1/* md5sum.c - Calculate RFC 1321 md5 hash and sha1 hash.
2 *
3 * Copyright 2012 Rob Landley <rob@landley.net>
4 *
5 * See http://refspecs.linuxfoundation.org/LSB_4.1.0/LSB-Core-generic/LSB-Core-generic/md5sum.html
6 * and http://www.ietf.org/rfc/rfc1321.txt
7 *
8 * They're combined this way to share infrastructure, and because md5sum is
9 * and LSB standard command, sha1sum is just a good idea.
10
11USE_MD5SUM(NEWTOY(md5sum, "b", TOYFLAG_USR|TOYFLAG_BIN))
12USE_SHA1SUM(NEWTOY(sha1sum, "b", TOYFLAG_USR|TOYFLAG_BIN))
13
14config MD5SUM
15  bool "md5sum"
16  default y
17  help
18    usage: md5sum [FILE]...
19
20    Calculate md5 hash for each input file, reading from stdin if none.
21    Output one hash (16 hex digits) for each input file, followed by
22    filename.
23
24    -b	brief (hash only, no filename)
25
26config SHA1SUM
27  bool "sha1sum"
28  default y
29  help
30    usage: sha1sum [FILE]...
31
32    calculate sha1 hash for each input file, reading from stdin if none.
33    Output one hash (20 hex digits) for each input file, followed by
34    filename.
35
36    -b	brief (hash only, no filename)
37*/
38
39#define FOR_md5sum
40#include "toys.h"
41
42GLOBALS(
43  unsigned state[5];
44  unsigned oldstate[5];
45  uint64_t count;
46  union {
47    char c[64];
48    unsigned i[16];
49  } buffer;
50)
51
52#define rol(value, bits) (((value) << (bits)) | ((value) >> (32 - (bits))))
53
54// for(i=0; i<64; i++) md5table[i] = abs(sin(i+1))*(1<<32);  But calculating
55// that involves not just floating point but pulling in -lm (and arguing with
56// C about whether 1<<32 is a valid thing to do on 32 bit platforms) so:
57
58static uint32_t md5table[64] = {
59  0xd76aa478, 0xe8c7b756, 0x242070db, 0xc1bdceee, 0xf57c0faf, 0x4787c62a,
60  0xa8304613, 0xfd469501, 0x698098d8, 0x8b44f7af, 0xffff5bb1, 0x895cd7be,
61  0x6b901122, 0xfd987193, 0xa679438e, 0x49b40821, 0xf61e2562, 0xc040b340,
62  0x265e5a51, 0xe9b6c7aa, 0xd62f105d, 0x02441453, 0xd8a1e681, 0xe7d3fbc8,
63  0x21e1cde6, 0xc33707d6, 0xf4d50d87, 0x455a14ed, 0xa9e3e905, 0xfcefa3f8,
64  0x676f02d9, 0x8d2a4c8a, 0xfffa3942, 0x8771f681, 0x6d9d6122, 0xfde5380c,
65  0xa4beea44, 0x4bdecfa9, 0xf6bb4b60, 0xbebfbc70, 0x289b7ec6, 0xeaa127fa,
66  0xd4ef3085, 0x04881d05, 0xd9d4d039, 0xe6db99e5, 0x1fa27cf8, 0xc4ac5665,
67  0xf4292244, 0x432aff97, 0xab9423a7, 0xfc93a039, 0x655b59c3, 0x8f0ccc92,
68  0xffeff47d, 0x85845dd1, 0x6fa87e4f, 0xfe2ce6e0, 0xa3014314, 0x4e0811a1,
69  0xf7537e82, 0xbd3af235, 0x2ad7d2bb, 0xeb86d391
70};
71
72static const uint8_t md5rot[64] = {
73  7, 12, 17, 22, 7, 12, 17, 22, 7, 12, 17, 22, 7, 12, 17, 22,
74  5,  9, 14, 20, 5,  9, 14, 20, 5,  9, 14, 20, 5,  9, 14, 20,
75  4, 11, 16, 23, 4, 11, 16, 23, 4, 11, 16, 23, 4, 11, 16, 23,
76  6, 10, 15, 21, 6, 10, 15, 21, 6, 10, 15, 21, 6, 10, 15, 21
77};
78
79// Mix next 64 bytes of data into md5 hash
80
81static void md5_transform(void)
82{
83  unsigned x[4], *b = TT.buffer.i;
84  int i;
85
86  memcpy(x, TT.state, sizeof(x));
87
88  for (i=0; i<64; i++) {
89    unsigned int in, temp, swap;
90    if (i<16) {
91      in = i;
92      temp = x[1];
93      temp = (temp & x[2]) | ((~temp) & x[3]);
94    } else if (i<32) {
95      in = (1+(5*i))&15;
96      temp = x[3];
97      temp = (x[1] & temp) | (x[2] & ~temp);
98    } else if (i<48) {
99      in = (3*i+5)&15;
100      temp = x[1] ^ x[2] ^ x[3];
101    } else {
102      in = (7*i)&15;
103      temp = x[2] ^ (x[1] | ~x[3]);
104    }
105    temp += x[0] + b[in] + md5table[i];
106    swap = x[3];
107    x[3] = x[2];
108    x[2] = x[1];
109    x[1] += rol(temp, md5rot[i]);
110    x[0] = swap;
111  }
112  for (i=0; i<4; i++) TT.state[i] += x[i];
113}
114
115// Mix next 64 bytes of data into sha1 hash.
116
117static const unsigned rconsts[]={0x5A827999,0x6ED9EBA1,0x8F1BBCDC,0xCA62C1D6};
118
119static void sha1_transform(void)
120{
121  int i, j, k, count;
122  unsigned *block = TT.buffer.i;
123  unsigned *rot[5], *temp;
124
125  // Copy context->state[] to working vars
126  for (i=0; i<5; i++) {
127    TT.oldstate[i] = TT.state[i];
128    rot[i] = TT.state + i;
129  }
130  // 4 rounds of 20 operations each.
131  for (i=count=0; i<4; i++) {
132    for (j=0; j<20; j++) {
133      unsigned work;
134
135      work = *rot[2] ^ *rot[3];
136      if (!i) work = (work & *rot[1]) ^ *rot[3];
137      else {
138        if (i==2) work = ((*rot[1]|*rot[2])&*rot[3])|(*rot[1]&*rot[2]);
139        else work ^= *rot[1];
140      }
141
142      if (!i && j<16)
143        work += block[count] = (rol(block[count],24)&0xFF00FF00)
144                             | (rol(block[count],8)&0x00FF00FF);
145      else
146        work += block[count&15] = rol(block[(count+13)&15]
147              ^ block[(count+8)&15] ^ block[(count+2)&15] ^ block[count&15], 1);
148      *rot[4] += work + rol(*rot[0],5) + rconsts[i];
149      *rot[1] = rol(*rot[1],30);
150
151      // Rotate by one for next time.
152      temp = rot[4];
153      for (k=4; k; k--) rot[k] = rot[k-1];
154      *rot = temp;
155      count++;
156    }
157  }
158  // Add the previous values of state[]
159  for (i=0; i<5; i++) TT.state[i] += TT.oldstate[i];
160}
161
162// Fill the 64-byte working buffer and call transform() when full.
163
164static void hash_update(char *data, unsigned int len, void (*transform)(void))
165{
166  unsigned int i, j;
167
168  j = TT.count & 63;
169  TT.count += len;
170
171  for (;;) {
172    // Grab next chunk of data, return if it's not enough to process a frame
173    i = 64 - j;
174    if (i>len) i = len;
175    memcpy(TT.buffer.c+j, data, i);
176    if (j+i != 64) break;
177
178    // Process a frame
179    if (IS_BIG_ENDIAN)
180      for (j=0; j<16; j++) TT.buffer.i[j] = SWAP_LE32(TT.buffer.i[j]);
181    transform();
182    j=0;
183    data += i;
184    len -= i;
185  }
186}
187
188// Callback for loopfiles()
189
190static void do_hash(int fd, char *name)
191{
192  uint64_t count;
193  int i, sha1=toys.which->name[0]=='s';;
194  char buf;
195  void (*transform)(void);
196
197  /* SHA1 initialization constants  (md5sum uses first 4) */
198  TT.state[0] = 0x67452301;
199  TT.state[1] = 0xEFCDAB89;
200  TT.state[2] = 0x98BADCFE;
201  TT.state[3] = 0x10325476;
202  TT.state[4] = 0xC3D2E1F0;
203  TT.count = 0;
204
205  transform = sha1 ? sha1_transform : md5_transform;
206  for (;;) {
207    i = read(fd, toybuf, sizeof(toybuf));
208    if (i<1) break;
209    hash_update(toybuf, i, transform);
210  }
211
212  count = TT.count << 3;
213
214  // End the message by appending a "1" bit to the data, ending with the
215  // message size (in bits, big endian), and adding enough zero bits in
216  // between to pad to the end of the next 64-byte frame.
217  //
218  // Since our input up to now has been in whole bytes, we can deal with
219  // bytes here too.
220
221  buf = 0x80;
222  do {
223    hash_update(&buf, 1, transform);
224    buf = 0;
225  } while ((TT.count & 63) != 56);
226  count = sha1 ? SWAP_BE64(count) : SWAP_LE64(count);
227  hash_update((void *)&count, 8, transform);
228
229  if (sha1)
230    for (i = 0; i < 20; i++)
231      printf("%02x", 255&(TT.state[i>>2] >> ((3-(i & 3)) * 8)));
232  else for (i=0; i<4; i++) printf("%08x", bswap_32(TT.state[i]));
233
234  // Wipe variables. Cryptographer paranoia.
235  memset(&TT, 0, sizeof(TT));
236
237  printf((toys.optflags & FLAG_b) ? "\n" : "  %s\n", name);
238}
239
240void md5sum_main(void)
241{
242  loopfiles(toys.optargs, do_hash);
243}
244
245void sha1sum_main(void)
246{
247  md5sum_main();
248}
249