4e7a1ff055
The performance measurements seem to be very out-of-date. Also, the idea for optimizing the case of an even modulus is interesting, but it isn't useful because we never use an even modulus. Change-Id: I012eb37638cda3c63db0e390c8c728f65b949e54 Reviewed-on: https://boringssl-review.googlesource.com/7733 Reviewed-by: David Benjamin <davidben@google.com>
1259 lines
35 KiB
C
1259 lines
35 KiB
C
/* 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) 1998-2005 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>
|
|
|
|
#include <assert.h>
|
|
#include <string.h>
|
|
|
|
#include <openssl/cpu.h>
|
|
#include <openssl/err.h>
|
|
#include <openssl/mem.h>
|
|
|
|
#include "internal.h"
|
|
|
|
|
|
#if !defined(OPENSSL_NO_ASM) && defined(OPENSSL_X86_64)
|
|
#define OPENSSL_BN_ASM_MONT5
|
|
#define RSAZ_ENABLED
|
|
|
|
#include "rsaz_exp.h"
|
|
|
|
void bn_mul_mont_gather5(BN_ULONG *rp, const BN_ULONG *ap, const void *table,
|
|
const BN_ULONG *np, const BN_ULONG *n0, int num,
|
|
int power);
|
|
void bn_scatter5(const BN_ULONG *inp, size_t num, void *table, size_t power);
|
|
void bn_gather5(BN_ULONG *out, size_t num, void *table, size_t power);
|
|
void bn_power5(BN_ULONG *rp, const BN_ULONG *ap, const void *table,
|
|
const BN_ULONG *np, const BN_ULONG *n0, int num, int power);
|
|
int bn_from_montgomery(BN_ULONG *rp, const BN_ULONG *ap,
|
|
const BN_ULONG *not_used, const BN_ULONG *np,
|
|
const BN_ULONG *n0, int num);
|
|
#endif
|
|
|
|
int BN_exp(BIGNUM *r, const BIGNUM *a, const BIGNUM *p, BN_CTX *ctx) {
|
|
int i, bits, ret = 0;
|
|
BIGNUM *v, *rr;
|
|
|
|
if ((p->flags & BN_FLG_CONSTTIME) != 0) {
|
|
/* BN_FLG_CONSTTIME only supported by BN_mod_exp_mont() */
|
|
OPENSSL_PUT_ERROR(BN, ERR_R_SHOULD_NOT_HAVE_BEEN_CALLED);
|
|
return 0;
|
|
}
|
|
|
|
BN_CTX_start(ctx);
|
|
if (r == a || r == p) {
|
|
rr = BN_CTX_get(ctx);
|
|
} else {
|
|
rr = r;
|
|
}
|
|
|
|
v = BN_CTX_get(ctx);
|
|
if (rr == NULL || v == NULL) {
|
|
goto err;
|
|
}
|
|
|
|
if (BN_copy(v, a) == NULL) {
|
|
goto err;
|
|
}
|
|
bits = BN_num_bits(p);
|
|
|
|
if (BN_is_odd(p)) {
|
|
if (BN_copy(rr, a) == NULL) {
|
|
goto err;
|
|
}
|
|
} else {
|
|
if (!BN_one(rr)) {
|
|
goto err;
|
|
}
|
|
}
|
|
|
|
for (i = 1; i < bits; i++) {
|
|
if (!BN_sqr(v, v, ctx)) {
|
|
goto err;
|
|
}
|
|
if (BN_is_bit_set(p, i)) {
|
|
if (!BN_mul(rr, rr, v, ctx)) {
|
|
goto err;
|
|
}
|
|
}
|
|
}
|
|
|
|
if (r != rr && !BN_copy(r, rr)) {
|
|
goto err;
|
|
}
|
|
ret = 1;
|
|
|
|
err:
|
|
BN_CTX_end(ctx);
|
|
return ret;
|
|
}
|
|
|
|
/* maximum precomputation table size for *variable* sliding windows */
|
|
#define TABLE_SIZE 32
|
|
|
|
typedef struct bn_recp_ctx_st {
|
|
BIGNUM N; /* the divisor */
|
|
BIGNUM Nr; /* the reciprocal */
|
|
int num_bits;
|
|
int shift;
|
|
int flags;
|
|
} BN_RECP_CTX;
|
|
|
|
static void BN_RECP_CTX_init(BN_RECP_CTX *recp) {
|
|
BN_init(&recp->N);
|
|
BN_init(&recp->Nr);
|
|
recp->num_bits = 0;
|
|
recp->shift = 0;
|
|
recp->flags = 0;
|
|
}
|
|
|
|
static void BN_RECP_CTX_free(BN_RECP_CTX *recp) {
|
|
if (recp == NULL) {
|
|
return;
|
|
}
|
|
|
|
BN_free(&recp->N);
|
|
BN_free(&recp->Nr);
|
|
}
|
|
|
|
static int BN_RECP_CTX_set(BN_RECP_CTX *recp, const BIGNUM *d, BN_CTX *ctx) {
|
|
if (!BN_copy(&(recp->N), d)) {
|
|
return 0;
|
|
}
|
|
BN_zero(&recp->Nr);
|
|
recp->num_bits = BN_num_bits(d);
|
|
recp->shift = 0;
|
|
|
|
return 1;
|
|
}
|
|
|
|
/* len is the expected size of the result We actually calculate with an extra
|
|
* word of precision, so we can do faster division if the remainder is not
|
|
* required.
|
|
* r := 2^len / m */
|
|
static int BN_reciprocal(BIGNUM *r, const BIGNUM *m, int len, BN_CTX *ctx) {
|
|
int ret = -1;
|
|
BIGNUM *t;
|
|
|
|
BN_CTX_start(ctx);
|
|
t = BN_CTX_get(ctx);
|
|
if (t == NULL) {
|
|
goto err;
|
|
}
|
|
|
|
if (!BN_set_bit(t, len)) {
|
|
goto err;
|
|
}
|
|
|
|
if (!BN_div(r, NULL, t, m, ctx)) {
|
|
goto err;
|
|
}
|
|
|
|
ret = len;
|
|
|
|
err:
|
|
BN_CTX_end(ctx);
|
|
return ret;
|
|
}
|
|
|
|
static int BN_div_recp(BIGNUM *dv, BIGNUM *rem, const BIGNUM *m,
|
|
BN_RECP_CTX *recp, BN_CTX *ctx) {
|
|
int i, j, ret = 0;
|
|
BIGNUM *a, *b, *d, *r;
|
|
|
|
BN_CTX_start(ctx);
|
|
a = BN_CTX_get(ctx);
|
|
b = BN_CTX_get(ctx);
|
|
if (dv != NULL) {
|
|
d = dv;
|
|
} else {
|
|
d = BN_CTX_get(ctx);
|
|
}
|
|
|
|
if (rem != NULL) {
|
|
r = rem;
|
|
} else {
|
|
r = BN_CTX_get(ctx);
|
|
}
|
|
|
|
if (a == NULL || b == NULL || d == NULL || r == NULL) {
|
|
goto err;
|
|
}
|
|
|
|
if (BN_ucmp(m, &recp->N) < 0) {
|
|
BN_zero(d);
|
|
if (!BN_copy(r, m)) {
|
|
goto err;
|
|
}
|
|
BN_CTX_end(ctx);
|
|
return 1;
|
|
}
|
|
|
|
/* We want the remainder
|
|
* Given input of ABCDEF / ab
|
|
* we need multiply ABCDEF by 3 digests of the reciprocal of ab */
|
|
|
|
/* i := max(BN_num_bits(m), 2*BN_num_bits(N)) */
|
|
i = BN_num_bits(m);
|
|
j = recp->num_bits << 1;
|
|
if (j > i) {
|
|
i = j;
|
|
}
|
|
|
|
/* Nr := round(2^i / N) */
|
|
if (i != recp->shift) {
|
|
recp->shift =
|
|
BN_reciprocal(&(recp->Nr), &(recp->N), i,
|
|
ctx); /* BN_reciprocal returns i, or -1 for an error */
|
|
}
|
|
|
|
if (recp->shift == -1) {
|
|
goto err;
|
|
}
|
|
|
|
/* d := |round(round(m / 2^BN_num_bits(N)) * recp->Nr / 2^(i -
|
|
* BN_num_bits(N)))|
|
|
* = |round(round(m / 2^BN_num_bits(N)) * round(2^i / N) / 2^(i -
|
|
* BN_num_bits(N)))|
|
|
* <= |(m / 2^BN_num_bits(N)) * (2^i / N) * (2^BN_num_bits(N) / 2^i)|
|
|
* = |m/N| */
|
|
if (!BN_rshift(a, m, recp->num_bits)) {
|
|
goto err;
|
|
}
|
|
if (!BN_mul(b, a, &(recp->Nr), ctx)) {
|
|
goto err;
|
|
}
|
|
if (!BN_rshift(d, b, i - recp->num_bits)) {
|
|
goto err;
|
|
}
|
|
d->neg = 0;
|
|
|
|
if (!BN_mul(b, &(recp->N), d, ctx)) {
|
|
goto err;
|
|
}
|
|
if (!BN_usub(r, m, b)) {
|
|
goto err;
|
|
}
|
|
r->neg = 0;
|
|
|
|
j = 0;
|
|
while (BN_ucmp(r, &(recp->N)) >= 0) {
|
|
if (j++ > 2) {
|
|
OPENSSL_PUT_ERROR(BN, BN_R_BAD_RECIPROCAL);
|
|
goto err;
|
|
}
|
|
if (!BN_usub(r, r, &(recp->N))) {
|
|
goto err;
|
|
}
|
|
if (!BN_add_word(d, 1)) {
|
|
goto err;
|
|
}
|
|
}
|
|
|
|
r->neg = BN_is_zero(r) ? 0 : m->neg;
|
|
d->neg = m->neg ^ recp->N.neg;
|
|
ret = 1;
|
|
|
|
err:
|
|
BN_CTX_end(ctx);
|
|
return ret;
|
|
}
|
|
|
|
static int BN_mod_mul_reciprocal(BIGNUM *r, const BIGNUM *x, const BIGNUM *y,
|
|
BN_RECP_CTX *recp, BN_CTX *ctx) {
|
|
int ret = 0;
|
|
BIGNUM *a;
|
|
const BIGNUM *ca;
|
|
|
|
BN_CTX_start(ctx);
|
|
a = BN_CTX_get(ctx);
|
|
if (a == NULL) {
|
|
goto err;
|
|
}
|
|
|
|
if (y != NULL) {
|
|
if (x == y) {
|
|
if (!BN_sqr(a, x, ctx)) {
|
|
goto err;
|
|
}
|
|
} else {
|
|
if (!BN_mul(a, x, y, ctx)) {
|
|
goto err;
|
|
}
|
|
}
|
|
ca = a;
|
|
} else {
|
|
ca = x; /* Just do the mod */
|
|
}
|
|
|
|
ret = BN_div_recp(NULL, r, ca, recp, ctx);
|
|
|
|
err:
|
|
BN_CTX_end(ctx);
|
|
return ret;
|
|
}
|
|
|
|
/* BN_window_bits_for_exponent_size -- macro for sliding window mod_exp
|
|
* functions
|
|
*
|
|
* For window size 'w' (w >= 2) and a random 'b' bits exponent, the number of
|
|
* multiplications is a constant plus on average
|
|
*
|
|
* 2^(w-1) + (b-w)/(w+1);
|
|
*
|
|
* here 2^(w-1) is for precomputing the table (we actually need entries only
|
|
* for windows that have the lowest bit set), and (b-w)/(w+1) is an
|
|
* approximation for the expected number of w-bit windows, not counting the
|
|
* first one.
|
|
*
|
|
* Thus we should use
|
|
*
|
|
* w >= 6 if b > 671
|
|
* w = 5 if 671 > b > 239
|
|
* w = 4 if 239 > b > 79
|
|
* w = 3 if 79 > b > 23
|
|
* w <= 2 if 23 > b
|
|
*
|
|
* (with draws in between). Very small exponents are often selected
|
|
* with low Hamming weight, so we use w = 1 for b <= 23. */
|
|
#define BN_window_bits_for_exponent_size(b) \
|
|
((b) > 671 ? 6 : \
|
|
(b) > 239 ? 5 : \
|
|
(b) > 79 ? 4 : \
|
|
(b) > 23 ? 3 : 1)
|
|
|
|
static int mod_exp_recp(BIGNUM *r, const BIGNUM *a, const BIGNUM *p,
|
|
const BIGNUM *m, BN_CTX *ctx) {
|
|
int i, j, bits, ret = 0, wstart, window;
|
|
int start = 1;
|
|
BIGNUM *aa;
|
|
/* Table of variables obtained from 'ctx' */
|
|
BIGNUM *val[TABLE_SIZE];
|
|
BN_RECP_CTX recp;
|
|
|
|
if (BN_get_flags(p, BN_FLG_CONSTTIME) != 0) {
|
|
/* BN_FLG_CONSTTIME only supported by BN_mod_exp_mont() */
|
|
OPENSSL_PUT_ERROR(BN, ERR_R_SHOULD_NOT_HAVE_BEEN_CALLED);
|
|
return 0;
|
|
}
|
|
|
|
bits = BN_num_bits(p);
|
|
|
|
if (bits == 0) {
|
|
/* x**0 mod 1 is still zero. */
|
|
if (BN_is_one(m)) {
|
|
BN_zero(r);
|
|
return 1;
|
|
}
|
|
return BN_one(r);
|
|
}
|
|
|
|
BN_CTX_start(ctx);
|
|
aa = BN_CTX_get(ctx);
|
|
val[0] = BN_CTX_get(ctx);
|
|
if (!aa || !val[0]) {
|
|
goto err;
|
|
}
|
|
|
|
BN_RECP_CTX_init(&recp);
|
|
if (m->neg) {
|
|
/* ignore sign of 'm' */
|
|
if (!BN_copy(aa, m)) {
|
|
goto err;
|
|
}
|
|
aa->neg = 0;
|
|
if (BN_RECP_CTX_set(&recp, aa, ctx) <= 0) {
|
|
goto err;
|
|
}
|
|
} else {
|
|
if (BN_RECP_CTX_set(&recp, m, ctx) <= 0) {
|
|
goto err;
|
|
}
|
|
}
|
|
|
|
if (!BN_nnmod(val[0], a, m, ctx)) {
|
|
goto err; /* 1 */
|
|
}
|
|
if (BN_is_zero(val[0])) {
|
|
BN_zero(r);
|
|
ret = 1;
|
|
goto err;
|
|
}
|
|
|
|
window = BN_window_bits_for_exponent_size(bits);
|
|
if (window > 1) {
|
|
if (!BN_mod_mul_reciprocal(aa, val[0], val[0], &recp, ctx)) {
|
|
goto err; /* 2 */
|
|
}
|
|
j = 1 << (window - 1);
|
|
for (i = 1; i < j; i++) {
|
|
if (((val[i] = BN_CTX_get(ctx)) == NULL) ||
|
|
!BN_mod_mul_reciprocal(val[i], val[i - 1], aa, &recp, ctx)) {
|
|
goto err;
|
|
}
|
|
}
|
|
}
|
|
|
|
start = 1; /* This is used to avoid multiplication etc
|
|
* when there is only the value '1' in the
|
|
* buffer. */
|
|
wstart = bits - 1; /* The top bit of the window */
|
|
|
|
if (!BN_one(r)) {
|
|
goto err;
|
|
}
|
|
|
|
for (;;) {
|
|
int wvalue; /* The 'value' of the window */
|
|
int wend; /* The bottom bit of the window */
|
|
|
|
if (BN_is_bit_set(p, wstart) == 0) {
|
|
if (!start) {
|
|
if (!BN_mod_mul_reciprocal(r, r, r, &recp, ctx)) {
|
|
goto err;
|
|
}
|
|
}
|
|
if (wstart == 0) {
|
|
break;
|
|
}
|
|
wstart--;
|
|
continue;
|
|
}
|
|
|
|
/* We now have wstart on a 'set' bit, we now need to work out
|
|
* how bit a window to do. To do this we need to scan
|
|
* forward until the last set bit before the end of the
|
|
* window */
|
|
wvalue = 1;
|
|
wend = 0;
|
|
for (i = 1; i < window; i++) {
|
|
if (wstart - i < 0) {
|
|
break;
|
|
}
|
|
if (BN_is_bit_set(p, wstart - i)) {
|
|
wvalue <<= (i - wend);
|
|
wvalue |= 1;
|
|
wend = i;
|
|
}
|
|
}
|
|
|
|
/* wend is the size of the current window */
|
|
j = wend + 1;
|
|
/* add the 'bytes above' */
|
|
if (!start) {
|
|
for (i = 0; i < j; i++) {
|
|
if (!BN_mod_mul_reciprocal(r, r, r, &recp, ctx)) {
|
|
goto err;
|
|
}
|
|
}
|
|
}
|
|
|
|
/* wvalue will be an odd number < 2^window */
|
|
if (!BN_mod_mul_reciprocal(r, r, val[wvalue >> 1], &recp, ctx)) {
|
|
goto err;
|
|
}
|
|
|
|
/* move the 'window' down further */
|
|
wstart -= wend + 1;
|
|
start = 0;
|
|
if (wstart < 0) {
|
|
break;
|
|
}
|
|
}
|
|
ret = 1;
|
|
|
|
err:
|
|
BN_CTX_end(ctx);
|
|
BN_RECP_CTX_free(&recp);
|
|
return ret;
|
|
}
|
|
|
|
int BN_mod_exp(BIGNUM *r, const BIGNUM *a, const BIGNUM *p, const BIGNUM *m,
|
|
BN_CTX *ctx) {
|
|
if (BN_is_odd(m)) {
|
|
return BN_mod_exp_mont(r, a, p, m, ctx, NULL);
|
|
}
|
|
|
|
return mod_exp_recp(r, a, p, m, ctx);
|
|
}
|
|
|
|
int BN_mod_exp_mont(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
|
|
const BIGNUM *m, BN_CTX *ctx, const BN_MONT_CTX *mont) {
|
|
int i, j, bits, ret = 0, wstart, window;
|
|
int start = 1;
|
|
BIGNUM *d, *r;
|
|
const BIGNUM *aa;
|
|
/* Table of variables obtained from 'ctx' */
|
|
BIGNUM *val[TABLE_SIZE];
|
|
BN_MONT_CTX *new_mont = NULL;
|
|
|
|
if (BN_get_flags(p, BN_FLG_CONSTTIME) != 0) {
|
|
return BN_mod_exp_mont_consttime(rr, a, p, m, ctx, mont);
|
|
}
|
|
|
|
if (!BN_is_odd(m)) {
|
|
OPENSSL_PUT_ERROR(BN, BN_R_CALLED_WITH_EVEN_MODULUS);
|
|
return 0;
|
|
}
|
|
bits = BN_num_bits(p);
|
|
if (bits == 0) {
|
|
/* x**0 mod 1 is still zero. */
|
|
if (BN_is_one(m)) {
|
|
BN_zero(rr);
|
|
return 1;
|
|
}
|
|
return BN_one(rr);
|
|
}
|
|
|
|
BN_CTX_start(ctx);
|
|
d = BN_CTX_get(ctx);
|
|
r = BN_CTX_get(ctx);
|
|
val[0] = BN_CTX_get(ctx);
|
|
if (!d || !r || !val[0]) {
|
|
goto err;
|
|
}
|
|
|
|
/* Allocate a montgomery context if it was not supplied by the caller. */
|
|
if (mont == NULL) {
|
|
new_mont = BN_MONT_CTX_new();
|
|
if (new_mont == NULL || !BN_MONT_CTX_set(new_mont, m, ctx)) {
|
|
goto err;
|
|
}
|
|
mont = new_mont;
|
|
}
|
|
|
|
if (a->neg || BN_ucmp(a, m) >= 0) {
|
|
if (!BN_nnmod(val[0], a, m, ctx)) {
|
|
goto err;
|
|
}
|
|
aa = val[0];
|
|
} else {
|
|
aa = a;
|
|
}
|
|
|
|
if (BN_is_zero(aa)) {
|
|
BN_zero(rr);
|
|
ret = 1;
|
|
goto err;
|
|
}
|
|
if (!BN_to_montgomery(val[0], aa, mont, ctx)) {
|
|
goto err; /* 1 */
|
|
}
|
|
|
|
window = BN_window_bits_for_exponent_size(bits);
|
|
if (window > 1) {
|
|
if (!BN_mod_mul_montgomery(d, val[0], val[0], mont, ctx)) {
|
|
goto err; /* 2 */
|
|
}
|
|
j = 1 << (window - 1);
|
|
for (i = 1; i < j; i++) {
|
|
if (((val[i] = BN_CTX_get(ctx)) == NULL) ||
|
|
!BN_mod_mul_montgomery(val[i], val[i - 1], d, mont, ctx)) {
|
|
goto err;
|
|
}
|
|
}
|
|
}
|
|
|
|
start = 1; /* This is used to avoid multiplication etc
|
|
* when there is only the value '1' in the
|
|
* buffer. */
|
|
wstart = bits - 1; /* The top bit of the window */
|
|
|
|
j = m->top; /* borrow j */
|
|
if (m->d[j - 1] & (((BN_ULONG)1) << (BN_BITS2 - 1))) {
|
|
if (bn_wexpand(r, j) == NULL) {
|
|
goto err;
|
|
}
|
|
/* 2^(top*BN_BITS2) - m */
|
|
r->d[0] = (0 - m->d[0]) & BN_MASK2;
|
|
for (i = 1; i < j; i++) {
|
|
r->d[i] = (~m->d[i]) & BN_MASK2;
|
|
}
|
|
r->top = j;
|
|
/* Upper words will be zero if the corresponding words of 'm'
|
|
* were 0xfff[...], so decrement r->top accordingly. */
|
|
bn_correct_top(r);
|
|
} else if (!BN_to_montgomery(r, BN_value_one(), mont, ctx)) {
|
|
goto err;
|
|
}
|
|
|
|
for (;;) {
|
|
int wvalue; /* The 'value' of the window */
|
|
int wend; /* The bottom bit of the window */
|
|
|
|
if (BN_is_bit_set(p, wstart) == 0) {
|
|
if (!start && !BN_mod_mul_montgomery(r, r, r, mont, ctx)) {
|
|
goto err;
|
|
}
|
|
if (wstart == 0) {
|
|
break;
|
|
}
|
|
wstart--;
|
|
continue;
|
|
}
|
|
|
|
/* We now have wstart on a 'set' bit, we now need to work out how bit a
|
|
* window to do. To do this we need to scan forward until the last set bit
|
|
* before the end of the window */
|
|
wvalue = 1;
|
|
wend = 0;
|
|
for (i = 1; i < window; i++) {
|
|
if (wstart - i < 0) {
|
|
break;
|
|
}
|
|
if (BN_is_bit_set(p, wstart - i)) {
|
|
wvalue <<= (i - wend);
|
|
wvalue |= 1;
|
|
wend = i;
|
|
}
|
|
}
|
|
|
|
/* wend is the size of the current window */
|
|
j = wend + 1;
|
|
/* add the 'bytes above' */
|
|
if (!start) {
|
|
for (i = 0; i < j; i++) {
|
|
if (!BN_mod_mul_montgomery(r, r, r, mont, ctx)) {
|
|
goto err;
|
|
}
|
|
}
|
|
}
|
|
|
|
/* wvalue will be an odd number < 2^window */
|
|
if (!BN_mod_mul_montgomery(r, r, val[wvalue >> 1], mont, ctx)) {
|
|
goto err;
|
|
}
|
|
|
|
/* move the 'window' down further */
|
|
wstart -= wend + 1;
|
|
start = 0;
|
|
if (wstart < 0) {
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (!BN_from_montgomery(rr, r, mont, ctx)) {
|
|
goto err;
|
|
}
|
|
ret = 1;
|
|
|
|
err:
|
|
BN_MONT_CTX_free(new_mont);
|
|
BN_CTX_end(ctx);
|
|
return ret;
|
|
}
|
|
|
|
/* BN_mod_exp_mont_consttime() stores the precomputed powers in a specific
|
|
* layout so that accessing any of these table values shows the same access
|
|
* pattern as far as cache lines are concerned. The following functions are
|
|
* used to transfer a BIGNUM from/to that table. */
|
|
static int copy_to_prebuf(const BIGNUM *b, int top, unsigned char *buf, int idx,
|
|
int window) {
|
|
int i, j;
|
|
const int width = 1 << window;
|
|
BN_ULONG *table = (BN_ULONG *) buf;
|
|
|
|
if (top > b->top) {
|
|
top = b->top; /* this works because 'buf' is explicitly zeroed */
|
|
}
|
|
|
|
for (i = 0, j = idx; i < top; i++, j += width) {
|
|
table[j] = b->d[i];
|
|
}
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int copy_from_prebuf(BIGNUM *b, int top, unsigned char *buf, int idx,
|
|
int window) {
|
|
int i, j;
|
|
const int width = 1 << window;
|
|
volatile BN_ULONG *table = (volatile BN_ULONG *)buf;
|
|
|
|
if (bn_wexpand(b, top) == NULL) {
|
|
return 0;
|
|
}
|
|
|
|
if (window <= 3) {
|
|
for (i = 0; i < top; i++, table += width) {
|
|
BN_ULONG acc = 0;
|
|
|
|
for (j = 0; j < width; j++) {
|
|
acc |= table[j] & ((BN_ULONG)0 - (constant_time_eq_int(j, idx) & 1));
|
|
}
|
|
|
|
b->d[i] = acc;
|
|
}
|
|
} else {
|
|
int xstride = 1 << (window - 2);
|
|
BN_ULONG y0, y1, y2, y3;
|
|
|
|
i = idx >> (window - 2); /* equivalent of idx / xstride */
|
|
idx &= xstride - 1; /* equivalent of idx % xstride */
|
|
|
|
y0 = (BN_ULONG)0 - (constant_time_eq_int(i, 0) & 1);
|
|
y1 = (BN_ULONG)0 - (constant_time_eq_int(i, 1) & 1);
|
|
y2 = (BN_ULONG)0 - (constant_time_eq_int(i, 2) & 1);
|
|
y3 = (BN_ULONG)0 - (constant_time_eq_int(i, 3) & 1);
|
|
|
|
for (i = 0; i < top; i++, table += width) {
|
|
BN_ULONG acc = 0;
|
|
|
|
for (j = 0; j < xstride; j++) {
|
|
acc |= ((table[j + 0 * xstride] & y0) | (table[j + 1 * xstride] & y1) |
|
|
(table[j + 2 * xstride] & y2) | (table[j + 3 * xstride] & y3)) &
|
|
((BN_ULONG)0 - (constant_time_eq_int(j, idx) & 1));
|
|
}
|
|
|
|
b->d[i] = acc;
|
|
}
|
|
}
|
|
|
|
b->top = top;
|
|
bn_correct_top(b);
|
|
return 1;
|
|
}
|
|
|
|
/* BN_mod_exp_mont_conttime is based on the assumption that the L1 data cache
|
|
* line width of the target processor is at least the following value. */
|
|
#define MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH (64)
|
|
#define MOD_EXP_CTIME_MIN_CACHE_LINE_MASK \
|
|
(MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH - 1)
|
|
|
|
/* Window sizes optimized for fixed window size modular exponentiation
|
|
* algorithm (BN_mod_exp_mont_consttime).
|
|
*
|
|
* To achieve the security goals of BN_mode_exp_mont_consttime, the maximum
|
|
* size of the window must not exceed
|
|
* log_2(MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH).
|
|
*
|
|
* Window size thresholds are defined for cache line sizes of 32 and 64, cache
|
|
* line sizes where log_2(32)=5 and log_2(64)=6 respectively. A window size of
|
|
* 7 should only be used on processors that have a 128 byte or greater cache
|
|
* line size. */
|
|
#if MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH == 64
|
|
|
|
#define BN_window_bits_for_ctime_exponent_size(b) \
|
|
((b) > 937 ? 6 : (b) > 306 ? 5 : (b) > 89 ? 4 : (b) > 22 ? 3 : 1)
|
|
#define BN_MAX_WINDOW_BITS_FOR_CTIME_EXPONENT_SIZE (6)
|
|
|
|
#elif MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH == 32
|
|
|
|
#define BN_window_bits_for_ctime_exponent_size(b) \
|
|
((b) > 306 ? 5 : (b) > 89 ? 4 : (b) > 22 ? 3 : 1)
|
|
#define BN_MAX_WINDOW_BITS_FOR_CTIME_EXPONENT_SIZE (5)
|
|
|
|
#endif
|
|
|
|
/* Given a pointer value, compute the next address that is a cache line
|
|
* multiple. */
|
|
#define MOD_EXP_CTIME_ALIGN(x_) \
|
|
((unsigned char *)(x_) + \
|
|
(MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH - \
|
|
(((size_t)(x_)) & (MOD_EXP_CTIME_MIN_CACHE_LINE_MASK))))
|
|
|
|
/* This variant of BN_mod_exp_mont() uses fixed windows and the special
|
|
* precomputation memory layout to limit data-dependency to a minimum
|
|
* to protect secret exponents (cf. the hyper-threading timing attacks
|
|
* pointed out by Colin Percival,
|
|
* http://www.daemonology.net/hyperthreading-considered-harmful/)
|
|
*/
|
|
int BN_mod_exp_mont_consttime(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
|
|
const BIGNUM *m, BN_CTX *ctx,
|
|
const BN_MONT_CTX *mont) {
|
|
int i, bits, ret = 0, window, wvalue;
|
|
int top;
|
|
BN_MONT_CTX *new_mont = NULL;
|
|
|
|
int numPowers;
|
|
unsigned char *powerbufFree = NULL;
|
|
int powerbufLen = 0;
|
|
unsigned char *powerbuf = NULL;
|
|
BIGNUM tmp, am;
|
|
|
|
if (!BN_is_odd(m)) {
|
|
OPENSSL_PUT_ERROR(BN, BN_R_CALLED_WITH_EVEN_MODULUS);
|
|
return 0;
|
|
}
|
|
|
|
top = m->top;
|
|
|
|
bits = BN_num_bits(p);
|
|
if (bits == 0) {
|
|
/* x**0 mod 1 is still zero. */
|
|
if (BN_is_one(m)) {
|
|
BN_zero(rr);
|
|
return 1;
|
|
}
|
|
return BN_one(rr);
|
|
}
|
|
|
|
/* Allocate a montgomery context if it was not supplied by the caller. */
|
|
if (mont == NULL) {
|
|
new_mont = BN_MONT_CTX_new();
|
|
if (new_mont == NULL || !BN_MONT_CTX_set(new_mont, m, ctx)) {
|
|
goto err;
|
|
}
|
|
mont = new_mont;
|
|
}
|
|
|
|
#ifdef RSAZ_ENABLED
|
|
/* If the size of the operands allow it, perform the optimized
|
|
* RSAZ exponentiation. For further information see
|
|
* crypto/bn/rsaz_exp.c and accompanying assembly modules. */
|
|
if ((16 == a->top) && (16 == p->top) && (BN_num_bits(m) == 1024) &&
|
|
rsaz_avx2_eligible()) {
|
|
if (NULL == bn_wexpand(rr, 16)) {
|
|
goto err;
|
|
}
|
|
RSAZ_1024_mod_exp_avx2(rr->d, a->d, p->d, m->d, mont->RR.d, mont->n0[0]);
|
|
rr->top = 16;
|
|
rr->neg = 0;
|
|
bn_correct_top(rr);
|
|
ret = 1;
|
|
goto err;
|
|
} else if ((8 == a->top) && (8 == p->top) && (BN_num_bits(m) == 512)) {
|
|
if (NULL == bn_wexpand(rr, 8)) {
|
|
goto err;
|
|
}
|
|
RSAZ_512_mod_exp(rr->d, a->d, p->d, m->d, mont->n0[0], mont->RR.d);
|
|
rr->top = 8;
|
|
rr->neg = 0;
|
|
bn_correct_top(rr);
|
|
ret = 1;
|
|
goto err;
|
|
}
|
|
#endif
|
|
|
|
/* Get the window size to use with size of p. */
|
|
window = BN_window_bits_for_ctime_exponent_size(bits);
|
|
#if defined(OPENSSL_BN_ASM_MONT5)
|
|
if (window >= 5) {
|
|
window = 5; /* ~5% improvement for RSA2048 sign, and even for RSA4096 */
|
|
/* reserve space for mont->N.d[] copy */
|
|
powerbufLen += top * sizeof(mont->N.d[0]);
|
|
}
|
|
#endif
|
|
|
|
/* Allocate a buffer large enough to hold all of the pre-computed
|
|
* powers of am, am itself and tmp.
|
|
*/
|
|
numPowers = 1 << window;
|
|
powerbufLen +=
|
|
sizeof(m->d[0]) *
|
|
(top * numPowers + ((2 * top) > numPowers ? (2 * top) : numPowers));
|
|
#ifdef alloca
|
|
if (powerbufLen < 3072) {
|
|
powerbufFree = alloca(powerbufLen + MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH);
|
|
} else
|
|
#endif
|
|
{
|
|
if ((powerbufFree = OPENSSL_malloc(
|
|
powerbufLen + MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH)) == NULL) {
|
|
goto err;
|
|
}
|
|
}
|
|
|
|
powerbuf = MOD_EXP_CTIME_ALIGN(powerbufFree);
|
|
memset(powerbuf, 0, powerbufLen);
|
|
|
|
#ifdef alloca
|
|
if (powerbufLen < 3072) {
|
|
powerbufFree = NULL;
|
|
}
|
|
#endif
|
|
|
|
/* lay down tmp and am right after powers table */
|
|
tmp.d = (BN_ULONG *)(powerbuf + sizeof(m->d[0]) * top * numPowers);
|
|
am.d = tmp.d + top;
|
|
tmp.top = am.top = 0;
|
|
tmp.dmax = am.dmax = top;
|
|
tmp.neg = am.neg = 0;
|
|
tmp.flags = am.flags = BN_FLG_STATIC_DATA;
|
|
|
|
/* prepare a^0 in Montgomery domain */
|
|
/* by Shay Gueron's suggestion */
|
|
if (m->d[top - 1] & (((BN_ULONG)1) << (BN_BITS2 - 1))) {
|
|
/* 2^(top*BN_BITS2) - m */
|
|
tmp.d[0] = (0 - m->d[0]) & BN_MASK2;
|
|
for (i = 1; i < top; i++) {
|
|
tmp.d[i] = (~m->d[i]) & BN_MASK2;
|
|
}
|
|
tmp.top = top;
|
|
} else if (!BN_to_montgomery(&tmp, BN_value_one(), mont, ctx)) {
|
|
goto err;
|
|
}
|
|
|
|
/* prepare a^1 in Montgomery domain */
|
|
if (a->neg || BN_ucmp(a, m) >= 0) {
|
|
if (!BN_mod(&am, a, m, ctx) ||
|
|
!BN_to_montgomery(&am, &am, mont, ctx)) {
|
|
goto err;
|
|
}
|
|
} else if (!BN_to_montgomery(&am, a, mont, ctx)) {
|
|
goto err;
|
|
}
|
|
|
|
#if defined(OPENSSL_BN_ASM_MONT5)
|
|
/* This optimization uses ideas from http://eprint.iacr.org/2011/239,
|
|
* specifically optimization of cache-timing attack countermeasures
|
|
* and pre-computation optimization. */
|
|
|
|
/* Dedicated window==4 case improves 512-bit RSA sign by ~15%, but as
|
|
* 512-bit RSA is hardly relevant, we omit it to spare size... */
|
|
if (window == 5 && top > 1) {
|
|
const BN_ULONG *n0 = mont->n0;
|
|
BN_ULONG *np;
|
|
|
|
/* BN_to_montgomery can contaminate words above .top
|
|
* [in BN_DEBUG[_DEBUG] build]... */
|
|
for (i = am.top; i < top; i++) {
|
|
am.d[i] = 0;
|
|
}
|
|
for (i = tmp.top; i < top; i++) {
|
|
tmp.d[i] = 0;
|
|
}
|
|
|
|
/* copy mont->N.d[] to improve cache locality */
|
|
for (np = am.d + top, i = 0; i < top; i++) {
|
|
np[i] = mont->N.d[i];
|
|
}
|
|
|
|
bn_scatter5(tmp.d, top, powerbuf, 0);
|
|
bn_scatter5(am.d, am.top, powerbuf, 1);
|
|
bn_mul_mont(tmp.d, am.d, am.d, np, n0, top);
|
|
bn_scatter5(tmp.d, top, powerbuf, 2);
|
|
|
|
/* same as above, but uses squaring for 1/2 of operations */
|
|
for (i = 4; i < 32; i *= 2) {
|
|
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
|
|
bn_scatter5(tmp.d, top, powerbuf, i);
|
|
}
|
|
for (i = 3; i < 8; i += 2) {
|
|
int j;
|
|
bn_mul_mont_gather5(tmp.d, am.d, powerbuf, np, n0, top, i - 1);
|
|
bn_scatter5(tmp.d, top, powerbuf, i);
|
|
for (j = 2 * i; j < 32; j *= 2) {
|
|
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
|
|
bn_scatter5(tmp.d, top, powerbuf, j);
|
|
}
|
|
}
|
|
for (; i < 16; i += 2) {
|
|
bn_mul_mont_gather5(tmp.d, am.d, powerbuf, np, n0, top, i - 1);
|
|
bn_scatter5(tmp.d, top, powerbuf, i);
|
|
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
|
|
bn_scatter5(tmp.d, top, powerbuf, 2 * i);
|
|
}
|
|
for (; i < 32; i += 2) {
|
|
bn_mul_mont_gather5(tmp.d, am.d, powerbuf, np, n0, top, i - 1);
|
|
bn_scatter5(tmp.d, top, powerbuf, i);
|
|
}
|
|
|
|
bits--;
|
|
for (wvalue = 0, i = bits % 5; i >= 0; i--, bits--) {
|
|
wvalue = (wvalue << 1) + BN_is_bit_set(p, bits);
|
|
}
|
|
bn_gather5(tmp.d, top, powerbuf, wvalue);
|
|
|
|
/* At this point |bits| is 4 mod 5 and at least -1. (|bits| is the first bit
|
|
* that has not been read yet.) */
|
|
assert(bits >= -1 && (bits == -1 || bits % 5 == 4));
|
|
|
|
/* Scan the exponent one window at a time starting from the most
|
|
* significant bits.
|
|
*/
|
|
if (top & 7) {
|
|
while (bits >= 0) {
|
|
for (wvalue = 0, i = 0; i < 5; i++, bits--) {
|
|
wvalue = (wvalue << 1) + BN_is_bit_set(p, bits);
|
|
}
|
|
|
|
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
|
|
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
|
|
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
|
|
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
|
|
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
|
|
bn_mul_mont_gather5(tmp.d, tmp.d, powerbuf, np, n0, top, wvalue);
|
|
}
|
|
} else {
|
|
const uint8_t *p_bytes = (const uint8_t *)p->d;
|
|
int max_bits = p->top * BN_BITS2;
|
|
assert(bits < max_bits);
|
|
/* |p = 0| has been handled as a special case, so |max_bits| is at least
|
|
* one word. */
|
|
assert(max_bits >= 64);
|
|
|
|
/* If the first bit to be read lands in the last byte, unroll the first
|
|
* iteration to avoid reading past the bounds of |p->d|. (After the first
|
|
* iteration, we are guaranteed to be past the last byte.) Note |bits|
|
|
* here is the top bit, inclusive. */
|
|
if (bits - 4 >= max_bits - 8) {
|
|
/* Read five bits from |bits-4| through |bits|, inclusive. */
|
|
wvalue = p_bytes[p->top * BN_BYTES - 1];
|
|
wvalue >>= (bits - 4) & 7;
|
|
wvalue &= 0x1f;
|
|
bits -= 5;
|
|
bn_power5(tmp.d, tmp.d, powerbuf, np, n0, top, wvalue);
|
|
}
|
|
while (bits >= 0) {
|
|
/* Read five bits from |bits-4| through |bits|, inclusive. */
|
|
int first_bit = bits - 4;
|
|
wvalue = *(const uint16_t *) (p_bytes + (first_bit >> 3));
|
|
wvalue >>= first_bit & 7;
|
|
wvalue &= 0x1f;
|
|
bits -= 5;
|
|
bn_power5(tmp.d, tmp.d, powerbuf, np, n0, top, wvalue);
|
|
}
|
|
}
|
|
|
|
ret = bn_from_montgomery(tmp.d, tmp.d, NULL, np, n0, top);
|
|
tmp.top = top;
|
|
bn_correct_top(&tmp);
|
|
if (ret) {
|
|
if (!BN_copy(rr, &tmp)) {
|
|
ret = 0;
|
|
}
|
|
goto err; /* non-zero ret means it's not error */
|
|
}
|
|
} else
|
|
#endif
|
|
{
|
|
if (!copy_to_prebuf(&tmp, top, powerbuf, 0, window) ||
|
|
!copy_to_prebuf(&am, top, powerbuf, 1, window)) {
|
|
goto err;
|
|
}
|
|
|
|
/* If the window size is greater than 1, then calculate
|
|
* val[i=2..2^winsize-1]. Powers are computed as a*a^(i-1)
|
|
* (even powers could instead be computed as (a^(i/2))^2
|
|
* to use the slight performance advantage of sqr over mul).
|
|
*/
|
|
if (window > 1) {
|
|
if (!BN_mod_mul_montgomery(&tmp, &am, &am, mont, ctx) ||
|
|
!copy_to_prebuf(&tmp, top, powerbuf, 2, window)) {
|
|
goto err;
|
|
}
|
|
for (i = 3; i < numPowers; i++) {
|
|
/* Calculate a^i = a^(i-1) * a */
|
|
if (!BN_mod_mul_montgomery(&tmp, &am, &tmp, mont, ctx) ||
|
|
!copy_to_prebuf(&tmp, top, powerbuf, i, window)) {
|
|
goto err;
|
|
}
|
|
}
|
|
}
|
|
|
|
bits--;
|
|
for (wvalue = 0, i = bits % window; i >= 0; i--, bits--) {
|
|
wvalue = (wvalue << 1) + BN_is_bit_set(p, bits);
|
|
}
|
|
if (!copy_from_prebuf(&tmp, top, powerbuf, wvalue, window)) {
|
|
goto err;
|
|
}
|
|
|
|
/* Scan the exponent one window at a time starting from the most
|
|
* significant bits.
|
|
*/
|
|
while (bits >= 0) {
|
|
wvalue = 0; /* The 'value' of the window */
|
|
|
|
/* Scan the window, squaring the result as we go */
|
|
for (i = 0; i < window; i++, bits--) {
|
|
if (!BN_mod_mul_montgomery(&tmp, &tmp, &tmp, mont, ctx)) {
|
|
goto err;
|
|
}
|
|
wvalue = (wvalue << 1) + BN_is_bit_set(p, bits);
|
|
}
|
|
|
|
/* Fetch the appropriate pre-computed value from the pre-buf */
|
|
if (!copy_from_prebuf(&am, top, powerbuf, wvalue, window)) {
|
|
goto err;
|
|
}
|
|
|
|
/* Multiply the result into the intermediate result */
|
|
if (!BN_mod_mul_montgomery(&tmp, &tmp, &am, mont, ctx)) {
|
|
goto err;
|
|
}
|
|
}
|
|
}
|
|
|
|
/* Convert the final result from montgomery to standard format */
|
|
if (!BN_from_montgomery(rr, &tmp, mont, ctx)) {
|
|
goto err;
|
|
}
|
|
ret = 1;
|
|
|
|
err:
|
|
BN_MONT_CTX_free(new_mont);
|
|
if (powerbuf != NULL) {
|
|
OPENSSL_cleanse(powerbuf, powerbufLen);
|
|
OPENSSL_free(powerbufFree);
|
|
}
|
|
return (ret);
|
|
}
|
|
|
|
int BN_mod_exp_mont_word(BIGNUM *rr, BN_ULONG a, const BIGNUM *p,
|
|
const BIGNUM *m, BN_CTX *ctx,
|
|
const BN_MONT_CTX *mont) {
|
|
BIGNUM a_bignum;
|
|
BN_init(&a_bignum);
|
|
|
|
int ret = 0;
|
|
|
|
if (!BN_set_word(&a_bignum, a)) {
|
|
OPENSSL_PUT_ERROR(BN, ERR_R_INTERNAL_ERROR);
|
|
goto err;
|
|
}
|
|
|
|
ret = BN_mod_exp_mont(rr, &a_bignum, p, m, ctx, mont);
|
|
|
|
err:
|
|
BN_free(&a_bignum);
|
|
|
|
return ret;
|
|
}
|
|
|
|
#define TABLE_SIZE 32
|
|
|
|
int BN_mod_exp2_mont(BIGNUM *rr, const BIGNUM *a1, const BIGNUM *p1,
|
|
const BIGNUM *a2, const BIGNUM *p2, const BIGNUM *m,
|
|
BN_CTX *ctx, const BN_MONT_CTX *mont) {
|
|
BIGNUM tmp;
|
|
BN_init(&tmp);
|
|
|
|
int ret = 0;
|
|
BN_MONT_CTX *new_mont = NULL;
|
|
|
|
/* Allocate a montgomery context if it was not supplied by the caller. */
|
|
if (mont == NULL) {
|
|
new_mont = BN_MONT_CTX_new();
|
|
if (new_mont == NULL || !BN_MONT_CTX_set(new_mont, m, ctx)) {
|
|
goto err;
|
|
}
|
|
mont = new_mont;
|
|
}
|
|
|
|
/* BN_mod_mul_montgomery removes one Montgomery factor, so passing one
|
|
* Montgomery-encoded and one non-Montgomery-encoded value gives a
|
|
* non-Montgomery-encoded result. */
|
|
if (!BN_mod_exp_mont(rr, a1, p1, m, ctx, mont) ||
|
|
!BN_mod_exp_mont(&tmp, a2, p2, m, ctx, mont) ||
|
|
!BN_to_montgomery(rr, rr, mont, ctx) ||
|
|
!BN_mod_mul_montgomery(rr, rr, &tmp, mont, ctx)) {
|
|
goto err;
|
|
}
|
|
|
|
ret = 1;
|
|
|
|
err:
|
|
BN_MONT_CTX_free(new_mont);
|
|
BN_free(&tmp);
|
|
|
|
return ret;
|
|
}
|