block-sha1/sha1.c

   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  * Ben Herrenschmidt reports that on PPC, the C version comes close
  32  * to the optimized asm with this (ie on PPC you don't want that
  33  * 'volatile', since there are lots of registers).
  34  *
  35  * On ARM we get the best code generation by forcing a full memory barrier
  36  * between each SHA_ROUND, otherwise gcc happily get wild with spilling and
  37  * the stack frame size simply explode and performance goes down the drain.
  38  */
  39
  40 #if defined(__i386__) || defined(__x86_64__)
  41   #define setW(x, val) (*(volatile unsigned int *)&W(x) = (val))
  42 #elif defined(__GNUC__) && defined(__arm__)
  43   #define setW(x, val) do { W(x) = (val); __asm__("":::"memory"); } while (0)
  44 #else
  45   #define setW(x, val) (W(x) = (val))
  46 #endif
  47
  48 /* This "rolls" over the 512-bit array */
  49 #define W(x) (array[(x)&15])
  50
  51 /*
  52  * Where do we get the source from? The first 16 iterations get it from
  53  * the input data, the next mix it from the 512-bit array.
  54  */
  55 #define SHA_SRC(t) get_be32((unsigned char *) block + (t)*4)
  56 #define SHA_MIX(t) SHA_ROL(W((t)+13) ^ W((t)+8) ^ W((t)+2) ^ W(t), 1)
  57
  58 #define SHA_ROUND(t, input, fn, constant, A, B, C, D, E) do { \
  59         unsigned int TEMP = input(t); setW(t, TEMP); \
  60         E += TEMP + SHA_ROL(A,5) + (fn) + (constant); \
  61         B = SHA_ROR(B, 2); } while (0)
  62
  63 #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 )
  64 #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 )
  65 #define T_20_39(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (B^C^D) , 0x6ed9eba1, A, B, C, D, E )
  66 #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 )
  67 #define T_60_79(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (B^C^D) ,  0xca62c1d6, A, B, C, D, E )
  68
  69 static void blk_SHA1_Block(blk_SHA_CTX *ctx, const void *block)
  70 {
  71         unsigned int A,B,C,D,E;
  72         unsigned int array[16];
  73
  74         A = ctx->H[0];
  75         B = ctx->H[1];
  76         C = ctx->H[2];
  77         D = ctx->H[3];
  78         E = ctx->H[4];
  79
  80         /* Round 1 - iterations 0-16 take their input from 'block' */
  81         T_0_15( 0, A, B, C, D, E);
  82         T_0_15( 1, E, A, B, C, D);
  83         T_0_15( 2, D, E, A, B, C);
  84         T_0_15( 3, C, D, E, A, B);
  85         T_0_15( 4, B, C, D, E, A);
  86         T_0_15( 5, A, B, C, D, E);
  87         T_0_15( 6, E, A, B, C, D);
  88         T_0_15( 7, D, E, A, B, C);
  89         T_0_15( 8, C, D, E, A, B);
  90         T_0_15( 9, B, C, D, E, A);
  91         T_0_15(10, A, B, C, D, E);
  92         T_0_15(11, E, A, B, C, D);
  93         T_0_15(12, D, E, A, B, C);
  94         T_0_15(13, C, D, E, A, B);
  95         T_0_15(14, B, C, D, E, A);
  96         T_0_15(15, A, B, C, D, E);
  97
  98         /* Round 1 - tail. Input from 512-bit mixing array */
  99         T_16_19(16, E, A, B, C, D);
 100         T_16_19(17, D, E, A, B, C);
 101         T_16_19(18, C, D, E, A, B);
 102         T_16_19(19, B, C, D, E, A);
 103
 104         /* Round 2 */
 105         T_20_39(20, A, B, C, D, E);
 106         T_20_39(21, E, A, B, C, D);
 107         T_20_39(22, D, E, A, B, C);
 108         T_20_39(23, C, D, E, A, B);
 109         T_20_39(24, B, C, D, E, A);
 110         T_20_39(25, A, B, C, D, E);
 111         T_20_39(26, E, A, B, C, D);
 112         T_20_39(27, D, E, A, B, C);
 113         T_20_39(28, C, D, E, A, B);
 114         T_20_39(29, B, C, D, E, A);
 115         T_20_39(30, A, B, C, D, E);
 116         T_20_39(31, E, A, B, C, D);
 117         T_20_39(32, D, E, A, B, C);
 118         T_20_39(33, C, D, E, A, B);
 119         T_20_39(34, B, C, D, E, A);
 120         T_20_39(35, A, B, C, D, E);
 121         T_20_39(36, E, A, B, C, D);
 122         T_20_39(37, D, E, A, B, C);
 123         T_20_39(38, C, D, E, A, B);
 124         T_20_39(39, B, C, D, E, A);
 125
 126         /* Round 3 */
 127         T_40_59(40, A, B, C, D, E);
 128         T_40_59(41, E, A, B, C, D);
 129         T_40_59(42, D, E, A, B, C);
 130         T_40_59(43, C, D, E, A, B);
 131         T_40_59(44, B, C, D, E, A);
 132         T_40_59(45, A, B, C, D, E);
 133         T_40_59(46, E, A, B, C, D);
 134         T_40_59(47, D, E, A, B, C);
 135         T_40_59(48, C, D, E, A, B);
 136         T_40_59(49, B, C, D, E, A);
 137         T_40_59(50, A, B, C, D, E);
 138         T_40_59(51, E, A, B, C, D);
 139         T_40_59(52, D, E, A, B, C);
 140         T_40_59(53, C, D, E, A, B);
 141         T_40_59(54, B, C, D, E, A);
 142         T_40_59(55, A, B, C, D, E);
 143         T_40_59(56, E, A, B, C, D);
 144         T_40_59(57, D, E, A, B, C);
 145         T_40_59(58, C, D, E, A, B);
 146         T_40_59(59, B, C, D, E, A);
 147
 148         /* Round 4 */
 149         T_60_79(60, A, B, C, D, E);
 150         T_60_79(61, E, A, B, C, D);
 151         T_60_79(62, D, E, A, B, C);
 152         T_60_79(63, C, D, E, A, B);
 153         T_60_79(64, B, C, D, E, A);
 154         T_60_79(65, A, B, C, D, E);
 155         T_60_79(66, E, A, B, C, D);
 156         T_60_79(67, D, E, A, B, C);
 157         T_60_79(68, C, D, E, A, B);
 158         T_60_79(69, B, C, D, E, A);
 159         T_60_79(70, A, B, C, D, E);
 160         T_60_79(71, E, A, B, C, D);
 161         T_60_79(72, D, E, A, B, C);
 162         T_60_79(73, C, D, E, A, B);
 163         T_60_79(74, B, C, D, E, A);
 164         T_60_79(75, A, B, C, D, E);
 165         T_60_79(76, E, A, B, C, D);
 166         T_60_79(77, D, E, A, B, C);
 167         T_60_79(78, C, D, E, A, B);
 168         T_60_79(79, B, C, D, E, A);
 169
 170         ctx->H[0] += A;
 171         ctx->H[1] += B;
 172         ctx->H[2] += C;
 173         ctx->H[3] += D;
 174         ctx->H[4] += E;
 175 }
 176
 177 void blk_SHA1_Init(blk_SHA_CTX *ctx)
 178 {
 179         ctx->size = 0;
 180
 181         /* Initialize H with the magic constants (see FIPS180 for constants) */
 182         ctx->H[0] = 0x67452301;
 183         ctx->H[1] = 0xefcdab89;
 184         ctx->H[2] = 0x98badcfe;
 185         ctx->H[3] = 0x10325476;
 186         ctx->H[4] = 0xc3d2e1f0;
 187 }
 188
 189 void blk_SHA1_Update(blk_SHA_CTX *ctx, const void *data, size_t len)
 190 {
 191         unsigned int lenW = ctx->size & 63;
 192
 193         ctx->size += len;
 194
 195         /* Read the data into W and process blocks as they get full */
 196         if (lenW) {
 197                 unsigned int left = 64 - lenW;
 198                 if (len < left)
 199                         left = len;
 200                 memcpy(lenW + (char *)ctx->W, data, left);
 201                 lenW = (lenW + left) & 63;
 202                 len -= left;
 203                 data = ((const char *)data + left);
 204                 if (lenW)
 205                         return;
 206                 blk_SHA1_Block(ctx, ctx->W);
 207         }
 208         while (len >= 64) {
 209                 blk_SHA1_Block(ctx, data);
 210                 data = ((const char *)data + 64);
 211                 len -= 64;
 212         }
 213         if (len)
 214                 memcpy(ctx->W, data, len);
 215 }
 216
 217 void blk_SHA1_Final(unsigned char hashout[20], blk_SHA_CTX *ctx)
 218 {
 219         static const unsigned char pad[64] = { 0x80 };
 220         unsigned int padlen[2];
 221         int i;
 222
 223         /* Pad with a binary 1 (ie 0x80), then zeroes, then length */
 224         padlen[0] = htonl((uint32_t)(ctx->size >> 29));
 225         padlen[1] = htonl((uint32_t)(ctx->size << 3));
 226
 227         i = ctx->size & 63;
 228         blk_SHA1_Update(ctx, pad, 1 + (63 & (55 - i)));
 229         blk_SHA1_Update(ctx, padlen, 8);
 230
 231         /* Output hash */
 232         for (i = 0; i < 5; i++)
 233                 put_be32(hashout + i * 4, ctx->H[i]);
 234 }