michael@0: /* vim:set ts=2 sw=2 et cindent: */
michael@0: /* This Source Code Form is subject to the terms of the Mozilla Public
michael@0:  * License, v. 2.0. If a copy of the MPL was not distributed with this
michael@0:  * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
michael@0: 
michael@0: /*
michael@0:  * "clean room" MD4 implementation (see RFC 1320)
michael@0:  */
michael@0: 
michael@0: #include <string.h>
michael@0: #include "md4.h"
michael@0: 
michael@0: typedef uint32_t Uint32;
michael@0: typedef uint8_t Uint8;
michael@0: 
michael@0: /* the "conditional" function */
michael@0: #define F(x,y,z) (((x) & (y)) | (~(x) & (z)))
michael@0: 
michael@0: /* the "majority" function */
michael@0: #define G(x,y,z) (((x) & (y)) | ((x) & (z)) | ((y) & (z)))
michael@0: 
michael@0: /* the "parity" function */
michael@0: #define H(x,y,z) ((x) ^ (y) ^ (z))
michael@0: 
michael@0: /* rotate n-bits to the left */
michael@0: #define ROTL(x,n) (((x) << (n)) | ((x) >> (0x20 - n)))
michael@0: 
michael@0: /* round 1: [abcd k s]: a = (a + F(b,c,d) + X[k]) <<< s */
michael@0: #define RD1(a,b,c,d,k,s) a += F(b,c,d) + X[k]; a = ROTL(a,s)
michael@0: 
michael@0: /* round 2: [abcd k s]: a = (a + G(b,c,d) + X[k] + MAGIC) <<< s */
michael@0: #define RD2(a,b,c,d,k,s) a += G(b,c,d) + X[k] + 0x5A827999; a = ROTL(a,s)
michael@0: 
michael@0: /* round 3: [abcd k s]: a = (a + H(b,c,d) + X[k] + MAGIC) <<< s */
michael@0: #define RD3(a,b,c,d,k,s) a += H(b,c,d) + X[k] + 0x6ED9EBA1; a = ROTL(a,s)
michael@0: 
michael@0: /* converts from word array to byte array, len is number of bytes */
michael@0: static void w2b(Uint8 *out, const Uint32 *in, Uint32 len)
michael@0: {
michael@0:   Uint8 *bp; const Uint32 *wp, *wpend;
michael@0: 
michael@0:   bp = out;
michael@0:   wp = in;
michael@0:   wpend = wp + (len >> 2);
michael@0: 
michael@0:   for (; wp != wpend; ++wp, bp += 4)
michael@0:   {
michael@0:     bp[0] = (Uint8) ((*wp      ) & 0xFF);
michael@0:     bp[1] = (Uint8) ((*wp >>  8) & 0xFF);
michael@0:     bp[2] = (Uint8) ((*wp >> 16) & 0xFF);
michael@0:     bp[3] = (Uint8) ((*wp >> 24) & 0xFF);
michael@0:   }
michael@0: }
michael@0: 
michael@0: /* converts from byte array to word array, len is number of bytes */
michael@0: static void b2w(Uint32 *out, const Uint8 *in, Uint32 len)
michael@0: {
michael@0:   Uint32 *wp; const Uint8 *bp, *bpend;
michael@0: 
michael@0:   wp = out;
michael@0:   bp = in;
michael@0:   bpend = in + len;
michael@0: 
michael@0:   for (; bp != bpend; bp += 4, ++wp)
michael@0:   {
michael@0:     *wp = (Uint32) (bp[0]      ) |
michael@0:           (Uint32) (bp[1] <<  8) |
michael@0:           (Uint32) (bp[2] << 16) |
michael@0:           (Uint32) (bp[3] << 24);
michael@0:   }
michael@0: }
michael@0: 
michael@0: /* update state: data is 64 bytes in length */
michael@0: static void md4step(Uint32 state[4], const Uint8 *data)
michael@0: {
michael@0:   Uint32 A, B, C, D, X[16];
michael@0: 
michael@0:   b2w(X, data, 64);
michael@0: 
michael@0:   A = state[0];
michael@0:   B = state[1];
michael@0:   C = state[2];
michael@0:   D = state[3];
michael@0: 
michael@0:   RD1(A,B,C,D, 0,3); RD1(D,A,B,C, 1,7); RD1(C,D,A,B, 2,11); RD1(B,C,D,A, 3,19);
michael@0:   RD1(A,B,C,D, 4,3); RD1(D,A,B,C, 5,7); RD1(C,D,A,B, 6,11); RD1(B,C,D,A, 7,19);
michael@0:   RD1(A,B,C,D, 8,3); RD1(D,A,B,C, 9,7); RD1(C,D,A,B,10,11); RD1(B,C,D,A,11,19);
michael@0:   RD1(A,B,C,D,12,3); RD1(D,A,B,C,13,7); RD1(C,D,A,B,14,11); RD1(B,C,D,A,15,19);
michael@0: 
michael@0:   RD2(A,B,C,D, 0,3); RD2(D,A,B,C, 4,5); RD2(C,D,A,B, 8, 9); RD2(B,C,D,A,12,13); 
michael@0:   RD2(A,B,C,D, 1,3); RD2(D,A,B,C, 5,5); RD2(C,D,A,B, 9, 9); RD2(B,C,D,A,13,13); 
michael@0:   RD2(A,B,C,D, 2,3); RD2(D,A,B,C, 6,5); RD2(C,D,A,B,10, 9); RD2(B,C,D,A,14,13); 
michael@0:   RD2(A,B,C,D, 3,3); RD2(D,A,B,C, 7,5); RD2(C,D,A,B,11, 9); RD2(B,C,D,A,15,13); 
michael@0: 
michael@0:   RD3(A,B,C,D, 0,3); RD3(D,A,B,C, 8,9); RD3(C,D,A,B, 4,11); RD3(B,C,D,A,12,15);
michael@0:   RD3(A,B,C,D, 2,3); RD3(D,A,B,C,10,9); RD3(C,D,A,B, 6,11); RD3(B,C,D,A,14,15);
michael@0:   RD3(A,B,C,D, 1,3); RD3(D,A,B,C, 9,9); RD3(C,D,A,B, 5,11); RD3(B,C,D,A,13,15);
michael@0:   RD3(A,B,C,D, 3,3); RD3(D,A,B,C,11,9); RD3(C,D,A,B, 7,11); RD3(B,C,D,A,15,15);
michael@0: 
michael@0:   state[0] += A;
michael@0:   state[1] += B;
michael@0:   state[2] += C;
michael@0:   state[3] += D;
michael@0: }
michael@0: 
michael@0: void md4sum(const Uint8 *input, Uint32 inputLen, Uint8 *result)
michael@0: {
michael@0:   Uint8 final[128];
michael@0:   Uint32 i, n, m, state[4];
michael@0: 
michael@0:   /* magic initial states */
michael@0:   state[0] = 0x67452301;
michael@0:   state[1] = 0xEFCDAB89;
michael@0:   state[2] = 0x98BADCFE;
michael@0:   state[3] = 0x10325476;
michael@0: 
michael@0:   /* compute number of complete 64-byte segments contained in input */
michael@0:   m = inputLen >> 6;
michael@0: 
michael@0:   /* digest first m segments */
michael@0:   for (i=0; i<m; ++i)
michael@0:     md4step(state, (input + (i << 6)));
michael@0: 
michael@0:   /* build final buffer */
michael@0:   n = inputLen % 64;
michael@0:   memcpy(final, input + (m << 6), n);
michael@0:   final[n] = 0x80;
michael@0:   memset(final + n + 1, 0, 120 - (n + 1));
michael@0: 
michael@0:   inputLen = inputLen << 3;
michael@0:   w2b(final + (n >= 56 ? 120 : 56), &inputLen, 4);
michael@0: 
michael@0:   md4step(state, final);
michael@0:   if (n >= 56)
michael@0:     md4step(state, final + 64);
michael@0: 
michael@0:   /* copy state to result */
michael@0:   w2b(result, state, 16);
michael@0: }