1 /* crypto/des/des_locl.h */
2 /* Copyright (C) 1995-1997 Eric Young (eay@cryptsoft.com)
5 * This package is an SSL implementation written
6 * by Eric Young (eay@cryptsoft.com).
7 * The implementation was written so as to conform with Netscapes SSL.
9 * This library is free for commercial and non-commercial use as long as
10 * the following conditions are aheared to. The following conditions
11 * apply to all code found in this distribution, be it the RC4, RSA,
12 * lhash, DES, etc., code; not just the SSL code. The SSL documentation
13 * included with this distribution is covered by the same copyright terms
14 * except that the holder is Tim Hudson (tjh@cryptsoft.com).
16 * Copyright remains Eric Young's, and as such any Copyright notices in
17 * the code are not to be removed.
18 * If this package is used in a product, Eric Young should be given attribution
19 * as the author of the parts of the library used.
20 * This can be in the form of a textual message at program startup or
21 * in documentation (online or textual) provided with the package.
23 * Redistribution and use in source and binary forms, with or without
24 * modification, are permitted provided that the following conditions
26 * 1. Redistributions of source code must retain the copyright
27 * notice, this list of conditions and the following disclaimer.
28 * 2. Redistributions in binary form must reproduce the above copyright
29 * notice, this list of conditions and the following disclaimer in the
30 * documentation and/or other materials provided with the distribution.
31 * 3. All advertising materials mentioning features or use of this software
32 * must display the following acknowledgement:
33 * "This product includes cryptographic software written by
34 * Eric Young (eay@cryptsoft.com)"
35 * The word 'cryptographic' can be left out if the rouines from the library
36 * being used are not cryptographic related :-).
37 * 4. If you include any Windows specific code (or a derivative thereof) from
38 * the apps directory (application code) you must include an acknowledgement:
39 * "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
41 * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
42 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
43 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
44 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
45 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
46 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
47 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
48 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
49 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
50 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
53 * The licence and distribution terms for any publically available version or
54 * derivative of this code cannot be changed. i.e. this code cannot simply be
55 * copied and put under another distribution licence
56 * [including the GNU Public Licence.]
59 #ifndef HEADER_DES_LOCL_H
60 #define HEADER_DES_LOCL_H
62 #if defined(WIN32) || defined(WIN16)
71 #include <openssl/opensslconf.h>
74 #if !defined(VMS) || defined(__DECC)
76 # include OPENSSL_UNISTD
83 #include <openssl/des.h>
85 #ifdef MSDOS /* Visual C++ 2.1 (Windows NT/95) */
92 #if defined(__STDC__) || defined(VMS) || defined(M_XENIX) || defined(MSDOS)
97 #define HALF_ITERATIONS 8
99 /* used in des_read and des_write */
100 #define MAXWRITE (1024*16)
101 #define BSIZE (MAXWRITE+4)
103 #define c2l(c,l) (l =((DES_LONG)(*((c)++))) , \
104 l|=((DES_LONG)(*((c)++)))<< 8L, \
105 l|=((DES_LONG)(*((c)++)))<<16L, \
106 l|=((DES_LONG)(*((c)++)))<<24L)
108 /* NOTE - c is not incremented as per c2l */
109 #define c2ln(c,l1,l2,n) { \
113 case 8: l2 =((DES_LONG)(*(--(c))))<<24L; \
114 case 7: l2|=((DES_LONG)(*(--(c))))<<16L; \
115 case 6: l2|=((DES_LONG)(*(--(c))))<< 8L; \
116 case 5: l2|=((DES_LONG)(*(--(c)))); \
117 case 4: l1 =((DES_LONG)(*(--(c))))<<24L; \
118 case 3: l1|=((DES_LONG)(*(--(c))))<<16L; \
119 case 2: l1|=((DES_LONG)(*(--(c))))<< 8L; \
120 case 1: l1|=((DES_LONG)(*(--(c)))); \
124 #define l2c(l,c) (*((c)++)=(unsigned char)(((l) )&0xff), \
125 *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \
126 *((c)++)=(unsigned char)(((l)>>16L)&0xff), \
127 *((c)++)=(unsigned char)(((l)>>24L)&0xff))
129 /* replacements for htonl and ntohl since I have no idea what to do
130 * when faced with machines with 8 byte longs. */
133 #define n2l(c,l) (l =((DES_LONG)(*((c)++)))<<24L, \
134 l|=((DES_LONG)(*((c)++)))<<16L, \
135 l|=((DES_LONG)(*((c)++)))<< 8L, \
136 l|=((DES_LONG)(*((c)++))))
138 #define l2n(l,c) (*((c)++)=(unsigned char)(((l)>>24L)&0xff), \
139 *((c)++)=(unsigned char)(((l)>>16L)&0xff), \
140 *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \
141 *((c)++)=(unsigned char)(((l) )&0xff))
143 /* NOTE - c is not incremented as per l2c */
144 #define l2cn(l1,l2,c,n) { \
147 case 8: *(--(c))=(unsigned char)(((l2)>>24L)&0xff); \
148 case 7: *(--(c))=(unsigned char)(((l2)>>16L)&0xff); \
149 case 6: *(--(c))=(unsigned char)(((l2)>> 8L)&0xff); \
150 case 5: *(--(c))=(unsigned char)(((l2) )&0xff); \
151 case 4: *(--(c))=(unsigned char)(((l1)>>24L)&0xff); \
152 case 3: *(--(c))=(unsigned char)(((l1)>>16L)&0xff); \
153 case 2: *(--(c))=(unsigned char)(((l1)>> 8L)&0xff); \
154 case 1: *(--(c))=(unsigned char)(((l1) )&0xff); \
159 #define ROTATE(a,n) (_lrotr(a,n))
161 #define ROTATE(a,n) (((a)>>(n))+((a)<<(32-(n))))
164 /* Don't worry about the LOAD_DATA() stuff, that is used by
165 * fcrypt() to add it's little bit to the front */
169 #define LOAD_DATA_tmp(R,S,u,t,E0,E1) \
170 { DES_LONG tmp; LOAD_DATA(R,S,u,t,E0,E1,tmp); }
172 #define LOAD_DATA(R,S,u,t,E0,E1,tmp) \
175 tmp=(u<<16); u^=R^s[S ]; u^=tmp; \
176 tmp=(t<<16); t^=R^s[S+1]; t^=tmp
178 #define LOAD_DATA_tmp(a,b,c,d,e,f) LOAD_DATA(a,b,c,d,e,f,g)
179 #define LOAD_DATA(R,S,u,t,E0,E1,tmp) \
184 /* The changes to this macro may help or hinder, depending on the
185 * compiler and the architecture. gcc2 always seems to do well :-).
186 * Inspired by Dana How <how@isl.stanford.edu>
187 * DO NOT use the alternative version on machines with 8 byte longs.
188 * It does not seem to work on the Alpha, even when DES_LONG is 4
189 * bytes, probably an issue of accessing non-word aligned objects :-( */
192 /* It recently occurred to me that 0^0^0^0^0^0^0 == 0, so there
193 * is no reason to not xor all the sub items together. This potentially
194 * saves a register since things can be xored directly into L */
196 #if defined(DES_RISC1) || defined(DES_RISC2)
198 #define D_ENCRYPT(LL,R,S) { \
199 unsigned int u1,u2,u3; \
200 LOAD_DATA(R,S,u,t,E0,E1,u1); \
206 LL^= *(const DES_LONG *)(des_SP +u1); \
207 LL^= *(const DES_LONG *)(des_SP+0x200+u2); \
211 LL^= *(const DES_LONG *)(des_SP+0x400+u1); \
212 LL^= *(const DES_LONG *)(des_SP+0x600+u3); \
217 LL^= *(const DES_LONG *)(des_SP+0x100+u1); \
218 LL^= *(const DES_LONG *)(des_SP+0x300+u2); \
222 LL^= *(const DES_LONG *)(des_SP+0x500+u1); \
223 LL^= *(const DES_LONG *)(des_SP+0x700+u3); }
226 #define D_ENCRYPT(LL,R,S) { \
227 unsigned int u1,u2,s1,s2; \
228 LOAD_DATA(R,S,u,t,E0,E1,u1); \
233 LL^= *(const DES_LONG *)(des_SP +u1); \
234 LL^= *(const DES_LONG *)(des_SP+0x200+u2); \
239 LL^= *(const DES_LONG *)(des_SP+0x400+s1); \
240 LL^= *(const DES_LONG *)(des_SP+0x600+s2); \
244 LL^= *(const DES_LONG *)(des_SP+0x100+u1); \
245 LL^= *(const DES_LONG *)(des_SP+0x300+u2); \
250 LL^= *(const DES_LONG *)(des_SP+0x500+s1); \
251 LL^= *(const DES_LONG *)(des_SP+0x700+s2); }
254 #define D_ENCRYPT(LL,R,S) { \
255 LOAD_DATA_tmp(R,S,u,t,E0,E1); \
258 *(const DES_LONG *)(des_SP +((u )&0xfc))^ \
259 *(const DES_LONG *)(des_SP+0x200+((u>> 8L)&0xfc))^ \
260 *(const DES_LONG *)(des_SP+0x400+((u>>16L)&0xfc))^ \
261 *(const DES_LONG *)(des_SP+0x600+((u>>24L)&0xfc))^ \
262 *(const DES_LONG *)(des_SP+0x100+((t )&0xfc))^ \
263 *(const DES_LONG *)(des_SP+0x300+((t>> 8L)&0xfc))^ \
264 *(const DES_LONG *)(des_SP+0x500+((t>>16L)&0xfc))^ \
265 *(const DES_LONG *)(des_SP+0x700+((t>>24L)&0xfc)); }
268 #else /* original version */
270 #if defined(DES_RISC1) || defined(DES_RISC2)
272 #define D_ENCRYPT(LL,R,S) {\
273 unsigned int u1,u2,u3; \
274 LOAD_DATA(R,S,u,t,E0,E1,u1); \
281 LL^=des_SPtrans[0][u1]; \
282 LL^=des_SPtrans[2][u2]; \
286 LL^=des_SPtrans[4][u1]; \
287 LL^=des_SPtrans[6][u3]; \
292 LL^=des_SPtrans[1][u1]; \
293 LL^=des_SPtrans[3][u2]; \
297 LL^=des_SPtrans[5][u1]; \
298 LL^=des_SPtrans[7][u3]; }
301 #define D_ENCRYPT(LL,R,S) {\
302 unsigned int u1,u2,s1,s2; \
303 LOAD_DATA(R,S,u,t,E0,E1,u1); \
309 LL^=des_SPtrans[0][u1]; \
310 LL^=des_SPtrans[2][u2]; \
315 LL^=des_SPtrans[4][s1]; \
316 LL^=des_SPtrans[6][s2]; \
320 LL^=des_SPtrans[1][u1]; \
321 LL^=des_SPtrans[3][u2]; \
326 LL^=des_SPtrans[5][s1]; \
327 LL^=des_SPtrans[7][s2]; }
332 #define D_ENCRYPT(LL,R,S) {\
333 LOAD_DATA_tmp(R,S,u,t,E0,E1); \
336 des_SPtrans[0][(u>> 2L)&0x3f]^ \
337 des_SPtrans[2][(u>>10L)&0x3f]^ \
338 des_SPtrans[4][(u>>18L)&0x3f]^ \
339 des_SPtrans[6][(u>>26L)&0x3f]^ \
340 des_SPtrans[1][(t>> 2L)&0x3f]^ \
341 des_SPtrans[3][(t>>10L)&0x3f]^ \
342 des_SPtrans[5][(t>>18L)&0x3f]^ \
343 des_SPtrans[7][(t>>26L)&0x3f]; }
348 * The problem is more of a geometric problem that random bit fiddling.
349 0 1 2 3 4 5 6 7 62 54 46 38 30 22 14 6
350 8 9 10 11 12 13 14 15 60 52 44 36 28 20 12 4
351 16 17 18 19 20 21 22 23 58 50 42 34 26 18 10 2
352 24 25 26 27 28 29 30 31 to 56 48 40 32 24 16 8 0
354 32 33 34 35 36 37 38 39 63 55 47 39 31 23 15 7
355 40 41 42 43 44 45 46 47 61 53 45 37 29 21 13 5
356 48 49 50 51 52 53 54 55 59 51 43 35 27 19 11 3
357 56 57 58 59 60 61 62 63 57 49 41 33 25 17 9 1
359 The output has been subject to swaps of the form
360 0 1 -> 3 1 but the odd and even bits have been put into
362 different words. The main trick is to remember that
363 t=((l>>size)^r)&(mask);
366 can be used to swap and move bits between words.
368 So l = 0 1 2 3 r = 16 17 18 19
370 8 9 10 11 24 25 26 27
371 12 13 14 15 28 29 30 31
372 becomes (for size == 2 and mask == 0x3333)
373 t = 2^16 3^17 -- -- l = 0 1 16 17 r = 2 3 18 19
374 6^20 7^21 -- -- 4 5 20 21 6 7 22 23
375 10^24 11^25 -- -- 8 9 24 25 10 11 24 25
376 14^28 15^29 -- -- 12 13 28 29 14 15 28 29
378 Thanks for hints from Richard Outerbridge - he told me IP&FP
379 could be done in 15 xor, 10 shifts and 5 ands.
380 When I finally started to think of the problem in 2D
381 I first got ~42 operations without xors. When I remembered
382 how to use xors :-) I got it to its final state.
384 #define PERM_OP(a,b,t,n,m) ((t)=((((a)>>(n))^(b))&(m)),\
390 register DES_LONG tt; \
391 PERM_OP(r,l,tt, 4,0x0f0f0f0fL); \
392 PERM_OP(l,r,tt,16,0x0000ffffL); \
393 PERM_OP(r,l,tt, 2,0x33333333L); \
394 PERM_OP(l,r,tt, 8,0x00ff00ffL); \
395 PERM_OP(r,l,tt, 1,0x55555555L); \
400 register DES_LONG tt; \
401 PERM_OP(l,r,tt, 1,0x55555555L); \
402 PERM_OP(r,l,tt, 8,0x00ff00ffL); \
403 PERM_OP(l,r,tt, 2,0x33333333L); \
404 PERM_OP(r,l,tt,16,0x0000ffffL); \
405 PERM_OP(l,r,tt, 4,0x0f0f0f0fL); \
408 OPENSSL_EXTERN const DES_LONG des_SPtrans[8][64];
410 void fcrypt_body(DES_LONG *out,des_key_schedule ks,
411 DES_LONG Eswap0, DES_LONG Eswap1);