xref: /illumos-gate/usr/src/lib/libnsl/des/des_soft.c (revision 141040e8a310da49386b596573e5dde5580572ec)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License, Version 1.0 only
6  * (the "License").  You may not use this file except in compliance
7  * with the License.
8  *
9  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10  * or http://www.opensolaris.org/os/licensing.
11  * See the License for the specific language governing permissions
12  * and limitations under the License.
13  *
14  * When distributing Covered Code, include this CDDL HEADER in each
15  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16  * If applicable, add the following below this CDDL HEADER, with the
17  * fields enclosed by brackets "[]" replaced with your own identifying
18  * information: Portions Copyright [yyyy] [name of copyright owner]
19  *
20  * CDDL HEADER END
21  */
22 
23 /*
24  * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
25  * Use is subject to license terms.
26  */
27 
28 /* Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T */
29 /* All Rights Reserved */
30 
31 /*
32  * Portions of this source code were derived from Berkeley 4.3 BSD
33  * under license from the Regents of the University of California.
34  */
35 
36 #pragma ident	"%Z%%M%	%I%	%E% SMI"
37 
38 /*
39  * Warning!  Things are arranged very carefully in this file to
40  * allow read-only data to be moved to the text segment.  The
41  * various DES tables must appear before any function definitions
42  * (this is arranged by including them immediately below) and partab
43  * must also appear before and function definitions
44  * This arrangement allows all data up through the first text to
45  * be moved to text.
46  */
47 
48 #include "mt.h"
49 #include <sys/types.h>
50 #include <des/softdes.h>
51 #include <des/desdata.h>
52 #ifdef sun
53 #include <sys/ioctl.h>
54 #include <sys/des.h>
55 #else
56 #include <des/des.h>
57 #endif
58 #include <rpcsvc/nis_dhext.h>
59 
60 /*
61  * Fast (?) software implementation of DES
62  * Has been seen going at 2000 bytes/sec on a Sun-2
63  * Works on a VAX too.
64  * Won't work without 8 bit chars and 32 bit longs
65  */
66 
67 #define	btst(k, b)	(k[b >> 3] & (0x80 >> (b & 07)))
68 #define	BIT28	(1<<28)
69 
70 static int	__des_encrypt(uchar_t *, struct deskeydata *);
71 static int	__des_setkey(uchar_t[8], struct deskeydata *, unsigned);
72 
73 
74 /*
75  * Table giving odd parity in the low bit for ASCII characters
76  */
77 const char partab[128] = {
78 	0x01, 0x01, 0x02, 0x02, 0x04, 0x04, 0x07, 0x07,
79 	0x08, 0x08, 0x0b, 0x0b, 0x0d, 0x0d, 0x0e, 0x0e,
80 	0x10, 0x10, 0x13, 0x13, 0x15, 0x15, 0x16, 0x16,
81 	0x19, 0x19, 0x1a, 0x1a, 0x1c, 0x1c, 0x1f, 0x1f,
82 	0x20, 0x20, 0x23, 0x23, 0x25, 0x25, 0x26, 0x26,
83 	0x29, 0x29, 0x2a, 0x2a, 0x2c, 0x2c, 0x2f, 0x2f,
84 	0x31, 0x31, 0x32, 0x32, 0x34, 0x34, 0x37, 0x37,
85 	0x38, 0x38, 0x3b, 0x3b, 0x3d, 0x3d, 0x3e, 0x3e,
86 	0x40, 0x40, 0x43, 0x43, 0x45, 0x45, 0x46, 0x46,
87 	0x49, 0x49, 0x4a, 0x4a, 0x4c, 0x4c, 0x4f, 0x4f,
88 	0x51, 0x51, 0x52, 0x52, 0x54, 0x54, 0x57, 0x57,
89 	0x58, 0x58, 0x5b, 0x5b, 0x5d, 0x5d, 0x5e, 0x5e,
90 	0x61, 0x61, 0x62, 0x62, 0x64, 0x64, 0x67, 0x67,
91 	0x68, 0x68, 0x6b, 0x6b, 0x6d, 0x6d, 0x6e, 0x6e,
92 	0x70, 0x70, 0x73, 0x73, 0x75, 0x75, 0x76, 0x76,
93 	0x79, 0x79, 0x7a, 0x7a, 0x7c, 0x7c, 0x7f, 0x7f,
94 };
95 
96 /*
97  * Add odd parity to low bit of 8 byte key
98  */
99 void
100 des_setparity(char *p)
101 {
102 	int i;
103 
104 	for (i = 0; i < 8; i++) {
105 		*p = partab[*p & 0x7f];
106 		p++;
107 	}
108 }
109 
110 static const unsigned char partab_g[256] = {
111 	0x01, 0x01, 0x02, 0x02, 0x04, 0x04, 0x07, 0x07,
112 	0x08, 0x08, 0x0b, 0x0b, 0x0d, 0x0d, 0x0e, 0x0e,
113 	0x10, 0x10, 0x13, 0x13, 0x15, 0x15, 0x16, 0x16,
114 	0x19, 0x19, 0x1a, 0x1a, 0x1c, 0x1c, 0x1f, 0x1f,
115 	0x20, 0x20, 0x23, 0x23, 0x25, 0x25, 0x26, 0x26,
116 	0x29, 0x29, 0x2a, 0x2a, 0x2c, 0x2c, 0x2f, 0x2f,
117 	0x31, 0x31, 0x32, 0x32, 0x34, 0x34, 0x37, 0x37,
118 	0x38, 0x38, 0x3b, 0x3b, 0x3d, 0x3d, 0x3e, 0x3e,
119 	0x40, 0x40, 0x43, 0x43, 0x45, 0x45, 0x46, 0x46,
120 	0x49, 0x49, 0x4a, 0x4a, 0x4c, 0x4c, 0x4f, 0x4f,
121 	0x51, 0x51, 0x52, 0x52, 0x54, 0x54, 0x57, 0x57,
122 	0x58, 0x58, 0x5b, 0x5b, 0x5d, 0x5d, 0x5e, 0x5e,
123 	0x61, 0x61, 0x62, 0x62, 0x64, 0x64, 0x67, 0x67,
124 	0x68, 0x68, 0x6b, 0x6b, 0x6d, 0x6d, 0x6e, 0x6e,
125 	0x70, 0x70, 0x73, 0x73, 0x75, 0x75, 0x76, 0x76,
126 	0x79, 0x79, 0x7a, 0x7a, 0x7c, 0x7c, 0x7f, 0x7f,
127 	0x80, 0x80, 0x83, 0x83, 0x85, 0x85, 0x86, 0x86,
128 	0x89, 0x89, 0x8a, 0x8a, 0x8c, 0x8c, 0x8f, 0x8f,
129 	0x91, 0x91, 0x92, 0x92, 0x94, 0x94, 0x97, 0x97,
130 	0x98, 0x98, 0x9b, 0x9b, 0x9d, 0x9d, 0x9e, 0x9e,
131 	0xa1, 0xa1, 0xa2, 0xa2, 0xa4, 0xa4, 0xa7, 0xa7,
132 	0xa8, 0xa8, 0xab, 0xab, 0xad, 0xad, 0xae, 0xae,
133 	0xb0, 0xb0, 0xb3, 0xb3, 0xb5, 0xb5, 0xb6, 0xb6,
134 	0xb9, 0xb9, 0xba, 0xba, 0xbc, 0xbc, 0xbf, 0xbf,
135 	0xc1, 0xc1, 0xc2, 0xc2, 0xc4, 0xc4, 0xc7, 0xc7,
136 	0xc8, 0xc8, 0xcb, 0xcb, 0xcd, 0xcd, 0xce, 0xce,
137 	0xd0, 0xd0, 0xd3, 0xd3, 0xd5, 0xd5, 0xd6, 0xd6,
138 	0xd9, 0xd9, 0xda, 0xda, 0xdc, 0xdc, 0xdf, 0xdf,
139 	0xe0, 0xe0, 0xe3, 0xe3, 0xe5, 0xe5, 0xe6, 0xe6,
140 	0xe9, 0xe9, 0xea, 0xea, 0xec, 0xec, 0xef, 0xef,
141 	0xf1, 0xf1, 0xf2, 0xf2, 0xf4, 0xf4, 0xf7, 0xf7,
142 	0xf8, 0xf8, 0xfb, 0xfb, 0xfd, 0xfd, 0xfe, 0xfe
143 };
144 
145 /*
146  * A corrected version of des_setparity (see bug 1149767).
147  */
148 void
149 des_setparity_g(des_block *p)
150 {
151 	int i;
152 
153 	for (i = 0; i < 8; i++) {
154 		(*p).c[i] = partab_g[(*p).c[i]];
155 	}
156 }
157 
158 /*
159  * Software encrypt or decrypt a block of data (multiple of 8 bytes)
160  * Do the CBC ourselves if needed.
161  */
162 int
163 __des_crypt(char *buf, unsigned len, struct desparams *desp)
164 {
165 /* EXPORT DELETE START */
166 	short i;
167 	unsigned mode;
168 	unsigned dir;
169 	char nextiv[8];
170 	struct deskeydata softkey;
171 
172 	mode = (unsigned)desp->des_mode;
173 	dir = (unsigned)desp->des_dir;
174 	(void) __des_setkey(desp->des_key, &softkey, dir);
175 	while (len != 0) {
176 		switch (mode) {
177 		case CBC:
178 			switch (dir) {
179 			case ENCRYPT:
180 				for (i = 0; i < 8; i++)
181 					buf[i] ^= desp->des_ivec[i];
182 				(void) __des_encrypt((uchar_t *)buf, &softkey);
183 				for (i = 0; i < 8; i++)
184 					desp->des_ivec[i] = buf[i];
185 				break;
186 			case DECRYPT:
187 				for (i = 0; i < 8; i++)
188 					nextiv[i] = buf[i];
189 				(void) __des_encrypt((uchar_t *)buf, &softkey);
190 				for (i = 0; i < 8; i++) {
191 					buf[i] ^= desp->des_ivec[i];
192 					desp->des_ivec[i] = nextiv[i];
193 				}
194 				break;
195 			}
196 			break;
197 		case ECB:
198 			(void) __des_encrypt((uchar_t *)buf, &softkey);
199 			break;
200 		}
201 		buf += 8;
202 		len -= 8;
203 	}
204 /* EXPORT DELETE END */
205 	return (1);
206 }
207 
208 
209 /*
210  * Set the key and direction for an encryption operation
211  * We build the 16 key entries here
212  */
213 static int
214 __des_setkey(uchar_t userkey[8], struct deskeydata *kd, unsigned dir)
215 {
216 /* EXPORT DELETE START */
217 	int32_t C, D;
218 	short i;
219 
220 	/*
221 	 * First, generate C and D by permuting
222 	 * the key. The low order bit of each
223 	 * 8-bit char is not used, so C and D are only 28
224 	 * bits apiece.
225 	 */
226 	{
227 		short bit;
228 		const short *pcc = PC1_C, *pcd = PC1_D;
229 
230 		C = D = 0;
231 		for (i = 0; i < 28; i++) {
232 			C <<= 1;
233 			D <<= 1;
234 			bit = *pcc++;
235 			if (btst(userkey, bit))
236 				C |= 1;
237 			bit = *pcd++;
238 			if (btst(userkey, bit))
239 				D |= 1;
240 		}
241 	}
242 	/*
243 	 * To generate Ki, rotate C and D according
244 	 * to schedule and pick up a permutation
245 	 * using PC2.
246 	 */
247 	for (i = 0; i < 16; i++) {
248 		chunk_t *c;
249 		short j, k, bit;
250 		uint32_t bbit;
251 
252 		/*
253 		 * Do the "left shift" (rotate)
254 		 * We know we always rotate by either 1 or 2 bits
255 		 * the shifts table tells us if its 2
256 		 */
257 		C <<= 1;
258 		if (C & BIT28)
259 			C |= 1;
260 		D <<= 1;
261 		if (D & BIT28)
262 			D |= 1;
263 		if (shifts[i]) {
264 			C <<= 1;
265 			if (C & BIT28)
266 				C |= 1;
267 			D <<= 1;
268 			if (D & BIT28)
269 				D |= 1;
270 		}
271 		/*
272 		 * get Ki. Note C and D are concatenated.
273 		 */
274 		bit = 0;
275 		switch (dir) {
276 		case ENCRYPT:
277 			c = &kd->keyval[i]; break;
278 		case DECRYPT:
279 			c = &kd->keyval[15 - i]; break;
280 		}
281 		c->long0 = 0;
282 		c->long1 = 0;
283 		bbit = (1 << 5) << 24;
284 		for (j = 0; j < 4; j++) {
285 			for (k = 0; k < 6; k++) {
286 				if (C & (BIT28 >> PC2_C[bit]))
287 					c->long0 |= bbit >> k;
288 				if (D & (BIT28 >> PC2_D[bit]))
289 					c->long1 |= bbit >> k;
290 				bit++;
291 			}
292 			bbit >>= 8;
293 		}
294 
295 	}
296 /* EXPORT DELETE END */
297 	return (1);
298 }
299 
300 
301 
302 /*
303  * Do an encryption operation
304  * Much pain is taken (with preprocessor) to avoid loops so the compiler
305  * can do address arithmetic instead of doing it at runtime.
306  * Note that the byte-to-chunk conversion is necessary to guarantee
307  * processor byte-order independence.
308  */
309 static int
310 __des_encrypt(uchar_t *data, struct deskeydata *kd)
311 {
312 /* EXPORT DELETE START */
313 	chunk_t work1, work2;
314 
315 	/*
316 	 * Initial permutation
317 	 * and byte to chunk conversion
318 	 */
319 	{
320 		const uint32_t *lp;
321 		uint32_t l0, l1, w;
322 		short i, pbit;
323 
324 		work1.byte0 = data[0];
325 		work1.byte1 = data[1];
326 		work1.byte2 = data[2];
327 		work1.byte3 = data[3];
328 		work1.byte4 = data[4];
329 		work1.byte5 = data[5];
330 		work1.byte6 = data[6];
331 		work1.byte7 = data[7];
332 		l0 = l1 = 0;
333 		w = work1.long0;
334 		for (lp = (uint32_t *)&longtab[0], i = 0; i < 32; i++) {
335 			if (w & *lp++) {
336 				pbit = IPtab[i];
337 				if (pbit < 32)
338 					l0 |= longtab[pbit];
339 				else
340 					l1 |= longtab[pbit-32];
341 			}
342 		}
343 		w = work1.long1;
344 		for (lp = (uint32_t *)&longtab[0], i = 32; i < 64; i++) {
345 			if (w & *lp++) {
346 				pbit = IPtab[i];
347 				if (pbit < 32)
348 					l0 |= longtab[pbit];
349 				else
350 					l1 |= longtab[pbit-32];
351 			}
352 		}
353 		work2.long0 = l0;
354 		work2.long1 = l1;
355 	}
356 
357 /*
358  * Expand 8 bits of 32 bit R to 48 bit R
359  */
360 #define	do_R_to_ER(op, b)	{			\
361 	const struct R_to_ER *p = &R_to_ER_tab[b][R.byte##b];	\
362 	e0 op p->l0;				\
363 	e1 op p->l1;				\
364 }
365 
366 /*
367  * Inner part of the algorithm:
368  * Expand R from 32 to 48 bits; xor key value;
369  * apply S boxes; permute 32 bits of output
370  */
371 /* BEGIN CSTYLED */
372 #define	do_F(iter, inR, outR) 	{			\
373 	chunk_t R, ER;					\
374 	uint32_t e0, e1;				\
375 	R.long0 = inR;					\
376 	do_R_to_ER(=, 0);				\
377 	do_R_to_ER(|=, 1);				\
378 	do_R_to_ER(|=, 2);				\
379 	do_R_to_ER(|=, 3);				\
380 	ER.long0 = e0 ^ kd->keyval[iter].long0;		\
381 	ER.long1 = e1 ^ kd->keyval[iter].long1;		\
382 	R.long0 =					\
383 		S_tab[0][ER.byte0] +			\
384 		S_tab[1][ER.byte1] +			\
385 		S_tab[2][ER.byte2] +			\
386 		S_tab[3][ER.byte3] +			\
387 		S_tab[4][ER.byte4] +			\
388 		S_tab[5][ER.byte5] +			\
389 		S_tab[6][ER.byte6] +			\
390 		S_tab[7][ER.byte7]; 			\
391 	outR =						\
392 		P_tab[0][R.byte0] +			\
393 		P_tab[1][R.byte1] +			\
394 		P_tab[2][R.byte2] +			\
395 		P_tab[3][R.byte3]; 			\
396 }
397 /* END CSTYLED */
398 
399 /*
400  * Do a cipher step
401  * Apply inner part; do xor and exchange of 32 bit parts
402  */
403 #define	cipher(iter, inR, inL, outR, outL)	{	\
404 	do_F(iter, inR, outR);				\
405 	outR ^= inL;					\
406 	outL = inR;					\
407 }
408 
409 	/*
410 	 * Apply the 16 ciphering steps
411 	 */
412 	{
413 		uint32_t r0, l0, r1, l1;
414 
415 		l0 = work2.long0;
416 		r0 = work2.long1;
417 		cipher(0, r0, l0, r1, l1);
418 		cipher(1, r1, l1, r0, l0);
419 		cipher(2, r0, l0, r1, l1);
420 		cipher(3, r1, l1, r0, l0);
421 		cipher(4, r0, l0, r1, l1);
422 		cipher(5, r1, l1, r0, l0);
423 		cipher(6, r0, l0, r1, l1);
424 		cipher(7, r1, l1, r0, l0);
425 		cipher(8, r0, l0, r1, l1);
426 		cipher(9, r1, l1, r0, l0);
427 		cipher(10, r0, l0, r1, l1);
428 		cipher(11, r1, l1, r0, l0);
429 		cipher(12, r0, l0, r1, l1);
430 		cipher(13, r1, l1, r0, l0);
431 		cipher(14, r0, l0, r1, l1);
432 		cipher(15, r1, l1, r0, l0);
433 		work1.long0 = r0;
434 		work1.long1 = l0;
435 	}
436 
437 	/*
438 	 * Final permutation
439 	 * and chunk to byte conversion
440 	 */
441 	{
442 		uint32_t *lp;
443 		uint32_t l0, l1, w;
444 		short i, pbit;
445 
446 		l0 = l1 = 0;
447 		w = work1.long0;
448 		for (lp = (uint32_t *)&longtab[0], i = 0; i < 32; i++) {
449 			if (w & *lp++) {
450 				pbit = FPtab[i];
451 				if (pbit < 32)
452 					l0 |= longtab[pbit];
453 				else
454 					l1 |= longtab[pbit-32];
455 			}
456 		}
457 		w = work1.long1;
458 		for (lp = (uint32_t *)&longtab[0], i = 32; i < 64; i++) {
459 			if (w & *lp++) {
460 				pbit = FPtab[i];
461 				if (pbit < 32)
462 					l0 |= longtab[pbit];
463 				else
464 					l1 |= longtab[pbit-32];
465 			}
466 		}
467 		work2.long0 = l0;
468 		work2.long1 = l1;
469 	}
470 	data[0] = work2.byte0;
471 	data[1] = work2.byte1;
472 	data[2] = work2.byte2;
473 	data[3] = work2.byte3;
474 	data[4] = work2.byte4;
475 	data[5] = work2.byte5;
476 	data[6] = work2.byte6;
477 	data[7] = work2.byte7;
478 
479 /* EXPORT DELETE END */
480 	return (1);
481 }
482