155714Skris/* crypto/rc4/rc4_enc.c */ 255714Skris/* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com) 355714Skris * All rights reserved. 455714Skris * 555714Skris * This package is an SSL implementation written 655714Skris * by Eric Young (eay@cryptsoft.com). 755714Skris * The implementation was written so as to conform with Netscapes SSL. 855714Skris * 955714Skris * This library is free for commercial and non-commercial use as long as 1055714Skris * the following conditions are aheared to. The following conditions 1155714Skris * apply to all code found in this distribution, be it the RC4, RSA, 1255714Skris * lhash, DES, etc., code; not just the SSL code. The SSL documentation 1355714Skris * included with this distribution is covered by the same copyright terms 1455714Skris * except that the holder is Tim Hudson (tjh@cryptsoft.com). 1555714Skris * 1655714Skris * Copyright remains Eric Young's, and as such any Copyright notices in 1755714Skris * the code are not to be removed. 1855714Skris * If this package is used in a product, Eric Young should be given attribution 1955714Skris * as the author of the parts of the library used. 2055714Skris * This can be in the form of a textual message at program startup or 2155714Skris * in documentation (online or textual) provided with the package. 2255714Skris * 2355714Skris * Redistribution and use in source and binary forms, with or without 2455714Skris * modification, are permitted provided that the following conditions 2555714Skris * are met: 2655714Skris * 1. Redistributions of source code must retain the copyright 2755714Skris * notice, this list of conditions and the following disclaimer. 2855714Skris * 2. Redistributions in binary form must reproduce the above copyright 2955714Skris * notice, this list of conditions and the following disclaimer in the 3055714Skris * documentation and/or other materials provided with the distribution. 3155714Skris * 3. All advertising materials mentioning features or use of this software 3255714Skris * must display the following acknowledgement: 3355714Skris * "This product includes cryptographic software written by 3455714Skris * Eric Young (eay@cryptsoft.com)" 3555714Skris * The word 'cryptographic' can be left out if the rouines from the library 3655714Skris * being used are not cryptographic related :-). 3755714Skris * 4. If you include any Windows specific code (or a derivative thereof) from 3855714Skris * the apps directory (application code) you must include an acknowledgement: 3955714Skris * "This product includes software written by Tim Hudson (tjh@cryptsoft.com)" 4055714Skris * 4155714Skris * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND 4255714Skris * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 4355714Skris * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 4455714Skris * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 4555714Skris * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 4655714Skris * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 4755714Skris * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 4855714Skris * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 4955714Skris * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 5055714Skris * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 5155714Skris * SUCH DAMAGE. 5255714Skris * 5355714Skris * The licence and distribution terms for any publically available version or 5455714Skris * derivative of this code cannot be changed. i.e. this code cannot simply be 5555714Skris * copied and put under another distribution licence 5655714Skris * [including the GNU Public Licence.] 5755714Skris */ 5855714Skris 5955714Skris#include <openssl/rc4.h> 6055714Skris#include "rc4_locl.h" 6155714Skris 6255714Skris/* RC4 as implemented from a posting from 6355714Skris * Newsgroups: sci.crypt 6455714Skris * From: sterndark@netcom.com (David Sterndark) 6555714Skris * Subject: RC4 Algorithm revealed. 6655714Skris * Message-ID: <sternCvKL4B.Hyy@netcom.com> 6755714Skris * Date: Wed, 14 Sep 1994 06:35:31 GMT 6855714Skris */ 6955714Skris 70238405Sjkimvoid RC4(RC4_KEY *key, size_t len, const unsigned char *indata, 7155714Skris unsigned char *outdata) 7255714Skris { 7355714Skris register RC4_INT *d; 7455714Skris register RC4_INT x,y,tx,ty; 75238405Sjkim size_t i; 7655714Skris 7755714Skris x=key->x; 7855714Skris y=key->y; 7955714Skris d=key->data; 8055714Skris 8159191Skris#if defined(RC4_CHUNK) 8259191Skris /* 8359191Skris * The original reason for implementing this(*) was the fact that 8459191Skris * pre-21164a Alpha CPUs don't have byte load/store instructions 8559191Skris * and e.g. a byte store has to be done with 64-bit load, shift, 8659191Skris * and, or and finally 64-bit store. Peaking data and operating 8759191Skris * at natural word size made it possible to reduce amount of 8859191Skris * instructions as well as to perform early read-ahead without 8959191Skris * suffering from RAW (read-after-write) hazard. This resulted 9059191Skris * in ~40%(**) performance improvement on 21064 box with gcc. 9159191Skris * But it's not only Alpha users who win here:-) Thanks to the 9259191Skris * early-n-wide read-ahead this implementation also exhibits 9359191Skris * >40% speed-up on SPARC and 20-30% on 64-bit MIPS (depending 9459191Skris * on sizeof(RC4_INT)). 9559191Skris * 9659191Skris * (*) "this" means code which recognizes the case when input 9759191Skris * and output pointers appear to be aligned at natural CPU 9859191Skris * word boundary 9959191Skris * (**) i.e. according to 'apps/openssl speed rc4' benchmark, 10059191Skris * crypto/rc4/rc4speed.c exhibits almost 70% speed-up... 10159191Skris * 10259191Skris * Cavets. 10359191Skris * 10459191Skris * - RC4_CHUNK="unsigned long long" should be a #1 choice for 10559191Skris * UltraSPARC. Unfortunately gcc generates very slow code 10659191Skris * (2.5-3 times slower than one generated by Sun's WorkShop 10759191Skris * C) and therefore gcc (at least 2.95 and earlier) should 10859191Skris * always be told that RC4_CHUNK="unsigned long". 10959191Skris * 11059191Skris * <appro@fy.chalmers.se> 11159191Skris */ 11259191Skris 11359191Skris# define RC4_STEP ( \ 11459191Skris x=(x+1) &0xff, \ 11559191Skris tx=d[x], \ 11659191Skris y=(tx+y)&0xff, \ 11759191Skris ty=d[y], \ 11859191Skris d[y]=tx, \ 11959191Skris d[x]=ty, \ 12059191Skris (RC4_CHUNK)d[(tx+ty)&0xff]\ 12159191Skris ) 12259191Skris 123238405Sjkim if ( ( ((size_t)indata & (sizeof(RC4_CHUNK)-1)) | 124238405Sjkim ((size_t)outdata & (sizeof(RC4_CHUNK)-1)) ) == 0 ) 12559191Skris { 12659191Skris RC4_CHUNK ichunk,otp; 12759191Skris const union { long one; char little; } is_endian = {1}; 12859191Skris 12959191Skris /* 13059191Skris * I reckon we can afford to implement both endian 13159191Skris * cases and to decide which way to take at run-time 13259191Skris * because the machine code appears to be very compact 13359191Skris * and redundant 1-2KB is perfectly tolerable (i.e. 13459191Skris * in case the compiler fails to eliminate it:-). By 13559191Skris * suggestion from Terrel Larson <terr@terralogic.net> 13659191Skris * who also stands for the is_endian union:-) 13759191Skris * 13859191Skris * Special notes. 13959191Skris * 14059191Skris * - is_endian is declared automatic as doing otherwise 14159191Skris * (declaring static) prevents gcc from eliminating 14259191Skris * the redundant code; 14359191Skris * - compilers (those I've tried) don't seem to have 14459191Skris * problems eliminating either the operators guarded 14559191Skris * by "if (sizeof(RC4_CHUNK)==8)" or the condition 14659191Skris * expressions themselves so I've got 'em to replace 14759191Skris * corresponding #ifdefs from the previous version; 14859191Skris * - I chose to let the redundant switch cases when 14959191Skris * sizeof(RC4_CHUNK)!=8 be (were also #ifdefed 15059191Skris * before); 15159191Skris * - in case you wonder "&(sizeof(RC4_CHUNK)*8-1)" in 15259191Skris * [LB]ESHFT guards against "shift is out of range" 15359191Skris * warnings when sizeof(RC4_CHUNK)!=8 15459191Skris * 15559191Skris * <appro@fy.chalmers.se> 15659191Skris */ 15759191Skris if (!is_endian.little) 15859191Skris { /* BIG-ENDIAN CASE */ 15959191Skris# define BESHFT(c) (((sizeof(RC4_CHUNK)-(c)-1)*8)&(sizeof(RC4_CHUNK)*8-1)) 160238405Sjkim for (;len&(0-sizeof(RC4_CHUNK));len-=sizeof(RC4_CHUNK)) 16159191Skris { 16259191Skris ichunk = *(RC4_CHUNK *)indata; 16359191Skris otp = RC4_STEP<<BESHFT(0); 16459191Skris otp |= RC4_STEP<<BESHFT(1); 16559191Skris otp |= RC4_STEP<<BESHFT(2); 16659191Skris otp |= RC4_STEP<<BESHFT(3); 16759191Skris if (sizeof(RC4_CHUNK)==8) 16859191Skris { 16959191Skris otp |= RC4_STEP<<BESHFT(4); 17059191Skris otp |= RC4_STEP<<BESHFT(5); 17159191Skris otp |= RC4_STEP<<BESHFT(6); 17259191Skris otp |= RC4_STEP<<BESHFT(7); 17359191Skris } 17459191Skris *(RC4_CHUNK *)outdata = otp^ichunk; 17559191Skris indata += sizeof(RC4_CHUNK); 17659191Skris outdata += sizeof(RC4_CHUNK); 17759191Skris } 17859191Skris if (len) 17959191Skris { 18059191Skris RC4_CHUNK mask=(RC4_CHUNK)-1, ochunk; 18159191Skris 18259191Skris ichunk = *(RC4_CHUNK *)indata; 18359191Skris ochunk = *(RC4_CHUNK *)outdata; 18459191Skris otp = 0; 18559191Skris i = BESHFT(0); 18659191Skris mask <<= (sizeof(RC4_CHUNK)-len)<<3; 18759191Skris switch (len&(sizeof(RC4_CHUNK)-1)) 18859191Skris { 18959191Skris case 7: otp = RC4_STEP<<i, i-=8; 19059191Skris case 6: otp |= RC4_STEP<<i, i-=8; 19159191Skris case 5: otp |= RC4_STEP<<i, i-=8; 19259191Skris case 4: otp |= RC4_STEP<<i, i-=8; 19359191Skris case 3: otp |= RC4_STEP<<i, i-=8; 19459191Skris case 2: otp |= RC4_STEP<<i, i-=8; 19559191Skris case 1: otp |= RC4_STEP<<i, i-=8; 19659191Skris case 0: ; /* 19759191Skris * it's never the case, 19859191Skris * but it has to be here 19959191Skris * for ultrix? 20059191Skris */ 20159191Skris } 20259191Skris ochunk &= ~mask; 20359191Skris ochunk |= (otp^ichunk) & mask; 20459191Skris *(RC4_CHUNK *)outdata = ochunk; 20559191Skris } 20659191Skris key->x=x; 20759191Skris key->y=y; 20859191Skris return; 20959191Skris } 21059191Skris else 21159191Skris { /* LITTLE-ENDIAN CASE */ 21259191Skris# define LESHFT(c) (((c)*8)&(sizeof(RC4_CHUNK)*8-1)) 213238405Sjkim for (;len&(0-sizeof(RC4_CHUNK));len-=sizeof(RC4_CHUNK)) 21459191Skris { 21559191Skris ichunk = *(RC4_CHUNK *)indata; 21659191Skris otp = RC4_STEP; 21759191Skris otp |= RC4_STEP<<8; 21859191Skris otp |= RC4_STEP<<16; 21959191Skris otp |= RC4_STEP<<24; 22059191Skris if (sizeof(RC4_CHUNK)==8) 22159191Skris { 22259191Skris otp |= RC4_STEP<<LESHFT(4); 22359191Skris otp |= RC4_STEP<<LESHFT(5); 22459191Skris otp |= RC4_STEP<<LESHFT(6); 22559191Skris otp |= RC4_STEP<<LESHFT(7); 22659191Skris } 22759191Skris *(RC4_CHUNK *)outdata = otp^ichunk; 22859191Skris indata += sizeof(RC4_CHUNK); 22959191Skris outdata += sizeof(RC4_CHUNK); 23059191Skris } 23159191Skris if (len) 23259191Skris { 23359191Skris RC4_CHUNK mask=(RC4_CHUNK)-1, ochunk; 23459191Skris 23559191Skris ichunk = *(RC4_CHUNK *)indata; 23659191Skris ochunk = *(RC4_CHUNK *)outdata; 23759191Skris otp = 0; 23859191Skris i = 0; 23959191Skris mask >>= (sizeof(RC4_CHUNK)-len)<<3; 24059191Skris switch (len&(sizeof(RC4_CHUNK)-1)) 24159191Skris { 24259191Skris case 7: otp = RC4_STEP, i+=8; 24359191Skris case 6: otp |= RC4_STEP<<i, i+=8; 24459191Skris case 5: otp |= RC4_STEP<<i, i+=8; 24559191Skris case 4: otp |= RC4_STEP<<i, i+=8; 24659191Skris case 3: otp |= RC4_STEP<<i, i+=8; 24759191Skris case 2: otp |= RC4_STEP<<i, i+=8; 24859191Skris case 1: otp |= RC4_STEP<<i, i+=8; 24959191Skris case 0: ; /* 25059191Skris * it's never the case, 25159191Skris * but it has to be here 25259191Skris * for ultrix? 25359191Skris */ 25459191Skris } 25559191Skris ochunk &= ~mask; 25659191Skris ochunk |= (otp^ichunk) & mask; 25759191Skris *(RC4_CHUNK *)outdata = ochunk; 25859191Skris } 25959191Skris key->x=x; 26059191Skris key->y=y; 26159191Skris return; 26259191Skris } 26359191Skris } 26459191Skris#endif 26555714Skris#define LOOP(in,out) \ 26655714Skris x=((x+1)&0xff); \ 26755714Skris tx=d[x]; \ 26855714Skris y=(tx+y)&0xff; \ 26955714Skris d[x]=ty=d[y]; \ 27055714Skris d[y]=tx; \ 27155714Skris (out) = d[(tx+ty)&0xff]^ (in); 27255714Skris 27355714Skris#ifndef RC4_INDEX 27455714Skris#define RC4_LOOP(a,b,i) LOOP(*((a)++),*((b)++)) 27555714Skris#else 27655714Skris#define RC4_LOOP(a,b,i) LOOP(a[i],b[i]) 27755714Skris#endif 27855714Skris 279238405Sjkim i=len>>3; 28055714Skris if (i) 28155714Skris { 28255714Skris for (;;) 28355714Skris { 28455714Skris RC4_LOOP(indata,outdata,0); 28555714Skris RC4_LOOP(indata,outdata,1); 28655714Skris RC4_LOOP(indata,outdata,2); 28755714Skris RC4_LOOP(indata,outdata,3); 28855714Skris RC4_LOOP(indata,outdata,4); 28955714Skris RC4_LOOP(indata,outdata,5); 29055714Skris RC4_LOOP(indata,outdata,6); 29155714Skris RC4_LOOP(indata,outdata,7); 29255714Skris#ifdef RC4_INDEX 29355714Skris indata+=8; 29455714Skris outdata+=8; 29555714Skris#endif 29655714Skris if (--i == 0) break; 29755714Skris } 29855714Skris } 299238405Sjkim i=len&0x07; 30055714Skris if (i) 30155714Skris { 30255714Skris for (;;) 30355714Skris { 30455714Skris RC4_LOOP(indata,outdata,0); if (--i == 0) break; 30555714Skris RC4_LOOP(indata,outdata,1); if (--i == 0) break; 30655714Skris RC4_LOOP(indata,outdata,2); if (--i == 0) break; 30755714Skris RC4_LOOP(indata,outdata,3); if (--i == 0) break; 30855714Skris RC4_LOOP(indata,outdata,4); if (--i == 0) break; 30955714Skris RC4_LOOP(indata,outdata,5); if (--i == 0) break; 31055714Skris RC4_LOOP(indata,outdata,6); if (--i == 0) break; 31155714Skris } 31255714Skris } 31355714Skris key->x=x; 31455714Skris key->y=y; 31555714Skris } 316