xref: /illumos-gate/usr/src/common/crypto/ecc/ecp_192.c (revision 7ecb637260da07a2d98353887f8a77793afc3cf1)
1 /*
2  * ***** BEGIN LICENSE BLOCK *****
3  * Version: MPL 1.1/GPL 2.0/LGPL 2.1
4  *
5  * The contents of this file are subject to the Mozilla Public License Version
6  * 1.1 (the "License"); you may not use this file except in compliance with
7  * the License. You may obtain a copy of the License at
8  * http://www.mozilla.org/MPL/
9  *
10  * Software distributed under the License is distributed on an "AS IS" basis,
11  * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
12  * for the specific language governing rights and limitations under the
13  * License.
14  *
15  * The Original Code is the elliptic curve math library for prime field curves.
16  *
17  * The Initial Developer of the Original Code is
18  * Sun Microsystems, Inc.
19  * Portions created by the Initial Developer are Copyright (C) 2003
20  * the Initial Developer. All Rights Reserved.
21  *
22  * Contributor(s):
23  *   Douglas Stebila <douglas@stebila.ca>, Sun Microsystems Laboratories
24  *
25  * Alternatively, the contents of this file may be used under the terms of
26  * either the GNU General Public License Version 2 or later (the "GPL"), or
27  * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
28  * in which case the provisions of the GPL or the LGPL are applicable instead
29  * of those above. If you wish to allow use of your version of this file only
30  * under the terms of either the GPL or the LGPL, and not to allow others to
31  * use your version of this file under the terms of the MPL, indicate your
32  * decision by deleting the provisions above and replace them with the notice
33  * and other provisions required by the GPL or the LGPL. If you do not delete
34  * the provisions above, a recipient may use your version of this file under
35  * the terms of any one of the MPL, the GPL or the LGPL.
36  *
37  * ***** END LICENSE BLOCK ***** */
38 /*
39  * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
40  * Use is subject to license terms.
41  *
42  * Sun elects to use this software under the MPL license.
43  */
44 
45 #include "ecp.h"
46 #include "mpi.h"
47 #include "mplogic.h"
48 #include "mpi-priv.h"
49 #ifndef _KERNEL
50 #include <stdlib.h>
51 #endif
52 
53 #define ECP192_DIGITS ECL_CURVE_DIGITS(192)
54 
55 /* Fast modular reduction for p192 = 2^192 - 2^64 - 1.  a can be r. Uses
56  * algorithm 7 from Brown, Hankerson, Lopez, Menezes. Software
57  * Implementation of the NIST Elliptic Curves over Prime Fields. */
58 mp_err
ec_GFp_nistp192_mod(const mp_int * a,mp_int * r,const GFMethod * meth)59 ec_GFp_nistp192_mod(const mp_int *a, mp_int *r, const GFMethod *meth)
60 {
61 	mp_err res = MP_OKAY;
62 	mp_size a_used = MP_USED(a);
63 	mp_digit r3;
64 #ifndef MPI_AMD64_ADD
65 	mp_digit carry;
66 #endif
67 #ifdef ECL_THIRTY_TWO_BIT
68 	mp_digit a5a = 0, a5b = 0, a4a = 0, a4b = 0, a3a = 0, a3b = 0;
69         mp_digit r0a, r0b, r1a, r1b, r2a, r2b;
70 #else
71 	mp_digit a5 = 0, a4 = 0, a3 = 0;
72         mp_digit r0, r1, r2;
73 #endif
74 
75 	/* reduction not needed if a is not larger than field size */
76 	if (a_used < ECP192_DIGITS) {
77 		if (a == r) {
78 			return MP_OKAY;
79 		}
80 		return mp_copy(a, r);
81 	}
82 
83 	/* for polynomials larger than twice the field size, use regular
84 	 * reduction */
85 	if (a_used > ECP192_DIGITS*2) {
86 		MP_CHECKOK(mp_mod(a, &meth->irr, r));
87 	} else {
88 		/* copy out upper words of a */
89 
90 #ifdef ECL_THIRTY_TWO_BIT
91 
92 		/* in all the math below,
93 		 * nXb is most signifiant, nXa is least significant */
94 		switch (a_used) {
95 		case 12:
96 			a5b = MP_DIGIT(a, 11);
97 			/* FALLTHROUGH */
98 		case 11:
99 			a5a = MP_DIGIT(a, 10);
100 			/* FALLTHROUGH */
101 		case 10:
102 			a4b = MP_DIGIT(a, 9);
103 			/* FALLTHROUGH */
104 		case 9:
105 			a4a = MP_DIGIT(a, 8);
106 			/* FALLTHROUGH */
107 		case 8:
108 			a3b = MP_DIGIT(a, 7);
109 			/* FALLTHROUGH */
110 		case 7:
111 			a3a = MP_DIGIT(a, 6);
112 		}
113 
114 
115                 r2b= MP_DIGIT(a, 5);
116                 r2a= MP_DIGIT(a, 4);
117                 r1b = MP_DIGIT(a, 3);
118                 r1a = MP_DIGIT(a, 2);
119                 r0b = MP_DIGIT(a, 1);
120                 r0a = MP_DIGIT(a, 0);
121 
122 		/* implement r = (a2,a1,a0)+(a5,a5,a5)+(a4,a4,0)+(0,a3,a3) */
123 		MP_ADD_CARRY(r0a, a3a, r0a, 0,    carry);
124 		MP_ADD_CARRY(r0b, a3b, r0b, carry, carry);
125 		MP_ADD_CARRY(r1a, a3a, r1a, carry, carry);
126 		MP_ADD_CARRY(r1b, a3b, r1b, carry, carry);
127 		MP_ADD_CARRY(r2a, a4a, r2a, carry, carry);
128 		MP_ADD_CARRY(r2b, a4b, r2b, carry, carry);
129 		r3 = carry; carry = 0;
130 		MP_ADD_CARRY(r0a, a5a, r0a, 0,     carry);
131 		MP_ADD_CARRY(r0b, a5b, r0b, carry, carry);
132 		MP_ADD_CARRY(r1a, a5a, r1a, carry, carry);
133 		MP_ADD_CARRY(r1b, a5b, r1b, carry, carry);
134 		MP_ADD_CARRY(r2a, a5a, r2a, carry, carry);
135 		MP_ADD_CARRY(r2b, a5b, r2b, carry, carry);
136 		r3 += carry;
137 		MP_ADD_CARRY(r1a, a4a, r1a, 0,     carry);
138 		MP_ADD_CARRY(r1b, a4b, r1b, carry, carry);
139 		MP_ADD_CARRY(r2a,   0, r2a, carry, carry);
140 		MP_ADD_CARRY(r2b,   0, r2b, carry, carry);
141 		r3 += carry;
142 
143 		/* reduce out the carry */
144 		while (r3) {
145 			MP_ADD_CARRY(r0a, r3, r0a, 0,     carry);
146 			MP_ADD_CARRY(r0b,  0, r0b, carry, carry);
147 			MP_ADD_CARRY(r1a, r3, r1a, carry, carry);
148 			MP_ADD_CARRY(r1b,  0, r1b, carry, carry);
149 			MP_ADD_CARRY(r2a,  0, r2a, carry, carry);
150 			MP_ADD_CARRY(r2b,  0, r2b, carry, carry);
151 			r3 = carry;
152 		}
153 
154 		/* check for final reduction */
155 		/*
156 		 * our field is 0xffffffffffffffff, 0xfffffffffffffffe,
157 		 * 0xffffffffffffffff. That means we can only be over and need
158 		 * one more reduction
159 		 *  if r2 == 0xffffffffffffffffff (same as r2+1 == 0)
160 		 *     and
161 		 *     r1 == 0xffffffffffffffffff   or
162 		 *     r1 == 0xfffffffffffffffffe and r0 = 0xfffffffffffffffff
163 		 * In all cases, we subtract the field (or add the 2's
164 		 * complement value (1,1,0)).  (r0, r1, r2)
165 		 */
166 		if (((r2b == 0xffffffff) && (r2a == 0xffffffff)
167 			&& (r1b == 0xffffffff) ) &&
168 			   ((r1a == 0xffffffff) ||
169 			    (r1a == 0xfffffffe) && (r0a == 0xffffffff) &&
170 					(r0b == 0xffffffff)) ) {
171 			/* do a quick subtract */
172 			MP_ADD_CARRY(r0a, 1, r0a, 0, carry);
173 			MP_ADD_CARRY(r0b, 0, r0b, carry, carry);
174 			r1a += 1 + carry;
175 			r1b = r2a = r2b = 0;
176 		}
177 
178 		/* set the lower words of r */
179 		if (a != r) {
180 			MP_CHECKOK(s_mp_pad(r, 6));
181 		}
182 		MP_DIGIT(r, 5) = r2b;
183 		MP_DIGIT(r, 4) = r2a;
184 		MP_DIGIT(r, 3) = r1b;
185 		MP_DIGIT(r, 2) = r1a;
186 		MP_DIGIT(r, 1) = r0b;
187 		MP_DIGIT(r, 0) = r0a;
188 		MP_USED(r) = 6;
189 #else
190 		switch (a_used) {
191 		case 6:
192 			a5 = MP_DIGIT(a, 5);
193 			/* FALLTHROUGH */
194 		case 5:
195 			a4 = MP_DIGIT(a, 4);
196 			/* FALLTHROUGH */
197 		case 4:
198 			a3 = MP_DIGIT(a, 3);
199 		}
200 
201                 r2 = MP_DIGIT(a, 2);
202                 r1 = MP_DIGIT(a, 1);
203                 r0 = MP_DIGIT(a, 0);
204 
205 		/* implement r = (a2,a1,a0)+(a5,a5,a5)+(a4,a4,0)+(0,a3,a3) */
206 #ifndef MPI_AMD64_ADD
207 		MP_ADD_CARRY(r0, a3, r0, 0,     carry);
208 		MP_ADD_CARRY(r1, a3, r1, carry, carry);
209 		MP_ADD_CARRY(r2, a4, r2, carry, carry);
210 		r3 = carry;
211 		MP_ADD_CARRY(r0, a5, r0, 0,     carry);
212 		MP_ADD_CARRY(r1, a5, r1, carry, carry);
213 		MP_ADD_CARRY(r2, a5, r2, carry, carry);
214 		r3 += carry;
215 		MP_ADD_CARRY(r1, a4, r1, 0,     carry);
216 		MP_ADD_CARRY(r2,  0, r2, carry, carry);
217 		r3 += carry;
218 
219 #else
220                 r2 = MP_DIGIT(a, 2);
221                 r1 = MP_DIGIT(a, 1);
222                 r0 = MP_DIGIT(a, 0);
223 
224                 /* set the lower words of r */
225                 __asm__ (
226                 "xorq   %3,%3           \n\t"
227                 "addq   %4,%0           \n\t"
228                 "adcq   %4,%1           \n\t"
229                 "adcq   %5,%2           \n\t"
230                 "adcq   $0,%3           \n\t"
231                 "addq   %6,%0           \n\t"
232                 "adcq   %6,%1           \n\t"
233                 "adcq   %6,%2           \n\t"
234                 "adcq   $0,%3           \n\t"
235                 "addq   %5,%1           \n\t"
236                 "adcq   $0,%2           \n\t"
237                 "adcq   $0,%3           \n\t"
238                 : "=r"(r0), "=r"(r1), "=r"(r2), "=r"(r3), "=r"(a3),
239 		  "=r"(a4), "=r"(a5)
240                 : "0" (r0), "1" (r1), "2" (r2), "3" (r3),
241 		  "4" (a3), "5" (a4), "6"(a5)
242                 : "%cc" );
243 #endif
244 
245 		/* reduce out the carry */
246 		while (r3) {
247 #ifndef MPI_AMD64_ADD
248 			MP_ADD_CARRY(r0, r3, r0, 0,     carry);
249 			MP_ADD_CARRY(r1, r3, r1, carry, carry);
250 			MP_ADD_CARRY(r2,  0, r2, carry, carry);
251 			r3 = carry;
252 #else
253 			a3=r3;
254               		__asm__ (
255                 	"xorq   %3,%3           \n\t"
256                 	"addq   %4,%0           \n\t"
257                 	"adcq   %4,%1           \n\t"
258                 	"adcq   $0,%2           \n\t"
259                 	"adcq   $0,%3           \n\t"
260                 	: "=r"(r0), "=r"(r1), "=r"(r2), "=r"(r3), "=r"(a3)
261                 	: "0" (r0), "1" (r1), "2" (r2), "3" (r3), "4"(a3)
262                 	: "%cc" );
263 #endif
264 		}
265 
266 		/* check for final reduction */
267 		/*
268 		 * our field is 0xffffffffffffffff, 0xfffffffffffffffe,
269 		 * 0xffffffffffffffff. That means we can only be over and need
270 		 * one more reduction
271 		 *  if r2 == 0xffffffffffffffffff (same as r2+1 == 0)
272 		 *     and
273 		 *     r1 == 0xffffffffffffffffff   or
274 		 *     r1 == 0xfffffffffffffffffe and r0 = 0xfffffffffffffffff
275 		 * In all cases, we subtract the field (or add the 2's
276 		 * complement value (1,1,0)).  (r0, r1, r2)
277 		 */
278 		if (r3 || ((r2 == MP_DIGIT_MAX) &&
279 		      ((r1 == MP_DIGIT_MAX) ||
280 			((r1 == (MP_DIGIT_MAX-1)) && (r0 == MP_DIGIT_MAX))))) {
281 			/* do a quick subtract */
282 			MP_ADD_CARRY(r0, 1, r0, 0, carry);
283 			r1 += 1 + carry;
284 			r2 = 0;
285 		}
286 		/* set the lower words of r */
287 		if (a != r) {
288 			MP_CHECKOK(s_mp_pad(r, 3));
289 		}
290 		MP_DIGIT(r, 2) = r2;
291 		MP_DIGIT(r, 1) = r1;
292 		MP_DIGIT(r, 0) = r0;
293 		MP_USED(r) = 3;
294 #endif
295 	}
296 
297 	s_mp_clamp(r);
298   CLEANUP:
299 	return res;
300 }
301 
302 #ifndef ECL_THIRTY_TWO_BIT
303 /* Compute the sum of 192 bit curves. Do the work in-line since the
304  * number of words are so small, we don't want to overhead of mp function
305  * calls.  Uses optimized modular reduction for p192.
306  */
307 mp_err
ec_GFp_nistp192_add(const mp_int * a,const mp_int * b,mp_int * r,const GFMethod * meth)308 ec_GFp_nistp192_add(const mp_int *a, const mp_int *b, mp_int *r,
309 			const GFMethod *meth)
310 {
311 	mp_err res = MP_OKAY;
312 	mp_digit a0 = 0, a1 = 0, a2 = 0;
313 	mp_digit r0 = 0, r1 = 0, r2 = 0;
314 	mp_digit carry;
315 
316 	switch(MP_USED(a)) {
317 	case 3:
318 		a2 = MP_DIGIT(a,2);
319 		/* FALLTHROUGH */
320 	case 2:
321 		a1 = MP_DIGIT(a,1);
322 		/* FALLTHROUGH */
323 	case 1:
324 		a0 = MP_DIGIT(a,0);
325 	}
326 	switch(MP_USED(b)) {
327 	case 3:
328 		r2 = MP_DIGIT(b,2);
329 		/* FALLTHROUGH */
330 	case 2:
331 		r1 = MP_DIGIT(b,1);
332 		/* FALLTHROUGH */
333 	case 1:
334 		r0 = MP_DIGIT(b,0);
335 	}
336 
337 #ifndef MPI_AMD64_ADD
338 	MP_ADD_CARRY(a0, r0, r0, 0,     carry);
339 	MP_ADD_CARRY(a1, r1, r1, carry, carry);
340 	MP_ADD_CARRY(a2, r2, r2, carry, carry);
341 #else
342 	__asm__ (
343                 "xorq   %3,%3           \n\t"
344                 "addq   %4,%0           \n\t"
345                 "adcq   %5,%1           \n\t"
346                 "adcq   %6,%2           \n\t"
347                 "adcq   $0,%3           \n\t"
348                 : "=r"(r0), "=r"(r1), "=r"(r2), "=r"(carry)
349                 : "r" (a0), "r" (a1), "r" (a2), "0" (r0),
350 		  "1" (r1), "2" (r2)
351                 : "%cc" );
352 #endif
353 
354 	/* Do quick 'subract' if we've gone over
355 	 * (add the 2's complement of the curve field) */
356 	if (carry || ((r2 == MP_DIGIT_MAX) &&
357 		      ((r1 == MP_DIGIT_MAX) ||
358 			((r1 == (MP_DIGIT_MAX-1)) && (r0 == MP_DIGIT_MAX))))) {
359 #ifndef MPI_AMD64_ADD
360 		MP_ADD_CARRY(r0, 1, r0, 0,     carry);
361 		MP_ADD_CARRY(r1, 1, r1, carry, carry);
362 		MP_ADD_CARRY(r2, 0, r2, carry, carry);
363 #else
364 		__asm__ (
365 			"addq   $1,%0           \n\t"
366 			"adcq   $1,%1           \n\t"
367 			"adcq   $0,%2           \n\t"
368 			: "=r"(r0), "=r"(r1), "=r"(r2)
369 			: "0" (r0), "1" (r1), "2" (r2)
370 			: "%cc" );
371 #endif
372 	}
373 
374 
375 	MP_CHECKOK(s_mp_pad(r, 3));
376 	MP_DIGIT(r, 2) = r2;
377 	MP_DIGIT(r, 1) = r1;
378 	MP_DIGIT(r, 0) = r0;
379 	MP_SIGN(r) = MP_ZPOS;
380 	MP_USED(r) = 3;
381 	s_mp_clamp(r);
382 
383 
384   CLEANUP:
385 	return res;
386 }
387 
388 /* Compute the diff of 192 bit curves. Do the work in-line since the
389  * number of words are so small, we don't want to overhead of mp function
390  * calls.  Uses optimized modular reduction for p192.
391  */
392 mp_err
ec_GFp_nistp192_sub(const mp_int * a,const mp_int * b,mp_int * r,const GFMethod * meth)393 ec_GFp_nistp192_sub(const mp_int *a, const mp_int *b, mp_int *r,
394 			const GFMethod *meth)
395 {
396 	mp_err res = MP_OKAY;
397 	mp_digit b0 = 0, b1 = 0, b2 = 0;
398 	mp_digit r0 = 0, r1 = 0, r2 = 0;
399 	mp_digit borrow;
400 
401 	switch(MP_USED(a)) {
402 	case 3:
403 		r2 = MP_DIGIT(a,2);
404 		/* FALLTHROUGH */
405 	case 2:
406 		r1 = MP_DIGIT(a,1);
407 		/* FALLTHROUGH */
408 	case 1:
409 		r0 = MP_DIGIT(a,0);
410 	}
411 
412 	switch(MP_USED(b)) {
413 	case 3:
414 		b2 = MP_DIGIT(b,2);
415 		/* FALLTHROUGH */
416 	case 2:
417 		b1 = MP_DIGIT(b,1);
418 		/* FALLTHROUGH */
419 	case 1:
420 		b0 = MP_DIGIT(b,0);
421 	}
422 
423 #ifndef MPI_AMD64_ADD
424 	MP_SUB_BORROW(r0, b0, r0, 0,     borrow);
425 	MP_SUB_BORROW(r1, b1, r1, borrow, borrow);
426 	MP_SUB_BORROW(r2, b2, r2, borrow, borrow);
427 #else
428 	__asm__ (
429                 "xorq   %3,%3           \n\t"
430                 "subq   %4,%0           \n\t"
431                 "sbbq   %5,%1           \n\t"
432                 "sbbq   %6,%2           \n\t"
433                 "adcq   $0,%3           \n\t"
434                 : "=r"(r0), "=r"(r1), "=r"(r2), "=r"(borrow)
435                 : "r" (b0), "r" (b1), "r" (b2), "0" (r0),
436 		  "1" (r1), "2" (r2)
437                 : "%cc" );
438 #endif
439 
440 	/* Do quick 'add' if we've gone under 0
441 	 * (subtract the 2's complement of the curve field) */
442 	if (borrow) {
443 #ifndef MPI_AMD64_ADD
444 		MP_SUB_BORROW(r0, 1, r0, 0,     borrow);
445 		MP_SUB_BORROW(r1, 1, r1, borrow, borrow);
446 		MP_SUB_BORROW(r2,  0, r2, borrow, borrow);
447 #else
448 		__asm__ (
449 			"subq   $1,%0           \n\t"
450 			"sbbq   $1,%1           \n\t"
451 			"sbbq   $0,%2           \n\t"
452 			: "=r"(r0), "=r"(r1), "=r"(r2)
453 			: "0" (r0), "1" (r1), "2" (r2)
454 			: "%cc" );
455 #endif
456 	}
457 
458 	MP_CHECKOK(s_mp_pad(r, 3));
459 	MP_DIGIT(r, 2) = r2;
460 	MP_DIGIT(r, 1) = r1;
461 	MP_DIGIT(r, 0) = r0;
462 	MP_SIGN(r) = MP_ZPOS;
463 	MP_USED(r) = 3;
464 	s_mp_clamp(r);
465 
466   CLEANUP:
467 	return res;
468 }
469 
470 #endif
471 
472 /* Compute the square of polynomial a, reduce modulo p192. Store the
473  * result in r.  r could be a.  Uses optimized modular reduction for p192.
474  */
475 mp_err
ec_GFp_nistp192_sqr(const mp_int * a,mp_int * r,const GFMethod * meth)476 ec_GFp_nistp192_sqr(const mp_int *a, mp_int *r, const GFMethod *meth)
477 {
478 	mp_err res = MP_OKAY;
479 
480 	MP_CHECKOK(mp_sqr(a, r));
481 	MP_CHECKOK(ec_GFp_nistp192_mod(r, r, meth));
482   CLEANUP:
483 	return res;
484 }
485 
486 /* Compute the product of two polynomials a and b, reduce modulo p192.
487  * Store the result in r.  r could be a or b; a could be b.  Uses
488  * optimized modular reduction for p192. */
489 mp_err
ec_GFp_nistp192_mul(const mp_int * a,const mp_int * b,mp_int * r,const GFMethod * meth)490 ec_GFp_nistp192_mul(const mp_int *a, const mp_int *b, mp_int *r,
491 					const GFMethod *meth)
492 {
493 	mp_err res = MP_OKAY;
494 
495 	MP_CHECKOK(mp_mul(a, b, r));
496 	MP_CHECKOK(ec_GFp_nistp192_mod(r, r, meth));
497   CLEANUP:
498 	return res;
499 }
500 
501 /* Divides two field elements. If a is NULL, then returns the inverse of
502  * b. */
503 mp_err
ec_GFp_nistp192_div(const mp_int * a,const mp_int * b,mp_int * r,const GFMethod * meth)504 ec_GFp_nistp192_div(const mp_int *a, const mp_int *b, mp_int *r,
505 		   const GFMethod *meth)
506 {
507 	mp_err res = MP_OKAY;
508 	mp_int t;
509 
510 	/* If a is NULL, then return the inverse of b, otherwise return a/b. */
511 	if (a == NULL) {
512 		return  mp_invmod(b, &meth->irr, r);
513 	} else {
514 		/* MPI doesn't support divmod, so we implement it using invmod and
515 		 * mulmod. */
516 		MP_CHECKOK(mp_init(&t, FLAG(b)));
517 		MP_CHECKOK(mp_invmod(b, &meth->irr, &t));
518 		MP_CHECKOK(mp_mul(a, &t, r));
519 		MP_CHECKOK(ec_GFp_nistp192_mod(r, r, meth));
520 	  CLEANUP:
521 		mp_clear(&t);
522 		return res;
523 	}
524 }
525 
526 /* Wire in fast field arithmetic and precomputation of base point for
527  * named curves. */
528 mp_err
ec_group_set_gfp192(ECGroup * group,ECCurveName name)529 ec_group_set_gfp192(ECGroup *group, ECCurveName name)
530 {
531 	if (name == ECCurve_NIST_P192) {
532 		group->meth->field_mod = &ec_GFp_nistp192_mod;
533 		group->meth->field_mul = &ec_GFp_nistp192_mul;
534 		group->meth->field_sqr = &ec_GFp_nistp192_sqr;
535 		group->meth->field_div = &ec_GFp_nistp192_div;
536 #ifndef ECL_THIRTY_TWO_BIT
537 		group->meth->field_add = &ec_GFp_nistp192_add;
538 		group->meth->field_sub = &ec_GFp_nistp192_sub;
539 #endif
540 	}
541 	return MP_OKAY;
542 }
543