spa.h revision 276081
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 */ 26 27#ifndef _SYS_SPA_H 28#define _SYS_SPA_H 29 30#include <sys/avl.h> 31#include <sys/zfs_context.h> 32#include <sys/nvpair.h> 33#include <sys/sysmacros.h> 34#include <sys/types.h> 35#include <sys/fs/zfs.h> 36 37#ifdef __cplusplus 38extern "C" { 39#endif 40 41/* 42 * Forward references that lots of things need. 43 */ 44typedef struct spa spa_t; 45typedef struct vdev vdev_t; 46typedef struct metaslab metaslab_t; 47typedef struct metaslab_group metaslab_group_t; 48typedef struct metaslab_class metaslab_class_t; 49typedef struct zio zio_t; 50typedef struct zilog zilog_t; 51typedef struct spa_aux_vdev spa_aux_vdev_t; 52typedef struct ddt ddt_t; 53typedef struct ddt_entry ddt_entry_t; 54struct dsl_pool; 55struct dsl_dataset; 56 57/* 58 * General-purpose 32-bit and 64-bit bitfield encodings. 59 */ 60#define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) 61#define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) 62#define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) 63#define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) 64 65#define BF32_GET(x, low, len) BF32_DECODE(x, low, len) 66#define BF64_GET(x, low, len) BF64_DECODE(x, low, len) 67 68#define BF32_SET(x, low, len, val) do { \ 69 ASSERT3U(val, <, 1U << (len)); \ 70 ASSERT3U(low + len, <=, 32); \ 71 (x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \ 72_NOTE(CONSTCOND) } while (0) 73 74#define BF64_SET(x, low, len, val) do { \ 75 ASSERT3U(val, <, 1ULL << (len)); \ 76 ASSERT3U(low + len, <=, 64); \ 77 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \ 78_NOTE(CONSTCOND) } while (0) 79 80#define BF32_GET_SB(x, low, len, shift, bias) \ 81 ((BF32_GET(x, low, len) + (bias)) << (shift)) 82#define BF64_GET_SB(x, low, len, shift, bias) \ 83 ((BF64_GET(x, low, len) + (bias)) << (shift)) 84 85#define BF32_SET_SB(x, low, len, shift, bias, val) do { \ 86 ASSERT(IS_P2ALIGNED(val, 1U << shift)); \ 87 ASSERT3S((val) >> (shift), >=, bias); \ 88 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \ 89_NOTE(CONSTCOND) } while (0) 90#define BF64_SET_SB(x, low, len, shift, bias, val) do { \ 91 ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \ 92 ASSERT3S((val) >> (shift), >=, bias); \ 93 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \ 94_NOTE(CONSTCOND) } while (0) 95 96/* 97 * We currently support block sizes from 512 bytes to 16MB. 98 * The benefits of larger blocks, and thus larger IO, need to be weighed 99 * against the cost of COWing a giant block to modify one byte, and the 100 * large latency of reading or writing a large block. 101 * 102 * Note that although blocks up to 16MB are supported, the recordsize 103 * property can not be set larger than zfs_max_recordsize (default 1MB). 104 * See the comment near zfs_max_recordsize in dsl_dataset.c for details. 105 * 106 * Note that although the LSIZE field of the blkptr_t can store sizes up 107 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to 108 * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB. 109 */ 110#define SPA_MINBLOCKSHIFT 9 111#define SPA_OLD_MAXBLOCKSHIFT 17 112#define SPA_MAXBLOCKSHIFT 24 113#define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 114#define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT) 115#define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 116 117/* 118 * Default maximum supported logical ashift. 119 * 120 * The current 8k allocation block size limit is due to the 8k 121 * aligned/sized operations performed by vdev_probe() on 122 * vdev_label->vl_pad2. Using another "safe region" for these tests 123 * would allow the limit to be raised to 16k, at the expense of 124 * only having 8 available uberblocks in the label area. 125 */ 126#define SPA_MAXASHIFT 13 127 128/* 129 * Default minimum supported logical ashift. 130 */ 131#define SPA_MINASHIFT SPA_MINBLOCKSHIFT 132 133/* 134 * Size of block to hold the configuration data (a packed nvlist) 135 */ 136#define SPA_CONFIG_BLOCKSIZE (1ULL << 14) 137 138/* 139 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. 140 * The ASIZE encoding should be at least 64 times larger (6 more bits) 141 * to support up to 4-way RAID-Z mirror mode with worst-case gang block 142 * overhead, three DVAs per bp, plus one more bit in case we do anything 143 * else that expands the ASIZE. 144 */ 145#define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ 146#define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ 147#define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ 148 149/* 150 * All SPA data is represented by 128-bit data virtual addresses (DVAs). 151 * The members of the dva_t should be considered opaque outside the SPA. 152 */ 153typedef struct dva { 154 uint64_t dva_word[2]; 155} dva_t; 156 157/* 158 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. 159 */ 160typedef struct zio_cksum { 161 uint64_t zc_word[4]; 162} zio_cksum_t; 163 164/* 165 * Each block is described by its DVAs, time of birth, checksum, etc. 166 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 167 * 168 * 64 56 48 40 32 24 16 8 0 169 * +-------+-------+-------+-------+-------+-------+-------+-------+ 170 * 0 | vdev1 | GRID | ASIZE | 171 * +-------+-------+-------+-------+-------+-------+-------+-------+ 172 * 1 |G| offset1 | 173 * +-------+-------+-------+-------+-------+-------+-------+-------+ 174 * 2 | vdev2 | GRID | ASIZE | 175 * +-------+-------+-------+-------+-------+-------+-------+-------+ 176 * 3 |G| offset2 | 177 * +-------+-------+-------+-------+-------+-------+-------+-------+ 178 * 4 | vdev3 | GRID | ASIZE | 179 * +-------+-------+-------+-------+-------+-------+-------+-------+ 180 * 5 |G| offset3 | 181 * +-------+-------+-------+-------+-------+-------+-------+-------+ 182 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 183 * +-------+-------+-------+-------+-------+-------+-------+-------+ 184 * 7 | padding | 185 * +-------+-------+-------+-------+-------+-------+-------+-------+ 186 * 8 | padding | 187 * +-------+-------+-------+-------+-------+-------+-------+-------+ 188 * 9 | physical birth txg | 189 * +-------+-------+-------+-------+-------+-------+-------+-------+ 190 * a | logical birth txg | 191 * +-------+-------+-------+-------+-------+-------+-------+-------+ 192 * b | fill count | 193 * +-------+-------+-------+-------+-------+-------+-------+-------+ 194 * c | checksum[0] | 195 * +-------+-------+-------+-------+-------+-------+-------+-------+ 196 * d | checksum[1] | 197 * +-------+-------+-------+-------+-------+-------+-------+-------+ 198 * e | checksum[2] | 199 * +-------+-------+-------+-------+-------+-------+-------+-------+ 200 * f | checksum[3] | 201 * +-------+-------+-------+-------+-------+-------+-------+-------+ 202 * 203 * Legend: 204 * 205 * vdev virtual device ID 206 * offset offset into virtual device 207 * LSIZE logical size 208 * PSIZE physical size (after compression) 209 * ASIZE allocated size (including RAID-Z parity and gang block headers) 210 * GRID RAID-Z layout information (reserved for future use) 211 * cksum checksum function 212 * comp compression function 213 * G gang block indicator 214 * B byteorder (endianness) 215 * D dedup 216 * X encryption (on version 30, which is not supported) 217 * E blkptr_t contains embedded data (see below) 218 * lvl level of indirection 219 * type DMU object type 220 * phys birth txg of block allocation; zero if same as logical birth txg 221 * log. birth transaction group in which the block was logically born 222 * fill count number of non-zero blocks under this bp 223 * checksum[4] 256-bit checksum of the data this bp describes 224 */ 225 226/* 227 * "Embedded" blkptr_t's don't actually point to a block, instead they 228 * have a data payload embedded in the blkptr_t itself. See the comment 229 * in blkptr.c for more details. 230 * 231 * The blkptr_t is laid out as follows: 232 * 233 * 64 56 48 40 32 24 16 8 0 234 * +-------+-------+-------+-------+-------+-------+-------+-------+ 235 * 0 | payload | 236 * 1 | payload | 237 * 2 | payload | 238 * 3 | payload | 239 * 4 | payload | 240 * 5 | payload | 241 * +-------+-------+-------+-------+-------+-------+-------+-------+ 242 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 243 * +-------+-------+-------+-------+-------+-------+-------+-------+ 244 * 7 | payload | 245 * 8 | payload | 246 * 9 | payload | 247 * +-------+-------+-------+-------+-------+-------+-------+-------+ 248 * a | logical birth txg | 249 * +-------+-------+-------+-------+-------+-------+-------+-------+ 250 * b | payload | 251 * c | payload | 252 * d | payload | 253 * e | payload | 254 * f | payload | 255 * +-------+-------+-------+-------+-------+-------+-------+-------+ 256 * 257 * Legend: 258 * 259 * payload contains the embedded data 260 * B (byteorder) byteorder (endianness) 261 * D (dedup) padding (set to zero) 262 * X encryption (set to zero; see above) 263 * E (embedded) set to one 264 * lvl indirection level 265 * type DMU object type 266 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 267 * comp compression function of payload 268 * PSIZE size of payload after compression, in bytes 269 * LSIZE logical size of payload, in bytes 270 * note that 25 bits is enough to store the largest 271 * "normal" BP's LSIZE (2^16 * 2^9) in bytes 272 * log. birth transaction group in which the block was logically born 273 * 274 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 275 * bp's they are stored in units of SPA_MINBLOCKSHIFT. 276 * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 277 * The B, D, X, lvl, type, and comp fields are stored the same as with normal 278 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 279 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 280 * other macros, as they assert that they are only used on BP's of the correct 281 * "embedded-ness". 282 */ 283 284#define BPE_GET_ETYPE(bp) \ 285 (ASSERT(BP_IS_EMBEDDED(bp)), \ 286 BF64_GET((bp)->blk_prop, 40, 8)) 287#define BPE_SET_ETYPE(bp, t) do { \ 288 ASSERT(BP_IS_EMBEDDED(bp)); \ 289 BF64_SET((bp)->blk_prop, 40, 8, t); \ 290_NOTE(CONSTCOND) } while (0) 291 292#define BPE_GET_LSIZE(bp) \ 293 (ASSERT(BP_IS_EMBEDDED(bp)), \ 294 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 295#define BPE_SET_LSIZE(bp, x) do { \ 296 ASSERT(BP_IS_EMBEDDED(bp)); \ 297 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 298_NOTE(CONSTCOND) } while (0) 299 300#define BPE_GET_PSIZE(bp) \ 301 (ASSERT(BP_IS_EMBEDDED(bp)), \ 302 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 303#define BPE_SET_PSIZE(bp, x) do { \ 304 ASSERT(BP_IS_EMBEDDED(bp)); \ 305 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 306_NOTE(CONSTCOND) } while (0) 307 308typedef enum bp_embedded_type { 309 BP_EMBEDDED_TYPE_DATA, 310 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ 311 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED 312} bp_embedded_type_t; 313 314#define BPE_NUM_WORDS 14 315#define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 316#define BPE_IS_PAYLOADWORD(bp, wp) \ 317 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 318 319#define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 320#define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 321 322/* 323 * A block is a hole when it has either 1) never been written to, or 324 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads 325 * without physically allocating disk space. Holes are represented in the 326 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is 327 * done through the BP_IS_HOLE macro. For holes, the logical size, level, 328 * DMU object type, and birth times are all also stored for holes that 329 * were written to at some point (i.e. were punched after having been filled). 330 */ 331typedef struct blkptr { 332 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 333 uint64_t blk_prop; /* size, compression, type, etc */ 334 uint64_t blk_pad[2]; /* Extra space for the future */ 335 uint64_t blk_phys_birth; /* txg when block was allocated */ 336 uint64_t blk_birth; /* transaction group at birth */ 337 uint64_t blk_fill; /* fill count */ 338 zio_cksum_t blk_cksum; /* 256-bit checksum */ 339} blkptr_t; 340 341/* 342 * Macros to get and set fields in a bp or DVA. 343 */ 344#define DVA_GET_ASIZE(dva) \ 345 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 346#define DVA_SET_ASIZE(dva, x) \ 347 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 348 SPA_MINBLOCKSHIFT, 0, x) 349 350#define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 351#define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 352 353#define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) 354#define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) 355 356#define DVA_GET_OFFSET(dva) \ 357 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 358#define DVA_SET_OFFSET(dva, x) \ 359 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 360 361#define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 362#define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 363 364#define BP_GET_LSIZE(bp) \ 365 (BP_IS_EMBEDDED(bp) ? \ 366 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 367 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 368#define BP_SET_LSIZE(bp, x) do { \ 369 ASSERT(!BP_IS_EMBEDDED(bp)); \ 370 BF64_SET_SB((bp)->blk_prop, \ 371 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 372_NOTE(CONSTCOND) } while (0) 373 374#define BP_GET_PSIZE(bp) \ 375 (BP_IS_EMBEDDED(bp) ? 0 : \ 376 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 377#define BP_SET_PSIZE(bp, x) do { \ 378 ASSERT(!BP_IS_EMBEDDED(bp)); \ 379 BF64_SET_SB((bp)->blk_prop, \ 380 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 381_NOTE(CONSTCOND) } while (0) 382 383#define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7) 384#define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x) 385 386#define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 387#define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x) 388 389#define BP_GET_CHECKSUM(bp) \ 390 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \ 391 BF64_GET((bp)->blk_prop, 40, 8)) 392#define BP_SET_CHECKSUM(bp, x) do { \ 393 ASSERT(!BP_IS_EMBEDDED(bp)); \ 394 BF64_SET((bp)->blk_prop, 40, 8, x); \ 395_NOTE(CONSTCOND) } while (0) 396 397#define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 398#define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 399 400#define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 401#define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 402 403#define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 404#define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 405 406#define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 407#define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 408 409#define BP_PHYSICAL_BIRTH(bp) \ 410 (BP_IS_EMBEDDED(bp) ? 0 : \ 411 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 412 413#define BP_SET_BIRTH(bp, logical, physical) \ 414{ \ 415 ASSERT(!BP_IS_EMBEDDED(bp)); \ 416 (bp)->blk_birth = (logical); \ 417 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \ 418} 419 420#define BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill) 421 422#define BP_GET_ASIZE(bp) \ 423 (BP_IS_EMBEDDED(bp) ? 0 : \ 424 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 425 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 426 DVA_GET_ASIZE(&(bp)->blk_dva[2])) 427 428#define BP_GET_UCSIZE(bp) \ 429 ((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \ 430 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)) 431 432#define BP_GET_NDVAS(bp) \ 433 (BP_IS_EMBEDDED(bp) ? 0 : \ 434 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 435 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 436 !!DVA_GET_ASIZE(&(bp)->blk_dva[2])) 437 438#define BP_COUNT_GANG(bp) \ 439 (BP_IS_EMBEDDED(bp) ? 0 : \ 440 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \ 441 DVA_GET_GANG(&(bp)->blk_dva[1]) + \ 442 DVA_GET_GANG(&(bp)->blk_dva[2]))) 443 444#define DVA_EQUAL(dva1, dva2) \ 445 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 446 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 447 448#define BP_EQUAL(bp1, bp2) \ 449 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \ 450 (bp1)->blk_birth == (bp2)->blk_birth && \ 451 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \ 452 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \ 453 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2])) 454 455#define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 456 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 457 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 458 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 459 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 460 461#define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 462 463#define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ 464{ \ 465 (zcp)->zc_word[0] = w0; \ 466 (zcp)->zc_word[1] = w1; \ 467 (zcp)->zc_word[2] = w2; \ 468 (zcp)->zc_word[3] = w3; \ 469} 470 471#define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0]) 472#define BP_IS_GANG(bp) \ 473 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp))) 474#define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 475 (dva)->dva_word[1] == 0ULL) 476#define BP_IS_HOLE(bp) \ 477 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp))) 478 479/* BP_IS_RAIDZ(bp) assumes no block compression */ 480#define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \ 481 BP_GET_PSIZE(bp)) 482 483#define BP_ZERO(bp) \ 484{ \ 485 (bp)->blk_dva[0].dva_word[0] = 0; \ 486 (bp)->blk_dva[0].dva_word[1] = 0; \ 487 (bp)->blk_dva[1].dva_word[0] = 0; \ 488 (bp)->blk_dva[1].dva_word[1] = 0; \ 489 (bp)->blk_dva[2].dva_word[0] = 0; \ 490 (bp)->blk_dva[2].dva_word[1] = 0; \ 491 (bp)->blk_prop = 0; \ 492 (bp)->blk_pad[0] = 0; \ 493 (bp)->blk_pad[1] = 0; \ 494 (bp)->blk_phys_birth = 0; \ 495 (bp)->blk_birth = 0; \ 496 (bp)->blk_fill = 0; \ 497 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 498} 499 500#if BYTE_ORDER == _BIG_ENDIAN 501#define ZFS_HOST_BYTEORDER (0ULL) 502#else 503#define ZFS_HOST_BYTEORDER (1ULL) 504#endif 505 506#define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) 507 508#define BP_SPRINTF_LEN 320 509 510/* 511 * This macro allows code sharing between zfs, libzpool, and mdb. 512 * 'func' is either snprintf() or mdb_snprintf(). 513 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line. 514 */ 515#define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \ 516{ \ 517 static const char *copyname[] = \ 518 { "zero", "single", "double", "triple" }; \ 519 int len = 0; \ 520 int copies = 0; \ 521 \ 522 if (bp == NULL) { \ 523 len += func(buf + len, size - len, "<NULL>"); \ 524 } else if (BP_IS_HOLE(bp)) { \ 525 len += func(buf + len, size - len, "<hole>"); \ 526 if (bp->blk_birth > 0) { \ 527 len += func(buf + len, size - len, \ 528 " birth=%lluL", \ 529 (u_longlong_t)bp->blk_birth); \ 530 } \ 531 } else if (BP_IS_EMBEDDED(bp)) { \ 532 len = func(buf + len, size - len, \ 533 "EMBEDDED [L%llu %s] et=%u %s " \ 534 "size=%llxL/%llxP birth=%lluL", \ 535 (u_longlong_t)BP_GET_LEVEL(bp), \ 536 type, \ 537 (int)BPE_GET_ETYPE(bp), \ 538 compress, \ 539 (u_longlong_t)BPE_GET_LSIZE(bp), \ 540 (u_longlong_t)BPE_GET_PSIZE(bp), \ 541 (u_longlong_t)bp->blk_birth); \ 542 } else { \ 543 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \ 544 const dva_t *dva = &bp->blk_dva[d]; \ 545 if (DVA_IS_VALID(dva)) \ 546 copies++; \ 547 len += func(buf + len, size - len, \ 548 "DVA[%d]=<%llu:%llx:%llx>%c", d, \ 549 (u_longlong_t)DVA_GET_VDEV(dva), \ 550 (u_longlong_t)DVA_GET_OFFSET(dva), \ 551 (u_longlong_t)DVA_GET_ASIZE(dva), \ 552 ws); \ 553 } \ 554 if (BP_IS_GANG(bp) && \ 555 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \ 556 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \ 557 copies--; \ 558 len += func(buf + len, size - len, \ 559 "[L%llu %s] %s %s %s %s %s %s%c" \ 560 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \ 561 "cksum=%llx:%llx:%llx:%llx", \ 562 (u_longlong_t)BP_GET_LEVEL(bp), \ 563 type, \ 564 checksum, \ 565 compress, \ 566 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \ 567 BP_IS_GANG(bp) ? "gang" : "contiguous", \ 568 BP_GET_DEDUP(bp) ? "dedup" : "unique", \ 569 copyname[copies], \ 570 ws, \ 571 (u_longlong_t)BP_GET_LSIZE(bp), \ 572 (u_longlong_t)BP_GET_PSIZE(bp), \ 573 (u_longlong_t)bp->blk_birth, \ 574 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \ 575 (u_longlong_t)BP_GET_FILL(bp), \ 576 ws, \ 577 (u_longlong_t)bp->blk_cksum.zc_word[0], \ 578 (u_longlong_t)bp->blk_cksum.zc_word[1], \ 579 (u_longlong_t)bp->blk_cksum.zc_word[2], \ 580 (u_longlong_t)bp->blk_cksum.zc_word[3]); \ 581 } \ 582 ASSERT(len < size); \ 583} 584 585#include <sys/dmu.h> 586 587#define BP_GET_BUFC_TYPE(bp) \ 588 (((BP_GET_LEVEL(bp) > 0) || (DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))) ? \ 589 ARC_BUFC_METADATA : ARC_BUFC_DATA) 590 591typedef enum spa_import_type { 592 SPA_IMPORT_EXISTING, 593 SPA_IMPORT_ASSEMBLE 594} spa_import_type_t; 595 596/* state manipulation functions */ 597extern int spa_open(const char *pool, spa_t **, void *tag); 598extern int spa_open_rewind(const char *pool, spa_t **, void *tag, 599 nvlist_t *policy, nvlist_t **config); 600extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot, 601 size_t buflen); 602extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props, 603 nvlist_t *zplprops); 604#if defined(sun) 605extern int spa_import_rootpool(char *devpath, char *devid); 606#else 607extern int spa_import_rootpool(const char *name); 608#endif 609extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props, 610 uint64_t flags); 611extern nvlist_t *spa_tryimport(nvlist_t *tryconfig); 612extern int spa_destroy(char *pool); 613extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force, 614 boolean_t hardforce); 615extern int spa_reset(char *pool); 616extern void spa_async_request(spa_t *spa, int flag); 617extern void spa_async_unrequest(spa_t *spa, int flag); 618extern void spa_async_suspend(spa_t *spa); 619extern void spa_async_resume(spa_t *spa); 620extern spa_t *spa_inject_addref(char *pool); 621extern void spa_inject_delref(spa_t *spa); 622extern void spa_scan_stat_init(spa_t *spa); 623extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps); 624 625#define SPA_ASYNC_CONFIG_UPDATE 0x01 626#define SPA_ASYNC_REMOVE 0x02 627#define SPA_ASYNC_PROBE 0x04 628#define SPA_ASYNC_RESILVER_DONE 0x08 629#define SPA_ASYNC_RESILVER 0x10 630#define SPA_ASYNC_AUTOEXPAND 0x20 631#define SPA_ASYNC_REMOVE_DONE 0x40 632#define SPA_ASYNC_REMOVE_STOP 0x80 633 634/* 635 * Controls the behavior of spa_vdev_remove(). 636 */ 637#define SPA_REMOVE_UNSPARE 0x01 638#define SPA_REMOVE_DONE 0x02 639 640/* device manipulation */ 641extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot); 642extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, 643 int replacing); 644extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, 645 int replace_done); 646extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare); 647extern boolean_t spa_vdev_remove_active(spa_t *spa); 648extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath); 649extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru); 650extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config, 651 nvlist_t *props, boolean_t exp); 652 653/* spare state (which is global across all pools) */ 654extern void spa_spare_add(vdev_t *vd); 655extern void spa_spare_remove(vdev_t *vd); 656extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt); 657extern void spa_spare_activate(vdev_t *vd); 658 659/* L2ARC state (which is global across all pools) */ 660extern void spa_l2cache_add(vdev_t *vd); 661extern void spa_l2cache_remove(vdev_t *vd); 662extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool); 663extern void spa_l2cache_activate(vdev_t *vd); 664extern void spa_l2cache_drop(spa_t *spa); 665 666/* scanning */ 667extern int spa_scan(spa_t *spa, pool_scan_func_t func); 668extern int spa_scan_stop(spa_t *spa); 669 670/* spa syncing */ 671extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */ 672extern void spa_sync_allpools(void); 673 674/* spa namespace global mutex */ 675extern kmutex_t spa_namespace_lock; 676 677/* 678 * SPA configuration functions in spa_config.c 679 */ 680 681#define SPA_CONFIG_UPDATE_POOL 0 682#define SPA_CONFIG_UPDATE_VDEVS 1 683 684extern void spa_config_sync(spa_t *, boolean_t, boolean_t); 685extern void spa_config_load(void); 686extern nvlist_t *spa_all_configs(uint64_t *); 687extern void spa_config_set(spa_t *spa, nvlist_t *config); 688extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, 689 int getstats); 690extern void spa_config_update(spa_t *spa, int what); 691 692/* 693 * Miscellaneous SPA routines in spa_misc.c 694 */ 695 696/* Namespace manipulation */ 697extern spa_t *spa_lookup(const char *name); 698extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot); 699extern void spa_remove(spa_t *spa); 700extern spa_t *spa_next(spa_t *prev); 701 702/* Refcount functions */ 703extern void spa_open_ref(spa_t *spa, void *tag); 704extern void spa_close(spa_t *spa, void *tag); 705extern boolean_t spa_refcount_zero(spa_t *spa); 706 707#define SCL_NONE 0x00 708#define SCL_CONFIG 0x01 709#define SCL_STATE 0x02 710#define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */ 711#define SCL_ALLOC 0x08 712#define SCL_ZIO 0x10 713#define SCL_FREE 0x20 714#define SCL_VDEV 0x40 715#define SCL_LOCKS 7 716#define SCL_ALL ((1 << SCL_LOCKS) - 1) 717#define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO) 718 719/* Pool configuration locks */ 720extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw); 721extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw); 722extern void spa_config_exit(spa_t *spa, int locks, void *tag); 723extern int spa_config_held(spa_t *spa, int locks, krw_t rw); 724 725/* Pool vdev add/remove lock */ 726extern uint64_t spa_vdev_enter(spa_t *spa); 727extern uint64_t spa_vdev_config_enter(spa_t *spa); 728extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, 729 int error, char *tag); 730extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error); 731 732/* Pool vdev state change lock */ 733extern void spa_vdev_state_enter(spa_t *spa, int oplock); 734extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error); 735 736/* Log state */ 737typedef enum spa_log_state { 738 SPA_LOG_UNKNOWN = 0, /* unknown log state */ 739 SPA_LOG_MISSING, /* missing log(s) */ 740 SPA_LOG_CLEAR, /* clear the log(s) */ 741 SPA_LOG_GOOD, /* log(s) are good */ 742} spa_log_state_t; 743 744extern spa_log_state_t spa_get_log_state(spa_t *spa); 745extern void spa_set_log_state(spa_t *spa, spa_log_state_t state); 746extern int spa_offline_log(spa_t *spa); 747 748/* Log claim callback */ 749extern void spa_claim_notify(zio_t *zio); 750 751/* Accessor functions */ 752extern boolean_t spa_shutting_down(spa_t *spa); 753extern struct dsl_pool *spa_get_dsl(spa_t *spa); 754extern boolean_t spa_is_initializing(spa_t *spa); 755extern blkptr_t *spa_get_rootblkptr(spa_t *spa); 756extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp); 757extern void spa_altroot(spa_t *, char *, size_t); 758extern int spa_sync_pass(spa_t *spa); 759extern char *spa_name(spa_t *spa); 760extern uint64_t spa_guid(spa_t *spa); 761extern uint64_t spa_load_guid(spa_t *spa); 762extern uint64_t spa_last_synced_txg(spa_t *spa); 763extern uint64_t spa_first_txg(spa_t *spa); 764extern uint64_t spa_syncing_txg(spa_t *spa); 765extern uint64_t spa_version(spa_t *spa); 766extern pool_state_t spa_state(spa_t *spa); 767extern spa_load_state_t spa_load_state(spa_t *spa); 768extern uint64_t spa_freeze_txg(spa_t *spa); 769extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize); 770extern uint64_t spa_get_dspace(spa_t *spa); 771extern uint64_t spa_get_slop_space(spa_t *spa); 772extern void spa_update_dspace(spa_t *spa); 773extern uint64_t spa_version(spa_t *spa); 774extern boolean_t spa_deflate(spa_t *spa); 775extern metaslab_class_t *spa_normal_class(spa_t *spa); 776extern metaslab_class_t *spa_log_class(spa_t *spa); 777extern int spa_max_replication(spa_t *spa); 778extern int spa_prev_software_version(spa_t *spa); 779extern int spa_busy(void); 780extern uint8_t spa_get_failmode(spa_t *spa); 781extern boolean_t spa_suspended(spa_t *spa); 782extern uint64_t spa_bootfs(spa_t *spa); 783extern uint64_t spa_delegation(spa_t *spa); 784extern objset_t *spa_meta_objset(spa_t *spa); 785extern uint64_t spa_deadman_synctime(spa_t *spa); 786 787/* Miscellaneous support routines */ 788extern void spa_activate_mos_feature(spa_t *spa, const char *feature, 789 dmu_tx_t *tx); 790extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature); 791extern int spa_rename(const char *oldname, const char *newname); 792extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid); 793extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid); 794extern char *spa_strdup(const char *); 795extern void spa_strfree(char *); 796extern uint64_t spa_get_random(uint64_t range); 797extern uint64_t spa_generate_guid(spa_t *spa); 798extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp); 799extern void spa_freeze(spa_t *spa); 800extern int spa_change_guid(spa_t *spa); 801extern void spa_upgrade(spa_t *spa, uint64_t version); 802extern void spa_evict_all(void); 803extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid, 804 boolean_t l2cache); 805extern boolean_t spa_has_spare(spa_t *, uint64_t guid); 806extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva); 807extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp); 808extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp); 809extern boolean_t spa_has_slogs(spa_t *spa); 810extern boolean_t spa_is_root(spa_t *spa); 811extern boolean_t spa_writeable(spa_t *spa); 812extern boolean_t spa_has_pending_synctask(spa_t *spa); 813extern int spa_maxblocksize(spa_t *spa); 814 815extern int spa_mode(spa_t *spa); 816extern uint64_t zfs_strtonum(const char *str, char **nptr); 817#define strtonum(str, nptr) zfs_strtonum((str), (nptr)) 818 819extern char *spa_his_ievent_table[]; 820 821extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx); 822extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read, 823 char *his_buf); 824extern int spa_history_log(spa_t *spa, const char *his_buf); 825extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl); 826extern void spa_history_log_version(spa_t *spa, const char *operation); 827extern void spa_history_log_internal(spa_t *spa, const char *operation, 828 dmu_tx_t *tx, const char *fmt, ...); 829extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op, 830 dmu_tx_t *tx, const char *fmt, ...); 831extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation, 832 dmu_tx_t *tx, const char *fmt, ...); 833 834/* error handling */ 835struct zbookmark_phys; 836extern void spa_log_error(spa_t *spa, zio_t *zio); 837extern void zfs_ereport_post(const char *cls, spa_t *spa, vdev_t *vd, 838 zio_t *zio, uint64_t stateoroffset, uint64_t length); 839extern void zfs_post_remove(spa_t *spa, vdev_t *vd); 840extern void zfs_post_state_change(spa_t *spa, vdev_t *vd); 841extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd); 842extern uint64_t spa_get_errlog_size(spa_t *spa); 843extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count); 844extern void spa_errlog_rotate(spa_t *spa); 845extern void spa_errlog_drain(spa_t *spa); 846extern void spa_errlog_sync(spa_t *spa, uint64_t txg); 847extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub); 848 849/* vdev cache */ 850extern void vdev_cache_stat_init(void); 851extern void vdev_cache_stat_fini(void); 852 853/* Initialization and termination */ 854extern void spa_init(int flags); 855extern void spa_fini(void); 856extern void spa_boot_init(); 857 858/* properties */ 859extern int spa_prop_set(spa_t *spa, nvlist_t *nvp); 860extern int spa_prop_get(spa_t *spa, nvlist_t **nvp); 861extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx); 862extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t); 863 864/* asynchronous event notification */ 865extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name); 866 867#ifdef ZFS_DEBUG 868#define dprintf_bp(bp, fmt, ...) do { \ 869 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 870 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \ 871 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \ 872 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \ 873 kmem_free(__blkbuf, BP_SPRINTF_LEN); \ 874 } \ 875_NOTE(CONSTCOND) } while (0) 876#else 877#define dprintf_bp(bp, fmt, ...) 878#endif 879 880extern boolean_t spa_debug_enabled(spa_t *spa); 881#define spa_dbgmsg(spa, ...) \ 882{ \ 883 if (spa_debug_enabled(spa)) \ 884 zfs_dbgmsg(__VA_ARGS__); \ 885} 886 887extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */ 888 889#ifdef __cplusplus 890} 891#endif 892 893#endif /* _SYS_SPA_H */ 894