2018-02-01 21:49:18 +00:00
|
|
|
/*
|
|
|
|
* Copyright 2013-2016 The OpenSSL Project Authors. All Rights Reserved.
|
|
|
|
* Copyright (c) 2012, Intel Corporation. All Rights Reserved.
|
|
|
|
*
|
|
|
|
* Licensed under the OpenSSL license (the "License"). You may not use
|
|
|
|
* this file except in compliance with the License. You can obtain a copy
|
|
|
|
* in the file LICENSE in the source distribution or at
|
|
|
|
* https://www.openssl.org/source/license.html
|
|
|
|
*
|
|
|
|
* Originally written by Shay Gueron (1, 2), and Vlad Krasnov (1)
|
|
|
|
* (1) Intel Corporation, Israel Development Center, Haifa, Israel
|
|
|
|
* (2) University of Haifa, Israel
|
|
|
|
*/
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2014-07-08 23:27:43 +01:00
|
|
|
#include <openssl/base.h>
|
|
|
|
|
2014-07-11 01:07:14 +01:00
|
|
|
#if !defined(OPENSSL_NO_ASM) && defined(OPENSSL_X86_64)
|
2014-07-08 23:27:43 +01:00
|
|
|
|
2014-06-20 20:00:00 +01:00
|
|
|
#include "rsaz_exp.h"
|
|
|
|
|
|
|
|
#include <openssl/mem.h>
|
|
|
|
|
2018-06-04 18:16:16 +01:00
|
|
|
#include "internal.h"
|
2017-04-28 22:47:06 +01:00
|
|
|
#include "../../internal.h"
|
2016-01-18 06:12:57 +00:00
|
|
|
|
|
|
|
|
2018-02-13 20:14:30 +00:00
|
|
|
// See crypto/bn/asm/rsaz-avx2.pl for further details.
|
|
|
|
void rsaz_1024_norm2red_avx2(void *red, const void *norm);
|
|
|
|
void rsaz_1024_mul_avx2(void *ret, const void *a, const void *b, const void *n,
|
|
|
|
BN_ULONG k);
|
|
|
|
void rsaz_1024_sqr_avx2(void *ret, const void *a, const void *n, BN_ULONG k,
|
|
|
|
int cnt);
|
|
|
|
void rsaz_1024_scatter5_avx2(void *tbl, const void *val, int i);
|
|
|
|
void rsaz_1024_gather5_avx2(void *val, const void *tbl, int i);
|
|
|
|
void rsaz_1024_red2norm_avx2(void *norm, const void *red);
|
|
|
|
|
2018-02-14 01:47:28 +00:00
|
|
|
// one is 1 in RSAZ's representation.
|
2018-02-13 20:14:30 +00:00
|
|
|
alignas(64) static const BN_ULONG one[40] = {
|
|
|
|
1, 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, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
|
2018-02-14 01:47:28 +00:00
|
|
|
// two80 is 2^80 in RSAZ's representation. Note RSAZ uses base 2^29, so this is
|
|
|
|
// 2^(29*2 + 22) = 2^80, not 2^(64*2 + 22).
|
2018-02-13 20:14:30 +00:00
|
|
|
alignas(64) static const BN_ULONG two80[40] = {
|
|
|
|
0, 0, 1 << 22, 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, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
|
2014-06-20 20:00:00 +01:00
|
|
|
|
|
|
|
void RSAZ_1024_mod_exp_avx2(BN_ULONG result_norm[16],
|
2018-10-24 23:08:00 +01:00
|
|
|
const BN_ULONG base_norm[16],
|
|
|
|
const BN_ULONG exponent[16],
|
|
|
|
const BN_ULONG m_norm[16], const BN_ULONG RR[16],
|
|
|
|
BN_ULONG k0,
|
|
|
|
BN_ULONG storage_words[MOD_EXP_CTIME_STORAGE_LEN]) {
|
|
|
|
OPENSSL_STATIC_ASSERT(MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH % 64 == 0,
|
|
|
|
"MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH is too small");
|
Replace |alloca| in |BN_mod_exp_mont_consttime|.
|alloca| is dangerous and poorly specified, according to any
description of |alloca|. It's also hard for some analysis tools to
reason about.
The code here assumed |alloca| is a macro, which isn't a valid
assumption. Depending on what which headers are included and what
toolchain is being used, |alloca| may or may not be defined as a macro,
and this might change over time if/when toolchains are updated. Or, we
might be doing static analysis and/or dynamic analysis with a different
configuration w.r.t. the availability of |alloca| than production
builds use.
Regardless, the |alloca| code path only kicked in when the inputs are
840 bits or smaller. Since the multi-prime RSA support was removed, for
interesting RSA key sizes the input will be at least 1024 bits and this
code path won't be triggered since powerbufLen will be larger than 3072
bytes in those cases. ECC inversion via Fermat's Little Theorem has its
own constant-time exponentiation so there are no cases where smaller
inputs need to be fast.
The RSAZ code avoids the |OPENSSL_malloc| for 2048-bit RSA keys.
Increasingly the RSAZ code won't be used though, since it will be
skipped over on Broadwell+ CPUs. Generalize the RSAZ stack allocation
to work for non-RSAZ code paths. In order to ensure this doesn't cause
too much stack usage on platforms where RSAZ wasn't already being used,
only do so on x86-64, which already has this large stack size
requirement due to RSAZ.
This change will make it easier to refactor |BN_mod_exp_mont_consttime|
to do that more safely and in a way that's more compatible with various
analysis tools.
This is also a step towards eliminating the |uintptr_t|-based alignment
hack.
Since this change increases the number of times |OPENSSL_free| is
skipped, I've added an explicit |OPENSSL_cleanse| to ensure the
zeroization is done. This should be done regardless of the other changes
here.
Change-Id: I8a161ce2720a26127e85fff7513f394883e50b2e
Reviewed-on: https://boringssl-review.googlesource.com/28584
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Reviewed-by: David Benjamin <davidben@google.com>
2018-05-17 04:24:20 +01:00
|
|
|
unsigned char *storage = (unsigned char *)storage_words;
|
|
|
|
assert((uintptr_t)storage % 64 == 0);
|
|
|
|
|
2018-02-13 20:14:30 +00:00
|
|
|
unsigned char *a_inv, *m, *result, *table_s = storage + (320 * 3),
|
|
|
|
*R2 = table_s; // borrow
|
|
|
|
if (((((uintptr_t)storage & 4095) + 320) >> 12) != 0) {
|
|
|
|
result = storage;
|
|
|
|
a_inv = storage + 320;
|
|
|
|
m = storage + (320 * 2); // should not cross page
|
|
|
|
} else {
|
|
|
|
m = storage; // should not cross page
|
|
|
|
result = storage + 320;
|
|
|
|
a_inv = storage + (320 * 2);
|
|
|
|
}
|
|
|
|
|
|
|
|
rsaz_1024_norm2red_avx2(m, m_norm);
|
|
|
|
rsaz_1024_norm2red_avx2(a_inv, base_norm);
|
|
|
|
rsaz_1024_norm2red_avx2(R2, RR);
|
|
|
|
|
2018-02-14 01:47:28 +00:00
|
|
|
// Convert |R2| from the usual radix, giving R = 2^1024, to RSAZ's radix,
|
|
|
|
// giving R = 2^(36*29) = 2^1044.
|
2018-02-13 20:14:30 +00:00
|
|
|
rsaz_1024_mul_avx2(R2, R2, R2, m, k0);
|
2018-02-14 01:47:28 +00:00
|
|
|
// R2 = 2^2048 * 2^2048 / 2^1044 = 2^3052
|
2018-02-13 20:14:30 +00:00
|
|
|
rsaz_1024_mul_avx2(R2, R2, two80, m, k0);
|
2018-02-14 01:47:28 +00:00
|
|
|
// R2 = 2^3052 * 2^80 / 2^1044 = 2^2088 = (2^1044)^2
|
2018-02-13 20:14:30 +00:00
|
|
|
|
|
|
|
// table[0] = 1
|
|
|
|
rsaz_1024_mul_avx2(result, R2, one, m, k0);
|
|
|
|
// table[1] = a_inv^1
|
|
|
|
rsaz_1024_mul_avx2(a_inv, a_inv, R2, m, k0);
|
|
|
|
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, a_inv, 1);
|
|
|
|
|
|
|
|
// table[2] = a_inv^2
|
|
|
|
rsaz_1024_sqr_avx2(result, a_inv, m, k0, 1);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 2);
|
2014-06-20 20:00:00 +01:00
|
|
|
#if 0
|
2018-02-14 01:47:28 +00:00
|
|
|
// This is almost 2x smaller and less than 1% slower.
|
2018-02-13 20:14:30 +00:00
|
|
|
for (int index = 3; index < 32; index++) {
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, index);
|
|
|
|
}
|
2014-06-20 20:00:00 +01:00
|
|
|
#else
|
2018-02-13 20:14:30 +00:00
|
|
|
// table[4] = a_inv^4
|
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 4);
|
|
|
|
// table[8] = a_inv^8
|
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 8);
|
|
|
|
// table[16] = a_inv^16
|
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 16);
|
|
|
|
// table[17] = a_inv^17
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 17);
|
|
|
|
|
|
|
|
// table[3]
|
|
|
|
rsaz_1024_gather5_avx2(result, table_s, 2);
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 3);
|
|
|
|
// table[6]
|
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 6);
|
|
|
|
// table[12]
|
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 12);
|
|
|
|
// table[24]
|
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 24);
|
|
|
|
// table[25]
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 25);
|
|
|
|
|
|
|
|
// table[5]
|
|
|
|
rsaz_1024_gather5_avx2(result, table_s, 4);
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 5);
|
|
|
|
// table[10]
|
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 10);
|
|
|
|
// table[20]
|
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 20);
|
|
|
|
// table[21]
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 21);
|
|
|
|
|
|
|
|
// table[7]
|
|
|
|
rsaz_1024_gather5_avx2(result, table_s, 6);
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 7);
|
|
|
|
// table[14]
|
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 14);
|
|
|
|
// table[28]
|
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 28);
|
|
|
|
// table[29]
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 29);
|
|
|
|
|
|
|
|
// table[9]
|
|
|
|
rsaz_1024_gather5_avx2(result, table_s, 8);
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 9);
|
|
|
|
// table[18]
|
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 18);
|
|
|
|
// table[19]
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 19);
|
|
|
|
|
|
|
|
// table[11]
|
|
|
|
rsaz_1024_gather5_avx2(result, table_s, 10);
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 11);
|
|
|
|
// table[22]
|
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 22);
|
|
|
|
// table[23]
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 23);
|
|
|
|
|
|
|
|
// table[13]
|
|
|
|
rsaz_1024_gather5_avx2(result, table_s, 12);
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 13);
|
|
|
|
// table[26]
|
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 26);
|
|
|
|
// table[27]
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 27);
|
|
|
|
|
|
|
|
// table[15]
|
|
|
|
rsaz_1024_gather5_avx2(result, table_s, 14);
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 15);
|
|
|
|
// table[30]
|
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 30);
|
|
|
|
// table[31]
|
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
rsaz_1024_scatter5_avx2(table_s, result, 31);
|
2014-06-20 20:00:00 +01:00
|
|
|
#endif
|
|
|
|
|
2018-02-13 20:14:30 +00:00
|
|
|
const uint8_t *p_str = (const uint8_t *)exponent;
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2018-02-13 20:14:30 +00:00
|
|
|
// load first window
|
|
|
|
int wvalue = p_str[127] >> 3;
|
|
|
|
rsaz_1024_gather5_avx2(result, table_s, wvalue);
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2018-02-13 20:14:30 +00:00
|
|
|
int index = 1014;
|
2018-02-14 01:47:28 +00:00
|
|
|
while (index > -1) { // Loop for the remaining 127 windows.
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2018-02-13 20:14:30 +00:00
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 5);
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2018-02-13 20:14:30 +00:00
|
|
|
uint16_t wvalue_16;
|
|
|
|
memcpy(&wvalue_16, &p_str[index / 8], sizeof(wvalue_16));
|
|
|
|
wvalue = wvalue_16;
|
|
|
|
wvalue = (wvalue >> (index % 8)) & 31;
|
|
|
|
index -= 5;
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2018-02-14 01:47:28 +00:00
|
|
|
rsaz_1024_gather5_avx2(a_inv, table_s, wvalue); // Borrow |a_inv|.
|
2018-02-13 20:14:30 +00:00
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
|
|
|
}
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2018-02-14 01:47:28 +00:00
|
|
|
// Square four times.
|
2018-02-13 20:14:30 +00:00
|
|
|
rsaz_1024_sqr_avx2(result, result, m, k0, 4);
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2018-02-13 20:14:30 +00:00
|
|
|
wvalue = p_str[0] & 15;
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2018-02-14 01:47:28 +00:00
|
|
|
rsaz_1024_gather5_avx2(a_inv, table_s, wvalue); // Borrow |a_inv|.
|
2018-02-13 20:14:30 +00:00
|
|
|
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2018-02-14 01:47:28 +00:00
|
|
|
// Convert from Montgomery.
|
2018-02-13 20:14:30 +00:00
|
|
|
rsaz_1024_mul_avx2(result, result, one, m, k0);
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2018-02-13 20:14:30 +00:00
|
|
|
rsaz_1024_red2norm_avx2(result_norm, result);
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2018-02-13 20:14:30 +00:00
|
|
|
OPENSSL_cleanse(storage, sizeof(storage));
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
2018-02-13 20:14:30 +00:00
|
|
|
#endif // OPENSSL_X86_64
|