67c9bd0723dd78672102fb984371b1e3fd803b3c
   1/*
   2 * Based on the Mozilla SHA1 (see mozilla-sha1/sha1.c),
   3 * optimized to do word accesses rather than byte accesses,
   4 * and to avoid unnecessary copies into the context array.
   5 */
   6
   7#include <string.h>
   8#include <arpa/inet.h>
   9
  10#include "sha1.h"
  11
  12#if defined(__i386__) || defined(__x86_64__)
  13
  14/*
  15 * Force usage of rol or ror by selecting the one with the smaller constant.
  16 * It _can_ generate slightly smaller code (a constant of 1 is special), but
  17 * perhaps more importantly it's possibly faster on any uarch that does a
  18 * rotate with a loop.
  19 */
  20
  21#define SHA_ASM(op, x, n) ({ unsigned int __res; __asm__(op " %1,%0":"=r" (__res):"i" (n), "0" (x)); __res; })
  22#define SHA_ROL(x,n)    SHA_ASM("rol", x, n)
  23#define SHA_ROR(x,n)    SHA_ASM("ror", x, n)
  24
  25#else
  26
  27#define SHA_ROT(X,l,r)  (((X) << (l)) | ((X) >> (r)))
  28#define SHA_ROL(X,n)    SHA_ROT(X,n,32-(n))
  29#define SHA_ROR(X,n)    SHA_ROT(X,32-(n),n)
  30
  31#endif
  32
  33/*
  34 * If you have 32 registers or more, the compiler can (and should)
  35 * try to change the array[] accesses into registers. However, on
  36 * machines with less than ~25 registers, that won't really work,
  37 * and at least gcc will make an unholy mess of it.
  38 *
  39 * So to avoid that mess which just slows things down, we force
  40 * the stores to memory to actually happen (we might be better off
  41 * with a 'W(t)=(val);asm("":"+m" (W(t))' there instead, as
  42 * suggested by Artur Skawina - that will also make gcc unable to
  43 * try to do the silly "optimize away loads" part because it won't
  44 * see what the value will be).
  45 *
  46 * Ben Herrenschmidt reports that on PPC, the C version comes close
  47 * to the optimized asm with this (ie on PPC you don't want that
  48 * 'volatile', since there are lots of registers).
  49 *
  50 * On ARM we get the best code generation by forcing a full memory barrier
  51 * between each SHA_ROUND, otherwise gcc happily get wild with spilling and
  52 * the stack frame size simply explode and performance goes down the drain.
  53 */
  54
  55#if defined(__i386__) || defined(__x86_64__)
  56  #define setW(x, val) (*(volatile unsigned int *)&W(x) = (val))
  57#elif defined(__arm__)
  58  #define setW(x, val) do { W(x) = (val); __asm__("":::"memory"); } while (0)
  59#else
  60  #define setW(x, val) (W(x) = (val))
  61#endif
  62
  63/* This "rolls" over the 512-bit array */
  64#define W(x) (array[(x)&15])
  65
  66/*
  67 * Where do we get the source from? The first 16 iterations get it from
  68 * the input data, the next mix it from the 512-bit array.
  69 */
  70#define SHA_SRC(t) htonl(data[t])
  71#define SHA_MIX(t) SHA_ROL(W(t+13) ^ W(t+8) ^ W(t+2) ^ W(t), 1)
  72
  73#define SHA_ROUND(t, input, fn, constant, A, B, C, D, E) do { \
  74        unsigned int TEMP = input(t); setW(t, TEMP); \
  75        E += TEMP + SHA_ROL(A,5) + (fn) + (constant); \
  76        B = SHA_ROR(B, 2); } while (0)
  77
  78#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 )
  79#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 )
  80#define T_20_39(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (B^C^D) , 0x6ed9eba1, A, B, C, D, E )
  81#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 )
  82#define T_60_79(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (B^C^D) ,  0xca62c1d6, A, B, C, D, E )
  83
  84static void blk_SHA1_Block(blk_SHA_CTX *ctx, const unsigned int *data)
  85{
  86        unsigned int A,B,C,D,E;
  87        unsigned int array[16];
  88
  89        A = ctx->H[0];
  90        B = ctx->H[1];
  91        C = ctx->H[2];
  92        D = ctx->H[3];
  93        E = ctx->H[4];
  94
  95        /* Round 1 - iterations 0-16 take their input from 'data' */
  96        T_0_15( 0, A, B, C, D, E);
  97        T_0_15( 1, E, A, B, C, D);
  98        T_0_15( 2, D, E, A, B, C);
  99        T_0_15( 3, C, D, E, A, B);
 100        T_0_15( 4, B, C, D, E, A);
 101        T_0_15( 5, A, B, C, D, E);
 102        T_0_15( 6, E, A, B, C, D);
 103        T_0_15( 7, D, E, A, B, C);
 104        T_0_15( 8, C, D, E, A, B);
 105        T_0_15( 9, B, C, D, E, A);
 106        T_0_15(10, A, B, C, D, E);
 107        T_0_15(11, E, A, B, C, D);
 108        T_0_15(12, D, E, A, B, C);
 109        T_0_15(13, C, D, E, A, B);
 110        T_0_15(14, B, C, D, E, A);
 111        T_0_15(15, A, B, C, D, E);
 112
 113        /* Round 1 - tail. Input from 512-bit mixing array */
 114        T_16_19(16, E, A, B, C, D);
 115        T_16_19(17, D, E, A, B, C);
 116        T_16_19(18, C, D, E, A, B);
 117        T_16_19(19, B, C, D, E, A);
 118
 119        /* Round 2 */
 120        T_20_39(20, A, B, C, D, E);
 121        T_20_39(21, E, A, B, C, D);
 122        T_20_39(22, D, E, A, B, C);
 123        T_20_39(23, C, D, E, A, B);
 124        T_20_39(24, B, C, D, E, A);
 125        T_20_39(25, A, B, C, D, E);
 126        T_20_39(26, E, A, B, C, D);
 127        T_20_39(27, D, E, A, B, C);
 128        T_20_39(28, C, D, E, A, B);
 129        T_20_39(29, B, C, D, E, A);
 130        T_20_39(30, A, B, C, D, E);
 131        T_20_39(31, E, A, B, C, D);
 132        T_20_39(32, D, E, A, B, C);
 133        T_20_39(33, C, D, E, A, B);
 134        T_20_39(34, B, C, D, E, A);
 135        T_20_39(35, A, B, C, D, E);
 136        T_20_39(36, E, A, B, C, D);
 137        T_20_39(37, D, E, A, B, C);
 138        T_20_39(38, C, D, E, A, B);
 139        T_20_39(39, B, C, D, E, A);
 140
 141        /* Round 3 */
 142        T_40_59(40, A, B, C, D, E);
 143        T_40_59(41, E, A, B, C, D);
 144        T_40_59(42, D, E, A, B, C);
 145        T_40_59(43, C, D, E, A, B);
 146        T_40_59(44, B, C, D, E, A);
 147        T_40_59(45, A, B, C, D, E);
 148        T_40_59(46, E, A, B, C, D);
 149        T_40_59(47, D, E, A, B, C);
 150        T_40_59(48, C, D, E, A, B);
 151        T_40_59(49, B, C, D, E, A);
 152        T_40_59(50, A, B, C, D, E);
 153        T_40_59(51, E, A, B, C, D);
 154        T_40_59(52, D, E, A, B, C);
 155        T_40_59(53, C, D, E, A, B);
 156        T_40_59(54, B, C, D, E, A);
 157        T_40_59(55, A, B, C, D, E);
 158        T_40_59(56, E, A, B, C, D);
 159        T_40_59(57, D, E, A, B, C);
 160        T_40_59(58, C, D, E, A, B);
 161        T_40_59(59, B, C, D, E, A);
 162
 163        /* Round 4 */
 164        T_60_79(60, A, B, C, D, E);
 165        T_60_79(61, E, A, B, C, D);
 166        T_60_79(62, D, E, A, B, C);
 167        T_60_79(63, C, D, E, A, B);
 168        T_60_79(64, B, C, D, E, A);
 169        T_60_79(65, A, B, C, D, E);
 170        T_60_79(66, E, A, B, C, D);
 171        T_60_79(67, D, E, A, B, C);
 172        T_60_79(68, C, D, E, A, B);
 173        T_60_79(69, B, C, D, E, A);
 174        T_60_79(70, A, B, C, D, E);
 175        T_60_79(71, E, A, B, C, D);
 176        T_60_79(72, D, E, A, B, C);
 177        T_60_79(73, C, D, E, A, B);
 178        T_60_79(74, B, C, D, E, A);
 179        T_60_79(75, A, B, C, D, E);
 180        T_60_79(76, E, A, B, C, D);
 181        T_60_79(77, D, E, A, B, C);
 182        T_60_79(78, C, D, E, A, B);
 183        T_60_79(79, B, C, D, E, A);
 184
 185        ctx->H[0] += A;
 186        ctx->H[1] += B;
 187        ctx->H[2] += C;
 188        ctx->H[3] += D;
 189        ctx->H[4] += E;
 190}
 191
 192void blk_SHA1_Init(blk_SHA_CTX *ctx)
 193{
 194        ctx->size = 0;
 195
 196        /* Initialize H with the magic constants (see FIPS180 for constants) */
 197        ctx->H[0] = 0x67452301;
 198        ctx->H[1] = 0xefcdab89;
 199        ctx->H[2] = 0x98badcfe;
 200        ctx->H[3] = 0x10325476;
 201        ctx->H[4] = 0xc3d2e1f0;
 202}
 203
 204void blk_SHA1_Update(blk_SHA_CTX *ctx, const void *data, unsigned long len)
 205{
 206        int lenW = ctx->size & 63;
 207
 208        ctx->size += len;
 209
 210        /* Read the data into W and process blocks as they get full */
 211        if (lenW) {
 212                int left = 64 - lenW;
 213                if (len < left)
 214                        left = len;
 215                memcpy(lenW + (char *)ctx->W, data, left);
 216                lenW = (lenW + left) & 63;
 217                len -= left;
 218                data += left;
 219                if (lenW)
 220                        return;
 221                blk_SHA1_Block(ctx, ctx->W);
 222        }
 223        while (len >= 64) {
 224                blk_SHA1_Block(ctx, data);
 225                data += 64;
 226                len -= 64;
 227        }
 228        if (len)
 229                memcpy(ctx->W, data, len);
 230}
 231
 232void blk_SHA1_Final(unsigned char hashout[20], blk_SHA_CTX *ctx)
 233{
 234        static const unsigned char pad[64] = { 0x80 };
 235        unsigned int padlen[2];
 236        int i;
 237
 238        /* Pad with a binary 1 (ie 0x80), then zeroes, then length */
 239        padlen[0] = htonl(ctx->size >> 29);
 240        padlen[1] = htonl(ctx->size << 3);
 241
 242        i = ctx->size & 63;
 243        blk_SHA1_Update(ctx, pad, 1+ (63 & (55 - i)));
 244        blk_SHA1_Update(ctx, padlen, 8);
 245
 246        /* Output hash */
 247        for (i = 0; i < 5; i++)
 248                ((unsigned int *)hashout)[i] = htonl(ctx->H[i]);
 249}