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 #define IS_P2ALIGNED(v, a) ((((uintptr_t)(v)) & ((uintptr_t)(a) - 1)) == 0) 98 99 /* 100 * General-purpose 32-bit and 64-bit bitfield encodings. 101 */ 102 #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) 103 #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) 104 #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) 105 #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) 106 107 #define BF32_GET(x, low, len) BF32_DECODE(x, low, len) 108 #define BF64_GET(x, low, len) BF64_DECODE(x, low, len) 109 110 #define BF32_SET(x, low, len, val) \ 111 ((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len)) 112 #define BF64_SET(x, low, len, val) \ 113 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)) 114 115 #define BF32_GET_SB(x, low, len, shift, bias) \ 116 ((BF32_GET(x, low, len) + (bias)) << (shift)) 117 #define BF64_GET_SB(x, low, len, shift, bias) \ 118 ((BF64_GET(x, low, len) + (bias)) << (shift)) 119 120 #define BF32_SET_SB(x, low, len, shift, bias, val) \ 121 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)) 122 #define BF64_SET_SB(x, low, len, shift, bias, val) \ 123 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)) 124 125 /* 126 * Macros to reverse byte order 127 */ 128 #define BSWAP_8(x) ((x) & 0xff) 129 #define BSWAP_16(x) ((BSWAP_8(x) << 8) | BSWAP_8((x) >> 8)) 130 #define BSWAP_32(x) ((BSWAP_16(x) << 16) | BSWAP_16((x) >> 16)) 131 #define BSWAP_64(x) ((BSWAP_32(x) << 32) | BSWAP_32((x) >> 32)) 132 133 #define SPA_MINBLOCKSHIFT 9 134 #define SPA_OLDMAXBLOCKSHIFT 17 135 #define SPA_MAXBLOCKSHIFT 24 136 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 137 #define SPA_OLDMAXBLOCKSIZE (1ULL << SPA_OLDMAXBLOCKSHIFT) 138 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 139 140 /* 141 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. 142 * The ASIZE encoding should be at least 64 times larger (6 more bits) 143 * to support up to 4-way RAID-Z mirror mode with worst-case gang block 144 * overhead, three DVAs per bp, plus one more bit in case we do anything 145 * else that expands the ASIZE. 146 */ 147 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ 148 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ 149 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ 150 151 /* 152 * All SPA data is represented by 128-bit data virtual addresses (DVAs). 153 * The members of the dva_t should be considered opaque outside the SPA. 154 */ 155 typedef struct dva { 156 uint64_t dva_word[2]; 157 } dva_t; 158 159 /* 160 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. 161 */ 162 typedef struct zio_cksum { 163 uint64_t zc_word[4]; 164 } zio_cksum_t; 165 166 /* 167 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept 168 * secret and is suitable for use in MAC algorithms as the key. 169 */ 170 typedef struct zio_cksum_salt { 171 uint8_t zcs_bytes[32]; 172 } zio_cksum_salt_t; 173 174 /* 175 * Each block is described by its DVAs, time of birth, checksum, etc. 176 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 177 * 178 * 64 56 48 40 32 24 16 8 0 179 * +-------+-------+-------+-------+-------+-------+-------+-------+ 180 * 0 | vdev1 | GRID | ASIZE | 181 * +-------+-------+-------+-------+-------+-------+-------+-------+ 182 * 1 |G| offset1 | 183 * +-------+-------+-------+-------+-------+-------+-------+-------+ 184 * 2 | vdev2 | GRID | ASIZE | 185 * +-------+-------+-------+-------+-------+-------+-------+-------+ 186 * 3 |G| offset2 | 187 * +-------+-------+-------+-------+-------+-------+-------+-------+ 188 * 4 | vdev3 | GRID | ASIZE | 189 * +-------+-------+-------+-------+-------+-------+-------+-------+ 190 * 5 |G| offset3 | 191 * +-------+-------+-------+-------+-------+-------+-------+-------+ 192 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 193 * +-------+-------+-------+-------+-------+-------+-------+-------+ 194 * 7 | padding | 195 * +-------+-------+-------+-------+-------+-------+-------+-------+ 196 * 8 | padding | 197 * +-------+-------+-------+-------+-------+-------+-------+-------+ 198 * 9 | physical birth txg | 199 * +-------+-------+-------+-------+-------+-------+-------+-------+ 200 * a | logical birth txg | 201 * +-------+-------+-------+-------+-------+-------+-------+-------+ 202 * b | fill count | 203 * +-------+-------+-------+-------+-------+-------+-------+-------+ 204 * c | checksum[0] | 205 * +-------+-------+-------+-------+-------+-------+-------+-------+ 206 * d | checksum[1] | 207 * +-------+-------+-------+-------+-------+-------+-------+-------+ 208 * e | checksum[2] | 209 * +-------+-------+-------+-------+-------+-------+-------+-------+ 210 * f | checksum[3] | 211 * +-------+-------+-------+-------+-------+-------+-------+-------+ 212 * 213 * Legend: 214 * 215 * vdev virtual device ID 216 * offset offset into virtual device 217 * LSIZE logical size 218 * PSIZE physical size (after compression) 219 * ASIZE allocated size (including RAID-Z parity and gang block headers) 220 * GRID RAID-Z layout information (reserved for future use) 221 * cksum checksum function 222 * comp compression function 223 * G gang block indicator 224 * B byteorder (endianness) 225 * D dedup 226 * X encryption (on version 30, which is not supported) 227 * E blkptr_t contains embedded data (see below) 228 * lvl level of indirection 229 * type DMU object type 230 * phys birth txg of block allocation; zero if same as logical birth txg 231 * log. birth transaction group in which the block was logically born 232 * fill count number of non-zero blocks under this bp 233 * checksum[4] 256-bit checksum of the data this bp describes 234 */ 235 236 /* 237 * "Embedded" blkptr_t's don't actually point to a block, instead they 238 * have a data payload embedded in the blkptr_t itself. See the comment 239 * in blkptr.c for more details. 240 * 241 * The blkptr_t is laid out as follows: 242 * 243 * 64 56 48 40 32 24 16 8 0 244 * +-------+-------+-------+-------+-------+-------+-------+-------+ 245 * 0 | payload | 246 * 1 | payload | 247 * 2 | payload | 248 * 3 | payload | 249 * 4 | payload | 250 * 5 | payload | 251 * +-------+-------+-------+-------+-------+-------+-------+-------+ 252 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 253 * +-------+-------+-------+-------+-------+-------+-------+-------+ 254 * 7 | payload | 255 * 8 | payload | 256 * 9 | payload | 257 * +-------+-------+-------+-------+-------+-------+-------+-------+ 258 * a | logical birth txg | 259 * +-------+-------+-------+-------+-------+-------+-------+-------+ 260 * b | payload | 261 * c | payload | 262 * d | payload | 263 * e | payload | 264 * f | payload | 265 * +-------+-------+-------+-------+-------+-------+-------+-------+ 266 * 267 * Legend: 268 * 269 * payload contains the embedded data 270 * B (byteorder) byteorder (endianness) 271 * D (dedup) padding (set to zero) 272 * X encryption (set to zero; see above) 273 * E (embedded) set to one 274 * lvl indirection level 275 * type DMU object type 276 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 277 * comp compression function of payload 278 * PSIZE size of payload after compression, in bytes 279 * LSIZE logical size of payload, in bytes 280 * note that 25 bits is enough to store the largest 281 * "normal" BP's LSIZE (2^16 * 2^9) in bytes 282 * log. birth transaction group in which the block was logically born 283 * 284 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 285 * bp's they are stored in units of SPA_MINBLOCKSHIFT. 286 * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 287 * The B, D, X, lvl, type, and comp fields are stored the same as with normal 288 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 289 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 290 * other macros, as they assert that they are only used on BP's of the correct 291 * "embedded-ness". 292 */ 293 294 #define BPE_GET_ETYPE(bp) \ 295 (ASSERT(BP_IS_EMBEDDED(bp)), \ 296 BF64_GET((bp)->blk_prop, 40, 8)) 297 #define BPE_SET_ETYPE(bp, t) do { \ 298 ASSERT(BP_IS_EMBEDDED(bp)); \ 299 BF64_SET((bp)->blk_prop, 40, 8, t); \ 300 _NOTE(CONSTCOND) } while (0) 301 302 #define BPE_GET_LSIZE(bp) \ 303 (ASSERT(BP_IS_EMBEDDED(bp)), \ 304 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 305 #define BPE_SET_LSIZE(bp, x) do { \ 306 ASSERT(BP_IS_EMBEDDED(bp)); \ 307 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 308 _NOTE(CONSTCOND) } while (0) 309 310 #define BPE_GET_PSIZE(bp) \ 311 (ASSERT(BP_IS_EMBEDDED(bp)), \ 312 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 313 #define BPE_SET_PSIZE(bp, x) do { \ 314 ASSERT(BP_IS_EMBEDDED(bp)); \ 315 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 316 _NOTE(CONSTCOND) } while (0) 317 318 typedef enum bp_embedded_type { 319 BP_EMBEDDED_TYPE_DATA, 320 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ 321 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED 322 } bp_embedded_type_t; 323 324 #define BPE_NUM_WORDS 14 325 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 326 #define BPE_IS_PAYLOADWORD(bp, wp) \ 327 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 328 329 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 330 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 331 332 typedef struct blkptr { 333 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 334 uint64_t blk_prop; /* size, compression, type, etc */ 335 uint64_t blk_pad[2]; /* Extra space for the future */ 336 uint64_t blk_phys_birth; /* txg when block was allocated */ 337 uint64_t blk_birth; /* transaction group at birth */ 338 uint64_t blk_fill; /* fill count */ 339 zio_cksum_t blk_cksum; /* 256-bit checksum */ 340 } blkptr_t; 341 342 /* 343 * Macros to get and set fields in a bp or DVA. 344 */ 345 #define DVA_GET_ASIZE(dva) \ 346 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 347 #define DVA_SET_ASIZE(dva, x) \ 348 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 349 SPA_MINBLOCKSHIFT, 0, x) 350 351 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 352 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 353 354 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) 355 #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) 356 357 #define DVA_GET_OFFSET(dva) \ 358 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 359 #define DVA_SET_OFFSET(dva, x) \ 360 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 361 362 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 363 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 364 365 #define BP_GET_LSIZE(bp) \ 366 (BP_IS_EMBEDDED(bp) ? \ 367 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 368 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 369 #define BP_SET_LSIZE(bp, x) do { \ 370 ASSERT(!BP_IS_EMBEDDED(bp)); \ 371 BF64_SET_SB((bp)->blk_prop, \ 372 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 373 _NOTE(CONSTCOND) } while (0) 374 375 #define BP_GET_PSIZE(bp) \ 376 BF64_GET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1) 377 #define BP_SET_PSIZE(bp, x) \ 378 BF64_SET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x) 379 380 #define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7) 381 #define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x) 382 383 #define BP_GET_CHECKSUM(bp) BF64_GET((bp)->blk_prop, 40, 8) 384 #define BP_SET_CHECKSUM(bp, x) BF64_SET((bp)->blk_prop, 40, 8, x) 385 386 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 387 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 388 389 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 390 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 391 392 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 393 394 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 395 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 396 397 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 398 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 399 400 #define BP_PHYSICAL_BIRTH(bp) \ 401 ((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 402 403 #define BP_SET_BIRTH(bp, logical, physical) \ 404 { \ 405 ASSERT(!BP_IS_EMBEDDED(bp)); \ 406 (bp)->blk_birth = (logical); \ 407 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \ 408 } 409 410 #define BP_GET_FILL(bp) \ 411 ((BP_IS_EMBEDDED(bp)) ? 1 : (bp)->blk_fill) 412 413 #define BP_SET_FILL(bp, fill) \ 414 { \ 415 (bp)->blk_fill = fill; \ 416 } 417 418 #define BP_GET_ASIZE(bp) \ 419 (DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 420 DVA_GET_ASIZE(&(bp)->blk_dva[2])) 421 422 #define BP_GET_UCSIZE(bp) \ 423 ((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \ 424 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)); 425 426 #define BP_GET_NDVAS(bp) \ 427 (!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 428 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 429 !!DVA_GET_ASIZE(&(bp)->blk_dva[2])) 430 431 #define DVA_EQUAL(dva1, dva2) \ 432 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 433 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 434 435 #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 436 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 437 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 438 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 439 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 440 441 442 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 443 444 #define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ 445 { \ 446 (zcp)->zc_word[0] = w0; \ 447 (zcp)->zc_word[1] = w1; \ 448 (zcp)->zc_word[2] = w2; \ 449 (zcp)->zc_word[3] = w3; \ 450 } 451 452 #define BP_IDENTITY(bp) (&(bp)->blk_dva[0]) 453 #define BP_IS_GANG(bp) DVA_GET_GANG(BP_IDENTITY(bp)) 454 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 455 (dva)->dva_word[1] == 0ULL) 456 #define BP_IS_HOLE(bp) DVA_IS_EMPTY(BP_IDENTITY(bp)) 457 #define BP_IS_OLDER(bp, txg) (!BP_IS_HOLE(bp) && (bp)->blk_birth < (txg)) 458 459 #define BP_ZERO(bp) \ 460 { \ 461 (bp)->blk_dva[0].dva_word[0] = 0; \ 462 (bp)->blk_dva[0].dva_word[1] = 0; \ 463 (bp)->blk_dva[1].dva_word[0] = 0; \ 464 (bp)->blk_dva[1].dva_word[1] = 0; \ 465 (bp)->blk_dva[2].dva_word[0] = 0; \ 466 (bp)->blk_dva[2].dva_word[1] = 0; \ 467 (bp)->blk_prop = 0; \ 468 (bp)->blk_pad[0] = 0; \ 469 (bp)->blk_pad[1] = 0; \ 470 (bp)->blk_phys_birth = 0; \ 471 (bp)->blk_birth = 0; \ 472 (bp)->blk_fill = 0; \ 473 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 474 } 475 476 #if BYTE_ORDER == _BIG_ENDIAN 477 #define ZFS_HOST_BYTEORDER (0ULL) 478 #else 479 #define ZFS_HOST_BYTEORDER (1ULL) 480 #endif 481 482 #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) 483 #define BPE_NUM_WORDS 14 484 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 485 #define BPE_IS_PAYLOADWORD(bp, wp) \ 486 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 487 488 /* 489 * Embedded checksum 490 */ 491 #define ZEC_MAGIC 0x210da7ab10c7a11ULL 492 493 typedef struct zio_eck { 494 uint64_t zec_magic; /* for validation, endianness */ 495 zio_cksum_t zec_cksum; /* 256-bit checksum */ 496 } zio_eck_t; 497 498 /* 499 * Gang block headers are self-checksumming and contain an array 500 * of block pointers. 501 */ 502 #define SPA_OLD_GANGBLOCKSIZE SPA_MINBLOCKSIZE 503 504 #define VDEV_RAIDZ_MAXPARITY 3 505 506 #define VDEV_PAD_SIZE (8 << 10) 507 /* 2 padding areas (vl_pad1 and vl_be) to skip */ 508 #define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2 509 #define VDEV_PHYS_SIZE (112 << 10) 510 #define VDEV_UBERBLOCK_RING (128 << 10) 511 512 /* 513 * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock 514 * ring when MMP is enabled. 515 */ 516 #define MMP_BLOCKS_PER_LABEL 1 517 518 /* The largest uberblock we support is 8k. */ 519 #define MAX_UBERBLOCK_SHIFT (13) 520 #define VDEV_UBERBLOCK_SHIFT(vd) \ 521 MIN(MAX((vd)->v_top->v_ashift, UBERBLOCK_SHIFT), MAX_UBERBLOCK_SHIFT) 522 #define VDEV_UBERBLOCK_COUNT(vd) \ 523 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd)) 524 #define VDEV_UBERBLOCK_OFFSET(vd, n) \ 525 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)]) 526 #define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd)) 527 528 #define ASHIFT_UBERBLOCK_SHIFT(ashift) \ 529 MIN(MAX(ashift, UBERBLOCK_SHIFT), \ 530 MAX_UBERBLOCK_SHIFT) 531 #define ASHIFT_UBERBLOCK_SIZE(ashift) \ 532 (1ULL << ASHIFT_UBERBLOCK_SHIFT(ashift)) 533 534 typedef struct vdev_phys { 535 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)]; 536 zio_eck_t vp_zbt; 537 } vdev_phys_t; 538 539 typedef enum vbe_vers { 540 /* The bootenv file is stored as ascii text in the envblock */ 541 VB_RAW = 0, 542 543 /* 544 * The bootenv file is converted to an nvlist and then packed into the 545 * envblock. 546 */ 547 VB_NVLIST = 1 548 } vbe_vers_t; 549 550 typedef struct vdev_boot_envblock { 551 uint64_t vbe_version; 552 char vbe_bootenv[VDEV_PAD_SIZE - sizeof (uint64_t) - 553 sizeof (zio_eck_t)]; 554 zio_eck_t vbe_zbt; 555 } vdev_boot_envblock_t; 556 557 _Static_assert(sizeof (vdev_boot_envblock_t) == VDEV_PAD_SIZE, 558 "bad size for vdev_boot_envblock_t"); 559 560 typedef struct vdev_label { 561 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */ 562 vdev_boot_envblock_t vl_be; /* 8K */ 563 vdev_phys_t vl_vdev_phys; /* 112K */ 564 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */ 565 } vdev_label_t; /* 256K total */ 566 567 /* 568 * vdev_dirty() flags 569 */ 570 #define VDD_METASLAB 0x01 571 #define VDD_DTL 0x02 572 573 /* 574 * Size and offset of embedded boot loader region on each label. 575 * The total size of the first two labels plus the boot area is 4MB. 576 */ 577 #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t)) 578 #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */ 579 580 /* 581 * Size of label regions at the start and end of each leaf device. 582 */ 583 #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE) 584 #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t)) 585 #define VDEV_LABELS 4 586 587 enum zio_checksum { 588 ZIO_CHECKSUM_INHERIT = 0, 589 ZIO_CHECKSUM_ON, 590 ZIO_CHECKSUM_OFF, 591 ZIO_CHECKSUM_LABEL, 592 ZIO_CHECKSUM_GANG_HEADER, 593 ZIO_CHECKSUM_ZILOG, 594 ZIO_CHECKSUM_FLETCHER_2, 595 ZIO_CHECKSUM_FLETCHER_4, 596 ZIO_CHECKSUM_SHA256, 597 ZIO_CHECKSUM_ZILOG2, 598 ZIO_CHECKSUM_NOPARITY, 599 ZIO_CHECKSUM_SHA512, 600 ZIO_CHECKSUM_SKEIN, 601 ZIO_CHECKSUM_EDONR, 602 ZIO_CHECKSUM_BLAKE3, 603 ZIO_CHECKSUM_FUNCTIONS 604 }; 605 606 #define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4 607 #define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON 608 609 enum zio_compress { 610 ZIO_COMPRESS_INHERIT = 0, 611 ZIO_COMPRESS_ON, 612 ZIO_COMPRESS_OFF, 613 ZIO_COMPRESS_LZJB, 614 ZIO_COMPRESS_EMPTY, 615 ZIO_COMPRESS_GZIP_1, 616 ZIO_COMPRESS_GZIP_2, 617 ZIO_COMPRESS_GZIP_3, 618 ZIO_COMPRESS_GZIP_4, 619 ZIO_COMPRESS_GZIP_5, 620 ZIO_COMPRESS_GZIP_6, 621 ZIO_COMPRESS_GZIP_7, 622 ZIO_COMPRESS_GZIP_8, 623 ZIO_COMPRESS_GZIP_9, 624 ZIO_COMPRESS_ZLE, 625 ZIO_COMPRESS_LZ4, 626 ZIO_COMPRESS_ZSTD, 627 ZIO_COMPRESS_FUNCTIONS 628 }; 629 630 enum zio_zstd_levels { 631 ZIO_ZSTD_LEVEL_INHERIT = 0, 632 ZIO_ZSTD_LEVEL_1, 633 #define ZIO_ZSTD_LEVEL_MIN ZIO_ZSTD_LEVEL_1 634 ZIO_ZSTD_LEVEL_2, 635 ZIO_ZSTD_LEVEL_3, 636 #define ZIO_ZSTD_LEVEL_DEFAULT ZIO_ZSTD_LEVEL_3 637 ZIO_ZSTD_LEVEL_4, 638 ZIO_ZSTD_LEVEL_5, 639 ZIO_ZSTD_LEVEL_6, 640 ZIO_ZSTD_LEVEL_7, 641 ZIO_ZSTD_LEVEL_8, 642 ZIO_ZSTD_LEVEL_9, 643 ZIO_ZSTD_LEVEL_10, 644 ZIO_ZSTD_LEVEL_11, 645 ZIO_ZSTD_LEVEL_12, 646 ZIO_ZSTD_LEVEL_13, 647 ZIO_ZSTD_LEVEL_14, 648 ZIO_ZSTD_LEVEL_15, 649 ZIO_ZSTD_LEVEL_16, 650 ZIO_ZSTD_LEVEL_17, 651 ZIO_ZSTD_LEVEL_18, 652 ZIO_ZSTD_LEVEL_19, 653 #define ZIO_ZSTD_LEVEL_MAX ZIO_ZSTD_LEVEL_19 654 ZIO_ZSTD_LEVEL_RESERVE = 101, /* Leave room for new positive levels */ 655 ZIO_ZSTD_LEVEL_FAST, /* Fast levels are negative */ 656 ZIO_ZSTD_LEVEL_FAST_1, 657 #define ZIO_ZSTD_LEVEL_FAST_DEFAULT ZIO_ZSTD_LEVEL_FAST_1 658 ZIO_ZSTD_LEVEL_FAST_2, 659 ZIO_ZSTD_LEVEL_FAST_3, 660 ZIO_ZSTD_LEVEL_FAST_4, 661 ZIO_ZSTD_LEVEL_FAST_5, 662 ZIO_ZSTD_LEVEL_FAST_6, 663 ZIO_ZSTD_LEVEL_FAST_7, 664 ZIO_ZSTD_LEVEL_FAST_8, 665 ZIO_ZSTD_LEVEL_FAST_9, 666 ZIO_ZSTD_LEVEL_FAST_10, 667 ZIO_ZSTD_LEVEL_FAST_20, 668 ZIO_ZSTD_LEVEL_FAST_30, 669 ZIO_ZSTD_LEVEL_FAST_40, 670 ZIO_ZSTD_LEVEL_FAST_50, 671 ZIO_ZSTD_LEVEL_FAST_60, 672 ZIO_ZSTD_LEVEL_FAST_70, 673 ZIO_ZSTD_LEVEL_FAST_80, 674 ZIO_ZSTD_LEVEL_FAST_90, 675 ZIO_ZSTD_LEVEL_FAST_100, 676 ZIO_ZSTD_LEVEL_FAST_500, 677 ZIO_ZSTD_LEVEL_FAST_1000, 678 #define ZIO_ZSTD_LEVEL_FAST_MAX ZIO_ZSTD_LEVEL_FAST_1000 679 ZIO_ZSTD_LEVEL_AUTO = 251, /* Reserved for future use */ 680 ZIO_ZSTD_LEVEL_LEVELS 681 }; 682 683 #define ZIO_COMPRESS_ON_VALUE ZIO_COMPRESS_LZJB 684 #define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF 685 686 /* 687 * On-disk version number. 688 */ 689 #define SPA_VERSION_1 1ULL 690 #define SPA_VERSION_2 2ULL 691 #define SPA_VERSION_3 3ULL 692 #define SPA_VERSION_4 4ULL 693 #define SPA_VERSION_5 5ULL 694 #define SPA_VERSION_6 6ULL 695 #define SPA_VERSION_7 7ULL 696 #define SPA_VERSION_8 8ULL 697 #define SPA_VERSION_9 9ULL 698 #define SPA_VERSION_10 10ULL 699 #define SPA_VERSION_11 11ULL 700 #define SPA_VERSION_12 12ULL 701 #define SPA_VERSION_13 13ULL 702 #define SPA_VERSION_14 14ULL 703 #define SPA_VERSION_15 15ULL 704 #define SPA_VERSION_16 16ULL 705 #define SPA_VERSION_17 17ULL 706 #define SPA_VERSION_18 18ULL 707 #define SPA_VERSION_19 19ULL 708 #define SPA_VERSION_20 20ULL 709 #define SPA_VERSION_21 21ULL 710 #define SPA_VERSION_22 22ULL 711 #define SPA_VERSION_23 23ULL 712 #define SPA_VERSION_24 24ULL 713 #define SPA_VERSION_25 25ULL 714 #define SPA_VERSION_26 26ULL 715 #define SPA_VERSION_27 27ULL 716 #define SPA_VERSION_28 28ULL 717 #define SPA_VERSION_5000 5000ULL 718 719 /* 720 * When bumping up SPA_VERSION, make sure GRUB ZFS understands the on-disk 721 * format change. Go to usr/src/grub/grub-0.97/stage2/{zfs-include/, fsys_zfs*}, 722 * and do the appropriate changes. Also bump the version number in 723 * usr/src/grub/capability. 724 */ 725 #define SPA_VERSION SPA_VERSION_5000 726 #define SPA_VERSION_STRING "5000" 727 728 /* 729 * Symbolic names for the changes that caused a SPA_VERSION switch. 730 * Used in the code when checking for presence or absence of a feature. 731 * Feel free to define multiple symbolic names for each version if there 732 * were multiple changes to on-disk structures during that version. 733 * 734 * NOTE: When checking the current SPA_VERSION in your code, be sure 735 * to use spa_version() since it reports the version of the 736 * last synced uberblock. Checking the in-flight version can 737 * be dangerous in some cases. 738 */ 739 #define SPA_VERSION_INITIAL SPA_VERSION_1 740 #define SPA_VERSION_DITTO_BLOCKS SPA_VERSION_2 741 #define SPA_VERSION_SPARES SPA_VERSION_3 742 #define SPA_VERSION_RAID6 SPA_VERSION_3 743 #define SPA_VERSION_BPLIST_ACCOUNT SPA_VERSION_3 744 #define SPA_VERSION_RAIDZ_DEFLATE SPA_VERSION_3 745 #define SPA_VERSION_DNODE_BYTES SPA_VERSION_3 746 #define SPA_VERSION_ZPOOL_HISTORY SPA_VERSION_4 747 #define SPA_VERSION_GZIP_COMPRESSION SPA_VERSION_5 748 #define SPA_VERSION_BOOTFS SPA_VERSION_6 749 #define SPA_VERSION_SLOGS SPA_VERSION_7 750 #define SPA_VERSION_DELEGATED_PERMS SPA_VERSION_8 751 #define SPA_VERSION_FUID SPA_VERSION_9 752 #define SPA_VERSION_REFRESERVATION SPA_VERSION_9 753 #define SPA_VERSION_REFQUOTA SPA_VERSION_9 754 #define SPA_VERSION_UNIQUE_ACCURATE SPA_VERSION_9 755 #define SPA_VERSION_L2CACHE SPA_VERSION_10 756 #define SPA_VERSION_NEXT_CLONES SPA_VERSION_11 757 #define SPA_VERSION_ORIGIN SPA_VERSION_11 758 #define SPA_VERSION_DSL_SCRUB SPA_VERSION_11 759 #define SPA_VERSION_SNAP_PROPS SPA_VERSION_12 760 #define SPA_VERSION_USED_BREAKDOWN SPA_VERSION_13 761 #define SPA_VERSION_PASSTHROUGH_X SPA_VERSION_14 762 #define SPA_VERSION_USERSPACE SPA_VERSION_15 763 #define SPA_VERSION_STMF_PROP SPA_VERSION_16 764 #define SPA_VERSION_RAIDZ3 SPA_VERSION_17 765 #define SPA_VERSION_USERREFS SPA_VERSION_18 766 #define SPA_VERSION_HOLES SPA_VERSION_19 767 #define SPA_VERSION_ZLE_COMPRESSION SPA_VERSION_20 768 #define SPA_VERSION_DEDUP SPA_VERSION_21 769 #define SPA_VERSION_RECVD_PROPS SPA_VERSION_22 770 #define SPA_VERSION_SLIM_ZIL SPA_VERSION_23 771 #define SPA_VERSION_SA SPA_VERSION_24 772 #define SPA_VERSION_SCAN SPA_VERSION_25 773 #define SPA_VERSION_DIR_CLONES SPA_VERSION_26 774 #define SPA_VERSION_DEADLISTS SPA_VERSION_26 775 #define SPA_VERSION_FAST_SNAP SPA_VERSION_27 776 #define SPA_VERSION_MULTI_REPLACE SPA_VERSION_28 777 #define SPA_VERSION_BEFORE_FEATURES SPA_VERSION_28 778 #define SPA_VERSION_FEATURES SPA_VERSION_5000 779 780 #define SPA_VERSION_IS_SUPPORTED(v) \ 781 (((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \ 782 ((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION)) 783 784 /* 785 * The following are configuration names used in the nvlist describing a pool's 786 * configuration. 787 */ 788 #define ZPOOL_CONFIG_VERSION "version" 789 #define ZPOOL_CONFIG_POOL_NAME "name" 790 #define ZPOOL_CONFIG_POOL_STATE "state" 791 #define ZPOOL_CONFIG_POOL_TXG "txg" 792 #define ZPOOL_CONFIG_POOL_GUID "pool_guid" 793 #define ZPOOL_CONFIG_CREATE_TXG "create_txg" 794 #define ZPOOL_CONFIG_TOP_GUID "top_guid" 795 #define ZPOOL_CONFIG_VDEV_TREE "vdev_tree" 796 #define ZPOOL_CONFIG_TYPE "type" 797 #define ZPOOL_CONFIG_CHILDREN "children" 798 #define ZPOOL_CONFIG_ID "id" 799 #define ZPOOL_CONFIG_GUID "guid" 800 #define ZPOOL_CONFIG_INDIRECT_OBJECT "com.delphix:indirect_object" 801 #define ZPOOL_CONFIG_INDIRECT_BIRTHS "com.delphix:indirect_births" 802 #define ZPOOL_CONFIG_PREV_INDIRECT_VDEV "com.delphix:prev_indirect_vdev" 803 #define ZPOOL_CONFIG_PATH "path" 804 #define ZPOOL_CONFIG_DEVID "devid" 805 #define ZPOOL_CONFIG_METASLAB_ARRAY "metaslab_array" 806 #define ZPOOL_CONFIG_METASLAB_SHIFT "metaslab_shift" 807 #define ZPOOL_CONFIG_ASHIFT "ashift" 808 #define ZPOOL_CONFIG_ASIZE "asize" 809 #define ZPOOL_CONFIG_DTL "DTL" 810 #define ZPOOL_CONFIG_STATS "stats" 811 #define ZPOOL_CONFIG_WHOLE_DISK "whole_disk" 812 #define ZPOOL_CONFIG_ERRCOUNT "error_count" 813 #define ZPOOL_CONFIG_NOT_PRESENT "not_present" 814 #define ZPOOL_CONFIG_SPARES "spares" 815 #define ZPOOL_CONFIG_IS_SPARE "is_spare" 816 #define ZPOOL_CONFIG_NPARITY "nparity" 817 #define ZPOOL_CONFIG_HOSTID "hostid" 818 #define ZPOOL_CONFIG_HOSTNAME "hostname" 819 #define ZPOOL_CONFIG_IS_LOG "is_log" 820 #define ZPOOL_CONFIG_TIMESTAMP "timestamp" /* not stored on disk */ 821 #define ZPOOL_CONFIG_FEATURES_FOR_READ "features_for_read" 822 #define ZPOOL_CONFIG_VDEV_CHILDREN "vdev_children" 823 824 /* 825 * The persistent vdev state is stored as separate values rather than a single 826 * 'vdev_state' entry. This is because a device can be in multiple states, such 827 * as offline and degraded. 828 */ 829 #define ZPOOL_CONFIG_OFFLINE "offline" 830 #define ZPOOL_CONFIG_FAULTED "faulted" 831 #define ZPOOL_CONFIG_DEGRADED "degraded" 832 #define ZPOOL_CONFIG_REMOVED "removed" 833 #define ZPOOL_CONFIG_FRU "fru" 834 #define ZPOOL_CONFIG_AUX_STATE "aux_state" 835 836 #define VDEV_TYPE_ROOT "root" 837 #define VDEV_TYPE_MIRROR "mirror" 838 #define VDEV_TYPE_REPLACING "replacing" 839 #define VDEV_TYPE_RAIDZ "raidz" 840 #define VDEV_TYPE_DISK "disk" 841 #define VDEV_TYPE_FILE "file" 842 #define VDEV_TYPE_MISSING "missing" 843 #define VDEV_TYPE_HOLE "hole" 844 #define VDEV_TYPE_SPARE "spare" 845 #define VDEV_TYPE_LOG "log" 846 #define VDEV_TYPE_L2CACHE "l2cache" 847 #define VDEV_TYPE_INDIRECT "indirect" 848 849 /* 850 * This is needed in userland to report the minimum necessary device size. 851 */ 852 #define SPA_MINDEVSIZE (64ULL << 20) 853 854 /* 855 * The location of the pool configuration repository, shared between kernel and 856 * userland. 857 */ 858 #define ZPOOL_CACHE "/boot/zfs/zpool.cache" 859 860 /* 861 * vdev states are ordered from least to most healthy. 862 * A vdev that's CANT_OPEN or below is considered unusable. 863 */ 864 typedef enum vdev_state { 865 VDEV_STATE_UNKNOWN = 0, /* Uninitialized vdev */ 866 VDEV_STATE_CLOSED, /* Not currently open */ 867 VDEV_STATE_OFFLINE, /* Not allowed to open */ 868 VDEV_STATE_REMOVED, /* Explicitly removed from system */ 869 VDEV_STATE_CANT_OPEN, /* Tried to open, but failed */ 870 VDEV_STATE_FAULTED, /* External request to fault device */ 871 VDEV_STATE_DEGRADED, /* Replicated vdev with unhealthy kids */ 872 VDEV_STATE_HEALTHY /* Presumed good */ 873 } vdev_state_t; 874 875 /* 876 * vdev aux states. When a vdev is in the CANT_OPEN state, the aux field 877 * of the vdev stats structure uses these constants to distinguish why. 878 */ 879 typedef enum vdev_aux { 880 VDEV_AUX_NONE, /* no error */ 881 VDEV_AUX_OPEN_FAILED, /* ldi_open_*() or vn_open() failed */ 882 VDEV_AUX_CORRUPT_DATA, /* bad label or disk contents */ 883 VDEV_AUX_NO_REPLICAS, /* insufficient number of replicas */ 884 VDEV_AUX_BAD_GUID_SUM, /* vdev guid sum doesn't match */ 885 VDEV_AUX_TOO_SMALL, /* vdev size is too small */ 886 VDEV_AUX_BAD_LABEL, /* the label is OK but invalid */ 887 VDEV_AUX_VERSION_NEWER, /* on-disk version is too new */ 888 VDEV_AUX_VERSION_OLDER, /* on-disk version is too old */ 889 VDEV_AUX_SPARED /* hot spare used in another pool */ 890 } vdev_aux_t; 891 892 /* 893 * pool state. The following states are written to disk as part of the normal 894 * SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE. The remaining states are 895 * software abstractions used at various levels to communicate pool state. 896 */ 897 typedef enum pool_state { 898 POOL_STATE_ACTIVE = 0, /* In active use */ 899 POOL_STATE_EXPORTED, /* Explicitly exported */ 900 POOL_STATE_DESTROYED, /* Explicitly destroyed */ 901 POOL_STATE_SPARE, /* Reserved for hot spare use */ 902 POOL_STATE_UNINITIALIZED, /* Internal spa_t state */ 903 POOL_STATE_UNAVAIL, /* Internal libzfs state */ 904 POOL_STATE_POTENTIALLY_ACTIVE /* Internal libzfs state */ 905 } pool_state_t; 906 907 /* 908 * The uberblock version is incremented whenever an incompatible on-disk 909 * format change is made to the SPA, DMU, or ZAP. 910 * 911 * Note: the first two fields should never be moved. When a storage pool 912 * is opened, the uberblock must be read off the disk before the version 913 * can be checked. If the ub_version field is moved, we may not detect 914 * version mismatch. If the ub_magic field is moved, applications that 915 * expect the magic number in the first word won't work. 916 */ 917 #define UBERBLOCK_MAGIC 0x00bab10c /* oo-ba-bloc! */ 918 #define UBERBLOCK_SHIFT 10 /* up to 1K */ 919 920 #define MMP_MAGIC 0xa11cea11 /* all-see-all */ 921 922 #define MMP_INTERVAL_VALID_BIT 0x01 923 #define MMP_SEQ_VALID_BIT 0x02 924 #define MMP_FAIL_INT_VALID_BIT 0x04 925 926 #define MMP_VALID(ubp) (ubp->ub_magic == UBERBLOCK_MAGIC && \ 927 ubp->ub_mmp_magic == MMP_MAGIC) 928 #define MMP_INTERVAL_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \ 929 MMP_INTERVAL_VALID_BIT)) 930 #define MMP_SEQ_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \ 931 MMP_SEQ_VALID_BIT)) 932 #define MMP_FAIL_INT_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \ 933 MMP_FAIL_INT_VALID_BIT)) 934 935 #define MMP_INTERVAL(ubp) ((ubp->ub_mmp_config & 0x00000000FFFFFF00) \ 936 >> 8) 937 #define MMP_SEQ(ubp) ((ubp->ub_mmp_config & 0x0000FFFF00000000) \ 938 >> 32) 939 #define MMP_FAIL_INT(ubp) ((ubp->ub_mmp_config & 0xFFFF000000000000) \ 940 >> 48) 941 942 typedef struct uberblock { 943 uint64_t ub_magic; /* UBERBLOCK_MAGIC */ 944 uint64_t ub_version; /* SPA_VERSION */ 945 uint64_t ub_txg; /* txg of last sync */ 946 uint64_t ub_guid_sum; /* sum of all vdev guids */ 947 uint64_t ub_timestamp; /* UTC time of last sync */ 948 blkptr_t ub_rootbp; /* MOS objset_phys_t */ 949 /* highest SPA_VERSION supported by software that wrote this txg */ 950 uint64_t ub_software_version; 951 /* Maybe missing in uberblocks we read, but always written */ 952 uint64_t ub_mmp_magic; 953 /* 954 * If ub_mmp_delay == 0 and ub_mmp_magic is valid, MMP is off. 955 * Otherwise, nanosec since last MMP write. 956 */ 957 uint64_t ub_mmp_delay; 958 959 /* 960 * The ub_mmp_config contains the multihost write interval, multihost 961 * fail intervals, sequence number for sub-second granularity, and 962 * valid bit mask. This layout is as follows: 963 * 964 * 64 56 48 40 32 24 16 8 0 965 * +-------+-------+-------+-------+-------+-------+-------+-------+ 966 * 0 | Fail Intervals| Seq | Write Interval (ms) | VALID | 967 * +-------+-------+-------+-------+-------+-------+-------+-------+ 968 * 969 * This allows a write_interval of (2^24/1000)s, over 4.5 hours 970 * 971 * VALID Bits: 972 * - 0x01 - Write Interval (ms) 973 * - 0x02 - Sequence number exists 974 * - 0x04 - Fail Intervals 975 * - 0xf8 - Reserved 976 */ 977 uint64_t ub_mmp_config; 978 979 /* 980 * ub_checkpoint_txg indicates two things about the current uberblock: 981 * 982 * 1] If it is not zero then this uberblock is a checkpoint. If it is 983 * zero, then this uberblock is not a checkpoint. 984 * 985 * 2] On checkpointed uberblocks, the value of ub_checkpoint_txg is 986 * the ub_txg that the uberblock had at the time we moved it to 987 * the MOS config. 988 * 989 * The field is set when we checkpoint the uberblock and continues to 990 * hold that value even after we've rewound (unlike the ub_txg that 991 * is reset to a higher value). 992 * 993 * Besides checks used to determine whether we are reopening the 994 * pool from a checkpointed uberblock [see spa_ld_select_uberblock()], 995 * the value of the field is used to determine which ZIL blocks have 996 * been allocated according to the ms_sm when we are rewinding to a 997 * checkpoint. Specifically, if blk_birth > ub_checkpoint_txg, then 998 * the ZIL block is not allocated [see uses of spa_min_claim_txg()]. 999 */ 1000 uint64_t ub_checkpoint_txg; 1001 } uberblock_t; 1002 1003 /* 1004 * Flags. 1005 */ 1006 #define DNODE_MUST_BE_ALLOCATED 1 1007 #define DNODE_MUST_BE_FREE 2 1008 1009 /* 1010 * Fixed constants. 1011 */ 1012 #define DNODE_SHIFT 9 /* 512 bytes */ 1013 #define DN_MIN_INDBLKSHIFT 12 /* 4k */ 1014 #define DN_MAX_INDBLKSHIFT 17 /* 128k */ 1015 #define DNODE_BLOCK_SHIFT 14 /* 16k */ 1016 #define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */ 1017 #define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */ 1018 #define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */ 1019 1020 /* 1021 * Derived constants. 1022 */ 1023 #define DNODE_MIN_SIZE (1 << DNODE_SHIFT) 1024 #define DNODE_MAX_SIZE (1 << DNODE_BLOCK_SHIFT) 1025 #define DNODE_BLOCK_SIZE (1 << DNODE_BLOCK_SHIFT) 1026 #define DNODE_MIN_SLOTS (DNODE_MIN_SIZE >> DNODE_SHIFT) 1027 #define DNODE_MAX_SLOTS (DNODE_MAX_SIZE >> DNODE_SHIFT) 1028 #define DN_BONUS_SIZE(dnsize) ((dnsize) - DNODE_CORE_SIZE - \ 1029 (1 << SPA_BLKPTRSHIFT)) 1030 #define DN_SLOTS_TO_BONUSLEN(slots) DN_BONUS_SIZE((slots) << DNODE_SHIFT) 1031 #define DN_OLD_MAX_BONUSLEN (DN_BONUS_SIZE(DNODE_MIN_SIZE)) 1032 #define DN_MAX_NBLKPTR ((DNODE_MIN_SIZE - DNODE_CORE_SIZE) >> \ 1033 SPA_BLKPTRSHIFT) 1034 #define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT) 1035 #define DN_ZERO_BONUSLEN (DN_BONUS_SIZE(DNODE_MAX_SIZE) + 1) 1036 1037 #define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT) 1038 #define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT) 1039 #define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT) 1040 1041 /* The +2 here is a cheesy way to round up */ 1042 #define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \ 1043 (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT))) 1044 1045 #define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \ 1046 (((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t)))) 1047 1048 #define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \ 1049 (dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT) 1050 1051 #define EPB(blkshift, typeshift) (1 << (blkshift - typeshift)) 1052 1053 /* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */ 1054 #define DNODE_FLAG_USED_BYTES (1<<0) 1055 #define DNODE_FLAG_USERUSED_ACCOUNTED (1<<1) 1056 1057 /* Does dnode have a SA spill blkptr in bonus? */ 1058 #define DNODE_FLAG_SPILL_BLKPTR (1<<2) 1059 1060 typedef struct dnode_phys { 1061 uint8_t dn_type; /* dmu_object_type_t */ 1062 uint8_t dn_indblkshift; /* ln2(indirect block size) */ 1063 uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */ 1064 uint8_t dn_nblkptr; /* length of dn_blkptr */ 1065 uint8_t dn_bonustype; /* type of data in bonus buffer */ 1066 uint8_t dn_checksum; /* ZIO_CHECKSUM type */ 1067 uint8_t dn_compress; /* ZIO_COMPRESS type */ 1068 uint8_t dn_flags; /* DNODE_FLAG_* */ 1069 uint16_t dn_datablkszsec; /* data block size in 512b sectors */ 1070 uint16_t dn_bonuslen; /* length of dn_bonus */ 1071 uint8_t dn_extra_slots; /* # of subsequent slots consumed */ 1072 uint8_t dn_pad2[3]; 1073 1074 /* accounting is protected by dn_dirty_mtx */ 1075 uint64_t dn_maxblkid; /* largest allocated block ID */ 1076 uint64_t dn_used; /* bytes (or sectors) of disk space */ 1077 1078 uint64_t dn_pad3[4]; 1079 1080 /* 1081 * The tail region is 448 bytes for a 512 byte dnode, and 1082 * correspondingly larger for larger dnode sizes. The spill 1083 * block pointer, when present, is always at the end of the tail 1084 * region. There are three ways this space may be used, using 1085 * a 512 byte dnode for this diagram: 1086 * 1087 * 0 64 128 192 256 320 384 448 (offset) 1088 * +---------------+---------------+---------------+-------+ 1089 * | dn_blkptr[0] | dn_blkptr[1] | dn_blkptr[2] | / | 1090 * +---------------+---------------+---------------+-------+ 1091 * | dn_blkptr[0] | dn_bonus[0..319] | 1092 * +---------------+-----------------------+---------------+ 1093 * | dn_blkptr[0] | dn_bonus[0..191] | dn_spill | 1094 * +---------------+-----------------------+---------------+ 1095 */ 1096 union { 1097 blkptr_t dn_blkptr[1+DN_OLD_MAX_BONUSLEN/sizeof (blkptr_t)]; 1098 struct { 1099 blkptr_t __dn_ignore1; 1100 uint8_t dn_bonus[DN_OLD_MAX_BONUSLEN]; 1101 }; 1102 struct { 1103 blkptr_t __dn_ignore2; 1104 uint8_t __dn_ignore3[DN_OLD_MAX_BONUSLEN - 1105 sizeof (blkptr_t)]; 1106 blkptr_t dn_spill; 1107 }; 1108 }; 1109 } dnode_phys_t; 1110 1111 #define DN_SPILL_BLKPTR(dnp) (blkptr_t *)((char *)(dnp) + \ 1112 (((dnp)->dn_extra_slots + 1) << DNODE_SHIFT) - (1 << SPA_BLKPTRSHIFT)) 1113 1114 typedef enum dmu_object_byteswap { 1115 DMU_BSWAP_UINT8, 1116 DMU_BSWAP_UINT16, 1117 DMU_BSWAP_UINT32, 1118 DMU_BSWAP_UINT64, 1119 DMU_BSWAP_ZAP, 1120 DMU_BSWAP_DNODE, 1121 DMU_BSWAP_OBJSET, 1122 DMU_BSWAP_ZNODE, 1123 DMU_BSWAP_OLDACL, 1124 DMU_BSWAP_ACL, 1125 /* 1126 * Allocating a new byteswap type number makes the on-disk format 1127 * incompatible with any other format that uses the same number. 1128 * 1129 * Data can usually be structured to work with one of the 1130 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. 1131 */ 1132 DMU_BSWAP_NUMFUNCS 1133 } dmu_object_byteswap_t; 1134 1135 #define DMU_OT_NEWTYPE 0x80 1136 #define DMU_OT_METADATA 0x40 1137 #define DMU_OT_BYTESWAP_MASK 0x3f 1138 1139 /* 1140 * Defines a uint8_t object type. Object types specify if the data 1141 * in the object is metadata (boolean) and how to byteswap the data 1142 * (dmu_object_byteswap_t). 1143 */ 1144 #define DMU_OT(byteswap, metadata) \ 1145 (DMU_OT_NEWTYPE | \ 1146 ((metadata) ? DMU_OT_METADATA : 0) | \ 1147 ((byteswap) & DMU_OT_BYTESWAP_MASK)) 1148 1149 typedef enum dmu_object_type { 1150 DMU_OT_NONE, 1151 /* general: */ 1152 DMU_OT_OBJECT_DIRECTORY, /* ZAP */ 1153 DMU_OT_OBJECT_ARRAY, /* UINT64 */ 1154 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ 1155 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ 1156 DMU_OT_BPOBJ, /* UINT64 */ 1157 DMU_OT_BPOBJ_HDR, /* UINT64 */ 1158 /* spa: */ 1159 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ 1160 DMU_OT_SPACE_MAP, /* UINT64 */ 1161 /* zil: */ 1162 DMU_OT_INTENT_LOG, /* UINT64 */ 1163 /* dmu: */ 1164 DMU_OT_DNODE, /* DNODE */ 1165 DMU_OT_OBJSET, /* OBJSET */ 1166 /* dsl: */ 1167 DMU_OT_DSL_DIR, /* UINT64 */ 1168 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ 1169 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ 1170 DMU_OT_DSL_PROPS, /* ZAP */ 1171 DMU_OT_DSL_DATASET, /* UINT64 */ 1172 /* zpl: */ 1173 DMU_OT_ZNODE, /* ZNODE */ 1174 DMU_OT_OLDACL, /* Old ACL */ 1175 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ 1176 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ 1177 DMU_OT_MASTER_NODE, /* ZAP */ 1178 DMU_OT_UNLINKED_SET, /* ZAP */ 1179 /* zvol: */ 1180 DMU_OT_ZVOL, /* UINT8 */ 1181 DMU_OT_ZVOL_PROP, /* ZAP */ 1182 /* other; for testing only! */ 1183 DMU_OT_PLAIN_OTHER, /* UINT8 */ 1184 DMU_OT_UINT64_OTHER, /* UINT64 */ 1185 DMU_OT_ZAP_OTHER, /* ZAP */ 1186 /* new object types: */ 1187 DMU_OT_ERROR_LOG, /* ZAP */ 1188 DMU_OT_SPA_HISTORY, /* UINT8 */ 1189 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ 1190 DMU_OT_POOL_PROPS, /* ZAP */ 1191 DMU_OT_DSL_PERMS, /* ZAP */ 1192 DMU_OT_ACL, /* ACL */ 1193 DMU_OT_SYSACL, /* SYSACL */ 1194 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ 1195 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ 1196 DMU_OT_NEXT_CLONES, /* ZAP */ 1197 DMU_OT_SCAN_QUEUE, /* ZAP */ 1198 DMU_OT_USERGROUP_USED, /* ZAP */ 1199 DMU_OT_USERGROUP_QUOTA, /* ZAP */ 1200 DMU_OT_USERREFS, /* ZAP */ 1201 DMU_OT_DDT_ZAP, /* ZAP */ 1202 DMU_OT_DDT_STATS, /* ZAP */ 1203 DMU_OT_SA, /* System attr */ 1204 DMU_OT_SA_MASTER_NODE, /* ZAP */ 1205 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ 1206 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ 1207 DMU_OT_SCAN_XLATE, /* ZAP */ 1208 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ 1209 DMU_OT_DEADLIST, /* ZAP */ 1210 DMU_OT_DEADLIST_HDR, /* UINT64 */ 1211 DMU_OT_DSL_CLONES, /* ZAP */ 1212 DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */ 1213 DMU_OT_NUMTYPES, 1214 1215 /* 1216 * Names for valid types declared with DMU_OT(). 1217 */ 1218 DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE), 1219 DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE), 1220 DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE), 1221 DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE), 1222 DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE), 1223 DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE), 1224 DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE), 1225 DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE), 1226 DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE), 1227 DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE) 1228 } dmu_object_type_t; 1229 1230 typedef enum dmu_objset_type { 1231 DMU_OST_NONE, 1232 DMU_OST_META, 1233 DMU_OST_ZFS, 1234 DMU_OST_ZVOL, 1235 DMU_OST_OTHER, /* For testing only! */ 1236 DMU_OST_ANY, /* Be careful! */ 1237 DMU_OST_NUMTYPES 1238 } dmu_objset_type_t; 1239 1240 #define ZAP_MAXVALUELEN (1024 * 8) 1241 1242 /* 1243 * header for all bonus and spill buffers. 1244 * The header has a fixed portion with a variable number 1245 * of "lengths" depending on the number of variable sized 1246 * attribues which are determined by the "layout number" 1247 */ 1248 1249 #define SA_MAGIC 0x2F505A /* ZFS SA */ 1250 typedef struct sa_hdr_phys { 1251 uint32_t sa_magic; 1252 uint16_t sa_layout_info; /* Encoded with hdrsize and layout number */ 1253 uint16_t sa_lengths[1]; /* optional sizes for variable length attrs */ 1254 /* ... Data follows the lengths. */ 1255 } sa_hdr_phys_t; 1256 1257 /* 1258 * sa_hdr_phys -> sa_layout_info 1259 * 1260 * 16 10 0 1261 * +--------+-------+ 1262 * | hdrsz |layout | 1263 * +--------+-------+ 1264 * 1265 * Bits 0-10 are the layout number 1266 * Bits 11-16 are the size of the header. 1267 * The hdrsize is the number * 8 1268 * 1269 * For example. 1270 * hdrsz of 1 ==> 8 byte header 1271 * 2 ==> 16 byte header 1272 * 1273 */ 1274 1275 #define SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10) 1276 #define SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 16, 3, 0) 1277 #define SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \ 1278 { \ 1279 BF32_SET_SB(x, 10, 6, 3, 0, size); \ 1280 BF32_SET(x, 0, 10, num); \ 1281 } 1282 1283 #define SA_ATTR_BSWAP(x) BF32_GET(x, 16, 8) 1284 #define SA_ATTR_LENGTH(x) BF32_GET(x, 24, 16) 1285 #define SA_ATTR_NUM(x) BF32_GET(x, 0, 16) 1286 #define SA_ATTR_ENCODE(x, attr, length, bswap) \ 1287 { \ 1288 BF64_SET(x, 24, 16, length); \ 1289 BF64_SET(x, 16, 8, bswap); \ 1290 BF64_SET(x, 0, 16, attr); \ 1291 } 1292 1293 #define SA_MODE_OFFSET 0 1294 #define SA_SIZE_OFFSET 8 1295 #define SA_GEN_OFFSET 16 1296 #define SA_UID_OFFSET 24 1297 #define SA_GID_OFFSET 32 1298 #define SA_PARENT_OFFSET 40 1299 #define SA_SYMLINK_OFFSET 160 1300 1301 #define SA_REGISTRY "REGISTRY" 1302 #define SA_LAYOUTS "LAYOUTS" 1303 1304 typedef enum sa_bswap_type { 1305 SA_UINT64_ARRAY, 1306 SA_UINT32_ARRAY, 1307 SA_UINT16_ARRAY, 1308 SA_UINT8_ARRAY, 1309 SA_ACL, 1310 } sa_bswap_type_t; 1311 1312 typedef uint16_t sa_attr_type_t; 1313 1314 #define ZIO_OBJSET_MAC_LEN 32 1315 1316 /* 1317 * Intent log header - this on disk structure holds fields to manage 1318 * the log. All fields are 64 bit to easily handle cross architectures. 1319 */ 1320 typedef struct zil_header { 1321 uint64_t zh_claim_txg; /* txg in which log blocks were claimed */ 1322 uint64_t zh_replay_seq; /* highest replayed sequence number */ 1323 blkptr_t zh_log; /* log chain */ 1324 uint64_t zh_claim_seq; /* highest claimed sequence number */ 1325 uint64_t zh_pad[5]; 1326 } zil_header_t; 1327 1328 #define OBJSET_PHYS_SIZE_V2 2048 1329 #define OBJSET_PHYS_SIZE_V3 4096 1330 1331 typedef struct objset_phys { 1332 dnode_phys_t os_meta_dnode; 1333 zil_header_t os_zil_header; 1334 uint64_t os_type; 1335 uint64_t os_flags; 1336 uint8_t os_portable_mac[ZIO_OBJSET_MAC_LEN]; 1337 uint8_t os_local_mac[ZIO_OBJSET_MAC_LEN]; 1338 char os_pad0[OBJSET_PHYS_SIZE_V2 - sizeof (dnode_phys_t)*3 - 1339 sizeof (zil_header_t) - sizeof (uint64_t)*2 - 1340 2*ZIO_OBJSET_MAC_LEN]; 1341 dnode_phys_t os_userused_dnode; 1342 dnode_phys_t os_groupused_dnode; 1343 dnode_phys_t os_projectused_dnode; 1344 char os_pad1[OBJSET_PHYS_SIZE_V3 - OBJSET_PHYS_SIZE_V2 - 1345 sizeof (dnode_phys_t)]; 1346 } objset_phys_t; 1347 1348 typedef struct space_map_phys { 1349 /* object number: not needed but kept for backwards compatibility */ 1350 uint64_t smp_object; 1351 1352 /* length of the object in bytes */ 1353 uint64_t smp_length; 1354 1355 /* space allocated from the map */ 1356 int64_t smp_alloc; 1357 } space_map_phys_t; 1358 1359 typedef enum { 1360 SM_ALLOC, 1361 SM_FREE 1362 } maptype_t; 1363 1364 /* one-word entry constants */ 1365 #define SM_DEBUG_PREFIX 2 1366 #define SM_OFFSET_BITS 47 1367 #define SM_RUN_BITS 15 1368 1369 /* two-word entry constants */ 1370 #define SM2_PREFIX 3 1371 #define SM2_OFFSET_BITS 63 1372 #define SM2_RUN_BITS 36 1373 1374 #define SM_PREFIX_DECODE(x) BF64_DECODE(x, 62, 2) 1375 #define SM_PREFIX_ENCODE(x) BF64_ENCODE(x, 62, 2) 1376 1377 #define SM_DEBUG_ACTION_DECODE(x) BF64_DECODE(x, 60, 2) 1378 #define SM_DEBUG_ACTION_ENCODE(x) BF64_ENCODE(x, 60, 2) 1379 #define SM_DEBUG_SYNCPASS_DECODE(x) BF64_DECODE(x, 50, 10) 1380 #define SM_DEBUG_SYNCPASS_ENCODE(x) BF64_ENCODE(x, 50, 10) 1381 #define SM_DEBUG_TXG_DECODE(x) BF64_DECODE(x, 0, 50) 1382 #define SM_DEBUG_TXG_ENCODE(x) BF64_ENCODE(x, 0, 50) 1383 1384 #define SM_OFFSET_DECODE(x) BF64_DECODE(x, 16, SM_OFFSET_BITS) 1385 #define SM_OFFSET_ENCODE(x) BF64_ENCODE(x, 16, SM_OFFSET_BITS) 1386 #define SM_TYPE_DECODE(x) BF64_DECODE(x, 15, 1) 1387 #define SM_TYPE_ENCODE(x) BF64_ENCODE(x, 15, 1) 1388 #define SM_RUN_DECODE(x) (BF64_DECODE(x, 0, SM_RUN_BITS) + 1) 1389 #define SM_RUN_ENCODE(x) BF64_ENCODE((x) - 1, 0, SM_RUN_BITS) 1390 #define SM_RUN_MAX SM_RUN_DECODE(~0ULL) 1391 #define SM_OFFSET_MAX SM_OFFSET_DECODE(~0ULL) 1392 1393 #define SM2_RUN_DECODE(x) (BF64_DECODE(x, 24, SM2_RUN_BITS) + 1) 1394 #define SM2_RUN_ENCODE(x) BF64_ENCODE((x) - 1, 24, SM2_RUN_BITS) 1395 #define SM2_VDEV_DECODE(x) BF64_DECODE(x, 0, 24) 1396 #define SM2_VDEV_ENCODE(x) BF64_ENCODE(x, 0, 24) 1397 #define SM2_TYPE_DECODE(x) BF64_DECODE(x, SM2_OFFSET_BITS, 1) 1398 #define SM2_TYPE_ENCODE(x) BF64_ENCODE(x, SM2_OFFSET_BITS, 1) 1399 #define SM2_OFFSET_DECODE(x) BF64_DECODE(x, 0, SM2_OFFSET_BITS) 1400 #define SM2_OFFSET_ENCODE(x) BF64_ENCODE(x, 0, SM2_OFFSET_BITS) 1401 #define SM2_RUN_MAX SM2_RUN_DECODE(~0ULL) 1402 #define SM2_OFFSET_MAX SM2_OFFSET_DECODE(~0ULL) 1403 1404 typedef enum dd_used { 1405 DD_USED_HEAD, 1406 DD_USED_SNAP, 1407 DD_USED_CHILD, 1408 DD_USED_CHILD_RSRV, 1409 DD_USED_REFRSRV, 1410 DD_USED_NUM 1411 } dd_used_t; 1412 1413 #define DD_FLAG_USED_BREAKDOWN (1 << 0) 1414 1415 typedef struct dsl_dir_phys { 1416 uint64_t dd_creation_time; /* not actually used */ 1417 uint64_t dd_head_dataset_obj; 1418 uint64_t dd_parent_obj; 1419 uint64_t dd_clone_parent_obj; 1420 uint64_t dd_child_dir_zapobj; 1421 /* 1422 * how much space our children are accounting for; for leaf 1423 * datasets, == physical space used by fs + snaps 1424 */ 1425 uint64_t dd_used_bytes; 1426 uint64_t dd_compressed_bytes; 1427 uint64_t dd_uncompressed_bytes; 1428 /* Administrative quota setting */ 1429 uint64_t dd_quota; 1430 /* Administrative reservation setting */ 1431 uint64_t dd_reserved; 1432 uint64_t dd_props_zapobj; 1433 uint64_t dd_pad[1]; 1434 uint64_t dd_flags; 1435 uint64_t dd_used_breakdown[DD_USED_NUM]; 1436 uint64_t dd_clones; 1437 uint64_t dd_pad1[13]; /* pad out to 256 bytes for good measure */ 1438 } dsl_dir_phys_t; 1439 1440 typedef struct dsl_dataset_phys { 1441 uint64_t ds_dir_obj; 1442 uint64_t ds_prev_snap_obj; 1443 uint64_t ds_prev_snap_txg; 1444 uint64_t ds_next_snap_obj; 1445 uint64_t ds_snapnames_zapobj; /* zap obj of snaps; ==0 for snaps */ 1446 uint64_t ds_num_children; /* clone/snap children; ==0 for head */ 1447 uint64_t ds_creation_time; /* seconds since 1970 */ 1448 uint64_t ds_creation_txg; 1449 uint64_t ds_deadlist_obj; 1450 uint64_t ds_used_bytes; 1451 uint64_t ds_compressed_bytes; 1452 uint64_t ds_uncompressed_bytes; 1453 uint64_t ds_unique_bytes; /* only relevant to snapshots */ 1454 /* 1455 * The ds_fsid_guid is a 56-bit ID that can change to avoid 1456 * collisions. The ds_guid is a 64-bit ID that will never 1457 * change, so there is a small probability that it will collide. 1458 */ 1459 uint64_t ds_fsid_guid; 1460 uint64_t ds_guid; 1461 uint64_t ds_flags; 1462 blkptr_t ds_bp; 1463 uint64_t ds_next_clones_obj; /* DMU_OT_DSL_CLONES */ 1464 uint64_t ds_props_obj; /* DMU_OT_DSL_PROPS for snaps */ 1465 uint64_t ds_userrefs_obj; /* DMU_OT_USERREFS */ 1466 uint64_t ds_pad[5]; /* pad out to 320 bytes for good measure */ 1467 } dsl_dataset_phys_t; 1468 1469 typedef struct dsl_deadlist_phys { 1470 uint64_t dl_used; 1471 uint64_t dl_comp; 1472 uint64_t dl_uncomp; 1473 uint64_t dl_pad[37]; /* pad out to 320b for future expansion */ 1474 } dsl_deadlist_phys_t; 1475 1476 #define BPOBJ_SIZE_V2 (6 * sizeof (uint64_t)) 1477 1478 typedef struct bpobj_phys { 1479 uint64_t bpo_num_blkptrs; 1480 uint64_t bpo_bytes; 1481 uint64_t bpo_comp; 1482 uint64_t bpo_uncomp; 1483 uint64_t bpo_subobjs; 1484 uint64_t bpo_num_subobjs; 1485 uint64_t bpo_num_freed; 1486 } bpobj_phys_t; 1487 1488 /* 1489 * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 1490 */ 1491 #define DMU_POOL_DIRECTORY_OBJECT 1 1492 #define DMU_POOL_CONFIG "config" 1493 #define DMU_POOL_FEATURES_FOR_READ "features_for_read" 1494 #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write" 1495 #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions" 1496 #define DMU_POOL_ROOT_DATASET "root_dataset" 1497 #define DMU_POOL_SYNC_BPLIST "sync_bplist" 1498 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 1499 #define DMU_POOL_ERRLOG_LAST "errlog_last" 1500 #define DMU_POOL_SPARES "spares" 1501 #define DMU_POOL_DEFLATE "deflate" 1502 #define DMU_POOL_HISTORY "history" 1503 #define DMU_POOL_PROPS "pool_props" 1504 #define DMU_POOL_FREE_BPOBJ "free_bpobj" 1505 #define DMU_POOL_BPTREE_OBJ "bptree_obj" 1506 #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj" 1507 #define DMU_POOL_TMP_USERREFS "tmp_userrefs" 1508 #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" 1509 #define DMU_POOL_REMOVING "com.delphix:removing" 1510 #define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj" 1511 #define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect" 1512 #define DMU_POOL_ZPOOL_CHECKPOINT "com.delphix:zpool_checkpoint" 1513 1514 #define ZAP_MAGIC 0x2F52AB2ABULL 1515 1516 #define FZAP_BLOCK_SHIFT(zap) ((zap)->zap_block_shift) 1517 1518 #define ZAP_MAXCD (uint32_t)(-1) 1519 #define ZAP_HASHBITS 28 1520 #define MZAP_ENT_LEN 64 1521 #define MZAP_ENT_MAX \ 1522 ((MZAP_MAX_BLKSZ - sizeof(mzap_phys_t)) / sizeof(mzap_ent_phys_t) + 1) 1523 #define MZAP_NAME_LEN (MZAP_ENT_LEN - 8 - 4 - 2) 1524 #define MZAP_MAX_BLKSZ SPA_OLDMAXBLOCKSIZE 1525 1526 typedef struct mzap_ent_phys { 1527 uint64_t mze_value; 1528 uint32_t mze_cd; 1529 uint16_t mze_pad; /* in case we want to chain them someday */ 1530 char mze_name[MZAP_NAME_LEN]; 1531 } mzap_ent_phys_t; 1532 1533 typedef struct mzap_phys { 1534 uint64_t mz_block_type; /* ZBT_MICRO */ 1535 uint64_t mz_salt; 1536 uint64_t mz_normflags; 1537 uint64_t mz_pad[5]; 1538 mzap_ent_phys_t mz_chunk[1]; 1539 /* actually variable size depending on block size */ 1540 } mzap_phys_t; 1541 1542 /* 1543 * The (fat) zap is stored in one object. It is an array of 1544 * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of: 1545 * 1546 * ptrtbl fits in first block: 1547 * [zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ... 1548 * 1549 * ptrtbl too big for first block: 1550 * [zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ... 1551 * 1552 */ 1553 1554 #define ZBT_LEAF ((1ULL << 63) + 0) 1555 #define ZBT_HEADER ((1ULL << 63) + 1) 1556 #define ZBT_MICRO ((1ULL << 63) + 3) 1557 /* any other values are ptrtbl blocks */ 1558 1559 /* 1560 * the embedded pointer table takes up half a block: 1561 * block size / entry size (2^3) / 2 1562 */ 1563 #define ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1) 1564 1565 /* 1566 * The embedded pointer table starts half-way through the block. Since 1567 * the pointer table itself is half the block, it starts at (64-bit) 1568 * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)). 1569 */ 1570 #define ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \ 1571 ((uint64_t *)(zap)->zap_phys) \ 1572 [(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))] 1573 1574 #define ZAP_HASH_IDX(hash, n) (((n) == 0) ? 0 : ((hash) >> (64 - (n)))) 1575 1576 /* 1577 * TAKE NOTE: 1578 * If zap_phys_t is modified, zap_byteswap() must be modified. 1579 */ 1580 typedef struct zap_phys { 1581 uint64_t zap_block_type; /* ZBT_HEADER */ 1582 uint64_t zap_magic; /* ZAP_MAGIC */ 1583 1584 struct zap_table_phys { 1585 uint64_t zt_blk; /* starting block number */ 1586 uint64_t zt_numblks; /* number of blocks */ 1587 uint64_t zt_shift; /* bits to index it */ 1588 uint64_t zt_nextblk; /* next (larger) copy start block */ 1589 uint64_t zt_blks_copied; /* number source blocks copied */ 1590 } zap_ptrtbl; 1591 1592 uint64_t zap_freeblk; /* the next free block */ 1593 uint64_t zap_num_leafs; /* number of leafs */ 1594 uint64_t zap_num_entries; /* number of entries */ 1595 uint64_t zap_salt; /* salt to stir into hash function */ 1596 uint64_t zap_normflags; /* flags for u8_textprep_str() */ 1597 uint64_t zap_flags; /* zap_flags_t */ 1598 /* 1599 * This structure is followed by padding, and then the embedded 1600 * pointer table. The embedded pointer table takes up second 1601 * half of the block. It is accessed using the 1602 * ZAP_EMBEDDED_PTRTBL_ENT() macro. 1603 */ 1604 } zap_phys_t; 1605 1606 typedef struct zap_table_phys zap_table_phys_t; 1607 1608 struct spa; 1609 typedef struct fat_zap { 1610 int zap_block_shift; /* block size shift */ 1611 zap_phys_t *zap_phys; 1612 const struct spa *zap_spa; 1613 const dnode_phys_t *zap_dnode; 1614 } fat_zap_t; 1615 1616 #define ZAP_LEAF_MAGIC 0x2AB1EAF 1617 1618 /* chunk size = 24 bytes */ 1619 #define ZAP_LEAF_CHUNKSIZE 24 1620 1621 /* 1622 * The amount of space available for chunks is: 1623 * block size (1<<l->l_bs) - hash entry size (2) * number of hash 1624 * entries - header space (2*chunksize) 1625 */ 1626 #define ZAP_LEAF_NUMCHUNKS(l) \ 1627 (((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \ 1628 ZAP_LEAF_CHUNKSIZE - 2) 1629 1630 /* 1631 * The amount of space within the chunk available for the array is: 1632 * chunk size - space for type (1) - space for next pointer (2) 1633 */ 1634 #define ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3) 1635 1636 #define ZAP_LEAF_ARRAY_NCHUNKS(bytes) \ 1637 (((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES) 1638 1639 /* 1640 * Low water mark: when there are only this many chunks free, start 1641 * growing the ptrtbl. Ideally, this should be larger than a 1642 * "reasonably-sized" entry. 20 chunks is more than enough for the 1643 * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value), 1644 * while still being only around 3% for 16k blocks. 1645 */ 1646 #define ZAP_LEAF_LOW_WATER (20) 1647 1648 /* 1649 * The leaf hash table has block size / 2^5 (32) number of entries, 1650 * which should be more than enough for the maximum number of entries, 1651 * which is less than block size / CHUNKSIZE (24) / minimum number of 1652 * chunks per entry (3). 1653 */ 1654 #define ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5) 1655 #define ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l)) 1656 1657 /* 1658 * The chunks start immediately after the hash table. The end of the 1659 * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a 1660 * chunk_t. 1661 */ 1662 #define ZAP_LEAF_CHUNK(l, idx) \ 1663 ((zap_leaf_chunk_t *)(void *) \ 1664 ((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx] 1665 #define ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry) 1666 1667 #define ZAP_LEAF_HASH(l, h) \ 1668 ((ZAP_LEAF_HASH_NUMENTRIES(l)-1) & \ 1669 ((h) >> \ 1670 (64 - ZAP_LEAF_HASH_SHIFT(l) - (l)->l_phys->l_hdr.lh_prefix_len))) 1671 #define ZAP_LEAF_HASH_ENTPTR(l, h) (&(l)->l_phys->l_hash[ZAP_LEAF_HASH(l, h)]) 1672 1673 typedef enum zap_chunk_type { 1674 ZAP_CHUNK_FREE = 253, 1675 ZAP_CHUNK_ENTRY = 252, 1676 ZAP_CHUNK_ARRAY = 251, 1677 ZAP_CHUNK_TYPE_MAX = 250 1678 } zap_chunk_type_t; 1679 1680 /* 1681 * TAKE NOTE: 1682 * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified. 1683 */ 1684 typedef struct zap_leaf_phys { 1685 struct zap_leaf_header { 1686 uint64_t lh_block_type; /* ZBT_LEAF */ 1687 uint64_t lh_pad1; 1688 uint64_t lh_prefix; /* hash prefix of this leaf */ 1689 uint32_t lh_magic; /* ZAP_LEAF_MAGIC */ 1690 uint16_t lh_nfree; /* number free chunks */ 1691 uint16_t lh_nentries; /* number of entries */ 1692 uint16_t lh_prefix_len; /* num bits used to id this */ 1693 1694 /* above is accessable to zap, below is zap_leaf private */ 1695 1696 uint16_t lh_freelist; /* chunk head of free list */ 1697 uint8_t lh_pad2[12]; 1698 } l_hdr; /* 2 24-byte chunks */ 1699 1700 /* 1701 * The header is followed by a hash table with 1702 * ZAP_LEAF_HASH_NUMENTRIES(zap) entries. The hash table is 1703 * followed by an array of ZAP_LEAF_NUMCHUNKS(zap) 1704 * zap_leaf_chunk structures. These structures are accessed 1705 * with the ZAP_LEAF_CHUNK() macro. 1706 */ 1707 1708 uint16_t l_hash[1]; 1709 } zap_leaf_phys_t; 1710 1711 typedef union zap_leaf_chunk { 1712 struct zap_leaf_entry { 1713 uint8_t le_type; /* always ZAP_CHUNK_ENTRY */ 1714 uint8_t le_value_intlen; /* size of ints */ 1715 uint16_t le_next; /* next entry in hash chain */ 1716 uint16_t le_name_chunk; /* first chunk of the name */ 1717 uint16_t le_name_numints; /* bytes in name, incl null */ 1718 uint16_t le_value_chunk; /* first chunk of the value */ 1719 uint16_t le_value_numints; /* value length in ints */ 1720 uint32_t le_cd; /* collision differentiator */ 1721 uint64_t le_hash; /* hash value of the name */ 1722 } l_entry; 1723 struct zap_leaf_array { 1724 uint8_t la_type; /* always ZAP_CHUNK_ARRAY */ 1725 uint8_t la_array[ZAP_LEAF_ARRAY_BYTES]; 1726 uint16_t la_next; /* next blk or CHAIN_END */ 1727 } l_array; 1728 struct zap_leaf_free { 1729 uint8_t lf_type; /* always ZAP_CHUNK_FREE */ 1730 uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES]; 1731 uint16_t lf_next; /* next in free list, or CHAIN_END */ 1732 } l_free; 1733 } zap_leaf_chunk_t; 1734 1735 typedef struct zap_leaf { 1736 int l_bs; /* block size shift */ 1737 zap_leaf_phys_t *l_phys; 1738 } zap_leaf_t; 1739 1740 #define ZAP_MAXNAMELEN 256 1741 #define ZAP_MAXVALUELEN (1024 * 8) 1742 1743 #define ACE_READ_DATA 0x00000001 /* file: read data */ 1744 #define ACE_LIST_DIRECTORY 0x00000001 /* dir: list files */ 1745 #define ACE_WRITE_DATA 0x00000002 /* file: write data */ 1746 #define ACE_ADD_FILE 0x00000002 /* dir: create file */ 1747 #define ACE_APPEND_DATA 0x00000004 /* file: append data */ 1748 #define ACE_ADD_SUBDIRECTORY 0x00000004 /* dir: create subdir */ 1749 #define ACE_READ_NAMED_ATTRS 0x00000008 /* FILE_READ_EA */ 1750 #define ACE_WRITE_NAMED_ATTRS 0x00000010 /* FILE_WRITE_EA */ 1751 #define ACE_EXECUTE 0x00000020 /* file: execute */ 1752 #define ACE_TRAVERSE 0x00000020 /* dir: lookup name */ 1753 #define ACE_DELETE_CHILD 0x00000040 /* dir: unlink child */ 1754 #define ACE_READ_ATTRIBUTES 0x00000080 /* (all) stat, etc. */ 1755 #define ACE_WRITE_ATTRIBUTES 0x00000100 /* (all) utimes, etc. */ 1756 #define ACE_DELETE 0x00010000 /* (all) unlink self */ 1757 #define ACE_READ_ACL 0x00020000 /* (all) getsecattr */ 1758 #define ACE_WRITE_ACL 0x00040000 /* (all) setsecattr */ 1759 #define ACE_WRITE_OWNER 0x00080000 /* (all) chown */ 1760 #define ACE_SYNCHRONIZE 0x00100000 /* (all) */ 1761 1762 #define ACE_FILE_INHERIT_ACE 0x0001 1763 #define ACE_DIRECTORY_INHERIT_ACE 0x0002 1764 #define ACE_NO_PROPAGATE_INHERIT_ACE 0x0004 1765 #define ACE_INHERIT_ONLY_ACE 0x0008 1766 #define ACE_SUCCESSFUL_ACCESS_ACE_FLAG 0x0010 1767 #define ACE_FAILED_ACCESS_ACE_FLAG 0x0020 1768 #define ACE_IDENTIFIER_GROUP 0x0040 1769 #define ACE_INHERITED_ACE 0x0080 1770 #define ACE_OWNER 0x1000 1771 #define ACE_GROUP 0x2000 1772 #define ACE_EVERYONE 0x4000 1773 1774 #define ACE_ACCESS_ALLOWED_ACE_TYPE 0x0000 1775 #define ACE_ACCESS_DENIED_ACE_TYPE 0x0001 1776 #define ACE_SYSTEM_AUDIT_ACE_TYPE 0x0002 1777 #define ACE_SYSTEM_ALARM_ACE_TYPE 0x0003 1778 1779 typedef struct zfs_ace_hdr { 1780 uint16_t z_type; 1781 uint16_t z_flags; 1782 uint32_t z_access_mask; 1783 } zfs_ace_hdr_t; 1784 1785 /* 1786 * Define special zfs pflags 1787 */ 1788 #define ZFS_XATTR 0x1 /* is an extended attribute */ 1789 #define ZFS_INHERIT_ACE 0x2 /* ace has inheritable ACEs */ 1790 #define ZFS_ACL_TRIVIAL 0x4 /* files ACL is trivial */ 1791 #define ZFS_ACL_OBJ_ACE 0x8 /* ACL has CMPLX Object ACE */ 1792 #define ZFS_ACL_PROTECTED 0x10 /* ACL protected */ 1793 #define ZFS_ACL_DEFAULTED 0x20 /* ACL should be defaulted */ 1794 #define ZFS_ACL_AUTO_INHERIT 0x40 /* ACL should be inherited */ 1795 #define ZFS_BONUS_SCANSTAMP 0x80 /* Scanstamp in bonus area */ 1796 #define ZFS_NO_EXECS_DENIED 0x100 /* exec was given to everyone */ 1797 1798 #define ZFS_READONLY 0x0000000100000000ull 1799 #define ZFS_HIDDEN 0x0000000200000000ull 1800 #define ZFS_SYSTEM 0x0000000400000000ull 1801 #define ZFS_ARCHIVE 0x0000000800000000ull 1802 #define ZFS_IMMUTABLE 0x0000001000000000ull 1803 #define ZFS_NOUNLINK 0x0000002000000000ull 1804 #define ZFS_APPENDONLY 0x0000004000000000ull 1805 #define ZFS_NODUMP 0x0000008000000000ull 1806 #define ZFS_OPAQUE 0x0000010000000000ull 1807 #define ZFS_AV_QUARANTINED 0x0000020000000000ull 1808 #define ZFS_AV_MODIFIED 0x0000040000000000ull 1809 #define ZFS_REPARSE 0x0000080000000000ull 1810 #define ZFS_OFFLINE 0x0000100000000000ull 1811 #define ZFS_SPARSE 0x0000200000000000ull 1812 1813 #define MASTER_NODE_OBJ 1 1814 1815 /* 1816 * special attributes for master node. 1817 */ 1818 1819 #define ZFS_FSID "FSID" 1820 #define ZFS_UNLINKED_SET "DELETE_QUEUE" 1821 #define ZFS_ROOT_OBJ "ROOT" 1822 #define ZPL_VERSION_OBJ "VERSION" 1823 #define ZFS_PROP_BLOCKPERPAGE "BLOCKPERPAGE" 1824 #define ZFS_PROP_NOGROWBLOCKS "NOGROWBLOCKS" 1825 #define ZFS_SA_ATTRS "SA_ATTRS" 1826 1827 #define ZFS_FLAG_BLOCKPERPAGE 0x1 1828 #define ZFS_FLAG_NOGROWBLOCKS 0x2 1829 1830 /* 1831 * ZPL version - rev'd whenever an incompatible on-disk format change 1832 * occurs. Independent of SPA/DMU/ZAP versioning. 1833 */ 1834 1835 #define ZPL_VERSION 1ULL 1836 1837 /* 1838 * The directory entry has the type (currently unused on Solaris) in the 1839 * top 4 bits, and the object number in the low 48 bits. The "middle" 1840 * 12 bits are unused. 1841 */ 1842 #define ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4) 1843 #define ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48) 1844 #define ZFS_DIRENT_MAKE(type, obj) (((uint64_t)type << 60) | obj) 1845 1846 typedef struct ace { 1847 uid_t a_who; /* uid or gid */ 1848 uint32_t a_access_mask; /* read,write,... */ 1849 uint16_t a_flags; /* see below */ 1850 uint16_t a_type; /* allow or deny */ 1851 } ace_t; 1852 1853 #define ACE_SLOT_CNT 6 1854 1855 typedef struct zfs_znode_acl { 1856 uint64_t z_acl_extern_obj; /* ext acl pieces */ 1857 uint32_t z_acl_count; /* Number of ACEs */ 1858 uint16_t z_acl_version; /* acl version */ 1859 uint16_t z_acl_pad; /* pad */ 1860 ace_t z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */ 1861 } zfs_znode_acl_t; 1862 1863 /* 1864 * This is the persistent portion of the znode. It is stored 1865 * in the "bonus buffer" of the file. Short symbolic links 1866 * are also stored in the bonus buffer. 1867 */ 1868 typedef struct znode_phys { 1869 uint64_t zp_atime[2]; /* 0 - last file access time */ 1870 uint64_t zp_mtime[2]; /* 16 - last file modification time */ 1871 uint64_t zp_ctime[2]; /* 32 - last file change time */ 1872 uint64_t zp_crtime[2]; /* 48 - creation time */ 1873 uint64_t zp_gen; /* 64 - generation (txg of creation) */ 1874 uint64_t zp_mode; /* 72 - file mode bits */ 1875 uint64_t zp_size; /* 80 - size of file */ 1876 uint64_t zp_parent; /* 88 - directory parent (`..') */ 1877 uint64_t zp_links; /* 96 - number of links to file */ 1878 uint64_t zp_xattr; /* 104 - DMU object for xattrs */ 1879 uint64_t zp_rdev; /* 112 - dev_t for VBLK & VCHR files */ 1880 uint64_t zp_flags; /* 120 - persistent flags */ 1881 uint64_t zp_uid; /* 128 - file owner */ 1882 uint64_t zp_gid; /* 136 - owning group */ 1883 uint64_t zp_pad[4]; /* 144 - future */ 1884 zfs_znode_acl_t zp_acl; /* 176 - 263 ACL */ 1885 /* 1886 * Data may pad out any remaining bytes in the znode buffer, eg: 1887 * 1888 * |<---------------------- dnode_phys (512) ------------------------>| 1889 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->| 1890 * |<---- znode (264) ---->|<---- data (56) ---->| 1891 * 1892 * At present, we only use this space to store symbolic links. 1893 */ 1894 } znode_phys_t; 1895 1896 /* 1897 * In-core vdev representation. 1898 */ 1899 struct vdev; 1900 struct spa; 1901 typedef int vdev_phys_read_t(struct vdev *, void *, off_t, void *, size_t); 1902 typedef int vdev_phys_write_t(struct vdev *, off_t, void *, size_t); 1903 typedef int vdev_read_t(struct vdev *, const blkptr_t *, void *, off_t, size_t); 1904 1905 typedef STAILQ_HEAD(vdev_list, vdev) vdev_list_t; 1906 1907 typedef struct vdev_indirect_mapping_entry_phys { 1908 /* 1909 * Decode with DVA_MAPPING_* macros. 1910 * Contains: 1911 * the source offset (low 63 bits) 1912 * the one-bit "mark", used for garbage collection (by zdb) 1913 */ 1914 uint64_t vimep_src; 1915 1916 /* 1917 * Note: the DVA's asize is 24 bits, and can thus store ranges 1918 * up to 8GB. 1919 */ 1920 dva_t vimep_dst; 1921 } vdev_indirect_mapping_entry_phys_t; 1922 1923 #define DVA_MAPPING_GET_SRC_OFFSET(vimep) \ 1924 BF64_GET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0) 1925 #define DVA_MAPPING_SET_SRC_OFFSET(vimep, x) \ 1926 BF64_SET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0, x) 1927 1928 /* 1929 * This is stored in the bonus buffer of the mapping object, see comment of 1930 * vdev_indirect_config for more details. 1931 */ 1932 typedef struct vdev_indirect_mapping_phys { 1933 uint64_t vimp_max_offset; 1934 uint64_t vimp_bytes_mapped; 1935 uint64_t vimp_num_entries; /* number of v_i_m_entry_phys_t's */ 1936 1937 /* 1938 * For each entry in the mapping object, this object contains an 1939 * entry representing the number of bytes of that mapping entry 1940 * that were no longer in use by the pool at the time this indirect 1941 * vdev was last condensed. 1942 */ 1943 uint64_t vimp_counts_object; 1944 } vdev_indirect_mapping_phys_t; 1945 1946 #define VDEV_INDIRECT_MAPPING_SIZE_V0 (3 * sizeof (uint64_t)) 1947 1948 typedef struct vdev_indirect_mapping { 1949 uint64_t vim_object; 1950 boolean_t vim_havecounts; 1951 1952 /* vim_entries segment offset currently in memory. */ 1953 uint64_t vim_entry_offset; 1954 /* vim_entries segment size. */ 1955 size_t vim_num_entries; 1956 1957 /* Needed by dnode_read() */ 1958 const void *vim_spa; 1959 dnode_phys_t *vim_dn; 1960 1961 /* 1962 * An ordered array of mapping entries, sorted by source offset. 1963 * Note that vim_entries is needed during a removal (and contains 1964 * mappings that have been synced to disk so far) to handle frees 1965 * from the removing device. 1966 */ 1967 vdev_indirect_mapping_entry_phys_t *vim_entries; 1968 objset_phys_t *vim_objset; 1969 vdev_indirect_mapping_phys_t *vim_phys; 1970 } vdev_indirect_mapping_t; 1971 1972 /* 1973 * On-disk indirect vdev state. 1974 * 1975 * An indirect vdev is described exclusively in the MOS config of a pool. 1976 * The config for an indirect vdev includes several fields, which are 1977 * accessed in memory by a vdev_indirect_config_t. 1978 */ 1979 typedef struct vdev_indirect_config { 1980 /* 1981 * Object (in MOS) which contains the indirect mapping. This object 1982 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by 1983 * vimep_src. The bonus buffer for this object is a 1984 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev 1985 * removal is initiated. 1986 * 1987 * Note that this object can be empty if none of the data on the vdev 1988 * has been copied yet. 1989 */ 1990 uint64_t vic_mapping_object; 1991 1992 /* 1993 * Object (in MOS) which contains the birth times for the mapping 1994 * entries. This object contains an array of 1995 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus 1996 * buffer for this object is a vdev_indirect_birth_phys_t. This object 1997 * is allocated when a vdev removal is initiated. 1998 * 1999 * Note that this object can be empty if none of the vdev has yet been 2000 * copied. 2001 */ 2002 uint64_t vic_births_object; 2003 2004 /* 2005 * This is the vdev ID which was removed previous to this vdev, or 2006 * UINT64_MAX if there are no previously removed vdevs. 2007 */ 2008 uint64_t vic_prev_indirect_vdev; 2009 } vdev_indirect_config_t; 2010 2011 typedef struct vdev { 2012 STAILQ_ENTRY(vdev) v_childlink; /* link in parent's child list */ 2013 vdev_list_t v_children; /* children of this vdev */ 2014 const char *v_name; /* vdev name */ 2015 uint64_t v_guid; /* vdev guid */ 2016 uint64_t v_label; /* label instantiated from (top vdev) */ 2017 uint64_t v_txg; /* most recent transaction (top vdev) */ 2018 uint64_t v_id; /* index in parent */ 2019 uint64_t v_psize; /* physical device capacity */ 2020 int v_ashift; /* offset to block shift */ 2021 int v_nparity; /* # parity for raidz */ 2022 struct vdev *v_top; /* parent vdev */ 2023 size_t v_nchildren; /* # children */ 2024 vdev_state_t v_state; /* current state */ 2025 vdev_phys_read_t *v_phys_read; /* read from raw leaf vdev */ 2026 vdev_phys_write_t *v_phys_write; /* write to raw leaf vdev */ 2027 vdev_read_t *v_read; /* read from vdev */ 2028 void *v_priv; /* data for read/write function */ 2029 boolean_t v_islog; 2030 struct spa *v_spa; /* link to spa */ 2031 /* 2032 * Values stored in the config for an indirect or removing vdev. 2033 */ 2034 vdev_indirect_config_t vdev_indirect_config; 2035 vdev_indirect_mapping_t *v_mapping; 2036 } vdev_t; 2037 2038 /* 2039 * In-core pool representation. 2040 */ 2041 typedef STAILQ_HEAD(spa_list, spa) spa_list_t; 2042 2043 typedef struct spa { 2044 STAILQ_ENTRY(spa) spa_link; /* link in global pool list */ 2045 char *spa_name; /* pool name */ 2046 uint64_t spa_guid; /* pool guid */ 2047 struct uberblock *spa_uberblock; /* best uberblock so far */ 2048 vdev_t *spa_root_vdev; /* toplevel vdev container */ 2049 objset_phys_t *spa_mos; /* MOS for this pool */ 2050 zio_cksum_salt_t spa_cksum_salt; /* secret salt for cksum */ 2051 void *spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS]; 2052 boolean_t spa_with_log; /* this pool has log */ 2053 2054 struct uberblock spa_uberblock_master; /* best uberblock so far */ 2055 objset_phys_t spa_mos_master; /* MOS for this pool */ 2056 struct uberblock spa_uberblock_checkpoint; /* checkpoint uberblock */ 2057 objset_phys_t spa_mos_checkpoint; /* Checkpoint MOS */ 2058 void *spa_bootenv; /* bootenv from pool label */ 2059 } spa_t; 2060 2061 /* IO related arguments. */ 2062 typedef struct zio { 2063 spa_t *io_spa; 2064 blkptr_t *io_bp; 2065 void *io_data; 2066 uint64_t io_size; 2067 uint64_t io_offset; 2068 2069 /* Stuff for the vdev stack */ 2070 vdev_t *io_vd; 2071 void *io_vsd; 2072 2073 int io_error; 2074 } zio_t; 2075 2076 extern void decode_embedded_bp_compressed(const blkptr_t *, void *); 2077 2078 #endif /* _ZFSIMPL_H_ */ 2079