Home | History | Annotate | Download | only in bn
      1 /* Copyright (C) 1995-1998 Eric Young (eay (at) cryptsoft.com)
      2  * All rights reserved.
      3  *
      4  * This package is an SSL implementation written
      5  * by Eric Young (eay (at) cryptsoft.com).
      6  * The implementation was written so as to conform with Netscapes SSL.
      7  *
      8  * This library is free for commercial and non-commercial use as long as
      9  * the following conditions are aheared to.  The following conditions
     10  * apply to all code found in this distribution, be it the RC4, RSA,
     11  * lhash, DES, etc., code; not just the SSL code.  The SSL documentation
     12  * included with this distribution is covered by the same copyright terms
     13  * except that the holder is Tim Hudson (tjh (at) cryptsoft.com).
     14  *
     15  * Copyright remains Eric Young's, and as such any Copyright notices in
     16  * the code are not to be removed.
     17  * If this package is used in a product, Eric Young should be given attribution
     18  * as the author of the parts of the library used.
     19  * This can be in the form of a textual message at program startup or
     20  * in documentation (online or textual) provided with the package.
     21  *
     22  * Redistribution and use in source and binary forms, with or without
     23  * modification, are permitted provided that the following conditions
     24  * are met:
     25  * 1. Redistributions of source code must retain the copyright
     26  *    notice, this list of conditions and the following disclaimer.
     27  * 2. Redistributions in binary form must reproduce the above copyright
     28  *    notice, this list of conditions and the following disclaimer in the
     29  *    documentation and/or other materials provided with the distribution.
     30  * 3. All advertising materials mentioning features or use of this software
     31  *    must display the following acknowledgement:
     32  *    "This product includes cryptographic software written by
     33  *     Eric Young (eay (at) cryptsoft.com)"
     34  *    The word 'cryptographic' can be left out if the rouines from the library
     35  *    being used are not cryptographic related :-).
     36  * 4. If you include any Windows specific code (or a derivative thereof) from
     37  *    the apps directory (application code) you must include an acknowledgement:
     38  *    "This product includes software written by Tim Hudson (tjh (at) cryptsoft.com)"
     39  *
     40  * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
     41  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     42  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     43  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
     44  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     45  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     46  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     47  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     48  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     49  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     50  * SUCH DAMAGE.
     51  *
     52  * The licence and distribution terms for any publically available version or
     53  * derivative of this code cannot be changed.  i.e. this code cannot simply be
     54  * copied and put under another distribution licence
     55  * [including the GNU Public Licence.] */
     56 
     57 #include <openssl/bn.h>
     58 
     59 #include <assert.h>
     60 #include <limits.h>
     61 
     62 #include <openssl/err.h>
     63 
     64 #include "internal.h"
     65 
     66 
     67 #if !defined(BN_ULLONG)
     68 // bn_div_words divides a double-width |h|,|l| by |d| and returns the result,
     69 // which must fit in a |BN_ULONG|.
     70 static BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d) {
     71   BN_ULONG dh, dl, q, ret = 0, th, tl, t;
     72   int i, count = 2;
     73 
     74   if (d == 0) {
     75     return BN_MASK2;
     76   }
     77 
     78   i = BN_num_bits_word(d);
     79   assert((i == BN_BITS2) || (h <= (BN_ULONG)1 << i));
     80 
     81   i = BN_BITS2 - i;
     82   if (h >= d) {
     83     h -= d;
     84   }
     85 
     86   if (i) {
     87     d <<= i;
     88     h = (h << i) | (l >> (BN_BITS2 - i));
     89     l <<= i;
     90   }
     91   dh = (d & BN_MASK2h) >> BN_BITS4;
     92   dl = (d & BN_MASK2l);
     93   for (;;) {
     94     if ((h >> BN_BITS4) == dh) {
     95       q = BN_MASK2l;
     96     } else {
     97       q = h / dh;
     98     }
     99 
    100     th = q * dh;
    101     tl = dl * q;
    102     for (;;) {
    103       t = h - th;
    104       if ((t & BN_MASK2h) ||
    105           ((tl) <= ((t << BN_BITS4) | ((l & BN_MASK2h) >> BN_BITS4)))) {
    106         break;
    107       }
    108       q--;
    109       th -= dh;
    110       tl -= dl;
    111     }
    112     t = (tl >> BN_BITS4);
    113     tl = (tl << BN_BITS4) & BN_MASK2h;
    114     th += t;
    115 
    116     if (l < tl) {
    117       th++;
    118     }
    119     l -= tl;
    120     if (h < th) {
    121       h += d;
    122       q--;
    123     }
    124     h -= th;
    125 
    126     if (--count == 0) {
    127       break;
    128     }
    129 
    130     ret = q << BN_BITS4;
    131     h = (h << BN_BITS4) | (l >> BN_BITS4);
    132     l = (l & BN_MASK2l) << BN_BITS4;
    133   }
    134 
    135   ret |= q;
    136   return ret;
    137 }
    138 #endif  // !defined(BN_ULLONG)
    139 
    140 static inline void bn_div_rem_words(BN_ULONG *quotient_out, BN_ULONG *rem_out,
    141                                     BN_ULONG n0, BN_ULONG n1, BN_ULONG d0) {
    142   // GCC and Clang generate function calls to |__udivdi3| and |__umoddi3| when
    143   // the |BN_ULLONG|-based C code is used.
    144   //
    145   // GCC bugs:
    146   //   * https://gcc.gnu.org/bugzilla/show_bug.cgi?id=14224
    147   //   * https://gcc.gnu.org/bugzilla/show_bug.cgi?id=43721
    148   //   * https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54183
    149   //   * https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58897
    150   //   * https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65668
    151   //
    152   // Clang bugs:
    153   //   * https://llvm.org/bugs/show_bug.cgi?id=6397
    154   //   * https://llvm.org/bugs/show_bug.cgi?id=12418
    155   //
    156   // These issues aren't specific to x86 and x86_64, so it might be worthwhile
    157   // to add more assembly language implementations.
    158 #if !defined(OPENSSL_NO_ASM) && defined(OPENSSL_X86) && \
    159     (defined(__GNUC__) || defined(__clang__))
    160   __asm__ volatile("divl %4"
    161                    : "=a"(*quotient_out), "=d"(*rem_out)
    162                    : "a"(n1), "d"(n0), "rm"(d0)
    163                    : "cc");
    164 #elif !defined(OPENSSL_NO_ASM) && defined(OPENSSL_X86_64) && \
    165     (defined(__GNUC__) || defined(__clang__))
    166   __asm__ volatile("divq %4"
    167                    : "=a"(*quotient_out), "=d"(*rem_out)
    168                    : "a"(n1), "d"(n0), "rm"(d0)
    169                    : "cc");
    170 #else
    171 #if defined(BN_ULLONG)
    172   BN_ULLONG n = (((BN_ULLONG)n0) << BN_BITS2) | n1;
    173   *quotient_out = (BN_ULONG)(n / d0);
    174 #else
    175   *quotient_out = bn_div_words(n0, n1, d0);
    176 #endif
    177   *rem_out = n1 - (*quotient_out * d0);
    178 #endif
    179 }
    180 
    181 // BN_div computes "quotient := numerator / divisor", rounding towards zero,
    182 // and sets up |rem| such that "quotient * divisor + rem = numerator" holds.
    183 //
    184 // Thus:
    185 //
    186 //     quotient->neg == numerator->neg ^ divisor->neg
    187 //        (unless the result is zero)
    188 //     rem->neg == numerator->neg
    189 //        (unless the remainder is zero)
    190 //
    191 // If |quotient| or |rem| is NULL, the respective value is not returned.
    192 //
    193 // This was specifically designed to contain fewer branches that may leak
    194 // sensitive information; see "New Branch Prediction Vulnerabilities in OpenSSL
    195 // and Necessary Software Countermeasures" by Onur Acmez, Shay Gueron, and
    196 // Jean-Pierre Seifert.
    197 int BN_div(BIGNUM *quotient, BIGNUM *rem, const BIGNUM *numerator,
    198            const BIGNUM *divisor, BN_CTX *ctx) {
    199   int norm_shift, loop;
    200   BIGNUM wnum;
    201   BN_ULONG *resp, *wnump;
    202   BN_ULONG d0, d1;
    203   int num_n, div_n;
    204 
    205   // Invalid zero-padding would have particularly bad consequences
    206   // so don't just rely on bn_check_top() here
    207   if ((numerator->top > 0 && numerator->d[numerator->top - 1] == 0) ||
    208       (divisor->top > 0 && divisor->d[divisor->top - 1] == 0)) {
    209     OPENSSL_PUT_ERROR(BN, BN_R_NOT_INITIALIZED);
    210     return 0;
    211   }
    212 
    213   if (BN_is_zero(divisor)) {
    214     OPENSSL_PUT_ERROR(BN, BN_R_DIV_BY_ZERO);
    215     return 0;
    216   }
    217 
    218   BN_CTX_start(ctx);
    219   BIGNUM *tmp = BN_CTX_get(ctx);
    220   BIGNUM *snum = BN_CTX_get(ctx);
    221   BIGNUM *sdiv = BN_CTX_get(ctx);
    222   BIGNUM *res = NULL;
    223   if (quotient == NULL) {
    224     res = BN_CTX_get(ctx);
    225   } else {
    226     res = quotient;
    227   }
    228   if (sdiv == NULL || res == NULL) {
    229     goto err;
    230   }
    231 
    232   // First we normalise the numbers
    233   norm_shift = BN_BITS2 - (BN_num_bits(divisor) % BN_BITS2);
    234   if (!BN_lshift(sdiv, divisor, norm_shift)) {
    235     goto err;
    236   }
    237   sdiv->neg = 0;
    238   norm_shift += BN_BITS2;
    239   if (!BN_lshift(snum, numerator, norm_shift)) {
    240     goto err;
    241   }
    242   snum->neg = 0;
    243 
    244   // Since we don't want to have special-case logic for the case where snum is
    245   // larger than sdiv, we pad snum with enough zeroes without changing its
    246   // value.
    247   if (snum->top <= sdiv->top + 1) {
    248     if (!bn_wexpand(snum, sdiv->top + 2)) {
    249       goto err;
    250     }
    251     for (int i = snum->top; i < sdiv->top + 2; i++) {
    252       snum->d[i] = 0;
    253     }
    254     snum->top = sdiv->top + 2;
    255   } else {
    256     if (!bn_wexpand(snum, snum->top + 1)) {
    257       goto err;
    258     }
    259     snum->d[snum->top] = 0;
    260     snum->top++;
    261   }
    262 
    263   div_n = sdiv->top;
    264   num_n = snum->top;
    265   loop = num_n - div_n;
    266   // Lets setup a 'window' into snum
    267   // This is the part that corresponds to the current
    268   // 'area' being divided
    269   wnum.neg = 0;
    270   wnum.d = &(snum->d[loop]);
    271   wnum.top = div_n;
    272   // only needed when BN_ucmp messes up the values between top and max
    273   wnum.dmax = snum->dmax - loop;  // so we don't step out of bounds
    274 
    275   // Get the top 2 words of sdiv
    276   // div_n=sdiv->top;
    277   d0 = sdiv->d[div_n - 1];
    278   d1 = (div_n == 1) ? 0 : sdiv->d[div_n - 2];
    279 
    280   // pointer to the 'top' of snum
    281   wnump = &(snum->d[num_n - 1]);
    282 
    283   // Setup to 'res'
    284   res->neg = (numerator->neg ^ divisor->neg);
    285   if (!bn_wexpand(res, loop + 1)) {
    286     goto err;
    287   }
    288   res->top = loop - 1;
    289   resp = &(res->d[loop - 1]);
    290 
    291   // space for temp
    292   if (!bn_wexpand(tmp, div_n + 1)) {
    293     goto err;
    294   }
    295 
    296   // if res->top == 0 then clear the neg value otherwise decrease
    297   // the resp pointer
    298   if (res->top == 0) {
    299     res->neg = 0;
    300   } else {
    301     resp--;
    302   }
    303 
    304   for (int i = 0; i < loop - 1; i++, wnump--, resp--) {
    305     BN_ULONG q, l0;
    306     // the first part of the loop uses the top two words of snum and sdiv to
    307     // calculate a BN_ULONG q such that | wnum - sdiv * q | < sdiv
    308     BN_ULONG n0, n1, rm = 0;
    309 
    310     n0 = wnump[0];
    311     n1 = wnump[-1];
    312     if (n0 == d0) {
    313       q = BN_MASK2;
    314     } else {
    315       // n0 < d0
    316       bn_div_rem_words(&q, &rm, n0, n1, d0);
    317 
    318 #ifdef BN_ULLONG
    319       BN_ULLONG t2 = (BN_ULLONG)d1 * q;
    320       for (;;) {
    321         if (t2 <= ((((BN_ULLONG)rm) << BN_BITS2) | wnump[-2])) {
    322           break;
    323         }
    324         q--;
    325         rm += d0;
    326         if (rm < d0) {
    327           break;  // don't let rm overflow
    328         }
    329         t2 -= d1;
    330       }
    331 #else  // !BN_ULLONG
    332       BN_ULONG t2l, t2h;
    333       BN_UMULT_LOHI(t2l, t2h, d1, q);
    334       for (;;) {
    335         if (t2h < rm ||
    336             (t2h == rm && t2l <= wnump[-2])) {
    337           break;
    338         }
    339         q--;
    340         rm += d0;
    341         if (rm < d0) {
    342           break;  // don't let rm overflow
    343         }
    344         if (t2l < d1) {
    345           t2h--;
    346         }
    347         t2l -= d1;
    348       }
    349 #endif  // !BN_ULLONG
    350     }
    351 
    352     l0 = bn_mul_words(tmp->d, sdiv->d, div_n, q);
    353     tmp->d[div_n] = l0;
    354     wnum.d--;
    355     // ingore top values of the bignums just sub the two
    356     // BN_ULONG arrays with bn_sub_words
    357     if (bn_sub_words(wnum.d, wnum.d, tmp->d, div_n + 1)) {
    358       // Note: As we have considered only the leading
    359       // two BN_ULONGs in the calculation of q, sdiv * q
    360       // might be greater than wnum (but then (q-1) * sdiv
    361       // is less or equal than wnum)
    362       q--;
    363       if (bn_add_words(wnum.d, wnum.d, sdiv->d, div_n)) {
    364         // we can't have an overflow here (assuming
    365         // that q != 0, but if q == 0 then tmp is
    366         // zero anyway)
    367         (*wnump)++;
    368       }
    369     }
    370     // store part of the result
    371     *resp = q;
    372   }
    373 
    374   bn_correct_top(snum);
    375 
    376   if (rem != NULL) {
    377     // Keep a copy of the neg flag in numerator because if |rem| == |numerator|
    378     // |BN_rshift| will overwrite it.
    379     int neg = numerator->neg;
    380     if (!BN_rshift(rem, snum, norm_shift)) {
    381       goto err;
    382     }
    383     if (!BN_is_zero(rem)) {
    384       rem->neg = neg;
    385     }
    386   }
    387 
    388   bn_correct_top(res);
    389   BN_CTX_end(ctx);
    390   return 1;
    391 
    392 err:
    393   BN_CTX_end(ctx);
    394   return 0;
    395 }
    396 
    397 int BN_nnmod(BIGNUM *r, const BIGNUM *m, const BIGNUM *d, BN_CTX *ctx) {
    398   if (!(BN_mod(r, m, d, ctx))) {
    399     return 0;
    400   }
    401   if (!r->neg) {
    402     return 1;
    403   }
    404 
    405   // now -|d| < r < 0, so we have to set r := r + |d|.
    406   return (d->neg ? BN_sub : BN_add)(r, r, d);
    407 }
    408 
    409 int BN_mod_add(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, const BIGNUM *m,
    410                BN_CTX *ctx) {
    411   if (!BN_add(r, a, b)) {
    412     return 0;
    413   }
    414   return BN_nnmod(r, r, m, ctx);
    415 }
    416 
    417 int BN_mod_add_quick(BIGNUM *r, const BIGNUM *a, const BIGNUM *b,
    418                      const BIGNUM *m) {
    419   if (!BN_uadd(r, a, b)) {
    420     return 0;
    421   }
    422   if (BN_ucmp(r, m) >= 0) {
    423     return BN_usub(r, r, m);
    424   }
    425   return 1;
    426 }
    427 
    428 int BN_mod_sub(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, const BIGNUM *m,
    429                BN_CTX *ctx) {
    430   if (!BN_sub(r, a, b)) {
    431     return 0;
    432   }
    433   return BN_nnmod(r, r, m, ctx);
    434 }
    435 
    436 // BN_mod_sub variant that may be used if both  a  and  b  are non-negative
    437 // and less than  m
    438 int BN_mod_sub_quick(BIGNUM *r, const BIGNUM *a, const BIGNUM *b,
    439                      const BIGNUM *m) {
    440   if (!BN_sub(r, a, b)) {
    441     return 0;
    442   }
    443   if (r->neg) {
    444     return BN_add(r, r, m);
    445   }
    446   return 1;
    447 }
    448 
    449 int BN_mod_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, const BIGNUM *m,
    450                BN_CTX *ctx) {
    451   BIGNUM *t;
    452   int ret = 0;
    453 
    454   BN_CTX_start(ctx);
    455   t = BN_CTX_get(ctx);
    456   if (t == NULL) {
    457     goto err;
    458   }
    459 
    460   if (a == b) {
    461     if (!BN_sqr(t, a, ctx)) {
    462       goto err;
    463     }
    464   } else {
    465     if (!BN_mul(t, a, b, ctx)) {
    466       goto err;
    467     }
    468   }
    469 
    470   if (!BN_nnmod(r, t, m, ctx)) {
    471     goto err;
    472   }
    473 
    474   ret = 1;
    475 
    476 err:
    477   BN_CTX_end(ctx);
    478   return ret;
    479 }
    480 
    481 int BN_mod_sqr(BIGNUM *r, const BIGNUM *a, const BIGNUM *m, BN_CTX *ctx) {
    482   if (!BN_sqr(r, a, ctx)) {
    483     return 0;
    484   }
    485 
    486   // r->neg == 0,  thus we don't need BN_nnmod
    487   return BN_mod(r, r, m, ctx);
    488 }
    489 
    490 int BN_mod_lshift(BIGNUM *r, const BIGNUM *a, int n, const BIGNUM *m,
    491                   BN_CTX *ctx) {
    492   BIGNUM *abs_m = NULL;
    493   int ret;
    494 
    495   if (!BN_nnmod(r, a, m, ctx)) {
    496     return 0;
    497   }
    498 
    499   if (m->neg) {
    500     abs_m = BN_dup(m);
    501     if (abs_m == NULL) {
    502       return 0;
    503     }
    504     abs_m->neg = 0;
    505   }
    506 
    507   ret = BN_mod_lshift_quick(r, r, n, (abs_m ? abs_m : m));
    508 
    509   BN_free(abs_m);
    510   return ret;
    511 }
    512 
    513 int BN_mod_lshift_quick(BIGNUM *r, const BIGNUM *a, int n, const BIGNUM *m) {
    514   if (r != a) {
    515     if (BN_copy(r, a) == NULL) {
    516       return 0;
    517     }
    518   }
    519 
    520   while (n > 0) {
    521     int max_shift;
    522 
    523     // 0 < r < m
    524     max_shift = BN_num_bits(m) - BN_num_bits(r);
    525     // max_shift >= 0
    526 
    527     if (max_shift < 0) {
    528       OPENSSL_PUT_ERROR(BN, BN_R_INPUT_NOT_REDUCED);
    529       return 0;
    530     }
    531 
    532     if (max_shift > n) {
    533       max_shift = n;
    534     }
    535 
    536     if (max_shift) {
    537       if (!BN_lshift(r, r, max_shift)) {
    538         return 0;
    539       }
    540       n -= max_shift;
    541     } else {
    542       if (!BN_lshift1(r, r)) {
    543         return 0;
    544       }
    545       --n;
    546     }
    547 
    548     // BN_num_bits(r) <= BN_num_bits(m)
    549     if (BN_cmp(r, m) >= 0) {
    550       if (!BN_sub(r, r, m)) {
    551         return 0;
    552       }
    553     }
    554   }
    555 
    556   return 1;
    557 }
    558 
    559 int BN_mod_lshift1(BIGNUM *r, const BIGNUM *a, const BIGNUM *m, BN_CTX *ctx) {
    560   if (!BN_lshift1(r, a)) {
    561     return 0;
    562   }
    563 
    564   return BN_nnmod(r, r, m, ctx);
    565 }
    566 
    567 int BN_mod_lshift1_quick(BIGNUM *r, const BIGNUM *a, const BIGNUM *m) {
    568   if (!BN_lshift1(r, a)) {
    569     return 0;
    570   }
    571   if (BN_cmp(r, m) >= 0) {
    572     return BN_sub(r, r, m);
    573   }
    574 
    575   return 1;
    576 }
    577 
    578 BN_ULONG BN_div_word(BIGNUM *a, BN_ULONG w) {
    579   BN_ULONG ret = 0;
    580   int i, j;
    581 
    582   if (!w) {
    583     // actually this an error (division by zero)
    584     return (BN_ULONG) - 1;
    585   }
    586 
    587   if (a->top == 0) {
    588     return 0;
    589   }
    590 
    591   // normalize input for |bn_div_rem_words|.
    592   j = BN_BITS2 - BN_num_bits_word(w);
    593   w <<= j;
    594   if (!BN_lshift(a, a, j)) {
    595     return (BN_ULONG) - 1;
    596   }
    597 
    598   for (i = a->top - 1; i >= 0; i--) {
    599     BN_ULONG l = a->d[i];
    600     BN_ULONG d;
    601     BN_ULONG unused_rem;
    602     bn_div_rem_words(&d, &unused_rem, ret, l, w);
    603     ret = l - (d * w);
    604     a->d[i] = d;
    605   }
    606 
    607   if ((a->top > 0) && (a->d[a->top - 1] == 0)) {
    608     a->top--;
    609   }
    610 
    611   if (a->top == 0) {
    612     a->neg = 0;
    613   }
    614 
    615   ret >>= j;
    616   return ret;
    617 }
    618 
    619 BN_ULONG BN_mod_word(const BIGNUM *a, BN_ULONG w) {
    620 #ifndef BN_CAN_DIVIDE_ULLONG
    621   BN_ULONG ret = 0;
    622 #else
    623   BN_ULLONG ret = 0;
    624 #endif
    625   int i;
    626 
    627   if (w == 0) {
    628     return (BN_ULONG) -1;
    629   }
    630 
    631 #ifndef BN_CAN_DIVIDE_ULLONG
    632   // If |w| is too long and we don't have |BN_ULLONG| division then we need to
    633   // fall back to using |BN_div_word|.
    634   if (w > ((BN_ULONG)1 << BN_BITS4)) {
    635     BIGNUM *tmp = BN_dup(a);
    636     if (tmp == NULL) {
    637       return (BN_ULONG)-1;
    638     }
    639     ret = BN_div_word(tmp, w);
    640     BN_free(tmp);
    641     return ret;
    642   }
    643 #endif
    644 
    645   for (i = a->top - 1; i >= 0; i--) {
    646 #ifndef BN_CAN_DIVIDE_ULLONG
    647     ret = ((ret << BN_BITS4) | ((a->d[i] >> BN_BITS4) & BN_MASK2l)) % w;
    648     ret = ((ret << BN_BITS4) | (a->d[i] & BN_MASK2l)) % w;
    649 #else
    650     ret = (BN_ULLONG)(((ret << (BN_ULLONG)BN_BITS2) | a->d[i]) % (BN_ULLONG)w);
    651 #endif
    652   }
    653   return (BN_ULONG)ret;
    654 }
    655 
    656 int BN_mod_pow2(BIGNUM *r, const BIGNUM *a, size_t e) {
    657   if (e == 0 || a->top == 0) {
    658     BN_zero(r);
    659     return 1;
    660   }
    661 
    662   size_t num_words = 1 + ((e - 1) / BN_BITS2);
    663 
    664   // If |a| definitely has less than |e| bits, just BN_copy.
    665   if ((size_t) a->top < num_words) {
    666     return BN_copy(r, a) != NULL;
    667   }
    668 
    669   // Otherwise, first make sure we have enough space in |r|.
    670   // Note that this will fail if num_words > INT_MAX.
    671   if (!bn_wexpand(r, num_words)) {
    672     return 0;
    673   }
    674 
    675   // Copy the content of |a| into |r|.
    676   OPENSSL_memcpy(r->d, a->d, num_words * sizeof(BN_ULONG));
    677 
    678   // If |e| isn't word-aligned, we have to mask off some of our bits.
    679   size_t top_word_exponent = e % (sizeof(BN_ULONG) * 8);
    680   if (top_word_exponent != 0) {
    681     r->d[num_words - 1] &= (((BN_ULONG) 1) << top_word_exponent) - 1;
    682   }
    683 
    684   // Fill in the remaining fields of |r|.
    685   r->neg = a->neg;
    686   r->top = (int) num_words;
    687   bn_correct_top(r);
    688   return 1;
    689 }
    690 
    691 int BN_nnmod_pow2(BIGNUM *r, const BIGNUM *a, size_t e) {
    692   if (!BN_mod_pow2(r, a, e)) {
    693     return 0;
    694   }
    695 
    696   // If the returned value was non-negative, we're done.
    697   if (BN_is_zero(r) || !r->neg) {
    698     return 1;
    699   }
    700 
    701   size_t num_words = 1 + (e - 1) / BN_BITS2;
    702 
    703   // Expand |r| to the size of our modulus.
    704   if (!bn_wexpand(r, num_words)) {
    705     return 0;
    706   }
    707 
    708   // Clear the upper words of |r|.
    709   OPENSSL_memset(&r->d[r->top], 0, (num_words - r->top) * BN_BYTES);
    710 
    711   // Set parameters of |r|.
    712   r->neg = 0;
    713   r->top = (int) num_words;
    714 
    715   // Now, invert every word. The idea here is that we want to compute 2^e-|x|,
    716   // which is actually equivalent to the twos-complement representation of |x|
    717   // in |e| bits, which is -x = ~x + 1.
    718   for (int i = 0; i < r->top; i++) {
    719     r->d[i] = ~r->d[i];
    720   }
    721 
    722   // If our exponent doesn't span the top word, we have to mask the rest.
    723   size_t top_word_exponent = e % BN_BITS2;
    724   if (top_word_exponent != 0) {
    725     r->d[r->top - 1] &= (((BN_ULONG) 1) << top_word_exponent) - 1;
    726   }
    727 
    728   // Keep the correct_top invariant for BN_add.
    729   bn_correct_top(r);
    730 
    731   // Finally, add one, for the reason described above.
    732   return BN_add(r, r, BN_value_one());
    733 }
    734