unit tests

Signed-off-by: Casey Lee <[email protected]>
This commit is contained in:
Casey Lee committed 2020-02-20 21:05:31 -05:00
1 parent be75ee20b1
commit 0582306861
466 files changed
+94682 -62525

No files matched your search

+9 -2
View File
@@ -2,8 +2,15 @@
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// Package cast5 implements CAST5, as defined in RFC 2144. CAST5 is a common
// OpenPGP cipher.
// Package cast5 implements CAST5, as defined in RFC 2144.
//
// CAST5 is a legacy cipher and its short block size makes it vulnerable to
// birthday bound attacks (see https://sweet32.info). It should only be used
// where compatibility with legacy systems, not security, is the goal.
//
// Deprecated: any new system should use AES (from crypto/aes, if necessary in
// an AEAD mode like crypto/cipher.NewGCM) or XChaCha20-Poly1305 (from
// golang.org/x/crypto/chacha20poly1305).
package cast5 // import "golang.org/x/crypto/cast5"
import "errors"
+17
View File
@@ -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)
}
+308
View File
@@ -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
+364
View File
@@ -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
}
+13
View File
@@ -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)
}
+16
View File
@@ -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)
}
+449
View File
@@ -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
+26
View File
@@ -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)
}
}
@@ -2,7 +2,7 @@
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// +build s390x,!gccgo,!appengine
// +build !gccgo,!appengine
#include "go_asm.h"
#include "textflag.h"
@@ -24,15 +24,6 @@ DATA ·constants<>+0x14(SB)/4, $0x3320646e
DATA ·constants<>+0x18(SB)/4, $0x79622d32
DATA ·constants<>+0x1c(SB)/4, $0x6b206574
// EXRL targets:
TEXT ·mvcSrcToBuf(SB), NOFRAME|NOSPLIT, $0
MVC $1, (R1), (R8)
RET
TEXT ·mvcBufToDst(SB), NOFRAME|NOSPLIT, $0
MVC $1, (R8), (R9)
RET
#define BSWAP V5
#define J0 V6
#define KEY0 V7
@@ -144,7 +135,7 @@ TEXT ·mvcBufToDst(SB), NOFRAME|NOSPLIT, $0
VMRHF v, w, c \ // c = {a[2], b[2], c[2], d[2]}
VMRLF v, w, d // d = {a[3], b[3], c[3], d[3]}
// func xorKeyStreamVX(dst, src []byte, key *[8]uint32, nonce *[3]uint32, counter *uint32, buf *[256]byte, len *int)
// func xorKeyStreamVX(dst, src []byte, key *[8]uint32, nonce *[3]uint32, counter *uint32)
TEXT ·xorKeyStreamVX(SB), NOSPLIT, $0
MOVD $·constants<>(SB), R1
MOVD dst+0(FP), R2 // R2=&dst[0]
@@ -152,25 +143,10 @@ TEXT ·xorKeyStreamVX(SB), NOSPLIT, $0
MOVD key+48(FP), R5 // R5=key
MOVD nonce+56(FP), R6 // R6=nonce
MOVD counter+64(FP), R7 // R7=counter
MOVD buf+72(FP), R8 // R8=buf
MOVD len+80(FP), R9 // R9=len
// load BSWAP and J0
VLM (R1), BSWAP, J0
// set up tail buffer
ADD $-1, R4, R12
MOVBZ R12, R12
CMPUBEQ R12, $255, aligned
MOVD R4, R1
AND $~255, R1
MOVD $(R3)(R1*1), R1
EXRL $·mvcSrcToBuf(SB), R12
MOVD $255, R0
SUB R12, R0
MOVD R0, (R9) // update len
aligned:
// setup
MOVD $95, R0
VLM (R5), KEY0, KEY1
@@ -217,9 +193,7 @@ loop:
// decrement length
ADD $-256, R4
BLT tail
continue:
// rearrange vectors
SHUFFLE(X0, X1, X2, X3, M0, M1, M2, M3)
ADDV(J0, X0, X1, X2, X3)
@@ -245,39 +219,6 @@ continue:
MOVD $256(R3), R3
CMPBNE R4, $0, chacha
CMPUBEQ R12, $255, return
EXRL $·mvcBufToDst(SB), R12 // len was updated during setup
return:
VSTEF $0, CTR, (R7)
RET
tail:
MOVD R2, R9
MOVD R8, R2
MOVD R8, R3
MOVD $0, R4
JMP continue
// func hasVectorFacility() bool
TEXT ·hasVectorFacility(SB), NOSPLIT, $24-1
MOVD $x-24(SP), R1
XC $24, 0(R1), 0(R1) // clear the storage
MOVD $2, R0 // R0 is the number of double words stored -1
WORD $0xB2B01000 // STFLE 0(R1)
XOR R0, R0 // reset the value of R0
MOVBZ z-8(SP), R1
AND $0x40, R1
BEQ novector
vectorinstalled:
// check if the vector instruction has been enabled
VLEIB $0, $0xF, V16
VLGVB $0, V16, R1
CMPBNE R1, $0xF, novector
MOVB $1, ret+0(FP) // have vx
RET
novector:
MOVB $0, ret+0(FP) // no vx
RET
@@ -4,9 +4,7 @@
package chacha20
import (
"runtime"
)
import "runtime"
// Platforms that have fast unaligned 32-bit little endian accesses.
const unaligned = runtime.GOARCH == "386" ||
-8
View File
@@ -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
-20
View File
@@ -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
-65
View File
@@ -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
+79 -818
View File
File diff suppressed because it is too large. Load diff
@@ -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
File diff suppressed because it is too large. Load diff
+11
View File
@@ -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)
}
-23
View File
@@ -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)
}
-73
View File
@@ -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
-169
View File
@@ -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
-132
View File
@@ -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
+5
View File
@@ -2,6 +2,11 @@
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// In Go 1.13, the ed25519 package was promoted to the standard library as
// crypto/ed25519, and this package became a wrapper for the standard library one.
//
// +build !go1.13
// Package ed25519 implements the Ed25519 signature algorithm. See
// https://ed25519.cr.yp.to/.
//
+73
View File
@@ -0,0 +1,73 @@
// Copyright 2019 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// +build go1.13
// Package ed25519 implements the Ed25519 signature algorithm. See
// https://ed25519.cr.yp.to/.
//
// These functions are also compatible with the “Ed25519” function defined in
// RFC 8032. However, unlike RFC 8032's formulation, this package's private key
// representation includes a public key suffix to make multiple signing
// operations with the same key more efficient. This package refers to the RFC
// 8032 private key as the “seed”.
//
// Beginning with Go 1.13, the functionality of this package was moved to the
// standard library as crypto/ed25519. This package only acts as a compatibility
// wrapper.
package ed25519
import (
"crypto/ed25519"
"io"
)
const (
// PublicKeySize is the size, in bytes, of public keys as used in this package.
PublicKeySize = 32
// PrivateKeySize is the size, in bytes, of private keys as used in this package.
PrivateKeySize = 64
// SignatureSize is the size, in bytes, of signatures generated and verified by this package.
SignatureSize = 64
// SeedSize is the size, in bytes, of private key seeds. These are the private key representations used by RFC 8032.
SeedSize = 32
)
// PublicKey is the type of Ed25519 public keys.
//
// This type is an alias for crypto/ed25519's PublicKey type.
// See the crypto/ed25519 package for the methods on this type.
type PublicKey = ed25519.PublicKey
// PrivateKey is the type of Ed25519 private keys. It implements crypto.Signer.
//
// This type is an alias for crypto/ed25519's PrivateKey type.
// See the crypto/ed25519 package for the methods on this type.
type PrivateKey = ed25519.PrivateKey
// GenerateKey generates a public/private key pair using entropy from rand.
// If rand is nil, crypto/rand.Reader will be used.
func GenerateKey(rand io.Reader) (PublicKey, PrivateKey, error) {
return ed25519.GenerateKey(rand)
}
// NewKeyFromSeed calculates a private key from a seed. It will panic if
// len(seed) is not SeedSize. This function is provided for interoperability
// with RFC 8032. RFC 8032's private keys correspond to seeds in this
// package.
func NewKeyFromSeed(seed []byte) PrivateKey {
return ed25519.NewKeyFromSeed(seed)
}
// Sign signs the message with privateKey and returns a signature. It will
// panic if len(privateKey) is not PrivateKeySize.
func Sign(privateKey PrivateKey, message []byte) []byte {
return ed25519.Sign(privateKey, message)
}
// Verify reports whether sig is a valid signature of message by publicKey. It
// will panic if len(publicKey) is not PublicKeySize.
func Verify(publicKey PublicKey, message, sig []byte) bool {
return ed25519.Verify(publicKey, message, sig)
}
-264
View File
@@ -1,264 +0,0 @@
// Copyright 2016 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// Package ChaCha20 implements the core ChaCha20 function as specified
// in https://tools.ietf.org/html/rfc7539#section-2.3.
package chacha20
import (
"crypto/cipher"
"encoding/binary"
"golang.org/x/crypto/internal/subtle"
)
// assert that *Cipher implements cipher.Stream
var _ cipher.Stream = (*Cipher)(nil)
// Cipher is a stateful instance of ChaCha20 using a particular key
// and nonce. A *Cipher implements the cipher.Stream interface.
type Cipher struct {
key [8]uint32
counter uint32 // incremented after each block
nonce [3]uint32
buf [bufSize]byte // buffer for unused keystream bytes
len int // number of unused keystream bytes at end of buf
}
// New creates a new ChaCha20 stream cipher with the given key and nonce.
// The initial counter value is set to 0.
func New(key [8]uint32, nonce [3]uint32) *Cipher {
return &Cipher{key: key, nonce: nonce}
}
// ChaCha20 constants spelling "expand 32-byte k"
const (
j0 uint32 = 0x61707865
j1 uint32 = 0x3320646e
j2 uint32 = 0x79622d32
j3 uint32 = 0x6b206574
)
func quarterRound(a, b, c, d uint32) (uint32, uint32, uint32, uint32) {
a += b
d ^= a
d = (d << 16) | (d >> 16)
c += d
b ^= c
b = (b << 12) | (b >> 20)
a += b
d ^= a
d = (d << 8) | (d >> 24)
c += d
b ^= c
b = (b << 7) | (b >> 25)
return a, b, c, d
}
// XORKeyStream XORs each byte in the given slice with a byte from the
// cipher's key stream. Dst and src must overlap entirely or not at all.
//
// If len(dst) < len(src), XORKeyStream will panic. It is acceptable
// to pass a dst bigger than src, and in that case, XORKeyStream will
// only update dst[:len(src)] and will not touch the rest of dst.
//
// Multiple calls to XORKeyStream behave as if the concatenation of
// the src buffers was passed in a single run. That is, Cipher
// maintains state and does not reset at each XORKeyStream call.
func (s *Cipher) XORKeyStream(dst, src []byte) {
if len(dst) < len(src) {
panic("chacha20: output smaller than input")
}
if subtle.InexactOverlap(dst[:len(src)], src) {
panic("chacha20: invalid buffer overlap")
}
// xor src with buffered keystream first
if s.len != 0 {
buf := s.buf[len(s.buf)-s.len:]
if len(src) < len(buf) {
buf = buf[:len(src)]
}
td, ts := dst[:len(buf)], src[:len(buf)] // BCE hint
for i, b := range buf {
td[i] = ts[i] ^ b
}
s.len -= len(buf)
if s.len != 0 {
return
}
s.buf = [len(s.buf)]byte{} // zero the empty buffer
src = src[len(buf):]
dst = dst[len(buf):]
}
if len(src) == 0 {
return
}
if haveAsm {
if uint64(len(src))+uint64(s.counter)*64 > (1<<38)-64 {
panic("chacha20: counter overflow")
}
s.xorKeyStreamAsm(dst, src)
return
}
// set up a 64-byte buffer to pad out the final block if needed
// (hoisted out of the main loop to avoid spills)
rem := len(src) % 64 // length of final block
fin := len(src) - rem // index of final block
if rem > 0 {
copy(s.buf[len(s.buf)-64:], src[fin:])
}
// pre-calculate most of the first round
s1, s5, s9, s13 := quarterRound(j1, s.key[1], s.key[5], s.nonce[0])
s2, s6, s10, s14 := quarterRound(j2, s.key[2], s.key[6], s.nonce[1])
s3, s7, s11, s15 := quarterRound(j3, s.key[3], s.key[7], s.nonce[2])
n := len(src)
src, dst = src[:n:n], dst[:n:n] // BCE hint
for i := 0; i < n; i += 64 {
// calculate the remainder of the first round
s0, s4, s8, s12 := quarterRound(j0, s.key[0], s.key[4], s.counter)
// execute the second round
x0, x5, x10, x15 := quarterRound(s0, s5, s10, s15)
x1, x6, x11, x12 := quarterRound(s1, s6, s11, s12)
x2, x7, x8, x13 := quarterRound(s2, s7, s8, s13)
x3, x4, x9, x14 := quarterRound(s3, s4, s9, s14)
// execute the remaining 18 rounds
for i := 0; i < 9; i++ {
x0, x4, x8, x12 = quarterRound(x0, x4, x8, x12)
x1, x5, x9, x13 = quarterRound(x1, x5, x9, x13)
x2, x6, x10, x14 = quarterRound(x2, x6, x10, x14)
x3, x7, x11, x15 = quarterRound(x3, x7, x11, x15)
x0, x5, x10, x15 = quarterRound(x0, x5, x10, x15)
x1, x6, x11, x12 = quarterRound(x1, x6, x11, x12)
x2, x7, x8, x13 = quarterRound(x2, x7, x8, x13)
x3, x4, x9, x14 = quarterRound(x3, x4, x9, x14)
}
x0 += j0
x1 += j1
x2 += j2
x3 += j3
x4 += s.key[0]
x5 += s.key[1]
x6 += s.key[2]
x7 += s.key[3]
x8 += s.key[4]
x9 += s.key[5]
x10 += s.key[6]
x11 += s.key[7]
x12 += s.counter
x13 += s.nonce[0]
x14 += s.nonce[1]
x15 += s.nonce[2]
// increment the counter
s.counter += 1
if s.counter == 0 {
panic("chacha20: counter overflow")
}
// pad to 64 bytes if needed
in, out := src[i:], dst[i:]
if i == fin {
// src[fin:] has already been copied into s.buf before
// the main loop
in, out = s.buf[len(s.buf)-64:], s.buf[len(s.buf)-64:]
}
in, out = in[:64], out[:64] // BCE hint
// XOR the key stream with the source and write out the result
xor(out[0:], in[0:], x0)
xor(out[4:], in[4:], x1)
xor(out[8:], in[8:], x2)
xor(out[12:], in[12:], x3)
xor(out[16:], in[16:], x4)
xor(out[20:], in[20:], x5)
xor(out[24:], in[24:], x6)
xor(out[28:], in[28:], x7)
xor(out[32:], in[32:], x8)
xor(out[36:], in[36:], x9)
xor(out[40:], in[40:], x10)
xor(out[44:], in[44:], x11)
xor(out[48:], in[48:], x12)
xor(out[52:], in[52:], x13)
xor(out[56:], in[56:], x14)
xor(out[60:], in[60:], x15)
}
// copy any trailing bytes out of the buffer and into dst
if rem != 0 {
s.len = 64 - rem
copy(dst[fin:], s.buf[len(s.buf)-64:])
}
}
// Advance discards bytes in the key stream until the next 64 byte block
// boundary is reached and updates the counter accordingly. If the key
// stream is already at a block boundary no bytes will be discarded and
// the counter will be unchanged.
func (s *Cipher) Advance() {
s.len -= s.len % 64
if s.len == 0 {
s.buf = [len(s.buf)]byte{}
}
}
// XORKeyStream crypts bytes from in to out using the given key and counters.
// In and out must overlap entirely or not at all. Counter contains the raw
// ChaCha20 counter bytes (i.e. block counter followed by nonce).
func XORKeyStream(out, in []byte, counter *[16]byte, key *[32]byte) {
s := Cipher{
key: [8]uint32{
binary.LittleEndian.Uint32(key[0:4]),
binary.LittleEndian.Uint32(key[4:8]),
binary.LittleEndian.Uint32(key[8:12]),
binary.LittleEndian.Uint32(key[12:16]),
binary.LittleEndian.Uint32(key[16:20]),
binary.LittleEndian.Uint32(key[20:24]),
binary.LittleEndian.Uint32(key[24:28]),
binary.LittleEndian.Uint32(key[28:32]),
},
nonce: [3]uint32{
binary.LittleEndian.Uint32(counter[4:8]),
binary.LittleEndian.Uint32(counter[8:12]),
binary.LittleEndian.Uint32(counter[12:16]),
},
counter: binary.LittleEndian.Uint32(counter[0:4]),
}
s.XORKeyStream(out, in)
}
// HChaCha20 uses the ChaCha20 core to generate a derived key from a key and a
// nonce. It should only be used as part of the XChaCha20 construction.
func HChaCha20(key *[8]uint32, nonce *[4]uint32) [8]uint32 {
x0, x1, x2, x3 := j0, j1, j2, j3
x4, x5, x6, x7 := key[0], key[1], key[2], key[3]
x8, x9, x10, x11 := key[4], key[5], key[6], key[7]
x12, x13, x14, x15 := nonce[0], nonce[1], nonce[2], nonce[3]
for i := 0; i < 10; i++ {
x0, x4, x8, x12 = quarterRound(x0, x4, x8, x12)
x1, x5, x9, x13 = quarterRound(x1, x5, x9, x13)
x2, x6, x10, x14 = quarterRound(x2, x6, x10, x14)
x3, x7, x11, x15 = quarterRound(x3, x7, x11, x15)
x0, x5, x10, x15 = quarterRound(x0, x5, x10, x15)
x1, x6, x11, x12 = quarterRound(x1, x6, x11, x12)
x2, x7, x8, x13 = quarterRound(x2, x7, x8, x13)
x3, x4, x9, x14 = quarterRound(x3, x4, x9, x14)
}
var out [8]uint32
out[0], out[1], out[2], out[3] = x0, x1, x2, x3
out[4], out[5], out[6], out[7] = x12, x13, x14, x15
return out
}
-16
View File
@@ -1,16 +0,0 @@
// Copyright 2018 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// +build !s390x gccgo appengine
package chacha20
const (
bufSize = 64
haveAsm = false
)
func (*Cipher) xorKeyStreamAsm(dst, src []byte) {
panic("not implemented")
}
-30
View File
@@ -1,30 +0,0 @@
// Copyright 2018 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// +build s390x,!gccgo,!appengine
package chacha20
var haveAsm = hasVectorFacility()
const bufSize = 256
// hasVectorFacility reports whether the machine supports the vector
// facility (vx).
// Implementation in asm_s390x.s.
func hasVectorFacility() bool
// xorKeyStreamVX is an assembly implementation of XORKeyStream. It must only
// be called when the vector facility is available.
// Implementation in asm_s390x.s.
//go:noescape
func xorKeyStreamVX(dst, src []byte, key *[8]uint32, nonce *[3]uint32, counter *uint32, buf *[256]byte, len *int)
func (c *Cipher) xorKeyStreamAsm(dst, src []byte) {
xorKeyStreamVX(dst, src, &c.key, &c.nonce, &c.counter, &c.buf, &c.len)
}
// EXRL targets, DO NOT CALL!
func mvcSrcToBuf()
func mvcBufToDst()
+3 -1
View File
@@ -76,7 +76,9 @@ func Encrypt(random io.Reader, pub *PublicKey, msg []byte) (c1, c2 *big.Int, err
// Bleichenbacher, Advances in Cryptology (Crypto '98),
func Decrypt(priv *PrivateKey, c1, c2 *big.Int) (msg []byte, err error) {
s := new(big.Int).Exp(c1, priv.X, priv.P)
s.ModInverse(s, priv.P)
if s.ModInverse(s, priv.P) == nil {
return nil, errors.New("elgamal: invalid private key")
}
s.Mul(s, c2)
s.Mod(s, priv.P)
em := s.Bytes()
+7 -7
View File
@@ -504,7 +504,7 @@ const defaultRSAKeyBits = 2048
// which may be empty but must not contain any of "()<>\x00".
// If config is nil, sensible defaults will be used.
func NewEntity(name, comment, email string, config *packet.Config) (*Entity, error) {
currentTime := config.Now()
creationTime := config.Now()
bits := defaultRSAKeyBits
if config != nil && config.RSABits != 0 {
@@ -525,8 +525,8 @@ func NewEntity(name, comment, email string, config *packet.Config) (*Entity, err
}
e := &Entity{
PrimaryKey: packet.NewRSAPublicKey(currentTime, &signingPriv.PublicKey),
PrivateKey: packet.NewRSAPrivateKey(currentTime, signingPriv),
PrimaryKey: packet.NewRSAPublicKey(creationTime, &signingPriv.PublicKey),
PrivateKey: packet.NewRSAPrivateKey(creationTime, signingPriv),
Identities: make(map[string]*Identity),
}
isPrimaryId := true
@@ -534,7 +534,7 @@ func NewEntity(name, comment, email string, config *packet.Config) (*Entity, err
Name: uid.Id,
UserId: uid,
SelfSignature: &packet.Signature{
CreationTime: currentTime,
CreationTime: creationTime,
SigType: packet.SigTypePositiveCert,
PubKeyAlgo: packet.PubKeyAlgoRSA,
Hash: config.Hash(),
@@ -563,10 +563,10 @@ func NewEntity(name, comment, email string, config *packet.Config) (*Entity, err
e.Subkeys = make([]Subkey, 1)
e.Subkeys[0] = Subkey{
PublicKey: packet.NewRSAPublicKey(currentTime, &encryptingPriv.PublicKey),
PrivateKey: packet.NewRSAPrivateKey(currentTime, encryptingPriv),
PublicKey: packet.NewRSAPublicKey(creationTime, &encryptingPriv.PublicKey),
PrivateKey: packet.NewRSAPrivateKey(creationTime, encryptingPriv),
Sig: &packet.Signature{
CreationTime: currentTime,
CreationTime: creationTime,
SigType: packet.SigTypeSubkeyBinding,
PubKeyAlgo: packet.PubKeyAlgoRSA,
Hash: config.Hash(),
+4 -2
View File
@@ -5,6 +5,7 @@
package packet
import (
"crypto"
"crypto/rsa"
"encoding/binary"
"io"
@@ -78,8 +79,9 @@ func (e *EncryptedKey) Decrypt(priv *PrivateKey, config *Config) error {
// padding oracle attacks.
switch priv.PubKeyAlgo {
case PubKeyAlgoRSA, PubKeyAlgoRSAEncryptOnly:
k := priv.PrivateKey.(*rsa.PrivateKey)
b, err = rsa.DecryptPKCS1v15(config.Random(), k, padToKeySize(&k.PublicKey, e.encryptedMPI1.bytes))
// Supports both *rsa.PrivateKey and crypto.Decrypter
k := priv.PrivateKey.(crypto.Decrypter)
b, err = k.Decrypt(config.Random(), padToKeySize(k.Public().(*rsa.PublicKey), e.encryptedMPI1.bytes), nil)
case PubKeyAlgoElGamal:
c1 := new(big.Int).SetBytes(e.encryptedMPI1.bytes)
c2 := new(big.Int).SetBytes(e.encryptedMPI2.bytes)
+14 -14
View File
@@ -31,54 +31,54 @@ type PrivateKey struct {
encryptedData []byte
cipher CipherFunction
s2k func(out, in []byte)
PrivateKey interface{} // An *{rsa|dsa|ecdsa}.PrivateKey or a crypto.Signer.
PrivateKey interface{} // An *{rsa|dsa|ecdsa}.PrivateKey or crypto.Signer/crypto.Decrypter (Decryptor RSA only).
sha1Checksum bool
iv []byte
}
func NewRSAPrivateKey(currentTime time.Time, priv *rsa.PrivateKey) *PrivateKey {
func NewRSAPrivateKey(creationTime time.Time, priv *rsa.PrivateKey) *PrivateKey {
pk := new(PrivateKey)
pk.PublicKey = *NewRSAPublicKey(currentTime, &priv.PublicKey)
pk.PublicKey = *NewRSAPublicKey(creationTime, &priv.PublicKey)
pk.PrivateKey = priv
return pk
}
func NewDSAPrivateKey(currentTime time.Time, priv *dsa.PrivateKey) *PrivateKey {
func NewDSAPrivateKey(creationTime time.Time, priv *dsa.PrivateKey) *PrivateKey {
pk := new(PrivateKey)
pk.PublicKey = *NewDSAPublicKey(currentTime, &priv.PublicKey)
pk.PublicKey = *NewDSAPublicKey(creationTime, &priv.PublicKey)
pk.PrivateKey = priv
return pk
}
func NewElGamalPrivateKey(currentTime time.Time, priv *elgamal.PrivateKey) *PrivateKey {
func NewElGamalPrivateKey(creationTime time.Time, priv *elgamal.PrivateKey) *PrivateKey {
pk := new(PrivateKey)
pk.PublicKey = *NewElGamalPublicKey(currentTime, &priv.PublicKey)
pk.PublicKey = *NewElGamalPublicKey(creationTime, &priv.PublicKey)
pk.PrivateKey = priv
return pk
}
func NewECDSAPrivateKey(currentTime time.Time, priv *ecdsa.PrivateKey) *PrivateKey {
func NewECDSAPrivateKey(creationTime time.Time, priv *ecdsa.PrivateKey) *PrivateKey {
pk := new(PrivateKey)
pk.PublicKey = *NewECDSAPublicKey(currentTime, &priv.PublicKey)
pk.PublicKey = *NewECDSAPublicKey(creationTime, &priv.PublicKey)
pk.PrivateKey = priv
return pk
}
// NewSignerPrivateKey creates a PrivateKey from a crypto.Signer that
// implements RSA or ECDSA.
func NewSignerPrivateKey(currentTime time.Time, signer crypto.Signer) *PrivateKey {
func NewSignerPrivateKey(creationTime time.Time, signer crypto.Signer) *PrivateKey {
pk := new(PrivateKey)
// In general, the public Keys should be used as pointers. We still
// type-switch on the values, for backwards-compatibility.
switch pubkey := signer.Public().(type) {
case *rsa.PublicKey:
pk.PublicKey = *NewRSAPublicKey(currentTime, pubkey)
pk.PublicKey = *NewRSAPublicKey(creationTime, pubkey)
case rsa.PublicKey:
pk.PublicKey = *NewRSAPublicKey(currentTime, &pubkey)
pk.PublicKey = *NewRSAPublicKey(creationTime, &pubkey)
case *ecdsa.PublicKey:
pk.PublicKey = *NewECDSAPublicKey(currentTime, pubkey)
pk.PublicKey = *NewECDSAPublicKey(creationTime, pubkey)
case ecdsa.PublicKey:
pk.PublicKey = *NewECDSAPublicKey(currentTime, &pubkey)
pk.PublicKey = *NewECDSAPublicKey(creationTime, &pubkey)
default:
panic("openpgp: unknown crypto.Signer type in NewSignerPrivateKey")
}
+39
View File
@@ -0,0 +1,39 @@
// Copyright 2019 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// +build !go1.13
package poly1305
// Generic fallbacks for the math/bits intrinsics, copied from
// src/math/bits/bits.go. They were added in Go 1.12, but Add64 and Sum64 had
// variable time fallbacks until Go 1.13.
func bitsAdd64(x, y, carry uint64) (sum, carryOut uint64) {
sum = x + y + carry
carryOut = ((x & y) | ((x | y) &^ sum)) >> 63
return
}
func bitsSub64(x, y, borrow uint64) (diff, borrowOut uint64) {
diff = x - y - borrow
borrowOut = ((^x & y) | (^(x ^ y) & diff)) >> 63
return
}
func bitsMul64(x, y uint64) (hi, lo uint64) {
const mask32 = 1<<32 - 1
x0 := x & mask32
x1 := x >> 32
y0 := y & mask32
y1 := y >> 32
w0 := x0 * y0
t := x1*y0 + w0>>32
w1 := t & mask32
w2 := t >> 32
w1 += x0 * y1
hi = x1*y1 + w2 + w1>>32
lo = x * y
return
}
+21
View File
@@ -0,0 +1,21 @@
// Copyright 2019 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// +build go1.13
package poly1305
import "math/bits"
func bitsAdd64(x, y, carry uint64) (sum, carryOut uint64) {
return bits.Add64(x, y, carry)
}
func bitsSub64(x, y, borrow uint64) (diff, borrowOut uint64) {
return bits.Sub64(x, y, borrow)
}
func bitsMul64(x, y uint64) (hi, lo uint64) {
return bits.Mul64(x, y)
}
+11
View File
@@ -0,0 +1,11 @@
// 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 !amd64,!ppc64le gccgo appengine
package poly1305
type mac struct{ macGeneric }
func newMAC(key *[32]byte) mac { return mac{newMACGeneric(key)} }
+73 -17
View File
@@ -2,21 +2,19 @@
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
/*
Package poly1305 implements Poly1305 one-time message authentication code as
specified in https://cr.yp.to/mac/poly1305-20050329.pdf.
Poly1305 is a fast, one-time authentication function. It is infeasible for an
attacker to generate an authenticator for a message without the key. However, a
key must only be used for a single message. Authenticating two different
messages with the same key allows an attacker to forge authenticators for other
messages with the same key.
Poly1305 was originally coupled with AES in order to make Poly1305-AES. AES was
used with a fixed key in order to generate one-time keys from an nonce.
However, in this package AES isn't used and the one-time key is specified
directly.
*/
// Package poly1305 implements Poly1305 one-time message authentication code as
// specified in https://cr.yp.to/mac/poly1305-20050329.pdf.
//
// Poly1305 is a fast, one-time authentication function. It is infeasible for an
// attacker to generate an authenticator for a message without the key. However, a
// key must only be used for a single message. Authenticating two different
// messages with the same key allows an attacker to forge authenticators for other
// messages with the same key.
//
// Poly1305 was originally coupled with AES in order to make Poly1305-AES. AES was
// used with a fixed key in order to generate one-time keys from an nonce.
// However, in this package AES isn't used and the one-time key is specified
// directly.
package poly1305 // import "golang.org/x/crypto/poly1305"
import "crypto/subtle"
@@ -24,10 +22,68 @@ import "crypto/subtle"
// TagSize is the size, in bytes, of a poly1305 authenticator.
const TagSize = 16
// Verify returns true if mac is a valid authenticator for m with the given
// key.
// Sum generates an authenticator for msg using a one-time key and puts the
// 16-byte result into out. Authenticating two different messages with the same
// key allows an attacker to forge messages at will.
func Sum(out *[16]byte, m []byte, key *[32]byte) {
sum(out, m, key)
}
// Verify returns true if mac is a valid authenticator for m with the given key.
func Verify(mac *[16]byte, m []byte, key *[32]byte) bool {
var tmp [16]byte
Sum(&tmp, m, key)
return subtle.ConstantTimeCompare(tmp[:], mac[:]) == 1
}
// New returns a new MAC computing an authentication
// tag of all data written to it with the given key.
// This allows writing the message progressively instead
// of passing it as a single slice. Common users should use
// the Sum function instead.
//
// The key must be unique for each message, as authenticating
// two different messages with the same key allows an attacker
// to forge messages at will.
func New(key *[32]byte) *MAC {
return &MAC{
mac: newMAC(key),
finalized: false,
}
}
// MAC is an io.Writer computing an authentication tag
// of the data written to it.
//
// MAC cannot be used like common hash.Hash implementations,
// because using a poly1305 key twice breaks its security.
// Therefore writing data to a running MAC after calling
// Sum causes it to panic.
type MAC struct {
mac // platform-dependent implementation
finalized bool
}
// Size returns the number of bytes Sum will return.
func (h *MAC) Size() int { return TagSize }
// Write adds more data to the running message authentication code.
// It never returns an error.
//
// It must not be called after the first call of Sum.
func (h *MAC) Write(p []byte) (n int, err error) {
if h.finalized {
panic("poly1305: write to MAC after Sum")
}
return h.mac.Write(p)
}
// Sum computes the authenticator of all data written to the
// message authentication code.
func (h *MAC) Sum(b []byte) []byte {
var mac [TagSize]byte
h.mac.Sum(&mac)
h.finalized = true
return append(b, mac[:]...)
}
+47 -11
View File
@@ -6,17 +6,53 @@
package poly1305
// This function is implemented in sum_amd64.s
//go:noescape
func poly1305(out *[16]byte, m *byte, mlen uint64, key *[32]byte)
func update(state *macState, msg []byte)
// Sum generates an authenticator for m using a one-time key and puts the
// 16-byte result into out. Authenticating two different messages with the same
// key allows an attacker to forge messages at will.
func Sum(out *[16]byte, m []byte, key *[32]byte) {
var mPtr *byte
if len(m) > 0 {
mPtr = &m[0]
}
poly1305(out, mPtr, uint64(len(m)), key)
func sum(out *[16]byte, m []byte, key *[32]byte) {
h := newMAC(key)
h.Write(m)
h.Sum(out)
}
func newMAC(key *[32]byte) (h mac) {
initialize(key, &h.r, &h.s)
return
}
// mac is a wrapper for macGeneric that redirects calls that would have gone to
// updateGeneric to update.
//
// Its Write and Sum methods are otherwise identical to the macGeneric ones, but
// using function pointers would carry a major performance cost.
type mac struct{ macGeneric }
func (h *mac) Write(p []byte) (int, error) {
nn := len(p)
if h.offset > 0 {
n := copy(h.buffer[h.offset:], p)
if h.offset+n < TagSize {
h.offset += n
return nn, nil
}
p = p[n:]
h.offset = 0
update(&h.macState, h.buffer[:])
}
if n := len(p) - (len(p) % TagSize); n > 0 {
update(&h.macState, p[:n])
p = p[n:]
}
if len(p) > 0 {
h.offset += copy(h.buffer[h.offset:], p)
}
return nn, nil
}
func (h *mac) Sum(out *[16]byte) {
state := h.macState
if h.offset > 0 {
update(&state, h.buffer[:h.offset])
}
finalize(out, &state.h, &state.s)
}
+13 -30
View File
@@ -54,24 +54,17 @@
ADCQ t3, h1; \
ADCQ $0, h2
DATA ·poly1305Mask<>+0x00(SB)/8, $0x0FFFFFFC0FFFFFFF
DATA ·poly1305Mask<>+0x08(SB)/8, $0x0FFFFFFC0FFFFFFC
GLOBL ·poly1305Mask<>(SB), RODATA, $16
// func update(state *[7]uint64, msg []byte)
TEXT ·update(SB), $0-32
MOVQ state+0(FP), DI
MOVQ msg_base+8(FP), SI
MOVQ msg_len+16(FP), R15
// func poly1305(out *[16]byte, m *byte, mlen uint64, key *[32]key)
TEXT ·poly1305(SB), $0-32
MOVQ out+0(FP), DI
MOVQ m+8(FP), SI
MOVQ mlen+16(FP), R15
MOVQ key+24(FP), AX
MOVQ 0(AX), R11
MOVQ 8(AX), R12
ANDQ ·poly1305Mask<>(SB), R11 // r0
ANDQ ·poly1305Mask<>+8(SB), R12 // r1
XORQ R8, R8 // h0
XORQ R9, R9 // h1
XORQ R10, R10 // h2
MOVQ 0(DI), R8 // h0
MOVQ 8(DI), R9 // h1
MOVQ 16(DI), R10 // h2
MOVQ 24(DI), R11 // r0
MOVQ 32(DI), R12 // r1
CMPQ R15, $16
JB bytes_between_0_and_15
@@ -109,17 +102,7 @@ flush_buffer:
JMP multiply
done:
MOVQ R8, AX
MOVQ R9, BX
SUBQ $0xFFFFFFFFFFFFFFFB, AX
SBBQ $0xFFFFFFFFFFFFFFFF, BX
SBBQ $3, R10
CMOVQCS R8, AX
CMOVQCS R9, BX
MOVQ key+24(FP), R8
ADDQ 16(R8), AX
ADCQ 24(R8), BX
MOVQ AX, 0(DI)
MOVQ BX, 8(DI)
MOVQ R8, 0(DI)
MOVQ R9, 8(DI)
MOVQ R10, 16(DI)
RET
+2 -5
View File
@@ -6,14 +6,11 @@
package poly1305
// This function is implemented in sum_arm.s
// poly1305_auth_armv6 is implemented in sum_arm.s
//go:noescape
func poly1305_auth_armv6(out *[16]byte, m *byte, mlen uint32, key *[32]byte)
// Sum generates an authenticator for m using a one-time key and puts the
// 16-byte result into out. Authenticating two different messages with the same
// key allows an attacker to forge messages at will.
func Sum(out *[16]byte, m []byte, key *[32]byte) {
func sum(out *[16]byte, m []byte, key *[32]byte) {
var mPtr *byte
if len(m) > 0 {
mPtr = &m[0]
+307
View File
@@ -0,0 +1,307 @@
// Copyright 2018 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// This file provides the generic implementation of Sum and MAC. Other files
// might provide optimized assembly implementations of some of this code.
package poly1305
import "encoding/binary"
// Poly1305 [RFC 7539] is a relatively simple algorithm: the authentication tag
// for a 64 bytes message is approximately
//
// s + m[0:16] * r⁴ + m[16:32] * r³ + m[32:48] * r² + m[48:64] * r mod 2¹³⁰ - 5
//
// for some secret r and s. It can be computed sequentially like
//
// for len(msg) > 0:
// h += read(msg, 16)
// h *= r
// h %= 2¹³⁰ - 5
// return h + s
//
// All the complexity is about doing performant constant-time math on numbers
// larger than any available numeric type.
func sumGeneric(out *[TagSize]byte, msg []byte, key *[32]byte) {
h := newMACGeneric(key)
h.Write(msg)
h.Sum(out)
}
func newMACGeneric(key *[32]byte) (h macGeneric) {
initialize(key, &h.r, &h.s)
return
}
// macState holds numbers in saturated 64-bit little-endian limbs. That is,
// the value of [x0, x1, x2] is x[0] + x[1] * 2⁶⁴ + x[2] * 2¹²⁸.
type macState struct {
// h is the main accumulator. It is to be interpreted modulo 2¹³⁰ - 5, but
// can grow larger during and after rounds.
h [3]uint64
// r and s are the private key components.
r [2]uint64
s [2]uint64
}
type macGeneric struct {
macState
buffer [TagSize]byte
offset int
}
// Write splits the incoming message into TagSize chunks, and passes them to
// update. It buffers incomplete chunks.
func (h *macGeneric) Write(p []byte) (int, error) {
nn := len(p)
if h.offset > 0 {
n := copy(h.buffer[h.offset:], p)
if h.offset+n < TagSize {
h.offset += n
return nn, nil
}
p = p[n:]
h.offset = 0
updateGeneric(&h.macState, h.buffer[:])
}
if n := len(p) - (len(p) % TagSize); n > 0 {
updateGeneric(&h.macState, p[:n])
p = p[n:]
}
if len(p) > 0 {
h.offset += copy(h.buffer[h.offset:], p)
}
return nn, nil
}
// Sum flushes the last incomplete chunk from the buffer, if any, and generates
// the MAC output. It does not modify its state, in order to allow for multiple
// calls to Sum, even if no Write is allowed after Sum.
func (h *macGeneric) Sum(out *[TagSize]byte) {
state := h.macState
if h.offset > 0 {
updateGeneric(&state, h.buffer[:h.offset])
}
finalize(out, &state.h, &state.s)
}
// [rMask0, rMask1] is the specified Poly1305 clamping mask in little-endian. It
// clears some bits of the secret coefficient to make it possible to implement
// multiplication more efficiently.
const (
rMask0 = 0x0FFFFFFC0FFFFFFF
rMask1 = 0x0FFFFFFC0FFFFFFC
)
func initialize(key *[32]byte, r, s *[2]uint64) {
r[0] = binary.LittleEndian.Uint64(key[0:8]) & rMask0
r[1] = binary.LittleEndian.Uint64(key[8:16]) & rMask1
s[0] = binary.LittleEndian.Uint64(key[16:24])
s[1] = binary.LittleEndian.Uint64(key[24:32])
}
// uint128 holds a 128-bit number as two 64-bit limbs, for use with the
// bits.Mul64 and bits.Add64 intrinsics.
type uint128 struct {
lo, hi uint64
}
func mul64(a, b uint64) uint128 {
hi, lo := bitsMul64(a, b)
return uint128{lo, hi}
}
func add128(a, b uint128) uint128 {
lo, c := bitsAdd64(a.lo, b.lo, 0)
hi, c := bitsAdd64(a.hi, b.hi, c)
if c != 0 {
panic("poly1305: unexpected overflow")
}
return uint128{lo, hi}
}
func shiftRightBy2(a uint128) uint128 {
a.lo = a.lo>>2 | (a.hi&3)<<62
a.hi = a.hi >> 2
return a
}
// updateGeneric absorbs msg into the state.h accumulator. For each chunk m of
// 128 bits of message, it computes
//
// h₊ = (h + m) * r mod 2¹³⁰ - 5
//
// If the msg length is not a multiple of TagSize, it assumes the last
// incomplete chunk is the final one.
func updateGeneric(state *macState, msg []byte) {
h0, h1, h2 := state.h[0], state.h[1], state.h[2]
r0, r1 := state.r[0], state.r[1]
for len(msg) > 0 {
var c uint64
// For the first step, h + m, we use a chain of bits.Add64 intrinsics.
// The resulting value of h might exceed 2¹³⁰ - 5, but will be partially
// reduced at the end of the multiplication below.
//
// The spec requires us to set a bit just above the message size, not to
// hide leading zeroes. For full chunks, that's 1 << 128, so we can just
// add 1 to the most significant (2¹²⁸) limb, h2.
if len(msg) >= TagSize {
h0, c = bitsAdd64(h0, binary.LittleEndian.Uint64(msg[0:8]), 0)
h1, c = bitsAdd64(h1, binary.LittleEndian.Uint64(msg[8:16]), c)
h2 += c + 1
msg = msg[TagSize:]
} else {
var buf [TagSize]byte
copy(buf[:], msg)
buf[len(msg)] = 1
h0, c = bitsAdd64(h0, binary.LittleEndian.Uint64(buf[0:8]), 0)
h1, c = bitsAdd64(h1, binary.LittleEndian.Uint64(buf[8:16]), c)
h2 += c
msg = nil
}
// Multiplication of big number limbs is similar to elementary school
// columnar multiplication. Instead of digits, there are 64-bit limbs.
//
// We are multiplying a 3 limbs number, h, by a 2 limbs number, r.
//
// h2 h1 h0 x
// r1 r0 =
// ----------------
// h2r0 h1r0 h0r0 <-- individual 128-bit products
// + h2r1 h1r1 h0r1
// ------------------------
// m3 m2 m1 m0 <-- result in 128-bit overlapping limbs
// ------------------------
// m3.hi m2.hi m1.hi m0.hi <-- carry propagation
// + m3.lo m2.lo m1.lo m0.lo
// -------------------------------
// t4 t3 t2 t1 t0 <-- final result in 64-bit limbs
//
// The main difference from pen-and-paper multiplication is that we do
// carry propagation in a separate step, as if we wrote two digit sums
// at first (the 128-bit limbs), and then carried the tens all at once.
h0r0 := mul64(h0, r0)
h1r0 := mul64(h1, r0)
h2r0 := mul64(h2, r0)
h0r1 := mul64(h0, r1)
h1r1 := mul64(h1, r1)
h2r1 := mul64(h2, r1)
// Since h2 is known to be at most 7 (5 + 1 + 1), and r0 and r1 have their
// top 4 bits cleared by rMask{0,1}, we know that their product is not going
// to overflow 64 bits, so we can ignore the high part of the products.
//
// This also means that the product doesn't have a fifth limb (t4).
if h2r0.hi != 0 {
panic("poly1305: unexpected overflow")
}
if h2r1.hi != 0 {
panic("poly1305: unexpected overflow")
}
m0 := h0r0
m1 := add128(h1r0, h0r1) // These two additions don't overflow thanks again
m2 := add128(h2r0, h1r1) // to the 4 masked bits at the top of r0 and r1.
m3 := h2r1
t0 := m0.lo
t1, c := bitsAdd64(m1.lo, m0.hi, 0)
t2, c := bitsAdd64(m2.lo, m1.hi, c)
t3, _ := bitsAdd64(m3.lo, m2.hi, c)
// Now we have the result as 4 64-bit limbs, and we need to reduce it
// modulo 2¹³⁰ - 5. The special shape of this Crandall prime lets us do
// a cheap partial reduction according to the reduction identity
//
// c * 2¹³⁰ + n = c * 5 + n mod 2¹³⁰ - 5
//
// because 2¹³⁰ = 5 mod 2¹³⁰ - 5. Partial reduction since the result is
// likely to be larger than 2¹³⁰ - 5, but still small enough to fit the
// assumptions we make about h in the rest of the code.
//
// See also https://speakerdeck.com/gtank/engineering-prime-numbers?slide=23
// We split the final result at the 2¹³⁰ mark into h and cc, the carry.
// Note that the carry bits are effectively shifted left by 2, in other
// words, cc = c * 4 for the c in the reduction identity.
h0, h1, h2 = t0, t1, t2&maskLow2Bits
cc := uint128{t2 & maskNotLow2Bits, t3}
// To add c * 5 to h, we first add cc = c * 4, and then add (cc >> 2) = c.
h0, c = bitsAdd64(h0, cc.lo, 0)
h1, c = bitsAdd64(h1, cc.hi, c)
h2 += c
cc = shiftRightBy2(cc)
h0, c = bitsAdd64(h0, cc.lo, 0)
h1, c = bitsAdd64(h1, cc.hi, c)
h2 += c
// h2 is at most 3 + 1 + 1 = 5, making the whole of h at most
//
// 5 * 2¹²⁸ + (2¹²⁸ - 1) = 6 * 2¹²⁸ - 1
}
state.h[0], state.h[1], state.h[2] = h0, h1, h2
}
const (
maskLow2Bits uint64 = 0x0000000000000003
maskNotLow2Bits uint64 = ^maskLow2Bits
)
// select64 returns x if v == 1 and y if v == 0, in constant time.
func select64(v, x, y uint64) uint64 { return ^(v-1)&x | (v-1)&y }
// [p0, p1, p2] is 2¹³⁰ - 5 in little endian order.
const (
p0 = 0xFFFFFFFFFFFFFFFB
p1 = 0xFFFFFFFFFFFFFFFF
p2 = 0x0000000000000003
)
// finalize completes the modular reduction of h and computes
//
// out = h + s mod 2¹²⁸
//
func finalize(out *[TagSize]byte, h *[3]uint64, s *[2]uint64) {
h0, h1, h2 := h[0], h[1], h[2]
// After the partial reduction in updateGeneric, h might be more than
// 2¹³⁰ - 5, but will be less than 2 * (2¹³⁰ - 5). To complete the reduction
// in constant time, we compute t = h - (2¹³⁰ - 5), and select h as the
// result if the subtraction underflows, and t otherwise.
hMinusP0, b := bitsSub64(h0, p0, 0)
hMinusP1, b := bitsSub64(h1, p1, b)
_, b = bitsSub64(h2, p2, b)
// h = h if h < p else h - p
h0 = select64(b, h0, hMinusP0)
h1 = select64(b, h1, hMinusP1)
// Finally, we compute the last Poly1305 step
//
// tag = h + s mod 2¹²⁸
//
// by just doing a wide addition with the 128 low bits of h and discarding
// the overflow.
h0, c := bitsAdd64(h0, s[0], 0)
h1, _ = bitsAdd64(h1, s[1], c)
binary.LittleEndian.PutUint64(out[0:8], h0)
binary.LittleEndian.PutUint64(out[8:16], h1)
}
+5 -6
View File
@@ -2,13 +2,12 @@
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// +build s390x,!go1.11 !arm,!amd64,!s390x gccgo appengine nacl
// +build s390x,!go1.11 !arm,!amd64,!s390x,!ppc64le gccgo appengine nacl
package poly1305
// Sum generates an authenticator for msg using a one-time key and puts the
// 16-byte result into out. Authenticating two different messages with the same
// key allows an attacker to forge messages at will.
func Sum(out *[TagSize]byte, msg []byte, key *[32]byte) {
sumGeneric(out, msg, key)
func sum(out *[TagSize]byte, msg []byte, key *[32]byte) {
h := newMAC(key)
h.Write(msg)
h.Sum(out)
}
+58
View File
@@ -0,0 +1,58 @@
// Copyright 2019 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// +build ppc64le,!gccgo,!appengine
package poly1305
//go:noescape
func update(state *macState, msg []byte)
func sum(out *[16]byte, m []byte, key *[32]byte) {
h := newMAC(key)
h.Write(m)
h.Sum(out)
}
func newMAC(key *[32]byte) (h mac) {
initialize(key, &h.r, &h.s)
return
}
// mac is a wrapper for macGeneric that redirects calls that would have gone to
// updateGeneric to update.
//
// Its Write and Sum methods are otherwise identical to the macGeneric ones, but
// using function pointers would carry a major performance cost.
type mac struct{ macGeneric }
func (h *mac) Write(p []byte) (int, error) {
nn := len(p)
if h.offset > 0 {
n := copy(h.buffer[h.offset:], p)
if h.offset+n < TagSize {
h.offset += n
return nn, nil
}
p = p[n:]
h.offset = 0
update(&h.macState, h.buffer[:])
}
if n := len(p) - (len(p) % TagSize); n > 0 {
update(&h.macState, p[:n])
p = p[n:]
}
if len(p) > 0 {
h.offset += copy(h.buffer[h.offset:], p)
}
return nn, nil
}
func (h *mac) Sum(out *[16]byte) {
state := h.macState
if h.offset > 0 {
update(&state, h.buffer[:h.offset])
}
finalize(out, &state.h, &state.s)
}
+181
View File
@@ -0,0 +1,181 @@
// Copyright 2019 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// +build ppc64le,!gccgo,!appengine
#include "textflag.h"
// This was ported from the amd64 implementation.
#define POLY1305_ADD(msg, h0, h1, h2, t0, t1, t2) \
MOVD (msg), t0; \
MOVD 8(msg), t1; \
MOVD $1, t2; \
ADDC t0, h0, h0; \
ADDE t1, h1, h1; \
ADDE t2, h2; \
ADD $16, msg
#define POLY1305_MUL(h0, h1, h2, r0, r1, t0, t1, t2, t3, t4, t5) \
MULLD r0, h0, t0; \
MULLD r0, h1, t4; \
MULHDU r0, h0, t1; \
MULHDU r0, h1, t5; \
ADDC t4, t1, t1; \
MULLD r0, h2, t2; \
ADDZE t5; \
MULHDU r1, h0, t4; \
MULLD r1, h0, h0; \
ADD t5, t2, t2; \
ADDC h0, t1, t1; \
MULLD h2, r1, t3; \
ADDZE t4, h0; \
MULHDU r1, h1, t5; \
MULLD r1, h1, t4; \
ADDC t4, t2, t2; \
ADDE t5, t3, t3; \
ADDC h0, t2, t2; \
MOVD $-4, t4; \
MOVD t0, h0; \
MOVD t1, h1; \
ADDZE t3; \
ANDCC $3, t2, h2; \
AND t2, t4, t0; \
ADDC t0, h0, h0; \
ADDE t3, h1, h1; \
SLD $62, t3, t4; \
SRD $2, t2; \
ADDZE h2; \
OR t4, t2, t2; \
SRD $2, t3; \
ADDC t2, h0, h0; \
ADDE t3, h1, h1; \
ADDZE h2
DATA ·poly1305Mask<>+0x00(SB)/8, $0x0FFFFFFC0FFFFFFF
DATA ·poly1305Mask<>+0x08(SB)/8, $0x0FFFFFFC0FFFFFFC
GLOBL ·poly1305Mask<>(SB), RODATA, $16
// func update(state *[7]uint64, msg []byte)
TEXT ·update(SB), $0-32
MOVD state+0(FP), R3
MOVD msg_base+8(FP), R4
MOVD msg_len+16(FP), R5
MOVD 0(R3), R8 // h0
MOVD 8(R3), R9 // h1
MOVD 16(R3), R10 // h2
MOVD 24(R3), R11 // r0
MOVD 32(R3), R12 // r1
CMP R5, $16
BLT bytes_between_0_and_15
loop:
POLY1305_ADD(R4, R8, R9, R10, R20, R21, R22)
multiply:
POLY1305_MUL(R8, R9, R10, R11, R12, R16, R17, R18, R14, R20, R21)
ADD $-16, R5
CMP R5, $16
BGE loop
bytes_between_0_and_15:
CMP $0, R5
BEQ done
MOVD $0, R16 // h0
MOVD $0, R17 // h1
flush_buffer:
CMP R5, $8
BLE just1
MOVD $8, R21
SUB R21, R5, R21
// Greater than 8 -- load the rightmost remaining bytes in msg
// and put into R17 (h1)
MOVD (R4)(R21), R17
MOVD $16, R22
// Find the offset to those bytes
SUB R5, R22, R22
SLD $3, R22
// Shift to get only the bytes in msg
SRD R22, R17, R17
// Put 1 at high end
MOVD $1, R23
SLD $3, R21
SLD R21, R23, R23
OR R23, R17, R17
// Remainder is 8
MOVD $8, R5
just1:
CMP R5, $8
BLT less8
// Exactly 8
MOVD (R4), R16
CMP $0, R17
// Check if we've already set R17; if not
// set 1 to indicate end of msg.
BNE carry
MOVD $1, R17
BR carry
less8:
MOVD $0, R16 // h0
MOVD $0, R22 // shift count
CMP R5, $4
BLT less4
MOVWZ (R4), R16
ADD $4, R4
ADD $-4, R5
MOVD $32, R22
less4:
CMP R5, $2
BLT less2
MOVHZ (R4), R21
SLD R22, R21, R21
OR R16, R21, R16
ADD $16, R22
ADD $-2, R5
ADD $2, R4
less2:
CMP $0, R5
BEQ insert1
MOVBZ (R4), R21
SLD R22, R21, R21
OR R16, R21, R16
ADD $8, R22
insert1:
// Insert 1 at end of msg
MOVD $1, R21
SLD R22, R21, R21
OR R16, R21, R16
carry:
// Add new values to h0, h1, h2
ADDC R16, R8
ADDE R17, R9
ADDE $0, R10
MOVD $16, R5
ADD R5, R4
BR multiply
done:
// Save h0, h1, h2 in state
MOVD R8, 0(R3)
MOVD R9, 8(R3)
MOVD R10, 16(R3)
RET
-139
View File
@@ -1,139 +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 poly1305
import "encoding/binary"
// sumGeneric generates an authenticator for msg using a one-time key and
// puts the 16-byte result into out. This is the generic implementation of
// Sum and should be called if no assembly implementation is available.
func sumGeneric(out *[TagSize]byte, msg []byte, key *[32]byte) {
var (
h0, h1, h2, h3, h4 uint32 // the hash accumulators
r0, r1, r2, r3, r4 uint64 // the r part of the key
)
r0 = uint64(binary.LittleEndian.Uint32(key[0:]) & 0x3ffffff)
r1 = uint64((binary.LittleEndian.Uint32(key[3:]) >> 2) & 0x3ffff03)
r2 = uint64((binary.LittleEndian.Uint32(key[6:]) >> 4) & 0x3ffc0ff)
r3 = uint64((binary.LittleEndian.Uint32(key[9:]) >> 6) & 0x3f03fff)
r4 = uint64((binary.LittleEndian.Uint32(key[12:]) >> 8) & 0x00fffff)
R1, R2, R3, R4 := r1*5, r2*5, r3*5, r4*5
for len(msg) >= TagSize {
// h += msg
h0 += binary.LittleEndian.Uint32(msg[0:]) & 0x3ffffff
h1 += (binary.LittleEndian.Uint32(msg[3:]) >> 2) & 0x3ffffff
h2 += (binary.LittleEndian.Uint32(msg[6:]) >> 4) & 0x3ffffff
h3 += (binary.LittleEndian.Uint32(msg[9:]) >> 6) & 0x3ffffff
h4 += (binary.LittleEndian.Uint32(msg[12:]) >> 8) | (1 << 24)
// h *= r
d0 := (uint64(h0) * r0) + (uint64(h1) * R4) + (uint64(h2) * R3) + (uint64(h3) * R2) + (uint64(h4) * R1)
d1 := (d0 >> 26) + (uint64(h0) * r1) + (uint64(h1) * r0) + (uint64(h2) * R4) + (uint64(h3) * R3) + (uint64(h4) * R2)
d2 := (d1 >> 26) + (uint64(h0) * r2) + (uint64(h1) * r1) + (uint64(h2) * r0) + (uint64(h3) * R4) + (uint64(h4) * R3)
d3 := (d2 >> 26) + (uint64(h0) * r3) + (uint64(h1) * r2) + (uint64(h2) * r1) + (uint64(h3) * r0) + (uint64(h4) * R4)
d4 := (d3 >> 26) + (uint64(h0) * r4) + (uint64(h1) * r3) + (uint64(h2) * r2) + (uint64(h3) * r1) + (uint64(h4) * r0)
// h %= p
h0 = uint32(d0) & 0x3ffffff
h1 = uint32(d1) & 0x3ffffff
h2 = uint32(d2) & 0x3ffffff
h3 = uint32(d3) & 0x3ffffff
h4 = uint32(d4) & 0x3ffffff
h0 += uint32(d4>>26) * 5
h1 += h0 >> 26
h0 = h0 & 0x3ffffff
msg = msg[TagSize:]
}
if len(msg) > 0 {
var block [TagSize]byte
off := copy(block[:], msg)
block[off] = 0x01
// h += msg
h0 += binary.LittleEndian.Uint32(block[0:]) & 0x3ffffff
h1 += (binary.LittleEndian.Uint32(block[3:]) >> 2) & 0x3ffffff
h2 += (binary.LittleEndian.Uint32(block[6:]) >> 4) & 0x3ffffff
h3 += (binary.LittleEndian.Uint32(block[9:]) >> 6) & 0x3ffffff
h4 += (binary.LittleEndian.Uint32(block[12:]) >> 8)
// h *= r
d0 := (uint64(h0) * r0) + (uint64(h1) * R4) + (uint64(h2) * R3) + (uint64(h3) * R2) + (uint64(h4) * R1)
d1 := (d0 >> 26) + (uint64(h0) * r1) + (uint64(h1) * r0) + (uint64(h2) * R4) + (uint64(h3) * R3) + (uint64(h4) * R2)
d2 := (d1 >> 26) + (uint64(h0) * r2) + (uint64(h1) * r1) + (uint64(h2) * r0) + (uint64(h3) * R4) + (uint64(h4) * R3)
d3 := (d2 >> 26) + (uint64(h0) * r3) + (uint64(h1) * r2) + (uint64(h2) * r1) + (uint64(h3) * r0) + (uint64(h4) * R4)
d4 := (d3 >> 26) + (uint64(h0) * r4) + (uint64(h1) * r3) + (uint64(h2) * r2) + (uint64(h3) * r1) + (uint64(h4) * r0)
// h %= p
h0 = uint32(d0) & 0x3ffffff
h1 = uint32(d1) & 0x3ffffff
h2 = uint32(d2) & 0x3ffffff
h3 = uint32(d3) & 0x3ffffff
h4 = uint32(d4) & 0x3ffffff
h0 += uint32(d4>>26) * 5
h1 += h0 >> 26
h0 = h0 & 0x3ffffff
}
// h %= p reduction
h2 += h1 >> 26
h1 &= 0x3ffffff
h3 += h2 >> 26
h2 &= 0x3ffffff
h4 += h3 >> 26
h3 &= 0x3ffffff
h0 += 5 * (h4 >> 26)
h4 &= 0x3ffffff
h1 += h0 >> 26
h0 &= 0x3ffffff
// h - p
t0 := h0 + 5
t1 := h1 + (t0 >> 26)
t2 := h2 + (t1 >> 26)
t3 := h3 + (t2 >> 26)
t4 := h4 + (t3 >> 26) - (1 << 26)
t0 &= 0x3ffffff
t1 &= 0x3ffffff
t2 &= 0x3ffffff
t3 &= 0x3ffffff
// select h if h < p else h - p
t_mask := (t4 >> 31) - 1
h_mask := ^t_mask
h0 = (h0 & h_mask) | (t0 & t_mask)
h1 = (h1 & h_mask) | (t1 & t_mask)
h2 = (h2 & h_mask) | (t2 & t_mask)
h3 = (h3 & h_mask) | (t3 & t_mask)
h4 = (h4 & h_mask) | (t4 & t_mask)
// h %= 2^128
h0 |= h1 << 26
h1 = ((h1 >> 6) | (h2 << 20))
h2 = ((h2 >> 12) | (h3 << 14))
h3 = ((h3 >> 18) | (h4 << 8))
// s: the s part of the key
// tag = (h + s) % (2^128)
t := uint64(h0) + uint64(binary.LittleEndian.Uint32(key[16:]))
h0 = uint32(t)
t = uint64(h1) + uint64(binary.LittleEndian.Uint32(key[20:])) + (t >> 32)
h1 = uint32(t)
t = uint64(h2) + uint64(binary.LittleEndian.Uint32(key[24:])) + (t >> 32)
h2 = uint32(t)
t = uint64(h3) + uint64(binary.LittleEndian.Uint32(key[28:])) + (t >> 32)
h3 = uint32(t)
binary.LittleEndian.PutUint32(out[0:], h0)
binary.LittleEndian.PutUint32(out[4:], h1)
binary.LittleEndian.PutUint32(out[8:], h2)
binary.LittleEndian.PutUint32(out[12:], h3)
}
+6 -16
View File
@@ -6,16 +6,9 @@
package poly1305
// hasVectorFacility reports whether the machine supports
// the vector facility (vx).
func hasVectorFacility() bool
// hasVMSLFacility reports whether the machine supports
// Vector Multiply Sum Logical (VMSL).
func hasVMSLFacility() bool
var hasVX = hasVectorFacility()
var hasVMSL = hasVMSLFacility()
import (
"golang.org/x/sys/cpu"
)
// poly1305vx is an assembly implementation of Poly1305 that uses vector
// instructions. It must only be called if the vector facility (vx) is
@@ -29,16 +22,13 @@ func poly1305vx(out *[16]byte, m *byte, mlen uint64, key *[32]byte)
//go:noescape
func poly1305vmsl(out *[16]byte, m *byte, mlen uint64, key *[32]byte)
// Sum generates an authenticator for m using a one-time key and puts the
// 16-byte result into out. Authenticating two different messages with the same
// key allows an attacker to forge messages at will.
func Sum(out *[16]byte, m []byte, key *[32]byte) {
if hasVX {
func sum(out *[16]byte, m []byte, key *[32]byte) {
if cpu.S390X.HasVX {
var mPtr *byte
if len(m) > 0 {
mPtr = &m[0]
}
if hasVMSL && len(m) > 256 {
if cpu.S390X.HasVXE && len(m) > 256 {
poly1305vmsl(out, mPtr, uint64(len(m)), key)
} else {
poly1305vx(out, mPtr, uint64(len(m)), key)
-22
View File
@@ -376,25 +376,3 @@ b1:
MOVD $0, R3
BR multiply
TEXT ·hasVectorFacility(SB), NOSPLIT, $24-1
MOVD $x-24(SP), R1
XC $24, 0(R1), 0(R1) // clear the storage
MOVD $2, R0 // R0 is the number of double words stored -1
WORD $0xB2B01000 // STFLE 0(R1)
XOR R0, R0 // reset the value of R0
MOVBZ z-8(SP), R1
AND $0x40, R1
BEQ novector
vectorinstalled:
// check if the vector instruction has been enabled
VLEIB $0, $0xF, V16
VLGVB $0, V16, R1
CMPBNE R1, $0xF, novector
MOVB $1, ret+0(FP) // have vx
RET
novector:
MOVB $0, ret+0(FP) // no vx
RET
-22
View File
@@ -907,25 +907,3 @@ square:
MULTIPLY(H0_0, H1_0, H2_0, H0_1, H1_1, H2_1, R_0, R_1, R_2, R5_1, R5_2, M0, M1, M2, M3, M4, M5, T_0, T_1, T_2, T_3, T_4, T_5, T_6, T_7, T_8, T_9)
REDUCE2(H0_0, H1_0, H2_0, M0, M1, M2, M3, M4, T_9, T_10, H0_1, M5)
BR next
TEXT ·hasVMSLFacility(SB), NOSPLIT, $24-1
MOVD $x-24(SP), R1
XC $24, 0(R1), 0(R1) // clear the storage
MOVD $2, R0 // R0 is the number of double words stored -1
WORD $0xB2B01000 // STFLE 0(R1)
XOR R0, R0 // reset the value of R0
MOVBZ z-8(SP), R1
AND $0x01, R1
BEQ novmsl
vectorinstalled:
// check if the vector instruction has been enabled
VLEIB $0, $0xF, V16
VLGVB $0, V16, R1
CMPBNE R1, $0xF, novmsl
MOVB $1, ret+0(FP) // have vx
RET
novmsl:
MOVB $0, ret+0(FP) // no vx
RET
+3
View File
@@ -541,6 +541,9 @@ func ServeAgent(agent Agent, c io.ReadWriter) error {
return err
}
l := binary.BigEndian.Uint32(length[:])
if l == 0 {
return fmt.Errorf("agent: request size is 0")
}
if l > maxAgentResponseBytes {
// We also cap requests.
return fmt.Errorf("agent: request too large: %d", l)
+23 -12
View File
@@ -17,12 +17,14 @@ import (
// These constants from [PROTOCOL.certkeys] represent the algorithm names
// for certificate types supported by this package.
const (
CertAlgoRSAv01 = "[email protected]"
CertAlgoDSAv01 = "[email protected]"
CertAlgoECDSA256v01 = "[email protected]"
CertAlgoECDSA384v01 = "[email protected]"
CertAlgoECDSA521v01 = "[email protected]"
CertAlgoED25519v01 = "ssh-ed25519[email protected]"
CertAlgoRSAv01 = "[email protected]"
CertAlgoDSAv01 = "[email protected]"
CertAlgoECDSA256v01 = "[email protected]"
CertAlgoECDSA384v01 = "[email protected]"
CertAlgoECDSA521v01 = "[email protected]"
CertAlgoSKECDSA256v01 = "sk-ecdsa-sha2-nistp256[email protected]"
CertAlgoED25519v01 = "[email protected]"
CertAlgoSKED25519v01 = "[email protected]"
)
// Certificate types distinguish between host and user
@@ -37,6 +39,7 @@ const (
type Signature struct {
Format string
Blob []byte
Rest []byte `ssh:"rest"`
}
// CertTimeInfinity can be used for OpenSSHCertV01.ValidBefore to indicate that
@@ -429,12 +432,14 @@ func (c *Certificate) SignCert(rand io.Reader, authority Signer) error {
}
var certAlgoNames = map[string]string{
KeyAlgoRSA: CertAlgoRSAv01,
KeyAlgoDSA: CertAlgoDSAv01,
KeyAlgoECDSA256: CertAlgoECDSA256v01,
KeyAlgoECDSA384: CertAlgoECDSA384v01,
KeyAlgoECDSA521: CertAlgoECDSA521v01,
KeyAlgoED25519: CertAlgoED25519v01,
KeyAlgoRSA: CertAlgoRSAv01,
KeyAlgoDSA: CertAlgoDSAv01,
KeyAlgoECDSA256: CertAlgoECDSA256v01,
KeyAlgoECDSA384: CertAlgoECDSA384v01,
KeyAlgoECDSA521: CertAlgoECDSA521v01,
KeyAlgoSKECDSA256: CertAlgoSKECDSA256v01,
KeyAlgoED25519: CertAlgoED25519v01,
KeyAlgoSKED25519: CertAlgoSKED25519v01,
}
// certToPrivAlgo returns the underlying algorithm for a certificate algorithm.
@@ -518,6 +523,12 @@ func parseSignatureBody(in []byte) (out *Signature, rest []byte, ok bool) {
return
}
switch out.Format {
case KeyAlgoSKECDSA256, CertAlgoSKECDSA256v01, KeyAlgoSKED25519, CertAlgoSKED25519v01:
out.Rest = in
return out, nil, ok
}
return out, in, ok
}
+43 -32
View File
@@ -16,9 +16,8 @@ import (
"hash"
"io"
"io/ioutil"
"math/bits"
"golang.org/x/crypto/internal/chacha20"
"golang.org/x/crypto/chacha20"
"golang.org/x/crypto/poly1305"
)
@@ -149,8 +148,8 @@ type streamPacketCipher struct {
macResult []byte
}
// readPacket reads and decrypt a single packet from the reader argument.
func (s *streamPacketCipher) readPacket(seqNum uint32, r io.Reader) ([]byte, error) {
// readCipherPacket reads and decrypt a single packet from the reader argument.
func (s *streamPacketCipher) readCipherPacket(seqNum uint32, r io.Reader) ([]byte, error) {
if _, err := io.ReadFull(r, s.prefix[:]); err != nil {
return nil, err
}
@@ -221,8 +220,8 @@ func (s *streamPacketCipher) readPacket(seqNum uint32, r io.Reader) ([]byte, err
return s.packetData[:length-paddingLength-1], nil
}
// writePacket encrypts and sends a packet of data to the writer argument
func (s *streamPacketCipher) writePacket(seqNum uint32, w io.Writer, rand io.Reader, packet []byte) error {
// writeCipherPacket encrypts and sends a packet of data to the writer argument
func (s *streamPacketCipher) writeCipherPacket(seqNum uint32, w io.Writer, rand io.Reader, packet []byte) error {
if len(packet) > maxPacket {
return errors.New("ssh: packet too large")
}
@@ -327,7 +326,7 @@ func newGCMCipher(key, iv, unusedMacKey []byte, unusedAlgs directionAlgorithms)
const gcmTagSize = 16
func (c *gcmCipher) writePacket(seqNum uint32, w io.Writer, rand io.Reader, packet []byte) error {
func (c *gcmCipher) writeCipherPacket(seqNum uint32, w io.Writer, rand io.Reader, packet []byte) error {
// Pad out to multiple of 16 bytes. This is different from the
// stream cipher because that encrypts the length too.
padding := byte(packetSizeMultiple - (1+len(packet))%packetSizeMultiple)
@@ -370,7 +369,7 @@ func (c *gcmCipher) incIV() {
}
}
func (c *gcmCipher) readPacket(seqNum uint32, r io.Reader) ([]byte, error) {
func (c *gcmCipher) readCipherPacket(seqNum uint32, r io.Reader) ([]byte, error) {
if _, err := io.ReadFull(r, c.prefix[:]); err != nil {
return nil, err
}
@@ -486,8 +485,8 @@ type cbcError string
func (e cbcError) Error() string { return string(e) }
func (c *cbcCipher) readPacket(seqNum uint32, r io.Reader) ([]byte, error) {
p, err := c.readPacketLeaky(seqNum, r)
func (c *cbcCipher) readCipherPacket(seqNum uint32, r io.Reader) ([]byte, error) {
p, err := c.readCipherPacketLeaky(seqNum, r)
if err != nil {
if _, ok := err.(cbcError); ok {
// Verification error: read a fixed amount of
@@ -500,7 +499,7 @@ func (c *cbcCipher) readPacket(seqNum uint32, r io.Reader) ([]byte, error) {
return p, err
}
func (c *cbcCipher) readPacketLeaky(seqNum uint32, r io.Reader) ([]byte, error) {
func (c *cbcCipher) readCipherPacketLeaky(seqNum uint32, r io.Reader) ([]byte, error) {
blockSize := c.decrypter.BlockSize()
// Read the header, which will include some of the subsequent data in the
@@ -576,7 +575,7 @@ func (c *cbcCipher) readPacketLeaky(seqNum uint32, r io.Reader) ([]byte, error)
return c.packetData[prefixLen:paddingStart], nil
}
func (c *cbcCipher) writePacket(seqNum uint32, w io.Writer, rand io.Reader, packet []byte) error {
func (c *cbcCipher) writeCipherPacket(seqNum uint32, w io.Writer, rand io.Reader, packet []byte) error {
effectiveBlockSize := maxUInt32(cbcMinPacketSizeMultiple, c.encrypter.BlockSize())
// Length of encrypted portion of the packet (header, payload, padding).
@@ -642,8 +641,8 @@ const chacha20Poly1305ID = "[email protected]"
// 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
func (c *chacha20Poly1305Cipher) readCipherPacket(seqNum uint32, r io.Reader) ([]byte, error) {
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 {
@@ -723,12 +726,16 @@ func (c *chacha20Poly1305Cipher) readPacket(seqNum uint32, r io.Reader) ([]byte,
return plain, nil
}
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
func (c *chacha20Poly1305Cipher) writeCipherPacket(seqNum uint32, w io.Writer, rand io.Reader, payload []byte) error {
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
+114
View File
@@ -523,3 +523,117 @@ func (r *retryableAuthMethod) method() string {
func RetryableAuthMethod(auth AuthMethod, maxTries int) AuthMethod {
return &retryableAuthMethod{authMethod: auth, maxTries: maxTries}
}
// GSSAPIWithMICAuthMethod is an AuthMethod with "gssapi-with-mic" authentication.
// See RFC 4462 section 3
// gssAPIClient is implementation of the GSSAPIClient interface, see the definition of the interface for details.
// target is the server host you want to log in to.
func GSSAPIWithMICAuthMethod(gssAPIClient GSSAPIClient, target string) AuthMethod {
if gssAPIClient == nil {
panic("gss-api client must be not nil with enable gssapi-with-mic")
}
return &gssAPIWithMICCallback{gssAPIClient: gssAPIClient, target: target}
}
type gssAPIWithMICCallback struct {
gssAPIClient GSSAPIClient
target string
}
func (g *gssAPIWithMICCallback) auth(session []byte, user string, c packetConn, rand io.Reader) (authResult, []string, error) {
m := &userAuthRequestMsg{
User: user,
Service: serviceSSH,
Method: g.method(),
}
// The GSS-API authentication method is initiated when the client sends an SSH_MSG_USERAUTH_REQUEST.
// See RFC 4462 section 3.2.
m.Payload = appendU32(m.Payload, 1)
m.Payload = appendString(m.Payload, string(krb5OID))
if err := c.writePacket(Marshal(m)); err != nil {
return authFailure, nil, err
}
// The server responds to the SSH_MSG_USERAUTH_REQUEST with either an
// SSH_MSG_USERAUTH_FAILURE if none of the mechanisms are supported or
// with an SSH_MSG_USERAUTH_GSSAPI_RESPONSE.
// See RFC 4462 section 3.3.
// OpenSSH supports Kerberos V5 mechanism only for GSS-API authentication,so I don't want to check
// selected mech if it is valid.
packet, err := c.readPacket()
if err != nil {
return authFailure, nil, err
}
userAuthGSSAPIResp := &userAuthGSSAPIResponse{}
if err := Unmarshal(packet, userAuthGSSAPIResp); err != nil {
return authFailure, nil, err
}
// Start the loop into the exchange token.
// See RFC 4462 section 3.4.
var token []byte
defer g.gssAPIClient.DeleteSecContext()
for {
// Initiates the establishment of a security context between the application and a remote peer.
nextToken, needContinue, err := g.gssAPIClient.InitSecContext("host@"+g.target, token, false)
if err != nil {
return authFailure, nil, err
}
if len(nextToken) > 0 {
if err := c.writePacket(Marshal(&userAuthGSSAPIToken{
Token: nextToken,
})); err != nil {
return authFailure, nil, err
}
}
if !needContinue {
break
}
packet, err = c.readPacket()
if err != nil {
return authFailure, nil, err
}
switch packet[0] {
case msgUserAuthFailure:
var msg userAuthFailureMsg
if err := Unmarshal(packet, &msg); err != nil {
return authFailure, nil, err
}
if msg.PartialSuccess {
return authPartialSuccess, msg.Methods, nil
}
return authFailure, msg.Methods, nil
case msgUserAuthGSSAPIError:
userAuthGSSAPIErrorResp := &userAuthGSSAPIError{}
if err := Unmarshal(packet, userAuthGSSAPIErrorResp); err != nil {
return authFailure, nil, err
}
return authFailure, nil, fmt.Errorf("GSS-API Error:\n"+
"Major Status: %d\n"+
"Minor Status: %d\n"+
"Error Message: %s\n", userAuthGSSAPIErrorResp.MajorStatus, userAuthGSSAPIErrorResp.MinorStatus,
userAuthGSSAPIErrorResp.Message)
case msgUserAuthGSSAPIToken:
userAuthGSSAPITokenReq := &userAuthGSSAPIToken{}
if err := Unmarshal(packet, userAuthGSSAPITokenReq); err != nil {
return authFailure, nil, err
}
token = userAuthGSSAPITokenReq.Token
}
}
// Binding Encryption Keys.
// See RFC 4462 section 3.5.
micField := buildMIC(string(session), user, "ssh-connection", "gssapi-with-mic")
micToken, err := g.gssAPIClient.GetMIC(micField)
if err != nil {
return authFailure, nil, err
}
if err := c.writePacket(Marshal(&userAuthGSSAPIMIC{
MIC: micToken,
})); err != nil {
return authFailure, nil, err
}
return handleAuthResponse(c)
}
func (g *gssAPIWithMICCallback) method() string {
return "gssapi-with-mic"
}
Loaded 50 of 466 files, more files were not shown because too many files have changed in this diff. Show more