etc/TODO: Update bidi-related items.
[emacs.git] / lib / md5.c
blobb7fad633364cf8c85c4941e589e24001be80d56f
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) 1995-1997, 1999-2001, 2005-2006, 2008-2011 Free Software
4 Foundation, Inc.
5 This file is part of the GNU C Library.
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.
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.
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. */
21 /* Written by Ulrich Drepper <drepper@gnu.ai.mit.edu>, 1995. */
23 #include <config.h>
25 #include "md5.h"
27 #include <stddef.h>
28 #include <stdlib.h>
29 #include <string.h>
30 #include <sys/types.h>
32 #if USE_UNLOCKED_IO
33 # include "unlocked-io.h"
34 #endif
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
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
59 #define BLOCKSIZE 32768
60 #if BLOCKSIZE % 64 != 0
61 # error "invalid BLOCKSIZE"
62 #endif
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, ... */ };
69 /* Initialize structure containing state of computation.
70 (RFC 1321, 3.3: Step 3) */
71 void
72 md5_init_ctx (struct md5_ctx *ctx)
74 ctx->A = 0x67452301;
75 ctx->B = 0xefcdab89;
76 ctx->C = 0x98badcfe;
77 ctx->D = 0x10325476;
79 ctx->total[0] = ctx->total[1] = 0;
80 ctx->buflen = 0;
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)
89 memcpy (cp, &v, sizeof v);
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)
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));
103 return resbuf;
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)
111 /* Take yet unprocessed bytes into account. */
112 uint32_t bytes = ctx->buflen;
113 size_t size = (bytes < 56) ? 64 / 4 : 64 * 2 / 4;
115 /* Now count remaining bytes. */
116 ctx->total[0] += bytes;
117 if (ctx->total[0] < bytes)
118 ++ctx->total[1];
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));
124 memcpy (&((char *) ctx->buffer)[bytes], fillbuf, (size - 2) * 4 - bytes);
126 /* Process last bytes. */
127 md5_process_block (ctx->buffer, size * 4, ctx);
129 return md5_read_ctx (ctx, resbuf);
132 /* Compute MD5 message digest for bytes read from STREAM. The
133 resulting message digest number will be written into the 16 bytes
134 beginning at RESBLOCK. */
136 md5_stream (FILE *stream, void *resblock)
138 struct md5_ctx ctx;
139 size_t sum;
141 char *buffer = malloc (BLOCKSIZE + 72);
142 if (!buffer)
143 return 1;
145 /* Initialize the computation context. */
146 md5_init_ctx (&ctx);
148 /* Iterate over full file contents. */
149 while (1)
151 /* We read the file in blocks of BLOCKSIZE bytes. One call of the
152 computation function processes the whole buffer so that with the
153 next round of the loop another block can be read. */
154 size_t n;
155 sum = 0;
157 /* Read block. Take care for partial reads. */
158 while (1)
160 n = fread (buffer + sum, 1, BLOCKSIZE - sum, stream);
162 sum += n;
164 if (sum == BLOCKSIZE)
165 break;
167 if (n == 0)
169 /* Check for the error flag IFF N == 0, so that we don't
170 exit the loop after a partial read due to e.g., EAGAIN
171 or EWOULDBLOCK. */
172 if (ferror (stream))
174 free (buffer);
175 return 1;
177 goto process_partial_block;
180 /* We've read at least one byte, so ignore errors. But always
181 check for EOF, since feof may be true even though N > 0.
182 Otherwise, we could end up calling fread after EOF. */
183 if (feof (stream))
184 goto process_partial_block;
187 /* Process buffer with BLOCKSIZE bytes. Note that
188 BLOCKSIZE % 64 == 0
190 md5_process_block (buffer, BLOCKSIZE, &ctx);
193 process_partial_block:
195 /* Process any remaining bytes. */
196 if (sum > 0)
197 md5_process_bytes (buffer, sum, &ctx);
199 /* Construct result in desired memory. */
200 md5_finish_ctx (&ctx, resblock);
201 free (buffer);
202 return 0;
205 /* Compute MD5 message digest for LEN bytes beginning at BUFFER. The
206 result is always in little endian byte order, so that a byte-wise
207 output yields to the wanted ASCII representation of the message
208 digest. */
209 void *
210 md5_buffer (const char *buffer, size_t len, void *resblock)
212 struct md5_ctx ctx;
214 /* Initialize the computation context. */
215 md5_init_ctx (&ctx);
217 /* Process whole buffer but last len % 64 bytes. */
218 md5_process_bytes (buffer, len, &ctx);
220 /* Put result in desired memory area. */
221 return md5_finish_ctx (&ctx, resblock);
225 void
226 md5_process_bytes (const void *buffer, size_t len, struct md5_ctx *ctx)
228 /* When we already have some bits in our internal buffer concatenate
229 both inputs first. */
230 if (ctx->buflen != 0)
232 size_t left_over = ctx->buflen;
233 size_t add = 128 - left_over > len ? len : 128 - left_over;
235 memcpy (&((char *) ctx->buffer)[left_over], buffer, add);
236 ctx->buflen += add;
238 if (ctx->buflen > 64)
240 md5_process_block (ctx->buffer, ctx->buflen & ~63, ctx);
242 ctx->buflen &= 63;
243 /* The regions in the following copy operation cannot overlap. */
244 memcpy (ctx->buffer,
245 &((char *) ctx->buffer)[(left_over + add) & ~63],
246 ctx->buflen);
249 buffer = (const char *) buffer + add;
250 len -= add;
253 /* Process available complete blocks. */
254 if (len >= 64)
256 #if !_STRING_ARCH_unaligned
257 # define alignof(type) offsetof (struct { char c; type x; }, x)
258 # define UNALIGNED_P(p) (((size_t) p) % alignof (uint32_t) != 0)
259 if (UNALIGNED_P (buffer))
260 while (len > 64)
262 md5_process_block (memcpy (ctx->buffer, buffer, 64), 64, ctx);
263 buffer = (const char *) buffer + 64;
264 len -= 64;
266 else
267 #endif
269 md5_process_block (buffer, len & ~63, ctx);
270 buffer = (const char *) buffer + (len & ~63);
271 len &= 63;
275 /* Move remaining bytes in internal buffer. */
276 if (len > 0)
278 size_t left_over = ctx->buflen;
280 memcpy (&((char *) ctx->buffer)[left_over], buffer, len);
281 left_over += len;
282 if (left_over >= 64)
284 md5_process_block (ctx->buffer, 64, ctx);
285 left_over -= 64;
286 memcpy (ctx->buffer, &ctx->buffer[16], left_over);
288 ctx->buflen = left_over;
293 /* These are the four functions used in the four steps of the MD5 algorithm
294 and defined in the RFC 1321. The first function is a little bit optimized
295 (as found in Colin Plumbs public domain implementation). */
296 /* #define FF(b, c, d) ((b & c) | (~b & d)) */
297 #define FF(b, c, d) (d ^ (b & (c ^ d)))
298 #define FG(b, c, d) FF (d, b, c)
299 #define FH(b, c, d) (b ^ c ^ d)
300 #define FI(b, c, d) (c ^ (b | ~d))
302 /* Process LEN bytes of BUFFER, accumulating context into CTX.
303 It is assumed that LEN % 64 == 0. */
305 void
306 md5_process_block (const void *buffer, size_t len, struct md5_ctx *ctx)
308 uint32_t correct_words[16];
309 const uint32_t *words = buffer;
310 size_t nwords = len / sizeof (uint32_t);
311 const uint32_t *endp = words + nwords;
312 uint32_t A = ctx->A;
313 uint32_t B = ctx->B;
314 uint32_t C = ctx->C;
315 uint32_t D = ctx->D;
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];
324 /* Process all bytes in the buffer with 64 bytes in each round of
325 the loop. */
326 while (words < endp)
328 uint32_t *cwp = correct_words;
329 uint32_t A_save = A;
330 uint32_t B_save = B;
331 uint32_t C_save = C;
332 uint32_t D_save = D;
334 /* First round: using the given function, the context and a constant
335 the next context is computed. Because the algorithms processing
336 unit is a 32-bit word and it is determined to work on words in
337 little endian byte order we perhaps have to change the byte order
338 before the computation. To reduce the work for the next steps
339 we store the swapped words in the array CORRECT_WORDS. */
341 #define OP(a, b, c, d, s, T) \
342 do \
344 a += FF (b, c, d) + (*cwp++ = SWAP (*words)) + T; \
345 ++words; \
346 CYCLIC (a, s); \
347 a += b; \
349 while (0)
351 /* It is unfortunate that C does not provide an operator for
352 cyclic rotation. Hope the C compiler is smart enough. */
353 #define CYCLIC(w, s) (w = (w << s) | (w >> (32 - s)))
355 /* Before we start, one word to the strange constants.
356 They are defined in RFC 1321 as
358 T[i] = (int) (4294967296.0 * fabs (sin (i))), i=1..64
360 Here is an equivalent invocation using Perl:
362 perl -e 'foreach(1..64){printf "0x%08x\n", int (4294967296 * abs (sin $_))}'
365 /* Round 1. */
366 OP (A, B, C, D, 7, 0xd76aa478);
367 OP (D, A, B, C, 12, 0xe8c7b756);
368 OP (C, D, A, B, 17, 0x242070db);
369 OP (B, C, D, A, 22, 0xc1bdceee);
370 OP (A, B, C, D, 7, 0xf57c0faf);
371 OP (D, A, B, C, 12, 0x4787c62a);
372 OP (C, D, A, B, 17, 0xa8304613);
373 OP (B, C, D, A, 22, 0xfd469501);
374 OP (A, B, C, D, 7, 0x698098d8);
375 OP (D, A, B, C, 12, 0x8b44f7af);
376 OP (C, D, A, B, 17, 0xffff5bb1);
377 OP (B, C, D, A, 22, 0x895cd7be);
378 OP (A, B, C, D, 7, 0x6b901122);
379 OP (D, A, B, C, 12, 0xfd987193);
380 OP (C, D, A, B, 17, 0xa679438e);
381 OP (B, C, D, A, 22, 0x49b40821);
383 /* For the second to fourth round we have the possibly swapped words
384 in CORRECT_WORDS. Redefine the macro to take an additional first
385 argument specifying the function to use. */
386 #undef OP
387 #define OP(f, a, b, c, d, k, s, T) \
388 do \
390 a += f (b, c, d) + correct_words[k] + T; \
391 CYCLIC (a, s); \
392 a += b; \
394 while (0)
396 /* Round 2. */
397 OP (FG, A, B, C, D, 1, 5, 0xf61e2562);
398 OP (FG, D, A, B, C, 6, 9, 0xc040b340);
399 OP (FG, C, D, A, B, 11, 14, 0x265e5a51);
400 OP (FG, B, C, D, A, 0, 20, 0xe9b6c7aa);
401 OP (FG, A, B, C, D, 5, 5, 0xd62f105d);
402 OP (FG, D, A, B, C, 10, 9, 0x02441453);
403 OP (FG, C, D, A, B, 15, 14, 0xd8a1e681);
404 OP (FG, B, C, D, A, 4, 20, 0xe7d3fbc8);
405 OP (FG, A, B, C, D, 9, 5, 0x21e1cde6);
406 OP (FG, D, A, B, C, 14, 9, 0xc33707d6);
407 OP (FG, C, D, A, B, 3, 14, 0xf4d50d87);
408 OP (FG, B, C, D, A, 8, 20, 0x455a14ed);
409 OP (FG, A, B, C, D, 13, 5, 0xa9e3e905);
410 OP (FG, D, A, B, C, 2, 9, 0xfcefa3f8);
411 OP (FG, C, D, A, B, 7, 14, 0x676f02d9);
412 OP (FG, B, C, D, A, 12, 20, 0x8d2a4c8a);
414 /* Round 3. */
415 OP (FH, A, B, C, D, 5, 4, 0xfffa3942);
416 OP (FH, D, A, B, C, 8, 11, 0x8771f681);
417 OP (FH, C, D, A, B, 11, 16, 0x6d9d6122);
418 OP (FH, B, C, D, A, 14, 23, 0xfde5380c);
419 OP (FH, A, B, C, D, 1, 4, 0xa4beea44);
420 OP (FH, D, A, B, C, 4, 11, 0x4bdecfa9);
421 OP (FH, C, D, A, B, 7, 16, 0xf6bb4b60);
422 OP (FH, B, C, D, A, 10, 23, 0xbebfbc70);
423 OP (FH, A, B, C, D, 13, 4, 0x289b7ec6);
424 OP (FH, D, A, B, C, 0, 11, 0xeaa127fa);
425 OP (FH, C, D, A, B, 3, 16, 0xd4ef3085);
426 OP (FH, B, C, D, A, 6, 23, 0x04881d05);
427 OP (FH, A, B, C, D, 9, 4, 0xd9d4d039);
428 OP (FH, D, A, B, C, 12, 11, 0xe6db99e5);
429 OP (FH, C, D, A, B, 15, 16, 0x1fa27cf8);
430 OP (FH, B, C, D, A, 2, 23, 0xc4ac5665);
432 /* Round 4. */
433 OP (FI, A, B, C, D, 0, 6, 0xf4292244);
434 OP (FI, D, A, B, C, 7, 10, 0x432aff97);
435 OP (FI, C, D, A, B, 14, 15, 0xab9423a7);
436 OP (FI, B, C, D, A, 5, 21, 0xfc93a039);
437 OP (FI, A, B, C, D, 12, 6, 0x655b59c3);
438 OP (FI, D, A, B, C, 3, 10, 0x8f0ccc92);
439 OP (FI, C, D, A, B, 10, 15, 0xffeff47d);
440 OP (FI, B, C, D, A, 1, 21, 0x85845dd1);
441 OP (FI, A, B, C, D, 8, 6, 0x6fa87e4f);
442 OP (FI, D, A, B, C, 15, 10, 0xfe2ce6e0);
443 OP (FI, C, D, A, B, 6, 15, 0xa3014314);
444 OP (FI, B, C, D, A, 13, 21, 0x4e0811a1);
445 OP (FI, A, B, C, D, 4, 6, 0xf7537e82);
446 OP (FI, D, A, B, C, 11, 10, 0xbd3af235);
447 OP (FI, C, D, A, B, 2, 15, 0x2ad7d2bb);
448 OP (FI, B, C, D, A, 9, 21, 0xeb86d391);
450 /* Add the starting values of the context. */
451 A += A_save;
452 B += B_save;
453 C += C_save;
454 D += D_save;
457 /* Put checksum in context given as argument. */
458 ctx->A = A;
459 ctx->B = B;
460 ctx->C = C;
461 ctx->D = D;