1
/* crypto/rc4/rc4_enc.c */
2
/* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
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.
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).
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.
23
* Redistribution and use in source and binary forms, with or without
24
* modification, are permitted provided that the following conditions
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)"
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
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.]
61
/* RC4 as implemented from a posting from
62
* Newsgroups: sci.crypt
63
* From: sterndark@netcom.com (David Sterndark)
64
* Subject: RC4 Algorithm revealed.
65
* Message-ID: <sternCvKL4B.Hyy@netcom.com>
66
* Date: Wed, 14 Sep 1994 06:35:31 GMT
69
void RC4(RC4_KEY *key, unsigned long len, const unsigned char *indata,
70
unsigned char *outdata)
73
register RC4_INT x,y,tx,ty;
80
#if defined(RC4_CHUNK)
82
* The original reason for implementing this(*) was the fact that
83
* pre-21164a Alpha CPUs don't have byte load/store instructions
84
* and e.g. a byte store has to be done with 64-bit load, shift,
85
* and, or and finally 64-bit store. Peaking data and operating
86
* at natural word size made it possible to reduce amount of
87
* instructions as well as to perform early read-ahead without
88
* suffering from RAW (read-after-write) hazard. This resulted
89
* in ~40%(**) performance improvement on 21064 box with gcc.
90
* But it's not only Alpha users who win here:-) Thanks to the
91
* early-n-wide read-ahead this implementation also exhibits
92
* >40% speed-up on SPARC and 20-30% on 64-bit MIPS (depending
93
* on sizeof(RC4_INT)).
95
* (*) "this" means code which recognizes the case when input
96
* and output pointers appear to be aligned at natural CPU
98
* (**) i.e. according to 'apps/openssl speed rc4' benchmark,
99
* crypto/rc4/rc4speed.c exhibits almost 70% speed-up...
103
* - RC4_CHUNK="unsigned long long" should be a #1 choice for
104
* UltraSPARC. Unfortunately gcc generates very slow code
105
* (2.5-3 times slower than one generated by Sun's WorkShop
106
* C) and therefore gcc (at least 2.95 and earlier) should
107
* always be told that RC4_CHUNK="unsigned long".
109
* <appro@fy.chalmers.se>
112
# define RC4_STEP ( \
119
(RC4_CHUNK)d[(tx+ty)&0xff]\
122
if ( ( ((unsigned long)indata & (sizeof(RC4_CHUNK)-1)) |
123
((unsigned long)outdata & (sizeof(RC4_CHUNK)-1)) ) == 0 )
125
RC4_CHUNK ichunk,otp;
126
const union { long one; char little; } is_endian = {1};
129
* I reckon we can afford to implement both endian
130
* cases and to decide which way to take at run-time
131
* because the machine code appears to be very compact
132
* and redundant 1-2KB is perfectly tolerable (i.e.
133
* in case the compiler fails to eliminate it:-). By
134
* suggestion from Terrel Larson <terr@terralogic.net>
135
* who also stands for the is_endian union:-)
139
* - is_endian is declared automatic as doing otherwise
140
* (declaring static) prevents gcc from eliminating
141
* the redundant code;
142
* - compilers (those I've tried) don't seem to have
143
* problems eliminating either the operators guarded
144
* by "if (sizeof(RC4_CHUNK)==8)" or the condition
145
* expressions themselves so I've got 'em to replace
146
* corresponding #ifdefs from the previous version;
147
* - I chose to let the redundant switch cases when
148
* sizeof(RC4_CHUNK)!=8 be (were also #ifdefed
150
* - in case you wonder "&(sizeof(RC4_CHUNK)*8-1)" in
151
* [LB]ESHFT guards against "shift is out of range"
152
* warnings when sizeof(RC4_CHUNK)!=8
154
* <appro@fy.chalmers.se>
156
if (!is_endian.little)
157
{ /* BIG-ENDIAN CASE */
158
# define BESHFT(c) (((sizeof(RC4_CHUNK)-(c)-1)*8)&(sizeof(RC4_CHUNK)*8-1))
159
for (;len&-sizeof(RC4_CHUNK);len-=sizeof(RC4_CHUNK))
161
ichunk = *(RC4_CHUNK *)indata;
162
otp = RC4_STEP<<BESHFT(0);
163
otp |= RC4_STEP<<BESHFT(1);
164
otp |= RC4_STEP<<BESHFT(2);
165
otp |= RC4_STEP<<BESHFT(3);
166
if (sizeof(RC4_CHUNK)==8)
168
otp |= RC4_STEP<<BESHFT(4);
169
otp |= RC4_STEP<<BESHFT(5);
170
otp |= RC4_STEP<<BESHFT(6);
171
otp |= RC4_STEP<<BESHFT(7);
173
*(RC4_CHUNK *)outdata = otp^ichunk;
174
indata += sizeof(RC4_CHUNK);
175
outdata += sizeof(RC4_CHUNK);
179
RC4_CHUNK mask=(RC4_CHUNK)-1, ochunk;
181
ichunk = *(RC4_CHUNK *)indata;
182
ochunk = *(RC4_CHUNK *)outdata;
185
mask <<= (sizeof(RC4_CHUNK)-len)<<3;
186
switch (len&(sizeof(RC4_CHUNK)-1))
188
case 7: otp = RC4_STEP<<i, i-=8;
189
case 6: otp |= RC4_STEP<<i, i-=8;
190
case 5: otp |= RC4_STEP<<i, i-=8;
191
case 4: otp |= RC4_STEP<<i, i-=8;
192
case 3: otp |= RC4_STEP<<i, i-=8;
193
case 2: otp |= RC4_STEP<<i, i-=8;
194
case 1: otp |= RC4_STEP<<i, i-=8;
196
* it's never the case,
197
* but it has to be here
202
ochunk |= (otp^ichunk) & mask;
203
*(RC4_CHUNK *)outdata = ochunk;
210
{ /* LITTLE-ENDIAN CASE */
211
# define LESHFT(c) (((c)*8)&(sizeof(RC4_CHUNK)*8-1))
212
for (;len&-sizeof(RC4_CHUNK);len-=sizeof(RC4_CHUNK))
214
ichunk = *(RC4_CHUNK *)indata;
219
if (sizeof(RC4_CHUNK)==8)
221
otp |= RC4_STEP<<LESHFT(4);
222
otp |= RC4_STEP<<LESHFT(5);
223
otp |= RC4_STEP<<LESHFT(6);
224
otp |= RC4_STEP<<LESHFT(7);
226
*(RC4_CHUNK *)outdata = otp^ichunk;
227
indata += sizeof(RC4_CHUNK);
228
outdata += sizeof(RC4_CHUNK);
232
RC4_CHUNK mask=(RC4_CHUNK)-1, ochunk;
234
ichunk = *(RC4_CHUNK *)indata;
235
ochunk = *(RC4_CHUNK *)outdata;
238
mask >>= (sizeof(RC4_CHUNK)-len)<<3;
239
switch (len&(sizeof(RC4_CHUNK)-1))
241
case 7: otp = RC4_STEP, i+=8;
242
case 6: otp |= RC4_STEP<<i, i+=8;
243
case 5: otp |= RC4_STEP<<i, i+=8;
244
case 4: otp |= RC4_STEP<<i, i+=8;
245
case 3: otp |= RC4_STEP<<i, i+=8;
246
case 2: otp |= RC4_STEP<<i, i+=8;
247
case 1: otp |= RC4_STEP<<i, i+=8;
249
* it's never the case,
250
* but it has to be here
255
ochunk |= (otp^ichunk) & mask;
256
*(RC4_CHUNK *)outdata = ochunk;
264
#define LOOP(in,out) \
270
(out) = d[(tx+ty)&0xff]^ (in);
273
#define RC4_LOOP(a,b,i) LOOP(*((a)++),*((b)++))
275
#define RC4_LOOP(a,b,i) LOOP(a[i],b[i])
283
RC4_LOOP(indata,outdata,0);
284
RC4_LOOP(indata,outdata,1);
285
RC4_LOOP(indata,outdata,2);
286
RC4_LOOP(indata,outdata,3);
287
RC4_LOOP(indata,outdata,4);
288
RC4_LOOP(indata,outdata,5);
289
RC4_LOOP(indata,outdata,6);
290
RC4_LOOP(indata,outdata,7);
303
RC4_LOOP(indata,outdata,0); if (--i == 0) break;
304
RC4_LOOP(indata,outdata,1); if (--i == 0) break;
305
RC4_LOOP(indata,outdata,2); if (--i == 0) break;
306
RC4_LOOP(indata,outdata,3); if (--i == 0) break;
307
RC4_LOOP(indata,outdata,4); if (--i == 0) break;
308
RC4_LOOP(indata,outdata,5); if (--i == 0) break;
309
RC4_LOOP(indata,outdata,6); if (--i == 0) break;