tweak defaults and turn on gsogro on linux by default

This reverts commit fa098c551a.

CHANGELOG.md

@@ -7,30 +7,12 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
-### Added
-
-- Experimental Linux UDP offload support: enable `listen.enable_gso` and
-  `listen.enable_gro` to activate UDP_SEGMENT batching and GRO receive
-  splitting. Includes automatic capability probing, per-packet fallbacks, and
-  runtime metrics/logs for visibility.
-- Optional Linux TUN `virtio_net_hdr` support: set `tun.enable_vnet_hdr` to
-  have Nebula negotiate VNET headers and offload flags so future batches can
-  be delivered to the kernel with metadata instead of per-packet writes.
-- Linux UDP send sharding can now be tuned with `listen.send_shards`; defaults
-  to `GOMAXPROCS` but can be increased to stripe heavy peers across more
-  goroutines.
-
 ### Changed
 
 - `default_local_cidr_any` now defaults to false, meaning that any firewall rule
   intended to target an `unsafe_routes` entry must explicitly declare it via the
   `local_cidr` field. This is almost always the intended behavior. This flag is
   deprecated and will be removed in a future release.
-- UDP receive path now enqueues into per-worker lock-free rings, restoring the
-  `listen.decrypt_workers`/`listen.decrypt_queue_depth` tuning knobs while
-  eliminating the mutex contention from the old shared channel.
-- Increased replay protection window to 32k packets so high-throughput links
-  tolerate larger bursts of reordering without tripping the anti-replay logic.
 
 ## [1.9.4] - 2024-09-09
 

bits.go

@@ -5,6 +5,7 @@ import (
 	"github.com/sirupsen/logrus"
 )
 
+// TODO: Pretty sure this is just all sorts of racy now, we need it to be atomic
 type Bits struct {
 	length             uint64
 	current            uint64
@@ -43,7 +44,7 @@ func (b *Bits) Check(l logrus.FieldLogger, i uint64) bool {
 	}
 
 	// Not within the window
-	l.Debugf("rejected a packet (top) %d %d\n", b.current, i)
+	l.Error("rejected a packet (top) %d %d\n", b.current, i)
 	return false
 }
 

cert/pem.go

@@ -1,8 +1,10 @@
 package cert
 
 import (
+	"encoding/hex"
 	"encoding/pem"
 	"fmt"
+	"time"
 
 	"golang.org/x/crypto/ed25519"
 )
@@ -138,6 +140,101 @@ func MarshalSigningPrivateKeyToPEM(curve Curve, b []byte) []byte {
 	}
 }
 
+// Backward compatibility functions for older API
+func MarshalX25519PublicKey(b []byte) []byte {
+	return MarshalPublicKeyToPEM(Curve_CURVE25519, b)
+}
+
+func MarshalX25519PrivateKey(b []byte) []byte {
+	return MarshalPrivateKeyToPEM(Curve_CURVE25519, b)
+}
+
+func MarshalPublicKey(curve Curve, b []byte) []byte {
+	return MarshalPublicKeyToPEM(curve, b)
+}
+
+func MarshalPrivateKey(curve Curve, b []byte) []byte {
+	return MarshalPrivateKeyToPEM(curve, b)
+}
+
+// NebulaCertificate is a compatibility wrapper for the old API
+type NebulaCertificate struct {
+	Details   NebulaCertificateDetails
+	Signature []byte
+	cert      Certificate
+}
+
+// NebulaCertificateDetails is a compatibility wrapper for certificate details
+type NebulaCertificateDetails struct {
+	Name      string
+	NotBefore time.Time
+	NotAfter  time.Time
+	PublicKey []byte
+	IsCA      bool
+	Issuer    []byte
+	Curve     Curve
+}
+
+// UnmarshalNebulaCertificateFromPEM provides backward compatibility with the old API
+func UnmarshalNebulaCertificateFromPEM(b []byte) (*NebulaCertificate, []byte, error) {
+	c, rest, err := UnmarshalCertificateFromPEM(b)
+	if err != nil {
+		return nil, rest, err
+	}
+
+	issuerBytes, err := func() ([]byte, error) {
+		issuer := c.Issuer()
+		if issuer == "" {
+			return nil, nil
+		}
+		decoded, err := hex.DecodeString(issuer)
+		if err != nil {
+			return nil, fmt.Errorf("failed to decode issuer fingerprint: %w", err)
+		}
+		return decoded, nil
+	}()
+	if err != nil {
+		return nil, rest, err
+	}
+
+	pubKey := c.PublicKey()
+	if pubKey != nil {
+		pubKey = append([]byte(nil), pubKey...)
+	}
+
+	sig := c.Signature()
+	if sig != nil {
+		sig = append([]byte(nil), sig...)
+	}
+
+	return &NebulaCertificate{
+		Details: NebulaCertificateDetails{
+			Name:      c.Name(),
+			NotBefore: c.NotBefore(),
+			NotAfter:  c.NotAfter(),
+			PublicKey: pubKey,
+			IsCA:      c.IsCA(),
+			Issuer:    issuerBytes,
+			Curve:     c.Curve(),
+		},
+		Signature: sig,
+		cert:      c,
+	}, rest, nil
+}
+
+// IssuerString returns the issuer in hex format for compatibility
+func (n *NebulaCertificate) IssuerString() string {
+	if n.Details.Issuer == nil {
+		return ""
+	}
+	return hex.EncodeToString(n.Details.Issuer)
+}
+
+// Certificate returns the underlying certificate (read-only)
+func (n *NebulaCertificate) Certificate() Certificate {
+	return n.cert
+}
+
 // UnmarshalPrivateKeyFromPEM will try to unmarshal the first pem block in a byte array, returning any non
 // consumed data or an error on failure
 func UnmarshalPrivateKeyFromPEM(b []byte) ([]byte, []byte, Curve, error) {

cmd/nebula-service/main.go

@@ -65,8 +65,16 @@ func main() {
 	}
 
 	if !*configTest {
-		ctrl.Start()
+		wait, err := ctrl.Start()
-		ctrl.ShutdownBlock()
+		if err != nil {
+			util.LogWithContextIfNeeded("Error while running", err, l)
+			os.Exit(1)
+		}
+
+		go ctrl.ShutdownBlock()
+		wait()
+
+		l.Info("Goodbye")
 	}
 
 	os.Exit(0)

cmd/nebula/main.go

@@ -3,6 +3,9 @@ package main
 import (
 	"flag"
 	"fmt"
+	"log"
+	"net/http"
+	_ "net/http/pprof"
 	"os"
 
 	"github.com/sirupsen/logrus"
@@ -58,10 +61,22 @@ func main() {
 		os.Exit(1)
 	}
 
+	go func() {
+		log.Println(http.ListenAndServe("0.0.0.0:6060", nil))
+	}()
+
 	if !*configTest {
-		ctrl.Start()
+		wait, err := ctrl.Start()
+		if err != nil {
+			util.LogWithContextIfNeeded("Error while running", err, l)
+			os.Exit(1)
+		}
+
+		go ctrl.ShutdownBlock()
 		notifyReady(l)
-		ctrl.ShutdownBlock()
+		wait()
+
+		l.Info("Goodbye")
 	}
 
 	os.Exit(0)

connection_state.go

@@ -13,10 +13,9 @@ import (
 	"github.com/slackhq/nebula/noiseutil"
 )
 
-// ReplayWindow controls the size of the sliding window used to detect replays.
+// TODO: In a 5Gbps test, 1024 is not sufficient. With a 1400 MTU this is about 1.4Gbps of window, assuming full packets.
-// High-bandwidth links with GRO/GSO can reorder more than a thousand packets in
+// 4092 should be sufficient for 5Gbps
-// flight, so keep this comfortably above the largest expected burst.
+const ReplayWindow = 4096
-const ReplayWindow = 32768
 
 type ConnectionState struct {
 	eKey           *NebulaCipherState

control.go

@@ -2,9 +2,11 @@ package nebula
 
 import (
 	"context"
+	"errors"
 	"net/netip"
 	"os"
 	"os/signal"
+	"sync"
 	"syscall"
 
 	"github.com/sirupsen/logrus"
@@ -13,6 +15,16 @@ import (
 	"github.com/slackhq/nebula/overlay"
 )
 
+type RunState int
+
+const (
+	Stopped  RunState = 0 // The control has yet to be started
+	Started  RunState = 1 // The control has been started
+	Stopping RunState = 2 // The control is stopping
+)
+
+var ErrAlreadyStarted = errors.New("nebula is already started")
+
 // Every interaction here needs to take extra care to copy memory and not return or use arguments "as is" when touching
 // core. This means copying IP objects, slices, de-referencing pointers and taking the actual value, etc
 
@@ -26,6 +38,9 @@ type controlHostLister interface {
 }
 
 type Control struct {
+	stateLock sync.Mutex
+	state     RunState
+
 	f                      *Interface
 	l                      *logrus.Logger
 	ctx                    context.Context
@@ -49,10 +64,21 @@ type ControlHostInfo struct {
 	CurrentRelaysThroughMe []netip.Addr     `json:"currentRelaysThroughMe"`
 }
 
-// Start actually runs nebula, this is a nonblocking call. To block use Control.ShutdownBlock()
+// Start actually runs nebula, this is a nonblocking call.
-func (c *Control) Start() {
+// The returned function can be used to wait for nebula to fully stop.
+func (c *Control) Start() (func(), error) {
+	c.stateLock.Lock()
+	if c.state != Stopped {
+		c.stateLock.Unlock()
+		return nil, ErrAlreadyStarted
+	}
+
 	// Activate the interface
-	c.f.activate()
+	err := c.f.activate()
+	if err != nil {
+		c.stateLock.Unlock()
+		return nil, err
+	}
 
 	// Call all the delayed funcs that waited patiently for the interface to be created.
 	if c.sshStart != nil {
@@ -72,15 +98,33 @@ func (c *Control) Start() {
 	}
 
 	// Start reading packets.
-	c.f.run()
+	c.state = Started
+	c.stateLock.Unlock()
+	return c.f.run(c.ctx)
+}
+
+func (c *Control) State() RunState {
+	c.stateLock.Lock()
+	defer c.stateLock.Unlock()
+	return c.state
 }
 
 func (c *Control) Context() context.Context {
 	return c.ctx
 }
 
-// Stop signals nebula to shutdown and close all tunnels, returns after the shutdown is complete
+// Stop is a non-blocking call that signals nebula to close all tunnels and shut down
 func (c *Control) Stop() {
+	c.stateLock.Lock()
+	if c.state != Started {
+		c.stateLock.Unlock()
+		// We are stopping or stopped already
+		return
+	}
+
+	c.state = Stopping
+	c.stateLock.Unlock()
+
 	// Stop the handshakeManager (and other services), to prevent new tunnels from
 	// being created while we're shutting them all down.
 	c.cancel()
@@ -89,7 +133,7 @@ func (c *Control) Stop() {
 	if err := c.f.Close(); err != nil {
 		c.l.WithError(err).Error("Close interface failed")
 	}
-	c.l.Info("Goodbye")
+	c.state = Stopped
 }
 
 // ShutdownBlock will listen for and block on term and interrupt signals, calling Control.Stop() once signalled

examples/config.yml

@@ -132,6 +132,13 @@ listen:
   # Sets the max number of packets to pull from the kernel for each syscall (under systems that support recvmmsg)
   # default is 64, does not support reload
   #batch: 64
+
+# Control batching between UDP and TUN pipelines
+#batch:
+#  inbound_size: 32        # packets to queue from UDP before handing to workers
+#  outbound_size: 32       # packets to queue from TUN before handing to workers
+#  flush_interval: 50us    # flush partially filled batches after this duration
+#  max_outstanding: 1028   # batches buffered per routine on each channel
   # Configure socket buffers for the udp side (outside), leave unset to use the system defaults. Values will be doubled by the kernel
   # Default is net.core.rmem_default and net.core.wmem_default (/proc/sys/net/core/rmem_default and /proc/sys/net/core/rmem_default)
   # Maximum is limited by memory in the system, SO_RCVBUFFORCE and SO_SNDBUFFORCE is used to avoid having to raise the system wide

firewall_test.go

@@ -392,7 +392,7 @@ func BenchmarkFirewallTable_match(b *testing.B) {
 		c := &cert.CachedCertificate{
 			Certificate: &dummyCert{
 				name:     "nope",
-				networks: []netip.Prefix{netip.MustParsePrefix("fd99::99/128")},
+				networks: []netip.Prefix{netip.MustParsePrefix("fd99:99/128")},
 			},
 			InvertedGroups: map[string]struct{}{"nope": {}},
 		}
@@ -692,50 +692,6 @@ func TestFirewall_DropConntrackReload(t *testing.T) {
 	assert.Equal(t, fw.Drop(p, false, &h, cp, nil), ErrNoMatchingRule)
 }
 
-func TestFirewall_DropIPSpoofing(t *testing.T) {
-	l := test.NewLogger()
-	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
-
-	c := cert.CachedCertificate{
-		Certificate: &dummyCert{
-			name:     "host-owner",
-			networks: []netip.Prefix{netip.MustParsePrefix("192.0.2.1/24")},
-		},
-	}
-
-	c1 := cert.CachedCertificate{
-		Certificate: &dummyCert{
-			name:           "host",
-			networks:       []netip.Prefix{netip.MustParsePrefix("192.0.2.2/24")},
-			unsafeNetworks: []netip.Prefix{netip.MustParsePrefix("198.51.100.0/24")},
-		},
-	}
-	h1 := HostInfo{
-		ConnectionState: &ConnectionState{
-			peerCert: &c1,
-		},
-		vpnAddrs: []netip.Addr{c1.Certificate.Networks()[0].Addr()},
-	}
-	h1.buildNetworks(c1.Certificate.Networks(), c1.Certificate.UnsafeNetworks())
-
-	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
-
-	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 1, 1, []string{}, "", netip.Prefix{}, netip.Prefix{}, "", ""))
-	cp := cert.NewCAPool()
-
-	// Packet spoofed by `c1`. Note that the remote addr is not a valid one.
-	p := firewall.Packet{
-		LocalAddr:  netip.MustParseAddr("192.0.2.1"),
-		RemoteAddr: netip.MustParseAddr("192.0.2.3"),
-		LocalPort:  1,
-		RemotePort: 1,
-		Protocol:   firewall.ProtoUDP,
-		Fragment:   false,
-	}
-	assert.Equal(t, fw.Drop(p, true, &h1, cp, nil), ErrInvalidRemoteIP)
-}
-
 func BenchmarkLookup(b *testing.B) {
 	ml := func(m map[string]struct{}, a [][]string) {
 		for n := 0; n < b.N; n++ {

go.mod

@@ -8,30 +8,30 @@ require (
 	github.com/armon/go-radix v1.0.0
 	github.com/cyberdelia/go-metrics-graphite v0.0.0-20161219230853-39f87cc3b432
 	github.com/flynn/noise v1.1.0
-	github.com/gaissmai/bart v0.25.0
+	github.com/gaissmai/bart v0.20.4
 	github.com/gogo/protobuf v1.3.2
 	github.com/google/gopacket v1.1.19
-	github.com/kardianos/service v1.2.4
+	github.com/kardianos/service v1.2.2
-	github.com/miekg/dns v1.1.68
+	github.com/miekg/dns v1.1.65
 	github.com/miekg/pkcs11 v1.1.2-0.20231115102856-9078ad6b9d4b
 	github.com/nbrownus/go-metrics-prometheus v0.0.0-20210712211119-974a6260965f
-	github.com/prometheus/client_golang v1.23.2
+	github.com/prometheus/client_golang v1.22.0
 	github.com/rcrowley/go-metrics v0.0.0-20201227073835-cf1acfcdf475
 	github.com/sirupsen/logrus v1.9.3
 	github.com/skip2/go-qrcode v0.0.0-20200617195104-da1b6568686e
 	github.com/stefanberger/go-pkcs11uri v0.0.0-20230803200340-78284954bff6
-	github.com/stretchr/testify v1.11.1
+	github.com/stretchr/testify v1.10.0
 	github.com/vishvananda/netlink v1.3.1
-	golang.org/x/crypto v0.43.0
+	golang.org/x/crypto v0.37.0
 	golang.org/x/exp v0.0.0-20230725093048-515e97ebf090
-	golang.org/x/net v0.45.0
+	golang.org/x/net v0.39.0
-	golang.org/x/sync v0.17.0
+	golang.org/x/sync v0.13.0
-	golang.org/x/sys v0.37.0
+	golang.org/x/sys v0.32.0
-	golang.org/x/term v0.36.0
+	golang.org/x/term v0.31.0
 	golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2
 	golang.zx2c4.com/wireguard v0.0.0-20230325221338-052af4a8072b
 	golang.zx2c4.com/wireguard/windows v0.5.3
-	google.golang.org/protobuf v1.36.8
+	google.golang.org/protobuf v1.36.6
 	gopkg.in/yaml.v3 v3.0.1
 	gvisor.dev/gvisor v0.0.0-20240423190808-9d7a357edefe
 )
@@ -43,12 +43,11 @@ require (
 	github.com/google/btree v1.1.2 // indirect
 	github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 // indirect
 	github.com/pmezard/go-difflib v1.0.0 // indirect
-	github.com/prometheus/client_model v0.6.2 // indirect
+	github.com/prometheus/client_model v0.6.1 // indirect
-	github.com/prometheus/common v0.66.1 // indirect
+	github.com/prometheus/common v0.62.0 // indirect
-	github.com/prometheus/procfs v0.16.1 // indirect
+	github.com/prometheus/procfs v0.15.1 // indirect
 	github.com/vishvananda/netns v0.0.5 // indirect
-	go.yaml.in/yaml/v2 v2.4.2 // indirect
+	golang.org/x/mod v0.23.0 // indirect
-	golang.org/x/mod v0.24.0 // indirect
 	golang.org/x/time v0.5.0 // indirect
-	golang.org/x/tools v0.33.0 // indirect
+	golang.org/x/tools v0.30.0 // indirect
 )

go.sum

@@ -24,8 +24,8 @@ github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/flynn/noise v1.1.0 h1:KjPQoQCEFdZDiP03phOvGi11+SVVhBG2wOWAorLsstg=
 github.com/flynn/noise v1.1.0/go.mod h1:xbMo+0i6+IGbYdJhF31t2eR1BIU0CYc12+BNAKwUTag=
-github.com/gaissmai/bart v0.25.0 h1:eqiokVPqM3F94vJ0bTHXHtH91S8zkKL+bKh+BsGOsJM=
+github.com/gaissmai/bart v0.20.4 h1:Ik47r1fy3jRVU+1eYzKSW3ho2UgBVTVnUS8O993584U=
-github.com/gaissmai/bart v0.25.0/go.mod h1:GREWQfTLRWz/c5FTOsIw+KkscuFkIV5t8Rp7Nd1Td5c=
+github.com/gaissmai/bart v0.20.4/go.mod h1:cEed+ge8dalcbpi8wtS9x9m2hn/fNJH5suhdGQOHnYk=
 github.com/go-kit/kit v0.8.0/go.mod h1:xBxKIO96dXMWWy0MnWVtmwkA9/13aqxPnvrjFYMA2as=
 github.com/go-kit/kit v0.9.0/go.mod h1:xBxKIO96dXMWWy0MnWVtmwkA9/13aqxPnvrjFYMA2as=
 github.com/go-kit/log v0.1.0/go.mod h1:zbhenjAZHb184qTLMA9ZjW7ThYL0H2mk7Q6pNt4vbaY=
@@ -64,8 +64,8 @@ github.com/json-iterator/go v1.1.10/go.mod h1:KdQUCv79m/52Kvf8AW2vK1V8akMuk1QjK/
 github.com/json-iterator/go v1.1.11/go.mod h1:KdQUCv79m/52Kvf8AW2vK1V8akMuk1QjK/uOdHXbAo4=
 github.com/julienschmidt/httprouter v1.2.0/go.mod h1:SYymIcj16QtmaHHD7aYtjjsJG7VTCxuUUipMqKk8s4w=
 github.com/julienschmidt/httprouter v1.3.0/go.mod h1:JR6WtHb+2LUe8TCKY3cZOxFyyO8IZAc4RVcycCCAKdM=
-github.com/kardianos/service v1.2.4 h1:XNlGtZOYNx2u91urOdg/Kfmc+gfmuIo1Dd3rEi2OgBk=
+github.com/kardianos/service v1.2.2 h1:ZvePhAHfvo0A7Mftk/tEzqEZ7Q4lgnR8sGz4xu1YX60=
-github.com/kardianos/service v1.2.4/go.mod h1:E4V9ufUuY82F7Ztlu1eN9VXWIQxg8NoLQlmFe0MtrXc=
+github.com/kardianos/service v1.2.2/go.mod h1:CIMRFEJVL+0DS1a3Nx06NaMn4Dz63Ng6O7dl0qH0zVM=
 github.com/kisielk/errcheck v1.5.0/go.mod h1:pFxgyoBC7bSaBwPgfKdkLd5X25qrDl4LWUI2bnpBCr8=
 github.com/kisielk/gotool v1.0.0/go.mod h1:XhKaO+MFFWcvkIS/tQcRk01m1F5IRFswLeQ+oQHNcck=
 github.com/klauspost/compress v1.18.0 h1:c/Cqfb0r+Yi+JtIEq73FWXVkRonBlf0CRNYc8Zttxdo=
@@ -83,8 +83,8 @@ github.com/kr/text v0.1.0/go.mod h1:4Jbv+DJW3UT/LiOwJeYQe1efqtUx/iVham/4vfdArNI=
 github.com/kylelemons/godebug v1.1.0 h1:RPNrshWIDI6G2gRW9EHilWtl7Z6Sb1BR0xunSBf0SNc=
 github.com/kylelemons/godebug v1.1.0/go.mod h1:9/0rRGxNHcop5bhtWyNeEfOS8JIWk580+fNqagV/RAw=
 github.com/matttproud/golang_protobuf_extensions v1.0.1/go.mod h1:D8He9yQNgCq6Z5Ld7szi9bcBfOoFv/3dc6xSMkL2PC0=
-github.com/miekg/dns v1.1.68 h1:jsSRkNozw7G/mnmXULynzMNIsgY2dHC8LO6U6Ij2JEA=
+github.com/miekg/dns v1.1.65 h1:0+tIPHzUW0GCge7IiK3guGP57VAw7hoPDfApjkMD1Fc=
-github.com/miekg/dns v1.1.68/go.mod h1:fujopn7TB3Pu3JM69XaawiU0wqjpL9/8xGop5UrTPps=
+github.com/miekg/dns v1.1.65/go.mod h1:Dzw9769uoKVaLuODMDZz9M6ynFU6Em65csPuoi8G0ck=
 github.com/miekg/pkcs11 v1.1.2-0.20231115102856-9078ad6b9d4b h1:J/AzCvg5z0Hn1rqZUJjpbzALUmkKX0Zwbc/i4fw7Sfk=
 github.com/miekg/pkcs11 v1.1.2-0.20231115102856-9078ad6b9d4b/go.mod h1:XsNlhZGX73bx86s2hdc/FuaLm2CPZJemRLMA+WTFxgs=
 github.com/modern-go/concurrent v0.0.0-20180228061459-e0a39a4cb421/go.mod h1:6dJC0mAP4ikYIbvyc7fijjWJddQyLn8Ig3JB5CqoB9Q=
@@ -106,24 +106,24 @@ github.com/prometheus/client_golang v0.9.1/go.mod h1:7SWBe2y4D6OKWSNQJUaRYU/AaXP
 github.com/prometheus/client_golang v1.0.0/go.mod h1:db9x61etRT2tGnBNRi70OPL5FsnadC4Ky3P0J6CfImo=
 github.com/prometheus/client_golang v1.7.1/go.mod h1:PY5Wy2awLA44sXw4AOSfFBetzPP4j5+D6mVACh+pe2M=
 github.com/prometheus/client_golang v1.11.0/go.mod h1:Z6t4BnS23TR94PD6BsDNk8yVqroYurpAkEiz0P2BEV0=
-github.com/prometheus/client_golang v1.23.2 h1:Je96obch5RDVy3FDMndoUsjAhG5Edi49h0RJWRi/o0o=
+github.com/prometheus/client_golang v1.22.0 h1:rb93p9lokFEsctTys46VnV1kLCDpVZ0a/Y92Vm0Zc6Q=
-github.com/prometheus/client_golang v1.23.2/go.mod h1:Tb1a6LWHB3/SPIzCoaDXI4I8UHKeFTEQ1YCr+0Gyqmg=
+github.com/prometheus/client_golang v1.22.0/go.mod h1:R7ljNsLXhuQXYZYtw6GAE9AZg8Y7vEW5scdCXrWRXC0=
 github.com/prometheus/client_model v0.0.0-20180712105110-5c3871d89910/go.mod h1:MbSGuTsp3dbXC40dX6PRTWyKYBIrTGTE9sqQNg2J8bo=
 github.com/prometheus/client_model v0.0.0-20190129233127-fd36f4220a90/go.mod h1:xMI15A0UPsDsEKsMN9yxemIoYk6Tm2C1GtYGdfGttqA=
 github.com/prometheus/client_model v0.2.0/go.mod h1:xMI15A0UPsDsEKsMN9yxemIoYk6Tm2C1GtYGdfGttqA=
-github.com/prometheus/client_model v0.6.2 h1:oBsgwpGs7iVziMvrGhE53c/GrLUsZdHnqNwqPLxwZyk=
+github.com/prometheus/client_model v0.6.1 h1:ZKSh/rekM+n3CeS952MLRAdFwIKqeY8b62p8ais2e9E=
-github.com/prometheus/client_model v0.6.2/go.mod h1:y3m2F6Gdpfy6Ut/GBsUqTWZqCUvMVzSfMLjcu6wAwpE=
+github.com/prometheus/client_model v0.6.1/go.mod h1:OrxVMOVHjw3lKMa8+x6HeMGkHMQyHDk9E3jmP2AmGiY=
 github.com/prometheus/common v0.4.1/go.mod h1:TNfzLD0ON7rHzMJeJkieUDPYmFC7Snx/y86RQel1bk4=
 github.com/prometheus/common v0.10.0/go.mod h1:Tlit/dnDKsSWFlCLTWaA1cyBgKHSMdTB80sz/V91rCo=
 github.com/prometheus/common v0.26.0/go.mod h1:M7rCNAaPfAosfx8veZJCuw84e35h3Cfd9VFqTh1DIvc=
-github.com/prometheus/common v0.66.1 h1:h5E0h5/Y8niHc5DlaLlWLArTQI7tMrsfQjHV+d9ZoGs=
+github.com/prometheus/common v0.62.0 h1:xasJaQlnWAeyHdUBeGjXmutelfJHWMRr+Fg4QszZ2Io=
-github.com/prometheus/common v0.66.1/go.mod h1:gcaUsgf3KfRSwHY4dIMXLPV0K/Wg1oZ8+SbZk/HH/dA=
+github.com/prometheus/common v0.62.0/go.mod h1:vyBcEuLSvWos9B1+CyL7JZ2up+uFzXhkqml0W5zIY1I=
 github.com/prometheus/procfs v0.0.0-20181005140218-185b4288413d/go.mod h1:c3At6R/oaqEKCNdg8wHV1ftS6bRYblBhIjjI8uT2IGk=
 github.com/prometheus/procfs v0.0.2/go.mod h1:TjEm7ze935MbeOT/UhFTIMYKhuLP4wbCsTZCD3I8kEA=
 github.com/prometheus/procfs v0.1.3/go.mod h1:lV6e/gmhEcM9IjHGsFOCxxuZ+z1YqCvr4OA4YeYWdaU=
 github.com/prometheus/procfs v0.6.0/go.mod h1:cz+aTbrPOrUb4q7XlbU9ygM+/jj0fzG6c1xBZuNvfVA=
-github.com/prometheus/procfs v0.16.1 h1:hZ15bTNuirocR6u0JZ6BAHHmwS1p8B4P6MRqxtzMyRg=
+github.com/prometheus/procfs v0.15.1 h1:YagwOFzUgYfKKHX6Dr+sHT7km/hxC76UB0learggepc=
-github.com/prometheus/procfs v0.16.1/go.mod h1:teAbpZRB1iIAJYREa1LsoWUXykVXA1KlTmWl8x/U+Is=
+github.com/prometheus/procfs v0.15.1/go.mod h1:fB45yRUv8NstnjriLhBQLuOUt+WW4BsoGhij/e3PBqk=
 github.com/rcrowley/go-metrics v0.0.0-20201227073835-cf1acfcdf475 h1:N/ElC8H3+5XpJzTSTfLsJV/mx9Q9g7kxmchpfZyxgzM=
 github.com/rcrowley/go-metrics v0.0.0-20201227073835-cf1acfcdf475/go.mod h1:bCqnVzQkZxMG4s8nGwiZ5l3QUCyqpo9Y+/ZMZ9VjZe4=
 github.com/rogpeppe/go-internal v1.10.0 h1:TMyTOH3F/DB16zRVcYyreMH6GnZZrwQVAoYjRBZyWFQ=
@@ -143,33 +143,29 @@ github.com/stretchr/testify v1.2.2/go.mod h1:a8OnRcib4nhh0OaRAV+Yts87kKdq0PP7pXf
 github.com/stretchr/testify v1.3.0/go.mod h1:M5WIy9Dh21IEIfnGCwXGc5bZfKNJtfHm1UVUgZn+9EI=
 github.com/stretchr/testify v1.4.0/go.mod h1:j7eGeouHqKxXV5pUuKE4zz7dFj8WfuZ+81PSLYec5m4=
 github.com/stretchr/testify v1.7.0/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
-github.com/stretchr/testify v1.11.1 h1:7s2iGBzp5EwR7/aIZr8ao5+dra3wiQyKjjFuvgVKu7U=
+github.com/stretchr/testify v1.10.0 h1:Xv5erBjTwe/5IxqUQTdXv5kgmIvbHo3QQyRwhJsOfJA=
-github.com/stretchr/testify v1.11.1/go.mod h1:wZwfW3scLgRK+23gO65QZefKpKQRnfz6sD981Nm4B6U=
+github.com/stretchr/testify v1.10.0/go.mod h1:r2ic/lqez/lEtzL7wO/rwa5dbSLXVDPFyf8C91i36aY=
 github.com/vishvananda/netlink v1.3.1 h1:3AEMt62VKqz90r0tmNhog0r/PpWKmrEShJU0wJW6bV0=
 github.com/vishvananda/netlink v1.3.1/go.mod h1:ARtKouGSTGchR8aMwmkzC0qiNPrrWO5JS/XMVl45+b4=
 github.com/vishvananda/netns v0.0.5 h1:DfiHV+j8bA32MFM7bfEunvT8IAqQ/NzSJHtcmW5zdEY=
 github.com/vishvananda/netns v0.0.5/go.mod h1:SpkAiCQRtJ6TvvxPnOSyH3BMl6unz3xZlaprSwhNNJM=
 github.com/yuin/goldmark v1.1.27/go.mod h1:3hX8gzYuyVAZsxl0MRgGTJEmQBFcNTphYh9decYSb74=
 github.com/yuin/goldmark v1.2.1/go.mod h1:3hX8gzYuyVAZsxl0MRgGTJEmQBFcNTphYh9decYSb74=
-go.uber.org/goleak v1.3.0 h1:2K3zAYmnTNqV73imy9J1T3WC+gmCePx2hEGkimedGto=
-go.uber.org/goleak v1.3.0/go.mod h1:CoHD4mav9JJNrW/WLlf7HGZPjdw8EucARQHekz1X6bE=
-go.yaml.in/yaml/v2 v2.4.2 h1:DzmwEr2rDGHl7lsFgAHxmNz/1NlQ7xLIrlN2h5d1eGI=
-go.yaml.in/yaml/v2 v2.4.2/go.mod h1:081UH+NErpNdqlCXm3TtEran0rJZGxAYx9hb/ELlsPU=
 golang.org/x/crypto v0.0.0-20180904163835-0709b304e793/go.mod h1:6SG95UA2DQfeDnfUPMdvaQW0Q7yPrPDi9nlGo2tz2b4=
 golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
 golang.org/x/crypto v0.0.0-20191011191535-87dc89f01550/go.mod h1:yigFU9vqHzYiE8UmvKecakEJjdnWj3jj499lnFckfCI=
 golang.org/x/crypto v0.0.0-20200622213623-75b288015ac9/go.mod h1:LzIPMQfyMNhhGPhUkYOs5KpL4U8rLKemX1yGLhDgUto=
 golang.org/x/crypto v0.0.0-20210322153248-0c34fe9e7dc2/go.mod h1:T9bdIzuCu7OtxOm1hfPfRQxPLYneinmdGuTeoZ9dtd4=
-golang.org/x/crypto v0.43.0 h1:dduJYIi3A3KOfdGOHX8AVZ/jGiyPa3IbBozJ5kNuE04=
+golang.org/x/crypto v0.37.0 h1:kJNSjF/Xp7kU0iB2Z+9viTPMW4EqqsrywMXLJOOsXSE=
-golang.org/x/crypto v0.43.0/go.mod h1:BFbav4mRNlXJL4wNeejLpWxB7wMbc79PdRGhWKncxR0=
+golang.org/x/crypto v0.37.0/go.mod h1:vg+k43peMZ0pUMhYmVAWysMK35e6ioLh3wB8ZCAfbVc=
 golang.org/x/exp v0.0.0-20230725093048-515e97ebf090 h1:Di6/M8l0O2lCLc6VVRWhgCiApHV8MnQurBnFSHsQtNY=
 golang.org/x/exp v0.0.0-20230725093048-515e97ebf090/go.mod h1:FXUEEKJgO7OQYeo8N01OfiKP8RXMtf6e8aTskBGqWdc=
 golang.org/x/lint v0.0.0-20200302205851-738671d3881b/go.mod h1:3xt1FjdF8hUf6vQPIChWIBhFzV8gjjsPE/fR3IyQdNY=
 golang.org/x/mod v0.1.1-0.20191105210325-c90efee705ee/go.mod h1:QqPTAvyqsEbceGzBzNggFXnrqF1CaUcvgkdR5Ot7KZg=
 golang.org/x/mod v0.2.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
 golang.org/x/mod v0.3.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
-golang.org/x/mod v0.24.0 h1:ZfthKaKaT4NrhGVZHO1/WDTwGES4De8KtWO0SIbNJMU=
+golang.org/x/mod v0.23.0 h1:Zb7khfcRGKk+kqfxFaP5tZqCnDZMjC5VtUBs87Hr6QM=
-golang.org/x/mod v0.24.0/go.mod h1:IXM97Txy2VM4PJ3gI61r1YEk/gAj6zAHN3AdZt6S9Ww=
+golang.org/x/mod v0.23.0/go.mod h1:6SkKJ3Xj0I0BrPOZoBy3bdMptDDU9oJrpohJ3eWZ1fY=
 golang.org/x/net v0.0.0-20180724234803-3673e40ba225/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
 golang.org/x/net v0.0.0-20181114220301-adae6a3d119a/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
 golang.org/x/net v0.0.0-20190108225652-1e06a53dbb7e/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
@@ -180,8 +176,8 @@ golang.org/x/net v0.0.0-20200226121028-0de0cce0169b/go.mod h1:z5CRVTTTmAJ677TzLL
 golang.org/x/net v0.0.0-20200625001655-4c5254603344/go.mod h1:/O7V0waA8r7cgGh81Ro3o1hOxt32SMVPicZroKQ2sZA=
 golang.org/x/net v0.0.0-20201021035429-f5854403a974/go.mod h1:sp8m0HH+o8qH0wwXwYZr8TS3Oi6o0r6Gce1SSxlDquU=
 golang.org/x/net v0.0.0-20210226172049-e18ecbb05110/go.mod h1:m0MpNAwzfU5UDzcl9v0D8zg8gWTRqZa9RBIspLL5mdg=
-golang.org/x/net v0.45.0 h1:RLBg5JKixCy82FtLJpeNlVM0nrSqpCRYzVU1n8kj0tM=
+golang.org/x/net v0.39.0 h1:ZCu7HMWDxpXpaiKdhzIfaltL9Lp31x/3fCP11bc6/fY=
-golang.org/x/net v0.45.0/go.mod h1:ECOoLqd5U3Lhyeyo/QDCEVQ4sNgYsqvCZ722XogGieY=
+golang.org/x/net v0.39.0/go.mod h1:X7NRbYVEA+ewNkCNyJ513WmMdQ3BineSwVtN2zD/d+E=
 golang.org/x/oauth2 v0.0.0-20190226205417-e64efc72b421/go.mod h1:gOpvHmFTYa4IltrdGE7lF6nIHvwfUNPOp7c8zoXwtLw=
 golang.org/x/sync v0.0.0-20181108010431-42b317875d0f/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20181221193216-37e7f081c4d4/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
@@ -189,8 +185,8 @@ golang.org/x/sync v0.0.0-20190423024810-112230192c58/go.mod h1:RxMgew5VJxzue5/jJ
 golang.org/x/sync v0.0.0-20190911185100-cd5d95a43a6e/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20201020160332-67f06af15bc9/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20201207232520-09787c993a3a/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
-golang.org/x/sync v0.17.0 h1:l60nONMj9l5drqw6jlhIELNv9I0A4OFgRsG9k2oT9Ug=
+golang.org/x/sync v0.13.0 h1:AauUjRAJ9OSnvULf/ARrrVywoJDy0YS2AwQ98I37610=
-golang.org/x/sync v0.17.0/go.mod h1:9KTHXmSnoGruLpwFjVSX0lNNA75CykiMECbovNTZqGI=
+golang.org/x/sync v0.13.0/go.mod h1:1dzgHSNfp02xaA81J2MS99Qcpr2w7fw1gpm99rleRqA=
 golang.org/x/sys v0.0.0-20180905080454-ebe1bf3edb33/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20181116152217-5ac8a444bdc5/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20190215142949-d0b11bdaac8a/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
@@ -201,17 +197,18 @@ golang.org/x/sys v0.0.0-20200323222414-85ca7c5b95cd/go.mod h1:h1NjWce9XRLGQEsW7w
 golang.org/x/sys v0.0.0-20200615200032-f1bc736245b1/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20200625212154-ddb9806d33ae/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20200930185726-fdedc70b468f/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
+golang.org/x/sys v0.0.0-20201015000850-e3ed0017c211/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20201119102817-f84b799fce68/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210124154548-22da62e12c0c/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210603081109-ebe580a85c40/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.0.0-20220715151400-c0bba94af5f8/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.2.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.10.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
-golang.org/x/sys v0.37.0 h1:fdNQudmxPjkdUTPnLn5mdQv7Zwvbvpaxqs831goi9kQ=
+golang.org/x/sys v0.32.0 h1:s77OFDvIQeibCmezSnk/q6iAfkdiQaJi4VzroCFrN20=
-golang.org/x/sys v0.37.0/go.mod h1:OgkHotnGiDImocRcuBABYBEXf8A9a87e/uXjp9XT3ks=
+golang.org/x/sys v0.32.0/go.mod h1:BJP2sWEmIv4KK5OTEluFJCKSidICx8ciO85XgH3Ak8k=
 golang.org/x/term v0.0.0-20201126162022-7de9c90e9dd1/go.mod h1:bj7SfCRtBDWHUb9snDiAeCFNEtKQo2Wmx5Cou7ajbmo=
-golang.org/x/term v0.36.0 h1:zMPR+aF8gfksFprF/Nc/rd1wRS1EI6nDBGyWAvDzx2Q=
+golang.org/x/term v0.31.0 h1:erwDkOK1Msy6offm1mOgvspSkslFnIGsFnxOKoufg3o=
-golang.org/x/term v0.36.0/go.mod h1:Qu394IJq6V6dCBRgwqshf3mPF85AqzYEzofzRdZkWss=
+golang.org/x/term v0.31.0/go.mod h1:R4BeIy7D95HzImkxGkTW1UQTtP54tio2RyHz7PwK0aw=
 golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
 golang.org/x/text v0.3.2/go.mod h1:bEr9sfX3Q8Zfm5fL9x+3itogRgK3+ptLWKqgva+5dAk=
 golang.org/x/text v0.3.3/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
@@ -222,8 +219,8 @@ golang.org/x/tools v0.0.0-20191119224855-298f0cb1881e/go.mod h1:b+2E5dAYhXwXZwtn
 golang.org/x/tools v0.0.0-20200130002326-2f3ba24bd6e7/go.mod h1:TB2adYChydJhpapKDTa4BR/hXlZSLoq2Wpct/0txZ28=
 golang.org/x/tools v0.0.0-20200619180055-7c47624df98f/go.mod h1:EkVYQZoAsY45+roYkvgYkIh4xh/qjgUK9TdY2XT94GE=
 golang.org/x/tools v0.0.0-20210106214847-113979e3529a/go.mod h1:emZCQorbCU4vsT4fOWvOPXz4eW1wZW4PmDk9uLelYpA=
-golang.org/x/tools v0.33.0 h1:4qz2S3zmRxbGIhDIAgjxvFutSvH5EfnsYrRBj0UI0bc=
+golang.org/x/tools v0.30.0 h1:BgcpHewrV5AUp2G9MebG4XPFI1E2W41zU1SaqVA9vJY=
-golang.org/x/tools v0.33.0/go.mod h1:CIJMaWEY88juyUfo7UbgPqbC8rU2OqfAV1h2Qp0oMYI=
+golang.org/x/tools v0.30.0/go.mod h1:c347cR/OJfw5TI+GfX7RUPNMdDRRbjvYTS0jPyvsVtY=
 golang.org/x/xerrors v0.0.0-20190717185122-a985d3407aa7/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20191011141410-1b5146add898/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20191204190536-9bdfabe68543/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
@@ -242,8 +239,8 @@ google.golang.org/protobuf v1.20.1-0.20200309200217-e05f789c0967/go.mod h1:A+miE
 google.golang.org/protobuf v1.21.0/go.mod h1:47Nbq4nVaFHyn7ilMalzfO3qCViNmqZ2kzikPIcrTAo=
 google.golang.org/protobuf v1.23.0/go.mod h1:EGpADcykh3NcUnDUJcl1+ZksZNG86OlYog2l/sGQquU=
 google.golang.org/protobuf v1.26.0-rc.1/go.mod h1:jlhhOSvTdKEhbULTjvd4ARK9grFBp09yW+WbY/TyQbw=
-google.golang.org/protobuf v1.36.8 h1:xHScyCOEuuwZEc6UtSOvPbAT4zRh0xcNRYekJwfqyMc=
+google.golang.org/protobuf v1.36.6 h1:z1NpPI8ku2WgiWnf+t9wTPsn6eP1L7ksHUlkfLvd9xY=
-google.golang.org/protobuf v1.36.8/go.mod h1:fuxRtAxBytpl4zzqUh6/eyUujkJdNiuEkXntxiD/uRU=
+google.golang.org/protobuf v1.36.6/go.mod h1:jduwjTPXsFjZGTmRluh+L6NjiWu7pchiJ2/5YcXBHnY=
 gopkg.in/alecthomas/kingpin.v2 v2.2.6/go.mod h1:FMv+mEhP44yOT+4EoQTLFTRgOQ1FBLkstjWtayDeSgw=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
 gopkg.in/check.v1 v1.0.0-20190902080502-41f04d3bba15/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=

hostmap.go

@@ -738,8 +738,7 @@ func (i *HostInfo) buildNetworks(networks, unsafeNetworks []netip.Prefix) {
 
 	i.networks = new(bart.Lite)
 	for _, network := range networks {
-		nprefix := netip.PrefixFrom(network.Addr(), network.Addr().BitLen())
+		i.networks.Insert(network)
-		i.networks.Insert(nprefix)
 	}
 
 	for _, network := range unsafeNetworks {

inside.go

@@ -11,19 +11,19 @@ import (
 	"github.com/slackhq/nebula/routing"
 )
 
-func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet, nb, out []byte, q int, localCache firewall.ConntrackCache) {
+func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet, nb, out []byte, queue func(netip.AddrPort, int), q int, localCache firewall.ConntrackCache) bool {
 	err := newPacket(packet, false, fwPacket)
 	if err != nil {
 		if f.l.Level >= logrus.DebugLevel {
 			f.l.WithField("packet", packet).Debugf("Error while validating outbound packet: %s", err)
 		}
-		return
+		return false
 	}
 
 	// Ignore local broadcast packets
 	if f.dropLocalBroadcast {
 		if f.myBroadcastAddrsTable.Contains(fwPacket.RemoteAddr) {
-			return
+			return false
 		}
 	}
 
@@ -40,12 +40,12 @@ func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet
 		}
 		// Otherwise, drop. On linux, we should never see these packets - Linux
 		// routes packets from the nebula addr to the nebula addr through the loopback device.
-		return
+		return false
 	}
 
 	// Ignore multicast packets
 	if f.dropMulticast && fwPacket.RemoteAddr.IsMulticast() {
-		return
+		return false
 	}
 
 	hostinfo, ready := f.getOrHandshakeConsiderRouting(fwPacket, func(hh *HandshakeHostInfo) {
@@ -59,26 +59,26 @@ func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet
 				WithField("fwPacket", fwPacket).
 				Debugln("dropping outbound packet, vpnAddr not in our vpn networks or in unsafe networks")
 		}
-		return
+		return false
 	}
 
 	if !ready {
-		return
+		return false
 	}
 
 	dropReason := f.firewall.Drop(*fwPacket, false, hostinfo, f.pki.GetCAPool(), localCache)
 	if dropReason == nil {
-		f.sendNoMetrics(header.Message, 0, hostinfo.ConnectionState, hostinfo, netip.AddrPort{}, packet, nb, out, q)
+		return f.sendNoMetrics(header.Message, 0, hostinfo.ConnectionState, hostinfo, netip.AddrPort{}, packet, nb, out, queue, q)
-
-	} else {
-		f.rejectInside(packet, out, q)
-		if f.l.Level >= logrus.DebugLevel {
-			hostinfo.logger(f.l).
-				WithField("fwPacket", fwPacket).
-				WithField("reason", dropReason).
-				Debugln("dropping outbound packet")
-		}
 	}
+
+	f.rejectInside(packet, out, q)
+	if f.l.Level >= logrus.DebugLevel {
+		hostinfo.logger(f.l).
+			WithField("fwPacket", fwPacket).
+			WithField("reason", dropReason).
+			Debugln("dropping outbound packet")
+	}
+	return false
 }
 
 func (f *Interface) rejectInside(packet []byte, out []byte, q int) {
@@ -117,7 +117,7 @@ func (f *Interface) rejectOutside(packet []byte, ci *ConnectionState, hostinfo *
 		return
 	}
 
-	f.sendNoMetrics(header.Message, 0, ci, hostinfo, netip.AddrPort{}, out, nb, packet, q)
+	_ = f.sendNoMetrics(header.Message, 0, ci, hostinfo, netip.AddrPort{}, out, nb, packet, nil, q)
 }
 
 // Handshake will attempt to initiate a tunnel with the provided vpn address if it is within our vpn networks. This is a no-op if the tunnel is already established or being established
@@ -228,7 +228,7 @@ func (f *Interface) sendMessageNow(t header.MessageType, st header.MessageSubTyp
 		return
 	}
 
-	f.sendNoMetrics(header.Message, st, hostinfo.ConnectionState, hostinfo, netip.AddrPort{}, p, nb, out, 0)
+	_ = f.sendNoMetrics(header.Message, st, hostinfo.ConnectionState, hostinfo, netip.AddrPort{}, p, nb, out, nil, 0)
 }
 
 // SendMessageToVpnAddr handles real addr:port lookup and sends to the current best known address for vpnAddr
@@ -258,12 +258,12 @@ func (f *Interface) SendMessageToHostInfo(t header.MessageType, st header.Messag
 
 func (f *Interface) send(t header.MessageType, st header.MessageSubType, ci *ConnectionState, hostinfo *HostInfo, p, nb, out []byte) {
 	f.messageMetrics.Tx(t, st, 1)
-	f.sendNoMetrics(t, st, ci, hostinfo, netip.AddrPort{}, p, nb, out, 0)
+	_ = f.sendNoMetrics(t, st, ci, hostinfo, netip.AddrPort{}, p, nb, out, nil, 0)
 }
 
 func (f *Interface) sendTo(t header.MessageType, st header.MessageSubType, ci *ConnectionState, hostinfo *HostInfo, remote netip.AddrPort, p, nb, out []byte) {
 	f.messageMetrics.Tx(t, st, 1)
-	f.sendNoMetrics(t, st, ci, hostinfo, remote, p, nb, out, 0)
+	_ = f.sendNoMetrics(t, st, ci, hostinfo, remote, p, nb, out, nil, 0)
 }
 
 // SendVia sends a payload through a Relay tunnel. No authentication or encryption is done
@@ -331,9 +331,12 @@ func (f *Interface) SendVia(via *HostInfo,
 	f.connectionManager.RelayUsed(relay.LocalIndex)
 }
 
-func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType, ci *ConnectionState, hostinfo *HostInfo, remote netip.AddrPort, p, nb, out []byte, q int) {
+// sendNoMetrics encrypts and writes/queues an outbound packet. It returns true
+// when the payload has been handed to a caller-provided queue (meaning the
+// caller is responsible for flushing it later).
+func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType, ci *ConnectionState, hostinfo *HostInfo, remote netip.AddrPort, p, nb, out []byte, queue func(netip.AddrPort, int), q int) bool {
 	if ci.eKey == nil {
-		return
+		return false
 	}
 	useRelay := !remote.IsValid() && !hostinfo.remote.IsValid()
 	fullOut := out
@@ -380,32 +383,39 @@ func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType
 			WithField("udpAddr", remote).WithField("counter", c).
 			WithField("attemptedCounter", c).
 			Error("Failed to encrypt outgoing packet")
-		return
+		return false
 	}
 
-	if remote.IsValid() {
+	dest := remote
-		err = f.writers[q].WriteTo(out, remote)
+	if !dest.IsValid() {
-		if err != nil {
+		dest = hostinfo.remote
-			hostinfo.logger(f.l).WithError(err).
-				WithField("udpAddr", remote).Error("Failed to write outgoing packet")
-		}
-	} else if hostinfo.remote.IsValid() {
-		err = f.writers[q].WriteTo(out, hostinfo.remote)
-		if err != nil {
-			hostinfo.logger(f.l).WithError(err).
-				WithField("udpAddr", remote).Error("Failed to write outgoing packet")
-		}
-	} else {
-		// Try to send via a relay
-		for _, relayIP := range hostinfo.relayState.CopyRelayIps() {
-			relayHostInfo, relay, err := f.hostMap.QueryVpnAddrsRelayFor(hostinfo.vpnAddrs, relayIP)
-			if err != nil {
-				hostinfo.relayState.DeleteRelay(relayIP)
-				hostinfo.logger(f.l).WithField("relay", relayIP).WithError(err).Info("sendNoMetrics failed to find HostInfo")
-				continue
-			}
-			f.SendVia(relayHostInfo, relay, out, nb, fullOut[:header.Len+len(out)], true)
-			break
-		}
 	}
+
+	if dest.IsValid() {
+		if queue != nil {
+			queue(dest, len(out))
+			return true
+		}
+
+		err = f.writers[q].WriteTo(out, dest)
+		if err != nil {
+			hostinfo.logger(f.l).WithError(err).
+				WithField("udpAddr", dest).Error("Failed to write outgoing packet")
+		}
+		return false
+	}
+
+	// Try to send via a relay
+	for _, relayIP := range hostinfo.relayState.CopyRelayIps() {
+		relayHostInfo, relay, err := f.hostMap.QueryVpnAddrsRelayFor(hostinfo.vpnAddrs, relayIP)
+		if err != nil {
+			hostinfo.relayState.DeleteRelay(relayIP)
+			hostinfo.logger(f.l).WithField("relay", relayIP).WithError(err).Info("sendNoMetrics failed to find HostInfo")
+			continue
+		}
+		f.SendVia(relayHostInfo, relay, out, nb, fullOut[:header.Len+len(out)], true)
+		break
+	}
+
+	return false
 }

interface.go

@@ -5,9 +5,7 @@ import (
 	"errors"
 	"fmt"
 	"io"
-	"math/bits"
 	"net/netip"
-	"os"
 	"runtime"
 	"sync"
 	"sync/atomic"
@@ -20,14 +18,21 @@ import (
 	"github.com/slackhq/nebula/firewall"
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/overlay"
+	"github.com/slackhq/nebula/packet"
 	"github.com/slackhq/nebula/udp"
 )
 
 const (
-	mtu                        = 9001
+	mtu = 9001
-	tunReadBufferSize          = mtu * 8
+
-	defaultDecryptWorkerFactor = 2
+	inboundBatchSizeDefault      = 128
-	defaultInboundQueueDepth   = 1024
+	outboundBatchSizeDefault     = 64
+	batchFlushIntervalDefault    = 12 * time.Microsecond
+	maxOutstandingBatchesDefault = 8
+	sendBatchSizeDefault         = 64
+	maxPendingPacketsDefault     = 32
+	maxPendingBytesDefault       = 64 * 1024
+	maxSendBufPerRoutineDefault  = 16
 )
 
 type InterfaceConfig struct {
@@ -54,9 +59,18 @@ type InterfaceConfig struct {
 	reQueryWait     time.Duration
 
 	ConntrackCacheTimeout time.Duration
+	BatchConfig           BatchConfig
 	l                     *logrus.Logger
-	DecryptWorkers        int
+}
-	DecryptQueueDepth     int
+
+type BatchConfig struct {
+	InboundBatchSize      int
+	OutboundBatchSize     int
+	FlushInterval         time.Duration
+	MaxOutstandingPerChan int
+	MaxPendingPackets     int
+	MaxPendingBytes       int
+	MaxSendBuffersPerChan int
 }
 
 type Interface struct {
@@ -96,172 +110,165 @@ type Interface struct {
 
 	writers []udp.Conn
 	readers []io.ReadWriteCloser
+	wg      sync.WaitGroup
 
 	metricHandshakes    metrics.Histogram
 	messageMetrics      *MessageMetrics
 	cachedPacketMetrics *cachedPacketMetrics
 
-	l              *logrus.Logger
+	l *logrus.Logger
-	ctx            context.Context
+
-	udpListenWG    sync.WaitGroup
+	inPool  sync.Pool
-	inboundPool    sync.Pool
+	inbound []chan *packetBatch
-	decryptWG      sync.WaitGroup
+
-	decryptQueues  []*inboundRing
+	outPool  sync.Pool
-	decryptWorkers int
+	outbound []chan *outboundBatch
-	decryptStates  []decryptWorkerState
+
-	decryptCounter atomic.Uint32
+	packetBatchPool   sync.Pool
+	outboundBatchPool sync.Pool
+
+	sendPool      sync.Pool
+	sendBufCache  [][]*[]byte
+	sendBatchSize int
+
+	inboundBatchSize      int
+	outboundBatchSize     int
+	batchFlushInterval    time.Duration
+	maxOutstandingPerChan int
+	maxPendingPackets     int
+	maxPendingBytes       int
+	maxSendBufPerRoutine  int
 }
 
-type inboundPacket struct {
+type outboundSend struct {
-	addr    netip.AddrPort
+	buf    *[]byte
-	payload []byte
+	length int
-	release func()
+	addr   netip.AddrPort
-	queue   int
 }
 
-type decryptWorkerState struct {
+type packetBatch struct {
-	queue  *inboundRing
+	packets []*packet.Packet
-	notify chan struct{}
 }
 
-type decryptContext struct {
+func newPacketBatch(capacity int) *packetBatch {
-	ctTicker *firewall.ConntrackCacheTicker
+	return &packetBatch{
-	plain    []byte
+		packets: make([]*packet.Packet, 0, capacity),
-	head     header.H
-	fwPacket firewall.Packet
-	light    *LightHouseHandler
-	nebula   []byte
-}
-
-type inboundCell struct {
-	seq atomic.Uint64
-	pkt *inboundPacket
-}
-
-type inboundRing struct {
-	mask       uint64
-	cells      []inboundCell
-	enqueuePos atomic.Uint64
-	dequeuePos atomic.Uint64
-}
-
-func newInboundRing(capacity int) *inboundRing {
-	if capacity < 2 {
-		capacity = 2
 	}
-	size := nextPowerOfTwo(uint32(capacity))
-	if size < 2 {
-		size = 2
-	}
-	ring := &inboundRing{
-		mask:  uint64(size - 1),
-		cells: make([]inboundCell, size),
-	}
-	for i := range ring.cells {
-		ring.cells[i].seq.Store(uint64(i))
-	}
-	return ring
 }
 
-func nextPowerOfTwo(v uint32) uint32 {
+func (b *packetBatch) add(p *packet.Packet) {
-	if v == 0 {
+	b.packets = append(b.packets, p)
-		return 1
-	}
-	return 1 << (32 - bits.LeadingZeros32(v-1))
 }
 
-func (r *inboundRing) Enqueue(pkt *inboundPacket) bool {
+func (b *packetBatch) reset() {
-	var cell *inboundCell
+	for i := range b.packets {
-	pos := r.enqueuePos.Load()
+		b.packets[i] = nil
-	for {
-		cell = &r.cells[pos&r.mask]
-		seq := cell.seq.Load()
-		diff := int64(seq) - int64(pos)
-		if diff == 0 {
-			if r.enqueuePos.CompareAndSwap(pos, pos+1) {
-				break
-			}
-		} else if diff < 0 {
-			return false
-		} else {
-			pos = r.enqueuePos.Load()
-		}
 	}
-	cell.pkt = pkt
+	b.packets = b.packets[:0]
-	cell.seq.Store(pos + 1)
-	return true
 }
 
-func (r *inboundRing) Dequeue() (*inboundPacket, bool) {
+func (f *Interface) getPacketBatch() *packetBatch {
-	var cell *inboundCell
+	if v := f.packetBatchPool.Get(); v != nil {
-	pos := r.dequeuePos.Load()
+		b := v.(*packetBatch)
-	for {
+		b.reset()
-		cell = &r.cells[pos&r.mask]
+		return b
-		seq := cell.seq.Load()
-		diff := int64(seq) - int64(pos+1)
-		if diff == 0 {
-			if r.dequeuePos.CompareAndSwap(pos, pos+1) {
-				break
-			}
-		} else if diff < 0 {
-			return nil, false
-		} else {
-			pos = r.dequeuePos.Load()
-		}
 	}
-	pkt := cell.pkt
+	return newPacketBatch(f.inboundBatchSize)
-	cell.pkt = nil
-	cell.seq.Store(pos + r.mask + 1)
-	return pkt, true
 }
 
-func (f *Interface) getInboundPacket() *inboundPacket {
+func (f *Interface) releasePacketBatch(b *packetBatch) {
-	if pkt, ok := f.inboundPool.Get().(*inboundPacket); ok && pkt != nil {
+	b.reset()
-		return pkt
+	f.packetBatchPool.Put(b)
-	}
-	return &inboundPacket{}
 }
 
-func (f *Interface) putInboundPacket(pkt *inboundPacket) {
+type outboundBatch struct {
-	if pkt == nil {
+	payloads []*[]byte
+}
+
+func newOutboundBatch(capacity int) *outboundBatch {
+	return &outboundBatch{payloads: make([]*[]byte, 0, capacity)}
+}
+
+func (b *outboundBatch) add(buf *[]byte) {
+	b.payloads = append(b.payloads, buf)
+}
+
+func (b *outboundBatch) reset() {
+	for i := range b.payloads {
+		b.payloads[i] = nil
+	}
+	b.payloads = b.payloads[:0]
+}
+
+func (f *Interface) getOutboundBatch() *outboundBatch {
+	if v := f.outboundBatchPool.Get(); v != nil {
+		b := v.(*outboundBatch)
+		b.reset()
+		return b
+	}
+	return newOutboundBatch(f.outboundBatchSize)
+}
+
+func (f *Interface) releaseOutboundBatch(b *outboundBatch) {
+	b.reset()
+	f.outboundBatchPool.Put(b)
+}
+
+func (f *Interface) getSendBuffer(q int) *[]byte {
+	cache := f.sendBufCache[q]
+	if n := len(cache); n > 0 {
+		buf := cache[n-1]
+		f.sendBufCache[q] = cache[:n-1]
+		*buf = (*buf)[:0]
+		return buf
+	}
+	if v := f.sendPool.Get(); v != nil {
+		buf := v.(*[]byte)
+		*buf = (*buf)[:0]
+		return buf
+	}
+	b := make([]byte, mtu)
+	return &b
+}
+
+func (f *Interface) releaseSendBuffer(q int, buf *[]byte) {
+	if buf == nil {
 		return
 	}
-	pkt.addr = netip.AddrPort{}
+	*buf = (*buf)[:0]
-	pkt.payload = nil
+	cache := f.sendBufCache[q]
-	pkt.release = nil
+	if len(cache) < f.maxSendBufPerRoutine {
-	pkt.queue = 0
+		f.sendBufCache[q] = append(cache, buf)
-	f.inboundPool.Put(pkt)
-}
-
-func newDecryptContext(f *Interface) *decryptContext {
-	return &decryptContext{
-		ctTicker: firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout),
-		plain:    make([]byte, udp.MTU),
-		head:     header.H{},
-		fwPacket: firewall.Packet{},
-		light:    f.lightHouse.NewRequestHandler(),
-		nebula:   make([]byte, 12, 12),
-	}
-}
-
-func (f *Interface) processInboundPacket(pkt *inboundPacket, ctx *decryptContext) {
-	if pkt == nil {
 		return
 	}
-	defer func() {
+	f.sendPool.Put(buf)
-		if pkt.release != nil {
+}
-			pkt.release()
-		}
-		f.putInboundPacket(pkt)
-	}()
 
-	ctx.head = header.H{}
+func (f *Interface) flushSendQueue(q int, pending *[]outboundSend, pendingBytes *int) {
-	ctx.fwPacket = firewall.Packet{}
+	if len(*pending) == 0 {
-	var cache firewall.ConntrackCache
+		return
-	if ctx.ctTicker != nil {
+	}
-		cache = ctx.ctTicker.Get(f.l)
+
+	batch := make([]udp.BatchPacket, len(*pending))
+	for i, entry := range *pending {
+		batch[i] = udp.BatchPacket{
+			Payload: (*entry.buf)[:entry.length],
+			Addr:    entry.addr,
+		}
+	}
+
+	sent, err := f.writers[q].WriteBatch(batch)
+	if err != nil {
+		f.l.WithError(err).WithField("sent", sent).Error("Failed to batch send packets")
+	}
+
+	for _, entry := range *pending {
+		f.releaseSendBuffer(q, entry.buf)
+	}
+	*pending = (*pending)[:0]
+	if pendingBytes != nil {
+		*pendingBytes = 0
 	}
-	f.readOutsidePackets(pkt.addr, nil, ctx.plain[:0], pkt.payload, &ctx.head, &ctx.fwPacket, ctx.light, ctx.nebula, pkt.queue, cache)
 }
 
 type EncWriter interface {
@@ -331,35 +338,29 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 	}
 
 	cs := c.pki.getCertState()
-	decryptWorkers := c.DecryptWorkers
-	if decryptWorkers < 0 {
-		decryptWorkers = 0
-	}
-	if decryptWorkers == 0 {
-		decryptWorkers = c.routines * defaultDecryptWorkerFactor
-		if decryptWorkers < c.routines {
-			decryptWorkers = c.routines
-		}
-	}
-	if decryptWorkers < 0 {
-		decryptWorkers = 0
-	}
-	if runtime.GOOS != "linux" {
-		decryptWorkers = 0
-	}
 
-	queueDepth := c.DecryptQueueDepth
+	bc := c.BatchConfig
-	if queueDepth <= 0 {
+	if bc.InboundBatchSize <= 0 {
-		queueDepth = defaultInboundQueueDepth
+		bc.InboundBatchSize = inboundBatchSizeDefault
 	}
-	minDepth := c.routines * 64
+	if bc.OutboundBatchSize <= 0 {
-	if minDepth <= 0 {
+		bc.OutboundBatchSize = outboundBatchSizeDefault
-		minDepth = 64
 	}
-	if queueDepth < minDepth {
+	if bc.FlushInterval <= 0 {
-		queueDepth = minDepth
+		bc.FlushInterval = batchFlushIntervalDefault
+	}
+	if bc.MaxOutstandingPerChan <= 0 {
+		bc.MaxOutstandingPerChan = maxOutstandingBatchesDefault
+	}
+	if bc.MaxPendingPackets <= 0 {
+		bc.MaxPendingPackets = maxPendingPacketsDefault
+	}
+	if bc.MaxPendingBytes <= 0 {
+		bc.MaxPendingBytes = maxPendingBytesDefault
+	}
+	if bc.MaxSendBuffersPerChan <= 0 {
+		bc.MaxSendBuffersPerChan = maxSendBufPerRoutineDefault
 	}
-
 	ifce := &Interface{
 		pki:                   c.pki,
 		hostMap:               c.HostMap,
@@ -392,38 +393,62 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 			dropped: metrics.GetOrRegisterCounter("hostinfo.cached_packets.dropped", nil),
 		},
 
-		l:              c.l,
+		inbound:  make([]chan *packetBatch, c.routines),
-		ctx:            ctx,
+		outbound: make([]chan *outboundBatch, c.routines),
-		inboundPool:    sync.Pool{New: func() any { return &inboundPacket{} }},
+
-		decryptWorkers: decryptWorkers,
+		l: c.l,
+
+		inboundBatchSize:      bc.InboundBatchSize,
+		outboundBatchSize:     bc.OutboundBatchSize,
+		batchFlushInterval:    bc.FlushInterval,
+		maxOutstandingPerChan: bc.MaxOutstandingPerChan,
+		maxPendingPackets:     bc.MaxPendingPackets,
+		maxPendingBytes:       bc.MaxPendingBytes,
+		maxSendBufPerRoutine:  bc.MaxSendBuffersPerChan,
+		sendBatchSize:         bc.OutboundBatchSize,
 	}
 
+	for i := 0; i < c.routines; i++ {
+		ifce.inbound[i] = make(chan *packetBatch, ifce.maxOutstandingPerChan)
+		ifce.outbound[i] = make(chan *outboundBatch, ifce.maxOutstandingPerChan)
+	}
+
+	ifce.inPool = sync.Pool{New: func() any {
+		return packet.New()
+	}}
+
+	ifce.outPool = sync.Pool{New: func() any {
+		t := make([]byte, mtu)
+		return &t
+	}}
+
+	ifce.packetBatchPool = sync.Pool{New: func() any {
+		return newPacketBatch(ifce.inboundBatchSize)
+	}}
+
+	ifce.outboundBatchPool = sync.Pool{New: func() any {
+		return newOutboundBatch(ifce.outboundBatchSize)
+	}}
+
+	ifce.sendPool = sync.Pool{New: func() any {
+		buf := make([]byte, mtu)
+		return &buf
+	}}
+	ifce.sendBufCache = make([][]*[]byte, c.routines)
+
 	ifce.tryPromoteEvery.Store(c.tryPromoteEvery)
 	ifce.reQueryEvery.Store(c.reQueryEvery)
 	ifce.reQueryWait.Store(int64(c.reQueryWait))
 
 	ifce.connectionManager.intf = ifce
 
-	if decryptWorkers > 0 {
-		ifce.decryptQueues = make([]*inboundRing, decryptWorkers)
-		ifce.decryptStates = make([]decryptWorkerState, decryptWorkers)
-		for i := 0; i < decryptWorkers; i++ {
-			queue := newInboundRing(queueDepth)
-			ifce.decryptQueues[i] = queue
-			ifce.decryptStates[i] = decryptWorkerState{
-				queue:  queue,
-				notify: make(chan struct{}, 1),
-			}
-		}
-	}
-
 	return ifce, nil
 }
 
 // activate creates the interface on the host. After the interface is created, any
 // other services that want to bind listeners to its IP may do so successfully. However,
 // the interface isn't going to process anything until run() is called.
-func (f *Interface) activate() {
+func (f *Interface) activate() error {
 	// actually turn on tun dev
 
 	addr, err := f.outside.LocalAddr()
@@ -444,94 +469,44 @@ func (f *Interface) activate() {
 		if i > 0 {
 			reader, err = f.inside.NewMultiQueueReader()
 			if err != nil {
-				f.l.Fatal(err)
+				return err
 			}
 		}
 		f.readers[i] = reader
 	}
 
-	if err := f.inside.Activate(); err != nil {
+	if err = f.inside.Activate(); err != nil {
 		f.inside.Close()
-		f.l.Fatal(err)
+		return err
 	}
+
+	return nil
 }
 
-func (f *Interface) startDecryptWorkers() {
+func (f *Interface) run(c context.Context) (func(), error) {
-	if f.decryptWorkers <= 0 || len(f.decryptQueues) == 0 {
-		return
-	}
-	f.decryptWG.Add(f.decryptWorkers)
-	for i := 0; i < f.decryptWorkers; i++ {
-		go f.decryptWorker(i)
-	}
-}
-
-func (f *Interface) decryptWorker(id int) {
-	defer f.decryptWG.Done()
-	if id < 0 || id >= len(f.decryptStates) {
-		return
-	}
-	state := f.decryptStates[id]
-	if state.queue == nil {
-		return
-	}
-	ctx := newDecryptContext(f)
-	for {
-		for {
-			pkt, ok := state.queue.Dequeue()
-			if !ok {
-				break
-			}
-			f.processInboundPacket(pkt, ctx)
-		}
-		if f.closed.Load() || f.ctx.Err() != nil {
-			for {
-				pkt, ok := state.queue.Dequeue()
-				if !ok {
-					return
-				}
-				f.processInboundPacket(pkt, ctx)
-			}
-		}
-		select {
-		case <-f.ctx.Done():
-		case <-state.notify:
-		}
-	}
-}
-
-func (f *Interface) notifyDecryptWorker(idx int) {
-	if idx < 0 || idx >= len(f.decryptStates) {
-		return
-	}
-	state := f.decryptStates[idx]
-	if state.notify == nil {
-		return
-	}
-	select {
-	case state.notify <- struct{}{}:
-	default:
-	}
-}
-
-func (f *Interface) run() {
-	f.startDecryptWorkers()
-	// Launch n queues to read packets from udp
-	f.udpListenWG.Add(f.routines)
 	for i := 0; i < f.routines; i++ {
+		// Launch n queues to read packets from udp
+		f.wg.Add(1)
 		go f.listenOut(i)
+
+		// Launch n queues to read packets from tun dev
+		f.wg.Add(1)
+		go f.listenIn(f.readers[i], i)
+
+		// Launch n queues to read packets from tun dev
+		f.wg.Add(1)
+		go f.workerIn(i, c)
+
+		// Launch n queues to read packets from tun dev
+		f.wg.Add(1)
+		go f.workerOut(i, c)
 	}
 
-	// Launch n queues to read packets from tun dev
+	return f.wg.Wait, nil
-	for i := 0; i < f.routines; i++ {
-		go f.listenIn(f.readers[i], i)
-	}
 }
 
 func (f *Interface) listenOut(i int) {
 	runtime.LockOSThread()
-	defer f.udpListenWG.Done()
-
 	var li udp.Conn
 	if i > 0 {
 		li = f.writers[i]
@@ -539,96 +514,176 @@ func (f *Interface) listenOut(i int) {
 		li = f.outside
 	}
 
-	useWorkers := f.decryptWorkers > 0 && len(f.decryptQueues) > 0
+	batch := f.getPacketBatch()
-	var (
+	lastFlush := time.Now()
-		inlineTicker  *firewall.ConntrackCacheTicker
-		inlineHandler *LightHouseHandler
-		inlinePlain   []byte
-		inlineHeader  header.H
-		inlinePacket  firewall.Packet
-		inlineNB      []byte
-		inlineCtx     *decryptContext
-	)
 
-	if useWorkers {
+	flush := func(force bool) {
-		inlineCtx = newDecryptContext(f)
+		if len(batch.packets) == 0 {
-	} else {
+			if force {
-		inlineTicker = firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
+				f.releasePacketBatch(batch)
-		inlineHandler = f.lightHouse.NewRequestHandler()
+			}
-		inlinePlain = make([]byte, udp.MTU)
+			return
-		inlineNB = make([]byte, 12, 12)
+		}
+
+		f.inbound[i] <- batch
+		batch = f.getPacketBatch()
+		lastFlush = time.Now()
 	}
 
-	li.ListenOut(func(fromUdpAddr netip.AddrPort, payload []byte, release func()) {
+	err := li.ListenOut(func(fromUdpAddr netip.AddrPort, payload []byte) {
-		if !useWorkers {
+		p := f.inPool.Get().(*packet.Packet)
-			if release != nil {
+		p.Payload = p.Payload[:mtu]
-				defer release()
+		copy(p.Payload, payload)
-			}
+		p.Payload = p.Payload[:len(payload)]
-			select {
+		p.Addr = fromUdpAddr
-			case <-f.ctx.Done():
+		batch.add(p)
-				return
-			default:
-			}
-			inlineHeader = header.H{}
-			inlinePacket = firewall.Packet{}
-			var cache firewall.ConntrackCache
-			if inlineTicker != nil {
-				cache = inlineTicker.Get(f.l)
-			}
-			f.readOutsidePackets(fromUdpAddr, nil, inlinePlain[:0], payload, &inlineHeader, &inlinePacket, inlineHandler, inlineNB, i, cache)
-			return
-		}
 
-		if f.ctx.Err() != nil {
+		if len(batch.packets) >= f.inboundBatchSize || time.Since(lastFlush) >= f.batchFlushInterval {
-			if release != nil {
+			flush(false)
-				release()
-			}
-			return
 		}
-
-		pkt := f.getInboundPacket()
-		pkt.addr = fromUdpAddr
-		pkt.payload = payload
-		pkt.release = release
-		pkt.queue = i
-
-		queueCount := len(f.decryptQueues)
-		if queueCount == 0 {
-			f.processInboundPacket(pkt, inlineCtx)
-			return
-		}
-		w := int(f.decryptCounter.Add(1)-1) % queueCount
-		if w < 0 || w >= queueCount || !f.decryptQueues[w].Enqueue(pkt) {
-			f.processInboundPacket(pkt, inlineCtx)
-			return
-		}
-		f.notifyDecryptWorker(w)
 	})
+
+	if len(batch.packets) > 0 {
+		f.inbound[i] <- batch
+	} else {
+		f.releasePacketBatch(batch)
+	}
+
+	if err != nil && !f.closed.Load() {
+		f.l.WithError(err).Error("Error while reading packet inbound packet, closing")
+		//TODO: Trigger Control to close
+	}
+
+	f.l.Debugf("underlay reader %v is done", i)
+	f.wg.Done()
 }
 
 func (f *Interface) listenIn(reader io.ReadWriteCloser, i int) {
 	runtime.LockOSThread()
 
-	packet := make([]byte, tunReadBufferSize)
+	batch := f.getOutboundBatch()
-	out := make([]byte, tunReadBufferSize)
+	lastFlush := time.Now()
-	fwPacket := &firewall.Packet{}
-	nb := make([]byte, 12, 12)
 
-	conntrackCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
+	flush := func(force bool) {
-
+		if len(batch.payloads) == 0 {
-	for {
+			if force {
-		n, err := reader.Read(packet)
+				f.releaseOutboundBatch(batch)
-		if err != nil {
-			if errors.Is(err, os.ErrClosed) && f.closed.Load() {
-				return
 			}
-
+			return
-			f.l.WithError(err).Error("Error while reading outbound packet")
-			// This only seems to happen when something fatal happens to the fd, so exit.
-			os.Exit(2)
 		}
 
-		f.consumeInsidePacket(packet[:n], fwPacket, nb, out, i, conntrackCache.Get(f.l))
+		f.outbound[i] <- batch
+		batch = f.getOutboundBatch()
+		lastFlush = time.Now()
+	}
+
+	for {
+		p := f.outPool.Get().(*[]byte)
+		*p = (*p)[:mtu]
+		n, err := reader.Read(*p)
+		if err != nil {
+			if !f.closed.Load() {
+				f.l.WithError(err).Error("Error while reading outbound packet, closing")
+				//TODO: Trigger Control to close
+			}
+			break
+		}
+
+		*p = (*p)[:n]
+		batch.add(p)
+
+		if len(batch.payloads) >= f.outboundBatchSize || time.Since(lastFlush) >= f.batchFlushInterval {
+			flush(false)
+		}
+	}
+
+	if len(batch.payloads) > 0 {
+		f.outbound[i] <- batch
+	} else {
+		f.releaseOutboundBatch(batch)
+	}
+
+	f.l.Debugf("overlay reader %v is done", i)
+	f.wg.Done()
+}
+
+func (f *Interface) workerIn(i int, ctx context.Context) {
+	lhh := f.lightHouse.NewRequestHandler()
+	conntrackCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
+	fwPacket2 := &firewall.Packet{}
+	nb2 := make([]byte, 12, 12)
+	result2 := make([]byte, mtu)
+	h := &header.H{}
+
+	for {
+		select {
+		case batch := <-f.inbound[i]:
+			for _, p := range batch.packets {
+				f.readOutsidePackets(p.Addr, nil, result2[:0], p.Payload, h, fwPacket2, lhh, nb2, i, conntrackCache.Get(f.l))
+				p.Payload = p.Payload[:mtu]
+				f.inPool.Put(p)
+			}
+			f.releasePacketBatch(batch)
+		case <-ctx.Done():
+			f.wg.Done()
+			return
+		}
+	}
+}
+
+func (f *Interface) workerOut(i int, ctx context.Context) {
+	conntrackCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
+	fwPacket1 := &firewall.Packet{}
+	nb1 := make([]byte, 12, 12)
+	pending := make([]outboundSend, 0, f.sendBatchSize)
+	pendingBytes := 0
+	maxPendingPackets := f.maxPendingPackets
+	if maxPendingPackets <= 0 {
+		maxPendingPackets = f.sendBatchSize
+	}
+	maxPendingBytes := f.maxPendingBytes
+	if maxPendingBytes <= 0 {
+		maxPendingBytes = f.sendBatchSize * mtu
+	}
+
+	for {
+		select {
+		case batch := <-f.outbound[i]:
+			for _, data := range batch.payloads {
+				sendBuf := f.getSendBuffer(i)
+				buf := (*sendBuf)[:0]
+				queue := func(addr netip.AddrPort, length int) {
+					if len(pending) >= maxPendingPackets || pendingBytes+length > maxPendingBytes {
+						f.flushSendQueue(i, &pending, &pendingBytes)
+					}
+					pending = append(pending, outboundSend{
+						buf:    sendBuf,
+						length: length,
+						addr:   addr,
+					})
+					pendingBytes += length
+					if len(pending) >= f.sendBatchSize || pendingBytes >= maxPendingBytes {
+						f.flushSendQueue(i, &pending, &pendingBytes)
+					}
+				}
+				sent := f.consumeInsidePacket(*data, fwPacket1, nb1, buf, queue, i, conntrackCache.Get(f.l))
+				if !sent {
+					f.releaseSendBuffer(i, sendBuf)
+				}
+				*data = (*data)[:mtu]
+				f.outPool.Put(data)
+			}
+			f.releaseOutboundBatch(batch)
+			if len(pending) > 0 {
+				f.flushSendQueue(i, &pending, &pendingBytes)
+			}
+		case <-ctx.Done():
+			if len(pending) > 0 {
+				f.flushSendQueue(i, &pending, &pendingBytes)
+			}
+			f.wg.Done()
+			return
+		}
 	}
 }
 
@@ -781,6 +836,7 @@ func (f *Interface) GetCertState() *CertState {
 func (f *Interface) Close() error {
 	f.closed.Store(true)
 
+	// Release the udp readers
 	for _, u := range f.writers {
 		err := u.Close()
 		if err != nil {
@@ -788,19 +844,13 @@ func (f *Interface) Close() error {
 		}
 	}
 
-	f.udpListenWG.Wait()
+	// Release the tun readers
-	if f.decryptWorkers > 0 {
+	for _, u := range f.readers {
-		for _, state := range f.decryptStates {
+		err := u.Close()
-			if state.notify != nil {
+		if err != nil {
-				select {
+			f.l.WithError(err).Error("Error while closing tun device")
-				case state.notify <- struct{}{}:
-				default:
-				}
-			}
 		}
-		f.decryptWG.Wait()
 	}
 
-	// Release the tun device
+	return nil
-	return f.inside.Close()
 }

main.go

@@ -120,8 +120,6 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		l.WithField("duration", conntrackCacheTimeout).Info("Using routine-local conntrack cache")
 	}
 
-	udp.SetDisableUDPCsum(c.GetBool("listen.disable_udp_checksum", false))
-
 	var tun overlay.Device
 	if !configTest {
 		c.CatchHUP(ctx)
@@ -166,7 +164,7 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 
 		for i := 0; i < routines; i++ {
 			l.Infof("listening on %v", netip.AddrPortFrom(listenHost, uint16(port)))
-			udpServer, err := udp.NewListener(l, listenHost, port, routines > 1, c.GetInt("listen.batch", 64))
+			udpServer, err := udp.NewListener(l, listenHost, port, routines > 1, c.GetInt("listen.batch", 128))
 			if err != nil {
 				return nil, util.NewContextualError("Failed to open udp listener", m{"queue": i}, err)
 			}
@@ -223,8 +221,15 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		}
 	}
 
-	decryptWorkers := c.GetInt("listen.decrypt_workers", 0)
+	batchCfg := BatchConfig{
-	decryptQueueDepth := c.GetInt("listen.decrypt_queue_depth", 0)
+		InboundBatchSize:      c.GetInt("batch.inbound_size", inboundBatchSizeDefault),
+		OutboundBatchSize:     c.GetInt("batch.outbound_size", outboundBatchSizeDefault),
+		FlushInterval:         c.GetDuration("batch.flush_interval", batchFlushIntervalDefault),
+		MaxOutstandingPerChan: c.GetInt("batch.max_outstanding", maxOutstandingBatchesDefault),
+		MaxPendingPackets:     c.GetInt("batch.max_pending_packets", 0),
+		MaxPendingBytes:       c.GetInt("batch.max_pending_bytes", 0),
+		MaxSendBuffersPerChan: c.GetInt("batch.max_send_buffers_per_routine", 0),
+	}
 
 	ifConfig := &InterfaceConfig{
 		HostMap:               hostMap,
@@ -247,9 +252,8 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		relayManager:          NewRelayManager(ctx, l, hostMap, c),
 		punchy:                punchy,
 		ConntrackCacheTimeout: conntrackCacheTimeout,
+		BatchConfig:           batchCfg,
 		l:                     l,
-		DecryptWorkers:        decryptWorkers,
-		DecryptQueueDepth:     decryptQueueDepth,
 	}
 
 	var ifce *Interface
@@ -291,14 +295,14 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 	}
 
 	return &Control{
-		ifce,
+		f:                      ifce,
-		l,
+		l:                      l,
-		ctx,
+		ctx:                    ctx,
-		cancel,
+		cancel:                 cancel,
-		sshStart,
+		sshStart:               sshStart,
-		statsStart,
+		statsStart:             statsStart,
-		dnsStart,
+		dnsStart:               dnsStart,
-		lightHouse.StartUpdateWorker,
+		lighthouseStart:        lightHouse.StartUpdateWorker,
-		connManager.Start,
+		connectionManagerStart: connManager.Start,
 	}, nil
 }

outside.go

@@ -29,7 +29,7 @@ func (f *Interface) readOutsidePackets(ip netip.AddrPort, via *ViaSender, out []
 		return
 	}
 
-	//l.Error("in packet ", header, packet[HeaderLen:])
+	//f.l.Error("in packet ", h)
 	if ip.IsValid() {
 		if f.myVpnNetworksTable.Contains(ip.Addr()) {
 			if f.l.Level >= logrus.DebugLevel {
@@ -245,6 +245,7 @@ func (f *Interface) handleHostRoaming(hostinfo *HostInfo, udpAddr netip.AddrPort
 			return
 		}
 
+		//TODO: Seems we have a bunch of stuff racing here, since we don't have a lock on hostinfo anymore we announce roaming in bursts
 		hostinfo.logger(f.l).WithField("udpAddr", hostinfo.remote).WithField("newAddr", udpAddr).
 			Info("Host roamed to new udp ip/port.")
 		hostinfo.lastRoam = time.Now()
@@ -470,13 +471,7 @@ func (f *Interface) decryptToTun(hostinfo *HostInfo, messageCounter uint64, out
 
 	out, err = hostinfo.ConnectionState.dKey.DecryptDanger(out, packet[:header.Len], packet[header.Len:], messageCounter, nb)
 	if err != nil {
-		hostinfo.logger(f.l).
+		hostinfo.logger(f.l).WithError(err).WithField("fwPacket", fwPacket).Error("Failed to decrypt packet")
-			WithError(err).
-			WithField("tag", "decrypt-debug").
-			WithField("remoteIndexLocal", hostinfo.localIndexId).
-			WithField("messageCounter", messageCounter).
-			WithField("packet_len", len(packet)).
-			Error("Failed to decrypt packet")
 		return false
 	}
 

overlay/tun.go

@@ -1,7 +1,6 @@
 package overlay
 
 import (
-	"fmt"
 	"net"
 	"net/netip"
 
@@ -82,41 +81,3 @@ func prefixToMask(prefix netip.Prefix) netip.Addr {
 	addr, _ := netip.AddrFromSlice(net.CIDRMask(prefix.Bits(), pLen))
 	return addr
 }
-
-func flipBytes(b []byte) []byte {
-	for i := 0; i < len(b); i++ {
-		b[i] ^= 0xFF
-	}
-	return b
-}
-func orBytes(a []byte, b []byte) []byte {
-	ret := make([]byte, len(a))
-	for i := 0; i < len(a); i++ {
-		ret[i] = a[i] | b[i]
-	}
-	return ret
-}
-
-func getBroadcast(cidr netip.Prefix) netip.Addr {
-	broadcast, _ := netip.AddrFromSlice(
-		orBytes(
-			cidr.Addr().AsSlice(),
-			flipBytes(prefixToMask(cidr).AsSlice()),
-		),
-	)
-	return broadcast
-}
-
-func selectGateway(dest netip.Prefix, gateways []netip.Prefix) (netip.Prefix, error) {
-	for _, gateway := range gateways {
-		if dest.Addr().Is4() && gateway.Addr().Is4() {
-			return gateway, nil
-		}
-
-		if dest.Addr().Is6() && gateway.Addr().Is6() {
-			return gateway, nil
-		}
-	}
-
-	return netip.Prefix{}, fmt.Errorf("no gateway found for %v in the list of vpn networks", dest)
-}

overlay/tun_darwin.go

@@ -294,6 +294,7 @@ func (t *tun) activate6(network netip.Prefix) error {
 			Vltime: 0xffffffff,
 			Pltime: 0xffffffff,
 		},
+		//TODO: CERT-V2 should we disable DAD (duplicate address detection) and mark this as a secured address?
 		Flags: _IN6_IFF_NODAD,
 	}
 

overlay/tun_freebsd.go

@@ -501,6 +501,30 @@ func (t *tun) deviceBytes() (o [16]byte) {
 	return
 }
 
+func flipBytes(b []byte) []byte {
+	for i := 0; i < len(b); i++ {
+		b[i] ^= 0xFF
+	}
+	return b
+}
+func orBytes(a []byte, b []byte) []byte {
+	ret := make([]byte, len(a))
+	for i := 0; i < len(a); i++ {
+		ret[i] = a[i] | b[i]
+	}
+	return ret
+}
+
+func getBroadcast(cidr netip.Prefix) netip.Addr {
+	broadcast, _ := netip.AddrFromSlice(
+		orBytes(
+			cidr.Addr().AsSlice(),
+			flipBytes(prefixToMask(cidr).AsSlice()),
+		),
+	)
+	return broadcast
+}
+
 func addRoute(prefix netip.Prefix, gateway netroute.Addr) error {
 	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
 	if err != nil {

overlay/tun_linux.go

@@ -25,17 +25,14 @@ import (
 
 type tun struct {
 	io.ReadWriteCloser
-	fd            int
+	fd          int
-	Device        string
+	Device      string
-	vpnNetworks   []netip.Prefix
+	vpnNetworks []netip.Prefix
-	MaxMTU        int
+	MaxMTU      int
-	DefaultMTU    int
+	DefaultMTU  int
-	TXQueueLen    int
+	TXQueueLen  int
-	deviceIndex   int
+	deviceIndex int
-	ioctlFd       uintptr
+	ioctlFd     uintptr
-	enableVnetHdr bool
-	vnetHdrLen    int
-	queues        []*tunQueue
 
 	Routes                    atomic.Pointer[[]Route]
 	routeTree                 atomic.Pointer[bart.Table[routing.Gateways]]
@@ -68,90 +65,10 @@ type ifreqQLEN struct {
 	pad   [8]byte
 }
 
-const (
-	virtioNetHdrLen     = 12
-	tunDefaultMaxPacket = 65536
-)
-
-type tunQueue struct {
-	file          *os.File
-	fd            int
-	enableVnetHdr bool
-	vnetHdrLen    int
-	maxPacket     int
-	writeScratch  []byte
-	readScratch   []byte
-	l             *logrus.Logger
-}
-
-func newTunQueue(file *os.File, enableVnetHdr bool, vnetHdrLen, maxPacket int, l *logrus.Logger) *tunQueue {
-	if maxPacket <= 0 {
-		maxPacket = tunDefaultMaxPacket
-	}
-	q := &tunQueue{
-		file:          file,
-		fd:            int(file.Fd()),
-		enableVnetHdr: enableVnetHdr,
-		vnetHdrLen:    vnetHdrLen,
-		maxPacket:     maxPacket,
-		l:             l,
-	}
-	if enableVnetHdr {
-		q.growReadScratch(maxPacket)
-	}
-	return q
-}
-
-func (q *tunQueue) growReadScratch(packetSize int) {
-	needed := q.vnetHdrLen + packetSize
-	if needed < q.vnetHdrLen+DefaultMTU {
-		needed = q.vnetHdrLen + DefaultMTU
-	}
-	if q.readScratch == nil || cap(q.readScratch) < needed {
-		q.readScratch = make([]byte, needed)
-	} else {
-		q.readScratch = q.readScratch[:needed]
-	}
-}
-
-func (q *tunQueue) setMaxPacket(packet int) {
-	if packet <= 0 {
-		packet = DefaultMTU
-	}
-	q.maxPacket = packet
-	if q.enableVnetHdr {
-		q.growReadScratch(packet)
-	}
-}
-
-func configureVnetHdr(fd int, hdrLen int, l *logrus.Logger) error {
-	features, err := unix.IoctlGetInt(fd, unix.TUNGETFEATURES)
-	if err == nil && features&unix.IFF_VNET_HDR == 0 {
-		return fmt.Errorf("kernel does not support IFF_VNET_HDR")
-	}
-	if err := unix.IoctlSetInt(fd, unix.TUNSETVNETHDRSZ, hdrLen); err != nil {
-		return err
-	}
-	offload := unix.TUN_F_CSUM | unix.TUN_F_UFO
-	if err := unix.IoctlSetInt(fd, unix.TUNSETOFFLOAD, offload); err != nil {
-		if l != nil {
-			l.WithError(err).Warn("Failed to enable TUN offload features")
-		}
-	}
-	return nil
-}
-
 func newTunFromFd(c *config.C, l *logrus.Logger, deviceFd int, vpnNetworks []netip.Prefix) (*tun, error) {
 	file := os.NewFile(uintptr(deviceFd), "/dev/net/tun")
-	enableVnetHdr := c.GetBool("tun.enable_vnet_hdr", false)
-	if enableVnetHdr {
-		if err := configureVnetHdr(deviceFd, virtioNetHdrLen, l); err != nil {
-			l.WithError(err).Warn("Failed to configure VNET header support on provided tun fd; disabling")
-			enableVnetHdr = false
-		}
-	}
 
-	t, err := newTunGeneric(c, l, file, vpnNetworks, enableVnetHdr)
+	t, err := newTunGeneric(c, l, file, vpnNetworks)
 	if err != nil {
 		return nil, err
 	}
@@ -189,25 +106,14 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueu
 	if multiqueue {
 		req.Flags |= unix.IFF_MULTI_QUEUE
 	}
-	enableVnetHdr := c.GetBool("tun.enable_vnet_hdr", false)
-	if enableVnetHdr {
-		req.Flags |= unix.IFF_VNET_HDR
-	}
 	copy(req.Name[:], c.GetString("tun.dev", ""))
 	if err = ioctl(uintptr(fd), uintptr(unix.TUNSETIFF), uintptr(unsafe.Pointer(&req))); err != nil {
 		return nil, err
 	}
 	name := strings.Trim(string(req.Name[:]), "\x00")
 
-	if enableVnetHdr {
-		if err := configureVnetHdr(fd, virtioNetHdrLen, l); err != nil {
-			l.WithError(err).Warn("Failed to configure VNET header support on tun device; disabling")
-			enableVnetHdr = false
-		}
-	}
-
 	file := os.NewFile(uintptr(fd), "/dev/net/tun")
-	t, err := newTunGeneric(c, l, file, vpnNetworks, enableVnetHdr)
+	t, err := newTunGeneric(c, l, file, vpnNetworks)
 	if err != nil {
 		return nil, err
 	}
@@ -217,30 +123,21 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueu
 	return t, nil
 }
 
-func newTunGeneric(c *config.C, l *logrus.Logger, file *os.File, vpnNetworks []netip.Prefix, enableVnetHdr bool) (*tun, error) {
+func newTunGeneric(c *config.C, l *logrus.Logger, file *os.File, vpnNetworks []netip.Prefix) (*tun, error) {
-	queue := newTunQueue(file, enableVnetHdr, virtioNetHdrLen, tunDefaultMaxPacket, l)
 	t := &tun{
-		ReadWriteCloser:           queue,
+		ReadWriteCloser:           file,
 		fd:                        int(file.Fd()),
 		vpnNetworks:               vpnNetworks,
 		TXQueueLen:                c.GetInt("tun.tx_queue", 500),
 		useSystemRoutes:           c.GetBool("tun.use_system_route_table", false),
 		useSystemRoutesBufferSize: c.GetInt("tun.use_system_route_table_buffer_size", 0),
 		l:                         l,
-		enableVnetHdr:             enableVnetHdr,
-		vnetHdrLen:                virtioNetHdrLen,
-		queues:                    []*tunQueue{queue},
 	}
 
 	err := t.reload(c, true)
 	if err != nil {
 		return nil, err
 	}
-	if enableVnetHdr {
-		for _, q := range t.queues {
-			q.setMaxPacket(t.MaxMTU)
-		}
-	}
 
 	c.RegisterReloadCallback(func(c *config.C) {
 		err := t.reload(c, false)
@@ -283,11 +180,6 @@ func (t *tun) reload(c *config.C, initial bool) error {
 
 	t.MaxMTU = newMaxMTU
 	t.DefaultMTU = newDefaultMTU
-	if t.enableVnetHdr {
-		for _, q := range t.queues {
-			q.setMaxPacket(t.MaxMTU)
-		}
-	}
 
 	// Teach nebula how to handle the routes before establishing them in the system table
 	oldRoutes := t.Routes.Swap(&routes)
@@ -332,87 +224,14 @@ func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
 
 	var req ifReq
 	req.Flags = uint16(unix.IFF_TUN | unix.IFF_NO_PI | unix.IFF_MULTI_QUEUE)
-	if t.enableVnetHdr {
-		req.Flags |= unix.IFF_VNET_HDR
-	}
 	copy(req.Name[:], t.Device)
 	if err = ioctl(uintptr(fd), uintptr(unix.TUNSETIFF), uintptr(unsafe.Pointer(&req))); err != nil {
 		return nil, err
 	}
 
 	file := os.NewFile(uintptr(fd), "/dev/net/tun")
-	queue := newTunQueue(file, t.enableVnetHdr, t.vnetHdrLen, t.MaxMTU, t.l)
-	if t.enableVnetHdr {
-		if err := configureVnetHdr(fd, t.vnetHdrLen, t.l); err != nil {
-			queue.enableVnetHdr = false
-		}
-	}
-	t.queues = append(t.queues, queue)
 
-	return queue, nil
+	return file, nil
-}
-
-func (q *tunQueue) Read(p []byte) (int, error) {
-	if !q.enableVnetHdr {
-		return q.file.Read(p)
-	}
-
-	if len(p)+q.vnetHdrLen > cap(q.readScratch) {
-		q.growReadScratch(len(p))
-	}
-
-	buf := q.readScratch[:cap(q.readScratch)]
-	n, err := q.file.Read(buf)
-	if n <= 0 {
-		return n, err
-	}
-	if n < q.vnetHdrLen {
-		if err == nil {
-			err = io.ErrUnexpectedEOF
-		}
-		return 0, err
-	}
-
-	payload := buf[q.vnetHdrLen:n]
-	if len(payload) > len(p) {
-		copy(p, payload[:len(p)])
-		if err == nil {
-			err = io.ErrShortBuffer
-		}
-		return len(p), err
-	}
-	copy(p, payload)
-	return len(payload), err
-}
-
-func (q *tunQueue) Write(b []byte) (int, error) {
-	if !q.enableVnetHdr {
-		return unix.Write(q.fd, b)
-	}
-
-	total := q.vnetHdrLen + len(b)
-	if cap(q.writeScratch) < total {
-		q.writeScratch = make([]byte, total)
-	} else {
-		q.writeScratch = q.writeScratch[:total]
-	}
-
-	for i := 0; i < q.vnetHdrLen; i++ {
-		q.writeScratch[i] = 0
-	}
-	copy(q.writeScratch[q.vnetHdrLen:], b)
-
-	n, err := unix.Write(q.fd, q.writeScratch)
-	if n >= q.vnetHdrLen {
-		n -= q.vnetHdrLen
-	} else {
-		n = 0
-	}
-	return n, err
-}
-
-func (q *tunQueue) Close() error {
-	return q.file.Close()
 }
 
 func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {

overlay/tun_netbsd.go

@@ -4,12 +4,13 @@
 package overlay
 
 import (
-	"errors"
 	"fmt"
 	"io"
 	"net/netip"
 	"os"
+	"os/exec"
 	"regexp"
+	"strconv"
 	"sync/atomic"
 	"syscall"
 	"unsafe"
@@ -19,42 +20,11 @@ import (
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/routing"
 	"github.com/slackhq/nebula/util"
-	netroute "golang.org/x/net/route"
-	"golang.org/x/sys/unix"
 )
 
-const (
+type ifreqDestroy struct {
-	SIOCAIFADDR_IN6 = 0x8080696b
+	Name [16]byte
-	TUNSIFHEAD      = 0x80047442
+	pad  [16]byte
-	TUNSIFMODE      = 0x80047458
-)
-
-type ifreqAlias4 struct {
-	Name     [unix.IFNAMSIZ]byte
-	Addr     unix.RawSockaddrInet4
-	DstAddr  unix.RawSockaddrInet4
-	MaskAddr unix.RawSockaddrInet4
-}
-
-type ifreqAlias6 struct {
-	Name       [unix.IFNAMSIZ]byte
-	Addr       unix.RawSockaddrInet6
-	DstAddr    unix.RawSockaddrInet6
-	PrefixMask unix.RawSockaddrInet6
-	Flags      uint32
-	Lifetime   addrLifetime
-}
-
-type ifreq struct {
-	Name [unix.IFNAMSIZ]byte
-	data int
-}
-
-type addrLifetime struct {
-	Expire    uint64
-	Preferred uint64
-	Vltime    uint32
-	Pltime    uint32
 }
 
 type tun struct {
@@ -64,18 +34,40 @@ type tun struct {
 	Routes      atomic.Pointer[[]Route]
 	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
 	l           *logrus.Logger
-	f           *os.File
+
-	fd          int
+	io.ReadWriteCloser
 }
 
-var deviceNameRE = regexp.MustCompile(`^tun[0-9]+$`)
+func (t *tun) Close() error {
+	if t.ReadWriteCloser != nil {
+		if err := t.ReadWriteCloser.Close(); err != nil {
+			return err
+		}
+
+		s, err := syscall.Socket(syscall.AF_INET, syscall.SOCK_DGRAM, syscall.IPPROTO_IP)
+		if err != nil {
+			return err
+		}
+		defer syscall.Close(s)
+
+		ifreq := ifreqDestroy{Name: t.deviceBytes()}
+
+		err = ioctl(uintptr(s), syscall.SIOCIFDESTROY, uintptr(unsafe.Pointer(&ifreq)))
+
+		return err
+	}
+	return nil
+}
 
 func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (*tun, error) {
 	return nil, fmt.Errorf("newTunFromFd not supported in NetBSD")
 }
 
+var deviceNameRE = regexp.MustCompile(`^tun[0-9]+$`)
+
 func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (*tun, error) {
 	// Try to open tun device
+	var file *os.File
 	var err error
 	deviceName := c.GetString("tun.dev", "")
 	if deviceName == "" {
@@ -85,23 +77,17 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 		return nil, fmt.Errorf("a device name in the format of /dev/tunN must be specified")
 	}
 
-	fd, err := unix.Open("/dev/"+deviceName, os.O_RDWR, 0)
+	file, err = os.OpenFile("/dev/"+deviceName, os.O_RDWR, 0)
 	if err != nil {
 		return nil, err
 	}
 
-	err = unix.SetNonblock(fd, true)
-	if err != nil {
-		l.WithError(err).Warn("Failed to set the tun device as nonblocking")
-	}
-
 	t := &tun{
-		f:           os.NewFile(uintptr(fd), ""),
+		ReadWriteCloser: file,
-		fd:          fd,
+		Device:          deviceName,
-		Device:      deviceName,
+		vpnNetworks:     vpnNetworks,
-		vpnNetworks: vpnNetworks,
+		MTU:             c.GetInt("tun.mtu", DefaultMTU),
-		MTU:         c.GetInt("tun.mtu", DefaultMTU),
+		l:               l,
-		l:           l,
 	}
 
 	err = t.reload(c, true)
@@ -119,225 +105,40 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 	return t, nil
 }
 
-func (t *tun) Close() error {
-	if t.f != nil {
-		if err := t.f.Close(); err != nil {
-			return fmt.Errorf("error closing tun file: %w", err)
-		}
-
-		// t.f.Close should have handled it for us but let's be extra sure
-		_ = unix.Close(t.fd)
-
-		s, err := syscall.Socket(syscall.AF_INET, syscall.SOCK_DGRAM, syscall.IPPROTO_IP)
-		if err != nil {
-			return err
-		}
-		defer syscall.Close(s)
-
-		ifr := ifreq{Name: t.deviceBytes()}
-		err = ioctl(uintptr(s), syscall.SIOCIFDESTROY, uintptr(unsafe.Pointer(&ifr)))
-		return err
-	}
-	return nil
-}
-
-func (t *tun) Read(to []byte) (int, error) {
-	rc, err := t.f.SyscallConn()
-	if err != nil {
-		return 0, fmt.Errorf("failed to get syscall conn for tun: %w", err)
-	}
-
-	var errno syscall.Errno
-	var n uintptr
-	err = rc.Read(func(fd uintptr) bool {
-		// first 4 bytes is protocol family, in network byte order
-		head := [4]byte{}
-		iovecs := []syscall.Iovec{
-			{&head[0], 4},
-			{&to[0], uint64(len(to))},
-		}
-
-		n, _, errno = syscall.Syscall(syscall.SYS_READV, fd, uintptr(unsafe.Pointer(&iovecs[0])), uintptr(2))
-		if errno.Temporary() {
-			// We got an EAGAIN, EINTR, or EWOULDBLOCK, go again
-			return false
-		}
-		return true
-	})
-	if err != nil {
-		if err == syscall.EBADF || err.Error() == "use of closed file" {
-			// Go doesn't export poll.ErrFileClosing but happily reports it to us so here we are
-			// https://github.com/golang/go/blob/master/src/internal/poll/fd_poll_runtime.go#L121
-			return 0, os.ErrClosed
-		}
-		return 0, fmt.Errorf("failed to make read call for tun: %w", err)
-	}
-
-	if errno != 0 {
-		return 0, fmt.Errorf("failed to make inner read call for tun: %w", errno)
-	}
-
-	// fix bytes read number to exclude header
-	bytesRead := int(n)
-	if bytesRead < 0 {
-		return bytesRead, nil
-	} else if bytesRead < 4 {
-		return 0, nil
-	} else {
-		return bytesRead - 4, nil
-	}
-}
-
-// Write is only valid for single threaded use
-func (t *tun) Write(from []byte) (int, error) {
-	if len(from) <= 1 {
-		return 0, syscall.EIO
-	}
-
-	ipVer := from[0] >> 4
-	var head [4]byte
-	// first 4 bytes is protocol family, in network byte order
-	if ipVer == 4 {
-		head[3] = syscall.AF_INET
-	} else if ipVer == 6 {
-		head[3] = syscall.AF_INET6
-	} else {
-		return 0, fmt.Errorf("unable to determine IP version from packet")
-	}
-
-	rc, err := t.f.SyscallConn()
-	if err != nil {
-		return 0, err
-	}
-
-	var errno syscall.Errno
-	var n uintptr
-	err = rc.Write(func(fd uintptr) bool {
-		iovecs := []syscall.Iovec{
-			{&head[0], 4},
-			{&from[0], uint64(len(from))},
-		}
-
-		n, _, errno = syscall.Syscall(syscall.SYS_WRITEV, fd, uintptr(unsafe.Pointer(&iovecs[0])), uintptr(2))
-		// According to NetBSD documentation for TUN, writes will only return errors in which
-		// this packet will never be delivered so just go on living life.
-		return true
-	})
-	if err != nil {
-		return 0, err
-	}
-
-	if errno != 0 {
-		return 0, errno
-	}
-
-	return int(n) - 4, err
-}
-
 func (t *tun) addIp(cidr netip.Prefix) error {
-	if cidr.Addr().Is4() {
+	var err error
-		var req ifreqAlias4
-		req.Name = t.deviceBytes()
-		req.Addr = unix.RawSockaddrInet4{
-			Len:    unix.SizeofSockaddrInet4,
-			Family: unix.AF_INET,
-			Addr:   cidr.Addr().As4(),
-		}
-		req.DstAddr = unix.RawSockaddrInet4{
-			Len:    unix.SizeofSockaddrInet4,
-			Family: unix.AF_INET,
-			Addr:   cidr.Addr().As4(),
-		}
-		req.MaskAddr = unix.RawSockaddrInet4{
-			Len:    unix.SizeofSockaddrInet4,
-			Family: unix.AF_INET,
-			Addr:   prefixToMask(cidr).As4(),
-		}
 
-		s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
+	// TODO use syscalls instead of exec.Command
-		if err != nil {
+	cmd := exec.Command("/sbin/ifconfig", t.Device, cidr.String(), cidr.Addr().String())
-			return err
+	t.l.Debug("command: ", cmd.String())
-		}
+	if err = cmd.Run(); err != nil {
-		defer syscall.Close(s)
+		return fmt.Errorf("failed to run 'ifconfig': %s", err)
-
-		if err := ioctl(uintptr(s), unix.SIOCAIFADDR, uintptr(unsafe.Pointer(&req))); err != nil {
-			return fmt.Errorf("failed to set tun address %s: %s", cidr.Addr(), err)
-		}
-
-		return nil
 	}
 
-	if cidr.Addr().Is6() {
+	cmd = exec.Command("/sbin/route", "-n", "add", "-net", cidr.String(), cidr.Addr().String())
-		var req ifreqAlias6
+	t.l.Debug("command: ", cmd.String())
-		req.Name = t.deviceBytes()
+	if err = cmd.Run(); err != nil {
-		req.Addr = unix.RawSockaddrInet6{
+		return fmt.Errorf("failed to run 'route add': %s", err)
-			Len:    unix.SizeofSockaddrInet6,
-			Family: unix.AF_INET6,
-			Addr:   cidr.Addr().As16(),
-		}
-		req.PrefixMask = unix.RawSockaddrInet6{
-			Len:    unix.SizeofSockaddrInet6,
-			Family: unix.AF_INET6,
-			Addr:   prefixToMask(cidr).As16(),
-		}
-		req.Lifetime = addrLifetime{
-			Vltime: 0xffffffff,
-			Pltime: 0xffffffff,
-		}
-
-		s, err := unix.Socket(unix.AF_INET6, unix.SOCK_DGRAM, unix.IPPROTO_IP)
-		if err != nil {
-			return err
-		}
-		defer syscall.Close(s)
-
-		if err := ioctl(uintptr(s), SIOCAIFADDR_IN6, uintptr(unsafe.Pointer(&req))); err != nil {
-			return fmt.Errorf("failed to set tun address %s: %s", cidr.Addr().String(), err)
-		}
-		return nil
 	}
 
-	return fmt.Errorf("unknown address type %v", cidr)
+	cmd = exec.Command("/sbin/ifconfig", t.Device, "mtu", strconv.Itoa(t.MTU))
-}
+	t.l.Debug("command: ", cmd.String())
-
+	if err = cmd.Run(); err != nil {
-func (t *tun) Activate() error {
+		return fmt.Errorf("failed to run 'ifconfig': %s", err)
-	mode := int32(unix.IFF_BROADCAST)
-	err := ioctl(uintptr(t.fd), TUNSIFMODE, uintptr(unsafe.Pointer(&mode)))
-	if err != nil {
-		return fmt.Errorf("failed to set tun device mode: %w", err)
-	}
-
-	v := 1
-	err = ioctl(uintptr(t.fd), TUNSIFHEAD, uintptr(unsafe.Pointer(&v)))
-	if err != nil {
-		return fmt.Errorf("failed to set tun device head: %w", err)
-	}
-
-	err = t.doIoctlByName(unix.SIOCSIFMTU, uint32(t.MTU))
-	if err != nil {
-		return fmt.Errorf("failed to set tun mtu: %w", err)
-	}
-
-	for i := range t.vpnNetworks {
-		err = t.addIp(t.vpnNetworks[i])
-		if err != nil {
-			return err
-		}
 	}
 
+	// Unsafe path routes
 	return t.addRoutes(false)
 }
 
-func (t *tun) doIoctlByName(ctl uintptr, value uint32) error {
+func (t *tun) Activate() error {
-	s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
+	for i := range t.vpnNetworks {
-	if err != nil {
+		err := t.addIp(t.vpnNetworks[i])
-		return err
+		if err != nil {
+			return err
+		}
 	}
-	defer syscall.Close(s)
+	return nil
-
-	ir := ifreq{Name: t.deviceBytes(), data: int(value)}
-	err = ioctl(uintptr(s), ctl, uintptr(unsafe.Pointer(&ir)))
-	return err
 }
 
 func (t *tun) reload(c *config.C, initial bool) error {
@@ -396,23 +197,21 @@ func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
 
 func (t *tun) addRoutes(logErrors bool) error {
 	routes := *t.Routes.Load()
-
 	for _, r := range routes {
 		if len(r.Via) == 0 || !r.Install {
 			// We don't allow route MTUs so only install routes with a via
 			continue
 		}
 
-		err := addRoute(r.Cidr, t.vpnNetworks)
+		cmd := exec.Command("/sbin/route", "-n", "add", "-net", r.Cidr.String(), t.vpnNetworks[0].Addr().String())
-		if err != nil {
+		t.l.Debug("command: ", cmd.String())
-			retErr := util.NewContextualError("Failed to add route", map[string]any{"route": r}, err)
+		if err := cmd.Run(); err != nil {
+			retErr := util.NewContextualError("failed to run 'route add' for unsafe_route", map[string]any{"route": r}, err)
 			if logErrors {
 				retErr.Log(t.l)
 			} else {
 				return retErr
 			}
-		} else {
-			t.l.WithField("route", r).Info("Added route")
 		}
 	}
 
@@ -425,8 +224,10 @@ func (t *tun) removeRoutes(routes []Route) error {
 			continue
 		}
 
-		err := delRoute(r.Cidr, t.vpnNetworks)
+		//TODO: CERT-V2 is this right?
-		if err != nil {
+		cmd := exec.Command("/sbin/route", "-n", "delete", "-net", r.Cidr.String(), t.vpnNetworks[0].Addr().String())
+		t.l.Debug("command: ", cmd.String())
+		if err := cmd.Run(); err != nil {
 			t.l.WithError(err).WithField("route", r).Error("Failed to remove route")
 		} else {
 			t.l.WithField("route", r).Info("Removed route")
@@ -441,109 +242,3 @@ func (t *tun) deviceBytes() (o [16]byte) {
 	}
 	return
 }
-
-func addRoute(prefix netip.Prefix, gateways []netip.Prefix) error {
-	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
-	if err != nil {
-		return fmt.Errorf("unable to create AF_ROUTE socket: %v", err)
-	}
-	defer unix.Close(sock)
-
-	route := &netroute.RouteMessage{
-		Version: unix.RTM_VERSION,
-		Type:    unix.RTM_ADD,
-		Flags:   unix.RTF_UP | unix.RTF_GATEWAY,
-		Seq:     1,
-	}
-
-	if prefix.Addr().Is4() {
-		gw, err := selectGateway(prefix, gateways)
-		if err != nil {
-			return err
-		}
-		route.Addrs = []netroute.Addr{
-			unix.RTAX_DST:     &netroute.Inet4Addr{IP: prefix.Masked().Addr().As4()},
-			unix.RTAX_NETMASK: &netroute.Inet4Addr{IP: prefixToMask(prefix).As4()},
-			unix.RTAX_GATEWAY: &netroute.Inet4Addr{IP: gw.Addr().As4()},
-		}
-	} else {
-		gw, err := selectGateway(prefix, gateways)
-		if err != nil {
-			return err
-		}
-		route.Addrs = []netroute.Addr{
-			unix.RTAX_DST:     &netroute.Inet6Addr{IP: prefix.Masked().Addr().As16()},
-			unix.RTAX_NETMASK: &netroute.Inet6Addr{IP: prefixToMask(prefix).As16()},
-			unix.RTAX_GATEWAY: &netroute.Inet6Addr{IP: gw.Addr().As16()},
-		}
-	}
-
-	data, err := route.Marshal()
-	if err != nil {
-		return fmt.Errorf("failed to create route.RouteMessage: %w", err)
-	}
-
-	_, err = unix.Write(sock, data[:])
-	if err != nil {
-		if errors.Is(err, unix.EEXIST) {
-			// Try to do a change
-			route.Type = unix.RTM_CHANGE
-			data, err = route.Marshal()
-			if err != nil {
-				return fmt.Errorf("failed to create route.RouteMessage for change: %w", err)
-			}
-			_, err = unix.Write(sock, data[:])
-			return err
-		}
-		return fmt.Errorf("failed to write route.RouteMessage to socket: %w", err)
-	}
-
-	return nil
-}
-
-func delRoute(prefix netip.Prefix, gateways []netip.Prefix) error {
-	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
-	if err != nil {
-		return fmt.Errorf("unable to create AF_ROUTE socket: %v", err)
-	}
-	defer unix.Close(sock)
-
-	route := netroute.RouteMessage{
-		Version: unix.RTM_VERSION,
-		Type:    unix.RTM_DELETE,
-		Seq:     1,
-	}
-
-	if prefix.Addr().Is4() {
-		gw, err := selectGateway(prefix, gateways)
-		if err != nil {
-			return err
-		}
-		route.Addrs = []netroute.Addr{
-			unix.RTAX_DST:     &netroute.Inet4Addr{IP: prefix.Masked().Addr().As4()},
-			unix.RTAX_NETMASK: &netroute.Inet4Addr{IP: prefixToMask(prefix).As4()},
-			unix.RTAX_GATEWAY: &netroute.Inet4Addr{IP: gw.Addr().As4()},
-		}
-	} else {
-		gw, err := selectGateway(prefix, gateways)
-		if err != nil {
-			return err
-		}
-		route.Addrs = []netroute.Addr{
-			unix.RTAX_DST:     &netroute.Inet6Addr{IP: prefix.Masked().Addr().As16()},
-			unix.RTAX_NETMASK: &netroute.Inet6Addr{IP: prefixToMask(prefix).As16()},
-			unix.RTAX_GATEWAY: &netroute.Inet6Addr{IP: gw.Addr().As16()},
-		}
-	}
-
-	data, err := route.Marshal()
-	if err != nil {
-		return fmt.Errorf("failed to create route.RouteMessage: %w", err)
-	}
-	_, err = unix.Write(sock, data[:])
-	if err != nil {
-		return fmt.Errorf("failed to write route.RouteMessage to socket: %w", err)
-	}
-
-	return nil
-}

overlay/tun_openbsd.go

@@ -4,50 +4,23 @@
 package overlay
 
 import (
-	"errors"
 	"fmt"
 	"io"
 	"net/netip"
 	"os"
+	"os/exec"
 	"regexp"
+	"strconv"
 	"sync/atomic"
 	"syscall"
-	"unsafe"
 
 	"github.com/gaissmai/bart"
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/routing"
 	"github.com/slackhq/nebula/util"
-	netroute "golang.org/x/net/route"
-	"golang.org/x/sys/unix"
 )
 
-const (
-	SIOCAIFADDR_IN6 = 0x8080691a
-)
-
-type ifreqAlias4 struct {
-	Name     [unix.IFNAMSIZ]byte
-	Addr     unix.RawSockaddrInet4
-	DstAddr  unix.RawSockaddrInet4
-	MaskAddr unix.RawSockaddrInet4
-}
-
-type ifreqAlias6 struct {
-	Name       [unix.IFNAMSIZ]byte
-	Addr       unix.RawSockaddrInet6
-	DstAddr    unix.RawSockaddrInet6
-	PrefixMask unix.RawSockaddrInet6
-	Flags      uint32
-	Lifetime   [2]uint32
-}
-
-type ifreq struct {
-	Name [unix.IFNAMSIZ]byte
-	data int
-}
-
 type tun struct {
 	Device      string
 	vpnNetworks []netip.Prefix
@@ -55,46 +28,48 @@ type tun struct {
 	Routes      atomic.Pointer[[]Route]
 	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
 	l           *logrus.Logger
-	f           *os.File
+
-	fd          int
+	io.ReadWriteCloser
+
 	// cache out buffer since we need to prepend 4 bytes for tun metadata
 	out []byte
 }
 
-var deviceNameRE = regexp.MustCompile(`^tun[0-9]+$`)
+func (t *tun) Close() error {
+	if t.ReadWriteCloser != nil {
+		return t.ReadWriteCloser.Close()
+	}
 
-func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (*tun, error) {
+	return nil
-	return nil, fmt.Errorf("newTunFromFd not supported in openbsd")
 }
 
+func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (*tun, error) {
+	return nil, fmt.Errorf("newTunFromFd not supported in OpenBSD")
+}
+
+var deviceNameRE = regexp.MustCompile(`^tun[0-9]+$`)
+
 func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (*tun, error) {
-	// Try to open tun device
-	var err error
 	deviceName := c.GetString("tun.dev", "")
 	if deviceName == "" {
-		return nil, fmt.Errorf("a device name in the format of /dev/tunN must be specified")
+		return nil, fmt.Errorf("a device name in the format of tunN must be specified")
-	}
-	if !deviceNameRE.MatchString(deviceName) {
-		return nil, fmt.Errorf("a device name in the format of /dev/tunN must be specified")
 	}
 
-	fd, err := unix.Open("/dev/"+deviceName, os.O_RDWR, 0)
+	if !deviceNameRE.MatchString(deviceName) {
+		return nil, fmt.Errorf("a device name in the format of tunN must be specified")
+	}
+
+	file, err := os.OpenFile("/dev/"+deviceName, os.O_RDWR, 0)
 	if err != nil {
 		return nil, err
 	}
 
-	err = unix.SetNonblock(fd, true)
-	if err != nil {
-		l.WithError(err).Warn("Failed to set the tun device as nonblocking")
-	}
-
 	t := &tun{
-		f:           os.NewFile(uintptr(fd), ""),
+		ReadWriteCloser: file,
-		fd:          fd,
+		Device:          deviceName,
-		Device:      deviceName,
+		vpnNetworks:     vpnNetworks,
-		vpnNetworks: vpnNetworks,
+		MTU:             c.GetInt("tun.mtu", DefaultMTU),
-		MTU:         c.GetInt("tun.mtu", DefaultMTU),
+		l:               l,
-		l:           l,
 	}
 
 	err = t.reload(c, true)
@@ -112,154 +87,6 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 	return t, nil
 }
 
-func (t *tun) Close() error {
-	if t.f != nil {
-		if err := t.f.Close(); err != nil {
-			return fmt.Errorf("error closing tun file: %w", err)
-		}
-
-		// t.f.Close should have handled it for us but let's be extra sure
-		_ = unix.Close(t.fd)
-	}
-	return nil
-}
-
-func (t *tun) Read(to []byte) (int, error) {
-	buf := make([]byte, len(to)+4)
-
-	n, err := t.f.Read(buf)
-
-	copy(to, buf[4:])
-	return n - 4, err
-}
-
-// Write is only valid for single threaded use
-func (t *tun) Write(from []byte) (int, error) {
-	buf := t.out
-	if cap(buf) < len(from)+4 {
-		buf = make([]byte, len(from)+4)
-		t.out = buf
-	}
-	buf = buf[:len(from)+4]
-
-	if len(from) == 0 {
-		return 0, syscall.EIO
-	}
-
-	// Determine the IP Family for the NULL L2 Header
-	ipVer := from[0] >> 4
-	if ipVer == 4 {
-		buf[3] = syscall.AF_INET
-	} else if ipVer == 6 {
-		buf[3] = syscall.AF_INET6
-	} else {
-		return 0, fmt.Errorf("unable to determine IP version from packet")
-	}
-
-	copy(buf[4:], from)
-
-	n, err := t.f.Write(buf)
-	return n - 4, err
-}
-
-func (t *tun) addIp(cidr netip.Prefix) error {
-	if cidr.Addr().Is4() {
-		var req ifreqAlias4
-		req.Name = t.deviceBytes()
-		req.Addr = unix.RawSockaddrInet4{
-			Len:    unix.SizeofSockaddrInet4,
-			Family: unix.AF_INET,
-			Addr:   cidr.Addr().As4(),
-		}
-		req.DstAddr = unix.RawSockaddrInet4{
-			Len:    unix.SizeofSockaddrInet4,
-			Family: unix.AF_INET,
-			Addr:   cidr.Addr().As4(),
-		}
-		req.MaskAddr = unix.RawSockaddrInet4{
-			Len:    unix.SizeofSockaddrInet4,
-			Family: unix.AF_INET,
-			Addr:   prefixToMask(cidr).As4(),
-		}
-
-		s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
-		if err != nil {
-			return err
-		}
-		defer syscall.Close(s)
-
-		if err := ioctl(uintptr(s), unix.SIOCAIFADDR, uintptr(unsafe.Pointer(&req))); err != nil {
-			return fmt.Errorf("failed to set tun address %s: %s", cidr.Addr(), err)
-		}
-
-		err = addRoute(cidr, t.vpnNetworks)
-		if err != nil {
-			return fmt.Errorf("failed to set route for vpn network %v: %w", cidr, err)
-		}
-
-		return nil
-	}
-
-	if cidr.Addr().Is6() {
-		var req ifreqAlias6
-		req.Name = t.deviceBytes()
-		req.Addr = unix.RawSockaddrInet6{
-			Len:    unix.SizeofSockaddrInet6,
-			Family: unix.AF_INET6,
-			Addr:   cidr.Addr().As16(),
-		}
-		req.PrefixMask = unix.RawSockaddrInet6{
-			Len:    unix.SizeofSockaddrInet6,
-			Family: unix.AF_INET6,
-			Addr:   prefixToMask(cidr).As16(),
-		}
-		req.Lifetime[0] = 0xffffffff
-		req.Lifetime[1] = 0xffffffff
-
-		s, err := unix.Socket(unix.AF_INET6, unix.SOCK_DGRAM, unix.IPPROTO_IP)
-		if err != nil {
-			return err
-		}
-		defer syscall.Close(s)
-
-		if err := ioctl(uintptr(s), SIOCAIFADDR_IN6, uintptr(unsafe.Pointer(&req))); err != nil {
-			return fmt.Errorf("failed to set tun address %s: %s", cidr.Addr().String(), err)
-		}
-
-		return nil
-	}
-
-	return fmt.Errorf("unknown address type %v", cidr)
-}
-
-func (t *tun) Activate() error {
-	err := t.doIoctlByName(unix.SIOCSIFMTU, uint32(t.MTU))
-	if err != nil {
-		return fmt.Errorf("failed to set tun mtu: %w", err)
-	}
-
-	for i := range t.vpnNetworks {
-		err = t.addIp(t.vpnNetworks[i])
-		if err != nil {
-			return err
-		}
-	}
-
-	return t.addRoutes(false)
-}
-
-func (t *tun) doIoctlByName(ctl uintptr, value uint32) error {
-	s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
-	if err != nil {
-		return err
-	}
-	defer syscall.Close(s)
-
-	ir := ifreq{Name: t.deviceBytes(), data: int(value)}
-	err = ioctl(uintptr(s), ctl, uintptr(unsafe.Pointer(&ir)))
-	return err
-}
-
 func (t *tun) reload(c *config.C, initial bool) error {
 	change, routes, err := getAllRoutesFromConfig(c, t.vpnNetworks, initial)
 	if err != nil {
@@ -297,42 +124,63 @@ func (t *tun) reload(c *config.C, initial bool) error {
 	return nil
 }
 
+func (t *tun) addIp(cidr netip.Prefix) error {
+	var err error
+	// TODO use syscalls instead of exec.Command
+	cmd := exec.Command("/sbin/ifconfig", t.Device, cidr.String(), cidr.Addr().String())
+	t.l.Debug("command: ", cmd.String())
+	if err = cmd.Run(); err != nil {
+		return fmt.Errorf("failed to run 'ifconfig': %s", err)
+	}
+
+	cmd = exec.Command("/sbin/ifconfig", t.Device, "mtu", strconv.Itoa(t.MTU))
+	t.l.Debug("command: ", cmd.String())
+	if err = cmd.Run(); err != nil {
+		return fmt.Errorf("failed to run 'ifconfig': %s", err)
+	}
+
+	cmd = exec.Command("/sbin/route", "-n", "add", "-inet", cidr.String(), cidr.Addr().String())
+	t.l.Debug("command: ", cmd.String())
+	if err = cmd.Run(); err != nil {
+		return fmt.Errorf("failed to run 'route add': %s", err)
+	}
+
+	// Unsafe path routes
+	return t.addRoutes(false)
+}
+
+func (t *tun) Activate() error {
+	for i := range t.vpnNetworks {
+		err := t.addIp(t.vpnNetworks[i])
+		if err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
 func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
 	r, _ := t.routeTree.Load().Lookup(ip)
 	return r
 }
 
-func (t *tun) Networks() []netip.Prefix {
-	return t.vpnNetworks
-}
-
-func (t *tun) Name() string {
-	return t.Device
-}
-
-func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
-	return nil, fmt.Errorf("TODO: multiqueue not implemented for openbsd")
-}
-
 func (t *tun) addRoutes(logErrors bool) error {
 	routes := *t.Routes.Load()
-
 	for _, r := range routes {
 		if len(r.Via) == 0 || !r.Install {
 			// We don't allow route MTUs so only install routes with a via
 			continue
 		}
-
+		//TODO: CERT-V2 is this right?
-		err := addRoute(r.Cidr, t.vpnNetworks)
+		cmd := exec.Command("/sbin/route", "-n", "add", "-inet", r.Cidr.String(), t.vpnNetworks[0].Addr().String())
-		if err != nil {
+		t.l.Debug("command: ", cmd.String())
-			retErr := util.NewContextualError("Failed to add route", map[string]any{"route": r}, err)
+		if err := cmd.Run(); err != nil {
+			retErr := util.NewContextualError("failed to run 'route add' for unsafe_route", map[string]any{"route": r}, err)
 			if logErrors {
 				retErr.Log(t.l)
 			} else {
 				return retErr
 			}
-		} else {
-			t.l.WithField("route", r).Info("Added route")
 		}
 	}
 
@@ -344,9 +192,10 @@ func (t *tun) removeRoutes(routes []Route) error {
 		if !r.Install {
 			continue
 		}
-
+		//TODO: CERT-V2 is this right?
-		err := delRoute(r.Cidr, t.vpnNetworks)
+		cmd := exec.Command("/sbin/route", "-n", "delete", "-inet", r.Cidr.String(), t.vpnNetworks[0].Addr().String())
-		if err != nil {
+		t.l.Debug("command: ", cmd.String())
+		if err := cmd.Run(); err != nil {
 			t.l.WithError(err).WithField("route", r).Error("Failed to remove route")
 		} else {
 			t.l.WithField("route", r).Info("Removed route")
@@ -355,115 +204,52 @@ func (t *tun) removeRoutes(routes []Route) error {
 	return nil
 }
 
-func (t *tun) deviceBytes() (o [16]byte) {
+func (t *tun) Networks() []netip.Prefix {
-	for i, c := range t.Device {
+	return t.vpnNetworks
-		o[i] = byte(c)
-	}
-	return
 }
 
-func addRoute(prefix netip.Prefix, gateways []netip.Prefix) error {
+func (t *tun) Name() string {
-	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
+	return t.Device
-	if err != nil {
+}
-		return fmt.Errorf("unable to create AF_ROUTE socket: %v", err)
-	}
-	defer unix.Close(sock)
 
-	route := &netroute.RouteMessage{
+func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
-		Version: unix.RTM_VERSION,
+	return nil, fmt.Errorf("TODO: multiqueue not implemented for freebsd")
-		Type:    unix.RTM_ADD,
+}
-		Flags:   unix.RTF_UP | unix.RTF_GATEWAY,
+
-		Seq:     1,
+func (t *tun) Read(to []byte) (int, error) {
+	buf := make([]byte, len(to)+4)
+
+	n, err := t.ReadWriteCloser.Read(buf)
+
+	copy(to, buf[4:])
+	return n - 4, err
+}
+
+// Write is only valid for single threaded use
+func (t *tun) Write(from []byte) (int, error) {
+	buf := t.out
+	if cap(buf) < len(from)+4 {
+		buf = make([]byte, len(from)+4)
+		t.out = buf
+	}
+	buf = buf[:len(from)+4]
+
+	if len(from) == 0 {
+		return 0, syscall.EIO
 	}
 
-	if prefix.Addr().Is4() {
+	// Determine the IP Family for the NULL L2 Header
-		gw, err := selectGateway(prefix, gateways)
+	ipVer := from[0] >> 4
-		if err != nil {
+	if ipVer == 4 {
-			return err
+		buf[3] = syscall.AF_INET
-		}
+	} else if ipVer == 6 {
-		route.Addrs = []netroute.Addr{
+		buf[3] = syscall.AF_INET6
-			unix.RTAX_DST:     &netroute.Inet4Addr{IP: prefix.Masked().Addr().As4()},
-			unix.RTAX_NETMASK: &netroute.Inet4Addr{IP: prefixToMask(prefix).As4()},
-			unix.RTAX_GATEWAY: &netroute.Inet4Addr{IP: gw.Addr().As4()},
-		}
 	} else {
-		gw, err := selectGateway(prefix, gateways)
+		return 0, fmt.Errorf("unable to determine IP version from packet")
-		if err != nil {
-			return err
-		}
-		route.Addrs = []netroute.Addr{
-			unix.RTAX_DST:     &netroute.Inet6Addr{IP: prefix.Masked().Addr().As16()},
-			unix.RTAX_NETMASK: &netroute.Inet6Addr{IP: prefixToMask(prefix).As16()},
-			unix.RTAX_GATEWAY: &netroute.Inet6Addr{IP: gw.Addr().As16()},
-		}
 	}
 
-	data, err := route.Marshal()
+	copy(buf[4:], from)
-	if err != nil {
-		return fmt.Errorf("failed to create route.RouteMessage: %w", err)
-	}
 
-	_, err = unix.Write(sock, data[:])
+	n, err := t.ReadWriteCloser.Write(buf)
-	if err != nil {
+	return n - 4, err
-		if errors.Is(err, unix.EEXIST) {
-			// Try to do a change
-			route.Type = unix.RTM_CHANGE
-			data, err = route.Marshal()
-			if err != nil {
-				return fmt.Errorf("failed to create route.RouteMessage for change: %w", err)
-			}
-			_, err = unix.Write(sock, data[:])
-			return err
-		}
-		return fmt.Errorf("failed to write route.RouteMessage to socket: %w", err)
-	}
-
-	return nil
-}
-
-func delRoute(prefix netip.Prefix, gateways []netip.Prefix) error {
-	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
-	if err != nil {
-		return fmt.Errorf("unable to create AF_ROUTE socket: %v", err)
-	}
-	defer unix.Close(sock)
-
-	route := netroute.RouteMessage{
-		Version: unix.RTM_VERSION,
-		Type:    unix.RTM_DELETE,
-		Seq:     1,
-	}
-
-	if prefix.Addr().Is4() {
-		gw, err := selectGateway(prefix, gateways)
-		if err != nil {
-			return err
-		}
-		route.Addrs = []netroute.Addr{
-			unix.RTAX_DST:     &netroute.Inet4Addr{IP: prefix.Masked().Addr().As4()},
-			unix.RTAX_NETMASK: &netroute.Inet4Addr{IP: prefixToMask(prefix).As4()},
-			unix.RTAX_GATEWAY: &netroute.Inet4Addr{IP: gw.Addr().As4()},
-		}
-	} else {
-		gw, err := selectGateway(prefix, gateways)
-		if err != nil {
-			return err
-		}
-		route.Addrs = []netroute.Addr{
-			unix.RTAX_DST:     &netroute.Inet6Addr{IP: prefix.Masked().Addr().As16()},
-			unix.RTAX_NETMASK: &netroute.Inet6Addr{IP: prefixToMask(prefix).As16()},
-			unix.RTAX_GATEWAY: &netroute.Inet6Addr{IP: gw.Addr().As16()},
-		}
-	}
-
-	data, err := route.Marshal()
-	if err != nil {
-		return fmt.Errorf("failed to create route.RouteMessage: %w", err)
-	}
-	_, err = unix.Write(sock, data[:])
-	if err != nil {
-		return fmt.Errorf("failed to write route.RouteMessage to socket: %w", err)
-	}
-
-	return nil
 }

packet/packet.go

@@ -0,0 +1,12 @@
+package packet
+
+import "net/netip"
+
+type Packet struct {
+	Payload []byte
+	Addr    netip.AddrPort
+}
+
+func New() *Packet {
+	return &Packet{Payload: make([]byte, 9001)}
+}

service/service.go

@@ -44,7 +44,10 @@ type Service struct {
 }
 
 func New(control *nebula.Control) (*Service, error) {
-	control.Start()
+	wait, err := control.Start()
+	if err != nil {
+		return nil, err
+	}
 
 	ctx := control.Context()
 	eg, ctx := errgroup.WithContext(ctx)
@@ -141,6 +144,12 @@ func New(control *nebula.Control) (*Service, error) {
 		}
 	})
 
+	// Add the nebula wait function to the group
+	eg.Go(func() error {
+		wait()
+		return nil
+	})
+
 	return &s, nil
 }
 

udp/config.go

@@ -1,16 +0,0 @@
-package udp
-
-import "sync/atomic"
-
-var disableUDPCsum atomic.Bool
-
-// SetDisableUDPCsum controls whether IPv4 UDP sockets opt out of kernel
-// checksum calculation via SO_NO_CHECK. Only applicable on platforms that
-// support the option (Linux). IPv6 always keeps the checksum enabled.
-func SetDisableUDPCsum(disable bool) {
-	disableUDPCsum.Store(disable)
-}
-
-func udpChecksumDisabled() bool {
-	return disableUDPCsum.Load()
-}

udp/conn.go

@@ -11,18 +11,23 @@ const MTU = 9001
 type EncReader func(
 	addr netip.AddrPort,
 	payload []byte,
-	release func(),
 )
 
 type Conn interface {
 	Rebind() error
 	LocalAddr() (netip.AddrPort, error)
-	ListenOut(r EncReader)
+	ListenOut(r EncReader) error
 	WriteTo(b []byte, addr netip.AddrPort) error
+	WriteBatch(pkts []BatchPacket) (int, error)
 	ReloadConfig(c *config.C)
 	Close() error
 }
 
+type BatchPacket struct {
+	Payload []byte
+	Addr    netip.AddrPort
+}
+
 type NoopConn struct{}
 
 func (NoopConn) Rebind() error {
@@ -31,12 +36,15 @@ func (NoopConn) Rebind() error {
 func (NoopConn) LocalAddr() (netip.AddrPort, error) {
 	return netip.AddrPort{}, nil
 }
-func (NoopConn) ListenOut(_ EncReader) {
+func (NoopConn) ListenOut(_ EncReader) error {
-	return
+	return nil
 }
 func (NoopConn) WriteTo(_ []byte, _ netip.AddrPort) error {
 	return nil
 }
+func (NoopConn) WriteBatch(_ []BatchPacket) (int, error) {
+	return 0, nil
+}
 func (NoopConn) ReloadConfig(_ *config.C) {
 	return
 }

udp/msghdr_helper_linux_32.go

@@ -1,25 +0,0 @@
-//go:build linux && (386 || amd64p32 || arm || mips || mipsle) && !android && !e2e_testing
-// +build linux
-// +build 386 amd64p32 arm mips mipsle
-// +build !android
-// +build !e2e_testing
-
-package udp
-
-import "golang.org/x/sys/unix"
-
-func controllen(n int) uint32 {
-	return uint32(n)
-}
-
-func setCmsgLen(h *unix.Cmsghdr, n int) {
-	h.Len = uint32(unix.CmsgLen(n))
-}
-
-func setIovecLen(v *unix.Iovec, n int) {
-	v.Len = uint32(n)
-}
-
-func setMsghdrIovlen(m *unix.Msghdr, n int) {
-	m.Iovlen = uint32(n)
-}

udp/msghdr_helper_linux_64.go

@@ -1,25 +0,0 @@
-//go:build linux && (amd64 || arm64 || ppc64 || ppc64le || mips64 || mips64le || s390x || riscv64 || loong64) && !android && !e2e_testing
-// +build linux
-// +build amd64 arm64 ppc64 ppc64le mips64 mips64le s390x riscv64 loong64
-// +build !android
-// +build !e2e_testing
-
-package udp
-
-import "golang.org/x/sys/unix"
-
-func controllen(n int) uint64 {
-	return uint64(n)
-}
-
-func setCmsgLen(h *unix.Cmsghdr, n int) {
-	h.Len = uint64(unix.CmsgLen(n))
-}
-
-func setIovecLen(v *unix.Iovec, n int) {
-	v.Len = uint64(n)
-}
-
-func setMsghdrIovlen(m *unix.Msghdr, n int) {
-	m.Iovlen = uint64(n)
-}

udp/sendmmsg_linux_32.go

@@ -1,25 +0,0 @@
-//go:build linux && (386 || amd64p32 || arm || mips || mipsle) && !android && !e2e_testing
-
-package udp
-
-import (
-	"unsafe"
-
-	"golang.org/x/sys/unix"
-)
-
-type linuxMmsgHdr struct {
-	Hdr unix.Msghdr
-	Len uint32
-}
-
-func sendmmsg(fd int, hdrs []linuxMmsgHdr, flags int) (int, error) {
-	if len(hdrs) == 0 {
-		return 0, nil
-	}
-	n, _, errno := unix.Syscall6(unix.SYS_SENDMMSG, uintptr(fd), uintptr(unsafe.Pointer(&hdrs[0])), uintptr(len(hdrs)), uintptr(flags), 0, 0)
-	if errno != 0 {
-		return int(n), errno
-	}
-	return int(n), nil
-}

udp/sendmmsg_linux_64.go

@@ -1,26 +0,0 @@
-//go:build linux && (amd64 || arm64 || ppc64 || ppc64le || mips64 || mips64le || s390x || riscv64 || loong64) && !android && !e2e_testing
-
-package udp
-
-import (
-	"unsafe"
-
-	"golang.org/x/sys/unix"
-)
-
-type linuxMmsgHdr struct {
-	Hdr unix.Msghdr
-	Len uint32
-	_   uint32
-}
-
-func sendmmsg(fd int, hdrs []linuxMmsgHdr, flags int) (int, error) {
-	if len(hdrs) == 0 {
-		return 0, nil
-	}
-	n, _, errno := unix.Syscall6(unix.SYS_SENDMMSG, uintptr(fd), uintptr(unsafe.Pointer(&hdrs[0])), uintptr(len(hdrs)), uintptr(flags), 0, 0)
-	if errno != 0 {
-		return int(n), errno
-	}
-	return int(n), nil
-}

udp/udp_darwin.go

@@ -140,6 +140,17 @@ func (u *StdConn) WriteTo(b []byte, ap netip.AddrPort) error {
 	}
 }
 
+func (u *StdConn) WriteBatch(pkts []BatchPacket) (int, error) {
+	sent := 0
+	for _, pkt := range pkts {
+		if err := u.WriteTo(pkt.Payload, pkt.Addr); err != nil {
+			return sent, err
+		}
+		sent++
+	}
+	return sent, nil
+}
+
 func (u *StdConn) LocalAddr() (netip.AddrPort, error) {
 	a := u.UDPConn.LocalAddr()
 
@@ -165,7 +176,7 @@ func NewUDPStatsEmitter(udpConns []Conn) func() {
 	return func() {}
 }
 
-func (u *StdConn) ListenOut(r EncReader) {
+func (u *StdConn) ListenOut(r EncReader) error {
 	buffer := make([]byte, MTU)
 
 	for {
@@ -174,14 +185,17 @@ func (u *StdConn) ListenOut(r EncReader) {
 		if err != nil {
 			if errors.Is(err, net.ErrClosed) {
 				u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
-				return
+				return err
 			}
 
 			u.l.WithError(err).Error("unexpected udp socket receive error")
+			continue
 		}
 
-		r(netip.AddrPortFrom(rua.Addr().Unmap(), rua.Port()), buffer[:n], nil)
+		r(netip.AddrPortFrom(rua.Addr().Unmap(), rua.Port()), buffer[:n])
 	}
+
+	return nil
 }
 
 func (u *StdConn) Rebind() error {

udp/udp_generic.go

@@ -42,6 +42,17 @@ func (u *GenericConn) WriteTo(b []byte, addr netip.AddrPort) error {
 	return err
 }
 
+func (u *GenericConn) WriteBatch(pkts []BatchPacket) (int, error) {
+	sent := 0
+	for _, pkt := range pkts {
+		if err := u.WriteTo(pkt.Payload, pkt.Addr); err != nil {
+			return sent, err
+		}
+		sent++
+	}
+	return sent, nil
+}
+
 func (u *GenericConn) LocalAddr() (netip.AddrPort, error) {
 	a := u.UDPConn.LocalAddr()
 
@@ -71,17 +82,16 @@ type rawMessage struct {
 	Len uint32
 }
 
-func (u *GenericConn) ListenOut(r EncReader) {
+func (u *GenericConn) ListenOut(r EncReader) error {
 	buffer := make([]byte, MTU)
 
 	for {
 		// Just read one packet at a time
 		n, rua, err := u.ReadFromUDPAddrPort(buffer)
 		if err != nil {
-			u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
+			return err
-			return
 		}
 
-		r(netip.AddrPortFrom(rua.Addr().Unmap(), rua.Port()), buffer[:n], nil)
+		r(netip.AddrPortFrom(rua.Addr().Unmap(), rua.Port()), buffer[:n])
 	}
 }

udp/udp_linux_32.go

@@ -30,29 +30,20 @@ type rawMessage struct {
 	Len uint32
 }
 
-func (u *StdConn) PrepareRawMessages(n int) ([]rawMessage, [][]byte, [][]byte, [][]byte) {
+func (u *StdConn) PrepareRawMessages(n int, bufSize int) ([]rawMessage, [][]byte, [][]byte) {
-	controlLen := int(u.controlLen.Load())
+	if bufSize <= 0 {
-
+		bufSize = MTU
+	}
 	msgs := make([]rawMessage, n)
 	buffers := make([][]byte, n)
 	names := make([][]byte, n)
 
-	var controls [][]byte
-	if controlLen > 0 {
-		controls = make([][]byte, n)
-	}
-
 	for i := range msgs {
-		size := int(u.groBufSize.Load())
+		buffers[i] = make([]byte, bufSize)
-		if size < MTU {
-			size = MTU
-		}
-		buf := u.borrowRxBuffer(size)
-		buffers[i] = buf
 		names[i] = make([]byte, unix.SizeofSockaddrInet6)
 
 		vs := []iovec{
-			{Base: &buf[0], Len: uint32(len(buf))},
+			{Base: &buffers[i][0], Len: uint32(len(buffers[i]))},
 		}
 
 		msgs[i].Hdr.Iov = &vs[0]
@@ -60,22 +51,39 @@ func (u *StdConn) PrepareRawMessages(n int) ([]rawMessage, [][]byte, [][]byte, [
 
 		msgs[i].Hdr.Name = &names[i][0]
 		msgs[i].Hdr.Namelen = uint32(len(names[i]))
-
-		if controlLen > 0 {
-			controls[i] = make([]byte, controlLen)
-			msgs[i].Hdr.Control = &controls[i][0]
-			msgs[i].Hdr.Controllen = controllen(len(controls[i]))
-		} else {
-			msgs[i].Hdr.Control = nil
-			msgs[i].Hdr.Controllen = controllen(0)
-		}
 	}
 
-	return msgs, buffers, names, controls
+	return msgs, buffers, names
 }
 
-func setIovecBase(msg *rawMessage, buf []byte) {
+func setRawMessageControl(msg *rawMessage, buf []byte) {
-	iov := (*iovec)(msg.Hdr.Iov)
+	if len(buf) == 0 {
-	iov.Base = &buf[0]
+		msg.Hdr.Control = nil
-	iov.Len = uint32(len(buf))
+		msg.Hdr.Controllen = 0
+		return
+	}
+	msg.Hdr.Control = &buf[0]
+	msg.Hdr.Controllen = uint32(len(buf))
+}
+
+func getRawMessageControlLen(msg *rawMessage) int {
+	return int(msg.Hdr.Controllen)
+}
+
+func getRawMessageFlags(msg *rawMessage) int {
+	return int(msg.Hdr.Flags)
+}
+
+func setCmsgLen(h *unix.Cmsghdr, l int) {
+	h.Len = uint32(l)
+}
+
+func setIovecSlice(iov *iovec, b []byte) {
+	if len(b) == 0 {
+		iov.Base = nil
+		iov.Len = 0
+		return
+	}
+	iov.Base = &b[0]
+	iov.Len = uint32(len(b))
 }

udp/udp_linux_64.go

@@ -33,50 +33,60 @@ type rawMessage struct {
 	Pad0 [4]byte
 }
 
-func (u *StdConn) PrepareRawMessages(n int) ([]rawMessage, [][]byte, [][]byte, [][]byte) {
+func (u *StdConn) PrepareRawMessages(n int, bufSize int) ([]rawMessage, [][]byte, [][]byte) {
-	controlLen := int(u.controlLen.Load())
+	if bufSize <= 0 {
-
+		bufSize = MTU
+	}
 	msgs := make([]rawMessage, n)
 	buffers := make([][]byte, n)
 	names := make([][]byte, n)
 
-	var controls [][]byte
-	if controlLen > 0 {
-		controls = make([][]byte, n)
-	}
-
 	for i := range msgs {
-		size := int(u.groBufSize.Load())
+		buffers[i] = make([]byte, bufSize)
-		if size < MTU {
-			size = MTU
-		}
-		buf := u.borrowRxBuffer(size)
-		buffers[i] = buf
 		names[i] = make([]byte, unix.SizeofSockaddrInet6)
 
-		vs := []iovec{{Base: &buf[0], Len: uint64(len(buf))}}
+		vs := []iovec{
+			{Base: &buffers[i][0], Len: uint64(len(buffers[i]))},
+		}
 
 		msgs[i].Hdr.Iov = &vs[0]
 		msgs[i].Hdr.Iovlen = uint64(len(vs))
 
 		msgs[i].Hdr.Name = &names[i][0]
 		msgs[i].Hdr.Namelen = uint32(len(names[i]))
-
-		if controlLen > 0 {
-			controls[i] = make([]byte, controlLen)
-			msgs[i].Hdr.Control = &controls[i][0]
-			msgs[i].Hdr.Controllen = controllen(len(controls[i]))
-		} else {
-			msgs[i].Hdr.Control = nil
-			msgs[i].Hdr.Controllen = controllen(0)
-		}
 	}
 
-	return msgs, buffers, names, controls
+	return msgs, buffers, names
 }
 
-func setIovecBase(msg *rawMessage, buf []byte) {
+func setRawMessageControl(msg *rawMessage, buf []byte) {
-	iov := (*iovec)(msg.Hdr.Iov)
+	if len(buf) == 0 {
-	iov.Base = &buf[0]
+		msg.Hdr.Control = nil
-	iov.Len = uint64(len(buf))
+		msg.Hdr.Controllen = 0
+		return
+	}
+	msg.Hdr.Control = &buf[0]
+	msg.Hdr.Controllen = uint64(len(buf))
+}
+
+func getRawMessageControlLen(msg *rawMessage) int {
+	return int(msg.Hdr.Controllen)
+}
+
+func getRawMessageFlags(msg *rawMessage) int {
+	return int(msg.Hdr.Flags)
+}
+
+func setCmsgLen(h *unix.Cmsghdr, l int) {
+	h.Len = uint64(l)
+}
+
+func setIovecSlice(iov *iovec, b []byte) {
+	if len(b) == 0 {
+		iov.Base = nil
+		iov.Len = 0
+		return
+	}
+	iov.Base = &b[0]
+	iov.Len = uint64(len(b))
 }

udp/udp_rio_windows.go

@@ -134,7 +134,7 @@ func (u *RIOConn) bind(sa windows.Sockaddr) error {
 	return nil
 }
 
-func (u *RIOConn) ListenOut(r EncReader) {
+func (u *RIOConn) ListenOut(r EncReader) error {
 	buffer := make([]byte, MTU)
 
 	for {
@@ -149,7 +149,7 @@ func (u *RIOConn) ListenOut(r EncReader) {
 			continue
 		}
 
-		r(netip.AddrPortFrom(netip.AddrFrom16(rua.Addr).Unmap(), (rua.Port>>8)|((rua.Port&0xff)<<8)), buffer[:n], nil)
+		r(netip.AddrPortFrom(netip.AddrFrom16(rua.Addr).Unmap(), (rua.Port>>8)|((rua.Port&0xff)<<8)), buffer[:n])
 	}
 }
 
@@ -304,6 +304,17 @@ func (u *RIOConn) WriteTo(buf []byte, ip netip.AddrPort) error {
 	return winrio.SendEx(u.rq, dataBuffer, 1, nil, addressBuffer, nil, nil, 0, 0)
 }
 
+func (u *RIOConn) WriteBatch(pkts []BatchPacket) (int, error) {
+	sent := 0
+	for _, pkt := range pkts {
+		if err := u.WriteTo(pkt.Payload, pkt.Addr); err != nil {
+			return sent, err
+		}
+		sent++
+	}
+	return sent, nil
+}
+
 func (u *RIOConn) LocalAddr() (netip.AddrPort, error) {
 	sa, err := windows.Getsockname(u.sock)
 	if err != nil {

udp/udp_tester.go

@@ -106,13 +106,24 @@ func (u *TesterConn) WriteTo(b []byte, addr netip.AddrPort) error {
 	return nil
 }
 
+func (u *TesterConn) WriteBatch(pkts []BatchPacket) (int, error) {
+	sent := 0
+	for _, pkt := range pkts {
+		if err := u.WriteTo(pkt.Payload, pkt.Addr); err != nil {
+			return sent, err
+		}
+		sent++
+	}
+	return sent, nil
+}
+
 func (u *TesterConn) ListenOut(r EncReader) {
 	for {
 		p, ok := <-u.RxPackets
 		if !ok {
 			return
 		}
-		r(p.From, p.Data, func() {})
+		r(p.From, p.Data)
 	}
 }