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 */ 80 typedef 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 */ 154 typedef 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 */ 161 typedef 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 */ 169 typedef 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) 316 317 typedef 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 */ 330 331 typedef 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 491 492 typedef 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)) 508 509 typedef 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)) 538 539 #define ASHIFT_UBERBLOCK_SHIFT(ashift) \ 540 MIN(MAX(ashift, UBERBLOCK_SHIFT), \ 541 MAX_UBERBLOCK_SHIFT) 542 #define ASHIFT_UBERBLOCK_SIZE(ashift) \ 543 (1ULL << ASHIFT_UBERBLOCK_SHIFT(ashift)) 544 545 typedef struct vdev_phys { 546 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)]; 547 zio_eck_t vp_zbt; 548 } vdev_phys_t; 549 550 typedef enum vbe_vers { 551 /* The bootenv file is stored as ascii text in the envblock */ 552 VB_RAW = 0, 553 554 /* 555 * The bootenv file is converted to an nvlist and then packed into the 556 * envblock. 557 */ 558 VB_NVLIST = 1 559 } vbe_vers_t; 560 561 typedef struct vdev_boot_envblock { 562 uint64_t vbe_version; 563 char vbe_bootenv[VDEV_PAD_SIZE - sizeof (uint64_t) - 564 sizeof (zio_eck_t)]; 565 zio_eck_t vbe_zbt; 566 } vdev_boot_envblock_t; 567 568 _Static_assert(sizeof (vdev_boot_envblock_t) == VDEV_PAD_SIZE, 569 "bad size for vdev_boot_envblock_t"); 570 571 typedef struct vdev_label { 572 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */ 573 vdev_boot_envblock_t vl_be; /* 8K */ 574 vdev_phys_t vl_vdev_phys; /* 112K */ 575 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */ 576 } vdev_label_t; /* 256K total */ 577 578 /* 579 * vdev_dirty() flags 580 */ 581 #define VDD_METASLAB 0x01 582 #define VDD_DTL 0x02 583 584 /* 585 * Size and offset of embedded boot loader region on each label. 586 * The total size of the first two labels plus the boot area is 4MB. 587 */ 588 #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t)) 589 #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */ 590 591 /* 592 * Size of label regions at the start and end of each leaf device. 593 */ 594 #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE) 595 #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t)) 596 #define VDEV_LABELS 4 597 598 enum zio_checksum { 599 ZIO_CHECKSUM_INHERIT = 0, 600 ZIO_CHECKSUM_ON, 601 ZIO_CHECKSUM_OFF, 602 ZIO_CHECKSUM_LABEL, 603 ZIO_CHECKSUM_GANG_HEADER, 604 ZIO_CHECKSUM_ZILOG, 605 ZIO_CHECKSUM_FLETCHER_2, 606 ZIO_CHECKSUM_FLETCHER_4, 607 ZIO_CHECKSUM_SHA256, 608 ZIO_CHECKSUM_ZILOG2, 609 ZIO_CHECKSUM_NOPARITY, 610 ZIO_CHECKSUM_SHA512, 611 ZIO_CHECKSUM_SKEIN, 612 ZIO_CHECKSUM_EDONR, 613 ZIO_CHECKSUM_BLAKE3, 614 ZIO_CHECKSUM_FUNCTIONS 615 }; 616 617 #define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4 618 #define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON 619 620 enum zio_compress { 621 ZIO_COMPRESS_INHERIT = 0, 622 ZIO_COMPRESS_ON, 623 ZIO_COMPRESS_OFF, 624 ZIO_COMPRESS_LZJB, 625 ZIO_COMPRESS_EMPTY, 626 ZIO_COMPRESS_GZIP_1, 627 ZIO_COMPRESS_GZIP_2, 628 ZIO_COMPRESS_GZIP_3, 629 ZIO_COMPRESS_GZIP_4, 630 ZIO_COMPRESS_GZIP_5, 631 ZIO_COMPRESS_GZIP_6, 632 ZIO_COMPRESS_GZIP_7, 633 ZIO_COMPRESS_GZIP_8, 634 ZIO_COMPRESS_GZIP_9, 635 ZIO_COMPRESS_ZLE, 636 ZIO_COMPRESS_LZ4, 637 ZIO_COMPRESS_ZSTD, 638 ZIO_COMPRESS_FUNCTIONS 639 }; 640 641 enum zio_zstd_levels { 642 ZIO_ZSTD_LEVEL_INHERIT = 0, 643 ZIO_ZSTD_LEVEL_1, 644 #define ZIO_ZSTD_LEVEL_MIN ZIO_ZSTD_LEVEL_1 645 ZIO_ZSTD_LEVEL_2, 646 ZIO_ZSTD_LEVEL_3, 647 #define ZIO_ZSTD_LEVEL_DEFAULT ZIO_ZSTD_LEVEL_3 648 ZIO_ZSTD_LEVEL_4, 649 ZIO_ZSTD_LEVEL_5, 650 ZIO_ZSTD_LEVEL_6, 651 ZIO_ZSTD_LEVEL_7, 652 ZIO_ZSTD_LEVEL_8, 653 ZIO_ZSTD_LEVEL_9, 654 ZIO_ZSTD_LEVEL_10, 655 ZIO_ZSTD_LEVEL_11, 656 ZIO_ZSTD_LEVEL_12, 657 ZIO_ZSTD_LEVEL_13, 658 ZIO_ZSTD_LEVEL_14, 659 ZIO_ZSTD_LEVEL_15, 660 ZIO_ZSTD_LEVEL_16, 661 ZIO_ZSTD_LEVEL_17, 662 ZIO_ZSTD_LEVEL_18, 663 ZIO_ZSTD_LEVEL_19, 664 #define ZIO_ZSTD_LEVEL_MAX ZIO_ZSTD_LEVEL_19 665 ZIO_ZSTD_LEVEL_RESERVE = 101, /* Leave room for new positive levels */ 666 ZIO_ZSTD_LEVEL_FAST, /* Fast levels are negative */ 667 ZIO_ZSTD_LEVEL_FAST_1, 668 #define ZIO_ZSTD_LEVEL_FAST_DEFAULT ZIO_ZSTD_LEVEL_FAST_1 669 ZIO_ZSTD_LEVEL_FAST_2, 670 ZIO_ZSTD_LEVEL_FAST_3, 671 ZIO_ZSTD_LEVEL_FAST_4, 672 ZIO_ZSTD_LEVEL_FAST_5, 673 ZIO_ZSTD_LEVEL_FAST_6, 674 ZIO_ZSTD_LEVEL_FAST_7, 675 ZIO_ZSTD_LEVEL_FAST_8, 676 ZIO_ZSTD_LEVEL_FAST_9, 677 ZIO_ZSTD_LEVEL_FAST_10, 678 ZIO_ZSTD_LEVEL_FAST_20, 679 ZIO_ZSTD_LEVEL_FAST_30, 680 ZIO_ZSTD_LEVEL_FAST_40, 681 ZIO_ZSTD_LEVEL_FAST_50, 682 ZIO_ZSTD_LEVEL_FAST_60, 683 ZIO_ZSTD_LEVEL_FAST_70, 684 ZIO_ZSTD_LEVEL_FAST_80, 685 ZIO_ZSTD_LEVEL_FAST_90, 686 ZIO_ZSTD_LEVEL_FAST_100, 687 ZIO_ZSTD_LEVEL_FAST_500, 688 ZIO_ZSTD_LEVEL_FAST_1000, 689 #define ZIO_ZSTD_LEVEL_FAST_MAX ZIO_ZSTD_LEVEL_FAST_1000 690 ZIO_ZSTD_LEVEL_AUTO = 251, /* Reserved for future use */ 691 ZIO_ZSTD_LEVEL_LEVELS 692 }; 693 694 #define ZIO_COMPRESS_ON_VALUE ZIO_COMPRESS_LZJB 695 #define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF 696 697 /* 698 * On-disk version number. 699 */ 700 #define SPA_VERSION_1 1ULL 701 #define SPA_VERSION_2 2ULL 702 #define SPA_VERSION_3 3ULL 703 #define SPA_VERSION_4 4ULL 704 #define SPA_VERSION_5 5ULL 705 #define SPA_VERSION_6 6ULL 706 #define SPA_VERSION_7 7ULL 707 #define SPA_VERSION_8 8ULL 708 #define SPA_VERSION_9 9ULL 709 #define SPA_VERSION_10 10ULL 710 #define SPA_VERSION_11 11ULL 711 #define SPA_VERSION_12 12ULL 712 #define SPA_VERSION_13 13ULL 713 #define SPA_VERSION_14 14ULL 714 #define SPA_VERSION_15 15ULL 715 #define SPA_VERSION_16 16ULL 716 #define SPA_VERSION_17 17ULL 717 #define SPA_VERSION_18 18ULL 718 #define SPA_VERSION_19 19ULL 719 #define SPA_VERSION_20 20ULL 720 #define SPA_VERSION_21 21ULL 721 #define SPA_VERSION_22 22ULL 722 #define SPA_VERSION_23 23ULL 723 #define SPA_VERSION_24 24ULL 724 #define SPA_VERSION_25 25ULL 725 #define SPA_VERSION_26 26ULL 726 #define SPA_VERSION_27 27ULL 727 #define SPA_VERSION_28 28ULL 728 #define SPA_VERSION_5000 5000ULL 729 730 /* 731 * When bumping up SPA_VERSION, make sure GRUB ZFS understands the on-disk 732 * format change. Go to usr/src/grub/grub-0.97/stage2/{zfs-include/, fsys_zfs*}, 733 * and do the appropriate changes. Also bump the version number in 734 * usr/src/grub/capability. 735 */ 736 #define SPA_VERSION SPA_VERSION_5000 737 #define SPA_VERSION_STRING "5000" 738 739 /* 740 * Symbolic names for the changes that caused a SPA_VERSION switch. 741 * Used in the code when checking for presence or absence of a feature. 742 * Feel free to define multiple symbolic names for each version if there 743 * were multiple changes to on-disk structures during that version. 744 * 745 * NOTE: When checking the current SPA_VERSION in your code, be sure 746 * to use spa_version() since it reports the version of the 747 * last synced uberblock. Checking the in-flight version can 748 * be dangerous in some cases. 749 */ 750 #define SPA_VERSION_INITIAL SPA_VERSION_1 751 #define SPA_VERSION_DITTO_BLOCKS SPA_VERSION_2 752 #define SPA_VERSION_SPARES SPA_VERSION_3 753 #define SPA_VERSION_RAID6 SPA_VERSION_3 754 #define SPA_VERSION_BPLIST_ACCOUNT SPA_VERSION_3 755 #define SPA_VERSION_RAIDZ_DEFLATE SPA_VERSION_3 756 #define SPA_VERSION_DNODE_BYTES SPA_VERSION_3 757 #define SPA_VERSION_ZPOOL_HISTORY SPA_VERSION_4 758 #define SPA_VERSION_GZIP_COMPRESSION SPA_VERSION_5 759 #define SPA_VERSION_BOOTFS SPA_VERSION_6 760 #define SPA_VERSION_SLOGS SPA_VERSION_7 761 #define SPA_VERSION_DELEGATED_PERMS SPA_VERSION_8 762 #define SPA_VERSION_FUID SPA_VERSION_9 763 #define SPA_VERSION_REFRESERVATION SPA_VERSION_9 764 #define SPA_VERSION_REFQUOTA SPA_VERSION_9 765 #define SPA_VERSION_UNIQUE_ACCURATE SPA_VERSION_9 766 #define SPA_VERSION_L2CACHE SPA_VERSION_10 767 #define SPA_VERSION_NEXT_CLONES SPA_VERSION_11 768 #define SPA_VERSION_ORIGIN SPA_VERSION_11 769 #define SPA_VERSION_DSL_SCRUB SPA_VERSION_11 770 #define SPA_VERSION_SNAP_PROPS SPA_VERSION_12 771 #define SPA_VERSION_USED_BREAKDOWN SPA_VERSION_13 772 #define SPA_VERSION_PASSTHROUGH_X SPA_VERSION_14 773 #define SPA_VERSION_USERSPACE SPA_VERSION_15 774 #define SPA_VERSION_STMF_PROP SPA_VERSION_16 775 #define SPA_VERSION_RAIDZ3 SPA_VERSION_17 776 #define SPA_VERSION_USERREFS SPA_VERSION_18 777 #define SPA_VERSION_HOLES SPA_VERSION_19 778 #define SPA_VERSION_ZLE_COMPRESSION SPA_VERSION_20 779 #define SPA_VERSION_DEDUP SPA_VERSION_21 780 #define SPA_VERSION_RECVD_PROPS SPA_VERSION_22 781 #define SPA_VERSION_SLIM_ZIL SPA_VERSION_23 782 #define SPA_VERSION_SA SPA_VERSION_24 783 #define SPA_VERSION_SCAN SPA_VERSION_25 784 #define SPA_VERSION_DIR_CLONES SPA_VERSION_26 785 #define SPA_VERSION_DEADLISTS SPA_VERSION_26 786 #define SPA_VERSION_FAST_SNAP SPA_VERSION_27 787 #define SPA_VERSION_MULTI_REPLACE SPA_VERSION_28 788 #define SPA_VERSION_BEFORE_FEATURES SPA_VERSION_28 789 #define SPA_VERSION_FEATURES SPA_VERSION_5000 790 791 #define SPA_VERSION_IS_SUPPORTED(v) \ 792 (((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \ 793 ((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION)) 794 795 /* 796 * The following are configuration names used in the nvlist describing a pool's 797 * configuration. 798 */ 799 #define ZPOOL_CONFIG_VERSION "version" 800 #define ZPOOL_CONFIG_POOL_NAME "name" 801 #define ZPOOL_CONFIG_POOL_STATE "state" 802 #define ZPOOL_CONFIG_POOL_TXG "txg" 803 #define ZPOOL_CONFIG_POOL_GUID "pool_guid" 804 #define ZPOOL_CONFIG_CREATE_TXG "create_txg" 805 #define ZPOOL_CONFIG_TOP_GUID "top_guid" 806 #define ZPOOL_CONFIG_VDEV_TREE "vdev_tree" 807 #define ZPOOL_CONFIG_TYPE "type" 808 #define ZPOOL_CONFIG_CHILDREN "children" 809 #define ZPOOL_CONFIG_ID "id" 810 #define ZPOOL_CONFIG_GUID "guid" 811 #define ZPOOL_CONFIG_INDIRECT_OBJECT "com.delphix:indirect_object" 812 #define ZPOOL_CONFIG_INDIRECT_BIRTHS "com.delphix:indirect_births" 813 #define ZPOOL_CONFIG_PREV_INDIRECT_VDEV "com.delphix:prev_indirect_vdev" 814 #define ZPOOL_CONFIG_PATH "path" 815 #define ZPOOL_CONFIG_DEVID "devid" 816 #define ZPOOL_CONFIG_METASLAB_ARRAY "metaslab_array" 817 #define ZPOOL_CONFIG_METASLAB_SHIFT "metaslab_shift" 818 #define ZPOOL_CONFIG_ASHIFT "ashift" 819 #define ZPOOL_CONFIG_ASIZE "asize" 820 #define ZPOOL_CONFIG_DTL "DTL" 821 #define ZPOOL_CONFIG_STATS "stats" 822 #define ZPOOL_CONFIG_WHOLE_DISK "whole_disk" 823 #define ZPOOL_CONFIG_ERRCOUNT "error_count" 824 #define ZPOOL_CONFIG_NOT_PRESENT "not_present" 825 #define ZPOOL_CONFIG_SPARES "spares" 826 #define ZPOOL_CONFIG_IS_SPARE "is_spare" 827 #define ZPOOL_CONFIG_NPARITY "nparity" 828 #define ZPOOL_CONFIG_HOSTID "hostid" 829 #define ZPOOL_CONFIG_HOSTNAME "hostname" 830 #define ZPOOL_CONFIG_IS_LOG "is_log" 831 #define ZPOOL_CONFIG_TIMESTAMP "timestamp" /* not stored on disk */ 832 #define ZPOOL_CONFIG_FEATURES_FOR_READ "features_for_read" 833 #define ZPOOL_CONFIG_VDEV_CHILDREN "vdev_children" 834 835 /* 836 * The persistent vdev state is stored as separate values rather than a single 837 * 'vdev_state' entry. This is because a device can be in multiple states, such 838 * as offline and degraded. 839 */ 840 #define ZPOOL_CONFIG_OFFLINE "offline" 841 #define ZPOOL_CONFIG_FAULTED "faulted" 842 #define ZPOOL_CONFIG_DEGRADED "degraded" 843 #define ZPOOL_CONFIG_REMOVED "removed" 844 #define ZPOOL_CONFIG_FRU "fru" 845 #define ZPOOL_CONFIG_AUX_STATE "aux_state" 846 847 #define VDEV_TYPE_ROOT "root" 848 #define VDEV_TYPE_MIRROR "mirror" 849 #define VDEV_TYPE_REPLACING "replacing" 850 #define VDEV_TYPE_RAIDZ "raidz" 851 #define VDEV_TYPE_DISK "disk" 852 #define VDEV_TYPE_FILE "file" 853 #define VDEV_TYPE_MISSING "missing" 854 #define VDEV_TYPE_HOLE "hole" 855 #define VDEV_TYPE_SPARE "spare" 856 #define VDEV_TYPE_LOG "log" 857 #define VDEV_TYPE_L2CACHE "l2cache" 858 #define VDEV_TYPE_INDIRECT "indirect" 859 860 /* 861 * This is needed in userland to report the minimum necessary device size. 862 */ 863 #define SPA_MINDEVSIZE (64ULL << 20) 864 865 /* 866 * The location of the pool configuration repository, shared between kernel and 867 * userland. 868 */ 869 #define ZPOOL_CACHE "/boot/zfs/zpool.cache" 870 871 /* 872 * vdev states are ordered from least to most healthy. 873 * A vdev that's CANT_OPEN or below is considered unusable. 874 */ 875 typedef enum vdev_state { 876 VDEV_STATE_UNKNOWN = 0, /* Uninitialized vdev */ 877 VDEV_STATE_CLOSED, /* Not currently open */ 878 VDEV_STATE_OFFLINE, /* Not allowed to open */ 879 VDEV_STATE_REMOVED, /* Explicitly removed from system */ 880 VDEV_STATE_CANT_OPEN, /* Tried to open, but failed */ 881 VDEV_STATE_FAULTED, /* External request to fault device */ 882 VDEV_STATE_DEGRADED, /* Replicated vdev with unhealthy kids */ 883 VDEV_STATE_HEALTHY /* Presumed good */ 884 } vdev_state_t; 885 886 /* 887 * vdev aux states. When a vdev is in the CANT_OPEN state, the aux field 888 * of the vdev stats structure uses these constants to distinguish why. 889 */ 890 typedef enum vdev_aux { 891 VDEV_AUX_NONE, /* no error */ 892 VDEV_AUX_OPEN_FAILED, /* ldi_open_*() or vn_open() failed */ 893 VDEV_AUX_CORRUPT_DATA, /* bad label or disk contents */ 894 VDEV_AUX_NO_REPLICAS, /* insufficient number of replicas */ 895 VDEV_AUX_BAD_GUID_SUM, /* vdev guid sum doesn't match */ 896 VDEV_AUX_TOO_SMALL, /* vdev size is too small */ 897 VDEV_AUX_BAD_LABEL, /* the label is OK but invalid */ 898 VDEV_AUX_VERSION_NEWER, /* on-disk version is too new */ 899 VDEV_AUX_VERSION_OLDER, /* on-disk version is too old */ 900 VDEV_AUX_SPARED /* hot spare used in another pool */ 901 } vdev_aux_t; 902 903 /* 904 * pool state. The following states are written to disk as part of the normal 905 * SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE. The remaining states are 906 * software abstractions used at various levels to communicate pool state. 907 */ 908 typedef enum pool_state { 909 POOL_STATE_ACTIVE = 0, /* In active use */ 910 POOL_STATE_EXPORTED, /* Explicitly exported */ 911 POOL_STATE_DESTROYED, /* Explicitly destroyed */ 912 POOL_STATE_SPARE, /* Reserved for hot spare use */ 913 POOL_STATE_UNINITIALIZED, /* Internal spa_t state */ 914 POOL_STATE_UNAVAIL, /* Internal libzfs state */ 915 POOL_STATE_POTENTIALLY_ACTIVE /* Internal libzfs state */ 916 } pool_state_t; 917 918 /* 919 * The uberblock version is incremented whenever an incompatible on-disk 920 * format change is made to the SPA, DMU, or ZAP. 921 * 922 * Note: the first two fields should never be moved. When a storage pool 923 * is opened, the uberblock must be read off the disk before the version 924 * can be checked. If the ub_version field is moved, we may not detect 925 * version mismatch. If the ub_magic field is moved, applications that 926 * expect the magic number in the first word won't work. 927 */ 928 #define UBERBLOCK_MAGIC 0x00bab10c /* oo-ba-bloc! */ 929 #define UBERBLOCK_SHIFT 10 /* up to 1K */ 930 931 #define MMP_MAGIC 0xa11cea11 /* all-see-all */ 932 933 #define MMP_INTERVAL_VALID_BIT 0x01 934 #define MMP_SEQ_VALID_BIT 0x02 935 #define MMP_FAIL_INT_VALID_BIT 0x04 936 937 #define MMP_VALID(ubp) (ubp->ub_magic == UBERBLOCK_MAGIC && \ 938 ubp->ub_mmp_magic == MMP_MAGIC) 939 #define MMP_INTERVAL_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \ 940 MMP_INTERVAL_VALID_BIT)) 941 #define MMP_SEQ_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \ 942 MMP_SEQ_VALID_BIT)) 943 #define MMP_FAIL_INT_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \ 944 MMP_FAIL_INT_VALID_BIT)) 945 946 #define MMP_INTERVAL(ubp) ((ubp->ub_mmp_config & 0x00000000FFFFFF00) \ 947 >> 8) 948 #define MMP_SEQ(ubp) ((ubp->ub_mmp_config & 0x0000FFFF00000000) \ 949 >> 32) 950 #define MMP_FAIL_INT(ubp) ((ubp->ub_mmp_config & 0xFFFF000000000000) \ 951 >> 48) 952 953 typedef struct uberblock { 954 uint64_t ub_magic; /* UBERBLOCK_MAGIC */ 955 uint64_t ub_version; /* SPA_VERSION */ 956 uint64_t ub_txg; /* txg of last sync */ 957 uint64_t ub_guid_sum; /* sum of all vdev guids */ 958 uint64_t ub_timestamp; /* UTC time of last sync */ 959 blkptr_t ub_rootbp; /* MOS objset_phys_t */ 960 /* highest SPA_VERSION supported by software that wrote this txg */ 961 uint64_t ub_software_version; 962 /* Maybe missing in uberblocks we read, but always written */ 963 uint64_t ub_mmp_magic; 964 /* 965 * If ub_mmp_delay == 0 and ub_mmp_magic is valid, MMP is off. 966 * Otherwise, nanosec since last MMP write. 967 */ 968 uint64_t ub_mmp_delay; 969 970 /* 971 * The ub_mmp_config contains the multihost write interval, multihost 972 * fail intervals, sequence number for sub-second granularity, and 973 * valid bit mask. This layout is as follows: 974 * 975 * 64 56 48 40 32 24 16 8 0 976 * +-------+-------+-------+-------+-------+-------+-------+-------+ 977 * 0 | Fail Intervals| Seq | Write Interval (ms) | VALID | 978 * +-------+-------+-------+-------+-------+-------+-------+-------+ 979 * 980 * This allows a write_interval of (2^24/1000)s, over 4.5 hours 981 * 982 * VALID Bits: 983 * - 0x01 - Write Interval (ms) 984 * - 0x02 - Sequence number exists 985 * - 0x04 - Fail Intervals 986 * - 0xf8 - Reserved 987 */ 988 uint64_t ub_mmp_config; 989 990 /* 991 * ub_checkpoint_txg indicates two things about the current uberblock: 992 * 993 * 1] If it is not zero then this uberblock is a checkpoint. If it is 994 * zero, then this uberblock is not a checkpoint. 995 * 996 * 2] On checkpointed uberblocks, the value of ub_checkpoint_txg is 997 * the ub_txg that the uberblock had at the time we moved it to 998 * the MOS config. 999 * 1000 * The field is set when we checkpoint the uberblock and continues to 1001 * hold that value even after we've rewound (unlike the ub_txg that 1002 * is reset to a higher value). 1003 * 1004 * Besides checks used to determine whether we are reopening the 1005 * pool from a checkpointed uberblock [see spa_ld_select_uberblock()], 1006 * the value of the field is used to determine which ZIL blocks have 1007 * been allocated according to the ms_sm when we are rewinding to a 1008 * checkpoint. Specifically, if blk_birth > ub_checkpoint_txg, then 1009 * the ZIL block is not allocated [see uses of spa_min_claim_txg()]. 1010 */ 1011 uint64_t ub_checkpoint_txg; 1012 } uberblock_t; 1013 1014 /* 1015 * Flags. 1016 */ 1017 #define DNODE_MUST_BE_ALLOCATED 1 1018 #define DNODE_MUST_BE_FREE 2 1019 1020 /* 1021 * Fixed constants. 1022 */ 1023 #define DNODE_SHIFT 9 /* 512 bytes */ 1024 #define DN_MIN_INDBLKSHIFT 12 /* 4k */ 1025 #define DN_MAX_INDBLKSHIFT 17 /* 128k */ 1026 #define DNODE_BLOCK_SHIFT 14 /* 16k */ 1027 #define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */ 1028 #define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */ 1029 #define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */ 1030 1031 /* 1032 * Derived constants. 1033 */ 1034 #define DNODE_MIN_SIZE (1 << DNODE_SHIFT) 1035 #define DNODE_MAX_SIZE (1 << DNODE_BLOCK_SHIFT) 1036 #define DNODE_BLOCK_SIZE (1 << DNODE_BLOCK_SHIFT) 1037 #define DNODE_MIN_SLOTS (DNODE_MIN_SIZE >> DNODE_SHIFT) 1038 #define DNODE_MAX_SLOTS (DNODE_MAX_SIZE >> DNODE_SHIFT) 1039 #define DN_BONUS_SIZE(dnsize) ((dnsize) - DNODE_CORE_SIZE - \ 1040 (1 << SPA_BLKPTRSHIFT)) 1041 #define DN_SLOTS_TO_BONUSLEN(slots) DN_BONUS_SIZE((slots) << DNODE_SHIFT) 1042 #define DN_OLD_MAX_BONUSLEN (DN_BONUS_SIZE(DNODE_MIN_SIZE)) 1043 #define DN_MAX_NBLKPTR ((DNODE_MIN_SIZE - DNODE_CORE_SIZE) >> \ 1044 SPA_BLKPTRSHIFT) 1045 #define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT) 1046 #define DN_ZERO_BONUSLEN (DN_BONUS_SIZE(DNODE_MAX_SIZE) + 1) 1047 1048 #define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT) 1049 #define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT) 1050 #define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT) 1051 1052 /* The +2 here is a cheesy way to round up */ 1053 #define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \ 1054 (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT))) 1055 1056 #define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \ 1057 (((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t)))) 1058 1059 #define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \ 1060 (dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT) 1061 1062 #define EPB(blkshift, typeshift) (1 << (blkshift - typeshift)) 1063 1064 /* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */ 1065 #define DNODE_FLAG_USED_BYTES (1<<0) 1066 #define DNODE_FLAG_USERUSED_ACCOUNTED (1<<1) 1067 1068 /* Does dnode have a SA spill blkptr in bonus? */ 1069 #define DNODE_FLAG_SPILL_BLKPTR (1<<2) 1070 1071 typedef struct dnode_phys { 1072 uint8_t dn_type; /* dmu_object_type_t */ 1073 uint8_t dn_indblkshift; /* ln2(indirect block size) */ 1074 uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */ 1075 uint8_t dn_nblkptr; /* length of dn_blkptr */ 1076 uint8_t dn_bonustype; /* type of data in bonus buffer */ 1077 uint8_t dn_checksum; /* ZIO_CHECKSUM type */ 1078 uint8_t dn_compress; /* ZIO_COMPRESS type */ 1079 uint8_t dn_flags; /* DNODE_FLAG_* */ 1080 uint16_t dn_datablkszsec; /* data block size in 512b sectors */ 1081 uint16_t dn_bonuslen; /* length of dn_bonus */ 1082 uint8_t dn_extra_slots; /* # of subsequent slots consumed */ 1083 uint8_t dn_pad2[3]; 1084 1085 /* accounting is protected by dn_dirty_mtx */ 1086 uint64_t dn_maxblkid; /* largest allocated block ID */ 1087 uint64_t dn_used; /* bytes (or sectors) of disk space */ 1088 1089 uint64_t dn_pad3[4]; 1090 1091 /* 1092 * The tail region is 448 bytes for a 512 byte dnode, and 1093 * correspondingly larger for larger dnode sizes. The spill 1094 * block pointer, when present, is always at the end of the tail 1095 * region. There are three ways this space may be used, using 1096 * a 512 byte dnode for this diagram: 1097 * 1098 * 0 64 128 192 256 320 384 448 (offset) 1099 * +---------------+---------------+---------------+-------+ 1100 * | dn_blkptr[0] | dn_blkptr[1] | dn_blkptr[2] | / | 1101 * +---------------+---------------+---------------+-------+ 1102 * | dn_blkptr[0] | dn_bonus[0..319] | 1103 * +---------------+-----------------------+---------------+ 1104 * | dn_blkptr[0] | dn_bonus[0..191] | dn_spill | 1105 * +---------------+-----------------------+---------------+ 1106 */ 1107 union { 1108 blkptr_t dn_blkptr[1+DN_OLD_MAX_BONUSLEN/sizeof (blkptr_t)]; 1109 struct { 1110 blkptr_t __dn_ignore1; 1111 uint8_t dn_bonus[DN_OLD_MAX_BONUSLEN]; 1112 }; 1113 struct { 1114 blkptr_t __dn_ignore2; 1115 uint8_t __dn_ignore3[DN_OLD_MAX_BONUSLEN - 1116 sizeof (blkptr_t)]; 1117 blkptr_t dn_spill; 1118 }; 1119 }; 1120 } dnode_phys_t; 1121 1122 #define DN_SPILL_BLKPTR(dnp) (blkptr_t *)((char *)(dnp) + \ 1123 (((dnp)->dn_extra_slots + 1) << DNODE_SHIFT) - (1 << SPA_BLKPTRSHIFT)) 1124 1125 typedef enum dmu_object_byteswap { 1126 DMU_BSWAP_UINT8, 1127 DMU_BSWAP_UINT16, 1128 DMU_BSWAP_UINT32, 1129 DMU_BSWAP_UINT64, 1130 DMU_BSWAP_ZAP, 1131 DMU_BSWAP_DNODE, 1132 DMU_BSWAP_OBJSET, 1133 DMU_BSWAP_ZNODE, 1134 DMU_BSWAP_OLDACL, 1135 DMU_BSWAP_ACL, 1136 /* 1137 * Allocating a new byteswap type number makes the on-disk format 1138 * incompatible with any other format that uses the same number. 1139 * 1140 * Data can usually be structured to work with one of the 1141 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. 1142 */ 1143 DMU_BSWAP_NUMFUNCS 1144 } dmu_object_byteswap_t; 1145 1146 #define DMU_OT_NEWTYPE 0x80 1147 #define DMU_OT_METADATA 0x40 1148 #define DMU_OT_BYTESWAP_MASK 0x3f 1149 1150 /* 1151 * Defines a uint8_t object type. Object types specify if the data 1152 * in the object is metadata (boolean) and how to byteswap the data 1153 * (dmu_object_byteswap_t). 1154 */ 1155 #define DMU_OT(byteswap, metadata) \ 1156 (DMU_OT_NEWTYPE | \ 1157 ((metadata) ? DMU_OT_METADATA : 0) | \ 1158 ((byteswap) & DMU_OT_BYTESWAP_MASK)) 1159 1160 typedef enum dmu_object_type { 1161 DMU_OT_NONE, 1162 /* general: */ 1163 DMU_OT_OBJECT_DIRECTORY, /* ZAP */ 1164 DMU_OT_OBJECT_ARRAY, /* UINT64 */ 1165 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ 1166 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ 1167 DMU_OT_BPOBJ, /* UINT64 */ 1168 DMU_OT_BPOBJ_HDR, /* UINT64 */ 1169 /* spa: */ 1170 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ 1171 DMU_OT_SPACE_MAP, /* UINT64 */ 1172 /* zil: */ 1173 DMU_OT_INTENT_LOG, /* UINT64 */ 1174 /* dmu: */ 1175 DMU_OT_DNODE, /* DNODE */ 1176 DMU_OT_OBJSET, /* OBJSET */ 1177 /* dsl: */ 1178 DMU_OT_DSL_DIR, /* UINT64 */ 1179 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ 1180 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ 1181 DMU_OT_DSL_PROPS, /* ZAP */ 1182 DMU_OT_DSL_DATASET, /* UINT64 */ 1183 /* zpl: */ 1184 DMU_OT_ZNODE, /* ZNODE */ 1185 DMU_OT_OLDACL, /* Old ACL */ 1186 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ 1187 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ 1188 DMU_OT_MASTER_NODE, /* ZAP */ 1189 DMU_OT_UNLINKED_SET, /* ZAP */ 1190 /* zvol: */ 1191 DMU_OT_ZVOL, /* UINT8 */ 1192 DMU_OT_ZVOL_PROP, /* ZAP */ 1193 /* other; for testing only! */ 1194 DMU_OT_PLAIN_OTHER, /* UINT8 */ 1195 DMU_OT_UINT64_OTHER, /* UINT64 */ 1196 DMU_OT_ZAP_OTHER, /* ZAP */ 1197 /* new object types: */ 1198 DMU_OT_ERROR_LOG, /* ZAP */ 1199 DMU_OT_SPA_HISTORY, /* UINT8 */ 1200 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ 1201 DMU_OT_POOL_PROPS, /* ZAP */ 1202 DMU_OT_DSL_PERMS, /* ZAP */ 1203 DMU_OT_ACL, /* ACL */ 1204 DMU_OT_SYSACL, /* SYSACL */ 1205 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ 1206 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ 1207 DMU_OT_NEXT_CLONES, /* ZAP */ 1208 DMU_OT_SCAN_QUEUE, /* ZAP */ 1209 DMU_OT_USERGROUP_USED, /* ZAP */ 1210 DMU_OT_USERGROUP_QUOTA, /* ZAP */ 1211 DMU_OT_USERREFS, /* ZAP */ 1212 DMU_OT_DDT_ZAP, /* ZAP */ 1213 DMU_OT_DDT_STATS, /* ZAP */ 1214 DMU_OT_SA, /* System attr */ 1215 DMU_OT_SA_MASTER_NODE, /* ZAP */ 1216 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ 1217 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ 1218 DMU_OT_SCAN_XLATE, /* ZAP */ 1219 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ 1220 DMU_OT_DEADLIST, /* ZAP */ 1221 DMU_OT_DEADLIST_HDR, /* UINT64 */ 1222 DMU_OT_DSL_CLONES, /* ZAP */ 1223 DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */ 1224 DMU_OT_NUMTYPES, 1225 1226 /* 1227 * Names for valid types declared with DMU_OT(). 1228 */ 1229 DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE), 1230 DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE), 1231 DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE), 1232 DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE), 1233 DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE), 1234 DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE), 1235 DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE), 1236 DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE), 1237 DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE), 1238 DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE) 1239 } dmu_object_type_t; 1240 1241 typedef enum dmu_objset_type { 1242 DMU_OST_NONE, 1243 DMU_OST_META, 1244 DMU_OST_ZFS, 1245 DMU_OST_ZVOL, 1246 DMU_OST_OTHER, /* For testing only! */ 1247 DMU_OST_ANY, /* Be careful! */ 1248 DMU_OST_NUMTYPES 1249 } dmu_objset_type_t; 1250 1251 #define ZAP_MAXVALUELEN (1024 * 8) 1252 1253 /* 1254 * header for all bonus and spill buffers. 1255 * The header has a fixed portion with a variable number 1256 * of "lengths" depending on the number of variable sized 1257 * attribues which are determined by the "layout number" 1258 */ 1259 1260 #define SA_MAGIC 0x2F505A /* ZFS SA */ 1261 typedef struct sa_hdr_phys { 1262 uint32_t sa_magic; 1263 uint16_t sa_layout_info; /* Encoded with hdrsize and layout number */ 1264 uint16_t sa_lengths[1]; /* optional sizes for variable length attrs */ 1265 /* ... Data follows the lengths. */ 1266 } sa_hdr_phys_t; 1267 1268 /* 1269 * sa_hdr_phys -> sa_layout_info 1270 * 1271 * 16 10 0 1272 * +--------+-------+ 1273 * | hdrsz |layout | 1274 * +--------+-------+ 1275 * 1276 * Bits 0-10 are the layout number 1277 * Bits 11-16 are the size of the header. 1278 * The hdrsize is the number * 8 1279 * 1280 * For example. 1281 * hdrsz of 1 ==> 8 byte header 1282 * 2 ==> 16 byte header 1283 * 1284 */ 1285 1286 #define SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10) 1287 #define SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 16, 3, 0) 1288 #define SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \ 1289 { \ 1290 BF32_SET_SB(x, 10, 6, 3, 0, size); \ 1291 BF32_SET(x, 0, 10, num); \ 1292 } 1293 1294 #define SA_ATTR_BSWAP(x) BF32_GET(x, 16, 8) 1295 #define SA_ATTR_LENGTH(x) BF32_GET(x, 24, 16) 1296 #define SA_ATTR_NUM(x) BF32_GET(x, 0, 16) 1297 #define SA_ATTR_ENCODE(x, attr, length, bswap) \ 1298 { \ 1299 BF64_SET(x, 24, 16, length); \ 1300 BF64_SET(x, 16, 8, bswap); \ 1301 BF64_SET(x, 0, 16, attr); \ 1302 } 1303 1304 #define SA_MODE_OFFSET 0 1305 #define SA_SIZE_OFFSET 8 1306 #define SA_GEN_OFFSET 16 1307 #define SA_UID_OFFSET 24 1308 #define SA_GID_OFFSET 32 1309 #define SA_PARENT_OFFSET 40 1310 #define SA_SYMLINK_OFFSET 160 1311 1312 #define SA_REGISTRY "REGISTRY" 1313 #define SA_LAYOUTS "LAYOUTS" 1314 1315 typedef enum sa_bswap_type { 1316 SA_UINT64_ARRAY, 1317 SA_UINT32_ARRAY, 1318 SA_UINT16_ARRAY, 1319 SA_UINT8_ARRAY, 1320 SA_ACL, 1321 } sa_bswap_type_t; 1322 1323 typedef uint16_t sa_attr_type_t; 1324 1325 #define ZIO_OBJSET_MAC_LEN 32 1326 1327 /* 1328 * Intent log header - this on disk structure holds fields to manage 1329 * the log. All fields are 64 bit to easily handle cross architectures. 1330 */ 1331 typedef struct zil_header { 1332 uint64_t zh_claim_txg; /* txg in which log blocks were claimed */ 1333 uint64_t zh_replay_seq; /* highest replayed sequence number */ 1334 blkptr_t zh_log; /* log chain */ 1335 uint64_t zh_claim_seq; /* highest claimed sequence number */ 1336 uint64_t zh_pad[5]; 1337 } zil_header_t; 1338 1339 #define OBJSET_PHYS_SIZE_V2 2048 1340 #define OBJSET_PHYS_SIZE_V3 4096 1341 1342 typedef struct objset_phys { 1343 dnode_phys_t os_meta_dnode; 1344 zil_header_t os_zil_header; 1345 uint64_t os_type; 1346 uint64_t os_flags; 1347 uint8_t os_portable_mac[ZIO_OBJSET_MAC_LEN]; 1348 uint8_t os_local_mac[ZIO_OBJSET_MAC_LEN]; 1349 char os_pad0[OBJSET_PHYS_SIZE_V2 - sizeof (dnode_phys_t)*3 - 1350 sizeof (zil_header_t) - sizeof (uint64_t)*2 - 1351 2*ZIO_OBJSET_MAC_LEN]; 1352 dnode_phys_t os_userused_dnode; 1353 dnode_phys_t os_groupused_dnode; 1354 dnode_phys_t os_projectused_dnode; 1355 char os_pad1[OBJSET_PHYS_SIZE_V3 - OBJSET_PHYS_SIZE_V2 - 1356 sizeof (dnode_phys_t)]; 1357 } objset_phys_t; 1358 1359 typedef struct space_map_phys { 1360 /* object number: not needed but kept for backwards compatibility */ 1361 uint64_t smp_object; 1362 1363 /* length of the object in bytes */ 1364 uint64_t smp_length; 1365 1366 /* space allocated from the map */ 1367 int64_t smp_alloc; 1368 } space_map_phys_t; 1369 1370 typedef enum { 1371 SM_ALLOC, 1372 SM_FREE 1373 } maptype_t; 1374 1375 /* one-word entry constants */ 1376 #define SM_DEBUG_PREFIX 2 1377 #define SM_OFFSET_BITS 47 1378 #define SM_RUN_BITS 15 1379 1380 /* two-word entry constants */ 1381 #define SM2_PREFIX 3 1382 #define SM2_OFFSET_BITS 63 1383 #define SM2_RUN_BITS 36 1384 1385 #define SM_PREFIX_DECODE(x) BF64_DECODE(x, 62, 2) 1386 #define SM_PREFIX_ENCODE(x) BF64_ENCODE(x, 62, 2) 1387 1388 #define SM_DEBUG_ACTION_DECODE(x) BF64_DECODE(x, 60, 2) 1389 #define SM_DEBUG_ACTION_ENCODE(x) BF64_ENCODE(x, 60, 2) 1390 #define SM_DEBUG_SYNCPASS_DECODE(x) BF64_DECODE(x, 50, 10) 1391 #define SM_DEBUG_SYNCPASS_ENCODE(x) BF64_ENCODE(x, 50, 10) 1392 #define SM_DEBUG_TXG_DECODE(x) BF64_DECODE(x, 0, 50) 1393 #define SM_DEBUG_TXG_ENCODE(x) BF64_ENCODE(x, 0, 50) 1394 1395 #define SM_OFFSET_DECODE(x) BF64_DECODE(x, 16, SM_OFFSET_BITS) 1396 #define SM_OFFSET_ENCODE(x) BF64_ENCODE(x, 16, SM_OFFSET_BITS) 1397 #define SM_TYPE_DECODE(x) BF64_DECODE(x, 15, 1) 1398 #define SM_TYPE_ENCODE(x) BF64_ENCODE(x, 15, 1) 1399 #define SM_RUN_DECODE(x) (BF64_DECODE(x, 0, SM_RUN_BITS) + 1) 1400 #define SM_RUN_ENCODE(x) BF64_ENCODE((x) - 1, 0, SM_RUN_BITS) 1401 #define SM_RUN_MAX SM_RUN_DECODE(~0ULL) 1402 #define SM_OFFSET_MAX SM_OFFSET_DECODE(~0ULL) 1403 1404 #define SM2_RUN_DECODE(x) (BF64_DECODE(x, 24, SM2_RUN_BITS) + 1) 1405 #define SM2_RUN_ENCODE(x) BF64_ENCODE((x) - 1, 24, SM2_RUN_BITS) 1406 #define SM2_VDEV_DECODE(x) BF64_DECODE(x, 0, 24) 1407 #define SM2_VDEV_ENCODE(x) BF64_ENCODE(x, 0, 24) 1408 #define SM2_TYPE_DECODE(x) BF64_DECODE(x, SM2_OFFSET_BITS, 1) 1409 #define SM2_TYPE_ENCODE(x) BF64_ENCODE(x, SM2_OFFSET_BITS, 1) 1410 #define SM2_OFFSET_DECODE(x) BF64_DECODE(x, 0, SM2_OFFSET_BITS) 1411 #define SM2_OFFSET_ENCODE(x) BF64_ENCODE(x, 0, SM2_OFFSET_BITS) 1412 #define SM2_RUN_MAX SM2_RUN_DECODE(~0ULL) 1413 #define SM2_OFFSET_MAX SM2_OFFSET_DECODE(~0ULL) 1414 1415 typedef enum dd_used { 1416 DD_USED_HEAD, 1417 DD_USED_SNAP, 1418 DD_USED_CHILD, 1419 DD_USED_CHILD_RSRV, 1420 DD_USED_REFRSRV, 1421 DD_USED_NUM 1422 } dd_used_t; 1423 1424 #define DD_FLAG_USED_BREAKDOWN (1 << 0) 1425 1426 typedef struct dsl_dir_phys { 1427 uint64_t dd_creation_time; /* not actually used */ 1428 uint64_t dd_head_dataset_obj; 1429 uint64_t dd_parent_obj; 1430 uint64_t dd_clone_parent_obj; 1431 uint64_t dd_child_dir_zapobj; 1432 /* 1433 * how much space our children are accounting for; for leaf 1434 * datasets, == physical space used by fs + snaps 1435 */ 1436 uint64_t dd_used_bytes; 1437 uint64_t dd_compressed_bytes; 1438 uint64_t dd_uncompressed_bytes; 1439 /* Administrative quota setting */ 1440 uint64_t dd_quota; 1441 /* Administrative reservation setting */ 1442 uint64_t dd_reserved; 1443 uint64_t dd_props_zapobj; 1444 uint64_t dd_pad[1]; 1445 uint64_t dd_flags; 1446 uint64_t dd_used_breakdown[DD_USED_NUM]; 1447 uint64_t dd_clones; 1448 uint64_t dd_pad1[13]; /* pad out to 256 bytes for good measure */ 1449 } dsl_dir_phys_t; 1450 1451 typedef struct dsl_dataset_phys { 1452 uint64_t ds_dir_obj; 1453 uint64_t ds_prev_snap_obj; 1454 uint64_t ds_prev_snap_txg; 1455 uint64_t ds_next_snap_obj; 1456 uint64_t ds_snapnames_zapobj; /* zap obj of snaps; ==0 for snaps */ 1457 uint64_t ds_num_children; /* clone/snap children; ==0 for head */ 1458 uint64_t ds_creation_time; /* seconds since 1970 */ 1459 uint64_t ds_creation_txg; 1460 uint64_t ds_deadlist_obj; 1461 uint64_t ds_used_bytes; 1462 uint64_t ds_compressed_bytes; 1463 uint64_t ds_uncompressed_bytes; 1464 uint64_t ds_unique_bytes; /* only relevant to snapshots */ 1465 /* 1466 * The ds_fsid_guid is a 56-bit ID that can change to avoid 1467 * collisions. The ds_guid is a 64-bit ID that will never 1468 * change, so there is a small probability that it will collide. 1469 */ 1470 uint64_t ds_fsid_guid; 1471 uint64_t ds_guid; 1472 uint64_t ds_flags; 1473 blkptr_t ds_bp; 1474 uint64_t ds_next_clones_obj; /* DMU_OT_DSL_CLONES */ 1475 uint64_t ds_props_obj; /* DMU_OT_DSL_PROPS for snaps */ 1476 uint64_t ds_userrefs_obj; /* DMU_OT_USERREFS */ 1477 uint64_t ds_pad[5]; /* pad out to 320 bytes for good measure */ 1478 } dsl_dataset_phys_t; 1479 1480 typedef struct dsl_deadlist_phys { 1481 uint64_t dl_used; 1482 uint64_t dl_comp; 1483 uint64_t dl_uncomp; 1484 uint64_t dl_pad[37]; /* pad out to 320b for future expansion */ 1485 } dsl_deadlist_phys_t; 1486 1487 #define BPOBJ_SIZE_V2 (6 * sizeof (uint64_t)) 1488 1489 typedef struct bpobj_phys { 1490 uint64_t bpo_num_blkptrs; 1491 uint64_t bpo_bytes; 1492 uint64_t bpo_comp; 1493 uint64_t bpo_uncomp; 1494 uint64_t bpo_subobjs; 1495 uint64_t bpo_num_subobjs; 1496 uint64_t bpo_num_freed; 1497 } bpobj_phys_t; 1498 1499 /* 1500 * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 1501 */ 1502 #define DMU_POOL_DIRECTORY_OBJECT 1 1503 #define DMU_POOL_CONFIG "config" 1504 #define DMU_POOL_FEATURES_FOR_READ "features_for_read" 1505 #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write" 1506 #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions" 1507 #define DMU_POOL_ROOT_DATASET "root_dataset" 1508 #define DMU_POOL_SYNC_BPLIST "sync_bplist" 1509 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 1510 #define DMU_POOL_ERRLOG_LAST "errlog_last" 1511 #define DMU_POOL_SPARES "spares" 1512 #define DMU_POOL_DEFLATE "deflate" 1513 #define DMU_POOL_HISTORY "history" 1514 #define DMU_POOL_PROPS "pool_props" 1515 #define DMU_POOL_FREE_BPOBJ "free_bpobj" 1516 #define DMU_POOL_BPTREE_OBJ "bptree_obj" 1517 #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj" 1518 #define DMU_POOL_TMP_USERREFS "tmp_userrefs" 1519 #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" 1520 #define DMU_POOL_REMOVING "com.delphix:removing" 1521 #define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj" 1522 #define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect" 1523 #define DMU_POOL_ZPOOL_CHECKPOINT "com.delphix:zpool_checkpoint" 1524 1525 #define ZAP_MAGIC 0x2F52AB2ABULL 1526 1527 #define FZAP_BLOCK_SHIFT(zap) ((zap)->zap_block_shift) 1528 1529 #define ZAP_MAXCD (uint32_t)(-1) 1530 #define ZAP_HASHBITS 28 1531 #define MZAP_ENT_LEN 64 1532 #define MZAP_ENT_MAX \ 1533 ((MZAP_MAX_BLKSZ - sizeof(mzap_phys_t)) / sizeof(mzap_ent_phys_t) + 1) 1534 #define MZAP_NAME_LEN (MZAP_ENT_LEN - 8 - 4 - 2) 1535 #define MZAP_MAX_BLKSZ SPA_OLDMAXBLOCKSIZE 1536 1537 typedef struct mzap_ent_phys { 1538 uint64_t mze_value; 1539 uint32_t mze_cd; 1540 uint16_t mze_pad; /* in case we want to chain them someday */ 1541 char mze_name[MZAP_NAME_LEN]; 1542 } mzap_ent_phys_t; 1543 1544 typedef struct mzap_phys { 1545 uint64_t mz_block_type; /* ZBT_MICRO */ 1546 uint64_t mz_salt; 1547 uint64_t mz_normflags; 1548 uint64_t mz_pad[5]; 1549 mzap_ent_phys_t mz_chunk[1]; 1550 /* actually variable size depending on block size */ 1551 } mzap_phys_t; 1552 1553 /* 1554 * The (fat) zap is stored in one object. It is an array of 1555 * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of: 1556 * 1557 * ptrtbl fits in first block: 1558 * [zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ... 1559 * 1560 * ptrtbl too big for first block: 1561 * [zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ... 1562 * 1563 */ 1564 1565 #define ZBT_LEAF ((1ULL << 63) + 0) 1566 #define ZBT_HEADER ((1ULL << 63) + 1) 1567 #define ZBT_MICRO ((1ULL << 63) + 3) 1568 /* any other values are ptrtbl blocks */ 1569 1570 /* 1571 * the embedded pointer table takes up half a block: 1572 * block size / entry size (2^3) / 2 1573 */ 1574 #define ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1) 1575 1576 /* 1577 * The embedded pointer table starts half-way through the block. Since 1578 * the pointer table itself is half the block, it starts at (64-bit) 1579 * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)). 1580 */ 1581 #define ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \ 1582 ((uint64_t *)(zap)->zap_phys) \ 1583 [(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))] 1584 1585 #define ZAP_HASH_IDX(hash, n) (((n) == 0) ? 0 : ((hash) >> (64 - (n)))) 1586 1587 /* 1588 * TAKE NOTE: 1589 * If zap_phys_t is modified, zap_byteswap() must be modified. 1590 */ 1591 typedef struct zap_phys { 1592 uint64_t zap_block_type; /* ZBT_HEADER */ 1593 uint64_t zap_magic; /* ZAP_MAGIC */ 1594 1595 struct zap_table_phys { 1596 uint64_t zt_blk; /* starting block number */ 1597 uint64_t zt_numblks; /* number of blocks */ 1598 uint64_t zt_shift; /* bits to index it */ 1599 uint64_t zt_nextblk; /* next (larger) copy start block */ 1600 uint64_t zt_blks_copied; /* number source blocks copied */ 1601 } zap_ptrtbl; 1602 1603 uint64_t zap_freeblk; /* the next free block */ 1604 uint64_t zap_num_leafs; /* number of leafs */ 1605 uint64_t zap_num_entries; /* number of entries */ 1606 uint64_t zap_salt; /* salt to stir into hash function */ 1607 uint64_t zap_normflags; /* flags for u8_textprep_str() */ 1608 uint64_t zap_flags; /* zap_flags_t */ 1609 /* 1610 * This structure is followed by padding, and then the embedded 1611 * pointer table. The embedded pointer table takes up second 1612 * half of the block. It is accessed using the 1613 * ZAP_EMBEDDED_PTRTBL_ENT() macro. 1614 */ 1615 } zap_phys_t; 1616 1617 typedef struct zap_table_phys zap_table_phys_t; 1618 1619 struct spa; 1620 typedef struct fat_zap { 1621 int zap_block_shift; /* block size shift */ 1622 zap_phys_t *zap_phys; 1623 const struct spa *zap_spa; 1624 const dnode_phys_t *zap_dnode; 1625 } fat_zap_t; 1626 1627 #define ZAP_LEAF_MAGIC 0x2AB1EAF 1628 1629 /* chunk size = 24 bytes */ 1630 #define ZAP_LEAF_CHUNKSIZE 24 1631 1632 /* 1633 * The amount of space available for chunks is: 1634 * block size (1<<l->l_bs) - hash entry size (2) * number of hash 1635 * entries - header space (2*chunksize) 1636 */ 1637 #define ZAP_LEAF_NUMCHUNKS(l) \ 1638 (((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \ 1639 ZAP_LEAF_CHUNKSIZE - 2) 1640 1641 /* 1642 * The amount of space within the chunk available for the array is: 1643 * chunk size - space for type (1) - space for next pointer (2) 1644 */ 1645 #define ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3) 1646 1647 #define ZAP_LEAF_ARRAY_NCHUNKS(bytes) \ 1648 (((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES) 1649 1650 /* 1651 * Low water mark: when there are only this many chunks free, start 1652 * growing the ptrtbl. Ideally, this should be larger than a 1653 * "reasonably-sized" entry. 20 chunks is more than enough for the 1654 * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value), 1655 * while still being only around 3% for 16k blocks. 1656 */ 1657 #define ZAP_LEAF_LOW_WATER (20) 1658 1659 /* 1660 * The leaf hash table has block size / 2^5 (32) number of entries, 1661 * which should be more than enough for the maximum number of entries, 1662 * which is less than block size / CHUNKSIZE (24) / minimum number of 1663 * chunks per entry (3). 1664 */ 1665 #define ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5) 1666 #define ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l)) 1667 1668 /* 1669 * The chunks start immediately after the hash table. The end of the 1670 * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a 1671 * chunk_t. 1672 */ 1673 #define ZAP_LEAF_CHUNK(l, idx) \ 1674 ((zap_leaf_chunk_t *)(void *) \ 1675 ((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx] 1676 #define ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry) 1677 1678 #define ZAP_LEAF_HASH(l, h) \ 1679 ((ZAP_LEAF_HASH_NUMENTRIES(l)-1) & \ 1680 ((h) >> \ 1681 (64 - ZAP_LEAF_HASH_SHIFT(l) - (l)->l_phys->l_hdr.lh_prefix_len))) 1682 #define ZAP_LEAF_HASH_ENTPTR(l, h) (&(l)->l_phys->l_hash[ZAP_LEAF_HASH(l, h)]) 1683 1684 typedef enum zap_chunk_type { 1685 ZAP_CHUNK_FREE = 253, 1686 ZAP_CHUNK_ENTRY = 252, 1687 ZAP_CHUNK_ARRAY = 251, 1688 ZAP_CHUNK_TYPE_MAX = 250 1689 } zap_chunk_type_t; 1690 1691 /* 1692 * TAKE NOTE: 1693 * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified. 1694 */ 1695 typedef struct zap_leaf_phys { 1696 struct zap_leaf_header { 1697 uint64_t lh_block_type; /* ZBT_LEAF */ 1698 uint64_t lh_pad1; 1699 uint64_t lh_prefix; /* hash prefix of this leaf */ 1700 uint32_t lh_magic; /* ZAP_LEAF_MAGIC */ 1701 uint16_t lh_nfree; /* number free chunks */ 1702 uint16_t lh_nentries; /* number of entries */ 1703 uint16_t lh_prefix_len; /* num bits used to id this */ 1704 1705 /* above is accessable to zap, below is zap_leaf private */ 1706 1707 uint16_t lh_freelist; /* chunk head of free list */ 1708 uint8_t lh_pad2[12]; 1709 } l_hdr; /* 2 24-byte chunks */ 1710 1711 /* 1712 * The header is followed by a hash table with 1713 * ZAP_LEAF_HASH_NUMENTRIES(zap) entries. The hash table is 1714 * followed by an array of ZAP_LEAF_NUMCHUNKS(zap) 1715 * zap_leaf_chunk structures. These structures are accessed 1716 * with the ZAP_LEAF_CHUNK() macro. 1717 */ 1718 1719 uint16_t l_hash[1]; 1720 } zap_leaf_phys_t; 1721 1722 typedef union zap_leaf_chunk { 1723 struct zap_leaf_entry { 1724 uint8_t le_type; /* always ZAP_CHUNK_ENTRY */ 1725 uint8_t le_value_intlen; /* size of ints */ 1726 uint16_t le_next; /* next entry in hash chain */ 1727 uint16_t le_name_chunk; /* first chunk of the name */ 1728 uint16_t le_name_numints; /* bytes in name, incl null */ 1729 uint16_t le_value_chunk; /* first chunk of the value */ 1730 uint16_t le_value_numints; /* value length in ints */ 1731 uint32_t le_cd; /* collision differentiator */ 1732 uint64_t le_hash; /* hash value of the name */ 1733 } l_entry; 1734 struct zap_leaf_array { 1735 uint8_t la_type; /* always ZAP_CHUNK_ARRAY */ 1736 uint8_t la_array[ZAP_LEAF_ARRAY_BYTES]; 1737 uint16_t la_next; /* next blk or CHAIN_END */ 1738 } l_array; 1739 struct zap_leaf_free { 1740 uint8_t lf_type; /* always ZAP_CHUNK_FREE */ 1741 uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES]; 1742 uint16_t lf_next; /* next in free list, or CHAIN_END */ 1743 } l_free; 1744 } zap_leaf_chunk_t; 1745 1746 typedef struct zap_leaf { 1747 int l_bs; /* block size shift */ 1748 zap_leaf_phys_t *l_phys; 1749 } zap_leaf_t; 1750 1751 #define ZAP_MAXNAMELEN 256 1752 #define ZAP_MAXVALUELEN (1024 * 8) 1753 1754 #define ACE_READ_DATA 0x00000001 /* file: read data */ 1755 #define ACE_LIST_DIRECTORY 0x00000001 /* dir: list files */ 1756 #define ACE_WRITE_DATA 0x00000002 /* file: write data */ 1757 #define ACE_ADD_FILE 0x00000002 /* dir: create file */ 1758 #define ACE_APPEND_DATA 0x00000004 /* file: append data */ 1759 #define ACE_ADD_SUBDIRECTORY 0x00000004 /* dir: create subdir */ 1760 #define ACE_READ_NAMED_ATTRS 0x00000008 /* FILE_READ_EA */ 1761 #define ACE_WRITE_NAMED_ATTRS 0x00000010 /* FILE_WRITE_EA */ 1762 #define ACE_EXECUTE 0x00000020 /* file: execute */ 1763 #define ACE_TRAVERSE 0x00000020 /* dir: lookup name */ 1764 #define ACE_DELETE_CHILD 0x00000040 /* dir: unlink child */ 1765 #define ACE_READ_ATTRIBUTES 0x00000080 /* (all) stat, etc. */ 1766 #define ACE_WRITE_ATTRIBUTES 0x00000100 /* (all) utimes, etc. */ 1767 #define ACE_DELETE 0x00010000 /* (all) unlink self */ 1768 #define ACE_READ_ACL 0x00020000 /* (all) getsecattr */ 1769 #define ACE_WRITE_ACL 0x00040000 /* (all) setsecattr */ 1770 #define ACE_WRITE_OWNER 0x00080000 /* (all) chown */ 1771 #define ACE_SYNCHRONIZE 0x00100000 /* (all) */ 1772 1773 #define ACE_FILE_INHERIT_ACE 0x0001 1774 #define ACE_DIRECTORY_INHERIT_ACE 0x0002 1775 #define ACE_NO_PROPAGATE_INHERIT_ACE 0x0004 1776 #define ACE_INHERIT_ONLY_ACE 0x0008 1777 #define ACE_SUCCESSFUL_ACCESS_ACE_FLAG 0x0010 1778 #define ACE_FAILED_ACCESS_ACE_FLAG 0x0020 1779 #define ACE_IDENTIFIER_GROUP 0x0040 1780 #define ACE_INHERITED_ACE 0x0080 1781 #define ACE_OWNER 0x1000 1782 #define ACE_GROUP 0x2000 1783 #define ACE_EVERYONE 0x4000 1784 1785 #define ACE_ACCESS_ALLOWED_ACE_TYPE 0x0000 1786 #define ACE_ACCESS_DENIED_ACE_TYPE 0x0001 1787 #define ACE_SYSTEM_AUDIT_ACE_TYPE 0x0002 1788 #define ACE_SYSTEM_ALARM_ACE_TYPE 0x0003 1789 1790 typedef struct zfs_ace_hdr { 1791 uint16_t z_type; 1792 uint16_t z_flags; 1793 uint32_t z_access_mask; 1794 } zfs_ace_hdr_t; 1795 1796 /* 1797 * Define special zfs pflags 1798 */ 1799 #define ZFS_XATTR 0x1 /* is an extended attribute */ 1800 #define ZFS_INHERIT_ACE 0x2 /* ace has inheritable ACEs */ 1801 #define ZFS_ACL_TRIVIAL 0x4 /* files ACL is trivial */ 1802 #define ZFS_ACL_OBJ_ACE 0x8 /* ACL has CMPLX Object ACE */ 1803 #define ZFS_ACL_PROTECTED 0x10 /* ACL protected */ 1804 #define ZFS_ACL_DEFAULTED 0x20 /* ACL should be defaulted */ 1805 #define ZFS_ACL_AUTO_INHERIT 0x40 /* ACL should be inherited */ 1806 #define ZFS_BONUS_SCANSTAMP 0x80 /* Scanstamp in bonus area */ 1807 #define ZFS_NO_EXECS_DENIED 0x100 /* exec was given to everyone */ 1808 1809 #define ZFS_READONLY 0x0000000100000000ull 1810 #define ZFS_HIDDEN 0x0000000200000000ull 1811 #define ZFS_SYSTEM 0x0000000400000000ull 1812 #define ZFS_ARCHIVE 0x0000000800000000ull 1813 #define ZFS_IMMUTABLE 0x0000001000000000ull 1814 #define ZFS_NOUNLINK 0x0000002000000000ull 1815 #define ZFS_APPENDONLY 0x0000004000000000ull 1816 #define ZFS_NODUMP 0x0000008000000000ull 1817 #define ZFS_OPAQUE 0x0000010000000000ull 1818 #define ZFS_AV_QUARANTINED 0x0000020000000000ull 1819 #define ZFS_AV_MODIFIED 0x0000040000000000ull 1820 #define ZFS_REPARSE 0x0000080000000000ull 1821 #define ZFS_OFFLINE 0x0000100000000000ull 1822 #define ZFS_SPARSE 0x0000200000000000ull 1823 1824 #define MASTER_NODE_OBJ 1 1825 1826 /* 1827 * special attributes for master node. 1828 */ 1829 1830 #define ZFS_FSID "FSID" 1831 #define ZFS_UNLINKED_SET "DELETE_QUEUE" 1832 #define ZFS_ROOT_OBJ "ROOT" 1833 #define ZPL_VERSION_OBJ "VERSION" 1834 #define ZFS_PROP_BLOCKPERPAGE "BLOCKPERPAGE" 1835 #define ZFS_PROP_NOGROWBLOCKS "NOGROWBLOCKS" 1836 #define ZFS_SA_ATTRS "SA_ATTRS" 1837 1838 #define ZFS_FLAG_BLOCKPERPAGE 0x1 1839 #define ZFS_FLAG_NOGROWBLOCKS 0x2 1840 1841 /* 1842 * ZPL version - rev'd whenever an incompatible on-disk format change 1843 * occurs. Independent of SPA/DMU/ZAP versioning. 1844 */ 1845 1846 #define ZPL_VERSION 1ULL 1847 1848 /* 1849 * The directory entry has the type (currently unused on Solaris) in the 1850 * top 4 bits, and the object number in the low 48 bits. The "middle" 1851 * 12 bits are unused. 1852 */ 1853 #define ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4) 1854 #define ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48) 1855 #define ZFS_DIRENT_MAKE(type, obj) (((uint64_t)type << 60) | obj) 1856 1857 typedef struct ace { 1858 uid_t a_who; /* uid or gid */ 1859 uint32_t a_access_mask; /* read,write,... */ 1860 uint16_t a_flags; /* see below */ 1861 uint16_t a_type; /* allow or deny */ 1862 } ace_t; 1863 1864 #define ACE_SLOT_CNT 6 1865 1866 typedef struct zfs_znode_acl { 1867 uint64_t z_acl_extern_obj; /* ext acl pieces */ 1868 uint32_t z_acl_count; /* Number of ACEs */ 1869 uint16_t z_acl_version; /* acl version */ 1870 uint16_t z_acl_pad; /* pad */ 1871 ace_t z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */ 1872 } zfs_znode_acl_t; 1873 1874 /* 1875 * This is the persistent portion of the znode. It is stored 1876 * in the "bonus buffer" of the file. Short symbolic links 1877 * are also stored in the bonus buffer. 1878 */ 1879 typedef struct znode_phys { 1880 uint64_t zp_atime[2]; /* 0 - last file access time */ 1881 uint64_t zp_mtime[2]; /* 16 - last file modification time */ 1882 uint64_t zp_ctime[2]; /* 32 - last file change time */ 1883 uint64_t zp_crtime[2]; /* 48 - creation time */ 1884 uint64_t zp_gen; /* 64 - generation (txg of creation) */ 1885 uint64_t zp_mode; /* 72 - file mode bits */ 1886 uint64_t zp_size; /* 80 - size of file */ 1887 uint64_t zp_parent; /* 88 - directory parent (`..') */ 1888 uint64_t zp_links; /* 96 - number of links to file */ 1889 uint64_t zp_xattr; /* 104 - DMU object for xattrs */ 1890 uint64_t zp_rdev; /* 112 - dev_t for VBLK & VCHR files */ 1891 uint64_t zp_flags; /* 120 - persistent flags */ 1892 uint64_t zp_uid; /* 128 - file owner */ 1893 uint64_t zp_gid; /* 136 - owning group */ 1894 uint64_t zp_pad[4]; /* 144 - future */ 1895 zfs_znode_acl_t zp_acl; /* 176 - 263 ACL */ 1896 /* 1897 * Data may pad out any remaining bytes in the znode buffer, eg: 1898 * 1899 * |<---------------------- dnode_phys (512) ------------------------>| 1900 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->| 1901 * |<---- znode (264) ---->|<---- data (56) ---->| 1902 * 1903 * At present, we only use this space to store symbolic links. 1904 */ 1905 } znode_phys_t; 1906 1907 /* 1908 * In-core vdev representation. 1909 */ 1910 struct vdev; 1911 struct spa; 1912 typedef int vdev_phys_read_t(struct vdev *, void *, off_t, void *, size_t); 1913 typedef int vdev_phys_write_t(struct vdev *, off_t, void *, size_t); 1914 typedef int vdev_read_t(struct vdev *, const blkptr_t *, void *, off_t, size_t); 1915 1916 typedef STAILQ_HEAD(vdev_list, vdev) vdev_list_t; 1917 1918 typedef struct vdev_indirect_mapping_entry_phys { 1919 /* 1920 * Decode with DVA_MAPPING_* macros. 1921 * Contains: 1922 * the source offset (low 63 bits) 1923 * the one-bit "mark", used for garbage collection (by zdb) 1924 */ 1925 uint64_t vimep_src; 1926 1927 /* 1928 * Note: the DVA's asize is 24 bits, and can thus store ranges 1929 * up to 8GB. 1930 */ 1931 dva_t vimep_dst; 1932 } vdev_indirect_mapping_entry_phys_t; 1933 1934 #define DVA_MAPPING_GET_SRC_OFFSET(vimep) \ 1935 BF64_GET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0) 1936 #define DVA_MAPPING_SET_SRC_OFFSET(vimep, x) \ 1937 BF64_SET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0, x) 1938 1939 /* 1940 * This is stored in the bonus buffer of the mapping object, see comment of 1941 * vdev_indirect_config for more details. 1942 */ 1943 typedef struct vdev_indirect_mapping_phys { 1944 uint64_t vimp_max_offset; 1945 uint64_t vimp_bytes_mapped; 1946 uint64_t vimp_num_entries; /* number of v_i_m_entry_phys_t's */ 1947 1948 /* 1949 * For each entry in the mapping object, this object contains an 1950 * entry representing the number of bytes of that mapping entry 1951 * that were no longer in use by the pool at the time this indirect 1952 * vdev was last condensed. 1953 */ 1954 uint64_t vimp_counts_object; 1955 } vdev_indirect_mapping_phys_t; 1956 1957 #define VDEV_INDIRECT_MAPPING_SIZE_V0 (3 * sizeof (uint64_t)) 1958 1959 typedef struct vdev_indirect_mapping { 1960 uint64_t vim_object; 1961 boolean_t vim_havecounts; 1962 1963 /* vim_entries segment offset currently in memory. */ 1964 uint64_t vim_entry_offset; 1965 /* vim_entries segment size. */ 1966 size_t vim_num_entries; 1967 1968 /* Needed by dnode_read() */ 1969 const void *vim_spa; 1970 dnode_phys_t *vim_dn; 1971 1972 /* 1973 * An ordered array of mapping entries, sorted by source offset. 1974 * Note that vim_entries is needed during a removal (and contains 1975 * mappings that have been synced to disk so far) to handle frees 1976 * from the removing device. 1977 */ 1978 vdev_indirect_mapping_entry_phys_t *vim_entries; 1979 objset_phys_t *vim_objset; 1980 vdev_indirect_mapping_phys_t *vim_phys; 1981 } vdev_indirect_mapping_t; 1982 1983 /* 1984 * On-disk indirect vdev state. 1985 * 1986 * An indirect vdev is described exclusively in the MOS config of a pool. 1987 * The config for an indirect vdev includes several fields, which are 1988 * accessed in memory by a vdev_indirect_config_t. 1989 */ 1990 typedef struct vdev_indirect_config { 1991 /* 1992 * Object (in MOS) which contains the indirect mapping. This object 1993 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by 1994 * vimep_src. The bonus buffer for this object is a 1995 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev 1996 * removal is initiated. 1997 * 1998 * Note that this object can be empty if none of the data on the vdev 1999 * has been copied yet. 2000 */ 2001 uint64_t vic_mapping_object; 2002 2003 /* 2004 * Object (in MOS) which contains the birth times for the mapping 2005 * entries. This object contains an array of 2006 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus 2007 * buffer for this object is a vdev_indirect_birth_phys_t. This object 2008 * is allocated when a vdev removal is initiated. 2009 * 2010 * Note that this object can be empty if none of the vdev has yet been 2011 * copied. 2012 */ 2013 uint64_t vic_births_object; 2014 2015 /* 2016 * This is the vdev ID which was removed previous to this vdev, or 2017 * UINT64_MAX if there are no previously removed vdevs. 2018 */ 2019 uint64_t vic_prev_indirect_vdev; 2020 } vdev_indirect_config_t; 2021 2022 typedef struct vdev { 2023 STAILQ_ENTRY(vdev) v_childlink; /* link in parent's child list */ 2024 vdev_list_t v_children; /* children of this vdev */ 2025 const char *v_name; /* vdev name */ 2026 uint64_t v_guid; /* vdev guid */ 2027 uint64_t v_label; /* label instantiated from (top vdev) */ 2028 uint64_t v_txg; /* most recent transaction (top vdev) */ 2029 uint64_t v_id; /* index in parent */ 2030 uint64_t v_psize; /* physical device capacity */ 2031 int v_ashift; /* offset to block shift */ 2032 int v_nparity; /* # parity for raidz */ 2033 struct vdev *v_top; /* parent vdev */ 2034 size_t v_nchildren; /* # children */ 2035 vdev_state_t v_state; /* current state */ 2036 vdev_phys_read_t *v_phys_read; /* read from raw leaf vdev */ 2037 vdev_phys_write_t *v_phys_write; /* write to raw leaf vdev */ 2038 vdev_read_t *v_read; /* read from vdev */ 2039 void *v_priv; /* data for read/write function */ 2040 boolean_t v_islog; 2041 struct spa *v_spa; /* link to spa */ 2042 /* 2043 * Values stored in the config for an indirect or removing vdev. 2044 */ 2045 vdev_indirect_config_t vdev_indirect_config; 2046 vdev_indirect_mapping_t *v_mapping; 2047 } vdev_t; 2048 2049 /* 2050 * In-core pool representation. 2051 */ 2052 typedef STAILQ_HEAD(spa_list, spa) spa_list_t; 2053 2054 typedef struct spa { 2055 STAILQ_ENTRY(spa) spa_link; /* link in global pool list */ 2056 char *spa_name; /* pool name */ 2057 uint64_t spa_guid; /* pool guid */ 2058 struct uberblock *spa_uberblock; /* best uberblock so far */ 2059 vdev_t *spa_root_vdev; /* toplevel vdev container */ 2060 objset_phys_t *spa_mos; /* MOS for this pool */ 2061 zio_cksum_salt_t spa_cksum_salt; /* secret salt for cksum */ 2062 void *spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS]; 2063 boolean_t spa_with_log; /* this pool has log */ 2064 2065 struct uberblock spa_uberblock_master; /* best uberblock so far */ 2066 objset_phys_t spa_mos_master; /* MOS for this pool */ 2067 struct uberblock spa_uberblock_checkpoint; /* checkpoint uberblock */ 2068 objset_phys_t spa_mos_checkpoint; /* Checkpoint MOS */ 2069 void *spa_bootenv; /* bootenv from pool label */ 2070 } spa_t; 2071 2072 /* IO related arguments. */ 2073 typedef struct zio { 2074 spa_t *io_spa; 2075 blkptr_t *io_bp; 2076 void *io_data; 2077 uint64_t io_size; 2078 uint64_t io_offset; 2079 2080 /* Stuff for the vdev stack */ 2081 vdev_t *io_vd; 2082 void *io_vsd; 2083 2084 int io_error; 2085 } zio_t; 2086 2087 extern void decode_embedded_bp_compressed(const blkptr_t *, void *); 2088 2089 #endif /* _ZFSIMPL_H_ */ 2090