1/* crypto/bf/bf_locl.h */
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_BF_LOCL_H
60#define HEADER_BF_LOCL_H
61#include <openssl/opensslconf.h> /* BF_PTR, BF_PTR2 */
62
63#undef c2l
64#define c2l(c,l)	(l =((unsigned long)(*((c)++)))    , \
65			 l|=((unsigned long)(*((c)++)))<< 8L, \
66			 l|=((unsigned long)(*((c)++)))<<16L, \
67			 l|=((unsigned long)(*((c)++)))<<24L)
68
69/* NOTE - c is not incremented as per c2l */
70#undef c2ln
71#define c2ln(c,l1,l2,n)	{ \
72			c+=n; \
73			l1=l2=0; \
74			switch (n) { \
75			case 8: l2 =((unsigned long)(*(--(c))))<<24L; \
76			case 7: l2|=((unsigned long)(*(--(c))))<<16L; \
77			case 6: l2|=((unsigned long)(*(--(c))))<< 8L; \
78			case 5: l2|=((unsigned long)(*(--(c))));     \
79			case 4: l1 =((unsigned long)(*(--(c))))<<24L; \
80			case 3: l1|=((unsigned long)(*(--(c))))<<16L; \
81			case 2: l1|=((unsigned long)(*(--(c))))<< 8L; \
82			case 1: l1|=((unsigned long)(*(--(c))));     \
83				} \
84			}
85
86#undef l2c
87#define l2c(l,c)	(*((c)++)=(unsigned char)(((l)     )&0xff), \
88			 *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \
89			 *((c)++)=(unsigned char)(((l)>>16L)&0xff), \
90			 *((c)++)=(unsigned char)(((l)>>24L)&0xff))
91
92/* NOTE - c is not incremented as per l2c */
93#undef l2cn
94#define l2cn(l1,l2,c,n)	{ \
95			c+=n; \
96			switch (n) { \
97			case 8: *(--(c))=(unsigned char)(((l2)>>24L)&0xff); \
98			case 7: *(--(c))=(unsigned char)(((l2)>>16L)&0xff); \
99			case 6: *(--(c))=(unsigned char)(((l2)>> 8L)&0xff); \
100			case 5: *(--(c))=(unsigned char)(((l2)     )&0xff); \
101			case 4: *(--(c))=(unsigned char)(((l1)>>24L)&0xff); \
102			case 3: *(--(c))=(unsigned char)(((l1)>>16L)&0xff); \
103			case 2: *(--(c))=(unsigned char)(((l1)>> 8L)&0xff); \
104			case 1: *(--(c))=(unsigned char)(((l1)     )&0xff); \
105				} \
106			}
107
108/* NOTE - c is not incremented as per n2l */
109#define n2ln(c,l1,l2,n)	{ \
110			c+=n; \
111			l1=l2=0; \
112			switch (n) { \
113			case 8: l2 =((unsigned long)(*(--(c))))    ; \
114			case 7: l2|=((unsigned long)(*(--(c))))<< 8; \
115			case 6: l2|=((unsigned long)(*(--(c))))<<16; \
116			case 5: l2|=((unsigned long)(*(--(c))))<<24; \
117			case 4: l1 =((unsigned long)(*(--(c))))    ; \
118			case 3: l1|=((unsigned long)(*(--(c))))<< 8; \
119			case 2: l1|=((unsigned long)(*(--(c))))<<16; \
120			case 1: l1|=((unsigned long)(*(--(c))))<<24; \
121				} \
122			}
123
124/* NOTE - c is not incremented as per l2n */
125#define l2nn(l1,l2,c,n)	{ \
126			c+=n; \
127			switch (n) { \
128			case 8: *(--(c))=(unsigned char)(((l2)    )&0xff); \
129			case 7: *(--(c))=(unsigned char)(((l2)>> 8)&0xff); \
130			case 6: *(--(c))=(unsigned char)(((l2)>>16)&0xff); \
131			case 5: *(--(c))=(unsigned char)(((l2)>>24)&0xff); \
132			case 4: *(--(c))=(unsigned char)(((l1)    )&0xff); \
133			case 3: *(--(c))=(unsigned char)(((l1)>> 8)&0xff); \
134			case 2: *(--(c))=(unsigned char)(((l1)>>16)&0xff); \
135			case 1: *(--(c))=(unsigned char)(((l1)>>24)&0xff); \
136				} \
137			}
138
139#undef n2l
140#define n2l(c,l)        (l =((unsigned long)(*((c)++)))<<24L, \
141                         l|=((unsigned long)(*((c)++)))<<16L, \
142                         l|=((unsigned long)(*((c)++)))<< 8L, \
143                         l|=((unsigned long)(*((c)++))))
144
145#undef l2n
146#define l2n(l,c)        (*((c)++)=(unsigned char)(((l)>>24L)&0xff), \
147                         *((c)++)=(unsigned char)(((l)>>16L)&0xff), \
148                         *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \
149                         *((c)++)=(unsigned char)(((l)     )&0xff))
150
151/* This is actually a big endian algorithm, the most significant byte
152 * is used to lookup array 0 */
153
154#if defined(BF_PTR2)
155
156/*
157 * This is basically a special Intel version. Point is that Intel
158 * doesn't have many registers, but offers a reach choice of addressing
159 * modes. So we spare some registers by directly traversing BF_KEY
160 * structure and hiring the most decorated addressing mode. The code
161 * generated by EGCS is *perfectly* competitive with assembler
162 * implementation!
163 */
164#define BF_ENC(LL,R,KEY,Pi) (\
165	LL^=KEY[Pi], \
166	t=  KEY[BF_ROUNDS+2 +   0 + ((R>>24)&0xFF)], \
167	t+= KEY[BF_ROUNDS+2 + 256 + ((R>>16)&0xFF)], \
168	t^= KEY[BF_ROUNDS+2 + 512 + ((R>>8 )&0xFF)], \
169	t+= KEY[BF_ROUNDS+2 + 768 + ((R    )&0xFF)], \
170	LL^=t \
171	)
172
173#elif defined(BF_PTR)
174
175#ifndef BF_LONG_LOG2
176#define BF_LONG_LOG2  2       /* default to BF_LONG being 32 bits */
177#endif
178#define BF_M  (0xFF<<BF_LONG_LOG2)
179#define BF_0  (24-BF_LONG_LOG2)
180#define BF_1  (16-BF_LONG_LOG2)
181#define BF_2  ( 8-BF_LONG_LOG2)
182#define BF_3  BF_LONG_LOG2 /* left shift */
183
184/*
185 * This is normally very good on RISC platforms where normally you
186 * have to explicitly "multiply" array index by sizeof(BF_LONG)
187 * in order to calculate the effective address. This implementation
188 * excuses CPU from this extra work. Power[PC] uses should have most
189 * fun as (R>>BF_i)&BF_M gets folded into a single instruction, namely
190 * rlwinm. So let'em double-check if their compiler does it.
191 */
192
193#define BF_ENC(LL,R,S,P) ( \
194	LL^=P, \
195	LL^= (((*(BF_LONG *)((unsigned char *)&(S[  0])+((R>>BF_0)&BF_M))+ \
196		*(BF_LONG *)((unsigned char *)&(S[256])+((R>>BF_1)&BF_M)))^ \
197		*(BF_LONG *)((unsigned char *)&(S[512])+((R>>BF_2)&BF_M)))+ \
198		*(BF_LONG *)((unsigned char *)&(S[768])+((R<<BF_3)&BF_M))) \
199	)
200#else
201
202/*
203 * This is a *generic* version. Seem to perform best on platforms that
204 * offer explicit support for extraction of 8-bit nibbles preferably
205 * complemented with "multiplying" of array index by sizeof(BF_LONG).
206 * For the moment of this writing the list comprises Alpha CPU featuring
207 * extbl and s[48]addq instructions.
208 */
209
210#define BF_ENC(LL,R,S,P) ( \
211	LL^=P, \
212	LL^=(((	S[       ((int)(R>>24)&0xff)] + \
213		S[0x0100+((int)(R>>16)&0xff)])^ \
214		S[0x0200+((int)(R>> 8)&0xff)])+ \
215		S[0x0300+((int)(R    )&0xff)])&0xffffffffL \
216	)
217#endif
218
219#endif
220