1 /* $OpenBSD: rc4_enc.c,v 1.14 2015/10/20 15:50:13 jsing 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 <machine/endian.h> 60 #include <openssl/rc4.h> 61 #include "rc4_locl.h" 62 63 /* RC4 as implemented from a posting from 64 * Newsgroups: sci.crypt 65 * From: sterndark@netcom.com (David Sterndark) 66 * Subject: RC4 Algorithm revealed. 67 * Message-ID: <sternCvKL4B.Hyy@netcom.com> 68 * Date: Wed, 14 Sep 1994 06:35:31 GMT 69 */ 70 71 void 72 RC4(RC4_KEY *key, size_t len, const unsigned char *indata, 73 unsigned char *outdata) 74 { 75 RC4_INT *d; 76 RC4_INT x, y,tx, ty; 77 size_t i; 78 79 x = key->x; 80 y = key->y; 81 d = key->data; 82 83 #if defined(RC4_CHUNK) 84 /* 85 * The original reason for implementing this(*) was the fact that 86 * pre-21164a Alpha CPUs don't have byte load/store instructions 87 * and e.g. a byte store has to be done with 64-bit load, shift, 88 * and, or and finally 64-bit store. Peaking data and operating 89 * at natural word size made it possible to reduce amount of 90 * instructions as well as to perform early read-ahead without 91 * suffering from RAW (read-after-write) hazard. This resulted 92 * in ~40%(**) performance improvement on 21064 box with gcc. 93 * But it's not only Alpha users who win here:-) Thanks to the 94 * early-n-wide read-ahead this implementation also exhibits 95 * >40% speed-up on SPARC and 20-30% on 64-bit MIPS (depending 96 * on sizeof(RC4_INT)). 97 * 98 * (*) "this" means code which recognizes the case when input 99 * and output pointers appear to be aligned at natural CPU 100 * word boundary 101 * (**) i.e. according to 'apps/openssl speed rc4' benchmark, 102 * crypto/rc4/rc4speed.c exhibits almost 70% speed-up... 103 * 104 * Caveats. 105 * 106 * - RC4_CHUNK="unsigned long long" should be a #1 choice for 107 * UltraSPARC. Unfortunately gcc generates very slow code 108 * (2.5-3 times slower than one generated by Sun's WorkShop 109 * C) and therefore gcc (at least 2.95 and earlier) should 110 * always be told that RC4_CHUNK="unsigned long". 111 * 112 * <appro@fy.chalmers.se> 113 */ 114 115 # define RC4_STEP ( \ 116 x=(x+1) &0xff, \ 117 tx=d[x], \ 118 y=(tx+y)&0xff, \ 119 ty=d[y], \ 120 d[y]=tx, \ 121 d[x]=ty, \ 122 (RC4_CHUNK)d[(tx+ty)&0xff]\ 123 ) 124 125 if ((((size_t)indata & (sizeof(RC4_CHUNK) - 1)) | 126 ((size_t)outdata & (sizeof(RC4_CHUNK) - 1))) == 0 ) { 127 RC4_CHUNK ichunk, otp; 128 129 /* 130 * I reckon we can afford to implement both endian 131 * cases and to decide which way to take at run-time 132 * because the machine code appears to be very compact 133 * and redundant 1-2KB is perfectly tolerable (i.e. 134 * in case the compiler fails to eliminate it:-). By 135 * suggestion from Terrel Larson <terr@terralogic.net>. 136 * 137 * Special notes. 138 * 139 * - compilers (those I've tried) don't seem to have 140 * problems eliminating either the operators guarded 141 * by "if (sizeof(RC4_CHUNK)==8)" or the condition 142 * expressions themselves so I've got 'em to replace 143 * corresponding #ifdefs from the previous version; 144 * - I chose to let the redundant switch cases when 145 * sizeof(RC4_CHUNK)!=8 be (were also #ifdefed 146 * before); 147 * - in case you wonder "&(sizeof(RC4_CHUNK)*8-1)" in 148 * [LB]ESHFT guards against "shift is out of range" 149 * warnings when sizeof(RC4_CHUNK)!=8 150 * 151 * <appro@fy.chalmers.se> 152 */ 153 if (BYTE_ORDER != LITTLE_ENDIAN) { /* BIG-ENDIAN CASE */ 154 # define BESHFT(c) (((sizeof(RC4_CHUNK)-(c)-1)*8)&(sizeof(RC4_CHUNK)*8-1)) 155 for (; len & (0 - sizeof(RC4_CHUNK)); len -= sizeof(RC4_CHUNK)) { 156 ichunk = *(RC4_CHUNK *)indata; 157 otp = RC4_STEP << BESHFT(0); 158 otp |= RC4_STEP << BESHFT(1); 159 otp |= RC4_STEP << BESHFT(2); 160 otp |= RC4_STEP << BESHFT(3); 161 if (sizeof(RC4_CHUNK) == 8) { 162 otp |= RC4_STEP << BESHFT(4); 163 otp |= RC4_STEP << BESHFT(5); 164 otp |= RC4_STEP << BESHFT(6); 165 otp |= RC4_STEP << BESHFT(7); 166 } 167 *(RC4_CHUNK *)outdata = otp^ichunk; 168 indata += sizeof(RC4_CHUNK); 169 outdata += sizeof(RC4_CHUNK); 170 } 171 } else { /* LITTLE-ENDIAN CASE */ 172 # define LESHFT(c) (((c)*8)&(sizeof(RC4_CHUNK)*8-1)) 173 for (; len & (0 - sizeof(RC4_CHUNK)); len -= sizeof(RC4_CHUNK)) { 174 ichunk = *(RC4_CHUNK *)indata; 175 otp = RC4_STEP; 176 otp |= RC4_STEP << 8; 177 otp |= RC4_STEP << 16; 178 otp |= RC4_STEP << 24; 179 if (sizeof(RC4_CHUNK) == 8) { 180 otp |= RC4_STEP << LESHFT(4); 181 otp |= RC4_STEP << LESHFT(5); 182 otp |= RC4_STEP << LESHFT(6); 183 otp |= RC4_STEP << LESHFT(7); 184 } 185 *(RC4_CHUNK *)outdata = otp ^ ichunk; 186 indata += sizeof(RC4_CHUNK); 187 outdata += sizeof(RC4_CHUNK); 188 } 189 } 190 } 191 #endif 192 #define LOOP(in,out) \ 193 x=((x+1)&0xff); \ 194 tx=d[x]; \ 195 y=(tx+y)&0xff; \ 196 d[x]=ty=d[y]; \ 197 d[y]=tx; \ 198 (out) = d[(tx+ty)&0xff]^ (in); 199 200 #ifndef RC4_INDEX 201 #define RC4_LOOP(a,b,i) LOOP(*((a)++),*((b)++)) 202 #else 203 #define RC4_LOOP(a,b,i) LOOP(a[i],b[i]) 204 #endif 205 206 i = len >> 3; 207 if (i) { 208 for (;;) { 209 RC4_LOOP(indata, outdata, 0); 210 RC4_LOOP(indata, outdata, 1); 211 RC4_LOOP(indata, outdata, 2); 212 RC4_LOOP(indata, outdata, 3); 213 RC4_LOOP(indata, outdata, 4); 214 RC4_LOOP(indata, outdata, 5); 215 RC4_LOOP(indata, outdata, 6); 216 RC4_LOOP(indata, outdata, 7); 217 #ifdef RC4_INDEX 218 indata += 8; 219 outdata += 8; 220 #endif 221 if (--i == 0) 222 break; 223 } 224 } 225 i = len&0x07; 226 if (i) { 227 for (;;) { 228 RC4_LOOP(indata, outdata, 0); 229 if (--i == 0) 230 break; 231 RC4_LOOP(indata, outdata, 1); 232 if (--i == 0) 233 break; 234 RC4_LOOP(indata, outdata, 2); 235 if (--i == 0) 236 break; 237 RC4_LOOP(indata, outdata, 3); 238 if (--i == 0) 239 break; 240 RC4_LOOP(indata, outdata, 4); 241 if (--i == 0) 242 break; 243 RC4_LOOP(indata, outdata, 5); 244 if (--i == 0) 245 break; 246 RC4_LOOP(indata, outdata, 6); 247 if (--i == 0) 248 break; 249 } 250 } 251 key->x = x; 252 key->y = y; 253 } 254