wearefair-grpc 1.3.1.pre.a

data/third_party/boringssl/crypto/bn/montgomery_inv.c

@@ -0,0 +1,160 @@
+/* Copyright 2016 Brian Smith.
+ *
+ * Permission to use, copy, modify, and/or distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
+ * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
+ * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
+ * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */
+
+#include <openssl/bn.h>
+
+#include <assert.h>
+
+#include "internal.h"
+#include "../internal.h"
+
+
+static uint64_t bn_neg_inv_mod_r_u64(uint64_t n);
+
+OPENSSL_COMPILE_ASSERT(BN_MONT_CTX_N0_LIMBS == 1 || BN_MONT_CTX_N0_LIMBS == 2,
+                       BN_MONT_CTX_N0_LIMBS_VALUE_INVALID);
+OPENSSL_COMPILE_ASSERT(sizeof(uint64_t) ==
+                       BN_MONT_CTX_N0_LIMBS * sizeof(BN_ULONG),
+                       BN_MONT_CTX_N0_LIMBS_DOES_NOT_MATCH_UINT64_T);
+
+/* LG_LITTLE_R is log_2(r). */
+#define LG_LITTLE_R (BN_MONT_CTX_N0_LIMBS * BN_BITS2)
+
+uint64_t bn_mont_n0(const BIGNUM *n) {
+  /* These conditions are checked by the caller, |BN_MONT_CTX_set|. */
+  assert(!BN_is_zero(n));
+  assert(!BN_is_negative(n));
+  assert(BN_is_odd(n));
+
+  /* r == 2**(BN_MONT_CTX_N0_LIMBS * BN_BITS2) and LG_LITTLE_R == lg(r). This
+   * ensures that we can do integer division by |r| by simply ignoring
+   * |BN_MONT_CTX_N0_LIMBS| limbs. Similarly, we can calculate values modulo
+   * |r| by just looking at the lowest |BN_MONT_CTX_N0_LIMBS| limbs. This is
+   * what makes Montgomery multiplication efficient.
+   *
+   * As shown in Algorithm 1 of "Fast Prime Field Elliptic Curve Cryptography
+   * with 256 Bit Primes" by Shay Gueron and Vlad Krasnov, in the loop of a
+   * multi-limb Montgomery multiplication of |a * b (mod n)|, given the
+   * unreduced product |t == a * b|, we repeatedly calculate:
+   *
+   *    t1 := t % r         |t1| is |t|'s lowest limb (see previous paragraph).
+   *    t2 := t1*n0*n
+   *    t3 := t + t2
+   *    t := t3 / r         copy all limbs of |t3| except the lowest to |t|.
+   *
+   * In the last step, it would only make sense to ignore the lowest limb of
+   * |t3| if it were zero. The middle steps ensure that this is the case:
+   *
+   *                            t3 ==  0 (mod r)
+   *                        t + t2 ==  0 (mod r)
+   *                   t + t1*n0*n ==  0 (mod r)
+   *                       t1*n0*n == -t (mod r)
+   *                        t*n0*n == -t (mod r)
+   *                          n0*n == -1 (mod r)
+   *                            n0 == -1/n (mod r)
+   *
+   * Thus, in each iteration of the loop, we multiply by the constant factor
+   * |n0|, the negative inverse of n (mod r). */
+
+  /* n_mod_r = n % r. As explained above, this is done by taking the lowest
+   * |BN_MONT_CTX_N0_LIMBS| limbs of |n|. */
+  uint64_t n_mod_r = n->d[0];
+#if BN_MONT_CTX_N0_LIMBS == 2
+  if (n->top > 1) {
+    n_mod_r |= (uint64_t)n->d[1] << BN_BITS2;
+  }
+#endif
+
+  return bn_neg_inv_mod_r_u64(n_mod_r);
+}
+
+/* bn_neg_inv_r_mod_n_u64 calculates the -1/n mod r; i.e. it calculates |v|
+ * such that u*r - v*n == 1. |r| is the constant defined in |bn_mont_n0|. |n|
+ * must be odd.
+ *
+ * This is derived from |xbinGCD| in Henry S. Warren, Jr.'s "Montgomery
+ * Multiplication" (http://www.hackersdelight.org/MontgomeryMultiplication.pdf).
+ * It is very similar to the MODULAR-INVERSE function in Stephen R. Dussé's and
+ * Burton S. Kaliski Jr.'s "A Cryptographic Library for the Motorola DSP56000"
+ * (http://link.springer.com/chapter/10.1007%2F3-540-46877-3_21).
+ *
+ * This is inspired by Joppe W. Bos's "Constant Time Modular Inversion"
+ * (http://www.joppebos.com/files/CTInversion.pdf) so that the inversion is
+ * constant-time with respect to |n|. We assume uint64_t additions,
+ * subtractions, shifts, and bitwise operations are all constant time, which
+ * may be a large leap of faith on 32-bit targets. We avoid division and
+ * multiplication, which tend to be the most problematic in terms of timing
+ * leaks.
+ *
+ * Most GCD implementations return values such that |u*r + v*n == 1|, so the
+ * caller would have to negate the resultant |v| for the purpose of Montgomery
+ * multiplication. This implementation does the negation implicitly by doing
+ * the computations as a difference instead of a sum. */
+static uint64_t bn_neg_inv_mod_r_u64(uint64_t n) {
+  assert(n % 2 == 1);
+
+  /* alpha == 2**(lg r - 1) == r / 2. */
+  static const uint64_t alpha = UINT64_C(1) << (LG_LITTLE_R - 1);
+
+  const uint64_t beta = n;
+
+  uint64_t u = 1;
+  uint64_t v = 0;
+
+  /* The invariant maintained from here on is:
+   * 2**(lg r - i) == u*2*alpha - v*beta. */
+  for (size_t i = 0; i < LG_LITTLE_R; ++i) {
+#if BN_BITS2 == 64 && defined(BN_ULLONG)
+    assert((BN_ULLONG)(1) << (LG_LITTLE_R - i) ==
+           ((BN_ULLONG)u * 2 * alpha) - ((BN_ULLONG)v * beta));
+#endif
+
+    /* Delete a common factor of 2 in u and v if |u| is even. Otherwise, set
+     * |u = (u + beta) / 2| and |v = (v / 2) + alpha|. */
+
+    uint64_t u_is_odd = UINT64_C(0) - (u & 1); /* Either 0xff..ff or 0. */
+
+    /* The addition can overflow, so use Dietz's method for it.
+     *
+     * Dietz calculates (x+y)/2 by (x⊕y)>>1 + x&y. This is valid for all
+     * (unsigned) x and y, even when x+y overflows. Evidence for 32-bit values
+     * (embedded in 64 bits to so that overflow can be ignored):
+     *
+     * (declare-fun x () (_ BitVec 64))
+     * (declare-fun y () (_ BitVec 64))
+     * (assert (let (
+     *    (one (_ bv1 64))
+     *    (thirtyTwo (_ bv32 64)))
+     *    (and
+     *      (bvult x (bvshl one thirtyTwo))
+     *      (bvult y (bvshl one thirtyTwo))
+     *      (not (=
+     *        (bvadd (bvlshr (bvxor x y) one) (bvand x y))
+     *        (bvlshr (bvadd x y) one)))
+     * )))
+     * (check-sat) */
+    uint64_t beta_if_u_is_odd = beta & u_is_odd; /* Either |beta| or 0. */
+    u = ((u ^ beta_if_u_is_odd) >> 1) + (u & beta_if_u_is_odd);
+
+    uint64_t alpha_if_u_is_odd = alpha & u_is_odd; /* Either |alpha| or 0. */
+    v = (v >> 1) + alpha_if_u_is_odd;
+  }
+
+  /* The invariant now shows that u*r - v*n == 1 since r == 2 * alpha. */
+#if BN_BITS2 == 64 && defined(BN_ULLONG)
+  assert(1 == ((BN_ULLONG)u * 2 * alpha) - ((BN_ULLONG)v * beta));
+#endif
+
+  return v;
+}

data/third_party/boringssl/crypto/bn/mul.c

@@ -0,0 +1,869 @@
+/* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
+ * All rights reserved.
+ *
+ * This package is an SSL implementation written
+ * by Eric Young (eay@cryptsoft.com).
+ * The implementation was written so as to conform with Netscapes SSL.
+ *
+ * This library is free for commercial and non-commercial use as long as
+ * the following conditions are aheared to.  The following conditions
+ * apply to all code found in this distribution, be it the RC4, RSA,
+ * lhash, DES, etc., code; not just the SSL code.  The SSL documentation
+ * included with this distribution is covered by the same copyright terms
+ * except that the holder is Tim Hudson (tjh@cryptsoft.com).
+ *
+ * Copyright remains Eric Young's, and as such any Copyright notices in
+ * the code are not to be removed.
+ * If this package is used in a product, Eric Young should be given attribution
+ * as the author of the parts of the library used.
+ * This can be in the form of a textual message at program startup or
+ * in documentation (online or textual) provided with the package.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. All advertising materials mentioning features or use of this software
+ *    must display the following acknowledgement:
+ *    "This product includes cryptographic software written by
+ *     Eric Young (eay@cryptsoft.com)"
+ *    The word 'cryptographic' can be left out if the rouines from the library
+ *    being used are not cryptographic related :-).
+ * 4. If you include any Windows specific code (or a derivative thereof) from
+ *    the apps directory (application code) you must include an acknowledgement:
+ *    "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
+ *
+ * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ * The licence and distribution terms for any publically available version or
+ * derivative of this code cannot be changed.  i.e. this code cannot simply be
+ * copied and put under another distribution licence
+ * [including the GNU Public Licence.] */
+
+#include <openssl/bn.h>
+
+#include <assert.h>
+#include <string.h>
+
+#include "internal.h"
+
+
+#define BN_MUL_RECURSIVE_SIZE_NORMAL 16
+#define BN_SQR_RECURSIVE_SIZE_NORMAL BN_MUL_RECURSIVE_SIZE_NORMAL
+
+
+static void bn_mul_normal(BN_ULONG *r, BN_ULONG *a, int na, BN_ULONG *b,
+                          int nb) {
+  BN_ULONG *rr;
+
+  if (na < nb) {
+    int itmp;
+    BN_ULONG *ltmp;
+
+    itmp = na;
+    na = nb;
+    nb = itmp;
+    ltmp = a;
+    a = b;
+    b = ltmp;
+  }
+  rr = &(r[na]);
+  if (nb <= 0) {
+    (void)bn_mul_words(r, a, na, 0);
+    return;
+  } else {
+    rr[0] = bn_mul_words(r, a, na, b[0]);
+  }
+
+  for (;;) {
+    if (--nb <= 0) {
+      return;
+    }
+    rr[1] = bn_mul_add_words(&(r[1]), a, na, b[1]);
+    if (--nb <= 0) {
+      return;
+    }
+    rr[2] = bn_mul_add_words(&(r[2]), a, na, b[2]);
+    if (--nb <= 0) {
+      return;
+    }
+    rr[3] = bn_mul_add_words(&(r[3]), a, na, b[3]);
+    if (--nb <= 0) {
+      return;
+    }
+    rr[4] = bn_mul_add_words(&(r[4]), a, na, b[4]);
+    rr += 4;
+    r += 4;
+    b += 4;
+  }
+}
+
+#if !defined(OPENSSL_X86) || defined(OPENSSL_NO_ASM)
+/* Here follows specialised variants of bn_add_words() and bn_sub_words(). They
+ * have the property performing operations on arrays of different sizes. The
+ * sizes of those arrays is expressed through cl, which is the common length (
+ * basicall, min(len(a),len(b)) ), and dl, which is the delta between the two
+ * lengths, calculated as len(a)-len(b). All lengths are the number of
+ * BN_ULONGs...  For the operations that require a result array as parameter,
+ * it must have the length cl+abs(dl). These functions should probably end up
+ * in bn_asm.c as soon as there are assembler counterparts for the systems that
+ * use assembler files.  */
+
+static BN_ULONG bn_sub_part_words(BN_ULONG *r, const BN_ULONG *a,
+                                  const BN_ULONG *b, int cl, int dl) {
+  BN_ULONG c, t;
+
+  assert(cl >= 0);
+  c = bn_sub_words(r, a, b, cl);
+
+  if (dl == 0) {
+    return c;
+  }
+
+  r += cl;
+  a += cl;
+  b += cl;
+
+  if (dl < 0) {
+    for (;;) {
+      t = b[0];
+      r[0] = (0 - t - c) & BN_MASK2;
+      if (t != 0) {
+        c = 1;
+      }
+      if (++dl >= 0) {
+        break;
+      }
+
+      t = b[1];
+      r[1] = (0 - t - c) & BN_MASK2;
+      if (t != 0) {
+        c = 1;
+      }
+      if (++dl >= 0) {
+        break;
+      }
+
+      t = b[2];
+      r[2] = (0 - t - c) & BN_MASK2;
+      if (t != 0) {
+        c = 1;
+      }
+      if (++dl >= 0) {
+        break;
+      }
+
+      t = b[3];
+      r[3] = (0 - t - c) & BN_MASK2;
+      if (t != 0) {
+        c = 1;
+      }
+      if (++dl >= 0) {
+        break;
+      }
+
+      b += 4;
+      r += 4;
+    }
+  } else {
+    int save_dl = dl;
+    while (c) {
+      t = a[0];
+      r[0] = (t - c) & BN_MASK2;
+      if (t != 0) {
+        c = 0;
+      }
+      if (--dl <= 0) {
+        break;
+      }
+
+      t = a[1];
+      r[1] = (t - c) & BN_MASK2;
+      if (t != 0) {
+        c = 0;
+      }
+      if (--dl <= 0) {
+        break;
+      }
+
+      t = a[2];
+      r[2] = (t - c) & BN_MASK2;
+      if (t != 0) {
+        c = 0;
+      }
+      if (--dl <= 0) {
+        break;
+      }
+
+      t = a[3];
+      r[3] = (t - c) & BN_MASK2;
+      if (t != 0) {
+        c = 0;
+      }
+      if (--dl <= 0) {
+        break;
+      }
+
+      save_dl = dl;
+      a += 4;
+      r += 4;
+    }
+    if (dl > 0) {
+      if (save_dl > dl) {
+        switch (save_dl - dl) {
+          case 1:
+            r[1] = a[1];
+            if (--dl <= 0) {
+              break;
+            }
+          case 2:
+            r[2] = a[2];
+            if (--dl <= 0) {
+              break;
+            }
+          case 3:
+            r[3] = a[3];
+            if (--dl <= 0) {
+              break;
+            }
+        }
+        a += 4;
+        r += 4;
+      }
+    }
+
+    if (dl > 0) {
+      for (;;) {
+        r[0] = a[0];
+        if (--dl <= 0) {
+          break;
+        }
+        r[1] = a[1];
+        if (--dl <= 0) {
+          break;
+        }
+        r[2] = a[2];
+        if (--dl <= 0) {
+          break;
+        }
+        r[3] = a[3];
+        if (--dl <= 0) {
+          break;
+        }
+
+        a += 4;
+        r += 4;
+      }
+    }
+  }
+
+  return c;
+}
+#else
+/* On other platforms the function is defined in asm. */
+BN_ULONG bn_sub_part_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b,
+                           int cl, int dl);
+#endif
+
+/* Karatsuba recursive multiplication algorithm
+ * (cf. Knuth, The Art of Computer Programming, Vol. 2) */
+
+/* r is 2*n2 words in size,
+ * a and b are both n2 words in size.
+ * n2 must be a power of 2.
+ * We multiply and return the result.
+ * t must be 2*n2 words in size
+ * We calculate
+ * a[0]*b[0]
+ * a[0]*b[0]+a[1]*b[1]+(a[0]-a[1])*(b[1]-b[0])
+ * a[1]*b[1]
+ */
+/* dnX may not be positive, but n2/2+dnX has to be */
+static void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2,
+                             int dna, int dnb, BN_ULONG *t) {
+  int n = n2 / 2, c1, c2;
+  int tna = n + dna, tnb = n + dnb;
+  unsigned int neg, zero;
+  BN_ULONG ln, lo, *p;
+
+  /* Only call bn_mul_comba 8 if n2 == 8 and the
+   * two arrays are complete [steve]
+   */
+  if (n2 == 8 && dna == 0 && dnb == 0) {
+    bn_mul_comba8(r, a, b);
+    return;
+  }
+
+  /* Else do normal multiply */
+  if (n2 < BN_MUL_RECURSIVE_SIZE_NORMAL) {
+    bn_mul_normal(r, a, n2 + dna, b, n2 + dnb);
+    if ((dna + dnb) < 0) {
+      memset(&r[2 * n2 + dna + dnb], 0, sizeof(BN_ULONG) * -(dna + dnb));
+    }
+    return;
+  }
+
+  /* r=(a[0]-a[1])*(b[1]-b[0]) */
+  c1 = bn_cmp_part_words(a, &(a[n]), tna, n - tna);
+  c2 = bn_cmp_part_words(&(b[n]), b, tnb, tnb - n);
+  zero = neg = 0;
+  switch (c1 * 3 + c2) {
+    case -4:
+      bn_sub_part_words(t, &(a[n]), a, tna, tna - n);       /* - */
+      bn_sub_part_words(&(t[n]), b, &(b[n]), tnb, n - tnb); /* - */
+      break;
+    case -3:
+      zero = 1;
+      break;
+    case -2:
+      bn_sub_part_words(t, &(a[n]), a, tna, tna - n);       /* - */
+      bn_sub_part_words(&(t[n]), &(b[n]), b, tnb, tnb - n); /* + */
+      neg = 1;
+      break;
+    case -1:
+    case 0:
+    case 1:
+      zero = 1;
+      break;
+    case 2:
+      bn_sub_part_words(t, a, &(a[n]), tna, n - tna);       /* + */
+      bn_sub_part_words(&(t[n]), b, &(b[n]), tnb, n - tnb); /* - */
+      neg = 1;
+      break;
+    case 3:
+      zero = 1;
+      break;
+    case 4:
+      bn_sub_part_words(t, a, &(a[n]), tna, n - tna);
+      bn_sub_part_words(&(t[n]), &(b[n]), b, tnb, tnb - n);
+      break;
+  }
+
+  if (n == 4 && dna == 0 && dnb == 0) {
+    /* XXX: bn_mul_comba4 could take extra args to do this well */
+    if (!zero) {
+      bn_mul_comba4(&(t[n2]), t, &(t[n]));
+    } else {
+      memset(&(t[n2]), 0, 8 * sizeof(BN_ULONG));
+    }
+
+    bn_mul_comba4(r, a, b);
+    bn_mul_comba4(&(r[n2]), &(a[n]), &(b[n]));
+  } else if (n == 8 && dna == 0 && dnb == 0) {
+    /* XXX: bn_mul_comba8 could take extra args to do this well */
+    if (!zero) {
+      bn_mul_comba8(&(t[n2]), t, &(t[n]));
+    } else {
+      memset(&(t[n2]), 0, 16 * sizeof(BN_ULONG));
+    }
+
+    bn_mul_comba8(r, a, b);
+    bn_mul_comba8(&(r[n2]), &(a[n]), &(b[n]));
+  } else {
+    p = &(t[n2 * 2]);
+    if (!zero) {
+      bn_mul_recursive(&(t[n2]), t, &(t[n]), n, 0, 0, p);
+    } else {
+      memset(&(t[n2]), 0, n2 * sizeof(BN_ULONG));
+    }
+    bn_mul_recursive(r, a, b, n, 0, 0, p);
+    bn_mul_recursive(&(r[n2]), &(a[n]), &(b[n]), n, dna, dnb, p);
+  }
+
+  /* t[32] holds (a[0]-a[1])*(b[1]-b[0]), c1 is the sign
+   * r[10] holds (a[0]*b[0])
+   * r[32] holds (b[1]*b[1]) */
+
+  c1 = (int)(bn_add_words(t, r, &(r[n2]), n2));
+
+  if (neg) {
+    /* if t[32] is negative */
+    c1 -= (int)(bn_sub_words(&(t[n2]), t, &(t[n2]), n2));
+  } else {
+    /* Might have a carry */
+    c1 += (int)(bn_add_words(&(t[n2]), &(t[n2]), t, n2));
+  }
+
+  /* t[32] holds (a[0]-a[1])*(b[1]-b[0])+(a[0]*b[0])+(a[1]*b[1])
+   * r[10] holds (a[0]*b[0])
+   * r[32] holds (b[1]*b[1])
+   * c1 holds the carry bits */
+  c1 += (int)(bn_add_words(&(r[n]), &(r[n]), &(t[n2]), n2));
+  if (c1) {
+    p = &(r[n + n2]);
+    lo = *p;
+    ln = (lo + c1) & BN_MASK2;
+    *p = ln;
+
+    /* The overflow will stop before we over write
+     * words we should not overwrite */
+    if (ln < (BN_ULONG)c1) {
+      do {
+        p++;
+        lo = *p;
+        ln = (lo + 1) & BN_MASK2;
+        *p = ln;
+      } while (ln == 0);
+    }
+  }
+}
+
+/* n+tn is the word length
+ * t needs to be n*4 is size, as does r */
+/* tnX may not be negative but less than n */
+static void bn_mul_part_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n,
+                                  int tna, int tnb, BN_ULONG *t) {
+  int i, j, n2 = n * 2;
+  int c1, c2, neg;
+  BN_ULONG ln, lo, *p;
+
+  if (n < 8) {
+    bn_mul_normal(r, a, n + tna, b, n + tnb);
+    return;
+  }
+
+  /* r=(a[0]-a[1])*(b[1]-b[0]) */
+  c1 = bn_cmp_part_words(a, &(a[n]), tna, n - tna);
+  c2 = bn_cmp_part_words(&(b[n]), b, tnb, tnb - n);
+  neg = 0;
+  switch (c1 * 3 + c2) {
+    case -4:
+      bn_sub_part_words(t, &(a[n]), a, tna, tna - n);       /* - */
+      bn_sub_part_words(&(t[n]), b, &(b[n]), tnb, n - tnb); /* - */
+      break;
+    case -3:
+    /* break; */
+    case -2:
+      bn_sub_part_words(t, &(a[n]), a, tna, tna - n);       /* - */
+      bn_sub_part_words(&(t[n]), &(b[n]), b, tnb, tnb - n); /* + */
+      neg = 1;
+      break;
+    case -1:
+    case 0:
+    case 1:
+    /* break; */
+    case 2:
+      bn_sub_part_words(t, a, &(a[n]), tna, n - tna);       /* + */
+      bn_sub_part_words(&(t[n]), b, &(b[n]), tnb, n - tnb); /* - */
+      neg = 1;
+      break;
+    case 3:
+    /* break; */
+    case 4:
+      bn_sub_part_words(t, a, &(a[n]), tna, n - tna);
+      bn_sub_part_words(&(t[n]), &(b[n]), b, tnb, tnb - n);
+      break;
+  }
+
+  if (n == 8) {
+    bn_mul_comba8(&(t[n2]), t, &(t[n]));
+    bn_mul_comba8(r, a, b);
+    bn_mul_normal(&(r[n2]), &(a[n]), tna, &(b[n]), tnb);
+    memset(&(r[n2 + tna + tnb]), 0, sizeof(BN_ULONG) * (n2 - tna - tnb));
+  } else {
+    p = &(t[n2 * 2]);
+    bn_mul_recursive(&(t[n2]), t, &(t[n]), n, 0, 0, p);
+    bn_mul_recursive(r, a, b, n, 0, 0, p);
+    i = n / 2;
+    /* If there is only a bottom half to the number,
+     * just do it */
+    if (tna > tnb) {
+      j = tna - i;
+    } else {
+      j = tnb - i;
+    }
+
+    if (j == 0) {
+      bn_mul_recursive(&(r[n2]), &(a[n]), &(b[n]), i, tna - i, tnb - i, p);
+      memset(&(r[n2 + i * 2]), 0, sizeof(BN_ULONG) * (n2 - i * 2));
+    } else if (j > 0) {
+      /* eg, n == 16, i == 8 and tn == 11 */
+      bn_mul_part_recursive(&(r[n2]), &(a[n]), &(b[n]), i, tna - i, tnb - i, p);
+      memset(&(r[n2 + tna + tnb]), 0, sizeof(BN_ULONG) * (n2 - tna - tnb));
+    } else {
+      /* (j < 0) eg, n == 16, i == 8 and tn == 5 */
+      memset(&(r[n2]), 0, sizeof(BN_ULONG) * n2);
+      if (tna < BN_MUL_RECURSIVE_SIZE_NORMAL &&
+          tnb < BN_MUL_RECURSIVE_SIZE_NORMAL) {
+        bn_mul_normal(&(r[n2]), &(a[n]), tna, &(b[n]), tnb);
+      } else {
+        for (;;) {
+          i /= 2;
+          /* these simplified conditions work
+           * exclusively because difference
+           * between tna and tnb is 1 or 0 */
+          if (i < tna || i < tnb) {
+            bn_mul_part_recursive(&(r[n2]), &(a[n]), &(b[n]), i, tna - i,
+                                  tnb - i, p);
+            break;
+          } else if (i == tna || i == tnb) {
+            bn_mul_recursive(&(r[n2]), &(a[n]), &(b[n]), i, tna - i, tnb - i,
+                             p);
+            break;
+          }
+        }
+      }
+    }
+  }
+
+  /* t[32] holds (a[0]-a[1])*(b[1]-b[0]), c1 is the sign
+   * r[10] holds (a[0]*b[0])
+   * r[32] holds (b[1]*b[1])
+   */
+
+  c1 = (int)(bn_add_words(t, r, &(r[n2]), n2));
+
+  if (neg) {
+    /* if t[32] is negative */
+    c1 -= (int)(bn_sub_words(&(t[n2]), t, &(t[n2]), n2));
+  } else {
+    /* Might have a carry */
+    c1 += (int)(bn_add_words(&(t[n2]), &(t[n2]), t, n2));
+  }
+
+  /* t[32] holds (a[0]-a[1])*(b[1]-b[0])+(a[0]*b[0])+(a[1]*b[1])
+   * r[10] holds (a[0]*b[0])
+   * r[32] holds (b[1]*b[1])
+   * c1 holds the carry bits */
+  c1 += (int)(bn_add_words(&(r[n]), &(r[n]), &(t[n2]), n2));
+  if (c1) {
+    p = &(r[n + n2]);
+    lo = *p;
+    ln = (lo + c1) & BN_MASK2;
+    *p = ln;
+
+    /* The overflow will stop before we over write
+     * words we should not overwrite */
+    if (ln < (BN_ULONG)c1) {
+      do {
+        p++;
+        lo = *p;
+        ln = (lo + 1) & BN_MASK2;
+        *p = ln;
+      } while (ln == 0);
+    }
+  }
+}
+
+int BN_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, BN_CTX *ctx) {
+  int ret = 0;
+  int top, al, bl;
+  BIGNUM *rr;
+  int i;
+  BIGNUM *t = NULL;
+  int j = 0, k;
+
+  al = a->top;
+  bl = b->top;
+
+  if ((al == 0) || (bl == 0)) {
+    BN_zero(r);
+    return 1;
+  }
+  top = al + bl;
+
+  BN_CTX_start(ctx);
+  if ((r == a) || (r == b)) {
+    if ((rr = BN_CTX_get(ctx)) == NULL) {
+      goto err;
+    }
+  } else {
+    rr = r;
+  }
+  rr->neg = a->neg ^ b->neg;
+
+  i = al - bl;
+  if (i == 0) {
+    if (al == 8) {
+      if (bn_wexpand(rr, 16) == NULL) {
+        goto err;
+      }
+      rr->top = 16;
+      bn_mul_comba8(rr->d, a->d, b->d);
+      goto end;
+    }
+  }
+
+  static const int kMulNormalSize = 16;
+  if (al >= kMulNormalSize && bl >= kMulNormalSize) {
+    if (i >= -1 && i <= 1) {
+      /* Find out the power of two lower or equal
+         to the longest of the two numbers */
+      if (i >= 0) {
+        j = BN_num_bits_word((BN_ULONG)al);
+      }
+      if (i == -1) {
+        j = BN_num_bits_word((BN_ULONG)bl);
+      }
+      j = 1 << (j - 1);
+      assert(j <= al || j <= bl);
+      k = j + j;
+      t = BN_CTX_get(ctx);
+      if (t == NULL) {
+        goto err;
+      }
+      if (al > j || bl > j) {
+        if (bn_wexpand(t, k * 4) == NULL) {
+          goto err;
+        }
+        if (bn_wexpand(rr, k * 4) == NULL) {
+          goto err;
+        }
+        bn_mul_part_recursive(rr->d, a->d, b->d, j, al - j, bl - j, t->d);
+      } else {
+        /* al <= j || bl <= j */
+        if (bn_wexpand(t, k * 2) == NULL) {
+          goto err;
+        }
+        if (bn_wexpand(rr, k * 2) == NULL) {
+          goto err;
+        }
+        bn_mul_recursive(rr->d, a->d, b->d, j, al - j, bl - j, t->d);
+      }
+      rr->top = top;
+      goto end;
+    }
+  }
+
+  if (bn_wexpand(rr, top) == NULL) {
+    goto err;
+  }
+  rr->top = top;
+  bn_mul_normal(rr->d, a->d, al, b->d, bl);
+
+end:
+  bn_correct_top(rr);
+  if (r != rr && !BN_copy(r, rr)) {
+    goto err;
+  }
+  ret = 1;
+
+err:
+  BN_CTX_end(ctx);
+  return ret;
+}
+
+/* tmp must have 2*n words */
+static void bn_sqr_normal(BN_ULONG *r, const BN_ULONG *a, int n, BN_ULONG *tmp) {
+  int i, j, max;
+  const BN_ULONG *ap;
+  BN_ULONG *rp;
+
+  max = n * 2;
+  ap = a;
+  rp = r;
+  rp[0] = rp[max - 1] = 0;
+  rp++;
+  j = n;
+
+  if (--j > 0) {
+    ap++;
+    rp[j] = bn_mul_words(rp, ap, j, ap[-1]);
+    rp += 2;
+  }
+
+  for (i = n - 2; i > 0; i--) {
+    j--;
+    ap++;
+    rp[j] = bn_mul_add_words(rp, ap, j, ap[-1]);
+    rp += 2;
+  }
+
+  bn_add_words(r, r, r, max);
+
+  /* There will not be a carry */
+
+  bn_sqr_words(tmp, a, n);
+
+  bn_add_words(r, r, tmp, max);
+}
+
+/* r is 2*n words in size,
+ * a and b are both n words in size.    (There's not actually a 'b' here ...)
+ * n must be a power of 2.
+ * We multiply and return the result.
+ * t must be 2*n words in size
+ * We calculate
+ * a[0]*b[0]
+ * a[0]*b[0]+a[1]*b[1]+(a[0]-a[1])*(b[1]-b[0])
+ * a[1]*b[1]
+ */
+static void bn_sqr_recursive(BN_ULONG *r, const BN_ULONG *a, int n2, BN_ULONG *t) {
+  int n = n2 / 2;
+  int zero, c1;
+  BN_ULONG ln, lo, *p;
+
+  if (n2 == 4) {
+    bn_sqr_comba4(r, a);
+    return;
+  } else if (n2 == 8) {
+    bn_sqr_comba8(r, a);
+    return;
+  }
+  if (n2 < BN_SQR_RECURSIVE_SIZE_NORMAL) {
+    bn_sqr_normal(r, a, n2, t);
+    return;
+  }
+  /* r=(a[0]-a[1])*(a[1]-a[0]) */
+  c1 = bn_cmp_words(a, &(a[n]), n);
+  zero = 0;
+  if (c1 > 0) {
+    bn_sub_words(t, a, &(a[n]), n);
+  } else if (c1 < 0) {
+    bn_sub_words(t, &(a[n]), a, n);
+  } else {
+    zero = 1;
+  }
+
+  /* The result will always be negative unless it is zero */
+  p = &(t[n2 * 2]);
+
+  if (!zero) {
+    bn_sqr_recursive(&(t[n2]), t, n, p);
+  } else {
+    memset(&(t[n2]), 0, n2 * sizeof(BN_ULONG));
+  }
+  bn_sqr_recursive(r, a, n, p);
+  bn_sqr_recursive(&(r[n2]), &(a[n]), n, p);
+
+  /* t[32] holds (a[0]-a[1])*(a[1]-a[0]), it is negative or zero
+   * r[10] holds (a[0]*b[0])
+   * r[32] holds (b[1]*b[1]) */
+
+  c1 = (int)(bn_add_words(t, r, &(r[n2]), n2));
+
+  /* t[32] is negative */
+  c1 -= (int)(bn_sub_words(&(t[n2]), t, &(t[n2]), n2));
+
+  /* t[32] holds (a[0]-a[1])*(a[1]-a[0])+(a[0]*a[0])+(a[1]*a[1])
+   * r[10] holds (a[0]*a[0])
+   * r[32] holds (a[1]*a[1])
+   * c1 holds the carry bits */
+  c1 += (int)(bn_add_words(&(r[n]), &(r[n]), &(t[n2]), n2));
+  if (c1) {
+    p = &(r[n + n2]);
+    lo = *p;
+    ln = (lo + c1) & BN_MASK2;
+    *p = ln;
+
+    /* The overflow will stop before we over write
+     * words we should not overwrite */
+    if (ln < (BN_ULONG)c1) {
+      do {
+        p++;
+        lo = *p;
+        ln = (lo + 1) & BN_MASK2;
+        *p = ln;
+      } while (ln == 0);
+    }
+  }
+}
+
+int BN_mul_word(BIGNUM *bn, BN_ULONG w) {
+  BN_ULONG ll;
+
+  w &= BN_MASK2;
+  if (!bn->top) {
+    return 1;
+  }
+
+  if (w == 0) {
+    BN_zero(bn);
+    return 1;
+  }
+
+  ll = bn_mul_words(bn->d, bn->d, bn->top, w);
+  if (ll) {
+    if (bn_wexpand(bn, bn->top + 1) == NULL) {
+      return 0;
+    }
+    bn->d[bn->top++] = ll;
+  }
+
+  return 1;
+}
+
+int BN_sqr(BIGNUM *r, const BIGNUM *a, BN_CTX *ctx) {
+  int max, al;
+  int ret = 0;
+  BIGNUM *tmp, *rr;
+
+  al = a->top;
+  if (al <= 0) {
+    r->top = 0;
+    r->neg = 0;
+    return 1;
+  }
+
+  BN_CTX_start(ctx);
+  rr = (a != r) ? r : BN_CTX_get(ctx);
+  tmp = BN_CTX_get(ctx);
+  if (!rr || !tmp) {
+    goto err;
+  }
+
+  max = 2 * al; /* Non-zero (from above) */
+  if (bn_wexpand(rr, max) == NULL) {
+    goto err;
+  }
+
+  if (al == 4) {
+    bn_sqr_comba4(rr->d, a->d);
+  } else if (al == 8) {
+    bn_sqr_comba8(rr->d, a->d);
+  } else {
+    if (al < BN_SQR_RECURSIVE_SIZE_NORMAL) {
+      BN_ULONG t[BN_SQR_RECURSIVE_SIZE_NORMAL * 2];
+      bn_sqr_normal(rr->d, a->d, al, t);
+    } else {
+      int j, k;
+
+      j = BN_num_bits_word((BN_ULONG)al);
+      j = 1 << (j - 1);
+      k = j + j;
+      if (al == j) {
+        if (bn_wexpand(tmp, k * 2) == NULL) {
+          goto err;
+        }
+        bn_sqr_recursive(rr->d, a->d, al, tmp->d);
+      } else {
+        if (bn_wexpand(tmp, max) == NULL) {
+          goto err;
+        }
+        bn_sqr_normal(rr->d, a->d, al, tmp->d);
+      }
+    }
+  }
+
+  rr->neg = 0;
+  /* If the most-significant half of the top word of 'a' is zero, then
+   * the square of 'a' will max-1 words. */
+  if (a->d[al - 1] == (a->d[al - 1] & BN_MASK2l)) {
+    rr->top = max - 1;
+  } else {
+    rr->top = max;
+  }
+
+  if (rr != r && !BN_copy(r, rr)) {
+    goto err;
+  }
+  ret = 1;
+
+err:
+  BN_CTX_end(ctx);
+  return ret;
+}