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