/*
- * Copyright 2014-2016 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2014-2017 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright (c) 2014, Intel Corporation. All Rights Reserved.
+ * Copyright (c) 2015, CloudFlare, Inc.
*
* Licensed under the OpenSSL license (the "License"). You may not use
* this file except in compliance with the License. You can obtain a copy
* in the file LICENSE in the source distribution or at
* https://www.openssl.org/source/license.html
+ *
+ * Originally written by Shay Gueron (1, 2), and Vlad Krasnov (1, 3)
+ * (1) Intel Corporation, Israel Development Center, Haifa, Israel
+ * (2) University of Haifa, Israel
+ * (3) CloudFlare, Inc.
+ *
+ * Reference:
+ * S.Gueron and V.Krasnov, "Fast Prime Field Elliptic Curve Cryptography with
+ * 256 Bit Primes"
*/
-/******************************************************************************
- * *
- * Copyright 2014 Intel Corporation *
- * *
- * Licensed under the Apache License, Version 2.0 (the "License"); *
- * you may not use this file except in compliance with the License. *
- * You may obtain a copy of the License at *
- * *
- * http://www.apache.org/licenses/LICENSE-2.0 *
- * *
- * Unless required by applicable law or agreed to in writing, software *
- * distributed under the License is distributed on an "AS IS" BASIS, *
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. *
- * See the License for the specific language governing permissions and *
- * limitations under the License. *
- * *
- ******************************************************************************
- * *
- * Developers and authors: *
- * Shay Gueron (1, 2), and Vlad Krasnov (1) *
- * (1) Intel Corporation, Israel Development Center *
- * (2) University of Haifa *
- * Reference: *
- * S.Gueron and V.Krasnov, "Fast Prime Field Elliptic Curve Cryptography with *
- * 256 Bit Primes" *
- * *
- ******************************************************************************/
-
#include <string.h>
#include "internal/cryptlib.h"
#include "internal/bn_int.h"
#include "ec_lcl.h"
+#include "internal/refcount.h"
#if BN_BITS2 != 64
# define TOBN(hi,lo) lo,hi
*/
PRECOMP256_ROW *precomp;
void *precomp_storage;
- int references;
+ CRYPTO_REF_COUNT references;
CRYPTO_RWLOCK *lock;
};
return res;
}
+/*
+ * For reference, this macro is used only when new ecp_nistz256 assembly
+ * module is being developed. For example, configure with
+ * -DECP_NISTZ256_REFERENCE_IMPLEMENTATION and implement only functions
+ * performing simplest arithmetic operations on 256-bit vectors. Then
+ * work on implementation of higher-level functions performing point
+ * operations. Then remove ECP_NISTZ256_REFERENCE_IMPLEMENTATION
+ * and never define it again. (The correct macro denoting presence of
+ * ecp_nistz256 module is ECP_NISTZ256_ASM.)
+ */
#ifndef ECP_NISTZ256_REFERENCE_IMPLEMENTATION
void ecp_nistz256_point_double(P256_POINT *r, const P256_POINT *a);
void ecp_nistz256_point_add(P256_POINT *r,
const BN_ULONG *in2_y = b->Y;
const BN_ULONG *in2_z = b->Z;
- /* We encode infinity as (0,0), which is not on the curve,
- * so it is OK. */
- in1infty = (in1_x[0] | in1_x[1] | in1_x[2] | in1_x[3] |
- in1_y[0] | in1_y[1] | in1_y[2] | in1_y[3]);
+ /*
+ * Infinity in encoded as (,,0)
+ */
+ in1infty = (in1_z[0] | in1_z[1] | in1_z[2] | in1_z[3]);
if (P256_LIMBS == 8)
- in1infty |= (in1_x[4] | in1_x[5] | in1_x[6] | in1_x[7] |
- in1_y[4] | in1_y[5] | in1_y[6] | in1_y[7]);
+ in1infty |= (in1_z[4] | in1_z[5] | in1_z[6] | in1_z[7]);
- in2infty = (in2_x[0] | in2_x[1] | in2_x[2] | in2_x[3] |
- in2_y[0] | in2_y[1] | in2_y[2] | in2_y[3]);
+ in2infty = (in2_z[0] | in2_z[1] | in2_z[2] | in2_z[3]);
if (P256_LIMBS == 8)
- in2infty |= (in2_x[4] | in2_x[5] | in2_x[6] | in2_x[7] |
- in2_y[4] | in2_y[5] | in2_y[6] | in2_y[7]);
+ in2infty |= (in2_z[4] | in2_z[5] | in2_z[6] | in2_z[7]);
in1infty = is_zero(in1infty);
in2infty = is_zero(in2infty);
const BN_ULONG *in2_y = b->Y;
/*
- * In affine representation we encode infty as (0,0), which is not on the
- * curve, so it is OK
+ * Infinity in encoded as (,,0)
*/
- in1infty = (in1_x[0] | in1_x[1] | in1_x[2] | in1_x[3] |
- in1_y[0] | in1_y[1] | in1_y[2] | in1_y[3]);
+ in1infty = (in1_z[0] | in1_z[1] | in1_z[2] | in1_z[3]);
if (P256_LIMBS == 8)
- in1infty |= (in1_x[4] | in1_x[5] | in1_x[6] | in1_x[7] |
- in1_y[4] | in1_y[5] | in1_y[6] | in1_y[7]);
+ in1infty |= (in1_z[4] | in1_z[5] | in1_z[6] | in1_z[7]);
+ /*
+ * In affine representation we encode infinity as (0,0), which is
+ * not on the curve, so it is OK
+ */
in2infty = (in2_x[0] | in2_x[1] | in2_x[2] | in2_x[3] |
in2_y[0] | in2_y[1] | in2_y[2] | in2_y[3]);
if (P256_LIMBS == 8)
}
/* Coordinates of G, for which we have precomputed tables */
-const static BN_ULONG def_xG[P256_LIMBS] = {
+static const BN_ULONG def_xG[P256_LIMBS] = {
TOBN(0x79e730d4, 0x18a9143c), TOBN(0x75ba95fc, 0x5fedb601),
TOBN(0x79fb732b, 0x77622510), TOBN(0x18905f76, 0xa53755c6)
};
-const static BN_ULONG def_yG[P256_LIMBS] = {
+static const BN_ULONG def_yG[P256_LIMBS] = {
TOBN(0xddf25357, 0xce95560a), TOBN(0x8b4ab8e4, 0xba19e45c),
TOBN(0xd2e88688, 0xdd21f325), TOBN(0x8571ff18, 0x25885d85)
};
*/
#if defined(ECP_NISTZ256_AVX2)
# if !(defined(__x86_64) || defined(__x86_64__) || \
- defined(_M_AMD64) || defined(_MX64)) || \
+ defined(_M_AMD64) || defined(_M_X64)) || \
!(defined(__GNUC__) || defined(_MSC_VER)) /* this is for ALIGN32 */
# undef ECP_NISTZ256_AVX2
# else
} else
#endif
{
+ BN_ULONG infty;
+
/* First window */
wvalue = (p_str[0] << 1) & mask;
idx += window_size;
ecp_nistz256_neg(p.p.Z, p.p.Y);
copy_conditional(p.p.Y, p.p.Z, wvalue & 1);
- memcpy(p.p.Z, ONE, sizeof(ONE));
+ /*
+ * Since affine infinity is encoded as (0,0) and
+ * Jacobian ias (,,0), we need to harmonize them
+ * by assigning "one" or zero to Z.
+ */
+ infty = (p.p.X[0] | p.p.X[1] | p.p.X[2] | p.p.X[3] |
+ p.p.Y[0] | p.p.Y[1] | p.p.Y[2] | p.p.Y[3]);
+ if (P256_LIMBS == 8)
+ infty |= (p.p.X[4] | p.p.X[5] | p.p.X[6] | p.p.X[7] |
+ p.p.Y[4] | p.p.Y[5] | p.p.Y[6] | p.p.Y[7]);
+
+ infty = 0 - is_zero(infty);
+ infty = ~infty;
+
+ p.p.Z[0] = ONE[0] & infty;
+ p.p.Z[1] = ONE[1] & infty;
+ p.p.Z[2] = ONE[2] & infty;
+ p.p.Z[3] = ONE[3] & infty;
+ if (P256_LIMBS == 8) {
+ p.p.Z[4] = ONE[4] & infty;
+ p.p.Z[5] = ONE[5] & infty;
+ p.p.Z[6] = ONE[6] & infty;
+ p.p.Z[7] = ONE[7] & infty;
+ }
for (i = 1; i < 37; i++) {
unsigned int off = (idx - 1) / 8;
{
int i;
if (p != NULL)
- CRYPTO_atomic_add(&p->references, 1, &i, p->lock);
+ CRYPTO_UP_REF(&p->references, &i, p->lock);
return p;
}
if (pre == NULL)
return;
- CRYPTO_atomic_add(&pre->references, -1, &i, pre->lock);
+ CRYPTO_DOWN_REF(&pre->references, &i, pre->lock);
REF_PRINT_COUNT("EC_nistz256", x);
if (i > 0)
return;
return HAVEPRECOMP(group, nistz256);
}
+#if defined(__x86_64) || defined(__x86_64__) || \
+ defined(_M_AMD64) || defined(_M_X64) || \
+ defined(__powerpc64__) || defined(_ARCH_PP64) || \
+ defined(__aarch64__)
+/*
+ * Montgomery mul modulo Order(P): res = a*b*2^-256 mod Order(P)
+ */
+void ecp_nistz256_ord_mul_mont(BN_ULONG res[P256_LIMBS],
+ const BN_ULONG a[P256_LIMBS],
+ const BN_ULONG b[P256_LIMBS]);
+void ecp_nistz256_ord_sqr_mont(BN_ULONG res[P256_LIMBS],
+ const BN_ULONG a[P256_LIMBS],
+ int rep);
+
+static int ecp_nistz256_inv_mod_ord(const EC_GROUP *group, BIGNUM *r,
+ BIGNUM *x, BN_CTX *ctx)
+{
+ /* RR = 2^512 mod ord(p256) */
+ static const BN_ULONG RR[P256_LIMBS] = { TOBN(0x83244c95,0xbe79eea2),
+ TOBN(0x4699799c,0x49bd6fa6),
+ TOBN(0x2845b239,0x2b6bec59),
+ TOBN(0x66e12d94,0xf3d95620) };
+ /* The constant 1 (unlike ONE that is one in Montgomery representation) */
+ static const BN_ULONG one[P256_LIMBS] = { TOBN(0,1),TOBN(0,0),
+ TOBN(0,0),TOBN(0,0) };
+ /* expLo - the low 128bit of the exponent we use (ord(p256) - 2),
+ * split into 4bit windows */
+ static const unsigned char expLo[32] = { 0xb,0xc,0xe,0x6,0xf,0xa,0xa,0xd,
+ 0xa,0x7,0x1,0x7,0x9,0xe,0x8,0x4,
+ 0xf,0x3,0xb,0x9,0xc,0xa,0xc,0x2,
+ 0xf,0xc,0x6,0x3,0x2,0x5,0x4,0xf };
+ /*
+ * We don't use entry 0 in the table, so we omit it and address
+ * with -1 offset.
+ */
+ BN_ULONG table[15][P256_LIMBS];
+ BN_ULONG out[P256_LIMBS], t[P256_LIMBS];
+ int i, ret = 0;
+
+ /*
+ * Catch allocation failure early.
+ */
+ if (bn_wexpand(r, P256_LIMBS) == NULL) {
+ ECerr(EC_F_ECP_NISTZ256_INV_MOD_ORD, ERR_R_BN_LIB);
+ goto err;
+ }
+
+ if ((BN_num_bits(x) > 256) || BN_is_negative(x)) {
+ BIGNUM *tmp;
+
+ if ((tmp = BN_CTX_get(ctx)) == NULL
+ || !BN_nnmod(tmp, x, group->order, ctx)) {
+ ECerr(EC_F_ECP_NISTZ256_INV_MOD_ORD, ERR_R_BN_LIB);
+ goto err;
+ }
+ x = tmp;
+ }
+
+ if (!ecp_nistz256_bignum_to_field_elem(t, x)) {
+ ECerr(EC_F_ECP_NISTZ256_INV_MOD_ORD, EC_R_COORDINATES_OUT_OF_RANGE);
+ goto err;
+ }
+
+ ecp_nistz256_ord_mul_mont(table[0], t, RR);
+ for (i = 2; i < 16; i += 2) {
+ ecp_nistz256_ord_sqr_mont(table[i-1], table[i/2-1], 1);
+ ecp_nistz256_ord_mul_mont(table[i], table[i-1], table[0]);
+ }
+
+ /*
+ * The top 128bit of the exponent are highly redudndant, so we
+ * perform an optimized flow
+ */
+ ecp_nistz256_ord_sqr_mont(t, table[15-1], 4); /* f0 */
+ ecp_nistz256_ord_mul_mont(t, t, table[15-1]); /* ff */
+
+ ecp_nistz256_ord_sqr_mont(out, t, 8); /* ff00 */
+ ecp_nistz256_ord_mul_mont(out, out, t); /* ffff */
+
+ ecp_nistz256_ord_sqr_mont(t, out, 16); /* ffff0000 */
+ ecp_nistz256_ord_mul_mont(t, t, out); /* ffffffff */
+
+ ecp_nistz256_ord_sqr_mont(out, t, 64); /* ffffffff0000000000000000 */
+ ecp_nistz256_ord_mul_mont(out, out, t); /* ffffffff00000000ffffffff */
+
+ ecp_nistz256_ord_sqr_mont(out, out, 32); /* ffffffff00000000ffffffff00000000 */
+ ecp_nistz256_ord_mul_mont(out, out, t); /* ffffffff00000000ffffffffffffffff */
+
+ /*
+ * The bottom 128 bit of the exponent are easier done with a table
+ */
+ for(i = 0; i < 32; i++) {
+ ecp_nistz256_ord_sqr_mont(out, out, 4);
+ /* The exponent is public, no need in constant-time access */
+ ecp_nistz256_ord_mul_mont(out, out, table[expLo[i]-1]);
+ }
+ ecp_nistz256_ord_mul_mont(out, out, one);
+
+ /*
+ * Can't fail, but check return code to be consistent anyway.
+ */
+ if (!bn_set_words(r, out, P256_LIMBS))
+ goto err;
+
+ ret = 1;
+err:
+ return ret;
+}
+#else
+# define ecp_nistz256_inv_mod_ord NULL
+#endif
+
const EC_METHOD *EC_GFp_nistz256_method(void)
{
static const EC_METHOD ret = {
ec_key_simple_generate_public_key,
0, /* keycopy */
0, /* keyfinish */
- ecdh_simple_compute_key
+ ecdh_simple_compute_key,
+ ecp_nistz256_inv_mod_ord /* can be #defined-ed NULL */
};
return &ret;