sha256.c 5.3 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204
  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. // Avoid bogus warnings in transform().
  24. #if (__GNUC__ == 4 && __GNUC_MINOR__ >= 2) || __GNUC__ > 4
  25. # pragma GCC diagnostic ignored "-Wuninitialized"
  26. #endif
  27. #include "check.h"
  28. // At least on x86, GCC is able to optimize this to a rotate instruction.
  29. #define rotr_32(num, amount) ((num) >> (amount) | (num) << (32 - (amount)))
  30. #define blk0(i) (W[i] = data[i])
  31. #define blk2(i) (W[i & 15] += s1(W[(i - 2) & 15]) + W[(i - 7) & 15] \
  32. + s0(W[(i - 15) & 15]))
  33. #define Ch(x, y, z) (z ^ (x & (y ^ z)))
  34. #define Maj(x, y, z) ((x & y) | (z & (x | y)))
  35. #define a(i) T[(0 - i) & 7]
  36. #define b(i) T[(1 - i) & 7]
  37. #define c(i) T[(2 - i) & 7]
  38. #define d(i) T[(3 - i) & 7]
  39. #define e(i) T[(4 - i) & 7]
  40. #define f(i) T[(5 - i) & 7]
  41. #define g(i) T[(6 - i) & 7]
  42. #define h(i) T[(7 - i) & 7]
  43. #define R(i) \
  44. h(i) += S1(e(i)) + Ch(e(i), f(i), g(i)) + SHA256_K[i + j] \
  45. + (j ? blk2(i) : blk0(i)); \
  46. d(i) += h(i); \
  47. h(i) += S0(a(i)) + Maj(a(i), b(i), c(i))
  48. #define S0(x) (rotr_32(x, 2) ^ rotr_32(x, 13) ^ rotr_32(x, 22))
  49. #define S1(x) (rotr_32(x, 6) ^ rotr_32(x, 11) ^ rotr_32(x, 25))
  50. #define s0(x) (rotr_32(x, 7) ^ rotr_32(x, 18) ^ (x >> 3))
  51. #define s1(x) (rotr_32(x, 17) ^ rotr_32(x, 19) ^ (x >> 10))
  52. static const uint32_t SHA256_K[64] = {
  53. 0x428A2F98, 0x71374491, 0xB5C0FBCF, 0xE9B5DBA5,
  54. 0x3956C25B, 0x59F111F1, 0x923F82A4, 0xAB1C5ED5,
  55. 0xD807AA98, 0x12835B01, 0x243185BE, 0x550C7DC3,
  56. 0x72BE5D74, 0x80DEB1FE, 0x9BDC06A7, 0xC19BF174,
  57. 0xE49B69C1, 0xEFBE4786, 0x0FC19DC6, 0x240CA1CC,
  58. 0x2DE92C6F, 0x4A7484AA, 0x5CB0A9DC, 0x76F988DA,
  59. 0x983E5152, 0xA831C66D, 0xB00327C8, 0xBF597FC7,
  60. 0xC6E00BF3, 0xD5A79147, 0x06CA6351, 0x14292967,
  61. 0x27B70A85, 0x2E1B2138, 0x4D2C6DFC, 0x53380D13,
  62. 0x650A7354, 0x766A0ABB, 0x81C2C92E, 0x92722C85,
  63. 0xA2BFE8A1, 0xA81A664B, 0xC24B8B70, 0xC76C51A3,
  64. 0xD192E819, 0xD6990624, 0xF40E3585, 0x106AA070,
  65. 0x19A4C116, 0x1E376C08, 0x2748774C, 0x34B0BCB5,
  66. 0x391C0CB3, 0x4ED8AA4A, 0x5B9CCA4F, 0x682E6FF3,
  67. 0x748F82EE, 0x78A5636F, 0x84C87814, 0x8CC70208,
  68. 0x90BEFFFA, 0xA4506CEB, 0xBEF9A3F7, 0xC67178F2,
  69. };
  70. static void
  71. transform(uint32_t state[8], const uint32_t data[16])
  72. {
  73. uint32_t W[16];
  74. uint32_t T[8];
  75. unsigned int j;
  76. // Copy state[] to working vars.
  77. memcpy(T, state, sizeof(T));
  78. // 64 operations, partially loop unrolled
  79. for (j = 0; j < 64; j += 16) {
  80. R( 0); R( 1); R( 2); R( 3);
  81. R( 4); R( 5); R( 6); R( 7);
  82. R( 8); R( 9); R(10); R(11);
  83. R(12); R(13); R(14); R(15);
  84. }
  85. // Add the working vars back into state[].
  86. state[0] += a(0);
  87. state[1] += b(0);
  88. state[2] += c(0);
  89. state[3] += d(0);
  90. state[4] += e(0);
  91. state[5] += f(0);
  92. state[6] += g(0);
  93. state[7] += h(0);
  94. }
  95. static void
  96. process(lzma_check_state *check)
  97. {
  98. #ifdef WORDS_BIGENDIAN
  99. transform(check->state.sha256.state, check->buffer.u32);
  100. #else
  101. uint32_t data[16];
  102. size_t i;
  103. for (i = 0; i < 16; ++i)
  104. data[i] = bswap32(check->buffer.u32[i]);
  105. transform(check->state.sha256.state, data);
  106. #endif
  107. return;
  108. }
  109. extern void
  110. lzma_sha256_init(lzma_check_state *check)
  111. {
  112. static const uint32_t s[8] = {
  113. 0x6A09E667, 0xBB67AE85, 0x3C6EF372, 0xA54FF53A,
  114. 0x510E527F, 0x9B05688C, 0x1F83D9AB, 0x5BE0CD19,
  115. };
  116. memcpy(check->state.sha256.state, s, sizeof(s));
  117. check->state.sha256.size = 0;
  118. return;
  119. }
  120. extern void
  121. lzma_sha256_update(const uint8_t *buf, size_t size, lzma_check_state *check)
  122. {
  123. // Copy the input data into a properly aligned temporary buffer.
  124. // This way we can be called with arbitrarily sized buffers
  125. // (no need to be multiple of 64 bytes), and the code works also
  126. // on architectures that don't allow unaligned memory access.
  127. while (size > 0) {
  128. const size_t copy_start = check->state.sha256.size & 0x3F;
  129. size_t copy_size = 64 - copy_start;
  130. if (copy_size > size)
  131. copy_size = size;
  132. memcpy(check->buffer.u8 + copy_start, buf, copy_size);
  133. buf += copy_size;
  134. size -= copy_size;
  135. check->state.sha256.size += copy_size;
  136. if ((check->state.sha256.size & 0x3F) == 0)
  137. process(check);
  138. }
  139. return;
  140. }
  141. extern void
  142. lzma_sha256_finish(lzma_check_state *check)
  143. {
  144. size_t i;
  145. // Add padding as described in RFC 3174 (it describes SHA-1 but
  146. // the same padding style is used for SHA-256 too).
  147. size_t pos = check->state.sha256.size & 0x3F;
  148. check->buffer.u8[pos++] = 0x80;
  149. while (pos != 64 - 8) {
  150. if (pos == 64) {
  151. process(check);
  152. pos = 0;
  153. }
  154. check->buffer.u8[pos++] = 0x00;
  155. }
  156. // Convert the message size from bytes to bits.
  157. check->state.sha256.size *= 8;
  158. check->buffer.u64[(64 - 8) / 8] = conv64be(check->state.sha256.size);
  159. process(check);
  160. for (i = 0; i < 8; ++i)
  161. check->buffer.u32[i] = conv32be(check->state.sha256.state[i]);
  162. return;
  163. }