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#define MAXNAMELEN 256 60 61/* CRC64 table */ 62#define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ 63 64/* 65 * Macros for various sorts of alignment and rounding when the alignment 66 * is known to be a power of 2. 67 */ 68#define P2ALIGN(x, align) ((x) & -(align)) 69#define P2PHASE(x, align) ((x) & ((align) - 1)) 70#define P2NPHASE(x, align) (-(x) & ((align) - 1)) 71#define P2ROUNDUP(x, align) (-(-(x) & -(align))) 72#define P2END(x, align) (-(~(x) & -(align))) 73#define P2PHASEUP(x, align, phase) ((phase) - (((phase) - (x)) & -(align))) 74#define P2BOUNDARY(off, len, align) (((off) ^ ((off) + (len) - 1)) > (align) - 1) 75 76/* 77 * General-purpose 32-bit and 64-bit bitfield encodings. 78 */ 79#define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) 80#define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) 81#define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) 82#define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) 83 84#define BF32_GET(x, low, len) BF32_DECODE(x, low, len) 85#define BF64_GET(x, low, len) BF64_DECODE(x, low, len) 86 87#define BF32_SET(x, low, len, val) \ 88 ((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len)) 89#define BF64_SET(x, low, len, val) \ 90 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)) 91 92#define BF32_GET_SB(x, low, len, shift, bias) \ 93 ((BF32_GET(x, low, len) + (bias)) << (shift)) 94#define BF64_GET_SB(x, low, len, shift, bias) \ 95 ((BF64_GET(x, low, len) + (bias)) << (shift)) 96 97#define BF32_SET_SB(x, low, len, shift, bias, val) \ 98 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)) 99#define BF64_SET_SB(x, low, len, shift, bias, val) \ 100 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)) 101 102/* 103 * Macros to reverse byte order 104 */ 105#define BSWAP_8(x) ((x) & 0xff) 106#define BSWAP_16(x) ((BSWAP_8(x) << 8) | BSWAP_8((x) >> 8)) 107#define BSWAP_32(x) ((BSWAP_16(x) << 16) | BSWAP_16((x) >> 16)) 108#define BSWAP_64(x) ((BSWAP_32(x) << 32) | BSWAP_32((x) >> 32)) 109 110/* 111 * We currently support nine block sizes, from 512 bytes to 128K. 112 * We could go higher, but the benefits are near-zero and the cost 113 * of COWing a giant block to modify one byte would become excessive. 114 */ 115#define SPA_MINBLOCKSHIFT 9 116#define SPA_MAXBLOCKSHIFT 17 117#define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 118#define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 119 120#define SPA_BLOCKSIZES (SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1) 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 */ 137typedef 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 */ 144typedef struct zio_cksum { 145 uint64_t zc_word[4]; 146} zio_cksum_t; 147 148/* 149 * Each block is described by its DVAs, time of birth, checksum, etc. 150 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 151 * 152 * 64 56 48 40 32 24 16 8 0 153 * +-------+-------+-------+-------+-------+-------+-------+-------+ 154 * 0 | vdev1 | GRID | ASIZE | 155 * +-------+-------+-------+-------+-------+-------+-------+-------+ 156 * 1 |G| offset1 | 157 * +-------+-------+-------+-------+-------+-------+-------+-------+ 158 * 2 | vdev2 | GRID | ASIZE | 159 * +-------+-------+-------+-------+-------+-------+-------+-------+ 160 * 3 |G| offset2 | 161 * +-------+-------+-------+-------+-------+-------+-------+-------+ 162 * 4 | vdev3 | GRID | ASIZE | 163 * +-------+-------+-------+-------+-------+-------+-------+-------+ 164 * 5 |G| offset3 | 165 * +-------+-------+-------+-------+-------+-------+-------+-------+ 166 * 6 |BDX|lvl| type | cksum | comp | PSIZE | LSIZE | 167 * +-------+-------+-------+-------+-------+-------+-------+-------+ 168 * 7 | padding | 169 * +-------+-------+-------+-------+-------+-------+-------+-------+ 170 * 8 | padding | 171 * +-------+-------+-------+-------+-------+-------+-------+-------+ 172 * 9 | physical birth txg | 173 * +-------+-------+-------+-------+-------+-------+-------+-------+ 174 * a | logical birth txg | 175 * +-------+-------+-------+-------+-------+-------+-------+-------+ 176 * b | fill count | 177 * +-------+-------+-------+-------+-------+-------+-------+-------+ 178 * c | checksum[0] | 179 * +-------+-------+-------+-------+-------+-------+-------+-------+ 180 * d | checksum[1] | 181 * +-------+-------+-------+-------+-------+-------+-------+-------+ 182 * e | checksum[2] | 183 * +-------+-------+-------+-------+-------+-------+-------+-------+ 184 * f | checksum[3] | 185 * +-------+-------+-------+-------+-------+-------+-------+-------+ 186 * 187 * Legend: 188 * 189 * vdev virtual device ID 190 * offset offset into virtual device 191 * LSIZE logical size 192 * PSIZE physical size (after compression) 193 * ASIZE allocated size (including RAID-Z parity and gang block headers) 194 * GRID RAID-Z layout information (reserved for future use) 195 * cksum checksum function 196 * comp compression function 197 * G gang block indicator 198 * B byteorder (endianness) 199 * D dedup 200 * X unused 201 * lvl level of indirection 202 * type DMU object type 203 * phys birth txg of block allocation; zero if same as logical birth txg 204 * log. birth transaction group in which the block was logically born 205 * fill count number of non-zero blocks under this bp 206 * checksum[4] 256-bit checksum of the data this bp describes 207 */ 208#define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 209#define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 210 211typedef struct blkptr { 212 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 213 uint64_t blk_prop; /* size, compression, type, etc */ 214 uint64_t blk_pad[2]; /* Extra space for the future */ 215 uint64_t blk_phys_birth; /* txg when block was allocated */ 216 uint64_t blk_birth; /* transaction group at birth */ 217 uint64_t blk_fill; /* fill count */ 218 zio_cksum_t blk_cksum; /* 256-bit checksum */ 219} blkptr_t; 220 221/* 222 * Macros to get and set fields in a bp or DVA. 223 */ 224#define DVA_GET_ASIZE(dva) \ 225 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 226#define DVA_SET_ASIZE(dva, x) \ 227 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 228 SPA_MINBLOCKSHIFT, 0, x) 229 230#define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 231#define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 232 233#define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) 234#define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) 235 236#define DVA_GET_OFFSET(dva) \ 237 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 238#define DVA_SET_OFFSET(dva, x) \ 239 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 240 241#define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 242#define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 243 244#define BP_GET_LSIZE(bp) \ 245 (BP_IS_HOLE(bp) ? 0 : \ 246 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 247#define BP_SET_LSIZE(bp, x) \ 248 BF64_SET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x) 249 250#define BP_GET_PSIZE(bp) \ 251 BF64_GET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1) 252#define BP_SET_PSIZE(bp, x) \ 253 BF64_SET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x) 254 255#define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 8) 256#define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 8, x) 257 258#define BP_GET_CHECKSUM(bp) BF64_GET((bp)->blk_prop, 40, 8) 259#define BP_SET_CHECKSUM(bp, x) BF64_SET((bp)->blk_prop, 40, 8, x) 260 261#define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 262#define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 263 264#define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 265#define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 266 267#define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 268#define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 269 270#define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 271#define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 272 273#define BP_PHYSICAL_BIRTH(bp) \ 274 ((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 275 276#define BP_GET_ASIZE(bp) \ 277 (DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 278 DVA_GET_ASIZE(&(bp)->blk_dva[2])) 279 280#define BP_GET_UCSIZE(bp) \ 281 ((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \ 282 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)); 283 284#define BP_GET_NDVAS(bp) \ 285 (!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 286 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 287 !!DVA_GET_ASIZE(&(bp)->blk_dva[2])) 288 289#define DVA_EQUAL(dva1, dva2) \ 290 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 291 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 292 293#define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 294 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 295 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 296 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 297 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 298 299 300#define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 301 302#define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ 303{ \ 304 (zcp)->zc_word[0] = w0; \ 305 (zcp)->zc_word[1] = w1; \ 306 (zcp)->zc_word[2] = w2; \ 307 (zcp)->zc_word[3] = w3; \ 308} 309 310#define BP_IDENTITY(bp) (&(bp)->blk_dva[0]) 311#define BP_IS_GANG(bp) DVA_GET_GANG(BP_IDENTITY(bp)) 312#define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 313 (dva)->dva_word[1] == 0ULL) 314#define BP_IS_HOLE(bp) DVA_IS_EMPTY(BP_IDENTITY(bp)) 315#define BP_IS_OLDER(bp, txg) (!BP_IS_HOLE(bp) && (bp)->blk_birth < (txg)) 316 317#define BP_ZERO(bp) \ 318{ \ 319 (bp)->blk_dva[0].dva_word[0] = 0; \ 320 (bp)->blk_dva[0].dva_word[1] = 0; \ 321 (bp)->blk_dva[1].dva_word[0] = 0; \ 322 (bp)->blk_dva[1].dva_word[1] = 0; \ 323 (bp)->blk_dva[2].dva_word[0] = 0; \ 324 (bp)->blk_dva[2].dva_word[1] = 0; \ 325 (bp)->blk_prop = 0; \ 326 (bp)->blk_pad[0] = 0; \ 327 (bp)->blk_pad[1] = 0; \ 328 (bp)->blk_phys_birth = 0; \ 329 (bp)->blk_birth = 0; \ 330 (bp)->blk_fill = 0; \ 331 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 332} 333 334/* 335 * Embedded checksum 336 */ 337#define ZEC_MAGIC 0x210da7ab10c7a11ULL 338 339typedef struct zio_eck { 340 uint64_t zec_magic; /* for validation, endianness */ 341 zio_cksum_t zec_cksum; /* 256-bit checksum */ 342} zio_eck_t; 343 344/* 345 * Gang block headers are self-checksumming and contain an array 346 * of block pointers. 347 */ 348#define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE 349#define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \ 350 sizeof (zio_eck_t)) / sizeof (blkptr_t)) 351#define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \ 352 sizeof (zio_eck_t) - \ 353 (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\ 354 sizeof (uint64_t)) 355 356typedef struct zio_gbh { 357 blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS]; 358 uint64_t zg_filler[SPA_GBH_FILLER]; 359 zio_eck_t zg_tail; 360} zio_gbh_phys_t; 361 362#define VDEV_RAIDZ_MAXPARITY 3 363 364#define VDEV_PAD_SIZE (8 << 10) 365/* 2 padding areas (vl_pad1 and vl_pad2) to skip */ 366#define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2 367#define VDEV_PHYS_SIZE (112 << 10) 368#define VDEV_UBERBLOCK_RING (128 << 10) 369 370#define VDEV_UBERBLOCK_SHIFT(vd) \ 371 MAX((vd)->v_top->v_ashift, UBERBLOCK_SHIFT) 372#define VDEV_UBERBLOCK_COUNT(vd) \ 373 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd)) 374#define VDEV_UBERBLOCK_OFFSET(vd, n) \ 375 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)]) 376#define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd)) 377 378typedef struct vdev_phys { 379 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)]; 380 zio_eck_t vp_zbt; 381} vdev_phys_t; 382 383typedef struct vdev_label { 384 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */ 385 char vl_pad2[VDEV_PAD_SIZE]; /* 8K */ 386 vdev_phys_t vl_vdev_phys; /* 112K */ 387 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */ 388} vdev_label_t; /* 256K total */ 389 390/* 391 * vdev_dirty() flags 392 */ 393#define VDD_METASLAB 0x01 394#define VDD_DTL 0x02 395 396/* 397 * Size and offset of embedded boot loader region on each label. 398 * The total size of the first two labels plus the boot area is 4MB. 399 */ 400#define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t)) 401#define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */ 402 403/* 404 * Size of label regions at the start and end of each leaf device. 405 */ 406#define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE) 407#define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t)) 408#define VDEV_LABELS 4 409 410enum zio_checksum { 411 ZIO_CHECKSUM_INHERIT = 0, 412 ZIO_CHECKSUM_ON, 413 ZIO_CHECKSUM_OFF, 414 ZIO_CHECKSUM_LABEL, 415 ZIO_CHECKSUM_GANG_HEADER, 416 ZIO_CHECKSUM_ZILOG, 417 ZIO_CHECKSUM_FLETCHER_2, 418 ZIO_CHECKSUM_FLETCHER_4, 419 ZIO_CHECKSUM_SHA256, 420 ZIO_CHECKSUM_ZILOG2, 421 ZIO_CHECKSUM_FUNCTIONS 422}; 423 424#define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4 425#define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON 426 427enum zio_compress { 428 ZIO_COMPRESS_INHERIT = 0, 429 ZIO_COMPRESS_ON, 430 ZIO_COMPRESS_OFF, 431 ZIO_COMPRESS_LZJB, 432 ZIO_COMPRESS_EMPTY, 433 ZIO_COMPRESS_GZIP_1, 434 ZIO_COMPRESS_GZIP_2, 435 ZIO_COMPRESS_GZIP_3, 436 ZIO_COMPRESS_GZIP_4, 437 ZIO_COMPRESS_GZIP_5, 438 ZIO_COMPRESS_GZIP_6, 439 ZIO_COMPRESS_GZIP_7, 440 ZIO_COMPRESS_GZIP_8, 441 ZIO_COMPRESS_GZIP_9, 442 ZIO_COMPRESS_ZLE, 443 ZIO_COMPRESS_LZ4, 444 ZIO_COMPRESS_FUNCTIONS 445}; 446 447#define ZIO_COMPRESS_ON_VALUE ZIO_COMPRESS_LZJB 448#define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF 449 450/* nvlist pack encoding */ 451#define NV_ENCODE_NATIVE 0 452#define NV_ENCODE_XDR 1 453 454typedef enum { 455 DATA_TYPE_UNKNOWN = 0, 456 DATA_TYPE_BOOLEAN, 457 DATA_TYPE_BYTE, 458 DATA_TYPE_INT16, 459 DATA_TYPE_UINT16, 460 DATA_TYPE_INT32, 461 DATA_TYPE_UINT32, 462 DATA_TYPE_INT64, 463 DATA_TYPE_UINT64, 464 DATA_TYPE_STRING, 465 DATA_TYPE_BYTE_ARRAY, 466 DATA_TYPE_INT16_ARRAY, 467 DATA_TYPE_UINT16_ARRAY, 468 DATA_TYPE_INT32_ARRAY, 469 DATA_TYPE_UINT32_ARRAY, 470 DATA_TYPE_INT64_ARRAY, 471 DATA_TYPE_UINT64_ARRAY, 472 DATA_TYPE_STRING_ARRAY, 473 DATA_TYPE_HRTIME, 474 DATA_TYPE_NVLIST, 475 DATA_TYPE_NVLIST_ARRAY, 476 DATA_TYPE_BOOLEAN_VALUE, 477 DATA_TYPE_INT8, 478 DATA_TYPE_UINT8, 479 DATA_TYPE_BOOLEAN_ARRAY, 480 DATA_TYPE_INT8_ARRAY, 481 DATA_TYPE_UINT8_ARRAY 482} data_type_t; 483 484/* 485 * On-disk version number. 486 */ 487#define SPA_VERSION_1 1ULL 488#define SPA_VERSION_2 2ULL 489#define SPA_VERSION_3 3ULL 490#define SPA_VERSION_4 4ULL 491#define SPA_VERSION_5 5ULL 492#define SPA_VERSION_6 6ULL 493#define SPA_VERSION_7 7ULL 494#define SPA_VERSION_8 8ULL 495#define SPA_VERSION_9 9ULL 496#define SPA_VERSION_10 10ULL 497#define SPA_VERSION_11 11ULL 498#define SPA_VERSION_12 12ULL 499#define SPA_VERSION_13 13ULL 500#define SPA_VERSION_14 14ULL 501#define SPA_VERSION_15 15ULL 502#define SPA_VERSION_16 16ULL 503#define SPA_VERSION_17 17ULL 504#define SPA_VERSION_18 18ULL 505#define SPA_VERSION_19 19ULL 506#define SPA_VERSION_20 20ULL 507#define SPA_VERSION_21 21ULL 508#define SPA_VERSION_22 22ULL 509#define SPA_VERSION_23 23ULL 510#define SPA_VERSION_24 24ULL 511#define SPA_VERSION_25 25ULL 512#define SPA_VERSION_26 26ULL 513#define SPA_VERSION_27 27ULL 514#define SPA_VERSION_28 28ULL 515#define SPA_VERSION_5000 5000ULL 516 517/* 518 * When bumping up SPA_VERSION, make sure GRUB ZFS understands the on-disk 519 * format change. Go to usr/src/grub/grub-0.97/stage2/{zfs-include/, fsys_zfs*}, 520 * and do the appropriate changes. Also bump the version number in 521 * usr/src/grub/capability. 522 */ 523#define SPA_VERSION SPA_VERSION_5000 524#define SPA_VERSION_STRING "5000" 525 526/* 527 * Symbolic names for the changes that caused a SPA_VERSION switch. 528 * Used in the code when checking for presence or absence of a feature. 529 * Feel free to define multiple symbolic names for each version if there 530 * were multiple changes to on-disk structures during that version. 531 * 532 * NOTE: When checking the current SPA_VERSION in your code, be sure 533 * to use spa_version() since it reports the version of the 534 * last synced uberblock. Checking the in-flight version can 535 * be dangerous in some cases. 536 */ 537#define SPA_VERSION_INITIAL SPA_VERSION_1 538#define SPA_VERSION_DITTO_BLOCKS SPA_VERSION_2 539#define SPA_VERSION_SPARES SPA_VERSION_3 540#define SPA_VERSION_RAID6 SPA_VERSION_3 541#define SPA_VERSION_BPLIST_ACCOUNT SPA_VERSION_3 542#define SPA_VERSION_RAIDZ_DEFLATE SPA_VERSION_3 543#define SPA_VERSION_DNODE_BYTES SPA_VERSION_3 544#define SPA_VERSION_ZPOOL_HISTORY SPA_VERSION_4 545#define SPA_VERSION_GZIP_COMPRESSION SPA_VERSION_5 546#define SPA_VERSION_BOOTFS SPA_VERSION_6 547#define SPA_VERSION_SLOGS SPA_VERSION_7 548#define SPA_VERSION_DELEGATED_PERMS SPA_VERSION_8 549#define SPA_VERSION_FUID SPA_VERSION_9 550#define SPA_VERSION_REFRESERVATION SPA_VERSION_9 551#define SPA_VERSION_REFQUOTA SPA_VERSION_9 552#define SPA_VERSION_UNIQUE_ACCURATE SPA_VERSION_9 553#define SPA_VERSION_L2CACHE SPA_VERSION_10 554#define SPA_VERSION_NEXT_CLONES SPA_VERSION_11 555#define SPA_VERSION_ORIGIN SPA_VERSION_11 556#define SPA_VERSION_DSL_SCRUB SPA_VERSION_11 557#define SPA_VERSION_SNAP_PROPS SPA_VERSION_12 558#define SPA_VERSION_USED_BREAKDOWN SPA_VERSION_13 559#define SPA_VERSION_PASSTHROUGH_X SPA_VERSION_14 560#define SPA_VERSION_USERSPACE SPA_VERSION_15 561#define SPA_VERSION_STMF_PROP SPA_VERSION_16 562#define SPA_VERSION_RAIDZ3 SPA_VERSION_17 563#define SPA_VERSION_USERREFS SPA_VERSION_18 564#define SPA_VERSION_HOLES SPA_VERSION_19 565#define SPA_VERSION_ZLE_COMPRESSION SPA_VERSION_20 566#define SPA_VERSION_DEDUP SPA_VERSION_21 567#define SPA_VERSION_RECVD_PROPS SPA_VERSION_22 568#define SPA_VERSION_SLIM_ZIL SPA_VERSION_23 569#define SPA_VERSION_SA SPA_VERSION_24 570#define SPA_VERSION_SCAN SPA_VERSION_25 571#define SPA_VERSION_DIR_CLONES SPA_VERSION_26 572#define SPA_VERSION_DEADLISTS SPA_VERSION_26 573#define SPA_VERSION_FAST_SNAP SPA_VERSION_27 574#define SPA_VERSION_MULTI_REPLACE SPA_VERSION_28 575#define SPA_VERSION_BEFORE_FEATURES SPA_VERSION_28 576#define SPA_VERSION_FEATURES SPA_VERSION_5000 577 578#define SPA_VERSION_IS_SUPPORTED(v) \ 579 (((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \ 580 ((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION)) 581 582/* 583 * The following are configuration names used in the nvlist describing a pool's 584 * configuration. 585 */ 586#define ZPOOL_CONFIG_VERSION "version" 587#define ZPOOL_CONFIG_POOL_NAME "name" 588#define ZPOOL_CONFIG_POOL_STATE "state" 589#define ZPOOL_CONFIG_POOL_TXG "txg" 590#define ZPOOL_CONFIG_POOL_GUID "pool_guid" 591#define ZPOOL_CONFIG_CREATE_TXG "create_txg" 592#define ZPOOL_CONFIG_TOP_GUID "top_guid" 593#define ZPOOL_CONFIG_VDEV_TREE "vdev_tree" 594#define ZPOOL_CONFIG_TYPE "type" 595#define ZPOOL_CONFIG_CHILDREN "children" 596#define ZPOOL_CONFIG_ID "id" 597#define ZPOOL_CONFIG_GUID "guid" 598#define ZPOOL_CONFIG_PATH "path" 599#define ZPOOL_CONFIG_DEVID "devid" 600#define ZPOOL_CONFIG_METASLAB_ARRAY "metaslab_array" 601#define ZPOOL_CONFIG_METASLAB_SHIFT "metaslab_shift" 602#define ZPOOL_CONFIG_ASHIFT "ashift" 603#define ZPOOL_CONFIG_ASIZE "asize" 604#define ZPOOL_CONFIG_DTL "DTL" 605#define ZPOOL_CONFIG_STATS "stats" 606#define ZPOOL_CONFIG_WHOLE_DISK "whole_disk" 607#define ZPOOL_CONFIG_ERRCOUNT "error_count" 608#define ZPOOL_CONFIG_NOT_PRESENT "not_present" 609#define ZPOOL_CONFIG_SPARES "spares" 610#define ZPOOL_CONFIG_IS_SPARE "is_spare" 611#define ZPOOL_CONFIG_NPARITY "nparity" 612#define ZPOOL_CONFIG_HOSTID "hostid" 613#define ZPOOL_CONFIG_HOSTNAME "hostname" 614#define ZPOOL_CONFIG_IS_LOG "is_log" 615#define ZPOOL_CONFIG_TIMESTAMP "timestamp" /* not stored on disk */ 616#define ZPOOL_CONFIG_FEATURES_FOR_READ "features_for_read" 617 618/* 619 * The persistent vdev state is stored as separate values rather than a single 620 * 'vdev_state' entry. This is because a device can be in multiple states, such 621 * as offline and degraded. 622 */ 623#define ZPOOL_CONFIG_OFFLINE "offline" 624#define ZPOOL_CONFIG_FAULTED "faulted" 625#define ZPOOL_CONFIG_DEGRADED "degraded" 626#define ZPOOL_CONFIG_REMOVED "removed" 627#define ZPOOL_CONFIG_FRU "fru" 628#define ZPOOL_CONFIG_AUX_STATE "aux_state" 629 630#define VDEV_TYPE_ROOT "root" 631#define VDEV_TYPE_MIRROR "mirror" 632#define VDEV_TYPE_REPLACING "replacing" 633#define VDEV_TYPE_RAIDZ "raidz" 634#define VDEV_TYPE_DISK "disk" 635#define VDEV_TYPE_FILE "file" 636#define VDEV_TYPE_MISSING "missing" 637#define VDEV_TYPE_HOLE "hole" 638#define VDEV_TYPE_SPARE "spare" 639#define VDEV_TYPE_LOG "log" 640#define VDEV_TYPE_L2CACHE "l2cache" 641 642/* 643 * This is needed in userland to report the minimum necessary device size. 644 */ 645#define SPA_MINDEVSIZE (64ULL << 20) 646 647/* 648 * The location of the pool configuration repository, shared between kernel and 649 * userland. 650 */ 651#define ZPOOL_CACHE "/boot/zfs/zpool.cache" 652 653/* 654 * vdev states are ordered from least to most healthy. 655 * A vdev that's CANT_OPEN or below is considered unusable. 656 */ 657typedef enum vdev_state { 658 VDEV_STATE_UNKNOWN = 0, /* Uninitialized vdev */ 659 VDEV_STATE_CLOSED, /* Not currently open */ 660 VDEV_STATE_OFFLINE, /* Not allowed to open */ 661 VDEV_STATE_REMOVED, /* Explicitly removed from system */ 662 VDEV_STATE_CANT_OPEN, /* Tried to open, but failed */ 663 VDEV_STATE_FAULTED, /* External request to fault device */ 664 VDEV_STATE_DEGRADED, /* Replicated vdev with unhealthy kids */ 665 VDEV_STATE_HEALTHY /* Presumed good */ 666} vdev_state_t; 667 668/* 669 * vdev aux states. When a vdev is in the CANT_OPEN state, the aux field 670 * of the vdev stats structure uses these constants to distinguish why. 671 */ 672typedef enum vdev_aux { 673 VDEV_AUX_NONE, /* no error */ 674 VDEV_AUX_OPEN_FAILED, /* ldi_open_*() or vn_open() failed */ 675 VDEV_AUX_CORRUPT_DATA, /* bad label or disk contents */ 676 VDEV_AUX_NO_REPLICAS, /* insufficient number of replicas */ 677 VDEV_AUX_BAD_GUID_SUM, /* vdev guid sum doesn't match */ 678 VDEV_AUX_TOO_SMALL, /* vdev size is too small */ 679 VDEV_AUX_BAD_LABEL, /* the label is OK but invalid */ 680 VDEV_AUX_VERSION_NEWER, /* on-disk version is too new */ 681 VDEV_AUX_VERSION_OLDER, /* on-disk version is too old */ 682 VDEV_AUX_SPARED /* hot spare used in another pool */ 683} vdev_aux_t; 684 685/* 686 * pool state. The following states are written to disk as part of the normal 687 * SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE. The remaining states are 688 * software abstractions used at various levels to communicate pool state. 689 */ 690typedef enum pool_state { 691 POOL_STATE_ACTIVE = 0, /* In active use */ 692 POOL_STATE_EXPORTED, /* Explicitly exported */ 693 POOL_STATE_DESTROYED, /* Explicitly destroyed */ 694 POOL_STATE_SPARE, /* Reserved for hot spare use */ 695 POOL_STATE_UNINITIALIZED, /* Internal spa_t state */ 696 POOL_STATE_UNAVAIL, /* Internal libzfs state */ 697 POOL_STATE_POTENTIALLY_ACTIVE /* Internal libzfs state */ 698} pool_state_t; 699 700/* 701 * The uberblock version is incremented whenever an incompatible on-disk 702 * format change is made to the SPA, DMU, or ZAP. 703 * 704 * Note: the first two fields should never be moved. When a storage pool 705 * is opened, the uberblock must be read off the disk before the version 706 * can be checked. If the ub_version field is moved, we may not detect 707 * version mismatch. If the ub_magic field is moved, applications that 708 * expect the magic number in the first word won't work. 709 */ 710#define UBERBLOCK_MAGIC 0x00bab10c /* oo-ba-bloc! */ 711#define UBERBLOCK_SHIFT 10 /* up to 1K */ 712 713struct uberblock { 714 uint64_t ub_magic; /* UBERBLOCK_MAGIC */ 715 uint64_t ub_version; /* SPA_VERSION */ 716 uint64_t ub_txg; /* txg of last sync */ 717 uint64_t ub_guid_sum; /* sum of all vdev guids */ 718 uint64_t ub_timestamp; /* UTC time of last sync */ 719 blkptr_t ub_rootbp; /* MOS objset_phys_t */ 720}; 721 722/* 723 * Flags. 724 */ 725#define DNODE_MUST_BE_ALLOCATED 1 726#define DNODE_MUST_BE_FREE 2 727 728/* 729 * Fixed constants. 730 */ 731#define DNODE_SHIFT 9 /* 512 bytes */ 732#define DN_MIN_INDBLKSHIFT 10 /* 1k */ 733#define DN_MAX_INDBLKSHIFT 14 /* 16k */ 734#define DNODE_BLOCK_SHIFT 14 /* 16k */ 735#define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */ 736#define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */ 737#define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */ 738 739/* 740 * Derived constants. 741 */ 742#define DNODE_SIZE (1 << DNODE_SHIFT) 743#define DN_MAX_NBLKPTR ((DNODE_SIZE - DNODE_CORE_SIZE) >> SPA_BLKPTRSHIFT) 744#define DN_MAX_BONUSLEN (DNODE_SIZE - DNODE_CORE_SIZE - (1 << SPA_BLKPTRSHIFT)) 745#define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT) 746 747#define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT) 748#define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT) 749#define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT) 750 751/* The +2 here is a cheesy way to round up */ 752#define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \ 753 (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT))) 754 755#define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \ 756 (((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t)))) 757 758#define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \ 759 (dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT) 760 761#define EPB(blkshift, typeshift) (1 << (blkshift - typeshift)) 762 763/* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */ 764#define DNODE_FLAG_USED_BYTES (1<<0) 765#define DNODE_FLAG_USERUSED_ACCOUNTED (1<<1) 766 767/* Does dnode have a SA spill blkptr in bonus? */ 768#define DNODE_FLAG_SPILL_BLKPTR (1<<2) 769 770typedef struct dnode_phys { 771 uint8_t dn_type; /* dmu_object_type_t */ 772 uint8_t dn_indblkshift; /* ln2(indirect block size) */ 773 uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */ 774 uint8_t dn_nblkptr; /* length of dn_blkptr */ 775 uint8_t dn_bonustype; /* type of data in bonus buffer */ 776 uint8_t dn_checksum; /* ZIO_CHECKSUM type */ 777 uint8_t dn_compress; /* ZIO_COMPRESS type */ 778 uint8_t dn_flags; /* DNODE_FLAG_* */ 779 uint16_t dn_datablkszsec; /* data block size in 512b sectors */ 780 uint16_t dn_bonuslen; /* length of dn_bonus */ 781 uint8_t dn_pad2[4]; 782 783 /* accounting is protected by dn_dirty_mtx */ 784 uint64_t dn_maxblkid; /* largest allocated block ID */ 785 uint64_t dn_used; /* bytes (or sectors) of disk space */ 786 787 uint64_t dn_pad3[4]; 788 789 blkptr_t dn_blkptr[1]; 790 uint8_t dn_bonus[DN_MAX_BONUSLEN - sizeof (blkptr_t)]; 791 blkptr_t dn_spill; 792} dnode_phys_t; 793 794typedef enum dmu_object_type { 795 DMU_OT_NONE, 796 /* general: */ 797 DMU_OT_OBJECT_DIRECTORY, /* ZAP */ 798 DMU_OT_OBJECT_ARRAY, /* UINT64 */ 799 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ 800 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ 801 DMU_OT_BPLIST, /* UINT64 */ 802 DMU_OT_BPLIST_HDR, /* UINT64 */ 803 /* spa: */ 804 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ 805 DMU_OT_SPACE_MAP, /* UINT64 */ 806 /* zil: */ 807 DMU_OT_INTENT_LOG, /* UINT64 */ 808 /* dmu: */ 809 DMU_OT_DNODE, /* DNODE */ 810 DMU_OT_OBJSET, /* OBJSET */ 811 /* dsl: */ 812 DMU_OT_DSL_DIR, /* UINT64 */ 813 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ 814 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ 815 DMU_OT_DSL_PROPS, /* ZAP */ 816 DMU_OT_DSL_DATASET, /* UINT64 */ 817 /* zpl: */ 818 DMU_OT_ZNODE, /* ZNODE */ 819 DMU_OT_OLDACL, /* Old ACL */ 820 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ 821 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ 822 DMU_OT_MASTER_NODE, /* ZAP */ 823 DMU_OT_UNLINKED_SET, /* ZAP */ 824 /* zvol: */ 825 DMU_OT_ZVOL, /* UINT8 */ 826 DMU_OT_ZVOL_PROP, /* ZAP */ 827 /* other; for testing only! */ 828 DMU_OT_PLAIN_OTHER, /* UINT8 */ 829 DMU_OT_UINT64_OTHER, /* UINT64 */ 830 DMU_OT_ZAP_OTHER, /* ZAP */ 831 /* new object types: */ 832 DMU_OT_ERROR_LOG, /* ZAP */ 833 DMU_OT_SPA_HISTORY, /* UINT8 */ 834 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ 835 DMU_OT_POOL_PROPS, /* ZAP */ 836 DMU_OT_DSL_PERMS, /* ZAP */ 837 DMU_OT_ACL, /* ACL */ 838 DMU_OT_SYSACL, /* SYSACL */ 839 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ 840 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ 841 DMU_OT_NEXT_CLONES, /* ZAP */ 842 DMU_OT_SCAN_QUEUE, /* ZAP */ 843 DMU_OT_USERGROUP_USED, /* ZAP */ 844 DMU_OT_USERGROUP_QUOTA, /* ZAP */ 845 DMU_OT_USERREFS, /* ZAP */ 846 DMU_OT_DDT_ZAP, /* ZAP */ 847 DMU_OT_DDT_STATS, /* ZAP */ 848 DMU_OT_SA, /* System attr */ 849 DMU_OT_SA_MASTER_NODE, /* ZAP */ 850 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ 851 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ 852 DMU_OT_SCAN_XLATE, /* ZAP */ 853 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ 854 DMU_OT_NUMTYPES 855} dmu_object_type_t; 856 857typedef enum dmu_objset_type { 858 DMU_OST_NONE, 859 DMU_OST_META, 860 DMU_OST_ZFS, 861 DMU_OST_ZVOL, 862 DMU_OST_OTHER, /* For testing only! */ 863 DMU_OST_ANY, /* Be careful! */ 864 DMU_OST_NUMTYPES 865} dmu_objset_type_t; 866 867/* 868 * header for all bonus and spill buffers. 869 * The header has a fixed portion with a variable number 870 * of "lengths" depending on the number of variable sized 871 * attribues which are determined by the "layout number" 872 */ 873 874#define SA_MAGIC 0x2F505A /* ZFS SA */ 875typedef struct sa_hdr_phys { 876 uint32_t sa_magic; 877 uint16_t sa_layout_info; /* Encoded with hdrsize and layout number */ 878 uint16_t sa_lengths[1]; /* optional sizes for variable length attrs */ 879 /* ... Data follows the lengths. */ 880} sa_hdr_phys_t; 881 882/* 883 * sa_hdr_phys -> sa_layout_info 884 * 885 * 16 10 0 886 * +--------+-------+ 887 * | hdrsz |layout | 888 * +--------+-------+ 889 * 890 * Bits 0-10 are the layout number 891 * Bits 11-16 are the size of the header. 892 * The hdrsize is the number * 8 893 * 894 * For example. 895 * hdrsz of 1 ==> 8 byte header 896 * 2 ==> 16 byte header 897 * 898 */ 899 900#define SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10) 901#define SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 16, 3, 0) 902#define SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \ 903{ \ 904 BF32_SET_SB(x, 10, 6, 3, 0, size); \ 905 BF32_SET(x, 0, 10, num); \ 906} 907 908#define SA_MODE_OFFSET 0 909#define SA_SIZE_OFFSET 8 910#define SA_GEN_OFFSET 16 911#define SA_UID_OFFSET 24 912#define SA_GID_OFFSET 32 913#define SA_PARENT_OFFSET 40 914 915/* 916 * Intent log header - this on disk structure holds fields to manage 917 * the log. All fields are 64 bit to easily handle cross architectures. 918 */ 919typedef struct zil_header { 920 uint64_t zh_claim_txg; /* txg in which log blocks were claimed */ 921 uint64_t zh_replay_seq; /* highest replayed sequence number */ 922 blkptr_t zh_log; /* log chain */ 923 uint64_t zh_claim_seq; /* highest claimed sequence number */ 924 uint64_t zh_pad[5]; 925} zil_header_t; 926 927#define OBJSET_PHYS_SIZE 2048 928 929typedef struct objset_phys { 930 dnode_phys_t os_meta_dnode; 931 zil_header_t os_zil_header; 932 uint64_t os_type; 933 uint64_t os_flags; 934 char os_pad[OBJSET_PHYS_SIZE - sizeof (dnode_phys_t)*3 - 935 sizeof (zil_header_t) - sizeof (uint64_t)*2]; 936 dnode_phys_t os_userused_dnode; 937 dnode_phys_t os_groupused_dnode; 938} objset_phys_t; 939 940typedef struct dsl_dir_phys { 941 uint64_t dd_creation_time; /* not actually used */ 942 uint64_t dd_head_dataset_obj; 943 uint64_t dd_parent_obj; 944 uint64_t dd_clone_parent_obj; 945 uint64_t dd_child_dir_zapobj; 946 /* 947 * how much space our children are accounting for; for leaf 948 * datasets, == physical space used by fs + snaps 949 */ 950 uint64_t dd_used_bytes; 951 uint64_t dd_compressed_bytes; 952 uint64_t dd_uncompressed_bytes; 953 /* Administrative quota setting */ 954 uint64_t dd_quota; 955 /* Administrative reservation setting */ 956 uint64_t dd_reserved; 957 uint64_t dd_props_zapobj; 958 uint64_t dd_pad[21]; /* pad out to 256 bytes for good measure */ 959} dsl_dir_phys_t; 960 961typedef struct dsl_dataset_phys { 962 uint64_t ds_dir_obj; 963 uint64_t ds_prev_snap_obj; 964 uint64_t ds_prev_snap_txg; 965 uint64_t ds_next_snap_obj; 966 uint64_t ds_snapnames_zapobj; /* zap obj of snaps; ==0 for snaps */ 967 uint64_t ds_num_children; /* clone/snap children; ==0 for head */ 968 uint64_t ds_creation_time; /* seconds since 1970 */ 969 uint64_t ds_creation_txg; 970 uint64_t ds_deadlist_obj; 971 uint64_t ds_used_bytes; 972 uint64_t ds_compressed_bytes; 973 uint64_t ds_uncompressed_bytes; 974 uint64_t ds_unique_bytes; /* only relevant to snapshots */ 975 /* 976 * The ds_fsid_guid is a 56-bit ID that can change to avoid 977 * collisions. The ds_guid is a 64-bit ID that will never 978 * change, so there is a small probability that it will collide. 979 */ 980 uint64_t ds_fsid_guid; 981 uint64_t ds_guid; 982 uint64_t ds_flags; 983 blkptr_t ds_bp; 984 uint64_t ds_pad[8]; /* pad out to 320 bytes for good measure */ 985} dsl_dataset_phys_t; 986 987/* 988 * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 989 */ 990#define DMU_POOL_DIRECTORY_OBJECT 1 991#define DMU_POOL_CONFIG "config" 992#define DMU_POOL_ROOT_DATASET "root_dataset" 993#define DMU_POOL_SYNC_BPLIST "sync_bplist" 994#define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 995#define DMU_POOL_ERRLOG_LAST "errlog_last" 996#define DMU_POOL_SPARES "spares" 997#define DMU_POOL_DEFLATE "deflate" 998#define DMU_POOL_HISTORY "history" 999#define DMU_POOL_PROPS "pool_props" 1000 1001#define ZAP_MAGIC 0x2F52AB2ABULL 1002 1003#define FZAP_BLOCK_SHIFT(zap) ((zap)->zap_block_shift) 1004 1005#define ZAP_MAXCD (uint32_t)(-1) 1006#define ZAP_HASHBITS 28 1007#define MZAP_ENT_LEN 64 1008#define MZAP_NAME_LEN (MZAP_ENT_LEN - 8 - 4 - 2) 1009#define MZAP_MAX_BLKSHIFT SPA_MAXBLOCKSHIFT 1010#define MZAP_MAX_BLKSZ (1 << MZAP_MAX_BLKSHIFT) 1011 1012typedef struct mzap_ent_phys { 1013 uint64_t mze_value; 1014 uint32_t mze_cd; 1015 uint16_t mze_pad; /* in case we want to chain them someday */ 1016 char mze_name[MZAP_NAME_LEN]; 1017} mzap_ent_phys_t; 1018 1019typedef struct mzap_phys { 1020 uint64_t mz_block_type; /* ZBT_MICRO */ 1021 uint64_t mz_salt; 1022 uint64_t mz_pad[6]; 1023 mzap_ent_phys_t mz_chunk[1]; 1024 /* actually variable size depending on block size */ 1025} mzap_phys_t; 1026 1027/* 1028 * The (fat) zap is stored in one object. It is an array of 1029 * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of: 1030 * 1031 * ptrtbl fits in first block: 1032 * [zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ... 1033 * 1034 * ptrtbl too big for first block: 1035 * [zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ... 1036 * 1037 */ 1038 1039#define ZBT_LEAF ((1ULL << 63) + 0) 1040#define ZBT_HEADER ((1ULL << 63) + 1) 1041#define ZBT_MICRO ((1ULL << 63) + 3) 1042/* any other values are ptrtbl blocks */ 1043 1044/* 1045 * the embedded pointer table takes up half a block: 1046 * block size / entry size (2^3) / 2 1047 */ 1048#define ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1) 1049 1050/* 1051 * The embedded pointer table starts half-way through the block. Since 1052 * the pointer table itself is half the block, it starts at (64-bit) 1053 * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)). 1054 */ 1055#define ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \ 1056 ((uint64_t *)(zap)->zap_phys) \ 1057 [(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))] 1058 1059/* 1060 * TAKE NOTE: 1061 * If zap_phys_t is modified, zap_byteswap() must be modified. 1062 */ 1063typedef struct zap_phys { 1064 uint64_t zap_block_type; /* ZBT_HEADER */ 1065 uint64_t zap_magic; /* ZAP_MAGIC */ 1066 1067 struct zap_table_phys { 1068 uint64_t zt_blk; /* starting block number */ 1069 uint64_t zt_numblks; /* number of blocks */ 1070 uint64_t zt_shift; /* bits to index it */ 1071 uint64_t zt_nextblk; /* next (larger) copy start block */ 1072 uint64_t zt_blks_copied; /* number source blocks copied */ 1073 } zap_ptrtbl; 1074 1075 uint64_t zap_freeblk; /* the next free block */ 1076 uint64_t zap_num_leafs; /* number of leafs */ 1077 uint64_t zap_num_entries; /* number of entries */ 1078 uint64_t zap_salt; /* salt to stir into hash function */ 1079 /* 1080 * This structure is followed by padding, and then the embedded 1081 * pointer table. The embedded pointer table takes up second 1082 * half of the block. It is accessed using the 1083 * ZAP_EMBEDDED_PTRTBL_ENT() macro. 1084 */ 1085} zap_phys_t; 1086 1087typedef struct zap_table_phys zap_table_phys_t; 1088 1089typedef struct fat_zap { 1090 int zap_block_shift; /* block size shift */ 1091 zap_phys_t *zap_phys; 1092} fat_zap_t; 1093 1094#define ZAP_LEAF_MAGIC 0x2AB1EAF 1095 1096/* chunk size = 24 bytes */ 1097#define ZAP_LEAF_CHUNKSIZE 24 1098 1099/* 1100 * The amount of space available for chunks is: 1101 * block size (1<<l->l_bs) - hash entry size (2) * number of hash 1102 * entries - header space (2*chunksize) 1103 */ 1104#define ZAP_LEAF_NUMCHUNKS(l) \ 1105 (((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \ 1106 ZAP_LEAF_CHUNKSIZE - 2) 1107 1108/* 1109 * The amount of space within the chunk available for the array is: 1110 * chunk size - space for type (1) - space for next pointer (2) 1111 */ 1112#define ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3) 1113 1114#define ZAP_LEAF_ARRAY_NCHUNKS(bytes) \ 1115 (((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES) 1116 1117/* 1118 * Low water mark: when there are only this many chunks free, start 1119 * growing the ptrtbl. Ideally, this should be larger than a 1120 * "reasonably-sized" entry. 20 chunks is more than enough for the 1121 * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value), 1122 * while still being only around 3% for 16k blocks. 1123 */ 1124#define ZAP_LEAF_LOW_WATER (20) 1125 1126/* 1127 * The leaf hash table has block size / 2^5 (32) number of entries, 1128 * which should be more than enough for the maximum number of entries, 1129 * which is less than block size / CHUNKSIZE (24) / minimum number of 1130 * chunks per entry (3). 1131 */ 1132#define ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5) 1133#define ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l)) 1134 1135/* 1136 * The chunks start immediately after the hash table. The end of the 1137 * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a 1138 * chunk_t. 1139 */ 1140#define ZAP_LEAF_CHUNK(l, idx) \ 1141 ((zap_leaf_chunk_t *) \ 1142 ((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx] 1143#define ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry) 1144 1145typedef enum zap_chunk_type { 1146 ZAP_CHUNK_FREE = 253, 1147 ZAP_CHUNK_ENTRY = 252, 1148 ZAP_CHUNK_ARRAY = 251, 1149 ZAP_CHUNK_TYPE_MAX = 250 1150} zap_chunk_type_t; 1151 1152/* 1153 * TAKE NOTE: 1154 * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified. 1155 */ 1156typedef struct zap_leaf_phys { 1157 struct zap_leaf_header { 1158 uint64_t lh_block_type; /* ZBT_LEAF */ 1159 uint64_t lh_pad1; 1160 uint64_t lh_prefix; /* hash prefix of this leaf */ 1161 uint32_t lh_magic; /* ZAP_LEAF_MAGIC */ 1162 uint16_t lh_nfree; /* number free chunks */ 1163 uint16_t lh_nentries; /* number of entries */ 1164 uint16_t lh_prefix_len; /* num bits used to id this */ 1165 1166/* above is accessable to zap, below is zap_leaf private */ 1167 1168 uint16_t lh_freelist; /* chunk head of free list */ 1169 uint8_t lh_pad2[12]; 1170 } l_hdr; /* 2 24-byte chunks */ 1171 1172 /* 1173 * The header is followed by a hash table with 1174 * ZAP_LEAF_HASH_NUMENTRIES(zap) entries. The hash table is 1175 * followed by an array of ZAP_LEAF_NUMCHUNKS(zap) 1176 * zap_leaf_chunk structures. These structures are accessed 1177 * with the ZAP_LEAF_CHUNK() macro. 1178 */ 1179 1180 uint16_t l_hash[1]; 1181} zap_leaf_phys_t; 1182 1183typedef union zap_leaf_chunk { 1184 struct zap_leaf_entry { 1185 uint8_t le_type; /* always ZAP_CHUNK_ENTRY */ 1186 uint8_t le_value_intlen; /* size of ints */ 1187 uint16_t le_next; /* next entry in hash chain */ 1188 uint16_t le_name_chunk; /* first chunk of the name */ 1189 uint16_t le_name_numints; /* bytes in name, incl null */ 1190 uint16_t le_value_chunk; /* first chunk of the value */ 1191 uint16_t le_value_numints; /* value length in ints */ 1192 uint32_t le_cd; /* collision differentiator */ 1193 uint64_t le_hash; /* hash value of the name */ 1194 } l_entry; 1195 struct zap_leaf_array { 1196 uint8_t la_type; /* always ZAP_CHUNK_ARRAY */ 1197 uint8_t la_array[ZAP_LEAF_ARRAY_BYTES]; 1198 uint16_t la_next; /* next blk or CHAIN_END */ 1199 } l_array; 1200 struct zap_leaf_free { 1201 uint8_t lf_type; /* always ZAP_CHUNK_FREE */ 1202 uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES]; 1203 uint16_t lf_next; /* next in free list, or CHAIN_END */ 1204 } l_free; 1205} zap_leaf_chunk_t; 1206 1207typedef struct zap_leaf { 1208 int l_bs; /* block size shift */ 1209 zap_leaf_phys_t *l_phys; 1210} zap_leaf_t; 1211 1212/* 1213 * Define special zfs pflags 1214 */ 1215#define ZFS_XATTR 0x1 /* is an extended attribute */ 1216#define ZFS_INHERIT_ACE 0x2 /* ace has inheritable ACEs */ 1217#define ZFS_ACL_TRIVIAL 0x4 /* files ACL is trivial */ 1218 1219#define MASTER_NODE_OBJ 1 1220 1221/* 1222 * special attributes for master node. 1223 */ 1224 1225#define ZFS_FSID "FSID" 1226#define ZFS_UNLINKED_SET "DELETE_QUEUE" 1227#define ZFS_ROOT_OBJ "ROOT" 1228#define ZPL_VERSION_OBJ "VERSION" 1229#define ZFS_PROP_BLOCKPERPAGE "BLOCKPERPAGE" 1230#define ZFS_PROP_NOGROWBLOCKS "NOGROWBLOCKS" 1231 1232#define ZFS_FLAG_BLOCKPERPAGE 0x1 1233#define ZFS_FLAG_NOGROWBLOCKS 0x2 1234 1235/* 1236 * ZPL version - rev'd whenever an incompatible on-disk format change 1237 * occurs. Independent of SPA/DMU/ZAP versioning. 1238 */ 1239 1240#define ZPL_VERSION 1ULL 1241 1242/* 1243 * The directory entry has the type (currently unused on Solaris) in the 1244 * top 4 bits, and the object number in the low 48 bits. The "middle" 1245 * 12 bits are unused. 1246 */ 1247#define ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4) 1248#define ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48) 1249#define ZFS_DIRENT_MAKE(type, obj) (((uint64_t)type << 60) | obj) 1250 1251typedef struct ace { 1252 uid_t a_who; /* uid or gid */ 1253 uint32_t a_access_mask; /* read,write,... */ 1254 uint16_t a_flags; /* see below */ 1255 uint16_t a_type; /* allow or deny */ 1256} ace_t; 1257 1258#define ACE_SLOT_CNT 6 1259 1260typedef struct zfs_znode_acl { 1261 uint64_t z_acl_extern_obj; /* ext acl pieces */ 1262 uint32_t z_acl_count; /* Number of ACEs */ 1263 uint16_t z_acl_version; /* acl version */ 1264 uint16_t z_acl_pad; /* pad */ 1265 ace_t z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */ 1266} zfs_znode_acl_t; 1267 1268/* 1269 * This is the persistent portion of the znode. It is stored 1270 * in the "bonus buffer" of the file. Short symbolic links 1271 * are also stored in the bonus buffer. 1272 */ 1273typedef struct znode_phys { 1274 uint64_t zp_atime[2]; /* 0 - last file access time */ 1275 uint64_t zp_mtime[2]; /* 16 - last file modification time */ 1276 uint64_t zp_ctime[2]; /* 32 - last file change time */ 1277 uint64_t zp_crtime[2]; /* 48 - creation time */ 1278 uint64_t zp_gen; /* 64 - generation (txg of creation) */ 1279 uint64_t zp_mode; /* 72 - file mode bits */ 1280 uint64_t zp_size; /* 80 - size of file */ 1281 uint64_t zp_parent; /* 88 - directory parent (`..') */ 1282 uint64_t zp_links; /* 96 - number of links to file */ 1283 uint64_t zp_xattr; /* 104 - DMU object for xattrs */ 1284 uint64_t zp_rdev; /* 112 - dev_t for VBLK & VCHR files */ 1285 uint64_t zp_flags; /* 120 - persistent flags */ 1286 uint64_t zp_uid; /* 128 - file owner */ 1287 uint64_t zp_gid; /* 136 - owning group */ 1288 uint64_t zp_pad[4]; /* 144 - future */ 1289 zfs_znode_acl_t zp_acl; /* 176 - 263 ACL */ 1290 /* 1291 * Data may pad out any remaining bytes in the znode buffer, eg: 1292 * 1293 * |<---------------------- dnode_phys (512) ------------------------>| 1294 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->| 1295 * |<---- znode (264) ---->|<---- data (56) ---->| 1296 * 1297 * At present, we only use this space to store symbolic links. 1298 */ 1299} znode_phys_t; 1300 1301/* 1302 * In-core vdev representation. 1303 */ 1304struct vdev; 1305typedef int vdev_phys_read_t(struct vdev *vdev, void *priv, 1306 off_t offset, void *buf, size_t bytes); 1307typedef int vdev_read_t(struct vdev *vdev, const blkptr_t *bp, 1308 void *buf, off_t offset, size_t bytes); 1309 1310typedef STAILQ_HEAD(vdev_list, vdev) vdev_list_t; 1311 1312typedef struct vdev { 1313 STAILQ_ENTRY(vdev) v_childlink; /* link in parent's child list */ 1314 STAILQ_ENTRY(vdev) v_alllink; /* link in global vdev list */ 1315 vdev_list_t v_children; /* children of this vdev */ 1316 const char *v_name; /* vdev name */ 1317 uint64_t v_guid; /* vdev guid */ 1318 int v_id; /* index in parent */ 1319 int v_ashift; /* offset to block shift */ 1320 int v_nparity; /* # parity for raidz */ 1321 struct vdev *v_top; /* parent vdev */ 1322 int v_nchildren; /* # children */ 1323 vdev_state_t v_state; /* current state */ 1324 vdev_phys_read_t *v_phys_read; /* read from raw leaf vdev */ 1325 vdev_read_t *v_read; /* read from vdev */ 1326 void *v_read_priv; /* private data for read function */ 1327} vdev_t; 1328 1329/* 1330 * In-core pool representation. 1331 */ 1332typedef STAILQ_HEAD(spa_list, spa) spa_list_t; 1333 1334typedef struct spa { 1335 STAILQ_ENTRY(spa) spa_link; /* link in global pool list */ 1336 char *spa_name; /* pool name */ 1337 uint64_t spa_guid; /* pool guid */ 1338 uint64_t spa_txg; /* most recent transaction */ 1339 struct uberblock spa_uberblock; /* best uberblock so far */ 1340 vdev_list_t spa_vdevs; /* list of all toplevel vdevs */ 1341 objset_phys_t spa_mos; /* MOS for this pool */ 1342 int spa_inited; /* initialized */ 1343} spa_t; 1344