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 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.
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 2, 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
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.  */
20 
21 /* Written by Ulrich Drepper <drepper@gnu.ai.mit.edu>, 1995.  */
22 
23 #ifdef HAVE_CONFIG_H
24 # include <config.h>
25 #endif
26 
27 #include <sys/types.h>
28 #include <stdlib.h>
29 #include <string.h>
30 
31 #include <md5.h>
32 
33 #ifdef _LIBC
34 # include <endian.h>
35 # if __BYTE_ORDER == __BIG_ENDIAN
36 #  define WORDS_BIGENDIAN 1
37 # endif
38 /* We need to keep the namespace clean so define the MD5 function
39    protected using leading __ .  */
40 # define md5_init_ctx __md5_init_ctx
41 # define md5_process_block __md5_process_block
42 # define md5_process_bytes __md5_process_bytes
43 # define md5_finish_ctx __md5_finish_ctx
44 # define md5_read_ctx __md5_read_ctx
45 # define md5_stream __md5_stream
46 # define md5_buffer __md5_buffer
47 #endif
48 
49 #ifdef WORDS_BIGENDIAN
50 # define SWAP(n)                                                        \
51     (((n) << 24) | (((n) & 0xff00) << 8) | (((n) >> 8) & 0xff00) | ((n) >> 24))
52 #else
53 # define SWAP(n) (n)
54 #endif
55 
56 #define BLOCKSIZE 4096
57 /* Ensure that BLOCKSIZE is a multiple of 64.  */
58 #if BLOCKSIZE % 64 != 0
59 /* FIXME-someday (soon?): use #error instead of this kludge.  */
60 "invalid BLOCKSIZE"
61 #endif
62 
63 /* This array contains the bytes used to pad the buffer to the next
64    64-byte boundary.  (RFC 1321, 3.1: Step 1)  */
65 static const unsigned char fillbuf[64] = { 0x80, 0 /* , 0, 0, ...  */ };
66 
67 
68 /* Initialize structure containing state of computation.
69    (RFC 1321, 3.3: Step 3)  */
70 void
md5_init_ctx(struct md5_ctx * ctx)71 md5_init_ctx (struct md5_ctx *ctx)
72 {
73   ctx->A = 0x67452301;
74   ctx->B = 0xefcdab89;
75   ctx->C = 0x98badcfe;
76   ctx->D = 0x10325476;
77 
78   ctx->total[0] = ctx->total[1] = 0;
79   ctx->buflen = 0;
80 }
81 
82 /* Put result from CTX in first 16 bytes following RESBUF.  The result
83    must be in little endian byte order.
84 
85    IMPORTANT: On some systems it is required that RESBUF is correctly
86    aligned for a 32 bits value.  */
87 void *
md5_read_ctx(const struct md5_ctx * ctx,void * resbuf)88 md5_read_ctx (const struct md5_ctx *ctx, void *resbuf)
89 {
90   ((md5_uint32 *) resbuf)[0] = SWAP (ctx->A);
91   ((md5_uint32 *) resbuf)[1] = SWAP (ctx->B);
92   ((md5_uint32 *) resbuf)[2] = SWAP (ctx->C);
93   ((md5_uint32 *) resbuf)[3] = SWAP (ctx->D);
94 
95   return resbuf;
96 }
97 
98 /* Process the remaining bytes in the internal buffer and the usual
99    prolog according to the standard and write the result to RESBUF.
100 
101    IMPORTANT: On some systems it is required that RESBUF is correctly
102    aligned for a 32 bits value.  */
103 void *
md5_finish_ctx(struct md5_ctx * ctx,void * resbuf)104 md5_finish_ctx (struct md5_ctx *ctx, void *resbuf)
105 {
106   /* Take yet unprocessed bytes into account.  */
107   md5_uint32 bytes = ctx->buflen;
108   size_t pad;
109 
110   /* Now count remaining bytes.  */
111   ctx->total[0] += bytes;
112   if (ctx->total[0] < bytes)
113     ++ctx->total[1];
114 
115   pad = bytes >= 56 ? 64 + 56 - bytes : 56 - bytes;
116   memcpy (&ctx->buffer[bytes], fillbuf, pad);
117 
118   /* Put the 64-bit file length in *bits* at the end of the buffer.  */
119   *(md5_uint32 *) &ctx->buffer[bytes + pad] = SWAP (ctx->total[0] << 3);
120   *(md5_uint32 *) &ctx->buffer[bytes + pad + 4] = SWAP ((ctx->total[1] << 3) |
121                                                         (ctx->total[0] >> 29));
122 
123   /* Process last bytes.  */
124   md5_process_block (ctx->buffer, bytes + pad + 8, ctx);
125 
126   return md5_read_ctx (ctx, resbuf);
127 }
128 
129 /* Compute MD5 message digest for bytes read from STREAM.  The
130    resulting message digest number will be written into the 16 bytes
131    beginning at RESBLOCK.  */
132 int
md5_stream(FILE * stream,void * resblock)133 md5_stream (FILE *stream, void *resblock)
134 {
135   struct md5_ctx ctx;
136   char buffer[BLOCKSIZE + 72];
137   size_t sum;
138 
139   /* Initialize the computation context.  */
140   md5_init_ctx (&ctx);
141 
142   /* Iterate over full file contents.  */
143   while (1)
144     {
145       /* We read the file in blocks of BLOCKSIZE bytes.  One call of the
146          computation function processes the whole buffer so that with the
147          next round of the loop another block can be read.  */
148       size_t n;
149       sum = 0;
150 
151       /* Read block.  Take care for partial reads.  */
152       while (1)
153         {
154           n = fread (buffer + sum, 1, BLOCKSIZE - sum, stream);
155 
156           sum += n;
157 
158           if (sum == BLOCKSIZE)
159             break;
160 
161           if (n == 0)
162             {
163               /* Check for the error flag IFF N == 0, so that we don't
164                  exit the loop after a partial read due to e.g., EAGAIN
165                  or EWOULDBLOCK.  */
166               if (ferror (stream))
167                 return 1;
168               goto process_partial_block;
169             }
170 
171           /* We've read at least one byte, so ignore errors.  But always
172              check for EOF, since feof may be true even though N > 0.
173              Otherwise, we could end up calling fread after EOF.  */
174           if (feof (stream))
175             goto process_partial_block;
176         }
177 
178       /* Process buffer with BLOCKSIZE bytes.  Note that
179                         BLOCKSIZE % 64 == 0
180        */
181       md5_process_block (buffer, BLOCKSIZE, &ctx);
182     }
183 
184  process_partial_block:;
185 
186   /* Process any remaining bytes.  */
187   if (sum > 0)
188     md5_process_bytes (buffer, sum, &ctx);
189 
190   /* Construct result in desired memory.  */
191   md5_finish_ctx (&ctx, resblock);
192   return 0;
193 }
194 
195 /* Compute MD5 message digest for LEN bytes beginning at BUFFER.  The
196    result is always in little endian byte order, so that a byte-wise
197    output yields to the wanted ASCII representation of the message
198    digest.  */
199 void *
md5_buffer(const char * buffer,size_t len,void * resblock)200 md5_buffer (const char *buffer, size_t len, void *resblock)
201 {
202   struct md5_ctx ctx;
203 
204   /* Initialize the computation context.  */
205   md5_init_ctx (&ctx);
206 
207   /* Process whole buffer but last len % 64 bytes.  */
208   md5_process_bytes (buffer, len, &ctx);
209 
210   /* Put result in desired memory area.  */
211   return md5_finish_ctx (&ctx, resblock);
212 }
213 
214 
215 void
md5_process_bytes(const void * buffer,size_t len,struct md5_ctx * ctx)216 md5_process_bytes ( const void *buffer, size_t len, struct md5_ctx *ctx)
217 {
218   /* When we already have some bits in our internal buffer concatenate
219      both inputs first.  */
220   if (ctx->buflen != 0)
221     {
222       size_t left_over = ctx->buflen;
223       size_t add = 128 - left_over > len ? len : 128 - left_over;
224 
225       memcpy (&ctx->buffer[left_over], buffer, add);
226       ctx->buflen += add;
227 
228       if (ctx->buflen > 64)
229         {
230           md5_process_block (ctx->buffer, ctx->buflen & ~63, ctx);
231 
232           ctx->buflen &= 63;
233           /* The regions in the following copy operation cannot overlap.  */
234           memcpy (ctx->buffer, &ctx->buffer[(left_over + add) & ~63],
235                   ctx->buflen);
236         }
237 
238       buffer = (const char *) buffer + add;
239       len -= add;
240     }
241 
242   /* Process available complete blocks.  */
243   if (len >= 64)
244     {
245 #if !_STRING_ARCH_unaligned
246 /* To check alignment gcc has an appropriate operator.  Other
247    compilers don't.  */
248 # if __GNUC__ >= 2
249 #  define UNALIGNED_P(p) (((md5_uintptr) p) % __alignof__ (md5_uint32) != 0)
250 # else
251 #  define UNALIGNED_P(p) (((md5_uintptr) p) % sizeof (md5_uint32) != 0)
252 # endif
253       if (UNALIGNED_P (buffer))
254         while (len > 64)
255           {
256             memcpy (ctx->buffer, buffer, 64);
257             md5_process_block (ctx->buffer, 64, ctx);
258             buffer = (const char *) buffer + 64;
259             len -= 64;
260           }
261       else
262 #endif
263         {
264           md5_process_block (buffer, len & ~63, ctx);
265           buffer = (const char *) buffer + (len & ~63);
266           len &= 63;
267         }
268     }
269 
270   /* Move remaining bytes in internal buffer.  */
271   if (len > 0)
272     {
273       size_t left_over = ctx->buflen;
274 
275       memcpy (&ctx->buffer[left_over], buffer, len);
276       left_over += len;
277       if (left_over >= 64)
278         {
279           md5_process_block (ctx->buffer, 64, ctx);
280           left_over -= 64;
281           memcpy (ctx->buffer, &ctx->buffer[64], left_over);
282         }
283       ctx->buflen = left_over;
284     }
285 }
286 
287 
288 /* These are the four functions used in the four steps of the MD5 algorithm
289    and defined in the RFC 1321.  The first function is a little bit optimized
290    (as found in Colin Plumbs public domain implementation).  */
291 /* #define FF(b, c, d) ((b & c) | (~b & d)) */
292 #define FF(b, c, d) (d ^ (b & (c ^ d)))
293 #define FG(b, c, d) FF (d, b, c)
294 #define FH(b, c, d) (b ^ c ^ d)
295 #define FI(b, c, d) (c ^ (b | ~d))
296 
297 /* Process LEN bytes of BUFFER, accumulating context into CTX.
298    It is assumed that LEN % 64 == 0.  */
299 
300 void
md5_process_block(const void * buffer,size_t len,struct md5_ctx * ctx)301 md5_process_block (const void *buffer, size_t len, struct md5_ctx *ctx)
302 {
303   md5_uint32 correct_words[16];
304   const md5_uint32 *words = (const md5_uint32 *) buffer;
305   size_t nwords = len / sizeof (md5_uint32);
306   const md5_uint32 *endp = words + nwords;
307   md5_uint32 A = ctx->A;
308   md5_uint32 B = ctx->B;
309   md5_uint32 C = ctx->C;
310   md5_uint32 D = ctx->D;
311 
312   /* First increment the byte count.  RFC 1321 specifies the possible
313      length of the file up to 2^64 bits.  Here we only compute the
314      number of bytes.  Do a double word increment.  */
315   ctx->total[0] += len;
316   if (ctx->total[0] < len)
317     ++ctx->total[1];
318 
319   /* Process all bytes in the buffer with 64 bytes in each round of
320      the loop.  */
321   while (words < endp)
322     {
323       md5_uint32 *cwp = correct_words;
324       md5_uint32 A_save = A;
325       md5_uint32 B_save = B;
326       md5_uint32 C_save = C;
327       md5_uint32 D_save = D;
328 
329       /* First round: using the given function, the context and a constant
330          the next context is computed.  Because the algorithms processing
331          unit is a 32-bit word and it is determined to work on words in
332          little endian byte order we perhaps have to change the byte order
333          before the computation.  To reduce the work for the next steps
334          we store the swapped words in the array CORRECT_WORDS.  */
335 
336 #define OP(a, b, c, d, s, T)                                            \
337       do                                                                \
338         {                                                               \
339           a += FF (b, c, d) + (*cwp++ = SWAP (*words)) + T;             \
340           ++words;                                                      \
341           a = rol (a, s);                                               \
342           a += b;                                                       \
343         }                                                               \
344       while (0)
345 
346       /* Before we start, one word to the strange constants.
347          They are defined in RFC 1321 as
348 
349          T[i] = (int) (4294967296.0 * fabs (sin (i))), i=1..64, or
350          perl -e 'foreach(1..64){printf "0x%08x\n", int (4294967296 * abs (sin $_))}'
351        */
352 
353       /* Round 1.  */
354       OP (A, B, C, D,  7, 0xd76aa478);
355       OP (D, A, B, C, 12, 0xe8c7b756);
356       OP (C, D, A, B, 17, 0x242070db);
357       OP (B, C, D, A, 22, 0xc1bdceee);
358       OP (A, B, C, D,  7, 0xf57c0faf);
359       OP (D, A, B, C, 12, 0x4787c62a);
360       OP (C, D, A, B, 17, 0xa8304613);
361       OP (B, C, D, A, 22, 0xfd469501);
362       OP (A, B, C, D,  7, 0x698098d8);
363       OP (D, A, B, C, 12, 0x8b44f7af);
364       OP (C, D, A, B, 17, 0xffff5bb1);
365       OP (B, C, D, A, 22, 0x895cd7be);
366       OP (A, B, C, D,  7, 0x6b901122);
367       OP (D, A, B, C, 12, 0xfd987193);
368       OP (C, D, A, B, 17, 0xa679438e);
369       OP (B, C, D, A, 22, 0x49b40821);
370 
371       /* For the second to fourth round we have the possibly swapped words
372          in CORRECT_WORDS.  Redefine the macro to take an additional first
373          argument specifying the function to use.  */
374 #undef OP
375 #define OP(f, a, b, c, d, k, s, T)                                      \
376       do                                                                \
377         {                                                               \
378           a += f (b, c, d) + correct_words[k] + T;                      \
379           a = rol (a, s);                                               \
380           a += b;                                                       \
381         }                                                               \
382       while (0)
383 
384       /* Round 2.  */
385       OP (FG, A, B, C, D,  1,  5, 0xf61e2562);
386       OP (FG, D, A, B, C,  6,  9, 0xc040b340);
387       OP (FG, C, D, A, B, 11, 14, 0x265e5a51);
388       OP (FG, B, C, D, A,  0, 20, 0xe9b6c7aa);
389       OP (FG, A, B, C, D,  5,  5, 0xd62f105d);
390       OP (FG, D, A, B, C, 10,  9, 0x02441453);
391       OP (FG, C, D, A, B, 15, 14, 0xd8a1e681);
392       OP (FG, B, C, D, A,  4, 20, 0xe7d3fbc8);
393       OP (FG, A, B, C, D,  9,  5, 0x21e1cde6);
394       OP (FG, D, A, B, C, 14,  9, 0xc33707d6);
395       OP (FG, C, D, A, B,  3, 14, 0xf4d50d87);
396       OP (FG, B, C, D, A,  8, 20, 0x455a14ed);
397       OP (FG, A, B, C, D, 13,  5, 0xa9e3e905);
398       OP (FG, D, A, B, C,  2,  9, 0xfcefa3f8);
399       OP (FG, C, D, A, B,  7, 14, 0x676f02d9);
400       OP (FG, B, C, D, A, 12, 20, 0x8d2a4c8a);
401 
402       /* Round 3.  */
403       OP (FH, A, B, C, D,  5,  4, 0xfffa3942);
404       OP (FH, D, A, B, C,  8, 11, 0x8771f681);
405       OP (FH, C, D, A, B, 11, 16, 0x6d9d6122);
406       OP (FH, B, C, D, A, 14, 23, 0xfde5380c);
407       OP (FH, A, B, C, D,  1,  4, 0xa4beea44);
408       OP (FH, D, A, B, C,  4, 11, 0x4bdecfa9);
409       OP (FH, C, D, A, B,  7, 16, 0xf6bb4b60);
410       OP (FH, B, C, D, A, 10, 23, 0xbebfbc70);
411       OP (FH, A, B, C, D, 13,  4, 0x289b7ec6);
412       OP (FH, D, A, B, C,  0, 11, 0xeaa127fa);
413       OP (FH, C, D, A, B,  3, 16, 0xd4ef3085);
414       OP (FH, B, C, D, A,  6, 23, 0x04881d05);
415       OP (FH, A, B, C, D,  9,  4, 0xd9d4d039);
416       OP (FH, D, A, B, C, 12, 11, 0xe6db99e5);
417       OP (FH, C, D, A, B, 15, 16, 0x1fa27cf8);
418       OP (FH, B, C, D, A,  2, 23, 0xc4ac5665);
419 
420       /* Round 4.  */
421       OP (FI, A, B, C, D,  0,  6, 0xf4292244);
422       OP (FI, D, A, B, C,  7, 10, 0x432aff97);
423       OP (FI, C, D, A, B, 14, 15, 0xab9423a7);
424       OP (FI, B, C, D, A,  5, 21, 0xfc93a039);
425       OP (FI, A, B, C, D, 12,  6, 0x655b59c3);
426       OP (FI, D, A, B, C,  3, 10, 0x8f0ccc92);
427       OP (FI, C, D, A, B, 10, 15, 0xffeff47d);
428       OP (FI, B, C, D, A,  1, 21, 0x85845dd1);
429       OP (FI, A, B, C, D,  8,  6, 0x6fa87e4f);
430       OP (FI, D, A, B, C, 15, 10, 0xfe2ce6e0);
431       OP (FI, C, D, A, B,  6, 15, 0xa3014314);
432       OP (FI, B, C, D, A, 13, 21, 0x4e0811a1);
433       OP (FI, A, B, C, D,  4,  6, 0xf7537e82);
434       OP (FI, D, A, B, C, 11, 10, 0xbd3af235);
435       OP (FI, C, D, A, B,  2, 15, 0x2ad7d2bb);
436       OP (FI, B, C, D, A,  9, 21, 0xeb86d391);
437 
438       /* Add the starting values of the context.  */
439       A += A_save;
440       B += B_save;
441       C += C_save;
442       D += D_save;
443     }
444 
445   /* Put checksum in context given as argument.  */
446   ctx->A = A;
447   ctx->B = B;
448   ctx->C = C;
449   ctx->D = D;
450 }
451