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 #define MAXNAMELEN 256 63 64 #define _NOTE(s) 65 66 /* CRC64 table */ 67 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ 68 69 /* 70 * Macros for various sorts of alignment and rounding when the alignment 71 * is known to be a power of 2. 72 */ 73 #define P2ALIGN(x, align) ((x) & -(align)) 74 #define P2PHASE(x, align) ((x) & ((align) - 1)) 75 #define P2NPHASE(x, align) (-(x) & ((align) - 1)) 76 #define P2ROUNDUP(x, align) (-(-(x) & -(align))) 77 #define P2END(x, align) (-(~(x) & -(align))) 78 #define P2PHASEUP(x, align, phase) ((phase) - (((phase) - (x)) & -(align))) 79 #define P2BOUNDARY(off, len, align) (((off) ^ ((off) + (len) - 1)) > (align) - 1) 80 81 /* 82 * General-purpose 32-bit and 64-bit bitfield encodings. 83 */ 84 #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) 85 #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) 86 #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) 87 #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) 88 89 #define BF32_GET(x, low, len) BF32_DECODE(x, low, len) 90 #define BF64_GET(x, low, len) BF64_DECODE(x, low, len) 91 92 #define BF32_SET(x, low, len, val) \ 93 ((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len)) 94 #define BF64_SET(x, low, len, val) \ 95 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)) 96 97 #define BF32_GET_SB(x, low, len, shift, bias) \ 98 ((BF32_GET(x, low, len) + (bias)) << (shift)) 99 #define BF64_GET_SB(x, low, len, shift, bias) \ 100 ((BF64_GET(x, low, len) + (bias)) << (shift)) 101 102 #define BF32_SET_SB(x, low, len, shift, bias, val) \ 103 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)) 104 #define BF64_SET_SB(x, low, len, shift, bias, val) \ 105 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)) 106 107 /* 108 * Macros to reverse byte order 109 */ 110 #define BSWAP_8(x) ((x) & 0xff) 111 #define BSWAP_16(x) ((BSWAP_8(x) << 8) | BSWAP_8((x) >> 8)) 112 #define BSWAP_32(x) ((BSWAP_16(x) << 16) | BSWAP_16((x) >> 16)) 113 #define BSWAP_64(x) ((BSWAP_32(x) << 32) | BSWAP_32((x) >> 32)) 114 115 #define SPA_MINBLOCKSHIFT 9 116 #define SPA_OLDMAXBLOCKSHIFT 17 117 #define SPA_MAXBLOCKSHIFT 24 118 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 119 #define SPA_OLDMAXBLOCKSIZE (1ULL << SPA_OLDMAXBLOCKSHIFT) 120 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 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 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept 150 * secret and is suitable for use in MAC algorithms as the key. 151 */ 152 typedef struct zio_cksum_salt { 153 uint8_t zcs_bytes[32]; 154 } zio_cksum_salt_t; 155 156 /* 157 * Each block is described by its DVAs, time of birth, checksum, etc. 158 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 159 * 160 * 64 56 48 40 32 24 16 8 0 161 * +-------+-------+-------+-------+-------+-------+-------+-------+ 162 * 0 | vdev1 | GRID | ASIZE | 163 * +-------+-------+-------+-------+-------+-------+-------+-------+ 164 * 1 |G| offset1 | 165 * +-------+-------+-------+-------+-------+-------+-------+-------+ 166 * 2 | vdev2 | GRID | ASIZE | 167 * +-------+-------+-------+-------+-------+-------+-------+-------+ 168 * 3 |G| offset2 | 169 * +-------+-------+-------+-------+-------+-------+-------+-------+ 170 * 4 | vdev3 | GRID | ASIZE | 171 * +-------+-------+-------+-------+-------+-------+-------+-------+ 172 * 5 |G| offset3 | 173 * +-------+-------+-------+-------+-------+-------+-------+-------+ 174 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 175 * +-------+-------+-------+-------+-------+-------+-------+-------+ 176 * 7 | padding | 177 * +-------+-------+-------+-------+-------+-------+-------+-------+ 178 * 8 | padding | 179 * +-------+-------+-------+-------+-------+-------+-------+-------+ 180 * 9 | physical birth txg | 181 * +-------+-------+-------+-------+-------+-------+-------+-------+ 182 * a | logical birth txg | 183 * +-------+-------+-------+-------+-------+-------+-------+-------+ 184 * b | fill count | 185 * +-------+-------+-------+-------+-------+-------+-------+-------+ 186 * c | checksum[0] | 187 * +-------+-------+-------+-------+-------+-------+-------+-------+ 188 * d | checksum[1] | 189 * +-------+-------+-------+-------+-------+-------+-------+-------+ 190 * e | checksum[2] | 191 * +-------+-------+-------+-------+-------+-------+-------+-------+ 192 * f | checksum[3] | 193 * +-------+-------+-------+-------+-------+-------+-------+-------+ 194 * 195 * Legend: 196 * 197 * vdev virtual device ID 198 * offset offset into virtual device 199 * LSIZE logical size 200 * PSIZE physical size (after compression) 201 * ASIZE allocated size (including RAID-Z parity and gang block headers) 202 * GRID RAID-Z layout information (reserved for future use) 203 * cksum checksum function 204 * comp compression function 205 * G gang block indicator 206 * B byteorder (endianness) 207 * D dedup 208 * X encryption (on version 30, which is not supported) 209 * E blkptr_t contains embedded data (see below) 210 * lvl level of indirection 211 * type DMU object type 212 * phys birth txg of block allocation; zero if same as logical birth txg 213 * log. birth transaction group in which the block was logically born 214 * fill count number of non-zero blocks under this bp 215 * checksum[4] 256-bit checksum of the data this bp describes 216 */ 217 218 /* 219 * "Embedded" blkptr_t's don't actually point to a block, instead they 220 * have a data payload embedded in the blkptr_t itself. See the comment 221 * in blkptr.c for more details. 222 * 223 * The blkptr_t is laid out as follows: 224 * 225 * 64 56 48 40 32 24 16 8 0 226 * +-------+-------+-------+-------+-------+-------+-------+-------+ 227 * 0 | payload | 228 * 1 | payload | 229 * 2 | payload | 230 * 3 | payload | 231 * 4 | payload | 232 * 5 | payload | 233 * +-------+-------+-------+-------+-------+-------+-------+-------+ 234 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 235 * +-------+-------+-------+-------+-------+-------+-------+-------+ 236 * 7 | payload | 237 * 8 | payload | 238 * 9 | payload | 239 * +-------+-------+-------+-------+-------+-------+-------+-------+ 240 * a | logical birth txg | 241 * +-------+-------+-------+-------+-------+-------+-------+-------+ 242 * b | payload | 243 * c | payload | 244 * d | payload | 245 * e | payload | 246 * f | payload | 247 * +-------+-------+-------+-------+-------+-------+-------+-------+ 248 * 249 * Legend: 250 * 251 * payload contains the embedded data 252 * B (byteorder) byteorder (endianness) 253 * D (dedup) padding (set to zero) 254 * X encryption (set to zero; see above) 255 * E (embedded) set to one 256 * lvl indirection level 257 * type DMU object type 258 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 259 * comp compression function of payload 260 * PSIZE size of payload after compression, in bytes 261 * LSIZE logical size of payload, in bytes 262 * note that 25 bits is enough to store the largest 263 * "normal" BP's LSIZE (2^16 * 2^9) in bytes 264 * log. birth transaction group in which the block was logically born 265 * 266 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 267 * bp's they are stored in units of SPA_MINBLOCKSHIFT. 268 * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 269 * The B, D, X, lvl, type, and comp fields are stored the same as with normal 270 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 271 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 272 * other macros, as they assert that they are only used on BP's of the correct 273 * "embedded-ness". 274 */ 275 276 #define BPE_GET_ETYPE(bp) \ 277 (ASSERT(BP_IS_EMBEDDED(bp)), \ 278 BF64_GET((bp)->blk_prop, 40, 8)) 279 #define BPE_SET_ETYPE(bp, t) do { \ 280 ASSERT(BP_IS_EMBEDDED(bp)); \ 281 BF64_SET((bp)->blk_prop, 40, 8, t); \ 282 _NOTE(CONSTCOND) } while (0) 283 284 #define BPE_GET_LSIZE(bp) \ 285 (ASSERT(BP_IS_EMBEDDED(bp)), \ 286 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 287 #define BPE_SET_LSIZE(bp, x) do { \ 288 ASSERT(BP_IS_EMBEDDED(bp)); \ 289 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 290 _NOTE(CONSTCOND) } while (0) 291 292 #define BPE_GET_PSIZE(bp) \ 293 (ASSERT(BP_IS_EMBEDDED(bp)), \ 294 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 295 #define BPE_SET_PSIZE(bp, x) do { \ 296 ASSERT(BP_IS_EMBEDDED(bp)); \ 297 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 298 _NOTE(CONSTCOND) } while (0) 299 300 typedef enum bp_embedded_type { 301 BP_EMBEDDED_TYPE_DATA, 302 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ 303 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED 304 } bp_embedded_type_t; 305 306 #define BPE_NUM_WORDS 14 307 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 308 #define BPE_IS_PAYLOADWORD(bp, wp) \ 309 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 310 311 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 312 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 313 314 typedef struct blkptr { 315 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 316 uint64_t blk_prop; /* size, compression, type, etc */ 317 uint64_t blk_pad[2]; /* Extra space for the future */ 318 uint64_t blk_phys_birth; /* txg when block was allocated */ 319 uint64_t blk_birth; /* transaction group at birth */ 320 uint64_t blk_fill; /* fill count */ 321 zio_cksum_t blk_cksum; /* 256-bit checksum */ 322 } blkptr_t; 323 324 /* 325 * Macros to get and set fields in a bp or DVA. 326 */ 327 #define DVA_GET_ASIZE(dva) \ 328 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 329 #define DVA_SET_ASIZE(dva, x) \ 330 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 331 SPA_MINBLOCKSHIFT, 0, x) 332 333 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 334 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 335 336 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) 337 #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) 338 339 #define DVA_GET_OFFSET(dva) \ 340 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 341 #define DVA_SET_OFFSET(dva, x) \ 342 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 343 344 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 345 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 346 347 #define BP_GET_LSIZE(bp) \ 348 (BP_IS_EMBEDDED(bp) ? \ 349 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 350 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 351 #define BP_SET_LSIZE(bp, x) do { \ 352 ASSERT(!BP_IS_EMBEDDED(bp)); \ 353 BF64_SET_SB((bp)->blk_prop, \ 354 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 355 _NOTE(CONSTCOND) } while (0) 356 357 #define BP_GET_PSIZE(bp) \ 358 BF64_GET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1) 359 #define BP_SET_PSIZE(bp, x) \ 360 BF64_SET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x) 361 362 #define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7) 363 #define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x) 364 365 #define BP_GET_CHECKSUM(bp) BF64_GET((bp)->blk_prop, 40, 8) 366 #define BP_SET_CHECKSUM(bp, x) BF64_SET((bp)->blk_prop, 40, 8, x) 367 368 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 369 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 370 371 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 372 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 373 374 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 375 376 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 377 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 378 379 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 380 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 381 382 #define BP_PHYSICAL_BIRTH(bp) \ 383 ((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 384 385 #define BP_GET_ASIZE(bp) \ 386 (DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 387 DVA_GET_ASIZE(&(bp)->blk_dva[2])) 388 389 #define BP_GET_UCSIZE(bp) \ 390 ((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \ 391 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)); 392 393 #define BP_GET_NDVAS(bp) \ 394 (!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 395 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 396 !!DVA_GET_ASIZE(&(bp)->blk_dva[2])) 397 398 #define DVA_EQUAL(dva1, dva2) \ 399 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 400 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 401 402 #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 403 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 404 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 405 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 406 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 407 408 409 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 410 411 #define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ 412 { \ 413 (zcp)->zc_word[0] = w0; \ 414 (zcp)->zc_word[1] = w1; \ 415 (zcp)->zc_word[2] = w2; \ 416 (zcp)->zc_word[3] = w3; \ 417 } 418 419 #define BP_IDENTITY(bp) (&(bp)->blk_dva[0]) 420 #define BP_IS_GANG(bp) DVA_GET_GANG(BP_IDENTITY(bp)) 421 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 422 (dva)->dva_word[1] == 0ULL) 423 #define BP_IS_HOLE(bp) DVA_IS_EMPTY(BP_IDENTITY(bp)) 424 #define BP_IS_OLDER(bp, txg) (!BP_IS_HOLE(bp) && (bp)->blk_birth < (txg)) 425 426 #define BP_ZERO(bp) \ 427 { \ 428 (bp)->blk_dva[0].dva_word[0] = 0; \ 429 (bp)->blk_dva[0].dva_word[1] = 0; \ 430 (bp)->blk_dva[1].dva_word[0] = 0; \ 431 (bp)->blk_dva[1].dva_word[1] = 0; \ 432 (bp)->blk_dva[2].dva_word[0] = 0; \ 433 (bp)->blk_dva[2].dva_word[1] = 0; \ 434 (bp)->blk_prop = 0; \ 435 (bp)->blk_pad[0] = 0; \ 436 (bp)->blk_pad[1] = 0; \ 437 (bp)->blk_phys_birth = 0; \ 438 (bp)->blk_birth = 0; \ 439 (bp)->blk_fill = 0; \ 440 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 441 } 442 443 #define BPE_NUM_WORDS 14 444 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 445 #define BPE_IS_PAYLOADWORD(bp, wp) \ 446 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 447 448 /* 449 * Embedded checksum 450 */ 451 #define ZEC_MAGIC 0x210da7ab10c7a11ULL 452 453 typedef struct zio_eck { 454 uint64_t zec_magic; /* for validation, endianness */ 455 zio_cksum_t zec_cksum; /* 256-bit checksum */ 456 } zio_eck_t; 457 458 /* 459 * Gang block headers are self-checksumming and contain an array 460 * of block pointers. 461 */ 462 #define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE 463 #define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \ 464 sizeof (zio_eck_t)) / sizeof (blkptr_t)) 465 #define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \ 466 sizeof (zio_eck_t) - \ 467 (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\ 468 sizeof (uint64_t)) 469 470 typedef struct zio_gbh { 471 blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS]; 472 uint64_t zg_filler[SPA_GBH_FILLER]; 473 zio_eck_t zg_tail; 474 } zio_gbh_phys_t; 475 476 #define VDEV_RAIDZ_MAXPARITY 3 477 478 #define VDEV_PAD_SIZE (8 << 10) 479 /* 2 padding areas (vl_pad1 and vl_pad2) to skip */ 480 #define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2 481 #define VDEV_PHYS_SIZE (112 << 10) 482 #define VDEV_UBERBLOCK_RING (128 << 10) 483 484 #define VDEV_UBERBLOCK_SHIFT(vd) \ 485 MAX((vd)->v_top->v_ashift, UBERBLOCK_SHIFT) 486 #define VDEV_UBERBLOCK_COUNT(vd) \ 487 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd)) 488 #define VDEV_UBERBLOCK_OFFSET(vd, n) \ 489 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)]) 490 #define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd)) 491 492 typedef struct vdev_phys { 493 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)]; 494 zio_eck_t vp_zbt; 495 } vdev_phys_t; 496 497 typedef struct vdev_label { 498 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */ 499 char vl_pad2[VDEV_PAD_SIZE]; /* 8K */ 500 vdev_phys_t vl_vdev_phys; /* 112K */ 501 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */ 502 } vdev_label_t; /* 256K total */ 503 504 /* 505 * vdev_dirty() flags 506 */ 507 #define VDD_METASLAB 0x01 508 #define VDD_DTL 0x02 509 510 /* 511 * Size and offset of embedded boot loader region on each label. 512 * The total size of the first two labels plus the boot area is 4MB. 513 */ 514 #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t)) 515 #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */ 516 517 /* 518 * Size of label regions at the start and end of each leaf device. 519 */ 520 #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE) 521 #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t)) 522 #define VDEV_LABELS 4 523 524 enum zio_checksum { 525 ZIO_CHECKSUM_INHERIT = 0, 526 ZIO_CHECKSUM_ON, 527 ZIO_CHECKSUM_OFF, 528 ZIO_CHECKSUM_LABEL, 529 ZIO_CHECKSUM_GANG_HEADER, 530 ZIO_CHECKSUM_ZILOG, 531 ZIO_CHECKSUM_FLETCHER_2, 532 ZIO_CHECKSUM_FLETCHER_4, 533 ZIO_CHECKSUM_SHA256, 534 ZIO_CHECKSUM_ZILOG2, 535 ZIO_CHECKSUM_NOPARITY, 536 ZIO_CHECKSUM_SHA512, 537 ZIO_CHECKSUM_SKEIN, 538 ZIO_CHECKSUM_EDONR, 539 ZIO_CHECKSUM_FUNCTIONS 540 }; 541 542 #define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4 543 #define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON 544 545 enum zio_compress { 546 ZIO_COMPRESS_INHERIT = 0, 547 ZIO_COMPRESS_ON, 548 ZIO_COMPRESS_OFF, 549 ZIO_COMPRESS_LZJB, 550 ZIO_COMPRESS_EMPTY, 551 ZIO_COMPRESS_GZIP_1, 552 ZIO_COMPRESS_GZIP_2, 553 ZIO_COMPRESS_GZIP_3, 554 ZIO_COMPRESS_GZIP_4, 555 ZIO_COMPRESS_GZIP_5, 556 ZIO_COMPRESS_GZIP_6, 557 ZIO_COMPRESS_GZIP_7, 558 ZIO_COMPRESS_GZIP_8, 559 ZIO_COMPRESS_GZIP_9, 560 ZIO_COMPRESS_ZLE, 561 ZIO_COMPRESS_LZ4, 562 ZIO_COMPRESS_FUNCTIONS 563 }; 564 565 #define ZIO_COMPRESS_ON_VALUE ZIO_COMPRESS_LZJB 566 #define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF 567 568 /* nvlist pack encoding */ 569 #define NV_ENCODE_NATIVE 0 570 #define NV_ENCODE_XDR 1 571 572 typedef enum { 573 DATA_TYPE_UNKNOWN = 0, 574 DATA_TYPE_BOOLEAN, 575 DATA_TYPE_BYTE, 576 DATA_TYPE_INT16, 577 DATA_TYPE_UINT16, 578 DATA_TYPE_INT32, 579 DATA_TYPE_UINT32, 580 DATA_TYPE_INT64, 581 DATA_TYPE_UINT64, 582 DATA_TYPE_STRING, 583 DATA_TYPE_BYTE_ARRAY, 584 DATA_TYPE_INT16_ARRAY, 585 DATA_TYPE_UINT16_ARRAY, 586 DATA_TYPE_INT32_ARRAY, 587 DATA_TYPE_UINT32_ARRAY, 588 DATA_TYPE_INT64_ARRAY, 589 DATA_TYPE_UINT64_ARRAY, 590 DATA_TYPE_STRING_ARRAY, 591 DATA_TYPE_HRTIME, 592 DATA_TYPE_NVLIST, 593 DATA_TYPE_NVLIST_ARRAY, 594 DATA_TYPE_BOOLEAN_VALUE, 595 DATA_TYPE_INT8, 596 DATA_TYPE_UINT8, 597 DATA_TYPE_BOOLEAN_ARRAY, 598 DATA_TYPE_INT8_ARRAY, 599 DATA_TYPE_UINT8_ARRAY 600 } data_type_t; 601 602 /* 603 * On-disk version number. 604 */ 605 #define SPA_VERSION_1 1ULL 606 #define SPA_VERSION_2 2ULL 607 #define SPA_VERSION_3 3ULL 608 #define SPA_VERSION_4 4ULL 609 #define SPA_VERSION_5 5ULL 610 #define SPA_VERSION_6 6ULL 611 #define SPA_VERSION_7 7ULL 612 #define SPA_VERSION_8 8ULL 613 #define SPA_VERSION_9 9ULL 614 #define SPA_VERSION_10 10ULL 615 #define SPA_VERSION_11 11ULL 616 #define SPA_VERSION_12 12ULL 617 #define SPA_VERSION_13 13ULL 618 #define SPA_VERSION_14 14ULL 619 #define SPA_VERSION_15 15ULL 620 #define SPA_VERSION_16 16ULL 621 #define SPA_VERSION_17 17ULL 622 #define SPA_VERSION_18 18ULL 623 #define SPA_VERSION_19 19ULL 624 #define SPA_VERSION_20 20ULL 625 #define SPA_VERSION_21 21ULL 626 #define SPA_VERSION_22 22ULL 627 #define SPA_VERSION_23 23ULL 628 #define SPA_VERSION_24 24ULL 629 #define SPA_VERSION_25 25ULL 630 #define SPA_VERSION_26 26ULL 631 #define SPA_VERSION_27 27ULL 632 #define SPA_VERSION_28 28ULL 633 #define SPA_VERSION_5000 5000ULL 634 635 /* 636 * When bumping up SPA_VERSION, make sure GRUB ZFS understands the on-disk 637 * format change. Go to usr/src/grub/grub-0.97/stage2/{zfs-include/, fsys_zfs*}, 638 * and do the appropriate changes. Also bump the version number in 639 * usr/src/grub/capability. 640 */ 641 #define SPA_VERSION SPA_VERSION_5000 642 #define SPA_VERSION_STRING "5000" 643 644 /* 645 * Symbolic names for the changes that caused a SPA_VERSION switch. 646 * Used in the code when checking for presence or absence of a feature. 647 * Feel free to define multiple symbolic names for each version if there 648 * were multiple changes to on-disk structures during that version. 649 * 650 * NOTE: When checking the current SPA_VERSION in your code, be sure 651 * to use spa_version() since it reports the version of the 652 * last synced uberblock. Checking the in-flight version can 653 * be dangerous in some cases. 654 */ 655 #define SPA_VERSION_INITIAL SPA_VERSION_1 656 #define SPA_VERSION_DITTO_BLOCKS SPA_VERSION_2 657 #define SPA_VERSION_SPARES SPA_VERSION_3 658 #define SPA_VERSION_RAID6 SPA_VERSION_3 659 #define SPA_VERSION_BPLIST_ACCOUNT SPA_VERSION_3 660 #define SPA_VERSION_RAIDZ_DEFLATE SPA_VERSION_3 661 #define SPA_VERSION_DNODE_BYTES SPA_VERSION_3 662 #define SPA_VERSION_ZPOOL_HISTORY SPA_VERSION_4 663 #define SPA_VERSION_GZIP_COMPRESSION SPA_VERSION_5 664 #define SPA_VERSION_BOOTFS SPA_VERSION_6 665 #define SPA_VERSION_SLOGS SPA_VERSION_7 666 #define SPA_VERSION_DELEGATED_PERMS SPA_VERSION_8 667 #define SPA_VERSION_FUID SPA_VERSION_9 668 #define SPA_VERSION_REFRESERVATION SPA_VERSION_9 669 #define SPA_VERSION_REFQUOTA SPA_VERSION_9 670 #define SPA_VERSION_UNIQUE_ACCURATE SPA_VERSION_9 671 #define SPA_VERSION_L2CACHE SPA_VERSION_10 672 #define SPA_VERSION_NEXT_CLONES SPA_VERSION_11 673 #define SPA_VERSION_ORIGIN SPA_VERSION_11 674 #define SPA_VERSION_DSL_SCRUB SPA_VERSION_11 675 #define SPA_VERSION_SNAP_PROPS SPA_VERSION_12 676 #define SPA_VERSION_USED_BREAKDOWN SPA_VERSION_13 677 #define SPA_VERSION_PASSTHROUGH_X SPA_VERSION_14 678 #define SPA_VERSION_USERSPACE SPA_VERSION_15 679 #define SPA_VERSION_STMF_PROP SPA_VERSION_16 680 #define SPA_VERSION_RAIDZ3 SPA_VERSION_17 681 #define SPA_VERSION_USERREFS SPA_VERSION_18 682 #define SPA_VERSION_HOLES SPA_VERSION_19 683 #define SPA_VERSION_ZLE_COMPRESSION SPA_VERSION_20 684 #define SPA_VERSION_DEDUP SPA_VERSION_21 685 #define SPA_VERSION_RECVD_PROPS SPA_VERSION_22 686 #define SPA_VERSION_SLIM_ZIL SPA_VERSION_23 687 #define SPA_VERSION_SA SPA_VERSION_24 688 #define SPA_VERSION_SCAN SPA_VERSION_25 689 #define SPA_VERSION_DIR_CLONES SPA_VERSION_26 690 #define SPA_VERSION_DEADLISTS SPA_VERSION_26 691 #define SPA_VERSION_FAST_SNAP SPA_VERSION_27 692 #define SPA_VERSION_MULTI_REPLACE SPA_VERSION_28 693 #define SPA_VERSION_BEFORE_FEATURES SPA_VERSION_28 694 #define SPA_VERSION_FEATURES SPA_VERSION_5000 695 696 #define SPA_VERSION_IS_SUPPORTED(v) \ 697 (((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \ 698 ((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION)) 699 700 /* 701 * The following are configuration names used in the nvlist describing a pool's 702 * configuration. 703 */ 704 #define ZPOOL_CONFIG_VERSION "version" 705 #define ZPOOL_CONFIG_POOL_NAME "name" 706 #define ZPOOL_CONFIG_POOL_STATE "state" 707 #define ZPOOL_CONFIG_POOL_TXG "txg" 708 #define ZPOOL_CONFIG_POOL_GUID "pool_guid" 709 #define ZPOOL_CONFIG_CREATE_TXG "create_txg" 710 #define ZPOOL_CONFIG_TOP_GUID "top_guid" 711 #define ZPOOL_CONFIG_VDEV_TREE "vdev_tree" 712 #define ZPOOL_CONFIG_TYPE "type" 713 #define ZPOOL_CONFIG_CHILDREN "children" 714 #define ZPOOL_CONFIG_ID "id" 715 #define ZPOOL_CONFIG_GUID "guid" 716 #define ZPOOL_CONFIG_PATH "path" 717 #define ZPOOL_CONFIG_DEVID "devid" 718 #define ZPOOL_CONFIG_PHYS_PATH "phys_path" 719 #define ZPOOL_CONFIG_METASLAB_ARRAY "metaslab_array" 720 #define ZPOOL_CONFIG_METASLAB_SHIFT "metaslab_shift" 721 #define ZPOOL_CONFIG_ASHIFT "ashift" 722 #define ZPOOL_CONFIG_ASIZE "asize" 723 #define ZPOOL_CONFIG_DTL "DTL" 724 #define ZPOOL_CONFIG_STATS "stats" 725 #define ZPOOL_CONFIG_WHOLE_DISK "whole_disk" 726 #define ZPOOL_CONFIG_ERRCOUNT "error_count" 727 #define ZPOOL_CONFIG_NOT_PRESENT "not_present" 728 #define ZPOOL_CONFIG_SPARES "spares" 729 #define ZPOOL_CONFIG_IS_SPARE "is_spare" 730 #define ZPOOL_CONFIG_NPARITY "nparity" 731 #define ZPOOL_CONFIG_HOSTID "hostid" 732 #define ZPOOL_CONFIG_HOSTNAME "hostname" 733 #define ZPOOL_CONFIG_IS_LOG "is_log" 734 #define ZPOOL_CONFIG_TIMESTAMP "timestamp" /* not stored on disk */ 735 #define ZPOOL_CONFIG_FEATURES_FOR_READ "features_for_read" 736 737 /* 738 * The persistent vdev state is stored as separate values rather than a single 739 * 'vdev_state' entry. This is because a device can be in multiple states, such 740 * as offline and degraded. 741 */ 742 #define ZPOOL_CONFIG_OFFLINE "offline" 743 #define ZPOOL_CONFIG_FAULTED "faulted" 744 #define ZPOOL_CONFIG_DEGRADED "degraded" 745 #define ZPOOL_CONFIG_REMOVED "removed" 746 #define ZPOOL_CONFIG_FRU "fru" 747 #define ZPOOL_CONFIG_AUX_STATE "aux_state" 748 749 #define VDEV_TYPE_ROOT "root" 750 #define VDEV_TYPE_MIRROR "mirror" 751 #define VDEV_TYPE_REPLACING "replacing" 752 #define VDEV_TYPE_RAIDZ "raidz" 753 #define VDEV_TYPE_DISK "disk" 754 #define VDEV_TYPE_FILE "file" 755 #define VDEV_TYPE_MISSING "missing" 756 #define VDEV_TYPE_HOLE "hole" 757 #define VDEV_TYPE_SPARE "spare" 758 #define VDEV_TYPE_LOG "log" 759 #define VDEV_TYPE_L2CACHE "l2cache" 760 761 /* 762 * This is needed in userland to report the minimum necessary device size. 763 */ 764 #define SPA_MINDEVSIZE (64ULL << 20) 765 766 /* 767 * The location of the pool configuration repository, shared between kernel and 768 * userland. 769 */ 770 #define ZPOOL_CACHE "/boot/zfs/zpool.cache" 771 772 /* 773 * vdev states are ordered from least to most healthy. 774 * A vdev that's CANT_OPEN or below is considered unusable. 775 */ 776 typedef enum vdev_state { 777 VDEV_STATE_UNKNOWN = 0, /* Uninitialized vdev */ 778 VDEV_STATE_CLOSED, /* Not currently open */ 779 VDEV_STATE_OFFLINE, /* Not allowed to open */ 780 VDEV_STATE_REMOVED, /* Explicitly removed from system */ 781 VDEV_STATE_CANT_OPEN, /* Tried to open, but failed */ 782 VDEV_STATE_FAULTED, /* External request to fault device */ 783 VDEV_STATE_DEGRADED, /* Replicated vdev with unhealthy kids */ 784 VDEV_STATE_HEALTHY /* Presumed good */ 785 } vdev_state_t; 786 787 /* 788 * vdev aux states. When a vdev is in the CANT_OPEN state, the aux field 789 * of the vdev stats structure uses these constants to distinguish why. 790 */ 791 typedef enum vdev_aux { 792 VDEV_AUX_NONE, /* no error */ 793 VDEV_AUX_OPEN_FAILED, /* ldi_open_*() or vn_open() failed */ 794 VDEV_AUX_CORRUPT_DATA, /* bad label or disk contents */ 795 VDEV_AUX_NO_REPLICAS, /* insufficient number of replicas */ 796 VDEV_AUX_BAD_GUID_SUM, /* vdev guid sum doesn't match */ 797 VDEV_AUX_TOO_SMALL, /* vdev size is too small */ 798 VDEV_AUX_BAD_LABEL, /* the label is OK but invalid */ 799 VDEV_AUX_VERSION_NEWER, /* on-disk version is too new */ 800 VDEV_AUX_VERSION_OLDER, /* on-disk version is too old */ 801 VDEV_AUX_SPARED /* hot spare used in another pool */ 802 } vdev_aux_t; 803 804 /* 805 * pool state. The following states are written to disk as part of the normal 806 * SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE. The remaining states are 807 * software abstractions used at various levels to communicate pool state. 808 */ 809 typedef enum pool_state { 810 POOL_STATE_ACTIVE = 0, /* In active use */ 811 POOL_STATE_EXPORTED, /* Explicitly exported */ 812 POOL_STATE_DESTROYED, /* Explicitly destroyed */ 813 POOL_STATE_SPARE, /* Reserved for hot spare use */ 814 POOL_STATE_UNINITIALIZED, /* Internal spa_t state */ 815 POOL_STATE_UNAVAIL, /* Internal libzfs state */ 816 POOL_STATE_POTENTIALLY_ACTIVE /* Internal libzfs state */ 817 } pool_state_t; 818 819 /* 820 * The uberblock version is incremented whenever an incompatible on-disk 821 * format change is made to the SPA, DMU, or ZAP. 822 * 823 * Note: the first two fields should never be moved. When a storage pool 824 * is opened, the uberblock must be read off the disk before the version 825 * can be checked. If the ub_version field is moved, we may not detect 826 * version mismatch. If the ub_magic field is moved, applications that 827 * expect the magic number in the first word won't work. 828 */ 829 #define UBERBLOCK_MAGIC 0x00bab10c /* oo-ba-bloc! */ 830 #define UBERBLOCK_SHIFT 10 /* up to 1K */ 831 832 struct uberblock { 833 uint64_t ub_magic; /* UBERBLOCK_MAGIC */ 834 uint64_t ub_version; /* SPA_VERSION */ 835 uint64_t ub_txg; /* txg of last sync */ 836 uint64_t ub_guid_sum; /* sum of all vdev guids */ 837 uint64_t ub_timestamp; /* UTC time of last sync */ 838 blkptr_t ub_rootbp; /* MOS objset_phys_t */ 839 }; 840 841 /* 842 * Flags. 843 */ 844 #define DNODE_MUST_BE_ALLOCATED 1 845 #define DNODE_MUST_BE_FREE 2 846 847 /* 848 * Fixed constants. 849 */ 850 #define DNODE_SHIFT 9 /* 512 bytes */ 851 #define DN_MIN_INDBLKSHIFT 12 /* 4k */ 852 #define DN_MAX_INDBLKSHIFT 14 /* 16k */ 853 #define DNODE_BLOCK_SHIFT 14 /* 16k */ 854 #define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */ 855 #define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */ 856 #define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */ 857 858 /* 859 * Derived constants. 860 */ 861 #define DNODE_MIN_SIZE (1 << DNODE_SHIFT) 862 #define DNODE_MAX_SIZE (1 << DNODE_BLOCK_SHIFT) 863 #define DNODE_BLOCK_SIZE (1 << DNODE_BLOCK_SHIFT) 864 #define DNODE_MIN_SLOTS (DNODE_MIN_SIZE >> DNODE_SHIFT) 865 #define DNODE_MAX_SLOTS (DNODE_MAX_SIZE >> DNODE_SHIFT) 866 #define DN_BONUS_SIZE(dnsize) ((dnsize) - DNODE_CORE_SIZE - \ 867 (1 << SPA_BLKPTRSHIFT)) 868 #define DN_SLOTS_TO_BONUSLEN(slots) DN_BONUS_SIZE((slots) << DNODE_SHIFT) 869 #define DN_OLD_MAX_BONUSLEN (DN_BONUS_SIZE(DNODE_MIN_SIZE)) 870 #define DN_MAX_NBLKPTR ((DNODE_MIN_SIZE - DNODE_CORE_SIZE) >> \ 871 SPA_BLKPTRSHIFT) 872 #define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT) 873 #define DN_ZERO_BONUSLEN (DN_BONUS_SIZE(DNODE_MAX_SIZE) + 1) 874 875 #define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT) 876 #define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT) 877 #define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT) 878 879 /* The +2 here is a cheesy way to round up */ 880 #define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \ 881 (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT))) 882 883 #define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \ 884 (((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t)))) 885 886 #define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \ 887 (dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT) 888 889 #define EPB(blkshift, typeshift) (1 << (blkshift - typeshift)) 890 891 /* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */ 892 #define DNODE_FLAG_USED_BYTES (1<<0) 893 #define DNODE_FLAG_USERUSED_ACCOUNTED (1<<1) 894 895 /* Does dnode have a SA spill blkptr in bonus? */ 896 #define DNODE_FLAG_SPILL_BLKPTR (1<<2) 897 898 typedef struct dnode_phys { 899 uint8_t dn_type; /* dmu_object_type_t */ 900 uint8_t dn_indblkshift; /* ln2(indirect block size) */ 901 uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */ 902 uint8_t dn_nblkptr; /* length of dn_blkptr */ 903 uint8_t dn_bonustype; /* type of data in bonus buffer */ 904 uint8_t dn_checksum; /* ZIO_CHECKSUM type */ 905 uint8_t dn_compress; /* ZIO_COMPRESS type */ 906 uint8_t dn_flags; /* DNODE_FLAG_* */ 907 uint16_t dn_datablkszsec; /* data block size in 512b sectors */ 908 uint16_t dn_bonuslen; /* length of dn_bonus */ 909 uint8_t dn_extra_slots; /* # of subsequent slots consumed */ 910 uint8_t dn_pad2[3]; 911 912 /* accounting is protected by dn_dirty_mtx */ 913 uint64_t dn_maxblkid; /* largest allocated block ID */ 914 uint64_t dn_used; /* bytes (or sectors) of disk space */ 915 916 uint64_t dn_pad3[4]; 917 918 /* 919 * The tail region is 448 bytes for a 512 byte dnode, and 920 * correspondingly larger for larger dnode sizes. The spill 921 * block pointer, when present, is always at the end of the tail 922 * region. There are three ways this space may be used, using 923 * a 512 byte dnode for this diagram: 924 * 925 * 0 64 128 192 256 320 384 448 (offset) 926 * +---------------+---------------+---------------+-------+ 927 * | dn_blkptr[0] | dn_blkptr[1] | dn_blkptr[2] | / | 928 * +---------------+---------------+---------------+-------+ 929 * | dn_blkptr[0] | dn_bonus[0..319] | 930 * +---------------+-----------------------+---------------+ 931 * | dn_blkptr[0] | dn_bonus[0..191] | dn_spill | 932 * +---------------+-----------------------+---------------+ 933 */ 934 union { 935 blkptr_t dn_blkptr[1+DN_OLD_MAX_BONUSLEN/sizeof (blkptr_t)]; 936 struct { 937 blkptr_t __dn_ignore1; 938 uint8_t dn_bonus[DN_OLD_MAX_BONUSLEN]; 939 }; 940 struct { 941 blkptr_t __dn_ignore2; 942 uint8_t __dn_ignore3[DN_OLD_MAX_BONUSLEN - 943 sizeof (blkptr_t)]; 944 blkptr_t dn_spill; 945 }; 946 }; 947 } dnode_phys_t; 948 949 #define DN_SPILL_BLKPTR(dnp) (blkptr_t *)((char *)(dnp) + \ 950 (((dnp)->dn_extra_slots + 1) << DNODE_SHIFT) - (1 << SPA_BLKPTRSHIFT)) 951 952 typedef enum dmu_object_byteswap { 953 DMU_BSWAP_UINT8, 954 DMU_BSWAP_UINT16, 955 DMU_BSWAP_UINT32, 956 DMU_BSWAP_UINT64, 957 DMU_BSWAP_ZAP, 958 DMU_BSWAP_DNODE, 959 DMU_BSWAP_OBJSET, 960 DMU_BSWAP_ZNODE, 961 DMU_BSWAP_OLDACL, 962 DMU_BSWAP_ACL, 963 /* 964 * Allocating a new byteswap type number makes the on-disk format 965 * incompatible with any other format that uses the same number. 966 * 967 * Data can usually be structured to work with one of the 968 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. 969 */ 970 DMU_BSWAP_NUMFUNCS 971 } dmu_object_byteswap_t; 972 973 #define DMU_OT_NEWTYPE 0x80 974 #define DMU_OT_METADATA 0x40 975 #define DMU_OT_BYTESWAP_MASK 0x3f 976 977 /* 978 * Defines a uint8_t object type. Object types specify if the data 979 * in the object is metadata (boolean) and how to byteswap the data 980 * (dmu_object_byteswap_t). 981 */ 982 #define DMU_OT(byteswap, metadata) \ 983 (DMU_OT_NEWTYPE | \ 984 ((metadata) ? DMU_OT_METADATA : 0) | \ 985 ((byteswap) & DMU_OT_BYTESWAP_MASK)) 986 987 typedef enum dmu_object_type { 988 DMU_OT_NONE, 989 /* general: */ 990 DMU_OT_OBJECT_DIRECTORY, /* ZAP */ 991 DMU_OT_OBJECT_ARRAY, /* UINT64 */ 992 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ 993 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ 994 DMU_OT_BPLIST, /* UINT64 */ 995 DMU_OT_BPLIST_HDR, /* UINT64 */ 996 /* spa: */ 997 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ 998 DMU_OT_SPACE_MAP, /* UINT64 */ 999 /* zil: */ 1000 DMU_OT_INTENT_LOG, /* UINT64 */ 1001 /* dmu: */ 1002 DMU_OT_DNODE, /* DNODE */ 1003 DMU_OT_OBJSET, /* OBJSET */ 1004 /* dsl: */ 1005 DMU_OT_DSL_DIR, /* UINT64 */ 1006 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ 1007 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ 1008 DMU_OT_DSL_PROPS, /* ZAP */ 1009 DMU_OT_DSL_DATASET, /* UINT64 */ 1010 /* zpl: */ 1011 DMU_OT_ZNODE, /* ZNODE */ 1012 DMU_OT_OLDACL, /* Old ACL */ 1013 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ 1014 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ 1015 DMU_OT_MASTER_NODE, /* ZAP */ 1016 DMU_OT_UNLINKED_SET, /* ZAP */ 1017 /* zvol: */ 1018 DMU_OT_ZVOL, /* UINT8 */ 1019 DMU_OT_ZVOL_PROP, /* ZAP */ 1020 /* other; for testing only! */ 1021 DMU_OT_PLAIN_OTHER, /* UINT8 */ 1022 DMU_OT_UINT64_OTHER, /* UINT64 */ 1023 DMU_OT_ZAP_OTHER, /* ZAP */ 1024 /* new object types: */ 1025 DMU_OT_ERROR_LOG, /* ZAP */ 1026 DMU_OT_SPA_HISTORY, /* UINT8 */ 1027 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ 1028 DMU_OT_POOL_PROPS, /* ZAP */ 1029 DMU_OT_DSL_PERMS, /* ZAP */ 1030 DMU_OT_ACL, /* ACL */ 1031 DMU_OT_SYSACL, /* SYSACL */ 1032 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ 1033 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ 1034 DMU_OT_NEXT_CLONES, /* ZAP */ 1035 DMU_OT_SCAN_QUEUE, /* ZAP */ 1036 DMU_OT_USERGROUP_USED, /* ZAP */ 1037 DMU_OT_USERGROUP_QUOTA, /* ZAP */ 1038 DMU_OT_USERREFS, /* ZAP */ 1039 DMU_OT_DDT_ZAP, /* ZAP */ 1040 DMU_OT_DDT_STATS, /* ZAP */ 1041 DMU_OT_SA, /* System attr */ 1042 DMU_OT_SA_MASTER_NODE, /* ZAP */ 1043 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ 1044 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ 1045 DMU_OT_SCAN_XLATE, /* ZAP */ 1046 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ 1047 DMU_OT_NUMTYPES, 1048 1049 /* 1050 * Names for valid types declared with DMU_OT(). 1051 */ 1052 DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE), 1053 DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE), 1054 DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE), 1055 DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE), 1056 DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE), 1057 DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE), 1058 DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE), 1059 DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE), 1060 DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE), 1061 DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE) 1062 } dmu_object_type_t; 1063 1064 typedef enum dmu_objset_type { 1065 DMU_OST_NONE, 1066 DMU_OST_META, 1067 DMU_OST_ZFS, 1068 DMU_OST_ZVOL, 1069 DMU_OST_OTHER, /* For testing only! */ 1070 DMU_OST_ANY, /* Be careful! */ 1071 DMU_OST_NUMTYPES 1072 } dmu_objset_type_t; 1073 1074 #define ZAP_MAXVALUELEN (1024 * 8) 1075 1076 /* 1077 * header for all bonus and spill buffers. 1078 * The header has a fixed portion with a variable number 1079 * of "lengths" depending on the number of variable sized 1080 * attribues which are determined by the "layout number" 1081 */ 1082 1083 #define SA_MAGIC 0x2F505A /* ZFS SA */ 1084 typedef struct sa_hdr_phys { 1085 uint32_t sa_magic; 1086 uint16_t sa_layout_info; /* Encoded with hdrsize and layout number */ 1087 uint16_t sa_lengths[1]; /* optional sizes for variable length attrs */ 1088 /* ... Data follows the lengths. */ 1089 } sa_hdr_phys_t; 1090 1091 /* 1092 * sa_hdr_phys -> sa_layout_info 1093 * 1094 * 16 10 0 1095 * +--------+-------+ 1096 * | hdrsz |layout | 1097 * +--------+-------+ 1098 * 1099 * Bits 0-10 are the layout number 1100 * Bits 11-16 are the size of the header. 1101 * The hdrsize is the number * 8 1102 * 1103 * For example. 1104 * hdrsz of 1 ==> 8 byte header 1105 * 2 ==> 16 byte header 1106 * 1107 */ 1108 1109 #define SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10) 1110 #define SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 16, 3, 0) 1111 #define SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \ 1112 { \ 1113 BF32_SET_SB(x, 10, 6, 3, 0, size); \ 1114 BF32_SET(x, 0, 10, num); \ 1115 } 1116 1117 #define SA_MODE_OFFSET 0 1118 #define SA_SIZE_OFFSET 8 1119 #define SA_GEN_OFFSET 16 1120 #define SA_UID_OFFSET 24 1121 #define SA_GID_OFFSET 32 1122 #define SA_PARENT_OFFSET 40 1123 #define SA_SYMLINK_OFFSET 160 1124 1125 /* 1126 * Intent log header - this on disk structure holds fields to manage 1127 * the log. All fields are 64 bit to easily handle cross architectures. 1128 */ 1129 typedef struct zil_header { 1130 uint64_t zh_claim_txg; /* txg in which log blocks were claimed */ 1131 uint64_t zh_replay_seq; /* highest replayed sequence number */ 1132 blkptr_t zh_log; /* log chain */ 1133 uint64_t zh_claim_seq; /* highest claimed sequence number */ 1134 uint64_t zh_pad[5]; 1135 } zil_header_t; 1136 1137 #define OBJSET_PHYS_SIZE 2048 1138 1139 typedef struct objset_phys { 1140 dnode_phys_t os_meta_dnode; 1141 zil_header_t os_zil_header; 1142 uint64_t os_type; 1143 uint64_t os_flags; 1144 char os_pad[OBJSET_PHYS_SIZE - sizeof (dnode_phys_t)*3 - 1145 sizeof (zil_header_t) - sizeof (uint64_t)*2]; 1146 dnode_phys_t os_userused_dnode; 1147 dnode_phys_t os_groupused_dnode; 1148 } objset_phys_t; 1149 1150 typedef struct dsl_dir_phys { 1151 uint64_t dd_creation_time; /* not actually used */ 1152 uint64_t dd_head_dataset_obj; 1153 uint64_t dd_parent_obj; 1154 uint64_t dd_clone_parent_obj; 1155 uint64_t dd_child_dir_zapobj; 1156 /* 1157 * how much space our children are accounting for; for leaf 1158 * datasets, == physical space used by fs + snaps 1159 */ 1160 uint64_t dd_used_bytes; 1161 uint64_t dd_compressed_bytes; 1162 uint64_t dd_uncompressed_bytes; 1163 /* Administrative quota setting */ 1164 uint64_t dd_quota; 1165 /* Administrative reservation setting */ 1166 uint64_t dd_reserved; 1167 uint64_t dd_props_zapobj; 1168 uint64_t dd_pad[21]; /* pad out to 256 bytes for good measure */ 1169 } dsl_dir_phys_t; 1170 1171 typedef struct dsl_dataset_phys { 1172 uint64_t ds_dir_obj; 1173 uint64_t ds_prev_snap_obj; 1174 uint64_t ds_prev_snap_txg; 1175 uint64_t ds_next_snap_obj; 1176 uint64_t ds_snapnames_zapobj; /* zap obj of snaps; ==0 for snaps */ 1177 uint64_t ds_num_children; /* clone/snap children; ==0 for head */ 1178 uint64_t ds_creation_time; /* seconds since 1970 */ 1179 uint64_t ds_creation_txg; 1180 uint64_t ds_deadlist_obj; 1181 uint64_t ds_used_bytes; 1182 uint64_t ds_compressed_bytes; 1183 uint64_t ds_uncompressed_bytes; 1184 uint64_t ds_unique_bytes; /* only relevant to snapshots */ 1185 /* 1186 * The ds_fsid_guid is a 56-bit ID that can change to avoid 1187 * collisions. The ds_guid is a 64-bit ID that will never 1188 * change, so there is a small probability that it will collide. 1189 */ 1190 uint64_t ds_fsid_guid; 1191 uint64_t ds_guid; 1192 uint64_t ds_flags; 1193 blkptr_t ds_bp; 1194 uint64_t ds_pad[8]; /* pad out to 320 bytes for good measure */ 1195 } dsl_dataset_phys_t; 1196 1197 /* 1198 * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 1199 */ 1200 #define DMU_POOL_DIRECTORY_OBJECT 1 1201 #define DMU_POOL_CONFIG "config" 1202 #define DMU_POOL_FEATURES_FOR_READ "features_for_read" 1203 #define DMU_POOL_ROOT_DATASET "root_dataset" 1204 #define DMU_POOL_SYNC_BPLIST "sync_bplist" 1205 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 1206 #define DMU_POOL_ERRLOG_LAST "errlog_last" 1207 #define DMU_POOL_SPARES "spares" 1208 #define DMU_POOL_DEFLATE "deflate" 1209 #define DMU_POOL_HISTORY "history" 1210 #define DMU_POOL_PROPS "pool_props" 1211 #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" 1212 1213 #define ZAP_MAGIC 0x2F52AB2ABULL 1214 1215 #define FZAP_BLOCK_SHIFT(zap) ((zap)->zap_block_shift) 1216 1217 #define ZAP_MAXCD (uint32_t)(-1) 1218 #define ZAP_HASHBITS 28 1219 #define MZAP_ENT_LEN 64 1220 #define MZAP_NAME_LEN (MZAP_ENT_LEN - 8 - 4 - 2) 1221 #define MZAP_MAX_BLKSHIFT SPA_MAXBLOCKSHIFT 1222 #define MZAP_MAX_BLKSZ (1 << MZAP_MAX_BLKSHIFT) 1223 1224 typedef struct mzap_ent_phys { 1225 uint64_t mze_value; 1226 uint32_t mze_cd; 1227 uint16_t mze_pad; /* in case we want to chain them someday */ 1228 char mze_name[MZAP_NAME_LEN]; 1229 } mzap_ent_phys_t; 1230 1231 typedef struct mzap_phys { 1232 uint64_t mz_block_type; /* ZBT_MICRO */ 1233 uint64_t mz_salt; 1234 uint64_t mz_pad[6]; 1235 mzap_ent_phys_t mz_chunk[1]; 1236 /* actually variable size depending on block size */ 1237 } mzap_phys_t; 1238 1239 /* 1240 * The (fat) zap is stored in one object. It is an array of 1241 * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of: 1242 * 1243 * ptrtbl fits in first block: 1244 * [zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ... 1245 * 1246 * ptrtbl too big for first block: 1247 * [zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ... 1248 * 1249 */ 1250 1251 #define ZBT_LEAF ((1ULL << 63) + 0) 1252 #define ZBT_HEADER ((1ULL << 63) + 1) 1253 #define ZBT_MICRO ((1ULL << 63) + 3) 1254 /* any other values are ptrtbl blocks */ 1255 1256 /* 1257 * the embedded pointer table takes up half a block: 1258 * block size / entry size (2^3) / 2 1259 */ 1260 #define ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1) 1261 1262 /* 1263 * The embedded pointer table starts half-way through the block. Since 1264 * the pointer table itself is half the block, it starts at (64-bit) 1265 * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)). 1266 */ 1267 #define ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \ 1268 ((uint64_t *)(zap)->zap_phys) \ 1269 [(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))] 1270 1271 /* 1272 * TAKE NOTE: 1273 * If zap_phys_t is modified, zap_byteswap() must be modified. 1274 */ 1275 typedef struct zap_phys { 1276 uint64_t zap_block_type; /* ZBT_HEADER */ 1277 uint64_t zap_magic; /* ZAP_MAGIC */ 1278 1279 struct zap_table_phys { 1280 uint64_t zt_blk; /* starting block number */ 1281 uint64_t zt_numblks; /* number of blocks */ 1282 uint64_t zt_shift; /* bits to index it */ 1283 uint64_t zt_nextblk; /* next (larger) copy start block */ 1284 uint64_t zt_blks_copied; /* number source blocks copied */ 1285 } zap_ptrtbl; 1286 1287 uint64_t zap_freeblk; /* the next free block */ 1288 uint64_t zap_num_leafs; /* number of leafs */ 1289 uint64_t zap_num_entries; /* number of entries */ 1290 uint64_t zap_salt; /* salt to stir into hash function */ 1291 /* 1292 * This structure is followed by padding, and then the embedded 1293 * pointer table. The embedded pointer table takes up second 1294 * half of the block. It is accessed using the 1295 * ZAP_EMBEDDED_PTRTBL_ENT() macro. 1296 */ 1297 } zap_phys_t; 1298 1299 typedef struct zap_table_phys zap_table_phys_t; 1300 1301 typedef struct fat_zap { 1302 int zap_block_shift; /* block size shift */ 1303 zap_phys_t *zap_phys; 1304 } fat_zap_t; 1305 1306 #define ZAP_LEAF_MAGIC 0x2AB1EAF 1307 1308 /* chunk size = 24 bytes */ 1309 #define ZAP_LEAF_CHUNKSIZE 24 1310 1311 /* 1312 * The amount of space available for chunks is: 1313 * block size (1<<l->l_bs) - hash entry size (2) * number of hash 1314 * entries - header space (2*chunksize) 1315 */ 1316 #define ZAP_LEAF_NUMCHUNKS(l) \ 1317 (((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \ 1318 ZAP_LEAF_CHUNKSIZE - 2) 1319 1320 /* 1321 * The amount of space within the chunk available for the array is: 1322 * chunk size - space for type (1) - space for next pointer (2) 1323 */ 1324 #define ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3) 1325 1326 #define ZAP_LEAF_ARRAY_NCHUNKS(bytes) \ 1327 (((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES) 1328 1329 /* 1330 * Low water mark: when there are only this many chunks free, start 1331 * growing the ptrtbl. Ideally, this should be larger than a 1332 * "reasonably-sized" entry. 20 chunks is more than enough for the 1333 * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value), 1334 * while still being only around 3% for 16k blocks. 1335 */ 1336 #define ZAP_LEAF_LOW_WATER (20) 1337 1338 /* 1339 * The leaf hash table has block size / 2^5 (32) number of entries, 1340 * which should be more than enough for the maximum number of entries, 1341 * which is less than block size / CHUNKSIZE (24) / minimum number of 1342 * chunks per entry (3). 1343 */ 1344 #define ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5) 1345 #define ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l)) 1346 1347 /* 1348 * The chunks start immediately after the hash table. The end of the 1349 * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a 1350 * chunk_t. 1351 */ 1352 #define ZAP_LEAF_CHUNK(l, idx) \ 1353 ((zap_leaf_chunk_t *) \ 1354 ((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx] 1355 #define ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry) 1356 1357 typedef enum zap_chunk_type { 1358 ZAP_CHUNK_FREE = 253, 1359 ZAP_CHUNK_ENTRY = 252, 1360 ZAP_CHUNK_ARRAY = 251, 1361 ZAP_CHUNK_TYPE_MAX = 250 1362 } zap_chunk_type_t; 1363 1364 /* 1365 * TAKE NOTE: 1366 * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified. 1367 */ 1368 typedef struct zap_leaf_phys { 1369 struct zap_leaf_header { 1370 uint64_t lh_block_type; /* ZBT_LEAF */ 1371 uint64_t lh_pad1; 1372 uint64_t lh_prefix; /* hash prefix of this leaf */ 1373 uint32_t lh_magic; /* ZAP_LEAF_MAGIC */ 1374 uint16_t lh_nfree; /* number free chunks */ 1375 uint16_t lh_nentries; /* number of entries */ 1376 uint16_t lh_prefix_len; /* num bits used to id this */ 1377 1378 /* above is accessable to zap, below is zap_leaf private */ 1379 1380 uint16_t lh_freelist; /* chunk head of free list */ 1381 uint8_t lh_pad2[12]; 1382 } l_hdr; /* 2 24-byte chunks */ 1383 1384 /* 1385 * The header is followed by a hash table with 1386 * ZAP_LEAF_HASH_NUMENTRIES(zap) entries. The hash table is 1387 * followed by an array of ZAP_LEAF_NUMCHUNKS(zap) 1388 * zap_leaf_chunk structures. These structures are accessed 1389 * with the ZAP_LEAF_CHUNK() macro. 1390 */ 1391 1392 uint16_t l_hash[1]; 1393 } zap_leaf_phys_t; 1394 1395 typedef union zap_leaf_chunk { 1396 struct zap_leaf_entry { 1397 uint8_t le_type; /* always ZAP_CHUNK_ENTRY */ 1398 uint8_t le_value_intlen; /* size of ints */ 1399 uint16_t le_next; /* next entry in hash chain */ 1400 uint16_t le_name_chunk; /* first chunk of the name */ 1401 uint16_t le_name_numints; /* bytes in name, incl null */ 1402 uint16_t le_value_chunk; /* first chunk of the value */ 1403 uint16_t le_value_numints; /* value length in ints */ 1404 uint32_t le_cd; /* collision differentiator */ 1405 uint64_t le_hash; /* hash value of the name */ 1406 } l_entry; 1407 struct zap_leaf_array { 1408 uint8_t la_type; /* always ZAP_CHUNK_ARRAY */ 1409 uint8_t la_array[ZAP_LEAF_ARRAY_BYTES]; 1410 uint16_t la_next; /* next blk or CHAIN_END */ 1411 } l_array; 1412 struct zap_leaf_free { 1413 uint8_t lf_type; /* always ZAP_CHUNK_FREE */ 1414 uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES]; 1415 uint16_t lf_next; /* next in free list, or CHAIN_END */ 1416 } l_free; 1417 } zap_leaf_chunk_t; 1418 1419 typedef struct zap_leaf { 1420 int l_bs; /* block size shift */ 1421 zap_leaf_phys_t *l_phys; 1422 } zap_leaf_t; 1423 1424 /* 1425 * Define special zfs pflags 1426 */ 1427 #define ZFS_XATTR 0x1 /* is an extended attribute */ 1428 #define ZFS_INHERIT_ACE 0x2 /* ace has inheritable ACEs */ 1429 #define ZFS_ACL_TRIVIAL 0x4 /* files ACL is trivial */ 1430 1431 #define MASTER_NODE_OBJ 1 1432 1433 /* 1434 * special attributes for master node. 1435 */ 1436 1437 #define ZFS_FSID "FSID" 1438 #define ZFS_UNLINKED_SET "DELETE_QUEUE" 1439 #define ZFS_ROOT_OBJ "ROOT" 1440 #define ZPL_VERSION_OBJ "VERSION" 1441 #define ZFS_PROP_BLOCKPERPAGE "BLOCKPERPAGE" 1442 #define ZFS_PROP_NOGROWBLOCKS "NOGROWBLOCKS" 1443 1444 #define ZFS_FLAG_BLOCKPERPAGE 0x1 1445 #define ZFS_FLAG_NOGROWBLOCKS 0x2 1446 1447 /* 1448 * ZPL version - rev'd whenever an incompatible on-disk format change 1449 * occurs. Independent of SPA/DMU/ZAP versioning. 1450 */ 1451 1452 #define ZPL_VERSION 1ULL 1453 1454 /* 1455 * The directory entry has the type (currently unused on Solaris) in the 1456 * top 4 bits, and the object number in the low 48 bits. The "middle" 1457 * 12 bits are unused. 1458 */ 1459 #define ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4) 1460 #define ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48) 1461 #define ZFS_DIRENT_MAKE(type, obj) (((uint64_t)type << 60) | obj) 1462 1463 typedef struct ace { 1464 uid_t a_who; /* uid or gid */ 1465 uint32_t a_access_mask; /* read,write,... */ 1466 uint16_t a_flags; /* see below */ 1467 uint16_t a_type; /* allow or deny */ 1468 } ace_t; 1469 1470 #define ACE_SLOT_CNT 6 1471 1472 typedef struct zfs_znode_acl { 1473 uint64_t z_acl_extern_obj; /* ext acl pieces */ 1474 uint32_t z_acl_count; /* Number of ACEs */ 1475 uint16_t z_acl_version; /* acl version */ 1476 uint16_t z_acl_pad; /* pad */ 1477 ace_t z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */ 1478 } zfs_znode_acl_t; 1479 1480 /* 1481 * This is the persistent portion of the znode. It is stored 1482 * in the "bonus buffer" of the file. Short symbolic links 1483 * are also stored in the bonus buffer. 1484 */ 1485 typedef struct znode_phys { 1486 uint64_t zp_atime[2]; /* 0 - last file access time */ 1487 uint64_t zp_mtime[2]; /* 16 - last file modification time */ 1488 uint64_t zp_ctime[2]; /* 32 - last file change time */ 1489 uint64_t zp_crtime[2]; /* 48 - creation time */ 1490 uint64_t zp_gen; /* 64 - generation (txg of creation) */ 1491 uint64_t zp_mode; /* 72 - file mode bits */ 1492 uint64_t zp_size; /* 80 - size of file */ 1493 uint64_t zp_parent; /* 88 - directory parent (`..') */ 1494 uint64_t zp_links; /* 96 - number of links to file */ 1495 uint64_t zp_xattr; /* 104 - DMU object for xattrs */ 1496 uint64_t zp_rdev; /* 112 - dev_t for VBLK & VCHR files */ 1497 uint64_t zp_flags; /* 120 - persistent flags */ 1498 uint64_t zp_uid; /* 128 - file owner */ 1499 uint64_t zp_gid; /* 136 - owning group */ 1500 uint64_t zp_pad[4]; /* 144 - future */ 1501 zfs_znode_acl_t zp_acl; /* 176 - 263 ACL */ 1502 /* 1503 * Data may pad out any remaining bytes in the znode buffer, eg: 1504 * 1505 * |<---------------------- dnode_phys (512) ------------------------>| 1506 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->| 1507 * |<---- znode (264) ---->|<---- data (56) ---->| 1508 * 1509 * At present, we only use this space to store symbolic links. 1510 */ 1511 } znode_phys_t; 1512 1513 /* 1514 * In-core vdev representation. 1515 */ 1516 struct vdev; 1517 struct spa; 1518 typedef int vdev_phys_read_t(struct vdev *vdev, void *priv, 1519 off_t offset, void *buf, size_t bytes); 1520 typedef int vdev_read_t(struct vdev *vdev, const blkptr_t *bp, 1521 void *buf, off_t offset, size_t bytes); 1522 1523 typedef STAILQ_HEAD(vdev_list, vdev) vdev_list_t; 1524 1525 typedef struct vdev { 1526 STAILQ_ENTRY(vdev) v_childlink; /* link in parent's child list */ 1527 STAILQ_ENTRY(vdev) v_alllink; /* link in global vdev list */ 1528 vdev_list_t v_children; /* children of this vdev */ 1529 const char *v_name; /* vdev name */ 1530 const char *v_phys_path; /* vdev bootpath */ 1531 const char *v_devid; /* vdev devid */ 1532 uint64_t v_guid; /* vdev guid */ 1533 int v_id; /* index in parent */ 1534 int v_ashift; /* offset to block shift */ 1535 int v_nparity; /* # parity for raidz */ 1536 struct vdev *v_top; /* parent vdev */ 1537 int v_nchildren; /* # children */ 1538 vdev_state_t v_state; /* current state */ 1539 vdev_phys_read_t *v_phys_read; /* read from raw leaf vdev */ 1540 vdev_read_t *v_read; /* read from vdev */ 1541 void *v_read_priv; /* private data for read function */ 1542 struct spa *spa; /* link to spa */ 1543 } vdev_t; 1544 1545 /* 1546 * In-core pool representation. 1547 */ 1548 typedef STAILQ_HEAD(spa_list, spa) spa_list_t; 1549 1550 typedef struct spa { 1551 STAILQ_ENTRY(spa) spa_link; /* link in global pool list */ 1552 char *spa_name; /* pool name */ 1553 uint64_t spa_guid; /* pool guid */ 1554 uint64_t spa_txg; /* most recent transaction */ 1555 struct uberblock spa_uberblock; /* best uberblock so far */ 1556 vdev_list_t spa_vdevs; /* list of all toplevel vdevs */ 1557 objset_phys_t spa_mos; /* MOS for this pool */ 1558 zio_cksum_salt_t spa_cksum_salt; /* secret salt for cksum */ 1559 void *spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS]; 1560 int spa_inited; /* initialized */ 1561 vdev_t *spa_boot_vdev; /* boot device for kernel */ 1562 } spa_t; 1563 1564 static void decode_embedded_bp_compressed(const blkptr_t *, void *); 1565