(collate_read): Ignore script lines as well when ignoring the whole category.
[glibc.git] / crypt / md5.c
blob922e7cc7e97352f81fde0a076fdb11886bbec7ce
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.
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.
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.
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. */
22 /* Written by Ulrich Drepper <drepper@gnu.ai.mit.edu>, 1995. */
24 #ifdef HAVE_CONFIG_H
25 # include <config.h>
26 #endif
28 #include <sys/types.h>
30 #if STDC_HEADERS || defined _LIBC
31 # include <stdlib.h>
32 # include <string.h>
33 #else
34 # ifndef HAVE_MEMCPY
35 # define memcpy(d, s, n) (bcopy ((s), (d), (n)), (d))
36 # endif
37 #endif
39 #include "md5.h"
41 #ifdef _LIBC
42 # include <endian.h>
43 # if __BYTE_ORDER == __BIG_ENDIAN
44 # define WORDS_BIGENDIAN 1
45 # endif
46 /* We need to keep the namespace clean so define the MD5 function
47 protected using leading __ . */
48 # define md5_init_ctx __md5_init_ctx
49 # define md5_process_block __md5_process_block
50 # define md5_process_bytes __md5_process_bytes
51 # define md5_finish_ctx __md5_finish_ctx
52 # define md5_read_ctx __md5_read_ctx
53 # define md5_stream __md5_stream
54 # define md5_buffer __md5_buffer
55 #endif
57 #ifdef WORDS_BIGENDIAN
58 # define SWAP(n) \
59 (((n) << 24) | (((n) & 0xff00) << 8) | (((n) >> 8) & 0xff00) | ((n) >> 24))
60 #else
61 # define SWAP(n) (n)
62 #endif
65 /* This array contains the bytes used to pad the buffer to the next
66 64-byte boundary. (RFC 1321, 3.1: Step 1) */
67 static const unsigned char fillbuf[64] = { 0x80, 0 /* , 0, 0, ... */ };
70 /* Initialize structure containing state of computation.
71 (RFC 1321, 3.3: Step 3) */
72 void
73 md5_init_ctx (ctx)
74 struct md5_ctx *ctx;
76 ctx->A = 0x67452301;
77 ctx->B = 0xefcdab89;
78 ctx->C = 0x98badcfe;
79 ctx->D = 0x10325476;
81 ctx->total[0] = ctx->total[1] = 0;
82 ctx->buflen = 0;
85 /* Put result from CTX in first 16 bytes following RESBUF. The result
86 must be in little endian byte order.
88 IMPORTANT: On some systems it is required that RESBUF is correctly
89 aligned for a 32 bits value. */
90 void *
91 md5_read_ctx (ctx, resbuf)
92 const struct md5_ctx *ctx;
93 void *resbuf;
95 ((md5_uint32 *) resbuf)[0] = SWAP (ctx->A);
96 ((md5_uint32 *) resbuf)[1] = SWAP (ctx->B);
97 ((md5_uint32 *) resbuf)[2] = SWAP (ctx->C);
98 ((md5_uint32 *) resbuf)[3] = SWAP (ctx->D);
100 return resbuf;
103 /* Process the remaining bytes in the internal buffer and the usual
104 prolog according to the standard and write the result to RESBUF.
106 IMPORTANT: On some systems it is required that RESBUF is correctly
107 aligned for a 32 bits value. */
108 void *
109 md5_finish_ctx (ctx, resbuf)
110 struct md5_ctx *ctx;
111 void *resbuf;
113 /* Take yet unprocessed bytes into account. */
114 md5_uint32 bytes = ctx->buflen;
115 size_t pad;
117 /* Now count remaining bytes. */
118 ctx->total[0] += bytes;
119 if (ctx->total[0] < bytes)
120 ++ctx->total[1];
122 pad = bytes >= 56 ? 64 + 56 - bytes : 56 - bytes;
123 memcpy (&ctx->buffer[bytes], fillbuf, pad);
125 /* Put the 64-bit file length in *bits* at the end of the buffer. */
126 *(md5_uint32 *) &ctx->buffer[bytes + pad] = SWAP (ctx->total[0] << 3);
127 *(md5_uint32 *) &ctx->buffer[bytes + pad + 4] = SWAP ((ctx->total[1] << 3) |
128 (ctx->total[0] >> 29));
130 /* Process last bytes. */
131 md5_process_block (ctx->buffer, bytes + pad + 8, ctx);
133 return md5_read_ctx (ctx, resbuf);
136 /* Compute MD5 message digest for bytes read from STREAM. The
137 resulting message digest number will be written into the 16 bytes
138 beginning at RESBLOCK. */
140 md5_stream (stream, resblock)
141 FILE *stream;
142 void *resblock;
144 /* Important: BLOCKSIZE must be a multiple of 64. */
145 #define BLOCKSIZE 4096
146 struct md5_ctx ctx;
147 char buffer[BLOCKSIZE + 72];
148 size_t sum;
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. */
165 n = fread (buffer + sum, 1, BLOCKSIZE - sum, stream);
167 sum += n;
169 while (sum < BLOCKSIZE && n != 0);
170 if (n == 0 && ferror (stream))
171 return 1;
173 /* If end of file is reached, end the loop. */
174 if (n == 0)
175 break;
177 /* Process buffer with BLOCKSIZE bytes. Note that
178 BLOCKSIZE % 64 == 0
180 md5_process_block (buffer, BLOCKSIZE, &ctx);
183 /* Add the last bytes if necessary. */
184 if (sum > 0)
185 md5_process_bytes (buffer, sum, &ctx);
187 /* Construct result in desired memory. */
188 md5_finish_ctx (&ctx, resblock);
189 return 0;
192 /* Compute MD5 message digest for LEN bytes beginning at BUFFER. The
193 result is always in little endian byte order, so that a byte-wise
194 output yields to the wanted ASCII representation of the message
195 digest. */
196 void *
197 md5_buffer (buffer, len, resblock)
198 const char *buffer;
199 size_t len;
200 void *resblock;
202 struct md5_ctx ctx;
204 /* Initialize the computation context. */
205 md5_init_ctx (&ctx);
207 /* Process whole buffer but last len % 64 bytes. */
208 md5_process_bytes (buffer, len, &ctx);
210 /* Put result in desired memory area. */
211 return md5_finish_ctx (&ctx, resblock);
215 void
216 md5_process_bytes (buffer, len, ctx)
217 const void *buffer;
218 size_t len;
219 struct md5_ctx *ctx;
221 /* When we already have some bits in our internal buffer concatenate
222 both inputs first. */
223 if (ctx->buflen != 0)
225 size_t left_over = ctx->buflen;
226 size_t add = 128 - left_over > len ? len : 128 - left_over;
228 memcpy (&ctx->buffer[left_over], buffer, add);
229 ctx->buflen += add;
231 if (ctx->buflen > 64)
233 md5_process_block (ctx->buffer, ctx->buflen & ~63, ctx);
235 ctx->buflen &= 63;
236 /* The regions in the following copy operation cannot overlap. */
237 memcpy (ctx->buffer, &ctx->buffer[(left_over + add) & ~63],
238 ctx->buflen);
241 buffer = (const char *) buffer + add;
242 len -= add;
245 /* Process available complete blocks. */
246 if (len >= 64)
248 #if !_STRING_ARCH_unaligned
249 /* To check alignment gcc has an appropriate operator. Other
250 compilers don't. */
251 # if __GNUC__ >= 2
252 # define UNALIGNED_P(p) (((md5_uintptr) p) % __alignof__ (md5_uint32) != 0)
253 # else
254 # define UNALIGNED_P(p) (((md5_uintptr) p) % sizeof (md5_uint32) != 0)
255 # endif
256 if (UNALIGNED_P (buffer))
257 while (len > 64)
259 md5_process_block (memcpy (ctx->buffer, buffer, 64), 64, ctx);
260 buffer = (const char *) buffer + 64;
261 len -= 64;
263 else
264 #endif
266 md5_process_block (buffer, len & ~63, ctx);
267 buffer = (const char *) buffer + (len & ~63);
268 len &= 63;
272 /* Move remaining bytes in internal buffer. */
273 if (len > 0)
275 size_t left_over = ctx->buflen;
277 memcpy (&ctx->buffer[left_over], buffer, len);
278 left_over += len;
279 if (left_over >= 64)
281 md5_process_block (ctx->buffer, 64, ctx);
282 left_over -= 64;
283 memcpy (ctx->buffer, &ctx->buffer[64], left_over);
285 ctx->buflen = left_over;
290 /* These are the four functions used in the four steps of the MD5 algorithm
291 and defined in the RFC 1321. The first function is a little bit optimized
292 (as found in Colin Plumbs public domain implementation). */
293 /* #define FF(b, c, d) ((b & c) | (~b & d)) */
294 #define FF(b, c, d) (d ^ (b & (c ^ d)))
295 #define FG(b, c, d) FF (d, b, c)
296 #define FH(b, c, d) (b ^ c ^ d)
297 #define FI(b, c, d) (c ^ (b | ~d))
299 /* Process LEN bytes of BUFFER, accumulating context into CTX.
300 It is assumed that LEN % 64 == 0. */
302 void
303 md5_process_block (buffer, len, ctx)
304 const void *buffer;
305 size_t len;
306 struct md5_ctx *ctx;
308 md5_uint32 correct_words[16];
309 const md5_uint32 *words = buffer;
310 size_t nwords = len / sizeof (md5_uint32);
311 const md5_uint32 *endp = words + nwords;
312 md5_uint32 A = ctx->A;
313 md5_uint32 B = ctx->B;
314 md5_uint32 C = ctx->C;
315 md5_uint32 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 md5_uint32 *cwp = correct_words;
329 md5_uint32 A_save = A;
330 md5_uint32 B_save = B;
331 md5_uint32 C_save = C;
332 md5_uint32 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
361 /* Round 1. */
362 OP (A, B, C, D, 7, 0xd76aa478);
363 OP (D, A, B, C, 12, 0xe8c7b756);
364 OP (C, D, A, B, 17, 0x242070db);
365 OP (B, C, D, A, 22, 0xc1bdceee);
366 OP (A, B, C, D, 7, 0xf57c0faf);
367 OP (D, A, B, C, 12, 0x4787c62a);
368 OP (C, D, A, B, 17, 0xa8304613);
369 OP (B, C, D, A, 22, 0xfd469501);
370 OP (A, B, C, D, 7, 0x698098d8);
371 OP (D, A, B, C, 12, 0x8b44f7af);
372 OP (C, D, A, B, 17, 0xffff5bb1);
373 OP (B, C, D, A, 22, 0x895cd7be);
374 OP (A, B, C, D, 7, 0x6b901122);
375 OP (D, A, B, C, 12, 0xfd987193);
376 OP (C, D, A, B, 17, 0xa679438e);
377 OP (B, C, D, A, 22, 0x49b40821);
379 /* For the second to fourth round we have the possibly swapped words
380 in CORRECT_WORDS. Redefine the macro to take an additional first
381 argument specifying the function to use. */
382 #undef OP
383 #define OP(f, a, b, c, d, k, s, T) \
384 do \
386 a += f (b, c, d) + correct_words[k] + T; \
387 CYCLIC (a, s); \
388 a += b; \
390 while (0)
392 /* Round 2. */
393 OP (FG, A, B, C, D, 1, 5, 0xf61e2562);
394 OP (FG, D, A, B, C, 6, 9, 0xc040b340);
395 OP (FG, C, D, A, B, 11, 14, 0x265e5a51);
396 OP (FG, B, C, D, A, 0, 20, 0xe9b6c7aa);
397 OP (FG, A, B, C, D, 5, 5, 0xd62f105d);
398 OP (FG, D, A, B, C, 10, 9, 0x02441453);
399 OP (FG, C, D, A, B, 15, 14, 0xd8a1e681);
400 OP (FG, B, C, D, A, 4, 20, 0xe7d3fbc8);
401 OP (FG, A, B, C, D, 9, 5, 0x21e1cde6);
402 OP (FG, D, A, B, C, 14, 9, 0xc33707d6);
403 OP (FG, C, D, A, B, 3, 14, 0xf4d50d87);
404 OP (FG, B, C, D, A, 8, 20, 0x455a14ed);
405 OP (FG, A, B, C, D, 13, 5, 0xa9e3e905);
406 OP (FG, D, A, B, C, 2, 9, 0xfcefa3f8);
407 OP (FG, C, D, A, B, 7, 14, 0x676f02d9);
408 OP (FG, B, C, D, A, 12, 20, 0x8d2a4c8a);
410 /* Round 3. */
411 OP (FH, A, B, C, D, 5, 4, 0xfffa3942);
412 OP (FH, D, A, B, C, 8, 11, 0x8771f681);
413 OP (FH, C, D, A, B, 11, 16, 0x6d9d6122);
414 OP (FH, B, C, D, A, 14, 23, 0xfde5380c);
415 OP (FH, A, B, C, D, 1, 4, 0xa4beea44);
416 OP (FH, D, A, B, C, 4, 11, 0x4bdecfa9);
417 OP (FH, C, D, A, B, 7, 16, 0xf6bb4b60);
418 OP (FH, B, C, D, A, 10, 23, 0xbebfbc70);
419 OP (FH, A, B, C, D, 13, 4, 0x289b7ec6);
420 OP (FH, D, A, B, C, 0, 11, 0xeaa127fa);
421 OP (FH, C, D, A, B, 3, 16, 0xd4ef3085);
422 OP (FH, B, C, D, A, 6, 23, 0x04881d05);
423 OP (FH, A, B, C, D, 9, 4, 0xd9d4d039);
424 OP (FH, D, A, B, C, 12, 11, 0xe6db99e5);
425 OP (FH, C, D, A, B, 15, 16, 0x1fa27cf8);
426 OP (FH, B, C, D, A, 2, 23, 0xc4ac5665);
428 /* Round 4. */
429 OP (FI, A, B, C, D, 0, 6, 0xf4292244);
430 OP (FI, D, A, B, C, 7, 10, 0x432aff97);
431 OP (FI, C, D, A, B, 14, 15, 0xab9423a7);
432 OP (FI, B, C, D, A, 5, 21, 0xfc93a039);
433 OP (FI, A, B, C, D, 12, 6, 0x655b59c3);
434 OP (FI, D, A, B, C, 3, 10, 0x8f0ccc92);
435 OP (FI, C, D, A, B, 10, 15, 0xffeff47d);
436 OP (FI, B, C, D, A, 1, 21, 0x85845dd1);
437 OP (FI, A, B, C, D, 8, 6, 0x6fa87e4f);
438 OP (FI, D, A, B, C, 15, 10, 0xfe2ce6e0);
439 OP (FI, C, D, A, B, 6, 15, 0xa3014314);
440 OP (FI, B, C, D, A, 13, 21, 0x4e0811a1);
441 OP (FI, A, B, C, D, 4, 6, 0xf7537e82);
442 OP (FI, D, A, B, C, 11, 10, 0xbd3af235);
443 OP (FI, C, D, A, B, 2, 15, 0x2ad7d2bb);
444 OP (FI, B, C, D, A, 9, 21, 0xeb86d391);
446 /* Add the starting values of the context. */
447 A += A_save;
448 B += B_save;
449 C += C_save;
450 D += D_save;
453 /* Put checksum in context given as argument. */
454 ctx->A = A;
455 ctx->B = B;
456 ctx->C = C;
457 ctx->D = D;