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
4 Free Software Foundation, Inc.
5 This file is part of the GNU C Library.
6
7 The GNU C Library is free software; you can redistribute it and/or
8 modify it under the terms of the GNU Lesser General Public
9 License as published by the Free Software Foundation; either
10 version 2.1 of the License, or (at your option) any later version.
11
12 The GNU C Library 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 GNU
15 Lesser General Public License for more details.
16
17 You should have received a copy of the GNU Lesser General Public
18 License along with the GNU C Library; if not, write to the Free
19 Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
20 02111-1307 USA. */
21
22 /* Written by Ulrich Drepper <drepper@gnu.ai.mit.edu>, 1995. */
23
24 #include <sys/types.h>
25
26 # include <stdlib.h>
27 # include <string.h>
28
29 #include "md5.h"
30
31 /* #ifdef _LIBC */
32 # include <endian.h>
33 # if __BYTE_ORDER == __BIG_ENDIAN
34 # define WORDS_BIGENDIAN 1
35 # endif
36 /* We need to keep the namespace clean so define the MD5 function
37 protected using leading __ . */
38 # define md5_init_ctx __md5_init_ctx
39 # define md5_process_block __md5_process_block
40 # define md5_process_bytes __md5_process_bytes
41 # define md5_finish_ctx __md5_finish_ctx
42 # define md5_read_ctx __md5_read_ctx
43 # define md5_stream __md5_stream
44 # define md5_buffer __md5_buffer
45 /* #endif */
46
47 #ifdef WORDS_BIGENDIAN
48 # define SWAP(n) \
49 (((n) << 24) | (((n) & 0xff00) << 8) | (((n) >> 8) & 0xff00) | ((n) >> 24))
50 #else
51 # define SWAP(n) (n)
52 #endif
53
54 void
55 md5_process_bytes (const void *buffer, size_t len, struct md5_ctx *ctx);
56 void
57 md5_process_block (const void *buffer, size_t len, struct md5_ctx *ctx);
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 /* Initialize structure containing state of computation.
65 (RFC 1321, 3.3: Step 3) */
66 void
md5_init_ctx(ctx)67 md5_init_ctx (ctx)
68 struct md5_ctx *ctx;
69 {
70 ctx->A = 0x67452301;
71 ctx->B = 0xefcdab89;
72 ctx->C = 0x98badcfe;
73 ctx->D = 0x10325476;
74
75 ctx->total[0] = ctx->total[1] = 0;
76 ctx->buflen = 0;
77 }
78
79 /* Put result from CTX in first 16 bytes following RESBUF. The result
80 must be in little endian byte order.
81
82 IMPORTANT: On some systems it is required that RESBUF is correctly
83 aligned for a 32 bits value. */
84 void *
md5_read_ctx(ctx,resbuf)85 md5_read_ctx (ctx, resbuf)
86 const struct md5_ctx *ctx;
87 void *resbuf;
88 {
89 ((md5_uint32 *) resbuf)[0] = SWAP (ctx->A);
90 ((md5_uint32 *) resbuf)[1] = SWAP (ctx->B);
91 ((md5_uint32 *) resbuf)[2] = SWAP (ctx->C);
92 ((md5_uint32 *) resbuf)[3] = SWAP (ctx->D);
93
94 return resbuf;
95 }
96
97 /* Process the remaining bytes in the internal buffer and the usual
98 prolog according to the standard and write the result to RESBUF.
99
100 IMPORTANT: On some systems it is required that RESBUF is correctly
101 aligned for a 32 bits value. */
102 void *
md5_finish_ctx(ctx,resbuf)103 md5_finish_ctx (ctx, resbuf)
104 struct md5_ctx *ctx;
105 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
md5_stream(stream,resblock)134 md5_stream (stream, resblock)
135 FILE *stream;
136 void *resblock;
137 {
138 /* Important: BLOCKSIZE must be a multiple of 64. */
139 #define BLOCKSIZE 4096
140 struct md5_ctx ctx;
141 char buffer[BLOCKSIZE + 72];
142 size_t sum;
143
144 /* Initialize the computation context. */
145 md5_init_ctx (&ctx);
146
147 /* Iterate over full file contents. */
148 while (1)
149 {
150 /* We read the file in blocks of BLOCKSIZE bytes. One call of the
151 computation function processes the whole buffer so that with the
152 next round of the loop another block can be read. */
153 size_t n;
154 sum = 0;
155
156 /* Read block. Take care for partial reads. */
157 do
158 {
159 n = fread (buffer + sum, 1, BLOCKSIZE - sum, stream);
160
161 sum += n;
162 }
163 while (sum < BLOCKSIZE && n != 0);
164 if (n == 0 && ferror (stream))
165 return 1;
166
167 /* If end of file is reached, end the loop. */
168 if (n == 0)
169 break;
170
171 /* Process buffer with BLOCKSIZE bytes. Note that
172 BLOCKSIZE % 64 == 0
173 */
174 md5_process_block (buffer, BLOCKSIZE, &ctx);
175 }
176
177 /* Add the last bytes if necessary. */
178 if (sum > 0)
179 md5_process_bytes (buffer, sum, &ctx);
180
181 /* Construct result in desired memory. */
182 md5_finish_ctx (&ctx, resblock);
183 return 0;
184 }
185
186 /* Compute MD5 message digest for LEN bytes beginning at BUFFER. The
187 result is always in little endian byte order, so that a byte-wise
188 output yields to the wanted ASCII representation of the message
189 digest. */
190 void *
md5_buffer(buffer,len,resblock)191 md5_buffer (buffer, len, resblock)
192 const char *buffer;
193 size_t len;
194 void *resblock;
195 {
196 struct md5_ctx ctx;
197
198 /* Initialize the computation context. */
199 md5_init_ctx (&ctx);
200
201 /* Process whole buffer but last len % 64 bytes. */
202 md5_process_bytes (buffer, len, &ctx);
203
204 /* Put result in desired memory area. */
205 return md5_finish_ctx (&ctx, resblock);
206 }
207
208
209 void
md5_process_bytes(buffer,len,ctx)210 md5_process_bytes (buffer, len, ctx)
211 const void *buffer;
212 size_t len;
213 struct md5_ctx *ctx;
214 {
215 /* When we already have some bits in our internal buffer concatenate
216 both inputs first. */
217 if (ctx->buflen != 0)
218 {
219 size_t left_over = ctx->buflen;
220 size_t add = 128 - left_over > len ? len : 128 - left_over;
221
222 memcpy (&ctx->buffer[left_over], buffer, add);
223 ctx->buflen += add;
224
225 if (ctx->buflen > 64)
226 {
227 md5_process_block (ctx->buffer, ctx->buflen & ~63, ctx);
228
229 ctx->buflen &= 63;
230 /* The regions in the following copy operation cannot overlap. */
231 memcpy (ctx->buffer, &ctx->buffer[(left_over + add) & ~63],
232 ctx->buflen);
233 }
234
235 buffer = (const char *) buffer + add;
236 len -= add;
237 }
238
239 /* Process available complete blocks. */
240 if (len >= 64)
241 {
242 #if !_STRING_ARCH_unaligned
243 /* To check alignment gcc has an appropriate operator. Other
244 compilers don't. */
245 # if __GNUC__ >= 2
246 # define UNALIGNED_P(p) (((md5_uintptr) p) % __alignof__ (md5_uint32) != 0)
247 # else
248 # define UNALIGNED_P(p) (((md5_uintptr) p) % sizeof (md5_uint32) != 0)
249 # endif
250 if (UNALIGNED_P (buffer))
251 while (len > 64)
252 {
253 md5_process_block (memcpy (ctx->buffer, buffer, 64), 64, ctx);
254 buffer = (const char *) buffer + 64;
255 len -= 64;
256 }
257 else
258 #endif
259 {
260 md5_process_block (buffer, len & ~63, ctx);
261 buffer = (const char *) buffer + (len & ~63);
262 len &= 63;
263 }
264 }
265
266 /* Move remaining bytes in internal buffer. */
267 if (len > 0)
268 {
269 size_t left_over = ctx->buflen;
270
271 memcpy (&ctx->buffer[left_over], buffer, len);
272 left_over += len;
273 if (left_over >= 64)
274 {
275 md5_process_block (ctx->buffer, 64, ctx);
276 left_over -= 64;
277 memcpy (ctx->buffer, &ctx->buffer[64], left_over);
278 }
279 ctx->buflen = left_over;
280 }
281 }
282
283
284 /* These are the four functions used in the four steps of the MD5 algorithm
285 and defined in the RFC 1321. The first function is a little bit optimized
286 (as found in Colin Plumbs public domain implementation). */
287 /* #define FF(b, c, d) ((b & c) | (~b & d)) */
288 #define FF(b, c, d) (d ^ (b & (c ^ d)))
289 #define FG(b, c, d) FF (d, b, c)
290 #define FH(b, c, d) (b ^ c ^ d)
291 #define FI(b, c, d) (c ^ (b | ~d))
292
293 /* Process LEN bytes of BUFFER, accumulating context into CTX.
294 It is assumed that LEN % 64 == 0. */
295
296 void
md5_process_block(buffer,len,ctx)297 md5_process_block (buffer, len, ctx)
298 const void *buffer;
299 size_t len;
300 struct md5_ctx *ctx;
301 {
302 md5_uint32 correct_words[16];
303 const md5_uint32 *words = buffer;
304 size_t nwords = len / sizeof (md5_uint32);
305 const md5_uint32 *endp = words + nwords;
306 md5_uint32 A = ctx->A;
307 md5_uint32 B = ctx->B;
308 md5_uint32 C = ctx->C;
309 md5_uint32 D = ctx->D;
310
311 /* First increment the byte count. RFC 1321 specifies the possible
312 length of the file up to 2^64 bits. Here we only compute the
313 number of bytes. Do a double word increment. */
314 ctx->total[0] += len;
315 if (ctx->total[0] < len)
316 ++ctx->total[1];
317
318 /* Process all bytes in the buffer with 64 bytes in each round of
319 the loop. */
320 while (words < endp)
321 {
322 md5_uint32 *cwp = correct_words;
323 md5_uint32 A_save = A;
324 md5_uint32 B_save = B;
325 md5_uint32 C_save = C;
326 md5_uint32 D_save = D;
327
328 /* First round: using the given function, the context and a constant
329 the next context is computed. Because the algorithms processing
330 unit is a 32-bit word and it is determined to work on words in
331 little endian byte order we perhaps have to change the byte order
332 before the computation. To reduce the work for the next steps
333 we store the swapped words in the array CORRECT_WORDS. */
334
335 #define OP(a, b, c, d, s, T) \
336 do \
337 { \
338 a += FF (b, c, d) + (*cwp++ = SWAP (*words)) + T; \
339 ++words; \
340 CYCLIC (a, s); \
341 a += b; \
342 } \
343 while (0)
344
345 /* It is unfortunate that C does not provide an operator for
346 cyclic rotation. Hope the C compiler is smart enough. */
347 #define CYCLIC(w, s) (w = (w << s) | (w >> (32 - s)))
348
349 /* Before we start, one word to the strange constants.
350 They are defined in RFC 1321 as
351
352 T[i] = (int) (4294967296.0 * fabs (sin (i))), i=1..64
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