Commit | Line | Data |
---|---|---|
942f733f PE |
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. | |
ba318903 | 3 | Copyright (C) 1995-1997, 1999-2001, 2005-2006, 2008-2014 Free Software |
942f733f PE |
4 | Foundation, Inc. |
5 | This file is part of the GNU C Library. | |
6 | ||
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. | |
11 | ||
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. | |
16 | ||
17 | You should have received a copy of the GNU General Public License | |
caf8a9b2 | 18 | along with this program; if not, see <http://www.gnu.org/licenses/>. */ |
942f733f PE |
19 | |
20 | /* Written by Ulrich Drepper <drepper@gnu.ai.mit.edu>, 1995. */ | |
21 | ||
22 | #include <config.h> | |
23 | ||
e9551b12 PE |
24 | #if HAVE_OPENSSL_MD5 |
25 | # define GL_OPENSSL_INLINE _GL_EXTERN_INLINE | |
26 | #endif | |
942f733f PE |
27 | #include "md5.h" |
28 | ||
caf8a9b2 PE |
29 | #include <stdalign.h> |
30 | #include <stdint.h> | |
942f733f PE |
31 | #include <stdlib.h> |
32 | #include <string.h> | |
33 | #include <sys/types.h> | |
34 | ||
35 | #if USE_UNLOCKED_IO | |
36 | # include "unlocked-io.h" | |
37 | #endif | |
38 | ||
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 | |
54 | ||
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 | |
61 | ||
62 | #define BLOCKSIZE 32768 | |
63 | #if BLOCKSIZE % 64 != 0 | |
64 | # error "invalid BLOCKSIZE" | |
65 | #endif | |
66 | ||
e9551b12 | 67 | #if ! HAVE_OPENSSL_MD5 |
942f733f PE |
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, ... */ }; | |
71 | ||
72 | ||
73 | /* Initialize structure containing state of computation. | |
74 | (RFC 1321, 3.3: Step 3) */ | |
75 | void | |
76 | md5_init_ctx (struct md5_ctx *ctx) | |
77 | { | |
78 | ctx->A = 0x67452301; | |
79 | ctx->B = 0xefcdab89; | |
80 | ctx->C = 0x98badcfe; | |
81 | ctx->D = 0x10325476; | |
82 | ||
83 | ctx->total[0] = ctx->total[1] = 0; | |
84 | ctx->buflen = 0; | |
85 | } | |
86 | ||
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 | |
5f90be1b | 89 | * (uint32_t *) cp = v */ |
f64898ab | 90 | static void |
5f90be1b | 91 | set_uint32 (char *cp, uint32_t v) |
942f733f PE |
92 | { |
93 | memcpy (cp, &v, sizeof v); | |
94 | } | |
95 | ||
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) | |
100 | { | |
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)); | |
106 | ||
107 | return resbuf; | |
108 | } | |
109 | ||
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) | |
114 | { | |
115 | /* Take yet unprocessed bytes into account. */ | |
5f90be1b | 116 | uint32_t bytes = ctx->buflen; |
942f733f PE |
117 | size_t size = (bytes < 56) ? 64 / 4 : 64 * 2 / 4; |
118 | ||
119 | /* Now count remaining bytes. */ | |
120 | ctx->total[0] += bytes; | |
121 | if (ctx->total[0] < bytes) | |
122 | ++ctx->total[1]; | |
123 | ||
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)); | |
127 | ||
128 | memcpy (&((char *) ctx->buffer)[bytes], fillbuf, (size - 2) * 4 - bytes); | |
129 | ||
130 | /* Process last bytes. */ | |
131 | md5_process_block (ctx->buffer, size * 4, ctx); | |
132 | ||
133 | return md5_read_ctx (ctx, resbuf); | |
134 | } | |
e9551b12 | 135 | #endif |
942f733f PE |
136 | |
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. */ | |
140 | int | |
141 | md5_stream (FILE *stream, void *resblock) | |
142 | { | |
143 | struct md5_ctx ctx; | |
144 | size_t sum; | |
145 | ||
146 | char *buffer = malloc (BLOCKSIZE + 72); | |
147 | if (!buffer) | |
148 | return 1; | |
149 | ||
150 | /* Initialize the computation context. */ | |
151 | md5_init_ctx (&ctx); | |
152 | ||
153 | /* Iterate over full file contents. */ | |
154 | while (1) | |
155 | { | |
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; | |
161 | ||
162 | /* Read block. Take care for partial reads. */ | |
163 | while (1) | |
164 | { | |
165 | n = fread (buffer + sum, 1, BLOCKSIZE - sum, stream); | |
166 | ||
167 | sum += n; | |
168 | ||
169 | if (sum == BLOCKSIZE) | |
170 | break; | |
171 | ||
172 | if (n == 0) | |
173 | { | |
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)) | |
178 | { | |
179 | free (buffer); | |
180 | return 1; | |
181 | } | |
182 | goto process_partial_block; | |
183 | } | |
184 | ||
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; | |
190 | } | |
191 | ||
192 | /* Process buffer with BLOCKSIZE bytes. Note that | |
193 | BLOCKSIZE % 64 == 0 | |
194 | */ | |
195 | md5_process_block (buffer, BLOCKSIZE, &ctx); | |
196 | } | |
197 | ||
198 | process_partial_block: | |
199 | ||
200 | /* Process any remaining bytes. */ | |
201 | if (sum > 0) | |
202 | md5_process_bytes (buffer, sum, &ctx); | |
203 | ||
204 | /* Construct result in desired memory. */ | |
205 | md5_finish_ctx (&ctx, resblock); | |
206 | free (buffer); | |
207 | return 0; | |
208 | } | |
209 | ||
e9551b12 | 210 | #if ! HAVE_OPENSSL_MD5 |
942f733f PE |
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) | |
217 | { | |
218 | struct md5_ctx ctx; | |
219 | ||
220 | /* Initialize the computation context. */ | |
221 | md5_init_ctx (&ctx); | |
222 | ||
223 | /* Process whole buffer but last len % 64 bytes. */ | |
224 | md5_process_bytes (buffer, len, &ctx); | |
225 | ||
226 | /* Put result in desired memory area. */ | |
227 | return md5_finish_ctx (&ctx, resblock); | |
228 | } | |
229 | ||
230 | ||
231 | void | |
232 | md5_process_bytes (const void *buffer, size_t len, struct md5_ctx *ctx) | |
233 | { | |
234 | /* When we already have some bits in our internal buffer concatenate | |
235 | both inputs first. */ | |
236 | if (ctx->buflen != 0) | |
237 | { | |
238 | size_t left_over = ctx->buflen; | |
239 | size_t add = 128 - left_over > len ? len : 128 - left_over; | |
240 | ||
241 | memcpy (&((char *) ctx->buffer)[left_over], buffer, add); | |
242 | ctx->buflen += add; | |
243 | ||
244 | if (ctx->buflen > 64) | |
245 | { | |
246 | md5_process_block (ctx->buffer, ctx->buflen & ~63, ctx); | |
247 | ||
248 | ctx->buflen &= 63; | |
249 | /* The regions in the following copy operation cannot overlap. */ | |
250 | memcpy (ctx->buffer, | |
251 | &((char *) ctx->buffer)[(left_over + add) & ~63], | |
252 | ctx->buflen); | |
253 | } | |
254 | ||
255 | buffer = (const char *) buffer + add; | |
256 | len -= add; | |
257 | } | |
258 | ||
259 | /* Process available complete blocks. */ | |
260 | if (len >= 64) | |
261 | { | |
262 | #if !_STRING_ARCH_unaligned | |
caf8a9b2 | 263 | # define UNALIGNED_P(p) ((uintptr_t) (p) % alignof (uint32_t) != 0) |
942f733f PE |
264 | if (UNALIGNED_P (buffer)) |
265 | while (len > 64) | |
266 | { | |
267 | md5_process_block (memcpy (ctx->buffer, buffer, 64), 64, ctx); | |
268 | buffer = (const char *) buffer + 64; | |
269 | len -= 64; | |
270 | } | |
271 | else | |
272 | #endif | |
273 | { | |
274 | md5_process_block (buffer, len & ~63, ctx); | |
275 | buffer = (const char *) buffer + (len & ~63); | |
276 | len &= 63; | |
277 | } | |
278 | } | |
279 | ||
280 | /* Move remaining bytes in internal buffer. */ | |
281 | if (len > 0) | |
282 | { | |
283 | size_t left_over = ctx->buflen; | |
284 | ||
285 | memcpy (&((char *) ctx->buffer)[left_over], buffer, len); | |
286 | left_over += len; | |
287 | if (left_over >= 64) | |
288 | { | |
289 | md5_process_block (ctx->buffer, 64, ctx); | |
290 | left_over -= 64; | |
291 | memcpy (ctx->buffer, &ctx->buffer[16], left_over); | |
292 | } | |
293 | ctx->buflen = left_over; | |
294 | } | |
295 | } | |
296 | ||
297 | ||
298 | /* These are the four functions used in the four steps of the MD5 algorithm | |
299 | and defined in the RFC 1321. The first function is a little bit optimized | |
300 | (as found in Colin Plumbs public domain implementation). */ | |
301 | /* #define FF(b, c, d) ((b & c) | (~b & d)) */ | |
302 | #define FF(b, c, d) (d ^ (b & (c ^ d))) | |
303 | #define FG(b, c, d) FF (d, b, c) | |
304 | #define FH(b, c, d) (b ^ c ^ d) | |
305 | #define FI(b, c, d) (c ^ (b | ~d)) | |
306 | ||
307 | /* Process LEN bytes of BUFFER, accumulating context into CTX. | |
308 | It is assumed that LEN % 64 == 0. */ | |
309 | ||
310 | void | |
311 | md5_process_block (const void *buffer, size_t len, struct md5_ctx *ctx) | |
312 | { | |
5f90be1b PE |
313 | uint32_t correct_words[16]; |
314 | const uint32_t *words = buffer; | |
315 | size_t nwords = len / sizeof (uint32_t); | |
316 | const uint32_t *endp = words + nwords; | |
317 | uint32_t A = ctx->A; | |
318 | uint32_t B = ctx->B; | |
319 | uint32_t C = ctx->C; | |
320 | uint32_t D = ctx->D; | |
caf8a9b2 | 321 | uint32_t lolen = len; |
942f733f PE |
322 | |
323 | /* First increment the byte count. RFC 1321 specifies the possible | |
324 | length of the file up to 2^64 bits. Here we only compute the | |
325 | number of bytes. Do a double word increment. */ | |
caf8a9b2 PE |
326 | ctx->total[0] += lolen; |
327 | ctx->total[1] += (len >> 31 >> 1) + (ctx->total[0] < lolen); | |
942f733f PE |
328 | |
329 | /* Process all bytes in the buffer with 64 bytes in each round of | |
330 | the loop. */ | |
331 | while (words < endp) | |
332 | { | |
5f90be1b PE |
333 | uint32_t *cwp = correct_words; |
334 | uint32_t A_save = A; | |
335 | uint32_t B_save = B; | |
336 | uint32_t C_save = C; | |
337 | uint32_t D_save = D; | |
942f733f PE |
338 | |
339 | /* First round: using the given function, the context and a constant | |
340 | the next context is computed. Because the algorithms processing | |
341 | unit is a 32-bit word and it is determined to work on words in | |
342 | little endian byte order we perhaps have to change the byte order | |
343 | before the computation. To reduce the work for the next steps | |
344 | we store the swapped words in the array CORRECT_WORDS. */ | |
345 | ||
346 | #define OP(a, b, c, d, s, T) \ | |
347 | do \ | |
348 | { \ | |
349 | a += FF (b, c, d) + (*cwp++ = SWAP (*words)) + T; \ | |
350 | ++words; \ | |
351 | CYCLIC (a, s); \ | |
352 | a += b; \ | |
353 | } \ | |
354 | while (0) | |
355 | ||
356 | /* It is unfortunate that C does not provide an operator for | |
357 | cyclic rotation. Hope the C compiler is smart enough. */ | |
358 | #define CYCLIC(w, s) (w = (w << s) | (w >> (32 - s))) | |
359 | ||
360 | /* Before we start, one word to the strange constants. | |
361 | They are defined in RFC 1321 as | |
362 | ||
363 | T[i] = (int) (4294967296.0 * fabs (sin (i))), i=1..64 | |
364 | ||
365 | Here is an equivalent invocation using Perl: | |
366 | ||
367 | perl -e 'foreach(1..64){printf "0x%08x\n", int (4294967296 * abs (sin $_))}' | |
368 | */ | |
369 | ||
370 | /* Round 1. */ | |
371 | OP (A, B, C, D, 7, 0xd76aa478); | |
372 | OP (D, A, B, C, 12, 0xe8c7b756); | |
373 | OP (C, D, A, B, 17, 0x242070db); | |
374 | OP (B, C, D, A, 22, 0xc1bdceee); | |
375 | OP (A, B, C, D, 7, 0xf57c0faf); | |
376 | OP (D, A, B, C, 12, 0x4787c62a); | |
377 | OP (C, D, A, B, 17, 0xa8304613); | |
378 | OP (B, C, D, A, 22, 0xfd469501); | |
379 | OP (A, B, C, D, 7, 0x698098d8); | |
380 | OP (D, A, B, C, 12, 0x8b44f7af); | |
381 | OP (C, D, A, B, 17, 0xffff5bb1); | |
382 | OP (B, C, D, A, 22, 0x895cd7be); | |
383 | OP (A, B, C, D, 7, 0x6b901122); | |
384 | OP (D, A, B, C, 12, 0xfd987193); | |
385 | OP (C, D, A, B, 17, 0xa679438e); | |
386 | OP (B, C, D, A, 22, 0x49b40821); | |
387 | ||
388 | /* For the second to fourth round we have the possibly swapped words | |
389 | in CORRECT_WORDS. Redefine the macro to take an additional first | |
390 | argument specifying the function to use. */ | |
391 | #undef OP | |
392 | #define OP(f, a, b, c, d, k, s, T) \ | |
393 | do \ | |
394 | { \ | |
395 | a += f (b, c, d) + correct_words[k] + T; \ | |
396 | CYCLIC (a, s); \ | |
397 | a += b; \ | |
398 | } \ | |
399 | while (0) | |
400 | ||
401 | /* Round 2. */ | |
402 | OP (FG, A, B, C, D, 1, 5, 0xf61e2562); | |
403 | OP (FG, D, A, B, C, 6, 9, 0xc040b340); | |
404 | OP (FG, C, D, A, B, 11, 14, 0x265e5a51); | |
405 | OP (FG, B, C, D, A, 0, 20, 0xe9b6c7aa); | |
406 | OP (FG, A, B, C, D, 5, 5, 0xd62f105d); | |
407 | OP (FG, D, A, B, C, 10, 9, 0x02441453); | |
408 | OP (FG, C, D, A, B, 15, 14, 0xd8a1e681); | |
409 | OP (FG, B, C, D, A, 4, 20, 0xe7d3fbc8); | |
410 | OP (FG, A, B, C, D, 9, 5, 0x21e1cde6); | |
411 | OP (FG, D, A, B, C, 14, 9, 0xc33707d6); | |
412 | OP (FG, C, D, A, B, 3, 14, 0xf4d50d87); | |
413 | OP (FG, B, C, D, A, 8, 20, 0x455a14ed); | |
414 | OP (FG, A, B, C, D, 13, 5, 0xa9e3e905); | |
415 | OP (FG, D, A, B, C, 2, 9, 0xfcefa3f8); | |
416 | OP (FG, C, D, A, B, 7, 14, 0x676f02d9); | |
417 | OP (FG, B, C, D, A, 12, 20, 0x8d2a4c8a); | |
418 | ||
419 | /* Round 3. */ | |
420 | OP (FH, A, B, C, D, 5, 4, 0xfffa3942); | |
421 | OP (FH, D, A, B, C, 8, 11, 0x8771f681); | |
422 | OP (FH, C, D, A, B, 11, 16, 0x6d9d6122); | |
423 | OP (FH, B, C, D, A, 14, 23, 0xfde5380c); | |
424 | OP (FH, A, B, C, D, 1, 4, 0xa4beea44); | |
425 | OP (FH, D, A, B, C, 4, 11, 0x4bdecfa9); | |
426 | OP (FH, C, D, A, B, 7, 16, 0xf6bb4b60); | |
427 | OP (FH, B, C, D, A, 10, 23, 0xbebfbc70); | |
428 | OP (FH, A, B, C, D, 13, 4, 0x289b7ec6); | |
429 | OP (FH, D, A, B, C, 0, 11, 0xeaa127fa); | |
430 | OP (FH, C, D, A, B, 3, 16, 0xd4ef3085); | |
431 | OP (FH, B, C, D, A, 6, 23, 0x04881d05); | |
432 | OP (FH, A, B, C, D, 9, 4, 0xd9d4d039); | |
433 | OP (FH, D, A, B, C, 12, 11, 0xe6db99e5); | |
434 | OP (FH, C, D, A, B, 15, 16, 0x1fa27cf8); | |
435 | OP (FH, B, C, D, A, 2, 23, 0xc4ac5665); | |
436 | ||
437 | /* Round 4. */ | |
438 | OP (FI, A, B, C, D, 0, 6, 0xf4292244); | |
439 | OP (FI, D, A, B, C, 7, 10, 0x432aff97); | |
440 | OP (FI, C, D, A, B, 14, 15, 0xab9423a7); | |
441 | OP (FI, B, C, D, A, 5, 21, 0xfc93a039); | |
442 | OP (FI, A, B, C, D, 12, 6, 0x655b59c3); | |
443 | OP (FI, D, A, B, C, 3, 10, 0x8f0ccc92); | |
444 | OP (FI, C, D, A, B, 10, 15, 0xffeff47d); | |
445 | OP (FI, B, C, D, A, 1, 21, 0x85845dd1); | |
446 | OP (FI, A, B, C, D, 8, 6, 0x6fa87e4f); | |
447 | OP (FI, D, A, B, C, 15, 10, 0xfe2ce6e0); | |
448 | OP (FI, C, D, A, B, 6, 15, 0xa3014314); | |
449 | OP (FI, B, C, D, A, 13, 21, 0x4e0811a1); | |
450 | OP (FI, A, B, C, D, 4, 6, 0xf7537e82); | |
451 | OP (FI, D, A, B, C, 11, 10, 0xbd3af235); | |
452 | OP (FI, C, D, A, B, 2, 15, 0x2ad7d2bb); | |
453 | OP (FI, B, C, D, A, 9, 21, 0xeb86d391); | |
454 | ||
455 | /* Add the starting values of the context. */ | |
456 | A += A_save; | |
457 | B += B_save; | |
458 | C += C_save; | |
459 | D += D_save; | |
460 | } | |
461 | ||
462 | /* Put checksum in context given as argument. */ | |
463 | ctx->A = A; | |
464 | ctx->B = B; | |
465 | ctx->C = C; | |
466 | ctx->D = D; | |
467 | } | |
e9551b12 | 468 | #endif |