diff --git a/crypto/ec/ec.h b/crypto/ec/ec.h
index 3601de2c0f86031e4ee70d7a7c4640935ba9fcea..99e161e753e03b218f478bde732f926b9e2d9515 100644
--- a/crypto/ec/ec.h
+++ b/crypto/ec/ec.h
@@ -245,6 +245,12 @@ int EC_GROUP_set_generator(EC_GROUP *group, const EC_POINT *generator, const BIG
  */
 const EC_POINT *EC_GROUP_get0_generator(const EC_GROUP *group);
 
+/** Returns the montgomery data for order(Generator)
+ *  \param  group  EC_GROUP object
+ *  \return the currently used generator (possibly NULL).
+*/
+BN_MONT_CTX *EC_GROUP_get_mont_data(const EC_GROUP *group);
+
 /** Gets the order of a EC_GROUP
  *  \param  group  EC_GROUP object
  *  \param  order  BIGNUM to which the order is copied
diff --git a/crypto/ec/ec_lcl.h b/crypto/ec/ec_lcl.h
index b0d48b6b5cc7f8a8af66523b24feda6cfaf6cd60..22b53d28a9f33a0b458ba26027501481d77f9b88 100644
--- a/crypto/ec/ec_lcl.h
+++ b/crypto/ec/ec_lcl.h
@@ -235,6 +235,8 @@ struct ec_group_st {
 	void *field_data1; /* method-specific (e.g., Montgomery structure) */
 	void *field_data2; /* method-specific */
 	int (*field_mod_func)(BIGNUM *, const BIGNUM *, const BIGNUM *,	BN_CTX *); /* method-specific */
+
+	BN_MONT_CTX *mont_data;    /* data for ECDSA inverse */
 } /* EC_GROUP */;
 
 struct ec_key_st {
@@ -444,3 +446,12 @@ void ec_GFp_nistp_points_make_affine_internal(size_t num, void *point_array,
 	void (*felem_contract)(void *out, const void *in));
 void ec_GFp_nistp_recode_scalar_bits(unsigned char *sign, unsigned char *digit, unsigned char in);
 #endif
+int ec_precompute_mont_data(EC_GROUP *);
+
+#ifdef ECP_NISTZ256_ASM
+/** Returns GFp methods using montgomery multiplication, with x86-64 optimized
+ * P256. See http://eprint.iacr.org/2013/816.
+ *  \return  EC_METHOD object
+ */
+const EC_METHOD *EC_GFp_nistz256_method(void);
+#endif
diff --git a/crypto/ec/ec_lib.c b/crypto/ec/ec_lib.c
index a295a84f2b0167ec1928b3a7293e951408a08ec8..0a182d4229e76cc11561930738162dba0b312e5b 100644
--- a/crypto/ec/ec_lib.c
+++ b/crypto/ec/ec_lib.c
@@ -101,6 +101,7 @@ EC_GROUP *EC_GROUP_new(const EC_METHOD *meth)
 	ret->meth = meth;
 
 	ret->extra_data = NULL;
+	ret->mont_data = NULL;
 
 	ret->generator = NULL;
 	BN_init(&ret->order);
@@ -132,6 +133,9 @@ void EC_GROUP_free(EC_GROUP *group)
 
 	EC_EX_DATA_free_all_data(&group->extra_data);
 
+	if (group->mont_data)
+		BN_MONT_CTX_free(group->mont_data);
+
 	if (group->generator != NULL)
 		EC_POINT_free(group->generator);
 	BN_free(&group->order);
@@ -155,6 +159,9 @@ void EC_GROUP_clear_free(EC_GROUP *group)
 
 	EC_EX_DATA_clear_free_all_data(&group->extra_data);
 
+	if (group->mont_data)
+		BN_MONT_CTX_free(group->mont_data);
+
 	if (group->generator != NULL)
 		EC_POINT_clear_free(group->generator);
 	BN_clear_free(&group->order);
@@ -200,6 +207,25 @@ int EC_GROUP_copy(EC_GROUP *dest, const EC_GROUP *src)
 			return 0;
 		}
 
+	if (src->mont_data != NULL)
+		{
+		if (dest->mont_data == NULL)
+			{
+			dest->mont_data = BN_MONT_CTX_new();
+			if (dest->mont_data == NULL) return 0;
+			}
+		if (!BN_MONT_CTX_copy(dest->mont_data, src->mont_data)) return 0;
+		}
+	else
+		{
+		/* src->generator == NULL */
+		if (dest->mont_data != NULL)
+			{
+			BN_MONT_CTX_free(dest->mont_data);
+			dest->mont_data = NULL;
+			}
+		}
+
 	if (src->generator != NULL)
 		{
 		if (dest->generator == NULL)
@@ -309,6 +335,11 @@ int EC_GROUP_set_generator(EC_GROUP *group, const EC_POINT *generator, const BIG
 	else
 		BN_zero(&group->cofactor);
 
+	/* We ignore the return value because some groups have an order with
+	 * factors of two, which makes the Montgomery setup fail.
+	 * |group->mont_data| will be NULL in this case. */
+	ec_precompute_mont_data(group);
+
 	return 1;
 	}
 
@@ -318,6 +349,10 @@ const EC_POINT *EC_GROUP_get0_generator(const EC_GROUP *group)
 	return group->generator;
 	}
 
+BN_MONT_CTX *EC_GROUP_get_mont_data(const EC_GROUP *group)
+	{
+	return group->mont_data;
+	}
 
 int EC_GROUP_get_order(const EC_GROUP *group, BIGNUM *order, BN_CTX *ctx)
 	{
@@ -1097,3 +1132,39 @@ int EC_GROUP_have_precompute_mult(const EC_GROUP *group)
 	else
 		return 0; /* cannot tell whether precomputation has been performed */
 	}
+
+/* ec_precompute_mont_data sets |group->mont_data| from |group->order| and
+ * returns one on success. On error it returns zero. */
+int ec_precompute_mont_data(EC_GROUP *group)
+	{
+	BN_CTX *ctx = BN_CTX_new();
+	int ret = 0;
+
+	if (group->mont_data)
+		{
+		BN_MONT_CTX_free(group->mont_data);
+		group->mont_data = NULL;
+		}
+
+	if (ctx == NULL)
+		goto err;
+
+	group->mont_data = BN_MONT_CTX_new();
+	if (!group->mont_data)
+		goto err;
+
+	if (!BN_MONT_CTX_set(group->mont_data, &group->order, ctx))
+		{
+		BN_MONT_CTX_free(group->mont_data);
+		group->mont_data = NULL;
+		goto err;
+		}
+
+	ret = 1;
+
+err:
+
+	if (ctx)
+		BN_CTX_free(ctx);
+	return ret;
+	}
diff --git a/crypto/ecdsa/ecs_ossl.c b/crypto/ecdsa/ecs_ossl.c
index adab1f74b41daf6e719ca1fdae1ba817085c7802..97541a24b5feb2801b42d781408fab5385702c53 100644
--- a/crypto/ecdsa/ecs_ossl.c
+++ b/crypto/ecdsa/ecs_ossl.c
@@ -219,11 +219,37 @@ static int ecdsa_sign_setup(EC_KEY *eckey, BN_CTX *ctx_in,
 	while (BN_is_zero(r));
 
 	/* compute the inverse of k */
-	if (!BN_mod_inverse(k, k, order, ctx))
-	{
-		ECDSAerr(ECDSA_F_ECDSA_SIGN_SETUP, ERR_R_BN_LIB);
-		goto err;	
-	}
+	if (EC_GROUP_get_mont_data(group) != NULL)
+		{
+		/* We want inverse in constant time, therefore we utilize the
+		 * fact order must be prime and use Fermats Little Theorem
+		 * instead. */
+		if (!BN_set_word(X, 2) )
+			{
+			ECDSAerr(ECDSA_F_ECDSA_SIGN_SETUP, ERR_R_BN_LIB);
+			goto err;
+			}
+		if (!BN_mod_sub(X, order, X, order, ctx))
+			{
+			ECDSAerr(ECDSA_F_ECDSA_SIGN_SETUP, ERR_R_BN_LIB);
+			goto err;
+			}
+		BN_set_flags(X, BN_FLG_CONSTTIME);
+		if (!BN_mod_exp_mont_consttime(k, k, X, order, ctx, EC_GROUP_get_mont_data(group)))
+			{
+			ECDSAerr(ECDSA_F_ECDSA_SIGN_SETUP, ERR_R_BN_LIB);
+			goto err;
+			}
+		}
+	else
+		{
+		if (!BN_mod_inverse(k, k, order, ctx))
+			{
+			ECDSAerr(ECDSA_F_ECDSA_SIGN_SETUP, ERR_R_BN_LIB);
+			goto err;	
+			}
+		}
+
 	/* clear old values if necessary */
 	if (*rp != NULL)
 		BN_clear_free(*rp);