1 /* 2 * Copyright (c) 2002 McAfee, Inc. 3 * All rights reserved. 4 * 5 * This software was developed for the FreeBSD Project by Marshall 6 * Kirk McKusick and McAfee Research,, the Security Research Division of 7 * McAfee, Inc. under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as 8 * part of the DARPA CHATS research program 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 /* 32 * CDDL HEADER START 33 * 34 * The contents of this file are subject to the terms of the 35 * Common Development and Distribution License (the "License"). 36 * You may not use this file except in compliance with the License. 37 * 38 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 39 * or http://www.opensolaris.org/os/licensing. 40 * See the License for the specific language governing permissions 41 * and limitations under the License. 42 * 43 * When distributing Covered Code, include this CDDL HEADER in each 44 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 45 * If applicable, add the following below this CDDL HEADER, with the 46 * fields enclosed by brackets "[]" replaced with your own identifying 47 * information: Portions Copyright [yyyy] [name of copyright owner] 48 * 49 * CDDL HEADER END 50 */ 51 /* 52 * Copyright 2009 Sun Microsystems, Inc. All rights reserved. 53 * Use is subject to license terms. 54 */ 55 /* 56 * Copyright 2013 by Saso Kiselkov. All rights reserved. 57 */ 58 /* 59 * Copyright (c) 2013 by Delphix. All rights reserved. 60 */ 61 62 #ifndef _ZFSIMPL_H_ 63 #define _ZFSIMPL_H_ 64 65 #include <sys/queue.h> 66 #include <sys/list.h> 67 #include <bootstrap.h> 68 69 #define MAXNAMELEN 256 70 71 #define _NOTE(s) 72 73 /* 74 * AVL comparator helpers 75 */ 76 #define AVL_ISIGN(a) (((a) > 0) - ((a) < 0)) 77 #define AVL_CMP(a, b) (((a) > (b)) - ((a) < (b))) 78 #define AVL_PCMP(a, b) \ 79 (((uintptr_t)(a) > (uintptr_t)(b)) - ((uintptr_t)(a) < (uintptr_t)(b))) 80 81 /* CRC64 table */ 82 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ 83 84 /* 85 * Macros for various sorts of alignment and rounding when the alignment 86 * is known to be a power of 2. 87 */ 88 #define P2ALIGN(x, align) ((x) & -(align)) 89 #define P2PHASE(x, align) ((x) & ((align) - 1)) 90 #define P2NPHASE(x, align) (-(x) & ((align) - 1)) 91 #define P2ROUNDUP(x, align) (-(-(x) & -(align))) 92 #define P2END(x, align) (-(~(x) & -(align))) 93 #define P2PHASEUP(x, align, phase) ((phase) - (((phase) - (x)) & -(align))) 94 #define P2BOUNDARY(off, len, align) \ 95 (((off) ^ ((off) + (len) - 1)) > (align) - 1) 96 97 /* 98 * General-purpose 32-bit and 64-bit bitfield encodings. 99 */ 100 #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) 101 #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) 102 #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) 103 #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) 104 105 #define BF32_GET(x, low, len) BF32_DECODE(x, low, len) 106 #define BF64_GET(x, low, len) BF64_DECODE(x, low, len) 107 108 #define BF32_SET(x, low, len, val) \ 109 ((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len)) 110 #define BF64_SET(x, low, len, val) \ 111 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)) 112 113 #define BF32_GET_SB(x, low, len, shift, bias) \ 114 ((BF32_GET(x, low, len) + (bias)) << (shift)) 115 #define BF64_GET_SB(x, low, len, shift, bias) \ 116 ((BF64_GET(x, low, len) + (bias)) << (shift)) 117 118 #define BF32_SET_SB(x, low, len, shift, bias, val) \ 119 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)) 120 #define BF64_SET_SB(x, low, len, shift, bias, val) \ 121 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)) 122 123 /* 124 * Macros to reverse byte order 125 */ 126 #define BSWAP_8(x) ((x) & 0xff) 127 #define BSWAP_16(x) ((BSWAP_8(x) << 8) | BSWAP_8((x) >> 8)) 128 #define BSWAP_32(x) ((BSWAP_16(x) << 16) | BSWAP_16((x) >> 16)) 129 #define BSWAP_64(x) ((BSWAP_32(x) << 32) | BSWAP_32((x) >> 32)) 130 131 #define SPA_MINBLOCKSHIFT 9 132 #define SPA_OLDMAXBLOCKSHIFT 17 133 #define SPA_MAXBLOCKSHIFT 24 134 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 135 #define SPA_OLDMAXBLOCKSIZE (1ULL << SPA_OLDMAXBLOCKSHIFT) 136 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 137 138 /* 139 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. 140 * The ASIZE encoding should be at least 64 times larger (6 more bits) 141 * to support up to 4-way RAID-Z mirror mode with worst-case gang block 142 * overhead, three DVAs per bp, plus one more bit in case we do anything 143 * else that expands the ASIZE. 144 */ 145 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ 146 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ 147 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ 148 149 /* 150 * All SPA data is represented by 128-bit data virtual addresses (DVAs). 151 * The members of the dva_t should be considered opaque outside the SPA. 152 */ 153 typedef struct dva { 154 uint64_t dva_word[2]; 155 } dva_t; 156 157 /* 158 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. 159 */ 160 typedef struct zio_cksum { 161 uint64_t zc_word[4]; 162 } zio_cksum_t; 163 164 /* 165 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept 166 * secret and is suitable for use in MAC algorithms as the key. 167 */ 168 typedef struct zio_cksum_salt { 169 uint8_t zcs_bytes[32]; 170 } zio_cksum_salt_t; 171 172 /* 173 * Each block is described by its DVAs, time of birth, checksum, etc. 174 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 175 * 176 * 64 56 48 40 32 24 16 8 0 177 * +-------+-------+-------+-------+-------+-------+-------+-------+ 178 * 0 | vdev1 | GRID | ASIZE | 179 * +-------+-------+-------+-------+-------+-------+-------+-------+ 180 * 1 |G| offset1 | 181 * +-------+-------+-------+-------+-------+-------+-------+-------+ 182 * 2 | vdev2 | GRID | ASIZE | 183 * +-------+-------+-------+-------+-------+-------+-------+-------+ 184 * 3 |G| offset2 | 185 * +-------+-------+-------+-------+-------+-------+-------+-------+ 186 * 4 | vdev3 | GRID | ASIZE | 187 * +-------+-------+-------+-------+-------+-------+-------+-------+ 188 * 5 |G| offset3 | 189 * +-------+-------+-------+-------+-------+-------+-------+-------+ 190 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 191 * +-------+-------+-------+-------+-------+-------+-------+-------+ 192 * 7 | padding | 193 * +-------+-------+-------+-------+-------+-------+-------+-------+ 194 * 8 | padding | 195 * +-------+-------+-------+-------+-------+-------+-------+-------+ 196 * 9 | physical birth txg | 197 * +-------+-------+-------+-------+-------+-------+-------+-------+ 198 * a | logical birth txg | 199 * +-------+-------+-------+-------+-------+-------+-------+-------+ 200 * b | fill count | 201 * +-------+-------+-------+-------+-------+-------+-------+-------+ 202 * c | checksum[0] | 203 * +-------+-------+-------+-------+-------+-------+-------+-------+ 204 * d | checksum[1] | 205 * +-------+-------+-------+-------+-------+-------+-------+-------+ 206 * e | checksum[2] | 207 * +-------+-------+-------+-------+-------+-------+-------+-------+ 208 * f | checksum[3] | 209 * +-------+-------+-------+-------+-------+-------+-------+-------+ 210 * 211 * Legend: 212 * 213 * vdev virtual device ID 214 * offset offset into virtual device 215 * LSIZE logical size 216 * PSIZE physical size (after compression) 217 * ASIZE allocated size (including RAID-Z parity and gang block headers) 218 * GRID RAID-Z layout information (reserved for future use) 219 * cksum checksum function 220 * comp compression function 221 * G gang block indicator 222 * B byteorder (endianness) 223 * D dedup 224 * X encryption (on version 30, which is not supported) 225 * E blkptr_t contains embedded data (see below) 226 * lvl level of indirection 227 * type DMU object type 228 * phys birth txg of block allocation; zero if same as logical birth txg 229 * log. birth transaction group in which the block was logically born 230 * fill count number of non-zero blocks under this bp 231 * checksum[4] 256-bit checksum of the data this bp describes 232 */ 233 234 /* 235 * "Embedded" blkptr_t's don't actually point to a block, instead they 236 * have a data payload embedded in the blkptr_t itself. See the comment 237 * in blkptr.c for more details. 238 * 239 * The blkptr_t is laid out as follows: 240 * 241 * 64 56 48 40 32 24 16 8 0 242 * +-------+-------+-------+-------+-------+-------+-------+-------+ 243 * 0 | payload | 244 * 1 | payload | 245 * 2 | payload | 246 * 3 | payload | 247 * 4 | payload | 248 * 5 | payload | 249 * +-------+-------+-------+-------+-------+-------+-------+-------+ 250 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 251 * +-------+-------+-------+-------+-------+-------+-------+-------+ 252 * 7 | payload | 253 * 8 | payload | 254 * 9 | payload | 255 * +-------+-------+-------+-------+-------+-------+-------+-------+ 256 * a | logical birth txg | 257 * +-------+-------+-------+-------+-------+-------+-------+-------+ 258 * b | payload | 259 * c | payload | 260 * d | payload | 261 * e | payload | 262 * f | payload | 263 * +-------+-------+-------+-------+-------+-------+-------+-------+ 264 * 265 * Legend: 266 * 267 * payload contains the embedded data 268 * B (byteorder) byteorder (endianness) 269 * D (dedup) padding (set to zero) 270 * X encryption (set to zero; see above) 271 * E (embedded) set to one 272 * lvl indirection level 273 * type DMU object type 274 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 275 * comp compression function of payload 276 * PSIZE size of payload after compression, in bytes 277 * LSIZE logical size of payload, in bytes 278 * note that 25 bits is enough to store the largest 279 * "normal" BP's LSIZE (2^16 * 2^9) in bytes 280 * log. birth transaction group in which the block was logically born 281 * 282 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 283 * bp's they are stored in units of SPA_MINBLOCKSHIFT. 284 * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 285 * The B, D, X, lvl, type, and comp fields are stored the same as with normal 286 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 287 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 288 * other macros, as they assert that they are only used on BP's of the correct 289 * "embedded-ness". 290 */ 291 292 #define BPE_GET_ETYPE(bp) \ 293 (ASSERT(BP_IS_EMBEDDED(bp)), \ 294 BF64_GET((bp)->blk_prop, 40, 8)) 295 #define BPE_SET_ETYPE(bp, t) do { \ 296 ASSERT(BP_IS_EMBEDDED(bp)); \ 297 BF64_SET((bp)->blk_prop, 40, 8, t); \ 298 _NOTE(CONSTCOND) } while (0) 299 300 #define BPE_GET_LSIZE(bp) \ 301 (ASSERT(BP_IS_EMBEDDED(bp)), \ 302 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 303 #define BPE_SET_LSIZE(bp, x) do { \ 304 ASSERT(BP_IS_EMBEDDED(bp)); \ 305 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 306 _NOTE(CONSTCOND) } while (0) 307 308 #define BPE_GET_PSIZE(bp) \ 309 (ASSERT(BP_IS_EMBEDDED(bp)), \ 310 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 311 #define BPE_SET_PSIZE(bp, x) do { \ 312 ASSERT(BP_IS_EMBEDDED(bp)); \ 313 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 314 _NOTE(CONSTCOND) } while (0) 315 316 typedef enum bp_embedded_type { 317 BP_EMBEDDED_TYPE_DATA, 318 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ 319 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED 320 } bp_embedded_type_t; 321 322 #define BPE_NUM_WORDS 14 323 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 324 #define BPE_IS_PAYLOADWORD(bp, wp) \ 325 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 326 327 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 328 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 329 330 typedef struct blkptr { 331 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 332 uint64_t blk_prop; /* size, compression, type, etc */ 333 uint64_t blk_pad[2]; /* Extra space for the future */ 334 uint64_t blk_phys_birth; /* txg when block was allocated */ 335 uint64_t blk_birth; /* transaction group at birth */ 336 uint64_t blk_fill; /* fill count */ 337 zio_cksum_t blk_cksum; /* 256-bit checksum */ 338 } blkptr_t; 339 340 /* 341 * Macros to get and set fields in a bp or DVA. 342 */ 343 #define DVA_GET_ASIZE(dva) \ 344 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 345 #define DVA_SET_ASIZE(dva, x) \ 346 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 347 SPA_MINBLOCKSHIFT, 0, x) 348 349 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 350 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 351 352 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) 353 #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) 354 355 #define DVA_GET_OFFSET(dva) \ 356 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 357 #define DVA_SET_OFFSET(dva, x) \ 358 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 359 360 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 361 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 362 363 #define BP_GET_LSIZE(bp) \ 364 (BP_IS_EMBEDDED(bp) ? \ 365 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 366 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 367 #define BP_SET_LSIZE(bp, x) do { \ 368 ASSERT(!BP_IS_EMBEDDED(bp)); \ 369 BF64_SET_SB((bp)->blk_prop, \ 370 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 371 _NOTE(CONSTCOND) } while (0) 372 373 #define BP_GET_PSIZE(bp) \ 374 BF64_GET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1) 375 #define BP_SET_PSIZE(bp, x) \ 376 BF64_SET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x) 377 378 #define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7) 379 #define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x) 380 381 #define BP_GET_CHECKSUM(bp) BF64_GET((bp)->blk_prop, 40, 8) 382 #define BP_SET_CHECKSUM(bp, x) BF64_SET((bp)->blk_prop, 40, 8, x) 383 384 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 385 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 386 387 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 388 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 389 390 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 391 392 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 393 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 394 395 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 396 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 397 398 #define BP_PHYSICAL_BIRTH(bp) \ 399 ((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 400 401 #define BP_GET_ASIZE(bp) \ 402 (DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 403 DVA_GET_ASIZE(&(bp)->blk_dva[2])) 404 405 #define BP_GET_UCSIZE(bp) \ 406 ((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \ 407 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)); 408 409 #define BP_GET_NDVAS(bp) \ 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 DVA_EQUAL(dva1, dva2) \ 415 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 416 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 417 418 #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 419 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 420 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 421 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 422 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 423 424 425 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 426 427 #define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ 428 { \ 429 (zcp)->zc_word[0] = w0; \ 430 (zcp)->zc_word[1] = w1; \ 431 (zcp)->zc_word[2] = w2; \ 432 (zcp)->zc_word[3] = w3; \ 433 } 434 435 #define BP_IDENTITY(bp) (&(bp)->blk_dva[0]) 436 #define BP_IS_GANG(bp) DVA_GET_GANG(BP_IDENTITY(bp)) 437 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 438 (dva)->dva_word[1] == 0ULL) 439 #define BP_IS_HOLE(bp) DVA_IS_EMPTY(BP_IDENTITY(bp)) 440 #define BP_IS_OLDER(bp, txg) (!BP_IS_HOLE(bp) && (bp)->blk_birth < (txg)) 441 442 #define BP_ZERO(bp) \ 443 { \ 444 (bp)->blk_dva[0].dva_word[0] = 0; \ 445 (bp)->blk_dva[0].dva_word[1] = 0; \ 446 (bp)->blk_dva[1].dva_word[0] = 0; \ 447 (bp)->blk_dva[1].dva_word[1] = 0; \ 448 (bp)->blk_dva[2].dva_word[0] = 0; \ 449 (bp)->blk_dva[2].dva_word[1] = 0; \ 450 (bp)->blk_prop = 0; \ 451 (bp)->blk_pad[0] = 0; \ 452 (bp)->blk_pad[1] = 0; \ 453 (bp)->blk_phys_birth = 0; \ 454 (bp)->blk_birth = 0; \ 455 (bp)->blk_fill = 0; \ 456 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 457 } 458 459 #if BYTE_ORDER == _BIG_ENDIAN 460 #define ZFS_HOST_BYTEORDER (0ULL) 461 #else 462 #define ZFS_HOST_BYTEORDER (1ULL) 463 #endif 464 465 #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) 466 #define BPE_NUM_WORDS 14 467 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 468 #define BPE_IS_PAYLOADWORD(bp, wp) \ 469 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 470 471 /* 472 * Embedded checksum 473 */ 474 #define ZEC_MAGIC 0x210da7ab10c7a11ULL 475 476 typedef struct zio_eck { 477 uint64_t zec_magic; /* for validation, endianness */ 478 zio_cksum_t zec_cksum; /* 256-bit checksum */ 479 } zio_eck_t; 480 481 /* 482 * Gang block headers are self-checksumming and contain an array 483 * of block pointers. 484 */ 485 #define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE 486 #define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \ 487 sizeof (zio_eck_t)) / sizeof (blkptr_t)) 488 #define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \ 489 sizeof (zio_eck_t) - \ 490 (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\ 491 sizeof (uint64_t)) 492 493 typedef struct zio_gbh { 494 blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS]; 495 uint64_t zg_filler[SPA_GBH_FILLER]; 496 zio_eck_t zg_tail; 497 } zio_gbh_phys_t; 498 499 #define VDEV_RAIDZ_MAXPARITY 3 500 501 #define VDEV_PAD_SIZE (8 << 10) 502 /* 2 padding areas (vl_pad1 and vl_be) to skip */ 503 #define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2 504 #define VDEV_PHYS_SIZE (112 << 10) 505 #define VDEV_UBERBLOCK_RING (128 << 10) 506 507 /* 508 * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock 509 * ring when MMP is enabled. 510 */ 511 #define MMP_BLOCKS_PER_LABEL 1 512 513 /* The largest uberblock we support is 8k. */ 514 #define MAX_UBERBLOCK_SHIFT (13) 515 #define VDEV_UBERBLOCK_SHIFT(vd) \ 516 MIN(MAX((vd)->v_top->v_ashift, UBERBLOCK_SHIFT), MAX_UBERBLOCK_SHIFT) 517 #define VDEV_UBERBLOCK_COUNT(vd) \ 518 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd)) 519 #define VDEV_UBERBLOCK_OFFSET(vd, n) \ 520 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)]) 521 #define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd)) 522 523 typedef struct vdev_phys { 524 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)]; 525 zio_eck_t vp_zbt; 526 } vdev_phys_t; 527 528 typedef enum vbe_vers { 529 /* The bootenv file is stored as ascii text in the envblock */ 530 VB_RAW = 0, 531 532 /* 533 * The bootenv file is converted to an nvlist and then packed into the 534 * envblock. 535 */ 536 VB_NVLIST = 1 537 } vbe_vers_t; 538 539 typedef struct vdev_boot_envblock { 540 uint64_t vbe_version; 541 char vbe_bootenv[VDEV_PAD_SIZE - sizeof (uint64_t) - 542 sizeof (zio_eck_t)]; 543 zio_eck_t vbe_zbt; 544 } vdev_boot_envblock_t; 545 546 CTASSERT(sizeof (vdev_boot_envblock_t) == VDEV_PAD_SIZE); 547 548 typedef struct vdev_label { 549 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */ 550 vdev_boot_envblock_t vl_be; /* 8K */ 551 vdev_phys_t vl_vdev_phys; /* 112K */ 552 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */ 553 } vdev_label_t; /* 256K total */ 554 555 /* 556 * vdev_dirty() flags 557 */ 558 #define VDD_METASLAB 0x01 559 #define VDD_DTL 0x02 560 561 /* 562 * Size and offset of embedded boot loader region on each label. 563 * The total size of the first two labels plus the boot area is 4MB. 564 */ 565 #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t)) 566 #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */ 567 568 /* 569 * Size of label regions at the start and end of each leaf device. 570 */ 571 #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE) 572 #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t)) 573 #define VDEV_LABELS 4 574 575 enum zio_checksum { 576 ZIO_CHECKSUM_INHERIT = 0, 577 ZIO_CHECKSUM_ON, 578 ZIO_CHECKSUM_OFF, 579 ZIO_CHECKSUM_LABEL, 580 ZIO_CHECKSUM_GANG_HEADER, 581 ZIO_CHECKSUM_ZILOG, 582 ZIO_CHECKSUM_FLETCHER_2, 583 ZIO_CHECKSUM_FLETCHER_4, 584 ZIO_CHECKSUM_SHA256, 585 ZIO_CHECKSUM_ZILOG2, 586 ZIO_CHECKSUM_NOPARITY, 587 ZIO_CHECKSUM_SHA512, 588 ZIO_CHECKSUM_SKEIN, 589 ZIO_CHECKSUM_EDONR, 590 ZIO_CHECKSUM_FUNCTIONS 591 }; 592 593 #define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4 594 #define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON 595 596 enum zio_compress { 597 ZIO_COMPRESS_INHERIT = 0, 598 ZIO_COMPRESS_ON, 599 ZIO_COMPRESS_OFF, 600 ZIO_COMPRESS_LZJB, 601 ZIO_COMPRESS_EMPTY, 602 ZIO_COMPRESS_GZIP_1, 603 ZIO_COMPRESS_GZIP_2, 604 ZIO_COMPRESS_GZIP_3, 605 ZIO_COMPRESS_GZIP_4, 606 ZIO_COMPRESS_GZIP_5, 607 ZIO_COMPRESS_GZIP_6, 608 ZIO_COMPRESS_GZIP_7, 609 ZIO_COMPRESS_GZIP_8, 610 ZIO_COMPRESS_GZIP_9, 611 ZIO_COMPRESS_ZLE, 612 ZIO_COMPRESS_LZ4, 613 ZIO_COMPRESS_FUNCTIONS 614 }; 615 616 #define ZIO_COMPRESS_ON_VALUE ZIO_COMPRESS_LZJB 617 #define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF 618 619 /* nvlist pack encoding */ 620 #define NV_ENCODE_NATIVE 0 621 #define NV_ENCODE_XDR 1 622 623 typedef enum { 624 DATA_TYPE_UNKNOWN = 0, 625 DATA_TYPE_BOOLEAN, 626 DATA_TYPE_BYTE, 627 DATA_TYPE_INT16, 628 DATA_TYPE_UINT16, 629 DATA_TYPE_INT32, 630 DATA_TYPE_UINT32, 631 DATA_TYPE_INT64, 632 DATA_TYPE_UINT64, 633 DATA_TYPE_STRING, 634 DATA_TYPE_BYTE_ARRAY, 635 DATA_TYPE_INT16_ARRAY, 636 DATA_TYPE_UINT16_ARRAY, 637 DATA_TYPE_INT32_ARRAY, 638 DATA_TYPE_UINT32_ARRAY, 639 DATA_TYPE_INT64_ARRAY, 640 DATA_TYPE_UINT64_ARRAY, 641 DATA_TYPE_STRING_ARRAY, 642 DATA_TYPE_HRTIME, 643 DATA_TYPE_NVLIST, 644 DATA_TYPE_NVLIST_ARRAY, 645 DATA_TYPE_BOOLEAN_VALUE, 646 DATA_TYPE_INT8, 647 DATA_TYPE_UINT8, 648 DATA_TYPE_BOOLEAN_ARRAY, 649 DATA_TYPE_INT8_ARRAY, 650 DATA_TYPE_UINT8_ARRAY 651 } data_type_t; 652 653 /* 654 * On-disk version number. 655 */ 656 #define SPA_VERSION_1 1ULL 657 #define SPA_VERSION_2 2ULL 658 #define SPA_VERSION_3 3ULL 659 #define SPA_VERSION_4 4ULL 660 #define SPA_VERSION_5 5ULL 661 #define SPA_VERSION_6 6ULL 662 #define SPA_VERSION_7 7ULL 663 #define SPA_VERSION_8 8ULL 664 #define SPA_VERSION_9 9ULL 665 #define SPA_VERSION_10 10ULL 666 #define SPA_VERSION_11 11ULL 667 #define SPA_VERSION_12 12ULL 668 #define SPA_VERSION_13 13ULL 669 #define SPA_VERSION_14 14ULL 670 #define SPA_VERSION_15 15ULL 671 #define SPA_VERSION_16 16ULL 672 #define SPA_VERSION_17 17ULL 673 #define SPA_VERSION_18 18ULL 674 #define SPA_VERSION_19 19ULL 675 #define SPA_VERSION_20 20ULL 676 #define SPA_VERSION_21 21ULL 677 #define SPA_VERSION_22 22ULL 678 #define SPA_VERSION_23 23ULL 679 #define SPA_VERSION_24 24ULL 680 #define SPA_VERSION_25 25ULL 681 #define SPA_VERSION_26 26ULL 682 #define SPA_VERSION_27 27ULL 683 #define SPA_VERSION_28 28ULL 684 #define SPA_VERSION_5000 5000ULL 685 686 /* 687 * When bumping up SPA_VERSION, make sure GRUB ZFS understands the on-disk 688 * format change. Go to usr/src/grub/grub-0.97/stage2/{zfs-include/, fsys_zfs*}, 689 * and do the appropriate changes. Also bump the version number in 690 * usr/src/grub/capability. 691 */ 692 #define SPA_VERSION SPA_VERSION_5000 693 #define SPA_VERSION_STRING "5000" 694 695 /* 696 * Symbolic names for the changes that caused a SPA_VERSION switch. 697 * Used in the code when checking for presence or absence of a feature. 698 * Feel free to define multiple symbolic names for each version if there 699 * were multiple changes to on-disk structures during that version. 700 * 701 * NOTE: When checking the current SPA_VERSION in your code, be sure 702 * to use spa_version() since it reports the version of the 703 * last synced uberblock. Checking the in-flight version can 704 * be dangerous in some cases. 705 */ 706 #define SPA_VERSION_INITIAL SPA_VERSION_1 707 #define SPA_VERSION_DITTO_BLOCKS SPA_VERSION_2 708 #define SPA_VERSION_SPARES SPA_VERSION_3 709 #define SPA_VERSION_RAID6 SPA_VERSION_3 710 #define SPA_VERSION_BPLIST_ACCOUNT SPA_VERSION_3 711 #define SPA_VERSION_RAIDZ_DEFLATE SPA_VERSION_3 712 #define SPA_VERSION_DNODE_BYTES SPA_VERSION_3 713 #define SPA_VERSION_ZPOOL_HISTORY SPA_VERSION_4 714 #define SPA_VERSION_GZIP_COMPRESSION SPA_VERSION_5 715 #define SPA_VERSION_BOOTFS SPA_VERSION_6 716 #define SPA_VERSION_SLOGS SPA_VERSION_7 717 #define SPA_VERSION_DELEGATED_PERMS SPA_VERSION_8 718 #define SPA_VERSION_FUID SPA_VERSION_9 719 #define SPA_VERSION_REFRESERVATION SPA_VERSION_9 720 #define SPA_VERSION_REFQUOTA SPA_VERSION_9 721 #define SPA_VERSION_UNIQUE_ACCURATE SPA_VERSION_9 722 #define SPA_VERSION_L2CACHE SPA_VERSION_10 723 #define SPA_VERSION_NEXT_CLONES SPA_VERSION_11 724 #define SPA_VERSION_ORIGIN SPA_VERSION_11 725 #define SPA_VERSION_DSL_SCRUB SPA_VERSION_11 726 #define SPA_VERSION_SNAP_PROPS SPA_VERSION_12 727 #define SPA_VERSION_USED_BREAKDOWN SPA_VERSION_13 728 #define SPA_VERSION_PASSTHROUGH_X SPA_VERSION_14 729 #define SPA_VERSION_USERSPACE SPA_VERSION_15 730 #define SPA_VERSION_STMF_PROP SPA_VERSION_16 731 #define SPA_VERSION_RAIDZ3 SPA_VERSION_17 732 #define SPA_VERSION_USERREFS SPA_VERSION_18 733 #define SPA_VERSION_HOLES SPA_VERSION_19 734 #define SPA_VERSION_ZLE_COMPRESSION SPA_VERSION_20 735 #define SPA_VERSION_DEDUP SPA_VERSION_21 736 #define SPA_VERSION_RECVD_PROPS SPA_VERSION_22 737 #define SPA_VERSION_SLIM_ZIL SPA_VERSION_23 738 #define SPA_VERSION_SA SPA_VERSION_24 739 #define SPA_VERSION_SCAN SPA_VERSION_25 740 #define SPA_VERSION_DIR_CLONES SPA_VERSION_26 741 #define SPA_VERSION_DEADLISTS SPA_VERSION_26 742 #define SPA_VERSION_FAST_SNAP SPA_VERSION_27 743 #define SPA_VERSION_MULTI_REPLACE SPA_VERSION_28 744 #define SPA_VERSION_BEFORE_FEATURES SPA_VERSION_28 745 #define SPA_VERSION_FEATURES SPA_VERSION_5000 746 747 #define SPA_VERSION_IS_SUPPORTED(v) \ 748 (((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \ 749 ((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION)) 750 751 /* 752 * The following are configuration names used in the nvlist describing a pool's 753 * configuration. 754 */ 755 #define ZPOOL_CONFIG_VERSION "version" 756 #define ZPOOL_CONFIG_POOL_NAME "name" 757 #define ZPOOL_CONFIG_POOL_STATE "state" 758 #define ZPOOL_CONFIG_POOL_TXG "txg" 759 #define ZPOOL_CONFIG_POOL_GUID "pool_guid" 760 #define ZPOOL_CONFIG_CREATE_TXG "create_txg" 761 #define ZPOOL_CONFIG_TOP_GUID "top_guid" 762 #define ZPOOL_CONFIG_VDEV_TREE "vdev_tree" 763 #define ZPOOL_CONFIG_TYPE "type" 764 #define ZPOOL_CONFIG_CHILDREN "children" 765 #define ZPOOL_CONFIG_ID "id" 766 #define ZPOOL_CONFIG_GUID "guid" 767 #define ZPOOL_CONFIG_INDIRECT_OBJECT "com.delphix:indirect_object" 768 #define ZPOOL_CONFIG_INDIRECT_BIRTHS "com.delphix:indirect_births" 769 #define ZPOOL_CONFIG_PREV_INDIRECT_VDEV "com.delphix:prev_indirect_vdev" 770 #define ZPOOL_CONFIG_PATH "path" 771 #define ZPOOL_CONFIG_DEVID "devid" 772 #define ZPOOL_CONFIG_PHYS_PATH "phys_path" 773 #define ZPOOL_CONFIG_METASLAB_ARRAY "metaslab_array" 774 #define ZPOOL_CONFIG_METASLAB_SHIFT "metaslab_shift" 775 #define ZPOOL_CONFIG_ASHIFT "ashift" 776 #define ZPOOL_CONFIG_ASIZE "asize" 777 #define ZPOOL_CONFIG_DTL "DTL" 778 #define ZPOOL_CONFIG_STATS "stats" 779 #define ZPOOL_CONFIG_WHOLE_DISK "whole_disk" 780 #define ZPOOL_CONFIG_ERRCOUNT "error_count" 781 #define ZPOOL_CONFIG_NOT_PRESENT "not_present" 782 #define ZPOOL_CONFIG_SPARES "spares" 783 #define ZPOOL_CONFIG_IS_SPARE "is_spare" 784 #define ZPOOL_CONFIG_NPARITY "nparity" 785 #define ZPOOL_CONFIG_HOSTID "hostid" 786 #define ZPOOL_CONFIG_HOSTNAME "hostname" 787 #define ZPOOL_CONFIG_IS_LOG "is_log" 788 #define ZPOOL_CONFIG_TIMESTAMP "timestamp" /* not stored on disk */ 789 #define ZPOOL_CONFIG_FEATURES_FOR_READ "features_for_read" 790 #define ZPOOL_CONFIG_VDEV_CHILDREN "vdev_children" 791 792 /* 793 * The persistent vdev state is stored as separate values rather than a single 794 * 'vdev_state' entry. This is because a device can be in multiple states, such 795 * as offline and degraded. 796 */ 797 #define ZPOOL_CONFIG_OFFLINE "offline" 798 #define ZPOOL_CONFIG_FAULTED "faulted" 799 #define ZPOOL_CONFIG_DEGRADED "degraded" 800 #define ZPOOL_CONFIG_REMOVED "removed" 801 #define ZPOOL_CONFIG_FRU "fru" 802 #define ZPOOL_CONFIG_AUX_STATE "aux_state" 803 804 #define VDEV_TYPE_ROOT "root" 805 #define VDEV_TYPE_MIRROR "mirror" 806 #define VDEV_TYPE_REPLACING "replacing" 807 #define VDEV_TYPE_RAIDZ "raidz" 808 #define VDEV_TYPE_DISK "disk" 809 #define VDEV_TYPE_FILE "file" 810 #define VDEV_TYPE_MISSING "missing" 811 #define VDEV_TYPE_HOLE "hole" 812 #define VDEV_TYPE_SPARE "spare" 813 #define VDEV_TYPE_LOG "log" 814 #define VDEV_TYPE_L2CACHE "l2cache" 815 #define VDEV_TYPE_INDIRECT "indirect" 816 817 /* 818 * This is needed in userland to report the minimum necessary device size. 819 */ 820 #define SPA_MINDEVSIZE (64ULL << 20) 821 822 /* 823 * The location of the pool configuration repository, shared between kernel and 824 * userland. 825 */ 826 #define ZPOOL_CACHE "/boot/zfs/zpool.cache" 827 828 /* 829 * vdev states are ordered from least to most healthy. 830 * A vdev that's CANT_OPEN or below is considered unusable. 831 */ 832 typedef enum vdev_state { 833 VDEV_STATE_UNKNOWN = 0, /* Uninitialized vdev */ 834 VDEV_STATE_CLOSED, /* Not currently open */ 835 VDEV_STATE_OFFLINE, /* Not allowed to open */ 836 VDEV_STATE_REMOVED, /* Explicitly removed from system */ 837 VDEV_STATE_CANT_OPEN, /* Tried to open, but failed */ 838 VDEV_STATE_FAULTED, /* External request to fault device */ 839 VDEV_STATE_DEGRADED, /* Replicated vdev with unhealthy kids */ 840 VDEV_STATE_HEALTHY /* Presumed good */ 841 } vdev_state_t; 842 843 /* 844 * vdev aux states. When a vdev is in the CANT_OPEN state, the aux field 845 * of the vdev stats structure uses these constants to distinguish why. 846 */ 847 typedef enum vdev_aux { 848 VDEV_AUX_NONE, /* no error */ 849 VDEV_AUX_OPEN_FAILED, /* ldi_open_*() or vn_open() failed */ 850 VDEV_AUX_CORRUPT_DATA, /* bad label or disk contents */ 851 VDEV_AUX_NO_REPLICAS, /* insufficient number of replicas */ 852 VDEV_AUX_BAD_GUID_SUM, /* vdev guid sum doesn't match */ 853 VDEV_AUX_TOO_SMALL, /* vdev size is too small */ 854 VDEV_AUX_BAD_LABEL, /* the label is OK but invalid */ 855 VDEV_AUX_VERSION_NEWER, /* on-disk version is too new */ 856 VDEV_AUX_VERSION_OLDER, /* on-disk version is too old */ 857 VDEV_AUX_SPARED /* hot spare used in another pool */ 858 } vdev_aux_t; 859 860 /* 861 * pool state. The following states are written to disk as part of the normal 862 * SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE. The remaining states are 863 * software abstractions used at various levels to communicate pool state. 864 */ 865 typedef enum pool_state { 866 POOL_STATE_ACTIVE = 0, /* In active use */ 867 POOL_STATE_EXPORTED, /* Explicitly exported */ 868 POOL_STATE_DESTROYED, /* Explicitly destroyed */ 869 POOL_STATE_SPARE, /* Reserved for hot spare use */ 870 POOL_STATE_UNINITIALIZED, /* Internal spa_t state */ 871 POOL_STATE_UNAVAIL, /* Internal libzfs state */ 872 POOL_STATE_POTENTIALLY_ACTIVE /* Internal libzfs state */ 873 } pool_state_t; 874 875 /* 876 * The uberblock version is incremented whenever an incompatible on-disk 877 * format change is made to the SPA, DMU, or ZAP. 878 * 879 * Note: the first two fields should never be moved. When a storage pool 880 * is opened, the uberblock must be read off the disk before the version 881 * can be checked. If the ub_version field is moved, we may not detect 882 * version mismatch. If the ub_magic field is moved, applications that 883 * expect the magic number in the first word won't work. 884 */ 885 #define UBERBLOCK_MAGIC 0x00bab10c /* oo-ba-bloc! */ 886 #define UBERBLOCK_SHIFT 10 /* up to 1K */ 887 #define MMP_MAGIC 0xa11cea11 /* all-see-all */ 888 889 #define MMP_INTERVAL_VALID_BIT 0x01 890 #define MMP_SEQ_VALID_BIT 0x02 891 #define MMP_FAIL_INT_VALID_BIT 0x04 892 893 #define MMP_VALID(ubp) (ubp->ub_magic == UBERBLOCK_MAGIC && \ 894 ubp->ub_mmp_magic == MMP_MAGIC) 895 #define MMP_INTERVAL_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \ 896 MMP_INTERVAL_VALID_BIT)) 897 #define MMP_SEQ_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \ 898 MMP_SEQ_VALID_BIT)) 899 #define MMP_FAIL_INT_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \ 900 MMP_FAIL_INT_VALID_BIT)) 901 902 #define MMP_INTERVAL(ubp) ((ubp->ub_mmp_config & 0x00000000FFFFFF00) \ 903 >> 8) 904 #define MMP_SEQ(ubp) ((ubp->ub_mmp_config & 0x0000FFFF00000000) \ 905 >> 32) 906 #define MMP_FAIL_INT(ubp) ((ubp->ub_mmp_config & 0xFFFF000000000000) \ 907 >> 48) 908 909 typedef struct uberblock { 910 uint64_t ub_magic; /* UBERBLOCK_MAGIC */ 911 uint64_t ub_version; /* SPA_VERSION */ 912 uint64_t ub_txg; /* txg of last sync */ 913 uint64_t ub_guid_sum; /* sum of all vdev guids */ 914 uint64_t ub_timestamp; /* UTC time of last sync */ 915 blkptr_t ub_rootbp; /* MOS objset_phys_t */ 916 /* highest SPA_VERSION supported by software that wrote this txg */ 917 uint64_t ub_software_version; 918 /* Maybe missing in uberblocks we read, but always written */ 919 uint64_t ub_mmp_magic; 920 /* 921 * If ub_mmp_delay == 0 and ub_mmp_magic is valid, MMP is off. 922 * Otherwise, nanosec since last MMP write. 923 */ 924 uint64_t ub_mmp_delay; 925 926 /* BEGIN CSTYLED */ 927 /* 928 * The ub_mmp_config contains the multihost write interval, multihost 929 * fail intervals, sequence number for sub-second granularity, and 930 * valid bit mask. This layout is as follows: 931 * 932 * 64 56 48 40 32 24 16 8 0 933 * +-------+-------+-------+-------+-------+-------+-------+-------+ 934 * 0 | Fail Intervals| Seq | Write Interval (ms) | VALID | 935 * +-------+-------+-------+-------+-------+-------+-------+-------+ 936 * 937 * This allows a write_interval of (2^24/1000)s, over 4.5 hours 938 * 939 * VALID Bits: 940 * - 0x01 - Write Interval (ms) 941 * - 0x02 - Sequence number exists 942 * - 0x04 - Fail Intervals 943 * - 0xf8 - Reserved 944 */ 945 /* END CSTYLED */ 946 uint64_t ub_mmp_config; 947 948 /* 949 * ub_checkpoint_txg indicates two things about the current uberblock: 950 * 951 * 1] If it is not zero then this uberblock is a checkpoint. If it is 952 * zero, then this uberblock is not a checkpoint. 953 * 954 * 2] On checkpointed uberblocks, the value of ub_checkpoint_txg is 955 * the ub_txg that the uberblock had at the time we moved it to 956 * the MOS config. 957 * 958 * The field is set when we checkpoint the uberblock and continues to 959 * hold that value even after we've rewound (unlike the ub_txg that 960 * is reset to a higher value). 961 * 962 * Besides checks used to determine whether we are reopening the 963 * pool from a checkpointed uberblock [see spa_ld_select_uberblock()], 964 * the value of the field is used to determine which ZIL blocks have 965 * been allocated according to the ms_sm when we are rewinding to a 966 * checkpoint. Specifically, if blk_birth > ub_checkpoint_txg, then 967 * the ZIL block is not allocated [see uses of spa_min_claim_txg()]. 968 */ 969 uint64_t ub_checkpoint_txg; 970 } uberblock_t; 971 972 /* 973 * Flags. 974 */ 975 #define DNODE_MUST_BE_ALLOCATED 1 976 #define DNODE_MUST_BE_FREE 2 977 978 /* 979 * Fixed constants. 980 */ 981 #define DNODE_SHIFT 9 /* 512 bytes */ 982 #define DN_MIN_INDBLKSHIFT 12 /* 4k */ 983 #define DN_MAX_INDBLKSHIFT 17 /* 128k */ 984 #define DNODE_BLOCK_SHIFT 14 /* 16k */ 985 #define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */ 986 #define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */ 987 #define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */ 988 989 /* 990 * Derived constants. 991 */ 992 #define DNODE_MIN_SIZE (1 << DNODE_SHIFT) 993 #define DNODE_MAX_SIZE (1 << DNODE_BLOCK_SHIFT) 994 #define DNODE_BLOCK_SIZE (1 << DNODE_BLOCK_SHIFT) 995 #define DNODE_MIN_SLOTS (DNODE_MIN_SIZE >> DNODE_SHIFT) 996 #define DNODE_MAX_SLOTS (DNODE_MAX_SIZE >> DNODE_SHIFT) 997 #define DN_BONUS_SIZE(dnsize) ((dnsize) - DNODE_CORE_SIZE - \ 998 (1 << SPA_BLKPTRSHIFT)) 999 #define DN_SLOTS_TO_BONUSLEN(slots) DN_BONUS_SIZE((slots) << DNODE_SHIFT) 1000 #define DN_OLD_MAX_BONUSLEN (DN_BONUS_SIZE(DNODE_MIN_SIZE)) 1001 #define DN_MAX_NBLKPTR ((DNODE_MIN_SIZE - DNODE_CORE_SIZE) >> \ 1002 SPA_BLKPTRSHIFT) 1003 #define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT) 1004 #define DN_ZERO_BONUSLEN (DN_BONUS_SIZE(DNODE_MAX_SIZE) + 1) 1005 1006 #define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT) 1007 #define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT) 1008 #define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT) 1009 1010 /* The +2 here is a cheesy way to round up */ 1011 #define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \ 1012 (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT))) 1013 1014 #define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \ 1015 (((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t)))) 1016 1017 #define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \ 1018 (dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT) 1019 1020 #define EPB(blkshift, typeshift) (1 << (blkshift - typeshift)) 1021 1022 /* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */ 1023 #define DNODE_FLAG_USED_BYTES (1<<0) 1024 #define DNODE_FLAG_USERUSED_ACCOUNTED (1<<1) 1025 1026 /* Does dnode have a SA spill blkptr in bonus? */ 1027 #define DNODE_FLAG_SPILL_BLKPTR (1<<2) 1028 1029 typedef struct dnode_phys { 1030 uint8_t dn_type; /* dmu_object_type_t */ 1031 uint8_t dn_indblkshift; /* ln2(indirect block size) */ 1032 uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */ 1033 uint8_t dn_nblkptr; /* length of dn_blkptr */ 1034 uint8_t dn_bonustype; /* type of data in bonus buffer */ 1035 uint8_t dn_checksum; /* ZIO_CHECKSUM type */ 1036 uint8_t dn_compress; /* ZIO_COMPRESS type */ 1037 uint8_t dn_flags; /* DNODE_FLAG_* */ 1038 uint16_t dn_datablkszsec; /* data block size in 512b sectors */ 1039 uint16_t dn_bonuslen; /* length of dn_bonus */ 1040 uint8_t dn_extra_slots; /* # of subsequent slots consumed */ 1041 uint8_t dn_pad2[3]; 1042 1043 /* accounting is protected by dn_dirty_mtx */ 1044 uint64_t dn_maxblkid; /* largest allocated block ID */ 1045 uint64_t dn_used; /* bytes (or sectors) of disk space */ 1046 1047 uint64_t dn_pad3[4]; 1048 1049 /* BEGIN CSTYLED */ 1050 /* 1051 * The tail region is 448 bytes for a 512 byte dnode, and 1052 * correspondingly larger for larger dnode sizes. The spill 1053 * block pointer, when present, is always at the end of the tail 1054 * region. There are three ways this space may be used, using 1055 * a 512 byte dnode for this diagram: 1056 * 1057 * 0 64 128 192 256 320 384 448 (offset) 1058 * +---------------+---------------+---------------+-------+ 1059 * | dn_blkptr[0] | dn_blkptr[1] | dn_blkptr[2] | / | 1060 * +---------------+---------------+---------------+-------+ 1061 * | dn_blkptr[0] | dn_bonus[0..319] | 1062 * +---------------+-----------------------+---------------+ 1063 * | dn_blkptr[0] | dn_bonus[0..191] | dn_spill | 1064 * +---------------+-----------------------+---------------+ 1065 */ 1066 /* END CSTYLED */ 1067 union { 1068 blkptr_t dn_blkptr[1+DN_OLD_MAX_BONUSLEN/sizeof (blkptr_t)]; 1069 struct { 1070 blkptr_t __dn_ignore1; 1071 uint8_t dn_bonus[DN_OLD_MAX_BONUSLEN]; 1072 }; 1073 struct { 1074 blkptr_t __dn_ignore2; 1075 uint8_t __dn_ignore3[DN_OLD_MAX_BONUSLEN - 1076 sizeof (blkptr_t)]; 1077 blkptr_t dn_spill; 1078 }; 1079 }; 1080 } dnode_phys_t; 1081 1082 #define DN_SPILL_BLKPTR(dnp) (blkptr_t *)((char *)(dnp) + \ 1083 (((dnp)->dn_extra_slots + 1) << DNODE_SHIFT) - (1 << SPA_BLKPTRSHIFT)) 1084 1085 typedef enum dmu_object_byteswap { 1086 DMU_BSWAP_UINT8, 1087 DMU_BSWAP_UINT16, 1088 DMU_BSWAP_UINT32, 1089 DMU_BSWAP_UINT64, 1090 DMU_BSWAP_ZAP, 1091 DMU_BSWAP_DNODE, 1092 DMU_BSWAP_OBJSET, 1093 DMU_BSWAP_ZNODE, 1094 DMU_BSWAP_OLDACL, 1095 DMU_BSWAP_ACL, 1096 /* 1097 * Allocating a new byteswap type number makes the on-disk format 1098 * incompatible with any other format that uses the same number. 1099 * 1100 * Data can usually be structured to work with one of the 1101 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. 1102 */ 1103 DMU_BSWAP_NUMFUNCS 1104 } dmu_object_byteswap_t; 1105 1106 #define DMU_OT_NEWTYPE 0x80 1107 #define DMU_OT_METADATA 0x40 1108 #define DMU_OT_BYTESWAP_MASK 0x3f 1109 1110 /* 1111 * Defines a uint8_t object type. Object types specify if the data 1112 * in the object is metadata (boolean) and how to byteswap the data 1113 * (dmu_object_byteswap_t). 1114 */ 1115 #define DMU_OT(byteswap, metadata) \ 1116 (DMU_OT_NEWTYPE | \ 1117 ((metadata) ? DMU_OT_METADATA : 0) | \ 1118 ((byteswap) & DMU_OT_BYTESWAP_MASK)) 1119 1120 typedef enum dmu_object_type { 1121 DMU_OT_NONE, 1122 /* general: */ 1123 DMU_OT_OBJECT_DIRECTORY, /* ZAP */ 1124 DMU_OT_OBJECT_ARRAY, /* UINT64 */ 1125 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ 1126 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ 1127 DMU_OT_BPLIST, /* UINT64 */ 1128 DMU_OT_BPLIST_HDR, /* UINT64 */ 1129 /* spa: */ 1130 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ 1131 DMU_OT_SPACE_MAP, /* UINT64 */ 1132 /* zil: */ 1133 DMU_OT_INTENT_LOG, /* UINT64 */ 1134 /* dmu: */ 1135 DMU_OT_DNODE, /* DNODE */ 1136 DMU_OT_OBJSET, /* OBJSET */ 1137 /* dsl: */ 1138 DMU_OT_DSL_DIR, /* UINT64 */ 1139 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ 1140 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ 1141 DMU_OT_DSL_PROPS, /* ZAP */ 1142 DMU_OT_DSL_DATASET, /* UINT64 */ 1143 /* zpl: */ 1144 DMU_OT_ZNODE, /* ZNODE */ 1145 DMU_OT_OLDACL, /* Old ACL */ 1146 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ 1147 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ 1148 DMU_OT_MASTER_NODE, /* ZAP */ 1149 DMU_OT_UNLINKED_SET, /* ZAP */ 1150 /* zvol: */ 1151 DMU_OT_ZVOL, /* UINT8 */ 1152 DMU_OT_ZVOL_PROP, /* ZAP */ 1153 /* other; for testing only! */ 1154 DMU_OT_PLAIN_OTHER, /* UINT8 */ 1155 DMU_OT_UINT64_OTHER, /* UINT64 */ 1156 DMU_OT_ZAP_OTHER, /* ZAP */ 1157 /* new object types: */ 1158 DMU_OT_ERROR_LOG, /* ZAP */ 1159 DMU_OT_SPA_HISTORY, /* UINT8 */ 1160 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ 1161 DMU_OT_POOL_PROPS, /* ZAP */ 1162 DMU_OT_DSL_PERMS, /* ZAP */ 1163 DMU_OT_ACL, /* ACL */ 1164 DMU_OT_SYSACL, /* SYSACL */ 1165 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ 1166 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ 1167 DMU_OT_NEXT_CLONES, /* ZAP */ 1168 DMU_OT_SCAN_QUEUE, /* ZAP */ 1169 DMU_OT_USERGROUP_USED, /* ZAP */ 1170 DMU_OT_USERGROUP_QUOTA, /* ZAP */ 1171 DMU_OT_USERREFS, /* ZAP */ 1172 DMU_OT_DDT_ZAP, /* ZAP */ 1173 DMU_OT_DDT_STATS, /* ZAP */ 1174 DMU_OT_SA, /* System attr */ 1175 DMU_OT_SA_MASTER_NODE, /* ZAP */ 1176 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ 1177 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ 1178 DMU_OT_SCAN_XLATE, /* ZAP */ 1179 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ 1180 DMU_OT_NUMTYPES, 1181 1182 /* 1183 * Names for valid types declared with DMU_OT(). 1184 */ 1185 DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE), 1186 DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE), 1187 DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE), 1188 DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE), 1189 DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE), 1190 DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE), 1191 DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE), 1192 DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE), 1193 DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE), 1194 DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE) 1195 } dmu_object_type_t; 1196 1197 typedef enum dmu_objset_type { 1198 DMU_OST_NONE, 1199 DMU_OST_META, 1200 DMU_OST_ZFS, 1201 DMU_OST_ZVOL, 1202 DMU_OST_OTHER, /* For testing only! */ 1203 DMU_OST_ANY, /* Be careful! */ 1204 DMU_OST_NUMTYPES 1205 } dmu_objset_type_t; 1206 1207 #define ZAP_MAXVALUELEN (1024 * 8) 1208 1209 /* 1210 * header for all bonus and spill buffers. 1211 * The header has a fixed portion with a variable number 1212 * of "lengths" depending on the number of variable sized 1213 * attribues which are determined by the "layout number" 1214 */ 1215 1216 #define SA_MAGIC 0x2F505A /* ZFS SA */ 1217 typedef struct sa_hdr_phys { 1218 uint32_t sa_magic; 1219 uint16_t sa_layout_info; /* Encoded with hdrsize and layout number */ 1220 uint16_t sa_lengths[1]; /* optional sizes for variable length attrs */ 1221 /* ... Data follows the lengths. */ 1222 } sa_hdr_phys_t; 1223 1224 /* 1225 * sa_hdr_phys -> sa_layout_info 1226 * 1227 * 16 10 0 1228 * +--------+-------+ 1229 * | hdrsz |layout | 1230 * +--------+-------+ 1231 * 1232 * Bits 0-10 are the layout number 1233 * Bits 11-16 are the size of the header. 1234 * The hdrsize is the number * 8 1235 * 1236 * For example. 1237 * hdrsz of 1 ==> 8 byte header 1238 * 2 ==> 16 byte header 1239 * 1240 */ 1241 1242 #define SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10) 1243 #define SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 16, 3, 0) 1244 #define SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \ 1245 { \ 1246 BF32_SET_SB(x, 10, 6, 3, 0, size); \ 1247 BF32_SET(x, 0, 10, num); \ 1248 } 1249 1250 #define SA_MODE_OFFSET 0 1251 #define SA_SIZE_OFFSET 8 1252 #define SA_GEN_OFFSET 16 1253 #define SA_UID_OFFSET 24 1254 #define SA_GID_OFFSET 32 1255 #define SA_PARENT_OFFSET 40 1256 #define SA_SYMLINK_OFFSET 160 1257 1258 #define ZIO_OBJSET_MAC_LEN 32 1259 1260 /* 1261 * Intent log header - this on disk structure holds fields to manage 1262 * the log. All fields are 64 bit to easily handle cross architectures. 1263 */ 1264 typedef struct zil_header { 1265 uint64_t zh_claim_txg; /* txg in which log blocks were claimed */ 1266 uint64_t zh_replay_seq; /* highest replayed sequence number */ 1267 blkptr_t zh_log; /* log chain */ 1268 uint64_t zh_claim_seq; /* highest claimed sequence number */ 1269 uint64_t zh_pad[5]; 1270 } zil_header_t; 1271 1272 #define OBJSET_PHYS_SIZE_V2 2048 1273 #define OBJSET_PHYS_SIZE_V3 4096 1274 1275 #define OBJSET_PHYS_PAD0_SIZE \ 1276 (OBJSET_PHYS_SIZE_V2 - sizeof (dnode_phys_t) * 3 - \ 1277 sizeof (zil_header_t) - sizeof (uint64_t) * 2 - \ 1278 2 * ZIO_OBJSET_MAC_LEN) 1279 #define OBJSET_PHYS_PAD1_SIZE \ 1280 (OBJSET_PHYS_SIZE_V3 - OBJSET_PHYS_SIZE_V2 - sizeof (dnode_phys_t)) 1281 1282 typedef struct objset_phys { 1283 dnode_phys_t os_meta_dnode; 1284 zil_header_t os_zil_header; 1285 uint64_t os_type; 1286 uint64_t os_flags; 1287 uint8_t os_portable_mac[ZIO_OBJSET_MAC_LEN]; 1288 uint8_t os_local_mac[ZIO_OBJSET_MAC_LEN]; 1289 char os_pad0[OBJSET_PHYS_PAD0_SIZE]; 1290 dnode_phys_t os_userused_dnode; 1291 dnode_phys_t os_groupused_dnode; 1292 dnode_phys_t os_projectused_dnode; 1293 char os_pad1[OBJSET_PHYS_PAD1_SIZE]; 1294 } objset_phys_t; 1295 1296 typedef struct dsl_dir_phys { 1297 uint64_t dd_creation_time; /* not actually used */ 1298 uint64_t dd_head_dataset_obj; 1299 uint64_t dd_parent_obj; 1300 uint64_t dd_clone_parent_obj; 1301 uint64_t dd_child_dir_zapobj; 1302 /* 1303 * how much space our children are accounting for; for leaf 1304 * datasets, == physical space used by fs + snaps 1305 */ 1306 uint64_t dd_used_bytes; 1307 uint64_t dd_compressed_bytes; 1308 uint64_t dd_uncompressed_bytes; 1309 /* Administrative quota setting */ 1310 uint64_t dd_quota; 1311 /* Administrative reservation setting */ 1312 uint64_t dd_reserved; 1313 uint64_t dd_props_zapobj; 1314 uint64_t dd_pad[21]; /* pad out to 256 bytes for good measure */ 1315 } dsl_dir_phys_t; 1316 1317 typedef struct dsl_dataset_phys { 1318 uint64_t ds_dir_obj; 1319 uint64_t ds_prev_snap_obj; 1320 uint64_t ds_prev_snap_txg; 1321 uint64_t ds_next_snap_obj; 1322 uint64_t ds_snapnames_zapobj; /* zap obj of snaps; ==0 for snaps */ 1323 uint64_t ds_num_children; /* clone/snap children; ==0 for head */ 1324 uint64_t ds_creation_time; /* seconds since 1970 */ 1325 uint64_t ds_creation_txg; 1326 uint64_t ds_deadlist_obj; 1327 uint64_t ds_used_bytes; 1328 uint64_t ds_compressed_bytes; 1329 uint64_t ds_uncompressed_bytes; 1330 uint64_t ds_unique_bytes; /* only relevant to snapshots */ 1331 /* 1332 * The ds_fsid_guid is a 56-bit ID that can change to avoid 1333 * collisions. The ds_guid is a 64-bit ID that will never 1334 * change, so there is a small probability that it will collide. 1335 */ 1336 uint64_t ds_fsid_guid; 1337 uint64_t ds_guid; 1338 uint64_t ds_flags; 1339 blkptr_t ds_bp; 1340 uint64_t ds_pad[8]; /* pad out to 320 bytes for good measure */ 1341 } dsl_dataset_phys_t; 1342 1343 /* 1344 * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 1345 */ 1346 #define DMU_POOL_DIRECTORY_OBJECT 1 1347 #define DMU_POOL_CONFIG "config" 1348 #define DMU_POOL_FEATURES_FOR_READ "features_for_read" 1349 #define DMU_POOL_ROOT_DATASET "root_dataset" 1350 #define DMU_POOL_SYNC_BPLIST "sync_bplist" 1351 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 1352 #define DMU_POOL_ERRLOG_LAST "errlog_last" 1353 #define DMU_POOL_SPARES "spares" 1354 #define DMU_POOL_DEFLATE "deflate" 1355 #define DMU_POOL_HISTORY "history" 1356 #define DMU_POOL_PROPS "pool_props" 1357 #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" 1358 #define DMU_POOL_REMOVING "com.delphix:removing" 1359 #define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj" 1360 #define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect" 1361 1362 #define ZAP_MAGIC 0x2F52AB2ABULL 1363 1364 #define FZAP_BLOCK_SHIFT(zap) ((zap)->zap_block_shift) 1365 1366 #define ZAP_MAXCD (uint32_t)(-1) 1367 #define ZAP_HASHBITS 28 1368 #define MZAP_ENT_LEN 64 1369 #define MZAP_NAME_LEN (MZAP_ENT_LEN - 8 - 4 - 2) 1370 #define MZAP_MAX_BLKSZ SPA_OLD_MAXBLOCKSIZE 1371 1372 typedef struct mzap_ent_phys { 1373 uint64_t mze_value; 1374 uint32_t mze_cd; 1375 uint16_t mze_pad; /* in case we want to chain them someday */ 1376 char mze_name[MZAP_NAME_LEN]; 1377 } mzap_ent_phys_t; 1378 1379 typedef struct mzap_phys { 1380 uint64_t mz_block_type; /* ZBT_MICRO */ 1381 uint64_t mz_salt; 1382 uint64_t mz_normflags; 1383 uint64_t mz_pad[5]; 1384 mzap_ent_phys_t mz_chunk[1]; 1385 /* actually variable size depending on block size */ 1386 } mzap_phys_t; 1387 1388 /* 1389 * The (fat) zap is stored in one object. It is an array of 1390 * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of: 1391 * 1392 * ptrtbl fits in first block: 1393 * [zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ... 1394 * 1395 * ptrtbl too big for first block: 1396 * [zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ... 1397 * 1398 */ 1399 1400 #define ZBT_LEAF ((1ULL << 63) + 0) 1401 #define ZBT_HEADER ((1ULL << 63) + 1) 1402 #define ZBT_MICRO ((1ULL << 63) + 3) 1403 /* any other values are ptrtbl blocks */ 1404 1405 /* 1406 * the embedded pointer table takes up half a block: 1407 * block size / entry size (2^3) / 2 1408 */ 1409 #define ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1) 1410 1411 /* 1412 * The embedded pointer table starts half-way through the block. Since 1413 * the pointer table itself is half the block, it starts at (64-bit) 1414 * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)). 1415 */ 1416 #define ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \ 1417 ((uint64_t *)(zap)->zap_phys) \ 1418 [(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))] 1419 1420 /* 1421 * TAKE NOTE: 1422 * If zap_phys_t is modified, zap_byteswap() must be modified. 1423 */ 1424 typedef struct zap_phys { 1425 uint64_t zap_block_type; /* ZBT_HEADER */ 1426 uint64_t zap_magic; /* ZAP_MAGIC */ 1427 1428 struct zap_table_phys { 1429 uint64_t zt_blk; /* starting block number */ 1430 uint64_t zt_numblks; /* number of blocks */ 1431 uint64_t zt_shift; /* bits to index it */ 1432 uint64_t zt_nextblk; /* next (larger) copy start block */ 1433 uint64_t zt_blks_copied; /* number source blocks copied */ 1434 } zap_ptrtbl; 1435 1436 uint64_t zap_freeblk; /* the next free block */ 1437 uint64_t zap_num_leafs; /* number of leafs */ 1438 uint64_t zap_num_entries; /* number of entries */ 1439 uint64_t zap_salt; /* salt to stir into hash function */ 1440 uint64_t zap_normflags; /* flags for u8_textprep_str() */ 1441 uint64_t zap_flags; /* zap_flags_t */ 1442 /* 1443 * This structure is followed by padding, and then the embedded 1444 * pointer table. The embedded pointer table takes up second 1445 * half of the block. It is accessed using the 1446 * ZAP_EMBEDDED_PTRTBL_ENT() macro. 1447 */ 1448 } zap_phys_t; 1449 1450 typedef struct zap_table_phys zap_table_phys_t; 1451 1452 struct spa; 1453 typedef struct fat_zap { 1454 int zap_block_shift; /* block size shift */ 1455 zap_phys_t *zap_phys; 1456 const struct spa *zap_spa; 1457 const dnode_phys_t *zap_dnode; 1458 } fat_zap_t; 1459 1460 #define ZAP_LEAF_MAGIC 0x2AB1EAF 1461 1462 /* chunk size = 24 bytes */ 1463 #define ZAP_LEAF_CHUNKSIZE 24 1464 1465 /* 1466 * The amount of space available for chunks is: 1467 * block size (1<<l->l_bs) - hash entry size (2) * number of hash 1468 * entries - header space (2*chunksize) 1469 */ 1470 #define ZAP_LEAF_NUMCHUNKS(l) \ 1471 (((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \ 1472 ZAP_LEAF_CHUNKSIZE - 2) 1473 1474 /* 1475 * The amount of space within the chunk available for the array is: 1476 * chunk size - space for type (1) - space for next pointer (2) 1477 */ 1478 #define ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3) 1479 1480 #define ZAP_LEAF_ARRAY_NCHUNKS(bytes) \ 1481 (((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES) 1482 1483 /* 1484 * Low water mark: when there are only this many chunks free, start 1485 * growing the ptrtbl. Ideally, this should be larger than a 1486 * "reasonably-sized" entry. 20 chunks is more than enough for the 1487 * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value), 1488 * while still being only around 3% for 16k blocks. 1489 */ 1490 #define ZAP_LEAF_LOW_WATER (20) 1491 1492 /* 1493 * The leaf hash table has block size / 2^5 (32) number of entries, 1494 * which should be more than enough for the maximum number of entries, 1495 * which is less than block size / CHUNKSIZE (24) / minimum number of 1496 * chunks per entry (3). 1497 */ 1498 #define ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5) 1499 #define ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l)) 1500 1501 /* 1502 * The chunks start immediately after the hash table. The end of the 1503 * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a 1504 * chunk_t. 1505 */ 1506 #define ZAP_LEAF_CHUNK(l, idx) \ 1507 ((zap_leaf_chunk_t *) \ 1508 ((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx] 1509 #define ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry) 1510 1511 typedef enum zap_chunk_type { 1512 ZAP_CHUNK_FREE = 253, 1513 ZAP_CHUNK_ENTRY = 252, 1514 ZAP_CHUNK_ARRAY = 251, 1515 ZAP_CHUNK_TYPE_MAX = 250 1516 } zap_chunk_type_t; 1517 1518 /* 1519 * TAKE NOTE: 1520 * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified. 1521 */ 1522 typedef struct zap_leaf_phys { 1523 struct zap_leaf_header { 1524 uint64_t lh_block_type; /* ZBT_LEAF */ 1525 uint64_t lh_pad1; 1526 uint64_t lh_prefix; /* hash prefix of this leaf */ 1527 uint32_t lh_magic; /* ZAP_LEAF_MAGIC */ 1528 uint16_t lh_nfree; /* number free chunks */ 1529 uint16_t lh_nentries; /* number of entries */ 1530 uint16_t lh_prefix_len; /* num bits used to id this */ 1531 1532 /* above is accessable to zap, below is zap_leaf private */ 1533 1534 uint16_t lh_freelist; /* chunk head of free list */ 1535 uint8_t lh_pad2[12]; 1536 } l_hdr; /* 2 24-byte chunks */ 1537 1538 /* 1539 * The header is followed by a hash table with 1540 * ZAP_LEAF_HASH_NUMENTRIES(zap) entries. The hash table is 1541 * followed by an array of ZAP_LEAF_NUMCHUNKS(zap) 1542 * zap_leaf_chunk structures. These structures are accessed 1543 * with the ZAP_LEAF_CHUNK() macro. 1544 */ 1545 1546 uint16_t l_hash[1]; 1547 } zap_leaf_phys_t; 1548 1549 typedef union zap_leaf_chunk { 1550 struct zap_leaf_entry { 1551 uint8_t le_type; /* always ZAP_CHUNK_ENTRY */ 1552 uint8_t le_value_intlen; /* size of ints */ 1553 uint16_t le_next; /* next entry in hash chain */ 1554 uint16_t le_name_chunk; /* first chunk of the name */ 1555 uint16_t le_name_numints; /* bytes in name, incl null */ 1556 uint16_t le_value_chunk; /* first chunk of the value */ 1557 uint16_t le_value_numints; /* value length in ints */ 1558 uint32_t le_cd; /* collision differentiator */ 1559 uint64_t le_hash; /* hash value of the name */ 1560 } l_entry; 1561 struct zap_leaf_array { 1562 uint8_t la_type; /* always ZAP_CHUNK_ARRAY */ 1563 uint8_t la_array[ZAP_LEAF_ARRAY_BYTES]; 1564 uint16_t la_next; /* next blk or CHAIN_END */ 1565 } l_array; 1566 struct zap_leaf_free { 1567 uint8_t lf_type; /* always ZAP_CHUNK_FREE */ 1568 uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES]; 1569 uint16_t lf_next; /* next in free list, or CHAIN_END */ 1570 } l_free; 1571 } zap_leaf_chunk_t; 1572 1573 typedef struct zap_leaf { 1574 int l_bs; /* block size shift */ 1575 zap_leaf_phys_t *l_phys; 1576 } zap_leaf_t; 1577 1578 /* 1579 * Define special zfs pflags 1580 */ 1581 #define ZFS_XATTR 0x1 /* is an extended attribute */ 1582 #define ZFS_INHERIT_ACE 0x2 /* ace has inheritable ACEs */ 1583 #define ZFS_ACL_TRIVIAL 0x4 /* files ACL is trivial */ 1584 1585 #define MASTER_NODE_OBJ 1 1586 1587 /* 1588 * special attributes for master node. 1589 */ 1590 1591 #define ZFS_FSID "FSID" 1592 #define ZFS_UNLINKED_SET "DELETE_QUEUE" 1593 #define ZFS_ROOT_OBJ "ROOT" 1594 #define ZPL_VERSION_OBJ "VERSION" 1595 #define ZFS_PROP_BLOCKPERPAGE "BLOCKPERPAGE" 1596 #define ZFS_PROP_NOGROWBLOCKS "NOGROWBLOCKS" 1597 1598 #define ZFS_FLAG_BLOCKPERPAGE 0x1 1599 #define ZFS_FLAG_NOGROWBLOCKS 0x2 1600 1601 /* 1602 * ZPL version - rev'd whenever an incompatible on-disk format change 1603 * occurs. Independent of SPA/DMU/ZAP versioning. 1604 */ 1605 1606 #define ZPL_VERSION 1ULL 1607 1608 /* 1609 * The directory entry has the type (currently unused on Solaris) in the 1610 * top 4 bits, and the object number in the low 48 bits. The "middle" 1611 * 12 bits are unused. 1612 */ 1613 #define ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4) 1614 #define ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48) 1615 #define ZFS_DIRENT_MAKE(type, obj) (((uint64_t)type << 60) | obj) 1616 1617 typedef struct ace { 1618 uid_t a_who; /* uid or gid */ 1619 uint32_t a_access_mask; /* read,write,... */ 1620 uint16_t a_flags; /* see below */ 1621 uint16_t a_type; /* allow or deny */ 1622 } ace_t; 1623 1624 #define ACE_SLOT_CNT 6 1625 1626 typedef struct zfs_znode_acl { 1627 uint64_t z_acl_extern_obj; /* ext acl pieces */ 1628 uint32_t z_acl_count; /* Number of ACEs */ 1629 uint16_t z_acl_version; /* acl version */ 1630 uint16_t z_acl_pad; /* pad */ 1631 ace_t z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */ 1632 } zfs_znode_acl_t; 1633 1634 /* 1635 * This is the persistent portion of the znode. It is stored 1636 * in the "bonus buffer" of the file. Short symbolic links 1637 * are also stored in the bonus buffer. 1638 */ 1639 typedef struct znode_phys { 1640 uint64_t zp_atime[2]; /* 0 - last file access time */ 1641 uint64_t zp_mtime[2]; /* 16 - last file modification time */ 1642 uint64_t zp_ctime[2]; /* 32 - last file change time */ 1643 uint64_t zp_crtime[2]; /* 48 - creation time */ 1644 uint64_t zp_gen; /* 64 - generation (txg of creation) */ 1645 uint64_t zp_mode; /* 72 - file mode bits */ 1646 uint64_t zp_size; /* 80 - size of file */ 1647 uint64_t zp_parent; /* 88 - directory parent (`..') */ 1648 uint64_t zp_links; /* 96 - number of links to file */ 1649 uint64_t zp_xattr; /* 104 - DMU object for xattrs */ 1650 uint64_t zp_rdev; /* 112 - dev_t for VBLK & VCHR files */ 1651 uint64_t zp_flags; /* 120 - persistent flags */ 1652 uint64_t zp_uid; /* 128 - file owner */ 1653 uint64_t zp_gid; /* 136 - owning group */ 1654 uint64_t zp_pad[4]; /* 144 - future */ 1655 zfs_znode_acl_t zp_acl; /* 176 - 263 ACL */ 1656 /* 1657 * Data may pad out any remaining bytes in the znode buffer, eg: 1658 * 1659 * |<---------------------- dnode_phys (512) ------------------------>| 1660 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->| 1661 * |<---- znode (264) ---->|<---- data (56) ---->| 1662 * 1663 * At present, we only use this space to store symbolic links. 1664 */ 1665 } znode_phys_t; 1666 1667 /* 1668 * In-core vdev representation. 1669 */ 1670 struct vdev; 1671 struct spa; 1672 typedef int vdev_phys_read_t(struct vdev *, void *, off_t, void *, size_t); 1673 typedef int vdev_phys_write_t(struct vdev *, off_t, void *, size_t); 1674 typedef int vdev_read_t(struct vdev *, const blkptr_t *, void *, off_t, size_t); 1675 1676 typedef STAILQ_HEAD(vdev_list, vdev) vdev_list_t; 1677 1678 typedef struct vdev_indirect_mapping_entry_phys { 1679 /* 1680 * Decode with DVA_MAPPING_* macros. 1681 * Contains: 1682 * the source offset (low 63 bits) 1683 * the one-bit "mark", used for garbage collection (by zdb) 1684 */ 1685 uint64_t vimep_src; 1686 1687 /* 1688 * Note: the DVA's asize is 24 bits, and can thus store ranges 1689 * up to 8GB. 1690 */ 1691 dva_t vimep_dst; 1692 } vdev_indirect_mapping_entry_phys_t; 1693 1694 #define DVA_MAPPING_GET_SRC_OFFSET(vimep) \ 1695 BF64_GET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0) 1696 #define DVA_MAPPING_SET_SRC_OFFSET(vimep, x) \ 1697 BF64_SET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0, x) 1698 1699 typedef struct vdev_indirect_mapping_entry { 1700 vdev_indirect_mapping_entry_phys_t vime_mapping; 1701 uint32_t vime_obsolete_count; 1702 list_node_t vime_node; 1703 } vdev_indirect_mapping_entry_t; 1704 1705 /* 1706 * This is stored in the bonus buffer of the mapping object, see comment of 1707 * vdev_indirect_config for more details. 1708 */ 1709 typedef struct vdev_indirect_mapping_phys { 1710 uint64_t vimp_max_offset; 1711 uint64_t vimp_bytes_mapped; 1712 uint64_t vimp_num_entries; /* number of v_i_m_entry_phys_t's */ 1713 1714 /* 1715 * For each entry in the mapping object, this object contains an 1716 * entry representing the number of bytes of that mapping entry 1717 * that were no longer in use by the pool at the time this indirect 1718 * vdev was last condensed. 1719 */ 1720 uint64_t vimp_counts_object; 1721 } vdev_indirect_mapping_phys_t; 1722 1723 #define VDEV_INDIRECT_MAPPING_SIZE_V0 (3 * sizeof (uint64_t)) 1724 1725 typedef struct vdev_indirect_mapping { 1726 uint64_t vim_object; 1727 boolean_t vim_havecounts; 1728 1729 /* vim_entries segment offset currently in memory. */ 1730 uint64_t vim_entry_offset; 1731 /* vim_entries segment size. */ 1732 size_t vim_num_entries; 1733 1734 /* Needed by dnode_read() */ 1735 const void *vim_spa; 1736 dnode_phys_t *vim_dn; 1737 1738 /* 1739 * An ordered array of mapping entries, sorted by source offset. 1740 * Note that vim_entries is needed during a removal (and contains 1741 * mappings that have been synced to disk so far) to handle frees 1742 * from the removing device. 1743 */ 1744 vdev_indirect_mapping_entry_phys_t *vim_entries; 1745 objset_phys_t *vim_objset; 1746 vdev_indirect_mapping_phys_t *vim_phys; 1747 } vdev_indirect_mapping_t; 1748 1749 /* 1750 * On-disk indirect vdev state. 1751 * 1752 * An indirect vdev is described exclusively in the MOS config of a pool. 1753 * The config for an indirect vdev includes several fields, which are 1754 * accessed in memory by a vdev_indirect_config_t. 1755 */ 1756 typedef struct vdev_indirect_config { 1757 /* 1758 * Object (in MOS) which contains the indirect mapping. This object 1759 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by 1760 * vimep_src. The bonus buffer for this object is a 1761 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev 1762 * removal is initiated. 1763 * 1764 * Note that this object can be empty if none of the data on the vdev 1765 * has been copied yet. 1766 */ 1767 uint64_t vic_mapping_object; 1768 1769 /* 1770 * Object (in MOS) which contains the birth times for the mapping 1771 * entries. This object contains an array of 1772 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus 1773 * buffer for this object is a vdev_indirect_birth_phys_t. This object 1774 * is allocated when a vdev removal is initiated. 1775 * 1776 * Note that this object can be empty if none of the vdev has yet been 1777 * copied. 1778 */ 1779 uint64_t vic_births_object; 1780 1781 /* 1782 * This is the vdev ID which was removed previous to this vdev, or 1783 * UINT64_MAX if there are no previously removed vdevs. 1784 */ 1785 uint64_t vic_prev_indirect_vdev; 1786 } vdev_indirect_config_t; 1787 1788 typedef struct vdev { 1789 STAILQ_ENTRY(vdev) v_childlink; /* link in parent's child list */ 1790 STAILQ_ENTRY(vdev) v_alllink; /* link in global vdev list */ 1791 vdev_list_t v_children; /* children of this vdev */ 1792 const char *v_name; /* vdev name */ 1793 const char *v_phys_path; /* vdev bootpath */ 1794 const char *v_devid; /* vdev devid */ 1795 uint64_t v_guid; /* vdev guid */ 1796 uint64_t v_id; /* index in parent */ 1797 uint64_t v_psize; /* physical device capacity */ 1798 int v_ashift; /* offset to block shift */ 1799 int v_nparity; /* # parity for raidz */ 1800 struct vdev *v_top; /* parent vdev */ 1801 size_t v_nchildren; /* # children */ 1802 vdev_state_t v_state; /* current state */ 1803 vdev_phys_read_t *v_phys_read; /* read from raw leaf vdev */ 1804 vdev_phys_write_t *v_phys_write; /* write to raw leaf vdev */ 1805 vdev_read_t *v_read; /* read from vdev */ 1806 void *v_priv; /* data for read/write function */ 1807 boolean_t v_islog; 1808 struct spa *v_spa; /* link to spa */ 1809 /* 1810 * Values stored in the config for an indirect or removing vdev. 1811 */ 1812 vdev_indirect_config_t vdev_indirect_config; 1813 vdev_indirect_mapping_t *v_mapping; 1814 } vdev_t; 1815 1816 /* 1817 * In-core pool representation. 1818 */ 1819 typedef STAILQ_HEAD(spa_list, spa) spa_list_t; 1820 1821 typedef struct spa { 1822 STAILQ_ENTRY(spa) spa_link; /* link in global pool list */ 1823 char *spa_name; /* pool name */ 1824 uint64_t spa_guid; /* pool guid */ 1825 uint64_t spa_txg; /* most recent transaction */ 1826 struct uberblock spa_uberblock; /* best uberblock so far */ 1827 vdev_t *spa_root_vdev; /* toplevel vdev container */ 1828 objset_phys_t spa_mos; /* MOS for this pool */ 1829 zio_cksum_salt_t spa_cksum_salt; /* secret salt for cksum */ 1830 void *spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS]; 1831 vdev_t *spa_boot_vdev; /* boot device for kernel */ 1832 boolean_t spa_with_log; /* this pool has log */ 1833 void *spa_bootenv; /* bootenv from pool label */ 1834 } spa_t; 1835 1836 /* IO related arguments. */ 1837 typedef struct zio { 1838 spa_t *io_spa; 1839 blkptr_t *io_bp; 1840 void *io_data; 1841 uint64_t io_size; 1842 uint64_t io_offset; 1843 1844 /* Stuff for the vdev stack */ 1845 vdev_t *io_vd; 1846 void *io_vsd; 1847 1848 int io_error; 1849 } zio_t; 1850 1851 extern void decode_embedded_bp_compressed(const blkptr_t *, void *); 1852 1853 #endif /* _ZFSIMPL_H_ */ 1854