Change one-line summary to reflect that shred does
[coreutils.git] / lib / md5.c
blob999e25ef3b9705e1ebebc9d29eeae8c43597fb14
1 /* md5.c - 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, 2001, 2003, 2004 Free Software Foundation, Inc.
4 NOTE: The canonical source of this file is maintained with the GNU C
5 Library. Bugs can be reported to bug-glibc@prep.ai.mit.edu.
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 2, 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., 59 Temple Place - Suite 330, Boston, MA 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 "md5.h"
29 #include <stddef.h>
30 #include <string.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 4096
60 /* Ensure that BLOCKSIZE is a multiple of 64. */
61 #if BLOCKSIZE % 64 != 0
62 /* FIXME-someday (soon?): use #error instead of this kludge. */
63 "invalid BLOCKSIZE"
64 #endif
66 /* This array contains the bytes used to pad the buffer to the next
67 64-byte boundary. (RFC 1321, 3.1: Step 1) */
68 static const unsigned char fillbuf[64] = { 0x80, 0 /* , 0, 0, ... */ };
71 /* Initialize structure containing state of computation.
72 (RFC 1321, 3.3: Step 3) */
73 void
74 md5_init_ctx (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 (const struct md5_ctx *ctx, void *resbuf)
93 ((md5_uint32 *) resbuf)[0] = SWAP (ctx->A);
94 ((md5_uint32 *) resbuf)[1] = SWAP (ctx->B);
95 ((md5_uint32 *) resbuf)[2] = SWAP (ctx->C);
96 ((md5_uint32 *) resbuf)[3] = SWAP (ctx->D);
98 return resbuf;
101 /* Process the remaining bytes in the internal buffer and the usual
102 prolog according to the standard and write the result to RESBUF.
104 IMPORTANT: On some systems it is required that RESBUF is correctly
105 aligned for a 32 bits value. */
106 void *
107 md5_finish_ctx (struct md5_ctx *ctx, void *resbuf)
109 /* Take yet unprocessed bytes into account. */
110 md5_uint32 bytes = ctx->buflen;
111 size_t pad;
113 /* Now count remaining bytes. */
114 ctx->total[0] += bytes;
115 if (ctx->total[0] < bytes)
116 ++ctx->total[1];
118 pad = bytes >= 56 ? 64 + 56 - bytes : 56 - bytes;
119 memcpy (&ctx->buffer[bytes], fillbuf, pad);
121 /* Put the 64-bit file length in *bits* at the end of the buffer. */
122 *(md5_uint32 *) &ctx->buffer[bytes + pad] = SWAP (ctx->total[0] << 3);
123 *(md5_uint32 *) &ctx->buffer[bytes + pad + 4] = SWAP ((ctx->total[1] << 3) |
124 (ctx->total[0] >> 29));
126 /* Process last bytes. */
127 md5_process_block (ctx->buffer, bytes + pad + 8, 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 char buffer[BLOCKSIZE + 72];
140 size_t sum;
142 /* Initialize the computation context. */
143 md5_init_ctx (&ctx);
145 /* Iterate over full file contents. */
146 while (1)
148 /* We read the file in blocks of BLOCKSIZE bytes. One call of the
149 computation function processes the whole buffer so that with the
150 next round of the loop another block can be read. */
151 size_t n;
152 sum = 0;
154 /* Read block. Take care for partial reads. */
155 while (1)
157 n = fread (buffer + sum, 1, BLOCKSIZE - sum, stream);
159 sum += n;
161 if (sum == BLOCKSIZE)
162 break;
164 if (n == 0)
166 /* Check for the error flag IFF N == 0, so that we don't
167 exit the loop after a partial read due to e.g., EAGAIN
168 or EWOULDBLOCK. */
169 if (ferror (stream))
170 return 1;
171 goto process_partial_block;
174 /* We've read at least one byte, so ignore errors. But always
175 check for EOF, since feof may be true even though N > 0.
176 Otherwise, we could end up calling fread after EOF. */
177 if (feof (stream))
178 goto process_partial_block;
181 /* Process buffer with BLOCKSIZE bytes. Note that
182 BLOCKSIZE % 64 == 0
184 md5_process_block (buffer, BLOCKSIZE, &ctx);
187 process_partial_block:;
189 /* Process any remaining bytes. */
190 if (sum > 0)
191 md5_process_bytes (buffer, sum, &ctx);
193 /* Construct result in desired memory. */
194 md5_finish_ctx (&ctx, resblock);
195 return 0;
198 /* Compute MD5 message digest for LEN bytes beginning at BUFFER. The
199 result is always in little endian byte order, so that a byte-wise
200 output yields to the wanted ASCII representation of the message
201 digest. */
202 void *
203 md5_buffer (const char *buffer, size_t len, void *resblock)
205 struct md5_ctx ctx;
207 /* Initialize the computation context. */
208 md5_init_ctx (&ctx);
210 /* Process whole buffer but last len % 64 bytes. */
211 md5_process_bytes (buffer, len, &ctx);
213 /* Put result in desired memory area. */
214 return md5_finish_ctx (&ctx, resblock);
218 void
219 md5_process_bytes (const void *buffer, size_t len, 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 # define alignof(type) offsetof (struct { char c; type x; }, x)
250 # define UNALIGNED_P(p) (((size_t) p) % alignof (md5_uint32) != 0)
251 if (UNALIGNED_P (buffer))
252 while (len > 64)
254 md5_process_block (memcpy (ctx->buffer, buffer, 64), 64, ctx);
255 buffer = (const char *) buffer + 64;
256 len -= 64;
258 else
259 #endif
261 md5_process_block (buffer, len & ~63, ctx);
262 buffer = (const char *) buffer + (len & ~63);
263 len &= 63;
267 /* Move remaining bytes in internal buffer. */
268 if (len > 0)
270 size_t left_over = ctx->buflen;
272 memcpy (&ctx->buffer[left_over], buffer, len);
273 left_over += len;
274 if (left_over >= 64)
276 md5_process_block (ctx->buffer, 64, ctx);
277 left_over -= 64;
278 memcpy (ctx->buffer, &ctx->buffer[64], left_over);
280 ctx->buflen = left_over;
285 /* These are the four functions used in the four steps of the MD5 algorithm
286 and defined in the RFC 1321. The first function is a little bit optimized
287 (as found in Colin Plumbs public domain implementation). */
288 /* #define FF(b, c, d) ((b & c) | (~b & d)) */
289 #define FF(b, c, d) (d ^ (b & (c ^ d)))
290 #define FG(b, c, d) FF (d, b, c)
291 #define FH(b, c, d) (b ^ c ^ d)
292 #define FI(b, c, d) (c ^ (b | ~d))
294 /* Process LEN bytes of BUFFER, accumulating context into CTX.
295 It is assumed that LEN % 64 == 0. */
297 void
298 md5_process_block (const void *buffer, size_t len, struct md5_ctx *ctx)
300 md5_uint32 correct_words[16];
301 const md5_uint32 *words = buffer;
302 size_t nwords = len / sizeof (md5_uint32);
303 const md5_uint32 *endp = words + nwords;
304 md5_uint32 A = ctx->A;
305 md5_uint32 B = ctx->B;
306 md5_uint32 C = ctx->C;
307 md5_uint32 D = ctx->D;
309 /* First increment the byte count. RFC 1321 specifies the possible
310 length of the file up to 2^64 bits. Here we only compute the
311 number of bytes. Do a double word increment. */
312 ctx->total[0] += len;
313 if (ctx->total[0] < len)
314 ++ctx->total[1];
316 /* Process all bytes in the buffer with 64 bytes in each round of
317 the loop. */
318 while (words < endp)
320 md5_uint32 *cwp = correct_words;
321 md5_uint32 A_save = A;
322 md5_uint32 B_save = B;
323 md5_uint32 C_save = C;
324 md5_uint32 D_save = D;
326 /* First round: using the given function, the context and a constant
327 the next context is computed. Because the algorithms processing
328 unit is a 32-bit word and it is determined to work on words in
329 little endian byte order we perhaps have to change the byte order
330 before the computation. To reduce the work for the next steps
331 we store the swapped words in the array CORRECT_WORDS. */
333 #define OP(a, b, c, d, s, T) \
334 do \
336 a += FF (b, c, d) + (*cwp++ = SWAP (*words)) + T; \
337 ++words; \
338 a = rol (a, s); \
339 a += b; \
341 while (0)
343 /* Before we start, one word to the strange constants.
344 They are defined in RFC 1321 as
346 T[i] = (int) (4294967296.0 * fabs (sin (i))), i=1..64, or
347 perl -e 'foreach(1..64){printf "0x%08x\n", int (4294967296 * abs (sin $_))}'
350 /* Round 1. */
351 OP (A, B, C, D, 7, 0xd76aa478);
352 OP (D, A, B, C, 12, 0xe8c7b756);
353 OP (C, D, A, B, 17, 0x242070db);
354 OP (B, C, D, A, 22, 0xc1bdceee);
355 OP (A, B, C, D, 7, 0xf57c0faf);
356 OP (D, A, B, C, 12, 0x4787c62a);
357 OP (C, D, A, B, 17, 0xa8304613);
358 OP (B, C, D, A, 22, 0xfd469501);
359 OP (A, B, C, D, 7, 0x698098d8);
360 OP (D, A, B, C, 12, 0x8b44f7af);
361 OP (C, D, A, B, 17, 0xffff5bb1);
362 OP (B, C, D, A, 22, 0x895cd7be);
363 OP (A, B, C, D, 7, 0x6b901122);
364 OP (D, A, B, C, 12, 0xfd987193);
365 OP (C, D, A, B, 17, 0xa679438e);
366 OP (B, C, D, A, 22, 0x49b40821);
368 /* For the second to fourth round we have the possibly swapped words
369 in CORRECT_WORDS. Redefine the macro to take an additional first
370 argument specifying the function to use. */
371 #undef OP
372 #define OP(f, a, b, c, d, k, s, T) \
373 do \
375 a += f (b, c, d) + correct_words[k] + T; \
376 a = rol (a, s); \
377 a += b; \
379 while (0)
381 /* Round 2. */
382 OP (FG, A, B, C, D, 1, 5, 0xf61e2562);
383 OP (FG, D, A, B, C, 6, 9, 0xc040b340);
384 OP (FG, C, D, A, B, 11, 14, 0x265e5a51);
385 OP (FG, B, C, D, A, 0, 20, 0xe9b6c7aa);
386 OP (FG, A, B, C, D, 5, 5, 0xd62f105d);
387 OP (FG, D, A, B, C, 10, 9, 0x02441453);
388 OP (FG, C, D, A, B, 15, 14, 0xd8a1e681);
389 OP (FG, B, C, D, A, 4, 20, 0xe7d3fbc8);
390 OP (FG, A, B, C, D, 9, 5, 0x21e1cde6);
391 OP (FG, D, A, B, C, 14, 9, 0xc33707d6);
392 OP (FG, C, D, A, B, 3, 14, 0xf4d50d87);
393 OP (FG, B, C, D, A, 8, 20, 0x455a14ed);
394 OP (FG, A, B, C, D, 13, 5, 0xa9e3e905);
395 OP (FG, D, A, B, C, 2, 9, 0xfcefa3f8);
396 OP (FG, C, D, A, B, 7, 14, 0x676f02d9);
397 OP (FG, B, C, D, A, 12, 20, 0x8d2a4c8a);
399 /* Round 3. */
400 OP (FH, A, B, C, D, 5, 4, 0xfffa3942);
401 OP (FH, D, A, B, C, 8, 11, 0x8771f681);
402 OP (FH, C, D, A, B, 11, 16, 0x6d9d6122);
403 OP (FH, B, C, D, A, 14, 23, 0xfde5380c);
404 OP (FH, A, B, C, D, 1, 4, 0xa4beea44);
405 OP (FH, D, A, B, C, 4, 11, 0x4bdecfa9);
406 OP (FH, C, D, A, B, 7, 16, 0xf6bb4b60);
407 OP (FH, B, C, D, A, 10, 23, 0xbebfbc70);
408 OP (FH, A, B, C, D, 13, 4, 0x289b7ec6);
409 OP (FH, D, A, B, C, 0, 11, 0xeaa127fa);
410 OP (FH, C, D, A, B, 3, 16, 0xd4ef3085);
411 OP (FH, B, C, D, A, 6, 23, 0x04881d05);
412 OP (FH, A, B, C, D, 9, 4, 0xd9d4d039);
413 OP (FH, D, A, B, C, 12, 11, 0xe6db99e5);
414 OP (FH, C, D, A, B, 15, 16, 0x1fa27cf8);
415 OP (FH, B, C, D, A, 2, 23, 0xc4ac5665);
417 /* Round 4. */
418 OP (FI, A, B, C, D, 0, 6, 0xf4292244);
419 OP (FI, D, A, B, C, 7, 10, 0x432aff97);
420 OP (FI, C, D, A, B, 14, 15, 0xab9423a7);
421 OP (FI, B, C, D, A, 5, 21, 0xfc93a039);
422 OP (FI, A, B, C, D, 12, 6, 0x655b59c3);
423 OP (FI, D, A, B, C, 3, 10, 0x8f0ccc92);
424 OP (FI, C, D, A, B, 10, 15, 0xffeff47d);
425 OP (FI, B, C, D, A, 1, 21, 0x85845dd1);
426 OP (FI, A, B, C, D, 8, 6, 0x6fa87e4f);
427 OP (FI, D, A, B, C, 15, 10, 0xfe2ce6e0);
428 OP (FI, C, D, A, B, 6, 15, 0xa3014314);
429 OP (FI, B, C, D, A, 13, 21, 0x4e0811a1);
430 OP (FI, A, B, C, D, 4, 6, 0xf7537e82);
431 OP (FI, D, A, B, C, 11, 10, 0xbd3af235);
432 OP (FI, C, D, A, B, 2, 15, 0x2ad7d2bb);
433 OP (FI, B, C, D, A, 9, 21, 0xeb86d391);
435 /* Add the starting values of the context. */
436 A += A_save;
437 B += B_save;
438 C += C_save;
439 D += D_save;
442 /* Put checksum in context given as argument. */
443 ctx->A = A;
444 ctx->B = B;
445 ctx->C = C;
446 ctx->D = D;