grpc 1.15.0 → 1.30.0

data/third_party/boringssl-with-bazel/src/crypto/fipsmodule/bn/montgomery_inv.c

@@ -0,0 +1,186 @@
+/* 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_STATIC_ASSERT(BN_MONT_CTX_N0_LIMBS == 1 || BN_MONT_CTX_N0_LIMBS == 2,
+                      "BN_MONT_CTX_N0_LIMBS value is invalid");
+OPENSSL_STATIC_ASSERT(sizeof(BN_ULONG) * BN_MONT_CTX_N0_LIMBS ==
+                          sizeof(uint64_t),
+                      "uint64_t is insufficient precision for n0");
+
+// 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| or
+  // |BN_MONT_CTX_new_consttime|.
+  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->width > 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;
+}
+
+int bn_mod_exp_base_2_consttime(BIGNUM *r, unsigned p, const BIGNUM *n,
+                                BN_CTX *ctx) {
+  assert(!BN_is_zero(n));
+  assert(!BN_is_negative(n));
+  assert(BN_is_odd(n));
+
+  BN_zero(r);
+
+  unsigned n_bits = BN_num_bits(n);
+  assert(n_bits != 0);
+  assert(p > n_bits);
+  if (n_bits == 1) {
+    return 1;
+  }
+
+  // Set |r| to the larger power of two smaller than |n|, then shift with
+  // reductions the rest of the way.
+  if (!BN_set_bit(r, n_bits - 1) ||
+      !bn_mod_lshift_consttime(r, r, p - (n_bits - 1), n, ctx)) {
+    return 0;
+  }
+
+  return 1;
+}

data/third_party/boringssl-with-bazel/src/crypto/fipsmodule/bn/mul.c

@@ -0,0 +1,749 @@
+/* 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 <stdlib.h>
+#include <string.h>
+
+#include <openssl/err.h>
+#include <openssl/mem.h>
+#include <openssl/type_check.h>
+
+#include "internal.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_abs_sub_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b,
+                             size_t num, BN_ULONG *tmp) {
+  BN_ULONG borrow = bn_sub_words(tmp, a, b, num);
+  bn_sub_words(r, b, a, num);
+  bn_select_words(r, 0 - borrow, r /* tmp < 0 */, tmp /* tmp >= 0 */, num);
+}
+
+static void bn_mul_normal(BN_ULONG *r, const BN_ULONG *a, size_t na,
+                          const BN_ULONG *b, size_t nb) {
+  if (na < nb) {
+    size_t itmp = na;
+    na = nb;
+    nb = itmp;
+    const BN_ULONG *ltmp = a;
+    a = b;
+    b = ltmp;
+  }
+  BN_ULONG *rr = &(r[na]);
+  if (nb == 0) {
+    OPENSSL_memset(r, 0, na * sizeof(BN_ULONG));
+    return;
+  }
+  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;
+  }
+}
+
+// bn_sub_part_words sets |r| to |a| - |b|. It returns the borrow bit, which is
+// one if the operation underflowed and zero otherwise. |cl| is the common
+// length, that is, the shorter of len(a) or len(b). |dl| is the delta length,
+// that is, len(a) - len(b). |r|'s length matches the larger of |a| and |b|, or
+// cl + abs(dl).
+//
+// TODO(davidben): Make this take |size_t|. The |cl| + |dl| calling convention
+// is confusing.
+static BN_ULONG bn_sub_part_words(BN_ULONG *r, const BN_ULONG *a,
+                                  const BN_ULONG *b, int cl, int dl) {
+  assert(cl >= 0);
+  BN_ULONG borrow = bn_sub_words(r, a, b, cl);
+  if (dl == 0) {
+    return borrow;
+  }
+
+  r += cl;
+  a += cl;
+  b += cl;
+
+  if (dl < 0) {
+    // |a| is shorter than |b|. Complete the subtraction as if the excess words
+    // in |a| were zeros.
+    dl = -dl;
+    for (int i = 0; i < dl; i++) {
+      r[i] = 0u - b[i] - borrow;
+      borrow |= r[i] != 0;
+    }
+  } else {
+    // |b| is shorter than |a|. Complete the subtraction as if the excess words
+    // in |b| were zeros.
+    for (int i = 0; i < dl; i++) {
+      // |r| and |a| may alias, so use a temporary.
+      BN_ULONG tmp = a[i];
+      r[i] = a[i] - borrow;
+      borrow = tmp < r[i];
+    }
+  }
+
+  return borrow;
+}
+
+// bn_abs_sub_part_words computes |r| = |a| - |b|, storing the absolute value
+// and returning a mask of all ones if the result was negative and all zeros if
+// the result was positive. |cl| and |dl| follow the |bn_sub_part_words| calling
+// convention.
+//
+// TODO(davidben): Make this take |size_t|. The |cl| + |dl| calling convention
+// is confusing.
+static BN_ULONG bn_abs_sub_part_words(BN_ULONG *r, const BN_ULONG *a,
+                                      const BN_ULONG *b, int cl, int dl,
+                                      BN_ULONG *tmp) {
+  BN_ULONG borrow = bn_sub_part_words(tmp, a, b, cl, dl);
+  bn_sub_part_words(r, b, a, cl, -dl);
+  int r_len = cl + (dl < 0 ? -dl : dl);
+  borrow = 0 - borrow;
+  bn_select_words(r, borrow, r /* tmp < 0 */, tmp /* tmp >= 0 */, r_len);
+  return borrow;
+}
+
+int bn_abs_sub_consttime(BIGNUM *r, const BIGNUM *a, const BIGNUM *b,
+                         BN_CTX *ctx) {
+  int cl = a->width < b->width ? a->width : b->width;
+  int dl = a->width - b->width;
+  int r_len = a->width < b->width ? b->width : a->width;
+  BN_CTX_start(ctx);
+  BIGNUM *tmp = BN_CTX_get(ctx);
+  int ok = tmp != NULL &&
+           bn_wexpand(r, r_len) &&
+           bn_wexpand(tmp, r_len);
+  if (ok) {
+    bn_abs_sub_part_words(r->d, a->d, b->d, cl, dl, tmp->d);
+    r->width = r_len;
+  }
+  BN_CTX_end(ctx);
+  return ok;
+}
+
+// Karatsuba recursive multiplication algorithm
+// (cf. Knuth, The Art of Computer Programming, Vol. 2)
+
+// bn_mul_recursive sets |r| to |a| * |b|, using |t| as scratch space. |r| has
+// length 2*|n2|, |a| has length |n2| + |dna|, |b| has length |n2| + |dnb|, and
+// |t| has length 4*|n2|. |n2| must be a power of two. Finally, we must have
+// -|BN_MUL_RECURSIVE_SIZE_NORMAL|/2 <= |dna| <= 0 and
+// -|BN_MUL_RECURSIVE_SIZE_NORMAL|/2 <= |dnb| <= 0.
+//
+// TODO(davidben): Simplify and |size_t| the calling convention around lengths
+// here.
+static void bn_mul_recursive(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b,
+                             int n2, int dna, int dnb, BN_ULONG *t) {
+  // |n2| is a power of two.
+  assert(n2 != 0 && (n2 & (n2 - 1)) == 0);
+  // Check |dna| and |dnb| are in range.
+  assert(-BN_MUL_RECURSIVE_SIZE_NORMAL/2 <= dna && dna <= 0);
+  assert(-BN_MUL_RECURSIVE_SIZE_NORMAL/2 <= dnb && dnb <= 0);
+
+  // 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) {
+      OPENSSL_memset(&r[2 * n2 + dna + dnb], 0,
+                     sizeof(BN_ULONG) * -(dna + dnb));
+    }
+    return;
+  }
+
+  // Split |a| and |b| into a0,a1 and b0,b1, where a0 and b0 have size |n|.
+  // Split |t| into t0,t1,t2,t3, each of size |n|, with the remaining 4*|n| used
+  // for recursive calls.
+  // Split |r| into r0,r1,r2,r3. We must contribute a0*b0 to r0,r1, a0*a1+b0*b1
+  // to r1,r2, and a1*b1 to r2,r3. The middle term we will compute as:
+  //
+  //   a0*a1 + b0*b1 = (a0 - a1)*(b1 - b0) + a1*b1 + a0*b0
+  //
+  // Note that we know |n| >= |BN_MUL_RECURSIVE_SIZE_NORMAL|/2 above, so
+  // |tna| and |tnb| are non-negative.
+  int n = n2 / 2, tna = n + dna, tnb = n + dnb;
+
+  // t0 = a0 - a1 and t1 = b1 - b0. The result will be multiplied, so we XOR
+  // their sign masks, giving the sign of (a0 - a1)*(b1 - b0). t0 and t1
+  // themselves store the absolute value.
+  BN_ULONG neg = bn_abs_sub_part_words(t, a, &a[n], tna, n - tna, &t[n2]);
+  neg ^= bn_abs_sub_part_words(&t[n], &b[n], b, tnb, tnb - n, &t[n2]);
+
+  // Compute:
+  // t2,t3 = t0 * t1 = |(a0 - a1)*(b1 - b0)|
+  // r0,r1 = a0 * b0
+  // r2,r3 = a1 * b1
+  if (n == 4 && dna == 0 && dnb == 0) {
+    bn_mul_comba4(&t[n2], t, &t[n]);
+
+    bn_mul_comba4(r, a, b);
+    bn_mul_comba4(&r[n2], &a[n], &b[n]);
+  } else if (n == 8 && dna == 0 && dnb == 0) {
+    bn_mul_comba8(&t[n2], t, &t[n]);
+
+    bn_mul_comba8(r, a, b);
+    bn_mul_comba8(&r[n2], &a[n], &b[n]);
+  } else {
+    BN_ULONG *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);
+    bn_mul_recursive(&r[n2], &a[n], &b[n], n, dna, dnb, p);
+  }
+
+  // t0,t1,c = r0,r1 + r2,r3 = a0*b0 + a1*b1
+  BN_ULONG c = bn_add_words(t, r, &r[n2], n2);
+
+  // t2,t3,c = t0,t1,c + neg*t2,t3 = (a0 - a1)*(b1 - b0) + a1*b1 + a0*b0.
+  // The second term is stored as the absolute value, so we do this with a
+  // constant-time select.
+  BN_ULONG c_neg = c - bn_sub_words(&t[n2 * 2], t, &t[n2], n2);
+  BN_ULONG c_pos = c + bn_add_words(&t[n2], t, &t[n2], n2);
+  bn_select_words(&t[n2], neg, &t[n2 * 2], &t[n2], n2);
+  OPENSSL_STATIC_ASSERT(sizeof(BN_ULONG) <= sizeof(crypto_word_t),
+                        "crypto_word_t is too small");
+  c = constant_time_select_w(neg, c_neg, c_pos);
+
+  // We now have our three components. Add them together.
+  // r1,r2,c = r1,r2 + t2,t3,c
+  c += bn_add_words(&r[n], &r[n], &t[n2], n2);
+
+  // Propagate the carry bit to the end.
+  for (int i = n + n2; i < n2 + n2; i++) {
+    BN_ULONG old = r[i];
+    r[i] = old + c;
+    c = r[i] < old;
+  }
+
+  // The product should fit without carries.
+  assert(c == 0);
+}
+
+// bn_mul_part_recursive sets |r| to |a| * |b|, using |t| as scratch space. |r|
+// has length 4*|n|, |a| has length |n| + |tna|, |b| has length |n| + |tnb|, and
+// |t| has length 8*|n|. |n| must be a power of two. Additionally, we must have
+// 0 <= tna < n and 0 <= tnb < n, and |tna| and |tnb| must differ by at most
+// one.
+//
+// TODO(davidben): Make this take |size_t| and perhaps the actual lengths of |a|
+// and |b|.
+static void bn_mul_part_recursive(BN_ULONG *r, const BN_ULONG *a,
+                                  const BN_ULONG *b, int n, int tna, int tnb,
+                                  BN_ULONG *t) {
+  // |n| is a power of two.
+  assert(n != 0 && (n & (n - 1)) == 0);
+  // Check |tna| and |tnb| are in range.
+  assert(0 <= tna && tna < n);
+  assert(0 <= tnb && tnb < n);
+  assert(-1 <= tna - tnb && tna - tnb <= 1);
+
+  int n2 = n * 2;
+  if (n < 8) {
+    bn_mul_normal(r, a, n + tna, b, n + tnb);
+    OPENSSL_memset(r + n2 + tna + tnb, 0, n2 - tna - tnb);
+    return;
+  }
+
+  // Split |a| and |b| into a0,a1 and b0,b1, where a0 and b0 have size |n|. |a1|
+  // and |b1| have size |tna| and |tnb|, respectively.
+  // Split |t| into t0,t1,t2,t3, each of size |n|, with the remaining 4*|n| used
+  // for recursive calls.
+  // Split |r| into r0,r1,r2,r3. We must contribute a0*b0 to r0,r1, a0*a1+b0*b1
+  // to r1,r2, and a1*b1 to r2,r3. The middle term we will compute as:
+  //
+  //   a0*a1 + b0*b1 = (a0 - a1)*(b1 - b0) + a1*b1 + a0*b0
+
+  // t0 = a0 - a1 and t1 = b1 - b0. The result will be multiplied, so we XOR
+  // their sign masks, giving the sign of (a0 - a1)*(b1 - b0). t0 and t1
+  // themselves store the absolute value.
+  BN_ULONG neg = bn_abs_sub_part_words(t, a, &a[n], tna, n - tna, &t[n2]);
+  neg ^= bn_abs_sub_part_words(&t[n], &b[n], b, tnb, tnb - n, &t[n2]);
+
+  // Compute:
+  // t2,t3 = t0 * t1 = |(a0 - a1)*(b1 - b0)|
+  // r0,r1 = a0 * b0
+  // r2,r3 = a1 * b1
+  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);
+    // |bn_mul_normal| only writes |tna| + |tna| words. Zero the rest.
+    OPENSSL_memset(&r[n2 + tna + tnb], 0, sizeof(BN_ULONG) * (n2 - tna - tnb));
+  } else {
+    BN_ULONG *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);
+
+    OPENSSL_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 {
+      int i = n;
+      for (;;) {
+        i /= 2;
+        if (i < tna || i < tnb) {
+          // E.g., n == 16, i == 8 and tna == 11. |tna| and |tnb| are within one
+          // of each other, so if |tna| is larger and tna > i, then we know
+          // tnb >= i, and this call is valid.
+          bn_mul_part_recursive(&r[n2], &a[n], &b[n], i, tna - i, tnb - i, p);
+          break;
+        }
+        if (i == tna || i == tnb) {
+          // If there is only a bottom half to the number, just do it. We know
+          // the larger of |tna - i| and |tnb - i| is zero. The other is zero or
+          // -1 by because of |tna| and |tnb| differ by at most one.
+          bn_mul_recursive(&r[n2], &a[n], &b[n], i, tna - i, tnb - i, p);
+          break;
+        }
+
+        // This loop will eventually terminate when |i| falls below
+        // |BN_MUL_RECURSIVE_SIZE_NORMAL| because we know one of |tna| and |tnb|
+        // exceeds that.
+      }
+    }
+  }
+
+  // t0,t1,c = r0,r1 + r2,r3 = a0*b0 + a1*b1
+  BN_ULONG c = bn_add_words(t, r, &r[n2], n2);
+
+  // t2,t3,c = t0,t1,c + neg*t2,t3 = (a0 - a1)*(b1 - b0) + a1*b1 + a0*b0.
+  // The second term is stored as the absolute value, so we do this with a
+  // constant-time select.
+  BN_ULONG c_neg = c - bn_sub_words(&t[n2 * 2], t, &t[n2], n2);
+  BN_ULONG c_pos = c + bn_add_words(&t[n2], t, &t[n2], n2);
+  bn_select_words(&t[n2], neg, &t[n2 * 2], &t[n2], n2);
+  OPENSSL_STATIC_ASSERT(sizeof(BN_ULONG) <= sizeof(crypto_word_t),
+                        "crypto_word_t is too small");
+  c = constant_time_select_w(neg, c_neg, c_pos);
+
+  // We now have our three components. Add them together.
+  // r1,r2,c = r1,r2 + t2,t3,c
+  c += bn_add_words(&r[n], &r[n], &t[n2], n2);
+
+  // Propagate the carry bit to the end.
+  for (int i = n + n2; i < n2 + n2; i++) {
+    BN_ULONG old = r[i];
+    r[i] = old + c;
+    c = r[i] < old;
+  }
+
+  // The product should fit without carries.
+  assert(c == 0);
+}
+
+// bn_mul_impl implements |BN_mul| and |bn_mul_consttime|. Note this function
+// breaks |BIGNUM| invariants and may return a negative zero. This is handled by
+// the callers.
+static int bn_mul_impl(BIGNUM *r, const BIGNUM *a, const BIGNUM *b,
+                       BN_CTX *ctx) {
+  int al = a->width;
+  int bl = b->width;
+  if (al == 0 || bl == 0) {
+    BN_zero(r);
+    return 1;
+  }
+
+  int ret = 0;
+  BIGNUM *rr;
+  BN_CTX_start(ctx);
+  if (r == a || r == b) {
+    rr = BN_CTX_get(ctx);
+    if (rr == NULL) {
+      goto err;
+    }
+  } else {
+    rr = r;
+  }
+  rr->neg = a->neg ^ b->neg;
+
+  int i = al - bl;
+  if (i == 0) {
+    if (al == 8) {
+      if (!bn_wexpand(rr, 16)) {
+        goto err;
+      }
+      rr->width = 16;
+      bn_mul_comba8(rr->d, a->d, b->d);
+      goto end;
+    }
+  }
+
+  int top = al + bl;
+  static const int kMulNormalSize = 16;
+  if (al >= kMulNormalSize && bl >= kMulNormalSize) {
+    if (-1 <= i && i <= 1) {
+      // Find the largest power of two less than or equal to the larger length.
+      int j;
+      if (i >= 0) {
+        j = BN_num_bits_word((BN_ULONG)al);
+      } else {
+        j = BN_num_bits_word((BN_ULONG)bl);
+      }
+      j = 1 << (j - 1);
+      assert(j <= al || j <= bl);
+      BIGNUM *t = BN_CTX_get(ctx);
+      if (t == NULL) {
+        goto err;
+      }
+      if (al > j || bl > j) {
+        // We know |al| and |bl| are at most one from each other, so if al > j,
+        // bl >= j, and vice versa. Thus we can use |bn_mul_part_recursive|.
+        //
+        // TODO(davidben): This codepath is almost unused in standard
+        // algorithms. Is this optimization necessary? See notes in
+        // https://boringssl-review.googlesource.com/q/I0bd604e2cd6a75c266f64476c23a730ca1721ea6
+        assert(al >= j && bl >= j);
+        if (!bn_wexpand(t, j * 8) ||
+            !bn_wexpand(rr, j * 4)) {
+          goto err;
+        }
+        bn_mul_part_recursive(rr->d, a->d, b->d, j, al - j, bl - j, t->d);
+      } else {
+        // al <= j && bl <= j. Additionally, we know j <= al or j <= bl, so one
+        // of al - j or bl - j is zero. The other, by the bound on |i| above, is
+        // zero or -1. Thus, we can use |bn_mul_recursive|.
+        if (!bn_wexpand(t, j * 4) ||
+            !bn_wexpand(rr, j * 2)) {
+          goto err;
+        }
+        bn_mul_recursive(rr->d, a->d, b->d, j, al - j, bl - j, t->d);
+      }
+      rr->width = top;
+      goto end;
+    }
+  }
+
+  if (!bn_wexpand(rr, top)) {
+    goto err;
+  }
+  rr->width = top;
+  bn_mul_normal(rr->d, a->d, al, b->d, bl);
+
+end:
+  if (r != rr && !BN_copy(r, rr)) {
+    goto err;
+  }
+  ret = 1;
+
+err:
+  BN_CTX_end(ctx);
+  return ret;
+}
+
+int BN_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, BN_CTX *ctx) {
+  if (!bn_mul_impl(r, a, b, ctx)) {
+    return 0;
+  }
+
+  // This additionally fixes any negative zeros created by |bn_mul_impl|.
+  bn_set_minimal_width(r);
+  return 1;
+}
+
+int bn_mul_consttime(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, BN_CTX *ctx) {
+  // Prevent negative zeros.
+  if (a->neg || b->neg) {
+    OPENSSL_PUT_ERROR(BN, BN_R_NEGATIVE_NUMBER);
+    return 0;
+  }
+
+  return bn_mul_impl(r, a, b, ctx);
+}
+
+void bn_mul_small(BN_ULONG *r, size_t num_r, const BN_ULONG *a, size_t num_a,
+                  const BN_ULONG *b, size_t num_b) {
+  if (num_r != num_a + num_b) {
+    abort();
+  }
+  // TODO(davidben): Should this call |bn_mul_comba4| too? |BN_mul| does not
+  // hit that code.
+  if (num_a == 8 && num_b == 8) {
+    bn_mul_comba8(r, a, b);
+  } else {
+    bn_mul_normal(r, a, num_a, b, num_b);
+  }
+}
+
+// tmp must have 2*n words
+static void bn_sqr_normal(BN_ULONG *r, const BN_ULONG *a, size_t n,
+                          BN_ULONG *tmp) {
+  if (n == 0) {
+    return;
+  }
+
+  size_t max = n * 2;
+  const BN_ULONG *ap = a;
+  BN_ULONG *rp = r;
+  rp[0] = rp[max - 1] = 0;
+  rp++;
+
+  // Compute the contribution of a[i] * a[j] for all i < j.
+  if (n > 1) {
+    ap++;
+    rp[n - 1] = bn_mul_words(rp, ap, n - 1, ap[-1]);
+    rp += 2;
+  }
+  if (n > 2) {
+    for (size_t i = n - 2; i > 0; i--) {
+      ap++;
+      rp[i] = bn_mul_add_words(rp, ap, i, ap[-1]);
+      rp += 2;
+    }
+  }
+
+  // The final result fits in |max| words, so none of the following operations
+  // will overflow.
+
+  // Double |r|, giving the contribution of a[i] * a[j] for all i != j.
+  bn_add_words(r, r, r, max);
+
+  // Add in the contribution of a[i] * a[i] for all i.
+  bn_sqr_words(tmp, a, n);
+  bn_add_words(r, r, tmp, max);
+}
+
+// bn_sqr_recursive sets |r| to |a|^2, using |t| as scratch space. |r| has
+// length 2*|n2|, |a| has length |n2|, and |t| has length 4*|n2|. |n2| must be
+// a power of two.
+static void bn_sqr_recursive(BN_ULONG *r, const BN_ULONG *a, size_t n2,
+                             BN_ULONG *t) {
+  // |n2| is a power of two.
+  assert(n2 != 0 && (n2 & (n2 - 1)) == 0);
+
+  if (n2 == 4) {
+    bn_sqr_comba4(r, a);
+    return;
+  }
+  if (n2 == 8) {
+    bn_sqr_comba8(r, a);
+    return;
+  }
+  if (n2 < BN_SQR_RECURSIVE_SIZE_NORMAL) {
+    bn_sqr_normal(r, a, n2, t);
+    return;
+  }
+
+  // Split |a| into a0,a1, each of size |n|.
+  // Split |t| into t0,t1,t2,t3, each of size |n|, with the remaining 4*|n| used
+  // for recursive calls.
+  // Split |r| into r0,r1,r2,r3. We must contribute a0^2 to r0,r1, 2*a0*a1 to
+  // r1,r2, and a1^2 to r2,r3.
+  size_t n = n2 / 2;
+  BN_ULONG *t_recursive = &t[n2 * 2];
+
+  // t0 = |a0 - a1|.
+  bn_abs_sub_words(t, a, &a[n], n, &t[n]);
+  // t2,t3 = t0^2 = |a0 - a1|^2 = a0^2 - 2*a0*a1 + a1^2
+  bn_sqr_recursive(&t[n2], t, n, t_recursive);
+
+  // r0,r1 = a0^2
+  bn_sqr_recursive(r, a, n, t_recursive);
+
+  // r2,r3 = a1^2
+  bn_sqr_recursive(&r[n2], &a[n], n, t_recursive);
+
+  // t0,t1,c = r0,r1 + r2,r3 = a0^2 + a1^2
+  BN_ULONG c = bn_add_words(t, r, &r[n2], n2);
+  // t2,t3,c = t0,t1,c - t2,t3 = 2*a0*a1
+  c -= bn_sub_words(&t[n2], t, &t[n2], n2);
+
+  // We now have our three components. Add them together.
+  // r1,r2,c = r1,r2 + t2,t3,c
+  c += bn_add_words(&r[n], &r[n], &t[n2], n2);
+
+  // Propagate the carry bit to the end.
+  for (size_t i = n + n2; i < n2 + n2; i++) {
+    BN_ULONG old = r[i];
+    r[i] = old + c;
+    c = r[i] < old;
+  }
+
+  // The square should fit without carries.
+  assert(c == 0);
+}
+
+int BN_mul_word(BIGNUM *bn, BN_ULONG w) {
+  if (!bn->width) {
+    return 1;
+  }
+
+  if (w == 0) {
+    BN_zero(bn);
+    return 1;
+  }
+
+  BN_ULONG ll = bn_mul_words(bn->d, bn->d, bn->width, w);
+  if (ll) {
+    if (!bn_wexpand(bn, bn->width + 1)) {
+      return 0;
+    }
+    bn->d[bn->width++] = ll;
+  }
+
+  return 1;
+}
+
+int bn_sqr_consttime(BIGNUM *r, const BIGNUM *a, BN_CTX *ctx) {
+  int al = a->width;
+  if (al <= 0) {
+    r->width = 0;
+    r->neg = 0;
+    return 1;
+  }
+
+  int ret = 0;
+  BN_CTX_start(ctx);
+  BIGNUM *rr = (a != r) ? r : BN_CTX_get(ctx);
+  BIGNUM *tmp = BN_CTX_get(ctx);
+  if (!rr || !tmp) {
+    goto err;
+  }
+
+  int max = 2 * al;  // Non-zero (from above)
+  if (!bn_wexpand(rr, max)) {
+    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 {
+      // If |al| is a power of two, we can use |bn_sqr_recursive|.
+      if (al != 0 && (al & (al - 1)) == 0) {
+        if (!bn_wexpand(tmp, al * 4)) {
+          goto err;
+        }
+        bn_sqr_recursive(rr->d, a->d, al, tmp->d);
+      } else {
+        if (!bn_wexpand(tmp, max)) {
+          goto err;
+        }
+        bn_sqr_normal(rr->d, a->d, al, tmp->d);
+      }
+    }
+  }
+
+  rr->neg = 0;
+  rr->width = max;
+
+  if (rr != r && !BN_copy(r, rr)) {
+    goto err;
+  }
+  ret = 1;
+
+err:
+  BN_CTX_end(ctx);
+  return ret;
+}
+
+int BN_sqr(BIGNUM *r, const BIGNUM *a, BN_CTX *ctx) {
+  if (!bn_sqr_consttime(r, a, ctx)) {
+    return 0;
+  }
+
+  bn_set_minimal_width(r);
+  return 1;
+}
+
+void bn_sqr_small(BN_ULONG *r, size_t num_r, const BN_ULONG *a, size_t num_a) {
+  if (num_r != 2 * num_a || num_a > BN_SMALL_MAX_WORDS) {
+    abort();
+  }
+  if (num_a == 4) {
+    bn_sqr_comba4(r, a);
+  } else if (num_a == 8) {
+    bn_sqr_comba8(r, a);
+  } else {
+    BN_ULONG tmp[2 * BN_SMALL_MAX_WORDS];
+    bn_sqr_normal(r, a, num_a, tmp);
+    OPENSSL_cleanse(tmp, 2 * num_a * sizeof(BN_ULONG));
+  }
+}