Update to v106r65 release.

byuu says:

This synchronizes bsnes/higan with many recent internal nall changes.

This will be the last WIP until I am situated in Japan. Apologies for the
bugfixes that didn't get applied yet, I ran out of time.
This commit is contained in:
Tim Allen
2018-10-04 20:11:23 +10:00
parent 336d20123f
commit 03b06257d3
75 changed files with 2242 additions and 1371 deletions

View File

@@ -4,64 +4,42 @@
#if defined(EC_REFERENCE)
#include <nall/elliptic-curve/modulo25519-reference.hpp>
#else
#include <nall/elliptic-curve/modulo25519.hpp>
#include <nall/elliptic-curve/modulo25519-optimized.hpp>
#endif
namespace nall { namespace EllipticCurve {
static const uint256_t L = (1_u256 << 252) + 27742317777372353535851937790883648493_u256;
struct Ed25519 {
Ed25519() {
field y = field(4) * field(5).reciprocal();
field x = recoverX(y);
point B{x, y, 1, x * y};
for(uint n : range(253)) {
Bscalar[n] = B;
B = edwardsDouble(B);
}
}
auto publicKey(uint256_t privateKey) const -> uint256_t {
auto H = uint512_t{Hash::SHA512(to_vector(privateKey)).output()};
auto a = clamp(H);
auto A = compress(scalarMultiplyB(modL(a)));
return A;
return compress(scalarMultiply(B, clamp(hash(privateKey)) % L));
}
auto sign(const vector<uint8_t>& message, uint256_t privateKey) const -> uint512_t {
auto H = uint512_t{Hash::SHA512(to_vector(privateKey)).output()};
auto a = clamp(H);
auto A = compress(scalarMultiplyB(modL(a)));
auto sign(array_view<uint8_t> message, uint256_t privateKey) const -> uint512_t {
uint512_t H = hash(privateKey);
uint256_t a = clamp(H) % L;
uint256_t A = compress(scalarMultiply(B, a));
Hash::SHA512 hash1;
hash1.input(to_vector(upper(H)));
hash1.input(message);
auto r = uint512_t{hash1.output()};
auto R = compress(scalarMultiplyB(modL(r)));
uint512_t r = hash(upper(H), message) % L;
uint256_t R = compress(scalarMultiply(B, r));
Hash::SHA512 hash2;
hash2.input(to_vector(R));
hash2.input(to_vector(A));
hash2.input(message);
uint512_t k = modL(uint512_t{hash2.output()});
uint256_t S = modL(k * a + r);
uint512_t k = hash(R, A, message) % L;
uint256_t S = (k * a + r) % L;
return uint512_t(S) << 256 | R;
}
auto verify(const vector<uint8_t>& message, uint512_t signature, uint256_t publicKey) const -> bool {
auto verify(array_view<uint8_t> message, uint512_t signature, uint256_t publicKey) const -> bool {
auto R = decompress(lower(signature));
auto A = decompress(publicKey);
if(!R || !A) return false;
uint256_t S = upper(signature);
Hash::SHA512 hash;
hash.input(to_vector(lower(signature)));
hash.input(to_vector(publicKey));
hash.input(message);
auto r = uint512_t{hash.output()};
uint256_t S = upper(signature) % L;
uint512_t r = hash(lower(signature), publicKey, message) % L;
auto p = scalarMultiplyB(modL(S));
auto q = edwardsAdd(R(), scalarMultiply(modL(r), A()));
auto p = scalarMultiply(B, S);
auto q = edwardsAdd(R(), scalarMultiply(A(), r));
if(!onCurve(p) || !onCurve(q)) return false;
if(p.x * q.z - q.x * p.z) return false;
if(p.y * q.z - q.y * p.z) return false;
@@ -71,31 +49,46 @@ struct Ed25519 {
private:
using field = Modulo25519;
struct point { field x, y, z, t; };
point Bscalar[253];
const field D = -field(121665) * field(121666).reciprocal();
const field D = -field(121665) * reciprocal(field(121666));
const point B = *decompress((field(4) * reciprocal(field(5)))());
const BarrettReduction<256> L = BarrettReduction<256>{EllipticCurve::L};
inline auto clamp(uint256_t p) const -> uint256_t {
p &= ((0_u256 - 1) >> 2) - 7;
p |= 1_u256 << 254;
return p;
inline auto input(Hash::SHA512&) const -> void {}
template<typename... P> inline auto input(Hash::SHA512& hash, uint256_t value, P&&... p) const -> void {
for(uint byte : range(32)) hash.input(uint8_t(value >> byte * 8));
input(hash, forward<P>(p)...);
}
inline auto recoverX(field y) const -> field {
field y2 = y.square();
field x = ((y2 - 1) * (D * y2 + 1).reciprocal()).squareRoot();
return x() & 1 ? -x : x;
template<typename... P> inline auto input(Hash::SHA512& hash, array_view<uint8_t> value, P&&... p) const -> void {
hash.input(value);
input(hash, forward<P>(p)...);
}
template<typename... P> inline auto hash(P&&... p) const -> uint512_t {
Hash::SHA512 hash;
input(hash, forward<P>(p)...);
uint512_t result;
for(auto byte : reverse(hash.output())) result = result << 8 | byte;
return result;
}
inline auto clamp(uint256_t p) const -> uint256_t {
p &= (1_u256 << 254) - 8;
p |= (1_u256 << 254);
return p;
}
inline auto onCurve(point p) const -> bool {
if(!p.z) return false;
if(p.x * p.y != p.z * p.t) return false;
if(p.y.square() - p.x.square() - p.z.square() - p.t.square() * D) return false;
if(p.x * p.y - p.z * p.t) return false;
if(square(p.y) - square(p.x) - square(p.z) - square(p.t) * D) return false;
return true;
}
inline auto decompress(uint256_t c) const -> maybe<point> {
field y = c & (1_u256 << 255) - 1;
field x = recoverX(y);
field y = c & ~0_u256 >> 1;
field x = squareRoot((square(y) - 1) * reciprocal(D * square(y) + 1));
if(c >> 255) x = -x;
point p{x, y, 1, x * y};
if(!onCurve(p)) return nothing;
@@ -103,18 +96,18 @@ private:
}
inline auto compress(point p) const -> uint256_t {
field r = p.z.reciprocal();
field r = reciprocal(p.z);
field x = p.x * r;
field y = p.y * r;
return (x() & 1) << 255 | (y() & ((0_u256 - 1) >> 1));
return (x & 1) << 255 | (y & ~0_u256 >> 1);
}
inline auto edwardsDouble(point p) const -> point {
field a = p.x.square();
field b = p.y.square();
field c = p.z.square();
field a = square(p.x);
field b = square(p.y);
field c = square(p.z);
field d = -a;
field e = (p.x + p.y).square() - a - b;
field e = square(p.x + p.y) - a - b;
field g = d + b;
field f = g - (c + c);
field h = d - b;
@@ -133,29 +126,16 @@ private:
return {e * f, g * h, f * g, e * h};
}
inline auto scalarMultiply(uint512_t e, point q) const -> point {
inline auto scalarMultiply(point q, uint256_t exponent) const -> point {
point p{0, 1, 1, 0}, c;
for(uint n : reverse(range(253))) {
for(uint bit : reverse(range(253))) {
p = edwardsDouble(p);
c = edwardsAdd(p, q);
bool bit = e >> n & 1;
cmove(bit, p.x, c.x);
cmove(bit, p.y, c.y);
cmove(bit, p.z, c.z);
cmove(bit, p.t, c.t);
}
return p;
}
inline auto scalarMultiplyB(uint512_t e) const -> point {
point p{0, 1, 1, 0}, c;
for(uint n : reverse(range(253))) {
bool bit = e >> n & 1;
c = edwardsAdd(p, Bscalar[n]);
cmove(bit, p.x, c.x);
cmove(bit, p.y, c.y);
cmove(bit, p.z, c.z);
cmove(bit, p.t, c.t);
bool condition = exponent >> bit & 1;
cmove(condition, p.x, c.x);
cmove(condition, p.y, c.y);
cmove(condition, p.z, c.z);
cmove(condition, p.t, c.t);
}
return p;
}