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)
75 #include OPENSSL_UNISTD
79 #include <openssl/des.h>
81 #ifdef MSDOS /* Visual C++ 2.1 (Windows NT/95) */
88 #if defined(__STDC__) || defined(VMS) || defined(M_XENIX) || defined(MSDOS)
93 #define HALF_ITERATIONS 8
95 /* used in des_read and des_write */
96 #define MAXWRITE (1024*16)
97 #define BSIZE (MAXWRITE+4)
99 #define c2l(c,l) (l =((DES_LONG)(*((c)++))) , \
100 l|=((DES_LONG)(*((c)++)))<< 8L, \
101 l|=((DES_LONG)(*((c)++)))<<16L, \
102 l|=((DES_LONG)(*((c)++)))<<24L)
104 /* NOTE - c is not incremented as per c2l */
105 #define c2ln(c,l1,l2,n) { \
109 case 8: l2 =((DES_LONG)(*(--(c))))<<24L; \
110 case 7: l2|=((DES_LONG)(*(--(c))))<<16L; \
111 case 6: l2|=((DES_LONG)(*(--(c))))<< 8L; \
112 case 5: l2|=((DES_LONG)(*(--(c)))); \
113 case 4: l1 =((DES_LONG)(*(--(c))))<<24L; \
114 case 3: l1|=((DES_LONG)(*(--(c))))<<16L; \
115 case 2: l1|=((DES_LONG)(*(--(c))))<< 8L; \
116 case 1: l1|=((DES_LONG)(*(--(c)))); \
120 #define l2c(l,c) (*((c)++)=(unsigned char)(((l) )&0xff), \
121 *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \
122 *((c)++)=(unsigned char)(((l)>>16L)&0xff), \
123 *((c)++)=(unsigned char)(((l)>>24L)&0xff))
125 /* replacements for htonl and ntohl since I have no idea what to do
126 * when faced with machines with 8 byte longs. */
129 #define n2l(c,l) (l =((DES_LONG)(*((c)++)))<<24L, \
130 l|=((DES_LONG)(*((c)++)))<<16L, \
131 l|=((DES_LONG)(*((c)++)))<< 8L, \
132 l|=((DES_LONG)(*((c)++))))
134 #define l2n(l,c) (*((c)++)=(unsigned char)(((l)>>24L)&0xff), \
135 *((c)++)=(unsigned char)(((l)>>16L)&0xff), \
136 *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \
137 *((c)++)=(unsigned char)(((l) )&0xff))
139 /* NOTE - c is not incremented as per l2c */
140 #define l2cn(l1,l2,c,n) { \
143 case 8: *(--(c))=(unsigned char)(((l2)>>24L)&0xff); \
144 case 7: *(--(c))=(unsigned char)(((l2)>>16L)&0xff); \
145 case 6: *(--(c))=(unsigned char)(((l2)>> 8L)&0xff); \
146 case 5: *(--(c))=(unsigned char)(((l2) )&0xff); \
147 case 4: *(--(c))=(unsigned char)(((l1)>>24L)&0xff); \
148 case 3: *(--(c))=(unsigned char)(((l1)>>16L)&0xff); \
149 case 2: *(--(c))=(unsigned char)(((l1)>> 8L)&0xff); \
150 case 1: *(--(c))=(unsigned char)(((l1) )&0xff); \
155 #define ROTATE(a,n) (_lrotr(a,n))
157 #define ROTATE(a,n) (((a)>>(n))+((a)<<(32-(n))))
160 /* Don't worry about the LOAD_DATA() stuff, that is used by
161 * fcrypt() to add it's little bit to the front */
165 #define LOAD_DATA_tmp(R,S,u,t,E0,E1) \
166 { DES_LONG tmp; LOAD_DATA(R,S,u,t,E0,E1,tmp); }
168 #define LOAD_DATA(R,S,u,t,E0,E1,tmp) \
171 tmp=(u<<16); u^=R^s[S ]; u^=tmp; \
172 tmp=(t<<16); t^=R^s[S+1]; t^=tmp
174 #define LOAD_DATA_tmp(a,b,c,d,e,f) LOAD_DATA(a,b,c,d,e,f,g)
175 #define LOAD_DATA(R,S,u,t,E0,E1,tmp) \
180 /* The changes to this macro may help or hinder, depending on the
181 * compiler and the achitecture. gcc2 always seems to do well :-).
182 * Inspired by Dana How <how@isl.stanford.edu>
183 * DO NOT use the alternative version on machines with 8 byte longs.
184 * It does not seem to work on the Alpha, even when DES_LONG is 4
185 * bytes, probably an issue of accessing non-word aligned objects :-( */
188 /* It recently occured to me that 0^0^0^0^0^0^0 == 0, so there
189 * is no reason to not xor all the sub items together. This potentially
190 * saves a register since things can be xored directly into L */
192 #if defined(DES_RISC1) || defined(DES_RISC2)
194 #define D_ENCRYPT(LL,R,S) { \
195 unsigned int u1,u2,u3; \
196 LOAD_DATA(R,S,u,t,E0,E1,u1); \
202 LL^= *(const DES_LONG *)(des_SP +u1); \
203 LL^= *(const DES_LONG *)(des_SP+0x200+u2); \
207 LL^= *(const DES_LONG *)(des_SP+0x400+u1); \
208 LL^= *(const DES_LONG *)(des_SP+0x600+u3); \
213 LL^= *(const DES_LONG *)(des_SP+0x100+u1); \
214 LL^= *(const DES_LONG *)(des_SP+0x300+u2); \
218 LL^= *(const DES_LONG *)(des_SP+0x500+u1); \
219 LL^= *(const DES_LONG *)(des_SP+0x700+u3); }
222 #define D_ENCRYPT(LL,R,S) { \
223 unsigned int u1,u2,s1,s2; \
224 LOAD_DATA(R,S,u,t,E0,E1,u1); \
229 LL^= *(const DES_LONG *)(des_SP +u1); \
230 LL^= *(const DES_LONG *)(des_SP+0x200+u2); \
235 LL^= *(const DES_LONG *)(des_SP+0x400+s1); \
236 LL^= *(const DES_LONG *)(des_SP+0x600+s2); \
240 LL^= *(const DES_LONG *)(des_SP+0x100+u1); \
241 LL^= *(const DES_LONG *)(des_SP+0x300+u2); \
246 LL^= *(const DES_LONG *)(des_SP+0x500+s1); \
247 LL^= *(const DES_LONG *)(des_SP+0x700+s2); }
250 #define D_ENCRYPT(LL,R,S) { \
251 LOAD_DATA_tmp(R,S,u,t,E0,E1); \
254 *(const DES_LONG *)(des_SP +((u )&0xfc))^ \
255 *(const DES_LONG *)(des_SP+0x200+((u>> 8L)&0xfc))^ \
256 *(const DES_LONG *)(des_SP+0x400+((u>>16L)&0xfc))^ \
257 *(const DES_LONG *)(des_SP+0x600+((u>>24L)&0xfc))^ \
258 *(const DES_LONG *)(des_SP+0x100+((t )&0xfc))^ \
259 *(const DES_LONG *)(des_SP+0x300+((t>> 8L)&0xfc))^ \
260 *(const DES_LONG *)(des_SP+0x500+((t>>16L)&0xfc))^ \
261 *(const DES_LONG *)(des_SP+0x700+((t>>24L)&0xfc)); }
264 #else /* original version */
266 #if defined(DES_RISC1) || defined(DES_RISC2)
268 #define D_ENCRYPT(LL,R,S) {\
269 unsigned int u1,u2,u3; \
270 LOAD_DATA(R,S,u,t,E0,E1,u1); \
277 LL^=des_SPtrans[0][u1]; \
278 LL^=des_SPtrans[2][u2]; \
282 LL^=des_SPtrans[4][u1]; \
283 LL^=des_SPtrans[6][u3]; \
288 LL^=des_SPtrans[1][u1]; \
289 LL^=des_SPtrans[3][u2]; \
293 LL^=des_SPtrans[5][u1]; \
294 LL^=des_SPtrans[7][u3]; }
297 #define D_ENCRYPT(LL,R,S) {\
298 unsigned int u1,u2,s1,s2; \
299 LOAD_DATA(R,S,u,t,E0,E1,u1); \
305 LL^=des_SPtrans[0][u1]; \
306 LL^=des_SPtrans[2][u2]; \
311 LL^=des_SPtrans[4][s1]; \
312 LL^=des_SPtrans[6][s2]; \
316 LL^=des_SPtrans[1][u1]; \
317 LL^=des_SPtrans[3][u2]; \
322 LL^=des_SPtrans[5][s1]; \
323 LL^=des_SPtrans[7][s2]; }
328 #define D_ENCRYPT(LL,R,S) {\
329 LOAD_DATA_tmp(R,S,u,t,E0,E1); \
332 des_SPtrans[0][(u>> 2L)&0x3f]^ \
333 des_SPtrans[2][(u>>10L)&0x3f]^ \
334 des_SPtrans[4][(u>>18L)&0x3f]^ \
335 des_SPtrans[6][(u>>26L)&0x3f]^ \
336 des_SPtrans[1][(t>> 2L)&0x3f]^ \
337 des_SPtrans[3][(t>>10L)&0x3f]^ \
338 des_SPtrans[5][(t>>18L)&0x3f]^ \
339 des_SPtrans[7][(t>>26L)&0x3f]; }
344 * The problem is more of a geometric problem that random bit fiddling.
345 0 1 2 3 4 5 6 7 62 54 46 38 30 22 14 6
346 8 9 10 11 12 13 14 15 60 52 44 36 28 20 12 4
347 16 17 18 19 20 21 22 23 58 50 42 34 26 18 10 2
348 24 25 26 27 28 29 30 31 to 56 48 40 32 24 16 8 0
350 32 33 34 35 36 37 38 39 63 55 47 39 31 23 15 7
351 40 41 42 43 44 45 46 47 61 53 45 37 29 21 13 5
352 48 49 50 51 52 53 54 55 59 51 43 35 27 19 11 3
353 56 57 58 59 60 61 62 63 57 49 41 33 25 17 9 1
355 The output has been subject to swaps of the form
356 0 1 -> 3 1 but the odd and even bits have been put into
358 different words. The main trick is to remember that
359 t=((l>>size)^r)&(mask);
362 can be used to swap and move bits between words.
364 So l = 0 1 2 3 r = 16 17 18 19
366 8 9 10 11 24 25 26 27
367 12 13 14 15 28 29 30 31
368 becomes (for size == 2 and mask == 0x3333)
369 t = 2^16 3^17 -- -- l = 0 1 16 17 r = 2 3 18 19
370 6^20 7^21 -- -- 4 5 20 21 6 7 22 23
371 10^24 11^25 -- -- 8 9 24 25 10 11 24 25
372 14^28 15^29 -- -- 12 13 28 29 14 15 28 29
374 Thanks for hints from Richard Outerbridge - he told me IP&FP
375 could be done in 15 xor, 10 shifts and 5 ands.
376 When I finally started to think of the problem in 2D
377 I first got ~42 operations without xors. When I remembered
378 how to use xors :-) I got it to its final state.
380 #define PERM_OP(a,b,t,n,m) ((t)=((((a)>>(n))^(b))&(m)),\
386 register DES_LONG tt; \
387 PERM_OP(r,l,tt, 4,0x0f0f0f0fL); \
388 PERM_OP(l,r,tt,16,0x0000ffffL); \
389 PERM_OP(r,l,tt, 2,0x33333333L); \
390 PERM_OP(l,r,tt, 8,0x00ff00ffL); \
391 PERM_OP(r,l,tt, 1,0x55555555L); \
396 register DES_LONG tt; \
397 PERM_OP(l,r,tt, 1,0x55555555L); \
398 PERM_OP(r,l,tt, 8,0x00ff00ffL); \
399 PERM_OP(l,r,tt, 2,0x33333333L); \
400 PERM_OP(r,l,tt,16,0x0000ffffL); \
401 PERM_OP(l,r,tt, 4,0x0f0f0f0fL); \
404 OPENSSL_EXTERN const DES_LONG des_SPtrans[8][64];
406 void fcrypt_body(DES_LONG *out,des_key_schedule ks,
407 DES_LONG Eswap0, DES_LONG Eswap1);