-/* crypto/ec/ecp_nistp521.c */
/*
- * Written by Adam Langley (Google) for the OpenSSL project
+ * Copyright 2011-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the Apache License 2.0 (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
*/
+
/* Copyright 2011 Google Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* work which got its smarts from Daniel J. Bernstein's work on the same.
*/
-#include <openssl/opensslconf.h>
-#ifndef OPENSSL_NO_EC_NISTP_64_GCC_128
-
-# ifndef OPENSSL_SYS_VMS
-# include <stdint.h>
-# else
-# include <inttypes.h>
-# endif
+#include <openssl/e_os2.h>
+#ifdef OPENSSL_NO_EC_NISTP_64_GCC_128
+NON_EMPTY_TRANSLATION_UNIT
+#else
# include <string.h>
# include <openssl/err.h>
# include "ec_lcl.h"
-# if defined(__GNUC__) && (__GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ >= 1))
+# if defined(__SIZEOF_INT128__) && __SIZEOF_INT128__==16
/* even with gcc, the typedef won't work for 32-bit platforms */
typedef __uint128_t uint128_t; /* nonstandard; implemented by gcc on 64-bit
* platforms */
# else
-# error "Need GCC 3.1 or later to define type uint128_t"
+# error "Your compiler doesn't appear to support 128-bit integer types"
# endif
typedef uint8_t u8;
typedef uint64_t u64;
-typedef int64_t s64;
/*
* The underlying field. P521 operates over GF(2^521-1). We can serialise an
(*((limb *) & out[58])) = in[8];
}
-/* To preserve endianness when using BN_bn2bin and BN_bin2bn */
-static void flip_endian(u8 *out, const u8 *in, unsigned len)
-{
- unsigned i;
- for (i = 0; i < len; ++i)
- out[i] = in[len - 1 - i];
-}
-
/* BN_to_felem converts an OpenSSL BIGNUM into an felem */
static int BN_to_felem(felem out, const BIGNUM *bn)
{
- felem_bytearray b_in;
felem_bytearray b_out;
- unsigned num_bytes;
+ int num_bytes;
- /* BN_bn2bin eats leading zeroes */
- memset(b_out, 0, sizeof b_out);
- num_bytes = BN_num_bytes(bn);
- if (num_bytes > sizeof b_out) {
+ if (BN_is_negative(bn)) {
ECerr(EC_F_BN_TO_FELEM, EC_R_BIGNUM_OUT_OF_RANGE);
return 0;
}
- if (BN_is_negative(bn)) {
+ num_bytes = BN_bn2lebinpad(bn, b_out, sizeof(b_out));
+ if (num_bytes < 0) {
ECerr(EC_F_BN_TO_FELEM, EC_R_BIGNUM_OUT_OF_RANGE);
return 0;
}
- num_bytes = BN_bn2bin(bn, b_in);
- flip_endian(b_out, b_in, num_bytes);
bin66_to_felem(out, b_out);
return 1;
}
/* felem_to_BN converts an felem into an OpenSSL BIGNUM */
static BIGNUM *felem_to_BN(BIGNUM *out, const felem in)
{
- felem_bytearray b_in, b_out;
- felem_to_bin66(b_in, in);
- flip_endian(b_out, b_in, sizeof b_out);
- return BN_bin2bn(b_out, sizeof b_out, out);
+ felem_bytearray b_out;
+ felem_to_bin66(b_out, in);
+ return BN_lebin2bn(b_out, sizeof(b_out), out);
}
/*-
static void felem_diff_128_64(largefelem out, const felem in)
{
/*
- * In order to prevent underflow, we add 0 mod p before subtracting.
+ * In order to prevent underflow, we add 64p mod p (which is equivalent
+ * to 0 mod p) before subtracting. p is 2^521 - 1, i.e. in binary a 521
+ * digit number with all bits set to 1. See "The representation of field
+ * elements" comment above for a description of how limbs are used to
+ * represent a number. 64p is represented with 8 limbs containing a number
+ * with 58 bits set and one limb with a number with 57 bits set.
*/
- static const limb two63m6 = (((limb) 1) << 62) - (((limb) 1) << 5);
- static const limb two63m5 = (((limb) 1) << 62) - (((limb) 1) << 4);
+ static const limb two63m6 = (((limb) 1) << 63) - (((limb) 1) << 6);
+ static const limb two63m5 = (((limb) 1) << 63) - (((limb) 1) << 5);
out[0] += two63m6 - in[0];
out[1] += two63m5 - in[1];
felem_scalar(inx2, in, 2);
felem_scalar(inx4, in, 4);
- /*-
- * We have many cases were we want to do
- * in[x] * in[y] +
- * in[y] * in[x]
- * This is obviously just
- * 2 * in[x] * in[y]
- * However, rather than do the doubling on the 128 bit result, we
- * double one of the inputs to the multiplication by reading from
- * |inx2| */
+ /*-
+ * We have many cases were we want to do
+ * in[x] * in[y] +
+ * in[y] * in[x]
+ * This is obviously just
+ * 2 * in[x] * in[y]
+ * However, rather than do the doubling on the 128 bit result, we
+ * double one of the inputs to the multiplication by reading from
+ * |inx2|
+ */
out[0] = ((uint128_t) in[0]) * in[0];
out[1] = ((uint128_t) in[0]) * inx2[1];
out[2] = ((uint128_t) in[0]) * inx2[2] + ((uint128_t) in[1]) * in[1];
out[3] = ((uint128_t) in[0]) * inx2[3] + ((uint128_t) in[1]) * inx2[2];
out[4] = ((uint128_t) in[0]) * inx2[4] +
- ((uint128_t) in[1]) * inx2[3] + ((uint128_t) in[2]) * in[2];
+ ((uint128_t) in[1]) * inx2[3] + ((uint128_t) in[2]) * in[2];
out[5] = ((uint128_t) in[0]) * inx2[5] +
- ((uint128_t) in[1]) * inx2[4] + ((uint128_t) in[2]) * inx2[3];
+ ((uint128_t) in[1]) * inx2[4] + ((uint128_t) in[2]) * inx2[3];
out[6] = ((uint128_t) in[0]) * inx2[6] +
- ((uint128_t) in[1]) * inx2[5] +
- ((uint128_t) in[2]) * inx2[4] + ((uint128_t) in[3]) * in[3];
+ ((uint128_t) in[1]) * inx2[5] +
+ ((uint128_t) in[2]) * inx2[4] + ((uint128_t) in[3]) * in[3];
out[7] = ((uint128_t) in[0]) * inx2[7] +
- ((uint128_t) in[1]) * inx2[6] +
- ((uint128_t) in[2]) * inx2[5] + ((uint128_t) in[3]) * inx2[4];
+ ((uint128_t) in[1]) * inx2[6] +
+ ((uint128_t) in[2]) * inx2[5] + ((uint128_t) in[3]) * inx2[4];
out[8] = ((uint128_t) in[0]) * inx2[8] +
- ((uint128_t) in[1]) * inx2[7] +
- ((uint128_t) in[2]) * inx2[6] +
- ((uint128_t) in[3]) * inx2[5] + ((uint128_t) in[4]) * in[4];
+ ((uint128_t) in[1]) * inx2[7] +
+ ((uint128_t) in[2]) * inx2[6] +
+ ((uint128_t) in[3]) * inx2[5] + ((uint128_t) in[4]) * in[4];
/*
* The remaining limbs fall above 2^521, with the first falling at 2^522.
/* 9 */
out[0] += ((uint128_t) in[1]) * inx4[8] +
- ((uint128_t) in[2]) * inx4[7] +
- ((uint128_t) in[3]) * inx4[6] + ((uint128_t) in[4]) * inx4[5];
+ ((uint128_t) in[2]) * inx4[7] +
+ ((uint128_t) in[3]) * inx4[6] + ((uint128_t) in[4]) * inx4[5];
/* 10 */
out[1] += ((uint128_t) in[2]) * inx4[8] +
- ((uint128_t) in[3]) * inx4[7] +
- ((uint128_t) in[4]) * inx4[6] + ((uint128_t) in[5]) * inx2[5];
+ ((uint128_t) in[3]) * inx4[7] +
+ ((uint128_t) in[4]) * inx4[6] + ((uint128_t) in[5]) * inx2[5];
/* 11 */
out[2] += ((uint128_t) in[3]) * inx4[8] +
- ((uint128_t) in[4]) * inx4[7] + ((uint128_t) in[5]) * inx4[6];
+ ((uint128_t) in[4]) * inx4[7] + ((uint128_t) in[5]) * inx4[6];
/* 12 */
out[3] += ((uint128_t) in[4]) * inx4[8] +
- ((uint128_t) in[5]) * inx4[7] + ((uint128_t) in[6]) * inx2[6];
+ ((uint128_t) in[5]) * inx4[7] + ((uint128_t) in[6]) * inx2[6];
/* 13 */
out[4] += ((uint128_t) in[5]) * inx4[8] + ((uint128_t) in[6]) * inx4[7];
out[0] = ((uint128_t) in1[0]) * in2[0];
- out[1] = ((uint128_t) in1[0]) * in2[1] + ((uint128_t) in1[1]) * in2[0];
+ out[1] = ((uint128_t) in1[0]) * in2[1] +
+ ((uint128_t) in1[1]) * in2[0];
out[2] = ((uint128_t) in1[0]) * in2[2] +
- ((uint128_t) in1[1]) * in2[1] + ((uint128_t) in1[2]) * in2[0];
+ ((uint128_t) in1[1]) * in2[1] +
+ ((uint128_t) in1[2]) * in2[0];
out[3] = ((uint128_t) in1[0]) * in2[3] +
- ((uint128_t) in1[1]) * in2[2] +
- ((uint128_t) in1[2]) * in2[1] + ((uint128_t) in1[3]) * in2[0];
+ ((uint128_t) in1[1]) * in2[2] +
+ ((uint128_t) in1[2]) * in2[1] +
+ ((uint128_t) in1[3]) * in2[0];
out[4] = ((uint128_t) in1[0]) * in2[4] +
- ((uint128_t) in1[1]) * in2[3] +
- ((uint128_t) in1[2]) * in2[2] +
- ((uint128_t) in1[3]) * in2[1] + ((uint128_t) in1[4]) * in2[0];
+ ((uint128_t) in1[1]) * in2[3] +
+ ((uint128_t) in1[2]) * in2[2] +
+ ((uint128_t) in1[3]) * in2[1] +
+ ((uint128_t) in1[4]) * in2[0];
out[5] = ((uint128_t) in1[0]) * in2[5] +
- ((uint128_t) in1[1]) * in2[4] +
- ((uint128_t) in1[2]) * in2[3] +
- ((uint128_t) in1[3]) * in2[2] +
- ((uint128_t) in1[4]) * in2[1] + ((uint128_t) in1[5]) * in2[0];
+ ((uint128_t) in1[1]) * in2[4] +
+ ((uint128_t) in1[2]) * in2[3] +
+ ((uint128_t) in1[3]) * in2[2] +
+ ((uint128_t) in1[4]) * in2[1] +
+ ((uint128_t) in1[5]) * in2[0];
out[6] = ((uint128_t) in1[0]) * in2[6] +
- ((uint128_t) in1[1]) * in2[5] +
- ((uint128_t) in1[2]) * in2[4] +
- ((uint128_t) in1[3]) * in2[3] +
- ((uint128_t) in1[4]) * in2[2] +
- ((uint128_t) in1[5]) * in2[1] + ((uint128_t) in1[6]) * in2[0];
+ ((uint128_t) in1[1]) * in2[5] +
+ ((uint128_t) in1[2]) * in2[4] +
+ ((uint128_t) in1[3]) * in2[3] +
+ ((uint128_t) in1[4]) * in2[2] +
+ ((uint128_t) in1[5]) * in2[1] +
+ ((uint128_t) in1[6]) * in2[0];
out[7] = ((uint128_t) in1[0]) * in2[7] +
- ((uint128_t) in1[1]) * in2[6] +
- ((uint128_t) in1[2]) * in2[5] +
- ((uint128_t) in1[3]) * in2[4] +
- ((uint128_t) in1[4]) * in2[3] +
- ((uint128_t) in1[5]) * in2[2] +
- ((uint128_t) in1[6]) * in2[1] + ((uint128_t) in1[7]) * in2[0];
+ ((uint128_t) in1[1]) * in2[6] +
+ ((uint128_t) in1[2]) * in2[5] +
+ ((uint128_t) in1[3]) * in2[4] +
+ ((uint128_t) in1[4]) * in2[3] +
+ ((uint128_t) in1[5]) * in2[2] +
+ ((uint128_t) in1[6]) * in2[1] +
+ ((uint128_t) in1[7]) * in2[0];
out[8] = ((uint128_t) in1[0]) * in2[8] +
- ((uint128_t) in1[1]) * in2[7] +
- ((uint128_t) in1[2]) * in2[6] +
- ((uint128_t) in1[3]) * in2[5] +
- ((uint128_t) in1[4]) * in2[4] +
- ((uint128_t) in1[5]) * in2[3] +
- ((uint128_t) in1[6]) * in2[2] +
- ((uint128_t) in1[7]) * in2[1] + ((uint128_t) in1[8]) * in2[0];
+ ((uint128_t) in1[1]) * in2[7] +
+ ((uint128_t) in1[2]) * in2[6] +
+ ((uint128_t) in1[3]) * in2[5] +
+ ((uint128_t) in1[4]) * in2[4] +
+ ((uint128_t) in1[5]) * in2[3] +
+ ((uint128_t) in1[6]) * in2[2] +
+ ((uint128_t) in1[7]) * in2[1] +
+ ((uint128_t) in1[8]) * in2[0];
/* See comment in felem_square about the use of in2x2 here */
out[0] += ((uint128_t) in1[1]) * in2x2[8] +
- ((uint128_t) in1[2]) * in2x2[7] +
- ((uint128_t) in1[3]) * in2x2[6] +
- ((uint128_t) in1[4]) * in2x2[5] +
- ((uint128_t) in1[5]) * in2x2[4] +
- ((uint128_t) in1[6]) * in2x2[3] +
- ((uint128_t) in1[7]) * in2x2[2] + ((uint128_t) in1[8]) * in2x2[1];
+ ((uint128_t) in1[2]) * in2x2[7] +
+ ((uint128_t) in1[3]) * in2x2[6] +
+ ((uint128_t) in1[4]) * in2x2[5] +
+ ((uint128_t) in1[5]) * in2x2[4] +
+ ((uint128_t) in1[6]) * in2x2[3] +
+ ((uint128_t) in1[7]) * in2x2[2] +
+ ((uint128_t) in1[8]) * in2x2[1];
out[1] += ((uint128_t) in1[2]) * in2x2[8] +
- ((uint128_t) in1[3]) * in2x2[7] +
- ((uint128_t) in1[4]) * in2x2[6] +
- ((uint128_t) in1[5]) * in2x2[5] +
- ((uint128_t) in1[6]) * in2x2[4] +
- ((uint128_t) in1[7]) * in2x2[3] + ((uint128_t) in1[8]) * in2x2[2];
+ ((uint128_t) in1[3]) * in2x2[7] +
+ ((uint128_t) in1[4]) * in2x2[6] +
+ ((uint128_t) in1[5]) * in2x2[5] +
+ ((uint128_t) in1[6]) * in2x2[4] +
+ ((uint128_t) in1[7]) * in2x2[3] +
+ ((uint128_t) in1[8]) * in2x2[2];
out[2] += ((uint128_t) in1[3]) * in2x2[8] +
- ((uint128_t) in1[4]) * in2x2[7] +
- ((uint128_t) in1[5]) * in2x2[6] +
- ((uint128_t) in1[6]) * in2x2[5] +
- ((uint128_t) in1[7]) * in2x2[4] + ((uint128_t) in1[8]) * in2x2[3];
+ ((uint128_t) in1[4]) * in2x2[7] +
+ ((uint128_t) in1[5]) * in2x2[6] +
+ ((uint128_t) in1[6]) * in2x2[5] +
+ ((uint128_t) in1[7]) * in2x2[4] +
+ ((uint128_t) in1[8]) * in2x2[3];
out[3] += ((uint128_t) in1[4]) * in2x2[8] +
- ((uint128_t) in1[5]) * in2x2[7] +
- ((uint128_t) in1[6]) * in2x2[6] +
- ((uint128_t) in1[7]) * in2x2[5] + ((uint128_t) in1[8]) * in2x2[4];
+ ((uint128_t) in1[5]) * in2x2[7] +
+ ((uint128_t) in1[6]) * in2x2[6] +
+ ((uint128_t) in1[7]) * in2x2[5] +
+ ((uint128_t) in1[8]) * in2x2[4];
out[4] += ((uint128_t) in1[5]) * in2x2[8] +
- ((uint128_t) in1[6]) * in2x2[7] +
- ((uint128_t) in1[7]) * in2x2[6] + ((uint128_t) in1[8]) * in2x2[5];
+ ((uint128_t) in1[6]) * in2x2[7] +
+ ((uint128_t) in1[7]) * in2x2[6] +
+ ((uint128_t) in1[8]) * in2x2[5];
out[5] += ((uint128_t) in1[6]) * in2x2[8] +
- ((uint128_t) in1[7]) * in2x2[7] + ((uint128_t) in1[8]) * in2x2[6];
+ ((uint128_t) in1[7]) * in2x2[7] +
+ ((uint128_t) in1[8]) * in2x2[6];
out[6] += ((uint128_t) in1[7]) * in2x2[8] +
- ((uint128_t) in1[8]) * in2x2[7];
+ ((uint128_t) in1[8]) * in2x2[7];
out[7] += ((uint128_t) in1[8]) * in2x2[8];
}
out[1] += ((limb) in[0]) >> 58;
out[1] += (((limb) (in[0] >> 64)) & bottom52bits) << 6;
- /*-
- * out[1] < 2^58 + 2^6 + 2^58
- * = 2^59 + 2^6
- */
+ /*-
+ * out[1] < 2^58 + 2^6 + 2^58
+ * = 2^59 + 2^6
+ */
out[2] += ((limb) (in[0] >> 64)) >> 52;
out[2] += ((limb) in[1]) >> 58;
out[8] += ((limb) in[7]) >> 58;
out[8] += (((limb) (in[7] >> 64)) & bottom52bits) << 6;
- /*-
- * out[x > 1] < 2^58 + 2^6 + 2^58 + 2^12
- * < 2^59 + 2^13
- */
+ /*-
+ * out[x > 1] < 2^58 + 2^6 + 2^58 + 2^12
+ * < 2^59 + 2^13
+ */
overflow1 = ((limb) (in[7] >> 64)) >> 52;
overflow1 += ((limb) in[8]) >> 58;
out[1] += out[0] >> 58;
out[0] &= bottom58bits;
- /*-
- * out[0] < 2^58
- * out[1] < 2^59 + 2^6 + 2^13 + 2^2
- * < 2^59 + 2^14
- */
+ /*-
+ * out[0] < 2^58
+ * out[1] < 2^59 + 2^6 + 2^13 + 2^2
+ * < 2^59 + 2^14
+ */
}
static void felem_square_reduce(felem out, const felem in)
* We know that ftmp[i] < 2^63, therefore the only way that the top bit
* can be set is if is_zero was 0 before the decrement.
*/
- is_zero = ((s64) is_zero) >> 63;
+ is_zero = 0 - (is_zero >> 63);
is_p = ftmp[0] ^ kPrime[0];
is_p |= ftmp[1] ^ kPrime[1];
is_p |= ftmp[8] ^ kPrime[8];
is_p--;
- is_p = ((s64) is_p) >> 63;
+ is_p = 0 - (is_p >> 63);
is_zero |= is_p;
return is_zero;
}
-static int felem_is_zero_int(const felem in)
+static int felem_is_zero_int(const void *in)
{
return (int)(felem_is_zero(in) & ((limb) 1));
}
is_p &= is_p << 4;
is_p &= is_p << 2;
is_p &= is_p << 1;
- is_p = ((s64) is_p) >> 63;
+ is_p = 0 - (is_p >> 63);
is_p = ~is_p;
/* is_p is 0 iff |out| == 2^521-1 and all ones otherwise */
is_greater |= is_greater << 4;
is_greater |= is_greater << 2;
is_greater |= is_greater << 1;
- is_greater = ((s64) is_greater) >> 63;
+ is_greater = 0 - (is_greater >> 63);
out[0] -= kPrime[0] & is_greater;
out[1] -= kPrime[1] & is_greater;
* coordinates */
/*-
- * point_double calcuates 2*(x_in, y_in, z_in)
+ * point_double calculates 2*(x_in, y_in, z_in)
*
* The method is taken from:
* http://hyperelliptic.org/EFD/g1p/auto-shortw-jacobian-3.html#doubling-dbl-2001-b
felem_scalar64(ftmp2, 3);
/* ftmp2[i] < 3*2^60 + 3*2^15 */
felem_mul(tmp, ftmp, ftmp2);
- /*-
- * tmp[i] < 17(3*2^121 + 3*2^76)
- * = 61*2^121 + 61*2^76
- * < 64*2^121 + 64*2^76
- * = 2^127 + 2^82
- * < 2^128
- */
+ /*-
+ * tmp[i] < 17(3*2^121 + 3*2^76)
+ * = 61*2^121 + 61*2^76
+ * < 64*2^121 + 64*2^76
+ * = 2^127 + 2^82
+ * < 2^128
+ */
felem_reduce(alpha, tmp);
/* x' = alpha^2 - 8*beta */
felem_diff64(beta, x_out);
/* beta[i] < 2^61 + 2^60 + 2^16 */
felem_mul(tmp, alpha, beta);
- /*-
- * tmp[i] < 17*((2^59 + 2^14)(2^61 + 2^60 + 2^16))
- * = 17*(2^120 + 2^75 + 2^119 + 2^74 + 2^75 + 2^30)
- * = 17*(2^120 + 2^119 + 2^76 + 2^74 + 2^30)
- * < 2^128
- */
+ /*-
+ * tmp[i] < 17*((2^59 + 2^14)(2^61 + 2^60 + 2^16))
+ * = 17*(2^120 + 2^75 + 2^119 + 2^74 + 2^75 + 2^30)
+ * = 17*(2^120 + 2^119 + 2^76 + 2^74 + 2^30)
+ * < 2^128
+ */
felem_square(tmp2, gamma);
- /*-
- * tmp2[i] < 17*(2^59 + 2^14)^2
- * = 17*(2^118 + 2^74 + 2^28)
- */
+ /*-
+ * tmp2[i] < 17*(2^59 + 2^14)^2
+ * = 17*(2^118 + 2^74 + 2^28)
+ */
felem_scalar128(tmp2, 8);
- /*-
- * tmp2[i] < 8*17*(2^118 + 2^74 + 2^28)
- * = 2^125 + 2^121 + 2^81 + 2^77 + 2^35 + 2^31
- * < 2^126
- */
+ /*-
+ * tmp2[i] < 8*17*(2^118 + 2^74 + 2^28)
+ * = 2^125 + 2^121 + 2^81 + 2^77 + 2^35 + 2^31
+ * < 2^126
+ */
felem_diff128(tmp, tmp2);
- /*-
- * tmp[i] < 2^127 - 2^69 + 17(2^120 + 2^119 + 2^76 + 2^74 + 2^30)
- * = 2^127 + 2^124 + 2^122 + 2^120 + 2^118 + 2^80 + 2^78 + 2^76 +
- * 2^74 + 2^69 + 2^34 + 2^30
- * < 2^128
- */
+ /*-
+ * tmp[i] < 2^127 - 2^69 + 17(2^120 + 2^119 + 2^76 + 2^74 + 2^30)
+ * = 2^127 + 2^124 + 2^122 + 2^120 + 2^118 + 2^80 + 2^78 + 2^76 +
+ * 2^74 + 2^69 + 2^34 + 2^30
+ * < 2^128
+ */
felem_reduce(y_out, tmp);
}
}
/*-
- * point_add calcuates (x1, y1, z1) + (x2, y2, z2)
+ * point_add calculates (x1, y1, z1) + (x2, y2, z2)
*
* The method is taken from
* http://hyperelliptic.org/EFD/g1p/auto-shortw-jacobian-3.html#addition-add-2007-bl,
* adapted for mixed addition (z2 = 1, or z2 = 0 for the point at infinity).
*
* This function includes a branch for checking whether the two input points
- * are equal (while not equal to the point at infinity). This case never
- * happens during single point multiplication, so there is no timing leak for
- * ECDH or ECDSA signing. */
+ * are equal (while not equal to the point at infinity). See comment below
+ * on constant-time.
+ */
static void point_add(felem x3, felem y3, felem z3,
const felem x1, const felem y1, const felem z1,
const int mixed, const felem x2, const felem y2,
/* ftmp5[i] < 2^61 */
if (x_equal && y_equal && !z1_is_zero && !z2_is_zero) {
+ /*
+ * This is obviously not constant-time but it will almost-never happen
+ * for ECDH / ECDSA. The case where it can happen is during scalar-mult
+ * where the intermediate value gets very close to the group order.
+ * Since |ec_GFp_nistp_recode_scalar_bits| produces signed digits for
+ * the scalar, it's possible for the intermediate value to be a small
+ * negative multiple of the base point, and for the final signed digit
+ * to be the same value. We believe that this only occurs for the scalar
+ * 1fffffffffffffffffffffffffffffffffffffffffffffffffffffffffff
+ * ffffffa51868783bf2f966b7fcc0148f709a5d03bb5c9b8899c47aebb6fb
+ * 71e913863f7, in that case the penultimate intermediate is -9G and
+ * the final digit is also -9G. Since this only happens for a single
+ * scalar, the timing leak is irrelevant. (Any attacker who wanted to
+ * check whether a secret scalar was that exact value, can already do
+ * so.)
+ */
point_double(x3, y3, z3, x1, y1, z1);
return;
}
* Tables for other points have table[i] = iG for i in 0 .. 16. */
/* gmul is the table of precomputed base points */
-static const felem gmul[16][3] = { {{0, 0, 0, 0, 0, 0, 0, 0, 0},
- {0, 0, 0, 0, 0, 0, 0, 0, 0},
- {0, 0, 0, 0, 0, 0, 0, 0, 0}},
+static const felem gmul[16][3] = {
+{{0, 0, 0, 0, 0, 0, 0, 0, 0},
+ {0, 0, 0, 0, 0, 0, 0, 0, 0},
+ {0, 0, 0, 0, 0, 0, 0, 0, 0}},
{{0x017e7e31c2e5bd66, 0x022cf0615a90a6fe, 0x00127a2ffa8de334,
0x01dfbf9d64a3f877, 0x006b4d3dbaa14b5e, 0x014fed487e0a2bd8,
0x015b4429c6481390, 0x03a73678fb2d988e, 0x00c6858e06b70404},
{
unsigned i, j;
limb *outlimbs = &out[0][0];
- memset(outlimbs, 0, 3 * sizeof(felem));
+
+ memset(out, 0, sizeof(*out) * 3);
for (i = 0; i < size; i++) {
const limb *inlimbs = &pre_comp[i][0][0];
u8 sign, digit;
/* set nq to the point at infinity */
- memset(nq, 0, 3 * sizeof(felem));
+ memset(nq, 0, sizeof(nq));
/*
* Loop over all scalars msb-to-lsb, interleaving additions of multiples
}
/* Precomputation for the group generator. */
-typedef struct {
+struct nistp521_pre_comp_st {
felem g_pre_comp[16][3];
- int references;
-} NISTP521_PRE_COMP;
+ CRYPTO_REF_COUNT references;
+ CRYPTO_RWLOCK *lock;
+};
const EC_METHOD *EC_GFp_nistp521_method(void)
{
ec_GFp_nistp521_group_set_curve,
ec_GFp_simple_group_get_curve,
ec_GFp_simple_group_get_degree,
+ ec_group_simple_order_bits,
ec_GFp_simple_group_check_discriminant,
ec_GFp_simple_point_init,
ec_GFp_simple_point_finish,
ec_GFp_nist_field_mul,
ec_GFp_nist_field_sqr,
0 /* field_div */ ,
+ ec_GFp_simple_field_inv,
0 /* field_encode */ ,
0 /* field_decode */ ,
- 0 /* field_set_to_one */
+ 0, /* field_set_to_one */
+ ec_key_simple_priv2oct,
+ ec_key_simple_oct2priv,
+ 0, /* set private */
+ ec_key_simple_generate_key,
+ ec_key_simple_check_key,
+ ec_key_simple_generate_public_key,
+ 0, /* keycopy */
+ 0, /* keyfinish */
+ ecdh_simple_compute_key,
+ ecdsa_simple_sign_setup,
+ ecdsa_simple_sign_sig,
+ ecdsa_simple_verify_sig,
+ 0, /* field_inverse_mod_ord */
+ 0, /* blind_coordinates */
+ 0, /* ladder_pre */
+ 0, /* ladder_step */
+ 0 /* ladder_post */
};
return &ret;
* FUNCTIONS TO MANAGE PRECOMPUTATION
*/
-static NISTP521_PRE_COMP *nistp521_pre_comp_new()
+static NISTP521_PRE_COMP *nistp521_pre_comp_new(void)
{
- NISTP521_PRE_COMP *ret = NULL;
- ret = (NISTP521_PRE_COMP *) OPENSSL_malloc(sizeof(NISTP521_PRE_COMP));
- if (!ret) {
+ NISTP521_PRE_COMP *ret = OPENSSL_zalloc(sizeof(*ret));
+
+ if (ret == NULL) {
ECerr(EC_F_NISTP521_PRE_COMP_NEW, ERR_R_MALLOC_FAILURE);
return ret;
}
- memset(ret->g_pre_comp, 0, sizeof(ret->g_pre_comp));
- ret->references = 1;
- return ret;
-}
-
-static void *nistp521_pre_comp_dup(void *src_)
-{
- NISTP521_PRE_COMP *src = src_;
- /* no need to actually copy, these objects never change! */
- CRYPTO_add(&src->references, 1, CRYPTO_LOCK_EC_PRE_COMP);
+ ret->references = 1;
- return src_;
+ ret->lock = CRYPTO_THREAD_lock_new();
+ if (ret->lock == NULL) {
+ ECerr(EC_F_NISTP521_PRE_COMP_NEW, ERR_R_MALLOC_FAILURE);
+ OPENSSL_free(ret);
+ return NULL;
+ }
+ return ret;
}
-static void nistp521_pre_comp_free(void *pre_)
+NISTP521_PRE_COMP *EC_nistp521_pre_comp_dup(NISTP521_PRE_COMP *p)
{
int i;
- NISTP521_PRE_COMP *pre = pre_;
-
- if (!pre)
- return;
-
- i = CRYPTO_add(&pre->references, -1, CRYPTO_LOCK_EC_PRE_COMP);
- if (i > 0)
- return;
-
- OPENSSL_free(pre);
+ if (p != NULL)
+ CRYPTO_UP_REF(&p->references, &i, p->lock);
+ return p;
}
-static void nistp521_pre_comp_clear_free(void *pre_)
+void EC_nistp521_pre_comp_free(NISTP521_PRE_COMP *p)
{
int i;
- NISTP521_PRE_COMP *pre = pre_;
- if (!pre)
+ if (p == NULL)
return;
- i = CRYPTO_add(&pre->references, -1, CRYPTO_LOCK_EC_PRE_COMP);
+ CRYPTO_DOWN_REF(&p->references, &i, p->lock);
+ REF_PRINT_COUNT("EC_nistp521", x);
if (i > 0)
return;
+ REF_ASSERT_ISNT(i < 0);
- OPENSSL_cleanse(pre, sizeof(*pre));
- OPENSSL_free(pre);
+ CRYPTO_THREAD_lock_free(p->lock);
+ OPENSSL_free(p);
}
/******************************************************************************/
BN_CTX *ctx)
{
int ret = 0;
- BN_CTX *new_ctx = NULL;
BIGNUM *curve_p, *curve_a, *curve_b;
+#ifndef FIPS_MODE
+ BN_CTX *new_ctx = NULL;
if (ctx == NULL)
- if ((ctx = new_ctx = BN_CTX_new()) == NULL)
- return 0;
+ ctx = new_ctx = BN_CTX_new();
+#endif
+ if (ctx == NULL)
+ return 0;
+
BN_CTX_start(ctx);
- if (((curve_p = BN_CTX_get(ctx)) == NULL) ||
- ((curve_a = BN_CTX_get(ctx)) == NULL) ||
- ((curve_b = BN_CTX_get(ctx)) == NULL))
+ curve_p = BN_CTX_get(ctx);
+ curve_a = BN_CTX_get(ctx);
+ curve_b = BN_CTX_get(ctx);
+ if (curve_b == NULL)
goto err;
BN_bin2bn(nistp521_curve_params[0], sizeof(felem_bytearray), curve_p);
BN_bin2bn(nistp521_curve_params[1], sizeof(felem_bytearray), curve_a);
ret = ec_GFp_simple_group_set_curve(group, p, a, b, ctx);
err:
BN_CTX_end(ctx);
- if (new_ctx != NULL)
- BN_CTX_free(new_ctx);
+#ifndef FIPS_MODE
+ BN_CTX_free(new_ctx);
+#endif
return ret;
}
EC_R_POINT_AT_INFINITY);
return 0;
}
- if ((!BN_to_felem(x_in, &point->X)) || (!BN_to_felem(y_in, &point->Y)) ||
- (!BN_to_felem(z1, &point->Z)))
+ if ((!BN_to_felem(x_in, point->X)) || (!BN_to_felem(y_in, point->Y)) ||
+ (!BN_to_felem(z1, point->Z)))
return 0;
felem_inv(z2, z1);
felem_square(tmp, z2);
sizeof(felem),
tmp_felems,
(void (*)(void *))felem_one,
- (int (*)(const void *))
felem_is_zero_int,
(void (*)(void *, const void *))
felem_assign,
int ret = 0;
int j;
int mixed = 0;
- BN_CTX *new_ctx = NULL;
BIGNUM *x, *y, *z, *tmp_scalar;
felem_bytearray g_secret;
felem_bytearray *secrets = NULL;
- felem(*pre_comp)[17][3] = NULL;
+ felem (*pre_comp)[17][3] = NULL;
felem *tmp_felems = NULL;
- felem_bytearray tmp;
- unsigned i, num_bytes;
+ unsigned i;
+ int num_bytes;
int have_pre_comp = 0;
size_t num_points = num;
felem x_in, y_in, z_in, x_out, y_out, z_out;
const EC_POINT *p = NULL;
const BIGNUM *p_scalar = NULL;
- if (ctx == NULL)
- if ((ctx = new_ctx = BN_CTX_new()) == NULL)
- return 0;
BN_CTX_start(ctx);
- if (((x = BN_CTX_get(ctx)) == NULL) ||
- ((y = BN_CTX_get(ctx)) == NULL) ||
- ((z = BN_CTX_get(ctx)) == NULL) ||
- ((tmp_scalar = BN_CTX_get(ctx)) == NULL))
+ x = BN_CTX_get(ctx);
+ y = BN_CTX_get(ctx);
+ z = BN_CTX_get(ctx);
+ tmp_scalar = BN_CTX_get(ctx);
+ if (tmp_scalar == NULL)
goto err;
if (scalar != NULL) {
- pre = EC_EX_DATA_get_data(group->extra_data,
- nistp521_pre_comp_dup,
- nistp521_pre_comp_free,
- nistp521_pre_comp_clear_free);
+ pre = group->pre_comp.nistp521;
if (pre)
/* we have precomputation, try to use it */
g_pre_comp = &pre->g_pre_comp[0];
*/
mixed = 1;
}
- secrets = OPENSSL_malloc(num_points * sizeof(felem_bytearray));
- pre_comp = OPENSSL_malloc(num_points * 17 * 3 * sizeof(felem));
+ secrets = OPENSSL_zalloc(sizeof(*secrets) * num_points);
+ pre_comp = OPENSSL_zalloc(sizeof(*pre_comp) * num_points);
if (mixed)
tmp_felems =
- OPENSSL_malloc((num_points * 17 + 1) * sizeof(felem));
+ OPENSSL_malloc(sizeof(*tmp_felems) * (num_points * 17 + 1));
if ((secrets == NULL) || (pre_comp == NULL)
|| (mixed && (tmp_felems == NULL))) {
ECerr(EC_F_EC_GFP_NISTP521_POINTS_MUL, ERR_R_MALLOC_FAILURE);
* we treat NULL scalars as 0, and NULL points as points at infinity,
* i.e., they contribute nothing to the linear combination
*/
- memset(secrets, 0, num_points * sizeof(felem_bytearray));
- memset(pre_comp, 0, num_points * 17 * 3 * sizeof(felem));
for (i = 0; i < num_points; ++i) {
- if (i == num)
+ if (i == num) {
/*
* we didn't have a valid precomputation, so we pick the
* generator
*/
- {
p = EC_GROUP_get0_generator(group);
p_scalar = scalar;
- } else
+ } else {
/* the i^th point */
- {
p = points[i];
p_scalar = scalars[i];
}
* this is an unusual input, and we don't guarantee
* constant-timeness
*/
- if (!BN_nnmod(tmp_scalar, p_scalar, &group->order, ctx)) {
+ if (!BN_nnmod(tmp_scalar, p_scalar, group->order, ctx)) {
ECerr(EC_F_EC_GFP_NISTP521_POINTS_MUL, ERR_R_BN_LIB);
goto err;
}
- num_bytes = BN_bn2bin(tmp_scalar, tmp);
- } else
- num_bytes = BN_bn2bin(p_scalar, tmp);
- flip_endian(secrets[i], tmp, num_bytes);
+ num_bytes = BN_bn2lebinpad(tmp_scalar,
+ secrets[i], sizeof(secrets[i]));
+ } else {
+ num_bytes = BN_bn2lebinpad(p_scalar,
+ secrets[i], sizeof(secrets[i]));
+ }
+ if (num_bytes < 0) {
+ ECerr(EC_F_EC_GFP_NISTP521_POINTS_MUL, ERR_R_BN_LIB);
+ goto err;
+ }
/* precompute multiples */
- if ((!BN_to_felem(x_out, &p->X)) ||
- (!BN_to_felem(y_out, &p->Y)) ||
- (!BN_to_felem(z_out, &p->Z)))
+ if ((!BN_to_felem(x_out, p->X)) ||
+ (!BN_to_felem(y_out, p->Y)) ||
+ (!BN_to_felem(z_out, p->Z)))
goto err;
memcpy(pre_comp[i][1][0], x_out, sizeof(felem));
memcpy(pre_comp[i][1][1], y_out, sizeof(felem));
* this is an unusual input, and we don't guarantee
* constant-timeness
*/
- if (!BN_nnmod(tmp_scalar, scalar, &group->order, ctx)) {
+ if (!BN_nnmod(tmp_scalar, scalar, group->order, ctx)) {
ECerr(EC_F_EC_GFP_NISTP521_POINTS_MUL, ERR_R_BN_LIB);
goto err;
}
- num_bytes = BN_bn2bin(tmp_scalar, tmp);
- } else
- num_bytes = BN_bn2bin(scalar, tmp);
- flip_endian(g_secret, tmp, num_bytes);
+ num_bytes = BN_bn2lebinpad(tmp_scalar, g_secret, sizeof(g_secret));
+ } else {
+ num_bytes = BN_bn2lebinpad(scalar, g_secret, sizeof(g_secret));
+ }
/* do the multiplication with generator precomputation */
batch_mul(x_out, y_out, z_out,
(const felem_bytearray(*))secrets, num_points,
g_secret,
mixed, (const felem(*)[17][3])pre_comp,
(const felem(*)[3])g_pre_comp);
- } else
+ } else {
/* do the multiplication without generator precomputation */
batch_mul(x_out, y_out, z_out,
(const felem_bytearray(*))secrets, num_points,
NULL, mixed, (const felem(*)[17][3])pre_comp, NULL);
+ }
/* reduce the output to its unique minimal representation */
felem_contract(x_in, x_out);
felem_contract(y_in, y_out);
err:
BN_CTX_end(ctx);
- if (generator != NULL)
- EC_POINT_free(generator);
- if (new_ctx != NULL)
- BN_CTX_free(new_ctx);
- if (secrets != NULL)
- OPENSSL_free(secrets);
- if (pre_comp != NULL)
- OPENSSL_free(pre_comp);
- if (tmp_felems != NULL)
- OPENSSL_free(tmp_felems);
+ EC_POINT_free(generator);
+ OPENSSL_free(secrets);
+ OPENSSL_free(pre_comp);
+ OPENSSL_free(tmp_felems);
return ret;
}
int ret = 0;
NISTP521_PRE_COMP *pre = NULL;
int i, j;
- BN_CTX *new_ctx = NULL;
BIGNUM *x, *y;
EC_POINT *generator = NULL;
felem tmp_felems[16];
+#ifndef FIPS_MODE
+ BN_CTX *new_ctx = NULL;
+#endif
/* throw away old precomputation */
- EC_EX_DATA_free_data(&group->extra_data, nistp521_pre_comp_dup,
- nistp521_pre_comp_free,
- nistp521_pre_comp_clear_free);
+ EC_pre_comp_free(group);
+
+#ifndef FIPS_MODE
if (ctx == NULL)
- if ((ctx = new_ctx = BN_CTX_new()) == NULL)
- return 0;
+ ctx = new_ctx = BN_CTX_new();
+#endif
+ if (ctx == NULL)
+ return 0;
+
BN_CTX_start(ctx);
- if (((x = BN_CTX_get(ctx)) == NULL) || ((y = BN_CTX_get(ctx)) == NULL))
+ x = BN_CTX_get(ctx);
+ y = BN_CTX_get(ctx);
+ if (y == NULL)
goto err;
/* get the generator */
if (group->generator == NULL)
goto err;
BN_bin2bn(nistp521_curve_params[3], sizeof(felem_bytearray), x);
BN_bin2bn(nistp521_curve_params[4], sizeof(felem_bytearray), y);
- if (!EC_POINT_set_affine_coordinates_GFp(group, generator, x, y, ctx))
+ if (!EC_POINT_set_affine_coordinates(group, generator, x, y, ctx))
goto err;
if ((pre = nistp521_pre_comp_new()) == NULL)
goto err;
*/
if (0 == EC_POINT_cmp(group, generator, group->generator, ctx)) {
memcpy(pre->g_pre_comp, gmul, sizeof(pre->g_pre_comp));
- ret = 1;
- goto err;
+ goto done;
}
- if ((!BN_to_felem(pre->g_pre_comp[1][0], &group->generator->X)) ||
- (!BN_to_felem(pre->g_pre_comp[1][1], &group->generator->Y)) ||
- (!BN_to_felem(pre->g_pre_comp[1][2], &group->generator->Z)))
+ if ((!BN_to_felem(pre->g_pre_comp[1][0], group->generator->X)) ||
+ (!BN_to_felem(pre->g_pre_comp[1][1], group->generator->Y)) ||
+ (!BN_to_felem(pre->g_pre_comp[1][2], group->generator->Z)))
goto err;
/* compute 2^130*G, 2^260*G, 2^390*G */
for (i = 1; i <= 4; i <<= 1) {
}
make_points_affine(15, &(pre->g_pre_comp[1]), tmp_felems);
- if (!EC_EX_DATA_set_data(&group->extra_data, pre, nistp521_pre_comp_dup,
- nistp521_pre_comp_free,
- nistp521_pre_comp_clear_free))
- goto err;
+ done:
+ SETPRECOMP(group, nistp521, pre);
ret = 1;
pre = NULL;
err:
BN_CTX_end(ctx);
- if (generator != NULL)
- EC_POINT_free(generator);
- if (new_ctx != NULL)
- BN_CTX_free(new_ctx);
- if (pre)
- nistp521_pre_comp_free(pre);
+ EC_POINT_free(generator);
+#ifndef FIPS_MODE
+ BN_CTX_free(new_ctx);
+#endif
+ EC_nistp521_pre_comp_free(pre);
return ret;
}
int ec_GFp_nistp521_have_precompute_mult(const EC_GROUP *group)
{
- if (EC_EX_DATA_get_data(group->extra_data, nistp521_pre_comp_dup,
- nistp521_pre_comp_free,
- nistp521_pre_comp_clear_free)
- != NULL)
- return 1;
- else
- return 0;
+ return HAVEPRECOMP(group, nistp521);
}
-#else
-static void *dummy = &dummy;
#endif
projects / openssl.git / blobdiff