xref: /freebsd/contrib/bearssl/src/symcipher/des_ct.c (revision 911f0260390e18cf85f3dbf2c719b593efdc1e3c)
1 /*
2  * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org>
3  *
4  * Permission is hereby granted, free of charge, to any person obtaining
5  * a copy of this software and associated documentation files (the
6  * "Software"), to deal in the Software without restriction, including
7  * without limitation the rights to use, copy, modify, merge, publish,
8  * distribute, sublicense, and/or sell copies of the Software, and to
9  * permit persons to whom the Software is furnished to do so, subject to
10  * the following conditions:
11  *
12  * The above copyright notice and this permission notice shall be
13  * included in all copies or substantial portions of the Software.
14  *
15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
16  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
17  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
18  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
19  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
20  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
21  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22  * SOFTWARE.
23  */
24 
25 #include "inner.h"
26 
27 /*
28  * During key schedule, we need to apply bit extraction PC-2 then permute
29  * things into our bitslice representation. PC-2 extracts 48 bits out
30  * of two 28-bit words (kl and kr), and we store these bits into two
31  * 32-bit words sk0 and sk1.
32  *
33  *  -- bit 16+x of sk0 comes from bit QL0[x] of kl
34  *  -- bit x of sk0 comes from bit QR0[x] of kr
35  *  -- bit 16+x of sk1 comes from bit QL1[x] of kl
36  *  -- bit x of sk1 comes from bit QR1[x] of kr
37  */
38 
39 static const unsigned char QL0[] = {
40 	17,  4, 27, 23, 13, 22,  7, 18,
41 	16, 24,  2, 20,  1,  8, 15, 26
42 };
43 
44 static const unsigned char QR0[] = {
45 	25, 19,  9,  1,  5, 11, 23,  8,
46 	17,  0, 22,  3,  6, 20, 27, 24
47 };
48 
49 static const unsigned char QL1[] = {
50 	28, 28, 14, 11, 28, 28, 25,  0,
51 	28, 28,  5,  9, 28, 28, 12, 21
52 };
53 
54 static const unsigned char QR1[] = {
55 	28, 28, 15,  4, 28, 28, 26, 16,
56 	28, 28, 12,  7, 28, 28, 10, 14
57 };
58 
59 /*
60  * 32-bit rotation. The C compiler is supposed to recognize it as a
61  * rotation and use the local architecture rotation opcode (if available).
62  */
63 static inline uint32_t
64 rotl(uint32_t x, int n)
65 {
66 	return (x << n) | (x >> (32 - n));
67 }
68 
69 /*
70  * Compute key schedule for 8 key bytes (produces 32 subkey words).
71  */
72 static void
73 keysched_unit(uint32_t *skey, const void *key)
74 {
75 	int i;
76 
77 	br_des_keysched_unit(skey, key);
78 
79 	/*
80 	 * Apply PC-2 + bitslicing.
81 	 */
82 	for (i = 0; i < 16; i ++) {
83 		uint32_t kl, kr, sk0, sk1;
84 		int j;
85 
86 		kl = skey[(i << 1) + 0];
87 		kr = skey[(i << 1) + 1];
88 		sk0 = 0;
89 		sk1 = 0;
90 		for (j = 0; j < 16; j ++) {
91 			sk0 <<= 1;
92 			sk1 <<= 1;
93 			sk0 |= ((kl >> QL0[j]) & (uint32_t)1) << 16;
94 			sk0 |= (kr >> QR0[j]) & (uint32_t)1;
95 			sk1 |= ((kl >> QL1[j]) & (uint32_t)1) << 16;
96 			sk1 |= (kr >> QR1[j]) & (uint32_t)1;
97 		}
98 
99 		skey[(i << 1) + 0] = sk0;
100 		skey[(i << 1) + 1] = sk1;
101 	}
102 
103 #if 0
104 		/*
105 		 * Speed-optimized version for PC-2 + bitslicing.
106 		 * (Unused. Kept for reference only.)
107 		 */
108 		sk0 = kl & (uint32_t)0x00100000;
109 		sk0 |= (kl & (uint32_t)0x08008000) << 2;
110 		sk0 |= (kl & (uint32_t)0x00400000) << 4;
111 		sk0 |= (kl & (uint32_t)0x00800000) << 5;
112 		sk0 |= (kl & (uint32_t)0x00040000) << 6;
113 		sk0 |= (kl & (uint32_t)0x00010000) << 7;
114 		sk0 |= (kl & (uint32_t)0x00000100) << 10;
115 		sk0 |= (kl & (uint32_t)0x00022000) << 14;
116 		sk0 |= (kl & (uint32_t)0x00000082) << 18;
117 		sk0 |= (kl & (uint32_t)0x00000004) << 19;
118 		sk0 |= (kl & (uint32_t)0x04000000) >> 10;
119 		sk0 |= (kl & (uint32_t)0x00000010) << 26;
120 		sk0 |= (kl & (uint32_t)0x01000000) >> 2;
121 
122 		sk0 |= kr & (uint32_t)0x00000100;
123 		sk0 |= (kr & (uint32_t)0x00000008) << 1;
124 		sk0 |= (kr & (uint32_t)0x00000200) << 4;
125 		sk0 |= rotl(kr & (uint32_t)0x08000021, 6);
126 		sk0 |= (kr & (uint32_t)0x01000000) >> 24;
127 		sk0 |= (kr & (uint32_t)0x00000002) << 11;
128 		sk0 |= (kr & (uint32_t)0x00100000) >> 18;
129 		sk0 |= (kr & (uint32_t)0x00400000) >> 17;
130 		sk0 |= (kr & (uint32_t)0x00800000) >> 14;
131 		sk0 |= (kr & (uint32_t)0x02020000) >> 10;
132 		sk0 |= (kr & (uint32_t)0x00080000) >> 5;
133 		sk0 |= (kr & (uint32_t)0x00000040) >> 3;
134 		sk0 |= (kr & (uint32_t)0x00000800) >> 1;
135 
136 		sk1 = kl & (uint32_t)0x02000000;
137 		sk1 |= (kl & (uint32_t)0x00001000) << 5;
138 		sk1 |= (kl & (uint32_t)0x00000200) << 11;
139 		sk1 |= (kl & (uint32_t)0x00004000) << 15;
140 		sk1 |= (kl & (uint32_t)0x00000020) << 16;
141 		sk1 |= (kl & (uint32_t)0x00000800) << 17;
142 		sk1 |= (kl & (uint32_t)0x00000001) << 24;
143 		sk1 |= (kl & (uint32_t)0x00200000) >> 5;
144 
145 		sk1 |= (kr & (uint32_t)0x00000010) << 8;
146 		sk1 |= (kr & (uint32_t)0x04000000) >> 17;
147 		sk1 |= (kr & (uint32_t)0x00004000) >> 14;
148 		sk1 |= (kr & (uint32_t)0x00000400) >> 9;
149 		sk1 |= (kr & (uint32_t)0x00010000) >> 8;
150 		sk1 |= (kr & (uint32_t)0x00001000) >> 7;
151 		sk1 |= (kr & (uint32_t)0x00000080) >> 3;
152 		sk1 |= (kr & (uint32_t)0x00008000) >> 2;
153 #endif
154 }
155 
156 /* see inner.h */
157 unsigned
158 br_des_ct_keysched(uint32_t *skey, const void *key, size_t key_len)
159 {
160 	switch (key_len) {
161 	case 8:
162 		keysched_unit(skey, key);
163 		return 1;
164 	case 16:
165 		keysched_unit(skey, key);
166 		keysched_unit(skey + 32, (const unsigned char *)key + 8);
167 		br_des_rev_skey(skey + 32);
168 		memcpy(skey + 64, skey, 32 * sizeof *skey);
169 		return 3;
170 	default:
171 		keysched_unit(skey, key);
172 		keysched_unit(skey + 32, (const unsigned char *)key + 8);
173 		br_des_rev_skey(skey + 32);
174 		keysched_unit(skey + 64, (const unsigned char *)key + 16);
175 		return 3;
176 	}
177 }
178 
179 /*
180  * DES confusion function. This function performs expansion E (32 to
181  * 48 bits), XOR with subkey, S-boxes, and permutation P.
182  */
183 static inline uint32_t
184 Fconf(uint32_t r0, const uint32_t *sk)
185 {
186 	/*
187 	 * Each 6->4 S-box is virtually turned into four 6->1 boxes; we
188 	 * thus end up with 32 boxes that we call "T-boxes" here. We will
189 	 * evaluate them with bitslice code.
190 	 *
191 	 * Each T-box is a circuit of multiplexers (sort of) and thus
192 	 * takes 70 inputs: the 6 actual T-box inputs, and 64 constants
193 	 * that describe the T-box output for all combinations of the
194 	 * 6 inputs. With this model, all T-boxes are identical (with
195 	 * distinct inputs) and thus can be executed in parallel with
196 	 * bitslice code.
197 	 *
198 	 * T-boxes are numbered from 0 to 31, in least-to-most
199 	 * significant order. Thus, S-box S1 corresponds to T-boxes 31,
200 	 * 30, 29 and 28, in that order. T-box 'n' is computed with the
201 	 * bits at rank 'n' in the 32-bit words.
202 	 *
203 	 * Words x0 to x5 contain the T-box inputs 0 to 5.
204 	 */
205 	uint32_t x0, x1, x2, x3, x4, x5, z0;
206 	uint32_t y0, y1, y2, y3, y4, y5, y6, y7, y8, y9;
207 	uint32_t y10, y11, y12, y13, y14, y15, y16, y17, y18, y19;
208 	uint32_t y20, y21, y22, y23, y24, y25, y26, y27, y28, y29;
209 	uint32_t y30;
210 
211 	/*
212 	 * Spread input bits over the 6 input words x*.
213 	 */
214 	x1 = r0 & (uint32_t)0x11111111;
215 	x2 = (r0 >> 1) & (uint32_t)0x11111111;
216 	x3 = (r0 >> 2) & (uint32_t)0x11111111;
217 	x4 = (r0 >> 3) & (uint32_t)0x11111111;
218 	x1 = (x1 << 4) - x1;
219 	x2 = (x2 << 4) - x2;
220 	x3 = (x3 << 4) - x3;
221 	x4 = (x4 << 4) - x4;
222 	x0 = (x4 << 4) | (x4 >> 28);
223 	x5 = (x1 >> 4) | (x1 << 28);
224 
225 	/*
226 	 * XOR with the subkey for this round.
227 	 */
228 	x0 ^= sk[0];
229 	x1 ^= sk[1];
230 	x2 ^= sk[2];
231 	x3 ^= sk[3];
232 	x4 ^= sk[4];
233 	x5 ^= sk[5];
234 
235 	/*
236 	 * The T-boxes are done in parallel, since they all use a
237 	 * "tree of multiplexer". We use "fake multiplexers":
238 	 *
239 	 *   y = a ^ (x & b)
240 	 *
241 	 * computes y as either 'a' (if x == 0) or 'a ^ b' (if x == 1).
242 	 */
243 	y0 = (uint32_t)0xEFA72C4D ^ (x0 & (uint32_t)0xEC7AC69C);
244 	y1 = (uint32_t)0xAEAAEDFF ^ (x0 & (uint32_t)0x500FB821);
245 	y2 = (uint32_t)0x37396665 ^ (x0 & (uint32_t)0x40EFA809);
246 	y3 = (uint32_t)0x68D7B833 ^ (x0 & (uint32_t)0xA5EC0B28);
247 	y4 = (uint32_t)0xC9C755BB ^ (x0 & (uint32_t)0x252CF820);
248 	y5 = (uint32_t)0x73FC3606 ^ (x0 & (uint32_t)0x40205801);
249 	y6 = (uint32_t)0xA2A0A918 ^ (x0 & (uint32_t)0xE220F929);
250 	y7 = (uint32_t)0x8222BD90 ^ (x0 & (uint32_t)0x44A3F9E1);
251 	y8 = (uint32_t)0xD6B6AC77 ^ (x0 & (uint32_t)0x794F104A);
252 	y9 = (uint32_t)0x3069300C ^ (x0 & (uint32_t)0x026F320B);
253 	y10 = (uint32_t)0x6CE0D5CC ^ (x0 & (uint32_t)0x7640B01A);
254 	y11 = (uint32_t)0x59A9A22D ^ (x0 & (uint32_t)0x238F1572);
255 	y12 = (uint32_t)0xAC6D0BD4 ^ (x0 & (uint32_t)0x7A63C083);
256 	y13 = (uint32_t)0x21C83200 ^ (x0 & (uint32_t)0x11CCA000);
257 	y14 = (uint32_t)0xA0E62188 ^ (x0 & (uint32_t)0x202F69AA);
258 	/* y15 = (uint32_t)0x00000000 ^ (x0 & (uint32_t)0x00000000); */
259 	y16 = (uint32_t)0xAF7D655A ^ (x0 & (uint32_t)0x51B33BE9);
260 	y17 = (uint32_t)0xF0168AA3 ^ (x0 & (uint32_t)0x3B0FE8AE);
261 	y18 = (uint32_t)0x90AA30C6 ^ (x0 & (uint32_t)0x90BF8816);
262 	y19 = (uint32_t)0x5AB2750A ^ (x0 & (uint32_t)0x09E34F9B);
263 	y20 = (uint32_t)0x5391BE65 ^ (x0 & (uint32_t)0x0103BE88);
264 	y21 = (uint32_t)0x93372BAF ^ (x0 & (uint32_t)0x49AC8E25);
265 	y22 = (uint32_t)0xF288210C ^ (x0 & (uint32_t)0x922C313D);
266 	y23 = (uint32_t)0x920AF5C0 ^ (x0 & (uint32_t)0x70EF31B0);
267 	y24 = (uint32_t)0x63D312C0 ^ (x0 & (uint32_t)0x6A707100);
268 	y25 = (uint32_t)0x537B3006 ^ (x0 & (uint32_t)0xB97C9011);
269 	y26 = (uint32_t)0xA2EFB0A5 ^ (x0 & (uint32_t)0xA320C959);
270 	y27 = (uint32_t)0xBC8F96A5 ^ (x0 & (uint32_t)0x6EA0AB4A);
271 	y28 = (uint32_t)0xFAD176A5 ^ (x0 & (uint32_t)0x6953DDF8);
272 	y29 = (uint32_t)0x665A14A3 ^ (x0 & (uint32_t)0xF74F3E2B);
273 	y30 = (uint32_t)0xF2EFF0CC ^ (x0 & (uint32_t)0xF0306CAD);
274 	/* y31 = (uint32_t)0x00000000 ^ (x0 & (uint32_t)0x00000000); */
275 
276 	y0 = y0 ^ (x1 & y1);
277 	y1 = y2 ^ (x1 & y3);
278 	y2 = y4 ^ (x1 & y5);
279 	y3 = y6 ^ (x1 & y7);
280 	y4 = y8 ^ (x1 & y9);
281 	y5 = y10 ^ (x1 & y11);
282 	y6 = y12 ^ (x1 & y13);
283 	y7 = y14; /* was: y14 ^ (x1 & y15) */
284 	y8 = y16 ^ (x1 & y17);
285 	y9 = y18 ^ (x1 & y19);
286 	y10 = y20 ^ (x1 & y21);
287 	y11 = y22 ^ (x1 & y23);
288 	y12 = y24 ^ (x1 & y25);
289 	y13 = y26 ^ (x1 & y27);
290 	y14 = y28 ^ (x1 & y29);
291 	y15 = y30; /* was: y30 ^ (x1 & y31) */
292 
293 	y0 = y0 ^ (x2 & y1);
294 	y1 = y2 ^ (x2 & y3);
295 	y2 = y4 ^ (x2 & y5);
296 	y3 = y6 ^ (x2 & y7);
297 	y4 = y8 ^ (x2 & y9);
298 	y5 = y10 ^ (x2 & y11);
299 	y6 = y12 ^ (x2 & y13);
300 	y7 = y14 ^ (x2 & y15);
301 
302 	y0 = y0 ^ (x3 & y1);
303 	y1 = y2 ^ (x3 & y3);
304 	y2 = y4 ^ (x3 & y5);
305 	y3 = y6 ^ (x3 & y7);
306 
307 	y0 = y0 ^ (x4 & y1);
308 	y1 = y2 ^ (x4 & y3);
309 
310 	y0 = y0 ^ (x5 & y1);
311 
312 	/*
313 	 * The P permutation:
314 	 * -- Each bit move is converted into a mask + left rotation.
315 	 * -- Rotations that use the same movement are coalesced together.
316 	 * -- Left and right shifts are used as alternatives to a rotation
317 	 * where appropriate (this will help architectures that do not have
318 	 * a rotation opcode).
319 	 */
320 	z0 = (y0 & (uint32_t)0x00000004) << 3;
321 	z0 |= (y0 & (uint32_t)0x00004000) << 4;
322 	z0 |= rotl(y0 & 0x12020120, 5);
323 	z0 |= (y0 & (uint32_t)0x00100000) << 6;
324 	z0 |= (y0 & (uint32_t)0x00008000) << 9;
325 	z0 |= (y0 & (uint32_t)0x04000000) >> 22;
326 	z0 |= (y0 & (uint32_t)0x00000001) << 11;
327 	z0 |= rotl(y0 & 0x20000200, 12);
328 	z0 |= (y0 & (uint32_t)0x00200000) >> 19;
329 	z0 |= (y0 & (uint32_t)0x00000040) << 14;
330 	z0 |= (y0 & (uint32_t)0x00010000) << 15;
331 	z0 |= (y0 & (uint32_t)0x00000002) << 16;
332 	z0 |= rotl(y0 & 0x40801800, 17);
333 	z0 |= (y0 & (uint32_t)0x00080000) >> 13;
334 	z0 |= (y0 & (uint32_t)0x00000010) << 21;
335 	z0 |= (y0 & (uint32_t)0x01000000) >> 10;
336 	z0 |= rotl(y0 & 0x88000008, 24);
337 	z0 |= (y0 & (uint32_t)0x00000480) >> 7;
338 	z0 |= (y0 & (uint32_t)0x00442000) >> 6;
339 	return z0;
340 }
341 
342 /*
343  * Process one block through 16 successive rounds, omitting the swap
344  * in the final round.
345  */
346 static void
347 process_block_unit(uint32_t *pl, uint32_t *pr, const uint32_t *sk_exp)
348 {
349 	int i;
350 	uint32_t l, r;
351 
352 	l = *pl;
353 	r = *pr;
354 	for (i = 0; i < 16; i ++) {
355 		uint32_t t;
356 
357 		t = l ^ Fconf(r, sk_exp);
358 		l = r;
359 		r = t;
360 		sk_exp += 6;
361 	}
362 	*pl = r;
363 	*pr = l;
364 }
365 
366 /* see inner.h */
367 void
368 br_des_ct_process_block(unsigned num_rounds,
369 	const uint32_t *sk_exp, void *block)
370 {
371 	unsigned char *buf;
372 	uint32_t l, r;
373 
374 	buf = block;
375 	l = br_dec32be(buf);
376 	r = br_dec32be(buf + 4);
377 	br_des_do_IP(&l, &r);
378 	while (num_rounds -- > 0) {
379 		process_block_unit(&l, &r, sk_exp);
380 		sk_exp += 96;
381 	}
382 	br_des_do_invIP(&l, &r);
383 	br_enc32be(buf, l);
384 	br_enc32be(buf + 4, r);
385 }
386 
387 /* see inner.h */
388 void
389 br_des_ct_skey_expand(uint32_t *sk_exp,
390 	unsigned num_rounds, const uint32_t *skey)
391 {
392 	num_rounds <<= 4;
393 	while (num_rounds -- > 0) {
394 		uint32_t v, w0, w1, w2, w3;
395 
396 		v = *skey ++;
397 		w0 = v & 0x11111111;
398 		w1 = (v >> 1) & 0x11111111;
399 		w2 = (v >> 2) & 0x11111111;
400 		w3 = (v >> 3) & 0x11111111;
401 		*sk_exp ++ = (w0 << 4) - w0;
402 		*sk_exp ++ = (w1 << 4) - w1;
403 		*sk_exp ++ = (w2 << 4) - w2;
404 		*sk_exp ++ = (w3 << 4) - w3;
405 		v = *skey ++;
406 		w0 = v & 0x11111111;
407 		w1 = (v >> 1) & 0x11111111;
408 		*sk_exp ++ = (w0 << 4) - w0;
409 		*sk_exp ++ = (w1 << 4) - w1;
410 	}
411 }
412