1: /*
2: * Copyright (c) 2007, Cameron Rich
3: *
4: * All rights reserved.
5: *
6: * Redistribution and use in source and binary forms, with or without
7: * modification, are permitted provided that the following conditions are met:
8: *
9: * * Redistributions of source code must retain the above copyright notice,
10: * this list of conditions and the following disclaimer.
11: * * Redistributions in binary form must reproduce the above copyright notice,
12: * this list of conditions and the following disclaimer in the documentation
13: * and/or other materials provided with the distribution.
14: * * Neither the name of the axTLS project nor the names of its contributors
15: * may be used to endorse or promote products derived from this software
16: * without specific prior written permission.
17: *
18: * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19: * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
20: * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
21: * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
22: * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
23: * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
24: * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
25: * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
26: * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
27: * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
28: * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29: */
30:
31: /**
32: * @defgroup bigint_api Big Integer API
33: * @brief The bigint implementation as used by the axTLS project.
34: *
35: * The bigint library is for RSA encryption/decryption as well as signing.
36: * This code tries to minimise use of malloc/free by maintaining a small
37: * cache. A bigint context may maintain state by being made "permanent".
38: * It be be later released with a bi_depermanent() and bi_free() call.
39: *
40: * It supports the following reduction techniques:
41: * - Classical
42: * - Barrett
43: * - Montgomery
44: *
45: * It also implements the following:
46: * - Karatsuba multiplication
47: * - Squaring
48: * - Sliding window exponentiation
49: * - Chinese Remainder Theorem (implemented in rsa.c).
50: *
51: * All the algorithms used are pretty standard, and designed for different
52: * data bus sizes. Negative numbers are not dealt with at all, so a subtraction
53: * may need to be tested for negativity.
54: *
55: * This library steals some ideas from Jef Poskanzer
56: * <http://cs.marlboro.edu/term/cs-fall02/algorithms/crypto/RSA/bigint>
57: * and GMP <http://www.swox.com/gmp>. It gets most of its implementation
58: * detail from "The Handbook of Applied Cryptography"
59: * <http://www.cacr.math.uwaterloo.ca/hac/about/chap14.pdf>
60: * @{
61: */
62:
63: #include <stdlib.h>
64: #include <limits.h>
65: #include <string.h>
66: #include <stdio.h>
67: #include <time.h>
68: #include "os_port.h"
69: #include "bigint.h"
70:
71: #define V1 v->comps[v->size-1] /**< v1 for division */
72: #define V2 v->comps[v->size-2] /**< v2 for division */
73: #define U(j) tmp_u->comps[tmp_u->size-j-1] /**< uj for division */
74: #define Q(j) quotient->comps[quotient->size-j-1] /**< qj for division */
75:
76: static bigint *bi_int_multiply(BI_CTX *ctx, bigint *bi, comp i);
77: static bigint *bi_int_divide(BI_CTX *ctx, bigint *biR, comp denom);
78: static bigint *alloc(BI_CTX *ctx, int size);
79: static bigint *trim(bigint *bi);
80: static void more_comps(bigint *bi, int n);
81: #if defined(CONFIG_BIGINT_KARATSUBA) || defined(CONFIG_BIGINT_BARRETT) || \
82: defined(CONFIG_BIGINT_MONTGOMERY)
83: static bigint *comp_right_shift(bigint *biR, int num_shifts);
84: static bigint *comp_left_shift(bigint *biR, int num_shifts);
85: #endif
86:
87: #ifdef CONFIG_BIGINT_CHECK_ON
88: static void check(const bigint *bi);
89: #else
90: #define check(A) /**< disappears in normal production mode */
91: #endif
92:
93:
94: /**
95: * @brief Start a new bigint context.
96: * @return A bigint context.
97: */
98: BI_CTX *bi_initialize(void)
99: {
100: /* calloc() sets everything to zero */
101: BI_CTX *ctx = (BI_CTX *)calloc(1, sizeof(BI_CTX));
102:
103: /* the radix */
104: ctx->bi_radix = alloc(ctx, 2);
105: ctx->bi_radix->comps[0] = 0;
106: ctx->bi_radix->comps[1] = 1;
107: bi_permanent(ctx->bi_radix);
108: return ctx;
109: }
110:
111: /**
112: * @brief Close the bigint context and free any resources.
113: *
114: * Free up any used memory - a check is done if all objects were not
115: * properly freed.
116: * @param ctx [in] The bigint session context.
117: */
118: void bi_terminate(BI_CTX *ctx)
119: {
120: bi_depermanent(ctx->bi_radix);
121: bi_free(ctx, ctx->bi_radix);
122:
123: if (ctx->active_count != 0)
124: {
125: #ifdef CONFIG_SSL_FULL_MODE
126: printf("bi_terminate: there were %d un-freed bigints\n",
127: ctx->active_count);
128: #endif
129: abort();
130: }
131:
132: bi_clear_cache(ctx);
133: free(ctx);
134: }
135:
136: /**
137: *@brief Clear the memory cache.
138: */
139: void bi_clear_cache(BI_CTX *ctx)
140: {
141: bigint *p, *pn;
142:
143: if (ctx->free_list == NULL)
144: return;
145:
146: for (p = ctx->free_list; p != NULL; p = pn)
147: {
148: pn = p->next;
149: free(p->comps);
150: free(p);
151: }
152:
153: ctx->free_count = 0;
154: ctx->free_list = NULL;
155: }
156:
157: /**
158: * @brief Increment the number of references to this object.
159: * It does not do a full copy.
160: * @param bi [in] The bigint to copy.
161: * @return A reference to the same bigint.
162: */
163: bigint *bi_copy(bigint *bi)
164: {
165: check(bi);
166: if (bi->refs != PERMANENT)
167: bi->refs++;
168: return bi;
169: }
170:
171: /**
172: * @brief Simply make a bigint object "unfreeable" if bi_free() is called on it.
173: *
174: * For this object to be freed, bi_depermanent() must be called.
175: * @param bi [in] The bigint to be made permanent.
176: */
177: void bi_permanent(bigint *bi)
178: {
179: check(bi);
180: if (bi->refs != 1)
181: {
182: #ifdef CONFIG_SSL_FULL_MODE
183: printf("bi_permanent: refs was not 1\n");
184: #endif
185: abort();
186: }
187:
188: bi->refs = PERMANENT;
189: }
190:
191: /**
192: * @brief Take a permanent object and make it eligible for freedom.
193: * @param bi [in] The bigint to be made back to temporary.
194: */
195: void bi_depermanent(bigint *bi)
196: {
197: check(bi);
198: if (bi->refs != PERMANENT)
199: {
200: #ifdef CONFIG_SSL_FULL_MODE
201: printf("bi_depermanent: bigint was not permanent\n");
202: #endif
203: abort();
204: }
205:
206: bi->refs = 1;
207: }
208:
209: /**
210: * @brief Free a bigint object so it can be used again.
211: *
212: * The memory itself it not actually freed, just tagged as being available
213: * @param ctx [in] The bigint session context.
214: * @param bi [in] The bigint to be freed.
215: */
216: void bi_free(BI_CTX *ctx, bigint *bi)
217: {
218: check(bi);
219: if (bi->refs == PERMANENT)
220: {
221: return;
222: }
223:
224: if (--bi->refs > 0)
225: {
226: return;
227: }
228:
229: bi->next = ctx->free_list;
230: ctx->free_list = bi;
231: ctx->free_count++;
232:
233: if (--ctx->active_count < 0)
234: {
235: #ifdef CONFIG_SSL_FULL_MODE
236: printf("bi_free: active_count went negative "
237: "- double-freed bigint?\n");
238: #endif
239: abort();
240: }
241: }
242:
243: /**
244: * @brief Convert an (unsigned) integer into a bigint.
245: * @param ctx [in] The bigint session context.
246: * @param i [in] The (unsigned) integer to be converted.
247: *
248: */
249: bigint *int_to_bi(BI_CTX *ctx, comp i)
250: {
251: bigint *biR = alloc(ctx, 1);
252: biR->comps[0] = i;
253: return biR;
254: }
255:
256: /**
257: * @brief Do a full copy of the bigint object.
258: * @param ctx [in] The bigint session context.
259: * @param bi [in] The bigint object to be copied.
260: */
261: bigint *bi_clone(BI_CTX *ctx, const bigint *bi)
262: {
263: bigint *biR = alloc(ctx, bi->size);
264: check(bi);
265: memcpy(biR->comps, bi->comps, bi->size*COMP_BYTE_SIZE);
266: return biR;
267: }
268:
269: /**
270: * @brief Perform an addition operation between two bigints.
271: * @param ctx [in] The bigint session context.
272: * @param bia [in] A bigint.
273: * @param bib [in] Another bigint.
274: * @return The result of the addition.
275: */
276: bigint *bi_add(BI_CTX *ctx, bigint *bia, bigint *bib)
277: {
278: int n;
279: comp carry = 0;
280: comp *pa, *pb;
281:
282: check(bia);
283: check(bib);
284:
285: n = max(bia->size, bib->size);
286: more_comps(bia, n+1);
287: more_comps(bib, n);
288: pa = bia->comps;
289: pb = bib->comps;
290:
291: do
292: {
293: comp sl, rl, cy1;
294: sl = *pa + *pb++;
295: rl = sl + carry;
296: cy1 = sl < *pa;
297: carry = cy1 | (rl < sl);
298: *pa++ = rl;
299: } while (--n != 0);
300:
301: *pa = carry; /* do overflow */
302: bi_free(ctx, bib);
303: return trim(bia);
304: }
305:
306: /**
307: * @brief Perform a subtraction operation between two bigints.
308: * @param ctx [in] The bigint session context.
309: * @param bia [in] A bigint.
310: * @param bib [in] Another bigint.
311: * @param is_negative [out] If defined, indicates that the result was negative.
312: * is_negative may be null.
313: * @return The result of the subtraction. The result is always positive.
314: */
315: bigint *bi_subtract(BI_CTX *ctx,
316: bigint *bia, bigint *bib, int *is_negative)
317: {
318: int n = bia->size;
319: comp *pa, *pb, carry = 0;
320:
321: check(bia);
322: check(bib);
323:
324: more_comps(bib, n);
325: pa = bia->comps;
326: pb = bib->comps;
327:
328: do
329: {
330: comp sl, rl, cy1;
331: sl = *pa - *pb++;
332: rl = sl - carry;
333: cy1 = sl > *pa;
334: carry = cy1 | (rl > sl);
335: *pa++ = rl;
336: } while (--n != 0);
337:
338: if (is_negative) /* indicate a negative result */
339: {
340: *is_negative = carry;
341: }
342:
343: bi_free(ctx, trim(bib)); /* put bib back to the way it was */
344: return trim(bia);
345: }
346:
347: /**
348: * Perform a multiply between a bigint an an (unsigned) integer
349: */
350: static bigint *bi_int_multiply(BI_CTX *ctx, bigint *bia, comp b)
351: {
352: int j = 0, n = bia->size;
353: bigint *biR = alloc(ctx, n + 1);
354: comp carry = 0;
355: comp *r = biR->comps;
356: comp *a = bia->comps;
357:
358: check(bia);
359:
360: /* clear things to start with */
361: memset(r, 0, ((n+1)*COMP_BYTE_SIZE));
362:
363: do
364: {
365: long_comp tmp = *r + (long_comp)a[j]*b + carry;
366: *r++ = (comp)tmp; /* downsize */
367: carry = (comp)(tmp >> COMP_BIT_SIZE);
368: } while (++j < n);
369:
370: *r = carry;
371: bi_free(ctx, bia);
372: return trim(biR);
373: }
374:
375: /**
376: * @brief Does both division and modulo calculations.
377: *
378: * Used extensively when doing classical reduction.
379: * @param ctx [in] The bigint session context.
380: * @param u [in] A bigint which is the numerator.
381: * @param v [in] Either the denominator or the modulus depending on the mode.
382: * @param is_mod [n] Determines if this is a normal division (0) or a reduction
383: * (1).
384: * @return The result of the division/reduction.
385: */
386: bigint *bi_divide(BI_CTX *ctx, bigint *u, bigint *v, int is_mod)
387: {
388: int n = v->size, m = u->size-n;
389: int j = 0, orig_u_size = u->size;
390: uint8_t mod_offset = ctx->mod_offset;
391: comp d;
392: bigint *quotient, *tmp_u;
393: comp q_dash;
394:
395: check(u);
396: check(v);
397:
398: /* if doing reduction and we are < mod, then return mod */
399: if (is_mod && bi_compare(v, u) > 0)
400: {
401: bi_free(ctx, v);
402: return u;
403: }
404:
405: quotient = alloc(ctx, m+1);
406: tmp_u = alloc(ctx, n+1);
407: v = trim(v); /* make sure we have no leading 0's */
408: d = (comp)((long_comp)COMP_RADIX/(V1+1));
409:
410: /* clear things to start with */
411: memset(quotient->comps, 0, ((quotient->size)*COMP_BYTE_SIZE));
412:
413: /* normalise */
414: if (d > 1)
415: {
416: u = bi_int_multiply(ctx, u, d);
417:
418: if (is_mod)
419: {
420: v = ctx->bi_normalised_mod[mod_offset];
421: }
422: else
423: {
424: v = bi_int_multiply(ctx, v, d);
425: }
426: }
427:
428: if (orig_u_size == u->size) /* new digit position u0 */
429: {
430: more_comps(u, orig_u_size + 1);
431: }
432:
433: do
434: {
435: /* get a temporary short version of u */
436: memcpy(tmp_u->comps, &u->comps[u->size-n-1-j], (n+1)*COMP_BYTE_SIZE);
437:
438: /* calculate q' */
439: if (U(0) == V1)
440: {
441: q_dash = COMP_RADIX-1;
442: }
443: else
444: {
445: q_dash = (comp)(((long_comp)U(0)*COMP_RADIX + U(1))/V1);
446:
447: if (v->size > 1 && V2)
448: {
449: /* we are implementing the following:
450: if (V2*q_dash > (((U(0)*COMP_RADIX + U(1) -
451: q_dash*V1)*COMP_RADIX) + U(2))) ... */
452: comp inner = (comp)((long_comp)COMP_RADIX*U(0) + U(1) -
453: (long_comp)q_dash*V1);
454: if ((long_comp)V2*q_dash > (long_comp)inner*COMP_RADIX + U(2))
455: {
456: q_dash--;
457: }
458: }
459: }
460:
461: /* multiply and subtract */
462: if (q_dash)
463: {
464: int is_negative;
465: tmp_u = bi_subtract(ctx, tmp_u,
466: bi_int_multiply(ctx, bi_copy(v), q_dash), &is_negative);
467: more_comps(tmp_u, n+1);
468:
469: Q(j) = q_dash;
470:
471: /* add back */
472: if (is_negative)
473: {
474: Q(j)--;
475: tmp_u = bi_add(ctx, tmp_u, bi_copy(v));
476:
477: /* lop off the carry */
478: tmp_u->size--;
479: v->size--;
480: }
481: }
482: else
483: {
484: Q(j) = 0;
485: }
486:
487: /* copy back to u */
488: memcpy(&u->comps[u->size-n-1-j], tmp_u->comps, (n+1)*COMP_BYTE_SIZE);
489: } while (++j <= m);
490:
491: bi_free(ctx, tmp_u);
492: bi_free(ctx, v);
493:
494: if (is_mod) /* get the remainder */
495: {
496: bi_free(ctx, quotient);
497: return bi_int_divide(ctx, trim(u), d);
498: }
499: else /* get the quotient */
500: {
501: bi_free(ctx, u);
502: return trim(quotient);
503: }
504: }
505:
506: /*
507: * Perform an integer divide on a bigint.
508: */
509: static bigint *bi_int_divide(BI_CTX *ctx, bigint *biR, comp denom)
510: {
511: int i = biR->size - 1;
512: long_comp r = 0;
513:
514: check(biR);
515:
516: do
517: {
518: r = (r<<COMP_BIT_SIZE) + biR->comps[i];
519: biR->comps[i] = (comp)(r / denom);
520: r %= denom;
521: } while (--i >= 0);
522:
523: return trim(biR);
524: }
525:
526: #ifdef CONFIG_BIGINT_MONTGOMERY
527: /**
528: * There is a need for the value of integer N' such that B^-1(B-1)-N^-1N'=1,
529: * where B^-1(B-1) mod N=1. Actually, only the least significant part of
530: * N' is needed, hence the definition N0'=N' mod b. We reproduce below the
531: * simple algorithm from an article by Dusse and Kaliski to efficiently
532: * find N0' from N0 and b */
533: static comp modular_inverse(bigint *bim)
534: {
535: int i;
536: comp t = 1;
537: comp two_2_i_minus_1 = 2; /* 2^(i-1) */
538: long_comp two_2_i = 4; /* 2^i */
539: comp N = bim->comps[0];
540:
541: for (i = 2; i <= COMP_BIT_SIZE; i++)
542: {
543: if ((long_comp)N*t%two_2_i >= two_2_i_minus_1)
544: {
545: t += two_2_i_minus_1;
546: }
547:
548: two_2_i_minus_1 <<= 1;
549: two_2_i <<= 1;
550: }
551:
552: return (comp)(COMP_RADIX-t);
553: }
554: #endif
555:
556: #if defined(CONFIG_BIGINT_KARATSUBA) || defined(CONFIG_BIGINT_BARRETT) || \
557: defined(CONFIG_BIGINT_MONTGOMERY)
558: /**
559: * Take each component and shift down (in terms of components)
560: */
561: static bigint *comp_right_shift(bigint *biR, int num_shifts)
562: {
563: int i = biR->size-num_shifts;
564: comp *x = biR->comps;
565: comp *y = &biR->comps[num_shifts];
566:
567: check(biR);
568:
569: if (i <= 0) /* have we completely right shifted? */
570: {
571: biR->comps[0] = 0; /* return 0 */
572: biR->size = 1;
573: return biR;
574: }
575:
576: do
577: {
578: *x++ = *y++;
579: } while (--i > 0);
580:
581: biR->size -= num_shifts;
582: return biR;
583: }
584:
585: /**
586: * Take each component and shift it up (in terms of components)
587: */
588: static bigint *comp_left_shift(bigint *biR, int num_shifts)
589: {
590: int i = biR->size-1;
591: comp *x, *y;
592:
593: check(biR);
594:
595: if (num_shifts <= 0)
596: {
597: return biR;
598: }
599:
600: more_comps(biR, biR->size + num_shifts);
601:
602: x = &biR->comps[i+num_shifts];
603: y = &biR->comps[i];
604:
605: do
606: {
607: *x-- = *y--;
608: } while (i--);
609:
610: memset(biR->comps, 0, num_shifts*COMP_BYTE_SIZE); /* zero LS comps */
611: return biR;
612: }
613: #endif
614:
615: /**
616: * @brief Allow a binary sequence to be imported as a bigint.
617: * @param ctx [in] The bigint session context.
618: * @param data [in] The data to be converted.
619: * @param size [in] The number of bytes of data.
620: * @return A bigint representing this data.
621: */
622: bigint *bi_import(BI_CTX *ctx, const uint8_t *data, int size)
623: {
624: bigint *biR = alloc(ctx, (size+COMP_BYTE_SIZE-1)/COMP_BYTE_SIZE);
625: int i, j = 0, offset = 0;
626:
627: memset(biR->comps, 0, biR->size*COMP_BYTE_SIZE);
628:
629: for (i = size-1; i >= 0; i--)
630: {
631: biR->comps[offset] += data[i] << (j*8);
632:
633: if (++j == COMP_BYTE_SIZE)
634: {
635: j = 0;
636: offset ++;
637: }
638: }
639:
640: return trim(biR);
641: }
642:
643: #ifdef CONFIG_SSL_FULL_MODE
644: /**
645: * @brief The testharness uses this code to import text hex-streams and
646: * convert them into bigints.
647: * @param ctx [in] The bigint session context.
648: * @param data [in] A string consisting of hex characters. The characters must
649: * be in upper case.
650: * @return A bigint representing this data.
651: */
652: bigint *bi_str_import(BI_CTX *ctx, const char *data)
653: {
654: int size = strlen(data);
655: bigint *biR = alloc(ctx, (size+COMP_NUM_NIBBLES-1)/COMP_NUM_NIBBLES);
656: int i, j = 0, offset = 0;
657: memset(biR->comps, 0, biR->size*COMP_BYTE_SIZE);
658:
659: for (i = size-1; i >= 0; i--)
660: {
661: int num = (data[i] <= '9') ? (data[i] - '0') : (data[i] - 'A' + 10);
662: biR->comps[offset] += num << (j*4);
663:
664: if (++j == COMP_NUM_NIBBLES)
665: {
666: j = 0;
667: offset ++;
668: }
669: }
670:
671: return biR;
672: }
673:
674: void bi_print(const char *label, bigint *x)
675: {
676: int i, j;
677:
678: if (x == NULL)
679: {
680: printf("%s: (null)\n", label);
681: return;
682: }
683:
684: printf("%s: (size %d)\n", label, x->size);
685: for (i = x->size-1; i >= 0; i--)
686: {
687: for (j = COMP_NUM_NIBBLES-1; j >= 0; j--)
688: {
689: comp mask = 0x0f << (j*4);
690: comp num = (x->comps[i] & mask) >> (j*4);
691: putc((num <= 9) ? (num + '0') : (num + 'A' - 10), stdout);
692: }
693: }
694:
695: printf("\n");
696: }
697: #endif
698:
699: /**
700: * @brief Take a bigint and convert it into a byte sequence.
701: *
702: * This is useful after a decrypt operation.
703: * @param ctx [in] The bigint session context.
704: * @param x [in] The bigint to be converted.
705: * @param data [out] The converted data as a byte stream.
706: * @param size [in] The maximum size of the byte stream. Unused bytes will be
707: * zeroed.
708: */
709: void bi_export(BI_CTX *ctx, bigint *x, uint8_t *data, int size)
710: {
711: int i, j, k = size-1;
712:
713: check(x);
714: memset(data, 0, size); /* ensure all leading 0's are cleared */
715:
716: for (i = 0; i < x->size; i++)
717: {
718: for (j = 0; j < COMP_BYTE_SIZE; j++)
719: {
720: comp mask = 0xff << (j*8);
721: int num = (x->comps[i] & mask) >> (j*8);
722: data[k--] = num;
723:
724: if (k < 0)
725: {
726: goto buf_done;
727: }
728: }
729: }
730: buf_done:
731:
732: bi_free(ctx, x);
733: }
734:
735: /**
736: * @brief Pre-calculate some of the expensive steps in reduction.
737: *
738: * This function should only be called once (normally when a session starts).
739: * When the session is over, bi_free_mod() should be called. bi_mod_power()
740: * relies on this function being called.
741: * @param ctx [in] The bigint session context.
742: * @param bim [in] The bigint modulus that will be used.
743: * @param mod_offset [in] There are three moduluii that can be stored - the
744: * standard modulus, and its two primes p and q. This offset refers to which
745: * modulus we are referring to.
746: * @see bi_free_mod(), bi_mod_power().
747: */
748: void bi_set_mod(BI_CTX *ctx, bigint *bim, int mod_offset)
749: {
750: int k = bim->size;
751: comp d = (comp)((long_comp)COMP_RADIX/(bim->comps[k-1]+1));
752: #ifdef CONFIG_BIGINT_MONTGOMERY
753: bigint *R, *R2;
754: #endif
755:
756: ctx->bi_mod[mod_offset] = bim;
757: bi_permanent(ctx->bi_mod[mod_offset]);
758: ctx->bi_normalised_mod[mod_offset] = bi_int_multiply(ctx, bim, d);
759: bi_permanent(ctx->bi_normalised_mod[mod_offset]);
760:
761: #if defined(CONFIG_BIGINT_MONTGOMERY)
762: /* set montgomery variables */
763: R = comp_left_shift(bi_clone(ctx, ctx->bi_radix), k-1); /* R */
764: R2 = comp_left_shift(bi_clone(ctx, ctx->bi_radix), k*2-1); /* R^2 */
765: ctx->bi_RR_mod_m[mod_offset] = bi_mod(ctx, R2); /* R^2 mod m */
766: ctx->bi_R_mod_m[mod_offset] = bi_mod(ctx, R); /* R mod m */
767:
768: bi_permanent(ctx->bi_RR_mod_m[mod_offset]);
769: bi_permanent(ctx->bi_R_mod_m[mod_offset]);
770:
771: ctx->N0_dash[mod_offset] = modular_inverse(ctx->bi_mod[mod_offset]);
772:
773: #elif defined (CONFIG_BIGINT_BARRETT)
774: ctx->bi_mu[mod_offset] =
775: bi_divide(ctx, comp_left_shift(
776: bi_clone(ctx, ctx->bi_radix), k*2-1), ctx->bi_mod[mod_offset], 0);
777: bi_permanent(ctx->bi_mu[mod_offset]);
778: #endif
779: }
780:
781: /**
782: * @brief Used when cleaning various bigints at the end of a session.
783: * @param ctx [in] The bigint session context.
784: * @param mod_offset [in] The offset to use.
785: * @see bi_set_mod().
786: */
787: void bi_free_mod(BI_CTX *ctx, int mod_offset)
788: {
789: bi_depermanent(ctx->bi_mod[mod_offset]);
790: bi_free(ctx, ctx->bi_mod[mod_offset]);
791: #if defined (CONFIG_BIGINT_MONTGOMERY)
792: bi_depermanent(ctx->bi_RR_mod_m[mod_offset]);
793: bi_depermanent(ctx->bi_R_mod_m[mod_offset]);
794: bi_free(ctx, ctx->bi_RR_mod_m[mod_offset]);
795: bi_free(ctx, ctx->bi_R_mod_m[mod_offset]);
796: #elif defined(CONFIG_BIGINT_BARRETT)
797: bi_depermanent(ctx->bi_mu[mod_offset]);
798: bi_free(ctx, ctx->bi_mu[mod_offset]);
799: #endif
800: bi_depermanent(ctx->bi_normalised_mod[mod_offset]);
801: bi_free(ctx, ctx->bi_normalised_mod[mod_offset]);
802: }
803:
804: /**
805: * Perform a standard multiplication between two bigints.
806: *
807: * Barrett reduction has no need for some parts of the product, so ignore bits
808: * of the multiply. This routine gives Barrett its big performance
809: * improvements over Classical/Montgomery reduction methods.
810: */
811: static bigint *regular_multiply(BI_CTX *ctx, bigint *bia, bigint *bib,
812: int inner_partial, int outer_partial)
813: {
814: int i = 0, j;
815: int n = bia->size;
816: int t = bib->size;
817: bigint *biR = alloc(ctx, n + t);
818: comp *sr = biR->comps;
819: comp *sa = bia->comps;
820: comp *sb = bib->comps;
821:
822: check(bia);
823: check(bib);
824:
825: /* clear things to start with */
826: memset(biR->comps, 0, ((n+t)*COMP_BYTE_SIZE));
827:
828: do
829: {
830: long_comp tmp;
831: comp carry = 0;
832: int r_index = i;
833: j = 0;
834:
835: if (outer_partial && outer_partial-i > 0 && outer_partial < n)
836: {
837: r_index = outer_partial-1;
838: j = outer_partial-i-1;
839: }
840:
841: do
842: {
843: if (inner_partial && r_index >= inner_partial)
844: {
845: break;
846: }
847:
848: tmp = sr[r_index] + ((long_comp)sa[j])*sb[i] + carry;
849: sr[r_index++] = (comp)tmp; /* downsize */
850: carry = tmp >> COMP_BIT_SIZE;
851: } while (++j < n);
852:
853: sr[r_index] = carry;
854: } while (++i < t);
855:
856: bi_free(ctx, bia);
857: bi_free(ctx, bib);
858: return trim(biR);
859: }
860:
861: #ifdef CONFIG_BIGINT_KARATSUBA
862: /*
863: * Karatsuba improves on regular multiplication due to only 3 multiplications
864: * being done instead of 4. The additional additions/subtractions are O(N)
865: * rather than O(N^2) and so for big numbers it saves on a few operations
866: */
867: static bigint *karatsuba(BI_CTX *ctx, bigint *bia, bigint *bib, int is_square)
868: {
869: bigint *x0, *x1;
870: bigint *p0, *p1, *p2;
871: int m;
872:
873: if (is_square)
874: {
875: m = (bia->size + 1)/2;
876: }
877: else
878: {
879: m = (max(bia->size, bib->size) + 1)/2;
880: }
881:
882: x0 = bi_clone(ctx, bia);
883: x0->size = m;
884: x1 = bi_clone(ctx, bia);
885: comp_right_shift(x1, m);
886: bi_free(ctx, bia);
887:
888: /* work out the 3 partial products */
889: if (is_square)
890: {
891: p0 = bi_square(ctx, bi_copy(x0));
892: p2 = bi_square(ctx, bi_copy(x1));
893: p1 = bi_square(ctx, bi_add(ctx, x0, x1));
894: }
895: else /* normal multiply */
896: {
897: bigint *y0, *y1;
898: y0 = bi_clone(ctx, bib);
899: y0->size = m;
900: y1 = bi_clone(ctx, bib);
901: comp_right_shift(y1, m);
902: bi_free(ctx, bib);
903:
904: p0 = bi_multiply(ctx, bi_copy(x0), bi_copy(y0));
905: p2 = bi_multiply(ctx, bi_copy(x1), bi_copy(y1));
906: p1 = bi_multiply(ctx, bi_add(ctx, x0, x1), bi_add(ctx, y0, y1));
907: }
908:
909: p1 = bi_subtract(ctx,
910: bi_subtract(ctx, p1, bi_copy(p2), NULL), bi_copy(p0), NULL);
911:
912: comp_left_shift(p1, m);
913: comp_left_shift(p2, 2*m);
914: return bi_add(ctx, p1, bi_add(ctx, p0, p2));
915: }
916: #endif
917:
918: /**
919: * @brief Perform a multiplication operation between two bigints.
920: * @param ctx [in] The bigint session context.
921: * @param bia [in] A bigint.
922: * @param bib [in] Another bigint.
923: * @return The result of the multiplication.
924: */
925: bigint *bi_multiply(BI_CTX *ctx, bigint *bia, bigint *bib)
926: {
927: check(bia);
928: check(bib);
929:
930: #ifdef CONFIG_BIGINT_KARATSUBA
931: if (min(bia->size, bib->size) < MUL_KARATSUBA_THRESH)
932: {
933: return regular_multiply(ctx, bia, bib, 0, 0);
934: }
935:
936: return karatsuba(ctx, bia, bib, 0);
937: #else
938: return regular_multiply(ctx, bia, bib, 0, 0);
939: #endif
940: }
941:
942: #ifdef CONFIG_BIGINT_SQUARE
943: /*
944: * Perform the actual square operion. It takes into account overflow.
945: */
946: static bigint *regular_square(BI_CTX *ctx, bigint *bi)
947: {
948: int t = bi->size;
949: int i = 0, j;
950: bigint *biR = alloc(ctx, t*2+1);
951: comp *w = biR->comps;
952: comp *x = bi->comps;
953: long_comp carry;
954: memset(w, 0, biR->size*COMP_BYTE_SIZE);
955:
956: do
957: {
958: long_comp tmp = w[2*i] + (long_comp)x[i]*x[i];
959: w[2*i] = (comp)tmp;
960: carry = tmp >> COMP_BIT_SIZE;
961:
962: for (j = i+1; j < t; j++)
963: {
964: uint8_t c = 0;
965: long_comp xx = (long_comp)x[i]*x[j];
966: if ((COMP_MAX-xx) < xx)
967: c = 1;
968:
969: tmp = (xx<<1);
970:
971: if ((COMP_MAX-tmp) < w[i+j])
972: c = 1;
973:
974: tmp += w[i+j];
975:
976: if ((COMP_MAX-tmp) < carry)
977: c = 1;
978:
979: tmp += carry;
980: w[i+j] = (comp)tmp;
981: carry = tmp >> COMP_BIT_SIZE;
982:
983: if (c)
984: carry += COMP_RADIX;
985: }
986:
987: tmp = w[i+t] + carry;
988: w[i+t] = (comp)tmp;
989: w[i+t+1] = tmp >> COMP_BIT_SIZE;
990: } while (++i < t);
991:
992: bi_free(ctx, bi);
993: return trim(biR);
994: }
995:
996: /**
997: * @brief Perform a square operation on a bigint.
998: * @param ctx [in] The bigint session context.
999: * @param bia [in] A bigint.
1000: * @return The result of the multiplication.
1001: */
1002: bigint *bi_square(BI_CTX *ctx, bigint *bia)
1003: {
1004: check(bia);
1005:
1006: #ifdef CONFIG_BIGINT_KARATSUBA
1007: if (bia->size < SQU_KARATSUBA_THRESH)
1008: {
1009: return regular_square(ctx, bia);
1010: }
1011:
1012: return karatsuba(ctx, bia, NULL, 1);
1013: #else
1014: return regular_square(ctx, bia);
1015: #endif
1016: }
1017: #endif
1018:
1019: /**
1020: * @brief Compare two bigints.
1021: * @param bia [in] A bigint.
1022: * @param bib [in] Another bigint.
1023: * @return -1 if smaller, 1 if larger and 0 if equal.
1024: */
1025: int bi_compare(bigint *bia, bigint *bib)
1026: {
1027: int r, i;
1028:
1029: check(bia);
1030: check(bib);
1031:
1032: if (bia->size > bib->size)
1033: r = 1;
1034: else if (bia->size < bib->size)
1035: r = -1;
1036: else
1037: {
1038: comp *a = bia->comps;
1039: comp *b = bib->comps;
1040:
1041: /* Same number of components. Compare starting from the high end
1042: * and working down. */
1043: r = 0;
1044: i = bia->size - 1;
1045:
1046: do
1047: {
1048: if (a[i] > b[i])
1049: {
1050: r = 1;
1051: break;
1052: }
1053: else if (a[i] < b[i])
1054: {
1055: r = -1;
1056: break;
1057: }
1058: } while (--i >= 0);
1059: }
1060:
1061: return r;
1062: }
1063:
1064: /*
1065: * Allocate and zero more components. Does not consume bi.
1066: */
1067: static void more_comps(bigint *bi, int n)
1068: {
1069: if (n > bi->max_comps)
1070: {
1071: bi->max_comps = max(bi->max_comps * 2, n);
1072: bi->comps = (comp*)realloc(bi->comps, bi->max_comps * COMP_BYTE_SIZE);
1073: }
1074:
1075: if (n > bi->size)
1076: {
1077: memset(&bi->comps[bi->size], 0, (n-bi->size)*COMP_BYTE_SIZE);
1078: }
1079:
1080: bi->size = n;
1081: }
1082:
1083: /*
1084: * Make a new empty bigint. It may just use an old one if one is available.
1085: * Otherwise get one off the heap.
1086: */
1087: static bigint *alloc(BI_CTX *ctx, int size)
1088: {
1089: bigint *biR;
1090:
1091: /* Can we recycle an old bigint? */
1092: if (ctx->free_list != NULL)
1093: {
1094: biR = ctx->free_list;
1095: ctx->free_list = biR->next;
1096: ctx->free_count--;
1097:
1098: if (biR->refs != 0)
1099: {
1100: #ifdef CONFIG_SSL_FULL_MODE
1101: printf("alloc: refs was not 0\n");
1102: #endif
1103: abort(); /* create a stack trace from a core dump */
1104: }
1105:
1106: more_comps(biR, size);
1107: }
1108: else
1109: {
1110: /* No free bigints available - create a new one. */
1111: biR = (bigint *)malloc(sizeof(bigint));
1112: biR->comps = (comp*)malloc(size * COMP_BYTE_SIZE);
1113: biR->max_comps = size; /* give some space to spare */
1114: }
1115:
1116: biR->size = size;
1117: biR->refs = 1;
1118: biR->next = NULL;
1119: ctx->active_count++;
1120: return biR;
1121: }
1122:
1123: /*
1124: * Work out the highest '1' bit in an exponent. Used when doing sliding-window
1125: * exponentiation.
1126: */
1127: static int find_max_exp_index(bigint *biexp)
1128: {
1129: int i = COMP_BIT_SIZE-1;
1130: comp shift = COMP_RADIX/2;
1131: comp test = biexp->comps[biexp->size-1]; /* assume no leading zeroes */
1132:
1133: check(biexp);
1134:
1135: do
1136: {
1137: if (test & shift)
1138: {
1139: return i+(biexp->size-1)*COMP_BIT_SIZE;
1140: }
1141:
1142: shift >>= 1;
1143: } while (i-- != 0);
1144:
1145: return -1; /* error - must have been a leading 0 */
1146: }
1147:
1148: /*
1149: * Is a particular bit is an exponent 1 or 0? Used when doing sliding-window
1150: * exponentiation.
1151: */
1152: static int exp_bit_is_one(bigint *biexp, int offset)
1153: {
1154: comp test = biexp->comps[offset / COMP_BIT_SIZE];
1155: int num_shifts = offset % COMP_BIT_SIZE;
1156: comp shift = 1;
1157: int i;
1158:
1159: check(biexp);
1160:
1161: for (i = 0; i < num_shifts; i++)
1162: {
1163: shift <<= 1;
1164: }
1165:
1166: return (test & shift) != 0;
1167: }
1168:
1169: #ifdef CONFIG_BIGINT_CHECK_ON
1170: /*
1171: * Perform a sanity check on bi.
1172: */
1173: static void check(const bigint *bi)
1174: {
1175: if (bi->refs <= 0)
1176: {
1177: printf("check: zero or negative refs in bigint\n");
1178: abort();
1179: }
1180:
1181: if (bi->next != NULL)
1182: {
1183: printf("check: attempt to use a bigint from "
1184: "the free list\n");
1185: abort();
1186: }
1187: }
1188: #endif
1189:
1190: /*
1191: * Delete any leading 0's (and allow for 0).
1192: */
1193: static bigint *trim(bigint *bi)
1194: {
1195: check(bi);
1196:
1197: while (bi->comps[bi->size-1] == 0 && bi->size > 1)
1198: {
1199: bi->size--;
1200: }
1201:
1202: return bi;
1203: }
1204:
1205: #if defined(CONFIG_BIGINT_MONTGOMERY)
1206: /**
1207: * @brief Perform a single montgomery reduction.
1208: * @param ctx [in] The bigint session context.
1209: * @param bixy [in] A bigint.
1210: * @return The result of the montgomery reduction.
1211: */
1212: bigint *bi_mont(BI_CTX *ctx, bigint *bixy)
1213: {
1214: int i = 0, n;
1215: uint8_t mod_offset = ctx->mod_offset;
1216: bigint *bim = ctx->bi_mod[mod_offset];
1217: comp mod_inv = ctx->N0_dash[mod_offset];
1218:
1219: check(bixy);
1220:
1221: if (ctx->use_classical) /* just use classical instead */
1222: {
1223: return bi_mod(ctx, bixy);
1224: }
1225:
1226: n = bim->size;
1227:
1228: do
1229: {
1230: bixy = bi_add(ctx, bixy, comp_left_shift(
1231: bi_int_multiply(ctx, bim, bixy->comps[i]*mod_inv), i));
1232: } while (++i < n);
1233:
1234: comp_right_shift(bixy, n);
1235:
1236: if (bi_compare(bixy, bim) >= 0)
1237: {
1238: bixy = bi_subtract(ctx, bixy, bim, NULL);
1239: }
1240:
1241: return bixy;
1242: }
1243:
1244: #elif defined(CONFIG_BIGINT_BARRETT)
1245: /*
1246: * Stomp on the most significant components to give the illusion of a "mod base
1247: * radix" operation
1248: */
1249: static bigint *comp_mod(bigint *bi, int mod)
1250: {
1251: check(bi);
1252:
1253: if (bi->size > mod)
1254: {
1255: bi->size = mod;
1256: }
1257:
1258: return bi;
1259: }
1260:
1261: /**
1262: * @brief Perform a single Barrett reduction.
1263: * @param ctx [in] The bigint session context.
1264: * @param bi [in] A bigint.
1265: * @return The result of the Barrett reduction.
1266: */
1267: bigint *bi_barrett(BI_CTX *ctx, bigint *bi)
1268: {
1269: bigint *q1, *q2, *q3, *r1, *r2, *r;
1270: uint8_t mod_offset = ctx->mod_offset;
1271: bigint *bim = ctx->bi_mod[mod_offset];
1272: int k = bim->size;
1273:
1274: check(bi);
1275: check(bim);
1276:
1277: /* use Classical method instead - Barrett cannot help here */
1278: if (bi->size > k*2)
1279: {
1280: return bi_mod(ctx, bi);
1281: }
1282:
1283: q1 = comp_right_shift(bi_clone(ctx, bi), k-1);
1284:
1285: /* do outer partial multiply */
1286: q2 = regular_multiply(ctx, q1, ctx->bi_mu[mod_offset], 0, k-1);
1287: q3 = comp_right_shift(q2, k+1);
1288: r1 = comp_mod(bi, k+1);
1289:
1290: /* do inner partial multiply */
1291: r2 = comp_mod(regular_multiply(ctx, q3, bim, k+1, 0), k+1);
1292: r = bi_subtract(ctx, r1, r2, NULL);
1293:
1294: /* if (r >= m) r = r - m; */
1295: if (bi_compare(r, bim) >= 0)
1296: {
1297: r = bi_subtract(ctx, r, bim, NULL);
1298: }
1299:
1300: return r;
1301: }
1302: #endif /* CONFIG_BIGINT_BARRETT */
1303:
1304: #ifdef CONFIG_BIGINT_SLIDING_WINDOW
1305: /*
1306: * Work out g1, g3, g5, g7... etc for the sliding-window algorithm
1307: */
1308: static void precompute_slide_window(BI_CTX *ctx, int window, bigint *g1)
1309: {
1310: int k = 1, i;
1311: bigint *g2;
1312:
1313: for (i = 0; i < window-1; i++) /* compute 2^(window-1) */
1314: {
1315: k <<= 1;
1316: }
1317:
1318: ctx->g = (bigint **)malloc(k*sizeof(bigint *));
1319: ctx->g[0] = bi_clone(ctx, g1);
1320: bi_permanent(ctx->g[0]);
1321: g2 = bi_residue(ctx, bi_square(ctx, ctx->g[0])); /* g^2 */
1322:
1323: for (i = 1; i < k; i++)
1324: {
1325: ctx->g[i] = bi_residue(ctx, bi_multiply(ctx, ctx->g[i-1], bi_copy(g2)));
1326: bi_permanent(ctx->g[i]);
1327: }
1328:
1329: bi_free(ctx, g2);
1330: ctx->window = k;
1331: }
1332: #endif
1333:
1334: /**
1335: * @brief Perform a modular exponentiation.
1336: *
1337: * This function requires bi_set_mod() to have been called previously. This is
1338: * one of the optimisations used for performance.
1339: * @param ctx [in] The bigint session context.
1340: * @param bi [in] The bigint on which to perform the mod power operation.
1341: * @param biexp [in] The bigint exponent.
1342: * @return The result of the mod exponentiation operation
1343: * @see bi_set_mod().
1344: */
1345: bigint *bi_mod_power(BI_CTX *ctx, bigint *bi, bigint *biexp)
1346: {
1347: int i = find_max_exp_index(biexp), j, window_size = 1;
1348: bigint *biR = int_to_bi(ctx, 1);
1349:
1350: #if defined(CONFIG_BIGINT_MONTGOMERY)
1351: uint8_t mod_offset = ctx->mod_offset;
1352: if (!ctx->use_classical)
1353: {
1354: /* preconvert */
1355: bi = bi_mont(ctx,
1356: bi_multiply(ctx, bi, ctx->bi_RR_mod_m[mod_offset])); /* x' */
1357: bi_free(ctx, biR);
1358: biR = ctx->bi_R_mod_m[mod_offset]; /* A */
1359: }
1360: #endif
1361:
1362: check(bi);
1363: check(biexp);
1364:
1365: #ifdef CONFIG_BIGINT_SLIDING_WINDOW
1366: for (j = i; j > 32; j /= 5) /* work out an optimum size */
1367: window_size++;
1368:
1369: /* work out the slide constants */
1370: precompute_slide_window(ctx, window_size, bi);
1371: #else /* just one constant */
1372: ctx->g = (bigint **)malloc(sizeof(bigint *));
1373: ctx->g[0] = bi_clone(ctx, bi);
1374: ctx->window = 1;
1375: bi_permanent(ctx->g[0]);
1376: #endif
1377:
1378: /* if sliding-window is off, then only one bit will be done at a time and
1379: * will reduce to standard left-to-right exponentiation */
1380: do
1381: {
1382: if (exp_bit_is_one(biexp, i))
1383: {
1384: int l = i-window_size+1;
1385: int part_exp = 0;
1386:
1387: if (l < 0) /* LSB of exponent will always be 1 */
1388: l = 0;
1389: else
1390: {
1391: while (exp_bit_is_one(biexp, l) == 0)
1392: l++; /* go back up */
1393: }
1394:
1395: /* build up the section of the exponent */
1396: for (j = i; j >= l; j--)
1397: {
1398: biR = bi_residue(ctx, bi_square(ctx, biR));
1399: if (exp_bit_is_one(biexp, j))
1400: part_exp++;
1401:
1402: if (j != l)
1403: part_exp <<= 1;
1404: }
1405:
1406: part_exp = (part_exp-1)/2; /* adjust for array */
1407: biR = bi_residue(ctx, bi_multiply(ctx, biR, ctx->g[part_exp]));
1408: i = l-1;
1409: }
1410: else /* square it */
1411: {
1412: biR = bi_residue(ctx, bi_square(ctx, biR));
1413: i--;
1414: }
1415: } while (i >= 0);
1416:
1417: /* cleanup */
1418: for (i = 0; i < ctx->window; i++)
1419: {
1420: bi_depermanent(ctx->g[i]);
1421: bi_free(ctx, ctx->g[i]);
1422: }
1423:
1424: free(ctx->g);
1425: bi_free(ctx, bi);
1426: bi_free(ctx, biexp);
1427: #if defined CONFIG_BIGINT_MONTGOMERY
1428: return ctx->use_classical ? biR : bi_mont(ctx, biR); /* convert back */
1429: #else /* CONFIG_BIGINT_CLASSICAL or CONFIG_BIGINT_BARRETT */
1430: return biR;
1431: #endif
1432: }
1433:
1434: #ifdef CONFIG_SSL_CERT_VERIFICATION
1435: /**
1436: * @brief Perform a modular exponentiation using a temporary modulus.
1437: *
1438: * We need this function to check the signatures of certificates. The modulus
1439: * of this function is temporary as it's just used for authentication.
1440: * @param ctx [in] The bigint session context.
1441: * @param bi [in] The bigint to perform the exp/mod.
1442: * @param bim [in] The temporary modulus.
1443: * @param biexp [in] The bigint exponent.
1444: * @return The result of the mod exponentiation operation
1445: * @see bi_set_mod().
1446: */
1447: bigint *bi_mod_power2(BI_CTX *ctx, bigint *bi, bigint *bim, bigint *biexp)
1448: {
1449: bigint *biR, *tmp_biR;
1450:
1451: /* Set up a temporary bigint context and transfer what we need between
1452: * them. We need to do this since we want to keep the original modulus
1453: * which is already in this context. This operation is only called when
1454: * doing peer verification, and so is not expensive :-) */
1455: BI_CTX *tmp_ctx = bi_initialize();
1456: bi_set_mod(tmp_ctx, bi_clone(tmp_ctx, bim), BIGINT_M_OFFSET);
1457: tmp_biR = bi_mod_power(tmp_ctx,
1458: bi_clone(tmp_ctx, bi),
1459: bi_clone(tmp_ctx, biexp));
1460: biR = bi_clone(ctx, tmp_biR);
1461: bi_free(tmp_ctx, tmp_biR);
1462: bi_free_mod(tmp_ctx, BIGINT_M_OFFSET);
1463: bi_terminate(tmp_ctx);
1464:
1465: bi_free(ctx, bi);
1466: bi_free(ctx, bim);
1467: bi_free(ctx, biexp);
1468: return biR;
1469: }
1470: #endif
1471:
1472: #ifdef CONFIG_BIGINT_CRT
1473: /**
1474: * @brief Use the Chinese Remainder Theorem to quickly perform RSA decrypts.
1475: *
1476: * @param ctx [in] The bigint session context.
1477: * @param bi [in] The bigint to perform the exp/mod.
1478: * @param dP [in] CRT's dP bigint
1479: * @param dQ [in] CRT's dQ bigint
1480: * @param p [in] CRT's p bigint
1481: * @param q [in] CRT's q bigint
1482: * @param qInv [in] CRT's qInv bigint
1483: * @return The result of the CRT operation
1484: */
1485: bigint *bi_crt(BI_CTX *ctx, bigint *bi,
1486: bigint *dP, bigint *dQ,
1487: bigint *p, bigint *q, bigint *qInv)
1488: {
1489: bigint *m1, *m2, *h;
1490:
1491: /* Montgomery has a condition the 0 < x, y < m and these products violate
1492: * that condition. So disable Montgomery when using CRT */
1493: #if defined(CONFIG_BIGINT_MONTGOMERY)
1494: ctx->use_classical = 1;
1495: #endif
1496: ctx->mod_offset = BIGINT_P_OFFSET;
1497: m1 = bi_mod_power(ctx, bi_copy(bi), dP);
1498:
1499: ctx->mod_offset = BIGINT_Q_OFFSET;
1500: m2 = bi_mod_power(ctx, bi, dQ);
1501:
1502: h = bi_subtract(ctx, bi_add(ctx, m1, p), bi_copy(m2), NULL);
1503: h = bi_multiply(ctx, h, qInv);
1504: ctx->mod_offset = BIGINT_P_OFFSET;
1505: h = bi_residue(ctx, h);
1506: #if defined(CONFIG_BIGINT_MONTGOMERY)
1507: ctx->use_classical = 0; /* reset for any further operation */
1508: #endif
1509: return bi_add(ctx, m2, bi_multiply(ctx, q, h));
1510: }
1511: #endif
1512: /** @} */
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>