sha1.c (13145B)
1 /* sha.c - Functions to compute SHA1 message digest of files or 2 memory blocks according to the NIST specification FIPS-180-1. 3 4 Copyright (C) 2000, 2001, 2003 Free Software Foundation, Inc. 5 6 This program is free software; you can redistribute it and/or modify it 7 under the terms of the GNU General Public License as published by the 8 Free Software Foundation; either version 2, or (at your option) any 9 later version. 10 11 This program is distributed in the hope that it will be useful, 12 but WITHOUT ANY WARRANTY; without even the implied warranty of 13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 14 GNU General Public License for more details. 15 16 You should have received a copy of the GNU General Public License 17 along with this program; if not, write to the Free Software Foundation, 18 Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ 19 20 /* Written by Scott G. Miller 21 Credits: 22 Robert Klep <robert@ilse.nl> -- Expansion function fix 23 */ 24 25 #ifdef HAVE_CONFIG_H 26 # include <config.h> 27 #endif 28 29 #include "sha1.h" 30 31 #include <sys/types.h> 32 33 #include <stdlib.h> 34 #include <string.h> 35 36 37 /* 38 Not-swap is a macro that does an endian swap on architectures that are 39 big-endian, as SHA needs some data in a little-endian format 40 */ 41 42 #ifdef WORDS_BIGENDIAN 43 # define NOTSWAP(n) (n) 44 # define SWAP(n) \ 45 (((n) << 24) | (((n) & 0xff00) << 8) | (((n) >> 8) & 0xff00) | ((n) >> 24)) 46 #else 47 # define NOTSWAP(n) \ 48 (((n) << 24) | (((n) & 0xff00) << 8) | (((n) >> 8) & 0xff00) | ((n) >> 24)) 49 # define SWAP(n) (n) 50 #endif 51 52 #define BLOCKSIZE 4096 53 /* Ensure that BLOCKSIZE is a multiple of 64. */ 54 #if BLOCKSIZE % 64 != 0 55 /* FIXME-someday (soon?): use #error instead of this kludge. */ 56 "invalid BLOCKSIZE" 57 #endif 58 59 /* This array contains the bytes used to pad the buffer to the next 60 64-byte boundary. (RFC 1321, 3.1: Step 1) */ 61 static const unsigned char fillbuf[64] = { 0x80, 0 /* , 0, 0, ... */ }; 62 63 64 /* 65 Takes a pointer to a 160 bit block of data (five 32 bit ints) and 66 intializes it to the start constants of the SHA1 algorithm. This 67 must be called before using hash in the call to sha_hash 68 */ 69 void 70 sha_init_ctx (struct sha_ctx *ctx) 71 { 72 ctx->A = 0x67452301; 73 ctx->B = 0xefcdab89; 74 ctx->C = 0x98badcfe; 75 ctx->D = 0x10325476; 76 ctx->E = 0xc3d2e1f0; 77 78 ctx->total[0] = ctx->total[1] = 0; 79 ctx->buflen = 0; 80 } 81 82 /* Put result from CTX in first 20 bytes following RESBUF. The result 83 must be in little endian byte order. 84 85 IMPORTANT: On some systems it is required that RESBUF is correctly 86 aligned for a 32 bits value. */ 87 void * 88 sha_read_ctx (const struct sha_ctx *ctx, void *resbuf) 89 { 90 ((uint32_t *) resbuf)[0] = NOTSWAP (ctx->A); 91 ((uint32_t *) resbuf)[1] = NOTSWAP (ctx->B); 92 ((uint32_t *) resbuf)[2] = NOTSWAP (ctx->C); 93 ((uint32_t *) resbuf)[3] = NOTSWAP (ctx->D); 94 ((uint32_t *) resbuf)[4] = NOTSWAP (ctx->E); 95 96 return resbuf; 97 } 98 99 /* Process the remaining bytes in the internal buffer and the usual 100 prolog according to the standard and write the result to RESBUF. 101 102 IMPORTANT: On some systems it is required that RESBUF is correctly 103 aligned for a 32 bits value. */ 104 void * 105 sha_finish_ctx (struct sha_ctx *ctx, void *resbuf) 106 { 107 /* Take yet unprocessed bytes into account. */ 108 uint32_t bytes = ctx->buflen; 109 size_t pad; 110 111 /* Now count remaining bytes. */ 112 ctx->total[0] += bytes; 113 if (ctx->total[0] < bytes) 114 ++ctx->total[1]; 115 116 pad = bytes >= 56 ? 64 + 56 - bytes : 56 - bytes; 117 memcpy (&ctx->buffer[bytes], fillbuf, pad); 118 119 /* Put the 64-bit file length in *bits* at the end of the buffer. */ 120 *(uint32_t *) &ctx->buffer[bytes + pad + 4] = NOTSWAP (ctx->total[0] << 3); 121 *(uint32_t *) &ctx->buffer[bytes + pad] = NOTSWAP ((ctx->total[1] << 3) | 122 (ctx->total[0] >> 29)); 123 124 /* Process last bytes. */ 125 sha_process_block (ctx->buffer, bytes + pad + 8, ctx); 126 127 return sha_read_ctx (ctx, resbuf); 128 } 129 130 /* Compute SHA1 message digest for bytes read from STREAM. The 131 resulting message digest number will be written into the 16 bytes 132 beginning at RESBLOCK. */ 133 int 134 sha_stream (FILE *stream, void *resblock) 135 { 136 struct sha_ctx ctx; 137 char buffer[BLOCKSIZE + 72]; 138 size_t sum; 139 140 /* Initialize the computation context. */ 141 sha_init_ctx (&ctx); 142 143 /* Iterate over full file contents. */ 144 while (1) 145 { 146 /* We read the file in blocks of BLOCKSIZE bytes. One call of the 147 computation function processes the whole buffer so that with the 148 next round of the loop another block can be read. */ 149 size_t n; 150 sum = 0; 151 152 /* Read block. Take care for partial reads. */ 153 while (1) 154 { 155 n = fread (buffer + sum, 1, BLOCKSIZE - sum, stream); 156 157 sum += n; 158 159 if (sum == BLOCKSIZE) 160 break; 161 162 if (n == 0) 163 { 164 /* Check for the error flag IFF N == 0, so that we don't 165 exit the loop after a partial read due to e.g., EAGAIN 166 or EWOULDBLOCK. */ 167 if (ferror (stream)) 168 return 1; 169 goto process_partial_block; 170 } 171 172 /* We've read at least one byte, so ignore errors. But always 173 check for EOF, since feof may be true even though N > 0. 174 Otherwise, we could end up calling fread after EOF. */ 175 if (feof (stream)) 176 goto process_partial_block; 177 } 178 179 /* Process buffer with BLOCKSIZE bytes. Note that 180 BLOCKSIZE % 64 == 0 181 */ 182 sha_process_block (buffer, BLOCKSIZE, &ctx); 183 } 184 185 process_partial_block:; 186 187 /* Process any remaining bytes. */ 188 if (sum > 0) 189 sha_process_bytes (buffer, sum, &ctx); 190 191 /* Construct result in desired memory. */ 192 sha_finish_ctx (&ctx, resblock); 193 return 0; 194 } 195 196 /* Compute MD5 message digest for LEN bytes beginning at BUFFER. The 197 result is always in little endian byte order, so that a byte-wise 198 output yields to the wanted ASCII representation of the message 199 digest. */ 200 void * 201 sha_buffer (const char *buffer, size_t len, void *resblock) 202 { 203 struct sha_ctx ctx; 204 205 /* Initialize the computation context. */ 206 sha_init_ctx (&ctx); 207 208 /* Process whole buffer but last len % 64 bytes. */ 209 sha_process_bytes (buffer, len, &ctx); 210 211 /* Put result in desired memory area. */ 212 return sha_finish_ctx (&ctx, resblock); 213 } 214 215 void 216 sha_process_bytes (const void *buffer, size_t len, struct sha_ctx *ctx) 217 { 218 /* When we already have some bits in our internal buffer concatenate 219 both inputs first. */ 220 if (ctx->buflen != 0) 221 { 222 size_t left_over = ctx->buflen; 223 size_t add = 128 - left_over > len ? len : 128 - left_over; 224 225 memcpy (&ctx->buffer[left_over], buffer, add); 226 ctx->buflen += add; 227 228 if (ctx->buflen > 64) 229 { 230 sha_process_block (ctx->buffer, ctx->buflen & ~63, ctx); 231 232 ctx->buflen &= 63; 233 /* The regions in the following copy operation cannot overlap. */ 234 memcpy (ctx->buffer, &ctx->buffer[(left_over + add) & ~63], 235 ctx->buflen); 236 } 237 238 buffer = (const char *) buffer + add; 239 len -= add; 240 } 241 242 /* Process available complete blocks. */ 243 if (len >= 64) 244 { 245 #if !_STRING_ARCH_unaligned 246 /* To check alignment gcc has an appropriate operator. Other 247 compilers don't. */ 248 # if __GNUC__ >= 2 249 # define UNALIGNED_P(p) (((uintptr_t) p) % __alignof__ (uint32_t) != 0) 250 # else 251 # define UNALIGNED_P(p) (((uintptr_t) p) % sizeof (uint32_t) != 0) 252 # endif 253 if (UNALIGNED_P (buffer)) 254 while (len > 64) 255 { 256 sha_process_block (memcpy (ctx->buffer, buffer, 64), 64, ctx); 257 buffer = (const char *) buffer + 64; 258 len -= 64; 259 } 260 else 261 #endif 262 { 263 sha_process_block (buffer, len & ~63, ctx); 264 buffer = (const char *) buffer + (len & ~63); 265 len &= 63; 266 } 267 } 268 269 /* Move remaining bytes in internal buffer. */ 270 if (len > 0) 271 { 272 size_t left_over = ctx->buflen; 273 274 memcpy (&ctx->buffer[left_over], buffer, len); 275 left_over += len; 276 if (left_over >= 64) 277 { 278 sha_process_block (ctx->buffer, 64, ctx); 279 left_over -= 64; 280 memcpy (ctx->buffer, &ctx->buffer[64], left_over); 281 } 282 ctx->buflen = left_over; 283 } 284 } 285 286 /* --- Code below is the primary difference between md5.c and sha.c --- */ 287 288 /* SHA1 round constants */ 289 #define K1 0x5a827999L 290 #define K2 0x6ed9eba1L 291 #define K3 0x8f1bbcdcL 292 #define K4 0xca62c1d6L 293 294 /* Round functions. Note that F2 is the same as F4. */ 295 #define F1(B,C,D) ( D ^ ( B & ( C ^ D ) ) ) 296 #define F2(B,C,D) (B ^ C ^ D) 297 #define F3(B,C,D) ( ( B & C ) | ( D & ( B | C ) ) ) 298 #define F4(B,C,D) (B ^ C ^ D) 299 300 /* Process LEN bytes of BUFFER, accumulating context into CTX. 301 It is assumed that LEN % 64 == 0. 302 Most of this code comes from GnuPG's cipher/sha1.c. */ 303 304 void 305 sha_process_block (const void *buffer, size_t len, struct sha_ctx *ctx) 306 { 307 const uint32_t *words = buffer; 308 size_t nwords = len / sizeof (uint32_t); 309 const uint32_t *endp = words + nwords; 310 uint32_t x[16]; 311 uint32_t a = ctx->A; 312 uint32_t b = ctx->B; 313 uint32_t c = ctx->C; 314 uint32_t d = ctx->D; 315 uint32_t e = ctx->E; 316 317 /* First increment the byte count. RFC 1321 specifies the possible 318 length of the file up to 2^64 bits. Here we only compute the 319 number of bytes. Do a double word increment. */ 320 ctx->total[0] += len; 321 if (ctx->total[0] < len) 322 ++ctx->total[1]; 323 324 #define M(I) ( tm = x[I&0x0f] ^ x[(I-14)&0x0f] \ 325 ^ x[(I-8)&0x0f] ^ x[(I-3)&0x0f] \ 326 , (x[I&0x0f] = rol(tm, 1)) ) 327 328 #define rol(x,n) ( ((x) << (n)) | ((x) >> (32-(n))) ) 329 330 #define R(A,B,C,D,E,F,K,M) do { E += rol( A, 5 ) \ 331 + F( B, C, D ) \ 332 + K \ 333 + M; \ 334 B = rol( B, 30 ); \ 335 } while(0) 336 337 while (words < endp) 338 { 339 uint32_t tm; 340 int t; 341 /* FIXME: see sha1.c for a better implementation. */ 342 for (t = 0; t < 16; t++) 343 { 344 x[t] = NOTSWAP (*words); 345 words++; 346 } 347 348 R( a, b, c, d, e, F1, K1, x[ 0] ); 349 R( e, a, b, c, d, F1, K1, x[ 1] ); 350 R( d, e, a, b, c, F1, K1, x[ 2] ); 351 R( c, d, e, a, b, F1, K1, x[ 3] ); 352 R( b, c, d, e, a, F1, K1, x[ 4] ); 353 R( a, b, c, d, e, F1, K1, x[ 5] ); 354 R( e, a, b, c, d, F1, K1, x[ 6] ); 355 R( d, e, a, b, c, F1, K1, x[ 7] ); 356 R( c, d, e, a, b, F1, K1, x[ 8] ); 357 R( b, c, d, e, a, F1, K1, x[ 9] ); 358 R( a, b, c, d, e, F1, K1, x[10] ); 359 R( e, a, b, c, d, F1, K1, x[11] ); 360 R( d, e, a, b, c, F1, K1, x[12] ); 361 R( c, d, e, a, b, F1, K1, x[13] ); 362 R( b, c, d, e, a, F1, K1, x[14] ); 363 R( a, b, c, d, e, F1, K1, x[15] ); 364 R( e, a, b, c, d, F1, K1, M(16) ); 365 R( d, e, a, b, c, F1, K1, M(17) ); 366 R( c, d, e, a, b, F1, K1, M(18) ); 367 R( b, c, d, e, a, F1, K1, M(19) ); 368 R( a, b, c, d, e, F2, K2, M(20) ); 369 R( e, a, b, c, d, F2, K2, M(21) ); 370 R( d, e, a, b, c, F2, K2, M(22) ); 371 R( c, d, e, a, b, F2, K2, M(23) ); 372 R( b, c, d, e, a, F2, K2, M(24) ); 373 R( a, b, c, d, e, F2, K2, M(25) ); 374 R( e, a, b, c, d, F2, K2, M(26) ); 375 R( d, e, a, b, c, F2, K2, M(27) ); 376 R( c, d, e, a, b, F2, K2, M(28) ); 377 R( b, c, d, e, a, F2, K2, M(29) ); 378 R( a, b, c, d, e, F2, K2, M(30) ); 379 R( e, a, b, c, d, F2, K2, M(31) ); 380 R( d, e, a, b, c, F2, K2, M(32) ); 381 R( c, d, e, a, b, F2, K2, M(33) ); 382 R( b, c, d, e, a, F2, K2, M(34) ); 383 R( a, b, c, d, e, F2, K2, M(35) ); 384 R( e, a, b, c, d, F2, K2, M(36) ); 385 R( d, e, a, b, c, F2, K2, M(37) ); 386 R( c, d, e, a, b, F2, K2, M(38) ); 387 R( b, c, d, e, a, F2, K2, M(39) ); 388 R( a, b, c, d, e, F3, K3, M(40) ); 389 R( e, a, b, c, d, F3, K3, M(41) ); 390 R( d, e, a, b, c, F3, K3, M(42) ); 391 R( c, d, e, a, b, F3, K3, M(43) ); 392 R( b, c, d, e, a, F3, K3, M(44) ); 393 R( a, b, c, d, e, F3, K3, M(45) ); 394 R( e, a, b, c, d, F3, K3, M(46) ); 395 R( d, e, a, b, c, F3, K3, M(47) ); 396 R( c, d, e, a, b, F3, K3, M(48) ); 397 R( b, c, d, e, a, F3, K3, M(49) ); 398 R( a, b, c, d, e, F3, K3, M(50) ); 399 R( e, a, b, c, d, F3, K3, M(51) ); 400 R( d, e, a, b, c, F3, K3, M(52) ); 401 R( c, d, e, a, b, F3, K3, M(53) ); 402 R( b, c, d, e, a, F3, K3, M(54) ); 403 R( a, b, c, d, e, F3, K3, M(55) ); 404 R( e, a, b, c, d, F3, K3, M(56) ); 405 R( d, e, a, b, c, F3, K3, M(57) ); 406 R( c, d, e, a, b, F3, K3, M(58) ); 407 R( b, c, d, e, a, F3, K3, M(59) ); 408 R( a, b, c, d, e, F4, K4, M(60) ); 409 R( e, a, b, c, d, F4, K4, M(61) ); 410 R( d, e, a, b, c, F4, K4, M(62) ); 411 R( c, d, e, a, b, F4, K4, M(63) ); 412 R( b, c, d, e, a, F4, K4, M(64) ); 413 R( a, b, c, d, e, F4, K4, M(65) ); 414 R( e, a, b, c, d, F4, K4, M(66) ); 415 R( d, e, a, b, c, F4, K4, M(67) ); 416 R( c, d, e, a, b, F4, K4, M(68) ); 417 R( b, c, d, e, a, F4, K4, M(69) ); 418 R( a, b, c, d, e, F4, K4, M(70) ); 419 R( e, a, b, c, d, F4, K4, M(71) ); 420 R( d, e, a, b, c, F4, K4, M(72) ); 421 R( c, d, e, a, b, F4, K4, M(73) ); 422 R( b, c, d, e, a, F4, K4, M(74) ); 423 R( a, b, c, d, e, F4, K4, M(75) ); 424 R( e, a, b, c, d, F4, K4, M(76) ); 425 R( d, e, a, b, c, F4, K4, M(77) ); 426 R( c, d, e, a, b, F4, K4, M(78) ); 427 R( b, c, d, e, a, F4, K4, M(79) ); 428 429 a = ctx->A += a; 430 b = ctx->B += b; 431 c = ctx->C += c; 432 d = ctx->D += d; 433 e = ctx->E += e; 434 } 435 }