spa.h revision 268657
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 nine block sizes, from 512 bytes to 128K. 98 * We could go higher, but the benefits are near-zero and the cost 99 * of COWing a giant block to modify one byte would become excessive. 100 */ 101#define SPA_MINBLOCKSHIFT 9 102#define SPA_MAXBLOCKSHIFT 17 103#define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 104#define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 105 106#define SPA_BLOCKSIZES (SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1) 107 108/* 109 * Default maximum supported logical ashift. 110 * 111 * The current 8k allocation block size limit is due to the 8k 112 * aligned/sized operations performed by vdev_probe() on 113 * vdev_label->vl_pad2. Using another "safe region" for these tests 114 * would allow the limit to be raised to 16k, at the expense of 115 * only having 8 available uberblocks in the label area. 116 */ 117#define SPA_MAXASHIFT 13 118 119/* 120 * Default minimum supported logical ashift. 121 */ 122#define SPA_MINASHIFT SPA_MINBLOCKSHIFT 123 124/* 125 * Size of block to hold the configuration data (a packed nvlist) 126 */ 127#define SPA_CONFIG_BLOCKSIZE (1ULL << 14) 128 129/* 130 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. 131 * The ASIZE encoding should be at least 64 times larger (6 more bits) 132 * to support up to 4-way RAID-Z mirror mode with worst-case gang block 133 * overhead, three DVAs per bp, plus one more bit in case we do anything 134 * else that expands the ASIZE. 135 */ 136#define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ 137#define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ 138#define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ 139 140/* 141 * All SPA data is represented by 128-bit data virtual addresses (DVAs). 142 * The members of the dva_t should be considered opaque outside the SPA. 143 */ 144typedef struct dva { 145 uint64_t dva_word[2]; 146} dva_t; 147 148/* 149 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. 150 */ 151typedef struct zio_cksum { 152 uint64_t zc_word[4]; 153} zio_cksum_t; 154 155/* 156 * Each block is described by its DVAs, time of birth, checksum, etc. 157 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 158 * 159 * 64 56 48 40 32 24 16 8 0 160 * +-------+-------+-------+-------+-------+-------+-------+-------+ 161 * 0 | vdev1 | GRID | ASIZE | 162 * +-------+-------+-------+-------+-------+-------+-------+-------+ 163 * 1 |G| offset1 | 164 * +-------+-------+-------+-------+-------+-------+-------+-------+ 165 * 2 | vdev2 | GRID | ASIZE | 166 * +-------+-------+-------+-------+-------+-------+-------+-------+ 167 * 3 |G| offset2 | 168 * +-------+-------+-------+-------+-------+-------+-------+-------+ 169 * 4 | vdev3 | GRID | ASIZE | 170 * +-------+-------+-------+-------+-------+-------+-------+-------+ 171 * 5 |G| offset3 | 172 * +-------+-------+-------+-------+-------+-------+-------+-------+ 173 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 174 * +-------+-------+-------+-------+-------+-------+-------+-------+ 175 * 7 | padding | 176 * +-------+-------+-------+-------+-------+-------+-------+-------+ 177 * 8 | padding | 178 * +-------+-------+-------+-------+-------+-------+-------+-------+ 179 * 9 | physical birth txg | 180 * +-------+-------+-------+-------+-------+-------+-------+-------+ 181 * a | logical birth txg | 182 * +-------+-------+-------+-------+-------+-------+-------+-------+ 183 * b | fill count | 184 * +-------+-------+-------+-------+-------+-------+-------+-------+ 185 * c | checksum[0] | 186 * +-------+-------+-------+-------+-------+-------+-------+-------+ 187 * d | checksum[1] | 188 * +-------+-------+-------+-------+-------+-------+-------+-------+ 189 * e | checksum[2] | 190 * +-------+-------+-------+-------+-------+-------+-------+-------+ 191 * f | checksum[3] | 192 * +-------+-------+-------+-------+-------+-------+-------+-------+ 193 * 194 * Legend: 195 * 196 * vdev virtual device ID 197 * offset offset into virtual device 198 * LSIZE logical size 199 * PSIZE physical size (after compression) 200 * ASIZE allocated size (including RAID-Z parity and gang block headers) 201 * GRID RAID-Z layout information (reserved for future use) 202 * cksum checksum function 203 * comp compression function 204 * G gang block indicator 205 * B byteorder (endianness) 206 * D dedup 207 * X encryption (on version 30, which is not supported) 208 * E blkptr_t contains embedded data (see below) 209 * lvl level of indirection 210 * type DMU object type 211 * phys birth txg of block allocation; zero if same as logical birth txg 212 * log. birth transaction group in which the block was logically born 213 * fill count number of non-zero blocks under this bp 214 * checksum[4] 256-bit checksum of the data this bp describes 215 */ 216 217/* 218 * "Embedded" blkptr_t's don't actually point to a block, instead they 219 * have a data payload embedded in the blkptr_t itself. See the comment 220 * in blkptr.c for more details. 221 * 222 * The blkptr_t is laid out as follows: 223 * 224 * 64 56 48 40 32 24 16 8 0 225 * +-------+-------+-------+-------+-------+-------+-------+-------+ 226 * 0 | payload | 227 * 1 | payload | 228 * 2 | payload | 229 * 3 | payload | 230 * 4 | payload | 231 * 5 | payload | 232 * +-------+-------+-------+-------+-------+-------+-------+-------+ 233 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 234 * +-------+-------+-------+-------+-------+-------+-------+-------+ 235 * 7 | payload | 236 * 8 | payload | 237 * 9 | payload | 238 * +-------+-------+-------+-------+-------+-------+-------+-------+ 239 * a | logical birth txg | 240 * +-------+-------+-------+-------+-------+-------+-------+-------+ 241 * b | payload | 242 * c | payload | 243 * d | payload | 244 * e | payload | 245 * f | payload | 246 * +-------+-------+-------+-------+-------+-------+-------+-------+ 247 * 248 * Legend: 249 * 250 * payload contains the embedded data 251 * B (byteorder) byteorder (endianness) 252 * D (dedup) padding (set to zero) 253 * X encryption (set to zero; see above) 254 * E (embedded) set to one 255 * lvl indirection level 256 * type DMU object type 257 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 258 * comp compression function of payload 259 * PSIZE size of payload after compression, in bytes 260 * LSIZE logical size of payload, in bytes 261 * note that 25 bits is enough to store the largest 262 * "normal" BP's LSIZE (2^16 * 2^9) in bytes 263 * log. birth transaction group in which the block was logically born 264 * 265 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 266 * bp's they are stored in units of SPA_MINBLOCKSHIFT. 267 * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 268 * The B, D, X, lvl, type, and comp fields are stored the same as with normal 269 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 270 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 271 * other macros, as they assert that they are only used on BP's of the correct 272 * "embedded-ness". 273 */ 274 275#define BPE_GET_ETYPE(bp) \ 276 (ASSERT(BP_IS_EMBEDDED(bp)), \ 277 BF64_GET((bp)->blk_prop, 40, 8)) 278#define BPE_SET_ETYPE(bp, t) do { \ 279 ASSERT(BP_IS_EMBEDDED(bp)); \ 280 BF64_SET((bp)->blk_prop, 40, 8, t); \ 281_NOTE(CONSTCOND) } while (0) 282 283#define BPE_GET_LSIZE(bp) \ 284 (ASSERT(BP_IS_EMBEDDED(bp)), \ 285 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 286#define BPE_SET_LSIZE(bp, x) do { \ 287 ASSERT(BP_IS_EMBEDDED(bp)); \ 288 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 289_NOTE(CONSTCOND) } while (0) 290 291#define BPE_GET_PSIZE(bp) \ 292 (ASSERT(BP_IS_EMBEDDED(bp)), \ 293 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 294#define BPE_SET_PSIZE(bp, x) do { \ 295 ASSERT(BP_IS_EMBEDDED(bp)); \ 296 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 297_NOTE(CONSTCOND) } while (0) 298 299typedef enum bp_embedded_type { 300 BP_EMBEDDED_TYPE_DATA, 301 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ 302 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED 303} bp_embedded_type_t; 304 305#define BPE_NUM_WORDS 14 306#define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 307#define BPE_IS_PAYLOADWORD(bp, wp) \ 308 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 309 310#define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 311#define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 312 313/* 314 * A block is a hole when it has either 1) never been written to, or 315 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads 316 * without physically allocating disk space. Holes are represented in the 317 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is 318 * done through the BP_IS_HOLE macro. For holes, the logical size, level, 319 * DMU object type, and birth times are all also stored for holes that 320 * were written to at some point (i.e. were punched after having been filled). 321 */ 322typedef struct blkptr { 323 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 324 uint64_t blk_prop; /* size, compression, type, etc */ 325 uint64_t blk_pad[2]; /* Extra space for the future */ 326 uint64_t blk_phys_birth; /* txg when block was allocated */ 327 uint64_t blk_birth; /* transaction group at birth */ 328 uint64_t blk_fill; /* fill count */ 329 zio_cksum_t blk_cksum; /* 256-bit checksum */ 330} blkptr_t; 331 332/* 333 * Macros to get and set fields in a bp or DVA. 334 */ 335#define DVA_GET_ASIZE(dva) \ 336 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 337#define DVA_SET_ASIZE(dva, x) \ 338 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 339 SPA_MINBLOCKSHIFT, 0, x) 340 341#define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 342#define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 343 344#define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) 345#define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) 346 347#define DVA_GET_OFFSET(dva) \ 348 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 349#define DVA_SET_OFFSET(dva, x) \ 350 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 351 352#define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 353#define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 354 355#define BP_GET_LSIZE(bp) \ 356 (BP_IS_EMBEDDED(bp) ? \ 357 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 358 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 359#define BP_SET_LSIZE(bp, x) do { \ 360 ASSERT(!BP_IS_EMBEDDED(bp)); \ 361 BF64_SET_SB((bp)->blk_prop, \ 362 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 363_NOTE(CONSTCOND) } while (0) 364 365#define BP_GET_PSIZE(bp) \ 366 (BP_IS_EMBEDDED(bp) ? 0 : \ 367 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 368#define BP_SET_PSIZE(bp, x) do { \ 369 ASSERT(!BP_IS_EMBEDDED(bp)); \ 370 BF64_SET_SB((bp)->blk_prop, \ 371 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 372_NOTE(CONSTCOND) } while (0) 373 374#define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7) 375#define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x) 376 377#define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 378#define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x) 379 380#define BP_GET_CHECKSUM(bp) \ 381 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \ 382 BF64_GET((bp)->blk_prop, 40, 8)) 383#define BP_SET_CHECKSUM(bp, x) do { \ 384 ASSERT(!BP_IS_EMBEDDED(bp)); \ 385 BF64_SET((bp)->blk_prop, 40, 8, x); \ 386_NOTE(CONSTCOND) } while (0) 387 388#define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 389#define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 390 391#define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 392#define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 393 394#define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 395#define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 396 397#define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 398#define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 399 400#define BP_PHYSICAL_BIRTH(bp) \ 401 (BP_IS_EMBEDDED(bp) ? 0 : \ 402 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 403 404#define BP_SET_BIRTH(bp, logical, physical) \ 405{ \ 406 ASSERT(!BP_IS_EMBEDDED(bp)); \ 407 (bp)->blk_birth = (logical); \ 408 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \ 409} 410 411#define BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill) 412 413#define BP_GET_ASIZE(bp) \ 414 (BP_IS_EMBEDDED(bp) ? 0 : \ 415 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 416 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 417 DVA_GET_ASIZE(&(bp)->blk_dva[2])) 418 419#define BP_GET_UCSIZE(bp) \ 420 ((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \ 421 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)) 422 423#define BP_GET_NDVAS(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_COUNT_GANG(bp) \ 430 (BP_IS_EMBEDDED(bp) ? 0 : \ 431 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \ 432 DVA_GET_GANG(&(bp)->blk_dva[1]) + \ 433 DVA_GET_GANG(&(bp)->blk_dva[2]))) 434 435#define DVA_EQUAL(dva1, dva2) \ 436 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 437 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 438 439#define BP_EQUAL(bp1, bp2) \ 440 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \ 441 (bp1)->blk_birth == (bp2)->blk_birth && \ 442 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \ 443 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \ 444 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2])) 445 446#define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 447 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 448 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 449 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 450 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 451 452#define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 453 454#define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ 455{ \ 456 (zcp)->zc_word[0] = w0; \ 457 (zcp)->zc_word[1] = w1; \ 458 (zcp)->zc_word[2] = w2; \ 459 (zcp)->zc_word[3] = w3; \ 460} 461 462#define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0]) 463#define BP_IS_GANG(bp) \ 464 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp))) 465#define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 466 (dva)->dva_word[1] == 0ULL) 467#define BP_IS_HOLE(bp) \ 468 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp))) 469 470/* BP_IS_RAIDZ(bp) assumes no block compression */ 471#define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \ 472 BP_GET_PSIZE(bp)) 473 474#define BP_ZERO(bp) \ 475{ \ 476 (bp)->blk_dva[0].dva_word[0] = 0; \ 477 (bp)->blk_dva[0].dva_word[1] = 0; \ 478 (bp)->blk_dva[1].dva_word[0] = 0; \ 479 (bp)->blk_dva[1].dva_word[1] = 0; \ 480 (bp)->blk_dva[2].dva_word[0] = 0; \ 481 (bp)->blk_dva[2].dva_word[1] = 0; \ 482 (bp)->blk_prop = 0; \ 483 (bp)->blk_pad[0] = 0; \ 484 (bp)->blk_pad[1] = 0; \ 485 (bp)->blk_phys_birth = 0; \ 486 (bp)->blk_birth = 0; \ 487 (bp)->blk_fill = 0; \ 488 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 489} 490 491#if BYTE_ORDER == _BIG_ENDIAN 492#define ZFS_HOST_BYTEORDER (0ULL) 493#else 494#define ZFS_HOST_BYTEORDER (1ULL) 495#endif 496 497#define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) 498 499#define BP_SPRINTF_LEN 320 500 501/* 502 * This macro allows code sharing between zfs, libzpool, and mdb. 503 * 'func' is either snprintf() or mdb_snprintf(). 504 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line. 505 */ 506#define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \ 507{ \ 508 static const char *copyname[] = \ 509 { "zero", "single", "double", "triple" }; \ 510 int len = 0; \ 511 int copies = 0; \ 512 \ 513 if (bp == NULL) { \ 514 len += func(buf + len, size - len, "<NULL>"); \ 515 } else if (BP_IS_HOLE(bp)) { \ 516 len += func(buf + len, size - len, "<hole>"); \ 517 if (bp->blk_birth > 0) { \ 518 len += func(buf + len, size - len, \ 519 " birth=%lluL", \ 520 (u_longlong_t)bp->blk_birth); \ 521 } \ 522 } else if (BP_IS_EMBEDDED(bp)) { \ 523 len = func(buf + len, size - len, \ 524 "EMBEDDED [L%llu %s] et=%u %s " \ 525 "size=%llxL/%llxP birth=%lluL", \ 526 (u_longlong_t)BP_GET_LEVEL(bp), \ 527 type, \ 528 (int)BPE_GET_ETYPE(bp), \ 529 compress, \ 530 (u_longlong_t)BPE_GET_LSIZE(bp), \ 531 (u_longlong_t)BPE_GET_PSIZE(bp), \ 532 (u_longlong_t)bp->blk_birth); \ 533 } else { \ 534 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \ 535 const dva_t *dva = &bp->blk_dva[d]; \ 536 if (DVA_IS_VALID(dva)) \ 537 copies++; \ 538 len += func(buf + len, size - len, \ 539 "DVA[%d]=<%llu:%llx:%llx>%c", d, \ 540 (u_longlong_t)DVA_GET_VDEV(dva), \ 541 (u_longlong_t)DVA_GET_OFFSET(dva), \ 542 (u_longlong_t)DVA_GET_ASIZE(dva), \ 543 ws); \ 544 } \ 545 if (BP_IS_GANG(bp) && \ 546 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \ 547 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \ 548 copies--; \ 549 len += func(buf + len, size - len, \ 550 "[L%llu %s] %s %s %s %s %s %s%c" \ 551 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \ 552 "cksum=%llx:%llx:%llx:%llx", \ 553 (u_longlong_t)BP_GET_LEVEL(bp), \ 554 type, \ 555 checksum, \ 556 compress, \ 557 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \ 558 BP_IS_GANG(bp) ? "gang" : "contiguous", \ 559 BP_GET_DEDUP(bp) ? "dedup" : "unique", \ 560 copyname[copies], \ 561 ws, \ 562 (u_longlong_t)BP_GET_LSIZE(bp), \ 563 (u_longlong_t)BP_GET_PSIZE(bp), \ 564 (u_longlong_t)bp->blk_birth, \ 565 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \ 566 (u_longlong_t)BP_GET_FILL(bp), \ 567 ws, \ 568 (u_longlong_t)bp->blk_cksum.zc_word[0], \ 569 (u_longlong_t)bp->blk_cksum.zc_word[1], \ 570 (u_longlong_t)bp->blk_cksum.zc_word[2], \ 571 (u_longlong_t)bp->blk_cksum.zc_word[3]); \ 572 } \ 573 ASSERT(len < size); \ 574} 575 576#include <sys/dmu.h> 577 578#define BP_GET_BUFC_TYPE(bp) \ 579 (((BP_GET_LEVEL(bp) > 0) || (DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))) ? \ 580 ARC_BUFC_METADATA : ARC_BUFC_DATA) 581 582typedef enum spa_import_type { 583 SPA_IMPORT_EXISTING, 584 SPA_IMPORT_ASSEMBLE 585} spa_import_type_t; 586 587/* state manipulation functions */ 588extern int spa_open(const char *pool, spa_t **, void *tag); 589extern int spa_open_rewind(const char *pool, spa_t **, void *tag, 590 nvlist_t *policy, nvlist_t **config); 591extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot, 592 size_t buflen); 593extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props, 594 nvlist_t *zplprops); 595#if defined(sun) 596extern int spa_import_rootpool(char *devpath, char *devid); 597#else 598extern int spa_import_rootpool(const char *name); 599#endif 600extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props, 601 uint64_t flags); 602extern nvlist_t *spa_tryimport(nvlist_t *tryconfig); 603extern int spa_destroy(char *pool); 604extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force, 605 boolean_t hardforce); 606extern int spa_reset(char *pool); 607extern void spa_async_request(spa_t *spa, int flag); 608extern void spa_async_unrequest(spa_t *spa, int flag); 609extern void spa_async_suspend(spa_t *spa); 610extern void spa_async_resume(spa_t *spa); 611extern spa_t *spa_inject_addref(char *pool); 612extern void spa_inject_delref(spa_t *spa); 613extern void spa_scan_stat_init(spa_t *spa); 614extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps); 615 616#define SPA_ASYNC_CONFIG_UPDATE 0x01 617#define SPA_ASYNC_REMOVE 0x02 618#define SPA_ASYNC_PROBE 0x04 619#define SPA_ASYNC_RESILVER_DONE 0x08 620#define SPA_ASYNC_RESILVER 0x10 621#define SPA_ASYNC_AUTOEXPAND 0x20 622#define SPA_ASYNC_REMOVE_DONE 0x40 623#define SPA_ASYNC_REMOVE_STOP 0x80 624 625/* 626 * Controls the behavior of spa_vdev_remove(). 627 */ 628#define SPA_REMOVE_UNSPARE 0x01 629#define SPA_REMOVE_DONE 0x02 630 631/* device manipulation */ 632extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot); 633extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, 634 int replacing); 635extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, 636 int replace_done); 637extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare); 638extern boolean_t spa_vdev_remove_active(spa_t *spa); 639extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath); 640extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru); 641extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config, 642 nvlist_t *props, boolean_t exp); 643 644/* spare state (which is global across all pools) */ 645extern void spa_spare_add(vdev_t *vd); 646extern void spa_spare_remove(vdev_t *vd); 647extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt); 648extern void spa_spare_activate(vdev_t *vd); 649 650/* L2ARC state (which is global across all pools) */ 651extern void spa_l2cache_add(vdev_t *vd); 652extern void spa_l2cache_remove(vdev_t *vd); 653extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool); 654extern void spa_l2cache_activate(vdev_t *vd); 655extern void spa_l2cache_drop(spa_t *spa); 656 657/* scanning */ 658extern int spa_scan(spa_t *spa, pool_scan_func_t func); 659extern int spa_scan_stop(spa_t *spa); 660 661/* spa syncing */ 662extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */ 663extern void spa_sync_allpools(void); 664 665/* spa namespace global mutex */ 666extern kmutex_t spa_namespace_lock; 667 668/* 669 * SPA configuration functions in spa_config.c 670 */ 671 672#define SPA_CONFIG_UPDATE_POOL 0 673#define SPA_CONFIG_UPDATE_VDEVS 1 674 675extern void spa_config_sync(spa_t *, boolean_t, boolean_t); 676extern void spa_config_load(void); 677extern nvlist_t *spa_all_configs(uint64_t *); 678extern void spa_config_set(spa_t *spa, nvlist_t *config); 679extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, 680 int getstats); 681extern void spa_config_update(spa_t *spa, int what); 682 683/* 684 * Miscellaneous SPA routines in spa_misc.c 685 */ 686 687/* Namespace manipulation */ 688extern spa_t *spa_lookup(const char *name); 689extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot); 690extern void spa_remove(spa_t *spa); 691extern spa_t *spa_next(spa_t *prev); 692 693/* Refcount functions */ 694extern void spa_open_ref(spa_t *spa, void *tag); 695extern void spa_close(spa_t *spa, void *tag); 696extern boolean_t spa_refcount_zero(spa_t *spa); 697 698#define SCL_NONE 0x00 699#define SCL_CONFIG 0x01 700#define SCL_STATE 0x02 701#define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */ 702#define SCL_ALLOC 0x08 703#define SCL_ZIO 0x10 704#define SCL_FREE 0x20 705#define SCL_VDEV 0x40 706#define SCL_LOCKS 7 707#define SCL_ALL ((1 << SCL_LOCKS) - 1) 708#define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO) 709 710/* Pool configuration locks */ 711extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw); 712extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw); 713extern void spa_config_exit(spa_t *spa, int locks, void *tag); 714extern int spa_config_held(spa_t *spa, int locks, krw_t rw); 715 716/* Pool vdev add/remove lock */ 717extern uint64_t spa_vdev_enter(spa_t *spa); 718extern uint64_t spa_vdev_config_enter(spa_t *spa); 719extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, 720 int error, char *tag); 721extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error); 722 723/* Pool vdev state change lock */ 724extern void spa_vdev_state_enter(spa_t *spa, int oplock); 725extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error); 726 727/* Log state */ 728typedef enum spa_log_state { 729 SPA_LOG_UNKNOWN = 0, /* unknown log state */ 730 SPA_LOG_MISSING, /* missing log(s) */ 731 SPA_LOG_CLEAR, /* clear the log(s) */ 732 SPA_LOG_GOOD, /* log(s) are good */ 733} spa_log_state_t; 734 735extern spa_log_state_t spa_get_log_state(spa_t *spa); 736extern void spa_set_log_state(spa_t *spa, spa_log_state_t state); 737extern int spa_offline_log(spa_t *spa); 738 739/* Log claim callback */ 740extern void spa_claim_notify(zio_t *zio); 741 742/* Accessor functions */ 743extern boolean_t spa_shutting_down(spa_t *spa); 744extern struct dsl_pool *spa_get_dsl(spa_t *spa); 745extern boolean_t spa_is_initializing(spa_t *spa); 746extern blkptr_t *spa_get_rootblkptr(spa_t *spa); 747extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp); 748extern void spa_altroot(spa_t *, char *, size_t); 749extern int spa_sync_pass(spa_t *spa); 750extern char *spa_name(spa_t *spa); 751extern uint64_t spa_guid(spa_t *spa); 752extern uint64_t spa_load_guid(spa_t *spa); 753extern uint64_t spa_last_synced_txg(spa_t *spa); 754extern uint64_t spa_first_txg(spa_t *spa); 755extern uint64_t spa_syncing_txg(spa_t *spa); 756extern uint64_t spa_version(spa_t *spa); 757extern pool_state_t spa_state(spa_t *spa); 758extern spa_load_state_t spa_load_state(spa_t *spa); 759extern uint64_t spa_freeze_txg(spa_t *spa); 760extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize); 761extern uint64_t spa_get_dspace(spa_t *spa); 762extern void spa_update_dspace(spa_t *spa); 763extern uint64_t spa_version(spa_t *spa); 764extern boolean_t spa_deflate(spa_t *spa); 765extern metaslab_class_t *spa_normal_class(spa_t *spa); 766extern metaslab_class_t *spa_log_class(spa_t *spa); 767extern int spa_max_replication(spa_t *spa); 768extern int spa_prev_software_version(spa_t *spa); 769extern int spa_busy(void); 770extern uint8_t spa_get_failmode(spa_t *spa); 771extern boolean_t spa_suspended(spa_t *spa); 772extern uint64_t spa_bootfs(spa_t *spa); 773extern uint64_t spa_delegation(spa_t *spa); 774extern objset_t *spa_meta_objset(spa_t *spa); 775extern uint64_t spa_deadman_synctime(spa_t *spa); 776 777/* Miscellaneous support routines */ 778extern void spa_activate_mos_feature(spa_t *spa, const char *feature, 779 dmu_tx_t *tx); 780extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature); 781extern int spa_rename(const char *oldname, const char *newname); 782extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid); 783extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid); 784extern char *spa_strdup(const char *); 785extern void spa_strfree(char *); 786extern uint64_t spa_get_random(uint64_t range); 787extern uint64_t spa_generate_guid(spa_t *spa); 788extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp); 789extern void spa_freeze(spa_t *spa); 790extern int spa_change_guid(spa_t *spa); 791extern void spa_upgrade(spa_t *spa, uint64_t version); 792extern void spa_evict_all(void); 793extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid, 794 boolean_t l2cache); 795extern boolean_t spa_has_spare(spa_t *, uint64_t guid); 796extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva); 797extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp); 798extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp); 799extern boolean_t spa_has_slogs(spa_t *spa); 800extern boolean_t spa_is_root(spa_t *spa); 801extern boolean_t spa_writeable(spa_t *spa); 802 803extern int spa_mode(spa_t *spa); 804extern uint64_t zfs_strtonum(const char *str, char **nptr); 805#define strtonum(str, nptr) zfs_strtonum((str), (nptr)) 806 807extern char *spa_his_ievent_table[]; 808 809extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx); 810extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read, 811 char *his_buf); 812extern int spa_history_log(spa_t *spa, const char *his_buf); 813extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl); 814extern void spa_history_log_version(spa_t *spa, const char *operation); 815extern void spa_history_log_internal(spa_t *spa, const char *operation, 816 dmu_tx_t *tx, const char *fmt, ...); 817extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op, 818 dmu_tx_t *tx, const char *fmt, ...); 819extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation, 820 dmu_tx_t *tx, const char *fmt, ...); 821 822/* error handling */ 823struct zbookmark_phys; 824extern void spa_log_error(spa_t *spa, zio_t *zio); 825extern void zfs_ereport_post(const char *cls, spa_t *spa, vdev_t *vd, 826 zio_t *zio, uint64_t stateoroffset, uint64_t length); 827extern void zfs_post_remove(spa_t *spa, vdev_t *vd); 828extern void zfs_post_state_change(spa_t *spa, vdev_t *vd); 829extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd); 830extern uint64_t spa_get_errlog_size(spa_t *spa); 831extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count); 832extern void spa_errlog_rotate(spa_t *spa); 833extern void spa_errlog_drain(spa_t *spa); 834extern void spa_errlog_sync(spa_t *spa, uint64_t txg); 835extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub); 836 837/* vdev cache */ 838extern void vdev_cache_stat_init(void); 839extern void vdev_cache_stat_fini(void); 840 841/* Initialization and termination */ 842extern void spa_init(int flags); 843extern void spa_fini(void); 844extern void spa_boot_init(); 845 846/* properties */ 847extern int spa_prop_set(spa_t *spa, nvlist_t *nvp); 848extern int spa_prop_get(spa_t *spa, nvlist_t **nvp); 849extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx); 850extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t); 851 852/* asynchronous event notification */ 853extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name); 854 855#ifdef ZFS_DEBUG 856#define dprintf_bp(bp, fmt, ...) do { \ 857 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 858 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \ 859 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \ 860 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \ 861 kmem_free(__blkbuf, BP_SPRINTF_LEN); \ 862 } \ 863_NOTE(CONSTCOND) } while (0) 864#else 865#define dprintf_bp(bp, fmt, ...) 866#endif 867 868extern boolean_t spa_debug_enabled(spa_t *spa); 869#define spa_dbgmsg(spa, ...) \ 870{ \ 871 if (spa_debug_enabled(spa)) \ 872 zfs_dbgmsg(__VA_ARGS__); \ 873} 874 875extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */ 876 877#ifdef __cplusplus 878} 879#endif 880 881#endif /* _SYS_SPA_H */ 882