Merge BoringSSL 884086e: Remove x86_64 x25519 assembly.
This commit is contained in:
Brian Smith
commit
82804e9460
3
build.rs
3
build.rs
@ -91,8 +91,6 @@ const RING_SRCS: &'static [(&'static [&'static str], &'static str)] = &[
|
||||
(&[X86], "crypto/fipsmodule/sha/asm/sha256-586.pl"),
|
||||
(&[X86], "crypto/fipsmodule/sha/asm/sha512-586.pl"),
|
||||
|
||||
(&[X86_64], "crypto/curve25519/x25519-x86_64.c"),
|
||||
|
||||
(&[X86_64], "crypto/fipsmodule/aes/asm/aes-x86_64.pl"),
|
||||
(&[X86_64], "crypto/fipsmodule/aes/asm/aesni-x86_64.pl"),
|
||||
(&[X86_64], "crypto/fipsmodule/aes/asm/bsaes-x86_64.pl"),
|
||||
@ -101,7 +99,6 @@ const RING_SRCS: &'static [(&'static [&'static str], &'static str)] = &[
|
||||
(&[X86_64], "crypto/fipsmodule/bn/asm/x86_64-mont5.pl"),
|
||||
(&[X86_64], "crypto/chacha/asm/chacha-x86_64.pl"),
|
||||
(&[NEVER], "crypto/cipher_extra/asm/aes128gcmsiv-x86_64.pl"),
|
||||
(&[X86_64], "crypto/curve25519/asm/x25519-asm-x86_64.S"),
|
||||
(&[X86_64], "crypto/fipsmodule/ec/asm/ecp_nistz256-x86_64.pl"),
|
||||
(&[X86_64], "crypto/fipsmodule/ec/asm/p256-x86_64-asm.pl"),
|
||||
(&[X86_64], "crypto/fipsmodule/modes/asm/aesni-gcm-x86_64.pl"),
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
@ -1,245 +0,0 @@
|
||||
/* Copyright (c) 2015, 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. */
|
||||
|
||||
// This code is mostly taken from the ref10 version of Ed25519 in SUPERCOP
|
||||
// 20141124 (http://bench.cr.yp.to/supercop.html). That code is released as
|
||||
// public domain but this file has the ISC license just to keep licencing
|
||||
// simple.
|
||||
//
|
||||
// The field functions are shared by Ed25519 and X25519 where possible.
|
||||
|
||||
#include <string.h>
|
||||
|
||||
#include "../internal.h"
|
||||
#include "../../third_party/fiat/internal.h"
|
||||
|
||||
|
||||
#if defined(BORINGSSL_X25519_X86_64)
|
||||
|
||||
typedef struct { uint64_t v[5]; } fe25519;
|
||||
|
||||
// These functions are defined in asm/x25519-x86_64.S.
|
||||
void GFp_x25519_x86_64_work_cswap(fe25519 *, uint64_t);
|
||||
void GFp_x25519_x86_64_mul(fe25519 *out, const fe25519 *a, const fe25519 *b);
|
||||
void GFp_x25519_x86_64_square(fe25519 *out, const fe25519 *a);
|
||||
void GFp_x25519_x86_64_freeze(fe25519 *);
|
||||
void GFp_x25519_x86_64_ladderstep(fe25519 *work);
|
||||
|
||||
static void fe25519_setint(fe25519 *r, unsigned v) {
|
||||
r->v[0] = v;
|
||||
r->v[1] = 0;
|
||||
r->v[2] = 0;
|
||||
r->v[3] = 0;
|
||||
r->v[4] = 0;
|
||||
}
|
||||
|
||||
// Assumes input x being reduced below 2^255
|
||||
static void fe25519_pack(unsigned char r[32], const fe25519 *x) {
|
||||
fe25519 t;
|
||||
t = *x;
|
||||
GFp_x25519_x86_64_freeze(&t);
|
||||
|
||||
r[0] = (uint8_t)(t.v[0] & 0xff);
|
||||
r[1] = (uint8_t)((t.v[0] >> 8) & 0xff);
|
||||
r[2] = (uint8_t)((t.v[0] >> 16) & 0xff);
|
||||
r[3] = (uint8_t)((t.v[0] >> 24) & 0xff);
|
||||
r[4] = (uint8_t)((t.v[0] >> 32) & 0xff);
|
||||
r[5] = (uint8_t)((t.v[0] >> 40) & 0xff);
|
||||
r[6] = (uint8_t)((t.v[0] >> 48));
|
||||
|
||||
r[6] ^= (uint8_t)((t.v[1] << 3) & 0xf8);
|
||||
r[7] = (uint8_t)((t.v[1] >> 5) & 0xff);
|
||||
r[8] = (uint8_t)((t.v[1] >> 13) & 0xff);
|
||||
r[9] = (uint8_t)((t.v[1] >> 21) & 0xff);
|
||||
r[10] = (uint8_t)((t.v[1] >> 29) & 0xff);
|
||||
r[11] = (uint8_t)((t.v[1] >> 37) & 0xff);
|
||||
r[12] = (uint8_t)((t.v[1] >> 45));
|
||||
|
||||
r[12] ^= (uint8_t)((t.v[2] << 6) & 0xc0);
|
||||
r[13] = (uint8_t)((t.v[2] >> 2) & 0xff);
|
||||
r[14] = (uint8_t)((t.v[2] >> 10) & 0xff);
|
||||
r[15] = (uint8_t)((t.v[2] >> 18) & 0xff);
|
||||
r[16] = (uint8_t)((t.v[2] >> 26) & 0xff);
|
||||
r[17] = (uint8_t)((t.v[2] >> 34) & 0xff);
|
||||
r[18] = (uint8_t)((t.v[2] >> 42) & 0xff);
|
||||
r[19] = (uint8_t)((t.v[2] >> 50));
|
||||
|
||||
r[19] ^= (uint8_t)((t.v[3] << 1) & 0xfe);
|
||||
r[20] = (uint8_t)((t.v[3] >> 7) & 0xff);
|
||||
r[21] = (uint8_t)((t.v[3] >> 15) & 0xff);
|
||||
r[22] = (uint8_t)((t.v[3] >> 23) & 0xff);
|
||||
r[23] = (uint8_t)((t.v[3] >> 31) & 0xff);
|
||||
r[24] = (uint8_t)((t.v[3] >> 39) & 0xff);
|
||||
r[25] = (uint8_t)((t.v[3] >> 47));
|
||||
|
||||
r[25] ^= (uint8_t)((t.v[4] << 4) & 0xf0);
|
||||
r[26] = (uint8_t)((t.v[4] >> 4) & 0xff);
|
||||
r[27] = (uint8_t)((t.v[4] >> 12) & 0xff);
|
||||
r[28] = (uint8_t)((t.v[4] >> 20) & 0xff);
|
||||
r[29] = (uint8_t)((t.v[4] >> 28) & 0xff);
|
||||
r[30] = (uint8_t)((t.v[4] >> 36) & 0xff);
|
||||
r[31] = (uint8_t)((t.v[4] >> 44));
|
||||
}
|
||||
|
||||
static void fe25519_unpack(fe25519 *r, const uint8_t x[32]) {
|
||||
r->v[0] = x[0];
|
||||
r->v[0] += (uint64_t)x[1] << 8;
|
||||
r->v[0] += (uint64_t)x[2] << 16;
|
||||
r->v[0] += (uint64_t)x[3] << 24;
|
||||
r->v[0] += (uint64_t)x[4] << 32;
|
||||
r->v[0] += (uint64_t)x[5] << 40;
|
||||
r->v[0] += ((uint64_t)x[6] & 7) << 48;
|
||||
|
||||
r->v[1] = x[6] >> 3;
|
||||
r->v[1] += (uint64_t)x[7] << 5;
|
||||
r->v[1] += (uint64_t)x[8] << 13;
|
||||
r->v[1] += (uint64_t)x[9] << 21;
|
||||
r->v[1] += (uint64_t)x[10] << 29;
|
||||
r->v[1] += (uint64_t)x[11] << 37;
|
||||
r->v[1] += ((uint64_t)x[12] & 63) << 45;
|
||||
|
||||
r->v[2] = x[12] >> 6;
|
||||
r->v[2] += (uint64_t)x[13] << 2;
|
||||
r->v[2] += (uint64_t)x[14] << 10;
|
||||
r->v[2] += (uint64_t)x[15] << 18;
|
||||
r->v[2] += (uint64_t)x[16] << 26;
|
||||
r->v[2] += (uint64_t)x[17] << 34;
|
||||
r->v[2] += (uint64_t)x[18] << 42;
|
||||
r->v[2] += ((uint64_t)x[19] & 1) << 50;
|
||||
|
||||
r->v[3] = x[19] >> 1;
|
||||
r->v[3] += (uint64_t)x[20] << 7;
|
||||
r->v[3] += (uint64_t)x[21] << 15;
|
||||
r->v[3] += (uint64_t)x[22] << 23;
|
||||
r->v[3] += (uint64_t)x[23] << 31;
|
||||
r->v[3] += (uint64_t)x[24] << 39;
|
||||
r->v[3] += ((uint64_t)x[25] & 15) << 47;
|
||||
|
||||
r->v[4] = x[25] >> 4;
|
||||
r->v[4] += (uint64_t)x[26] << 4;
|
||||
r->v[4] += (uint64_t)x[27] << 12;
|
||||
r->v[4] += (uint64_t)x[28] << 20;
|
||||
r->v[4] += (uint64_t)x[29] << 28;
|
||||
r->v[4] += (uint64_t)x[30] << 36;
|
||||
r->v[4] += ((uint64_t)x[31] & 127) << 44;
|
||||
}
|
||||
|
||||
static void fe25519_invert(fe25519 *r, const fe25519 *x) {
|
||||
fe25519 z2;
|
||||
fe25519 z9;
|
||||
fe25519 z11;
|
||||
fe25519 z2_5_0;
|
||||
fe25519 z2_10_0;
|
||||
fe25519 z2_20_0;
|
||||
fe25519 z2_50_0;
|
||||
fe25519 z2_100_0;
|
||||
fe25519 t;
|
||||
int i;
|
||||
|
||||
/* 2 */ GFp_x25519_x86_64_square(&z2, x);
|
||||
/* 4 */ GFp_x25519_x86_64_square(&t, &z2);
|
||||
/* 8 */ GFp_x25519_x86_64_square(&t, &t);
|
||||
/* 9 */ GFp_x25519_x86_64_mul(&z9, &t, x);
|
||||
/* 11 */ GFp_x25519_x86_64_mul(&z11, &z9, &z2);
|
||||
/* 22 */ GFp_x25519_x86_64_square(&t, &z11);
|
||||
/* 2^5 - 2^0 = 31 */ GFp_x25519_x86_64_mul(&z2_5_0, &t, &z9);
|
||||
|
||||
/* 2^6 - 2^1 */ GFp_x25519_x86_64_square(&t, &z2_5_0);
|
||||
/* 2^20 - 2^10 */ for (i = 1; i < 5; i++) { GFp_x25519_x86_64_square(&t, &t); }
|
||||
/* 2^10 - 2^0 */ GFp_x25519_x86_64_mul(&z2_10_0, &t, &z2_5_0);
|
||||
|
||||
/* 2^11 - 2^1 */ GFp_x25519_x86_64_square(&t, &z2_10_0);
|
||||
/* 2^20 - 2^10 */ for (i = 1; i < 10; i++) { GFp_x25519_x86_64_square(&t, &t); }
|
||||
/* 2^20 - 2^0 */ GFp_x25519_x86_64_mul(&z2_20_0, &t, &z2_10_0);
|
||||
|
||||
/* 2^21 - 2^1 */ GFp_x25519_x86_64_square(&t, &z2_20_0);
|
||||
/* 2^40 - 2^20 */ for (i = 1; i < 20; i++) { GFp_x25519_x86_64_square(&t, &t); }
|
||||
/* 2^40 - 2^0 */ GFp_x25519_x86_64_mul(&t, &t, &z2_20_0);
|
||||
|
||||
/* 2^41 - 2^1 */ GFp_x25519_x86_64_square(&t, &t);
|
||||
/* 2^50 - 2^10 */ for (i = 1; i < 10; i++) { GFp_x25519_x86_64_square(&t, &t); }
|
||||
/* 2^50 - 2^0 */ GFp_x25519_x86_64_mul(&z2_50_0, &t, &z2_10_0);
|
||||
|
||||
/* 2^51 - 2^1 */ GFp_x25519_x86_64_square(&t, &z2_50_0);
|
||||
/* 2^100 - 2^50 */ for (i = 1; i < 50; i++) { GFp_x25519_x86_64_square(&t, &t); }
|
||||
/* 2^100 - 2^0 */ GFp_x25519_x86_64_mul(&z2_100_0, &t, &z2_50_0);
|
||||
|
||||
/* 2^101 - 2^1 */ GFp_x25519_x86_64_square(&t, &z2_100_0);
|
||||
/* 2^200 - 2^100 */ for (i = 1; i < 100; i++) {
|
||||
GFp_x25519_x86_64_square(&t, &t);
|
||||
}
|
||||
/* 2^200 - 2^0 */ GFp_x25519_x86_64_mul(&t, &t, &z2_100_0);
|
||||
|
||||
/* 2^201 - 2^1 */ GFp_x25519_x86_64_square(&t, &t);
|
||||
/* 2^250 - 2^50 */ for (i = 1; i < 50; i++) { GFp_x25519_x86_64_square(&t, &t); }
|
||||
/* 2^250 - 2^0 */ GFp_x25519_x86_64_mul(&t, &t, &z2_50_0);
|
||||
|
||||
/* 2^251 - 2^1 */ GFp_x25519_x86_64_square(&t, &t);
|
||||
/* 2^252 - 2^2 */ GFp_x25519_x86_64_square(&t, &t);
|
||||
/* 2^253 - 2^3 */ GFp_x25519_x86_64_square(&t, &t);
|
||||
|
||||
/* 2^254 - 2^4 */ GFp_x25519_x86_64_square(&t, &t);
|
||||
|
||||
/* 2^255 - 2^5 */ GFp_x25519_x86_64_square(&t, &t);
|
||||
/* 2^255 - 21 */ GFp_x25519_x86_64_mul(r, &t, &z11);
|
||||
}
|
||||
|
||||
static void mladder(fe25519 *xr, fe25519 *zr, const uint8_t s[32]) {
|
||||
fe25519 work[5];
|
||||
|
||||
work[0] = *xr;
|
||||
fe25519_setint(work + 1, 1);
|
||||
fe25519_setint(work + 2, 0);
|
||||
work[3] = *xr;
|
||||
fe25519_setint(work + 4, 1);
|
||||
|
||||
int i, j;
|
||||
uint8_t prevbit = 0;
|
||||
|
||||
j = 6;
|
||||
for (i = 31; i >= 0; i--) {
|
||||
while (j >= 0) {
|
||||
const uint8_t bit = 1 & (s[i] >> j);
|
||||
const uint64_t swap = bit ^ prevbit;
|
||||
prevbit = bit;
|
||||
GFp_x25519_x86_64_work_cswap(work + 1, swap);
|
||||
GFp_x25519_x86_64_ladderstep(work);
|
||||
j -= 1;
|
||||
}
|
||||
j = 7;
|
||||
}
|
||||
|
||||
*xr = work[1];
|
||||
*zr = work[2];
|
||||
}
|
||||
|
||||
void GFp_x25519_x86_64(uint8_t out[32], const uint8_t scalar[32],
|
||||
const uint8_t point[32]) {
|
||||
uint8_t e[32];
|
||||
memcpy(e, scalar, sizeof(e));
|
||||
|
||||
e[0] &= 248;
|
||||
e[31] &= 127;
|
||||
e[31] |= 64;
|
||||
|
||||
fe25519 t;
|
||||
fe25519 z;
|
||||
fe25519_unpack(&t, point);
|
||||
mladder(&t, &z, e);
|
||||
fe25519_invert(&z, &z);
|
||||
GFp_x25519_x86_64_mul(&t, &t, &z);
|
||||
fe25519_pack(out, &t);
|
||||
}
|
||||
|
||||
#endif // BORINGSSL_X25519_X86_64
|
||||
34
third_party/fiat/curve25519.c
vendored
34
third_party/fiat/curve25519.c
vendored
@ -513,8 +513,6 @@ static void fe_sq_tt(fe *h, const fe *f) {
|
||||
fe_sqr_impl(h->v, f->v);
|
||||
}
|
||||
|
||||
#if !defined(BORINGSSL_X25519_X86_64)
|
||||
|
||||
// Replace (f,g) with (g,f) if b == 1;
|
||||
// replace (f,g) with (f,g) if b == 0.
|
||||
//
|
||||
@ -590,8 +588,6 @@ static void fe_mul121666(fe *h, const fe_loose *f) {
|
||||
assert_fe(h->v);
|
||||
}
|
||||
|
||||
#endif // !BORINGSSL_X25519_X86_64
|
||||
|
||||
// Adapted from Fiat-synthesized |fe_sub_impl| with |out| = 0.
|
||||
static void fe_neg_impl(uint64_t out[5], const uint64_t in2[5]) {
|
||||
{ const uint64_t x10 = 0;
|
||||
@ -1202,8 +1198,6 @@ static void fe_sq_tt(fe *h, const fe *f) {
|
||||
fe_sqr_impl(h->v, f->v);
|
||||
}
|
||||
|
||||
#if !defined(BORINGSSL_X25519_X86_64)
|
||||
|
||||
// Replace (f,g) with (g,f) if b == 1;
|
||||
// replace (f,g) with (f,g) if b == 0.
|
||||
//
|
||||
@ -1343,8 +1337,6 @@ static void fe_mul121666(fe *h, const fe_loose *f) {
|
||||
assert_fe(h->v);
|
||||
}
|
||||
|
||||
#endif // !BORINGSSL_X25519_X86_64
|
||||
|
||||
// Adapted from Fiat-synthesized |fe_sub_impl| with |out| = 0.
|
||||
static void fe_neg_impl(uint32_t out[10], const uint32_t in2[10]) {
|
||||
{ const uint32_t x20 = 0;
|
||||
@ -2866,14 +2858,6 @@ static void sc_muladd(uint8_t *s, const uint8_t *a, const uint8_t *b,
|
||||
s[31] = s11 >> 17;
|
||||
}
|
||||
|
||||
#if defined(BORINGSSL_X25519_X86_64)
|
||||
|
||||
static void x25519_scalar_mult(uint8_t out[32], const uint8_t scalar[32],
|
||||
const uint8_t point[32]) {
|
||||
GFp_x25519_x86_64(out, scalar, point);
|
||||
}
|
||||
|
||||
#else
|
||||
|
||||
static void x25519_scalar_mult_generic(uint8_t out[32],
|
||||
const uint8_t scalar[32],
|
||||
@ -2966,8 +2950,6 @@ static void x25519_scalar_mult(uint8_t out[32], const uint8_t scalar[32],
|
||||
x25519_scalar_mult_generic(out, scalar, point);
|
||||
}
|
||||
|
||||
#endif // BORINGSSL_X25519_X86_64
|
||||
|
||||
void GFp_x25519_scalar_mult(uint8_t out[32], const uint8_t scalar[32],
|
||||
const uint8_t point[32]) {
|
||||
x25519_scalar_mult(out, scalar, point);
|
||||
@ -2977,20 +2959,6 @@ void GFp_x25519_scalar_mult(uint8_t out[32], const uint8_t scalar[32],
|
||||
void GFp_x25519_public_from_private(uint8_t out_public_value[32],
|
||||
const uint8_t private_key[32]);
|
||||
|
||||
#if defined(BORINGSSL_X25519_X86_64)
|
||||
|
||||
// When |BORINGSSL_X25519_X86_64| is set, base point multiplication is done with
|
||||
// the Montgomery ladder because it's faster. Otherwise it's done using the
|
||||
// Ed25519 tables.
|
||||
|
||||
void GFp_x25519_public_from_private(uint8_t out_public_value[32],
|
||||
const uint8_t private_key[32]) {
|
||||
static const uint8_t kMongomeryBasePoint[32] = {9};
|
||||
GFp_x25519_scalar_mult(out_public_value, private_key, kMongomeryBasePoint);
|
||||
}
|
||||
|
||||
#else
|
||||
|
||||
void GFp_x25519_public_from_private(uint8_t out_public_value[32],
|
||||
const uint8_t private_key[32]) {
|
||||
#if defined(BORINGSSL_X25519_NEON)
|
||||
@ -3019,8 +2987,6 @@ void GFp_x25519_public_from_private(uint8_t out_public_value[32],
|
||||
fe_tobytes(out_public_value, &zminusy_inv);
|
||||
}
|
||||
|
||||
#endif // BORINGSSL_X25519_X86_64
|
||||
|
||||
void GFp_x25519_fe_invert(fe *out, const fe *z) {
|
||||
fe_invert(out, z);
|
||||
}
|
||||
|
||||
11
third_party/fiat/internal.h
vendored
11
third_party/fiat/internal.h
vendored
@ -37,17 +37,6 @@ extern "C" {
|
||||
#include "../../crypto/internal.h"
|
||||
|
||||
|
||||
#if defined(OPENSSL_X86_64) && !defined(OPENSSL_SMALL) && \
|
||||
!defined(OPENSSL_WINDOWS) && !defined(OPENSSL_NO_ASM)
|
||||
// This isn't compatible with Windows because the asm code makes use of the red
|
||||
// zone, which Windows doesn't support.
|
||||
#define BORINGSSL_X25519_X86_64
|
||||
|
||||
void GFp_x25519_x86_64(uint8_t out[32], const uint8_t scalar[32],
|
||||
const uint8_t point[32]);
|
||||
#endif
|
||||
|
||||
|
||||
#if defined(OPENSSL_ARM) && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_APPLE)
|
||||
#define BORINGSSL_X25519_NEON
|
||||
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user