TUN vectorised reads/writes (#1145)

This PR updates the Wireguard dependency and updates to use new
vectorised reads/writes, which should reduce the number of syscalls and
improve performance.

This will only make a difference on Linux as this is the only platform
for which the Wireguard TUN library supports vectorised reads/writes.
For other platforms, single reads and writes will be performed as usual.

---------

Co-authored-by: Neil Alexander <neilalexander@users.noreply.github.com>

go.mod

@@ -17,7 +17,7 @@ require (
 	golang.org/x/net v0.25.0
 	golang.org/x/sys v0.20.0
 	golang.org/x/text v0.15.0
-	golang.zx2c4.com/wireguard v0.0.0-20230223181233-21636207a675
+	golang.zx2c4.com/wireguard v0.0.0-20231211153847-12269c276173
 	golang.zx2c4.com/wireguard/windows v0.5.3
 )
 

go.sum

@@ -27,6 +27,8 @@ github.com/golang/protobuf v1.5.3 h1:KhyjKVUg7Usr/dYsdSqoFveMYd5ko72D+zANwlG1mmg
 github.com/golang/protobuf v1.5.3/go.mod h1:XVQd3VNwM+JqD3oG2Ue2ip4fOMUkwXdXDdiuN0vRsmY=
 github.com/gologme/log v1.3.0 h1:l781G4dE+pbigClDSDzSaaYKtiueHCILUa/qSDsmHAo=
 github.com/gologme/log v1.3.0/go.mod h1:yKT+DvIPdDdDoPtqFrFxheooyVmoqi0BAsw+erN3wA4=
+github.com/google/btree v1.0.1 h1:gK4Kx5IaGY9CD5sPJ36FHiBJ6ZXl0kilRiiCj+jdYp4=
+github.com/google/btree v1.0.1/go.mod h1:xXMiIv4Fb/0kKde4SpL7qlzvu5cMJDRkFDxJfI9uaxA=
 github.com/google/go-cmp v0.6.0 h1:ofyhxvXcZhMsU5ulbFiLKl/XBFqE1GSq7atu8tAmTRI=
 github.com/google/go-cmp v0.6.0/go.mod h1:17dUlkBOakJ0+DkrSSNjCkIjxS6bF9zb3elmeNGIjoY=
 github.com/google/pprof v0.0.0-20210407192527-94a9f03dee38 h1:yAJXTCF9TqKcTiHJAE8dj7HMvPfh66eeA2JYW7eFpSE=
@@ -137,8 +139,8 @@ golang.org/x/tools v0.21.0/go.mod h1:aiJjzUbINMkxbQROHiO6hDPo2LHcIPhhQsa9DLh0yGk
 golang.org/x/xerrors v0.0.0-20190717185122-a985d3407aa7/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2 h1:B82qJJgjvYKsXS9jeunTOisW56dUokqW/FOteYJJ/yg=
 golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2/go.mod h1:deeaetjYA+DHMHg+sMSMI58GrEteJUUzzw7en6TJQcI=
-golang.zx2c4.com/wireguard v0.0.0-20230223181233-21636207a675 h1:/J/RVnr7ng4fWPRH3xa4WtBJ1Jp+Auu4YNLmGiPv5QU=
+golang.zx2c4.com/wireguard v0.0.0-20231211153847-12269c276173 h1:/jFs0duh4rdb8uIfPMv78iAJGcPKDeqAFnaLBropIC4=
-golang.zx2c4.com/wireguard v0.0.0-20230223181233-21636207a675/go.mod h1:whfbyDBt09xhCYQWtO2+3UVjlaq6/9hDZrjg2ZE6SyA=
+golang.zx2c4.com/wireguard v0.0.0-20231211153847-12269c276173/go.mod h1:tkCQ4FQXmpAgYVh++1cq16/dH4QJtmvpRv19DWGAHSA=
 golang.zx2c4.com/wireguard/windows v0.5.3 h1:On6j2Rpn3OEMXqBq00QEDC7bWSZrPIHKIus8eIuExIE=
 golang.zx2c4.com/wireguard/windows v0.5.3/go.mod h1:9TEe8TJmtwyQebdFwAkEWOPr3prrtqm+REGFifP60hI=
 google.golang.org/protobuf v1.28.0 h1:w43yiav+6bVFTBQFZX0r7ipe9JQ1QsbMgHwbBziscLw=
@@ -147,3 +149,5 @@ gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8
 gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
 gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
 gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
+gvisor.dev/gvisor v0.0.0-20230927004350-cbd86285d259 h1:TbRPT0HtzFP3Cno1zZo7yPzEEnfu8EjLfl6IU9VfqkQ=
+gvisor.dev/gvisor v0.0.0-20230927004350-cbd86285d259/go.mod h1:AVgIgHMwK63XvmAzWG9vLQ41YnVHN0du0tEC46fI7yY=

src/tun/iface.go

@@ -1,42 +1,60 @@
 package tun
 
-const TUN_OFFSET_BYTES = 4
+const TUN_OFFSET_BYTES = 80 // sizeof(virtio_net_hdr)
 
 func (tun *TunAdapter) read() {
-	var buf [TUN_OFFSET_BYTES + 65535]byte
+	vs := tun.iface.BatchSize()
+	bufs := make([][]byte, vs)
+	sizes := make([]int, vs)
+	for i := range bufs {
+		bufs[i] = make([]byte, TUN_OFFSET_BYTES+65535)
+	}
 	for {
-		n, err := tun.iface.Read(buf[:], TUN_OFFSET_BYTES)
+		n, err := tun.iface.Read(bufs, sizes, TUN_OFFSET_BYTES)
-		if n <= TUN_OFFSET_BYTES || err != nil {
+		if err != nil {
 			tun.log.Errorln("Error reading TUN:", err)
-			ferr := tun.iface.Flush()
-			if ferr != nil {
-				tun.log.Errorln("Unable to flush packets:", ferr)
-			}
 			return
 		}
-		begin := TUN_OFFSET_BYTES
+		for i, b := range bufs[:n] {
-		end := begin + n
+			if _, err := tun.rwc.Write(b[TUN_OFFSET_BYTES : TUN_OFFSET_BYTES+sizes[i]]); err != nil {
-		bs := buf[begin:end]
+				tun.log.Debugln("Unable to send packet:", err)
-		if _, err := tun.rwc.Write(bs); err != nil {
+			}
-			tun.log.Debugln("Unable to send packet:", err)
 		}
 	}
 }
 
-func (tun *TunAdapter) write() {
+func (tun *TunAdapter) queue() {
-	var buf [TUN_OFFSET_BYTES + 65535]byte
 	for {
-		bs := buf[TUN_OFFSET_BYTES:]
+		p := bufPool.Get().([]byte)[:bufPoolSize]
-		n, err := tun.rwc.Read(bs)
+		n, err := tun.rwc.Read(p)
 		if err != nil {
 			tun.log.Errorln("Exiting TUN writer due to core read error:", err)
 			return
 		}
+		tun.ch <- p[:n]
+	}
+}
+
+func (tun *TunAdapter) write() {
+	vs := cap(tun.ch)
+	bufs := make([][]byte, vs)
+	for i := range bufs {
+		bufs[i] = make([]byte, TUN_OFFSET_BYTES+65535)
+	}
+	for {
+		n := len(tun.ch)
+		if n == 0 {
+			n = 1 // Nothing queued up yet, wait for it instead
+		}
+		for i := 0; i < n; i++ {
+			msg := <-tun.ch
+			bufs[i] = append(bufs[i][:TUN_OFFSET_BYTES], msg...)
+			bufPool.Put(msg) // nolint:staticcheck
+		}
 		if !tun.isEnabled {
 			continue // Nothing to do, the tun isn't enabled
 		}
-		bs = buf[:TUN_OFFSET_BYTES+n]
+		if _, err := tun.iface.Write(bufs[:n], TUN_OFFSET_BYTES); err != nil {
-		if _, err = tun.iface.Write(bs, TUN_OFFSET_BYTES); err != nil {
 			tun.Act(nil, func() {
 				if !tun.isOpen {
 					tun.log.Errorln("TUN iface write error:", err)

src/tun/tun.go

@@ -10,9 +10,11 @@ import (
 	"fmt"
 	"io"
 	"net"
+	"sync"
+	"time"
 
 	"github.com/Arceliar/phony"
-	"golang.zx2c4.com/wireguard/tun"
+	wgtun "golang.zx2c4.com/wireguard/tun"
 
 	"github.com/yggdrasil-network/yggdrasil-go/src/address"
 	"github.com/yggdrasil-network/yggdrasil-go/src/config"
@@ -39,7 +41,7 @@ type TunAdapter struct {
 	addr        address.Address
 	subnet      address.Subnet
 	mtu         uint64
-	iface       tun.Device
+	iface       wgtun.Device
 	phony.Inbox // Currently only used for _handlePacket from the reader, TODO: all the stuff that currently needs a mutex below
 	isOpen      bool
 	isEnabled   bool // Used by the writer to drop sessionTraffic if not enabled
@@ -48,6 +50,7 @@ type TunAdapter struct {
 		name InterfaceName
 		mtu  InterfaceMTU
 	}
+	ch chan []byte
 }
 
 // Gets the maximum supported MTU for the platform based on the defaults in
@@ -62,6 +65,20 @@ func getSupportedMTU(mtu uint64) uint64 {
 	return mtu
 }
 
+func waitForTUNUp(ch <-chan wgtun.Event) bool {
+	t := time.After(time.Second * 5)
+	for {
+		select {
+		case ev := <-ch:
+			if ev == wgtun.EventUp {
+				return true
+			}
+		case <-t:
+			return false
+		}
+	}
+}
+
 // Name returns the name of the adapter, e.g. "tun0". On Windows, this may
 // return a canonical adapter name instead.
 func (tun *TunAdapter) Name() string {
@@ -145,6 +162,8 @@ func (tun *TunAdapter) _start() error {
 	tun.rwc.SetMTU(tun.MTU())
 	tun.isOpen = true
 	tun.isEnabled = true
+	tun.ch = make(chan []byte, tun.iface.BatchSize())
+	go tun.queue()
 	go tun.read()
 	go tun.write()
 	return nil
@@ -178,3 +197,12 @@ func (tun *TunAdapter) _stop() error {
 	}
 	return nil
 }
+
+const bufPoolSize = TUN_OFFSET_BYTES + 65535
+
+var bufPool = sync.Pool{
+	New: func() any {
+		b := [bufPoolSize]byte{}
+		return b[:]
+	},
+}

src/tun/tun_bsd.go

@@ -80,6 +80,9 @@ func (tun *TunAdapter) setup(ifname string, addr string, mtu uint64) error {
 	if err != nil {
 		return fmt.Errorf("failed to create TUN: %w", err)
 	}
+	if !waitForTUNUp(iface.Events()) {
+		return fmt.Errorf("TUN did not come up in time")
+	}
 	tun.iface = iface
 	if mtu, err := iface.MTU(); err == nil {
 		tun.mtu = getSupportedMTU(uint64(mtu))

src/tun/tun_darwin.go

@@ -27,6 +27,9 @@ func (tun *TunAdapter) setup(ifname string, addr string, mtu uint64) error {
 	if err != nil {
 		return fmt.Errorf("failed to create TUN: %w", err)
 	}
+	if !waitForTUNUp(iface.Events()) {
+		return fmt.Errorf("TUN did not come up in time")
+	}
 	tun.iface = iface
 	if m, err := iface.MTU(); err == nil {
 		tun.mtu = getSupportedMTU(uint64(m))
@@ -55,6 +58,9 @@ func (tun *TunAdapter) setupFD(fd int32, addr string, mtu uint64) error {
 		unix.Close(dfd)
 		return fmt.Errorf("failed to create TUN from FD: %w", err)
 	}
+	if !waitForTUNUp(iface.Events()) {
+		return fmt.Errorf("TUN did not come up in time")
+	}
 	tun.iface = iface
 	if m, err := iface.MTU(); err == nil {
 		tun.mtu = getSupportedMTU(uint64(m))

src/tun/tun_linux.go

@@ -21,6 +21,9 @@ func (tun *TunAdapter) setup(ifname string, addr string, mtu uint64) error {
 	if err != nil {
 		return fmt.Errorf("failed to create TUN: %w", err)
 	}
+	if !waitForTUNUp(iface.Events()) {
+		return fmt.Errorf("TUN did not come up in time")
+	}
 	tun.iface = iface
 	if mtu, err := iface.MTU(); err == nil {
 		tun.mtu = getSupportedMTU(uint64(mtu))

src/tun/tun_other.go

@@ -18,6 +18,9 @@ func (tun *TunAdapter) setup(ifname string, addr string, mtu uint64) error {
 	if err != nil {
 		return fmt.Errorf("failed to create TUN: %w", err)
 	}
+	if !waitForTUNUp(iface.Events()) {
+		return fmt.Errorf("TUN did not come up in time")
+	}
 	tun.iface = iface
 	if mtu, err := iface.MTU(); err == nil {
 		tun.mtu = getSupportedMTU(uint64(mtu))

src/tun/tun_windows.go

@@ -34,6 +34,9 @@ func (tun *TunAdapter) setup(ifname string, addr string, mtu uint64) error {
 		if iface, err = wgtun.CreateTUNWithRequestedGUID(ifname, &guid, int(mtu)); err != nil {
 			return err
 		}
+		if !waitForTUNUp(iface.Events()) {
+			return fmt.Errorf("TUN did not come up in time")
+		}
 		tun.iface = iface
 		if addr != "" {
 			if err = tun.setupAddress(addr); err != nil {