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 2013 Saso Kiselkov. All rights reserved. 24 * Copyright (c) 2011, 2014 by Delphix. All rights reserved. 25 * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 26 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. 27 */ 28 29 #ifndef _SYS_SPA_H 30 #define _SYS_SPA_H 31 32 #include <sys/avl.h> 33 #include <sys/zfs_context.h> 34 #include <sys/nvpair.h> 35 #include <sys/sysmacros.h> 36 #include <sys/types.h> 37 #include <sys/fs/zfs.h> 38 39 #ifdef __cplusplus 40 extern "C" { 41 #endif 42 43 /* 44 * Forward references that lots of things need. 45 */ 46 typedef struct spa spa_t; 47 typedef struct vdev vdev_t; 48 typedef struct metaslab metaslab_t; 49 typedef struct metaslab_group metaslab_group_t; 50 typedef struct metaslab_class metaslab_class_t; 51 typedef struct zio zio_t; 52 typedef struct zilog zilog_t; 53 typedef struct spa_aux_vdev spa_aux_vdev_t; 54 typedef struct ddt ddt_t; 55 typedef struct ddt_entry ddt_entry_t; 56 struct dsl_pool; 57 struct dsl_dataset; 58 59 /* 60 * General-purpose 32-bit and 64-bit bitfield encodings. 61 */ 62 #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) 63 #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) 64 #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) 65 #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) 66 67 #define BF32_GET(x, low, len) BF32_DECODE(x, low, len) 68 #define BF64_GET(x, low, len) BF64_DECODE(x, low, len) 69 70 #define BF32_SET(x, low, len, val) do { \ 71 ASSERT3U(val, <, 1U << (len)); \ 72 ASSERT3U(low + len, <=, 32); \ 73 (x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \ 74 _NOTE(CONSTCOND) } while (0) 75 76 #define BF64_SET(x, low, len, val) do { \ 77 ASSERT3U(val, <, 1ULL << (len)); \ 78 ASSERT3U(low + len, <=, 64); \ 79 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \ 80 _NOTE(CONSTCOND) } while (0) 81 82 #define BF32_GET_SB(x, low, len, shift, bias) \ 83 ((BF32_GET(x, low, len) + (bias)) << (shift)) 84 #define BF64_GET_SB(x, low, len, shift, bias) \ 85 ((BF64_GET(x, low, len) + (bias)) << (shift)) 86 87 #define BF32_SET_SB(x, low, len, shift, bias, val) do { \ 88 ASSERT(IS_P2ALIGNED(val, 1U << shift)); \ 89 ASSERT3S((val) >> (shift), >=, bias); \ 90 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \ 91 _NOTE(CONSTCOND) } while (0) 92 #define BF64_SET_SB(x, low, len, shift, bias, val) do { \ 93 ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \ 94 ASSERT3S((val) >> (shift), >=, bias); \ 95 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \ 96 _NOTE(CONSTCOND) } while (0) 97 98 /* 99 * We currently support block sizes from 512 bytes to 16MB. 100 * The benefits of larger blocks, and thus larger IO, need to be weighed 101 * against the cost of COWing a giant block to modify one byte, and the 102 * large latency of reading or writing a large block. 103 * 104 * Note that although blocks up to 16MB are supported, the recordsize 105 * property can not be set larger than zfs_max_recordsize (default 1MB). 106 * See the comment near zfs_max_recordsize in dsl_dataset.c for details. 107 * 108 * Note that although the LSIZE field of the blkptr_t can store sizes up 109 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to 110 * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB. 111 */ 112 #define SPA_MINBLOCKSHIFT 9 113 #define SPA_OLD_MAXBLOCKSHIFT 17 114 #define SPA_MAXBLOCKSHIFT 24 115 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 116 #define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT) 117 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 118 119 /* 120 * Size of block to hold the configuration data (a packed nvlist) 121 */ 122 #define SPA_CONFIG_BLOCKSIZE (1ULL << 14) 123 124 /* 125 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. 126 * The ASIZE encoding should be at least 64 times larger (6 more bits) 127 * to support up to 4-way RAID-Z mirror mode with worst-case gang block 128 * overhead, three DVAs per bp, plus one more bit in case we do anything 129 * else that expands the ASIZE. 130 */ 131 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ 132 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ 133 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ 134 135 /* 136 * All SPA data is represented by 128-bit data virtual addresses (DVAs). 137 * The members of the dva_t should be considered opaque outside the SPA. 138 */ 139 typedef struct dva { 140 uint64_t dva_word[2]; 141 } dva_t; 142 143 /* 144 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. 145 */ 146 typedef struct zio_cksum { 147 uint64_t zc_word[4]; 148 } zio_cksum_t; 149 150 /* 151 * Each block is described by its DVAs, time of birth, checksum, etc. 152 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 153 * 154 * 64 56 48 40 32 24 16 8 0 155 * +-------+-------+-------+-------+-------+-------+-------+-------+ 156 * 0 | vdev1 | GRID | ASIZE | 157 * +-------+-------+-------+-------+-------+-------+-------+-------+ 158 * 1 |G| offset1 | 159 * +-------+-------+-------+-------+-------+-------+-------+-------+ 160 * 2 | vdev2 | GRID | ASIZE | 161 * +-------+-------+-------+-------+-------+-------+-------+-------+ 162 * 3 |G| offset2 | 163 * +-------+-------+-------+-------+-------+-------+-------+-------+ 164 * 4 | vdev3 | GRID | ASIZE | 165 * +-------+-------+-------+-------+-------+-------+-------+-------+ 166 * 5 |G| offset3 | 167 * +-------+-------+-------+-------+-------+-------+-------+-------+ 168 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 169 * +-------+-------+-------+-------+-------+-------+-------+-------+ 170 * 7 | padding | 171 * +-------+-------+-------+-------+-------+-------+-------+-------+ 172 * 8 | padding | 173 * +-------+-------+-------+-------+-------+-------+-------+-------+ 174 * 9 | physical birth txg | 175 * +-------+-------+-------+-------+-------+-------+-------+-------+ 176 * a | logical birth txg | 177 * +-------+-------+-------+-------+-------+-------+-------+-------+ 178 * b | fill count | 179 * +-------+-------+-------+-------+-------+-------+-------+-------+ 180 * c | checksum[0] | 181 * +-------+-------+-------+-------+-------+-------+-------+-------+ 182 * d | checksum[1] | 183 * +-------+-------+-------+-------+-------+-------+-------+-------+ 184 * e | checksum[2] | 185 * +-------+-------+-------+-------+-------+-------+-------+-------+ 186 * f | checksum[3] | 187 * +-------+-------+-------+-------+-------+-------+-------+-------+ 188 * 189 * Legend: 190 * 191 * vdev virtual device ID 192 * offset offset into virtual device 193 * LSIZE logical size 194 * PSIZE physical size (after compression) 195 * ASIZE allocated size (including RAID-Z parity and gang block headers) 196 * GRID RAID-Z layout information (reserved for future use) 197 * cksum checksum function 198 * comp compression function 199 * G gang block indicator 200 * B byteorder (endianness) 201 * D dedup 202 * X encryption (on version 30, which is not supported) 203 * E blkptr_t contains embedded data (see below) 204 * lvl level of indirection 205 * type DMU object type 206 * phys birth txg of block allocation; zero if same as logical birth txg 207 * log. birth transaction group in which the block was logically born 208 * fill count number of non-zero blocks under this bp 209 * checksum[4] 256-bit checksum of the data this bp describes 210 */ 211 212 /* 213 * "Embedded" blkptr_t's don't actually point to a block, instead they 214 * have a data payload embedded in the blkptr_t itself. See the comment 215 * in blkptr.c for more details. 216 * 217 * The blkptr_t is laid out as follows: 218 * 219 * 64 56 48 40 32 24 16 8 0 220 * +-------+-------+-------+-------+-------+-------+-------+-------+ 221 * 0 | payload | 222 * 1 | payload | 223 * 2 | payload | 224 * 3 | payload | 225 * 4 | payload | 226 * 5 | payload | 227 * +-------+-------+-------+-------+-------+-------+-------+-------+ 228 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 229 * +-------+-------+-------+-------+-------+-------+-------+-------+ 230 * 7 | payload | 231 * 8 | payload | 232 * 9 | payload | 233 * +-------+-------+-------+-------+-------+-------+-------+-------+ 234 * a | logical birth txg | 235 * +-------+-------+-------+-------+-------+-------+-------+-------+ 236 * b | payload | 237 * c | payload | 238 * d | payload | 239 * e | payload | 240 * f | payload | 241 * +-------+-------+-------+-------+-------+-------+-------+-------+ 242 * 243 * Legend: 244 * 245 * payload contains the embedded data 246 * B (byteorder) byteorder (endianness) 247 * D (dedup) padding (set to zero) 248 * X encryption (set to zero; see above) 249 * E (embedded) set to one 250 * lvl indirection level 251 * type DMU object type 252 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 253 * comp compression function of payload 254 * PSIZE size of payload after compression, in bytes 255 * LSIZE logical size of payload, in bytes 256 * note that 25 bits is enough to store the largest 257 * "normal" BP's LSIZE (2^16 * 2^9) in bytes 258 * log. birth transaction group in which the block was logically born 259 * 260 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 261 * bp's they are stored in units of SPA_MINBLOCKSHIFT. 262 * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 263 * The B, D, X, lvl, type, and comp fields are stored the same as with normal 264 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 265 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 266 * other macros, as they assert that they are only used on BP's of the correct 267 * "embedded-ness". 268 */ 269 270 #define BPE_GET_ETYPE(bp) \ 271 (ASSERT(BP_IS_EMBEDDED(bp)), \ 272 BF64_GET((bp)->blk_prop, 40, 8)) 273 #define BPE_SET_ETYPE(bp, t) do { \ 274 ASSERT(BP_IS_EMBEDDED(bp)); \ 275 BF64_SET((bp)->blk_prop, 40, 8, t); \ 276 _NOTE(CONSTCOND) } while (0) 277 278 #define BPE_GET_LSIZE(bp) \ 279 (ASSERT(BP_IS_EMBEDDED(bp)), \ 280 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 281 #define BPE_SET_LSIZE(bp, x) do { \ 282 ASSERT(BP_IS_EMBEDDED(bp)); \ 283 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 284 _NOTE(CONSTCOND) } while (0) 285 286 #define BPE_GET_PSIZE(bp) \ 287 (ASSERT(BP_IS_EMBEDDED(bp)), \ 288 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 289 #define BPE_SET_PSIZE(bp, x) do { \ 290 ASSERT(BP_IS_EMBEDDED(bp)); \ 291 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 292 _NOTE(CONSTCOND) } while (0) 293 294 typedef enum bp_embedded_type { 295 BP_EMBEDDED_TYPE_DATA, 296 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ 297 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED 298 } bp_embedded_type_t; 299 300 #define BPE_NUM_WORDS 14 301 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 302 #define BPE_IS_PAYLOADWORD(bp, wp) \ 303 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 304 305 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 306 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 307 308 /* 309 * A block is a hole when it has either 1) never been written to, or 310 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads 311 * without physically allocating disk space. Holes are represented in the 312 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is 313 * done through the BP_IS_HOLE macro. For holes, the logical size, level, 314 * DMU object type, and birth times are all also stored for holes that 315 * were written to at some point (i.e. were punched after having been filled). 316 */ 317 typedef struct blkptr { 318 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 319 uint64_t blk_prop; /* size, compression, type, etc */ 320 uint64_t blk_pad[2]; /* Extra space for the future */ 321 uint64_t blk_phys_birth; /* txg when block was allocated */ 322 uint64_t blk_birth; /* transaction group at birth */ 323 uint64_t blk_fill; /* fill count */ 324 zio_cksum_t blk_cksum; /* 256-bit checksum */ 325 } blkptr_t; 326 327 /* 328 * Macros to get and set fields in a bp or DVA. 329 */ 330 #define DVA_GET_ASIZE(dva) \ 331 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 332 #define DVA_SET_ASIZE(dva, x) \ 333 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 334 SPA_MINBLOCKSHIFT, 0, x) 335 336 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 337 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 338 339 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) 340 #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) 341 342 #define DVA_GET_OFFSET(dva) \ 343 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 344 #define DVA_SET_OFFSET(dva, x) \ 345 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 346 347 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 348 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 349 350 #define BP_GET_LSIZE(bp) \ 351 (BP_IS_EMBEDDED(bp) ? \ 352 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 353 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 354 #define BP_SET_LSIZE(bp, x) do { \ 355 ASSERT(!BP_IS_EMBEDDED(bp)); \ 356 BF64_SET_SB((bp)->blk_prop, \ 357 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 358 _NOTE(CONSTCOND) } while (0) 359 360 #define BP_GET_PSIZE(bp) \ 361 (BP_IS_EMBEDDED(bp) ? 0 : \ 362 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 363 #define BP_SET_PSIZE(bp, x) do { \ 364 ASSERT(!BP_IS_EMBEDDED(bp)); \ 365 BF64_SET_SB((bp)->blk_prop, \ 366 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 367 _NOTE(CONSTCOND) } while (0) 368 369 #define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7) 370 #define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x) 371 372 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 373 #define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x) 374 375 #define BP_GET_CHECKSUM(bp) \ 376 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \ 377 BF64_GET((bp)->blk_prop, 40, 8)) 378 #define BP_SET_CHECKSUM(bp, x) do { \ 379 ASSERT(!BP_IS_EMBEDDED(bp)); \ 380 BF64_SET((bp)->blk_prop, 40, 8, x); \ 381 _NOTE(CONSTCOND) } while (0) 382 383 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 384 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 385 386 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 387 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 388 389 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 390 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 391 392 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 393 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 394 395 #define BP_PHYSICAL_BIRTH(bp) \ 396 (BP_IS_EMBEDDED(bp) ? 0 : \ 397 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 398 399 #define BP_SET_BIRTH(bp, logical, physical) \ 400 { \ 401 ASSERT(!BP_IS_EMBEDDED(bp)); \ 402 (bp)->blk_birth = (logical); \ 403 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \ 404 } 405 406 #define BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill) 407 408 #define BP_GET_ASIZE(bp) \ 409 (BP_IS_EMBEDDED(bp) ? 0 : \ 410 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 411 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 412 DVA_GET_ASIZE(&(bp)->blk_dva[2])) 413 414 #define BP_GET_UCSIZE(bp) \ 415 ((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \ 416 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)) 417 418 #define BP_GET_NDVAS(bp) \ 419 (BP_IS_EMBEDDED(bp) ? 0 : \ 420 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 421 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 422 !!DVA_GET_ASIZE(&(bp)->blk_dva[2])) 423 424 #define BP_COUNT_GANG(bp) \ 425 (BP_IS_EMBEDDED(bp) ? 0 : \ 426 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \ 427 DVA_GET_GANG(&(bp)->blk_dva[1]) + \ 428 DVA_GET_GANG(&(bp)->blk_dva[2]))) 429 430 #define DVA_EQUAL(dva1, dva2) \ 431 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 432 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 433 434 #define BP_EQUAL(bp1, bp2) \ 435 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \ 436 (bp1)->blk_birth == (bp2)->blk_birth && \ 437 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \ 438 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \ 439 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2])) 440 441 #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 442 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 443 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 444 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 445 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 446 447 #define ZIO_CHECKSUM_IS_ZERO(zc) \ 448 (0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \ 449 (zc)->zc_word[2] | (zc)->zc_word[3])) 450 451 #define ZIO_CHECKSUM_BSWAP(_zc) \ 452 do { \ 453 zio_cksum_t *zc = (_zc); \ 454 zc->zc_word[0] = BSWAP_64(zc->zc_word[0]); \ 455 zc->zc_word[1] = BSWAP_64(zc->zc_word[1]); \ 456 zc->zc_word[2] = BSWAP_64(zc->zc_word[2]); \ 457 zc->zc_word[3] = BSWAP_64(zc->zc_word[3]); \ 458 _NOTE(NOTREACHED) \ 459 _NOTE(CONSTCOND) \ 460 } while (0) 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 #ifdef _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, \ 527 "HOLE [L%llu %s] " \ 528 "size=%llxL birth=%lluL", \ 529 (u_longlong_t)BP_GET_LEVEL(bp), \ 530 type, \ 531 (u_longlong_t)BP_GET_LSIZE(bp), \ 532 (u_longlong_t)bp->blk_birth); \ 533 } else if (BP_IS_EMBEDDED(bp)) { \ 534 len = func(buf + len, size - len, \ 535 "EMBEDDED [L%llu %s] et=%u %s " \ 536 "size=%llxL/%llxP birth=%lluL", \ 537 (u_longlong_t)BP_GET_LEVEL(bp), \ 538 type, \ 539 (int)BPE_GET_ETYPE(bp), \ 540 compress, \ 541 (u_longlong_t)BPE_GET_LSIZE(bp), \ 542 (u_longlong_t)BPE_GET_PSIZE(bp), \ 543 (u_longlong_t)bp->blk_birth); \ 544 } else { \ 545 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \ 546 const dva_t *dva = &bp->blk_dva[d]; \ 547 if (DVA_IS_VALID(dva)) \ 548 copies++; \ 549 len += func(buf + len, size - len, \ 550 "DVA[%d]=<%llu:%llx:%llx>%c", d, \ 551 (u_longlong_t)DVA_GET_VDEV(dva), \ 552 (u_longlong_t)DVA_GET_OFFSET(dva), \ 553 (u_longlong_t)DVA_GET_ASIZE(dva), \ 554 ws); \ 555 } \ 556 if (BP_IS_GANG(bp) && \ 557 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \ 558 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \ 559 copies--; \ 560 len += func(buf + len, size - len, \ 561 "[L%llu %s] %s %s %s %s %s %s%c" \ 562 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \ 563 "cksum=%llx:%llx:%llx:%llx", \ 564 (u_longlong_t)BP_GET_LEVEL(bp), \ 565 type, \ 566 checksum, \ 567 compress, \ 568 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \ 569 BP_IS_GANG(bp) ? "gang" : "contiguous", \ 570 BP_GET_DEDUP(bp) ? "dedup" : "unique", \ 571 copyname[copies], \ 572 ws, \ 573 (u_longlong_t)BP_GET_LSIZE(bp), \ 574 (u_longlong_t)BP_GET_PSIZE(bp), \ 575 (u_longlong_t)bp->blk_birth, \ 576 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \ 577 (u_longlong_t)BP_GET_FILL(bp), \ 578 ws, \ 579 (u_longlong_t)bp->blk_cksum.zc_word[0], \ 580 (u_longlong_t)bp->blk_cksum.zc_word[1], \ 581 (u_longlong_t)bp->blk_cksum.zc_word[2], \ 582 (u_longlong_t)bp->blk_cksum.zc_word[3]); \ 583 } \ 584 ASSERT(len < size); \ 585 } 586 587 #include <sys/dmu.h> 588 589 #define BP_GET_BUFC_TYPE(bp) \ 590 (((BP_GET_LEVEL(bp) > 0) || (DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))) ? \ 591 ARC_BUFC_METADATA : ARC_BUFC_DATA) 592 593 typedef enum spa_import_type { 594 SPA_IMPORT_EXISTING, 595 SPA_IMPORT_ASSEMBLE 596 } spa_import_type_t; 597 598 /* state manipulation functions */ 599 extern int spa_open(const char *pool, spa_t **, void *tag); 600 extern int spa_open_rewind(const char *pool, spa_t **, void *tag, 601 nvlist_t *policy, nvlist_t **config); 602 extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot, 603 size_t buflen); 604 extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props, 605 nvlist_t *zplprops); 606 extern int spa_import_rootpool(char *devpath, char *devid); 607 extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props, 608 uint64_t flags); 609 extern nvlist_t *spa_tryimport(nvlist_t *tryconfig); 610 extern int spa_destroy(char *pool); 611 extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force, 612 boolean_t hardforce); 613 extern int spa_reset(char *pool); 614 extern void spa_async_request(spa_t *spa, int flag); 615 extern void spa_async_unrequest(spa_t *spa, int flag); 616 extern void spa_async_suspend(spa_t *spa); 617 extern void spa_async_resume(spa_t *spa); 618 extern spa_t *spa_inject_addref(char *pool); 619 extern void spa_inject_delref(spa_t *spa); 620 extern void spa_scan_stat_init(spa_t *spa); 621 extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps); 622 623 #define SPA_ASYNC_CONFIG_UPDATE 0x01 624 #define SPA_ASYNC_REMOVE 0x02 625 #define SPA_ASYNC_PROBE 0x04 626 #define SPA_ASYNC_RESILVER_DONE 0x08 627 #define SPA_ASYNC_RESILVER 0x10 628 #define SPA_ASYNC_AUTOEXPAND 0x20 629 #define SPA_ASYNC_REMOVE_DONE 0x40 630 #define SPA_ASYNC_REMOVE_STOP 0x80 631 #define SPA_ASYNC_L2CACHE_REBUILD 0x100 632 633 /* 634 * Controls the behavior of spa_vdev_remove(). 635 */ 636 #define SPA_REMOVE_UNSPARE 0x01 637 #define SPA_REMOVE_DONE 0x02 638 639 /* device manipulation */ 640 extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot); 641 extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, 642 int replacing); 643 extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, 644 int replace_done); 645 extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare); 646 extern boolean_t spa_vdev_remove_active(spa_t *spa); 647 extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath); 648 extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru); 649 extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config, 650 nvlist_t *props, boolean_t exp); 651 652 /* spare state (which is global across all pools) */ 653 extern void spa_spare_add(vdev_t *vd); 654 extern void spa_spare_remove(vdev_t *vd); 655 extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt); 656 extern void spa_spare_activate(vdev_t *vd); 657 658 /* L2ARC state (which is global across all pools) */ 659 extern void spa_l2cache_add(vdev_t *vd); 660 extern void spa_l2cache_remove(vdev_t *vd); 661 extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool); 662 extern void spa_l2cache_activate(vdev_t *vd); 663 extern void spa_l2cache_drop(spa_t *spa); 664 665 /* scanning */ 666 extern int spa_scan(spa_t *spa, pool_scan_func_t func); 667 extern int spa_scan_stop(spa_t *spa); 668 669 /* spa syncing */ 670 extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */ 671 extern void spa_sync_allpools(void); 672 673 /* spa namespace global mutex */ 674 extern kmutex_t spa_namespace_lock; 675 676 /* 677 * SPA configuration functions in spa_config.c 678 */ 679 680 #define SPA_CONFIG_UPDATE_POOL 0 681 #define SPA_CONFIG_UPDATE_VDEVS 1 682 683 extern void spa_config_sync(spa_t *, boolean_t, boolean_t); 684 extern void spa_config_load(void); 685 extern nvlist_t *spa_all_configs(uint64_t *); 686 extern void spa_config_set(spa_t *spa, nvlist_t *config); 687 extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, 688 int getstats); 689 extern void spa_config_update(spa_t *spa, int what); 690 691 /* 692 * Miscellaneous SPA routines in spa_misc.c 693 */ 694 695 /* Namespace manipulation */ 696 extern spa_t *spa_lookup(const char *name); 697 extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot); 698 extern void spa_remove(spa_t *spa); 699 extern spa_t *spa_next(spa_t *prev); 700 701 /* Refcount functions */ 702 extern void spa_open_ref(spa_t *spa, void *tag); 703 extern void spa_close(spa_t *spa, void *tag); 704 extern void spa_async_close(spa_t *spa, void *tag); 705 extern 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 */ 720 extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw); 721 extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw); 722 extern void spa_config_exit(spa_t *spa, int locks, void *tag); 723 extern int spa_config_held(spa_t *spa, int locks, krw_t rw); 724 725 /* Pool vdev add/remove lock */ 726 extern uint64_t spa_vdev_enter(spa_t *spa); 727 extern uint64_t spa_vdev_config_enter(spa_t *spa); 728 extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, 729 int error, char *tag); 730 extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error); 731 732 /* Pool vdev state change lock */ 733 extern void spa_vdev_state_enter(spa_t *spa, int oplock); 734 extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error); 735 736 /* Log state */ 737 typedef 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; 743 744 extern spa_log_state_t spa_get_log_state(spa_t *spa); 745 extern void spa_set_log_state(spa_t *spa, spa_log_state_t state); 746 extern int spa_offline_log(spa_t *spa); 747 748 /* Log claim callback */ 749 extern void spa_claim_notify(zio_t *zio); 750 751 /* Accessor functions */ 752 extern boolean_t spa_shutting_down(spa_t *spa); 753 extern struct dsl_pool *spa_get_dsl(spa_t *spa); 754 extern boolean_t spa_is_initializing(spa_t *spa); 755 extern blkptr_t *spa_get_rootblkptr(spa_t *spa); 756 extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp); 757 extern void spa_altroot(spa_t *, char *, size_t); 758 extern int spa_sync_pass(spa_t *spa); 759 extern char *spa_name(spa_t *spa); 760 extern uint64_t spa_guid(spa_t *spa); 761 extern uint64_t spa_load_guid(spa_t *spa); 762 extern uint64_t spa_last_synced_txg(spa_t *spa); 763 extern uint64_t spa_first_txg(spa_t *spa); 764 extern uint64_t spa_syncing_txg(spa_t *spa); 765 extern uint64_t spa_version(spa_t *spa); 766 extern pool_state_t spa_state(spa_t *spa); 767 extern spa_load_state_t spa_load_state(spa_t *spa); 768 extern uint64_t spa_freeze_txg(spa_t *spa); 769 extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize); 770 extern uint64_t spa_get_dspace(spa_t *spa); 771 extern uint64_t spa_get_slop_space(spa_t *spa); 772 extern void spa_update_dspace(spa_t *spa); 773 extern uint64_t spa_version(spa_t *spa); 774 extern boolean_t spa_deflate(spa_t *spa); 775 extern metaslab_class_t *spa_normal_class(spa_t *spa); 776 extern metaslab_class_t *spa_log_class(spa_t *spa); 777 extern void spa_evicting_os_register(spa_t *, objset_t *os); 778 extern void spa_evicting_os_deregister(spa_t *, objset_t *os); 779 extern void spa_evicting_os_wait(spa_t *spa); 780 extern int spa_max_replication(spa_t *spa); 781 extern int spa_prev_software_version(spa_t *spa); 782 extern int spa_busy(void); 783 extern uint8_t spa_get_failmode(spa_t *spa); 784 extern boolean_t spa_suspended(spa_t *spa); 785 extern uint64_t spa_bootfs(spa_t *spa); 786 extern uint64_t spa_delegation(spa_t *spa); 787 extern objset_t *spa_meta_objset(spa_t *spa); 788 extern uint64_t spa_deadman_synctime(spa_t *spa); 789 790 /* Miscellaneous support routines */ 791 extern void spa_activate_mos_feature(spa_t *spa, const char *feature, 792 dmu_tx_t *tx); 793 extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature); 794 extern int spa_rename(const char *oldname, const char *newname); 795 extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid); 796 extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid); 797 extern char *spa_strdup(const char *); 798 extern void spa_strfree(char *); 799 extern uint64_t spa_get_random(uint64_t range); 800 extern uint64_t spa_generate_guid(spa_t *spa); 801 extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp); 802 extern void spa_freeze(spa_t *spa); 803 extern int spa_change_guid(spa_t *spa); 804 extern void spa_upgrade(spa_t *spa, uint64_t version); 805 extern void spa_evict_all(void); 806 extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid, 807 boolean_t l2cache); 808 extern boolean_t spa_has_spare(spa_t *, uint64_t guid); 809 extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva); 810 extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp); 811 extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp); 812 extern boolean_t spa_has_slogs(spa_t *spa); 813 extern boolean_t spa_is_root(spa_t *spa); 814 extern boolean_t spa_writeable(spa_t *spa); 815 extern boolean_t spa_has_pending_synctask(spa_t *spa); 816 extern int spa_maxblocksize(spa_t *spa); 817 extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp); 818 819 extern int spa_mode(spa_t *spa); 820 extern uint64_t strtonum(const char *str, char **nptr); 821 822 extern char *spa_his_ievent_table[]; 823 824 extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx); 825 extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read, 826 char *his_buf); 827 extern int spa_history_log(spa_t *spa, const char *his_buf); 828 extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl); 829 extern void spa_history_log_version(spa_t *spa, const char *operation); 830 extern void spa_history_log_internal(spa_t *spa, const char *operation, 831 dmu_tx_t *tx, const char *fmt, ...); 832 extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op, 833 dmu_tx_t *tx, const char *fmt, ...); 834 extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation, 835 dmu_tx_t *tx, const char *fmt, ...); 836 837 /* error handling */ 838 struct zbookmark_phys; 839 extern void spa_log_error(spa_t *spa, zio_t *zio); 840 extern void zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd, 841 zio_t *zio, uint64_t stateoroffset, uint64_t length); 842 extern void zfs_post_remove(spa_t *spa, vdev_t *vd); 843 extern void zfs_post_state_change(spa_t *spa, vdev_t *vd); 844 extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd); 845 extern uint64_t spa_get_errlog_size(spa_t *spa); 846 extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count); 847 extern void spa_errlog_rotate(spa_t *spa); 848 extern void spa_errlog_drain(spa_t *spa); 849 extern void spa_errlog_sync(spa_t *spa, uint64_t txg); 850 extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub); 851 852 /* vdev cache */ 853 extern void vdev_cache_stat_init(void); 854 extern void vdev_cache_stat_fini(void); 855 856 /* Initialization and termination */ 857 extern void spa_init(int flags); 858 extern void spa_fini(void); 859 extern void spa_boot_init(); 860 861 /* properties */ 862 extern int spa_prop_set(spa_t *spa, nvlist_t *nvp); 863 extern int spa_prop_get(spa_t *spa, nvlist_t **nvp); 864 extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx); 865 extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t); 866 867 /* asynchronous event notification */ 868 extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name); 869 870 #ifdef ZFS_DEBUG 871 #define dprintf_bp(bp, fmt, ...) do { \ 872 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 873 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \ 874 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \ 875 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \ 876 kmem_free(__blkbuf, BP_SPRINTF_LEN); \ 877 } \ 878 _NOTE(CONSTCOND) } while (0) 879 #else 880 #define dprintf_bp(bp, fmt, ...) 881 #endif 882 883 extern boolean_t spa_debug_enabled(spa_t *spa); 884 #define spa_dbgmsg(spa, ...) \ 885 { \ 886 if (spa_debug_enabled(spa)) \ 887 zfs_dbgmsg(__VA_ARGS__); \ 888 } 889 890 extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */ 891 892 #ifdef __cplusplus 893 } 894 #endif 895 896 #endif /* _SYS_SPA_H */ 897