zfsimpl.h revision 328866
1251881Speter/*- 2251881Speter * Copyright (c) 2002 McAfee, Inc. 3251881Speter * All rights reserved. 4251881Speter * 5251881Speter * This software was developed for the FreeBSD Project by Marshall 6251881Speter * Kirk McKusick and McAfee Research,, the Security Research Division of 7251881Speter * McAfee, Inc. under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as 8251881Speter * part of the DARPA CHATS research program 9251881Speter * 10251881Speter * Redistribution and use in source and binary forms, with or without 11251881Speter * modification, are permitted provided that the following conditions 12251881Speter * are met: 13251881Speter * 1. Redistributions of source code must retain the above copyright 14251881Speter * notice, this list of conditions and the following disclaimer. 15251881Speter * 2. Redistributions in binary form must reproduce the above copyright 16251881Speter * notice, this list of conditions and the following disclaimer in the 17251881Speter * documentation and/or other materials provided with the distribution. 18251881Speter * 19251881Speter * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 20251881Speter * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21251881Speter * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22251881Speter * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 23251881Speter * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24251881Speter * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25251881Speter * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26251881Speter * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27251881Speter * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28251881Speter * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29251881Speter * SUCH DAMAGE. 30251881Speter */ 31251881Speter/* 32251881Speter * CDDL HEADER START 33251881Speter * 34251881Speter * The contents of this file are subject to the terms of the 35251881Speter * Common Development and Distribution License (the "License"). 36251881Speter * You may not use this file except in compliance with the License. 37251881Speter * 38251881Speter * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 39251881Speter * or http://www.opensolaris.org/os/licensing. 40251881Speter * See the License for the specific language governing permissions 41251881Speter * and limitations under the License. 42251881Speter * 43251881Speter * When distributing Covered Code, include this CDDL HEADER in each 44251881Speter * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 45251881Speter * If applicable, add the following below this CDDL HEADER, with the 46251881Speter * fields enclosed by brackets "[]" replaced with your own identifying 47251881Speter * information: Portions Copyright [yyyy] [name of copyright owner] 48251881Speter * 49251881Speter * CDDL HEADER END 50251881Speter */ 51251881Speter/* 52251881Speter * Copyright 2009 Sun Microsystems, Inc. All rights reserved. 53251881Speter * Use is subject to license terms. 54251881Speter */ 55251881Speter/* 56251881Speter * Copyright 2013 by Saso Kiselkov. All rights reserved. 57251881Speter */ 58251881Speter/* 59289180Speter * Copyright (c) 2013 by Delphix. All rights reserved. 60289180Speter */ 61289180Speter 62289180Speter#define MAXNAMELEN 256 63251881Speter 64251881Speter#define _NOTE(s) 65251881Speter 66251881Spetertypedef enum { B_FALSE, B_TRUE } boolean_t; 67251881Speter 68251881Speter/* CRC64 table */ 69251881Speter#define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ 70251881Speter 71251881Speter/* 72251881Speter * Macros for various sorts of alignment and rounding when the alignment 73251881Speter * is known to be a power of 2. 74251881Speter */ 75251881Speter#define P2ALIGN(x, align) ((x) & -(align)) 76251881Speter#define P2PHASE(x, align) ((x) & ((align) - 1)) 77251881Speter#define P2NPHASE(x, align) (-(x) & ((align) - 1)) 78251881Speter#define P2ROUNDUP(x, align) (-(-(x) & -(align))) 79251881Speter#define P2END(x, align) (-(~(x) & -(align))) 80251881Speter#define P2PHASEUP(x, align, phase) ((phase) - (((phase) - (x)) & -(align))) 81251881Speter#define P2BOUNDARY(off, len, align) (((off) ^ ((off) + (len) - 1)) > (align) - 1) 82251881Speter 83251881Speter/* 84251881Speter * General-purpose 32-bit and 64-bit bitfield encodings. 85251881Speter */ 86251881Speter#define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) 87251881Speter#define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) 88251881Speter#define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) 89251881Speter#define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) 90251881Speter 91251881Speter#define BF32_GET(x, low, len) BF32_DECODE(x, low, len) 92251881Speter#define BF64_GET(x, low, len) BF64_DECODE(x, low, len) 93251881Speter 94251881Speter#define BF32_SET(x, low, len, val) \ 95251881Speter ((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len)) 96251881Speter#define BF64_SET(x, low, len, val) \ 97251881Speter ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)) 98251881Speter 99251881Speter#define BF32_GET_SB(x, low, len, shift, bias) \ 100251881Speter ((BF32_GET(x, low, len) + (bias)) << (shift)) 101251881Speter#define BF64_GET_SB(x, low, len, shift, bias) \ 102251881Speter ((BF64_GET(x, low, len) + (bias)) << (shift)) 103251881Speter 104251881Speter#define BF32_SET_SB(x, low, len, shift, bias, val) \ 105251881Speter BF32_SET(x, low, len, ((val) >> (shift)) - (bias)) 106251881Speter#define BF64_SET_SB(x, low, len, shift, bias, val) \ 107251881Speter BF64_SET(x, low, len, ((val) >> (shift)) - (bias)) 108251881Speter 109251881Speter/* 110251881Speter * Macros to reverse byte order 111251881Speter */ 112251881Speter#define BSWAP_8(x) ((x) & 0xff) 113251881Speter#define BSWAP_16(x) ((BSWAP_8(x) << 8) | BSWAP_8((x) >> 8)) 114257936Speter#define BSWAP_32(x) ((BSWAP_16(x) << 16) | BSWAP_16((x) >> 16)) 115251881Speter#define BSWAP_64(x) ((BSWAP_32(x) << 32) | BSWAP_32((x) >> 32)) 116251881Speter 117251881Speter#define SPA_MINBLOCKSHIFT 9 118289180Speter#define SPA_OLDMAXBLOCKSHIFT 17 119289180Speter#define SPA_MAXBLOCKSHIFT 24 120289180Speter#define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 121289180Speter#define SPA_OLDMAXBLOCKSIZE (1ULL << SPA_OLDMAXBLOCKSHIFT) 122251881Speter#define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 123251881Speter 124251881Speter/* 125251881Speter * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. 126289180Speter * The ASIZE encoding should be at least 64 times larger (6 more bits) 127289180Speter * to support up to 4-way RAID-Z mirror mode with worst-case gang block 128251881Speter * overhead, three DVAs per bp, plus one more bit in case we do anything 129251881Speter * else that expands the ASIZE. 130251881Speter */ 131251881Speter#define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ 132251881Speter#define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ 133251881Speter#define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ 134251881Speter 135251881Speter/* 136251881Speter * All SPA data is represented by 128-bit data virtual addresses (DVAs). 137251881Speter * The members of the dva_t should be considered opaque outside the SPA. 138251881Speter */ 139251881Spetertypedef struct dva { 140251881Speter uint64_t dva_word[2]; 141251881Speter} dva_t; 142251881Speter 143251881Speter/* 144251881Speter * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. 145251881Speter */ 146251881Spetertypedef struct zio_cksum { 147251881Speter uint64_t zc_word[4]; 148251881Speter} zio_cksum_t; 149251881Speter 150251881Speter/* 151289180Speter * Some checksums/hashes need a 256-bit initialization salt. This salt is kept 152251881Speter * secret and is suitable for use in MAC algorithms as the key. 153251881Speter */ 154251881Spetertypedef struct zio_cksum_salt { 155251881Speter uint8_t zcs_bytes[32]; 156251881Speter} zio_cksum_salt_t; 157289180Speter 158251881Speter/* 159251881Speter * Each block is described by its DVAs, time of birth, checksum, etc. 160251881Speter * The word-by-word, bit-by-bit layout of the blkptr is as follows: 161289180Speter * 162251881Speter * 64 56 48 40 32 24 16 8 0 163251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 164251881Speter * 0 | vdev1 | GRID | ASIZE | 165251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 166251881Speter * 1 |G| offset1 | 167251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 168251881Speter * 2 | vdev2 | GRID | ASIZE | 169251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 170251881Speter * 3 |G| offset2 | 171251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 172289180Speter * 4 | vdev3 | GRID | ASIZE | 173289180Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 174289180Speter * 5 |G| offset3 | 175289180Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 176289180Speter * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 177251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 178251881Speter * 7 | padding | 179251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 180251881Speter * 8 | padding | 181251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 182251881Speter * 9 | physical birth txg | 183251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 184251881Speter * a | logical birth txg | 185251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 186251881Speter * b | fill count | 187251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 188251881Speter * c | checksum[0] | 189251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 190251881Speter * d | checksum[1] | 191289180Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 192251881Speter * e | checksum[2] | 193251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 194251881Speter * f | checksum[3] | 195251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 196289180Speter * 197289180Speter * Legend: 198289180Speter * 199251881Speter * vdev virtual device ID 200251881Speter * offset offset into virtual device 201251881Speter * LSIZE logical size 202251881Speter * PSIZE physical size (after compression) 203251881Speter * ASIZE allocated size (including RAID-Z parity and gang block headers) 204289180Speter * GRID RAID-Z layout information (reserved for future use) 205289180Speter * cksum checksum function 206251881Speter * comp compression function 207251881Speter * G gang block indicator 208289180Speter * B byteorder (endianness) 209289180Speter * D dedup 210289180Speter * X encryption (on version 30, which is not supported) 211289180Speter * E blkptr_t contains embedded data (see below) 212289180Speter * lvl level of indirection 213251881Speter * type DMU object type 214289180Speter * phys birth txg of block allocation; zero if same as logical birth txg 215289180Speter * log. birth transaction group in which the block was logically born 216289180Speter * fill count number of non-zero blocks under this bp 217251881Speter * checksum[4] 256-bit checksum of the data this bp describes 218289180Speter */ 219289180Speter 220289180Speter/* 221251881Speter * "Embedded" blkptr_t's don't actually point to a block, instead they 222251881Speter * have a data payload embedded in the blkptr_t itself. See the comment 223289180Speter * in blkptr.c for more details. 224251881Speter * 225251881Speter * The blkptr_t is laid out as follows: 226251881Speter * 227251881Speter * 64 56 48 40 32 24 16 8 0 228251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 229251881Speter * 0 | payload | 230251881Speter * 1 | payload | 231251881Speter * 2 | payload | 232251881Speter * 3 | payload | 233251881Speter * 4 | payload | 234251881Speter * 5 | payload | 235251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 236251881Speter * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 237251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 238289180Speter * 7 | payload | 239251881Speter * 8 | payload | 240251881Speter * 9 | payload | 241289180Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 242289180Speter * a | logical birth txg | 243251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 244289180Speter * b | payload | 245289180Speter * c | payload | 246251881Speter * d | payload | 247289180Speter * e | payload | 248289180Speter * f | payload | 249251881Speter * +-------+-------+-------+-------+-------+-------+-------+-------+ 250251881Speter * 251289180Speter * Legend: 252289180Speter * 253289180Speter * payload contains the embedded data 254289180Speter * B (byteorder) byteorder (endianness) 255289180Speter * D (dedup) padding (set to zero) 256289180Speter * X encryption (set to zero; see above) 257289180Speter * E (embedded) set to one 258289180Speter * lvl indirection level 259251881Speter * type DMU object type 260289180Speter * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 261251881Speter * comp compression function of payload 262251881Speter * PSIZE size of payload after compression, in bytes 263289180Speter * LSIZE logical size of payload, in bytes 264251881Speter * note that 25 bits is enough to store the largest 265251881Speter * "normal" BP's LSIZE (2^16 * 2^9) in bytes 266251881Speter * log. birth transaction group in which the block was logically born 267251881Speter * 268289180Speter * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 269289180Speter * bp's they are stored in units of SPA_MINBLOCKSHIFT. 270289180Speter * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 271289180Speter * The B, D, X, lvl, type, and comp fields are stored the same as with normal 272251881Speter * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 273251881Speter * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 274251881Speter * other macros, as they assert that they are only used on BP's of the correct 275289180Speter * "embedded-ness". 276251881Speter */ 277289180Speter 278251881Speter#define BPE_GET_ETYPE(bp) \ 279289180Speter (ASSERT(BP_IS_EMBEDDED(bp)), \ 280251881Speter BF64_GET((bp)->blk_prop, 40, 8)) 281289180Speter#define BPE_SET_ETYPE(bp, t) do { \ 282251881Speter ASSERT(BP_IS_EMBEDDED(bp)); \ 283289180Speter BF64_SET((bp)->blk_prop, 40, 8, t); \ 284251881Speter_NOTE(CONSTCOND) } while (0) 285251881Speter 286289180Speter#define BPE_GET_LSIZE(bp) \ 287251881Speter (ASSERT(BP_IS_EMBEDDED(bp)), \ 288289180Speter BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 289289180Speter#define BPE_SET_LSIZE(bp, x) do { \ 290251881Speter ASSERT(BP_IS_EMBEDDED(bp)); \ 291289180Speter BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 292289180Speter_NOTE(CONSTCOND) } while (0) 293289180Speter 294289180Speter#define BPE_GET_PSIZE(bp) \ 295289180Speter (ASSERT(BP_IS_EMBEDDED(bp)), \ 296289180Speter BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 297251881Speter#define BPE_SET_PSIZE(bp, x) do { \ 298289180Speter ASSERT(BP_IS_EMBEDDED(bp)); \ 299289180Speter BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 300289180Speter_NOTE(CONSTCOND) } while (0) 301289180Speter 302289180Spetertypedef enum bp_embedded_type { 303289180Speter BP_EMBEDDED_TYPE_DATA, 304289180Speter BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ 305289180Speter NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED 306289180Speter} bp_embedded_type_t; 307289180Speter 308289180Speter#define BPE_NUM_WORDS 14 309289180Speter#define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 310289180Speter#define BPE_IS_PAYLOADWORD(bp, wp) \ 311289180Speter ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 312289180Speter 313289180Speter#define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 314289180Speter#define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 315289180Speter 316289180Spetertypedef struct blkptr { 317289180Speter dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 318289180Speter uint64_t blk_prop; /* size, compression, type, etc */ 319289180Speter uint64_t blk_pad[2]; /* Extra space for the future */ 320289180Speter uint64_t blk_phys_birth; /* txg when block was allocated */ 321251881Speter uint64_t blk_birth; /* transaction group at birth */ 322 uint64_t blk_fill; /* fill count */ 323 zio_cksum_t blk_cksum; /* 256-bit checksum */ 324} blkptr_t; 325 326/* 327 * Macros to get and set fields in a bp or DVA. 328 */ 329#define DVA_GET_ASIZE(dva) \ 330 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 331#define DVA_SET_ASIZE(dva, x) \ 332 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 333 SPA_MINBLOCKSHIFT, 0, x) 334 335#define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 336#define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 337 338#define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) 339#define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) 340 341#define DVA_GET_OFFSET(dva) \ 342 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 343#define DVA_SET_OFFSET(dva, x) \ 344 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 345 346#define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 347#define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 348 349#define BP_GET_LSIZE(bp) \ 350 (BP_IS_EMBEDDED(bp) ? \ 351 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 352 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 353#define BP_SET_LSIZE(bp, x) do { \ 354 ASSERT(!BP_IS_EMBEDDED(bp)); \ 355 BF64_SET_SB((bp)->blk_prop, \ 356 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 357_NOTE(CONSTCOND) } while (0) 358 359#define BP_GET_PSIZE(bp) \ 360 BF64_GET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1) 361#define BP_SET_PSIZE(bp, x) \ 362 BF64_SET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x) 363 364#define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7) 365#define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x) 366 367#define BP_GET_CHECKSUM(bp) BF64_GET((bp)->blk_prop, 40, 8) 368#define BP_SET_CHECKSUM(bp, x) BF64_SET((bp)->blk_prop, 40, 8, x) 369 370#define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 371#define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 372 373#define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 374#define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 375 376#define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 377 378#define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 379#define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 380 381#define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 382#define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 383 384#define BP_PHYSICAL_BIRTH(bp) \ 385 ((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 386 387#define BP_GET_ASIZE(bp) \ 388 (DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 389 DVA_GET_ASIZE(&(bp)->blk_dva[2])) 390 391#define BP_GET_UCSIZE(bp) \ 392 ((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \ 393 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)); 394 395#define BP_GET_NDVAS(bp) \ 396 (!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 397 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 398 !!DVA_GET_ASIZE(&(bp)->blk_dva[2])) 399 400#define DVA_EQUAL(dva1, dva2) \ 401 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 402 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 403 404#define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 405 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 406 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 407 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 408 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 409 410 411#define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 412 413#define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ 414{ \ 415 (zcp)->zc_word[0] = w0; \ 416 (zcp)->zc_word[1] = w1; \ 417 (zcp)->zc_word[2] = w2; \ 418 (zcp)->zc_word[3] = w3; \ 419} 420 421#define BP_IDENTITY(bp) (&(bp)->blk_dva[0]) 422#define BP_IS_GANG(bp) DVA_GET_GANG(BP_IDENTITY(bp)) 423#define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 424 (dva)->dva_word[1] == 0ULL) 425#define BP_IS_HOLE(bp) DVA_IS_EMPTY(BP_IDENTITY(bp)) 426#define BP_IS_OLDER(bp, txg) (!BP_IS_HOLE(bp) && (bp)->blk_birth < (txg)) 427 428#define BP_ZERO(bp) \ 429{ \ 430 (bp)->blk_dva[0].dva_word[0] = 0; \ 431 (bp)->blk_dva[0].dva_word[1] = 0; \ 432 (bp)->blk_dva[1].dva_word[0] = 0; \ 433 (bp)->blk_dva[1].dva_word[1] = 0; \ 434 (bp)->blk_dva[2].dva_word[0] = 0; \ 435 (bp)->blk_dva[2].dva_word[1] = 0; \ 436 (bp)->blk_prop = 0; \ 437 (bp)->blk_pad[0] = 0; \ 438 (bp)->blk_pad[1] = 0; \ 439 (bp)->blk_phys_birth = 0; \ 440 (bp)->blk_birth = 0; \ 441 (bp)->blk_fill = 0; \ 442 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 443} 444 445#define BPE_NUM_WORDS 14 446#define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 447#define BPE_IS_PAYLOADWORD(bp, wp) \ 448 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 449 450/* 451 * Embedded checksum 452 */ 453#define ZEC_MAGIC 0x210da7ab10c7a11ULL 454 455typedef struct zio_eck { 456 uint64_t zec_magic; /* for validation, endianness */ 457 zio_cksum_t zec_cksum; /* 256-bit checksum */ 458} zio_eck_t; 459 460/* 461 * Gang block headers are self-checksumming and contain an array 462 * of block pointers. 463 */ 464#define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE 465#define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \ 466 sizeof (zio_eck_t)) / sizeof (blkptr_t)) 467#define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \ 468 sizeof (zio_eck_t) - \ 469 (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\ 470 sizeof (uint64_t)) 471 472typedef struct zio_gbh { 473 blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS]; 474 uint64_t zg_filler[SPA_GBH_FILLER]; 475 zio_eck_t zg_tail; 476} zio_gbh_phys_t; 477 478#define VDEV_RAIDZ_MAXPARITY 3 479 480#define VDEV_PAD_SIZE (8 << 10) 481/* 2 padding areas (vl_pad1 and vl_pad2) to skip */ 482#define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2 483#define VDEV_PHYS_SIZE (112 << 10) 484#define VDEV_UBERBLOCK_RING (128 << 10) 485 486#define VDEV_UBERBLOCK_SHIFT(vd) \ 487 MAX((vd)->v_top->v_ashift, UBERBLOCK_SHIFT) 488#define VDEV_UBERBLOCK_COUNT(vd) \ 489 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd)) 490#define VDEV_UBERBLOCK_OFFSET(vd, n) \ 491 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)]) 492#define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd)) 493 494typedef struct vdev_phys { 495 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)]; 496 zio_eck_t vp_zbt; 497} vdev_phys_t; 498 499typedef struct vdev_label { 500 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */ 501 char vl_pad2[VDEV_PAD_SIZE]; /* 8K */ 502 vdev_phys_t vl_vdev_phys; /* 112K */ 503 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */ 504} vdev_label_t; /* 256K total */ 505 506/* 507 * vdev_dirty() flags 508 */ 509#define VDD_METASLAB 0x01 510#define VDD_DTL 0x02 511 512/* 513 * Size and offset of embedded boot loader region on each label. 514 * The total size of the first two labels plus the boot area is 4MB. 515 */ 516#define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t)) 517#define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */ 518 519/* 520 * Size of label regions at the start and end of each leaf device. 521 */ 522#define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE) 523#define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t)) 524#define VDEV_LABELS 4 525 526enum zio_checksum { 527 ZIO_CHECKSUM_INHERIT = 0, 528 ZIO_CHECKSUM_ON, 529 ZIO_CHECKSUM_OFF, 530 ZIO_CHECKSUM_LABEL, 531 ZIO_CHECKSUM_GANG_HEADER, 532 ZIO_CHECKSUM_ZILOG, 533 ZIO_CHECKSUM_FLETCHER_2, 534 ZIO_CHECKSUM_FLETCHER_4, 535 ZIO_CHECKSUM_SHA256, 536 ZIO_CHECKSUM_ZILOG2, 537 ZIO_CHECKSUM_NOPARITY, 538 ZIO_CHECKSUM_SHA512, 539 ZIO_CHECKSUM_SKEIN, 540 ZIO_CHECKSUM_EDONR, 541 ZIO_CHECKSUM_FUNCTIONS 542}; 543 544#define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4 545#define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON 546 547enum zio_compress { 548 ZIO_COMPRESS_INHERIT = 0, 549 ZIO_COMPRESS_ON, 550 ZIO_COMPRESS_OFF, 551 ZIO_COMPRESS_LZJB, 552 ZIO_COMPRESS_EMPTY, 553 ZIO_COMPRESS_GZIP_1, 554 ZIO_COMPRESS_GZIP_2, 555 ZIO_COMPRESS_GZIP_3, 556 ZIO_COMPRESS_GZIP_4, 557 ZIO_COMPRESS_GZIP_5, 558 ZIO_COMPRESS_GZIP_6, 559 ZIO_COMPRESS_GZIP_7, 560 ZIO_COMPRESS_GZIP_8, 561 ZIO_COMPRESS_GZIP_9, 562 ZIO_COMPRESS_ZLE, 563 ZIO_COMPRESS_LZ4, 564 ZIO_COMPRESS_FUNCTIONS 565}; 566 567#define ZIO_COMPRESS_ON_VALUE ZIO_COMPRESS_LZJB 568#define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF 569 570/* nvlist pack encoding */ 571#define NV_ENCODE_NATIVE 0 572#define NV_ENCODE_XDR 1 573 574typedef enum { 575 DATA_TYPE_UNKNOWN = 0, 576 DATA_TYPE_BOOLEAN, 577 DATA_TYPE_BYTE, 578 DATA_TYPE_INT16, 579 DATA_TYPE_UINT16, 580 DATA_TYPE_INT32, 581 DATA_TYPE_UINT32, 582 DATA_TYPE_INT64, 583 DATA_TYPE_UINT64, 584 DATA_TYPE_STRING, 585 DATA_TYPE_BYTE_ARRAY, 586 DATA_TYPE_INT16_ARRAY, 587 DATA_TYPE_UINT16_ARRAY, 588 DATA_TYPE_INT32_ARRAY, 589 DATA_TYPE_UINT32_ARRAY, 590 DATA_TYPE_INT64_ARRAY, 591 DATA_TYPE_UINT64_ARRAY, 592 DATA_TYPE_STRING_ARRAY, 593 DATA_TYPE_HRTIME, 594 DATA_TYPE_NVLIST, 595 DATA_TYPE_NVLIST_ARRAY, 596 DATA_TYPE_BOOLEAN_VALUE, 597 DATA_TYPE_INT8, 598 DATA_TYPE_UINT8, 599 DATA_TYPE_BOOLEAN_ARRAY, 600 DATA_TYPE_INT8_ARRAY, 601 DATA_TYPE_UINT8_ARRAY 602} data_type_t; 603 604/* 605 * On-disk version number. 606 */ 607#define SPA_VERSION_1 1ULL 608#define SPA_VERSION_2 2ULL 609#define SPA_VERSION_3 3ULL 610#define SPA_VERSION_4 4ULL 611#define SPA_VERSION_5 5ULL 612#define SPA_VERSION_6 6ULL 613#define SPA_VERSION_7 7ULL 614#define SPA_VERSION_8 8ULL 615#define SPA_VERSION_9 9ULL 616#define SPA_VERSION_10 10ULL 617#define SPA_VERSION_11 11ULL 618#define SPA_VERSION_12 12ULL 619#define SPA_VERSION_13 13ULL 620#define SPA_VERSION_14 14ULL 621#define SPA_VERSION_15 15ULL 622#define SPA_VERSION_16 16ULL 623#define SPA_VERSION_17 17ULL 624#define SPA_VERSION_18 18ULL 625#define SPA_VERSION_19 19ULL 626#define SPA_VERSION_20 20ULL 627#define SPA_VERSION_21 21ULL 628#define SPA_VERSION_22 22ULL 629#define SPA_VERSION_23 23ULL 630#define SPA_VERSION_24 24ULL 631#define SPA_VERSION_25 25ULL 632#define SPA_VERSION_26 26ULL 633#define SPA_VERSION_27 27ULL 634#define SPA_VERSION_28 28ULL 635#define SPA_VERSION_5000 5000ULL 636 637/* 638 * When bumping up SPA_VERSION, make sure GRUB ZFS understands the on-disk 639 * format change. Go to usr/src/grub/grub-0.97/stage2/{zfs-include/, fsys_zfs*}, 640 * and do the appropriate changes. Also bump the version number in 641 * usr/src/grub/capability. 642 */ 643#define SPA_VERSION SPA_VERSION_5000 644#define SPA_VERSION_STRING "5000" 645 646/* 647 * Symbolic names for the changes that caused a SPA_VERSION switch. 648 * Used in the code when checking for presence or absence of a feature. 649 * Feel free to define multiple symbolic names for each version if there 650 * were multiple changes to on-disk structures during that version. 651 * 652 * NOTE: When checking the current SPA_VERSION in your code, be sure 653 * to use spa_version() since it reports the version of the 654 * last synced uberblock. Checking the in-flight version can 655 * be dangerous in some cases. 656 */ 657#define SPA_VERSION_INITIAL SPA_VERSION_1 658#define SPA_VERSION_DITTO_BLOCKS SPA_VERSION_2 659#define SPA_VERSION_SPARES SPA_VERSION_3 660#define SPA_VERSION_RAID6 SPA_VERSION_3 661#define SPA_VERSION_BPLIST_ACCOUNT SPA_VERSION_3 662#define SPA_VERSION_RAIDZ_DEFLATE SPA_VERSION_3 663#define SPA_VERSION_DNODE_BYTES SPA_VERSION_3 664#define SPA_VERSION_ZPOOL_HISTORY SPA_VERSION_4 665#define SPA_VERSION_GZIP_COMPRESSION SPA_VERSION_5 666#define SPA_VERSION_BOOTFS SPA_VERSION_6 667#define SPA_VERSION_SLOGS SPA_VERSION_7 668#define SPA_VERSION_DELEGATED_PERMS SPA_VERSION_8 669#define SPA_VERSION_FUID SPA_VERSION_9 670#define SPA_VERSION_REFRESERVATION SPA_VERSION_9 671#define SPA_VERSION_REFQUOTA SPA_VERSION_9 672#define SPA_VERSION_UNIQUE_ACCURATE SPA_VERSION_9 673#define SPA_VERSION_L2CACHE SPA_VERSION_10 674#define SPA_VERSION_NEXT_CLONES SPA_VERSION_11 675#define SPA_VERSION_ORIGIN SPA_VERSION_11 676#define SPA_VERSION_DSL_SCRUB SPA_VERSION_11 677#define SPA_VERSION_SNAP_PROPS SPA_VERSION_12 678#define SPA_VERSION_USED_BREAKDOWN SPA_VERSION_13 679#define SPA_VERSION_PASSTHROUGH_X SPA_VERSION_14 680#define SPA_VERSION_USERSPACE SPA_VERSION_15 681#define SPA_VERSION_STMF_PROP SPA_VERSION_16 682#define SPA_VERSION_RAIDZ3 SPA_VERSION_17 683#define SPA_VERSION_USERREFS SPA_VERSION_18 684#define SPA_VERSION_HOLES SPA_VERSION_19 685#define SPA_VERSION_ZLE_COMPRESSION SPA_VERSION_20 686#define SPA_VERSION_DEDUP SPA_VERSION_21 687#define SPA_VERSION_RECVD_PROPS SPA_VERSION_22 688#define SPA_VERSION_SLIM_ZIL SPA_VERSION_23 689#define SPA_VERSION_SA SPA_VERSION_24 690#define SPA_VERSION_SCAN SPA_VERSION_25 691#define SPA_VERSION_DIR_CLONES SPA_VERSION_26 692#define SPA_VERSION_DEADLISTS SPA_VERSION_26 693#define SPA_VERSION_FAST_SNAP SPA_VERSION_27 694#define SPA_VERSION_MULTI_REPLACE SPA_VERSION_28 695#define SPA_VERSION_BEFORE_FEATURES SPA_VERSION_28 696#define SPA_VERSION_FEATURES SPA_VERSION_5000 697 698#define SPA_VERSION_IS_SUPPORTED(v) \ 699 (((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \ 700 ((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION)) 701 702/* 703 * The following are configuration names used in the nvlist describing a pool's 704 * configuration. 705 */ 706#define ZPOOL_CONFIG_VERSION "version" 707#define ZPOOL_CONFIG_POOL_NAME "name" 708#define ZPOOL_CONFIG_POOL_STATE "state" 709#define ZPOOL_CONFIG_POOL_TXG "txg" 710#define ZPOOL_CONFIG_POOL_GUID "pool_guid" 711#define ZPOOL_CONFIG_CREATE_TXG "create_txg" 712#define ZPOOL_CONFIG_TOP_GUID "top_guid" 713#define ZPOOL_CONFIG_VDEV_TREE "vdev_tree" 714#define ZPOOL_CONFIG_TYPE "type" 715#define ZPOOL_CONFIG_CHILDREN "children" 716#define ZPOOL_CONFIG_ID "id" 717#define ZPOOL_CONFIG_GUID "guid" 718#define ZPOOL_CONFIG_PATH "path" 719#define ZPOOL_CONFIG_DEVID "devid" 720#define ZPOOL_CONFIG_METASLAB_ARRAY "metaslab_array" 721#define ZPOOL_CONFIG_METASLAB_SHIFT "metaslab_shift" 722#define ZPOOL_CONFIG_ASHIFT "ashift" 723#define ZPOOL_CONFIG_ASIZE "asize" 724#define ZPOOL_CONFIG_DTL "DTL" 725#define ZPOOL_CONFIG_STATS "stats" 726#define ZPOOL_CONFIG_WHOLE_DISK "whole_disk" 727#define ZPOOL_CONFIG_ERRCOUNT "error_count" 728#define ZPOOL_CONFIG_NOT_PRESENT "not_present" 729#define ZPOOL_CONFIG_SPARES "spares" 730#define ZPOOL_CONFIG_IS_SPARE "is_spare" 731#define ZPOOL_CONFIG_NPARITY "nparity" 732#define ZPOOL_CONFIG_HOSTID "hostid" 733#define ZPOOL_CONFIG_HOSTNAME "hostname" 734#define ZPOOL_CONFIG_IS_LOG "is_log" 735#define ZPOOL_CONFIG_TIMESTAMP "timestamp" /* not stored on disk */ 736#define ZPOOL_CONFIG_FEATURES_FOR_READ "features_for_read" 737 738/* 739 * The persistent vdev state is stored as separate values rather than a single 740 * 'vdev_state' entry. This is because a device can be in multiple states, such 741 * as offline and degraded. 742 */ 743#define ZPOOL_CONFIG_OFFLINE "offline" 744#define ZPOOL_CONFIG_FAULTED "faulted" 745#define ZPOOL_CONFIG_DEGRADED "degraded" 746#define ZPOOL_CONFIG_REMOVED "removed" 747#define ZPOOL_CONFIG_FRU "fru" 748#define ZPOOL_CONFIG_AUX_STATE "aux_state" 749 750#define VDEV_TYPE_ROOT "root" 751#define VDEV_TYPE_MIRROR "mirror" 752#define VDEV_TYPE_REPLACING "replacing" 753#define VDEV_TYPE_RAIDZ "raidz" 754#define VDEV_TYPE_DISK "disk" 755#define VDEV_TYPE_FILE "file" 756#define VDEV_TYPE_MISSING "missing" 757#define VDEV_TYPE_HOLE "hole" 758#define VDEV_TYPE_SPARE "spare" 759#define VDEV_TYPE_LOG "log" 760#define VDEV_TYPE_L2CACHE "l2cache" 761 762/* 763 * This is needed in userland to report the minimum necessary device size. 764 */ 765#define SPA_MINDEVSIZE (64ULL << 20) 766 767/* 768 * The location of the pool configuration repository, shared between kernel and 769 * userland. 770 */ 771#define ZPOOL_CACHE "/boot/zfs/zpool.cache" 772 773/* 774 * vdev states are ordered from least to most healthy. 775 * A vdev that's CANT_OPEN or below is considered unusable. 776 */ 777typedef enum vdev_state { 778 VDEV_STATE_UNKNOWN = 0, /* Uninitialized vdev */ 779 VDEV_STATE_CLOSED, /* Not currently open */ 780 VDEV_STATE_OFFLINE, /* Not allowed to open */ 781 VDEV_STATE_REMOVED, /* Explicitly removed from system */ 782 VDEV_STATE_CANT_OPEN, /* Tried to open, but failed */ 783 VDEV_STATE_FAULTED, /* External request to fault device */ 784 VDEV_STATE_DEGRADED, /* Replicated vdev with unhealthy kids */ 785 VDEV_STATE_HEALTHY /* Presumed good */ 786} vdev_state_t; 787 788/* 789 * vdev aux states. When a vdev is in the CANT_OPEN state, the aux field 790 * of the vdev stats structure uses these constants to distinguish why. 791 */ 792typedef enum vdev_aux { 793 VDEV_AUX_NONE, /* no error */ 794 VDEV_AUX_OPEN_FAILED, /* ldi_open_*() or vn_open() failed */ 795 VDEV_AUX_CORRUPT_DATA, /* bad label or disk contents */ 796 VDEV_AUX_NO_REPLICAS, /* insufficient number of replicas */ 797 VDEV_AUX_BAD_GUID_SUM, /* vdev guid sum doesn't match */ 798 VDEV_AUX_TOO_SMALL, /* vdev size is too small */ 799 VDEV_AUX_BAD_LABEL, /* the label is OK but invalid */ 800 VDEV_AUX_VERSION_NEWER, /* on-disk version is too new */ 801 VDEV_AUX_VERSION_OLDER, /* on-disk version is too old */ 802 VDEV_AUX_SPARED /* hot spare used in another pool */ 803} vdev_aux_t; 804 805/* 806 * pool state. The following states are written to disk as part of the normal 807 * SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE. The remaining states are 808 * software abstractions used at various levels to communicate pool state. 809 */ 810typedef enum pool_state { 811 POOL_STATE_ACTIVE = 0, /* In active use */ 812 POOL_STATE_EXPORTED, /* Explicitly exported */ 813 POOL_STATE_DESTROYED, /* Explicitly destroyed */ 814 POOL_STATE_SPARE, /* Reserved for hot spare use */ 815 POOL_STATE_UNINITIALIZED, /* Internal spa_t state */ 816 POOL_STATE_UNAVAIL, /* Internal libzfs state */ 817 POOL_STATE_POTENTIALLY_ACTIVE /* Internal libzfs state */ 818} pool_state_t; 819 820/* 821 * The uberblock version is incremented whenever an incompatible on-disk 822 * format change is made to the SPA, DMU, or ZAP. 823 * 824 * Note: the first two fields should never be moved. When a storage pool 825 * is opened, the uberblock must be read off the disk before the version 826 * can be checked. If the ub_version field is moved, we may not detect 827 * version mismatch. If the ub_magic field is moved, applications that 828 * expect the magic number in the first word won't work. 829 */ 830#define UBERBLOCK_MAGIC 0x00bab10c /* oo-ba-bloc! */ 831#define UBERBLOCK_SHIFT 10 /* up to 1K */ 832 833struct uberblock { 834 uint64_t ub_magic; /* UBERBLOCK_MAGIC */ 835 uint64_t ub_version; /* SPA_VERSION */ 836 uint64_t ub_txg; /* txg of last sync */ 837 uint64_t ub_guid_sum; /* sum of all vdev guids */ 838 uint64_t ub_timestamp; /* UTC time of last sync */ 839 blkptr_t ub_rootbp; /* MOS objset_phys_t */ 840}; 841 842/* 843 * Flags. 844 */ 845#define DNODE_MUST_BE_ALLOCATED 1 846#define DNODE_MUST_BE_FREE 2 847 848/* 849 * Fixed constants. 850 */ 851#define DNODE_SHIFT 9 /* 512 bytes */ 852#define DN_MIN_INDBLKSHIFT 12 /* 4k */ 853#define DN_MAX_INDBLKSHIFT 14 /* 16k */ 854#define DNODE_BLOCK_SHIFT 14 /* 16k */ 855#define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */ 856#define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */ 857#define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */ 858 859/* 860 * Derived constants. 861 */ 862#define DNODE_SIZE (1 << DNODE_SHIFT) 863#define DN_MAX_NBLKPTR ((DNODE_SIZE - DNODE_CORE_SIZE) >> SPA_BLKPTRSHIFT) 864#define DN_MAX_BONUSLEN (DNODE_SIZE - DNODE_CORE_SIZE - (1 << SPA_BLKPTRSHIFT)) 865#define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT) 866 867#define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT) 868#define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT) 869#define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT) 870 871/* The +2 here is a cheesy way to round up */ 872#define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \ 873 (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT))) 874 875#define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \ 876 (((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t)))) 877 878#define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \ 879 (dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT) 880 881#define EPB(blkshift, typeshift) (1 << (blkshift - typeshift)) 882 883/* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */ 884#define DNODE_FLAG_USED_BYTES (1<<0) 885#define DNODE_FLAG_USERUSED_ACCOUNTED (1<<1) 886 887/* Does dnode have a SA spill blkptr in bonus? */ 888#define DNODE_FLAG_SPILL_BLKPTR (1<<2) 889 890typedef struct dnode_phys { 891 uint8_t dn_type; /* dmu_object_type_t */ 892 uint8_t dn_indblkshift; /* ln2(indirect block size) */ 893 uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */ 894 uint8_t dn_nblkptr; /* length of dn_blkptr */ 895 uint8_t dn_bonustype; /* type of data in bonus buffer */ 896 uint8_t dn_checksum; /* ZIO_CHECKSUM type */ 897 uint8_t dn_compress; /* ZIO_COMPRESS type */ 898 uint8_t dn_flags; /* DNODE_FLAG_* */ 899 uint16_t dn_datablkszsec; /* data block size in 512b sectors */ 900 uint16_t dn_bonuslen; /* length of dn_bonus */ 901 uint8_t dn_pad2[4]; 902 903 /* accounting is protected by dn_dirty_mtx */ 904 uint64_t dn_maxblkid; /* largest allocated block ID */ 905 uint64_t dn_used; /* bytes (or sectors) of disk space */ 906 907 uint64_t dn_pad3[4]; 908 909 blkptr_t dn_blkptr[1]; 910 uint8_t dn_bonus[DN_MAX_BONUSLEN - sizeof (blkptr_t)]; 911 blkptr_t dn_spill; 912} dnode_phys_t; 913 914typedef enum dmu_object_byteswap { 915 DMU_BSWAP_UINT8, 916 DMU_BSWAP_UINT16, 917 DMU_BSWAP_UINT32, 918 DMU_BSWAP_UINT64, 919 DMU_BSWAP_ZAP, 920 DMU_BSWAP_DNODE, 921 DMU_BSWAP_OBJSET, 922 DMU_BSWAP_ZNODE, 923 DMU_BSWAP_OLDACL, 924 DMU_BSWAP_ACL, 925 /* 926 * Allocating a new byteswap type number makes the on-disk format 927 * incompatible with any other format that uses the same number. 928 * 929 * Data can usually be structured to work with one of the 930 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. 931 */ 932 DMU_BSWAP_NUMFUNCS 933} dmu_object_byteswap_t; 934 935#define DMU_OT_NEWTYPE 0x80 936#define DMU_OT_METADATA 0x40 937#define DMU_OT_BYTESWAP_MASK 0x3f 938 939/* 940 * Defines a uint8_t object type. Object types specify if the data 941 * in the object is metadata (boolean) and how to byteswap the data 942 * (dmu_object_byteswap_t). 943 */ 944#define DMU_OT(byteswap, metadata) \ 945 (DMU_OT_NEWTYPE | \ 946 ((metadata) ? DMU_OT_METADATA : 0) | \ 947 ((byteswap) & DMU_OT_BYTESWAP_MASK)) 948 949typedef enum dmu_object_type { 950 DMU_OT_NONE, 951 /* general: */ 952 DMU_OT_OBJECT_DIRECTORY, /* ZAP */ 953 DMU_OT_OBJECT_ARRAY, /* UINT64 */ 954 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ 955 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ 956 DMU_OT_BPLIST, /* UINT64 */ 957 DMU_OT_BPLIST_HDR, /* UINT64 */ 958 /* spa: */ 959 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ 960 DMU_OT_SPACE_MAP, /* UINT64 */ 961 /* zil: */ 962 DMU_OT_INTENT_LOG, /* UINT64 */ 963 /* dmu: */ 964 DMU_OT_DNODE, /* DNODE */ 965 DMU_OT_OBJSET, /* OBJSET */ 966 /* dsl: */ 967 DMU_OT_DSL_DIR, /* UINT64 */ 968 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ 969 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ 970 DMU_OT_DSL_PROPS, /* ZAP */ 971 DMU_OT_DSL_DATASET, /* UINT64 */ 972 /* zpl: */ 973 DMU_OT_ZNODE, /* ZNODE */ 974 DMU_OT_OLDACL, /* Old ACL */ 975 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ 976 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ 977 DMU_OT_MASTER_NODE, /* ZAP */ 978 DMU_OT_UNLINKED_SET, /* ZAP */ 979 /* zvol: */ 980 DMU_OT_ZVOL, /* UINT8 */ 981 DMU_OT_ZVOL_PROP, /* ZAP */ 982 /* other; for testing only! */ 983 DMU_OT_PLAIN_OTHER, /* UINT8 */ 984 DMU_OT_UINT64_OTHER, /* UINT64 */ 985 DMU_OT_ZAP_OTHER, /* ZAP */ 986 /* new object types: */ 987 DMU_OT_ERROR_LOG, /* ZAP */ 988 DMU_OT_SPA_HISTORY, /* UINT8 */ 989 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ 990 DMU_OT_POOL_PROPS, /* ZAP */ 991 DMU_OT_DSL_PERMS, /* ZAP */ 992 DMU_OT_ACL, /* ACL */ 993 DMU_OT_SYSACL, /* SYSACL */ 994 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ 995 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ 996 DMU_OT_NEXT_CLONES, /* ZAP */ 997 DMU_OT_SCAN_QUEUE, /* ZAP */ 998 DMU_OT_USERGROUP_USED, /* ZAP */ 999 DMU_OT_USERGROUP_QUOTA, /* ZAP */ 1000 DMU_OT_USERREFS, /* ZAP */ 1001 DMU_OT_DDT_ZAP, /* ZAP */ 1002 DMU_OT_DDT_STATS, /* ZAP */ 1003 DMU_OT_SA, /* System attr */ 1004 DMU_OT_SA_MASTER_NODE, /* ZAP */ 1005 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ 1006 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ 1007 DMU_OT_SCAN_XLATE, /* ZAP */ 1008 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ 1009 DMU_OT_NUMTYPES, 1010 1011 /* 1012 * Names for valid types declared with DMU_OT(). 1013 */ 1014 DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE), 1015 DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE), 1016 DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE), 1017 DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE), 1018 DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE), 1019 DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE), 1020 DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE), 1021 DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE), 1022 DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE), 1023 DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE) 1024} dmu_object_type_t; 1025 1026typedef enum dmu_objset_type { 1027 DMU_OST_NONE, 1028 DMU_OST_META, 1029 DMU_OST_ZFS, 1030 DMU_OST_ZVOL, 1031 DMU_OST_OTHER, /* For testing only! */ 1032 DMU_OST_ANY, /* Be careful! */ 1033 DMU_OST_NUMTYPES 1034} dmu_objset_type_t; 1035 1036/* 1037 * header for all bonus and spill buffers. 1038 * The header has a fixed portion with a variable number 1039 * of "lengths" depending on the number of variable sized 1040 * attribues which are determined by the "layout number" 1041 */ 1042 1043#define SA_MAGIC 0x2F505A /* ZFS SA */ 1044typedef struct sa_hdr_phys { 1045 uint32_t sa_magic; 1046 uint16_t sa_layout_info; /* Encoded with hdrsize and layout number */ 1047 uint16_t sa_lengths[1]; /* optional sizes for variable length attrs */ 1048 /* ... Data follows the lengths. */ 1049} sa_hdr_phys_t; 1050 1051/* 1052 * sa_hdr_phys -> sa_layout_info 1053 * 1054 * 16 10 0 1055 * +--------+-------+ 1056 * | hdrsz |layout | 1057 * +--------+-------+ 1058 * 1059 * Bits 0-10 are the layout number 1060 * Bits 11-16 are the size of the header. 1061 * The hdrsize is the number * 8 1062 * 1063 * For example. 1064 * hdrsz of 1 ==> 8 byte header 1065 * 2 ==> 16 byte header 1066 * 1067 */ 1068 1069#define SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10) 1070#define SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 16, 3, 0) 1071#define SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \ 1072{ \ 1073 BF32_SET_SB(x, 10, 6, 3, 0, size); \ 1074 BF32_SET(x, 0, 10, num); \ 1075} 1076 1077#define SA_MODE_OFFSET 0 1078#define SA_SIZE_OFFSET 8 1079#define SA_GEN_OFFSET 16 1080#define SA_UID_OFFSET 24 1081#define SA_GID_OFFSET 32 1082#define SA_PARENT_OFFSET 40 1083#define SA_SYMLINK_OFFSET 160 1084 1085/* 1086 * Intent log header - this on disk structure holds fields to manage 1087 * the log. All fields are 64 bit to easily handle cross architectures. 1088 */ 1089typedef struct zil_header { 1090 uint64_t zh_claim_txg; /* txg in which log blocks were claimed */ 1091 uint64_t zh_replay_seq; /* highest replayed sequence number */ 1092 blkptr_t zh_log; /* log chain */ 1093 uint64_t zh_claim_seq; /* highest claimed sequence number */ 1094 uint64_t zh_pad[5]; 1095} zil_header_t; 1096 1097#define OBJSET_PHYS_SIZE 2048 1098 1099typedef struct objset_phys { 1100 dnode_phys_t os_meta_dnode; 1101 zil_header_t os_zil_header; 1102 uint64_t os_type; 1103 uint64_t os_flags; 1104 char os_pad[OBJSET_PHYS_SIZE - sizeof (dnode_phys_t)*3 - 1105 sizeof (zil_header_t) - sizeof (uint64_t)*2]; 1106 dnode_phys_t os_userused_dnode; 1107 dnode_phys_t os_groupused_dnode; 1108} objset_phys_t; 1109 1110typedef struct dsl_dir_phys { 1111 uint64_t dd_creation_time; /* not actually used */ 1112 uint64_t dd_head_dataset_obj; 1113 uint64_t dd_parent_obj; 1114 uint64_t dd_clone_parent_obj; 1115 uint64_t dd_child_dir_zapobj; 1116 /* 1117 * how much space our children are accounting for; for leaf 1118 * datasets, == physical space used by fs + snaps 1119 */ 1120 uint64_t dd_used_bytes; 1121 uint64_t dd_compressed_bytes; 1122 uint64_t dd_uncompressed_bytes; 1123 /* Administrative quota setting */ 1124 uint64_t dd_quota; 1125 /* Administrative reservation setting */ 1126 uint64_t dd_reserved; 1127 uint64_t dd_props_zapobj; 1128 uint64_t dd_pad[21]; /* pad out to 256 bytes for good measure */ 1129} dsl_dir_phys_t; 1130 1131typedef struct dsl_dataset_phys { 1132 uint64_t ds_dir_obj; 1133 uint64_t ds_prev_snap_obj; 1134 uint64_t ds_prev_snap_txg; 1135 uint64_t ds_next_snap_obj; 1136 uint64_t ds_snapnames_zapobj; /* zap obj of snaps; ==0 for snaps */ 1137 uint64_t ds_num_children; /* clone/snap children; ==0 for head */ 1138 uint64_t ds_creation_time; /* seconds since 1970 */ 1139 uint64_t ds_creation_txg; 1140 uint64_t ds_deadlist_obj; 1141 uint64_t ds_used_bytes; 1142 uint64_t ds_compressed_bytes; 1143 uint64_t ds_uncompressed_bytes; 1144 uint64_t ds_unique_bytes; /* only relevant to snapshots */ 1145 /* 1146 * The ds_fsid_guid is a 56-bit ID that can change to avoid 1147 * collisions. The ds_guid is a 64-bit ID that will never 1148 * change, so there is a small probability that it will collide. 1149 */ 1150 uint64_t ds_fsid_guid; 1151 uint64_t ds_guid; 1152 uint64_t ds_flags; 1153 blkptr_t ds_bp; 1154 uint64_t ds_pad[8]; /* pad out to 320 bytes for good measure */ 1155} dsl_dataset_phys_t; 1156 1157/* 1158 * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 1159 */ 1160#define DMU_POOL_DIRECTORY_OBJECT 1 1161#define DMU_POOL_CONFIG "config" 1162#define DMU_POOL_FEATURES_FOR_READ "features_for_read" 1163#define DMU_POOL_ROOT_DATASET "root_dataset" 1164#define DMU_POOL_SYNC_BPLIST "sync_bplist" 1165#define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 1166#define DMU_POOL_ERRLOG_LAST "errlog_last" 1167#define DMU_POOL_SPARES "spares" 1168#define DMU_POOL_DEFLATE "deflate" 1169#define DMU_POOL_HISTORY "history" 1170#define DMU_POOL_PROPS "pool_props" 1171#define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" 1172 1173#define ZAP_MAGIC 0x2F52AB2ABULL 1174 1175#define FZAP_BLOCK_SHIFT(zap) ((zap)->zap_block_shift) 1176 1177#define ZAP_MAXCD (uint32_t)(-1) 1178#define ZAP_HASHBITS 28 1179#define MZAP_ENT_LEN 64 1180#define MZAP_NAME_LEN (MZAP_ENT_LEN - 8 - 4 - 2) 1181#define MZAP_MAX_BLKSHIFT SPA_MAXBLOCKSHIFT 1182#define MZAP_MAX_BLKSZ (1 << MZAP_MAX_BLKSHIFT) 1183 1184typedef struct mzap_ent_phys { 1185 uint64_t mze_value; 1186 uint32_t mze_cd; 1187 uint16_t mze_pad; /* in case we want to chain them someday */ 1188 char mze_name[MZAP_NAME_LEN]; 1189} mzap_ent_phys_t; 1190 1191typedef struct mzap_phys { 1192 uint64_t mz_block_type; /* ZBT_MICRO */ 1193 uint64_t mz_salt; 1194 uint64_t mz_pad[6]; 1195 mzap_ent_phys_t mz_chunk[1]; 1196 /* actually variable size depending on block size */ 1197} mzap_phys_t; 1198 1199/* 1200 * The (fat) zap is stored in one object. It is an array of 1201 * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of: 1202 * 1203 * ptrtbl fits in first block: 1204 * [zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ... 1205 * 1206 * ptrtbl too big for first block: 1207 * [zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ... 1208 * 1209 */ 1210 1211#define ZBT_LEAF ((1ULL << 63) + 0) 1212#define ZBT_HEADER ((1ULL << 63) + 1) 1213#define ZBT_MICRO ((1ULL << 63) + 3) 1214/* any other values are ptrtbl blocks */ 1215 1216/* 1217 * the embedded pointer table takes up half a block: 1218 * block size / entry size (2^3) / 2 1219 */ 1220#define ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1) 1221 1222/* 1223 * The embedded pointer table starts half-way through the block. Since 1224 * the pointer table itself is half the block, it starts at (64-bit) 1225 * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)). 1226 */ 1227#define ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \ 1228 ((uint64_t *)(zap)->zap_phys) \ 1229 [(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))] 1230 1231/* 1232 * TAKE NOTE: 1233 * If zap_phys_t is modified, zap_byteswap() must be modified. 1234 */ 1235typedef struct zap_phys { 1236 uint64_t zap_block_type; /* ZBT_HEADER */ 1237 uint64_t zap_magic; /* ZAP_MAGIC */ 1238 1239 struct zap_table_phys { 1240 uint64_t zt_blk; /* starting block number */ 1241 uint64_t zt_numblks; /* number of blocks */ 1242 uint64_t zt_shift; /* bits to index it */ 1243 uint64_t zt_nextblk; /* next (larger) copy start block */ 1244 uint64_t zt_blks_copied; /* number source blocks copied */ 1245 } zap_ptrtbl; 1246 1247 uint64_t zap_freeblk; /* the next free block */ 1248 uint64_t zap_num_leafs; /* number of leafs */ 1249 uint64_t zap_num_entries; /* number of entries */ 1250 uint64_t zap_salt; /* salt to stir into hash function */ 1251 /* 1252 * This structure is followed by padding, and then the embedded 1253 * pointer table. The embedded pointer table takes up second 1254 * half of the block. It is accessed using the 1255 * ZAP_EMBEDDED_PTRTBL_ENT() macro. 1256 */ 1257} zap_phys_t; 1258 1259typedef struct zap_table_phys zap_table_phys_t; 1260 1261typedef struct fat_zap { 1262 int zap_block_shift; /* block size shift */ 1263 zap_phys_t *zap_phys; 1264} fat_zap_t; 1265 1266#define ZAP_LEAF_MAGIC 0x2AB1EAF 1267 1268/* chunk size = 24 bytes */ 1269#define ZAP_LEAF_CHUNKSIZE 24 1270 1271/* 1272 * The amount of space available for chunks is: 1273 * block size (1<<l->l_bs) - hash entry size (2) * number of hash 1274 * entries - header space (2*chunksize) 1275 */ 1276#define ZAP_LEAF_NUMCHUNKS(l) \ 1277 (((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \ 1278 ZAP_LEAF_CHUNKSIZE - 2) 1279 1280/* 1281 * The amount of space within the chunk available for the array is: 1282 * chunk size - space for type (1) - space for next pointer (2) 1283 */ 1284#define ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3) 1285 1286#define ZAP_LEAF_ARRAY_NCHUNKS(bytes) \ 1287 (((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES) 1288 1289/* 1290 * Low water mark: when there are only this many chunks free, start 1291 * growing the ptrtbl. Ideally, this should be larger than a 1292 * "reasonably-sized" entry. 20 chunks is more than enough for the 1293 * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value), 1294 * while still being only around 3% for 16k blocks. 1295 */ 1296#define ZAP_LEAF_LOW_WATER (20) 1297 1298/* 1299 * The leaf hash table has block size / 2^5 (32) number of entries, 1300 * which should be more than enough for the maximum number of entries, 1301 * which is less than block size / CHUNKSIZE (24) / minimum number of 1302 * chunks per entry (3). 1303 */ 1304#define ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5) 1305#define ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l)) 1306 1307/* 1308 * The chunks start immediately after the hash table. The end of the 1309 * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a 1310 * chunk_t. 1311 */ 1312#define ZAP_LEAF_CHUNK(l, idx) \ 1313 ((zap_leaf_chunk_t *) \ 1314 ((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx] 1315#define ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry) 1316 1317typedef enum zap_chunk_type { 1318 ZAP_CHUNK_FREE = 253, 1319 ZAP_CHUNK_ENTRY = 252, 1320 ZAP_CHUNK_ARRAY = 251, 1321 ZAP_CHUNK_TYPE_MAX = 250 1322} zap_chunk_type_t; 1323 1324/* 1325 * TAKE NOTE: 1326 * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified. 1327 */ 1328typedef struct zap_leaf_phys { 1329 struct zap_leaf_header { 1330 uint64_t lh_block_type; /* ZBT_LEAF */ 1331 uint64_t lh_pad1; 1332 uint64_t lh_prefix; /* hash prefix of this leaf */ 1333 uint32_t lh_magic; /* ZAP_LEAF_MAGIC */ 1334 uint16_t lh_nfree; /* number free chunks */ 1335 uint16_t lh_nentries; /* number of entries */ 1336 uint16_t lh_prefix_len; /* num bits used to id this */ 1337 1338/* above is accessable to zap, below is zap_leaf private */ 1339 1340 uint16_t lh_freelist; /* chunk head of free list */ 1341 uint8_t lh_pad2[12]; 1342 } l_hdr; /* 2 24-byte chunks */ 1343 1344 /* 1345 * The header is followed by a hash table with 1346 * ZAP_LEAF_HASH_NUMENTRIES(zap) entries. The hash table is 1347 * followed by an array of ZAP_LEAF_NUMCHUNKS(zap) 1348 * zap_leaf_chunk structures. These structures are accessed 1349 * with the ZAP_LEAF_CHUNK() macro. 1350 */ 1351 1352 uint16_t l_hash[1]; 1353} zap_leaf_phys_t; 1354 1355typedef union zap_leaf_chunk { 1356 struct zap_leaf_entry { 1357 uint8_t le_type; /* always ZAP_CHUNK_ENTRY */ 1358 uint8_t le_value_intlen; /* size of ints */ 1359 uint16_t le_next; /* next entry in hash chain */ 1360 uint16_t le_name_chunk; /* first chunk of the name */ 1361 uint16_t le_name_numints; /* bytes in name, incl null */ 1362 uint16_t le_value_chunk; /* first chunk of the value */ 1363 uint16_t le_value_numints; /* value length in ints */ 1364 uint32_t le_cd; /* collision differentiator */ 1365 uint64_t le_hash; /* hash value of the name */ 1366 } l_entry; 1367 struct zap_leaf_array { 1368 uint8_t la_type; /* always ZAP_CHUNK_ARRAY */ 1369 uint8_t la_array[ZAP_LEAF_ARRAY_BYTES]; 1370 uint16_t la_next; /* next blk or CHAIN_END */ 1371 } l_array; 1372 struct zap_leaf_free { 1373 uint8_t lf_type; /* always ZAP_CHUNK_FREE */ 1374 uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES]; 1375 uint16_t lf_next; /* next in free list, or CHAIN_END */ 1376 } l_free; 1377} zap_leaf_chunk_t; 1378 1379typedef struct zap_leaf { 1380 int l_bs; /* block size shift */ 1381 zap_leaf_phys_t *l_phys; 1382} zap_leaf_t; 1383 1384/* 1385 * Define special zfs pflags 1386 */ 1387#define ZFS_XATTR 0x1 /* is an extended attribute */ 1388#define ZFS_INHERIT_ACE 0x2 /* ace has inheritable ACEs */ 1389#define ZFS_ACL_TRIVIAL 0x4 /* files ACL is trivial */ 1390 1391#define MASTER_NODE_OBJ 1 1392 1393/* 1394 * special attributes for master node. 1395 */ 1396 1397#define ZFS_FSID "FSID" 1398#define ZFS_UNLINKED_SET "DELETE_QUEUE" 1399#define ZFS_ROOT_OBJ "ROOT" 1400#define ZPL_VERSION_OBJ "VERSION" 1401#define ZFS_PROP_BLOCKPERPAGE "BLOCKPERPAGE" 1402#define ZFS_PROP_NOGROWBLOCKS "NOGROWBLOCKS" 1403 1404#define ZFS_FLAG_BLOCKPERPAGE 0x1 1405#define ZFS_FLAG_NOGROWBLOCKS 0x2 1406 1407/* 1408 * ZPL version - rev'd whenever an incompatible on-disk format change 1409 * occurs. Independent of SPA/DMU/ZAP versioning. 1410 */ 1411 1412#define ZPL_VERSION 1ULL 1413 1414/* 1415 * The directory entry has the type (currently unused on Solaris) in the 1416 * top 4 bits, and the object number in the low 48 bits. The "middle" 1417 * 12 bits are unused. 1418 */ 1419#define ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4) 1420#define ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48) 1421#define ZFS_DIRENT_MAKE(type, obj) (((uint64_t)type << 60) | obj) 1422 1423typedef struct ace { 1424 uid_t a_who; /* uid or gid */ 1425 uint32_t a_access_mask; /* read,write,... */ 1426 uint16_t a_flags; /* see below */ 1427 uint16_t a_type; /* allow or deny */ 1428} ace_t; 1429 1430#define ACE_SLOT_CNT 6 1431 1432typedef struct zfs_znode_acl { 1433 uint64_t z_acl_extern_obj; /* ext acl pieces */ 1434 uint32_t z_acl_count; /* Number of ACEs */ 1435 uint16_t z_acl_version; /* acl version */ 1436 uint16_t z_acl_pad; /* pad */ 1437 ace_t z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */ 1438} zfs_znode_acl_t; 1439 1440/* 1441 * This is the persistent portion of the znode. It is stored 1442 * in the "bonus buffer" of the file. Short symbolic links 1443 * are also stored in the bonus buffer. 1444 */ 1445typedef struct znode_phys { 1446 uint64_t zp_atime[2]; /* 0 - last file access time */ 1447 uint64_t zp_mtime[2]; /* 16 - last file modification time */ 1448 uint64_t zp_ctime[2]; /* 32 - last file change time */ 1449 uint64_t zp_crtime[2]; /* 48 - creation time */ 1450 uint64_t zp_gen; /* 64 - generation (txg of creation) */ 1451 uint64_t zp_mode; /* 72 - file mode bits */ 1452 uint64_t zp_size; /* 80 - size of file */ 1453 uint64_t zp_parent; /* 88 - directory parent (`..') */ 1454 uint64_t zp_links; /* 96 - number of links to file */ 1455 uint64_t zp_xattr; /* 104 - DMU object for xattrs */ 1456 uint64_t zp_rdev; /* 112 - dev_t for VBLK & VCHR files */ 1457 uint64_t zp_flags; /* 120 - persistent flags */ 1458 uint64_t zp_uid; /* 128 - file owner */ 1459 uint64_t zp_gid; /* 136 - owning group */ 1460 uint64_t zp_pad[4]; /* 144 - future */ 1461 zfs_znode_acl_t zp_acl; /* 176 - 263 ACL */ 1462 /* 1463 * Data may pad out any remaining bytes in the znode buffer, eg: 1464 * 1465 * |<---------------------- dnode_phys (512) ------------------------>| 1466 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->| 1467 * |<---- znode (264) ---->|<---- data (56) ---->| 1468 * 1469 * At present, we only use this space to store symbolic links. 1470 */ 1471} znode_phys_t; 1472 1473/* 1474 * In-core vdev representation. 1475 */ 1476struct vdev; 1477struct spa; 1478typedef int vdev_phys_read_t(struct vdev *vdev, void *priv, 1479 off_t offset, void *buf, size_t bytes); 1480typedef int vdev_read_t(struct vdev *vdev, const blkptr_t *bp, 1481 void *buf, off_t offset, size_t bytes); 1482 1483typedef STAILQ_HEAD(vdev_list, vdev) vdev_list_t; 1484 1485typedef struct vdev { 1486 STAILQ_ENTRY(vdev) v_childlink; /* link in parent's child list */ 1487 STAILQ_ENTRY(vdev) v_alllink; /* link in global vdev list */ 1488 vdev_list_t v_children; /* children of this vdev */ 1489 const char *v_name; /* vdev name */ 1490 uint64_t v_guid; /* vdev guid */ 1491 int v_id; /* index in parent */ 1492 int v_ashift; /* offset to block shift */ 1493 int v_nparity; /* # parity for raidz */ 1494 struct vdev *v_top; /* parent vdev */ 1495 int v_nchildren; /* # children */ 1496 vdev_state_t v_state; /* current state */ 1497 vdev_phys_read_t *v_phys_read; /* read from raw leaf vdev */ 1498 vdev_read_t *v_read; /* read from vdev */ 1499 void *v_read_priv; /* private data for read function */ 1500 struct spa *spa; /* link to spa */ 1501} vdev_t; 1502 1503/* 1504 * In-core pool representation. 1505 */ 1506typedef STAILQ_HEAD(spa_list, spa) spa_list_t; 1507 1508typedef struct spa { 1509 STAILQ_ENTRY(spa) spa_link; /* link in global pool list */ 1510 char *spa_name; /* pool name */ 1511 uint64_t spa_guid; /* pool guid */ 1512 uint64_t spa_txg; /* most recent transaction */ 1513 struct uberblock spa_uberblock; /* best uberblock so far */ 1514 vdev_list_t spa_vdevs; /* list of all toplevel vdevs */ 1515 objset_phys_t spa_mos; /* MOS for this pool */ 1516 zio_cksum_salt_t spa_cksum_salt; /* secret salt for cksum */ 1517 void *spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS]; 1518 int spa_inited; /* initialized */ 1519} spa_t; 1520 1521static void decode_embedded_bp_compressed(const blkptr_t *, void *); 1522