1 /////////////////////////////////////////////////////////////////////////////// 2 // 3 /// \file sha256.c 4 /// \brief SHA-256 5 /// 6 /// \todo Crypto++ has x86 ASM optimizations. They use SSE so if they 7 /// are imported to liblzma, SSE instructions need to be used 8 /// conditionally to keep the code working on older boxes. 9 // 10 // This code is based on the code found from 7-Zip, which has a modified 11 // version of the SHA-256 found from Crypto++ <http://www.cryptopp.com/>. 12 // The code was modified a little to fit into liblzma. 13 // 14 // Authors: Kevin Springle 15 // Wei Dai 16 // Igor Pavlov 17 // Lasse Collin 18 // 19 // This file has been put into the public domain. 20 // You can do whatever you want with this file. 21 // 22 /////////////////////////////////////////////////////////////////////////////// 23 24 #include "check.h" 25 26 // Rotate a uint32_t. GCC can optimize this to a rotate instruction 27 // at least on x86. 28 static inline uint32_t 29 rotr_32(uint32_t num, unsigned amount) 30 { 31 return (num >> amount) | (num << (32 - amount)); 32 } 33 34 #define blk0(i) (W[i] = conv32be(data[i])) 35 #define blk2(i) (W[i & 15] += s1(W[(i - 2) & 15]) + W[(i - 7) & 15] \ 36 + s0(W[(i - 15) & 15])) 37 38 #define Ch(x, y, z) (z ^ (x & (y ^ z))) 39 #define Maj(x, y, z) ((x & (y ^ z)) + (y & z)) 40 41 #define a(i) T[(0 - i) & 7] 42 #define b(i) T[(1 - i) & 7] 43 #define c(i) T[(2 - i) & 7] 44 #define d(i) T[(3 - i) & 7] 45 #define e(i) T[(4 - i) & 7] 46 #define f(i) T[(5 - i) & 7] 47 #define g(i) T[(6 - i) & 7] 48 #define h(i) T[(7 - i) & 7] 49 50 #define R(i, j, blk) \ 51 h(i) += S1(e(i)) + Ch(e(i), f(i), g(i)) + SHA256_K[i + j] + blk; \ 52 d(i) += h(i); \ 53 h(i) += S0(a(i)) + Maj(a(i), b(i), c(i)) 54 #define R0(i) R(i, 0, blk0(i)) 55 #define R2(i) R(i, j, blk2(i)) 56 57 #define S0(x) rotr_32(x ^ rotr_32(x ^ rotr_32(x, 9), 11), 2) 58 #define S1(x) rotr_32(x ^ rotr_32(x ^ rotr_32(x, 14), 5), 6) 59 #define s0(x) (rotr_32(x ^ rotr_32(x, 11), 7) ^ (x >> 3)) 60 #define s1(x) (rotr_32(x ^ rotr_32(x, 2), 17) ^ (x >> 10)) 61 62 63 static const uint32_t SHA256_K[64] = { 64 0x428A2F98, 0x71374491, 0xB5C0FBCF, 0xE9B5DBA5, 65 0x3956C25B, 0x59F111F1, 0x923F82A4, 0xAB1C5ED5, 66 0xD807AA98, 0x12835B01, 0x243185BE, 0x550C7DC3, 67 0x72BE5D74, 0x80DEB1FE, 0x9BDC06A7, 0xC19BF174, 68 0xE49B69C1, 0xEFBE4786, 0x0FC19DC6, 0x240CA1CC, 69 0x2DE92C6F, 0x4A7484AA, 0x5CB0A9DC, 0x76F988DA, 70 0x983E5152, 0xA831C66D, 0xB00327C8, 0xBF597FC7, 71 0xC6E00BF3, 0xD5A79147, 0x06CA6351, 0x14292967, 72 0x27B70A85, 0x2E1B2138, 0x4D2C6DFC, 0x53380D13, 73 0x650A7354, 0x766A0ABB, 0x81C2C92E, 0x92722C85, 74 0xA2BFE8A1, 0xA81A664B, 0xC24B8B70, 0xC76C51A3, 75 0xD192E819, 0xD6990624, 0xF40E3585, 0x106AA070, 76 0x19A4C116, 0x1E376C08, 0x2748774C, 0x34B0BCB5, 77 0x391C0CB3, 0x4ED8AA4A, 0x5B9CCA4F, 0x682E6FF3, 78 0x748F82EE, 0x78A5636F, 0x84C87814, 0x8CC70208, 79 0x90BEFFFA, 0xA4506CEB, 0xBEF9A3F7, 0xC67178F2, 80 }; 81 82 83 static void 84 transform(uint32_t state[8], const uint32_t data[16]) 85 { 86 uint32_t W[16]; 87 uint32_t T[8]; 88 89 // Copy state[] to working vars. 90 memcpy(T, state, sizeof(T)); 91 92 // The first 16 operations unrolled 93 R0( 0); R0( 1); R0( 2); R0( 3); 94 R0( 4); R0( 5); R0( 6); R0( 7); 95 R0( 8); R0( 9); R0(10); R0(11); 96 R0(12); R0(13); R0(14); R0(15); 97 98 // The remaining 48 operations partially unrolled 99 for (unsigned int j = 16; j < 64; j += 16) { 100 R2( 0); R2( 1); R2( 2); R2( 3); 101 R2( 4); R2( 5); R2( 6); R2( 7); 102 R2( 8); R2( 9); R2(10); R2(11); 103 R2(12); R2(13); R2(14); R2(15); 104 } 105 106 // Add the working vars back into state[]. 107 state[0] += a(0); 108 state[1] += b(0); 109 state[2] += c(0); 110 state[3] += d(0); 111 state[4] += e(0); 112 state[5] += f(0); 113 state[6] += g(0); 114 state[7] += h(0); 115 } 116 117 118 static void 119 process(lzma_check_state *check) 120 { 121 transform(check->state.sha256.state, check->buffer.u32); 122 return; 123 } 124 125 126 extern void 127 lzma_sha256_init(lzma_check_state *check) 128 { 129 static const uint32_t s[8] = { 130 0x6A09E667, 0xBB67AE85, 0x3C6EF372, 0xA54FF53A, 131 0x510E527F, 0x9B05688C, 0x1F83D9AB, 0x5BE0CD19, 132 }; 133 134 memcpy(check->state.sha256.state, s, sizeof(s)); 135 check->state.sha256.size = 0; 136 137 return; 138 } 139 140 141 extern void 142 lzma_sha256_update(const uint8_t *buf, size_t size, lzma_check_state *check) 143 { 144 // Copy the input data into a properly aligned temporary buffer. 145 // This way we can be called with arbitrarily sized buffers 146 // (no need to be multiple of 64 bytes), and the code works also 147 // on architectures that don't allow unaligned memory access. 148 while (size > 0) { 149 const size_t copy_start = check->state.sha256.size & 0x3F; 150 size_t copy_size = 64 - copy_start; 151 if (copy_size > size) 152 copy_size = size; 153 154 memcpy(check->buffer.u8 + copy_start, buf, copy_size); 155 156 buf += copy_size; 157 size -= copy_size; 158 check->state.sha256.size += copy_size; 159 160 if ((check->state.sha256.size & 0x3F) == 0) 161 process(check); 162 } 163 164 return; 165 } 166 167 168 extern void 169 lzma_sha256_finish(lzma_check_state *check) 170 { 171 // Add padding as described in RFC 3174 (it describes SHA-1 but 172 // the same padding style is used for SHA-256 too). 173 size_t pos = check->state.sha256.size & 0x3F; 174 check->buffer.u8[pos++] = 0x80; 175 176 while (pos != 64 - 8) { 177 if (pos == 64) { 178 process(check); 179 pos = 0; 180 } 181 182 check->buffer.u8[pos++] = 0x00; 183 } 184 185 // Convert the message size from bytes to bits. 186 check->state.sha256.size *= 8; 187 188 check->buffer.u64[(64 - 8) / 8] = conv64be(check->state.sha256.size); 189 190 process(check); 191 192 for (size_t i = 0; i < 8; ++i) 193 check->buffer.u32[i] = conv32be(check->state.sha256.state[i]); 194 195 return; 196 } 197