1/*
2 * Copyright (c) 2007, 2011, Oracle and/or its affiliates. All rights reserved.
3 * Use is subject to license terms.
4 *
5 * This library is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU Lesser General Public
7 * License as published by the Free Software Foundation; either
8 * version 2.1 of the License, or (at your option) any later version.
9 *
10 * This library is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * Lesser General Public License for more details.
14 *
15 * You should have received a copy of the GNU Lesser General Public License
16 * along with this library; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 */
23
24/* *********************************************************************
25 *
26 * The Original Code is the elliptic curve math library for prime field curves.
27 *
28 * The Initial Developer of the Original Code is
29 * Sun Microsystems, Inc.
30 * Portions created by the Initial Developer are Copyright (C) 2003
31 * the Initial Developer. All Rights Reserved.
32 *
33 * Contributor(s):
34 * Douglas Stebila <douglas@stebila.ca>, Sun Microsystems Laboratories
35 *
36 *********************************************************************** */
37
38#include "ecp.h"
39#include "mpi.h"
40#include "mplogic.h"
41#include "mpi-priv.h"
42#ifndef _KERNEL
43#include <stdlib.h>
44#endif
45
46#define ECP192_DIGITS ECL_CURVE_DIGITS(192)
47
48/* Fast modular reduction for p192 = 2^192 - 2^64 - 1. a can be r. Uses
49 * algorithm 7 from Brown, Hankerson, Lopez, Menezes. Software
50 * Implementation of the NIST Elliptic Curves over Prime Fields. */
51mp_err
52ec_GFp_nistp192_mod(const mp_int *a, mp_int *r, const GFMethod *meth)
53{
54 mp_err res = MP_OKAY;
55 mp_size a_used = MP_USED(a);
56 mp_digit r3;
57#ifndef MPI_AMD64_ADD
58 mp_digit carry;
59#endif
60#ifdef ECL_THIRTY_TWO_BIT
61 mp_digit a5a = 0, a5b = 0, a4a = 0, a4b = 0, a3a = 0, a3b = 0;
62 mp_digit r0a, r0b, r1a, r1b, r2a, r2b;
63#else
64 mp_digit a5 = 0, a4 = 0, a3 = 0;
65 mp_digit r0, r1, r2;
66#endif
67
68 /* reduction not needed if a is not larger than field size */
69 if (a_used < ECP192_DIGITS) {
70 if (a == r) {
71 return MP_OKAY;
72 }
73 return mp_copy(a, r);
74 }
75
76 /* for polynomials larger than twice the field size, use regular
77 * reduction */
78 if (a_used > ECP192_DIGITS*2) {
79 MP_CHECKOK(mp_mod(a, &meth->irr, r));
80 } else {
81 /* copy out upper words of a */
82
83#ifdef ECL_THIRTY_TWO_BIT
84
85 /* in all the math below,
86 * nXb is most signifiant, nXa is least significant */
87 switch (a_used) {
88 case 12:
89 a5b = MP_DIGIT(a, 11);
90 case 11:
91 a5a = MP_DIGIT(a, 10);
92 case 10:
93 a4b = MP_DIGIT(a, 9);
94 case 9:
95 a4a = MP_DIGIT(a, 8);
96 case 8:
97 a3b = MP_DIGIT(a, 7);
98 case 7:
99 a3a = MP_DIGIT(a, 6);
100 }
101
102
103 r2b= MP_DIGIT(a, 5);
104 r2a= MP_DIGIT(a, 4);
105 r1b = MP_DIGIT(a, 3);
106 r1a = MP_DIGIT(a, 2);
107 r0b = MP_DIGIT(a, 1);
108 r0a = MP_DIGIT(a, 0);
109
110 /* implement r = (a2,a1,a0)+(a5,a5,a5)+(a4,a4,0)+(0,a3,a3) */
111 MP_ADD_CARRY(r0a, a3a, r0a, 0, carry);
112 MP_ADD_CARRY(r0b, a3b, r0b, carry, carry);
113 MP_ADD_CARRY(r1a, a3a, r1a, carry, carry);
114 MP_ADD_CARRY(r1b, a3b, r1b, carry, carry);
115 MP_ADD_CARRY(r2a, a4a, r2a, carry, carry);
116 MP_ADD_CARRY(r2b, a4b, r2b, carry, carry);
117 r3 = carry; carry = 0;
118 MP_ADD_CARRY(r0a, a5a, r0a, 0, carry);
119 MP_ADD_CARRY(r0b, a5b, r0b, carry, carry);
120 MP_ADD_CARRY(r1a, a5a, r1a, carry, carry);
121 MP_ADD_CARRY(r1b, a5b, r1b, carry, carry);
122 MP_ADD_CARRY(r2a, a5a, r2a, carry, carry);
123 MP_ADD_CARRY(r2b, a5b, r2b, carry, carry);
124 r3 += carry;
125 MP_ADD_CARRY(r1a, a4a, r1a, 0, carry);
126 MP_ADD_CARRY(r1b, a4b, r1b, carry, carry);
127 MP_ADD_CARRY(r2a, 0, r2a, carry, carry);
128 MP_ADD_CARRY(r2b, 0, r2b, carry, carry);
129 r3 += carry;
130
131 /* reduce out the carry */
132 while (r3) {
133 MP_ADD_CARRY(r0a, r3, r0a, 0, carry);
134 MP_ADD_CARRY(r0b, 0, r0b, carry, carry);
135 MP_ADD_CARRY(r1a, r3, r1a, carry, carry);
136 MP_ADD_CARRY(r1b, 0, r1b, carry, carry);
137 MP_ADD_CARRY(r2a, 0, r2a, carry, carry);
138 MP_ADD_CARRY(r2b, 0, r2b, carry, carry);
139 r3 = carry;
140 }
141
142 /* check for final reduction */
143 /*
144 * our field is 0xffffffffffffffff, 0xfffffffffffffffe,
145 * 0xffffffffffffffff. That means we can only be over and need
146 * one more reduction
147 * if r2 == 0xffffffffffffffffff (same as r2+1 == 0)
148 * and
149 * r1 == 0xffffffffffffffffff or
150 * r1 == 0xfffffffffffffffffe and r0 = 0xfffffffffffffffff
151 * In all cases, we subtract the field (or add the 2's
152 * complement value (1,1,0)). (r0, r1, r2)
153 */
154 if (((r2b == 0xffffffff) && (r2a == 0xffffffff)
155 && (r1b == 0xffffffff) ) &&
156 ((r1a == 0xffffffff) ||
157 (r1a == 0xfffffffe) && (r0a == 0xffffffff) &&
158 (r0b == 0xffffffff)) ) {
159 /* do a quick subtract */
160 MP_ADD_CARRY(r0a, 1, r0a, 0, carry);
161 r0b += carry;
162 r1a = r1b = r2a = r2b = 0;
163 }
164
165 /* set the lower words of r */
166 if (a != r) {
167 MP_CHECKOK(s_mp_pad(r, 6));
168 }
169 MP_DIGIT(r, 5) = r2b;
170 MP_DIGIT(r, 4) = r2a;
171 MP_DIGIT(r, 3) = r1b;
172 MP_DIGIT(r, 2) = r1a;
173 MP_DIGIT(r, 1) = r0b;
174 MP_DIGIT(r, 0) = r0a;
175 MP_USED(r) = 6;
176#else
177 switch (a_used) {
178 case 6:
179 a5 = MP_DIGIT(a, 5);
180 case 5:
181 a4 = MP_DIGIT(a, 4);
182 case 4:
183 a3 = MP_DIGIT(a, 3);
184 }
185
186 r2 = MP_DIGIT(a, 2);
187 r1 = MP_DIGIT(a, 1);
188 r0 = MP_DIGIT(a, 0);
189
190 /* implement r = (a2,a1,a0)+(a5,a5,a5)+(a4,a4,0)+(0,a3,a3) */
191#ifndef MPI_AMD64_ADD
192 MP_ADD_CARRY_ZERO(r0, a3, r0, carry);
193 MP_ADD_CARRY(r1, a3, r1, carry, carry);
194 MP_ADD_CARRY(r2, a4, r2, carry, carry);
195 r3 = carry;
196 MP_ADD_CARRY_ZERO(r0, a5, r0, carry);
197 MP_ADD_CARRY(r1, a5, r1, carry, carry);
198 MP_ADD_CARRY(r2, a5, r2, carry, carry);
199 r3 += carry;
200 MP_ADD_CARRY_ZERO(r1, a4, r1, carry);
201 MP_ADD_CARRY(r2, 0, r2, carry, carry);
202 r3 += carry;
203
204#else
205 r2 = MP_DIGIT(a, 2);
206 r1 = MP_DIGIT(a, 1);
207 r0 = MP_DIGIT(a, 0);
208
209 /* set the lower words of r */
210 __asm__ (
211 "xorq %3,%3 \n\t"
212 "addq %4,%0 \n\t"
213 "adcq %4,%1 \n\t"
214 "adcq %5,%2 \n\t"
215 "adcq $0,%3 \n\t"
216 "addq %6,%0 \n\t"
217 "adcq %6,%1 \n\t"
218 "adcq %6,%2 \n\t"
219 "adcq $0,%3 \n\t"
220 "addq %5,%1 \n\t"
221 "adcq $0,%2 \n\t"
222 "adcq $0,%3 \n\t"
223 : "=r"(r0), "=r"(r1), "=r"(r2), "=r"(r3), "=r"(a3),
224 "=r"(a4), "=r"(a5)
225 : "0" (r0), "1" (r1), "2" (r2), "3" (r3),
226 "4" (a3), "5" (a4), "6"(a5)
227 : "%cc" );
228#endif
229
230 /* reduce out the carry */
231 while (r3) {
232#ifndef MPI_AMD64_ADD
233 MP_ADD_CARRY_ZERO(r0, r3, r0, carry);
234 MP_ADD_CARRY(r1, r3, r1, carry, carry);
235 MP_ADD_CARRY(r2, 0, r2, carry, carry);
236 r3 = carry;
237#else
238 a3=r3;
239 __asm__ (
240 "xorq %3,%3 \n\t"
241 "addq %4,%0 \n\t"
242 "adcq %4,%1 \n\t"
243 "adcq $0,%2 \n\t"
244 "adcq $0,%3 \n\t"
245 : "=r"(r0), "=r"(r1), "=r"(r2), "=r"(r3), "=r"(a3)
246 : "0" (r0), "1" (r1), "2" (r2), "3" (r3), "4"(a3)
247 : "%cc" );
248#endif
249 }
250
251 /* check for final reduction */
252 /*
253 * our field is 0xffffffffffffffff, 0xfffffffffffffffe,
254 * 0xffffffffffffffff. That means we can only be over and need
255 * one more reduction
256 * if r2 == 0xffffffffffffffffff (same as r2+1 == 0)
257 * and
258 * r1 == 0xffffffffffffffffff or
259 * r1 == 0xfffffffffffffffffe and r0 = 0xfffffffffffffffff
260 * In all cases, we subtract the field (or add the 2's
261 * complement value (1,1,0)). (r0, r1, r2)
262 */
263 if (r3 || ((r2 == MP_DIGIT_MAX) &&
264 ((r1 == MP_DIGIT_MAX) ||
265 ((r1 == (MP_DIGIT_MAX-1)) && (r0 == MP_DIGIT_MAX))))) {
266 /* do a quick subtract */
267 r0++;
268 r1 = r2 = 0;
269 }
270 /* set the lower words of r */
271 if (a != r) {
272 MP_CHECKOK(s_mp_pad(r, 3));
273 }
274 MP_DIGIT(r, 2) = r2;
275 MP_DIGIT(r, 1) = r1;
276 MP_DIGIT(r, 0) = r0;
277 MP_USED(r) = 3;
278#endif
279 }
280
281 CLEANUP:
282 return res;
283}
284
285#ifndef ECL_THIRTY_TWO_BIT
286/* Compute the sum of 192 bit curves. Do the work in-line since the
287 * number of words are so small, we don't want to overhead of mp function
288 * calls. Uses optimized modular reduction for p192.
289 */
290mp_err
291ec_GFp_nistp192_add(const mp_int *a, const mp_int *b, mp_int *r,
292 const GFMethod *meth)
293{
294 mp_err res = MP_OKAY;
295 mp_digit a0 = 0, a1 = 0, a2 = 0;
296 mp_digit r0 = 0, r1 = 0, r2 = 0;
297 mp_digit carry;
298
299 switch(MP_USED(a)) {
300 case 3:
301 a2 = MP_DIGIT(a,2);
302 case 2:
303 a1 = MP_DIGIT(a,1);
304 case 1:
305 a0 = MP_DIGIT(a,0);
306 }
307 switch(MP_USED(b)) {
308 case 3:
309 r2 = MP_DIGIT(b,2);
310 case 2:
311 r1 = MP_DIGIT(b,1);
312 case 1:
313 r0 = MP_DIGIT(b,0);
314 }
315
316#ifndef MPI_AMD64_ADD
317 MP_ADD_CARRY_ZERO(a0, r0, r0, carry);
318 MP_ADD_CARRY(a1, r1, r1, carry, carry);
319 MP_ADD_CARRY(a2, r2, r2, carry, carry);
320#else
321 __asm__ (
322 "xorq %3,%3 \n\t"
323 "addq %4,%0 \n\t"
324 "adcq %5,%1 \n\t"
325 "adcq %6,%2 \n\t"
326 "adcq $0,%3 \n\t"
327 : "=r"(r0), "=r"(r1), "=r"(r2), "=r"(carry)
328 : "r" (a0), "r" (a1), "r" (a2), "0" (r0),
329 "1" (r1), "2" (r2)
330 : "%cc" );
331#endif
332
333 /* Do quick 'subract' if we've gone over
334 * (add the 2's complement of the curve field) */
335 if (carry || ((r2 == MP_DIGIT_MAX) &&
336 ((r1 == MP_DIGIT_MAX) ||
337 ((r1 == (MP_DIGIT_MAX-1)) && (r0 == MP_DIGIT_MAX))))) {
338#ifndef MPI_AMD64_ADD
339 MP_ADD_CARRY_ZERO(r0, 1, r0, carry);
340 MP_ADD_CARRY(r1, 1, r1, carry, carry);
341 MP_ADD_CARRY(r2, 0, r2, carry, carry);
342#else
343 __asm__ (
344 "addq $1,%0 \n\t"
345 "adcq $1,%1 \n\t"
346 "adcq $0,%2 \n\t"
347 : "=r"(r0), "=r"(r1), "=r"(r2)
348 : "0" (r0), "1" (r1), "2" (r2)
349 : "%cc" );
350#endif
351 }
352
353
354 MP_CHECKOK(s_mp_pad(r, 3));
355 MP_DIGIT(r, 2) = r2;
356 MP_DIGIT(r, 1) = r1;
357 MP_DIGIT(r, 0) = r0;
358 MP_SIGN(r) = MP_ZPOS;
359 MP_USED(r) = 3;
360 s_mp_clamp(r);
361
362
363 CLEANUP:
364 return res;
365}
366
367/* Compute the diff of 192 bit curves. Do the work in-line since the
368 * number of words are so small, we don't want to overhead of mp function
369 * calls. Uses optimized modular reduction for p192.
370 */
371mp_err
372ec_GFp_nistp192_sub(const mp_int *a, const mp_int *b, mp_int *r,
373 const GFMethod *meth)
374{
375 mp_err res = MP_OKAY;
376 mp_digit b0 = 0, b1 = 0, b2 = 0;
377 mp_digit r0 = 0, r1 = 0, r2 = 0;
378 mp_digit borrow;
379
380 switch(MP_USED(a)) {
381 case 3:
382 r2 = MP_DIGIT(a,2);
383 case 2:
384 r1 = MP_DIGIT(a,1);
385 case 1:
386 r0 = MP_DIGIT(a,0);
387 }
388
389 switch(MP_USED(b)) {
390 case 3:
391 b2 = MP_DIGIT(b,2);
392 case 2:
393 b1 = MP_DIGIT(b,1);
394 case 1:
395 b0 = MP_DIGIT(b,0);
396 }
397
398#ifndef MPI_AMD64_ADD
399 MP_SUB_BORROW(r0, b0, r0, 0, borrow);
400 MP_SUB_BORROW(r1, b1, r1, borrow, borrow);
401 MP_SUB_BORROW(r2, b2, r2, borrow, borrow);
402#else
403 __asm__ (
404 "xorq %3,%3 \n\t"
405 "subq %4,%0 \n\t"
406 "sbbq %5,%1 \n\t"
407 "sbbq %6,%2 \n\t"
408 "adcq $0,%3 \n\t"
409 : "=r"(r0), "=r"(r1), "=r"(r2), "=r"(borrow)
410 : "r" (b0), "r" (b1), "r" (b2), "0" (r0),
411 "1" (r1), "2" (r2)
412 : "%cc" );
413#endif
414
415 /* Do quick 'add' if we've gone under 0
416 * (subtract the 2's complement of the curve field) */
417 if (borrow) {
418#ifndef MPI_AMD64_ADD
419 MP_SUB_BORROW(r0, 1, r0, 0, borrow);
420 MP_SUB_BORROW(r1, 1, r1, borrow, borrow);
421 MP_SUB_BORROW(r2, 0, r2, borrow, borrow);
422#else
423 __asm__ (
424 "subq $1,%0 \n\t"
425 "sbbq $1,%1 \n\t"
426 "sbbq $0,%2 \n\t"
427 : "=r"(r0), "=r"(r1), "=r"(r2)
428 : "0" (r0), "1" (r1), "2" (r2)
429 : "%cc" );
430#endif
431 }
432
433 MP_CHECKOK(s_mp_pad(r, 3));
434 MP_DIGIT(r, 2) = r2;
435 MP_DIGIT(r, 1) = r1;
436 MP_DIGIT(r, 0) = r0;
437 MP_SIGN(r) = MP_ZPOS;
438 MP_USED(r) = 3;
439 s_mp_clamp(r);
440
441 CLEANUP:
442 return res;
443}
444
445#endif
446
447/* Compute the square of polynomial a, reduce modulo p192. Store the
448 * result in r. r could be a. Uses optimized modular reduction for p192.
449 */
450mp_err
451ec_GFp_nistp192_sqr(const mp_int *a, mp_int *r, const GFMethod *meth)
452{
453 mp_err res = MP_OKAY;
454
455 MP_CHECKOK(mp_sqr(a, r));
456 MP_CHECKOK(ec_GFp_nistp192_mod(r, r, meth));
457 CLEANUP:
458 return res;
459}
460
461/* Compute the product of two polynomials a and b, reduce modulo p192.
462 * Store the result in r. r could be a or b; a could be b. Uses
463 * optimized modular reduction for p192. */
464mp_err
465ec_GFp_nistp192_mul(const mp_int *a, const mp_int *b, mp_int *r,
466 const GFMethod *meth)
467{
468 mp_err res = MP_OKAY;
469
470 MP_CHECKOK(mp_mul(a, b, r));
471 MP_CHECKOK(ec_GFp_nistp192_mod(r, r, meth));
472 CLEANUP:
473 return res;
474}
475
476/* Divides two field elements. If a is NULL, then returns the inverse of
477 * b. */
478mp_err
479ec_GFp_nistp192_div(const mp_int *a, const mp_int *b, mp_int *r,
480 const GFMethod *meth)
481{
482 mp_err res = MP_OKAY;
483 mp_int t;
484
485 /* If a is NULL, then return the inverse of b, otherwise return a/b. */
486 if (a == NULL) {
487 return mp_invmod(b, &meth->irr, r);
488 } else {
489 /* MPI doesn't support divmod, so we implement it using invmod and
490 * mulmod. */
491 MP_CHECKOK(mp_init(&t, FLAG(b)));
492 MP_CHECKOK(mp_invmod(b, &meth->irr, &t));
493 MP_CHECKOK(mp_mul(a, &t, r));
494 MP_CHECKOK(ec_GFp_nistp192_mod(r, r, meth));
495 CLEANUP:
496 mp_clear(&t);
497 return res;
498 }
499}
500
501/* Wire in fast field arithmetic and precomputation of base point for
502 * named curves. */
503mp_err
504ec_group_set_gfp192(ECGroup *group, ECCurveName name)
505{
506 if (name == ECCurve_NIST_P192) {
507 group->meth->field_mod = &ec_GFp_nistp192_mod;
508 group->meth->field_mul = &ec_GFp_nistp192_mul;
509 group->meth->field_sqr = &ec_GFp_nistp192_sqr;
510 group->meth->field_div = &ec_GFp_nistp192_div;
511#ifndef ECL_THIRTY_TWO_BIT
512 group->meth->field_add = &ec_GFp_nistp192_add;
513 group->meth->field_sub = &ec_GFp_nistp192_sub;
514#endif
515 }
516 return MP_OKAY;
517}
518