1 | /* $NetBSD: des_locl.h,v 1.4 2001/09/09 11:01:02 tls Exp $ */ |
2 | /* $KAME: des_locl.h,v 1.6 2000/11/06 13:58:09 itojun Exp $ */ |
3 | |
4 | /* crypto/des/des_locl.h */ |
5 | /* Copyright (C) 1995-1997 Eric Young (eay@mincom.oz.au) |
6 | * All rights reserved. |
7 | * |
8 | * This file is part of an SSL implementation written |
9 | * by Eric Young (eay@mincom.oz.au). |
10 | * The implementation was written so as to conform with Netscapes SSL |
11 | * specification. This library and applications are |
12 | * FREE FOR COMMERCIAL AND NON-COMMERCIAL USE |
13 | * as long as the following conditions are aheared to. |
14 | * |
15 | * Copyright remains Eric Young's, and as such any Copyright notices in |
16 | * the code are not to be removed. If this code is used in a product, |
17 | * Eric Young should be given attribution as the author of the parts used. |
18 | * This can be in the form of a textual message at program startup or |
19 | * in documentation (online or textual) provided with the package. |
20 | * |
21 | * Redistribution and use in source and binary forms, with or without |
22 | * modification, are permitted provided that the following conditions |
23 | * are met: |
24 | * 1. Redistributions of source code must retain the copyright |
25 | * notice, this list of conditions and the following disclaimer. |
26 | * 2. Redistributions in binary form must reproduce the above copyright |
27 | * notice, this list of conditions and the following disclaimer in the |
28 | * documentation and/or other materials provided with the distribution. |
29 | * 3. All advertising materials mentioning features or use of this software |
30 | * must display the following acknowledgement: |
31 | * This product includes software developed by Eric Young (eay@mincom.oz.au) |
32 | * |
33 | * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND |
34 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
35 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
36 | * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE |
37 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
38 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
39 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
40 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
41 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
42 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
43 | * SUCH DAMAGE. |
44 | * |
45 | * The licence and distribution terms for any publically available version or |
46 | * derivative of this code cannot be changed. i.e. this code cannot simply be |
47 | * copied and put under another distribution licence |
48 | * [including the GNU Public Licence.] |
49 | */ |
50 | |
51 | #ifndef HEADER_DES_LOCL_H |
52 | #define |
53 | |
54 | #include <crypto/des/des.h> |
55 | |
56 | #undef DES_PTR |
57 | |
58 | #ifdef __STDC__ |
59 | #undef NOPROTO |
60 | #endif |
61 | |
62 | #define ITERATIONS 16 |
63 | #define HALF_ITERATIONS 8 |
64 | |
65 | /* used in des_read and des_write */ |
66 | #define MAXWRITE (1024*16) |
67 | #define BSIZE (MAXWRITE+4) |
68 | |
69 | #define c2l(c,l) (l =((DES_LONG)(*((c)++))) , \ |
70 | l|=((DES_LONG)(*((c)++)))<< 8L, \ |
71 | l|=((DES_LONG)(*((c)++)))<<16L, \ |
72 | l|=((DES_LONG)(*((c)++)))<<24L) |
73 | |
74 | /* NOTE - c is not incremented as per c2l */ |
75 | #define c2ln(c,l1,l2,n) { \ |
76 | c+=n; \ |
77 | l1=l2=0; \ |
78 | switch (n) { \ |
79 | case 8: l2 =((DES_LONG)(*(--(c))))<<24L; \ |
80 | case 7: l2|=((DES_LONG)(*(--(c))))<<16L; \ |
81 | case 6: l2|=((DES_LONG)(*(--(c))))<< 8L; \ |
82 | case 5: l2|=((DES_LONG)(*(--(c)))); \ |
83 | case 4: l1 =((DES_LONG)(*(--(c))))<<24L; \ |
84 | case 3: l1|=((DES_LONG)(*(--(c))))<<16L; \ |
85 | case 2: l1|=((DES_LONG)(*(--(c))))<< 8L; \ |
86 | case 1: l1|=((DES_LONG)(*(--(c)))); \ |
87 | } \ |
88 | } |
89 | |
90 | #define l2c(l,c) (*((c)++)=(unsigned char)(((l) )&0xff), \ |
91 | *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \ |
92 | *((c)++)=(unsigned char)(((l)>>16L)&0xff), \ |
93 | *((c)++)=(unsigned char)(((l)>>24L)&0xff)) |
94 | |
95 | /* replacements for htonl and ntohl since I have no idea what to do |
96 | * when faced with machines with 8 byte longs. */ |
97 | #define HDRSIZE 4 |
98 | |
99 | #define n2l(c,l) (l =((DES_LONG)(*((c)++)))<<24L, \ |
100 | l|=((DES_LONG)(*((c)++)))<<16L, \ |
101 | l|=((DES_LONG)(*((c)++)))<< 8L, \ |
102 | l|=((DES_LONG)(*((c)++)))) |
103 | |
104 | #define l2n(l,c) (*((c)++)=(unsigned char)(((l)>>24L)&0xff), \ |
105 | *((c)++)=(unsigned char)(((l)>>16L)&0xff), \ |
106 | *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \ |
107 | *((c)++)=(unsigned char)(((l) )&0xff)) |
108 | |
109 | /* NOTE - c is not incremented as per l2c */ |
110 | #define l2cn(l1,l2,c,n) { \ |
111 | c+=n; \ |
112 | switch (n) { \ |
113 | case 8: *(--(c))=(unsigned char)(((l2)>>24L)&0xff); \ |
114 | case 7: *(--(c))=(unsigned char)(((l2)>>16L)&0xff); \ |
115 | case 6: *(--(c))=(unsigned char)(((l2)>> 8L)&0xff); \ |
116 | case 5: *(--(c))=(unsigned char)(((l2) )&0xff); \ |
117 | case 4: *(--(c))=(unsigned char)(((l1)>>24L)&0xff); \ |
118 | case 3: *(--(c))=(unsigned char)(((l1)>>16L)&0xff); \ |
119 | case 2: *(--(c))=(unsigned char)(((l1)>> 8L)&0xff); \ |
120 | case 1: *(--(c))=(unsigned char)(((l1) )&0xff); \ |
121 | } \ |
122 | } |
123 | |
124 | #define ROTATE(a,n) (((a)>>(n))+((a)<<(32-(n)))) |
125 | |
126 | #define LOAD_DATA_tmp(a,b,c,d,e,f) LOAD_DATA(a,b,c,d,e,f,g) |
127 | #define LOAD_DATA(R,S,u,t,E0,E1,tmp) \ |
128 | u=R^s[S ]; \ |
129 | t=R^s[S+1] |
130 | |
131 | /* The changes to this macro may help or hinder, depending on the |
132 | * compiler and the achitecture. gcc2 always seems to do well :-). |
133 | * Inspired by Dana How <how@isl.stanford.edu> |
134 | * DO NOT use the alternative version on machines with 8 byte longs. |
135 | * It does not seem to work on the Alpha, even when DES_LONG is 4 |
136 | * bytes, probably an issue of accessing non-word aligned objects :-( */ |
137 | #ifdef DES_PTR |
138 | |
139 | /* It recently occurred to me that 0^0^0^0^0^0^0 == 0, so there |
140 | * is no reason to not xor all the sub items together. This potentially |
141 | * saves a register since things can be xored directly into L */ |
142 | |
143 | #if defined(DES_RISC1) || defined(DES_RISC2) |
144 | #ifdef DES_RISC1 |
145 | #define D_ENCRYPT(LL,R,S) { \ |
146 | unsigned int u1,u2,u3; \ |
147 | LOAD_DATA(R,S,u,t,E0,E1,u1); \ |
148 | u2=(int)u>>8L; \ |
149 | u1=(int)u&0xfc; \ |
150 | u2&=0xfc; \ |
151 | t=ROTATE(t,4); \ |
152 | u>>=16L; \ |
153 | LL^= *(const DES_LONG *)(des_SP +u1); \ |
154 | LL^= *(const DES_LONG *)(des_SP+0x200+u2); \ |
155 | u3=(int)(u>>8L); \ |
156 | u1=(int)u&0xfc; \ |
157 | u3&=0xfc; \ |
158 | LL^= *(const DES_LONG *)(des_SP+0x400+u1); \ |
159 | LL^= *(const DES_LONG *)(des_SP+0x600+u3); \ |
160 | u2=(int)t>>8L; \ |
161 | u1=(int)t&0xfc; \ |
162 | u2&=0xfc; \ |
163 | t>>=16L; \ |
164 | LL^= *(const DES_LONG *)(des_SP+0x100+u1); \ |
165 | LL^= *(const DES_LONG *)(des_SP+0x300+u2); \ |
166 | u3=(int)t>>8L; \ |
167 | u1=(int)t&0xfc; \ |
168 | u3&=0xfc; \ |
169 | LL^= *(const DES_LONG *)(des_SP+0x500+u1); \ |
170 | LL^= *(const DES_LONG *)(des_SP+0x700+u3); } |
171 | #endif /* DES_RISC1 */ |
172 | #ifdef DES_RISC2 |
173 | #define D_ENCRYPT(LL,R,S) { \ |
174 | unsigned int u1,u2,s1,s2; \ |
175 | LOAD_DATA(R,S,u,t,E0,E1,u1); \ |
176 | u2=(int)u>>8L; \ |
177 | u1=(int)u&0xfc; \ |
178 | u2&=0xfc; \ |
179 | t=ROTATE(t,4); \ |
180 | LL^= *(const DES_LONG *)(des_SP +u1); \ |
181 | LL^= *(const DES_LONG *)(des_SP+0x200+u2); \ |
182 | s1=(int)(u>>16L); \ |
183 | s2=(int)(u>>24L); \ |
184 | s1&=0xfc; \ |
185 | s2&=0xfc; \ |
186 | LL^= *(const DES_LONG *)(des_SP+0x400+s1); \ |
187 | LL^= *(const DES_LONG *)(des_SP+0x600+s2); \ |
188 | u2=(int)t>>8L; \ |
189 | u1=(int)t&0xfc; \ |
190 | u2&=0xfc; \ |
191 | LL^= *(const DES_LONG *)(des_SP+0x100+u1); \ |
192 | LL^= *(const DES_LONG *)(des_SP+0x300+u2); \ |
193 | s1=(int)(t>>16L); \ |
194 | s2=(int)(t>>24L); \ |
195 | s1&=0xfc; \ |
196 | s2&=0xfc; \ |
197 | LL^= *(const DES_LONG *)(des_SP+0x400+s1); \ |
198 | LL^= *(const DES_LONG *)(des_SP+0x600+s2); \ |
199 | u2=(int)t>>8L; \ |
200 | u1=(int)t&0xfc; \ |
201 | u2&=0xfc; \ |
202 | LL^= *(const DES_LONG *)(des_SP+0x100+u1); \ |
203 | LL^= *(const DES_LONG *)(des_SP+0x300+u2); \ |
204 | s1=(int)(t>>16L); \ |
205 | s2=(int)(t>>24L); \ |
206 | s1&=0xfc; \ |
207 | s2&=0xfc; \ |
208 | LL^= *(const DES_LONG *)(des_SP+0x500+s1); \ |
209 | LL^= *(const DES_LONG *)(des_SP+0x700+s2); } |
210 | #endif /* DES_RISC2 */ |
211 | #else /* DES_RISC1 || DES_RISC2 */ |
212 | #define D_ENCRYPT(LL,R,S) { \ |
213 | LOAD_DATA_tmp(R,S,u,t,E0,E1); \ |
214 | t=ROTATE(t,4); \ |
215 | LL^= \ |
216 | *(const DES_LONG *)(des_SP +((u )&0xfc))^ \ |
217 | *(const DES_LONG *)(des_SP+0x200+((u>> 8L)&0xfc))^ \ |
218 | *(const DES_LONG *)(des_SP+0x400+((u>>16L)&0xfc))^ \ |
219 | *(const DES_LONG *)(des_SP+0x600+((u>>24L)&0xfc))^ \ |
220 | *(const DES_LONG *)(des_SP+0x100+((t )&0xfc))^ \ |
221 | *(const DES_LONG *)(des_SP+0x300+((t>> 8L)&0xfc))^ \ |
222 | *(const DES_LONG *)(des_SP+0x500+((t>>16L)&0xfc))^ \ |
223 | *(const DES_LONG *)(des_SP+0x700+((t>>24L)&0xfc)); } |
224 | #endif /* DES_RISC1 || DES_RISC2 */ |
225 | #else /* original version */ |
226 | |
227 | #if defined(DES_RISC1) || defined(DES_RISC2) |
228 | #ifdef DES_RISC1 |
229 | #define D_ENCRYPT(LL,R,S) {\ |
230 | unsigned int u1,u2,u3; \ |
231 | LOAD_DATA(R,S,u,t,E0,E1,u1); \ |
232 | u>>=2L; \ |
233 | t=ROTATE(t,6); \ |
234 | u2=(int)u>>8L; \ |
235 | u1=(int)u&0x3f; \ |
236 | u2&=0x3f; \ |
237 | u>>=16L; \ |
238 | LL^=des_SPtrans[0][u1]; \ |
239 | LL^=des_SPtrans[2][u2]; \ |
240 | u3=(int)u>>8L; \ |
241 | u1=(int)u&0x3f; \ |
242 | u3&=0x3f; \ |
243 | LL^=des_SPtrans[4][u1]; \ |
244 | LL^=des_SPtrans[6][u3]; \ |
245 | u2=(int)t>>8L; \ |
246 | u1=(int)t&0x3f; \ |
247 | u2&=0x3f; \ |
248 | t>>=16L; \ |
249 | LL^=des_SPtrans[1][u1]; \ |
250 | LL^=des_SPtrans[3][u2]; \ |
251 | u3=(int)t>>8L; \ |
252 | u1=(int)t&0x3f; \ |
253 | u3&=0x3f; \ |
254 | LL^=des_SPtrans[5][u1]; \ |
255 | LL^=des_SPtrans[7][u3]; } |
256 | #endif /* DES_RISC1 */ |
257 | #ifdef DES_RISC2 |
258 | #define D_ENCRYPT(LL,R,S) {\ |
259 | unsigned int u1,u2,s1,s2; \ |
260 | LOAD_DATA(R,S,u,t,E0,E1,u1); \ |
261 | u>>=2L; \ |
262 | t=ROTATE(t,6); \ |
263 | u2=(int)u>>8L; \ |
264 | u1=(int)u&0x3f; \ |
265 | u2&=0x3f; \ |
266 | LL^=des_SPtrans[0][u1]; \ |
267 | LL^=des_SPtrans[2][u2]; \ |
268 | s1=(int)u>>16L; \ |
269 | s2=(int)u>>24L; \ |
270 | s1&=0x3f; \ |
271 | s2&=0x3f; \ |
272 | LL^=des_SPtrans[4][s1]; \ |
273 | LL^=des_SPtrans[6][s2]; \ |
274 | u2=(int)t>>8L; \ |
275 | u1=(int)t&0x3f; \ |
276 | u2&=0x3f; \ |
277 | LL^=des_SPtrans[1][u1]; \ |
278 | LL^=des_SPtrans[3][u2]; \ |
279 | s1=(int)t>>16; \ |
280 | s2=(int)t>>24L; \ |
281 | s1&=0x3f; \ |
282 | s2&=0x3f; \ |
283 | LL^=des_SPtrans[5][s1]; \ |
284 | LL^=des_SPtrans[7][s2]; } |
285 | #endif /* DES_RISC2 */ |
286 | |
287 | #else /* DES_RISC1 || DES_RISC2 */ |
288 | |
289 | #define D_ENCRYPT(LL,R,S) {\ |
290 | LOAD_DATA_tmp(R,S,u,t,E0,E1); \ |
291 | t=ROTATE(t,4); \ |
292 | LL^=\ |
293 | des_SPtrans[0][(u>> 2L)&0x3f]^ \ |
294 | des_SPtrans[2][(u>>10L)&0x3f]^ \ |
295 | des_SPtrans[4][(u>>18L)&0x3f]^ \ |
296 | des_SPtrans[6][(u>>26L)&0x3f]^ \ |
297 | des_SPtrans[1][(t>> 2L)&0x3f]^ \ |
298 | des_SPtrans[3][(t>>10L)&0x3f]^ \ |
299 | des_SPtrans[5][(t>>18L)&0x3f]^ \ |
300 | des_SPtrans[7][(t>>26L)&0x3f]; } |
301 | #endif /* DES_RISC1 || DES_RISC2 */ |
302 | #endif /* DES_PTR */ |
303 | |
304 | /* IP and FP |
305 | * The problem is more of a geometric problem that random bit fiddling. |
306 | 0 1 2 3 4 5 6 7 62 54 46 38 30 22 14 6 |
307 | 8 9 10 11 12 13 14 15 60 52 44 36 28 20 12 4 |
308 | 16 17 18 19 20 21 22 23 58 50 42 34 26 18 10 2 |
309 | 24 25 26 27 28 29 30 31 to 56 48 40 32 24 16 8 0 |
310 | |
311 | 32 33 34 35 36 37 38 39 63 55 47 39 31 23 15 7 |
312 | 40 41 42 43 44 45 46 47 61 53 45 37 29 21 13 5 |
313 | 48 49 50 51 52 53 54 55 59 51 43 35 27 19 11 3 |
314 | 56 57 58 59 60 61 62 63 57 49 41 33 25 17 9 1 |
315 | |
316 | The output has been subject to swaps of the form |
317 | 0 1 -> 3 1 but the odd and even bits have been put into |
318 | 2 3 2 0 |
319 | different words. The main trick is to remember that |
320 | t=((l>>size)^r)&(mask); |
321 | r^=t; |
322 | l^=(t<<size); |
323 | can be used to swap and move bits between words. |
324 | |
325 | So l = 0 1 2 3 r = 16 17 18 19 |
326 | 4 5 6 7 20 21 22 23 |
327 | 8 9 10 11 24 25 26 27 |
328 | 12 13 14 15 28 29 30 31 |
329 | becomes (for size == 2 and mask == 0x3333) |
330 | t = 2^16 3^17 -- -- l = 0 1 16 17 r = 2 3 18 19 |
331 | 6^20 7^21 -- -- 4 5 20 21 6 7 22 23 |
332 | 10^24 11^25 -- -- 8 9 24 25 10 11 24 25 |
333 | 14^28 15^29 -- -- 12 13 28 29 14 15 28 29 |
334 | |
335 | Thanks for hints from Richard Outerbridge - he told me IP&FP |
336 | could be done in 15 xor, 10 shifts and 5 ands. |
337 | When I finally started to think of the problem in 2D |
338 | I first got ~42 operations without xors. When I remembered |
339 | how to use xors :-) I got it to its final state. |
340 | */ |
341 | #define PERM_OP(a,b,t,n,m) ((t)=((((a)>>(n))^(b))&(m)),\ |
342 | (b)^=(t),\ |
343 | (a)^=((t)<<(n))) |
344 | |
345 | #define IP(l,r) \ |
346 | { \ |
347 | register DES_LONG tt; \ |
348 | PERM_OP(r,l,tt, 4,0x0f0f0f0fL); \ |
349 | PERM_OP(l,r,tt,16,0x0000ffffL); \ |
350 | PERM_OP(r,l,tt, 2,0x33333333L); \ |
351 | PERM_OP(l,r,tt, 8,0x00ff00ffL); \ |
352 | PERM_OP(r,l,tt, 1,0x55555555L); \ |
353 | } |
354 | |
355 | #define FP(l,r) \ |
356 | { \ |
357 | register DES_LONG tt; \ |
358 | PERM_OP(l,r,tt, 1,0x55555555L); \ |
359 | PERM_OP(r,l,tt, 8,0x00ff00ffL); \ |
360 | PERM_OP(l,r,tt, 2,0x33333333L); \ |
361 | PERM_OP(r,l,tt,16,0x0000ffffL); \ |
362 | PERM_OP(l,r,tt, 4,0x0f0f0f0fL); \ |
363 | } |
364 | #endif |
365 | |