2014-06-20 20:00:00 +01:00
|
|
|
/* 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.]
|
|
|
|
*/
|
|
|
|
/* ====================================================================
|
|
|
|
* Copyright (c) 1998-2006 The OpenSSL Project. All rights reserved.
|
|
|
|
*
|
|
|
|
* 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 above 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 acknowledgment:
|
|
|
|
* "This product includes software developed by the OpenSSL Project
|
|
|
|
* for use in the OpenSSL Toolkit. (http://www.openssl.org/)"
|
|
|
|
*
|
|
|
|
* 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
|
|
|
|
* endorse or promote products derived from this software without
|
|
|
|
* prior written permission. For written permission, please contact
|
|
|
|
* openssl-core@openssl.org.
|
|
|
|
*
|
|
|
|
* 5. Products derived from this software may not be called "OpenSSL"
|
|
|
|
* nor may "OpenSSL" appear in their names without prior written
|
|
|
|
* permission of the OpenSSL Project.
|
|
|
|
*
|
|
|
|
* 6. Redistributions of any form whatsoever must retain the following
|
|
|
|
* acknowledgment:
|
|
|
|
* "This product includes software developed by the OpenSSL Project
|
|
|
|
* for use in the OpenSSL Toolkit (http://www.openssl.org/)"
|
|
|
|
*
|
|
|
|
* THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
|
|
|
|
* EXPRESSED 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 OpenSSL PROJECT OR
|
|
|
|
* ITS 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.
|
|
|
|
* ====================================================================
|
|
|
|
*
|
|
|
|
* This product includes cryptographic software written by Eric Young
|
|
|
|
* (eay@cryptsoft.com). This product includes software written by Tim
|
|
|
|
* Hudson (tjh@cryptsoft.com). */
|
|
|
|
|
|
|
|
#include <openssl/bn.h>
|
|
|
|
|
2016-08-20 00:58:40 +01:00
|
|
|
#include <assert.h>
|
2015-01-31 01:08:37 +00:00
|
|
|
#include <string.h>
|
|
|
|
|
2015-08-11 22:10:55 +01:00
|
|
|
#include <openssl/err.h>
|
2014-06-20 20:00:00 +01:00
|
|
|
#include <openssl/mem.h>
|
|
|
|
#include <openssl/thread.h>
|
|
|
|
|
|
|
|
#include "internal.h"
|
2017-04-28 22:47:06 +01:00
|
|
|
#include "../../internal.h"
|
2014-06-20 20:00:00 +01:00
|
|
|
|
|
|
|
|
2015-11-06 22:20:21 +00:00
|
|
|
#if !defined(OPENSSL_NO_ASM) && \
|
|
|
|
(defined(OPENSSL_X86) || defined(OPENSSL_X86_64) || \
|
|
|
|
defined(OPENSSL_ARM) || defined(OPENSSL_AARCH64))
|
2014-06-20 20:00:00 +01:00
|
|
|
#define OPENSSL_BN_ASM_MONT
|
|
|
|
#endif
|
|
|
|
|
Avoid the error case in |bn_mul_mont|.
On 32-bit x86, |bn_mul_mont| returns 0 when the modulus has less than
four limbs. Instead of calling |bn_mul_mont| and then falling back to
the |BN_mul|+|BN_from_montgomery_word| path for small moduli, just
avoid calling |bn_mul_mont| at all for small moduli.
This allows us to more clearly understand exactly when the fallback
code path, which is a timing side channel, is taken. This change makes
it easier to start minimizing this side channel.
The limit is set at 128 bits, which is four limbs on 32-bit and two
limbs on 64-bit platforms. Do this consistently on all platforms even
though it seems to be needed only for 32-bit x86, to minimize platform
variance: every platform uses the same cut-off in terms of input size.
128 bits is small enough to allow even questionably small curves, like
secp128r1, to use the |bn_mul_mont| path, and is way too small for RSA
and FFDH, so this change shouldn't have any security impact other than
the positive impact of simplifying the control flow.
Change-Id: I9b68ae33dc2c86b54ed4294839c7eca6a1dc11c0
Reviewed-on: https://boringssl-review.googlesource.com/14084
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
2017-03-03 19:07:32 +00:00
|
|
|
static int bn_mod_mul_montgomery_fallback(BIGNUM *r, const BIGNUM *a,
|
|
|
|
const BIGNUM *b,
|
|
|
|
const BN_MONT_CTX *mont, BN_CTX *ctx);
|
|
|
|
|
|
|
|
|
2014-06-20 20:00:00 +01:00
|
|
|
BN_MONT_CTX *BN_MONT_CTX_new(void) {
|
|
|
|
BN_MONT_CTX *ret = OPENSSL_malloc(sizeof(BN_MONT_CTX));
|
|
|
|
|
|
|
|
if (ret == NULL) {
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2016-12-13 06:07:13 +00:00
|
|
|
OPENSSL_memset(ret, 0, sizeof(BN_MONT_CTX));
|
2015-10-26 15:31:51 +00:00
|
|
|
BN_init(&ret->RR);
|
|
|
|
BN_init(&ret->N);
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2015-10-26 15:31:51 +00:00
|
|
|
return ret;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
void BN_MONT_CTX_free(BN_MONT_CTX *mont) {
|
|
|
|
if (mont == NULL) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
BN_free(&mont->RR);
|
|
|
|
BN_free(&mont->N);
|
2015-10-26 15:31:51 +00:00
|
|
|
OPENSSL_free(mont);
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
2015-11-03 15:40:23 +00:00
|
|
|
BN_MONT_CTX *BN_MONT_CTX_copy(BN_MONT_CTX *to, const BN_MONT_CTX *from) {
|
2014-06-20 20:00:00 +01:00
|
|
|
if (to == from) {
|
|
|
|
return to;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!BN_copy(&to->RR, &from->RR) ||
|
2015-11-12 04:36:23 +00:00
|
|
|
!BN_copy(&to->N, &from->N)) {
|
2014-06-20 20:00:00 +01:00
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
to->n0[0] = from->n0[0];
|
|
|
|
to->n0[1] = from->n0[1];
|
|
|
|
return to;
|
|
|
|
}
|
|
|
|
|
2016-07-30 03:19:46 +01:00
|
|
|
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(BN_ULONG) * BN_MONT_CTX_N0_LIMBS ==
|
|
|
|
sizeof(uint64_t), BN_MONT_CTX_set_64_bit_mismatch);
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2016-07-30 03:19:46 +01:00
|
|
|
int BN_MONT_CTX_set(BN_MONT_CTX *mont, const BIGNUM *mod, BN_CTX *ctx) {
|
2015-08-11 22:10:55 +01:00
|
|
|
if (BN_is_zero(mod)) {
|
|
|
|
OPENSSL_PUT_ERROR(BN, BN_R_DIV_BY_ZERO);
|
|
|
|
return 0;
|
|
|
|
}
|
2016-07-30 03:19:46 +01:00
|
|
|
if (!BN_is_odd(mod)) {
|
|
|
|
OPENSSL_PUT_ERROR(BN, BN_R_CALLED_WITH_EVEN_MODULUS);
|
|
|
|
return 0;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
2016-07-30 03:19:46 +01:00
|
|
|
if (BN_is_negative(mod)) {
|
|
|
|
OPENSSL_PUT_ERROR(BN, BN_R_NEGATIVE_NUMBER);
|
|
|
|
return 0;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
2017-08-18 19:06:02 +01:00
|
|
|
// Save the modulus.
|
2016-07-30 03:19:46 +01:00
|
|
|
if (!BN_copy(&mont->N, mod)) {
|
|
|
|
OPENSSL_PUT_ERROR(BN, ERR_R_INTERNAL_ERROR);
|
|
|
|
return 0;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
2017-08-18 19:06:02 +01:00
|
|
|
// Find n0 such that n0 * N == -1 (mod r).
|
|
|
|
//
|
|
|
|
// Only certain BN_BITS2<=32 platforms actually make use of n0[1]. For the
|
|
|
|
// others, we could use a shorter R value and use faster |BN_ULONG|-based
|
|
|
|
// math instead of |uint64_t|-based math, which would be double-precision.
|
|
|
|
// However, currently only the assembler files know which is which.
|
2016-07-30 03:19:46 +01:00
|
|
|
uint64_t n0 = bn_mont_n0(mod);
|
|
|
|
mont->n0[0] = (BN_ULONG)n0;
|
|
|
|
#if BN_MONT_CTX_N0_LIMBS == 2
|
|
|
|
mont->n0[1] = (BN_ULONG)(n0 >> BN_BITS2);
|
2014-06-20 20:00:00 +01:00
|
|
|
#else
|
|
|
|
mont->n0[1] = 0;
|
|
|
|
#endif
|
|
|
|
|
2017-08-18 19:06:02 +01:00
|
|
|
// Save RR = R**2 (mod N). R is the smallest power of 2**BN_BITS such that R
|
|
|
|
// > mod. Even though the assembly on some 32-bit platforms works with 64-bit
|
|
|
|
// values, using |BN_BITS2| here, rather than |BN_MONT_CTX_N0_LIMBS *
|
|
|
|
// BN_BITS2|, is correct because R**2 will still be a multiple of the latter
|
|
|
|
// as |BN_MONT_CTX_N0_LIMBS| is either one or two.
|
|
|
|
//
|
|
|
|
// XXX: This is not constant time with respect to |mont->N|, but it should
|
|
|
|
// be.
|
2016-07-30 03:19:46 +01:00
|
|
|
unsigned lgBigR = (BN_num_bits(mod) + (BN_BITS2 - 1)) / BN_BITS2 * BN_BITS2;
|
2016-08-20 00:58:40 +01:00
|
|
|
if (!bn_mod_exp_base_2_vartime(&mont->RR, lgBigR * 2, &mont->N)) {
|
2016-07-30 03:19:46 +01:00
|
|
|
return 0;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
2016-07-30 03:19:46 +01:00
|
|
|
return 1;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
2016-03-25 20:11:04 +00:00
|
|
|
int BN_MONT_CTX_set_locked(BN_MONT_CTX **pmont, CRYPTO_MUTEX *lock,
|
|
|
|
const BIGNUM *mod, BN_CTX *bn_ctx) {
|
2015-04-13 19:04:14 +01:00
|
|
|
CRYPTO_MUTEX_lock_read(lock);
|
|
|
|
BN_MONT_CTX *ctx = *pmont;
|
2016-05-24 16:28:36 +01:00
|
|
|
CRYPTO_MUTEX_unlock_read(lock);
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2015-04-13 19:04:14 +01:00
|
|
|
if (ctx) {
|
2016-03-25 20:11:04 +00:00
|
|
|
return 1;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
2015-04-13 19:04:14 +01:00
|
|
|
CRYPTO_MUTEX_lock_write(lock);
|
|
|
|
ctx = *pmont;
|
|
|
|
if (ctx) {
|
|
|
|
goto out;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2015-04-13 19:04:14 +01:00
|
|
|
ctx = BN_MONT_CTX_new();
|
|
|
|
if (ctx == NULL) {
|
|
|
|
goto out;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
2015-04-13 19:04:14 +01:00
|
|
|
if (!BN_MONT_CTX_set(ctx, mod, bn_ctx)) {
|
|
|
|
BN_MONT_CTX_free(ctx);
|
|
|
|
ctx = NULL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
*pmont = ctx;
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2015-04-13 19:04:14 +01:00
|
|
|
out:
|
2016-05-24 16:28:36 +01:00
|
|
|
CRYPTO_MUTEX_unlock_write(lock);
|
2016-03-25 20:11:04 +00:00
|
|
|
return ctx != NULL;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
int BN_to_montgomery(BIGNUM *ret, const BIGNUM *a, const BN_MONT_CTX *mont,
|
|
|
|
BN_CTX *ctx) {
|
|
|
|
return BN_mod_mul_montgomery(ret, a, &mont->RR, mont, ctx);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int BN_from_montgomery_word(BIGNUM *ret, BIGNUM *r,
|
|
|
|
const BN_MONT_CTX *mont) {
|
|
|
|
BN_ULONG *ap, *np, *rp, n0, v, carry;
|
|
|
|
int nl, max, i;
|
|
|
|
|
2016-03-28 18:54:48 +01:00
|
|
|
const BIGNUM *n = &mont->N;
|
2014-06-20 20:00:00 +01:00
|
|
|
nl = n->top;
|
|
|
|
if (nl == 0) {
|
|
|
|
ret->top = 0;
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2017-08-18 19:06:02 +01:00
|
|
|
max = (2 * nl); // carry is stored separately
|
2017-04-21 16:26:30 +01:00
|
|
|
if (!bn_wexpand(r, max)) {
|
2014-06-20 20:00:00 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
r->neg ^= n->neg;
|
|
|
|
np = n->d;
|
|
|
|
rp = r->d;
|
|
|
|
|
2017-08-18 19:06:02 +01:00
|
|
|
// clear the top words of T
|
2014-06-20 20:00:00 +01:00
|
|
|
if (max > r->top) {
|
2016-12-13 06:07:13 +00:00
|
|
|
OPENSSL_memset(&rp[r->top], 0, (max - r->top) * sizeof(BN_ULONG));
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
r->top = max;
|
|
|
|
n0 = mont->n0[0];
|
|
|
|
|
|
|
|
for (carry = 0, i = 0; i < nl; i++, rp++) {
|
|
|
|
v = bn_mul_add_words(rp, np, nl, (rp[0] * n0) & BN_MASK2);
|
|
|
|
v = (v + carry + rp[nl]) & BN_MASK2;
|
|
|
|
carry |= (v != rp[nl]);
|
|
|
|
carry &= (v <= rp[nl]);
|
|
|
|
rp[nl] = v;
|
|
|
|
}
|
|
|
|
|
2017-04-21 16:26:30 +01:00
|
|
|
if (!bn_wexpand(ret, nl)) {
|
2014-06-20 20:00:00 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
ret->top = nl;
|
|
|
|
ret->neg = r->neg;
|
|
|
|
|
|
|
|
rp = ret->d;
|
|
|
|
ap = &(r->d[nl]);
|
|
|
|
|
|
|
|
{
|
|
|
|
BN_ULONG *nrp;
|
2016-03-28 18:54:48 +01:00
|
|
|
uintptr_t m;
|
2014-06-20 20:00:00 +01:00
|
|
|
|
|
|
|
v = bn_sub_words(rp, ap, np, nl) - carry;
|
2017-08-18 19:06:02 +01:00
|
|
|
// if subtraction result is real, then trick unconditional memcpy below to
|
|
|
|
// perform in-place "refresh" instead of actual copy.
|
2016-03-28 18:54:48 +01:00
|
|
|
m = (0u - (uintptr_t)v);
|
|
|
|
nrp = (BN_ULONG *)(((uintptr_t)rp & ~m) | ((uintptr_t)ap & m));
|
2014-06-20 20:00:00 +01:00
|
|
|
|
|
|
|
for (i = 0, nl -= 4; i < nl; i += 4) {
|
|
|
|
BN_ULONG t1, t2, t3, t4;
|
|
|
|
|
|
|
|
t1 = nrp[i + 0];
|
|
|
|
t2 = nrp[i + 1];
|
|
|
|
t3 = nrp[i + 2];
|
|
|
|
ap[i + 0] = 0;
|
|
|
|
t4 = nrp[i + 3];
|
|
|
|
ap[i + 1] = 0;
|
|
|
|
rp[i + 0] = t1;
|
|
|
|
ap[i + 2] = 0;
|
|
|
|
rp[i + 1] = t2;
|
|
|
|
ap[i + 3] = 0;
|
|
|
|
rp[i + 2] = t3;
|
|
|
|
rp[i + 3] = t4;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (nl += 4; i < nl; i++) {
|
|
|
|
rp[i] = nrp[i], ap[i] = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
bn_correct_top(r);
|
|
|
|
bn_correct_top(ret);
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2016-03-26 04:12:13 +00:00
|
|
|
int BN_from_montgomery(BIGNUM *r, const BIGNUM *a, const BN_MONT_CTX *mont,
|
2014-06-20 20:00:00 +01:00
|
|
|
BN_CTX *ctx) {
|
2016-03-26 04:12:13 +00:00
|
|
|
int ret = 0;
|
2014-06-20 20:00:00 +01:00
|
|
|
BIGNUM *t;
|
|
|
|
|
|
|
|
BN_CTX_start(ctx);
|
|
|
|
t = BN_CTX_get(ctx);
|
2016-03-26 04:12:13 +00:00
|
|
|
if (t == NULL ||
|
|
|
|
!BN_copy(t, a)) {
|
|
|
|
goto err;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
2016-03-26 04:12:13 +00:00
|
|
|
ret = BN_from_montgomery_word(r, t, mont);
|
|
|
|
|
|
|
|
err:
|
2014-06-20 20:00:00 +01:00
|
|
|
BN_CTX_end(ctx);
|
|
|
|
|
2016-03-26 04:12:13 +00:00
|
|
|
return ret;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
int BN_mod_mul_montgomery(BIGNUM *r, const BIGNUM *a, const BIGNUM *b,
|
|
|
|
const BN_MONT_CTX *mont, BN_CTX *ctx) {
|
Avoid the error case in |bn_mul_mont|.
On 32-bit x86, |bn_mul_mont| returns 0 when the modulus has less than
four limbs. Instead of calling |bn_mul_mont| and then falling back to
the |BN_mul|+|BN_from_montgomery_word| path for small moduli, just
avoid calling |bn_mul_mont| at all for small moduli.
This allows us to more clearly understand exactly when the fallback
code path, which is a timing side channel, is taken. This change makes
it easier to start minimizing this side channel.
The limit is set at 128 bits, which is four limbs on 32-bit and two
limbs on 64-bit platforms. Do this consistently on all platforms even
though it seems to be needed only for 32-bit x86, to minimize platform
variance: every platform uses the same cut-off in terms of input size.
128 bits is small enough to allow even questionably small curves, like
secp128r1, to use the |bn_mul_mont| path, and is way too small for RSA
and FFDH, so this change shouldn't have any security impact other than
the positive impact of simplifying the control flow.
Change-Id: I9b68ae33dc2c86b54ed4294839c7eca6a1dc11c0
Reviewed-on: https://boringssl-review.googlesource.com/14084
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
2017-03-03 19:07:32 +00:00
|
|
|
#if !defined(OPENSSL_BN_ASM_MONT)
|
|
|
|
return bn_mod_mul_montgomery_fallback(r, a, b, mont, ctx);
|
|
|
|
#else
|
2014-06-20 20:00:00 +01:00
|
|
|
int num = mont->N.top;
|
|
|
|
|
2017-08-18 19:06:02 +01:00
|
|
|
// |bn_mul_mont| requires at least 128 bits of limbs, at least for x86.
|
Avoid the error case in |bn_mul_mont|.
On 32-bit x86, |bn_mul_mont| returns 0 when the modulus has less than
four limbs. Instead of calling |bn_mul_mont| and then falling back to
the |BN_mul|+|BN_from_montgomery_word| path for small moduli, just
avoid calling |bn_mul_mont| at all for small moduli.
This allows us to more clearly understand exactly when the fallback
code path, which is a timing side channel, is taken. This change makes
it easier to start minimizing this side channel.
The limit is set at 128 bits, which is four limbs on 32-bit and two
limbs on 64-bit platforms. Do this consistently on all platforms even
though it seems to be needed only for 32-bit x86, to minimize platform
variance: every platform uses the same cut-off in terms of input size.
128 bits is small enough to allow even questionably small curves, like
secp128r1, to use the |bn_mul_mont| path, and is way too small for RSA
and FFDH, so this change shouldn't have any security impact other than
the positive impact of simplifying the control flow.
Change-Id: I9b68ae33dc2c86b54ed4294839c7eca6a1dc11c0
Reviewed-on: https://boringssl-review.googlesource.com/14084
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
2017-03-03 19:07:32 +00:00
|
|
|
if (num < (128 / BN_BITS2) ||
|
|
|
|
a->top != num ||
|
|
|
|
b->top != num) {
|
|
|
|
return bn_mod_mul_montgomery_fallback(r, a, b, mont, ctx);
|
|
|
|
}
|
|
|
|
|
2017-04-21 16:26:30 +01:00
|
|
|
if (!bn_wexpand(r, num)) {
|
Avoid the error case in |bn_mul_mont|.
On 32-bit x86, |bn_mul_mont| returns 0 when the modulus has less than
four limbs. Instead of calling |bn_mul_mont| and then falling back to
the |BN_mul|+|BN_from_montgomery_word| path for small moduli, just
avoid calling |bn_mul_mont| at all for small moduli.
This allows us to more clearly understand exactly when the fallback
code path, which is a timing side channel, is taken. This change makes
it easier to start minimizing this side channel.
The limit is set at 128 bits, which is four limbs on 32-bit and two
limbs on 64-bit platforms. Do this consistently on all platforms even
though it seems to be needed only for 32-bit x86, to minimize platform
variance: every platform uses the same cut-off in terms of input size.
128 bits is small enough to allow even questionably small curves, like
secp128r1, to use the |bn_mul_mont| path, and is way too small for RSA
and FFDH, so this change shouldn't have any security impact other than
the positive impact of simplifying the control flow.
Change-Id: I9b68ae33dc2c86b54ed4294839c7eca6a1dc11c0
Reviewed-on: https://boringssl-review.googlesource.com/14084
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
2017-03-03 19:07:32 +00:00
|
|
|
return 0;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
Avoid the error case in |bn_mul_mont|.
On 32-bit x86, |bn_mul_mont| returns 0 when the modulus has less than
four limbs. Instead of calling |bn_mul_mont| and then falling back to
the |BN_mul|+|BN_from_montgomery_word| path for small moduli, just
avoid calling |bn_mul_mont| at all for small moduli.
This allows us to more clearly understand exactly when the fallback
code path, which is a timing side channel, is taken. This change makes
it easier to start minimizing this side channel.
The limit is set at 128 bits, which is four limbs on 32-bit and two
limbs on 64-bit platforms. Do this consistently on all platforms even
though it seems to be needed only for 32-bit x86, to minimize platform
variance: every platform uses the same cut-off in terms of input size.
128 bits is small enough to allow even questionably small curves, like
secp128r1, to use the |bn_mul_mont| path, and is way too small for RSA
and FFDH, so this change shouldn't have any security impact other than
the positive impact of simplifying the control flow.
Change-Id: I9b68ae33dc2c86b54ed4294839c7eca6a1dc11c0
Reviewed-on: https://boringssl-review.googlesource.com/14084
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
2017-03-03 19:07:32 +00:00
|
|
|
if (!bn_mul_mont(r->d, a->d, b->d, mont->N.d, mont->n0, num)) {
|
2017-08-18 19:06:02 +01:00
|
|
|
// The check above ensures this won't happen.
|
Avoid the error case in |bn_mul_mont|.
On 32-bit x86, |bn_mul_mont| returns 0 when the modulus has less than
four limbs. Instead of calling |bn_mul_mont| and then falling back to
the |BN_mul|+|BN_from_montgomery_word| path for small moduli, just
avoid calling |bn_mul_mont| at all for small moduli.
This allows us to more clearly understand exactly when the fallback
code path, which is a timing side channel, is taken. This change makes
it easier to start minimizing this side channel.
The limit is set at 128 bits, which is four limbs on 32-bit and two
limbs on 64-bit platforms. Do this consistently on all platforms even
though it seems to be needed only for 32-bit x86, to minimize platform
variance: every platform uses the same cut-off in terms of input size.
128 bits is small enough to allow even questionably small curves, like
secp128r1, to use the |bn_mul_mont| path, and is way too small for RSA
and FFDH, so this change shouldn't have any security impact other than
the positive impact of simplifying the control flow.
Change-Id: I9b68ae33dc2c86b54ed4294839c7eca6a1dc11c0
Reviewed-on: https://boringssl-review.googlesource.com/14084
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
2017-03-03 19:07:32 +00:00
|
|
|
assert(0);
|
|
|
|
OPENSSL_PUT_ERROR(BN, ERR_R_INTERNAL_ERROR);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
r->neg = a->neg ^ b->neg;
|
|
|
|
r->top = num;
|
|
|
|
bn_correct_top(r);
|
|
|
|
|
|
|
|
return 1;
|
2014-06-20 20:00:00 +01:00
|
|
|
#endif
|
Avoid the error case in |bn_mul_mont|.
On 32-bit x86, |bn_mul_mont| returns 0 when the modulus has less than
four limbs. Instead of calling |bn_mul_mont| and then falling back to
the |BN_mul|+|BN_from_montgomery_word| path for small moduli, just
avoid calling |bn_mul_mont| at all for small moduli.
This allows us to more clearly understand exactly when the fallback
code path, which is a timing side channel, is taken. This change makes
it easier to start minimizing this side channel.
The limit is set at 128 bits, which is four limbs on 32-bit and two
limbs on 64-bit platforms. Do this consistently on all platforms even
though it seems to be needed only for 32-bit x86, to minimize platform
variance: every platform uses the same cut-off in terms of input size.
128 bits is small enough to allow even questionably small curves, like
secp128r1, to use the |bn_mul_mont| path, and is way too small for RSA
and FFDH, so this change shouldn't have any security impact other than
the positive impact of simplifying the control flow.
Change-Id: I9b68ae33dc2c86b54ed4294839c7eca6a1dc11c0
Reviewed-on: https://boringssl-review.googlesource.com/14084
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
2017-03-03 19:07:32 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static int bn_mod_mul_montgomery_fallback(BIGNUM *r, const BIGNUM *a,
|
|
|
|
const BIGNUM *b,
|
|
|
|
const BN_MONT_CTX *mont,
|
|
|
|
BN_CTX *ctx) {
|
|
|
|
int ret = 0;
|
2014-06-20 20:00:00 +01:00
|
|
|
|
|
|
|
BN_CTX_start(ctx);
|
Avoid the error case in |bn_mul_mont|.
On 32-bit x86, |bn_mul_mont| returns 0 when the modulus has less than
four limbs. Instead of calling |bn_mul_mont| and then falling back to
the |BN_mul|+|BN_from_montgomery_word| path for small moduli, just
avoid calling |bn_mul_mont| at all for small moduli.
This allows us to more clearly understand exactly when the fallback
code path, which is a timing side channel, is taken. This change makes
it easier to start minimizing this side channel.
The limit is set at 128 bits, which is four limbs on 32-bit and two
limbs on 64-bit platforms. Do this consistently on all platforms even
though it seems to be needed only for 32-bit x86, to minimize platform
variance: every platform uses the same cut-off in terms of input size.
128 bits is small enough to allow even questionably small curves, like
secp128r1, to use the |bn_mul_mont| path, and is way too small for RSA
and FFDH, so this change shouldn't have any security impact other than
the positive impact of simplifying the control flow.
Change-Id: I9b68ae33dc2c86b54ed4294839c7eca6a1dc11c0
Reviewed-on: https://boringssl-review.googlesource.com/14084
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
2017-03-03 19:07:32 +00:00
|
|
|
BIGNUM *tmp = BN_CTX_get(ctx);
|
2014-06-20 20:00:00 +01:00
|
|
|
if (tmp == NULL) {
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (a == b) {
|
|
|
|
if (!BN_sqr(tmp, a, ctx)) {
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
if (!BN_mul(tmp, a, b, ctx)) {
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2017-08-18 19:06:02 +01:00
|
|
|
// reduce from aRR to aR
|
2014-06-20 20:00:00 +01:00
|
|
|
if (!BN_from_montgomery_word(r, tmp, mont)) {
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = 1;
|
|
|
|
|
|
|
|
err:
|
|
|
|
BN_CTX_end(ctx);
|
|
|
|
return ret;
|
|
|
|
}
|