2002-09-02 Jakub Jelinek <jakub@redhat.com>
[glibc.git] / crypt / md5.c
blob04bce5ab378539c033e1b3ee62ddbbfae33e0816
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 Free Software Foundation, Inc.
4 This file is part of the GNU C Library.
6 The GNU C Library is free software; you can redistribute it and/or
7 modify it under the terms of the GNU Lesser General Public
8 License as published by the Free Software Foundation; either
9 version 2.1 of the License, or (at your option) any later version.
11 The GNU C Library is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
16 You should have received a copy of the GNU Lesser General Public
17 License along with the GNU C Library; if not, write to the Free
18 Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
19 02111-1307 USA. */
21 /* Written by Ulrich Drepper <drepper@gnu.ai.mit.edu>, 1995. */
23 #ifdef HAVE_CONFIG_H
24 # include <config.h>
25 #endif
27 #include <sys/types.h>
29 #if STDC_HEADERS || defined _LIBC
30 # include <stdlib.h>
31 # include <string.h>
32 #else
33 # ifndef HAVE_MEMCPY
34 # define memcpy(d, s, n) bcopy ((s), (d), (n))
35 # endif
36 #endif
38 #include "md5.h"
40 #ifdef _LIBC
41 # include <endian.h>
42 # if __BYTE_ORDER == __BIG_ENDIAN
43 # define WORDS_BIGENDIAN 1
44 # endif
45 /* We need to keep the namespace clean so define the MD5 function
46 protected using leading __ . */
47 # define md5_init_ctx __md5_init_ctx
48 # define md5_process_block __md5_process_block
49 # define md5_process_bytes __md5_process_bytes
50 # define md5_finish_ctx __md5_finish_ctx
51 # define md5_read_ctx __md5_read_ctx
52 # define md5_stream __md5_stream
53 # define md5_buffer __md5_buffer
54 #endif
56 #ifdef WORDS_BIGENDIAN
57 # define SWAP(n) \
58 (((n) << 24) | (((n) & 0xff00) << 8) | (((n) >> 8) & 0xff00) | ((n) >> 24))
59 #else
60 # define SWAP(n) (n)
61 #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 (ctx)
73 struct md5_ctx *ctx;
75 ctx->A = 0x67452301;
76 ctx->B = 0xefcdab89;
77 ctx->C = 0x98badcfe;
78 ctx->D = 0x10325476;
80 ctx->total[0] = ctx->total[1] = 0;
81 ctx->buflen = 0;
84 /* Put result from CTX in first 16 bytes following RESBUF. The result
85 must be in little endian byte order.
87 IMPORTANT: On some systems it is required that RESBUF is correctly
88 aligned for a 32 bits value. */
89 void *
90 md5_read_ctx (ctx, resbuf)
91 const struct md5_ctx *ctx;
92 void *resbuf;
94 ((md5_uint32 *) resbuf)[0] = SWAP (ctx->A);
95 ((md5_uint32 *) resbuf)[1] = SWAP (ctx->B);
96 ((md5_uint32 *) resbuf)[2] = SWAP (ctx->C);
97 ((md5_uint32 *) resbuf)[3] = SWAP (ctx->D);
99 return resbuf;
102 /* Process the remaining bytes in the internal buffer and the usual
103 prolog according to the standard and write the result to RESBUF.
105 IMPORTANT: On some systems it is required that RESBUF is correctly
106 aligned for a 32 bits value. */
107 void *
108 md5_finish_ctx (ctx, resbuf)
109 struct md5_ctx *ctx;
110 void *resbuf;
112 /* Take yet unprocessed bytes into account. */
113 md5_uint32 bytes = ctx->buflen;
114 size_t pad;
116 /* Now count remaining bytes. */
117 ctx->total[0] += bytes;
118 if (ctx->total[0] < bytes)
119 ++ctx->total[1];
121 pad = bytes >= 56 ? 64 + 56 - bytes : 56 - bytes;
122 memcpy (&ctx->buffer[bytes], fillbuf, pad);
124 /* Put the 64-bit file length in *bits* at the end of the buffer. */
125 *(md5_uint32 *) &ctx->buffer[bytes + pad] = SWAP (ctx->total[0] << 3);
126 *(md5_uint32 *) &ctx->buffer[bytes + pad + 4] = SWAP ((ctx->total[1] << 3) |
127 (ctx->total[0] >> 29));
129 /* Process last bytes. */
130 md5_process_block (ctx->buffer, bytes + pad + 8, ctx);
132 return md5_read_ctx (ctx, resbuf);
135 /* Compute MD5 message digest for bytes read from STREAM. The
136 resulting message digest number will be written into the 16 bytes
137 beginning at RESBLOCK. */
139 md5_stream (stream, resblock)
140 FILE *stream;
141 void *resblock;
143 /* Important: BLOCKSIZE must be a multiple of 64. */
144 #define BLOCKSIZE 4096
145 struct md5_ctx ctx;
146 char buffer[BLOCKSIZE + 72];
147 size_t sum;
149 /* Initialize the computation context. */
150 md5_init_ctx (&ctx);
152 /* Iterate over full file contents. */
153 while (1)
155 /* We read the file in blocks of BLOCKSIZE bytes. One call of the
156 computation function processes the whole buffer so that with the
157 next round of the loop another block can be read. */
158 size_t n;
159 sum = 0;
161 /* Read block. Take care for partial reads. */
164 n = fread (buffer + sum, 1, BLOCKSIZE - sum, stream);
166 sum += n;
168 while (sum < BLOCKSIZE && n != 0);
169 if (n == 0 && ferror (stream))
170 return 1;
172 /* If end of file is reached, end the loop. */
173 if (n == 0)
174 break;
176 /* Process buffer with BLOCKSIZE bytes. Note that
177 BLOCKSIZE % 64 == 0
179 md5_process_block (buffer, BLOCKSIZE, &ctx);
182 /* Add the last bytes if necessary. */
183 if (sum > 0)
184 md5_process_bytes (buffer, sum, &ctx);
186 /* Construct result in desired memory. */
187 md5_finish_ctx (&ctx, resblock);
188 return 0;
191 /* Compute MD5 message digest for LEN bytes beginning at BUFFER. The
192 result is always in little endian byte order, so that a byte-wise
193 output yields to the wanted ASCII representation of the message
194 digest. */
195 void *
196 md5_buffer (buffer, len, resblock)
197 const char *buffer;
198 size_t len;
199 void *resblock;
201 struct md5_ctx ctx;
203 /* Initialize the computation context. */
204 md5_init_ctx (&ctx);
206 /* Process whole buffer but last len % 64 bytes. */
207 md5_process_bytes (buffer, len, &ctx);
209 /* Put result in desired memory area. */
210 return md5_finish_ctx (&ctx, resblock);
214 void
215 md5_process_bytes (buffer, len, ctx)
216 const void *buffer;
217 size_t len;
218 struct md5_ctx *ctx;
220 /* When we already have some bits in our internal buffer concatenate
221 both inputs first. */
222 if (ctx->buflen != 0)
224 size_t left_over = ctx->buflen;
225 size_t add = 128 - left_over > len ? len : 128 - left_over;
227 memcpy (&ctx->buffer[left_over], buffer, add);
228 ctx->buflen += add;
230 if (ctx->buflen > 64)
232 md5_process_block (ctx->buffer, ctx->buflen & ~63, ctx);
234 ctx->buflen &= 63;
235 /* The regions in the following copy operation cannot overlap. */
236 memcpy (ctx->buffer, &ctx->buffer[(left_over + add) & ~63],
237 ctx->buflen);
240 buffer = (const char *) buffer + add;
241 len -= add;
244 /* Process available complete blocks. */
245 if (len >= 64)
247 #if !_STRING_ARCH_unaligned
248 /* To check alignment gcc has an appropriate operator. Other
249 compilers don't. */
250 # if __GNUC__ >= 2
251 # define UNALIGNED_P(p) (((md5_uintptr) p) % __alignof__ (md5_uint32) != 0)
252 # else
253 # define UNALIGNED_P(p) (((md5_uintptr) p) % sizeof (md5_uint32) != 0)
254 # endif
255 if (UNALIGNED_P (buffer))
256 while (len > 64)
258 md5_process_block (memcpy (ctx->buffer, buffer, 64), 64, ctx);
259 buffer = (const char *) buffer + 64;
260 len -= 64;
262 else
263 #endif
265 md5_process_block (buffer, len & ~63, ctx);
266 buffer = (const char *) buffer + (len & ~63);
267 len &= 63;
271 /* Move remaining bytes in internal buffer. */
272 if (len > 0)
274 size_t left_over = ctx->buflen;
276 memcpy (&ctx->buffer[left_over], buffer, len);
277 left_over += len;
278 if (left_over >= 64)
280 md5_process_block (ctx->buffer, 64, ctx);
281 left_over -= 64;
282 memcpy (ctx->buffer, &ctx->buffer[64], left_over);
284 ctx->buflen = left_over;
289 /* These are the four functions used in the four steps of the MD5 algorithm
290 and defined in the RFC 1321. The first function is a little bit optimized
291 (as found in Colin Plumbs public domain implementation). */
292 /* #define FF(b, c, d) ((b & c) | (~b & d)) */
293 #define FF(b, c, d) (d ^ (b & (c ^ d)))
294 #define FG(b, c, d) FF (d, b, c)
295 #define FH(b, c, d) (b ^ c ^ d)
296 #define FI(b, c, d) (c ^ (b | ~d))
298 /* Process LEN bytes of BUFFER, accumulating context into CTX.
299 It is assumed that LEN % 64 == 0. */
301 void
302 md5_process_block (buffer, len, ctx)
303 const void *buffer;
304 size_t len;
305 struct md5_ctx *ctx;
307 md5_uint32 correct_words[16];
308 const md5_uint32 *words = buffer;
309 size_t nwords = len / sizeof (md5_uint32);
310 const md5_uint32 *endp = words + nwords;
311 md5_uint32 A = ctx->A;
312 md5_uint32 B = ctx->B;
313 md5_uint32 C = ctx->C;
314 md5_uint32 D = ctx->D;
316 /* First increment the byte count. RFC 1321 specifies the possible
317 length of the file up to 2^64 bits. Here we only compute the
318 number of bytes. Do a double word increment. */
319 ctx->total[0] += len;
320 if (ctx->total[0] < len)
321 ++ctx->total[1];
323 /* Process all bytes in the buffer with 64 bytes in each round of
324 the loop. */
325 while (words < endp)
327 md5_uint32 *cwp = correct_words;
328 md5_uint32 A_save = A;
329 md5_uint32 B_save = B;
330 md5_uint32 C_save = C;
331 md5_uint32 D_save = D;
333 /* First round: using the given function, the context and a constant
334 the next context is computed. Because the algorithms processing
335 unit is a 32-bit word and it is determined to work on words in
336 little endian byte order we perhaps have to change the byte order
337 before the computation. To reduce the work for the next steps
338 we store the swapped words in the array CORRECT_WORDS. */
340 #define OP(a, b, c, d, s, T) \
341 do \
343 a += FF (b, c, d) + (*cwp++ = SWAP (*words)) + T; \
344 ++words; \
345 CYCLIC (a, s); \
346 a += b; \
348 while (0)
350 /* It is unfortunate that C does not provide an operator for
351 cyclic rotation. Hope the C compiler is smart enough. */
352 #define CYCLIC(w, s) (w = (w << s) | (w >> (32 - s)))
354 /* Before we start, one word to the strange constants.
355 They are defined in RFC 1321 as
357 T[i] = (int) (4294967296.0 * fabs (sin (i))), i=1..64
360 /* Round 1. */
361 OP (A, B, C, D, 7, 0xd76aa478);
362 OP (D, A, B, C, 12, 0xe8c7b756);
363 OP (C, D, A, B, 17, 0x242070db);
364 OP (B, C, D, A, 22, 0xc1bdceee);
365 OP (A, B, C, D, 7, 0xf57c0faf);
366 OP (D, A, B, C, 12, 0x4787c62a);
367 OP (C, D, A, B, 17, 0xa8304613);
368 OP (B, C, D, A, 22, 0xfd469501);
369 OP (A, B, C, D, 7, 0x698098d8);
370 OP (D, A, B, C, 12, 0x8b44f7af);
371 OP (C, D, A, B, 17, 0xffff5bb1);
372 OP (B, C, D, A, 22, 0x895cd7be);
373 OP (A, B, C, D, 7, 0x6b901122);
374 OP (D, A, B, C, 12, 0xfd987193);
375 OP (C, D, A, B, 17, 0xa679438e);
376 OP (B, C, D, A, 22, 0x49b40821);
378 /* For the second to fourth round we have the possibly swapped words
379 in CORRECT_WORDS. Redefine the macro to take an additional first
380 argument specifying the function to use. */
381 #undef OP
382 #define OP(f, a, b, c, d, k, s, T) \
383 do \
385 a += f (b, c, d) + correct_words[k] + T; \
386 CYCLIC (a, s); \
387 a += b; \
389 while (0)
391 /* Round 2. */
392 OP (FG, A, B, C, D, 1, 5, 0xf61e2562);
393 OP (FG, D, A, B, C, 6, 9, 0xc040b340);
394 OP (FG, C, D, A, B, 11, 14, 0x265e5a51);
395 OP (FG, B, C, D, A, 0, 20, 0xe9b6c7aa);
396 OP (FG, A, B, C, D, 5, 5, 0xd62f105d);
397 OP (FG, D, A, B, C, 10, 9, 0x02441453);
398 OP (FG, C, D, A, B, 15, 14, 0xd8a1e681);
399 OP (FG, B, C, D, A, 4, 20, 0xe7d3fbc8);
400 OP (FG, A, B, C, D, 9, 5, 0x21e1cde6);
401 OP (FG, D, A, B, C, 14, 9, 0xc33707d6);
402 OP (FG, C, D, A, B, 3, 14, 0xf4d50d87);
403 OP (FG, B, C, D, A, 8, 20, 0x455a14ed);
404 OP (FG, A, B, C, D, 13, 5, 0xa9e3e905);
405 OP (FG, D, A, B, C, 2, 9, 0xfcefa3f8);
406 OP (FG, C, D, A, B, 7, 14, 0x676f02d9);
407 OP (FG, B, C, D, A, 12, 20, 0x8d2a4c8a);
409 /* Round 3. */
410 OP (FH, A, B, C, D, 5, 4, 0xfffa3942);
411 OP (FH, D, A, B, C, 8, 11, 0x8771f681);
412 OP (FH, C, D, A, B, 11, 16, 0x6d9d6122);
413 OP (FH, B, C, D, A, 14, 23, 0xfde5380c);
414 OP (FH, A, B, C, D, 1, 4, 0xa4beea44);
415 OP (FH, D, A, B, C, 4, 11, 0x4bdecfa9);
416 OP (FH, C, D, A, B, 7, 16, 0xf6bb4b60);
417 OP (FH, B, C, D, A, 10, 23, 0xbebfbc70);
418 OP (FH, A, B, C, D, 13, 4, 0x289b7ec6);
419 OP (FH, D, A, B, C, 0, 11, 0xeaa127fa);
420 OP (FH, C, D, A, B, 3, 16, 0xd4ef3085);
421 OP (FH, B, C, D, A, 6, 23, 0x04881d05);
422 OP (FH, A, B, C, D, 9, 4, 0xd9d4d039);
423 OP (FH, D, A, B, C, 12, 11, 0xe6db99e5);
424 OP (FH, C, D, A, B, 15, 16, 0x1fa27cf8);
425 OP (FH, B, C, D, A, 2, 23, 0xc4ac5665);
427 /* Round 4. */
428 OP (FI, A, B, C, D, 0, 6, 0xf4292244);
429 OP (FI, D, A, B, C, 7, 10, 0x432aff97);
430 OP (FI, C, D, A, B, 14, 15, 0xab9423a7);
431 OP (FI, B, C, D, A, 5, 21, 0xfc93a039);
432 OP (FI, A, B, C, D, 12, 6, 0x655b59c3);
433 OP (FI, D, A, B, C, 3, 10, 0x8f0ccc92);
434 OP (FI, C, D, A, B, 10, 15, 0xffeff47d);
435 OP (FI, B, C, D, A, 1, 21, 0x85845dd1);
436 OP (FI, A, B, C, D, 8, 6, 0x6fa87e4f);
437 OP (FI, D, A, B, C, 15, 10, 0xfe2ce6e0);
438 OP (FI, C, D, A, B, 6, 15, 0xa3014314);
439 OP (FI, B, C, D, A, 13, 21, 0x4e0811a1);
440 OP (FI, A, B, C, D, 4, 6, 0xf7537e82);
441 OP (FI, D, A, B, C, 11, 10, 0xbd3af235);
442 OP (FI, C, D, A, B, 2, 15, 0x2ad7d2bb);
443 OP (FI, B, C, D, A, 9, 21, 0xeb86d391);
445 /* Add the starting values of the context. */
446 A += A_save;
447 B += B_save;
448 C += C_save;
449 D += D_save;
452 /* Put checksum in context given as argument. */
453 ctx->A = A;
454 ctx->B = B;
455 ctx->C = C;
456 ctx->D = D;