1 /* md5.c - Functions to compute MD5 message digest of files or memory blocks 2 according to the definition of MD5 in RFC 1321 from April 1992. 3 Copyright (C) 1995, 1996, 2001, 2003, 2004, 2005 Free Software Foundation, Inc. 4 NOTE: The canonical source of this file is maintained with the GNU C 5 Library. Bugs can be reported to bug-glibc@prep.ai.mit.edu. 6 7 This program is free software; you can redistribute it and/or modify it 8 under the terms of the GNU General Public License as published by the 9 Free Software Foundation; either version 2, or (at your option) any 10 later version. 11 12 This program is distributed in the hope that it will be useful, 13 but WITHOUT ANY WARRANTY; without even the implied warranty of 14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 15 GNU General Public License for more details. 16 17 You should have received a copy of the GNU General Public License 18 along with this program; if not, write to the Free Software Foundation, 19 Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */ 20 21 /* Written by Ulrich Drepper <drepper@gnu.ai.mit.edu>, 1995. */ 22 23 #ifdef HAVE_CONFIG_H 24 # include <config.h> 25 #endif 26 27 #include "md5.h" 28 29 #include <stddef.h> 30 #include <string.h> 31 32 #if USE_UNLOCKED_IO 33 # include "unlocked-io.h" 34 #endif 35 36 #ifdef _LIBC 37 # include <endian.h> 38 # if __BYTE_ORDER == __BIG_ENDIAN 39 # define WORDS_BIGENDIAN 1 40 # endif 41 /* We need to keep the namespace clean so define the MD5 function 42 protected using leading __ . */ 43 # define md5_init_ctx __md5_init_ctx 44 # define md5_process_block __md5_process_block 45 # define md5_process_bytes __md5_process_bytes 46 # define md5_finish_ctx __md5_finish_ctx 47 # define md5_read_ctx __md5_read_ctx 48 # define md5_stream __md5_stream 49 # define md5_buffer __md5_buffer 50 #endif 51 52 #ifdef WORDS_BIGENDIAN 53 # define SWAP(n) \ 54 (((n) << 24) | (((n) & 0xff00) << 8) | (((n) >> 8) & 0xff00) | ((n) >> 24)) 55 #else 56 # define SWAP(n) (n) 57 #endif 58 59 #define BLOCKSIZE 4096 60 #if BLOCKSIZE % 64 != 0 61 # error "invalid BLOCKSIZE" 62 #endif 63 64 /* This array contains the bytes used to pad the buffer to the next 65 64-byte boundary. (RFC 1321, 3.1: Step 1) */ 66 static const unsigned char fillbuf[64] = { 0x80, 0 /* , 0, 0, ... */ }; 67 68 69 /* Initialize structure containing state of computation. 70 (RFC 1321, 3.3: Step 3) */ 71 void 72 md5_init_ctx (struct md5_ctx *ctx) 73 { 74 ctx->A = 0x67452301; 75 ctx->B = 0xefcdab89; 76 ctx->C = 0x98badcfe; 77 ctx->D = 0x10325476; 78 79 ctx->total[0] = ctx->total[1] = 0; 80 ctx->buflen = 0; 81 } 82 83 /* Put result from CTX in first 16 bytes following RESBUF. The result 84 must be in little endian byte order. 85 86 IMPORTANT: On some systems it is required that RESBUF is correctly 87 aligned for a 32 bits value. */ 88 void * 89 md5_read_ctx (const struct md5_ctx *ctx, void *resbuf) 90 { 91 ((md5_uint32 *) resbuf)[0] = SWAP (ctx->A); 92 ((md5_uint32 *) resbuf)[1] = SWAP (ctx->B); 93 ((md5_uint32 *) resbuf)[2] = SWAP (ctx->C); 94 ((md5_uint32 *) resbuf)[3] = SWAP (ctx->D); 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 md5_finish_ctx (struct md5_ctx *ctx, void *resbuf) 106 { 107 /* Take yet unprocessed bytes into account. */ 108 md5_uint32 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 *(md5_uint32 *) &ctx->buffer[bytes + pad] = SWAP (ctx->total[0] << 3); 121 *(md5_uint32 *) &ctx->buffer[bytes + pad + 4] = SWAP ((ctx->total[1] << 3) | 122 (ctx->total[0] >> 29)); 123 124 /* Process last bytes. */ 125 md5_process_block (ctx->buffer, bytes + pad + 8, ctx); 126 127 return md5_read_ctx (ctx, resbuf); 128 } 129 130 /* Compute MD5 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 md5_stream (FILE *stream, void *resblock) 135 { 136 struct md5_ctx ctx; 137 char buffer[BLOCKSIZE + 72]; 138 size_t sum; 139 140 /* Initialize the computation context. */ 141 md5_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 md5_process_block (buffer, BLOCKSIZE, &ctx); 183 } 184 185 process_partial_block:; 186 187 /* Process any remaining bytes. */ 188 if (sum > 0) 189 md5_process_bytes (buffer, sum, &ctx); 190 191 /* Construct result in desired memory. */ 192 md5_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 md5_buffer (const char *buffer, size_t len, void *resblock) 202 { 203 struct md5_ctx ctx; 204 205 /* Initialize the computation context. */ 206 md5_init_ctx (&ctx); 207 208 /* Process whole buffer but last len % 64 bytes. */ 209 md5_process_bytes (buffer, len, &ctx); 210 211 /* Put result in desired memory area. */ 212 return md5_finish_ctx (&ctx, resblock); 213 } 214 215 216 void 217 md5_process_bytes (const void *buffer, size_t len, struct md5_ctx *ctx) 218 { 219 /* When we already have some bits in our internal buffer concatenate 220 both inputs first. */ 221 if (ctx->buflen != 0) 222 { 223 size_t left_over = ctx->buflen; 224 size_t add = 128 - left_over > len ? len : 128 - left_over; 225 226 memcpy (&ctx->buffer[left_over], buffer, add); 227 ctx->buflen += add; 228 229 if (ctx->buflen > 64) 230 { 231 md5_process_block (ctx->buffer, ctx->buflen & ~63, ctx); 232 233 ctx->buflen &= 63; 234 /* The regions in the following copy operation cannot overlap. */ 235 memcpy (ctx->buffer, &ctx->buffer[(left_over + add) & ~63], 236 ctx->buflen); 237 } 238 239 buffer = (const char *) buffer + add; 240 len -= add; 241 } 242 243 /* Process available complete blocks. */ 244 if (len >= 64) 245 { 246 #if !_STRING_ARCH_unaligned 247 # define alignof(type) offsetof (struct { char c; type x; }, x) 248 # define UNALIGNED_P(p) (((size_t) p) % alignof (md5_uint32) != 0) 249 if (UNALIGNED_P (buffer)) 250 while (len > 64) 251 { 252 md5_process_block (memcpy (ctx->buffer, buffer, 64), 64, ctx); 253 buffer = (const char *) buffer + 64; 254 len -= 64; 255 } 256 else 257 #endif 258 { 259 md5_process_block (buffer, len & ~63, ctx); 260 buffer = (const char *) buffer + (len & ~63); 261 len &= 63; 262 } 263 } 264 265 /* Move remaining bytes in internal buffer. */ 266 if (len > 0) 267 { 268 size_t left_over = ctx->buflen; 269 270 memcpy (&ctx->buffer[left_over], buffer, len); 271 left_over += len; 272 if (left_over >= 64) 273 { 274 md5_process_block (ctx->buffer, 64, ctx); 275 left_over -= 64; 276 memcpy (ctx->buffer, &ctx->buffer[64], left_over); 277 } 278 ctx->buflen = left_over; 279 } 280 } 281 282 283 /* These are the four functions used in the four steps of the MD5 algorithm 284 and defined in the RFC 1321. The first function is a little bit optimized 285 (as found in Colin Plumbs public domain implementation). */ 286 /* #define FF(b, c, d) ((b & c) | (~b & d)) */ 287 #define FF(b, c, d) (d ^ (b & (c ^ d))) 288 #define FG(b, c, d) FF (d, b, c) 289 #define FH(b, c, d) (b ^ c ^ d) 290 #define FI(b, c, d) (c ^ (b | ~d)) 291 292 /* Process LEN bytes of BUFFER, accumulating context into CTX. 293 It is assumed that LEN % 64 == 0. */ 294 295 void 296 md5_process_block (const void *buffer, size_t len, struct md5_ctx *ctx) 297 { 298 md5_uint32 correct_words[16]; 299 const md5_uint32 *words = buffer; 300 size_t nwords = len / sizeof (md5_uint32); 301 const md5_uint32 *endp = words + nwords; 302 md5_uint32 A = ctx->A; 303 md5_uint32 B = ctx->B; 304 md5_uint32 C = ctx->C; 305 md5_uint32 D = ctx->D; 306 307 /* First increment the byte count. RFC 1321 specifies the possible 308 length of the file up to 2^64 bits. Here we only compute the 309 number of bytes. Do a double word increment. */ 310 ctx->total[0] += len; 311 if (ctx->total[0] < len) 312 ++ctx->total[1]; 313 314 /* Process all bytes in the buffer with 64 bytes in each round of 315 the loop. */ 316 while (words < endp) 317 { 318 md5_uint32 *cwp = correct_words; 319 md5_uint32 A_save = A; 320 md5_uint32 B_save = B; 321 md5_uint32 C_save = C; 322 md5_uint32 D_save = D; 323 324 /* First round: using the given function, the context and a constant 325 the next context is computed. Because the algorithms processing 326 unit is a 32-bit word and it is determined to work on words in 327 little endian byte order we perhaps have to change the byte order 328 before the computation. To reduce the work for the next steps 329 we store the swapped words in the array CORRECT_WORDS. */ 330 331 #define OP(a, b, c, d, s, T) \ 332 do \ 333 { \ 334 a += FF (b, c, d) + (*cwp++ = SWAP (*words)) + T; \ 335 ++words; \ 336 CYCLIC (a, s); \ 337 a += b; \ 338 } \ 339 while (0) 340 341 /* It is unfortunate that C does not provide an operator for 342 cyclic rotation. Hope the C compiler is smart enough. */ 343 #define CYCLIC(w, s) (w = (w << s) | (w >> (32 - s))) 344 345 /* Before we start, one word to the strange constants. 346 They are defined in RFC 1321 as 347 348 T[i] = (int) (4294967296.0 * fabs (sin (i))), i=1..64 349 350 Here is an equivalent invocation using Perl: 351 352 perl -e 'foreach(1..64){printf "0x%08x\n", int (4294967296 * abs (sin $_))}' 353 */ 354 355 /* Round 1. */ 356 OP (A, B, C, D, 7, 0xd76aa478); 357 OP (D, A, B, C, 12, 0xe8c7b756); 358 OP (C, D, A, B, 17, 0x242070db); 359 OP (B, C, D, A, 22, 0xc1bdceee); 360 OP (A, B, C, D, 7, 0xf57c0faf); 361 OP (D, A, B, C, 12, 0x4787c62a); 362 OP (C, D, A, B, 17, 0xa8304613); 363 OP (B, C, D, A, 22, 0xfd469501); 364 OP (A, B, C, D, 7, 0x698098d8); 365 OP (D, A, B, C, 12, 0x8b44f7af); 366 OP (C, D, A, B, 17, 0xffff5bb1); 367 OP (B, C, D, A, 22, 0x895cd7be); 368 OP (A, B, C, D, 7, 0x6b901122); 369 OP (D, A, B, C, 12, 0xfd987193); 370 OP (C, D, A, B, 17, 0xa679438e); 371 OP (B, C, D, A, 22, 0x49b40821); 372 373 /* For the second to fourth round we have the possibly swapped words 374 in CORRECT_WORDS. Redefine the macro to take an additional first 375 argument specifying the function to use. */ 376 #undef OP 377 #define OP(f, a, b, c, d, k, s, T) \ 378 do \ 379 { \ 380 a += f (b, c, d) + correct_words[k] + T; \ 381 CYCLIC (a, s); \ 382 a += b; \ 383 } \ 384 while (0) 385 386 /* Round 2. */ 387 OP (FG, A, B, C, D, 1, 5, 0xf61e2562); 388 OP (FG, D, A, B, C, 6, 9, 0xc040b340); 389 OP (FG, C, D, A, B, 11, 14, 0x265e5a51); 390 OP (FG, B, C, D, A, 0, 20, 0xe9b6c7aa); 391 OP (FG, A, B, C, D, 5, 5, 0xd62f105d); 392 OP (FG, D, A, B, C, 10, 9, 0x02441453); 393 OP (FG, C, D, A, B, 15, 14, 0xd8a1e681); 394 OP (FG, B, C, D, A, 4, 20, 0xe7d3fbc8); 395 OP (FG, A, B, C, D, 9, 5, 0x21e1cde6); 396 OP (FG, D, A, B, C, 14, 9, 0xc33707d6); 397 OP (FG, C, D, A, B, 3, 14, 0xf4d50d87); 398 OP (FG, B, C, D, A, 8, 20, 0x455a14ed); 399 OP (FG, A, B, C, D, 13, 5, 0xa9e3e905); 400 OP (FG, D, A, B, C, 2, 9, 0xfcefa3f8); 401 OP (FG, C, D, A, B, 7, 14, 0x676f02d9); 402 OP (FG, B, C, D, A, 12, 20, 0x8d2a4c8a); 403 404 /* Round 3. */ 405 OP (FH, A, B, C, D, 5, 4, 0xfffa3942); 406 OP (FH, D, A, B, C, 8, 11, 0x8771f681); 407 OP (FH, C, D, A, B, 11, 16, 0x6d9d6122); 408 OP (FH, B, C, D, A, 14, 23, 0xfde5380c); 409 OP (FH, A, B, C, D, 1, 4, 0xa4beea44); 410 OP (FH, D, A, B, C, 4, 11, 0x4bdecfa9); 411 OP (FH, C, D, A, B, 7, 16, 0xf6bb4b60); 412 OP (FH, B, C, D, A, 10, 23, 0xbebfbc70); 413 OP (FH, A, B, C, D, 13, 4, 0x289b7ec6); 414 OP (FH, D, A, B, C, 0, 11, 0xeaa127fa); 415 OP (FH, C, D, A, B, 3, 16, 0xd4ef3085); 416 OP (FH, B, C, D, A, 6, 23, 0x04881d05); 417 OP (FH, A, B, C, D, 9, 4, 0xd9d4d039); 418 OP (FH, D, A, B, C, 12, 11, 0xe6db99e5); 419 OP (FH, C, D, A, B, 15, 16, 0x1fa27cf8); 420 OP (FH, B, C, D, A, 2, 23, 0xc4ac5665); 421 422 /* Round 4. */ 423 OP (FI, A, B, C, D, 0, 6, 0xf4292244); 424 OP (FI, D, A, B, C, 7, 10, 0x432aff97); 425 OP (FI, C, D, A, B, 14, 15, 0xab9423a7); 426 OP (FI, B, C, D, A, 5, 21, 0xfc93a039); 427 OP (FI, A, B, C, D, 12, 6, 0x655b59c3); 428 OP (FI, D, A, B, C, 3, 10, 0x8f0ccc92); 429 OP (FI, C, D, A, B, 10, 15, 0xffeff47d); 430 OP (FI, B, C, D, A, 1, 21, 0x85845dd1); 431 OP (FI, A, B, C, D, 8, 6, 0x6fa87e4f); 432 OP (FI, D, A, B, C, 15, 10, 0xfe2ce6e0); 433 OP (FI, C, D, A, B, 6, 15, 0xa3014314); 434 OP (FI, B, C, D, A, 13, 21, 0x4e0811a1); 435 OP (FI, A, B, C, D, 4, 6, 0xf7537e82); 436 OP (FI, D, A, B, C, 11, 10, 0xbd3af235); 437 OP (FI, C, D, A, B, 2, 15, 0x2ad7d2bb); 438 OP (FI, B, C, D, A, 9, 21, 0xeb86d391); 439 440 /* Add the starting values of the context. */ 441 A += A_save; 442 B += B_save; 443 C += C_save; 444 D += D_save; 445 } 446 447 /* Put checksum in context given as argument. */ 448 ctx->A = A; 449 ctx->B = B; 450 ctx->C = C; 451 ctx->D = D; 452 } 453