1 /* 2 * GRUB -- GRand Unified Bootloader 3 * Copyright (C) 1999,2000,2001,2002,2003,2004 Free Software Foundation, Inc. 4 * 5 * This program is free software; you can redistribute it and/or modify 6 * it under the terms of the GNU General Public License as published by 7 * the Free Software Foundation; either version 2 of the License, or 8 * (at your option) any later version. 9 * 10 * This program is distributed in the hope that it will be useful, 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 13 * GNU General Public License for more details. 14 * 15 * You should have received a copy of the GNU General Public License 16 * along with this program; If not, see <http://www.gnu.org/licenses/>. 17 */ 18 /* 19 * Copyright 2010 Sun Microsystems, Inc. All rights reserved. 20 * Use is subject to license terms. 21 */ 22 23 #ifndef _SYS_SPA_H 24 #define _SYS_SPA_H 25 26 /* 27 * General-purpose 32-bit and 64-bit bitfield encodings. 28 */ 29 #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) 30 #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) 31 #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) 32 #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) 33 34 #define BF32_GET(x, low, len) BF32_DECODE(x, low, len) 35 #define BF64_GET(x, low, len) BF64_DECODE(x, low, len) 36 37 #define BF32_SET(x, low, len, val) \ 38 ((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len)) 39 #define BF64_SET(x, low, len, val) \ 40 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)) 41 42 #define BF32_GET_SB(x, low, len, shift, bias) \ 43 ((BF32_GET(x, low, len) + (bias)) << (shift)) 44 #define BF64_GET_SB(x, low, len, shift, bias) \ 45 ((BF64_GET(x, low, len) + (bias)) << (shift)) 46 47 #define BF32_SET_SB(x, low, len, shift, bias, val) \ 48 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)) 49 #define BF64_SET_SB(x, low, len, shift, bias, val) \ 50 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)) 51 52 /* 53 * We currently support nine block sizes, from 512 bytes to 128K. 54 * We could go higher, but the benefits are near-zero and the cost 55 * of COWing a giant block to modify one byte would become excessive. 56 */ 57 #define SPA_MINBLOCKSHIFT 9 58 #define SPA_MAXBLOCKSHIFT 17 59 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 60 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 61 62 #define SPA_BLOCKSIZES (SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1) 63 64 /* 65 * Size of block to hold the configuration data (a packed nvlist) 66 */ 67 #define SPA_CONFIG_BLOCKSIZE (1 << 14) 68 69 /* 70 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. 71 * The ASIZE encoding should be at least 64 times larger (6 more bits) 72 * to support up to 4-way RAID-Z mirror mode with worst-case gang block 73 * overhead, three DVAs per bp, plus one more bit in case we do anything 74 * else that expands the ASIZE. 75 */ 76 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ 77 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ 78 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ 79 80 /* 81 * All SPA data is represented by 128-bit data virtual addresses (DVAs). 82 * The members of the dva_t should be considered opaque outside the SPA. 83 */ 84 typedef struct dva { 85 uint64_t dva_word[2]; 86 } dva_t; 87 88 /* 89 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. 90 */ 91 typedef struct zio_cksum { 92 uint64_t zc_word[4]; 93 } zio_cksum_t; 94 95 /* 96 * Each block is described by its DVAs, time of birth, checksum, etc. 97 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 98 * 99 * 64 56 48 40 32 24 16 8 0 100 * +-------+-------+-------+-------+-------+-------+-------+-------+ 101 * 0 | vdev1 | GRID | ASIZE | 102 * +-------+-------+-------+-------+-------+-------+-------+-------+ 103 * 1 |G| offset1 | 104 * +-------+-------+-------+-------+-------+-------+-------+-------+ 105 * 2 | vdev2 | GRID | ASIZE | 106 * +-------+-------+-------+-------+-------+-------+-------+-------+ 107 * 3 |G| offset2 | 108 * +-------+-------+-------+-------+-------+-------+-------+-------+ 109 * 4 | vdev3 | GRID | ASIZE | 110 * +-------+-------+-------+-------+-------+-------+-------+-------+ 111 * 5 |G| offset3 | 112 * +-------+-------+-------+-------+-------+-------+-------+-------+ 113 * 6 |BDX|lvl| type | cksum | comp | PSIZE | LSIZE | 114 * +-------+-------+-------+-------+-------+-------+-------+-------+ 115 * 7 | padding | 116 * +-------+-------+-------+-------+-------+-------+-------+-------+ 117 * 8 | padding | 118 * +-------+-------+-------+-------+-------+-------+-------+-------+ 119 * 9 | physical birth txg | 120 * +-------+-------+-------+-------+-------+-------+-------+-------+ 121 * a | logical birth txg | 122 * +-------+-------+-------+-------+-------+-------+-------+-------+ 123 * b | fill count | 124 * +-------+-------+-------+-------+-------+-------+-------+-------+ 125 * c | checksum[0] | 126 * +-------+-------+-------+-------+-------+-------+-------+-------+ 127 * d | checksum[1] | 128 * +-------+-------+-------+-------+-------+-------+-------+-------+ 129 * e | checksum[2] | 130 * +-------+-------+-------+-------+-------+-------+-------+-------+ 131 * f | checksum[3] | 132 * +-------+-------+-------+-------+-------+-------+-------+-------+ 133 * 134 * Legend: 135 * 136 * vdev virtual device ID 137 * offset offset into virtual device 138 * LSIZE logical size 139 * PSIZE physical size (after compression) 140 * ASIZE allocated size (including RAID-Z parity and gang block headers) 141 * GRID RAID-Z layout information (reserved for future use) 142 * cksum checksum function 143 * comp compression function 144 * G gang block indicator 145 * B byteorder (endianness) 146 * D dedup 147 * X unused 148 * lvl level of indirection 149 * type DMU object type 150 * phys birth txg of block allocation; zero if same as logical birth txg 151 * log. birth transaction group in which the block was logically born 152 * fill count number of non-zero blocks under this bp 153 * checksum[4] 256-bit checksum of the data this bp describes 154 */ 155 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 156 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 157 158 typedef struct blkptr { 159 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 160 uint64_t blk_prop; /* size, compression, type, etc */ 161 uint64_t blk_pad[2]; /* Extra space for the future */ 162 uint64_t blk_phys_birth; /* txg when block was allocated */ 163 uint64_t blk_birth; /* transaction group at birth */ 164 uint64_t blk_fill; /* fill count */ 165 zio_cksum_t blk_cksum; /* 256-bit checksum */ 166 } blkptr_t; 167 168 /* 169 * Macros to get and set fields in a bp or DVA. 170 */ 171 #define DVA_GET_ASIZE(dva) \ 172 BF64_GET_SB((dva)->dva_word[0], 0, 24, SPA_MINBLOCKSHIFT, 0) 173 #define DVA_SET_ASIZE(dva, x) \ 174 BF64_SET_SB((dva)->dva_word[0], 0, 24, SPA_MINBLOCKSHIFT, 0, x) 175 176 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 177 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 178 179 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) 180 #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) 181 182 #define DVA_GET_OFFSET(dva) \ 183 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 184 #define DVA_SET_OFFSET(dva, x) \ 185 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 186 187 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 188 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 189 190 #define BP_GET_LSIZE(bp) \ 191 BF64_GET_SB((bp)->blk_prop, 0, 16, SPA_MINBLOCKSHIFT, 1) 192 #define BP_SET_LSIZE(bp, x) \ 193 BF64_SET_SB((bp)->blk_prop, 0, 16, SPA_MINBLOCKSHIFT, 1, x) 194 195 #define BP_GET_PSIZE(bp) \ 196 BF64_GET_SB((bp)->blk_prop, 16, 16, SPA_MINBLOCKSHIFT, 1) 197 #define BP_SET_PSIZE(bp, x) \ 198 BF64_SET_SB((bp)->blk_prop, 16, 16, SPA_MINBLOCKSHIFT, 1, x) 199 200 #define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 8) 201 #define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 8, x) 202 203 #define BP_GET_CHECKSUM(bp) BF64_GET((bp)->blk_prop, 40, 8) 204 #define BP_SET_CHECKSUM(bp, x) BF64_SET((bp)->blk_prop, 40, 8, x) 205 206 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 207 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 208 209 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 210 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 211 212 #define BP_GET_PROP_BIT_61(bp) BF64_GET((bp)->blk_prop, 61, 1) 213 #define BP_SET_PROP_BIT_61(bp, x) BF64_SET((bp)->blk_prop, 61, 1, x) 214 215 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 216 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 217 218 #define BP_GET_BYTEORDER(bp) (0 - BF64_GET((bp)->blk_prop, 63, 1)) 219 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 220 221 #define BP_PHYSICAL_BIRTH(bp) \ 222 ((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 223 224 #define BP_SET_BIRTH(bp, logical, physical) \ 225 { \ 226 (bp)->blk_birth = (logical); \ 227 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \ 228 } 229 230 #define BP_GET_ASIZE(bp) \ 231 (DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 232 DVA_GET_ASIZE(&(bp)->blk_dva[2])) 233 234 #define BP_GET_UCSIZE(bp) \ 235 ((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \ 236 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)); 237 238 #define BP_GET_NDVAS(bp) \ 239 (!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 240 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 241 !!DVA_GET_ASIZE(&(bp)->blk_dva[2])) 242 243 #define BP_COUNT_GANG(bp) \ 244 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \ 245 DVA_GET_GANG(&(bp)->blk_dva[1]) + \ 246 DVA_GET_GANG(&(bp)->blk_dva[2])) 247 248 #define DVA_EQUAL(dva1, dva2) \ 249 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 250 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 251 252 #define BP_EQUAL(bp1, bp2) \ 253 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \ 254 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \ 255 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \ 256 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2])) 257 258 #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ 259 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ 260 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ 261 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ 262 ((zc1).zc_word[3] - (zc2).zc_word[3]))) 263 264 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 265 266 #define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ 267 { \ 268 (zcp)->zc_word[0] = w0; \ 269 (zcp)->zc_word[1] = w1; \ 270 (zcp)->zc_word[2] = w2; \ 271 (zcp)->zc_word[3] = w3; \ 272 } 273 274 #define BP_IDENTITY(bp) (&(bp)->blk_dva[0]) 275 #define BP_IS_GANG(bp) DVA_GET_GANG(BP_IDENTITY(bp)) 276 #define BP_IS_HOLE(bp) ((bp)->blk_birth == 0) 277 278 /* BP_IS_RAIDZ(bp) assumes no block compression */ 279 #define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \ 280 BP_GET_PSIZE(bp)) 281 282 #define BP_ZERO(bp) \ 283 { \ 284 (bp)->blk_dva[0].dva_word[0] = 0; \ 285 (bp)->blk_dva[0].dva_word[1] = 0; \ 286 (bp)->blk_dva[1].dva_word[0] = 0; \ 287 (bp)->blk_dva[1].dva_word[1] = 0; \ 288 (bp)->blk_dva[2].dva_word[0] = 0; \ 289 (bp)->blk_dva[2].dva_word[1] = 0; \ 290 (bp)->blk_prop = 0; \ 291 (bp)->blk_pad[0] = 0; \ 292 (bp)->blk_pad[1] = 0; \ 293 (bp)->blk_phys_birth = 0; \ 294 (bp)->blk_birth = 0; \ 295 (bp)->blk_fill = 0; \ 296 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 297 } 298 299 /* 300 * Note: the byteorder is either 0 or -1, both of which are palindromes. 301 * This simplifies the endianness handling a bit. 302 */ 303 #ifdef _BIG_ENDIAN 304 #define ZFS_HOST_BYTEORDER (0ULL) 305 #else 306 #define ZFS_HOST_BYTEORDER (-1ULL) 307 #endif 308 309 #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) 310 311 #define BP_SPRINTF_LEN 320 312 313 #endif /* _SYS_SPA_H */ 314