1 /* $OpenBSD: des_locl.h,v 1.17 2014/08/18 19:15:34 bcook Exp $ */
2 /* Copyright (C) 1995-1997 Eric Young (eay@cryptsoft.com)
3  * All rights reserved.
4  *
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.
8  *
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).
15  *
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.
22  *
23  * Redistribution and use in source and binary forms, with or without
24  * modification, are permitted provided that the following conditions
25  * are met:
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)"
40  *
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
51  * SUCH DAMAGE.
52  *
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.]
57  */
58 
59 #ifndef HEADER_DES_LOCL_H
60 #define HEADER_DES_LOCL_H
61 
62 #include <math.h>
63 #include <stdint.h>
64 #include <stdio.h>
65 #include <stdlib.h>
66 #include <string.h>
67 #include <unistd.h>
68 
69 #include <openssl/opensslconf.h>
70 
71 #include <openssl/des.h>
72 
73 #define ITERATIONS 16
74 #define HALF_ITERATIONS 8
75 
76 /* used in des_read and des_write */
77 #define MAXWRITE	(1024*16)
78 #define BSIZE		(MAXWRITE+4)
79 
80 #define c2l(c,l)	(l =((DES_LONG)(*((c)++)))    , \
81 			 l|=((DES_LONG)(*((c)++)))<< 8L, \
82 			 l|=((DES_LONG)(*((c)++)))<<16L, \
83 			 l|=((DES_LONG)(*((c)++)))<<24L)
84 
85 /* NOTE - c is not incremented as per c2l */
86 #define c2ln(c,l1,l2,n)	{ \
87 			c+=n; \
88 			l1=l2=0; \
89 			switch (n) { \
90 			case 8: l2 =((DES_LONG)(*(--(c))))<<24L; \
91 			case 7: l2|=((DES_LONG)(*(--(c))))<<16L; \
92 			case 6: l2|=((DES_LONG)(*(--(c))))<< 8L; \
93 			case 5: l2|=((DES_LONG)(*(--(c))));     \
94 			case 4: l1 =((DES_LONG)(*(--(c))))<<24L; \
95 			case 3: l1|=((DES_LONG)(*(--(c))))<<16L; \
96 			case 2: l1|=((DES_LONG)(*(--(c))))<< 8L; \
97 			case 1: l1|=((DES_LONG)(*(--(c))));     \
98 				} \
99 			}
100 
101 #define l2c(l,c)	(*((c)++)=(unsigned char)(((l)     )&0xff), \
102 			 *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \
103 			 *((c)++)=(unsigned char)(((l)>>16L)&0xff), \
104 			 *((c)++)=(unsigned char)(((l)>>24L)&0xff))
105 
106 /* replacements for htonl and ntohl since I have no idea what to do
107  * when faced with machines with 8 byte longs. */
108 #define HDRSIZE 4
109 
110 #define n2l(c,l)	(l =((DES_LONG)(*((c)++)))<<24L, \
111 			 l|=((DES_LONG)(*((c)++)))<<16L, \
112 			 l|=((DES_LONG)(*((c)++)))<< 8L, \
113 			 l|=((DES_LONG)(*((c)++))))
114 
115 #define l2n(l,c)	(*((c)++)=(unsigned char)(((l)>>24L)&0xff), \
116 			 *((c)++)=(unsigned char)(((l)>>16L)&0xff), \
117 			 *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \
118 			 *((c)++)=(unsigned char)(((l)     )&0xff))
119 
120 /* NOTE - c is not incremented as per l2c */
121 #define l2cn(l1,l2,c,n)	{ \
122 			c+=n; \
123 			switch (n) { \
124 			case 8: *(--(c))=(unsigned char)(((l2)>>24L)&0xff); \
125 			case 7: *(--(c))=(unsigned char)(((l2)>>16L)&0xff); \
126 			case 6: *(--(c))=(unsigned char)(((l2)>> 8L)&0xff); \
127 			case 5: *(--(c))=(unsigned char)(((l2)     )&0xff); \
128 			case 4: *(--(c))=(unsigned char)(((l1)>>24L)&0xff); \
129 			case 3: *(--(c))=(unsigned char)(((l1)>>16L)&0xff); \
130 			case 2: *(--(c))=(unsigned char)(((l1)>> 8L)&0xff); \
131 			case 1: *(--(c))=(unsigned char)(((l1)     )&0xff); \
132 				} \
133 			}
134 
135 static inline uint32_t ROTATE(uint32_t a, uint32_t n)
136 {
137 	return (a>>n)+(a<<(32-n));
138 }
139 
140 /* Don't worry about the LOAD_DATA() stuff, that is used by
141  * fcrypt() to add it's little bit to the front */
142 
143 #ifdef DES_FCRYPT
144 
145 #define LOAD_DATA_tmp(R,S,u,t,E0,E1) \
146 	{ DES_LONG tmp; LOAD_DATA(R,S,u,t,E0,E1,tmp); }
147 
148 #define LOAD_DATA(R,S,u,t,E0,E1,tmp) \
149 	t=R^(R>>16L); \
150 	u=t&E0; t&=E1; \
151 	tmp=(u<<16); u^=R^s[S  ]; u^=tmp; \
152 	tmp=(t<<16); t^=R^s[S+1]; t^=tmp
153 #else
154 #define LOAD_DATA_tmp(a,b,c,d,e,f) LOAD_DATA(a,b,c,d,e,f,g)
155 #define LOAD_DATA(R,S,u,t,E0,E1,tmp) \
156 	u=R^s[S  ]; \
157 	t=R^s[S+1]
158 #endif
159 
160 /* The changes to this macro may help or hinder, depending on the
161  * compiler and the architecture.  gcc2 always seems to do well :-).
162  * Inspired by Dana How <how@isl.stanford.edu>
163  * DO NOT use the alternative version on machines with 8 byte longs.
164  * It does not seem to work on the Alpha, even when DES_LONG is 4
165  * bytes, probably an issue of accessing non-word aligned objects :-( */
166 #ifdef DES_PTR
167 
168 /* It recently occurred to me that 0^0^0^0^0^0^0 == 0, so there
169  * is no reason to not xor all the sub items together.  This potentially
170  * saves a register since things can be xored directly into L */
171 
172 #if defined(DES_RISC1) || defined(DES_RISC2)
173 #ifdef DES_RISC1
174 #define D_ENCRYPT(LL,R,S) { \
175 	unsigned int u1,u2,u3; \
176 	LOAD_DATA(R,S,u,t,E0,E1,u1); \
177 	u2=(int)u>>8L; \
178 	u1=(int)u&0xfc; \
179 	u2&=0xfc; \
180 	t=ROTATE(t,4); \
181 	u>>=16L; \
182 	LL^= *(const DES_LONG *)(des_SP      +u1); \
183 	LL^= *(const DES_LONG *)(des_SP+0x200+u2); \
184 	u3=(int)(u>>8L); \
185 	u1=(int)u&0xfc; \
186 	u3&=0xfc; \
187 	LL^= *(const DES_LONG *)(des_SP+0x400+u1); \
188 	LL^= *(const DES_LONG *)(des_SP+0x600+u3); \
189 	u2=(int)t>>8L; \
190 	u1=(int)t&0xfc; \
191 	u2&=0xfc; \
192 	t>>=16L; \
193 	LL^= *(const DES_LONG *)(des_SP+0x100+u1); \
194 	LL^= *(const DES_LONG *)(des_SP+0x300+u2); \
195 	u3=(int)t>>8L; \
196 	u1=(int)t&0xfc; \
197 	u3&=0xfc; \
198 	LL^= *(const DES_LONG *)(des_SP+0x500+u1); \
199 	LL^= *(const DES_LONG *)(des_SP+0x700+u3); }
200 #endif
201 #ifdef DES_RISC2
202 #define D_ENCRYPT(LL,R,S) { \
203 	unsigned int u1,u2,s1,s2; \
204 	LOAD_DATA(R,S,u,t,E0,E1,u1); \
205 	u2=(int)u>>8L; \
206 	u1=(int)u&0xfc; \
207 	u2&=0xfc; \
208 	t=ROTATE(t,4); \
209 	LL^= *(const DES_LONG *)(des_SP      +u1); \
210 	LL^= *(const DES_LONG *)(des_SP+0x200+u2); \
211 	s1=(int)(u>>16L); \
212 	s2=(int)(u>>24L); \
213 	s1&=0xfc; \
214 	s2&=0xfc; \
215 	LL^= *(const DES_LONG *)(des_SP+0x400+s1); \
216 	LL^= *(const DES_LONG *)(des_SP+0x600+s2); \
217 	u2=(int)t>>8L; \
218 	u1=(int)t&0xfc; \
219 	u2&=0xfc; \
220 	LL^= *(const DES_LONG *)(des_SP+0x100+u1); \
221 	LL^= *(const DES_LONG *)(des_SP+0x300+u2); \
222 	s1=(int)(t>>16L); \
223 	s2=(int)(t>>24L); \
224 	s1&=0xfc; \
225 	s2&=0xfc; \
226 	LL^= *(const DES_LONG *)(des_SP+0x500+s1); \
227 	LL^= *(const DES_LONG *)(des_SP+0x700+s2); }
228 #endif
229 #else
230 #define D_ENCRYPT(LL,R,S) { \
231 	LOAD_DATA_tmp(R,S,u,t,E0,E1); \
232 	t=ROTATE(t,4); \
233 	LL^= \
234 	*(const DES_LONG *)(des_SP      +((u     )&0xfc))^ \
235 	*(const DES_LONG *)(des_SP+0x200+((u>> 8L)&0xfc))^ \
236 	*(const DES_LONG *)(des_SP+0x400+((u>>16L)&0xfc))^ \
237 	*(const DES_LONG *)(des_SP+0x600+((u>>24L)&0xfc))^ \
238 	*(const DES_LONG *)(des_SP+0x100+((t     )&0xfc))^ \
239 	*(const DES_LONG *)(des_SP+0x300+((t>> 8L)&0xfc))^ \
240 	*(const DES_LONG *)(des_SP+0x500+((t>>16L)&0xfc))^ \
241 	*(const DES_LONG *)(des_SP+0x700+((t>>24L)&0xfc)); }
242 #endif
243 
244 #else /* original version */
245 
246 #if defined(DES_RISC1) || defined(DES_RISC2)
247 #ifdef DES_RISC1
248 #define D_ENCRYPT(LL,R,S) {\
249 	unsigned int u1,u2,u3; \
250 	LOAD_DATA(R,S,u,t,E0,E1,u1); \
251 	u>>=2L; \
252 	t=ROTATE(t,6); \
253 	u2=(int)u>>8L; \
254 	u1=(int)u&0x3f; \
255 	u2&=0x3f; \
256 	u>>=16L; \
257 	LL^=DES_SPtrans[0][u1]; \
258 	LL^=DES_SPtrans[2][u2]; \
259 	u3=(int)u>>8L; \
260 	u1=(int)u&0x3f; \
261 	u3&=0x3f; \
262 	LL^=DES_SPtrans[4][u1]; \
263 	LL^=DES_SPtrans[6][u3]; \
264 	u2=(int)t>>8L; \
265 	u1=(int)t&0x3f; \
266 	u2&=0x3f; \
267 	t>>=16L; \
268 	LL^=DES_SPtrans[1][u1]; \
269 	LL^=DES_SPtrans[3][u2]; \
270 	u3=(int)t>>8L; \
271 	u1=(int)t&0x3f; \
272 	u3&=0x3f; \
273 	LL^=DES_SPtrans[5][u1]; \
274 	LL^=DES_SPtrans[7][u3]; }
275 #endif
276 #ifdef DES_RISC2
277 #define D_ENCRYPT(LL,R,S) {\
278 	unsigned int u1,u2,s1,s2; \
279 	LOAD_DATA(R,S,u,t,E0,E1,u1); \
280 	u>>=2L; \
281 	t=ROTATE(t,6); \
282 	u2=(int)u>>8L; \
283 	u1=(int)u&0x3f; \
284 	u2&=0x3f; \
285 	LL^=DES_SPtrans[0][u1]; \
286 	LL^=DES_SPtrans[2][u2]; \
287 	s1=(int)u>>16L; \
288 	s2=(int)u>>24L; \
289 	s1&=0x3f; \
290 	s2&=0x3f; \
291 	LL^=DES_SPtrans[4][s1]; \
292 	LL^=DES_SPtrans[6][s2]; \
293 	u2=(int)t>>8L; \
294 	u1=(int)t&0x3f; \
295 	u2&=0x3f; \
296 	LL^=DES_SPtrans[1][u1]; \
297 	LL^=DES_SPtrans[3][u2]; \
298 	s1=(int)t>>16; \
299 	s2=(int)t>>24L; \
300 	s1&=0x3f; \
301 	s2&=0x3f; \
302 	LL^=DES_SPtrans[5][s1]; \
303 	LL^=DES_SPtrans[7][s2]; }
304 #endif
305 
306 #else
307 
308 #define D_ENCRYPT(LL,R,S) {\
309 	LOAD_DATA_tmp(R,S,u,t,E0,E1); \
310 	t=ROTATE(t,4); \
311 	LL^=\
312 		DES_SPtrans[0][(u>> 2L)&0x3f]^ \
313 		DES_SPtrans[2][(u>>10L)&0x3f]^ \
314 		DES_SPtrans[4][(u>>18L)&0x3f]^ \
315 		DES_SPtrans[6][(u>>26L)&0x3f]^ \
316 		DES_SPtrans[1][(t>> 2L)&0x3f]^ \
317 		DES_SPtrans[3][(t>>10L)&0x3f]^ \
318 		DES_SPtrans[5][(t>>18L)&0x3f]^ \
319 		DES_SPtrans[7][(t>>26L)&0x3f]; }
320 #endif
321 #endif
322 
323 	/* IP and FP
324 	 * The problem is more of a geometric problem that random bit fiddling.
325 	 0  1  2  3  4  5  6  7      62 54 46 38 30 22 14  6
326 	 8  9 10 11 12 13 14 15      60 52 44 36 28 20 12  4
327 	16 17 18 19 20 21 22 23      58 50 42 34 26 18 10  2
328 	24 25 26 27 28 29 30 31  to  56 48 40 32 24 16  8  0
329 
330 	32 33 34 35 36 37 38 39      63 55 47 39 31 23 15  7
331 	40 41 42 43 44 45 46 47      61 53 45 37 29 21 13  5
332 	48 49 50 51 52 53 54 55      59 51 43 35 27 19 11  3
333 	56 57 58 59 60 61 62 63      57 49 41 33 25 17  9  1
334 
335 	The output has been subject to swaps of the form
336 	0 1 -> 3 1 but the odd and even bits have been put into
337 	2 3    2 0
338 	different words.  The main trick is to remember that
339 	t=((l>>size)^r)&(mask);
340 	r^=t;
341 	l^=(t<<size);
342 	can be used to swap and move bits between words.
343 
344 	So l =  0  1  2  3  r = 16 17 18 19
345 	        4  5  6  7      20 21 22 23
346 	        8  9 10 11      24 25 26 27
347 	       12 13 14 15      28 29 30 31
348 	becomes (for size == 2 and mask == 0x3333)
349 	   t =   2^16  3^17 -- --   l =  0  1 16 17  r =  2  3 18 19
350 		 6^20  7^21 -- --        4  5 20 21       6  7 22 23
351 		10^24 11^25 -- --        8  9 24 25      10 11 24 25
352 		14^28 15^29 -- --       12 13 28 29      14 15 28 29
353 
354 	Thanks for hints from Richard Outerbridge - he told me IP&FP
355 	could be done in 15 xor, 10 shifts and 5 ands.
356 	When I finally started to think of the problem in 2D
357 	I first got ~42 operations without xors.  When I remembered
358 	how to use xors :-) I got it to its final state.
359 	*/
360 #define PERM_OP(a,b,t,n,m) ((t)=((((a)>>(n))^(b))&(m)),\
361 	(b)^=(t),\
362 	(a)^=((t)<<(n)))
363 
364 #define IP(l,r) \
365 	{ \
366 	DES_LONG tt; \
367 	PERM_OP(r,l,tt, 4,0x0f0f0f0fL); \
368 	PERM_OP(l,r,tt,16,0x0000ffffL); \
369 	PERM_OP(r,l,tt, 2,0x33333333L); \
370 	PERM_OP(l,r,tt, 8,0x00ff00ffL); \
371 	PERM_OP(r,l,tt, 1,0x55555555L); \
372 	}
373 
374 #define FP(l,r) \
375 	{ \
376 	DES_LONG tt; \
377 	PERM_OP(l,r,tt, 1,0x55555555L); \
378 	PERM_OP(r,l,tt, 8,0x00ff00ffL); \
379 	PERM_OP(l,r,tt, 2,0x33333333L); \
380 	PERM_OP(r,l,tt,16,0x0000ffffL); \
381 	PERM_OP(l,r,tt, 4,0x0f0f0f0fL); \
382 	}
383 
384 extern const DES_LONG DES_SPtrans[8][64];
385 
386 void fcrypt_body(DES_LONG *out,DES_key_schedule *ks,
387 		 DES_LONG Eswap0, DES_LONG Eswap1);
388 
389 #ifdef OPENSSL_SMALL_FOOTPRINT
390 #undef DES_UNROLL
391 #endif
392 #endif
393