security/nss/lib/freebl/mpi/mp_gf2m.c

Wed, 31 Dec 2014 06:09:35 +0100

author
Michael Schloh von Bennewitz <michael@schloh.com>
date
Wed, 31 Dec 2014 06:09:35 +0100
changeset 0
6474c204b198
permissions
-rw-r--r--

Cloned upstream origin tor-browser at tor-browser-31.3.0esr-4.5-1-build1
revision ID fc1c9ff7c1b2defdbc039f12214767608f46423f for hacking purpose.

michael@0 1 /* This Source Code Form is subject to the terms of the Mozilla Public
michael@0 2 * License, v. 2.0. If a copy of the MPL was not distributed with this
michael@0 3 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
michael@0 4
michael@0 5 #include "mp_gf2m.h"
michael@0 6 #include "mp_gf2m-priv.h"
michael@0 7 #include "mplogic.h"
michael@0 8 #include "mpi-priv.h"
michael@0 9
michael@0 10 const mp_digit mp_gf2m_sqr_tb[16] =
michael@0 11 {
michael@0 12 0, 1, 4, 5, 16, 17, 20, 21,
michael@0 13 64, 65, 68, 69, 80, 81, 84, 85
michael@0 14 };
michael@0 15
michael@0 16 /* Multiply two binary polynomials mp_digits a, b.
michael@0 17 * Result is a polynomial with degree < 2 * MP_DIGIT_BITS - 1.
michael@0 18 * Output in two mp_digits rh, rl.
michael@0 19 */
michael@0 20 #if MP_DIGIT_BITS == 32
michael@0 21 void
michael@0 22 s_bmul_1x1(mp_digit *rh, mp_digit *rl, const mp_digit a, const mp_digit b)
michael@0 23 {
michael@0 24 register mp_digit h, l, s;
michael@0 25 mp_digit tab[8], top2b = a >> 30;
michael@0 26 register mp_digit a1, a2, a4;
michael@0 27
michael@0 28 a1 = a & (0x3FFFFFFF); a2 = a1 << 1; a4 = a2 << 1;
michael@0 29
michael@0 30 tab[0] = 0; tab[1] = a1; tab[2] = a2; tab[3] = a1^a2;
michael@0 31 tab[4] = a4; tab[5] = a1^a4; tab[6] = a2^a4; tab[7] = a1^a2^a4;
michael@0 32
michael@0 33 s = tab[b & 0x7]; l = s;
michael@0 34 s = tab[b >> 3 & 0x7]; l ^= s << 3; h = s >> 29;
michael@0 35 s = tab[b >> 6 & 0x7]; l ^= s << 6; h ^= s >> 26;
michael@0 36 s = tab[b >> 9 & 0x7]; l ^= s << 9; h ^= s >> 23;
michael@0 37 s = tab[b >> 12 & 0x7]; l ^= s << 12; h ^= s >> 20;
michael@0 38 s = tab[b >> 15 & 0x7]; l ^= s << 15; h ^= s >> 17;
michael@0 39 s = tab[b >> 18 & 0x7]; l ^= s << 18; h ^= s >> 14;
michael@0 40 s = tab[b >> 21 & 0x7]; l ^= s << 21; h ^= s >> 11;
michael@0 41 s = tab[b >> 24 & 0x7]; l ^= s << 24; h ^= s >> 8;
michael@0 42 s = tab[b >> 27 & 0x7]; l ^= s << 27; h ^= s >> 5;
michael@0 43 s = tab[b >> 30 ]; l ^= s << 30; h ^= s >> 2;
michael@0 44
michael@0 45 /* compensate for the top two bits of a */
michael@0 46
michael@0 47 if (top2b & 01) { l ^= b << 30; h ^= b >> 2; }
michael@0 48 if (top2b & 02) { l ^= b << 31; h ^= b >> 1; }
michael@0 49
michael@0 50 *rh = h; *rl = l;
michael@0 51 }
michael@0 52 #else
michael@0 53 void
michael@0 54 s_bmul_1x1(mp_digit *rh, mp_digit *rl, const mp_digit a, const mp_digit b)
michael@0 55 {
michael@0 56 register mp_digit h, l, s;
michael@0 57 mp_digit tab[16], top3b = a >> 61;
michael@0 58 register mp_digit a1, a2, a4, a8;
michael@0 59
michael@0 60 a1 = a & (0x1FFFFFFFFFFFFFFFULL); a2 = a1 << 1;
michael@0 61 a4 = a2 << 1; a8 = a4 << 1;
michael@0 62 tab[ 0] = 0; tab[ 1] = a1; tab[ 2] = a2; tab[ 3] = a1^a2;
michael@0 63 tab[ 4] = a4; tab[ 5] = a1^a4; tab[ 6] = a2^a4; tab[ 7] = a1^a2^a4;
michael@0 64 tab[ 8] = a8; tab[ 9] = a1^a8; tab[10] = a2^a8; tab[11] = a1^a2^a8;
michael@0 65 tab[12] = a4^a8; tab[13] = a1^a4^a8; tab[14] = a2^a4^a8; tab[15] = a1^a2^a4^a8;
michael@0 66
michael@0 67 s = tab[b & 0xF]; l = s;
michael@0 68 s = tab[b >> 4 & 0xF]; l ^= s << 4; h = s >> 60;
michael@0 69 s = tab[b >> 8 & 0xF]; l ^= s << 8; h ^= s >> 56;
michael@0 70 s = tab[b >> 12 & 0xF]; l ^= s << 12; h ^= s >> 52;
michael@0 71 s = tab[b >> 16 & 0xF]; l ^= s << 16; h ^= s >> 48;
michael@0 72 s = tab[b >> 20 & 0xF]; l ^= s << 20; h ^= s >> 44;
michael@0 73 s = tab[b >> 24 & 0xF]; l ^= s << 24; h ^= s >> 40;
michael@0 74 s = tab[b >> 28 & 0xF]; l ^= s << 28; h ^= s >> 36;
michael@0 75 s = tab[b >> 32 & 0xF]; l ^= s << 32; h ^= s >> 32;
michael@0 76 s = tab[b >> 36 & 0xF]; l ^= s << 36; h ^= s >> 28;
michael@0 77 s = tab[b >> 40 & 0xF]; l ^= s << 40; h ^= s >> 24;
michael@0 78 s = tab[b >> 44 & 0xF]; l ^= s << 44; h ^= s >> 20;
michael@0 79 s = tab[b >> 48 & 0xF]; l ^= s << 48; h ^= s >> 16;
michael@0 80 s = tab[b >> 52 & 0xF]; l ^= s << 52; h ^= s >> 12;
michael@0 81 s = tab[b >> 56 & 0xF]; l ^= s << 56; h ^= s >> 8;
michael@0 82 s = tab[b >> 60 ]; l ^= s << 60; h ^= s >> 4;
michael@0 83
michael@0 84 /* compensate for the top three bits of a */
michael@0 85
michael@0 86 if (top3b & 01) { l ^= b << 61; h ^= b >> 3; }
michael@0 87 if (top3b & 02) { l ^= b << 62; h ^= b >> 2; }
michael@0 88 if (top3b & 04) { l ^= b << 63; h ^= b >> 1; }
michael@0 89
michael@0 90 *rh = h; *rl = l;
michael@0 91 }
michael@0 92 #endif
michael@0 93
michael@0 94 /* Compute xor-multiply of two binary polynomials (a1, a0) x (b1, b0)
michael@0 95 * result is a binary polynomial in 4 mp_digits r[4].
michael@0 96 * The caller MUST ensure that r has the right amount of space allocated.
michael@0 97 */
michael@0 98 void
michael@0 99 s_bmul_2x2(mp_digit *r, const mp_digit a1, const mp_digit a0, const mp_digit b1,
michael@0 100 const mp_digit b0)
michael@0 101 {
michael@0 102 mp_digit m1, m0;
michael@0 103 /* r[3] = h1, r[2] = h0; r[1] = l1; r[0] = l0 */
michael@0 104 s_bmul_1x1(r+3, r+2, a1, b1);
michael@0 105 s_bmul_1x1(r+1, r, a0, b0);
michael@0 106 s_bmul_1x1(&m1, &m0, a0 ^ a1, b0 ^ b1);
michael@0 107 /* Correction on m1 ^= l1 ^ h1; m0 ^= l0 ^ h0; */
michael@0 108 r[2] ^= m1 ^ r[1] ^ r[3]; /* h0 ^= m1 ^ l1 ^ h1; */
michael@0 109 r[1] = r[3] ^ r[2] ^ r[0] ^ m1 ^ m0; /* l1 ^= l0 ^ h0 ^ m0; */
michael@0 110 }
michael@0 111
michael@0 112 /* Compute xor-multiply of two binary polynomials (a2, a1, a0) x (b2, b1, b0)
michael@0 113 * result is a binary polynomial in 6 mp_digits r[6].
michael@0 114 * The caller MUST ensure that r has the right amount of space allocated.
michael@0 115 */
michael@0 116 void
michael@0 117 s_bmul_3x3(mp_digit *r, const mp_digit a2, const mp_digit a1, const mp_digit a0,
michael@0 118 const mp_digit b2, const mp_digit b1, const mp_digit b0)
michael@0 119 {
michael@0 120 mp_digit zm[4];
michael@0 121
michael@0 122 s_bmul_1x1(r+5, r+4, a2, b2); /* fill top 2 words */
michael@0 123 s_bmul_2x2(zm, a1, a2^a0, b1, b2^b0); /* fill middle 4 words */
michael@0 124 s_bmul_2x2(r, a1, a0, b1, b0); /* fill bottom 4 words */
michael@0 125
michael@0 126 zm[3] ^= r[3];
michael@0 127 zm[2] ^= r[2];
michael@0 128 zm[1] ^= r[1] ^ r[5];
michael@0 129 zm[0] ^= r[0] ^ r[4];
michael@0 130
michael@0 131 r[5] ^= zm[3];
michael@0 132 r[4] ^= zm[2];
michael@0 133 r[3] ^= zm[1];
michael@0 134 r[2] ^= zm[0];
michael@0 135 }
michael@0 136
michael@0 137 /* Compute xor-multiply of two binary polynomials (a3, a2, a1, a0) x (b3, b2, b1, b0)
michael@0 138 * result is a binary polynomial in 8 mp_digits r[8].
michael@0 139 * The caller MUST ensure that r has the right amount of space allocated.
michael@0 140 */
michael@0 141 void s_bmul_4x4(mp_digit *r, const mp_digit a3, const mp_digit a2, const mp_digit a1,
michael@0 142 const mp_digit a0, const mp_digit b3, const mp_digit b2, const mp_digit b1,
michael@0 143 const mp_digit b0)
michael@0 144 {
michael@0 145 mp_digit zm[4];
michael@0 146
michael@0 147 s_bmul_2x2(r+4, a3, a2, b3, b2); /* fill top 4 words */
michael@0 148 s_bmul_2x2(zm, a3^a1, a2^a0, b3^b1, b2^b0); /* fill middle 4 words */
michael@0 149 s_bmul_2x2(r, a1, a0, b1, b0); /* fill bottom 4 words */
michael@0 150
michael@0 151 zm[3] ^= r[3] ^ r[7];
michael@0 152 zm[2] ^= r[2] ^ r[6];
michael@0 153 zm[1] ^= r[1] ^ r[5];
michael@0 154 zm[0] ^= r[0] ^ r[4];
michael@0 155
michael@0 156 r[5] ^= zm[3];
michael@0 157 r[4] ^= zm[2];
michael@0 158 r[3] ^= zm[1];
michael@0 159 r[2] ^= zm[0];
michael@0 160 }
michael@0 161
michael@0 162 /* Compute addition of two binary polynomials a and b,
michael@0 163 * store result in c; c could be a or b, a and b could be equal;
michael@0 164 * c is the bitwise XOR of a and b.
michael@0 165 */
michael@0 166 mp_err
michael@0 167 mp_badd(const mp_int *a, const mp_int *b, mp_int *c)
michael@0 168 {
michael@0 169 mp_digit *pa, *pb, *pc;
michael@0 170 mp_size ix;
michael@0 171 mp_size used_pa, used_pb;
michael@0 172 mp_err res = MP_OKAY;
michael@0 173
michael@0 174 /* Add all digits up to the precision of b. If b had more
michael@0 175 * precision than a initially, swap a, b first
michael@0 176 */
michael@0 177 if (MP_USED(a) >= MP_USED(b)) {
michael@0 178 pa = MP_DIGITS(a);
michael@0 179 pb = MP_DIGITS(b);
michael@0 180 used_pa = MP_USED(a);
michael@0 181 used_pb = MP_USED(b);
michael@0 182 } else {
michael@0 183 pa = MP_DIGITS(b);
michael@0 184 pb = MP_DIGITS(a);
michael@0 185 used_pa = MP_USED(b);
michael@0 186 used_pb = MP_USED(a);
michael@0 187 }
michael@0 188
michael@0 189 /* Make sure c has enough precision for the output value */
michael@0 190 MP_CHECKOK( s_mp_pad(c, used_pa) );
michael@0 191
michael@0 192 /* Do word-by-word xor */
michael@0 193 pc = MP_DIGITS(c);
michael@0 194 for (ix = 0; ix < used_pb; ix++) {
michael@0 195 (*pc++) = (*pa++) ^ (*pb++);
michael@0 196 }
michael@0 197
michael@0 198 /* Finish the rest of digits until we're actually done */
michael@0 199 for (; ix < used_pa; ++ix) {
michael@0 200 *pc++ = *pa++;
michael@0 201 }
michael@0 202
michael@0 203 MP_USED(c) = used_pa;
michael@0 204 MP_SIGN(c) = ZPOS;
michael@0 205 s_mp_clamp(c);
michael@0 206
michael@0 207 CLEANUP:
michael@0 208 return res;
michael@0 209 }
michael@0 210
michael@0 211 #define s_mp_div2(a) MP_CHECKOK( mpl_rsh((a), (a), 1) );
michael@0 212
michael@0 213 /* Compute binary polynomial multiply d = a * b */
michael@0 214 static void
michael@0 215 s_bmul_d(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *d)
michael@0 216 {
michael@0 217 mp_digit a_i, a0b0, a1b1, carry = 0;
michael@0 218 while (a_len--) {
michael@0 219 a_i = *a++;
michael@0 220 s_bmul_1x1(&a1b1, &a0b0, a_i, b);
michael@0 221 *d++ = a0b0 ^ carry;
michael@0 222 carry = a1b1;
michael@0 223 }
michael@0 224 *d = carry;
michael@0 225 }
michael@0 226
michael@0 227 /* Compute binary polynomial xor multiply accumulate d ^= a * b */
michael@0 228 static void
michael@0 229 s_bmul_d_add(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *d)
michael@0 230 {
michael@0 231 mp_digit a_i, a0b0, a1b1, carry = 0;
michael@0 232 while (a_len--) {
michael@0 233 a_i = *a++;
michael@0 234 s_bmul_1x1(&a1b1, &a0b0, a_i, b);
michael@0 235 *d++ ^= a0b0 ^ carry;
michael@0 236 carry = a1b1;
michael@0 237 }
michael@0 238 *d ^= carry;
michael@0 239 }
michael@0 240
michael@0 241 /* Compute binary polynomial xor multiply c = a * b.
michael@0 242 * All parameters may be identical.
michael@0 243 */
michael@0 244 mp_err
michael@0 245 mp_bmul(const mp_int *a, const mp_int *b, mp_int *c)
michael@0 246 {
michael@0 247 mp_digit *pb, b_i;
michael@0 248 mp_int tmp;
michael@0 249 mp_size ib, a_used, b_used;
michael@0 250 mp_err res = MP_OKAY;
michael@0 251
michael@0 252 MP_DIGITS(&tmp) = 0;
michael@0 253
michael@0 254 ARGCHK(a != NULL && b != NULL && c != NULL, MP_BADARG);
michael@0 255
michael@0 256 if (a == c) {
michael@0 257 MP_CHECKOK( mp_init_copy(&tmp, a) );
michael@0 258 if (a == b)
michael@0 259 b = &tmp;
michael@0 260 a = &tmp;
michael@0 261 } else if (b == c) {
michael@0 262 MP_CHECKOK( mp_init_copy(&tmp, b) );
michael@0 263 b = &tmp;
michael@0 264 }
michael@0 265
michael@0 266 if (MP_USED(a) < MP_USED(b)) {
michael@0 267 const mp_int *xch = b; /* switch a and b if b longer */
michael@0 268 b = a;
michael@0 269 a = xch;
michael@0 270 }
michael@0 271
michael@0 272 MP_USED(c) = 1; MP_DIGIT(c, 0) = 0;
michael@0 273 MP_CHECKOK( s_mp_pad(c, USED(a) + USED(b)) );
michael@0 274
michael@0 275 pb = MP_DIGITS(b);
michael@0 276 s_bmul_d(MP_DIGITS(a), MP_USED(a), *pb++, MP_DIGITS(c));
michael@0 277
michael@0 278 /* Outer loop: Digits of b */
michael@0 279 a_used = MP_USED(a);
michael@0 280 b_used = MP_USED(b);
michael@0 281 MP_USED(c) = a_used + b_used;
michael@0 282 for (ib = 1; ib < b_used; ib++) {
michael@0 283 b_i = *pb++;
michael@0 284
michael@0 285 /* Inner product: Digits of a */
michael@0 286 if (b_i)
michael@0 287 s_bmul_d_add(MP_DIGITS(a), a_used, b_i, MP_DIGITS(c) + ib);
michael@0 288 else
michael@0 289 MP_DIGIT(c, ib + a_used) = b_i;
michael@0 290 }
michael@0 291
michael@0 292 s_mp_clamp(c);
michael@0 293
michael@0 294 SIGN(c) = ZPOS;
michael@0 295
michael@0 296 CLEANUP:
michael@0 297 mp_clear(&tmp);
michael@0 298 return res;
michael@0 299 }
michael@0 300
michael@0 301
michael@0 302 /* Compute modular reduction of a and store result in r.
michael@0 303 * r could be a.
michael@0 304 * For modular arithmetic, the irreducible polynomial f(t) is represented
michael@0 305 * as an array of int[], where f(t) is of the form:
michael@0 306 * f(t) = t^p[0] + t^p[1] + ... + t^p[k]
michael@0 307 * where m = p[0] > p[1] > ... > p[k] = 0.
michael@0 308 */
michael@0 309 mp_err
michael@0 310 mp_bmod(const mp_int *a, const unsigned int p[], mp_int *r)
michael@0 311 {
michael@0 312 int j, k;
michael@0 313 int n, dN, d0, d1;
michael@0 314 mp_digit zz, *z, tmp;
michael@0 315 mp_size used;
michael@0 316 mp_err res = MP_OKAY;
michael@0 317
michael@0 318 /* The algorithm does the reduction in place in r,
michael@0 319 * if a != r, copy a into r first so reduction can be done in r
michael@0 320 */
michael@0 321 if (a != r) {
michael@0 322 MP_CHECKOK( mp_copy(a, r) );
michael@0 323 }
michael@0 324 z = MP_DIGITS(r);
michael@0 325
michael@0 326 /* start reduction */
michael@0 327 /*dN = p[0] / MP_DIGIT_BITS; */
michael@0 328 dN = p[0] >> MP_DIGIT_BITS_LOG_2;
michael@0 329 used = MP_USED(r);
michael@0 330
michael@0 331 for (j = used - 1; j > dN;) {
michael@0 332
michael@0 333 zz = z[j];
michael@0 334 if (zz == 0) {
michael@0 335 j--; continue;
michael@0 336 }
michael@0 337 z[j] = 0;
michael@0 338
michael@0 339 for (k = 1; p[k] > 0; k++) {
michael@0 340 /* reducing component t^p[k] */
michael@0 341 n = p[0] - p[k];
michael@0 342 /*d0 = n % MP_DIGIT_BITS; */
michael@0 343 d0 = n & MP_DIGIT_BITS_MASK;
michael@0 344 d1 = MP_DIGIT_BITS - d0;
michael@0 345 /*n /= MP_DIGIT_BITS; */
michael@0 346 n >>= MP_DIGIT_BITS_LOG_2;
michael@0 347 z[j-n] ^= (zz>>d0);
michael@0 348 if (d0)
michael@0 349 z[j-n-1] ^= (zz<<d1);
michael@0 350 }
michael@0 351
michael@0 352 /* reducing component t^0 */
michael@0 353 n = dN;
michael@0 354 /*d0 = p[0] % MP_DIGIT_BITS;*/
michael@0 355 d0 = p[0] & MP_DIGIT_BITS_MASK;
michael@0 356 d1 = MP_DIGIT_BITS - d0;
michael@0 357 z[j-n] ^= (zz >> d0);
michael@0 358 if (d0)
michael@0 359 z[j-n-1] ^= (zz << d1);
michael@0 360
michael@0 361 }
michael@0 362
michael@0 363 /* final round of reduction */
michael@0 364 while (j == dN) {
michael@0 365
michael@0 366 /* d0 = p[0] % MP_DIGIT_BITS; */
michael@0 367 d0 = p[0] & MP_DIGIT_BITS_MASK;
michael@0 368 zz = z[dN] >> d0;
michael@0 369 if (zz == 0) break;
michael@0 370 d1 = MP_DIGIT_BITS - d0;
michael@0 371
michael@0 372 /* clear up the top d1 bits */
michael@0 373 if (d0) {
michael@0 374 z[dN] = (z[dN] << d1) >> d1;
michael@0 375 } else {
michael@0 376 z[dN] = 0;
michael@0 377 }
michael@0 378 *z ^= zz; /* reduction t^0 component */
michael@0 379
michael@0 380 for (k = 1; p[k] > 0; k++) {
michael@0 381 /* reducing component t^p[k]*/
michael@0 382 /* n = p[k] / MP_DIGIT_BITS; */
michael@0 383 n = p[k] >> MP_DIGIT_BITS_LOG_2;
michael@0 384 /* d0 = p[k] % MP_DIGIT_BITS; */
michael@0 385 d0 = p[k] & MP_DIGIT_BITS_MASK;
michael@0 386 d1 = MP_DIGIT_BITS - d0;
michael@0 387 z[n] ^= (zz << d0);
michael@0 388 tmp = zz >> d1;
michael@0 389 if (d0 && tmp)
michael@0 390 z[n+1] ^= tmp;
michael@0 391 }
michael@0 392 }
michael@0 393
michael@0 394 s_mp_clamp(r);
michael@0 395 CLEANUP:
michael@0 396 return res;
michael@0 397 }
michael@0 398
michael@0 399 /* Compute the product of two polynomials a and b, reduce modulo p,
michael@0 400 * Store the result in r. r could be a or b; a could be b.
michael@0 401 */
michael@0 402 mp_err
michael@0 403 mp_bmulmod(const mp_int *a, const mp_int *b, const unsigned int p[], mp_int *r)
michael@0 404 {
michael@0 405 mp_err res;
michael@0 406
michael@0 407 if (a == b) return mp_bsqrmod(a, p, r);
michael@0 408 if ((res = mp_bmul(a, b, r) ) != MP_OKAY)
michael@0 409 return res;
michael@0 410 return mp_bmod(r, p, r);
michael@0 411 }
michael@0 412
michael@0 413 /* Compute binary polynomial squaring c = a*a mod p .
michael@0 414 * Parameter r and a can be identical.
michael@0 415 */
michael@0 416
michael@0 417 mp_err
michael@0 418 mp_bsqrmod(const mp_int *a, const unsigned int p[], mp_int *r)
michael@0 419 {
michael@0 420 mp_digit *pa, *pr, a_i;
michael@0 421 mp_int tmp;
michael@0 422 mp_size ia, a_used;
michael@0 423 mp_err res;
michael@0 424
michael@0 425 ARGCHK(a != NULL && r != NULL, MP_BADARG);
michael@0 426 MP_DIGITS(&tmp) = 0;
michael@0 427
michael@0 428 if (a == r) {
michael@0 429 MP_CHECKOK( mp_init_copy(&tmp, a) );
michael@0 430 a = &tmp;
michael@0 431 }
michael@0 432
michael@0 433 MP_USED(r) = 1; MP_DIGIT(r, 0) = 0;
michael@0 434 MP_CHECKOK( s_mp_pad(r, 2*USED(a)) );
michael@0 435
michael@0 436 pa = MP_DIGITS(a);
michael@0 437 pr = MP_DIGITS(r);
michael@0 438 a_used = MP_USED(a);
michael@0 439 MP_USED(r) = 2 * a_used;
michael@0 440
michael@0 441 for (ia = 0; ia < a_used; ia++) {
michael@0 442 a_i = *pa++;
michael@0 443 *pr++ = gf2m_SQR0(a_i);
michael@0 444 *pr++ = gf2m_SQR1(a_i);
michael@0 445 }
michael@0 446
michael@0 447 MP_CHECKOK( mp_bmod(r, p, r) );
michael@0 448 s_mp_clamp(r);
michael@0 449 SIGN(r) = ZPOS;
michael@0 450
michael@0 451 CLEANUP:
michael@0 452 mp_clear(&tmp);
michael@0 453 return res;
michael@0 454 }
michael@0 455
michael@0 456 /* Compute binary polynomial y/x mod p, y divided by x, reduce modulo p.
michael@0 457 * Store the result in r. r could be x or y, and x could equal y.
michael@0 458 * Uses algorithm Modular_Division_GF(2^m) from
michael@0 459 * Chang-Shantz, S. "From Euclid's GCD to Montgomery Multiplication to
michael@0 460 * the Great Divide".
michael@0 461 */
michael@0 462 int
michael@0 463 mp_bdivmod(const mp_int *y, const mp_int *x, const mp_int *pp,
michael@0 464 const unsigned int p[], mp_int *r)
michael@0 465 {
michael@0 466 mp_int aa, bb, uu;
michael@0 467 mp_int *a, *b, *u, *v;
michael@0 468 mp_err res = MP_OKAY;
michael@0 469
michael@0 470 MP_DIGITS(&aa) = 0;
michael@0 471 MP_DIGITS(&bb) = 0;
michael@0 472 MP_DIGITS(&uu) = 0;
michael@0 473
michael@0 474 MP_CHECKOK( mp_init_copy(&aa, x) );
michael@0 475 MP_CHECKOK( mp_init_copy(&uu, y) );
michael@0 476 MP_CHECKOK( mp_init_copy(&bb, pp) );
michael@0 477 MP_CHECKOK( s_mp_pad(r, USED(pp)) );
michael@0 478 MP_USED(r) = 1; MP_DIGIT(r, 0) = 0;
michael@0 479
michael@0 480 a = &aa; b= &bb; u=&uu; v=r;
michael@0 481 /* reduce x and y mod p */
michael@0 482 MP_CHECKOK( mp_bmod(a, p, a) );
michael@0 483 MP_CHECKOK( mp_bmod(u, p, u) );
michael@0 484
michael@0 485 while (!mp_isodd(a)) {
michael@0 486 s_mp_div2(a);
michael@0 487 if (mp_isodd(u)) {
michael@0 488 MP_CHECKOK( mp_badd(u, pp, u) );
michael@0 489 }
michael@0 490 s_mp_div2(u);
michael@0 491 }
michael@0 492
michael@0 493 do {
michael@0 494 if (mp_cmp_mag(b, a) > 0) {
michael@0 495 MP_CHECKOK( mp_badd(b, a, b) );
michael@0 496 MP_CHECKOK( mp_badd(v, u, v) );
michael@0 497 do {
michael@0 498 s_mp_div2(b);
michael@0 499 if (mp_isodd(v)) {
michael@0 500 MP_CHECKOK( mp_badd(v, pp, v) );
michael@0 501 }
michael@0 502 s_mp_div2(v);
michael@0 503 } while (!mp_isodd(b));
michael@0 504 }
michael@0 505 else if ((MP_DIGIT(a,0) == 1) && (MP_USED(a) == 1))
michael@0 506 break;
michael@0 507 else {
michael@0 508 MP_CHECKOK( mp_badd(a, b, a) );
michael@0 509 MP_CHECKOK( mp_badd(u, v, u) );
michael@0 510 do {
michael@0 511 s_mp_div2(a);
michael@0 512 if (mp_isodd(u)) {
michael@0 513 MP_CHECKOK( mp_badd(u, pp, u) );
michael@0 514 }
michael@0 515 s_mp_div2(u);
michael@0 516 } while (!mp_isodd(a));
michael@0 517 }
michael@0 518 } while (1);
michael@0 519
michael@0 520 MP_CHECKOK( mp_copy(u, r) );
michael@0 521
michael@0 522 CLEANUP:
michael@0 523 mp_clear(&aa);
michael@0 524 mp_clear(&bb);
michael@0 525 mp_clear(&uu);
michael@0 526 return res;
michael@0 527
michael@0 528 }
michael@0 529
michael@0 530 /* Convert the bit-string representation of a polynomial a into an array
michael@0 531 * of integers corresponding to the bits with non-zero coefficient.
michael@0 532 * Up to max elements of the array will be filled. Return value is total
michael@0 533 * number of coefficients that would be extracted if array was large enough.
michael@0 534 */
michael@0 535 int
michael@0 536 mp_bpoly2arr(const mp_int *a, unsigned int p[], int max)
michael@0 537 {
michael@0 538 int i, j, k;
michael@0 539 mp_digit top_bit, mask;
michael@0 540
michael@0 541 top_bit = 1;
michael@0 542 top_bit <<= MP_DIGIT_BIT - 1;
michael@0 543
michael@0 544 for (k = 0; k < max; k++) p[k] = 0;
michael@0 545 k = 0;
michael@0 546
michael@0 547 for (i = MP_USED(a) - 1; i >= 0; i--) {
michael@0 548 mask = top_bit;
michael@0 549 for (j = MP_DIGIT_BIT - 1; j >= 0; j--) {
michael@0 550 if (MP_DIGITS(a)[i] & mask) {
michael@0 551 if (k < max) p[k] = MP_DIGIT_BIT * i + j;
michael@0 552 k++;
michael@0 553 }
michael@0 554 mask >>= 1;
michael@0 555 }
michael@0 556 }
michael@0 557
michael@0 558 return k;
michael@0 559 }
michael@0 560
michael@0 561 /* Convert the coefficient array representation of a polynomial to a
michael@0 562 * bit-string. The array must be terminated by 0.
michael@0 563 */
michael@0 564 mp_err
michael@0 565 mp_barr2poly(const unsigned int p[], mp_int *a)
michael@0 566 {
michael@0 567
michael@0 568 mp_err res = MP_OKAY;
michael@0 569 int i;
michael@0 570
michael@0 571 mp_zero(a);
michael@0 572 for (i = 0; p[i] > 0; i++) {
michael@0 573 MP_CHECKOK( mpl_set_bit(a, p[i], 1) );
michael@0 574 }
michael@0 575 MP_CHECKOK( mpl_set_bit(a, 0, 1) );
michael@0 576
michael@0 577 CLEANUP:
michael@0 578 return res;
michael@0 579 }

mercurial