| 1 | /* |
| 2 | * SHA1 routine optimized to do word accesses rather than byte accesses, |
| 3 | * and to avoid unnecessary copies into the context array. |
| 4 | * |
| 5 | * This was initially based on the Mozilla SHA1 implementation, although |
| 6 | * none of the original Mozilla code remains. |
| 7 | */ |
| 8 | |
| 9 | /* this is only to get definitions for memcpy(), ntohl() and htonl() */ |
| 10 | #include "../git-compat-util.h" |
| 11 | |
| 12 | #include "sha1.h" |
| 13 | |
| 14 | #define SHA_ROT(X,l,r) (((X) << (l)) | ((X) >> (r))) |
| 15 | #define SHA_ROL(X,n) SHA_ROT(X,n,32-(n)) |
| 16 | #define SHA_ROR(X,n) SHA_ROT(X,32-(n),n) |
| 17 | |
| 18 | /* |
| 19 | * If you have 32 registers or more, the compiler can (and should) |
| 20 | * try to change the array[] accesses into registers. However, on |
| 21 | * machines with less than ~25 registers, that won't really work, |
| 22 | * and at least gcc will make an unholy mess of it. |
| 23 | * |
| 24 | * So to avoid that mess which just slows things down, we force |
| 25 | * the stores to memory to actually happen (we might be better off |
| 26 | * with a 'W(t)=(val);asm("":"+m" (W(t))' there instead, as |
| 27 | * suggested by Artur Skawina - that will also make gcc unable to |
| 28 | * try to do the silly "optimize away loads" part because it won't |
| 29 | * see what the value will be). |
| 30 | * |
| 31 | * On ARM we get the best code generation by forcing a full memory barrier |
| 32 | * between each SHA_ROUND, otherwise gcc happily get wild with spilling and |
| 33 | * the stack frame size simply explode and performance goes down the drain. |
| 34 | */ |
| 35 | |
| 36 | #if defined(__i386__) || defined(__x86_64__) |
| 37 | #define setW(x, val) (*(volatile unsigned int *)&W(x) = (val)) |
| 38 | #elif defined(__GNUC__) && defined(__arm__) |
| 39 | #define setW(x, val) do { W(x) = (val); __asm__("":::"memory"); } while (0) |
| 40 | #else |
| 41 | #define setW(x, val) (W(x) = (val)) |
| 42 | #endif |
| 43 | |
| 44 | /* This "rolls" over the 512-bit array */ |
| 45 | #define W(x) (array[(x)&15]) |
| 46 | |
| 47 | /* |
| 48 | * Where do we get the source from? The first 16 iterations get it from |
| 49 | * the input data, the next mix it from the 512-bit array. |
| 50 | */ |
| 51 | #define SHA_SRC(t) get_be32((unsigned char *) block + (t)*4) |
| 52 | #define SHA_MIX(t) SHA_ROL(W((t)+13) ^ W((t)+8) ^ W((t)+2) ^ W(t), 1) |
| 53 | |
| 54 | #define SHA_ROUND(t, input, fn, constant, A, B, C, D, E) do { \ |
| 55 | unsigned int TEMP = input(t); setW(t, TEMP); \ |
| 56 | E += TEMP + SHA_ROL(A,5) + (fn) + (constant); \ |
| 57 | B = SHA_ROR(B, 2); } while (0) |
| 58 | |
| 59 | #define T_0_15(t, A, B, C, D, E) SHA_ROUND(t, SHA_SRC, (((C^D)&B)^D) , 0x5a827999, A, B, C, D, E ) |
| 60 | #define T_16_19(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (((C^D)&B)^D) , 0x5a827999, A, B, C, D, E ) |
| 61 | #define T_20_39(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (B^C^D) , 0x6ed9eba1, A, B, C, D, E ) |
| 62 | #define T_40_59(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, ((B&C)+(D&(B^C))) , 0x8f1bbcdc, A, B, C, D, E ) |
| 63 | #define T_60_79(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (B^C^D) , 0xca62c1d6, A, B, C, D, E ) |
| 64 | |
| 65 | static void blk_SHA1_Block(blk_SHA_CTX *ctx, const void *block) |
| 66 | { |
| 67 | unsigned int A,B,C,D,E; |
| 68 | unsigned int array[16]; |
| 69 | |
| 70 | A = ctx->H[0]; |
| 71 | B = ctx->H[1]; |
| 72 | C = ctx->H[2]; |
| 73 | D = ctx->H[3]; |
| 74 | E = ctx->H[4]; |
| 75 | |
| 76 | /* Round 1 - iterations 0-16 take their input from 'block' */ |
| 77 | T_0_15( 0, A, B, C, D, E); |
| 78 | T_0_15( 1, E, A, B, C, D); |
| 79 | T_0_15( 2, D, E, A, B, C); |
| 80 | T_0_15( 3, C, D, E, A, B); |
| 81 | T_0_15( 4, B, C, D, E, A); |
| 82 | T_0_15( 5, A, B, C, D, E); |
| 83 | T_0_15( 6, E, A, B, C, D); |
| 84 | T_0_15( 7, D, E, A, B, C); |
| 85 | T_0_15( 8, C, D, E, A, B); |
| 86 | T_0_15( 9, B, C, D, E, A); |
| 87 | T_0_15(10, A, B, C, D, E); |
| 88 | T_0_15(11, E, A, B, C, D); |
| 89 | T_0_15(12, D, E, A, B, C); |
| 90 | T_0_15(13, C, D, E, A, B); |
| 91 | T_0_15(14, B, C, D, E, A); |
| 92 | T_0_15(15, A, B, C, D, E); |
| 93 | |
| 94 | /* Round 1 - tail. Input from 512-bit mixing array */ |
| 95 | T_16_19(16, E, A, B, C, D); |
| 96 | T_16_19(17, D, E, A, B, C); |
| 97 | T_16_19(18, C, D, E, A, B); |
| 98 | T_16_19(19, B, C, D, E, A); |
| 99 | |
| 100 | /* Round 2 */ |
| 101 | T_20_39(20, A, B, C, D, E); |
| 102 | T_20_39(21, E, A, B, C, D); |
| 103 | T_20_39(22, D, E, A, B, C); |
| 104 | T_20_39(23, C, D, E, A, B); |
| 105 | T_20_39(24, B, C, D, E, A); |
| 106 | T_20_39(25, A, B, C, D, E); |
| 107 | T_20_39(26, E, A, B, C, D); |
| 108 | T_20_39(27, D, E, A, B, C); |
| 109 | T_20_39(28, C, D, E, A, B); |
| 110 | T_20_39(29, B, C, D, E, A); |
| 111 | T_20_39(30, A, B, C, D, E); |
| 112 | T_20_39(31, E, A, B, C, D); |
| 113 | T_20_39(32, D, E, A, B, C); |
| 114 | T_20_39(33, C, D, E, A, B); |
| 115 | T_20_39(34, B, C, D, E, A); |
| 116 | T_20_39(35, A, B, C, D, E); |
| 117 | T_20_39(36, E, A, B, C, D); |
| 118 | T_20_39(37, D, E, A, B, C); |
| 119 | T_20_39(38, C, D, E, A, B); |
| 120 | T_20_39(39, B, C, D, E, A); |
| 121 | |
| 122 | /* Round 3 */ |
| 123 | T_40_59(40, A, B, C, D, E); |
| 124 | T_40_59(41, E, A, B, C, D); |
| 125 | T_40_59(42, D, E, A, B, C); |
| 126 | T_40_59(43, C, D, E, A, B); |
| 127 | T_40_59(44, B, C, D, E, A); |
| 128 | T_40_59(45, A, B, C, D, E); |
| 129 | T_40_59(46, E, A, B, C, D); |
| 130 | T_40_59(47, D, E, A, B, C); |
| 131 | T_40_59(48, C, D, E, A, B); |
| 132 | T_40_59(49, B, C, D, E, A); |
| 133 | T_40_59(50, A, B, C, D, E); |
| 134 | T_40_59(51, E, A, B, C, D); |
| 135 | T_40_59(52, D, E, A, B, C); |
| 136 | T_40_59(53, C, D, E, A, B); |
| 137 | T_40_59(54, B, C, D, E, A); |
| 138 | T_40_59(55, A, B, C, D, E); |
| 139 | T_40_59(56, E, A, B, C, D); |
| 140 | T_40_59(57, D, E, A, B, C); |
| 141 | T_40_59(58, C, D, E, A, B); |
| 142 | T_40_59(59, B, C, D, E, A); |
| 143 | |
| 144 | /* Round 4 */ |
| 145 | T_60_79(60, A, B, C, D, E); |
| 146 | T_60_79(61, E, A, B, C, D); |
| 147 | T_60_79(62, D, E, A, B, C); |
| 148 | T_60_79(63, C, D, E, A, B); |
| 149 | T_60_79(64, B, C, D, E, A); |
| 150 | T_60_79(65, A, B, C, D, E); |
| 151 | T_60_79(66, E, A, B, C, D); |
| 152 | T_60_79(67, D, E, A, B, C); |
| 153 | T_60_79(68, C, D, E, A, B); |
| 154 | T_60_79(69, B, C, D, E, A); |
| 155 | T_60_79(70, A, B, C, D, E); |
| 156 | T_60_79(71, E, A, B, C, D); |
| 157 | T_60_79(72, D, E, A, B, C); |
| 158 | T_60_79(73, C, D, E, A, B); |
| 159 | T_60_79(74, B, C, D, E, A); |
| 160 | T_60_79(75, A, B, C, D, E); |
| 161 | T_60_79(76, E, A, B, C, D); |
| 162 | T_60_79(77, D, E, A, B, C); |
| 163 | T_60_79(78, C, D, E, A, B); |
| 164 | T_60_79(79, B, C, D, E, A); |
| 165 | |
| 166 | ctx->H[0] += A; |
| 167 | ctx->H[1] += B; |
| 168 | ctx->H[2] += C; |
| 169 | ctx->H[3] += D; |
| 170 | ctx->H[4] += E; |
| 171 | } |
| 172 | |
| 173 | void blk_SHA1_Init(blk_SHA_CTX *ctx) |
| 174 | { |
| 175 | ctx->size = 0; |
| 176 | |
| 177 | /* Initialize H with the magic constants (see FIPS180 for constants) */ |
| 178 | ctx->H[0] = 0x67452301; |
| 179 | ctx->H[1] = 0xefcdab89; |
| 180 | ctx->H[2] = 0x98badcfe; |
| 181 | ctx->H[3] = 0x10325476; |
| 182 | ctx->H[4] = 0xc3d2e1f0; |
| 183 | } |
| 184 | |
| 185 | void blk_SHA1_Update(blk_SHA_CTX *ctx, const void *data, size_t len) |
| 186 | { |
| 187 | unsigned int lenW = ctx->size & 63; |
| 188 | |
| 189 | ctx->size += len; |
| 190 | |
| 191 | /* Read the data into W and process blocks as they get full */ |
| 192 | if (lenW) { |
| 193 | unsigned int left = 64 - lenW; |
| 194 | if (len < left) |
| 195 | left = len; |
| 196 | memcpy(lenW + (char *)ctx->W, data, left); |
| 197 | lenW = (lenW + left) & 63; |
| 198 | len -= left; |
| 199 | data = ((const char *)data + left); |
| 200 | if (lenW) |
| 201 | return; |
| 202 | blk_SHA1_Block(ctx, ctx->W); |
| 203 | } |
| 204 | while (len >= 64) { |
| 205 | blk_SHA1_Block(ctx, data); |
| 206 | data = ((const char *)data + 64); |
| 207 | len -= 64; |
| 208 | } |
| 209 | if (len) |
| 210 | memcpy(ctx->W, data, len); |
| 211 | } |
| 212 | |
| 213 | void blk_SHA1_Final(unsigned char hashout[20], blk_SHA_CTX *ctx) |
| 214 | { |
| 215 | static const unsigned char pad[64] = { 0x80 }; |
| 216 | unsigned int padlen[2]; |
| 217 | int i; |
| 218 | |
| 219 | /* Pad with a binary 1 (ie 0x80), then zeroes, then length */ |
| 220 | padlen[0] = htonl((uint32_t)(ctx->size >> 29)); |
| 221 | padlen[1] = htonl((uint32_t)(ctx->size << 3)); |
| 222 | |
| 223 | i = ctx->size & 63; |
| 224 | blk_SHA1_Update(ctx, pad, 1 + (63 & (55 - i))); |
| 225 | blk_SHA1_Update(ctx, padlen, 8); |
| 226 | |
| 227 | /* Output hash */ |
| 228 | for (i = 0; i < 5; i++) |
| 229 | put_be32(hashout + i * 4, ctx->H[i]); |
| 230 | } |