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, 2016 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 */ 29 30 #ifndef _SYS_SPA_H 31 #define _SYS_SPA_H 32 33 #include <sys/avl.h> 34 #include <sys/zfs_context.h> 35 #include <sys/nvpair.h> 36 #include <sys/sysmacros.h> 37 #include <sys/types.h> 38 #include <sys/fs/zfs.h> 39 #include <sys/dmu.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 #define SPA_COMPRESSBITS 7 138 139 /* 140 * All SPA data is represented by 128-bit data virtual addresses (DVAs). 141 * The members of the dva_t should be considered opaque outside the SPA. 142 */ 143 typedef struct dva { 144 uint64_t dva_word[2]; 145 } dva_t; 146 147 /* 148 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. 149 */ 150 typedef struct zio_cksum { 151 uint64_t zc_word[4]; 152 } zio_cksum_t; 153 154 /* 155 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept 156 * secret and is suitable for use in MAC algorithms as the key. 157 */ 158 typedef struct zio_cksum_salt { 159 uint8_t zcs_bytes[32]; 160 } zio_cksum_salt_t; 161 162 /* 163 * Each block is described by its DVAs, time of birth, checksum, etc. 164 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 165 * 166 * 64 56 48 40 32 24 16 8 0 167 * +-------+-------+-------+-------+-------+-------+-------+-------+ 168 * 0 | vdev1 | GRID | ASIZE | 169 * +-------+-------+-------+-------+-------+-------+-------+-------+ 170 * 1 |G| offset1 | 171 * +-------+-------+-------+-------+-------+-------+-------+-------+ 172 * 2 | vdev2 | GRID | ASIZE | 173 * +-------+-------+-------+-------+-------+-------+-------+-------+ 174 * 3 |G| offset2 | 175 * +-------+-------+-------+-------+-------+-------+-------+-------+ 176 * 4 | vdev3 | GRID | ASIZE | 177 * +-------+-------+-------+-------+-------+-------+-------+-------+ 178 * 5 |G| offset3 | 179 * +-------+-------+-------+-------+-------+-------+-------+-------+ 180 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 181 * +-------+-------+-------+-------+-------+-------+-------+-------+ 182 * 7 | padding | 183 * +-------+-------+-------+-------+-------+-------+-------+-------+ 184 * 8 | padding | 185 * +-------+-------+-------+-------+-------+-------+-------+-------+ 186 * 9 | physical birth txg | 187 * +-------+-------+-------+-------+-------+-------+-------+-------+ 188 * a | logical birth txg | 189 * +-------+-------+-------+-------+-------+-------+-------+-------+ 190 * b | fill count | 191 * +-------+-------+-------+-------+-------+-------+-------+-------+ 192 * c | checksum[0] | 193 * +-------+-------+-------+-------+-------+-------+-------+-------+ 194 * d | checksum[1] | 195 * +-------+-------+-------+-------+-------+-------+-------+-------+ 196 * e | checksum[2] | 197 * +-------+-------+-------+-------+-------+-------+-------+-------+ 198 * f | checksum[3] | 199 * +-------+-------+-------+-------+-------+-------+-------+-------+ 200 * 201 * Legend: 202 * 203 * vdev virtual device ID 204 * offset offset into virtual device 205 * LSIZE logical size 206 * PSIZE physical size (after compression) 207 * ASIZE allocated size (including RAID-Z parity and gang block headers) 208 * GRID RAID-Z layout information (reserved for future use) 209 * cksum checksum function 210 * comp compression function 211 * G gang block indicator 212 * B byteorder (endianness) 213 * D dedup 214 * X encryption (on version 30, which is not supported) 215 * E blkptr_t contains embedded data (see below) 216 * lvl level of indirection 217 * type DMU object type 218 * phys birth txg of block allocation; zero if same as logical birth txg 219 * log. birth transaction group in which the block was logically born 220 * fill count number of non-zero blocks under this bp 221 * checksum[4] 256-bit checksum of the data this bp describes 222 */ 223 224 /* 225 * "Embedded" blkptr_t's don't actually point to a block, instead they 226 * have a data payload embedded in the blkptr_t itself. See the comment 227 * in blkptr.c for more details. 228 * 229 * The blkptr_t is laid out as follows: 230 * 231 * 64 56 48 40 32 24 16 8 0 232 * +-------+-------+-------+-------+-------+-------+-------+-------+ 233 * 0 | payload | 234 * 1 | payload | 235 * 2 | payload | 236 * 3 | payload | 237 * 4 | payload | 238 * 5 | payload | 239 * +-------+-------+-------+-------+-------+-------+-------+-------+ 240 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 241 * +-------+-------+-------+-------+-------+-------+-------+-------+ 242 * 7 | payload | 243 * 8 | payload | 244 * 9 | payload | 245 * +-------+-------+-------+-------+-------+-------+-------+-------+ 246 * a | logical birth txg | 247 * +-------+-------+-------+-------+-------+-------+-------+-------+ 248 * b | payload | 249 * c | payload | 250 * d | payload | 251 * e | payload | 252 * f | payload | 253 * +-------+-------+-------+-------+-------+-------+-------+-------+ 254 * 255 * Legend: 256 * 257 * payload contains the embedded data 258 * B (byteorder) byteorder (endianness) 259 * D (dedup) padding (set to zero) 260 * X encryption (set to zero; see above) 261 * E (embedded) set to one 262 * lvl indirection level 263 * type DMU object type 264 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 265 * comp compression function of payload 266 * PSIZE size of payload after compression, in bytes 267 * LSIZE logical size of payload, in bytes 268 * note that 25 bits is enough to store the largest 269 * "normal" BP's LSIZE (2^16 * 2^9) in bytes 270 * log. birth transaction group in which the block was logically born 271 * 272 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 273 * bp's they are stored in units of SPA_MINBLOCKSHIFT. 274 * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 275 * The B, D, X, lvl, type, and comp fields are stored the same as with normal 276 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 277 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 278 * other macros, as they assert that they are only used on BP's of the correct 279 * "embedded-ness". 280 */ 281 282 #define BPE_GET_ETYPE(bp) \ 283 (ASSERT(BP_IS_EMBEDDED(bp)), \ 284 BF64_GET((bp)->blk_prop, 40, 8)) 285 #define BPE_SET_ETYPE(bp, t) do { \ 286 ASSERT(BP_IS_EMBEDDED(bp)); \ 287 BF64_SET((bp)->blk_prop, 40, 8, t); \ 288 _NOTE(CONSTCOND) } while (0) 289 290 #define BPE_GET_LSIZE(bp) \ 291 (ASSERT(BP_IS_EMBEDDED(bp)), \ 292 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 293 #define BPE_SET_LSIZE(bp, x) do { \ 294 ASSERT(BP_IS_EMBEDDED(bp)); \ 295 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 296 _NOTE(CONSTCOND) } while (0) 297 298 #define BPE_GET_PSIZE(bp) \ 299 (ASSERT(BP_IS_EMBEDDED(bp)), \ 300 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 301 #define BPE_SET_PSIZE(bp, x) do { \ 302 ASSERT(BP_IS_EMBEDDED(bp)); \ 303 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 304 _NOTE(CONSTCOND) } while (0) 305 306 typedef enum bp_embedded_type { 307 BP_EMBEDDED_TYPE_DATA, 308 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ 309 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED 310 } bp_embedded_type_t; 311 312 #define BPE_NUM_WORDS 14 313 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 314 #define BPE_IS_PAYLOADWORD(bp, wp) \ 315 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 316 317 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 318 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 319 320 /* 321 * A block is a hole when it has either 1) never been written to, or 322 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads 323 * without physically allocating disk space. Holes are represented in the 324 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is 325 * done through the BP_IS_HOLE macro. For holes, the logical size, level, 326 * DMU object type, and birth times are all also stored for holes that 327 * were written to at some point (i.e. were punched after having been filled). 328 */ 329 typedef struct blkptr { 330 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 331 uint64_t blk_prop; /* size, compression, type, etc */ 332 uint64_t blk_pad[2]; /* Extra space for the future */ 333 uint64_t blk_phys_birth; /* txg when block was allocated */ 334 uint64_t blk_birth; /* transaction group at birth */ 335 uint64_t blk_fill; /* fill count */ 336 zio_cksum_t blk_cksum; /* 256-bit checksum */ 337 } blkptr_t; 338 339 /* 340 * Macros to get and set fields in a bp or DVA. 341 */ 342 #define DVA_GET_ASIZE(dva) \ 343 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 344 #define DVA_SET_ASIZE(dva, x) \ 345 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 346 SPA_MINBLOCKSHIFT, 0, x) 347 348 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 349 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 350 351 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) 352 #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) 353 354 #define DVA_GET_OFFSET(dva) \ 355 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 356 #define DVA_SET_OFFSET(dva, x) \ 357 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 358 359 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 360 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 361 362 #define BP_GET_LSIZE(bp) \ 363 (BP_IS_EMBEDDED(bp) ? \ 364 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 365 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 366 #define BP_SET_LSIZE(bp, x) do { \ 367 ASSERT(!BP_IS_EMBEDDED(bp)); \ 368 BF64_SET_SB((bp)->blk_prop, \ 369 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 370 _NOTE(CONSTCOND) } while (0) 371 372 #define BP_GET_PSIZE(bp) \ 373 (BP_IS_EMBEDDED(bp) ? 0 : \ 374 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 375 #define BP_SET_PSIZE(bp, x) do { \ 376 ASSERT(!BP_IS_EMBEDDED(bp)); \ 377 BF64_SET_SB((bp)->blk_prop, \ 378 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 379 _NOTE(CONSTCOND) } while (0) 380 381 #define BP_GET_COMPRESS(bp) \ 382 BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS) 383 #define BP_SET_COMPRESS(bp, x) \ 384 BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x) 385 386 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 387 #define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x) 388 389 #define BP_GET_CHECKSUM(bp) \ 390 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \ 391 BF64_GET((bp)->blk_prop, 40, 8)) 392 #define BP_SET_CHECKSUM(bp, x) do { \ 393 ASSERT(!BP_IS_EMBEDDED(bp)); \ 394 BF64_SET((bp)->blk_prop, 40, 8, x); \ 395 _NOTE(CONSTCOND) } while (0) 396 397 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 398 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 399 400 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 401 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 402 403 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 404 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 405 406 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 407 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 408 409 #define BP_PHYSICAL_BIRTH(bp) \ 410 (BP_IS_EMBEDDED(bp) ? 0 : \ 411 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 412 413 #define BP_SET_BIRTH(bp, logical, physical) \ 414 { \ 415 ASSERT(!BP_IS_EMBEDDED(bp)); \ 416 (bp)->blk_birth = (logical); \ 417 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \ 418 } 419 420 #define BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill) 421 422 #define BP_GET_ASIZE(bp) \ 423 (BP_IS_EMBEDDED(bp) ? 0 : \ 424 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 425 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 426 DVA_GET_ASIZE(&(bp)->blk_dva[2])) 427 428 #define BP_GET_UCSIZE(bp) \ 429 ((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \ 430 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)) 431 432 #define BP_GET_NDVAS(bp) \ 433 (BP_IS_EMBEDDED(bp) ? 0 : \ 434 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 435 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 436 !!DVA_GET_ASIZE(&(bp)->blk_dva[2])) 437 438 #define BP_COUNT_GANG(bp) \ 439 (BP_IS_EMBEDDED(bp) ? 0 : \ 440 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \ 441 DVA_GET_GANG(&(bp)->blk_dva[1]) + \ 442 DVA_GET_GANG(&(bp)->blk_dva[2]))) 443 444 #define DVA_EQUAL(dva1, dva2) \ 445 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 446 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 447 448 #define BP_EQUAL(bp1, bp2) \ 449 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \ 450 (bp1)->blk_birth == (bp2)->blk_birth && \ 451 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \ 452 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \ 453 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2])) 454 455 #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 456 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 457 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 458 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 459 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 460 461 #define ZIO_CHECKSUM_IS_ZERO(zc) \ 462 (0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \ 463 (zc)->zc_word[2] | (zc)->zc_word[3])) 464 465 #define ZIO_CHECKSUM_BSWAP(zcp) \ 466 { \ 467 (zcp)->zc_word[0] = BSWAP_64((zcp)->zc_word[0]); \ 468 (zcp)->zc_word[1] = BSWAP_64((zcp)->zc_word[1]); \ 469 (zcp)->zc_word[2] = BSWAP_64((zcp)->zc_word[2]); \ 470 (zcp)->zc_word[3] = BSWAP_64((zcp)->zc_word[3]); \ 471 } 472 473 474 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 475 476 #define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ 477 { \ 478 (zcp)->zc_word[0] = w0; \ 479 (zcp)->zc_word[1] = w1; \ 480 (zcp)->zc_word[2] = w2; \ 481 (zcp)->zc_word[3] = w3; \ 482 } 483 484 #define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0]) 485 #define BP_IS_GANG(bp) \ 486 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp))) 487 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 488 (dva)->dva_word[1] == 0ULL) 489 #define BP_IS_HOLE(bp) \ 490 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp))) 491 492 /* BP_IS_RAIDZ(bp) assumes no block compression */ 493 #define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \ 494 BP_GET_PSIZE(bp)) 495 496 #define BP_ZERO(bp) \ 497 { \ 498 (bp)->blk_dva[0].dva_word[0] = 0; \ 499 (bp)->blk_dva[0].dva_word[1] = 0; \ 500 (bp)->blk_dva[1].dva_word[0] = 0; \ 501 (bp)->blk_dva[1].dva_word[1] = 0; \ 502 (bp)->blk_dva[2].dva_word[0] = 0; \ 503 (bp)->blk_dva[2].dva_word[1] = 0; \ 504 (bp)->blk_prop = 0; \ 505 (bp)->blk_pad[0] = 0; \ 506 (bp)->blk_pad[1] = 0; \ 507 (bp)->blk_phys_birth = 0; \ 508 (bp)->blk_birth = 0; \ 509 (bp)->blk_fill = 0; \ 510 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 511 } 512 513 #ifdef _BIG_ENDIAN 514 #define ZFS_HOST_BYTEORDER (0ULL) 515 #else 516 #define ZFS_HOST_BYTEORDER (1ULL) 517 #endif 518 519 #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) 520 521 #define BP_SPRINTF_LEN 320 522 523 /* 524 * This macro allows code sharing between zfs, libzpool, and mdb. 525 * 'func' is either snprintf() or mdb_snprintf(). 526 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line. 527 */ 528 #define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \ 529 { \ 530 static const char *copyname[] = \ 531 { "zero", "single", "double", "triple" }; \ 532 int len = 0; \ 533 int copies = 0; \ 534 \ 535 if (bp == NULL) { \ 536 len += func(buf + len, size - len, "<NULL>"); \ 537 } else if (BP_IS_HOLE(bp)) { \ 538 len += func(buf + len, size - len, \ 539 "HOLE [L%llu %s] " \ 540 "size=%llxL birth=%lluL", \ 541 (u_longlong_t)BP_GET_LEVEL(bp), \ 542 type, \ 543 (u_longlong_t)BP_GET_LSIZE(bp), \ 544 (u_longlong_t)bp->blk_birth); \ 545 } else if (BP_IS_EMBEDDED(bp)) { \ 546 len = func(buf + len, size - len, \ 547 "EMBEDDED [L%llu %s] et=%u %s " \ 548 "size=%llxL/%llxP birth=%lluL", \ 549 (u_longlong_t)BP_GET_LEVEL(bp), \ 550 type, \ 551 (int)BPE_GET_ETYPE(bp), \ 552 compress, \ 553 (u_longlong_t)BPE_GET_LSIZE(bp), \ 554 (u_longlong_t)BPE_GET_PSIZE(bp), \ 555 (u_longlong_t)bp->blk_birth); \ 556 } else { \ 557 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \ 558 const dva_t *dva = &bp->blk_dva[d]; \ 559 if (DVA_IS_VALID(dva)) \ 560 copies++; \ 561 len += func(buf + len, size - len, \ 562 "DVA[%d]=<%llu:%llx:%llx>%c", d, \ 563 (u_longlong_t)DVA_GET_VDEV(dva), \ 564 (u_longlong_t)DVA_GET_OFFSET(dva), \ 565 (u_longlong_t)DVA_GET_ASIZE(dva), \ 566 ws); \ 567 } \ 568 if (BP_IS_GANG(bp) && \ 569 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \ 570 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \ 571 copies--; \ 572 len += func(buf + len, size - len, \ 573 "[L%llu %s] %s %s %s %s %s %s%c" \ 574 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \ 575 "cksum=%llx:%llx:%llx:%llx", \ 576 (u_longlong_t)BP_GET_LEVEL(bp), \ 577 type, \ 578 checksum, \ 579 compress, \ 580 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \ 581 BP_IS_GANG(bp) ? "gang" : "contiguous", \ 582 BP_GET_DEDUP(bp) ? "dedup" : "unique", \ 583 copyname[copies], \ 584 ws, \ 585 (u_longlong_t)BP_GET_LSIZE(bp), \ 586 (u_longlong_t)BP_GET_PSIZE(bp), \ 587 (u_longlong_t)bp->blk_birth, \ 588 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \ 589 (u_longlong_t)BP_GET_FILL(bp), \ 590 ws, \ 591 (u_longlong_t)bp->blk_cksum.zc_word[0], \ 592 (u_longlong_t)bp->blk_cksum.zc_word[1], \ 593 (u_longlong_t)bp->blk_cksum.zc_word[2], \ 594 (u_longlong_t)bp->blk_cksum.zc_word[3]); \ 595 } \ 596 ASSERT(len < size); \ 597 } 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 641 642 /* 643 * Controls the behavior of spa_vdev_remove(). 644 */ 645 #define SPA_REMOVE_UNSPARE 0x01 646 #define SPA_REMOVE_DONE 0x02 647 648 /* device manipulation */ 649 extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot); 650 extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, 651 int replacing); 652 extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, 653 int replace_done); 654 extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare); 655 extern boolean_t spa_vdev_remove_active(spa_t *spa); 656 extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath); 657 extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru); 658 extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config, 659 nvlist_t *props, boolean_t exp); 660 661 /* spare state (which is global across all pools) */ 662 extern void spa_spare_add(vdev_t *vd); 663 extern void spa_spare_remove(vdev_t *vd); 664 extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt); 665 extern void spa_spare_activate(vdev_t *vd); 666 667 /* L2ARC state (which is global across all pools) */ 668 extern void spa_l2cache_add(vdev_t *vd); 669 extern void spa_l2cache_remove(vdev_t *vd); 670 extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool); 671 extern void spa_l2cache_activate(vdev_t *vd); 672 extern void spa_l2cache_drop(spa_t *spa); 673 674 /* scanning */ 675 extern int spa_scan(spa_t *spa, pool_scan_func_t func); 676 extern int spa_scan_stop(spa_t *spa); 677 678 /* spa syncing */ 679 extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */ 680 extern void spa_sync_allpools(void); 681 682 /* spa namespace global mutex */ 683 extern kmutex_t spa_namespace_lock; 684 685 /* 686 * SPA configuration functions in spa_config.c 687 */ 688 689 #define SPA_CONFIG_UPDATE_POOL 0 690 #define SPA_CONFIG_UPDATE_VDEVS 1 691 692 extern void spa_config_sync(spa_t *, boolean_t, boolean_t); 693 extern void spa_config_load(void); 694 extern nvlist_t *spa_all_configs(uint64_t *); 695 extern void spa_config_set(spa_t *spa, nvlist_t *config); 696 extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, 697 int getstats); 698 extern void spa_config_update(spa_t *spa, int what); 699 700 /* 701 * Miscellaneous SPA routines in spa_misc.c 702 */ 703 704 /* Namespace manipulation */ 705 extern spa_t *spa_lookup(const char *name); 706 extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot); 707 extern void spa_remove(spa_t *spa); 708 extern spa_t *spa_next(spa_t *prev); 709 710 /* Refcount functions */ 711 extern void spa_open_ref(spa_t *spa, void *tag); 712 extern void spa_close(spa_t *spa, void *tag); 713 extern void spa_async_close(spa_t *spa, void *tag); 714 extern boolean_t spa_refcount_zero(spa_t *spa); 715 716 #define SCL_NONE 0x00 717 #define SCL_CONFIG 0x01 718 #define SCL_STATE 0x02 719 #define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */ 720 #define SCL_ALLOC 0x08 721 #define SCL_ZIO 0x10 722 #define SCL_FREE 0x20 723 #define SCL_VDEV 0x40 724 #define SCL_LOCKS 7 725 #define SCL_ALL ((1 << SCL_LOCKS) - 1) 726 #define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO) 727 728 /* Pool configuration locks */ 729 extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw); 730 extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw); 731 extern void spa_config_exit(spa_t *spa, int locks, void *tag); 732 extern int spa_config_held(spa_t *spa, int locks, krw_t rw); 733 734 /* Pool vdev add/remove lock */ 735 extern uint64_t spa_vdev_enter(spa_t *spa); 736 extern uint64_t spa_vdev_config_enter(spa_t *spa); 737 extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, 738 int error, char *tag); 739 extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error); 740 741 /* Pool vdev state change lock */ 742 extern void spa_vdev_state_enter(spa_t *spa, int oplock); 743 extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error); 744 745 /* Log state */ 746 typedef enum spa_log_state { 747 SPA_LOG_UNKNOWN = 0, /* unknown log state */ 748 SPA_LOG_MISSING, /* missing log(s) */ 749 SPA_LOG_CLEAR, /* clear the log(s) */ 750 SPA_LOG_GOOD, /* log(s) are good */ 751 } spa_log_state_t; 752 753 extern spa_log_state_t spa_get_log_state(spa_t *spa); 754 extern void spa_set_log_state(spa_t *spa, spa_log_state_t state); 755 extern int spa_offline_log(spa_t *spa); 756 757 /* Log claim callback */ 758 extern void spa_claim_notify(zio_t *zio); 759 760 /* Accessor functions */ 761 extern boolean_t spa_shutting_down(spa_t *spa); 762 extern struct dsl_pool *spa_get_dsl(spa_t *spa); 763 extern boolean_t spa_is_initializing(spa_t *spa); 764 extern blkptr_t *spa_get_rootblkptr(spa_t *spa); 765 extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp); 766 extern void spa_altroot(spa_t *, char *, size_t); 767 extern int spa_sync_pass(spa_t *spa); 768 extern char *spa_name(spa_t *spa); 769 extern uint64_t spa_guid(spa_t *spa); 770 extern uint64_t spa_load_guid(spa_t *spa); 771 extern uint64_t spa_last_synced_txg(spa_t *spa); 772 extern uint64_t spa_first_txg(spa_t *spa); 773 extern uint64_t spa_syncing_txg(spa_t *spa); 774 extern uint64_t spa_version(spa_t *spa); 775 extern pool_state_t spa_state(spa_t *spa); 776 extern spa_load_state_t spa_load_state(spa_t *spa); 777 extern uint64_t spa_freeze_txg(spa_t *spa); 778 extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize); 779 extern uint64_t spa_get_dspace(spa_t *spa); 780 extern uint64_t spa_get_slop_space(spa_t *spa); 781 extern void spa_update_dspace(spa_t *spa); 782 extern uint64_t spa_version(spa_t *spa); 783 extern boolean_t spa_deflate(spa_t *spa); 784 extern metaslab_class_t *spa_normal_class(spa_t *spa); 785 extern metaslab_class_t *spa_log_class(spa_t *spa); 786 extern void spa_evicting_os_register(spa_t *, objset_t *os); 787 extern void spa_evicting_os_deregister(spa_t *, objset_t *os); 788 extern void spa_evicting_os_wait(spa_t *spa); 789 extern int spa_max_replication(spa_t *spa); 790 extern int spa_prev_software_version(spa_t *spa); 791 extern int spa_busy(void); 792 extern uint8_t spa_get_failmode(spa_t *spa); 793 extern boolean_t spa_suspended(spa_t *spa); 794 extern uint64_t spa_bootfs(spa_t *spa); 795 extern uint64_t spa_delegation(spa_t *spa); 796 extern objset_t *spa_meta_objset(spa_t *spa); 797 extern uint64_t spa_deadman_synctime(spa_t *spa); 798 799 /* Miscellaneous support routines */ 800 extern void spa_activate_mos_feature(spa_t *spa, const char *feature, 801 dmu_tx_t *tx); 802 extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature); 803 extern int spa_rename(const char *oldname, const char *newname); 804 extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid); 805 extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid); 806 extern char *spa_strdup(const char *); 807 extern void spa_strfree(char *); 808 extern uint64_t spa_get_random(uint64_t range); 809 extern uint64_t spa_generate_guid(spa_t *spa); 810 extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp); 811 extern void spa_freeze(spa_t *spa); 812 extern int spa_change_guid(spa_t *spa); 813 extern void spa_upgrade(spa_t *spa, uint64_t version); 814 extern void spa_evict_all(void); 815 extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid, 816 boolean_t l2cache); 817 extern boolean_t spa_has_spare(spa_t *, uint64_t guid); 818 extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva); 819 extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp); 820 extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp); 821 extern boolean_t spa_has_slogs(spa_t *spa); 822 extern boolean_t spa_is_root(spa_t *spa); 823 extern boolean_t spa_writeable(spa_t *spa); 824 extern boolean_t spa_has_pending_synctask(spa_t *spa); 825 extern int spa_maxblocksize(spa_t *spa); 826 extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp); 827 828 extern int spa_mode(spa_t *spa); 829 extern uint64_t strtonum(const char *str, char **nptr); 830 831 extern char *spa_his_ievent_table[]; 832 833 extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx); 834 extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read, 835 char *his_buf); 836 extern int spa_history_log(spa_t *spa, const char *his_buf); 837 extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl); 838 extern void spa_history_log_version(spa_t *spa, const char *operation); 839 extern void spa_history_log_internal(spa_t *spa, const char *operation, 840 dmu_tx_t *tx, const char *fmt, ...); 841 extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op, 842 dmu_tx_t *tx, const char *fmt, ...); 843 extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation, 844 dmu_tx_t *tx, const char *fmt, ...); 845 846 /* error handling */ 847 struct zbookmark_phys; 848 extern void spa_log_error(spa_t *spa, zio_t *zio); 849 extern void zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd, 850 zio_t *zio, uint64_t stateoroffset, uint64_t length); 851 extern void zfs_post_remove(spa_t *spa, vdev_t *vd); 852 extern void zfs_post_state_change(spa_t *spa, vdev_t *vd); 853 extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd); 854 extern uint64_t spa_get_errlog_size(spa_t *spa); 855 extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count); 856 extern void spa_errlog_rotate(spa_t *spa); 857 extern void spa_errlog_drain(spa_t *spa); 858 extern void spa_errlog_sync(spa_t *spa, uint64_t txg); 859 extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub); 860 861 /* vdev cache */ 862 extern void vdev_cache_stat_init(void); 863 extern void vdev_cache_stat_fini(void); 864 865 /* Initialization and termination */ 866 extern void spa_init(int flags); 867 extern void spa_fini(void); 868 extern void spa_boot_init(); 869 870 /* properties */ 871 extern int spa_prop_set(spa_t *spa, nvlist_t *nvp); 872 extern int spa_prop_get(spa_t *spa, nvlist_t **nvp); 873 extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx); 874 extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t); 875 876 /* asynchronous event notification */ 877 extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name); 878 879 #ifdef ZFS_DEBUG 880 #define dprintf_bp(bp, fmt, ...) do { \ 881 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 882 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \ 883 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \ 884 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \ 885 kmem_free(__blkbuf, BP_SPRINTF_LEN); \ 886 } \ 887 _NOTE(CONSTCOND) } while (0) 888 #else 889 #define dprintf_bp(bp, fmt, ...) 890 #endif 891 892 extern boolean_t spa_debug_enabled(spa_t *spa); 893 #define spa_dbgmsg(spa, ...) \ 894 { \ 895 if (spa_debug_enabled(spa)) \ 896 zfs_dbgmsg(__VA_ARGS__); \ 897 } 898 899 extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */ 900 901 #ifdef __cplusplus 902 } 903 #endif 904 905 #endif /* _SYS_SPA_H */ 906