1 /* crypto/des/des_locl.org */
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 /* WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING
61 * Always modify des_locl.org since des_locl.h is automatically generated from
62 * it during SSLeay configuration.
64 * WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING
67 #ifndef HEADER_DES_LOCL_H
68 #define HEADER_DES_LOCL_H
70 #if defined(WIN32) || defined(WIN16)
78 #include "hw_assist.h"
81 #ifndef DES_DEFAULT_OPTIONS
82 /* the following is tweaked from a config script, that is why it is a
83 * protected undef/define */
88 /* This helps C compiler generate the correct code for multiple functional
89 * units. It reduces register dependancies at the expense of 2 more
99 #if defined(DES_RISC1) && defined(DES_RISC2)
100 YOU SHOULD NOT HAVE BOTH DES_RISC1 AND DES_RISC2 DEFINED!!!!!
103 /* Unroll the inner loop, this sometimes helps, sometimes hinders.
104 * Very mucy CPU dependant */
109 /* These default values were supplied by
110 * Peter Gutman <pgut001@cs.auckland.ac.nz>
111 * They are only used if nothing else has been defined */
112 #if !defined(DES_PTR) && !defined(DES_RISC1) && !defined(DES_RISC2) && !defined(DES_UNROLL)
113 /* Special defines which change the way the code is built depending on the
114 CPU and OS. For SGI machines you can use _MIPS_SZLONG (32 or 64) to find
115 even newer MIPS CPU's, but at the moment one size fits all for
116 optimization options. Older Sparc's work better with only UNROLL, but
117 there's no way to tell at compile time what it is you're running on */
119 #if defined( sun ) /* Newer Sparc's */
123 #elif defined( __ultrix ) /* Older MIPS */
127 #elif defined( __osf1__ ) /* Alpha */
130 #elif defined ( _AIX ) /* RS6000 */
132 #elif defined( __hpux ) /* HP-PA */
134 #elif defined( __aux ) /* 68K */
136 #elif defined( __dgux ) /* 88K (but P6 in latest boxes) */
138 #elif defined( __sgi ) /* Newer MIPS */
142 #elif defined( i386 ) /* x86 boxes, should be gcc */
146 #elif defined( __sh3__ ) /* Hitachi SH3 */
148 #elif defined(mcf5307) /* ColdFire 5307 */
150 #elif defined(mcf5200) /* ColdFire 52xx */
151 #define DES_PTR /* Not definitely sure this is best */
153 /* These are the defaults that were previously applied so I've preserved
159 #endif /* Systems-specific speed defines */
162 #endif /* DES_DEFAULT_OPTIONS */
164 #ifdef MSDOS /* Visual C++ 2.1 (Windows NT/95) */
175 #if defined(__STDC__) || defined(VMS) || defined(M_XENIX) || defined(MSDOS)
179 #include <linux/string.h>
206 #define srandom(s) srand(s)
210 #define ITERATIONS 16
211 #define HALF_ITERATIONS 8
213 /* used in des_read and des_write */
214 #define MAXWRITE (1024*16)
215 #define BSIZE (MAXWRITE+4)
217 #define c2l(c,l) (l =((DES_LONG)(*((c)++))) , \
218 l|=((DES_LONG)(*((c)++)))<< 8L, \
219 l|=((DES_LONG)(*((c)++)))<<16L, \
220 l|=((DES_LONG)(*((c)++)))<<24L)
222 /* NOTE - c is not incremented as per c2l */
223 #define c2ln(c,l1,l2,n) { \
227 case 8: l2 =((DES_LONG)(*(--(c))))<<24L; \
228 case 7: l2|=((DES_LONG)(*(--(c))))<<16L; \
229 case 6: l2|=((DES_LONG)(*(--(c))))<< 8L; \
230 case 5: l2|=((DES_LONG)(*(--(c)))); \
231 case 4: l1 =((DES_LONG)(*(--(c))))<<24L; \
232 case 3: l1|=((DES_LONG)(*(--(c))))<<16L; \
233 case 2: l1|=((DES_LONG)(*(--(c))))<< 8L; \
234 case 1: l1|=((DES_LONG)(*(--(c)))); \
238 #define l2c(l,c) (*((c)++)=(unsigned char)(((l) )&0xff), \
239 *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \
240 *((c)++)=(unsigned char)(((l)>>16L)&0xff), \
241 *((c)++)=(unsigned char)(((l)>>24L)&0xff))
243 /* replacements for htonl and ntohl since I have no idea what to do
244 * when faced with machines with 8 byte longs. */
247 #define n2l(c,l) (l =((DES_LONG)(*((c)++)))<<24L, \
248 l|=((DES_LONG)(*((c)++)))<<16L, \
249 l|=((DES_LONG)(*((c)++)))<< 8L, \
250 l|=((DES_LONG)(*((c)++))))
252 #define l2n(l,c) (*((c)++)=(unsigned char)(((l)>>24L)&0xff), \
253 *((c)++)=(unsigned char)(((l)>>16L)&0xff), \
254 *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \
255 *((c)++)=(unsigned char)(((l) )&0xff))
257 /* NOTE - c is not incremented as per l2c */
258 #define l2cn(l1,l2,c,n) { \
261 case 8: *(--(c))=(unsigned char)(((l2)>>24L)&0xff); \
262 case 7: *(--(c))=(unsigned char)(((l2)>>16L)&0xff); \
263 case 6: *(--(c))=(unsigned char)(((l2)>> 8L)&0xff); \
264 case 5: *(--(c))=(unsigned char)(((l2) )&0xff); \
265 case 4: *(--(c))=(unsigned char)(((l1)>>24L)&0xff); \
266 case 3: *(--(c))=(unsigned char)(((l1)>>16L)&0xff); \
267 case 2: *(--(c))=(unsigned char)(((l1)>> 8L)&0xff); \
268 case 1: *(--(c))=(unsigned char)(((l1) )&0xff); \
273 #define ROTATE(a,n) (_lrotr(a,n))
275 #define ROTATE(a,n) (((a)>>(n))+((a)<<(32-(n))))
278 /* Don't worry about the LOAD_DATA() stuff, that is used by
279 * fcrypt() to add it's little bit to the front */
283 #define LOAD_DATA_tmp(R,S,u,t,E0,E1) \
284 { DES_LONG tmp; LOAD_DATA(R,S,u,t,E0,E1,tmp); }
286 #define LOAD_DATA(R,S,u,t,E0,E1,tmp) \
289 tmp=(u<<16); u^=R^s[S ]; u^=tmp; \
290 tmp=(t<<16); t^=R^s[S+1]; t^=tmp
292 #define LOAD_DATA_tmp(a,b,c,d,e,f) LOAD_DATA(a,b,c,d,e,f,g)
293 #define LOAD_DATA(R,S,u,t,E0,E1,tmp) \
298 /* The changes to this macro may help or hinder, depending on the
299 * compiler and the achitecture. gcc2 always seems to do well :-).
300 * Inspired by Dana How <how@isl.stanford.edu>
301 * DO NOT use the alternative version on machines with 8 byte longs.
302 * It does not seem to work on the Alpha, even when DES_LONG is 4
303 * bytes, probably an issue of accessing non-word aligned objects :-( */
306 /* It recently occured to me that 0^0^0^0^0^0^0 == 0, so there
307 * is no reason to not xor all the sub items together. This potentially
308 * saves a register since things can be xored directly into L */
310 #if defined(DES_RISC1) || defined(DES_RISC2)
312 #define D_ENCRYPT(LL,R,S) { \
313 unsigned int u1,u2,u3; \
314 LOAD_DATA(R,S,u,t,E0,E1,u1); \
320 LL^= *(DES_LONG *)((unsigned char *)des_SP +u1); \
321 LL^= *(DES_LONG *)((unsigned char *)des_SP+0x200+u2); \
325 LL^= *(DES_LONG *)((unsigned char *)des_SP+0x400+u1); \
326 LL^= *(DES_LONG *)((unsigned char *)des_SP+0x600+u3); \
331 LL^= *(DES_LONG *)((unsigned char *)des_SP+0x100+u1); \
332 LL^= *(DES_LONG *)((unsigned char *)des_SP+0x300+u2); \
336 LL^= *(DES_LONG *)((unsigned char *)des_SP+0x500+u1); \
337 LL^= *(DES_LONG *)((unsigned char *)des_SP+0x700+u3); }
340 #define D_ENCRYPT(LL,R,S) { \
341 unsigned int u1,u2,s1,s2; \
342 LOAD_DATA(R,S,u,t,E0,E1,u1); \
347 LL^= *(DES_LONG *)((unsigned char *)des_SP +u1); \
348 LL^= *(DES_LONG *)((unsigned char *)des_SP+0x200+u2); \
353 LL^= *(DES_LONG *)((unsigned char *)des_SP+0x400+s1); \
354 LL^= *(DES_LONG *)((unsigned char *)des_SP+0x600+s2); \
358 LL^= *(DES_LONG *)((unsigned char *)des_SP+0x100+u1); \
359 LL^= *(DES_LONG *)((unsigned char *)des_SP+0x300+u2); \
364 LL^= *(DES_LONG *)((unsigned char *)des_SP+0x500+s1); \
365 LL^= *(DES_LONG *)((unsigned char *)des_SP+0x700+s2); }
368 #define D_ENCRYPT(LL,R,S) { \
369 LOAD_DATA_tmp(R,S,u,t,E0,E1); \
372 *(DES_LONG *)((unsigned char *)des_SP +((u )&0xfc))^ \
373 *(DES_LONG *)((unsigned char *)des_SP+0x200+((u>> 8L)&0xfc))^ \
374 *(DES_LONG *)((unsigned char *)des_SP+0x400+((u>>16L)&0xfc))^ \
375 *(DES_LONG *)((unsigned char *)des_SP+0x600+((u>>24L)&0xfc))^ \
376 *(DES_LONG *)((unsigned char *)des_SP+0x100+((t )&0xfc))^ \
377 *(DES_LONG *)((unsigned char *)des_SP+0x300+((t>> 8L)&0xfc))^ \
378 *(DES_LONG *)((unsigned char *)des_SP+0x500+((t>>16L)&0xfc))^ \
379 *(DES_LONG *)((unsigned char *)des_SP+0x700+((t>>24L)&0xfc)); }
382 #else /* original version */
384 #if defined(DES_RISC1) || defined(DES_RISC2)
386 #define D_ENCRYPT(LL,R,S) {\
387 unsigned int u1,u2,u3; \
388 LOAD_DATA(R,S,u,t,E0,E1,u1); \
395 LL^=des_SPtrans[0][u1]; \
396 LL^=des_SPtrans[2][u2]; \
400 LL^=des_SPtrans[4][u1]; \
401 LL^=des_SPtrans[6][u3]; \
406 LL^=des_SPtrans[1][u1]; \
407 LL^=des_SPtrans[3][u2]; \
411 LL^=des_SPtrans[5][u1]; \
412 LL^=des_SPtrans[7][u3]; }
415 #define D_ENCRYPT(LL,R,S) {\
416 unsigned int u1,u2,s1,s2; \
417 LOAD_DATA(R,S,u,t,E0,E1,u1); \
423 LL^=des_SPtrans[0][u1]; \
424 LL^=des_SPtrans[2][u2]; \
429 LL^=des_SPtrans[4][s1]; \
430 LL^=des_SPtrans[6][s2]; \
434 LL^=des_SPtrans[1][u1]; \
435 LL^=des_SPtrans[3][u2]; \
440 LL^=des_SPtrans[5][s1]; \
441 LL^=des_SPtrans[7][s2]; }
446 #define D_ENCRYPT(LL,R,S) {\
447 LOAD_DATA_tmp(R,S,u,t,E0,E1); \
450 des_SPtrans[0][(u>> 2L)&0x3f]^ \
451 des_SPtrans[2][(u>>10L)&0x3f]^ \
452 des_SPtrans[4][(u>>18L)&0x3f]^ \
453 des_SPtrans[6][(u>>26L)&0x3f]^ \
454 des_SPtrans[1][(t>> 2L)&0x3f]^ \
455 des_SPtrans[3][(t>>10L)&0x3f]^ \
456 des_SPtrans[5][(t>>18L)&0x3f]^ \
457 des_SPtrans[7][(t>>26L)&0x3f]; }
462 * The problem is more of a geometric problem that random bit fiddling.
463 0 1 2 3 4 5 6 7 62 54 46 38 30 22 14 6
464 8 9 10 11 12 13 14 15 60 52 44 36 28 20 12 4
465 16 17 18 19 20 21 22 23 58 50 42 34 26 18 10 2
466 24 25 26 27 28 29 30 31 to 56 48 40 32 24 16 8 0
468 32 33 34 35 36 37 38 39 63 55 47 39 31 23 15 7
469 40 41 42 43 44 45 46 47 61 53 45 37 29 21 13 5
470 48 49 50 51 52 53 54 55 59 51 43 35 27 19 11 3
471 56 57 58 59 60 61 62 63 57 49 41 33 25 17 9 1
473 The output has been subject to swaps of the form
474 0 1 -> 3 1 but the odd and even bits have been put into
476 different words. The main trick is to remember that
477 t=((l>>size)^r)&(mask);
480 can be used to swap and move bits between words.
482 So l = 0 1 2 3 r = 16 17 18 19
484 8 9 10 11 24 25 26 27
485 12 13 14 15 28 29 30 31
486 becomes (for size == 2 and mask == 0x3333)
487 t = 2^16 3^17 -- -- l = 0 1 16 17 r = 2 3 18 19
488 6^20 7^21 -- -- 4 5 20 21 6 7 22 23
489 10^24 11^25 -- -- 8 9 24 25 10 11 24 25
490 14^28 15^29 -- -- 12 13 28 29 14 15 28 29
492 Thanks for hints from Richard Outerbridge - he told me IP&FP
493 could be done in 15 xor, 10 shifts and 5 ands.
494 When I finally started to think of the problem in 2D
495 I first got ~42 operations without xors. When I remembered
496 how to use xors :-) I got it to its final state.
498 #define PERM_OP(a,b,t,n,m) ((t)=((((a)>>(n))^(b))&(m)),\
504 register DES_LONG tt; \
505 PERM_OP(r,l,tt, 4,0x0f0f0f0fL); \
506 PERM_OP(l,r,tt,16,0x0000ffffL); \
507 PERM_OP(r,l,tt, 2,0x33333333L); \
508 PERM_OP(l,r,tt, 8,0x00ff00ffL); \
509 PERM_OP(r,l,tt, 1,0x55555555L); \
514 register DES_LONG tt; \
515 PERM_OP(l,r,tt, 1,0x55555555L); \
516 PERM_OP(r,l,tt, 8,0x00ff00ffL); \
517 PERM_OP(l,r,tt, 2,0x33333333L); \
518 PERM_OP(r,l,tt,16,0x0000ffffL); \
519 PERM_OP(l,r,tt, 4,0x0f0f0f0fL); \
522 extern const DES_LONG des_SPtrans[8][64];
526 void fcrypt_body(DES_LONG *out,des_key_schedule ks,
527 DES_LONG Eswap0, DES_LONG Eswap1);
531 #endif /* NO_FCRYPT */