spa.h revision 288549
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 */ 27 28#ifndef _SYS_SPA_H 29#define _SYS_SPA_H 30 31#include <sys/avl.h> 32#include <sys/zfs_context.h> 33#include <sys/nvpair.h> 34#include <sys/sysmacros.h> 35#include <sys/types.h> 36#include <sys/fs/zfs.h> 37 38#ifdef __cplusplus 39extern "C" { 40#endif 41 42/* 43 * Forward references that lots of things need. 44 */ 45typedef struct spa spa_t; 46typedef struct vdev vdev_t; 47typedef struct metaslab metaslab_t; 48typedef struct metaslab_group metaslab_group_t; 49typedef struct metaslab_class metaslab_class_t; 50typedef struct zio zio_t; 51typedef struct zilog zilog_t; 52typedef struct spa_aux_vdev spa_aux_vdev_t; 53typedef struct ddt ddt_t; 54typedef struct ddt_entry ddt_entry_t; 55struct dsl_pool; 56struct dsl_dataset; 57 58/* 59 * General-purpose 32-bit and 64-bit bitfield encodings. 60 */ 61#define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) 62#define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) 63#define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) 64#define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) 65 66#define BF32_GET(x, low, len) BF32_DECODE(x, low, len) 67#define BF64_GET(x, low, len) BF64_DECODE(x, low, len) 68 69#define BF32_SET(x, low, len, val) do { \ 70 ASSERT3U(val, <, 1U << (len)); \ 71 ASSERT3U(low + len, <=, 32); \ 72 (x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \ 73_NOTE(CONSTCOND) } while (0) 74 75#define BF64_SET(x, low, len, val) do { \ 76 ASSERT3U(val, <, 1ULL << (len)); \ 77 ASSERT3U(low + len, <=, 64); \ 78 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \ 79_NOTE(CONSTCOND) } while (0) 80 81#define BF32_GET_SB(x, low, len, shift, bias) \ 82 ((BF32_GET(x, low, len) + (bias)) << (shift)) 83#define BF64_GET_SB(x, low, len, shift, bias) \ 84 ((BF64_GET(x, low, len) + (bias)) << (shift)) 85 86#define BF32_SET_SB(x, low, len, shift, bias, val) do { \ 87 ASSERT(IS_P2ALIGNED(val, 1U << shift)); \ 88 ASSERT3S((val) >> (shift), >=, bias); \ 89 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \ 90_NOTE(CONSTCOND) } while (0) 91#define BF64_SET_SB(x, low, len, shift, bias, val) do { \ 92 ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \ 93 ASSERT3S((val) >> (shift), >=, bias); \ 94 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \ 95_NOTE(CONSTCOND) } while (0) 96 97/* 98 * We currently support block sizes from 512 bytes to 16MB. 99 * The benefits of larger blocks, and thus larger IO, need to be weighed 100 * against the cost of COWing a giant block to modify one byte, and the 101 * large latency of reading or writing a large block. 102 * 103 * Note that although blocks up to 16MB are supported, the recordsize 104 * property can not be set larger than zfs_max_recordsize (default 1MB). 105 * See the comment near zfs_max_recordsize in dsl_dataset.c for details. 106 * 107 * Note that although the LSIZE field of the blkptr_t can store sizes up 108 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to 109 * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB. 110 */ 111#define SPA_MINBLOCKSHIFT 9 112#define SPA_OLD_MAXBLOCKSHIFT 17 113#define SPA_MAXBLOCKSHIFT 24 114#define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 115#define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT) 116#define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 117 118/* 119 * Default maximum supported logical ashift. 120 * 121 * The current 8k allocation block size limit is due to the 8k 122 * aligned/sized operations performed by vdev_probe() on 123 * vdev_label->vl_pad2. Using another "safe region" for these tests 124 * would allow the limit to be raised to 16k, at the expense of 125 * only having 8 available uberblocks in the label area. 126 */ 127#define SPA_MAXASHIFT 13 128 129/* 130 * Default minimum supported logical ashift. 131 */ 132#define SPA_MINASHIFT SPA_MINBLOCKSHIFT 133 134/* 135 * Size of block to hold the configuration data (a packed nvlist) 136 */ 137#define SPA_CONFIG_BLOCKSIZE (1ULL << 14) 138 139/* 140 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. 141 * The ASIZE encoding should be at least 64 times larger (6 more bits) 142 * to support up to 4-way RAID-Z mirror mode with worst-case gang block 143 * overhead, three DVAs per bp, plus one more bit in case we do anything 144 * else that expands the ASIZE. 145 */ 146#define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ 147#define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ 148#define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ 149 150/* 151 * All SPA data is represented by 128-bit data virtual addresses (DVAs). 152 * The members of the dva_t should be considered opaque outside the SPA. 153 */ 154typedef struct dva { 155 uint64_t dva_word[2]; 156} dva_t; 157 158/* 159 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. 160 */ 161typedef struct zio_cksum { 162 uint64_t zc_word[4]; 163} zio_cksum_t; 164 165/* 166 * Each block is described by its DVAs, time of birth, checksum, etc. 167 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 168 * 169 * 64 56 48 40 32 24 16 8 0 170 * +-------+-------+-------+-------+-------+-------+-------+-------+ 171 * 0 | vdev1 | GRID | ASIZE | 172 * +-------+-------+-------+-------+-------+-------+-------+-------+ 173 * 1 |G| offset1 | 174 * +-------+-------+-------+-------+-------+-------+-------+-------+ 175 * 2 | vdev2 | GRID | ASIZE | 176 * +-------+-------+-------+-------+-------+-------+-------+-------+ 177 * 3 |G| offset2 | 178 * +-------+-------+-------+-------+-------+-------+-------+-------+ 179 * 4 | vdev3 | GRID | ASIZE | 180 * +-------+-------+-------+-------+-------+-------+-------+-------+ 181 * 5 |G| offset3 | 182 * +-------+-------+-------+-------+-------+-------+-------+-------+ 183 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 184 * +-------+-------+-------+-------+-------+-------+-------+-------+ 185 * 7 | padding | 186 * +-------+-------+-------+-------+-------+-------+-------+-------+ 187 * 8 | padding | 188 * +-------+-------+-------+-------+-------+-------+-------+-------+ 189 * 9 | physical birth txg | 190 * +-------+-------+-------+-------+-------+-------+-------+-------+ 191 * a | logical birth txg | 192 * +-------+-------+-------+-------+-------+-------+-------+-------+ 193 * b | fill count | 194 * +-------+-------+-------+-------+-------+-------+-------+-------+ 195 * c | checksum[0] | 196 * +-------+-------+-------+-------+-------+-------+-------+-------+ 197 * d | checksum[1] | 198 * +-------+-------+-------+-------+-------+-------+-------+-------+ 199 * e | checksum[2] | 200 * +-------+-------+-------+-------+-------+-------+-------+-------+ 201 * f | checksum[3] | 202 * +-------+-------+-------+-------+-------+-------+-------+-------+ 203 * 204 * Legend: 205 * 206 * vdev virtual device ID 207 * offset offset into virtual device 208 * LSIZE logical size 209 * PSIZE physical size (after compression) 210 * ASIZE allocated size (including RAID-Z parity and gang block headers) 211 * GRID RAID-Z layout information (reserved for future use) 212 * cksum checksum function 213 * comp compression function 214 * G gang block indicator 215 * B byteorder (endianness) 216 * D dedup 217 * X encryption (on version 30, which is not supported) 218 * E blkptr_t contains embedded data (see below) 219 * lvl level of indirection 220 * type DMU object type 221 * phys birth txg of block allocation; zero if same as logical birth txg 222 * log. birth transaction group in which the block was logically born 223 * fill count number of non-zero blocks under this bp 224 * checksum[4] 256-bit checksum of the data this bp describes 225 */ 226 227/* 228 * "Embedded" blkptr_t's don't actually point to a block, instead they 229 * have a data payload embedded in the blkptr_t itself. See the comment 230 * in blkptr.c for more details. 231 * 232 * The blkptr_t is laid out as follows: 233 * 234 * 64 56 48 40 32 24 16 8 0 235 * +-------+-------+-------+-------+-------+-------+-------+-------+ 236 * 0 | payload | 237 * 1 | payload | 238 * 2 | payload | 239 * 3 | payload | 240 * 4 | payload | 241 * 5 | payload | 242 * +-------+-------+-------+-------+-------+-------+-------+-------+ 243 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 244 * +-------+-------+-------+-------+-------+-------+-------+-------+ 245 * 7 | payload | 246 * 8 | payload | 247 * 9 | payload | 248 * +-------+-------+-------+-------+-------+-------+-------+-------+ 249 * a | logical birth txg | 250 * +-------+-------+-------+-------+-------+-------+-------+-------+ 251 * b | payload | 252 * c | payload | 253 * d | payload | 254 * e | payload | 255 * f | payload | 256 * +-------+-------+-------+-------+-------+-------+-------+-------+ 257 * 258 * Legend: 259 * 260 * payload contains the embedded data 261 * B (byteorder) byteorder (endianness) 262 * D (dedup) padding (set to zero) 263 * X encryption (set to zero; see above) 264 * E (embedded) set to one 265 * lvl indirection level 266 * type DMU object type 267 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 268 * comp compression function of payload 269 * PSIZE size of payload after compression, in bytes 270 * LSIZE logical size of payload, in bytes 271 * note that 25 bits is enough to store the largest 272 * "normal" BP's LSIZE (2^16 * 2^9) in bytes 273 * log. birth transaction group in which the block was logically born 274 * 275 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 276 * bp's they are stored in units of SPA_MINBLOCKSHIFT. 277 * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 278 * The B, D, X, lvl, type, and comp fields are stored the same as with normal 279 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 280 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 281 * other macros, as they assert that they are only used on BP's of the correct 282 * "embedded-ness". 283 */ 284 285#define BPE_GET_ETYPE(bp) \ 286 (ASSERT(BP_IS_EMBEDDED(bp)), \ 287 BF64_GET((bp)->blk_prop, 40, 8)) 288#define BPE_SET_ETYPE(bp, t) do { \ 289 ASSERT(BP_IS_EMBEDDED(bp)); \ 290 BF64_SET((bp)->blk_prop, 40, 8, t); \ 291_NOTE(CONSTCOND) } while (0) 292 293#define BPE_GET_LSIZE(bp) \ 294 (ASSERT(BP_IS_EMBEDDED(bp)), \ 295 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 296#define BPE_SET_LSIZE(bp, x) do { \ 297 ASSERT(BP_IS_EMBEDDED(bp)); \ 298 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 299_NOTE(CONSTCOND) } while (0) 300 301#define BPE_GET_PSIZE(bp) \ 302 (ASSERT(BP_IS_EMBEDDED(bp)), \ 303 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 304#define BPE_SET_PSIZE(bp, x) do { \ 305 ASSERT(BP_IS_EMBEDDED(bp)); \ 306 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 307_NOTE(CONSTCOND) } while (0) 308 309typedef enum bp_embedded_type { 310 BP_EMBEDDED_TYPE_DATA, 311 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ 312 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED 313} bp_embedded_type_t; 314 315#define BPE_NUM_WORDS 14 316#define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 317#define BPE_IS_PAYLOADWORD(bp, wp) \ 318 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 319 320#define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 321#define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 322 323/* 324 * A block is a hole when it has either 1) never been written to, or 325 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads 326 * without physically allocating disk space. Holes are represented in the 327 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is 328 * done through the BP_IS_HOLE macro. For holes, the logical size, level, 329 * DMU object type, and birth times are all also stored for holes that 330 * were written to at some point (i.e. were punched after having been filled). 331 */ 332typedef struct blkptr { 333 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 334 uint64_t blk_prop; /* size, compression, type, etc */ 335 uint64_t blk_pad[2]; /* Extra space for the future */ 336 uint64_t blk_phys_birth; /* txg when block was allocated */ 337 uint64_t blk_birth; /* transaction group at birth */ 338 uint64_t blk_fill; /* fill count */ 339 zio_cksum_t blk_cksum; /* 256-bit checksum */ 340} blkptr_t; 341 342/* 343 * Macros to get and set fields in a bp or DVA. 344 */ 345#define DVA_GET_ASIZE(dva) \ 346 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 347#define DVA_SET_ASIZE(dva, x) \ 348 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 349 SPA_MINBLOCKSHIFT, 0, x) 350 351#define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 352#define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 353 354#define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) 355#define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) 356 357#define DVA_GET_OFFSET(dva) \ 358 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 359#define DVA_SET_OFFSET(dva, x) \ 360 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 361 362#define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 363#define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 364 365#define BP_GET_LSIZE(bp) \ 366 (BP_IS_EMBEDDED(bp) ? \ 367 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 368 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 369#define BP_SET_LSIZE(bp, x) do { \ 370 ASSERT(!BP_IS_EMBEDDED(bp)); \ 371 BF64_SET_SB((bp)->blk_prop, \ 372 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 373_NOTE(CONSTCOND) } while (0) 374 375#define BP_GET_PSIZE(bp) \ 376 (BP_IS_EMBEDDED(bp) ? 0 : \ 377 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 378#define BP_SET_PSIZE(bp, x) do { \ 379 ASSERT(!BP_IS_EMBEDDED(bp)); \ 380 BF64_SET_SB((bp)->blk_prop, \ 381 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 382_NOTE(CONSTCOND) } while (0) 383 384#define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7) 385#define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x) 386 387#define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 388#define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x) 389 390#define BP_GET_CHECKSUM(bp) \ 391 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \ 392 BF64_GET((bp)->blk_prop, 40, 8)) 393#define BP_SET_CHECKSUM(bp, x) do { \ 394 ASSERT(!BP_IS_EMBEDDED(bp)); \ 395 BF64_SET((bp)->blk_prop, 40, 8, x); \ 396_NOTE(CONSTCOND) } while (0) 397 398#define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 399#define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 400 401#define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 402#define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 403 404#define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 405#define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 406 407#define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 408#define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 409 410#define BP_PHYSICAL_BIRTH(bp) \ 411 (BP_IS_EMBEDDED(bp) ? 0 : \ 412 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 413 414#define BP_SET_BIRTH(bp, logical, physical) \ 415{ \ 416 ASSERT(!BP_IS_EMBEDDED(bp)); \ 417 (bp)->blk_birth = (logical); \ 418 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \ 419} 420 421#define BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill) 422 423#define BP_GET_ASIZE(bp) \ 424 (BP_IS_EMBEDDED(bp) ? 0 : \ 425 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 426 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 427 DVA_GET_ASIZE(&(bp)->blk_dva[2])) 428 429#define BP_GET_UCSIZE(bp) \ 430 ((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \ 431 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)) 432 433#define BP_GET_NDVAS(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_COUNT_GANG(bp) \ 440 (BP_IS_EMBEDDED(bp) ? 0 : \ 441 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \ 442 DVA_GET_GANG(&(bp)->blk_dva[1]) + \ 443 DVA_GET_GANG(&(bp)->blk_dva[2]))) 444 445#define DVA_EQUAL(dva1, dva2) \ 446 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 447 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 448 449#define BP_EQUAL(bp1, bp2) \ 450 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \ 451 (bp1)->blk_birth == (bp2)->blk_birth && \ 452 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \ 453 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \ 454 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2])) 455 456#define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 457 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 458 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 459 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 460 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 461 462#define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 463 464#define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ 465{ \ 466 (zcp)->zc_word[0] = w0; \ 467 (zcp)->zc_word[1] = w1; \ 468 (zcp)->zc_word[2] = w2; \ 469 (zcp)->zc_word[3] = w3; \ 470} 471 472#define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0]) 473#define BP_IS_GANG(bp) \ 474 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp))) 475#define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 476 (dva)->dva_word[1] == 0ULL) 477#define BP_IS_HOLE(bp) \ 478 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp))) 479 480/* BP_IS_RAIDZ(bp) assumes no block compression */ 481#define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \ 482 BP_GET_PSIZE(bp)) 483 484#define BP_ZERO(bp) \ 485{ \ 486 (bp)->blk_dva[0].dva_word[0] = 0; \ 487 (bp)->blk_dva[0].dva_word[1] = 0; \ 488 (bp)->blk_dva[1].dva_word[0] = 0; \ 489 (bp)->blk_dva[1].dva_word[1] = 0; \ 490 (bp)->blk_dva[2].dva_word[0] = 0; \ 491 (bp)->blk_dva[2].dva_word[1] = 0; \ 492 (bp)->blk_prop = 0; \ 493 (bp)->blk_pad[0] = 0; \ 494 (bp)->blk_pad[1] = 0; \ 495 (bp)->blk_phys_birth = 0; \ 496 (bp)->blk_birth = 0; \ 497 (bp)->blk_fill = 0; \ 498 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 499} 500 501#if BYTE_ORDER == _BIG_ENDIAN 502#define ZFS_HOST_BYTEORDER (0ULL) 503#else 504#define ZFS_HOST_BYTEORDER (1ULL) 505#endif 506 507#define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) 508 509#define BP_SPRINTF_LEN 320 510 511/* 512 * This macro allows code sharing between zfs, libzpool, and mdb. 513 * 'func' is either snprintf() or mdb_snprintf(). 514 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line. 515 */ 516#define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \ 517{ \ 518 static const char *copyname[] = \ 519 { "zero", "single", "double", "triple" }; \ 520 int len = 0; \ 521 int copies = 0; \ 522 \ 523 if (bp == NULL) { \ 524 len += func(buf + len, size - len, "<NULL>"); \ 525 } else if (BP_IS_HOLE(bp)) { \ 526 len += func(buf + len, size - len, "<hole>"); \ 527 if (bp->blk_birth > 0) { \ 528 len += func(buf + len, size - len, \ 529 " birth=%lluL", \ 530 (u_longlong_t)bp->blk_birth); \ 531 } \ 532 } else if (BP_IS_EMBEDDED(bp)) { \ 533 len = func(buf + len, size - len, \ 534 "EMBEDDED [L%llu %s] et=%u %s " \ 535 "size=%llxL/%llxP birth=%lluL", \ 536 (u_longlong_t)BP_GET_LEVEL(bp), \ 537 type, \ 538 (int)BPE_GET_ETYPE(bp), \ 539 compress, \ 540 (u_longlong_t)BPE_GET_LSIZE(bp), \ 541 (u_longlong_t)BPE_GET_PSIZE(bp), \ 542 (u_longlong_t)bp->blk_birth); \ 543 } else { \ 544 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \ 545 const dva_t *dva = &bp->blk_dva[d]; \ 546 if (DVA_IS_VALID(dva)) \ 547 copies++; \ 548 len += func(buf + len, size - len, \ 549 "DVA[%d]=<%llu:%llx:%llx>%c", d, \ 550 (u_longlong_t)DVA_GET_VDEV(dva), \ 551 (u_longlong_t)DVA_GET_OFFSET(dva), \ 552 (u_longlong_t)DVA_GET_ASIZE(dva), \ 553 ws); \ 554 } \ 555 if (BP_IS_GANG(bp) && \ 556 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \ 557 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \ 558 copies--; \ 559 len += func(buf + len, size - len, \ 560 "[L%llu %s] %s %s %s %s %s %s%c" \ 561 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \ 562 "cksum=%llx:%llx:%llx:%llx", \ 563 (u_longlong_t)BP_GET_LEVEL(bp), \ 564 type, \ 565 checksum, \ 566 compress, \ 567 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \ 568 BP_IS_GANG(bp) ? "gang" : "contiguous", \ 569 BP_GET_DEDUP(bp) ? "dedup" : "unique", \ 570 copyname[copies], \ 571 ws, \ 572 (u_longlong_t)BP_GET_LSIZE(bp), \ 573 (u_longlong_t)BP_GET_PSIZE(bp), \ 574 (u_longlong_t)bp->blk_birth, \ 575 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \ 576 (u_longlong_t)BP_GET_FILL(bp), \ 577 ws, \ 578 (u_longlong_t)bp->blk_cksum.zc_word[0], \ 579 (u_longlong_t)bp->blk_cksum.zc_word[1], \ 580 (u_longlong_t)bp->blk_cksum.zc_word[2], \ 581 (u_longlong_t)bp->blk_cksum.zc_word[3]); \ 582 } \ 583 ASSERT(len < size); \ 584} 585 586#include <sys/dmu.h> 587 588#define BP_GET_BUFC_TYPE(bp) \ 589 (((BP_GET_LEVEL(bp) > 0) || (DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))) ? \ 590 ARC_BUFC_METADATA : ARC_BUFC_DATA) 591 592typedef enum spa_import_type { 593 SPA_IMPORT_EXISTING, 594 SPA_IMPORT_ASSEMBLE 595} spa_import_type_t; 596 597/* state manipulation functions */ 598extern int spa_open(const char *pool, spa_t **, void *tag); 599extern int spa_open_rewind(const char *pool, spa_t **, void *tag, 600 nvlist_t *policy, nvlist_t **config); 601extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot, 602 size_t buflen); 603extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props, 604 nvlist_t *zplprops); 605#if defined(sun) 606extern int spa_import_rootpool(char *devpath, char *devid); 607#else 608extern int spa_import_rootpool(const char *name); 609#endif 610extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props, 611 uint64_t flags); 612extern nvlist_t *spa_tryimport(nvlist_t *tryconfig); 613extern int spa_destroy(char *pool); 614extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force, 615 boolean_t hardforce); 616extern int spa_reset(char *pool); 617extern void spa_async_request(spa_t *spa, int flag); 618extern void spa_async_unrequest(spa_t *spa, int flag); 619extern void spa_async_suspend(spa_t *spa); 620extern void spa_async_resume(spa_t *spa); 621extern spa_t *spa_inject_addref(char *pool); 622extern void spa_inject_delref(spa_t *spa); 623extern void spa_scan_stat_init(spa_t *spa); 624extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps); 625 626#define SPA_ASYNC_CONFIG_UPDATE 0x01 627#define SPA_ASYNC_REMOVE 0x02 628#define SPA_ASYNC_PROBE 0x04 629#define SPA_ASYNC_RESILVER_DONE 0x08 630#define SPA_ASYNC_RESILVER 0x10 631#define SPA_ASYNC_AUTOEXPAND 0x20 632#define SPA_ASYNC_REMOVE_DONE 0x40 633#define SPA_ASYNC_REMOVE_STOP 0x80 634 635/* 636 * Controls the behavior of spa_vdev_remove(). 637 */ 638#define SPA_REMOVE_UNSPARE 0x01 639#define SPA_REMOVE_DONE 0x02 640 641/* device manipulation */ 642extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot); 643extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, 644 int replacing); 645extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, 646 int replace_done); 647extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare); 648extern boolean_t spa_vdev_remove_active(spa_t *spa); 649extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath); 650extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru); 651extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config, 652 nvlist_t *props, boolean_t exp); 653 654/* spare state (which is global across all pools) */ 655extern void spa_spare_add(vdev_t *vd); 656extern void spa_spare_remove(vdev_t *vd); 657extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt); 658extern void spa_spare_activate(vdev_t *vd); 659 660/* L2ARC state (which is global across all pools) */ 661extern void spa_l2cache_add(vdev_t *vd); 662extern void spa_l2cache_remove(vdev_t *vd); 663extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool); 664extern void spa_l2cache_activate(vdev_t *vd); 665extern void spa_l2cache_drop(spa_t *spa); 666 667/* scanning */ 668extern int spa_scan(spa_t *spa, pool_scan_func_t func); 669extern int spa_scan_stop(spa_t *spa); 670 671/* spa syncing */ 672extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */ 673extern void spa_sync_allpools(void); 674 675/* spa namespace global mutex */ 676extern kmutex_t spa_namespace_lock; 677 678/* 679 * SPA configuration functions in spa_config.c 680 */ 681 682#define SPA_CONFIG_UPDATE_POOL 0 683#define SPA_CONFIG_UPDATE_VDEVS 1 684 685extern void spa_config_sync(spa_t *, boolean_t, boolean_t); 686extern void spa_config_load(void); 687extern nvlist_t *spa_all_configs(uint64_t *); 688extern void spa_config_set(spa_t *spa, nvlist_t *config); 689extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, 690 int getstats); 691extern void spa_config_update(spa_t *spa, int what); 692 693/* 694 * Miscellaneous SPA routines in spa_misc.c 695 */ 696 697/* Namespace manipulation */ 698extern spa_t *spa_lookup(const char *name); 699extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot); 700extern void spa_remove(spa_t *spa); 701extern spa_t *spa_next(spa_t *prev); 702 703/* Refcount functions */ 704extern void spa_open_ref(spa_t *spa, void *tag); 705extern void spa_close(spa_t *spa, void *tag); 706extern void spa_async_close(spa_t *spa, void *tag); 707extern boolean_t spa_refcount_zero(spa_t *spa); 708 709#define SCL_NONE 0x00 710#define SCL_CONFIG 0x01 711#define SCL_STATE 0x02 712#define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */ 713#define SCL_ALLOC 0x08 714#define SCL_ZIO 0x10 715#define SCL_FREE 0x20 716#define SCL_VDEV 0x40 717#define SCL_LOCKS 7 718#define SCL_ALL ((1 << SCL_LOCKS) - 1) 719#define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO) 720 721/* Pool configuration locks */ 722extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw); 723extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw); 724extern void spa_config_exit(spa_t *spa, int locks, void *tag); 725extern int spa_config_held(spa_t *spa, int locks, krw_t rw); 726 727/* Pool vdev add/remove lock */ 728extern uint64_t spa_vdev_enter(spa_t *spa); 729extern uint64_t spa_vdev_config_enter(spa_t *spa); 730extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, 731 int error, char *tag); 732extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error); 733 734/* Pool vdev state change lock */ 735extern void spa_vdev_state_enter(spa_t *spa, int oplock); 736extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error); 737 738/* Log state */ 739typedef enum spa_log_state { 740 SPA_LOG_UNKNOWN = 0, /* unknown log state */ 741 SPA_LOG_MISSING, /* missing log(s) */ 742 SPA_LOG_CLEAR, /* clear the log(s) */ 743 SPA_LOG_GOOD, /* log(s) are good */ 744} spa_log_state_t; 745 746extern spa_log_state_t spa_get_log_state(spa_t *spa); 747extern void spa_set_log_state(spa_t *spa, spa_log_state_t state); 748extern int spa_offline_log(spa_t *spa); 749 750/* Log claim callback */ 751extern void spa_claim_notify(zio_t *zio); 752 753/* Accessor functions */ 754extern boolean_t spa_shutting_down(spa_t *spa); 755extern struct dsl_pool *spa_get_dsl(spa_t *spa); 756extern boolean_t spa_is_initializing(spa_t *spa); 757extern blkptr_t *spa_get_rootblkptr(spa_t *spa); 758extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp); 759extern void spa_altroot(spa_t *, char *, size_t); 760extern int spa_sync_pass(spa_t *spa); 761extern char *spa_name(spa_t *spa); 762extern uint64_t spa_guid(spa_t *spa); 763extern uint64_t spa_load_guid(spa_t *spa); 764extern uint64_t spa_last_synced_txg(spa_t *spa); 765extern uint64_t spa_first_txg(spa_t *spa); 766extern uint64_t spa_syncing_txg(spa_t *spa); 767extern uint64_t spa_version(spa_t *spa); 768extern pool_state_t spa_state(spa_t *spa); 769extern spa_load_state_t spa_load_state(spa_t *spa); 770extern uint64_t spa_freeze_txg(spa_t *spa); 771extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize); 772extern uint64_t spa_get_dspace(spa_t *spa); 773extern uint64_t spa_get_slop_space(spa_t *spa); 774extern void spa_update_dspace(spa_t *spa); 775extern uint64_t spa_version(spa_t *spa); 776extern boolean_t spa_deflate(spa_t *spa); 777extern metaslab_class_t *spa_normal_class(spa_t *spa); 778extern metaslab_class_t *spa_log_class(spa_t *spa); 779extern void spa_evicting_os_register(spa_t *, objset_t *os); 780extern void spa_evicting_os_deregister(spa_t *, objset_t *os); 781extern void spa_evicting_os_wait(spa_t *spa); 782extern int spa_max_replication(spa_t *spa); 783extern int spa_prev_software_version(spa_t *spa); 784extern int spa_busy(void); 785extern uint8_t spa_get_failmode(spa_t *spa); 786extern boolean_t spa_suspended(spa_t *spa); 787extern uint64_t spa_bootfs(spa_t *spa); 788extern uint64_t spa_delegation(spa_t *spa); 789extern objset_t *spa_meta_objset(spa_t *spa); 790extern uint64_t spa_deadman_synctime(spa_t *spa); 791 792/* Miscellaneous support routines */ 793extern void spa_activate_mos_feature(spa_t *spa, const char *feature, 794 dmu_tx_t *tx); 795extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature); 796extern int spa_rename(const char *oldname, const char *newname); 797extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid); 798extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid); 799extern char *spa_strdup(const char *); 800extern void spa_strfree(char *); 801extern uint64_t spa_get_random(uint64_t range); 802extern uint64_t spa_generate_guid(spa_t *spa); 803extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp); 804extern void spa_freeze(spa_t *spa); 805extern int spa_change_guid(spa_t *spa); 806extern void spa_upgrade(spa_t *spa, uint64_t version); 807extern void spa_evict_all(void); 808extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid, 809 boolean_t l2cache); 810extern boolean_t spa_has_spare(spa_t *, uint64_t guid); 811extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva); 812extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp); 813extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp); 814extern boolean_t spa_has_slogs(spa_t *spa); 815extern boolean_t spa_is_root(spa_t *spa); 816extern boolean_t spa_writeable(spa_t *spa); 817extern boolean_t spa_has_pending_synctask(spa_t *spa); 818extern int spa_maxblocksize(spa_t *spa); 819extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp); 820 821extern int spa_mode(spa_t *spa); 822extern uint64_t zfs_strtonum(const char *str, char **nptr); 823#define strtonum(str, nptr) zfs_strtonum((str), (nptr)) 824 825extern char *spa_his_ievent_table[]; 826 827extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx); 828extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read, 829 char *his_buf); 830extern int spa_history_log(spa_t *spa, const char *his_buf); 831extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl); 832extern void spa_history_log_version(spa_t *spa, const char *operation); 833extern void spa_history_log_internal(spa_t *spa, const char *operation, 834 dmu_tx_t *tx, const char *fmt, ...); 835extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op, 836 dmu_tx_t *tx, const char *fmt, ...); 837extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation, 838 dmu_tx_t *tx, const char *fmt, ...); 839 840/* error handling */ 841struct zbookmark_phys; 842extern void spa_log_error(spa_t *spa, zio_t *zio); 843extern void zfs_ereport_post(const char *cls, spa_t *spa, vdev_t *vd, 844 zio_t *zio, uint64_t stateoroffset, uint64_t length); 845extern void zfs_post_remove(spa_t *spa, vdev_t *vd); 846extern void zfs_post_state_change(spa_t *spa, vdev_t *vd); 847extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd); 848extern uint64_t spa_get_errlog_size(spa_t *spa); 849extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count); 850extern void spa_errlog_rotate(spa_t *spa); 851extern void spa_errlog_drain(spa_t *spa); 852extern void spa_errlog_sync(spa_t *spa, uint64_t txg); 853extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub); 854 855/* vdev cache */ 856extern void vdev_cache_stat_init(void); 857extern void vdev_cache_stat_fini(void); 858 859/* Initialization and termination */ 860extern void spa_init(int flags); 861extern void spa_fini(void); 862extern void spa_boot_init(); 863 864/* properties */ 865extern int spa_prop_set(spa_t *spa, nvlist_t *nvp); 866extern int spa_prop_get(spa_t *spa, nvlist_t **nvp); 867extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx); 868extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t); 869 870/* asynchronous event notification */ 871extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name); 872 873#ifdef ZFS_DEBUG 874#define dprintf_bp(bp, fmt, ...) do { \ 875 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 876 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \ 877 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \ 878 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \ 879 kmem_free(__blkbuf, BP_SPRINTF_LEN); \ 880 } \ 881_NOTE(CONSTCOND) } while (0) 882#else 883#define dprintf_bp(bp, fmt, ...) 884#endif 885 886extern boolean_t spa_debug_enabled(spa_t *spa); 887#define spa_dbgmsg(spa, ...) \ 888{ \ 889 if (spa_debug_enabled(spa)) \ 890 zfs_dbgmsg(__VA_ARGS__); \ 891} 892 893extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */ 894 895#ifdef __cplusplus 896} 897#endif 898 899#endif /* _SYS_SPA_H */ 900