTE(CONSTCOND) } while (0) 295 296 #define BPE_GET_PSIZE(bp) \ 297 (ASSERT(BP_IS_EMBEDDED(bp)), \ 298 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 299 #define BPE_SET_PSIZE(bp, x) do { \ 300 ASSERT(BP_IS_EMBEDDED(bp)); \ 301 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 302 _NOTE(CONSTCOND) } while (0) 303 304 typedef enum bp_embedded_type { 305 BP_EMBEDDED_TYPE_DATA, 306 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ 307 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED 308 } bp_embedded_type_t; 309 310 #define BPE_NUM_WORDS 14 311 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 312 #define BPE_IS_PAYLOADWORD(bp, wp) \ 313 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 314 315 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 316 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 317 318 /* 319 * A block is a hole when it has either 1) never been written to, or 320 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads 321 * without physically allocating disk space. Holes are represented in the 322 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is 323 * done through the BP_IS_HOLE macro. For holes, the logical size, level, 324 * DMU object type, and birth times are all also stored for holes that 325 * were written to at some point (i.e. were punched after having been filled). 326 */ 327 typedef struct blkptr { 328 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 329 uint64_t blk_prop; /* size, compression, type, etc */ 330 uint64_t blk_pad[2]; /* Extra space for the future */ 331 uint64_t blk_phys_birth; /* txg when block was allocated */ 332 uint64_t blk_birth; /* transaction group at birth */ 333 uint64_t blk_fill; /* fill count */ 334 zio_cksum_t blk_cksum; /* 256-bit checksum */ 335 } blkptr_t; 336 337 /* 338 * Macros to get and set fields in a bp or DVA. 339 */ 340 #define DVA_GET_ASIZE(dva) \ 341 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 342 #define DVA_SET_ASIZE(dva, x) \ 343 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 344 SPA_MINBLOCKSHIFT, 0, x) 345 346 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 347 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 348 349 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) 350 #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) 351 352 #define DVA_GET_OFFSET(dva) \ 353 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 354 #define DVA_SET_OFFSET(dva, x) \ 355 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 356 357 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 358 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 359 360 #define BP_GET_LSIZE(bp) \ 361 (BP_IS_EMBEDDED(bp) ? \ 362 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 363 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 364 #define BP_SET_LSIZE(bp, x) do { \ 365 ASSERT(!BP_IS_EMBEDDED(bp)); \ 366 BF64_SET_SB((bp)->blk_prop, \ 367 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 368 _NOTE(CONSTCOND) } while (0) 369 370 #define BP_GET_PSIZE(bp) \ 371 (BP_IS_EMBEDDED(bp) ? 0 : \ 372 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 373 #define BP_SET_PSIZE(bp, x) do { \ 374 ASSERT(!BP_IS_EMBEDDED(bp)); \ 375 BF64_SET_SB((bp)->blk_prop, \ 376 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 377 _NOTE(CONSTCOND) } while (0) 378 379 #define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7) 380 #define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x) 381 382 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 383 #define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x) 384 385 #define BP_GET_CHECKSUM(bp) \ 386 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \ 387 BF64_GET((bp)->blk_prop, 40, 8)) 388 #define BP_SET_CHECKSUM(bp, x) do { \ 389 ASSERT(!BP_IS_EMBEDDED(bp)); \ 390 BF64_SET((bp)->blk_prop, 40, 8, x); \ 391 _NOTE(CONSTCOND) } while (0) 392 393 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 394 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 395 396 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 397 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 398 399 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 400 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 401 402 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 403 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 404 405 #define BP_PHYSICAL_BIRTH(bp) \ 406 (BP_IS_EMBEDDED(bp) ? 0 : \ 407 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 408 409 #define BP_SET_BIRTH(bp, logical, physical) \ 410 { \ 411 ASSERT(!BP_IS_EMBEDDED(bp)); \ 412 (bp)->blk_birth = (logical); \ 413 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \ 414 } 415 416 #define BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill) 417 418 #define BP_GET_ASIZE(bp) \ 419 (BP_IS_EMBEDDED(bp) ? 0 : \ 420 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 421 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 422 DVA_GET_ASIZE(&(bp)->blk_dva[2])) 423 424 #define BP_GET_UCSIZE(bp) \ 425 ((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \ 426 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)) 427 428 #define BP_GET_NDVAS(bp) \ 429 (BP_IS_EMBEDDED(bp) ? 0 : \ 430 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 431 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 432 !!DVA_GET_ASIZE(&(bp)->blk_dva[2])) 433 434 #define BP_COUNT_GANG(bp) \ 435 (BP_IS_EMBEDDED(bp) ? 0 : \ 436 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \ 437 DVA_GET_GANG(&(bp)->blk_dva[1]) + \ 438 DVA_GET_GANG(&(bp)->blk_dva[2]))) 439 440 #define DVA_EQUAL(dva1, dva2) \ 441 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 442 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 443 444 #define BP_EQUAL(bp1, bp2) \ 445 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \ 446 (bp1)->blk_birth == (bp2)->blk_birth && \ 447 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \ 448 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \ 449 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2])) 450 451 #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 452 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 453 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 454 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 455 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 456 457 #define ZIO_CHECKSUM_IS_ZERO(zc) \ 458 (0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \ 459 (zc)->zc_word[2] | (zc)->zc_word[3])) 460 461 #define ZIO_CHECKSUM_BSWAP(_zc) \ 462 do { \ 463 zio_cksum_t *zc = (_zc); \ 464 zc->zc_word[0] = BSWAP_64(zc->zc_word[0]); \ 465 zc->zc_word[1] = BSWAP_64(zc->zc_word[1]); \ 466 zc->zc_word[2] = BSWAP_64(zc->zc_word[2]); \ 467 zc->zc_word[3] = BSWAP_64(zc->zc_word[3]); \ 468 _NOTE(NOTREACHED) \ 469 _NOTE(CONSTCOND) \ 470 } while (0) 471 472 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 473 474 #define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ 475 { \ 476 (zcp)->zc_word[0] = w0; \ 477 (zcp)->zc_word[1] = w1; \ 478 (zcp)->zc_word[2] = w2; \ 479 (zcp)->zc_word[3] = w3; \ 480 } 481 482 #define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0]) 483 #define BP_IS_GANG(bp) \ 484 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp))) 485 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 486 (dva)->dva_word[1] == 0ULL) 487 #define BP_IS_HOLE(bp) \ 488 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp))) 489 490 /* BP_IS_RAIDZ(bp) assumes no block compression */ 491 #define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \ 492 BP_GET_PSIZE(bp)) 493 494 #define BP_ZERO(bp) \ 495 { \ 496 (bp)->blk_dva[0].dva_word[0] = 0; \ 497 (bp)->blk_dva[0].dva_word[1] = 0; \ 498 (bp)->blk_dva[1].dva_word[0] = 0; \ 499 (bp)->blk_dva[1].dva_word[1] = 0; \ 500 (bp)->blk_dva[2].dva_word[0] = 0; \ 501 (bp)->blk_dva[2].dva_word[1] = 0; \ 502 (bp)->blk_prop = 0; \ 503 (bp)->blk_pad[0] = 0; \ 504 (bp)->blk_pad[1] = 0; \ 505 (bp)->blk_phys_birth = 0; \ 506 (bp)->blk_birth = 0; \ 507 (bp)->blk_fill = 0; \ 508 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 509 } 510 511 #ifdef _BIG_ENDIAN 512 #define ZFS_HOST_BYTEORDER (0ULL) 513 #else 514 #define ZFS_HOST_BYTEORDER (1ULL) 515 #endif 516 517 #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) 518 519 #define BP_SPRINTF_LEN 320 520 521 /* 522 * This macro allows code sharing between zfs, libzpool, and mdb. 523 * 'func' is either snprintf() or mdb_snprintf(). 524 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line. 525 */ 526 #define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \ 527 { \ 528 static const char *copyname[] = \ 529 { "zero", "single", "double", "triple" }; \ 530 int len = 0; \ 531 int copies = 0; \ 532 \ 533 if (bp == NULL) { \ 534 len += func(buf + len, size - len, "<NULL>"); \ 535 } else if (BP_IS_HOLE(bp)) { \ 536 len += func(buf + len, size - len, \ 537 "HOLE [L%llu %s] " \ 538 "size=%llxL birth=%lluL", \ 539 (u_longlong_t)BP_GET_LEVEL(bp), \ 540 type, \ 541 (u_longlong_t)BP_GET_LSIZE(bp), \ 542 (u_longlong_t)bp->blk_birth); \ 543 } else if (BP_IS_EMBEDDED(bp)) { \ 544 len = func(buf + len, size - len, \ 545 "EMBEDDED [L%llu %s] et=%u %s " \ 546 "size=%llxL/%llxP birth=%lluL", \ 547 (u_longlong_t)BP_GET_LEVEL(bp), \ 548 type, \ 549 (int)BPE_GET_ETYPE(bp), \ 550 compress, \ 551 (u_longlong_t)BPE_GET_LSIZE(bp), \ 552 (u_longlong_t)BPE_GET_PSIZE(bp), \ 553 (u_longlong_t)bp->blk_birth); \ 554 } else { \ 555 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \ 556 const dva_t *dva = &bp->blk_dva[d]; \ 557 if (DVA_IS_VALID(dva)) \ 558 copies++; \ 559 len += func(buf + len, size - len, \ 560 "DVA[%d]=<%llu:%llx:%llx>%c", d, \ 561 (u_longlong_t)DVA_GET_VDEV(dva), \ 562 (u_longlong_t)DVA_GET_OFFSET(dva), \ 563 (u_longlong_t)DVA_GET_ASIZE(dva), \ 564 ws); \ 565 } \ 566 if (BP_IS_GANG(bp) && \ 567 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \ 568 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \ 569 copies--; \ 570 len += func(buf + len, size - len, \ 571 "[L%llu %s] %s %s %s %s %s %s%c" \ 572 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \ 573 "cksum=%llx:%llx:%llx:%llx", \ 574 (u_longlong_t)BP_GET_LEVEL(bp), \ 575 type, \ 576 checksum, \ 577 compress, \ 578 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \ 579 BP_IS_GANG(bp) ? "gang" : "contiguous", \ 580 BP_GET_DEDUP(bp) ? "dedup" : "unique", \ 581 copyname[copies], \ 582 ws, \ 583 (u_longlong_t)BP_GET_LSIZE(bp), \ 584 (u_longlong_t)BP_GET_PSIZE(bp), \ 585 (u_longlong_t)bp->blk_birth, \ 586 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \ 587 (u_longlong_t)BP_GET_FILL(bp), \ 588 ws, \ 589 (u_longlong_t)bp->blk_cksum.zc_word[0], \ 590 (u_longlong_t)bp->blk_cksum.zc_word[1], \ 591 (u_longlong_t)bp->blk_cksum.zc_word[2], \ 592 (u_longlong_t)bp->blk_cksum.zc_word[3]); \ 593 } \ 594 ASSERT(len < size); \ 595 } 596 597 #include <sys/dmu.h> 598 599 #define BP_GET_BUFC_TYPE(bp) \ 600 (((BP_GET_LEVEL(bp) > 0) || (DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))) ? \ 601 ARC_BUFC_METADATA : ARC_BUFC_DATA) 602 603 typedef enum spa_import_type { 604 SPA_IMPORT_EXISTING, 605 SPA_IMPORT_ASSEMBLE 606 } spa_import_type_t; 607 608 /* state manipulation functions */ 609 extern int spa_open(const char *pool, spa_t **, void *tag); 610 extern int spa_open_rewind(const char *pool, spa_t **, void *tag, 611 nvlist_t *policy, nvlist_t **config); 612 extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot, 613 size_t buflen); 614 extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props, 615 nvlist_t *zplprops); 616 extern int spa_import_rootpool(char *devpath, char *devid); 617 extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props, 618 uint64_t flags); 619 extern nvlist_t *spa_tryimport(nvlist_t *tryconfig); 620 extern int spa_destroy(char *pool); 621 extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force, 622 boolean_t hardforce); 623 extern int spa_reset(char *pool); 624 extern void spa_async_request(spa_t *spa, int flag); 625 extern void spa_async_unrequest(spa_t *spa, int flag); 626 extern void spa_async_suspend(spa_t *spa); 627 extern void spa_async_resume(spa_t *spa); 628 extern spa_t *spa_inject_addref(char *pool); 629 extern void spa_inject_delref(spa_t *spa); 630 extern void spa_scan_stat_init(spa_t *spa); 631 extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps); 632 633 #define SPA_ASYNC_CONFIG_UPDATE 0x01 634 #define SPA_ASYNC_REMOVE 0x02 635 #define SPA_ASYNC_PROBE 0x04 636 #define SPA_ASYNC_RESILVER_DONE 0x08 637 #define SPA_ASYNC_RESILVER 0x10 638 #define SPA_ASYNC_AUTOEXPAND 0x20 639 #define SPA_ASYNC_REMOVE_DONE 0x40 640 #define SPA_ASYNC_REMOVE_STOP 0x80 641 #define SPA_ASYNC_L2CACHE_REBUILD 0x100 642 643 /* 644 * Controls the behavior of spa_vdev_remove(). 645 */ 646 #define SPA_REMOVE_UNSPARE 0x01 647 #define SPA_REMOVE_DONE 0x02 648 649 /* device manipulation */ 650 extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot); 651 extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, 652 int replacing); 653 extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, 654 int replace_done); 655 extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare); 656 extern boolean_t spa_vdev_remove_active(spa_t *spa); 657 extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath); 658 extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru); 659 extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config, 660 nvlist_t *props, boolean_t exp); 661 662 /* spare state (which is global across all pools) */ 663 extern void spa_spare_add(vdev_t *vd); 664 extern void spa_spare_remove(vdev_t *vd); 665 extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt); 666 extern void spa_spare_activate(vdev_t *vd); 667 668 /* L2ARC state (which is global across all pools) */ 669 extern void spa_l2cache_add(vdev_t *vd); 670 extern void spa_l2cache_remove(vdev_t *vd); 671 extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool); 672 extern void spa_l2cache_activate(vdev_t *vd); 673 extern void spa_l2cache_drop(spa_t *spa); 674 675 /* scanning */ 676 extern int spa_scan(spa_t *spa, pool_scan_func_t func); 677 extern int spa_scan_stop(spa_t *spa); 678 679 /* spa syncing */ 680 extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */ 681 extern void spa_sync_allpools(void); 682 683 /* spa namespace global mutex */ 684 extern kmutex_t spa_namespace_lock; 685 686 /* 687 * SPA configuration functions in spa_config.c 688 */ 689 690 #define SPA_CONFIG_UPDATE_POOL 0 691 #define SPA_CONFIG_UPDATE_VDEVS 1 692 693 extern void spa_config_sync(spa_t *, boolean_t, boolean_t); 694 extern void spa_config_load(void); 695 extern nvlist_t *spa_all_configs(uint64_t *); 696 extern void spa_config_set(spa_t *spa, nvlist_t *config); 697 extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, 698 int getstats); 699 extern void spa_config_update(spa_t *spa, int what); 700 701 /* 702 * Miscellaneous SPA routines in spa_misc.c 703 */ 704 705 /* Namespace manipulation */ 706 extern spa_t *spa_lookup(const char *name); 707 extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot); 708 extern void spa_remove(spa_t *spa); 709 extern spa_t *spa_next(spa_t *prev); 710 711 /* Refcount functions */ 712 extern void spa_open_ref(spa_t *spa, void *tag); 713 extern void spa_close(spa_t *spa, void *tag); 714 extern void spa_async_close(spa_t *spa, void *tag); 715 extern boolean_t spa_refcount_zero(spa_t *spa); 716 717 #define SCL_NONE 0x00 718 #define SCL_CONFIG 0x01 719 #define SCL_STATE 0x02 720 #define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */ 721 #define SCL_ALLOC 0x08 722 #define SCL_ZIO 0x10 723 #define SCL_FREE 0x20 724 #define SCL_VDEV 0x40 725 #define SCL_LOCKS 7 726 #define SCL_ALL ((1 << SCL_LOCKS) - 1) 727 #define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO) 728 729 /* Pool configuration locks */ 730 extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw); 731 extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw); 732 extern void spa_config_exit(spa_t *spa, int locks, void *tag); 733 extern int spa_config_held(spa_t *spa, int locks, krw_t rw); 734 735 /* Pool vdev add/remove lock */ 736 extern uint64_t spa_vdev_enter(spa_t *spa); 737 extern uint64_t spa_vdev_config_enter(spa_t *spa); 738 extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, 739 int error, char *tag); 740 extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error); 741 742 /* Pool vdev state change lock */ 743 extern void spa_vdev_state_enter(spa_t *spa, int oplock); 744 extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error); 745 746 /* Log state */ 747 typedef enum spa_log_state { 748 SPA_LOG_UNKNOWN = 0, /* unknown log state */ 749 SPA_LOG_MISSING, /* missing log(s) */ 750 SPA_LOG_CLEAR, /* clear the log(s) */ 751 SPA_LOG_GOOD, /* log(s) are good */ 752 } spa_log_state_t; 753 754 extern spa_log_state_t spa_get_log_state(spa_t *spa); 755 extern void spa_set_log_state(spa_t *spa, spa_log_state_t state); 756 extern int spa_offline_log(spa_t *spa); 757 758 /* Log claim callback */ 759 extern void spa_claim_notify(zio_t *zio); 760 761 /* Accessor functions */ 762 extern boolean_t spa_shutting_down(spa_t *spa); 763 extern struct dsl_pool *spa_get_dsl(spa_t *spa); 764 extern boolean_t spa_is_initializing(spa_t *spa); 765 extern blkptr_t *spa_get_rootblkptr(spa_t *spa); 766 extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp); 767 extern void spa_altroot(spa_t *, char *, size_t); 768 extern int spa_sync_pass(spa_t *spa); 769 extern char *spa_name(spa_t *spa); 770 extern uint64_t spa_guid(spa_t *spa); 771 extern uint64_t spa_load_guid(spa_t *spa); 772 extern uint64_t spa_last_synced_txg(spa_t *spa); 773 extern uint64_t spa_first_txg(spa_t *spa); 774 extern uint64_t spa_syncing_txg(spa_t *spa); 775 extern uint64_t spa_version(spa_t *spa); 776 extern pool_state_t spa_state(spa_t *spa); 777 extern spa_load_state_t spa_load_state(spa_t *spa); 778 extern uint64_t spa_freeze_txg(spa_t *spa); 779 extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize); 780 extern uint64_t spa_get_dspace(spa_t *spa); 781 extern uint64_t spa_get_slop_space(spa_t *spa); 782 extern void spa_update_dspace(spa_t *spa); 783 extern uint64_t spa_version(spa_t *spa); 784 extern boolean_t spa_deflate(spa_t *spa); 785 extern metaslab_class_t *spa_normal_class(spa_t *spa); 786 extern metaslab_class_t *spa_log_class(spa_t *spa); 787 extern void spa_evicting_os_register(spa_t *, objset_t *os); 788 extern void spa_evicting_os_deregister(spa_t *, objset_t *os); 789 extern void spa_evicting_os_wait(spa_t *spa); 790 extern int spa_max_replication(spa_t *spa); 791 extern int spa_prev_software_version(spa_t *spa); 792 extern int spa_busy(void); 793 extern uint8_t spa_get_failmode(spa_t *spa); 794 extern boolean_t spa_suspended(spa_t *spa); 795 extern uint64_t spa_bootfs(spa_t *spa); 796 extern uint64_t spa_delegation(spa_t *spa); 797 extern objset_t *spa_meta_objset(spa_t *spa); 798 extern uint64_t spa_deadman_synctime(spa_t *spa); 799 800 /* Miscellaneous support routines */ 801 extern void spa_activate_mos_feature(spa_t *spa, const char *feature, 802 dmu_tx_t *tx); 803 extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature); 804 extern int spa_rename(const char *oldname, const char *newname); 805 extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid); 806 extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid); 807 extern char *spa_strdup(const char *); 808 extern void spa_strfree(char *); 809 extern uint64_t spa_get_random(uint64_t range); 810 extern uint64_t spa_generate_guid(spa_t *spa); 811 extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp); 812 extern void spa_freeze(spa_t *spa); 813 extern int spa_change_guid(spa_t *spa); 814 extern void spa_upgrade(spa_t *spa, uint64_t version); 815 extern void spa_evict_all(void); 816 extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid, 817 boolean_t l2cache); 818 extern boolean_t spa_has_spare(spa_t *, uint64_t guid); 819 extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva); 820 extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp); 821 extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp); 822 extern boolean_t spa_has_slogs(spa_t *spa); 823 extern boolean_t spa_is_root(spa_t *spa); 824 extern boolean_t spa_writeable(spa_t *spa); 825 extern boolean_t spa_has_pending_synctask(spa_t *spa); 826 extern int spa_maxblocksize(spa_t *spa); 827 extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp); 828 829 extern int spa_mode(spa_t *spa); 830 extern uint64_t strtonum(const char *str, char **nptr); 831 832 extern char *spa_his_ievent_table[]; 833 834 extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx); 835 extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read, 836 char *his_buf); 837 extern int spa_history_log(spa_t *spa, const char *his_buf); 838 extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl); 839 extern void spa_history_log_version(spa_t *spa, const char *operation); 840 extern void spa_history_log_internal(spa_t *spa, const char *operation, 841 dmu_tx_t *tx, const char *fmt, ...); 842 extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op, 843 dmu_tx_t *tx, const char *fmt, ...); 844 extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation, 845 dmu_tx_t *tx, const char *fmt, ...); 846 847 /* error handling */ 848 struct zbookmark_phys; 849 extern void spa_log_error(spa_t *spa, zio_t *zio); 850 extern void zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd, 851 zio_t *zio, uint64_t stateoroffset, uint64_t length); 852 extern void zfs_post_remove(spa_t *spa, vdev_t *vd); 853 extern void zfs_post_state_change(spa_t *spa, vdev_t *vd); 854 extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd); 855 extern uint64_t spa_get_errlog_size(spa_t *spa); 856 extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count); 857 extern void spa_errlog_rotate(spa_t *spa); 858 extern void spa_errlog_drain(spa_t *spa); 859 extern void spa_errlog_sync(spa_t *spa, uint64_t txg); 860 extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub); 861 862 /* vdev cache */ 863 extern void vdev_cache_stat_init(void); 864 extern void vdev_cache_stat_fini(void); 865 866 /* Initialization and termination */ 867 extern void spa_init(int flags); 868 extern void spa_fini(void); 869 extern void spa_boot_init(); 870 871 /* properties */ 872 extern int spa_prop_set(spa_t *spa, nvlist_t *nvp); 873 extern int spa_prop_get(spa_t *spa, nvlist_t **nvp); 874 extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx); 875 extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t); 876 877 /* asynchronous event notification */ 878 extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name); 879 880 #ifdef ZFS_DEBUG 881 #define dprintf_bp(bp, fmt, ...) do { \ 882 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 883 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \ 884 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \ 885 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \ 886 kmem_free(__blkbuf, BP_SPRINTF_LEN); \ 887 } \ 888 _NOTE(CONSTCOND) } while (0) 889 #else 890 #define dprintf_bp(bp, fmt, ...) 891 #endif 892 893 extern boolean_t spa_debug_enabled(spa_t *spa); 894 #define spa_dbgmsg(spa, ...) \ 895 { \ 896 if (spa_debug_enabled(spa)) \ 897 zfs_dbgmsg(__VA_ARGS__); \ 898 } 899 900 extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */ 901 902 #ifdef __cplusplus 903 } 904 #endif 905 906 #endif /* _SYS_SPA_H */ 907