cryptlib  3.4.1
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Properties Macros
rc4enc.c
Go to the documentation of this file.
1 /* crypto/rc4/rc4_enc.c */
2 /* Copyright (C) 1995-1998 Eric Young ([email protected])
3  * All rights reserved.
4  *
5  * This package is an SSL implementation written
6  * by Eric Young ([email protected]).
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 ([email protected]).
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 ([email protected])"
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 ([email protected])"
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 #if defined( INC_ALL )
60  #include "osconfig.h"
61  #include "rc4.h"
62  #include "rc4locl.h"
63 #else
64  #include "crypt/osconfig.h"
65  #include "crypt/rc4.h"
66  #include "crypt/rc4locl.h"
67 #endif /* Compiler-specific includes */
68 
69 /* RC4 as implemented from a posting from
70  * Newsgroups: sci.crypt
71  * From: [email protected] (David Sterndark)
72  * Subject: RC4 Algorithm revealed.
73  * Message-ID: <[email protected]>
74  * Date: Wed, 14 Sep 1994 06:35:31 GMT
75  */
76 
77 void RC4(RC4_KEY *key, unsigned long len, const unsigned char *indata,
78  unsigned char *outdata)
79  {
80  register RC4_INT *d;
81  register RC4_INT x,y,tx,ty;
82  int i;
83 
84  x=key->x;
85  y=key->y;
86  d=key->data;
87 
88 #if defined(RC4_CHUNK)
89  /*
90  * The original reason for implementing this(*) was the fact that
91  * pre-21164a Alpha CPUs don't have byte load/store instructions
92  * and e.g. a byte store has to be done with 64-bit load, shift,
93  * and, or and finally 64-bit store. Peaking data and operating
94  * at natural word size made it possible to reduce amount of
95  * instructions as well as to perform early read-ahead without
96  * suffering from RAW (read-after-write) hazard. This resulted
97  * in ~40%(**) performance improvement on 21064 box with gcc.
98  * But it's not only Alpha users who win here:-) Thanks to the
99  * early-n-wide read-ahead this implementation also exhibits
100  * >40% speed-up on SPARC and 20-30% on 64-bit MIPS (depending
101  * on sizeof(RC4_INT)).
102  *
103  * (*) "this" means code which recognizes the case when input
104  * and output pointers appear to be aligned at natural CPU
105  * word boundary
106  * (**) i.e. according to 'apps/openssl speed rc4' benchmark,
107  * crypto/rc4/rc4speed.c exhibits almost 70% speed-up...
108  *
109  * Cavets.
110  *
111  * - RC4_CHUNK="unsigned long long" should be a #1 choice for
112  * UltraSPARC. Unfortunately gcc generates very slow code
113  * (2.5-3 times slower than one generated by Sun's WorkShop
114  * C) and therefore gcc (at least 2.95 and earlier) should
115  * always be told that RC4_CHUNK="unsigned long".
116  *
118  */
119 
120 # define RC4_STEP ( \
121  x=(x+1) &0xff, \
122  tx=d[x], \
123  y=(tx+y)&0xff, \
124  ty=d[y], \
125  d[y]=tx, \
126  d[x]=ty, \
127  (RC4_CHUNK)d[(tx+ty)&0xff]\
128  )
129 
130  if ( ( ((unsigned long)indata & (sizeof(RC4_CHUNK)-1)) |
131  ((unsigned long)outdata & (sizeof(RC4_CHUNK)-1)) ) == 0 )
132  {
133  RC4_CHUNK ichunk,otp;
134  const union { long one; char little; } is_endian = {1};
135 
136  /*
137  * I reckon we can afford to implement both endian
138  * cases and to decide which way to take at run-time
139  * because the machine code appears to be very compact
140  * and redundant 1-2KB is perfectly tolerable (i.e.
141  * in case the compiler fails to eliminate it:-). By
142  * suggestion from Terrel Larson <[email protected]>
143  * who also stands for the is_endian union:-)
144  *
145  * Special notes.
146  *
147  * - is_endian is declared automatic as doing otherwise
148  * (declaring static) prevents gcc from eliminating
149  * the redundant code;
150  * - compilers (those I've tried) don't seem to have
151  * problems eliminating either the operators guarded
152  * by "if (sizeof(RC4_CHUNK)==8)" or the condition
153  * expressions themselves so I've got 'em to replace
154  * corresponding #ifdefs from the previous version;
155  * - I chose to let the redundant switch cases when
156  * sizeof(RC4_CHUNK)!=8 be (were also #ifdefed
157  * before);
158  * - in case you wonder "&(sizeof(RC4_CHUNK)*8-1)" in
159  * [LB]ESHFT guards against "shift is out of range"
160  * warnings when sizeof(RC4_CHUNK)!=8
161  *
163  */
164  if (!is_endian.little)
165  { /* BIG-ENDIAN CASE */
166 # define BESHFT(c) (((sizeof(RC4_CHUNK)-(c)-1)*8)&(sizeof(RC4_CHUNK)*8-1))
167  for (;len&-sizeof(RC4_CHUNK);len-=sizeof(RC4_CHUNK))
168  {
169  ichunk = *(RC4_CHUNK *)indata;
170  otp = RC4_STEP<<BESHFT(0);
171  otp |= RC4_STEP<<BESHFT(1);
172  otp |= RC4_STEP<<BESHFT(2);
173  otp |= RC4_STEP<<BESHFT(3);
174  if (sizeof(RC4_CHUNK)==8)
175  {
176  otp |= RC4_STEP<<BESHFT(4);
177  otp |= RC4_STEP<<BESHFT(5);
178  otp |= RC4_STEP<<BESHFT(6);
179  otp |= RC4_STEP<<BESHFT(7);
180  }
181  *(RC4_CHUNK *)outdata = otp^ichunk;
182  indata += sizeof(RC4_CHUNK);
183  outdata += sizeof(RC4_CHUNK);
184  }
185  if (len)
186  {
187  RC4_CHUNK mask=(RC4_CHUNK)-1, ochunk;
188 
189  ichunk = *(RC4_CHUNK *)indata;
190  ochunk = *(RC4_CHUNK *)outdata;
191  otp = 0;
192  i = BESHFT(0);
193  mask <<= (sizeof(RC4_CHUNK)-len)<<3;
194  switch (len&(sizeof(RC4_CHUNK)-1))
195  {
196  case 7: otp = RC4_STEP<<i, i-=8;
197  case 6: otp |= RC4_STEP<<i, i-=8;
198  case 5: otp |= RC4_STEP<<i, i-=8;
199  case 4: otp |= RC4_STEP<<i, i-=8;
200  case 3: otp |= RC4_STEP<<i, i-=8;
201  case 2: otp |= RC4_STEP<<i, i-=8;
202  case 1: otp |= RC4_STEP<<i, i-=8;
203  case 0: ; /*
204  * it's never the case,
205  * but it has to be here
206  * for ultrix?
207  */
208  }
209  ochunk &= ~mask;
210  ochunk |= (otp^ichunk) & mask;
211  *(RC4_CHUNK *)outdata = ochunk;
212  }
213  key->x=x;
214  key->y=y;
215  return;
216  }
217  else
218  { /* LITTLE-ENDIAN CASE */
219 # define LESHFT(c) (((c)*8)&(sizeof(RC4_CHUNK)*8-1))
220  for (;len&-sizeof(RC4_CHUNK);len-=sizeof(RC4_CHUNK))
221  {
222  ichunk = *(RC4_CHUNK *)indata;
223  otp = RC4_STEP;
224  otp |= RC4_STEP<<8;
225  otp |= RC4_STEP<<16;
226  otp |= RC4_STEP<<24;
227  if (sizeof(RC4_CHUNK)==8)
228  {
229  otp |= RC4_STEP<<LESHFT(4);
230  otp |= RC4_STEP<<LESHFT(5);
231  otp |= RC4_STEP<<LESHFT(6);
232  otp |= RC4_STEP<<LESHFT(7);
233  }
234  *(RC4_CHUNK *)outdata = otp^ichunk;
235  indata += sizeof(RC4_CHUNK);
236  outdata += sizeof(RC4_CHUNK);
237  }
238  if (len)
239  {
240  RC4_CHUNK mask=(RC4_CHUNK)-1, ochunk;
241 
242  ichunk = *(RC4_CHUNK *)indata;
243  ochunk = *(RC4_CHUNK *)outdata;
244  otp = 0;
245  i = 0;
246  mask >>= (sizeof(RC4_CHUNK)-len)<<3;
247  switch (len&(sizeof(RC4_CHUNK)-1))
248  {
249  case 7: otp = RC4_STEP, i+=8;
250  case 6: otp |= RC4_STEP<<i, i+=8;
251  case 5: otp |= RC4_STEP<<i, i+=8;
252  case 4: otp |= RC4_STEP<<i, i+=8;
253  case 3: otp |= RC4_STEP<<i, i+=8;
254  case 2: otp |= RC4_STEP<<i, i+=8;
255  case 1: otp |= RC4_STEP<<i, i+=8;
256  case 0: ; /*
257  * it's never the case,
258  * but it has to be here
259  * for ultrix?
260  */
261  }
262  ochunk &= ~mask;
263  ochunk |= (otp^ichunk) & mask;
264  *(RC4_CHUNK *)outdata = ochunk;
265  }
266  key->x=x;
267  key->y=y;
268  return;
269  }
270  }
271 #endif
272 #define LOOP(in,out) \
273  x=((x+1)&0xff); \
274  tx=d[x]; \
275  y=(tx+y)&0xff; \
276  d[x]=ty=d[y]; \
277  d[y]=tx; \
278  (out) = d[(tx+ty)&0xff]^ (in);
279 
280 #ifndef RC4_INDEX
281 #define RC4_LOOP(a,b,i) LOOP(*((a)++),*((b)++))
282 #else
283 #define RC4_LOOP(a,b,i) LOOP(a[i],b[i])
284 #endif
285 
286  i=(int)(len>>3L);
287  if (i)
288  {
289  for (;;)
290  {
291  RC4_LOOP(indata,outdata,0);
292  RC4_LOOP(indata,outdata,1);
293  RC4_LOOP(indata,outdata,2);
294  RC4_LOOP(indata,outdata,3);
295  RC4_LOOP(indata,outdata,4);
296  RC4_LOOP(indata,outdata,5);
297  RC4_LOOP(indata,outdata,6);
298  RC4_LOOP(indata,outdata,7);
299 #ifdef RC4_INDEX
300  indata+=8;
301  outdata+=8;
302 #endif
303  if (--i == 0) break;
304  }
305  }
306  i=(int)len&0x07;
307  if (i)
308  {
309  for (;;)
310  {
311  RC4_LOOP(indata,outdata,0); if (--i == 0) break;
312  RC4_LOOP(indata,outdata,1); if (--i == 0) break;
313  RC4_LOOP(indata,outdata,2); if (--i == 0) break;
314  RC4_LOOP(indata,outdata,3); if (--i == 0) break;
315  RC4_LOOP(indata,outdata,4); if (--i == 0) break;
316  RC4_LOOP(indata,outdata,5); if (--i == 0) break;
317  RC4_LOOP(indata,outdata,6); if (--i == 0) break;
318  }
319  }
320  key->x=x;
321  key->y=y;
322  }