TUN-3090: Upgrade crypto dep

go.mod

@@ -58,7 +58,7 @@ require (
 	github.com/stretchr/testify v1.3.0
 	github.com/tinylib/msgp v1.1.0 // indirect
 	github.com/xo/dburl v0.0.0-20191005012637-293c3298d6c0
-	golang.org/x/crypto v0.0.0-20191011191535-87dc89f01550
+	golang.org/x/crypto v0.0.0-20200604202706-70a84ac30bf9
 	golang.org/x/net v0.0.0-20191014212845-da9a3fd4c582
 	golang.org/x/oauth2 v0.0.0-20190604053449-0f29369cfe45 // indirect
 	golang.org/x/sync v0.0.0-20190423024810-112230192c58

go.sum

@@ -204,6 +204,8 @@ golang.org/x/crypto v0.0.0-20190325154230-a5d413f7728c/go.mod h1:djNgcEr1/C05ACk
 golang.org/x/crypto v0.0.0-20190701094942-4def268fd1a4/go.mod h1:yigFU9vqHzYiE8UmvKecakEJjdnWj3jj499lnFckfCI=
 golang.org/x/crypto v0.0.0-20191011191535-87dc89f01550 h1:ObdrDkeb4kJdCP557AjRjq69pTHfNouLtWZG7j9rPN8=
 golang.org/x/crypto v0.0.0-20191011191535-87dc89f01550/go.mod h1:yigFU9vqHzYiE8UmvKecakEJjdnWj3jj499lnFckfCI=
+golang.org/x/crypto v0.0.0-20200604202706-70a84ac30bf9 h1:vEg9joUBmeBcK9iSJftGNf3coIG4HqZElCPehJsfAYM=
+golang.org/x/crypto v0.0.0-20200604202706-70a84ac30bf9/go.mod h1:LzIPMQfyMNhhGPhUkYOs5KpL4U8rLKemX1yGLhDgUto=
 golang.org/x/exp v0.0.0-20190121172915-509febef88a4/go.mod h1:CJ0aWSM057203Lf6IL+f9T1iT9GByDxfZKAQTCR3kQA=
 golang.org/x/lint v0.0.0-20181026193005-c67002cb31c3/go.mod h1:UVdnD1Gm6xHRNCYTkRU2/jEulfH38KcIWyp/GAMgvoE=
 golang.org/x/lint v0.0.0-20190227174305-5b3e6a55c961/go.mod h1:wehouNa3lNwaWXcvxsM5YxQ5yQlVC4a0KAMCusXpPoU=

vendor/golang.org/x/crypto/blake2b/blake2b.go

@@ -0,0 +1,291 @@
+// Copyright 2016 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 blake2b implements the BLAKE2b hash algorithm defined by RFC 7693
+// and the extendable output function (XOF) BLAKE2Xb.
+//
+// BLAKE2b is optimized for 64-bit platforms—including NEON-enabled ARMs—and
+// produces digests of any size between 1 and 64 bytes.
+// For a detailed specification of BLAKE2b see https://blake2.net/blake2.pdf
+// and for BLAKE2Xb see https://blake2.net/blake2x.pdf
+//
+// If you aren't sure which function you need, use BLAKE2b (Sum512 or New512).
+// If you need a secret-key MAC (message authentication code), use the New512
+// function with a non-nil key.
+//
+// BLAKE2X is a construction to compute hash values larger than 64 bytes. It
+// can produce hash values between 0 and 4 GiB.
+package blake2b
+
+import (
+	"encoding/binary"
+	"errors"
+	"hash"
+)
+
+const (
+	// The blocksize of BLAKE2b in bytes.
+	BlockSize = 128
+	// The hash size of BLAKE2b-512 in bytes.
+	Size = 64
+	// The hash size of BLAKE2b-384 in bytes.
+	Size384 = 48
+	// The hash size of BLAKE2b-256 in bytes.
+	Size256 = 32
+)
+
+var (
+	useAVX2 bool
+	useAVX  bool
+	useSSE4 bool
+)
+
+var (
+	errKeySize  = errors.New("blake2b: invalid key size")
+	errHashSize = errors.New("blake2b: invalid hash size")
+)
+
+var iv = [8]uint64{
+	0x6a09e667f3bcc908, 0xbb67ae8584caa73b, 0x3c6ef372fe94f82b, 0xa54ff53a5f1d36f1,
+	0x510e527fade682d1, 0x9b05688c2b3e6c1f, 0x1f83d9abfb41bd6b, 0x5be0cd19137e2179,
+}
+
+// Sum512 returns the BLAKE2b-512 checksum of the data.
+func Sum512(data []byte) [Size]byte {
+	var sum [Size]byte
+	checkSum(&sum, Size, data)
+	return sum
+}
+
+// Sum384 returns the BLAKE2b-384 checksum of the data.
+func Sum384(data []byte) [Size384]byte {
+	var sum [Size]byte
+	var sum384 [Size384]byte
+	checkSum(&sum, Size384, data)
+	copy(sum384[:], sum[:Size384])
+	return sum384
+}
+
+// Sum256 returns the BLAKE2b-256 checksum of the data.
+func Sum256(data []byte) [Size256]byte {
+	var sum [Size]byte
+	var sum256 [Size256]byte
+	checkSum(&sum, Size256, data)
+	copy(sum256[:], sum[:Size256])
+	return sum256
+}
+
+// New512 returns a new hash.Hash computing the BLAKE2b-512 checksum. A non-nil
+// key turns the hash into a MAC. The key must be between zero and 64 bytes long.
+func New512(key []byte) (hash.Hash, error) { return newDigest(Size, key) }
+
+// New384 returns a new hash.Hash computing the BLAKE2b-384 checksum. A non-nil
+// key turns the hash into a MAC. The key must be between zero and 64 bytes long.
+func New384(key []byte) (hash.Hash, error) { return newDigest(Size384, key) }
+
+// New256 returns a new hash.Hash computing the BLAKE2b-256 checksum. A non-nil
+// key turns the hash into a MAC. The key must be between zero and 64 bytes long.
+func New256(key []byte) (hash.Hash, error) { return newDigest(Size256, key) }
+
+// New returns a new hash.Hash computing the BLAKE2b checksum with a custom length.
+// A non-nil key turns the hash into a MAC. The key must be between zero and 64 bytes long.
+// The hash size can be a value between 1 and 64 but it is highly recommended to use
+// values equal or greater than:
+// - 32 if BLAKE2b is used as a hash function (The key is zero bytes long).
+// - 16 if BLAKE2b is used as a MAC function (The key is at least 16 bytes long).
+// When the key is nil, the returned hash.Hash implements BinaryMarshaler
+// and BinaryUnmarshaler for state (de)serialization as documented by hash.Hash.
+func New(size int, key []byte) (hash.Hash, error) { return newDigest(size, key) }
+
+func newDigest(hashSize int, key []byte) (*digest, error) {
+	if hashSize < 1 || hashSize > Size {
+		return nil, errHashSize
+	}
+	if len(key) > Size {
+		return nil, errKeySize
+	}
+	d := &digest{
+		size:   hashSize,
+		keyLen: len(key),
+	}
+	copy(d.key[:], key)
+	d.Reset()
+	return d, nil
+}
+
+func checkSum(sum *[Size]byte, hashSize int, data []byte) {
+	h := iv
+	h[0] ^= uint64(hashSize) | (1 << 16) | (1 << 24)
+	var c [2]uint64
+
+	if length := len(data); length > BlockSize {
+		n := length &^ (BlockSize - 1)
+		if length == n {
+			n -= BlockSize
+		}
+		hashBlocks(&h, &c, 0, data[:n])
+		data = data[n:]
+	}
+
+	var block [BlockSize]byte
+	offset := copy(block[:], data)
+	remaining := uint64(BlockSize - offset)
+	if c[0] < remaining {
+		c[1]--
+	}
+	c[0] -= remaining
+
+	hashBlocks(&h, &c, 0xFFFFFFFFFFFFFFFF, block[:])
+
+	for i, v := range h[:(hashSize+7)/8] {
+		binary.LittleEndian.PutUint64(sum[8*i:], v)
+	}
+}
+
+type digest struct {
+	h      [8]uint64
+	c      [2]uint64
+	size   int
+	block  [BlockSize]byte
+	offset int
+
+	key    [BlockSize]byte
+	keyLen int
+}
+
+const (
+	magic         = "b2b"
+	marshaledSize = len(magic) + 8*8 + 2*8 + 1 + BlockSize + 1
+)
+
+func (d *digest) MarshalBinary() ([]byte, error) {
+	if d.keyLen != 0 {
+		return nil, errors.New("crypto/blake2b: cannot marshal MACs")
+	}
+	b := make([]byte, 0, marshaledSize)
+	b = append(b, magic...)
+	for i := 0; i < 8; i++ {
+		b = appendUint64(b, d.h[i])
+	}
+	b = appendUint64(b, d.c[0])
+	b = appendUint64(b, d.c[1])
+	// Maximum value for size is 64
+	b = append(b, byte(d.size))
+	b = append(b, d.block[:]...)
+	b = append(b, byte(d.offset))
+	return b, nil
+}
+
+func (d *digest) UnmarshalBinary(b []byte) error {
+	if len(b) < len(magic) || string(b[:len(magic)]) != magic {
+		return errors.New("crypto/blake2b: invalid hash state identifier")
+	}
+	if len(b) != marshaledSize {
+		return errors.New("crypto/blake2b: invalid hash state size")
+	}
+	b = b[len(magic):]
+	for i := 0; i < 8; i++ {
+		b, d.h[i] = consumeUint64(b)
+	}
+	b, d.c[0] = consumeUint64(b)
+	b, d.c[1] = consumeUint64(b)
+	d.size = int(b[0])
+	b = b[1:]
+	copy(d.block[:], b[:BlockSize])
+	b = b[BlockSize:]
+	d.offset = int(b[0])
+	return nil
+}
+
+func (d *digest) BlockSize() int { return BlockSize }
+
+func (d *digest) Size() int { return d.size }
+
+func (d *digest) Reset() {
+	d.h = iv
+	d.h[0] ^= uint64(d.size) | (uint64(d.keyLen) << 8) | (1 << 16) | (1 << 24)
+	d.offset, d.c[0], d.c[1] = 0, 0, 0
+	if d.keyLen > 0 {
+		d.block = d.key
+		d.offset = BlockSize
+	}
+}
+
+func (d *digest) Write(p []byte) (n int, err error) {
+	n = len(p)
+
+	if d.offset > 0 {
+		remaining := BlockSize - d.offset
+		if n <= remaining {
+			d.offset += copy(d.block[d.offset:], p)
+			return
+		}
+		copy(d.block[d.offset:], p[:remaining])
+		hashBlocks(&d.h, &d.c, 0, d.block[:])
+		d.offset = 0
+		p = p[remaining:]
+	}
+
+	if length := len(p); length > BlockSize {
+		nn := length &^ (BlockSize - 1)
+		if length == nn {
+			nn -= BlockSize
+		}
+		hashBlocks(&d.h, &d.c, 0, p[:nn])
+		p = p[nn:]
+	}
+
+	if len(p) > 0 {
+		d.offset += copy(d.block[:], p)
+	}
+
+	return
+}
+
+func (d *digest) Sum(sum []byte) []byte {
+	var hash [Size]byte
+	d.finalize(&hash)
+	return append(sum, hash[:d.size]...)
+}
+
+func (d *digest) finalize(hash *[Size]byte) {
+	var block [BlockSize]byte
+	copy(block[:], d.block[:d.offset])
+	remaining := uint64(BlockSize - d.offset)
+
+	c := d.c
+	if c[0] < remaining {
+		c[1]--
+	}
+	c[0] -= remaining
+
+	h := d.h
+	hashBlocks(&h, &c, 0xFFFFFFFFFFFFFFFF, block[:])
+
+	for i, v := range h {
+		binary.LittleEndian.PutUint64(hash[8*i:], v)
+	}
+}
+
+func appendUint64(b []byte, x uint64) []byte {
+	var a [8]byte
+	binary.BigEndian.PutUint64(a[:], x)
+	return append(b, a[:]...)
+}
+
+func appendUint32(b []byte, x uint32) []byte {
+	var a [4]byte
+	binary.BigEndian.PutUint32(a[:], x)
+	return append(b, a[:]...)
+}
+
+func consumeUint64(b []byte) ([]byte, uint64) {
+	x := binary.BigEndian.Uint64(b)
+	return b[8:], x
+}
+
+func consumeUint32(b []byte) ([]byte, uint32) {
+	x := binary.BigEndian.Uint32(b)
+	return b[4:], x
+}

vendor/golang.org/x/crypto/blake2b/blake2bAVX2_amd64.go

@@ -0,0 +1,37 @@
+// Copyright 2016 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.
+
+// +build go1.7,amd64,!gccgo,!appengine
+
+package blake2b
+
+import "golang.org/x/sys/cpu"
+
+func init() {
+	useAVX2 = cpu.X86.HasAVX2
+	useAVX = cpu.X86.HasAVX
+	useSSE4 = cpu.X86.HasSSE41
+}
+
+//go:noescape
+func hashBlocksAVX2(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+
+//go:noescape
+func hashBlocksAVX(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+
+//go:noescape
+func hashBlocksSSE4(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+
+func hashBlocks(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte) {
+	switch {
+	case useAVX2:
+		hashBlocksAVX2(h, c, flag, blocks)
+	case useAVX:
+		hashBlocksAVX(h, c, flag, blocks)
+	case useSSE4:
+		hashBlocksSSE4(h, c, flag, blocks)
+	default:
+		hashBlocksGeneric(h, c, flag, blocks)
+	}
+}

vendor/golang.org/x/crypto/blake2b/blake2bAVX2_amd64.s

@@ -0,0 +1,750 @@
+// Copyright 2016 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.
+
+// +build go1.7,amd64,!gccgo,!appengine
+
+#include "textflag.h"
+
+DATA ·AVX2_iv0<>+0x00(SB)/8, $0x6a09e667f3bcc908
+DATA ·AVX2_iv0<>+0x08(SB)/8, $0xbb67ae8584caa73b
+DATA ·AVX2_iv0<>+0x10(SB)/8, $0x3c6ef372fe94f82b
+DATA ·AVX2_iv0<>+0x18(SB)/8, $0xa54ff53a5f1d36f1
+GLOBL ·AVX2_iv0<>(SB), (NOPTR+RODATA), $32
+
+DATA ·AVX2_iv1<>+0x00(SB)/8, $0x510e527fade682d1
+DATA ·AVX2_iv1<>+0x08(SB)/8, $0x9b05688c2b3e6c1f
+DATA ·AVX2_iv1<>+0x10(SB)/8, $0x1f83d9abfb41bd6b
+DATA ·AVX2_iv1<>+0x18(SB)/8, $0x5be0cd19137e2179
+GLOBL ·AVX2_iv1<>(SB), (NOPTR+RODATA), $32
+
+DATA ·AVX2_c40<>+0x00(SB)/8, $0x0201000706050403
+DATA ·AVX2_c40<>+0x08(SB)/8, $0x0a09080f0e0d0c0b
+DATA ·AVX2_c40<>+0x10(SB)/8, $0x0201000706050403
+DATA ·AVX2_c40<>+0x18(SB)/8, $0x0a09080f0e0d0c0b
+GLOBL ·AVX2_c40<>(SB), (NOPTR+RODATA), $32
+
+DATA ·AVX2_c48<>+0x00(SB)/8, $0x0100070605040302
+DATA ·AVX2_c48<>+0x08(SB)/8, $0x09080f0e0d0c0b0a
+DATA ·AVX2_c48<>+0x10(SB)/8, $0x0100070605040302
+DATA ·AVX2_c48<>+0x18(SB)/8, $0x09080f0e0d0c0b0a
+GLOBL ·AVX2_c48<>(SB), (NOPTR+RODATA), $32
+
+DATA ·AVX_iv0<>+0x00(SB)/8, $0x6a09e667f3bcc908
+DATA ·AVX_iv0<>+0x08(SB)/8, $0xbb67ae8584caa73b
+GLOBL ·AVX_iv0<>(SB), (NOPTR+RODATA), $16
+
+DATA ·AVX_iv1<>+0x00(SB)/8, $0x3c6ef372fe94f82b
+DATA ·AVX_iv1<>+0x08(SB)/8, $0xa54ff53a5f1d36f1
+GLOBL ·AVX_iv1<>(SB), (NOPTR+RODATA), $16
+
+DATA ·AVX_iv2<>+0x00(SB)/8, $0x510e527fade682d1
+DATA ·AVX_iv2<>+0x08(SB)/8, $0x9b05688c2b3e6c1f
+GLOBL ·AVX_iv2<>(SB), (NOPTR+RODATA), $16
+
+DATA ·AVX_iv3<>+0x00(SB)/8, $0x1f83d9abfb41bd6b
+DATA ·AVX_iv3<>+0x08(SB)/8, $0x5be0cd19137e2179
+GLOBL ·AVX_iv3<>(SB), (NOPTR+RODATA), $16
+
+DATA ·AVX_c40<>+0x00(SB)/8, $0x0201000706050403
+DATA ·AVX_c40<>+0x08(SB)/8, $0x0a09080f0e0d0c0b
+GLOBL ·AVX_c40<>(SB), (NOPTR+RODATA), $16
+
+DATA ·AVX_c48<>+0x00(SB)/8, $0x0100070605040302
+DATA ·AVX_c48<>+0x08(SB)/8, $0x09080f0e0d0c0b0a
+GLOBL ·AVX_c48<>(SB), (NOPTR+RODATA), $16
+
+#define VPERMQ_0x39_Y1_Y1 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xc9; BYTE $0x39
+#define VPERMQ_0x93_Y1_Y1 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xc9; BYTE $0x93
+#define VPERMQ_0x4E_Y2_Y2 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xd2; BYTE $0x4e
+#define VPERMQ_0x93_Y3_Y3 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xdb; BYTE $0x93
+#define VPERMQ_0x39_Y3_Y3 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xdb; BYTE $0x39
+
+#define ROUND_AVX2(m0, m1, m2, m3, t, c40, c48) \
+	VPADDQ  m0, Y0, Y0;   \
+	VPADDQ  Y1, Y0, Y0;   \
+	VPXOR   Y0, Y3, Y3;   \
+	VPSHUFD $-79, Y3, Y3; \
+	VPADDQ  Y3, Y2, Y2;   \
+	VPXOR   Y2, Y1, Y1;   \
+	VPSHUFB c40, Y1, Y1;  \
+	VPADDQ  m1, Y0, Y0;   \
+	VPADDQ  Y1, Y0, Y0;   \
+	VPXOR   Y0, Y3, Y3;   \
+	VPSHUFB c48, Y3, Y3;  \
+	VPADDQ  Y3, Y2, Y2;   \
+	VPXOR   Y2, Y1, Y1;   \
+	VPADDQ  Y1, Y1, t;    \
+	VPSRLQ  $63, Y1, Y1;  \
+	VPXOR   t, Y1, Y1;    \
+	VPERMQ_0x39_Y1_Y1;    \
+	VPERMQ_0x4E_Y2_Y2;    \
+	VPERMQ_0x93_Y3_Y3;    \
+	VPADDQ  m2, Y0, Y0;   \
+	VPADDQ  Y1, Y0, Y0;   \
+	VPXOR   Y0, Y3, Y3;   \
+	VPSHUFD $-79, Y3, Y3; \
+	VPADDQ  Y3, Y2, Y2;   \
+	VPXOR   Y2, Y1, Y1;   \
+	VPSHUFB c40, Y1, Y1;  \
+	VPADDQ  m3, Y0, Y0;   \
+	VPADDQ  Y1, Y0, Y0;   \
+	VPXOR   Y0, Y3, Y3;   \
+	VPSHUFB c48, Y3, Y3;  \
+	VPADDQ  Y3, Y2, Y2;   \
+	VPXOR   Y2, Y1, Y1;   \
+	VPADDQ  Y1, Y1, t;    \
+	VPSRLQ  $63, Y1, Y1;  \
+	VPXOR   t, Y1, Y1;    \
+	VPERMQ_0x39_Y3_Y3;    \
+	VPERMQ_0x4E_Y2_Y2;    \
+	VPERMQ_0x93_Y1_Y1
+
+#define VMOVQ_SI_X11_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x1E
+#define VMOVQ_SI_X12_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x26
+#define VMOVQ_SI_X13_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x2E
+#define VMOVQ_SI_X14_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x36
+#define VMOVQ_SI_X15_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x3E
+
+#define VMOVQ_SI_X11(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x5E; BYTE $n
+#define VMOVQ_SI_X12(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x66; BYTE $n
+#define VMOVQ_SI_X13(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x6E; BYTE $n
+#define VMOVQ_SI_X14(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x76; BYTE $n
+#define VMOVQ_SI_X15(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x7E; BYTE $n
+
+#define VPINSRQ_1_SI_X11_0 BYTE $0xC4; BYTE $0x63; BYTE $0xA1; BYTE $0x22; BYTE $0x1E; BYTE $0x01
+#define VPINSRQ_1_SI_X12_0 BYTE $0xC4; BYTE $0x63; BYTE $0x99; BYTE $0x22; BYTE $0x26; BYTE $0x01
+#define VPINSRQ_1_SI_X13_0 BYTE $0xC4; BYTE $0x63; BYTE $0x91; BYTE $0x22; BYTE $0x2E; BYTE $0x01
+#define VPINSRQ_1_SI_X14_0 BYTE $0xC4; BYTE $0x63; BYTE $0x89; BYTE $0x22; BYTE $0x36; BYTE $0x01
+#define VPINSRQ_1_SI_X15_0 BYTE $0xC4; BYTE $0x63; BYTE $0x81; BYTE $0x22; BYTE $0x3E; BYTE $0x01
+
+#define VPINSRQ_1_SI_X11(n) BYTE $0xC4; BYTE $0x63; BYTE $0xA1; BYTE $0x22; BYTE $0x5E; BYTE $n; BYTE $0x01
+#define VPINSRQ_1_SI_X12(n) BYTE $0xC4; BYTE $0x63; BYTE $0x99; BYTE $0x22; BYTE $0x66; BYTE $n; BYTE $0x01
+#define VPINSRQ_1_SI_X13(n) BYTE $0xC4; BYTE $0x63; BYTE $0x91; BYTE $0x22; BYTE $0x6E; BYTE $n; BYTE $0x01
+#define VPINSRQ_1_SI_X14(n) BYTE $0xC4; BYTE $0x63; BYTE $0x89; BYTE $0x22; BYTE $0x76; BYTE $n; BYTE $0x01
+#define VPINSRQ_1_SI_X15(n) BYTE $0xC4; BYTE $0x63; BYTE $0x81; BYTE $0x22; BYTE $0x7E; BYTE $n; BYTE $0x01
+
+#define VMOVQ_R8_X15 BYTE $0xC4; BYTE $0x41; BYTE $0xF9; BYTE $0x6E; BYTE $0xF8
+#define VPINSRQ_1_R9_X15 BYTE $0xC4; BYTE $0x43; BYTE $0x81; BYTE $0x22; BYTE $0xF9; BYTE $0x01
+
+// load msg: Y12 = (i0, i1, i2, i3)
+// i0, i1, i2, i3 must not be 0
+#define LOAD_MSG_AVX2_Y12(i0, i1, i2, i3) \
+	VMOVQ_SI_X12(i0*8);           \
+	VMOVQ_SI_X11(i2*8);           \
+	VPINSRQ_1_SI_X12(i1*8);       \
+	VPINSRQ_1_SI_X11(i3*8);       \
+	VINSERTI128 $1, X11, Y12, Y12
+
+// load msg: Y13 = (i0, i1, i2, i3)
+// i0, i1, i2, i3 must not be 0
+#define LOAD_MSG_AVX2_Y13(i0, i1, i2, i3) \
+	VMOVQ_SI_X13(i0*8);           \
+	VMOVQ_SI_X11(i2*8);           \
+	VPINSRQ_1_SI_X13(i1*8);       \
+	VPINSRQ_1_SI_X11(i3*8);       \
+	VINSERTI128 $1, X11, Y13, Y13
+
+// load msg: Y14 = (i0, i1, i2, i3)
+// i0, i1, i2, i3 must not be 0
+#define LOAD_MSG_AVX2_Y14(i0, i1, i2, i3) \
+	VMOVQ_SI_X14(i0*8);           \
+	VMOVQ_SI_X11(i2*8);           \
+	VPINSRQ_1_SI_X14(i1*8);       \
+	VPINSRQ_1_SI_X11(i3*8);       \
+	VINSERTI128 $1, X11, Y14, Y14
+
+// load msg: Y15 = (i0, i1, i2, i3)
+// i0, i1, i2, i3 must not be 0
+#define LOAD_MSG_AVX2_Y15(i0, i1, i2, i3) \
+	VMOVQ_SI_X15(i0*8);           \
+	VMOVQ_SI_X11(i2*8);           \
+	VPINSRQ_1_SI_X15(i1*8);       \
+	VPINSRQ_1_SI_X11(i3*8);       \
+	VINSERTI128 $1, X11, Y15, Y15
+
+#define LOAD_MSG_AVX2_0_2_4_6_1_3_5_7_8_10_12_14_9_11_13_15() \
+	VMOVQ_SI_X12_0;                   \
+	VMOVQ_SI_X11(4*8);                \
+	VPINSRQ_1_SI_X12(2*8);            \
+	VPINSRQ_1_SI_X11(6*8);            \
+	VINSERTI128 $1, X11, Y12, Y12;    \
+	LOAD_MSG_AVX2_Y13(1, 3, 5, 7);    \
+	LOAD_MSG_AVX2_Y14(8, 10, 12, 14); \
+	LOAD_MSG_AVX2_Y15(9, 11, 13, 15)
+
+#define LOAD_MSG_AVX2_14_4_9_13_10_8_15_6_1_0_11_5_12_2_7_3() \
+	LOAD_MSG_AVX2_Y12(14, 4, 9, 13); \
+	LOAD_MSG_AVX2_Y13(10, 8, 15, 6); \
+	VMOVQ_SI_X11(11*8);              \
+	VPSHUFD     $0x4E, 0*8(SI), X14; \
+	VPINSRQ_1_SI_X11(5*8);           \
+	VINSERTI128 $1, X11, Y14, Y14;   \
+	LOAD_MSG_AVX2_Y15(12, 2, 7, 3)
+
+#define LOAD_MSG_AVX2_11_12_5_15_8_0_2_13_10_3_7_9_14_6_1_4() \
+	VMOVQ_SI_X11(5*8);              \
+	VMOVDQU     11*8(SI), X12;      \
+	VPINSRQ_1_SI_X11(15*8);         \
+	VINSERTI128 $1, X11, Y12, Y12;  \
+	VMOVQ_SI_X13(8*8);              \
+	VMOVQ_SI_X11(2*8);              \
+	VPINSRQ_1_SI_X13_0;             \
+	VPINSRQ_1_SI_X11(13*8);         \
+	VINSERTI128 $1, X11, Y13, Y13;  \
+	LOAD_MSG_AVX2_Y14(10, 3, 7, 9); \
+	LOAD_MSG_AVX2_Y15(14, 6, 1, 4)
+
+#define LOAD_MSG_AVX2_7_3_13_11_9_1_12_14_2_5_4_15_6_10_0_8() \
+	LOAD_MSG_AVX2_Y12(7, 3, 13, 11); \
+	LOAD_MSG_AVX2_Y13(9, 1, 12, 14); \
+	LOAD_MSG_AVX2_Y14(2, 5, 4, 15);  \
+	VMOVQ_SI_X15(6*8);               \
+	VMOVQ_SI_X11_0;                  \
+	VPINSRQ_1_SI_X15(10*8);          \
+	VPINSRQ_1_SI_X11(8*8);           \
+	VINSERTI128 $1, X11, Y15, Y15
+
+#define LOAD_MSG_AVX2_9_5_2_10_0_7_4_15_14_11_6_3_1_12_8_13() \
+	LOAD_MSG_AVX2_Y12(9, 5, 2, 10);  \
+	VMOVQ_SI_X13_0;                  \
+	VMOVQ_SI_X11(4*8);               \
+	VPINSRQ_1_SI_X13(7*8);           \
+	VPINSRQ_1_SI_X11(15*8);          \
+	VINSERTI128 $1, X11, Y13, Y13;   \
+	LOAD_MSG_AVX2_Y14(14, 11, 6, 3); \
+	LOAD_MSG_AVX2_Y15(1, 12, 8, 13)
+
+#define LOAD_MSG_AVX2_2_6_0_8_12_10_11_3_4_7_15_1_13_5_14_9() \
+	VMOVQ_SI_X12(2*8);                \
+	VMOVQ_SI_X11_0;                   \
+	VPINSRQ_1_SI_X12(6*8);            \
+	VPINSRQ_1_SI_X11(8*8);            \
+	VINSERTI128 $1, X11, Y12, Y12;    \
+	LOAD_MSG_AVX2_Y13(12, 10, 11, 3); \
+	LOAD_MSG_AVX2_Y14(4, 7, 15, 1);   \
+	LOAD_MSG_AVX2_Y15(13, 5, 14, 9)
+
+#define LOAD_MSG_AVX2_12_1_14_4_5_15_13_10_0_6_9_8_7_3_2_11() \
+	LOAD_MSG_AVX2_Y12(12, 1, 14, 4);  \
+	LOAD_MSG_AVX2_Y13(5, 15, 13, 10); \
+	VMOVQ_SI_X14_0;                   \
+	VPSHUFD     $0x4E, 8*8(SI), X11;  \
+	VPINSRQ_1_SI_X14(6*8);            \
+	VINSERTI128 $1, X11, Y14, Y14;    \
+	LOAD_MSG_AVX2_Y15(7, 3, 2, 11)
+
+#define LOAD_MSG_AVX2_13_7_12_3_11_14_1_9_5_15_8_2_0_4_6_10() \
+	LOAD_MSG_AVX2_Y12(13, 7, 12, 3); \
+	LOAD_MSG_AVX2_Y13(11, 14, 1, 9); \
+	LOAD_MSG_AVX2_Y14(5, 15, 8, 2);  \
+	VMOVQ_SI_X15_0;                  \
+	VMOVQ_SI_X11(6*8);               \
+	VPINSRQ_1_SI_X15(4*8);           \
+	VPINSRQ_1_SI_X11(10*8);          \
+	VINSERTI128 $1, X11, Y15, Y15
+
+#define LOAD_MSG_AVX2_6_14_11_0_15_9_3_8_12_13_1_10_2_7_4_5() \
+	VMOVQ_SI_X12(6*8);              \
+	VMOVQ_SI_X11(11*8);             \
+	VPINSRQ_1_SI_X12(14*8);         \
+	VPINSRQ_1_SI_X11_0;             \
+	VINSERTI128 $1, X11, Y12, Y12;  \
+	LOAD_MSG_AVX2_Y13(15, 9, 3, 8); \
+	VMOVQ_SI_X11(1*8);              \
+	VMOVDQU     12*8(SI), X14;      \
+	VPINSRQ_1_SI_X11(10*8);         \
+	VINSERTI128 $1, X11, Y14, Y14;  \
+	VMOVQ_SI_X15(2*8);              \
+	VMOVDQU     4*8(SI), X11;       \
+	VPINSRQ_1_SI_X15(7*8);          \
+	VINSERTI128 $1, X11, Y15, Y15
+
+#define LOAD_MSG_AVX2_10_8_7_1_2_4_6_5_15_9_3_13_11_14_12_0() \
+	LOAD_MSG_AVX2_Y12(10, 8, 7, 1);  \
+	VMOVQ_SI_X13(2*8);               \
+	VPSHUFD     $0x4E, 5*8(SI), X11; \
+	VPINSRQ_1_SI_X13(4*8);           \
+	VINSERTI128 $1, X11, Y13, Y13;   \
+	LOAD_MSG_AVX2_Y14(15, 9, 3, 13); \
+	VMOVQ_SI_X15(11*8);              \
+	VMOVQ_SI_X11(12*8);              \
+	VPINSRQ_1_SI_X15(14*8);          \
+	VPINSRQ_1_SI_X11_0;              \
+	VINSERTI128 $1, X11, Y15, Y15
+
+// func hashBlocksAVX2(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+TEXT ·hashBlocksAVX2(SB), 4, $320-48 // frame size = 288 + 32 byte alignment
+	MOVQ h+0(FP), AX
+	MOVQ c+8(FP), BX
+	MOVQ flag+16(FP), CX
+	MOVQ blocks_base+24(FP), SI
+	MOVQ blocks_len+32(FP), DI
+
+	MOVQ SP, DX
+	MOVQ SP, R9
+	ADDQ $31, R9
+	ANDQ $~31, R9
+	MOVQ R9, SP
+
+	MOVQ CX, 16(SP)
+	XORQ CX, CX
+	MOVQ CX, 24(SP)
+
+	VMOVDQU ·AVX2_c40<>(SB), Y4
+	VMOVDQU ·AVX2_c48<>(SB), Y5
+
+	VMOVDQU 0(AX), Y8
+	VMOVDQU 32(AX), Y9
+	VMOVDQU ·AVX2_iv0<>(SB), Y6
+	VMOVDQU ·AVX2_iv1<>(SB), Y7
+
+	MOVQ 0(BX), R8
+	MOVQ 8(BX), R9
+	MOVQ R9, 8(SP)
+
+loop:
+	ADDQ $128, R8
+	MOVQ R8, 0(SP)
+	CMPQ R8, $128
+	JGE  noinc
+	INCQ R9
+	MOVQ R9, 8(SP)
+
+noinc:
+	VMOVDQA Y8, Y0
+	VMOVDQA Y9, Y1
+	VMOVDQA Y6, Y2
+	VPXOR   0(SP), Y7, Y3
+
+	LOAD_MSG_AVX2_0_2_4_6_1_3_5_7_8_10_12_14_9_11_13_15()
+	VMOVDQA Y12, 32(SP)
+	VMOVDQA Y13, 64(SP)
+	VMOVDQA Y14, 96(SP)
+	VMOVDQA Y15, 128(SP)
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_14_4_9_13_10_8_15_6_1_0_11_5_12_2_7_3()
+	VMOVDQA Y12, 160(SP)
+	VMOVDQA Y13, 192(SP)
+	VMOVDQA Y14, 224(SP)
+	VMOVDQA Y15, 256(SP)
+
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_11_12_5_15_8_0_2_13_10_3_7_9_14_6_1_4()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_7_3_13_11_9_1_12_14_2_5_4_15_6_10_0_8()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_9_5_2_10_0_7_4_15_14_11_6_3_1_12_8_13()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_2_6_0_8_12_10_11_3_4_7_15_1_13_5_14_9()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_12_1_14_4_5_15_13_10_0_6_9_8_7_3_2_11()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_13_7_12_3_11_14_1_9_5_15_8_2_0_4_6_10()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_6_14_11_0_15_9_3_8_12_13_1_10_2_7_4_5()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_10_8_7_1_2_4_6_5_15_9_3_13_11_14_12_0()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+
+	ROUND_AVX2(32(SP), 64(SP), 96(SP), 128(SP), Y10, Y4, Y5)
+	ROUND_AVX2(160(SP), 192(SP), 224(SP), 256(SP), Y10, Y4, Y5)
+
+	VPXOR Y0, Y8, Y8
+	VPXOR Y1, Y9, Y9
+	VPXOR Y2, Y8, Y8
+	VPXOR Y3, Y9, Y9
+
+	LEAQ 128(SI), SI
+	SUBQ $128, DI
+	JNE  loop
+
+	MOVQ R8, 0(BX)
+	MOVQ R9, 8(BX)
+
+	VMOVDQU Y8, 0(AX)
+	VMOVDQU Y9, 32(AX)
+	VZEROUPPER
+
+	MOVQ DX, SP
+	RET
+
+#define VPUNPCKLQDQ_X2_X2_X15 BYTE $0xC5; BYTE $0x69; BYTE $0x6C; BYTE $0xFA
+#define VPUNPCKLQDQ_X3_X3_X15 BYTE $0xC5; BYTE $0x61; BYTE $0x6C; BYTE $0xFB
+#define VPUNPCKLQDQ_X7_X7_X15 BYTE $0xC5; BYTE $0x41; BYTE $0x6C; BYTE $0xFF
+#define VPUNPCKLQDQ_X13_X13_X15 BYTE $0xC4; BYTE $0x41; BYTE $0x11; BYTE $0x6C; BYTE $0xFD
+#define VPUNPCKLQDQ_X14_X14_X15 BYTE $0xC4; BYTE $0x41; BYTE $0x09; BYTE $0x6C; BYTE $0xFE
+
+#define VPUNPCKHQDQ_X15_X2_X2 BYTE $0xC4; BYTE $0xC1; BYTE $0x69; BYTE $0x6D; BYTE $0xD7
+#define VPUNPCKHQDQ_X15_X3_X3 BYTE $0xC4; BYTE $0xC1; BYTE $0x61; BYTE $0x6D; BYTE $0xDF
+#define VPUNPCKHQDQ_X15_X6_X6 BYTE $0xC4; BYTE $0xC1; BYTE $0x49; BYTE $0x6D; BYTE $0xF7
+#define VPUNPCKHQDQ_X15_X7_X7 BYTE $0xC4; BYTE $0xC1; BYTE $0x41; BYTE $0x6D; BYTE $0xFF
+#define VPUNPCKHQDQ_X15_X3_X2 BYTE $0xC4; BYTE $0xC1; BYTE $0x61; BYTE $0x6D; BYTE $0xD7
+#define VPUNPCKHQDQ_X15_X7_X6 BYTE $0xC4; BYTE $0xC1; BYTE $0x41; BYTE $0x6D; BYTE $0xF7
+#define VPUNPCKHQDQ_X15_X13_X3 BYTE $0xC4; BYTE $0xC1; BYTE $0x11; BYTE $0x6D; BYTE $0xDF
+#define VPUNPCKHQDQ_X15_X13_X7 BYTE $0xC4; BYTE $0xC1; BYTE $0x11; BYTE $0x6D; BYTE $0xFF
+
+#define SHUFFLE_AVX() \
+	VMOVDQA X6, X13;         \
+	VMOVDQA X2, X14;         \
+	VMOVDQA X4, X6;          \
+	VPUNPCKLQDQ_X13_X13_X15; \
+	VMOVDQA X5, X4;          \
+	VMOVDQA X6, X5;          \
+	VPUNPCKHQDQ_X15_X7_X6;   \
+	VPUNPCKLQDQ_X7_X7_X15;   \
+	VPUNPCKHQDQ_X15_X13_X7;  \
+	VPUNPCKLQDQ_X3_X3_X15;   \
+	VPUNPCKHQDQ_X15_X2_X2;   \
+	VPUNPCKLQDQ_X14_X14_X15; \
+	VPUNPCKHQDQ_X15_X3_X3;   \
+
+#define SHUFFLE_AVX_INV() \
+	VMOVDQA X2, X13;         \
+	VMOVDQA X4, X14;         \
+	VPUNPCKLQDQ_X2_X2_X15;   \
+	VMOVDQA X5, X4;          \
+	VPUNPCKHQDQ_X15_X3_X2;   \
+	VMOVDQA X14, X5;         \
+	VPUNPCKLQDQ_X3_X3_X15;   \
+	VMOVDQA X6, X14;         \
+	VPUNPCKHQDQ_X15_X13_X3;  \
+	VPUNPCKLQDQ_X7_X7_X15;   \
+	VPUNPCKHQDQ_X15_X6_X6;   \
+	VPUNPCKLQDQ_X14_X14_X15; \
+	VPUNPCKHQDQ_X15_X7_X7;   \
+
+#define HALF_ROUND_AVX(v0, v1, v2, v3, v4, v5, v6, v7, m0, m1, m2, m3, t0, c40, c48) \
+	VPADDQ  m0, v0, v0;   \
+	VPADDQ  v2, v0, v0;   \
+	VPADDQ  m1, v1, v1;   \
+	VPADDQ  v3, v1, v1;   \
+	VPXOR   v0, v6, v6;   \
+	VPXOR   v1, v7, v7;   \
+	VPSHUFD $-79, v6, v6; \
+	VPSHUFD $-79, v7, v7; \
+	VPADDQ  v6, v4, v4;   \
+	VPADDQ  v7, v5, v5;   \
+	VPXOR   v4, v2, v2;   \
+	VPXOR   v5, v3, v3;   \
+	VPSHUFB c40, v2, v2;  \
+	VPSHUFB c40, v3, v3;  \
+	VPADDQ  m2, v0, v0;   \
+	VPADDQ  v2, v0, v0;   \
+	VPADDQ  m3, v1, v1;   \
+	VPADDQ  v3, v1, v1;   \
+	VPXOR   v0, v6, v6;   \
+	VPXOR   v1, v7, v7;   \
+	VPSHUFB c48, v6, v6;  \
+	VPSHUFB c48, v7, v7;  \
+	VPADDQ  v6, v4, v4;   \
+	VPADDQ  v7, v5, v5;   \
+	VPXOR   v4, v2, v2;   \
+	VPXOR   v5, v3, v3;   \
+	VPADDQ  v2, v2, t0;   \
+	VPSRLQ  $63, v2, v2;  \
+	VPXOR   t0, v2, v2;   \
+	VPADDQ  v3, v3, t0;   \
+	VPSRLQ  $63, v3, v3;  \
+	VPXOR   t0, v3, v3
+
+// load msg: X12 = (i0, i1), X13 = (i2, i3), X14 = (i4, i5), X15 = (i6, i7)
+// i0, i1, i2, i3, i4, i5, i6, i7 must not be 0
+#define LOAD_MSG_AVX(i0, i1, i2, i3, i4, i5, i6, i7) \
+	VMOVQ_SI_X12(i0*8);     \
+	VMOVQ_SI_X13(i2*8);     \
+	VMOVQ_SI_X14(i4*8);     \
+	VMOVQ_SI_X15(i6*8);     \
+	VPINSRQ_1_SI_X12(i1*8); \
+	VPINSRQ_1_SI_X13(i3*8); \
+	VPINSRQ_1_SI_X14(i5*8); \
+	VPINSRQ_1_SI_X15(i7*8)
+
+// load msg: X12 = (0, 2), X13 = (4, 6), X14 = (1, 3), X15 = (5, 7)
+#define LOAD_MSG_AVX_0_2_4_6_1_3_5_7() \
+	VMOVQ_SI_X12_0;        \
+	VMOVQ_SI_X13(4*8);     \
+	VMOVQ_SI_X14(1*8);     \
+	VMOVQ_SI_X15(5*8);     \
+	VPINSRQ_1_SI_X12(2*8); \
+	VPINSRQ_1_SI_X13(6*8); \
+	VPINSRQ_1_SI_X14(3*8); \
+	VPINSRQ_1_SI_X15(7*8)
+
+// load msg: X12 = (1, 0), X13 = (11, 5), X14 = (12, 2), X15 = (7, 3)
+#define LOAD_MSG_AVX_1_0_11_5_12_2_7_3() \
+	VPSHUFD $0x4E, 0*8(SI), X12; \
+	VMOVQ_SI_X13(11*8);          \
+	VMOVQ_SI_X14(12*8);          \
+	VMOVQ_SI_X15(7*8);           \
+	VPINSRQ_1_SI_X13(5*8);       \
+	VPINSRQ_1_SI_X14(2*8);       \
+	VPINSRQ_1_SI_X15(3*8)
+
+// load msg: X12 = (11, 12), X13 = (5, 15), X14 = (8, 0), X15 = (2, 13)
+#define LOAD_MSG_AVX_11_12_5_15_8_0_2_13() \
+	VMOVDQU 11*8(SI), X12;  \
+	VMOVQ_SI_X13(5*8);      \
+	VMOVQ_SI_X14(8*8);      \
+	VMOVQ_SI_X15(2*8);      \
+	VPINSRQ_1_SI_X13(15*8); \
+	VPINSRQ_1_SI_X14_0;     \
+	VPINSRQ_1_SI_X15(13*8)
+
+// load msg: X12 = (2, 5), X13 = (4, 15), X14 = (6, 10), X15 = (0, 8)
+#define LOAD_MSG_AVX_2_5_4_15_6_10_0_8() \
+	VMOVQ_SI_X12(2*8);      \
+	VMOVQ_SI_X13(4*8);      \
+	VMOVQ_SI_X14(6*8);      \
+	VMOVQ_SI_X15_0;         \
+	VPINSRQ_1_SI_X12(5*8);  \
+	VPINSRQ_1_SI_X13(15*8); \
+	VPINSRQ_1_SI_X14(10*8); \
+	VPINSRQ_1_SI_X15(8*8)
+
+// load msg: X12 = (9, 5), X13 = (2, 10), X14 = (0, 7), X15 = (4, 15)
+#define LOAD_MSG_AVX_9_5_2_10_0_7_4_15() \
+	VMOVQ_SI_X12(9*8);      \
+	VMOVQ_SI_X13(2*8);      \
+	VMOVQ_SI_X14_0;         \
+	VMOVQ_SI_X15(4*8);      \
+	VPINSRQ_1_SI_X12(5*8);  \
+	VPINSRQ_1_SI_X13(10*8); \
+	VPINSRQ_1_SI_X14(7*8);  \
+	VPINSRQ_1_SI_X15(15*8)
+
+// load msg: X12 = (2, 6), X13 = (0, 8), X14 = (12, 10), X15 = (11, 3)
+#define LOAD_MSG_AVX_2_6_0_8_12_10_11_3() \
+	VMOVQ_SI_X12(2*8);      \
+	VMOVQ_SI_X13_0;         \
+	VMOVQ_SI_X14(12*8);     \
+	VMOVQ_SI_X15(11*8);     \
+	VPINSRQ_1_SI_X12(6*8);  \
+	VPINSRQ_1_SI_X13(8*8);  \
+	VPINSRQ_1_SI_X14(10*8); \
+	VPINSRQ_1_SI_X15(3*8)
+
+// load msg: X12 = (0, 6), X13 = (9, 8), X14 = (7, 3), X15 = (2, 11)
+#define LOAD_MSG_AVX_0_6_9_8_7_3_2_11() \
+	MOVQ    0*8(SI), X12;        \
+	VPSHUFD $0x4E, 8*8(SI), X13; \
+	MOVQ    7*8(SI), X14;        \
+	MOVQ    2*8(SI), X15;        \
+	VPINSRQ_1_SI_X12(6*8);       \
+	VPINSRQ_1_SI_X14(3*8);       \
+	VPINSRQ_1_SI_X15(11*8)
+
+// load msg: X12 = (6, 14), X13 = (11, 0), X14 = (15, 9), X15 = (3, 8)
+#define LOAD_MSG_AVX_6_14_11_0_15_9_3_8() \
+	MOVQ 6*8(SI), X12;      \
+	MOVQ 11*8(SI), X13;     \
+	MOVQ 15*8(SI), X14;     \
+	MOVQ 3*8(SI), X15;      \
+	VPINSRQ_1_SI_X12(14*8); \
+	VPINSRQ_1_SI_X13_0;     \
+	VPINSRQ_1_SI_X14(9*8);  \
+	VPINSRQ_1_SI_X15(8*8)
+
+// load msg: X12 = (5, 15), X13 = (8, 2), X14 = (0, 4), X15 = (6, 10)
+#define LOAD_MSG_AVX_5_15_8_2_0_4_6_10() \
+	MOVQ 5*8(SI), X12;      \
+	MOVQ 8*8(SI), X13;      \
+	MOVQ 0*8(SI), X14;      \
+	MOVQ 6*8(SI), X15;      \
+	VPINSRQ_1_SI_X12(15*8); \
+	VPINSRQ_1_SI_X13(2*8);  \
+	VPINSRQ_1_SI_X14(4*8);  \
+	VPINSRQ_1_SI_X15(10*8)
+
+// load msg: X12 = (12, 13), X13 = (1, 10), X14 = (2, 7), X15 = (4, 5)
+#define LOAD_MSG_AVX_12_13_1_10_2_7_4_5() \
+	VMOVDQU 12*8(SI), X12;  \
+	MOVQ    1*8(SI), X13;   \
+	MOVQ    2*8(SI), X14;   \
+	VPINSRQ_1_SI_X13(10*8); \
+	VPINSRQ_1_SI_X14(7*8);  \
+	VMOVDQU 4*8(SI), X15
+
+// load msg: X12 = (15, 9), X13 = (3, 13), X14 = (11, 14), X15 = (12, 0)
+#define LOAD_MSG_AVX_15_9_3_13_11_14_12_0() \
+	MOVQ 15*8(SI), X12;     \
+	MOVQ 3*8(SI), X13;      \
+	MOVQ 11*8(SI), X14;     \
+	MOVQ 12*8(SI), X15;     \
+	VPINSRQ_1_SI_X12(9*8);  \
+	VPINSRQ_1_SI_X13(13*8); \
+	VPINSRQ_1_SI_X14(14*8); \
+	VPINSRQ_1_SI_X15_0
+
+// func hashBlocksAVX(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+TEXT ·hashBlocksAVX(SB), 4, $288-48 // frame size = 272 + 16 byte alignment
+	MOVQ h+0(FP), AX
+	MOVQ c+8(FP), BX
+	MOVQ flag+16(FP), CX
+	MOVQ blocks_base+24(FP), SI
+	MOVQ blocks_len+32(FP), DI
+
+	MOVQ SP, BP
+	MOVQ SP, R9
+	ADDQ $15, R9
+	ANDQ $~15, R9
+	MOVQ R9, SP
+
+	VMOVDQU ·AVX_c40<>(SB), X0
+	VMOVDQU ·AVX_c48<>(SB), X1
+	VMOVDQA X0, X8
+	VMOVDQA X1, X9
+
+	VMOVDQU ·AVX_iv3<>(SB), X0
+	VMOVDQA X0, 0(SP)
+	XORQ    CX, 0(SP)          // 0(SP) = ·AVX_iv3 ^ (CX || 0)
+
+	VMOVDQU 0(AX), X10
+	VMOVDQU 16(AX), X11
+	VMOVDQU 32(AX), X2
+	VMOVDQU 48(AX), X3
+
+	MOVQ 0(BX), R8
+	MOVQ 8(BX), R9
+
+loop:
+	ADDQ $128, R8
+	CMPQ R8, $128
+	JGE  noinc
+	INCQ R9
+
+noinc:
+	VMOVQ_R8_X15
+	VPINSRQ_1_R9_X15
+
+	VMOVDQA X10, X0
+	VMOVDQA X11, X1
+	VMOVDQU ·AVX_iv0<>(SB), X4
+	VMOVDQU ·AVX_iv1<>(SB), X5
+	VMOVDQU ·AVX_iv2<>(SB), X6
+
+	VPXOR   X15, X6, X6
+	VMOVDQA 0(SP), X7
+
+	LOAD_MSG_AVX_0_2_4_6_1_3_5_7()
+	VMOVDQA X12, 16(SP)
+	VMOVDQA X13, 32(SP)
+	VMOVDQA X14, 48(SP)
+	VMOVDQA X15, 64(SP)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX(8, 10, 12, 14, 9, 11, 13, 15)
+	VMOVDQA X12, 80(SP)
+	VMOVDQA X13, 96(SP)
+	VMOVDQA X14, 112(SP)
+	VMOVDQA X15, 128(SP)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX(14, 4, 9, 13, 10, 8, 15, 6)
+	VMOVDQA X12, 144(SP)
+	VMOVDQA X13, 160(SP)
+	VMOVDQA X14, 176(SP)
+	VMOVDQA X15, 192(SP)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_1_0_11_5_12_2_7_3()
+	VMOVDQA X12, 208(SP)
+	VMOVDQA X13, 224(SP)
+	VMOVDQA X14, 240(SP)
+	VMOVDQA X15, 256(SP)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX_11_12_5_15_8_0_2_13()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX(10, 3, 7, 9, 14, 6, 1, 4)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX(7, 3, 13, 11, 9, 1, 12, 14)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_2_5_4_15_6_10_0_8()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX_9_5_2_10_0_7_4_15()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX(14, 11, 6, 3, 1, 12, 8, 13)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX_2_6_0_8_12_10_11_3()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX(4, 7, 15, 1, 13, 5, 14, 9)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX(12, 1, 14, 4, 5, 15, 13, 10)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_0_6_9_8_7_3_2_11()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX(13, 7, 12, 3, 11, 14, 1, 9)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_5_15_8_2_0_4_6_10()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX_6_14_11_0_15_9_3_8()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_12_13_1_10_2_7_4_5()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX(10, 8, 7, 1, 2, 4, 6, 5)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_15_9_3_13_11_14_12_0()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 16(SP), 32(SP), 48(SP), 64(SP), X15, X8, X9)
+	SHUFFLE_AVX()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 80(SP), 96(SP), 112(SP), 128(SP), X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 144(SP), 160(SP), 176(SP), 192(SP), X15, X8, X9)
+	SHUFFLE_AVX()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 208(SP), 224(SP), 240(SP), 256(SP), X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	VMOVDQU 32(AX), X14
+	VMOVDQU 48(AX), X15
+	VPXOR   X0, X10, X10
+	VPXOR   X1, X11, X11
+	VPXOR   X2, X14, X14
+	VPXOR   X3, X15, X15
+	VPXOR   X4, X10, X10
+	VPXOR   X5, X11, X11
+	VPXOR   X6, X14, X2
+	VPXOR   X7, X15, X3
+	VMOVDQU X2, 32(AX)
+	VMOVDQU X3, 48(AX)
+
+	LEAQ 128(SI), SI
+	SUBQ $128, DI
+	JNE  loop
+
+	VMOVDQU X10, 0(AX)
+	VMOVDQU X11, 16(AX)
+
+	MOVQ R8, 0(BX)
+	MOVQ R9, 8(BX)
+	VZEROUPPER
+
+	MOVQ BP, SP
+	RET

vendor/golang.org/x/crypto/blake2b/blake2b_amd64.go

@@ -0,0 +1,24 @@
+// Copyright 2016 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.
+
+// +build !go1.7,amd64,!gccgo,!appengine
+
+package blake2b
+
+import "golang.org/x/sys/cpu"
+
+func init() {
+	useSSE4 = cpu.X86.HasSSE41
+}
+
+//go:noescape
+func hashBlocksSSE4(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+
+func hashBlocks(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte) {
+	if useSSE4 {
+		hashBlocksSSE4(h, c, flag, blocks)
+	} else {
+		hashBlocksGeneric(h, c, flag, blocks)
+	}
+}

vendor/golang.org/x/crypto/blake2b/blake2b_amd64.s

@@ -0,0 +1,281 @@
+// Copyright 2016 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.
+
+// +build amd64,!gccgo,!appengine
+
+#include "textflag.h"
+
+DATA ·iv0<>+0x00(SB)/8, $0x6a09e667f3bcc908
+DATA ·iv0<>+0x08(SB)/8, $0xbb67ae8584caa73b
+GLOBL ·iv0<>(SB), (NOPTR+RODATA), $16
+
+DATA ·iv1<>+0x00(SB)/8, $0x3c6ef372fe94f82b
+DATA ·iv1<>+0x08(SB)/8, $0xa54ff53a5f1d36f1
+GLOBL ·iv1<>(SB), (NOPTR+RODATA), $16
+
+DATA ·iv2<>+0x00(SB)/8, $0x510e527fade682d1
+DATA ·iv2<>+0x08(SB)/8, $0x9b05688c2b3e6c1f
+GLOBL ·iv2<>(SB), (NOPTR+RODATA), $16
+
+DATA ·iv3<>+0x00(SB)/8, $0x1f83d9abfb41bd6b
+DATA ·iv3<>+0x08(SB)/8, $0x5be0cd19137e2179
+GLOBL ·iv3<>(SB), (NOPTR+RODATA), $16
+
+DATA ·c40<>+0x00(SB)/8, $0x0201000706050403
+DATA ·c40<>+0x08(SB)/8, $0x0a09080f0e0d0c0b
+GLOBL ·c40<>(SB), (NOPTR+RODATA), $16
+
+DATA ·c48<>+0x00(SB)/8, $0x0100070605040302
+DATA ·c48<>+0x08(SB)/8, $0x09080f0e0d0c0b0a
+GLOBL ·c48<>(SB), (NOPTR+RODATA), $16
+
+#define SHUFFLE(v2, v3, v4, v5, v6, v7, t1, t2) \
+	MOVO       v4, t1; \
+	MOVO       v5, v4; \
+	MOVO       t1, v5; \
+	MOVO       v6, t1; \
+	PUNPCKLQDQ v6, t2; \
+	PUNPCKHQDQ v7, v6; \
+	PUNPCKHQDQ t2, v6; \
+	PUNPCKLQDQ v7, t2; \
+	MOVO       t1, v7; \
+	MOVO       v2, t1; \
+	PUNPCKHQDQ t2, v7; \
+	PUNPCKLQDQ v3, t2; \
+	PUNPCKHQDQ t2, v2; \
+	PUNPCKLQDQ t1, t2; \
+	PUNPCKHQDQ t2, v3
+
+#define SHUFFLE_INV(v2, v3, v4, v5, v6, v7, t1, t2) \
+	MOVO       v4, t1; \
+	MOVO       v5, v4; \
+	MOVO       t1, v5; \
+	MOVO       v2, t1; \
+	PUNPCKLQDQ v2, t2; \
+	PUNPCKHQDQ v3, v2; \
+	PUNPCKHQDQ t2, v2; \
+	PUNPCKLQDQ v3, t2; \
+	MOVO       t1, v3; \
+	MOVO       v6, t1; \
+	PUNPCKHQDQ t2, v3; \
+	PUNPCKLQDQ v7, t2; \
+	PUNPCKHQDQ t2, v6; \
+	PUNPCKLQDQ t1, t2; \
+	PUNPCKHQDQ t2, v7
+
+#define HALF_ROUND(v0, v1, v2, v3, v4, v5, v6, v7, m0, m1, m2, m3, t0, c40, c48) \
+	PADDQ  m0, v0;        \
+	PADDQ  m1, v1;        \
+	PADDQ  v2, v0;        \
+	PADDQ  v3, v1;        \
+	PXOR   v0, v6;        \
+	PXOR   v1, v7;        \
+	PSHUFD $0xB1, v6, v6; \
+	PSHUFD $0xB1, v7, v7; \
+	PADDQ  v6, v4;        \
+	PADDQ  v7, v5;        \
+	PXOR   v4, v2;        \
+	PXOR   v5, v3;        \
+	PSHUFB c40, v2;       \
+	PSHUFB c40, v3;       \
+	PADDQ  m2, v0;        \
+	PADDQ  m3, v1;        \
+	PADDQ  v2, v0;        \
+	PADDQ  v3, v1;        \
+	PXOR   v0, v6;        \
+	PXOR   v1, v7;        \
+	PSHUFB c48, v6;       \
+	PSHUFB c48, v7;       \
+	PADDQ  v6, v4;        \
+	PADDQ  v7, v5;        \
+	PXOR   v4, v2;        \
+	PXOR   v5, v3;        \
+	MOVOU  v2, t0;        \
+	PADDQ  v2, t0;        \
+	PSRLQ  $63, v2;       \
+	PXOR   t0, v2;        \
+	MOVOU  v3, t0;        \
+	PADDQ  v3, t0;        \
+	PSRLQ  $63, v3;       \
+	PXOR   t0, v3
+
+#define LOAD_MSG(m0, m1, m2, m3, src, i0, i1, i2, i3, i4, i5, i6, i7) \
+	MOVQ   i0*8(src), m0;     \
+	PINSRQ $1, i1*8(src), m0; \
+	MOVQ   i2*8(src), m1;     \
+	PINSRQ $1, i3*8(src), m1; \
+	MOVQ   i4*8(src), m2;     \
+	PINSRQ $1, i5*8(src), m2; \
+	MOVQ   i6*8(src), m3;     \
+	PINSRQ $1, i7*8(src), m3
+
+// func hashBlocksSSE4(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+TEXT ·hashBlocksSSE4(SB), 4, $288-48 // frame size = 272 + 16 byte alignment
+	MOVQ h+0(FP), AX
+	MOVQ c+8(FP), BX
+	MOVQ flag+16(FP), CX
+	MOVQ blocks_base+24(FP), SI
+	MOVQ blocks_len+32(FP), DI
+
+	MOVQ SP, BP
+	MOVQ SP, R9
+	ADDQ $15, R9
+	ANDQ $~15, R9
+	MOVQ R9, SP
+
+	MOVOU ·iv3<>(SB), X0
+	MOVO  X0, 0(SP)
+	XORQ  CX, 0(SP)     // 0(SP) = ·iv3 ^ (CX || 0)
+
+	MOVOU ·c40<>(SB), X13
+	MOVOU ·c48<>(SB), X14
+
+	MOVOU 0(AX), X12
+	MOVOU 16(AX), X15
+
+	MOVQ 0(BX), R8
+	MOVQ 8(BX), R9
+
+loop:
+	ADDQ $128, R8
+	CMPQ R8, $128
+	JGE  noinc
+	INCQ R9
+
+noinc:
+	MOVQ R8, X8
+	PINSRQ $1, R9, X8
+
+	MOVO X12, X0
+	MOVO X15, X1
+	MOVOU 32(AX), X2
+	MOVOU 48(AX), X3
+	MOVOU ·iv0<>(SB), X4
+	MOVOU ·iv1<>(SB), X5
+	MOVOU ·iv2<>(SB), X6
+
+	PXOR X8, X6
+	MOVO 0(SP), X7
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 0, 2, 4, 6, 1, 3, 5, 7)
+	MOVO X8, 16(SP)
+	MOVO X9, 32(SP)
+	MOVO X10, 48(SP)
+	MOVO X11, 64(SP)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 8, 10, 12, 14, 9, 11, 13, 15)
+	MOVO X8, 80(SP)
+	MOVO X9, 96(SP)
+	MOVO X10, 112(SP)
+	MOVO X11, 128(SP)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 14, 4, 9, 13, 10, 8, 15, 6)
+	MOVO X8, 144(SP)
+	MOVO X9, 160(SP)
+	MOVO X10, 176(SP)
+	MOVO X11, 192(SP)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 1, 0, 11, 5, 12, 2, 7, 3)
+	MOVO X8, 208(SP)
+	MOVO X9, 224(SP)
+	MOVO X10, 240(SP)
+	MOVO X11, 256(SP)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 11, 12, 5, 15, 8, 0, 2, 13)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 10, 3, 7, 9, 14, 6, 1, 4)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 7, 3, 13, 11, 9, 1, 12, 14)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 2, 5, 4, 15, 6, 10, 0, 8)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 9, 5, 2, 10, 0, 7, 4, 15)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 14, 11, 6, 3, 1, 12, 8, 13)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 2, 6, 0, 8, 12, 10, 11, 3)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 4, 7, 15, 1, 13, 5, 14, 9)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 12, 1, 14, 4, 5, 15, 13, 10)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 0, 6, 9, 8, 7, 3, 2, 11)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 13, 7, 12, 3, 11, 14, 1, 9)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 5, 15, 8, 2, 0, 4, 6, 10)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 6, 14, 11, 0, 15, 9, 3, 8)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 12, 13, 1, 10, 2, 7, 4, 5)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 10, 8, 7, 1, 2, 4, 6, 5)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 15, 9, 3, 13, 11, 14, 12, 0)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 16(SP), 32(SP), 48(SP), 64(SP), X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 80(SP), 96(SP), 112(SP), 128(SP), X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 144(SP), 160(SP), 176(SP), 192(SP), X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 208(SP), 224(SP), 240(SP), 256(SP), X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	MOVOU 32(AX), X10
+	MOVOU 48(AX), X11
+	PXOR  X0, X12
+	PXOR  X1, X15
+	PXOR  X2, X10
+	PXOR  X3, X11
+	PXOR  X4, X12
+	PXOR  X5, X15
+	PXOR  X6, X10
+	PXOR  X7, X11
+	MOVOU X10, 32(AX)
+	MOVOU X11, 48(AX)
+
+	LEAQ 128(SI), SI
+	SUBQ $128, DI
+	JNE  loop
+
+	MOVOU X12, 0(AX)
+	MOVOU X15, 16(AX)
+
+	MOVQ R8, 0(BX)
+	MOVQ R9, 8(BX)
+
+	MOVQ BP, SP
+	RET

vendor/golang.org/x/crypto/blake2b/blake2b_generic.go

@@ -0,0 +1,182 @@
+// Copyright 2016 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 blake2b
+
+import (
+	"encoding/binary"
+	"math/bits"
+)
+
+// the precomputed values for BLAKE2b
+// there are 12 16-byte arrays - one for each round
+// the entries are calculated from the sigma constants.
+var precomputed = [12][16]byte{
+	{0, 2, 4, 6, 1, 3, 5, 7, 8, 10, 12, 14, 9, 11, 13, 15},
+	{14, 4, 9, 13, 10, 8, 15, 6, 1, 0, 11, 5, 12, 2, 7, 3},
+	{11, 12, 5, 15, 8, 0, 2, 13, 10, 3, 7, 9, 14, 6, 1, 4},
+	{7, 3, 13, 11, 9, 1, 12, 14, 2, 5, 4, 15, 6, 10, 0, 8},
+	{9, 5, 2, 10, 0, 7, 4, 15, 14, 11, 6, 3, 1, 12, 8, 13},
+	{2, 6, 0, 8, 12, 10, 11, 3, 4, 7, 15, 1, 13, 5, 14, 9},
+	{12, 1, 14, 4, 5, 15, 13, 10, 0, 6, 9, 8, 7, 3, 2, 11},
+	{13, 7, 12, 3, 11, 14, 1, 9, 5, 15, 8, 2, 0, 4, 6, 10},
+	{6, 14, 11, 0, 15, 9, 3, 8, 12, 13, 1, 10, 2, 7, 4, 5},
+	{10, 8, 7, 1, 2, 4, 6, 5, 15, 9, 3, 13, 11, 14, 12, 0},
+	{0, 2, 4, 6, 1, 3, 5, 7, 8, 10, 12, 14, 9, 11, 13, 15}, // equal to the first
+	{14, 4, 9, 13, 10, 8, 15, 6, 1, 0, 11, 5, 12, 2, 7, 3}, // equal to the second
+}
+
+func hashBlocksGeneric(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte) {
+	var m [16]uint64
+	c0, c1 := c[0], c[1]
+
+	for i := 0; i < len(blocks); {
+		c0 += BlockSize
+		if c0 < BlockSize {
+			c1++
+		}
+
+		v0, v1, v2, v3, v4, v5, v6, v7 := h[0], h[1], h[2], h[3], h[4], h[5], h[6], h[7]
+		v8, v9, v10, v11, v12, v13, v14, v15 := iv[0], iv[1], iv[2], iv[3], iv[4], iv[5], iv[6], iv[7]
+		v12 ^= c0
+		v13 ^= c1
+		v14 ^= flag
+
+		for j := range m {
+			m[j] = binary.LittleEndian.Uint64(blocks[i:])
+			i += 8
+		}
+
+		for j := range precomputed {
+			s := &(precomputed[j])
+
+			v0 += m[s[0]]
+			v0 += v4
+			v12 ^= v0
+			v12 = bits.RotateLeft64(v12, -32)
+			v8 += v12
+			v4 ^= v8
+			v4 = bits.RotateLeft64(v4, -24)
+			v1 += m[s[1]]
+			v1 += v5
+			v13 ^= v1
+			v13 = bits.RotateLeft64(v13, -32)
+			v9 += v13
+			v5 ^= v9
+			v5 = bits.RotateLeft64(v5, -24)
+			v2 += m[s[2]]
+			v2 += v6
+			v14 ^= v2
+			v14 = bits.RotateLeft64(v14, -32)
+			v10 += v14
+			v6 ^= v10
+			v6 = bits.RotateLeft64(v6, -24)
+			v3 += m[s[3]]
+			v3 += v7
+			v15 ^= v3
+			v15 = bits.RotateLeft64(v15, -32)
+			v11 += v15
+			v7 ^= v11
+			v7 = bits.RotateLeft64(v7, -24)
+
+			v0 += m[s[4]]
+			v0 += v4
+			v12 ^= v0
+			v12 = bits.RotateLeft64(v12, -16)
+			v8 += v12
+			v4 ^= v8
+			v4 = bits.RotateLeft64(v4, -63)
+			v1 += m[s[5]]
+			v1 += v5
+			v13 ^= v1
+			v13 = bits.RotateLeft64(v13, -16)
+			v9 += v13
+			v5 ^= v9
+			v5 = bits.RotateLeft64(v5, -63)
+			v2 += m[s[6]]
+			v2 += v6
+			v14 ^= v2
+			v14 = bits.RotateLeft64(v14, -16)
+			v10 += v14
+			v6 ^= v10
+			v6 = bits.RotateLeft64(v6, -63)
+			v3 += m[s[7]]
+			v3 += v7
+			v15 ^= v3
+			v15 = bits.RotateLeft64(v15, -16)
+			v11 += v15
+			v7 ^= v11
+			v7 = bits.RotateLeft64(v7, -63)
+
+			v0 += m[s[8]]
+			v0 += v5
+			v15 ^= v0
+			v15 = bits.RotateLeft64(v15, -32)
+			v10 += v15
+			v5 ^= v10
+			v5 = bits.RotateLeft64(v5, -24)
+			v1 += m[s[9]]
+			v1 += v6
+			v12 ^= v1
+			v12 = bits.RotateLeft64(v12, -32)
+			v11 += v12
+			v6 ^= v11
+			v6 = bits.RotateLeft64(v6, -24)
+			v2 += m[s[10]]
+			v2 += v7
+			v13 ^= v2
+			v13 = bits.RotateLeft64(v13, -32)
+			v8 += v13
+			v7 ^= v8
+			v7 = bits.RotateLeft64(v7, -24)
+			v3 += m[s[11]]
+			v3 += v4
+			v14 ^= v3
+			v14 = bits.RotateLeft64(v14, -32)
+			v9 += v14
+			v4 ^= v9
+			v4 = bits.RotateLeft64(v4, -24)
+
+			v0 += m[s[12]]
+			v0 += v5
+			v15 ^= v0
+			v15 = bits.RotateLeft64(v15, -16)
+			v10 += v15
+			v5 ^= v10
+			v5 = bits.RotateLeft64(v5, -63)
+			v1 += m[s[13]]
+			v1 += v6
+			v12 ^= v1
+			v12 = bits.RotateLeft64(v12, -16)
+			v11 += v12
+			v6 ^= v11
+			v6 = bits.RotateLeft64(v6, -63)
+			v2 += m[s[14]]
+			v2 += v7
+			v13 ^= v2
+			v13 = bits.RotateLeft64(v13, -16)
+			v8 += v13
+			v7 ^= v8
+			v7 = bits.RotateLeft64(v7, -63)
+			v3 += m[s[15]]
+			v3 += v4
+			v14 ^= v3
+			v14 = bits.RotateLeft64(v14, -16)
+			v9 += v14
+			v4 ^= v9
+			v4 = bits.RotateLeft64(v4, -63)
+
+		}
+
+		h[0] ^= v0 ^ v8
+		h[1] ^= v1 ^ v9
+		h[2] ^= v2 ^ v10
+		h[3] ^= v3 ^ v11
+		h[4] ^= v4 ^ v12
+		h[5] ^= v5 ^ v13
+		h[6] ^= v6 ^ v14
+		h[7] ^= v7 ^ v15
+	}
+	c[0], c[1] = c0, c1
+}

vendor/golang.org/x/crypto/blake2b/blake2b_ref.go

@@ -0,0 +1,11 @@
+// Copyright 2016 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.
+
+// +build !amd64 appengine gccgo
+
+package blake2b
+
+func hashBlocks(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte) {
+	hashBlocksGeneric(h, c, flag, blocks)
+}

vendor/golang.org/x/crypto/blake2b/blake2x.go

@@ -0,0 +1,177 @@
+// Copyright 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 blake2b
+
+import (
+	"encoding/binary"
+	"errors"
+	"io"
+)
+
+// XOF defines the interface to hash functions that
+// support arbitrary-length output.
+type XOF interface {
+	// Write absorbs more data into the hash's state. It panics if called
+	// after Read.
+	io.Writer
+
+	// Read reads more output from the hash. It returns io.EOF if the limit
+	// has been reached.
+	io.Reader
+
+	// Clone returns a copy of the XOF in its current state.
+	Clone() XOF
+
+	// Reset resets the XOF to its initial state.
+	Reset()
+}
+
+// OutputLengthUnknown can be used as the size argument to NewXOF to indicate
+// the length of the output is not known in advance.
+const OutputLengthUnknown = 0
+
+// magicUnknownOutputLength is a magic value for the output size that indicates
+// an unknown number of output bytes.
+const magicUnknownOutputLength = (1 << 32) - 1
+
+// maxOutputLength is the absolute maximum number of bytes to produce when the
+// number of output bytes is unknown.
+const maxOutputLength = (1 << 32) * 64
+
+// NewXOF creates a new variable-output-length hash. The hash either produce a
+// known number of bytes (1 <= size < 2**32-1), or an unknown number of bytes
+// (size == OutputLengthUnknown). In the latter case, an absolute limit of
+// 256GiB applies.
+//
+// A non-nil key turns the hash into a MAC. The key must between
+// zero and 32 bytes long.
+func NewXOF(size uint32, key []byte) (XOF, error) {
+	if len(key) > Size {
+		return nil, errKeySize
+	}
+	if size == magicUnknownOutputLength {
+		// 2^32-1 indicates an unknown number of bytes and thus isn't a
+		// valid length.
+		return nil, errors.New("blake2b: XOF length too large")
+	}
+	if size == OutputLengthUnknown {
+		size = magicUnknownOutputLength
+	}
+	x := &xof{
+		d: digest{
+			size:   Size,
+			keyLen: len(key),
+		},
+		length: size,
+	}
+	copy(x.d.key[:], key)
+	x.Reset()
+	return x, nil
+}
+
+type xof struct {
+	d                digest
+	length           uint32
+	remaining        uint64
+	cfg, root, block [Size]byte
+	offset           int
+	nodeOffset       uint32
+	readMode         bool
+}
+
+func (x *xof) Write(p []byte) (n int, err error) {
+	if x.readMode {
+		panic("blake2b: write to XOF after read")
+	}
+	return x.d.Write(p)
+}
+
+func (x *xof) Clone() XOF {
+	clone := *x
+	return &clone
+}
+
+func (x *xof) Reset() {
+	x.cfg[0] = byte(Size)
+	binary.LittleEndian.PutUint32(x.cfg[4:], uint32(Size)) // leaf length
+	binary.LittleEndian.PutUint32(x.cfg[12:], x.length)    // XOF length
+	x.cfg[17] = byte(Size)                                 // inner hash size
+
+	x.d.Reset()
+	x.d.h[1] ^= uint64(x.length) << 32
+
+	x.remaining = uint64(x.length)
+	if x.remaining == magicUnknownOutputLength {
+		x.remaining = maxOutputLength
+	}
+	x.offset, x.nodeOffset = 0, 0
+	x.readMode = false
+}
+
+func (x *xof) Read(p []byte) (n int, err error) {
+	if !x.readMode {
+		x.d.finalize(&x.root)
+		x.readMode = true
+	}
+
+	if x.remaining == 0 {
+		return 0, io.EOF
+	}
+
+	n = len(p)
+	if uint64(n) > x.remaining {
+		n = int(x.remaining)
+		p = p[:n]
+	}
+
+	if x.offset > 0 {
+		blockRemaining := Size - x.offset
+		if n < blockRemaining {
+			x.offset += copy(p, x.block[x.offset:])
+			x.remaining -= uint64(n)
+			return
+		}
+		copy(p, x.block[x.offset:])
+		p = p[blockRemaining:]
+		x.offset = 0
+		x.remaining -= uint64(blockRemaining)
+	}
+
+	for len(p) >= Size {
+		binary.LittleEndian.PutUint32(x.cfg[8:], x.nodeOffset)
+		x.nodeOffset++
+
+		x.d.initConfig(&x.cfg)
+		x.d.Write(x.root[:])
+		x.d.finalize(&x.block)
+
+		copy(p, x.block[:])
+		p = p[Size:]
+		x.remaining -= uint64(Size)
+	}
+
+	if todo := len(p); todo > 0 {
+		if x.remaining < uint64(Size) {
+			x.cfg[0] = byte(x.remaining)
+		}
+		binary.LittleEndian.PutUint32(x.cfg[8:], x.nodeOffset)
+		x.nodeOffset++
+
+		x.d.initConfig(&x.cfg)
+		x.d.Write(x.root[:])
+		x.d.finalize(&x.block)
+
+		x.offset = copy(p, x.block[:todo])
+		x.remaining -= uint64(todo)
+	}
+	return
+}
+
+func (d *digest) initConfig(cfg *[Size]byte) {
+	d.offset, d.c[0], d.c[1] = 0, 0, 0
+	for i := range d.h {
+		d.h[i] = iv[i] ^ binary.LittleEndian.Uint64(cfg[i*8:])
+	}
+}

vendor/golang.org/x/crypto/blake2b/register.go

@@ -0,0 +1,32 @@
+// Copyright 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.
+
+// +build go1.9
+
+package blake2b
+
+import (
+	"crypto"
+	"hash"
+)
+
+func init() {
+	newHash256 := func() hash.Hash {
+		h, _ := New256(nil)
+		return h
+	}
+	newHash384 := func() hash.Hash {
+		h, _ := New384(nil)
+		return h
+	}
+
+	newHash512 := func() hash.Hash {
+		h, _ := New512(nil)
+		return h
+	}
+
+	crypto.RegisterHash(crypto.BLAKE2b_256, newHash256)
+	crypto.RegisterHash(crypto.BLAKE2b_384, newHash384)
+	crypto.RegisterHash(crypto.BLAKE2b_512, newHash512)
+}

vendor/golang.org/x/crypto/blowfish/block.go

@@ -0,0 +1,159 @@
+// Copyright 2010 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 blowfish
+
+// getNextWord returns the next big-endian uint32 value from the byte slice
+// at the given position in a circular manner, updating the position.
+func getNextWord(b []byte, pos *int) uint32 {
+	var w uint32
+	j := *pos
+	for i := 0; i < 4; i++ {
+		w = w<<8 | uint32(b[j])
+		j++
+		if j >= len(b) {
+			j = 0
+		}
+	}
+	*pos = j
+	return w
+}
+
+// ExpandKey performs a key expansion on the given *Cipher. Specifically, it
+// performs the Blowfish algorithm's key schedule which sets up the *Cipher's
+// pi and substitution tables for calls to Encrypt. This is used, primarily,
+// by the bcrypt package to reuse the Blowfish key schedule during its
+// set up. It's unlikely that you need to use this directly.
+func ExpandKey(key []byte, c *Cipher) {
+	j := 0
+	for i := 0; i < 18; i++ {
+		// Using inlined getNextWord for performance.
+		var d uint32
+		for k := 0; k < 4; k++ {
+			d = d<<8 | uint32(key[j])
+			j++
+			if j >= len(key) {
+				j = 0
+			}
+		}
+		c.p[i] ^= d
+	}
+
+	var l, r uint32
+	for i := 0; i < 18; i += 2 {
+		l, r = encryptBlock(l, r, c)
+		c.p[i], c.p[i+1] = l, r
+	}
+
+	for i := 0; i < 256; i += 2 {
+		l, r = encryptBlock(l, r, c)
+		c.s0[i], c.s0[i+1] = l, r
+	}
+	for i := 0; i < 256; i += 2 {
+		l, r = encryptBlock(l, r, c)
+		c.s1[i], c.s1[i+1] = l, r
+	}
+	for i := 0; i < 256; i += 2 {
+		l, r = encryptBlock(l, r, c)
+		c.s2[i], c.s2[i+1] = l, r
+	}
+	for i := 0; i < 256; i += 2 {
+		l, r = encryptBlock(l, r, c)
+		c.s3[i], c.s3[i+1] = l, r
+	}
+}
+
+// This is similar to ExpandKey, but folds the salt during the key
+// schedule. While ExpandKey is essentially expandKeyWithSalt with an all-zero
+// salt passed in, reusing ExpandKey turns out to be a place of inefficiency
+// and specializing it here is useful.
+func expandKeyWithSalt(key []byte, salt []byte, c *Cipher) {
+	j := 0
+	for i := 0; i < 18; i++ {
+		c.p[i] ^= getNextWord(key, &j)
+	}
+
+	j = 0
+	var l, r uint32
+	for i := 0; i < 18; i += 2 {
+		l ^= getNextWord(salt, &j)
+		r ^= getNextWord(salt, &j)
+		l, r = encryptBlock(l, r, c)
+		c.p[i], c.p[i+1] = l, r
+	}
+
+	for i := 0; i < 256; i += 2 {
+		l ^= getNextWord(salt, &j)
+		r ^= getNextWord(salt, &j)
+		l, r = encryptBlock(l, r, c)
+		c.s0[i], c.s0[i+1] = l, r
+	}
+
+	for i := 0; i < 256; i += 2 {
+		l ^= getNextWord(salt, &j)
+		r ^= getNextWord(salt, &j)
+		l, r = encryptBlock(l, r, c)
+		c.s1[i], c.s1[i+1] = l, r
+	}
+
+	for i := 0; i < 256; i += 2 {
+		l ^= getNextWord(salt, &j)
+		r ^= getNextWord(salt, &j)
+		l, r = encryptBlock(l, r, c)
+		c.s2[i], c.s2[i+1] = l, r
+	}
+
+	for i := 0; i < 256; i += 2 {
+		l ^= getNextWord(salt, &j)
+		r ^= getNextWord(salt, &j)
+		l, r = encryptBlock(l, r, c)
+		c.s3[i], c.s3[i+1] = l, r
+	}
+}
+
+func encryptBlock(l, r uint32, c *Cipher) (uint32, uint32) {
+	xl, xr := l, r
+	xl ^= c.p[0]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[1]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[2]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[3]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[4]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[5]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[6]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[7]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[8]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[9]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[10]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[11]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[12]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[13]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[14]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[15]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[16]
+	xr ^= c.p[17]
+	return xr, xl
+}
+
+func decryptBlock(l, r uint32, c *Cipher) (uint32, uint32) {
+	xl, xr := l, r
+	xl ^= c.p[17]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[16]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[15]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[14]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[13]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[12]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[11]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[10]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[9]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[8]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[7]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[6]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[5]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[4]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[3]
+	xr ^= ((c.s0[byte(xl>>24)] + c.s1[byte(xl>>16)]) ^ c.s2[byte(xl>>8)]) + c.s3[byte(xl)] ^ c.p[2]
+	xl ^= ((c.s0[byte(xr>>24)] + c.s1[byte(xr>>16)]) ^ c.s2[byte(xr>>8)]) + c.s3[byte(xr)] ^ c.p[1]
+	xr ^= c.p[0]
+	return xr, xl
+}

vendor/golang.org/x/crypto/blowfish/cipher.go

@@ -0,0 +1,99 @@
+// Copyright 2010 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 blowfish implements Bruce Schneier's Blowfish encryption algorithm.
+//
+// Blowfish is a legacy cipher and its short block size makes it vulnerable to
+// birthday bound attacks (see https://sweet32.info). It should only be used
+// where compatibility with legacy systems, not security, is the goal.
+//
+// Deprecated: any new system should use AES (from crypto/aes, if necessary in
+// an AEAD mode like crypto/cipher.NewGCM) or XChaCha20-Poly1305 (from
+// golang.org/x/crypto/chacha20poly1305).
+package blowfish // import "golang.org/x/crypto/blowfish"
+
+// The code is a port of Bruce Schneier's C implementation.
+// See https://www.schneier.com/blowfish.html.
+
+import "strconv"
+
+// The Blowfish block size in bytes.
+const BlockSize = 8
+
+// A Cipher is an instance of Blowfish encryption using a particular key.
+type Cipher struct {
+	p              [18]uint32
+	s0, s1, s2, s3 [256]uint32
+}
+
+type KeySizeError int
+
+func (k KeySizeError) Error() string {
+	return "crypto/blowfish: invalid key size " + strconv.Itoa(int(k))
+}
+
+// NewCipher creates and returns a Cipher.
+// The key argument should be the Blowfish key, from 1 to 56 bytes.
+func NewCipher(key []byte) (*Cipher, error) {
+	var result Cipher
+	if k := len(key); k < 1 || k > 56 {
+		return nil, KeySizeError(k)
+	}
+	initCipher(&result)
+	ExpandKey(key, &result)
+	return &result, nil
+}
+
+// NewSaltedCipher creates a returns a Cipher that folds a salt into its key
+// schedule. For most purposes, NewCipher, instead of NewSaltedCipher, is
+// sufficient and desirable. For bcrypt compatibility, the key can be over 56
+// bytes.
+func NewSaltedCipher(key, salt []byte) (*Cipher, error) {
+	if len(salt) == 0 {
+		return NewCipher(key)
+	}
+	var result Cipher
+	if k := len(key); k < 1 {
+		return nil, KeySizeError(k)
+	}
+	initCipher(&result)
+	expandKeyWithSalt(key, salt, &result)
+	return &result, nil
+}
+
+// BlockSize returns the Blowfish block size, 8 bytes.
+// It is necessary to satisfy the Block interface in the
+// package "crypto/cipher".
+func (c *Cipher) BlockSize() int { return BlockSize }
+
+// Encrypt encrypts the 8-byte buffer src using the key k
+// and stores the result in dst.
+// Note that for amounts of data larger than a block,
+// it is not safe to just call Encrypt on successive blocks;
+// instead, use an encryption mode like CBC (see crypto/cipher/cbc.go).
+func (c *Cipher) Encrypt(dst, src []byte) {
+	l := uint32(src[0])<<24 | uint32(src[1])<<16 | uint32(src[2])<<8 | uint32(src[3])
+	r := uint32(src[4])<<24 | uint32(src[5])<<16 | uint32(src[6])<<8 | uint32(src[7])
+	l, r = encryptBlock(l, r, c)
+	dst[0], dst[1], dst[2], dst[3] = byte(l>>24), byte(l>>16), byte(l>>8), byte(l)
+	dst[4], dst[5], dst[6], dst[7] = byte(r>>24), byte(r>>16), byte(r>>8), byte(r)
+}
+
+// Decrypt decrypts the 8-byte buffer src using the key k
+// and stores the result in dst.
+func (c *Cipher) Decrypt(dst, src []byte) {
+	l := uint32(src[0])<<24 | uint32(src[1])<<16 | uint32(src[2])<<8 | uint32(src[3])
+	r := uint32(src[4])<<24 | uint32(src[5])<<16 | uint32(src[6])<<8 | uint32(src[7])
+	l, r = decryptBlock(l, r, c)
+	dst[0], dst[1], dst[2], dst[3] = byte(l>>24), byte(l>>16), byte(l>>8), byte(l)
+	dst[4], dst[5], dst[6], dst[7] = byte(r>>24), byte(r>>16), byte(r>>8), byte(r)
+}
+
+func initCipher(c *Cipher) {
+	copy(c.p[0:], p[0:])
+	copy(c.s0[0:], s0[0:])
+	copy(c.s1[0:], s1[0:])
+	copy(c.s2[0:], s2[0:])
+	copy(c.s3[0:], s3[0:])
+}

vendor/golang.org/x/crypto/blowfish/const.go

@@ -0,0 +1,199 @@
+// Copyright 2010 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.
+
+// The startup permutation array and substitution boxes.
+// They are the hexadecimal digits of PI; see:
+// https://www.schneier.com/code/constants.txt.
+
+package blowfish
+
+var s0 = [256]uint32{
+	0xd1310ba6, 0x98dfb5ac, 0x2ffd72db, 0xd01adfb7, 0xb8e1afed, 0x6a267e96,
+	0xba7c9045, 0xf12c7f99, 0x24a19947, 0xb3916cf7, 0x0801f2e2, 0x858efc16,
+	0x636920d8, 0x71574e69, 0xa458fea3, 0xf4933d7e, 0x0d95748f, 0x728eb658,
+	0x718bcd58, 0x82154aee, 0x7b54a41d, 0xc25a59b5, 0x9c30d539, 0x2af26013,
+	0xc5d1b023, 0x286085f0, 0xca417918, 0xb8db38ef, 0x8e79dcb0, 0x603a180e,
+	0x6c9e0e8b, 0xb01e8a3e, 0xd71577c1, 0xbd314b27, 0x78af2fda, 0x55605c60,
+	0xe65525f3, 0xaa55ab94, 0x57489862, 0x63e81440, 0x55ca396a, 0x2aab10b6,
+	0xb4cc5c34, 0x1141e8ce, 0xa15486af, 0x7c72e993, 0xb3ee1411, 0x636fbc2a,
+	0x2ba9c55d, 0x741831f6, 0xce5c3e16, 0x9b87931e, 0xafd6ba33, 0x6c24cf5c,
+	0x7a325381, 0x28958677, 0x3b8f4898, 0x6b4bb9af, 0xc4bfe81b, 0x66282193,
+	0x61d809cc, 0xfb21a991, 0x487cac60, 0x5dec8032, 0xef845d5d, 0xe98575b1,
+	0xdc262302, 0xeb651b88, 0x23893e81, 0xd396acc5, 0x0f6d6ff3, 0x83f44239,
+	0x2e0b4482, 0xa4842004, 0x69c8f04a, 0x9e1f9b5e, 0x21c66842, 0xf6e96c9a,
+	0x670c9c61, 0xabd388f0, 0x6a51a0d2, 0xd8542f68, 0x960fa728, 0xab5133a3,
+	0x6eef0b6c, 0x137a3be4, 0xba3bf050, 0x7efb2a98, 0xa1f1651d, 0x39af0176,
+	0x66ca593e, 0x82430e88, 0x8cee8619, 0x456f9fb4, 0x7d84a5c3, 0x3b8b5ebe,
+	0xe06f75d8, 0x85c12073, 0x401a449f, 0x56c16aa6, 0x4ed3aa62, 0x363f7706,
+	0x1bfedf72, 0x429b023d, 0x37d0d724, 0xd00a1248, 0xdb0fead3, 0x49f1c09b,
+	0x075372c9, 0x80991b7b, 0x25d479d8, 0xf6e8def7, 0xe3fe501a, 0xb6794c3b,
+	0x976ce0bd, 0x04c006ba, 0xc1a94fb6, 0x409f60c4, 0x5e5c9ec2, 0x196a2463,
+	0x68fb6faf, 0x3e6c53b5, 0x1339b2eb, 0x3b52ec6f, 0x6dfc511f, 0x9b30952c,
+	0xcc814544, 0xaf5ebd09, 0xbee3d004, 0xde334afd, 0x660f2807, 0x192e4bb3,
+	0xc0cba857, 0x45c8740f, 0xd20b5f39, 0xb9d3fbdb, 0x5579c0bd, 0x1a60320a,
+	0xd6a100c6, 0x402c7279, 0x679f25fe, 0xfb1fa3cc, 0x8ea5e9f8, 0xdb3222f8,
+	0x3c7516df, 0xfd616b15, 0x2f501ec8, 0xad0552ab, 0x323db5fa, 0xfd238760,
+	0x53317b48, 0x3e00df82, 0x9e5c57bb, 0xca6f8ca0, 0x1a87562e, 0xdf1769db,
+	0xd542a8f6, 0x287effc3, 0xac6732c6, 0x8c4f5573, 0x695b27b0, 0xbbca58c8,
+	0xe1ffa35d, 0xb8f011a0, 0x10fa3d98, 0xfd2183b8, 0x4afcb56c, 0x2dd1d35b,
+	0x9a53e479, 0xb6f84565, 0xd28e49bc, 0x4bfb9790, 0xe1ddf2da, 0xa4cb7e33,
+	0x62fb1341, 0xcee4c6e8, 0xef20cada, 0x36774c01, 0xd07e9efe, 0x2bf11fb4,
+	0x95dbda4d, 0xae909198, 0xeaad8e71, 0x6b93d5a0, 0xd08ed1d0, 0xafc725e0,
+	0x8e3c5b2f, 0x8e7594b7, 0x8ff6e2fb, 0xf2122b64, 0x8888b812, 0x900df01c,
+	0x4fad5ea0, 0x688fc31c, 0xd1cff191, 0xb3a8c1ad, 0x2f2f2218, 0xbe0e1777,
+	0xea752dfe, 0x8b021fa1, 0xe5a0cc0f, 0xb56f74e8, 0x18acf3d6, 0xce89e299,
+	0xb4a84fe0, 0xfd13e0b7, 0x7cc43b81, 0xd2ada8d9, 0x165fa266, 0x80957705,
+	0x93cc7314, 0x211a1477, 0xe6ad2065, 0x77b5fa86, 0xc75442f5, 0xfb9d35cf,
+	0xebcdaf0c, 0x7b3e89a0, 0xd6411bd3, 0xae1e7e49, 0x00250e2d, 0x2071b35e,
+	0x226800bb, 0x57b8e0af, 0x2464369b, 0xf009b91e, 0x5563911d, 0x59dfa6aa,
+	0x78c14389, 0xd95a537f, 0x207d5ba2, 0x02e5b9c5, 0x83260376, 0x6295cfa9,
+	0x11c81968, 0x4e734a41, 0xb3472dca, 0x7b14a94a, 0x1b510052, 0x9a532915,
+	0xd60f573f, 0xbc9bc6e4, 0x2b60a476, 0x81e67400, 0x08ba6fb5, 0x571be91f,
+	0xf296ec6b, 0x2a0dd915, 0xb6636521, 0xe7b9f9b6, 0xff34052e, 0xc5855664,
+	0x53b02d5d, 0xa99f8fa1, 0x08ba4799, 0x6e85076a,
+}
+
+var s1 = [256]uint32{
+	0x4b7a70e9, 0xb5b32944, 0xdb75092e, 0xc4192623, 0xad6ea6b0, 0x49a7df7d,
+	0x9cee60b8, 0x8fedb266, 0xecaa8c71, 0x699a17ff, 0x5664526c, 0xc2b19ee1,
+	0x193602a5, 0x75094c29, 0xa0591340, 0xe4183a3e, 0x3f54989a, 0x5b429d65,
+	0x6b8fe4d6, 0x99f73fd6, 0xa1d29c07, 0xefe830f5, 0x4d2d38e6, 0xf0255dc1,
+	0x4cdd2086, 0x8470eb26, 0x6382e9c6, 0x021ecc5e, 0x09686b3f, 0x3ebaefc9,
+	0x3c971814, 0x6b6a70a1, 0x687f3584, 0x52a0e286, 0xb79c5305, 0xaa500737,
+	0x3e07841c, 0x7fdeae5c, 0x8e7d44ec, 0x5716f2b8, 0xb03ada37, 0xf0500c0d,
+	0xf01c1f04, 0x0200b3ff, 0xae0cf51a, 0x3cb574b2, 0x25837a58, 0xdc0921bd,
+	0xd19113f9, 0x7ca92ff6, 0x94324773, 0x22f54701, 0x3ae5e581, 0x37c2dadc,
+	0xc8b57634, 0x9af3dda7, 0xa9446146, 0x0fd0030e, 0xecc8c73e, 0xa4751e41,
+	0xe238cd99, 0x3bea0e2f, 0x3280bba1, 0x183eb331, 0x4e548b38, 0x4f6db908,
+	0x6f420d03, 0xf60a04bf, 0x2cb81290, 0x24977c79, 0x5679b072, 0xbcaf89af,
+	0xde9a771f, 0xd9930810, 0xb38bae12, 0xdccf3f2e, 0x5512721f, 0x2e6b7124,
+	0x501adde6, 0x9f84cd87, 0x7a584718, 0x7408da17, 0xbc9f9abc, 0xe94b7d8c,
+	0xec7aec3a, 0xdb851dfa, 0x63094366, 0xc464c3d2, 0xef1c1847, 0x3215d908,
+	0xdd433b37, 0x24c2ba16, 0x12a14d43, 0x2a65c451, 0x50940002, 0x133ae4dd,
+	0x71dff89e, 0x10314e55, 0x81ac77d6, 0x5f11199b, 0x043556f1, 0xd7a3c76b,
+	0x3c11183b, 0x5924a509, 0xf28fe6ed, 0x97f1fbfa, 0x9ebabf2c, 0x1e153c6e,
+	0x86e34570, 0xeae96fb1, 0x860e5e0a, 0x5a3e2ab3, 0x771fe71c, 0x4e3d06fa,
+	0x2965dcb9, 0x99e71d0f, 0x803e89d6, 0x5266c825, 0x2e4cc978, 0x9c10b36a,
+	0xc6150eba, 0x94e2ea78, 0xa5fc3c53, 0x1e0a2df4, 0xf2f74ea7, 0x361d2b3d,
+	0x1939260f, 0x19c27960, 0x5223a708, 0xf71312b6, 0xebadfe6e, 0xeac31f66,
+	0xe3bc4595, 0xa67bc883, 0xb17f37d1, 0x018cff28, 0xc332ddef, 0xbe6c5aa5,
+	0x65582185, 0x68ab9802, 0xeecea50f, 0xdb2f953b, 0x2aef7dad, 0x5b6e2f84,
+	0x1521b628, 0x29076170, 0xecdd4775, 0x619f1510, 0x13cca830, 0xeb61bd96,
+	0x0334fe1e, 0xaa0363cf, 0xb5735c90, 0x4c70a239, 0xd59e9e0b, 0xcbaade14,
+	0xeecc86bc, 0x60622ca7, 0x9cab5cab, 0xb2f3846e, 0x648b1eaf, 0x19bdf0ca,
+	0xa02369b9, 0x655abb50, 0x40685a32, 0x3c2ab4b3, 0x319ee9d5, 0xc021b8f7,
+	0x9b540b19, 0x875fa099, 0x95f7997e, 0x623d7da8, 0xf837889a, 0x97e32d77,
+	0x11ed935f, 0x16681281, 0x0e358829, 0xc7e61fd6, 0x96dedfa1, 0x7858ba99,
+	0x57f584a5, 0x1b227263, 0x9b83c3ff, 0x1ac24696, 0xcdb30aeb, 0x532e3054,
+	0x8fd948e4, 0x6dbc3128, 0x58ebf2ef, 0x34c6ffea, 0xfe28ed61, 0xee7c3c73,
+	0x5d4a14d9, 0xe864b7e3, 0x42105d14, 0x203e13e0, 0x45eee2b6, 0xa3aaabea,
+	0xdb6c4f15, 0xfacb4fd0, 0xc742f442, 0xef6abbb5, 0x654f3b1d, 0x41cd2105,
+	0xd81e799e, 0x86854dc7, 0xe44b476a, 0x3d816250, 0xcf62a1f2, 0x5b8d2646,
+	0xfc8883a0, 0xc1c7b6a3, 0x7f1524c3, 0x69cb7492, 0x47848a0b, 0x5692b285,
+	0x095bbf00, 0xad19489d, 0x1462b174, 0x23820e00, 0x58428d2a, 0x0c55f5ea,
+	0x1dadf43e, 0x233f7061, 0x3372f092, 0x8d937e41, 0xd65fecf1, 0x6c223bdb,
+	0x7cde3759, 0xcbee7460, 0x4085f2a7, 0xce77326e, 0xa6078084, 0x19f8509e,
+	0xe8efd855, 0x61d99735, 0xa969a7aa, 0xc50c06c2, 0x5a04abfc, 0x800bcadc,
+	0x9e447a2e, 0xc3453484, 0xfdd56705, 0x0e1e9ec9, 0xdb73dbd3, 0x105588cd,
+	0x675fda79, 0xe3674340, 0xc5c43465, 0x713e38d8, 0x3d28f89e, 0xf16dff20,
+	0x153e21e7, 0x8fb03d4a, 0xe6e39f2b, 0xdb83adf7,
+}
+
+var s2 = [256]uint32{
+	0xe93d5a68, 0x948140f7, 0xf64c261c, 0x94692934, 0x411520f7, 0x7602d4f7,
+	0xbcf46b2e, 0xd4a20068, 0xd4082471, 0x3320f46a, 0x43b7d4b7, 0x500061af,
+	0x1e39f62e, 0x97244546, 0x14214f74, 0xbf8b8840, 0x4d95fc1d, 0x96b591af,
+	0x70f4ddd3, 0x66a02f45, 0xbfbc09ec, 0x03bd9785, 0x7fac6dd0, 0x31cb8504,
+	0x96eb27b3, 0x55fd3941, 0xda2547e6, 0xabca0a9a, 0x28507825, 0x530429f4,
+	0x0a2c86da, 0xe9b66dfb, 0x68dc1462, 0xd7486900, 0x680ec0a4, 0x27a18dee,
+	0x4f3ffea2, 0xe887ad8c, 0xb58ce006, 0x7af4d6b6, 0xaace1e7c, 0xd3375fec,
+	0xce78a399, 0x406b2a42, 0x20fe9e35, 0xd9f385b9, 0xee39d7ab, 0x3b124e8b,
+	0x1dc9faf7, 0x4b6d1856, 0x26a36631, 0xeae397b2, 0x3a6efa74, 0xdd5b4332,
+	0x6841e7f7, 0xca7820fb, 0xfb0af54e, 0xd8feb397, 0x454056ac, 0xba489527,
+	0x55533a3a, 0x20838d87, 0xfe6ba9b7, 0xd096954b, 0x55a867bc, 0xa1159a58,
+	0xcca92963, 0x99e1db33, 0xa62a4a56, 0x3f3125f9, 0x5ef47e1c, 0x9029317c,
+	0xfdf8e802, 0x04272f70, 0x80bb155c, 0x05282ce3, 0x95c11548, 0xe4c66d22,
+	0x48c1133f, 0xc70f86dc, 0x07f9c9ee, 0x41041f0f, 0x404779a4, 0x5d886e17,
+	0x325f51eb, 0xd59bc0d1, 0xf2bcc18f, 0x41113564, 0x257b7834, 0x602a9c60,
+	0xdff8e8a3, 0x1f636c1b, 0x0e12b4c2, 0x02e1329e, 0xaf664fd1, 0xcad18115,
+	0x6b2395e0, 0x333e92e1, 0x3b240b62, 0xeebeb922, 0x85b2a20e, 0xe6ba0d99,
+	0xde720c8c, 0x2da2f728, 0xd0127845, 0x95b794fd, 0x647d0862, 0xe7ccf5f0,
+	0x5449a36f, 0x877d48fa, 0xc39dfd27, 0xf33e8d1e, 0x0a476341, 0x992eff74,
+	0x3a6f6eab, 0xf4f8fd37, 0xa812dc60, 0xa1ebddf8, 0x991be14c, 0xdb6e6b0d,
+	0xc67b5510, 0x6d672c37, 0x2765d43b, 0xdcd0e804, 0xf1290dc7, 0xcc00ffa3,
+	0xb5390f92, 0x690fed0b, 0x667b9ffb, 0xcedb7d9c, 0xa091cf0b, 0xd9155ea3,
+	0xbb132f88, 0x515bad24, 0x7b9479bf, 0x763bd6eb, 0x37392eb3, 0xcc115979,
+	0x8026e297, 0xf42e312d, 0x6842ada7, 0xc66a2b3b, 0x12754ccc, 0x782ef11c,
+	0x6a124237, 0xb79251e7, 0x06a1bbe6, 0x4bfb6350, 0x1a6b1018, 0x11caedfa,
+	0x3d25bdd8, 0xe2e1c3c9, 0x44421659, 0x0a121386, 0xd90cec6e, 0xd5abea2a,
+	0x64af674e, 0xda86a85f, 0xbebfe988, 0x64e4c3fe, 0x9dbc8057, 0xf0f7c086,
+	0x60787bf8, 0x6003604d, 0xd1fd8346, 0xf6381fb0, 0x7745ae04, 0xd736fccc,
+	0x83426b33, 0xf01eab71, 0xb0804187, 0x3c005e5f, 0x77a057be, 0xbde8ae24,
+	0x55464299, 0xbf582e61, 0x4e58f48f, 0xf2ddfda2, 0xf474ef38, 0x8789bdc2,
+	0x5366f9c3, 0xc8b38e74, 0xb475f255, 0x46fcd9b9, 0x7aeb2661, 0x8b1ddf84,
+	0x846a0e79, 0x915f95e2, 0x466e598e, 0x20b45770, 0x8cd55591, 0xc902de4c,
+	0xb90bace1, 0xbb8205d0, 0x11a86248, 0x7574a99e, 0xb77f19b6, 0xe0a9dc09,
+	0x662d09a1, 0xc4324633, 0xe85a1f02, 0x09f0be8c, 0x4a99a025, 0x1d6efe10,
+	0x1ab93d1d, 0x0ba5a4df, 0xa186f20f, 0x2868f169, 0xdcb7da83, 0x573906fe,
+	0xa1e2ce9b, 0x4fcd7f52, 0x50115e01, 0xa70683fa, 0xa002b5c4, 0x0de6d027,
+	0x9af88c27, 0x773f8641, 0xc3604c06, 0x61a806b5, 0xf0177a28, 0xc0f586e0,
+	0x006058aa, 0x30dc7d62, 0x11e69ed7, 0x2338ea63, 0x53c2dd94, 0xc2c21634,
+	0xbbcbee56, 0x90bcb6de, 0xebfc7da1, 0xce591d76, 0x6f05e409, 0x4b7c0188,
+	0x39720a3d, 0x7c927c24, 0x86e3725f, 0x724d9db9, 0x1ac15bb4, 0xd39eb8fc,
+	0xed545578, 0x08fca5b5, 0xd83d7cd3, 0x4dad0fc4, 0x1e50ef5e, 0xb161e6f8,
+	0xa28514d9, 0x6c51133c, 0x6fd5c7e7, 0x56e14ec4, 0x362abfce, 0xddc6c837,
+	0xd79a3234, 0x92638212, 0x670efa8e, 0x406000e0,
+}
+
+var s3 = [256]uint32{
+	0x3a39ce37, 0xd3faf5cf, 0xabc27737, 0x5ac52d1b, 0x5cb0679e, 0x4fa33742,
+	0xd3822740, 0x99bc9bbe, 0xd5118e9d, 0xbf0f7315, 0xd62d1c7e, 0xc700c47b,
+	0xb78c1b6b, 0x21a19045, 0xb26eb1be, 0x6a366eb4, 0x5748ab2f, 0xbc946e79,
+	0xc6a376d2, 0x6549c2c8, 0x530ff8ee, 0x468dde7d, 0xd5730a1d, 0x4cd04dc6,
+	0x2939bbdb, 0xa9ba4650, 0xac9526e8, 0xbe5ee304, 0xa1fad5f0, 0x6a2d519a,
+	0x63ef8ce2, 0x9a86ee22, 0xc089c2b8, 0x43242ef6, 0xa51e03aa, 0x9cf2d0a4,
+	0x83c061ba, 0x9be96a4d, 0x8fe51550, 0xba645bd6, 0x2826a2f9, 0xa73a3ae1,
+	0x4ba99586, 0xef5562e9, 0xc72fefd3, 0xf752f7da, 0x3f046f69, 0x77fa0a59,
+	0x80e4a915, 0x87b08601, 0x9b09e6ad, 0x3b3ee593, 0xe990fd5a, 0x9e34d797,
+	0x2cf0b7d9, 0x022b8b51, 0x96d5ac3a, 0x017da67d, 0xd1cf3ed6, 0x7c7d2d28,
+	0x1f9f25cf, 0xadf2b89b, 0x5ad6b472, 0x5a88f54c, 0xe029ac71, 0xe019a5e6,
+	0x47b0acfd, 0xed93fa9b, 0xe8d3c48d, 0x283b57cc, 0xf8d56629, 0x79132e28,
+	0x785f0191, 0xed756055, 0xf7960e44, 0xe3d35e8c, 0x15056dd4, 0x88f46dba,
+	0x03a16125, 0x0564f0bd, 0xc3eb9e15, 0x3c9057a2, 0x97271aec, 0xa93a072a,
+	0x1b3f6d9b, 0x1e6321f5, 0xf59c66fb, 0x26dcf319, 0x7533d928, 0xb155fdf5,
+	0x03563482, 0x8aba3cbb, 0x28517711, 0xc20ad9f8, 0xabcc5167, 0xccad925f,
+	0x4de81751, 0x3830dc8e, 0x379d5862, 0x9320f991, 0xea7a90c2, 0xfb3e7bce,
+	0x5121ce64, 0x774fbe32, 0xa8b6e37e, 0xc3293d46, 0x48de5369, 0x6413e680,
+	0xa2ae0810, 0xdd6db224, 0x69852dfd, 0x09072166, 0xb39a460a, 0x6445c0dd,
+	0x586cdecf, 0x1c20c8ae, 0x5bbef7dd, 0x1b588d40, 0xccd2017f, 0x6bb4e3bb,
+	0xdda26a7e, 0x3a59ff45, 0x3e350a44, 0xbcb4cdd5, 0x72eacea8, 0xfa6484bb,
+	0x8d6612ae, 0xbf3c6f47, 0xd29be463, 0x542f5d9e, 0xaec2771b, 0xf64e6370,
+	0x740e0d8d, 0xe75b1357, 0xf8721671, 0xaf537d5d, 0x4040cb08, 0x4eb4e2cc,
+	0x34d2466a, 0x0115af84, 0xe1b00428, 0x95983a1d, 0x06b89fb4, 0xce6ea048,
+	0x6f3f3b82, 0x3520ab82, 0x011a1d4b, 0x277227f8, 0x611560b1, 0xe7933fdc,
+	0xbb3a792b, 0x344525bd, 0xa08839e1, 0x51ce794b, 0x2f32c9b7, 0xa01fbac9,
+	0xe01cc87e, 0xbcc7d1f6, 0xcf0111c3, 0xa1e8aac7, 0x1a908749, 0xd44fbd9a,
+	0xd0dadecb, 0xd50ada38, 0x0339c32a, 0xc6913667, 0x8df9317c, 0xe0b12b4f,
+	0xf79e59b7, 0x43f5bb3a, 0xf2d519ff, 0x27d9459c, 0xbf97222c, 0x15e6fc2a,
+	0x0f91fc71, 0x9b941525, 0xfae59361, 0xceb69ceb, 0xc2a86459, 0x12baa8d1,
+	0xb6c1075e, 0xe3056a0c, 0x10d25065, 0xcb03a442, 0xe0ec6e0e, 0x1698db3b,
+	0x4c98a0be, 0x3278e964, 0x9f1f9532, 0xe0d392df, 0xd3a0342b, 0x8971f21e,
+	0x1b0a7441, 0x4ba3348c, 0xc5be7120, 0xc37632d8, 0xdf359f8d, 0x9b992f2e,
+	0xe60b6f47, 0x0fe3f11d, 0xe54cda54, 0x1edad891, 0xce6279cf, 0xcd3e7e6f,
+	0x1618b166, 0xfd2c1d05, 0x848fd2c5, 0xf6fb2299, 0xf523f357, 0xa6327623,
+	0x93a83531, 0x56cccd02, 0xacf08162, 0x5a75ebb5, 0x6e163697, 0x88d273cc,
+	0xde966292, 0x81b949d0, 0x4c50901b, 0x71c65614, 0xe6c6c7bd, 0x327a140a,
+	0x45e1d006, 0xc3f27b9a, 0xc9aa53fd, 0x62a80f00, 0xbb25bfe2, 0x35bdd2f6,
+	0x71126905, 0xb2040222, 0xb6cbcf7c, 0xcd769c2b, 0x53113ec0, 0x1640e3d3,
+	0x38abbd60, 0x2547adf0, 0xba38209c, 0xf746ce76, 0x77afa1c5, 0x20756060,
+	0x85cbfe4e, 0x8ae88dd8, 0x7aaaf9b0, 0x4cf9aa7e, 0x1948c25c, 0x02fb8a8c,
+	0x01c36ae4, 0xd6ebe1f9, 0x90d4f869, 0xa65cdea0, 0x3f09252d, 0xc208e69f,
+	0xb74e6132, 0xce77e25b, 0x578fdfe3, 0x3ac372e6,
+}
+
+var p = [18]uint32{
+	0x243f6a88, 0x85a308d3, 0x13198a2e, 0x03707344, 0xa4093822, 0x299f31d0,
+	0x082efa98, 0xec4e6c89, 0x452821e6, 0x38d01377, 0xbe5466cf, 0x34e90c6c,
+	0xc0ac29b7, 0xc97c50dd, 0x3f84d5b5, 0xb5470917, 0x9216d5d9, 0x8979fb1b,
+}

vendor/golang.org/x/crypto/chacha20/chacha_arm64.go

@@ -0,0 +1,16 @@
+// 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.
+
+// +build go1.11,!gccgo,!purego
+
+package chacha20
+
+const bufSize = 256
+
+//go:noescape
+func xorKeyStreamVX(dst, src []byte, key *[8]uint32, nonce *[3]uint32, counter *uint32)
+
+func (c *Cipher) xorKeyStreamBlocks(dst, src []byte) {
+	xorKeyStreamVX(dst, src, &c.key, &c.nonce, &c.counter)
+}

vendor/golang.org/x/crypto/chacha20/chacha_arm64.s

@@ -2,8 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build go1.11
+// +build go1.11,!gccgo,!purego
-// +build !gccgo,!appengine
 
 #include "textflag.h"
 

vendor/golang.org/x/crypto/chacha20/chacha_generic.go

@@ -0,0 +1,398 @@
+// Copyright 2016 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 chacha20 implements the ChaCha20 and XChaCha20 encryption algorithms
+// as specified in RFC 8439 and draft-irtf-cfrg-xchacha-01.
+package chacha20
+
+import (
+	"crypto/cipher"
+	"encoding/binary"
+	"errors"
+	"math/bits"
+
+	"golang.org/x/crypto/internal/subtle"
+)
+
+const (
+	// KeySize is the size of the key used by this cipher, in bytes.
+	KeySize = 32
+
+	// NonceSize is the size of the nonce used with the standard variant of this
+	// cipher, in bytes.
+	//
+	// Note that this is too short to be safely generated at random if the same
+	// key is reused more than 2³² times.
+	NonceSize = 12
+
+	// NonceSizeX is the size of the nonce used with the XChaCha20 variant of
+	// this cipher, in bytes.
+	NonceSizeX = 24
+)
+
+// Cipher is a stateful instance of ChaCha20 or XChaCha20 using a particular key
+// and nonce. A *Cipher implements the cipher.Stream interface.
+type Cipher struct {
+	// The ChaCha20 state is 16 words: 4 constant, 8 of key, 1 of counter
+	// (incremented after each block), and 3 of nonce.
+	key     [8]uint32
+	counter uint32
+	nonce   [3]uint32
+
+	// The last len bytes of buf are leftover key stream bytes from the previous
+	// XORKeyStream invocation. The size of buf depends on how many blocks are
+	// computed at a time by xorKeyStreamBlocks.
+	buf [bufSize]byte
+	len int
+
+	// overflow is set when the counter overflowed, no more blocks can be
+	// generated, and the next XORKeyStream call should panic.
+	overflow bool
+
+	// The counter-independent results of the first round are cached after they
+	// are computed the first time.
+	precompDone      bool
+	p1, p5, p9, p13  uint32
+	p2, p6, p10, p14 uint32
+	p3, p7, p11, p15 uint32
+}
+
+var _ cipher.Stream = (*Cipher)(nil)
+
+// NewUnauthenticatedCipher creates a new ChaCha20 stream cipher with the given
+// 32 bytes key and a 12 or 24 bytes nonce. If a nonce of 24 bytes is provided,
+// the XChaCha20 construction will be used. It returns an error if key or nonce
+// have any other length.
+//
+// Note that ChaCha20, like all stream ciphers, is not authenticated and allows
+// attackers to silently tamper with the plaintext. For this reason, it is more
+// appropriate as a building block than as a standalone encryption mechanism.
+// Instead, consider using package golang.org/x/crypto/chacha20poly1305.
+func NewUnauthenticatedCipher(key, nonce []byte) (*Cipher, error) {
+	// This function is split into a wrapper so that the Cipher allocation will
+	// be inlined, and depending on how the caller uses the return value, won't
+	// escape to the heap.
+	c := &Cipher{}
+	return newUnauthenticatedCipher(c, key, nonce)
+}
+
+func newUnauthenticatedCipher(c *Cipher, key, nonce []byte) (*Cipher, error) {
+	if len(key) != KeySize {
+		return nil, errors.New("chacha20: wrong key size")
+	}
+	if len(nonce) == NonceSizeX {
+		// XChaCha20 uses the ChaCha20 core to mix 16 bytes of the nonce into a
+		// derived key, allowing it to operate on a nonce of 24 bytes. See
+		// draft-irtf-cfrg-xchacha-01, Section 2.3.
+		key, _ = HChaCha20(key, nonce[0:16])
+		cNonce := make([]byte, NonceSize)
+		copy(cNonce[4:12], nonce[16:24])
+		nonce = cNonce
+	} else if len(nonce) != NonceSize {
+		return nil, errors.New("chacha20: wrong nonce size")
+	}
+
+	key, nonce = key[:KeySize], nonce[:NonceSize] // bounds check elimination hint
+	c.key = [8]uint32{
+		binary.LittleEndian.Uint32(key[0:4]),
+		binary.LittleEndian.Uint32(key[4:8]),
+		binary.LittleEndian.Uint32(key[8:12]),
+		binary.LittleEndian.Uint32(key[12:16]),
+		binary.LittleEndian.Uint32(key[16:20]),
+		binary.LittleEndian.Uint32(key[20:24]),
+		binary.LittleEndian.Uint32(key[24:28]),
+		binary.LittleEndian.Uint32(key[28:32]),
+	}
+	c.nonce = [3]uint32{
+		binary.LittleEndian.Uint32(nonce[0:4]),
+		binary.LittleEndian.Uint32(nonce[4:8]),
+		binary.LittleEndian.Uint32(nonce[8:12]),
+	}
+	return c, nil
+}
+
+// The constant first 4 words of the ChaCha20 state.
+const (
+	j0 uint32 = 0x61707865 // expa
+	j1 uint32 = 0x3320646e // nd 3
+	j2 uint32 = 0x79622d32 // 2-by
+	j3 uint32 = 0x6b206574 // te k
+)
+
+const blockSize = 64
+
+// quarterRound is the core of ChaCha20. It shuffles the bits of 4 state words.
+// It's executed 4 times for each of the 20 ChaCha20 rounds, operating on all 16
+// words each round, in columnar or diagonal groups of 4 at a time.
+func quarterRound(a, b, c, d uint32) (uint32, uint32, uint32, uint32) {
+	a += b
+	d ^= a
+	d = bits.RotateLeft32(d, 16)
+	c += d
+	b ^= c
+	b = bits.RotateLeft32(b, 12)
+	a += b
+	d ^= a
+	d = bits.RotateLeft32(d, 8)
+	c += d
+	b ^= c
+	b = bits.RotateLeft32(b, 7)
+	return a, b, c, d
+}
+
+// SetCounter sets the Cipher counter. The next invocation of XORKeyStream will
+// behave as if (64 * counter) bytes had been encrypted so far.
+//
+// To prevent accidental counter reuse, SetCounter panics if counter is less
+// than the current value.
+//
+// Note that the execution time of XORKeyStream is not independent of the
+// counter value.
+func (s *Cipher) SetCounter(counter uint32) {
+	// Internally, s may buffer multiple blocks, which complicates this
+	// implementation slightly. When checking whether the counter has rolled
+	// back, we must use both s.counter and s.len to determine how many blocks
+	// we have already output.
+	outputCounter := s.counter - uint32(s.len)/blockSize
+	if s.overflow || counter < outputCounter {
+		panic("chacha20: SetCounter attempted to rollback counter")
+	}
+
+	// In the general case, we set the new counter value and reset s.len to 0,
+	// causing the next call to XORKeyStream to refill the buffer. However, if
+	// we're advancing within the existing buffer, we can save work by simply
+	// setting s.len.
+	if counter < s.counter {
+		s.len = int(s.counter-counter) * blockSize
+	} else {
+		s.counter = counter
+		s.len = 0
+	}
+}
+
+// XORKeyStream XORs each byte in the given slice with a byte from the
+// cipher's key stream. Dst and src must overlap entirely or not at all.
+//
+// If len(dst) < len(src), XORKeyStream will panic. It is acceptable
+// to pass a dst bigger than src, and in that case, XORKeyStream will
+// only update dst[:len(src)] and will not touch the rest of dst.
+//
+// Multiple calls to XORKeyStream behave as if the concatenation of
+// the src buffers was passed in a single run. That is, Cipher
+// maintains state and does not reset at each XORKeyStream call.
+func (s *Cipher) XORKeyStream(dst, src []byte) {
+	if len(src) == 0 {
+		return
+	}
+	if len(dst) < len(src) {
+		panic("chacha20: output smaller than input")
+	}
+	dst = dst[:len(src)]
+	if subtle.InexactOverlap(dst, src) {
+		panic("chacha20: invalid buffer overlap")
+	}
+
+	// First, drain any remaining key stream from a previous XORKeyStream.
+	if s.len != 0 {
+		keyStream := s.buf[bufSize-s.len:]
+		if len(src) < len(keyStream) {
+			keyStream = keyStream[:len(src)]
+		}
+		_ = src[len(keyStream)-1] // bounds check elimination hint
+		for i, b := range keyStream {
+			dst[i] = src[i] ^ b
+		}
+		s.len -= len(keyStream)
+		dst, src = dst[len(keyStream):], src[len(keyStream):]
+	}
+	if len(src) == 0 {
+		return
+	}
+
+	// If we'd need to let the counter overflow and keep generating output,
+	// panic immediately. If instead we'd only reach the last block, remember
+	// not to generate any more output after the buffer is drained.
+	numBlocks := (uint64(len(src)) + blockSize - 1) / blockSize
+	if s.overflow || uint64(s.counter)+numBlocks > 1<<32 {
+		panic("chacha20: counter overflow")
+	} else if uint64(s.counter)+numBlocks == 1<<32 {
+		s.overflow = true
+	}
+
+	// xorKeyStreamBlocks implementations expect input lengths that are a
+	// multiple of bufSize. Platform-specific ones process multiple blocks at a
+	// time, so have bufSizes that are a multiple of blockSize.
+
+	full := len(src) - len(src)%bufSize
+	if full > 0 {
+		s.xorKeyStreamBlocks(dst[:full], src[:full])
+	}
+	dst, src = dst[full:], src[full:]
+
+	// If using a multi-block xorKeyStreamBlocks would overflow, use the generic
+	// one that does one block at a time.
+	const blocksPerBuf = bufSize / blockSize
+	if uint64(s.counter)+blocksPerBuf > 1<<32 {
+		s.buf = [bufSize]byte{}
+		numBlocks := (len(src) + blockSize - 1) / blockSize
+		buf := s.buf[bufSize-numBlocks*blockSize:]
+		copy(buf, src)
+		s.xorKeyStreamBlocksGeneric(buf, buf)
+		s.len = len(buf) - copy(dst, buf)
+		return
+	}
+
+	// If we have a partial (multi-)block, pad it for xorKeyStreamBlocks, and
+	// keep the leftover keystream for the next XORKeyStream invocation.
+	if len(src) > 0 {
+		s.buf = [bufSize]byte{}
+		copy(s.buf[:], src)
+		s.xorKeyStreamBlocks(s.buf[:], s.buf[:])
+		s.len = bufSize - copy(dst, s.buf[:])
+	}
+}
+
+func (s *Cipher) xorKeyStreamBlocksGeneric(dst, src []byte) {
+	if len(dst) != len(src) || len(dst)%blockSize != 0 {
+		panic("chacha20: internal error: wrong dst and/or src length")
+	}
+
+	// To generate each block of key stream, the initial cipher state
+	// (represented below) is passed through 20 rounds of shuffling,
+	// alternatively applying quarterRounds by columns (like 1, 5, 9, 13)
+	// or by diagonals (like 1, 6, 11, 12).
+	//
+	//      0:cccccccc   1:cccccccc   2:cccccccc   3:cccccccc
+	//      4:kkkkkkkk   5:kkkkkkkk   6:kkkkkkkk   7:kkkkkkkk
+	//      8:kkkkkkkk   9:kkkkkkkk  10:kkkkkkkk  11:kkkkkkkk
+	//     12:bbbbbbbb  13:nnnnnnnn  14:nnnnnnnn  15:nnnnnnnn
+	//
+	//            c=constant k=key b=blockcount n=nonce
+	var (
+		c0, c1, c2, c3   = j0, j1, j2, j3
+		c4, c5, c6, c7   = s.key[0], s.key[1], s.key[2], s.key[3]
+		c8, c9, c10, c11 = s.key[4], s.key[5], s.key[6], s.key[7]
+		_, c13, c14, c15 = s.counter, s.nonce[0], s.nonce[1], s.nonce[2]
+	)
+
+	// Three quarters of the first round don't depend on the counter, so we can
+	// calculate them here, and reuse them for multiple blocks in the loop, and
+	// for future XORKeyStream invocations.
+	if !s.precompDone {
+		s.p1, s.p5, s.p9, s.p13 = quarterRound(c1, c5, c9, c13)
+		s.p2, s.p6, s.p10, s.p14 = quarterRound(c2, c6, c10, c14)
+		s.p3, s.p7, s.p11, s.p15 = quarterRound(c3, c7, c11, c15)
+		s.precompDone = true
+	}
+
+	// A condition of len(src) > 0 would be sufficient, but this also
+	// acts as a bounds check elimination hint.
+	for len(src) >= 64 && len(dst) >= 64 {
+		// The remainder of the first column round.
+		fcr0, fcr4, fcr8, fcr12 := quarterRound(c0, c4, c8, s.counter)
+
+		// The second diagonal round.
+		x0, x5, x10, x15 := quarterRound(fcr0, s.p5, s.p10, s.p15)
+		x1, x6, x11, x12 := quarterRound(s.p1, s.p6, s.p11, fcr12)
+		x2, x7, x8, x13 := quarterRound(s.p2, s.p7, fcr8, s.p13)
+		x3, x4, x9, x14 := quarterRound(s.p3, fcr4, s.p9, s.p14)
+
+		// The remaining 18 rounds.
+		for i := 0; i < 9; i++ {
+			// Column round.
+			x0, x4, x8, x12 = quarterRound(x0, x4, x8, x12)
+			x1, x5, x9, x13 = quarterRound(x1, x5, x9, x13)
+			x2, x6, x10, x14 = quarterRound(x2, x6, x10, x14)
+			x3, x7, x11, x15 = quarterRound(x3, x7, x11, x15)
+
+			// Diagonal round.
+			x0, x5, x10, x15 = quarterRound(x0, x5, x10, x15)
+			x1, x6, x11, x12 = quarterRound(x1, x6, x11, x12)
+			x2, x7, x8, x13 = quarterRound(x2, x7, x8, x13)
+			x3, x4, x9, x14 = quarterRound(x3, x4, x9, x14)
+		}
+
+		// Add back the initial state to generate the key stream, then
+		// XOR the key stream with the source and write out the result.
+		addXor(dst[0:4], src[0:4], x0, c0)
+		addXor(dst[4:8], src[4:8], x1, c1)
+		addXor(dst[8:12], src[8:12], x2, c2)
+		addXor(dst[12:16], src[12:16], x3, c3)
+		addXor(dst[16:20], src[16:20], x4, c4)
+		addXor(dst[20:24], src[20:24], x5, c5)
+		addXor(dst[24:28], src[24:28], x6, c6)
+		addXor(dst[28:32], src[28:32], x7, c7)
+		addXor(dst[32:36], src[32:36], x8, c8)
+		addXor(dst[36:40], src[36:40], x9, c9)
+		addXor(dst[40:44], src[40:44], x10, c10)
+		addXor(dst[44:48], src[44:48], x11, c11)
+		addXor(dst[48:52], src[48:52], x12, s.counter)
+		addXor(dst[52:56], src[52:56], x13, c13)
+		addXor(dst[56:60], src[56:60], x14, c14)
+		addXor(dst[60:64], src[60:64], x15, c15)
+
+		s.counter += 1
+
+		src, dst = src[blockSize:], dst[blockSize:]
+	}
+}
+
+// HChaCha20 uses the ChaCha20 core to generate a derived key from a 32 bytes
+// key and a 16 bytes nonce. It returns an error if key or nonce have any other
+// length. It is used as part of the XChaCha20 construction.
+func HChaCha20(key, nonce []byte) ([]byte, error) {
+	// This function is split into a wrapper so that the slice allocation will
+	// be inlined, and depending on how the caller uses the return value, won't
+	// escape to the heap.
+	out := make([]byte, 32)
+	return hChaCha20(out, key, nonce)
+}
+
+func hChaCha20(out, key, nonce []byte) ([]byte, error) {
+	if len(key) != KeySize {
+		return nil, errors.New("chacha20: wrong HChaCha20 key size")
+	}
+	if len(nonce) != 16 {
+		return nil, errors.New("chacha20: wrong HChaCha20 nonce size")
+	}
+
+	x0, x1, x2, x3 := j0, j1, j2, j3
+	x4 := binary.LittleEndian.Uint32(key[0:4])
+	x5 := binary.LittleEndian.Uint32(key[4:8])
+	x6 := binary.LittleEndian.Uint32(key[8:12])
+	x7 := binary.LittleEndian.Uint32(key[12:16])
+	x8 := binary.LittleEndian.Uint32(key[16:20])
+	x9 := binary.LittleEndian.Uint32(key[20:24])
+	x10 := binary.LittleEndian.Uint32(key[24:28])
+	x11 := binary.LittleEndian.Uint32(key[28:32])
+	x12 := binary.LittleEndian.Uint32(nonce[0:4])
+	x13 := binary.LittleEndian.Uint32(nonce[4:8])
+	x14 := binary.LittleEndian.Uint32(nonce[8:12])
+	x15 := binary.LittleEndian.Uint32(nonce[12:16])
+
+	for i := 0; i < 10; i++ {
+		// Diagonal round.
+		x0, x4, x8, x12 = quarterRound(x0, x4, x8, x12)
+		x1, x5, x9, x13 = quarterRound(x1, x5, x9, x13)
+		x2, x6, x10, x14 = quarterRound(x2, x6, x10, x14)
+		x3, x7, x11, x15 = quarterRound(x3, x7, x11, x15)
+
+		// Column round.
+		x0, x5, x10, x15 = quarterRound(x0, x5, x10, x15)
+		x1, x6, x11, x12 = quarterRound(x1, x6, x11, x12)
+		x2, x7, x8, x13 = quarterRound(x2, x7, x8, x13)
+		x3, x4, x9, x14 = quarterRound(x3, x4, x9, x14)
+	}
+
+	_ = out[31] // bounds check elimination hint
+	binary.LittleEndian.PutUint32(out[0:4], x0)
+	binary.LittleEndian.PutUint32(out[4:8], x1)
+	binary.LittleEndian.PutUint32(out[8:12], x2)
+	binary.LittleEndian.PutUint32(out[12:16], x3)
+	binary.LittleEndian.PutUint32(out[16:20], x12)
+	binary.LittleEndian.PutUint32(out[20:24], x13)
+	binary.LittleEndian.PutUint32(out[24:28], x14)
+	binary.LittleEndian.PutUint32(out[28:32], x15)
+	return out, nil
+}

vendor/golang.org/x/crypto/chacha20/chacha_noasm.go

@@ -2,15 +2,12 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build !arm64,!s390x,!ppc64le arm64,!go1.11 gccgo appengine
+// +build !arm64,!s390x,!ppc64le arm64,!go1.11 gccgo purego
 
 package chacha20
 
-const (
+const bufSize = blockSize
-	bufSize = 64
-	haveAsm = false
-)
 
-func (*Cipher) xorKeyStreamAsm(dst, src []byte) {
+func (s *Cipher) xorKeyStreamBlocks(dst, src []byte) {
-	panic("not implemented")
+	s.xorKeyStreamBlocksGeneric(dst, src)
 }

vendor/golang.org/x/crypto/chacha20/chacha_ppc64le.go

@@ -0,0 +1,16 @@
+// 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.
+
+// +build !gccgo,!purego
+
+package chacha20
+
+const bufSize = 256
+
+//go:noescape
+func chaCha20_ctr32_vsx(out, inp *byte, len int, key *[8]uint32, counter *uint32)
+
+func (c *Cipher) xorKeyStreamBlocks(dst, src []byte) {
+	chaCha20_ctr32_vsx(&dst[0], &src[0], len(src), &c.key, &c.counter)
+}

vendor/golang.org/x/crypto/chacha20/chacha_ppc64le.s

@@ -19,7 +19,7 @@
 // The differences in this and the original implementation are
 // due to the calling conventions and initialization of constants.
 
-// +build ppc64le,!gccgo,!appengine
+// +build !gccgo,!purego
 
 #include "textflag.h"
 
@@ -31,24 +31,7 @@
 #define TMP  R15
 
 #define CONSTBASE  R16
-
+#define BLOCKS R17
-#define X0   R11
-#define X1   R12
-#define X2   R14
-#define X3   R15
-#define X4   R16
-#define X5   R17
-#define X6   R18
-#define X7   R19
-#define X8   R20
-#define X9   R21
-#define X10  R22
-#define X11  R23
-#define X12  R24
-#define X13  R25
-#define X14  R26
-#define X15  R27
-
 
 DATA consts<>+0x00(SB)/8, $0x3320646e61707865
 DATA consts<>+0x08(SB)/8, $0x6b20657479622d32
@@ -72,13 +55,13 @@ DATA consts<>+0x90(SB)/8, $0x0000000100000000
 DATA consts<>+0x98(SB)/8, $0x0000000300000002
 GLOBL consts<>(SB), RODATA, $0xa0
 
-//func chaCha20_ctr32_vsx(out, inp []byte, len int, key *[32]byte, counter *[16]byte)
+//func chaCha20_ctr32_vsx(out, inp *byte, len int, key *[8]uint32, counter *uint32)
 TEXT ·chaCha20_ctr32_vsx(SB),NOSPLIT,$64-40
 	MOVD out+0(FP), OUT
 	MOVD inp+8(FP), INP
 	MOVD len+16(FP), LEN
 	MOVD key+24(FP), KEY
-	MOVD cnt+32(FP), CNT
+	MOVD counter+32(FP), CNT
 
 	// Addressing for constants
 	MOVD $consts<>+0x00(SB), CONSTBASE
@@ -86,6 +69,7 @@ TEXT ·chaCha20_ctr32_vsx(SB),NOSPLIT,$64-40
 	MOVD $32, R9
 	MOVD $48, R10
 	MOVD $64, R11
+	SRD $6, LEN, BLOCKS
 	// V16
 	LXVW4X (CONSTBASE)(R0), VS48
 	ADD $80,CONSTBASE
@@ -429,9 +413,9 @@ loop_vsx:
 	BNE  loop_outer_vsx
 
 done_vsx:
-	// Increment counter by 4
+	// Increment counter by number of 64 byte blocks
 	MOVD (CNT), R14
-	ADD  $4, R14
+	ADD  BLOCKS, R14
 	MOVD R14, (CNT)
 	RET
 

vendor/golang.org/x/crypto/chacha20/chacha_s390x.go

@@ -0,0 +1,26 @@
+// 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.
+
+// +build !gccgo,!purego
+
+package chacha20
+
+import "golang.org/x/sys/cpu"
+
+var haveAsm = cpu.S390X.HasVX
+
+const bufSize = 256
+
+// xorKeyStreamVX is an assembly implementation of XORKeyStream. It must only
+// be called when the vector facility is available. Implementation in asm_s390x.s.
+//go:noescape
+func xorKeyStreamVX(dst, src []byte, key *[8]uint32, nonce *[3]uint32, counter *uint32)
+
+func (c *Cipher) xorKeyStreamBlocks(dst, src []byte) {
+	if cpu.S390X.HasVX {
+		xorKeyStreamVX(dst, src, &c.key, &c.nonce, &c.counter)
+	} else {
+		c.xorKeyStreamBlocksGeneric(dst, src)
+	}
+}

vendor/golang.org/x/crypto/chacha20/chacha_s390x.s

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build s390x,!gccgo,!appengine
+// +build !gccgo,!purego
 
 #include "go_asm.h"
 #include "textflag.h"
@@ -24,15 +24,6 @@ DATA ·constants<>+0x14(SB)/4, $0x3320646e
 DATA ·constants<>+0x18(SB)/4, $0x79622d32
 DATA ·constants<>+0x1c(SB)/4, $0x6b206574
 
-// EXRL targets:
-TEXT ·mvcSrcToBuf(SB), NOFRAME|NOSPLIT, $0
-	MVC $1, (R1), (R8)
-	RET
-
-TEXT ·mvcBufToDst(SB), NOFRAME|NOSPLIT, $0
-	MVC $1, (R8), (R9)
-	RET
-
 #define BSWAP V5
 #define J0    V6
 #define KEY0  V7
@@ -144,7 +135,7 @@ TEXT ·mvcBufToDst(SB), NOFRAME|NOSPLIT, $0
 	VMRHF v, w, c \ // c = {a[2], b[2], c[2], d[2]}
 	VMRLF v, w, d // d = {a[3], b[3], c[3], d[3]}
 
-// func xorKeyStreamVX(dst, src []byte, key *[8]uint32, nonce *[3]uint32, counter *uint32, buf *[256]byte, len *int)
+// func xorKeyStreamVX(dst, src []byte, key *[8]uint32, nonce *[3]uint32, counter *uint32)
 TEXT ·xorKeyStreamVX(SB), NOSPLIT, $0
 	MOVD $·constants<>(SB), R1
 	MOVD dst+0(FP), R2         // R2=&dst[0]
@@ -152,25 +143,10 @@ TEXT ·xorKeyStreamVX(SB), NOSPLIT, $0
 	MOVD key+48(FP), R5        // R5=key
 	MOVD nonce+56(FP), R6      // R6=nonce
 	MOVD counter+64(FP), R7    // R7=counter
-	MOVD buf+72(FP), R8        // R8=buf
-	MOVD len+80(FP), R9        // R9=len
 
 	// load BSWAP and J0
 	VLM (R1), BSWAP, J0
 
-	// set up tail buffer
-	ADD     $-1, R4, R12
-	MOVBZ   R12, R12
-	CMPUBEQ R12, $255, aligned
-	MOVD    R4, R1
-	AND     $~255, R1
-	MOVD    $(R3)(R1*1), R1
-	EXRL    $·mvcSrcToBuf(SB), R12
-	MOVD    $255, R0
-	SUB     R12, R0
-	MOVD    R0, (R9)               // update len
-
-aligned:
 	// setup
 	MOVD  $95, R0
 	VLM   (R5), KEY0, KEY1
@@ -217,9 +193,7 @@ loop:
 
 	// decrement length
 	ADD $-256, R4
-	BLT tail
 
-continue:
 	// rearrange vectors
 	SHUFFLE(X0, X1, X2, X3, M0, M1, M2, M3)
 	ADDV(J0, X0, X1, X2, X3)
@@ -245,16 +219,6 @@ continue:
 	MOVD $256(R3), R3
 
 	CMPBNE  R4, $0, chacha
-	CMPUBEQ R12, $255, return
-	EXRL    $·mvcBufToDst(SB), R12 // len was updated during setup
 
-return:
 	VSTEF $0, CTR, (R7)
 	RET
-
-tail:
-	MOVD R2, R9
-	MOVD R8, R2
-	MOVD R8, R3
-	MOVD $0, R4
-	JMP  continue

vendor/golang.org/x/crypto/chacha20/xor.go

@@ -4,9 +4,7 @@
 
 package chacha20
 
-import (
+import "runtime"
-	"runtime"
-)
 
 // Platforms that have fast unaligned 32-bit little endian accesses.
 const unaligned = runtime.GOARCH == "386" ||
@@ -15,10 +13,10 @@ const unaligned = runtime.GOARCH == "386" ||
 	runtime.GOARCH == "ppc64le" ||
 	runtime.GOARCH == "s390x"
 
-// xor reads a little endian uint32 from src, XORs it with u and
+// addXor reads a little endian uint32 from src, XORs it with (a + b) and
 // places the result in little endian byte order in dst.
-func xor(dst, src []byte, u uint32) {
+func addXor(dst, src []byte, a, b uint32) {
-	_, _ = src[3], dst[3] // eliminate bounds checks
+	_, _ = src[3], dst[3] // bounds check elimination hint
 	if unaligned {
 		// The compiler should optimize this code into
 		// 32-bit unaligned little endian loads and stores.
@@ -29,15 +27,16 @@ func xor(dst, src []byte, u uint32) {
 		v |= uint32(src[1]) << 8
 		v |= uint32(src[2]) << 16
 		v |= uint32(src[3]) << 24
-		v ^= u
+		v ^= a + b
 		dst[0] = byte(v)
 		dst[1] = byte(v >> 8)
 		dst[2] = byte(v >> 16)
 		dst[3] = byte(v >> 24)
 	} else {
-		dst[0] = src[0] ^ byte(u)
+		a += b
-		dst[1] = src[1] ^ byte(u>>8)
+		dst[0] = src[0] ^ byte(a)
-		dst[2] = src[2] ^ byte(u>>16)
+		dst[1] = src[1] ^ byte(a>>8)
-		dst[3] = src[3] ^ byte(u>>24)
+		dst[2] = src[2] ^ byte(a>>16)
+		dst[3] = src[3] ^ byte(a>>24)
 	}
 }

vendor/golang.org/x/crypto/curve25519/const_amd64.h

@@ -1,8 +0,0 @@
-// 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.
-
-// This code was translated into a form compatible with 6a from the public
-// domain sources in SUPERCOP: https://bench.cr.yp.to/supercop.html
-
-#define REDMASK51     0x0007FFFFFFFFFFFF

vendor/golang.org/x/crypto/curve25519/const_amd64.s

@@ -1,20 +0,0 @@
-// 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.
-
-// This code was translated into a form compatible with 6a from the public
-// domain sources in SUPERCOP: https://bench.cr.yp.to/supercop.html
-
-// +build amd64,!gccgo,!appengine
-
-// These constants cannot be encoded in non-MOVQ immediates.
-// We access them directly from memory instead.
-
-DATA ·_121666_213(SB)/8, $996687872
-GLOBL ·_121666_213(SB), 8, $8
-
-DATA ·_2P0(SB)/8, $0xFFFFFFFFFFFDA
-GLOBL ·_2P0(SB), 8, $8
-
-DATA ·_2P1234(SB)/8, $0xFFFFFFFFFFFFE
-GLOBL ·_2P1234(SB), 8, $8

vendor/golang.org/x/crypto/curve25519/cswap_amd64.s

@@ -1,65 +0,0 @@
-// 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.
-
-// +build amd64,!gccgo,!appengine
-
-// func cswap(inout *[4][5]uint64, v uint64)
-TEXT ·cswap(SB),7,$0
-	MOVQ inout+0(FP),DI
-	MOVQ v+8(FP),SI
-
-	SUBQ $1, SI
-	NOTQ SI
-	MOVQ SI, X15
-	PSHUFD $0x44, X15, X15
-
-	MOVOU 0(DI), X0
-	MOVOU 16(DI), X2
-	MOVOU 32(DI), X4
-	MOVOU 48(DI), X6
-	MOVOU 64(DI), X8
-	MOVOU 80(DI), X1
-	MOVOU 96(DI), X3
-	MOVOU 112(DI), X5
-	MOVOU 128(DI), X7
-	MOVOU 144(DI), X9
-
-	MOVO X1, X10
-	MOVO X3, X11
-	MOVO X5, X12
-	MOVO X7, X13
-	MOVO X9, X14
-
-	PXOR X0, X10
-	PXOR X2, X11
-	PXOR X4, X12
-	PXOR X6, X13
-	PXOR X8, X14
-	PAND X15, X10
-	PAND X15, X11
-	PAND X15, X12
-	PAND X15, X13
-	PAND X15, X14
-	PXOR X10, X0
-	PXOR X10, X1
-	PXOR X11, X2
-	PXOR X11, X3
-	PXOR X12, X4
-	PXOR X12, X5
-	PXOR X13, X6
-	PXOR X13, X7
-	PXOR X14, X8
-	PXOR X14, X9
-
-	MOVOU X0, 0(DI)
-	MOVOU X2, 16(DI)
-	MOVOU X4, 32(DI)
-	MOVOU X6, 48(DI)
-	MOVOU X8, 64(DI)
-	MOVOU X1, 80(DI)
-	MOVOU X3, 96(DI)
-	MOVOU X5, 112(DI)
-	MOVOU X7, 128(DI)
-	MOVOU X9, 144(DI)
-	RET

vendor/golang.org/x/crypto/curve25519/curve25519.go

@@ -1,834 +1,95 @@
-// Copyright 2013 The Go Authors. All rights reserved.
+// 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.
 
-// We have an implementation in amd64 assembly so this code is only run on
+// Package curve25519 provides an implementation of the X25519 function, which
-// non-amd64 platforms. The amd64 assembly does not support gccgo.
+// performs scalar multiplication on the elliptic curve known as Curve25519.
-// +build !amd64 gccgo appengine
+// See RFC 7748.
-
+package curve25519 // import "golang.org/x/crypto/curve25519"
-package curve25519
 
 import (
-	"encoding/binary"
+	"crypto/subtle"
+	"fmt"
 )
 
-// This code is a port of the public domain, "ref10" implementation of
+// ScalarMult sets dst to the product scalar * point.
-// curve25519 from SUPERCOP 20130419 by D. J. Bernstein.
+//
+// Deprecated: when provided a low-order point, ScalarMult will set dst to all
+// 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)
+}
 
-// fieldElement represents an element of the field GF(2^255 - 19). An element
+// ScalarBaseMult sets dst to the product scalar * base where base is the
-// t, entries t[0]...t[9], represents the integer t[0]+2^26 t[1]+2^51 t[2]+2^77
+// standard generator.
-// t[3]+2^102 t[4]+...+2^230 t[9]. Bounds on each t[i] vary depending on
+//
-// context.
+// It is recommended to use the X25519 function with Basepoint instead, as
-type fieldElement [10]int32
+// copying into fixed size arrays can lead to unexpected bugs.
+func ScalarBaseMult(dst, scalar *[32]byte) {
+	ScalarMult(dst, scalar, &basePoint)
+}
 
-func feZero(fe *fieldElement) {
+const (
-	for i := range fe {
+	// ScalarSize is the size of the scalar input to X25519.
-		fe[i] = 0
+	ScalarSize = 32
+	// PointSize is the size of the point input to X25519.
+	PointSize = 32
+)
+
+// Basepoint is the canonical Curve25519 generator.
+var Basepoint []byte
+
+var basePoint = [32]byte{9, 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}
+
+func init() { Basepoint = basePoint[:] }
+
+func checkBasepoint() {
+	if subtle.ConstantTimeCompare(Basepoint, []byte{
+		0x09, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+	}) != 1 {
+		panic("curve25519: global Basepoint value was modified")
 	}
 }
 
-func feOne(fe *fieldElement) {
+// X25519 returns the result of the scalar multiplication (scalar * point),
-	feZero(fe)
+// according to RFC 7748, Section 5. scalar, point and the return value are
-	fe[0] = 1
+// slices of 32 bytes.
+//
+// scalar can be generated at random, for example with crypto/rand. point should
+// be either Basepoint or the output of another X25519 call.
+//
+// If point is Basepoint (but not if it's a different slice with the same
+// contents) a precomputed implementation might be used for performance.
+func X25519(scalar, point []byte) ([]byte, error) {
+	// Outline the body of function, to let the allocation be inlined in the
+	// caller, and possibly avoid escaping to the heap.
+	var dst [32]byte
+	return x25519(&dst, scalar, point)
 }
 
-func feAdd(dst, a, b *fieldElement) {
+func x25519(dst *[32]byte, scalar, point []byte) ([]byte, error) {
-	for i := range dst {
+	var in [32]byte
-		dst[i] = a[i] + b[i]
+	if l := len(scalar); l != 32 {
+		return nil, fmt.Errorf("bad scalar length: %d, expected %d", l, 32)
 	}
-}
+	if l := len(point); l != 32 {
-
+		return nil, fmt.Errorf("bad point length: %d, expected %d", l, 32)
-func feSub(dst, a, b *fieldElement) {
+	}
-	for i := range dst {
+	copy(in[:], scalar)
-		dst[i] = a[i] - b[i]
+	if &point[0] == &Basepoint[0] {
-	}
+		checkBasepoint()
-}
+		ScalarBaseMult(dst, &in)
-
+	} else {
-func feCopy(dst, src *fieldElement) {
+		var base, zero [32]byte
-	for i := range dst {
+		copy(base[:], point)
-		dst[i] = src[i]
+		ScalarMult(dst, &in, &base)
-	}
+		if subtle.ConstantTimeCompare(dst[:], zero[:]) == 1 {
-}
+			return nil, fmt.Errorf("bad input point: low order point")
-
+		}
-// feCSwap replaces (f,g) with (g,f) if b == 1; replaces (f,g) with (f,g) if b == 0.
+	}
-//
+	return dst[:], nil
-// Preconditions: b in {0,1}.
-func feCSwap(f, g *fieldElement, b int32) {
-	b = -b
-	for i := range f {
-		t := b & (f[i] ^ g[i])
-		f[i] ^= t
-		g[i] ^= t
-	}
-}
-
-// load3 reads a 24-bit, little-endian value from in.
-func load3(in []byte) int64 {
-	var r int64
-	r = int64(in[0])
-	r |= int64(in[1]) << 8
-	r |= int64(in[2]) << 16
-	return r
-}
-
-// load4 reads a 32-bit, little-endian value from in.
-func load4(in []byte) int64 {
-	return int64(binary.LittleEndian.Uint32(in))
-}
-
-func feFromBytes(dst *fieldElement, src *[32]byte) {
-	h0 := load4(src[:])
-	h1 := load3(src[4:]) << 6
-	h2 := load3(src[7:]) << 5
-	h3 := load3(src[10:]) << 3
-	h4 := load3(src[13:]) << 2
-	h5 := load4(src[16:])
-	h6 := load3(src[20:]) << 7
-	h7 := load3(src[23:]) << 5
-	h8 := load3(src[26:]) << 4
-	h9 := (load3(src[29:]) & 0x7fffff) << 2
-
-	var carry [10]int64
-	carry[9] = (h9 + 1<<24) >> 25
-	h0 += carry[9] * 19
-	h9 -= carry[9] << 25
-	carry[1] = (h1 + 1<<24) >> 25
-	h2 += carry[1]
-	h1 -= carry[1] << 25
-	carry[3] = (h3 + 1<<24) >> 25
-	h4 += carry[3]
-	h3 -= carry[3] << 25
-	carry[5] = (h5 + 1<<24) >> 25
-	h6 += carry[5]
-	h5 -= carry[5] << 25
-	carry[7] = (h7 + 1<<24) >> 25
-	h8 += carry[7]
-	h7 -= carry[7] << 25
-
-	carry[0] = (h0 + 1<<25) >> 26
-	h1 += carry[0]
-	h0 -= carry[0] << 26
-	carry[2] = (h2 + 1<<25) >> 26
-	h3 += carry[2]
-	h2 -= carry[2] << 26
-	carry[4] = (h4 + 1<<25) >> 26
-	h5 += carry[4]
-	h4 -= carry[4] << 26
-	carry[6] = (h6 + 1<<25) >> 26
-	h7 += carry[6]
-	h6 -= carry[6] << 26
-	carry[8] = (h8 + 1<<25) >> 26
-	h9 += carry[8]
-	h8 -= carry[8] << 26
-
-	dst[0] = int32(h0)
-	dst[1] = int32(h1)
-	dst[2] = int32(h2)
-	dst[3] = int32(h3)
-	dst[4] = int32(h4)
-	dst[5] = int32(h5)
-	dst[6] = int32(h6)
-	dst[7] = int32(h7)
-	dst[8] = int32(h8)
-	dst[9] = int32(h9)
-}
-
-// feToBytes marshals h to s.
-// Preconditions:
-//   |h| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
-//
-// Write p=2^255-19; q=floor(h/p).
-// Basic claim: q = floor(2^(-255)(h + 19 2^(-25)h9 + 2^(-1))).
-//
-// Proof:
-//   Have |h|<=p so |q|<=1 so |19^2 2^(-255) q|<1/4.
-//   Also have |h-2^230 h9|<2^230 so |19 2^(-255)(h-2^230 h9)|<1/4.
-//
-//   Write y=2^(-1)-19^2 2^(-255)q-19 2^(-255)(h-2^230 h9).
-//   Then 0<y<1.
-//
-//   Write r=h-pq.
-//   Have 0<=r<=p-1=2^255-20.
-//   Thus 0<=r+19(2^-255)r<r+19(2^-255)2^255<=2^255-1.
-//
-//   Write x=r+19(2^-255)r+y.
-//   Then 0<x<2^255 so floor(2^(-255)x) = 0 so floor(q+2^(-255)x) = q.
-//
-//   Have q+2^(-255)x = 2^(-255)(h + 19 2^(-25) h9 + 2^(-1))
-//   so floor(2^(-255)(h + 19 2^(-25) h9 + 2^(-1))) = q.
-func feToBytes(s *[32]byte, h *fieldElement) {
-	var carry [10]int32
-
-	q := (19*h[9] + (1 << 24)) >> 25
-	q = (h[0] + q) >> 26
-	q = (h[1] + q) >> 25
-	q = (h[2] + q) >> 26
-	q = (h[3] + q) >> 25
-	q = (h[4] + q) >> 26
-	q = (h[5] + q) >> 25
-	q = (h[6] + q) >> 26
-	q = (h[7] + q) >> 25
-	q = (h[8] + q) >> 26
-	q = (h[9] + q) >> 25
-
-	// Goal: Output h-(2^255-19)q, which is between 0 and 2^255-20.
-	h[0] += 19 * q
-	// Goal: Output h-2^255 q, which is between 0 and 2^255-20.
-
-	carry[0] = h[0] >> 26
-	h[1] += carry[0]
-	h[0] -= carry[0] << 26
-	carry[1] = h[1] >> 25
-	h[2] += carry[1]
-	h[1] -= carry[1] << 25
-	carry[2] = h[2] >> 26
-	h[3] += carry[2]
-	h[2] -= carry[2] << 26
-	carry[3] = h[3] >> 25
-	h[4] += carry[3]
-	h[3] -= carry[3] << 25
-	carry[4] = h[4] >> 26
-	h[5] += carry[4]
-	h[4] -= carry[4] << 26
-	carry[5] = h[5] >> 25
-	h[6] += carry[5]
-	h[5] -= carry[5] << 25
-	carry[6] = h[6] >> 26
-	h[7] += carry[6]
-	h[6] -= carry[6] << 26
-	carry[7] = h[7] >> 25
-	h[8] += carry[7]
-	h[7] -= carry[7] << 25
-	carry[8] = h[8] >> 26
-	h[9] += carry[8]
-	h[8] -= carry[8] << 26
-	carry[9] = h[9] >> 25
-	h[9] -= carry[9] << 25
-	// h10 = carry9
-
-	// Goal: Output h[0]+...+2^255 h10-2^255 q, which is between 0 and 2^255-20.
-	// Have h[0]+...+2^230 h[9] between 0 and 2^255-1;
-	// evidently 2^255 h10-2^255 q = 0.
-	// Goal: Output h[0]+...+2^230 h[9].
-
-	s[0] = byte(h[0] >> 0)
-	s[1] = byte(h[0] >> 8)
-	s[2] = byte(h[0] >> 16)
-	s[3] = byte((h[0] >> 24) | (h[1] << 2))
-	s[4] = byte(h[1] >> 6)
-	s[5] = byte(h[1] >> 14)
-	s[6] = byte((h[1] >> 22) | (h[2] << 3))
-	s[7] = byte(h[2] >> 5)
-	s[8] = byte(h[2] >> 13)
-	s[9] = byte((h[2] >> 21) | (h[3] << 5))
-	s[10] = byte(h[3] >> 3)
-	s[11] = byte(h[3] >> 11)
-	s[12] = byte((h[3] >> 19) | (h[4] << 6))
-	s[13] = byte(h[4] >> 2)
-	s[14] = byte(h[4] >> 10)
-	s[15] = byte(h[4] >> 18)
-	s[16] = byte(h[5] >> 0)
-	s[17] = byte(h[5] >> 8)
-	s[18] = byte(h[5] >> 16)
-	s[19] = byte((h[5] >> 24) | (h[6] << 1))
-	s[20] = byte(h[6] >> 7)
-	s[21] = byte(h[6] >> 15)
-	s[22] = byte((h[6] >> 23) | (h[7] << 3))
-	s[23] = byte(h[7] >> 5)
-	s[24] = byte(h[7] >> 13)
-	s[25] = byte((h[7] >> 21) | (h[8] << 4))
-	s[26] = byte(h[8] >> 4)
-	s[27] = byte(h[8] >> 12)
-	s[28] = byte((h[8] >> 20) | (h[9] << 6))
-	s[29] = byte(h[9] >> 2)
-	s[30] = byte(h[9] >> 10)
-	s[31] = byte(h[9] >> 18)
-}
-
-// feMul calculates h = f * g
-// Can overlap h with f or g.
-//
-// Preconditions:
-//    |f| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
-//    |g| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
-//
-// Postconditions:
-//    |h| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
-//
-// Notes on implementation strategy:
-//
-// Using schoolbook multiplication.
-// Karatsuba would save a little in some cost models.
-//
-// Most multiplications by 2 and 19 are 32-bit precomputations;
-// cheaper than 64-bit postcomputations.
-//
-// There is one remaining multiplication by 19 in the carry chain;
-// one *19 precomputation can be merged into this,
-// but the resulting data flow is considerably less clean.
-//
-// There are 12 carries below.
-// 10 of them are 2-way parallelizable and vectorizable.
-// Can get away with 11 carries, but then data flow is much deeper.
-//
-// With tighter constraints on inputs can squeeze carries into int32.
-func feMul(h, f, g *fieldElement) {
-	f0 := f[0]
-	f1 := f[1]
-	f2 := f[2]
-	f3 := f[3]
-	f4 := f[4]
-	f5 := f[5]
-	f6 := f[6]
-	f7 := f[7]
-	f8 := f[8]
-	f9 := f[9]
-	g0 := g[0]
-	g1 := g[1]
-	g2 := g[2]
-	g3 := g[3]
-	g4 := g[4]
-	g5 := g[5]
-	g6 := g[6]
-	g7 := g[7]
-	g8 := g[8]
-	g9 := g[9]
-	g1_19 := 19 * g1 // 1.4*2^29
-	g2_19 := 19 * g2 // 1.4*2^30; still ok
-	g3_19 := 19 * g3
-	g4_19 := 19 * g4
-	g5_19 := 19 * g5
-	g6_19 := 19 * g6
-	g7_19 := 19 * g7
-	g8_19 := 19 * g8
-	g9_19 := 19 * g9
-	f1_2 := 2 * f1
-	f3_2 := 2 * f3
-	f5_2 := 2 * f5
-	f7_2 := 2 * f7
-	f9_2 := 2 * f9
-	f0g0 := int64(f0) * int64(g0)
-	f0g1 := int64(f0) * int64(g1)
-	f0g2 := int64(f0) * int64(g2)
-	f0g3 := int64(f0) * int64(g3)
-	f0g4 := int64(f0) * int64(g4)
-	f0g5 := int64(f0) * int64(g5)
-	f0g6 := int64(f0) * int64(g6)
-	f0g7 := int64(f0) * int64(g7)
-	f0g8 := int64(f0) * int64(g8)
-	f0g9 := int64(f0) * int64(g9)
-	f1g0 := int64(f1) * int64(g0)
-	f1g1_2 := int64(f1_2) * int64(g1)
-	f1g2 := int64(f1) * int64(g2)
-	f1g3_2 := int64(f1_2) * int64(g3)
-	f1g4 := int64(f1) * int64(g4)
-	f1g5_2 := int64(f1_2) * int64(g5)
-	f1g6 := int64(f1) * int64(g6)
-	f1g7_2 := int64(f1_2) * int64(g7)
-	f1g8 := int64(f1) * int64(g8)
-	f1g9_38 := int64(f1_2) * int64(g9_19)
-	f2g0 := int64(f2) * int64(g0)
-	f2g1 := int64(f2) * int64(g1)
-	f2g2 := int64(f2) * int64(g2)
-	f2g3 := int64(f2) * int64(g3)
-	f2g4 := int64(f2) * int64(g4)
-	f2g5 := int64(f2) * int64(g5)
-	f2g6 := int64(f2) * int64(g6)
-	f2g7 := int64(f2) * int64(g7)
-	f2g8_19 := int64(f2) * int64(g8_19)
-	f2g9_19 := int64(f2) * int64(g9_19)
-	f3g0 := int64(f3) * int64(g0)
-	f3g1_2 := int64(f3_2) * int64(g1)
-	f3g2 := int64(f3) * int64(g2)
-	f3g3_2 := int64(f3_2) * int64(g3)
-	f3g4 := int64(f3) * int64(g4)
-	f3g5_2 := int64(f3_2) * int64(g5)
-	f3g6 := int64(f3) * int64(g6)
-	f3g7_38 := int64(f3_2) * int64(g7_19)
-	f3g8_19 := int64(f3) * int64(g8_19)
-	f3g9_38 := int64(f3_2) * int64(g9_19)
-	f4g0 := int64(f4) * int64(g0)
-	f4g1 := int64(f4) * int64(g1)
-	f4g2 := int64(f4) * int64(g2)
-	f4g3 := int64(f4) * int64(g3)
-	f4g4 := int64(f4) * int64(g4)
-	f4g5 := int64(f4) * int64(g5)
-	f4g6_19 := int64(f4) * int64(g6_19)
-	f4g7_19 := int64(f4) * int64(g7_19)
-	f4g8_19 := int64(f4) * int64(g8_19)
-	f4g9_19 := int64(f4) * int64(g9_19)
-	f5g0 := int64(f5) * int64(g0)
-	f5g1_2 := int64(f5_2) * int64(g1)
-	f5g2 := int64(f5) * int64(g2)
-	f5g3_2 := int64(f5_2) * int64(g3)
-	f5g4 := int64(f5) * int64(g4)
-	f5g5_38 := int64(f5_2) * int64(g5_19)
-	f5g6_19 := int64(f5) * int64(g6_19)
-	f5g7_38 := int64(f5_2) * int64(g7_19)
-	f5g8_19 := int64(f5) * int64(g8_19)
-	f5g9_38 := int64(f5_2) * int64(g9_19)
-	f6g0 := int64(f6) * int64(g0)
-	f6g1 := int64(f6) * int64(g1)
-	f6g2 := int64(f6) * int64(g2)
-	f6g3 := int64(f6) * int64(g3)
-	f6g4_19 := int64(f6) * int64(g4_19)
-	f6g5_19 := int64(f6) * int64(g5_19)
-	f6g6_19 := int64(f6) * int64(g6_19)
-	f6g7_19 := int64(f6) * int64(g7_19)
-	f6g8_19 := int64(f6) * int64(g8_19)
-	f6g9_19 := int64(f6) * int64(g9_19)
-	f7g0 := int64(f7) * int64(g0)
-	f7g1_2 := int64(f7_2) * int64(g1)
-	f7g2 := int64(f7) * int64(g2)
-	f7g3_38 := int64(f7_2) * int64(g3_19)
-	f7g4_19 := int64(f7) * int64(g4_19)
-	f7g5_38 := int64(f7_2) * int64(g5_19)
-	f7g6_19 := int64(f7) * int64(g6_19)
-	f7g7_38 := int64(f7_2) * int64(g7_19)
-	f7g8_19 := int64(f7) * int64(g8_19)
-	f7g9_38 := int64(f7_2) * int64(g9_19)
-	f8g0 := int64(f8) * int64(g0)
-	f8g1 := int64(f8) * int64(g1)
-	f8g2_19 := int64(f8) * int64(g2_19)
-	f8g3_19 := int64(f8) * int64(g3_19)
-	f8g4_19 := int64(f8) * int64(g4_19)
-	f8g5_19 := int64(f8) * int64(g5_19)
-	f8g6_19 := int64(f8) * int64(g6_19)
-	f8g7_19 := int64(f8) * int64(g7_19)
-	f8g8_19 := int64(f8) * int64(g8_19)
-	f8g9_19 := int64(f8) * int64(g9_19)
-	f9g0 := int64(f9) * int64(g0)
-	f9g1_38 := int64(f9_2) * int64(g1_19)
-	f9g2_19 := int64(f9) * int64(g2_19)
-	f9g3_38 := int64(f9_2) * int64(g3_19)
-	f9g4_19 := int64(f9) * int64(g4_19)
-	f9g5_38 := int64(f9_2) * int64(g5_19)
-	f9g6_19 := int64(f9) * int64(g6_19)
-	f9g7_38 := int64(f9_2) * int64(g7_19)
-	f9g8_19 := int64(f9) * int64(g8_19)
-	f9g9_38 := int64(f9_2) * int64(g9_19)
-	h0 := f0g0 + f1g9_38 + f2g8_19 + f3g7_38 + f4g6_19 + f5g5_38 + f6g4_19 + f7g3_38 + f8g2_19 + f9g1_38
-	h1 := f0g1 + f1g0 + f2g9_19 + f3g8_19 + f4g7_19 + f5g6_19 + f6g5_19 + f7g4_19 + f8g3_19 + f9g2_19
-	h2 := f0g2 + f1g1_2 + f2g0 + f3g9_38 + f4g8_19 + f5g7_38 + f6g6_19 + f7g5_38 + f8g4_19 + f9g3_38
-	h3 := f0g3 + f1g2 + f2g1 + f3g0 + f4g9_19 + f5g8_19 + f6g7_19 + f7g6_19 + f8g5_19 + f9g4_19
-	h4 := f0g4 + f1g3_2 + f2g2 + f3g1_2 + f4g0 + f5g9_38 + f6g8_19 + f7g7_38 + f8g6_19 + f9g5_38
-	h5 := f0g5 + f1g4 + f2g3 + f3g2 + f4g1 + f5g0 + f6g9_19 + f7g8_19 + f8g7_19 + f9g6_19
-	h6 := f0g6 + f1g5_2 + f2g4 + f3g3_2 + f4g2 + f5g1_2 + f6g0 + f7g9_38 + f8g8_19 + f9g7_38
-	h7 := f0g7 + f1g6 + f2g5 + f3g4 + f4g3 + f5g2 + f6g1 + f7g0 + f8g9_19 + f9g8_19
-	h8 := f0g8 + f1g7_2 + f2g6 + f3g5_2 + f4g4 + f5g3_2 + f6g2 + f7g1_2 + f8g0 + f9g9_38
-	h9 := f0g9 + f1g8 + f2g7 + f3g6 + f4g5 + f5g4 + f6g3 + f7g2 + f8g1 + f9g0
-	var carry [10]int64
-
-	// |h0| <= (1.1*1.1*2^52*(1+19+19+19+19)+1.1*1.1*2^50*(38+38+38+38+38))
-	//   i.e. |h0| <= 1.2*2^59; narrower ranges for h2, h4, h6, h8
-	// |h1| <= (1.1*1.1*2^51*(1+1+19+19+19+19+19+19+19+19))
-	//   i.e. |h1| <= 1.5*2^58; narrower ranges for h3, h5, h7, h9
-
-	carry[0] = (h0 + (1 << 25)) >> 26
-	h1 += carry[0]
-	h0 -= carry[0] << 26
-	carry[4] = (h4 + (1 << 25)) >> 26
-	h5 += carry[4]
-	h4 -= carry[4] << 26
-	// |h0| <= 2^25
-	// |h4| <= 2^25
-	// |h1| <= 1.51*2^58
-	// |h5| <= 1.51*2^58
-
-	carry[1] = (h1 + (1 << 24)) >> 25
-	h2 += carry[1]
-	h1 -= carry[1] << 25
-	carry[5] = (h5 + (1 << 24)) >> 25
-	h6 += carry[5]
-	h5 -= carry[5] << 25
-	// |h1| <= 2^24; from now on fits into int32
-	// |h5| <= 2^24; from now on fits into int32
-	// |h2| <= 1.21*2^59
-	// |h6| <= 1.21*2^59
-
-	carry[2] = (h2 + (1 << 25)) >> 26
-	h3 += carry[2]
-	h2 -= carry[2] << 26
-	carry[6] = (h6 + (1 << 25)) >> 26
-	h7 += carry[6]
-	h6 -= carry[6] << 26
-	// |h2| <= 2^25; from now on fits into int32 unchanged
-	// |h6| <= 2^25; from now on fits into int32 unchanged
-	// |h3| <= 1.51*2^58
-	// |h7| <= 1.51*2^58
-
-	carry[3] = (h3 + (1 << 24)) >> 25
-	h4 += carry[3]
-	h3 -= carry[3] << 25
-	carry[7] = (h7 + (1 << 24)) >> 25
-	h8 += carry[7]
-	h7 -= carry[7] << 25
-	// |h3| <= 2^24; from now on fits into int32 unchanged
-	// |h7| <= 2^24; from now on fits into int32 unchanged
-	// |h4| <= 1.52*2^33
-	// |h8| <= 1.52*2^33
-
-	carry[4] = (h4 + (1 << 25)) >> 26
-	h5 += carry[4]
-	h4 -= carry[4] << 26
-	carry[8] = (h8 + (1 << 25)) >> 26
-	h9 += carry[8]
-	h8 -= carry[8] << 26
-	// |h4| <= 2^25; from now on fits into int32 unchanged
-	// |h8| <= 2^25; from now on fits into int32 unchanged
-	// |h5| <= 1.01*2^24
-	// |h9| <= 1.51*2^58
-
-	carry[9] = (h9 + (1 << 24)) >> 25
-	h0 += carry[9] * 19
-	h9 -= carry[9] << 25
-	// |h9| <= 2^24; from now on fits into int32 unchanged
-	// |h0| <= 1.8*2^37
-
-	carry[0] = (h0 + (1 << 25)) >> 26
-	h1 += carry[0]
-	h0 -= carry[0] << 26
-	// |h0| <= 2^25; from now on fits into int32 unchanged
-	// |h1| <= 1.01*2^24
-
-	h[0] = int32(h0)
-	h[1] = int32(h1)
-	h[2] = int32(h2)
-	h[3] = int32(h3)
-	h[4] = int32(h4)
-	h[5] = int32(h5)
-	h[6] = int32(h6)
-	h[7] = int32(h7)
-	h[8] = int32(h8)
-	h[9] = int32(h9)
-}
-
-// feSquare calculates h = f*f. Can overlap h with f.
-//
-// Preconditions:
-//    |f| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
-//
-// Postconditions:
-//    |h| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
-func feSquare(h, f *fieldElement) {
-	f0 := f[0]
-	f1 := f[1]
-	f2 := f[2]
-	f3 := f[3]
-	f4 := f[4]
-	f5 := f[5]
-	f6 := f[6]
-	f7 := f[7]
-	f8 := f[8]
-	f9 := f[9]
-	f0_2 := 2 * f0
-	f1_2 := 2 * f1
-	f2_2 := 2 * f2
-	f3_2 := 2 * f3
-	f4_2 := 2 * f4
-	f5_2 := 2 * f5
-	f6_2 := 2 * f6
-	f7_2 := 2 * f7
-	f5_38 := 38 * f5 // 1.31*2^30
-	f6_19 := 19 * f6 // 1.31*2^30
-	f7_38 := 38 * f7 // 1.31*2^30
-	f8_19 := 19 * f8 // 1.31*2^30
-	f9_38 := 38 * f9 // 1.31*2^30
-	f0f0 := int64(f0) * int64(f0)
-	f0f1_2 := int64(f0_2) * int64(f1)
-	f0f2_2 := int64(f0_2) * int64(f2)
-	f0f3_2 := int64(f0_2) * int64(f3)
-	f0f4_2 := int64(f0_2) * int64(f4)
-	f0f5_2 := int64(f0_2) * int64(f5)
-	f0f6_2 := int64(f0_2) * int64(f6)
-	f0f7_2 := int64(f0_2) * int64(f7)
-	f0f8_2 := int64(f0_2) * int64(f8)
-	f0f9_2 := int64(f0_2) * int64(f9)
-	f1f1_2 := int64(f1_2) * int64(f1)
-	f1f2_2 := int64(f1_2) * int64(f2)
-	f1f3_4 := int64(f1_2) * int64(f3_2)
-	f1f4_2 := int64(f1_2) * int64(f4)
-	f1f5_4 := int64(f1_2) * int64(f5_2)
-	f1f6_2 := int64(f1_2) * int64(f6)
-	f1f7_4 := int64(f1_2) * int64(f7_2)
-	f1f8_2 := int64(f1_2) * int64(f8)
-	f1f9_76 := int64(f1_2) * int64(f9_38)
-	f2f2 := int64(f2) * int64(f2)
-	f2f3_2 := int64(f2_2) * int64(f3)
-	f2f4_2 := int64(f2_2) * int64(f4)
-	f2f5_2 := int64(f2_2) * int64(f5)
-	f2f6_2 := int64(f2_2) * int64(f6)
-	f2f7_2 := int64(f2_2) * int64(f7)
-	f2f8_38 := int64(f2_2) * int64(f8_19)
-	f2f9_38 := int64(f2) * int64(f9_38)
-	f3f3_2 := int64(f3_2) * int64(f3)
-	f3f4_2 := int64(f3_2) * int64(f4)
-	f3f5_4 := int64(f3_2) * int64(f5_2)
-	f3f6_2 := int64(f3_2) * int64(f6)
-	f3f7_76 := int64(f3_2) * int64(f7_38)
-	f3f8_38 := int64(f3_2) * int64(f8_19)
-	f3f9_76 := int64(f3_2) * int64(f9_38)
-	f4f4 := int64(f4) * int64(f4)
-	f4f5_2 := int64(f4_2) * int64(f5)
-	f4f6_38 := int64(f4_2) * int64(f6_19)
-	f4f7_38 := int64(f4) * int64(f7_38)
-	f4f8_38 := int64(f4_2) * int64(f8_19)
-	f4f9_38 := int64(f4) * int64(f9_38)
-	f5f5_38 := int64(f5) * int64(f5_38)
-	f5f6_38 := int64(f5_2) * int64(f6_19)
-	f5f7_76 := int64(f5_2) * int64(f7_38)
-	f5f8_38 := int64(f5_2) * int64(f8_19)
-	f5f9_76 := int64(f5_2) * int64(f9_38)
-	f6f6_19 := int64(f6) * int64(f6_19)
-	f6f7_38 := int64(f6) * int64(f7_38)
-	f6f8_38 := int64(f6_2) * int64(f8_19)
-	f6f9_38 := int64(f6) * int64(f9_38)
-	f7f7_38 := int64(f7) * int64(f7_38)
-	f7f8_38 := int64(f7_2) * int64(f8_19)
-	f7f9_76 := int64(f7_2) * int64(f9_38)
-	f8f8_19 := int64(f8) * int64(f8_19)
-	f8f9_38 := int64(f8) * int64(f9_38)
-	f9f9_38 := int64(f9) * int64(f9_38)
-	h0 := f0f0 + f1f9_76 + f2f8_38 + f3f7_76 + f4f6_38 + f5f5_38
-	h1 := f0f1_2 + f2f9_38 + f3f8_38 + f4f7_38 + f5f6_38
-	h2 := f0f2_2 + f1f1_2 + f3f9_76 + f4f8_38 + f5f7_76 + f6f6_19
-	h3 := f0f3_2 + f1f2_2 + f4f9_38 + f5f8_38 + f6f7_38
-	h4 := f0f4_2 + f1f3_4 + f2f2 + f5f9_76 + f6f8_38 + f7f7_38
-	h5 := f0f5_2 + f1f4_2 + f2f3_2 + f6f9_38 + f7f8_38
-	h6 := f0f6_2 + f1f5_4 + f2f4_2 + f3f3_2 + f7f9_76 + f8f8_19
-	h7 := f0f7_2 + f1f6_2 + f2f5_2 + f3f4_2 + f8f9_38
-	h8 := f0f8_2 + f1f7_4 + f2f6_2 + f3f5_4 + f4f4 + f9f9_38
-	h9 := f0f9_2 + f1f8_2 + f2f7_2 + f3f6_2 + f4f5_2
-	var carry [10]int64
-
-	carry[0] = (h0 + (1 << 25)) >> 26
-	h1 += carry[0]
-	h0 -= carry[0] << 26
-	carry[4] = (h4 + (1 << 25)) >> 26
-	h5 += carry[4]
-	h4 -= carry[4] << 26
-
-	carry[1] = (h1 + (1 << 24)) >> 25
-	h2 += carry[1]
-	h1 -= carry[1] << 25
-	carry[5] = (h5 + (1 << 24)) >> 25
-	h6 += carry[5]
-	h5 -= carry[5] << 25
-
-	carry[2] = (h2 + (1 << 25)) >> 26
-	h3 += carry[2]
-	h2 -= carry[2] << 26
-	carry[6] = (h6 + (1 << 25)) >> 26
-	h7 += carry[6]
-	h6 -= carry[6] << 26
-
-	carry[3] = (h3 + (1 << 24)) >> 25
-	h4 += carry[3]
-	h3 -= carry[3] << 25
-	carry[7] = (h7 + (1 << 24)) >> 25
-	h8 += carry[7]
-	h7 -= carry[7] << 25
-
-	carry[4] = (h4 + (1 << 25)) >> 26
-	h5 += carry[4]
-	h4 -= carry[4] << 26
-	carry[8] = (h8 + (1 << 25)) >> 26
-	h9 += carry[8]
-	h8 -= carry[8] << 26
-
-	carry[9] = (h9 + (1 << 24)) >> 25
-	h0 += carry[9] * 19
-	h9 -= carry[9] << 25
-
-	carry[0] = (h0 + (1 << 25)) >> 26
-	h1 += carry[0]
-	h0 -= carry[0] << 26
-
-	h[0] = int32(h0)
-	h[1] = int32(h1)
-	h[2] = int32(h2)
-	h[3] = int32(h3)
-	h[4] = int32(h4)
-	h[5] = int32(h5)
-	h[6] = int32(h6)
-	h[7] = int32(h7)
-	h[8] = int32(h8)
-	h[9] = int32(h9)
-}
-
-// feMul121666 calculates h = f * 121666. Can overlap h with f.
-//
-// Preconditions:
-//    |f| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
-//
-// Postconditions:
-//    |h| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
-func feMul121666(h, f *fieldElement) {
-	h0 := int64(f[0]) * 121666
-	h1 := int64(f[1]) * 121666
-	h2 := int64(f[2]) * 121666
-	h3 := int64(f[3]) * 121666
-	h4 := int64(f[4]) * 121666
-	h5 := int64(f[5]) * 121666
-	h6 := int64(f[6]) * 121666
-	h7 := int64(f[7]) * 121666
-	h8 := int64(f[8]) * 121666
-	h9 := int64(f[9]) * 121666
-	var carry [10]int64
-
-	carry[9] = (h9 + (1 << 24)) >> 25
-	h0 += carry[9] * 19
-	h9 -= carry[9] << 25
-	carry[1] = (h1 + (1 << 24)) >> 25
-	h2 += carry[1]
-	h1 -= carry[1] << 25
-	carry[3] = (h3 + (1 << 24)) >> 25
-	h4 += carry[3]
-	h3 -= carry[3] << 25
-	carry[5] = (h5 + (1 << 24)) >> 25
-	h6 += carry[5]
-	h5 -= carry[5] << 25
-	carry[7] = (h7 + (1 << 24)) >> 25
-	h8 += carry[7]
-	h7 -= carry[7] << 25
-
-	carry[0] = (h0 + (1 << 25)) >> 26
-	h1 += carry[0]
-	h0 -= carry[0] << 26
-	carry[2] = (h2 + (1 << 25)) >> 26
-	h3 += carry[2]
-	h2 -= carry[2] << 26
-	carry[4] = (h4 + (1 << 25)) >> 26
-	h5 += carry[4]
-	h4 -= carry[4] << 26
-	carry[6] = (h6 + (1 << 25)) >> 26
-	h7 += carry[6]
-	h6 -= carry[6] << 26
-	carry[8] = (h8 + (1 << 25)) >> 26
-	h9 += carry[8]
-	h8 -= carry[8] << 26
-
-	h[0] = int32(h0)
-	h[1] = int32(h1)
-	h[2] = int32(h2)
-	h[3] = int32(h3)
-	h[4] = int32(h4)
-	h[5] = int32(h5)
-	h[6] = int32(h6)
-	h[7] = int32(h7)
-	h[8] = int32(h8)
-	h[9] = int32(h9)
-}
-
-// feInvert sets out = z^-1.
-func feInvert(out, z *fieldElement) {
-	var t0, t1, t2, t3 fieldElement
-	var i int
-
-	feSquare(&t0, z)
-	for i = 1; i < 1; i++ {
-		feSquare(&t0, &t0)
-	}
-	feSquare(&t1, &t0)
-	for i = 1; i < 2; i++ {
-		feSquare(&t1, &t1)
-	}
-	feMul(&t1, z, &t1)
-	feMul(&t0, &t0, &t1)
-	feSquare(&t2, &t0)
-	for i = 1; i < 1; i++ {
-		feSquare(&t2, &t2)
-	}
-	feMul(&t1, &t1, &t2)
-	feSquare(&t2, &t1)
-	for i = 1; i < 5; i++ {
-		feSquare(&t2, &t2)
-	}
-	feMul(&t1, &t2, &t1)
-	feSquare(&t2, &t1)
-	for i = 1; i < 10; i++ {
-		feSquare(&t2, &t2)
-	}
-	feMul(&t2, &t2, &t1)
-	feSquare(&t3, &t2)
-	for i = 1; i < 20; i++ {
-		feSquare(&t3, &t3)
-	}
-	feMul(&t2, &t3, &t2)
-	feSquare(&t2, &t2)
-	for i = 1; i < 10; i++ {
-		feSquare(&t2, &t2)
-	}
-	feMul(&t1, &t2, &t1)
-	feSquare(&t2, &t1)
-	for i = 1; i < 50; i++ {
-		feSquare(&t2, &t2)
-	}
-	feMul(&t2, &t2, &t1)
-	feSquare(&t3, &t2)
-	for i = 1; i < 100; i++ {
-		feSquare(&t3, &t3)
-	}
-	feMul(&t2, &t3, &t2)
-	feSquare(&t2, &t2)
-	for i = 1; i < 50; i++ {
-		feSquare(&t2, &t2)
-	}
-	feMul(&t1, &t2, &t1)
-	feSquare(&t1, &t1)
-	for i = 1; i < 5; i++ {
-		feSquare(&t1, &t1)
-	}
-	feMul(out, &t1, &t0)
-}
-
-func scalarMult(out, in, base *[32]byte) {
-	var e [32]byte
-
-	copy(e[:], in[:])
-	e[0] &= 248
-	e[31] &= 127
-	e[31] |= 64
-
-	var x1, x2, z2, x3, z3, tmp0, tmp1 fieldElement
-	feFromBytes(&x1, base)
-	feOne(&x2)
-	feCopy(&x3, &x1)
-	feOne(&z3)
-
-	swap := int32(0)
-	for pos := 254; pos >= 0; pos-- {
-		b := e[pos/8] >> uint(pos&7)
-		b &= 1
-		swap ^= int32(b)
-		feCSwap(&x2, &x3, swap)
-		feCSwap(&z2, &z3, swap)
-		swap = int32(b)
-
-		feSub(&tmp0, &x3, &z3)
-		feSub(&tmp1, &x2, &z2)
-		feAdd(&x2, &x2, &z2)
-		feAdd(&z2, &x3, &z3)
-		feMul(&z3, &tmp0, &x2)
-		feMul(&z2, &z2, &tmp1)
-		feSquare(&tmp0, &tmp1)
-		feSquare(&tmp1, &x2)
-		feAdd(&x3, &z3, &z2)
-		feSub(&z2, &z3, &z2)
-		feMul(&x2, &tmp1, &tmp0)
-		feSub(&tmp1, &tmp1, &tmp0)
-		feSquare(&z2, &z2)
-		feMul121666(&z3, &tmp1)
-		feSquare(&x3, &x3)
-		feAdd(&tmp0, &tmp0, &z3)
-		feMul(&z3, &x1, &z2)
-		feMul(&z2, &tmp1, &tmp0)
-	}
-
-	feCSwap(&x2, &x3, swap)
-	feCSwap(&z2, &z3, swap)
-
-	feInvert(&z2, &z2)
-	feMul(&x2, &x2, &z2)
-	feToBytes(out, &x2)
 }

vendor/golang.org/x/crypto/curve25519/curve25519_amd64.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build amd64,!gccgo,!appengine
+// +build amd64,!gccgo,!appengine,!purego
 
 package curve25519
 

vendor/golang.org/x/crypto/curve25519/curve25519_amd64.s

@@ -5,9 +5,84 @@
 // This code was translated into a form compatible with 6a from the public
 // domain sources in SUPERCOP: https://bench.cr.yp.to/supercop.html
 
-// +build amd64,!gccgo,!appengine
+// +build amd64,!gccgo,!appengine,!purego
 
-#include "const_amd64.h"
+#define REDMASK51     0x0007FFFFFFFFFFFF
+
+// These constants cannot be encoded in non-MOVQ immediates.
+// We access them directly from memory instead.
+
+DATA ·_121666_213(SB)/8, $996687872
+GLOBL ·_121666_213(SB), 8, $8
+
+DATA ·_2P0(SB)/8, $0xFFFFFFFFFFFDA
+GLOBL ·_2P0(SB), 8, $8
+
+DATA ·_2P1234(SB)/8, $0xFFFFFFFFFFFFE
+GLOBL ·_2P1234(SB), 8, $8
+
+// func freeze(inout *[5]uint64)
+TEXT ·freeze(SB),7,$0-8
+	MOVQ inout+0(FP), DI
+
+	MOVQ 0(DI),SI
+	MOVQ 8(DI),DX
+	MOVQ 16(DI),CX
+	MOVQ 24(DI),R8
+	MOVQ 32(DI),R9
+	MOVQ $REDMASK51,AX
+	MOVQ AX,R10
+	SUBQ $18,R10
+	MOVQ $3,R11
+REDUCELOOP:
+	MOVQ SI,R12
+	SHRQ $51,R12
+	ANDQ AX,SI
+	ADDQ R12,DX
+	MOVQ DX,R12
+	SHRQ $51,R12
+	ANDQ AX,DX
+	ADDQ R12,CX
+	MOVQ CX,R12
+	SHRQ $51,R12
+	ANDQ AX,CX
+	ADDQ R12,R8
+	MOVQ R8,R12
+	SHRQ $51,R12
+	ANDQ AX,R8
+	ADDQ R12,R9
+	MOVQ R9,R12
+	SHRQ $51,R12
+	ANDQ AX,R9
+	IMUL3Q $19,R12,R12
+	ADDQ R12,SI
+	SUBQ $1,R11
+	JA REDUCELOOP
+	MOVQ $1,R12
+	CMPQ R10,SI
+	CMOVQLT R11,R12
+	CMPQ AX,DX
+	CMOVQNE R11,R12
+	CMPQ AX,CX
+	CMOVQNE R11,R12
+	CMPQ AX,R8
+	CMOVQNE R11,R12
+	CMPQ AX,R9
+	CMOVQNE R11,R12
+	NEGQ R12
+	ANDQ R12,AX
+	ANDQ R12,R10
+	SUBQ R10,SI
+	SUBQ AX,DX
+	SUBQ AX,CX
+	SUBQ AX,R8
+	SUBQ AX,R9
+	MOVQ SI,0(DI)
+	MOVQ DX,8(DI)
+	MOVQ CX,16(DI)
+	MOVQ R8,24(DI)
+	MOVQ R9,32(DI)
+	RET
 
 // func ladderstep(inout *[5][5]uint64)
 TEXT ·ladderstep(SB),0,$296-8
@@ -1375,3 +1450,344 @@ TEXT ·ladderstep(SB),0,$296-8
 	MOVQ AX,104(DI)
 	MOVQ R10,112(DI)
 	RET
+
+// func cswap(inout *[4][5]uint64, v uint64)
+TEXT ·cswap(SB),7,$0
+	MOVQ inout+0(FP),DI
+	MOVQ v+8(FP),SI
+
+	SUBQ $1, SI
+	NOTQ SI
+	MOVQ SI, X15
+	PSHUFD $0x44, X15, X15
+
+	MOVOU 0(DI), X0
+	MOVOU 16(DI), X2
+	MOVOU 32(DI), X4
+	MOVOU 48(DI), X6
+	MOVOU 64(DI), X8
+	MOVOU 80(DI), X1
+	MOVOU 96(DI), X3
+	MOVOU 112(DI), X5
+	MOVOU 128(DI), X7
+	MOVOU 144(DI), X9
+
+	MOVO X1, X10
+	MOVO X3, X11
+	MOVO X5, X12
+	MOVO X7, X13
+	MOVO X9, X14
+
+	PXOR X0, X10
+	PXOR X2, X11
+	PXOR X4, X12
+	PXOR X6, X13
+	PXOR X8, X14
+	PAND X15, X10
+	PAND X15, X11
+	PAND X15, X12
+	PAND X15, X13
+	PAND X15, X14
+	PXOR X10, X0
+	PXOR X10, X1
+	PXOR X11, X2
+	PXOR X11, X3
+	PXOR X12, X4
+	PXOR X12, X5
+	PXOR X13, X6
+	PXOR X13, X7
+	PXOR X14, X8
+	PXOR X14, X9
+
+	MOVOU X0, 0(DI)
+	MOVOU X2, 16(DI)
+	MOVOU X4, 32(DI)
+	MOVOU X6, 48(DI)
+	MOVOU X8, 64(DI)
+	MOVOU X1, 80(DI)
+	MOVOU X3, 96(DI)
+	MOVOU X5, 112(DI)
+	MOVOU X7, 128(DI)
+	MOVOU X9, 144(DI)
+	RET
+
+// func mul(dest, a, b *[5]uint64)
+TEXT ·mul(SB),0,$16-24
+	MOVQ dest+0(FP), DI
+	MOVQ a+8(FP), SI
+	MOVQ b+16(FP), DX
+
+	MOVQ DX,CX
+	MOVQ 24(SI),DX
+	IMUL3Q $19,DX,AX
+	MOVQ AX,0(SP)
+	MULQ 16(CX)
+	MOVQ AX,R8
+	MOVQ DX,R9
+	MOVQ 32(SI),DX
+	IMUL3Q $19,DX,AX
+	MOVQ AX,8(SP)
+	MULQ 8(CX)
+	ADDQ AX,R8
+	ADCQ DX,R9
+	MOVQ 0(SI),AX
+	MULQ 0(CX)
+	ADDQ AX,R8
+	ADCQ DX,R9
+	MOVQ 0(SI),AX
+	MULQ 8(CX)
+	MOVQ AX,R10
+	MOVQ DX,R11
+	MOVQ 0(SI),AX
+	MULQ 16(CX)
+	MOVQ AX,R12
+	MOVQ DX,R13
+	MOVQ 0(SI),AX
+	MULQ 24(CX)
+	MOVQ AX,R14
+	MOVQ DX,R15
+	MOVQ 0(SI),AX
+	MULQ 32(CX)
+	MOVQ AX,BX
+	MOVQ DX,BP
+	MOVQ 8(SI),AX
+	MULQ 0(CX)
+	ADDQ AX,R10
+	ADCQ DX,R11
+	MOVQ 8(SI),AX
+	MULQ 8(CX)
+	ADDQ AX,R12
+	ADCQ DX,R13
+	MOVQ 8(SI),AX
+	MULQ 16(CX)
+	ADDQ AX,R14
+	ADCQ DX,R15
+	MOVQ 8(SI),AX
+	MULQ 24(CX)
+	ADDQ AX,BX
+	ADCQ DX,BP
+	MOVQ 8(SI),DX
+	IMUL3Q $19,DX,AX
+	MULQ 32(CX)
+	ADDQ AX,R8
+	ADCQ DX,R9
+	MOVQ 16(SI),AX
+	MULQ 0(CX)
+	ADDQ AX,R12
+	ADCQ DX,R13
+	MOVQ 16(SI),AX
+	MULQ 8(CX)
+	ADDQ AX,R14
+	ADCQ DX,R15
+	MOVQ 16(SI),AX
+	MULQ 16(CX)
+	ADDQ AX,BX
+	ADCQ DX,BP
+	MOVQ 16(SI),DX
+	IMUL3Q $19,DX,AX
+	MULQ 24(CX)
+	ADDQ AX,R8
+	ADCQ DX,R9
+	MOVQ 16(SI),DX
+	IMUL3Q $19,DX,AX
+	MULQ 32(CX)
+	ADDQ AX,R10
+	ADCQ DX,R11
+	MOVQ 24(SI),AX
+	MULQ 0(CX)
+	ADDQ AX,R14
+	ADCQ DX,R15
+	MOVQ 24(SI),AX
+	MULQ 8(CX)
+	ADDQ AX,BX
+	ADCQ DX,BP
+	MOVQ 0(SP),AX
+	MULQ 24(CX)
+	ADDQ AX,R10
+	ADCQ DX,R11
+	MOVQ 0(SP),AX
+	MULQ 32(CX)
+	ADDQ AX,R12
+	ADCQ DX,R13
+	MOVQ 32(SI),AX
+	MULQ 0(CX)
+	ADDQ AX,BX
+	ADCQ DX,BP
+	MOVQ 8(SP),AX
+	MULQ 16(CX)
+	ADDQ AX,R10
+	ADCQ DX,R11
+	MOVQ 8(SP),AX
+	MULQ 24(CX)
+	ADDQ AX,R12
+	ADCQ DX,R13
+	MOVQ 8(SP),AX
+	MULQ 32(CX)
+	ADDQ AX,R14
+	ADCQ DX,R15
+	MOVQ $REDMASK51,SI
+	SHLQ $13,R8,R9
+	ANDQ SI,R8
+	SHLQ $13,R10,R11
+	ANDQ SI,R10
+	ADDQ R9,R10
+	SHLQ $13,R12,R13
+	ANDQ SI,R12
+	ADDQ R11,R12
+	SHLQ $13,R14,R15
+	ANDQ SI,R14
+	ADDQ R13,R14
+	SHLQ $13,BX,BP
+	ANDQ SI,BX
+	ADDQ R15,BX
+	IMUL3Q $19,BP,DX
+	ADDQ DX,R8
+	MOVQ R8,DX
+	SHRQ $51,DX
+	ADDQ R10,DX
+	MOVQ DX,CX
+	SHRQ $51,DX
+	ANDQ SI,R8
+	ADDQ R12,DX
+	MOVQ DX,R9
+	SHRQ $51,DX
+	ANDQ SI,CX
+	ADDQ R14,DX
+	MOVQ DX,AX
+	SHRQ $51,DX
+	ANDQ SI,R9
+	ADDQ BX,DX
+	MOVQ DX,R10
+	SHRQ $51,DX
+	ANDQ SI,AX
+	IMUL3Q $19,DX,DX
+	ADDQ DX,R8
+	ANDQ SI,R10
+	MOVQ R8,0(DI)
+	MOVQ CX,8(DI)
+	MOVQ R9,16(DI)
+	MOVQ AX,24(DI)
+	MOVQ R10,32(DI)
+	RET
+
+// func square(out, in *[5]uint64)
+TEXT ·square(SB),7,$0-16
+	MOVQ out+0(FP), DI
+	MOVQ in+8(FP), SI
+
+	MOVQ 0(SI),AX
+	MULQ 0(SI)
+	MOVQ AX,CX
+	MOVQ DX,R8
+	MOVQ 0(SI),AX
+	SHLQ $1,AX
+	MULQ 8(SI)
+	MOVQ AX,R9
+	MOVQ DX,R10
+	MOVQ 0(SI),AX
+	SHLQ $1,AX
+	MULQ 16(SI)
+	MOVQ AX,R11
+	MOVQ DX,R12
+	MOVQ 0(SI),AX
+	SHLQ $1,AX
+	MULQ 24(SI)
+	MOVQ AX,R13
+	MOVQ DX,R14
+	MOVQ 0(SI),AX
+	SHLQ $1,AX
+	MULQ 32(SI)
+	MOVQ AX,R15
+	MOVQ DX,BX
+	MOVQ 8(SI),AX
+	MULQ 8(SI)
+	ADDQ AX,R11
+	ADCQ DX,R12
+	MOVQ 8(SI),AX
+	SHLQ $1,AX
+	MULQ 16(SI)
+	ADDQ AX,R13
+	ADCQ DX,R14
+	MOVQ 8(SI),AX
+	SHLQ $1,AX
+	MULQ 24(SI)
+	ADDQ AX,R15
+	ADCQ DX,BX
+	MOVQ 8(SI),DX
+	IMUL3Q $38,DX,AX
+	MULQ 32(SI)
+	ADDQ AX,CX
+	ADCQ DX,R8
+	MOVQ 16(SI),AX
+	MULQ 16(SI)
+	ADDQ AX,R15
+	ADCQ DX,BX
+	MOVQ 16(SI),DX
+	IMUL3Q $38,DX,AX
+	MULQ 24(SI)
+	ADDQ AX,CX
+	ADCQ DX,R8
+	MOVQ 16(SI),DX
+	IMUL3Q $38,DX,AX
+	MULQ 32(SI)
+	ADDQ AX,R9
+	ADCQ DX,R10
+	MOVQ 24(SI),DX
+	IMUL3Q $19,DX,AX
+	MULQ 24(SI)
+	ADDQ AX,R9
+	ADCQ DX,R10
+	MOVQ 24(SI),DX
+	IMUL3Q $38,DX,AX
+	MULQ 32(SI)
+	ADDQ AX,R11
+	ADCQ DX,R12
+	MOVQ 32(SI),DX
+	IMUL3Q $19,DX,AX
+	MULQ 32(SI)
+	ADDQ AX,R13
+	ADCQ DX,R14
+	MOVQ $REDMASK51,SI
+	SHLQ $13,CX,R8
+	ANDQ SI,CX
+	SHLQ $13,R9,R10
+	ANDQ SI,R9
+	ADDQ R8,R9
+	SHLQ $13,R11,R12
+	ANDQ SI,R11
+	ADDQ R10,R11
+	SHLQ $13,R13,R14
+	ANDQ SI,R13
+	ADDQ R12,R13
+	SHLQ $13,R15,BX
+	ANDQ SI,R15
+	ADDQ R14,R15
+	IMUL3Q $19,BX,DX
+	ADDQ DX,CX
+	MOVQ CX,DX
+	SHRQ $51,DX
+	ADDQ R9,DX
+	ANDQ SI,CX
+	MOVQ DX,R8
+	SHRQ $51,DX
+	ADDQ R11,DX
+	ANDQ SI,R8
+	MOVQ DX,R9
+	SHRQ $51,DX
+	ADDQ R13,DX
+	ANDQ SI,R9
+	MOVQ DX,AX
+	SHRQ $51,DX
+	ADDQ R15,DX
+	ANDQ SI,AX
+	MOVQ DX,R10
+	SHRQ $51,DX
+	IMUL3Q $19,DX,DX
+	ADDQ DX,CX
+	ANDQ SI,R10
+	MOVQ CX,0(DI)
+	MOVQ R8,8(DI)
+	MOVQ R9,16(DI)
+	MOVQ AX,24(DI)
+	MOVQ R10,32(DI)
+	RET

vendor/golang.org/x/crypto/curve25519/curve25519_generic.go

@@ -0,0 +1,828 @@
+// Copyright 2013 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 curve25519
+
+import "encoding/binary"
+
+// This code is a port of the public domain, "ref10" implementation of
+// curve25519 from SUPERCOP 20130419 by D. J. Bernstein.
+
+// fieldElement represents an element of the field GF(2^255 - 19). An element
+// t, entries t[0]...t[9], represents the integer t[0]+2^26 t[1]+2^51 t[2]+2^77
+// t[3]+2^102 t[4]+...+2^230 t[9]. Bounds on each t[i] vary depending on
+// context.
+type fieldElement [10]int32
+
+func feZero(fe *fieldElement) {
+	for i := range fe {
+		fe[i] = 0
+	}
+}
+
+func feOne(fe *fieldElement) {
+	feZero(fe)
+	fe[0] = 1
+}
+
+func feAdd(dst, a, b *fieldElement) {
+	for i := range dst {
+		dst[i] = a[i] + b[i]
+	}
+}
+
+func feSub(dst, a, b *fieldElement) {
+	for i := range dst {
+		dst[i] = a[i] - b[i]
+	}
+}
+
+func feCopy(dst, src *fieldElement) {
+	for i := range dst {
+		dst[i] = src[i]
+	}
+}
+
+// feCSwap replaces (f,g) with (g,f) if b == 1; replaces (f,g) with (f,g) if b == 0.
+//
+// Preconditions: b in {0,1}.
+func feCSwap(f, g *fieldElement, b int32) {
+	b = -b
+	for i := range f {
+		t := b & (f[i] ^ g[i])
+		f[i] ^= t
+		g[i] ^= t
+	}
+}
+
+// load3 reads a 24-bit, little-endian value from in.
+func load3(in []byte) int64 {
+	var r int64
+	r = int64(in[0])
+	r |= int64(in[1]) << 8
+	r |= int64(in[2]) << 16
+	return r
+}
+
+// load4 reads a 32-bit, little-endian value from in.
+func load4(in []byte) int64 {
+	return int64(binary.LittleEndian.Uint32(in))
+}
+
+func feFromBytes(dst *fieldElement, src *[32]byte) {
+	h0 := load4(src[:])
+	h1 := load3(src[4:]) << 6
+	h2 := load3(src[7:]) << 5
+	h3 := load3(src[10:]) << 3
+	h4 := load3(src[13:]) << 2
+	h5 := load4(src[16:])
+	h6 := load3(src[20:]) << 7
+	h7 := load3(src[23:]) << 5
+	h8 := load3(src[26:]) << 4
+	h9 := (load3(src[29:]) & 0x7fffff) << 2
+
+	var carry [10]int64
+	carry[9] = (h9 + 1<<24) >> 25
+	h0 += carry[9] * 19
+	h9 -= carry[9] << 25
+	carry[1] = (h1 + 1<<24) >> 25
+	h2 += carry[1]
+	h1 -= carry[1] << 25
+	carry[3] = (h3 + 1<<24) >> 25
+	h4 += carry[3]
+	h3 -= carry[3] << 25
+	carry[5] = (h5 + 1<<24) >> 25
+	h6 += carry[5]
+	h5 -= carry[5] << 25
+	carry[7] = (h7 + 1<<24) >> 25
+	h8 += carry[7]
+	h7 -= carry[7] << 25
+
+	carry[0] = (h0 + 1<<25) >> 26
+	h1 += carry[0]
+	h0 -= carry[0] << 26
+	carry[2] = (h2 + 1<<25) >> 26
+	h3 += carry[2]
+	h2 -= carry[2] << 26
+	carry[4] = (h4 + 1<<25) >> 26
+	h5 += carry[4]
+	h4 -= carry[4] << 26
+	carry[6] = (h6 + 1<<25) >> 26
+	h7 += carry[6]
+	h6 -= carry[6] << 26
+	carry[8] = (h8 + 1<<25) >> 26
+	h9 += carry[8]
+	h8 -= carry[8] << 26
+
+	dst[0] = int32(h0)
+	dst[1] = int32(h1)
+	dst[2] = int32(h2)
+	dst[3] = int32(h3)
+	dst[4] = int32(h4)
+	dst[5] = int32(h5)
+	dst[6] = int32(h6)
+	dst[7] = int32(h7)
+	dst[8] = int32(h8)
+	dst[9] = int32(h9)
+}
+
+// feToBytes marshals h to s.
+// Preconditions:
+//   |h| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
+//
+// Write p=2^255-19; q=floor(h/p).
+// Basic claim: q = floor(2^(-255)(h + 19 2^(-25)h9 + 2^(-1))).
+//
+// Proof:
+//   Have |h|<=p so |q|<=1 so |19^2 2^(-255) q|<1/4.
+//   Also have |h-2^230 h9|<2^230 so |19 2^(-255)(h-2^230 h9)|<1/4.
+//
+//   Write y=2^(-1)-19^2 2^(-255)q-19 2^(-255)(h-2^230 h9).
+//   Then 0<y<1.
+//
+//   Write r=h-pq.
+//   Have 0<=r<=p-1=2^255-20.
+//   Thus 0<=r+19(2^-255)r<r+19(2^-255)2^255<=2^255-1.
+//
+//   Write x=r+19(2^-255)r+y.
+//   Then 0<x<2^255 so floor(2^(-255)x) = 0 so floor(q+2^(-255)x) = q.
+//
+//   Have q+2^(-255)x = 2^(-255)(h + 19 2^(-25) h9 + 2^(-1))
+//   so floor(2^(-255)(h + 19 2^(-25) h9 + 2^(-1))) = q.
+func feToBytes(s *[32]byte, h *fieldElement) {
+	var carry [10]int32
+
+	q := (19*h[9] + (1 << 24)) >> 25
+	q = (h[0] + q) >> 26
+	q = (h[1] + q) >> 25
+	q = (h[2] + q) >> 26
+	q = (h[3] + q) >> 25
+	q = (h[4] + q) >> 26
+	q = (h[5] + q) >> 25
+	q = (h[6] + q) >> 26
+	q = (h[7] + q) >> 25
+	q = (h[8] + q) >> 26
+	q = (h[9] + q) >> 25
+
+	// Goal: Output h-(2^255-19)q, which is between 0 and 2^255-20.
+	h[0] += 19 * q
+	// Goal: Output h-2^255 q, which is between 0 and 2^255-20.
+
+	carry[0] = h[0] >> 26
+	h[1] += carry[0]
+	h[0] -= carry[0] << 26
+	carry[1] = h[1] >> 25
+	h[2] += carry[1]
+	h[1] -= carry[1] << 25
+	carry[2] = h[2] >> 26
+	h[3] += carry[2]
+	h[2] -= carry[2] << 26
+	carry[3] = h[3] >> 25
+	h[4] += carry[3]
+	h[3] -= carry[3] << 25
+	carry[4] = h[4] >> 26
+	h[5] += carry[4]
+	h[4] -= carry[4] << 26
+	carry[5] = h[5] >> 25
+	h[6] += carry[5]
+	h[5] -= carry[5] << 25
+	carry[6] = h[6] >> 26
+	h[7] += carry[6]
+	h[6] -= carry[6] << 26
+	carry[7] = h[7] >> 25
+	h[8] += carry[7]
+	h[7] -= carry[7] << 25
+	carry[8] = h[8] >> 26
+	h[9] += carry[8]
+	h[8] -= carry[8] << 26
+	carry[9] = h[9] >> 25
+	h[9] -= carry[9] << 25
+	// h10 = carry9
+
+	// Goal: Output h[0]+...+2^255 h10-2^255 q, which is between 0 and 2^255-20.
+	// Have h[0]+...+2^230 h[9] between 0 and 2^255-1;
+	// evidently 2^255 h10-2^255 q = 0.
+	// Goal: Output h[0]+...+2^230 h[9].
+
+	s[0] = byte(h[0] >> 0)
+	s[1] = byte(h[0] >> 8)
+	s[2] = byte(h[0] >> 16)
+	s[3] = byte((h[0] >> 24) | (h[1] << 2))
+	s[4] = byte(h[1] >> 6)
+	s[5] = byte(h[1] >> 14)
+	s[6] = byte((h[1] >> 22) | (h[2] << 3))
+	s[7] = byte(h[2] >> 5)
+	s[8] = byte(h[2] >> 13)
+	s[9] = byte((h[2] >> 21) | (h[3] << 5))
+	s[10] = byte(h[3] >> 3)
+	s[11] = byte(h[3] >> 11)
+	s[12] = byte((h[3] >> 19) | (h[4] << 6))
+	s[13] = byte(h[4] >> 2)
+	s[14] = byte(h[4] >> 10)
+	s[15] = byte(h[4] >> 18)
+	s[16] = byte(h[5] >> 0)
+	s[17] = byte(h[5] >> 8)
+	s[18] = byte(h[5] >> 16)
+	s[19] = byte((h[5] >> 24) | (h[6] << 1))
+	s[20] = byte(h[6] >> 7)
+	s[21] = byte(h[6] >> 15)
+	s[22] = byte((h[6] >> 23) | (h[7] << 3))
+	s[23] = byte(h[7] >> 5)
+	s[24] = byte(h[7] >> 13)
+	s[25] = byte((h[7] >> 21) | (h[8] << 4))
+	s[26] = byte(h[8] >> 4)
+	s[27] = byte(h[8] >> 12)
+	s[28] = byte((h[8] >> 20) | (h[9] << 6))
+	s[29] = byte(h[9] >> 2)
+	s[30] = byte(h[9] >> 10)
+	s[31] = byte(h[9] >> 18)
+}
+
+// feMul calculates h = f * g
+// Can overlap h with f or g.
+//
+// Preconditions:
+//    |f| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
+//    |g| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
+//
+// Postconditions:
+//    |h| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
+//
+// Notes on implementation strategy:
+//
+// Using schoolbook multiplication.
+// Karatsuba would save a little in some cost models.
+//
+// Most multiplications by 2 and 19 are 32-bit precomputations;
+// cheaper than 64-bit postcomputations.
+//
+// There is one remaining multiplication by 19 in the carry chain;
+// one *19 precomputation can be merged into this,
+// but the resulting data flow is considerably less clean.
+//
+// There are 12 carries below.
+// 10 of them are 2-way parallelizable and vectorizable.
+// Can get away with 11 carries, but then data flow is much deeper.
+//
+// With tighter constraints on inputs can squeeze carries into int32.
+func feMul(h, f, g *fieldElement) {
+	f0 := f[0]
+	f1 := f[1]
+	f2 := f[2]
+	f3 := f[3]
+	f4 := f[4]
+	f5 := f[5]
+	f6 := f[6]
+	f7 := f[7]
+	f8 := f[8]
+	f9 := f[9]
+	g0 := g[0]
+	g1 := g[1]
+	g2 := g[2]
+	g3 := g[3]
+	g4 := g[4]
+	g5 := g[5]
+	g6 := g[6]
+	g7 := g[7]
+	g8 := g[8]
+	g9 := g[9]
+	g1_19 := 19 * g1 // 1.4*2^29
+	g2_19 := 19 * g2 // 1.4*2^30; still ok
+	g3_19 := 19 * g3
+	g4_19 := 19 * g4
+	g5_19 := 19 * g5
+	g6_19 := 19 * g6
+	g7_19 := 19 * g7
+	g8_19 := 19 * g8
+	g9_19 := 19 * g9
+	f1_2 := 2 * f1
+	f3_2 := 2 * f3
+	f5_2 := 2 * f5
+	f7_2 := 2 * f7
+	f9_2 := 2 * f9
+	f0g0 := int64(f0) * int64(g0)
+	f0g1 := int64(f0) * int64(g1)
+	f0g2 := int64(f0) * int64(g2)
+	f0g3 := int64(f0) * int64(g3)
+	f0g4 := int64(f0) * int64(g4)
+	f0g5 := int64(f0) * int64(g5)
+	f0g6 := int64(f0) * int64(g6)
+	f0g7 := int64(f0) * int64(g7)
+	f0g8 := int64(f0) * int64(g8)
+	f0g9 := int64(f0) * int64(g9)
+	f1g0 := int64(f1) * int64(g0)
+	f1g1_2 := int64(f1_2) * int64(g1)
+	f1g2 := int64(f1) * int64(g2)
+	f1g3_2 := int64(f1_2) * int64(g3)
+	f1g4 := int64(f1) * int64(g4)
+	f1g5_2 := int64(f1_2) * int64(g5)
+	f1g6 := int64(f1) * int64(g6)
+	f1g7_2 := int64(f1_2) * int64(g7)
+	f1g8 := int64(f1) * int64(g8)
+	f1g9_38 := int64(f1_2) * int64(g9_19)
+	f2g0 := int64(f2) * int64(g0)
+	f2g1 := int64(f2) * int64(g1)
+	f2g2 := int64(f2) * int64(g2)
+	f2g3 := int64(f2) * int64(g3)
+	f2g4 := int64(f2) * int64(g4)
+	f2g5 := int64(f2) * int64(g5)
+	f2g6 := int64(f2) * int64(g6)
+	f2g7 := int64(f2) * int64(g7)
+	f2g8_19 := int64(f2) * int64(g8_19)
+	f2g9_19 := int64(f2) * int64(g9_19)
+	f3g0 := int64(f3) * int64(g0)
+	f3g1_2 := int64(f3_2) * int64(g1)
+	f3g2 := int64(f3) * int64(g2)
+	f3g3_2 := int64(f3_2) * int64(g3)
+	f3g4 := int64(f3) * int64(g4)
+	f3g5_2 := int64(f3_2) * int64(g5)
+	f3g6 := int64(f3) * int64(g6)
+	f3g7_38 := int64(f3_2) * int64(g7_19)
+	f3g8_19 := int64(f3) * int64(g8_19)
+	f3g9_38 := int64(f3_2) * int64(g9_19)
+	f4g0 := int64(f4) * int64(g0)
+	f4g1 := int64(f4) * int64(g1)
+	f4g2 := int64(f4) * int64(g2)
+	f4g3 := int64(f4) * int64(g3)
+	f4g4 := int64(f4) * int64(g4)
+	f4g5 := int64(f4) * int64(g5)
+	f4g6_19 := int64(f4) * int64(g6_19)
+	f4g7_19 := int64(f4) * int64(g7_19)
+	f4g8_19 := int64(f4) * int64(g8_19)
+	f4g9_19 := int64(f4) * int64(g9_19)
+	f5g0 := int64(f5) * int64(g0)
+	f5g1_2 := int64(f5_2) * int64(g1)
+	f5g2 := int64(f5) * int64(g2)
+	f5g3_2 := int64(f5_2) * int64(g3)
+	f5g4 := int64(f5) * int64(g4)
+	f5g5_38 := int64(f5_2) * int64(g5_19)
+	f5g6_19 := int64(f5) * int64(g6_19)
+	f5g7_38 := int64(f5_2) * int64(g7_19)
+	f5g8_19 := int64(f5) * int64(g8_19)
+	f5g9_38 := int64(f5_2) * int64(g9_19)
+	f6g0 := int64(f6) * int64(g0)
+	f6g1 := int64(f6) * int64(g1)
+	f6g2 := int64(f6) * int64(g2)
+	f6g3 := int64(f6) * int64(g3)
+	f6g4_19 := int64(f6) * int64(g4_19)
+	f6g5_19 := int64(f6) * int64(g5_19)
+	f6g6_19 := int64(f6) * int64(g6_19)
+	f6g7_19 := int64(f6) * int64(g7_19)
+	f6g8_19 := int64(f6) * int64(g8_19)
+	f6g9_19 := int64(f6) * int64(g9_19)
+	f7g0 := int64(f7) * int64(g0)
+	f7g1_2 := int64(f7_2) * int64(g1)
+	f7g2 := int64(f7) * int64(g2)
+	f7g3_38 := int64(f7_2) * int64(g3_19)
+	f7g4_19 := int64(f7) * int64(g4_19)
+	f7g5_38 := int64(f7_2) * int64(g5_19)
+	f7g6_19 := int64(f7) * int64(g6_19)
+	f7g7_38 := int64(f7_2) * int64(g7_19)
+	f7g8_19 := int64(f7) * int64(g8_19)
+	f7g9_38 := int64(f7_2) * int64(g9_19)
+	f8g0 := int64(f8) * int64(g0)
+	f8g1 := int64(f8) * int64(g1)
+	f8g2_19 := int64(f8) * int64(g2_19)
+	f8g3_19 := int64(f8) * int64(g3_19)
+	f8g4_19 := int64(f8) * int64(g4_19)
+	f8g5_19 := int64(f8) * int64(g5_19)
+	f8g6_19 := int64(f8) * int64(g6_19)
+	f8g7_19 := int64(f8) * int64(g7_19)
+	f8g8_19 := int64(f8) * int64(g8_19)
+	f8g9_19 := int64(f8) * int64(g9_19)
+	f9g0 := int64(f9) * int64(g0)
+	f9g1_38 := int64(f9_2) * int64(g1_19)
+	f9g2_19 := int64(f9) * int64(g2_19)
+	f9g3_38 := int64(f9_2) * int64(g3_19)
+	f9g4_19 := int64(f9) * int64(g4_19)
+	f9g5_38 := int64(f9_2) * int64(g5_19)
+	f9g6_19 := int64(f9) * int64(g6_19)
+	f9g7_38 := int64(f9_2) * int64(g7_19)
+	f9g8_19 := int64(f9) * int64(g8_19)
+	f9g9_38 := int64(f9_2) * int64(g9_19)
+	h0 := f0g0 + f1g9_38 + f2g8_19 + f3g7_38 + f4g6_19 + f5g5_38 + f6g4_19 + f7g3_38 + f8g2_19 + f9g1_38
+	h1 := f0g1 + f1g0 + f2g9_19 + f3g8_19 + f4g7_19 + f5g6_19 + f6g5_19 + f7g4_19 + f8g3_19 + f9g2_19
+	h2 := f0g2 + f1g1_2 + f2g0 + f3g9_38 + f4g8_19 + f5g7_38 + f6g6_19 + f7g5_38 + f8g4_19 + f9g3_38
+	h3 := f0g3 + f1g2 + f2g1 + f3g0 + f4g9_19 + f5g8_19 + f6g7_19 + f7g6_19 + f8g5_19 + f9g4_19
+	h4 := f0g4 + f1g3_2 + f2g2 + f3g1_2 + f4g0 + f5g9_38 + f6g8_19 + f7g7_38 + f8g6_19 + f9g5_38
+	h5 := f0g5 + f1g4 + f2g3 + f3g2 + f4g1 + f5g0 + f6g9_19 + f7g8_19 + f8g7_19 + f9g6_19
+	h6 := f0g6 + f1g5_2 + f2g4 + f3g3_2 + f4g2 + f5g1_2 + f6g0 + f7g9_38 + f8g8_19 + f9g7_38
+	h7 := f0g7 + f1g6 + f2g5 + f3g4 + f4g3 + f5g2 + f6g1 + f7g0 + f8g9_19 + f9g8_19
+	h8 := f0g8 + f1g7_2 + f2g6 + f3g5_2 + f4g4 + f5g3_2 + f6g2 + f7g1_2 + f8g0 + f9g9_38
+	h9 := f0g9 + f1g8 + f2g7 + f3g6 + f4g5 + f5g4 + f6g3 + f7g2 + f8g1 + f9g0
+	var carry [10]int64
+
+	// |h0| <= (1.1*1.1*2^52*(1+19+19+19+19)+1.1*1.1*2^50*(38+38+38+38+38))
+	//   i.e. |h0| <= 1.2*2^59; narrower ranges for h2, h4, h6, h8
+	// |h1| <= (1.1*1.1*2^51*(1+1+19+19+19+19+19+19+19+19))
+	//   i.e. |h1| <= 1.5*2^58; narrower ranges for h3, h5, h7, h9
+
+	carry[0] = (h0 + (1 << 25)) >> 26
+	h1 += carry[0]
+	h0 -= carry[0] << 26
+	carry[4] = (h4 + (1 << 25)) >> 26
+	h5 += carry[4]
+	h4 -= carry[4] << 26
+	// |h0| <= 2^25
+	// |h4| <= 2^25
+	// |h1| <= 1.51*2^58
+	// |h5| <= 1.51*2^58
+
+	carry[1] = (h1 + (1 << 24)) >> 25
+	h2 += carry[1]
+	h1 -= carry[1] << 25
+	carry[5] = (h5 + (1 << 24)) >> 25
+	h6 += carry[5]
+	h5 -= carry[5] << 25
+	// |h1| <= 2^24; from now on fits into int32
+	// |h5| <= 2^24; from now on fits into int32
+	// |h2| <= 1.21*2^59
+	// |h6| <= 1.21*2^59
+
+	carry[2] = (h2 + (1 << 25)) >> 26
+	h3 += carry[2]
+	h2 -= carry[2] << 26
+	carry[6] = (h6 + (1 << 25)) >> 26
+	h7 += carry[6]
+	h6 -= carry[6] << 26
+	// |h2| <= 2^25; from now on fits into int32 unchanged
+	// |h6| <= 2^25; from now on fits into int32 unchanged
+	// |h3| <= 1.51*2^58
+	// |h7| <= 1.51*2^58
+
+	carry[3] = (h3 + (1 << 24)) >> 25
+	h4 += carry[3]
+	h3 -= carry[3] << 25
+	carry[7] = (h7 + (1 << 24)) >> 25
+	h8 += carry[7]
+	h7 -= carry[7] << 25
+	// |h3| <= 2^24; from now on fits into int32 unchanged
+	// |h7| <= 2^24; from now on fits into int32 unchanged
+	// |h4| <= 1.52*2^33
+	// |h8| <= 1.52*2^33
+
+	carry[4] = (h4 + (1 << 25)) >> 26
+	h5 += carry[4]
+	h4 -= carry[4] << 26
+	carry[8] = (h8 + (1 << 25)) >> 26
+	h9 += carry[8]
+	h8 -= carry[8] << 26
+	// |h4| <= 2^25; from now on fits into int32 unchanged
+	// |h8| <= 2^25; from now on fits into int32 unchanged
+	// |h5| <= 1.01*2^24
+	// |h9| <= 1.51*2^58
+
+	carry[9] = (h9 + (1 << 24)) >> 25
+	h0 += carry[9] * 19
+	h9 -= carry[9] << 25
+	// |h9| <= 2^24; from now on fits into int32 unchanged
+	// |h0| <= 1.8*2^37
+
+	carry[0] = (h0 + (1 << 25)) >> 26
+	h1 += carry[0]
+	h0 -= carry[0] << 26
+	// |h0| <= 2^25; from now on fits into int32 unchanged
+	// |h1| <= 1.01*2^24
+
+	h[0] = int32(h0)
+	h[1] = int32(h1)
+	h[2] = int32(h2)
+	h[3] = int32(h3)
+	h[4] = int32(h4)
+	h[5] = int32(h5)
+	h[6] = int32(h6)
+	h[7] = int32(h7)
+	h[8] = int32(h8)
+	h[9] = int32(h9)
+}
+
+// feSquare calculates h = f*f. Can overlap h with f.
+//
+// Preconditions:
+//    |f| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
+//
+// Postconditions:
+//    |h| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
+func feSquare(h, f *fieldElement) {
+	f0 := f[0]
+	f1 := f[1]
+	f2 := f[2]
+	f3 := f[3]
+	f4 := f[4]
+	f5 := f[5]
+	f6 := f[6]
+	f7 := f[7]
+	f8 := f[8]
+	f9 := f[9]
+	f0_2 := 2 * f0
+	f1_2 := 2 * f1
+	f2_2 := 2 * f2
+	f3_2 := 2 * f3
+	f4_2 := 2 * f4
+	f5_2 := 2 * f5
+	f6_2 := 2 * f6
+	f7_2 := 2 * f7
+	f5_38 := 38 * f5 // 1.31*2^30
+	f6_19 := 19 * f6 // 1.31*2^30
+	f7_38 := 38 * f7 // 1.31*2^30
+	f8_19 := 19 * f8 // 1.31*2^30
+	f9_38 := 38 * f9 // 1.31*2^30
+	f0f0 := int64(f0) * int64(f0)
+	f0f1_2 := int64(f0_2) * int64(f1)
+	f0f2_2 := int64(f0_2) * int64(f2)
+	f0f3_2 := int64(f0_2) * int64(f3)
+	f0f4_2 := int64(f0_2) * int64(f4)
+	f0f5_2 := int64(f0_2) * int64(f5)
+	f0f6_2 := int64(f0_2) * int64(f6)
+	f0f7_2 := int64(f0_2) * int64(f7)
+	f0f8_2 := int64(f0_2) * int64(f8)
+	f0f9_2 := int64(f0_2) * int64(f9)
+	f1f1_2 := int64(f1_2) * int64(f1)
+	f1f2_2 := int64(f1_2) * int64(f2)
+	f1f3_4 := int64(f1_2) * int64(f3_2)
+	f1f4_2 := int64(f1_2) * int64(f4)
+	f1f5_4 := int64(f1_2) * int64(f5_2)
+	f1f6_2 := int64(f1_2) * int64(f6)
+	f1f7_4 := int64(f1_2) * int64(f7_2)
+	f1f8_2 := int64(f1_2) * int64(f8)
+	f1f9_76 := int64(f1_2) * int64(f9_38)
+	f2f2 := int64(f2) * int64(f2)
+	f2f3_2 := int64(f2_2) * int64(f3)
+	f2f4_2 := int64(f2_2) * int64(f4)
+	f2f5_2 := int64(f2_2) * int64(f5)
+	f2f6_2 := int64(f2_2) * int64(f6)
+	f2f7_2 := int64(f2_2) * int64(f7)
+	f2f8_38 := int64(f2_2) * int64(f8_19)
+	f2f9_38 := int64(f2) * int64(f9_38)
+	f3f3_2 := int64(f3_2) * int64(f3)
+	f3f4_2 := int64(f3_2) * int64(f4)
+	f3f5_4 := int64(f3_2) * int64(f5_2)
+	f3f6_2 := int64(f3_2) * int64(f6)
+	f3f7_76 := int64(f3_2) * int64(f7_38)
+	f3f8_38 := int64(f3_2) * int64(f8_19)
+	f3f9_76 := int64(f3_2) * int64(f9_38)
+	f4f4 := int64(f4) * int64(f4)
+	f4f5_2 := int64(f4_2) * int64(f5)
+	f4f6_38 := int64(f4_2) * int64(f6_19)
+	f4f7_38 := int64(f4) * int64(f7_38)
+	f4f8_38 := int64(f4_2) * int64(f8_19)
+	f4f9_38 := int64(f4) * int64(f9_38)
+	f5f5_38 := int64(f5) * int64(f5_38)
+	f5f6_38 := int64(f5_2) * int64(f6_19)
+	f5f7_76 := int64(f5_2) * int64(f7_38)
+	f5f8_38 := int64(f5_2) * int64(f8_19)
+	f5f9_76 := int64(f5_2) * int64(f9_38)
+	f6f6_19 := int64(f6) * int64(f6_19)
+	f6f7_38 := int64(f6) * int64(f7_38)
+	f6f8_38 := int64(f6_2) * int64(f8_19)
+	f6f9_38 := int64(f6) * int64(f9_38)
+	f7f7_38 := int64(f7) * int64(f7_38)
+	f7f8_38 := int64(f7_2) * int64(f8_19)
+	f7f9_76 := int64(f7_2) * int64(f9_38)
+	f8f8_19 := int64(f8) * int64(f8_19)
+	f8f9_38 := int64(f8) * int64(f9_38)
+	f9f9_38 := int64(f9) * int64(f9_38)
+	h0 := f0f0 + f1f9_76 + f2f8_38 + f3f7_76 + f4f6_38 + f5f5_38
+	h1 := f0f1_2 + f2f9_38 + f3f8_38 + f4f7_38 + f5f6_38
+	h2 := f0f2_2 + f1f1_2 + f3f9_76 + f4f8_38 + f5f7_76 + f6f6_19
+	h3 := f0f3_2 + f1f2_2 + f4f9_38 + f5f8_38 + f6f7_38
+	h4 := f0f4_2 + f1f3_4 + f2f2 + f5f9_76 + f6f8_38 + f7f7_38
+	h5 := f0f5_2 + f1f4_2 + f2f3_2 + f6f9_38 + f7f8_38
+	h6 := f0f6_2 + f1f5_4 + f2f4_2 + f3f3_2 + f7f9_76 + f8f8_19
+	h7 := f0f7_2 + f1f6_2 + f2f5_2 + f3f4_2 + f8f9_38
+	h8 := f0f8_2 + f1f7_4 + f2f6_2 + f3f5_4 + f4f4 + f9f9_38
+	h9 := f0f9_2 + f1f8_2 + f2f7_2 + f3f6_2 + f4f5_2
+	var carry [10]int64
+
+	carry[0] = (h0 + (1 << 25)) >> 26
+	h1 += carry[0]
+	h0 -= carry[0] << 26
+	carry[4] = (h4 + (1 << 25)) >> 26
+	h5 += carry[4]
+	h4 -= carry[4] << 26
+
+	carry[1] = (h1 + (1 << 24)) >> 25
+	h2 += carry[1]
+	h1 -= carry[1] << 25
+	carry[5] = (h5 + (1 << 24)) >> 25
+	h6 += carry[5]
+	h5 -= carry[5] << 25
+
+	carry[2] = (h2 + (1 << 25)) >> 26
+	h3 += carry[2]
+	h2 -= carry[2] << 26
+	carry[6] = (h6 + (1 << 25)) >> 26
+	h7 += carry[6]
+	h6 -= carry[6] << 26
+
+	carry[3] = (h3 + (1 << 24)) >> 25
+	h4 += carry[3]
+	h3 -= carry[3] << 25
+	carry[7] = (h7 + (1 << 24)) >> 25
+	h8 += carry[7]
+	h7 -= carry[7] << 25
+
+	carry[4] = (h4 + (1 << 25)) >> 26
+	h5 += carry[4]
+	h4 -= carry[4] << 26
+	carry[8] = (h8 + (1 << 25)) >> 26
+	h9 += carry[8]
+	h8 -= carry[8] << 26
+
+	carry[9] = (h9 + (1 << 24)) >> 25
+	h0 += carry[9] * 19
+	h9 -= carry[9] << 25
+
+	carry[0] = (h0 + (1 << 25)) >> 26
+	h1 += carry[0]
+	h0 -= carry[0] << 26
+
+	h[0] = int32(h0)
+	h[1] = int32(h1)
+	h[2] = int32(h2)
+	h[3] = int32(h3)
+	h[4] = int32(h4)
+	h[5] = int32(h5)
+	h[6] = int32(h6)
+	h[7] = int32(h7)
+	h[8] = int32(h8)
+	h[9] = int32(h9)
+}
+
+// feMul121666 calculates h = f * 121666. Can overlap h with f.
+//
+// Preconditions:
+//    |f| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
+//
+// Postconditions:
+//    |h| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
+func feMul121666(h, f *fieldElement) {
+	h0 := int64(f[0]) * 121666
+	h1 := int64(f[1]) * 121666
+	h2 := int64(f[2]) * 121666
+	h3 := int64(f[3]) * 121666
+	h4 := int64(f[4]) * 121666
+	h5 := int64(f[5]) * 121666
+	h6 := int64(f[6]) * 121666
+	h7 := int64(f[7]) * 121666
+	h8 := int64(f[8]) * 121666
+	h9 := int64(f[9]) * 121666
+	var carry [10]int64
+
+	carry[9] = (h9 + (1 << 24)) >> 25
+	h0 += carry[9] * 19
+	h9 -= carry[9] << 25
+	carry[1] = (h1 + (1 << 24)) >> 25
+	h2 += carry[1]
+	h1 -= carry[1] << 25
+	carry[3] = (h3 + (1 << 24)) >> 25
+	h4 += carry[3]
+	h3 -= carry[3] << 25
+	carry[5] = (h5 + (1 << 24)) >> 25
+	h6 += carry[5]
+	h5 -= carry[5] << 25
+	carry[7] = (h7 + (1 << 24)) >> 25
+	h8 += carry[7]
+	h7 -= carry[7] << 25
+
+	carry[0] = (h0 + (1 << 25)) >> 26
+	h1 += carry[0]
+	h0 -= carry[0] << 26
+	carry[2] = (h2 + (1 << 25)) >> 26
+	h3 += carry[2]
+	h2 -= carry[2] << 26
+	carry[4] = (h4 + (1 << 25)) >> 26
+	h5 += carry[4]
+	h4 -= carry[4] << 26
+	carry[6] = (h6 + (1 << 25)) >> 26
+	h7 += carry[6]
+	h6 -= carry[6] << 26
+	carry[8] = (h8 + (1 << 25)) >> 26
+	h9 += carry[8]
+	h8 -= carry[8] << 26
+
+	h[0] = int32(h0)
+	h[1] = int32(h1)
+	h[2] = int32(h2)
+	h[3] = int32(h3)
+	h[4] = int32(h4)
+	h[5] = int32(h5)
+	h[6] = int32(h6)
+	h[7] = int32(h7)
+	h[8] = int32(h8)
+	h[9] = int32(h9)
+}
+
+// feInvert sets out = z^-1.
+func feInvert(out, z *fieldElement) {
+	var t0, t1, t2, t3 fieldElement
+	var i int
+
+	feSquare(&t0, z)
+	for i = 1; i < 1; i++ {
+		feSquare(&t0, &t0)
+	}
+	feSquare(&t1, &t0)
+	for i = 1; i < 2; i++ {
+		feSquare(&t1, &t1)
+	}
+	feMul(&t1, z, &t1)
+	feMul(&t0, &t0, &t1)
+	feSquare(&t2, &t0)
+	for i = 1; i < 1; i++ {
+		feSquare(&t2, &t2)
+	}
+	feMul(&t1, &t1, &t2)
+	feSquare(&t2, &t1)
+	for i = 1; i < 5; i++ {
+		feSquare(&t2, &t2)
+	}
+	feMul(&t1, &t2, &t1)
+	feSquare(&t2, &t1)
+	for i = 1; i < 10; i++ {
+		feSquare(&t2, &t2)
+	}
+	feMul(&t2, &t2, &t1)
+	feSquare(&t3, &t2)
+	for i = 1; i < 20; i++ {
+		feSquare(&t3, &t3)
+	}
+	feMul(&t2, &t3, &t2)
+	feSquare(&t2, &t2)
+	for i = 1; i < 10; i++ {
+		feSquare(&t2, &t2)
+	}
+	feMul(&t1, &t2, &t1)
+	feSquare(&t2, &t1)
+	for i = 1; i < 50; i++ {
+		feSquare(&t2, &t2)
+	}
+	feMul(&t2, &t2, &t1)
+	feSquare(&t3, &t2)
+	for i = 1; i < 100; i++ {
+		feSquare(&t3, &t3)
+	}
+	feMul(&t2, &t3, &t2)
+	feSquare(&t2, &t2)
+	for i = 1; i < 50; i++ {
+		feSquare(&t2, &t2)
+	}
+	feMul(&t1, &t2, &t1)
+	feSquare(&t1, &t1)
+	for i = 1; i < 5; i++ {
+		feSquare(&t1, &t1)
+	}
+	feMul(out, &t1, &t0)
+}
+
+func scalarMultGeneric(out, in, base *[32]byte) {
+	var e [32]byte
+
+	copy(e[:], in[:])
+	e[0] &= 248
+	e[31] &= 127
+	e[31] |= 64
+
+	var x1, x2, z2, x3, z3, tmp0, tmp1 fieldElement
+	feFromBytes(&x1, base)
+	feOne(&x2)
+	feCopy(&x3, &x1)
+	feOne(&z3)
+
+	swap := int32(0)
+	for pos := 254; pos >= 0; pos-- {
+		b := e[pos/8] >> uint(pos&7)
+		b &= 1
+		swap ^= int32(b)
+		feCSwap(&x2, &x3, swap)
+		feCSwap(&z2, &z3, swap)
+		swap = int32(b)
+
+		feSub(&tmp0, &x3, &z3)
+		feSub(&tmp1, &x2, &z2)
+		feAdd(&x2, &x2, &z2)
+		feAdd(&z2, &x3, &z3)
+		feMul(&z3, &tmp0, &x2)
+		feMul(&z2, &z2, &tmp1)
+		feSquare(&tmp0, &tmp1)
+		feSquare(&tmp1, &x2)
+		feAdd(&x3, &z3, &z2)
+		feSub(&z2, &z3, &z2)
+		feMul(&x2, &tmp1, &tmp0)
+		feSub(&tmp1, &tmp1, &tmp0)
+		feSquare(&z2, &z2)
+		feMul121666(&z3, &tmp1)
+		feSquare(&x3, &x3)
+		feAdd(&tmp0, &tmp0, &z3)
+		feMul(&z3, &x1, &z2)
+		feMul(&z2, &tmp1, &tmp0)
+	}
+
+	feCSwap(&x2, &x3, swap)
+	feCSwap(&z2, &z3, swap)
+
+	feInvert(&z2, &z2)
+	feMul(&x2, &x2, &z2)
+	feToBytes(out, &x2)
+}

vendor/golang.org/x/crypto/curve25519/curve25519_noasm.go

@@ -0,0 +1,11 @@
+// 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.
+
+// +build !amd64 gccgo appengine purego
+
+package curve25519
+
+func scalarMult(out, in, base *[32]byte) {
+	scalarMultGeneric(out, in, base)
+}

vendor/golang.org/x/crypto/curve25519/doc.go

@@ -1,23 +0,0 @@
-// 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 curve25519 provides an implementation of scalar multiplication on
-// the elliptic curve known as curve25519. See https://cr.yp.to/ecdh.html
-package curve25519 // import "golang.org/x/crypto/curve25519"
-
-// basePoint is the x coordinate of the generator of the curve.
-var basePoint = [32]byte{9, 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}
-
-// ScalarMult sets dst to the product in*base where dst and base are the x
-// coordinates of group points and all values are in little-endian form.
-func ScalarMult(dst, in, base *[32]byte) {
-	scalarMult(dst, in, base)
-}
-
-// ScalarBaseMult sets dst to the product in*base where dst and base are the x
-// coordinates of group points, base is the standard generator and all values
-// are in little-endian form.
-func ScalarBaseMult(dst, in *[32]byte) {
-	ScalarMult(dst, in, &basePoint)
-}

vendor/golang.org/x/crypto/curve25519/freeze_amd64.s

@@ -1,73 +0,0 @@
-// 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.
-
-// This code was translated into a form compatible with 6a from the public
-// domain sources in SUPERCOP: https://bench.cr.yp.to/supercop.html
-
-// +build amd64,!gccgo,!appengine
-
-#include "const_amd64.h"
-
-// func freeze(inout *[5]uint64)
-TEXT ·freeze(SB),7,$0-8
-	MOVQ inout+0(FP), DI
-
-	MOVQ 0(DI),SI
-	MOVQ 8(DI),DX
-	MOVQ 16(DI),CX
-	MOVQ 24(DI),R8
-	MOVQ 32(DI),R9
-	MOVQ $REDMASK51,AX
-	MOVQ AX,R10
-	SUBQ $18,R10
-	MOVQ $3,R11
-REDUCELOOP:
-	MOVQ SI,R12
-	SHRQ $51,R12
-	ANDQ AX,SI
-	ADDQ R12,DX
-	MOVQ DX,R12
-	SHRQ $51,R12
-	ANDQ AX,DX
-	ADDQ R12,CX
-	MOVQ CX,R12
-	SHRQ $51,R12
-	ANDQ AX,CX
-	ADDQ R12,R8
-	MOVQ R8,R12
-	SHRQ $51,R12
-	ANDQ AX,R8
-	ADDQ R12,R9
-	MOVQ R9,R12
-	SHRQ $51,R12
-	ANDQ AX,R9
-	IMUL3Q $19,R12,R12
-	ADDQ R12,SI
-	SUBQ $1,R11
-	JA REDUCELOOP
-	MOVQ $1,R12
-	CMPQ R10,SI
-	CMOVQLT R11,R12
-	CMPQ AX,DX
-	CMOVQNE R11,R12
-	CMPQ AX,CX
-	CMOVQNE R11,R12
-	CMPQ AX,R8
-	CMOVQNE R11,R12
-	CMPQ AX,R9
-	CMOVQNE R11,R12
-	NEGQ R12
-	ANDQ R12,AX
-	ANDQ R12,R10
-	SUBQ R10,SI
-	SUBQ AX,DX
-	SUBQ AX,CX
-	SUBQ AX,R8
-	SUBQ AX,R9
-	MOVQ SI,0(DI)
-	MOVQ DX,8(DI)
-	MOVQ CX,16(DI)
-	MOVQ R8,24(DI)
-	MOVQ R9,32(DI)
-	RET

vendor/golang.org/x/crypto/curve25519/mul_amd64.s

@@ -1,169 +0,0 @@
-// 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.
-
-// This code was translated into a form compatible with 6a from the public
-// domain sources in SUPERCOP: https://bench.cr.yp.to/supercop.html
-
-// +build amd64,!gccgo,!appengine
-
-#include "const_amd64.h"
-
-// func mul(dest, a, b *[5]uint64)
-TEXT ·mul(SB),0,$16-24
-	MOVQ dest+0(FP), DI
-	MOVQ a+8(FP), SI
-	MOVQ b+16(FP), DX
-
-	MOVQ DX,CX
-	MOVQ 24(SI),DX
-	IMUL3Q $19,DX,AX
-	MOVQ AX,0(SP)
-	MULQ 16(CX)
-	MOVQ AX,R8
-	MOVQ DX,R9
-	MOVQ 32(SI),DX
-	IMUL3Q $19,DX,AX
-	MOVQ AX,8(SP)
-	MULQ 8(CX)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 0(SI),AX
-	MULQ 0(CX)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 0(SI),AX
-	MULQ 8(CX)
-	MOVQ AX,R10
-	MOVQ DX,R11
-	MOVQ 0(SI),AX
-	MULQ 16(CX)
-	MOVQ AX,R12
-	MOVQ DX,R13
-	MOVQ 0(SI),AX
-	MULQ 24(CX)
-	MOVQ AX,R14
-	MOVQ DX,R15
-	MOVQ 0(SI),AX
-	MULQ 32(CX)
-	MOVQ AX,BX
-	MOVQ DX,BP
-	MOVQ 8(SI),AX
-	MULQ 0(CX)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 8(SI),AX
-	MULQ 8(CX)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 8(SI),AX
-	MULQ 16(CX)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 8(SI),AX
-	MULQ 24(CX)
-	ADDQ AX,BX
-	ADCQ DX,BP
-	MOVQ 8(SI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 32(CX)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 16(SI),AX
-	MULQ 0(CX)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 16(SI),AX
-	MULQ 8(CX)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 16(SI),AX
-	MULQ 16(CX)
-	ADDQ AX,BX
-	ADCQ DX,BP
-	MOVQ 16(SI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 24(CX)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 16(SI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 32(CX)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 24(SI),AX
-	MULQ 0(CX)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 24(SI),AX
-	MULQ 8(CX)
-	ADDQ AX,BX
-	ADCQ DX,BP
-	MOVQ 0(SP),AX
-	MULQ 24(CX)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 0(SP),AX
-	MULQ 32(CX)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 32(SI),AX
-	MULQ 0(CX)
-	ADDQ AX,BX
-	ADCQ DX,BP
-	MOVQ 8(SP),AX
-	MULQ 16(CX)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 8(SP),AX
-	MULQ 24(CX)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 8(SP),AX
-	MULQ 32(CX)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ $REDMASK51,SI
-	SHLQ $13,R8,R9
-	ANDQ SI,R8
-	SHLQ $13,R10,R11
-	ANDQ SI,R10
-	ADDQ R9,R10
-	SHLQ $13,R12,R13
-	ANDQ SI,R12
-	ADDQ R11,R12
-	SHLQ $13,R14,R15
-	ANDQ SI,R14
-	ADDQ R13,R14
-	SHLQ $13,BX,BP
-	ANDQ SI,BX
-	ADDQ R15,BX
-	IMUL3Q $19,BP,DX
-	ADDQ DX,R8
-	MOVQ R8,DX
-	SHRQ $51,DX
-	ADDQ R10,DX
-	MOVQ DX,CX
-	SHRQ $51,DX
-	ANDQ SI,R8
-	ADDQ R12,DX
-	MOVQ DX,R9
-	SHRQ $51,DX
-	ANDQ SI,CX
-	ADDQ R14,DX
-	MOVQ DX,AX
-	SHRQ $51,DX
-	ANDQ SI,R9
-	ADDQ BX,DX
-	MOVQ DX,R10
-	SHRQ $51,DX
-	ANDQ SI,AX
-	IMUL3Q $19,DX,DX
-	ADDQ DX,R8
-	ANDQ SI,R10
-	MOVQ R8,0(DI)
-	MOVQ CX,8(DI)
-	MOVQ R9,16(DI)
-	MOVQ AX,24(DI)
-	MOVQ R10,32(DI)
-	RET

vendor/golang.org/x/crypto/curve25519/square_amd64.s

@@ -1,132 +0,0 @@
-// 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.
-
-// This code was translated into a form compatible with 6a from the public
-// domain sources in SUPERCOP: https://bench.cr.yp.to/supercop.html
-
-// +build amd64,!gccgo,!appengine
-
-#include "const_amd64.h"
-
-// func square(out, in *[5]uint64)
-TEXT ·square(SB),7,$0-16
-	MOVQ out+0(FP), DI
-	MOVQ in+8(FP), SI
-
-	MOVQ 0(SI),AX
-	MULQ 0(SI)
-	MOVQ AX,CX
-	MOVQ DX,R8
-	MOVQ 0(SI),AX
-	SHLQ $1,AX
-	MULQ 8(SI)
-	MOVQ AX,R9
-	MOVQ DX,R10
-	MOVQ 0(SI),AX
-	SHLQ $1,AX
-	MULQ 16(SI)
-	MOVQ AX,R11
-	MOVQ DX,R12
-	MOVQ 0(SI),AX
-	SHLQ $1,AX
-	MULQ 24(SI)
-	MOVQ AX,R13
-	MOVQ DX,R14
-	MOVQ 0(SI),AX
-	SHLQ $1,AX
-	MULQ 32(SI)
-	MOVQ AX,R15
-	MOVQ DX,BX
-	MOVQ 8(SI),AX
-	MULQ 8(SI)
-	ADDQ AX,R11
-	ADCQ DX,R12
-	MOVQ 8(SI),AX
-	SHLQ $1,AX
-	MULQ 16(SI)
-	ADDQ AX,R13
-	ADCQ DX,R14
-	MOVQ 8(SI),AX
-	SHLQ $1,AX
-	MULQ 24(SI)
-	ADDQ AX,R15
-	ADCQ DX,BX
-	MOVQ 8(SI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 32(SI)
-	ADDQ AX,CX
-	ADCQ DX,R8
-	MOVQ 16(SI),AX
-	MULQ 16(SI)
-	ADDQ AX,R15
-	ADCQ DX,BX
-	MOVQ 16(SI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 24(SI)
-	ADDQ AX,CX
-	ADCQ DX,R8
-	MOVQ 16(SI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 32(SI)
-	ADDQ AX,R9
-	ADCQ DX,R10
-	MOVQ 24(SI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 24(SI)
-	ADDQ AX,R9
-	ADCQ DX,R10
-	MOVQ 24(SI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 32(SI)
-	ADDQ AX,R11
-	ADCQ DX,R12
-	MOVQ 32(SI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 32(SI)
-	ADDQ AX,R13
-	ADCQ DX,R14
-	MOVQ $REDMASK51,SI
-	SHLQ $13,CX,R8
-	ANDQ SI,CX
-	SHLQ $13,R9,R10
-	ANDQ SI,R9
-	ADDQ R8,R9
-	SHLQ $13,R11,R12
-	ANDQ SI,R11
-	ADDQ R10,R11
-	SHLQ $13,R13,R14
-	ANDQ SI,R13
-	ADDQ R12,R13
-	SHLQ $13,R15,BX
-	ANDQ SI,R15
-	ADDQ R14,R15
-	IMUL3Q $19,BX,DX
-	ADDQ DX,CX
-	MOVQ CX,DX
-	SHRQ $51,DX
-	ADDQ R9,DX
-	ANDQ SI,CX
-	MOVQ DX,R8
-	SHRQ $51,DX
-	ADDQ R11,DX
-	ANDQ SI,R8
-	MOVQ DX,R9
-	SHRQ $51,DX
-	ADDQ R13,DX
-	ANDQ SI,R9
-	MOVQ DX,AX
-	SHRQ $51,DX
-	ADDQ R15,DX
-	ANDQ SI,AX
-	MOVQ DX,R10
-	SHRQ $51,DX
-	IMUL3Q $19,DX,DX
-	ADDQ DX,CX
-	ANDQ SI,R10
-	MOVQ CX,0(DI)
-	MOVQ R8,8(DI)
-	MOVQ R9,16(DI)
-	MOVQ AX,24(DI)
-	MOVQ R10,32(DI)
-	RET

vendor/golang.org/x/crypto/internal/chacha20/chacha_arm64.go

@@ -1,31 +0,0 @@
-// 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.
-
-// +build go1.11
-// +build !gccgo
-
-package chacha20
-
-const (
-	haveAsm = true
-	bufSize = 256
-)
-
-//go:noescape
-func xorKeyStreamVX(dst, src []byte, key *[8]uint32, nonce *[3]uint32, counter *uint32)
-
-func (c *Cipher) xorKeyStreamAsm(dst, src []byte) {
-
-	if len(src) >= bufSize {
-		xorKeyStreamVX(dst, src, &c.key, &c.nonce, &c.counter)
-	}
-
-	if len(src)%bufSize != 0 {
-		i := len(src) - len(src)%bufSize
-		c.buf = [bufSize]byte{}
-		copy(c.buf[:], src[i:])
-		xorKeyStreamVX(c.buf[:], c.buf[:], &c.key, &c.nonce, &c.counter)
-		c.len = bufSize - copy(dst[i:], c.buf[:len(src)%bufSize])
-	}
-}

vendor/golang.org/x/crypto/internal/chacha20/chacha_generic.go

@@ -1,264 +0,0 @@
-// Copyright 2016 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 ChaCha20 implements the core ChaCha20 function as specified
-// in https://tools.ietf.org/html/rfc7539#section-2.3.
-package chacha20
-
-import (
-	"crypto/cipher"
-	"encoding/binary"
-
-	"golang.org/x/crypto/internal/subtle"
-)
-
-// assert that *Cipher implements cipher.Stream
-var _ cipher.Stream = (*Cipher)(nil)
-
-// Cipher is a stateful instance of ChaCha20 using a particular key
-// and nonce. A *Cipher implements the cipher.Stream interface.
-type Cipher struct {
-	key     [8]uint32
-	counter uint32 // incremented after each block
-	nonce   [3]uint32
-	buf     [bufSize]byte // buffer for unused keystream bytes
-	len     int           // number of unused keystream bytes at end of buf
-}
-
-// New creates a new ChaCha20 stream cipher with the given key and nonce.
-// The initial counter value is set to 0.
-func New(key [8]uint32, nonce [3]uint32) *Cipher {
-	return &Cipher{key: key, nonce: nonce}
-}
-
-// ChaCha20 constants spelling "expand 32-byte k"
-const (
-	j0 uint32 = 0x61707865
-	j1 uint32 = 0x3320646e
-	j2 uint32 = 0x79622d32
-	j3 uint32 = 0x6b206574
-)
-
-func quarterRound(a, b, c, d uint32) (uint32, uint32, uint32, uint32) {
-	a += b
-	d ^= a
-	d = (d << 16) | (d >> 16)
-	c += d
-	b ^= c
-	b = (b << 12) | (b >> 20)
-	a += b
-	d ^= a
-	d = (d << 8) | (d >> 24)
-	c += d
-	b ^= c
-	b = (b << 7) | (b >> 25)
-	return a, b, c, d
-}
-
-// XORKeyStream XORs each byte in the given slice with a byte from the
-// cipher's key stream. Dst and src must overlap entirely or not at all.
-//
-// If len(dst) < len(src), XORKeyStream will panic. It is acceptable
-// to pass a dst bigger than src, and in that case, XORKeyStream will
-// only update dst[:len(src)] and will not touch the rest of dst.
-//
-// Multiple calls to XORKeyStream behave as if the concatenation of
-// the src buffers was passed in a single run. That is, Cipher
-// maintains state and does not reset at each XORKeyStream call.
-func (s *Cipher) XORKeyStream(dst, src []byte) {
-	if len(dst) < len(src) {
-		panic("chacha20: output smaller than input")
-	}
-	if subtle.InexactOverlap(dst[:len(src)], src) {
-		panic("chacha20: invalid buffer overlap")
-	}
-
-	// xor src with buffered keystream first
-	if s.len != 0 {
-		buf := s.buf[len(s.buf)-s.len:]
-		if len(src) < len(buf) {
-			buf = buf[:len(src)]
-		}
-		td, ts := dst[:len(buf)], src[:len(buf)] // BCE hint
-		for i, b := range buf {
-			td[i] = ts[i] ^ b
-		}
-		s.len -= len(buf)
-		if s.len != 0 {
-			return
-		}
-		s.buf = [len(s.buf)]byte{} // zero the empty buffer
-		src = src[len(buf):]
-		dst = dst[len(buf):]
-	}
-
-	if len(src) == 0 {
-		return
-	}
-	if haveAsm {
-		if uint64(len(src))+uint64(s.counter)*64 > (1<<38)-64 {
-			panic("chacha20: counter overflow")
-		}
-		s.xorKeyStreamAsm(dst, src)
-		return
-	}
-
-	// set up a 64-byte buffer to pad out the final block if needed
-	// (hoisted out of the main loop to avoid spills)
-	rem := len(src) % 64  // length of final block
-	fin := len(src) - rem // index of final block
-	if rem > 0 {
-		copy(s.buf[len(s.buf)-64:], src[fin:])
-	}
-
-	// pre-calculate most of the first round
-	s1, s5, s9, s13 := quarterRound(j1, s.key[1], s.key[5], s.nonce[0])
-	s2, s6, s10, s14 := quarterRound(j2, s.key[2], s.key[6], s.nonce[1])
-	s3, s7, s11, s15 := quarterRound(j3, s.key[3], s.key[7], s.nonce[2])
-
-	n := len(src)
-	src, dst = src[:n:n], dst[:n:n] // BCE hint
-	for i := 0; i < n; i += 64 {
-		// calculate the remainder of the first round
-		s0, s4, s8, s12 := quarterRound(j0, s.key[0], s.key[4], s.counter)
-
-		// execute the second round
-		x0, x5, x10, x15 := quarterRound(s0, s5, s10, s15)
-		x1, x6, x11, x12 := quarterRound(s1, s6, s11, s12)
-		x2, x7, x8, x13 := quarterRound(s2, s7, s8, s13)
-		x3, x4, x9, x14 := quarterRound(s3, s4, s9, s14)
-
-		// execute the remaining 18 rounds
-		for i := 0; i < 9; i++ {
-			x0, x4, x8, x12 = quarterRound(x0, x4, x8, x12)
-			x1, x5, x9, x13 = quarterRound(x1, x5, x9, x13)
-			x2, x6, x10, x14 = quarterRound(x2, x6, x10, x14)
-			x3, x7, x11, x15 = quarterRound(x3, x7, x11, x15)
-
-			x0, x5, x10, x15 = quarterRound(x0, x5, x10, x15)
-			x1, x6, x11, x12 = quarterRound(x1, x6, x11, x12)
-			x2, x7, x8, x13 = quarterRound(x2, x7, x8, x13)
-			x3, x4, x9, x14 = quarterRound(x3, x4, x9, x14)
-		}
-
-		x0 += j0
-		x1 += j1
-		x2 += j2
-		x3 += j3
-
-		x4 += s.key[0]
-		x5 += s.key[1]
-		x6 += s.key[2]
-		x7 += s.key[3]
-		x8 += s.key[4]
-		x9 += s.key[5]
-		x10 += s.key[6]
-		x11 += s.key[7]
-
-		x12 += s.counter
-		x13 += s.nonce[0]
-		x14 += s.nonce[1]
-		x15 += s.nonce[2]
-
-		// increment the counter
-		s.counter += 1
-		if s.counter == 0 {
-			panic("chacha20: counter overflow")
-		}
-
-		// pad to 64 bytes if needed
-		in, out := src[i:], dst[i:]
-		if i == fin {
-			// src[fin:] has already been copied into s.buf before
-			// the main loop
-			in, out = s.buf[len(s.buf)-64:], s.buf[len(s.buf)-64:]
-		}
-		in, out = in[:64], out[:64] // BCE hint
-
-		// XOR the key stream with the source and write out the result
-		xor(out[0:], in[0:], x0)
-		xor(out[4:], in[4:], x1)
-		xor(out[8:], in[8:], x2)
-		xor(out[12:], in[12:], x3)
-		xor(out[16:], in[16:], x4)
-		xor(out[20:], in[20:], x5)
-		xor(out[24:], in[24:], x6)
-		xor(out[28:], in[28:], x7)
-		xor(out[32:], in[32:], x8)
-		xor(out[36:], in[36:], x9)
-		xor(out[40:], in[40:], x10)
-		xor(out[44:], in[44:], x11)
-		xor(out[48:], in[48:], x12)
-		xor(out[52:], in[52:], x13)
-		xor(out[56:], in[56:], x14)
-		xor(out[60:], in[60:], x15)
-	}
-	// copy any trailing bytes out of the buffer and into dst
-	if rem != 0 {
-		s.len = 64 - rem
-		copy(dst[fin:], s.buf[len(s.buf)-64:])
-	}
-}
-
-// Advance discards bytes in the key stream until the next 64 byte block
-// boundary is reached and updates the counter accordingly. If the key
-// stream is already at a block boundary no bytes will be discarded and
-// the counter will be unchanged.
-func (s *Cipher) Advance() {
-	s.len -= s.len % 64
-	if s.len == 0 {
-		s.buf = [len(s.buf)]byte{}
-	}
-}
-
-// XORKeyStream crypts bytes from in to out using the given key and counters.
-// In and out must overlap entirely or not at all. Counter contains the raw
-// ChaCha20 counter bytes (i.e. block counter followed by nonce).
-func XORKeyStream(out, in []byte, counter *[16]byte, key *[32]byte) {
-	s := Cipher{
-		key: [8]uint32{
-			binary.LittleEndian.Uint32(key[0:4]),
-			binary.LittleEndian.Uint32(key[4:8]),
-			binary.LittleEndian.Uint32(key[8:12]),
-			binary.LittleEndian.Uint32(key[12:16]),
-			binary.LittleEndian.Uint32(key[16:20]),
-			binary.LittleEndian.Uint32(key[20:24]),
-			binary.LittleEndian.Uint32(key[24:28]),
-			binary.LittleEndian.Uint32(key[28:32]),
-		},
-		nonce: [3]uint32{
-			binary.LittleEndian.Uint32(counter[4:8]),
-			binary.LittleEndian.Uint32(counter[8:12]),
-			binary.LittleEndian.Uint32(counter[12:16]),
-		},
-		counter: binary.LittleEndian.Uint32(counter[0:4]),
-	}
-	s.XORKeyStream(out, in)
-}
-
-// HChaCha20 uses the ChaCha20 core to generate a derived key from a key and a
-// nonce. It should only be used as part of the XChaCha20 construction.
-func HChaCha20(key *[8]uint32, nonce *[4]uint32) [8]uint32 {
-	x0, x1, x2, x3 := j0, j1, j2, j3
-	x4, x5, x6, x7 := key[0], key[1], key[2], key[3]
-	x8, x9, x10, x11 := key[4], key[5], key[6], key[7]
-	x12, x13, x14, x15 := nonce[0], nonce[1], nonce[2], nonce[3]
-
-	for i := 0; i < 10; i++ {
-		x0, x4, x8, x12 = quarterRound(x0, x4, x8, x12)
-		x1, x5, x9, x13 = quarterRound(x1, x5, x9, x13)
-		x2, x6, x10, x14 = quarterRound(x2, x6, x10, x14)
-		x3, x7, x11, x15 = quarterRound(x3, x7, x11, x15)
-
-		x0, x5, x10, x15 = quarterRound(x0, x5, x10, x15)
-		x1, x6, x11, x12 = quarterRound(x1, x6, x11, x12)
-		x2, x7, x8, x13 = quarterRound(x2, x7, x8, x13)
-		x3, x4, x9, x14 = quarterRound(x3, x4, x9, x14)
-	}
-
-	var out [8]uint32
-	out[0], out[1], out[2], out[3] = x0, x1, x2, x3
-	out[4], out[5], out[6], out[7] = x12, x13, x14, x15
-	return out
-}

vendor/golang.org/x/crypto/internal/chacha20/chacha_ppc64le.go

@@ -1,53 +0,0 @@
-// 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.
-
-// +build ppc64le,!gccgo,!appengine
-
-package chacha20
-
-import (
-	"encoding/binary"
-)
-
-var haveAsm = true
-
-const bufSize = 256
-
-//go:noescape
-func chaCha20_ctr32_vsx(out, inp *byte, len int, key *[8]uint32, counter *uint32)
-
-func (c *Cipher) xorKeyStreamAsm(dst, src []byte) {
-	// This implementation can handle buffers that aren't multiples of
-	// 256.
-	if len(src) >= bufSize {
-		chaCha20_ctr32_vsx(&dst[0], &src[0], len(src), &c.key, &c.counter)
-	} else if len(src)%bufSize != 0 {
-		chaCha20_ctr32_vsx(&c.buf[0], &c.buf[0], bufSize, &c.key, &c.counter)
-		start := len(src) - len(src)%bufSize
-		ts, td, tb := src[start:], dst[start:], c.buf[:]
-		// Unroll loop to XOR 32 bytes per iteration.
-		for i := 0; i < len(ts)-32; i += 32 {
-			td, tb = td[:len(ts)], tb[:len(ts)] // bounds check elimination
-			s0 := binary.LittleEndian.Uint64(ts[0:8])
-			s1 := binary.LittleEndian.Uint64(ts[8:16])
-			s2 := binary.LittleEndian.Uint64(ts[16:24])
-			s3 := binary.LittleEndian.Uint64(ts[24:32])
-			b0 := binary.LittleEndian.Uint64(tb[0:8])
-			b1 := binary.LittleEndian.Uint64(tb[8:16])
-			b2 := binary.LittleEndian.Uint64(tb[16:24])
-			b3 := binary.LittleEndian.Uint64(tb[24:32])
-			binary.LittleEndian.PutUint64(td[0:8], s0^b0)
-			binary.LittleEndian.PutUint64(td[8:16], s1^b1)
-			binary.LittleEndian.PutUint64(td[16:24], s2^b2)
-			binary.LittleEndian.PutUint64(td[24:32], s3^b3)
-			ts, td, tb = ts[32:], td[32:], tb[32:]
-		}
-		td, tb = td[:len(ts)], tb[:len(ts)] // bounds check elimination
-		for i, v := range ts {
-			td[i] = tb[i] ^ v
-		}
-		c.len = bufSize - (len(src) % bufSize)
-	}
-
-}

vendor/golang.org/x/crypto/internal/chacha20/chacha_s390x.go

@@ -1,29 +0,0 @@
-// 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.
-
-// +build s390x,!gccgo,!appengine
-
-package chacha20
-
-import (
-	"golang.org/x/sys/cpu"
-)
-
-var haveAsm = cpu.S390X.HasVX
-
-const bufSize = 256
-
-// xorKeyStreamVX is an assembly implementation of XORKeyStream. It must only
-// be called when the vector facility is available.
-// Implementation in asm_s390x.s.
-//go:noescape
-func xorKeyStreamVX(dst, src []byte, key *[8]uint32, nonce *[3]uint32, counter *uint32, buf *[256]byte, len *int)
-
-func (c *Cipher) xorKeyStreamAsm(dst, src []byte) {
-	xorKeyStreamVX(dst, src, &c.key, &c.nonce, &c.counter, &c.buf, &c.len)
-}
-
-// EXRL targets, DO NOT CALL!
-func mvcSrcToBuf()
-func mvcBufToDst()

vendor/golang.org/x/crypto/nacl/box/box.go

@@ -31,19 +31,30 @@ Thus large amounts of data should be chunked so that each message is small.
 chunk size.
 
 This package is interoperable with NaCl: https://nacl.cr.yp.to/box.html.
+Anonymous sealing/opening is an extension of NaCl defined by and interoperable
+with libsodium:
+https://libsodium.gitbook.io/doc/public-key_cryptography/sealed_boxes.
 */
 package box // import "golang.org/x/crypto/nacl/box"
 
 import (
+	cryptorand "crypto/rand"
 	"io"
 
+	"golang.org/x/crypto/blake2b"
 	"golang.org/x/crypto/curve25519"
 	"golang.org/x/crypto/nacl/secretbox"
 	"golang.org/x/crypto/salsa20/salsa"
 )
 
-// Overhead is the number of bytes of overhead when boxing a message.
+const (
-const Overhead = secretbox.Overhead
+	// Overhead is the number of bytes of overhead when boxing a message.
+	Overhead = secretbox.Overhead
+
+	// AnonymousOverhead is the number of bytes of overhead when using anonymous
+	// sealed boxes.
+	AnonymousOverhead = Overhead + 32
+)
 
 // GenerateKey generates a new public/private key pair suitable for use with
 // Seal and Open.
@@ -101,3 +112,71 @@ func Open(out, box []byte, nonce *[24]byte, peersPublicKey, privateKey *[32]byte
 func OpenAfterPrecomputation(out, box []byte, nonce *[24]byte, sharedKey *[32]byte) ([]byte, bool) {
 	return secretbox.Open(out, box, nonce, sharedKey)
 }
+
+// SealAnonymous appends an encrypted and authenticated copy of message to out,
+// which will be AnonymousOverhead bytes longer than the original and must not
+// overlap it. This differs from Seal in that the sender is not required to
+// provide a private key.
+func SealAnonymous(out, message []byte, recipient *[32]byte, rand io.Reader) ([]byte, error) {
+	if rand == nil {
+		rand = cryptorand.Reader
+	}
+	ephemeralPub, ephemeralPriv, err := GenerateKey(rand)
+	if err != nil {
+		return nil, err
+	}
+
+	var nonce [24]byte
+	if err := sealNonce(ephemeralPub, recipient, &nonce); err != nil {
+		return nil, err
+	}
+
+	if total := len(out) + AnonymousOverhead + len(message); cap(out) < total {
+		original := out
+		out = make([]byte, 0, total)
+		out = append(out, original...)
+	}
+	out = append(out, ephemeralPub[:]...)
+
+	return Seal(out, message, &nonce, recipient, ephemeralPriv), nil
+}
+
+// OpenAnonymous authenticates and decrypts a box produced by SealAnonymous and
+// appends the message to out, which must not overlap box. The output will be
+// AnonymousOverhead bytes smaller than box.
+func OpenAnonymous(out, box []byte, publicKey, privateKey *[32]byte) (message []byte, ok bool) {
+	if len(box) < AnonymousOverhead {
+		return nil, false
+	}
+
+	var ephemeralPub [32]byte
+	copy(ephemeralPub[:], box[:32])
+
+	var nonce [24]byte
+	if err := sealNonce(&ephemeralPub, publicKey, &nonce); err != nil {
+		return nil, false
+	}
+
+	return Open(out, box[32:], &nonce, &ephemeralPub, privateKey)
+}
+
+// sealNonce generates a 24 byte nonce that is a blake2b digest of the
+// ephemeral public key and the receiver's public key.
+func sealNonce(ephemeralPub, peersPublicKey *[32]byte, nonce *[24]byte) error {
+	h, err := blake2b.New(24, nil)
+	if err != nil {
+		return err
+	}
+
+	if _, err = h.Write(ephemeralPub[:]); err != nil {
+		return err
+	}
+
+	if _, err = h.Write(peersPublicKey[:]); err != nil {
+		return err
+	}
+
+	h.Sum(nonce[:0])
+
+	return nil
+}

vendor/golang.org/x/crypto/poly1305/bits_compat.go

@@ -0,0 +1,39 @@
+// 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.
+
+// +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
+}

vendor/golang.org/x/crypto/poly1305/bits_go1.13.go

@@ -0,0 +1,21 @@
+// 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.
+
+// +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)
+}

vendor/golang.org/x/crypto/poly1305/mac_noasm.go

@@ -2,10 +2,8 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build !amd64,!ppc64le gccgo appengine
+// +build !amd64,!ppc64le,!s390x gccgo purego
 
 package poly1305
 
 type mac struct{ macGeneric }
-
-func newMAC(key *[32]byte) mac { return mac{newMACGeneric(key)} }

vendor/golang.org/x/crypto/poly1305/poly1305.go

@@ -22,8 +22,16 @@ import "crypto/subtle"
 // TagSize is the size, in bytes, of a poly1305 authenticator.
 const TagSize = 16
 
-// Verify returns true if mac is a valid authenticator for m with the given
+// Sum generates an authenticator for msg using a one-time key and puts the
-// key.
+// 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)
@@ -40,10 +48,9 @@ func Verify(mac *[16]byte, m []byte, key *[32]byte) bool {
 // two different messages with the same key allows an attacker
 // to forge messages at will.
 func New(key *[32]byte) *MAC {
-	return &MAC{
+	m := &MAC{}
-		mac:       newMAC(key),
+	initialize(key, &m.macState)
-		finalized: false,
+	return m
-	}
 }
 
 // MAC is an io.Writer computing an authentication tag
@@ -52,7 +59,7 @@ func New(key *[32]byte) *MAC {
 // 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 causes it to panic.
+// Sum or Verify causes it to panic.
 type MAC struct {
 	mac // platform-dependent implementation
 
@@ -65,10 +72,10 @@ 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.
+// 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")
+		panic("poly1305: write to MAC after Sum or Verify")
 	}
 	return h.mac.Write(p)
 }
@@ -81,3 +88,12 @@ func (h *MAC) Sum(b []byte) []byte {
 	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
+}

vendor/golang.org/x/crypto/poly1305/sum_amd64.go

@@ -2,67 +2,46 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build amd64,!gccgo,!appengine
+// +build !gccgo,!purego
 
 package poly1305
 
 //go:noescape
-func initialize(state *[7]uint64, key *[32]byte)
+func update(state *macState, msg []byte)
 
-//go:noescape
+// mac is a wrapper for macGeneric that redirects calls that would have gone to
-func update(state *[7]uint64, msg []byte)
+// 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 }
 
-//go:noescape
+func (h *mac) Write(p []byte) (int, error) {
-func finalize(tag *[TagSize]byte, state *[7]uint64)
+	nn := len(p)
-
-// Sum generates an authenticator for m 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 := newMAC(key)
-	h.Write(m)
-	h.Sum(out)
-}
-
-func newMAC(key *[32]byte) (h mac) {
-	initialize(&h.state, key)
-	return
-}
-
-type mac struct {
-	state [7]uint64 // := uint64{ h0, h1, h2, r0, r1, pad0, pad1 }
-
-	buffer [TagSize]byte
-	offset int
-}
-
-func (h *mac) Write(p []byte) (n int, err error) {
-	n = len(p)
 	if h.offset > 0 {
-		remaining := TagSize - h.offset
+		n := copy(h.buffer[h.offset:], p)
-		if n < remaining {
+		if h.offset+n < TagSize {
-			h.offset += copy(h.buffer[h.offset:], p)
+			h.offset += n
-			return n, nil
+			return nn, nil
 		}
-		copy(h.buffer[h.offset:], p[:remaining])
+		p = p[n:]
-		p = p[remaining:]
 		h.offset = 0
-		update(&h.state, h.buffer[:])
+		update(&h.macState, h.buffer[:])
 	}
-	if nn := len(p) - (len(p) % TagSize); nn > 0 {
+	if n := len(p) - (len(p) % TagSize); n > 0 {
-		update(&h.state, p[:nn])
+		update(&h.macState, p[:n])
-		p = p[nn:]
+		p = p[n:]
 	}
 	if len(p) > 0 {
 		h.offset += copy(h.buffer[h.offset:], p)
 	}
-	return n, nil
+	return nn, nil
 }
 
 func (h *mac) Sum(out *[16]byte) {
-	state := h.state
+	state := h.macState
 	if h.offset > 0 {
 		update(&state, h.buffer[:h.offset])
 	}
-	finalize(out, &state)
+	finalize(out, &state.h, &state.s)
 }

vendor/golang.org/x/crypto/poly1305/sum_amd64.s

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build amd64,!gccgo,!appengine
+// +build !gccgo,!purego
 
 #include "textflag.h"
 
@@ -54,10 +54,6 @@
 	ADCQ  t3, h1;                  \
 	ADCQ  $0, h2
 
-DATA ·poly1305Mask<>+0x00(SB)/8, $0x0FFFFFFC0FFFFFFF
-DATA ·poly1305Mask<>+0x08(SB)/8, $0x0FFFFFFC0FFFFFFC
-GLOBL ·poly1305Mask<>(SB), RODATA, $16
-
 // func update(state *[7]uint64, msg []byte)
 TEXT ·update(SB), $0-32
 	MOVQ state+0(FP), DI
@@ -110,39 +106,3 @@ done:
 	MOVQ R9, 8(DI)
 	MOVQ R10, 16(DI)
 	RET
-
-// func initialize(state *[7]uint64, key *[32]byte)
-TEXT ·initialize(SB), $0-16
-	MOVQ state+0(FP), DI
-	MOVQ key+8(FP), SI
-
-	// state[0...7] is initialized with zero
-	MOVOU 0(SI), X0
-	MOVOU 16(SI), X1
-	MOVOU ·poly1305Mask<>(SB), X2
-	PAND  X2, X0
-	MOVOU X0, 24(DI)
-	MOVOU X1, 40(DI)
-	RET
-
-// func finalize(tag *[TagSize]byte, state *[7]uint64)
-TEXT ·finalize(SB), $0-16
-	MOVQ tag+0(FP), DI
-	MOVQ state+8(FP), SI
-
-	MOVQ    0(SI), AX
-	MOVQ    8(SI), BX
-	MOVQ    16(SI), CX
-	MOVQ    AX, R8
-	MOVQ    BX, R9
-	SUBQ    $0xFFFFFFFFFFFFFFFB, AX
-	SBBQ    $0xFFFFFFFFFFFFFFFF, BX
-	SBBQ    $3, CX
-	CMOVQCS R8, AX
-	CMOVQCS R9, BX
-	ADDQ    40(SI), AX
-	ADCQ    48(SI), BX
-
-	MOVQ AX, 0(DI)
-	MOVQ BX, 8(DI)
-	RET

vendor/golang.org/x/crypto/poly1305/sum_arm.go

@@ -1,22 +0,0 @@
-// Copyright 2015 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.
-
-// +build arm,!gccgo,!appengine,!nacl
-
-package poly1305
-
-// This function is implemented in sum_arm.s
-//go:noescape
-func poly1305_auth_armv6(out *[16]byte, m *byte, mlen uint32, key *[32]byte)
-
-// Sum generates an authenticator for m 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) {
-	var mPtr *byte
-	if len(m) > 0 {
-		mPtr = &m[0]
-	}
-	poly1305_auth_armv6(out, mPtr, uint32(len(m)), key)
-}

vendor/golang.org/x/crypto/poly1305/sum_arm.s

@@ -1,427 +0,0 @@
-// Copyright 2015 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.
-
-// +build arm,!gccgo,!appengine,!nacl
-
-#include "textflag.h"
-
-// This code was translated into a form compatible with 5a from the public
-// domain source by Andrew Moon: github.com/floodyberry/poly1305-opt/blob/master/app/extensions/poly1305.
-
-DATA ·poly1305_init_constants_armv6<>+0x00(SB)/4, $0x3ffffff
-DATA ·poly1305_init_constants_armv6<>+0x04(SB)/4, $0x3ffff03
-DATA ·poly1305_init_constants_armv6<>+0x08(SB)/4, $0x3ffc0ff
-DATA ·poly1305_init_constants_armv6<>+0x0c(SB)/4, $0x3f03fff
-DATA ·poly1305_init_constants_armv6<>+0x10(SB)/4, $0x00fffff
-GLOBL ·poly1305_init_constants_armv6<>(SB), 8, $20
-
-// Warning: the linker may use R11 to synthesize certain instructions. Please
-// take care and verify that no synthetic instructions use it.
-
-TEXT poly1305_init_ext_armv6<>(SB), NOSPLIT, $0
-	// Needs 16 bytes of stack and 64 bytes of space pointed to by R0.  (It
-	// might look like it's only 60 bytes of space but the final four bytes
-	// will be written by another function.) We need to skip over four
-	// bytes of stack because that's saving the value of 'g'.
-	ADD       $4, R13, R8
-	MOVM.IB   [R4-R7], (R8)
-	MOVM.IA.W (R1), [R2-R5]
-	MOVW      $·poly1305_init_constants_armv6<>(SB), R7
-	MOVW      R2, R8
-	MOVW      R2>>26, R9
-	MOVW      R3>>20, g
-	MOVW      R4>>14, R11
-	MOVW      R5>>8, R12
-	ORR       R3<<6, R9, R9
-	ORR       R4<<12, g, g
-	ORR       R5<<18, R11, R11
-	MOVM.IA   (R7), [R2-R6]
-	AND       R8, R2, R2
-	AND       R9, R3, R3
-	AND       g, R4, R4
-	AND       R11, R5, R5
-	AND       R12, R6, R6
-	MOVM.IA.W [R2-R6], (R0)
-	EOR       R2, R2, R2
-	EOR       R3, R3, R3
-	EOR       R4, R4, R4
-	EOR       R5, R5, R5
-	EOR       R6, R6, R6
-	MOVM.IA.W [R2-R6], (R0)
-	MOVM.IA.W (R1), [R2-R5]
-	MOVM.IA   [R2-R6], (R0)
-	ADD       $20, R13, R0
-	MOVM.DA   (R0), [R4-R7]
-	RET
-
-#define MOVW_UNALIGNED(Rsrc, Rdst, Rtmp, offset) \
-	MOVBU (offset+0)(Rsrc), Rtmp; \
-	MOVBU Rtmp, (offset+0)(Rdst); \
-	MOVBU (offset+1)(Rsrc), Rtmp; \
-	MOVBU Rtmp, (offset+1)(Rdst); \
-	MOVBU (offset+2)(Rsrc), Rtmp; \
-	MOVBU Rtmp, (offset+2)(Rdst); \
-	MOVBU (offset+3)(Rsrc), Rtmp; \
-	MOVBU Rtmp, (offset+3)(Rdst)
-
-TEXT poly1305_blocks_armv6<>(SB), NOSPLIT, $0
-	// Needs 24 bytes of stack for saved registers and then 88 bytes of
-	// scratch space after that. We assume that 24 bytes at (R13) have
-	// already been used: four bytes for the link register saved in the
-	// prelude of poly1305_auth_armv6, four bytes for saving the value of g
-	// in that function and 16 bytes of scratch space used around
-	// poly1305_finish_ext_armv6_skip1.
-	ADD     $24, R13, R12
-	MOVM.IB [R4-R8, R14], (R12)
-	MOVW    R0, 88(R13)
-	MOVW    R1, 92(R13)
-	MOVW    R2, 96(R13)
-	MOVW    R1, R14
-	MOVW    R2, R12
-	MOVW    56(R0), R8
-	WORD    $0xe1180008                // TST R8, R8 not working see issue 5921
-	EOR     R6, R6, R6
-	MOVW.EQ $(1<<24), R6
-	MOVW    R6, 84(R13)
-	ADD     $116, R13, g
-	MOVM.IA (R0), [R0-R9]
-	MOVM.IA [R0-R4], (g)
-	CMP     $16, R12
-	BLO     poly1305_blocks_armv6_done
-
-poly1305_blocks_armv6_mainloop:
-	WORD    $0xe31e0003                            // TST R14, #3 not working see issue 5921
-	BEQ     poly1305_blocks_armv6_mainloop_aligned
-	ADD     $100, R13, g
-	MOVW_UNALIGNED(R14, g, R0, 0)
-	MOVW_UNALIGNED(R14, g, R0, 4)
-	MOVW_UNALIGNED(R14, g, R0, 8)
-	MOVW_UNALIGNED(R14, g, R0, 12)
-	MOVM.IA (g), [R0-R3]
-	ADD     $16, R14
-	B       poly1305_blocks_armv6_mainloop_loaded
-
-poly1305_blocks_armv6_mainloop_aligned:
-	MOVM.IA.W (R14), [R0-R3]
-
-poly1305_blocks_armv6_mainloop_loaded:
-	MOVW    R0>>26, g
-	MOVW    R1>>20, R11
-	MOVW    R2>>14, R12
-	MOVW    R14, 92(R13)
-	MOVW    R3>>8, R4
-	ORR     R1<<6, g, g
-	ORR     R2<<12, R11, R11
-	ORR     R3<<18, R12, R12
-	BIC     $0xfc000000, R0, R0
-	BIC     $0xfc000000, g, g
-	MOVW    84(R13), R3
-	BIC     $0xfc000000, R11, R11
-	BIC     $0xfc000000, R12, R12
-	ADD     R0, R5, R5
-	ADD     g, R6, R6
-	ORR     R3, R4, R4
-	ADD     R11, R7, R7
-	ADD     $116, R13, R14
-	ADD     R12, R8, R8
-	ADD     R4, R9, R9
-	MOVM.IA (R14), [R0-R4]
-	MULLU   R4, R5, (R11, g)
-	MULLU   R3, R5, (R14, R12)
-	MULALU  R3, R6, (R11, g)
-	MULALU  R2, R6, (R14, R12)
-	MULALU  R2, R7, (R11, g)
-	MULALU  R1, R7, (R14, R12)
-	ADD     R4<<2, R4, R4
-	ADD     R3<<2, R3, R3
-	MULALU  R1, R8, (R11, g)
-	MULALU  R0, R8, (R14, R12)
-	MULALU  R0, R9, (R11, g)
-	MULALU  R4, R9, (R14, R12)
-	MOVW    g, 76(R13)
-	MOVW    R11, 80(R13)
-	MOVW    R12, 68(R13)
-	MOVW    R14, 72(R13)
-	MULLU   R2, R5, (R11, g)
-	MULLU   R1, R5, (R14, R12)
-	MULALU  R1, R6, (R11, g)
-	MULALU  R0, R6, (R14, R12)
-	MULALU  R0, R7, (R11, g)
-	MULALU  R4, R7, (R14, R12)
-	ADD     R2<<2, R2, R2
-	ADD     R1<<2, R1, R1
-	MULALU  R4, R8, (R11, g)
-	MULALU  R3, R8, (R14, R12)
-	MULALU  R3, R9, (R11, g)
-	MULALU  R2, R9, (R14, R12)
-	MOVW    g, 60(R13)
-	MOVW    R11, 64(R13)
-	MOVW    R12, 52(R13)
-	MOVW    R14, 56(R13)
-	MULLU   R0, R5, (R11, g)
-	MULALU  R4, R6, (R11, g)
-	MULALU  R3, R7, (R11, g)
-	MULALU  R2, R8, (R11, g)
-	MULALU  R1, R9, (R11, g)
-	ADD     $52, R13, R0
-	MOVM.IA (R0), [R0-R7]
-	MOVW    g>>26, R12
-	MOVW    R4>>26, R14
-	ORR     R11<<6, R12, R12
-	ORR     R5<<6, R14, R14
-	BIC     $0xfc000000, g, g
-	BIC     $0xfc000000, R4, R4
-	ADD.S   R12, R0, R0
-	ADC     $0, R1, R1
-	ADD.S   R14, R6, R6
-	ADC     $0, R7, R7
-	MOVW    R0>>26, R12
-	MOVW    R6>>26, R14
-	ORR     R1<<6, R12, R12
-	ORR     R7<<6, R14, R14
-	BIC     $0xfc000000, R0, R0
-	BIC     $0xfc000000, R6, R6
-	ADD     R14<<2, R14, R14
-	ADD.S   R12, R2, R2
-	ADC     $0, R3, R3
-	ADD     R14, g, g
-	MOVW    R2>>26, R12
-	MOVW    g>>26, R14
-	ORR     R3<<6, R12, R12
-	BIC     $0xfc000000, g, R5
-	BIC     $0xfc000000, R2, R7
-	ADD     R12, R4, R4
-	ADD     R14, R0, R0
-	MOVW    R4>>26, R12
-	BIC     $0xfc000000, R4, R8
-	ADD     R12, R6, R9
-	MOVW    96(R13), R12
-	MOVW    92(R13), R14
-	MOVW    R0, R6
-	CMP     $32, R12
-	SUB     $16, R12, R12
-	MOVW    R12, 96(R13)
-	BHS     poly1305_blocks_armv6_mainloop
-
-poly1305_blocks_armv6_done:
-	MOVW    88(R13), R12
-	MOVW    R5, 20(R12)
-	MOVW    R6, 24(R12)
-	MOVW    R7, 28(R12)
-	MOVW    R8, 32(R12)
-	MOVW    R9, 36(R12)
-	ADD     $48, R13, R0
-	MOVM.DA (R0), [R4-R8, R14]
-	RET
-
-#define MOVHUP_UNALIGNED(Rsrc, Rdst, Rtmp) \
-	MOVBU.P 1(Rsrc), Rtmp; \
-	MOVBU.P Rtmp, 1(Rdst); \
-	MOVBU.P 1(Rsrc), Rtmp; \
-	MOVBU.P Rtmp, 1(Rdst)
-
-#define MOVWP_UNALIGNED(Rsrc, Rdst, Rtmp) \
-	MOVHUP_UNALIGNED(Rsrc, Rdst, Rtmp); \
-	MOVHUP_UNALIGNED(Rsrc, Rdst, Rtmp)
-
-// func poly1305_auth_armv6(out *[16]byte, m *byte, mlen uint32, key *[32]key)
-TEXT ·poly1305_auth_armv6(SB), $196-16
-	// The value 196, just above, is the sum of 64 (the size of the context
-	// structure) and 132 (the amount of stack needed).
-	//
-	// At this point, the stack pointer (R13) has been moved down. It
-	// points to the saved link register and there's 196 bytes of free
-	// space above it.
-	//
-	// The stack for this function looks like:
-	//
-	// +---------------------
-	// |
-	// | 64 bytes of context structure
-	// |
-	// +---------------------
-	// |
-	// | 112 bytes for poly1305_blocks_armv6
-	// |
-	// +---------------------
-	// | 16 bytes of final block, constructed at
-	// | poly1305_finish_ext_armv6_skip8
-	// +---------------------
-	// | four bytes of saved 'g'
-	// +---------------------
-	// | lr, saved by prelude    <- R13 points here
-	// +---------------------
-	MOVW g, 4(R13)
-
-	MOVW out+0(FP), R4
-	MOVW m+4(FP), R5
-	MOVW mlen+8(FP), R6
-	MOVW key+12(FP), R7
-
-	ADD  $136, R13, R0 // 136 = 4 + 4 + 16 + 112
-	MOVW R7, R1
-
-	// poly1305_init_ext_armv6 will write to the stack from R13+4, but
-	// that's ok because none of the other values have been written yet.
-	BL    poly1305_init_ext_armv6<>(SB)
-	BIC.S $15, R6, R2
-	BEQ   poly1305_auth_armv6_noblocks
-	ADD   $136, R13, R0
-	MOVW  R5, R1
-	ADD   R2, R5, R5
-	SUB   R2, R6, R6
-	BL    poly1305_blocks_armv6<>(SB)
-
-poly1305_auth_armv6_noblocks:
-	ADD  $136, R13, R0
-	MOVW R5, R1
-	MOVW R6, R2
-	MOVW R4, R3
-
-	MOVW  R0, R5
-	MOVW  R1, R6
-	MOVW  R2, R7
-	MOVW  R3, R8
-	AND.S R2, R2, R2
-	BEQ   poly1305_finish_ext_armv6_noremaining
-	EOR   R0, R0
-	ADD   $8, R13, R9                           // 8 = offset to 16 byte scratch space
-	MOVW  R0, (R9)
-	MOVW  R0, 4(R9)
-	MOVW  R0, 8(R9)
-	MOVW  R0, 12(R9)
-	WORD  $0xe3110003                           // TST R1, #3 not working see issue 5921
-	BEQ   poly1305_finish_ext_armv6_aligned
-	WORD  $0xe3120008                           // TST R2, #8 not working see issue 5921
-	BEQ   poly1305_finish_ext_armv6_skip8
-	MOVWP_UNALIGNED(R1, R9, g)
-	MOVWP_UNALIGNED(R1, R9, g)
-
-poly1305_finish_ext_armv6_skip8:
-	WORD $0xe3120004                     // TST $4, R2 not working see issue 5921
-	BEQ  poly1305_finish_ext_armv6_skip4
-	MOVWP_UNALIGNED(R1, R9, g)
-
-poly1305_finish_ext_armv6_skip4:
-	WORD $0xe3120002                     // TST $2, R2 not working see issue 5921
-	BEQ  poly1305_finish_ext_armv6_skip2
-	MOVHUP_UNALIGNED(R1, R9, g)
-	B    poly1305_finish_ext_armv6_skip2
-
-poly1305_finish_ext_armv6_aligned:
-	WORD      $0xe3120008                             // TST R2, #8 not working see issue 5921
-	BEQ       poly1305_finish_ext_armv6_skip8_aligned
-	MOVM.IA.W (R1), [g-R11]
-	MOVM.IA.W [g-R11], (R9)
-
-poly1305_finish_ext_armv6_skip8_aligned:
-	WORD   $0xe3120004                             // TST $4, R2 not working see issue 5921
-	BEQ    poly1305_finish_ext_armv6_skip4_aligned
-	MOVW.P 4(R1), g
-	MOVW.P g, 4(R9)
-
-poly1305_finish_ext_armv6_skip4_aligned:
-	WORD    $0xe3120002                     // TST $2, R2 not working see issue 5921
-	BEQ     poly1305_finish_ext_armv6_skip2
-	MOVHU.P 2(R1), g
-	MOVH.P  g, 2(R9)
-
-poly1305_finish_ext_armv6_skip2:
-	WORD    $0xe3120001                     // TST $1, R2 not working see issue 5921
-	BEQ     poly1305_finish_ext_armv6_skip1
-	MOVBU.P 1(R1), g
-	MOVBU.P g, 1(R9)
-
-poly1305_finish_ext_armv6_skip1:
-	MOVW  $1, R11
-	MOVBU R11, 0(R9)
-	MOVW  R11, 56(R5)
-	MOVW  R5, R0
-	ADD   $8, R13, R1
-	MOVW  $16, R2
-	BL    poly1305_blocks_armv6<>(SB)
-
-poly1305_finish_ext_armv6_noremaining:
-	MOVW      20(R5), R0
-	MOVW      24(R5), R1
-	MOVW      28(R5), R2
-	MOVW      32(R5), R3
-	MOVW      36(R5), R4
-	MOVW      R4>>26, R12
-	BIC       $0xfc000000, R4, R4
-	ADD       R12<<2, R12, R12
-	ADD       R12, R0, R0
-	MOVW      R0>>26, R12
-	BIC       $0xfc000000, R0, R0
-	ADD       R12, R1, R1
-	MOVW      R1>>26, R12
-	BIC       $0xfc000000, R1, R1
-	ADD       R12, R2, R2
-	MOVW      R2>>26, R12
-	BIC       $0xfc000000, R2, R2
-	ADD       R12, R3, R3
-	MOVW      R3>>26, R12
-	BIC       $0xfc000000, R3, R3
-	ADD       R12, R4, R4
-	ADD       $5, R0, R6
-	MOVW      R6>>26, R12
-	BIC       $0xfc000000, R6, R6
-	ADD       R12, R1, R7
-	MOVW      R7>>26, R12
-	BIC       $0xfc000000, R7, R7
-	ADD       R12, R2, g
-	MOVW      g>>26, R12
-	BIC       $0xfc000000, g, g
-	ADD       R12, R3, R11
-	MOVW      $-(1<<26), R12
-	ADD       R11>>26, R12, R12
-	BIC       $0xfc000000, R11, R11
-	ADD       R12, R4, R9
-	MOVW      R9>>31, R12
-	SUB       $1, R12
-	AND       R12, R6, R6
-	AND       R12, R7, R7
-	AND       R12, g, g
-	AND       R12, R11, R11
-	AND       R12, R9, R9
-	MVN       R12, R12
-	AND       R12, R0, R0
-	AND       R12, R1, R1
-	AND       R12, R2, R2
-	AND       R12, R3, R3
-	AND       R12, R4, R4
-	ORR       R6, R0, R0
-	ORR       R7, R1, R1
-	ORR       g, R2, R2
-	ORR       R11, R3, R3
-	ORR       R9, R4, R4
-	ORR       R1<<26, R0, R0
-	MOVW      R1>>6, R1
-	ORR       R2<<20, R1, R1
-	MOVW      R2>>12, R2
-	ORR       R3<<14, R2, R2
-	MOVW      R3>>18, R3
-	ORR       R4<<8, R3, R3
-	MOVW      40(R5), R6
-	MOVW      44(R5), R7
-	MOVW      48(R5), g
-	MOVW      52(R5), R11
-	ADD.S     R6, R0, R0
-	ADC.S     R7, R1, R1
-	ADC.S     g, R2, R2
-	ADC.S     R11, R3, R3
-	MOVM.IA   [R0-R3], (R8)
-	MOVW      R5, R12
-	EOR       R0, R0, R0
-	EOR       R1, R1, R1
-	EOR       R2, R2, R2
-	EOR       R3, R3, R3
-	EOR       R4, R4, R4
-	EOR       R5, R5, R5
-	EOR       R6, R6, R6
-	EOR       R7, R7, R7
-	MOVM.IA.W [R0-R7], (R12)
-	MOVM.IA   [R0-R7], (R12)
-	MOVW      4(R13), g
-	RET

vendor/golang.org/x/crypto/poly1305/sum_generic.go

@@ -2,171 +2,309 @@
 // 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"
 
-const (
+// Poly1305 [RFC 7539] is a relatively simple algorithm: the authentication tag
-	msgBlock   = uint32(1 << 24)
+// for a 64 bytes message is approximately
-	finalBlock = uint32(0)
+//
-)
+//     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.
 
-// sumGeneric generates an authenticator for msg using a one-time key and
-// puts the 16-byte result into out. This is the generic implementation of
-// Sum and should be called if no assembly implementation is available.
 func sumGeneric(out *[TagSize]byte, msg []byte, key *[32]byte) {
 	h := newMACGeneric(key)
 	h.Write(msg)
 	h.Sum(out)
 }
 
-func newMACGeneric(key *[32]byte) (h macGeneric) {
+func newMACGeneric(key *[32]byte) macGeneric {
-	h.r[0] = binary.LittleEndian.Uint32(key[0:]) & 0x3ffffff
+	m := macGeneric{}
-	h.r[1] = (binary.LittleEndian.Uint32(key[3:]) >> 2) & 0x3ffff03
+	initialize(key, &m.macState)
-	h.r[2] = (binary.LittleEndian.Uint32(key[6:]) >> 4) & 0x3ffc0ff
+	return m
-	h.r[3] = (binary.LittleEndian.Uint32(key[9:]) >> 6) & 0x3f03fff
+}
-	h.r[4] = (binary.LittleEndian.Uint32(key[12:]) >> 8) & 0x00fffff
 
-	h.s[0] = binary.LittleEndian.Uint32(key[16:])
+// macState holds numbers in saturated 64-bit little-endian limbs. That is,
-	h.s[1] = binary.LittleEndian.Uint32(key[20:])
+// the value of [x0, x1, x2] is x[0] + x[1] * 2⁶⁴ + x[2] * 2¹²⁸.
-	h.s[2] = binary.LittleEndian.Uint32(key[24:])
+type macState struct {
-	h.s[3] = binary.LittleEndian.Uint32(key[28:])
+	// h is the main accumulator. It is to be interpreted modulo 2¹³⁰ - 5, but
-	return
+	// 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 {
-	h, r [5]uint32
+	macState
-	s    [4]uint32
 
 	buffer [TagSize]byte
 	offset int
 }
 
-func (h *macGeneric) Write(p []byte) (n int, err error) {
+// Write splits the incoming message into TagSize chunks, and passes them to
-	n = len(p)
+// update. It buffers incomplete chunks.
+func (h *macGeneric) Write(p []byte) (int, error) {
+	nn := len(p)
 	if h.offset > 0 {
-		remaining := TagSize - h.offset
+		n := copy(h.buffer[h.offset:], p)
-		if n < remaining {
+		if h.offset+n < TagSize {
-			h.offset += copy(h.buffer[h.offset:], p)
+			h.offset += n
-			return n, nil
+			return nn, nil
 		}
-		copy(h.buffer[h.offset:], p[:remaining])
+		p = p[n:]
-		p = p[remaining:]
 		h.offset = 0
-		updateGeneric(h.buffer[:], msgBlock, &(h.h), &(h.r))
+		updateGeneric(&h.macState, h.buffer[:])
 	}
-	if nn := len(p) - (len(p) % TagSize); nn > 0 {
+	if n := len(p) - (len(p) % TagSize); n > 0 {
-		updateGeneric(p, msgBlock, &(h.h), &(h.r))
+		updateGeneric(&h.macState, p[:n])
-		p = p[nn:]
+		p = p[n:]
 	}
 	if len(p) > 0 {
 		h.offset += copy(h.buffer[h.offset:], p)
 	}
-	return n, nil
+	return nn, nil
 }
 
-func (h *macGeneric) Sum(out *[16]byte) {
+// Sum flushes the last incomplete chunk from the buffer, if any, and generates
-	H, R := h.h, h.r
+// 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 {
-		var buffer [TagSize]byte
+		updateGeneric(&state, h.buffer[:h.offset])
-		copy(buffer[:], h.buffer[:h.offset])
-		buffer[h.offset] = 1 // invariant: h.offset < TagSize
-		updateGeneric(buffer[:], finalBlock, &H, &R)
 	}
-	finalizeGeneric(out, &H, &(h.s))
+	finalize(out, &state.h, &state.s)
 }
 
-func updateGeneric(msg []byte, flag uint32, h, r *[5]uint32) {
+// [rMask0, rMask1] is the specified Poly1305 clamping mask in little-endian. It
-	h0, h1, h2, h3, h4 := h[0], h[1], h[2], h[3], h[4]
+// clears some bits of the secret coefficient to make it possible to implement
-	r0, r1, r2, r3, r4 := uint64(r[0]), uint64(r[1]), uint64(r[2]), uint64(r[3]), uint64(r[4])
+// multiplication more efficiently.
-	R1, R2, R3, R4 := r1*5, r2*5, r3*5, r4*5
+const (
+	rMask0 = 0x0FFFFFFC0FFFFFFF
+	rMask1 = 0x0FFFFFFC0FFFFFFC
+)
 
-	for len(msg) >= TagSize {
+// initialize loads the 256-bit key into the two 128-bit secret values r and s.
-		// h += msg
+func initialize(key *[32]byte, m *macState) {
-		h0 += binary.LittleEndian.Uint32(msg[0:]) & 0x3ffffff
+	m.r[0] = binary.LittleEndian.Uint64(key[0:8]) & rMask0
-		h1 += (binary.LittleEndian.Uint32(msg[3:]) >> 2) & 0x3ffffff
+	m.r[1] = binary.LittleEndian.Uint64(key[8:16]) & rMask1
-		h2 += (binary.LittleEndian.Uint32(msg[6:]) >> 4) & 0x3ffffff
+	m.s[0] = binary.LittleEndian.Uint64(key[16:24])
-		h3 += (binary.LittleEndian.Uint32(msg[9:]) >> 6) & 0x3ffffff
+	m.s[1] = binary.LittleEndian.Uint64(key[24:32])
-		h4 += (binary.LittleEndian.Uint32(msg[12:]) >> 8) | flag
+}
 
-		// h *= r
+// uint128 holds a 128-bit number as two 64-bit limbs, for use with the
-		d0 := (uint64(h0) * r0) + (uint64(h1) * R4) + (uint64(h2) * R3) + (uint64(h3) * R2) + (uint64(h4) * R1)
+// bits.Mul64 and bits.Add64 intrinsics.
-		d1 := (d0 >> 26) + (uint64(h0) * r1) + (uint64(h1) * r0) + (uint64(h2) * R4) + (uint64(h3) * R3) + (uint64(h4) * R2)
+type uint128 struct {
-		d2 := (d1 >> 26) + (uint64(h0) * r2) + (uint64(h1) * r1) + (uint64(h2) * r0) + (uint64(h3) * R4) + (uint64(h4) * R3)
+	lo, hi uint64
-		d3 := (d2 >> 26) + (uint64(h0) * r3) + (uint64(h1) * r2) + (uint64(h2) * r1) + (uint64(h3) * r0) + (uint64(h4) * R4)
+}
-		d4 := (d3 >> 26) + (uint64(h0) * r4) + (uint64(h1) * r3) + (uint64(h2) * r2) + (uint64(h3) * r1) + (uint64(h4) * r0)
 
-		// h %= p
+func mul64(a, b uint64) uint128 {
-		h0 = uint32(d0) & 0x3ffffff
+	hi, lo := bitsMul64(a, b)
-		h1 = uint32(d1) & 0x3ffffff
+	return uint128{lo, hi}
-		h2 = uint32(d2) & 0x3ffffff
+}
-		h3 = uint32(d3) & 0x3ffffff
-		h4 = uint32(d4) & 0x3ffffff
 
-		h0 += uint32(d4>>26) * 5
+func add128(a, b uint128) uint128 {
-		h1 += h0 >> 26
+	lo, c := bitsAdd64(a.lo, b.lo, 0)
-		h0 = h0 & 0x3ffffff
+	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
 		}
 
-	h[0], h[1], h[2], h[3], h[4] = h0, h1, h2, h3, h4
+		// 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
 }
 
-func finalizeGeneric(out *[TagSize]byte, h *[5]uint32, s *[4]uint32) {
+const (
-	h0, h1, h2, h3, h4 := h[0], h[1], h[2], h[3], h[4]
+	maskLow2Bits    uint64 = 0x0000000000000003
+	maskNotLow2Bits uint64 = ^maskLow2Bits
+)
 
-	// h %= p reduction
+// select64 returns x if v == 1 and y if v == 0, in constant time.
-	h2 += h1 >> 26
+func select64(v, x, y uint64) uint64 { return ^(v-1)&x | (v-1)&y }
-	h1 &= 0x3ffffff
-	h3 += h2 >> 26
-	h2 &= 0x3ffffff
-	h4 += h3 >> 26
-	h3 &= 0x3ffffff
-	h0 += 5 * (h4 >> 26)
-	h4 &= 0x3ffffff
-	h1 += h0 >> 26
-	h0 &= 0x3ffffff
 
-	// h - p
+// [p0, p1, p2] is 2¹³⁰ - 5 in little endian order.
-	t0 := h0 + 5
+const (
-	t1 := h1 + (t0 >> 26)
+	p0 = 0xFFFFFFFFFFFFFFFB
-	t2 := h2 + (t1 >> 26)
+	p1 = 0xFFFFFFFFFFFFFFFF
-	t3 := h3 + (t2 >> 26)
+	p2 = 0x0000000000000003
-	t4 := h4 + (t3 >> 26) - (1 << 26)
+)
-	t0 &= 0x3ffffff
-	t1 &= 0x3ffffff
-	t2 &= 0x3ffffff
-	t3 &= 0x3ffffff
 
-	// select h if h < p else h - p
+// finalize completes the modular reduction of h and computes
-	t_mask := (t4 >> 31) - 1
+//
-	h_mask := ^t_mask
+//     out = h + s  mod  2¹²⁸
-	h0 = (h0 & h_mask) | (t0 & t_mask)
+//
-	h1 = (h1 & h_mask) | (t1 & t_mask)
+func finalize(out *[TagSize]byte, h *[3]uint64, s *[2]uint64) {
-	h2 = (h2 & h_mask) | (t2 & t_mask)
+	h0, h1, h2 := h[0], h[1], h[2]
-	h3 = (h3 & h_mask) | (t3 & t_mask)
-	h4 = (h4 & h_mask) | (t4 & t_mask)
 
-	// h %= 2^128
+	// After the partial reduction in updateGeneric, h might be more than
-	h0 |= h1 << 26
+	// 2¹³⁰ - 5, but will be less than 2 * (2¹³⁰ - 5). To complete the reduction
-	h1 = ((h1 >> 6) | (h2 << 20))
+	// in constant time, we compute t = h - (2¹³⁰ - 5), and select h as the
-	h2 = ((h2 >> 12) | (h3 << 14))
+	// result if the subtraction underflows, and t otherwise.
-	h3 = ((h3 >> 18) | (h4 << 8))
 
-	// s: the s part of the key
+	hMinusP0, b := bitsSub64(h0, p0, 0)
-	// tag = (h + s) % (2^128)
+	hMinusP1, b := bitsSub64(h1, p1, b)
-	t := uint64(h0) + uint64(s[0])
+	_, b = bitsSub64(h2, p2, b)
-	h0 = uint32(t)
-	t = uint64(h1) + uint64(s[1]) + (t >> 32)
-	h1 = uint32(t)
-	t = uint64(h2) + uint64(s[2]) + (t >> 32)
-	h2 = uint32(t)
-	t = uint64(h3) + uint64(s[3]) + (t >> 32)
-	h3 = uint32(t)
 
-	binary.LittleEndian.PutUint32(out[0:], h0)
+	// h = h if h < p else h - p
-	binary.LittleEndian.PutUint32(out[4:], h1)
+	h0 = select64(b, h0, hMinusP0)
-	binary.LittleEndian.PutUint32(out[8:], h2)
+	h1 = select64(b, h1, hMinusP1)
-	binary.LittleEndian.PutUint32(out[12:], h3)
+
+	// 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)
 }

vendor/golang.org/x/crypto/poly1305/sum_noasm.go

@@ -1,16 +0,0 @@
-// 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.
-
-// +build s390x,!go1.11 !arm,!amd64,!s390x,!ppc64le gccgo appengine nacl
-
-package poly1305
-
-// 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 *[TagSize]byte, msg []byte, key *[32]byte) {
-	h := newMAC(key)
-	h.Write(msg)
-	h.Sum(out)
-}

vendor/golang.org/x/crypto/poly1305/sum_ppc64le.go

@@ -2,67 +2,46 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build ppc64le,!gccgo,!appengine
+// +build !gccgo,!purego
 
 package poly1305
 
 //go:noescape
-func initialize(state *[7]uint64, key *[32]byte)
+func update(state *macState, msg []byte)
 
-//go:noescape
+// mac is a wrapper for macGeneric that redirects calls that would have gone to
-func update(state *[7]uint64, msg []byte)
+// 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 }
 
-//go:noescape
+func (h *mac) Write(p []byte) (int, error) {
-func finalize(tag *[TagSize]byte, state *[7]uint64)
+	nn := len(p)
-
-// Sum generates an authenticator for m 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 := newMAC(key)
-	h.Write(m)
-	h.Sum(out)
-}
-
-func newMAC(key *[32]byte) (h mac) {
-	initialize(&h.state, key)
-	return
-}
-
-type mac struct {
-	state [7]uint64 // := uint64{ h0, h1, h2, r0, r1, pad0, pad1 }
-
-	buffer [TagSize]byte
-	offset int
-}
-
-func (h *mac) Write(p []byte) (n int, err error) {
-	n = len(p)
 	if h.offset > 0 {
-		remaining := TagSize - h.offset
+		n := copy(h.buffer[h.offset:], p)
-		if n < remaining {
+		if h.offset+n < TagSize {
-			h.offset += copy(h.buffer[h.offset:], p)
+			h.offset += n
-			return n, nil
+			return nn, nil
 		}
-		copy(h.buffer[h.offset:], p[:remaining])
+		p = p[n:]
-		p = p[remaining:]
 		h.offset = 0
-		update(&h.state, h.buffer[:])
+		update(&h.macState, h.buffer[:])
 	}
-	if nn := len(p) - (len(p) % TagSize); nn > 0 {
+	if n := len(p) - (len(p) % TagSize); n > 0 {
-		update(&h.state, p[:nn])
+		update(&h.macState, p[:n])
-		p = p[nn:]
+		p = p[n:]
 	}
 	if len(p) > 0 {
 		h.offset += copy(h.buffer[h.offset:], p)
 	}
-	return n, nil
+	return nn, nil
 }
 
 func (h *mac) Sum(out *[16]byte) {
-	state := h.state
+	state := h.macState
 	if h.offset > 0 {
 		update(&state, h.buffer[:h.offset])
 	}
-	finalize(out, &state)
+	finalize(out, &state.h, &state.s)
 }

vendor/golang.org/x/crypto/poly1305/sum_ppc64le.s

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build ppc64le,!gccgo,!appengine
+// +build !gccgo,!purego
 
 #include "textflag.h"
 
@@ -58,7 +58,6 @@ DATA ·poly1305Mask<>+0x08(SB)/8, $0x0FFFFFFC0FFFFFFC
 GLOBL ·poly1305Mask<>(SB), RODATA, $16
 
 // func update(state *[7]uint64, msg []byte)
-
 TEXT ·update(SB), $0-32
 	MOVD state+0(FP), R3
 	MOVD msg_base+8(FP), R4
@@ -180,68 +179,3 @@ done:
 	MOVD R9, 8(R3)
 	MOVD R10, 16(R3)
 	RET
-
-// func initialize(state *[7]uint64, key *[32]byte)
-TEXT ·initialize(SB), $0-16
-	MOVD state+0(FP), R3
-	MOVD key+8(FP), R4
-
-	// state[0...7] is initialized with zero
-	// Load key
-	MOVD 0(R4), R5
-	MOVD 8(R4), R6
-	MOVD 16(R4), R7
-	MOVD 24(R4), R8
-
-	// Address of key mask
-	MOVD $·poly1305Mask<>(SB), R9
-
-	// Save original key in state
-	MOVD R7, 40(R3)
-	MOVD R8, 48(R3)
-
-	// Get mask
-	MOVD (R9), R7
-	MOVD 8(R9), R8
-
-	// And with key
-	AND R5, R7, R5
-	AND R6, R8, R6
-
-	// Save masked key in state
-	MOVD R5, 24(R3)
-	MOVD R6, 32(R3)
-	RET
-
-// func finalize(tag *[TagSize]byte, state *[7]uint64)
-TEXT ·finalize(SB), $0-16
-	MOVD tag+0(FP), R3
-	MOVD state+8(FP), R4
-
-	// Get h0, h1, h2 from state
-	MOVD 0(R4), R5
-	MOVD 8(R4), R6
-	MOVD 16(R4), R7
-
-	// Save h0, h1
-	MOVD  R5, R8
-	MOVD  R6, R9
-	MOVD  $3, R20
-	MOVD  $-1, R21
-	SUBC  $-5, R5
-	SUBE  R21, R6
-	SUBE  R20, R7
-	MOVD  $0, R21
-	SUBZE R21
-
-	// Check for carry
-	CMP  $0, R21
-	ISEL $2, R5, R8, R5
-	ISEL $2, R6, R9, R6
-	MOVD 40(R4), R8
-	MOVD 48(R4), R9
-	ADDC R8, R5
-	ADDE R9, R6
-	MOVD R5, 0(R3)
-	MOVD R6, 8(R3)
-	RET

vendor/golang.org/x/crypto/poly1305/sum_s390x.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build s390x,go1.11,!gccgo,!appengine
+// +build !gccgo,!purego
 
 package poly1305
 
@@ -10,33 +10,66 @@ import (
 	"golang.org/x/sys/cpu"
 )
 
-// poly1305vx is an assembly implementation of Poly1305 that uses vector
+// 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 poly1305vx(out *[16]byte, m *byte, mlen uint64, key *[32]byte)
+func updateVX(state *macState, msg []byte)
 
-// poly1305vmsl is an assembly implementation of Poly1305 that uses vector
+// mac is a replacement for macGeneric that uses a larger buffer and redirects
-// instructions, including VMSL. It must only be called if the vector facility (vx) is
+// calls that would have gone to updateGeneric to updateVX if the vector
-// available and if VMSL is supported.
+// facility is installed.
-//go:noescape
+//
-func poly1305vmsl(out *[16]byte, m *byte, mlen uint64, key *[32]byte)
+// 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
 
-// Sum generates an authenticator for m using a one-time key and puts the
+	buffer [16 * TagSize]byte // size must be a multiple of block size (16)
-// 16-byte result into out. Authenticating two different messages with the same
+	offset int
-// key allows an attacker to forge messages at will.
+}
-func Sum(out *[16]byte, m []byte, key *[32]byte) {
+
-	if cpu.S390X.HasVX {
+func (h *mac) Write(p []byte) (int, error) {
-		var mPtr *byte
+	nn := len(p)
-		if len(m) > 0 {
+	if h.offset > 0 {
-			mPtr = &m[0]
+		n := copy(h.buffer[h.offset:], p)
-		}
+		if h.offset+n < len(h.buffer) {
-		if cpu.S390X.HasVXE && len(m) > 256 {
+			h.offset += n
-			poly1305vmsl(out, mPtr, uint64(len(m)), key)
+			return nn, nil
-		} else {
+		}
-			poly1305vx(out, mPtr, uint64(len(m)), key)
+		p = p[n:]
-		}
+		h.offset = 0
-	} else {
+		if cpu.S390X.HasVX {
-		sumGeneric(out, m, key)
+			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)
 }

vendor/golang.org/x/crypto/poly1305/sum_s390x.s

@@ -2,115 +2,187 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build s390x,go1.11,!gccgo,!appengine
+// +build !gccgo,!purego
 
 #include "textflag.h"
 
-// Implementation of Poly1305 using the vector facility (vx).
+// This implementation of Poly1305 uses the vector facility (vx)
+// to process up to 2 blocks (32 bytes) per iteration using an
+// algorithm based on the one described in:
+//
+// NEON crypto, Daniel J. Bernstein & Peter Schwabe
+// https://cryptojedi.org/papers/neoncrypto-20120320.pdf
+//
+// This algorithm uses 5 26-bit limbs to represent a 130-bit
+// value. These limbs are, for the most part, zero extended and
+// placed into 64-bit vector register elements. Each vector
+// register is 128-bits wide and so holds 2 of these elements.
+// Using 26-bit limbs allows us plenty of headroom to accomodate
+// accumulations before and after multiplication without
+// overflowing either 32-bits (before multiplication) or 64-bits
+// (after multiplication).
+//
+// In order to parallelise the operations required to calculate
+// the sum we use two separate accumulators and then sum those
+// in an extra final step. For compatibility with the generic
+// implementation we perform this summation at the end of every
+// updateVX call.
+//
+// To use two accumulators we must multiply the message blocks
+// by r² rather than r. Only the final message block should be
+// multiplied by r.
+//
+// Example:
+//
+// We want to calculate the sum (h) for a 64 byte message (m):
+//
+//   h = m[0:16]r⁴ + m[16:32]r³ + m[32:48]r² + m[48:64]r
+//
+// To do this we split the calculation into the even indices
+// and odd indices of the message. These form our SIMD 'lanes':
+//
+//   h = m[ 0:16]r⁴ + m[32:48]r² +   <- lane 0
+//       m[16:32]r³ + m[48:64]r      <- lane 1
+//
+// To calculate this iteratively we refactor so that both lanes
+// are written in terms of r² and r:
+//
+//   h = (m[ 0:16]r² + m[32:48])r² + <- lane 0
+//       (m[16:32]r² + m[48:64])r    <- lane 1
+//                ^             ^
+//                |             coefficients for second iteration
+//                coefficients for first iteration
+//
+// So in this case we would have two iterations. In the first
+// both lanes are multiplied by r². In the second only the
+// first lane is multiplied by r² and the second lane is
+// instead multiplied by r. This gives use the odd and even
+// powers of r that we need from the original equation.
+//
+// Notation:
+//
+//   h - accumulator
+//   r - key
+//   m - message
+//
+//   [a, b]       - SIMD register holding two 64-bit values
+//   [a, b, c, d] - SIMD register holding four 32-bit values
+//   xᵢ[n]        - limb n of variable x with bit width i
+//
+// Limbs are expressed in little endian order, so for 26-bit
+// limbs x₂₆[4] will be the most significant limb and x₂₆[0]
+// will be the least significant limb.
 
-// constants
+// masking constants
-#define MOD26 V0
+#define MOD24 V0 // [0x0000000000ffffff, 0x0000000000ffffff] - mask low 24-bits
-#define EX0   V1
+#define MOD26 V1 // [0x0000000003ffffff, 0x0000000003ffffff] - mask low 26-bits
-#define EX1   V2
-#define EX2   V3
 
-// temporaries
+// expansion constants (see EXPAND macro)
-#define T_0 V4
+#define EX0 V2
-#define T_1 V5
+#define EX1 V3
-#define T_2 V6
+#define EX2 V4
-#define T_3 V7
-#define T_4 V8
 
-// key (r)
+// key (r², r or 1 depending on context)
-#define R_0  V9
+#define R_0 V5
-#define R_1  V10
+#define R_1 V6
-#define R_2  V11
+#define R_2 V7
-#define R_3  V12
+#define R_3 V8
-#define R_4  V13
+#define R_4 V9
-#define R5_1 V14
-#define R5_2 V15
-#define R5_3 V16
-#define R5_4 V17
-#define RSAVE_0 R5
-#define RSAVE_1 R6
-#define RSAVE_2 R7
-#define RSAVE_3 R8
-#define RSAVE_4 R9
-#define R5SAVE_1 V28
-#define R5SAVE_2 V29
-#define R5SAVE_3 V30
-#define R5SAVE_4 V31
 
-// message block
+// precalculated coefficients (5r², 5r or 0 depending on context)
-#define F_0 V18
+#define R5_1 V10
-#define F_1 V19
+#define R5_2 V11
-#define F_2 V20
+#define R5_3 V12
-#define F_3 V21
+#define R5_4 V13
-#define F_4 V22
 
-// accumulator
+// message block (m)
-#define H_0 V23
+#define M_0 V14
-#define H_1 V24
+#define M_1 V15
-#define H_2 V25
+#define M_2 V16
-#define H_3 V26
+#define M_3 V17
-#define H_4 V27
+#define M_4 V18
 
-GLOBL ·keyMask<>(SB), RODATA, $16
+// accumulator (h)
-DATA ·keyMask<>+0(SB)/8, $0xffffff0ffcffff0f
+#define H_0 V19
-DATA ·keyMask<>+8(SB)/8, $0xfcffff0ffcffff0f
+#define H_1 V20
+#define H_2 V21
+#define H_3 V22
+#define H_4 V23
 
-GLOBL ·bswapMask<>(SB), RODATA, $16
+// temporary registers (for short-lived values)
-DATA ·bswapMask<>+0(SB)/8, $0x0f0e0d0c0b0a0908
+#define T_0 V24
-DATA ·bswapMask<>+8(SB)/8, $0x0706050403020100
+#define T_1 V25
+#define T_2 V26
+#define T_3 V27
+#define T_4 V28
 
-GLOBL ·constants<>(SB), RODATA, $64
+GLOBL ·constants<>(SB), RODATA, $0x30
-// MOD26
-DATA ·constants<>+0(SB)/8, $0x3ffffff
-DATA ·constants<>+8(SB)/8, $0x3ffffff
 // EX0
-DATA ·constants<>+16(SB)/8, $0x0006050403020100
+DATA ·constants<>+0x00(SB)/8, $0x0006050403020100
-DATA ·constants<>+24(SB)/8, $0x1016151413121110
+DATA ·constants<>+0x08(SB)/8, $0x1016151413121110
 // EX1
-DATA ·constants<>+32(SB)/8, $0x060c0b0a09080706
+DATA ·constants<>+0x10(SB)/8, $0x060c0b0a09080706
-DATA ·constants<>+40(SB)/8, $0x161c1b1a19181716
+DATA ·constants<>+0x18(SB)/8, $0x161c1b1a19181716
 // EX2
-DATA ·constants<>+48(SB)/8, $0x0d0d0d0d0d0f0e0d
+DATA ·constants<>+0x20(SB)/8, $0x0d0d0d0d0d0f0e0d
-DATA ·constants<>+56(SB)/8, $0x1d1d1d1d1d1f1e1d
+DATA ·constants<>+0x28(SB)/8, $0x1d1d1d1d1d1f1e1d
 
-// h = (f*g) % (2**130-5) [partial reduction]
+// MULTIPLY multiplies each lane of f and g, partially reduced
+// modulo 2¹³⁰ - 5. The result, h, consists of partial products
+// in each lane that need to be reduced further to produce the
+// final result.
+//
+//   h₁₃₀ = (f₁₃₀g₁₃₀) % 2¹³⁰ + (5f₁₃₀g₁₃₀) / 2¹³⁰
+//
+// Note that the multiplication by 5 of the high bits is
+// achieved by precalculating the multiplication of four of the
+// g coefficients by 5. These are g51-g54.
 #define MULTIPLY(f0, f1, f2, f3, f4, g0, g1, g2, g3, g4, g51, g52, g53, g54, h0, h1, h2, h3, h4) \
 	VMLOF  f0, g0, h0        \
-	VMLOF  f0, g1, h1        \
-	VMLOF  f0, g2, h2        \
 	VMLOF  f0, g3, h3        \
+	VMLOF  f0, g1, h1        \
 	VMLOF  f0, g4, h4        \
+	VMLOF  f0, g2, h2        \
 	VMLOF  f1, g54, T_0      \
-	VMLOF  f1, g0, T_1       \
-	VMLOF  f1, g1, T_2       \
 	VMLOF  f1, g2, T_3       \
+	VMLOF  f1, g0, T_1       \
 	VMLOF  f1, g3, T_4       \
+	VMLOF  f1, g1, T_2       \
 	VMALOF f2, g53, h0, h0   \
-	VMALOF f2, g54, h1, h1   \
-	VMALOF f2, g0, h2, h2    \
 	VMALOF f2, g1, h3, h3    \
+	VMALOF f2, g54, h1, h1   \
 	VMALOF f2, g2, h4, h4    \
+	VMALOF f2, g0, h2, h2    \
 	VMALOF f3, g52, T_0, T_0 \
-	VMALOF f3, g53, T_1, T_1 \
-	VMALOF f3, g54, T_2, T_2 \
 	VMALOF f3, g0, T_3, T_3  \
+	VMALOF f3, g53, T_1, T_1 \
 	VMALOF f3, g1, T_4, T_4  \
+	VMALOF f3, g54, T_2, T_2 \
 	VMALOF f4, g51, h0, h0   \
-	VMALOF f4, g52, h1, h1   \
-	VMALOF f4, g53, h2, h2   \
 	VMALOF f4, g54, h3, h3   \
+	VMALOF f4, g52, h1, h1   \
 	VMALOF f4, g0, h4, h4    \
+	VMALOF f4, g53, h2, h2   \
 	VAG    T_0, h0, h0       \
-	VAG    T_1, h1, h1       \
-	VAG    T_2, h2, h2       \
 	VAG    T_3, h3, h3       \
-	VAG    T_4, h4, h4
+	VAG    T_1, h1, h1       \
+	VAG    T_4, h4, h4       \
+	VAG    T_2, h2, h2
 
-// carry h0->h1 h3->h4, h1->h2 h4->h0, h0->h1 h2->h3, h3->h4
+// REDUCE performs the following carry operations in four
+// stages, as specified in Bernstein & Schwabe:
+//
+//   1: h₂₆[0]->h₂₆[1] h₂₆[3]->h₂₆[4]
+//   2: h₂₆[1]->h₂₆[2] h₂₆[4]->h₂₆[0]
+//   3: h₂₆[0]->h₂₆[1] h₂₆[2]->h₂₆[3]
+//   4: h₂₆[3]->h₂₆[4]
+//
+// The result is that all of the limbs are limited to 26-bits
+// except for h₂₆[1] and h₂₆[4] which are limited to 27-bits.
+//
+// Note that although each limb is aligned at 26-bit intervals
+// they may contain values that exceed 2²⁶ - 1, hence the need
+// to carry the excess bits in each limb.
 #define REDUCE(h0, h1, h2, h3, h4) \
 	VESRLG $26, h0, T_0  \
 	VESRLG $26, h3, T_1  \
@@ -136,144 +208,155 @@ DATA ·constants<>+56(SB)/8, $0x1d1d1d1d1d1f1e1d
 	VN     MOD26, h3, h3 \
 	VAG    T_2, h4, h4
 
-// expand in0 into d[0] and in1 into d[1]
+// EXPAND splits the 128-bit little-endian values in0 and in1
+// into 26-bit big-endian limbs and places the results into
+// the first and second lane of d₂₆[0:4] respectively.
+//
+// The EX0, EX1 and EX2 constants are arrays of byte indices
+// for permutation. The permutation both reverses the bytes
+// in the input and ensures the bytes are copied into the
+// destination limb ready to be shifted into their final
+// position.
 #define EXPAND(in0, in1, d0, d1, d2, d3, d4) \
-	VGBM   $0x0707, d1       \ // d1=tmp
-	VPERM  in0, in1, EX2, d4 \
 	VPERM  in0, in1, EX0, d0 \
 	VPERM  in0, in1, EX1, d2 \
-	VN     d1, d4, d4        \
+	VPERM  in0, in1, EX2, d4 \
 	VESRLG $26, d0, d1       \
 	VESRLG $30, d2, d3       \
 	VESRLG $4, d2, d2        \
-	VN     MOD26, d0, d0     \
+	VN     MOD26, d0, d0     \ // [in0₂₆[0], in1₂₆[0]]
-	VN     MOD26, d1, d1     \
+	VN     MOD26, d3, d3     \ // [in0₂₆[3], in1₂₆[3]]
-	VN     MOD26, d2, d2     \
+	VN     MOD26, d1, d1     \ // [in0₂₆[1], in1₂₆[1]]
-	VN     MOD26, d3, d3
+	VN     MOD24, d4, d4     \ // [in0₂₆[4], in1₂₆[4]]
+	VN     MOD26, d2, d2     // [in0₂₆[2], in1₂₆[2]]
 
-// pack h4:h0 into h1:h0 (no carry)
+// func updateVX(state *macState, msg []byte)
-#define PACK(h0, h1, h2, h3, h4) \
+TEXT ·updateVX(SB), NOSPLIT, $0
-	VESLG $26, h1, h1  \
+	MOVD state+0(FP), R1
-	VESLG $26, h3, h3  \
+	LMG  msg+8(FP), R2, R3 // R2=msg_base, R3=msg_len
-	VO    h0, h1, h0   \
-	VO    h2, h3, h2   \
-	VESLG $4, h2, h2   \
-	VLEIB $7, $48, h1  \
-	VSLB  h1, h2, h2   \
-	VO    h0, h2, h0   \
-	VLEIB $7, $104, h1 \
-	VSLB  h1, h4, h3   \
-	VO    h3, h0, h0   \
-	VLEIB $7, $24, h1  \
-	VSRLB h1, h4, h1
 
-// if h > 2**130-5 then h -= 2**130-5
+	// load EX0, EX1 and EX2
-#define MOD(h0, h1, t0, t1, t2) \
-	VZERO t0          \
-	VLEIG $1, $5, t0  \
-	VACCQ h0, t0, t1  \
-	VAQ   h0, t0, t0  \
-	VONE  t2          \
-	VLEIG $1, $-4, t2 \
-	VAQ   t2, t1, t1  \
-	VACCQ h1, t1, t1  \
-	VONE  t2          \
-	VAQ   t2, t1, t1  \
-	VN    h0, t1, t2  \
-	VNC   t0, t1, t1  \
-	VO    t1, t2, h0
-
-// func poly1305vx(out *[16]byte, m *byte, mlen uint64, key *[32]key)
-TEXT ·poly1305vx(SB), $0-32
-	// This code processes up to 2 blocks (32 bytes) per iteration
-	// using the algorithm described in:
-	// NEON crypto, Daniel J. Bernstein & Peter Schwabe
-	// https://cryptojedi.org/papers/neoncrypto-20120320.pdf
-	LMG out+0(FP), R1, R4 // R1=out, R2=m, R3=mlen, R4=key
-
-	// load MOD26, EX0, EX1 and EX2
 	MOVD $·constants<>(SB), R5
-	VLM  (R5), MOD26, EX2
+	VLM  (R5), EX0, EX2
 
-	// setup r
+	// generate masks
-	VL   (R4), T_0
+	VGMG $(64-24), $63, MOD24 // [0x00ffffff, 0x00ffffff]
-	MOVD $·keyMask<>(SB), R6
+	VGMG $(64-26), $63, MOD26 // [0x03ffffff, 0x03ffffff]
-	VL   (R6), T_1
-	VN   T_0, T_1, T_0
-	EXPAND(T_0, T_0, R_0, R_1, R_2, R_3, R_4)
 
-	// setup r*5
+	// load h (accumulator) and r (key) from state
-	VLEIG $0, $5, T_0
+	VZERO T_1               // [0, 0]
-	VLEIG $1, $5, T_0
+	VL    0(R1), T_0        // [h₆₄[0], h₆₄[1]]
+	VLEG  $0, 16(R1), T_1   // [h₆₄[2], 0]
+	VL    24(R1), T_2       // [r₆₄[0], r₆₄[1]]
+	VPDI  $0, T_0, T_2, T_3 // [h₆₄[0], r₆₄[0]]
+	VPDI  $5, T_0, T_2, T_4 // [h₆₄[1], r₆₄[1]]
 
-	// store r (for final block)
+	// unpack h and r into 26-bit limbs
-	VMLOF T_0, R_1, R5SAVE_1
+	// note: h₆₄[2] may have the low 3 bits set, so h₂₆[4] is a 27-bit value
-	VMLOF T_0, R_2, R5SAVE_2
+	VN     MOD26, T_3, H_0            // [h₂₆[0], r₂₆[0]]
-	VMLOF T_0, R_3, R5SAVE_3
+	VZERO  H_1                        // [0, 0]
-	VMLOF T_0, R_4, R5SAVE_4
+	VZERO  H_3                        // [0, 0]
-	VLGVG $0, R_0, RSAVE_0
+	VGMG   $(64-12-14), $(63-12), T_0 // [0x03fff000, 0x03fff000] - 26-bit mask with low 12 bits masked out
-	VLGVG $0, R_1, RSAVE_1
+	VESLG  $24, T_1, T_1              // [h₆₄[2]<<24, 0]
-	VLGVG $0, R_2, RSAVE_2
+	VERIMG $-26&63, T_3, MOD26, H_1   // [h₂₆[1], r₂₆[1]]
-	VLGVG $0, R_3, RSAVE_3
+	VESRLG $+52&63, T_3, H_2          // [h₂₆[2], r₂₆[2]] - low 12 bits only
-	VLGVG $0, R_4, RSAVE_4
+	VERIMG $-14&63, T_4, MOD26, H_3   // [h₂₆[1], r₂₆[1]]
+	VESRLG $40, T_4, H_4              // [h₂₆[4], r₂₆[4]] - low 24 bits only
+	VERIMG $+12&63, T_4, T_0, H_2     // [h₂₆[2], r₂₆[2]] - complete
+	VO     T_1, H_4, H_4              // [h₂₆[4], r₂₆[4]] - complete
 
-	// skip r**2 calculation
+	// replicate r across all 4 vector elements
+	VREPF $3, H_0, R_0 // [r₂₆[0], r₂₆[0], r₂₆[0], r₂₆[0]]
+	VREPF $3, H_1, R_1 // [r₂₆[1], r₂₆[1], r₂₆[1], r₂₆[1]]
+	VREPF $3, H_2, R_2 // [r₂₆[2], r₂₆[2], r₂₆[2], r₂₆[2]]
+	VREPF $3, H_3, R_3 // [r₂₆[3], r₂₆[3], r₂₆[3], r₂₆[3]]
+	VREPF $3, H_4, R_4 // [r₂₆[4], r₂₆[4], r₂₆[4], r₂₆[4]]
+
+	// zero out lane 1 of h
+	VLEIG $1, $0, H_0 // [h₂₆[0], 0]
+	VLEIG $1, $0, H_1 // [h₂₆[1], 0]
+	VLEIG $1, $0, H_2 // [h₂₆[2], 0]
+	VLEIG $1, $0, H_3 // [h₂₆[3], 0]
+	VLEIG $1, $0, H_4 // [h₂₆[4], 0]
+
+	// calculate 5r (ignore least significant limb)
+	VREPIF $5, T_0
+	VMLF   T_0, R_1, R5_1 // [5r₂₆[1], 5r₂₆[1], 5r₂₆[1], 5r₂₆[1]]
+	VMLF   T_0, R_2, R5_2 // [5r₂₆[2], 5r₂₆[2], 5r₂₆[2], 5r₂₆[2]]
+	VMLF   T_0, R_3, R5_3 // [5r₂₆[3], 5r₂₆[3], 5r₂₆[3], 5r₂₆[3]]
+	VMLF   T_0, R_4, R5_4 // [5r₂₆[4], 5r₂₆[4], 5r₂₆[4], 5r₂₆[4]]
+
+	// skip r² calculation if we are only calculating one block
 	CMPBLE R3, $16, skip
 
-	// calculate r**2
+	// calculate r²
-	MULTIPLY(R_0, R_1, R_2, R_3, R_4, R_0, R_1, R_2, R_3, R_4, R5SAVE_1, R5SAVE_2, R5SAVE_3, R5SAVE_4, H_0, H_1, H_2, H_3, H_4)
+	MULTIPLY(R_0, R_1, R_2, R_3, R_4, R_0, R_1, R_2, R_3, R_4, R5_1, R5_2, R5_3, R5_4, M_0, M_1, M_2, M_3, M_4)
-	REDUCE(H_0, H_1, H_2, H_3, H_4)
+	REDUCE(M_0, M_1, M_2, M_3, M_4)
-	VLEIG $0, $5, T_0
+	VGBM   $0x0f0f, T_0
-	VLEIG $1, $5, T_0
+	VERIMG $0, M_0, T_0, R_0 // [r₂₆[0], r²₂₆[0], r₂₆[0], r²₂₆[0]]
-	VMLOF T_0, H_1, R5_1
+	VERIMG $0, M_1, T_0, R_1 // [r₂₆[1], r²₂₆[1], r₂₆[1], r²₂₆[1]]
-	VMLOF T_0, H_2, R5_2
+	VERIMG $0, M_2, T_0, R_2 // [r₂₆[2], r²₂₆[2], r₂₆[2], r²₂₆[2]]
-	VMLOF T_0, H_3, R5_3
+	VERIMG $0, M_3, T_0, R_3 // [r₂₆[3], r²₂₆[3], r₂₆[3], r²₂₆[3]]
-	VMLOF T_0, H_4, R5_4
+	VERIMG $0, M_4, T_0, R_4 // [r₂₆[4], r²₂₆[4], r₂₆[4], r²₂₆[4]]
-	VLR   H_0, R_0
-	VLR   H_1, R_1
-	VLR   H_2, R_2
-	VLR   H_3, R_3
-	VLR   H_4, R_4
 
-	// initialize h
+	// calculate 5r² (ignore least significant limb)
-	VZERO H_0
+	VREPIF $5, T_0
-	VZERO H_1
+	VMLF   T_0, R_1, R5_1 // [5r₂₆[1], 5r²₂₆[1], 5r₂₆[1], 5r²₂₆[1]]
-	VZERO H_2
+	VMLF   T_0, R_2, R5_2 // [5r₂₆[2], 5r²₂₆[2], 5r₂₆[2], 5r²₂₆[2]]
-	VZERO H_3
+	VMLF   T_0, R_3, R5_3 // [5r₂₆[3], 5r²₂₆[3], 5r₂₆[3], 5r²₂₆[3]]
-	VZERO H_4
+	VMLF   T_0, R_4, R5_4 // [5r₂₆[4], 5r²₂₆[4], 5r₂₆[4], 5r²₂₆[4]]
 
 loop:
-	CMPBLE R3, $32, b2
+	CMPBLE R3, $32, b2 // 2 or fewer blocks remaining, need to change key coefficients
+
+	// load next 2 blocks from message
 	VLM (R2), T_0, T_1
+
+	// update message slice
 	SUB  $32, R3
 	MOVD $32(R2), R2
-	EXPAND(T_0, T_1, F_0, F_1, F_2, F_3, F_4)
+
-	VLEIB  $4, $1, F_4
+	// unpack message blocks into 26-bit big-endian limbs
-	VLEIB  $12, $1, F_4
+	EXPAND(T_0, T_1, M_0, M_1, M_2, M_3, M_4)
+
+	// add 2¹²⁸ to each message block value
+	VLEIB $4, $1, M_4
+	VLEIB $12, $1, M_4
 
 multiply:
-	VAG    H_0, F_0, F_0
+	// accumulate the incoming message
-	VAG    H_1, F_1, F_1
+	VAG H_0, M_0, M_0
-	VAG    H_2, F_2, F_2
+	VAG H_3, M_3, M_3
-	VAG    H_3, F_3, F_3
+	VAG H_1, M_1, M_1
-	VAG    H_4, F_4, F_4
+	VAG H_4, M_4, M_4
-	MULTIPLY(F_0, F_1, F_2, F_3, F_4, R_0, R_1, R_2, R_3, R_4, R5_1, R5_2, R5_3, R5_4, H_0, H_1, H_2, H_3, H_4)
+	VAG H_2, M_2, M_2
+
+	// multiply the accumulator by the key coefficient
+	MULTIPLY(M_0, M_1, M_2, M_3, M_4, R_0, R_1, R_2, R_3, R_4, R5_1, R5_2, R5_3, R5_4, H_0, H_1, H_2, H_3, H_4)
+
+	// carry and partially reduce the partial products
 	REDUCE(H_0, H_1, H_2, H_3, H_4)
+
 	CMPBNE R3, $0, loop
 
 finish:
-	// sum vectors
+	// sum lane 0 and lane 1 and put the result in lane 1
 	VZERO  T_0
 	VSUMQG H_0, T_0, H_0
-	VSUMQG H_1, T_0, H_1
-	VSUMQG H_2, T_0, H_2
 	VSUMQG H_3, T_0, H_3
+	VSUMQG H_1, T_0, H_1
 	VSUMQG H_4, T_0, H_4
+	VSUMQG H_2, T_0, H_2
 
-	// h may be >= 2*(2**130-5) so we need to reduce it again
+	// reduce again after summation
+	// TODO(mundaym): there might be a more efficient way to do this
+	// now that we only have 1 active lane. For example, we could
+	// simultaneously pack the values as we reduce them.
 	REDUCE(H_0, H_1, H_2, H_3, H_4)
 
-	// carry h1->h4
+	// carry h[1] through to h[4] so that only h[4] can exceed 2²⁶ - 1
+	// TODO(mundaym): in testing this final carry was unnecessary.
+	// Needs a proof before it can be removed though.
 	VESRLG $26, H_1, T_1
 	VN     MOD26, H_1, H_1
 	VAQ    T_1, H_2, H_2
@@ -284,95 +367,137 @@ finish:
 	VN     MOD26, H_3, H_3
 	VAQ    T_3, H_4, H_4
 
-	// h is now < 2*(2**130-5)
+	// h is now < 2(2¹³⁰ - 5)
-	// pack h into h1 (hi) and h0 (lo)
+	// Pack each lane in h₂₆[0:4] into h₁₂₈[0:1].
-	PACK(H_0, H_1, H_2, H_3, H_4)
+	VESLG $26, H_1, H_1
-
+	VESLG $26, H_3, H_3
-	// if h > 2**130-5 then h -= 2**130-5
+	VO    H_0, H_1, H_0
-	MOD(H_0, H_1, T_0, T_1, T_2)
+	VO    H_2, H_3, H_2
-
+	VESLG $4, H_2, H_2
-	// h += s
+	VLEIB $7, $48, H_1
-	MOVD  $·bswapMask<>(SB), R5
+	VSLB  H_1, H_2, H_2
-	VL    (R5), T_1
+	VO    H_0, H_2, H_0
-	VL    16(R4), T_0
+	VLEIB $7, $104, H_1
-	VPERM T_0, T_0, T_1, T_0    // reverse bytes (to big)
+	VSLB  H_1, H_4, H_3
-	VAQ   T_0, H_0, H_0
+	VO    H_3, H_0, H_0
-	VPERM H_0, H_0, T_1, H_0    // reverse bytes (to little)
+	VLEIB $7, $24, H_1
-	VST   H_0, (R1)
+	VSRLB H_1, H_4, H_1
 
+	// update state
+	VSTEG $1, H_0, 0(R1)
+	VSTEG $0, H_0, 8(R1)
+	VSTEG $1, H_1, 16(R1)
 	RET
 
-b2:
+b2:  // 2 or fewer blocks remaining
 	CMPBLE R3, $16, b1
 
-	// 2 blocks remaining
+	// Load the 2 remaining blocks (17-32 bytes remaining).
-	SUB    $17, R3
+	MOVD $-17(R3), R0    // index of final byte to load modulo 16
-	VL     (R2), T_0
+	VL   (R2), T_0       // load full 16 byte block
-	VLL    R3, 16(R2), T_1
+	VLL  R0, 16(R2), T_1 // load final (possibly partial) block and pad with zeros to 16 bytes
-	ADD    $1, R3
-	MOVBZ  $1, R0
-	CMPBEQ R3, $16, 2(PC)
-	VLVGB  R3, R0, T_1
-	EXPAND(T_0, T_1, F_0, F_1, F_2, F_3, F_4)
-	CMPBNE R3, $16, 2(PC)
-	VLEIB  $12, $1, F_4
-	VLEIB  $4, $1, F_4
 
-	// setup [r²,r]
+	// The Poly1305 algorithm requires that a 1 bit be appended to
-	VLVGG $1, RSAVE_0, R_0
+	// each message block. If the final block is less than 16 bytes
-	VLVGG $1, RSAVE_1, R_1
+	// long then it is easiest to insert the 1 before the message
-	VLVGG $1, RSAVE_2, R_2
+	// block is split into 26-bit limbs. If, on the other hand, the
-	VLVGG $1, RSAVE_3, R_3
+	// final message block is 16 bytes long then we append the 1 bit
-	VLVGG $1, RSAVE_4, R_4
+	// after expansion as normal.
-	VPDI  $0, R5_1, R5SAVE_1, R5_1
+	MOVBZ  $1, R0
-	VPDI  $0, R5_2, R5SAVE_2, R5_2
+	MOVD   $-16(R3), R3   // index of byte in last block to insert 1 at (could be 16)
-	VPDI  $0, R5_3, R5SAVE_3, R5_3
+	CMPBEQ R3, $16, 2(PC) // skip the insertion if the final block is 16 bytes long
-	VPDI  $0, R5_4, R5SAVE_4, R5_4
+	VLVGB  R3, R0, T_1    // insert 1 into the byte at index R3
+
+	// Split both blocks into 26-bit limbs in the appropriate lanes.
+	EXPAND(T_0, T_1, M_0, M_1, M_2, M_3, M_4)
+
+	// Append a 1 byte to the end of the second to last block.
+	VLEIB $4, $1, M_4
+
+	// Append a 1 byte to the end of the last block only if it is a
+	// full 16 byte block.
+	CMPBNE R3, $16, 2(PC)
+	VLEIB  $12, $1, M_4
+
+	// Finally, set up the coefficients for the final multiplication.
+	// We have previously saved r and 5r in the 32-bit even indexes
+	// of the R_[0-4] and R5_[1-4] coefficient registers.
+	//
+	// We want lane 0 to be multiplied by r² so that can be kept the
+	// same. We want lane 1 to be multiplied by r so we need to move
+	// the saved r value into the 32-bit odd index in lane 1 by
+	// rotating the 64-bit lane by 32.
+	VGBM   $0x00ff, T_0         // [0, 0xffffffffffffffff] - mask lane 1 only
+	VERIMG $32, R_0, T_0, R_0   // [_,  r²₂₆[0], _,  r₂₆[0]]
+	VERIMG $32, R_1, T_0, R_1   // [_,  r²₂₆[1], _,  r₂₆[1]]
+	VERIMG $32, R_2, T_0, R_2   // [_,  r²₂₆[2], _,  r₂₆[2]]
+	VERIMG $32, R_3, T_0, R_3   // [_,  r²₂₆[3], _,  r₂₆[3]]
+	VERIMG $32, R_4, T_0, R_4   // [_,  r²₂₆[4], _,  r₂₆[4]]
+	VERIMG $32, R5_1, T_0, R5_1 // [_, 5r²₂₆[1], _, 5r₂₆[1]]
+	VERIMG $32, R5_2, T_0, R5_2 // [_, 5r²₂₆[2], _, 5r₂₆[2]]
+	VERIMG $32, R5_3, T_0, R5_3 // [_, 5r²₂₆[3], _, 5r₂₆[3]]
+	VERIMG $32, R5_4, T_0, R5_4 // [_, 5r²₂₆[4], _, 5r₂₆[4]]
 
 	MOVD $0, R3
 	BR   multiply
 
 skip:
-	VZERO H_0
-	VZERO H_1
-	VZERO H_2
-	VZERO H_3
-	VZERO H_4
-
 	CMPBEQ R3, $0, finish
 
-b1:
+b1:  // 1 block remaining
-	// 1 block remaining
+
-	SUB    $1, R3
+	// Load the final block (1-16 bytes). This will be placed into
-	VLL    R3, (R2), T_0
+	// lane 0.
-	ADD    $1, R3
+	MOVD $-1(R3), R0
+	VLL  R0, (R2), T_0 // pad to 16 bytes with zeros
+
+	// The Poly1305 algorithm requires that a 1 bit be appended to
+	// each message block. If the final block is less than 16 bytes
+	// long then it is easiest to insert the 1 before the message
+	// block is split into 26-bit limbs. If, on the other hand, the
+	// final message block is 16 bytes long then we append the 1 bit
+	// after expansion as normal.
 	MOVBZ  $1, R0
 	CMPBEQ R3, $16, 2(PC)
 	VLVGB  R3, R0, T_0
-	VZERO  T_1
-	EXPAND(T_0, T_1, F_0, F_1, F_2, F_3, F_4)
-	CMPBNE R3, $16, 2(PC)
-	VLEIB  $4, $1, F_4
-	VLEIG  $1, $1, R_0
-	VZERO  R_1
-	VZERO  R_2
-	VZERO  R_3
-	VZERO  R_4
-	VZERO  R5_1
-	VZERO  R5_2
-	VZERO  R5_3
-	VZERO  R5_4
 
-	// setup [r, 1]
+	// Set the message block in lane 1 to the value 0 so that it
-	VLVGG $0, RSAVE_0, R_0
+	// can be accumulated without affecting the final result.
-	VLVGG $0, RSAVE_1, R_1
+	VZERO T_1
-	VLVGG $0, RSAVE_2, R_2
+
-	VLVGG $0, RSAVE_3, R_3
+	// Split the final message block into 26-bit limbs in lane 0.
-	VLVGG $0, RSAVE_4, R_4
+	// Lane 1 will be contain 0.
-	VPDI  $0, R5SAVE_1, R5_1, R5_1
+	EXPAND(T_0, T_1, M_0, M_1, M_2, M_3, M_4)
-	VPDI  $0, R5SAVE_2, R5_2, R5_2
+
-	VPDI  $0, R5SAVE_3, R5_3, R5_3
+	// Append a 1 byte to the end of the last block only if it is a
-	VPDI  $0, R5SAVE_4, R5_4, R5_4
+	// full 16 byte block.
+	CMPBNE R3, $16, 2(PC)
+	VLEIB  $4, $1, M_4
+
+	// We have previously saved r and 5r in the 32-bit even indexes
+	// of the R_[0-4] and R5_[1-4] coefficient registers.
+	//
+	// We want lane 0 to be multiplied by r so we need to move the
+	// saved r value into the 32-bit odd index in lane 0. We want
+	// lane 1 to be set to the value 1. This makes multiplication
+	// a no-op. We do this by setting lane 1 in every register to 0
+	// and then just setting the 32-bit index 3 in R_0 to 1.
+	VZERO T_0
+	MOVD  $0, R0
+	MOVD  $0x10111213, R12
+	VLVGP R12, R0, T_1         // [_, 0x10111213, _, 0x00000000]
+	VPERM T_0, R_0, T_1, R_0   // [_,  r₂₆[0], _, 0]
+	VPERM T_0, R_1, T_1, R_1   // [_,  r₂₆[1], _, 0]
+	VPERM T_0, R_2, T_1, R_2   // [_,  r₂₆[2], _, 0]
+	VPERM T_0, R_3, T_1, R_3   // [_,  r₂₆[3], _, 0]
+	VPERM T_0, R_4, T_1, R_4   // [_,  r₂₆[4], _, 0]
+	VPERM T_0, R5_1, T_1, R5_1 // [_, 5r₂₆[1], _, 0]
+	VPERM T_0, R5_2, T_1, R5_2 // [_, 5r₂₆[2], _, 0]
+	VPERM T_0, R5_3, T_1, R5_3 // [_, 5r₂₆[3], _, 0]
+	VPERM T_0, R5_4, T_1, R5_4 // [_, 5r₂₆[4], _, 0]
+
+	// Set the value of lane 1 to be 1.
+	VLEIF $3, $1, R_0 // [_,  r₂₆[0], _, 1]
 
 	MOVD $0, R3
 	BR   multiply

vendor/golang.org/x/crypto/poly1305/sum_vmsl_s390x.s

@@ -1,909 +0,0 @@
-// 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.
-
-// +build s390x,go1.11,!gccgo,!appengine
-
-#include "textflag.h"
-
-// Implementation of Poly1305 using the vector facility (vx) and the VMSL instruction.
-
-// constants
-#define EX0   V1
-#define EX1   V2
-#define EX2   V3
-
-// temporaries
-#define T_0 V4
-#define T_1 V5
-#define T_2 V6
-#define T_3 V7
-#define T_4 V8
-#define T_5 V9
-#define T_6 V10
-#define T_7 V11
-#define T_8 V12
-#define T_9 V13
-#define T_10 V14
-
-// r**2 & r**4
-#define R_0  V15
-#define R_1  V16
-#define R_2  V17
-#define R5_1 V18
-#define R5_2 V19
-// key (r)
-#define RSAVE_0 R7
-#define RSAVE_1 R8
-#define RSAVE_2 R9
-#define R5SAVE_1 R10
-#define R5SAVE_2 R11
-
-// message block
-#define M0 V20
-#define M1 V21
-#define M2 V22
-#define M3 V23
-#define M4 V24
-#define M5 V25
-
-// accumulator
-#define H0_0 V26
-#define H1_0 V27
-#define H2_0 V28
-#define H0_1 V29
-#define H1_1 V30
-#define H2_1 V31
-
-GLOBL ·keyMask<>(SB), RODATA, $16
-DATA ·keyMask<>+0(SB)/8, $0xffffff0ffcffff0f
-DATA ·keyMask<>+8(SB)/8, $0xfcffff0ffcffff0f
-
-GLOBL ·bswapMask<>(SB), RODATA, $16
-DATA ·bswapMask<>+0(SB)/8, $0x0f0e0d0c0b0a0908
-DATA ·bswapMask<>+8(SB)/8, $0x0706050403020100
-
-GLOBL ·constants<>(SB), RODATA, $48
-// EX0
-DATA ·constants<>+0(SB)/8, $0x18191a1b1c1d1e1f
-DATA ·constants<>+8(SB)/8, $0x0000050403020100
-// EX1
-DATA ·constants<>+16(SB)/8, $0x18191a1b1c1d1e1f
-DATA ·constants<>+24(SB)/8, $0x00000a0908070605
-// EX2
-DATA ·constants<>+32(SB)/8, $0x18191a1b1c1d1e1f
-DATA ·constants<>+40(SB)/8, $0x0000000f0e0d0c0b
-
-GLOBL ·c<>(SB), RODATA, $48
-// EX0
-DATA ·c<>+0(SB)/8, $0x0000050403020100
-DATA ·c<>+8(SB)/8, $0x0000151413121110
-// EX1
-DATA ·c<>+16(SB)/8, $0x00000a0908070605
-DATA ·c<>+24(SB)/8, $0x00001a1918171615
-// EX2
-DATA ·c<>+32(SB)/8, $0x0000000f0e0d0c0b
-DATA ·c<>+40(SB)/8, $0x0000001f1e1d1c1b
-
-GLOBL ·reduce<>(SB), RODATA, $32
-// 44 bit
-DATA ·reduce<>+0(SB)/8, $0x0
-DATA ·reduce<>+8(SB)/8, $0xfffffffffff
-// 42 bit
-DATA ·reduce<>+16(SB)/8, $0x0
-DATA ·reduce<>+24(SB)/8, $0x3ffffffffff
-
-// h = (f*g) % (2**130-5) [partial reduction]
-// uses T_0...T_9 temporary registers
-// input: m02_0, m02_1, m02_2, m13_0, m13_1, m13_2, r_0, r_1, r_2, r5_1, r5_2, m4_0, m4_1, m4_2, m5_0, m5_1, m5_2
-// temp: t0, t1, t2, t3, t4, t5, t6, t7, t8, t9
-// output: m02_0, m02_1, m02_2, m13_0, m13_1, m13_2
-#define MULTIPLY(m02_0, m02_1, m02_2, m13_0, m13_1, m13_2, r_0, r_1, r_2, r5_1, r5_2, m4_0, m4_1, m4_2, m5_0, m5_1, m5_2, t0, t1, t2, t3, t4, t5, t6, t7, t8, t9) \
-	\ // Eliminate the dependency for the last 2 VMSLs
-	VMSLG m02_0, r_2, m4_2, m4_2                       \
-	VMSLG m13_0, r_2, m5_2, m5_2                       \ // 8 VMSLs pipelined
-	VMSLG m02_0, r_0, m4_0, m4_0                       \
-	VMSLG m02_1, r5_2, V0, T_0                         \
-	VMSLG m02_0, r_1, m4_1, m4_1                       \
-	VMSLG m02_1, r_0, V0, T_1                          \
-	VMSLG m02_1, r_1, V0, T_2                          \
-	VMSLG m02_2, r5_1, V0, T_3                         \
-	VMSLG m02_2, r5_2, V0, T_4                         \
-	VMSLG m13_0, r_0, m5_0, m5_0                       \
-	VMSLG m13_1, r5_2, V0, T_5                         \
-	VMSLG m13_0, r_1, m5_1, m5_1                       \
-	VMSLG m13_1, r_0, V0, T_6                          \
-	VMSLG m13_1, r_1, V0, T_7                          \
-	VMSLG m13_2, r5_1, V0, T_8                         \
-	VMSLG m13_2, r5_2, V0, T_9                         \
-	VMSLG m02_2, r_0, m4_2, m4_2                       \
-	VMSLG m13_2, r_0, m5_2, m5_2                       \
-	VAQ   m4_0, T_0, m02_0                             \
-	VAQ   m4_1, T_1, m02_1                             \
-	VAQ   m5_0, T_5, m13_0                             \
-	VAQ   m5_1, T_6, m13_1                             \
-	VAQ   m02_0, T_3, m02_0                            \
-	VAQ   m02_1, T_4, m02_1                            \
-	VAQ   m13_0, T_8, m13_0                            \
-	VAQ   m13_1, T_9, m13_1                            \
-	VAQ   m4_2, T_2, m02_2                             \
-	VAQ   m5_2, T_7, m13_2                             \
-
-// SQUARE uses three limbs of r and r_2*5 to output square of r
-// uses T_1, T_5 and T_7 temporary registers
-// input: r_0, r_1, r_2, r5_2
-// temp: TEMP0, TEMP1, TEMP2
-// output: p0, p1, p2
-#define SQUARE(r_0, r_1, r_2, r5_2, p0, p1, p2, TEMP0, TEMP1, TEMP2) \
-	VMSLG r_0, r_0, p0, p0     \
-	VMSLG r_1, r5_2, V0, TEMP0 \
-	VMSLG r_2, r5_2, p1, p1    \
-	VMSLG r_0, r_1, V0, TEMP1  \
-	VMSLG r_1, r_1, p2, p2     \
-	VMSLG r_0, r_2, V0, TEMP2  \
-	VAQ   TEMP0, p0, p0        \
-	VAQ   TEMP1, p1, p1        \
-	VAQ   TEMP2, p2, p2        \
-	VAQ   TEMP0, p0, p0        \
-	VAQ   TEMP1, p1, p1        \
-	VAQ   TEMP2, p2, p2        \
-
-// carry h0->h1->h2->h0 || h3->h4->h5->h3
-// uses T_2, T_4, T_5, T_7, T_8, T_9
-//       t6,  t7,  t8,  t9, t10, t11
-// input: h0, h1, h2, h3, h4, h5
-// temp: t0, t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11
-// output: h0, h1, h2, h3, h4, h5
-#define REDUCE(h0, h1, h2, h3, h4, h5, t0, t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11) \
-	VLM    (R12), t6, t7  \ // 44 and 42 bit clear mask
-	VLEIB  $7, $0x28, t10 \ // 5 byte shift mask
-	VREPIB $4, t8         \ // 4 bit shift mask
-	VREPIB $2, t11        \ // 2 bit shift mask
-	VSRLB  t10, h0, t0    \ // h0 byte shift
-	VSRLB  t10, h1, t1    \ // h1 byte shift
-	VSRLB  t10, h2, t2    \ // h2 byte shift
-	VSRLB  t10, h3, t3    \ // h3 byte shift
-	VSRLB  t10, h4, t4    \ // h4 byte shift
-	VSRLB  t10, h5, t5    \ // h5 byte shift
-	VSRL   t8, t0, t0     \ // h0 bit shift
-	VSRL   t8, t1, t1     \ // h2 bit shift
-	VSRL   t11, t2, t2    \ // h2 bit shift
-	VSRL   t8, t3, t3     \ // h3 bit shift
-	VSRL   t8, t4, t4     \ // h4 bit shift
-	VESLG  $2, t2, t9     \ // h2 carry x5
-	VSRL   t11, t5, t5    \ // h5 bit shift
-	VN     t6, h0, h0     \ // h0 clear carry
-	VAQ    t2, t9, t2     \ // h2 carry x5
-	VESLG  $2, t5, t9     \ // h5 carry x5
-	VN     t6, h1, h1     \ // h1 clear carry
-	VN     t7, h2, h2     \ // h2 clear carry
-	VAQ    t5, t9, t5     \ // h5 carry x5
-	VN     t6, h3, h3     \ // h3 clear carry
-	VN     t6, h4, h4     \ // h4 clear carry
-	VN     t7, h5, h5     \ // h5 clear carry
-	VAQ    t0, h1, h1     \ // h0->h1
-	VAQ    t3, h4, h4     \ // h3->h4
-	VAQ    t1, h2, h2     \ // h1->h2
-	VAQ    t4, h5, h5     \ // h4->h5
-	VAQ    t2, h0, h0     \ // h2->h0
-	VAQ    t5, h3, h3     \ // h5->h3
-	VREPG  $1, t6, t6     \ // 44 and 42 bit masks across both halves
-	VREPG  $1, t7, t7     \
-	VSLDB  $8, h0, h0, h0 \ // set up [h0/1/2, h3/4/5]
-	VSLDB  $8, h1, h1, h1 \
-	VSLDB  $8, h2, h2, h2 \
-	VO     h0, h3, h3     \
-	VO     h1, h4, h4     \
-	VO     h2, h5, h5     \
-	VESRLG $44, h3, t0    \ // 44 bit shift right
-	VESRLG $44, h4, t1    \
-	VESRLG $42, h5, t2    \
-	VN     t6, h3, h3     \ // clear carry bits
-	VN     t6, h4, h4     \
-	VN     t7, h5, h5     \
-	VESLG  $2, t2, t9     \ // multiply carry by 5
-	VAQ    t9, t2, t2     \
-	VAQ    t0, h4, h4     \
-	VAQ    t1, h5, h5     \
-	VAQ    t2, h3, h3     \
-
-// carry h0->h1->h2->h0
-// input: h0, h1, h2
-// temp: t0, t1, t2, t3, t4, t5, t6, t7, t8
-// output: h0, h1, h2
-#define REDUCE2(h0, h1, h2, t0, t1, t2, t3, t4, t5, t6, t7, t8) \
-	VLEIB  $7, $0x28, t3 \ // 5 byte shift mask
-	VREPIB $4, t4        \ // 4 bit shift mask
-	VREPIB $2, t7        \ // 2 bit shift mask
-	VGBM   $0x003F, t5   \ // mask to clear carry bits
-	VSRLB  t3, h0, t0    \
-	VSRLB  t3, h1, t1    \
-	VSRLB  t3, h2, t2    \
-	VESRLG $4, t5, t5    \ // 44 bit clear mask
-	VSRL   t4, t0, t0    \
-	VSRL   t4, t1, t1    \
-	VSRL   t7, t2, t2    \
-	VESRLG $2, t5, t6    \ // 42 bit clear mask
-	VESLG  $2, t2, t8    \
-	VAQ    t8, t2, t2    \
-	VN     t5, h0, h0    \
-	VN     t5, h1, h1    \
-	VN     t6, h2, h2    \
-	VAQ    t0, h1, h1    \
-	VAQ    t1, h2, h2    \
-	VAQ    t2, h0, h0    \
-	VSRLB  t3, h0, t0    \
-	VSRLB  t3, h1, t1    \
-	VSRLB  t3, h2, t2    \
-	VSRL   t4, t0, t0    \
-	VSRL   t4, t1, t1    \
-	VSRL   t7, t2, t2    \
-	VN     t5, h0, h0    \
-	VN     t5, h1, h1    \
-	VESLG  $2, t2, t8    \
-	VN     t6, h2, h2    \
-	VAQ    t0, h1, h1    \
-	VAQ    t8, t2, t2    \
-	VAQ    t1, h2, h2    \
-	VAQ    t2, h0, h0    \
-
-// expands two message blocks into the lower halfs of the d registers
-// moves the contents of the d registers into upper halfs
-// input: in1, in2, d0, d1, d2, d3, d4, d5
-// temp: TEMP0, TEMP1, TEMP2, TEMP3
-// output: d0, d1, d2, d3, d4, d5
-#define EXPACC(in1, in2, d0, d1, d2, d3, d4, d5, TEMP0, TEMP1, TEMP2, TEMP3) \
-	VGBM   $0xff3f, TEMP0      \
-	VGBM   $0xff1f, TEMP1      \
-	VESLG  $4, d1, TEMP2       \
-	VESLG  $4, d4, TEMP3       \
-	VESRLG $4, TEMP0, TEMP0    \
-	VPERM  in1, d0, EX0, d0    \
-	VPERM  in2, d3, EX0, d3    \
-	VPERM  in1, d2, EX2, d2    \
-	VPERM  in2, d5, EX2, d5    \
-	VPERM  in1, TEMP2, EX1, d1 \
-	VPERM  in2, TEMP3, EX1, d4 \
-	VN     TEMP0, d0, d0       \
-	VN     TEMP0, d3, d3       \
-	VESRLG $4, d1, d1          \
-	VESRLG $4, d4, d4          \
-	VN     TEMP1, d2, d2       \
-	VN     TEMP1, d5, d5       \
-	VN     TEMP0, d1, d1       \
-	VN     TEMP0, d4, d4       \
-
-// expands one message block into the lower halfs of the d registers
-// moves the contents of the d registers into upper halfs
-// input: in, d0, d1, d2
-// temp: TEMP0, TEMP1, TEMP2
-// output: d0, d1, d2
-#define EXPACC2(in, d0, d1, d2, TEMP0, TEMP1, TEMP2) \
-	VGBM   $0xff3f, TEMP0     \
-	VESLG  $4, d1, TEMP2      \
-	VGBM   $0xff1f, TEMP1     \
-	VPERM  in, d0, EX0, d0    \
-	VESRLG $4, TEMP0, TEMP0   \
-	VPERM  in, d2, EX2, d2    \
-	VPERM  in, TEMP2, EX1, d1 \
-	VN     TEMP0, d0, d0      \
-	VN     TEMP1, d2, d2      \
-	VESRLG $4, d1, d1         \
-	VN     TEMP0, d1, d1      \
-
-// pack h2:h0 into h1:h0 (no carry)
-// input: h0, h1, h2
-// output: h0, h1, h2
-#define PACK(h0, h1, h2) \
-	VMRLG  h1, h2, h2  \ // copy h1 to upper half h2
-	VESLG  $44, h1, h1 \ // shift limb 1 44 bits, leaving 20
-	VO     h0, h1, h0  \ // combine h0 with 20 bits from limb 1
-	VESRLG $20, h2, h1 \ // put top 24 bits of limb 1 into h1
-	VLEIG  $1, $0, h1  \ // clear h2 stuff from lower half of h1
-	VO     h0, h1, h0  \ // h0 now has 88 bits (limb 0 and 1)
-	VLEIG  $0, $0, h2  \ // clear upper half of h2
-	VESRLG $40, h2, h1 \ // h1 now has upper two bits of result
-	VLEIB  $7, $88, h1 \ // for byte shift (11 bytes)
-	VSLB   h1, h2, h2  \ // shift h2 11 bytes to the left
-	VO     h0, h2, h0  \ // combine h0 with 20 bits from limb 1
-	VLEIG  $0, $0, h1  \ // clear upper half of h1
-
-// if h > 2**130-5 then h -= 2**130-5
-// input: h0, h1
-// temp: t0, t1, t2
-// output: h0
-#define MOD(h0, h1, t0, t1, t2) \
-	VZERO t0          \
-	VLEIG $1, $5, t0  \
-	VACCQ h0, t0, t1  \
-	VAQ   h0, t0, t0  \
-	VONE  t2          \
-	VLEIG $1, $-4, t2 \
-	VAQ   t2, t1, t1  \
-	VACCQ h1, t1, t1  \
-	VONE  t2          \
-	VAQ   t2, t1, t1  \
-	VN    h0, t1, t2  \
-	VNC   t0, t1, t1  \
-	VO    t1, t2, h0  \
-
-// func poly1305vmsl(out *[16]byte, m *byte, mlen uint64, key *[32]key)
-TEXT ·poly1305vmsl(SB), $0-32
-	// This code processes 6 + up to 4 blocks (32 bytes) per iteration
-	// using the algorithm described in:
-	// NEON crypto, Daniel J. Bernstein & Peter Schwabe
-	// https://cryptojedi.org/papers/neoncrypto-20120320.pdf
-	// And as moddified for VMSL as described in
-	// Accelerating Poly1305 Cryptographic Message Authentication on the z14
-	// O'Farrell et al, CASCON 2017, p48-55
-	// https://ibm.ent.box.com/s/jf9gedj0e9d2vjctfyh186shaztavnht
-
-	LMG   out+0(FP), R1, R4 // R1=out, R2=m, R3=mlen, R4=key
-	VZERO V0                // c
-
-	// load EX0, EX1 and EX2
-	MOVD $·constants<>(SB), R5
-	VLM  (R5), EX0, EX2        // c
-
-	// setup r
-	VL    (R4), T_0
-	MOVD  $·keyMask<>(SB), R6
-	VL    (R6), T_1
-	VN    T_0, T_1, T_0
-	VZERO T_2                 // limbs for r
-	VZERO T_3
-	VZERO T_4
-	EXPACC2(T_0, T_2, T_3, T_4, T_1, T_5, T_7)
-
-	// T_2, T_3, T_4: [0, r]
-
-	// setup r*20
-	VLEIG $0, $0, T_0
-	VLEIG $1, $20, T_0       // T_0: [0, 20]
-	VZERO T_5
-	VZERO T_6
-	VMSLG T_0, T_3, T_5, T_5
-	VMSLG T_0, T_4, T_6, T_6
-
-	// store r for final block in GR
-	VLGVG $1, T_2, RSAVE_0  // c
-	VLGVG $1, T_3, RSAVE_1  // c
-	VLGVG $1, T_4, RSAVE_2  // c
-	VLGVG $1, T_5, R5SAVE_1 // c
-	VLGVG $1, T_6, R5SAVE_2 // c
-
-	// initialize h
-	VZERO H0_0
-	VZERO H1_0
-	VZERO H2_0
-	VZERO H0_1
-	VZERO H1_1
-	VZERO H2_1
-
-	// initialize pointer for reduce constants
-	MOVD $·reduce<>(SB), R12
-
-	// calculate r**2 and 20*(r**2)
-	VZERO R_0
-	VZERO R_1
-	VZERO R_2
-	SQUARE(T_2, T_3, T_4, T_6, R_0, R_1, R_2, T_1, T_5, T_7)
-	REDUCE2(R_0, R_1, R_2, M0, M1, M2, M3, M4, R5_1, R5_2, M5, T_1)
-	VZERO R5_1
-	VZERO R5_2
-	VMSLG T_0, R_1, R5_1, R5_1
-	VMSLG T_0, R_2, R5_2, R5_2
-
-	// skip r**4 calculation if 3 blocks or less
-	CMPBLE R3, $48, b4
-
-	// calculate r**4 and 20*(r**4)
-	VZERO T_8
-	VZERO T_9
-	VZERO T_10
-	SQUARE(R_0, R_1, R_2, R5_2, T_8, T_9, T_10, T_1, T_5, T_7)
-	REDUCE2(T_8, T_9, T_10, M0, M1, M2, M3, M4, T_2, T_3, M5, T_1)
-	VZERO T_2
-	VZERO T_3
-	VMSLG T_0, T_9, T_2, T_2
-	VMSLG T_0, T_10, T_3, T_3
-
-	// put r**2 to the right and r**4 to the left of R_0, R_1, R_2
-	VSLDB $8, T_8, T_8, T_8
-	VSLDB $8, T_9, T_9, T_9
-	VSLDB $8, T_10, T_10, T_10
-	VSLDB $8, T_2, T_2, T_2
-	VSLDB $8, T_3, T_3, T_3
-
-	VO T_8, R_0, R_0
-	VO T_9, R_1, R_1
-	VO T_10, R_2, R_2
-	VO T_2, R5_1, R5_1
-	VO T_3, R5_2, R5_2
-
-	CMPBLE R3, $80, load // less than or equal to 5 blocks in message
-
-	// 6(or 5+1) blocks
-	SUB    $81, R3
-	VLM    (R2), M0, M4
-	VLL    R3, 80(R2), M5
-	ADD    $1, R3
-	MOVBZ  $1, R0
-	CMPBGE R3, $16, 2(PC)
-	VLVGB  R3, R0, M5
-	MOVD   $96(R2), R2
-	EXPACC(M0, M1, H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, T_0, T_1, T_2, T_3)
-	EXPACC(M2, M3, H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, T_0, T_1, T_2, T_3)
-	VLEIB  $2, $1, H2_0
-	VLEIB  $2, $1, H2_1
-	VLEIB  $10, $1, H2_0
-	VLEIB  $10, $1, H2_1
-
-	VZERO  M0
-	VZERO  M1
-	VZERO  M2
-	VZERO  M3
-	VZERO  T_4
-	VZERO  T_10
-	EXPACC(M4, M5, M0, M1, M2, M3, T_4, T_10, T_0, T_1, T_2, T_3)
-	VLR    T_4, M4
-	VLEIB  $10, $1, M2
-	CMPBLT R3, $16, 2(PC)
-	VLEIB  $10, $1, T_10
-	MULTIPLY(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, R_0, R_1, R_2, R5_1, R5_2, M0, M1, M2, M3, M4, T_10, T_0, T_1, T_2, T_3, T_4, T_5, T_6, T_7, T_8, T_9)
-	REDUCE(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, T_10, M0, M1, M2, M3, M4, T_4, T_5, T_2, T_7, T_8, T_9)
-	VMRHG  V0, H0_1, H0_0
-	VMRHG  V0, H1_1, H1_0
-	VMRHG  V0, H2_1, H2_0
-	VMRLG  V0, H0_1, H0_1
-	VMRLG  V0, H1_1, H1_1
-	VMRLG  V0, H2_1, H2_1
-
-	SUB    $16, R3
-	CMPBLE R3, $0, square
-
-load:
-	// load EX0, EX1 and EX2
-	MOVD $·c<>(SB), R5
-	VLM  (R5), EX0, EX2
-
-loop:
-	CMPBLE R3, $64, add // b4	// last 4 or less blocks left
-
-	// next 4 full blocks
-	VLM  (R2), M2, M5
-	SUB  $64, R3
-	MOVD $64(R2), R2
-	REDUCE(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, T_10, M0, M1, T_0, T_1, T_3, T_4, T_5, T_2, T_7, T_8, T_9)
-
-	// expacc in-lined to create [m2, m3] limbs
-	VGBM   $0x3f3f, T_0     // 44 bit clear mask
-	VGBM   $0x1f1f, T_1     // 40 bit clear mask
-	VPERM  M2, M3, EX0, T_3
-	VESRLG $4, T_0, T_0     // 44 bit clear mask ready
-	VPERM  M2, M3, EX1, T_4
-	VPERM  M2, M3, EX2, T_5
-	VN     T_0, T_3, T_3
-	VESRLG $4, T_4, T_4
-	VN     T_1, T_5, T_5
-	VN     T_0, T_4, T_4
-	VMRHG  H0_1, T_3, H0_0
-	VMRHG  H1_1, T_4, H1_0
-	VMRHG  H2_1, T_5, H2_0
-	VMRLG  H0_1, T_3, H0_1
-	VMRLG  H1_1, T_4, H1_1
-	VMRLG  H2_1, T_5, H2_1
-	VLEIB  $10, $1, H2_0
-	VLEIB  $10, $1, H2_1
-	VPERM  M4, M5, EX0, T_3
-	VPERM  M4, M5, EX1, T_4
-	VPERM  M4, M5, EX2, T_5
-	VN     T_0, T_3, T_3
-	VESRLG $4, T_4, T_4
-	VN     T_1, T_5, T_5
-	VN     T_0, T_4, T_4
-	VMRHG  V0, T_3, M0
-	VMRHG  V0, T_4, M1
-	VMRHG  V0, T_5, M2
-	VMRLG  V0, T_3, M3
-	VMRLG  V0, T_4, M4
-	VMRLG  V0, T_5, M5
-	VLEIB  $10, $1, M2
-	VLEIB  $10, $1, M5
-
-	MULTIPLY(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, R_0, R_1, R_2, R5_1, R5_2, M0, M1, M2, M3, M4, M5, T_0, T_1, T_2, T_3, T_4, T_5, T_6, T_7, T_8, T_9)
-	CMPBNE R3, $0, loop
-	REDUCE(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, T_10, M0, M1, M3, M4, M5, T_4, T_5, T_2, T_7, T_8, T_9)
-	VMRHG  V0, H0_1, H0_0
-	VMRHG  V0, H1_1, H1_0
-	VMRHG  V0, H2_1, H2_0
-	VMRLG  V0, H0_1, H0_1
-	VMRLG  V0, H1_1, H1_1
-	VMRLG  V0, H2_1, H2_1
-
-	// load EX0, EX1, EX2
-	MOVD $·constants<>(SB), R5
-	VLM  (R5), EX0, EX2
-
-	// sum vectors
-	VAQ H0_0, H0_1, H0_0
-	VAQ H1_0, H1_1, H1_0
-	VAQ H2_0, H2_1, H2_0
-
-	// h may be >= 2*(2**130-5) so we need to reduce it again
-	// M0...M4 are used as temps here
-	REDUCE2(H0_0, H1_0, H2_0, M0, M1, M2, M3, M4, T_9, T_10, H0_1, M5)
-
-next:  // carry h1->h2
-	VLEIB  $7, $0x28, T_1
-	VREPIB $4, T_2
-	VGBM   $0x003F, T_3
-	VESRLG $4, T_3
-
-	// byte shift
-	VSRLB T_1, H1_0, T_4
-
-	// bit shift
-	VSRL T_2, T_4, T_4
-
-	// clear h1 carry bits
-	VN T_3, H1_0, H1_0
-
-	// add carry
-	VAQ T_4, H2_0, H2_0
-
-	// h is now < 2*(2**130-5)
-	// pack h into h1 (hi) and h0 (lo)
-	PACK(H0_0, H1_0, H2_0)
-
-	// if h > 2**130-5 then h -= 2**130-5
-	MOD(H0_0, H1_0, T_0, T_1, T_2)
-
-	// h += s
-	MOVD  $·bswapMask<>(SB), R5
-	VL    (R5), T_1
-	VL    16(R4), T_0
-	VPERM T_0, T_0, T_1, T_0    // reverse bytes (to big)
-	VAQ   T_0, H0_0, H0_0
-	VPERM H0_0, H0_0, T_1, H0_0 // reverse bytes (to little)
-	VST   H0_0, (R1)
-	RET
-
-add:
-	// load EX0, EX1, EX2
-	MOVD $·constants<>(SB), R5
-	VLM  (R5), EX0, EX2
-
-	REDUCE(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, T_10, M0, M1, M3, M4, M5, T_4, T_5, T_2, T_7, T_8, T_9)
-	VMRHG  V0, H0_1, H0_0
-	VMRHG  V0, H1_1, H1_0
-	VMRHG  V0, H2_1, H2_0
-	VMRLG  V0, H0_1, H0_1
-	VMRLG  V0, H1_1, H1_1
-	VMRLG  V0, H2_1, H2_1
-	CMPBLE R3, $64, b4
-
-b4:
-	CMPBLE R3, $48, b3 // 3 blocks or less
-
-	// 4(3+1) blocks remaining
-	SUB    $49, R3
-	VLM    (R2), M0, M2
-	VLL    R3, 48(R2), M3
-	ADD    $1, R3
-	MOVBZ  $1, R0
-	CMPBEQ R3, $16, 2(PC)
-	VLVGB  R3, R0, M3
-	MOVD   $64(R2), R2
-	EXPACC(M0, M1, H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, T_0, T_1, T_2, T_3)
-	VLEIB  $10, $1, H2_0
-	VLEIB  $10, $1, H2_1
-	VZERO  M0
-	VZERO  M1
-	VZERO  M4
-	VZERO  M5
-	VZERO  T_4
-	VZERO  T_10
-	EXPACC(M2, M3, M0, M1, M4, M5, T_4, T_10, T_0, T_1, T_2, T_3)
-	VLR    T_4, M2
-	VLEIB  $10, $1, M4
-	CMPBNE R3, $16, 2(PC)
-	VLEIB  $10, $1, T_10
-	MULTIPLY(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, R_0, R_1, R_2, R5_1, R5_2, M0, M1, M4, M5, M2, T_10, T_0, T_1, T_2, T_3, T_4, T_5, T_6, T_7, T_8, T_9)
-	REDUCE(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, T_10, M0, M1, M3, M4, M5, T_4, T_5, T_2, T_7, T_8, T_9)
-	VMRHG  V0, H0_1, H0_0
-	VMRHG  V0, H1_1, H1_0
-	VMRHG  V0, H2_1, H2_0
-	VMRLG  V0, H0_1, H0_1
-	VMRLG  V0, H1_1, H1_1
-	VMRLG  V0, H2_1, H2_1
-	SUB    $16, R3
-	CMPBLE R3, $0, square // this condition must always hold true!
-
-b3:
-	CMPBLE R3, $32, b2
-
-	// 3 blocks remaining
-
-	// setup [r²,r]
-	VSLDB $8, R_0, R_0, R_0
-	VSLDB $8, R_1, R_1, R_1
-	VSLDB $8, R_2, R_2, R_2
-	VSLDB $8, R5_1, R5_1, R5_1
-	VSLDB $8, R5_2, R5_2, R5_2
-
-	VLVGG $1, RSAVE_0, R_0
-	VLVGG $1, RSAVE_1, R_1
-	VLVGG $1, RSAVE_2, R_2
-	VLVGG $1, R5SAVE_1, R5_1
-	VLVGG $1, R5SAVE_2, R5_2
-
-	// setup [h0, h1]
-	VSLDB $8, H0_0, H0_0, H0_0
-	VSLDB $8, H1_0, H1_0, H1_0
-	VSLDB $8, H2_0, H2_0, H2_0
-	VO    H0_1, H0_0, H0_0
-	VO    H1_1, H1_0, H1_0
-	VO    H2_1, H2_0, H2_0
-	VZERO H0_1
-	VZERO H1_1
-	VZERO H2_1
-
-	VZERO M0
-	VZERO M1
-	VZERO M2
-	VZERO M3
-	VZERO M4
-	VZERO M5
-
-	// H*[r**2, r]
-	MULTIPLY(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, R_0, R_1, R_2, R5_1, R5_2, M0, M1, M2, M3, M4, M5, T_0, T_1, T_2, T_3, T_4, T_5, T_6, T_7, T_8, T_9)
-	REDUCE2(H0_0, H1_0, H2_0, M0, M1, M2, M3, M4, H0_1, H1_1, T_10, M5)
-
-	SUB    $33, R3
-	VLM    (R2), M0, M1
-	VLL    R3, 32(R2), M2
-	ADD    $1, R3
-	MOVBZ  $1, R0
-	CMPBEQ R3, $16, 2(PC)
-	VLVGB  R3, R0, M2
-
-	// H += m0
-	VZERO T_1
-	VZERO T_2
-	VZERO T_3
-	EXPACC2(M0, T_1, T_2, T_3, T_4, T_5, T_6)
-	VLEIB $10, $1, T_3
-	VAG   H0_0, T_1, H0_0
-	VAG   H1_0, T_2, H1_0
-	VAG   H2_0, T_3, H2_0
-
-	VZERO M0
-	VZERO M3
-	VZERO M4
-	VZERO M5
-	VZERO T_10
-
-	// (H+m0)*r
-	MULTIPLY(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, R_0, R_1, R_2, R5_1, R5_2, M0, M3, M4, M5, V0, T_10, T_0, T_1, T_2, T_3, T_4, T_5, T_6, T_7, T_8, T_9)
-	REDUCE2(H0_0, H1_0, H2_0, M0, M3, M4, M5, T_10, H0_1, H1_1, H2_1, T_9)
-
-	// H += m1
-	VZERO V0
-	VZERO T_1
-	VZERO T_2
-	VZERO T_3
-	EXPACC2(M1, T_1, T_2, T_3, T_4, T_5, T_6)
-	VLEIB $10, $1, T_3
-	VAQ   H0_0, T_1, H0_0
-	VAQ   H1_0, T_2, H1_0
-	VAQ   H2_0, T_3, H2_0
-	REDUCE2(H0_0, H1_0, H2_0, M0, M3, M4, M5, T_9, H0_1, H1_1, H2_1, T_10)
-
-	// [H, m2] * [r**2, r]
-	EXPACC2(M2, H0_0, H1_0, H2_0, T_1, T_2, T_3)
-	CMPBNE R3, $16, 2(PC)
-	VLEIB  $10, $1, H2_0
-	VZERO  M0
-	VZERO  M1
-	VZERO  M2
-	VZERO  M3
-	VZERO  M4
-	VZERO  M5
-	MULTIPLY(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, R_0, R_1, R_2, R5_1, R5_2, M0, M1, M2, M3, M4, M5, T_0, T_1, T_2, T_3, T_4, T_5, T_6, T_7, T_8, T_9)
-	REDUCE2(H0_0, H1_0, H2_0, M0, M1, M2, M3, M4, H0_1, H1_1, M5, T_10)
-	SUB    $16, R3
-	CMPBLE R3, $0, next   // this condition must always hold true!
-
-b2:
-	CMPBLE R3, $16, b1
-
-	// 2 blocks remaining
-
-	// setup [r²,r]
-	VSLDB $8, R_0, R_0, R_0
-	VSLDB $8, R_1, R_1, R_1
-	VSLDB $8, R_2, R_2, R_2
-	VSLDB $8, R5_1, R5_1, R5_1
-	VSLDB $8, R5_2, R5_2, R5_2
-
-	VLVGG $1, RSAVE_0, R_0
-	VLVGG $1, RSAVE_1, R_1
-	VLVGG $1, RSAVE_2, R_2
-	VLVGG $1, R5SAVE_1, R5_1
-	VLVGG $1, R5SAVE_2, R5_2
-
-	// setup [h0, h1]
-	VSLDB $8, H0_0, H0_0, H0_0
-	VSLDB $8, H1_0, H1_0, H1_0
-	VSLDB $8, H2_0, H2_0, H2_0
-	VO    H0_1, H0_0, H0_0
-	VO    H1_1, H1_0, H1_0
-	VO    H2_1, H2_0, H2_0
-	VZERO H0_1
-	VZERO H1_1
-	VZERO H2_1
-
-	VZERO M0
-	VZERO M1
-	VZERO M2
-	VZERO M3
-	VZERO M4
-	VZERO M5
-
-	// H*[r**2, r]
-	MULTIPLY(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, R_0, R_1, R_2, R5_1, R5_2, M0, M1, M2, M3, M4, M5, T_0, T_1, T_2, T_3, T_4, T_5, T_6, T_7, T_8, T_9)
-	REDUCE(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, T_10, M0, M1, M2, M3, M4, T_4, T_5, T_2, T_7, T_8, T_9)
-	VMRHG V0, H0_1, H0_0
-	VMRHG V0, H1_1, H1_0
-	VMRHG V0, H2_1, H2_0
-	VMRLG V0, H0_1, H0_1
-	VMRLG V0, H1_1, H1_1
-	VMRLG V0, H2_1, H2_1
-
-	// move h to the left and 0s at the right
-	VSLDB $8, H0_0, H0_0, H0_0
-	VSLDB $8, H1_0, H1_0, H1_0
-	VSLDB $8, H2_0, H2_0, H2_0
-
-	// get message blocks and append 1 to start
-	SUB    $17, R3
-	VL     (R2), M0
-	VLL    R3, 16(R2), M1
-	ADD    $1, R3
-	MOVBZ  $1, R0
-	CMPBEQ R3, $16, 2(PC)
-	VLVGB  R3, R0, M1
-	VZERO  T_6
-	VZERO  T_7
-	VZERO  T_8
-	EXPACC2(M0, T_6, T_7, T_8, T_1, T_2, T_3)
-	EXPACC2(M1, T_6, T_7, T_8, T_1, T_2, T_3)
-	VLEIB  $2, $1, T_8
-	CMPBNE R3, $16, 2(PC)
-	VLEIB  $10, $1, T_8
-
-	// add [m0, m1] to h
-	VAG H0_0, T_6, H0_0
-	VAG H1_0, T_7, H1_0
-	VAG H2_0, T_8, H2_0
-
-	VZERO M2
-	VZERO M3
-	VZERO M4
-	VZERO M5
-	VZERO T_10
-	VZERO M0
-
-	// at this point R_0 .. R5_2 look like [r**2, r]
-	MULTIPLY(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, R_0, R_1, R_2, R5_1, R5_2, M2, M3, M4, M5, T_10, M0, T_0, T_1, T_2, T_3, T_4, T_5, T_6, T_7, T_8, T_9)
-	REDUCE2(H0_0, H1_0, H2_0, M2, M3, M4, M5, T_9, H0_1, H1_1, H2_1, T_10)
-	SUB    $16, R3, R3
-	CMPBLE R3, $0, next
-
-b1:
-	CMPBLE R3, $0, next
-
-	// 1 block remaining
-
-	// setup [r²,r]
-	VSLDB $8, R_0, R_0, R_0
-	VSLDB $8, R_1, R_1, R_1
-	VSLDB $8, R_2, R_2, R_2
-	VSLDB $8, R5_1, R5_1, R5_1
-	VSLDB $8, R5_2, R5_2, R5_2
-
-	VLVGG $1, RSAVE_0, R_0
-	VLVGG $1, RSAVE_1, R_1
-	VLVGG $1, RSAVE_2, R_2
-	VLVGG $1, R5SAVE_1, R5_1
-	VLVGG $1, R5SAVE_2, R5_2
-
-	// setup [h0, h1]
-	VSLDB $8, H0_0, H0_0, H0_0
-	VSLDB $8, H1_0, H1_0, H1_0
-	VSLDB $8, H2_0, H2_0, H2_0
-	VO    H0_1, H0_0, H0_0
-	VO    H1_1, H1_0, H1_0
-	VO    H2_1, H2_0, H2_0
-	VZERO H0_1
-	VZERO H1_1
-	VZERO H2_1
-
-	VZERO M0
-	VZERO M1
-	VZERO M2
-	VZERO M3
-	VZERO M4
-	VZERO M5
-
-	// H*[r**2, r]
-	MULTIPLY(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, R_0, R_1, R_2, R5_1, R5_2, M0, M1, M2, M3, M4, M5, T_0, T_1, T_2, T_3, T_4, T_5, T_6, T_7, T_8, T_9)
-	REDUCE2(H0_0, H1_0, H2_0, M0, M1, M2, M3, M4, T_9, T_10, H0_1, M5)
-
-	// set up [0, m0] limbs
-	SUB    $1, R3
-	VLL    R3, (R2), M0
-	ADD    $1, R3
-	MOVBZ  $1, R0
-	CMPBEQ R3, $16, 2(PC)
-	VLVGB  R3, R0, M0
-	VZERO  T_1
-	VZERO  T_2
-	VZERO  T_3
-	EXPACC2(M0, T_1, T_2, T_3, T_4, T_5, T_6)// limbs: [0, m]
-	CMPBNE R3, $16, 2(PC)
-	VLEIB  $10, $1, T_3
-
-	// h+m0
-	VAQ H0_0, T_1, H0_0
-	VAQ H1_0, T_2, H1_0
-	VAQ H2_0, T_3, H2_0
-
-	VZERO M0
-	VZERO M1
-	VZERO M2
-	VZERO M3
-	VZERO M4
-	VZERO M5
-	MULTIPLY(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, R_0, R_1, R_2, R5_1, R5_2, M0, M1, M2, M3, M4, M5, T_0, T_1, T_2, T_3, T_4, T_5, T_6, T_7, T_8, T_9)
-	REDUCE2(H0_0, H1_0, H2_0, M0, M1, M2, M3, M4, T_9, T_10, H0_1, M5)
-
-	BR next
-
-square:
-	// setup [r²,r]
-	VSLDB $8, R_0, R_0, R_0
-	VSLDB $8, R_1, R_1, R_1
-	VSLDB $8, R_2, R_2, R_2
-	VSLDB $8, R5_1, R5_1, R5_1
-	VSLDB $8, R5_2, R5_2, R5_2
-
-	VLVGG $1, RSAVE_0, R_0
-	VLVGG $1, RSAVE_1, R_1
-	VLVGG $1, RSAVE_2, R_2
-	VLVGG $1, R5SAVE_1, R5_1
-	VLVGG $1, R5SAVE_2, R5_2
-
-	// setup [h0, h1]
-	VSLDB $8, H0_0, H0_0, H0_0
-	VSLDB $8, H1_0, H1_0, H1_0
-	VSLDB $8, H2_0, H2_0, H2_0
-	VO    H0_1, H0_0, H0_0
-	VO    H1_1, H1_0, H1_0
-	VO    H2_1, H2_0, H2_0
-	VZERO H0_1
-	VZERO H1_1
-	VZERO H2_1
-
-	VZERO M0
-	VZERO M1
-	VZERO M2
-	VZERO M3
-	VZERO M4
-	VZERO M5
-
-	// (h0*r**2) + (h1*r)
-	MULTIPLY(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, R_0, R_1, R_2, R5_1, R5_2, M0, M1, M2, M3, M4, M5, T_0, T_1, T_2, T_3, T_4, T_5, T_6, T_7, T_8, T_9)
-	REDUCE2(H0_0, H1_0, H2_0, M0, M1, M2, M3, M4, T_9, T_10, H0_1, M5)
-	BR next

vendor/golang.org/x/crypto/ssh/certs.go

@@ -22,7 +22,9 @@ const (
 	CertAlgoECDSA256v01   = "ecdsa-sha2-nistp256-cert-v01@openssh.com"
 	CertAlgoECDSA384v01   = "ecdsa-sha2-nistp384-cert-v01@openssh.com"
 	CertAlgoECDSA521v01   = "ecdsa-sha2-nistp521-cert-v01@openssh.com"
+	CertAlgoSKECDSA256v01 = "sk-ecdsa-sha2-nistp256-cert-v01@openssh.com"
 	CertAlgoED25519v01    = "ssh-ed25519-cert-v01@openssh.com"
+	CertAlgoSKED25519v01  = "sk-ssh-ed25519-cert-v01@openssh.com"
 )
 
 // Certificate types distinguish between host and user
@@ -37,6 +39,7 @@ const (
 type Signature struct {
 	Format string
 	Blob   []byte
+	Rest   []byte `ssh:"rest"`
 }
 
 // CertTimeInfinity can be used for OpenSSHCertV01.ValidBefore to indicate that
@@ -411,8 +414,8 @@ func (c *CertChecker) CheckCert(principal string, cert *Certificate) error {
 	return nil
 }
 
-// SignCert sets c.SignatureKey to the authority's public key and stores a
+// SignCert signs the certificate with an authority, setting the Nonce,
-// Signature, by authority, in the certificate.
+// SignatureKey, and Signature fields.
 func (c *Certificate) SignCert(rand io.Reader, authority Signer) error {
 	c.Nonce = make([]byte, 32)
 	if _, err := io.ReadFull(rand, c.Nonce); err != nil {
@@ -434,7 +437,9 @@ var certAlgoNames = map[string]string{
 	KeyAlgoECDSA256:   CertAlgoECDSA256v01,
 	KeyAlgoECDSA384:   CertAlgoECDSA384v01,
 	KeyAlgoECDSA521:   CertAlgoECDSA521v01,
+	KeyAlgoSKECDSA256: CertAlgoSKECDSA256v01,
 	KeyAlgoED25519:    CertAlgoED25519v01,
+	KeyAlgoSKED25519:  CertAlgoSKED25519v01,
 }
 
 // certToPrivAlgo returns the underlying algorithm for a certificate algorithm.
@@ -518,6 +523,12 @@ func parseSignatureBody(in []byte) (out *Signature, rest []byte, ok bool) {
 		return
 	}
 
+	switch out.Format {
+	case KeyAlgoSKECDSA256, CertAlgoSKECDSA256v01, KeyAlgoSKED25519, CertAlgoSKED25519v01:
+		out.Rest = in
+		return out, nil, ok
+	}
+
 	return out, in, ok
 }
 

vendor/golang.org/x/crypto/ssh/cipher.go

@@ -16,9 +16,8 @@ import (
 	"hash"
 	"io"
 	"io/ioutil"
-	"math/bits"
 
-	"golang.org/x/crypto/internal/chacha20"
+	"golang.org/x/crypto/chacha20"
 	"golang.org/x/crypto/poly1305"
 )
 
@@ -120,7 +119,7 @@ var cipherModes = map[string]*cipherMode{
 	chacha20Poly1305ID: {64, 0, newChaCha20Cipher},
 
 	// CBC mode is insecure and so is not included in the default config.
-	// (See http://www.isg.rhul.ac.uk/~kp/SandPfinal.pdf). If absolutely
+	// (See https://www.ieee-security.org/TC/SP2013/papers/4977a526.pdf). If absolutely
 	// needed, it's possible to specify a custom Config to enable it.
 	// You should expect that an active attacker can recover plaintext if
 	// you do.
@@ -642,8 +641,8 @@ const chacha20Poly1305ID = "chacha20-poly1305@openssh.com"
 // the methods here also implement padding, which RFC4253 Section 6
 // also requires of stream ciphers.
 type chacha20Poly1305Cipher struct {
-	lengthKey  [8]uint32
+	lengthKey  [32]byte
-	contentKey [8]uint32
+	contentKey [32]byte
 	buf        []byte
 }
 
@@ -656,21 +655,21 @@ func newChaCha20Cipher(key, unusedIV, unusedMACKey []byte, unusedAlgs directionA
 		buf: make([]byte, 256),
 	}
 
-	for i := range c.contentKey {
+	copy(c.contentKey[:], key[:32])
-		c.contentKey[i] = binary.LittleEndian.Uint32(key[i*4 : (i+1)*4])
+	copy(c.lengthKey[:], key[32:])
-	}
-	for i := range c.lengthKey {
-		c.lengthKey[i] = binary.LittleEndian.Uint32(key[(i+8)*4 : (i+9)*4])
-	}
 	return c, nil
 }
 
 func (c *chacha20Poly1305Cipher) readCipherPacket(seqNum uint32, r io.Reader) ([]byte, error) {
-	nonce := [3]uint32{0, 0, bits.ReverseBytes32(seqNum)}
+	nonce := make([]byte, 12)
-	s := chacha20.New(c.contentKey, nonce)
+	binary.BigEndian.PutUint32(nonce[8:], seqNum)
-	var polyKey [32]byte
+	s, err := chacha20.NewUnauthenticatedCipher(c.contentKey[:], nonce)
+	if err != nil {
+		return nil, err
+	}
+	var polyKey, discardBuf [32]byte
 	s.XORKeyStream(polyKey[:], polyKey[:])
-	s.Advance() // skip next 32 bytes
+	s.XORKeyStream(discardBuf[:], discardBuf[:]) // skip the next 32 bytes
 
 	encryptedLength := c.buf[:4]
 	if _, err := io.ReadFull(r, encryptedLength); err != nil {
@@ -678,7 +677,11 @@ func (c *chacha20Poly1305Cipher) readCipherPacket(seqNum uint32, r io.Reader) ([
 	}
 
 	var lenBytes [4]byte
-	chacha20.New(c.lengthKey, nonce).XORKeyStream(lenBytes[:], encryptedLength)
+	ls, err := chacha20.NewUnauthenticatedCipher(c.lengthKey[:], nonce)
+	if err != nil {
+		return nil, err
+	}
+	ls.XORKeyStream(lenBytes[:], encryptedLength)
 
 	length := binary.BigEndian.Uint32(lenBytes[:])
 	if length > maxPacket {
@@ -724,11 +727,15 @@ func (c *chacha20Poly1305Cipher) readCipherPacket(seqNum uint32, r io.Reader) ([
 }
 
 func (c *chacha20Poly1305Cipher) writeCipherPacket(seqNum uint32, w io.Writer, rand io.Reader, payload []byte) error {
-	nonce := [3]uint32{0, 0, bits.ReverseBytes32(seqNum)}
+	nonce := make([]byte, 12)
-	s := chacha20.New(c.contentKey, nonce)
+	binary.BigEndian.PutUint32(nonce[8:], seqNum)
-	var polyKey [32]byte
+	s, err := chacha20.NewUnauthenticatedCipher(c.contentKey[:], nonce)
+	if err != nil {
+		return err
+	}
+	var polyKey, discardBuf [32]byte
 	s.XORKeyStream(polyKey[:], polyKey[:])
-	s.Advance() // skip next 32 bytes
+	s.XORKeyStream(discardBuf[:], discardBuf[:]) // skip the next 32 bytes
 
 	// There is no blocksize, so fall back to multiple of 8 byte
 	// padding, as described in RFC 4253, Sec 6.
@@ -748,7 +755,11 @@ func (c *chacha20Poly1305Cipher) writeCipherPacket(seqNum uint32, w io.Writer, r
 	}
 
 	binary.BigEndian.PutUint32(c.buf, uint32(1+len(payload)+padding))
-	chacha20.New(c.lengthKey, nonce).XORKeyStream(c.buf, c.buf[:4])
+	ls, err := chacha20.NewUnauthenticatedCipher(c.lengthKey[:], nonce)
+	if err != nil {
+		return err
+	}
+	ls.XORKeyStream(c.buf, c.buf[:4])
 	c.buf[4] = byte(padding)
 	copy(c.buf[5:], payload)
 	packetEnd := 5 + len(payload) + padding

vendor/golang.org/x/crypto/ssh/internal/bcrypt_pbkdf/bcrypt_pbkdf.go

@@ -0,0 +1,93 @@
+// Copyright 2014 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 bcrypt_pbkdf implements bcrypt_pbkdf(3) from OpenBSD.
+//
+// See https://flak.tedunangst.com/post/bcrypt-pbkdf and
+// https://cvsweb.openbsd.org/cgi-bin/cvsweb/src/lib/libutil/bcrypt_pbkdf.c.
+package bcrypt_pbkdf
+
+import (
+	"crypto/sha512"
+	"errors"
+	"golang.org/x/crypto/blowfish"
+)
+
+const blockSize = 32
+
+// Key derives a key from the password, salt and rounds count, returning a
+// []byte of length keyLen that can be used as cryptographic key.
+func Key(password, salt []byte, rounds, keyLen int) ([]byte, error) {
+	if rounds < 1 {
+		return nil, errors.New("bcrypt_pbkdf: number of rounds is too small")
+	}
+	if len(password) == 0 {
+		return nil, errors.New("bcrypt_pbkdf: empty password")
+	}
+	if len(salt) == 0 || len(salt) > 1<<20 {
+		return nil, errors.New("bcrypt_pbkdf: bad salt length")
+	}
+	if keyLen > 1024 {
+		return nil, errors.New("bcrypt_pbkdf: keyLen is too large")
+	}
+
+	numBlocks := (keyLen + blockSize - 1) / blockSize
+	key := make([]byte, numBlocks*blockSize)
+
+	h := sha512.New()
+	h.Write(password)
+	shapass := h.Sum(nil)
+
+	shasalt := make([]byte, 0, sha512.Size)
+	cnt, tmp := make([]byte, 4), make([]byte, blockSize)
+	for block := 1; block <= numBlocks; block++ {
+		h.Reset()
+		h.Write(salt)
+		cnt[0] = byte(block >> 24)
+		cnt[1] = byte(block >> 16)
+		cnt[2] = byte(block >> 8)
+		cnt[3] = byte(block)
+		h.Write(cnt)
+		bcryptHash(tmp, shapass, h.Sum(shasalt))
+
+		out := make([]byte, blockSize)
+		copy(out, tmp)
+		for i := 2; i <= rounds; i++ {
+			h.Reset()
+			h.Write(tmp)
+			bcryptHash(tmp, shapass, h.Sum(shasalt))
+			for j := 0; j < len(out); j++ {
+				out[j] ^= tmp[j]
+			}
+		}
+
+		for i, v := range out {
+			key[i*numBlocks+(block-1)] = v
+		}
+	}
+	return key[:keyLen], nil
+}
+
+var magic = []byte("OxychromaticBlowfishSwatDynamite")
+
+func bcryptHash(out, shapass, shasalt []byte) {
+	c, err := blowfish.NewSaltedCipher(shapass, shasalt)
+	if err != nil {
+		panic(err)
+	}
+	for i := 0; i < 64; i++ {
+		blowfish.ExpandKey(shasalt, c)
+		blowfish.ExpandKey(shapass, c)
+	}
+	copy(out, magic)
+	for i := 0; i < 32; i += 8 {
+		for j := 0; j < 64; j++ {
+			c.Encrypt(out[i:i+8], out[i:i+8])
+		}
+	}
+	// Swap bytes due to different endianness.
+	for i := 0; i < 32; i += 4 {
+		out[i+3], out[i+2], out[i+1], out[i] = out[i], out[i+1], out[i+2], out[i+3]
+	}
+}

vendor/golang.org/x/crypto/ssh/kex.go

@@ -212,7 +212,7 @@ func (group *dhGroup) Server(c packetConn, randSource io.Reader, magics *handsha
 		HostKey:   hostKeyBytes,
 		Signature: sig,
 		Hash:      crypto.SHA1,
-	}, nil
+	}, err
 }
 
 // ecdh performs Elliptic Curve Diffie-Hellman key exchange as
@@ -572,7 +572,7 @@ func (gex *dhGEXSHA) diffieHellman(theirPublic, myPrivate *big.Int) (*big.Int, e
 	return new(big.Int).Exp(theirPublic, myPrivate, gex.p), nil
 }
 
-func (gex *dhGEXSHA) Client(c packetConn, randSource io.Reader, magics *handshakeMagics) (*kexResult, error) {
+func (gex dhGEXSHA) Client(c packetConn, randSource io.Reader, magics *handshakeMagics) (*kexResult, error) {
 	// Send GexRequest
 	kexDHGexRequest := kexDHGexRequestMsg{
 		MinBits:      dhGroupExchangeMinimumBits,
@@ -677,7 +677,7 @@ func (gex *dhGEXSHA) Client(c packetConn, randSource io.Reader, magics *handshak
 // Server half implementation of the Diffie Hellman Key Exchange with SHA1 and SHA256.
 //
 // This is a minimal implementation to satisfy the automated tests.
-func (gex *dhGEXSHA) Server(c packetConn, randSource io.Reader, magics *handshakeMagics, priv Signer) (result *kexResult, err error) {
+func (gex dhGEXSHA) Server(c packetConn, randSource io.Reader, magics *handshakeMagics, priv Signer) (result *kexResult, err error) {
 	// Receive GexRequest
 	packet, err := c.readPacket()
 	if err != nil {

vendor/golang.org/x/crypto/ssh/keys.go

@@ -7,6 +7,8 @@ package ssh
 import (
 	"bytes"
 	"crypto"
+	"crypto/aes"
+	"crypto/cipher"
 	"crypto/dsa"
 	"crypto/ecdsa"
 	"crypto/elliptic"
@@ -25,6 +27,7 @@ import (
 	"strings"
 
 	"golang.org/x/crypto/ed25519"
+	"golang.org/x/crypto/ssh/internal/bcrypt_pbkdf"
 )
 
 // These constants represent the algorithm names for key types supported by this
@@ -33,9 +36,11 @@ const (
 	KeyAlgoRSA        = "ssh-rsa"
 	KeyAlgoDSA        = "ssh-dss"
 	KeyAlgoECDSA256   = "ecdsa-sha2-nistp256"
+	KeyAlgoSKECDSA256 = "sk-ecdsa-sha2-nistp256@openssh.com"
 	KeyAlgoECDSA384   = "ecdsa-sha2-nistp384"
 	KeyAlgoECDSA521   = "ecdsa-sha2-nistp521"
 	KeyAlgoED25519    = "ssh-ed25519"
+	KeyAlgoSKED25519  = "sk-ssh-ed25519@openssh.com"
 )
 
 // These constants represent non-default signature algorithms that are supported
@@ -58,9 +63,13 @@ func parsePubKey(in []byte, algo string) (pubKey PublicKey, rest []byte, err err
 		return parseDSA(in)
 	case KeyAlgoECDSA256, KeyAlgoECDSA384, KeyAlgoECDSA521:
 		return parseECDSA(in)
+	case KeyAlgoSKECDSA256:
+		return parseSKECDSA(in)
 	case KeyAlgoED25519:
 		return parseED25519(in)
-	case CertAlgoRSAv01, CertAlgoDSAv01, CertAlgoECDSA256v01, CertAlgoECDSA384v01, CertAlgoECDSA521v01, CertAlgoED25519v01:
+	case KeyAlgoSKED25519:
+		return parseSKEd25519(in)
+	case CertAlgoRSAv01, CertAlgoDSAv01, CertAlgoECDSA256v01, CertAlgoECDSA384v01, CertAlgoECDSA521v01, CertAlgoSKECDSA256v01, CertAlgoED25519v01, CertAlgoSKED25519v01:
 		cert, err := parseCert(in, certToPrivAlgo(algo))
 		if err != nil {
 			return nil, nil, err
@@ -553,9 +562,11 @@ func parseED25519(in []byte) (out PublicKey, rest []byte, err error) {
 		return nil, nil, err
 	}
 
-	key := ed25519.PublicKey(w.KeyBytes)
+	if l := len(w.KeyBytes); l != ed25519.PublicKeySize {
+		return nil, nil, fmt.Errorf("invalid size %d for Ed25519 public key", l)
+	}
 
-	return (ed25519PublicKey)(key), w.Rest, nil
+	return ed25519PublicKey(w.KeyBytes), w.Rest, nil
 }
 
 func (k ed25519PublicKey) Marshal() []byte {
@@ -573,9 +584,11 @@ func (k ed25519PublicKey) Verify(b []byte, sig *Signature) error {
 	if sig.Format != k.Type() {
 		return fmt.Errorf("ssh: signature type %s for key type %s", sig.Format, k.Type())
 	}
+	if l := len(k); l != ed25519.PublicKeySize {
+		return fmt.Errorf("ssh: invalid size %d for Ed25519 public key", l)
+	}
 
-	edKey := (ed25519.PublicKey)(k)
+	if ok := ed25519.Verify(ed25519.PublicKey(k), b, sig.Blob); !ok {
-	if ok := ed25519.Verify(edKey, b, sig.Blob); !ok {
 		return errors.New("ssh: signature did not verify")
 	}
 
@@ -685,6 +698,224 @@ func (k *ecdsaPublicKey) CryptoPublicKey() crypto.PublicKey {
 	return (*ecdsa.PublicKey)(k)
 }
 
+// skFields holds the additional fields present in U2F/FIDO2 signatures.
+// See openssh/PROTOCOL.u2f 'SSH U2F Signatures' for details.
+type skFields struct {
+	// Flags contains U2F/FIDO2 flags such as 'user present'
+	Flags byte
+	// Counter is a monotonic signature counter which can be
+	// used to detect concurrent use of a private key, should
+	// it be extracted from hardware.
+	Counter uint32
+}
+
+type skECDSAPublicKey struct {
+	// application is a URL-like string, typically "ssh:" for SSH.
+	// see openssh/PROTOCOL.u2f for details.
+	application string
+	ecdsa.PublicKey
+}
+
+func (k *skECDSAPublicKey) Type() string {
+	return KeyAlgoSKECDSA256
+}
+
+func (k *skECDSAPublicKey) nistID() string {
+	return "nistp256"
+}
+
+func parseSKECDSA(in []byte) (out PublicKey, rest []byte, err error) {
+	var w struct {
+		Curve       string
+		KeyBytes    []byte
+		Application string
+		Rest        []byte `ssh:"rest"`
+	}
+
+	if err := Unmarshal(in, &w); err != nil {
+		return nil, nil, err
+	}
+
+	key := new(skECDSAPublicKey)
+	key.application = w.Application
+
+	if w.Curve != "nistp256" {
+		return nil, nil, errors.New("ssh: unsupported curve")
+	}
+	key.Curve = elliptic.P256()
+
+	key.X, key.Y = elliptic.Unmarshal(key.Curve, w.KeyBytes)
+	if key.X == nil || key.Y == nil {
+		return nil, nil, errors.New("ssh: invalid curve point")
+	}
+
+	return key, w.Rest, nil
+}
+
+func (k *skECDSAPublicKey) Marshal() []byte {
+	// See RFC 5656, section 3.1.
+	keyBytes := elliptic.Marshal(k.Curve, k.X, k.Y)
+	w := struct {
+		Name        string
+		ID          string
+		Key         []byte
+		Application string
+	}{
+		k.Type(),
+		k.nistID(),
+		keyBytes,
+		k.application,
+	}
+
+	return Marshal(&w)
+}
+
+func (k *skECDSAPublicKey) Verify(data []byte, sig *Signature) error {
+	if sig.Format != k.Type() {
+		return fmt.Errorf("ssh: signature type %s for key type %s", sig.Format, k.Type())
+	}
+
+	h := ecHash(k.Curve).New()
+	h.Write([]byte(k.application))
+	appDigest := h.Sum(nil)
+
+	h.Reset()
+	h.Write(data)
+	dataDigest := h.Sum(nil)
+
+	var ecSig struct {
+		R *big.Int
+		S *big.Int
+	}
+	if err := Unmarshal(sig.Blob, &ecSig); err != nil {
+		return err
+	}
+
+	var skf skFields
+	if err := Unmarshal(sig.Rest, &skf); err != nil {
+		return err
+	}
+
+	blob := struct {
+		ApplicationDigest []byte `ssh:"rest"`
+		Flags             byte
+		Counter           uint32
+		MessageDigest     []byte `ssh:"rest"`
+	}{
+		appDigest,
+		skf.Flags,
+		skf.Counter,
+		dataDigest,
+	}
+
+	original := Marshal(blob)
+
+	h.Reset()
+	h.Write(original)
+	digest := h.Sum(nil)
+
+	if ecdsa.Verify((*ecdsa.PublicKey)(&k.PublicKey), digest, ecSig.R, ecSig.S) {
+		return nil
+	}
+	return errors.New("ssh: signature did not verify")
+}
+
+type skEd25519PublicKey struct {
+	// application is a URL-like string, typically "ssh:" for SSH.
+	// see openssh/PROTOCOL.u2f for details.
+	application string
+	ed25519.PublicKey
+}
+
+func (k *skEd25519PublicKey) Type() string {
+	return KeyAlgoSKED25519
+}
+
+func parseSKEd25519(in []byte) (out PublicKey, rest []byte, err error) {
+	var w struct {
+		KeyBytes    []byte
+		Application string
+		Rest        []byte `ssh:"rest"`
+	}
+
+	if err := Unmarshal(in, &w); err != nil {
+		return nil, nil, err
+	}
+
+	if l := len(w.KeyBytes); l != ed25519.PublicKeySize {
+		return nil, nil, fmt.Errorf("invalid size %d for Ed25519 public key", l)
+	}
+
+	key := new(skEd25519PublicKey)
+	key.application = w.Application
+	key.PublicKey = ed25519.PublicKey(w.KeyBytes)
+
+	return key, w.Rest, nil
+}
+
+func (k *skEd25519PublicKey) Marshal() []byte {
+	w := struct {
+		Name        string
+		KeyBytes    []byte
+		Application string
+	}{
+		KeyAlgoSKED25519,
+		[]byte(k.PublicKey),
+		k.application,
+	}
+	return Marshal(&w)
+}
+
+func (k *skEd25519PublicKey) Verify(data []byte, sig *Signature) error {
+	if sig.Format != k.Type() {
+		return fmt.Errorf("ssh: signature type %s for key type %s", sig.Format, k.Type())
+	}
+	if l := len(k.PublicKey); l != ed25519.PublicKeySize {
+		return fmt.Errorf("invalid size %d for Ed25519 public key", l)
+	}
+
+	h := sha256.New()
+	h.Write([]byte(k.application))
+	appDigest := h.Sum(nil)
+
+	h.Reset()
+	h.Write(data)
+	dataDigest := h.Sum(nil)
+
+	var edSig struct {
+		Signature []byte `ssh:"rest"`
+	}
+
+	if err := Unmarshal(sig.Blob, &edSig); err != nil {
+		return err
+	}
+
+	var skf skFields
+	if err := Unmarshal(sig.Rest, &skf); err != nil {
+		return err
+	}
+
+	blob := struct {
+		ApplicationDigest []byte `ssh:"rest"`
+		Flags             byte
+		Counter           uint32
+		MessageDigest     []byte `ssh:"rest"`
+	}{
+		appDigest,
+		skf.Flags,
+		skf.Counter,
+		dataDigest,
+	}
+
+	original := Marshal(blob)
+
+	if ok := ed25519.Verify(k.PublicKey, original, edSig.Signature); !ok {
+		return errors.New("ssh: signature did not verify")
+	}
+
+	return nil
+}
+
 // NewSignerFromKey takes an *rsa.PrivateKey, *dsa.PrivateKey,
 // *ecdsa.PrivateKey or any other crypto.Signer and returns a
 // corresponding Signer instance. ECDSA keys must use P-256, P-384 or
@@ -830,14 +1061,18 @@ func NewPublicKey(key interface{}) (PublicKey, error) {
 	case *dsa.PublicKey:
 		return (*dsaPublicKey)(key), nil
 	case ed25519.PublicKey:
-		return (ed25519PublicKey)(key), nil
+		if l := len(key); l != ed25519.PublicKeySize {
+			return nil, fmt.Errorf("ssh: invalid size %d for Ed25519 public key", l)
+		}
+		return ed25519PublicKey(key), nil
 	default:
 		return nil, fmt.Errorf("ssh: unsupported key type %T", key)
 	}
 }
 
 // ParsePrivateKey returns a Signer from a PEM encoded private key. It supports
-// the same keys as ParseRawPrivateKey.
+// the same keys as ParseRawPrivateKey. If the private key is encrypted, it
+// will return a PassphraseMissingError.
 func ParsePrivateKey(pemBytes []byte) (Signer, error) {
 	key, err := ParseRawPrivateKey(pemBytes)
 	if err != nil {
@@ -850,8 +1085,8 @@ func ParsePrivateKey(pemBytes []byte) (Signer, error) {
 // ParsePrivateKeyWithPassphrase returns a Signer from a PEM encoded private
 // key and passphrase. It supports the same keys as
 // ParseRawPrivateKeyWithPassphrase.
-func ParsePrivateKeyWithPassphrase(pemBytes, passPhrase []byte) (Signer, error) {
+func ParsePrivateKeyWithPassphrase(pemBytes, passphrase []byte) (Signer, error) {
-	key, err := ParseRawPrivateKeyWithPassphrase(pemBytes, passPhrase)
+	key, err := ParseRawPrivateKeyWithPassphrase(pemBytes, passphrase)
 	if err != nil {
 		return nil, err
 	}
@@ -867,8 +1102,21 @@ func encryptedBlock(block *pem.Block) bool {
 	return strings.Contains(block.Headers["Proc-Type"], "ENCRYPTED")
 }
 
+// A PassphraseMissingError indicates that parsing this private key requires a
+// passphrase. Use ParsePrivateKeyWithPassphrase.
+type PassphraseMissingError struct {
+	// PublicKey will be set if the private key format includes an unencrypted
+	// public key along with the encrypted private key.
+	PublicKey PublicKey
+}
+
+func (*PassphraseMissingError) Error() string {
+	return "ssh: this private key is passphrase protected"
+}
+
 // ParseRawPrivateKey returns a private key from a PEM encoded private key. It
-// supports RSA (PKCS#1), PKCS#8, DSA (OpenSSL), and ECDSA private keys.
+// supports RSA (PKCS#1), PKCS#8, DSA (OpenSSL), and ECDSA private keys. If the
+// private key is encrypted, it will return a PassphraseMissingError.
 func ParseRawPrivateKey(pemBytes []byte) (interface{}, error) {
 	block, _ := pem.Decode(pemBytes)
 	if block == nil {
@@ -876,7 +1124,7 @@ func ParseRawPrivateKey(pemBytes []byte) (interface{}, error) {
 	}
 
 	if encryptedBlock(block) {
-		return nil, errors.New("ssh: cannot decode encrypted private keys")
+		return nil, &PassphraseMissingError{}
 	}
 
 	switch block.Type {
@@ -890,34 +1138,36 @@ func ParseRawPrivateKey(pemBytes []byte) (interface{}, error) {
 	case "DSA PRIVATE KEY":
 		return ParseDSAPrivateKey(block.Bytes)
 	case "OPENSSH PRIVATE KEY":
-		return parseOpenSSHPrivateKey(block.Bytes)
+		return parseOpenSSHPrivateKey(block.Bytes, unencryptedOpenSSHKey)
 	default:
 		return nil, fmt.Errorf("ssh: unsupported key type %q", block.Type)
 	}
 }
 
 // ParseRawPrivateKeyWithPassphrase returns a private key decrypted with
-// passphrase from a PEM encoded private key. If wrong passphrase, return
+// passphrase from a PEM encoded private key. If the passphrase is wrong, it
-// x509.IncorrectPasswordError.
+// will return x509.IncorrectPasswordError.
-func ParseRawPrivateKeyWithPassphrase(pemBytes, passPhrase []byte) (interface{}, error) {
+func ParseRawPrivateKeyWithPassphrase(pemBytes, passphrase []byte) (interface{}, error) {
 	block, _ := pem.Decode(pemBytes)
 	if block == nil {
 		return nil, errors.New("ssh: no key found")
 	}
-	buf := block.Bytes
 
-	if encryptedBlock(block) {
+	if block.Type == "OPENSSH PRIVATE KEY" {
-		if x509.IsEncryptedPEMBlock(block) {
+		return parseOpenSSHPrivateKey(block.Bytes, passphraseProtectedOpenSSHKey(passphrase))
-			var err error
+	}
-			buf, err = x509.DecryptPEMBlock(block, passPhrase)
+
+	if !encryptedBlock(block) || !x509.IsEncryptedPEMBlock(block) {
+		return nil, errors.New("ssh: not an encrypted key")
+	}
+
+	buf, err := x509.DecryptPEMBlock(block, passphrase)
 	if err != nil {
 		if err == x509.IncorrectPasswordError {
 			return nil, err
 		}
 		return nil, fmt.Errorf("ssh: cannot decode encrypted private keys: %v", err)
 	}
-		}
-	}
 
 	switch block.Type {
 	case "RSA PRIVATE KEY":
@@ -926,8 +1176,6 @@ func ParseRawPrivateKeyWithPassphrase(pemBytes, passPhrase []byte) (interface{},
 		return x509.ParseECPrivateKey(buf)
 	case "DSA PRIVATE KEY":
 		return ParseDSAPrivateKey(buf)
-	case "OPENSSH PRIVATE KEY":
-		return parseOpenSSHPrivateKey(buf)
 	default:
 		return nil, fmt.Errorf("ssh: unsupported key type %q", block.Type)
 	}
@@ -965,9 +1213,68 @@ func ParseDSAPrivateKey(der []byte) (*dsa.PrivateKey, error) {
 	}, nil
 }
 
-// Implemented based on the documentation at
+func unencryptedOpenSSHKey(cipherName, kdfName, kdfOpts string, privKeyBlock []byte) ([]byte, error) {
-// https://github.com/openssh/openssh-portable/blob/master/PROTOCOL.key
+	if kdfName != "none" || cipherName != "none" {
-func parseOpenSSHPrivateKey(key []byte) (crypto.PrivateKey, error) {
+		return nil, &PassphraseMissingError{}
+	}
+	if kdfOpts != "" {
+		return nil, errors.New("ssh: invalid openssh private key")
+	}
+	return privKeyBlock, nil
+}
+
+func passphraseProtectedOpenSSHKey(passphrase []byte) openSSHDecryptFunc {
+	return func(cipherName, kdfName, kdfOpts string, privKeyBlock []byte) ([]byte, error) {
+		if kdfName == "none" || cipherName == "none" {
+			return nil, errors.New("ssh: key is not password protected")
+		}
+		if kdfName != "bcrypt" {
+			return nil, fmt.Errorf("ssh: unknown KDF %q, only supports %q", kdfName, "bcrypt")
+		}
+
+		var opts struct {
+			Salt   string
+			Rounds uint32
+		}
+		if err := Unmarshal([]byte(kdfOpts), &opts); err != nil {
+			return nil, err
+		}
+
+		k, err := bcrypt_pbkdf.Key(passphrase, []byte(opts.Salt), int(opts.Rounds), 32+16)
+		if err != nil {
+			return nil, err
+		}
+		key, iv := k[:32], k[32:]
+
+		c, err := aes.NewCipher(key)
+		if err != nil {
+			return nil, err
+		}
+		switch cipherName {
+		case "aes256-ctr":
+			ctr := cipher.NewCTR(c, iv)
+			ctr.XORKeyStream(privKeyBlock, privKeyBlock)
+		case "aes256-cbc":
+			if len(privKeyBlock)%c.BlockSize() != 0 {
+				return nil, fmt.Errorf("ssh: invalid encrypted private key length, not a multiple of the block size")
+			}
+			cbc := cipher.NewCBCDecrypter(c, iv)
+			cbc.CryptBlocks(privKeyBlock, privKeyBlock)
+		default:
+			return nil, fmt.Errorf("ssh: unknown cipher %q, only supports %q or %q", cipherName, "aes256-ctr", "aes256-cbc")
+		}
+
+		return privKeyBlock, nil
+	}
+}
+
+type openSSHDecryptFunc func(CipherName, KdfName, KdfOpts string, PrivKeyBlock []byte) ([]byte, error)
+
+// parseOpenSSHPrivateKey parses an OpenSSH private key, using the decrypt
+// function to unwrap the encrypted portion. unencryptedOpenSSHKey can be used
+// as the decrypt function to parse an unencrypted private key. See
+// https://github.com/openssh/openssh-portable/blob/master/PROTOCOL.key.
+func parseOpenSSHPrivateKey(key []byte, decrypt openSSHDecryptFunc) (crypto.PrivateKey, error) {
 	const magic = "openssh-key-v1\x00"
 	if len(key) < len(magic) || string(key[:len(magic)]) != magic {
 		return nil, errors.New("ssh: invalid openssh private key format")
@@ -986,9 +1293,22 @@ func parseOpenSSHPrivateKey(key []byte) (crypto.PrivateKey, error) {
 	if err := Unmarshal(remaining, &w); err != nil {
 		return nil, err
 	}
+	if w.NumKeys != 1 {
+		// We only support single key files, and so does OpenSSH.
+		// https://github.com/openssh/openssh-portable/blob/4103a3ec7/sshkey.c#L4171
+		return nil, errors.New("ssh: multi-key files are not supported")
+	}
 
-	if w.KdfName != "none" || w.CipherName != "none" {
+	privKeyBlock, err := decrypt(w.CipherName, w.KdfName, w.KdfOpts, w.PrivKeyBlock)
-		return nil, errors.New("ssh: cannot decode encrypted private keys")
+	if err != nil {
+		if err, ok := err.(*PassphraseMissingError); ok {
+			pub, errPub := ParsePublicKey(w.PubKey)
+			if errPub != nil {
+				return nil, fmt.Errorf("ssh: failed to parse embedded public key: %v", errPub)
+			}
+			err.PublicKey = pub
+		}
+		return nil, err
 	}
 
 	pk1 := struct {
@@ -998,15 +1318,13 @@ func parseOpenSSHPrivateKey(key []byte) (crypto.PrivateKey, error) {
 		Rest    []byte `ssh:"rest"`
 	}{}
 
-	if err := Unmarshal(w.PrivKeyBlock, &pk1); err != nil {
+	if err := Unmarshal(privKeyBlock, &pk1); err != nil || pk1.Check1 != pk1.Check2 {
-		return nil, err
+		if w.CipherName != "none" {
+			return nil, x509.IncorrectPasswordError
+		}
+		return nil, errors.New("ssh: malformed OpenSSH key")
 	}
 
-	if pk1.Check1 != pk1.Check2 {
-		return nil, errors.New("ssh: checkint mismatch")
-	}
-
-	// we only handle ed25519 and rsa keys currently
 	switch pk1.Keytype {
 	case KeyAlgoRSA:
 		// https://github.com/openssh/openssh-portable/blob/master/sshkey.c#L2760-L2773
@@ -1025,10 +1343,8 @@ func parseOpenSSHPrivateKey(key []byte) (crypto.PrivateKey, error) {
 			return nil, err
 		}
 
-		for i, b := range key.Pad {
+		if err := checkOpenSSHKeyPadding(key.Pad); err != nil {
-			if int(b) != i+1 {
+			return nil, err
-				return nil, errors.New("ssh: padding not as expected")
-			}
 		}
 
 		pk := &rsa.PrivateKey{
@@ -1063,20 +1379,78 @@ func parseOpenSSHPrivateKey(key []byte) (crypto.PrivateKey, error) {
 			return nil, errors.New("ssh: private key unexpected length")
 		}
 
-		for i, b := range key.Pad {
+		if err := checkOpenSSHKeyPadding(key.Pad); err != nil {
-			if int(b) != i+1 {
+			return nil, err
-				return nil, errors.New("ssh: padding not as expected")
-			}
 		}
 
 		pk := ed25519.PrivateKey(make([]byte, ed25519.PrivateKeySize))
 		copy(pk, key.Priv)
 		return &pk, nil
+	case KeyAlgoECDSA256, KeyAlgoECDSA384, KeyAlgoECDSA521:
+		key := struct {
+			Curve   string
+			Pub     []byte
+			D       *big.Int
+			Comment string
+			Pad     []byte `ssh:"rest"`
+		}{}
+
+		if err := Unmarshal(pk1.Rest, &key); err != nil {
+			return nil, err
+		}
+
+		if err := checkOpenSSHKeyPadding(key.Pad); err != nil {
+			return nil, err
+		}
+
+		var curve elliptic.Curve
+		switch key.Curve {
+		case "nistp256":
+			curve = elliptic.P256()
+		case "nistp384":
+			curve = elliptic.P384()
+		case "nistp521":
+			curve = elliptic.P521()
+		default:
+			return nil, errors.New("ssh: unhandled elliptic curve: " + key.Curve)
+		}
+
+		X, Y := elliptic.Unmarshal(curve, key.Pub)
+		if X == nil || Y == nil {
+			return nil, errors.New("ssh: failed to unmarshal public key")
+		}
+
+		if key.D.Cmp(curve.Params().N) >= 0 {
+			return nil, errors.New("ssh: scalar is out of range")
+		}
+
+		x, y := curve.ScalarBaseMult(key.D.Bytes())
+		if x.Cmp(X) != 0 || y.Cmp(Y) != 0 {
+			return nil, errors.New("ssh: public key does not match private key")
+		}
+
+		return &ecdsa.PrivateKey{
+			PublicKey: ecdsa.PublicKey{
+				Curve: curve,
+				X:     X,
+				Y:     Y,
+			},
+			D: key.D,
+		}, nil
 	default:
 		return nil, errors.New("ssh: unhandled key type")
 	}
 }
 
+func checkOpenSSHKeyPadding(pad []byte) error {
+	for i, b := range pad {
+		if int(b) != i+1 {
+			return errors.New("ssh: padding not as expected")
+		}
+	}
+	return nil
+}
+
 // FingerprintLegacyMD5 returns the user presentation of the key's
 // fingerprint as described by RFC 4716 section 4.
 func FingerprintLegacyMD5(pubKey PublicKey) string {

vendor/golang.org/x/crypto/ssh/mux.go

@@ -240,7 +240,7 @@ func (m *mux) onePacket() error {
 	id := binary.BigEndian.Uint32(packet[1:])
 	ch := m.chanList.getChan(id)
 	if ch == nil {
-		return fmt.Errorf("ssh: invalid channel %d", id)
+		return m.handleUnknownChannelPacket(id, packet)
 	}
 
 	return ch.handlePacket(packet)
@@ -328,3 +328,24 @@ func (m *mux) openChannel(chanType string, extra []byte) (*channel, error) {
 		return nil, fmt.Errorf("ssh: unexpected packet in response to channel open: %T", msg)
 	}
 }
+
+func (m *mux) handleUnknownChannelPacket(id uint32, packet []byte) error {
+	msg, err := decode(packet)
+	if err != nil {
+		return err
+	}
+
+	switch msg := msg.(type) {
+	// RFC 4254 section 5.4 says unrecognized channel requests should
+	// receive a failure response.
+	case *channelRequestMsg:
+		if msg.WantReply {
+			return m.sendMessage(channelRequestFailureMsg{
+				PeersID: msg.PeersID,
+			})
+		}
+		return nil
+	default:
+		return fmt.Errorf("ssh: invalid channel %d", id)
+	}
+}

vendor/golang.org/x/crypto/ssh/server.go

@@ -284,8 +284,8 @@ func (s *connection) serverHandshake(config *ServerConfig) (*Permissions, error)
 
 func isAcceptableAlgo(algo string) bool {
 	switch algo {
-	case KeyAlgoRSA, KeyAlgoDSA, KeyAlgoECDSA256, KeyAlgoECDSA384, KeyAlgoECDSA521, KeyAlgoED25519,
+	case KeyAlgoRSA, KeyAlgoDSA, KeyAlgoECDSA256, KeyAlgoECDSA384, KeyAlgoECDSA521, KeyAlgoSKECDSA256, KeyAlgoED25519, KeyAlgoSKED25519,
-		CertAlgoRSAv01, CertAlgoDSAv01, CertAlgoECDSA256v01, CertAlgoECDSA384v01, CertAlgoECDSA521v01, CertAlgoED25519v01:
+		CertAlgoRSAv01, CertAlgoDSAv01, CertAlgoECDSA256v01, CertAlgoECDSA384v01, CertAlgoECDSA521v01, CertAlgoSKECDSA256v01, CertAlgoED25519v01, CertAlgoSKED25519v01:
 		return true
 	}
 	return false

vendor/golang.org/x/crypto/ssh/terminal/terminal.go

@@ -7,6 +7,7 @@ package terminal
 import (
 	"bytes"
 	"io"
+	"runtime"
 	"strconv"
 	"sync"
 	"unicode/utf8"
@@ -112,6 +113,7 @@ func NewTerminal(c io.ReadWriter, prompt string) *Terminal {
 }
 
 const (
+	keyCtrlC     = 3
 	keyCtrlD     = 4
 	keyCtrlU     = 21
 	keyEnter     = '\r'
@@ -150,8 +152,12 @@ func bytesToKey(b []byte, pasteActive bool) (rune, []byte) {
 		switch b[0] {
 		case 1: // ^A
 			return keyHome, b[1:]
+		case 2: // ^B
+			return keyLeft, b[1:]
 		case 5: // ^E
 			return keyEnd, b[1:]
+		case 6: // ^F
+			return keyRight, b[1:]
 		case 8: // ^H
 			return keyBackspace, b[1:]
 		case 11: // ^K
@@ -737,6 +743,9 @@ func (t *Terminal) readLine() (line string, err error) {
 						return "", io.EOF
 					}
 				}
+				if key == keyCtrlC {
+					return "", io.EOF
+				}
 				if key == keyPasteStart {
 					t.pasteActive = true
 					if len(t.line) == 0 {
@@ -939,6 +948,8 @@ func (s *stRingBuffer) NthPreviousEntry(n int) (value string, ok bool) {
 // readPasswordLine reads from reader until it finds \n or io.EOF.
 // The slice returned does not include the \n.
 // readPasswordLine also ignores any \r it finds.
+// Windows uses \r as end of line. So, on Windows, readPasswordLine
+// reads until it finds \r and ignores any \n it finds during processing.
 func readPasswordLine(reader io.Reader) ([]byte, error) {
 	var buf [1]byte
 	var ret []byte
@@ -947,10 +958,20 @@ func readPasswordLine(reader io.Reader) ([]byte, error) {
 		n, err := reader.Read(buf[:])
 		if n > 0 {
 			switch buf[0] {
+			case '\b':
+				if len(ret) > 0 {
+					ret = ret[:len(ret)-1]
+				}
 			case '\n':
+				if runtime.GOOS != "windows" {
 					return ret, nil
+				}
+				// otherwise ignore \n
 			case '\r':
-				// remove \r from passwords on Windows
+				if runtime.GOOS == "windows" {
+					return ret, nil
+				}
+				// otherwise ignore \r
 			default:
 				ret = append(ret, buf[0])
 			}

vendor/golang.org/x/crypto/ssh/terminal/util_windows.go

@@ -85,8 +85,8 @@ func ReadPassword(fd int) ([]byte, error) {
 	}
 	old := st
 
-	st &^= (windows.ENABLE_ECHO_INPUT)
+	st &^= (windows.ENABLE_ECHO_INPUT | windows.ENABLE_LINE_INPUT)
-	st |= (windows.ENABLE_PROCESSED_INPUT | windows.ENABLE_LINE_INPUT | windows.ENABLE_PROCESSED_OUTPUT)
+	st |= (windows.ENABLE_PROCESSED_OUTPUT | windows.ENABLE_PROCESSED_INPUT)
 	if err := windows.SetConsoleMode(windows.Handle(fd), st); err != nil {
 		return nil, err
 	}

vendor/modules.txt

@@ -194,11 +194,13 @@ github.com/stretchr/testify/assert
 github.com/stretchr/testify/require
 # github.com/xo/dburl v0.0.0-20191005012637-293c3298d6c0
 github.com/xo/dburl
-# golang.org/x/crypto v0.0.0-20191011191535-87dc89f01550
+# golang.org/x/crypto v0.0.0-20200604202706-70a84ac30bf9
+golang.org/x/crypto/blake2b
+golang.org/x/crypto/blowfish
+golang.org/x/crypto/chacha20
 golang.org/x/crypto/curve25519
 golang.org/x/crypto/ed25519
 golang.org/x/crypto/ed25519/internal/edwards25519
-golang.org/x/crypto/internal/chacha20
 golang.org/x/crypto/internal/subtle
 golang.org/x/crypto/md4
 golang.org/x/crypto/nacl/box
@@ -207,6 +209,7 @@ golang.org/x/crypto/pbkdf2
 golang.org/x/crypto/poly1305
 golang.org/x/crypto/salsa20/salsa
 golang.org/x/crypto/ssh
+golang.org/x/crypto/ssh/internal/bcrypt_pbkdf
 golang.org/x/crypto/ssh/terminal
 # golang.org/x/net v0.0.0-20191014212845-da9a3fd4c582
 golang.org/x/net/bpf