1/* 2 * Minimal code for RSA support from LibTomMath 0.41 3 * http://libtom.org/ 4 * http://libtom.org/files/ltm-0.41.tar.bz2 5 * This library was released in public domain by Tom St Denis. 6 * 7 * The combination in this file may not use all of the optimized algorithms 8 * from LibTomMath and may be considerable slower than the LibTomMath with its 9 * default settings. The main purpose of having this version here is to make it 10 * easier to build bignum.c wrapper without having to install and build an 11 * external library. 12 * 13 * If CONFIG_INTERNAL_LIBTOMMATH is defined, bignum.c includes this 14 * libtommath.c file instead of using the external LibTomMath library. 15 */ 16 17#ifndef CHAR_BIT 18#define CHAR_BIT 8 19#endif 20 21#define BN_MP_INVMOD_C 22#define BN_S_MP_EXPTMOD_C /* Note: #undef in tommath_superclass.h; this would 23 * require BN_MP_EXPTMOD_FAST_C instead */ 24#define BN_S_MP_MUL_DIGS_C 25#define BN_MP_INVMOD_SLOW_C 26#define BN_S_MP_SQR_C 27#define BN_S_MP_MUL_HIGH_DIGS_C /* Note: #undef in tommath_superclass.h; this 28 * would require other than mp_reduce */ 29 30#ifdef LTM_FAST 31 32/* Use faster div at the cost of about 1 kB */ 33#define BN_MP_MUL_D_C 34 35/* Include faster exptmod (Montgomery) at the cost of about 2.5 kB in code */ 36#define BN_MP_EXPTMOD_FAST_C 37#define BN_MP_MONTGOMERY_SETUP_C 38#define BN_FAST_MP_MONTGOMERY_REDUCE_C 39#define BN_MP_MONTGOMERY_CALC_NORMALIZATION_C 40#define BN_MP_MUL_2_C 41 42/* Include faster sqr at the cost of about 0.5 kB in code */ 43#define BN_FAST_S_MP_SQR_C 44 45/* About 0.25 kB of code, but ~1.7kB of stack space! */ 46#define BN_FAST_S_MP_MUL_DIGS_C 47 48#else /* LTM_FAST */ 49 50#define BN_MP_DIV_SMALL 51#define BN_MP_INIT_MULTI_C 52#define BN_MP_CLEAR_MULTI_C 53#define BN_MP_ABS_C 54#endif /* LTM_FAST */ 55 56/* Current uses do not require support for negative exponent in exptmod, so we 57 * can save about 1.5 kB in leaving out invmod. */ 58#define LTM_NO_NEG_EXP 59 60/* from tommath.h */ 61 62#ifndef MIN 63 #define MIN(x,y) ((x)<(y)?(x):(y)) 64#endif 65 66#ifndef MAX 67 #define MAX(x,y) ((x)>(y)?(x):(y)) 68#endif 69 70#define OPT_CAST(x) 71 72#ifdef __x86_64__ 73typedef unsigned long mp_digit; 74typedef unsigned long mp_word __attribute__((mode(TI))); 75 76#define DIGIT_BIT 60 77#define MP_64BIT 78#else 79typedef unsigned long mp_digit; 80typedef u64 mp_word; 81 82#define DIGIT_BIT 28 83#define MP_28BIT 84#endif 85 86 87#define XMALLOC os_malloc 88#define XFREE os_free 89#define XREALLOC os_realloc 90 91 92#define MP_MASK ((((mp_digit)1)<<((mp_digit)DIGIT_BIT))-((mp_digit)1)) 93 94#define MP_LT -1 /* less than */ 95#define MP_EQ 0 /* equal to */ 96#define MP_GT 1 /* greater than */ 97 98#define MP_ZPOS 0 /* positive integer */ 99#define MP_NEG 1 /* negative */ 100 101#define MP_OKAY 0 /* ok result */ 102#define MP_MEM -2 /* out of mem */ 103#define MP_VAL -3 /* invalid input */ 104 105#define MP_YES 1 /* yes response */ 106#define MP_NO 0 /* no response */ 107 108typedef int mp_err; 109 110/* define this to use lower memory usage routines (exptmods mostly) */ 111#define MP_LOW_MEM 112 113/* default precision */ 114#ifndef MP_PREC 115 #ifndef MP_LOW_MEM 116 #define MP_PREC 32 /* default digits of precision */ 117 #else 118 #define MP_PREC 8 /* default digits of precision */ 119 #endif 120#endif 121 122/* size of comba arrays, should be at least 2 * 2**(BITS_PER_WORD - BITS_PER_DIGIT*2) */ 123#define MP_WARRAY (1 << (sizeof(mp_word) * CHAR_BIT - 2 * DIGIT_BIT + 1)) 124 125/* the infamous mp_int structure */ 126typedef struct { 127 int used, alloc, sign; 128 mp_digit *dp; 129} mp_int; 130 131 132/* ---> Basic Manipulations <--- */ 133#define mp_iszero(a) (((a)->used == 0) ? MP_YES : MP_NO) 134#define mp_iseven(a) (((a)->used > 0 && (((a)->dp[0] & 1) == 0)) ? MP_YES : MP_NO) 135#define mp_isodd(a) (((a)->used > 0 && (((a)->dp[0] & 1) == 1)) ? MP_YES : MP_NO) 136 137 138/* prototypes for copied functions */ 139#define s_mp_mul(a, b, c) s_mp_mul_digs(a, b, c, (a)->used + (b)->used + 1) 140static int s_mp_exptmod(mp_int * G, mp_int * X, mp_int * P, mp_int * Y, int redmode); 141static int s_mp_mul_digs (mp_int * a, mp_int * b, mp_int * c, int digs); 142static int s_mp_sqr(mp_int * a, mp_int * b); 143static int s_mp_mul_high_digs(mp_int * a, mp_int * b, mp_int * c, int digs); 144 145#ifdef BN_FAST_S_MP_MUL_DIGS_C 146static int fast_s_mp_mul_digs (mp_int * a, mp_int * b, mp_int * c, int digs); 147#endif 148 149#ifdef BN_MP_INIT_MULTI_C 150static int mp_init_multi(mp_int *mp, ...); 151#endif 152#ifdef BN_MP_CLEAR_MULTI_C 153static void mp_clear_multi(mp_int *mp, ...); 154#endif 155static int mp_lshd(mp_int * a, int b); 156static void mp_set(mp_int * a, mp_digit b); 157static void mp_clamp(mp_int * a); 158static void mp_exch(mp_int * a, mp_int * b); 159static void mp_rshd(mp_int * a, int b); 160static void mp_zero(mp_int * a); 161static int mp_mod_2d(mp_int * a, int b, mp_int * c); 162static int mp_div_2d(mp_int * a, int b, mp_int * c, mp_int * d); 163static int mp_init_copy(mp_int * a, mp_int * b); 164static int mp_mul_2d(mp_int * a, int b, mp_int * c); 165#ifndef LTM_NO_NEG_EXP 166static int mp_div_2(mp_int * a, mp_int * b); 167static int mp_invmod(mp_int * a, mp_int * b, mp_int * c); 168static int mp_invmod_slow(mp_int * a, mp_int * b, mp_int * c); 169#endif /* LTM_NO_NEG_EXP */ 170static int mp_copy(mp_int * a, mp_int * b); 171static int mp_count_bits(mp_int * a); 172static int mp_div(mp_int * a, mp_int * b, mp_int * c, mp_int * d); 173static int mp_mod(mp_int * a, mp_int * b, mp_int * c); 174static int mp_grow(mp_int * a, int size); 175static int mp_cmp_mag(mp_int * a, mp_int * b); 176#ifdef BN_MP_ABS_C 177static int mp_abs(mp_int * a, mp_int * b); 178#endif 179static int mp_sqr(mp_int * a, mp_int * b); 180static int mp_reduce_2k_l(mp_int *a, mp_int *n, mp_int *d); 181static int mp_reduce_2k_setup_l(mp_int *a, mp_int *d); 182static int mp_2expt(mp_int * a, int b); 183static int mp_reduce_setup(mp_int * a, mp_int * b); 184static int mp_reduce(mp_int * x, mp_int * m, mp_int * mu); 185static int mp_init_size(mp_int * a, int size); 186#ifdef BN_MP_EXPTMOD_FAST_C 187static int mp_exptmod_fast (mp_int * G, mp_int * X, mp_int * P, mp_int * Y, int redmode); 188#endif /* BN_MP_EXPTMOD_FAST_C */ 189#ifdef BN_FAST_S_MP_SQR_C 190static int fast_s_mp_sqr (mp_int * a, mp_int * b); 191#endif /* BN_FAST_S_MP_SQR_C */ 192#ifdef BN_MP_MUL_D_C 193static int mp_mul_d (mp_int * a, mp_digit b, mp_int * c); 194#endif /* BN_MP_MUL_D_C */ 195 196 197 198/* functions from bn_<func name>.c */ 199 200 201/* reverse an array, used for radix code */ 202static void bn_reverse (unsigned char *s, int len) 203{ 204 int ix, iy; 205 unsigned char t; 206 207 ix = 0; 208 iy = len - 1; 209 while (ix < iy) { 210 t = s[ix]; 211 s[ix] = s[iy]; 212 s[iy] = t; 213 ++ix; 214 --iy; 215 } 216} 217 218 219/* low level addition, based on HAC pp.594, Algorithm 14.7 */ 220static int s_mp_add (mp_int * a, mp_int * b, mp_int * c) 221{ 222 mp_int *x; 223 int olduse, res, min, max; 224 225 /* find sizes, we let |a| <= |b| which means we have to sort 226 * them. "x" will point to the input with the most digits 227 */ 228 if (a->used > b->used) { 229 min = b->used; 230 max = a->used; 231 x = a; 232 } else { 233 min = a->used; 234 max = b->used; 235 x = b; 236 } 237 238 /* init result */ 239 if (c->alloc < max + 1) { 240 if ((res = mp_grow (c, max + 1)) != MP_OKAY) { 241 return res; 242 } 243 } 244 245 /* get old used digit count and set new one */ 246 olduse = c->used; 247 c->used = max + 1; 248 249 { 250 register mp_digit u, *tmpa, *tmpb, *tmpc; 251 register int i; 252 253 /* alias for digit pointers */ 254 255 /* first input */ 256 tmpa = a->dp; 257 258 /* second input */ 259 tmpb = b->dp; 260 261 /* destination */ 262 tmpc = c->dp; 263 264 /* zero the carry */ 265 u = 0; 266 for (i = 0; i < min; i++) { 267 /* Compute the sum at one digit, T[i] = A[i] + B[i] + U */ 268 *tmpc = *tmpa++ + *tmpb++ + u; 269 270 /* U = carry bit of T[i] */ 271 u = *tmpc >> ((mp_digit)DIGIT_BIT); 272 273 /* take away carry bit from T[i] */ 274 *tmpc++ &= MP_MASK; 275 } 276 277 /* now copy higher words if any, that is in A+B 278 * if A or B has more digits add those in 279 */ 280 if (min != max) { 281 for (; i < max; i++) { 282 /* T[i] = X[i] + U */ 283 *tmpc = x->dp[i] + u; 284 285 /* U = carry bit of T[i] */ 286 u = *tmpc >> ((mp_digit)DIGIT_BIT); 287 288 /* take away carry bit from T[i] */ 289 *tmpc++ &= MP_MASK; 290 } 291 } 292 293 /* add carry */ 294 *tmpc++ = u; 295 296 /* clear digits above oldused */ 297 for (i = c->used; i < olduse; i++) { 298 *tmpc++ = 0; 299 } 300 } 301 302 mp_clamp (c); 303 return MP_OKAY; 304} 305 306 307/* low level subtraction (assumes |a| > |b|), HAC pp.595 Algorithm 14.9 */ 308static int s_mp_sub (mp_int * a, mp_int * b, mp_int * c) 309{ 310 int olduse, res, min, max; 311 312 /* find sizes */ 313 min = b->used; 314 max = a->used; 315 316 /* init result */ 317 if (c->alloc < max) { 318 if ((res = mp_grow (c, max)) != MP_OKAY) { 319 return res; 320 } 321 } 322 olduse = c->used; 323 c->used = max; 324 325 { 326 register mp_digit u, *tmpa, *tmpb, *tmpc; 327 register int i; 328 329 /* alias for digit pointers */ 330 tmpa = a->dp; 331 tmpb = b->dp; 332 tmpc = c->dp; 333 334 /* set carry to zero */ 335 u = 0; 336 for (i = 0; i < min; i++) { 337 /* T[i] = A[i] - B[i] - U */ 338 *tmpc = *tmpa++ - *tmpb++ - u; 339 340 /* U = carry bit of T[i] 341 * Note this saves performing an AND operation since 342 * if a carry does occur it will propagate all the way to the 343 * MSB. As a result a single shift is enough to get the carry 344 */ 345 u = *tmpc >> ((mp_digit)(CHAR_BIT * sizeof (mp_digit) - 1)); 346 347 /* Clear carry from T[i] */ 348 *tmpc++ &= MP_MASK; 349 } 350 351 /* now copy higher words if any, e.g. if A has more digits than B */ 352 for (; i < max; i++) { 353 /* T[i] = A[i] - U */ 354 *tmpc = *tmpa++ - u; 355 356 /* U = carry bit of T[i] */ 357 u = *tmpc >> ((mp_digit)(CHAR_BIT * sizeof (mp_digit) - 1)); 358 359 /* Clear carry from T[i] */ 360 *tmpc++ &= MP_MASK; 361 } 362 363 /* clear digits above used (since we may not have grown result above) */ 364 for (i = c->used; i < olduse; i++) { 365 *tmpc++ = 0; 366 } 367 } 368 369 mp_clamp (c); 370 return MP_OKAY; 371} 372 373 374/* init a new mp_int */ 375static int mp_init (mp_int * a) 376{ 377 int i; 378 379 /* allocate memory required and clear it */ 380 a->dp = OPT_CAST(mp_digit) XMALLOC (sizeof (mp_digit) * MP_PREC); 381 if (a->dp == NULL) { 382 return MP_MEM; 383 } 384 385 /* set the digits to zero */ 386 for (i = 0; i < MP_PREC; i++) { 387 a->dp[i] = 0; 388 } 389 390 /* set the used to zero, allocated digits to the default precision 391 * and sign to positive */ 392 a->used = 0; 393 a->alloc = MP_PREC; 394 a->sign = MP_ZPOS; 395 396 return MP_OKAY; 397} 398 399 400/* clear one (frees) */ 401static void mp_clear (mp_int * a) 402{ 403 int i; 404 405 /* only do anything if a hasn't been freed previously */ 406 if (a->dp != NULL) { 407 /* first zero the digits */ 408 for (i = 0; i < a->used; i++) { 409 a->dp[i] = 0; 410 } 411 412 /* free ram */ 413 XFREE(a->dp); 414 415 /* reset members to make debugging easier */ 416 a->dp = NULL; 417 a->alloc = a->used = 0; 418 a->sign = MP_ZPOS; 419 } 420} 421 422 423/* high level addition (handles signs) */ 424static int mp_add (mp_int * a, mp_int * b, mp_int * c) 425{ 426 int sa, sb, res; 427 428 /* get sign of both inputs */ 429 sa = a->sign; 430 sb = b->sign; 431 432 /* handle two cases, not four */ 433 if (sa == sb) { 434 /* both positive or both negative */ 435 /* add their magnitudes, copy the sign */ 436 c->sign = sa; 437 res = s_mp_add (a, b, c); 438 } else { 439 /* one positive, the other negative */ 440 /* subtract the one with the greater magnitude from */ 441 /* the one of the lesser magnitude. The result gets */ 442 /* the sign of the one with the greater magnitude. */ 443 if (mp_cmp_mag (a, b) == MP_LT) { 444 c->sign = sb; 445 res = s_mp_sub (b, a, c); 446 } else { 447 c->sign = sa; 448 res = s_mp_sub (a, b, c); 449 } 450 } 451 return res; 452} 453 454 455/* high level subtraction (handles signs) */ 456static int mp_sub (mp_int * a, mp_int * b, mp_int * c) 457{ 458 int sa, sb, res; 459 460 sa = a->sign; 461 sb = b->sign; 462 463 if (sa != sb) { 464 /* subtract a negative from a positive, OR */ 465 /* subtract a positive from a negative. */ 466 /* In either case, ADD their magnitudes, */ 467 /* and use the sign of the first number. */ 468 c->sign = sa; 469 res = s_mp_add (a, b, c); 470 } else { 471 /* subtract a positive from a positive, OR */ 472 /* subtract a negative from a negative. */ 473 /* First, take the difference between their */ 474 /* magnitudes, then... */ 475 if (mp_cmp_mag (a, b) != MP_LT) { 476 /* Copy the sign from the first */ 477 c->sign = sa; 478 /* The first has a larger or equal magnitude */ 479 res = s_mp_sub (a, b, c); 480 } else { 481 /* The result has the *opposite* sign from */ 482 /* the first number. */ 483 c->sign = (sa == MP_ZPOS) ? MP_NEG : MP_ZPOS; 484 /* The second has a larger magnitude */ 485 res = s_mp_sub (b, a, c); 486 } 487 } 488 return res; 489} 490 491 492/* high level multiplication (handles sign) */ 493static int mp_mul (mp_int * a, mp_int * b, mp_int * c) 494{ 495 int res, neg; 496 neg = (a->sign == b->sign) ? MP_ZPOS : MP_NEG; 497 498 /* use Toom-Cook? */ 499#ifdef BN_MP_TOOM_MUL_C 500 if (MIN (a->used, b->used) >= TOOM_MUL_CUTOFF) { 501 res = mp_toom_mul(a, b, c); 502 } else 503#endif 504#ifdef BN_MP_KARATSUBA_MUL_C 505 /* use Karatsuba? */ 506 if (MIN (a->used, b->used) >= KARATSUBA_MUL_CUTOFF) { 507 res = mp_karatsuba_mul (a, b, c); 508 } else 509#endif 510 { 511 /* can we use the fast multiplier? 512 * 513 * The fast multiplier can be used if the output will 514 * have less than MP_WARRAY digits and the number of 515 * digits won't affect carry propagation 516 */ 517#ifdef BN_FAST_S_MP_MUL_DIGS_C 518 int digs = a->used + b->used + 1; 519 520 if ((digs < MP_WARRAY) && 521 MIN(a->used, b->used) <= 522 (1 << ((CHAR_BIT * sizeof (mp_word)) - (2 * DIGIT_BIT)))) { 523 res = fast_s_mp_mul_digs (a, b, c, digs); 524 } else 525#endif 526#ifdef BN_S_MP_MUL_DIGS_C 527 res = s_mp_mul (a, b, c); /* uses s_mp_mul_digs */ 528#else 529#error mp_mul could fail 530 res = MP_VAL; 531#endif 532 533 } 534 c->sign = (c->used > 0) ? neg : MP_ZPOS; 535 return res; 536} 537 538 539/* d = a * b (mod c) */ 540static int mp_mulmod (mp_int * a, mp_int * b, mp_int * c, mp_int * d) 541{ 542 int res; 543 mp_int t; 544 545 if ((res = mp_init (&t)) != MP_OKAY) { 546 return res; 547 } 548 549 if ((res = mp_mul (a, b, &t)) != MP_OKAY) { 550 mp_clear (&t); 551 return res; 552 } 553 res = mp_mod (&t, c, d); 554 mp_clear (&t); 555 return res; 556} 557 558 559/* c = a mod b, 0 <= c < b */ 560static int mp_mod (mp_int * a, mp_int * b, mp_int * c) 561{ 562 mp_int t; 563 int res; 564 565 if ((res = mp_init (&t)) != MP_OKAY) { 566 return res; 567 } 568 569 if ((res = mp_div (a, b, NULL, &t)) != MP_OKAY) { 570 mp_clear (&t); 571 return res; 572 } 573 574 if (t.sign != b->sign) { 575 res = mp_add (b, &t, c); 576 } else { 577 res = MP_OKAY; 578 mp_exch (&t, c); 579 } 580 581 mp_clear (&t); 582 return res; 583} 584 585 586/* this is a shell function that calls either the normal or Montgomery 587 * exptmod functions. Originally the call to the montgomery code was 588 * embedded in the normal function but that wasted a lot of stack space 589 * for nothing (since 99% of the time the Montgomery code would be called) 590 */ 591static int mp_exptmod (mp_int * G, mp_int * X, mp_int * P, mp_int * Y) 592{ 593 int dr; 594 595 /* modulus P must be positive */ 596 if (P->sign == MP_NEG) { 597 return MP_VAL; 598 } 599 600 /* if exponent X is negative we have to recurse */ 601 if (X->sign == MP_NEG) { 602#ifdef LTM_NO_NEG_EXP 603 return MP_VAL; 604#else /* LTM_NO_NEG_EXP */ 605#ifdef BN_MP_INVMOD_C 606 mp_int tmpG, tmpX; 607 int err; 608 609 /* first compute 1/G mod P */ 610 if ((err = mp_init(&tmpG)) != MP_OKAY) { 611 return err; 612 } 613 if ((err = mp_invmod(G, P, &tmpG)) != MP_OKAY) { 614 mp_clear(&tmpG); 615 return err; 616 } 617 618 /* now get |X| */ 619 if ((err = mp_init(&tmpX)) != MP_OKAY) { 620 mp_clear(&tmpG); 621 return err; 622 } 623 if ((err = mp_abs(X, &tmpX)) != MP_OKAY) { 624 mp_clear_multi(&tmpG, &tmpX, NULL); 625 return err; 626 } 627 628 /* and now compute (1/G)**|X| instead of G**X [X < 0] */ 629 err = mp_exptmod(&tmpG, &tmpX, P, Y); 630 mp_clear_multi(&tmpG, &tmpX, NULL); 631 return err; 632#else 633#error mp_exptmod would always fail 634 /* no invmod */ 635 return MP_VAL; 636#endif 637#endif /* LTM_NO_NEG_EXP */ 638 } 639 640/* modified diminished radix reduction */ 641#if defined(BN_MP_REDUCE_IS_2K_L_C) && defined(BN_MP_REDUCE_2K_L_C) && defined(BN_S_MP_EXPTMOD_C) 642 if (mp_reduce_is_2k_l(P) == MP_YES) { 643 return s_mp_exptmod(G, X, P, Y, 1); 644 } 645#endif 646 647#ifdef BN_MP_DR_IS_MODULUS_C 648 /* is it a DR modulus? */ 649 dr = mp_dr_is_modulus(P); 650#else 651 /* default to no */ 652 dr = 0; 653#endif 654 655#ifdef BN_MP_REDUCE_IS_2K_C 656 /* if not, is it a unrestricted DR modulus? */ 657 if (dr == 0) { 658 dr = mp_reduce_is_2k(P) << 1; 659 } 660#endif 661 662 /* if the modulus is odd or dr != 0 use the montgomery method */ 663#ifdef BN_MP_EXPTMOD_FAST_C 664 if (mp_isodd (P) == 1 || dr != 0) { 665 return mp_exptmod_fast (G, X, P, Y, dr); 666 } else { 667#endif 668#ifdef BN_S_MP_EXPTMOD_C 669 /* otherwise use the generic Barrett reduction technique */ 670 return s_mp_exptmod (G, X, P, Y, 0); 671#else 672#error mp_exptmod could fail 673 /* no exptmod for evens */ 674 return MP_VAL; 675#endif 676#ifdef BN_MP_EXPTMOD_FAST_C 677 } 678#endif 679 if (dr == 0) { 680 /* avoid compiler warnings about possibly unused variable */ 681 } 682} 683 684 685/* compare two ints (signed)*/ 686static int mp_cmp (mp_int * a, mp_int * b) 687{ 688 /* compare based on sign */ 689 if (a->sign != b->sign) { 690 if (a->sign == MP_NEG) { 691 return MP_LT; 692 } else { 693 return MP_GT; 694 } 695 } 696 697 /* compare digits */ 698 if (a->sign == MP_NEG) { 699 /* if negative compare opposite direction */ 700 return mp_cmp_mag(b, a); 701 } else { 702 return mp_cmp_mag(a, b); 703 } 704} 705 706 707/* compare a digit */ 708static int mp_cmp_d(mp_int * a, mp_digit b) 709{ 710 /* compare based on sign */ 711 if (a->sign == MP_NEG) { 712 return MP_LT; 713 } 714 715 /* compare based on magnitude */ 716 if (a->used > 1) { 717 return MP_GT; 718 } 719 720 /* compare the only digit of a to b */ 721 if (a->dp[0] > b) { 722 return MP_GT; 723 } else if (a->dp[0] < b) { 724 return MP_LT; 725 } else { 726 return MP_EQ; 727 } 728} 729 730 731#ifndef LTM_NO_NEG_EXP 732/* hac 14.61, pp608 */ 733static int mp_invmod (mp_int * a, mp_int * b, mp_int * c) 734{ 735 /* b cannot be negative */ 736 if (b->sign == MP_NEG || mp_iszero(b) == 1) { 737 return MP_VAL; 738 } 739 740#ifdef BN_FAST_MP_INVMOD_C 741 /* if the modulus is odd we can use a faster routine instead */ 742 if (mp_isodd (b) == 1) { 743 return fast_mp_invmod (a, b, c); 744 } 745#endif 746 747#ifdef BN_MP_INVMOD_SLOW_C 748 return mp_invmod_slow(a, b, c); 749#endif 750 751#ifndef BN_FAST_MP_INVMOD_C 752#ifndef BN_MP_INVMOD_SLOW_C 753#error mp_invmod would always fail 754#endif 755#endif 756 return MP_VAL; 757} 758#endif /* LTM_NO_NEG_EXP */ 759 760 761/* get the size for an unsigned equivalent */ 762static int mp_unsigned_bin_size (mp_int * a) 763{ 764 int size = mp_count_bits (a); 765 return (size / 8 + ((size & 7) != 0 ? 1 : 0)); 766} 767 768 769#ifndef LTM_NO_NEG_EXP 770/* hac 14.61, pp608 */ 771static int mp_invmod_slow (mp_int * a, mp_int * b, mp_int * c) 772{ 773 mp_int x, y, u, v, A, B, C, D; 774 int res; 775 776 /* b cannot be negative */ 777 if (b->sign == MP_NEG || mp_iszero(b) == 1) { 778 return MP_VAL; 779 } 780 781 /* init temps */ 782 if ((res = mp_init_multi(&x, &y, &u, &v, 783 &A, &B, &C, &D, NULL)) != MP_OKAY) { 784 return res; 785 } 786 787 /* x = a, y = b */ 788 if ((res = mp_mod(a, b, &x)) != MP_OKAY) { 789 goto LBL_ERR; 790 } 791 if ((res = mp_copy (b, &y)) != MP_OKAY) { 792 goto LBL_ERR; 793 } 794 795 /* 2. [modified] if x,y are both even then return an error! */ 796 if (mp_iseven (&x) == 1 && mp_iseven (&y) == 1) { 797 res = MP_VAL; 798 goto LBL_ERR; 799 } 800 801 /* 3. u=x, v=y, A=1, B=0, C=0,D=1 */ 802 if ((res = mp_copy (&x, &u)) != MP_OKAY) { 803 goto LBL_ERR; 804 } 805 if ((res = mp_copy (&y, &v)) != MP_OKAY) { 806 goto LBL_ERR; 807 } 808 mp_set (&A, 1); 809 mp_set (&D, 1); 810 811top: 812 /* 4. while u is even do */ 813 while (mp_iseven (&u) == 1) { 814 /* 4.1 u = u/2 */ 815 if ((res = mp_div_2 (&u, &u)) != MP_OKAY) { 816 goto LBL_ERR; 817 } 818 /* 4.2 if A or B is odd then */ 819 if (mp_isodd (&A) == 1 || mp_isodd (&B) == 1) { 820 /* A = (A+y)/2, B = (B-x)/2 */ 821 if ((res = mp_add (&A, &y, &A)) != MP_OKAY) { 822 goto LBL_ERR; 823 } 824 if ((res = mp_sub (&B, &x, &B)) != MP_OKAY) { 825 goto LBL_ERR; 826 } 827 } 828 /* A = A/2, B = B/2 */ 829 if ((res = mp_div_2 (&A, &A)) != MP_OKAY) { 830 goto LBL_ERR; 831 } 832 if ((res = mp_div_2 (&B, &B)) != MP_OKAY) { 833 goto LBL_ERR; 834 } 835 } 836 837 /* 5. while v is even do */ 838 while (mp_iseven (&v) == 1) { 839 /* 5.1 v = v/2 */ 840 if ((res = mp_div_2 (&v, &v)) != MP_OKAY) { 841 goto LBL_ERR; 842 } 843 /* 5.2 if C or D is odd then */ 844 if (mp_isodd (&C) == 1 || mp_isodd (&D) == 1) { 845 /* C = (C+y)/2, D = (D-x)/2 */ 846 if ((res = mp_add (&C, &y, &C)) != MP_OKAY) { 847 goto LBL_ERR; 848 } 849 if ((res = mp_sub (&D, &x, &D)) != MP_OKAY) { 850 goto LBL_ERR; 851 } 852 } 853 /* C = C/2, D = D/2 */ 854 if ((res = mp_div_2 (&C, &C)) != MP_OKAY) { 855 goto LBL_ERR; 856 } 857 if ((res = mp_div_2 (&D, &D)) != MP_OKAY) { 858 goto LBL_ERR; 859 } 860 } 861 862 /* 6. if u >= v then */ 863 if (mp_cmp (&u, &v) != MP_LT) { 864 /* u = u - v, A = A - C, B = B - D */ 865 if ((res = mp_sub (&u, &v, &u)) != MP_OKAY) { 866 goto LBL_ERR; 867 } 868 869 if ((res = mp_sub (&A, &C, &A)) != MP_OKAY) { 870 goto LBL_ERR; 871 } 872 873 if ((res = mp_sub (&B, &D, &B)) != MP_OKAY) { 874 goto LBL_ERR; 875 } 876 } else { 877 /* v - v - u, C = C - A, D = D - B */ 878 if ((res = mp_sub (&v, &u, &v)) != MP_OKAY) { 879 goto LBL_ERR; 880 } 881 882 if ((res = mp_sub (&C, &A, &C)) != MP_OKAY) { 883 goto LBL_ERR; 884 } 885 886 if ((res = mp_sub (&D, &B, &D)) != MP_OKAY) { 887 goto LBL_ERR; 888 } 889 } 890 891 /* if not zero goto step 4 */ 892 if (mp_iszero (&u) == 0) 893 goto top; 894 895 /* now a = C, b = D, gcd == g*v */ 896 897 /* if v != 1 then there is no inverse */ 898 if (mp_cmp_d (&v, 1) != MP_EQ) { 899 res = MP_VAL; 900 goto LBL_ERR; 901 } 902 903 /* if its too low */ 904 while (mp_cmp_d(&C, 0) == MP_LT) { 905 if ((res = mp_add(&C, b, &C)) != MP_OKAY) { 906 goto LBL_ERR; 907 } 908 } 909 910 /* too big */ 911 while (mp_cmp_mag(&C, b) != MP_LT) { 912 if ((res = mp_sub(&C, b, &C)) != MP_OKAY) { 913 goto LBL_ERR; 914 } 915 } 916 917 /* C is now the inverse */ 918 mp_exch (&C, c); 919 res = MP_OKAY; 920LBL_ERR:mp_clear_multi (&x, &y, &u, &v, &A, &B, &C, &D, NULL); 921 return res; 922} 923#endif /* LTM_NO_NEG_EXP */ 924 925 926/* compare maginitude of two ints (unsigned) */ 927static int mp_cmp_mag (mp_int * a, mp_int * b) 928{ 929 int n; 930 mp_digit *tmpa, *tmpb; 931 932 /* compare based on # of non-zero digits */ 933 if (a->used > b->used) { 934 return MP_GT; 935 } 936 937 if (a->used < b->used) { 938 return MP_LT; 939 } 940 941 /* alias for a */ 942 tmpa = a->dp + (a->used - 1); 943 944 /* alias for b */ 945 tmpb = b->dp + (a->used - 1); 946 947 /* compare based on digits */ 948 for (n = 0; n < a->used; ++n, --tmpa, --tmpb) { 949 if (*tmpa > *tmpb) { 950 return MP_GT; 951 } 952 953 if (*tmpa < *tmpb) { 954 return MP_LT; 955 } 956 } 957 return MP_EQ; 958} 959 960 961/* reads a unsigned char array, assumes the msb is stored first [big endian] */ 962static int mp_read_unsigned_bin (mp_int * a, const unsigned char *b, int c) 963{ 964 int res; 965 966 /* make sure there are at least two digits */ 967 if (a->alloc < 2) { 968 if ((res = mp_grow(a, 2)) != MP_OKAY) { 969 return res; 970 } 971 } 972 973 /* zero the int */ 974 mp_zero (a); 975 976 /* read the bytes in */ 977 while (c-- > 0) { 978 if ((res = mp_mul_2d (a, 8, a)) != MP_OKAY) { 979 return res; 980 } 981 982#ifndef MP_8BIT 983 a->dp[0] |= *b++; 984 a->used += 1; 985#else 986 a->dp[0] = (*b & MP_MASK); 987 a->dp[1] |= ((*b++ >> 7U) & 1); 988 a->used += 2; 989#endif 990 } 991 mp_clamp (a); 992 return MP_OKAY; 993} 994 995 996/* store in unsigned [big endian] format */ 997static int mp_to_unsigned_bin (mp_int * a, unsigned char *b) 998{ 999 int x, res; 1000 mp_int t; 1001 1002 if ((res = mp_init_copy (&t, a)) != MP_OKAY) { 1003 return res; 1004 } 1005 1006 x = 0; 1007 while (mp_iszero (&t) == 0) { 1008#ifndef MP_8BIT 1009 b[x++] = (unsigned char) (t.dp[0] & 255); 1010#else 1011 b[x++] = (unsigned char) (t.dp[0] | ((t.dp[1] & 0x01) << 7)); 1012#endif 1013 if ((res = mp_div_2d (&t, 8, &t, NULL)) != MP_OKAY) { 1014 mp_clear (&t); 1015 return res; 1016 } 1017 } 1018 bn_reverse (b, x); 1019 mp_clear (&t); 1020 return MP_OKAY; 1021} 1022 1023 1024/* shift right by a certain bit count (store quotient in c, optional remainder in d) */ 1025static int mp_div_2d (mp_int * a, int b, mp_int * c, mp_int * d) 1026{ 1027 mp_digit D, r, rr; 1028 int x, res; 1029 mp_int t; 1030 1031 1032 /* if the shift count is <= 0 then we do no work */ 1033 if (b <= 0) { 1034 res = mp_copy (a, c); 1035 if (d != NULL) { 1036 mp_zero (d); 1037 } 1038 return res; 1039 } 1040 1041 if ((res = mp_init (&t)) != MP_OKAY) { 1042 return res; 1043 } 1044 1045 /* get the remainder */ 1046 if (d != NULL) { 1047 if ((res = mp_mod_2d (a, b, &t)) != MP_OKAY) { 1048 mp_clear (&t); 1049 return res; 1050 } 1051 } 1052 1053 /* copy */ 1054 if ((res = mp_copy (a, c)) != MP_OKAY) { 1055 mp_clear (&t); 1056 return res; 1057 } 1058 1059 /* shift by as many digits in the bit count */ 1060 if (b >= (int)DIGIT_BIT) { 1061 mp_rshd (c, b / DIGIT_BIT); 1062 } 1063 1064 /* shift any bit count < DIGIT_BIT */ 1065 D = (mp_digit) (b % DIGIT_BIT); 1066 if (D != 0) { 1067 register mp_digit *tmpc, mask, shift; 1068 1069 /* mask */ 1070 mask = (((mp_digit)1) << D) - 1; 1071 1072 /* shift for lsb */ 1073 shift = DIGIT_BIT - D; 1074 1075 /* alias */ 1076 tmpc = c->dp + (c->used - 1); 1077 1078 /* carry */ 1079 r = 0; 1080 for (x = c->used - 1; x >= 0; x--) { 1081 /* get the lower bits of this word in a temp */ 1082 rr = *tmpc & mask; 1083 1084 /* shift the current word and mix in the carry bits from the previous word */ 1085 *tmpc = (*tmpc >> D) | (r << shift); 1086 --tmpc; 1087 1088 /* set the carry to the carry bits of the current word found above */ 1089 r = rr; 1090 } 1091 } 1092 mp_clamp (c); 1093 if (d != NULL) { 1094 mp_exch (&t, d); 1095 } 1096 mp_clear (&t); 1097 return MP_OKAY; 1098} 1099 1100 1101static int mp_init_copy (mp_int * a, mp_int * b) 1102{ 1103 int res; 1104 1105 if ((res = mp_init (a)) != MP_OKAY) { 1106 return res; 1107 } 1108 return mp_copy (b, a); 1109} 1110 1111 1112/* set to zero */ 1113static void mp_zero (mp_int * a) 1114{ 1115 int n; 1116 mp_digit *tmp; 1117 1118 a->sign = MP_ZPOS; 1119 a->used = 0; 1120 1121 tmp = a->dp; 1122 for (n = 0; n < a->alloc; n++) { 1123 *tmp++ = 0; 1124 } 1125} 1126 1127 1128/* copy, b = a */ 1129static int mp_copy (mp_int * a, mp_int * b) 1130{ 1131 int res, n; 1132 1133 /* if dst == src do nothing */ 1134 if (a == b) { 1135 return MP_OKAY; 1136 } 1137 1138 /* grow dest */ 1139 if (b->alloc < a->used) { 1140 if ((res = mp_grow (b, a->used)) != MP_OKAY) { 1141 return res; 1142 } 1143 } 1144 1145 /* zero b and copy the parameters over */ 1146 { 1147 register mp_digit *tmpa, *tmpb; 1148 1149 /* pointer aliases */ 1150 1151 /* source */ 1152 tmpa = a->dp; 1153 1154 /* destination */ 1155 tmpb = b->dp; 1156 1157 /* copy all the digits */ 1158 for (n = 0; n < a->used; n++) { 1159 *tmpb++ = *tmpa++; 1160 } 1161 1162 /* clear high digits */ 1163 for (; n < b->used; n++) { 1164 *tmpb++ = 0; 1165 } 1166 } 1167 1168 /* copy used count and sign */ 1169 b->used = a->used; 1170 b->sign = a->sign; 1171 return MP_OKAY; 1172} 1173 1174 1175/* shift right a certain amount of digits */ 1176static void mp_rshd (mp_int * a, int b) 1177{ 1178 int x; 1179 1180 /* if b <= 0 then ignore it */ 1181 if (b <= 0) { 1182 return; 1183 } 1184 1185 /* if b > used then simply zero it and return */ 1186 if (a->used <= b) { 1187 mp_zero (a); 1188 return; 1189 } 1190 1191 { 1192 register mp_digit *bottom, *top; 1193 1194 /* shift the digits down */ 1195 1196 /* bottom */ 1197 bottom = a->dp; 1198 1199 /* top [offset into digits] */ 1200 top = a->dp + b; 1201 1202 /* this is implemented as a sliding window where 1203 * the window is b-digits long and digits from 1204 * the top of the window are copied to the bottom 1205 * 1206 * e.g. 1207 1208 b-2 | b-1 | b0 | b1 | b2 | ... | bb | ----> 1209 /\ | ----> 1210 \-------------------/ ----> 1211 */ 1212 for (x = 0; x < (a->used - b); x++) { 1213 *bottom++ = *top++; 1214 } 1215 1216 /* zero the top digits */ 1217 for (; x < a->used; x++) { 1218 *bottom++ = 0; 1219 } 1220 } 1221 1222 /* remove excess digits */ 1223 a->used -= b; 1224} 1225 1226 1227/* swap the elements of two integers, for cases where you can't simply swap the 1228 * mp_int pointers around 1229 */ 1230static void mp_exch (mp_int * a, mp_int * b) 1231{ 1232 mp_int t; 1233 1234 t = *a; 1235 *a = *b; 1236 *b = t; 1237} 1238 1239 1240/* trim unused digits 1241 * 1242 * This is used to ensure that leading zero digits are 1243 * trimed and the leading "used" digit will be non-zero 1244 * Typically very fast. Also fixes the sign if there 1245 * are no more leading digits 1246 */ 1247static void mp_clamp (mp_int * a) 1248{ 1249 /* decrease used while the most significant digit is 1250 * zero. 1251 */ 1252 while (a->used > 0 && a->dp[a->used - 1] == 0) { 1253 --(a->used); 1254 } 1255 1256 /* reset the sign flag if used == 0 */ 1257 if (a->used == 0) { 1258 a->sign = MP_ZPOS; 1259 } 1260} 1261 1262 1263/* grow as required */ 1264static int mp_grow (mp_int * a, int size) 1265{ 1266 int i; 1267 mp_digit *tmp; 1268 1269 /* if the alloc size is smaller alloc more ram */ 1270 if (a->alloc < size) { 1271 /* ensure there are always at least MP_PREC digits extra on top */ 1272 size += (MP_PREC * 2) - (size % MP_PREC); 1273 1274 /* reallocate the array a->dp 1275 * 1276 * We store the return in a temporary variable 1277 * in case the operation failed we don't want 1278 * to overwrite the dp member of a. 1279 */ 1280 tmp = OPT_CAST(mp_digit) XREALLOC (a->dp, sizeof (mp_digit) * size); 1281 if (tmp == NULL) { 1282 /* reallocation failed but "a" is still valid [can be freed] */ 1283 return MP_MEM; 1284 } 1285 1286 /* reallocation succeeded so set a->dp */ 1287 a->dp = tmp; 1288 1289 /* zero excess digits */ 1290 i = a->alloc; 1291 a->alloc = size; 1292 for (; i < a->alloc; i++) { 1293 a->dp[i] = 0; 1294 } 1295 } 1296 return MP_OKAY; 1297} 1298 1299 1300#ifdef BN_MP_ABS_C 1301/* b = |a| 1302 * 1303 * Simple function copies the input and fixes the sign to positive 1304 */ 1305static int mp_abs (mp_int * a, mp_int * b) 1306{ 1307 int res; 1308 1309 /* copy a to b */ 1310 if (a != b) { 1311 if ((res = mp_copy (a, b)) != MP_OKAY) { 1312 return res; 1313 } 1314 } 1315 1316 /* force the sign of b to positive */ 1317 b->sign = MP_ZPOS; 1318 1319 return MP_OKAY; 1320} 1321#endif 1322 1323 1324/* set to a digit */ 1325static void mp_set (mp_int * a, mp_digit b) 1326{ 1327 mp_zero (a); 1328 a->dp[0] = b & MP_MASK; 1329 a->used = (a->dp[0] != 0) ? 1 : 0; 1330} 1331 1332 1333#ifndef LTM_NO_NEG_EXP 1334/* b = a/2 */ 1335static int mp_div_2(mp_int * a, mp_int * b) 1336{ 1337 int x, res, oldused; 1338 1339 /* copy */ 1340 if (b->alloc < a->used) { 1341 if ((res = mp_grow (b, a->used)) != MP_OKAY) { 1342 return res; 1343 } 1344 } 1345 1346 oldused = b->used; 1347 b->used = a->used; 1348 { 1349 register mp_digit r, rr, *tmpa, *tmpb; 1350 1351 /* source alias */ 1352 tmpa = a->dp + b->used - 1; 1353 1354 /* dest alias */ 1355 tmpb = b->dp + b->used - 1; 1356 1357 /* carry */ 1358 r = 0; 1359 for (x = b->used - 1; x >= 0; x--) { 1360 /* get the carry for the next iteration */ 1361 rr = *tmpa & 1; 1362 1363 /* shift the current digit, add in carry and store */ 1364 *tmpb-- = (*tmpa-- >> 1) | (r << (DIGIT_BIT - 1)); 1365 1366 /* forward carry to next iteration */ 1367 r = rr; 1368 } 1369 1370 /* zero excess digits */ 1371 tmpb = b->dp + b->used; 1372 for (x = b->used; x < oldused; x++) { 1373 *tmpb++ = 0; 1374 } 1375 } 1376 b->sign = a->sign; 1377 mp_clamp (b); 1378 return MP_OKAY; 1379} 1380#endif /* LTM_NO_NEG_EXP */ 1381 1382 1383/* shift left by a certain bit count */ 1384static int mp_mul_2d (mp_int * a, int b, mp_int * c) 1385{ 1386 mp_digit d; 1387 int res; 1388 1389 /* copy */ 1390 if (a != c) { 1391 if ((res = mp_copy (a, c)) != MP_OKAY) { 1392 return res; 1393 } 1394 } 1395 1396 if (c->alloc < (int)(c->used + b/DIGIT_BIT + 1)) { 1397 if ((res = mp_grow (c, c->used + b / DIGIT_BIT + 1)) != MP_OKAY) { 1398 return res; 1399 } 1400 } 1401 1402 /* shift by as many digits in the bit count */ 1403 if (b >= (int)DIGIT_BIT) { 1404 if ((res = mp_lshd (c, b / DIGIT_BIT)) != MP_OKAY) { 1405 return res; 1406 } 1407 } 1408 1409 /* shift any bit count < DIGIT_BIT */ 1410 d = (mp_digit) (b % DIGIT_BIT); 1411 if (d != 0) { 1412 register mp_digit *tmpc, shift, mask, r, rr; 1413 register int x; 1414 1415 /* bitmask for carries */ 1416 mask = (((mp_digit)1) << d) - 1; 1417 1418 /* shift for msbs */ 1419 shift = DIGIT_BIT - d; 1420 1421 /* alias */ 1422 tmpc = c->dp; 1423 1424 /* carry */ 1425 r = 0; 1426 for (x = 0; x < c->used; x++) { 1427 /* get the higher bits of the current word */ 1428 rr = (*tmpc >> shift) & mask; 1429 1430 /* shift the current word and OR in the carry */ 1431 *tmpc = ((*tmpc << d) | r) & MP_MASK; 1432 ++tmpc; 1433 1434 /* set the carry to the carry bits of the current word */ 1435 r = rr; 1436 } 1437 1438 /* set final carry */ 1439 if (r != 0) { 1440 c->dp[(c->used)++] = r; 1441 } 1442 } 1443 mp_clamp (c); 1444 return MP_OKAY; 1445} 1446 1447 1448#ifdef BN_MP_INIT_MULTI_C 1449static int mp_init_multi(mp_int *mp, ...) 1450{ 1451 mp_err res = MP_OKAY; /* Assume ok until proven otherwise */ 1452 int n = 0; /* Number of ok inits */ 1453 mp_int* cur_arg = mp; 1454 va_list args; 1455 1456 va_start(args, mp); /* init args to next argument from caller */ 1457 while (cur_arg != NULL) { 1458 if (mp_init(cur_arg) != MP_OKAY) { 1459 /* Oops - error! Back-track and mp_clear what we already 1460 succeeded in init-ing, then return error. 1461 */ 1462 va_list clean_args; 1463 1464 /* end the current list */ 1465 va_end(args); 1466 1467 /* now start cleaning up */ 1468 cur_arg = mp; 1469 va_start(clean_args, mp); 1470 while (n--) { 1471 mp_clear(cur_arg); 1472 cur_arg = va_arg(clean_args, mp_int*); 1473 } 1474 va_end(clean_args); 1475 res = MP_MEM; 1476 break; 1477 } 1478 n++; 1479 cur_arg = va_arg(args, mp_int*); 1480 } 1481 va_end(args); 1482 return res; /* Assumed ok, if error flagged above. */ 1483} 1484#endif 1485 1486 1487#ifdef BN_MP_CLEAR_MULTI_C 1488static void mp_clear_multi(mp_int *mp, ...) 1489{ 1490 mp_int* next_mp = mp; 1491 va_list args; 1492 va_start(args, mp); 1493 while (next_mp != NULL) { 1494 mp_clear(next_mp); 1495 next_mp = va_arg(args, mp_int*); 1496 } 1497 va_end(args); 1498} 1499#endif 1500 1501 1502/* shift left a certain amount of digits */ 1503static int mp_lshd (mp_int * a, int b) 1504{ 1505 int x, res; 1506 1507 /* if its less than zero return */ 1508 if (b <= 0) { 1509 return MP_OKAY; 1510 } 1511 1512 /* grow to fit the new digits */ 1513 if (a->alloc < a->used + b) { 1514 if ((res = mp_grow (a, a->used + b)) != MP_OKAY) { 1515 return res; 1516 } 1517 } 1518 1519 { 1520 register mp_digit *top, *bottom; 1521 1522 /* increment the used by the shift amount then copy upwards */ 1523 a->used += b; 1524 1525 /* top */ 1526 top = a->dp + a->used - 1; 1527 1528 /* base */ 1529 bottom = a->dp + a->used - 1 - b; 1530 1531 /* much like mp_rshd this is implemented using a sliding window 1532 * except the window goes the otherway around. Copying from 1533 * the bottom to the top. see bn_mp_rshd.c for more info. 1534 */ 1535 for (x = a->used - 1; x >= b; x--) { 1536 *top-- = *bottom--; 1537 } 1538 1539 /* zero the lower digits */ 1540 top = a->dp; 1541 for (x = 0; x < b; x++) { 1542 *top++ = 0; 1543 } 1544 } 1545 return MP_OKAY; 1546} 1547 1548 1549/* returns the number of bits in an int */ 1550static int mp_count_bits (mp_int * a) 1551{ 1552 int r; 1553 mp_digit q; 1554 1555 /* shortcut */ 1556 if (a->used == 0) { 1557 return 0; 1558 } 1559 1560 /* get number of digits and add that */ 1561 r = (a->used - 1) * DIGIT_BIT; 1562 1563 /* take the last digit and count the bits in it */ 1564 q = a->dp[a->used - 1]; 1565 while (q > ((mp_digit) 0)) { 1566 ++r; 1567 q >>= ((mp_digit) 1); 1568 } 1569 return r; 1570} 1571 1572 1573/* calc a value mod 2**b */ 1574static int mp_mod_2d (mp_int * a, int b, mp_int * c) 1575{ 1576 int x, res; 1577 1578 /* if b is <= 0 then zero the int */ 1579 if (b <= 0) { 1580 mp_zero (c); 1581 return MP_OKAY; 1582 } 1583 1584 /* if the modulus is larger than the value than return */ 1585 if (b >= (int) (a->used * DIGIT_BIT)) { 1586 res = mp_copy (a, c); 1587 return res; 1588 } 1589 1590 /* copy */ 1591 if ((res = mp_copy (a, c)) != MP_OKAY) { 1592 return res; 1593 } 1594 1595 /* zero digits above the last digit of the modulus */ 1596 for (x = (b / DIGIT_BIT) + ((b % DIGIT_BIT) == 0 ? 0 : 1); x < c->used; x++) { 1597 c->dp[x] = 0; 1598 } 1599 /* clear the digit that is not completely outside/inside the modulus */ 1600 c->dp[b / DIGIT_BIT] &= 1601 (mp_digit) ((((mp_digit) 1) << (((mp_digit) b) % DIGIT_BIT)) - ((mp_digit) 1)); 1602 mp_clamp (c); 1603 return MP_OKAY; 1604} 1605 1606 1607#ifdef BN_MP_DIV_SMALL 1608 1609/* slower bit-bang division... also smaller */ 1610static int mp_div(mp_int * a, mp_int * b, mp_int * c, mp_int * d) 1611{ 1612 mp_int ta, tb, tq, q; 1613 int res, n, n2; 1614 1615 /* is divisor zero ? */ 1616 if (mp_iszero (b) == 1) { 1617 return MP_VAL; 1618 } 1619 1620 /* if a < b then q=0, r = a */ 1621 if (mp_cmp_mag (a, b) == MP_LT) { 1622 if (d != NULL) { 1623 res = mp_copy (a, d); 1624 } else { 1625 res = MP_OKAY; 1626 } 1627 if (c != NULL) { 1628 mp_zero (c); 1629 } 1630 return res; 1631 } 1632 1633 /* init our temps */ 1634 if ((res = mp_init_multi(&ta, &tb, &tq, &q, NULL) != MP_OKAY)) { 1635 return res; 1636 } 1637 1638 1639 mp_set(&tq, 1); 1640 n = mp_count_bits(a) - mp_count_bits(b); 1641 if (((res = mp_abs(a, &ta)) != MP_OKAY) || 1642 ((res = mp_abs(b, &tb)) != MP_OKAY) || 1643 ((res = mp_mul_2d(&tb, n, &tb)) != MP_OKAY) || 1644 ((res = mp_mul_2d(&tq, n, &tq)) != MP_OKAY)) { 1645 goto LBL_ERR; 1646 } 1647 1648 while (n-- >= 0) { 1649 if (mp_cmp(&tb, &ta) != MP_GT) { 1650 if (((res = mp_sub(&ta, &tb, &ta)) != MP_OKAY) || 1651 ((res = mp_add(&q, &tq, &q)) != MP_OKAY)) { 1652 goto LBL_ERR; 1653 } 1654 } 1655 if (((res = mp_div_2d(&tb, 1, &tb, NULL)) != MP_OKAY) || 1656 ((res = mp_div_2d(&tq, 1, &tq, NULL)) != MP_OKAY)) { 1657 goto LBL_ERR; 1658 } 1659 } 1660 1661 /* now q == quotient and ta == remainder */ 1662 n = a->sign; 1663 n2 = (a->sign == b->sign ? MP_ZPOS : MP_NEG); 1664 if (c != NULL) { 1665 mp_exch(c, &q); 1666 c->sign = (mp_iszero(c) == MP_YES) ? MP_ZPOS : n2; 1667 } 1668 if (d != NULL) { 1669 mp_exch(d, &ta); 1670 d->sign = (mp_iszero(d) == MP_YES) ? MP_ZPOS : n; 1671 } 1672LBL_ERR: 1673 mp_clear_multi(&ta, &tb, &tq, &q, NULL); 1674 return res; 1675} 1676 1677#else 1678 1679/* integer signed division. 1680 * c*b + d == a [e.g. a/b, c=quotient, d=remainder] 1681 * HAC pp.598 Algorithm 14.20 1682 * 1683 * Note that the description in HAC is horribly 1684 * incomplete. For example, it doesn't consider 1685 * the case where digits are removed from 'x' in 1686 * the inner loop. It also doesn't consider the 1687 * case that y has fewer than three digits, etc.. 1688 * 1689 * The overall algorithm is as described as 1690 * 14.20 from HAC but fixed to treat these cases. 1691*/ 1692static int mp_div (mp_int * a, mp_int * b, mp_int * c, mp_int * d) 1693{ 1694 mp_int q, x, y, t1, t2; 1695 int res, n, t, i, norm, neg; 1696 1697 /* is divisor zero ? */ 1698 if (mp_iszero (b) == 1) { 1699 return MP_VAL; 1700 } 1701 1702 /* if a < b then q=0, r = a */ 1703 if (mp_cmp_mag (a, b) == MP_LT) { 1704 if (d != NULL) { 1705 res = mp_copy (a, d); 1706 } else { 1707 res = MP_OKAY; 1708 } 1709 if (c != NULL) { 1710 mp_zero (c); 1711 } 1712 return res; 1713 } 1714 1715 if ((res = mp_init_size (&q, a->used + 2)) != MP_OKAY) { 1716 return res; 1717 } 1718 q.used = a->used + 2; 1719 1720 if ((res = mp_init (&t1)) != MP_OKAY) { 1721 goto LBL_Q; 1722 } 1723 1724 if ((res = mp_init (&t2)) != MP_OKAY) { 1725 goto LBL_T1; 1726 } 1727 1728 if ((res = mp_init_copy (&x, a)) != MP_OKAY) { 1729 goto LBL_T2; 1730 } 1731 1732 if ((res = mp_init_copy (&y, b)) != MP_OKAY) { 1733 goto LBL_X; 1734 } 1735 1736 /* fix the sign */ 1737 neg = (a->sign == b->sign) ? MP_ZPOS : MP_NEG; 1738 x.sign = y.sign = MP_ZPOS; 1739 1740 /* normalize both x and y, ensure that y >= b/2, [b == 2**DIGIT_BIT] */ 1741 norm = mp_count_bits(&y) % DIGIT_BIT; 1742 if (norm < (int)(DIGIT_BIT-1)) { 1743 norm = (DIGIT_BIT-1) - norm; 1744 if ((res = mp_mul_2d (&x, norm, &x)) != MP_OKAY) { 1745 goto LBL_Y; 1746 } 1747 if ((res = mp_mul_2d (&y, norm, &y)) != MP_OKAY) { 1748 goto LBL_Y; 1749 } 1750 } else { 1751 norm = 0; 1752 } 1753 1754 /* note hac does 0 based, so if used==5 then its 0,1,2,3,4, e.g. use 4 */ 1755 n = x.used - 1; 1756 t = y.used - 1; 1757 1758 /* while (x >= y*b**n-t) do { q[n-t] += 1; x -= y*b**{n-t} } */ 1759 if ((res = mp_lshd (&y, n - t)) != MP_OKAY) { /* y = y*b**{n-t} */ 1760 goto LBL_Y; 1761 } 1762 1763 while (mp_cmp (&x, &y) != MP_LT) { 1764 ++(q.dp[n - t]); 1765 if ((res = mp_sub (&x, &y, &x)) != MP_OKAY) { 1766 goto LBL_Y; 1767 } 1768 } 1769 1770 /* reset y by shifting it back down */ 1771 mp_rshd (&y, n - t); 1772 1773 /* step 3. for i from n down to (t + 1) */ 1774 for (i = n; i >= (t + 1); i--) { 1775 if (i > x.used) { 1776 continue; 1777 } 1778 1779 /* step 3.1 if xi == yt then set q{i-t-1} to b-1, 1780 * otherwise set q{i-t-1} to (xi*b + x{i-1})/yt */ 1781 if (x.dp[i] == y.dp[t]) { 1782 q.dp[i - t - 1] = ((((mp_digit)1) << DIGIT_BIT) - 1); 1783 } else { 1784 mp_word tmp; 1785 tmp = ((mp_word) x.dp[i]) << ((mp_word) DIGIT_BIT); 1786 tmp |= ((mp_word) x.dp[i - 1]); 1787 tmp /= ((mp_word) y.dp[t]); 1788 if (tmp > (mp_word) MP_MASK) 1789 tmp = MP_MASK; 1790 q.dp[i - t - 1] = (mp_digit) (tmp & (mp_word) (MP_MASK)); 1791 } 1792 1793 /* while (q{i-t-1} * (yt * b + y{t-1})) > 1794 xi * b**2 + xi-1 * b + xi-2 1795 1796 do q{i-t-1} -= 1; 1797 */ 1798 q.dp[i - t - 1] = (q.dp[i - t - 1] + 1) & MP_MASK; 1799 do { 1800 q.dp[i - t - 1] = (q.dp[i - t - 1] - 1) & MP_MASK; 1801 1802 /* find left hand */ 1803 mp_zero (&t1); 1804 t1.dp[0] = (t - 1 < 0) ? 0 : y.dp[t - 1]; 1805 t1.dp[1] = y.dp[t]; 1806 t1.used = 2; 1807 if ((res = mp_mul_d (&t1, q.dp[i - t - 1], &t1)) != MP_OKAY) { 1808 goto LBL_Y; 1809 } 1810 1811 /* find right hand */ 1812 t2.dp[0] = (i - 2 < 0) ? 0 : x.dp[i - 2]; 1813 t2.dp[1] = (i - 1 < 0) ? 0 : x.dp[i - 1]; 1814 t2.dp[2] = x.dp[i]; 1815 t2.used = 3; 1816 } while (mp_cmp_mag(&t1, &t2) == MP_GT); 1817 1818 /* step 3.3 x = x - q{i-t-1} * y * b**{i-t-1} */ 1819 if ((res = mp_mul_d (&y, q.dp[i - t - 1], &t1)) != MP_OKAY) { 1820 goto LBL_Y; 1821 } 1822 1823 if ((res = mp_lshd (&t1, i - t - 1)) != MP_OKAY) { 1824 goto LBL_Y; 1825 } 1826 1827 if ((res = mp_sub (&x, &t1, &x)) != MP_OKAY) { 1828 goto LBL_Y; 1829 } 1830 1831 /* if x < 0 then { x = x + y*b**{i-t-1}; q{i-t-1} -= 1; } */ 1832 if (x.sign == MP_NEG) { 1833 if ((res = mp_copy (&y, &t1)) != MP_OKAY) { 1834 goto LBL_Y; 1835 } 1836 if ((res = mp_lshd (&t1, i - t - 1)) != MP_OKAY) { 1837 goto LBL_Y; 1838 } 1839 if ((res = mp_add (&x, &t1, &x)) != MP_OKAY) { 1840 goto LBL_Y; 1841 } 1842 1843 q.dp[i - t - 1] = (q.dp[i - t - 1] - 1UL) & MP_MASK; 1844 } 1845 } 1846 1847 /* now q is the quotient and x is the remainder 1848 * [which we have to normalize] 1849 */ 1850 1851 /* get sign before writing to c */ 1852 x.sign = x.used == 0 ? MP_ZPOS : a->sign; 1853 1854 if (c != NULL) { 1855 mp_clamp (&q); 1856 mp_exch (&q, c); 1857 c->sign = neg; 1858 } 1859 1860 if (d != NULL) { 1861 mp_div_2d (&x, norm, &x, NULL); 1862 mp_exch (&x, d); 1863 } 1864 1865 res = MP_OKAY; 1866 1867LBL_Y:mp_clear (&y); 1868LBL_X:mp_clear (&x); 1869LBL_T2:mp_clear (&t2); 1870LBL_T1:mp_clear (&t1); 1871LBL_Q:mp_clear (&q); 1872 return res; 1873} 1874 1875#endif 1876 1877 1878#ifdef MP_LOW_MEM 1879 #define TAB_SIZE 32 1880#else 1881 #define TAB_SIZE 256 1882#endif 1883 1884static int s_mp_exptmod (mp_int * G, mp_int * X, mp_int * P, mp_int * Y, int redmode) 1885{ 1886 mp_int M[TAB_SIZE], res, mu; 1887 mp_digit buf; 1888 int err, bitbuf, bitcpy, bitcnt, mode, digidx, x, y, winsize; 1889 int (*redux)(mp_int*,mp_int*,mp_int*); 1890 1891 /* find window size */ 1892 x = mp_count_bits (X); 1893 if (x <= 7) { 1894 winsize = 2; 1895 } else if (x <= 36) { 1896 winsize = 3; 1897 } else if (x <= 140) { 1898 winsize = 4; 1899 } else if (x <= 450) { 1900 winsize = 5; 1901 } else if (x <= 1303) { 1902 winsize = 6; 1903 } else if (x <= 3529) { 1904 winsize = 7; 1905 } else { 1906 winsize = 8; 1907 } 1908 1909#ifdef MP_LOW_MEM 1910 if (winsize > 5) { 1911 winsize = 5; 1912 } 1913#endif 1914 1915 /* init M array */ 1916 /* init first cell */ 1917 if ((err = mp_init(&M[1])) != MP_OKAY) { 1918 return err; 1919 } 1920 1921 /* now init the second half of the array */ 1922 for (x = 1<<(winsize-1); x < (1 << winsize); x++) { 1923 if ((err = mp_init(&M[x])) != MP_OKAY) { 1924 for (y = 1<<(winsize-1); y < x; y++) { 1925 mp_clear (&M[y]); 1926 } 1927 mp_clear(&M[1]); 1928 return err; 1929 } 1930 } 1931 1932 /* create mu, used for Barrett reduction */ 1933 if ((err = mp_init (&mu)) != MP_OKAY) { 1934 goto LBL_M; 1935 } 1936 1937 if (redmode == 0) { 1938 if ((err = mp_reduce_setup (&mu, P)) != MP_OKAY) { 1939 goto LBL_MU; 1940 } 1941 redux = mp_reduce; 1942 } else { 1943 if ((err = mp_reduce_2k_setup_l (P, &mu)) != MP_OKAY) { 1944 goto LBL_MU; 1945 } 1946 redux = mp_reduce_2k_l; 1947 } 1948 1949 /* create M table 1950 * 1951 * The M table contains powers of the base, 1952 * e.g. M[x] = G**x mod P 1953 * 1954 * The first half of the table is not 1955 * computed though accept for M[0] and M[1] 1956 */ 1957 if ((err = mp_mod (G, P, &M[1])) != MP_OKAY) { 1958 goto LBL_MU; 1959 } 1960 1961 /* compute the value at M[1<<(winsize-1)] by squaring 1962 * M[1] (winsize-1) times 1963 */ 1964 if ((err = mp_copy (&M[1], &M[1 << (winsize - 1)])) != MP_OKAY) { 1965 goto LBL_MU; 1966 } 1967 1968 for (x = 0; x < (winsize - 1); x++) { 1969 /* square it */ 1970 if ((err = mp_sqr (&M[1 << (winsize - 1)], 1971 &M[1 << (winsize - 1)])) != MP_OKAY) { 1972 goto LBL_MU; 1973 } 1974 1975 /* reduce modulo P */ 1976 if ((err = redux (&M[1 << (winsize - 1)], P, &mu)) != MP_OKAY) { 1977 goto LBL_MU; 1978 } 1979 } 1980 1981 /* create upper table, that is M[x] = M[x-1] * M[1] (mod P) 1982 * for x = (2**(winsize - 1) + 1) to (2**winsize - 1) 1983 */ 1984 for (x = (1 << (winsize - 1)) + 1; x < (1 << winsize); x++) { 1985 if ((err = mp_mul (&M[x - 1], &M[1], &M[x])) != MP_OKAY) { 1986 goto LBL_MU; 1987 } 1988 if ((err = redux (&M[x], P, &mu)) != MP_OKAY) { 1989 goto LBL_MU; 1990 } 1991 } 1992 1993 /* setup result */ 1994 if ((err = mp_init (&res)) != MP_OKAY) { 1995 goto LBL_MU; 1996 } 1997 mp_set (&res, 1); 1998 1999 /* set initial mode and bit cnt */ 2000 mode = 0; 2001 bitcnt = 1; 2002 buf = 0; 2003 digidx = X->used - 1; 2004 bitcpy = 0; 2005 bitbuf = 0; 2006 2007 for (;;) { 2008 /* grab next digit as required */ 2009 if (--bitcnt == 0) { 2010 /* if digidx == -1 we are out of digits */ 2011 if (digidx == -1) { 2012 break; 2013 } 2014 /* read next digit and reset the bitcnt */ 2015 buf = X->dp[digidx--]; 2016 bitcnt = (int) DIGIT_BIT; 2017 } 2018 2019 /* grab the next msb from the exponent */ 2020 y = (buf >> (mp_digit)(DIGIT_BIT - 1)) & 1; 2021 buf <<= (mp_digit)1; 2022 2023 /* if the bit is zero and mode == 0 then we ignore it 2024 * These represent the leading zero bits before the first 1 bit 2025 * in the exponent. Technically this opt is not required but it 2026 * does lower the # of trivial squaring/reductions used 2027 */ 2028 if (mode == 0 && y == 0) { 2029 continue; 2030 } 2031 2032 /* if the bit is zero and mode == 1 then we square */ 2033 if (mode == 1 && y == 0) { 2034 if ((err = mp_sqr (&res, &res)) != MP_OKAY) { 2035 goto LBL_RES; 2036 } 2037 if ((err = redux (&res, P, &mu)) != MP_OKAY) { 2038 goto LBL_RES; 2039 } 2040 continue; 2041 } 2042 2043 /* else we add it to the window */ 2044 bitbuf |= (y << (winsize - ++bitcpy)); 2045 mode = 2; 2046 2047 if (bitcpy == winsize) { 2048 /* ok window is filled so square as required and multiply */ 2049 /* square first */ 2050 for (x = 0; x < winsize; x++) { 2051 if ((err = mp_sqr (&res, &res)) != MP_OKAY) { 2052 goto LBL_RES; 2053 } 2054 if ((err = redux (&res, P, &mu)) != MP_OKAY) { 2055 goto LBL_RES; 2056 } 2057 } 2058 2059 /* then multiply */ 2060 if ((err = mp_mul (&res, &M[bitbuf], &res)) != MP_OKAY) { 2061 goto LBL_RES; 2062 } 2063 if ((err = redux (&res, P, &mu)) != MP_OKAY) { 2064 goto LBL_RES; 2065 } 2066 2067 /* empty window and reset */ 2068 bitcpy = 0; 2069 bitbuf = 0; 2070 mode = 1; 2071 } 2072 } 2073 2074 /* if bits remain then square/multiply */ 2075 if (mode == 2 && bitcpy > 0) { 2076 /* square then multiply if the bit is set */ 2077 for (x = 0; x < bitcpy; x++) { 2078 if ((err = mp_sqr (&res, &res)) != MP_OKAY) { 2079 goto LBL_RES; 2080 } 2081 if ((err = redux (&res, P, &mu)) != MP_OKAY) { 2082 goto LBL_RES; 2083 } 2084 2085 bitbuf <<= 1; 2086 if ((bitbuf & (1 << winsize)) != 0) { 2087 /* then multiply */ 2088 if ((err = mp_mul (&res, &M[1], &res)) != MP_OKAY) { 2089 goto LBL_RES; 2090 } 2091 if ((err = redux (&res, P, &mu)) != MP_OKAY) { 2092 goto LBL_RES; 2093 } 2094 } 2095 } 2096 } 2097 2098 mp_exch (&res, Y); 2099 err = MP_OKAY; 2100LBL_RES:mp_clear (&res); 2101LBL_MU:mp_clear (&mu); 2102LBL_M: 2103 mp_clear(&M[1]); 2104 for (x = 1<<(winsize-1); x < (1 << winsize); x++) { 2105 mp_clear (&M[x]); 2106 } 2107 return err; 2108} 2109 2110 2111/* computes b = a*a */ 2112static int mp_sqr (mp_int * a, mp_int * b) 2113{ 2114 int res; 2115 2116#ifdef BN_MP_TOOM_SQR_C 2117 /* use Toom-Cook? */ 2118 if (a->used >= TOOM_SQR_CUTOFF) { 2119 res = mp_toom_sqr(a, b); 2120 /* Karatsuba? */ 2121 } else 2122#endif 2123#ifdef BN_MP_KARATSUBA_SQR_C 2124if (a->used >= KARATSUBA_SQR_CUTOFF) { 2125 res = mp_karatsuba_sqr (a, b); 2126 } else 2127#endif 2128 { 2129#ifdef BN_FAST_S_MP_SQR_C 2130 /* can we use the fast comba multiplier? */ 2131 if ((a->used * 2 + 1) < MP_WARRAY && 2132 a->used < 2133 (1 << (sizeof(mp_word) * CHAR_BIT - 2*DIGIT_BIT - 1))) { 2134 res = fast_s_mp_sqr (a, b); 2135 } else 2136#endif 2137#ifdef BN_S_MP_SQR_C 2138 res = s_mp_sqr (a, b); 2139#else 2140#error mp_sqr could fail 2141 res = MP_VAL; 2142#endif 2143 } 2144 b->sign = MP_ZPOS; 2145 return res; 2146} 2147 2148 2149/* reduces a modulo n where n is of the form 2**p - d 2150 This differs from reduce_2k since "d" can be larger 2151 than a single digit. 2152*/ 2153static int mp_reduce_2k_l(mp_int *a, mp_int *n, mp_int *d) 2154{ 2155 mp_int q; 2156 int p, res; 2157 2158 if ((res = mp_init(&q)) != MP_OKAY) { 2159 return res; 2160 } 2161 2162 p = mp_count_bits(n); 2163top: 2164 /* q = a/2**p, a = a mod 2**p */ 2165 if ((res = mp_div_2d(a, p, &q, a)) != MP_OKAY) { 2166 goto ERR; 2167 } 2168 2169 /* q = q * d */ 2170 if ((res = mp_mul(&q, d, &q)) != MP_OKAY) { 2171 goto ERR; 2172 } 2173 2174 /* a = a + q */ 2175 if ((res = s_mp_add(a, &q, a)) != MP_OKAY) { 2176 goto ERR; 2177 } 2178 2179 if (mp_cmp_mag(a, n) != MP_LT) { 2180 s_mp_sub(a, n, a); 2181 goto top; 2182 } 2183 2184ERR: 2185 mp_clear(&q); 2186 return res; 2187} 2188 2189 2190/* determines the setup value */ 2191static int mp_reduce_2k_setup_l(mp_int *a, mp_int *d) 2192{ 2193 int res; 2194 mp_int tmp; 2195 2196 if ((res = mp_init(&tmp)) != MP_OKAY) { 2197 return res; 2198 } 2199 2200 if ((res = mp_2expt(&tmp, mp_count_bits(a))) != MP_OKAY) { 2201 goto ERR; 2202 } 2203 2204 if ((res = s_mp_sub(&tmp, a, d)) != MP_OKAY) { 2205 goto ERR; 2206 } 2207 2208ERR: 2209 mp_clear(&tmp); 2210 return res; 2211} 2212 2213 2214/* computes a = 2**b 2215 * 2216 * Simple algorithm which zeroes the int, grows it then just sets one bit 2217 * as required. 2218 */ 2219static int mp_2expt (mp_int * a, int b) 2220{ 2221 int res; 2222 2223 /* zero a as per default */ 2224 mp_zero (a); 2225 2226 /* grow a to accommodate the single bit */ 2227 if ((res = mp_grow (a, b / DIGIT_BIT + 1)) != MP_OKAY) { 2228 return res; 2229 } 2230 2231 /* set the used count of where the bit will go */ 2232 a->used = b / DIGIT_BIT + 1; 2233 2234 /* put the single bit in its place */ 2235 a->dp[b / DIGIT_BIT] = ((mp_digit)1) << (b % DIGIT_BIT); 2236 2237 return MP_OKAY; 2238} 2239 2240 2241/* pre-calculate the value required for Barrett reduction 2242 * For a given modulus "b" it calulates the value required in "a" 2243 */ 2244static int mp_reduce_setup (mp_int * a, mp_int * b) 2245{ 2246 int res; 2247 2248 if ((res = mp_2expt (a, b->used * 2 * DIGIT_BIT)) != MP_OKAY) { 2249 return res; 2250 } 2251 return mp_div (a, b, a, NULL); 2252} 2253 2254 2255/* reduces x mod m, assumes 0 < x < m**2, mu is 2256 * precomputed via mp_reduce_setup. 2257 * From HAC pp.604 Algorithm 14.42 2258 */ 2259static int mp_reduce (mp_int * x, mp_int * m, mp_int * mu) 2260{ 2261 mp_int q; 2262 int res, um = m->used; 2263 2264 /* q = x */ 2265 if ((res = mp_init_copy (&q, x)) != MP_OKAY) { 2266 return res; 2267 } 2268 2269 /* q1 = x / b**(k-1) */ 2270 mp_rshd (&q, um - 1); 2271 2272 /* according to HAC this optimization is ok */ 2273 if (((unsigned long) um) > (((mp_digit)1) << (DIGIT_BIT - 1))) { 2274 if ((res = mp_mul (&q, mu, &q)) != MP_OKAY) { 2275 goto CLEANUP; 2276 } 2277 } else { 2278#ifdef BN_S_MP_MUL_HIGH_DIGS_C 2279 if ((res = s_mp_mul_high_digs (&q, mu, &q, um)) != MP_OKAY) { 2280 goto CLEANUP; 2281 } 2282#elif defined(BN_FAST_S_MP_MUL_HIGH_DIGS_C) 2283 if ((res = fast_s_mp_mul_high_digs (&q, mu, &q, um)) != MP_OKAY) { 2284 goto CLEANUP; 2285 } 2286#else 2287 { 2288#error mp_reduce would always fail 2289 res = MP_VAL; 2290 goto CLEANUP; 2291 } 2292#endif 2293 } 2294 2295 /* q3 = q2 / b**(k+1) */ 2296 mp_rshd (&q, um + 1); 2297 2298 /* x = x mod b**(k+1), quick (no division) */ 2299 if ((res = mp_mod_2d (x, DIGIT_BIT * (um + 1), x)) != MP_OKAY) { 2300 goto CLEANUP; 2301 } 2302 2303 /* q = q * m mod b**(k+1), quick (no division) */ 2304 if ((res = s_mp_mul_digs (&q, m, &q, um + 1)) != MP_OKAY) { 2305 goto CLEANUP; 2306 } 2307 2308 /* x = x - q */ 2309 if ((res = mp_sub (x, &q, x)) != MP_OKAY) { 2310 goto CLEANUP; 2311 } 2312 2313 /* If x < 0, add b**(k+1) to it */ 2314 if (mp_cmp_d (x, 0) == MP_LT) { 2315 mp_set (&q, 1); 2316 if ((res = mp_lshd (&q, um + 1)) != MP_OKAY) { 2317 goto CLEANUP; 2318 } 2319 if ((res = mp_add (x, &q, x)) != MP_OKAY) { 2320 goto CLEANUP; 2321 } 2322 } 2323 2324 /* Back off if it's too big */ 2325 while (mp_cmp (x, m) != MP_LT) { 2326 if ((res = s_mp_sub (x, m, x)) != MP_OKAY) { 2327 goto CLEANUP; 2328 } 2329 } 2330 2331CLEANUP: 2332 mp_clear (&q); 2333 2334 return res; 2335} 2336 2337 2338/* multiplies |a| * |b| and only computes up to digs digits of result 2339 * HAC pp. 595, Algorithm 14.12 Modified so you can control how 2340 * many digits of output are created. 2341 */ 2342static int s_mp_mul_digs (mp_int * a, mp_int * b, mp_int * c, int digs) 2343{ 2344 mp_int t; 2345 int res, pa, pb, ix, iy; 2346 mp_digit u; 2347 mp_word r; 2348 mp_digit tmpx, *tmpt, *tmpy; 2349 2350#ifdef BN_FAST_S_MP_MUL_DIGS_C 2351 /* can we use the fast multiplier? */ 2352 if (((digs) < MP_WARRAY) && 2353 MIN (a->used, b->used) < 2354 (1 << ((CHAR_BIT * sizeof (mp_word)) - (2 * DIGIT_BIT)))) { 2355 return fast_s_mp_mul_digs (a, b, c, digs); 2356 } 2357#endif 2358 2359 if ((res = mp_init_size (&t, digs)) != MP_OKAY) { 2360 return res; 2361 } 2362 t.used = digs; 2363 2364 /* compute the digits of the product directly */ 2365 pa = a->used; 2366 for (ix = 0; ix < pa; ix++) { 2367 /* set the carry to zero */ 2368 u = 0; 2369 2370 /* limit ourselves to making digs digits of output */ 2371 pb = MIN (b->used, digs - ix); 2372 2373 /* setup some aliases */ 2374 /* copy of the digit from a used within the nested loop */ 2375 tmpx = a->dp[ix]; 2376 2377 /* an alias for the destination shifted ix places */ 2378 tmpt = t.dp + ix; 2379 2380 /* an alias for the digits of b */ 2381 tmpy = b->dp; 2382 2383 /* compute the columns of the output and propagate the carry */ 2384 for (iy = 0; iy < pb; iy++) { 2385 /* compute the column as a mp_word */ 2386 r = ((mp_word)*tmpt) + 2387 ((mp_word)tmpx) * ((mp_word)*tmpy++) + 2388 ((mp_word) u); 2389 2390 /* the new column is the lower part of the result */ 2391 *tmpt++ = (mp_digit) (r & ((mp_word) MP_MASK)); 2392 2393 /* get the carry word from the result */ 2394 u = (mp_digit) (r >> ((mp_word) DIGIT_BIT)); 2395 } 2396 /* set carry if it is placed below digs */ 2397 if (ix + iy < digs) { 2398 *tmpt = u; 2399 } 2400 } 2401 2402 mp_clamp (&t); 2403 mp_exch (&t, c); 2404 2405 mp_clear (&t); 2406 return MP_OKAY; 2407} 2408 2409 2410#ifdef BN_FAST_S_MP_MUL_DIGS_C 2411/* Fast (comba) multiplier 2412 * 2413 * This is the fast column-array [comba] multiplier. It is 2414 * designed to compute the columns of the product first 2415 * then handle the carries afterwards. This has the effect 2416 * of making the nested loops that compute the columns very 2417 * simple and schedulable on super-scalar processors. 2418 * 2419 * This has been modified to produce a variable number of 2420 * digits of output so if say only a half-product is required 2421 * you don't have to compute the upper half (a feature 2422 * required for fast Barrett reduction). 2423 * 2424 * Based on Algorithm 14.12 on pp.595 of HAC. 2425 * 2426 */ 2427static int fast_s_mp_mul_digs (mp_int * a, mp_int * b, mp_int * c, int digs) 2428{ 2429 int olduse, res, pa, ix, iz; 2430 mp_digit W[MP_WARRAY]; 2431 register mp_word _W; 2432 2433 /* grow the destination as required */ 2434 if (c->alloc < digs) { 2435 if ((res = mp_grow (c, digs)) != MP_OKAY) { 2436 return res; 2437 } 2438 } 2439 2440 /* number of output digits to produce */ 2441 pa = MIN(digs, a->used + b->used); 2442 2443 /* clear the carry */ 2444 _W = 0; 2445 for (ix = 0; ix < pa; ix++) { 2446 int tx, ty; 2447 int iy; 2448 mp_digit *tmpx, *tmpy; 2449 2450 /* get offsets into the two bignums */ 2451 ty = MIN(b->used-1, ix); 2452 tx = ix - ty; 2453 2454 /* setup temp aliases */ 2455 tmpx = a->dp + tx; 2456 tmpy = b->dp + ty; 2457 2458 /* this is the number of times the loop will iterrate, essentially 2459 while (tx++ < a->used && ty-- >= 0) { ... } 2460 */ 2461 iy = MIN(a->used-tx, ty+1); 2462 2463 /* execute loop */ 2464 for (iz = 0; iz < iy; ++iz) { 2465 _W += ((mp_word)*tmpx++)*((mp_word)*tmpy--); 2466 2467 } 2468 2469 /* store term */ 2470 W[ix] = ((mp_digit)_W) & MP_MASK; 2471 2472 /* make next carry */ 2473 _W = _W >> ((mp_word)DIGIT_BIT); 2474 } 2475 2476 /* setup dest */ 2477 olduse = c->used; 2478 c->used = pa; 2479 2480 { 2481 register mp_digit *tmpc; 2482 tmpc = c->dp; 2483 for (ix = 0; ix < pa+1; ix++) { 2484 /* now extract the previous digit [below the carry] */ 2485 *tmpc++ = W[ix]; 2486 } 2487 2488 /* clear unused digits [that existed in the old copy of c] */ 2489 for (; ix < olduse; ix++) { 2490 *tmpc++ = 0; 2491 } 2492 } 2493 mp_clamp (c); 2494 return MP_OKAY; 2495} 2496#endif /* BN_FAST_S_MP_MUL_DIGS_C */ 2497 2498 2499/* init an mp_init for a given size */ 2500static int mp_init_size (mp_int * a, int size) 2501{ 2502 int x; 2503 2504 /* pad size so there are always extra digits */ 2505 size += (MP_PREC * 2) - (size % MP_PREC); 2506 2507 /* alloc mem */ 2508 a->dp = OPT_CAST(mp_digit) XMALLOC (sizeof (mp_digit) * size); 2509 if (a->dp == NULL) { 2510 return MP_MEM; 2511 } 2512 2513 /* set the members */ 2514 a->used = 0; 2515 a->alloc = size; 2516 a->sign = MP_ZPOS; 2517 2518 /* zero the digits */ 2519 for (x = 0; x < size; x++) { 2520 a->dp[x] = 0; 2521 } 2522 2523 return MP_OKAY; 2524} 2525 2526 2527/* low level squaring, b = a*a, HAC pp.596-597, Algorithm 14.16 */ 2528static int s_mp_sqr (mp_int * a, mp_int * b) 2529{ 2530 mp_int t; 2531 int res, ix, iy, pa; 2532 mp_word r; 2533 mp_digit u, tmpx, *tmpt; 2534 2535 pa = a->used; 2536 if ((res = mp_init_size (&t, 2*pa + 1)) != MP_OKAY) { 2537 return res; 2538 } 2539 2540 /* default used is maximum possible size */ 2541 t.used = 2*pa + 1; 2542 2543 for (ix = 0; ix < pa; ix++) { 2544 /* first calculate the digit at 2*ix */ 2545 /* calculate double precision result */ 2546 r = ((mp_word) t.dp[2*ix]) + 2547 ((mp_word)a->dp[ix])*((mp_word)a->dp[ix]); 2548 2549 /* store lower part in result */ 2550 t.dp[ix+ix] = (mp_digit) (r & ((mp_word) MP_MASK)); 2551 2552 /* get the carry */ 2553 u = (mp_digit)(r >> ((mp_word) DIGIT_BIT)); 2554 2555 /* left hand side of A[ix] * A[iy] */ 2556 tmpx = a->dp[ix]; 2557 2558 /* alias for where to store the results */ 2559 tmpt = t.dp + (2*ix + 1); 2560 2561 for (iy = ix + 1; iy < pa; iy++) { 2562 /* first calculate the product */ 2563 r = ((mp_word)tmpx) * ((mp_word)a->dp[iy]); 2564 2565 /* now calculate the double precision result, note we use 2566 * addition instead of *2 since it's easier to optimize 2567 */ 2568 r = ((mp_word) *tmpt) + r + r + ((mp_word) u); 2569 2570 /* store lower part */ 2571 *tmpt++ = (mp_digit) (r & ((mp_word) MP_MASK)); 2572 2573 /* get carry */ 2574 u = (mp_digit)(r >> ((mp_word) DIGIT_BIT)); 2575 } 2576 /* propagate upwards */ 2577 while (u != ((mp_digit) 0)) { 2578 r = ((mp_word) *tmpt) + ((mp_word) u); 2579 *tmpt++ = (mp_digit) (r & ((mp_word) MP_MASK)); 2580 u = (mp_digit)(r >> ((mp_word) DIGIT_BIT)); 2581 } 2582 } 2583 2584 mp_clamp (&t); 2585 mp_exch (&t, b); 2586 mp_clear (&t); 2587 return MP_OKAY; 2588} 2589 2590 2591/* multiplies |a| * |b| and does not compute the lower digs digits 2592 * [meant to get the higher part of the product] 2593 */ 2594static int s_mp_mul_high_digs (mp_int * a, mp_int * b, mp_int * c, int digs) 2595{ 2596 mp_int t; 2597 int res, pa, pb, ix, iy; 2598 mp_digit u; 2599 mp_word r; 2600 mp_digit tmpx, *tmpt, *tmpy; 2601 2602 /* can we use the fast multiplier? */ 2603#ifdef BN_FAST_S_MP_MUL_HIGH_DIGS_C 2604 if (((a->used + b->used + 1) < MP_WARRAY) 2605 && MIN (a->used, b->used) < (1 << ((CHAR_BIT * sizeof (mp_word)) - (2 * DIGIT_BIT)))) { 2606 return fast_s_mp_mul_high_digs (a, b, c, digs); 2607 } 2608#endif 2609 2610 if ((res = mp_init_size (&t, a->used + b->used + 1)) != MP_OKAY) { 2611 return res; 2612 } 2613 t.used = a->used + b->used + 1; 2614 2615 pa = a->used; 2616 pb = b->used; 2617 for (ix = 0; ix < pa; ix++) { 2618 /* clear the carry */ 2619 u = 0; 2620 2621 /* left hand side of A[ix] * B[iy] */ 2622 tmpx = a->dp[ix]; 2623 2624 /* alias to the address of where the digits will be stored */ 2625 tmpt = &(t.dp[digs]); 2626 2627 /* alias for where to read the right hand side from */ 2628 tmpy = b->dp + (digs - ix); 2629 2630 for (iy = digs - ix; iy < pb; iy++) { 2631 /* calculate the double precision result */ 2632 r = ((mp_word)*tmpt) + 2633 ((mp_word)tmpx) * ((mp_word)*tmpy++) + 2634 ((mp_word) u); 2635 2636 /* get the lower part */ 2637 *tmpt++ = (mp_digit) (r & ((mp_word) MP_MASK)); 2638 2639 /* carry the carry */ 2640 u = (mp_digit) (r >> ((mp_word) DIGIT_BIT)); 2641 } 2642 *tmpt = u; 2643 } 2644 mp_clamp (&t); 2645 mp_exch (&t, c); 2646 mp_clear (&t); 2647 return MP_OKAY; 2648} 2649 2650 2651#ifdef BN_MP_MONTGOMERY_SETUP_C 2652/* setups the montgomery reduction stuff */ 2653static int 2654mp_montgomery_setup (mp_int * n, mp_digit * rho) 2655{ 2656 mp_digit x, b; 2657 2658/* fast inversion mod 2**k 2659 * 2660 * Based on the fact that 2661 * 2662 * XA = 1 (mod 2**n) => (X(2-XA)) A = 1 (mod 2**2n) 2663 * => 2*X*A - X*X*A*A = 1 2664 * => 2*(1) - (1) = 1 2665 */ 2666 b = n->dp[0]; 2667 2668 if ((b & 1) == 0) { 2669 return MP_VAL; 2670 } 2671 2672 x = (((b + 2) & 4) << 1) + b; /* here x*a==1 mod 2**4 */ 2673 x *= 2 - b * x; /* here x*a==1 mod 2**8 */ 2674#if !defined(MP_8BIT) 2675 x *= 2 - b * x; /* here x*a==1 mod 2**16 */ 2676#endif 2677#if defined(MP_64BIT) || !(defined(MP_8BIT) || defined(MP_16BIT)) 2678 x *= 2 - b * x; /* here x*a==1 mod 2**32 */ 2679#endif 2680#ifdef MP_64BIT 2681 x *= 2 - b * x; /* here x*a==1 mod 2**64 */ 2682#endif 2683 2684 /* rho = -1/m mod b */ 2685 *rho = (unsigned long)(((mp_word)1 << ((mp_word) DIGIT_BIT)) - x) & MP_MASK; 2686 2687 return MP_OKAY; 2688} 2689#endif 2690 2691 2692#ifdef BN_FAST_MP_MONTGOMERY_REDUCE_C 2693/* computes xR**-1 == x (mod N) via Montgomery Reduction 2694 * 2695 * This is an optimized implementation of montgomery_reduce 2696 * which uses the comba method to quickly calculate the columns of the 2697 * reduction. 2698 * 2699 * Based on Algorithm 14.32 on pp.601 of HAC. 2700*/ 2701static int fast_mp_montgomery_reduce (mp_int * x, mp_int * n, mp_digit rho) 2702{ 2703 int ix, res, olduse; 2704 mp_word W[MP_WARRAY]; 2705 2706 /* get old used count */ 2707 olduse = x->used; 2708 2709 /* grow a as required */ 2710 if (x->alloc < n->used + 1) { 2711 if ((res = mp_grow (x, n->used + 1)) != MP_OKAY) { 2712 return res; 2713 } 2714 } 2715 2716 /* first we have to get the digits of the input into 2717 * an array of double precision words W[...] 2718 */ 2719 { 2720 register mp_word *_W; 2721 register mp_digit *tmpx; 2722 2723 /* alias for the W[] array */ 2724 _W = W; 2725 2726 /* alias for the digits of x*/ 2727 tmpx = x->dp; 2728 2729 /* copy the digits of a into W[0..a->used-1] */ 2730 for (ix = 0; ix < x->used; ix++) { 2731 *_W++ = *tmpx++; 2732 } 2733 2734 /* zero the high words of W[a->used..m->used*2] */ 2735 for (; ix < n->used * 2 + 1; ix++) { 2736 *_W++ = 0; 2737 } 2738 } 2739 2740 /* now we proceed to zero successive digits 2741 * from the least significant upwards 2742 */ 2743 for (ix = 0; ix < n->used; ix++) { 2744 /* mu = ai * m' mod b 2745 * 2746 * We avoid a double precision multiplication (which isn't required) 2747 * by casting the value down to a mp_digit. Note this requires 2748 * that W[ix-1] have the carry cleared (see after the inner loop) 2749 */ 2750 register mp_digit mu; 2751 mu = (mp_digit) (((W[ix] & MP_MASK) * rho) & MP_MASK); 2752 2753 /* a = a + mu * m * b**i 2754 * 2755 * This is computed in place and on the fly. The multiplication 2756 * by b**i is handled by offseting which columns the results 2757 * are added to. 2758 * 2759 * Note the comba method normally doesn't handle carries in the 2760 * inner loop In this case we fix the carry from the previous 2761 * column since the Montgomery reduction requires digits of the 2762 * result (so far) [see above] to work. This is 2763 * handled by fixing up one carry after the inner loop. The 2764 * carry fixups are done in order so after these loops the 2765 * first m->used words of W[] have the carries fixed 2766 */ 2767 { 2768 register int iy; 2769 register mp_digit *tmpn; 2770 register mp_word *_W; 2771 2772 /* alias for the digits of the modulus */ 2773 tmpn = n->dp; 2774 2775 /* Alias for the columns set by an offset of ix */ 2776 _W = W + ix; 2777 2778 /* inner loop */ 2779 for (iy = 0; iy < n->used; iy++) { 2780 *_W++ += ((mp_word)mu) * ((mp_word)*tmpn++); 2781 } 2782 } 2783 2784 /* now fix carry for next digit, W[ix+1] */ 2785 W[ix + 1] += W[ix] >> ((mp_word) DIGIT_BIT); 2786 } 2787 2788 /* now we have to propagate the carries and 2789 * shift the words downward [all those least 2790 * significant digits we zeroed]. 2791 */ 2792 { 2793 register mp_digit *tmpx; 2794 register mp_word *_W, *_W1; 2795 2796 /* nox fix rest of carries */ 2797 2798 /* alias for current word */ 2799 _W1 = W + ix; 2800 2801 /* alias for next word, where the carry goes */ 2802 _W = W + ++ix; 2803 2804 for (; ix <= n->used * 2 + 1; ix++) { 2805 *_W++ += *_W1++ >> ((mp_word) DIGIT_BIT); 2806 } 2807 2808 /* copy out, A = A/b**n 2809 * 2810 * The result is A/b**n but instead of converting from an 2811 * array of mp_word to mp_digit than calling mp_rshd 2812 * we just copy them in the right order 2813 */ 2814 2815 /* alias for destination word */ 2816 tmpx = x->dp; 2817 2818 /* alias for shifted double precision result */ 2819 _W = W + n->used; 2820 2821 for (ix = 0; ix < n->used + 1; ix++) { 2822 *tmpx++ = (mp_digit)(*_W++ & ((mp_word) MP_MASK)); 2823 } 2824 2825 /* zero oldused digits, if the input a was larger than 2826 * m->used+1 we'll have to clear the digits 2827 */ 2828 for (; ix < olduse; ix++) { 2829 *tmpx++ = 0; 2830 } 2831 } 2832 2833 /* set the max used and clamp */ 2834 x->used = n->used + 1; 2835 mp_clamp (x); 2836 2837 /* if A >= m then A = A - m */ 2838 if (mp_cmp_mag (x, n) != MP_LT) { 2839 return s_mp_sub (x, n, x); 2840 } 2841 return MP_OKAY; 2842} 2843#endif 2844 2845 2846#ifdef BN_MP_MUL_2_C 2847/* b = a*2 */ 2848static int mp_mul_2(mp_int * a, mp_int * b) 2849{ 2850 int x, res, oldused; 2851 2852 /* grow to accommodate result */ 2853 if (b->alloc < a->used + 1) { 2854 if ((res = mp_grow (b, a->used + 1)) != MP_OKAY) { 2855 return res; 2856 } 2857 } 2858 2859 oldused = b->used; 2860 b->used = a->used; 2861 2862 { 2863 register mp_digit r, rr, *tmpa, *tmpb; 2864 2865 /* alias for source */ 2866 tmpa = a->dp; 2867 2868 /* alias for dest */ 2869 tmpb = b->dp; 2870 2871 /* carry */ 2872 r = 0; 2873 for (x = 0; x < a->used; x++) { 2874 2875 /* get what will be the *next* carry bit from the 2876 * MSB of the current digit 2877 */ 2878 rr = *tmpa >> ((mp_digit)(DIGIT_BIT - 1)); 2879 2880 /* now shift up this digit, add in the carry [from the previous] */ 2881 *tmpb++ = ((*tmpa++ << ((mp_digit)1)) | r) & MP_MASK; 2882 2883 /* copy the carry that would be from the source 2884 * digit into the next iteration 2885 */ 2886 r = rr; 2887 } 2888 2889 /* new leading digit? */ 2890 if (r != 0) { 2891 /* add a MSB which is always 1 at this point */ 2892 *tmpb = 1; 2893 ++(b->used); 2894 } 2895 2896 /* now zero any excess digits on the destination 2897 * that we didn't write to 2898 */ 2899 tmpb = b->dp + b->used; 2900 for (x = b->used; x < oldused; x++) { 2901 *tmpb++ = 0; 2902 } 2903 } 2904 b->sign = a->sign; 2905 return MP_OKAY; 2906} 2907#endif 2908 2909 2910#ifdef BN_MP_MONTGOMERY_CALC_NORMALIZATION_C 2911/* 2912 * shifts with subtractions when the result is greater than b. 2913 * 2914 * The method is slightly modified to shift B unconditionally up to just under 2915 * the leading bit of b. This saves a lot of multiple precision shifting. 2916 */ 2917static int mp_montgomery_calc_normalization (mp_int * a, mp_int * b) 2918{ 2919 int x, bits, res; 2920 2921 /* how many bits of last digit does b use */ 2922 bits = mp_count_bits (b) % DIGIT_BIT; 2923 2924 if (b->used > 1) { 2925 if ((res = mp_2expt (a, (b->used - 1) * DIGIT_BIT + bits - 1)) != MP_OKAY) { 2926 return res; 2927 } 2928 } else { 2929 mp_set(a, 1); 2930 bits = 1; 2931 } 2932 2933 2934 /* now compute C = A * B mod b */ 2935 for (x = bits - 1; x < (int)DIGIT_BIT; x++) { 2936 if ((res = mp_mul_2 (a, a)) != MP_OKAY) { 2937 return res; 2938 } 2939 if (mp_cmp_mag (a, b) != MP_LT) { 2940 if ((res = s_mp_sub (a, b, a)) != MP_OKAY) { 2941 return res; 2942 } 2943 } 2944 } 2945 2946 return MP_OKAY; 2947} 2948#endif 2949 2950 2951#ifdef BN_MP_EXPTMOD_FAST_C 2952/* computes Y == G**X mod P, HAC pp.616, Algorithm 14.85 2953 * 2954 * Uses a left-to-right k-ary sliding window to compute the modular exponentiation. 2955 * The value of k changes based on the size of the exponent. 2956 * 2957 * Uses Montgomery or Diminished Radix reduction [whichever appropriate] 2958 */ 2959 2960static int mp_exptmod_fast (mp_int * G, mp_int * X, mp_int * P, mp_int * Y, int redmode) 2961{ 2962 mp_int M[TAB_SIZE], res; 2963 mp_digit buf, mp; 2964 int err, bitbuf, bitcpy, bitcnt, mode, digidx, x, y, winsize; 2965 2966 /* use a pointer to the reduction algorithm. This allows us to use 2967 * one of many reduction algorithms without modding the guts of 2968 * the code with if statements everywhere. 2969 */ 2970 int (*redux)(mp_int*,mp_int*,mp_digit); 2971 2972 /* find window size */ 2973 x = mp_count_bits (X); 2974 if (x <= 7) { 2975 winsize = 2; 2976 } else if (x <= 36) { 2977 winsize = 3; 2978 } else if (x <= 140) { 2979 winsize = 4; 2980 } else if (x <= 450) { 2981 winsize = 5; 2982 } else if (x <= 1303) { 2983 winsize = 6; 2984 } else if (x <= 3529) { 2985 winsize = 7; 2986 } else { 2987 winsize = 8; 2988 } 2989 2990#ifdef MP_LOW_MEM 2991 if (winsize > 5) { 2992 winsize = 5; 2993 } 2994#endif 2995 2996 /* init M array */ 2997 /* init first cell */ 2998 if ((err = mp_init(&M[1])) != MP_OKAY) { 2999 return err; 3000 } 3001 3002 /* now init the second half of the array */ 3003 for (x = 1<<(winsize-1); x < (1 << winsize); x++) { 3004 if ((err = mp_init(&M[x])) != MP_OKAY) { 3005 for (y = 1<<(winsize-1); y < x; y++) { 3006 mp_clear (&M[y]); 3007 } 3008 mp_clear(&M[1]); 3009 return err; 3010 } 3011 } 3012 3013 /* determine and setup reduction code */ 3014 if (redmode == 0) { 3015#ifdef BN_MP_MONTGOMERY_SETUP_C 3016 /* now setup montgomery */ 3017 if ((err = mp_montgomery_setup (P, &mp)) != MP_OKAY) { 3018 goto LBL_M; 3019 } 3020#else 3021 err = MP_VAL; 3022 goto LBL_M; 3023#endif 3024 3025 /* automatically pick the comba one if available (saves quite a few calls/ifs) */ 3026#ifdef BN_FAST_MP_MONTGOMERY_REDUCE_C 3027 if (((P->used * 2 + 1) < MP_WARRAY) && 3028 P->used < (1 << ((CHAR_BIT * sizeof (mp_word)) - (2 * DIGIT_BIT)))) { 3029 redux = fast_mp_montgomery_reduce; 3030 } else 3031#endif 3032 { 3033#ifdef BN_MP_MONTGOMERY_REDUCE_C 3034 /* use slower baseline Montgomery method */ 3035 redux = mp_montgomery_reduce; 3036#else 3037 err = MP_VAL; 3038 goto LBL_M; 3039#endif 3040 } 3041 } else if (redmode == 1) { 3042#if defined(BN_MP_DR_SETUP_C) && defined(BN_MP_DR_REDUCE_C) 3043 /* setup DR reduction for moduli of the form B**k - b */ 3044 mp_dr_setup(P, &mp); 3045 redux = mp_dr_reduce; 3046#else 3047 err = MP_VAL; 3048 goto LBL_M; 3049#endif 3050 } else { 3051#if defined(BN_MP_REDUCE_2K_SETUP_C) && defined(BN_MP_REDUCE_2K_C) 3052 /* setup DR reduction for moduli of the form 2**k - b */ 3053 if ((err = mp_reduce_2k_setup(P, &mp)) != MP_OKAY) { 3054 goto LBL_M; 3055 } 3056 redux = mp_reduce_2k; 3057#else 3058 err = MP_VAL; 3059 goto LBL_M; 3060#endif 3061 } 3062 3063 /* setup result */ 3064 if ((err = mp_init (&res)) != MP_OKAY) { 3065 goto LBL_M; 3066 } 3067 3068 /* create M table 3069 * 3070 3071 * 3072 * The first half of the table is not computed though accept for M[0] and M[1] 3073 */ 3074 3075 if (redmode == 0) { 3076#ifdef BN_MP_MONTGOMERY_CALC_NORMALIZATION_C 3077 /* now we need R mod m */ 3078 if ((err = mp_montgomery_calc_normalization (&res, P)) != MP_OKAY) { 3079 goto LBL_RES; 3080 } 3081#else 3082 err = MP_VAL; 3083 goto LBL_RES; 3084#endif 3085 3086 /* now set M[1] to G * R mod m */ 3087 if ((err = mp_mulmod (G, &res, P, &M[1])) != MP_OKAY) { 3088 goto LBL_RES; 3089 } 3090 } else { 3091 mp_set(&res, 1); 3092 if ((err = mp_mod(G, P, &M[1])) != MP_OKAY) { 3093 goto LBL_RES; 3094 } 3095 } 3096 3097 /* compute the value at M[1<<(winsize-1)] by squaring M[1] (winsize-1) times */ 3098 if ((err = mp_copy (&M[1], &M[1 << (winsize - 1)])) != MP_OKAY) { 3099 goto LBL_RES; 3100 } 3101 3102 for (x = 0; x < (winsize - 1); x++) { 3103 if ((err = mp_sqr (&M[1 << (winsize - 1)], &M[1 << (winsize - 1)])) != MP_OKAY) { 3104 goto LBL_RES; 3105 } 3106 if ((err = redux (&M[1 << (winsize - 1)], P, mp)) != MP_OKAY) { 3107 goto LBL_RES; 3108 } 3109 } 3110 3111 /* create upper table */ 3112 for (x = (1 << (winsize - 1)) + 1; x < (1 << winsize); x++) { 3113 if ((err = mp_mul (&M[x - 1], &M[1], &M[x])) != MP_OKAY) { 3114 goto LBL_RES; 3115 } 3116 if ((err = redux (&M[x], P, mp)) != MP_OKAY) { 3117 goto LBL_RES; 3118 } 3119 } 3120 3121 /* set initial mode and bit cnt */ 3122 mode = 0; 3123 bitcnt = 1; 3124 buf = 0; 3125 digidx = X->used - 1; 3126 bitcpy = 0; 3127 bitbuf = 0; 3128 3129 for (;;) { 3130 /* grab next digit as required */ 3131 if (--bitcnt == 0) { 3132 /* if digidx == -1 we are out of digits so break */ 3133 if (digidx == -1) { 3134 break; 3135 } 3136 /* read next digit and reset bitcnt */ 3137 buf = X->dp[digidx--]; 3138 bitcnt = (int)DIGIT_BIT; 3139 } 3140 3141 /* grab the next msb from the exponent */ 3142 y = (mp_digit)(buf >> (DIGIT_BIT - 1)) & 1; 3143 buf <<= (mp_digit)1; 3144 3145 /* if the bit is zero and mode == 0 then we ignore it 3146 * These represent the leading zero bits before the first 1 bit 3147 * in the exponent. Technically this opt is not required but it 3148 * does lower the # of trivial squaring/reductions used 3149 */ 3150 if (mode == 0 && y == 0) { 3151 continue; 3152 } 3153 3154 /* if the bit is zero and mode == 1 then we square */ 3155 if (mode == 1 && y == 0) { 3156 if ((err = mp_sqr (&res, &res)) != MP_OKAY) { 3157 goto LBL_RES; 3158 } 3159 if ((err = redux (&res, P, mp)) != MP_OKAY) { 3160 goto LBL_RES; 3161 } 3162 continue; 3163 } 3164 3165 /* else we add it to the window */ 3166 bitbuf |= (y << (winsize - ++bitcpy)); 3167 mode = 2; 3168 3169 if (bitcpy == winsize) { 3170 /* ok window is filled so square as required and multiply */ 3171 /* square first */ 3172 for (x = 0; x < winsize; x++) { 3173 if ((err = mp_sqr (&res, &res)) != MP_OKAY) { 3174 goto LBL_RES; 3175 } 3176 if ((err = redux (&res, P, mp)) != MP_OKAY) { 3177 goto LBL_RES; 3178 } 3179 } 3180 3181 /* then multiply */ 3182 if ((err = mp_mul (&res, &M[bitbuf], &res)) != MP_OKAY) { 3183 goto LBL_RES; 3184 } 3185 if ((err = redux (&res, P, mp)) != MP_OKAY) { 3186 goto LBL_RES; 3187 } 3188 3189 /* empty window and reset */ 3190 bitcpy = 0; 3191 bitbuf = 0; 3192 mode = 1; 3193 } 3194 } 3195 3196 /* if bits remain then square/multiply */ 3197 if (mode == 2 && bitcpy > 0) { 3198 /* square then multiply if the bit is set */ 3199 for (x = 0; x < bitcpy; x++) { 3200 if ((err = mp_sqr (&res, &res)) != MP_OKAY) { 3201 goto LBL_RES; 3202 } 3203 if ((err = redux (&res, P, mp)) != MP_OKAY) { 3204 goto LBL_RES; 3205 } 3206 3207 /* get next bit of the window */ 3208 bitbuf <<= 1; 3209 if ((bitbuf & (1 << winsize)) != 0) { 3210 /* then multiply */ 3211 if ((err = mp_mul (&res, &M[1], &res)) != MP_OKAY) { 3212 goto LBL_RES; 3213 } 3214 if ((err = redux (&res, P, mp)) != MP_OKAY) { 3215 goto LBL_RES; 3216 } 3217 } 3218 } 3219 } 3220 3221 if (redmode == 0) { 3222 /* fixup result if Montgomery reduction is used 3223 * recall that any value in a Montgomery system is 3224 * actually multiplied by R mod n. So we have 3225 * to reduce one more time to cancel out the factor 3226 * of R. 3227 */ 3228 if ((err = redux(&res, P, mp)) != MP_OKAY) { 3229 goto LBL_RES; 3230 } 3231 } 3232 3233 /* swap res with Y */ 3234 mp_exch (&res, Y); 3235 err = MP_OKAY; 3236LBL_RES:mp_clear (&res); 3237LBL_M: 3238 mp_clear(&M[1]); 3239 for (x = 1<<(winsize-1); x < (1 << winsize); x++) { 3240 mp_clear (&M[x]); 3241 } 3242 return err; 3243} 3244#endif 3245 3246 3247#ifdef BN_FAST_S_MP_SQR_C 3248/* the jist of squaring... 3249 * you do like mult except the offset of the tmpx [one that 3250 * starts closer to zero] can't equal the offset of tmpy. 3251 * So basically you set up iy like before then you min it with 3252 * (ty-tx) so that it never happens. You double all those 3253 * you add in the inner loop 3254 3255After that loop you do the squares and add them in. 3256*/ 3257 3258static int fast_s_mp_sqr (mp_int * a, mp_int * b) 3259{ 3260 int olduse, res, pa, ix, iz; 3261 mp_digit W[MP_WARRAY], *tmpx; 3262 mp_word W1; 3263 3264 /* grow the destination as required */ 3265 pa = a->used + a->used; 3266 if (b->alloc < pa) { 3267 if ((res = mp_grow (b, pa)) != MP_OKAY) { 3268 return res; 3269 } 3270 } 3271 3272 /* number of output digits to produce */ 3273 W1 = 0; 3274 for (ix = 0; ix < pa; ix++) { 3275 int tx, ty, iy; 3276 mp_word _W; 3277 mp_digit *tmpy; 3278 3279 /* clear counter */ 3280 _W = 0; 3281 3282 /* get offsets into the two bignums */ 3283 ty = MIN(a->used-1, ix); 3284 tx = ix - ty; 3285 3286 /* setup temp aliases */ 3287 tmpx = a->dp + tx; 3288 tmpy = a->dp + ty; 3289 3290 /* this is the number of times the loop will iterrate, essentially 3291 while (tx++ < a->used && ty-- >= 0) { ... } 3292 */ 3293 iy = MIN(a->used-tx, ty+1); 3294 3295 /* now for squaring tx can never equal ty 3296 * we halve the distance since they approach at a rate of 2x 3297 * and we have to round because odd cases need to be executed 3298 */ 3299 iy = MIN(iy, (ty-tx+1)>>1); 3300 3301 /* execute loop */ 3302 for (iz = 0; iz < iy; iz++) { 3303 _W += ((mp_word)*tmpx++)*((mp_word)*tmpy--); 3304 } 3305 3306 /* double the inner product and add carry */ 3307 _W = _W + _W + W1; 3308 3309 /* even columns have the square term in them */ 3310 if ((ix&1) == 0) { 3311 _W += ((mp_word)a->dp[ix>>1])*((mp_word)a->dp[ix>>1]); 3312 } 3313 3314 /* store it */ 3315 W[ix] = (mp_digit)(_W & MP_MASK); 3316 3317 /* make next carry */ 3318 W1 = _W >> ((mp_word)DIGIT_BIT); 3319 } 3320 3321 /* setup dest */ 3322 olduse = b->used; 3323 b->used = a->used+a->used; 3324 3325 { 3326 mp_digit *tmpb; 3327 tmpb = b->dp; 3328 for (ix = 0; ix < pa; ix++) { 3329 *tmpb++ = W[ix] & MP_MASK; 3330 } 3331 3332 /* clear unused digits [that existed in the old copy of c] */ 3333 for (; ix < olduse; ix++) { 3334 *tmpb++ = 0; 3335 } 3336 } 3337 mp_clamp (b); 3338 return MP_OKAY; 3339} 3340#endif 3341 3342 3343#ifdef BN_MP_MUL_D_C 3344/* multiply by a digit */ 3345static int 3346mp_mul_d (mp_int * a, mp_digit b, mp_int * c) 3347{ 3348 mp_digit u, *tmpa, *tmpc; 3349 mp_word r; 3350 int ix, res, olduse; 3351 3352 /* make sure c is big enough to hold a*b */ 3353 if (c->alloc < a->used + 1) { 3354 if ((res = mp_grow (c, a->used + 1)) != MP_OKAY) { 3355 return res; 3356 } 3357 } 3358 3359 /* get the original destinations used count */ 3360 olduse = c->used; 3361 3362 /* set the sign */ 3363 c->sign = a->sign; 3364 3365 /* alias for a->dp [source] */ 3366 tmpa = a->dp; 3367 3368 /* alias for c->dp [dest] */ 3369 tmpc = c->dp; 3370 3371 /* zero carry */ 3372 u = 0; 3373 3374 /* compute columns */ 3375 for (ix = 0; ix < a->used; ix++) { 3376 /* compute product and carry sum for this term */ 3377 r = ((mp_word) u) + ((mp_word)*tmpa++) * ((mp_word)b); 3378 3379 /* mask off higher bits to get a single digit */ 3380 *tmpc++ = (mp_digit) (r & ((mp_word) MP_MASK)); 3381 3382 /* send carry into next iteration */ 3383 u = (mp_digit) (r >> ((mp_word) DIGIT_BIT)); 3384 } 3385 3386 /* store final carry [if any] and increment ix offset */ 3387 *tmpc++ = u; 3388 ++ix; 3389 3390 /* now zero digits above the top */ 3391 while (ix++ < olduse) { 3392 *tmpc++ = 0; 3393 } 3394 3395 /* set used count */ 3396 c->used = a->used + 1; 3397 mp_clamp(c); 3398 3399 return MP_OKAY; 3400} 3401#endif 3402