spa.h revision 297112
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, 2014 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 40#ifdef __cplusplus 41extern "C" { 42#endif 43 44/* 45 * Forward references that lots of things need. 46 */ 47typedef struct spa spa_t; 48typedef struct vdev vdev_t; 49typedef struct metaslab metaslab_t; 50typedef struct metaslab_group metaslab_group_t; 51typedef struct metaslab_class metaslab_class_t; 52typedef struct zio zio_t; 53typedef struct zilog zilog_t; 54typedef struct spa_aux_vdev spa_aux_vdev_t; 55typedef struct ddt ddt_t; 56typedef struct ddt_entry ddt_entry_t; 57struct dsl_pool; 58struct dsl_dataset; 59 60/* 61 * General-purpose 32-bit and 64-bit bitfield encodings. 62 */ 63#define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) 64#define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) 65#define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) 66#define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) 67 68#define BF32_GET(x, low, len) BF32_DECODE(x, low, len) 69#define BF64_GET(x, low, len) BF64_DECODE(x, low, len) 70 71#define BF32_SET(x, low, len, val) do { \ 72 ASSERT3U(val, <, 1U << (len)); \ 73 ASSERT3U(low + len, <=, 32); \ 74 (x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \ 75_NOTE(CONSTCOND) } while (0) 76 77#define BF64_SET(x, low, len, val) do { \ 78 ASSERT3U(val, <, 1ULL << (len)); \ 79 ASSERT3U(low + len, <=, 64); \ 80 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \ 81_NOTE(CONSTCOND) } while (0) 82 83#define BF32_GET_SB(x, low, len, shift, bias) \ 84 ((BF32_GET(x, low, len) + (bias)) << (shift)) 85#define BF64_GET_SB(x, low, len, shift, bias) \ 86 ((BF64_GET(x, low, len) + (bias)) << (shift)) 87 88#define BF32_SET_SB(x, low, len, shift, bias, val) do { \ 89 ASSERT(IS_P2ALIGNED(val, 1U << shift)); \ 90 ASSERT3S((val) >> (shift), >=, bias); \ 91 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \ 92_NOTE(CONSTCOND) } while (0) 93#define BF64_SET_SB(x, low, len, shift, bias, val) do { \ 94 ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \ 95 ASSERT3S((val) >> (shift), >=, bias); \ 96 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \ 97_NOTE(CONSTCOND) } while (0) 98 99/* 100 * We currently support block sizes from 512 bytes to 16MB. 101 * The benefits of larger blocks, and thus larger IO, need to be weighed 102 * against the cost of COWing a giant block to modify one byte, and the 103 * large latency of reading or writing a large block. 104 * 105 * Note that although blocks up to 16MB are supported, the recordsize 106 * property can not be set larger than zfs_max_recordsize (default 1MB). 107 * See the comment near zfs_max_recordsize in dsl_dataset.c for details. 108 * 109 * Note that although the LSIZE field of the blkptr_t can store sizes up 110 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to 111 * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB. 112 */ 113#define SPA_MINBLOCKSHIFT 9 114#define SPA_OLD_MAXBLOCKSHIFT 17 115#define SPA_MAXBLOCKSHIFT 24 116#define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 117#define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT) 118#define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 119 120/* 121 * Default maximum supported logical ashift. 122 * 123 * The current 8k allocation block size limit is due to the 8k 124 * aligned/sized operations performed by vdev_probe() on 125 * vdev_label->vl_pad2. Using another "safe region" for these tests 126 * would allow the limit to be raised to 16k, at the expense of 127 * only having 8 available uberblocks in the label area. 128 */ 129#define SPA_MAXASHIFT 13 130 131/* 132 * Default minimum supported logical ashift. 133 */ 134#define SPA_MINASHIFT SPA_MINBLOCKSHIFT 135 136/* 137 * Size of block to hold the configuration data (a packed nvlist) 138 */ 139#define SPA_CONFIG_BLOCKSIZE (1ULL << 14) 140 141/* 142 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. 143 * The ASIZE encoding should be at least 64 times larger (6 more bits) 144 * to support up to 4-way RAID-Z mirror mode with worst-case gang block 145 * overhead, three DVAs per bp, plus one more bit in case we do anything 146 * else that expands the ASIZE. 147 */ 148#define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ 149#define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ 150#define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ 151 152/* 153 * All SPA data is represented by 128-bit data virtual addresses (DVAs). 154 * The members of the dva_t should be considered opaque outside the SPA. 155 */ 156typedef struct dva { 157 uint64_t dva_word[2]; 158} dva_t; 159 160/* 161 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. 162 */ 163typedef struct zio_cksum { 164 uint64_t zc_word[4]; 165} zio_cksum_t; 166 167/* 168 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept 169 * secret and is suitable for use in MAC algorithms as the key. 170 */ 171typedef struct zio_cksum_salt { 172 uint8_t zcs_bytes[32]; 173} zio_cksum_salt_t; 174 175/* 176 * Each block is described by its DVAs, time of birth, checksum, etc. 177 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 178 * 179 * 64 56 48 40 32 24 16 8 0 180 * +-------+-------+-------+-------+-------+-------+-------+-------+ 181 * 0 | vdev1 | GRID | ASIZE | 182 * +-------+-------+-------+-------+-------+-------+-------+-------+ 183 * 1 |G| offset1 | 184 * +-------+-------+-------+-------+-------+-------+-------+-------+ 185 * 2 | vdev2 | GRID | ASIZE | 186 * +-------+-------+-------+-------+-------+-------+-------+-------+ 187 * 3 |G| offset2 | 188 * +-------+-------+-------+-------+-------+-------+-------+-------+ 189 * 4 | vdev3 | GRID | ASIZE | 190 * +-------+-------+-------+-------+-------+-------+-------+-------+ 191 * 5 |G| offset3 | 192 * +-------+-------+-------+-------+-------+-------+-------+-------+ 193 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 194 * +-------+-------+-------+-------+-------+-------+-------+-------+ 195 * 7 | padding | 196 * +-------+-------+-------+-------+-------+-------+-------+-------+ 197 * 8 | padding | 198 * +-------+-------+-------+-------+-------+-------+-------+-------+ 199 * 9 | physical birth txg | 200 * +-------+-------+-------+-------+-------+-------+-------+-------+ 201 * a | logical birth txg | 202 * +-------+-------+-------+-------+-------+-------+-------+-------+ 203 * b | fill count | 204 * +-------+-------+-------+-------+-------+-------+-------+-------+ 205 * c | checksum[0] | 206 * +-------+-------+-------+-------+-------+-------+-------+-------+ 207 * d | checksum[1] | 208 * +-------+-------+-------+-------+-------+-------+-------+-------+ 209 * e | checksum[2] | 210 * +-------+-------+-------+-------+-------+-------+-------+-------+ 211 * f | checksum[3] | 212 * +-------+-------+-------+-------+-------+-------+-------+-------+ 213 * 214 * Legend: 215 * 216 * vdev virtual device ID 217 * offset offset into virtual device 218 * LSIZE logical size 219 * PSIZE physical size (after compression) 220 * ASIZE allocated size (including RAID-Z parity and gang block headers) 221 * GRID RAID-Z layout information (reserved for future use) 222 * cksum checksum function 223 * comp compression function 224 * G gang block indicator 225 * B byteorder (endianness) 226 * D dedup 227 * X encryption (on version 30, which is not supported) 228 * E blkptr_t contains embedded data (see below) 229 * lvl level of indirection 230 * type DMU object type 231 * phys birth txg of block allocation; zero if same as logical birth txg 232 * log. birth transaction group in which the block was logically born 233 * fill count number of non-zero blocks under this bp 234 * checksum[4] 256-bit checksum of the data this bp describes 235 */ 236 237/* 238 * "Embedded" blkptr_t's don't actually point to a block, instead they 239 * have a data payload embedded in the blkptr_t itself. See the comment 240 * in blkptr.c for more details. 241 * 242 * The blkptr_t is laid out as follows: 243 * 244 * 64 56 48 40 32 24 16 8 0 245 * +-------+-------+-------+-------+-------+-------+-------+-------+ 246 * 0 | payload | 247 * 1 | payload | 248 * 2 | payload | 249 * 3 | payload | 250 * 4 | payload | 251 * 5 | payload | 252 * +-------+-------+-------+-------+-------+-------+-------+-------+ 253 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 254 * +-------+-------+-------+-------+-------+-------+-------+-------+ 255 * 7 | payload | 256 * 8 | payload | 257 * 9 | payload | 258 * +-------+-------+-------+-------+-------+-------+-------+-------+ 259 * a | logical birth txg | 260 * +-------+-------+-------+-------+-------+-------+-------+-------+ 261 * b | payload | 262 * c | payload | 263 * d | payload | 264 * e | payload | 265 * f | payload | 266 * +-------+-------+-------+-------+-------+-------+-------+-------+ 267 * 268 * Legend: 269 * 270 * payload contains the embedded data 271 * B (byteorder) byteorder (endianness) 272 * D (dedup) padding (set to zero) 273 * X encryption (set to zero; see above) 274 * E (embedded) set to one 275 * lvl indirection level 276 * type DMU object type 277 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 278 * comp compression function of payload 279 * PSIZE size of payload after compression, in bytes 280 * LSIZE logical size of payload, in bytes 281 * note that 25 bits is enough to store the largest 282 * "normal" BP's LSIZE (2^16 * 2^9) in bytes 283 * log. birth transaction group in which the block was logically born 284 * 285 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 286 * bp's they are stored in units of SPA_MINBLOCKSHIFT. 287 * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 288 * The B, D, X, lvl, type, and comp fields are stored the same as with normal 289 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 290 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 291 * other macros, as they assert that they are only used on BP's of the correct 292 * "embedded-ness". 293 */ 294 295#define BPE_GET_ETYPE(bp) \ 296 (ASSERT(BP_IS_EMBEDDED(bp)), \ 297 BF64_GET((bp)->blk_prop, 40, 8)) 298#define BPE_SET_ETYPE(bp, t) do { \ 299 ASSERT(BP_IS_EMBEDDED(bp)); \ 300 BF64_SET((bp)->blk_prop, 40, 8, t); \ 301_NOTE(CONSTCOND) } while (0) 302 303#define BPE_GET_LSIZE(bp) \ 304 (ASSERT(BP_IS_EMBEDDED(bp)), \ 305 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 306#define BPE_SET_LSIZE(bp, x) do { \ 307 ASSERT(BP_IS_EMBEDDED(bp)); \ 308 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 309_NOTE(CONSTCOND) } while (0) 310 311#define BPE_GET_PSIZE(bp) \ 312 (ASSERT(BP_IS_EMBEDDED(bp)), \ 313 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 314#define BPE_SET_PSIZE(bp, x) do { \ 315 ASSERT(BP_IS_EMBEDDED(bp)); \ 316 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 317_NOTE(CONSTCOND) } while (0) 318 319typedef enum bp_embedded_type { 320 BP_EMBEDDED_TYPE_DATA, 321 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ 322 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED 323} bp_embedded_type_t; 324 325#define BPE_NUM_WORDS 14 326#define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 327#define BPE_IS_PAYLOADWORD(bp, wp) \ 328 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 329 330#define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 331#define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 332 333/* 334 * A block is a hole when it has either 1) never been written to, or 335 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads 336 * without physically allocating disk space. Holes are represented in the 337 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is 338 * done through the BP_IS_HOLE macro. For holes, the logical size, level, 339 * DMU object type, and birth times are all also stored for holes that 340 * were written to at some point (i.e. were punched after having been filled). 341 */ 342typedef struct blkptr { 343 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 344 uint64_t blk_prop; /* size, compression, type, etc */ 345 uint64_t blk_pad[2]; /* Extra space for the future */ 346 uint64_t blk_phys_birth; /* txg when block was allocated */ 347 uint64_t blk_birth; /* transaction group at birth */ 348 uint64_t blk_fill; /* fill count */ 349 zio_cksum_t blk_cksum; /* 256-bit checksum */ 350} blkptr_t; 351 352/* 353 * Macros to get and set fields in a bp or DVA. 354 */ 355#define DVA_GET_ASIZE(dva) \ 356 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 357#define DVA_SET_ASIZE(dva, x) \ 358 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 359 SPA_MINBLOCKSHIFT, 0, x) 360 361#define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 362#define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 363 364#define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) 365#define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) 366 367#define DVA_GET_OFFSET(dva) \ 368 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 369#define DVA_SET_OFFSET(dva, x) \ 370 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 371 372#define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 373#define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 374 375#define BP_GET_LSIZE(bp) \ 376 (BP_IS_EMBEDDED(bp) ? \ 377 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 378 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 379#define BP_SET_LSIZE(bp, x) do { \ 380 ASSERT(!BP_IS_EMBEDDED(bp)); \ 381 BF64_SET_SB((bp)->blk_prop, \ 382 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 383_NOTE(CONSTCOND) } while (0) 384 385#define BP_GET_PSIZE(bp) \ 386 (BP_IS_EMBEDDED(bp) ? 0 : \ 387 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 388#define BP_SET_PSIZE(bp, x) do { \ 389 ASSERT(!BP_IS_EMBEDDED(bp)); \ 390 BF64_SET_SB((bp)->blk_prop, \ 391 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 392_NOTE(CONSTCOND) } while (0) 393 394#define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7) 395#define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x) 396 397#define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 398#define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x) 399 400#define BP_GET_CHECKSUM(bp) \ 401 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \ 402 BF64_GET((bp)->blk_prop, 40, 8)) 403#define BP_SET_CHECKSUM(bp, x) do { \ 404 ASSERT(!BP_IS_EMBEDDED(bp)); \ 405 BF64_SET((bp)->blk_prop, 40, 8, x); \ 406_NOTE(CONSTCOND) } while (0) 407 408#define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 409#define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 410 411#define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 412#define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 413 414#define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 415#define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 416 417#define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 418#define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 419 420#define BP_PHYSICAL_BIRTH(bp) \ 421 (BP_IS_EMBEDDED(bp) ? 0 : \ 422 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 423 424#define BP_SET_BIRTH(bp, logical, physical) \ 425{ \ 426 ASSERT(!BP_IS_EMBEDDED(bp)); \ 427 (bp)->blk_birth = (logical); \ 428 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \ 429} 430 431#define BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill) 432 433#define BP_GET_ASIZE(bp) \ 434 (BP_IS_EMBEDDED(bp) ? 0 : \ 435 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 436 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 437 DVA_GET_ASIZE(&(bp)->blk_dva[2])) 438 439#define BP_GET_UCSIZE(bp) \ 440 ((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \ 441 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)) 442 443#define BP_GET_NDVAS(bp) \ 444 (BP_IS_EMBEDDED(bp) ? 0 : \ 445 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 446 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 447 !!DVA_GET_ASIZE(&(bp)->blk_dva[2])) 448 449#define BP_COUNT_GANG(bp) \ 450 (BP_IS_EMBEDDED(bp) ? 0 : \ 451 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \ 452 DVA_GET_GANG(&(bp)->blk_dva[1]) + \ 453 DVA_GET_GANG(&(bp)->blk_dva[2]))) 454 455#define DVA_EQUAL(dva1, dva2) \ 456 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 457 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 458 459#define BP_EQUAL(bp1, bp2) \ 460 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \ 461 (bp1)->blk_birth == (bp2)->blk_birth && \ 462 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \ 463 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \ 464 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2])) 465 466#define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 467 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 468 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 469 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 470 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 471 472#define ZIO_CHECKSUM_IS_ZERO(zc) \ 473 (0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \ 474 (zc)->zc_word[2] | (zc)->zc_word[3])) 475 476#define ZIO_CHECKSUM_BSWAP(zcp) \ 477{ \ 478 (zcp)->zc_word[0] = BSWAP_64((zcp)->zc_word[0]); \ 479 (zcp)->zc_word[1] = BSWAP_64((zcp)->zc_word[1]); \ 480 (zcp)->zc_word[2] = BSWAP_64((zcp)->zc_word[2]); \ 481 (zcp)->zc_word[3] = BSWAP_64((zcp)->zc_word[3]); \ 482} 483 484 485#define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 486 487#define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ 488{ \ 489 (zcp)->zc_word[0] = w0; \ 490 (zcp)->zc_word[1] = w1; \ 491 (zcp)->zc_word[2] = w2; \ 492 (zcp)->zc_word[3] = w3; \ 493} 494 495#define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0]) 496#define BP_IS_GANG(bp) \ 497 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp))) 498#define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 499 (dva)->dva_word[1] == 0ULL) 500#define BP_IS_HOLE(bp) \ 501 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp))) 502 503/* BP_IS_RAIDZ(bp) assumes no block compression */ 504#define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \ 505 BP_GET_PSIZE(bp)) 506 507#define BP_ZERO(bp) \ 508{ \ 509 (bp)->blk_dva[0].dva_word[0] = 0; \ 510 (bp)->blk_dva[0].dva_word[1] = 0; \ 511 (bp)->blk_dva[1].dva_word[0] = 0; \ 512 (bp)->blk_dva[1].dva_word[1] = 0; \ 513 (bp)->blk_dva[2].dva_word[0] = 0; \ 514 (bp)->blk_dva[2].dva_word[1] = 0; \ 515 (bp)->blk_prop = 0; \ 516 (bp)->blk_pad[0] = 0; \ 517 (bp)->blk_pad[1] = 0; \ 518 (bp)->blk_phys_birth = 0; \ 519 (bp)->blk_birth = 0; \ 520 (bp)->blk_fill = 0; \ 521 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 522} 523 524#if BYTE_ORDER == _BIG_ENDIAN 525#define ZFS_HOST_BYTEORDER (0ULL) 526#else 527#define ZFS_HOST_BYTEORDER (1ULL) 528#endif 529 530#define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) 531 532#define BP_SPRINTF_LEN 320 533 534/* 535 * This macro allows code sharing between zfs, libzpool, and mdb. 536 * 'func' is either snprintf() or mdb_snprintf(). 537 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line. 538 */ 539#define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \ 540{ \ 541 static const char *copyname[] = \ 542 { "zero", "single", "double", "triple" }; \ 543 int len = 0; \ 544 int copies = 0; \ 545 \ 546 if (bp == NULL) { \ 547 len += func(buf + len, size - len, "<NULL>"); \ 548 } else if (BP_IS_HOLE(bp)) { \ 549 len += func(buf + len, size - len, \ 550 "HOLE [L%llu %s] " \ 551 "size=%llxL birth=%lluL", \ 552 (u_longlong_t)BP_GET_LEVEL(bp), \ 553 type, \ 554 (u_longlong_t)BP_GET_LSIZE(bp), \ 555 (u_longlong_t)bp->blk_birth); \ 556 } else if (BP_IS_EMBEDDED(bp)) { \ 557 len = func(buf + len, size - len, \ 558 "EMBEDDED [L%llu %s] et=%u %s " \ 559 "size=%llxL/%llxP birth=%lluL", \ 560 (u_longlong_t)BP_GET_LEVEL(bp), \ 561 type, \ 562 (int)BPE_GET_ETYPE(bp), \ 563 compress, \ 564 (u_longlong_t)BPE_GET_LSIZE(bp), \ 565 (u_longlong_t)BPE_GET_PSIZE(bp), \ 566 (u_longlong_t)bp->blk_birth); \ 567 } else { \ 568 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \ 569 const dva_t *dva = &bp->blk_dva[d]; \ 570 if (DVA_IS_VALID(dva)) \ 571 copies++; \ 572 len += func(buf + len, size - len, \ 573 "DVA[%d]=<%llu:%llx:%llx>%c", d, \ 574 (u_longlong_t)DVA_GET_VDEV(dva), \ 575 (u_longlong_t)DVA_GET_OFFSET(dva), \ 576 (u_longlong_t)DVA_GET_ASIZE(dva), \ 577 ws); \ 578 } \ 579 if (BP_IS_GANG(bp) && \ 580 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \ 581 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \ 582 copies--; \ 583 len += func(buf + len, size - len, \ 584 "[L%llu %s] %s %s %s %s %s %s%c" \ 585 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \ 586 "cksum=%llx:%llx:%llx:%llx", \ 587 (u_longlong_t)BP_GET_LEVEL(bp), \ 588 type, \ 589 checksum, \ 590 compress, \ 591 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \ 592 BP_IS_GANG(bp) ? "gang" : "contiguous", \ 593 BP_GET_DEDUP(bp) ? "dedup" : "unique", \ 594 copyname[copies], \ 595 ws, \ 596 (u_longlong_t)BP_GET_LSIZE(bp), \ 597 (u_longlong_t)BP_GET_PSIZE(bp), \ 598 (u_longlong_t)bp->blk_birth, \ 599 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \ 600 (u_longlong_t)BP_GET_FILL(bp), \ 601 ws, \ 602 (u_longlong_t)bp->blk_cksum.zc_word[0], \ 603 (u_longlong_t)bp->blk_cksum.zc_word[1], \ 604 (u_longlong_t)bp->blk_cksum.zc_word[2], \ 605 (u_longlong_t)bp->blk_cksum.zc_word[3]); \ 606 } \ 607 ASSERT(len < size); \ 608} 609 610#include <sys/dmu.h> 611 612#define BP_GET_BUFC_TYPE(bp) \ 613 (((BP_GET_LEVEL(bp) > 0) || (DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))) ? \ 614 ARC_BUFC_METADATA : ARC_BUFC_DATA) 615 616typedef enum spa_import_type { 617 SPA_IMPORT_EXISTING, 618 SPA_IMPORT_ASSEMBLE 619} spa_import_type_t; 620 621/* state manipulation functions */ 622extern int spa_open(const char *pool, spa_t **, void *tag); 623extern int spa_open_rewind(const char *pool, spa_t **, void *tag, 624 nvlist_t *policy, nvlist_t **config); 625extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot, 626 size_t buflen); 627extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props, 628 nvlist_t *zplprops); 629#ifdef illumos 630extern int spa_import_rootpool(char *devpath, char *devid); 631#else 632extern int spa_import_rootpool(const char *name); 633#endif 634extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props, 635 uint64_t flags); 636extern nvlist_t *spa_tryimport(nvlist_t *tryconfig); 637extern int spa_destroy(char *pool); 638extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force, 639 boolean_t hardforce); 640extern int spa_reset(char *pool); 641extern void spa_async_request(spa_t *spa, int flag); 642extern void spa_async_unrequest(spa_t *spa, int flag); 643extern void spa_async_suspend(spa_t *spa); 644extern void spa_async_resume(spa_t *spa); 645extern spa_t *spa_inject_addref(char *pool); 646extern void spa_inject_delref(spa_t *spa); 647extern void spa_scan_stat_init(spa_t *spa); 648extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps); 649 650#define SPA_ASYNC_CONFIG_UPDATE 0x01 651#define SPA_ASYNC_REMOVE 0x02 652#define SPA_ASYNC_PROBE 0x04 653#define SPA_ASYNC_RESILVER_DONE 0x08 654#define SPA_ASYNC_RESILVER 0x10 655#define SPA_ASYNC_AUTOEXPAND 0x20 656#define SPA_ASYNC_REMOVE_DONE 0x40 657#define SPA_ASYNC_REMOVE_STOP 0x80 658 659/* 660 * Controls the behavior of spa_vdev_remove(). 661 */ 662#define SPA_REMOVE_UNSPARE 0x01 663#define SPA_REMOVE_DONE 0x02 664 665/* device manipulation */ 666extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot); 667extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, 668 int replacing); 669extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, 670 int replace_done); 671extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare); 672extern boolean_t spa_vdev_remove_active(spa_t *spa); 673extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath); 674extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru); 675extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config, 676 nvlist_t *props, boolean_t exp); 677 678/* spare state (which is global across all pools) */ 679extern void spa_spare_add(vdev_t *vd); 680extern void spa_spare_remove(vdev_t *vd); 681extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt); 682extern void spa_spare_activate(vdev_t *vd); 683 684/* L2ARC state (which is global across all pools) */ 685extern void spa_l2cache_add(vdev_t *vd); 686extern void spa_l2cache_remove(vdev_t *vd); 687extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool); 688extern void spa_l2cache_activate(vdev_t *vd); 689extern void spa_l2cache_drop(spa_t *spa); 690 691/* scanning */ 692extern int spa_scan(spa_t *spa, pool_scan_func_t func); 693extern int spa_scan_stop(spa_t *spa); 694 695/* spa syncing */ 696extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */ 697extern void spa_sync_allpools(void); 698 699/* spa namespace global mutex */ 700extern kmutex_t spa_namespace_lock; 701 702/* 703 * SPA configuration functions in spa_config.c 704 */ 705 706#define SPA_CONFIG_UPDATE_POOL 0 707#define SPA_CONFIG_UPDATE_VDEVS 1 708 709extern void spa_config_sync(spa_t *, boolean_t, boolean_t); 710extern void spa_config_load(void); 711extern nvlist_t *spa_all_configs(uint64_t *); 712extern void spa_config_set(spa_t *spa, nvlist_t *config); 713extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, 714 int getstats); 715extern void spa_config_update(spa_t *spa, int what); 716 717/* 718 * Miscellaneous SPA routines in spa_misc.c 719 */ 720 721/* Namespace manipulation */ 722extern spa_t *spa_lookup(const char *name); 723extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot); 724extern void spa_remove(spa_t *spa); 725extern spa_t *spa_next(spa_t *prev); 726 727/* Refcount functions */ 728extern void spa_open_ref(spa_t *spa, void *tag); 729extern void spa_close(spa_t *spa, void *tag); 730extern void spa_async_close(spa_t *spa, void *tag); 731extern boolean_t spa_refcount_zero(spa_t *spa); 732 733#define SCL_NONE 0x00 734#define SCL_CONFIG 0x01 735#define SCL_STATE 0x02 736#define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */ 737#define SCL_ALLOC 0x08 738#define SCL_ZIO 0x10 739#define SCL_FREE 0x20 740#define SCL_VDEV 0x40 741#define SCL_LOCKS 7 742#define SCL_ALL ((1 << SCL_LOCKS) - 1) 743#define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO) 744 745/* Pool configuration locks */ 746extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw); 747extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw); 748extern void spa_config_exit(spa_t *spa, int locks, void *tag); 749extern int spa_config_held(spa_t *spa, int locks, krw_t rw); 750 751/* Pool vdev add/remove lock */ 752extern uint64_t spa_vdev_enter(spa_t *spa); 753extern uint64_t spa_vdev_config_enter(spa_t *spa); 754extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, 755 int error, char *tag); 756extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error); 757 758/* Pool vdev state change lock */ 759extern void spa_vdev_state_enter(spa_t *spa, int oplock); 760extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error); 761 762/* Log state */ 763typedef enum spa_log_state { 764 SPA_LOG_UNKNOWN = 0, /* unknown log state */ 765 SPA_LOG_MISSING, /* missing log(s) */ 766 SPA_LOG_CLEAR, /* clear the log(s) */ 767 SPA_LOG_GOOD, /* log(s) are good */ 768} spa_log_state_t; 769 770extern spa_log_state_t spa_get_log_state(spa_t *spa); 771extern void spa_set_log_state(spa_t *spa, spa_log_state_t state); 772extern int spa_offline_log(spa_t *spa); 773 774/* Log claim callback */ 775extern void spa_claim_notify(zio_t *zio); 776 777/* Accessor functions */ 778extern boolean_t spa_shutting_down(spa_t *spa); 779extern struct dsl_pool *spa_get_dsl(spa_t *spa); 780extern boolean_t spa_is_initializing(spa_t *spa); 781extern blkptr_t *spa_get_rootblkptr(spa_t *spa); 782extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp); 783extern void spa_altroot(spa_t *, char *, size_t); 784extern int spa_sync_pass(spa_t *spa); 785extern char *spa_name(spa_t *spa); 786extern uint64_t spa_guid(spa_t *spa); 787extern uint64_t spa_load_guid(spa_t *spa); 788extern uint64_t spa_last_synced_txg(spa_t *spa); 789extern uint64_t spa_first_txg(spa_t *spa); 790extern uint64_t spa_syncing_txg(spa_t *spa); 791extern uint64_t spa_version(spa_t *spa); 792extern pool_state_t spa_state(spa_t *spa); 793extern spa_load_state_t spa_load_state(spa_t *spa); 794extern uint64_t spa_freeze_txg(spa_t *spa); 795extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize); 796extern uint64_t spa_get_dspace(spa_t *spa); 797extern uint64_t spa_get_slop_space(spa_t *spa); 798extern void spa_update_dspace(spa_t *spa); 799extern uint64_t spa_version(spa_t *spa); 800extern boolean_t spa_deflate(spa_t *spa); 801extern metaslab_class_t *spa_normal_class(spa_t *spa); 802extern metaslab_class_t *spa_log_class(spa_t *spa); 803extern void spa_evicting_os_register(spa_t *, objset_t *os); 804extern void spa_evicting_os_deregister(spa_t *, objset_t *os); 805extern void spa_evicting_os_wait(spa_t *spa); 806extern int spa_max_replication(spa_t *spa); 807extern int spa_prev_software_version(spa_t *spa); 808extern int spa_busy(void); 809extern uint8_t spa_get_failmode(spa_t *spa); 810extern boolean_t spa_suspended(spa_t *spa); 811extern uint64_t spa_bootfs(spa_t *spa); 812extern uint64_t spa_delegation(spa_t *spa); 813extern objset_t *spa_meta_objset(spa_t *spa); 814extern uint64_t spa_deadman_synctime(spa_t *spa); 815 816/* Miscellaneous support routines */ 817extern void spa_activate_mos_feature(spa_t *spa, const char *feature, 818 dmu_tx_t *tx); 819extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature); 820extern int spa_rename(const char *oldname, const char *newname); 821extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid); 822extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid); 823extern char *spa_strdup(const char *); 824extern void spa_strfree(char *); 825extern uint64_t spa_get_random(uint64_t range); 826extern uint64_t spa_generate_guid(spa_t *spa); 827extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp); 828extern void spa_freeze(spa_t *spa); 829extern int spa_change_guid(spa_t *spa); 830extern void spa_upgrade(spa_t *spa, uint64_t version); 831extern void spa_evict_all(void); 832extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid, 833 boolean_t l2cache); 834extern boolean_t spa_has_spare(spa_t *, uint64_t guid); 835extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva); 836extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp); 837extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp); 838extern boolean_t spa_has_slogs(spa_t *spa); 839extern boolean_t spa_is_root(spa_t *spa); 840extern boolean_t spa_writeable(spa_t *spa); 841extern boolean_t spa_has_pending_synctask(spa_t *spa); 842extern int spa_maxblocksize(spa_t *spa); 843extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp); 844 845extern int spa_mode(spa_t *spa); 846extern uint64_t zfs_strtonum(const char *str, char **nptr); 847#define strtonum(str, nptr) zfs_strtonum((str), (nptr)) 848 849extern char *spa_his_ievent_table[]; 850 851extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx); 852extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read, 853 char *his_buf); 854extern int spa_history_log(spa_t *spa, const char *his_buf); 855extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl); 856extern void spa_history_log_version(spa_t *spa, const char *operation); 857extern void spa_history_log_internal(spa_t *spa, const char *operation, 858 dmu_tx_t *tx, const char *fmt, ...); 859extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op, 860 dmu_tx_t *tx, const char *fmt, ...); 861extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation, 862 dmu_tx_t *tx, const char *fmt, ...); 863 864/* error handling */ 865struct zbookmark_phys; 866extern void spa_log_error(spa_t *spa, zio_t *zio); 867extern void zfs_ereport_post(const char *cls, spa_t *spa, vdev_t *vd, 868 zio_t *zio, uint64_t stateoroffset, uint64_t length); 869extern void zfs_post_remove(spa_t *spa, vdev_t *vd); 870extern void zfs_post_state_change(spa_t *spa, vdev_t *vd); 871extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd); 872extern uint64_t spa_get_errlog_size(spa_t *spa); 873extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count); 874extern void spa_errlog_rotate(spa_t *spa); 875extern void spa_errlog_drain(spa_t *spa); 876extern void spa_errlog_sync(spa_t *spa, uint64_t txg); 877extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub); 878 879/* vdev cache */ 880extern void vdev_cache_stat_init(void); 881extern void vdev_cache_stat_fini(void); 882 883/* Initialization and termination */ 884extern void spa_init(int flags); 885extern void spa_fini(void); 886extern void spa_boot_init(); 887 888/* properties */ 889extern int spa_prop_set(spa_t *spa, nvlist_t *nvp); 890extern int spa_prop_get(spa_t *spa, nvlist_t **nvp); 891extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx); 892extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t); 893 894/* asynchronous event notification */ 895extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name); 896 897#ifdef ZFS_DEBUG 898#define dprintf_bp(bp, fmt, ...) do { \ 899 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 900 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \ 901 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \ 902 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \ 903 kmem_free(__blkbuf, BP_SPRINTF_LEN); \ 904 } \ 905_NOTE(CONSTCOND) } while (0) 906#else 907#define dprintf_bp(bp, fmt, ...) 908#endif 909 910extern boolean_t spa_debug_enabled(spa_t *spa); 911#define spa_dbgmsg(spa, ...) \ 912{ \ 913 if (spa_debug_enabled(spa)) \ 914 zfs_dbgmsg(__VA_ARGS__); \ 915} 916 917extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */ 918 919#ifdef __cplusplus 920} 921#endif 922 923#endif /* _SYS_SPA_H */ 924