diff options
Diffstat (limited to 'libgo/go/golang.org/x/crypto')
19 files changed, 1451 insertions, 5 deletions
diff --git a/libgo/go/golang.org/x/crypto/chacha20poly1305/chacha20poly1305.go b/libgo/go/golang.org/x/crypto/chacha20poly1305/chacha20poly1305.go index 0d7bac3..93da732 100644 --- a/libgo/go/golang.org/x/crypto/chacha20poly1305/chacha20poly1305.go +++ b/libgo/go/golang.org/x/crypto/chacha20poly1305/chacha20poly1305.go @@ -26,6 +26,10 @@ const ( // NonceSizeX is the size of the nonce used with the XChaCha20-Poly1305 // variant of this AEAD, in bytes. NonceSizeX = 24 + + // Overhead is the size of the Poly1305 authentication tag, and the + // difference between a ciphertext length and its plaintext. + Overhead = 16 ) type chacha20poly1305 struct { @@ -47,7 +51,7 @@ func (c *chacha20poly1305) NonceSize() int { } func (c *chacha20poly1305) Overhead() int { - return 16 + return Overhead } func (c *chacha20poly1305) Seal(dst, nonce, plaintext, additionalData []byte) []byte { diff --git a/libgo/go/golang.org/x/crypto/chacha20poly1305/chacha20poly1305_generic.go b/libgo/go/golang.org/x/crypto/chacha20poly1305/chacha20poly1305_generic.go index fe191d3..96b2fd8 100644 --- a/libgo/go/golang.org/x/crypto/chacha20poly1305/chacha20poly1305_generic.go +++ b/libgo/go/golang.org/x/crypto/chacha20poly1305/chacha20poly1305_generic.go @@ -8,8 +8,8 @@ import ( "encoding/binary" "golang.org/x/crypto/chacha20" + "golang.org/x/crypto/internal/poly1305" "golang.org/x/crypto/internal/subtle" - "golang.org/x/crypto/poly1305" ) func writeWithPadding(p *poly1305.MAC, b []byte) { diff --git a/libgo/go/golang.org/x/crypto/chacha20poly1305/xchacha20poly1305.go b/libgo/go/golang.org/x/crypto/chacha20poly1305/xchacha20poly1305.go index d9d46b9..1cebfe94 100644 --- a/libgo/go/golang.org/x/crypto/chacha20poly1305/xchacha20poly1305.go +++ b/libgo/go/golang.org/x/crypto/chacha20poly1305/xchacha20poly1305.go @@ -35,7 +35,7 @@ func (*xchacha20poly1305) NonceSize() int { } func (*xchacha20poly1305) Overhead() int { - return 16 + return Overhead } func (x *xchacha20poly1305) Seal(dst, nonce, plaintext, additionalData []byte) []byte { diff --git a/libgo/go/golang.org/x/crypto/cryptobyte/asn1.go b/libgo/go/golang.org/x/crypto/cryptobyte/asn1.go index 83c776d..3a1674a 100644 --- a/libgo/go/golang.org/x/crypto/cryptobyte/asn1.go +++ b/libgo/go/golang.org/x/crypto/cryptobyte/asn1.go @@ -407,7 +407,12 @@ func (s *String) ReadASN1Enum(out *int) bool { func (s *String) readBase128Int(out *int) bool { ret := 0 for i := 0; len(*s) > 0; i++ { - if i == 4 { + if i == 5 { + return false + } + // Avoid overflowing int on a 32-bit platform. + // We don't want different behavior based on the architecture. + if ret >= 1<<(31-7) { return false } ret <<= 7 diff --git a/libgo/go/golang.org/x/crypto/curve25519/curve25519.go b/libgo/go/golang.org/x/crypto/curve25519/curve25519.go index 4b9a655..cda3fdd 100644 --- a/libgo/go/golang.org/x/crypto/curve25519/curve25519.go +++ b/libgo/go/golang.org/x/crypto/curve25519/curve25519.go @@ -10,6 +10,8 @@ package curve25519 // import "golang.org/x/crypto/curve25519" import ( "crypto/subtle" "fmt" + + "golang.org/x/crypto/curve25519/internal/field" ) // ScalarMult sets dst to the product scalar * point. @@ -18,7 +20,55 @@ import ( // zeroes, irrespective of the scalar. Instead, use the X25519 function, which // will return an error. func ScalarMult(dst, scalar, point *[32]byte) { - scalarMult(dst, scalar, point) + var e [32]byte + + copy(e[:], scalar[:]) + e[0] &= 248 + e[31] &= 127 + e[31] |= 64 + + var x1, x2, z2, x3, z3, tmp0, tmp1 field.Element + x1.SetBytes(point[:]) + x2.One() + x3.Set(&x1) + z3.One() + + swap := 0 + for pos := 254; pos >= 0; pos-- { + b := e[pos/8] >> uint(pos&7) + b &= 1 + swap ^= int(b) + x2.Swap(&x3, swap) + z2.Swap(&z3, swap) + swap = int(b) + + tmp0.Subtract(&x3, &z3) + tmp1.Subtract(&x2, &z2) + x2.Add(&x2, &z2) + z2.Add(&x3, &z3) + z3.Multiply(&tmp0, &x2) + z2.Multiply(&z2, &tmp1) + tmp0.Square(&tmp1) + tmp1.Square(&x2) + x3.Add(&z3, &z2) + z2.Subtract(&z3, &z2) + x2.Multiply(&tmp1, &tmp0) + tmp1.Subtract(&tmp1, &tmp0) + z2.Square(&z2) + + z3.Mult32(&tmp1, 121666) + x3.Square(&x3) + tmp0.Add(&tmp0, &z3) + z3.Multiply(&x1, &z2) + z2.Multiply(&tmp1, &tmp0) + } + + x2.Swap(&x3, swap) + z2.Swap(&z3, swap) + + z2.Invert(&z2) + x2.Multiply(&x2, &z2) + copy(dst[:], x2.Bytes()) } // ScalarBaseMult sets dst to the product scalar * base where base is the diff --git a/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe.go b/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe.go new file mode 100644 index 0000000..ca841ad --- /dev/null +++ b/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe.go @@ -0,0 +1,416 @@ +// Copyright (c) 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package field implements fast arithmetic modulo 2^255-19. +package field + +import ( + "crypto/subtle" + "encoding/binary" + "math/bits" +) + +// Element represents an element of the field GF(2^255-19). Note that this +// is not a cryptographically secure group, and should only be used to interact +// with edwards25519.Point coordinates. +// +// This type works similarly to math/big.Int, and all arguments and receivers +// are allowed to alias. +// +// The zero value is a valid zero element. +type Element struct { + // An element t represents the integer + // t.l0 + t.l1*2^51 + t.l2*2^102 + t.l3*2^153 + t.l4*2^204 + // + // Between operations, all limbs are expected to be lower than 2^52. + l0 uint64 + l1 uint64 + l2 uint64 + l3 uint64 + l4 uint64 +} + +const maskLow51Bits uint64 = (1 << 51) - 1 + +var feZero = &Element{0, 0, 0, 0, 0} + +// Zero sets v = 0, and returns v. +func (v *Element) Zero() *Element { + *v = *feZero + return v +} + +var feOne = &Element{1, 0, 0, 0, 0} + +// One sets v = 1, and returns v. +func (v *Element) One() *Element { + *v = *feOne + return v +} + +// reduce reduces v modulo 2^255 - 19 and returns it. +func (v *Element) reduce() *Element { + v.carryPropagate() + + // After the light reduction we now have a field element representation + // v < 2^255 + 2^13 * 19, but need v < 2^255 - 19. + + // If v >= 2^255 - 19, then v + 19 >= 2^255, which would overflow 2^255 - 1, + // generating a carry. That is, c will be 0 if v < 2^255 - 19, and 1 otherwise. + c := (v.l0 + 19) >> 51 + c = (v.l1 + c) >> 51 + c = (v.l2 + c) >> 51 + c = (v.l3 + c) >> 51 + c = (v.l4 + c) >> 51 + + // If v < 2^255 - 19 and c = 0, this will be a no-op. Otherwise, it's + // effectively applying the reduction identity to the carry. + v.l0 += 19 * c + + v.l1 += v.l0 >> 51 + v.l0 = v.l0 & maskLow51Bits + v.l2 += v.l1 >> 51 + v.l1 = v.l1 & maskLow51Bits + v.l3 += v.l2 >> 51 + v.l2 = v.l2 & maskLow51Bits + v.l4 += v.l3 >> 51 + v.l3 = v.l3 & maskLow51Bits + // no additional carry + v.l4 = v.l4 & maskLow51Bits + + return v +} + +// Add sets v = a + b, and returns v. +func (v *Element) Add(a, b *Element) *Element { + v.l0 = a.l0 + b.l0 + v.l1 = a.l1 + b.l1 + v.l2 = a.l2 + b.l2 + v.l3 = a.l3 + b.l3 + v.l4 = a.l4 + b.l4 + // Using the generic implementation here is actually faster than the + // assembly. Probably because the body of this function is so simple that + // the compiler can figure out better optimizations by inlining the carry + // propagation. TODO + return v.carryPropagateGeneric() +} + +// Subtract sets v = a - b, and returns v. +func (v *Element) Subtract(a, b *Element) *Element { + // We first add 2 * p, to guarantee the subtraction won't underflow, and + // then subtract b (which can be up to 2^255 + 2^13 * 19). + v.l0 = (a.l0 + 0xFFFFFFFFFFFDA) - b.l0 + v.l1 = (a.l1 + 0xFFFFFFFFFFFFE) - b.l1 + v.l2 = (a.l2 + 0xFFFFFFFFFFFFE) - b.l2 + v.l3 = (a.l3 + 0xFFFFFFFFFFFFE) - b.l3 + v.l4 = (a.l4 + 0xFFFFFFFFFFFFE) - b.l4 + return v.carryPropagate() +} + +// Negate sets v = -a, and returns v. +func (v *Element) Negate(a *Element) *Element { + return v.Subtract(feZero, a) +} + +// Invert sets v = 1/z mod p, and returns v. +// +// If z == 0, Invert returns v = 0. +func (v *Element) Invert(z *Element) *Element { + // Inversion is implemented as exponentiation with exponent p − 2. It uses the + // same sequence of 255 squarings and 11 multiplications as [Curve25519]. + var z2, z9, z11, z2_5_0, z2_10_0, z2_20_0, z2_50_0, z2_100_0, t Element + + z2.Square(z) // 2 + t.Square(&z2) // 4 + t.Square(&t) // 8 + z9.Multiply(&t, z) // 9 + z11.Multiply(&z9, &z2) // 11 + t.Square(&z11) // 22 + z2_5_0.Multiply(&t, &z9) // 31 = 2^5 - 2^0 + + t.Square(&z2_5_0) // 2^6 - 2^1 + for i := 0; i < 4; i++ { + t.Square(&t) // 2^10 - 2^5 + } + z2_10_0.Multiply(&t, &z2_5_0) // 2^10 - 2^0 + + t.Square(&z2_10_0) // 2^11 - 2^1 + for i := 0; i < 9; i++ { + t.Square(&t) // 2^20 - 2^10 + } + z2_20_0.Multiply(&t, &z2_10_0) // 2^20 - 2^0 + + t.Square(&z2_20_0) // 2^21 - 2^1 + for i := 0; i < 19; i++ { + t.Square(&t) // 2^40 - 2^20 + } + t.Multiply(&t, &z2_20_0) // 2^40 - 2^0 + + t.Square(&t) // 2^41 - 2^1 + for i := 0; i < 9; i++ { + t.Square(&t) // 2^50 - 2^10 + } + z2_50_0.Multiply(&t, &z2_10_0) // 2^50 - 2^0 + + t.Square(&z2_50_0) // 2^51 - 2^1 + for i := 0; i < 49; i++ { + t.Square(&t) // 2^100 - 2^50 + } + z2_100_0.Multiply(&t, &z2_50_0) // 2^100 - 2^0 + + t.Square(&z2_100_0) // 2^101 - 2^1 + for i := 0; i < 99; i++ { + t.Square(&t) // 2^200 - 2^100 + } + t.Multiply(&t, &z2_100_0) // 2^200 - 2^0 + + t.Square(&t) // 2^201 - 2^1 + for i := 0; i < 49; i++ { + t.Square(&t) // 2^250 - 2^50 + } + t.Multiply(&t, &z2_50_0) // 2^250 - 2^0 + + t.Square(&t) // 2^251 - 2^1 + t.Square(&t) // 2^252 - 2^2 + t.Square(&t) // 2^253 - 2^3 + t.Square(&t) // 2^254 - 2^4 + t.Square(&t) // 2^255 - 2^5 + + return v.Multiply(&t, &z11) // 2^255 - 21 +} + +// Set sets v = a, and returns v. +func (v *Element) Set(a *Element) *Element { + *v = *a + return v +} + +// SetBytes sets v to x, which must be a 32-byte little-endian encoding. +// +// Consistent with RFC 7748, the most significant bit (the high bit of the +// last byte) is ignored, and non-canonical values (2^255-19 through 2^255-1) +// are accepted. Note that this is laxer than specified by RFC 8032. +func (v *Element) SetBytes(x []byte) *Element { + if len(x) != 32 { + panic("edwards25519: invalid field element input size") + } + + // Bits 0:51 (bytes 0:8, bits 0:64, shift 0, mask 51). + v.l0 = binary.LittleEndian.Uint64(x[0:8]) + v.l0 &= maskLow51Bits + // Bits 51:102 (bytes 6:14, bits 48:112, shift 3, mask 51). + v.l1 = binary.LittleEndian.Uint64(x[6:14]) >> 3 + v.l1 &= maskLow51Bits + // Bits 102:153 (bytes 12:20, bits 96:160, shift 6, mask 51). + v.l2 = binary.LittleEndian.Uint64(x[12:20]) >> 6 + v.l2 &= maskLow51Bits + // Bits 153:204 (bytes 19:27, bits 152:216, shift 1, mask 51). + v.l3 = binary.LittleEndian.Uint64(x[19:27]) >> 1 + v.l3 &= maskLow51Bits + // Bits 204:251 (bytes 24:32, bits 192:256, shift 12, mask 51). + // Note: not bytes 25:33, shift 4, to avoid overread. + v.l4 = binary.LittleEndian.Uint64(x[24:32]) >> 12 + v.l4 &= maskLow51Bits + + return v +} + +// Bytes returns the canonical 32-byte little-endian encoding of v. +func (v *Element) Bytes() []byte { + // This function is outlined to make the allocations inline in the caller + // rather than happen on the heap. + var out [32]byte + return v.bytes(&out) +} + +func (v *Element) bytes(out *[32]byte) []byte { + t := *v + t.reduce() + + var buf [8]byte + for i, l := range [5]uint64{t.l0, t.l1, t.l2, t.l3, t.l4} { + bitsOffset := i * 51 + binary.LittleEndian.PutUint64(buf[:], l<<uint(bitsOffset%8)) + for i, bb := range buf { + off := bitsOffset/8 + i + if off >= len(out) { + break + } + out[off] |= bb + } + } + + return out[:] +} + +// Equal returns 1 if v and u are equal, and 0 otherwise. +func (v *Element) Equal(u *Element) int { + sa, sv := u.Bytes(), v.Bytes() + return subtle.ConstantTimeCompare(sa, sv) +} + +// mask64Bits returns 0xffffffff if cond is 1, and 0 otherwise. +func mask64Bits(cond int) uint64 { return ^(uint64(cond) - 1) } + +// Select sets v to a if cond == 1, and to b if cond == 0. +func (v *Element) Select(a, b *Element, cond int) *Element { + m := mask64Bits(cond) + v.l0 = (m & a.l0) | (^m & b.l0) + v.l1 = (m & a.l1) | (^m & b.l1) + v.l2 = (m & a.l2) | (^m & b.l2) + v.l3 = (m & a.l3) | (^m & b.l3) + v.l4 = (m & a.l4) | (^m & b.l4) + return v +} + +// Swap swaps v and u if cond == 1 or leaves them unchanged if cond == 0, and returns v. +func (v *Element) Swap(u *Element, cond int) { + m := mask64Bits(cond) + t := m & (v.l0 ^ u.l0) + v.l0 ^= t + u.l0 ^= t + t = m & (v.l1 ^ u.l1) + v.l1 ^= t + u.l1 ^= t + t = m & (v.l2 ^ u.l2) + v.l2 ^= t + u.l2 ^= t + t = m & (v.l3 ^ u.l3) + v.l3 ^= t + u.l3 ^= t + t = m & (v.l4 ^ u.l4) + v.l4 ^= t + u.l4 ^= t +} + +// IsNegative returns 1 if v is negative, and 0 otherwise. +func (v *Element) IsNegative() int { + return int(v.Bytes()[0] & 1) +} + +// Absolute sets v to |u|, and returns v. +func (v *Element) Absolute(u *Element) *Element { + return v.Select(new(Element).Negate(u), u, u.IsNegative()) +} + +// Multiply sets v = x * y, and returns v. +func (v *Element) Multiply(x, y *Element) *Element { + feMul(v, x, y) + return v +} + +// Square sets v = x * x, and returns v. +func (v *Element) Square(x *Element) *Element { + feSquare(v, x) + return v +} + +// Mult32 sets v = x * y, and returns v. +func (v *Element) Mult32(x *Element, y uint32) *Element { + x0lo, x0hi := mul51(x.l0, y) + x1lo, x1hi := mul51(x.l1, y) + x2lo, x2hi := mul51(x.l2, y) + x3lo, x3hi := mul51(x.l3, y) + x4lo, x4hi := mul51(x.l4, y) + v.l0 = x0lo + 19*x4hi // carried over per the reduction identity + v.l1 = x1lo + x0hi + v.l2 = x2lo + x1hi + v.l3 = x3lo + x2hi + v.l4 = x4lo + x3hi + // The hi portions are going to be only 32 bits, plus any previous excess, + // so we can skip the carry propagation. + return v +} + +// mul51 returns lo + hi * 2⁵¹ = a * b. +func mul51(a uint64, b uint32) (lo uint64, hi uint64) { + mh, ml := bits.Mul64(a, uint64(b)) + lo = ml & maskLow51Bits + hi = (mh << 13) | (ml >> 51) + return +} + +// Pow22523 set v = x^((p-5)/8), and returns v. (p-5)/8 is 2^252-3. +func (v *Element) Pow22523(x *Element) *Element { + var t0, t1, t2 Element + + t0.Square(x) // x^2 + t1.Square(&t0) // x^4 + t1.Square(&t1) // x^8 + t1.Multiply(x, &t1) // x^9 + t0.Multiply(&t0, &t1) // x^11 + t0.Square(&t0) // x^22 + t0.Multiply(&t1, &t0) // x^31 + t1.Square(&t0) // x^62 + for i := 1; i < 5; i++ { // x^992 + t1.Square(&t1) + } + t0.Multiply(&t1, &t0) // x^1023 -> 1023 = 2^10 - 1 + t1.Square(&t0) // 2^11 - 2 + for i := 1; i < 10; i++ { // 2^20 - 2^10 + t1.Square(&t1) + } + t1.Multiply(&t1, &t0) // 2^20 - 1 + t2.Square(&t1) // 2^21 - 2 + for i := 1; i < 20; i++ { // 2^40 - 2^20 + t2.Square(&t2) + } + t1.Multiply(&t2, &t1) // 2^40 - 1 + t1.Square(&t1) // 2^41 - 2 + for i := 1; i < 10; i++ { // 2^50 - 2^10 + t1.Square(&t1) + } + t0.Multiply(&t1, &t0) // 2^50 - 1 + t1.Square(&t0) // 2^51 - 2 + for i := 1; i < 50; i++ { // 2^100 - 2^50 + t1.Square(&t1) + } + t1.Multiply(&t1, &t0) // 2^100 - 1 + t2.Square(&t1) // 2^101 - 2 + for i := 1; i < 100; i++ { // 2^200 - 2^100 + t2.Square(&t2) + } + t1.Multiply(&t2, &t1) // 2^200 - 1 + t1.Square(&t1) // 2^201 - 2 + for i := 1; i < 50; i++ { // 2^250 - 2^50 + t1.Square(&t1) + } + t0.Multiply(&t1, &t0) // 2^250 - 1 + t0.Square(&t0) // 2^251 - 2 + t0.Square(&t0) // 2^252 - 4 + return v.Multiply(&t0, x) // 2^252 - 3 -> x^(2^252-3) +} + +// sqrtM1 is 2^((p-1)/4), which squared is equal to -1 by Euler's Criterion. +var sqrtM1 = &Element{1718705420411056, 234908883556509, + 2233514472574048, 2117202627021982, 765476049583133} + +// SqrtRatio sets r to the non-negative square root of the ratio of u and v. +// +// If u/v is square, SqrtRatio returns r and 1. If u/v is not square, SqrtRatio +// sets r according to Section 4.3 of draft-irtf-cfrg-ristretto255-decaf448-00, +// and returns r and 0. +func (r *Element) SqrtRatio(u, v *Element) (rr *Element, wasSquare int) { + var a, b Element + + // r = (u * v3) * (u * v7)^((p-5)/8) + v2 := a.Square(v) + uv3 := b.Multiply(u, b.Multiply(v2, v)) + uv7 := a.Multiply(uv3, a.Square(v2)) + r.Multiply(uv3, r.Pow22523(uv7)) + + check := a.Multiply(v, a.Square(r)) // check = v * r^2 + + uNeg := b.Negate(u) + correctSignSqrt := check.Equal(u) + flippedSignSqrt := check.Equal(uNeg) + flippedSignSqrtI := check.Equal(uNeg.Multiply(uNeg, sqrtM1)) + + rPrime := b.Multiply(r, sqrtM1) // r_prime = SQRT_M1 * r + // r = CT_SELECT(r_prime IF flipped_sign_sqrt | flipped_sign_sqrt_i ELSE r) + r.Select(rPrime, r, flippedSignSqrt|flippedSignSqrtI) + + r.Absolute(r) // Choose the nonnegative square root. + return r, correctSignSqrt | flippedSignSqrt +} diff --git a/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe_amd64.go b/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe_amd64.go new file mode 100644 index 0000000..8fe5839 --- /dev/null +++ b/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe_amd64.go @@ -0,0 +1,14 @@ +// Code generated by command: go run fe_amd64_asm.go -out ../fe_amd64.s -stubs ../fe_amd64.go -pkg field. DO NOT EDIT. + +//go:build amd64 && gc && !purego +// +build amd64,gc,!purego + +package field + +// feMul sets out = a * b. It works like feMulGeneric. +//go:noescape +func feMul(out *Element, a *Element, b *Element) + +// feSquare sets out = a * a. It works like feSquareGeneric. +//go:noescape +func feSquare(out *Element, a *Element) diff --git a/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe_amd64_noasm.go b/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe_amd64_noasm.go new file mode 100644 index 0000000..ddb6c9b --- /dev/null +++ b/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe_amd64_noasm.go @@ -0,0 +1,12 @@ +// Copyright (c) 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !amd64 || !gc || purego +// +build !amd64 !gc purego + +package field + +func feMul(v, x, y *Element) { feMulGeneric(v, x, y) } + +func feSquare(v, x *Element) { feSquareGeneric(v, x) } diff --git a/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe_arm64.go b/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe_arm64.go new file mode 100644 index 0000000..af459ef --- /dev/null +++ b/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe_arm64.go @@ -0,0 +1,16 @@ +// Copyright (c) 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build arm64 && gc && !purego +// +build arm64,gc,!purego + +package field + +//go:noescape +func carryPropagate(v *Element) + +func (v *Element) carryPropagate() *Element { + carryPropagate(v) + return v +} diff --git a/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe_arm64_noasm.go b/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe_arm64_noasm.go new file mode 100644 index 0000000..234a5b2 --- /dev/null +++ b/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe_arm64_noasm.go @@ -0,0 +1,12 @@ +// Copyright (c) 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !arm64 || !gc || purego +// +build !arm64 !gc purego + +package field + +func (v *Element) carryPropagate() *Element { + return v.carryPropagateGeneric() +} diff --git a/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe_generic.go b/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe_generic.go new file mode 100644 index 0000000..7b5b78c --- /dev/null +++ b/libgo/go/golang.org/x/crypto/curve25519/internal/field/fe_generic.go @@ -0,0 +1,264 @@ +// Copyright (c) 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package field + +import "math/bits" + +// uint128 holds a 128-bit number as two 64-bit limbs, for use with the +// bits.Mul64 and bits.Add64 intrinsics. +type uint128 struct { + lo, hi uint64 +} + +// mul64 returns a * b. +func mul64(a, b uint64) uint128 { + hi, lo := bits.Mul64(a, b) + return uint128{lo, hi} +} + +// addMul64 returns v + a * b. +func addMul64(v uint128, a, b uint64) uint128 { + hi, lo := bits.Mul64(a, b) + lo, c := bits.Add64(lo, v.lo, 0) + hi, _ = bits.Add64(hi, v.hi, c) + return uint128{lo, hi} +} + +// shiftRightBy51 returns a >> 51. a is assumed to be at most 115 bits. +func shiftRightBy51(a uint128) uint64 { + return (a.hi << (64 - 51)) | (a.lo >> 51) +} + +func feMulGeneric(v, a, b *Element) { + a0 := a.l0 + a1 := a.l1 + a2 := a.l2 + a3 := a.l3 + a4 := a.l4 + + b0 := b.l0 + b1 := b.l1 + b2 := b.l2 + b3 := b.l3 + b4 := b.l4 + + // Limb multiplication works like pen-and-paper columnar multiplication, but + // with 51-bit limbs instead of digits. + // + // a4 a3 a2 a1 a0 x + // b4 b3 b2 b1 b0 = + // ------------------------ + // a4b0 a3b0 a2b0 a1b0 a0b0 + + // a4b1 a3b1 a2b1 a1b1 a0b1 + + // a4b2 a3b2 a2b2 a1b2 a0b2 + + // a4b3 a3b3 a2b3 a1b3 a0b3 + + // a4b4 a3b4 a2b4 a1b4 a0b4 = + // ---------------------------------------------- + // r8 r7 r6 r5 r4 r3 r2 r1 r0 + // + // We can then use the reduction identity (a * 2²⁵⁵ + b = a * 19 + b) to + // reduce the limbs that would overflow 255 bits. r5 * 2²⁵⁵ becomes 19 * r5, + // r6 * 2³⁰⁶ becomes 19 * r6 * 2⁵¹, etc. + // + // Reduction can be carried out simultaneously to multiplication. For + // example, we do not compute r5: whenever the result of a multiplication + // belongs to r5, like a1b4, we multiply it by 19 and add the result to r0. + // + // a4b0 a3b0 a2b0 a1b0 a0b0 + + // a3b1 a2b1 a1b1 a0b1 19×a4b1 + + // a2b2 a1b2 a0b2 19×a4b2 19×a3b2 + + // a1b3 a0b3 19×a4b3 19×a3b3 19×a2b3 + + // a0b4 19×a4b4 19×a3b4 19×a2b4 19×a1b4 = + // -------------------------------------- + // r4 r3 r2 r1 r0 + // + // Finally we add up the columns into wide, overlapping limbs. + + a1_19 := a1 * 19 + a2_19 := a2 * 19 + a3_19 := a3 * 19 + a4_19 := a4 * 19 + + // r0 = a0×b0 + 19×(a1×b4 + a2×b3 + a3×b2 + a4×b1) + r0 := mul64(a0, b0) + r0 = addMul64(r0, a1_19, b4) + r0 = addMul64(r0, a2_19, b3) + r0 = addMul64(r0, a3_19, b2) + r0 = addMul64(r0, a4_19, b1) + + // r1 = a0×b1 + a1×b0 + 19×(a2×b4 + a3×b3 + a4×b2) + r1 := mul64(a0, b1) + r1 = addMul64(r1, a1, b0) + r1 = addMul64(r1, a2_19, b4) + r1 = addMul64(r1, a3_19, b3) + r1 = addMul64(r1, a4_19, b2) + + // r2 = a0×b2 + a1×b1 + a2×b0 + 19×(a3×b4 + a4×b3) + r2 := mul64(a0, b2) + r2 = addMul64(r2, a1, b1) + r2 = addMul64(r2, a2, b0) + r2 = addMul64(r2, a3_19, b4) + r2 = addMul64(r2, a4_19, b3) + + // r3 = a0×b3 + a1×b2 + a2×b1 + a3×b0 + 19×a4×b4 + r3 := mul64(a0, b3) + r3 = addMul64(r3, a1, b2) + r3 = addMul64(r3, a2, b1) + r3 = addMul64(r3, a3, b0) + r3 = addMul64(r3, a4_19, b4) + + // r4 = a0×b4 + a1×b3 + a2×b2 + a3×b1 + a4×b0 + r4 := mul64(a0, b4) + r4 = addMul64(r4, a1, b3) + r4 = addMul64(r4, a2, b2) + r4 = addMul64(r4, a3, b1) + r4 = addMul64(r4, a4, b0) + + // After the multiplication, we need to reduce (carry) the five coefficients + // to obtain a result with limbs that are at most slightly larger than 2⁵¹, + // to respect the Element invariant. + // + // Overall, the reduction works the same as carryPropagate, except with + // wider inputs: we take the carry for each coefficient by shifting it right + // by 51, and add it to the limb above it. The top carry is multiplied by 19 + // according to the reduction identity and added to the lowest limb. + // + // The largest coefficient (r0) will be at most 111 bits, which guarantees + // that all carries are at most 111 - 51 = 60 bits, which fits in a uint64. + // + // r0 = a0×b0 + 19×(a1×b4 + a2×b3 + a3×b2 + a4×b1) + // r0 < 2⁵²×2⁵² + 19×(2⁵²×2⁵² + 2⁵²×2⁵² + 2⁵²×2⁵² + 2⁵²×2⁵²) + // r0 < (1 + 19 × 4) × 2⁵² × 2⁵² + // r0 < 2⁷ × 2⁵² × 2⁵² + // r0 < 2¹¹¹ + // + // Moreover, the top coefficient (r4) is at most 107 bits, so c4 is at most + // 56 bits, and c4 * 19 is at most 61 bits, which again fits in a uint64 and + // allows us to easily apply the reduction identity. + // + // r4 = a0×b4 + a1×b3 + a2×b2 + a3×b1 + a4×b0 + // r4 < 5 × 2⁵² × 2⁵² + // r4 < 2¹⁰⁷ + // + + c0 := shiftRightBy51(r0) + c1 := shiftRightBy51(r1) + c2 := shiftRightBy51(r2) + c3 := shiftRightBy51(r3) + c4 := shiftRightBy51(r4) + + rr0 := r0.lo&maskLow51Bits + c4*19 + rr1 := r1.lo&maskLow51Bits + c0 + rr2 := r2.lo&maskLow51Bits + c1 + rr3 := r3.lo&maskLow51Bits + c2 + rr4 := r4.lo&maskLow51Bits + c3 + + // Now all coefficients fit into 64-bit registers but are still too large to + // be passed around as a Element. We therefore do one last carry chain, + // where the carries will be small enough to fit in the wiggle room above 2⁵¹. + *v = Element{rr0, rr1, rr2, rr3, rr4} + v.carryPropagate() +} + +func feSquareGeneric(v, a *Element) { + l0 := a.l0 + l1 := a.l1 + l2 := a.l2 + l3 := a.l3 + l4 := a.l4 + + // Squaring works precisely like multiplication above, but thanks to its + // symmetry we get to group a few terms together. + // + // l4 l3 l2 l1 l0 x + // l4 l3 l2 l1 l0 = + // ------------------------ + // l4l0 l3l0 l2l0 l1l0 l0l0 + + // l4l1 l3l1 l2l1 l1l1 l0l1 + + // l4l2 l3l2 l2l2 l1l2 l0l2 + + // l4l3 l3l3 l2l3 l1l3 l0l3 + + // l4l4 l3l4 l2l4 l1l4 l0l4 = + // ---------------------------------------------- + // r8 r7 r6 r5 r4 r3 r2 r1 r0 + // + // l4l0 l3l0 l2l0 l1l0 l0l0 + + // l3l1 l2l1 l1l1 l0l1 19×l4l1 + + // l2l2 l1l2 l0l2 19×l4l2 19×l3l2 + + // l1l3 l0l3 19×l4l3 19×l3l3 19×l2l3 + + // l0l4 19×l4l4 19×l3l4 19×l2l4 19×l1l4 = + // -------------------------------------- + // r4 r3 r2 r1 r0 + // + // With precomputed 2×, 19×, and 2×19× terms, we can compute each limb with + // only three Mul64 and four Add64, instead of five and eight. + + l0_2 := l0 * 2 + l1_2 := l1 * 2 + + l1_38 := l1 * 38 + l2_38 := l2 * 38 + l3_38 := l3 * 38 + + l3_19 := l3 * 19 + l4_19 := l4 * 19 + + // r0 = l0×l0 + 19×(l1×l4 + l2×l3 + l3×l2 + l4×l1) = l0×l0 + 19×2×(l1×l4 + l2×l3) + r0 := mul64(l0, l0) + r0 = addMul64(r0, l1_38, l4) + r0 = addMul64(r0, l2_38, l3) + + // r1 = l0×l1 + l1×l0 + 19×(l2×l4 + l3×l3 + l4×l2) = 2×l0×l1 + 19×2×l2×l4 + 19×l3×l3 + r1 := mul64(l0_2, l1) + r1 = addMul64(r1, l2_38, l4) + r1 = addMul64(r1, l3_19, l3) + + // r2 = l0×l2 + l1×l1 + l2×l0 + 19×(l3×l4 + l4×l3) = 2×l0×l2 + l1×l1 + 19×2×l3×l4 + r2 := mul64(l0_2, l2) + r2 = addMul64(r2, l1, l1) + r2 = addMul64(r2, l3_38, l4) + + // r3 = l0×l3 + l1×l2 + l2×l1 + l3×l0 + 19×l4×l4 = 2×l0×l3 + 2×l1×l2 + 19×l4×l4 + r3 := mul64(l0_2, l3) + r3 = addMul64(r3, l1_2, l2) + r3 = addMul64(r3, l4_19, l4) + + // r4 = l0×l4 + l1×l3 + l2×l2 + l3×l1 + l4×l0 = 2×l0×l4 + 2×l1×l3 + l2×l2 + r4 := mul64(l0_2, l4) + r4 = addMul64(r4, l1_2, l3) + r4 = addMul64(r4, l2, l2) + + c0 := shiftRightBy51(r0) + c1 := shiftRightBy51(r1) + c2 := shiftRightBy51(r2) + c3 := shiftRightBy51(r3) + c4 := shiftRightBy51(r4) + + rr0 := r0.lo&maskLow51Bits + c4*19 + rr1 := r1.lo&maskLow51Bits + c0 + rr2 := r2.lo&maskLow51Bits + c1 + rr3 := r3.lo&maskLow51Bits + c2 + rr4 := r4.lo&maskLow51Bits + c3 + + *v = Element{rr0, rr1, rr2, rr3, rr4} + v.carryPropagate() +} + +// carryPropagate brings the limbs below 52 bits by applying the reduction +// identity (a * 2²⁵⁵ + b = a * 19 + b) to the l4 carry. TODO inline +func (v *Element) carryPropagateGeneric() *Element { + c0 := v.l0 >> 51 + c1 := v.l1 >> 51 + c2 := v.l2 >> 51 + c3 := v.l3 >> 51 + c4 := v.l4 >> 51 + + v.l0 = v.l0&maskLow51Bits + c4*19 + v.l1 = v.l1&maskLow51Bits + c0 + v.l2 = v.l2&maskLow51Bits + c1 + v.l3 = v.l3&maskLow51Bits + c2 + v.l4 = v.l4&maskLow51Bits + c3 + + return v +} diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/bits_compat.go b/libgo/go/golang.org/x/crypto/internal/poly1305/bits_compat.go new file mode 100644 index 0000000..45b5c966 --- /dev/null +++ b/libgo/go/golang.org/x/crypto/internal/poly1305/bits_compat.go @@ -0,0 +1,40 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !go1.13 +// +build !go1.13 + +package poly1305 + +// Generic fallbacks for the math/bits intrinsics, copied from +// src/math/bits/bits.go. They were added in Go 1.12, but Add64 and Sum64 had +// variable time fallbacks until Go 1.13. + +func bitsAdd64(x, y, carry uint64) (sum, carryOut uint64) { + sum = x + y + carry + carryOut = ((x & y) | ((x | y) &^ sum)) >> 63 + return +} + +func bitsSub64(x, y, borrow uint64) (diff, borrowOut uint64) { + diff = x - y - borrow + borrowOut = ((^x & y) | (^(x ^ y) & diff)) >> 63 + return +} + +func bitsMul64(x, y uint64) (hi, lo uint64) { + const mask32 = 1<<32 - 1 + x0 := x & mask32 + x1 := x >> 32 + y0 := y & mask32 + y1 := y >> 32 + w0 := x0 * y0 + t := x1*y0 + w0>>32 + w1 := t & mask32 + w2 := t >> 32 + w1 += x0 * y1 + hi = x1*y1 + w2 + w1>>32 + lo = x * y + return +} diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/bits_go1.13.go b/libgo/go/golang.org/x/crypto/internal/poly1305/bits_go1.13.go new file mode 100644 index 0000000..ed52b34 --- /dev/null +++ b/libgo/go/golang.org/x/crypto/internal/poly1305/bits_go1.13.go @@ -0,0 +1,22 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build go1.13 +// +build go1.13 + +package poly1305 + +import "math/bits" + +func bitsAdd64(x, y, carry uint64) (sum, carryOut uint64) { + return bits.Add64(x, y, carry) +} + +func bitsSub64(x, y, borrow uint64) (diff, borrowOut uint64) { + return bits.Sub64(x, y, borrow) +} + +func bitsMul64(x, y uint64) (hi, lo uint64) { + return bits.Mul64(x, y) +} diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/mac_noasm.go b/libgo/go/golang.org/x/crypto/internal/poly1305/mac_noasm.go new file mode 100644 index 0000000..f184b67 --- /dev/null +++ b/libgo/go/golang.org/x/crypto/internal/poly1305/mac_noasm.go @@ -0,0 +1,10 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build (!amd64 && !ppc64le && !s390x) || !gc || purego +// +build !amd64,!ppc64le,!s390x !gc purego + +package poly1305 + +type mac struct{ macGeneric } diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/poly1305.go b/libgo/go/golang.org/x/crypto/internal/poly1305/poly1305.go new file mode 100644 index 0000000..4aaea81 --- /dev/null +++ b/libgo/go/golang.org/x/crypto/internal/poly1305/poly1305.go @@ -0,0 +1,99 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package poly1305 implements Poly1305 one-time message authentication code as +// specified in https://cr.yp.to/mac/poly1305-20050329.pdf. +// +// Poly1305 is a fast, one-time authentication function. It is infeasible for an +// attacker to generate an authenticator for a message without the key. However, a +// key must only be used for a single message. Authenticating two different +// messages with the same key allows an attacker to forge authenticators for other +// messages with the same key. +// +// Poly1305 was originally coupled with AES in order to make Poly1305-AES. AES was +// used with a fixed key in order to generate one-time keys from an nonce. +// However, in this package AES isn't used and the one-time key is specified +// directly. +package poly1305 + +import "crypto/subtle" + +// TagSize is the size, in bytes, of a poly1305 authenticator. +const TagSize = 16 + +// Sum generates an authenticator for msg using a one-time key and puts the +// 16-byte result into out. Authenticating two different messages with the same +// key allows an attacker to forge messages at will. +func Sum(out *[16]byte, m []byte, key *[32]byte) { + h := New(key) + h.Write(m) + h.Sum(out[:0]) +} + +// Verify returns true if mac is a valid authenticator for m with the given key. +func Verify(mac *[16]byte, m []byte, key *[32]byte) bool { + var tmp [16]byte + Sum(&tmp, m, key) + return subtle.ConstantTimeCompare(tmp[:], mac[:]) == 1 +} + +// New returns a new MAC computing an authentication +// tag of all data written to it with the given key. +// This allows writing the message progressively instead +// of passing it as a single slice. Common users should use +// the Sum function instead. +// +// The key must be unique for each message, as authenticating +// two different messages with the same key allows an attacker +// to forge messages at will. +func New(key *[32]byte) *MAC { + m := &MAC{} + initialize(key, &m.macState) + return m +} + +// MAC is an io.Writer computing an authentication tag +// of the data written to it. +// +// MAC cannot be used like common hash.Hash implementations, +// because using a poly1305 key twice breaks its security. +// Therefore writing data to a running MAC after calling +// Sum or Verify causes it to panic. +type MAC struct { + mac // platform-dependent implementation + + finalized bool +} + +// Size returns the number of bytes Sum will return. +func (h *MAC) Size() int { return TagSize } + +// Write adds more data to the running message authentication code. +// It never returns an error. +// +// It must not be called after the first call of Sum or Verify. +func (h *MAC) Write(p []byte) (n int, err error) { + if h.finalized { + panic("poly1305: write to MAC after Sum or Verify") + } + return h.mac.Write(p) +} + +// Sum computes the authenticator of all data written to the +// message authentication code. +func (h *MAC) Sum(b []byte) []byte { + var mac [TagSize]byte + h.mac.Sum(&mac) + h.finalized = true + return append(b, mac[:]...) +} + +// Verify returns whether the authenticator of all data written to +// the message authentication code matches the expected value. +func (h *MAC) Verify(expected []byte) bool { + var mac [TagSize]byte + h.mac.Sum(&mac) + h.finalized = true + return subtle.ConstantTimeCompare(expected, mac[:]) == 1 +} diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/sum_amd64.go b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_amd64.go new file mode 100644 index 0000000..6d52233 --- /dev/null +++ b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_amd64.go @@ -0,0 +1,48 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build gc && !purego +// +build gc,!purego + +package poly1305 + +//go:noescape +func update(state *macState, msg []byte) + +// mac is a wrapper for macGeneric that redirects calls that would have gone to +// updateGeneric to update. +// +// Its Write and Sum methods are otherwise identical to the macGeneric ones, but +// using function pointers would carry a major performance cost. +type mac struct{ macGeneric } + +func (h *mac) Write(p []byte) (int, error) { + nn := len(p) + if h.offset > 0 { + n := copy(h.buffer[h.offset:], p) + if h.offset+n < TagSize { + h.offset += n + return nn, nil + } + p = p[n:] + h.offset = 0 + update(&h.macState, h.buffer[:]) + } + if n := len(p) - (len(p) % TagSize); n > 0 { + update(&h.macState, p[:n]) + p = p[n:] + } + if len(p) > 0 { + h.offset += copy(h.buffer[h.offset:], p) + } + return nn, nil +} + +func (h *mac) Sum(out *[16]byte) { + state := h.macState + if h.offset > 0 { + update(&state, h.buffer[:h.offset]) + } + finalize(out, &state.h, &state.s) +} diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/sum_generic.go b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_generic.go new file mode 100644 index 0000000..c942a65 --- /dev/null +++ b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_generic.go @@ -0,0 +1,310 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file provides the generic implementation of Sum and MAC. Other files +// might provide optimized assembly implementations of some of this code. + +package poly1305 + +import "encoding/binary" + +// Poly1305 [RFC 7539] is a relatively simple algorithm: the authentication tag +// for a 64 bytes message is approximately +// +// s + m[0:16] * r⁴ + m[16:32] * r³ + m[32:48] * r² + m[48:64] * r mod 2¹³⁰ - 5 +// +// for some secret r and s. It can be computed sequentially like +// +// for len(msg) > 0: +// h += read(msg, 16) +// h *= r +// h %= 2¹³⁰ - 5 +// return h + s +// +// All the complexity is about doing performant constant-time math on numbers +// larger than any available numeric type. + +func sumGeneric(out *[TagSize]byte, msg []byte, key *[32]byte) { + h := newMACGeneric(key) + h.Write(msg) + h.Sum(out) +} + +func newMACGeneric(key *[32]byte) macGeneric { + m := macGeneric{} + initialize(key, &m.macState) + return m +} + +// macState holds numbers in saturated 64-bit little-endian limbs. That is, +// the value of [x0, x1, x2] is x[0] + x[1] * 2⁶⁴ + x[2] * 2¹²⁸. +type macState struct { + // h is the main accumulator. It is to be interpreted modulo 2¹³⁰ - 5, but + // can grow larger during and after rounds. It must, however, remain below + // 2 * (2¹³⁰ - 5). + h [3]uint64 + // r and s are the private key components. + r [2]uint64 + s [2]uint64 +} + +type macGeneric struct { + macState + + buffer [TagSize]byte + offset int +} + +// Write splits the incoming message into TagSize chunks, and passes them to +// update. It buffers incomplete chunks. +func (h *macGeneric) Write(p []byte) (int, error) { + nn := len(p) + if h.offset > 0 { + n := copy(h.buffer[h.offset:], p) + if h.offset+n < TagSize { + h.offset += n + return nn, nil + } + p = p[n:] + h.offset = 0 + updateGeneric(&h.macState, h.buffer[:]) + } + if n := len(p) - (len(p) % TagSize); n > 0 { + updateGeneric(&h.macState, p[:n]) + p = p[n:] + } + if len(p) > 0 { + h.offset += copy(h.buffer[h.offset:], p) + } + return nn, nil +} + +// Sum flushes the last incomplete chunk from the buffer, if any, and generates +// the MAC output. It does not modify its state, in order to allow for multiple +// calls to Sum, even if no Write is allowed after Sum. +func (h *macGeneric) Sum(out *[TagSize]byte) { + state := h.macState + if h.offset > 0 { + updateGeneric(&state, h.buffer[:h.offset]) + } + finalize(out, &state.h, &state.s) +} + +// [rMask0, rMask1] is the specified Poly1305 clamping mask in little-endian. It +// clears some bits of the secret coefficient to make it possible to implement +// multiplication more efficiently. +const ( + rMask0 = 0x0FFFFFFC0FFFFFFF + rMask1 = 0x0FFFFFFC0FFFFFFC +) + +// initialize loads the 256-bit key into the two 128-bit secret values r and s. +func initialize(key *[32]byte, m *macState) { + m.r[0] = binary.LittleEndian.Uint64(key[0:8]) & rMask0 + m.r[1] = binary.LittleEndian.Uint64(key[8:16]) & rMask1 + m.s[0] = binary.LittleEndian.Uint64(key[16:24]) + m.s[1] = binary.LittleEndian.Uint64(key[24:32]) +} + +// uint128 holds a 128-bit number as two 64-bit limbs, for use with the +// bits.Mul64 and bits.Add64 intrinsics. +type uint128 struct { + lo, hi uint64 +} + +func mul64(a, b uint64) uint128 { + hi, lo := bitsMul64(a, b) + return uint128{lo, hi} +} + +func add128(a, b uint128) uint128 { + lo, c := bitsAdd64(a.lo, b.lo, 0) + hi, c := bitsAdd64(a.hi, b.hi, c) + if c != 0 { + panic("poly1305: unexpected overflow") + } + return uint128{lo, hi} +} + +func shiftRightBy2(a uint128) uint128 { + a.lo = a.lo>>2 | (a.hi&3)<<62 + a.hi = a.hi >> 2 + return a +} + +// updateGeneric absorbs msg into the state.h accumulator. For each chunk m of +// 128 bits of message, it computes +// +// h₊ = (h + m) * r mod 2¹³⁰ - 5 +// +// If the msg length is not a multiple of TagSize, it assumes the last +// incomplete chunk is the final one. +func updateGeneric(state *macState, msg []byte) { + h0, h1, h2 := state.h[0], state.h[1], state.h[2] + r0, r1 := state.r[0], state.r[1] + + for len(msg) > 0 { + var c uint64 + + // For the first step, h + m, we use a chain of bits.Add64 intrinsics. + // The resulting value of h might exceed 2¹³⁰ - 5, but will be partially + // reduced at the end of the multiplication below. + // + // The spec requires us to set a bit just above the message size, not to + // hide leading zeroes. For full chunks, that's 1 << 128, so we can just + // add 1 to the most significant (2¹²⁸) limb, h2. + if len(msg) >= TagSize { + h0, c = bitsAdd64(h0, binary.LittleEndian.Uint64(msg[0:8]), 0) + h1, c = bitsAdd64(h1, binary.LittleEndian.Uint64(msg[8:16]), c) + h2 += c + 1 + + msg = msg[TagSize:] + } else { + var buf [TagSize]byte + copy(buf[:], msg) + buf[len(msg)] = 1 + + h0, c = bitsAdd64(h0, binary.LittleEndian.Uint64(buf[0:8]), 0) + h1, c = bitsAdd64(h1, binary.LittleEndian.Uint64(buf[8:16]), c) + h2 += c + + msg = nil + } + + // Multiplication of big number limbs is similar to elementary school + // columnar multiplication. Instead of digits, there are 64-bit limbs. + // + // We are multiplying a 3 limbs number, h, by a 2 limbs number, r. + // + // h2 h1 h0 x + // r1 r0 = + // ---------------- + // h2r0 h1r0 h0r0 <-- individual 128-bit products + // + h2r1 h1r1 h0r1 + // ------------------------ + // m3 m2 m1 m0 <-- result in 128-bit overlapping limbs + // ------------------------ + // m3.hi m2.hi m1.hi m0.hi <-- carry propagation + // + m3.lo m2.lo m1.lo m0.lo + // ------------------------------- + // t4 t3 t2 t1 t0 <-- final result in 64-bit limbs + // + // The main difference from pen-and-paper multiplication is that we do + // carry propagation in a separate step, as if we wrote two digit sums + // at first (the 128-bit limbs), and then carried the tens all at once. + + h0r0 := mul64(h0, r0) + h1r0 := mul64(h1, r0) + h2r0 := mul64(h2, r0) + h0r1 := mul64(h0, r1) + h1r1 := mul64(h1, r1) + h2r1 := mul64(h2, r1) + + // Since h2 is known to be at most 7 (5 + 1 + 1), and r0 and r1 have their + // top 4 bits cleared by rMask{0,1}, we know that their product is not going + // to overflow 64 bits, so we can ignore the high part of the products. + // + // This also means that the product doesn't have a fifth limb (t4). + if h2r0.hi != 0 { + panic("poly1305: unexpected overflow") + } + if h2r1.hi != 0 { + panic("poly1305: unexpected overflow") + } + + m0 := h0r0 + m1 := add128(h1r0, h0r1) // These two additions don't overflow thanks again + m2 := add128(h2r0, h1r1) // to the 4 masked bits at the top of r0 and r1. + m3 := h2r1 + + t0 := m0.lo + t1, c := bitsAdd64(m1.lo, m0.hi, 0) + t2, c := bitsAdd64(m2.lo, m1.hi, c) + t3, _ := bitsAdd64(m3.lo, m2.hi, c) + + // Now we have the result as 4 64-bit limbs, and we need to reduce it + // modulo 2¹³⁰ - 5. The special shape of this Crandall prime lets us do + // a cheap partial reduction according to the reduction identity + // + // c * 2¹³⁰ + n = c * 5 + n mod 2¹³⁰ - 5 + // + // because 2¹³⁰ = 5 mod 2¹³⁰ - 5. Partial reduction since the result is + // likely to be larger than 2¹³⁰ - 5, but still small enough to fit the + // assumptions we make about h in the rest of the code. + // + // See also https://speakerdeck.com/gtank/engineering-prime-numbers?slide=23 + + // We split the final result at the 2¹³⁰ mark into h and cc, the carry. + // Note that the carry bits are effectively shifted left by 2, in other + // words, cc = c * 4 for the c in the reduction identity. + h0, h1, h2 = t0, t1, t2&maskLow2Bits + cc := uint128{t2 & maskNotLow2Bits, t3} + + // To add c * 5 to h, we first add cc = c * 4, and then add (cc >> 2) = c. + + h0, c = bitsAdd64(h0, cc.lo, 0) + h1, c = bitsAdd64(h1, cc.hi, c) + h2 += c + + cc = shiftRightBy2(cc) + + h0, c = bitsAdd64(h0, cc.lo, 0) + h1, c = bitsAdd64(h1, cc.hi, c) + h2 += c + + // h2 is at most 3 + 1 + 1 = 5, making the whole of h at most + // + // 5 * 2¹²⁸ + (2¹²⁸ - 1) = 6 * 2¹²⁸ - 1 + } + + state.h[0], state.h[1], state.h[2] = h0, h1, h2 +} + +const ( + maskLow2Bits uint64 = 0x0000000000000003 + maskNotLow2Bits uint64 = ^maskLow2Bits +) + +// select64 returns x if v == 1 and y if v == 0, in constant time. +func select64(v, x, y uint64) uint64 { return ^(v-1)&x | (v-1)&y } + +// [p0, p1, p2] is 2¹³⁰ - 5 in little endian order. +const ( + p0 = 0xFFFFFFFFFFFFFFFB + p1 = 0xFFFFFFFFFFFFFFFF + p2 = 0x0000000000000003 +) + +// finalize completes the modular reduction of h and computes +// +// out = h + s mod 2¹²⁸ +// +func finalize(out *[TagSize]byte, h *[3]uint64, s *[2]uint64) { + h0, h1, h2 := h[0], h[1], h[2] + + // After the partial reduction in updateGeneric, h might be more than + // 2¹³⁰ - 5, but will be less than 2 * (2¹³⁰ - 5). To complete the reduction + // in constant time, we compute t = h - (2¹³⁰ - 5), and select h as the + // result if the subtraction underflows, and t otherwise. + + hMinusP0, b := bitsSub64(h0, p0, 0) + hMinusP1, b := bitsSub64(h1, p1, b) + _, b = bitsSub64(h2, p2, b) + + // h = h if h < p else h - p + h0 = select64(b, h0, hMinusP0) + h1 = select64(b, h1, hMinusP1) + + // Finally, we compute the last Poly1305 step + // + // tag = h + s mod 2¹²⁸ + // + // by just doing a wide addition with the 128 low bits of h and discarding + // the overflow. + h0, c := bitsAdd64(h0, s[0], 0) + h1, _ = bitsAdd64(h1, s[1], c) + + binary.LittleEndian.PutUint64(out[0:8], h0) + binary.LittleEndian.PutUint64(out[8:16], h1) +} diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/sum_ppc64le.go b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_ppc64le.go new file mode 100644 index 0000000..4a06994 --- /dev/null +++ b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_ppc64le.go @@ -0,0 +1,48 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build gc && !purego +// +build gc,!purego + +package poly1305 + +//go:noescape +func update(state *macState, msg []byte) + +// mac is a wrapper for macGeneric that redirects calls that would have gone to +// updateGeneric to update. +// +// Its Write and Sum methods are otherwise identical to the macGeneric ones, but +// using function pointers would carry a major performance cost. +type mac struct{ macGeneric } + +func (h *mac) Write(p []byte) (int, error) { + nn := len(p) + if h.offset > 0 { + n := copy(h.buffer[h.offset:], p) + if h.offset+n < TagSize { + h.offset += n + return nn, nil + } + p = p[n:] + h.offset = 0 + update(&h.macState, h.buffer[:]) + } + if n := len(p) - (len(p) % TagSize); n > 0 { + update(&h.macState, p[:n]) + p = p[n:] + } + if len(p) > 0 { + h.offset += copy(h.buffer[h.offset:], p) + } + return nn, nil +} + +func (h *mac) Sum(out *[16]byte) { + state := h.macState + if h.offset > 0 { + update(&state, h.buffer[:h.offset]) + } + finalize(out, &state.h, &state.s) +} diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/sum_s390x.go b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_s390x.go new file mode 100644 index 0000000..62cc9f8 --- /dev/null +++ b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_s390x.go @@ -0,0 +1,76 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build gc && !purego +// +build gc,!purego + +package poly1305 + +import ( + "golang.org/x/sys/cpu" +) + +// updateVX is an assembly implementation of Poly1305 that uses vector +// instructions. It must only be called if the vector facility (vx) is +// available. +//go:noescape +func updateVX(state *macState, msg []byte) + +// mac is a replacement for macGeneric that uses a larger buffer and redirects +// calls that would have gone to updateGeneric to updateVX if the vector +// facility is installed. +// +// A larger buffer is required for good performance because the vector +// implementation has a higher fixed cost per call than the generic +// implementation. +type mac struct { + macState + + buffer [16 * TagSize]byte // size must be a multiple of block size (16) + offset int +} + +func (h *mac) Write(p []byte) (int, error) { + nn := len(p) + if h.offset > 0 { + n := copy(h.buffer[h.offset:], p) + if h.offset+n < len(h.buffer) { + h.offset += n + return nn, nil + } + p = p[n:] + h.offset = 0 + if cpu.S390X.HasVX { + updateVX(&h.macState, h.buffer[:]) + } else { + updateGeneric(&h.macState, h.buffer[:]) + } + } + + tail := len(p) % len(h.buffer) // number of bytes to copy into buffer + body := len(p) - tail // number of bytes to process now + if body > 0 { + if cpu.S390X.HasVX { + updateVX(&h.macState, p[:body]) + } else { + updateGeneric(&h.macState, p[:body]) + } + } + h.offset = copy(h.buffer[:], p[body:]) // copy tail bytes - can be 0 + return nn, nil +} + +func (h *mac) Sum(out *[TagSize]byte) { + state := h.macState + remainder := h.buffer[:h.offset] + + // Use the generic implementation if we have 2 or fewer blocks left + // to sum. The vector implementation has a higher startup time. + if cpu.S390X.HasVX && len(remainder) > 2*TagSize { + updateVX(&state, remainder) + } else if len(remainder) > 0 { + updateGeneric(&state, remainder) + } + finalize(out, &state.h, &state.s) +} |