xref: /titanic_50/usr/src/lib/libnsl/des/des_soft.c (revision 84ba300aaa958c8e8427c2ec66a932d86bee71c4)
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 /*
37  * Warning!  Things are arranged very carefully in this file to
38  * allow read-only data to be moved to the text segment.  The
39  * various DES tables must appear before any function definitions
40  * (this is arranged by including them immediately below) and partab
41  * must also appear before and function definitions
42  * This arrangement allows all data up through the first text to
43  * be moved to text.
44  */
45 
46 #include "mt.h"
47 #include <sys/types.h>
48 #include <des/softdes.h>
49 #include <des/desdata.h>
50 #ifdef sun
51 #include <sys/ioctl.h>
52 #include <sys/des.h>
53 #else
54 #include <des/des.h>
55 #endif
56 #include <rpcsvc/nis_dhext.h>
57 
58 /*
59  * Fast (?) software implementation of DES
60  * Has been seen going at 2000 bytes/sec on a Sun-2
61  * Works on a VAX too.
62  * Won't work without 8 bit chars and 32 bit longs
63  */
64 
65 #define	btst(k, b)	(k[b >> 3] & (0x80 >> (b & 07)))
66 #define	BIT28	(1<<28)
67 
68 static int	__des_encrypt(uchar_t *, struct deskeydata *);
69 static int	__des_setkey(uchar_t[8], struct deskeydata *, unsigned);
70 
71 
72 /*
73  * Table giving odd parity in the low bit for ASCII characters
74  */
75 const char partab[128] = {
76 	0x01, 0x01, 0x02, 0x02, 0x04, 0x04, 0x07, 0x07,
77 	0x08, 0x08, 0x0b, 0x0b, 0x0d, 0x0d, 0x0e, 0x0e,
78 	0x10, 0x10, 0x13, 0x13, 0x15, 0x15, 0x16, 0x16,
79 	0x19, 0x19, 0x1a, 0x1a, 0x1c, 0x1c, 0x1f, 0x1f,
80 	0x20, 0x20, 0x23, 0x23, 0x25, 0x25, 0x26, 0x26,
81 	0x29, 0x29, 0x2a, 0x2a, 0x2c, 0x2c, 0x2f, 0x2f,
82 	0x31, 0x31, 0x32, 0x32, 0x34, 0x34, 0x37, 0x37,
83 	0x38, 0x38, 0x3b, 0x3b, 0x3d, 0x3d, 0x3e, 0x3e,
84 	0x40, 0x40, 0x43, 0x43, 0x45, 0x45, 0x46, 0x46,
85 	0x49, 0x49, 0x4a, 0x4a, 0x4c, 0x4c, 0x4f, 0x4f,
86 	0x51, 0x51, 0x52, 0x52, 0x54, 0x54, 0x57, 0x57,
87 	0x58, 0x58, 0x5b, 0x5b, 0x5d, 0x5d, 0x5e, 0x5e,
88 	0x61, 0x61, 0x62, 0x62, 0x64, 0x64, 0x67, 0x67,
89 	0x68, 0x68, 0x6b, 0x6b, 0x6d, 0x6d, 0x6e, 0x6e,
90 	0x70, 0x70, 0x73, 0x73, 0x75, 0x75, 0x76, 0x76,
91 	0x79, 0x79, 0x7a, 0x7a, 0x7c, 0x7c, 0x7f, 0x7f,
92 };
93 
94 /*
95  * Add odd parity to low bit of 8 byte key
96  */
97 void
98 des_setparity(char *p)
99 {
100 	int i;
101 
102 	for (i = 0; i < 8; i++) {
103 		*p = partab[*p & 0x7f];
104 		p++;
105 	}
106 }
107 
108 static const unsigned char partab_g[256] = {
109 	0x01, 0x01, 0x02, 0x02, 0x04, 0x04, 0x07, 0x07,
110 	0x08, 0x08, 0x0b, 0x0b, 0x0d, 0x0d, 0x0e, 0x0e,
111 	0x10, 0x10, 0x13, 0x13, 0x15, 0x15, 0x16, 0x16,
112 	0x19, 0x19, 0x1a, 0x1a, 0x1c, 0x1c, 0x1f, 0x1f,
113 	0x20, 0x20, 0x23, 0x23, 0x25, 0x25, 0x26, 0x26,
114 	0x29, 0x29, 0x2a, 0x2a, 0x2c, 0x2c, 0x2f, 0x2f,
115 	0x31, 0x31, 0x32, 0x32, 0x34, 0x34, 0x37, 0x37,
116 	0x38, 0x38, 0x3b, 0x3b, 0x3d, 0x3d, 0x3e, 0x3e,
117 	0x40, 0x40, 0x43, 0x43, 0x45, 0x45, 0x46, 0x46,
118 	0x49, 0x49, 0x4a, 0x4a, 0x4c, 0x4c, 0x4f, 0x4f,
119 	0x51, 0x51, 0x52, 0x52, 0x54, 0x54, 0x57, 0x57,
120 	0x58, 0x58, 0x5b, 0x5b, 0x5d, 0x5d, 0x5e, 0x5e,
121 	0x61, 0x61, 0x62, 0x62, 0x64, 0x64, 0x67, 0x67,
122 	0x68, 0x68, 0x6b, 0x6b, 0x6d, 0x6d, 0x6e, 0x6e,
123 	0x70, 0x70, 0x73, 0x73, 0x75, 0x75, 0x76, 0x76,
124 	0x79, 0x79, 0x7a, 0x7a, 0x7c, 0x7c, 0x7f, 0x7f,
125 	0x80, 0x80, 0x83, 0x83, 0x85, 0x85, 0x86, 0x86,
126 	0x89, 0x89, 0x8a, 0x8a, 0x8c, 0x8c, 0x8f, 0x8f,
127 	0x91, 0x91, 0x92, 0x92, 0x94, 0x94, 0x97, 0x97,
128 	0x98, 0x98, 0x9b, 0x9b, 0x9d, 0x9d, 0x9e, 0x9e,
129 	0xa1, 0xa1, 0xa2, 0xa2, 0xa4, 0xa4, 0xa7, 0xa7,
130 	0xa8, 0xa8, 0xab, 0xab, 0xad, 0xad, 0xae, 0xae,
131 	0xb0, 0xb0, 0xb3, 0xb3, 0xb5, 0xb5, 0xb6, 0xb6,
132 	0xb9, 0xb9, 0xba, 0xba, 0xbc, 0xbc, 0xbf, 0xbf,
133 	0xc1, 0xc1, 0xc2, 0xc2, 0xc4, 0xc4, 0xc7, 0xc7,
134 	0xc8, 0xc8, 0xcb, 0xcb, 0xcd, 0xcd, 0xce, 0xce,
135 	0xd0, 0xd0, 0xd3, 0xd3, 0xd5, 0xd5, 0xd6, 0xd6,
136 	0xd9, 0xd9, 0xda, 0xda, 0xdc, 0xdc, 0xdf, 0xdf,
137 	0xe0, 0xe0, 0xe3, 0xe3, 0xe5, 0xe5, 0xe6, 0xe6,
138 	0xe9, 0xe9, 0xea, 0xea, 0xec, 0xec, 0xef, 0xef,
139 	0xf1, 0xf1, 0xf2, 0xf2, 0xf4, 0xf4, 0xf7, 0xf7,
140 	0xf8, 0xf8, 0xfb, 0xfb, 0xfd, 0xfd, 0xfe, 0xfe
141 };
142 
143 /*
144  * A corrected version of des_setparity (see bug 1149767).
145  */
146 void
147 des_setparity_g(des_block *p)
148 {
149 	int i;
150 
151 	for (i = 0; i < 8; i++) {
152 		(*p).c[i] = partab_g[(*p).c[i]];
153 	}
154 }
155 
156 /*
157  * Software encrypt or decrypt a block of data (multiple of 8 bytes)
158  * Do the CBC ourselves if needed.
159  */
160 int
161 __des_crypt(char *buf, unsigned len, struct desparams *desp)
162 {
163 	short i;
164 	unsigned mode;
165 	unsigned dir;
166 	char nextiv[8];
167 	struct deskeydata softkey;
168 
169 	mode = (unsigned)desp->des_mode;
170 	dir = (unsigned)desp->des_dir;
171 	(void) __des_setkey(desp->des_key, &softkey, dir);
172 	while (len != 0) {
173 		switch (mode) {
174 		case CBC:
175 			switch (dir) {
176 			case ENCRYPT:
177 				for (i = 0; i < 8; i++)
178 					buf[i] ^= desp->des_ivec[i];
179 				(void) __des_encrypt((uchar_t *)buf, &softkey);
180 				for (i = 0; i < 8; i++)
181 					desp->des_ivec[i] = buf[i];
182 				break;
183 			case DECRYPT:
184 				for (i = 0; i < 8; i++)
185 					nextiv[i] = buf[i];
186 				(void) __des_encrypt((uchar_t *)buf, &softkey);
187 				for (i = 0; i < 8; i++) {
188 					buf[i] ^= desp->des_ivec[i];
189 					desp->des_ivec[i] = nextiv[i];
190 				}
191 				break;
192 			}
193 			break;
194 		case ECB:
195 			(void) __des_encrypt((uchar_t *)buf, &softkey);
196 			break;
197 		}
198 		buf += 8;
199 		len -= 8;
200 	}
201 	return (1);
202 }
203 
204 
205 /*
206  * Set the key and direction for an encryption operation
207  * We build the 16 key entries here
208  */
209 static int
210 __des_setkey(uchar_t userkey[8], struct deskeydata *kd, unsigned dir)
211 {
212 	int32_t C, D;
213 	short i;
214 
215 	/*
216 	 * First, generate C and D by permuting
217 	 * the key. The low order bit of each
218 	 * 8-bit char is not used, so C and D are only 28
219 	 * bits apiece.
220 	 */
221 	{
222 		short bit;
223 		const short *pcc = PC1_C, *pcd = PC1_D;
224 
225 		C = D = 0;
226 		for (i = 0; i < 28; i++) {
227 			C <<= 1;
228 			D <<= 1;
229 			bit = *pcc++;
230 			if (btst(userkey, bit))
231 				C |= 1;
232 			bit = *pcd++;
233 			if (btst(userkey, bit))
234 				D |= 1;
235 		}
236 	}
237 	/*
238 	 * To generate Ki, rotate C and D according
239 	 * to schedule and pick up a permutation
240 	 * using PC2.
241 	 */
242 	for (i = 0; i < 16; i++) {
243 		chunk_t *c;
244 		short j, k, bit;
245 		uint32_t bbit;
246 
247 		/*
248 		 * Do the "left shift" (rotate)
249 		 * We know we always rotate by either 1 or 2 bits
250 		 * the shifts table tells us if its 2
251 		 */
252 		C <<= 1;
253 		if (C & BIT28)
254 			C |= 1;
255 		D <<= 1;
256 		if (D & BIT28)
257 			D |= 1;
258 		if (shifts[i]) {
259 			C <<= 1;
260 			if (C & BIT28)
261 				C |= 1;
262 			D <<= 1;
263 			if (D & BIT28)
264 				D |= 1;
265 		}
266 		/*
267 		 * get Ki. Note C and D are concatenated.
268 		 */
269 		bit = 0;
270 		switch (dir) {
271 		case ENCRYPT:
272 			c = &kd->keyval[i]; break;
273 		case DECRYPT:
274 			c = &kd->keyval[15 - i]; break;
275 		}
276 		c->long0 = 0;
277 		c->long1 = 0;
278 		bbit = (1 << 5) << 24;
279 		for (j = 0; j < 4; j++) {
280 			for (k = 0; k < 6; k++) {
281 				if (C & (BIT28 >> PC2_C[bit]))
282 					c->long0 |= bbit >> k;
283 				if (D & (BIT28 >> PC2_D[bit]))
284 					c->long1 |= bbit >> k;
285 				bit++;
286 			}
287 			bbit >>= 8;
288 		}
289 
290 	}
291 	return (1);
292 }
293 
294 
295 
296 /*
297  * Do an encryption operation
298  * Much pain is taken (with preprocessor) to avoid loops so the compiler
299  * can do address arithmetic instead of doing it at runtime.
300  * Note that the byte-to-chunk conversion is necessary to guarantee
301  * processor byte-order independence.
302  */
303 static int
304 __des_encrypt(uchar_t *data, struct deskeydata *kd)
305 {
306 	chunk_t work1, work2;
307 
308 	/*
309 	 * Initial permutation
310 	 * and byte to chunk conversion
311 	 */
312 	{
313 		const uint32_t *lp;
314 		uint32_t l0, l1, w;
315 		short i, pbit;
316 
317 		work1.byte0 = data[0];
318 		work1.byte1 = data[1];
319 		work1.byte2 = data[2];
320 		work1.byte3 = data[3];
321 		work1.byte4 = data[4];
322 		work1.byte5 = data[5];
323 		work1.byte6 = data[6];
324 		work1.byte7 = data[7];
325 		l0 = l1 = 0;
326 		w = work1.long0;
327 		for (lp = (uint32_t *)&longtab[0], i = 0; i < 32; i++) {
328 			if (w & *lp++) {
329 				pbit = IPtab[i];
330 				if (pbit < 32)
331 					l0 |= longtab[pbit];
332 				else
333 					l1 |= longtab[pbit-32];
334 			}
335 		}
336 		w = work1.long1;
337 		for (lp = (uint32_t *)&longtab[0], i = 32; i < 64; i++) {
338 			if (w & *lp++) {
339 				pbit = IPtab[i];
340 				if (pbit < 32)
341 					l0 |= longtab[pbit];
342 				else
343 					l1 |= longtab[pbit-32];
344 			}
345 		}
346 		work2.long0 = l0;
347 		work2.long1 = l1;
348 	}
349 
350 /*
351  * Expand 8 bits of 32 bit R to 48 bit R
352  */
353 #define	do_R_to_ER(op, b)	{			\
354 	const struct R_to_ER *p = &R_to_ER_tab[b][R.byte##b];	\
355 	e0 op p->l0;				\
356 	e1 op p->l1;				\
357 }
358 
359 /*
360  * Inner part of the algorithm:
361  * Expand R from 32 to 48 bits; xor key value;
362  * apply S boxes; permute 32 bits of output
363  */
364 /* BEGIN CSTYLED */
365 #define	do_F(iter, inR, outR) 	{			\
366 	chunk_t R, ER;					\
367 	uint32_t e0, e1;				\
368 	R.long0 = inR;					\
369 	do_R_to_ER(=, 0);				\
370 	do_R_to_ER(|=, 1);				\
371 	do_R_to_ER(|=, 2);				\
372 	do_R_to_ER(|=, 3);				\
373 	ER.long0 = e0 ^ kd->keyval[iter].long0;		\
374 	ER.long1 = e1 ^ kd->keyval[iter].long1;		\
375 	R.long0 =					\
376 		S_tab[0][ER.byte0] +			\
377 		S_tab[1][ER.byte1] +			\
378 		S_tab[2][ER.byte2] +			\
379 		S_tab[3][ER.byte3] +			\
380 		S_tab[4][ER.byte4] +			\
381 		S_tab[5][ER.byte5] +			\
382 		S_tab[6][ER.byte6] +			\
383 		S_tab[7][ER.byte7]; 			\
384 	outR =						\
385 		P_tab[0][R.byte0] +			\
386 		P_tab[1][R.byte1] +			\
387 		P_tab[2][R.byte2] +			\
388 		P_tab[3][R.byte3]; 			\
389 }
390 /* END CSTYLED */
391 
392 /*
393  * Do a cipher step
394  * Apply inner part; do xor and exchange of 32 bit parts
395  */
396 #define	cipher(iter, inR, inL, outR, outL)	{	\
397 	do_F(iter, inR, outR);				\
398 	outR ^= inL;					\
399 	outL = inR;					\
400 }
401 
402 	/*
403 	 * Apply the 16 ciphering steps
404 	 */
405 	{
406 		uint32_t r0, l0, r1, l1;
407 
408 		l0 = work2.long0;
409 		r0 = work2.long1;
410 		cipher(0, r0, l0, r1, l1);
411 		cipher(1, r1, l1, r0, l0);
412 		cipher(2, r0, l0, r1, l1);
413 		cipher(3, r1, l1, r0, l0);
414 		cipher(4, r0, l0, r1, l1);
415 		cipher(5, r1, l1, r0, l0);
416 		cipher(6, r0, l0, r1, l1);
417 		cipher(7, r1, l1, r0, l0);
418 		cipher(8, r0, l0, r1, l1);
419 		cipher(9, r1, l1, r0, l0);
420 		cipher(10, r0, l0, r1, l1);
421 		cipher(11, r1, l1, r0, l0);
422 		cipher(12, r0, l0, r1, l1);
423 		cipher(13, r1, l1, r0, l0);
424 		cipher(14, r0, l0, r1, l1);
425 		cipher(15, r1, l1, r0, l0);
426 		work1.long0 = r0;
427 		work1.long1 = l0;
428 	}
429 
430 	/*
431 	 * Final permutation
432 	 * and chunk to byte conversion
433 	 */
434 	{
435 		uint32_t *lp;
436 		uint32_t l0, l1, w;
437 		short i, pbit;
438 
439 		l0 = l1 = 0;
440 		w = work1.long0;
441 		for (lp = (uint32_t *)&longtab[0], i = 0; i < 32; i++) {
442 			if (w & *lp++) {
443 				pbit = FPtab[i];
444 				if (pbit < 32)
445 					l0 |= longtab[pbit];
446 				else
447 					l1 |= longtab[pbit-32];
448 			}
449 		}
450 		w = work1.long1;
451 		for (lp = (uint32_t *)&longtab[0], i = 32; i < 64; i++) {
452 			if (w & *lp++) {
453 				pbit = FPtab[i];
454 				if (pbit < 32)
455 					l0 |= longtab[pbit];
456 				else
457 					l1 |= longtab[pbit-32];
458 			}
459 		}
460 		work2.long0 = l0;
461 		work2.long1 = l1;
462 	}
463 	data[0] = work2.byte0;
464 	data[1] = work2.byte1;
465 	data[2] = work2.byte2;
466 	data[3] = work2.byte3;
467 	data[4] = work2.byte4;
468 	data[5] = work2.byte5;
469 	data[6] = work2.byte6;
470 	data[7] = work2.byte7;
471 
472 	return (1);
473 }
474