1 /* $OpenBSD: rc4_enc.c,v 1.17 2021/11/09 18:40:21 bcook Exp $ */
2 /* Copyright (C) 1995-1998 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 #include <endian.h>
60
61 #include <openssl/rc4.h>
62 #include "rc4_locl.h"
63
64 /* RC4 as implemented from a posting from
65 * Newsgroups: sci.crypt
66 * From: sterndark@netcom.com (David Sterndark)
67 * Subject: RC4 Algorithm revealed.
68 * Message-ID: <sternCvKL4B.Hyy@netcom.com>
69 * Date: Wed, 14 Sep 1994 06:35:31 GMT
70 */
71
72 void
RC4(RC4_KEY * key,size_t len,const unsigned char * indata,unsigned char * outdata)73 RC4(RC4_KEY *key, size_t len, const unsigned char *indata,
74 unsigned char *outdata)
75 {
76 RC4_INT *d;
77 RC4_INT x, y,tx, ty;
78 size_t i;
79
80 x = key->x;
81 y = key->y;
82 d = key->data;
83
84 #if defined(RC4_CHUNK)
85 /*
86 * The original reason for implementing this(*) was the fact that
87 * pre-21164a Alpha CPUs don't have byte load/store instructions
88 * and e.g. a byte store has to be done with 64-bit load, shift,
89 * and, or and finally 64-bit store. Peaking data and operating
90 * at natural word size made it possible to reduce amount of
91 * instructions as well as to perform early read-ahead without
92 * suffering from RAW (read-after-write) hazard. This resulted
93 * in ~40%(**) performance improvement on 21064 box with gcc.
94 * But it's not only Alpha users who win here:-) Thanks to the
95 * early-n-wide read-ahead this implementation also exhibits
96 * >40% speed-up on SPARC and 20-30% on 64-bit MIPS (depending
97 * on sizeof(RC4_INT)).
98 *
99 * (*) "this" means code which recognizes the case when input
100 * and output pointers appear to be aligned at natural CPU
101 * word boundary
102 * (**) i.e. according to 'apps/openssl speed rc4' benchmark,
103 * crypto/rc4/rc4speed.c exhibits almost 70% speed-up...
104 *
105 * Caveats.
106 *
107 * - RC4_CHUNK="unsigned long long" should be a #1 choice for
108 * UltraSPARC. Unfortunately gcc generates very slow code
109 * (2.5-3 times slower than one generated by Sun's WorkShop
110 * C) and therefore gcc (at least 2.95 and earlier) should
111 * always be told that RC4_CHUNK="unsigned long".
112 *
113 * <appro@fy.chalmers.se>
114 */
115
116 # define RC4_STEP ( \
117 x=(x+1) &0xff, \
118 tx=d[x], \
119 y=(tx+y)&0xff, \
120 ty=d[y], \
121 d[y]=tx, \
122 d[x]=ty, \
123 (RC4_CHUNK)d[(tx+ty)&0xff]\
124 )
125
126 if ((((size_t)indata & (sizeof(RC4_CHUNK) - 1)) |
127 ((size_t)outdata & (sizeof(RC4_CHUNK) - 1))) == 0 ) {
128 RC4_CHUNK ichunk, otp;
129
130 /*
131 * I reckon we can afford to implement both endian
132 * cases and to decide which way to take at run-time
133 * because the machine code appears to be very compact
134 * and redundant 1-2KB is perfectly tolerable (i.e.
135 * in case the compiler fails to eliminate it:-). By
136 * suggestion from Terrel Larson <terr@terralogic.net>.
137 *
138 * Special notes.
139 *
140 * - compilers (those I've tried) don't seem to have
141 * problems eliminating either the operators guarded
142 * by "if (sizeof(RC4_CHUNK)==8)" or the condition
143 * expressions themselves so I've got 'em to replace
144 * corresponding #ifdefs from the previous version;
145 * - I chose to let the redundant switch cases when
146 * sizeof(RC4_CHUNK)!=8 be (were also #ifdefed
147 * before);
148 * - in case you wonder "&(sizeof(RC4_CHUNK)*8-1)" in
149 * [LB]ESHFT guards against "shift is out of range"
150 * warnings when sizeof(RC4_CHUNK)!=8
151 *
152 * <appro@fy.chalmers.se>
153 */
154 #if BYTE_ORDER == BIG_ENDIAN
155 # define BESHFT(c) (((sizeof(RC4_CHUNK)-(c)-1)*8)&(sizeof(RC4_CHUNK)*8-1))
156 for (; len & (0 - sizeof(RC4_CHUNK)); len -= sizeof(RC4_CHUNK)) {
157 ichunk = *(RC4_CHUNK *)indata;
158 otp = RC4_STEP << BESHFT(0);
159 otp |= RC4_STEP << BESHFT(1);
160 otp |= RC4_STEP << BESHFT(2);
161 otp |= RC4_STEP << BESHFT(3);
162 if (sizeof(RC4_CHUNK) == 8) {
163 otp |= RC4_STEP << BESHFT(4);
164 otp |= RC4_STEP << BESHFT(5);
165 otp |= RC4_STEP << BESHFT(6);
166 otp |= RC4_STEP << BESHFT(7);
167 }
168 *(RC4_CHUNK *)outdata = otp^ichunk;
169 indata += sizeof(RC4_CHUNK);
170 outdata += sizeof(RC4_CHUNK);
171 }
172 #else
173 # define LESHFT(c) (((c)*8)&(sizeof(RC4_CHUNK)*8-1))
174 for (; len & (0 - sizeof(RC4_CHUNK)); len -= sizeof(RC4_CHUNK)) {
175 ichunk = *(RC4_CHUNK *)indata;
176 otp = RC4_STEP;
177 otp |= RC4_STEP << 8;
178 otp |= RC4_STEP << 16;
179 otp |= RC4_STEP << 24;
180 if (sizeof(RC4_CHUNK) == 8) {
181 otp |= RC4_STEP << LESHFT(4);
182 otp |= RC4_STEP << LESHFT(5);
183 otp |= RC4_STEP << LESHFT(6);
184 otp |= RC4_STEP << LESHFT(7);
185 }
186 *(RC4_CHUNK *)outdata = otp ^ ichunk;
187 indata += sizeof(RC4_CHUNK);
188 outdata += sizeof(RC4_CHUNK);
189 }
190 #endif
191 }
192 #endif
193 #define LOOP(in,out) \
194 x=((x+1)&0xff); \
195 tx=d[x]; \
196 y=(tx+y)&0xff; \
197 d[x]=ty=d[y]; \
198 d[y]=tx; \
199 (out) = d[(tx+ty)&0xff]^ (in);
200
201 #ifndef RC4_INDEX
202 #define RC4_LOOP(a,b,i) LOOP(*((a)++),*((b)++))
203 #else
204 #define RC4_LOOP(a,b,i) LOOP(a[i],b[i])
205 #endif
206
207 i = len >> 3;
208 if (i) {
209 for (;;) {
210 RC4_LOOP(indata, outdata, 0);
211 RC4_LOOP(indata, outdata, 1);
212 RC4_LOOP(indata, outdata, 2);
213 RC4_LOOP(indata, outdata, 3);
214 RC4_LOOP(indata, outdata, 4);
215 RC4_LOOP(indata, outdata, 5);
216 RC4_LOOP(indata, outdata, 6);
217 RC4_LOOP(indata, outdata, 7);
218 #ifdef RC4_INDEX
219 indata += 8;
220 outdata += 8;
221 #endif
222 if (--i == 0)
223 break;
224 }
225 }
226 i = len&0x07;
227 if (i) {
228 for (;;) {
229 RC4_LOOP(indata, outdata, 0);
230 if (--i == 0)
231 break;
232 RC4_LOOP(indata, outdata, 1);
233 if (--i == 0)
234 break;
235 RC4_LOOP(indata, outdata, 2);
236 if (--i == 0)
237 break;
238 RC4_LOOP(indata, outdata, 3);
239 if (--i == 0)
240 break;
241 RC4_LOOP(indata, outdata, 4);
242 if (--i == 0)
243 break;
244 RC4_LOOP(indata, outdata, 5);
245 if (--i == 0)
246 break;
247 RC4_LOOP(indata, outdata, 6);
248 if (--i == 0)
249 break;
250 }
251 }
252 key->x = x;
253 key->y = y;
254 }
255