Grand test fixup (#138)

* start fixing up tests

* fix up tests + automate with drone

* fiddle with linting

* messing about with drone.yml

* some more fiddling

* hmmm

* add cache

* add vendor directory

* verbose

* ci updates

* update some little things

* update sig

vendor/golang.org/x/crypto/AUTHORS

@@ -0,0 +1,3 @@
+# This source code refers to The Go Authors for copyright purposes.
+# The master list of authors is in the main Go distribution,
+# visible at https://tip.golang.org/AUTHORS.

vendor/golang.org/x/crypto/CONTRIBUTORS

@@ -0,0 +1,3 @@
+# This source code was written by the Go contributors.
+# The master list of contributors is in the main Go distribution,
+# visible at https://tip.golang.org/CONTRIBUTORS.

vendor/golang.org/x/crypto/LICENSE

@@ -0,0 +1,27 @@
+Copyright (c) 2009 The Go Authors. All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+
+   * Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+   * Redistributions in binary form must reproduce the above
+copyright notice, this list of conditions and the following disclaimer
+in the documentation and/or other materials provided with the
+distribution.
+   * Neither the name of Google Inc. nor the names of its
+contributors may be used to endorse or promote products derived from
+this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

vendor/golang.org/x/crypto/PATENTS

@@ -0,0 +1,22 @@
+Additional IP Rights Grant (Patents)
+
+"This implementation" means the copyrightable works distributed by
+Google as part of the Go project.
+
+Google hereby grants to You a perpetual, worldwide, non-exclusive,
+no-charge, royalty-free, irrevocable (except as stated in this section)
+patent license to make, have made, use, offer to sell, sell, import,
+transfer and otherwise run, modify and propagate the contents of this
+implementation of Go, where such license applies only to those patent
+claims, both currently owned or controlled by Google and acquired in
+the future, licensable by Google that are necessarily infringed by this
+implementation of Go.  This grant does not include claims that would be
+infringed only as a consequence of further modification of this
+implementation.  If you or your agent or exclusive licensee institute or
+order or agree to the institution of patent litigation against any
+entity (including a cross-claim or counterclaim in a lawsuit) alleging
+that this implementation of Go or any code incorporated within this
+implementation of Go constitutes direct or contributory patent
+infringement, or inducement of patent infringement, then any patent
+rights granted to you under this License for this implementation of Go
+shall terminate as of the date such litigation is filed.

vendor/golang.org/x/crypto/acme/autocert/cache.go

@@ -0,0 +1,136 @@
+// 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 autocert
+
+import (
+	"context"
+	"errors"
+	"io/ioutil"
+	"os"
+	"path/filepath"
+)
+
+// ErrCacheMiss is returned when a certificate is not found in cache.
+var ErrCacheMiss = errors.New("acme/autocert: certificate cache miss")
+
+// Cache is used by Manager to store and retrieve previously obtained certificates
+// and other account data as opaque blobs.
+//
+// Cache implementations should not rely on the key naming pattern. Keys can
+// include any printable ASCII characters, except the following: \/:*?"<>|
+type Cache interface {
+	// Get returns a certificate data for the specified key.
+	// If there's no such key, Get returns ErrCacheMiss.
+	Get(ctx context.Context, key string) ([]byte, error)
+
+	// Put stores the data in the cache under the specified key.
+	// Underlying implementations may use any data storage format,
+	// as long as the reverse operation, Get, results in the original data.
+	Put(ctx context.Context, key string, data []byte) error
+
+	// Delete removes a certificate data from the cache under the specified key.
+	// If there's no such key in the cache, Delete returns nil.
+	Delete(ctx context.Context, key string) error
+}
+
+// DirCache implements Cache using a directory on the local filesystem.
+// If the directory does not exist, it will be created with 0700 permissions.
+type DirCache string
+
+// Get reads a certificate data from the specified file name.
+func (d DirCache) Get(ctx context.Context, name string) ([]byte, error) {
+	name = filepath.Join(string(d), name)
+	var (
+		data []byte
+		err  error
+		done = make(chan struct{})
+	)
+	go func() {
+		data, err = ioutil.ReadFile(name)
+		close(done)
+	}()
+	select {
+	case <-ctx.Done():
+		return nil, ctx.Err()
+	case <-done:
+	}
+	if os.IsNotExist(err) {
+		return nil, ErrCacheMiss
+	}
+	return data, err
+}
+
+// Put writes the certificate data to the specified file name.
+// The file will be created with 0600 permissions.
+func (d DirCache) Put(ctx context.Context, name string, data []byte) error {
+	if err := os.MkdirAll(string(d), 0700); err != nil {
+		return err
+	}
+
+	done := make(chan struct{})
+	var err error
+	go func() {
+		defer close(done)
+		var tmp string
+		if tmp, err = d.writeTempFile(name, data); err != nil {
+			return
+		}
+		defer os.Remove(tmp)
+		select {
+		case <-ctx.Done():
+			// Don't overwrite the file if the context was canceled.
+		default:
+			newName := filepath.Join(string(d), name)
+			err = os.Rename(tmp, newName)
+		}
+	}()
+	select {
+	case <-ctx.Done():
+		return ctx.Err()
+	case <-done:
+	}
+	return err
+}
+
+// Delete removes the specified file name.
+func (d DirCache) Delete(ctx context.Context, name string) error {
+	name = filepath.Join(string(d), name)
+	var (
+		err  error
+		done = make(chan struct{})
+	)
+	go func() {
+		err = os.Remove(name)
+		close(done)
+	}()
+	select {
+	case <-ctx.Done():
+		return ctx.Err()
+	case <-done:
+	}
+	if err != nil && !os.IsNotExist(err) {
+		return err
+	}
+	return nil
+}
+
+// writeTempFile writes b to a temporary file, closes the file and returns its path.
+func (d DirCache) writeTempFile(prefix string, b []byte) (name string, reterr error) {
+	// TempFile uses 0600 permissions
+	f, err := ioutil.TempFile(string(d), prefix)
+	if err != nil {
+		return "", err
+	}
+	defer func() {
+		if reterr != nil {
+			os.Remove(f.Name())
+		}
+	}()
+	if _, err := f.Write(b); err != nil {
+		f.Close()
+		return "", err
+	}
+	return f.Name(), f.Close()
+}

vendor/golang.org/x/crypto/acme/autocert/listener.go

@@ -0,0 +1,155 @@
+// Copyright 2017 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package autocert
+
+import (
+	"crypto/tls"
+	"log"
+	"net"
+	"os"
+	"path/filepath"
+	"runtime"
+	"time"
+)
+
+// NewListener returns a net.Listener that listens on the standard TLS
+// port (443) on all interfaces and returns *tls.Conn connections with
+// LetsEncrypt certificates for the provided domain or domains.
+//
+// It enables one-line HTTPS servers:
+//
+//     log.Fatal(http.Serve(autocert.NewListener("example.com"), handler))
+//
+// NewListener is a convenience function for a common configuration.
+// More complex or custom configurations can use the autocert.Manager
+// type instead.
+//
+// Use of this function implies acceptance of the LetsEncrypt Terms of
+// Service. If domains is not empty, the provided domains are passed
+// to HostWhitelist. If domains is empty, the listener will do
+// LetsEncrypt challenges for any requested domain, which is not
+// recommended.
+//
+// Certificates are cached in a "golang-autocert" directory under an
+// operating system-specific cache or temp directory. This may not
+// be suitable for servers spanning multiple machines.
+//
+// The returned listener uses a *tls.Config that enables HTTP/2, and
+// should only be used with servers that support HTTP/2.
+//
+// The returned Listener also enables TCP keep-alives on the accepted
+// connections. The returned *tls.Conn are returned before their TLS
+// handshake has completed.
+func NewListener(domains ...string) net.Listener {
+	m := &Manager{
+		Prompt: AcceptTOS,
+	}
+	if len(domains) > 0 {
+		m.HostPolicy = HostWhitelist(domains...)
+	}
+	dir := cacheDir()
+	if err := os.MkdirAll(dir, 0700); err != nil {
+		log.Printf("warning: autocert.NewListener not using a cache: %v", err)
+	} else {
+		m.Cache = DirCache(dir)
+	}
+	return m.Listener()
+}
+
+// Listener listens on the standard TLS port (443) on all interfaces
+// and returns a net.Listener returning *tls.Conn connections.
+//
+// The returned listener uses a *tls.Config that enables HTTP/2, and
+// should only be used with servers that support HTTP/2.
+//
+// The returned Listener also enables TCP keep-alives on the accepted
+// connections. The returned *tls.Conn are returned before their TLS
+// handshake has completed.
+//
+// Unlike NewListener, it is the caller's responsibility to initialize
+// the Manager m's Prompt, Cache, HostPolicy, and other desired options.
+func (m *Manager) Listener() net.Listener {
+	ln := &listener{
+		conf: m.TLSConfig(),
+	}
+	ln.tcpListener, ln.tcpListenErr = net.Listen("tcp", ":443")
+	return ln
+}
+
+type listener struct {
+	conf *tls.Config
+
+	tcpListener  net.Listener
+	tcpListenErr error
+}
+
+func (ln *listener) Accept() (net.Conn, error) {
+	if ln.tcpListenErr != nil {
+		return nil, ln.tcpListenErr
+	}
+	conn, err := ln.tcpListener.Accept()
+	if err != nil {
+		return nil, err
+	}
+	tcpConn := conn.(*net.TCPConn)
+
+	// Because Listener is a convenience function, help out with
+	// this too.  This is not possible for the caller to set once
+	// we return a *tcp.Conn wrapping an inaccessible net.Conn.
+	// If callers don't want this, they can do things the manual
+	// way and tweak as needed. But this is what net/http does
+	// itself, so copy that. If net/http changes, we can change
+	// here too.
+	tcpConn.SetKeepAlive(true)
+	tcpConn.SetKeepAlivePeriod(3 * time.Minute)
+
+	return tls.Server(tcpConn, ln.conf), nil
+}
+
+func (ln *listener) Addr() net.Addr {
+	if ln.tcpListener != nil {
+		return ln.tcpListener.Addr()
+	}
+	// net.Listen failed. Return something non-nil in case callers
+	// call Addr before Accept:
+	return &net.TCPAddr{IP: net.IP{0, 0, 0, 0}, Port: 443}
+}
+
+func (ln *listener) Close() error {
+	if ln.tcpListenErr != nil {
+		return ln.tcpListenErr
+	}
+	return ln.tcpListener.Close()
+}
+
+func homeDir() string {
+	if runtime.GOOS == "windows" {
+		return os.Getenv("HOMEDRIVE") + os.Getenv("HOMEPATH")
+	}
+	if h := os.Getenv("HOME"); h != "" {
+		return h
+	}
+	return "/"
+}
+
+func cacheDir() string {
+	const base = "golang-autocert"
+	switch runtime.GOOS {
+	case "darwin":
+		return filepath.Join(homeDir(), "Library", "Caches", base)
+	case "windows":
+		for _, ev := range []string{"APPDATA", "CSIDL_APPDATA", "TEMP", "TMP"} {
+			if v := os.Getenv(ev); v != "" {
+				return filepath.Join(v, base)
+			}
+		}
+		// Worst case:
+		return filepath.Join(homeDir(), base)
+	}
+	if xdg := os.Getenv("XDG_CACHE_HOME"); xdg != "" {
+		return filepath.Join(xdg, base)
+	}
+	return filepath.Join(homeDir(), ".cache", base)
+}

vendor/golang.org/x/crypto/acme/autocert/renewal.go

@@ -0,0 +1,141 @@
+// 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 autocert
+
+import (
+	"context"
+	"crypto"
+	"sync"
+	"time"
+)
+
+// renewJitter is the maximum deviation from Manager.RenewBefore.
+const renewJitter = time.Hour
+
+// domainRenewal tracks the state used by the periodic timers
+// renewing a single domain's cert.
+type domainRenewal struct {
+	m   *Manager
+	ck  certKey
+	key crypto.Signer
+
+	timerMu sync.Mutex
+	timer   *time.Timer
+}
+
+// start starts a cert renewal timer at the time
+// defined by the certificate expiration time exp.
+//
+// If the timer is already started, calling start is a noop.
+func (dr *domainRenewal) start(exp time.Time) {
+	dr.timerMu.Lock()
+	defer dr.timerMu.Unlock()
+	if dr.timer != nil {
+		return
+	}
+	dr.timer = time.AfterFunc(dr.next(exp), dr.renew)
+}
+
+// stop stops the cert renewal timer.
+// If the timer is already stopped, calling stop is a noop.
+func (dr *domainRenewal) stop() {
+	dr.timerMu.Lock()
+	defer dr.timerMu.Unlock()
+	if dr.timer == nil {
+		return
+	}
+	dr.timer.Stop()
+	dr.timer = nil
+}
+
+// renew is called periodically by a timer.
+// The first renew call is kicked off by dr.start.
+func (dr *domainRenewal) renew() {
+	dr.timerMu.Lock()
+	defer dr.timerMu.Unlock()
+	if dr.timer == nil {
+		return
+	}
+
+	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Minute)
+	defer cancel()
+	// TODO: rotate dr.key at some point?
+	next, err := dr.do(ctx)
+	if err != nil {
+		next = renewJitter / 2
+		next += time.Duration(pseudoRand.int63n(int64(next)))
+	}
+	dr.timer = time.AfterFunc(next, dr.renew)
+	testDidRenewLoop(next, err)
+}
+
+// updateState locks and replaces the relevant Manager.state item with the given
+// state. It additionally updates dr.key with the given state's key.
+func (dr *domainRenewal) updateState(state *certState) {
+	dr.m.stateMu.Lock()
+	defer dr.m.stateMu.Unlock()
+	dr.key = state.key
+	dr.m.state[dr.ck] = state
+}
+
+// do is similar to Manager.createCert but it doesn't lock a Manager.state item.
+// Instead, it requests a new certificate independently and, upon success,
+// replaces dr.m.state item with a new one and updates cache for the given domain.
+//
+// It may lock and update the Manager.state if the expiration date of the currently
+// cached cert is far enough in the future.
+//
+// The returned value is a time interval after which the renewal should occur again.
+func (dr *domainRenewal) do(ctx context.Context) (time.Duration, error) {
+	// a race is likely unavoidable in a distributed environment
+	// but we try nonetheless
+	if tlscert, err := dr.m.cacheGet(ctx, dr.ck); err == nil {
+		next := dr.next(tlscert.Leaf.NotAfter)
+		if next > dr.m.renewBefore()+renewJitter {
+			signer, ok := tlscert.PrivateKey.(crypto.Signer)
+			if ok {
+				state := &certState{
+					key:  signer,
+					cert: tlscert.Certificate,
+					leaf: tlscert.Leaf,
+				}
+				dr.updateState(state)
+				return next, nil
+			}
+		}
+	}
+
+	der, leaf, err := dr.m.authorizedCert(ctx, dr.key, dr.ck)
+	if err != nil {
+		return 0, err
+	}
+	state := &certState{
+		key:  dr.key,
+		cert: der,
+		leaf: leaf,
+	}
+	tlscert, err := state.tlscert()
+	if err != nil {
+		return 0, err
+	}
+	if err := dr.m.cachePut(ctx, dr.ck, tlscert); err != nil {
+		return 0, err
+	}
+	dr.updateState(state)
+	return dr.next(leaf.NotAfter), nil
+}
+
+func (dr *domainRenewal) next(expiry time.Time) time.Duration {
+	d := expiry.Sub(dr.m.now()) - dr.m.renewBefore()
+	// add a bit of randomness to renew deadline
+	n := pseudoRand.int63n(int64(renewJitter))
+	d -= time.Duration(n)
+	if d < 0 {
+		return 0
+	}
+	return d
+}
+
+var testDidRenewLoop = func(next time.Duration, err error) {}

vendor/golang.org/x/crypto/acme/http.go

@@ -0,0 +1,321 @@
+// 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.
+
+package acme
+
+import (
+	"bytes"
+	"context"
+	"crypto"
+	"crypto/rand"
+	"encoding/json"
+	"fmt"
+	"io/ioutil"
+	"math/big"
+	"net/http"
+	"strconv"
+	"strings"
+	"time"
+)
+
+// retryTimer encapsulates common logic for retrying unsuccessful requests.
+// It is not safe for concurrent use.
+type retryTimer struct {
+	// backoffFn provides backoff delay sequence for retries.
+	// See Client.RetryBackoff doc comment.
+	backoffFn func(n int, r *http.Request, res *http.Response) time.Duration
+	// n is the current retry attempt.
+	n int
+}
+
+func (t *retryTimer) inc() {
+	t.n++
+}
+
+// backoff pauses the current goroutine as described in Client.RetryBackoff.
+func (t *retryTimer) backoff(ctx context.Context, r *http.Request, res *http.Response) error {
+	d := t.backoffFn(t.n, r, res)
+	if d <= 0 {
+		return fmt.Errorf("acme: no more retries for %s; tried %d time(s)", r.URL, t.n)
+	}
+	wakeup := time.NewTimer(d)
+	defer wakeup.Stop()
+	select {
+	case <-ctx.Done():
+		return ctx.Err()
+	case <-wakeup.C:
+		return nil
+	}
+}
+
+func (c *Client) retryTimer() *retryTimer {
+	f := c.RetryBackoff
+	if f == nil {
+		f = defaultBackoff
+	}
+	return &retryTimer{backoffFn: f}
+}
+
+// defaultBackoff provides default Client.RetryBackoff implementation
+// using a truncated exponential backoff algorithm,
+// as described in Client.RetryBackoff.
+//
+// The n argument is always bounded between 1 and 30.
+// The returned value is always greater than 0.
+func defaultBackoff(n int, r *http.Request, res *http.Response) time.Duration {
+	const max = 10 * time.Second
+	var jitter time.Duration
+	if x, err := rand.Int(rand.Reader, big.NewInt(1000)); err == nil {
+		// Set the minimum to 1ms to avoid a case where
+		// an invalid Retry-After value is parsed into 0 below,
+		// resulting in the 0 returned value which would unintentionally
+		// stop the retries.
+		jitter = (1 + time.Duration(x.Int64())) * time.Millisecond
+	}
+	if v, ok := res.Header["Retry-After"]; ok {
+		return retryAfter(v[0]) + jitter
+	}
+
+	if n < 1 {
+		n = 1
+	}
+	if n > 30 {
+		n = 30
+	}
+	d := time.Duration(1<<uint(n-1))*time.Second + jitter
+	if d > max {
+		return max
+	}
+	return d
+}
+
+// retryAfter parses a Retry-After HTTP header value,
+// trying to convert v into an int (seconds) or use http.ParseTime otherwise.
+// It returns zero value if v cannot be parsed.
+func retryAfter(v string) time.Duration {
+	if i, err := strconv.Atoi(v); err == nil {
+		return time.Duration(i) * time.Second
+	}
+	t, err := http.ParseTime(v)
+	if err != nil {
+		return 0
+	}
+	return t.Sub(timeNow())
+}
+
+// resOkay is a function that reports whether the provided response is okay.
+// It is expected to keep the response body unread.
+type resOkay func(*http.Response) bool
+
+// wantStatus returns a function which reports whether the code
+// matches the status code of a response.
+func wantStatus(codes ...int) resOkay {
+	return func(res *http.Response) bool {
+		for _, code := range codes {
+			if code == res.StatusCode {
+				return true
+			}
+		}
+		return false
+	}
+}
+
+// get issues an unsigned GET request to the specified URL.
+// It returns a non-error value only when ok reports true.
+//
+// get retries unsuccessful attempts according to c.RetryBackoff
+// until the context is done or a non-retriable error is received.
+func (c *Client) get(ctx context.Context, url string, ok resOkay) (*http.Response, error) {
+	retry := c.retryTimer()
+	for {
+		req, err := http.NewRequest("GET", url, nil)
+		if err != nil {
+			return nil, err
+		}
+		res, err := c.doNoRetry(ctx, req)
+		switch {
+		case err != nil:
+			return nil, err
+		case ok(res):
+			return res, nil
+		case isRetriable(res.StatusCode):
+			retry.inc()
+			resErr := responseError(res)
+			res.Body.Close()
+			// Ignore the error value from retry.backoff
+			// and return the one from last retry, as received from the CA.
+			if retry.backoff(ctx, req, res) != nil {
+				return nil, resErr
+			}
+		default:
+			defer res.Body.Close()
+			return nil, responseError(res)
+		}
+	}
+}
+
+// postAsGet is POST-as-GET, a replacement for GET in RFC8555
+// as described in https://tools.ietf.org/html/rfc8555#section-6.3.
+// It makes a POST request in KID form with zero JWS payload.
+// See nopayload doc comments in jws.go.
+func (c *Client) postAsGet(ctx context.Context, url string, ok resOkay) (*http.Response, error) {
+	return c.post(ctx, nil, url, noPayload, ok)
+}
+
+// post issues a signed POST request in JWS format using the provided key
+// to the specified URL. If key is nil, c.Key is used instead.
+// It returns a non-error value only when ok reports true.
+//
+// post retries unsuccessful attempts according to c.RetryBackoff
+// until the context is done or a non-retriable error is received.
+// It uses postNoRetry to make individual requests.
+func (c *Client) post(ctx context.Context, key crypto.Signer, url string, body interface{}, ok resOkay) (*http.Response, error) {
+	retry := c.retryTimer()
+	for {
+		res, req, err := c.postNoRetry(ctx, key, url, body)
+		if err != nil {
+			return nil, err
+		}
+		if ok(res) {
+			return res, nil
+		}
+		resErr := responseError(res)
+		res.Body.Close()
+		switch {
+		// Check for bad nonce before isRetriable because it may have been returned
+		// with an unretriable response code such as 400 Bad Request.
+		case isBadNonce(resErr):
+			// Consider any previously stored nonce values to be invalid.
+			c.clearNonces()
+		case !isRetriable(res.StatusCode):
+			return nil, resErr
+		}
+		retry.inc()
+		// Ignore the error value from retry.backoff
+		// and return the one from last retry, as received from the CA.
+		if err := retry.backoff(ctx, req, res); err != nil {
+			return nil, resErr
+		}
+	}
+}
+
+// postNoRetry signs the body with the given key and POSTs it to the provided url.
+// It is used by c.post to retry unsuccessful attempts.
+// The body argument must be JSON-serializable.
+//
+// If key argument is nil, c.Key is used to sign the request.
+// If key argument is nil and c.accountKID returns a non-zero keyID,
+// the request is sent in KID form. Otherwise, JWK form is used.
+//
+// In practice, when interfacing with RFC-compliant CAs most requests are sent in KID form
+// and JWK is used only when KID is unavailable: new account endpoint and certificate
+// revocation requests authenticated by a cert key.
+// See jwsEncodeJSON for other details.
+func (c *Client) postNoRetry(ctx context.Context, key crypto.Signer, url string, body interface{}) (*http.Response, *http.Request, error) {
+	kid := noKeyID
+	if key == nil {
+		key = c.Key
+		kid = c.accountKID(ctx)
+	}
+	nonce, err := c.popNonce(ctx, url)
+	if err != nil {
+		return nil, nil, err
+	}
+	b, err := jwsEncodeJSON(body, key, kid, nonce, url)
+	if err != nil {
+		return nil, nil, err
+	}
+	req, err := http.NewRequest("POST", url, bytes.NewReader(b))
+	if err != nil {
+		return nil, nil, err
+	}
+	req.Header.Set("Content-Type", "application/jose+json")
+	res, err := c.doNoRetry(ctx, req)
+	if err != nil {
+		return nil, nil, err
+	}
+	c.addNonce(res.Header)
+	return res, req, nil
+}
+
+// doNoRetry issues a request req, replacing its context (if any) with ctx.
+func (c *Client) doNoRetry(ctx context.Context, req *http.Request) (*http.Response, error) {
+	req.Header.Set("User-Agent", c.userAgent())
+	res, err := c.httpClient().Do(req.WithContext(ctx))
+	if err != nil {
+		select {
+		case <-ctx.Done():
+			// Prefer the unadorned context error.
+			// (The acme package had tests assuming this, previously from ctxhttp's
+			// behavior, predating net/http supporting contexts natively)
+			// TODO(bradfitz): reconsider this in the future. But for now this
+			// requires no test updates.
+			return nil, ctx.Err()
+		default:
+			return nil, err
+		}
+	}
+	return res, nil
+}
+
+func (c *Client) httpClient() *http.Client {
+	if c.HTTPClient != nil {
+		return c.HTTPClient
+	}
+	return http.DefaultClient
+}
+
+// packageVersion is the version of the module that contains this package, for
+// sending as part of the User-Agent header. It's set in version_go112.go.
+var packageVersion string
+
+// userAgent returns the User-Agent header value. It includes the package name,
+// the module version (if available), and the c.UserAgent value (if set).
+func (c *Client) userAgent() string {
+	ua := "golang.org/x/crypto/acme"
+	if packageVersion != "" {
+		ua += "@" + packageVersion
+	}
+	if c.UserAgent != "" {
+		ua = c.UserAgent + " " + ua
+	}
+	return ua
+}
+
+// isBadNonce reports whether err is an ACME "badnonce" error.
+func isBadNonce(err error) bool {
+	// According to the spec badNonce is urn:ietf:params:acme:error:badNonce.
+	// However, ACME servers in the wild return their versions of the error.
+	// See https://tools.ietf.org/html/draft-ietf-acme-acme-02#section-5.4
+	// and https://github.com/letsencrypt/boulder/blob/0e07eacb/docs/acme-divergences.md#section-66.
+	ae, ok := err.(*Error)
+	return ok && strings.HasSuffix(strings.ToLower(ae.ProblemType), ":badnonce")
+}
+
+// isRetriable reports whether a request can be retried
+// based on the response status code.
+//
+// Note that a "bad nonce" error is returned with a non-retriable 400 Bad Request code.
+// Callers should parse the response and check with isBadNonce.
+func isRetriable(code int) bool {
+	return code <= 399 || code >= 500 || code == http.StatusTooManyRequests
+}
+
+// responseError creates an error of Error type from resp.
+func responseError(resp *http.Response) error {
+	// don't care if ReadAll returns an error:
+	// json.Unmarshal will fail in that case anyway
+	b, _ := ioutil.ReadAll(resp.Body)
+	e := &wireError{Status: resp.StatusCode}
+	if err := json.Unmarshal(b, e); err != nil {
+		// this is not a regular error response:
+		// populate detail with anything we received,
+		// e.Status will already contain HTTP response code value
+		e.Detail = string(b)
+		if e.Detail == "" {
+			e.Detail = resp.Status
+		}
+	}
+	return e.error(resp.Header)
+}

vendor/golang.org/x/crypto/acme/jws.go

@@ -0,0 +1,229 @@
+// Copyright 2015 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package acme
+
+import (
+	"crypto"
+	"crypto/ecdsa"
+	"crypto/hmac"
+	"crypto/rand"
+	"crypto/rsa"
+	"crypto/sha256"
+	_ "crypto/sha512" // need for EC keys
+	"encoding/asn1"
+	"encoding/base64"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"math/big"
+)
+
+// keyID is the account identity provided by a CA during registration.
+type keyID string
+
+// noKeyID indicates that jwsEncodeJSON should compute and use JWK instead of a KID.
+// See jwsEncodeJSON for details.
+const noKeyID = keyID("")
+
+// noPayload indicates jwsEncodeJSON will encode zero-length octet string
+// in a JWS request. This is called POST-as-GET in RFC 8555 and is used to make
+// authenticated GET requests via POSTing with an empty payload.
+// See https://tools.ietf.org/html/rfc8555#section-6.3 for more details.
+const noPayload = ""
+
+// jsonWebSignature can be easily serialized into a JWS following
+// https://tools.ietf.org/html/rfc7515#section-3.2.
+type jsonWebSignature struct {
+	Protected string `json:"protected"`
+	Payload   string `json:"payload"`
+	Sig       string `json:"signature"`
+}
+
+// jwsEncodeJSON signs claimset using provided key and a nonce.
+// The result is serialized in JSON format containing either kid or jwk
+// fields based on the provided keyID value.
+//
+// If kid is non-empty, its quoted value is inserted in the protected head
+// as "kid" field value. Otherwise, JWK is computed using jwkEncode and inserted
+// as "jwk" field value. The "jwk" and "kid" fields are mutually exclusive.
+//
+// See https://tools.ietf.org/html/rfc7515#section-7.
+func jwsEncodeJSON(claimset interface{}, key crypto.Signer, kid keyID, nonce, url string) ([]byte, error) {
+	alg, sha := jwsHasher(key.Public())
+	if alg == "" || !sha.Available() {
+		return nil, ErrUnsupportedKey
+	}
+	var phead string
+	switch kid {
+	case noKeyID:
+		jwk, err := jwkEncode(key.Public())
+		if err != nil {
+			return nil, err
+		}
+		phead = fmt.Sprintf(`{"alg":%q,"jwk":%s,"nonce":%q,"url":%q}`, alg, jwk, nonce, url)
+	default:
+		phead = fmt.Sprintf(`{"alg":%q,"kid":%q,"nonce":%q,"url":%q}`, alg, kid, nonce, url)
+	}
+	phead = base64.RawURLEncoding.EncodeToString([]byte(phead))
+	var payload string
+	if claimset != noPayload {
+		cs, err := json.Marshal(claimset)
+		if err != nil {
+			return nil, err
+		}
+		payload = base64.RawURLEncoding.EncodeToString(cs)
+	}
+	hash := sha.New()
+	hash.Write([]byte(phead + "." + payload))
+	sig, err := jwsSign(key, sha, hash.Sum(nil))
+	if err != nil {
+		return nil, err
+	}
+	enc := jsonWebSignature{
+		Protected: phead,
+		Payload:   payload,
+		Sig:       base64.RawURLEncoding.EncodeToString(sig),
+	}
+	return json.Marshal(&enc)
+}
+
+// jwsWithMAC creates and signs a JWS using the given key and the HS256
+// algorithm. kid and url are included in the protected header. rawPayload
+// should not be base64-URL-encoded.
+func jwsWithMAC(key []byte, kid, url string, rawPayload []byte) (*jsonWebSignature, error) {
+	if len(key) == 0 {
+		return nil, errors.New("acme: cannot sign JWS with an empty MAC key")
+	}
+	header := struct {
+		Algorithm string `json:"alg"`
+		KID       string `json:"kid"`
+		URL       string `json:"url,omitempty"`
+	}{
+		// Only HMAC-SHA256 is supported.
+		Algorithm: "HS256",
+		KID:       kid,
+		URL:       url,
+	}
+	rawProtected, err := json.Marshal(header)
+	if err != nil {
+		return nil, err
+	}
+	protected := base64.RawURLEncoding.EncodeToString(rawProtected)
+	payload := base64.RawURLEncoding.EncodeToString(rawPayload)
+
+	h := hmac.New(sha256.New, key)
+	if _, err := h.Write([]byte(protected + "." + payload)); err != nil {
+		return nil, err
+	}
+	mac := h.Sum(nil)
+
+	return &jsonWebSignature{
+		Protected: protected,
+		Payload:   payload,
+		Sig:       base64.RawURLEncoding.EncodeToString(mac),
+	}, nil
+}
+
+// jwkEncode encodes public part of an RSA or ECDSA key into a JWK.
+// The result is also suitable for creating a JWK thumbprint.
+// https://tools.ietf.org/html/rfc7517
+func jwkEncode(pub crypto.PublicKey) (string, error) {
+	switch pub := pub.(type) {
+	case *rsa.PublicKey:
+		// https://tools.ietf.org/html/rfc7518#section-6.3.1
+		n := pub.N
+		e := big.NewInt(int64(pub.E))
+		// Field order is important.
+		// See https://tools.ietf.org/html/rfc7638#section-3.3 for details.
+		return fmt.Sprintf(`{"e":"%s","kty":"RSA","n":"%s"}`,
+			base64.RawURLEncoding.EncodeToString(e.Bytes()),
+			base64.RawURLEncoding.EncodeToString(n.Bytes()),
+		), nil
+	case *ecdsa.PublicKey:
+		// https://tools.ietf.org/html/rfc7518#section-6.2.1
+		p := pub.Curve.Params()
+		n := p.BitSize / 8
+		if p.BitSize%8 != 0 {
+			n++
+		}
+		x := pub.X.Bytes()
+		if n > len(x) {
+			x = append(make([]byte, n-len(x)), x...)
+		}
+		y := pub.Y.Bytes()
+		if n > len(y) {
+			y = append(make([]byte, n-len(y)), y...)
+		}
+		// Field order is important.
+		// See https://tools.ietf.org/html/rfc7638#section-3.3 for details.
+		return fmt.Sprintf(`{"crv":"%s","kty":"EC","x":"%s","y":"%s"}`,
+			p.Name,
+			base64.RawURLEncoding.EncodeToString(x),
+			base64.RawURLEncoding.EncodeToString(y),
+		), nil
+	}
+	return "", ErrUnsupportedKey
+}
+
+// jwsSign signs the digest using the given key.
+// The hash is unused for ECDSA keys.
+func jwsSign(key crypto.Signer, hash crypto.Hash, digest []byte) ([]byte, error) {
+	switch pub := key.Public().(type) {
+	case *rsa.PublicKey:
+		return key.Sign(rand.Reader, digest, hash)
+	case *ecdsa.PublicKey:
+		sigASN1, err := key.Sign(rand.Reader, digest, hash)
+		if err != nil {
+			return nil, err
+		}
+
+		var rs struct{ R, S *big.Int }
+		if _, err := asn1.Unmarshal(sigASN1, &rs); err != nil {
+			return nil, err
+		}
+
+		rb, sb := rs.R.Bytes(), rs.S.Bytes()
+		size := pub.Params().BitSize / 8
+		if size%8 > 0 {
+			size++
+		}
+		sig := make([]byte, size*2)
+		copy(sig[size-len(rb):], rb)
+		copy(sig[size*2-len(sb):], sb)
+		return sig, nil
+	}
+	return nil, ErrUnsupportedKey
+}
+
+// jwsHasher indicates suitable JWS algorithm name and a hash function
+// to use for signing a digest with the provided key.
+// It returns ("", 0) if the key is not supported.
+func jwsHasher(pub crypto.PublicKey) (string, crypto.Hash) {
+	switch pub := pub.(type) {
+	case *rsa.PublicKey:
+		return "RS256", crypto.SHA256
+	case *ecdsa.PublicKey:
+		switch pub.Params().Name {
+		case "P-256":
+			return "ES256", crypto.SHA256
+		case "P-384":
+			return "ES384", crypto.SHA384
+		case "P-521":
+			return "ES512", crypto.SHA512
+		}
+	}
+	return "", 0
+}
+
+// JWKThumbprint creates a JWK thumbprint out of pub
+// as specified in https://tools.ietf.org/html/rfc7638.
+func JWKThumbprint(pub crypto.PublicKey) (string, error) {
+	jwk, err := jwkEncode(pub)
+	if err != nil {
+		return "", err
+	}
+	b := sha256.Sum256([]byte(jwk))
+	return base64.RawURLEncoding.EncodeToString(b[:]), nil
+}

vendor/golang.org/x/crypto/acme/rfc8555.go

@@ -0,0 +1,412 @@
+// 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.
+
+package acme
+
+import (
+	"context"
+	"crypto"
+	"encoding/base64"
+	"encoding/json"
+	"encoding/pem"
+	"errors"
+	"fmt"
+	"io"
+	"io/ioutil"
+	"net/http"
+	"time"
+)
+
+// DeactivateReg permanently disables an existing account associated with c.Key.
+// A deactivated account can no longer request certificate issuance or access
+// resources related to the account, such as orders or authorizations.
+//
+// It only works with CAs implementing RFC 8555.
+func (c *Client) DeactivateReg(ctx context.Context) error {
+	url := string(c.accountKID(ctx))
+	if url == "" {
+		return ErrNoAccount
+	}
+	req := json.RawMessage(`{"status": "deactivated"}`)
+	res, err := c.post(ctx, nil, url, req, wantStatus(http.StatusOK))
+	if err != nil {
+		return err
+	}
+	res.Body.Close()
+	return nil
+}
+
+// registerRFC is equivalent to c.Register but for CAs implementing RFC 8555.
+// It expects c.Discover to have already been called.
+func (c *Client) registerRFC(ctx context.Context, acct *Account, prompt func(tosURL string) bool) (*Account, error) {
+	c.cacheMu.Lock() // guard c.kid access
+	defer c.cacheMu.Unlock()
+
+	req := struct {
+		TermsAgreed            bool              `json:"termsOfServiceAgreed,omitempty"`
+		Contact                []string          `json:"contact,omitempty"`
+		ExternalAccountBinding *jsonWebSignature `json:"externalAccountBinding,omitempty"`
+	}{
+		Contact: acct.Contact,
+	}
+	if c.dir.Terms != "" {
+		req.TermsAgreed = prompt(c.dir.Terms)
+	}
+
+	// set 'externalAccountBinding' field if requested
+	if acct.ExternalAccountBinding != nil {
+		eabJWS, err := c.encodeExternalAccountBinding(acct.ExternalAccountBinding)
+		if err != nil {
+			return nil, fmt.Errorf("acme: failed to encode external account binding: %v", err)
+		}
+		req.ExternalAccountBinding = eabJWS
+	}
+
+	res, err := c.post(ctx, c.Key, c.dir.RegURL, req, wantStatus(
+		http.StatusOK,      // account with this key already registered
+		http.StatusCreated, // new account created
+	))
+	if err != nil {
+		return nil, err
+	}
+
+	defer res.Body.Close()
+	a, err := responseAccount(res)
+	if err != nil {
+		return nil, err
+	}
+	// Cache Account URL even if we return an error to the caller.
+	// It is by all means a valid and usable "kid" value for future requests.
+	c.kid = keyID(a.URI)
+	if res.StatusCode == http.StatusOK {
+		return nil, ErrAccountAlreadyExists
+	}
+	return a, nil
+}
+
+// encodeExternalAccountBinding will encode an external account binding stanza
+// as described in https://tools.ietf.org/html/rfc8555#section-7.3.4.
+func (c *Client) encodeExternalAccountBinding(eab *ExternalAccountBinding) (*jsonWebSignature, error) {
+	jwk, err := jwkEncode(c.Key.Public())
+	if err != nil {
+		return nil, err
+	}
+	return jwsWithMAC(eab.Key, eab.KID, c.dir.RegURL, []byte(jwk))
+}
+
+// updateRegRFC is equivalent to c.UpdateReg but for CAs implementing RFC 8555.
+// It expects c.Discover to have already been called.
+func (c *Client) updateRegRFC(ctx context.Context, a *Account) (*Account, error) {
+	url := string(c.accountKID(ctx))
+	if url == "" {
+		return nil, ErrNoAccount
+	}
+	req := struct {
+		Contact []string `json:"contact,omitempty"`
+	}{
+		Contact: a.Contact,
+	}
+	res, err := c.post(ctx, nil, url, req, wantStatus(http.StatusOK))
+	if err != nil {
+		return nil, err
+	}
+	defer res.Body.Close()
+	return responseAccount(res)
+}
+
+// getGegRFC is equivalent to c.GetReg but for CAs implementing RFC 8555.
+// It expects c.Discover to have already been called.
+func (c *Client) getRegRFC(ctx context.Context) (*Account, error) {
+	req := json.RawMessage(`{"onlyReturnExisting": true}`)
+	res, err := c.post(ctx, c.Key, c.dir.RegURL, req, wantStatus(http.StatusOK))
+	if e, ok := err.(*Error); ok && e.ProblemType == "urn:ietf:params:acme:error:accountDoesNotExist" {
+		return nil, ErrNoAccount
+	}
+	if err != nil {
+		return nil, err
+	}
+
+	defer res.Body.Close()
+	return responseAccount(res)
+}
+
+func responseAccount(res *http.Response) (*Account, error) {
+	var v struct {
+		Status  string
+		Contact []string
+		Orders  string
+	}
+	if err := json.NewDecoder(res.Body).Decode(&v); err != nil {
+		return nil, fmt.Errorf("acme: invalid account response: %v", err)
+	}
+	return &Account{
+		URI:       res.Header.Get("Location"),
+		Status:    v.Status,
+		Contact:   v.Contact,
+		OrdersURL: v.Orders,
+	}, nil
+}
+
+// AuthorizeOrder initiates the order-based application for certificate issuance,
+// as opposed to pre-authorization in Authorize.
+// It is only supported by CAs implementing RFC 8555.
+//
+// The caller then needs to fetch each authorization with GetAuthorization,
+// identify those with StatusPending status and fulfill a challenge using Accept.
+// Once all authorizations are satisfied, the caller will typically want to poll
+// order status using WaitOrder until it's in StatusReady state.
+// To finalize the order and obtain a certificate, the caller submits a CSR with CreateOrderCert.
+func (c *Client) AuthorizeOrder(ctx context.Context, id []AuthzID, opt ...OrderOption) (*Order, error) {
+	dir, err := c.Discover(ctx)
+	if err != nil {
+		return nil, err
+	}
+
+	req := struct {
+		Identifiers []wireAuthzID `json:"identifiers"`
+		NotBefore   string        `json:"notBefore,omitempty"`
+		NotAfter    string        `json:"notAfter,omitempty"`
+	}{}
+	for _, v := range id {
+		req.Identifiers = append(req.Identifiers, wireAuthzID{
+			Type:  v.Type,
+			Value: v.Value,
+		})
+	}
+	for _, o := range opt {
+		switch o := o.(type) {
+		case orderNotBeforeOpt:
+			req.NotBefore = time.Time(o).Format(time.RFC3339)
+		case orderNotAfterOpt:
+			req.NotAfter = time.Time(o).Format(time.RFC3339)
+		default:
+			// Package's fault if we let this happen.
+			panic(fmt.Sprintf("unsupported order option type %T", o))
+		}
+	}
+
+	res, err := c.post(ctx, nil, dir.OrderURL, req, wantStatus(http.StatusCreated))
+	if err != nil {
+		return nil, err
+	}
+	defer res.Body.Close()
+	return responseOrder(res)
+}
+
+// GetOrder retrives an order identified by the given URL.
+// For orders created with AuthorizeOrder, the url value is Order.URI.
+//
+// If a caller needs to poll an order until its status is final,
+// see the WaitOrder method.
+func (c *Client) GetOrder(ctx context.Context, url string) (*Order, error) {
+	if _, err := c.Discover(ctx); err != nil {
+		return nil, err
+	}
+
+	res, err := c.postAsGet(ctx, url, wantStatus(http.StatusOK))
+	if err != nil {
+		return nil, err
+	}
+	defer res.Body.Close()
+	return responseOrder(res)
+}
+
+// WaitOrder polls an order from the given URL until it is in one of the final states,
+// StatusReady, StatusValid or StatusInvalid, the CA responded with a non-retryable error
+// or the context is done.
+//
+// It returns a non-nil Order only if its Status is StatusReady or StatusValid.
+// In all other cases WaitOrder returns an error.
+// If the Status is StatusInvalid, the returned error is of type *OrderError.
+func (c *Client) WaitOrder(ctx context.Context, url string) (*Order, error) {
+	if _, err := c.Discover(ctx); err != nil {
+		return nil, err
+	}
+	for {
+		res, err := c.postAsGet(ctx, url, wantStatus(http.StatusOK))
+		if err != nil {
+			return nil, err
+		}
+		o, err := responseOrder(res)
+		res.Body.Close()
+		switch {
+		case err != nil:
+			// Skip and retry.
+		case o.Status == StatusInvalid:
+			return nil, &OrderError{OrderURL: o.URI, Status: o.Status}
+		case o.Status == StatusReady || o.Status == StatusValid:
+			return o, nil
+		}
+
+		d := retryAfter(res.Header.Get("Retry-After"))
+		if d == 0 {
+			// Default retry-after.
+			// Same reasoning as in WaitAuthorization.
+			d = time.Second
+		}
+		t := time.NewTimer(d)
+		select {
+		case <-ctx.Done():
+			t.Stop()
+			return nil, ctx.Err()
+		case <-t.C:
+			// Retry.
+		}
+	}
+}
+
+func responseOrder(res *http.Response) (*Order, error) {
+	var v struct {
+		Status         string
+		Expires        time.Time
+		Identifiers    []wireAuthzID
+		NotBefore      time.Time
+		NotAfter       time.Time
+		Error          *wireError
+		Authorizations []string
+		Finalize       string
+		Certificate    string
+	}
+	if err := json.NewDecoder(res.Body).Decode(&v); err != nil {
+		return nil, fmt.Errorf("acme: error reading order: %v", err)
+	}
+	o := &Order{
+		URI:         res.Header.Get("Location"),
+		Status:      v.Status,
+		Expires:     v.Expires,
+		NotBefore:   v.NotBefore,
+		NotAfter:    v.NotAfter,
+		AuthzURLs:   v.Authorizations,
+		FinalizeURL: v.Finalize,
+		CertURL:     v.Certificate,
+	}
+	for _, id := range v.Identifiers {
+		o.Identifiers = append(o.Identifiers, AuthzID{Type: id.Type, Value: id.Value})
+	}
+	if v.Error != nil {
+		o.Error = v.Error.error(nil /* headers */)
+	}
+	return o, nil
+}
+
+// CreateOrderCert submits the CSR (Certificate Signing Request) to a CA at the specified URL.
+// The URL is the FinalizeURL field of an Order created with AuthorizeOrder.
+//
+// If the bundle argument is true, the returned value also contain the CA (issuer)
+// certificate chain. Otherwise, only a leaf certificate is returned.
+// The returned URL can be used to re-fetch the certificate using FetchCert.
+//
+// This method is only supported by CAs implementing RFC 8555. See CreateCert for pre-RFC CAs.
+//
+// CreateOrderCert returns an error if the CA's response is unreasonably large.
+// Callers are encouraged to parse the returned value to ensure the certificate is valid and has the expected features.
+func (c *Client) CreateOrderCert(ctx context.Context, url string, csr []byte, bundle bool) (der [][]byte, certURL string, err error) {
+	if _, err := c.Discover(ctx); err != nil { // required by c.accountKID
+		return nil, "", err
+	}
+
+	// RFC describes this as "finalize order" request.
+	req := struct {
+		CSR string `json:"csr"`
+	}{
+		CSR: base64.RawURLEncoding.EncodeToString(csr),
+	}
+	res, err := c.post(ctx, nil, url, req, wantStatus(http.StatusOK))
+	if err != nil {
+		return nil, "", err
+	}
+	defer res.Body.Close()
+	o, err := responseOrder(res)
+	if err != nil {
+		return nil, "", err
+	}
+
+	// Wait for CA to issue the cert if they haven't.
+	if o.Status != StatusValid {
+		o, err = c.WaitOrder(ctx, o.URI)
+	}
+	if err != nil {
+		return nil, "", err
+	}
+	// The only acceptable status post finalize and WaitOrder is "valid".
+	if o.Status != StatusValid {
+		return nil, "", &OrderError{OrderURL: o.URI, Status: o.Status}
+	}
+	crt, err := c.fetchCertRFC(ctx, o.CertURL, bundle)
+	return crt, o.CertURL, err
+}
+
+// fetchCertRFC downloads issued certificate from the given URL.
+// It expects the CA to respond with PEM-encoded certificate chain.
+//
+// The URL argument is the CertURL field of Order.
+func (c *Client) fetchCertRFC(ctx context.Context, url string, bundle bool) ([][]byte, error) {
+	res, err := c.postAsGet(ctx, url, wantStatus(http.StatusOK))
+	if err != nil {
+		return nil, err
+	}
+	defer res.Body.Close()
+
+	// Get all the bytes up to a sane maximum.
+	// Account very roughly for base64 overhead.
+	const max = maxCertChainSize + maxCertChainSize/33
+	b, err := ioutil.ReadAll(io.LimitReader(res.Body, max+1))
+	if err != nil {
+		return nil, fmt.Errorf("acme: fetch cert response stream: %v", err)
+	}
+	if len(b) > max {
+		return nil, errors.New("acme: certificate chain is too big")
+	}
+
+	// Decode PEM chain.
+	var chain [][]byte
+	for {
+		var p *pem.Block
+		p, b = pem.Decode(b)
+		if p == nil {
+			break
+		}
+		if p.Type != "CERTIFICATE" {
+			return nil, fmt.Errorf("acme: invalid PEM cert type %q", p.Type)
+		}
+
+		chain = append(chain, p.Bytes)
+		if !bundle {
+			return chain, nil
+		}
+		if len(chain) > maxChainLen {
+			return nil, errors.New("acme: certificate chain is too long")
+		}
+	}
+	if len(chain) == 0 {
+		return nil, errors.New("acme: certificate chain is empty")
+	}
+	return chain, nil
+}
+
+// sends a cert revocation request in either JWK form when key is non-nil or KID form otherwise.
+func (c *Client) revokeCertRFC(ctx context.Context, key crypto.Signer, cert []byte, reason CRLReasonCode) error {
+	req := &struct {
+		Cert   string `json:"certificate"`
+		Reason int    `json:"reason"`
+	}{
+		Cert:   base64.RawURLEncoding.EncodeToString(cert),
+		Reason: int(reason),
+	}
+	res, err := c.post(ctx, key, c.dir.RevokeURL, req, wantStatus(http.StatusOK))
+	if err != nil {
+		if isAlreadyRevoked(err) {
+			// Assume it is not an error to revoke an already revoked cert.
+			return nil
+		}
+		return err
+	}
+	defer res.Body.Close()
+	return nil
+}
+
+func isAlreadyRevoked(err error) bool {
+	e, ok := err.(*Error)
+	return ok && e.ProblemType == "urn:ietf:params:acme:error:alreadyRevoked"
+}

vendor/golang.org/x/crypto/acme/types.go

@@ -0,0 +1,622 @@
+// 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 acme
+
+import (
+	"crypto"
+	"crypto/x509"
+	"errors"
+	"fmt"
+	"net/http"
+	"strings"
+	"time"
+)
+
+// ACME status values of Account, Order, Authorization and Challenge objects.
+// See https://tools.ietf.org/html/rfc8555#section-7.1.6 for details.
+const (
+	StatusDeactivated = "deactivated"
+	StatusExpired     = "expired"
+	StatusInvalid     = "invalid"
+	StatusPending     = "pending"
+	StatusProcessing  = "processing"
+	StatusReady       = "ready"
+	StatusRevoked     = "revoked"
+	StatusUnknown     = "unknown"
+	StatusValid       = "valid"
+)
+
+// CRLReasonCode identifies the reason for a certificate revocation.
+type CRLReasonCode int
+
+// CRL reason codes as defined in RFC 5280.
+const (
+	CRLReasonUnspecified          CRLReasonCode = 0
+	CRLReasonKeyCompromise        CRLReasonCode = 1
+	CRLReasonCACompromise         CRLReasonCode = 2
+	CRLReasonAffiliationChanged   CRLReasonCode = 3
+	CRLReasonSuperseded           CRLReasonCode = 4
+	CRLReasonCessationOfOperation CRLReasonCode = 5
+	CRLReasonCertificateHold      CRLReasonCode = 6
+	CRLReasonRemoveFromCRL        CRLReasonCode = 8
+	CRLReasonPrivilegeWithdrawn   CRLReasonCode = 9
+	CRLReasonAACompromise         CRLReasonCode = 10
+)
+
+var (
+	// ErrUnsupportedKey is returned when an unsupported key type is encountered.
+	ErrUnsupportedKey = errors.New("acme: unknown key type; only RSA and ECDSA are supported")
+
+	// ErrAccountAlreadyExists indicates that the Client's key has already been registered
+	// with the CA. It is returned by Register method.
+	ErrAccountAlreadyExists = errors.New("acme: account already exists")
+
+	// ErrNoAccount indicates that the Client's key has not been registered with the CA.
+	ErrNoAccount = errors.New("acme: account does not exist")
+)
+
+// A Subproblem describes an ACME subproblem as reported in an Error.
+type Subproblem struct {
+	// Type is a URI reference that identifies the problem type,
+	// typically in a "urn:acme:error:xxx" form.
+	Type string
+	// Detail is a human-readable explanation specific to this occurrence of the problem.
+	Detail string
+	// Instance indicates a URL that the client should direct a human user to visit
+	// in order for instructions on how to agree to the updated Terms of Service.
+	// In such an event CA sets StatusCode to 403, Type to
+	// "urn:ietf:params:acme:error:userActionRequired", and adds a Link header with relation
+	// "terms-of-service" containing the latest TOS URL.
+	Instance string
+	// Identifier may contain the ACME identifier that the error is for.
+	Identifier *AuthzID
+}
+
+func (sp Subproblem) String() string {
+	str := fmt.Sprintf("%s: ", sp.Type)
+	if sp.Identifier != nil {
+		str += fmt.Sprintf("[%s: %s] ", sp.Identifier.Type, sp.Identifier.Value)
+	}
+	str += sp.Detail
+	return str
+}
+
+// Error is an ACME error, defined in Problem Details for HTTP APIs doc
+// http://tools.ietf.org/html/draft-ietf-appsawg-http-problem.
+type Error struct {
+	// StatusCode is The HTTP status code generated by the origin server.
+	StatusCode int
+	// ProblemType is a URI reference that identifies the problem type,
+	// typically in a "urn:acme:error:xxx" form.
+	ProblemType string
+	// Detail is a human-readable explanation specific to this occurrence of the problem.
+	Detail string
+	// Instance indicates a URL that the client should direct a human user to visit
+	// in order for instructions on how to agree to the updated Terms of Service.
+	// In such an event CA sets StatusCode to 403, ProblemType to
+	// "urn:ietf:params:acme:error:userActionRequired" and a Link header with relation
+	// "terms-of-service" containing the latest TOS URL.
+	Instance string
+	// Header is the original server error response headers.
+	// It may be nil.
+	Header http.Header
+	// Subproblems may contain more detailed information about the individual problems
+	// that caused the error. This field is only sent by RFC 8555 compatible ACME
+	// servers. Defined in RFC 8555 Section 6.7.1.
+	Subproblems []Subproblem
+}
+
+func (e *Error) Error() string {
+	str := fmt.Sprintf("%d %s: %s", e.StatusCode, e.ProblemType, e.Detail)
+	if len(e.Subproblems) > 0 {
+		str += fmt.Sprintf("; subproblems:")
+		for _, sp := range e.Subproblems {
+			str += fmt.Sprintf("\n\t%s", sp)
+		}
+	}
+	return str
+}
+
+// AuthorizationError indicates that an authorization for an identifier
+// did not succeed.
+// It contains all errors from Challenge items of the failed Authorization.
+type AuthorizationError struct {
+	// URI uniquely identifies the failed Authorization.
+	URI string
+
+	// Identifier is an AuthzID.Value of the failed Authorization.
+	Identifier string
+
+	// Errors is a collection of non-nil error values of Challenge items
+	// of the failed Authorization.
+	Errors []error
+}
+
+func (a *AuthorizationError) Error() string {
+	e := make([]string, len(a.Errors))
+	for i, err := range a.Errors {
+		e[i] = err.Error()
+	}
+
+	if a.Identifier != "" {
+		return fmt.Sprintf("acme: authorization error for %s: %s", a.Identifier, strings.Join(e, "; "))
+	}
+
+	return fmt.Sprintf("acme: authorization error: %s", strings.Join(e, "; "))
+}
+
+// OrderError is returned from Client's order related methods.
+// It indicates the order is unusable and the clients should start over with
+// AuthorizeOrder.
+//
+// The clients can still fetch the order object from CA using GetOrder
+// to inspect its state.
+type OrderError struct {
+	OrderURL string
+	Status   string
+}
+
+func (oe *OrderError) Error() string {
+	return fmt.Sprintf("acme: order %s status: %s", oe.OrderURL, oe.Status)
+}
+
+// RateLimit reports whether err represents a rate limit error and
+// any Retry-After duration returned by the server.
+//
+// See the following for more details on rate limiting:
+// https://tools.ietf.org/html/draft-ietf-acme-acme-05#section-5.6
+func RateLimit(err error) (time.Duration, bool) {
+	e, ok := err.(*Error)
+	if !ok {
+		return 0, false
+	}
+	// Some CA implementations may return incorrect values.
+	// Use case-insensitive comparison.
+	if !strings.HasSuffix(strings.ToLower(e.ProblemType), ":ratelimited") {
+		return 0, false
+	}
+	if e.Header == nil {
+		return 0, true
+	}
+	return retryAfter(e.Header.Get("Retry-After")), true
+}
+
+// Account is a user account. It is associated with a private key.
+// Non-RFC 8555 fields are empty when interfacing with a compliant CA.
+type Account struct {
+	// URI is the account unique ID, which is also a URL used to retrieve
+	// account data from the CA.
+	// When interfacing with RFC 8555-compliant CAs, URI is the "kid" field
+	// value in JWS signed requests.
+	URI string
+
+	// Contact is a slice of contact info used during registration.
+	// See https://tools.ietf.org/html/rfc8555#section-7.3 for supported
+	// formats.
+	Contact []string
+
+	// Status indicates current account status as returned by the CA.
+	// Possible values are StatusValid, StatusDeactivated, and StatusRevoked.
+	Status string
+
+	// OrdersURL is a URL from which a list of orders submitted by this account
+	// can be fetched.
+	OrdersURL string
+
+	// The terms user has agreed to.
+	// A value not matching CurrentTerms indicates that the user hasn't agreed
+	// to the actual Terms of Service of the CA.
+	//
+	// It is non-RFC 8555 compliant. Package users can store the ToS they agree to
+	// during Client's Register call in the prompt callback function.
+	AgreedTerms string
+
+	// Actual terms of a CA.
+	//
+	// It is non-RFC 8555 compliant. Use Directory's Terms field.
+	// When a CA updates their terms and requires an account agreement,
+	// a URL at which instructions to do so is available in Error's Instance field.
+	CurrentTerms string
+
+	// Authz is the authorization URL used to initiate a new authz flow.
+	//
+	// It is non-RFC 8555 compliant. Use Directory's AuthzURL or OrderURL.
+	Authz string
+
+	// Authorizations is a URI from which a list of authorizations
+	// granted to this account can be fetched via a GET request.
+	//
+	// It is non-RFC 8555 compliant and is obsoleted by OrdersURL.
+	Authorizations string
+
+	// Certificates is a URI from which a list of certificates
+	// issued for this account can be fetched via a GET request.
+	//
+	// It is non-RFC 8555 compliant and is obsoleted by OrdersURL.
+	Certificates string
+
+	// ExternalAccountBinding represents an arbitrary binding to an account of
+	// the CA which the ACME server is tied to.
+	// See https://tools.ietf.org/html/rfc8555#section-7.3.4 for more details.
+	ExternalAccountBinding *ExternalAccountBinding
+}
+
+// ExternalAccountBinding contains the data needed to form a request with
+// an external account binding.
+// See https://tools.ietf.org/html/rfc8555#section-7.3.4 for more details.
+type ExternalAccountBinding struct {
+	// KID is the Key ID of the symmetric MAC key that the CA provides to
+	// identify an external account from ACME.
+	KID string
+
+	// Key is the bytes of the symmetric key that the CA provides to identify
+	// the account. Key must correspond to the KID.
+	Key []byte
+}
+
+func (e *ExternalAccountBinding) String() string {
+	return fmt.Sprintf("&{KID: %q, Key: redacted}", e.KID)
+}
+
+// Directory is ACME server discovery data.
+// See https://tools.ietf.org/html/rfc8555#section-7.1.1 for more details.
+type Directory struct {
+	// NonceURL indicates an endpoint where to fetch fresh nonce values from.
+	NonceURL string
+
+	// RegURL is an account endpoint URL, allowing for creating new accounts.
+	// Pre-RFC 8555 CAs also allow modifying existing accounts at this URL.
+	RegURL string
+
+	// OrderURL is used to initiate the certificate issuance flow
+	// as described in RFC 8555.
+	OrderURL string
+
+	// AuthzURL is used to initiate identifier pre-authorization flow.
+	// Empty string indicates the flow is unsupported by the CA.
+	AuthzURL string
+
+	// CertURL is a new certificate issuance endpoint URL.
+	// It is non-RFC 8555 compliant and is obsoleted by OrderURL.
+	CertURL string
+
+	// RevokeURL is used to initiate a certificate revocation flow.
+	RevokeURL string
+
+	// KeyChangeURL allows to perform account key rollover flow.
+	KeyChangeURL string
+
+	// Term is a URI identifying the current terms of service.
+	Terms string
+
+	// Website is an HTTP or HTTPS URL locating a website
+	// providing more information about the ACME server.
+	Website string
+
+	// CAA consists of lowercase hostname elements, which the ACME server
+	// recognises as referring to itself for the purposes of CAA record validation
+	// as defined in RFC6844.
+	CAA []string
+
+	// ExternalAccountRequired indicates that the CA requires for all account-related
+	// requests to include external account binding information.
+	ExternalAccountRequired bool
+}
+
+// rfcCompliant reports whether the ACME server implements RFC 8555.
+// Note that some servers may have incomplete RFC implementation
+// even if the returned value is true.
+// If rfcCompliant reports false, the server most likely implements draft-02.
+func (d *Directory) rfcCompliant() bool {
+	return d.OrderURL != ""
+}
+
+// Order represents a client's request for a certificate.
+// It tracks the request flow progress through to issuance.
+type Order struct {
+	// URI uniquely identifies an order.
+	URI string
+
+	// Status represents the current status of the order.
+	// It indicates which action the client should take.
+	//
+	// Possible values are StatusPending, StatusReady, StatusProcessing, StatusValid and StatusInvalid.
+	// Pending means the CA does not believe that the client has fulfilled the requirements.
+	// Ready indicates that the client has fulfilled all the requirements and can submit a CSR
+	// to obtain a certificate. This is done with Client's CreateOrderCert.
+	// Processing means the certificate is being issued.
+	// Valid indicates the CA has issued the certificate. It can be downloaded
+	// from the Order's CertURL. This is done with Client's FetchCert.
+	// Invalid means the certificate will not be issued. Users should consider this order
+	// abandoned.
+	Status string
+
+	// Expires is the timestamp after which CA considers this order invalid.
+	Expires time.Time
+
+	// Identifiers contains all identifier objects which the order pertains to.
+	Identifiers []AuthzID
+
+	// NotBefore is the requested value of the notBefore field in the certificate.
+	NotBefore time.Time
+
+	// NotAfter is the requested value of the notAfter field in the certificate.
+	NotAfter time.Time
+
+	// AuthzURLs represents authorizations to complete before a certificate
+	// for identifiers specified in the order can be issued.
+	// It also contains unexpired authorizations that the client has completed
+	// in the past.
+	//
+	// Authorization objects can be fetched using Client's GetAuthorization method.
+	//
+	// The required authorizations are dictated by CA policies.
+	// There may not be a 1:1 relationship between the identifiers and required authorizations.
+	// Required authorizations can be identified by their StatusPending status.
+	//
+	// For orders in the StatusValid or StatusInvalid state these are the authorizations
+	// which were completed.
+	AuthzURLs []string
+
+	// FinalizeURL is the endpoint at which a CSR is submitted to obtain a certificate
+	// once all the authorizations are satisfied.
+	FinalizeURL string
+
+	// CertURL points to the certificate that has been issued in response to this order.
+	CertURL string
+
+	// The error that occurred while processing the order as received from a CA, if any.
+	Error *Error
+}
+
+// OrderOption allows customizing Client.AuthorizeOrder call.
+type OrderOption interface {
+	privateOrderOpt()
+}
+
+// WithOrderNotBefore sets order's NotBefore field.
+func WithOrderNotBefore(t time.Time) OrderOption {
+	return orderNotBeforeOpt(t)
+}
+
+// WithOrderNotAfter sets order's NotAfter field.
+func WithOrderNotAfter(t time.Time) OrderOption {
+	return orderNotAfterOpt(t)
+}
+
+type orderNotBeforeOpt time.Time
+
+func (orderNotBeforeOpt) privateOrderOpt() {}
+
+type orderNotAfterOpt time.Time
+
+func (orderNotAfterOpt) privateOrderOpt() {}
+
+// Authorization encodes an authorization response.
+type Authorization struct {
+	// URI uniquely identifies a authorization.
+	URI string
+
+	// Status is the current status of an authorization.
+	// Possible values are StatusPending, StatusValid, StatusInvalid, StatusDeactivated,
+	// StatusExpired and StatusRevoked.
+	Status string
+
+	// Identifier is what the account is authorized to represent.
+	Identifier AuthzID
+
+	// The timestamp after which the CA considers the authorization invalid.
+	Expires time.Time
+
+	// Wildcard is true for authorizations of a wildcard domain name.
+	Wildcard bool
+
+	// Challenges that the client needs to fulfill in order to prove possession
+	// of the identifier (for pending authorizations).
+	// For valid authorizations, the challenge that was validated.
+	// For invalid authorizations, the challenge that was attempted and failed.
+	//
+	// RFC 8555 compatible CAs require users to fuflfill only one of the challenges.
+	Challenges []*Challenge
+
+	// A collection of sets of challenges, each of which would be sufficient
+	// to prove possession of the identifier.
+	// Clients must complete a set of challenges that covers at least one set.
+	// Challenges are identified by their indices in the challenges array.
+	// If this field is empty, the client needs to complete all challenges.
+	//
+	// This field is unused in RFC 8555.
+	Combinations [][]int
+}
+
+// AuthzID is an identifier that an account is authorized to represent.
+type AuthzID struct {
+	Type  string // The type of identifier, "dns" or "ip".
+	Value string // The identifier itself, e.g. "example.org".
+}
+
+// DomainIDs creates a slice of AuthzID with "dns" identifier type.
+func DomainIDs(names ...string) []AuthzID {
+	a := make([]AuthzID, len(names))
+	for i, v := range names {
+		a[i] = AuthzID{Type: "dns", Value: v}
+	}
+	return a
+}
+
+// IPIDs creates a slice of AuthzID with "ip" identifier type.
+// Each element of addr is textual form of an address as defined
+// in RFC1123 Section 2.1 for IPv4 and in RFC5952 Section 4 for IPv6.
+func IPIDs(addr ...string) []AuthzID {
+	a := make([]AuthzID, len(addr))
+	for i, v := range addr {
+		a[i] = AuthzID{Type: "ip", Value: v}
+	}
+	return a
+}
+
+// wireAuthzID is ACME JSON representation of authorization identifier objects.
+type wireAuthzID struct {
+	Type  string `json:"type"`
+	Value string `json:"value"`
+}
+
+// wireAuthz is ACME JSON representation of Authorization objects.
+type wireAuthz struct {
+	Identifier   wireAuthzID
+	Status       string
+	Expires      time.Time
+	Wildcard     bool
+	Challenges   []wireChallenge
+	Combinations [][]int
+	Error        *wireError
+}
+
+func (z *wireAuthz) authorization(uri string) *Authorization {
+	a := &Authorization{
+		URI:          uri,
+		Status:       z.Status,
+		Identifier:   AuthzID{Type: z.Identifier.Type, Value: z.Identifier.Value},
+		Expires:      z.Expires,
+		Wildcard:     z.Wildcard,
+		Challenges:   make([]*Challenge, len(z.Challenges)),
+		Combinations: z.Combinations, // shallow copy
+	}
+	for i, v := range z.Challenges {
+		a.Challenges[i] = v.challenge()
+	}
+	return a
+}
+
+func (z *wireAuthz) error(uri string) *AuthorizationError {
+	err := &AuthorizationError{
+		URI:        uri,
+		Identifier: z.Identifier.Value,
+	}
+
+	if z.Error != nil {
+		err.Errors = append(err.Errors, z.Error.error(nil))
+	}
+
+	for _, raw := range z.Challenges {
+		if raw.Error != nil {
+			err.Errors = append(err.Errors, raw.Error.error(nil))
+		}
+	}
+
+	return err
+}
+
+// Challenge encodes a returned CA challenge.
+// Its Error field may be non-nil if the challenge is part of an Authorization
+// with StatusInvalid.
+type Challenge struct {
+	// Type is the challenge type, e.g. "http-01", "tls-alpn-01", "dns-01".
+	Type string
+
+	// URI is where a challenge response can be posted to.
+	URI string
+
+	// Token is a random value that uniquely identifies the challenge.
+	Token string
+
+	// Status identifies the status of this challenge.
+	// In RFC 8555, possible values are StatusPending, StatusProcessing, StatusValid,
+	// and StatusInvalid.
+	Status string
+
+	// Validated is the time at which the CA validated this challenge.
+	// Always zero value in pre-RFC 8555.
+	Validated time.Time
+
+	// Error indicates the reason for an authorization failure
+	// when this challenge was used.
+	// The type of a non-nil value is *Error.
+	Error error
+}
+
+// wireChallenge is ACME JSON challenge representation.
+type wireChallenge struct {
+	URL       string `json:"url"` // RFC
+	URI       string `json:"uri"` // pre-RFC
+	Type      string
+	Token     string
+	Status    string
+	Validated time.Time
+	Error     *wireError
+}
+
+func (c *wireChallenge) challenge() *Challenge {
+	v := &Challenge{
+		URI:    c.URL,
+		Type:   c.Type,
+		Token:  c.Token,
+		Status: c.Status,
+	}
+	if v.URI == "" {
+		v.URI = c.URI // c.URL was empty; use legacy
+	}
+	if v.Status == "" {
+		v.Status = StatusPending
+	}
+	if c.Error != nil {
+		v.Error = c.Error.error(nil)
+	}
+	return v
+}
+
+// wireError is a subset of fields of the Problem Details object
+// as described in https://tools.ietf.org/html/rfc7807#section-3.1.
+type wireError struct {
+	Status      int
+	Type        string
+	Detail      string
+	Instance    string
+	Subproblems []Subproblem
+}
+
+func (e *wireError) error(h http.Header) *Error {
+	err := &Error{
+		StatusCode:  e.Status,
+		ProblemType: e.Type,
+		Detail:      e.Detail,
+		Instance:    e.Instance,
+		Header:      h,
+		Subproblems: e.Subproblems,
+	}
+	return err
+}
+
+// CertOption is an optional argument type for the TLS ChallengeCert methods for
+// customizing a temporary certificate for TLS-based challenges.
+type CertOption interface {
+	privateCertOpt()
+}
+
+// WithKey creates an option holding a private/public key pair.
+// The private part signs a certificate, and the public part represents the signee.
+func WithKey(key crypto.Signer) CertOption {
+	return &certOptKey{key}
+}
+
+type certOptKey struct {
+	key crypto.Signer
+}
+
+func (*certOptKey) privateCertOpt() {}
+
+// WithTemplate creates an option for specifying a certificate template.
+// See x509.CreateCertificate for template usage details.
+//
+// In TLS ChallengeCert methods, the template is also used as parent,
+// resulting in a self-signed certificate.
+// The DNSNames field of t is always overwritten for tls-sni challenge certs.
+func WithTemplate(t *x509.Certificate) CertOption {
+	return (*certOptTemplate)(t)
+}
+
+type certOptTemplate x509.Certificate
+
+func (*certOptTemplate) privateCertOpt() {}

vendor/golang.org/x/crypto/acme/version_go112.go

@@ -0,0 +1,28 @@
+// Copyright 2019 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build go1.12
+// +build go1.12
+
+package acme
+
+import "runtime/debug"
+
+func init() {
+	// Set packageVersion if the binary was built in modules mode and x/crypto
+	// was not replaced with a different module.
+	info, ok := debug.ReadBuildInfo()
+	if !ok {
+		return
+	}
+	for _, m := range info.Deps {
+		if m.Path != "golang.org/x/crypto" {
+			continue
+		}
+		if m.Replace == nil {
+			packageVersion = m.Version
+		}
+		break
+	}
+}

vendor/golang.org/x/crypto/bcrypt/base64.go

@@ -0,0 +1,35 @@
+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package bcrypt
+
+import "encoding/base64"
+
+const alphabet = "./ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789"
+
+var bcEncoding = base64.NewEncoding(alphabet)
+
+func base64Encode(src []byte) []byte {
+	n := bcEncoding.EncodedLen(len(src))
+	dst := make([]byte, n)
+	bcEncoding.Encode(dst, src)
+	for dst[n-1] == '=' {
+		n--
+	}
+	return dst[:n]
+}
+
+func base64Decode(src []byte) ([]byte, error) {
+	numOfEquals := 4 - (len(src) % 4)
+	for i := 0; i < numOfEquals; i++ {
+		src = append(src, '=')
+	}
+
+	dst := make([]byte, bcEncoding.DecodedLen(len(src)))
+	n, err := bcEncoding.Decode(dst, src)
+	if err != nil {
+		return nil, err
+	}
+	return dst[:n], nil
+}

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

@@ -0,0 +1,295 @@
+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Package bcrypt implements Provos and Mazières's bcrypt adaptive hashing
+// algorithm. See http://www.usenix.org/event/usenix99/provos/provos.pdf
+package bcrypt // import "golang.org/x/crypto/bcrypt"
+
+// The code is a port of Provos and Mazières's C implementation.
+import (
+	"crypto/rand"
+	"crypto/subtle"
+	"errors"
+	"fmt"
+	"io"
+	"strconv"
+
+	"golang.org/x/crypto/blowfish"
+)
+
+const (
+	MinCost     int = 4  // the minimum allowable cost as passed in to GenerateFromPassword
+	MaxCost     int = 31 // the maximum allowable cost as passed in to GenerateFromPassword
+	DefaultCost int = 10 // the cost that will actually be set if a cost below MinCost is passed into GenerateFromPassword
+)
+
+// The error returned from CompareHashAndPassword when a password and hash do
+// not match.
+var ErrMismatchedHashAndPassword = errors.New("crypto/bcrypt: hashedPassword is not the hash of the given password")
+
+// The error returned from CompareHashAndPassword when a hash is too short to
+// be a bcrypt hash.
+var ErrHashTooShort = errors.New("crypto/bcrypt: hashedSecret too short to be a bcrypted password")
+
+// The error returned from CompareHashAndPassword when a hash was created with
+// a bcrypt algorithm newer than this implementation.
+type HashVersionTooNewError byte
+
+func (hv HashVersionTooNewError) Error() string {
+	return fmt.Sprintf("crypto/bcrypt: bcrypt algorithm version '%c' requested is newer than current version '%c'", byte(hv), majorVersion)
+}
+
+// The error returned from CompareHashAndPassword when a hash starts with something other than '$'
+type InvalidHashPrefixError byte
+
+func (ih InvalidHashPrefixError) Error() string {
+	return fmt.Sprintf("crypto/bcrypt: bcrypt hashes must start with '$', but hashedSecret started with '%c'", byte(ih))
+}
+
+type InvalidCostError int
+
+func (ic InvalidCostError) Error() string {
+	return fmt.Sprintf("crypto/bcrypt: cost %d is outside allowed range (%d,%d)", int(ic), int(MinCost), int(MaxCost))
+}
+
+const (
+	majorVersion       = '2'
+	minorVersion       = 'a'
+	maxSaltSize        = 16
+	maxCryptedHashSize = 23
+	encodedSaltSize    = 22
+	encodedHashSize    = 31
+	minHashSize        = 59
+)
+
+// magicCipherData is an IV for the 64 Blowfish encryption calls in
+// bcrypt(). It's the string "OrpheanBeholderScryDoubt" in big-endian bytes.
+var magicCipherData = []byte{
+	0x4f, 0x72, 0x70, 0x68,
+	0x65, 0x61, 0x6e, 0x42,
+	0x65, 0x68, 0x6f, 0x6c,
+	0x64, 0x65, 0x72, 0x53,
+	0x63, 0x72, 0x79, 0x44,
+	0x6f, 0x75, 0x62, 0x74,
+}
+
+type hashed struct {
+	hash  []byte
+	salt  []byte
+	cost  int // allowed range is MinCost to MaxCost
+	major byte
+	minor byte
+}
+
+// GenerateFromPassword returns the bcrypt hash of the password at the given
+// cost. If the cost given is less than MinCost, the cost will be set to
+// DefaultCost, instead. Use CompareHashAndPassword, as defined in this package,
+// to compare the returned hashed password with its cleartext version.
+func GenerateFromPassword(password []byte, cost int) ([]byte, error) {
+	p, err := newFromPassword(password, cost)
+	if err != nil {
+		return nil, err
+	}
+	return p.Hash(), nil
+}
+
+// CompareHashAndPassword compares a bcrypt hashed password with its possible
+// plaintext equivalent. Returns nil on success, or an error on failure.
+func CompareHashAndPassword(hashedPassword, password []byte) error {
+	p, err := newFromHash(hashedPassword)
+	if err != nil {
+		return err
+	}
+
+	otherHash, err := bcrypt(password, p.cost, p.salt)
+	if err != nil {
+		return err
+	}
+
+	otherP := &hashed{otherHash, p.salt, p.cost, p.major, p.minor}
+	if subtle.ConstantTimeCompare(p.Hash(), otherP.Hash()) == 1 {
+		return nil
+	}
+
+	return ErrMismatchedHashAndPassword
+}
+
+// Cost returns the hashing cost used to create the given hashed
+// password. When, in the future, the hashing cost of a password system needs
+// to be increased in order to adjust for greater computational power, this
+// function allows one to establish which passwords need to be updated.
+func Cost(hashedPassword []byte) (int, error) {
+	p, err := newFromHash(hashedPassword)
+	if err != nil {
+		return 0, err
+	}
+	return p.cost, nil
+}
+
+func newFromPassword(password []byte, cost int) (*hashed, error) {
+	if cost < MinCost {
+		cost = DefaultCost
+	}
+	p := new(hashed)
+	p.major = majorVersion
+	p.minor = minorVersion
+
+	err := checkCost(cost)
+	if err != nil {
+		return nil, err
+	}
+	p.cost = cost
+
+	unencodedSalt := make([]byte, maxSaltSize)
+	_, err = io.ReadFull(rand.Reader, unencodedSalt)
+	if err != nil {
+		return nil, err
+	}
+
+	p.salt = base64Encode(unencodedSalt)
+	hash, err := bcrypt(password, p.cost, p.salt)
+	if err != nil {
+		return nil, err
+	}
+	p.hash = hash
+	return p, err
+}
+
+func newFromHash(hashedSecret []byte) (*hashed, error) {
+	if len(hashedSecret) < minHashSize {
+		return nil, ErrHashTooShort
+	}
+	p := new(hashed)
+	n, err := p.decodeVersion(hashedSecret)
+	if err != nil {
+		return nil, err
+	}
+	hashedSecret = hashedSecret[n:]
+	n, err = p.decodeCost(hashedSecret)
+	if err != nil {
+		return nil, err
+	}
+	hashedSecret = hashedSecret[n:]
+
+	// The "+2" is here because we'll have to append at most 2 '=' to the salt
+	// when base64 decoding it in expensiveBlowfishSetup().
+	p.salt = make([]byte, encodedSaltSize, encodedSaltSize+2)
+	copy(p.salt, hashedSecret[:encodedSaltSize])
+
+	hashedSecret = hashedSecret[encodedSaltSize:]
+	p.hash = make([]byte, len(hashedSecret))
+	copy(p.hash, hashedSecret)
+
+	return p, nil
+}
+
+func bcrypt(password []byte, cost int, salt []byte) ([]byte, error) {
+	cipherData := make([]byte, len(magicCipherData))
+	copy(cipherData, magicCipherData)
+
+	c, err := expensiveBlowfishSetup(password, uint32(cost), salt)
+	if err != nil {
+		return nil, err
+	}
+
+	for i := 0; i < 24; i += 8 {
+		for j := 0; j < 64; j++ {
+			c.Encrypt(cipherData[i:i+8], cipherData[i:i+8])
+		}
+	}
+
+	// Bug compatibility with C bcrypt implementations. We only encode 23 of
+	// the 24 bytes encrypted.
+	hsh := base64Encode(cipherData[:maxCryptedHashSize])
+	return hsh, nil
+}
+
+func expensiveBlowfishSetup(key []byte, cost uint32, salt []byte) (*blowfish.Cipher, error) {
+	csalt, err := base64Decode(salt)
+	if err != nil {
+		return nil, err
+	}
+
+	// Bug compatibility with C bcrypt implementations. They use the trailing
+	// NULL in the key string during expansion.
+	// We copy the key to prevent changing the underlying array.
+	ckey := append(key[:len(key):len(key)], 0)
+
+	c, err := blowfish.NewSaltedCipher(ckey, csalt)
+	if err != nil {
+		return nil, err
+	}
+
+	var i, rounds uint64
+	rounds = 1 << cost
+	for i = 0; i < rounds; i++ {
+		blowfish.ExpandKey(ckey, c)
+		blowfish.ExpandKey(csalt, c)
+	}
+
+	return c, nil
+}
+
+func (p *hashed) Hash() []byte {
+	arr := make([]byte, 60)
+	arr[0] = '$'
+	arr[1] = p.major
+	n := 2
+	if p.minor != 0 {
+		arr[2] = p.minor
+		n = 3
+	}
+	arr[n] = '$'
+	n++
+	copy(arr[n:], []byte(fmt.Sprintf("%02d", p.cost)))
+	n += 2
+	arr[n] = '$'
+	n++
+	copy(arr[n:], p.salt)
+	n += encodedSaltSize
+	copy(arr[n:], p.hash)
+	n += encodedHashSize
+	return arr[:n]
+}
+
+func (p *hashed) decodeVersion(sbytes []byte) (int, error) {
+	if sbytes[0] != '$' {
+		return -1, InvalidHashPrefixError(sbytes[0])
+	}
+	if sbytes[1] > majorVersion {
+		return -1, HashVersionTooNewError(sbytes[1])
+	}
+	p.major = sbytes[1]
+	n := 3
+	if sbytes[2] != '$' {
+		p.minor = sbytes[2]
+		n++
+	}
+	return n, nil
+}
+
+// sbytes should begin where decodeVersion left off.
+func (p *hashed) decodeCost(sbytes []byte) (int, error) {
+	cost, err := strconv.Atoi(string(sbytes[0:2]))
+	if err != nil {
+		return -1, err
+	}
+	err = checkCost(cost)
+	if err != nil {
+		return -1, err
+	}
+	p.cost = cost
+	return 3, nil
+}
+
+func (p *hashed) String() string {
+	return fmt.Sprintf("&{hash: %#v, salt: %#v, cost: %d, major: %c, minor: %c}", string(p.hash), p.salt, p.cost, p.major, p.minor)
+}
+
+func checkCost(cost int) error {
+	if cost < MinCost || cost > MaxCost {
+		return InvalidCostError(cost)
+	}
+	return nil
+}

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

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

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

@@ -0,0 +1,38 @@
+// 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.
+
+//go:build go1.7 && amd64 && gc && !purego
+// +build go1.7,amd64,gc,!purego
+
+package blake2b
+
+import "golang.org/x/sys/cpu"
+
+func init() {
+	useAVX2 = cpu.X86.HasAVX2
+	useAVX = cpu.X86.HasAVX
+	useSSE4 = cpu.X86.HasSSE41
+}
+
+//go:noescape
+func hashBlocksAVX2(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+
+//go:noescape
+func hashBlocksAVX(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+
+//go:noescape
+func hashBlocksSSE4(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+
+func hashBlocks(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte) {
+	switch {
+	case useAVX2:
+		hashBlocksAVX2(h, c, flag, blocks)
+	case useAVX:
+		hashBlocksAVX(h, c, flag, blocks)
+	case useSSE4:
+		hashBlocksSSE4(h, c, flag, blocks)
+	default:
+		hashBlocksGeneric(h, c, flag, blocks)
+	}
+}

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

@@ -0,0 +1,745 @@
+// 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.
+
+//go:build go1.7 && amd64 && gc && !purego
+// +build go1.7,amd64,gc,!purego
+
+#include "textflag.h"
+
+DATA ·AVX2_iv0<>+0x00(SB)/8, $0x6a09e667f3bcc908
+DATA ·AVX2_iv0<>+0x08(SB)/8, $0xbb67ae8584caa73b
+DATA ·AVX2_iv0<>+0x10(SB)/8, $0x3c6ef372fe94f82b
+DATA ·AVX2_iv0<>+0x18(SB)/8, $0xa54ff53a5f1d36f1
+GLOBL ·AVX2_iv0<>(SB), (NOPTR+RODATA), $32
+
+DATA ·AVX2_iv1<>+0x00(SB)/8, $0x510e527fade682d1
+DATA ·AVX2_iv1<>+0x08(SB)/8, $0x9b05688c2b3e6c1f
+DATA ·AVX2_iv1<>+0x10(SB)/8, $0x1f83d9abfb41bd6b
+DATA ·AVX2_iv1<>+0x18(SB)/8, $0x5be0cd19137e2179
+GLOBL ·AVX2_iv1<>(SB), (NOPTR+RODATA), $32
+
+DATA ·AVX2_c40<>+0x00(SB)/8, $0x0201000706050403
+DATA ·AVX2_c40<>+0x08(SB)/8, $0x0a09080f0e0d0c0b
+DATA ·AVX2_c40<>+0x10(SB)/8, $0x0201000706050403
+DATA ·AVX2_c40<>+0x18(SB)/8, $0x0a09080f0e0d0c0b
+GLOBL ·AVX2_c40<>(SB), (NOPTR+RODATA), $32
+
+DATA ·AVX2_c48<>+0x00(SB)/8, $0x0100070605040302
+DATA ·AVX2_c48<>+0x08(SB)/8, $0x09080f0e0d0c0b0a
+DATA ·AVX2_c48<>+0x10(SB)/8, $0x0100070605040302
+DATA ·AVX2_c48<>+0x18(SB)/8, $0x09080f0e0d0c0b0a
+GLOBL ·AVX2_c48<>(SB), (NOPTR+RODATA), $32
+
+DATA ·AVX_iv0<>+0x00(SB)/8, $0x6a09e667f3bcc908
+DATA ·AVX_iv0<>+0x08(SB)/8, $0xbb67ae8584caa73b
+GLOBL ·AVX_iv0<>(SB), (NOPTR+RODATA), $16
+
+DATA ·AVX_iv1<>+0x00(SB)/8, $0x3c6ef372fe94f82b
+DATA ·AVX_iv1<>+0x08(SB)/8, $0xa54ff53a5f1d36f1
+GLOBL ·AVX_iv1<>(SB), (NOPTR+RODATA), $16
+
+DATA ·AVX_iv2<>+0x00(SB)/8, $0x510e527fade682d1
+DATA ·AVX_iv2<>+0x08(SB)/8, $0x9b05688c2b3e6c1f
+GLOBL ·AVX_iv2<>(SB), (NOPTR+RODATA), $16
+
+DATA ·AVX_iv3<>+0x00(SB)/8, $0x1f83d9abfb41bd6b
+DATA ·AVX_iv3<>+0x08(SB)/8, $0x5be0cd19137e2179
+GLOBL ·AVX_iv3<>(SB), (NOPTR+RODATA), $16
+
+DATA ·AVX_c40<>+0x00(SB)/8, $0x0201000706050403
+DATA ·AVX_c40<>+0x08(SB)/8, $0x0a09080f0e0d0c0b
+GLOBL ·AVX_c40<>(SB), (NOPTR+RODATA), $16
+
+DATA ·AVX_c48<>+0x00(SB)/8, $0x0100070605040302
+DATA ·AVX_c48<>+0x08(SB)/8, $0x09080f0e0d0c0b0a
+GLOBL ·AVX_c48<>(SB), (NOPTR+RODATA), $16
+
+#define VPERMQ_0x39_Y1_Y1 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xc9; BYTE $0x39
+#define VPERMQ_0x93_Y1_Y1 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xc9; BYTE $0x93
+#define VPERMQ_0x4E_Y2_Y2 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xd2; BYTE $0x4e
+#define VPERMQ_0x93_Y3_Y3 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xdb; BYTE $0x93
+#define VPERMQ_0x39_Y3_Y3 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xdb; BYTE $0x39
+
+#define ROUND_AVX2(m0, m1, m2, m3, t, c40, c48) \
+	VPADDQ  m0, Y0, Y0;   \
+	VPADDQ  Y1, Y0, Y0;   \
+	VPXOR   Y0, Y3, Y3;   \
+	VPSHUFD $-79, Y3, Y3; \
+	VPADDQ  Y3, Y2, Y2;   \
+	VPXOR   Y2, Y1, Y1;   \
+	VPSHUFB c40, Y1, Y1;  \
+	VPADDQ  m1, Y0, Y0;   \
+	VPADDQ  Y1, Y0, Y0;   \
+	VPXOR   Y0, Y3, Y3;   \
+	VPSHUFB c48, Y3, Y3;  \
+	VPADDQ  Y3, Y2, Y2;   \
+	VPXOR   Y2, Y1, Y1;   \
+	VPADDQ  Y1, Y1, t;    \
+	VPSRLQ  $63, Y1, Y1;  \
+	VPXOR   t, Y1, Y1;    \
+	VPERMQ_0x39_Y1_Y1;    \
+	VPERMQ_0x4E_Y2_Y2;    \
+	VPERMQ_0x93_Y3_Y3;    \
+	VPADDQ  m2, Y0, Y0;   \
+	VPADDQ  Y1, Y0, Y0;   \
+	VPXOR   Y0, Y3, Y3;   \
+	VPSHUFD $-79, Y3, Y3; \
+	VPADDQ  Y3, Y2, Y2;   \
+	VPXOR   Y2, Y1, Y1;   \
+	VPSHUFB c40, Y1, Y1;  \
+	VPADDQ  m3, Y0, Y0;   \
+	VPADDQ  Y1, Y0, Y0;   \
+	VPXOR   Y0, Y3, Y3;   \
+	VPSHUFB c48, Y3, Y3;  \
+	VPADDQ  Y3, Y2, Y2;   \
+	VPXOR   Y2, Y1, Y1;   \
+	VPADDQ  Y1, Y1, t;    \
+	VPSRLQ  $63, Y1, Y1;  \
+	VPXOR   t, Y1, Y1;    \
+	VPERMQ_0x39_Y3_Y3;    \
+	VPERMQ_0x4E_Y2_Y2;    \
+	VPERMQ_0x93_Y1_Y1
+
+#define VMOVQ_SI_X11_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x1E
+#define VMOVQ_SI_X12_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x26
+#define VMOVQ_SI_X13_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x2E
+#define VMOVQ_SI_X14_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x36
+#define VMOVQ_SI_X15_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x3E
+
+#define VMOVQ_SI_X11(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x5E; BYTE $n
+#define VMOVQ_SI_X12(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x66; BYTE $n
+#define VMOVQ_SI_X13(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x6E; BYTE $n
+#define VMOVQ_SI_X14(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x76; BYTE $n
+#define VMOVQ_SI_X15(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x7E; BYTE $n
+
+#define VPINSRQ_1_SI_X11_0 BYTE $0xC4; BYTE $0x63; BYTE $0xA1; BYTE $0x22; BYTE $0x1E; BYTE $0x01
+#define VPINSRQ_1_SI_X12_0 BYTE $0xC4; BYTE $0x63; BYTE $0x99; BYTE $0x22; BYTE $0x26; BYTE $0x01
+#define VPINSRQ_1_SI_X13_0 BYTE $0xC4; BYTE $0x63; BYTE $0x91; BYTE $0x22; BYTE $0x2E; BYTE $0x01
+#define VPINSRQ_1_SI_X14_0 BYTE $0xC4; BYTE $0x63; BYTE $0x89; BYTE $0x22; BYTE $0x36; BYTE $0x01
+#define VPINSRQ_1_SI_X15_0 BYTE $0xC4; BYTE $0x63; BYTE $0x81; BYTE $0x22; BYTE $0x3E; BYTE $0x01
+
+#define VPINSRQ_1_SI_X11(n) BYTE $0xC4; BYTE $0x63; BYTE $0xA1; BYTE $0x22; BYTE $0x5E; BYTE $n; BYTE $0x01
+#define VPINSRQ_1_SI_X12(n) BYTE $0xC4; BYTE $0x63; BYTE $0x99; BYTE $0x22; BYTE $0x66; BYTE $n; BYTE $0x01
+#define VPINSRQ_1_SI_X13(n) BYTE $0xC4; BYTE $0x63; BYTE $0x91; BYTE $0x22; BYTE $0x6E; BYTE $n; BYTE $0x01
+#define VPINSRQ_1_SI_X14(n) BYTE $0xC4; BYTE $0x63; BYTE $0x89; BYTE $0x22; BYTE $0x76; BYTE $n; BYTE $0x01
+#define VPINSRQ_1_SI_X15(n) BYTE $0xC4; BYTE $0x63; BYTE $0x81; BYTE $0x22; BYTE $0x7E; BYTE $n; BYTE $0x01
+
+#define VMOVQ_R8_X15 BYTE $0xC4; BYTE $0x41; BYTE $0xF9; BYTE $0x6E; BYTE $0xF8
+#define VPINSRQ_1_R9_X15 BYTE $0xC4; BYTE $0x43; BYTE $0x81; BYTE $0x22; BYTE $0xF9; BYTE $0x01
+
+// load msg: Y12 = (i0, i1, i2, i3)
+// i0, i1, i2, i3 must not be 0
+#define LOAD_MSG_AVX2_Y12(i0, i1, i2, i3) \
+	VMOVQ_SI_X12(i0*8);           \
+	VMOVQ_SI_X11(i2*8);           \
+	VPINSRQ_1_SI_X12(i1*8);       \
+	VPINSRQ_1_SI_X11(i3*8);       \
+	VINSERTI128 $1, X11, Y12, Y12
+
+// load msg: Y13 = (i0, i1, i2, i3)
+// i0, i1, i2, i3 must not be 0
+#define LOAD_MSG_AVX2_Y13(i0, i1, i2, i3) \
+	VMOVQ_SI_X13(i0*8);           \
+	VMOVQ_SI_X11(i2*8);           \
+	VPINSRQ_1_SI_X13(i1*8);       \
+	VPINSRQ_1_SI_X11(i3*8);       \
+	VINSERTI128 $1, X11, Y13, Y13
+
+// load msg: Y14 = (i0, i1, i2, i3)
+// i0, i1, i2, i3 must not be 0
+#define LOAD_MSG_AVX2_Y14(i0, i1, i2, i3) \
+	VMOVQ_SI_X14(i0*8);           \
+	VMOVQ_SI_X11(i2*8);           \
+	VPINSRQ_1_SI_X14(i1*8);       \
+	VPINSRQ_1_SI_X11(i3*8);       \
+	VINSERTI128 $1, X11, Y14, Y14
+
+// load msg: Y15 = (i0, i1, i2, i3)
+// i0, i1, i2, i3 must not be 0
+#define LOAD_MSG_AVX2_Y15(i0, i1, i2, i3) \
+	VMOVQ_SI_X15(i0*8);           \
+	VMOVQ_SI_X11(i2*8);           \
+	VPINSRQ_1_SI_X15(i1*8);       \
+	VPINSRQ_1_SI_X11(i3*8);       \
+	VINSERTI128 $1, X11, Y15, Y15
+
+#define LOAD_MSG_AVX2_0_2_4_6_1_3_5_7_8_10_12_14_9_11_13_15() \
+	VMOVQ_SI_X12_0;                   \
+	VMOVQ_SI_X11(4*8);                \
+	VPINSRQ_1_SI_X12(2*8);            \
+	VPINSRQ_1_SI_X11(6*8);            \
+	VINSERTI128 $1, X11, Y12, Y12;    \
+	LOAD_MSG_AVX2_Y13(1, 3, 5, 7);    \
+	LOAD_MSG_AVX2_Y14(8, 10, 12, 14); \
+	LOAD_MSG_AVX2_Y15(9, 11, 13, 15)
+
+#define LOAD_MSG_AVX2_14_4_9_13_10_8_15_6_1_0_11_5_12_2_7_3() \
+	LOAD_MSG_AVX2_Y12(14, 4, 9, 13); \
+	LOAD_MSG_AVX2_Y13(10, 8, 15, 6); \
+	VMOVQ_SI_X11(11*8);              \
+	VPSHUFD     $0x4E, 0*8(SI), X14; \
+	VPINSRQ_1_SI_X11(5*8);           \
+	VINSERTI128 $1, X11, Y14, Y14;   \
+	LOAD_MSG_AVX2_Y15(12, 2, 7, 3)
+
+#define LOAD_MSG_AVX2_11_12_5_15_8_0_2_13_10_3_7_9_14_6_1_4() \
+	VMOVQ_SI_X11(5*8);              \
+	VMOVDQU     11*8(SI), X12;      \
+	VPINSRQ_1_SI_X11(15*8);         \
+	VINSERTI128 $1, X11, Y12, Y12;  \
+	VMOVQ_SI_X13(8*8);              \
+	VMOVQ_SI_X11(2*8);              \
+	VPINSRQ_1_SI_X13_0;             \
+	VPINSRQ_1_SI_X11(13*8);         \
+	VINSERTI128 $1, X11, Y13, Y13;  \
+	LOAD_MSG_AVX2_Y14(10, 3, 7, 9); \
+	LOAD_MSG_AVX2_Y15(14, 6, 1, 4)
+
+#define LOAD_MSG_AVX2_7_3_13_11_9_1_12_14_2_5_4_15_6_10_0_8() \
+	LOAD_MSG_AVX2_Y12(7, 3, 13, 11); \
+	LOAD_MSG_AVX2_Y13(9, 1, 12, 14); \
+	LOAD_MSG_AVX2_Y14(2, 5, 4, 15);  \
+	VMOVQ_SI_X15(6*8);               \
+	VMOVQ_SI_X11_0;                  \
+	VPINSRQ_1_SI_X15(10*8);          \
+	VPINSRQ_1_SI_X11(8*8);           \
+	VINSERTI128 $1, X11, Y15, Y15
+
+#define LOAD_MSG_AVX2_9_5_2_10_0_7_4_15_14_11_6_3_1_12_8_13() \
+	LOAD_MSG_AVX2_Y12(9, 5, 2, 10);  \
+	VMOVQ_SI_X13_0;                  \
+	VMOVQ_SI_X11(4*8);               \
+	VPINSRQ_1_SI_X13(7*8);           \
+	VPINSRQ_1_SI_X11(15*8);          \
+	VINSERTI128 $1, X11, Y13, Y13;   \
+	LOAD_MSG_AVX2_Y14(14, 11, 6, 3); \
+	LOAD_MSG_AVX2_Y15(1, 12, 8, 13)
+
+#define LOAD_MSG_AVX2_2_6_0_8_12_10_11_3_4_7_15_1_13_5_14_9() \
+	VMOVQ_SI_X12(2*8);                \
+	VMOVQ_SI_X11_0;                   \
+	VPINSRQ_1_SI_X12(6*8);            \
+	VPINSRQ_1_SI_X11(8*8);            \
+	VINSERTI128 $1, X11, Y12, Y12;    \
+	LOAD_MSG_AVX2_Y13(12, 10, 11, 3); \
+	LOAD_MSG_AVX2_Y14(4, 7, 15, 1);   \
+	LOAD_MSG_AVX2_Y15(13, 5, 14, 9)
+
+#define LOAD_MSG_AVX2_12_1_14_4_5_15_13_10_0_6_9_8_7_3_2_11() \
+	LOAD_MSG_AVX2_Y12(12, 1, 14, 4);  \
+	LOAD_MSG_AVX2_Y13(5, 15, 13, 10); \
+	VMOVQ_SI_X14_0;                   \
+	VPSHUFD     $0x4E, 8*8(SI), X11;  \
+	VPINSRQ_1_SI_X14(6*8);            \
+	VINSERTI128 $1, X11, Y14, Y14;    \
+	LOAD_MSG_AVX2_Y15(7, 3, 2, 11)
+
+#define LOAD_MSG_AVX2_13_7_12_3_11_14_1_9_5_15_8_2_0_4_6_10() \
+	LOAD_MSG_AVX2_Y12(13, 7, 12, 3); \
+	LOAD_MSG_AVX2_Y13(11, 14, 1, 9); \
+	LOAD_MSG_AVX2_Y14(5, 15, 8, 2);  \
+	VMOVQ_SI_X15_0;                  \
+	VMOVQ_SI_X11(6*8);               \
+	VPINSRQ_1_SI_X15(4*8);           \
+	VPINSRQ_1_SI_X11(10*8);          \
+	VINSERTI128 $1, X11, Y15, Y15
+
+#define LOAD_MSG_AVX2_6_14_11_0_15_9_3_8_12_13_1_10_2_7_4_5() \
+	VMOVQ_SI_X12(6*8);              \
+	VMOVQ_SI_X11(11*8);             \
+	VPINSRQ_1_SI_X12(14*8);         \
+	VPINSRQ_1_SI_X11_0;             \
+	VINSERTI128 $1, X11, Y12, Y12;  \
+	LOAD_MSG_AVX2_Y13(15, 9, 3, 8); \
+	VMOVQ_SI_X11(1*8);              \
+	VMOVDQU     12*8(SI), X14;      \
+	VPINSRQ_1_SI_X11(10*8);         \
+	VINSERTI128 $1, X11, Y14, Y14;  \
+	VMOVQ_SI_X15(2*8);              \
+	VMOVDQU     4*8(SI), X11;       \
+	VPINSRQ_1_SI_X15(7*8);          \
+	VINSERTI128 $1, X11, Y15, Y15
+
+#define LOAD_MSG_AVX2_10_8_7_1_2_4_6_5_15_9_3_13_11_14_12_0() \
+	LOAD_MSG_AVX2_Y12(10, 8, 7, 1);  \
+	VMOVQ_SI_X13(2*8);               \
+	VPSHUFD     $0x4E, 5*8(SI), X11; \
+	VPINSRQ_1_SI_X13(4*8);           \
+	VINSERTI128 $1, X11, Y13, Y13;   \
+	LOAD_MSG_AVX2_Y14(15, 9, 3, 13); \
+	VMOVQ_SI_X15(11*8);              \
+	VMOVQ_SI_X11(12*8);              \
+	VPINSRQ_1_SI_X15(14*8);          \
+	VPINSRQ_1_SI_X11_0;              \
+	VINSERTI128 $1, X11, Y15, Y15
+
+// func hashBlocksAVX2(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+TEXT ·hashBlocksAVX2(SB), 4, $320-48 // frame size = 288 + 32 byte alignment
+	MOVQ h+0(FP), AX
+	MOVQ c+8(FP), BX
+	MOVQ flag+16(FP), CX
+	MOVQ blocks_base+24(FP), SI
+	MOVQ blocks_len+32(FP), DI
+
+	MOVQ SP, DX
+	ADDQ $31, DX
+	ANDQ $~31, DX
+
+	MOVQ CX, 16(DX)
+	XORQ CX, CX
+	MOVQ CX, 24(DX)
+
+	VMOVDQU ·AVX2_c40<>(SB), Y4
+	VMOVDQU ·AVX2_c48<>(SB), Y5
+
+	VMOVDQU 0(AX), Y8
+	VMOVDQU 32(AX), Y9
+	VMOVDQU ·AVX2_iv0<>(SB), Y6
+	VMOVDQU ·AVX2_iv1<>(SB), Y7
+
+	MOVQ 0(BX), R8
+	MOVQ 8(BX), R9
+	MOVQ R9, 8(DX)
+
+loop:
+	ADDQ $128, R8
+	MOVQ R8, 0(DX)
+	CMPQ R8, $128
+	JGE  noinc
+	INCQ R9
+	MOVQ R9, 8(DX)
+
+noinc:
+	VMOVDQA Y8, Y0
+	VMOVDQA Y9, Y1
+	VMOVDQA Y6, Y2
+	VPXOR   0(DX), Y7, Y3
+
+	LOAD_MSG_AVX2_0_2_4_6_1_3_5_7_8_10_12_14_9_11_13_15()
+	VMOVDQA Y12, 32(DX)
+	VMOVDQA Y13, 64(DX)
+	VMOVDQA Y14, 96(DX)
+	VMOVDQA Y15, 128(DX)
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_14_4_9_13_10_8_15_6_1_0_11_5_12_2_7_3()
+	VMOVDQA Y12, 160(DX)
+	VMOVDQA Y13, 192(DX)
+	VMOVDQA Y14, 224(DX)
+	VMOVDQA Y15, 256(DX)
+
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_11_12_5_15_8_0_2_13_10_3_7_9_14_6_1_4()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_7_3_13_11_9_1_12_14_2_5_4_15_6_10_0_8()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_9_5_2_10_0_7_4_15_14_11_6_3_1_12_8_13()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_2_6_0_8_12_10_11_3_4_7_15_1_13_5_14_9()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_12_1_14_4_5_15_13_10_0_6_9_8_7_3_2_11()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_13_7_12_3_11_14_1_9_5_15_8_2_0_4_6_10()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_6_14_11_0_15_9_3_8_12_13_1_10_2_7_4_5()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_10_8_7_1_2_4_6_5_15_9_3_13_11_14_12_0()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+
+	ROUND_AVX2(32(DX), 64(DX), 96(DX), 128(DX), Y10, Y4, Y5)
+	ROUND_AVX2(160(DX), 192(DX), 224(DX), 256(DX), Y10, Y4, Y5)
+
+	VPXOR Y0, Y8, Y8
+	VPXOR Y1, Y9, Y9
+	VPXOR Y2, Y8, Y8
+	VPXOR Y3, Y9, Y9
+
+	LEAQ 128(SI), SI
+	SUBQ $128, DI
+	JNE  loop
+
+	MOVQ R8, 0(BX)
+	MOVQ R9, 8(BX)
+
+	VMOVDQU Y8, 0(AX)
+	VMOVDQU Y9, 32(AX)
+	VZEROUPPER
+
+	RET
+
+#define VPUNPCKLQDQ_X2_X2_X15 BYTE $0xC5; BYTE $0x69; BYTE $0x6C; BYTE $0xFA
+#define VPUNPCKLQDQ_X3_X3_X15 BYTE $0xC5; BYTE $0x61; BYTE $0x6C; BYTE $0xFB
+#define VPUNPCKLQDQ_X7_X7_X15 BYTE $0xC5; BYTE $0x41; BYTE $0x6C; BYTE $0xFF
+#define VPUNPCKLQDQ_X13_X13_X15 BYTE $0xC4; BYTE $0x41; BYTE $0x11; BYTE $0x6C; BYTE $0xFD
+#define VPUNPCKLQDQ_X14_X14_X15 BYTE $0xC4; BYTE $0x41; BYTE $0x09; BYTE $0x6C; BYTE $0xFE
+
+#define VPUNPCKHQDQ_X15_X2_X2 BYTE $0xC4; BYTE $0xC1; BYTE $0x69; BYTE $0x6D; BYTE $0xD7
+#define VPUNPCKHQDQ_X15_X3_X3 BYTE $0xC4; BYTE $0xC1; BYTE $0x61; BYTE $0x6D; BYTE $0xDF
+#define VPUNPCKHQDQ_X15_X6_X6 BYTE $0xC4; BYTE $0xC1; BYTE $0x49; BYTE $0x6D; BYTE $0xF7
+#define VPUNPCKHQDQ_X15_X7_X7 BYTE $0xC4; BYTE $0xC1; BYTE $0x41; BYTE $0x6D; BYTE $0xFF
+#define VPUNPCKHQDQ_X15_X3_X2 BYTE $0xC4; BYTE $0xC1; BYTE $0x61; BYTE $0x6D; BYTE $0xD7
+#define VPUNPCKHQDQ_X15_X7_X6 BYTE $0xC4; BYTE $0xC1; BYTE $0x41; BYTE $0x6D; BYTE $0xF7
+#define VPUNPCKHQDQ_X15_X13_X3 BYTE $0xC4; BYTE $0xC1; BYTE $0x11; BYTE $0x6D; BYTE $0xDF
+#define VPUNPCKHQDQ_X15_X13_X7 BYTE $0xC4; BYTE $0xC1; BYTE $0x11; BYTE $0x6D; BYTE $0xFF
+
+#define SHUFFLE_AVX() \
+	VMOVDQA X6, X13;         \
+	VMOVDQA X2, X14;         \
+	VMOVDQA X4, X6;          \
+	VPUNPCKLQDQ_X13_X13_X15; \
+	VMOVDQA X5, X4;          \
+	VMOVDQA X6, X5;          \
+	VPUNPCKHQDQ_X15_X7_X6;   \
+	VPUNPCKLQDQ_X7_X7_X15;   \
+	VPUNPCKHQDQ_X15_X13_X7;  \
+	VPUNPCKLQDQ_X3_X3_X15;   \
+	VPUNPCKHQDQ_X15_X2_X2;   \
+	VPUNPCKLQDQ_X14_X14_X15; \
+	VPUNPCKHQDQ_X15_X3_X3;   \
+
+#define SHUFFLE_AVX_INV() \
+	VMOVDQA X2, X13;         \
+	VMOVDQA X4, X14;         \
+	VPUNPCKLQDQ_X2_X2_X15;   \
+	VMOVDQA X5, X4;          \
+	VPUNPCKHQDQ_X15_X3_X2;   \
+	VMOVDQA X14, X5;         \
+	VPUNPCKLQDQ_X3_X3_X15;   \
+	VMOVDQA X6, X14;         \
+	VPUNPCKHQDQ_X15_X13_X3;  \
+	VPUNPCKLQDQ_X7_X7_X15;   \
+	VPUNPCKHQDQ_X15_X6_X6;   \
+	VPUNPCKLQDQ_X14_X14_X15; \
+	VPUNPCKHQDQ_X15_X7_X7;   \
+
+#define HALF_ROUND_AVX(v0, v1, v2, v3, v4, v5, v6, v7, m0, m1, m2, m3, t0, c40, c48) \
+	VPADDQ  m0, v0, v0;   \
+	VPADDQ  v2, v0, v0;   \
+	VPADDQ  m1, v1, v1;   \
+	VPADDQ  v3, v1, v1;   \
+	VPXOR   v0, v6, v6;   \
+	VPXOR   v1, v7, v7;   \
+	VPSHUFD $-79, v6, v6; \
+	VPSHUFD $-79, v7, v7; \
+	VPADDQ  v6, v4, v4;   \
+	VPADDQ  v7, v5, v5;   \
+	VPXOR   v4, v2, v2;   \
+	VPXOR   v5, v3, v3;   \
+	VPSHUFB c40, v2, v2;  \
+	VPSHUFB c40, v3, v3;  \
+	VPADDQ  m2, v0, v0;   \
+	VPADDQ  v2, v0, v0;   \
+	VPADDQ  m3, v1, v1;   \
+	VPADDQ  v3, v1, v1;   \
+	VPXOR   v0, v6, v6;   \
+	VPXOR   v1, v7, v7;   \
+	VPSHUFB c48, v6, v6;  \
+	VPSHUFB c48, v7, v7;  \
+	VPADDQ  v6, v4, v4;   \
+	VPADDQ  v7, v5, v5;   \
+	VPXOR   v4, v2, v2;   \
+	VPXOR   v5, v3, v3;   \
+	VPADDQ  v2, v2, t0;   \
+	VPSRLQ  $63, v2, v2;  \
+	VPXOR   t0, v2, v2;   \
+	VPADDQ  v3, v3, t0;   \
+	VPSRLQ  $63, v3, v3;  \
+	VPXOR   t0, v3, v3
+
+// load msg: X12 = (i0, i1), X13 = (i2, i3), X14 = (i4, i5), X15 = (i6, i7)
+// i0, i1, i2, i3, i4, i5, i6, i7 must not be 0
+#define LOAD_MSG_AVX(i0, i1, i2, i3, i4, i5, i6, i7) \
+	VMOVQ_SI_X12(i0*8);     \
+	VMOVQ_SI_X13(i2*8);     \
+	VMOVQ_SI_X14(i4*8);     \
+	VMOVQ_SI_X15(i6*8);     \
+	VPINSRQ_1_SI_X12(i1*8); \
+	VPINSRQ_1_SI_X13(i3*8); \
+	VPINSRQ_1_SI_X14(i5*8); \
+	VPINSRQ_1_SI_X15(i7*8)
+
+// load msg: X12 = (0, 2), X13 = (4, 6), X14 = (1, 3), X15 = (5, 7)
+#define LOAD_MSG_AVX_0_2_4_6_1_3_5_7() \
+	VMOVQ_SI_X12_0;        \
+	VMOVQ_SI_X13(4*8);     \
+	VMOVQ_SI_X14(1*8);     \
+	VMOVQ_SI_X15(5*8);     \
+	VPINSRQ_1_SI_X12(2*8); \
+	VPINSRQ_1_SI_X13(6*8); \
+	VPINSRQ_1_SI_X14(3*8); \
+	VPINSRQ_1_SI_X15(7*8)
+
+// load msg: X12 = (1, 0), X13 = (11, 5), X14 = (12, 2), X15 = (7, 3)
+#define LOAD_MSG_AVX_1_0_11_5_12_2_7_3() \
+	VPSHUFD $0x4E, 0*8(SI), X12; \
+	VMOVQ_SI_X13(11*8);          \
+	VMOVQ_SI_X14(12*8);          \
+	VMOVQ_SI_X15(7*8);           \
+	VPINSRQ_1_SI_X13(5*8);       \
+	VPINSRQ_1_SI_X14(2*8);       \
+	VPINSRQ_1_SI_X15(3*8)
+
+// load msg: X12 = (11, 12), X13 = (5, 15), X14 = (8, 0), X15 = (2, 13)
+#define LOAD_MSG_AVX_11_12_5_15_8_0_2_13() \
+	VMOVDQU 11*8(SI), X12;  \
+	VMOVQ_SI_X13(5*8);      \
+	VMOVQ_SI_X14(8*8);      \
+	VMOVQ_SI_X15(2*8);      \
+	VPINSRQ_1_SI_X13(15*8); \
+	VPINSRQ_1_SI_X14_0;     \
+	VPINSRQ_1_SI_X15(13*8)
+
+// load msg: X12 = (2, 5), X13 = (4, 15), X14 = (6, 10), X15 = (0, 8)
+#define LOAD_MSG_AVX_2_5_4_15_6_10_0_8() \
+	VMOVQ_SI_X12(2*8);      \
+	VMOVQ_SI_X13(4*8);      \
+	VMOVQ_SI_X14(6*8);      \
+	VMOVQ_SI_X15_0;         \
+	VPINSRQ_1_SI_X12(5*8);  \
+	VPINSRQ_1_SI_X13(15*8); \
+	VPINSRQ_1_SI_X14(10*8); \
+	VPINSRQ_1_SI_X15(8*8)
+
+// load msg: X12 = (9, 5), X13 = (2, 10), X14 = (0, 7), X15 = (4, 15)
+#define LOAD_MSG_AVX_9_5_2_10_0_7_4_15() \
+	VMOVQ_SI_X12(9*8);      \
+	VMOVQ_SI_X13(2*8);      \
+	VMOVQ_SI_X14_0;         \
+	VMOVQ_SI_X15(4*8);      \
+	VPINSRQ_1_SI_X12(5*8);  \
+	VPINSRQ_1_SI_X13(10*8); \
+	VPINSRQ_1_SI_X14(7*8);  \
+	VPINSRQ_1_SI_X15(15*8)
+
+// load msg: X12 = (2, 6), X13 = (0, 8), X14 = (12, 10), X15 = (11, 3)
+#define LOAD_MSG_AVX_2_6_0_8_12_10_11_3() \
+	VMOVQ_SI_X12(2*8);      \
+	VMOVQ_SI_X13_0;         \
+	VMOVQ_SI_X14(12*8);     \
+	VMOVQ_SI_X15(11*8);     \
+	VPINSRQ_1_SI_X12(6*8);  \
+	VPINSRQ_1_SI_X13(8*8);  \
+	VPINSRQ_1_SI_X14(10*8); \
+	VPINSRQ_1_SI_X15(3*8)
+
+// load msg: X12 = (0, 6), X13 = (9, 8), X14 = (7, 3), X15 = (2, 11)
+#define LOAD_MSG_AVX_0_6_9_8_7_3_2_11() \
+	MOVQ    0*8(SI), X12;        \
+	VPSHUFD $0x4E, 8*8(SI), X13; \
+	MOVQ    7*8(SI), X14;        \
+	MOVQ    2*8(SI), X15;        \
+	VPINSRQ_1_SI_X12(6*8);       \
+	VPINSRQ_1_SI_X14(3*8);       \
+	VPINSRQ_1_SI_X15(11*8)
+
+// load msg: X12 = (6, 14), X13 = (11, 0), X14 = (15, 9), X15 = (3, 8)
+#define LOAD_MSG_AVX_6_14_11_0_15_9_3_8() \
+	MOVQ 6*8(SI), X12;      \
+	MOVQ 11*8(SI), X13;     \
+	MOVQ 15*8(SI), X14;     \
+	MOVQ 3*8(SI), X15;      \
+	VPINSRQ_1_SI_X12(14*8); \
+	VPINSRQ_1_SI_X13_0;     \
+	VPINSRQ_1_SI_X14(9*8);  \
+	VPINSRQ_1_SI_X15(8*8)
+
+// load msg: X12 = (5, 15), X13 = (8, 2), X14 = (0, 4), X15 = (6, 10)
+#define LOAD_MSG_AVX_5_15_8_2_0_4_6_10() \
+	MOVQ 5*8(SI), X12;      \
+	MOVQ 8*8(SI), X13;      \
+	MOVQ 0*8(SI), X14;      \
+	MOVQ 6*8(SI), X15;      \
+	VPINSRQ_1_SI_X12(15*8); \
+	VPINSRQ_1_SI_X13(2*8);  \
+	VPINSRQ_1_SI_X14(4*8);  \
+	VPINSRQ_1_SI_X15(10*8)
+
+// load msg: X12 = (12, 13), X13 = (1, 10), X14 = (2, 7), X15 = (4, 5)
+#define LOAD_MSG_AVX_12_13_1_10_2_7_4_5() \
+	VMOVDQU 12*8(SI), X12;  \
+	MOVQ    1*8(SI), X13;   \
+	MOVQ    2*8(SI), X14;   \
+	VPINSRQ_1_SI_X13(10*8); \
+	VPINSRQ_1_SI_X14(7*8);  \
+	VMOVDQU 4*8(SI), X15
+
+// load msg: X12 = (15, 9), X13 = (3, 13), X14 = (11, 14), X15 = (12, 0)
+#define LOAD_MSG_AVX_15_9_3_13_11_14_12_0() \
+	MOVQ 15*8(SI), X12;     \
+	MOVQ 3*8(SI), X13;      \
+	MOVQ 11*8(SI), X14;     \
+	MOVQ 12*8(SI), X15;     \
+	VPINSRQ_1_SI_X12(9*8);  \
+	VPINSRQ_1_SI_X13(13*8); \
+	VPINSRQ_1_SI_X14(14*8); \
+	VPINSRQ_1_SI_X15_0
+
+// func hashBlocksAVX(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+TEXT ·hashBlocksAVX(SB), 4, $288-48 // frame size = 272 + 16 byte alignment
+	MOVQ h+0(FP), AX
+	MOVQ c+8(FP), BX
+	MOVQ flag+16(FP), CX
+	MOVQ blocks_base+24(FP), SI
+	MOVQ blocks_len+32(FP), DI
+
+	MOVQ SP, R10
+	ADDQ $15, R10
+	ANDQ $~15, R10
+
+	VMOVDQU ·AVX_c40<>(SB), X0
+	VMOVDQU ·AVX_c48<>(SB), X1
+	VMOVDQA X0, X8
+	VMOVDQA X1, X9
+
+	VMOVDQU ·AVX_iv3<>(SB), X0
+	VMOVDQA X0, 0(R10)
+	XORQ    CX, 0(R10)          // 0(R10) = ·AVX_iv3 ^ (CX || 0)
+
+	VMOVDQU 0(AX), X10
+	VMOVDQU 16(AX), X11
+	VMOVDQU 32(AX), X2
+	VMOVDQU 48(AX), X3
+
+	MOVQ 0(BX), R8
+	MOVQ 8(BX), R9
+
+loop:
+	ADDQ $128, R8
+	CMPQ R8, $128
+	JGE  noinc
+	INCQ R9
+
+noinc:
+	VMOVQ_R8_X15
+	VPINSRQ_1_R9_X15
+
+	VMOVDQA X10, X0
+	VMOVDQA X11, X1
+	VMOVDQU ·AVX_iv0<>(SB), X4
+	VMOVDQU ·AVX_iv1<>(SB), X5
+	VMOVDQU ·AVX_iv2<>(SB), X6
+
+	VPXOR   X15, X6, X6
+	VMOVDQA 0(R10), X7
+
+	LOAD_MSG_AVX_0_2_4_6_1_3_5_7()
+	VMOVDQA X12, 16(R10)
+	VMOVDQA X13, 32(R10)
+	VMOVDQA X14, 48(R10)
+	VMOVDQA X15, 64(R10)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX(8, 10, 12, 14, 9, 11, 13, 15)
+	VMOVDQA X12, 80(R10)
+	VMOVDQA X13, 96(R10)
+	VMOVDQA X14, 112(R10)
+	VMOVDQA X15, 128(R10)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX(14, 4, 9, 13, 10, 8, 15, 6)
+	VMOVDQA X12, 144(R10)
+	VMOVDQA X13, 160(R10)
+	VMOVDQA X14, 176(R10)
+	VMOVDQA X15, 192(R10)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_1_0_11_5_12_2_7_3()
+	VMOVDQA X12, 208(R10)
+	VMOVDQA X13, 224(R10)
+	VMOVDQA X14, 240(R10)
+	VMOVDQA X15, 256(R10)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX_11_12_5_15_8_0_2_13()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX(10, 3, 7, 9, 14, 6, 1, 4)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX(7, 3, 13, 11, 9, 1, 12, 14)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_2_5_4_15_6_10_0_8()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX_9_5_2_10_0_7_4_15()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX(14, 11, 6, 3, 1, 12, 8, 13)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX_2_6_0_8_12_10_11_3()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX(4, 7, 15, 1, 13, 5, 14, 9)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX(12, 1, 14, 4, 5, 15, 13, 10)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_0_6_9_8_7_3_2_11()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX(13, 7, 12, 3, 11, 14, 1, 9)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_5_15_8_2_0_4_6_10()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX_6_14_11_0_15_9_3_8()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_12_13_1_10_2_7_4_5()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX(10, 8, 7, 1, 2, 4, 6, 5)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_15_9_3_13_11_14_12_0()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 16(R10), 32(R10), 48(R10), 64(R10), X15, X8, X9)
+	SHUFFLE_AVX()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 80(R10), 96(R10), 112(R10), 128(R10), X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 144(R10), 160(R10), 176(R10), 192(R10), X15, X8, X9)
+	SHUFFLE_AVX()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 208(R10), 224(R10), 240(R10), 256(R10), X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	VMOVDQU 32(AX), X14
+	VMOVDQU 48(AX), X15
+	VPXOR   X0, X10, X10
+	VPXOR   X1, X11, X11
+	VPXOR   X2, X14, X14
+	VPXOR   X3, X15, X15
+	VPXOR   X4, X10, X10
+	VPXOR   X5, X11, X11
+	VPXOR   X6, X14, X2
+	VPXOR   X7, X15, X3
+	VMOVDQU X2, 32(AX)
+	VMOVDQU X3, 48(AX)
+
+	LEAQ 128(SI), SI
+	SUBQ $128, DI
+	JNE  loop
+
+	VMOVDQU X10, 0(AX)
+	VMOVDQU X11, 16(AX)
+
+	MOVQ R8, 0(BX)
+	MOVQ R9, 8(BX)
+	VZEROUPPER
+
+	RET

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

@@ -0,0 +1,25 @@
+// 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.
+
+//go:build !go1.7 && amd64 && gc && !purego
+// +build !go1.7,amd64,gc,!purego
+
+package blake2b
+
+import "golang.org/x/sys/cpu"
+
+func init() {
+	useSSE4 = cpu.X86.HasSSE41
+}
+
+//go:noescape
+func hashBlocksSSE4(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+
+func hashBlocks(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte) {
+	if useSSE4 {
+		hashBlocksSSE4(h, c, flag, blocks)
+	} else {
+		hashBlocksGeneric(h, c, flag, blocks)
+	}
+}

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

@@ -0,0 +1,279 @@
+// 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.
+
+//go:build amd64 && gc && !purego
+// +build amd64,gc,!purego
+
+#include "textflag.h"
+
+DATA ·iv0<>+0x00(SB)/8, $0x6a09e667f3bcc908
+DATA ·iv0<>+0x08(SB)/8, $0xbb67ae8584caa73b
+GLOBL ·iv0<>(SB), (NOPTR+RODATA), $16
+
+DATA ·iv1<>+0x00(SB)/8, $0x3c6ef372fe94f82b
+DATA ·iv1<>+0x08(SB)/8, $0xa54ff53a5f1d36f1
+GLOBL ·iv1<>(SB), (NOPTR+RODATA), $16
+
+DATA ·iv2<>+0x00(SB)/8, $0x510e527fade682d1
+DATA ·iv2<>+0x08(SB)/8, $0x9b05688c2b3e6c1f
+GLOBL ·iv2<>(SB), (NOPTR+RODATA), $16
+
+DATA ·iv3<>+0x00(SB)/8, $0x1f83d9abfb41bd6b
+DATA ·iv3<>+0x08(SB)/8, $0x5be0cd19137e2179
+GLOBL ·iv3<>(SB), (NOPTR+RODATA), $16
+
+DATA ·c40<>+0x00(SB)/8, $0x0201000706050403
+DATA ·c40<>+0x08(SB)/8, $0x0a09080f0e0d0c0b
+GLOBL ·c40<>(SB), (NOPTR+RODATA), $16
+
+DATA ·c48<>+0x00(SB)/8, $0x0100070605040302
+DATA ·c48<>+0x08(SB)/8, $0x09080f0e0d0c0b0a
+GLOBL ·c48<>(SB), (NOPTR+RODATA), $16
+
+#define SHUFFLE(v2, v3, v4, v5, v6, v7, t1, t2) \
+	MOVO       v4, t1; \
+	MOVO       v5, v4; \
+	MOVO       t1, v5; \
+	MOVO       v6, t1; \
+	PUNPCKLQDQ v6, t2; \
+	PUNPCKHQDQ v7, v6; \
+	PUNPCKHQDQ t2, v6; \
+	PUNPCKLQDQ v7, t2; \
+	MOVO       t1, v7; \
+	MOVO       v2, t1; \
+	PUNPCKHQDQ t2, v7; \
+	PUNPCKLQDQ v3, t2; \
+	PUNPCKHQDQ t2, v2; \
+	PUNPCKLQDQ t1, t2; \
+	PUNPCKHQDQ t2, v3
+
+#define SHUFFLE_INV(v2, v3, v4, v5, v6, v7, t1, t2) \
+	MOVO       v4, t1; \
+	MOVO       v5, v4; \
+	MOVO       t1, v5; \
+	MOVO       v2, t1; \
+	PUNPCKLQDQ v2, t2; \
+	PUNPCKHQDQ v3, v2; \
+	PUNPCKHQDQ t2, v2; \
+	PUNPCKLQDQ v3, t2; \
+	MOVO       t1, v3; \
+	MOVO       v6, t1; \
+	PUNPCKHQDQ t2, v3; \
+	PUNPCKLQDQ v7, t2; \
+	PUNPCKHQDQ t2, v6; \
+	PUNPCKLQDQ t1, t2; \
+	PUNPCKHQDQ t2, v7
+
+#define HALF_ROUND(v0, v1, v2, v3, v4, v5, v6, v7, m0, m1, m2, m3, t0, c40, c48) \
+	PADDQ  m0, v0;        \
+	PADDQ  m1, v1;        \
+	PADDQ  v2, v0;        \
+	PADDQ  v3, v1;        \
+	PXOR   v0, v6;        \
+	PXOR   v1, v7;        \
+	PSHUFD $0xB1, v6, v6; \
+	PSHUFD $0xB1, v7, v7; \
+	PADDQ  v6, v4;        \
+	PADDQ  v7, v5;        \
+	PXOR   v4, v2;        \
+	PXOR   v5, v3;        \
+	PSHUFB c40, v2;       \
+	PSHUFB c40, v3;       \
+	PADDQ  m2, v0;        \
+	PADDQ  m3, v1;        \
+	PADDQ  v2, v0;        \
+	PADDQ  v3, v1;        \
+	PXOR   v0, v6;        \
+	PXOR   v1, v7;        \
+	PSHUFB c48, v6;       \
+	PSHUFB c48, v7;       \
+	PADDQ  v6, v4;        \
+	PADDQ  v7, v5;        \
+	PXOR   v4, v2;        \
+	PXOR   v5, v3;        \
+	MOVOU  v2, t0;        \
+	PADDQ  v2, t0;        \
+	PSRLQ  $63, v2;       \
+	PXOR   t0, v2;        \
+	MOVOU  v3, t0;        \
+	PADDQ  v3, t0;        \
+	PSRLQ  $63, v3;       \
+	PXOR   t0, v3
+
+#define LOAD_MSG(m0, m1, m2, m3, src, i0, i1, i2, i3, i4, i5, i6, i7) \
+	MOVQ   i0*8(src), m0;     \
+	PINSRQ $1, i1*8(src), m0; \
+	MOVQ   i2*8(src), m1;     \
+	PINSRQ $1, i3*8(src), m1; \
+	MOVQ   i4*8(src), m2;     \
+	PINSRQ $1, i5*8(src), m2; \
+	MOVQ   i6*8(src), m3;     \
+	PINSRQ $1, i7*8(src), m3
+
+// func hashBlocksSSE4(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+TEXT ·hashBlocksSSE4(SB), 4, $288-48 // frame size = 272 + 16 byte alignment
+	MOVQ h+0(FP), AX
+	MOVQ c+8(FP), BX
+	MOVQ flag+16(FP), CX
+	MOVQ blocks_base+24(FP), SI
+	MOVQ blocks_len+32(FP), DI
+
+	MOVQ SP, R10
+	ADDQ $15, R10
+	ANDQ $~15, R10
+
+	MOVOU ·iv3<>(SB), X0
+	MOVO  X0, 0(R10)
+	XORQ  CX, 0(R10)     // 0(R10) = ·iv3 ^ (CX || 0)
+
+	MOVOU ·c40<>(SB), X13
+	MOVOU ·c48<>(SB), X14
+
+	MOVOU 0(AX), X12
+	MOVOU 16(AX), X15
+
+	MOVQ 0(BX), R8
+	MOVQ 8(BX), R9
+
+loop:
+	ADDQ $128, R8
+	CMPQ R8, $128
+	JGE  noinc
+	INCQ R9
+
+noinc:
+	MOVQ R8, X8
+	PINSRQ $1, R9, X8
+
+	MOVO X12, X0
+	MOVO X15, X1
+	MOVOU 32(AX), X2
+	MOVOU 48(AX), X3
+	MOVOU ·iv0<>(SB), X4
+	MOVOU ·iv1<>(SB), X5
+	MOVOU ·iv2<>(SB), X6
+
+	PXOR X8, X6
+	MOVO 0(R10), X7
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 0, 2, 4, 6, 1, 3, 5, 7)
+	MOVO X8, 16(R10)
+	MOVO X9, 32(R10)
+	MOVO X10, 48(R10)
+	MOVO X11, 64(R10)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 8, 10, 12, 14, 9, 11, 13, 15)
+	MOVO X8, 80(R10)
+	MOVO X9, 96(R10)
+	MOVO X10, 112(R10)
+	MOVO X11, 128(R10)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 14, 4, 9, 13, 10, 8, 15, 6)
+	MOVO X8, 144(R10)
+	MOVO X9, 160(R10)
+	MOVO X10, 176(R10)
+	MOVO X11, 192(R10)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 1, 0, 11, 5, 12, 2, 7, 3)
+	MOVO X8, 208(R10)
+	MOVO X9, 224(R10)
+	MOVO X10, 240(R10)
+	MOVO X11, 256(R10)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 11, 12, 5, 15, 8, 0, 2, 13)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 10, 3, 7, 9, 14, 6, 1, 4)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 7, 3, 13, 11, 9, 1, 12, 14)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 2, 5, 4, 15, 6, 10, 0, 8)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 9, 5, 2, 10, 0, 7, 4, 15)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 14, 11, 6, 3, 1, 12, 8, 13)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 2, 6, 0, 8, 12, 10, 11, 3)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 4, 7, 15, 1, 13, 5, 14, 9)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 12, 1, 14, 4, 5, 15, 13, 10)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 0, 6, 9, 8, 7, 3, 2, 11)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 13, 7, 12, 3, 11, 14, 1, 9)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 5, 15, 8, 2, 0, 4, 6, 10)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 6, 14, 11, 0, 15, 9, 3, 8)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 12, 13, 1, 10, 2, 7, 4, 5)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 10, 8, 7, 1, 2, 4, 6, 5)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 15, 9, 3, 13, 11, 14, 12, 0)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 16(R10), 32(R10), 48(R10), 64(R10), X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 80(R10), 96(R10), 112(R10), 128(R10), X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 144(R10), 160(R10), 176(R10), 192(R10), X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 208(R10), 224(R10), 240(R10), 256(R10), X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	MOVOU 32(AX), X10
+	MOVOU 48(AX), X11
+	PXOR  X0, X12
+	PXOR  X1, X15
+	PXOR  X2, X10
+	PXOR  X3, X11
+	PXOR  X4, X12
+	PXOR  X5, X15
+	PXOR  X6, X10
+	PXOR  X7, X11
+	MOVOU X10, 32(AX)
+	MOVOU X11, 48(AX)
+
+	LEAQ 128(SI), SI
+	SUBQ $128, DI
+	JNE  loop
+
+	MOVOU X12, 0(AX)
+	MOVOU X15, 16(AX)
+
+	MOVQ R8, 0(BX)
+	MOVQ R9, 8(BX)
+
+	RET

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

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

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

@@ -0,0 +1,12 @@
+// 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.
+
+//go:build !amd64 || purego || !gc
+// +build !amd64 purego !gc
+
+package blake2b
+
+func hashBlocks(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte) {
+	hashBlocksGeneric(h, c, flag, blocks)
+}

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

@@ -0,0 +1,177 @@
+// Copyright 2017 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package blake2b
+
+import (
+	"encoding/binary"
+	"errors"
+	"io"
+)
+
+// XOF defines the interface to hash functions that
+// support arbitrary-length output.
+type XOF interface {
+	// Write absorbs more data into the hash's state. It panics if called
+	// after Read.
+	io.Writer
+
+	// Read reads more output from the hash. It returns io.EOF if the limit
+	// has been reached.
+	io.Reader
+
+	// Clone returns a copy of the XOF in its current state.
+	Clone() XOF
+
+	// Reset resets the XOF to its initial state.
+	Reset()
+}
+
+// OutputLengthUnknown can be used as the size argument to NewXOF to indicate
+// the length of the output is not known in advance.
+const OutputLengthUnknown = 0
+
+// magicUnknownOutputLength is a magic value for the output size that indicates
+// an unknown number of output bytes.
+const magicUnknownOutputLength = (1 << 32) - 1
+
+// maxOutputLength is the absolute maximum number of bytes to produce when the
+// number of output bytes is unknown.
+const maxOutputLength = (1 << 32) * 64
+
+// NewXOF creates a new variable-output-length hash. The hash either produce a
+// known number of bytes (1 <= size < 2**32-1), or an unknown number of bytes
+// (size == OutputLengthUnknown). In the latter case, an absolute limit of
+// 256GiB applies.
+//
+// A non-nil key turns the hash into a MAC. The key must between
+// zero and 32 bytes long.
+func NewXOF(size uint32, key []byte) (XOF, error) {
+	if len(key) > Size {
+		return nil, errKeySize
+	}
+	if size == magicUnknownOutputLength {
+		// 2^32-1 indicates an unknown number of bytes and thus isn't a
+		// valid length.
+		return nil, errors.New("blake2b: XOF length too large")
+	}
+	if size == OutputLengthUnknown {
+		size = magicUnknownOutputLength
+	}
+	x := &xof{
+		d: digest{
+			size:   Size,
+			keyLen: len(key),
+		},
+		length: size,
+	}
+	copy(x.d.key[:], key)
+	x.Reset()
+	return x, nil
+}
+
+type xof struct {
+	d                digest
+	length           uint32
+	remaining        uint64
+	cfg, root, block [Size]byte
+	offset           int
+	nodeOffset       uint32
+	readMode         bool
+}
+
+func (x *xof) Write(p []byte) (n int, err error) {
+	if x.readMode {
+		panic("blake2b: write to XOF after read")
+	}
+	return x.d.Write(p)
+}
+
+func (x *xof) Clone() XOF {
+	clone := *x
+	return &clone
+}
+
+func (x *xof) Reset() {
+	x.cfg[0] = byte(Size)
+	binary.LittleEndian.PutUint32(x.cfg[4:], uint32(Size)) // leaf length
+	binary.LittleEndian.PutUint32(x.cfg[12:], x.length)    // XOF length
+	x.cfg[17] = byte(Size)                                 // inner hash size
+
+	x.d.Reset()
+	x.d.h[1] ^= uint64(x.length) << 32
+
+	x.remaining = uint64(x.length)
+	if x.remaining == magicUnknownOutputLength {
+		x.remaining = maxOutputLength
+	}
+	x.offset, x.nodeOffset = 0, 0
+	x.readMode = false
+}
+
+func (x *xof) Read(p []byte) (n int, err error) {
+	if !x.readMode {
+		x.d.finalize(&x.root)
+		x.readMode = true
+	}
+
+	if x.remaining == 0 {
+		return 0, io.EOF
+	}
+
+	n = len(p)
+	if uint64(n) > x.remaining {
+		n = int(x.remaining)
+		p = p[:n]
+	}
+
+	if x.offset > 0 {
+		blockRemaining := Size - x.offset
+		if n < blockRemaining {
+			x.offset += copy(p, x.block[x.offset:])
+			x.remaining -= uint64(n)
+			return
+		}
+		copy(p, x.block[x.offset:])
+		p = p[blockRemaining:]
+		x.offset = 0
+		x.remaining -= uint64(blockRemaining)
+	}
+
+	for len(p) >= Size {
+		binary.LittleEndian.PutUint32(x.cfg[8:], x.nodeOffset)
+		x.nodeOffset++
+
+		x.d.initConfig(&x.cfg)
+		x.d.Write(x.root[:])
+		x.d.finalize(&x.block)
+
+		copy(p, x.block[:])
+		p = p[Size:]
+		x.remaining -= uint64(Size)
+	}
+
+	if todo := len(p); todo > 0 {
+		if x.remaining < uint64(Size) {
+			x.cfg[0] = byte(x.remaining)
+		}
+		binary.LittleEndian.PutUint32(x.cfg[8:], x.nodeOffset)
+		x.nodeOffset++
+
+		x.d.initConfig(&x.cfg)
+		x.d.Write(x.root[:])
+		x.d.finalize(&x.block)
+
+		x.offset = copy(p, x.block[:todo])
+		x.remaining -= uint64(todo)
+	}
+	return
+}
+
+func (d *digest) initConfig(cfg *[Size]byte) {
+	d.offset, d.c[0], d.c[1] = 0, 0, 0
+	for i := range d.h {
+		d.h[i] = iv[i] ^ binary.LittleEndian.Uint64(cfg[i*8:])
+	}
+}

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

@@ -0,0 +1,33 @@
+// Copyright 2017 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build go1.9
+// +build go1.9
+
+package blake2b
+
+import (
+	"crypto"
+	"hash"
+)
+
+func init() {
+	newHash256 := func() hash.Hash {
+		h, _ := New256(nil)
+		return h
+	}
+	newHash384 := func() hash.Hash {
+		h, _ := New384(nil)
+		return h
+	}
+
+	newHash512 := func() hash.Hash {
+		h, _ := New512(nil)
+		return h
+	}
+
+	crypto.RegisterHash(crypto.BLAKE2b_256, newHash256)
+	crypto.RegisterHash(crypto.BLAKE2b_384, newHash384)
+	crypto.RegisterHash(crypto.BLAKE2b_512, newHash512)
+}

vendor/golang.org/x/crypto/blake2s/blake2s.go

@@ -0,0 +1,246 @@
+// 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 blake2s implements the BLAKE2s hash algorithm defined by RFC 7693
+// and the extendable output function (XOF) BLAKE2Xs.
+//
+// BLAKE2s is optimized for 8- to 32-bit platforms and produces digests of any
+// size between 1 and 32 bytes.
+// For a detailed specification of BLAKE2s see https://blake2.net/blake2.pdf
+// and for BLAKE2Xs see https://blake2.net/blake2x.pdf
+//
+// If you aren't sure which function you need, use BLAKE2s (Sum256 or New256).
+// If you need a secret-key MAC (message authentication code), use the New256
+// function with a non-nil key.
+//
+// BLAKE2X is a construction to compute hash values larger than 32 bytes. It
+// can produce hash values between 0 and 65535 bytes.
+package blake2s // import "golang.org/x/crypto/blake2s"
+
+import (
+	"encoding/binary"
+	"errors"
+	"hash"
+)
+
+const (
+	// The blocksize of BLAKE2s in bytes.
+	BlockSize = 64
+
+	// The hash size of BLAKE2s-256 in bytes.
+	Size = 32
+
+	// The hash size of BLAKE2s-128 in bytes.
+	Size128 = 16
+)
+
+var errKeySize = errors.New("blake2s: invalid key size")
+
+var iv = [8]uint32{
+	0x6a09e667, 0xbb67ae85, 0x3c6ef372, 0xa54ff53a,
+	0x510e527f, 0x9b05688c, 0x1f83d9ab, 0x5be0cd19,
+}
+
+// Sum256 returns the BLAKE2s-256 checksum of the data.
+func Sum256(data []byte) [Size]byte {
+	var sum [Size]byte
+	checkSum(&sum, Size, data)
+	return sum
+}
+
+// New256 returns a new hash.Hash computing the BLAKE2s-256 checksum. A non-nil
+// key turns the hash into a MAC. The key must between zero and 32 bytes long.
+// When the key is nil, the returned hash.Hash implements BinaryMarshaler
+// and BinaryUnmarshaler for state (de)serialization as documented by hash.Hash.
+func New256(key []byte) (hash.Hash, error) { return newDigest(Size, key) }
+
+// New128 returns a new hash.Hash computing the BLAKE2s-128 checksum given a
+// non-empty key. Note that a 128-bit digest is too small to be secure as a
+// cryptographic hash and should only be used as a MAC, thus the key argument
+// is not optional.
+func New128(key []byte) (hash.Hash, error) {
+	if len(key) == 0 {
+		return nil, errors.New("blake2s: a key is required for a 128-bit hash")
+	}
+	return newDigest(Size128, key)
+}
+
+func newDigest(hashSize int, key []byte) (*digest, error) {
+	if len(key) > Size {
+		return nil, errKeySize
+	}
+	d := &digest{
+		size:   hashSize,
+		keyLen: len(key),
+	}
+	copy(d.key[:], key)
+	d.Reset()
+	return d, nil
+}
+
+func checkSum(sum *[Size]byte, hashSize int, data []byte) {
+	var (
+		h [8]uint32
+		c [2]uint32
+	)
+
+	h = iv
+	h[0] ^= uint32(hashSize) | (1 << 16) | (1 << 24)
+
+	if length := len(data); length > BlockSize {
+		n := length &^ (BlockSize - 1)
+		if length == n {
+			n -= BlockSize
+		}
+		hashBlocks(&h, &c, 0, data[:n])
+		data = data[n:]
+	}
+
+	var block [BlockSize]byte
+	offset := copy(block[:], data)
+	remaining := uint32(BlockSize - offset)
+
+	if c[0] < remaining {
+		c[1]--
+	}
+	c[0] -= remaining
+
+	hashBlocks(&h, &c, 0xFFFFFFFF, block[:])
+
+	for i, v := range h {
+		binary.LittleEndian.PutUint32(sum[4*i:], v)
+	}
+}
+
+type digest struct {
+	h      [8]uint32
+	c      [2]uint32
+	size   int
+	block  [BlockSize]byte
+	offset int
+
+	key    [BlockSize]byte
+	keyLen int
+}
+
+const (
+	magic         = "b2s"
+	marshaledSize = len(magic) + 8*4 + 2*4 + 1 + BlockSize + 1
+)
+
+func (d *digest) MarshalBinary() ([]byte, error) {
+	if d.keyLen != 0 {
+		return nil, errors.New("crypto/blake2s: cannot marshal MACs")
+	}
+	b := make([]byte, 0, marshaledSize)
+	b = append(b, magic...)
+	for i := 0; i < 8; i++ {
+		b = appendUint32(b, d.h[i])
+	}
+	b = appendUint32(b, d.c[0])
+	b = appendUint32(b, d.c[1])
+	// Maximum value for size is 32
+	b = append(b, byte(d.size))
+	b = append(b, d.block[:]...)
+	b = append(b, byte(d.offset))
+	return b, nil
+}
+
+func (d *digest) UnmarshalBinary(b []byte) error {
+	if len(b) < len(magic) || string(b[:len(magic)]) != magic {
+		return errors.New("crypto/blake2s: invalid hash state identifier")
+	}
+	if len(b) != marshaledSize {
+		return errors.New("crypto/blake2s: invalid hash state size")
+	}
+	b = b[len(magic):]
+	for i := 0; i < 8; i++ {
+		b, d.h[i] = consumeUint32(b)
+	}
+	b, d.c[0] = consumeUint32(b)
+	b, d.c[1] = consumeUint32(b)
+	d.size = int(b[0])
+	b = b[1:]
+	copy(d.block[:], b[:BlockSize])
+	b = b[BlockSize:]
+	d.offset = int(b[0])
+	return nil
+}
+
+func (d *digest) BlockSize() int { return BlockSize }
+
+func (d *digest) Size() int { return d.size }
+
+func (d *digest) Reset() {
+	d.h = iv
+	d.h[0] ^= uint32(d.size) | (uint32(d.keyLen) << 8) | (1 << 16) | (1 << 24)
+	d.offset, d.c[0], d.c[1] = 0, 0, 0
+	if d.keyLen > 0 {
+		d.block = d.key
+		d.offset = BlockSize
+	}
+}
+
+func (d *digest) Write(p []byte) (n int, err error) {
+	n = len(p)
+
+	if d.offset > 0 {
+		remaining := BlockSize - d.offset
+		if n <= remaining {
+			d.offset += copy(d.block[d.offset:], p)
+			return
+		}
+		copy(d.block[d.offset:], p[:remaining])
+		hashBlocks(&d.h, &d.c, 0, d.block[:])
+		d.offset = 0
+		p = p[remaining:]
+	}
+
+	if length := len(p); length > BlockSize {
+		nn := length &^ (BlockSize - 1)
+		if length == nn {
+			nn -= BlockSize
+		}
+		hashBlocks(&d.h, &d.c, 0, p[:nn])
+		p = p[nn:]
+	}
+
+	d.offset += copy(d.block[:], p)
+	return
+}
+
+func (d *digest) Sum(sum []byte) []byte {
+	var hash [Size]byte
+	d.finalize(&hash)
+	return append(sum, hash[:d.size]...)
+}
+
+func (d *digest) finalize(hash *[Size]byte) {
+	var block [BlockSize]byte
+	h := d.h
+	c := d.c
+
+	copy(block[:], d.block[:d.offset])
+	remaining := uint32(BlockSize - d.offset)
+	if c[0] < remaining {
+		c[1]--
+	}
+	c[0] -= remaining
+
+	hashBlocks(&h, &c, 0xFFFFFFFF, block[:])
+	for i, v := range h {
+		binary.LittleEndian.PutUint32(hash[4*i:], v)
+	}
+}
+
+func appendUint32(b []byte, x uint32) []byte {
+	var a [4]byte
+	binary.BigEndian.PutUint32(a[:], x)
+	return append(b, a[:]...)
+}
+
+func consumeUint32(b []byte) ([]byte, uint32) {
+	x := binary.BigEndian.Uint32(b)
+	return b[4:], x
+}

vendor/golang.org/x/crypto/blake2s/blake2s_386.go

@@ -0,0 +1,33 @@
+// 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.
+
+//go:build 386 && gc && !purego
+// +build 386,gc,!purego
+
+package blake2s
+
+import "golang.org/x/sys/cpu"
+
+var (
+	useSSE4  = false
+	useSSSE3 = cpu.X86.HasSSSE3
+	useSSE2  = cpu.X86.HasSSE2
+)
+
+//go:noescape
+func hashBlocksSSE2(h *[8]uint32, c *[2]uint32, flag uint32, blocks []byte)
+
+//go:noescape
+func hashBlocksSSSE3(h *[8]uint32, c *[2]uint32, flag uint32, blocks []byte)
+
+func hashBlocks(h *[8]uint32, c *[2]uint32, flag uint32, blocks []byte) {
+	switch {
+	case useSSSE3:
+		hashBlocksSSSE3(h, c, flag, blocks)
+	case useSSE2:
+		hashBlocksSSE2(h, c, flag, blocks)
+	default:
+		hashBlocksGeneric(h, c, flag, blocks)
+	}
+}

vendor/golang.org/x/crypto/blake2s/blake2s_386.s

@@ -0,0 +1,430 @@
+// 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.
+
+//go:build 386 && gc && !purego
+// +build 386,gc,!purego
+
+#include "textflag.h"
+
+DATA iv0<>+0x00(SB)/4, $0x6a09e667
+DATA iv0<>+0x04(SB)/4, $0xbb67ae85
+DATA iv0<>+0x08(SB)/4, $0x3c6ef372
+DATA iv0<>+0x0c(SB)/4, $0xa54ff53a
+GLOBL iv0<>(SB), (NOPTR+RODATA), $16
+
+DATA iv1<>+0x00(SB)/4, $0x510e527f
+DATA iv1<>+0x04(SB)/4, $0x9b05688c
+DATA iv1<>+0x08(SB)/4, $0x1f83d9ab
+DATA iv1<>+0x0c(SB)/4, $0x5be0cd19
+GLOBL iv1<>(SB), (NOPTR+RODATA), $16
+
+DATA rol16<>+0x00(SB)/8, $0x0504070601000302
+DATA rol16<>+0x08(SB)/8, $0x0D0C0F0E09080B0A
+GLOBL rol16<>(SB), (NOPTR+RODATA), $16
+
+DATA rol8<>+0x00(SB)/8, $0x0407060500030201
+DATA rol8<>+0x08(SB)/8, $0x0C0F0E0D080B0A09
+GLOBL rol8<>(SB), (NOPTR+RODATA), $16
+
+DATA counter<>+0x00(SB)/8, $0x40
+DATA counter<>+0x08(SB)/8, $0x0
+GLOBL counter<>(SB), (NOPTR+RODATA), $16
+
+#define ROTL_SSE2(n, t, v) \
+	MOVO  v, t;       \
+	PSLLL $n, t;      \
+	PSRLL $(32-n), v; \
+	PXOR  t, v
+
+#define ROTL_SSSE3(c, v) \
+	PSHUFB c, v
+
+#define ROUND_SSE2(v0, v1, v2, v3, m0, m1, m2, m3, t) \
+	PADDL  m0, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSE2(16, t, v3); \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(20, t, v1); \
+	PADDL  m1, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSE2(24, t, v3); \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(25, t, v1); \
+	PSHUFL $0x39, v1, v1; \
+	PSHUFL $0x4E, v2, v2; \
+	PSHUFL $0x93, v3, v3; \
+	PADDL  m2, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSE2(16, t, v3); \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(20, t, v1); \
+	PADDL  m3, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSE2(24, t, v3); \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(25, t, v1); \
+	PSHUFL $0x39, v3, v3; \
+	PSHUFL $0x4E, v2, v2; \
+	PSHUFL $0x93, v1, v1
+
+#define ROUND_SSSE3(v0, v1, v2, v3, m0, m1, m2, m3, t, c16, c8) \
+	PADDL  m0, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSSE3(c16, v3);  \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(20, t, v1); \
+	PADDL  m1, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSSE3(c8, v3);   \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(25, t, v1); \
+	PSHUFL $0x39, v1, v1; \
+	PSHUFL $0x4E, v2, v2; \
+	PSHUFL $0x93, v3, v3; \
+	PADDL  m2, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSSE3(c16, v3);  \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(20, t, v1); \
+	PADDL  m3, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSSE3(c8, v3);   \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(25, t, v1); \
+	PSHUFL $0x39, v3, v3; \
+	PSHUFL $0x4E, v2, v2; \
+	PSHUFL $0x93, v1, v1
+
+#define PRECOMPUTE(dst, off, src, t) \
+	MOVL 0*4(src), t;          \
+	MOVL t, 0*4+off+0(dst);    \
+	MOVL t, 9*4+off+64(dst);   \
+	MOVL t, 5*4+off+128(dst);  \
+	MOVL t, 14*4+off+192(dst); \
+	MOVL t, 4*4+off+256(dst);  \
+	MOVL t, 2*4+off+320(dst);  \
+	MOVL t, 8*4+off+384(dst);  \
+	MOVL t, 12*4+off+448(dst); \
+	MOVL t, 3*4+off+512(dst);  \
+	MOVL t, 15*4+off+576(dst); \
+	MOVL 1*4(src), t;          \
+	MOVL t, 4*4+off+0(dst);    \
+	MOVL t, 8*4+off+64(dst);   \
+	MOVL t, 14*4+off+128(dst); \
+	MOVL t, 5*4+off+192(dst);  \
+	MOVL t, 12*4+off+256(dst); \
+	MOVL t, 11*4+off+320(dst); \
+	MOVL t, 1*4+off+384(dst);  \
+	MOVL t, 6*4+off+448(dst);  \
+	MOVL t, 10*4+off+512(dst); \
+	MOVL t, 3*4+off+576(dst);  \
+	MOVL 2*4(src), t;          \
+	MOVL t, 1*4+off+0(dst);    \
+	MOVL t, 13*4+off+64(dst);  \
+	MOVL t, 6*4+off+128(dst);  \
+	MOVL t, 8*4+off+192(dst);  \
+	MOVL t, 2*4+off+256(dst);  \
+	MOVL t, 0*4+off+320(dst);  \
+	MOVL t, 14*4+off+384(dst); \
+	MOVL t, 11*4+off+448(dst); \
+	MOVL t, 12*4+off+512(dst); \
+	MOVL t, 4*4+off+576(dst);  \
+	MOVL 3*4(src), t;          \
+	MOVL t, 5*4+off+0(dst);    \
+	MOVL t, 15*4+off+64(dst);  \
+	MOVL t, 9*4+off+128(dst);  \
+	MOVL t, 1*4+off+192(dst);  \
+	MOVL t, 11*4+off+256(dst); \
+	MOVL t, 7*4+off+320(dst);  \
+	MOVL t, 13*4+off+384(dst); \
+	MOVL t, 3*4+off+448(dst);  \
+	MOVL t, 6*4+off+512(dst);  \
+	MOVL t, 10*4+off+576(dst); \
+	MOVL 4*4(src), t;          \
+	MOVL t, 2*4+off+0(dst);    \
+	MOVL t, 1*4+off+64(dst);   \
+	MOVL t, 15*4+off+128(dst); \
+	MOVL t, 10*4+off+192(dst); \
+	MOVL t, 6*4+off+256(dst);  \
+	MOVL t, 8*4+off+320(dst);  \
+	MOVL t, 3*4+off+384(dst);  \
+	MOVL t, 13*4+off+448(dst); \
+	MOVL t, 14*4+off+512(dst); \
+	MOVL t, 5*4+off+576(dst);  \
+	MOVL 5*4(src), t;          \
+	MOVL t, 6*4+off+0(dst);    \
+	MOVL t, 11*4+off+64(dst);  \
+	MOVL t, 2*4+off+128(dst);  \
+	MOVL t, 9*4+off+192(dst);  \
+	MOVL t, 1*4+off+256(dst);  \
+	MOVL t, 13*4+off+320(dst); \
+	MOVL t, 4*4+off+384(dst);  \
+	MOVL t, 8*4+off+448(dst);  \
+	MOVL t, 15*4+off+512(dst); \
+	MOVL t, 7*4+off+576(dst);  \
+	MOVL 6*4(src), t;          \
+	MOVL t, 3*4+off+0(dst);    \
+	MOVL t, 7*4+off+64(dst);   \
+	MOVL t, 13*4+off+128(dst); \
+	MOVL t, 12*4+off+192(dst); \
+	MOVL t, 10*4+off+256(dst); \
+	MOVL t, 1*4+off+320(dst);  \
+	MOVL t, 9*4+off+384(dst);  \
+	MOVL t, 14*4+off+448(dst); \
+	MOVL t, 0*4+off+512(dst);  \
+	MOVL t, 6*4+off+576(dst);  \
+	MOVL 7*4(src), t;          \
+	MOVL t, 7*4+off+0(dst);    \
+	MOVL t, 14*4+off+64(dst);  \
+	MOVL t, 10*4+off+128(dst); \
+	MOVL t, 0*4+off+192(dst);  \
+	MOVL t, 5*4+off+256(dst);  \
+	MOVL t, 9*4+off+320(dst);  \
+	MOVL t, 12*4+off+384(dst); \
+	MOVL t, 1*4+off+448(dst);  \
+	MOVL t, 13*4+off+512(dst); \
+	MOVL t, 2*4+off+576(dst);  \
+	MOVL 8*4(src), t;          \
+	MOVL t, 8*4+off+0(dst);    \
+	MOVL t, 5*4+off+64(dst);   \
+	MOVL t, 4*4+off+128(dst);  \
+	MOVL t, 15*4+off+192(dst); \
+	MOVL t, 14*4+off+256(dst); \
+	MOVL t, 3*4+off+320(dst);  \
+	MOVL t, 11*4+off+384(dst); \
+	MOVL t, 10*4+off+448(dst); \
+	MOVL t, 7*4+off+512(dst);  \
+	MOVL t, 1*4+off+576(dst);  \
+	MOVL 9*4(src), t;          \
+	MOVL t, 12*4+off+0(dst);   \
+	MOVL t, 2*4+off+64(dst);   \
+	MOVL t, 11*4+off+128(dst); \
+	MOVL t, 4*4+off+192(dst);  \
+	MOVL t, 0*4+off+256(dst);  \
+	MOVL t, 15*4+off+320(dst); \
+	MOVL t, 10*4+off+384(dst); \
+	MOVL t, 7*4+off+448(dst);  \
+	MOVL t, 5*4+off+512(dst);  \
+	MOVL t, 9*4+off+576(dst);  \
+	MOVL 10*4(src), t;         \
+	MOVL t, 9*4+off+0(dst);    \
+	MOVL t, 4*4+off+64(dst);   \
+	MOVL t, 8*4+off+128(dst);  \
+	MOVL t, 13*4+off+192(dst); \
+	MOVL t, 3*4+off+256(dst);  \
+	MOVL t, 5*4+off+320(dst);  \
+	MOVL t, 7*4+off+384(dst);  \
+	MOVL t, 15*4+off+448(dst); \
+	MOVL t, 11*4+off+512(dst); \
+	MOVL t, 0*4+off+576(dst);  \
+	MOVL 11*4(src), t;         \
+	MOVL t, 13*4+off+0(dst);   \
+	MOVL t, 10*4+off+64(dst);  \
+	MOVL t, 0*4+off+128(dst);  \
+	MOVL t, 3*4+off+192(dst);  \
+	MOVL t, 9*4+off+256(dst);  \
+	MOVL t, 6*4+off+320(dst);  \
+	MOVL t, 15*4+off+384(dst); \
+	MOVL t, 4*4+off+448(dst);  \
+	MOVL t, 2*4+off+512(dst);  \
+	MOVL t, 12*4+off+576(dst); \
+	MOVL 12*4(src), t;         \
+	MOVL t, 10*4+off+0(dst);   \
+	MOVL t, 12*4+off+64(dst);  \
+	MOVL t, 1*4+off+128(dst);  \
+	MOVL t, 6*4+off+192(dst);  \
+	MOVL t, 13*4+off+256(dst); \
+	MOVL t, 4*4+off+320(dst);  \
+	MOVL t, 0*4+off+384(dst);  \
+	MOVL t, 2*4+off+448(dst);  \
+	MOVL t, 8*4+off+512(dst);  \
+	MOVL t, 14*4+off+576(dst); \
+	MOVL 13*4(src), t;         \
+	MOVL t, 14*4+off+0(dst);   \
+	MOVL t, 3*4+off+64(dst);   \
+	MOVL t, 7*4+off+128(dst);  \
+	MOVL t, 2*4+off+192(dst);  \
+	MOVL t, 15*4+off+256(dst); \
+	MOVL t, 12*4+off+320(dst); \
+	MOVL t, 6*4+off+384(dst);  \
+	MOVL t, 0*4+off+448(dst);  \
+	MOVL t, 9*4+off+512(dst);  \
+	MOVL t, 11*4+off+576(dst); \
+	MOVL 14*4(src), t;         \
+	MOVL t, 11*4+off+0(dst);   \
+	MOVL t, 0*4+off+64(dst);   \
+	MOVL t, 12*4+off+128(dst); \
+	MOVL t, 7*4+off+192(dst);  \
+	MOVL t, 8*4+off+256(dst);  \
+	MOVL t, 14*4+off+320(dst); \
+	MOVL t, 2*4+off+384(dst);  \
+	MOVL t, 5*4+off+448(dst);  \
+	MOVL t, 1*4+off+512(dst);  \
+	MOVL t, 13*4+off+576(dst); \
+	MOVL 15*4(src), t;         \
+	MOVL t, 15*4+off+0(dst);   \
+	MOVL t, 6*4+off+64(dst);   \
+	MOVL t, 3*4+off+128(dst);  \
+	MOVL t, 11*4+off+192(dst); \
+	MOVL t, 7*4+off+256(dst);  \
+	MOVL t, 10*4+off+320(dst); \
+	MOVL t, 5*4+off+384(dst);  \
+	MOVL t, 9*4+off+448(dst);  \
+	MOVL t, 4*4+off+512(dst);  \
+	MOVL t, 8*4+off+576(dst)
+
+// func hashBlocksSSE2(h *[8]uint32, c *[2]uint32, flag uint32, blocks []byte)
+TEXT ·hashBlocksSSE2(SB), 0, $672-24 // frame = 656 + 16 byte alignment
+	MOVL h+0(FP), AX
+	MOVL c+4(FP), BX
+	MOVL flag+8(FP), CX
+	MOVL blocks_base+12(FP), SI
+	MOVL blocks_len+16(FP), DX
+
+	MOVL SP, DI
+	ADDL $15, DI
+	ANDL $~15, DI
+
+	MOVL CX, 8(DI)
+	MOVL 0(BX), CX
+	MOVL CX, 0(DI)
+	MOVL 4(BX), CX
+	MOVL CX, 4(DI)
+	XORL CX, CX
+	MOVL CX, 12(DI)
+
+	MOVOU 0(AX), X0
+	MOVOU 16(AX), X1
+	MOVOU counter<>(SB), X2
+
+loop:
+	MOVO  X0, X4
+	MOVO  X1, X5
+	MOVOU iv0<>(SB), X6
+	MOVOU iv1<>(SB), X7
+
+	MOVO  0(DI), X3
+	PADDQ X2, X3
+	PXOR  X3, X7
+	MOVO  X3, 0(DI)
+
+	PRECOMPUTE(DI, 16, SI, CX)
+	ROUND_SSE2(X4, X5, X6, X7, 16(DI), 32(DI), 48(DI), 64(DI), X3)
+	ROUND_SSE2(X4, X5, X6, X7, 16+64(DI), 32+64(DI), 48+64(DI), 64+64(DI), X3)
+	ROUND_SSE2(X4, X5, X6, X7, 16+128(DI), 32+128(DI), 48+128(DI), 64+128(DI), X3)
+	ROUND_SSE2(X4, X5, X6, X7, 16+192(DI), 32+192(DI), 48+192(DI), 64+192(DI), X3)
+	ROUND_SSE2(X4, X5, X6, X7, 16+256(DI), 32+256(DI), 48+256(DI), 64+256(DI), X3)
+	ROUND_SSE2(X4, X5, X6, X7, 16+320(DI), 32+320(DI), 48+320(DI), 64+320(DI), X3)
+	ROUND_SSE2(X4, X5, X6, X7, 16+384(DI), 32+384(DI), 48+384(DI), 64+384(DI), X3)
+	ROUND_SSE2(X4, X5, X6, X7, 16+448(DI), 32+448(DI), 48+448(DI), 64+448(DI), X3)
+	ROUND_SSE2(X4, X5, X6, X7, 16+512(DI), 32+512(DI), 48+512(DI), 64+512(DI), X3)
+	ROUND_SSE2(X4, X5, X6, X7, 16+576(DI), 32+576(DI), 48+576(DI), 64+576(DI), X3)
+
+	PXOR X4, X0
+	PXOR X5, X1
+	PXOR X6, X0
+	PXOR X7, X1
+
+	LEAL 64(SI), SI
+	SUBL $64, DX
+	JNE  loop
+
+	MOVL 0(DI), CX
+	MOVL CX, 0(BX)
+	MOVL 4(DI), CX
+	MOVL CX, 4(BX)
+
+	MOVOU X0, 0(AX)
+	MOVOU X1, 16(AX)
+
+	RET
+
+// func hashBlocksSSSE3(h *[8]uint32, c *[2]uint32, flag uint32, blocks []byte)
+TEXT ·hashBlocksSSSE3(SB), 0, $704-24 // frame = 688 + 16 byte alignment
+	MOVL h+0(FP), AX
+	MOVL c+4(FP), BX
+	MOVL flag+8(FP), CX
+	MOVL blocks_base+12(FP), SI
+	MOVL blocks_len+16(FP), DX
+
+	MOVL SP, DI
+	ADDL $15, DI
+	ANDL $~15, DI
+
+	MOVL CX, 8(DI)
+	MOVL 0(BX), CX
+	MOVL CX, 0(DI)
+	MOVL 4(BX), CX
+	MOVL CX, 4(DI)
+	XORL CX, CX
+	MOVL CX, 12(DI)
+
+	MOVOU 0(AX), X0
+	MOVOU 16(AX), X1
+	MOVOU counter<>(SB), X2
+
+loop:
+	MOVO  X0, 656(DI)
+	MOVO  X1, 672(DI)
+	MOVO  X0, X4
+	MOVO  X1, X5
+	MOVOU iv0<>(SB), X6
+	MOVOU iv1<>(SB), X7
+
+	MOVO  0(DI), X3
+	PADDQ X2, X3
+	PXOR  X3, X7
+	MOVO  X3, 0(DI)
+
+	MOVOU rol16<>(SB), X0
+	MOVOU rol8<>(SB), X1
+
+	PRECOMPUTE(DI, 16, SI, CX)
+	ROUND_SSSE3(X4, X5, X6, X7, 16(DI), 32(DI), 48(DI), 64(DI), X3, X0, X1)
+	ROUND_SSSE3(X4, X5, X6, X7, 16+64(DI), 32+64(DI), 48+64(DI), 64+64(DI), X3, X0, X1)
+	ROUND_SSSE3(X4, X5, X6, X7, 16+128(DI), 32+128(DI), 48+128(DI), 64+128(DI), X3, X0, X1)
+	ROUND_SSSE3(X4, X5, X6, X7, 16+192(DI), 32+192(DI), 48+192(DI), 64+192(DI), X3, X0, X1)
+	ROUND_SSSE3(X4, X5, X6, X7, 16+256(DI), 32+256(DI), 48+256(DI), 64+256(DI), X3, X0, X1)
+	ROUND_SSSE3(X4, X5, X6, X7, 16+320(DI), 32+320(DI), 48+320(DI), 64+320(DI), X3, X0, X1)
+	ROUND_SSSE3(X4, X5, X6, X7, 16+384(DI), 32+384(DI), 48+384(DI), 64+384(DI), X3, X0, X1)
+	ROUND_SSSE3(X4, X5, X6, X7, 16+448(DI), 32+448(DI), 48+448(DI), 64+448(DI), X3, X0, X1)
+	ROUND_SSSE3(X4, X5, X6, X7, 16+512(DI), 32+512(DI), 48+512(DI), 64+512(DI), X3, X0, X1)
+	ROUND_SSSE3(X4, X5, X6, X7, 16+576(DI), 32+576(DI), 48+576(DI), 64+576(DI), X3, X0, X1)
+
+	MOVO 656(DI), X0
+	MOVO 672(DI), X1
+	PXOR X4, X0
+	PXOR X5, X1
+	PXOR X6, X0
+	PXOR X7, X1
+
+	LEAL 64(SI), SI
+	SUBL $64, DX
+	JNE  loop
+
+	MOVL 0(DI), CX
+	MOVL CX, 0(BX)
+	MOVL 4(DI), CX
+	MOVL CX, 4(BX)
+
+	MOVOU X0, 0(AX)
+	MOVOU X1, 16(AX)
+
+	RET

vendor/golang.org/x/crypto/blake2s/blake2s_amd64.go

@@ -0,0 +1,38 @@
+// 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.
+
+//go:build amd64 && gc && !purego
+// +build amd64,gc,!purego
+
+package blake2s
+
+import "golang.org/x/sys/cpu"
+
+var (
+	useSSE4  = cpu.X86.HasSSE41
+	useSSSE3 = cpu.X86.HasSSSE3
+	useSSE2  = cpu.X86.HasSSE2
+)
+
+//go:noescape
+func hashBlocksSSE2(h *[8]uint32, c *[2]uint32, flag uint32, blocks []byte)
+
+//go:noescape
+func hashBlocksSSSE3(h *[8]uint32, c *[2]uint32, flag uint32, blocks []byte)
+
+//go:noescape
+func hashBlocksSSE4(h *[8]uint32, c *[2]uint32, flag uint32, blocks []byte)
+
+func hashBlocks(h *[8]uint32, c *[2]uint32, flag uint32, blocks []byte) {
+	switch {
+	case useSSE4:
+		hashBlocksSSE4(h, c, flag, blocks)
+	case useSSSE3:
+		hashBlocksSSSE3(h, c, flag, blocks)
+	case useSSE2:
+		hashBlocksSSE2(h, c, flag, blocks)
+	default:
+		hashBlocksGeneric(h, c, flag, blocks)
+	}
+}

vendor/golang.org/x/crypto/blake2s/blake2s_amd64.s

@@ -0,0 +1,433 @@
+// 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.
+
+//go:build amd64 && gc && !purego
+// +build amd64,gc,!purego
+
+#include "textflag.h"
+
+DATA iv0<>+0x00(SB)/4, $0x6a09e667
+DATA iv0<>+0x04(SB)/4, $0xbb67ae85
+DATA iv0<>+0x08(SB)/4, $0x3c6ef372
+DATA iv0<>+0x0c(SB)/4, $0xa54ff53a
+GLOBL iv0<>(SB), (NOPTR+RODATA), $16
+
+DATA iv1<>+0x00(SB)/4, $0x510e527f
+DATA iv1<>+0x04(SB)/4, $0x9b05688c
+DATA iv1<>+0x08(SB)/4, $0x1f83d9ab
+DATA iv1<>+0x0c(SB)/4, $0x5be0cd19
+GLOBL iv1<>(SB), (NOPTR+RODATA), $16
+
+DATA rol16<>+0x00(SB)/8, $0x0504070601000302
+DATA rol16<>+0x08(SB)/8, $0x0D0C0F0E09080B0A
+GLOBL rol16<>(SB), (NOPTR+RODATA), $16
+
+DATA rol8<>+0x00(SB)/8, $0x0407060500030201
+DATA rol8<>+0x08(SB)/8, $0x0C0F0E0D080B0A09
+GLOBL rol8<>(SB), (NOPTR+RODATA), $16
+
+DATA counter<>+0x00(SB)/8, $0x40
+DATA counter<>+0x08(SB)/8, $0x0
+GLOBL counter<>(SB), (NOPTR+RODATA), $16
+
+#define ROTL_SSE2(n, t, v) \
+	MOVO  v, t;       \
+	PSLLL $n, t;      \
+	PSRLL $(32-n), v; \
+	PXOR  t, v
+
+#define ROTL_SSSE3(c, v) \
+	PSHUFB c, v
+
+#define ROUND_SSE2(v0, v1, v2, v3, m0, m1, m2, m3, t) \
+	PADDL  m0, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSE2(16, t, v3); \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(20, t, v1); \
+	PADDL  m1, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSE2(24, t, v3); \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(25, t, v1); \
+	PSHUFL $0x39, v1, v1; \
+	PSHUFL $0x4E, v2, v2; \
+	PSHUFL $0x93, v3, v3; \
+	PADDL  m2, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSE2(16, t, v3); \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(20, t, v1); \
+	PADDL  m3, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSE2(24, t, v3); \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(25, t, v1); \
+	PSHUFL $0x39, v3, v3; \
+	PSHUFL $0x4E, v2, v2; \
+	PSHUFL $0x93, v1, v1
+
+#define ROUND_SSSE3(v0, v1, v2, v3, m0, m1, m2, m3, t, c16, c8) \
+	PADDL  m0, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSSE3(c16, v3);  \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(20, t, v1); \
+	PADDL  m1, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSSE3(c8, v3);   \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(25, t, v1); \
+	PSHUFL $0x39, v1, v1; \
+	PSHUFL $0x4E, v2, v2; \
+	PSHUFL $0x93, v3, v3; \
+	PADDL  m2, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSSE3(c16, v3);  \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(20, t, v1); \
+	PADDL  m3, v0;        \
+	PADDL  v1, v0;        \
+	PXOR   v0, v3;        \
+	ROTL_SSSE3(c8, v3);   \
+	PADDL  v3, v2;        \
+	PXOR   v2, v1;        \
+	ROTL_SSE2(25, t, v1); \
+	PSHUFL $0x39, v3, v3; \
+	PSHUFL $0x4E, v2, v2; \
+	PSHUFL $0x93, v1, v1
+
+
+#define LOAD_MSG_SSE4(m0, m1, m2, m3, src, i0, i1, i2, i3, i4, i5, i6, i7, i8, i9, i10, i11, i12, i13, i14, i15) \
+	MOVL   i0*4(src), m0;      \
+	PINSRD $1, i1*4(src), m0;  \
+	PINSRD $2, i2*4(src), m0;  \
+	PINSRD $3, i3*4(src), m0;  \
+	MOVL   i4*4(src), m1;      \
+	PINSRD $1, i5*4(src), m1;  \
+	PINSRD $2, i6*4(src), m1;  \
+	PINSRD $3, i7*4(src), m1;  \
+	MOVL   i8*4(src), m2;      \
+	PINSRD $1, i9*4(src), m2;  \
+	PINSRD $2, i10*4(src), m2; \
+	PINSRD $3, i11*4(src), m2; \
+	MOVL   i12*4(src), m3;     \
+	PINSRD $1, i13*4(src), m3; \
+	PINSRD $2, i14*4(src), m3; \
+	PINSRD $3, i15*4(src), m3
+
+#define PRECOMPUTE_MSG(dst, off, src, R8, R9, R10, R11, R12, R13, R14, R15) \
+	MOVQ 0*4(src), R8;           \
+	MOVQ 2*4(src), R9;           \
+	MOVQ 4*4(src), R10;          \
+	MOVQ 6*4(src), R11;          \
+	MOVQ 8*4(src), R12;          \
+	MOVQ 10*4(src), R13;         \
+	MOVQ 12*4(src), R14;         \
+	MOVQ 14*4(src), R15;         \
+	                             \
+	MOVL R8, 0*4+off+0(dst);     \
+	MOVL R8, 9*4+off+64(dst);    \
+	MOVL R8, 5*4+off+128(dst);   \
+	MOVL R8, 14*4+off+192(dst);  \
+	MOVL R8, 4*4+off+256(dst);   \
+	MOVL R8, 2*4+off+320(dst);   \
+	MOVL R8, 8*4+off+384(dst);   \
+	MOVL R8, 12*4+off+448(dst);  \
+	MOVL R8, 3*4+off+512(dst);   \
+	MOVL R8, 15*4+off+576(dst);  \
+	SHRQ $32, R8;                \
+	MOVL R8, 4*4+off+0(dst);     \
+	MOVL R8, 8*4+off+64(dst);    \
+	MOVL R8, 14*4+off+128(dst);  \
+	MOVL R8, 5*4+off+192(dst);   \
+	MOVL R8, 12*4+off+256(dst);  \
+	MOVL R8, 11*4+off+320(dst);  \
+	MOVL R8, 1*4+off+384(dst);   \
+	MOVL R8, 6*4+off+448(dst);   \
+	MOVL R8, 10*4+off+512(dst);  \
+	MOVL R8, 3*4+off+576(dst);   \
+	                             \
+	MOVL R9, 1*4+off+0(dst);     \
+	MOVL R9, 13*4+off+64(dst);   \
+	MOVL R9, 6*4+off+128(dst);   \
+	MOVL R9, 8*4+off+192(dst);   \
+	MOVL R9, 2*4+off+256(dst);   \
+	MOVL R9, 0*4+off+320(dst);   \
+	MOVL R9, 14*4+off+384(dst);  \
+	MOVL R9, 11*4+off+448(dst);  \
+	MOVL R9, 12*4+off+512(dst);  \
+	MOVL R9, 4*4+off+576(dst);   \
+	SHRQ $32, R9;                \
+	MOVL R9, 5*4+off+0(dst);     \
+	MOVL R9, 15*4+off+64(dst);   \
+	MOVL R9, 9*4+off+128(dst);   \
+	MOVL R9, 1*4+off+192(dst);   \
+	MOVL R9, 11*4+off+256(dst);  \
+	MOVL R9, 7*4+off+320(dst);   \
+	MOVL R9, 13*4+off+384(dst);  \
+	MOVL R9, 3*4+off+448(dst);   \
+	MOVL R9, 6*4+off+512(dst);   \
+	MOVL R9, 10*4+off+576(dst);  \
+	                             \
+	MOVL R10, 2*4+off+0(dst);    \
+	MOVL R10, 1*4+off+64(dst);   \
+	MOVL R10, 15*4+off+128(dst); \
+	MOVL R10, 10*4+off+192(dst); \
+	MOVL R10, 6*4+off+256(dst);  \
+	MOVL R10, 8*4+off+320(dst);  \
+	MOVL R10, 3*4+off+384(dst);  \
+	MOVL R10, 13*4+off+448(dst); \
+	MOVL R10, 14*4+off+512(dst); \
+	MOVL R10, 5*4+off+576(dst);  \
+	SHRQ $32, R10;               \
+	MOVL R10, 6*4+off+0(dst);    \
+	MOVL R10, 11*4+off+64(dst);  \
+	MOVL R10, 2*4+off+128(dst);  \
+	MOVL R10, 9*4+off+192(dst);  \
+	MOVL R10, 1*4+off+256(dst);  \
+	MOVL R10, 13*4+off+320(dst); \
+	MOVL R10, 4*4+off+384(dst);  \
+	MOVL R10, 8*4+off+448(dst);  \
+	MOVL R10, 15*4+off+512(dst); \
+	MOVL R10, 7*4+off+576(dst);  \
+	                             \
+	MOVL R11, 3*4+off+0(dst);    \
+	MOVL R11, 7*4+off+64(dst);   \
+	MOVL R11, 13*4+off+128(dst); \
+	MOVL R11, 12*4+off+192(dst); \
+	MOVL R11, 10*4+off+256(dst); \
+	MOVL R11, 1*4+off+320(dst);  \
+	MOVL R11, 9*4+off+384(dst);  \
+	MOVL R11, 14*4+off+448(dst); \
+	MOVL R11, 0*4+off+512(dst);  \
+	MOVL R11, 6*4+off+576(dst);  \
+	SHRQ $32, R11;               \
+	MOVL R11, 7*4+off+0(dst);    \
+	MOVL R11, 14*4+off+64(dst);  \
+	MOVL R11, 10*4+off+128(dst); \
+	MOVL R11, 0*4+off+192(dst);  \
+	MOVL R11, 5*4+off+256(dst);  \
+	MOVL R11, 9*4+off+320(dst);  \
+	MOVL R11, 12*4+off+384(dst); \
+	MOVL R11, 1*4+off+448(dst);  \
+	MOVL R11, 13*4+off+512(dst); \
+	MOVL R11, 2*4+off+576(dst);  \
+	                             \
+	MOVL R12, 8*4+off+0(dst);    \
+	MOVL R12, 5*4+off+64(dst);   \
+	MOVL R12, 4*4+off+128(dst);  \
+	MOVL R12, 15*4+off+192(dst); \
+	MOVL R12, 14*4+off+256(dst); \
+	MOVL R12, 3*4+off+320(dst);  \
+	MOVL R12, 11*4+off+384(dst); \
+	MOVL R12, 10*4+off+448(dst); \
+	MOVL R12, 7*4+off+512(dst);  \
+	MOVL R12, 1*4+off+576(dst);  \
+	SHRQ $32, R12;               \
+	MOVL R12, 12*4+off+0(dst);   \
+	MOVL R12, 2*4+off+64(dst);   \
+	MOVL R12, 11*4+off+128(dst); \
+	MOVL R12, 4*4+off+192(dst);  \
+	MOVL R12, 0*4+off+256(dst);  \
+	MOVL R12, 15*4+off+320(dst); \
+	MOVL R12, 10*4+off+384(dst); \
+	MOVL R12, 7*4+off+448(dst);  \
+	MOVL R12, 5*4+off+512(dst);  \
+	MOVL R12, 9*4+off+576(dst);  \
+	                             \
+	MOVL R13, 9*4+off+0(dst);    \
+	MOVL R13, 4*4+off+64(dst);   \
+	MOVL R13, 8*4+off+128(dst);  \
+	MOVL R13, 13*4+off+192(dst); \
+	MOVL R13, 3*4+off+256(dst);  \
+	MOVL R13, 5*4+off+320(dst);  \
+	MOVL R13, 7*4+off+384(dst);  \
+	MOVL R13, 15*4+off+448(dst); \
+	MOVL R13, 11*4+off+512(dst); \
+	MOVL R13, 0*4+off+576(dst);  \
+	SHRQ $32, R13;               \
+	MOVL R13, 13*4+off+0(dst);   \
+	MOVL R13, 10*4+off+64(dst);  \
+	MOVL R13, 0*4+off+128(dst);  \
+	MOVL R13, 3*4+off+192(dst);  \
+	MOVL R13, 9*4+off+256(dst);  \
+	MOVL R13, 6*4+off+320(dst);  \
+	MOVL R13, 15*4+off+384(dst); \
+	MOVL R13, 4*4+off+448(dst);  \
+	MOVL R13, 2*4+off+512(dst);  \
+	MOVL R13, 12*4+off+576(dst); \
+	                             \
+	MOVL R14, 10*4+off+0(dst);   \
+	MOVL R14, 12*4+off+64(dst);  \
+	MOVL R14, 1*4+off+128(dst);  \
+	MOVL R14, 6*4+off+192(dst);  \
+	MOVL R14, 13*4+off+256(dst); \
+	MOVL R14, 4*4+off+320(dst);  \
+	MOVL R14, 0*4+off+384(dst);  \
+	MOVL R14, 2*4+off+448(dst);  \
+	MOVL R14, 8*4+off+512(dst);  \
+	MOVL R14, 14*4+off+576(dst); \
+	SHRQ $32, R14;               \
+	MOVL R14, 14*4+off+0(dst);   \
+	MOVL R14, 3*4+off+64(dst);   \
+	MOVL R14, 7*4+off+128(dst);  \
+	MOVL R14, 2*4+off+192(dst);  \
+	MOVL R14, 15*4+off+256(dst); \
+	MOVL R14, 12*4+off+320(dst); \
+	MOVL R14, 6*4+off+384(dst);  \
+	MOVL R14, 0*4+off+448(dst);  \
+	MOVL R14, 9*4+off+512(dst);  \
+	MOVL R14, 11*4+off+576(dst); \
+	                             \
+	MOVL R15, 11*4+off+0(dst);   \
+	MOVL R15, 0*4+off+64(dst);   \
+	MOVL R15, 12*4+off+128(dst); \
+	MOVL R15, 7*4+off+192(dst);  \
+	MOVL R15, 8*4+off+256(dst);  \
+	MOVL R15, 14*4+off+320(dst); \
+	MOVL R15, 2*4+off+384(dst);  \
+	MOVL R15, 5*4+off+448(dst);  \
+	MOVL R15, 1*4+off+512(dst);  \
+	MOVL R15, 13*4+off+576(dst); \
+	SHRQ $32, R15;               \
+	MOVL R15, 15*4+off+0(dst);   \
+	MOVL R15, 6*4+off+64(dst);   \
+	MOVL R15, 3*4+off+128(dst);  \
+	MOVL R15, 11*4+off+192(dst); \
+	MOVL R15, 7*4+off+256(dst);  \
+	MOVL R15, 10*4+off+320(dst); \
+	MOVL R15, 5*4+off+384(dst);  \
+	MOVL R15, 9*4+off+448(dst);  \
+	MOVL R15, 4*4+off+512(dst);  \
+	MOVL R15, 8*4+off+576(dst)
+
+#define BLAKE2s_SSE2() \
+	PRECOMPUTE_MSG(BP, 16, SI, R8, R9, R10, R11, R12, R13, R14, R15);               \
+	ROUND_SSE2(X4, X5, X6, X7, 16(BP), 32(BP), 48(BP), 64(BP), X8);                 \
+	ROUND_SSE2(X4, X5, X6, X7, 16+64(BP), 32+64(BP), 48+64(BP), 64+64(BP), X8);     \
+	ROUND_SSE2(X4, X5, X6, X7, 16+128(BP), 32+128(BP), 48+128(BP), 64+128(BP), X8); \
+	ROUND_SSE2(X4, X5, X6, X7, 16+192(BP), 32+192(BP), 48+192(BP), 64+192(BP), X8); \
+	ROUND_SSE2(X4, X5, X6, X7, 16+256(BP), 32+256(BP), 48+256(BP), 64+256(BP), X8); \
+	ROUND_SSE2(X4, X5, X6, X7, 16+320(BP), 32+320(BP), 48+320(BP), 64+320(BP), X8); \
+	ROUND_SSE2(X4, X5, X6, X7, 16+384(BP), 32+384(BP), 48+384(BP), 64+384(BP), X8); \
+	ROUND_SSE2(X4, X5, X6, X7, 16+448(BP), 32+448(BP), 48+448(BP), 64+448(BP), X8); \
+	ROUND_SSE2(X4, X5, X6, X7, 16+512(BP), 32+512(BP), 48+512(BP), 64+512(BP), X8); \
+	ROUND_SSE2(X4, X5, X6, X7, 16+576(BP), 32+576(BP), 48+576(BP), 64+576(BP), X8)
+
+#define BLAKE2s_SSSE3() \
+	PRECOMPUTE_MSG(BP, 16, SI, R8, R9, R10, R11, R12, R13, R14, R15);                          \
+	ROUND_SSSE3(X4, X5, X6, X7, 16(BP), 32(BP), 48(BP), 64(BP), X8, X13, X14);                 \
+	ROUND_SSSE3(X4, X5, X6, X7, 16+64(BP), 32+64(BP), 48+64(BP), 64+64(BP), X8, X13, X14);     \
+	ROUND_SSSE3(X4, X5, X6, X7, 16+128(BP), 32+128(BP), 48+128(BP), 64+128(BP), X8, X13, X14); \
+	ROUND_SSSE3(X4, X5, X6, X7, 16+192(BP), 32+192(BP), 48+192(BP), 64+192(BP), X8, X13, X14); \
+	ROUND_SSSE3(X4, X5, X6, X7, 16+256(BP), 32+256(BP), 48+256(BP), 64+256(BP), X8, X13, X14); \
+	ROUND_SSSE3(X4, X5, X6, X7, 16+320(BP), 32+320(BP), 48+320(BP), 64+320(BP), X8, X13, X14); \
+	ROUND_SSSE3(X4, X5, X6, X7, 16+384(BP), 32+384(BP), 48+384(BP), 64+384(BP), X8, X13, X14); \
+	ROUND_SSSE3(X4, X5, X6, X7, 16+448(BP), 32+448(BP), 48+448(BP), 64+448(BP), X8, X13, X14); \
+	ROUND_SSSE3(X4, X5, X6, X7, 16+512(BP), 32+512(BP), 48+512(BP), 64+512(BP), X8, X13, X14); \
+	ROUND_SSSE3(X4, X5, X6, X7, 16+576(BP), 32+576(BP), 48+576(BP), 64+576(BP), X8, X13, X14)
+
+#define BLAKE2s_SSE4() \
+	LOAD_MSG_SSE4(X8, X9, X10, X11, SI, 0, 2, 4, 6, 1, 3, 5, 7, 8, 10, 12, 14, 9, 11, 13, 15); \
+	ROUND_SSSE3(X4, X5, X6, X7, X8, X9, X10, X11, X8, X13, X14);                               \
+	LOAD_MSG_SSE4(X8, X9, X10, X11, SI, 14, 4, 9, 13, 10, 8, 15, 6, 1, 0, 11, 5, 12, 2, 7, 3); \
+	ROUND_SSSE3(X4, X5, X6, X7, X8, X9, X10, X11, X8, X13, X14);                               \
+	LOAD_MSG_SSE4(X8, X9, X10, X11, SI, 11, 12, 5, 15, 8, 0, 2, 13, 10, 3, 7, 9, 14, 6, 1, 4); \
+	ROUND_SSSE3(X4, X5, X6, X7, X8, X9, X10, X11, X8, X13, X14);                               \
+	LOAD_MSG_SSE4(X8, X9, X10, X11, SI, 7, 3, 13, 11, 9, 1, 12, 14, 2, 5, 4, 15, 6, 10, 0, 8); \
+	ROUND_SSSE3(X4, X5, X6, X7, X8, X9, X10, X11, X8, X13, X14);                               \
+	LOAD_MSG_SSE4(X8, X9, X10, X11, SI, 9, 5, 2, 10, 0, 7, 4, 15, 14, 11, 6, 3, 1, 12, 8, 13); \
+	ROUND_SSSE3(X4, X5, X6, X7, X8, X9, X10, X11, X8, X13, X14);                               \
+	LOAD_MSG_SSE4(X8, X9, X10, X11, SI, 2, 6, 0, 8, 12, 10, 11, 3, 4, 7, 15, 1, 13, 5, 14, 9); \
+	ROUND_SSSE3(X4, X5, X6, X7, X8, X9, X10, X11, X8, X13, X14);                               \
+	LOAD_MSG_SSE4(X8, X9, X10, X11, SI, 12, 1, 14, 4, 5, 15, 13, 10, 0, 6, 9, 8, 7, 3, 2, 11); \
+	ROUND_SSSE3(X4, X5, X6, X7, X8, X9, X10, X11, X8, X13, X14);                               \
+	LOAD_MSG_SSE4(X8, X9, X10, X11, SI, 13, 7, 12, 3, 11, 14, 1, 9, 5, 15, 8, 2, 0, 4, 6, 10); \
+	ROUND_SSSE3(X4, X5, X6, X7, X8, X9, X10, X11, X8, X13, X14);                               \
+	LOAD_MSG_SSE4(X8, X9, X10, X11, SI, 6, 14, 11, 0, 15, 9, 3, 8, 12, 13, 1, 10, 2, 7, 4, 5); \
+	ROUND_SSSE3(X4, X5, X6, X7, X8, X9, X10, X11, X8, X13, X14);                               \
+	LOAD_MSG_SSE4(X8, X9, X10, X11, SI, 10, 8, 7, 1, 2, 4, 6, 5, 15, 9, 3, 13, 11, 14, 12, 0); \
+	ROUND_SSSE3(X4, X5, X6, X7, X8, X9, X10, X11, X8, X13, X14)
+
+#define HASH_BLOCKS(h, c, flag, blocks_base, blocks_len, BLAKE2s_FUNC) \
+	MOVQ  h, AX;                   \
+	MOVQ  c, BX;                   \
+	MOVL  flag, CX;                \
+	MOVQ  blocks_base, SI;         \
+	MOVQ  blocks_len, DX;          \
+	                               \
+	MOVQ  SP, BP;                  \
+	ADDQ  $15, BP;                 \
+	ANDQ  $~15, BP;                \
+	                               \
+	MOVQ  0(BX), R9;               \
+	MOVQ  R9, 0(BP);               \
+	MOVQ  CX, 8(BP);               \
+	                               \
+	MOVOU 0(AX), X0;               \
+	MOVOU 16(AX), X1;              \
+	MOVOU iv0<>(SB), X2;           \
+	MOVOU iv1<>(SB), X3            \
+	                               \
+	MOVOU counter<>(SB), X12;      \
+	MOVOU rol16<>(SB), X13;        \
+	MOVOU rol8<>(SB), X14;         \
+	MOVO  0(BP), X15;              \
+	                               \
+	loop:                          \
+	MOVO  X0, X4;                  \
+	MOVO  X1, X5;                  \
+	MOVO  X2, X6;                  \
+	MOVO  X3, X7;                  \
+	                               \
+	PADDQ X12, X15;                \
+	PXOR  X15, X7;                 \
+	                               \
+	BLAKE2s_FUNC();                \
+	                               \
+	PXOR  X4, X0;                  \
+	PXOR  X5, X1;                  \
+	PXOR  X6, X0;                  \
+	PXOR  X7, X1;                  \
+	                               \
+	LEAQ  64(SI), SI;              \
+	SUBQ  $64, DX;                 \
+	JNE   loop;                    \
+	                               \
+	MOVO  X15, 0(BP);              \
+	MOVQ  0(BP), R9;               \
+	MOVQ  R9, 0(BX);               \
+	                               \
+	MOVOU X0, 0(AX);               \
+	MOVOU X1, 16(AX)
+
+// func hashBlocksSSE2(h *[8]uint32, c *[2]uint32, flag uint32, blocks []byte)
+TEXT ·hashBlocksSSE2(SB), 0, $672-48 // frame = 656 + 16 byte alignment
+	HASH_BLOCKS(h+0(FP), c+8(FP), flag+16(FP), blocks_base+24(FP), blocks_len+32(FP), BLAKE2s_SSE2)
+	RET
+
+// func hashBlocksSSSE3(h *[8]uint32, c *[2]uint32, flag uint32, blocks []byte)
+TEXT ·hashBlocksSSSE3(SB), 0, $672-48 // frame = 656 + 16 byte alignment
+	HASH_BLOCKS(h+0(FP), c+8(FP), flag+16(FP), blocks_base+24(FP), blocks_len+32(FP), BLAKE2s_SSSE3)
+	RET
+
+// func hashBlocksSSE4(h *[8]uint32, c *[2]uint32, flag uint32, blocks []byte)
+TEXT ·hashBlocksSSE4(SB), 0, $32-48 // frame = 16 + 16 byte alignment
+	HASH_BLOCKS(h+0(FP), c+8(FP), flag+16(FP), blocks_base+24(FP), blocks_len+32(FP), BLAKE2s_SSE4)
+	RET

vendor/golang.org/x/crypto/blake2s/blake2s_generic.go

@@ -0,0 +1,178 @@
+// 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 blake2s
+
+import (
+	"math/bits"
+)
+
+// the precomputed values for BLAKE2s
+// there are 10 16-byte arrays - one for each round
+// the entries are calculated from the sigma constants.
+var precomputed = [10][16]byte{
+	{0, 2, 4, 6, 1, 3, 5, 7, 8, 10, 12, 14, 9, 11, 13, 15},
+	{14, 4, 9, 13, 10, 8, 15, 6, 1, 0, 11, 5, 12, 2, 7, 3},
+	{11, 12, 5, 15, 8, 0, 2, 13, 10, 3, 7, 9, 14, 6, 1, 4},
+	{7, 3, 13, 11, 9, 1, 12, 14, 2, 5, 4, 15, 6, 10, 0, 8},
+	{9, 5, 2, 10, 0, 7, 4, 15, 14, 11, 6, 3, 1, 12, 8, 13},
+	{2, 6, 0, 8, 12, 10, 11, 3, 4, 7, 15, 1, 13, 5, 14, 9},
+	{12, 1, 14, 4, 5, 15, 13, 10, 0, 6, 9, 8, 7, 3, 2, 11},
+	{13, 7, 12, 3, 11, 14, 1, 9, 5, 15, 8, 2, 0, 4, 6, 10},
+	{6, 14, 11, 0, 15, 9, 3, 8, 12, 13, 1, 10, 2, 7, 4, 5},
+	{10, 8, 7, 1, 2, 4, 6, 5, 15, 9, 3, 13, 11, 14, 12, 0},
+}
+
+func hashBlocksGeneric(h *[8]uint32, c *[2]uint32, flag uint32, blocks []byte) {
+	var m [16]uint32
+	c0, c1 := c[0], c[1]
+
+	for i := 0; i < len(blocks); {
+		c0 += BlockSize
+		if c0 < BlockSize {
+			c1++
+		}
+
+		v0, v1, v2, v3, v4, v5, v6, v7 := h[0], h[1], h[2], h[3], h[4], h[5], h[6], h[7]
+		v8, v9, v10, v11, v12, v13, v14, v15 := iv[0], iv[1], iv[2], iv[3], iv[4], iv[5], iv[6], iv[7]
+		v12 ^= c0
+		v13 ^= c1
+		v14 ^= flag
+
+		for j := range m {
+			m[j] = uint32(blocks[i]) | uint32(blocks[i+1])<<8 | uint32(blocks[i+2])<<16 | uint32(blocks[i+3])<<24
+			i += 4
+		}
+
+		for k := range precomputed {
+			s := &(precomputed[k])
+
+			v0 += m[s[0]]
+			v0 += v4
+			v12 ^= v0
+			v12 = bits.RotateLeft32(v12, -16)
+			v8 += v12
+			v4 ^= v8
+			v4 = bits.RotateLeft32(v4, -12)
+			v1 += m[s[1]]
+			v1 += v5
+			v13 ^= v1
+			v13 = bits.RotateLeft32(v13, -16)
+			v9 += v13
+			v5 ^= v9
+			v5 = bits.RotateLeft32(v5, -12)
+			v2 += m[s[2]]
+			v2 += v6
+			v14 ^= v2
+			v14 = bits.RotateLeft32(v14, -16)
+			v10 += v14
+			v6 ^= v10
+			v6 = bits.RotateLeft32(v6, -12)
+			v3 += m[s[3]]
+			v3 += v7
+			v15 ^= v3
+			v15 = bits.RotateLeft32(v15, -16)
+			v11 += v15
+			v7 ^= v11
+			v7 = bits.RotateLeft32(v7, -12)
+
+			v0 += m[s[4]]
+			v0 += v4
+			v12 ^= v0
+			v12 = bits.RotateLeft32(v12, -8)
+			v8 += v12
+			v4 ^= v8
+			v4 = bits.RotateLeft32(v4, -7)
+			v1 += m[s[5]]
+			v1 += v5
+			v13 ^= v1
+			v13 = bits.RotateLeft32(v13, -8)
+			v9 += v13
+			v5 ^= v9
+			v5 = bits.RotateLeft32(v5, -7)
+			v2 += m[s[6]]
+			v2 += v6
+			v14 ^= v2
+			v14 = bits.RotateLeft32(v14, -8)
+			v10 += v14
+			v6 ^= v10
+			v6 = bits.RotateLeft32(v6, -7)
+			v3 += m[s[7]]
+			v3 += v7
+			v15 ^= v3
+			v15 = bits.RotateLeft32(v15, -8)
+			v11 += v15
+			v7 ^= v11
+			v7 = bits.RotateLeft32(v7, -7)
+
+			v0 += m[s[8]]
+			v0 += v5
+			v15 ^= v0
+			v15 = bits.RotateLeft32(v15, -16)
+			v10 += v15
+			v5 ^= v10
+			v5 = bits.RotateLeft32(v5, -12)
+			v1 += m[s[9]]
+			v1 += v6
+			v12 ^= v1
+			v12 = bits.RotateLeft32(v12, -16)
+			v11 += v12
+			v6 ^= v11
+			v6 = bits.RotateLeft32(v6, -12)
+			v2 += m[s[10]]
+			v2 += v7
+			v13 ^= v2
+			v13 = bits.RotateLeft32(v13, -16)
+			v8 += v13
+			v7 ^= v8
+			v7 = bits.RotateLeft32(v7, -12)
+			v3 += m[s[11]]
+			v3 += v4
+			v14 ^= v3
+			v14 = bits.RotateLeft32(v14, -16)
+			v9 += v14
+			v4 ^= v9
+			v4 = bits.RotateLeft32(v4, -12)
+
+			v0 += m[s[12]]
+			v0 += v5
+			v15 ^= v0
+			v15 = bits.RotateLeft32(v15, -8)
+			v10 += v15
+			v5 ^= v10
+			v5 = bits.RotateLeft32(v5, -7)
+			v1 += m[s[13]]
+			v1 += v6
+			v12 ^= v1
+			v12 = bits.RotateLeft32(v12, -8)
+			v11 += v12
+			v6 ^= v11
+			v6 = bits.RotateLeft32(v6, -7)
+			v2 += m[s[14]]
+			v2 += v7
+			v13 ^= v2
+			v13 = bits.RotateLeft32(v13, -8)
+			v8 += v13
+			v7 ^= v8
+			v7 = bits.RotateLeft32(v7, -7)
+			v3 += m[s[15]]
+			v3 += v4
+			v14 ^= v3
+			v14 = bits.RotateLeft32(v14, -8)
+			v9 += v14
+			v4 ^= v9
+			v4 = bits.RotateLeft32(v4, -7)
+		}
+
+		h[0] ^= v0 ^ v8
+		h[1] ^= v1 ^ v9
+		h[2] ^= v2 ^ v10
+		h[3] ^= v3 ^ v11
+		h[4] ^= v4 ^ v12
+		h[5] ^= v5 ^ v13
+		h[6] ^= v6 ^ v14
+		h[7] ^= v7 ^ v15
+	}
+	c[0], c[1] = c0, c1
+}

vendor/golang.org/x/crypto/blake2s/blake2s_ref.go

@@ -0,0 +1,18 @@
+// 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.
+
+//go:build (!amd64 && !386) || !gc || purego
+// +build !amd64,!386 !gc purego
+
+package blake2s
+
+var (
+	useSSE4  = false
+	useSSSE3 = false
+	useSSE2  = false
+)
+
+func hashBlocks(h *[8]uint32, c *[2]uint32, flag uint32, blocks []byte) {
+	hashBlocksGeneric(h, c, flag, blocks)
+}

vendor/golang.org/x/crypto/blake2s/blake2x.go

@@ -0,0 +1,178 @@
+// Copyright 2017 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package blake2s
+
+import (
+	"encoding/binary"
+	"errors"
+	"io"
+)
+
+// XOF defines the interface to hash functions that
+// support arbitrary-length output.
+type XOF interface {
+	// Write absorbs more data into the hash's state. It panics if called
+	// after Read.
+	io.Writer
+
+	// Read reads more output from the hash. It returns io.EOF if the limit
+	// has been reached.
+	io.Reader
+
+	// Clone returns a copy of the XOF in its current state.
+	Clone() XOF
+
+	// Reset resets the XOF to its initial state.
+	Reset()
+}
+
+// OutputLengthUnknown can be used as the size argument to NewXOF to indicate
+// the length of the output is not known in advance.
+const OutputLengthUnknown = 0
+
+// magicUnknownOutputLength is a magic value for the output size that indicates
+// an unknown number of output bytes.
+const magicUnknownOutputLength = 65535
+
+// maxOutputLength is the absolute maximum number of bytes to produce when the
+// number of output bytes is unknown.
+const maxOutputLength = (1 << 32) * 32
+
+// NewXOF creates a new variable-output-length hash. The hash either produce a
+// known number of bytes (1 <= size < 65535), or an unknown number of bytes
+// (size == OutputLengthUnknown). In the latter case, an absolute limit of
+// 128GiB applies.
+//
+// A non-nil key turns the hash into a MAC. The key must between
+// zero and 32 bytes long.
+func NewXOF(size uint16, key []byte) (XOF, error) {
+	if len(key) > Size {
+		return nil, errKeySize
+	}
+	if size == magicUnknownOutputLength {
+		// 2^16-1 indicates an unknown number of bytes and thus isn't a
+		// valid length.
+		return nil, errors.New("blake2s: XOF length too large")
+	}
+	if size == OutputLengthUnknown {
+		size = magicUnknownOutputLength
+	}
+	x := &xof{
+		d: digest{
+			size:   Size,
+			keyLen: len(key),
+		},
+		length: size,
+	}
+	copy(x.d.key[:], key)
+	x.Reset()
+	return x, nil
+}
+
+type xof struct {
+	d                digest
+	length           uint16
+	remaining        uint64
+	cfg, root, block [Size]byte
+	offset           int
+	nodeOffset       uint32
+	readMode         bool
+}
+
+func (x *xof) Write(p []byte) (n int, err error) {
+	if x.readMode {
+		panic("blake2s: write to XOF after read")
+	}
+	return x.d.Write(p)
+}
+
+func (x *xof) Clone() XOF {
+	clone := *x
+	return &clone
+}
+
+func (x *xof) Reset() {
+	x.cfg[0] = byte(Size)
+	binary.LittleEndian.PutUint32(x.cfg[4:], uint32(Size)) // leaf length
+	binary.LittleEndian.PutUint16(x.cfg[12:], x.length)    // XOF length
+	x.cfg[15] = byte(Size)                                 // inner hash size
+
+	x.d.Reset()
+	x.d.h[3] ^= uint32(x.length)
+
+	x.remaining = uint64(x.length)
+	if x.remaining == magicUnknownOutputLength {
+		x.remaining = maxOutputLength
+	}
+	x.offset, x.nodeOffset = 0, 0
+	x.readMode = false
+}
+
+func (x *xof) Read(p []byte) (n int, err error) {
+	if !x.readMode {
+		x.d.finalize(&x.root)
+		x.readMode = true
+	}
+
+	if x.remaining == 0 {
+		return 0, io.EOF
+	}
+
+	n = len(p)
+	if uint64(n) > x.remaining {
+		n = int(x.remaining)
+		p = p[:n]
+	}
+
+	if x.offset > 0 {
+		blockRemaining := Size - x.offset
+		if n < blockRemaining {
+			x.offset += copy(p, x.block[x.offset:])
+			x.remaining -= uint64(n)
+			return
+		}
+		copy(p, x.block[x.offset:])
+		p = p[blockRemaining:]
+		x.offset = 0
+		x.remaining -= uint64(blockRemaining)
+	}
+
+	for len(p) >= Size {
+		binary.LittleEndian.PutUint32(x.cfg[8:], x.nodeOffset)
+		x.nodeOffset++
+
+		x.d.initConfig(&x.cfg)
+		x.d.Write(x.root[:])
+		x.d.finalize(&x.block)
+
+		copy(p, x.block[:])
+		p = p[Size:]
+		x.remaining -= uint64(Size)
+	}
+
+	if todo := len(p); todo > 0 {
+		if x.remaining < uint64(Size) {
+			x.cfg[0] = byte(x.remaining)
+		}
+		binary.LittleEndian.PutUint32(x.cfg[8:], x.nodeOffset)
+		x.nodeOffset++
+
+		x.d.initConfig(&x.cfg)
+		x.d.Write(x.root[:])
+		x.d.finalize(&x.block)
+
+		x.offset = copy(p, x.block[:todo])
+		x.remaining -= uint64(todo)
+	}
+
+	return
+}
+
+func (d *digest) initConfig(cfg *[Size]byte) {
+	d.offset, d.c[0], d.c[1] = 0, 0, 0
+	for i := range d.h {
+		d.h[i] = iv[i] ^ binary.LittleEndian.Uint32(cfg[i*4:])
+	}
+}

vendor/golang.org/x/crypto/blake2s/register.go

@@ -0,0 +1,22 @@
+// Copyright 2017 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build go1.9
+// +build go1.9
+
+package blake2s
+
+import (
+	"crypto"
+	"hash"
+)
+
+func init() {
+	newHash256 := func() hash.Hash {
+		h, _ := New256(nil)
+		return h
+	}
+
+	crypto.RegisterHash(crypto.BLAKE2s_256, newHash256)
+}

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

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

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

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

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

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

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

@@ -0,0 +1,17 @@
+// 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.
+
+//go:build go1.11 && gc && !purego
+// +build go1.11,gc,!purego
+
+package chacha20
+
+const bufSize = 256
+
+//go:noescape
+func xorKeyStreamVX(dst, src []byte, key *[8]uint32, nonce *[3]uint32, counter *uint32)
+
+func (c *Cipher) xorKeyStreamBlocks(dst, src []byte) {
+	xorKeyStreamVX(dst, src, &c.key, &c.nonce, &c.counter)
+}

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

@@ -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.
+
+//go:build go1.11 && gc && !purego
+// +build go1.11,gc,!purego
+
+#include "textflag.h"
+
+#define NUM_ROUNDS 10
+
+// func xorKeyStreamVX(dst, src []byte, key *[8]uint32, nonce *[3]uint32, counter *uint32)
+TEXT ·xorKeyStreamVX(SB), NOSPLIT, $0
+	MOVD	dst+0(FP), R1
+	MOVD	src+24(FP), R2
+	MOVD	src_len+32(FP), R3
+	MOVD	key+48(FP), R4
+	MOVD	nonce+56(FP), R6
+	MOVD	counter+64(FP), R7
+
+	MOVD	$·constants(SB), R10
+	MOVD	$·incRotMatrix(SB), R11
+
+	MOVW	(R7), R20
+
+	AND	$~255, R3, R13
+	ADD	R2, R13, R12 // R12 for block end
+	AND	$255, R3, R13
+loop:
+	MOVD	$NUM_ROUNDS, R21
+	VLD1	(R11), [V30.S4, V31.S4]
+
+	// load contants
+	// VLD4R (R10), [V0.S4, V1.S4, V2.S4, V3.S4]
+	WORD	$0x4D60E940
+
+	// load keys
+	// VLD4R 16(R4), [V4.S4, V5.S4, V6.S4, V7.S4]
+	WORD	$0x4DFFE884
+	// VLD4R 16(R4), [V8.S4, V9.S4, V10.S4, V11.S4]
+	WORD	$0x4DFFE888
+	SUB	$32, R4
+
+	// load counter + nonce
+	// VLD1R (R7), [V12.S4]
+	WORD	$0x4D40C8EC
+
+	// VLD3R (R6), [V13.S4, V14.S4, V15.S4]
+	WORD	$0x4D40E8CD
+
+	// update counter
+	VADD	V30.S4, V12.S4, V12.S4
+
+chacha:
+	// V0..V3 += V4..V7
+	// V12..V15 <<<= ((V12..V15 XOR V0..V3), 16)
+	VADD	V0.S4, V4.S4, V0.S4
+	VADD	V1.S4, V5.S4, V1.S4
+	VADD	V2.S4, V6.S4, V2.S4
+	VADD	V3.S4, V7.S4, V3.S4
+	VEOR	V12.B16, V0.B16, V12.B16
+	VEOR	V13.B16, V1.B16, V13.B16
+	VEOR	V14.B16, V2.B16, V14.B16
+	VEOR	V15.B16, V3.B16, V15.B16
+	VREV32	V12.H8, V12.H8
+	VREV32	V13.H8, V13.H8
+	VREV32	V14.H8, V14.H8
+	VREV32	V15.H8, V15.H8
+	// V8..V11 += V12..V15
+	// V4..V7 <<<= ((V4..V7 XOR V8..V11), 12)
+	VADD	V8.S4, V12.S4, V8.S4
+	VADD	V9.S4, V13.S4, V9.S4
+	VADD	V10.S4, V14.S4, V10.S4
+	VADD	V11.S4, V15.S4, V11.S4
+	VEOR	V8.B16, V4.B16, V16.B16
+	VEOR	V9.B16, V5.B16, V17.B16
+	VEOR	V10.B16, V6.B16, V18.B16
+	VEOR	V11.B16, V7.B16, V19.B16
+	VSHL	$12, V16.S4, V4.S4
+	VSHL	$12, V17.S4, V5.S4
+	VSHL	$12, V18.S4, V6.S4
+	VSHL	$12, V19.S4, V7.S4
+	VSRI	$20, V16.S4, V4.S4
+	VSRI	$20, V17.S4, V5.S4
+	VSRI	$20, V18.S4, V6.S4
+	VSRI	$20, V19.S4, V7.S4
+
+	// V0..V3 += V4..V7
+	// V12..V15 <<<= ((V12..V15 XOR V0..V3), 8)
+	VADD	V0.S4, V4.S4, V0.S4
+	VADD	V1.S4, V5.S4, V1.S4
+	VADD	V2.S4, V6.S4, V2.S4
+	VADD	V3.S4, V7.S4, V3.S4
+	VEOR	V12.B16, V0.B16, V12.B16
+	VEOR	V13.B16, V1.B16, V13.B16
+	VEOR	V14.B16, V2.B16, V14.B16
+	VEOR	V15.B16, V3.B16, V15.B16
+	VTBL	V31.B16, [V12.B16], V12.B16
+	VTBL	V31.B16, [V13.B16], V13.B16
+	VTBL	V31.B16, [V14.B16], V14.B16
+	VTBL	V31.B16, [V15.B16], V15.B16
+
+	// V8..V11 += V12..V15
+	// V4..V7 <<<= ((V4..V7 XOR V8..V11), 7)
+	VADD	V12.S4, V8.S4, V8.S4
+	VADD	V13.S4, V9.S4, V9.S4
+	VADD	V14.S4, V10.S4, V10.S4
+	VADD	V15.S4, V11.S4, V11.S4
+	VEOR	V8.B16, V4.B16, V16.B16
+	VEOR	V9.B16, V5.B16, V17.B16
+	VEOR	V10.B16, V6.B16, V18.B16
+	VEOR	V11.B16, V7.B16, V19.B16
+	VSHL	$7, V16.S4, V4.S4
+	VSHL	$7, V17.S4, V5.S4
+	VSHL	$7, V18.S4, V6.S4
+	VSHL	$7, V19.S4, V7.S4
+	VSRI	$25, V16.S4, V4.S4
+	VSRI	$25, V17.S4, V5.S4
+	VSRI	$25, V18.S4, V6.S4
+	VSRI	$25, V19.S4, V7.S4
+
+	// V0..V3 += V5..V7, V4
+	// V15,V12-V14 <<<= ((V15,V12-V14 XOR V0..V3), 16)
+	VADD	V0.S4, V5.S4, V0.S4
+	VADD	V1.S4, V6.S4, V1.S4
+	VADD	V2.S4, V7.S4, V2.S4
+	VADD	V3.S4, V4.S4, V3.S4
+	VEOR	V15.B16, V0.B16, V15.B16
+	VEOR	V12.B16, V1.B16, V12.B16
+	VEOR	V13.B16, V2.B16, V13.B16
+	VEOR	V14.B16, V3.B16, V14.B16
+	VREV32	V12.H8, V12.H8
+	VREV32	V13.H8, V13.H8
+	VREV32	V14.H8, V14.H8
+	VREV32	V15.H8, V15.H8
+
+	// V10 += V15; V5 <<<= ((V10 XOR V5), 12)
+	// ...
+	VADD	V15.S4, V10.S4, V10.S4
+	VADD	V12.S4, V11.S4, V11.S4
+	VADD	V13.S4, V8.S4, V8.S4
+	VADD	V14.S4, V9.S4, V9.S4
+	VEOR	V10.B16, V5.B16, V16.B16
+	VEOR	V11.B16, V6.B16, V17.B16
+	VEOR	V8.B16, V7.B16, V18.B16
+	VEOR	V9.B16, V4.B16, V19.B16
+	VSHL	$12, V16.S4, V5.S4
+	VSHL	$12, V17.S4, V6.S4
+	VSHL	$12, V18.S4, V7.S4
+	VSHL	$12, V19.S4, V4.S4
+	VSRI	$20, V16.S4, V5.S4
+	VSRI	$20, V17.S4, V6.S4
+	VSRI	$20, V18.S4, V7.S4
+	VSRI	$20, V19.S4, V4.S4
+
+	// V0 += V5; V15 <<<= ((V0 XOR V15), 8)
+	// ...
+	VADD	V5.S4, V0.S4, V0.S4
+	VADD	V6.S4, V1.S4, V1.S4
+	VADD	V7.S4, V2.S4, V2.S4
+	VADD	V4.S4, V3.S4, V3.S4
+	VEOR	V0.B16, V15.B16, V15.B16
+	VEOR	V1.B16, V12.B16, V12.B16
+	VEOR	V2.B16, V13.B16, V13.B16
+	VEOR	V3.B16, V14.B16, V14.B16
+	VTBL	V31.B16, [V12.B16], V12.B16
+	VTBL	V31.B16, [V13.B16], V13.B16
+	VTBL	V31.B16, [V14.B16], V14.B16
+	VTBL	V31.B16, [V15.B16], V15.B16
+
+	// V10 += V15; V5 <<<= ((V10 XOR V5), 7)
+	// ...
+	VADD	V15.S4, V10.S4, V10.S4
+	VADD	V12.S4, V11.S4, V11.S4
+	VADD	V13.S4, V8.S4, V8.S4
+	VADD	V14.S4, V9.S4, V9.S4
+	VEOR	V10.B16, V5.B16, V16.B16
+	VEOR	V11.B16, V6.B16, V17.B16
+	VEOR	V8.B16, V7.B16, V18.B16
+	VEOR	V9.B16, V4.B16, V19.B16
+	VSHL	$7, V16.S4, V5.S4
+	VSHL	$7, V17.S4, V6.S4
+	VSHL	$7, V18.S4, V7.S4
+	VSHL	$7, V19.S4, V4.S4
+	VSRI	$25, V16.S4, V5.S4
+	VSRI	$25, V17.S4, V6.S4
+	VSRI	$25, V18.S4, V7.S4
+	VSRI	$25, V19.S4, V4.S4
+
+	SUB	$1, R21
+	CBNZ	R21, chacha
+
+	// VLD4R (R10), [V16.S4, V17.S4, V18.S4, V19.S4]
+	WORD	$0x4D60E950
+
+	// VLD4R 16(R4), [V20.S4, V21.S4, V22.S4, V23.S4]
+	WORD	$0x4DFFE894
+	VADD	V30.S4, V12.S4, V12.S4
+	VADD	V16.S4, V0.S4, V0.S4
+	VADD	V17.S4, V1.S4, V1.S4
+	VADD	V18.S4, V2.S4, V2.S4
+	VADD	V19.S4, V3.S4, V3.S4
+	// VLD4R 16(R4), [V24.S4, V25.S4, V26.S4, V27.S4]
+	WORD	$0x4DFFE898
+	// restore R4
+	SUB	$32, R4
+
+	// load counter + nonce
+	// VLD1R (R7), [V28.S4]
+	WORD	$0x4D40C8FC
+	// VLD3R (R6), [V29.S4, V30.S4, V31.S4]
+	WORD	$0x4D40E8DD
+
+	VADD	V20.S4, V4.S4, V4.S4
+	VADD	V21.S4, V5.S4, V5.S4
+	VADD	V22.S4, V6.S4, V6.S4
+	VADD	V23.S4, V7.S4, V7.S4
+	VADD	V24.S4, V8.S4, V8.S4
+	VADD	V25.S4, V9.S4, V9.S4
+	VADD	V26.S4, V10.S4, V10.S4
+	VADD	V27.S4, V11.S4, V11.S4
+	VADD	V28.S4, V12.S4, V12.S4
+	VADD	V29.S4, V13.S4, V13.S4
+	VADD	V30.S4, V14.S4, V14.S4
+	VADD	V31.S4, V15.S4, V15.S4
+
+	VZIP1	V1.S4, V0.S4, V16.S4
+	VZIP2	V1.S4, V0.S4, V17.S4
+	VZIP1	V3.S4, V2.S4, V18.S4
+	VZIP2	V3.S4, V2.S4, V19.S4
+	VZIP1	V5.S4, V4.S4, V20.S4
+	VZIP2	V5.S4, V4.S4, V21.S4
+	VZIP1	V7.S4, V6.S4, V22.S4
+	VZIP2	V7.S4, V6.S4, V23.S4
+	VZIP1	V9.S4, V8.S4, V24.S4
+	VZIP2	V9.S4, V8.S4, V25.S4
+	VZIP1	V11.S4, V10.S4, V26.S4
+	VZIP2	V11.S4, V10.S4, V27.S4
+	VZIP1	V13.S4, V12.S4, V28.S4
+	VZIP2	V13.S4, V12.S4, V29.S4
+	VZIP1	V15.S4, V14.S4, V30.S4
+	VZIP2	V15.S4, V14.S4, V31.S4
+	VZIP1	V18.D2, V16.D2, V0.D2
+	VZIP2	V18.D2, V16.D2, V4.D2
+	VZIP1	V19.D2, V17.D2, V8.D2
+	VZIP2	V19.D2, V17.D2, V12.D2
+	VLD1.P	64(R2), [V16.B16, V17.B16, V18.B16, V19.B16]
+
+	VZIP1	V22.D2, V20.D2, V1.D2
+	VZIP2	V22.D2, V20.D2, V5.D2
+	VZIP1	V23.D2, V21.D2, V9.D2
+	VZIP2	V23.D2, V21.D2, V13.D2
+	VLD1.P	64(R2), [V20.B16, V21.B16, V22.B16, V23.B16]
+	VZIP1	V26.D2, V24.D2, V2.D2
+	VZIP2	V26.D2, V24.D2, V6.D2
+	VZIP1	V27.D2, V25.D2, V10.D2
+	VZIP2	V27.D2, V25.D2, V14.D2
+	VLD1.P	64(R2), [V24.B16, V25.B16, V26.B16, V27.B16]
+	VZIP1	V30.D2, V28.D2, V3.D2
+	VZIP2	V30.D2, V28.D2, V7.D2
+	VZIP1	V31.D2, V29.D2, V11.D2
+	VZIP2	V31.D2, V29.D2, V15.D2
+	VLD1.P	64(R2), [V28.B16, V29.B16, V30.B16, V31.B16]
+	VEOR	V0.B16, V16.B16, V16.B16
+	VEOR	V1.B16, V17.B16, V17.B16
+	VEOR	V2.B16, V18.B16, V18.B16
+	VEOR	V3.B16, V19.B16, V19.B16
+	VST1.P	[V16.B16, V17.B16, V18.B16, V19.B16], 64(R1)
+	VEOR	V4.B16, V20.B16, V20.B16
+	VEOR	V5.B16, V21.B16, V21.B16
+	VEOR	V6.B16, V22.B16, V22.B16
+	VEOR	V7.B16, V23.B16, V23.B16
+	VST1.P	[V20.B16, V21.B16, V22.B16, V23.B16], 64(R1)
+	VEOR	V8.B16, V24.B16, V24.B16
+	VEOR	V9.B16, V25.B16, V25.B16
+	VEOR	V10.B16, V26.B16, V26.B16
+	VEOR	V11.B16, V27.B16, V27.B16
+	VST1.P	[V24.B16, V25.B16, V26.B16, V27.B16], 64(R1)
+	VEOR	V12.B16, V28.B16, V28.B16
+	VEOR	V13.B16, V29.B16, V29.B16
+	VEOR	V14.B16, V30.B16, V30.B16
+	VEOR	V15.B16, V31.B16, V31.B16
+	VST1.P	[V28.B16, V29.B16, V30.B16, V31.B16], 64(R1)
+
+	ADD	$4, R20
+	MOVW	R20, (R7) // update counter
+
+	CMP	R2, R12
+	BGT	loop
+
+	RET
+
+
+DATA	·constants+0x00(SB)/4, $0x61707865
+DATA	·constants+0x04(SB)/4, $0x3320646e
+DATA	·constants+0x08(SB)/4, $0x79622d32
+DATA	·constants+0x0c(SB)/4, $0x6b206574
+GLOBL	·constants(SB), NOPTR|RODATA, $32
+
+DATA	·incRotMatrix+0x00(SB)/4, $0x00000000
+DATA	·incRotMatrix+0x04(SB)/4, $0x00000001
+DATA	·incRotMatrix+0x08(SB)/4, $0x00000002
+DATA	·incRotMatrix+0x0c(SB)/4, $0x00000003
+DATA	·incRotMatrix+0x10(SB)/4, $0x02010003
+DATA	·incRotMatrix+0x14(SB)/4, $0x06050407
+DATA	·incRotMatrix+0x18(SB)/4, $0x0A09080B
+DATA	·incRotMatrix+0x1c(SB)/4, $0x0E0D0C0F
+GLOBL	·incRotMatrix(SB), NOPTR|RODATA, $32

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

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

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

@@ -0,0 +1,14 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build (!arm64 && !s390x && !ppc64le) || (arm64 && !go1.11) || !gc || purego
+// +build !arm64,!s390x,!ppc64le arm64,!go1.11 !gc purego
+
+package chacha20
+
+const bufSize = blockSize
+
+func (s *Cipher) xorKeyStreamBlocks(dst, src []byte) {
+	s.xorKeyStreamBlocksGeneric(dst, src)
+}

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

@@ -0,0 +1,17 @@
+// Copyright 2019 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build gc && !purego
+// +build gc,!purego
+
+package chacha20
+
+const bufSize = 256
+
+//go:noescape
+func chaCha20_ctr32_vsx(out, inp *byte, len int, key *[8]uint32, counter *uint32)
+
+func (c *Cipher) xorKeyStreamBlocks(dst, src []byte) {
+	chaCha20_ctr32_vsx(&dst[0], &src[0], len(src), &c.key, &c.counter)
+}

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

@@ -0,0 +1,450 @@
+// 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.
+
+//go:build gc && !purego
+// +build gc,!purego
+
+#include "textflag.h"
+
+#define OUT  R3
+#define INP  R4
+#define LEN  R5
+#define KEY  R6
+#define CNT  R7
+#define TMP  R15
+
+#define CONSTBASE  R16
+#define BLOCKS R17
+
+DATA consts<>+0x00(SB)/8, $0x3320646e61707865
+DATA consts<>+0x08(SB)/8, $0x6b20657479622d32
+DATA consts<>+0x10(SB)/8, $0x0000000000000001
+DATA consts<>+0x18(SB)/8, $0x0000000000000000
+DATA consts<>+0x20(SB)/8, $0x0000000000000004
+DATA consts<>+0x28(SB)/8, $0x0000000000000000
+DATA consts<>+0x30(SB)/8, $0x0a0b08090e0f0c0d
+DATA consts<>+0x38(SB)/8, $0x0203000106070405
+DATA consts<>+0x40(SB)/8, $0x090a0b080d0e0f0c
+DATA consts<>+0x48(SB)/8, $0x0102030005060704
+DATA consts<>+0x50(SB)/8, $0x6170786561707865
+DATA consts<>+0x58(SB)/8, $0x6170786561707865
+DATA consts<>+0x60(SB)/8, $0x3320646e3320646e
+DATA consts<>+0x68(SB)/8, $0x3320646e3320646e
+DATA consts<>+0x70(SB)/8, $0x79622d3279622d32
+DATA consts<>+0x78(SB)/8, $0x79622d3279622d32
+DATA consts<>+0x80(SB)/8, $0x6b2065746b206574
+DATA consts<>+0x88(SB)/8, $0x6b2065746b206574
+DATA consts<>+0x90(SB)/8, $0x0000000100000000
+DATA consts<>+0x98(SB)/8, $0x0000000300000002
+GLOBL consts<>(SB), RODATA, $0xa0
+
+//func chaCha20_ctr32_vsx(out, inp *byte, len int, key *[8]uint32, counter *uint32)
+TEXT ·chaCha20_ctr32_vsx(SB),NOSPLIT,$64-40
+	MOVD out+0(FP), OUT
+	MOVD inp+8(FP), INP
+	MOVD len+16(FP), LEN
+	MOVD key+24(FP), KEY
+	MOVD counter+32(FP), CNT
+
+	// Addressing for constants
+	MOVD $consts<>+0x00(SB), CONSTBASE
+	MOVD $16, R8
+	MOVD $32, R9
+	MOVD $48, R10
+	MOVD $64, R11
+	SRD $6, LEN, BLOCKS
+	// V16
+	LXVW4X (CONSTBASE)(R0), VS48
+	ADD $80,CONSTBASE
+
+	// Load key into V17,V18
+	LXVW4X (KEY)(R0), VS49
+	LXVW4X (KEY)(R8), VS50
+
+	// Load CNT, NONCE into V19
+	LXVW4X (CNT)(R0), VS51
+
+	// Clear V27
+	VXOR V27, V27, V27
+
+	// V28
+	LXVW4X (CONSTBASE)(R11), VS60
+
+	// splat slot from V19 -> V26
+	VSPLTW $0, V19, V26
+
+	VSLDOI $4, V19, V27, V19
+	VSLDOI $12, V27, V19, V19
+
+	VADDUWM V26, V28, V26
+
+	MOVD $10, R14
+	MOVD R14, CTR
+
+loop_outer_vsx:
+	// V0, V1, V2, V3
+	LXVW4X (R0)(CONSTBASE), VS32
+	LXVW4X (R8)(CONSTBASE), VS33
+	LXVW4X (R9)(CONSTBASE), VS34
+	LXVW4X (R10)(CONSTBASE), VS35
+
+	// splat values from V17, V18 into V4-V11
+	VSPLTW $0, V17, V4
+	VSPLTW $1, V17, V5
+	VSPLTW $2, V17, V6
+	VSPLTW $3, V17, V7
+	VSPLTW $0, V18, V8
+	VSPLTW $1, V18, V9
+	VSPLTW $2, V18, V10
+	VSPLTW $3, V18, V11
+
+	// VOR
+	VOR V26, V26, V12
+
+	// splat values from V19 -> V13, V14, V15
+	VSPLTW $1, V19, V13
+	VSPLTW $2, V19, V14
+	VSPLTW $3, V19, V15
+
+	// splat   const values
+	VSPLTISW $-16, V27
+	VSPLTISW $12, V28
+	VSPLTISW $8, V29
+	VSPLTISW $7, V30
+
+loop_vsx:
+	VADDUWM V0, V4, V0
+	VADDUWM V1, V5, V1
+	VADDUWM V2, V6, V2
+	VADDUWM V3, V7, V3
+
+	VXOR V12, V0, V12
+	VXOR V13, V1, V13
+	VXOR V14, V2, V14
+	VXOR V15, V3, V15
+
+	VRLW V12, V27, V12
+	VRLW V13, V27, V13
+	VRLW V14, V27, V14
+	VRLW V15, V27, V15
+
+	VADDUWM V8, V12, V8
+	VADDUWM V9, V13, V9
+	VADDUWM V10, V14, V10
+	VADDUWM V11, V15, V11
+
+	VXOR V4, V8, V4
+	VXOR V5, V9, V5
+	VXOR V6, V10, V6
+	VXOR V7, V11, V7
+
+	VRLW V4, V28, V4
+	VRLW V5, V28, V5
+	VRLW V6, V28, V6
+	VRLW V7, V28, V7
+
+	VADDUWM V0, V4, V0
+	VADDUWM V1, V5, V1
+	VADDUWM V2, V6, V2
+	VADDUWM V3, V7, V3
+
+	VXOR V12, V0, V12
+	VXOR V13, V1, V13
+	VXOR V14, V2, V14
+	VXOR V15, V3, V15
+
+	VRLW V12, V29, V12
+	VRLW V13, V29, V13
+	VRLW V14, V29, V14
+	VRLW V15, V29, V15
+
+	VADDUWM V8, V12, V8
+	VADDUWM V9, V13, V9
+	VADDUWM V10, V14, V10
+	VADDUWM V11, V15, V11
+
+	VXOR V4, V8, V4
+	VXOR V5, V9, V5
+	VXOR V6, V10, V6
+	VXOR V7, V11, V7
+
+	VRLW V4, V30, V4
+	VRLW V5, V30, V5
+	VRLW V6, V30, V6
+	VRLW V7, V30, V7
+
+	VADDUWM V0, V5, V0
+	VADDUWM V1, V6, V1
+	VADDUWM V2, V7, V2
+	VADDUWM V3, V4, V3
+
+	VXOR V15, V0, V15
+	VXOR V12, V1, V12
+	VXOR V13, V2, V13
+	VXOR V14, V3, V14
+
+	VRLW V15, V27, V15
+	VRLW V12, V27, V12
+	VRLW V13, V27, V13
+	VRLW V14, V27, V14
+
+	VADDUWM V10, V15, V10
+	VADDUWM V11, V12, V11
+	VADDUWM V8, V13, V8
+	VADDUWM V9, V14, V9
+
+	VXOR V5, V10, V5
+	VXOR V6, V11, V6
+	VXOR V7, V8, V7
+	VXOR V4, V9, V4
+
+	VRLW V5, V28, V5
+	VRLW V6, V28, V6
+	VRLW V7, V28, V7
+	VRLW V4, V28, V4
+
+	VADDUWM V0, V5, V0
+	VADDUWM V1, V6, V1
+	VADDUWM V2, V7, V2
+	VADDUWM V3, V4, V3
+
+	VXOR V15, V0, V15
+	VXOR V12, V1, V12
+	VXOR V13, V2, V13
+	VXOR V14, V3, V14
+
+	VRLW V15, V29, V15
+	VRLW V12, V29, V12
+	VRLW V13, V29, V13
+	VRLW V14, V29, V14
+
+	VADDUWM V10, V15, V10
+	VADDUWM V11, V12, V11
+	VADDUWM V8, V13, V8
+	VADDUWM V9, V14, V9
+
+	VXOR V5, V10, V5
+	VXOR V6, V11, V6
+	VXOR V7, V8, V7
+	VXOR V4, V9, V4
+
+	VRLW V5, V30, V5
+	VRLW V6, V30, V6
+	VRLW V7, V30, V7
+	VRLW V4, V30, V4
+	BC   16, LT, loop_vsx
+
+	VADDUWM V12, V26, V12
+
+	WORD $0x13600F8C		// VMRGEW V0, V1, V27
+	WORD $0x13821F8C		// VMRGEW V2, V3, V28
+
+	WORD $0x10000E8C		// VMRGOW V0, V1, V0
+	WORD $0x10421E8C		// VMRGOW V2, V3, V2
+
+	WORD $0x13A42F8C		// VMRGEW V4, V5, V29
+	WORD $0x13C63F8C		// VMRGEW V6, V7, V30
+
+	XXPERMDI VS32, VS34, $0, VS33
+	XXPERMDI VS32, VS34, $3, VS35
+	XXPERMDI VS59, VS60, $0, VS32
+	XXPERMDI VS59, VS60, $3, VS34
+
+	WORD $0x10842E8C		// VMRGOW V4, V5, V4
+	WORD $0x10C63E8C		// VMRGOW V6, V7, V6
+
+	WORD $0x13684F8C		// VMRGEW V8, V9, V27
+	WORD $0x138A5F8C		// VMRGEW V10, V11, V28
+
+	XXPERMDI VS36, VS38, $0, VS37
+	XXPERMDI VS36, VS38, $3, VS39
+	XXPERMDI VS61, VS62, $0, VS36
+	XXPERMDI VS61, VS62, $3, VS38
+
+	WORD $0x11084E8C		// VMRGOW V8, V9, V8
+	WORD $0x114A5E8C		// VMRGOW V10, V11, V10
+
+	WORD $0x13AC6F8C		// VMRGEW V12, V13, V29
+	WORD $0x13CE7F8C		// VMRGEW V14, V15, V30
+
+	XXPERMDI VS40, VS42, $0, VS41
+	XXPERMDI VS40, VS42, $3, VS43
+	XXPERMDI VS59, VS60, $0, VS40
+	XXPERMDI VS59, VS60, $3, VS42
+
+	WORD $0x118C6E8C		// VMRGOW V12, V13, V12
+	WORD $0x11CE7E8C		// VMRGOW V14, V15, V14
+
+	VSPLTISW $4, V27
+	VADDUWM V26, V27, V26
+
+	XXPERMDI VS44, VS46, $0, VS45
+	XXPERMDI VS44, VS46, $3, VS47
+	XXPERMDI VS61, VS62, $0, VS44
+	XXPERMDI VS61, VS62, $3, VS46
+
+	VADDUWM V0, V16, V0
+	VADDUWM V4, V17, V4
+	VADDUWM V8, V18, V8
+	VADDUWM V12, V19, V12
+
+	CMPU LEN, $64
+	BLT tail_vsx
+
+	// Bottom of loop
+	LXVW4X (INP)(R0), VS59
+	LXVW4X (INP)(R8), VS60
+	LXVW4X (INP)(R9), VS61
+	LXVW4X (INP)(R10), VS62
+
+	VXOR V27, V0, V27
+	VXOR V28, V4, V28
+	VXOR V29, V8, V29
+	VXOR V30, V12, V30
+
+	STXVW4X VS59, (OUT)(R0)
+	STXVW4X VS60, (OUT)(R8)
+	ADD     $64, INP
+	STXVW4X VS61, (OUT)(R9)
+	ADD     $-64, LEN
+	STXVW4X VS62, (OUT)(R10)
+	ADD     $64, OUT
+	BEQ     done_vsx
+
+	VADDUWM V1, V16, V0
+	VADDUWM V5, V17, V4
+	VADDUWM V9, V18, V8
+	VADDUWM V13, V19, V12
+
+	CMPU  LEN, $64
+	BLT   tail_vsx
+
+	LXVW4X (INP)(R0), VS59
+	LXVW4X (INP)(R8), VS60
+	LXVW4X (INP)(R9), VS61
+	LXVW4X (INP)(R10), VS62
+	VXOR   V27, V0, V27
+
+	VXOR V28, V4, V28
+	VXOR V29, V8, V29
+	VXOR V30, V12, V30
+
+	STXVW4X VS59, (OUT)(R0)
+	STXVW4X VS60, (OUT)(R8)
+	ADD     $64, INP
+	STXVW4X VS61, (OUT)(R9)
+	ADD     $-64, LEN
+	STXVW4X VS62, (OUT)(V10)
+	ADD     $64, OUT
+	BEQ     done_vsx
+
+	VADDUWM V2, V16, V0
+	VADDUWM V6, V17, V4
+	VADDUWM V10, V18, V8
+	VADDUWM V14, V19, V12
+
+	CMPU LEN, $64
+	BLT  tail_vsx
+
+	LXVW4X (INP)(R0), VS59
+	LXVW4X (INP)(R8), VS60
+	LXVW4X (INP)(R9), VS61
+	LXVW4X (INP)(R10), VS62
+
+	VXOR V27, V0, V27
+	VXOR V28, V4, V28
+	VXOR V29, V8, V29
+	VXOR V30, V12, V30
+
+	STXVW4X VS59, (OUT)(R0)
+	STXVW4X VS60, (OUT)(R8)
+	ADD     $64, INP
+	STXVW4X VS61, (OUT)(R9)
+	ADD     $-64, LEN
+	STXVW4X VS62, (OUT)(R10)
+	ADD     $64, OUT
+	BEQ     done_vsx
+
+	VADDUWM V3, V16, V0
+	VADDUWM V7, V17, V4
+	VADDUWM V11, V18, V8
+	VADDUWM V15, V19, V12
+
+	CMPU  LEN, $64
+	BLT   tail_vsx
+
+	LXVW4X (INP)(R0), VS59
+	LXVW4X (INP)(R8), VS60
+	LXVW4X (INP)(R9), VS61
+	LXVW4X (INP)(R10), VS62
+
+	VXOR V27, V0, V27
+	VXOR V28, V4, V28
+	VXOR V29, V8, V29
+	VXOR V30, V12, V30
+
+	STXVW4X VS59, (OUT)(R0)
+	STXVW4X VS60, (OUT)(R8)
+	ADD     $64, INP
+	STXVW4X VS61, (OUT)(R9)
+	ADD     $-64, LEN
+	STXVW4X VS62, (OUT)(R10)
+	ADD     $64, OUT
+
+	MOVD $10, R14
+	MOVD R14, CTR
+	BNE  loop_outer_vsx
+
+done_vsx:
+	// Increment counter by number of 64 byte blocks
+	MOVD (CNT), R14
+	ADD  BLOCKS, R14
+	MOVD R14, (CNT)
+	RET
+
+tail_vsx:
+	ADD  $32, R1, R11
+	MOVD LEN, CTR
+
+	// Save values on stack to copy from
+	STXVW4X VS32, (R11)(R0)
+	STXVW4X VS36, (R11)(R8)
+	STXVW4X VS40, (R11)(R9)
+	STXVW4X VS44, (R11)(R10)
+	ADD $-1, R11, R12
+	ADD $-1, INP
+	ADD $-1, OUT
+
+looptail_vsx:
+	// Copying the result to OUT
+	// in bytes.
+	MOVBZU 1(R12), KEY
+	MOVBZU 1(INP), TMP
+	XOR    KEY, TMP, KEY
+	MOVBU  KEY, 1(OUT)
+	BC     16, LT, looptail_vsx
+
+	// Clear the stack values
+	STXVW4X VS48, (R11)(R0)
+	STXVW4X VS48, (R11)(R8)
+	STXVW4X VS48, (R11)(R9)
+	STXVW4X VS48, (R11)(R10)
+	BR      done_vsx

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

@@ -0,0 +1,27 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build gc && !purego
+// +build gc,!purego
+
+package chacha20
+
+import "golang.org/x/sys/cpu"
+
+var haveAsm = cpu.S390X.HasVX
+
+const bufSize = 256
+
+// xorKeyStreamVX is an assembly implementation of XORKeyStream. It must only
+// be called when the vector facility is available. Implementation in asm_s390x.s.
+//go:noescape
+func xorKeyStreamVX(dst, src []byte, key *[8]uint32, nonce *[3]uint32, counter *uint32)
+
+func (c *Cipher) xorKeyStreamBlocks(dst, src []byte) {
+	if cpu.S390X.HasVX {
+		xorKeyStreamVX(dst, src, &c.key, &c.nonce, &c.counter)
+	} else {
+		c.xorKeyStreamBlocksGeneric(dst, src)
+	}
+}

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

@@ -0,0 +1,225 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build gc && !purego
+// +build gc,!purego
+
+#include "go_asm.h"
+#include "textflag.h"
+
+// This is an implementation of the ChaCha20 encryption algorithm as
+// specified in RFC 7539. It uses vector instructions to compute
+// 4 keystream blocks in parallel (256 bytes) which are then XORed
+// with the bytes in the input slice.
+
+GLOBL ·constants<>(SB), RODATA|NOPTR, $32
+// BSWAP: swap bytes in each 4-byte element
+DATA ·constants<>+0x00(SB)/4, $0x03020100
+DATA ·constants<>+0x04(SB)/4, $0x07060504
+DATA ·constants<>+0x08(SB)/4, $0x0b0a0908
+DATA ·constants<>+0x0c(SB)/4, $0x0f0e0d0c
+// J0: [j0, j1, j2, j3]
+DATA ·constants<>+0x10(SB)/4, $0x61707865
+DATA ·constants<>+0x14(SB)/4, $0x3320646e
+DATA ·constants<>+0x18(SB)/4, $0x79622d32
+DATA ·constants<>+0x1c(SB)/4, $0x6b206574
+
+#define BSWAP V5
+#define J0    V6
+#define KEY0  V7
+#define KEY1  V8
+#define NONCE V9
+#define CTR   V10
+#define M0    V11
+#define M1    V12
+#define M2    V13
+#define M3    V14
+#define INC   V15
+#define X0    V16
+#define X1    V17
+#define X2    V18
+#define X3    V19
+#define X4    V20
+#define X5    V21
+#define X6    V22
+#define X7    V23
+#define X8    V24
+#define X9    V25
+#define X10   V26
+#define X11   V27
+#define X12   V28
+#define X13   V29
+#define X14   V30
+#define X15   V31
+
+#define NUM_ROUNDS 20
+
+#define ROUND4(a0, a1, a2, a3, b0, b1, b2, b3, c0, c1, c2, c3, d0, d1, d2, d3) \
+	VAF    a1, a0, a0  \
+	VAF    b1, b0, b0  \
+	VAF    c1, c0, c0  \
+	VAF    d1, d0, d0  \
+	VX     a0, a2, a2  \
+	VX     b0, b2, b2  \
+	VX     c0, c2, c2  \
+	VX     d0, d2, d2  \
+	VERLLF $16, a2, a2 \
+	VERLLF $16, b2, b2 \
+	VERLLF $16, c2, c2 \
+	VERLLF $16, d2, d2 \
+	VAF    a2, a3, a3  \
+	VAF    b2, b3, b3  \
+	VAF    c2, c3, c3  \
+	VAF    d2, d3, d3  \
+	VX     a3, a1, a1  \
+	VX     b3, b1, b1  \
+	VX     c3, c1, c1  \
+	VX     d3, d1, d1  \
+	VERLLF $12, a1, a1 \
+	VERLLF $12, b1, b1 \
+	VERLLF $12, c1, c1 \
+	VERLLF $12, d1, d1 \
+	VAF    a1, a0, a0  \
+	VAF    b1, b0, b0  \
+	VAF    c1, c0, c0  \
+	VAF    d1, d0, d0  \
+	VX     a0, a2, a2  \
+	VX     b0, b2, b2  \
+	VX     c0, c2, c2  \
+	VX     d0, d2, d2  \
+	VERLLF $8, a2, a2  \
+	VERLLF $8, b2, b2  \
+	VERLLF $8, c2, c2  \
+	VERLLF $8, d2, d2  \
+	VAF    a2, a3, a3  \
+	VAF    b2, b3, b3  \
+	VAF    c2, c3, c3  \
+	VAF    d2, d3, d3  \
+	VX     a3, a1, a1  \
+	VX     b3, b1, b1  \
+	VX     c3, c1, c1  \
+	VX     d3, d1, d1  \
+	VERLLF $7, a1, a1  \
+	VERLLF $7, b1, b1  \
+	VERLLF $7, c1, c1  \
+	VERLLF $7, d1, d1
+
+#define PERMUTE(mask, v0, v1, v2, v3) \
+	VPERM v0, v0, mask, v0 \
+	VPERM v1, v1, mask, v1 \
+	VPERM v2, v2, mask, v2 \
+	VPERM v3, v3, mask, v3
+
+#define ADDV(x, v0, v1, v2, v3) \
+	VAF x, v0, v0 \
+	VAF x, v1, v1 \
+	VAF x, v2, v2 \
+	VAF x, v3, v3
+
+#define XORV(off, dst, src, v0, v1, v2, v3) \
+	VLM  off(src), M0, M3          \
+	PERMUTE(BSWAP, v0, v1, v2, v3) \
+	VX   v0, M0, M0                \
+	VX   v1, M1, M1                \
+	VX   v2, M2, M2                \
+	VX   v3, M3, M3                \
+	VSTM M0, M3, off(dst)
+
+#define SHUFFLE(a, b, c, d, t, u, v, w) \
+	VMRHF a, c, t \ // t = {a[0], c[0], a[1], c[1]}
+	VMRHF b, d, u \ // u = {b[0], d[0], b[1], d[1]}
+	VMRLF a, c, v \ // v = {a[2], c[2], a[3], c[3]}
+	VMRLF b, d, w \ // w = {b[2], d[2], b[3], d[3]}
+	VMRHF t, u, a \ // a = {a[0], b[0], c[0], d[0]}
+	VMRLF t, u, b \ // b = {a[1], b[1], c[1], d[1]}
+	VMRHF v, w, c \ // c = {a[2], b[2], c[2], d[2]}
+	VMRLF v, w, d // d = {a[3], b[3], c[3], d[3]}
+
+// func xorKeyStreamVX(dst, src []byte, key *[8]uint32, nonce *[3]uint32, counter *uint32)
+TEXT ·xorKeyStreamVX(SB), NOSPLIT, $0
+	MOVD $·constants<>(SB), R1
+	MOVD dst+0(FP), R2         // R2=&dst[0]
+	LMG  src+24(FP), R3, R4    // R3=&src[0] R4=len(src)
+	MOVD key+48(FP), R5        // R5=key
+	MOVD nonce+56(FP), R6      // R6=nonce
+	MOVD counter+64(FP), R7    // R7=counter
+
+	// load BSWAP and J0
+	VLM (R1), BSWAP, J0
+
+	// setup
+	MOVD  $95, R0
+	VLM   (R5), KEY0, KEY1
+	VLL   R0, (R6), NONCE
+	VZERO M0
+	VLEIB $7, $32, M0
+	VSRLB M0, NONCE, NONCE
+
+	// initialize counter values
+	VLREPF (R7), CTR
+	VZERO  INC
+	VLEIF  $1, $1, INC
+	VLEIF  $2, $2, INC
+	VLEIF  $3, $3, INC
+	VAF    INC, CTR, CTR
+	VREPIF $4, INC
+
+chacha:
+	VREPF $0, J0, X0
+	VREPF $1, J0, X1
+	VREPF $2, J0, X2
+	VREPF $3, J0, X3
+	VREPF $0, KEY0, X4
+	VREPF $1, KEY0, X5
+	VREPF $2, KEY0, X6
+	VREPF $3, KEY0, X7
+	VREPF $0, KEY1, X8
+	VREPF $1, KEY1, X9
+	VREPF $2, KEY1, X10
+	VREPF $3, KEY1, X11
+	VLR   CTR, X12
+	VREPF $1, NONCE, X13
+	VREPF $2, NONCE, X14
+	VREPF $3, NONCE, X15
+
+	MOVD $(NUM_ROUNDS/2), R1
+
+loop:
+	ROUND4(X0, X4, X12,  X8, X1, X5, X13,  X9, X2, X6, X14, X10, X3, X7, X15, X11)
+	ROUND4(X0, X5, X15, X10, X1, X6, X12, X11, X2, X7, X13, X8,  X3, X4, X14, X9)
+
+	ADD $-1, R1
+	BNE loop
+
+	// decrement length
+	ADD $-256, R4
+
+	// rearrange vectors
+	SHUFFLE(X0, X1, X2, X3, M0, M1, M2, M3)
+	ADDV(J0, X0, X1, X2, X3)
+	SHUFFLE(X4, X5, X6, X7, M0, M1, M2, M3)
+	ADDV(KEY0, X4, X5, X6, X7)
+	SHUFFLE(X8, X9, X10, X11, M0, M1, M2, M3)
+	ADDV(KEY1, X8, X9, X10, X11)
+	VAF CTR, X12, X12
+	SHUFFLE(X12, X13, X14, X15, M0, M1, M2, M3)
+	ADDV(NONCE, X12, X13, X14, X15)
+
+	// increment counters
+	VAF INC, CTR, CTR
+
+	// xor keystream with plaintext
+	XORV(0*64, R2, R3, X0, X4,  X8, X12)
+	XORV(1*64, R2, R3, X1, X5,  X9, X13)
+	XORV(2*64, R2, R3, X2, X6, X10, X14)
+	XORV(3*64, R2, R3, X3, X7, X11, X15)
+
+	// increment pointers
+	MOVD $256(R2), R2
+	MOVD $256(R3), R3
+
+	CMPBNE  R4, $0, chacha
+
+	VSTEF $0, CTR, (R7)
+	RET

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

@@ -0,0 +1,42 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found src the LICENSE file.
+
+package chacha20
+
+import "runtime"
+
+// Platforms that have fast unaligned 32-bit little endian accesses.
+const unaligned = runtime.GOARCH == "386" ||
+	runtime.GOARCH == "amd64" ||
+	runtime.GOARCH == "arm64" ||
+	runtime.GOARCH == "ppc64le" ||
+	runtime.GOARCH == "s390x"
+
+// addXor reads a little endian uint32 from src, XORs it with (a + b) and
+// places the result in little endian byte order in dst.
+func addXor(dst, src []byte, a, b uint32) {
+	_, _ = src[3], dst[3] // bounds check elimination hint
+	if unaligned {
+		// The compiler should optimize this code into
+		// 32-bit unaligned little endian loads and stores.
+		// TODO: delete once the compiler does a reliably
+		// good job with the generic code below.
+		// See issue #25111 for more details.
+		v := uint32(src[0])
+		v |= uint32(src[1]) << 8
+		v |= uint32(src[2]) << 16
+		v |= uint32(src[3]) << 24
+		v ^= a + b
+		dst[0] = byte(v)
+		dst[1] = byte(v >> 8)
+		dst[2] = byte(v >> 16)
+		dst[3] = byte(v >> 24)
+	} else {
+		a += b
+		dst[0] = src[0] ^ byte(a)
+		dst[1] = src[1] ^ byte(a>>8)
+		dst[2] = src[2] ^ byte(a>>16)
+		dst[3] = src[3] ^ byte(a>>24)
+	}
+}

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

@@ -0,0 +1,145 @@
+// 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.
+
+// Package curve25519 provides an implementation of the X25519 function, which
+// performs scalar multiplication on the elliptic curve known as Curve25519.
+// See RFC 7748.
+package curve25519 // import "golang.org/x/crypto/curve25519"
+
+import (
+	"crypto/subtle"
+	"fmt"
+
+	"golang.org/x/crypto/curve25519/internal/field"
+)
+
+// ScalarMult sets dst to the product scalar * point.
+//
+// Deprecated: when provided a low-order point, ScalarMult will set dst to all
+// zeroes, irrespective of the scalar. Instead, use the X25519 function, which
+// will return an error.
+func ScalarMult(dst, scalar, point *[32]byte) {
+	var e [32]byte
+
+	copy(e[:], scalar[:])
+	e[0] &= 248
+	e[31] &= 127
+	e[31] |= 64
+
+	var x1, x2, z2, x3, z3, tmp0, tmp1 field.Element
+	x1.SetBytes(point[:])
+	x2.One()
+	x3.Set(&x1)
+	z3.One()
+
+	swap := 0
+	for pos := 254; pos >= 0; pos-- {
+		b := e[pos/8] >> uint(pos&7)
+		b &= 1
+		swap ^= int(b)
+		x2.Swap(&x3, swap)
+		z2.Swap(&z3, swap)
+		swap = int(b)
+
+		tmp0.Subtract(&x3, &z3)
+		tmp1.Subtract(&x2, &z2)
+		x2.Add(&x2, &z2)
+		z2.Add(&x3, &z3)
+		z3.Multiply(&tmp0, &x2)
+		z2.Multiply(&z2, &tmp1)
+		tmp0.Square(&tmp1)
+		tmp1.Square(&x2)
+		x3.Add(&z3, &z2)
+		z2.Subtract(&z3, &z2)
+		x2.Multiply(&tmp1, &tmp0)
+		tmp1.Subtract(&tmp1, &tmp0)
+		z2.Square(&z2)
+
+		z3.Mult32(&tmp1, 121666)
+		x3.Square(&x3)
+		tmp0.Add(&tmp0, &z3)
+		z3.Multiply(&x1, &z2)
+		z2.Multiply(&tmp1, &tmp0)
+	}
+
+	x2.Swap(&x3, swap)
+	z2.Swap(&z3, swap)
+
+	z2.Invert(&z2)
+	x2.Multiply(&x2, &z2)
+	copy(dst[:], x2.Bytes())
+}
+
+// ScalarBaseMult sets dst to the product scalar * base where base is the
+// standard generator.
+//
+// It is recommended to use the X25519 function with Basepoint instead, as
+// copying into fixed size arrays can lead to unexpected bugs.
+func ScalarBaseMult(dst, scalar *[32]byte) {
+	ScalarMult(dst, scalar, &basePoint)
+}
+
+const (
+	// ScalarSize is the size of the scalar input to X25519.
+	ScalarSize = 32
+	// PointSize is the size of the point input to X25519.
+	PointSize = 32
+)
+
+// Basepoint is the canonical Curve25519 generator.
+var Basepoint []byte
+
+var basePoint = [32]byte{9, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
+
+func init() { Basepoint = basePoint[:] }
+
+func checkBasepoint() {
+	if subtle.ConstantTimeCompare(Basepoint, []byte{
+		0x09, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+	}) != 1 {
+		panic("curve25519: global Basepoint value was modified")
+	}
+}
+
+// X25519 returns the result of the scalar multiplication (scalar * point),
+// according to RFC 7748, Section 5. scalar, point and the return value are
+// slices of 32 bytes.
+//
+// scalar can be generated at random, for example with crypto/rand. point should
+// be either Basepoint or the output of another X25519 call.
+//
+// If point is Basepoint (but not if it's a different slice with the same
+// contents) a precomputed implementation might be used for performance.
+func X25519(scalar, point []byte) ([]byte, error) {
+	// Outline the body of function, to let the allocation be inlined in the
+	// caller, and possibly avoid escaping to the heap.
+	var dst [32]byte
+	return x25519(&dst, scalar, point)
+}
+
+func x25519(dst *[32]byte, scalar, point []byte) ([]byte, error) {
+	var in [32]byte
+	if l := len(scalar); l != 32 {
+		return nil, fmt.Errorf("bad scalar length: %d, expected %d", l, 32)
+	}
+	if l := len(point); l != 32 {
+		return nil, fmt.Errorf("bad point length: %d, expected %d", l, 32)
+	}
+	copy(in[:], scalar)
+	if &point[0] == &Basepoint[0] {
+		checkBasepoint()
+		ScalarBaseMult(dst, &in)
+	} else {
+		var base, zero [32]byte
+		copy(base[:], point)
+		ScalarMult(dst, &in, &base)
+		if subtle.ConstantTimeCompare(dst[:], zero[:]) == 1 {
+			return nil, fmt.Errorf("bad input point: low order point")
+		}
+	}
+	return dst[:], nil
+}

vendor/golang.org/x/crypto/curve25519/internal/field/README

@@ -0,0 +1,7 @@
+This package is kept in sync with crypto/ed25519/internal/edwards25519/field in
+the standard library.
+
+If there are any changes in the standard library that need to be synced to this
+package, run sync.sh. It will not overwrite any local changes made since the
+previous sync, so it's ok to land changes in this package first, and then sync
+to the standard library later.

vendor/golang.org/x/crypto/curve25519/internal/field/fe.go

@@ -0,0 +1,416 @@
+// Copyright (c) 2017 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Package field implements fast arithmetic modulo 2^255-19.
+package field
+
+import (
+	"crypto/subtle"
+	"encoding/binary"
+	"math/bits"
+)
+
+// Element represents an element of the field GF(2^255-19). Note that this
+// is not a cryptographically secure group, and should only be used to interact
+// with edwards25519.Point coordinates.
+//
+// This type works similarly to math/big.Int, and all arguments and receivers
+// are allowed to alias.
+//
+// The zero value is a valid zero element.
+type Element struct {
+	// An element t represents the integer
+	//     t.l0 + t.l1*2^51 + t.l2*2^102 + t.l3*2^153 + t.l4*2^204
+	//
+	// Between operations, all limbs are expected to be lower than 2^52.
+	l0 uint64
+	l1 uint64
+	l2 uint64
+	l3 uint64
+	l4 uint64
+}
+
+const maskLow51Bits uint64 = (1 << 51) - 1
+
+var feZero = &Element{0, 0, 0, 0, 0}
+
+// Zero sets v = 0, and returns v.
+func (v *Element) Zero() *Element {
+	*v = *feZero
+	return v
+}
+
+var feOne = &Element{1, 0, 0, 0, 0}
+
+// One sets v = 1, and returns v.
+func (v *Element) One() *Element {
+	*v = *feOne
+	return v
+}
+
+// reduce reduces v modulo 2^255 - 19 and returns it.
+func (v *Element) reduce() *Element {
+	v.carryPropagate()
+
+	// After the light reduction we now have a field element representation
+	// v < 2^255 + 2^13 * 19, but need v < 2^255 - 19.
+
+	// If v >= 2^255 - 19, then v + 19 >= 2^255, which would overflow 2^255 - 1,
+	// generating a carry. That is, c will be 0 if v < 2^255 - 19, and 1 otherwise.
+	c := (v.l0 + 19) >> 51
+	c = (v.l1 + c) >> 51
+	c = (v.l2 + c) >> 51
+	c = (v.l3 + c) >> 51
+	c = (v.l4 + c) >> 51
+
+	// If v < 2^255 - 19 and c = 0, this will be a no-op. Otherwise, it's
+	// effectively applying the reduction identity to the carry.
+	v.l0 += 19 * c
+
+	v.l1 += v.l0 >> 51
+	v.l0 = v.l0 & maskLow51Bits
+	v.l2 += v.l1 >> 51
+	v.l1 = v.l1 & maskLow51Bits
+	v.l3 += v.l2 >> 51
+	v.l2 = v.l2 & maskLow51Bits
+	v.l4 += v.l3 >> 51
+	v.l3 = v.l3 & maskLow51Bits
+	// no additional carry
+	v.l4 = v.l4 & maskLow51Bits
+
+	return v
+}
+
+// Add sets v = a + b, and returns v.
+func (v *Element) Add(a, b *Element) *Element {
+	v.l0 = a.l0 + b.l0
+	v.l1 = a.l1 + b.l1
+	v.l2 = a.l2 + b.l2
+	v.l3 = a.l3 + b.l3
+	v.l4 = a.l4 + b.l4
+	// Using the generic implementation here is actually faster than the
+	// assembly. Probably because the body of this function is so simple that
+	// the compiler can figure out better optimizations by inlining the carry
+	// propagation. TODO
+	return v.carryPropagateGeneric()
+}
+
+// Subtract sets v = a - b, and returns v.
+func (v *Element) Subtract(a, b *Element) *Element {
+	// We first add 2 * p, to guarantee the subtraction won't underflow, and
+	// then subtract b (which can be up to 2^255 + 2^13 * 19).
+	v.l0 = (a.l0 + 0xFFFFFFFFFFFDA) - b.l0
+	v.l1 = (a.l1 + 0xFFFFFFFFFFFFE) - b.l1
+	v.l2 = (a.l2 + 0xFFFFFFFFFFFFE) - b.l2
+	v.l3 = (a.l3 + 0xFFFFFFFFFFFFE) - b.l3
+	v.l4 = (a.l4 + 0xFFFFFFFFFFFFE) - b.l4
+	return v.carryPropagate()
+}
+
+// Negate sets v = -a, and returns v.
+func (v *Element) Negate(a *Element) *Element {
+	return v.Subtract(feZero, a)
+}
+
+// Invert sets v = 1/z mod p, and returns v.
+//
+// If z == 0, Invert returns v = 0.
+func (v *Element) Invert(z *Element) *Element {
+	// Inversion is implemented as exponentiation with exponent p − 2. It uses the
+	// same sequence of 255 squarings and 11 multiplications as [Curve25519].
+	var z2, z9, z11, z2_5_0, z2_10_0, z2_20_0, z2_50_0, z2_100_0, t Element
+
+	z2.Square(z)             // 2
+	t.Square(&z2)            // 4
+	t.Square(&t)             // 8
+	z9.Multiply(&t, z)       // 9
+	z11.Multiply(&z9, &z2)   // 11
+	t.Square(&z11)           // 22
+	z2_5_0.Multiply(&t, &z9) // 31 = 2^5 - 2^0
+
+	t.Square(&z2_5_0) // 2^6 - 2^1
+	for i := 0; i < 4; i++ {
+		t.Square(&t) // 2^10 - 2^5
+	}
+	z2_10_0.Multiply(&t, &z2_5_0) // 2^10 - 2^0
+
+	t.Square(&z2_10_0) // 2^11 - 2^1
+	for i := 0; i < 9; i++ {
+		t.Square(&t) // 2^20 - 2^10
+	}
+	z2_20_0.Multiply(&t, &z2_10_0) // 2^20 - 2^0
+
+	t.Square(&z2_20_0) // 2^21 - 2^1
+	for i := 0; i < 19; i++ {
+		t.Square(&t) // 2^40 - 2^20
+	}
+	t.Multiply(&t, &z2_20_0) // 2^40 - 2^0
+
+	t.Square(&t) // 2^41 - 2^1
+	for i := 0; i < 9; i++ {
+		t.Square(&t) // 2^50 - 2^10
+	}
+	z2_50_0.Multiply(&t, &z2_10_0) // 2^50 - 2^0
+
+	t.Square(&z2_50_0) // 2^51 - 2^1
+	for i := 0; i < 49; i++ {
+		t.Square(&t) // 2^100 - 2^50
+	}
+	z2_100_0.Multiply(&t, &z2_50_0) // 2^100 - 2^0
+
+	t.Square(&z2_100_0) // 2^101 - 2^1
+	for i := 0; i < 99; i++ {
+		t.Square(&t) // 2^200 - 2^100
+	}
+	t.Multiply(&t, &z2_100_0) // 2^200 - 2^0
+
+	t.Square(&t) // 2^201 - 2^1
+	for i := 0; i < 49; i++ {
+		t.Square(&t) // 2^250 - 2^50
+	}
+	t.Multiply(&t, &z2_50_0) // 2^250 - 2^0
+
+	t.Square(&t) // 2^251 - 2^1
+	t.Square(&t) // 2^252 - 2^2
+	t.Square(&t) // 2^253 - 2^3
+	t.Square(&t) // 2^254 - 2^4
+	t.Square(&t) // 2^255 - 2^5
+
+	return v.Multiply(&t, &z11) // 2^255 - 21
+}
+
+// Set sets v = a, and returns v.
+func (v *Element) Set(a *Element) *Element {
+	*v = *a
+	return v
+}
+
+// SetBytes sets v to x, which must be a 32-byte little-endian encoding.
+//
+// Consistent with RFC 7748, the most significant bit (the high bit of the
+// last byte) is ignored, and non-canonical values (2^255-19 through 2^255-1)
+// are accepted. Note that this is laxer than specified by RFC 8032.
+func (v *Element) SetBytes(x []byte) *Element {
+	if len(x) != 32 {
+		panic("edwards25519: invalid field element input size")
+	}
+
+	// Bits 0:51 (bytes 0:8, bits 0:64, shift 0, mask 51).
+	v.l0 = binary.LittleEndian.Uint64(x[0:8])
+	v.l0 &= maskLow51Bits
+	// Bits 51:102 (bytes 6:14, bits 48:112, shift 3, mask 51).
+	v.l1 = binary.LittleEndian.Uint64(x[6:14]) >> 3
+	v.l1 &= maskLow51Bits
+	// Bits 102:153 (bytes 12:20, bits 96:160, shift 6, mask 51).
+	v.l2 = binary.LittleEndian.Uint64(x[12:20]) >> 6
+	v.l2 &= maskLow51Bits
+	// Bits 153:204 (bytes 19:27, bits 152:216, shift 1, mask 51).
+	v.l3 = binary.LittleEndian.Uint64(x[19:27]) >> 1
+	v.l3 &= maskLow51Bits
+	// Bits 204:251 (bytes 24:32, bits 192:256, shift 12, mask 51).
+	// Note: not bytes 25:33, shift 4, to avoid overread.
+	v.l4 = binary.LittleEndian.Uint64(x[24:32]) >> 12
+	v.l4 &= maskLow51Bits
+
+	return v
+}
+
+// Bytes returns the canonical 32-byte little-endian encoding of v.
+func (v *Element) Bytes() []byte {
+	// This function is outlined to make the allocations inline in the caller
+	// rather than happen on the heap.
+	var out [32]byte
+	return v.bytes(&out)
+}
+
+func (v *Element) bytes(out *[32]byte) []byte {
+	t := *v
+	t.reduce()
+
+	var buf [8]byte
+	for i, l := range [5]uint64{t.l0, t.l1, t.l2, t.l3, t.l4} {
+		bitsOffset := i * 51
+		binary.LittleEndian.PutUint64(buf[:], l<<uint(bitsOffset%8))
+		for i, bb := range buf {
+			off := bitsOffset/8 + i
+			if off >= len(out) {
+				break
+			}
+			out[off] |= bb
+		}
+	}
+
+	return out[:]
+}
+
+// Equal returns 1 if v and u are equal, and 0 otherwise.
+func (v *Element) Equal(u *Element) int {
+	sa, sv := u.Bytes(), v.Bytes()
+	return subtle.ConstantTimeCompare(sa, sv)
+}
+
+// mask64Bits returns 0xffffffff if cond is 1, and 0 otherwise.
+func mask64Bits(cond int) uint64 { return ^(uint64(cond) - 1) }
+
+// Select sets v to a if cond == 1, and to b if cond == 0.
+func (v *Element) Select(a, b *Element, cond int) *Element {
+	m := mask64Bits(cond)
+	v.l0 = (m & a.l0) | (^m & b.l0)
+	v.l1 = (m & a.l1) | (^m & b.l1)
+	v.l2 = (m & a.l2) | (^m & b.l2)
+	v.l3 = (m & a.l3) | (^m & b.l3)
+	v.l4 = (m & a.l4) | (^m & b.l4)
+	return v
+}
+
+// Swap swaps v and u if cond == 1 or leaves them unchanged if cond == 0, and returns v.
+func (v *Element) Swap(u *Element, cond int) {
+	m := mask64Bits(cond)
+	t := m & (v.l0 ^ u.l0)
+	v.l0 ^= t
+	u.l0 ^= t
+	t = m & (v.l1 ^ u.l1)
+	v.l1 ^= t
+	u.l1 ^= t
+	t = m & (v.l2 ^ u.l2)
+	v.l2 ^= t
+	u.l2 ^= t
+	t = m & (v.l3 ^ u.l3)
+	v.l3 ^= t
+	u.l3 ^= t
+	t = m & (v.l4 ^ u.l4)
+	v.l4 ^= t
+	u.l4 ^= t
+}
+
+// IsNegative returns 1 if v is negative, and 0 otherwise.
+func (v *Element) IsNegative() int {
+	return int(v.Bytes()[0] & 1)
+}
+
+// Absolute sets v to |u|, and returns v.
+func (v *Element) Absolute(u *Element) *Element {
+	return v.Select(new(Element).Negate(u), u, u.IsNegative())
+}
+
+// Multiply sets v = x * y, and returns v.
+func (v *Element) Multiply(x, y *Element) *Element {
+	feMul(v, x, y)
+	return v
+}
+
+// Square sets v = x * x, and returns v.
+func (v *Element) Square(x *Element) *Element {
+	feSquare(v, x)
+	return v
+}
+
+// Mult32 sets v = x * y, and returns v.
+func (v *Element) Mult32(x *Element, y uint32) *Element {
+	x0lo, x0hi := mul51(x.l0, y)
+	x1lo, x1hi := mul51(x.l1, y)
+	x2lo, x2hi := mul51(x.l2, y)
+	x3lo, x3hi := mul51(x.l3, y)
+	x4lo, x4hi := mul51(x.l4, y)
+	v.l0 = x0lo + 19*x4hi // carried over per the reduction identity
+	v.l1 = x1lo + x0hi
+	v.l2 = x2lo + x1hi
+	v.l3 = x3lo + x2hi
+	v.l4 = x4lo + x3hi
+	// The hi portions are going to be only 32 bits, plus any previous excess,
+	// so we can skip the carry propagation.
+	return v
+}
+
+// mul51 returns lo + hi * 2⁵¹ = a * b.
+func mul51(a uint64, b uint32) (lo uint64, hi uint64) {
+	mh, ml := bits.Mul64(a, uint64(b))
+	lo = ml & maskLow51Bits
+	hi = (mh << 13) | (ml >> 51)
+	return
+}
+
+// Pow22523 set v = x^((p-5)/8), and returns v. (p-5)/8 is 2^252-3.
+func (v *Element) Pow22523(x *Element) *Element {
+	var t0, t1, t2 Element
+
+	t0.Square(x)             // x^2
+	t1.Square(&t0)           // x^4
+	t1.Square(&t1)           // x^8
+	t1.Multiply(x, &t1)      // x^9
+	t0.Multiply(&t0, &t1)    // x^11
+	t0.Square(&t0)           // x^22
+	t0.Multiply(&t1, &t0)    // x^31
+	t1.Square(&t0)           // x^62
+	for i := 1; i < 5; i++ { // x^992
+		t1.Square(&t1)
+	}
+	t0.Multiply(&t1, &t0)     // x^1023 -> 1023 = 2^10 - 1
+	t1.Square(&t0)            // 2^11 - 2
+	for i := 1; i < 10; i++ { // 2^20 - 2^10
+		t1.Square(&t1)
+	}
+	t1.Multiply(&t1, &t0)     // 2^20 - 1
+	t2.Square(&t1)            // 2^21 - 2
+	for i := 1; i < 20; i++ { // 2^40 - 2^20
+		t2.Square(&t2)
+	}
+	t1.Multiply(&t2, &t1)     // 2^40 - 1
+	t1.Square(&t1)            // 2^41 - 2
+	for i := 1; i < 10; i++ { // 2^50 - 2^10
+		t1.Square(&t1)
+	}
+	t0.Multiply(&t1, &t0)     // 2^50 - 1
+	t1.Square(&t0)            // 2^51 - 2
+	for i := 1; i < 50; i++ { // 2^100 - 2^50
+		t1.Square(&t1)
+	}
+	t1.Multiply(&t1, &t0)      // 2^100 - 1
+	t2.Square(&t1)             // 2^101 - 2
+	for i := 1; i < 100; i++ { // 2^200 - 2^100
+		t2.Square(&t2)
+	}
+	t1.Multiply(&t2, &t1)     // 2^200 - 1
+	t1.Square(&t1)            // 2^201 - 2
+	for i := 1; i < 50; i++ { // 2^250 - 2^50
+		t1.Square(&t1)
+	}
+	t0.Multiply(&t1, &t0)     // 2^250 - 1
+	t0.Square(&t0)            // 2^251 - 2
+	t0.Square(&t0)            // 2^252 - 4
+	return v.Multiply(&t0, x) // 2^252 - 3 -> x^(2^252-3)
+}
+
+// sqrtM1 is 2^((p-1)/4), which squared is equal to -1 by Euler's Criterion.
+var sqrtM1 = &Element{1718705420411056, 234908883556509,
+	2233514472574048, 2117202627021982, 765476049583133}
+
+// SqrtRatio sets r to the non-negative square root of the ratio of u and v.
+//
+// If u/v is square, SqrtRatio returns r and 1. If u/v is not square, SqrtRatio
+// sets r according to Section 4.3 of draft-irtf-cfrg-ristretto255-decaf448-00,
+// and returns r and 0.
+func (r *Element) SqrtRatio(u, v *Element) (rr *Element, wasSquare int) {
+	var a, b Element
+
+	// r = (u * v3) * (u * v7)^((p-5)/8)
+	v2 := a.Square(v)
+	uv3 := b.Multiply(u, b.Multiply(v2, v))
+	uv7 := a.Multiply(uv3, a.Square(v2))
+	r.Multiply(uv3, r.Pow22523(uv7))
+
+	check := a.Multiply(v, a.Square(r)) // check = v * r^2
+
+	uNeg := b.Negate(u)
+	correctSignSqrt := check.Equal(u)
+	flippedSignSqrt := check.Equal(uNeg)
+	flippedSignSqrtI := check.Equal(uNeg.Multiply(uNeg, sqrtM1))
+
+	rPrime := b.Multiply(r, sqrtM1) // r_prime = SQRT_M1 * r
+	// r = CT_SELECT(r_prime IF flipped_sign_sqrt | flipped_sign_sqrt_i ELSE r)
+	r.Select(rPrime, r, flippedSignSqrt|flippedSignSqrtI)
+
+	r.Absolute(r) // Choose the nonnegative square root.
+	return r, correctSignSqrt | flippedSignSqrt
+}

vendor/golang.org/x/crypto/curve25519/internal/field/fe_amd64.go

@@ -0,0 +1,13 @@
+// Code generated by command: go run fe_amd64_asm.go -out ../fe_amd64.s -stubs ../fe_amd64.go -pkg field. DO NOT EDIT.
+
+// +build amd64,gc,!purego
+
+package field
+
+// feMul sets out = a * b. It works like feMulGeneric.
+//go:noescape
+func feMul(out *Element, a *Element, b *Element)
+
+// feSquare sets out = a * a. It works like feSquareGeneric.
+//go:noescape
+func feSquare(out *Element, a *Element)

vendor/golang.org/x/crypto/curve25519/internal/field/fe_amd64.s

@@ -0,0 +1,379 @@
+// Code generated by command: go run fe_amd64_asm.go -out ../fe_amd64.s -stubs ../fe_amd64.go -pkg field. DO NOT EDIT.
+
+//go:build amd64 && gc && !purego
+// +build amd64,gc,!purego
+
+#include "textflag.h"
+
+// func feMul(out *Element, a *Element, b *Element)
+TEXT ·feMul(SB), NOSPLIT, $0-24
+	MOVQ a+8(FP), CX
+	MOVQ b+16(FP), BX
+
+	// r0 = a0×b0
+	MOVQ (CX), AX
+	MULQ (BX)
+	MOVQ AX, DI
+	MOVQ DX, SI
+
+	// r0 += 19×a1×b4
+	MOVQ   8(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   32(BX)
+	ADDQ   AX, DI
+	ADCQ   DX, SI
+
+	// r0 += 19×a2×b3
+	MOVQ   16(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   24(BX)
+	ADDQ   AX, DI
+	ADCQ   DX, SI
+
+	// r0 += 19×a3×b2
+	MOVQ   24(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   16(BX)
+	ADDQ   AX, DI
+	ADCQ   DX, SI
+
+	// r0 += 19×a4×b1
+	MOVQ   32(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   8(BX)
+	ADDQ   AX, DI
+	ADCQ   DX, SI
+
+	// r1 = a0×b1
+	MOVQ (CX), AX
+	MULQ 8(BX)
+	MOVQ AX, R9
+	MOVQ DX, R8
+
+	// r1 += a1×b0
+	MOVQ 8(CX), AX
+	MULQ (BX)
+	ADDQ AX, R9
+	ADCQ DX, R8
+
+	// r1 += 19×a2×b4
+	MOVQ   16(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   32(BX)
+	ADDQ   AX, R9
+	ADCQ   DX, R8
+
+	// r1 += 19×a3×b3
+	MOVQ   24(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   24(BX)
+	ADDQ   AX, R9
+	ADCQ   DX, R8
+
+	// r1 += 19×a4×b2
+	MOVQ   32(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   16(BX)
+	ADDQ   AX, R9
+	ADCQ   DX, R8
+
+	// r2 = a0×b2
+	MOVQ (CX), AX
+	MULQ 16(BX)
+	MOVQ AX, R11
+	MOVQ DX, R10
+
+	// r2 += a1×b1
+	MOVQ 8(CX), AX
+	MULQ 8(BX)
+	ADDQ AX, R11
+	ADCQ DX, R10
+
+	// r2 += a2×b0
+	MOVQ 16(CX), AX
+	MULQ (BX)
+	ADDQ AX, R11
+	ADCQ DX, R10
+
+	// r2 += 19×a3×b4
+	MOVQ   24(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   32(BX)
+	ADDQ   AX, R11
+	ADCQ   DX, R10
+
+	// r2 += 19×a4×b3
+	MOVQ   32(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   24(BX)
+	ADDQ   AX, R11
+	ADCQ   DX, R10
+
+	// r3 = a0×b3
+	MOVQ (CX), AX
+	MULQ 24(BX)
+	MOVQ AX, R13
+	MOVQ DX, R12
+
+	// r3 += a1×b2
+	MOVQ 8(CX), AX
+	MULQ 16(BX)
+	ADDQ AX, R13
+	ADCQ DX, R12
+
+	// r3 += a2×b1
+	MOVQ 16(CX), AX
+	MULQ 8(BX)
+	ADDQ AX, R13
+	ADCQ DX, R12
+
+	// r3 += a3×b0
+	MOVQ 24(CX), AX
+	MULQ (BX)
+	ADDQ AX, R13
+	ADCQ DX, R12
+
+	// r3 += 19×a4×b4
+	MOVQ   32(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   32(BX)
+	ADDQ   AX, R13
+	ADCQ   DX, R12
+
+	// r4 = a0×b4
+	MOVQ (CX), AX
+	MULQ 32(BX)
+	MOVQ AX, R15
+	MOVQ DX, R14
+
+	// r4 += a1×b3
+	MOVQ 8(CX), AX
+	MULQ 24(BX)
+	ADDQ AX, R15
+	ADCQ DX, R14
+
+	// r4 += a2×b2
+	MOVQ 16(CX), AX
+	MULQ 16(BX)
+	ADDQ AX, R15
+	ADCQ DX, R14
+
+	// r4 += a3×b1
+	MOVQ 24(CX), AX
+	MULQ 8(BX)
+	ADDQ AX, R15
+	ADCQ DX, R14
+
+	// r4 += a4×b0
+	MOVQ 32(CX), AX
+	MULQ (BX)
+	ADDQ AX, R15
+	ADCQ DX, R14
+
+	// First reduction chain
+	MOVQ   $0x0007ffffffffffff, AX
+	SHLQ   $0x0d, DI, SI
+	SHLQ   $0x0d, R9, R8
+	SHLQ   $0x0d, R11, R10
+	SHLQ   $0x0d, R13, R12
+	SHLQ   $0x0d, R15, R14
+	ANDQ   AX, DI
+	IMUL3Q $0x13, R14, R14
+	ADDQ   R14, DI
+	ANDQ   AX, R9
+	ADDQ   SI, R9
+	ANDQ   AX, R11
+	ADDQ   R8, R11
+	ANDQ   AX, R13
+	ADDQ   R10, R13
+	ANDQ   AX, R15
+	ADDQ   R12, R15
+
+	// Second reduction chain (carryPropagate)
+	MOVQ   DI, SI
+	SHRQ   $0x33, SI
+	MOVQ   R9, R8
+	SHRQ   $0x33, R8
+	MOVQ   R11, R10
+	SHRQ   $0x33, R10
+	MOVQ   R13, R12
+	SHRQ   $0x33, R12
+	MOVQ   R15, R14
+	SHRQ   $0x33, R14
+	ANDQ   AX, DI
+	IMUL3Q $0x13, R14, R14
+	ADDQ   R14, DI
+	ANDQ   AX, R9
+	ADDQ   SI, R9
+	ANDQ   AX, R11
+	ADDQ   R8, R11
+	ANDQ   AX, R13
+	ADDQ   R10, R13
+	ANDQ   AX, R15
+	ADDQ   R12, R15
+
+	// Store output
+	MOVQ out+0(FP), AX
+	MOVQ DI, (AX)
+	MOVQ R9, 8(AX)
+	MOVQ R11, 16(AX)
+	MOVQ R13, 24(AX)
+	MOVQ R15, 32(AX)
+	RET
+
+// func feSquare(out *Element, a *Element)
+TEXT ·feSquare(SB), NOSPLIT, $0-16
+	MOVQ a+8(FP), CX
+
+	// r0 = l0×l0
+	MOVQ (CX), AX
+	MULQ (CX)
+	MOVQ AX, SI
+	MOVQ DX, BX
+
+	// r0 += 38×l1×l4
+	MOVQ   8(CX), AX
+	IMUL3Q $0x26, AX, AX
+	MULQ   32(CX)
+	ADDQ   AX, SI
+	ADCQ   DX, BX
+
+	// r0 += 38×l2×l3
+	MOVQ   16(CX), AX
+	IMUL3Q $0x26, AX, AX
+	MULQ   24(CX)
+	ADDQ   AX, SI
+	ADCQ   DX, BX
+
+	// r1 = 2×l0×l1
+	MOVQ (CX), AX
+	SHLQ $0x01, AX
+	MULQ 8(CX)
+	MOVQ AX, R8
+	MOVQ DX, DI
+
+	// r1 += 38×l2×l4
+	MOVQ   16(CX), AX
+	IMUL3Q $0x26, AX, AX
+	MULQ   32(CX)
+	ADDQ   AX, R8
+	ADCQ   DX, DI
+
+	// r1 += 19×l3×l3
+	MOVQ   24(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   24(CX)
+	ADDQ   AX, R8
+	ADCQ   DX, DI
+
+	// r2 = 2×l0×l2
+	MOVQ (CX), AX
+	SHLQ $0x01, AX
+	MULQ 16(CX)
+	MOVQ AX, R10
+	MOVQ DX, R9
+
+	// r2 += l1×l1
+	MOVQ 8(CX), AX
+	MULQ 8(CX)
+	ADDQ AX, R10
+	ADCQ DX, R9
+
+	// r2 += 38×l3×l4
+	MOVQ   24(CX), AX
+	IMUL3Q $0x26, AX, AX
+	MULQ   32(CX)
+	ADDQ   AX, R10
+	ADCQ   DX, R9
+
+	// r3 = 2×l0×l3
+	MOVQ (CX), AX
+	SHLQ $0x01, AX
+	MULQ 24(CX)
+	MOVQ AX, R12
+	MOVQ DX, R11
+
+	// r3 += 2×l1×l2
+	MOVQ   8(CX), AX
+	IMUL3Q $0x02, AX, AX
+	MULQ   16(CX)
+	ADDQ   AX, R12
+	ADCQ   DX, R11
+
+	// r3 += 19×l4×l4
+	MOVQ   32(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   32(CX)
+	ADDQ   AX, R12
+	ADCQ   DX, R11
+
+	// r4 = 2×l0×l4
+	MOVQ (CX), AX
+	SHLQ $0x01, AX
+	MULQ 32(CX)
+	MOVQ AX, R14
+	MOVQ DX, R13
+
+	// r4 += 2×l1×l3
+	MOVQ   8(CX), AX
+	IMUL3Q $0x02, AX, AX
+	MULQ   24(CX)
+	ADDQ   AX, R14
+	ADCQ   DX, R13
+
+	// r4 += l2×l2
+	MOVQ 16(CX), AX
+	MULQ 16(CX)
+	ADDQ AX, R14
+	ADCQ DX, R13
+
+	// First reduction chain
+	MOVQ   $0x0007ffffffffffff, AX
+	SHLQ   $0x0d, SI, BX
+	SHLQ   $0x0d, R8, DI
+	SHLQ   $0x0d, R10, R9
+	SHLQ   $0x0d, R12, R11
+	SHLQ   $0x0d, R14, R13
+	ANDQ   AX, SI
+	IMUL3Q $0x13, R13, R13
+	ADDQ   R13, SI
+	ANDQ   AX, R8
+	ADDQ   BX, R8
+	ANDQ   AX, R10
+	ADDQ   DI, R10
+	ANDQ   AX, R12
+	ADDQ   R9, R12
+	ANDQ   AX, R14
+	ADDQ   R11, R14
+
+	// Second reduction chain (carryPropagate)
+	MOVQ   SI, BX
+	SHRQ   $0x33, BX
+	MOVQ   R8, DI
+	SHRQ   $0x33, DI
+	MOVQ   R10, R9
+	SHRQ   $0x33, R9
+	MOVQ   R12, R11
+	SHRQ   $0x33, R11
+	MOVQ   R14, R13
+	SHRQ   $0x33, R13
+	ANDQ   AX, SI
+	IMUL3Q $0x13, R13, R13
+	ADDQ   R13, SI
+	ANDQ   AX, R8
+	ADDQ   BX, R8
+	ANDQ   AX, R10
+	ADDQ   DI, R10
+	ANDQ   AX, R12
+	ADDQ   R9, R12
+	ANDQ   AX, R14
+	ADDQ   R11, R14
+
+	// Store output
+	MOVQ out+0(FP), AX
+	MOVQ SI, (AX)
+	MOVQ R8, 8(AX)
+	MOVQ R10, 16(AX)
+	MOVQ R12, 24(AX)
+	MOVQ R14, 32(AX)
+	RET

vendor/golang.org/x/crypto/curve25519/internal/field/fe_amd64_noasm.go

@@ -0,0 +1,12 @@
+// Copyright (c) 2019 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build !amd64 || !gc || purego
+// +build !amd64 !gc purego
+
+package field
+
+func feMul(v, x, y *Element) { feMulGeneric(v, x, y) }
+
+func feSquare(v, x *Element) { feSquareGeneric(v, x) }

vendor/golang.org/x/crypto/curve25519/internal/field/fe_arm64.go

@@ -0,0 +1,16 @@
+// Copyright (c) 2020 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build arm64 && gc && !purego
+// +build arm64,gc,!purego
+
+package field
+
+//go:noescape
+func carryPropagate(v *Element)
+
+func (v *Element) carryPropagate() *Element {
+	carryPropagate(v)
+	return v
+}

vendor/golang.org/x/crypto/curve25519/internal/field/fe_arm64.s

@@ -0,0 +1,43 @@
+// Copyright (c) 2020 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build arm64 && gc && !purego
+// +build arm64,gc,!purego
+
+#include "textflag.h"
+
+// carryPropagate works exactly like carryPropagateGeneric and uses the
+// same AND, ADD, and LSR+MADD instructions emitted by the compiler, but
+// avoids loading R0-R4 twice and uses LDP and STP.
+//
+// See https://golang.org/issues/43145 for the main compiler issue.
+//
+// func carryPropagate(v *Element)
+TEXT ·carryPropagate(SB),NOFRAME|NOSPLIT,$0-8
+	MOVD v+0(FP), R20
+
+	LDP 0(R20), (R0, R1)
+	LDP 16(R20), (R2, R3)
+	MOVD 32(R20), R4
+
+	AND $0x7ffffffffffff, R0, R10
+	AND $0x7ffffffffffff, R1, R11
+	AND $0x7ffffffffffff, R2, R12
+	AND $0x7ffffffffffff, R3, R13
+	AND $0x7ffffffffffff, R4, R14
+
+	ADD R0>>51, R11, R11
+	ADD R1>>51, R12, R12
+	ADD R2>>51, R13, R13
+	ADD R3>>51, R14, R14
+	// R4>>51 * 19 + R10 -> R10
+	LSR $51, R4, R21
+	MOVD $19, R22
+	MADD R22, R10, R21, R10
+
+	STP (R10, R11), 0(R20)
+	STP (R12, R13), 16(R20)
+	MOVD R14, 32(R20)
+
+	RET

vendor/golang.org/x/crypto/curve25519/internal/field/fe_arm64_noasm.go

@@ -0,0 +1,12 @@
+// Copyright (c) 2021 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build !arm64 || !gc || purego
+// +build !arm64 !gc purego
+
+package field
+
+func (v *Element) carryPropagate() *Element {
+	return v.carryPropagateGeneric()
+}

vendor/golang.org/x/crypto/curve25519/internal/field/fe_generic.go

@@ -0,0 +1,264 @@
+// Copyright (c) 2017 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package field
+
+import "math/bits"
+
+// uint128 holds a 128-bit number as two 64-bit limbs, for use with the
+// bits.Mul64 and bits.Add64 intrinsics.
+type uint128 struct {
+	lo, hi uint64
+}
+
+// mul64 returns a * b.
+func mul64(a, b uint64) uint128 {
+	hi, lo := bits.Mul64(a, b)
+	return uint128{lo, hi}
+}
+
+// addMul64 returns v + a * b.
+func addMul64(v uint128, a, b uint64) uint128 {
+	hi, lo := bits.Mul64(a, b)
+	lo, c := bits.Add64(lo, v.lo, 0)
+	hi, _ = bits.Add64(hi, v.hi, c)
+	return uint128{lo, hi}
+}
+
+// shiftRightBy51 returns a >> 51. a is assumed to be at most 115 bits.
+func shiftRightBy51(a uint128) uint64 {
+	return (a.hi << (64 - 51)) | (a.lo >> 51)
+}
+
+func feMulGeneric(v, a, b *Element) {
+	a0 := a.l0
+	a1 := a.l1
+	a2 := a.l2
+	a3 := a.l3
+	a4 := a.l4
+
+	b0 := b.l0
+	b1 := b.l1
+	b2 := b.l2
+	b3 := b.l3
+	b4 := b.l4
+
+	// Limb multiplication works like pen-and-paper columnar multiplication, but
+	// with 51-bit limbs instead of digits.
+	//
+	//                          a4   a3   a2   a1   a0  x
+	//                          b4   b3   b2   b1   b0  =
+	//                         ------------------------
+	//                        a4b0 a3b0 a2b0 a1b0 a0b0  +
+	//                   a4b1 a3b1 a2b1 a1b1 a0b1       +
+	//              a4b2 a3b2 a2b2 a1b2 a0b2            +
+	//         a4b3 a3b3 a2b3 a1b3 a0b3                 +
+	//    a4b4 a3b4 a2b4 a1b4 a0b4                      =
+	//   ----------------------------------------------
+	//      r8   r7   r6   r5   r4   r3   r2   r1   r0
+	//
+	// We can then use the reduction identity (a * 2²⁵⁵ + b = a * 19 + b) to
+	// reduce the limbs that would overflow 255 bits. r5 * 2²⁵⁵ becomes 19 * r5,
+	// r6 * 2³⁰⁶ becomes 19 * r6 * 2⁵¹, etc.
+	//
+	// Reduction can be carried out simultaneously to multiplication. For
+	// example, we do not compute r5: whenever the result of a multiplication
+	// belongs to r5, like a1b4, we multiply it by 19 and add the result to r0.
+	//
+	//            a4b0    a3b0    a2b0    a1b0    a0b0  +
+	//            a3b1    a2b1    a1b1    a0b1 19×a4b1  +
+	//            a2b2    a1b2    a0b2 19×a4b2 19×a3b2  +
+	//            a1b3    a0b3 19×a4b3 19×a3b3 19×a2b3  +
+	//            a0b4 19×a4b4 19×a3b4 19×a2b4 19×a1b4  =
+	//           --------------------------------------
+	//              r4      r3      r2      r1      r0
+	//
+	// Finally we add up the columns into wide, overlapping limbs.
+
+	a1_19 := a1 * 19
+	a2_19 := a2 * 19
+	a3_19 := a3 * 19
+	a4_19 := a4 * 19
+
+	// r0 = a0×b0 + 19×(a1×b4 + a2×b3 + a3×b2 + a4×b1)
+	r0 := mul64(a0, b0)
+	r0 = addMul64(r0, a1_19, b4)
+	r0 = addMul64(r0, a2_19, b3)
+	r0 = addMul64(r0, a3_19, b2)
+	r0 = addMul64(r0, a4_19, b1)
+
+	// r1 = a0×b1 + a1×b0 + 19×(a2×b4 + a3×b3 + a4×b2)
+	r1 := mul64(a0, b1)
+	r1 = addMul64(r1, a1, b0)
+	r1 = addMul64(r1, a2_19, b4)
+	r1 = addMul64(r1, a3_19, b3)
+	r1 = addMul64(r1, a4_19, b2)
+
+	// r2 = a0×b2 + a1×b1 + a2×b0 + 19×(a3×b4 + a4×b3)
+	r2 := mul64(a0, b2)
+	r2 = addMul64(r2, a1, b1)
+	r2 = addMul64(r2, a2, b0)
+	r2 = addMul64(r2, a3_19, b4)
+	r2 = addMul64(r2, a4_19, b3)
+
+	// r3 = a0×b3 + a1×b2 + a2×b1 + a3×b0 + 19×a4×b4
+	r3 := mul64(a0, b3)
+	r3 = addMul64(r3, a1, b2)
+	r3 = addMul64(r3, a2, b1)
+	r3 = addMul64(r3, a3, b0)
+	r3 = addMul64(r3, a4_19, b4)
+
+	// r4 = a0×b4 + a1×b3 + a2×b2 + a3×b1 + a4×b0
+	r4 := mul64(a0, b4)
+	r4 = addMul64(r4, a1, b3)
+	r4 = addMul64(r4, a2, b2)
+	r4 = addMul64(r4, a3, b1)
+	r4 = addMul64(r4, a4, b0)
+
+	// After the multiplication, we need to reduce (carry) the five coefficients
+	// to obtain a result with limbs that are at most slightly larger than 2⁵¹,
+	// to respect the Element invariant.
+	//
+	// Overall, the reduction works the same as carryPropagate, except with
+	// wider inputs: we take the carry for each coefficient by shifting it right
+	// by 51, and add it to the limb above it. The top carry is multiplied by 19
+	// according to the reduction identity and added to the lowest limb.
+	//
+	// The largest coefficient (r0) will be at most 111 bits, which guarantees
+	// that all carries are at most 111 - 51 = 60 bits, which fits in a uint64.
+	//
+	//     r0 = a0×b0 + 19×(a1×b4 + a2×b3 + a3×b2 + a4×b1)
+	//     r0 < 2⁵²×2⁵² + 19×(2⁵²×2⁵² + 2⁵²×2⁵² + 2⁵²×2⁵² + 2⁵²×2⁵²)
+	//     r0 < (1 + 19 × 4) × 2⁵² × 2⁵²
+	//     r0 < 2⁷ × 2⁵² × 2⁵²
+	//     r0 < 2¹¹¹
+	//
+	// Moreover, the top coefficient (r4) is at most 107 bits, so c4 is at most
+	// 56 bits, and c4 * 19 is at most 61 bits, which again fits in a uint64 and
+	// allows us to easily apply the reduction identity.
+	//
+	//     r4 = a0×b4 + a1×b3 + a2×b2 + a3×b1 + a4×b0
+	//     r4 < 5 × 2⁵² × 2⁵²
+	//     r4 < 2¹⁰⁷
+	//
+
+	c0 := shiftRightBy51(r0)
+	c1 := shiftRightBy51(r1)
+	c2 := shiftRightBy51(r2)
+	c3 := shiftRightBy51(r3)
+	c4 := shiftRightBy51(r4)
+
+	rr0 := r0.lo&maskLow51Bits + c4*19
+	rr1 := r1.lo&maskLow51Bits + c0
+	rr2 := r2.lo&maskLow51Bits + c1
+	rr3 := r3.lo&maskLow51Bits + c2
+	rr4 := r4.lo&maskLow51Bits + c3
+
+	// Now all coefficients fit into 64-bit registers but are still too large to
+	// be passed around as a Element. We therefore do one last carry chain,
+	// where the carries will be small enough to fit in the wiggle room above 2⁵¹.
+	*v = Element{rr0, rr1, rr2, rr3, rr4}
+	v.carryPropagate()
+}
+
+func feSquareGeneric(v, a *Element) {
+	l0 := a.l0
+	l1 := a.l1
+	l2 := a.l2
+	l3 := a.l3
+	l4 := a.l4
+
+	// Squaring works precisely like multiplication above, but thanks to its
+	// symmetry we get to group a few terms together.
+	//
+	//                          l4   l3   l2   l1   l0  x
+	//                          l4   l3   l2   l1   l0  =
+	//                         ------------------------
+	//                        l4l0 l3l0 l2l0 l1l0 l0l0  +
+	//                   l4l1 l3l1 l2l1 l1l1 l0l1       +
+	//              l4l2 l3l2 l2l2 l1l2 l0l2            +
+	//         l4l3 l3l3 l2l3 l1l3 l0l3                 +
+	//    l4l4 l3l4 l2l4 l1l4 l0l4                      =
+	//   ----------------------------------------------
+	//      r8   r7   r6   r5   r4   r3   r2   r1   r0
+	//
+	//            l4l0    l3l0    l2l0    l1l0    l0l0  +
+	//            l3l1    l2l1    l1l1    l0l1 19×l4l1  +
+	//            l2l2    l1l2    l0l2 19×l4l2 19×l3l2  +
+	//            l1l3    l0l3 19×l4l3 19×l3l3 19×l2l3  +
+	//            l0l4 19×l4l4 19×l3l4 19×l2l4 19×l1l4  =
+	//           --------------------------------------
+	//              r4      r3      r2      r1      r0
+	//
+	// With precomputed 2×, 19×, and 2×19× terms, we can compute each limb with
+	// only three Mul64 and four Add64, instead of five and eight.
+
+	l0_2 := l0 * 2
+	l1_2 := l1 * 2
+
+	l1_38 := l1 * 38
+	l2_38 := l2 * 38
+	l3_38 := l3 * 38
+
+	l3_19 := l3 * 19
+	l4_19 := l4 * 19
+
+	// r0 = l0×l0 + 19×(l1×l4 + l2×l3 + l3×l2 + l4×l1) = l0×l0 + 19×2×(l1×l4 + l2×l3)
+	r0 := mul64(l0, l0)
+	r0 = addMul64(r0, l1_38, l4)
+	r0 = addMul64(r0, l2_38, l3)
+
+	// r1 = l0×l1 + l1×l0 + 19×(l2×l4 + l3×l3 + l4×l2) = 2×l0×l1 + 19×2×l2×l4 + 19×l3×l3
+	r1 := mul64(l0_2, l1)
+	r1 = addMul64(r1, l2_38, l4)
+	r1 = addMul64(r1, l3_19, l3)
+
+	// r2 = l0×l2 + l1×l1 + l2×l0 + 19×(l3×l4 + l4×l3) = 2×l0×l2 + l1×l1 + 19×2×l3×l4
+	r2 := mul64(l0_2, l2)
+	r2 = addMul64(r2, l1, l1)
+	r2 = addMul64(r2, l3_38, l4)
+
+	// r3 = l0×l3 + l1×l2 + l2×l1 + l3×l0 + 19×l4×l4 = 2×l0×l3 + 2×l1×l2 + 19×l4×l4
+	r3 := mul64(l0_2, l3)
+	r3 = addMul64(r3, l1_2, l2)
+	r3 = addMul64(r3, l4_19, l4)
+
+	// r4 = l0×l4 + l1×l3 + l2×l2 + l3×l1 + l4×l0 = 2×l0×l4 + 2×l1×l3 + l2×l2
+	r4 := mul64(l0_2, l4)
+	r4 = addMul64(r4, l1_2, l3)
+	r4 = addMul64(r4, l2, l2)
+
+	c0 := shiftRightBy51(r0)
+	c1 := shiftRightBy51(r1)
+	c2 := shiftRightBy51(r2)
+	c3 := shiftRightBy51(r3)
+	c4 := shiftRightBy51(r4)
+
+	rr0 := r0.lo&maskLow51Bits + c4*19
+	rr1 := r1.lo&maskLow51Bits + c0
+	rr2 := r2.lo&maskLow51Bits + c1
+	rr3 := r3.lo&maskLow51Bits + c2
+	rr4 := r4.lo&maskLow51Bits + c3
+
+	*v = Element{rr0, rr1, rr2, rr3, rr4}
+	v.carryPropagate()
+}
+
+// carryPropagate brings the limbs below 52 bits by applying the reduction
+// identity (a * 2²⁵⁵ + b = a * 19 + b) to the l4 carry. TODO inline
+func (v *Element) carryPropagateGeneric() *Element {
+	c0 := v.l0 >> 51
+	c1 := v.l1 >> 51
+	c2 := v.l2 >> 51
+	c3 := v.l3 >> 51
+	c4 := v.l4 >> 51
+
+	v.l0 = v.l0&maskLow51Bits + c4*19
+	v.l1 = v.l1&maskLow51Bits + c0
+	v.l2 = v.l2&maskLow51Bits + c1
+	v.l3 = v.l3&maskLow51Bits + c2
+	v.l4 = v.l4&maskLow51Bits + c3
+
+	return v
+}

vendor/golang.org/x/crypto/curve25519/internal/field/sync.checkpoint

@@ -0,0 +1 @@
+b0c49ae9f59d233526f8934262c5bbbe14d4358d

vendor/golang.org/x/crypto/curve25519/internal/field/sync.sh

@@ -0,0 +1,19 @@
+#! /bin/bash
+set -euo pipefail
+
+cd "$(git rev-parse --show-toplevel)"
+
+STD_PATH=src/crypto/ed25519/internal/edwards25519/field
+LOCAL_PATH=curve25519/internal/field
+LAST_SYNC_REF=$(cat $LOCAL_PATH/sync.checkpoint)
+
+git fetch https://go.googlesource.com/go master
+
+if git diff --quiet $LAST_SYNC_REF:$STD_PATH FETCH_HEAD:$STD_PATH; then
+    echo "No changes."
+else
+    NEW_REF=$(git rev-parse FETCH_HEAD | tee $LOCAL_PATH/sync.checkpoint)
+    echo "Applying changes from $LAST_SYNC_REF to $NEW_REF..."
+    git diff $LAST_SYNC_REF:$STD_PATH FETCH_HEAD:$STD_PATH | \
+        git apply -3 --directory=$LOCAL_PATH
+fi

vendor/golang.org/x/crypto/ed25519/ed25519.go

@@ -0,0 +1,223 @@
+// 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.
+
+// 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.
+//
+//go:build !go1.13
+// +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”.
+package ed25519
+
+// This code is a port of the public domain, “ref10” implementation of ed25519
+// from SUPERCOP.
+
+import (
+	"bytes"
+	"crypto"
+	cryptorand "crypto/rand"
+	"crypto/sha512"
+	"errors"
+	"io"
+	"strconv"
+
+	"golang.org/x/crypto/ed25519/internal/edwards25519"
+)
+
+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.
+type PublicKey []byte
+
+// PrivateKey is the type of Ed25519 private keys. It implements crypto.Signer.
+type PrivateKey []byte
+
+// Public returns the PublicKey corresponding to priv.
+func (priv PrivateKey) Public() crypto.PublicKey {
+	publicKey := make([]byte, PublicKeySize)
+	copy(publicKey, priv[32:])
+	return PublicKey(publicKey)
+}
+
+// Seed returns the private key seed corresponding to priv. It is provided for
+// interoperability with RFC 8032. RFC 8032's private keys correspond to seeds
+// in this package.
+func (priv PrivateKey) Seed() []byte {
+	seed := make([]byte, SeedSize)
+	copy(seed, priv[:32])
+	return seed
+}
+
+// Sign signs the given message with priv.
+// Ed25519 performs two passes over messages to be signed and therefore cannot
+// handle pre-hashed messages. Thus opts.HashFunc() must return zero to
+// indicate the message hasn't been hashed. This can be achieved by passing
+// crypto.Hash(0) as the value for opts.
+func (priv PrivateKey) Sign(rand io.Reader, message []byte, opts crypto.SignerOpts) (signature []byte, err error) {
+	if opts.HashFunc() != crypto.Hash(0) {
+		return nil, errors.New("ed25519: cannot sign hashed message")
+	}
+
+	return Sign(priv, message), nil
+}
+
+// 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) {
+	if rand == nil {
+		rand = cryptorand.Reader
+	}
+
+	seed := make([]byte, SeedSize)
+	if _, err := io.ReadFull(rand, seed); err != nil {
+		return nil, nil, err
+	}
+
+	privateKey := NewKeyFromSeed(seed)
+	publicKey := make([]byte, PublicKeySize)
+	copy(publicKey, privateKey[32:])
+
+	return publicKey, privateKey, nil
+}
+
+// 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 {
+	if l := len(seed); l != SeedSize {
+		panic("ed25519: bad seed length: " + strconv.Itoa(l))
+	}
+
+	digest := sha512.Sum512(seed)
+	digest[0] &= 248
+	digest[31] &= 127
+	digest[31] |= 64
+
+	var A edwards25519.ExtendedGroupElement
+	var hBytes [32]byte
+	copy(hBytes[:], digest[:])
+	edwards25519.GeScalarMultBase(&A, &hBytes)
+	var publicKeyBytes [32]byte
+	A.ToBytes(&publicKeyBytes)
+
+	privateKey := make([]byte, PrivateKeySize)
+	copy(privateKey, seed)
+	copy(privateKey[32:], publicKeyBytes[:])
+
+	return privateKey
+}
+
+// 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 {
+	if l := len(privateKey); l != PrivateKeySize {
+		panic("ed25519: bad private key length: " + strconv.Itoa(l))
+	}
+
+	h := sha512.New()
+	h.Write(privateKey[:32])
+
+	var digest1, messageDigest, hramDigest [64]byte
+	var expandedSecretKey [32]byte
+	h.Sum(digest1[:0])
+	copy(expandedSecretKey[:], digest1[:])
+	expandedSecretKey[0] &= 248
+	expandedSecretKey[31] &= 63
+	expandedSecretKey[31] |= 64
+
+	h.Reset()
+	h.Write(digest1[32:])
+	h.Write(message)
+	h.Sum(messageDigest[:0])
+
+	var messageDigestReduced [32]byte
+	edwards25519.ScReduce(&messageDigestReduced, &messageDigest)
+	var R edwards25519.ExtendedGroupElement
+	edwards25519.GeScalarMultBase(&R, &messageDigestReduced)
+
+	var encodedR [32]byte
+	R.ToBytes(&encodedR)
+
+	h.Reset()
+	h.Write(encodedR[:])
+	h.Write(privateKey[32:])
+	h.Write(message)
+	h.Sum(hramDigest[:0])
+	var hramDigestReduced [32]byte
+	edwards25519.ScReduce(&hramDigestReduced, &hramDigest)
+
+	var s [32]byte
+	edwards25519.ScMulAdd(&s, &hramDigestReduced, &expandedSecretKey, &messageDigestReduced)
+
+	signature := make([]byte, SignatureSize)
+	copy(signature[:], encodedR[:])
+	copy(signature[32:], s[:])
+
+	return signature
+}
+
+// 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 {
+	if l := len(publicKey); l != PublicKeySize {
+		panic("ed25519: bad public key length: " + strconv.Itoa(l))
+	}
+
+	if len(sig) != SignatureSize || sig[63]&224 != 0 {
+		return false
+	}
+
+	var A edwards25519.ExtendedGroupElement
+	var publicKeyBytes [32]byte
+	copy(publicKeyBytes[:], publicKey)
+	if !A.FromBytes(&publicKeyBytes) {
+		return false
+	}
+	edwards25519.FeNeg(&A.X, &A.X)
+	edwards25519.FeNeg(&A.T, &A.T)
+
+	h := sha512.New()
+	h.Write(sig[:32])
+	h.Write(publicKey[:])
+	h.Write(message)
+	var digest [64]byte
+	h.Sum(digest[:0])
+
+	var hReduced [32]byte
+	edwards25519.ScReduce(&hReduced, &digest)
+
+	var R edwards25519.ProjectiveGroupElement
+	var s [32]byte
+	copy(s[:], sig[32:])
+
+	// https://tools.ietf.org/html/rfc8032#section-5.1.7 requires that s be in
+	// the range [0, order) in order to prevent signature malleability.
+	if !edwards25519.ScMinimal(&s) {
+		return false
+	}
+
+	edwards25519.GeDoubleScalarMultVartime(&R, &hReduced, &A, &s)
+
+	var checkR [32]byte
+	R.ToBytes(&checkR)
+	return bytes.Equal(sig[:32], checkR[:])
+}

vendor/golang.org/x/crypto/ed25519/ed25519_go113.go

@@ -0,0 +1,74 @@
+// Copyright 2019 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build go1.13
+// +build go1.13
+
+// Package 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)
+}

vendor/golang.org/x/crypto/internal/subtle/aliasing.go

@@ -0,0 +1,33 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build !purego
+// +build !purego
+
+// Package subtle implements functions that are often useful in cryptographic
+// code but require careful thought to use correctly.
+package subtle // import "golang.org/x/crypto/internal/subtle"
+
+import "unsafe"
+
+// AnyOverlap reports whether x and y share memory at any (not necessarily
+// corresponding) index. The memory beyond the slice length is ignored.
+func AnyOverlap(x, y []byte) bool {
+	return len(x) > 0 && len(y) > 0 &&
+		uintptr(unsafe.Pointer(&x[0])) <= uintptr(unsafe.Pointer(&y[len(y)-1])) &&
+		uintptr(unsafe.Pointer(&y[0])) <= uintptr(unsafe.Pointer(&x[len(x)-1]))
+}
+
+// InexactOverlap reports whether x and y share memory at any non-corresponding
+// index. The memory beyond the slice length is ignored. Note that x and y can
+// have different lengths and still not have any inexact overlap.
+//
+// InexactOverlap can be used to implement the requirements of the crypto/cipher
+// AEAD, Block, BlockMode and Stream interfaces.
+func InexactOverlap(x, y []byte) bool {
+	if len(x) == 0 || len(y) == 0 || &x[0] == &y[0] {
+		return false
+	}
+	return AnyOverlap(x, y)
+}

vendor/golang.org/x/crypto/internal/subtle/aliasing_purego.go

@@ -0,0 +1,36 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build purego
+// +build purego
+
+// Package subtle implements functions that are often useful in cryptographic
+// code but require careful thought to use correctly.
+package subtle // import "golang.org/x/crypto/internal/subtle"
+
+// This is the Google App Engine standard variant based on reflect
+// because the unsafe package and cgo are disallowed.
+
+import "reflect"
+
+// AnyOverlap reports whether x and y share memory at any (not necessarily
+// corresponding) index. The memory beyond the slice length is ignored.
+func AnyOverlap(x, y []byte) bool {
+	return len(x) > 0 && len(y) > 0 &&
+		reflect.ValueOf(&x[0]).Pointer() <= reflect.ValueOf(&y[len(y)-1]).Pointer() &&
+		reflect.ValueOf(&y[0]).Pointer() <= reflect.ValueOf(&x[len(x)-1]).Pointer()
+}
+
+// InexactOverlap reports whether x and y share memory at any non-corresponding
+// index. The memory beyond the slice length is ignored. Note that x and y can
+// have different lengths and still not have any inexact overlap.
+//
+// InexactOverlap can be used to implement the requirements of the crypto/cipher
+// AEAD, Block, BlockMode and Stream interfaces.
+func InexactOverlap(x, y []byte) bool {
+	if len(x) == 0 || len(y) == 0 || &x[0] == &y[0] {
+		return false
+	}
+	return AnyOverlap(x, y)
+}

vendor/golang.org/x/crypto/pbkdf2/pbkdf2.go

@@ -0,0 +1,77 @@
+// 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 pbkdf2 implements the key derivation function PBKDF2 as defined in RFC
+2898 / PKCS #5 v2.0.
+
+A key derivation function is useful when encrypting data based on a password
+or any other not-fully-random data. It uses a pseudorandom function to derive
+a secure encryption key based on the password.
+
+While v2.0 of the standard defines only one pseudorandom function to use,
+HMAC-SHA1, the drafted v2.1 specification allows use of all five FIPS Approved
+Hash Functions SHA-1, SHA-224, SHA-256, SHA-384 and SHA-512 for HMAC. To
+choose, you can pass the `New` functions from the different SHA packages to
+pbkdf2.Key.
+*/
+package pbkdf2 // import "golang.org/x/crypto/pbkdf2"
+
+import (
+	"crypto/hmac"
+	"hash"
+)
+
+// Key derives a key from the password, salt and iteration count, returning a
+// []byte of length keylen that can be used as cryptographic key. The key is
+// derived based on the method described as PBKDF2 with the HMAC variant using
+// the supplied hash function.
+//
+// For example, to use a HMAC-SHA-1 based PBKDF2 key derivation function, you
+// can get a derived key for e.g. AES-256 (which needs a 32-byte key) by
+// doing:
+//
+// 	dk := pbkdf2.Key([]byte("some password"), salt, 4096, 32, sha1.New)
+//
+// Remember to get a good random salt. At least 8 bytes is recommended by the
+// RFC.
+//
+// Using a higher iteration count will increase the cost of an exhaustive
+// search but will also make derivation proportionally slower.
+func Key(password, salt []byte, iter, keyLen int, h func() hash.Hash) []byte {
+	prf := hmac.New(h, password)
+	hashLen := prf.Size()
+	numBlocks := (keyLen + hashLen - 1) / hashLen
+
+	var buf [4]byte
+	dk := make([]byte, 0, numBlocks*hashLen)
+	U := make([]byte, hashLen)
+	for block := 1; block <= numBlocks; block++ {
+		// N.B.: || means concatenation, ^ means XOR
+		// for each block T_i = U_1 ^ U_2 ^ ... ^ U_iter
+		// U_1 = PRF(password, salt || uint(i))
+		prf.Reset()
+		prf.Write(salt)
+		buf[0] = byte(block >> 24)
+		buf[1] = byte(block >> 16)
+		buf[2] = byte(block >> 8)
+		buf[3] = byte(block)
+		prf.Write(buf[:4])
+		dk = prf.Sum(dk)
+		T := dk[len(dk)-hashLen:]
+		copy(U, T)
+
+		// U_n = PRF(password, U_(n-1))
+		for n := 2; n <= iter; n++ {
+			prf.Reset()
+			prf.Write(U)
+			U = U[:0]
+			U = prf.Sum(U)
+			for x := range U {
+				T[x] ^= U[x]
+			}
+		}
+	}
+	return dk[:keyLen]
+}

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

@@ -0,0 +1,40 @@
+// Copyright 2019 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build !go1.13
+// +build !go1.13
+
+package poly1305
+
+// Generic fallbacks for the math/bits intrinsics, copied from
+// src/math/bits/bits.go. They were added in Go 1.12, but Add64 and Sum64 had
+// variable time fallbacks until Go 1.13.
+
+func bitsAdd64(x, y, carry uint64) (sum, carryOut uint64) {
+	sum = x + y + carry
+	carryOut = ((x & y) | ((x | y) &^ sum)) >> 63
+	return
+}
+
+func bitsSub64(x, y, borrow uint64) (diff, borrowOut uint64) {
+	diff = x - y - borrow
+	borrowOut = ((^x & y) | (^(x ^ y) & diff)) >> 63
+	return
+}
+
+func bitsMul64(x, y uint64) (hi, lo uint64) {
+	const mask32 = 1<<32 - 1
+	x0 := x & mask32
+	x1 := x >> 32
+	y0 := y & mask32
+	y1 := y >> 32
+	w0 := x0 * y0
+	t := x1*y0 + w0>>32
+	w1 := t & mask32
+	w2 := t >> 32
+	w1 += x0 * y1
+	hi = x1*y1 + w2 + w1>>32
+	lo = x * y
+	return
+}

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

@@ -0,0 +1,22 @@
+// Copyright 2019 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build go1.13
+// +build go1.13
+
+package poly1305
+
+import "math/bits"
+
+func bitsAdd64(x, y, carry uint64) (sum, carryOut uint64) {
+	return bits.Add64(x, y, carry)
+}
+
+func bitsSub64(x, y, borrow uint64) (diff, borrowOut uint64) {
+	return bits.Sub64(x, y, borrow)
+}
+
+func bitsMul64(x, y uint64) (hi, lo uint64) {
+	return bits.Mul64(x, y)
+}

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

@@ -0,0 +1,10 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build (!amd64 && !ppc64le && !s390x) || !gc || purego
+// +build !amd64,!ppc64le,!s390x !gc purego
+
+package poly1305
+
+type mac struct{ macGeneric }

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

@@ -0,0 +1,99 @@
+// Copyright 2012 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Package poly1305 implements Poly1305 one-time message authentication code as
+// specified in https://cr.yp.to/mac/poly1305-20050329.pdf.
+//
+// Poly1305 is a fast, one-time authentication function. It is infeasible for an
+// attacker to generate an authenticator for a message without the key. However, a
+// key must only be used for a single message. Authenticating two different
+// messages with the same key allows an attacker to forge authenticators for other
+// messages with the same key.
+//
+// Poly1305 was originally coupled with AES in order to make Poly1305-AES. AES was
+// used with a fixed key in order to generate one-time keys from an nonce.
+// However, in this package AES isn't used and the one-time key is specified
+// directly.
+package poly1305 // import "golang.org/x/crypto/poly1305"
+
+import "crypto/subtle"
+
+// TagSize is the size, in bytes, of a poly1305 authenticator.
+const TagSize = 16
+
+// Sum generates an authenticator for msg using a one-time key and puts the
+// 16-byte result into out. Authenticating two different messages with the same
+// key allows an attacker to forge messages at will.
+func Sum(out *[16]byte, m []byte, key *[32]byte) {
+	h := New(key)
+	h.Write(m)
+	h.Sum(out[:0])
+}
+
+// Verify returns true if mac is a valid authenticator for m with the given key.
+func Verify(mac *[16]byte, m []byte, key *[32]byte) bool {
+	var tmp [16]byte
+	Sum(&tmp, m, key)
+	return subtle.ConstantTimeCompare(tmp[:], mac[:]) == 1
+}
+
+// New returns a new MAC computing an authentication
+// tag of all data written to it with the given key.
+// This allows writing the message progressively instead
+// of passing it as a single slice. Common users should use
+// the Sum function instead.
+//
+// The key must be unique for each message, as authenticating
+// two different messages with the same key allows an attacker
+// to forge messages at will.
+func New(key *[32]byte) *MAC {
+	m := &MAC{}
+	initialize(key, &m.macState)
+	return m
+}
+
+// MAC is an io.Writer computing an authentication tag
+// of the data written to it.
+//
+// MAC cannot be used like common hash.Hash implementations,
+// because using a poly1305 key twice breaks its security.
+// Therefore writing data to a running MAC after calling
+// Sum or Verify causes it to panic.
+type MAC struct {
+	mac // platform-dependent implementation
+
+	finalized bool
+}
+
+// Size returns the number of bytes Sum will return.
+func (h *MAC) Size() int { return TagSize }
+
+// Write adds more data to the running message authentication code.
+// It never returns an error.
+//
+// It must not be called after the first call of Sum or Verify.
+func (h *MAC) Write(p []byte) (n int, err error) {
+	if h.finalized {
+		panic("poly1305: write to MAC after Sum or Verify")
+	}
+	return h.mac.Write(p)
+}
+
+// Sum computes the authenticator of all data written to the
+// message authentication code.
+func (h *MAC) Sum(b []byte) []byte {
+	var mac [TagSize]byte
+	h.mac.Sum(&mac)
+	h.finalized = true
+	return append(b, mac[:]...)
+}
+
+// Verify returns whether the authenticator of all data written to
+// the message authentication code matches the expected value.
+func (h *MAC) Verify(expected []byte) bool {
+	var mac [TagSize]byte
+	h.mac.Sum(&mac)
+	h.finalized = true
+	return subtle.ConstantTimeCompare(expected, mac[:]) == 1
+}

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

@@ -0,0 +1,48 @@
+// Copyright 2012 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build gc && !purego
+// +build gc,!purego
+
+package poly1305
+
+//go:noescape
+func update(state *macState, msg []byte)
+
+// mac is a wrapper for macGeneric that redirects calls that would have gone to
+// updateGeneric to update.
+//
+// Its Write and Sum methods are otherwise identical to the macGeneric ones, but
+// using function pointers would carry a major performance cost.
+type mac struct{ macGeneric }
+
+func (h *mac) Write(p []byte) (int, error) {
+	nn := len(p)
+	if h.offset > 0 {
+		n := copy(h.buffer[h.offset:], p)
+		if h.offset+n < TagSize {
+			h.offset += n
+			return nn, nil
+		}
+		p = p[n:]
+		h.offset = 0
+		update(&h.macState, h.buffer[:])
+	}
+	if n := len(p) - (len(p) % TagSize); n > 0 {
+		update(&h.macState, p[:n])
+		p = p[n:]
+	}
+	if len(p) > 0 {
+		h.offset += copy(h.buffer[h.offset:], p)
+	}
+	return nn, nil
+}
+
+func (h *mac) Sum(out *[16]byte) {
+	state := h.macState
+	if h.offset > 0 {
+		update(&state, h.buffer[:h.offset])
+	}
+	finalize(out, &state.h, &state.s)
+}

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

@@ -0,0 +1,109 @@
+// Copyright 2012 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build gc && !purego
+// +build gc,!purego
+
+#include "textflag.h"
+
+#define POLY1305_ADD(msg, h0, h1, h2) \
+	ADDQ 0(msg), h0;  \
+	ADCQ 8(msg), h1;  \
+	ADCQ $1, h2;      \
+	LEAQ 16(msg), msg
+
+#define POLY1305_MUL(h0, h1, h2, r0, r1, t0, t1, t2, t3) \
+	MOVQ  r0, AX;                  \
+	MULQ  h0;                      \
+	MOVQ  AX, t0;                  \
+	MOVQ  DX, t1;                  \
+	MOVQ  r0, AX;                  \
+	MULQ  h1;                      \
+	ADDQ  AX, t1;                  \
+	ADCQ  $0, DX;                  \
+	MOVQ  r0, t2;                  \
+	IMULQ h2, t2;                  \
+	ADDQ  DX, t2;                  \
+	                               \
+	MOVQ  r1, AX;                  \
+	MULQ  h0;                      \
+	ADDQ  AX, t1;                  \
+	ADCQ  $0, DX;                  \
+	MOVQ  DX, h0;                  \
+	MOVQ  r1, t3;                  \
+	IMULQ h2, t3;                  \
+	MOVQ  r1, AX;                  \
+	MULQ  h1;                      \
+	ADDQ  AX, t2;                  \
+	ADCQ  DX, t3;                  \
+	ADDQ  h0, t2;                  \
+	ADCQ  $0, t3;                  \
+	                               \
+	MOVQ  t0, h0;                  \
+	MOVQ  t1, h1;                  \
+	MOVQ  t2, h2;                  \
+	ANDQ  $3, h2;                  \
+	MOVQ  t2, t0;                  \
+	ANDQ  $0xFFFFFFFFFFFFFFFC, t0; \
+	ADDQ  t0, h0;                  \
+	ADCQ  t3, h1;                  \
+	ADCQ  $0, h2;                  \
+	SHRQ  $2, t3, t2;              \
+	SHRQ  $2, t3;                  \
+	ADDQ  t2, h0;                  \
+	ADCQ  t3, h1;                  \
+	ADCQ  $0, h2
+
+// 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
+
+	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
+
+loop:
+	POLY1305_ADD(SI, R8, R9, R10)
+
+multiply:
+	POLY1305_MUL(R8, R9, R10, R11, R12, BX, CX, R13, R14)
+	SUBQ $16, R15
+	CMPQ R15, $16
+	JAE  loop
+
+bytes_between_0_and_15:
+	TESTQ R15, R15
+	JZ    done
+	MOVQ  $1, BX
+	XORQ  CX, CX
+	XORQ  R13, R13
+	ADDQ  R15, SI
+
+flush_buffer:
+	SHLQ $8, BX, CX
+	SHLQ $8, BX
+	MOVB -1(SI), R13
+	XORQ R13, BX
+	DECQ SI
+	DECQ R15
+	JNZ  flush_buffer
+
+	ADDQ BX, R8
+	ADCQ CX, R9
+	ADCQ $0, R10
+	MOVQ $16, R15
+	JMP  multiply
+
+done:
+	MOVQ R8, 0(DI)
+	MOVQ R9, 8(DI)
+	MOVQ R10, 16(DI)
+	RET

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

@@ -0,0 +1,310 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// This file provides the generic implementation of Sum and MAC. Other files
+// might provide optimized assembly implementations of some of this code.
+
+package poly1305
+
+import "encoding/binary"
+
+// Poly1305 [RFC 7539] is a relatively simple algorithm: the authentication tag
+// for a 64 bytes message is approximately
+//
+//     s + m[0:16] * r⁴ + m[16:32] * r³ + m[32:48] * r² + m[48:64] * r  mod  2¹³⁰ - 5
+//
+// for some secret r and s. It can be computed sequentially like
+//
+//     for len(msg) > 0:
+//         h += read(msg, 16)
+//         h *= r
+//         h %= 2¹³⁰ - 5
+//     return h + s
+//
+// All the complexity is about doing performant constant-time math on numbers
+// larger than any available numeric type.
+
+func sumGeneric(out *[TagSize]byte, msg []byte, key *[32]byte) {
+	h := newMACGeneric(key)
+	h.Write(msg)
+	h.Sum(out)
+}
+
+func newMACGeneric(key *[32]byte) macGeneric {
+	m := macGeneric{}
+	initialize(key, &m.macState)
+	return m
+}
+
+// macState holds numbers in saturated 64-bit little-endian limbs. That is,
+// the value of [x0, x1, x2] is x[0] + x[1] * 2⁶⁴ + x[2] * 2¹²⁸.
+type macState struct {
+	// h is the main accumulator. It is to be interpreted modulo 2¹³⁰ - 5, but
+	// can grow larger during and after rounds. It must, however, remain below
+	// 2 * (2¹³⁰ - 5).
+	h [3]uint64
+	// r and s are the private key components.
+	r [2]uint64
+	s [2]uint64
+}
+
+type macGeneric struct {
+	macState
+
+	buffer [TagSize]byte
+	offset int
+}
+
+// Write splits the incoming message into TagSize chunks, and passes them to
+// update. It buffers incomplete chunks.
+func (h *macGeneric) Write(p []byte) (int, error) {
+	nn := len(p)
+	if h.offset > 0 {
+		n := copy(h.buffer[h.offset:], p)
+		if h.offset+n < TagSize {
+			h.offset += n
+			return nn, nil
+		}
+		p = p[n:]
+		h.offset = 0
+		updateGeneric(&h.macState, h.buffer[:])
+	}
+	if n := len(p) - (len(p) % TagSize); n > 0 {
+		updateGeneric(&h.macState, p[:n])
+		p = p[n:]
+	}
+	if len(p) > 0 {
+		h.offset += copy(h.buffer[h.offset:], p)
+	}
+	return nn, nil
+}
+
+// Sum flushes the last incomplete chunk from the buffer, if any, and generates
+// the MAC output. It does not modify its state, in order to allow for multiple
+// calls to Sum, even if no Write is allowed after Sum.
+func (h *macGeneric) Sum(out *[TagSize]byte) {
+	state := h.macState
+	if h.offset > 0 {
+		updateGeneric(&state, h.buffer[:h.offset])
+	}
+	finalize(out, &state.h, &state.s)
+}
+
+// [rMask0, rMask1] is the specified Poly1305 clamping mask in little-endian. It
+// clears some bits of the secret coefficient to make it possible to implement
+// multiplication more efficiently.
+const (
+	rMask0 = 0x0FFFFFFC0FFFFFFF
+	rMask1 = 0x0FFFFFFC0FFFFFFC
+)
+
+// initialize loads the 256-bit key into the two 128-bit secret values r and s.
+func initialize(key *[32]byte, m *macState) {
+	m.r[0] = binary.LittleEndian.Uint64(key[0:8]) & rMask0
+	m.r[1] = binary.LittleEndian.Uint64(key[8:16]) & rMask1
+	m.s[0] = binary.LittleEndian.Uint64(key[16:24])
+	m.s[1] = binary.LittleEndian.Uint64(key[24:32])
+}
+
+// uint128 holds a 128-bit number as two 64-bit limbs, for use with the
+// bits.Mul64 and bits.Add64 intrinsics.
+type uint128 struct {
+	lo, hi uint64
+}
+
+func mul64(a, b uint64) uint128 {
+	hi, lo := bitsMul64(a, b)
+	return uint128{lo, hi}
+}
+
+func add128(a, b uint128) uint128 {
+	lo, c := bitsAdd64(a.lo, b.lo, 0)
+	hi, c := bitsAdd64(a.hi, b.hi, c)
+	if c != 0 {
+		panic("poly1305: unexpected overflow")
+	}
+	return uint128{lo, hi}
+}
+
+func shiftRightBy2(a uint128) uint128 {
+	a.lo = a.lo>>2 | (a.hi&3)<<62
+	a.hi = a.hi >> 2
+	return a
+}
+
+// updateGeneric absorbs msg into the state.h accumulator. For each chunk m of
+// 128 bits of message, it computes
+//
+//     h₊ = (h + m) * r  mod  2¹³⁰ - 5
+//
+// If the msg length is not a multiple of TagSize, it assumes the last
+// incomplete chunk is the final one.
+func updateGeneric(state *macState, msg []byte) {
+	h0, h1, h2 := state.h[0], state.h[1], state.h[2]
+	r0, r1 := state.r[0], state.r[1]
+
+	for len(msg) > 0 {
+		var c uint64
+
+		// For the first step, h + m, we use a chain of bits.Add64 intrinsics.
+		// The resulting value of h might exceed 2¹³⁰ - 5, but will be partially
+		// reduced at the end of the multiplication below.
+		//
+		// The spec requires us to set a bit just above the message size, not to
+		// hide leading zeroes. For full chunks, that's 1 << 128, so we can just
+		// add 1 to the most significant (2¹²⁸) limb, h2.
+		if len(msg) >= TagSize {
+			h0, c = bitsAdd64(h0, binary.LittleEndian.Uint64(msg[0:8]), 0)
+			h1, c = bitsAdd64(h1, binary.LittleEndian.Uint64(msg[8:16]), c)
+			h2 += c + 1
+
+			msg = msg[TagSize:]
+		} else {
+			var buf [TagSize]byte
+			copy(buf[:], msg)
+			buf[len(msg)] = 1
+
+			h0, c = bitsAdd64(h0, binary.LittleEndian.Uint64(buf[0:8]), 0)
+			h1, c = bitsAdd64(h1, binary.LittleEndian.Uint64(buf[8:16]), c)
+			h2 += c
+
+			msg = nil
+		}
+
+		// Multiplication of big number limbs is similar to elementary school
+		// columnar multiplication. Instead of digits, there are 64-bit limbs.
+		//
+		// We are multiplying a 3 limbs number, h, by a 2 limbs number, r.
+		//
+		//                        h2    h1    h0  x
+		//                              r1    r0  =
+		//                       ----------------
+		//                      h2r0  h1r0  h0r0     <-- individual 128-bit products
+		//            +   h2r1  h1r1  h0r1
+		//               ------------------------
+		//                 m3    m2    m1    m0      <-- result in 128-bit overlapping limbs
+		//               ------------------------
+		//         m3.hi m2.hi m1.hi m0.hi           <-- carry propagation
+		//     +         m3.lo m2.lo m1.lo m0.lo
+		//        -------------------------------
+		//           t4    t3    t2    t1    t0      <-- final result in 64-bit limbs
+		//
+		// The main difference from pen-and-paper multiplication is that we do
+		// carry propagation in a separate step, as if we wrote two digit sums
+		// at first (the 128-bit limbs), and then carried the tens all at once.
+
+		h0r0 := mul64(h0, r0)
+		h1r0 := mul64(h1, r0)
+		h2r0 := mul64(h2, r0)
+		h0r1 := mul64(h0, r1)
+		h1r1 := mul64(h1, r1)
+		h2r1 := mul64(h2, r1)
+
+		// Since h2 is known to be at most 7 (5 + 1 + 1), and r0 and r1 have their
+		// top 4 bits cleared by rMask{0,1}, we know that their product is not going
+		// to overflow 64 bits, so we can ignore the high part of the products.
+		//
+		// This also means that the product doesn't have a fifth limb (t4).
+		if h2r0.hi != 0 {
+			panic("poly1305: unexpected overflow")
+		}
+		if h2r1.hi != 0 {
+			panic("poly1305: unexpected overflow")
+		}
+
+		m0 := h0r0
+		m1 := add128(h1r0, h0r1) // These two additions don't overflow thanks again
+		m2 := add128(h2r0, h1r1) // to the 4 masked bits at the top of r0 and r1.
+		m3 := h2r1
+
+		t0 := m0.lo
+		t1, c := bitsAdd64(m1.lo, m0.hi, 0)
+		t2, c := bitsAdd64(m2.lo, m1.hi, c)
+		t3, _ := bitsAdd64(m3.lo, m2.hi, c)
+
+		// Now we have the result as 4 64-bit limbs, and we need to reduce it
+		// modulo 2¹³⁰ - 5. The special shape of this Crandall prime lets us do
+		// a cheap partial reduction according to the reduction identity
+		//
+		//     c * 2¹³⁰ + n  =  c * 5 + n  mod  2¹³⁰ - 5
+		//
+		// because 2¹³⁰ = 5 mod 2¹³⁰ - 5. Partial reduction since the result is
+		// likely to be larger than 2¹³⁰ - 5, but still small enough to fit the
+		// assumptions we make about h in the rest of the code.
+		//
+		// See also https://speakerdeck.com/gtank/engineering-prime-numbers?slide=23
+
+		// We split the final result at the 2¹³⁰ mark into h and cc, the carry.
+		// Note that the carry bits are effectively shifted left by 2, in other
+		// words, cc = c * 4 for the c in the reduction identity.
+		h0, h1, h2 = t0, t1, t2&maskLow2Bits
+		cc := uint128{t2 & maskNotLow2Bits, t3}
+
+		// To add c * 5 to h, we first add cc = c * 4, and then add (cc >> 2) = c.
+
+		h0, c = bitsAdd64(h0, cc.lo, 0)
+		h1, c = bitsAdd64(h1, cc.hi, c)
+		h2 += c
+
+		cc = shiftRightBy2(cc)
+
+		h0, c = bitsAdd64(h0, cc.lo, 0)
+		h1, c = bitsAdd64(h1, cc.hi, c)
+		h2 += c
+
+		// h2 is at most 3 + 1 + 1 = 5, making the whole of h at most
+		//
+		//     5 * 2¹²⁸ + (2¹²⁸ - 1) = 6 * 2¹²⁸ - 1
+	}
+
+	state.h[0], state.h[1], state.h[2] = h0, h1, h2
+}
+
+const (
+	maskLow2Bits    uint64 = 0x0000000000000003
+	maskNotLow2Bits uint64 = ^maskLow2Bits
+)
+
+// select64 returns x if v == 1 and y if v == 0, in constant time.
+func select64(v, x, y uint64) uint64 { return ^(v-1)&x | (v-1)&y }
+
+// [p0, p1, p2] is 2¹³⁰ - 5 in little endian order.
+const (
+	p0 = 0xFFFFFFFFFFFFFFFB
+	p1 = 0xFFFFFFFFFFFFFFFF
+	p2 = 0x0000000000000003
+)
+
+// finalize completes the modular reduction of h and computes
+//
+//     out = h + s  mod  2¹²⁸
+//
+func finalize(out *[TagSize]byte, h *[3]uint64, s *[2]uint64) {
+	h0, h1, h2 := h[0], h[1], h[2]
+
+	// After the partial reduction in updateGeneric, h might be more than
+	// 2¹³⁰ - 5, but will be less than 2 * (2¹³⁰ - 5). To complete the reduction
+	// in constant time, we compute t = h - (2¹³⁰ - 5), and select h as the
+	// result if the subtraction underflows, and t otherwise.
+
+	hMinusP0, b := bitsSub64(h0, p0, 0)
+	hMinusP1, b := bitsSub64(h1, p1, b)
+	_, b = bitsSub64(h2, p2, b)
+
+	// h = h if h < p else h - p
+	h0 = select64(b, h0, hMinusP0)
+	h1 = select64(b, h1, hMinusP1)
+
+	// Finally, we compute the last Poly1305 step
+	//
+	//     tag = h + s  mod  2¹²⁸
+	//
+	// by just doing a wide addition with the 128 low bits of h and discarding
+	// the overflow.
+	h0, c := bitsAdd64(h0, s[0], 0)
+	h1, _ = bitsAdd64(h1, s[1], c)
+
+	binary.LittleEndian.PutUint64(out[0:8], h0)
+	binary.LittleEndian.PutUint64(out[8:16], h1)
+}

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

@@ -0,0 +1,48 @@
+// Copyright 2019 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build gc && !purego
+// +build gc,!purego
+
+package poly1305
+
+//go:noescape
+func update(state *macState, msg []byte)
+
+// mac is a wrapper for macGeneric that redirects calls that would have gone to
+// updateGeneric to update.
+//
+// Its Write and Sum methods are otherwise identical to the macGeneric ones, but
+// using function pointers would carry a major performance cost.
+type mac struct{ macGeneric }
+
+func (h *mac) Write(p []byte) (int, error) {
+	nn := len(p)
+	if h.offset > 0 {
+		n := copy(h.buffer[h.offset:], p)
+		if h.offset+n < TagSize {
+			h.offset += n
+			return nn, nil
+		}
+		p = p[n:]
+		h.offset = 0
+		update(&h.macState, h.buffer[:])
+	}
+	if n := len(p) - (len(p) % TagSize); n > 0 {
+		update(&h.macState, p[:n])
+		p = p[n:]
+	}
+	if len(p) > 0 {
+		h.offset += copy(h.buffer[h.offset:], p)
+	}
+	return nn, nil
+}
+
+func (h *mac) Sum(out *[16]byte) {
+	state := h.macState
+	if h.offset > 0 {
+		update(&state, h.buffer[:h.offset])
+	}
+	finalize(out, &state.h, &state.s)
+}

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

@@ -0,0 +1,182 @@
+// Copyright 2019 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build gc && !purego
+// +build gc,!purego
+
+#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  R5, $0
+	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 R17, $0
+
+	// 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   R5, $0
+	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
+	ADDZE R10, 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

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

@@ -0,0 +1,76 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build gc && !purego
+// +build gc,!purego
+
+package poly1305
+
+import (
+	"golang.org/x/sys/cpu"
+)
+
+// updateVX is an assembly implementation of Poly1305 that uses vector
+// instructions. It must only be called if the vector facility (vx) is
+// available.
+//go:noescape
+func updateVX(state *macState, msg []byte)
+
+// mac is a replacement for macGeneric that uses a larger buffer and redirects
+// calls that would have gone to updateGeneric to updateVX if the vector
+// facility is installed.
+//
+// A larger buffer is required for good performance because the vector
+// implementation has a higher fixed cost per call than the generic
+// implementation.
+type mac struct {
+	macState
+
+	buffer [16 * TagSize]byte // size must be a multiple of block size (16)
+	offset int
+}
+
+func (h *mac) Write(p []byte) (int, error) {
+	nn := len(p)
+	if h.offset > 0 {
+		n := copy(h.buffer[h.offset:], p)
+		if h.offset+n < len(h.buffer) {
+			h.offset += n
+			return nn, nil
+		}
+		p = p[n:]
+		h.offset = 0
+		if cpu.S390X.HasVX {
+			updateVX(&h.macState, h.buffer[:])
+		} else {
+			updateGeneric(&h.macState, h.buffer[:])
+		}
+	}
+
+	tail := len(p) % len(h.buffer) // number of bytes to copy into buffer
+	body := len(p) - tail          // number of bytes to process now
+	if body > 0 {
+		if cpu.S390X.HasVX {
+			updateVX(&h.macState, p[:body])
+		} else {
+			updateGeneric(&h.macState, p[:body])
+		}
+	}
+	h.offset = copy(h.buffer[:], p[body:]) // copy tail bytes - can be 0
+	return nn, nil
+}
+
+func (h *mac) Sum(out *[TagSize]byte) {
+	state := h.macState
+	remainder := h.buffer[:h.offset]
+
+	// Use the generic implementation if we have 2 or fewer blocks left
+	// to sum. The vector implementation has a higher startup time.
+	if cpu.S390X.HasVX && len(remainder) > 2*TagSize {
+		updateVX(&state, remainder)
+	} else if len(remainder) > 0 {
+		updateGeneric(&state, remainder)
+	}
+	finalize(out, &state.h, &state.s)
+}

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

@@ -0,0 +1,504 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build gc && !purego
+// +build gc,!purego
+
+#include "textflag.h"
+
+// This implementation of Poly1305 uses the vector facility (vx)
+// to process up to 2 blocks (32 bytes) per iteration using an
+// algorithm based on the one described in:
+//
+// NEON crypto, Daniel J. Bernstein & Peter Schwabe
+// https://cryptojedi.org/papers/neoncrypto-20120320.pdf
+//
+// This algorithm uses 5 26-bit limbs to represent a 130-bit
+// value. These limbs are, for the most part, zero extended and
+// placed into 64-bit vector register elements. Each vector
+// register is 128-bits wide and so holds 2 of these elements.
+// Using 26-bit limbs allows us plenty of headroom to accomodate
+// accumulations before and after multiplication without
+// overflowing either 32-bits (before multiplication) or 64-bits
+// (after multiplication).
+//
+// In order to parallelise the operations required to calculate
+// the sum we use two separate accumulators and then sum those
+// in an extra final step. For compatibility with the generic
+// implementation we perform this summation at the end of every
+// updateVX call.
+//
+// To use two accumulators we must multiply the message blocks
+// by r² rather than r. Only the final message block should be
+// multiplied by r.
+//
+// Example:
+//
+// We want to calculate the sum (h) for a 64 byte message (m):
+//
+//   h = m[0:16]r⁴ + m[16:32]r³ + m[32:48]r² + m[48:64]r
+//
+// To do this we split the calculation into the even indices
+// and odd indices of the message. These form our SIMD 'lanes':
+//
+//   h = m[ 0:16]r⁴ + m[32:48]r² +   <- lane 0
+//       m[16:32]r³ + m[48:64]r      <- lane 1
+//
+// To calculate this iteratively we refactor so that both lanes
+// are written in terms of r² and r:
+//
+//   h = (m[ 0:16]r² + m[32:48])r² + <- lane 0
+//       (m[16:32]r² + m[48:64])r    <- lane 1
+//                ^             ^
+//                |             coefficients for second iteration
+//                coefficients for first iteration
+//
+// So in this case we would have two iterations. In the first
+// both lanes are multiplied by r². In the second only the
+// first lane is multiplied by r² and the second lane is
+// instead multiplied by r. This gives use the odd and even
+// powers of r that we need from the original equation.
+//
+// Notation:
+//
+//   h - accumulator
+//   r - key
+//   m - message
+//
+//   [a, b]       - SIMD register holding two 64-bit values
+//   [a, b, c, d] - SIMD register holding four 32-bit values
+//   xᵢ[n]        - limb n of variable x with bit width i
+//
+// Limbs are expressed in little endian order, so for 26-bit
+// limbs x₂₆[4] will be the most significant limb and x₂₆[0]
+// will be the least significant limb.
+
+// masking constants
+#define MOD24 V0 // [0x0000000000ffffff, 0x0000000000ffffff] - mask low 24-bits
+#define MOD26 V1 // [0x0000000003ffffff, 0x0000000003ffffff] - mask low 26-bits
+
+// expansion constants (see EXPAND macro)
+#define EX0 V2
+#define EX1 V3
+#define EX2 V4
+
+// key (r², r or 1 depending on context)
+#define R_0 V5
+#define R_1 V6
+#define R_2 V7
+#define R_3 V8
+#define R_4 V9
+
+// precalculated coefficients (5r², 5r or 0 depending on context)
+#define R5_1 V10
+#define R5_2 V11
+#define R5_3 V12
+#define R5_4 V13
+
+// message block (m)
+#define M_0 V14
+#define M_1 V15
+#define M_2 V16
+#define M_3 V17
+#define M_4 V18
+
+// accumulator (h)
+#define H_0 V19
+#define H_1 V20
+#define H_2 V21
+#define H_3 V22
+#define H_4 V23
+
+// temporary registers (for short-lived values)
+#define T_0 V24
+#define T_1 V25
+#define T_2 V26
+#define T_3 V27
+#define T_4 V28
+
+GLOBL ·constants<>(SB), RODATA, $0x30
+// EX0
+DATA ·constants<>+0x00(SB)/8, $0x0006050403020100
+DATA ·constants<>+0x08(SB)/8, $0x1016151413121110
+// EX1
+DATA ·constants<>+0x10(SB)/8, $0x060c0b0a09080706
+DATA ·constants<>+0x18(SB)/8, $0x161c1b1a19181716
+// EX2
+DATA ·constants<>+0x20(SB)/8, $0x0d0d0d0d0d0f0e0d
+DATA ·constants<>+0x28(SB)/8, $0x1d1d1d1d1d1f1e1d
+
+// MULTIPLY multiplies each lane of f and g, partially reduced
+// modulo 2¹³⁰ - 5. The result, h, consists of partial products
+// in each lane that need to be reduced further to produce the
+// final result.
+//
+//   h₁₃₀ = (f₁₃₀g₁₃₀) % 2¹³⁰ + (5f₁₃₀g₁₃₀) / 2¹³⁰
+//
+// Note that the multiplication by 5 of the high bits is
+// achieved by precalculating the multiplication of four of the
+// g coefficients by 5. These are g51-g54.
+#define MULTIPLY(f0, f1, f2, f3, f4, g0, g1, g2, g3, g4, g51, g52, g53, g54, h0, h1, h2, h3, h4) \
+	VMLOF  f0, g0, h0        \
+	VMLOF  f0, g3, h3        \
+	VMLOF  f0, g1, h1        \
+	VMLOF  f0, g4, h4        \
+	VMLOF  f0, g2, h2        \
+	VMLOF  f1, g54, T_0      \
+	VMLOF  f1, g2, T_3       \
+	VMLOF  f1, g0, T_1       \
+	VMLOF  f1, g3, T_4       \
+	VMLOF  f1, g1, T_2       \
+	VMALOF f2, g53, h0, h0   \
+	VMALOF f2, g1, h3, h3    \
+	VMALOF f2, g54, h1, h1   \
+	VMALOF f2, g2, h4, h4    \
+	VMALOF f2, g0, h2, h2    \
+	VMALOF f3, g52, T_0, T_0 \
+	VMALOF f3, g0, T_3, T_3  \
+	VMALOF f3, g53, T_1, T_1 \
+	VMALOF f3, g1, T_4, T_4  \
+	VMALOF f3, g54, T_2, T_2 \
+	VMALOF f4, g51, h0, h0   \
+	VMALOF f4, g54, h3, h3   \
+	VMALOF f4, g52, h1, h1   \
+	VMALOF f4, g0, h4, h4    \
+	VMALOF f4, g53, h2, h2   \
+	VAG    T_0, h0, h0       \
+	VAG    T_3, h3, h3       \
+	VAG    T_1, h1, h1       \
+	VAG    T_4, h4, h4       \
+	VAG    T_2, h2, h2
+
+// REDUCE performs the following carry operations in four
+// stages, as specified in Bernstein & Schwabe:
+//
+//   1: h₂₆[0]->h₂₆[1] h₂₆[3]->h₂₆[4]
+//   2: h₂₆[1]->h₂₆[2] h₂₆[4]->h₂₆[0]
+//   3: h₂₆[0]->h₂₆[1] h₂₆[2]->h₂₆[3]
+//   4: h₂₆[3]->h₂₆[4]
+//
+// The result is that all of the limbs are limited to 26-bits
+// except for h₂₆[1] and h₂₆[4] which are limited to 27-bits.
+//
+// Note that although each limb is aligned at 26-bit intervals
+// they may contain values that exceed 2²⁶ - 1, hence the need
+// to carry the excess bits in each limb.
+#define REDUCE(h0, h1, h2, h3, h4) \
+	VESRLG $26, h0, T_0  \
+	VESRLG $26, h3, T_1  \
+	VN     MOD26, h0, h0 \
+	VN     MOD26, h3, h3 \
+	VAG    T_0, h1, h1   \
+	VAG    T_1, h4, h4   \
+	VESRLG $26, h1, T_2  \
+	VESRLG $26, h4, T_3  \
+	VN     MOD26, h1, h1 \
+	VN     MOD26, h4, h4 \
+	VESLG  $2, T_3, T_4  \
+	VAG    T_3, T_4, T_4 \
+	VAG    T_2, h2, h2   \
+	VAG    T_4, h0, h0   \
+	VESRLG $26, h2, T_0  \
+	VESRLG $26, h0, T_1  \
+	VN     MOD26, h2, h2 \
+	VN     MOD26, h0, h0 \
+	VAG    T_0, h3, h3   \
+	VAG    T_1, h1, h1   \
+	VESRLG $26, h3, T_2  \
+	VN     MOD26, h3, h3 \
+	VAG    T_2, h4, h4
+
+// EXPAND splits the 128-bit little-endian values in0 and in1
+// into 26-bit big-endian limbs and places the results into
+// the first and second lane of d₂₆[0:4] respectively.
+//
+// The EX0, EX1 and EX2 constants are arrays of byte indices
+// for permutation. The permutation both reverses the bytes
+// in the input and ensures the bytes are copied into the
+// destination limb ready to be shifted into their final
+// position.
+#define EXPAND(in0, in1, d0, d1, d2, d3, d4) \
+	VPERM  in0, in1, EX0, d0 \
+	VPERM  in0, in1, EX1, d2 \
+	VPERM  in0, in1, EX2, d4 \
+	VESRLG $26, d0, d1       \
+	VESRLG $30, d2, d3       \
+	VESRLG $4, d2, d2        \
+	VN     MOD26, d0, d0     \ // [in0₂₆[0], in1₂₆[0]]
+	VN     MOD26, d3, d3     \ // [in0₂₆[3], in1₂₆[3]]
+	VN     MOD26, d1, d1     \ // [in0₂₆[1], in1₂₆[1]]
+	VN     MOD24, d4, d4     \ // [in0₂₆[4], in1₂₆[4]]
+	VN     MOD26, d2, d2     // [in0₂₆[2], in1₂₆[2]]
+
+// func updateVX(state *macState, msg []byte)
+TEXT ·updateVX(SB), NOSPLIT, $0
+	MOVD state+0(FP), R1
+	LMG  msg+8(FP), R2, R3 // R2=msg_base, R3=msg_len
+
+	// load EX0, EX1 and EX2
+	MOVD $·constants<>(SB), R5
+	VLM  (R5), EX0, EX2
+
+	// generate masks
+	VGMG $(64-24), $63, MOD24 // [0x00ffffff, 0x00ffffff]
+	VGMG $(64-26), $63, MOD26 // [0x03ffffff, 0x03ffffff]
+
+	// load h (accumulator) and r (key) from state
+	VZERO T_1               // [0, 0]
+	VL    0(R1), T_0        // [h₆₄[0], h₆₄[1]]
+	VLEG  $0, 16(R1), T_1   // [h₆₄[2], 0]
+	VL    24(R1), T_2       // [r₆₄[0], r₆₄[1]]
+	VPDI  $0, T_0, T_2, T_3 // [h₆₄[0], r₆₄[0]]
+	VPDI  $5, T_0, T_2, T_4 // [h₆₄[1], r₆₄[1]]
+
+	// unpack h and r into 26-bit limbs
+	// note: h₆₄[2] may have the low 3 bits set, so h₂₆[4] is a 27-bit value
+	VN     MOD26, T_3, H_0            // [h₂₆[0], r₂₆[0]]
+	VZERO  H_1                        // [0, 0]
+	VZERO  H_3                        // [0, 0]
+	VGMG   $(64-12-14), $(63-12), T_0 // [0x03fff000, 0x03fff000] - 26-bit mask with low 12 bits masked out
+	VESLG  $24, T_1, T_1              // [h₆₄[2]<<24, 0]
+	VERIMG $-26&63, T_3, MOD26, H_1   // [h₂₆[1], r₂₆[1]]
+	VESRLG $+52&63, T_3, H_2          // [h₂₆[2], r₂₆[2]] - low 12 bits only
+	VERIMG $-14&63, T_4, MOD26, H_3   // [h₂₆[1], r₂₆[1]]
+	VESRLG $40, T_4, H_4              // [h₂₆[4], r₂₆[4]] - low 24 bits only
+	VERIMG $+12&63, T_4, T_0, H_2     // [h₂₆[2], r₂₆[2]] - complete
+	VO     T_1, H_4, H_4              // [h₂₆[4], r₂₆[4]] - complete
+
+	// replicate r across all 4 vector elements
+	VREPF $3, H_0, R_0 // [r₂₆[0], r₂₆[0], r₂₆[0], r₂₆[0]]
+	VREPF $3, H_1, R_1 // [r₂₆[1], r₂₆[1], r₂₆[1], r₂₆[1]]
+	VREPF $3, H_2, R_2 // [r₂₆[2], r₂₆[2], r₂₆[2], r₂₆[2]]
+	VREPF $3, H_3, R_3 // [r₂₆[3], r₂₆[3], r₂₆[3], r₂₆[3]]
+	VREPF $3, H_4, R_4 // [r₂₆[4], r₂₆[4], r₂₆[4], r₂₆[4]]
+
+	// zero out lane 1 of h
+	VLEIG $1, $0, H_0 // [h₂₆[0], 0]
+	VLEIG $1, $0, H_1 // [h₂₆[1], 0]
+	VLEIG $1, $0, H_2 // [h₂₆[2], 0]
+	VLEIG $1, $0, H_3 // [h₂₆[3], 0]
+	VLEIG $1, $0, H_4 // [h₂₆[4], 0]
+
+	// calculate 5r (ignore least significant limb)
+	VREPIF $5, T_0
+	VMLF   T_0, R_1, R5_1 // [5r₂₆[1], 5r₂₆[1], 5r₂₆[1], 5r₂₆[1]]
+	VMLF   T_0, R_2, R5_2 // [5r₂₆[2], 5r₂₆[2], 5r₂₆[2], 5r₂₆[2]]
+	VMLF   T_0, R_3, R5_3 // [5r₂₆[3], 5r₂₆[3], 5r₂₆[3], 5r₂₆[3]]
+	VMLF   T_0, R_4, R5_4 // [5r₂₆[4], 5r₂₆[4], 5r₂₆[4], 5r₂₆[4]]
+
+	// skip r² calculation if we are only calculating one block
+	CMPBLE R3, $16, skip
+
+	// calculate r²
+	MULTIPLY(R_0, R_1, R_2, R_3, R_4, R_0, R_1, R_2, R_3, R_4, R5_1, R5_2, R5_3, R5_4, M_0, M_1, M_2, M_3, M_4)
+	REDUCE(M_0, M_1, M_2, M_3, M_4)
+	VGBM   $0x0f0f, T_0
+	VERIMG $0, M_0, T_0, R_0 // [r₂₆[0], r²₂₆[0], r₂₆[0], r²₂₆[0]]
+	VERIMG $0, M_1, T_0, R_1 // [r₂₆[1], r²₂₆[1], r₂₆[1], r²₂₆[1]]
+	VERIMG $0, M_2, T_0, R_2 // [r₂₆[2], r²₂₆[2], r₂₆[2], r²₂₆[2]]
+	VERIMG $0, M_3, T_0, R_3 // [r₂₆[3], r²₂₆[3], r₂₆[3], r²₂₆[3]]
+	VERIMG $0, M_4, T_0, R_4 // [r₂₆[4], r²₂₆[4], r₂₆[4], r²₂₆[4]]
+
+	// calculate 5r² (ignore least significant limb)
+	VREPIF $5, T_0
+	VMLF   T_0, R_1, R5_1 // [5r₂₆[1], 5r²₂₆[1], 5r₂₆[1], 5r²₂₆[1]]
+	VMLF   T_0, R_2, R5_2 // [5r₂₆[2], 5r²₂₆[2], 5r₂₆[2], 5r²₂₆[2]]
+	VMLF   T_0, R_3, R5_3 // [5r₂₆[3], 5r²₂₆[3], 5r₂₆[3], 5r²₂₆[3]]
+	VMLF   T_0, R_4, R5_4 // [5r₂₆[4], 5r²₂₆[4], 5r₂₆[4], 5r²₂₆[4]]
+
+loop:
+	CMPBLE R3, $32, b2 // 2 or fewer blocks remaining, need to change key coefficients
+
+	// load next 2 blocks from message
+	VLM (R2), T_0, T_1
+
+	// update message slice
+	SUB  $32, R3
+	MOVD $32(R2), R2
+
+	// unpack message blocks into 26-bit big-endian limbs
+	EXPAND(T_0, T_1, M_0, M_1, M_2, M_3, M_4)
+
+	// add 2¹²⁸ to each message block value
+	VLEIB $4, $1, M_4
+	VLEIB $12, $1, M_4
+
+multiply:
+	// accumulate the incoming message
+	VAG H_0, M_0, M_0
+	VAG H_3, M_3, M_3
+	VAG H_1, M_1, M_1
+	VAG H_4, M_4, M_4
+	VAG H_2, M_2, M_2
+
+	// multiply the accumulator by the key coefficient
+	MULTIPLY(M_0, M_1, M_2, M_3, M_4, R_0, R_1, R_2, R_3, R_4, R5_1, R5_2, R5_3, R5_4, H_0, H_1, H_2, H_3, H_4)
+
+	// carry and partially reduce the partial products
+	REDUCE(H_0, H_1, H_2, H_3, H_4)
+
+	CMPBNE R3, $0, loop
+
+finish:
+	// sum lane 0 and lane 1 and put the result in lane 1
+	VZERO  T_0
+	VSUMQG H_0, T_0, H_0
+	VSUMQG H_3, T_0, H_3
+	VSUMQG H_1, T_0, H_1
+	VSUMQG H_4, T_0, H_4
+	VSUMQG H_2, T_0, H_2
+
+	// reduce again after summation
+	// TODO(mundaym): there might be a more efficient way to do this
+	// now that we only have 1 active lane. For example, we could
+	// simultaneously pack the values as we reduce them.
+	REDUCE(H_0, H_1, H_2, H_3, H_4)
+
+	// carry h[1] through to h[4] so that only h[4] can exceed 2²⁶ - 1
+	// TODO(mundaym): in testing this final carry was unnecessary.
+	// Needs a proof before it can be removed though.
+	VESRLG $26, H_1, T_1
+	VN     MOD26, H_1, H_1
+	VAQ    T_1, H_2, H_2
+	VESRLG $26, H_2, T_2
+	VN     MOD26, H_2, H_2
+	VAQ    T_2, H_3, H_3
+	VESRLG $26, H_3, T_3
+	VN     MOD26, H_3, H_3
+	VAQ    T_3, H_4, H_4
+
+	// h is now < 2(2¹³⁰ - 5)
+	// Pack each lane in h₂₆[0:4] into h₁₂₈[0:1].
+	VESLG $26, H_1, H_1
+	VESLG $26, H_3, H_3
+	VO    H_0, H_1, H_0
+	VO    H_2, H_3, H_2
+	VESLG $4, H_2, H_2
+	VLEIB $7, $48, H_1
+	VSLB  H_1, H_2, H_2
+	VO    H_0, H_2, H_0
+	VLEIB $7, $104, H_1
+	VSLB  H_1, H_4, H_3
+	VO    H_3, H_0, H_0
+	VLEIB $7, $24, H_1
+	VSRLB H_1, H_4, H_1
+
+	// update state
+	VSTEG $1, H_0, 0(R1)
+	VSTEG $0, H_0, 8(R1)
+	VSTEG $1, H_1, 16(R1)
+	RET
+
+b2:  // 2 or fewer blocks remaining
+	CMPBLE R3, $16, b1
+
+	// Load the 2 remaining blocks (17-32 bytes remaining).
+	MOVD $-17(R3), R0    // index of final byte to load modulo 16
+	VL   (R2), T_0       // load full 16 byte block
+	VLL  R0, 16(R2), T_1 // load final (possibly partial) block and pad with zeros to 16 bytes
+
+	// The Poly1305 algorithm requires that a 1 bit be appended to
+	// each message block. If the final block is less than 16 bytes
+	// long then it is easiest to insert the 1 before the message
+	// block is split into 26-bit limbs. If, on the other hand, the
+	// final message block is 16 bytes long then we append the 1 bit
+	// after expansion as normal.
+	MOVBZ  $1, R0
+	MOVD   $-16(R3), R3   // index of byte in last block to insert 1 at (could be 16)
+	CMPBEQ R3, $16, 2(PC) // skip the insertion if the final block is 16 bytes long
+	VLVGB  R3, R0, T_1    // insert 1 into the byte at index R3
+
+	// Split both blocks into 26-bit limbs in the appropriate lanes.
+	EXPAND(T_0, T_1, M_0, M_1, M_2, M_3, M_4)
+
+	// Append a 1 byte to the end of the second to last block.
+	VLEIB $4, $1, M_4
+
+	// Append a 1 byte to the end of the last block only if it is a
+	// full 16 byte block.
+	CMPBNE R3, $16, 2(PC)
+	VLEIB  $12, $1, M_4
+
+	// Finally, set up the coefficients for the final multiplication.
+	// We have previously saved r and 5r in the 32-bit even indexes
+	// of the R_[0-4] and R5_[1-4] coefficient registers.
+	//
+	// We want lane 0 to be multiplied by r² so that can be kept the
+	// same. We want lane 1 to be multiplied by r so we need to move
+	// the saved r value into the 32-bit odd index in lane 1 by
+	// rotating the 64-bit lane by 32.
+	VGBM   $0x00ff, T_0         // [0, 0xffffffffffffffff] - mask lane 1 only
+	VERIMG $32, R_0, T_0, R_0   // [_,  r²₂₆[0], _,  r₂₆[0]]
+	VERIMG $32, R_1, T_0, R_1   // [_,  r²₂₆[1], _,  r₂₆[1]]
+	VERIMG $32, R_2, T_0, R_2   // [_,  r²₂₆[2], _,  r₂₆[2]]
+	VERIMG $32, R_3, T_0, R_3   // [_,  r²₂₆[3], _,  r₂₆[3]]
+	VERIMG $32, R_4, T_0, R_4   // [_,  r²₂₆[4], _,  r₂₆[4]]
+	VERIMG $32, R5_1, T_0, R5_1 // [_, 5r²₂₆[1], _, 5r₂₆[1]]
+	VERIMG $32, R5_2, T_0, R5_2 // [_, 5r²₂₆[2], _, 5r₂₆[2]]
+	VERIMG $32, R5_3, T_0, R5_3 // [_, 5r²₂₆[3], _, 5r₂₆[3]]
+	VERIMG $32, R5_4, T_0, R5_4 // [_, 5r²₂₆[4], _, 5r₂₆[4]]
+
+	MOVD $0, R3
+	BR   multiply
+
+skip:
+	CMPBEQ R3, $0, finish
+
+b1:  // 1 block remaining
+
+	// Load the final block (1-16 bytes). This will be placed into
+	// lane 0.
+	MOVD $-1(R3), R0
+	VLL  R0, (R2), T_0 // pad to 16 bytes with zeros
+
+	// The Poly1305 algorithm requires that a 1 bit be appended to
+	// each message block. If the final block is less than 16 bytes
+	// long then it is easiest to insert the 1 before the message
+	// block is split into 26-bit limbs. If, on the other hand, the
+	// final message block is 16 bytes long then we append the 1 bit
+	// after expansion as normal.
+	MOVBZ  $1, R0
+	CMPBEQ R3, $16, 2(PC)
+	VLVGB  R3, R0, T_0
+
+	// Set the message block in lane 1 to the value 0 so that it
+	// can be accumulated without affecting the final result.
+	VZERO T_1
+
+	// Split the final message block into 26-bit limbs in lane 0.
+	// Lane 1 will be contain 0.
+	EXPAND(T_0, T_1, M_0, M_1, M_2, M_3, M_4)
+
+	// Append a 1 byte to the end of the last block only if it is a
+	// full 16 byte block.
+	CMPBNE R3, $16, 2(PC)
+	VLEIB  $4, $1, M_4
+
+	// We have previously saved r and 5r in the 32-bit even indexes
+	// of the R_[0-4] and R5_[1-4] coefficient registers.
+	//
+	// We want lane 0 to be multiplied by r so we need to move the
+	// saved r value into the 32-bit odd index in lane 0. We want
+	// lane 1 to be set to the value 1. This makes multiplication
+	// a no-op. We do this by setting lane 1 in every register to 0
+	// and then just setting the 32-bit index 3 in R_0 to 1.
+	VZERO T_0
+	MOVD  $0, R0
+	MOVD  $0x10111213, R12
+	VLVGP R12, R0, T_1         // [_, 0x10111213, _, 0x00000000]
+	VPERM T_0, R_0, T_1, R_0   // [_,  r₂₆[0], _, 0]
+	VPERM T_0, R_1, T_1, R_1   // [_,  r₂₆[1], _, 0]
+	VPERM T_0, R_2, T_1, R_2   // [_,  r₂₆[2], _, 0]
+	VPERM T_0, R_3, T_1, R_3   // [_,  r₂₆[3], _, 0]
+	VPERM T_0, R_4, T_1, R_4   // [_,  r₂₆[4], _, 0]
+	VPERM T_0, R5_1, T_1, R5_1 // [_, 5r₂₆[1], _, 0]
+	VPERM T_0, R5_2, T_1, R5_2 // [_, 5r₂₆[2], _, 0]
+	VPERM T_0, R5_3, T_1, R5_3 // [_, 5r₂₆[3], _, 0]
+	VPERM T_0, R5_4, T_1, R5_4 // [_, 5r₂₆[4], _, 0]
+
+	// Set the value of lane 1 to be 1.
+	VLEIF $3, $1, R_0 // [_,  r₂₆[0], _, 1]
+
+	MOVD $0, R3
+	BR   multiply

vendor/golang.org/x/crypto/ripemd160/ripemd160.go

@@ -0,0 +1,124 @@
+// Copyright 2010 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Package ripemd160 implements the RIPEMD-160 hash algorithm.
+//
+// Deprecated: RIPEMD-160 is a legacy hash and should not be used for new
+// applications. Also, this package does not and will not provide an optimized
+// implementation. Instead, use a modern hash like SHA-256 (from crypto/sha256).
+package ripemd160 // import "golang.org/x/crypto/ripemd160"
+
+// RIPEMD-160 is designed by Hans Dobbertin, Antoon Bosselaers, and Bart
+// Preneel with specifications available at:
+// http://homes.esat.kuleuven.be/~cosicart/pdf/AB-9601/AB-9601.pdf.
+
+import (
+	"crypto"
+	"hash"
+)
+
+func init() {
+	crypto.RegisterHash(crypto.RIPEMD160, New)
+}
+
+// The size of the checksum in bytes.
+const Size = 20
+
+// The block size of the hash algorithm in bytes.
+const BlockSize = 64
+
+const (
+	_s0 = 0x67452301
+	_s1 = 0xefcdab89
+	_s2 = 0x98badcfe
+	_s3 = 0x10325476
+	_s4 = 0xc3d2e1f0
+)
+
+// digest represents the partial evaluation of a checksum.
+type digest struct {
+	s  [5]uint32       // running context
+	x  [BlockSize]byte // temporary buffer
+	nx int             // index into x
+	tc uint64          // total count of bytes processed
+}
+
+func (d *digest) Reset() {
+	d.s[0], d.s[1], d.s[2], d.s[3], d.s[4] = _s0, _s1, _s2, _s3, _s4
+	d.nx = 0
+	d.tc = 0
+}
+
+// New returns a new hash.Hash computing the checksum.
+func New() hash.Hash {
+	result := new(digest)
+	result.Reset()
+	return result
+}
+
+func (d *digest) Size() int { return Size }
+
+func (d *digest) BlockSize() int { return BlockSize }
+
+func (d *digest) Write(p []byte) (nn int, err error) {
+	nn = len(p)
+	d.tc += uint64(nn)
+	if d.nx > 0 {
+		n := len(p)
+		if n > BlockSize-d.nx {
+			n = BlockSize - d.nx
+		}
+		for i := 0; i < n; i++ {
+			d.x[d.nx+i] = p[i]
+		}
+		d.nx += n
+		if d.nx == BlockSize {
+			_Block(d, d.x[0:])
+			d.nx = 0
+		}
+		p = p[n:]
+	}
+	n := _Block(d, p)
+	p = p[n:]
+	if len(p) > 0 {
+		d.nx = copy(d.x[:], p)
+	}
+	return
+}
+
+func (d0 *digest) Sum(in []byte) []byte {
+	// Make a copy of d0 so that caller can keep writing and summing.
+	d := *d0
+
+	// Padding.  Add a 1 bit and 0 bits until 56 bytes mod 64.
+	tc := d.tc
+	var tmp [64]byte
+	tmp[0] = 0x80
+	if tc%64 < 56 {
+		d.Write(tmp[0 : 56-tc%64])
+	} else {
+		d.Write(tmp[0 : 64+56-tc%64])
+	}
+
+	// Length in bits.
+	tc <<= 3
+	for i := uint(0); i < 8; i++ {
+		tmp[i] = byte(tc >> (8 * i))
+	}
+	d.Write(tmp[0:8])
+
+	if d.nx != 0 {
+		panic("d.nx != 0")
+	}
+
+	var digest [Size]byte
+	for i, s := range d.s {
+		digest[i*4] = byte(s)
+		digest[i*4+1] = byte(s >> 8)
+		digest[i*4+2] = byte(s >> 16)
+		digest[i*4+3] = byte(s >> 24)
+	}
+
+	return append(in, digest[:]...)
+}

vendor/golang.org/x/crypto/ripemd160/ripemd160block.go

@@ -0,0 +1,165 @@
+// Copyright 2010 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// RIPEMD-160 block step.
+// In its own file so that a faster assembly or C version
+// can be substituted easily.
+
+package ripemd160
+
+import (
+	"math/bits"
+)
+
+// work buffer indices and roll amounts for one line
+var _n = [80]uint{
+	0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
+	7, 4, 13, 1, 10, 6, 15, 3, 12, 0, 9, 5, 2, 14, 11, 8,
+	3, 10, 14, 4, 9, 15, 8, 1, 2, 7, 0, 6, 13, 11, 5, 12,
+	1, 9, 11, 10, 0, 8, 12, 4, 13, 3, 7, 15, 14, 5, 6, 2,
+	4, 0, 5, 9, 7, 12, 2, 10, 14, 1, 3, 8, 11, 6, 15, 13,
+}
+
+var _r = [80]uint{
+	11, 14, 15, 12, 5, 8, 7, 9, 11, 13, 14, 15, 6, 7, 9, 8,
+	7, 6, 8, 13, 11, 9, 7, 15, 7, 12, 15, 9, 11, 7, 13, 12,
+	11, 13, 6, 7, 14, 9, 13, 15, 14, 8, 13, 6, 5, 12, 7, 5,
+	11, 12, 14, 15, 14, 15, 9, 8, 9, 14, 5, 6, 8, 6, 5, 12,
+	9, 15, 5, 11, 6, 8, 13, 12, 5, 12, 13, 14, 11, 8, 5, 6,
+}
+
+// same for the other parallel one
+var n_ = [80]uint{
+	5, 14, 7, 0, 9, 2, 11, 4, 13, 6, 15, 8, 1, 10, 3, 12,
+	6, 11, 3, 7, 0, 13, 5, 10, 14, 15, 8, 12, 4, 9, 1, 2,
+	15, 5, 1, 3, 7, 14, 6, 9, 11, 8, 12, 2, 10, 0, 4, 13,
+	8, 6, 4, 1, 3, 11, 15, 0, 5, 12, 2, 13, 9, 7, 10, 14,
+	12, 15, 10, 4, 1, 5, 8, 7, 6, 2, 13, 14, 0, 3, 9, 11,
+}
+
+var r_ = [80]uint{
+	8, 9, 9, 11, 13, 15, 15, 5, 7, 7, 8, 11, 14, 14, 12, 6,
+	9, 13, 15, 7, 12, 8, 9, 11, 7, 7, 12, 7, 6, 15, 13, 11,
+	9, 7, 15, 11, 8, 6, 6, 14, 12, 13, 5, 14, 13, 13, 7, 5,
+	15, 5, 8, 11, 14, 14, 6, 14, 6, 9, 12, 9, 12, 5, 15, 8,
+	8, 5, 12, 9, 12, 5, 14, 6, 8, 13, 6, 5, 15, 13, 11, 11,
+}
+
+func _Block(md *digest, p []byte) int {
+	n := 0
+	var x [16]uint32
+	var alpha, beta uint32
+	for len(p) >= BlockSize {
+		a, b, c, d, e := md.s[0], md.s[1], md.s[2], md.s[3], md.s[4]
+		aa, bb, cc, dd, ee := a, b, c, d, e
+		j := 0
+		for i := 0; i < 16; i++ {
+			x[i] = uint32(p[j]) | uint32(p[j+1])<<8 | uint32(p[j+2])<<16 | uint32(p[j+3])<<24
+			j += 4
+		}
+
+		// round 1
+		i := 0
+		for i < 16 {
+			alpha = a + (b ^ c ^ d) + x[_n[i]]
+			s := int(_r[i])
+			alpha = bits.RotateLeft32(alpha, s) + e
+			beta = bits.RotateLeft32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+
+			// parallel line
+			alpha = aa + (bb ^ (cc | ^dd)) + x[n_[i]] + 0x50a28be6
+			s = int(r_[i])
+			alpha = bits.RotateLeft32(alpha, s) + ee
+			beta = bits.RotateLeft32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+
+			i++
+		}
+
+		// round 2
+		for i < 32 {
+			alpha = a + (b&c | ^b&d) + x[_n[i]] + 0x5a827999
+			s := int(_r[i])
+			alpha = bits.RotateLeft32(alpha, s) + e
+			beta = bits.RotateLeft32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+
+			// parallel line
+			alpha = aa + (bb&dd | cc&^dd) + x[n_[i]] + 0x5c4dd124
+			s = int(r_[i])
+			alpha = bits.RotateLeft32(alpha, s) + ee
+			beta = bits.RotateLeft32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+
+			i++
+		}
+
+		// round 3
+		for i < 48 {
+			alpha = a + (b | ^c ^ d) + x[_n[i]] + 0x6ed9eba1
+			s := int(_r[i])
+			alpha = bits.RotateLeft32(alpha, s) + e
+			beta = bits.RotateLeft32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+
+			// parallel line
+			alpha = aa + (bb | ^cc ^ dd) + x[n_[i]] + 0x6d703ef3
+			s = int(r_[i])
+			alpha = bits.RotateLeft32(alpha, s) + ee
+			beta = bits.RotateLeft32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+
+			i++
+		}
+
+		// round 4
+		for i < 64 {
+			alpha = a + (b&d | c&^d) + x[_n[i]] + 0x8f1bbcdc
+			s := int(_r[i])
+			alpha = bits.RotateLeft32(alpha, s) + e
+			beta = bits.RotateLeft32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+
+			// parallel line
+			alpha = aa + (bb&cc | ^bb&dd) + x[n_[i]] + 0x7a6d76e9
+			s = int(r_[i])
+			alpha = bits.RotateLeft32(alpha, s) + ee
+			beta = bits.RotateLeft32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+
+			i++
+		}
+
+		// round 5
+		for i < 80 {
+			alpha = a + (b ^ (c | ^d)) + x[_n[i]] + 0xa953fd4e
+			s := int(_r[i])
+			alpha = bits.RotateLeft32(alpha, s) + e
+			beta = bits.RotateLeft32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+
+			// parallel line
+			alpha = aa + (bb ^ cc ^ dd) + x[n_[i]]
+			s = int(r_[i])
+			alpha = bits.RotateLeft32(alpha, s) + ee
+			beta = bits.RotateLeft32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+
+			i++
+		}
+
+		// combine results
+		dd += c + md.s[1]
+		md.s[1] = md.s[2] + d + ee
+		md.s[2] = md.s[3] + e + aa
+		md.s[3] = md.s[4] + a + bb
+		md.s[4] = md.s[0] + b + cc
+		md.s[0] = dd
+
+		p = p[BlockSize:]
+		n += BlockSize
+	}
+	return n
+}

vendor/golang.org/x/crypto/sha3/doc.go

@@ -0,0 +1,66 @@
+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Package sha3 implements the SHA-3 fixed-output-length hash functions and
+// the SHAKE variable-output-length hash functions defined by FIPS-202.
+//
+// Both types of hash function use the "sponge" construction and the Keccak
+// permutation. For a detailed specification see http://keccak.noekeon.org/
+//
+//
+// Guidance
+//
+// If you aren't sure what function you need, use SHAKE256 with at least 64
+// bytes of output. The SHAKE instances are faster than the SHA3 instances;
+// the latter have to allocate memory to conform to the hash.Hash interface.
+//
+// If you need a secret-key MAC (message authentication code), prepend the
+// secret key to the input, hash with SHAKE256 and read at least 32 bytes of
+// output.
+//
+//
+// Security strengths
+//
+// The SHA3-x (x equals 224, 256, 384, or 512) functions have a security
+// strength against preimage attacks of x bits. Since they only produce "x"
+// bits of output, their collision-resistance is only "x/2" bits.
+//
+// The SHAKE-256 and -128 functions have a generic security strength of 256 and
+// 128 bits against all attacks, provided that at least 2x bits of their output
+// is used.  Requesting more than 64 or 32 bytes of output, respectively, does
+// not increase the collision-resistance of the SHAKE functions.
+//
+//
+// The sponge construction
+//
+// A sponge builds a pseudo-random function from a public pseudo-random
+// permutation, by applying the permutation to a state of "rate + capacity"
+// bytes, but hiding "capacity" of the bytes.
+//
+// A sponge starts out with a zero state. To hash an input using a sponge, up
+// to "rate" bytes of the input are XORed into the sponge's state. The sponge
+// is then "full" and the permutation is applied to "empty" it. This process is
+// repeated until all the input has been "absorbed". The input is then padded.
+// The digest is "squeezed" from the sponge in the same way, except that output
+// is copied out instead of input being XORed in.
+//
+// A sponge is parameterized by its generic security strength, which is equal
+// to half its capacity; capacity + rate is equal to the permutation's width.
+// Since the KeccakF-1600 permutation is 1600 bits (200 bytes) wide, this means
+// that the security strength of a sponge instance is equal to (1600 - bitrate) / 2.
+//
+//
+// Recommendations
+//
+// The SHAKE functions are recommended for most new uses. They can produce
+// output of arbitrary length. SHAKE256, with an output length of at least
+// 64 bytes, provides 256-bit security against all attacks.  The Keccak team
+// recommends it for most applications upgrading from SHA2-512. (NIST chose a
+// much stronger, but much slower, sponge instance for SHA3-512.)
+//
+// The SHA-3 functions are "drop-in" replacements for the SHA-2 functions.
+// They produce output of the same length, with the same security strengths
+// against all attacks. This means, in particular, that SHA3-256 only has
+// 128-bit collision resistance, because its output length is 32 bytes.
+package sha3 // import "golang.org/x/crypto/sha3"

vendor/golang.org/x/crypto/sha3/hashes.go

@@ -0,0 +1,97 @@
+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package sha3
+
+// This file provides functions for creating instances of the SHA-3
+// and SHAKE hash functions, as well as utility functions for hashing
+// bytes.
+
+import (
+	"hash"
+)
+
+// New224 creates a new SHA3-224 hash.
+// Its generic security strength is 224 bits against preimage attacks,
+// and 112 bits against collision attacks.
+func New224() hash.Hash {
+	if h := new224Asm(); h != nil {
+		return h
+	}
+	return &state{rate: 144, outputLen: 28, dsbyte: 0x06}
+}
+
+// New256 creates a new SHA3-256 hash.
+// Its generic security strength is 256 bits against preimage attacks,
+// and 128 bits against collision attacks.
+func New256() hash.Hash {
+	if h := new256Asm(); h != nil {
+		return h
+	}
+	return &state{rate: 136, outputLen: 32, dsbyte: 0x06}
+}
+
+// New384 creates a new SHA3-384 hash.
+// Its generic security strength is 384 bits against preimage attacks,
+// and 192 bits against collision attacks.
+func New384() hash.Hash {
+	if h := new384Asm(); h != nil {
+		return h
+	}
+	return &state{rate: 104, outputLen: 48, dsbyte: 0x06}
+}
+
+// New512 creates a new SHA3-512 hash.
+// Its generic security strength is 512 bits against preimage attacks,
+// and 256 bits against collision attacks.
+func New512() hash.Hash {
+	if h := new512Asm(); h != nil {
+		return h
+	}
+	return &state{rate: 72, outputLen: 64, dsbyte: 0x06}
+}
+
+// NewLegacyKeccak256 creates a new Keccak-256 hash.
+//
+// Only use this function if you require compatibility with an existing cryptosystem
+// that uses non-standard padding. All other users should use New256 instead.
+func NewLegacyKeccak256() hash.Hash { return &state{rate: 136, outputLen: 32, dsbyte: 0x01} }
+
+// NewLegacyKeccak512 creates a new Keccak-512 hash.
+//
+// Only use this function if you require compatibility with an existing cryptosystem
+// that uses non-standard padding. All other users should use New512 instead.
+func NewLegacyKeccak512() hash.Hash { return &state{rate: 72, outputLen: 64, dsbyte: 0x01} }
+
+// Sum224 returns the SHA3-224 digest of the data.
+func Sum224(data []byte) (digest [28]byte) {
+	h := New224()
+	h.Write(data)
+	h.Sum(digest[:0])
+	return
+}
+
+// Sum256 returns the SHA3-256 digest of the data.
+func Sum256(data []byte) (digest [32]byte) {
+	h := New256()
+	h.Write(data)
+	h.Sum(digest[:0])
+	return
+}
+
+// Sum384 returns the SHA3-384 digest of the data.
+func Sum384(data []byte) (digest [48]byte) {
+	h := New384()
+	h.Write(data)
+	h.Sum(digest[:0])
+	return
+}
+
+// Sum512 returns the SHA3-512 digest of the data.
+func Sum512(data []byte) (digest [64]byte) {
+	h := New512()
+	h.Write(data)
+	h.Sum(digest[:0])
+	return
+}

vendor/golang.org/x/crypto/sha3/hashes_generic.go

@@ -0,0 +1,28 @@
+// Copyright 2017 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build !gc || purego || !s390x
+// +build !gc purego !s390x
+
+package sha3
+
+import (
+	"hash"
+)
+
+// new224Asm returns an assembly implementation of SHA3-224 if available,
+// otherwise it returns nil.
+func new224Asm() hash.Hash { return nil }
+
+// new256Asm returns an assembly implementation of SHA3-256 if available,
+// otherwise it returns nil.
+func new256Asm() hash.Hash { return nil }
+
+// new384Asm returns an assembly implementation of SHA3-384 if available,
+// otherwise it returns nil.
+func new384Asm() hash.Hash { return nil }
+
+// new512Asm returns an assembly implementation of SHA3-512 if available,
+// otherwise it returns nil.
+func new512Asm() hash.Hash { return nil }

vendor/golang.org/x/crypto/sha3/keccakf.go

@@ -0,0 +1,413 @@
+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build !amd64 || purego || !gc
+// +build !amd64 purego !gc
+
+package sha3
+
+// rc stores the round constants for use in the ι step.
+var rc = [24]uint64{
+	0x0000000000000001,
+	0x0000000000008082,
+	0x800000000000808A,
+	0x8000000080008000,
+	0x000000000000808B,
+	0x0000000080000001,
+	0x8000000080008081,
+	0x8000000000008009,
+	0x000000000000008A,
+	0x0000000000000088,
+	0x0000000080008009,
+	0x000000008000000A,
+	0x000000008000808B,
+	0x800000000000008B,
+	0x8000000000008089,
+	0x8000000000008003,
+	0x8000000000008002,
+	0x8000000000000080,
+	0x000000000000800A,
+	0x800000008000000A,
+	0x8000000080008081,
+	0x8000000000008080,
+	0x0000000080000001,
+	0x8000000080008008,
+}
+
+// keccakF1600 applies the Keccak permutation to a 1600b-wide
+// state represented as a slice of 25 uint64s.
+func keccakF1600(a *[25]uint64) {
+	// Implementation translated from Keccak-inplace.c
+	// in the keccak reference code.
+	var t, bc0, bc1, bc2, bc3, bc4, d0, d1, d2, d3, d4 uint64
+
+	for i := 0; i < 24; i += 4 {
+		// Combines the 5 steps in each round into 2 steps.
+		// Unrolls 4 rounds per loop and spreads some steps across rounds.
+
+		// Round 1
+		bc0 = a[0] ^ a[5] ^ a[10] ^ a[15] ^ a[20]
+		bc1 = a[1] ^ a[6] ^ a[11] ^ a[16] ^ a[21]
+		bc2 = a[2] ^ a[7] ^ a[12] ^ a[17] ^ a[22]
+		bc3 = a[3] ^ a[8] ^ a[13] ^ a[18] ^ a[23]
+		bc4 = a[4] ^ a[9] ^ a[14] ^ a[19] ^ a[24]
+		d0 = bc4 ^ (bc1<<1 | bc1>>63)
+		d1 = bc0 ^ (bc2<<1 | bc2>>63)
+		d2 = bc1 ^ (bc3<<1 | bc3>>63)
+		d3 = bc2 ^ (bc4<<1 | bc4>>63)
+		d4 = bc3 ^ (bc0<<1 | bc0>>63)
+
+		bc0 = a[0] ^ d0
+		t = a[6] ^ d1
+		bc1 = t<<44 | t>>(64-44)
+		t = a[12] ^ d2
+		bc2 = t<<43 | t>>(64-43)
+		t = a[18] ^ d3
+		bc3 = t<<21 | t>>(64-21)
+		t = a[24] ^ d4
+		bc4 = t<<14 | t>>(64-14)
+		a[0] = bc0 ^ (bc2 &^ bc1) ^ rc[i]
+		a[6] = bc1 ^ (bc3 &^ bc2)
+		a[12] = bc2 ^ (bc4 &^ bc3)
+		a[18] = bc3 ^ (bc0 &^ bc4)
+		a[24] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[10] ^ d0
+		bc2 = t<<3 | t>>(64-3)
+		t = a[16] ^ d1
+		bc3 = t<<45 | t>>(64-45)
+		t = a[22] ^ d2
+		bc4 = t<<61 | t>>(64-61)
+		t = a[3] ^ d3
+		bc0 = t<<28 | t>>(64-28)
+		t = a[9] ^ d4
+		bc1 = t<<20 | t>>(64-20)
+		a[10] = bc0 ^ (bc2 &^ bc1)
+		a[16] = bc1 ^ (bc3 &^ bc2)
+		a[22] = bc2 ^ (bc4 &^ bc3)
+		a[3] = bc3 ^ (bc0 &^ bc4)
+		a[9] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[20] ^ d0
+		bc4 = t<<18 | t>>(64-18)
+		t = a[1] ^ d1
+		bc0 = t<<1 | t>>(64-1)
+		t = a[7] ^ d2
+		bc1 = t<<6 | t>>(64-6)
+		t = a[13] ^ d3
+		bc2 = t<<25 | t>>(64-25)
+		t = a[19] ^ d4
+		bc3 = t<<8 | t>>(64-8)
+		a[20] = bc0 ^ (bc2 &^ bc1)
+		a[1] = bc1 ^ (bc3 &^ bc2)
+		a[7] = bc2 ^ (bc4 &^ bc3)
+		a[13] = bc3 ^ (bc0 &^ bc4)
+		a[19] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[5] ^ d0
+		bc1 = t<<36 | t>>(64-36)
+		t = a[11] ^ d1
+		bc2 = t<<10 | t>>(64-10)
+		t = a[17] ^ d2
+		bc3 = t<<15 | t>>(64-15)
+		t = a[23] ^ d3
+		bc4 = t<<56 | t>>(64-56)
+		t = a[4] ^ d4
+		bc0 = t<<27 | t>>(64-27)
+		a[5] = bc0 ^ (bc2 &^ bc1)
+		a[11] = bc1 ^ (bc3 &^ bc2)
+		a[17] = bc2 ^ (bc4 &^ bc3)
+		a[23] = bc3 ^ (bc0 &^ bc4)
+		a[4] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[15] ^ d0
+		bc3 = t<<41 | t>>(64-41)
+		t = a[21] ^ d1
+		bc4 = t<<2 | t>>(64-2)
+		t = a[2] ^ d2
+		bc0 = t<<62 | t>>(64-62)
+		t = a[8] ^ d3
+		bc1 = t<<55 | t>>(64-55)
+		t = a[14] ^ d4
+		bc2 = t<<39 | t>>(64-39)
+		a[15] = bc0 ^ (bc2 &^ bc1)
+		a[21] = bc1 ^ (bc3 &^ bc2)
+		a[2] = bc2 ^ (bc4 &^ bc3)
+		a[8] = bc3 ^ (bc0 &^ bc4)
+		a[14] = bc4 ^ (bc1 &^ bc0)
+
+		// Round 2
+		bc0 = a[0] ^ a[5] ^ a[10] ^ a[15] ^ a[20]
+		bc1 = a[1] ^ a[6] ^ a[11] ^ a[16] ^ a[21]
+		bc2 = a[2] ^ a[7] ^ a[12] ^ a[17] ^ a[22]
+		bc3 = a[3] ^ a[8] ^ a[13] ^ a[18] ^ a[23]
+		bc4 = a[4] ^ a[9] ^ a[14] ^ a[19] ^ a[24]
+		d0 = bc4 ^ (bc1<<1 | bc1>>63)
+		d1 = bc0 ^ (bc2<<1 | bc2>>63)
+		d2 = bc1 ^ (bc3<<1 | bc3>>63)
+		d3 = bc2 ^ (bc4<<1 | bc4>>63)
+		d4 = bc3 ^ (bc0<<1 | bc0>>63)
+
+		bc0 = a[0] ^ d0
+		t = a[16] ^ d1
+		bc1 = t<<44 | t>>(64-44)
+		t = a[7] ^ d2
+		bc2 = t<<43 | t>>(64-43)
+		t = a[23] ^ d3
+		bc3 = t<<21 | t>>(64-21)
+		t = a[14] ^ d4
+		bc4 = t<<14 | t>>(64-14)
+		a[0] = bc0 ^ (bc2 &^ bc1) ^ rc[i+1]
+		a[16] = bc1 ^ (bc3 &^ bc2)
+		a[7] = bc2 ^ (bc4 &^ bc3)
+		a[23] = bc3 ^ (bc0 &^ bc4)
+		a[14] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[20] ^ d0
+		bc2 = t<<3 | t>>(64-3)
+		t = a[11] ^ d1
+		bc3 = t<<45 | t>>(64-45)
+		t = a[2] ^ d2
+		bc4 = t<<61 | t>>(64-61)
+		t = a[18] ^ d3
+		bc0 = t<<28 | t>>(64-28)
+		t = a[9] ^ d4
+		bc1 = t<<20 | t>>(64-20)
+		a[20] = bc0 ^ (bc2 &^ bc1)
+		a[11] = bc1 ^ (bc3 &^ bc2)
+		a[2] = bc2 ^ (bc4 &^ bc3)
+		a[18] = bc3 ^ (bc0 &^ bc4)
+		a[9] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[15] ^ d0
+		bc4 = t<<18 | t>>(64-18)
+		t = a[6] ^ d1
+		bc0 = t<<1 | t>>(64-1)
+		t = a[22] ^ d2
+		bc1 = t<<6 | t>>(64-6)
+		t = a[13] ^ d3
+		bc2 = t<<25 | t>>(64-25)
+		t = a[4] ^ d4
+		bc3 = t<<8 | t>>(64-8)
+		a[15] = bc0 ^ (bc2 &^ bc1)
+		a[6] = bc1 ^ (bc3 &^ bc2)
+		a[22] = bc2 ^ (bc4 &^ bc3)
+		a[13] = bc3 ^ (bc0 &^ bc4)
+		a[4] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[10] ^ d0
+		bc1 = t<<36 | t>>(64-36)
+		t = a[1] ^ d1
+		bc2 = t<<10 | t>>(64-10)
+		t = a[17] ^ d2
+		bc3 = t<<15 | t>>(64-15)
+		t = a[8] ^ d3
+		bc4 = t<<56 | t>>(64-56)
+		t = a[24] ^ d4
+		bc0 = t<<27 | t>>(64-27)
+		a[10] = bc0 ^ (bc2 &^ bc1)
+		a[1] = bc1 ^ (bc3 &^ bc2)
+		a[17] = bc2 ^ (bc4 &^ bc3)
+		a[8] = bc3 ^ (bc0 &^ bc4)
+		a[24] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[5] ^ d0
+		bc3 = t<<41 | t>>(64-41)
+		t = a[21] ^ d1
+		bc4 = t<<2 | t>>(64-2)
+		t = a[12] ^ d2
+		bc0 = t<<62 | t>>(64-62)
+		t = a[3] ^ d3
+		bc1 = t<<55 | t>>(64-55)
+		t = a[19] ^ d4
+		bc2 = t<<39 | t>>(64-39)
+		a[5] = bc0 ^ (bc2 &^ bc1)
+		a[21] = bc1 ^ (bc3 &^ bc2)
+		a[12] = bc2 ^ (bc4 &^ bc3)
+		a[3] = bc3 ^ (bc0 &^ bc4)
+		a[19] = bc4 ^ (bc1 &^ bc0)
+
+		// Round 3
+		bc0 = a[0] ^ a[5] ^ a[10] ^ a[15] ^ a[20]
+		bc1 = a[1] ^ a[6] ^ a[11] ^ a[16] ^ a[21]
+		bc2 = a[2] ^ a[7] ^ a[12] ^ a[17] ^ a[22]
+		bc3 = a[3] ^ a[8] ^ a[13] ^ a[18] ^ a[23]
+		bc4 = a[4] ^ a[9] ^ a[14] ^ a[19] ^ a[24]
+		d0 = bc4 ^ (bc1<<1 | bc1>>63)
+		d1 = bc0 ^ (bc2<<1 | bc2>>63)
+		d2 = bc1 ^ (bc3<<1 | bc3>>63)
+		d3 = bc2 ^ (bc4<<1 | bc4>>63)
+		d4 = bc3 ^ (bc0<<1 | bc0>>63)
+
+		bc0 = a[0] ^ d0
+		t = a[11] ^ d1
+		bc1 = t<<44 | t>>(64-44)
+		t = a[22] ^ d2
+		bc2 = t<<43 | t>>(64-43)
+		t = a[8] ^ d3
+		bc3 = t<<21 | t>>(64-21)
+		t = a[19] ^ d4
+		bc4 = t<<14 | t>>(64-14)
+		a[0] = bc0 ^ (bc2 &^ bc1) ^ rc[i+2]
+		a[11] = bc1 ^ (bc3 &^ bc2)
+		a[22] = bc2 ^ (bc4 &^ bc3)
+		a[8] = bc3 ^ (bc0 &^ bc4)
+		a[19] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[15] ^ d0
+		bc2 = t<<3 | t>>(64-3)
+		t = a[1] ^ d1
+		bc3 = t<<45 | t>>(64-45)
+		t = a[12] ^ d2
+		bc4 = t<<61 | t>>(64-61)
+		t = a[23] ^ d3
+		bc0 = t<<28 | t>>(64-28)
+		t = a[9] ^ d4
+		bc1 = t<<20 | t>>(64-20)
+		a[15] = bc0 ^ (bc2 &^ bc1)
+		a[1] = bc1 ^ (bc3 &^ bc2)
+		a[12] = bc2 ^ (bc4 &^ bc3)
+		a[23] = bc3 ^ (bc0 &^ bc4)
+		a[9] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[5] ^ d0
+		bc4 = t<<18 | t>>(64-18)
+		t = a[16] ^ d1
+		bc0 = t<<1 | t>>(64-1)
+		t = a[2] ^ d2
+		bc1 = t<<6 | t>>(64-6)
+		t = a[13] ^ d3
+		bc2 = t<<25 | t>>(64-25)
+		t = a[24] ^ d4
+		bc3 = t<<8 | t>>(64-8)
+		a[5] = bc0 ^ (bc2 &^ bc1)
+		a[16] = bc1 ^ (bc3 &^ bc2)
+		a[2] = bc2 ^ (bc4 &^ bc3)
+		a[13] = bc3 ^ (bc0 &^ bc4)
+		a[24] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[20] ^ d0
+		bc1 = t<<36 | t>>(64-36)
+		t = a[6] ^ d1
+		bc2 = t<<10 | t>>(64-10)
+		t = a[17] ^ d2
+		bc3 = t<<15 | t>>(64-15)
+		t = a[3] ^ d3
+		bc4 = t<<56 | t>>(64-56)
+		t = a[14] ^ d4
+		bc0 = t<<27 | t>>(64-27)
+		a[20] = bc0 ^ (bc2 &^ bc1)
+		a[6] = bc1 ^ (bc3 &^ bc2)
+		a[17] = bc2 ^ (bc4 &^ bc3)
+		a[3] = bc3 ^ (bc0 &^ bc4)
+		a[14] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[10] ^ d0
+		bc3 = t<<41 | t>>(64-41)
+		t = a[21] ^ d1
+		bc4 = t<<2 | t>>(64-2)
+		t = a[7] ^ d2
+		bc0 = t<<62 | t>>(64-62)
+		t = a[18] ^ d3
+		bc1 = t<<55 | t>>(64-55)
+		t = a[4] ^ d4
+		bc2 = t<<39 | t>>(64-39)
+		a[10] = bc0 ^ (bc2 &^ bc1)
+		a[21] = bc1 ^ (bc3 &^ bc2)
+		a[7] = bc2 ^ (bc4 &^ bc3)
+		a[18] = bc3 ^ (bc0 &^ bc4)
+		a[4] = bc4 ^ (bc1 &^ bc0)
+
+		// Round 4
+		bc0 = a[0] ^ a[5] ^ a[10] ^ a[15] ^ a[20]
+		bc1 = a[1] ^ a[6] ^ a[11] ^ a[16] ^ a[21]
+		bc2 = a[2] ^ a[7] ^ a[12] ^ a[17] ^ a[22]
+		bc3 = a[3] ^ a[8] ^ a[13] ^ a[18] ^ a[23]
+		bc4 = a[4] ^ a[9] ^ a[14] ^ a[19] ^ a[24]
+		d0 = bc4 ^ (bc1<<1 | bc1>>63)
+		d1 = bc0 ^ (bc2<<1 | bc2>>63)
+		d2 = bc1 ^ (bc3<<1 | bc3>>63)
+		d3 = bc2 ^ (bc4<<1 | bc4>>63)
+		d4 = bc3 ^ (bc0<<1 | bc0>>63)
+
+		bc0 = a[0] ^ d0
+		t = a[1] ^ d1
+		bc1 = t<<44 | t>>(64-44)
+		t = a[2] ^ d2
+		bc2 = t<<43 | t>>(64-43)
+		t = a[3] ^ d3
+		bc3 = t<<21 | t>>(64-21)
+		t = a[4] ^ d4
+		bc4 = t<<14 | t>>(64-14)
+		a[0] = bc0 ^ (bc2 &^ bc1) ^ rc[i+3]
+		a[1] = bc1 ^ (bc3 &^ bc2)
+		a[2] = bc2 ^ (bc4 &^ bc3)
+		a[3] = bc3 ^ (bc0 &^ bc4)
+		a[4] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[5] ^ d0
+		bc2 = t<<3 | t>>(64-3)
+		t = a[6] ^ d1
+		bc3 = t<<45 | t>>(64-45)
+		t = a[7] ^ d2
+		bc4 = t<<61 | t>>(64-61)
+		t = a[8] ^ d3
+		bc0 = t<<28 | t>>(64-28)
+		t = a[9] ^ d4
+		bc1 = t<<20 | t>>(64-20)
+		a[5] = bc0 ^ (bc2 &^ bc1)
+		a[6] = bc1 ^ (bc3 &^ bc2)
+		a[7] = bc2 ^ (bc4 &^ bc3)
+		a[8] = bc3 ^ (bc0 &^ bc4)
+		a[9] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[10] ^ d0
+		bc4 = t<<18 | t>>(64-18)
+		t = a[11] ^ d1
+		bc0 = t<<1 | t>>(64-1)
+		t = a[12] ^ d2
+		bc1 = t<<6 | t>>(64-6)
+		t = a[13] ^ d3
+		bc2 = t<<25 | t>>(64-25)
+		t = a[14] ^ d4
+		bc3 = t<<8 | t>>(64-8)
+		a[10] = bc0 ^ (bc2 &^ bc1)
+		a[11] = bc1 ^ (bc3 &^ bc2)
+		a[12] = bc2 ^ (bc4 &^ bc3)
+		a[13] = bc3 ^ (bc0 &^ bc4)
+		a[14] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[15] ^ d0
+		bc1 = t<<36 | t>>(64-36)
+		t = a[16] ^ d1
+		bc2 = t<<10 | t>>(64-10)
+		t = a[17] ^ d2
+		bc3 = t<<15 | t>>(64-15)
+		t = a[18] ^ d3
+		bc4 = t<<56 | t>>(64-56)
+		t = a[19] ^ d4
+		bc0 = t<<27 | t>>(64-27)
+		a[15] = bc0 ^ (bc2 &^ bc1)
+		a[16] = bc1 ^ (bc3 &^ bc2)
+		a[17] = bc2 ^ (bc4 &^ bc3)
+		a[18] = bc3 ^ (bc0 &^ bc4)
+		a[19] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[20] ^ d0
+		bc3 = t<<41 | t>>(64-41)
+		t = a[21] ^ d1
+		bc4 = t<<2 | t>>(64-2)
+		t = a[22] ^ d2
+		bc0 = t<<62 | t>>(64-62)
+		t = a[23] ^ d3
+		bc1 = t<<55 | t>>(64-55)
+		t = a[24] ^ d4
+		bc2 = t<<39 | t>>(64-39)
+		a[20] = bc0 ^ (bc2 &^ bc1)
+		a[21] = bc1 ^ (bc3 &^ bc2)
+		a[22] = bc2 ^ (bc4 &^ bc3)
+		a[23] = bc3 ^ (bc0 &^ bc4)
+		a[24] = bc4 ^ (bc1 &^ bc0)
+	}
+}

vendor/golang.org/x/crypto/sha3/keccakf_amd64.go

@@ -0,0 +1,14 @@
+// Copyright 2015 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build amd64 && !purego && gc
+// +build amd64,!purego,gc
+
+package sha3
+
+// This function is implemented in keccakf_amd64.s.
+
+//go:noescape
+
+func keccakF1600(a *[25]uint64)

vendor/golang.org/x/crypto/sha3/keccakf_amd64.s

@@ -0,0 +1,391 @@
+// Copyright 2015 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build amd64 && !purego && gc
+// +build amd64,!purego,gc
+
+// This code was translated into a form compatible with 6a from the public
+// domain sources at https://github.com/gvanas/KeccakCodePackage
+
+// Offsets in state
+#define _ba  (0*8)
+#define _be  (1*8)
+#define _bi  (2*8)
+#define _bo  (3*8)
+#define _bu  (4*8)
+#define _ga  (5*8)
+#define _ge  (6*8)
+#define _gi  (7*8)
+#define _go  (8*8)
+#define _gu  (9*8)
+#define _ka (10*8)
+#define _ke (11*8)
+#define _ki (12*8)
+#define _ko (13*8)
+#define _ku (14*8)
+#define _ma (15*8)
+#define _me (16*8)
+#define _mi (17*8)
+#define _mo (18*8)
+#define _mu (19*8)
+#define _sa (20*8)
+#define _se (21*8)
+#define _si (22*8)
+#define _so (23*8)
+#define _su (24*8)
+
+// Temporary registers
+#define rT1  AX
+
+// Round vars
+#define rpState DI
+#define rpStack SP
+
+#define rDa BX
+#define rDe CX
+#define rDi DX
+#define rDo R8
+#define rDu R9
+
+#define rBa R10
+#define rBe R11
+#define rBi R12
+#define rBo R13
+#define rBu R14
+
+#define rCa SI
+#define rCe BP
+#define rCi rBi
+#define rCo rBo
+#define rCu R15
+
+#define MOVQ_RBI_RCE MOVQ rBi, rCe
+#define XORQ_RT1_RCA XORQ rT1, rCa
+#define XORQ_RT1_RCE XORQ rT1, rCe
+#define XORQ_RBA_RCU XORQ rBa, rCu
+#define XORQ_RBE_RCU XORQ rBe, rCu
+#define XORQ_RDU_RCU XORQ rDu, rCu
+#define XORQ_RDA_RCA XORQ rDa, rCa
+#define XORQ_RDE_RCE XORQ rDe, rCe
+
+#define mKeccakRound(iState, oState, rc, B_RBI_RCE, G_RT1_RCA, G_RT1_RCE, G_RBA_RCU, K_RT1_RCA, K_RT1_RCE, K_RBA_RCU, M_RT1_RCA, M_RT1_RCE, M_RBE_RCU, S_RDU_RCU, S_RDA_RCA, S_RDE_RCE) \
+	/* Prepare round */    \
+	MOVQ rCe, rDa;         \
+	ROLQ $1, rDa;          \
+	                       \
+	MOVQ _bi(iState), rCi; \
+	XORQ _gi(iState), rDi; \
+	XORQ rCu, rDa;         \
+	XORQ _ki(iState), rCi; \
+	XORQ _mi(iState), rDi; \
+	XORQ rDi, rCi;         \
+	                       \
+	MOVQ rCi, rDe;         \
+	ROLQ $1, rDe;          \
+	                       \
+	MOVQ _bo(iState), rCo; \
+	XORQ _go(iState), rDo; \
+	XORQ rCa, rDe;         \
+	XORQ _ko(iState), rCo; \
+	XORQ _mo(iState), rDo; \
+	XORQ rDo, rCo;         \
+	                       \
+	MOVQ rCo, rDi;         \
+	ROLQ $1, rDi;          \
+	                       \
+	MOVQ rCu, rDo;         \
+	XORQ rCe, rDi;         \
+	ROLQ $1, rDo;          \
+	                       \
+	MOVQ rCa, rDu;         \
+	XORQ rCi, rDo;         \
+	ROLQ $1, rDu;          \
+	                       \
+	/* Result b */         \
+	MOVQ _ba(iState), rBa; \
+	MOVQ _ge(iState), rBe; \
+	XORQ rCo, rDu;         \
+	MOVQ _ki(iState), rBi; \
+	MOVQ _mo(iState), rBo; \
+	MOVQ _su(iState), rBu; \
+	XORQ rDe, rBe;         \
+	ROLQ $44, rBe;         \
+	XORQ rDi, rBi;         \
+	XORQ rDa, rBa;         \
+	ROLQ $43, rBi;         \
+	                       \
+	MOVQ rBe, rCa;         \
+	MOVQ rc, rT1;          \
+	ORQ  rBi, rCa;         \
+	XORQ rBa, rT1;         \
+	XORQ rT1, rCa;         \
+	MOVQ rCa, _ba(oState); \
+	                       \
+	XORQ rDu, rBu;         \
+	ROLQ $14, rBu;         \
+	MOVQ rBa, rCu;         \
+	ANDQ rBe, rCu;         \
+	XORQ rBu, rCu;         \
+	MOVQ rCu, _bu(oState); \
+	                       \
+	XORQ rDo, rBo;         \
+	ROLQ $21, rBo;         \
+	MOVQ rBo, rT1;         \
+	ANDQ rBu, rT1;         \
+	XORQ rBi, rT1;         \
+	MOVQ rT1, _bi(oState); \
+	                       \
+	NOTQ rBi;              \
+	ORQ  rBa, rBu;         \
+	ORQ  rBo, rBi;         \
+	XORQ rBo, rBu;         \
+	XORQ rBe, rBi;         \
+	MOVQ rBu, _bo(oState); \
+	MOVQ rBi, _be(oState); \
+	B_RBI_RCE;             \
+	                       \
+	/* Result g */         \
+	MOVQ _gu(iState), rBe; \
+	XORQ rDu, rBe;         \
+	MOVQ _ka(iState), rBi; \
+	ROLQ $20, rBe;         \
+	XORQ rDa, rBi;         \
+	ROLQ $3, rBi;          \
+	MOVQ _bo(iState), rBa; \
+	MOVQ rBe, rT1;         \
+	ORQ  rBi, rT1;         \
+	XORQ rDo, rBa;         \
+	MOVQ _me(iState), rBo; \
+	MOVQ _si(iState), rBu; \
+	ROLQ $28, rBa;         \
+	XORQ rBa, rT1;         \
+	MOVQ rT1, _ga(oState); \
+	G_RT1_RCA;             \
+	                       \
+	XORQ rDe, rBo;         \
+	ROLQ $45, rBo;         \
+	MOVQ rBi, rT1;         \
+	ANDQ rBo, rT1;         \
+	XORQ rBe, rT1;         \
+	MOVQ rT1, _ge(oState); \
+	G_RT1_RCE;             \
+	                       \
+	XORQ rDi, rBu;         \
+	ROLQ $61, rBu;         \
+	MOVQ rBu, rT1;         \
+	ORQ  rBa, rT1;         \
+	XORQ rBo, rT1;         \
+	MOVQ rT1, _go(oState); \
+	                       \
+	ANDQ rBe, rBa;         \
+	XORQ rBu, rBa;         \
+	MOVQ rBa, _gu(oState); \
+	NOTQ rBu;              \
+	G_RBA_RCU;             \
+	                       \
+	ORQ  rBu, rBo;         \
+	XORQ rBi, rBo;         \
+	MOVQ rBo, _gi(oState); \
+	                       \
+	/* Result k */         \
+	MOVQ _be(iState), rBa; \
+	MOVQ _gi(iState), rBe; \
+	MOVQ _ko(iState), rBi; \
+	MOVQ _mu(iState), rBo; \
+	MOVQ _sa(iState), rBu; \
+	XORQ rDi, rBe;         \
+	ROLQ $6, rBe;          \
+	XORQ rDo, rBi;         \
+	ROLQ $25, rBi;         \
+	MOVQ rBe, rT1;         \
+	ORQ  rBi, rT1;         \
+	XORQ rDe, rBa;         \
+	ROLQ $1, rBa;          \
+	XORQ rBa, rT1;         \
+	MOVQ rT1, _ka(oState); \
+	K_RT1_RCA;             \
+	                       \
+	XORQ rDu, rBo;         \
+	ROLQ $8, rBo;          \
+	MOVQ rBi, rT1;         \
+	ANDQ rBo, rT1;         \
+	XORQ rBe, rT1;         \
+	MOVQ rT1, _ke(oState); \
+	K_RT1_RCE;             \
+	                       \
+	XORQ rDa, rBu;         \
+	ROLQ $18, rBu;         \
+	NOTQ rBo;              \
+	MOVQ rBo, rT1;         \
+	ANDQ rBu, rT1;         \
+	XORQ rBi, rT1;         \
+	MOVQ rT1, _ki(oState); \
+	                       \
+	MOVQ rBu, rT1;         \
+	ORQ  rBa, rT1;         \
+	XORQ rBo, rT1;         \
+	MOVQ rT1, _ko(oState); \
+	                       \
+	ANDQ rBe, rBa;         \
+	XORQ rBu, rBa;         \
+	MOVQ rBa, _ku(oState); \
+	K_RBA_RCU;             \
+	                       \
+	/* Result m */         \
+	MOVQ _ga(iState), rBe; \
+	XORQ rDa, rBe;         \
+	MOVQ _ke(iState), rBi; \
+	ROLQ $36, rBe;         \
+	XORQ rDe, rBi;         \
+	MOVQ _bu(iState), rBa; \
+	ROLQ $10, rBi;         \
+	MOVQ rBe, rT1;         \
+	MOVQ _mi(iState), rBo; \
+	ANDQ rBi, rT1;         \
+	XORQ rDu, rBa;         \
+	MOVQ _so(iState), rBu; \
+	ROLQ $27, rBa;         \
+	XORQ rBa, rT1;         \
+	MOVQ rT1, _ma(oState); \
+	M_RT1_RCA;             \
+	                       \
+	XORQ rDi, rBo;         \
+	ROLQ $15, rBo;         \
+	MOVQ rBi, rT1;         \
+	ORQ  rBo, rT1;         \
+	XORQ rBe, rT1;         \
+	MOVQ rT1, _me(oState); \
+	M_RT1_RCE;             \
+	                       \
+	XORQ rDo, rBu;         \
+	ROLQ $56, rBu;         \
+	NOTQ rBo;              \
+	MOVQ rBo, rT1;         \
+	ORQ  rBu, rT1;         \
+	XORQ rBi, rT1;         \
+	MOVQ rT1, _mi(oState); \
+	                       \
+	ORQ  rBa, rBe;         \
+	XORQ rBu, rBe;         \
+	MOVQ rBe, _mu(oState); \
+	                       \
+	ANDQ rBa, rBu;         \
+	XORQ rBo, rBu;         \
+	MOVQ rBu, _mo(oState); \
+	M_RBE_RCU;             \
+	                       \
+	/* Result s */         \
+	MOVQ _bi(iState), rBa; \
+	MOVQ _go(iState), rBe; \
+	MOVQ _ku(iState), rBi; \
+	XORQ rDi, rBa;         \
+	MOVQ _ma(iState), rBo; \
+	ROLQ $62, rBa;         \
+	XORQ rDo, rBe;         \
+	MOVQ _se(iState), rBu; \
+	ROLQ $55, rBe;         \
+	                       \
+	XORQ rDu, rBi;         \
+	MOVQ rBa, rDu;         \
+	XORQ rDe, rBu;         \
+	ROLQ $2, rBu;          \
+	ANDQ rBe, rDu;         \
+	XORQ rBu, rDu;         \
+	MOVQ rDu, _su(oState); \
+	                       \
+	ROLQ $39, rBi;         \
+	S_RDU_RCU;             \
+	NOTQ rBe;              \
+	XORQ rDa, rBo;         \
+	MOVQ rBe, rDa;         \
+	ANDQ rBi, rDa;         \
+	XORQ rBa, rDa;         \
+	MOVQ rDa, _sa(oState); \
+	S_RDA_RCA;             \
+	                       \
+	ROLQ $41, rBo;         \
+	MOVQ rBi, rDe;         \
+	ORQ  rBo, rDe;         \
+	XORQ rBe, rDe;         \
+	MOVQ rDe, _se(oState); \
+	S_RDE_RCE;             \
+	                       \
+	MOVQ rBo, rDi;         \
+	MOVQ rBu, rDo;         \
+	ANDQ rBu, rDi;         \
+	ORQ  rBa, rDo;         \
+	XORQ rBi, rDi;         \
+	XORQ rBo, rDo;         \
+	MOVQ rDi, _si(oState); \
+	MOVQ rDo, _so(oState)  \
+
+// func keccakF1600(state *[25]uint64)
+TEXT ·keccakF1600(SB), 0, $200-8
+	MOVQ state+0(FP), rpState
+
+	// Convert the user state into an internal state
+	NOTQ _be(rpState)
+	NOTQ _bi(rpState)
+	NOTQ _go(rpState)
+	NOTQ _ki(rpState)
+	NOTQ _mi(rpState)
+	NOTQ _sa(rpState)
+
+	// Execute the KeccakF permutation
+	MOVQ _ba(rpState), rCa
+	MOVQ _be(rpState), rCe
+	MOVQ _bu(rpState), rCu
+
+	XORQ _ga(rpState), rCa
+	XORQ _ge(rpState), rCe
+	XORQ _gu(rpState), rCu
+
+	XORQ _ka(rpState), rCa
+	XORQ _ke(rpState), rCe
+	XORQ _ku(rpState), rCu
+
+	XORQ _ma(rpState), rCa
+	XORQ _me(rpState), rCe
+	XORQ _mu(rpState), rCu
+
+	XORQ _sa(rpState), rCa
+	XORQ _se(rpState), rCe
+	MOVQ _si(rpState), rDi
+	MOVQ _so(rpState), rDo
+	XORQ _su(rpState), rCu
+
+	mKeccakRound(rpState, rpStack, $0x0000000000000001, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpStack, rpState, $0x0000000000008082, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpState, rpStack, $0x800000000000808a, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpStack, rpState, $0x8000000080008000, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpState, rpStack, $0x000000000000808b, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpStack, rpState, $0x0000000080000001, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpState, rpStack, $0x8000000080008081, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpStack, rpState, $0x8000000000008009, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpState, rpStack, $0x000000000000008a, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpStack, rpState, $0x0000000000000088, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpState, rpStack, $0x0000000080008009, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpStack, rpState, $0x000000008000000a, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpState, rpStack, $0x000000008000808b, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpStack, rpState, $0x800000000000008b, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpState, rpStack, $0x8000000000008089, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpStack, rpState, $0x8000000000008003, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpState, rpStack, $0x8000000000008002, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpStack, rpState, $0x8000000000000080, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpState, rpStack, $0x000000000000800a, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpStack, rpState, $0x800000008000000a, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpState, rpStack, $0x8000000080008081, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpStack, rpState, $0x8000000000008080, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpState, rpStack, $0x0000000080000001, MOVQ_RBI_RCE, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBA_RCU, XORQ_RT1_RCA, XORQ_RT1_RCE, XORQ_RBE_RCU, XORQ_RDU_RCU, XORQ_RDA_RCA, XORQ_RDE_RCE)
+	mKeccakRound(rpStack, rpState, $0x8000000080008008, NOP, NOP, NOP, NOP, NOP, NOP, NOP, NOP, NOP, NOP, NOP, NOP, NOP)
+
+	// Revert the internal state to the user state
+	NOTQ _be(rpState)
+	NOTQ _bi(rpState)
+	NOTQ _go(rpState)
+	NOTQ _ki(rpState)
+	NOTQ _mi(rpState)
+	NOTQ _sa(rpState)
+
+	RET

vendor/golang.org/x/crypto/sha3/register.go

@@ -0,0 +1,19 @@
+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build go1.4
+// +build go1.4
+
+package sha3
+
+import (
+	"crypto"
+)
+
+func init() {
+	crypto.RegisterHash(crypto.SHA3_224, New224)
+	crypto.RegisterHash(crypto.SHA3_256, New256)
+	crypto.RegisterHash(crypto.SHA3_384, New384)
+	crypto.RegisterHash(crypto.SHA3_512, New512)
+}

vendor/golang.org/x/crypto/sha3/sha3.go

@@ -0,0 +1,193 @@
+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package sha3
+
+// spongeDirection indicates the direction bytes are flowing through the sponge.
+type spongeDirection int
+
+const (
+	// spongeAbsorbing indicates that the sponge is absorbing input.
+	spongeAbsorbing spongeDirection = iota
+	// spongeSqueezing indicates that the sponge is being squeezed.
+	spongeSqueezing
+)
+
+const (
+	// maxRate is the maximum size of the internal buffer. SHAKE-256
+	// currently needs the largest buffer.
+	maxRate = 168
+)
+
+type state struct {
+	// Generic sponge components.
+	a    [25]uint64 // main state of the hash
+	buf  []byte     // points into storage
+	rate int        // the number of bytes of state to use
+
+	// dsbyte contains the "domain separation" bits and the first bit of
+	// the padding. Sections 6.1 and 6.2 of [1] separate the outputs of the
+	// SHA-3 and SHAKE functions by appending bitstrings to the message.
+	// Using a little-endian bit-ordering convention, these are "01" for SHA-3
+	// and "1111" for SHAKE, or 00000010b and 00001111b, respectively. Then the
+	// padding rule from section 5.1 is applied to pad the message to a multiple
+	// of the rate, which involves adding a "1" bit, zero or more "0" bits, and
+	// a final "1" bit. We merge the first "1" bit from the padding into dsbyte,
+	// giving 00000110b (0x06) and 00011111b (0x1f).
+	// [1] http://csrc.nist.gov/publications/drafts/fips-202/fips_202_draft.pdf
+	//     "Draft FIPS 202: SHA-3 Standard: Permutation-Based Hash and
+	//      Extendable-Output Functions (May 2014)"
+	dsbyte byte
+
+	storage storageBuf
+
+	// Specific to SHA-3 and SHAKE.
+	outputLen int             // the default output size in bytes
+	state     spongeDirection // whether the sponge is absorbing or squeezing
+}
+
+// BlockSize returns the rate of sponge underlying this hash function.
+func (d *state) BlockSize() int { return d.rate }
+
+// Size returns the output size of the hash function in bytes.
+func (d *state) Size() int { return d.outputLen }
+
+// Reset clears the internal state by zeroing the sponge state and
+// the byte buffer, and setting Sponge.state to absorbing.
+func (d *state) Reset() {
+	// Zero the permutation's state.
+	for i := range d.a {
+		d.a[i] = 0
+	}
+	d.state = spongeAbsorbing
+	d.buf = d.storage.asBytes()[:0]
+}
+
+func (d *state) clone() *state {
+	ret := *d
+	if ret.state == spongeAbsorbing {
+		ret.buf = ret.storage.asBytes()[:len(ret.buf)]
+	} else {
+		ret.buf = ret.storage.asBytes()[d.rate-cap(d.buf) : d.rate]
+	}
+
+	return &ret
+}
+
+// permute applies the KeccakF-1600 permutation. It handles
+// any input-output buffering.
+func (d *state) permute() {
+	switch d.state {
+	case spongeAbsorbing:
+		// If we're absorbing, we need to xor the input into the state
+		// before applying the permutation.
+		xorIn(d, d.buf)
+		d.buf = d.storage.asBytes()[:0]
+		keccakF1600(&d.a)
+	case spongeSqueezing:
+		// If we're squeezing, we need to apply the permutatin before
+		// copying more output.
+		keccakF1600(&d.a)
+		d.buf = d.storage.asBytes()[:d.rate]
+		copyOut(d, d.buf)
+	}
+}
+
+// pads appends the domain separation bits in dsbyte, applies
+// the multi-bitrate 10..1 padding rule, and permutes the state.
+func (d *state) padAndPermute(dsbyte byte) {
+	if d.buf == nil {
+		d.buf = d.storage.asBytes()[:0]
+	}
+	// Pad with this instance's domain-separator bits. We know that there's
+	// at least one byte of space in d.buf because, if it were full,
+	// permute would have been called to empty it. dsbyte also contains the
+	// first one bit for the padding. See the comment in the state struct.
+	d.buf = append(d.buf, dsbyte)
+	zerosStart := len(d.buf)
+	d.buf = d.storage.asBytes()[:d.rate]
+	for i := zerosStart; i < d.rate; i++ {
+		d.buf[i] = 0
+	}
+	// This adds the final one bit for the padding. Because of the way that
+	// bits are numbered from the LSB upwards, the final bit is the MSB of
+	// the last byte.
+	d.buf[d.rate-1] ^= 0x80
+	// Apply the permutation
+	d.permute()
+	d.state = spongeSqueezing
+	d.buf = d.storage.asBytes()[:d.rate]
+	copyOut(d, d.buf)
+}
+
+// Write absorbs more data into the hash's state. It produces an error
+// if more data is written to the ShakeHash after writing
+func (d *state) Write(p []byte) (written int, err error) {
+	if d.state != spongeAbsorbing {
+		panic("sha3: write to sponge after read")
+	}
+	if d.buf == nil {
+		d.buf = d.storage.asBytes()[:0]
+	}
+	written = len(p)
+
+	for len(p) > 0 {
+		if len(d.buf) == 0 && len(p) >= d.rate {
+			// The fast path; absorb a full "rate" bytes of input and apply the permutation.
+			xorIn(d, p[:d.rate])
+			p = p[d.rate:]
+			keccakF1600(&d.a)
+		} else {
+			// The slow path; buffer the input until we can fill the sponge, and then xor it in.
+			todo := d.rate - len(d.buf)
+			if todo > len(p) {
+				todo = len(p)
+			}
+			d.buf = append(d.buf, p[:todo]...)
+			p = p[todo:]
+
+			// If the sponge is full, apply the permutation.
+			if len(d.buf) == d.rate {
+				d.permute()
+			}
+		}
+	}
+
+	return
+}
+
+// Read squeezes an arbitrary number of bytes from the sponge.
+func (d *state) Read(out []byte) (n int, err error) {
+	// If we're still absorbing, pad and apply the permutation.
+	if d.state == spongeAbsorbing {
+		d.padAndPermute(d.dsbyte)
+	}
+
+	n = len(out)
+
+	// Now, do the squeezing.
+	for len(out) > 0 {
+		n := copy(out, d.buf)
+		d.buf = d.buf[n:]
+		out = out[n:]
+
+		// Apply the permutation if we've squeezed the sponge dry.
+		if len(d.buf) == 0 {
+			d.permute()
+		}
+	}
+
+	return
+}
+
+// Sum applies padding to the hash state and then squeezes out the desired
+// number of output bytes.
+func (d *state) Sum(in []byte) []byte {
+	// Make a copy of the original hash so that caller can keep writing
+	// and summing.
+	dup := d.clone()
+	hash := make([]byte, dup.outputLen)
+	dup.Read(hash)
+	return append(in, hash...)
+}

vendor/golang.org/x/crypto/sha3/sha3_s390x.go

@@ -0,0 +1,285 @@
+// Copyright 2017 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build gc && !purego
+// +build gc,!purego
+
+package sha3
+
+// This file contains code for using the 'compute intermediate
+// message digest' (KIMD) and 'compute last message digest' (KLMD)
+// instructions to compute SHA-3 and SHAKE hashes on IBM Z.
+
+import (
+	"hash"
+
+	"golang.org/x/sys/cpu"
+)
+
+// codes represent 7-bit KIMD/KLMD function codes as defined in
+// the Principles of Operation.
+type code uint64
+
+const (
+	// function codes for KIMD/KLMD
+	sha3_224  code = 32
+	sha3_256       = 33
+	sha3_384       = 34
+	sha3_512       = 35
+	shake_128      = 36
+	shake_256      = 37
+	nopad          = 0x100
+)
+
+// kimd is a wrapper for the 'compute intermediate message digest' instruction.
+// src must be a multiple of the rate for the given function code.
+//go:noescape
+func kimd(function code, chain *[200]byte, src []byte)
+
+// klmd is a wrapper for the 'compute last message digest' instruction.
+// src padding is handled by the instruction.
+//go:noescape
+func klmd(function code, chain *[200]byte, dst, src []byte)
+
+type asmState struct {
+	a         [200]byte       // 1600 bit state
+	buf       []byte          // care must be taken to ensure cap(buf) is a multiple of rate
+	rate      int             // equivalent to block size
+	storage   [3072]byte      // underlying storage for buf
+	outputLen int             // output length if fixed, 0 if not
+	function  code            // KIMD/KLMD function code
+	state     spongeDirection // whether the sponge is absorbing or squeezing
+}
+
+func newAsmState(function code) *asmState {
+	var s asmState
+	s.function = function
+	switch function {
+	case sha3_224:
+		s.rate = 144
+		s.outputLen = 28
+	case sha3_256:
+		s.rate = 136
+		s.outputLen = 32
+	case sha3_384:
+		s.rate = 104
+		s.outputLen = 48
+	case sha3_512:
+		s.rate = 72
+		s.outputLen = 64
+	case shake_128:
+		s.rate = 168
+	case shake_256:
+		s.rate = 136
+	default:
+		panic("sha3: unrecognized function code")
+	}
+
+	// limit s.buf size to a multiple of s.rate
+	s.resetBuf()
+	return &s
+}
+
+func (s *asmState) clone() *asmState {
+	c := *s
+	c.buf = c.storage[:len(s.buf):cap(s.buf)]
+	return &c
+}
+
+// copyIntoBuf copies b into buf. It will panic if there is not enough space to
+// store all of b.
+func (s *asmState) copyIntoBuf(b []byte) {
+	bufLen := len(s.buf)
+	s.buf = s.buf[:len(s.buf)+len(b)]
+	copy(s.buf[bufLen:], b)
+}
+
+// resetBuf points buf at storage, sets the length to 0 and sets cap to be a
+// multiple of the rate.
+func (s *asmState) resetBuf() {
+	max := (cap(s.storage) / s.rate) * s.rate
+	s.buf = s.storage[:0:max]
+}
+
+// Write (via the embedded io.Writer interface) adds more data to the running hash.
+// It never returns an error.
+func (s *asmState) Write(b []byte) (int, error) {
+	if s.state != spongeAbsorbing {
+		panic("sha3: write to sponge after read")
+	}
+	length := len(b)
+	for len(b) > 0 {
+		if len(s.buf) == 0 && len(b) >= cap(s.buf) {
+			// Hash the data directly and push any remaining bytes
+			// into the buffer.
+			remainder := len(b) % s.rate
+			kimd(s.function, &s.a, b[:len(b)-remainder])
+			if remainder != 0 {
+				s.copyIntoBuf(b[len(b)-remainder:])
+			}
+			return length, nil
+		}
+
+		if len(s.buf) == cap(s.buf) {
+			// flush the buffer
+			kimd(s.function, &s.a, s.buf)
+			s.buf = s.buf[:0]
+		}
+
+		// copy as much as we can into the buffer
+		n := len(b)
+		if len(b) > cap(s.buf)-len(s.buf) {
+			n = cap(s.buf) - len(s.buf)
+		}
+		s.copyIntoBuf(b[:n])
+		b = b[n:]
+	}
+	return length, nil
+}
+
+// Read squeezes an arbitrary number of bytes from the sponge.
+func (s *asmState) Read(out []byte) (n int, err error) {
+	n = len(out)
+
+	// need to pad if we were absorbing
+	if s.state == spongeAbsorbing {
+		s.state = spongeSqueezing
+
+		// write hash directly into out if possible
+		if len(out)%s.rate == 0 {
+			klmd(s.function, &s.a, out, s.buf) // len(out) may be 0
+			s.buf = s.buf[:0]
+			return
+		}
+
+		// write hash into buffer
+		max := cap(s.buf)
+		if max > len(out) {
+			max = (len(out)/s.rate)*s.rate + s.rate
+		}
+		klmd(s.function, &s.a, s.buf[:max], s.buf)
+		s.buf = s.buf[:max]
+	}
+
+	for len(out) > 0 {
+		// flush the buffer
+		if len(s.buf) != 0 {
+			c := copy(out, s.buf)
+			out = out[c:]
+			s.buf = s.buf[c:]
+			continue
+		}
+
+		// write hash directly into out if possible
+		if len(out)%s.rate == 0 {
+			klmd(s.function|nopad, &s.a, out, nil)
+			return
+		}
+
+		// write hash into buffer
+		s.resetBuf()
+		if cap(s.buf) > len(out) {
+			s.buf = s.buf[:(len(out)/s.rate)*s.rate+s.rate]
+		}
+		klmd(s.function|nopad, &s.a, s.buf, nil)
+	}
+	return
+}
+
+// Sum appends the current hash to b and returns the resulting slice.
+// It does not change the underlying hash state.
+func (s *asmState) Sum(b []byte) []byte {
+	if s.outputLen == 0 {
+		panic("sha3: cannot call Sum on SHAKE functions")
+	}
+
+	// Copy the state to preserve the original.
+	a := s.a
+
+	// Hash the buffer. Note that we don't clear it because we
+	// aren't updating the state.
+	klmd(s.function, &a, nil, s.buf)
+	return append(b, a[:s.outputLen]...)
+}
+
+// Reset resets the Hash to its initial state.
+func (s *asmState) Reset() {
+	for i := range s.a {
+		s.a[i] = 0
+	}
+	s.resetBuf()
+	s.state = spongeAbsorbing
+}
+
+// Size returns the number of bytes Sum will return.
+func (s *asmState) Size() int {
+	return s.outputLen
+}
+
+// BlockSize returns the hash's underlying block size.
+// The Write method must be able to accept any amount
+// of data, but it may operate more efficiently if all writes
+// are a multiple of the block size.
+func (s *asmState) BlockSize() int {
+	return s.rate
+}
+
+// Clone returns a copy of the ShakeHash in its current state.
+func (s *asmState) Clone() ShakeHash {
+	return s.clone()
+}
+
+// new224Asm returns an assembly implementation of SHA3-224 if available,
+// otherwise it returns nil.
+func new224Asm() hash.Hash {
+	if cpu.S390X.HasSHA3 {
+		return newAsmState(sha3_224)
+	}
+	return nil
+}
+
+// new256Asm returns an assembly implementation of SHA3-256 if available,
+// otherwise it returns nil.
+func new256Asm() hash.Hash {
+	if cpu.S390X.HasSHA3 {
+		return newAsmState(sha3_256)
+	}
+	return nil
+}
+
+// new384Asm returns an assembly implementation of SHA3-384 if available,
+// otherwise it returns nil.
+func new384Asm() hash.Hash {
+	if cpu.S390X.HasSHA3 {
+		return newAsmState(sha3_384)
+	}
+	return nil
+}
+
+// new512Asm returns an assembly implementation of SHA3-512 if available,
+// otherwise it returns nil.
+func new512Asm() hash.Hash {
+	if cpu.S390X.HasSHA3 {
+		return newAsmState(sha3_512)
+	}
+	return nil
+}
+
+// newShake128Asm returns an assembly implementation of SHAKE-128 if available,
+// otherwise it returns nil.
+func newShake128Asm() ShakeHash {
+	if cpu.S390X.HasSHA3 {
+		return newAsmState(shake_128)
+	}
+	return nil
+}
+
+// newShake256Asm returns an assembly implementation of SHAKE-256 if available,
+// otherwise it returns nil.
+func newShake256Asm() ShakeHash {
+	if cpu.S390X.HasSHA3 {
+		return newAsmState(shake_256)
+	}
+	return nil
+}

vendor/golang.org/x/crypto/sha3/sha3_s390x.s

@@ -0,0 +1,34 @@
+// Copyright 2017 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build gc && !purego
+// +build gc,!purego
+
+#include "textflag.h"
+
+// func kimd(function code, chain *[200]byte, src []byte)
+TEXT ·kimd(SB), NOFRAME|NOSPLIT, $0-40
+	MOVD function+0(FP), R0
+	MOVD chain+8(FP), R1
+	LMG  src+16(FP), R2, R3 // R2=base, R3=len
+
+continue:
+	WORD $0xB93E0002 // KIMD --, R2
+	BVS  continue    // continue if interrupted
+	MOVD $0, R0      // reset R0 for pre-go1.8 compilers
+	RET
+
+// func klmd(function code, chain *[200]byte, dst, src []byte)
+TEXT ·klmd(SB), NOFRAME|NOSPLIT, $0-64
+	// TODO: SHAKE support
+	MOVD function+0(FP), R0
+	MOVD chain+8(FP), R1
+	LMG  dst+16(FP), R2, R3 // R2=base, R3=len
+	LMG  src+40(FP), R4, R5 // R4=base, R5=len
+
+continue:
+	WORD $0xB93F0024 // KLMD R2, R4
+	BVS  continue    // continue if interrupted
+	MOVD $0, R0      // reset R0 for pre-go1.8 compilers
+	RET

vendor/golang.org/x/crypto/sha3/shake.go

@@ -0,0 +1,173 @@
+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package sha3
+
+// This file defines the ShakeHash interface, and provides
+// functions for creating SHAKE and cSHAKE instances, as well as utility
+// functions for hashing bytes to arbitrary-length output.
+//
+//
+// SHAKE implementation is based on FIPS PUB 202 [1]
+// cSHAKE implementations is based on NIST SP 800-185 [2]
+//
+// [1] https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf
+// [2] https://doi.org/10.6028/NIST.SP.800-185
+
+import (
+	"encoding/binary"
+	"io"
+)
+
+// ShakeHash defines the interface to hash functions that
+// support arbitrary-length output.
+type ShakeHash interface {
+	// Write absorbs more data into the hash's state. It panics if input is
+	// written to it after output has been read from it.
+	io.Writer
+
+	// Read reads more output from the hash; reading affects the hash's
+	// state. (ShakeHash.Read is thus very different from Hash.Sum)
+	// It never returns an error.
+	io.Reader
+
+	// Clone returns a copy of the ShakeHash in its current state.
+	Clone() ShakeHash
+
+	// Reset resets the ShakeHash to its initial state.
+	Reset()
+}
+
+// cSHAKE specific context
+type cshakeState struct {
+	*state // SHA-3 state context and Read/Write operations
+
+	// initBlock is the cSHAKE specific initialization set of bytes. It is initialized
+	// by newCShake function and stores concatenation of N followed by S, encoded
+	// by the method specified in 3.3 of [1].
+	// It is stored here in order for Reset() to be able to put context into
+	// initial state.
+	initBlock []byte
+}
+
+// Consts for configuring initial SHA-3 state
+const (
+	dsbyteShake  = 0x1f
+	dsbyteCShake = 0x04
+	rate128      = 168
+	rate256      = 136
+)
+
+func bytepad(input []byte, w int) []byte {
+	// leftEncode always returns max 9 bytes
+	buf := make([]byte, 0, 9+len(input)+w)
+	buf = append(buf, leftEncode(uint64(w))...)
+	buf = append(buf, input...)
+	padlen := w - (len(buf) % w)
+	return append(buf, make([]byte, padlen)...)
+}
+
+func leftEncode(value uint64) []byte {
+	var b [9]byte
+	binary.BigEndian.PutUint64(b[1:], value)
+	// Trim all but last leading zero bytes
+	i := byte(1)
+	for i < 8 && b[i] == 0 {
+		i++
+	}
+	// Prepend number of encoded bytes
+	b[i-1] = 9 - i
+	return b[i-1:]
+}
+
+func newCShake(N, S []byte, rate int, dsbyte byte) ShakeHash {
+	c := cshakeState{state: &state{rate: rate, dsbyte: dsbyte}}
+
+	// leftEncode returns max 9 bytes
+	c.initBlock = make([]byte, 0, 9*2+len(N)+len(S))
+	c.initBlock = append(c.initBlock, leftEncode(uint64(len(N)*8))...)
+	c.initBlock = append(c.initBlock, N...)
+	c.initBlock = append(c.initBlock, leftEncode(uint64(len(S)*8))...)
+	c.initBlock = append(c.initBlock, S...)
+	c.Write(bytepad(c.initBlock, c.rate))
+	return &c
+}
+
+// Reset resets the hash to initial state.
+func (c *cshakeState) Reset() {
+	c.state.Reset()
+	c.Write(bytepad(c.initBlock, c.rate))
+}
+
+// Clone returns copy of a cSHAKE context within its current state.
+func (c *cshakeState) Clone() ShakeHash {
+	b := make([]byte, len(c.initBlock))
+	copy(b, c.initBlock)
+	return &cshakeState{state: c.clone(), initBlock: b}
+}
+
+// Clone returns copy of SHAKE context within its current state.
+func (c *state) Clone() ShakeHash {
+	return c.clone()
+}
+
+// NewShake128 creates a new SHAKE128 variable-output-length ShakeHash.
+// Its generic security strength is 128 bits against all attacks if at
+// least 32 bytes of its output are used.
+func NewShake128() ShakeHash {
+	if h := newShake128Asm(); h != nil {
+		return h
+	}
+	return &state{rate: rate128, dsbyte: dsbyteShake}
+}
+
+// NewShake256 creates a new SHAKE256 variable-output-length ShakeHash.
+// Its generic security strength is 256 bits against all attacks if
+// at least 64 bytes of its output are used.
+func NewShake256() ShakeHash {
+	if h := newShake256Asm(); h != nil {
+		return h
+	}
+	return &state{rate: rate256, dsbyte: dsbyteShake}
+}
+
+// NewCShake128 creates a new instance of cSHAKE128 variable-output-length ShakeHash,
+// a customizable variant of SHAKE128.
+// N is used to define functions based on cSHAKE, it can be empty when plain cSHAKE is
+// desired. S is a customization byte string used for domain separation - two cSHAKE
+// computations on same input with different S yield unrelated outputs.
+// When N and S are both empty, this is equivalent to NewShake128.
+func NewCShake128(N, S []byte) ShakeHash {
+	if len(N) == 0 && len(S) == 0 {
+		return NewShake128()
+	}
+	return newCShake(N, S, rate128, dsbyteCShake)
+}
+
+// NewCShake256 creates a new instance of cSHAKE256 variable-output-length ShakeHash,
+// a customizable variant of SHAKE256.
+// N is used to define functions based on cSHAKE, it can be empty when plain cSHAKE is
+// desired. S is a customization byte string used for domain separation - two cSHAKE
+// computations on same input with different S yield unrelated outputs.
+// When N and S are both empty, this is equivalent to NewShake256.
+func NewCShake256(N, S []byte) ShakeHash {
+	if len(N) == 0 && len(S) == 0 {
+		return NewShake256()
+	}
+	return newCShake(N, S, rate256, dsbyteCShake)
+}
+
+// ShakeSum128 writes an arbitrary-length digest of data into hash.
+func ShakeSum128(hash, data []byte) {
+	h := NewShake128()
+	h.Write(data)
+	h.Read(hash)
+}
+
+// ShakeSum256 writes an arbitrary-length digest of data into hash.
+func ShakeSum256(hash, data []byte) {
+	h := NewShake256()
+	h.Write(data)
+	h.Read(hash)
+}

vendor/golang.org/x/crypto/sha3/shake_generic.go

@@ -0,0 +1,20 @@
+// Copyright 2017 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build !gc || purego || !s390x
+// +build !gc purego !s390x
+
+package sha3
+
+// newShake128Asm returns an assembly implementation of SHAKE-128 if available,
+// otherwise it returns nil.
+func newShake128Asm() ShakeHash {
+	return nil
+}
+
+// newShake256Asm returns an assembly implementation of SHAKE-256 if available,
+// otherwise it returns nil.
+func newShake256Asm() ShakeHash {
+	return nil
+}

vendor/golang.org/x/crypto/sha3/xor.go

@@ -0,0 +1,24 @@
+// Copyright 2015 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build (!amd64 && !386 && !ppc64le) || purego
+// +build !amd64,!386,!ppc64le purego
+
+package sha3
+
+// A storageBuf is an aligned array of maxRate bytes.
+type storageBuf [maxRate]byte
+
+func (b *storageBuf) asBytes() *[maxRate]byte {
+	return (*[maxRate]byte)(b)
+}
+
+var (
+	xorIn            = xorInGeneric
+	copyOut          = copyOutGeneric
+	xorInUnaligned   = xorInGeneric
+	copyOutUnaligned = copyOutGeneric
+)
+
+const xorImplementationUnaligned = "generic"

vendor/golang.org/x/crypto/sha3/xor_generic.go

@@ -0,0 +1,28 @@
+// Copyright 2015 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package sha3
+
+import "encoding/binary"
+
+// xorInGeneric xors the bytes in buf into the state; it
+// makes no non-portable assumptions about memory layout
+// or alignment.
+func xorInGeneric(d *state, buf []byte) {
+	n := len(buf) / 8
+
+	for i := 0; i < n; i++ {
+		a := binary.LittleEndian.Uint64(buf)
+		d.a[i] ^= a
+		buf = buf[8:]
+	}
+}
+
+// copyOutGeneric copies uint64s to a byte buffer.
+func copyOutGeneric(d *state, b []byte) {
+	for i := 0; len(b) >= 8; i++ {
+		binary.LittleEndian.PutUint64(b, d.a[i])
+		b = b[8:]
+	}
+}

vendor/golang.org/x/crypto/sha3/xor_unaligned.go

@@ -0,0 +1,68 @@
+// Copyright 2015 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build (amd64 || 386 || ppc64le) && !purego
+// +build amd64 386 ppc64le
+// +build !purego
+
+package sha3
+
+import "unsafe"
+
+// A storageBuf is an aligned array of maxRate bytes.
+type storageBuf [maxRate / 8]uint64
+
+func (b *storageBuf) asBytes() *[maxRate]byte {
+	return (*[maxRate]byte)(unsafe.Pointer(b))
+}
+
+// xorInUnaligned uses unaligned reads and writes to update d.a to contain d.a
+// XOR buf.
+func xorInUnaligned(d *state, buf []byte) {
+	n := len(buf)
+	bw := (*[maxRate / 8]uint64)(unsafe.Pointer(&buf[0]))[: n/8 : n/8]
+	if n >= 72 {
+		d.a[0] ^= bw[0]
+		d.a[1] ^= bw[1]
+		d.a[2] ^= bw[2]
+		d.a[3] ^= bw[3]
+		d.a[4] ^= bw[4]
+		d.a[5] ^= bw[5]
+		d.a[6] ^= bw[6]
+		d.a[7] ^= bw[7]
+		d.a[8] ^= bw[8]
+	}
+	if n >= 104 {
+		d.a[9] ^= bw[9]
+		d.a[10] ^= bw[10]
+		d.a[11] ^= bw[11]
+		d.a[12] ^= bw[12]
+	}
+	if n >= 136 {
+		d.a[13] ^= bw[13]
+		d.a[14] ^= bw[14]
+		d.a[15] ^= bw[15]
+		d.a[16] ^= bw[16]
+	}
+	if n >= 144 {
+		d.a[17] ^= bw[17]
+	}
+	if n >= 168 {
+		d.a[18] ^= bw[18]
+		d.a[19] ^= bw[19]
+		d.a[20] ^= bw[20]
+	}
+}
+
+func copyOutUnaligned(d *state, buf []byte) {
+	ab := (*[maxRate]uint8)(unsafe.Pointer(&d.a[0]))
+	copy(buf, ab[:])
+}
+
+var (
+	xorIn   = xorInUnaligned
+	copyOut = copyOutUnaligned
+)
+
+const xorImplementationUnaligned = "unaligned"

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

@@ -0,0 +1,97 @@
+// 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 ssh
+
+import (
+	"io"
+	"sync"
+)
+
+// buffer provides a linked list buffer for data exchange
+// between producer and consumer. Theoretically the buffer is
+// of unlimited capacity as it does no allocation of its own.
+type buffer struct {
+	// protects concurrent access to head, tail and closed
+	*sync.Cond
+
+	head *element // the buffer that will be read first
+	tail *element // the buffer that will be read last
+
+	closed bool
+}
+
+// An element represents a single link in a linked list.
+type element struct {
+	buf  []byte
+	next *element
+}
+
+// newBuffer returns an empty buffer that is not closed.
+func newBuffer() *buffer {
+	e := new(element)
+	b := &buffer{
+		Cond: newCond(),
+		head: e,
+		tail: e,
+	}
+	return b
+}
+
+// write makes buf available for Read to receive.
+// buf must not be modified after the call to write.
+func (b *buffer) write(buf []byte) {
+	b.Cond.L.Lock()
+	e := &element{buf: buf}
+	b.tail.next = e
+	b.tail = e
+	b.Cond.Signal()
+	b.Cond.L.Unlock()
+}
+
+// eof closes the buffer. Reads from the buffer once all
+// the data has been consumed will receive io.EOF.
+func (b *buffer) eof() {
+	b.Cond.L.Lock()
+	b.closed = true
+	b.Cond.Signal()
+	b.Cond.L.Unlock()
+}
+
+// Read reads data from the internal buffer in buf.  Reads will block
+// if no data is available, or until the buffer is closed.
+func (b *buffer) Read(buf []byte) (n int, err error) {
+	b.Cond.L.Lock()
+	defer b.Cond.L.Unlock()
+
+	for len(buf) > 0 {
+		// if there is data in b.head, copy it
+		if len(b.head.buf) > 0 {
+			r := copy(buf, b.head.buf)
+			buf, b.head.buf = buf[r:], b.head.buf[r:]
+			n += r
+			continue
+		}
+		// if there is a next buffer, make it the head
+		if len(b.head.buf) == 0 && b.head != b.tail {
+			b.head = b.head.next
+			continue
+		}
+
+		// if at least one byte has been copied, return
+		if n > 0 {
+			break
+		}
+
+		// if nothing was read, and there is nothing outstanding
+		// check to see if the buffer is closed.
+		if b.closed {
+			err = io.EOF
+			break
+		}
+		// out of buffers, wait for producer
+		b.Cond.Wait()
+	}
+	return
+}

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

@@ -0,0 +1,546 @@
+// 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 ssh
+
+import (
+	"bytes"
+	"errors"
+	"fmt"
+	"io"
+	"net"
+	"sort"
+	"time"
+)
+
+// These constants from [PROTOCOL.certkeys] represent the algorithm names
+// for certificate types supported by this package.
+const (
+	CertAlgoRSAv01        = "ssh-rsa-cert-v01@openssh.com"
+	CertAlgoDSAv01        = "ssh-dss-cert-v01@openssh.com"
+	CertAlgoECDSA256v01   = "ecdsa-sha2-nistp256-cert-v01@openssh.com"
+	CertAlgoECDSA384v01   = "ecdsa-sha2-nistp384-cert-v01@openssh.com"
+	CertAlgoECDSA521v01   = "ecdsa-sha2-nistp521-cert-v01@openssh.com"
+	CertAlgoSKECDSA256v01 = "sk-ecdsa-sha2-nistp256-cert-v01@openssh.com"
+	CertAlgoED25519v01    = "ssh-ed25519-cert-v01@openssh.com"
+	CertAlgoSKED25519v01  = "sk-ssh-ed25519-cert-v01@openssh.com"
+)
+
+// Certificate types distinguish between host and user
+// certificates. The values can be set in the CertType field of
+// Certificate.
+const (
+	UserCert = 1
+	HostCert = 2
+)
+
+// Signature represents a cryptographic signature.
+type Signature struct {
+	Format string
+	Blob   []byte
+	Rest   []byte `ssh:"rest"`
+}
+
+// CertTimeInfinity can be used for OpenSSHCertV01.ValidBefore to indicate that
+// a certificate does not expire.
+const CertTimeInfinity = 1<<64 - 1
+
+// An Certificate represents an OpenSSH certificate as defined in
+// [PROTOCOL.certkeys]?rev=1.8. The Certificate type implements the
+// PublicKey interface, so it can be unmarshaled using
+// ParsePublicKey.
+type Certificate struct {
+	Nonce           []byte
+	Key             PublicKey
+	Serial          uint64
+	CertType        uint32
+	KeyId           string
+	ValidPrincipals []string
+	ValidAfter      uint64
+	ValidBefore     uint64
+	Permissions
+	Reserved     []byte
+	SignatureKey PublicKey
+	Signature    *Signature
+}
+
+// genericCertData holds the key-independent part of the certificate data.
+// Overall, certificates contain an nonce, public key fields and
+// key-independent fields.
+type genericCertData struct {
+	Serial          uint64
+	CertType        uint32
+	KeyId           string
+	ValidPrincipals []byte
+	ValidAfter      uint64
+	ValidBefore     uint64
+	CriticalOptions []byte
+	Extensions      []byte
+	Reserved        []byte
+	SignatureKey    []byte
+	Signature       []byte
+}
+
+func marshalStringList(namelist []string) []byte {
+	var to []byte
+	for _, name := range namelist {
+		s := struct{ N string }{name}
+		to = append(to, Marshal(&s)...)
+	}
+	return to
+}
+
+type optionsTuple struct {
+	Key   string
+	Value []byte
+}
+
+type optionsTupleValue struct {
+	Value string
+}
+
+// serialize a map of critical options or extensions
+// issue #10569 - per [PROTOCOL.certkeys] and SSH implementation,
+// we need two length prefixes for a non-empty string value
+func marshalTuples(tups map[string]string) []byte {
+	keys := make([]string, 0, len(tups))
+	for key := range tups {
+		keys = append(keys, key)
+	}
+	sort.Strings(keys)
+
+	var ret []byte
+	for _, key := range keys {
+		s := optionsTuple{Key: key}
+		if value := tups[key]; len(value) > 0 {
+			s.Value = Marshal(&optionsTupleValue{value})
+		}
+		ret = append(ret, Marshal(&s)...)
+	}
+	return ret
+}
+
+// issue #10569 - per [PROTOCOL.certkeys] and SSH implementation,
+// we need two length prefixes for a non-empty option value
+func parseTuples(in []byte) (map[string]string, error) {
+	tups := map[string]string{}
+	var lastKey string
+	var haveLastKey bool
+
+	for len(in) > 0 {
+		var key, val, extra []byte
+		var ok bool
+
+		if key, in, ok = parseString(in); !ok {
+			return nil, errShortRead
+		}
+		keyStr := string(key)
+		// according to [PROTOCOL.certkeys], the names must be in
+		// lexical order.
+		if haveLastKey && keyStr <= lastKey {
+			return nil, fmt.Errorf("ssh: certificate options are not in lexical order")
+		}
+		lastKey, haveLastKey = keyStr, true
+		// the next field is a data field, which if non-empty has a string embedded
+		if val, in, ok = parseString(in); !ok {
+			return nil, errShortRead
+		}
+		if len(val) > 0 {
+			val, extra, ok = parseString(val)
+			if !ok {
+				return nil, errShortRead
+			}
+			if len(extra) > 0 {
+				return nil, fmt.Errorf("ssh: unexpected trailing data after certificate option value")
+			}
+			tups[keyStr] = string(val)
+		} else {
+			tups[keyStr] = ""
+		}
+	}
+	return tups, nil
+}
+
+func parseCert(in []byte, privAlgo string) (*Certificate, error) {
+	nonce, rest, ok := parseString(in)
+	if !ok {
+		return nil, errShortRead
+	}
+
+	key, rest, err := parsePubKey(rest, privAlgo)
+	if err != nil {
+		return nil, err
+	}
+
+	var g genericCertData
+	if err := Unmarshal(rest, &g); err != nil {
+		return nil, err
+	}
+
+	c := &Certificate{
+		Nonce:       nonce,
+		Key:         key,
+		Serial:      g.Serial,
+		CertType:    g.CertType,
+		KeyId:       g.KeyId,
+		ValidAfter:  g.ValidAfter,
+		ValidBefore: g.ValidBefore,
+	}
+
+	for principals := g.ValidPrincipals; len(principals) > 0; {
+		principal, rest, ok := parseString(principals)
+		if !ok {
+			return nil, errShortRead
+		}
+		c.ValidPrincipals = append(c.ValidPrincipals, string(principal))
+		principals = rest
+	}
+
+	c.CriticalOptions, err = parseTuples(g.CriticalOptions)
+	if err != nil {
+		return nil, err
+	}
+	c.Extensions, err = parseTuples(g.Extensions)
+	if err != nil {
+		return nil, err
+	}
+	c.Reserved = g.Reserved
+	k, err := ParsePublicKey(g.SignatureKey)
+	if err != nil {
+		return nil, err
+	}
+
+	c.SignatureKey = k
+	c.Signature, rest, ok = parseSignatureBody(g.Signature)
+	if !ok || len(rest) > 0 {
+		return nil, errors.New("ssh: signature parse error")
+	}
+
+	return c, nil
+}
+
+type openSSHCertSigner struct {
+	pub    *Certificate
+	signer Signer
+}
+
+type algorithmOpenSSHCertSigner struct {
+	*openSSHCertSigner
+	algorithmSigner AlgorithmSigner
+}
+
+// NewCertSigner returns a Signer that signs with the given Certificate, whose
+// private key is held by signer. It returns an error if the public key in cert
+// doesn't match the key used by signer.
+func NewCertSigner(cert *Certificate, signer Signer) (Signer, error) {
+	if bytes.Compare(cert.Key.Marshal(), signer.PublicKey().Marshal()) != 0 {
+		return nil, errors.New("ssh: signer and cert have different public key")
+	}
+
+	if algorithmSigner, ok := signer.(AlgorithmSigner); ok {
+		return &algorithmOpenSSHCertSigner{
+			&openSSHCertSigner{cert, signer}, algorithmSigner}, nil
+	} else {
+		return &openSSHCertSigner{cert, signer}, nil
+	}
+}
+
+func (s *openSSHCertSigner) Sign(rand io.Reader, data []byte) (*Signature, error) {
+	return s.signer.Sign(rand, data)
+}
+
+func (s *openSSHCertSigner) PublicKey() PublicKey {
+	return s.pub
+}
+
+func (s *algorithmOpenSSHCertSigner) SignWithAlgorithm(rand io.Reader, data []byte, algorithm string) (*Signature, error) {
+	return s.algorithmSigner.SignWithAlgorithm(rand, data, algorithm)
+}
+
+const sourceAddressCriticalOption = "source-address"
+
+// CertChecker does the work of verifying a certificate. Its methods
+// can be plugged into ClientConfig.HostKeyCallback and
+// ServerConfig.PublicKeyCallback. For the CertChecker to work,
+// minimally, the IsAuthority callback should be set.
+type CertChecker struct {
+	// SupportedCriticalOptions lists the CriticalOptions that the
+	// server application layer understands. These are only used
+	// for user certificates.
+	SupportedCriticalOptions []string
+
+	// IsUserAuthority should return true if the key is recognized as an
+	// authority for the given user certificate. This allows for
+	// certificates to be signed by other certificates. This must be set
+	// if this CertChecker will be checking user certificates.
+	IsUserAuthority func(auth PublicKey) bool
+
+	// IsHostAuthority should report whether the key is recognized as
+	// an authority for this host. This allows for certificates to be
+	// signed by other keys, and for those other keys to only be valid
+	// signers for particular hostnames. This must be set if this
+	// CertChecker will be checking host certificates.
+	IsHostAuthority func(auth PublicKey, address string) bool
+
+	// Clock is used for verifying time stamps. If nil, time.Now
+	// is used.
+	Clock func() time.Time
+
+	// UserKeyFallback is called when CertChecker.Authenticate encounters a
+	// public key that is not a certificate. It must implement validation
+	// of user keys or else, if nil, all such keys are rejected.
+	UserKeyFallback func(conn ConnMetadata, key PublicKey) (*Permissions, error)
+
+	// HostKeyFallback is called when CertChecker.CheckHostKey encounters a
+	// public key that is not a certificate. It must implement host key
+	// validation or else, if nil, all such keys are rejected.
+	HostKeyFallback HostKeyCallback
+
+	// IsRevoked is called for each certificate so that revocation checking
+	// can be implemented. It should return true if the given certificate
+	// is revoked and false otherwise. If nil, no certificates are
+	// considered to have been revoked.
+	IsRevoked func(cert *Certificate) bool
+}
+
+// CheckHostKey checks a host key certificate. This method can be
+// plugged into ClientConfig.HostKeyCallback.
+func (c *CertChecker) CheckHostKey(addr string, remote net.Addr, key PublicKey) error {
+	cert, ok := key.(*Certificate)
+	if !ok {
+		if c.HostKeyFallback != nil {
+			return c.HostKeyFallback(addr, remote, key)
+		}
+		return errors.New("ssh: non-certificate host key")
+	}
+	if cert.CertType != HostCert {
+		return fmt.Errorf("ssh: certificate presented as a host key has type %d", cert.CertType)
+	}
+	if !c.IsHostAuthority(cert.SignatureKey, addr) {
+		return fmt.Errorf("ssh: no authorities for hostname: %v", addr)
+	}
+
+	hostname, _, err := net.SplitHostPort(addr)
+	if err != nil {
+		return err
+	}
+
+	// Pass hostname only as principal for host certificates (consistent with OpenSSH)
+	return c.CheckCert(hostname, cert)
+}
+
+// Authenticate checks a user certificate. Authenticate can be used as
+// a value for ServerConfig.PublicKeyCallback.
+func (c *CertChecker) Authenticate(conn ConnMetadata, pubKey PublicKey) (*Permissions, error) {
+	cert, ok := pubKey.(*Certificate)
+	if !ok {
+		if c.UserKeyFallback != nil {
+			return c.UserKeyFallback(conn, pubKey)
+		}
+		return nil, errors.New("ssh: normal key pairs not accepted")
+	}
+
+	if cert.CertType != UserCert {
+		return nil, fmt.Errorf("ssh: cert has type %d", cert.CertType)
+	}
+	if !c.IsUserAuthority(cert.SignatureKey) {
+		return nil, fmt.Errorf("ssh: certificate signed by unrecognized authority")
+	}
+
+	if err := c.CheckCert(conn.User(), cert); err != nil {
+		return nil, err
+	}
+
+	return &cert.Permissions, nil
+}
+
+// CheckCert checks CriticalOptions, ValidPrincipals, revocation, timestamp and
+// the signature of the certificate.
+func (c *CertChecker) CheckCert(principal string, cert *Certificate) error {
+	if c.IsRevoked != nil && c.IsRevoked(cert) {
+		return fmt.Errorf("ssh: certificate serial %d revoked", cert.Serial)
+	}
+
+	for opt := range cert.CriticalOptions {
+		// sourceAddressCriticalOption will be enforced by
+		// serverAuthenticate
+		if opt == sourceAddressCriticalOption {
+			continue
+		}
+
+		found := false
+		for _, supp := range c.SupportedCriticalOptions {
+			if supp == opt {
+				found = true
+				break
+			}
+		}
+		if !found {
+			return fmt.Errorf("ssh: unsupported critical option %q in certificate", opt)
+		}
+	}
+
+	if len(cert.ValidPrincipals) > 0 {
+		// By default, certs are valid for all users/hosts.
+		found := false
+		for _, p := range cert.ValidPrincipals {
+			if p == principal {
+				found = true
+				break
+			}
+		}
+		if !found {
+			return fmt.Errorf("ssh: principal %q not in the set of valid principals for given certificate: %q", principal, cert.ValidPrincipals)
+		}
+	}
+
+	clock := c.Clock
+	if clock == nil {
+		clock = time.Now
+	}
+
+	unixNow := clock().Unix()
+	if after := int64(cert.ValidAfter); after < 0 || unixNow < int64(cert.ValidAfter) {
+		return fmt.Errorf("ssh: cert is not yet valid")
+	}
+	if before := int64(cert.ValidBefore); cert.ValidBefore != uint64(CertTimeInfinity) && (unixNow >= before || before < 0) {
+		return fmt.Errorf("ssh: cert has expired")
+	}
+	if err := cert.SignatureKey.Verify(cert.bytesForSigning(), cert.Signature); err != nil {
+		return fmt.Errorf("ssh: certificate signature does not verify")
+	}
+
+	return nil
+}
+
+// SignCert signs the certificate with an authority, setting the Nonce,
+// SignatureKey, and Signature fields.
+func (c *Certificate) SignCert(rand io.Reader, authority Signer) error {
+	c.Nonce = make([]byte, 32)
+	if _, err := io.ReadFull(rand, c.Nonce); err != nil {
+		return err
+	}
+	c.SignatureKey = authority.PublicKey()
+
+	sig, err := authority.Sign(rand, c.bytesForSigning())
+	if err != nil {
+		return err
+	}
+	c.Signature = sig
+	return nil
+}
+
+var certAlgoNames = map[string]string{
+	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.
+// Panics if a non-certificate algorithm is passed.
+func certToPrivAlgo(algo string) string {
+	for privAlgo, pubAlgo := range certAlgoNames {
+		if pubAlgo == algo {
+			return privAlgo
+		}
+	}
+	panic("unknown cert algorithm")
+}
+
+func (cert *Certificate) bytesForSigning() []byte {
+	c2 := *cert
+	c2.Signature = nil
+	out := c2.Marshal()
+	// Drop trailing signature length.
+	return out[:len(out)-4]
+}
+
+// Marshal serializes c into OpenSSH's wire format. It is part of the
+// PublicKey interface.
+func (c *Certificate) Marshal() []byte {
+	generic := genericCertData{
+		Serial:          c.Serial,
+		CertType:        c.CertType,
+		KeyId:           c.KeyId,
+		ValidPrincipals: marshalStringList(c.ValidPrincipals),
+		ValidAfter:      uint64(c.ValidAfter),
+		ValidBefore:     uint64(c.ValidBefore),
+		CriticalOptions: marshalTuples(c.CriticalOptions),
+		Extensions:      marshalTuples(c.Extensions),
+		Reserved:        c.Reserved,
+		SignatureKey:    c.SignatureKey.Marshal(),
+	}
+	if c.Signature != nil {
+		generic.Signature = Marshal(c.Signature)
+	}
+	genericBytes := Marshal(&generic)
+	keyBytes := c.Key.Marshal()
+	_, keyBytes, _ = parseString(keyBytes)
+	prefix := Marshal(&struct {
+		Name  string
+		Nonce []byte
+		Key   []byte `ssh:"rest"`
+	}{c.Type(), c.Nonce, keyBytes})
+
+	result := make([]byte, 0, len(prefix)+len(genericBytes))
+	result = append(result, prefix...)
+	result = append(result, genericBytes...)
+	return result
+}
+
+// Type returns the key name. It is part of the PublicKey interface.
+func (c *Certificate) Type() string {
+	algo, ok := certAlgoNames[c.Key.Type()]
+	if !ok {
+		panic("unknown cert key type " + c.Key.Type())
+	}
+	return algo
+}
+
+// Verify verifies a signature against the certificate's public
+// key. It is part of the PublicKey interface.
+func (c *Certificate) Verify(data []byte, sig *Signature) error {
+	return c.Key.Verify(data, sig)
+}
+
+func parseSignatureBody(in []byte) (out *Signature, rest []byte, ok bool) {
+	format, in, ok := parseString(in)
+	if !ok {
+		return
+	}
+
+	out = &Signature{
+		Format: string(format),
+	}
+
+	if out.Blob, in, ok = parseString(in); !ok {
+		return
+	}
+
+	switch out.Format {
+	case KeyAlgoSKECDSA256, CertAlgoSKECDSA256v01, KeyAlgoSKED25519, CertAlgoSKED25519v01:
+		out.Rest = in
+		return out, nil, ok
+	}
+
+	return out, in, ok
+}
+
+func parseSignature(in []byte) (out *Signature, rest []byte, ok bool) {
+	sigBytes, rest, ok := parseString(in)
+	if !ok {
+		return
+	}
+
+	out, trailing, ok := parseSignatureBody(sigBytes)
+	if !ok || len(trailing) > 0 {
+		return nil, nil, false
+	}
+	return
+}

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

@@ -0,0 +1,633 @@
+// Copyright 2011 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 ssh
+
+import (
+	"encoding/binary"
+	"errors"
+	"fmt"
+	"io"
+	"log"
+	"sync"
+)
+
+const (
+	minPacketLength = 9
+	// channelMaxPacket contains the maximum number of bytes that will be
+	// sent in a single packet. As per RFC 4253, section 6.1, 32k is also
+	// the minimum.
+	channelMaxPacket = 1 << 15
+	// We follow OpenSSH here.
+	channelWindowSize = 64 * channelMaxPacket
+)
+
+// NewChannel represents an incoming request to a channel. It must either be
+// accepted for use by calling Accept, or rejected by calling Reject.
+type NewChannel interface {
+	// Accept accepts the channel creation request. It returns the Channel
+	// and a Go channel containing SSH requests. The Go channel must be
+	// serviced otherwise the Channel will hang.
+	Accept() (Channel, <-chan *Request, error)
+
+	// Reject rejects the channel creation request. After calling
+	// this, no other methods on the Channel may be called.
+	Reject(reason RejectionReason, message string) error
+
+	// ChannelType returns the type of the channel, as supplied by the
+	// client.
+	ChannelType() string
+
+	// ExtraData returns the arbitrary payload for this channel, as supplied
+	// by the client. This data is specific to the channel type.
+	ExtraData() []byte
+}
+
+// A Channel is an ordered, reliable, flow-controlled, duplex stream
+// that is multiplexed over an SSH connection.
+type Channel interface {
+	// Read reads up to len(data) bytes from the channel.
+	Read(data []byte) (int, error)
+
+	// Write writes len(data) bytes to the channel.
+	Write(data []byte) (int, error)
+
+	// Close signals end of channel use. No data may be sent after this
+	// call.
+	Close() error
+
+	// CloseWrite signals the end of sending in-band
+	// data. Requests may still be sent, and the other side may
+	// still send data
+	CloseWrite() error
+
+	// SendRequest sends a channel request.  If wantReply is true,
+	// it will wait for a reply and return the result as a
+	// boolean, otherwise the return value will be false. Channel
+	// requests are out-of-band messages so they may be sent even
+	// if the data stream is closed or blocked by flow control.
+	// If the channel is closed before a reply is returned, io.EOF
+	// is returned.
+	SendRequest(name string, wantReply bool, payload []byte) (bool, error)
+
+	// Stderr returns an io.ReadWriter that writes to this channel
+	// with the extended data type set to stderr. Stderr may
+	// safely be read and written from a different goroutine than
+	// Read and Write respectively.
+	Stderr() io.ReadWriter
+}
+
+// Request is a request sent outside of the normal stream of
+// data. Requests can either be specific to an SSH channel, or they
+// can be global.
+type Request struct {
+	Type      string
+	WantReply bool
+	Payload   []byte
+
+	ch  *channel
+	mux *mux
+}
+
+// Reply sends a response to a request. It must be called for all requests
+// where WantReply is true and is a no-op otherwise. The payload argument is
+// ignored for replies to channel-specific requests.
+func (r *Request) Reply(ok bool, payload []byte) error {
+	if !r.WantReply {
+		return nil
+	}
+
+	if r.ch == nil {
+		return r.mux.ackRequest(ok, payload)
+	}
+
+	return r.ch.ackRequest(ok)
+}
+
+// RejectionReason is an enumeration used when rejecting channel creation
+// requests. See RFC 4254, section 5.1.
+type RejectionReason uint32
+
+const (
+	Prohibited RejectionReason = iota + 1
+	ConnectionFailed
+	UnknownChannelType
+	ResourceShortage
+)
+
+// String converts the rejection reason to human readable form.
+func (r RejectionReason) String() string {
+	switch r {
+	case Prohibited:
+		return "administratively prohibited"
+	case ConnectionFailed:
+		return "connect failed"
+	case UnknownChannelType:
+		return "unknown channel type"
+	case ResourceShortage:
+		return "resource shortage"
+	}
+	return fmt.Sprintf("unknown reason %d", int(r))
+}
+
+func min(a uint32, b int) uint32 {
+	if a < uint32(b) {
+		return a
+	}
+	return uint32(b)
+}
+
+type channelDirection uint8
+
+const (
+	channelInbound channelDirection = iota
+	channelOutbound
+)
+
+// channel is an implementation of the Channel interface that works
+// with the mux class.
+type channel struct {
+	// R/O after creation
+	chanType          string
+	extraData         []byte
+	localId, remoteId uint32
+
+	// maxIncomingPayload and maxRemotePayload are the maximum
+	// payload sizes of normal and extended data packets for
+	// receiving and sending, respectively. The wire packet will
+	// be 9 or 13 bytes larger (excluding encryption overhead).
+	maxIncomingPayload uint32
+	maxRemotePayload   uint32
+
+	mux *mux
+
+	// decided is set to true if an accept or reject message has been sent
+	// (for outbound channels) or received (for inbound channels).
+	decided bool
+
+	// direction contains either channelOutbound, for channels created
+	// locally, or channelInbound, for channels created by the peer.
+	direction channelDirection
+
+	// Pending internal channel messages.
+	msg chan interface{}
+
+	// Since requests have no ID, there can be only one request
+	// with WantReply=true outstanding.  This lock is held by a
+	// goroutine that has such an outgoing request pending.
+	sentRequestMu sync.Mutex
+
+	incomingRequests chan *Request
+
+	sentEOF bool
+
+	// thread-safe data
+	remoteWin  window
+	pending    *buffer
+	extPending *buffer
+
+	// windowMu protects myWindow, the flow-control window.
+	windowMu sync.Mutex
+	myWindow uint32
+
+	// writeMu serializes calls to mux.conn.writePacket() and
+	// protects sentClose and packetPool. This mutex must be
+	// different from windowMu, as writePacket can block if there
+	// is a key exchange pending.
+	writeMu   sync.Mutex
+	sentClose bool
+
+	// packetPool has a buffer for each extended channel ID to
+	// save allocations during writes.
+	packetPool map[uint32][]byte
+}
+
+// writePacket sends a packet. If the packet is a channel close, it updates
+// sentClose. This method takes the lock c.writeMu.
+func (ch *channel) writePacket(packet []byte) error {
+	ch.writeMu.Lock()
+	if ch.sentClose {
+		ch.writeMu.Unlock()
+		return io.EOF
+	}
+	ch.sentClose = (packet[0] == msgChannelClose)
+	err := ch.mux.conn.writePacket(packet)
+	ch.writeMu.Unlock()
+	return err
+}
+
+func (ch *channel) sendMessage(msg interface{}) error {
+	if debugMux {
+		log.Printf("send(%d): %#v", ch.mux.chanList.offset, msg)
+	}
+
+	p := Marshal(msg)
+	binary.BigEndian.PutUint32(p[1:], ch.remoteId)
+	return ch.writePacket(p)
+}
+
+// WriteExtended writes data to a specific extended stream. These streams are
+// used, for example, for stderr.
+func (ch *channel) WriteExtended(data []byte, extendedCode uint32) (n int, err error) {
+	if ch.sentEOF {
+		return 0, io.EOF
+	}
+	// 1 byte message type, 4 bytes remoteId, 4 bytes data length
+	opCode := byte(msgChannelData)
+	headerLength := uint32(9)
+	if extendedCode > 0 {
+		headerLength += 4
+		opCode = msgChannelExtendedData
+	}
+
+	ch.writeMu.Lock()
+	packet := ch.packetPool[extendedCode]
+	// We don't remove the buffer from packetPool, so
+	// WriteExtended calls from different goroutines will be
+	// flagged as errors by the race detector.
+	ch.writeMu.Unlock()
+
+	for len(data) > 0 {
+		space := min(ch.maxRemotePayload, len(data))
+		if space, err = ch.remoteWin.reserve(space); err != nil {
+			return n, err
+		}
+		if want := headerLength + space; uint32(cap(packet)) < want {
+			packet = make([]byte, want)
+		} else {
+			packet = packet[:want]
+		}
+
+		todo := data[:space]
+
+		packet[0] = opCode
+		binary.BigEndian.PutUint32(packet[1:], ch.remoteId)
+		if extendedCode > 0 {
+			binary.BigEndian.PutUint32(packet[5:], uint32(extendedCode))
+		}
+		binary.BigEndian.PutUint32(packet[headerLength-4:], uint32(len(todo)))
+		copy(packet[headerLength:], todo)
+		if err = ch.writePacket(packet); err != nil {
+			return n, err
+		}
+
+		n += len(todo)
+		data = data[len(todo):]
+	}
+
+	ch.writeMu.Lock()
+	ch.packetPool[extendedCode] = packet
+	ch.writeMu.Unlock()
+
+	return n, err
+}
+
+func (ch *channel) handleData(packet []byte) error {
+	headerLen := 9
+	isExtendedData := packet[0] == msgChannelExtendedData
+	if isExtendedData {
+		headerLen = 13
+	}
+	if len(packet) < headerLen {
+		// malformed data packet
+		return parseError(packet[0])
+	}
+
+	var extended uint32
+	if isExtendedData {
+		extended = binary.BigEndian.Uint32(packet[5:])
+	}
+
+	length := binary.BigEndian.Uint32(packet[headerLen-4 : headerLen])
+	if length == 0 {
+		return nil
+	}
+	if length > ch.maxIncomingPayload {
+		// TODO(hanwen): should send Disconnect?
+		return errors.New("ssh: incoming packet exceeds maximum payload size")
+	}
+
+	data := packet[headerLen:]
+	if length != uint32(len(data)) {
+		return errors.New("ssh: wrong packet length")
+	}
+
+	ch.windowMu.Lock()
+	if ch.myWindow < length {
+		ch.windowMu.Unlock()
+		// TODO(hanwen): should send Disconnect with reason?
+		return errors.New("ssh: remote side wrote too much")
+	}
+	ch.myWindow -= length
+	ch.windowMu.Unlock()
+
+	if extended == 1 {
+		ch.extPending.write(data)
+	} else if extended > 0 {
+		// discard other extended data.
+	} else {
+		ch.pending.write(data)
+	}
+	return nil
+}
+
+func (c *channel) adjustWindow(n uint32) error {
+	c.windowMu.Lock()
+	// Since myWindow is managed on our side, and can never exceed
+	// the initial window setting, we don't worry about overflow.
+	c.myWindow += uint32(n)
+	c.windowMu.Unlock()
+	return c.sendMessage(windowAdjustMsg{
+		AdditionalBytes: uint32(n),
+	})
+}
+
+func (c *channel) ReadExtended(data []byte, extended uint32) (n int, err error) {
+	switch extended {
+	case 1:
+		n, err = c.extPending.Read(data)
+	case 0:
+		n, err = c.pending.Read(data)
+	default:
+		return 0, fmt.Errorf("ssh: extended code %d unimplemented", extended)
+	}
+
+	if n > 0 {
+		err = c.adjustWindow(uint32(n))
+		// sendWindowAdjust can return io.EOF if the remote
+		// peer has closed the connection, however we want to
+		// defer forwarding io.EOF to the caller of Read until
+		// the buffer has been drained.
+		if n > 0 && err == io.EOF {
+			err = nil
+		}
+	}
+
+	return n, err
+}
+
+func (c *channel) close() {
+	c.pending.eof()
+	c.extPending.eof()
+	close(c.msg)
+	close(c.incomingRequests)
+	c.writeMu.Lock()
+	// This is not necessary for a normal channel teardown, but if
+	// there was another error, it is.
+	c.sentClose = true
+	c.writeMu.Unlock()
+	// Unblock writers.
+	c.remoteWin.close()
+}
+
+// responseMessageReceived is called when a success or failure message is
+// received on a channel to check that such a message is reasonable for the
+// given channel.
+func (ch *channel) responseMessageReceived() error {
+	if ch.direction == channelInbound {
+		return errors.New("ssh: channel response message received on inbound channel")
+	}
+	if ch.decided {
+		return errors.New("ssh: duplicate response received for channel")
+	}
+	ch.decided = true
+	return nil
+}
+
+func (ch *channel) handlePacket(packet []byte) error {
+	switch packet[0] {
+	case msgChannelData, msgChannelExtendedData:
+		return ch.handleData(packet)
+	case msgChannelClose:
+		ch.sendMessage(channelCloseMsg{PeersID: ch.remoteId})
+		ch.mux.chanList.remove(ch.localId)
+		ch.close()
+		return nil
+	case msgChannelEOF:
+		// RFC 4254 is mute on how EOF affects dataExt messages but
+		// it is logical to signal EOF at the same time.
+		ch.extPending.eof()
+		ch.pending.eof()
+		return nil
+	}
+
+	decoded, err := decode(packet)
+	if err != nil {
+		return err
+	}
+
+	switch msg := decoded.(type) {
+	case *channelOpenFailureMsg:
+		if err := ch.responseMessageReceived(); err != nil {
+			return err
+		}
+		ch.mux.chanList.remove(msg.PeersID)
+		ch.msg <- msg
+	case *channelOpenConfirmMsg:
+		if err := ch.responseMessageReceived(); err != nil {
+			return err
+		}
+		if msg.MaxPacketSize < minPacketLength || msg.MaxPacketSize > 1<<31 {
+			return fmt.Errorf("ssh: invalid MaxPacketSize %d from peer", msg.MaxPacketSize)
+		}
+		ch.remoteId = msg.MyID
+		ch.maxRemotePayload = msg.MaxPacketSize
+		ch.remoteWin.add(msg.MyWindow)
+		ch.msg <- msg
+	case *windowAdjustMsg:
+		if !ch.remoteWin.add(msg.AdditionalBytes) {
+			return fmt.Errorf("ssh: invalid window update for %d bytes", msg.AdditionalBytes)
+		}
+	case *channelRequestMsg:
+		req := Request{
+			Type:      msg.Request,
+			WantReply: msg.WantReply,
+			Payload:   msg.RequestSpecificData,
+			ch:        ch,
+		}
+
+		ch.incomingRequests <- &req
+	default:
+		ch.msg <- msg
+	}
+	return nil
+}
+
+func (m *mux) newChannel(chanType string, direction channelDirection, extraData []byte) *channel {
+	ch := &channel{
+		remoteWin:        window{Cond: newCond()},
+		myWindow:         channelWindowSize,
+		pending:          newBuffer(),
+		extPending:       newBuffer(),
+		direction:        direction,
+		incomingRequests: make(chan *Request, chanSize),
+		msg:              make(chan interface{}, chanSize),
+		chanType:         chanType,
+		extraData:        extraData,
+		mux:              m,
+		packetPool:       make(map[uint32][]byte),
+	}
+	ch.localId = m.chanList.add(ch)
+	return ch
+}
+
+var errUndecided = errors.New("ssh: must Accept or Reject channel")
+var errDecidedAlready = errors.New("ssh: can call Accept or Reject only once")
+
+type extChannel struct {
+	code uint32
+	ch   *channel
+}
+
+func (e *extChannel) Write(data []byte) (n int, err error) {
+	return e.ch.WriteExtended(data, e.code)
+}
+
+func (e *extChannel) Read(data []byte) (n int, err error) {
+	return e.ch.ReadExtended(data, e.code)
+}
+
+func (ch *channel) Accept() (Channel, <-chan *Request, error) {
+	if ch.decided {
+		return nil, nil, errDecidedAlready
+	}
+	ch.maxIncomingPayload = channelMaxPacket
+	confirm := channelOpenConfirmMsg{
+		PeersID:       ch.remoteId,
+		MyID:          ch.localId,
+		MyWindow:      ch.myWindow,
+		MaxPacketSize: ch.maxIncomingPayload,
+	}
+	ch.decided = true
+	if err := ch.sendMessage(confirm); err != nil {
+		return nil, nil, err
+	}
+
+	return ch, ch.incomingRequests, nil
+}
+
+func (ch *channel) Reject(reason RejectionReason, message string) error {
+	if ch.decided {
+		return errDecidedAlready
+	}
+	reject := channelOpenFailureMsg{
+		PeersID:  ch.remoteId,
+		Reason:   reason,
+		Message:  message,
+		Language: "en",
+	}
+	ch.decided = true
+	return ch.sendMessage(reject)
+}
+
+func (ch *channel) Read(data []byte) (int, error) {
+	if !ch.decided {
+		return 0, errUndecided
+	}
+	return ch.ReadExtended(data, 0)
+}
+
+func (ch *channel) Write(data []byte) (int, error) {
+	if !ch.decided {
+		return 0, errUndecided
+	}
+	return ch.WriteExtended(data, 0)
+}
+
+func (ch *channel) CloseWrite() error {
+	if !ch.decided {
+		return errUndecided
+	}
+	ch.sentEOF = true
+	return ch.sendMessage(channelEOFMsg{
+		PeersID: ch.remoteId})
+}
+
+func (ch *channel) Close() error {
+	if !ch.decided {
+		return errUndecided
+	}
+
+	return ch.sendMessage(channelCloseMsg{
+		PeersID: ch.remoteId})
+}
+
+// Extended returns an io.ReadWriter that sends and receives data on the given,
+// SSH extended stream. Such streams are used, for example, for stderr.
+func (ch *channel) Extended(code uint32) io.ReadWriter {
+	if !ch.decided {
+		return nil
+	}
+	return &extChannel{code, ch}
+}
+
+func (ch *channel) Stderr() io.ReadWriter {
+	return ch.Extended(1)
+}
+
+func (ch *channel) SendRequest(name string, wantReply bool, payload []byte) (bool, error) {
+	if !ch.decided {
+		return false, errUndecided
+	}
+
+	if wantReply {
+		ch.sentRequestMu.Lock()
+		defer ch.sentRequestMu.Unlock()
+	}
+
+	msg := channelRequestMsg{
+		PeersID:             ch.remoteId,
+		Request:             name,
+		WantReply:           wantReply,
+		RequestSpecificData: payload,
+	}
+
+	if err := ch.sendMessage(msg); err != nil {
+		return false, err
+	}
+
+	if wantReply {
+		m, ok := (<-ch.msg)
+		if !ok {
+			return false, io.EOF
+		}
+		switch m.(type) {
+		case *channelRequestFailureMsg:
+			return false, nil
+		case *channelRequestSuccessMsg:
+			return true, nil
+		default:
+			return false, fmt.Errorf("ssh: unexpected response to channel request: %#v", m)
+		}
+	}
+
+	return false, nil
+}
+
+// ackRequest either sends an ack or nack to the channel request.
+func (ch *channel) ackRequest(ok bool) error {
+	if !ch.decided {
+		return errUndecided
+	}
+
+	var msg interface{}
+	if !ok {
+		msg = channelRequestFailureMsg{
+			PeersID: ch.remoteId,
+		}
+	} else {
+		msg = channelRequestSuccessMsg{
+			PeersID: ch.remoteId,
+		}
+	}
+	return ch.sendMessage(msg)
+}
+
+func (ch *channel) ChannelType() string {
+	return ch.chanType
+}
+
+func (ch *channel) ExtraData() []byte {
+	return ch.extraData
+}

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

@@ -0,0 +1,781 @@
+// Copyright 2011 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 ssh
+
+import (
+	"crypto/aes"
+	"crypto/cipher"
+	"crypto/des"
+	"crypto/rc4"
+	"crypto/subtle"
+	"encoding/binary"
+	"errors"
+	"fmt"
+	"hash"
+	"io"
+	"io/ioutil"
+
+	"golang.org/x/crypto/chacha20"
+	"golang.org/x/crypto/poly1305"
+)
+
+const (
+	packetSizeMultiple = 16 // TODO(huin) this should be determined by the cipher.
+
+	// RFC 4253 section 6.1 defines a minimum packet size of 32768 that implementations
+	// MUST be able to process (plus a few more kilobytes for padding and mac). The RFC
+	// indicates implementations SHOULD be able to handle larger packet sizes, but then
+	// waffles on about reasonable limits.
+	//
+	// OpenSSH caps their maxPacket at 256kB so we choose to do
+	// the same. maxPacket is also used to ensure that uint32
+	// length fields do not overflow, so it should remain well
+	// below 4G.
+	maxPacket = 256 * 1024
+)
+
+// noneCipher implements cipher.Stream and provides no encryption. It is used
+// by the transport before the first key-exchange.
+type noneCipher struct{}
+
+func (c noneCipher) XORKeyStream(dst, src []byte) {
+	copy(dst, src)
+}
+
+func newAESCTR(key, iv []byte) (cipher.Stream, error) {
+	c, err := aes.NewCipher(key)
+	if err != nil {
+		return nil, err
+	}
+	return cipher.NewCTR(c, iv), nil
+}
+
+func newRC4(key, iv []byte) (cipher.Stream, error) {
+	return rc4.NewCipher(key)
+}
+
+type cipherMode struct {
+	keySize int
+	ivSize  int
+	create  func(key, iv []byte, macKey []byte, algs directionAlgorithms) (packetCipher, error)
+}
+
+func streamCipherMode(skip int, createFunc func(key, iv []byte) (cipher.Stream, error)) func(key, iv []byte, macKey []byte, algs directionAlgorithms) (packetCipher, error) {
+	return func(key, iv, macKey []byte, algs directionAlgorithms) (packetCipher, error) {
+		stream, err := createFunc(key, iv)
+		if err != nil {
+			return nil, err
+		}
+
+		var streamDump []byte
+		if skip > 0 {
+			streamDump = make([]byte, 512)
+		}
+
+		for remainingToDump := skip; remainingToDump > 0; {
+			dumpThisTime := remainingToDump
+			if dumpThisTime > len(streamDump) {
+				dumpThisTime = len(streamDump)
+			}
+			stream.XORKeyStream(streamDump[:dumpThisTime], streamDump[:dumpThisTime])
+			remainingToDump -= dumpThisTime
+		}
+
+		mac := macModes[algs.MAC].new(macKey)
+		return &streamPacketCipher{
+			mac:       mac,
+			etm:       macModes[algs.MAC].etm,
+			macResult: make([]byte, mac.Size()),
+			cipher:    stream,
+		}, nil
+	}
+}
+
+// cipherModes documents properties of supported ciphers. Ciphers not included
+// are not supported and will not be negotiated, even if explicitly requested in
+// ClientConfig.Crypto.Ciphers.
+var cipherModes = map[string]*cipherMode{
+	// Ciphers from RFC4344, which introduced many CTR-based ciphers. Algorithms
+	// are defined in the order specified in the RFC.
+	"aes128-ctr": {16, aes.BlockSize, streamCipherMode(0, newAESCTR)},
+	"aes192-ctr": {24, aes.BlockSize, streamCipherMode(0, newAESCTR)},
+	"aes256-ctr": {32, aes.BlockSize, streamCipherMode(0, newAESCTR)},
+
+	// Ciphers from RFC4345, which introduces security-improved arcfour ciphers.
+	// They are defined in the order specified in the RFC.
+	"arcfour128": {16, 0, streamCipherMode(1536, newRC4)},
+	"arcfour256": {32, 0, streamCipherMode(1536, newRC4)},
+
+	// Cipher defined in RFC 4253, which describes SSH Transport Layer Protocol.
+	// Note that this cipher is not safe, as stated in RFC 4253: "Arcfour (and
+	// RC4) has problems with weak keys, and should be used with caution."
+	// RFC4345 introduces improved versions of Arcfour.
+	"arcfour": {16, 0, streamCipherMode(0, newRC4)},
+
+	// AEAD ciphers
+	gcmCipherID:        {16, 12, newGCMCipher},
+	chacha20Poly1305ID: {64, 0, newChaCha20Cipher},
+
+	// CBC mode is insecure and so is not included in the default config.
+	// (See https://www.ieee-security.org/TC/SP2013/papers/4977a526.pdf). If absolutely
+	// needed, it's possible to specify a custom Config to enable it.
+	// You should expect that an active attacker can recover plaintext if
+	// you do.
+	aes128cbcID: {16, aes.BlockSize, newAESCBCCipher},
+
+	// 3des-cbc is insecure and is not included in the default
+	// config.
+	tripledescbcID: {24, des.BlockSize, newTripleDESCBCCipher},
+}
+
+// prefixLen is the length of the packet prefix that contains the packet length
+// and number of padding bytes.
+const prefixLen = 5
+
+// streamPacketCipher is a packetCipher using a stream cipher.
+type streamPacketCipher struct {
+	mac    hash.Hash
+	cipher cipher.Stream
+	etm    bool
+
+	// The following members are to avoid per-packet allocations.
+	prefix      [prefixLen]byte
+	seqNumBytes [4]byte
+	padding     [2 * packetSizeMultiple]byte
+	packetData  []byte
+	macResult   []byte
+}
+
+// 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
+	}
+
+	var encryptedPaddingLength [1]byte
+	if s.mac != nil && s.etm {
+		copy(encryptedPaddingLength[:], s.prefix[4:5])
+		s.cipher.XORKeyStream(s.prefix[4:5], s.prefix[4:5])
+	} else {
+		s.cipher.XORKeyStream(s.prefix[:], s.prefix[:])
+	}
+
+	length := binary.BigEndian.Uint32(s.prefix[0:4])
+	paddingLength := uint32(s.prefix[4])
+
+	var macSize uint32
+	if s.mac != nil {
+		s.mac.Reset()
+		binary.BigEndian.PutUint32(s.seqNumBytes[:], seqNum)
+		s.mac.Write(s.seqNumBytes[:])
+		if s.etm {
+			s.mac.Write(s.prefix[:4])
+			s.mac.Write(encryptedPaddingLength[:])
+		} else {
+			s.mac.Write(s.prefix[:])
+		}
+		macSize = uint32(s.mac.Size())
+	}
+
+	if length <= paddingLength+1 {
+		return nil, errors.New("ssh: invalid packet length, packet too small")
+	}
+
+	if length > maxPacket {
+		return nil, errors.New("ssh: invalid packet length, packet too large")
+	}
+
+	// the maxPacket check above ensures that length-1+macSize
+	// does not overflow.
+	if uint32(cap(s.packetData)) < length-1+macSize {
+		s.packetData = make([]byte, length-1+macSize)
+	} else {
+		s.packetData = s.packetData[:length-1+macSize]
+	}
+
+	if _, err := io.ReadFull(r, s.packetData); err != nil {
+		return nil, err
+	}
+	mac := s.packetData[length-1:]
+	data := s.packetData[:length-1]
+
+	if s.mac != nil && s.etm {
+		s.mac.Write(data)
+	}
+
+	s.cipher.XORKeyStream(data, data)
+
+	if s.mac != nil {
+		if !s.etm {
+			s.mac.Write(data)
+		}
+		s.macResult = s.mac.Sum(s.macResult[:0])
+		if subtle.ConstantTimeCompare(s.macResult, mac) != 1 {
+			return nil, errors.New("ssh: MAC failure")
+		}
+	}
+
+	return s.packetData[:length-paddingLength-1], nil
+}
+
+// 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")
+	}
+
+	aadlen := 0
+	if s.mac != nil && s.etm {
+		// packet length is not encrypted for EtM modes
+		aadlen = 4
+	}
+
+	paddingLength := packetSizeMultiple - (prefixLen+len(packet)-aadlen)%packetSizeMultiple
+	if paddingLength < 4 {
+		paddingLength += packetSizeMultiple
+	}
+
+	length := len(packet) + 1 + paddingLength
+	binary.BigEndian.PutUint32(s.prefix[:], uint32(length))
+	s.prefix[4] = byte(paddingLength)
+	padding := s.padding[:paddingLength]
+	if _, err := io.ReadFull(rand, padding); err != nil {
+		return err
+	}
+
+	if s.mac != nil {
+		s.mac.Reset()
+		binary.BigEndian.PutUint32(s.seqNumBytes[:], seqNum)
+		s.mac.Write(s.seqNumBytes[:])
+
+		if s.etm {
+			// For EtM algorithms, the packet length must stay unencrypted,
+			// but the following data (padding length) must be encrypted
+			s.cipher.XORKeyStream(s.prefix[4:5], s.prefix[4:5])
+		}
+
+		s.mac.Write(s.prefix[:])
+
+		if !s.etm {
+			// For non-EtM algorithms, the algorithm is applied on unencrypted data
+			s.mac.Write(packet)
+			s.mac.Write(padding)
+		}
+	}
+
+	if !(s.mac != nil && s.etm) {
+		// For EtM algorithms, the padding length has already been encrypted
+		// and the packet length must remain unencrypted
+		s.cipher.XORKeyStream(s.prefix[:], s.prefix[:])
+	}
+
+	s.cipher.XORKeyStream(packet, packet)
+	s.cipher.XORKeyStream(padding, padding)
+
+	if s.mac != nil && s.etm {
+		// For EtM algorithms, packet and padding must be encrypted
+		s.mac.Write(packet)
+		s.mac.Write(padding)
+	}
+
+	if _, err := w.Write(s.prefix[:]); err != nil {
+		return err
+	}
+	if _, err := w.Write(packet); err != nil {
+		return err
+	}
+	if _, err := w.Write(padding); err != nil {
+		return err
+	}
+
+	if s.mac != nil {
+		s.macResult = s.mac.Sum(s.macResult[:0])
+		if _, err := w.Write(s.macResult); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+type gcmCipher struct {
+	aead   cipher.AEAD
+	prefix [4]byte
+	iv     []byte
+	buf    []byte
+}
+
+func newGCMCipher(key, iv, unusedMacKey []byte, unusedAlgs directionAlgorithms) (packetCipher, error) {
+	c, err := aes.NewCipher(key)
+	if err != nil {
+		return nil, err
+	}
+
+	aead, err := cipher.NewGCM(c)
+	if err != nil {
+		return nil, err
+	}
+
+	return &gcmCipher{
+		aead: aead,
+		iv:   iv,
+	}, nil
+}
+
+const gcmTagSize = 16
+
+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)
+	if padding < 4 {
+		padding += packetSizeMultiple
+	}
+
+	length := uint32(len(packet) + int(padding) + 1)
+	binary.BigEndian.PutUint32(c.prefix[:], length)
+	if _, err := w.Write(c.prefix[:]); err != nil {
+		return err
+	}
+
+	if cap(c.buf) < int(length) {
+		c.buf = make([]byte, length)
+	} else {
+		c.buf = c.buf[:length]
+	}
+
+	c.buf[0] = padding
+	copy(c.buf[1:], packet)
+	if _, err := io.ReadFull(rand, c.buf[1+len(packet):]); err != nil {
+		return err
+	}
+	c.buf = c.aead.Seal(c.buf[:0], c.iv, c.buf, c.prefix[:])
+	if _, err := w.Write(c.buf); err != nil {
+		return err
+	}
+	c.incIV()
+
+	return nil
+}
+
+func (c *gcmCipher) incIV() {
+	for i := 4 + 7; i >= 4; i-- {
+		c.iv[i]++
+		if c.iv[i] != 0 {
+			break
+		}
+	}
+}
+
+func (c *gcmCipher) readCipherPacket(seqNum uint32, r io.Reader) ([]byte, error) {
+	if _, err := io.ReadFull(r, c.prefix[:]); err != nil {
+		return nil, err
+	}
+	length := binary.BigEndian.Uint32(c.prefix[:])
+	if length > maxPacket {
+		return nil, errors.New("ssh: max packet length exceeded")
+	}
+
+	if cap(c.buf) < int(length+gcmTagSize) {
+		c.buf = make([]byte, length+gcmTagSize)
+	} else {
+		c.buf = c.buf[:length+gcmTagSize]
+	}
+
+	if _, err := io.ReadFull(r, c.buf); err != nil {
+		return nil, err
+	}
+
+	plain, err := c.aead.Open(c.buf[:0], c.iv, c.buf, c.prefix[:])
+	if err != nil {
+		return nil, err
+	}
+	c.incIV()
+
+	padding := plain[0]
+	if padding < 4 {
+		// padding is a byte, so it automatically satisfies
+		// the maximum size, which is 255.
+		return nil, fmt.Errorf("ssh: illegal padding %d", padding)
+	}
+
+	if int(padding+1) >= len(plain) {
+		return nil, fmt.Errorf("ssh: padding %d too large", padding)
+	}
+	plain = plain[1 : length-uint32(padding)]
+	return plain, nil
+}
+
+// cbcCipher implements aes128-cbc cipher defined in RFC 4253 section 6.1
+type cbcCipher struct {
+	mac       hash.Hash
+	macSize   uint32
+	decrypter cipher.BlockMode
+	encrypter cipher.BlockMode
+
+	// The following members are to avoid per-packet allocations.
+	seqNumBytes [4]byte
+	packetData  []byte
+	macResult   []byte
+
+	// Amount of data we should still read to hide which
+	// verification error triggered.
+	oracleCamouflage uint32
+}
+
+func newCBCCipher(c cipher.Block, key, iv, macKey []byte, algs directionAlgorithms) (packetCipher, error) {
+	cbc := &cbcCipher{
+		mac:        macModes[algs.MAC].new(macKey),
+		decrypter:  cipher.NewCBCDecrypter(c, iv),
+		encrypter:  cipher.NewCBCEncrypter(c, iv),
+		packetData: make([]byte, 1024),
+	}
+	if cbc.mac != nil {
+		cbc.macSize = uint32(cbc.mac.Size())
+	}
+
+	return cbc, nil
+}
+
+func newAESCBCCipher(key, iv, macKey []byte, algs directionAlgorithms) (packetCipher, error) {
+	c, err := aes.NewCipher(key)
+	if err != nil {
+		return nil, err
+	}
+
+	cbc, err := newCBCCipher(c, key, iv, macKey, algs)
+	if err != nil {
+		return nil, err
+	}
+
+	return cbc, nil
+}
+
+func newTripleDESCBCCipher(key, iv, macKey []byte, algs directionAlgorithms) (packetCipher, error) {
+	c, err := des.NewTripleDESCipher(key)
+	if err != nil {
+		return nil, err
+	}
+
+	cbc, err := newCBCCipher(c, key, iv, macKey, algs)
+	if err != nil {
+		return nil, err
+	}
+
+	return cbc, nil
+}
+
+func maxUInt32(a, b int) uint32 {
+	if a > b {
+		return uint32(a)
+	}
+	return uint32(b)
+}
+
+const (
+	cbcMinPacketSizeMultiple = 8
+	cbcMinPacketSize         = 16
+	cbcMinPaddingSize        = 4
+)
+
+// cbcError represents a verification error that may leak information.
+type cbcError string
+
+func (e cbcError) Error() string { return string(e) }
+
+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
+			// data, to make distinguishing between
+			// failing MAC and failing length check more
+			// difficult.
+			io.CopyN(ioutil.Discard, r, int64(c.oracleCamouflage))
+		}
+	}
+	return p, err
+}
+
+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
+	// case of block ciphers - this is copied back to the payload later.
+	// How many bytes of payload/padding will be read with this first read.
+	firstBlockLength := uint32((prefixLen + blockSize - 1) / blockSize * blockSize)
+	firstBlock := c.packetData[:firstBlockLength]
+	if _, err := io.ReadFull(r, firstBlock); err != nil {
+		return nil, err
+	}
+
+	c.oracleCamouflage = maxPacket + 4 + c.macSize - firstBlockLength
+
+	c.decrypter.CryptBlocks(firstBlock, firstBlock)
+	length := binary.BigEndian.Uint32(firstBlock[:4])
+	if length > maxPacket {
+		return nil, cbcError("ssh: packet too large")
+	}
+	if length+4 < maxUInt32(cbcMinPacketSize, blockSize) {
+		// The minimum size of a packet is 16 (or the cipher block size, whichever
+		// is larger) bytes.
+		return nil, cbcError("ssh: packet too small")
+	}
+	// The length of the packet (including the length field but not the MAC) must
+	// be a multiple of the block size or 8, whichever is larger.
+	if (length+4)%maxUInt32(cbcMinPacketSizeMultiple, blockSize) != 0 {
+		return nil, cbcError("ssh: invalid packet length multiple")
+	}
+
+	paddingLength := uint32(firstBlock[4])
+	if paddingLength < cbcMinPaddingSize || length <= paddingLength+1 {
+		return nil, cbcError("ssh: invalid packet length")
+	}
+
+	// Positions within the c.packetData buffer:
+	macStart := 4 + length
+	paddingStart := macStart - paddingLength
+
+	// Entire packet size, starting before length, ending at end of mac.
+	entirePacketSize := macStart + c.macSize
+
+	// Ensure c.packetData is large enough for the entire packet data.
+	if uint32(cap(c.packetData)) < entirePacketSize {
+		// Still need to upsize and copy, but this should be rare at runtime, only
+		// on upsizing the packetData buffer.
+		c.packetData = make([]byte, entirePacketSize)
+		copy(c.packetData, firstBlock)
+	} else {
+		c.packetData = c.packetData[:entirePacketSize]
+	}
+
+	n, err := io.ReadFull(r, c.packetData[firstBlockLength:])
+	if err != nil {
+		return nil, err
+	}
+	c.oracleCamouflage -= uint32(n)
+
+	remainingCrypted := c.packetData[firstBlockLength:macStart]
+	c.decrypter.CryptBlocks(remainingCrypted, remainingCrypted)
+
+	mac := c.packetData[macStart:]
+	if c.mac != nil {
+		c.mac.Reset()
+		binary.BigEndian.PutUint32(c.seqNumBytes[:], seqNum)
+		c.mac.Write(c.seqNumBytes[:])
+		c.mac.Write(c.packetData[:macStart])
+		c.macResult = c.mac.Sum(c.macResult[:0])
+		if subtle.ConstantTimeCompare(c.macResult, mac) != 1 {
+			return nil, cbcError("ssh: MAC failure")
+		}
+	}
+
+	return c.packetData[prefixLen:paddingStart], nil
+}
+
+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).
+	// Enforce minimum padding and packet size.
+	encLength := maxUInt32(prefixLen+len(packet)+cbcMinPaddingSize, cbcMinPaddingSize)
+	// Enforce block size.
+	encLength = (encLength + effectiveBlockSize - 1) / effectiveBlockSize * effectiveBlockSize
+
+	length := encLength - 4
+	paddingLength := int(length) - (1 + len(packet))
+
+	// Overall buffer contains: header, payload, padding, mac.
+	// Space for the MAC is reserved in the capacity but not the slice length.
+	bufferSize := encLength + c.macSize
+	if uint32(cap(c.packetData)) < bufferSize {
+		c.packetData = make([]byte, encLength, bufferSize)
+	} else {
+		c.packetData = c.packetData[:encLength]
+	}
+
+	p := c.packetData
+
+	// Packet header.
+	binary.BigEndian.PutUint32(p, length)
+	p = p[4:]
+	p[0] = byte(paddingLength)
+
+	// Payload.
+	p = p[1:]
+	copy(p, packet)
+
+	// Padding.
+	p = p[len(packet):]
+	if _, err := io.ReadFull(rand, p); err != nil {
+		return err
+	}
+
+	if c.mac != nil {
+		c.mac.Reset()
+		binary.BigEndian.PutUint32(c.seqNumBytes[:], seqNum)
+		c.mac.Write(c.seqNumBytes[:])
+		c.mac.Write(c.packetData)
+		// The MAC is now appended into the capacity reserved for it earlier.
+		c.packetData = c.mac.Sum(c.packetData)
+	}
+
+	c.encrypter.CryptBlocks(c.packetData[:encLength], c.packetData[:encLength])
+
+	if _, err := w.Write(c.packetData); err != nil {
+		return err
+	}
+
+	return nil
+}
+
+const chacha20Poly1305ID = "chacha20-poly1305@openssh.com"
+
+// chacha20Poly1305Cipher implements the chacha20-poly1305@openssh.com
+// AEAD, which is described here:
+//
+//   https://tools.ietf.org/html/draft-josefsson-ssh-chacha20-poly1305-openssh-00
+//
+// the methods here also implement padding, which RFC4253 Section 6
+// also requires of stream ciphers.
+type chacha20Poly1305Cipher struct {
+	lengthKey  [32]byte
+	contentKey [32]byte
+	buf        []byte
+}
+
+func newChaCha20Cipher(key, unusedIV, unusedMACKey []byte, unusedAlgs directionAlgorithms) (packetCipher, error) {
+	if len(key) != 64 {
+		panic(len(key))
+	}
+
+	c := &chacha20Poly1305Cipher{
+		buf: make([]byte, 256),
+	}
+
+	copy(c.contentKey[:], key[:32])
+	copy(c.lengthKey[:], key[32:])
+	return c, nil
+}
+
+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.XORKeyStream(discardBuf[:], discardBuf[:]) // skip the next 32 bytes
+
+	encryptedLength := c.buf[:4]
+	if _, err := io.ReadFull(r, encryptedLength); err != nil {
+		return nil, err
+	}
+
+	var lenBytes [4]byte
+	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 {
+		return nil, errors.New("ssh: invalid packet length, packet too large")
+	}
+
+	contentEnd := 4 + length
+	packetEnd := contentEnd + poly1305.TagSize
+	if uint32(cap(c.buf)) < packetEnd {
+		c.buf = make([]byte, packetEnd)
+		copy(c.buf[:], encryptedLength)
+	} else {
+		c.buf = c.buf[:packetEnd]
+	}
+
+	if _, err := io.ReadFull(r, c.buf[4:packetEnd]); err != nil {
+		return nil, err
+	}
+
+	var mac [poly1305.TagSize]byte
+	copy(mac[:], c.buf[contentEnd:packetEnd])
+	if !poly1305.Verify(&mac, c.buf[:contentEnd], &polyKey) {
+		return nil, errors.New("ssh: MAC failure")
+	}
+
+	plain := c.buf[4:contentEnd]
+	s.XORKeyStream(plain, plain)
+
+	padding := plain[0]
+	if padding < 4 {
+		// padding is a byte, so it automatically satisfies
+		// the maximum size, which is 255.
+		return nil, fmt.Errorf("ssh: illegal padding %d", padding)
+	}
+
+	if int(padding)+1 >= len(plain) {
+		return nil, fmt.Errorf("ssh: padding %d too large", padding)
+	}
+
+	plain = plain[1 : len(plain)-int(padding)]
+
+	return plain, nil
+}
+
+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.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.
+	const packetSizeMultiple = 8
+
+	padding := packetSizeMultiple - (1+len(payload))%packetSizeMultiple
+	if padding < 4 {
+		padding += packetSizeMultiple
+	}
+
+	// size (4 bytes), padding (1), payload, padding, tag.
+	totalLength := 4 + 1 + len(payload) + padding + poly1305.TagSize
+	if cap(c.buf) < totalLength {
+		c.buf = make([]byte, totalLength)
+	} else {
+		c.buf = c.buf[:totalLength]
+	}
+
+	binary.BigEndian.PutUint32(c.buf, uint32(1+len(payload)+padding))
+	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
+	if _, err := io.ReadFull(rand, c.buf[5+len(payload):packetEnd]); err != nil {
+		return err
+	}
+
+	s.XORKeyStream(c.buf[4:], c.buf[4:packetEnd])
+
+	var mac [poly1305.TagSize]byte
+	poly1305.Sum(&mac, c.buf[:packetEnd], &polyKey)
+
+	copy(c.buf[packetEnd:], mac[:])
+
+	if _, err := w.Write(c.buf); err != nil {
+		return err
+	}
+	return nil
+}

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

@@ -0,0 +1,278 @@
+// Copyright 2011 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 ssh
+
+import (
+	"bytes"
+	"errors"
+	"fmt"
+	"net"
+	"os"
+	"sync"
+	"time"
+)
+
+// Client implements a traditional SSH client that supports shells,
+// subprocesses, TCP port/streamlocal forwarding and tunneled dialing.
+type Client struct {
+	Conn
+
+	handleForwardsOnce sync.Once // guards calling (*Client).handleForwards
+
+	forwards        forwardList // forwarded tcpip connections from the remote side
+	mu              sync.Mutex
+	channelHandlers map[string]chan NewChannel
+}
+
+// HandleChannelOpen returns a channel on which NewChannel requests
+// for the given type are sent. If the type already is being handled,
+// nil is returned. The channel is closed when the connection is closed.
+func (c *Client) HandleChannelOpen(channelType string) <-chan NewChannel {
+	c.mu.Lock()
+	defer c.mu.Unlock()
+	if c.channelHandlers == nil {
+		// The SSH channel has been closed.
+		c := make(chan NewChannel)
+		close(c)
+		return c
+	}
+
+	ch := c.channelHandlers[channelType]
+	if ch != nil {
+		return nil
+	}
+
+	ch = make(chan NewChannel, chanSize)
+	c.channelHandlers[channelType] = ch
+	return ch
+}
+
+// NewClient creates a Client on top of the given connection.
+func NewClient(c Conn, chans <-chan NewChannel, reqs <-chan *Request) *Client {
+	conn := &Client{
+		Conn:            c,
+		channelHandlers: make(map[string]chan NewChannel, 1),
+	}
+
+	go conn.handleGlobalRequests(reqs)
+	go conn.handleChannelOpens(chans)
+	go func() {
+		conn.Wait()
+		conn.forwards.closeAll()
+	}()
+	return conn
+}
+
+// NewClientConn establishes an authenticated SSH connection using c
+// as the underlying transport.  The Request and NewChannel channels
+// must be serviced or the connection will hang.
+func NewClientConn(c net.Conn, addr string, config *ClientConfig) (Conn, <-chan NewChannel, <-chan *Request, error) {
+	fullConf := *config
+	fullConf.SetDefaults()
+	if fullConf.HostKeyCallback == nil {
+		c.Close()
+		return nil, nil, nil, errors.New("ssh: must specify HostKeyCallback")
+	}
+
+	conn := &connection{
+		sshConn: sshConn{conn: c, user: fullConf.User},
+	}
+
+	if err := conn.clientHandshake(addr, &fullConf); err != nil {
+		c.Close()
+		return nil, nil, nil, fmt.Errorf("ssh: handshake failed: %v", err)
+	}
+	conn.mux = newMux(conn.transport)
+	return conn, conn.mux.incomingChannels, conn.mux.incomingRequests, nil
+}
+
+// clientHandshake performs the client side key exchange. See RFC 4253 Section
+// 7.
+func (c *connection) clientHandshake(dialAddress string, config *ClientConfig) error {
+	if config.ClientVersion != "" {
+		c.clientVersion = []byte(config.ClientVersion)
+	} else {
+		c.clientVersion = []byte(packageVersion)
+	}
+	var err error
+	c.serverVersion, err = exchangeVersions(c.sshConn.conn, c.clientVersion)
+	if err != nil {
+		return err
+	}
+
+	c.transport = newClientTransport(
+		newTransport(c.sshConn.conn, config.Rand, true /* is client */),
+		c.clientVersion, c.serverVersion, config, dialAddress, c.sshConn.RemoteAddr())
+	if err := c.transport.waitSession(); err != nil {
+		return err
+	}
+
+	c.sessionID = c.transport.getSessionID()
+	return c.clientAuthenticate(config)
+}
+
+// verifyHostKeySignature verifies the host key obtained in the key
+// exchange.
+func verifyHostKeySignature(hostKey PublicKey, result *kexResult) error {
+	sig, rest, ok := parseSignatureBody(result.Signature)
+	if len(rest) > 0 || !ok {
+		return errors.New("ssh: signature parse error")
+	}
+
+	return hostKey.Verify(result.H, sig)
+}
+
+// NewSession opens a new Session for this client. (A session is a remote
+// execution of a program.)
+func (c *Client) NewSession() (*Session, error) {
+	ch, in, err := c.OpenChannel("session", nil)
+	if err != nil {
+		return nil, err
+	}
+	return newSession(ch, in)
+}
+
+func (c *Client) handleGlobalRequests(incoming <-chan *Request) {
+	for r := range incoming {
+		// This handles keepalive messages and matches
+		// the behaviour of OpenSSH.
+		r.Reply(false, nil)
+	}
+}
+
+// handleChannelOpens channel open messages from the remote side.
+func (c *Client) handleChannelOpens(in <-chan NewChannel) {
+	for ch := range in {
+		c.mu.Lock()
+		handler := c.channelHandlers[ch.ChannelType()]
+		c.mu.Unlock()
+
+		if handler != nil {
+			handler <- ch
+		} else {
+			ch.Reject(UnknownChannelType, fmt.Sprintf("unknown channel type: %v", ch.ChannelType()))
+		}
+	}
+
+	c.mu.Lock()
+	for _, ch := range c.channelHandlers {
+		close(ch)
+	}
+	c.channelHandlers = nil
+	c.mu.Unlock()
+}
+
+// Dial starts a client connection to the given SSH server. It is a
+// convenience function that connects to the given network address,
+// initiates the SSH handshake, and then sets up a Client.  For access
+// to incoming channels and requests, use net.Dial with NewClientConn
+// instead.
+func Dial(network, addr string, config *ClientConfig) (*Client, error) {
+	conn, err := net.DialTimeout(network, addr, config.Timeout)
+	if err != nil {
+		return nil, err
+	}
+	c, chans, reqs, err := NewClientConn(conn, addr, config)
+	if err != nil {
+		return nil, err
+	}
+	return NewClient(c, chans, reqs), nil
+}
+
+// HostKeyCallback is the function type used for verifying server
+// keys.  A HostKeyCallback must return nil if the host key is OK, or
+// an error to reject it. It receives the hostname as passed to Dial
+// or NewClientConn. The remote address is the RemoteAddr of the
+// net.Conn underlying the SSH connection.
+type HostKeyCallback func(hostname string, remote net.Addr, key PublicKey) error
+
+// BannerCallback is the function type used for treat the banner sent by
+// the server. A BannerCallback receives the message sent by the remote server.
+type BannerCallback func(message string) error
+
+// A ClientConfig structure is used to configure a Client. It must not be
+// modified after having been passed to an SSH function.
+type ClientConfig struct {
+	// Config contains configuration that is shared between clients and
+	// servers.
+	Config
+
+	// User contains the username to authenticate as.
+	User string
+
+	// Auth contains possible authentication methods to use with the
+	// server. Only the first instance of a particular RFC 4252 method will
+	// be used during authentication.
+	Auth []AuthMethod
+
+	// HostKeyCallback is called during the cryptographic
+	// handshake to validate the server's host key. The client
+	// configuration must supply this callback for the connection
+	// to succeed. The functions InsecureIgnoreHostKey or
+	// FixedHostKey can be used for simplistic host key checks.
+	HostKeyCallback HostKeyCallback
+
+	// BannerCallback is called during the SSH dance to display a custom
+	// server's message. The client configuration can supply this callback to
+	// handle it as wished. The function BannerDisplayStderr can be used for
+	// simplistic display on Stderr.
+	BannerCallback BannerCallback
+
+	// ClientVersion contains the version identification string that will
+	// be used for the connection. If empty, a reasonable default is used.
+	ClientVersion string
+
+	// HostKeyAlgorithms lists the key types that the client will
+	// accept from the server as host key, in order of
+	// preference. If empty, a reasonable default is used. Any
+	// string returned from PublicKey.Type method may be used, or
+	// any of the CertAlgoXxxx and KeyAlgoXxxx constants.
+	HostKeyAlgorithms []string
+
+	// Timeout is the maximum amount of time for the TCP connection to establish.
+	//
+	// A Timeout of zero means no timeout.
+	Timeout time.Duration
+}
+
+// InsecureIgnoreHostKey returns a function that can be used for
+// ClientConfig.HostKeyCallback to accept any host key. It should
+// not be used for production code.
+func InsecureIgnoreHostKey() HostKeyCallback {
+	return func(hostname string, remote net.Addr, key PublicKey) error {
+		return nil
+	}
+}
+
+type fixedHostKey struct {
+	key PublicKey
+}
+
+func (f *fixedHostKey) check(hostname string, remote net.Addr, key PublicKey) error {
+	if f.key == nil {
+		return fmt.Errorf("ssh: required host key was nil")
+	}
+	if !bytes.Equal(key.Marshal(), f.key.Marshal()) {
+		return fmt.Errorf("ssh: host key mismatch")
+	}
+	return nil
+}
+
+// FixedHostKey returns a function for use in
+// ClientConfig.HostKeyCallback to accept only a specific host key.
+func FixedHostKey(key PublicKey) HostKeyCallback {
+	hk := &fixedHostKey{key}
+	return hk.check
+}
+
+// BannerDisplayStderr returns a function that can be used for
+// ClientConfig.BannerCallback to display banners on os.Stderr.
+func BannerDisplayStderr() BannerCallback {
+	return func(banner string) error {
+		_, err := os.Stderr.WriteString(banner)
+
+		return err
+	}
+}

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

@@ -0,0 +1,641 @@
+// Copyright 2011 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 ssh
+
+import (
+	"bytes"
+	"errors"
+	"fmt"
+	"io"
+)
+
+type authResult int
+
+const (
+	authFailure authResult = iota
+	authPartialSuccess
+	authSuccess
+)
+
+// clientAuthenticate authenticates with the remote server. See RFC 4252.
+func (c *connection) clientAuthenticate(config *ClientConfig) error {
+	// initiate user auth session
+	if err := c.transport.writePacket(Marshal(&serviceRequestMsg{serviceUserAuth})); err != nil {
+		return err
+	}
+	packet, err := c.transport.readPacket()
+	if err != nil {
+		return err
+	}
+	var serviceAccept serviceAcceptMsg
+	if err := Unmarshal(packet, &serviceAccept); err != nil {
+		return err
+	}
+
+	// during the authentication phase the client first attempts the "none" method
+	// then any untried methods suggested by the server.
+	var tried []string
+	var lastMethods []string
+
+	sessionID := c.transport.getSessionID()
+	for auth := AuthMethod(new(noneAuth)); auth != nil; {
+		ok, methods, err := auth.auth(sessionID, config.User, c.transport, config.Rand)
+		if err != nil {
+			return err
+		}
+		if ok == authSuccess {
+			// success
+			return nil
+		} else if ok == authFailure {
+			if m := auth.method(); !contains(tried, m) {
+				tried = append(tried, m)
+			}
+		}
+		if methods == nil {
+			methods = lastMethods
+		}
+		lastMethods = methods
+
+		auth = nil
+
+	findNext:
+		for _, a := range config.Auth {
+			candidateMethod := a.method()
+			if contains(tried, candidateMethod) {
+				continue
+			}
+			for _, meth := range methods {
+				if meth == candidateMethod {
+					auth = a
+					break findNext
+				}
+			}
+		}
+	}
+	return fmt.Errorf("ssh: unable to authenticate, attempted methods %v, no supported methods remain", tried)
+}
+
+func contains(list []string, e string) bool {
+	for _, s := range list {
+		if s == e {
+			return true
+		}
+	}
+	return false
+}
+
+// An AuthMethod represents an instance of an RFC 4252 authentication method.
+type AuthMethod interface {
+	// auth authenticates user over transport t.
+	// Returns true if authentication is successful.
+	// If authentication is not successful, a []string of alternative
+	// method names is returned. If the slice is nil, it will be ignored
+	// and the previous set of possible methods will be reused.
+	auth(session []byte, user string, p packetConn, rand io.Reader) (authResult, []string, error)
+
+	// method returns the RFC 4252 method name.
+	method() string
+}
+
+// "none" authentication, RFC 4252 section 5.2.
+type noneAuth int
+
+func (n *noneAuth) auth(session []byte, user string, c packetConn, rand io.Reader) (authResult, []string, error) {
+	if err := c.writePacket(Marshal(&userAuthRequestMsg{
+		User:    user,
+		Service: serviceSSH,
+		Method:  "none",
+	})); err != nil {
+		return authFailure, nil, err
+	}
+
+	return handleAuthResponse(c)
+}
+
+func (n *noneAuth) method() string {
+	return "none"
+}
+
+// passwordCallback is an AuthMethod that fetches the password through
+// a function call, e.g. by prompting the user.
+type passwordCallback func() (password string, err error)
+
+func (cb passwordCallback) auth(session []byte, user string, c packetConn, rand io.Reader) (authResult, []string, error) {
+	type passwordAuthMsg struct {
+		User     string `sshtype:"50"`
+		Service  string
+		Method   string
+		Reply    bool
+		Password string
+	}
+
+	pw, err := cb()
+	// REVIEW NOTE: is there a need to support skipping a password attempt?
+	// The program may only find out that the user doesn't have a password
+	// when prompting.
+	if err != nil {
+		return authFailure, nil, err
+	}
+
+	if err := c.writePacket(Marshal(&passwordAuthMsg{
+		User:     user,
+		Service:  serviceSSH,
+		Method:   cb.method(),
+		Reply:    false,
+		Password: pw,
+	})); err != nil {
+		return authFailure, nil, err
+	}
+
+	return handleAuthResponse(c)
+}
+
+func (cb passwordCallback) method() string {
+	return "password"
+}
+
+// Password returns an AuthMethod using the given password.
+func Password(secret string) AuthMethod {
+	return passwordCallback(func() (string, error) { return secret, nil })
+}
+
+// PasswordCallback returns an AuthMethod that uses a callback for
+// fetching a password.
+func PasswordCallback(prompt func() (secret string, err error)) AuthMethod {
+	return passwordCallback(prompt)
+}
+
+type publickeyAuthMsg struct {
+	User    string `sshtype:"50"`
+	Service string
+	Method  string
+	// HasSig indicates to the receiver packet that the auth request is signed and
+	// should be used for authentication of the request.
+	HasSig   bool
+	Algoname string
+	PubKey   []byte
+	// Sig is tagged with "rest" so Marshal will exclude it during
+	// validateKey
+	Sig []byte `ssh:"rest"`
+}
+
+// publicKeyCallback is an AuthMethod that uses a set of key
+// pairs for authentication.
+type publicKeyCallback func() ([]Signer, error)
+
+func (cb publicKeyCallback) method() string {
+	return "publickey"
+}
+
+func (cb publicKeyCallback) auth(session []byte, user string, c packetConn, rand io.Reader) (authResult, []string, error) {
+	// Authentication is performed by sending an enquiry to test if a key is
+	// acceptable to the remote. If the key is acceptable, the client will
+	// attempt to authenticate with the valid key.  If not the client will repeat
+	// the process with the remaining keys.
+
+	signers, err := cb()
+	if err != nil {
+		return authFailure, nil, err
+	}
+	var methods []string
+	for _, signer := range signers {
+		ok, err := validateKey(signer.PublicKey(), user, c)
+		if err != nil {
+			return authFailure, nil, err
+		}
+		if !ok {
+			continue
+		}
+
+		pub := signer.PublicKey()
+		pubKey := pub.Marshal()
+		sign, err := signer.Sign(rand, buildDataSignedForAuth(session, userAuthRequestMsg{
+			User:    user,
+			Service: serviceSSH,
+			Method:  cb.method(),
+		}, []byte(pub.Type()), pubKey))
+		if err != nil {
+			return authFailure, nil, err
+		}
+
+		// manually wrap the serialized signature in a string
+		s := Marshal(sign)
+		sig := make([]byte, stringLength(len(s)))
+		marshalString(sig, s)
+		msg := publickeyAuthMsg{
+			User:     user,
+			Service:  serviceSSH,
+			Method:   cb.method(),
+			HasSig:   true,
+			Algoname: pub.Type(),
+			PubKey:   pubKey,
+			Sig:      sig,
+		}
+		p := Marshal(&msg)
+		if err := c.writePacket(p); err != nil {
+			return authFailure, nil, err
+		}
+		var success authResult
+		success, methods, err = handleAuthResponse(c)
+		if err != nil {
+			return authFailure, nil, err
+		}
+
+		// If authentication succeeds or the list of available methods does not
+		// contain the "publickey" method, do not attempt to authenticate with any
+		// other keys.  According to RFC 4252 Section 7, the latter can occur when
+		// additional authentication methods are required.
+		if success == authSuccess || !containsMethod(methods, cb.method()) {
+			return success, methods, err
+		}
+	}
+
+	return authFailure, methods, nil
+}
+
+func containsMethod(methods []string, method string) bool {
+	for _, m := range methods {
+		if m == method {
+			return true
+		}
+	}
+
+	return false
+}
+
+// validateKey validates the key provided is acceptable to the server.
+func validateKey(key PublicKey, user string, c packetConn) (bool, error) {
+	pubKey := key.Marshal()
+	msg := publickeyAuthMsg{
+		User:     user,
+		Service:  serviceSSH,
+		Method:   "publickey",
+		HasSig:   false,
+		Algoname: key.Type(),
+		PubKey:   pubKey,
+	}
+	if err := c.writePacket(Marshal(&msg)); err != nil {
+		return false, err
+	}
+
+	return confirmKeyAck(key, c)
+}
+
+func confirmKeyAck(key PublicKey, c packetConn) (bool, error) {
+	pubKey := key.Marshal()
+	algoname := key.Type()
+
+	for {
+		packet, err := c.readPacket()
+		if err != nil {
+			return false, err
+		}
+		switch packet[0] {
+		case msgUserAuthBanner:
+			if err := handleBannerResponse(c, packet); err != nil {
+				return false, err
+			}
+		case msgUserAuthPubKeyOk:
+			var msg userAuthPubKeyOkMsg
+			if err := Unmarshal(packet, &msg); err != nil {
+				return false, err
+			}
+			if msg.Algo != algoname || !bytes.Equal(msg.PubKey, pubKey) {
+				return false, nil
+			}
+			return true, nil
+		case msgUserAuthFailure:
+			return false, nil
+		default:
+			return false, unexpectedMessageError(msgUserAuthSuccess, packet[0])
+		}
+	}
+}
+
+// PublicKeys returns an AuthMethod that uses the given key
+// pairs.
+func PublicKeys(signers ...Signer) AuthMethod {
+	return publicKeyCallback(func() ([]Signer, error) { return signers, nil })
+}
+
+// PublicKeysCallback returns an AuthMethod that runs the given
+// function to obtain a list of key pairs.
+func PublicKeysCallback(getSigners func() (signers []Signer, err error)) AuthMethod {
+	return publicKeyCallback(getSigners)
+}
+
+// handleAuthResponse returns whether the preceding authentication request succeeded
+// along with a list of remaining authentication methods to try next and
+// an error if an unexpected response was received.
+func handleAuthResponse(c packetConn) (authResult, []string, error) {
+	for {
+		packet, err := c.readPacket()
+		if err != nil {
+			return authFailure, nil, err
+		}
+
+		switch packet[0] {
+		case msgUserAuthBanner:
+			if err := handleBannerResponse(c, packet); err != nil {
+				return authFailure, nil, err
+			}
+		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 msgUserAuthSuccess:
+			return authSuccess, nil, nil
+		default:
+			return authFailure, nil, unexpectedMessageError(msgUserAuthSuccess, packet[0])
+		}
+	}
+}
+
+func handleBannerResponse(c packetConn, packet []byte) error {
+	var msg userAuthBannerMsg
+	if err := Unmarshal(packet, &msg); err != nil {
+		return err
+	}
+
+	transport, ok := c.(*handshakeTransport)
+	if !ok {
+		return nil
+	}
+
+	if transport.bannerCallback != nil {
+		return transport.bannerCallback(msg.Message)
+	}
+
+	return nil
+}
+
+// KeyboardInteractiveChallenge should print questions, optionally
+// disabling echoing (e.g. for passwords), and return all the answers.
+// Challenge may be called multiple times in a single session. After
+// successful authentication, the server may send a challenge with no
+// questions, for which the user and instruction messages should be
+// printed.  RFC 4256 section 3.3 details how the UI should behave for
+// both CLI and GUI environments.
+type KeyboardInteractiveChallenge func(user, instruction string, questions []string, echos []bool) (answers []string, err error)
+
+// KeyboardInteractive returns an AuthMethod using a prompt/response
+// sequence controlled by the server.
+func KeyboardInteractive(challenge KeyboardInteractiveChallenge) AuthMethod {
+	return challenge
+}
+
+func (cb KeyboardInteractiveChallenge) method() string {
+	return "keyboard-interactive"
+}
+
+func (cb KeyboardInteractiveChallenge) auth(session []byte, user string, c packetConn, rand io.Reader) (authResult, []string, error) {
+	type initiateMsg struct {
+		User       string `sshtype:"50"`
+		Service    string
+		Method     string
+		Language   string
+		Submethods string
+	}
+
+	if err := c.writePacket(Marshal(&initiateMsg{
+		User:    user,
+		Service: serviceSSH,
+		Method:  "keyboard-interactive",
+	})); err != nil {
+		return authFailure, nil, err
+	}
+
+	for {
+		packet, err := c.readPacket()
+		if err != nil {
+			return authFailure, nil, err
+		}
+
+		// like handleAuthResponse, but with less options.
+		switch packet[0] {
+		case msgUserAuthBanner:
+			if err := handleBannerResponse(c, packet); err != nil {
+				return authFailure, nil, err
+			}
+			continue
+		case msgUserAuthInfoRequest:
+			// OK
+		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 msgUserAuthSuccess:
+			return authSuccess, nil, nil
+		default:
+			return authFailure, nil, unexpectedMessageError(msgUserAuthInfoRequest, packet[0])
+		}
+
+		var msg userAuthInfoRequestMsg
+		if err := Unmarshal(packet, &msg); err != nil {
+			return authFailure, nil, err
+		}
+
+		// Manually unpack the prompt/echo pairs.
+		rest := msg.Prompts
+		var prompts []string
+		var echos []bool
+		for i := 0; i < int(msg.NumPrompts); i++ {
+			prompt, r, ok := parseString(rest)
+			if !ok || len(r) == 0 {
+				return authFailure, nil, errors.New("ssh: prompt format error")
+			}
+			prompts = append(prompts, string(prompt))
+			echos = append(echos, r[0] != 0)
+			rest = r[1:]
+		}
+
+		if len(rest) != 0 {
+			return authFailure, nil, errors.New("ssh: extra data following keyboard-interactive pairs")
+		}
+
+		answers, err := cb(msg.User, msg.Instruction, prompts, echos)
+		if err != nil {
+			return authFailure, nil, err
+		}
+
+		if len(answers) != len(prompts) {
+			return authFailure, nil, fmt.Errorf("ssh: incorrect number of answers from keyboard-interactive callback %d (expected %d)", len(answers), len(prompts))
+		}
+		responseLength := 1 + 4
+		for _, a := range answers {
+			responseLength += stringLength(len(a))
+		}
+		serialized := make([]byte, responseLength)
+		p := serialized
+		p[0] = msgUserAuthInfoResponse
+		p = p[1:]
+		p = marshalUint32(p, uint32(len(answers)))
+		for _, a := range answers {
+			p = marshalString(p, []byte(a))
+		}
+
+		if err := c.writePacket(serialized); err != nil {
+			return authFailure, nil, err
+		}
+	}
+}
+
+type retryableAuthMethod struct {
+	authMethod AuthMethod
+	maxTries   int
+}
+
+func (r *retryableAuthMethod) auth(session []byte, user string, c packetConn, rand io.Reader) (ok authResult, methods []string, err error) {
+	for i := 0; r.maxTries <= 0 || i < r.maxTries; i++ {
+		ok, methods, err = r.authMethod.auth(session, user, c, rand)
+		if ok != authFailure || err != nil { // either success, partial success or error terminate
+			return ok, methods, err
+		}
+	}
+	return ok, methods, err
+}
+
+func (r *retryableAuthMethod) method() string {
+	return r.authMethod.method()
+}
+
+// RetryableAuthMethod is a decorator for other auth methods enabling them to
+// be retried up to maxTries before considering that AuthMethod itself failed.
+// If maxTries is <= 0, will retry indefinitely
+//
+// This is useful for interactive clients using challenge/response type
+// authentication (e.g. Keyboard-Interactive, Password, etc) where the user
+// could mistype their response resulting in the server issuing a
+// SSH_MSG_USERAUTH_FAILURE (rfc4252 #8 [password] and rfc4256 #3.4
+// [keyboard-interactive]); Without this decorator, the non-retryable
+// AuthMethod would be removed from future consideration, and never tried again
+// (and so the user would never be able to retry their entry).
+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"
+}

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

@@ -0,0 +1,404 @@
+// Copyright 2011 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 ssh
+
+import (
+	"crypto"
+	"crypto/rand"
+	"fmt"
+	"io"
+	"math"
+	"sync"
+
+	_ "crypto/sha1"
+	_ "crypto/sha256"
+	_ "crypto/sha512"
+)
+
+// These are string constants in the SSH protocol.
+const (
+	compressionNone = "none"
+	serviceUserAuth = "ssh-userauth"
+	serviceSSH      = "ssh-connection"
+)
+
+// supportedCiphers lists ciphers we support but might not recommend.
+var supportedCiphers = []string{
+	"aes128-ctr", "aes192-ctr", "aes256-ctr",
+	"aes128-gcm@openssh.com",
+	chacha20Poly1305ID,
+	"arcfour256", "arcfour128", "arcfour",
+	aes128cbcID,
+	tripledescbcID,
+}
+
+// preferredCiphers specifies the default preference for ciphers.
+var preferredCiphers = []string{
+	"aes128-gcm@openssh.com",
+	chacha20Poly1305ID,
+	"aes128-ctr", "aes192-ctr", "aes256-ctr",
+}
+
+// supportedKexAlgos specifies the supported key-exchange algorithms in
+// preference order.
+var supportedKexAlgos = []string{
+	kexAlgoCurve25519SHA256,
+	// P384 and P521 are not constant-time yet, but since we don't
+	// reuse ephemeral keys, using them for ECDH should be OK.
+	kexAlgoECDH256, kexAlgoECDH384, kexAlgoECDH521,
+	kexAlgoDH14SHA1, kexAlgoDH1SHA1,
+}
+
+// serverForbiddenKexAlgos contains key exchange algorithms, that are forbidden
+// for the server half.
+var serverForbiddenKexAlgos = map[string]struct{}{
+	kexAlgoDHGEXSHA1:   {}, // server half implementation is only minimal to satisfy the automated tests
+	kexAlgoDHGEXSHA256: {}, // server half implementation is only minimal to satisfy the automated tests
+}
+
+// preferredKexAlgos specifies the default preference for key-exchange algorithms
+// in preference order.
+var preferredKexAlgos = []string{
+	kexAlgoCurve25519SHA256,
+	kexAlgoECDH256, kexAlgoECDH384, kexAlgoECDH521,
+	kexAlgoDH14SHA1,
+}
+
+// supportedHostKeyAlgos specifies the supported host-key algorithms (i.e. methods
+// of authenticating servers) in preference order.
+var supportedHostKeyAlgos = []string{
+	CertAlgoRSAv01, CertAlgoDSAv01, CertAlgoECDSA256v01,
+	CertAlgoECDSA384v01, CertAlgoECDSA521v01, CertAlgoED25519v01,
+
+	KeyAlgoECDSA256, KeyAlgoECDSA384, KeyAlgoECDSA521,
+	KeyAlgoRSA, KeyAlgoDSA,
+
+	KeyAlgoED25519,
+}
+
+// supportedMACs specifies a default set of MAC algorithms in preference order.
+// This is based on RFC 4253, section 6.4, but with hmac-md5 variants removed
+// because they have reached the end of their useful life.
+var supportedMACs = []string{
+	"hmac-sha2-256-etm@openssh.com", "hmac-sha2-256", "hmac-sha1", "hmac-sha1-96",
+}
+
+var supportedCompressions = []string{compressionNone}
+
+// hashFuncs keeps the mapping of supported algorithms to their respective
+// hashes needed for signature verification.
+var hashFuncs = map[string]crypto.Hash{
+	KeyAlgoRSA:          crypto.SHA1,
+	KeyAlgoDSA:          crypto.SHA1,
+	KeyAlgoECDSA256:     crypto.SHA256,
+	KeyAlgoECDSA384:     crypto.SHA384,
+	KeyAlgoECDSA521:     crypto.SHA512,
+	CertAlgoRSAv01:      crypto.SHA1,
+	CertAlgoDSAv01:      crypto.SHA1,
+	CertAlgoECDSA256v01: crypto.SHA256,
+	CertAlgoECDSA384v01: crypto.SHA384,
+	CertAlgoECDSA521v01: crypto.SHA512,
+}
+
+// unexpectedMessageError results when the SSH message that we received didn't
+// match what we wanted.
+func unexpectedMessageError(expected, got uint8) error {
+	return fmt.Errorf("ssh: unexpected message type %d (expected %d)", got, expected)
+}
+
+// parseError results from a malformed SSH message.
+func parseError(tag uint8) error {
+	return fmt.Errorf("ssh: parse error in message type %d", tag)
+}
+
+func findCommon(what string, client []string, server []string) (common string, err error) {
+	for _, c := range client {
+		for _, s := range server {
+			if c == s {
+				return c, nil
+			}
+		}
+	}
+	return "", fmt.Errorf("ssh: no common algorithm for %s; client offered: %v, server offered: %v", what, client, server)
+}
+
+// directionAlgorithms records algorithm choices in one direction (either read or write)
+type directionAlgorithms struct {
+	Cipher      string
+	MAC         string
+	Compression string
+}
+
+// rekeyBytes returns a rekeying intervals in bytes.
+func (a *directionAlgorithms) rekeyBytes() int64 {
+	// According to RFC4344 block ciphers should rekey after
+	// 2^(BLOCKSIZE/4) blocks. For all AES flavors BLOCKSIZE is
+	// 128.
+	switch a.Cipher {
+	case "aes128-ctr", "aes192-ctr", "aes256-ctr", gcmCipherID, aes128cbcID:
+		return 16 * (1 << 32)
+
+	}
+
+	// For others, stick with RFC4253 recommendation to rekey after 1 Gb of data.
+	return 1 << 30
+}
+
+type algorithms struct {
+	kex     string
+	hostKey string
+	w       directionAlgorithms
+	r       directionAlgorithms
+}
+
+func findAgreedAlgorithms(isClient bool, clientKexInit, serverKexInit *kexInitMsg) (algs *algorithms, err error) {
+	result := &algorithms{}
+
+	result.kex, err = findCommon("key exchange", clientKexInit.KexAlgos, serverKexInit.KexAlgos)
+	if err != nil {
+		return
+	}
+
+	result.hostKey, err = findCommon("host key", clientKexInit.ServerHostKeyAlgos, serverKexInit.ServerHostKeyAlgos)
+	if err != nil {
+		return
+	}
+
+	stoc, ctos := &result.w, &result.r
+	if isClient {
+		ctos, stoc = stoc, ctos
+	}
+
+	ctos.Cipher, err = findCommon("client to server cipher", clientKexInit.CiphersClientServer, serverKexInit.CiphersClientServer)
+	if err != nil {
+		return
+	}
+
+	stoc.Cipher, err = findCommon("server to client cipher", clientKexInit.CiphersServerClient, serverKexInit.CiphersServerClient)
+	if err != nil {
+		return
+	}
+
+	ctos.MAC, err = findCommon("client to server MAC", clientKexInit.MACsClientServer, serverKexInit.MACsClientServer)
+	if err != nil {
+		return
+	}
+
+	stoc.MAC, err = findCommon("server to client MAC", clientKexInit.MACsServerClient, serverKexInit.MACsServerClient)
+	if err != nil {
+		return
+	}
+
+	ctos.Compression, err = findCommon("client to server compression", clientKexInit.CompressionClientServer, serverKexInit.CompressionClientServer)
+	if err != nil {
+		return
+	}
+
+	stoc.Compression, err = findCommon("server to client compression", clientKexInit.CompressionServerClient, serverKexInit.CompressionServerClient)
+	if err != nil {
+		return
+	}
+
+	return result, nil
+}
+
+// If rekeythreshold is too small, we can't make any progress sending
+// stuff.
+const minRekeyThreshold uint64 = 256
+
+// Config contains configuration data common to both ServerConfig and
+// ClientConfig.
+type Config struct {
+	// Rand provides the source of entropy for cryptographic
+	// primitives. If Rand is nil, the cryptographic random reader
+	// in package crypto/rand will be used.
+	Rand io.Reader
+
+	// The maximum number of bytes sent or received after which a
+	// new key is negotiated. It must be at least 256. If
+	// unspecified, a size suitable for the chosen cipher is used.
+	RekeyThreshold uint64
+
+	// The allowed key exchanges algorithms. If unspecified then a
+	// default set of algorithms is used.
+	KeyExchanges []string
+
+	// The allowed cipher algorithms. If unspecified then a sensible
+	// default is used.
+	Ciphers []string
+
+	// The allowed MAC algorithms. If unspecified then a sensible default
+	// is used.
+	MACs []string
+}
+
+// SetDefaults sets sensible values for unset fields in config. This is
+// exported for testing: Configs passed to SSH functions are copied and have
+// default values set automatically.
+func (c *Config) SetDefaults() {
+	if c.Rand == nil {
+		c.Rand = rand.Reader
+	}
+	if c.Ciphers == nil {
+		c.Ciphers = preferredCiphers
+	}
+	var ciphers []string
+	for _, c := range c.Ciphers {
+		if cipherModes[c] != nil {
+			// reject the cipher if we have no cipherModes definition
+			ciphers = append(ciphers, c)
+		}
+	}
+	c.Ciphers = ciphers
+
+	if c.KeyExchanges == nil {
+		c.KeyExchanges = preferredKexAlgos
+	}
+
+	if c.MACs == nil {
+		c.MACs = supportedMACs
+	}
+
+	if c.RekeyThreshold == 0 {
+		// cipher specific default
+	} else if c.RekeyThreshold < minRekeyThreshold {
+		c.RekeyThreshold = minRekeyThreshold
+	} else if c.RekeyThreshold >= math.MaxInt64 {
+		// Avoid weirdness if somebody uses -1 as a threshold.
+		c.RekeyThreshold = math.MaxInt64
+	}
+}
+
+// buildDataSignedForAuth returns the data that is signed in order to prove
+// possession of a private key. See RFC 4252, section 7.
+func buildDataSignedForAuth(sessionID []byte, req userAuthRequestMsg, algo, pubKey []byte) []byte {
+	data := struct {
+		Session []byte
+		Type    byte
+		User    string
+		Service string
+		Method  string
+		Sign    bool
+		Algo    []byte
+		PubKey  []byte
+	}{
+		sessionID,
+		msgUserAuthRequest,
+		req.User,
+		req.Service,
+		req.Method,
+		true,
+		algo,
+		pubKey,
+	}
+	return Marshal(data)
+}
+
+func appendU16(buf []byte, n uint16) []byte {
+	return append(buf, byte(n>>8), byte(n))
+}
+
+func appendU32(buf []byte, n uint32) []byte {
+	return append(buf, byte(n>>24), byte(n>>16), byte(n>>8), byte(n))
+}
+
+func appendU64(buf []byte, n uint64) []byte {
+	return append(buf,
+		byte(n>>56), byte(n>>48), byte(n>>40), byte(n>>32),
+		byte(n>>24), byte(n>>16), byte(n>>8), byte(n))
+}
+
+func appendInt(buf []byte, n int) []byte {
+	return appendU32(buf, uint32(n))
+}
+
+func appendString(buf []byte, s string) []byte {
+	buf = appendU32(buf, uint32(len(s)))
+	buf = append(buf, s...)
+	return buf
+}
+
+func appendBool(buf []byte, b bool) []byte {
+	if b {
+		return append(buf, 1)
+	}
+	return append(buf, 0)
+}
+
+// newCond is a helper to hide the fact that there is no usable zero
+// value for sync.Cond.
+func newCond() *sync.Cond { return sync.NewCond(new(sync.Mutex)) }
+
+// window represents the buffer available to clients
+// wishing to write to a channel.
+type window struct {
+	*sync.Cond
+	win          uint32 // RFC 4254 5.2 says the window size can grow to 2^32-1
+	writeWaiters int
+	closed       bool
+}
+
+// add adds win to the amount of window available
+// for consumers.
+func (w *window) add(win uint32) bool {
+	// a zero sized window adjust is a noop.
+	if win == 0 {
+		return true
+	}
+	w.L.Lock()
+	if w.win+win < win {
+		w.L.Unlock()
+		return false
+	}
+	w.win += win
+	// It is unusual that multiple goroutines would be attempting to reserve
+	// window space, but not guaranteed. Use broadcast to notify all waiters
+	// that additional window is available.
+	w.Broadcast()
+	w.L.Unlock()
+	return true
+}
+
+// close sets the window to closed, so all reservations fail
+// immediately.
+func (w *window) close() {
+	w.L.Lock()
+	w.closed = true
+	w.Broadcast()
+	w.L.Unlock()
+}
+
+// reserve reserves win from the available window capacity.
+// If no capacity remains, reserve will block. reserve may
+// return less than requested.
+func (w *window) reserve(win uint32) (uint32, error) {
+	var err error
+	w.L.Lock()
+	w.writeWaiters++
+	w.Broadcast()
+	for w.win == 0 && !w.closed {
+		w.Wait()
+	}
+	w.writeWaiters--
+	if w.win < win {
+		win = w.win
+	}
+	w.win -= win
+	if w.closed {
+		err = io.EOF
+	}
+	w.L.Unlock()
+	return win, err
+}
+
+// waitWriterBlocked waits until some goroutine is blocked for further
+// writes. It is used in tests only.
+func (w *window) waitWriterBlocked() {
+	w.Cond.L.Lock()
+	for w.writeWaiters == 0 {
+		w.Cond.Wait()
+	}
+	w.Cond.L.Unlock()
+}