spa.h revision 290757
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 */ 28 29#ifndef _SYS_SPA_H 30#define _SYS_SPA_H 31 32#include <sys/avl.h> 33#include <sys/zfs_context.h> 34#include <sys/nvpair.h> 35#include <sys/sysmacros.h> 36#include <sys/types.h> 37#include <sys/fs/zfs.h> 38 39#ifdef __cplusplus 40extern "C" { 41#endif 42 43/* 44 * Forward references that lots of things need. 45 */ 46typedef struct spa spa_t; 47typedef struct vdev vdev_t; 48typedef struct metaslab metaslab_t; 49typedef struct metaslab_group metaslab_group_t; 50typedef struct metaslab_class metaslab_class_t; 51typedef struct zio zio_t; 52typedef struct zilog zilog_t; 53typedef struct spa_aux_vdev spa_aux_vdev_t; 54typedef struct ddt ddt_t; 55typedef struct ddt_entry ddt_entry_t; 56struct dsl_pool; 57struct dsl_dataset; 58 59/* 60 * General-purpose 32-bit and 64-bit bitfield encodings. 61 */ 62#define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) 63#define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) 64#define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) 65#define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) 66 67#define BF32_GET(x, low, len) BF32_DECODE(x, low, len) 68#define BF64_GET(x, low, len) BF64_DECODE(x, low, len) 69 70#define BF32_SET(x, low, len, val) do { \ 71 ASSERT3U(val, <, 1U << (len)); \ 72 ASSERT3U(low + len, <=, 32); \ 73 (x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \ 74_NOTE(CONSTCOND) } while (0) 75 76#define BF64_SET(x, low, len, val) do { \ 77 ASSERT3U(val, <, 1ULL << (len)); \ 78 ASSERT3U(low + len, <=, 64); \ 79 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \ 80_NOTE(CONSTCOND) } while (0) 81 82#define BF32_GET_SB(x, low, len, shift, bias) \ 83 ((BF32_GET(x, low, len) + (bias)) << (shift)) 84#define BF64_GET_SB(x, low, len, shift, bias) \ 85 ((BF64_GET(x, low, len) + (bias)) << (shift)) 86 87#define BF32_SET_SB(x, low, len, shift, bias, val) do { \ 88 ASSERT(IS_P2ALIGNED(val, 1U << shift)); \ 89 ASSERT3S((val) >> (shift), >=, bias); \ 90 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \ 91_NOTE(CONSTCOND) } while (0) 92#define BF64_SET_SB(x, low, len, shift, bias, val) do { \ 93 ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \ 94 ASSERT3S((val) >> (shift), >=, bias); \ 95 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \ 96_NOTE(CONSTCOND) } while (0) 97 98/* 99 * We currently support block sizes from 512 bytes to 16MB. 100 * The benefits of larger blocks, and thus larger IO, need to be weighed 101 * against the cost of COWing a giant block to modify one byte, and the 102 * large latency of reading or writing a large block. 103 * 104 * Note that although blocks up to 16MB are supported, the recordsize 105 * property can not be set larger than zfs_max_recordsize (default 1MB). 106 * See the comment near zfs_max_recordsize in dsl_dataset.c for details. 107 * 108 * Note that although the LSIZE field of the blkptr_t can store sizes up 109 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to 110 * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB. 111 */ 112#define SPA_MINBLOCKSHIFT 9 113#define SPA_OLD_MAXBLOCKSHIFT 17 114#define SPA_MAXBLOCKSHIFT 24 115#define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 116#define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT) 117#define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 118 119/* 120 * Default maximum supported logical ashift. 121 * 122 * The current 8k allocation block size limit is due to the 8k 123 * aligned/sized operations performed by vdev_probe() on 124 * vdev_label->vl_pad2. Using another "safe region" for these tests 125 * would allow the limit to be raised to 16k, at the expense of 126 * only having 8 available uberblocks in the label area. 127 */ 128#define SPA_MAXASHIFT 13 129 130/* 131 * Default minimum supported logical ashift. 132 */ 133#define SPA_MINASHIFT SPA_MINBLOCKSHIFT 134 135/* 136 * Size of block to hold the configuration data (a packed nvlist) 137 */ 138#define SPA_CONFIG_BLOCKSIZE (1ULL << 14) 139 140/* 141 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. 142 * The ASIZE encoding should be at least 64 times larger (6 more bits) 143 * to support up to 4-way RAID-Z mirror mode with worst-case gang block 144 * overhead, three DVAs per bp, plus one more bit in case we do anything 145 * else that expands the ASIZE. 146 */ 147#define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ 148#define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ 149#define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ 150 151/* 152 * All SPA data is represented by 128-bit data virtual addresses (DVAs). 153 * The members of the dva_t should be considered opaque outside the SPA. 154 */ 155typedef struct dva { 156 uint64_t dva_word[2]; 157} dva_t; 158 159/* 160 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. 161 */ 162typedef struct zio_cksum { 163 uint64_t zc_word[4]; 164} zio_cksum_t; 165 166/* 167 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept 168 * secret and is suitable for use in MAC algorithms as the key. 169 */ 170typedef struct zio_cksum_salt { 171 uint8_t zcs_bytes[32]; 172} zio_cksum_salt_t; 173 174/* 175 * Each block is described by its DVAs, time of birth, checksum, etc. 176 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 177 * 178 * 64 56 48 40 32 24 16 8 0 179 * +-------+-------+-------+-------+-------+-------+-------+-------+ 180 * 0 | vdev1 | GRID | ASIZE | 181 * +-------+-------+-------+-------+-------+-------+-------+-------+ 182 * 1 |G| offset1 | 183 * +-------+-------+-------+-------+-------+-------+-------+-------+ 184 * 2 | vdev2 | GRID | ASIZE | 185 * +-------+-------+-------+-------+-------+-------+-------+-------+ 186 * 3 |G| offset2 | 187 * +-------+-------+-------+-------+-------+-------+-------+-------+ 188 * 4 | vdev3 | GRID | ASIZE | 189 * +-------+-------+-------+-------+-------+-------+-------+-------+ 190 * 5 |G| offset3 | 191 * +-------+-------+-------+-------+-------+-------+-------+-------+ 192 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 193 * +-------+-------+-------+-------+-------+-------+-------+-------+ 194 * 7 | padding | 195 * +-------+-------+-------+-------+-------+-------+-------+-------+ 196 * 8 | padding | 197 * +-------+-------+-------+-------+-------+-------+-------+-------+ 198 * 9 | physical birth txg | 199 * +-------+-------+-------+-------+-------+-------+-------+-------+ 200 * a | logical birth txg | 201 * +-------+-------+-------+-------+-------+-------+-------+-------+ 202 * b | fill count | 203 * +-------+-------+-------+-------+-------+-------+-------+-------+ 204 * c | checksum[0] | 205 * +-------+-------+-------+-------+-------+-------+-------+-------+ 206 * d | checksum[1] | 207 * +-------+-------+-------+-------+-------+-------+-------+-------+ 208 * e | checksum[2] | 209 * +-------+-------+-------+-------+-------+-------+-------+-------+ 210 * f | checksum[3] | 211 * +-------+-------+-------+-------+-------+-------+-------+-------+ 212 * 213 * Legend: 214 * 215 * vdev virtual device ID 216 * offset offset into virtual device 217 * LSIZE logical size 218 * PSIZE physical size (after compression) 219 * ASIZE allocated size (including RAID-Z parity and gang block headers) 220 * GRID RAID-Z layout information (reserved for future use) 221 * cksum checksum function 222 * comp compression function 223 * G gang block indicator 224 * B byteorder (endianness) 225 * D dedup 226 * X encryption (on version 30, which is not supported) 227 * E blkptr_t contains embedded data (see below) 228 * lvl level of indirection 229 * type DMU object type 230 * phys birth txg of block allocation; zero if same as logical birth txg 231 * log. birth transaction group in which the block was logically born 232 * fill count number of non-zero blocks under this bp 233 * checksum[4] 256-bit checksum of the data this bp describes 234 */ 235 236/* 237 * "Embedded" blkptr_t's don't actually point to a block, instead they 238 * have a data payload embedded in the blkptr_t itself. See the comment 239 * in blkptr.c for more details. 240 * 241 * The blkptr_t is laid out as follows: 242 * 243 * 64 56 48 40 32 24 16 8 0 244 * +-------+-------+-------+-------+-------+-------+-------+-------+ 245 * 0 | payload | 246 * 1 | payload | 247 * 2 | payload | 248 * 3 | payload | 249 * 4 | payload | 250 * 5 | payload | 251 * +-------+-------+-------+-------+-------+-------+-------+-------+ 252 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 253 * +-------+-------+-------+-------+-------+-------+-------+-------+ 254 * 7 | payload | 255 * 8 | payload | 256 * 9 | payload | 257 * +-------+-------+-------+-------+-------+-------+-------+-------+ 258 * a | logical birth txg | 259 * +-------+-------+-------+-------+-------+-------+-------+-------+ 260 * b | payload | 261 * c | payload | 262 * d | payload | 263 * e | payload | 264 * f | payload | 265 * +-------+-------+-------+-------+-------+-------+-------+-------+ 266 * 267 * Legend: 268 * 269 * payload contains the embedded data 270 * B (byteorder) byteorder (endianness) 271 * D (dedup) padding (set to zero) 272 * X encryption (set to zero; see above) 273 * E (embedded) set to one 274 * lvl indirection level 275 * type DMU object type 276 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 277 * comp compression function of payload 278 * PSIZE size of payload after compression, in bytes 279 * LSIZE logical size of payload, in bytes 280 * note that 25 bits is enough to store the largest 281 * "normal" BP's LSIZE (2^16 * 2^9) in bytes 282 * log. birth transaction group in which the block was logically born 283 * 284 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 285 * bp's they are stored in units of SPA_MINBLOCKSHIFT. 286 * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 287 * The B, D, X, lvl, type, and comp fields are stored the same as with normal 288 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 289 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 290 * other macros, as they assert that they are only used on BP's of the correct 291 * "embedded-ness". 292 */ 293 294#define BPE_GET_ETYPE(bp) \ 295 (ASSERT(BP_IS_EMBEDDED(bp)), \ 296 BF64_GET((bp)->blk_prop, 40, 8)) 297#define BPE_SET_ETYPE(bp, t) do { \ 298 ASSERT(BP_IS_EMBEDDED(bp)); \ 299 BF64_SET((bp)->blk_prop, 40, 8, t); \ 300_NOTE(CONSTCOND) } while (0) 301 302#define BPE_GET_LSIZE(bp) \ 303 (ASSERT(BP_IS_EMBEDDED(bp)), \ 304 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 305#define BPE_SET_LSIZE(bp, x) do { \ 306 ASSERT(BP_IS_EMBEDDED(bp)); \ 307 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 308_NOTE(CONSTCOND) } while (0) 309 310#define BPE_GET_PSIZE(bp) \ 311 (ASSERT(BP_IS_EMBEDDED(bp)), \ 312 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 313#define BPE_SET_PSIZE(bp, x) do { \ 314 ASSERT(BP_IS_EMBEDDED(bp)); \ 315 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 316_NOTE(CONSTCOND) } while (0) 317 318typedef enum bp_embedded_type { 319 BP_EMBEDDED_TYPE_DATA, 320 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ 321 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED 322} bp_embedded_type_t; 323 324#define BPE_NUM_WORDS 14 325#define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 326#define BPE_IS_PAYLOADWORD(bp, wp) \ 327 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 328 329#define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 330#define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 331 332/* 333 * A block is a hole when it has either 1) never been written to, or 334 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads 335 * without physically allocating disk space. Holes are represented in the 336 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is 337 * done through the BP_IS_HOLE macro. For holes, the logical size, level, 338 * DMU object type, and birth times are all also stored for holes that 339 * were written to at some point (i.e. were punched after having been filled). 340 */ 341typedef struct blkptr { 342 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 343 uint64_t blk_prop; /* size, compression, type, etc */ 344 uint64_t blk_pad[2]; /* Extra space for the future */ 345 uint64_t blk_phys_birth; /* txg when block was allocated */ 346 uint64_t blk_birth; /* transaction group at birth */ 347 uint64_t blk_fill; /* fill count */ 348 zio_cksum_t blk_cksum; /* 256-bit checksum */ 349} blkptr_t; 350 351/* 352 * Macros to get and set fields in a bp or DVA. 353 */ 354#define DVA_GET_ASIZE(dva) \ 355 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 356#define DVA_SET_ASIZE(dva, x) \ 357 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 358 SPA_MINBLOCKSHIFT, 0, x) 359 360#define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 361#define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 362 363#define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) 364#define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) 365 366#define DVA_GET_OFFSET(dva) \ 367 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 368#define DVA_SET_OFFSET(dva, x) \ 369 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 370 371#define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 372#define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 373 374#define BP_GET_LSIZE(bp) \ 375 (BP_IS_EMBEDDED(bp) ? \ 376 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 377 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 378#define BP_SET_LSIZE(bp, x) do { \ 379 ASSERT(!BP_IS_EMBEDDED(bp)); \ 380 BF64_SET_SB((bp)->blk_prop, \ 381 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 382_NOTE(CONSTCOND) } while (0) 383 384#define BP_GET_PSIZE(bp) \ 385 (BP_IS_EMBEDDED(bp) ? 0 : \ 386 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 387#define BP_SET_PSIZE(bp, x) do { \ 388 ASSERT(!BP_IS_EMBEDDED(bp)); \ 389 BF64_SET_SB((bp)->blk_prop, \ 390 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 391_NOTE(CONSTCOND) } while (0) 392 393#define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7) 394#define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x) 395 396#define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 397#define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x) 398 399#define BP_GET_CHECKSUM(bp) \ 400 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \ 401 BF64_GET((bp)->blk_prop, 40, 8)) 402#define BP_SET_CHECKSUM(bp, x) do { \ 403 ASSERT(!BP_IS_EMBEDDED(bp)); \ 404 BF64_SET((bp)->blk_prop, 40, 8, x); \ 405_NOTE(CONSTCOND) } while (0) 406 407#define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 408#define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 409 410#define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 411#define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 412 413#define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 414#define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 415 416#define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 417#define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 418 419#define BP_PHYSICAL_BIRTH(bp) \ 420 (BP_IS_EMBEDDED(bp) ? 0 : \ 421 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 422 423#define BP_SET_BIRTH(bp, logical, physical) \ 424{ \ 425 ASSERT(!BP_IS_EMBEDDED(bp)); \ 426 (bp)->blk_birth = (logical); \ 427 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \ 428} 429 430#define BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill) 431 432#define BP_GET_ASIZE(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_GET_UCSIZE(bp) \ 439 ((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \ 440 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)) 441 442#define BP_GET_NDVAS(bp) \ 443 (BP_IS_EMBEDDED(bp) ? 0 : \ 444 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 445 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 446 !!DVA_GET_ASIZE(&(bp)->blk_dva[2])) 447 448#define BP_COUNT_GANG(bp) \ 449 (BP_IS_EMBEDDED(bp) ? 0 : \ 450 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \ 451 DVA_GET_GANG(&(bp)->blk_dva[1]) + \ 452 DVA_GET_GANG(&(bp)->blk_dva[2]))) 453 454#define DVA_EQUAL(dva1, dva2) \ 455 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 456 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 457 458#define BP_EQUAL(bp1, bp2) \ 459 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \ 460 (bp1)->blk_birth == (bp2)->blk_birth && \ 461 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \ 462 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \ 463 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2])) 464 465#define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 466 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 467 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 468 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 469 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 470 471#define ZIO_CHECKSUM_IS_ZERO(zc) \ 472 (0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \ 473 (zc)->zc_word[2] | (zc)->zc_word[3])) 474 475#define ZIO_CHECKSUM_BSWAP(zcp) \ 476{ \ 477 (zcp)->zc_word[0] = BSWAP_64((zcp)->zc_word[0]); \ 478 (zcp)->zc_word[1] = BSWAP_64((zcp)->zc_word[1]); \ 479 (zcp)->zc_word[2] = BSWAP_64((zcp)->zc_word[2]); \ 480 (zcp)->zc_word[3] = BSWAP_64((zcp)->zc_word[3]); \ 481} 482 483 484#define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 485 486#define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ 487{ \ 488 (zcp)->zc_word[0] = w0; \ 489 (zcp)->zc_word[1] = w1; \ 490 (zcp)->zc_word[2] = w2; \ 491 (zcp)->zc_word[3] = w3; \ 492} 493 494#define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0]) 495#define BP_IS_GANG(bp) \ 496 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp))) 497#define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 498 (dva)->dva_word[1] == 0ULL) 499#define BP_IS_HOLE(bp) \ 500 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp))) 501 502/* BP_IS_RAIDZ(bp) assumes no block compression */ 503#define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \ 504 BP_GET_PSIZE(bp)) 505 506#define BP_ZERO(bp) \ 507{ \ 508 (bp)->blk_dva[0].dva_word[0] = 0; \ 509 (bp)->blk_dva[0].dva_word[1] = 0; \ 510 (bp)->blk_dva[1].dva_word[0] = 0; \ 511 (bp)->blk_dva[1].dva_word[1] = 0; \ 512 (bp)->blk_dva[2].dva_word[0] = 0; \ 513 (bp)->blk_dva[2].dva_word[1] = 0; \ 514 (bp)->blk_prop = 0; \ 515 (bp)->blk_pad[0] = 0; \ 516 (bp)->blk_pad[1] = 0; \ 517 (bp)->blk_phys_birth = 0; \ 518 (bp)->blk_birth = 0; \ 519 (bp)->blk_fill = 0; \ 520 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 521} 522 523#if BYTE_ORDER == _BIG_ENDIAN 524#define ZFS_HOST_BYTEORDER (0ULL) 525#else 526#define ZFS_HOST_BYTEORDER (1ULL) 527#endif 528 529#define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) 530 531#define BP_SPRINTF_LEN 320 532 533/* 534 * This macro allows code sharing between zfs, libzpool, and mdb. 535 * 'func' is either snprintf() or mdb_snprintf(). 536 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line. 537 */ 538#define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \ 539{ \ 540 static const char *copyname[] = \ 541 { "zero", "single", "double", "triple" }; \ 542 int len = 0; \ 543 int copies = 0; \ 544 \ 545 if (bp == NULL) { \ 546 len += func(buf + len, size - len, "<NULL>"); \ 547 } else if (BP_IS_HOLE(bp)) { \ 548 len += func(buf + len, size - len, \ 549 "HOLE [L%llu %s] " \ 550 "size=%llxL birth=%lluL", \ 551 (u_longlong_t)BP_GET_LEVEL(bp), \ 552 type, \ 553 (u_longlong_t)BP_GET_LSIZE(bp), \ 554 (u_longlong_t)bp->blk_birth); \ 555 } else if (BP_IS_EMBEDDED(bp)) { \ 556 len = func(buf + len, size - len, \ 557 "EMBEDDED [L%llu %s] et=%u %s " \ 558 "size=%llxL/%llxP birth=%lluL", \ 559 (u_longlong_t)BP_GET_LEVEL(bp), \ 560 type, \ 561 (int)BPE_GET_ETYPE(bp), \ 562 compress, \ 563 (u_longlong_t)BPE_GET_LSIZE(bp), \ 564 (u_longlong_t)BPE_GET_PSIZE(bp), \ 565 (u_longlong_t)bp->blk_birth); \ 566 } else { \ 567 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \ 568 const dva_t *dva = &bp->blk_dva[d]; \ 569 if (DVA_IS_VALID(dva)) \ 570 copies++; \ 571 len += func(buf + len, size - len, \ 572 "DVA[%d]=<%llu:%llx:%llx>%c", d, \ 573 (u_longlong_t)DVA_GET_VDEV(dva), \ 574 (u_longlong_t)DVA_GET_OFFSET(dva), \ 575 (u_longlong_t)DVA_GET_ASIZE(dva), \ 576 ws); \ 577 } \ 578 if (BP_IS_GANG(bp) && \ 579 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \ 580 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \ 581 copies--; \ 582 len += func(buf + len, size - len, \ 583 "[L%llu %s] %s %s %s %s %s %s%c" \ 584 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \ 585 "cksum=%llx:%llx:%llx:%llx", \ 586 (u_longlong_t)BP_GET_LEVEL(bp), \ 587 type, \ 588 checksum, \ 589 compress, \ 590 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \ 591 BP_IS_GANG(bp) ? "gang" : "contiguous", \ 592 BP_GET_DEDUP(bp) ? "dedup" : "unique", \ 593 copyname[copies], \ 594 ws, \ 595 (u_longlong_t)BP_GET_LSIZE(bp), \ 596 (u_longlong_t)BP_GET_PSIZE(bp), \ 597 (u_longlong_t)bp->blk_birth, \ 598 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \ 599 (u_longlong_t)BP_GET_FILL(bp), \ 600 ws, \ 601 (u_longlong_t)bp->blk_cksum.zc_word[0], \ 602 (u_longlong_t)bp->blk_cksum.zc_word[1], \ 603 (u_longlong_t)bp->blk_cksum.zc_word[2], \ 604 (u_longlong_t)bp->blk_cksum.zc_word[3]); \ 605 } \ 606 ASSERT(len < size); \ 607} 608 609#include <sys/dmu.h> 610 611#define BP_GET_BUFC_TYPE(bp) \ 612 (((BP_GET_LEVEL(bp) > 0) || (DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))) ? \ 613 ARC_BUFC_METADATA : ARC_BUFC_DATA) 614 615typedef enum spa_import_type { 616 SPA_IMPORT_EXISTING, 617 SPA_IMPORT_ASSEMBLE 618} spa_import_type_t; 619 620/* state manipulation functions */ 621extern int spa_open(const char *pool, spa_t **, void *tag); 622extern int spa_open_rewind(const char *pool, spa_t **, void *tag, 623 nvlist_t *policy, nvlist_t **config); 624extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot, 625 size_t buflen); 626extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props, 627 nvlist_t *zplprops); 628#if defined(sun) 629extern int spa_import_rootpool(char *devpath, char *devid); 630#else 631extern int spa_import_rootpool(const char *name); 632#endif 633extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props, 634 uint64_t flags); 635extern nvlist_t *spa_tryimport(nvlist_t *tryconfig); 636extern int spa_destroy(char *pool); 637extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force, 638 boolean_t hardforce); 639extern int spa_reset(char *pool); 640extern void spa_async_request(spa_t *spa, int flag); 641extern void spa_async_unrequest(spa_t *spa, int flag); 642extern void spa_async_suspend(spa_t *spa); 643extern void spa_async_resume(spa_t *spa); 644extern spa_t *spa_inject_addref(char *pool); 645extern void spa_inject_delref(spa_t *spa); 646extern void spa_scan_stat_init(spa_t *spa); 647extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps); 648 649#define SPA_ASYNC_CONFIG_UPDATE 0x01 650#define SPA_ASYNC_REMOVE 0x02 651#define SPA_ASYNC_PROBE 0x04 652#define SPA_ASYNC_RESILVER_DONE 0x08 653#define SPA_ASYNC_RESILVER 0x10 654#define SPA_ASYNC_AUTOEXPAND 0x20 655#define SPA_ASYNC_REMOVE_DONE 0x40 656#define SPA_ASYNC_REMOVE_STOP 0x80 657 658/* 659 * Controls the behavior of spa_vdev_remove(). 660 */ 661#define SPA_REMOVE_UNSPARE 0x01 662#define SPA_REMOVE_DONE 0x02 663 664/* device manipulation */ 665extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot); 666extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, 667 int replacing); 668extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, 669 int replace_done); 670extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare); 671extern boolean_t spa_vdev_remove_active(spa_t *spa); 672extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath); 673extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru); 674extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config, 675 nvlist_t *props, boolean_t exp); 676 677/* spare state (which is global across all pools) */ 678extern void spa_spare_add(vdev_t *vd); 679extern void spa_spare_remove(vdev_t *vd); 680extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt); 681extern void spa_spare_activate(vdev_t *vd); 682 683/* L2ARC state (which is global across all pools) */ 684extern void spa_l2cache_add(vdev_t *vd); 685extern void spa_l2cache_remove(vdev_t *vd); 686extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool); 687extern void spa_l2cache_activate(vdev_t *vd); 688extern void spa_l2cache_drop(spa_t *spa); 689 690/* scanning */ 691extern int spa_scan(spa_t *spa, pool_scan_func_t func); 692extern int spa_scan_stop(spa_t *spa); 693 694/* spa syncing */ 695extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */ 696extern void spa_sync_allpools(void); 697 698/* spa namespace global mutex */ 699extern kmutex_t spa_namespace_lock; 700 701/* 702 * SPA configuration functions in spa_config.c 703 */ 704 705#define SPA_CONFIG_UPDATE_POOL 0 706#define SPA_CONFIG_UPDATE_VDEVS 1 707 708extern void spa_config_sync(spa_t *, boolean_t, boolean_t); 709extern void spa_config_load(void); 710extern nvlist_t *spa_all_configs(uint64_t *); 711extern void spa_config_set(spa_t *spa, nvlist_t *config); 712extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, 713 int getstats); 714extern void spa_config_update(spa_t *spa, int what); 715 716/* 717 * Miscellaneous SPA routines in spa_misc.c 718 */ 719 720/* Namespace manipulation */ 721extern spa_t *spa_lookup(const char *name); 722extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot); 723extern void spa_remove(spa_t *spa); 724extern spa_t *spa_next(spa_t *prev); 725 726/* Refcount functions */ 727extern void spa_open_ref(spa_t *spa, void *tag); 728extern void spa_close(spa_t *spa, void *tag); 729extern void spa_async_close(spa_t *spa, void *tag); 730extern boolean_t spa_refcount_zero(spa_t *spa); 731 732#define SCL_NONE 0x00 733#define SCL_CONFIG 0x01 734#define SCL_STATE 0x02 735#define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */ 736#define SCL_ALLOC 0x08 737#define SCL_ZIO 0x10 738#define SCL_FREE 0x20 739#define SCL_VDEV 0x40 740#define SCL_LOCKS 7 741#define SCL_ALL ((1 << SCL_LOCKS) - 1) 742#define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO) 743 744/* Pool configuration locks */ 745extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw); 746extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw); 747extern void spa_config_exit(spa_t *spa, int locks, void *tag); 748extern int spa_config_held(spa_t *spa, int locks, krw_t rw); 749 750/* Pool vdev add/remove lock */ 751extern uint64_t spa_vdev_enter(spa_t *spa); 752extern uint64_t spa_vdev_config_enter(spa_t *spa); 753extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, 754 int error, char *tag); 755extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error); 756 757/* Pool vdev state change lock */ 758extern void spa_vdev_state_enter(spa_t *spa, int oplock); 759extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error); 760 761/* Log state */ 762typedef enum spa_log_state { 763 SPA_LOG_UNKNOWN = 0, /* unknown log state */ 764 SPA_LOG_MISSING, /* missing log(s) */ 765 SPA_LOG_CLEAR, /* clear the log(s) */ 766 SPA_LOG_GOOD, /* log(s) are good */ 767} spa_log_state_t; 768 769extern spa_log_state_t spa_get_log_state(spa_t *spa); 770extern void spa_set_log_state(spa_t *spa, spa_log_state_t state); 771extern int spa_offline_log(spa_t *spa); 772 773/* Log claim callback */ 774extern void spa_claim_notify(zio_t *zio); 775 776/* Accessor functions */ 777extern boolean_t spa_shutting_down(spa_t *spa); 778extern struct dsl_pool *spa_get_dsl(spa_t *spa); 779extern boolean_t spa_is_initializing(spa_t *spa); 780extern blkptr_t *spa_get_rootblkptr(spa_t *spa); 781extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp); 782extern void spa_altroot(spa_t *, char *, size_t); 783extern int spa_sync_pass(spa_t *spa); 784extern char *spa_name(spa_t *spa); 785extern uint64_t spa_guid(spa_t *spa); 786extern uint64_t spa_load_guid(spa_t *spa); 787extern uint64_t spa_last_synced_txg(spa_t *spa); 788extern uint64_t spa_first_txg(spa_t *spa); 789extern uint64_t spa_syncing_txg(spa_t *spa); 790extern uint64_t spa_version(spa_t *spa); 791extern pool_state_t spa_state(spa_t *spa); 792extern spa_load_state_t spa_load_state(spa_t *spa); 793extern uint64_t spa_freeze_txg(spa_t *spa); 794extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize); 795extern uint64_t spa_get_dspace(spa_t *spa); 796extern uint64_t spa_get_slop_space(spa_t *spa); 797extern void spa_update_dspace(spa_t *spa); 798extern uint64_t spa_version(spa_t *spa); 799extern boolean_t spa_deflate(spa_t *spa); 800extern metaslab_class_t *spa_normal_class(spa_t *spa); 801extern metaslab_class_t *spa_log_class(spa_t *spa); 802extern void spa_evicting_os_register(spa_t *, objset_t *os); 803extern void spa_evicting_os_deregister(spa_t *, objset_t *os); 804extern void spa_evicting_os_wait(spa_t *spa); 805extern int spa_max_replication(spa_t *spa); 806extern int spa_prev_software_version(spa_t *spa); 807extern int spa_busy(void); 808extern uint8_t spa_get_failmode(spa_t *spa); 809extern boolean_t spa_suspended(spa_t *spa); 810extern uint64_t spa_bootfs(spa_t *spa); 811extern uint64_t spa_delegation(spa_t *spa); 812extern objset_t *spa_meta_objset(spa_t *spa); 813extern uint64_t spa_deadman_synctime(spa_t *spa); 814 815/* Miscellaneous support routines */ 816extern void spa_activate_mos_feature(spa_t *spa, const char *feature, 817 dmu_tx_t *tx); 818extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature); 819extern int spa_rename(const char *oldname, const char *newname); 820extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid); 821extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid); 822extern char *spa_strdup(const char *); 823extern void spa_strfree(char *); 824extern uint64_t spa_get_random(uint64_t range); 825extern uint64_t spa_generate_guid(spa_t *spa); 826extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp); 827extern void spa_freeze(spa_t *spa); 828extern int spa_change_guid(spa_t *spa); 829extern void spa_upgrade(spa_t *spa, uint64_t version); 830extern void spa_evict_all(void); 831extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid, 832 boolean_t l2cache); 833extern boolean_t spa_has_spare(spa_t *, uint64_t guid); 834extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva); 835extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp); 836extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp); 837extern boolean_t spa_has_slogs(spa_t *spa); 838extern boolean_t spa_is_root(spa_t *spa); 839extern boolean_t spa_writeable(spa_t *spa); 840extern boolean_t spa_has_pending_synctask(spa_t *spa); 841extern int spa_maxblocksize(spa_t *spa); 842extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp); 843 844extern int spa_mode(spa_t *spa); 845extern uint64_t zfs_strtonum(const char *str, char **nptr); 846#define strtonum(str, nptr) zfs_strtonum((str), (nptr)) 847 848extern char *spa_his_ievent_table[]; 849 850extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx); 851extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read, 852 char *his_buf); 853extern int spa_history_log(spa_t *spa, const char *his_buf); 854extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl); 855extern void spa_history_log_version(spa_t *spa, const char *operation); 856extern void spa_history_log_internal(spa_t *spa, const char *operation, 857 dmu_tx_t *tx, const char *fmt, ...); 858extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op, 859 dmu_tx_t *tx, const char *fmt, ...); 860extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation, 861 dmu_tx_t *tx, const char *fmt, ...); 862 863/* error handling */ 864struct zbookmark_phys; 865extern void spa_log_error(spa_t *spa, zio_t *zio); 866extern void zfs_ereport_post(const char *cls, spa_t *spa, vdev_t *vd, 867 zio_t *zio, uint64_t stateoroffset, uint64_t length); 868extern void zfs_post_remove(spa_t *spa, vdev_t *vd); 869extern void zfs_post_state_change(spa_t *spa, vdev_t *vd); 870extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd); 871extern uint64_t spa_get_errlog_size(spa_t *spa); 872extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count); 873extern void spa_errlog_rotate(spa_t *spa); 874extern void spa_errlog_drain(spa_t *spa); 875extern void spa_errlog_sync(spa_t *spa, uint64_t txg); 876extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub); 877 878/* vdev cache */ 879extern void vdev_cache_stat_init(void); 880extern void vdev_cache_stat_fini(void); 881 882/* Initialization and termination */ 883extern void spa_init(int flags); 884extern void spa_fini(void); 885extern void spa_boot_init(); 886 887/* properties */ 888extern int spa_prop_set(spa_t *spa, nvlist_t *nvp); 889extern int spa_prop_get(spa_t *spa, nvlist_t **nvp); 890extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx); 891extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t); 892 893/* asynchronous event notification */ 894extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name); 895 896#ifdef ZFS_DEBUG 897#define dprintf_bp(bp, fmt, ...) do { \ 898 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 899 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \ 900 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \ 901 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \ 902 kmem_free(__blkbuf, BP_SPRINTF_LEN); \ 903 } \ 904_NOTE(CONSTCOND) } while (0) 905#else 906#define dprintf_bp(bp, fmt, ...) 907#endif 908 909extern boolean_t spa_debug_enabled(spa_t *spa); 910#define spa_dbgmsg(spa, ...) \ 911{ \ 912 if (spa_debug_enabled(spa)) \ 913 zfs_dbgmsg(__VA_ARGS__); \ 914} 915 916extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */ 917 918#ifdef __cplusplus 919} 920#endif 921 922#endif /* _SYS_SPA_H */ 923