Use x/crypto/chacha20, x/crypto/poly1305, and x/crypto/curve25519

psiphon/common/crypto/ssh/cipher.go

@@ -16,10 +16,9 @@ import (
 	"hash"
 	"io"
 	"io/ioutil"
-	"math/bits"
 
-	"github.com/Psiphon-Labs/psiphon-tunnel-core/psiphon/common/crypto/internal/chacha20"
-	"github.com/Psiphon-Labs/psiphon-tunnel-core/psiphon/common/crypto/poly1305"
+	"golang.org/x/crypto/chacha20"
+	"golang.org/x/crypto/poly1305"
 )
 
 const (
@@ -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
-	contentKey [8]uint32
+	lengthKey  [32]byte
+	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 {
-		c.contentKey[i] = binary.LittleEndian.Uint32(key[i*4 : (i+1)*4])
-	}
-	for i := range c.lengthKey {
-		c.lengthKey[i] = binary.LittleEndian.Uint32(key[(i+8)*4 : (i+9)*4])
-	}
+	copy(c.contentKey[:], key[:32])
+	copy(c.lengthKey[:], key[32:])
 	return c, nil
 }
 
 func (c *chacha20Poly1305Cipher) readPacket(seqNum uint32, r io.Reader) ([]byte, error) {
-	nonce := [3]uint32{0, 0, bits.ReverseBytes32(seqNum)}
-	s := chacha20.New(c.contentKey, nonce)
-	var polyKey [32]byte
+	nonce := make([]byte, 12)
+	binary.BigEndian.PutUint32(nonce[8:], seqNum)
+	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) readPacket(seqNum uint32, r io.Reader) ([]byte,
 	}
 
 	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) readPacket(seqNum uint32, r io.Reader) ([]byte,
 }
 
 func (c *chacha20Poly1305Cipher) writePacket(seqNum uint32, w io.Writer, rand io.Reader, payload []byte) error {
-	nonce := [3]uint32{0, 0, bits.ReverseBytes32(seqNum)}
-	s := chacha20.New(c.contentKey, nonce)
-	var polyKey [32]byte
+	nonce := make([]byte, 12)
+	binary.BigEndian.PutUint32(nonce[8:], seqNum)
+	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) writePacket(seqNum uint32, w io.Writer, rand io
 	}
 
 	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

psiphon/common/crypto/ssh/kex.go

@@ -14,7 +14,7 @@ import (
 	"io"
 	"math/big"
 
-	"github.com/Psiphon-Labs/psiphon-tunnel-core/psiphon/common/crypto/curve25519"
+	"golang.org/x/crypto/curve25519"
 )
 
 const (

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

@@ -0,0 +1,17 @@
+// 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,!appengine
+
+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

@@ -0,0 +1,308 @@
+// 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,!appengine
+
+#include "textflag.h"
+
+#define NUM_ROUNDS 10
+
+// func xorKeyStreamVX(dst, src []byte, key *[8]uint32, nonce *[3]uint32, counter *uint32)
+TEXT ·xorKeyStreamVX(SB), NOSPLIT, $0
+	MOVD	dst+0(FP), R1
+	MOVD	src+24(FP), R2
+	MOVD	src_len+32(FP), R3
+	MOVD	key+48(FP), R4
+	MOVD	nonce+56(FP), R6
+	MOVD	counter+64(FP), R7
+
+	MOVD	$·constants(SB), R10
+	MOVD	$·incRotMatrix(SB), R11
+
+	MOVW	(R7), R20
+
+	AND	$~255, R3, R13
+	ADD	R2, R13, R12 // R12 for block end
+	AND	$255, R3, R13
+loop:
+	MOVD	$NUM_ROUNDS, R21
+	VLD1	(R11), [V30.S4, V31.S4]
+
+	// load contants
+	// VLD4R (R10), [V0.S4, V1.S4, V2.S4, V3.S4]
+	WORD	$0x4D60E940
+
+	// load keys
+	// VLD4R 16(R4), [V4.S4, V5.S4, V6.S4, V7.S4]
+	WORD	$0x4DFFE884
+	// VLD4R 16(R4), [V8.S4, V9.S4, V10.S4, V11.S4]
+	WORD	$0x4DFFE888
+	SUB	$32, R4
+
+	// load counter + nonce
+	// VLD1R (R7), [V12.S4]
+	WORD	$0x4D40C8EC
+
+	// VLD3R (R6), [V13.S4, V14.S4, V15.S4]
+	WORD	$0x4D40E8CD
+
+	// update counter
+	VADD	V30.S4, V12.S4, V12.S4
+
+chacha:
+	// V0..V3 += V4..V7
+	// V12..V15 <<<= ((V12..V15 XOR V0..V3), 16)
+	VADD	V0.S4, V4.S4, V0.S4
+	VADD	V1.S4, V5.S4, V1.S4
+	VADD	V2.S4, V6.S4, V2.S4
+	VADD	V3.S4, V7.S4, V3.S4
+	VEOR	V12.B16, V0.B16, V12.B16
+	VEOR	V13.B16, V1.B16, V13.B16
+	VEOR	V14.B16, V2.B16, V14.B16
+	VEOR	V15.B16, V3.B16, V15.B16
+	VREV32	V12.H8, V12.H8
+	VREV32	V13.H8, V13.H8
+	VREV32	V14.H8, V14.H8
+	VREV32	V15.H8, V15.H8
+	// V8..V11 += V12..V15
+	// V4..V7 <<<= ((V4..V7 XOR V8..V11), 12)
+	VADD	V8.S4, V12.S4, V8.S4
+	VADD	V9.S4, V13.S4, V9.S4
+	VADD	V10.S4, V14.S4, V10.S4
+	VADD	V11.S4, V15.S4, V11.S4
+	VEOR	V8.B16, V4.B16, V16.B16
+	VEOR	V9.B16, V5.B16, V17.B16
+	VEOR	V10.B16, V6.B16, V18.B16
+	VEOR	V11.B16, V7.B16, V19.B16
+	VSHL	$12, V16.S4, V4.S4
+	VSHL	$12, V17.S4, V5.S4
+	VSHL	$12, V18.S4, V6.S4
+	VSHL	$12, V19.S4, V7.S4
+	VSRI	$20, V16.S4, V4.S4
+	VSRI	$20, V17.S4, V5.S4
+	VSRI	$20, V18.S4, V6.S4
+	VSRI	$20, V19.S4, V7.S4
+
+	// V0..V3 += V4..V7
+	// V12..V15 <<<= ((V12..V15 XOR V0..V3), 8)
+	VADD	V0.S4, V4.S4, V0.S4
+	VADD	V1.S4, V5.S4, V1.S4
+	VADD	V2.S4, V6.S4, V2.S4
+	VADD	V3.S4, V7.S4, V3.S4
+	VEOR	V12.B16, V0.B16, V12.B16
+	VEOR	V13.B16, V1.B16, V13.B16
+	VEOR	V14.B16, V2.B16, V14.B16
+	VEOR	V15.B16, V3.B16, V15.B16
+	VTBL	V31.B16, [V12.B16], V12.B16
+	VTBL	V31.B16, [V13.B16], V13.B16
+	VTBL	V31.B16, [V14.B16], V14.B16
+	VTBL	V31.B16, [V15.B16], V15.B16
+
+	// V8..V11 += V12..V15
+	// V4..V7 <<<= ((V4..V7 XOR V8..V11), 7)
+	VADD	V12.S4, V8.S4, V8.S4
+	VADD	V13.S4, V9.S4, V9.S4
+	VADD	V14.S4, V10.S4, V10.S4
+	VADD	V15.S4, V11.S4, V11.S4
+	VEOR	V8.B16, V4.B16, V16.B16
+	VEOR	V9.B16, V5.B16, V17.B16
+	VEOR	V10.B16, V6.B16, V18.B16
+	VEOR	V11.B16, V7.B16, V19.B16
+	VSHL	$7, V16.S4, V4.S4
+	VSHL	$7, V17.S4, V5.S4
+	VSHL	$7, V18.S4, V6.S4
+	VSHL	$7, V19.S4, V7.S4
+	VSRI	$25, V16.S4, V4.S4
+	VSRI	$25, V17.S4, V5.S4
+	VSRI	$25, V18.S4, V6.S4
+	VSRI	$25, V19.S4, V7.S4
+
+	// V0..V3 += V5..V7, V4
+	// V15,V12-V14 <<<= ((V15,V12-V14 XOR V0..V3), 16)
+	VADD	V0.S4, V5.S4, V0.S4
+	VADD	V1.S4, V6.S4, V1.S4
+	VADD	V2.S4, V7.S4, V2.S4
+	VADD	V3.S4, V4.S4, V3.S4
+	VEOR	V15.B16, V0.B16, V15.B16
+	VEOR	V12.B16, V1.B16, V12.B16
+	VEOR	V13.B16, V2.B16, V13.B16
+	VEOR	V14.B16, V3.B16, V14.B16
+	VREV32	V12.H8, V12.H8
+	VREV32	V13.H8, V13.H8
+	VREV32	V14.H8, V14.H8
+	VREV32	V15.H8, V15.H8
+
+	// V10 += V15; V5 <<<= ((V10 XOR V5), 12)
+	// ...
+	VADD	V15.S4, V10.S4, V10.S4
+	VADD	V12.S4, V11.S4, V11.S4
+	VADD	V13.S4, V8.S4, V8.S4
+	VADD	V14.S4, V9.S4, V9.S4
+	VEOR	V10.B16, V5.B16, V16.B16
+	VEOR	V11.B16, V6.B16, V17.B16
+	VEOR	V8.B16, V7.B16, V18.B16
+	VEOR	V9.B16, V4.B16, V19.B16
+	VSHL	$12, V16.S4, V5.S4
+	VSHL	$12, V17.S4, V6.S4
+	VSHL	$12, V18.S4, V7.S4
+	VSHL	$12, V19.S4, V4.S4
+	VSRI	$20, V16.S4, V5.S4
+	VSRI	$20, V17.S4, V6.S4
+	VSRI	$20, V18.S4, V7.S4
+	VSRI	$20, V19.S4, V4.S4
+
+	// V0 += V5; V15 <<<= ((V0 XOR V15), 8)
+	// ...
+	VADD	V5.S4, V0.S4, V0.S4
+	VADD	V6.S4, V1.S4, V1.S4
+	VADD	V7.S4, V2.S4, V2.S4
+	VADD	V4.S4, V3.S4, V3.S4
+	VEOR	V0.B16, V15.B16, V15.B16
+	VEOR	V1.B16, V12.B16, V12.B16
+	VEOR	V2.B16, V13.B16, V13.B16
+	VEOR	V3.B16, V14.B16, V14.B16
+	VTBL	V31.B16, [V12.B16], V12.B16
+	VTBL	V31.B16, [V13.B16], V13.B16
+	VTBL	V31.B16, [V14.B16], V14.B16
+	VTBL	V31.B16, [V15.B16], V15.B16
+
+	// V10 += V15; V5 <<<= ((V10 XOR V5), 7)
+	// ...
+	VADD	V15.S4, V10.S4, V10.S4
+	VADD	V12.S4, V11.S4, V11.S4
+	VADD	V13.S4, V8.S4, V8.S4
+	VADD	V14.S4, V9.S4, V9.S4
+	VEOR	V10.B16, V5.B16, V16.B16
+	VEOR	V11.B16, V6.B16, V17.B16
+	VEOR	V8.B16, V7.B16, V18.B16
+	VEOR	V9.B16, V4.B16, V19.B16
+	VSHL	$7, V16.S4, V5.S4
+	VSHL	$7, V17.S4, V6.S4
+	VSHL	$7, V18.S4, V7.S4
+	VSHL	$7, V19.S4, V4.S4
+	VSRI	$25, V16.S4, V5.S4
+	VSRI	$25, V17.S4, V6.S4
+	VSRI	$25, V18.S4, V7.S4
+	VSRI	$25, V19.S4, V4.S4
+
+	SUB	$1, R21
+	CBNZ	R21, chacha
+
+	// VLD4R (R10), [V16.S4, V17.S4, V18.S4, V19.S4]
+	WORD	$0x4D60E950
+
+	// VLD4R 16(R4), [V20.S4, V21.S4, V22.S4, V23.S4]
+	WORD	$0x4DFFE894
+	VADD	V30.S4, V12.S4, V12.S4
+	VADD	V16.S4, V0.S4, V0.S4
+	VADD	V17.S4, V1.S4, V1.S4
+	VADD	V18.S4, V2.S4, V2.S4
+	VADD	V19.S4, V3.S4, V3.S4
+	// VLD4R 16(R4), [V24.S4, V25.S4, V26.S4, V27.S4]
+	WORD	$0x4DFFE898
+	// restore R4
+	SUB	$32, R4
+
+	// load counter + nonce
+	// VLD1R (R7), [V28.S4]
+	WORD	$0x4D40C8FC
+	// VLD3R (R6), [V29.S4, V30.S4, V31.S4]
+	WORD	$0x4D40E8DD
+
+	VADD	V20.S4, V4.S4, V4.S4
+	VADD	V21.S4, V5.S4, V5.S4
+	VADD	V22.S4, V6.S4, V6.S4
+	VADD	V23.S4, V7.S4, V7.S4
+	VADD	V24.S4, V8.S4, V8.S4
+	VADD	V25.S4, V9.S4, V9.S4
+	VADD	V26.S4, V10.S4, V10.S4
+	VADD	V27.S4, V11.S4, V11.S4
+	VADD	V28.S4, V12.S4, V12.S4
+	VADD	V29.S4, V13.S4, V13.S4
+	VADD	V30.S4, V14.S4, V14.S4
+	VADD	V31.S4, V15.S4, V15.S4
+
+	VZIP1	V1.S4, V0.S4, V16.S4
+	VZIP2	V1.S4, V0.S4, V17.S4
+	VZIP1	V3.S4, V2.S4, V18.S4
+	VZIP2	V3.S4, V2.S4, V19.S4
+	VZIP1	V5.S4, V4.S4, V20.S4
+	VZIP2	V5.S4, V4.S4, V21.S4
+	VZIP1	V7.S4, V6.S4, V22.S4
+	VZIP2	V7.S4, V6.S4, V23.S4
+	VZIP1	V9.S4, V8.S4, V24.S4
+	VZIP2	V9.S4, V8.S4, V25.S4
+	VZIP1	V11.S4, V10.S4, V26.S4
+	VZIP2	V11.S4, V10.S4, V27.S4
+	VZIP1	V13.S4, V12.S4, V28.S4
+	VZIP2	V13.S4, V12.S4, V29.S4
+	VZIP1	V15.S4, V14.S4, V30.S4
+	VZIP2	V15.S4, V14.S4, V31.S4
+	VZIP1	V18.D2, V16.D2, V0.D2
+	VZIP2	V18.D2, V16.D2, V4.D2
+	VZIP1	V19.D2, V17.D2, V8.D2
+	VZIP2	V19.D2, V17.D2, V12.D2
+	VLD1.P	64(R2), [V16.B16, V17.B16, V18.B16, V19.B16]
+
+	VZIP1	V22.D2, V20.D2, V1.D2
+	VZIP2	V22.D2, V20.D2, V5.D2
+	VZIP1	V23.D2, V21.D2, V9.D2
+	VZIP2	V23.D2, V21.D2, V13.D2
+	VLD1.P	64(R2), [V20.B16, V21.B16, V22.B16, V23.B16]
+	VZIP1	V26.D2, V24.D2, V2.D2
+	VZIP2	V26.D2, V24.D2, V6.D2
+	VZIP1	V27.D2, V25.D2, V10.D2
+	VZIP2	V27.D2, V25.D2, V14.D2
+	VLD1.P	64(R2), [V24.B16, V25.B16, V26.B16, V27.B16]
+	VZIP1	V30.D2, V28.D2, V3.D2
+	VZIP2	V30.D2, V28.D2, V7.D2
+	VZIP1	V31.D2, V29.D2, V11.D2
+	VZIP2	V31.D2, V29.D2, V15.D2
+	VLD1.P	64(R2), [V28.B16, V29.B16, V30.B16, V31.B16]
+	VEOR	V0.B16, V16.B16, V16.B16
+	VEOR	V1.B16, V17.B16, V17.B16
+	VEOR	V2.B16, V18.B16, V18.B16
+	VEOR	V3.B16, V19.B16, V19.B16
+	VST1.P	[V16.B16, V17.B16, V18.B16, V19.B16], 64(R1)
+	VEOR	V4.B16, V20.B16, V20.B16
+	VEOR	V5.B16, V21.B16, V21.B16
+	VEOR	V6.B16, V22.B16, V22.B16
+	VEOR	V7.B16, V23.B16, V23.B16
+	VST1.P	[V20.B16, V21.B16, V22.B16, V23.B16], 64(R1)
+	VEOR	V8.B16, V24.B16, V24.B16
+	VEOR	V9.B16, V25.B16, V25.B16
+	VEOR	V10.B16, V26.B16, V26.B16
+	VEOR	V11.B16, V27.B16, V27.B16
+	VST1.P	[V24.B16, V25.B16, V26.B16, V27.B16], 64(R1)
+	VEOR	V12.B16, V28.B16, V28.B16
+	VEOR	V13.B16, V29.B16, V29.B16
+	VEOR	V14.B16, V30.B16, V30.B16
+	VEOR	V15.B16, V31.B16, V31.B16
+	VST1.P	[V28.B16, V29.B16, V30.B16, V31.B16], 64(R1)
+
+	ADD	$4, R20
+	MOVW	R20, (R7) // update counter
+
+	CMP	R2, R12
+	BGT	loop
+
+	RET
+
+
+DATA	·constants+0x00(SB)/4, $0x61707865
+DATA	·constants+0x04(SB)/4, $0x3320646e
+DATA	·constants+0x08(SB)/4, $0x79622d32
+DATA	·constants+0x0c(SB)/4, $0x6b206574
+GLOBL	·constants(SB), NOPTR|RODATA, $32
+
+DATA	·incRotMatrix+0x00(SB)/4, $0x00000000
+DATA	·incRotMatrix+0x04(SB)/4, $0x00000001
+DATA	·incRotMatrix+0x08(SB)/4, $0x00000002
+DATA	·incRotMatrix+0x0c(SB)/4, $0x00000003
+DATA	·incRotMatrix+0x10(SB)/4, $0x02010003
+DATA	·incRotMatrix+0x14(SB)/4, $0x06050407
+DATA	·incRotMatrix+0x18(SB)/4, $0x0A09080B
+DATA	·incRotMatrix+0x1c(SB)/4, $0x0E0D0C0F
+GLOBL	·incRotMatrix(SB), NOPTR|RODATA, $32

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

@@ -0,0 +1,364 @@
+// 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.
+	buf [bufSize]byte
+	len int
+
+	// 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")
+	}
+
+	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
+}
+
+// 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)
+		src = src[len(keyStream):]
+		dst = dst[len(keyStream):]
+	}
+
+	const blocksPerBuf = bufSize / blockSize
+	numBufs := (uint64(len(src)) + bufSize - 1) / bufSize
+	if uint64(s.counter)+numBufs*blocksPerBuf >= 1<<32 {
+		panic("chacha20: counter overflow")
+	}
+
+	// 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.
+
+	rem := len(src) % bufSize
+	full := len(src) - rem
+
+	if full > 0 {
+		s.xorKeyStreamBlocks(dst[:full], src[:full])
+	}
+
+	// If we have a partial (multi-)block, pad it for xorKeyStreamBlocks, and
+	// keep the leftover keystream for the next XORKeyStream invocation.
+	if rem > 0 {
+		s.buf = [bufSize]byte{}
+		copy(s.buf[:], src[full:])
+		s.xorKeyStreamBlocks(s.buf[:], s.buf[:])
+		s.len = bufSize - copy(dst[full:], 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
+	}
+
+	for i := 0; i < len(src); i += blockSize {
+		// 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)
+		}
+
+		// Finally, add back the initial state to generate the key stream.
+		x0 += c0
+		x1 += c1
+		x2 += c2
+		x3 += c3
+		x4 += c4
+		x5 += c5
+		x6 += c6
+		x7 += c7
+		x8 += c8
+		x9 += c9
+		x10 += c10
+		x11 += c11
+		x12 += s.counter
+		x13 += c13
+		x14 += c14
+		x15 += c15
+
+		s.counter += 1
+		if s.counter == 0 {
+			panic("chacha20: internal error: counter overflow")
+		}
+
+		in, out := src[i:], dst[i:]
+		in, out = in[:blockSize], out[:blockSize] // bounds check elimination 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)
+	}
+}
+
+// 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

@@ -0,0 +1,13 @@
+// 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 !arm64,!s390x,!ppc64le arm64,!go1.11 gccgo appengine
+
+package chacha20
+
+const bufSize = blockSize
+
+func (s *Cipher) xorKeyStreamBlocks(dst, src []byte) {
+	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,!appengine
+
+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

@@ -0,0 +1,449 @@
+// 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.
+
+// Based on CRYPTOGAMS code with the following comment:
+// # ====================================================================
+// # Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+// # project. The module is, however, dual licensed under OpenSSL and
+// # CRYPTOGAMS licenses depending on where you obtain it. For further
+// # details see http://www.openssl.org/~appro/cryptogams/.
+// # ====================================================================
+
+// Code for the perl script that generates the ppc64 assembler
+// can be found in the cryptogams repository at the link below. It is based on
+// the original from openssl.
+
+// https://github.com/dot-asm/cryptogams/commit/a60f5b50ed908e91
+
+// The differences in this and the original implementation are
+// due to the calling conventions and initialization of constants.
+
+// +build !gccgo,!appengine
+
+#include "textflag.h"
+
+#define OUT  R3
+#define INP  R4
+#define LEN  R5
+#define KEY  R6
+#define CNT  R7
+#define TMP  R15
+
+#define CONSTBASE  R16
+#define BLOCKS R17
+
+DATA consts<>+0x00(SB)/8, $0x3320646e61707865
+DATA consts<>+0x08(SB)/8, $0x6b20657479622d32
+DATA consts<>+0x10(SB)/8, $0x0000000000000001
+DATA consts<>+0x18(SB)/8, $0x0000000000000000
+DATA consts<>+0x20(SB)/8, $0x0000000000000004
+DATA consts<>+0x28(SB)/8, $0x0000000000000000
+DATA consts<>+0x30(SB)/8, $0x0a0b08090e0f0c0d
+DATA consts<>+0x38(SB)/8, $0x0203000106070405
+DATA consts<>+0x40(SB)/8, $0x090a0b080d0e0f0c
+DATA consts<>+0x48(SB)/8, $0x0102030005060704
+DATA consts<>+0x50(SB)/8, $0x6170786561707865
+DATA consts<>+0x58(SB)/8, $0x6170786561707865
+DATA consts<>+0x60(SB)/8, $0x3320646e3320646e
+DATA consts<>+0x68(SB)/8, $0x3320646e3320646e
+DATA consts<>+0x70(SB)/8, $0x79622d3279622d32
+DATA consts<>+0x78(SB)/8, $0x79622d3279622d32
+DATA consts<>+0x80(SB)/8, $0x6b2065746b206574
+DATA consts<>+0x88(SB)/8, $0x6b2065746b206574
+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 *[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 counter+32(FP), CNT
+
+	// Addressing for constants
+	MOVD $consts<>+0x00(SB), CONSTBASE
+	MOVD $16, R8
+	MOVD $32, R9
+	MOVD $48, R10
+	MOVD $64, R11
+	SRD $6, LEN, BLOCKS
+	// V16
+	LXVW4X (CONSTBASE)(R0), VS48
+	ADD $80,CONSTBASE
+
+	// Load key into V17,V18
+	LXVW4X (KEY)(R0), VS49
+	LXVW4X (KEY)(R8), VS50
+
+	// Load CNT, NONCE into V19
+	LXVW4X (CNT)(R0), VS51
+
+	// Clear V27
+	VXOR V27, V27, V27
+
+	// V28
+	LXVW4X (CONSTBASE)(R11), VS60
+
+	// splat slot from V19 -> V26
+	VSPLTW $0, V19, V26
+
+	VSLDOI $4, V19, V27, V19
+	VSLDOI $12, V27, V19, V19
+
+	VADDUWM V26, V28, V26
+
+	MOVD $10, R14
+	MOVD R14, CTR
+
+loop_outer_vsx:
+	// V0, V1, V2, V3
+	LXVW4X (R0)(CONSTBASE), VS32
+	LXVW4X (R8)(CONSTBASE), VS33
+	LXVW4X (R9)(CONSTBASE), VS34
+	LXVW4X (R10)(CONSTBASE), VS35
+
+	// splat values from V17, V18 into V4-V11
+	VSPLTW $0, V17, V4
+	VSPLTW $1, V17, V5
+	VSPLTW $2, V17, V6
+	VSPLTW $3, V17, V7
+	VSPLTW $0, V18, V8
+	VSPLTW $1, V18, V9
+	VSPLTW $2, V18, V10
+	VSPLTW $3, V18, V11
+
+	// VOR
+	VOR V26, V26, V12
+
+	// splat values from V19 -> V13, V14, V15
+	VSPLTW $1, V19, V13
+	VSPLTW $2, V19, V14
+	VSPLTW $3, V19, V15
+
+	// splat   const values
+	VSPLTISW $-16, V27
+	VSPLTISW $12, V28
+	VSPLTISW $8, V29
+	VSPLTISW $7, V30
+
+loop_vsx:
+	VADDUWM V0, V4, V0
+	VADDUWM V1, V5, V1
+	VADDUWM V2, V6, V2
+	VADDUWM V3, V7, V3
+
+	VXOR V12, V0, V12
+	VXOR V13, V1, V13
+	VXOR V14, V2, V14
+	VXOR V15, V3, V15
+
+	VRLW V12, V27, V12
+	VRLW V13, V27, V13
+	VRLW V14, V27, V14
+	VRLW V15, V27, V15
+
+	VADDUWM V8, V12, V8
+	VADDUWM V9, V13, V9
+	VADDUWM V10, V14, V10
+	VADDUWM V11, V15, V11
+
+	VXOR V4, V8, V4
+	VXOR V5, V9, V5
+	VXOR V6, V10, V6
+	VXOR V7, V11, V7
+
+	VRLW V4, V28, V4
+	VRLW V5, V28, V5
+	VRLW V6, V28, V6
+	VRLW V7, V28, V7
+
+	VADDUWM V0, V4, V0
+	VADDUWM V1, V5, V1
+	VADDUWM V2, V6, V2
+	VADDUWM V3, V7, V3
+
+	VXOR V12, V0, V12
+	VXOR V13, V1, V13
+	VXOR V14, V2, V14
+	VXOR V15, V3, V15
+
+	VRLW V12, V29, V12
+	VRLW V13, V29, V13
+	VRLW V14, V29, V14
+	VRLW V15, V29, V15
+
+	VADDUWM V8, V12, V8
+	VADDUWM V9, V13, V9
+	VADDUWM V10, V14, V10
+	VADDUWM V11, V15, V11
+
+	VXOR V4, V8, V4
+	VXOR V5, V9, V5
+	VXOR V6, V10, V6
+	VXOR V7, V11, V7
+
+	VRLW V4, V30, V4
+	VRLW V5, V30, V5
+	VRLW V6, V30, V6
+	VRLW V7, V30, V7
+
+	VADDUWM V0, V5, V0
+	VADDUWM V1, V6, V1
+	VADDUWM V2, V7, V2
+	VADDUWM V3, V4, V3
+
+	VXOR V15, V0, V15
+	VXOR V12, V1, V12
+	VXOR V13, V2, V13
+	VXOR V14, V3, V14
+
+	VRLW V15, V27, V15
+	VRLW V12, V27, V12
+	VRLW V13, V27, V13
+	VRLW V14, V27, V14
+
+	VADDUWM V10, V15, V10
+	VADDUWM V11, V12, V11
+	VADDUWM V8, V13, V8
+	VADDUWM V9, V14, V9
+
+	VXOR V5, V10, V5
+	VXOR V6, V11, V6
+	VXOR V7, V8, V7
+	VXOR V4, V9, V4
+
+	VRLW V5, V28, V5
+	VRLW V6, V28, V6
+	VRLW V7, V28, V7
+	VRLW V4, V28, V4
+
+	VADDUWM V0, V5, V0
+	VADDUWM V1, V6, V1
+	VADDUWM V2, V7, V2
+	VADDUWM V3, V4, V3
+
+	VXOR V15, V0, V15
+	VXOR V12, V1, V12
+	VXOR V13, V2, V13
+	VXOR V14, V3, V14
+
+	VRLW V15, V29, V15
+	VRLW V12, V29, V12
+	VRLW V13, V29, V13
+	VRLW V14, V29, V14
+
+	VADDUWM V10, V15, V10
+	VADDUWM V11, V12, V11
+	VADDUWM V8, V13, V8
+	VADDUWM V9, V14, V9
+
+	VXOR V5, V10, V5
+	VXOR V6, V11, V6
+	VXOR V7, V8, V7
+	VXOR V4, V9, V4
+
+	VRLW V5, V30, V5
+	VRLW V6, V30, V6
+	VRLW V7, V30, V7
+	VRLW V4, V30, V4
+	BC   16, LT, loop_vsx
+
+	VADDUWM V12, V26, V12
+
+	WORD $0x13600F8C		// VMRGEW V0, V1, V27
+	WORD $0x13821F8C		// VMRGEW V2, V3, V28
+
+	WORD $0x10000E8C		// VMRGOW V0, V1, V0
+	WORD $0x10421E8C		// VMRGOW V2, V3, V2
+
+	WORD $0x13A42F8C		// VMRGEW V4, V5, V29
+	WORD $0x13C63F8C		// VMRGEW V6, V7, V30
+
+	XXPERMDI VS32, VS34, $0, VS33
+	XXPERMDI VS32, VS34, $3, VS35
+	XXPERMDI VS59, VS60, $0, VS32
+	XXPERMDI VS59, VS60, $3, VS34
+
+	WORD $0x10842E8C		// VMRGOW V4, V5, V4
+	WORD $0x10C63E8C		// VMRGOW V6, V7, V6
+
+	WORD $0x13684F8C		// VMRGEW V8, V9, V27
+	WORD $0x138A5F8C		// VMRGEW V10, V11, V28
+
+	XXPERMDI VS36, VS38, $0, VS37
+	XXPERMDI VS36, VS38, $3, VS39
+	XXPERMDI VS61, VS62, $0, VS36
+	XXPERMDI VS61, VS62, $3, VS38
+
+	WORD $0x11084E8C		// VMRGOW V8, V9, V8
+	WORD $0x114A5E8C		// VMRGOW V10, V11, V10
+
+	WORD $0x13AC6F8C		// VMRGEW V12, V13, V29
+	WORD $0x13CE7F8C		// VMRGEW V14, V15, V30
+
+	XXPERMDI VS40, VS42, $0, VS41
+	XXPERMDI VS40, VS42, $3, VS43
+	XXPERMDI VS59, VS60, $0, VS40
+	XXPERMDI VS59, VS60, $3, VS42
+
+	WORD $0x118C6E8C		// VMRGOW V12, V13, V12
+	WORD $0x11CE7E8C		// VMRGOW V14, V15, V14
+
+	VSPLTISW $4, V27
+	VADDUWM V26, V27, V26
+
+	XXPERMDI VS44, VS46, $0, VS45
+	XXPERMDI VS44, VS46, $3, VS47
+	XXPERMDI VS61, VS62, $0, VS44
+	XXPERMDI VS61, VS62, $3, VS46
+
+	VADDUWM V0, V16, V0
+	VADDUWM V4, V17, V4
+	VADDUWM V8, V18, V8
+	VADDUWM V12, V19, V12
+
+	CMPU LEN, $64
+	BLT tail_vsx
+
+	// Bottom of loop
+	LXVW4X (INP)(R0), VS59
+	LXVW4X (INP)(R8), VS60
+	LXVW4X (INP)(R9), VS61
+	LXVW4X (INP)(R10), VS62
+
+	VXOR V27, V0, V27
+	VXOR V28, V4, V28
+	VXOR V29, V8, V29
+	VXOR V30, V12, V30
+
+	STXVW4X VS59, (OUT)(R0)
+	STXVW4X VS60, (OUT)(R8)
+	ADD     $64, INP
+	STXVW4X VS61, (OUT)(R9)
+	ADD     $-64, LEN
+	STXVW4X VS62, (OUT)(R10)
+	ADD     $64, OUT
+	BEQ     done_vsx
+
+	VADDUWM V1, V16, V0
+	VADDUWM V5, V17, V4
+	VADDUWM V9, V18, V8
+	VADDUWM V13, V19, V12
+
+	CMPU  LEN, $64
+	BLT   tail_vsx
+
+	LXVW4X (INP)(R0), VS59
+	LXVW4X (INP)(R8), VS60
+	LXVW4X (INP)(R9), VS61
+	LXVW4X (INP)(R10), VS62
+	VXOR   V27, V0, V27
+
+	VXOR V28, V4, V28
+	VXOR V29, V8, V29
+	VXOR V30, V12, V30
+
+	STXVW4X VS59, (OUT)(R0)
+	STXVW4X VS60, (OUT)(R8)
+	ADD     $64, INP
+	STXVW4X VS61, (OUT)(R9)
+	ADD     $-64, LEN
+	STXVW4X VS62, (OUT)(V10)
+	ADD     $64, OUT
+	BEQ     done_vsx
+
+	VADDUWM V2, V16, V0
+	VADDUWM V6, V17, V4
+	VADDUWM V10, V18, V8
+	VADDUWM V14, V19, V12
+
+	CMPU LEN, $64
+	BLT  tail_vsx
+
+	LXVW4X (INP)(R0), VS59
+	LXVW4X (INP)(R8), VS60
+	LXVW4X (INP)(R9), VS61
+	LXVW4X (INP)(R10), VS62
+
+	VXOR V27, V0, V27
+	VXOR V28, V4, V28
+	VXOR V29, V8, V29
+	VXOR V30, V12, V30
+
+	STXVW4X VS59, (OUT)(R0)
+	STXVW4X VS60, (OUT)(R8)
+	ADD     $64, INP
+	STXVW4X VS61, (OUT)(R9)
+	ADD     $-64, LEN
+	STXVW4X VS62, (OUT)(R10)
+	ADD     $64, OUT
+	BEQ     done_vsx
+
+	VADDUWM V3, V16, V0
+	VADDUWM V7, V17, V4
+	VADDUWM V11, V18, V8
+	VADDUWM V15, V19, V12
+
+	CMPU  LEN, $64
+	BLT   tail_vsx
+
+	LXVW4X (INP)(R0), VS59
+	LXVW4X (INP)(R8), VS60
+	LXVW4X (INP)(R9), VS61
+	LXVW4X (INP)(R10), VS62
+
+	VXOR V27, V0, V27
+	VXOR V28, V4, V28
+	VXOR V29, V8, V29
+	VXOR V30, V12, V30
+
+	STXVW4X VS59, (OUT)(R0)
+	STXVW4X VS60, (OUT)(R8)
+	ADD     $64, INP
+	STXVW4X VS61, (OUT)(R9)
+	ADD     $-64, LEN
+	STXVW4X VS62, (OUT)(R10)
+	ADD     $64, OUT
+
+	MOVD $10, R14
+	MOVD R14, CTR
+	BNE  loop_outer_vsx
+
+done_vsx:
+	// Increment counter by number of 64 byte blocks
+	MOVD (CNT), R14
+	ADD  BLOCKS, R14
+	MOVD R14, (CNT)
+	RET
+
+tail_vsx:
+	ADD  $32, R1, R11
+	MOVD LEN, CTR
+
+	// Save values on stack to copy from
+	STXVW4X VS32, (R11)(R0)
+	STXVW4X VS36, (R11)(R8)
+	STXVW4X VS40, (R11)(R9)
+	STXVW4X VS44, (R11)(R10)
+	ADD $-1, R11, R12
+	ADD $-1, INP
+	ADD $-1, OUT
+
+looptail_vsx:
+	// Copying the result to OUT
+	// in bytes.
+	MOVBZU 1(R12), KEY
+	MOVBZU 1(INP), TMP
+	XOR    KEY, TMP, KEY
+	MOVBU  KEY, 1(OUT)
+	BC     16, LT, looptail_vsx
+
+	// Clear the stack values
+	STXVW4X VS48, (R11)(R0)
+	STXVW4X VS48, (R11)(R8)
+	STXVW4X VS48, (R11)(R9)
+	STXVW4X VS48, (R11)(R10)
+	BR      done_vsx

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,!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)
+
+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

@@ -0,0 +1,224 @@
+// 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,!appengine
+
+#include "go_asm.h"
+#include "textflag.h"
+
+// This is an implementation of the ChaCha20 encryption algorithm as
+// specified in RFC 7539. It uses vector instructions to compute
+// 4 keystream blocks in parallel (256 bytes) which are then XORed
+// with the bytes in the input slice.
+
+GLOBL ·constants<>(SB), RODATA|NOPTR, $32
+// BSWAP: swap bytes in each 4-byte element
+DATA ·constants<>+0x00(SB)/4, $0x03020100
+DATA ·constants<>+0x04(SB)/4, $0x07060504
+DATA ·constants<>+0x08(SB)/4, $0x0b0a0908
+DATA ·constants<>+0x0c(SB)/4, $0x0f0e0d0c
+// J0: [j0, j1, j2, j3]
+DATA ·constants<>+0x10(SB)/4, $0x61707865
+DATA ·constants<>+0x14(SB)/4, $0x3320646e
+DATA ·constants<>+0x18(SB)/4, $0x79622d32
+DATA ·constants<>+0x1c(SB)/4, $0x6b206574
+
+#define BSWAP V5
+#define J0    V6
+#define KEY0  V7
+#define KEY1  V8
+#define NONCE V9
+#define CTR   V10
+#define M0    V11
+#define M1    V12
+#define M2    V13
+#define M3    V14
+#define INC   V15
+#define X0    V16
+#define X1    V17
+#define X2    V18
+#define X3    V19
+#define X4    V20
+#define X5    V21
+#define X6    V22
+#define X7    V23
+#define X8    V24
+#define X9    V25
+#define X10   V26
+#define X11   V27
+#define X12   V28
+#define X13   V29
+#define X14   V30
+#define X15   V31
+
+#define NUM_ROUNDS 20
+
+#define ROUND4(a0, a1, a2, a3, b0, b1, b2, b3, c0, c1, c2, c3, d0, d1, d2, d3) \
+	VAF    a1, a0, a0  \
+	VAF    b1, b0, b0  \
+	VAF    c1, c0, c0  \
+	VAF    d1, d0, d0  \
+	VX     a0, a2, a2  \
+	VX     b0, b2, b2  \
+	VX     c0, c2, c2  \
+	VX     d0, d2, d2  \
+	VERLLF $16, a2, a2 \
+	VERLLF $16, b2, b2 \
+	VERLLF $16, c2, c2 \
+	VERLLF $16, d2, d2 \
+	VAF    a2, a3, a3  \
+	VAF    b2, b3, b3  \
+	VAF    c2, c3, c3  \
+	VAF    d2, d3, d3  \
+	VX     a3, a1, a1  \
+	VX     b3, b1, b1  \
+	VX     c3, c1, c1  \
+	VX     d3, d1, d1  \
+	VERLLF $12, a1, a1 \
+	VERLLF $12, b1, b1 \
+	VERLLF $12, c1, c1 \
+	VERLLF $12, d1, d1 \
+	VAF    a1, a0, a0  \
+	VAF    b1, b0, b0  \
+	VAF    c1, c0, c0  \
+	VAF    d1, d0, d0  \
+	VX     a0, a2, a2  \
+	VX     b0, b2, b2  \
+	VX     c0, c2, c2  \
+	VX     d0, d2, d2  \
+	VERLLF $8, a2, a2  \
+	VERLLF $8, b2, b2  \
+	VERLLF $8, c2, c2  \
+	VERLLF $8, d2, d2  \
+	VAF    a2, a3, a3  \
+	VAF    b2, b3, b3  \
+	VAF    c2, c3, c3  \
+	VAF    d2, d3, d3  \
+	VX     a3, a1, a1  \
+	VX     b3, b1, b1  \
+	VX     c3, c1, c1  \
+	VX     d3, d1, d1  \
+	VERLLF $7, a1, a1  \
+	VERLLF $7, b1, b1  \
+	VERLLF $7, c1, c1  \
+	VERLLF $7, d1, d1
+
+#define PERMUTE(mask, v0, v1, v2, v3) \
+	VPERM v0, v0, mask, v0 \
+	VPERM v1, v1, mask, v1 \
+	VPERM v2, v2, mask, v2 \
+	VPERM v3, v3, mask, v3
+
+#define ADDV(x, v0, v1, v2, v3) \
+	VAF x, v0, v0 \
+	VAF x, v1, v1 \
+	VAF x, v2, v2 \
+	VAF x, v3, v3
+
+#define XORV(off, dst, src, v0, v1, v2, v3) \
+	VLM  off(src), M0, M3          \
+	PERMUTE(BSWAP, v0, v1, v2, v3) \
+	VX   v0, M0, M0                \
+	VX   v1, M1, M1                \
+	VX   v2, M2, M2                \
+	VX   v3, M3, M3                \
+	VSTM M0, M3, off(dst)
+
+#define SHUFFLE(a, b, c, d, t, u, v, w) \
+	VMRHF a, c, t \ // t = {a[0], c[0], a[1], c[1]}
+	VMRHF b, d, u \ // u = {b[0], d[0], b[1], d[1]}
+	VMRLF a, c, v \ // v = {a[2], c[2], a[3], c[3]}
+	VMRLF b, d, w \ // w = {b[2], d[2], b[3], d[3]}
+	VMRHF t, u, a \ // a = {a[0], b[0], c[0], d[0]}
+	VMRLF t, u, b \ // b = {a[1], b[1], c[1], d[1]}
+	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)
+TEXT ·xorKeyStreamVX(SB), NOSPLIT, $0
+	MOVD $·constants<>(SB), R1
+	MOVD dst+0(FP), R2         // R2=&dst[0]
+	LMG  src+24(FP), R3, R4    // R3=&src[0] R4=len(src)
+	MOVD key+48(FP), R5        // R5=key
+	MOVD nonce+56(FP), R6      // R6=nonce
+	MOVD counter+64(FP), R7    // R7=counter
+
+	// load BSWAP and J0
+	VLM (R1), BSWAP, J0
+
+	// setup
+	MOVD  $95, R0
+	VLM   (R5), KEY0, KEY1
+	VLL   R0, (R6), NONCE
+	VZERO M0
+	VLEIB $7, $32, M0
+	VSRLB M0, NONCE, NONCE
+
+	// initialize counter values
+	VLREPF (R7), CTR
+	VZERO  INC
+	VLEIF  $1, $1, INC
+	VLEIF  $2, $2, INC
+	VLEIF  $3, $3, INC
+	VAF    INC, CTR, CTR
+	VREPIF $4, INC
+
+chacha:
+	VREPF $0, J0, X0
+	VREPF $1, J0, X1
+	VREPF $2, J0, X2
+	VREPF $3, J0, X3
+	VREPF $0, KEY0, X4
+	VREPF $1, KEY0, X5
+	VREPF $2, KEY0, X6
+	VREPF $3, KEY0, X7
+	VREPF $0, KEY1, X8
+	VREPF $1, KEY1, X9
+	VREPF $2, KEY1, X10
+	VREPF $3, KEY1, X11
+	VLR   CTR, X12
+	VREPF $1, NONCE, X13
+	VREPF $2, NONCE, X14
+	VREPF $3, NONCE, X15
+
+	MOVD $(NUM_ROUNDS/2), R1
+
+loop:
+	ROUND4(X0, X4, X12,  X8, X1, X5, X13,  X9, X2, X6, X14, X10, X3, X7, X15, X11)
+	ROUND4(X0, X5, X15, X10, X1, X6, X12, X11, X2, X7, X13, X8,  X3, X4, X14, X9)
+
+	ADD $-1, R1
+	BNE loop
+
+	// decrement length
+	ADD $-256, R4
+
+	// rearrange vectors
+	SHUFFLE(X0, X1, X2, X3, M0, M1, M2, M3)
+	ADDV(J0, X0, X1, X2, X3)
+	SHUFFLE(X4, X5, X6, X7, M0, M1, M2, M3)
+	ADDV(KEY0, X4, X5, X6, X7)
+	SHUFFLE(X8, X9, X10, X11, M0, M1, M2, M3)
+	ADDV(KEY1, X8, X9, X10, X11)
+	VAF CTR, X12, X12
+	SHUFFLE(X12, X13, X14, X15, M0, M1, M2, M3)
+	ADDV(NONCE, X12, X13, X14, X15)
+
+	// increment counters
+	VAF INC, CTR, CTR
+
+	// xor keystream with plaintext
+	XORV(0*64, R2, R3, X0, X4,  X8, X12)
+	XORV(1*64, R2, R3, X1, X5,  X9, X13)
+	XORV(2*64, R2, R3, X2, X6, X10, X14)
+	XORV(3*64, R2, R3, X3, X7, X11, X15)
+
+	// increment pointers
+	MOVD $256(R2), R2
+	MOVD $256(R3), R3
+
+	CMPBNE  R4, $0, chacha
+
+	VSTEF $0, CTR, (R7)
+	RET

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

@@ -0,0 +1,41 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found src the LICENSE file.
+
+package chacha20
+
+import "runtime"
+
+// Platforms that have fast unaligned 32-bit little endian accesses.
+const unaligned = runtime.GOARCH == "386" ||
+	runtime.GOARCH == "amd64" ||
+	runtime.GOARCH == "arm64" ||
+	runtime.GOARCH == "ppc64le" ||
+	runtime.GOARCH == "s390x"
+
+// xor reads a little endian uint32 from src, XORs it with u and
+// places the result in little endian byte order in dst.
+func xor(dst, src []byte, u uint32) {
+	_, _ = src[3], dst[3] // eliminate bounds checks
+	if unaligned {
+		// The compiler should optimize this code into
+		// 32-bit unaligned little endian loads and stores.
+		// TODO: delete once the compiler does a reliably
+		// good job with the generic code below.
+		// See issue #25111 for more details.
+		v := uint32(src[0])
+		v |= uint32(src[1]) << 8
+		v |= uint32(src[2]) << 16
+		v |= uint32(src[3]) << 24
+		v ^= u
+		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)
+		dst[1] = src[1] ^ byte(u>>8)
+		dst[2] = src[2] ^ byte(u>>16)
+		dst[3] = src[3] ^ byte(u>>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
-// non-amd64 platforms. The amd64 assembly does not support gccgo.
-// +build !amd64 gccgo appengine
-
-package curve25519
+// Package curve25519 provides an implementation of the X25519 function, which
+// performs scalar multiplication on the elliptic curve known as Curve25519.
+// See RFC 7748.
+package curve25519 // import "golang.org/x/crypto/curve25519"
 
 import (
-	"encoding/binary"
+	"crypto/subtle"
+	"fmt"
 )
 
-// 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.
+// ScalarMult sets dst to the product scalar * point.
 //
-//   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)
+// 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)
 }
 
-// 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.
+// ScalarBaseMult sets dst to the product scalar * base where base is the
+// standard generator.
 //
-// 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)
+// It is recommended to use the X25519 function with Basepoint instead, as
+// copying into fixed size arrays can lead to unexpected bugs.
+func ScalarBaseMult(dst, scalar *[32]byte) {
+	ScalarMult(dst, scalar, &basePoint)
 }
 
-// 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
+const (
+	// ScalarSize is the size of the scalar input to X25519.
+	ScalarSize = 32
+	// PointSize is the size of the point input to X25519.
+	PointSize = 32
+)
 
-	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
+// Basepoint is the canonical Curve25519 generator.
+var Basepoint []byte
 
-	carry[9] = (h9 + (1 << 24)) >> 25
-	h0 += carry[9] * 19
-	h9 -= carry[9] << 25
+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}
 
-	carry[0] = (h0 + (1 << 25)) >> 26
-	h1 += carry[0]
-	h0 -= carry[0] << 26
+func init() { Basepoint = basePoint[:] }
 
-	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)
+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")
+	}
 }
 
-// feMul121666 calculates h = f * 121666. Can overlap h with f.
+// X25519 returns the result of the scalar multiplication (scalar * point),
+// according to RFC 7748, Section 5. scalar, point and the return value are
+// slices of 32 bytes.
 //
-// Preconditions:
-//    |f| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
+// scalar can be generated at random, for example with crypto/rand. point should
+// be either Basepoint or the output of another X25519 call.
 //
-// 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)
+// 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 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)
+func x25519(dst *[32]byte, scalar, point []byte) ([]byte, error) {
+	var in [32]byte
+	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)
+	}
+	copy(in[:], scalar)
+	if &point[0] == &Basepoint[0] {
+		checkBasepoint()
+		ScalarBaseMult(dst, &in)
+	} else {
+		var base, zero [32]byte
+		copy(base[:], point)
+		ScalarMult(dst, &in, &base)
+		if subtle.ConstantTimeCompare(dst[:], zero[:]) == 1 {
+			return nil, fmt.Errorf("bad input point: low order point")
+		}
+	}
+	return dst[:], nil
 }

vendor/golang.org/x/crypto/curve25519/mont25519_amd64.go → 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/ladderstep_amd64.s → 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
@@ -121,18 +196,18 @@ TEXT ·ladderstep(SB),0,$296-8
 	ADDQ AX,R12
 	ADCQ DX,R13
 	MOVQ $REDMASK51,DX
-	SHLQ $13,CX:SI
+	SHLQ $13,SI,CX
 	ANDQ DX,SI
-	SHLQ $13,R9:R8
+	SHLQ $13,R8,R9
 	ANDQ DX,R8
 	ADDQ CX,R8
-	SHLQ $13,R11:R10
+	SHLQ $13,R10,R11
 	ANDQ DX,R10
 	ADDQ R9,R10
-	SHLQ $13,R13:R12
+	SHLQ $13,R12,R13
 	ANDQ DX,R12
 	ADDQ R11,R12
-	SHLQ $13,R15:R14
+	SHLQ $13,R14,R15
 	ANDQ DX,R14
 	ADDQ R13,R14
 	IMUL3Q $19,R15,CX
@@ -236,18 +311,18 @@ TEXT ·ladderstep(SB),0,$296-8
 	ADDQ AX,R12
 	ADCQ DX,R13
 	MOVQ $REDMASK51,DX
-	SHLQ $13,CX:SI
+	SHLQ $13,SI,CX
 	ANDQ DX,SI
-	SHLQ $13,R9:R8
+	SHLQ $13,R8,R9
 	ANDQ DX,R8
 	ADDQ CX,R8
-	SHLQ $13,R11:R10
+	SHLQ $13,R10,R11
 	ANDQ DX,R10
 	ADDQ R9,R10
-	SHLQ $13,R13:R12
+	SHLQ $13,R12,R13
 	ANDQ DX,R12
 	ADDQ R11,R12
-	SHLQ $13,R15:R14
+	SHLQ $13,R14,R15
 	ANDQ DX,R14
 	ADDQ R13,R14
 	IMUL3Q $19,R15,CX
@@ -441,18 +516,18 @@ TEXT ·ladderstep(SB),0,$296-8
 	ADDQ AX,R12
 	ADCQ DX,R13
 	MOVQ $REDMASK51,DX
-	SHLQ $13,CX:SI
+	SHLQ $13,SI,CX
 	ANDQ DX,SI
-	SHLQ $13,R9:R8
+	SHLQ $13,R8,R9
 	ANDQ DX,R8
 	ADDQ CX,R8
-	SHLQ $13,R11:R10
+	SHLQ $13,R10,R11
 	ANDQ DX,R10
 	ADDQ R9,R10
-	SHLQ $13,R13:R12
+	SHLQ $13,R12,R13
 	ANDQ DX,R12
 	ADDQ R11,R12
-	SHLQ $13,R15:R14
+	SHLQ $13,R14,R15
 	ANDQ DX,R14
 	ADDQ R13,R14
 	IMUL3Q $19,R15,CX
@@ -591,18 +666,18 @@ TEXT ·ladderstep(SB),0,$296-8
 	ADDQ AX,R12
 	ADCQ DX,R13
 	MOVQ $REDMASK51,DX
-	SHLQ $13,CX:SI
+	SHLQ $13,SI,CX
 	ANDQ DX,SI
-	SHLQ $13,R9:R8
+	SHLQ $13,R8,R9
 	ANDQ DX,R8
 	ADDQ CX,R8
-	SHLQ $13,R11:R10
+	SHLQ $13,R10,R11
 	ANDQ DX,R10
 	ADDQ R9,R10
-	SHLQ $13,R13:R12
+	SHLQ $13,R12,R13
 	ANDQ DX,R12
 	ADDQ R11,R12
-	SHLQ $13,R15:R14
+	SHLQ $13,R14,R15
 	ANDQ DX,R14
 	ADDQ R13,R14
 	IMUL3Q $19,R15,CX
@@ -731,18 +806,18 @@ TEXT ·ladderstep(SB),0,$296-8
 	ADDQ AX,R12
 	ADCQ DX,R13
 	MOVQ $REDMASK51,DX
-	SHLQ $13,CX:SI
+	SHLQ $13,SI,CX
 	ANDQ DX,SI
-	SHLQ $13,R9:R8
+	SHLQ $13,R8,R9
 	ANDQ DX,R8
 	ADDQ CX,R8
-	SHLQ $13,R11:R10
+	SHLQ $13,R10,R11
 	ANDQ DX,R10
 	ADDQ R9,R10
-	SHLQ $13,R13:R12
+	SHLQ $13,R12,R13
 	ANDQ DX,R12
 	ADDQ R11,R12
-	SHLQ $13,R15:R14
+	SHLQ $13,R14,R15
 	ANDQ DX,R14
 	ADDQ R13,R14
 	IMUL3Q $19,R15,CX
@@ -846,18 +921,18 @@ TEXT ·ladderstep(SB),0,$296-8
 	ADDQ AX,R12
 	ADCQ DX,R13
 	MOVQ $REDMASK51,DX
-	SHLQ $13,CX:SI
+	SHLQ $13,SI,CX
 	ANDQ DX,SI
-	SHLQ $13,R9:R8
+	SHLQ $13,R8,R9
 	ANDQ DX,R8
 	ADDQ CX,R8
-	SHLQ $13,R11:R10
+	SHLQ $13,R10,R11
 	ANDQ DX,R10
 	ADDQ R9,R10
-	SHLQ $13,R13:R12
+	SHLQ $13,R12,R13
 	ANDQ DX,R12
 	ADDQ R11,R12
-	SHLQ $13,R15:R14
+	SHLQ $13,R14,R15
 	ANDQ DX,R14
 	ADDQ R13,R14
 	IMUL3Q $19,R15,CX
@@ -996,18 +1071,18 @@ TEXT ·ladderstep(SB),0,$296-8
 	ADDQ AX,R12
 	ADCQ DX,R13
 	MOVQ $REDMASK51,DX
-	SHLQ $13,CX:SI
+	SHLQ $13,SI,CX
 	ANDQ DX,SI
-	SHLQ $13,R9:R8
+	SHLQ $13,R8,R9
 	ANDQ DX,R8
 	ADDQ CX,R8
-	SHLQ $13,R11:R10
+	SHLQ $13,R10,R11
 	ANDQ DX,R10
 	ADDQ R9,R10
-	SHLQ $13,R13:R12
+	SHLQ $13,R12,R13
 	ANDQ DX,R12
 	ADDQ R11,R12
-	SHLQ $13,R15:R14
+	SHLQ $13,R14,R15
 	ANDQ DX,R14
 	ADDQ R13,R14
 	IMUL3Q $19,R15,CX
@@ -1146,18 +1221,18 @@ TEXT ·ladderstep(SB),0,$296-8
 	ADDQ AX,R12
 	ADCQ DX,R13
 	MOVQ $REDMASK51,DX
-	SHLQ $13,CX:SI
+	SHLQ $13,SI,CX
 	ANDQ DX,SI
-	SHLQ $13,R9:R8
+	SHLQ $13,R8,R9
 	ANDQ DX,R8
 	ADDQ CX,R8
-	SHLQ $13,R11:R10
+	SHLQ $13,R10,R11
 	ANDQ DX,R10
 	ADDQ R9,R10
-	SHLQ $13,R13:R12
+	SHLQ $13,R12,R13
 	ANDQ DX,R12
 	ADDQ R11,R12
-	SHLQ $13,R15:R14
+	SHLQ $13,R14,R15
 	ANDQ DX,R14
 	ADDQ R13,R14
 	IMUL3Q $19,R15,CX
@@ -1332,18 +1407,18 @@ TEXT ·ladderstep(SB),0,$296-8
 	ADDQ AX,R12
 	ADCQ DX,R13
 	MOVQ $REDMASK51,DX
-	SHLQ $13,CX:SI
+	SHLQ $13,SI,CX
 	ANDQ DX,SI
-	SHLQ $13,R9:R8
+	SHLQ $13,R8,R9
 	ANDQ DX,R8
 	ADDQ CX,R8
-	SHLQ $13,R11:R10
+	SHLQ $13,R10,R11
 	ANDQ DX,R10
 	ADDQ R9,R10
-	SHLQ $13,R13:R12
+	SHLQ $13,R12,R13
 	ANDQ DX,R12
 	ADDQ R11,R12
-	SHLQ $13,R15:R14
+	SHLQ $13,R14,R15
 	ANDQ DX,R14
 	ADDQ R13,R14
 	IMUL3Q $19,R15,CX
@@ -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,R9:R8
-	ANDQ SI,R8
-	SHLQ $13,R11:R10
-	ANDQ SI,R10
-	ADDQ R9,R10
-	SHLQ $13,R13:R12
-	ANDQ SI,R12
-	ADDQ R11,R12
-	SHLQ $13,R15:R14
-	ANDQ SI,R14
-	ADDQ R13,R14
-	SHLQ $13,BP:BX
-	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,R8:CX
-	ANDQ SI,CX
-	SHLQ $13,R10:R9
-	ANDQ SI,R9
-	ADDQ R8,R9
-	SHLQ $13,R12:R11
-	ANDQ SI,R11
-	ADDQ R10,R11
-	SHLQ $13,R14:R13
-	ANDQ SI,R13
-	ADDQ R12,R13
-	SHLQ $13,BX:R15
-	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/vendor.json

@@ -603,6 +603,12 @@
 			"revision": "1d94cc7ab1c630336ab82ccb9c9cda72a875c382",
 			"revisionTime": "2020-02-12T18:29:22Z"
 		},
+		{
+			"checksumSHA1": "EmvPYN4QYc2M/Q28smEpKJ/r5Sg=",
+			"path": "golang.org/x/crypto/chacha20",
+			"revision": "1d94cc7ab1c630336ab82ccb9c9cda72a875c382",
+			"revisionTime": "2020-02-12T18:29:22Z"
+		},
 		{
 			"checksumSHA1": "9TPZ7plxFmlYtMEv2LLXRCEQg7c=",
 			"path": "golang.org/x/crypto/chacha20poly1305",
@@ -622,10 +628,10 @@
 			"revisionTime": "2019-02-05T21:23:42Z"
 		},
 		{
-			"checksumSHA1": "JjkVVfbdvjH+FbVgKPrv0+WtpFw=",
+			"checksumSHA1": "fP1hCOmaJM9ZH0dZi6dw/+uAbR8=",
 			"path": "golang.org/x/crypto/curve25519",
-			"revision": "a1f597ede03a7bef967a422b5b3a5bd08805a01e",
-			"revisionTime": "2019-02-05T21:23:42Z"
+			"revision": "1d94cc7ab1c630336ab82ccb9c9cda72a875c382",
+			"revisionTime": "2020-02-12T18:29:22Z"
 		},
 		{
 			"checksumSHA1": "ELSEW2KG0p3oua5lIxl1xW2oFBo=",