* etc/AUTHORS: Update.
[emacs.git] / lib / md5.c
blob9dc915e2cf0f9d322f91729991a81b35e1b19237
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, 2008-2018 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, see <https://www.gnu.org/licenses/>. */
20 /* Written by Ulrich Drepper <drepper@gnu.ai.mit.edu>, 1995. */
22 #include <config.h>
24 #if HAVE_OPENSSL_MD5
25 # define GL_OPENSSL_INLINE _GL_EXTERN_INLINE
26 #endif
27 #include "md5.h"
29 #include <stdalign.h>
30 #include <stdint.h>
31 #include <stdlib.h>
32 #include <string.h>
33 #include <sys/types.h>
35 #if USE_UNLOCKED_IO
36 # include "unlocked-io.h"
37 #endif
39 #ifdef _LIBC
40 # include <endian.h>
41 # if __BYTE_ORDER == __BIG_ENDIAN
42 # define WORDS_BIGENDIAN 1
43 # endif
44 /* We need to keep the namespace clean so define the MD5 function
45 protected using leading __ . */
46 # define md5_init_ctx __md5_init_ctx
47 # define md5_process_block __md5_process_block
48 # define md5_process_bytes __md5_process_bytes
49 # define md5_finish_ctx __md5_finish_ctx
50 # define md5_read_ctx __md5_read_ctx
51 # define md5_stream __md5_stream
52 # define md5_buffer __md5_buffer
53 #endif
55 #ifdef WORDS_BIGENDIAN
56 # define SWAP(n) \
57 (((n) << 24) | (((n) & 0xff00) << 8) | (((n) >> 8) & 0xff00) | ((n) >> 24))
58 #else
59 # define SWAP(n) (n)
60 #endif
62 #define BLOCKSIZE 32768
63 #if BLOCKSIZE % 64 != 0
64 # error "invalid BLOCKSIZE"
65 #endif
67 #if ! HAVE_OPENSSL_MD5
68 /* This array contains the bytes used to pad the buffer to the next
69 64-byte boundary. (RFC 1321, 3.1: Step 1) */
70 static const unsigned char fillbuf[64] = { 0x80, 0 /* , 0, 0, ... */ };
73 /* Initialize structure containing state of computation.
74 (RFC 1321, 3.3: Step 3) */
75 void
76 md5_init_ctx (struct md5_ctx *ctx)
78 ctx->A = 0x67452301;
79 ctx->B = 0xefcdab89;
80 ctx->C = 0x98badcfe;
81 ctx->D = 0x10325476;
83 ctx->total[0] = ctx->total[1] = 0;
84 ctx->buflen = 0;
87 /* Copy the 4 byte value from v into the memory location pointed to by *cp,
88 If your architecture allows unaligned access this is equivalent to
89 * (uint32_t *) cp = v */
90 static void
91 set_uint32 (char *cp, uint32_t v)
93 memcpy (cp, &v, sizeof v);
96 /* Put result from CTX in first 16 bytes following RESBUF. The result
97 must be in little endian byte order. */
98 void *
99 md5_read_ctx (const struct md5_ctx *ctx, void *resbuf)
101 char *r = resbuf;
102 set_uint32 (r + 0 * sizeof ctx->A, SWAP (ctx->A));
103 set_uint32 (r + 1 * sizeof ctx->B, SWAP (ctx->B));
104 set_uint32 (r + 2 * sizeof ctx->C, SWAP (ctx->C));
105 set_uint32 (r + 3 * sizeof ctx->D, SWAP (ctx->D));
107 return resbuf;
110 /* Process the remaining bytes in the internal buffer and the usual
111 prolog according to the standard and write the result to RESBUF. */
112 void *
113 md5_finish_ctx (struct md5_ctx *ctx, void *resbuf)
115 /* Take yet unprocessed bytes into account. */
116 uint32_t bytes = ctx->buflen;
117 size_t size = (bytes < 56) ? 64 / 4 : 64 * 2 / 4;
119 /* Now count remaining bytes. */
120 ctx->total[0] += bytes;
121 if (ctx->total[0] < bytes)
122 ++ctx->total[1];
124 /* Put the 64-bit file length in *bits* at the end of the buffer. */
125 ctx->buffer[size - 2] = SWAP (ctx->total[0] << 3);
126 ctx->buffer[size - 1] = SWAP ((ctx->total[1] << 3) | (ctx->total[0] >> 29));
128 memcpy (&((char *) ctx->buffer)[bytes], fillbuf, (size - 2) * 4 - bytes);
130 /* Process last bytes. */
131 md5_process_block (ctx->buffer, size * 4, ctx);
133 return md5_read_ctx (ctx, resbuf);
135 #endif
137 /* Compute MD5 message digest for bytes read from STREAM. The
138 resulting message digest number will be written into the 16 bytes
139 beginning at RESBLOCK. */
141 md5_stream (FILE *stream, void *resblock)
143 struct md5_ctx ctx;
144 size_t sum;
146 char *buffer = malloc (BLOCKSIZE + 72);
147 if (!buffer)
148 return 1;
150 /* Initialize the computation context. */
151 md5_init_ctx (&ctx);
153 /* Iterate over full file contents. */
154 while (1)
156 /* We read the file in blocks of BLOCKSIZE bytes. One call of the
157 computation function processes the whole buffer so that with the
158 next round of the loop another block can be read. */
159 size_t n;
160 sum = 0;
162 /* Read block. Take care for partial reads. */
163 while (1)
165 n = fread (buffer + sum, 1, BLOCKSIZE - sum, stream);
167 sum += n;
169 if (sum == BLOCKSIZE)
170 break;
172 if (n == 0)
174 /* Check for the error flag IFF N == 0, so that we don't
175 exit the loop after a partial read due to e.g., EAGAIN
176 or EWOULDBLOCK. */
177 if (ferror (stream))
179 free (buffer);
180 return 1;
182 goto process_partial_block;
185 /* We've read at least one byte, so ignore errors. But always
186 check for EOF, since feof may be true even though N > 0.
187 Otherwise, we could end up calling fread after EOF. */
188 if (feof (stream))
189 goto process_partial_block;
192 /* Process buffer with BLOCKSIZE bytes. Note that
193 BLOCKSIZE % 64 == 0
195 md5_process_block (buffer, BLOCKSIZE, &ctx);
198 process_partial_block:
200 /* Process any remaining bytes. */
201 if (sum > 0)
202 md5_process_bytes (buffer, sum, &ctx);
204 /* Construct result in desired memory. */
205 md5_finish_ctx (&ctx, resblock);
206 free (buffer);
207 return 0;
210 #if ! HAVE_OPENSSL_MD5
211 /* Compute MD5 message digest for LEN bytes beginning at BUFFER. The
212 result is always in little endian byte order, so that a byte-wise
213 output yields to the wanted ASCII representation of the message
214 digest. */
215 void *
216 md5_buffer (const char *buffer, size_t len, void *resblock)
218 struct md5_ctx ctx;
220 /* Initialize the computation context. */
221 md5_init_ctx (&ctx);
223 /* Process whole buffer but last len % 64 bytes. */
224 md5_process_bytes (buffer, len, &ctx);
226 /* Put result in desired memory area. */
227 return md5_finish_ctx (&ctx, resblock);
231 void
232 md5_process_bytes (const void *buffer, size_t len, struct md5_ctx *ctx)
234 /* When we already have some bits in our internal buffer concatenate
235 both inputs first. */
236 if (ctx->buflen != 0)
238 size_t left_over = ctx->buflen;
239 size_t add = 128 - left_over > len ? len : 128 - left_over;
241 memcpy (&((char *) ctx->buffer)[left_over], buffer, add);
242 ctx->buflen += add;
244 if (ctx->buflen > 64)
246 md5_process_block (ctx->buffer, ctx->buflen & ~63, ctx);
248 ctx->buflen &= 63;
249 /* The regions in the following copy operation cannot overlap,
250 because ctx->buflen < 64 ≤ (left_over + add) & ~63. */
251 memcpy (ctx->buffer,
252 &((char *) ctx->buffer)[(left_over + add) & ~63],
253 ctx->buflen);
256 buffer = (const char *) buffer + add;
257 len -= add;
260 /* Process available complete blocks. */
261 if (len >= 64)
263 #if !(_STRING_ARCH_unaligned || _STRING_INLINE_unaligned)
264 # define UNALIGNED_P(p) ((uintptr_t) (p) % alignof (uint32_t) != 0)
265 if (UNALIGNED_P (buffer))
266 while (len > 64)
268 md5_process_block (memcpy (ctx->buffer, buffer, 64), 64, ctx);
269 buffer = (const char *) buffer + 64;
270 len -= 64;
272 else
273 #endif
275 md5_process_block (buffer, len & ~63, ctx);
276 buffer = (const char *) buffer + (len & ~63);
277 len &= 63;
281 /* Move remaining bytes in internal buffer. */
282 if (len > 0)
284 size_t left_over = ctx->buflen;
286 memcpy (&((char *) ctx->buffer)[left_over], buffer, len);
287 left_over += len;
288 if (left_over >= 64)
290 md5_process_block (ctx->buffer, 64, ctx);
291 left_over -= 64;
292 /* The regions in the following copy operation cannot overlap,
293 because left_over ≤ 64. */
294 memcpy (ctx->buffer, &ctx->buffer[16], left_over);
296 ctx->buflen = left_over;
301 /* These are the four functions used in the four steps of the MD5 algorithm
302 and defined in the RFC 1321. The first function is a little bit optimized
303 (as found in Colin Plumbs public domain implementation). */
304 /* #define FF(b, c, d) ((b & c) | (~b & d)) */
305 #define FF(b, c, d) (d ^ (b & (c ^ d)))
306 #define FG(b, c, d) FF (d, b, c)
307 #define FH(b, c, d) (b ^ c ^ d)
308 #define FI(b, c, d) (c ^ (b | ~d))
310 /* Process LEN bytes of BUFFER, accumulating context into CTX.
311 It is assumed that LEN % 64 == 0. */
313 void
314 md5_process_block (const void *buffer, size_t len, struct md5_ctx *ctx)
316 uint32_t correct_words[16];
317 const uint32_t *words = buffer;
318 size_t nwords = len / sizeof (uint32_t);
319 const uint32_t *endp = words + nwords;
320 uint32_t A = ctx->A;
321 uint32_t B = ctx->B;
322 uint32_t C = ctx->C;
323 uint32_t D = ctx->D;
324 uint32_t lolen = len;
326 /* First increment the byte count. RFC 1321 specifies the possible
327 length of the file up to 2^64 bits. Here we only compute the
328 number of bytes. Do a double word increment. */
329 ctx->total[0] += lolen;
330 ctx->total[1] += (len >> 31 >> 1) + (ctx->total[0] < lolen);
332 /* Process all bytes in the buffer with 64 bytes in each round of
333 the loop. */
334 while (words < endp)
336 uint32_t *cwp = correct_words;
337 uint32_t A_save = A;
338 uint32_t B_save = B;
339 uint32_t C_save = C;
340 uint32_t D_save = D;
342 /* First round: using the given function, the context and a constant
343 the next context is computed. Because the algorithms processing
344 unit is a 32-bit word and it is determined to work on words in
345 little endian byte order we perhaps have to change the byte order
346 before the computation. To reduce the work for the next steps
347 we store the swapped words in the array CORRECT_WORDS. */
349 #define OP(a, b, c, d, s, T) \
350 do \
352 a += FF (b, c, d) + (*cwp++ = SWAP (*words)) + T; \
353 ++words; \
354 CYCLIC (a, s); \
355 a += b; \
357 while (0)
359 /* It is unfortunate that C does not provide an operator for
360 cyclic rotation. Hope the C compiler is smart enough. */
361 #define CYCLIC(w, s) (w = (w << s) | (w >> (32 - s)))
363 /* Before we start, one word to the strange constants.
364 They are defined in RFC 1321 as
366 T[i] = (int) (4294967296.0 * fabs (sin (i))), i=1..64
368 Here is an equivalent invocation using Perl:
370 perl -e 'foreach(1..64){printf "0x%08x\n", int (4294967296 * abs (sin $_))}'
373 /* Round 1. */
374 OP (A, B, C, D, 7, 0xd76aa478);
375 OP (D, A, B, C, 12, 0xe8c7b756);
376 OP (C, D, A, B, 17, 0x242070db);
377 OP (B, C, D, A, 22, 0xc1bdceee);
378 OP (A, B, C, D, 7, 0xf57c0faf);
379 OP (D, A, B, C, 12, 0x4787c62a);
380 OP (C, D, A, B, 17, 0xa8304613);
381 OP (B, C, D, A, 22, 0xfd469501);
382 OP (A, B, C, D, 7, 0x698098d8);
383 OP (D, A, B, C, 12, 0x8b44f7af);
384 OP (C, D, A, B, 17, 0xffff5bb1);
385 OP (B, C, D, A, 22, 0x895cd7be);
386 OP (A, B, C, D, 7, 0x6b901122);
387 OP (D, A, B, C, 12, 0xfd987193);
388 OP (C, D, A, B, 17, 0xa679438e);
389 OP (B, C, D, A, 22, 0x49b40821);
391 /* For the second to fourth round we have the possibly swapped words
392 in CORRECT_WORDS. Redefine the macro to take an additional first
393 argument specifying the function to use. */
394 #undef OP
395 #define OP(f, a, b, c, d, k, s, T) \
396 do \
398 a += f (b, c, d) + correct_words[k] + T; \
399 CYCLIC (a, s); \
400 a += b; \
402 while (0)
404 /* Round 2. */
405 OP (FG, A, B, C, D, 1, 5, 0xf61e2562);
406 OP (FG, D, A, B, C, 6, 9, 0xc040b340);
407 OP (FG, C, D, A, B, 11, 14, 0x265e5a51);
408 OP (FG, B, C, D, A, 0, 20, 0xe9b6c7aa);
409 OP (FG, A, B, C, D, 5, 5, 0xd62f105d);
410 OP (FG, D, A, B, C, 10, 9, 0x02441453);
411 OP (FG, C, D, A, B, 15, 14, 0xd8a1e681);
412 OP (FG, B, C, D, A, 4, 20, 0xe7d3fbc8);
413 OP (FG, A, B, C, D, 9, 5, 0x21e1cde6);
414 OP (FG, D, A, B, C, 14, 9, 0xc33707d6);
415 OP (FG, C, D, A, B, 3, 14, 0xf4d50d87);
416 OP (FG, B, C, D, A, 8, 20, 0x455a14ed);
417 OP (FG, A, B, C, D, 13, 5, 0xa9e3e905);
418 OP (FG, D, A, B, C, 2, 9, 0xfcefa3f8);
419 OP (FG, C, D, A, B, 7, 14, 0x676f02d9);
420 OP (FG, B, C, D, A, 12, 20, 0x8d2a4c8a);
422 /* Round 3. */
423 OP (FH, A, B, C, D, 5, 4, 0xfffa3942);
424 OP (FH, D, A, B, C, 8, 11, 0x8771f681);
425 OP (FH, C, D, A, B, 11, 16, 0x6d9d6122);
426 OP (FH, B, C, D, A, 14, 23, 0xfde5380c);
427 OP (FH, A, B, C, D, 1, 4, 0xa4beea44);
428 OP (FH, D, A, B, C, 4, 11, 0x4bdecfa9);
429 OP (FH, C, D, A, B, 7, 16, 0xf6bb4b60);
430 OP (FH, B, C, D, A, 10, 23, 0xbebfbc70);
431 OP (FH, A, B, C, D, 13, 4, 0x289b7ec6);
432 OP (FH, D, A, B, C, 0, 11, 0xeaa127fa);
433 OP (FH, C, D, A, B, 3, 16, 0xd4ef3085);
434 OP (FH, B, C, D, A, 6, 23, 0x04881d05);
435 OP (FH, A, B, C, D, 9, 4, 0xd9d4d039);
436 OP (FH, D, A, B, C, 12, 11, 0xe6db99e5);
437 OP (FH, C, D, A, B, 15, 16, 0x1fa27cf8);
438 OP (FH, B, C, D, A, 2, 23, 0xc4ac5665);
440 /* Round 4. */
441 OP (FI, A, B, C, D, 0, 6, 0xf4292244);
442 OP (FI, D, A, B, C, 7, 10, 0x432aff97);
443 OP (FI, C, D, A, B, 14, 15, 0xab9423a7);
444 OP (FI, B, C, D, A, 5, 21, 0xfc93a039);
445 OP (FI, A, B, C, D, 12, 6, 0x655b59c3);
446 OP (FI, D, A, B, C, 3, 10, 0x8f0ccc92);
447 OP (FI, C, D, A, B, 10, 15, 0xffeff47d);
448 OP (FI, B, C, D, A, 1, 21, 0x85845dd1);
449 OP (FI, A, B, C, D, 8, 6, 0x6fa87e4f);
450 OP (FI, D, A, B, C, 15, 10, 0xfe2ce6e0);
451 OP (FI, C, D, A, B, 6, 15, 0xa3014314);
452 OP (FI, B, C, D, A, 13, 21, 0x4e0811a1);
453 OP (FI, A, B, C, D, 4, 6, 0xf7537e82);
454 OP (FI, D, A, B, C, 11, 10, 0xbd3af235);
455 OP (FI, C, D, A, B, 2, 15, 0x2ad7d2bb);
456 OP (FI, B, C, D, A, 9, 21, 0xeb86d391);
458 /* Add the starting values of the context. */
459 A += A_save;
460 B += B_save;
461 C += C_save;
462 D += D_save;
465 /* Put checksum in context given as argument. */
466 ctx->A = A;
467 ctx->B = B;
468 ctx->C = C;
469 ctx->D = D;
471 #endif
474 * Hey Emacs!
475 * Local Variables:
476 * coding: utf-8
477 * End: