xref: /openbsd-src/lib/libcrypto/bn/bn_mul.c (revision fc405d53b73a2d73393cb97f684863d17b583e38)
1 /* $OpenBSD: bn_mul.c,v 1.37 2023/04/19 10:51:22 jsing Exp $ */
2 /* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
3  * All rights reserved.
4  *
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.
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 (tjh@cryptsoft.com).
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 (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)"
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 #include <assert.h>
60 #include <stdio.h>
61 #include <string.h>
62 
63 #include <openssl/opensslconf.h>
64 
65 #include "bn_arch.h"
66 #include "bn_internal.h"
67 #include "bn_local.h"
68 
69 /*
70  * bn_mul_comba4() computes r[] = a[] * b[] using Comba multiplication
71  * (https://everything2.com/title/Comba+multiplication), where a and b are both
72  * four word arrays, producing an eight word array result.
73  */
74 #ifndef HAVE_BN_MUL_COMBA4
75 void
76 bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
77 {
78 	BN_ULONG c0, c1, c2;
79 
80 	bn_mulw_addtw(a[0], b[0],  0,  0,  0, &c2, &c1, &r[0]);
81 
82 	bn_mulw_addtw(a[0], b[1],  0, c2, c1, &c2, &c1, &c0);
83 	bn_mulw_addtw(a[1], b[0], c2, c1, c0, &c2, &c1, &r[1]);
84 
85 	bn_mulw_addtw(a[2], b[0],  0, c2, c1, &c2, &c1, &c0);
86 	bn_mulw_addtw(a[1], b[1], c2, c1, c0, &c2, &c1, &c0);
87 	bn_mulw_addtw(a[0], b[2], c2, c1, c0, &c2, &c1, &r[2]);
88 
89 	bn_mulw_addtw(a[0], b[3],  0, c2, c1, &c2, &c1, &c0);
90 	bn_mulw_addtw(a[1], b[2], c2, c1, c0, &c2, &c1, &c0);
91 	bn_mulw_addtw(a[2], b[1], c2, c1, c0, &c2, &c1, &c0);
92 	bn_mulw_addtw(a[3], b[0], c2, c1, c0, &c2, &c1, &r[3]);
93 
94 	bn_mulw_addtw(a[3], b[1],  0, c2, c1, &c2, &c1, &c0);
95 	bn_mulw_addtw(a[2], b[2], c2, c1, c0, &c2, &c1, &c0);
96 	bn_mulw_addtw(a[1], b[3], c2, c1, c0, &c2, &c1, &r[4]);
97 
98 	bn_mulw_addtw(a[2], b[3],  0, c2, c1, &c2, &c1, &c0);
99 	bn_mulw_addtw(a[3], b[2], c2, c1, c0, &c2, &c1, &r[5]);
100 
101 	bn_mulw_addtw(a[3], b[3],  0, c2, c1, &c2, &r[7], &r[6]);
102 }
103 #endif
104 
105 /*
106  * bn_mul_comba8() computes r[] = a[] * b[] using Comba multiplication
107  * (https://everything2.com/title/Comba+multiplication), where a and b are both
108  * eight word arrays, producing a 16 word array result.
109  */
110 #ifndef HAVE_BN_MUL_COMBA8
111 void
112 bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
113 {
114 	BN_ULONG c0, c1, c2;
115 
116 	bn_mulw_addtw(a[0], b[0],  0,  0,  0, &c2, &c1, &r[0]);
117 
118 	bn_mulw_addtw(a[0], b[1],  0, c2, c1, &c2, &c1, &c0);
119 	bn_mulw_addtw(a[1], b[0], c2, c1, c0, &c2, &c1, &r[1]);
120 
121 	bn_mulw_addtw(a[2], b[0],  0, c2, c1, &c2, &c1, &c0);
122 	bn_mulw_addtw(a[1], b[1], c2, c1, c0, &c2, &c1, &c0);
123 	bn_mulw_addtw(a[0], b[2], c2, c1, c0, &c2, &c1, &r[2]);
124 
125 	bn_mulw_addtw(a[0], b[3],  0, c2, c1, &c2, &c1, &c0);
126 	bn_mulw_addtw(a[1], b[2], c2, c1, c0, &c2, &c1, &c0);
127 	bn_mulw_addtw(a[2], b[1], c2, c1, c0, &c2, &c1, &c0);
128 	bn_mulw_addtw(a[3], b[0], c2, c1, c0, &c2, &c1, &r[3]);
129 
130 	bn_mulw_addtw(a[4], b[0],  0, c2, c1, &c2, &c1, &c0);
131 	bn_mulw_addtw(a[3], b[1], c2, c1, c0, &c2, &c1, &c0);
132 	bn_mulw_addtw(a[2], b[2], c2, c1, c0, &c2, &c1, &c0);
133 	bn_mulw_addtw(a[1], b[3], c2, c1, c0, &c2, &c1, &c0);
134 	bn_mulw_addtw(a[0], b[4], c2, c1, c0, &c2, &c1, &r[4]);
135 
136 	bn_mulw_addtw(a[0], b[5],  0, c2, c1, &c2, &c1, &c0);
137 	bn_mulw_addtw(a[1], b[4], c2, c1, c0, &c2, &c1, &c0);
138 	bn_mulw_addtw(a[2], b[3], c2, c1, c0, &c2, &c1, &c0);
139 	bn_mulw_addtw(a[3], b[2], c2, c1, c0, &c2, &c1, &c0);
140 	bn_mulw_addtw(a[4], b[1], c2, c1, c0, &c2, &c1, &c0);
141 	bn_mulw_addtw(a[5], b[0], c2, c1, c0, &c2, &c1, &r[5]);
142 
143 	bn_mulw_addtw(a[6], b[0],  0, c2, c1, &c2, &c1, &c0);
144 	bn_mulw_addtw(a[5], b[1], c2, c1, c0, &c2, &c1, &c0);
145 	bn_mulw_addtw(a[4], b[2], c2, c1, c0, &c2, &c1, &c0);
146 	bn_mulw_addtw(a[3], b[3], c2, c1, c0, &c2, &c1, &c0);
147 	bn_mulw_addtw(a[2], b[4], c2, c1, c0, &c2, &c1, &c0);
148 	bn_mulw_addtw(a[1], b[5], c2, c1, c0, &c2, &c1, &c0);
149 	bn_mulw_addtw(a[0], b[6], c2, c1, c0, &c2, &c1, &r[6]);
150 
151 	bn_mulw_addtw(a[0], b[7],  0, c2, c1, &c2, &c1, &c0);
152 	bn_mulw_addtw(a[1], b[6], c2, c1, c0, &c2, &c1, &c0);
153 	bn_mulw_addtw(a[2], b[5], c2, c1, c0, &c2, &c1, &c0);
154 	bn_mulw_addtw(a[3], b[4], c2, c1, c0, &c2, &c1, &c0);
155 	bn_mulw_addtw(a[4], b[3], c2, c1, c0, &c2, &c1, &c0);
156 	bn_mulw_addtw(a[5], b[2], c2, c1, c0, &c2, &c1, &c0);
157 	bn_mulw_addtw(a[6], b[1], c2, c1, c0, &c2, &c1, &c0);
158 	bn_mulw_addtw(a[7], b[0], c2, c1, c0, &c2, &c1, &r[7]);
159 
160 	bn_mulw_addtw(a[7], b[1],  0, c2, c1, &c2, &c1, &c0);
161 	bn_mulw_addtw(a[6], b[2], c2, c1, c0, &c2, &c1, &c0);
162 	bn_mulw_addtw(a[5], b[3], c2, c1, c0, &c2, &c1, &c0);
163 	bn_mulw_addtw(a[4], b[4], c2, c1, c0, &c2, &c1, &c0);
164 	bn_mulw_addtw(a[3], b[5], c2, c1, c0, &c2, &c1, &c0);
165 	bn_mulw_addtw(a[2], b[6], c2, c1, c0, &c2, &c1, &c0);
166 	bn_mulw_addtw(a[1], b[7], c2, c1, c0, &c2, &c1, &r[8]);
167 
168 	bn_mulw_addtw(a[2], b[7],  0, c2, c1, &c2, &c1, &c0);
169 	bn_mulw_addtw(a[3], b[6], c2, c1, c0, &c2, &c1, &c0);
170 	bn_mulw_addtw(a[4], b[5], c2, c1, c0, &c2, &c1, &c0);
171 	bn_mulw_addtw(a[5], b[4], c2, c1, c0, &c2, &c1, &c0);
172 	bn_mulw_addtw(a[6], b[3], c2, c1, c0, &c2, &c1, &c0);
173 	bn_mulw_addtw(a[7], b[2], c2, c1, c0, &c2, &c1, &r[9]);
174 
175 	bn_mulw_addtw(a[7], b[3],  0, c2, c1, &c2, &c1, &c0);
176 	bn_mulw_addtw(a[6], b[4], c2, c1, c0, &c2, &c1, &c0);
177 	bn_mulw_addtw(a[5], b[5], c2, c1, c0, &c2, &c1, &c0);
178 	bn_mulw_addtw(a[4], b[6], c2, c1, c0, &c2, &c1, &c0);
179 	bn_mulw_addtw(a[3], b[7], c2, c1, c0, &c2, &c1, &r[10]);
180 
181 	bn_mulw_addtw(a[4], b[7],  0, c2, c1, &c2, &c1, &c0);
182 	bn_mulw_addtw(a[5], b[6], c2, c1, c0, &c2, &c1, &c0);
183 	bn_mulw_addtw(a[6], b[5], c2, c1, c0, &c2, &c1, &c0);
184 	bn_mulw_addtw(a[7], b[4], c2, c1, c0, &c2, &c1, &r[11]);
185 
186 	bn_mulw_addtw(a[7], b[5],  0, c2, c1, &c2, &c1, &c0);
187 	bn_mulw_addtw(a[6], b[6], c2, c1, c0, &c2, &c1, &c0);
188 	bn_mulw_addtw(a[5], b[7], c2, c1, c0, &c2, &c1, &r[12]);
189 
190 	bn_mulw_addtw(a[6], b[7],  0, c2, c1, &c2, &c1, &c0);
191 	bn_mulw_addtw(a[7], b[6], c2, c1, c0, &c2, &c1, &r[13]);
192 
193 	bn_mulw_addtw(a[7], b[7],  0, c2, c1, &c2, &r[15], &r[14]);
194 }
195 #endif
196 
197 /*
198  * bn_mul_words() computes (carry:r[i]) = a[i] * w + carry, where a is an array
199  * of words and w is a single word. This should really be called bn_mulw_words()
200  * since only one input is an array. This is used as a step in the multiplication
201  * of word arrays.
202  */
203 #ifndef HAVE_BN_MUL_WORDS
204 BN_ULONG
205 bn_mul_words(BN_ULONG *r, const BN_ULONG *a, int num, BN_ULONG w)
206 {
207 	BN_ULONG carry = 0;
208 
209 	assert(num >= 0);
210 	if (num <= 0)
211 		return 0;
212 
213 #ifndef OPENSSL_SMALL_FOOTPRINT
214 	while (num & ~3) {
215 		bn_mulw_addw(a[0], w, carry, &carry, &r[0]);
216 		bn_mulw_addw(a[1], w, carry, &carry, &r[1]);
217 		bn_mulw_addw(a[2], w, carry, &carry, &r[2]);
218 		bn_mulw_addw(a[3], w, carry, &carry, &r[3]);
219 		a += 4;
220 		r += 4;
221 		num -= 4;
222 	}
223 #endif
224 	while (num) {
225 		bn_mulw_addw(a[0], w, carry, &carry, &r[0]);
226 		a++;
227 		r++;
228 		num--;
229 	}
230 	return carry;
231 }
232 #endif
233 
234 /*
235  * bn_mul_add_words() computes (carry:r[i]) = a[i] * w + r[i] + carry, where
236  * a is an array of words and w is a single word. This should really be called
237  * bn_mulw_add_words() since only one input is an array. This is used as a step
238  * in the multiplication of word arrays.
239  */
240 #ifndef HAVE_BN_MUL_ADD_WORDS
241 BN_ULONG
242 bn_mul_add_words(BN_ULONG *r, const BN_ULONG *a, int num, BN_ULONG w)
243 {
244 	BN_ULONG carry = 0;
245 
246 	assert(num >= 0);
247 	if (num <= 0)
248 		return 0;
249 
250 #ifndef OPENSSL_SMALL_FOOTPRINT
251 	while (num & ~3) {
252 		bn_mulw_addw_addw(a[0], w, r[0], carry, &carry, &r[0]);
253 		bn_mulw_addw_addw(a[1], w, r[1], carry, &carry, &r[1]);
254 		bn_mulw_addw_addw(a[2], w, r[2], carry, &carry, &r[2]);
255 		bn_mulw_addw_addw(a[3], w, r[3], carry, &carry, &r[3]);
256 		a += 4;
257 		r += 4;
258 		num -= 4;
259 	}
260 #endif
261 	while (num) {
262 		bn_mulw_addw_addw(a[0], w, r[0], carry, &carry, &r[0]);
263 		a++;
264 		r++;
265 		num--;
266 	}
267 
268 	return carry;
269 }
270 #endif
271 
272 void
273 bn_mul_normal(BN_ULONG *r, BN_ULONG *a, int na, BN_ULONG *b, int nb)
274 {
275 	BN_ULONG *rr;
276 
277 
278 	if (na < nb) {
279 		int itmp;
280 		BN_ULONG *ltmp;
281 
282 		itmp = na;
283 		na = nb;
284 		nb = itmp;
285 		ltmp = a;
286 		a = b;
287 		b = ltmp;
288 
289 	}
290 	rr = &(r[na]);
291 	if (nb <= 0) {
292 		(void)bn_mul_words(r, a, na, 0);
293 		return;
294 	} else
295 		rr[0] = bn_mul_words(r, a, na, b[0]);
296 
297 	for (;;) {
298 		if (--nb <= 0)
299 			return;
300 		rr[1] = bn_mul_add_words(&(r[1]), a, na, b[1]);
301 		if (--nb <= 0)
302 			return;
303 		rr[2] = bn_mul_add_words(&(r[2]), a, na, b[2]);
304 		if (--nb <= 0)
305 			return;
306 		rr[3] = bn_mul_add_words(&(r[3]), a, na, b[3]);
307 		if (--nb <= 0)
308 			return;
309 		rr[4] = bn_mul_add_words(&(r[4]), a, na, b[4]);
310 		rr += 4;
311 		r += 4;
312 		b += 4;
313 	}
314 }
315 
316 
317 #ifndef HAVE_BN_MUL
318 int
319 bn_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, int rn, BN_CTX *ctx)
320 {
321 	bn_mul_normal(r->d, a->d, a->top, b->d, b->top);
322 
323 	return 1;
324 }
325 
326 #endif /* HAVE_BN_MUL */
327 
328 int
329 BN_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, BN_CTX *ctx)
330 {
331 	BIGNUM *rr;
332 	int rn;
333 	int ret = 0;
334 
335 	BN_CTX_start(ctx);
336 
337 	if (BN_is_zero(a) || BN_is_zero(b)) {
338 		BN_zero(r);
339 		goto done;
340 	}
341 
342 	rr = r;
343 	if (rr == a || rr == b)
344 		rr = BN_CTX_get(ctx);
345 	if (rr == NULL)
346 		goto err;
347 
348 	rn = a->top + b->top;
349 	if (rn < a->top)
350 		goto err;
351 	if (!bn_wexpand(rr, rn))
352 		goto err;
353 
354 	if (a->top == 4 && b->top == 4) {
355 		bn_mul_comba4(rr->d, a->d, b->d);
356 	} else if (a->top == 8 && b->top == 8) {
357 		bn_mul_comba8(rr->d, a->d, b->d);
358 	} else {
359 		if (!bn_mul(rr, a, b, rn, ctx))
360 			goto err;
361 	}
362 
363 	rr->top = rn;
364 	bn_correct_top(rr);
365 
366 	BN_set_negative(rr, a->neg ^ b->neg);
367 
368 	if (!bn_copy(r, rr))
369 		goto err;
370  done:
371 	ret = 1;
372  err:
373 	BN_CTX_end(ctx);
374 
375 	return ret;
376 }
377