libextra/gl/md5.c

   1 /* 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, 1997, 1999, 2000, 2001, 2005, 2006, 2008, 2009,
   4    2010 Free Software Foundation, Inc.
   5    This file is part of the GNU C Library.
   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 3, 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 #include <config.h>
  24
  25 #include "md5.h"
  26
  27 #include <stddef.h>
  28 #include <stdlib.h>
  29 #include <string.h>
  30 #include <sys/types.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 32768
  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 /* Copy the 4 byte value from v into the memory location pointed to by *cp,
  84    If your architecture allows unaligned access this is equivalent to
  85    * (uint32_t *) cp = v  */
  86 static inline void
  87 set_uint32 (char *cp, uint32_t v)
  88 {
  89   memcpy (cp, &v, sizeof v);
  90 }
  91
  92 /* Put result from CTX in first 16 bytes following RESBUF.  The result
  93    must be in little endian byte order.  */
  94 void *
  95 md5_read_ctx (const struct md5_ctx *ctx, void *resbuf)
  96 {
  97   char *r = resbuf;
  98   set_uint32 (r + 0 * sizeof ctx->A, SWAP (ctx->A));
  99   set_uint32 (r + 1 * sizeof ctx->B, SWAP (ctx->B));
 100   set_uint32 (r + 2 * sizeof ctx->C, SWAP (ctx->C));
 101   set_uint32 (r + 3 * sizeof ctx->D, SWAP (ctx->D));
 102
 103   return resbuf;
 104 }
 105
 106 /* Process the remaining bytes in the internal buffer and the usual
 107    prolog according to the standard and write the result to RESBUF.  */
 108 void *
 109 md5_finish_ctx (struct md5_ctx *ctx, void *resbuf)
 110 {
 111   /* Take yet unprocessed bytes into account.  */
 112   uint32_t bytes = ctx->buflen;
 113   size_t size = (bytes < 56) ? 64 / 4 : 64 * 2 / 4;
 114
 115   /* Now count remaining bytes.  */
 116   ctx->total[0] += bytes;
 117   if (ctx->total[0] < bytes)
 118     ++ctx->total[1];
 119
 120   /* Put the 64-bit file length in *bits* at the end of the buffer.  */
 121   ctx->buffer[size - 2] = SWAP (ctx->total[0] << 3);
 122   ctx->buffer[size - 1] = SWAP ((ctx->total[1] << 3) | (ctx->total[0] >> 29));
 123
 124   memcpy (&((char *) ctx->buffer)[bytes], fillbuf, (size - 2) * 4 - bytes);
 125
 126   /* Process last bytes.  */
 127   md5_process_block (ctx->buffer, size * 4, ctx);
 128
 129   return md5_read_ctx (ctx, resbuf);
 130 }
 131
 132 #if 0 /* Not needed by GnuTLS, and it has a large stack frame. */
 133
 134 /* Compute MD5 message digest for bytes read from STREAM.  The
 135    resulting message digest number will be written into the 16 bytes
 136    beginning at RESBLOCK.  */
 137 int
 138 md5_stream (FILE *stream, void *resblock)
 139 {
 140   struct md5_ctx ctx;
 141   size_t sum;
 142
 143   char *buffer = malloc (BLOCKSIZE + 72);
 144   if (!buffer)
 145     return 1;
 146
 147   /* Initialize the computation context.  */
 148   md5_init_ctx (&ctx);
 149
 150   /* Iterate over full file contents.  */
 151   while (1)
 152     {
 153       /* We read the file in blocks of BLOCKSIZE bytes.  One call of the
 154          computation function processes the whole buffer so that with the
 155          next round of the loop another block can be read.  */
 156       size_t n;
 157       sum = 0;
 158
 159       /* Read block.  Take care for partial reads.  */
 160       while (1)
 161         {
 162           n = fread (buffer + sum, 1, BLOCKSIZE - sum, stream);
 163
 164           sum += n;
 165
 166           if (sum == BLOCKSIZE)
 167             break;
 168
 169           if (n == 0)
 170             {
 171               /* Check for the error flag IFF N == 0, so that we don't
 172                  exit the loop after a partial read due to e.g., EAGAIN
 173                  or EWOULDBLOCK.  */
 174               if (ferror (stream))
 175                 {
 176                   free (buffer);
 177                   return 1;
 178                 }
 179               goto process_partial_block;
 180             }
 181
 182           /* We've read at least one byte, so ignore errors.  But always
 183              check for EOF, since feof may be true even though N > 0.
 184              Otherwise, we could end up calling fread after EOF.  */
 185           if (feof (stream))
 186             goto process_partial_block;
 187         }
 188
 189       /* Process buffer with BLOCKSIZE bytes.  Note that
 190          BLOCKSIZE % 64 == 0
 191        */
 192       md5_process_block (buffer, BLOCKSIZE, &ctx);
 193     }
 194
 195 process_partial_block:
 196
 197   /* Process any remaining bytes.  */
 198   if (sum > 0)
 199     md5_process_bytes (buffer, sum, &ctx);
 200
 201   /* Construct result in desired memory.  */
 202   md5_finish_ctx (&ctx, resblock);
 203   free (buffer);
 204   return 0;
 205 }
 206 #endif
 207
 208 /* Compute MD5 message digest for LEN bytes beginning at BUFFER.  The
 209    result is always in little endian byte order, so that a byte-wise
 210    output yields to the wanted ASCII representation of the message
 211    digest.  */
 212 void *
 213 md5_buffer (const char *buffer, size_t len, void *resblock)
 214 {
 215   struct md5_ctx ctx;
 216
 217   /* Initialize the computation context.  */
 218   md5_init_ctx (&ctx);
 219
 220   /* Process whole buffer but last len % 64 bytes.  */
 221   md5_process_bytes (buffer, len, &ctx);
 222
 223   /* Put result in desired memory area.  */
 224   return md5_finish_ctx (&ctx, resblock);
 225 }
 226
 227
 228 void
 229 md5_process_bytes (const void *buffer, size_t len, struct md5_ctx *ctx)
 230 {
 231   /* When we already have some bits in our internal buffer concatenate
 232      both inputs first.  */
 233   if (ctx->buflen != 0)
 234     {
 235       size_t left_over = ctx->buflen;
 236       size_t add = 128 - left_over > len ? len : 128 - left_over;
 237
 238       memcpy (&((char *) ctx->buffer)[left_over], buffer, add);
 239       ctx->buflen += add;
 240
 241       if (ctx->buflen > 64)
 242         {
 243           md5_process_block (ctx->buffer, ctx->buflen & ~63, ctx);
 244
 245           ctx->buflen &= 63;
 246           /* The regions in the following copy operation cannot overlap.  */
 247           memcpy (ctx->buffer,
 248                   &((char *) ctx->buffer)[(left_over + add) & ~63],
 249                   ctx->buflen);
 250         }
 251
 252       buffer = (const char *) buffer + add;
 253       len -= add;
 254     }
 255
 256   /* Process available complete blocks.  */
 257   if (len >= 64)
 258     {
 259 #if !_STRING_ARCH_unaligned
 260 # define alignof(type) offsetof (struct { char c; type x; }, x)
 261 # define UNALIGNED_P(p) (((size_t) p) % alignof (uint32_t) != 0)
 262       if (UNALIGNED_P (buffer))
 263         while (len > 64)
 264           {
 265             md5_process_block (memcpy (ctx->buffer, buffer, 64), 64, ctx);
 266             buffer = (const char *) buffer + 64;
 267             len -= 64;
 268           }
 269       else
 270 #endif
 271         {
 272           md5_process_block (buffer, len & ~63, ctx);
 273           buffer = (const char *) buffer + (len & ~63);
 274           len &= 63;
 275         }
 276     }
 277
 278   /* Move remaining bytes in internal buffer.  */
 279   if (len > 0)
 280     {
 281       size_t left_over = ctx->buflen;
 282
 283       memcpy (&((char *) ctx->buffer)[left_over], buffer, len);
 284       left_over += len;
 285       if (left_over >= 64)
 286         {
 287           md5_process_block (ctx->buffer, 64, ctx);
 288           left_over -= 64;
 289           memcpy (ctx->buffer, &ctx->buffer[16], left_over);
 290         }
 291       ctx->buflen = left_over;
 292     }
 293 }
 294
 295
 296 /* These are the four functions used in the four steps of the MD5 algorithm
 297    and defined in the RFC 1321.  The first function is a little bit optimized
 298    (as found in Colin Plumbs public domain implementation).  */
 299 /* #define FF(b, c, d) ((b & c) | (~b & d)) */
 300 #define FF(b, c, d) (d ^ (b & (c ^ d)))
 301 #define FG(b, c, d) FF (d, b, c)
 302 #define FH(b, c, d) (b ^ c ^ d)
 303 #define FI(b, c, d) (c ^ (b | ~d))
 304
 305 /* Process LEN bytes of BUFFER, accumulating context into CTX.
 306    It is assumed that LEN % 64 == 0.  */
 307
 308 void
 309 md5_process_block (const void *buffer, size_t len, struct md5_ctx *ctx)
 310 {
 311   uint32_t correct_words[16];
 312   const uint32_t *words = buffer;
 313   size_t nwords = len / sizeof (uint32_t);
 314   const uint32_t *endp = words + nwords;
 315   uint32_t A = ctx->A;
 316   uint32_t B = ctx->B;
 317   uint32_t C = ctx->C;
 318   uint32_t D = ctx->D;
 319
 320   /* First increment the byte count.  RFC 1321 specifies the possible
 321      length of the file up to 2^64 bits.  Here we only compute the
 322      number of bytes.  Do a double word increment.  */
 323   ctx->total[0] += len;
 324   if (ctx->total[0] < len)
 325     ++ctx->total[1];
 326
 327   /* Process all bytes in the buffer with 64 bytes in each round of
 328      the loop.  */
 329   while (words < endp)
 330     {
 331       uint32_t *cwp = correct_words;
 332       uint32_t A_save = A;
 333       uint32_t B_save = B;
 334       uint32_t C_save = C;
 335       uint32_t D_save = D;
 336
 337       /* First round: using the given function, the context and a constant
 338          the next context is computed.  Because the algorithms processing
 339          unit is a 32-bit word and it is determined to work on words in
 340          little endian byte order we perhaps have to change the byte order
 341          before the computation.  To reduce the work for the next steps
 342          we store the swapped words in the array CORRECT_WORDS.  */
 343
 344 #define OP(a, b, c, d, s, T)                                            \
 345       do                                                                \
 346         {                                                               \
 347           a += FF (b, c, d) + (*cwp++ = SWAP (*words)) + T;             \
 348           ++words;                                                      \
 349           CYCLIC (a, s);                                                \
 350           a += b;                                                       \
 351         }                                                               \
 352       while (0)
 353
 354       /* It is unfortunate that C does not provide an operator for
 355          cyclic rotation.  Hope the C compiler is smart enough.  */
 356 #define CYCLIC(w, s) (w = (w << s) | (w >> (32 - s)))
 357
 358       /* Before we start, one word to the strange constants.
 359          They are defined in RFC 1321 as
 360
 361          T[i] = (int) (4294967296.0 * fabs (sin (i))), i=1..64
 362
 363          Here is an equivalent invocation using Perl:
 364
 365          perl -e 'foreach(1..64){printf "0x%08x\n", int (4294967296 * abs (sin $_))}'
 366        */
 367
 368       /* Round 1.  */
 369       OP (A, B, C, D, 7, 0xd76aa478);
 370       OP (D, A, B, C, 12, 0xe8c7b756);
 371       OP (C, D, A, B, 17, 0x242070db);
 372       OP (B, C, D, A, 22, 0xc1bdceee);
 373       OP (A, B, C, D, 7, 0xf57c0faf);
 374       OP (D, A, B, C, 12, 0x4787c62a);
 375       OP (C, D, A, B, 17, 0xa8304613);
 376       OP (B, C, D, A, 22, 0xfd469501);
 377       OP (A, B, C, D, 7, 0x698098d8);
 378       OP (D, A, B, C, 12, 0x8b44f7af);
 379       OP (C, D, A, B, 17, 0xffff5bb1);
 380       OP (B, C, D, A, 22, 0x895cd7be);
 381       OP (A, B, C, D, 7, 0x6b901122);
 382       OP (D, A, B, C, 12, 0xfd987193);
 383       OP (C, D, A, B, 17, 0xa679438e);
 384       OP (B, C, D, A, 22, 0x49b40821);
 385
 386       /* For the second to fourth round we have the possibly swapped words
 387          in CORRECT_WORDS.  Redefine the macro to take an additional first
 388          argument specifying the function to use.  */
 389 #undef OP
 390 #define OP(f, a, b, c, d, k, s, T)                                      \
 391       do                                                                \
 392         {                                                               \
 393           a += f (b, c, d) + correct_words[k] + T;                      \
 394           CYCLIC (a, s);                                                \
 395           a += b;                                                       \
 396         }                                                               \
 397       while (0)
 398
 399       /* Round 2.  */
 400       OP (FG, A, B, C, D, 1, 5, 0xf61e2562);
 401       OP (FG, D, A, B, C, 6, 9, 0xc040b340);
 402       OP (FG, C, D, A, B, 11, 14, 0x265e5a51);
 403       OP (FG, B, C, D, A, 0, 20, 0xe9b6c7aa);
 404       OP (FG, A, B, C, D, 5, 5, 0xd62f105d);
 405       OP (FG, D, A, B, C, 10, 9, 0x02441453);
 406       OP (FG, C, D, A, B, 15, 14, 0xd8a1e681);
 407       OP (FG, B, C, D, A, 4, 20, 0xe7d3fbc8);
 408       OP (FG, A, B, C, D, 9, 5, 0x21e1cde6);
 409       OP (FG, D, A, B, C, 14, 9, 0xc33707d6);
 410       OP (FG, C, D, A, B, 3, 14, 0xf4d50d87);
 411       OP (FG, B, C, D, A, 8, 20, 0x455a14ed);
 412       OP (FG, A, B, C, D, 13, 5, 0xa9e3e905);
 413       OP (FG, D, A, B, C, 2, 9, 0xfcefa3f8);
 414       OP (FG, C, D, A, B, 7, 14, 0x676f02d9);
 415       OP (FG, B, C, D, A, 12, 20, 0x8d2a4c8a);
 416
 417       /* Round 3.  */
 418       OP (FH, A, B, C, D, 5, 4, 0xfffa3942);
 419       OP (FH, D, A, B, C, 8, 11, 0x8771f681);
 420       OP (FH, C, D, A, B, 11, 16, 0x6d9d6122);
 421       OP (FH, B, C, D, A, 14, 23, 0xfde5380c);
 422       OP (FH, A, B, C, D, 1, 4, 0xa4beea44);
 423       OP (FH, D, A, B, C, 4, 11, 0x4bdecfa9);
 424       OP (FH, C, D, A, B, 7, 16, 0xf6bb4b60);
 425       OP (FH, B, C, D, A, 10, 23, 0xbebfbc70);
 426       OP (FH, A, B, C, D, 13, 4, 0x289b7ec6);
 427       OP (FH, D, A, B, C, 0, 11, 0xeaa127fa);
 428       OP (FH, C, D, A, B, 3, 16, 0xd4ef3085);
 429       OP (FH, B, C, D, A, 6, 23, 0x04881d05);
 430       OP (FH, A, B, C, D, 9, 4, 0xd9d4d039);
 431       OP (FH, D, A, B, C, 12, 11, 0xe6db99e5);
 432       OP (FH, C, D, A, B, 15, 16, 0x1fa27cf8);
 433       OP (FH, B, C, D, A, 2, 23, 0xc4ac5665);
 434
 435       /* Round 4.  */
 436       OP (FI, A, B, C, D, 0, 6, 0xf4292244);
 437       OP (FI, D, A, B, C, 7, 10, 0x432aff97);
 438       OP (FI, C, D, A, B, 14, 15, 0xab9423a7);
 439       OP (FI, B, C, D, A, 5, 21, 0xfc93a039);
 440       OP (FI, A, B, C, D, 12, 6, 0x655b59c3);
 441       OP (FI, D, A, B, C, 3, 10, 0x8f0ccc92);
 442       OP (FI, C, D, A, B, 10, 15, 0xffeff47d);
 443       OP (FI, B, C, D, A, 1, 21, 0x85845dd1);
 444       OP (FI, A, B, C, D, 8, 6, 0x6fa87e4f);
 445       OP (FI, D, A, B, C, 15, 10, 0xfe2ce6e0);
 446       OP (FI, C, D, A, B, 6, 15, 0xa3014314);
 447       OP (FI, B, C, D, A, 13, 21, 0x4e0811a1);
 448       OP (FI, A, B, C, D, 4, 6, 0xf7537e82);
 449       OP (FI, D, A, B, C, 11, 10, 0xbd3af235);
 450       OP (FI, C, D, A, B, 2, 15, 0x2ad7d2bb);
 451       OP (FI, B, C, D, A, 9, 21, 0xeb86d391);
 452
 453       /* Add the starting values of the context.  */
 454       A += A_save;
 455       B += B_save;
 456       C += C_save;
 457       D += D_save;
 458     }
 459
 460   /* Put checksum in context given as argument.  */
 461   ctx->A = A;
 462   ctx->B = B;
 463   ctx->C = C;
 464   ctx->D = D;
 465 }