1/*- 2 * Copyright (c) 2002 McAfee, Inc. 3 * All rights reserved. 4 * 5 * This software was developed for the FreeBSD Project by Marshall 6 * Kirk McKusick and McAfee Research,, the Security Research Division of 7 * McAfee, Inc. under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as 8 * part of the DARPA CHATS research program 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31/* 32 * CDDL HEADER START 33 * 34 * The contents of this file are subject to the terms of the 35 * Common Development and Distribution License (the "License"). 36 * You may not use this file except in compliance with the License. 37 * 38 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 39 * or http://www.opensolaris.org/os/licensing. 40 * See the License for the specific language governing permissions 41 * and limitations under the License. 42 * 43 * When distributing Covered Code, include this CDDL HEADER in each 44 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 45 * If applicable, add the following below this CDDL HEADER, with the 46 * fields enclosed by brackets "[]" replaced with your own identifying 47 * information: Portions Copyright [yyyy] [name of copyright owner] 48 * 49 * CDDL HEADER END 50 */ 51/* 52 * Copyright 2009 Sun Microsystems, Inc. All rights reserved. 53 * Use is subject to license terms. 54 */ 55/* 56 * Copyright 2013 by Saso Kiselkov. All rights reserved. 57 */ 58/* 59 * Copyright (c) 2020 by Delphix. All rights reserved. 60 */ 61 62#include <sys/queue.h> 63 64#ifndef _ZFSIMPL_H_ 65#define _ZFSIMPL_H_ 66 67#define MAXNAMELEN 256 68 69#define _NOTE(s) 70 71/* 72 * AVL comparator helpers 73 */ 74#define AVL_ISIGN(a) (((a) > 0) - ((a) < 0)) 75#define AVL_CMP(a, b) (((a) > (b)) - ((a) < (b))) 76#define AVL_PCMP(a, b) \ 77 (((uintptr_t)(a) > (uintptr_t)(b)) - ((uintptr_t)(a) < (uintptr_t)(b))) 78 79#if !defined(NEED_SOLARIS_BOOLEAN) /* Only defined when we'll define this elsewhere */ 80typedef enum { B_FALSE, B_TRUE } boolean_t; 81#endif 82 83/* CRC64 table */ 84#define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ 85 86/* 87 * Macros for various sorts of alignment and rounding when the alignment 88 * is known to be a power of 2. 89 */ 90#define P2ALIGN(x, align) ((x) & -(align)) 91#define P2PHASE(x, align) ((x) & ((align) - 1)) 92#define P2NPHASE(x, align) (-(x) & ((align) - 1)) 93#define P2ROUNDUP(x, align) (-(-(x) & -(align))) 94#define P2END(x, align) (-(~(x) & -(align))) 95#define P2PHASEUP(x, align, phase) ((phase) - (((phase) - (x)) & -(align))) 96#define P2BOUNDARY(off, len, align) (((off) ^ ((off) + (len) - 1)) > (align) - 1) 97 98/* 99 * General-purpose 32-bit and 64-bit bitfield encodings. 100 */ 101#define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) 102#define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) 103#define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) 104#define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) 105 106#define BF32_GET(x, low, len) BF32_DECODE(x, low, len) 107#define BF64_GET(x, low, len) BF64_DECODE(x, low, len) 108 109#define BF32_SET(x, low, len, val) \ 110 ((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len)) 111#define BF64_SET(x, low, len, val) \ 112 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)) 113 114#define BF32_GET_SB(x, low, len, shift, bias) \ 115 ((BF32_GET(x, low, len) + (bias)) << (shift)) 116#define BF64_GET_SB(x, low, len, shift, bias) \ 117 ((BF64_GET(x, low, len) + (bias)) << (shift)) 118 119#define BF32_SET_SB(x, low, len, shift, bias, val) \ 120 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)) 121#define BF64_SET_SB(x, low, len, shift, bias, val) \ 122 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)) 123 124/* 125 * Macros to reverse byte order 126 */ 127#define BSWAP_8(x) ((x) & 0xff) 128#define BSWAP_16(x) ((BSWAP_8(x) << 8) | BSWAP_8((x) >> 8)) 129#define BSWAP_32(x) ((BSWAP_16(x) << 16) | BSWAP_16((x) >> 16)) 130#define BSWAP_64(x) ((BSWAP_32(x) << 32) | BSWAP_32((x) >> 32)) 131 132#define SPA_MINBLOCKSHIFT 9 133#define SPA_OLDMAXBLOCKSHIFT 17 134#define SPA_MAXBLOCKSHIFT 24 135#define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 136#define SPA_OLDMAXBLOCKSIZE (1ULL << SPA_OLDMAXBLOCKSHIFT) 137#define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 138 139/* 140 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. 141 * The ASIZE encoding should be at least 64 times larger (6 more bits) 142 * to support up to 4-way RAID-Z mirror mode with worst-case gang block 143 * overhead, three DVAs per bp, plus one more bit in case we do anything 144 * else that expands the ASIZE. 145 */ 146#define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ 147#define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ 148#define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ 149 150/* 151 * All SPA data is represented by 128-bit data virtual addresses (DVAs). 152 * The members of the dva_t should be considered opaque outside the SPA. 153 */ 154typedef struct dva { 155 uint64_t dva_word[2]; 156} dva_t; 157 158/* 159 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. 160 */ 161typedef struct zio_cksum { 162 uint64_t zc_word[4]; 163} zio_cksum_t; 164 165/* 166 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept 167 * secret and is suitable for use in MAC algorithms as the key. 168 */ 169typedef struct zio_cksum_salt { 170 uint8_t zcs_bytes[32]; 171} zio_cksum_salt_t; 172 173/* 174 * Each block is described by its DVAs, time of birth, checksum, etc. 175 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 176 * 177 * 64 56 48 40 32 24 16 8 0 178 * +-------+-------+-------+-------+-------+-------+-------+-------+ 179 * 0 | vdev1 | GRID | ASIZE | 180 * +-------+-------+-------+-------+-------+-------+-------+-------+ 181 * 1 |G| offset1 | 182 * +-------+-------+-------+-------+-------+-------+-------+-------+ 183 * 2 | vdev2 | GRID | ASIZE | 184 * +-------+-------+-------+-------+-------+-------+-------+-------+ 185 * 3 |G| offset2 | 186 * +-------+-------+-------+-------+-------+-------+-------+-------+ 187 * 4 | vdev3 | GRID | ASIZE | 188 * +-------+-------+-------+-------+-------+-------+-------+-------+ 189 * 5 |G| offset3 | 190 * +-------+-------+-------+-------+-------+-------+-------+-------+ 191 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 192 * +-------+-------+-------+-------+-------+-------+-------+-------+ 193 * 7 | padding | 194 * +-------+-------+-------+-------+-------+-------+-------+-------+ 195 * 8 | padding | 196 * +-------+-------+-------+-------+-------+-------+-------+-------+ 197 * 9 | physical birth txg | 198 * +-------+-------+-------+-------+-------+-------+-------+-------+ 199 * a | logical birth txg | 200 * +-------+-------+-------+-------+-------+-------+-------+-------+ 201 * b | fill count | 202 * +-------+-------+-------+-------+-------+-------+-------+-------+ 203 * c | checksum[0] | 204 * +-------+-------+-------+-------+-------+-------+-------+-------+ 205 * d | checksum[1] | 206 * +-------+-------+-------+-------+-------+-------+-------+-------+ 207 * e | checksum[2] | 208 * +-------+-------+-------+-------+-------+-------+-------+-------+ 209 * f | checksum[3] | 210 * +-------+-------+-------+-------+-------+-------+-------+-------+ 211 * 212 * Legend: 213 * 214 * vdev virtual device ID 215 * offset offset into virtual device 216 * LSIZE logical size 217 * PSIZE physical size (after compression) 218 * ASIZE allocated size (including RAID-Z parity and gang block headers) 219 * GRID RAID-Z layout information (reserved for future use) 220 * cksum checksum function 221 * comp compression function 222 * G gang block indicator 223 * B byteorder (endianness) 224 * D dedup 225 * X encryption (on version 30, which is not supported) 226 * E blkptr_t contains embedded data (see below) 227 * lvl level of indirection 228 * type DMU object type 229 * phys birth txg of block allocation; zero if same as logical birth txg 230 * log. birth transaction group in which the block was logically born 231 * fill count number of non-zero blocks under this bp 232 * checksum[4] 256-bit checksum of the data this bp describes 233 */ 234 235/* 236 * "Embedded" blkptr_t's don't actually point to a block, instead they 237 * have a data payload embedded in the blkptr_t itself. See the comment 238 * in blkptr.c for more details. 239 * 240 * The blkptr_t is laid out as follows: 241 * 242 * 64 56 48 40 32 24 16 8 0 243 * +-------+-------+-------+-------+-------+-------+-------+-------+ 244 * 0 | payload | 245 * 1 | payload | 246 * 2 | payload | 247 * 3 | payload | 248 * 4 | payload | 249 * 5 | payload | 250 * +-------+-------+-------+-------+-------+-------+-------+-------+ 251 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 252 * +-------+-------+-------+-------+-------+-------+-------+-------+ 253 * 7 | payload | 254 * 8 | payload | 255 * 9 | payload | 256 * +-------+-------+-------+-------+-------+-------+-------+-------+ 257 * a | logical birth txg | 258 * +-------+-------+-------+-------+-------+-------+-------+-------+ 259 * b | payload | 260 * c | payload | 261 * d | payload | 262 * e | payload | 263 * f | payload | 264 * +-------+-------+-------+-------+-------+-------+-------+-------+ 265 * 266 * Legend: 267 * 268 * payload contains the embedded data 269 * B (byteorder) byteorder (endianness) 270 * D (dedup) padding (set to zero) 271 * X encryption (set to zero; see above) 272 * E (embedded) set to one 273 * lvl indirection level 274 * type DMU object type 275 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 276 * comp compression function of payload 277 * PSIZE size of payload after compression, in bytes 278 * LSIZE logical size of payload, in bytes 279 * note that 25 bits is enough to store the largest 280 * "normal" BP's LSIZE (2^16 * 2^9) in bytes 281 * log. birth transaction group in which the block was logically born 282 * 283 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 284 * bp's they are stored in units of SPA_MINBLOCKSHIFT. 285 * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 286 * The B, D, X, lvl, type, and comp fields are stored the same as with normal 287 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 288 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 289 * other macros, as they assert that they are only used on BP's of the correct 290 * "embedded-ness". 291 */ 292 293#define BPE_GET_ETYPE(bp) \ 294 (ASSERT(BP_IS_EMBEDDED(bp)), \ 295 BF64_GET((bp)->blk_prop, 40, 8)) 296#define BPE_SET_ETYPE(bp, t) do { \ 297 ASSERT(BP_IS_EMBEDDED(bp)); \ 298 BF64_SET((bp)->blk_prop, 40, 8, t); \ 299_NOTE(CONSTCOND) } while (0) 300 301#define BPE_GET_LSIZE(bp) \ 302 (ASSERT(BP_IS_EMBEDDED(bp)), \ 303 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 304#define BPE_SET_LSIZE(bp, x) do { \ 305 ASSERT(BP_IS_EMBEDDED(bp)); \ 306 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 307_NOTE(CONSTCOND) } while (0) 308 309#define BPE_GET_PSIZE(bp) \ 310 (ASSERT(BP_IS_EMBEDDED(bp)), \ 311 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 312#define BPE_SET_PSIZE(bp, x) do { \ 313 ASSERT(BP_IS_EMBEDDED(bp)); \ 314 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 315_NOTE(CONSTCOND) } while (0) 316 317typedef enum bp_embedded_type { 318 BP_EMBEDDED_TYPE_DATA, 319 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ 320 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED 321} bp_embedded_type_t; 322 323#define BPE_NUM_WORDS 14 324#define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 325#define BPE_IS_PAYLOADWORD(bp, wp) \ 326 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 327 328#define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 329#define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 330 331typedef struct blkptr { 332 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 333 uint64_t blk_prop; /* size, compression, type, etc */ 334 uint64_t blk_pad[2]; /* Extra space for the future */ 335 uint64_t blk_phys_birth; /* txg when block was allocated */ 336 uint64_t blk_birth; /* transaction group at birth */ 337 uint64_t blk_fill; /* fill count */ 338 zio_cksum_t blk_cksum; /* 256-bit checksum */ 339} blkptr_t; 340 341/* 342 * Macros to get and set fields in a bp or DVA. 343 */ 344#define DVA_GET_ASIZE(dva) \ 345 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 346#define DVA_SET_ASIZE(dva, x) \ 347 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 348 SPA_MINBLOCKSHIFT, 0, x) 349 350#define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 351#define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 352 353#define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) 354#define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) 355 356#define DVA_GET_OFFSET(dva) \ 357 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 358#define DVA_SET_OFFSET(dva, x) \ 359 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 360 361#define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 362#define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 363 364#define BP_GET_LSIZE(bp) \ 365 (BP_IS_EMBEDDED(bp) ? \ 366 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 367 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 368#define BP_SET_LSIZE(bp, x) do { \ 369 ASSERT(!BP_IS_EMBEDDED(bp)); \ 370 BF64_SET_SB((bp)->blk_prop, \ 371 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 372_NOTE(CONSTCOND) } while (0) 373 374#define BP_GET_PSIZE(bp) \ 375 BF64_GET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1) 376#define BP_SET_PSIZE(bp, x) \ 377 BF64_SET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x) 378 379#define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7) 380#define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x) 381 382#define BP_GET_CHECKSUM(bp) BF64_GET((bp)->blk_prop, 40, 8) 383#define BP_SET_CHECKSUM(bp, x) BF64_SET((bp)->blk_prop, 40, 8, x) 384 385#define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 386#define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 387 388#define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 389#define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 390 391#define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 392 393#define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 394#define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 395 396#define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 397#define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 398 399#define BP_PHYSICAL_BIRTH(bp) \ 400 ((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 401 402#define BP_SET_BIRTH(bp, logical, physical) \ 403{ \ 404 ASSERT(!BP_IS_EMBEDDED(bp)); \ 405 (bp)->blk_birth = (logical); \ 406 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \ 407} 408 409#define BP_GET_FILL(bp) \ 410 ((BP_IS_EMBEDDED(bp)) ? 1 : (bp)->blk_fill) 411 412#define BP_SET_FILL(bp, fill) \ 413{ \ 414 (bp)->blk_fill = fill; \ 415} 416 417#define BP_GET_ASIZE(bp) \ 418 (DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 419 DVA_GET_ASIZE(&(bp)->blk_dva[2])) 420 421#define BP_GET_UCSIZE(bp) \ 422 ((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \ 423 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)); 424 425#define BP_GET_NDVAS(bp) \ 426 (!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 427 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 428 !!DVA_GET_ASIZE(&(bp)->blk_dva[2])) 429 430#define DVA_EQUAL(dva1, dva2) \ 431 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 432 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 433 434#define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 435 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 436 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 437 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 438 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 439 440 441#define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 442 443#define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ 444{ \ 445 (zcp)->zc_word[0] = w0; \ 446 (zcp)->zc_word[1] = w1; \ 447 (zcp)->zc_word[2] = w2; \ 448 (zcp)->zc_word[3] = w3; \ 449} 450 451#define BP_IDENTITY(bp) (&(bp)->blk_dva[0]) 452#define BP_IS_GANG(bp) DVA_GET_GANG(BP_IDENTITY(bp)) 453#define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 454 (dva)->dva_word[1] == 0ULL) 455#define BP_IS_HOLE(bp) DVA_IS_EMPTY(BP_IDENTITY(bp)) 456#define BP_IS_OLDER(bp, txg) (!BP_IS_HOLE(bp) && (bp)->blk_birth < (txg)) 457 458#define BP_ZERO(bp) \ 459{ \ 460 (bp)->blk_dva[0].dva_word[0] = 0; \ 461 (bp)->blk_dva[0].dva_word[1] = 0; \ 462 (bp)->blk_dva[1].dva_word[0] = 0; \ 463 (bp)->blk_dva[1].dva_word[1] = 0; \ 464 (bp)->blk_dva[2].dva_word[0] = 0; \ 465 (bp)->blk_dva[2].dva_word[1] = 0; \ 466 (bp)->blk_prop = 0; \ 467 (bp)->blk_pad[0] = 0; \ 468 (bp)->blk_pad[1] = 0; \ 469 (bp)->blk_phys_birth = 0; \ 470 (bp)->blk_birth = 0; \ 471 (bp)->blk_fill = 0; \ 472 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 473} 474 475#if BYTE_ORDER == _BIG_ENDIAN 476#define ZFS_HOST_BYTEORDER (0ULL) 477#else 478#define ZFS_HOST_BYTEORDER (1ULL) 479#endif 480 481#define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) 482#define BPE_NUM_WORDS 14 483#define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 484#define BPE_IS_PAYLOADWORD(bp, wp) \ 485 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 486 487/* 488 * Embedded checksum 489 */ 490#define ZEC_MAGIC 0x210da7ab10c7a11ULL 491 492typedef struct zio_eck { 493 uint64_t zec_magic; /* for validation, endianness */ 494 zio_cksum_t zec_cksum; /* 256-bit checksum */ 495} zio_eck_t; 496 497/* 498 * Gang block headers are self-checksumming and contain an array 499 * of block pointers. 500 */ 501#define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE 502#define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \ 503 sizeof (zio_eck_t)) / sizeof (blkptr_t)) 504#define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \ 505 sizeof (zio_eck_t) - \ 506 (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\ 507 sizeof (uint64_t)) 508 509typedef struct zio_gbh { 510 blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS]; 511 uint64_t zg_filler[SPA_GBH_FILLER]; 512 zio_eck_t zg_tail; 513} zio_gbh_phys_t; 514 515#define VDEV_RAIDZ_MAXPARITY 3 516 517#define VDEV_PAD_SIZE (8 << 10) 518/* 2 padding areas (vl_pad1 and vl_be) to skip */ 519#define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2 520#define VDEV_PHYS_SIZE (112 << 10) 521#define VDEV_UBERBLOCK_RING (128 << 10) 522 523/* 524 * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock 525 * ring when MMP is enabled. 526 */ 527#define MMP_BLOCKS_PER_LABEL 1 528 529/* The largest uberblock we support is 8k. */ 530#define MAX_UBERBLOCK_SHIFT (13) 531#define VDEV_UBERBLOCK_SHIFT(vd) \ 532 MIN(MAX((vd)->v_top->v_ashift, UBERBLOCK_SHIFT), MAX_UBERBLOCK_SHIFT) 533#define VDEV_UBERBLOCK_COUNT(vd) \ 534 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd)) 535#define VDEV_UBERBLOCK_OFFSET(vd, n) \ 536 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)]) 537#define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd)) 538 539typedef struct vdev_phys { 540 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)]; 541 zio_eck_t vp_zbt; 542} vdev_phys_t; 543 544typedef enum vbe_vers { 545 /* The bootenv file is stored as ascii text in the envblock */ 546 VB_RAW = 0, 547 548 /* 549 * The bootenv file is converted to an nvlist and then packed into the 550 * envblock. 551 */ 552 VB_NVLIST = 1 553} vbe_vers_t; 554 555typedef struct vdev_boot_envblock { 556 uint64_t vbe_version; 557 char vbe_bootenv[VDEV_PAD_SIZE - sizeof (uint64_t) - 558 sizeof (zio_eck_t)]; 559 zio_eck_t vbe_zbt; 560} vdev_boot_envblock_t; 561 562_Static_assert(sizeof (vdev_boot_envblock_t) == VDEV_PAD_SIZE, 563 "bad size for vdev_boot_envblock_t"); 564 565typedef struct vdev_label { 566 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */ 567 vdev_boot_envblock_t vl_be; /* 8K */ 568 vdev_phys_t vl_vdev_phys; /* 112K */ 569 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */ 570} vdev_label_t; /* 256K total */ 571 572/* 573 * vdev_dirty() flags 574 */ 575#define VDD_METASLAB 0x01 576#define VDD_DTL 0x02 577 578/* 579 * Size and offset of embedded boot loader region on each label. 580 * The total size of the first two labels plus the boot area is 4MB. 581 */ 582#define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t)) 583#define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */ 584 585/* 586 * Size of label regions at the start and end of each leaf device. 587 */ 588#define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE) 589#define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t)) 590#define VDEV_LABELS 4 591 592enum zio_checksum { 593 ZIO_CHECKSUM_INHERIT = 0, 594 ZIO_CHECKSUM_ON, 595 ZIO_CHECKSUM_OFF, 596 ZIO_CHECKSUM_LABEL, 597 ZIO_CHECKSUM_GANG_HEADER, 598 ZIO_CHECKSUM_ZILOG, 599 ZIO_CHECKSUM_FLETCHER_2, 600 ZIO_CHECKSUM_FLETCHER_4, 601 ZIO_CHECKSUM_SHA256, 602 ZIO_CHECKSUM_ZILOG2, 603 ZIO_CHECKSUM_NOPARITY, 604 ZIO_CHECKSUM_SHA512, 605 ZIO_CHECKSUM_SKEIN, 606 ZIO_CHECKSUM_EDONR, 607 ZIO_CHECKSUM_BLAKE3, 608 ZIO_CHECKSUM_FUNCTIONS 609}; 610 611#define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4 612#define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON 613 614enum zio_compress { 615 ZIO_COMPRESS_INHERIT = 0, 616 ZIO_COMPRESS_ON, 617 ZIO_COMPRESS_OFF, 618 ZIO_COMPRESS_LZJB, 619 ZIO_COMPRESS_EMPTY, 620 ZIO_COMPRESS_GZIP_1, 621 ZIO_COMPRESS_GZIP_2, 622 ZIO_COMPRESS_GZIP_3, 623 ZIO_COMPRESS_GZIP_4, 624 ZIO_COMPRESS_GZIP_5, 625 ZIO_COMPRESS_GZIP_6, 626 ZIO_COMPRESS_GZIP_7, 627 ZIO_COMPRESS_GZIP_8, 628 ZIO_COMPRESS_GZIP_9, 629 ZIO_COMPRESS_ZLE, 630 ZIO_COMPRESS_LZ4, 631 ZIO_COMPRESS_ZSTD, 632 ZIO_COMPRESS_FUNCTIONS 633}; 634 635enum zio_zstd_levels { 636 ZIO_ZSTD_LEVEL_INHERIT = 0, 637 ZIO_ZSTD_LEVEL_1, 638#define ZIO_ZSTD_LEVEL_MIN ZIO_ZSTD_LEVEL_1 639 ZIO_ZSTD_LEVEL_2, 640 ZIO_ZSTD_LEVEL_3, 641#define ZIO_ZSTD_LEVEL_DEFAULT ZIO_ZSTD_LEVEL_3 642 ZIO_ZSTD_LEVEL_4, 643 ZIO_ZSTD_LEVEL_5, 644 ZIO_ZSTD_LEVEL_6, 645 ZIO_ZSTD_LEVEL_7, 646 ZIO_ZSTD_LEVEL_8, 647 ZIO_ZSTD_LEVEL_9, 648 ZIO_ZSTD_LEVEL_10, 649 ZIO_ZSTD_LEVEL_11, 650 ZIO_ZSTD_LEVEL_12, 651 ZIO_ZSTD_LEVEL_13, 652 ZIO_ZSTD_LEVEL_14, 653 ZIO_ZSTD_LEVEL_15, 654 ZIO_ZSTD_LEVEL_16, 655 ZIO_ZSTD_LEVEL_17, 656 ZIO_ZSTD_LEVEL_18, 657 ZIO_ZSTD_LEVEL_19, 658#define ZIO_ZSTD_LEVEL_MAX ZIO_ZSTD_LEVEL_19 659 ZIO_ZSTD_LEVEL_RESERVE = 101, /* Leave room for new positive levels */ 660 ZIO_ZSTD_LEVEL_FAST, /* Fast levels are negative */ 661 ZIO_ZSTD_LEVEL_FAST_1, 662#define ZIO_ZSTD_LEVEL_FAST_DEFAULT ZIO_ZSTD_LEVEL_FAST_1 663 ZIO_ZSTD_LEVEL_FAST_2, 664 ZIO_ZSTD_LEVEL_FAST_3, 665 ZIO_ZSTD_LEVEL_FAST_4, 666 ZIO_ZSTD_LEVEL_FAST_5, 667 ZIO_ZSTD_LEVEL_FAST_6, 668 ZIO_ZSTD_LEVEL_FAST_7, 669 ZIO_ZSTD_LEVEL_FAST_8, 670 ZIO_ZSTD_LEVEL_FAST_9, 671 ZIO_ZSTD_LEVEL_FAST_10, 672 ZIO_ZSTD_LEVEL_FAST_20, 673 ZIO_ZSTD_LEVEL_FAST_30, 674 ZIO_ZSTD_LEVEL_FAST_40, 675 ZIO_ZSTD_LEVEL_FAST_50, 676 ZIO_ZSTD_LEVEL_FAST_60, 677 ZIO_ZSTD_LEVEL_FAST_70, 678 ZIO_ZSTD_LEVEL_FAST_80, 679 ZIO_ZSTD_LEVEL_FAST_90, 680 ZIO_ZSTD_LEVEL_FAST_100, 681 ZIO_ZSTD_LEVEL_FAST_500, 682 ZIO_ZSTD_LEVEL_FAST_1000, 683#define ZIO_ZSTD_LEVEL_FAST_MAX ZIO_ZSTD_LEVEL_FAST_1000 684 ZIO_ZSTD_LEVEL_AUTO = 251, /* Reserved for future use */ 685 ZIO_ZSTD_LEVEL_LEVELS 686}; 687 688#define ZIO_COMPRESS_ON_VALUE ZIO_COMPRESS_LZJB 689#define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF 690 691/* 692 * On-disk version number. 693 */ 694#define SPA_VERSION_1 1ULL 695#define SPA_VERSION_2 2ULL 696#define SPA_VERSION_3 3ULL 697#define SPA_VERSION_4 4ULL 698#define SPA_VERSION_5 5ULL 699#define SPA_VERSION_6 6ULL 700#define SPA_VERSION_7 7ULL 701#define SPA_VERSION_8 8ULL 702#define SPA_VERSION_9 9ULL 703#define SPA_VERSION_10 10ULL 704#define SPA_VERSION_11 11ULL 705#define SPA_VERSION_12 12ULL 706#define SPA_VERSION_13 13ULL 707#define SPA_VERSION_14 14ULL 708#define SPA_VERSION_15 15ULL 709#define SPA_VERSION_16 16ULL 710#define SPA_VERSION_17 17ULL 711#define SPA_VERSION_18 18ULL 712#define SPA_VERSION_19 19ULL 713#define SPA_VERSION_20 20ULL 714#define SPA_VERSION_21 21ULL 715#define SPA_VERSION_22 22ULL 716#define SPA_VERSION_23 23ULL 717#define SPA_VERSION_24 24ULL 718#define SPA_VERSION_25 25ULL 719#define SPA_VERSION_26 26ULL 720#define SPA_VERSION_27 27ULL 721#define SPA_VERSION_28 28ULL 722#define SPA_VERSION_5000 5000ULL 723 724/* 725 * When bumping up SPA_VERSION, make sure GRUB ZFS understands the on-disk 726 * format change. Go to usr/src/grub/grub-0.97/stage2/{zfs-include/, fsys_zfs*}, 727 * and do the appropriate changes. Also bump the version number in 728 * usr/src/grub/capability. 729 */ 730#define SPA_VERSION SPA_VERSION_5000 731#define SPA_VERSION_STRING "5000" 732 733/* 734 * Symbolic names for the changes that caused a SPA_VERSION switch. 735 * Used in the code when checking for presence or absence of a feature. 736 * Feel free to define multiple symbolic names for each version if there 737 * were multiple changes to on-disk structures during that version. 738 * 739 * NOTE: When checking the current SPA_VERSION in your code, be sure 740 * to use spa_version() since it reports the version of the 741 * last synced uberblock. Checking the in-flight version can 742 * be dangerous in some cases. 743 */ 744#define SPA_VERSION_INITIAL SPA_VERSION_1 745#define SPA_VERSION_DITTO_BLOCKS SPA_VERSION_2 746#define SPA_VERSION_SPARES SPA_VERSION_3 747#define SPA_VERSION_RAID6 SPA_VERSION_3 748#define SPA_VERSION_BPLIST_ACCOUNT SPA_VERSION_3 749#define SPA_VERSION_RAIDZ_DEFLATE SPA_VERSION_3 750#define SPA_VERSION_DNODE_BYTES SPA_VERSION_3 751#define SPA_VERSION_ZPOOL_HISTORY SPA_VERSION_4 752#define SPA_VERSION_GZIP_COMPRESSION SPA_VERSION_5 753#define SPA_VERSION_BOOTFS SPA_VERSION_6 754#define SPA_VERSION_SLOGS SPA_VERSION_7 755#define SPA_VERSION_DELEGATED_PERMS SPA_VERSION_8 756#define SPA_VERSION_FUID SPA_VERSION_9 757#define SPA_VERSION_REFRESERVATION SPA_VERSION_9 758#define SPA_VERSION_REFQUOTA SPA_VERSION_9 759#define SPA_VERSION_UNIQUE_ACCURATE SPA_VERSION_9 760#define SPA_VERSION_L2CACHE SPA_VERSION_10 761#define SPA_VERSION_NEXT_CLONES SPA_VERSION_11 762#define SPA_VERSION_ORIGIN SPA_VERSION_11 763#define SPA_VERSION_DSL_SCRUB SPA_VERSION_11 764#define SPA_VERSION_SNAP_PROPS SPA_VERSION_12 765#define SPA_VERSION_USED_BREAKDOWN SPA_VERSION_13 766#define SPA_VERSION_PASSTHROUGH_X SPA_VERSION_14 767#define SPA_VERSION_USERSPACE SPA_VERSION_15 768#define SPA_VERSION_STMF_PROP SPA_VERSION_16 769#define SPA_VERSION_RAIDZ3 SPA_VERSION_17 770#define SPA_VERSION_USERREFS SPA_VERSION_18 771#define SPA_VERSION_HOLES SPA_VERSION_19 772#define SPA_VERSION_ZLE_COMPRESSION SPA_VERSION_20 773#define SPA_VERSION_DEDUP SPA_VERSION_21 774#define SPA_VERSION_RECVD_PROPS SPA_VERSION_22 775#define SPA_VERSION_SLIM_ZIL SPA_VERSION_23 776#define SPA_VERSION_SA SPA_VERSION_24 777#define SPA_VERSION_SCAN SPA_VERSION_25 778#define SPA_VERSION_DIR_CLONES SPA_VERSION_26 779#define SPA_VERSION_DEADLISTS SPA_VERSION_26 780#define SPA_VERSION_FAST_SNAP SPA_VERSION_27 781#define SPA_VERSION_MULTI_REPLACE SPA_VERSION_28 782#define SPA_VERSION_BEFORE_FEATURES SPA_VERSION_28 783#define SPA_VERSION_FEATURES SPA_VERSION_5000 784 785#define SPA_VERSION_IS_SUPPORTED(v) \ 786 (((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \ 787 ((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION)) 788 789/* 790 * The following are configuration names used in the nvlist describing a pool's 791 * configuration. 792 */ 793#define ZPOOL_CONFIG_VERSION "version" 794#define ZPOOL_CONFIG_POOL_NAME "name" 795#define ZPOOL_CONFIG_POOL_STATE "state" 796#define ZPOOL_CONFIG_POOL_TXG "txg" 797#define ZPOOL_CONFIG_POOL_GUID "pool_guid" 798#define ZPOOL_CONFIG_CREATE_TXG "create_txg" 799#define ZPOOL_CONFIG_TOP_GUID "top_guid" 800#define ZPOOL_CONFIG_VDEV_TREE "vdev_tree" 801#define ZPOOL_CONFIG_TYPE "type" 802#define ZPOOL_CONFIG_CHILDREN "children" 803#define ZPOOL_CONFIG_ID "id" 804#define ZPOOL_CONFIG_GUID "guid" 805#define ZPOOL_CONFIG_INDIRECT_OBJECT "com.delphix:indirect_object" 806#define ZPOOL_CONFIG_INDIRECT_BIRTHS "com.delphix:indirect_births" 807#define ZPOOL_CONFIG_PREV_INDIRECT_VDEV "com.delphix:prev_indirect_vdev" 808#define ZPOOL_CONFIG_PATH "path" 809#define ZPOOL_CONFIG_DEVID "devid" 810#define ZPOOL_CONFIG_METASLAB_ARRAY "metaslab_array" 811#define ZPOOL_CONFIG_METASLAB_SHIFT "metaslab_shift" 812#define ZPOOL_CONFIG_ASHIFT "ashift" 813#define ZPOOL_CONFIG_ASIZE "asize" 814#define ZPOOL_CONFIG_DTL "DTL" 815#define ZPOOL_CONFIG_STATS "stats" 816#define ZPOOL_CONFIG_WHOLE_DISK "whole_disk" 817#define ZPOOL_CONFIG_ERRCOUNT "error_count" 818#define ZPOOL_CONFIG_NOT_PRESENT "not_present" 819#define ZPOOL_CONFIG_SPARES "spares" 820#define ZPOOL_CONFIG_IS_SPARE "is_spare" 821#define ZPOOL_CONFIG_NPARITY "nparity" 822#define ZPOOL_CONFIG_HOSTID "hostid" 823#define ZPOOL_CONFIG_HOSTNAME "hostname" 824#define ZPOOL_CONFIG_IS_LOG "is_log" 825#define ZPOOL_CONFIG_TIMESTAMP "timestamp" /* not stored on disk */ 826#define ZPOOL_CONFIG_FEATURES_FOR_READ "features_for_read" 827#define ZPOOL_CONFIG_VDEV_CHILDREN "vdev_children" 828 829/* 830 * The persistent vdev state is stored as separate values rather than a single 831 * 'vdev_state' entry. This is because a device can be in multiple states, such 832 * as offline and degraded. 833 */ 834#define ZPOOL_CONFIG_OFFLINE "offline" 835#define ZPOOL_CONFIG_FAULTED "faulted" 836#define ZPOOL_CONFIG_DEGRADED "degraded" 837#define ZPOOL_CONFIG_REMOVED "removed" 838#define ZPOOL_CONFIG_FRU "fru" 839#define ZPOOL_CONFIG_AUX_STATE "aux_state" 840 841#define VDEV_TYPE_ROOT "root" 842#define VDEV_TYPE_MIRROR "mirror" 843#define VDEV_TYPE_REPLACING "replacing" 844#define VDEV_TYPE_RAIDZ "raidz" 845#define VDEV_TYPE_DISK "disk" 846#define VDEV_TYPE_FILE "file" 847#define VDEV_TYPE_MISSING "missing" 848#define VDEV_TYPE_HOLE "hole" 849#define VDEV_TYPE_SPARE "spare" 850#define VDEV_TYPE_LOG "log" 851#define VDEV_TYPE_L2CACHE "l2cache" 852#define VDEV_TYPE_INDIRECT "indirect" 853 854/* 855 * This is needed in userland to report the minimum necessary device size. 856 */ 857#define SPA_MINDEVSIZE (64ULL << 20) 858 859/* 860 * The location of the pool configuration repository, shared between kernel and 861 * userland. 862 */ 863#define ZPOOL_CACHE "/boot/zfs/zpool.cache" 864 865/* 866 * vdev states are ordered from least to most healthy. 867 * A vdev that's CANT_OPEN or below is considered unusable. 868 */ 869typedef enum vdev_state { 870 VDEV_STATE_UNKNOWN = 0, /* Uninitialized vdev */ 871 VDEV_STATE_CLOSED, /* Not currently open */ 872 VDEV_STATE_OFFLINE, /* Not allowed to open */ 873 VDEV_STATE_REMOVED, /* Explicitly removed from system */ 874 VDEV_STATE_CANT_OPEN, /* Tried to open, but failed */ 875 VDEV_STATE_FAULTED, /* External request to fault device */ 876 VDEV_STATE_DEGRADED, /* Replicated vdev with unhealthy kids */ 877 VDEV_STATE_HEALTHY /* Presumed good */ 878} vdev_state_t; 879 880/* 881 * vdev aux states. When a vdev is in the CANT_OPEN state, the aux field 882 * of the vdev stats structure uses these constants to distinguish why. 883 */ 884typedef enum vdev_aux { 885 VDEV_AUX_NONE, /* no error */ 886 VDEV_AUX_OPEN_FAILED, /* ldi_open_*() or vn_open() failed */ 887 VDEV_AUX_CORRUPT_DATA, /* bad label or disk contents */ 888 VDEV_AUX_NO_REPLICAS, /* insufficient number of replicas */ 889 VDEV_AUX_BAD_GUID_SUM, /* vdev guid sum doesn't match */ 890 VDEV_AUX_TOO_SMALL, /* vdev size is too small */ 891 VDEV_AUX_BAD_LABEL, /* the label is OK but invalid */ 892 VDEV_AUX_VERSION_NEWER, /* on-disk version is too new */ 893 VDEV_AUX_VERSION_OLDER, /* on-disk version is too old */ 894 VDEV_AUX_SPARED /* hot spare used in another pool */ 895} vdev_aux_t; 896 897/* 898 * pool state. The following states are written to disk as part of the normal 899 * SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE. The remaining states are 900 * software abstractions used at various levels to communicate pool state. 901 */ 902typedef enum pool_state { 903 POOL_STATE_ACTIVE = 0, /* In active use */ 904 POOL_STATE_EXPORTED, /* Explicitly exported */ 905 POOL_STATE_DESTROYED, /* Explicitly destroyed */ 906 POOL_STATE_SPARE, /* Reserved for hot spare use */ 907 POOL_STATE_UNINITIALIZED, /* Internal spa_t state */ 908 POOL_STATE_UNAVAIL, /* Internal libzfs state */ 909 POOL_STATE_POTENTIALLY_ACTIVE /* Internal libzfs state */ 910} pool_state_t; 911 912/* 913 * The uberblock version is incremented whenever an incompatible on-disk 914 * format change is made to the SPA, DMU, or ZAP. 915 * 916 * Note: the first two fields should never be moved. When a storage pool 917 * is opened, the uberblock must be read off the disk before the version 918 * can be checked. If the ub_version field is moved, we may not detect 919 * version mismatch. If the ub_magic field is moved, applications that 920 * expect the magic number in the first word won't work. 921 */ 922#define UBERBLOCK_MAGIC 0x00bab10c /* oo-ba-bloc! */ 923#define UBERBLOCK_SHIFT 10 /* up to 1K */ 924 925#define MMP_MAGIC 0xa11cea11 /* all-see-all */ 926 927#define MMP_INTERVAL_VALID_BIT 0x01 928#define MMP_SEQ_VALID_BIT 0x02 929#define MMP_FAIL_INT_VALID_BIT 0x04 930 931#define MMP_VALID(ubp) (ubp->ub_magic == UBERBLOCK_MAGIC && \ 932 ubp->ub_mmp_magic == MMP_MAGIC) 933#define MMP_INTERVAL_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \ 934 MMP_INTERVAL_VALID_BIT)) 935#define MMP_SEQ_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \ 936 MMP_SEQ_VALID_BIT)) 937#define MMP_FAIL_INT_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \ 938 MMP_FAIL_INT_VALID_BIT)) 939 940#define MMP_INTERVAL(ubp) ((ubp->ub_mmp_config & 0x00000000FFFFFF00) \ 941 >> 8) 942#define MMP_SEQ(ubp) ((ubp->ub_mmp_config & 0x0000FFFF00000000) \ 943 >> 32) 944#define MMP_FAIL_INT(ubp) ((ubp->ub_mmp_config & 0xFFFF000000000000) \ 945 >> 48) 946 947typedef struct uberblock { 948 uint64_t ub_magic; /* UBERBLOCK_MAGIC */ 949 uint64_t ub_version; /* SPA_VERSION */ 950 uint64_t ub_txg; /* txg of last sync */ 951 uint64_t ub_guid_sum; /* sum of all vdev guids */ 952 uint64_t ub_timestamp; /* UTC time of last sync */ 953 blkptr_t ub_rootbp; /* MOS objset_phys_t */ 954 /* highest SPA_VERSION supported by software that wrote this txg */ 955 uint64_t ub_software_version; 956 /* Maybe missing in uberblocks we read, but always written */ 957 uint64_t ub_mmp_magic; 958 /* 959 * If ub_mmp_delay == 0 and ub_mmp_magic is valid, MMP is off. 960 * Otherwise, nanosec since last MMP write. 961 */ 962 uint64_t ub_mmp_delay; 963 964 /* 965 * The ub_mmp_config contains the multihost write interval, multihost 966 * fail intervals, sequence number for sub-second granularity, and 967 * valid bit mask. This layout is as follows: 968 * 969 * 64 56 48 40 32 24 16 8 0 970 * +-------+-------+-------+-------+-------+-------+-------+-------+ 971 * 0 | Fail Intervals| Seq | Write Interval (ms) | VALID | 972 * +-------+-------+-------+-------+-------+-------+-------+-------+ 973 * 974 * This allows a write_interval of (2^24/1000)s, over 4.5 hours 975 * 976 * VALID Bits: 977 * - 0x01 - Write Interval (ms) 978 * - 0x02 - Sequence number exists 979 * - 0x04 - Fail Intervals 980 * - 0xf8 - Reserved 981 */ 982 uint64_t ub_mmp_config; 983 984 /* 985 * ub_checkpoint_txg indicates two things about the current uberblock: 986 * 987 * 1] If it is not zero then this uberblock is a checkpoint. If it is 988 * zero, then this uberblock is not a checkpoint. 989 * 990 * 2] On checkpointed uberblocks, the value of ub_checkpoint_txg is 991 * the ub_txg that the uberblock had at the time we moved it to 992 * the MOS config. 993 * 994 * The field is set when we checkpoint the uberblock and continues to 995 * hold that value even after we've rewound (unlike the ub_txg that 996 * is reset to a higher value). 997 * 998 * Besides checks used to determine whether we are reopening the 999 * pool from a checkpointed uberblock [see spa_ld_select_uberblock()], 1000 * the value of the field is used to determine which ZIL blocks have 1001 * been allocated according to the ms_sm when we are rewinding to a 1002 * checkpoint. Specifically, if blk_birth > ub_checkpoint_txg, then 1003 * the ZIL block is not allocated [see uses of spa_min_claim_txg()]. 1004 */ 1005 uint64_t ub_checkpoint_txg; 1006} uberblock_t; 1007 1008/* 1009 * Flags. 1010 */ 1011#define DNODE_MUST_BE_ALLOCATED 1 1012#define DNODE_MUST_BE_FREE 2 1013 1014/* 1015 * Fixed constants. 1016 */ 1017#define DNODE_SHIFT 9 /* 512 bytes */ 1018#define DN_MIN_INDBLKSHIFT 12 /* 4k */ 1019#define DN_MAX_INDBLKSHIFT 17 /* 128k */ 1020#define DNODE_BLOCK_SHIFT 14 /* 16k */ 1021#define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */ 1022#define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */ 1023#define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */ 1024 1025/* 1026 * Derived constants. 1027 */ 1028#define DNODE_MIN_SIZE (1 << DNODE_SHIFT) 1029#define DNODE_MAX_SIZE (1 << DNODE_BLOCK_SHIFT) 1030#define DNODE_BLOCK_SIZE (1 << DNODE_BLOCK_SHIFT) 1031#define DNODE_MIN_SLOTS (DNODE_MIN_SIZE >> DNODE_SHIFT) 1032#define DNODE_MAX_SLOTS (DNODE_MAX_SIZE >> DNODE_SHIFT) 1033#define DN_BONUS_SIZE(dnsize) ((dnsize) - DNODE_CORE_SIZE - \ 1034 (1 << SPA_BLKPTRSHIFT)) 1035#define DN_SLOTS_TO_BONUSLEN(slots) DN_BONUS_SIZE((slots) << DNODE_SHIFT) 1036#define DN_OLD_MAX_BONUSLEN (DN_BONUS_SIZE(DNODE_MIN_SIZE)) 1037#define DN_MAX_NBLKPTR ((DNODE_MIN_SIZE - DNODE_CORE_SIZE) >> \ 1038 SPA_BLKPTRSHIFT) 1039#define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT) 1040#define DN_ZERO_BONUSLEN (DN_BONUS_SIZE(DNODE_MAX_SIZE) + 1) 1041 1042#define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT) 1043#define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT) 1044#define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT) 1045 1046/* The +2 here is a cheesy way to round up */ 1047#define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \ 1048 (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT))) 1049 1050#define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \ 1051 (((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t)))) 1052 1053#define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \ 1054 (dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT) 1055 1056#define EPB(blkshift, typeshift) (1 << (blkshift - typeshift)) 1057 1058/* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */ 1059#define DNODE_FLAG_USED_BYTES (1<<0) 1060#define DNODE_FLAG_USERUSED_ACCOUNTED (1<<1) 1061 1062/* Does dnode have a SA spill blkptr in bonus? */ 1063#define DNODE_FLAG_SPILL_BLKPTR (1<<2) 1064 1065typedef struct dnode_phys { 1066 uint8_t dn_type; /* dmu_object_type_t */ 1067 uint8_t dn_indblkshift; /* ln2(indirect block size) */ 1068 uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */ 1069 uint8_t dn_nblkptr; /* length of dn_blkptr */ 1070 uint8_t dn_bonustype; /* type of data in bonus buffer */ 1071 uint8_t dn_checksum; /* ZIO_CHECKSUM type */ 1072 uint8_t dn_compress; /* ZIO_COMPRESS type */ 1073 uint8_t dn_flags; /* DNODE_FLAG_* */ 1074 uint16_t dn_datablkszsec; /* data block size in 512b sectors */ 1075 uint16_t dn_bonuslen; /* length of dn_bonus */ 1076 uint8_t dn_extra_slots; /* # of subsequent slots consumed */ 1077 uint8_t dn_pad2[3]; 1078 1079 /* accounting is protected by dn_dirty_mtx */ 1080 uint64_t dn_maxblkid; /* largest allocated block ID */ 1081 uint64_t dn_used; /* bytes (or sectors) of disk space */ 1082 1083 uint64_t dn_pad3[4]; 1084 1085 /* 1086 * The tail region is 448 bytes for a 512 byte dnode, and 1087 * correspondingly larger for larger dnode sizes. The spill 1088 * block pointer, when present, is always at the end of the tail 1089 * region. There are three ways this space may be used, using 1090 * a 512 byte dnode for this diagram: 1091 * 1092 * 0 64 128 192 256 320 384 448 (offset) 1093 * +---------------+---------------+---------------+-------+ 1094 * | dn_blkptr[0] | dn_blkptr[1] | dn_blkptr[2] | / | 1095 * +---------------+---------------+---------------+-------+ 1096 * | dn_blkptr[0] | dn_bonus[0..319] | 1097 * +---------------+-----------------------+---------------+ 1098 * | dn_blkptr[0] | dn_bonus[0..191] | dn_spill | 1099 * +---------------+-----------------------+---------------+ 1100 */ 1101 union { 1102 blkptr_t dn_blkptr[1+DN_OLD_MAX_BONUSLEN/sizeof (blkptr_t)]; 1103 struct { 1104 blkptr_t __dn_ignore1; 1105 uint8_t dn_bonus[DN_OLD_MAX_BONUSLEN]; 1106 }; 1107 struct { 1108 blkptr_t __dn_ignore2; 1109 uint8_t __dn_ignore3[DN_OLD_MAX_BONUSLEN - 1110 sizeof (blkptr_t)]; 1111 blkptr_t dn_spill; 1112 }; 1113 }; 1114} dnode_phys_t; 1115 1116#define DN_SPILL_BLKPTR(dnp) (blkptr_t *)((char *)(dnp) + \ 1117 (((dnp)->dn_extra_slots + 1) << DNODE_SHIFT) - (1 << SPA_BLKPTRSHIFT)) 1118 1119typedef enum dmu_object_byteswap { 1120 DMU_BSWAP_UINT8, 1121 DMU_BSWAP_UINT16, 1122 DMU_BSWAP_UINT32, 1123 DMU_BSWAP_UINT64, 1124 DMU_BSWAP_ZAP, 1125 DMU_BSWAP_DNODE, 1126 DMU_BSWAP_OBJSET, 1127 DMU_BSWAP_ZNODE, 1128 DMU_BSWAP_OLDACL, 1129 DMU_BSWAP_ACL, 1130 /* 1131 * Allocating a new byteswap type number makes the on-disk format 1132 * incompatible with any other format that uses the same number. 1133 * 1134 * Data can usually be structured to work with one of the 1135 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. 1136 */ 1137 DMU_BSWAP_NUMFUNCS 1138} dmu_object_byteswap_t; 1139 1140#define DMU_OT_NEWTYPE 0x80 1141#define DMU_OT_METADATA 0x40 1142#define DMU_OT_BYTESWAP_MASK 0x3f 1143 1144/* 1145 * Defines a uint8_t object type. Object types specify if the data 1146 * in the object is metadata (boolean) and how to byteswap the data 1147 * (dmu_object_byteswap_t). 1148 */ 1149#define DMU_OT(byteswap, metadata) \ 1150 (DMU_OT_NEWTYPE | \ 1151 ((metadata) ? DMU_OT_METADATA : 0) | \ 1152 ((byteswap) & DMU_OT_BYTESWAP_MASK)) 1153 1154typedef enum dmu_object_type { 1155 DMU_OT_NONE, 1156 /* general: */ 1157 DMU_OT_OBJECT_DIRECTORY, /* ZAP */ 1158 DMU_OT_OBJECT_ARRAY, /* UINT64 */ 1159 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ 1160 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ 1161 DMU_OT_BPOBJ, /* UINT64 */ 1162 DMU_OT_BPOBJ_HDR, /* UINT64 */ 1163 /* spa: */ 1164 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ 1165 DMU_OT_SPACE_MAP, /* UINT64 */ 1166 /* zil: */ 1167 DMU_OT_INTENT_LOG, /* UINT64 */ 1168 /* dmu: */ 1169 DMU_OT_DNODE, /* DNODE */ 1170 DMU_OT_OBJSET, /* OBJSET */ 1171 /* dsl: */ 1172 DMU_OT_DSL_DIR, /* UINT64 */ 1173 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ 1174 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ 1175 DMU_OT_DSL_PROPS, /* ZAP */ 1176 DMU_OT_DSL_DATASET, /* UINT64 */ 1177 /* zpl: */ 1178 DMU_OT_ZNODE, /* ZNODE */ 1179 DMU_OT_OLDACL, /* Old ACL */ 1180 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ 1181 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ 1182 DMU_OT_MASTER_NODE, /* ZAP */ 1183 DMU_OT_UNLINKED_SET, /* ZAP */ 1184 /* zvol: */ 1185 DMU_OT_ZVOL, /* UINT8 */ 1186 DMU_OT_ZVOL_PROP, /* ZAP */ 1187 /* other; for testing only! */ 1188 DMU_OT_PLAIN_OTHER, /* UINT8 */ 1189 DMU_OT_UINT64_OTHER, /* UINT64 */ 1190 DMU_OT_ZAP_OTHER, /* ZAP */ 1191 /* new object types: */ 1192 DMU_OT_ERROR_LOG, /* ZAP */ 1193 DMU_OT_SPA_HISTORY, /* UINT8 */ 1194 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ 1195 DMU_OT_POOL_PROPS, /* ZAP */ 1196 DMU_OT_DSL_PERMS, /* ZAP */ 1197 DMU_OT_ACL, /* ACL */ 1198 DMU_OT_SYSACL, /* SYSACL */ 1199 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ 1200 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ 1201 DMU_OT_NEXT_CLONES, /* ZAP */ 1202 DMU_OT_SCAN_QUEUE, /* ZAP */ 1203 DMU_OT_USERGROUP_USED, /* ZAP */ 1204 DMU_OT_USERGROUP_QUOTA, /* ZAP */ 1205 DMU_OT_USERREFS, /* ZAP */ 1206 DMU_OT_DDT_ZAP, /* ZAP */ 1207 DMU_OT_DDT_STATS, /* ZAP */ 1208 DMU_OT_SA, /* System attr */ 1209 DMU_OT_SA_MASTER_NODE, /* ZAP */ 1210 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ 1211 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ 1212 DMU_OT_SCAN_XLATE, /* ZAP */ 1213 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ 1214 DMU_OT_DEADLIST, /* ZAP */ 1215 DMU_OT_DEADLIST_HDR, /* UINT64 */ 1216 DMU_OT_DSL_CLONES, /* ZAP */ 1217 DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */ 1218 DMU_OT_NUMTYPES, 1219 1220 /* 1221 * Names for valid types declared with DMU_OT(). 1222 */ 1223 DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE), 1224 DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE), 1225 DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE), 1226 DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE), 1227 DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE), 1228 DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE), 1229 DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE), 1230 DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE), 1231 DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE), 1232 DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE) 1233} dmu_object_type_t; 1234 1235typedef enum dmu_objset_type { 1236 DMU_OST_NONE, 1237 DMU_OST_META, 1238 DMU_OST_ZFS, 1239 DMU_OST_ZVOL, 1240 DMU_OST_OTHER, /* For testing only! */ 1241 DMU_OST_ANY, /* Be careful! */ 1242 DMU_OST_NUMTYPES 1243} dmu_objset_type_t; 1244 1245#define ZAP_MAXVALUELEN (1024 * 8) 1246 1247/* 1248 * header for all bonus and spill buffers. 1249 * The header has a fixed portion with a variable number 1250 * of "lengths" depending on the number of variable sized 1251 * attribues which are determined by the "layout number" 1252 */ 1253 1254#define SA_MAGIC 0x2F505A /* ZFS SA */ 1255typedef struct sa_hdr_phys { 1256 uint32_t sa_magic; 1257 uint16_t sa_layout_info; /* Encoded with hdrsize and layout number */ 1258 uint16_t sa_lengths[1]; /* optional sizes for variable length attrs */ 1259 /* ... Data follows the lengths. */ 1260} sa_hdr_phys_t; 1261 1262/* 1263 * sa_hdr_phys -> sa_layout_info 1264 * 1265 * 16 10 0 1266 * +--------+-------+ 1267 * | hdrsz |layout | 1268 * +--------+-------+ 1269 * 1270 * Bits 0-10 are the layout number 1271 * Bits 11-16 are the size of the header. 1272 * The hdrsize is the number * 8 1273 * 1274 * For example. 1275 * hdrsz of 1 ==> 8 byte header 1276 * 2 ==> 16 byte header 1277 * 1278 */ 1279 1280#define SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10) 1281#define SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 16, 3, 0) 1282#define SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \ 1283{ \ 1284 BF32_SET_SB(x, 10, 6, 3, 0, size); \ 1285 BF32_SET(x, 0, 10, num); \ 1286} 1287 1288#define SA_ATTR_BSWAP(x) BF32_GET(x, 16, 8) 1289#define SA_ATTR_LENGTH(x) BF32_GET(x, 24, 16) 1290#define SA_ATTR_NUM(x) BF32_GET(x, 0, 16) 1291#define SA_ATTR_ENCODE(x, attr, length, bswap) \ 1292{ \ 1293 BF64_SET(x, 24, 16, length); \ 1294 BF64_SET(x, 16, 8, bswap); \ 1295 BF64_SET(x, 0, 16, attr); \ 1296} 1297 1298#define SA_MODE_OFFSET 0 1299#define SA_SIZE_OFFSET 8 1300#define SA_GEN_OFFSET 16 1301#define SA_UID_OFFSET 24 1302#define SA_GID_OFFSET 32 1303#define SA_PARENT_OFFSET 40 1304#define SA_SYMLINK_OFFSET 160 1305 1306#define SA_REGISTRY "REGISTRY" 1307#define SA_LAYOUTS "LAYOUTS" 1308 1309typedef enum sa_bswap_type { 1310 SA_UINT64_ARRAY, 1311 SA_UINT32_ARRAY, 1312 SA_UINT16_ARRAY, 1313 SA_UINT8_ARRAY, 1314 SA_ACL, 1315} sa_bswap_type_t; 1316 1317typedef uint16_t sa_attr_type_t; 1318 1319#define ZIO_OBJSET_MAC_LEN 32 1320 1321/* 1322 * Intent log header - this on disk structure holds fields to manage 1323 * the log. All fields are 64 bit to easily handle cross architectures. 1324 */ 1325typedef struct zil_header { 1326 uint64_t zh_claim_txg; /* txg in which log blocks were claimed */ 1327 uint64_t zh_replay_seq; /* highest replayed sequence number */ 1328 blkptr_t zh_log; /* log chain */ 1329 uint64_t zh_claim_seq; /* highest claimed sequence number */ 1330 uint64_t zh_pad[5]; 1331} zil_header_t; 1332 1333#define OBJSET_PHYS_SIZE_V2 2048 1334#define OBJSET_PHYS_SIZE_V3 4096 1335 1336typedef struct objset_phys { 1337 dnode_phys_t os_meta_dnode; 1338 zil_header_t os_zil_header; 1339 uint64_t os_type; 1340 uint64_t os_flags; 1341 uint8_t os_portable_mac[ZIO_OBJSET_MAC_LEN]; 1342 uint8_t os_local_mac[ZIO_OBJSET_MAC_LEN]; 1343 char os_pad0[OBJSET_PHYS_SIZE_V2 - sizeof (dnode_phys_t)*3 - 1344 sizeof (zil_header_t) - sizeof (uint64_t)*2 - 1345 2*ZIO_OBJSET_MAC_LEN]; 1346 dnode_phys_t os_userused_dnode; 1347 dnode_phys_t os_groupused_dnode; 1348 dnode_phys_t os_projectused_dnode; 1349 char os_pad1[OBJSET_PHYS_SIZE_V3 - OBJSET_PHYS_SIZE_V2 - 1350 sizeof (dnode_phys_t)]; 1351} objset_phys_t; 1352 1353typedef struct space_map_phys { 1354 /* object number: not needed but kept for backwards compatibility */ 1355 uint64_t smp_object; 1356 1357 /* length of the object in bytes */ 1358 uint64_t smp_length; 1359 1360 /* space allocated from the map */ 1361 int64_t smp_alloc; 1362} space_map_phys_t; 1363 1364typedef enum { 1365 SM_ALLOC, 1366 SM_FREE 1367} maptype_t; 1368 1369/* one-word entry constants */ 1370#define SM_DEBUG_PREFIX 2 1371#define SM_OFFSET_BITS 47 1372#define SM_RUN_BITS 15 1373 1374/* two-word entry constants */ 1375#define SM2_PREFIX 3 1376#define SM2_OFFSET_BITS 63 1377#define SM2_RUN_BITS 36 1378 1379#define SM_PREFIX_DECODE(x) BF64_DECODE(x, 62, 2) 1380#define SM_PREFIX_ENCODE(x) BF64_ENCODE(x, 62, 2) 1381 1382#define SM_DEBUG_ACTION_DECODE(x) BF64_DECODE(x, 60, 2) 1383#define SM_DEBUG_ACTION_ENCODE(x) BF64_ENCODE(x, 60, 2) 1384#define SM_DEBUG_SYNCPASS_DECODE(x) BF64_DECODE(x, 50, 10) 1385#define SM_DEBUG_SYNCPASS_ENCODE(x) BF64_ENCODE(x, 50, 10) 1386#define SM_DEBUG_TXG_DECODE(x) BF64_DECODE(x, 0, 50) 1387#define SM_DEBUG_TXG_ENCODE(x) BF64_ENCODE(x, 0, 50) 1388 1389#define SM_OFFSET_DECODE(x) BF64_DECODE(x, 16, SM_OFFSET_BITS) 1390#define SM_OFFSET_ENCODE(x) BF64_ENCODE(x, 16, SM_OFFSET_BITS) 1391#define SM_TYPE_DECODE(x) BF64_DECODE(x, 15, 1) 1392#define SM_TYPE_ENCODE(x) BF64_ENCODE(x, 15, 1) 1393#define SM_RUN_DECODE(x) (BF64_DECODE(x, 0, SM_RUN_BITS) + 1) 1394#define SM_RUN_ENCODE(x) BF64_ENCODE((x) - 1, 0, SM_RUN_BITS) 1395#define SM_RUN_MAX SM_RUN_DECODE(~0ULL) 1396#define SM_OFFSET_MAX SM_OFFSET_DECODE(~0ULL) 1397 1398#define SM2_RUN_DECODE(x) (BF64_DECODE(x, 24, SM2_RUN_BITS) + 1) 1399#define SM2_RUN_ENCODE(x) BF64_ENCODE((x) - 1, 24, SM2_RUN_BITS) 1400#define SM2_VDEV_DECODE(x) BF64_DECODE(x, 0, 24) 1401#define SM2_VDEV_ENCODE(x) BF64_ENCODE(x, 0, 24) 1402#define SM2_TYPE_DECODE(x) BF64_DECODE(x, SM2_OFFSET_BITS, 1) 1403#define SM2_TYPE_ENCODE(x) BF64_ENCODE(x, SM2_OFFSET_BITS, 1) 1404#define SM2_OFFSET_DECODE(x) BF64_DECODE(x, 0, SM2_OFFSET_BITS) 1405#define SM2_OFFSET_ENCODE(x) BF64_ENCODE(x, 0, SM2_OFFSET_BITS) 1406#define SM2_RUN_MAX SM2_RUN_DECODE(~0ULL) 1407#define SM2_OFFSET_MAX SM2_OFFSET_DECODE(~0ULL) 1408 1409typedef enum dd_used { 1410 DD_USED_HEAD, 1411 DD_USED_SNAP, 1412 DD_USED_CHILD, 1413 DD_USED_CHILD_RSRV, 1414 DD_USED_REFRSRV, 1415 DD_USED_NUM 1416} dd_used_t; 1417 1418#define DD_FLAG_USED_BREAKDOWN (1 << 0) 1419 1420typedef struct dsl_dir_phys { 1421 uint64_t dd_creation_time; /* not actually used */ 1422 uint64_t dd_head_dataset_obj; 1423 uint64_t dd_parent_obj; 1424 uint64_t dd_clone_parent_obj; 1425 uint64_t dd_child_dir_zapobj; 1426 /* 1427 * how much space our children are accounting for; for leaf 1428 * datasets, == physical space used by fs + snaps 1429 */ 1430 uint64_t dd_used_bytes; 1431 uint64_t dd_compressed_bytes; 1432 uint64_t dd_uncompressed_bytes; 1433 /* Administrative quota setting */ 1434 uint64_t dd_quota; 1435 /* Administrative reservation setting */ 1436 uint64_t dd_reserved; 1437 uint64_t dd_props_zapobj; 1438 uint64_t dd_pad[1]; 1439 uint64_t dd_flags; 1440 uint64_t dd_used_breakdown[DD_USED_NUM]; 1441 uint64_t dd_clones; 1442 uint64_t dd_pad1[13]; /* pad out to 256 bytes for good measure */ 1443} dsl_dir_phys_t; 1444 1445typedef struct dsl_dataset_phys { 1446 uint64_t ds_dir_obj; 1447 uint64_t ds_prev_snap_obj; 1448 uint64_t ds_prev_snap_txg; 1449 uint64_t ds_next_snap_obj; 1450 uint64_t ds_snapnames_zapobj; /* zap obj of snaps; ==0 for snaps */ 1451 uint64_t ds_num_children; /* clone/snap children; ==0 for head */ 1452 uint64_t ds_creation_time; /* seconds since 1970 */ 1453 uint64_t ds_creation_txg; 1454 uint64_t ds_deadlist_obj; 1455 uint64_t ds_used_bytes; 1456 uint64_t ds_compressed_bytes; 1457 uint64_t ds_uncompressed_bytes; 1458 uint64_t ds_unique_bytes; /* only relevant to snapshots */ 1459 /* 1460 * The ds_fsid_guid is a 56-bit ID that can change to avoid 1461 * collisions. The ds_guid is a 64-bit ID that will never 1462 * change, so there is a small probability that it will collide. 1463 */ 1464 uint64_t ds_fsid_guid; 1465 uint64_t ds_guid; 1466 uint64_t ds_flags; 1467 blkptr_t ds_bp; 1468 uint64_t ds_next_clones_obj; /* DMU_OT_DSL_CLONES */ 1469 uint64_t ds_props_obj; /* DMU_OT_DSL_PROPS for snaps */ 1470 uint64_t ds_userrefs_obj; /* DMU_OT_USERREFS */ 1471 uint64_t ds_pad[5]; /* pad out to 320 bytes for good measure */ 1472} dsl_dataset_phys_t; 1473 1474typedef struct dsl_deadlist_phys { 1475 uint64_t dl_used; 1476 uint64_t dl_comp; 1477 uint64_t dl_uncomp; 1478 uint64_t dl_pad[37]; /* pad out to 320b for future expansion */ 1479} dsl_deadlist_phys_t; 1480 1481#define BPOBJ_SIZE_V2 (6 * sizeof (uint64_t)) 1482 1483typedef struct bpobj_phys { 1484 uint64_t bpo_num_blkptrs; 1485 uint64_t bpo_bytes; 1486 uint64_t bpo_comp; 1487 uint64_t bpo_uncomp; 1488 uint64_t bpo_subobjs; 1489 uint64_t bpo_num_subobjs; 1490 uint64_t bpo_num_freed; 1491} bpobj_phys_t; 1492 1493/* 1494 * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 1495 */ 1496#define DMU_POOL_DIRECTORY_OBJECT 1 1497#define DMU_POOL_CONFIG "config" 1498#define DMU_POOL_FEATURES_FOR_READ "features_for_read" 1499#define DMU_POOL_FEATURES_FOR_WRITE "features_for_write" 1500#define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions" 1501#define DMU_POOL_ROOT_DATASET "root_dataset" 1502#define DMU_POOL_SYNC_BPLIST "sync_bplist" 1503#define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 1504#define DMU_POOL_ERRLOG_LAST "errlog_last" 1505#define DMU_POOL_SPARES "spares" 1506#define DMU_POOL_DEFLATE "deflate" 1507#define DMU_POOL_HISTORY "history" 1508#define DMU_POOL_PROPS "pool_props" 1509#define DMU_POOL_FREE_BPOBJ "free_bpobj" 1510#define DMU_POOL_BPTREE_OBJ "bptree_obj" 1511#define DMU_POOL_EMPTY_BPOBJ "empty_bpobj" 1512#define DMU_POOL_TMP_USERREFS "tmp_userrefs" 1513#define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" 1514#define DMU_POOL_REMOVING "com.delphix:removing" 1515#define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj" 1516#define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect" 1517#define DMU_POOL_ZPOOL_CHECKPOINT "com.delphix:zpool_checkpoint" 1518 1519#define ZAP_MAGIC 0x2F52AB2ABULL 1520 1521#define FZAP_BLOCK_SHIFT(zap) ((zap)->zap_block_shift) 1522 1523#define ZAP_MAXCD (uint32_t)(-1) 1524#define ZAP_HASHBITS 28 1525#define MZAP_ENT_LEN 64 1526#define MZAP_ENT_MAX \ 1527 ((MZAP_MAX_BLKSZ - sizeof(mzap_phys_t)) / sizeof(mzap_ent_phys_t) + 1) 1528#define MZAP_NAME_LEN (MZAP_ENT_LEN - 8 - 4 - 2) 1529#define MZAP_MAX_BLKSZ SPA_OLDMAXBLOCKSIZE 1530 1531typedef struct mzap_ent_phys { 1532 uint64_t mze_value; 1533 uint32_t mze_cd; 1534 uint16_t mze_pad; /* in case we want to chain them someday */ 1535 char mze_name[MZAP_NAME_LEN]; 1536} mzap_ent_phys_t; 1537 1538typedef struct mzap_phys { 1539 uint64_t mz_block_type; /* ZBT_MICRO */ 1540 uint64_t mz_salt; 1541 uint64_t mz_normflags; 1542 uint64_t mz_pad[5]; 1543 mzap_ent_phys_t mz_chunk[1]; 1544 /* actually variable size depending on block size */ 1545} mzap_phys_t; 1546 1547/* 1548 * The (fat) zap is stored in one object. It is an array of 1549 * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of: 1550 * 1551 * ptrtbl fits in first block: 1552 * [zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ... 1553 * 1554 * ptrtbl too big for first block: 1555 * [zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ... 1556 * 1557 */ 1558 1559#define ZBT_LEAF ((1ULL << 63) + 0) 1560#define ZBT_HEADER ((1ULL << 63) + 1) 1561#define ZBT_MICRO ((1ULL << 63) + 3) 1562/* any other values are ptrtbl blocks */ 1563 1564/* 1565 * the embedded pointer table takes up half a block: 1566 * block size / entry size (2^3) / 2 1567 */ 1568#define ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1) 1569 1570/* 1571 * The embedded pointer table starts half-way through the block. Since 1572 * the pointer table itself is half the block, it starts at (64-bit) 1573 * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)). 1574 */ 1575#define ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \ 1576 ((uint64_t *)(zap)->zap_phys) \ 1577 [(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))] 1578 1579#define ZAP_HASH_IDX(hash, n) (((n) == 0) ? 0 : ((hash) >> (64 - (n)))) 1580 1581/* 1582 * TAKE NOTE: 1583 * If zap_phys_t is modified, zap_byteswap() must be modified. 1584 */ 1585typedef struct zap_phys { 1586 uint64_t zap_block_type; /* ZBT_HEADER */ 1587 uint64_t zap_magic; /* ZAP_MAGIC */ 1588 1589 struct zap_table_phys { 1590 uint64_t zt_blk; /* starting block number */ 1591 uint64_t zt_numblks; /* number of blocks */ 1592 uint64_t zt_shift; /* bits to index it */ 1593 uint64_t zt_nextblk; /* next (larger) copy start block */ 1594 uint64_t zt_blks_copied; /* number source blocks copied */ 1595 } zap_ptrtbl; 1596 1597 uint64_t zap_freeblk; /* the next free block */ 1598 uint64_t zap_num_leafs; /* number of leafs */ 1599 uint64_t zap_num_entries; /* number of entries */ 1600 uint64_t zap_salt; /* salt to stir into hash function */ 1601 uint64_t zap_normflags; /* flags for u8_textprep_str() */ 1602 uint64_t zap_flags; /* zap_flags_t */ 1603 /* 1604 * This structure is followed by padding, and then the embedded 1605 * pointer table. The embedded pointer table takes up second 1606 * half of the block. It is accessed using the 1607 * ZAP_EMBEDDED_PTRTBL_ENT() macro. 1608 */ 1609} zap_phys_t; 1610 1611typedef struct zap_table_phys zap_table_phys_t; 1612 1613struct spa; 1614typedef struct fat_zap { 1615 int zap_block_shift; /* block size shift */ 1616 zap_phys_t *zap_phys; 1617 const struct spa *zap_spa; 1618 const dnode_phys_t *zap_dnode; 1619} fat_zap_t; 1620 1621#define ZAP_LEAF_MAGIC 0x2AB1EAF 1622 1623/* chunk size = 24 bytes */ 1624#define ZAP_LEAF_CHUNKSIZE 24 1625 1626/* 1627 * The amount of space available for chunks is: 1628 * block size (1<<l->l_bs) - hash entry size (2) * number of hash 1629 * entries - header space (2*chunksize) 1630 */ 1631#define ZAP_LEAF_NUMCHUNKS(l) \ 1632 (((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \ 1633 ZAP_LEAF_CHUNKSIZE - 2) 1634 1635/* 1636 * The amount of space within the chunk available for the array is: 1637 * chunk size - space for type (1) - space for next pointer (2) 1638 */ 1639#define ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3) 1640 1641#define ZAP_LEAF_ARRAY_NCHUNKS(bytes) \ 1642 (((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES) 1643 1644/* 1645 * Low water mark: when there are only this many chunks free, start 1646 * growing the ptrtbl. Ideally, this should be larger than a 1647 * "reasonably-sized" entry. 20 chunks is more than enough for the 1648 * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value), 1649 * while still being only around 3% for 16k blocks. 1650 */ 1651#define ZAP_LEAF_LOW_WATER (20) 1652 1653/* 1654 * The leaf hash table has block size / 2^5 (32) number of entries, 1655 * which should be more than enough for the maximum number of entries, 1656 * which is less than block size / CHUNKSIZE (24) / minimum number of 1657 * chunks per entry (3). 1658 */ 1659#define ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5) 1660#define ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l)) 1661 1662/* 1663 * The chunks start immediately after the hash table. The end of the 1664 * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a 1665 * chunk_t. 1666 */ 1667#define ZAP_LEAF_CHUNK(l, idx) \ 1668 ((zap_leaf_chunk_t *)(void *) \ 1669 ((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx] 1670#define ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry) 1671 1672#define ZAP_LEAF_HASH(l, h) \ 1673 ((ZAP_LEAF_HASH_NUMENTRIES(l)-1) & \ 1674 ((h) >> \ 1675 (64 - ZAP_LEAF_HASH_SHIFT(l) - (l)->l_phys->l_hdr.lh_prefix_len))) 1676#define ZAP_LEAF_HASH_ENTPTR(l, h) (&(l)->l_phys->l_hash[ZAP_LEAF_HASH(l, h)]) 1677 1678typedef enum zap_chunk_type { 1679 ZAP_CHUNK_FREE = 253, 1680 ZAP_CHUNK_ENTRY = 252, 1681 ZAP_CHUNK_ARRAY = 251, 1682 ZAP_CHUNK_TYPE_MAX = 250 1683} zap_chunk_type_t; 1684 1685/* 1686 * TAKE NOTE: 1687 * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified. 1688 */ 1689typedef struct zap_leaf_phys { 1690 struct zap_leaf_header { 1691 uint64_t lh_block_type; /* ZBT_LEAF */ 1692 uint64_t lh_pad1; 1693 uint64_t lh_prefix; /* hash prefix of this leaf */ 1694 uint32_t lh_magic; /* ZAP_LEAF_MAGIC */ 1695 uint16_t lh_nfree; /* number free chunks */ 1696 uint16_t lh_nentries; /* number of entries */ 1697 uint16_t lh_prefix_len; /* num bits used to id this */ 1698 1699/* above is accessable to zap, below is zap_leaf private */ 1700 1701 uint16_t lh_freelist; /* chunk head of free list */ 1702 uint8_t lh_pad2[12]; 1703 } l_hdr; /* 2 24-byte chunks */ 1704 1705 /* 1706 * The header is followed by a hash table with 1707 * ZAP_LEAF_HASH_NUMENTRIES(zap) entries. The hash table is 1708 * followed by an array of ZAP_LEAF_NUMCHUNKS(zap) 1709 * zap_leaf_chunk structures. These structures are accessed 1710 * with the ZAP_LEAF_CHUNK() macro. 1711 */ 1712 1713 uint16_t l_hash[1]; 1714} zap_leaf_phys_t; 1715 1716typedef union zap_leaf_chunk { 1717 struct zap_leaf_entry { 1718 uint8_t le_type; /* always ZAP_CHUNK_ENTRY */ 1719 uint8_t le_value_intlen; /* size of ints */ 1720 uint16_t le_next; /* next entry in hash chain */ 1721 uint16_t le_name_chunk; /* first chunk of the name */ 1722 uint16_t le_name_numints; /* bytes in name, incl null */ 1723 uint16_t le_value_chunk; /* first chunk of the value */ 1724 uint16_t le_value_numints; /* value length in ints */ 1725 uint32_t le_cd; /* collision differentiator */ 1726 uint64_t le_hash; /* hash value of the name */ 1727 } l_entry; 1728 struct zap_leaf_array { 1729 uint8_t la_type; /* always ZAP_CHUNK_ARRAY */ 1730 uint8_t la_array[ZAP_LEAF_ARRAY_BYTES]; 1731 uint16_t la_next; /* next blk or CHAIN_END */ 1732 } l_array; 1733 struct zap_leaf_free { 1734 uint8_t lf_type; /* always ZAP_CHUNK_FREE */ 1735 uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES]; 1736 uint16_t lf_next; /* next in free list, or CHAIN_END */ 1737 } l_free; 1738} zap_leaf_chunk_t; 1739 1740typedef struct zap_leaf { 1741 int l_bs; /* block size shift */ 1742 zap_leaf_phys_t *l_phys; 1743} zap_leaf_t; 1744 1745#define ZAP_MAXNAMELEN 256 1746#define ZAP_MAXVALUELEN (1024 * 8) 1747 1748#define ACE_READ_DATA 0x00000001 /* file: read data */ 1749#define ACE_LIST_DIRECTORY 0x00000001 /* dir: list files */ 1750#define ACE_WRITE_DATA 0x00000002 /* file: write data */ 1751#define ACE_ADD_FILE 0x00000002 /* dir: create file */ 1752#define ACE_APPEND_DATA 0x00000004 /* file: append data */ 1753#define ACE_ADD_SUBDIRECTORY 0x00000004 /* dir: create subdir */ 1754#define ACE_READ_NAMED_ATTRS 0x00000008 /* FILE_READ_EA */ 1755#define ACE_WRITE_NAMED_ATTRS 0x00000010 /* FILE_WRITE_EA */ 1756#define ACE_EXECUTE 0x00000020 /* file: execute */ 1757#define ACE_TRAVERSE 0x00000020 /* dir: lookup name */ 1758#define ACE_DELETE_CHILD 0x00000040 /* dir: unlink child */ 1759#define ACE_READ_ATTRIBUTES 0x00000080 /* (all) stat, etc. */ 1760#define ACE_WRITE_ATTRIBUTES 0x00000100 /* (all) utimes, etc. */ 1761#define ACE_DELETE 0x00010000 /* (all) unlink self */ 1762#define ACE_READ_ACL 0x00020000 /* (all) getsecattr */ 1763#define ACE_WRITE_ACL 0x00040000 /* (all) setsecattr */ 1764#define ACE_WRITE_OWNER 0x00080000 /* (all) chown */ 1765#define ACE_SYNCHRONIZE 0x00100000 /* (all) */ 1766 1767#define ACE_FILE_INHERIT_ACE 0x0001 1768#define ACE_DIRECTORY_INHERIT_ACE 0x0002 1769#define ACE_NO_PROPAGATE_INHERIT_ACE 0x0004 1770#define ACE_INHERIT_ONLY_ACE 0x0008 1771#define ACE_SUCCESSFUL_ACCESS_ACE_FLAG 0x0010 1772#define ACE_FAILED_ACCESS_ACE_FLAG 0x0020 1773#define ACE_IDENTIFIER_GROUP 0x0040 1774#define ACE_INHERITED_ACE 0x0080 1775#define ACE_OWNER 0x1000 1776#define ACE_GROUP 0x2000 1777#define ACE_EVERYONE 0x4000 1778 1779#define ACE_ACCESS_ALLOWED_ACE_TYPE 0x0000 1780#define ACE_ACCESS_DENIED_ACE_TYPE 0x0001 1781#define ACE_SYSTEM_AUDIT_ACE_TYPE 0x0002 1782#define ACE_SYSTEM_ALARM_ACE_TYPE 0x0003 1783 1784typedef struct zfs_ace_hdr { 1785 uint16_t z_type; 1786 uint16_t z_flags; 1787 uint32_t z_access_mask; 1788} zfs_ace_hdr_t; 1789 1790/* 1791 * Define special zfs pflags 1792 */ 1793#define ZFS_XATTR 0x1 /* is an extended attribute */ 1794#define ZFS_INHERIT_ACE 0x2 /* ace has inheritable ACEs */ 1795#define ZFS_ACL_TRIVIAL 0x4 /* files ACL is trivial */ 1796#define ZFS_ACL_OBJ_ACE 0x8 /* ACL has CMPLX Object ACE */ 1797#define ZFS_ACL_PROTECTED 0x10 /* ACL protected */ 1798#define ZFS_ACL_DEFAULTED 0x20 /* ACL should be defaulted */ 1799#define ZFS_ACL_AUTO_INHERIT 0x40 /* ACL should be inherited */ 1800#define ZFS_BONUS_SCANSTAMP 0x80 /* Scanstamp in bonus area */ 1801#define ZFS_NO_EXECS_DENIED 0x100 /* exec was given to everyone */ 1802 1803#define ZFS_READONLY 0x0000000100000000ull 1804#define ZFS_HIDDEN 0x0000000200000000ull 1805#define ZFS_SYSTEM 0x0000000400000000ull 1806#define ZFS_ARCHIVE 0x0000000800000000ull 1807#define ZFS_IMMUTABLE 0x0000001000000000ull 1808#define ZFS_NOUNLINK 0x0000002000000000ull 1809#define ZFS_APPENDONLY 0x0000004000000000ull 1810#define ZFS_NODUMP 0x0000008000000000ull 1811#define ZFS_OPAQUE 0x0000010000000000ull 1812#define ZFS_AV_QUARANTINED 0x0000020000000000ull 1813#define ZFS_AV_MODIFIED 0x0000040000000000ull 1814#define ZFS_REPARSE 0x0000080000000000ull 1815#define ZFS_OFFLINE 0x0000100000000000ull 1816#define ZFS_SPARSE 0x0000200000000000ull 1817 1818#define MASTER_NODE_OBJ 1 1819 1820/* 1821 * special attributes for master node. 1822 */ 1823 1824#define ZFS_FSID "FSID" 1825#define ZFS_UNLINKED_SET "DELETE_QUEUE" 1826#define ZFS_ROOT_OBJ "ROOT" 1827#define ZPL_VERSION_OBJ "VERSION" 1828#define ZFS_PROP_BLOCKPERPAGE "BLOCKPERPAGE" 1829#define ZFS_PROP_NOGROWBLOCKS "NOGROWBLOCKS" 1830#define ZFS_SA_ATTRS "SA_ATTRS" 1831 1832#define ZFS_FLAG_BLOCKPERPAGE 0x1 1833#define ZFS_FLAG_NOGROWBLOCKS 0x2 1834 1835/* 1836 * ZPL version - rev'd whenever an incompatible on-disk format change 1837 * occurs. Independent of SPA/DMU/ZAP versioning. 1838 */ 1839 1840#define ZPL_VERSION 1ULL 1841 1842/* 1843 * The directory entry has the type (currently unused on Solaris) in the 1844 * top 4 bits, and the object number in the low 48 bits. The "middle" 1845 * 12 bits are unused. 1846 */ 1847#define ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4) 1848#define ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48) 1849#define ZFS_DIRENT_MAKE(type, obj) (((uint64_t)type << 60) | obj) 1850 1851typedef struct ace { 1852 uid_t a_who; /* uid or gid */ 1853 uint32_t a_access_mask; /* read,write,... */ 1854 uint16_t a_flags; /* see below */ 1855 uint16_t a_type; /* allow or deny */ 1856} ace_t; 1857 1858#define ACE_SLOT_CNT 6 1859 1860typedef struct zfs_znode_acl { 1861 uint64_t z_acl_extern_obj; /* ext acl pieces */ 1862 uint32_t z_acl_count; /* Number of ACEs */ 1863 uint16_t z_acl_version; /* acl version */ 1864 uint16_t z_acl_pad; /* pad */ 1865 ace_t z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */ 1866} zfs_znode_acl_t; 1867 1868/* 1869 * This is the persistent portion of the znode. It is stored 1870 * in the "bonus buffer" of the file. Short symbolic links 1871 * are also stored in the bonus buffer. 1872 */ 1873typedef struct znode_phys { 1874 uint64_t zp_atime[2]; /* 0 - last file access time */ 1875 uint64_t zp_mtime[2]; /* 16 - last file modification time */ 1876 uint64_t zp_ctime[2]; /* 32 - last file change time */ 1877 uint64_t zp_crtime[2]; /* 48 - creation time */ 1878 uint64_t zp_gen; /* 64 - generation (txg of creation) */ 1879 uint64_t zp_mode; /* 72 - file mode bits */ 1880 uint64_t zp_size; /* 80 - size of file */ 1881 uint64_t zp_parent; /* 88 - directory parent (`..') */ 1882 uint64_t zp_links; /* 96 - number of links to file */ 1883 uint64_t zp_xattr; /* 104 - DMU object for xattrs */ 1884 uint64_t zp_rdev; /* 112 - dev_t for VBLK & VCHR files */ 1885 uint64_t zp_flags; /* 120 - persistent flags */ 1886 uint64_t zp_uid; /* 128 - file owner */ 1887 uint64_t zp_gid; /* 136 - owning group */ 1888 uint64_t zp_pad[4]; /* 144 - future */ 1889 zfs_znode_acl_t zp_acl; /* 176 - 263 ACL */ 1890 /* 1891 * Data may pad out any remaining bytes in the znode buffer, eg: 1892 * 1893 * |<---------------------- dnode_phys (512) ------------------------>| 1894 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->| 1895 * |<---- znode (264) ---->|<---- data (56) ---->| 1896 * 1897 * At present, we only use this space to store symbolic links. 1898 */ 1899} znode_phys_t; 1900 1901/* 1902 * In-core vdev representation. 1903 */ 1904struct vdev; 1905struct spa; 1906typedef int vdev_phys_read_t(struct vdev *, void *, off_t, void *, size_t); 1907typedef int vdev_phys_write_t(struct vdev *, off_t, void *, size_t); 1908typedef int vdev_read_t(struct vdev *, const blkptr_t *, void *, off_t, size_t); 1909 1910typedef STAILQ_HEAD(vdev_list, vdev) vdev_list_t; 1911 1912typedef struct vdev_indirect_mapping_entry_phys { 1913 /* 1914 * Decode with DVA_MAPPING_* macros. 1915 * Contains: 1916 * the source offset (low 63 bits) 1917 * the one-bit "mark", used for garbage collection (by zdb) 1918 */ 1919 uint64_t vimep_src; 1920 1921 /* 1922 * Note: the DVA's asize is 24 bits, and can thus store ranges 1923 * up to 8GB. 1924 */ 1925 dva_t vimep_dst; 1926} vdev_indirect_mapping_entry_phys_t; 1927 1928#define DVA_MAPPING_GET_SRC_OFFSET(vimep) \ 1929 BF64_GET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0) 1930#define DVA_MAPPING_SET_SRC_OFFSET(vimep, x) \ 1931 BF64_SET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0, x) 1932 1933/* 1934 * This is stored in the bonus buffer of the mapping object, see comment of 1935 * vdev_indirect_config for more details. 1936 */ 1937typedef struct vdev_indirect_mapping_phys { 1938 uint64_t vimp_max_offset; 1939 uint64_t vimp_bytes_mapped; 1940 uint64_t vimp_num_entries; /* number of v_i_m_entry_phys_t's */ 1941 1942 /* 1943 * For each entry in the mapping object, this object contains an 1944 * entry representing the number of bytes of that mapping entry 1945 * that were no longer in use by the pool at the time this indirect 1946 * vdev was last condensed. 1947 */ 1948 uint64_t vimp_counts_object; 1949} vdev_indirect_mapping_phys_t; 1950 1951#define VDEV_INDIRECT_MAPPING_SIZE_V0 (3 * sizeof (uint64_t)) 1952 1953typedef struct vdev_indirect_mapping { 1954 uint64_t vim_object; 1955 boolean_t vim_havecounts; 1956 1957 /* vim_entries segment offset currently in memory. */ 1958 uint64_t vim_entry_offset; 1959 /* vim_entries segment size. */ 1960 size_t vim_num_entries; 1961 1962 /* Needed by dnode_read() */ 1963 const void *vim_spa; 1964 dnode_phys_t *vim_dn; 1965 1966 /* 1967 * An ordered array of mapping entries, sorted by source offset. 1968 * Note that vim_entries is needed during a removal (and contains 1969 * mappings that have been synced to disk so far) to handle frees 1970 * from the removing device. 1971 */ 1972 vdev_indirect_mapping_entry_phys_t *vim_entries; 1973 objset_phys_t *vim_objset; 1974 vdev_indirect_mapping_phys_t *vim_phys; 1975} vdev_indirect_mapping_t; 1976 1977/* 1978 * On-disk indirect vdev state. 1979 * 1980 * An indirect vdev is described exclusively in the MOS config of a pool. 1981 * The config for an indirect vdev includes several fields, which are 1982 * accessed in memory by a vdev_indirect_config_t. 1983 */ 1984typedef struct vdev_indirect_config { 1985 /* 1986 * Object (in MOS) which contains the indirect mapping. This object 1987 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by 1988 * vimep_src. The bonus buffer for this object is a 1989 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev 1990 * removal is initiated. 1991 * 1992 * Note that this object can be empty if none of the data on the vdev 1993 * has been copied yet. 1994 */ 1995 uint64_t vic_mapping_object; 1996 1997 /* 1998 * Object (in MOS) which contains the birth times for the mapping 1999 * entries. This object contains an array of 2000 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus 2001 * buffer for this object is a vdev_indirect_birth_phys_t. This object 2002 * is allocated when a vdev removal is initiated. 2003 * 2004 * Note that this object can be empty if none of the vdev has yet been 2005 * copied. 2006 */ 2007 uint64_t vic_births_object; 2008 2009/* 2010 * This is the vdev ID which was removed previous to this vdev, or 2011 * UINT64_MAX if there are no previously removed vdevs. 2012 */ 2013 uint64_t vic_prev_indirect_vdev; 2014} vdev_indirect_config_t; 2015 2016typedef struct vdev { 2017 STAILQ_ENTRY(vdev) v_childlink; /* link in parent's child list */ 2018 STAILQ_ENTRY(vdev) v_alllink; /* link in global vdev list */ 2019 vdev_list_t v_children; /* children of this vdev */ 2020 const char *v_name; /* vdev name */ 2021 uint64_t v_guid; /* vdev guid */ 2022 uint64_t v_id; /* index in parent */ 2023 uint64_t v_psize; /* physical device capacity */ 2024 int v_ashift; /* offset to block shift */ 2025 int v_nparity; /* # parity for raidz */ 2026 struct vdev *v_top; /* parent vdev */ 2027 size_t v_nchildren; /* # children */ 2028 vdev_state_t v_state; /* current state */ 2029 vdev_phys_read_t *v_phys_read; /* read from raw leaf vdev */ 2030 vdev_phys_write_t *v_phys_write; /* write to raw leaf vdev */ 2031 vdev_read_t *v_read; /* read from vdev */ 2032 void *v_priv; /* data for read/write function */ 2033 boolean_t v_islog; 2034 struct spa *v_spa; /* link to spa */ 2035 /* 2036 * Values stored in the config for an indirect or removing vdev. 2037 */ 2038 vdev_indirect_config_t vdev_indirect_config; 2039 vdev_indirect_mapping_t *v_mapping; 2040} vdev_t; 2041 2042/* 2043 * In-core pool representation. 2044 */ 2045typedef STAILQ_HEAD(spa_list, spa) spa_list_t; 2046 2047typedef struct spa { 2048 STAILQ_ENTRY(spa) spa_link; /* link in global pool list */ 2049 char *spa_name; /* pool name */ 2050 uint64_t spa_guid; /* pool guid */ 2051 uint64_t spa_txg; /* most recent transaction */ 2052 struct uberblock *spa_uberblock; /* best uberblock so far */ 2053 vdev_t *spa_root_vdev; /* toplevel vdev container */ 2054 objset_phys_t *spa_mos; /* MOS for this pool */ 2055 zio_cksum_salt_t spa_cksum_salt; /* secret salt for cksum */ 2056 void *spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS]; 2057 boolean_t spa_with_log; /* this pool has log */ 2058 2059 struct uberblock spa_uberblock_master; /* best uberblock so far */ 2060 objset_phys_t spa_mos_master; /* MOS for this pool */ 2061 struct uberblock spa_uberblock_checkpoint; /* checkpoint uberblock */ 2062 objset_phys_t spa_mos_checkpoint; /* Checkpoint MOS */ 2063 void *spa_bootenv; /* bootenv from pool label */ 2064} spa_t; 2065 2066/* IO related arguments. */ 2067typedef struct zio { 2068 spa_t *io_spa; 2069 blkptr_t *io_bp; 2070 void *io_data; 2071 uint64_t io_size; 2072 uint64_t io_offset; 2073 2074 /* Stuff for the vdev stack */ 2075 vdev_t *io_vd; 2076 void *io_vsd; 2077 2078 int io_error; 2079} zio_t; 2080 2081extern void decode_embedded_bp_compressed(const blkptr_t *, void *); 2082 2083#endif /* _ZFSIMPL_H_ */ 2084