Add support for SIKE/p503 post-quantum KEM
Based on Microsoft's implementation available on github:
Source: https://github.com/Microsoft/PQCrypto-SIDH
Commit: 77044b76181eb61c744ac8eb7ddc7a8fe72f6919
Following changes has been applied
* In intel assembly, use MOV instead of MOVQ:
Intel instruction reference in the Intel Software Developer's Manual
volume 2A, the MOVQ has 4 forms. None of them mentions moving
literal to GPR, hence "movq $rax, 0x0" is wrong. Instead, on 64bit
system, MOV can be used.
* Some variables were wrongly zero-initialized (as per C99 spec)
* Move constant values to .RODATA segment, as keeping them in .TEXT
segment is not compatible with XOM.
* Fixes issue in arm64 code related to the fact that compiler doesn't
reserve enough space for the linker to relocate address of a global
variable when used by 'ldr' instructions. Solution is to use 'adrp'
followed by 'add' instruction. Relocations for 'adrp' and 'add'
instructions is generated by prefixing the label with :pg_hi21:
and :lo12: respectively.
* Enable MULX and ADX. Code from MS doesn't support PIC. MULX can't
reference global variable directly. Instead RIP-relative addressing
can be used. This improves performance around 10%-13% on SkyLake
* Check if CPU supports BMI2 and ADOX instruction at runtime. On AMD64
optimized implementation of montgomery multiplication and reduction
have 2 implementations - faster one takes advantage of BMI2
instruction set introduced in Haswell and ADOX introduced in
Broadwell. Thanks to OPENSSL_ia32cap_P it can be decided at runtime
which implementation to choose. As CPU configuration is static by
nature, branch predictor will be correct most of the time and hence
this check very often has no cost.
* Reuse some utilities from boringssl instead of reimplementing them.
This includes things like:
* definition of a limb size (use crypto_word_t instead of digit_t)
* use functions for checking in constant time if value is 0 and/or
less then
* #define's used for conditional compilation
* Use SSE2 for conditional swap on vector registers. Improves
performance a little bit.
* Fix f2elm_t definition. Code imported from MSR defines f2elm_t type as
a array of arrays. This decays to a pointer to an array (when passing
as an argument). In C, one can't assign const pointer to an array with
non-const pointer to an array. Seems it violates 6.7.3/8 from C99
(same for C11). This problem occures in GCC 6, only when -pedantic
flag is specified and it occures always in GCC 4.9 (debian jessie).
* Fix definition of eval_3_isog. Second argument in eval_3_isog mustn't be
const. Similar reason as above.
* Use HMAC-SHA256 instead of cSHAKE-256 to avoid upstreaming cSHAKE
and SHA3 code.
* Add speed and unit tests for SIKE.
Change-Id: I22f0bb1f9edff314a35cd74b48e8c4962568e330
2019-03-06 18:19:25 +00:00
|
|
|
/* Copyright (c) 2018, Google Inc.
|
|
|
|
*
|
|
|
|
* Permission to use, copy, modify, and/or distribute this software for any
|
|
|
|
* purpose with or without fee is hereby granted, provided that the above
|
|
|
|
* copyright notice and this permission notice appear in all copies.
|
|
|
|
*
|
|
|
|
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
|
|
|
|
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
|
|
|
|
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
|
|
|
|
* SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
|
|
|
|
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
|
|
|
|
* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
|
|
|
|
* CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */
|
|
|
|
|
|
|
|
#include <stdint.h>
|
|
|
|
|
|
|
|
#include <gtest/gtest.h>
|
|
|
|
|
|
|
|
#include "../../crypto/test/abi_test.h"
|
|
|
|
#include "sike.h"
|
|
|
|
#include "fpx.h"
|
|
|
|
|
|
|
|
TEST(SIKE, RoundTrip) {
|
|
|
|
uint8_t sk[SIKEp503_PRV_BYTESZ] = {0};
|
|
|
|
uint8_t pk[SIKEp503_PUB_BYTESZ] = {0};
|
|
|
|
uint8_t ct[SIKEp503_CT_BYTESZ] = {0};
|
|
|
|
uint8_t ss_enc[SIKEp503_SS_BYTESZ] = {0};
|
|
|
|
uint8_t ss_dec[SIKEp503_SS_BYTESZ] = {0};
|
|
|
|
|
|
|
|
EXPECT_EQ(SIKE_keypair(sk, pk), 1);
|
|
|
|
SIKE_encaps(ss_enc, ct, pk);
|
|
|
|
SIKE_decaps(ss_dec, ct, pk, sk);
|
|
|
|
|
|
|
|
EXPECT_EQ(memcmp(ss_enc, ss_dec, SIKEp503_SS_BYTESZ), 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(SIKE, Decapsulation) {
|
|
|
|
const uint8_t sk[SIKEp503_PRV_BYTESZ] = {
|
|
|
|
0xDB, 0xAF, 0x2C, 0x89, 0xCA, 0x5A, 0xD4, 0x9D, 0x4F, 0x13,
|
|
|
|
0x40, 0xDF, 0x2D, 0xB1, 0x5F, 0x4C, 0x91, 0xA7, 0x1F, 0x0B,
|
|
|
|
0x29, 0x15, 0x01, 0x59, 0xBC, 0x5F, 0x0B, 0x4A, 0x03, 0x27,
|
|
|
|
0x6F, 0x18};
|
|
|
|
|
|
|
|
const uint8_t pk[SIKEp503_PUB_BYTESZ] = {
|
|
|
|
0x07, 0xAA, 0x51, 0x45, 0x3E, 0x1F, 0x53, 0x2A, 0x0A, 0x05,
|
|
|
|
0x46, 0xF6, 0x54, 0x7F, 0x5D, 0x56, 0xD6, 0x76, 0xD3, 0xEA,
|
|
|
|
0x4B, 0x6B, 0x01, 0x9B, 0x11, 0x72, 0x6F, 0x75, 0xEA, 0x34,
|
|
|
|
0x3C, 0x28, 0x2C, 0x36, 0xFD, 0x77, 0xDA, 0xBE, 0xB6, 0x20,
|
|
|
|
0x18, 0xC1, 0x93, 0x98, 0x18, 0x86, 0x30, 0x2F, 0x2E, 0xD2,
|
|
|
|
0x00, 0x61, 0xFF, 0xAE, 0x78, 0xAE, 0xFB, 0x6F, 0x32, 0xAC,
|
|
|
|
0x06, 0xBF, 0x35, 0xF6, 0xF7, 0x5B, 0x98, 0x26, 0x95, 0xC2,
|
|
|
|
0xD8, 0xD6, 0x1C, 0x0E, 0x47, 0xDA, 0x76, 0xCE, 0xB5, 0xF1,
|
|
|
|
0x19, 0xCC, 0x01, 0xE1, 0x17, 0xA9, 0x62, 0xF7, 0x82, 0x6C,
|
|
|
|
0x25, 0x51, 0x25, 0xAE, 0xFE, 0xE3, 0xE2, 0xE1, 0x35, 0xAE,
|
|
|
|
0x2E, 0x8F, 0x38, 0xE0, 0x7C, 0x74, 0x3C, 0x1D, 0x39, 0x91,
|
|
|
|
0x1B, 0xC7, 0x9F, 0x8E, 0x33, 0x4E, 0x84, 0x19, 0xB8, 0xD9,
|
|
|
|
0xC2, 0x71, 0x35, 0x02, 0x47, 0x3E, 0x79, 0xEF, 0x47, 0xE1,
|
|
|
|
0xD8, 0x21, 0x96, 0x1F, 0x11, 0x59, 0x39, 0x34, 0x76, 0xEF,
|
|
|
|
0x3E, 0xB7, 0x4E, 0xFB, 0x7C, 0x55, 0xA1, 0x85, 0xAA, 0xAB,
|
|
|
|
0xAD, 0xF0, 0x09, 0xCB, 0xD1, 0xE3, 0x7C, 0x4F, 0x5D, 0x2D,
|
|
|
|
0xE1, 0x13, 0xF0, 0x71, 0xD9, 0xE5, 0xF6, 0xAF, 0x7F, 0xC1,
|
|
|
|
0x27, 0x95, 0x8D, 0x52, 0xD5, 0x96, 0x42, 0x38, 0x41, 0xF7,
|
|
|
|
0x24, 0x3F, 0x3A, 0xB5, 0x7E, 0x11, 0xE4, 0xF9, 0x33, 0xEE,
|
|
|
|
0x4D, 0xBE, 0x74, 0x48, 0xF9, 0x98, 0x04, 0x01, 0x16, 0xEB,
|
|
|
|
0xA9, 0x0D, 0x61, 0xC6, 0xFD, 0x4C, 0xCF, 0x98, 0x84, 0x4A,
|
|
|
|
0x94, 0xAC, 0x69, 0x2C, 0x02, 0x8B, 0xE3, 0xD1, 0x41, 0x0D,
|
|
|
|
0xF2, 0x2D, 0x46, 0x1F, 0x57, 0x1C, 0x77, 0x86, 0x18, 0xE3,
|
|
|
|
0x63, 0xDE, 0xF3, 0xE3, 0x02, 0x30, 0x54, 0x73, 0xAE, 0xC2,
|
|
|
|
0x32, 0xA2, 0xCE, 0xEB, 0xCF, 0x81, 0x46, 0x54, 0x5C, 0xF4,
|
|
|
|
0x5D, 0x2A, 0x03, 0x5D, 0x9C, 0xAE, 0xE0, 0x60, 0x03, 0x80,
|
|
|
|
0x11, 0x30, 0xA5, 0xAA, 0xD1, 0x75, 0x67, 0xE0, 0x1C, 0x2B,
|
|
|
|
0x6B, 0x5D, 0x83, 0xDE, 0x92, 0x9B, 0x0E, 0xD7, 0x11, 0x0F,
|
|
|
|
0x00, 0xC4, 0x59, 0xE4, 0x81, 0x04, 0x3B, 0xEE, 0x5C, 0x04,
|
|
|
|
0xD1, 0x0E, 0xD0, 0x67, 0xF5, 0xCC, 0xAA, 0x72, 0x73, 0xEA,
|
|
|
|
0xC4, 0x76, 0x99, 0x3B, 0x4C, 0x90, 0x2F, 0xCB, 0xD8, 0x0A,
|
|
|
|
0x5B, 0xEC, 0x0E, 0x0E, 0x1F, 0x59, 0xEA, 0x14, 0x8D, 0x34,
|
|
|
|
0x53, 0x65, 0x4C, 0x1A, 0x59, 0xA8, 0x95, 0x66, 0x60, 0xBB,
|
|
|
|
0xC4, 0xCC, 0x32, 0xA9, 0x8D, 0x2A, 0xAA, 0x14, 0x6F, 0x0F,
|
|
|
|
0x81, 0x4D, 0x32, 0x02, 0xFD, 0x33, 0x58, 0x42, 0xCF, 0xF3,
|
|
|
|
0x67, 0xD0, 0x9F, 0x0B, 0xB1, 0xCC, 0x18, 0xA5, 0xC4, 0x19,
|
|
|
|
0xB6, 0x00, 0xED, 0xFA, 0x32, 0x1A, 0x5F, 0x67, 0xC8, 0xC3,
|
|
|
|
0xEB, 0x0D, 0xB5, 0x9A, 0x36, 0x47, 0x82, 0x00};
|
|
|
|
|
|
|
|
const uint8_t ct_exp[SIKEp503_CT_BYTESZ] = {
|
|
|
|
0xE6, 0xB7, 0xE5, 0x7B, 0xA9, 0x19, 0xD1, 0x2C, 0xB8, 0x5C,
|
|
|
|
0x7B, 0x66, 0x74, 0xB0, 0x71, 0xA1, 0xFF, 0x71, 0x7F, 0x4B,
|
|
|
|
0xB5, 0xA6, 0xAF, 0x48, 0x32, 0x52, 0xD5, 0x82, 0xEE, 0x8A,
|
|
|
|
0xBB, 0x08, 0x1E, 0xF6, 0xAC, 0x91, 0xA2, 0xCB, 0x6B, 0x6A,
|
|
|
|
0x09, 0x2B, 0xD9, 0xC6, 0x27, 0xD6, 0x3A, 0x6B, 0x8D, 0xFC,
|
|
|
|
0xB8, 0x90, 0x8F, 0x72, 0xB3, 0xFA, 0x7D, 0x34, 0x7A, 0xC4,
|
|
|
|
0x7E, 0xE3, 0x30, 0xC5, 0xA0, 0xFE, 0x3D, 0x43, 0x14, 0x4E,
|
|
|
|
0x3A, 0x14, 0x76, 0x3E, 0xFB, 0xDF, 0xE3, 0xA8, 0xE3, 0x5E,
|
|
|
|
0x38, 0xF2, 0xE0, 0x39, 0x67, 0x60, 0xFD, 0xFB, 0xB4, 0x19,
|
|
|
|
0xCD, 0xE1, 0x93, 0xA2, 0x06, 0xCC, 0x65, 0xCD, 0x6E, 0xC8,
|
|
|
|
0xB4, 0x5E, 0x41, 0x4B, 0x6C, 0xA5, 0xF4, 0xE4, 0x9D, 0x52,
|
|
|
|
0x8C, 0x25, 0x60, 0xDD, 0x3D, 0xA9, 0x7F, 0xF2, 0x88, 0xC1,
|
|
|
|
0x0C, 0xEE, 0x97, 0xE0, 0xE7, 0x3B, 0xB7, 0xD3, 0x6F, 0x28,
|
|
|
|
0x79, 0x2F, 0x50, 0xB2, 0x4F, 0x74, 0x3A, 0x0C, 0x88, 0x27,
|
|
|
|
0x98, 0x3A, 0x27, 0xD3, 0x26, 0x83, 0x59, 0x49, 0x81, 0x5B,
|
|
|
|
0x0D, 0xA7, 0x0C, 0x4F, 0xEF, 0xFB, 0x1E, 0xAF, 0xE9, 0xD2,
|
|
|
|
0x1C, 0x10, 0x25, 0xEC, 0x9E, 0xFA, 0x57, 0x36, 0xAA, 0x3F,
|
|
|
|
0xC1, 0xA3, 0x2C, 0xE9, 0xB5, 0xC9, 0xED, 0x72, 0x51, 0x4C,
|
|
|
|
0x02, 0xB4, 0x7B, 0xB3, 0xED, 0x9F, 0x45, 0x03, 0x34, 0xAC,
|
|
|
|
0x9A, 0x9E, 0x62, 0x5F, 0x82, 0x7A, 0x77, 0x34, 0xF9, 0x21,
|
|
|
|
0x94, 0xD2, 0x38, 0x3D, 0x05, 0xF0, 0x8A, 0x60, 0x1C, 0xB7,
|
|
|
|
0x1D, 0xF5, 0xB7, 0x53, 0x77, 0xD3, 0x9D, 0x3D, 0x70, 0x6A,
|
|
|
|
0xCB, 0x18, 0x20, 0x6B, 0x29, 0x17, 0x3A, 0x6D, 0xA1, 0xB2,
|
|
|
|
0x64, 0xDB, 0x6C, 0xE6, 0x1A, 0x95, 0xA7, 0xF4, 0x1A, 0x78,
|
|
|
|
0x1D, 0xA2, 0x40, 0x15, 0x41, 0x59, 0xDD, 0xEE, 0x23, 0x57,
|
|
|
|
0xCE, 0x36, 0x0D, 0x55, 0xBD, 0xB8, 0xFD, 0x0F, 0x35, 0xBD,
|
|
|
|
0x5B, 0x92, 0xD6, 0x1C, 0x84, 0x8C, 0x32, 0x64, 0xA6, 0x5C,
|
|
|
|
0x45, 0x18, 0x07, 0x6B, 0xF9, 0xA9, 0x43, 0x9A, 0x83, 0xCD,
|
|
|
|
0xB5, 0xB3, 0xD9, 0x17, 0x99, 0x2C, 0x2A, 0x8B, 0xE0, 0x8E,
|
|
|
|
0xAF, 0xA6, 0x4C, 0x95, 0xBB, 0x70, 0x60, 0x1A, 0x3A, 0x97,
|
|
|
|
0xAA, 0x2F, 0x3D, 0x22, 0x83, 0xB7, 0x4F, 0x59, 0xED, 0x3F,
|
|
|
|
0x4E, 0xF4, 0x19, 0xC6, 0x25, 0x0B, 0x0A, 0x5E, 0x21, 0xB9,
|
|
|
|
0x91, 0xB8, 0x19, 0x84, 0x48, 0x78, 0xCE, 0x27, 0xBF, 0x41,
|
|
|
|
0x89, 0xF6, 0x30, 0xFD, 0x6B, 0xD9, 0xB8, 0x1D, 0x72, 0x8A,
|
|
|
|
0x56, 0xCC, 0x2F, 0x82, 0xE4, 0x46, 0x4D, 0x75, 0xD8, 0x92,
|
|
|
|
0xE6, 0x9C, 0xCC, 0xD2, 0xCD, 0x35, 0xE4, 0xFC, 0x2A, 0x85,
|
|
|
|
0x6B, 0xA9, 0xB2, 0x27, 0xC9, 0xA1, 0xFF, 0xB3, 0x96, 0x3E,
|
|
|
|
0x59, 0xF6, 0x4C, 0x66, 0x56, 0x2E, 0xF5, 0x1B, 0x97, 0x32,
|
|
|
|
0xB0, 0x71, 0x5A, 0x9C, 0x50, 0x4B, 0x6F, 0xC4, 0xCA, 0x94,
|
|
|
|
0x75, 0x37, 0x46, 0x10, 0x12, 0x2F, 0x4F, 0xA3, 0x82, 0xCD,
|
|
|
|
0xBD, 0x7C};
|
|
|
|
|
|
|
|
const uint8_t ss_exp[SIKEp503_SS_BYTESZ] = {
|
|
|
|
0x74, 0x3D, 0x25, 0x36, 0x00, 0x24, 0x63, 0x1A, 0x39, 0x1A,
|
|
|
|
0xB4, 0xAD, 0x01, 0x17, 0x78, 0xE9};
|
|
|
|
|
|
|
|
uint8_t ss_dec[SIKEp503_SS_BYTESZ] = {0};
|
|
|
|
SIKE_decaps(ss_dec, ct_exp, pk, sk);
|
|
|
|
EXPECT_EQ(memcmp(ss_dec, ss_exp, sizeof(ss_exp)), 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
// SIKE_encaps and SIKE_keypair doesn't return zeros.
|
|
|
|
TEST(SIKE, NonZero) {
|
|
|
|
uint8_t sk[SIKEp503_PRV_BYTESZ] = {0};
|
|
|
|
uint8_t pk[SIKEp503_PUB_BYTESZ] = {0};
|
|
|
|
uint8_t ct[SIKEp503_CT_BYTESZ] = {0};
|
|
|
|
uint8_t ss[SIKEp503_SS_BYTESZ] = {0};
|
|
|
|
|
|
|
|
// Check secret and public key returned by SIKE_keypair
|
|
|
|
EXPECT_EQ(SIKE_keypair(sk, pk), 1);
|
|
|
|
uint8_t tmp = 0;
|
|
|
|
for (size_t i=0; i<sizeof(sk); i++) tmp|=sk[i];
|
|
|
|
EXPECT_NE(tmp, 0);
|
|
|
|
|
|
|
|
tmp = 0;
|
|
|
|
for (size_t i=0; i<sizeof(pk); i++) tmp|=pk[i];
|
|
|
|
EXPECT_NE(tmp, 0);
|
|
|
|
|
|
|
|
// Check shared secret and ciphertext returned by SIKE_encaps
|
|
|
|
SIKE_encaps(ss, ct, pk);
|
|
|
|
tmp = 0;
|
|
|
|
for (size_t i=0; i<sizeof(ct); i++) tmp|=ct[i];
|
|
|
|
EXPECT_NE(tmp, 0);
|
|
|
|
|
|
|
|
tmp = 0;
|
|
|
|
for (size_t i=0; i<sizeof(ss); i++) tmp|=ss[i];
|
|
|
|
EXPECT_NE(tmp, 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(SIKE, Negative) {
|
|
|
|
uint8_t sk[SIKEp503_PRV_BYTESZ] = {0};
|
|
|
|
uint8_t pk[SIKEp503_PUB_BYTESZ] = {0};
|
|
|
|
uint8_t ct[SIKEp503_CT_BYTESZ] = {0};
|
|
|
|
uint8_t ss_enc[SIKEp503_SS_BYTESZ] = {0};
|
|
|
|
uint8_t ss_dec[SIKEp503_SS_BYTESZ] = {0};
|
|
|
|
|
|
|
|
EXPECT_EQ(SIKE_keypair(sk, pk), 1);
|
|
|
|
SIKE_encaps(ss_enc, ct, pk);
|
|
|
|
|
|
|
|
// Change cipertext
|
|
|
|
uint8_t ct_tmp[SIKEp503_CT_BYTESZ] = {0};
|
|
|
|
memcpy(ct_tmp, ct, sizeof(ct));
|
|
|
|
ct_tmp[0] = ~ct_tmp[0];
|
|
|
|
SIKE_decaps(ss_dec, ct_tmp, pk, sk);
|
|
|
|
EXPECT_NE(memcmp(ss_enc, ss_dec, SIKEp503_SS_BYTESZ), 0);
|
|
|
|
|
|
|
|
// Change secret key
|
|
|
|
uint8_t sk_tmp[SIKEp503_PRV_BYTESZ] = {0};
|
|
|
|
memcpy(sk_tmp, sk, sizeof(sk));
|
|
|
|
sk_tmp[0] = ~sk_tmp[0];
|
|
|
|
SIKE_decaps(ss_dec, ct, pk, sk_tmp);
|
|
|
|
EXPECT_NE(memcmp(ss_enc, ss_dec, SIKEp503_SS_BYTESZ), 0);
|
|
|
|
|
|
|
|
// Change public key
|
|
|
|
uint8_t pk_tmp[SIKEp503_PUB_BYTESZ] = {0};
|
|
|
|
memcpy(pk_tmp, pk, sizeof(pk));
|
|
|
|
pk_tmp[0] = ~pk_tmp[0];
|
|
|
|
SIKE_decaps(ss_dec, ct, pk_tmp, sk);
|
|
|
|
EXPECT_NE(memcmp(ss_enc, ss_dec, SIKEp503_SS_BYTESZ), 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
#if defined(SUPPORTS_ABI_TEST) && defined(OPENSSL_X86_64)
|
|
|
|
TEST(SIKE, ABI) {
|
|
|
|
felm_t a, b, c;
|
2019-04-19 12:58:50 +01:00
|
|
|
dfelm_t d, e, f;
|
Add support for SIKE/p503 post-quantum KEM
Based on Microsoft's implementation available on github:
Source: https://github.com/Microsoft/PQCrypto-SIDH
Commit: 77044b76181eb61c744ac8eb7ddc7a8fe72f6919
Following changes has been applied
* In intel assembly, use MOV instead of MOVQ:
Intel instruction reference in the Intel Software Developer's Manual
volume 2A, the MOVQ has 4 forms. None of them mentions moving
literal to GPR, hence "movq $rax, 0x0" is wrong. Instead, on 64bit
system, MOV can be used.
* Some variables were wrongly zero-initialized (as per C99 spec)
* Move constant values to .RODATA segment, as keeping them in .TEXT
segment is not compatible with XOM.
* Fixes issue in arm64 code related to the fact that compiler doesn't
reserve enough space for the linker to relocate address of a global
variable when used by 'ldr' instructions. Solution is to use 'adrp'
followed by 'add' instruction. Relocations for 'adrp' and 'add'
instructions is generated by prefixing the label with :pg_hi21:
and :lo12: respectively.
* Enable MULX and ADX. Code from MS doesn't support PIC. MULX can't
reference global variable directly. Instead RIP-relative addressing
can be used. This improves performance around 10%-13% on SkyLake
* Check if CPU supports BMI2 and ADOX instruction at runtime. On AMD64
optimized implementation of montgomery multiplication and reduction
have 2 implementations - faster one takes advantage of BMI2
instruction set introduced in Haswell and ADOX introduced in
Broadwell. Thanks to OPENSSL_ia32cap_P it can be decided at runtime
which implementation to choose. As CPU configuration is static by
nature, branch predictor will be correct most of the time and hence
this check very often has no cost.
* Reuse some utilities from boringssl instead of reimplementing them.
This includes things like:
* definition of a limb size (use crypto_word_t instead of digit_t)
* use functions for checking in constant time if value is 0 and/or
less then
* #define's used for conditional compilation
* Use SSE2 for conditional swap on vector registers. Improves
performance a little bit.
* Fix f2elm_t definition. Code imported from MSR defines f2elm_t type as
a array of arrays. This decays to a pointer to an array (when passing
as an argument). In C, one can't assign const pointer to an array with
non-const pointer to an array. Seems it violates 6.7.3/8 from C99
(same for C11). This problem occures in GCC 6, only when -pedantic
flag is specified and it occures always in GCC 4.9 (debian jessie).
* Fix definition of eval_3_isog. Second argument in eval_3_isog mustn't be
const. Similar reason as above.
* Use HMAC-SHA256 instead of cSHAKE-256 to avoid upstreaming cSHAKE
and SHA3 code.
* Add speed and unit tests for SIKE.
Change-Id: I22f0bb1f9edff314a35cd74b48e8c4962568e330
2019-03-06 18:19:25 +00:00
|
|
|
CHECK_ABI(sike_fpadd, a, b, c);
|
|
|
|
CHECK_ABI(sike_fpsub, a, b, c);
|
|
|
|
CHECK_ABI(sike_mpmul, a, b, d);
|
|
|
|
CHECK_ABI(sike_fprdc, d, a);
|
|
|
|
CHECK_ABI(sike_mpadd_asm, a, b, c);
|
2019-04-19 12:58:50 +01:00
|
|
|
CHECK_ABI(sike_mpsubx2_asm, d, e, f);
|
|
|
|
CHECK_ABI(sike_mpdblsubx2_asm, d, e, f);
|
Add support for SIKE/p503 post-quantum KEM
Based on Microsoft's implementation available on github:
Source: https://github.com/Microsoft/PQCrypto-SIDH
Commit: 77044b76181eb61c744ac8eb7ddc7a8fe72f6919
Following changes has been applied
* In intel assembly, use MOV instead of MOVQ:
Intel instruction reference in the Intel Software Developer's Manual
volume 2A, the MOVQ has 4 forms. None of them mentions moving
literal to GPR, hence "movq $rax, 0x0" is wrong. Instead, on 64bit
system, MOV can be used.
* Some variables were wrongly zero-initialized (as per C99 spec)
* Move constant values to .RODATA segment, as keeping them in .TEXT
segment is not compatible with XOM.
* Fixes issue in arm64 code related to the fact that compiler doesn't
reserve enough space for the linker to relocate address of a global
variable when used by 'ldr' instructions. Solution is to use 'adrp'
followed by 'add' instruction. Relocations for 'adrp' and 'add'
instructions is generated by prefixing the label with :pg_hi21:
and :lo12: respectively.
* Enable MULX and ADX. Code from MS doesn't support PIC. MULX can't
reference global variable directly. Instead RIP-relative addressing
can be used. This improves performance around 10%-13% on SkyLake
* Check if CPU supports BMI2 and ADOX instruction at runtime. On AMD64
optimized implementation of montgomery multiplication and reduction
have 2 implementations - faster one takes advantage of BMI2
instruction set introduced in Haswell and ADOX introduced in
Broadwell. Thanks to OPENSSL_ia32cap_P it can be decided at runtime
which implementation to choose. As CPU configuration is static by
nature, branch predictor will be correct most of the time and hence
this check very often has no cost.
* Reuse some utilities from boringssl instead of reimplementing them.
This includes things like:
* definition of a limb size (use crypto_word_t instead of digit_t)
* use functions for checking in constant time if value is 0 and/or
less then
* #define's used for conditional compilation
* Use SSE2 for conditional swap on vector registers. Improves
performance a little bit.
* Fix f2elm_t definition. Code imported from MSR defines f2elm_t type as
a array of arrays. This decays to a pointer to an array (when passing
as an argument). In C, one can't assign const pointer to an array with
non-const pointer to an array. Seems it violates 6.7.3/8 from C99
(same for C11). This problem occures in GCC 6, only when -pedantic
flag is specified and it occures always in GCC 4.9 (debian jessie).
* Fix definition of eval_3_isog. Second argument in eval_3_isog mustn't be
const. Similar reason as above.
* Use HMAC-SHA256 instead of cSHAKE-256 to avoid upstreaming cSHAKE
and SHA3 code.
* Add speed and unit tests for SIKE.
Change-Id: I22f0bb1f9edff314a35cd74b48e8c4962568e330
2019-03-06 18:19:25 +00:00
|
|
|
}
|
|
|
|
#endif // SUPPORTS_ABI_TEST && X86_64
|