]> git.ipfire.org Git - thirdparty/openssl.git/blob - crypto/des/des_locl.h
230a30f1e579ae3ceefc7e9b8086e410b2f49a19
[thirdparty/openssl.git] / crypto / des / des_locl.h
1 /*
2 * Copyright 1995-2016 The OpenSSL Project Authors. All Rights Reserved.
3 *
4 * Licensed under the Apache License 2.0 (the "License"). You may not use
5 * this file except in compliance with the License. You can obtain a copy
6 * in the file LICENSE in the source distribution or at
7 * https://www.openssl.org/source/license.html
8 */
9
10 #ifndef HEADER_DES_LOCL_H
11 # define HEADER_DES_LOCL_H
12
13 # include <openssl/e_os2.h>
14
15 # include <stdio.h>
16 # include <stdlib.h>
17 # include <string.h>
18
19 # include <openssl/des.h>
20
21 # ifdef OPENSSL_BUILD_SHLIBCRYPTO
22 # undef OPENSSL_EXTERN
23 # define OPENSSL_EXTERN OPENSSL_EXPORT
24 # endif
25
26 # define ITERATIONS 16
27 # define HALF_ITERATIONS 8
28
29 # define c2l(c,l) (l =((DES_LONG)(*((c)++))) , \
30 l|=((DES_LONG)(*((c)++)))<< 8L, \
31 l|=((DES_LONG)(*((c)++)))<<16L, \
32 l|=((DES_LONG)(*((c)++)))<<24L)
33
34 /* NOTE - c is not incremented as per c2l */
35 # define c2ln(c,l1,l2,n) { \
36 c+=n; \
37 l1=l2=0; \
38 switch (n) { \
39 case 8: l2 =((DES_LONG)(*(--(c))))<<24L; \
40 /* fall thru */ \
41 case 7: l2|=((DES_LONG)(*(--(c))))<<16L; \
42 /* fall thru */ \
43 case 6: l2|=((DES_LONG)(*(--(c))))<< 8L; \
44 /* fall thru */ \
45 case 5: l2|=((DES_LONG)(*(--(c)))); \
46 /* fall thru */ \
47 case 4: l1 =((DES_LONG)(*(--(c))))<<24L; \
48 /* fall thru */ \
49 case 3: l1|=((DES_LONG)(*(--(c))))<<16L; \
50 /* fall thru */ \
51 case 2: l1|=((DES_LONG)(*(--(c))))<< 8L; \
52 /* fall thru */ \
53 case 1: l1|=((DES_LONG)(*(--(c)))); \
54 } \
55 }
56
57 # define l2c(l,c) (*((c)++)=(unsigned char)(((l) )&0xff), \
58 *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \
59 *((c)++)=(unsigned char)(((l)>>16L)&0xff), \
60 *((c)++)=(unsigned char)(((l)>>24L)&0xff))
61
62 /*
63 * replacements for htonl and ntohl since I have no idea what to do when
64 * faced with machines with 8 byte longs.
65 */
66
67 # define n2l(c,l) (l =((DES_LONG)(*((c)++)))<<24L, \
68 l|=((DES_LONG)(*((c)++)))<<16L, \
69 l|=((DES_LONG)(*((c)++)))<< 8L, \
70 l|=((DES_LONG)(*((c)++))))
71
72 # define l2n(l,c) (*((c)++)=(unsigned char)(((l)>>24L)&0xff), \
73 *((c)++)=(unsigned char)(((l)>>16L)&0xff), \
74 *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \
75 *((c)++)=(unsigned char)(((l) )&0xff))
76
77 /* NOTE - c is not incremented as per l2c */
78 # define l2cn(l1,l2,c,n) { \
79 c+=n; \
80 switch (n) { \
81 case 8: *(--(c))=(unsigned char)(((l2)>>24L)&0xff); \
82 /* fall thru */ \
83 case 7: *(--(c))=(unsigned char)(((l2)>>16L)&0xff); \
84 /* fall thru */ \
85 case 6: *(--(c))=(unsigned char)(((l2)>> 8L)&0xff); \
86 /* fall thru */ \
87 case 5: *(--(c))=(unsigned char)(((l2) )&0xff); \
88 /* fall thru */ \
89 case 4: *(--(c))=(unsigned char)(((l1)>>24L)&0xff); \
90 /* fall thru */ \
91 case 3: *(--(c))=(unsigned char)(((l1)>>16L)&0xff); \
92 /* fall thru */ \
93 case 2: *(--(c))=(unsigned char)(((l1)>> 8L)&0xff); \
94 /* fall thru */ \
95 case 1: *(--(c))=(unsigned char)(((l1) )&0xff); \
96 } \
97 }
98
99 # if defined(_MSC_VER)
100 # define ROTATE(a,n) (_lrotr(a,n))
101 # elif defined(__ICC)
102 # define ROTATE(a,n) (_rotr(a,n))
103 # elif defined(__GNUC__) && __GNUC__>=2 && !defined(__STRICT_ANSI__) && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM) && !defined(PEDANTIC)
104 # if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__)
105 # define ROTATE(a,n) ({ register unsigned int ret; \
106 asm ("rorl %1,%0" \
107 : "=r"(ret) \
108 : "I"(n),"0"(a) \
109 : "cc"); \
110 ret; \
111 })
112 # endif
113 # endif
114 # ifndef ROTATE
115 # define ROTATE(a,n) (((a)>>(n))+((a)<<(32-(n))))
116 # endif
117
118 /*
119 * Don't worry about the LOAD_DATA() stuff, that is used by fcrypt() to add
120 * it's little bit to the front
121 */
122
123 # ifdef DES_FCRYPT
124
125 # define LOAD_DATA_tmp(R,S,u,t,E0,E1) \
126 { DES_LONG tmp; LOAD_DATA(R,S,u,t,E0,E1,tmp); }
127
128 # define LOAD_DATA(R,S,u,t,E0,E1,tmp) \
129 t=R^(R>>16L); \
130 u=t&E0; t&=E1; \
131 tmp=(u<<16); u^=R^s[S ]; u^=tmp; \
132 tmp=(t<<16); t^=R^s[S+1]; t^=tmp
133 # else
134 # define LOAD_DATA_tmp(a,b,c,d,e,f) LOAD_DATA(a,b,c,d,e,f,g)
135 # define LOAD_DATA(R,S,u,t,E0,E1,tmp) \
136 u=R^s[S ]; \
137 t=R^s[S+1]
138 # endif
139
140 /*
141 * It recently occurred to me that 0^0^0^0^0^0^0 == 0, so there is no reason
142 * to not xor all the sub items together. This potentially saves a register
143 * since things can be xored directly into L
144 */
145
146 # define D_ENCRYPT(LL,R,S) { \
147 LOAD_DATA_tmp(R,S,u,t,E0,E1); \
148 t=ROTATE(t,4); \
149 LL^= \
150 DES_SPtrans[0][(u>> 2L)&0x3f]^ \
151 DES_SPtrans[2][(u>>10L)&0x3f]^ \
152 DES_SPtrans[4][(u>>18L)&0x3f]^ \
153 DES_SPtrans[6][(u>>26L)&0x3f]^ \
154 DES_SPtrans[1][(t>> 2L)&0x3f]^ \
155 DES_SPtrans[3][(t>>10L)&0x3f]^ \
156 DES_SPtrans[5][(t>>18L)&0x3f]^ \
157 DES_SPtrans[7][(t>>26L)&0x3f]; }
158
159 /*-
160 * IP and FP
161 * The problem is more of a geometric problem that random bit fiddling.
162 0 1 2 3 4 5 6 7 62 54 46 38 30 22 14 6
163 8 9 10 11 12 13 14 15 60 52 44 36 28 20 12 4
164 16 17 18 19 20 21 22 23 58 50 42 34 26 18 10 2
165 24 25 26 27 28 29 30 31 to 56 48 40 32 24 16 8 0
166
167 32 33 34 35 36 37 38 39 63 55 47 39 31 23 15 7
168 40 41 42 43 44 45 46 47 61 53 45 37 29 21 13 5
169 48 49 50 51 52 53 54 55 59 51 43 35 27 19 11 3
170 56 57 58 59 60 61 62 63 57 49 41 33 25 17 9 1
171
172 The output has been subject to swaps of the form
173 0 1 -> 3 1 but the odd and even bits have been put into
174 2 3 2 0
175 different words. The main trick is to remember that
176 t=((l>>size)^r)&(mask);
177 r^=t;
178 l^=(t<<size);
179 can be used to swap and move bits between words.
180
181 So l = 0 1 2 3 r = 16 17 18 19
182 4 5 6 7 20 21 22 23
183 8 9 10 11 24 25 26 27
184 12 13 14 15 28 29 30 31
185 becomes (for size == 2 and mask == 0x3333)
186 t = 2^16 3^17 -- -- l = 0 1 16 17 r = 2 3 18 19
187 6^20 7^21 -- -- 4 5 20 21 6 7 22 23
188 10^24 11^25 -- -- 8 9 24 25 10 11 24 25
189 14^28 15^29 -- -- 12 13 28 29 14 15 28 29
190
191 Thanks for hints from Richard Outerbridge - he told me IP&FP
192 could be done in 15 xor, 10 shifts and 5 ands.
193 When I finally started to think of the problem in 2D
194 I first got ~42 operations without xors. When I remembered
195 how to use xors :-) I got it to its final state.
196 */
197 # define PERM_OP(a,b,t,n,m) ((t)=((((a)>>(n))^(b))&(m)),\
198 (b)^=(t),\
199 (a)^=((t)<<(n)))
200
201 # define IP(l,r) \
202 { \
203 register DES_LONG tt; \
204 PERM_OP(r,l,tt, 4,0x0f0f0f0fL); \
205 PERM_OP(l,r,tt,16,0x0000ffffL); \
206 PERM_OP(r,l,tt, 2,0x33333333L); \
207 PERM_OP(l,r,tt, 8,0x00ff00ffL); \
208 PERM_OP(r,l,tt, 1,0x55555555L); \
209 }
210
211 # define FP(l,r) \
212 { \
213 register DES_LONG tt; \
214 PERM_OP(l,r,tt, 1,0x55555555L); \
215 PERM_OP(r,l,tt, 8,0x00ff00ffL); \
216 PERM_OP(l,r,tt, 2,0x33333333L); \
217 PERM_OP(r,l,tt,16,0x0000ffffL); \
218 PERM_OP(l,r,tt, 4,0x0f0f0f0fL); \
219 }
220
221 extern const DES_LONG DES_SPtrans[8][64];
222
223 void fcrypt_body(DES_LONG *out, DES_key_schedule *ks,
224 DES_LONG Eswap0, DES_LONG Eswap1);
225
226 #endif