more incompatibilities, was this a good idea at all?

* Fix recv_error receipt limit allowance

* backport #1463 recv_error behavior changes

---------

Co-authored-by: JackDoan <me@jackdoan.com>

CHANGELOG.md

@@ -7,6 +7,35 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Changed
+
+- Various dependency updates.
+
+## [1.9.7] - 2025-10-10
+
+### Security
+
+- Fix an issue where Nebula could incorrectly accept and process a packet from an erroneous source IP when the sender's
+  certificate is configured with unsafe_routes (cert v1/v2) or multiple IPs (cert v2). (#1494)
+
+### Changed
+
+- Disable sending `recv_error` messages when a packet is received outside the allowable counter window. (#1459)
+- Improve error messages and remove some unnecessary fatal conditions in the Windows and generic udp listener. (#1543)
+
+## [1.9.6] - 2025-7-15
+
+### Added
+
+- Support dropping inactive tunnels. This is disabled by default in this release but can be enabled with `tunnels.drop_inactive`. See example config for more details. (#1413)
+
+### Fixed
+
+- Fix Darwin freeze due to presence of some Network Extensions (#1426)
+- Ensure the same relay tunnel is always used when multiple relay tunnels are present (#1422)
+- Fix Windows freeze due to ICMP error handling (#1412)
+- Fix relay migration panic (#1403)
+
 ## [1.9.5] - 2024-12-05
 
 ### Added
@@ -674,7 +703,9 @@ created.)
 
 - Initial public release.
 
-[Unreleased]: https://github.com/slackhq/nebula/compare/v1.9.5...HEAD
+[Unreleased]: https://github.com/slackhq/nebula/compare/v1.9.7...HEAD
+[1.9.7]: https://github.com/slackhq/nebula/releases/tag/v1.9.7
+[1.9.6]: https://github.com/slackhq/nebula/releases/tag/v1.9.6
 [1.9.5]: https://github.com/slackhq/nebula/releases/tag/v1.9.5
 [1.9.4]: https://github.com/slackhq/nebula/releases/tag/v1.9.4
 [1.9.3]: https://github.com/slackhq/nebula/releases/tag/v1.9.3

Makefile

@@ -133,6 +133,8 @@ build/linux-mips-softfloat/%: LDFLAGS += -s -w
 # boringcrypto
 build/linux-amd64-boringcrypto/%: GOENV += GOEXPERIMENT=boringcrypto CGO_ENABLED=1
 build/linux-arm64-boringcrypto/%: GOENV += GOEXPERIMENT=boringcrypto CGO_ENABLED=1
+build/linux-amd64-boringcrypto/%: LDFLAGS += -checklinkname=0
+build/linux-arm64-boringcrypto/%: LDFLAGS += -checklinkname=0
 
 build/%/nebula: .FORCE
 	GOOS=$(firstword $(subst -, , $*)) \
@@ -166,7 +168,7 @@ test:
 	go test -v ./...
 
 test-boringcrypto:
-	GOEXPERIMENT=boringcrypto CGO_ENABLED=1 go test -v ./...
+	GOEXPERIMENT=boringcrypto CGO_ENABLED=1 go test -ldflags "-checklinkname=0" -v ./...
 
 test-cov-html:
 	go test -coverprofile=coverage.out

connection_manager.go

@@ -4,13 +4,16 @@ import (
 	"bytes"
 	"context"
 	"encoding/binary"
+	"fmt"
 	"net/netip"
 	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/rcrowley/go-metrics"
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
+	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/header"
 )
 
@@ -27,130 +30,124 @@ const (
 )
 
 type connectionManager struct {
-	in     map[uint32]struct{}
-	inLock *sync.RWMutex
-
-	out     map[uint32]struct{}
-	outLock *sync.RWMutex
-
 	// relayUsed holds which relay localIndexs are in use
 	relayUsed     map[uint32]struct{}
 	relayUsedLock *sync.RWMutex
 
-	hostMap                 *HostMap
+	hostMap      *HostMap
-	trafficTimer            *LockingTimerWheel[uint32]
+	trafficTimer *LockingTimerWheel[uint32]
-	intf                    *Interface
+	intf         *Interface
-	pendingDeletion         map[uint32]struct{}
+	punchy       *Punchy
-	punchy                  *Punchy
+
+	// Configuration settings
 	checkInterval           time.Duration
 	pendingDeletionInterval time.Duration
-	metricsTxPunchy         metrics.Counter
+	inactivityTimeout       atomic.Int64
+	dropInactive            atomic.Bool
+
+	metricsTxPunchy metrics.Counter
 
 	l *logrus.Logger
 }
 
-func newConnectionManager(ctx context.Context, l *logrus.Logger, intf *Interface, checkInterval, pendingDeletionInterval time.Duration, punchy *Punchy) *connectionManager {
+func newConnectionManagerFromConfig(l *logrus.Logger, c *config.C, hm *HostMap, p *Punchy) *connectionManager {
-	var max time.Duration
+	cm := &connectionManager{
-	if checkInterval < pendingDeletionInterval {
+		hostMap:         hm,
-		max = pendingDeletionInterval
+		l:               l,
-	} else {
+		punchy:          p,
-		max = checkInterval
+		relayUsed:       make(map[uint32]struct{}),
+		relayUsedLock:   &sync.RWMutex{},
+		metricsTxPunchy: metrics.GetOrRegisterCounter("messages.tx.punchy", nil),
 	}
 
-	nc := &connectionManager{
+	cm.reload(c, true)
-		hostMap:                 intf.hostMap,
+	c.RegisterReloadCallback(func(c *config.C) {
-		in:                      make(map[uint32]struct{}),
+		cm.reload(c, false)
-		inLock:                  &sync.RWMutex{},
+	})
-		out:                     make(map[uint32]struct{}),
-		outLock:                 &sync.RWMutex{},
-		relayUsed:               make(map[uint32]struct{}),
-		relayUsedLock:           &sync.RWMutex{},
-		trafficTimer:            NewLockingTimerWheel[uint32](time.Millisecond*500, max),
-		intf:                    intf,
-		pendingDeletion:         make(map[uint32]struct{}),
-		checkInterval:           checkInterval,
-		pendingDeletionInterval: pendingDeletionInterval,
-		punchy:                  punchy,
-		metricsTxPunchy:         metrics.GetOrRegisterCounter("messages.tx.punchy", nil),
-		l:                       l,
-	}
 
-	nc.Start(ctx)
+	return cm
-	return nc
 }
 
-func (n *connectionManager) In(localIndex uint32) {
+func (cm *connectionManager) reload(c *config.C, initial bool) {
-	n.inLock.RLock()
+	if initial {
-	// If this already exists, return
+		cm.checkInterval = time.Duration(c.GetInt("timers.connection_alive_interval", 5)) * time.Second
-	if _, ok := n.in[localIndex]; ok {
+		cm.pendingDeletionInterval = time.Duration(c.GetInt("timers.pending_deletion_interval", 10)) * time.Second
-		n.inLock.RUnlock()
+
-		return
+		// We want at least a minimum resolution of 500ms per tick so that we can hit these intervals
+		// pretty close to their configured duration.
+		// The inactivity duration is checked each time a hostinfo ticks through so we don't need the wheel to contain it.
+		minDuration := min(time.Millisecond*500, cm.checkInterval, cm.pendingDeletionInterval)
+		maxDuration := max(cm.checkInterval, cm.pendingDeletionInterval)
+		cm.trafficTimer = NewLockingTimerWheel[uint32](minDuration, maxDuration)
+	}
+
+	if initial || c.HasChanged("tunnels.inactivity_timeout") {
+		old := cm.getInactivityTimeout()
+		cm.inactivityTimeout.Store((int64)(c.GetDuration("tunnels.inactivity_timeout", 10*time.Minute)))
+		if !initial {
+			cm.l.WithField("oldDuration", old).
+				WithField("newDuration", cm.getInactivityTimeout()).
+				Info("Inactivity timeout has changed")
+		}
+	}
+
+	if initial || c.HasChanged("tunnels.drop_inactive") {
+		old := cm.dropInactive.Load()
+		cm.dropInactive.Store(c.GetBool("tunnels.drop_inactive", false))
+		if !initial {
+			cm.l.WithField("oldBool", old).
+				WithField("newBool", cm.dropInactive.Load()).
+				Info("Drop inactive setting has changed")
+		}
 	}
-	n.inLock.RUnlock()
-	n.inLock.Lock()
-	n.in[localIndex] = struct{}{}
-	n.inLock.Unlock()
 }
 
-func (n *connectionManager) Out(localIndex uint32) {
+func (cm *connectionManager) getInactivityTimeout() time.Duration {
-	n.outLock.RLock()
+	return (time.Duration)(cm.inactivityTimeout.Load())
-	// If this already exists, return
-	if _, ok := n.out[localIndex]; ok {
-		n.outLock.RUnlock()
-		return
-	}
-	n.outLock.RUnlock()
-	n.outLock.Lock()
-	n.out[localIndex] = struct{}{}
-	n.outLock.Unlock()
 }
 
-func (n *connectionManager) RelayUsed(localIndex uint32) {
+func (cm *connectionManager) In(h *HostInfo) {
-	n.relayUsedLock.RLock()
+	h.in.Store(true)
+}
+
+func (cm *connectionManager) Out(h *HostInfo) {
+	h.out.Store(true)
+}
+
+func (cm *connectionManager) RelayUsed(localIndex uint32) {
+	cm.relayUsedLock.RLock()
 	// If this already exists, return
-	if _, ok := n.relayUsed[localIndex]; ok {
+	if _, ok := cm.relayUsed[localIndex]; ok {
-		n.relayUsedLock.RUnlock()
+		cm.relayUsedLock.RUnlock()
 		return
 	}
-	n.relayUsedLock.RUnlock()
+	cm.relayUsedLock.RUnlock()
-	n.relayUsedLock.Lock()
+	cm.relayUsedLock.Lock()
-	n.relayUsed[localIndex] = struct{}{}
+	cm.relayUsed[localIndex] = struct{}{}
-	n.relayUsedLock.Unlock()
+	cm.relayUsedLock.Unlock()
 }
 
 // getAndResetTrafficCheck returns if there was any inbound or outbound traffic within the last tick and
 // resets the state for this local index
-func (n *connectionManager) getAndResetTrafficCheck(localIndex uint32) (bool, bool) {
+func (cm *connectionManager) getAndResetTrafficCheck(h *HostInfo, now time.Time) (bool, bool) {
-	n.inLock.Lock()
+	in := h.in.Swap(false)
-	n.outLock.Lock()
+	out := h.out.Swap(false)
-	_, in := n.in[localIndex]
+	if in || out {
-	_, out := n.out[localIndex]
+		h.lastUsed = now
-	delete(n.in, localIndex)
+	}
-	delete(n.out, localIndex)
-	n.inLock.Unlock()
-	n.outLock.Unlock()
 	return in, out
 }
 
-func (n *connectionManager) AddTrafficWatch(localIndex uint32) {
+// AddTrafficWatch must be called for every new HostInfo.
-	// Use a write lock directly because it should be incredibly rare that we are ever already tracking this index
+// We will continue to monitor the HostInfo until the tunnel is dropped.
-	n.outLock.Lock()
+func (cm *connectionManager) AddTrafficWatch(h *HostInfo) {
-	if _, ok := n.out[localIndex]; ok {
+	if h.out.Swap(true) == false {
-		n.outLock.Unlock()
+		cm.trafficTimer.Add(h.localIndexId, cm.checkInterval)
-		return
 	}
-	n.out[localIndex] = struct{}{}
-	n.trafficTimer.Add(localIndex, n.checkInterval)
-	n.outLock.Unlock()
 }
 
-func (n *connectionManager) Start(ctx context.Context) {
+func (cm *connectionManager) Start(ctx context.Context) {
-	go n.Run(ctx)
+	clockSource := time.NewTicker(cm.trafficTimer.t.tickDuration)
-}
-
-func (n *connectionManager) Run(ctx context.Context) {
-	//TODO: this tick should be based on the min wheel tick? Check firewall
-	clockSource := time.NewTicker(500 * time.Millisecond)
 	defer clockSource.Stop()
 
 	p := []byte("")
@@ -163,61 +160,61 @@ func (n *connectionManager) Run(ctx context.Context) {
 			return
 
 		case now := <-clockSource.C:
-			n.trafficTimer.Advance(now)
+			cm.trafficTimer.Advance(now)
 			for {
-				localIndex, has := n.trafficTimer.Purge()
+				localIndex, has := cm.trafficTimer.Purge()
 				if !has {
 					break
 				}
 
-				n.doTrafficCheck(localIndex, p, nb, out, now)
+				cm.doTrafficCheck(localIndex, p, nb, out, now)
 			}
 		}
 	}
 }
 
-func (n *connectionManager) doTrafficCheck(localIndex uint32, p, nb, out []byte, now time.Time) {
+func (cm *connectionManager) doTrafficCheck(localIndex uint32, p, nb, out []byte, now time.Time) {
-	decision, hostinfo, primary := n.makeTrafficDecision(localIndex, now)
+	decision, hostinfo, primary := cm.makeTrafficDecision(localIndex, now)
 
 	switch decision {
 	case deleteTunnel:
-		if n.hostMap.DeleteHostInfo(hostinfo) {
+		if cm.hostMap.DeleteHostInfo(hostinfo) {
 			// Only clearing the lighthouse cache if this is the last hostinfo for this vpn ip in the hostmap
-			n.intf.lightHouse.DeleteVpnIp(hostinfo.vpnIp)
+			cm.intf.lightHouse.DeleteVpnIp(hostinfo.vpnIp)
 		}
 
 	case closeTunnel:
-		n.intf.sendCloseTunnel(hostinfo)
+		cm.intf.sendCloseTunnel(hostinfo)
-		n.intf.closeTunnel(hostinfo)
+		cm.intf.closeTunnel(hostinfo)
 
 	case swapPrimary:
-		n.swapPrimary(hostinfo, primary)
+		cm.swapPrimary(hostinfo, primary)
 
 	case migrateRelays:
-		n.migrateRelayUsed(hostinfo, primary)
+		cm.migrateRelayUsed(hostinfo, primary)
 
 	case tryRehandshake:
-		n.tryRehandshake(hostinfo)
+		cm.tryRehandshake(hostinfo)
 
 	case sendTestPacket:
-		n.intf.SendMessageToHostInfo(header.Test, header.TestRequest, hostinfo, p, nb, out)
+		cm.intf.SendMessageToHostInfo(header.Test, header.TestRequest, hostinfo, p, nb, out)
 	}
 
-	n.resetRelayTrafficCheck(hostinfo)
+	cm.resetRelayTrafficCheck(hostinfo)
 }
 
-func (n *connectionManager) resetRelayTrafficCheck(hostinfo *HostInfo) {
+func (cm *connectionManager) resetRelayTrafficCheck(hostinfo *HostInfo) {
 	if hostinfo != nil {
-		n.relayUsedLock.Lock()
+		cm.relayUsedLock.Lock()
-		defer n.relayUsedLock.Unlock()
+		defer cm.relayUsedLock.Unlock()
 		// No need to migrate any relays, delete usage info now.
 		for _, idx := range hostinfo.relayState.CopyRelayForIdxs() {
-			delete(n.relayUsed, idx)
+			delete(cm.relayUsed, idx)
 		}
 	}
 }
 
-func (n *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo) {
+func (cm *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo) {
 	relayFor := oldhostinfo.relayState.CopyAllRelayFor()
 
 	for _, r := range relayFor {
@@ -227,46 +224,51 @@ func (n *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo)
 		var relayFrom netip.Addr
 		var relayTo netip.Addr
 		switch {
-		case ok && existing.State == Established:
+		case ok:
-			// This relay already exists in newhostinfo, then do nothing.
+			switch existing.State {
-			continue
+			case Established, PeerRequested, Disestablished:
-		case ok && existing.State == Requested:
+				// This relay already exists in newhostinfo, then do nothing.
-			// The relay exists in a Requested state; re-send the request
+				continue
-			index = existing.LocalIndex
+			case Requested:
-			switch r.Type {
+				// The relayed connection exists in a Requested state; re-send the request
-			case TerminalType:
+				index = existing.LocalIndex
-				relayFrom = n.intf.myVpnNet.Addr()
+				switch r.Type {
-				relayTo = existing.PeerIp
+				case TerminalType:
-			case ForwardingType:
+					relayFrom = cm.intf.myVpnNet.Addr()
-				relayFrom = existing.PeerIp
+					relayTo = existing.PeerIp
-				relayTo = newhostinfo.vpnIp
+				case ForwardingType:
-			default:
+					relayFrom = existing.PeerIp
-				// should never happen
+					relayTo = newhostinfo.vpnIp
+				default:
+					// should never happen
+					panic(fmt.Sprintf("Migrating unknown relay type: %v", r.Type))
+				}
 			}
 		case !ok:
-			n.relayUsedLock.RLock()
+			cm.relayUsedLock.RLock()
-			if _, relayUsed := n.relayUsed[r.LocalIndex]; !relayUsed {
+			if _, relayUsed := cm.relayUsed[r.LocalIndex]; !relayUsed {
 				// The relay hasn't been used; don't migrate it.
-				n.relayUsedLock.RUnlock()
+				cm.relayUsedLock.RUnlock()
 				continue
 			}
-			n.relayUsedLock.RUnlock()
+			cm.relayUsedLock.RUnlock()
 			// The relay doesn't exist at all; create some relay state and send the request.
 			var err error
-			index, err = AddRelay(n.l, newhostinfo, n.hostMap, r.PeerIp, nil, r.Type, Requested)
+			index, err = AddRelay(cm.l, newhostinfo, cm.hostMap, r.PeerIp, nil, r.Type, Requested)
 			if err != nil {
-				n.l.WithError(err).Error("failed to migrate relay to new hostinfo")
+				cm.l.WithError(err).Error("failed to migrate relay to new hostinfo")
 				continue
 			}
 			switch r.Type {
 			case TerminalType:
-				relayFrom = n.intf.myVpnNet.Addr()
+				relayFrom = cm.intf.myVpnNet.Addr()
 				relayTo = r.PeerIp
 			case ForwardingType:
 				relayFrom = r.PeerIp
 				relayTo = newhostinfo.vpnIp
 			default:
 				// should never happen
+				panic(fmt.Sprintf("Migrating unknown relay type: %v", r.Type))
 			}
 		}
 
@@ -283,10 +285,10 @@ func (n *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo)
 		}
 		msg, err := req.Marshal()
 		if err != nil {
-			n.l.WithError(err).Error("failed to marshal Control message to migrate relay")
+			cm.l.WithError(err).Error("failed to marshal Control message to migrate relay")
 		} else {
-			n.intf.SendMessageToHostInfo(header.Control, 0, newhostinfo, msg, make([]byte, 12), make([]byte, mtu))
+			cm.intf.SendMessageToHostInfo(header.Control, 0, newhostinfo, msg, make([]byte, 12), make([]byte, mtu))
-			n.l.WithFields(logrus.Fields{
+			cm.l.WithFields(logrus.Fields{
 				"relayFrom":           req.RelayFromIp,
 				"relayTo":             req.RelayToIp,
 				"initiatorRelayIndex": req.InitiatorRelayIndex,
@@ -297,46 +299,45 @@ func (n *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo)
 	}
 }
 
-func (n *connectionManager) makeTrafficDecision(localIndex uint32, now time.Time) (trafficDecision, *HostInfo, *HostInfo) {
+func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Time) (trafficDecision, *HostInfo, *HostInfo) {
-	n.hostMap.RLock()
+	// Read lock the main hostmap to order decisions based on tunnels being the primary tunnel
-	defer n.hostMap.RUnlock()
+	cm.hostMap.RLock()
+	defer cm.hostMap.RUnlock()
 
-	hostinfo := n.hostMap.Indexes[localIndex]
+	hostinfo := cm.hostMap.Indexes[localIndex]
 	if hostinfo == nil {
-		n.l.WithField("localIndex", localIndex).Debugf("Not found in hostmap")
+		cm.l.WithField("localIndex", localIndex).Debugln("Not found in hostmap")
-		delete(n.pendingDeletion, localIndex)
 		return doNothing, nil, nil
 	}
 
-	if n.isInvalidCertificate(now, hostinfo) {
+	if cm.isInvalidCertificate(now, hostinfo) {
-		delete(n.pendingDeletion, hostinfo.localIndexId)
 		return closeTunnel, hostinfo, nil
 	}
 
-	primary := n.hostMap.Hosts[hostinfo.vpnIp]
+	primary := cm.hostMap.Hosts[hostinfo.vpnIp]
 	mainHostInfo := true
 	if primary != nil && primary != hostinfo {
 		mainHostInfo = false
 	}
 
 	// Check for traffic on this hostinfo
-	inTraffic, outTraffic := n.getAndResetTrafficCheck(localIndex)
+	inTraffic, outTraffic := cm.getAndResetTrafficCheck(hostinfo, now)
 
 	// A hostinfo is determined alive if there is incoming traffic
 	if inTraffic {
 		decision := doNothing
-		if n.l.Level >= logrus.DebugLevel {
+		if cm.l.Level >= logrus.DebugLevel {
-			hostinfo.logger(n.l).
+			hostinfo.logger(cm.l).
 				WithField("tunnelCheck", m{"state": "alive", "method": "passive"}).
 				Debug("Tunnel status")
 		}
-		delete(n.pendingDeletion, hostinfo.localIndexId)
+		hostinfo.pendingDeletion.Store(false)
 
 		if mainHostInfo {
 			decision = tryRehandshake
 
 		} else {
-			if n.shouldSwapPrimary(hostinfo, primary) {
+			if cm.shouldSwapPrimary(hostinfo, primary) {
 				decision = swapPrimary
 			} else {
 				// migrate the relays to the primary, if in use.
@@ -344,46 +345,55 @@ func (n *connectionManager) makeTrafficDecision(localIndex uint32, now time.Time
 			}
 		}
 
-		n.trafficTimer.Add(hostinfo.localIndexId, n.checkInterval)
+		cm.trafficTimer.Add(hostinfo.localIndexId, cm.checkInterval)
 
 		if !outTraffic {
 			// Send a punch packet to keep the NAT state alive
-			n.sendPunch(hostinfo)
+			cm.sendPunch(hostinfo)
 		}
 
 		return decision, hostinfo, primary
 	}
 
-	if _, ok := n.pendingDeletion[hostinfo.localIndexId]; ok {
+	if hostinfo.pendingDeletion.Load() {
 		// We have already sent a test packet and nothing was returned, this hostinfo is dead
-		hostinfo.logger(n.l).
+		hostinfo.logger(cm.l).
 			WithField("tunnelCheck", m{"state": "dead", "method": "active"}).
 			Info("Tunnel status")
 
-		delete(n.pendingDeletion, hostinfo.localIndexId)
 		return deleteTunnel, hostinfo, nil
 	}
 
 	decision := doNothing
 	if hostinfo != nil && hostinfo.ConnectionState != nil && mainHostInfo {
 		if !outTraffic {
+			inactiveFor, isInactive := cm.isInactive(hostinfo, now)
+			if isInactive {
+				// Tunnel is inactive, tear it down
+				hostinfo.logger(cm.l).
+					WithField("inactiveDuration", inactiveFor).
+					WithField("primary", mainHostInfo).
+					Info("Dropping tunnel due to inactivity")
+
+				return closeTunnel, hostinfo, primary
+			}
+
 			// If we aren't sending or receiving traffic then its an unused tunnel and we don't to test the tunnel.
 			// Just maintain NAT state if configured to do so.
-			n.sendPunch(hostinfo)
+			cm.sendPunch(hostinfo)
-			n.trafficTimer.Add(hostinfo.localIndexId, n.checkInterval)
+			cm.trafficTimer.Add(hostinfo.localIndexId, cm.checkInterval)
 			return doNothing, nil, nil
-
 		}
 
-		if n.punchy.GetTargetEverything() {
+		if cm.punchy.GetTargetEverything() {
 			// This is similar to the old punchy behavior with a slight optimization.
 			// We aren't receiving traffic but we are sending it, punch on all known
 			// ips in case we need to re-prime NAT state
-			n.sendPunch(hostinfo)
+			cm.sendPunch(hostinfo)
 		}
 
-		if n.l.Level >= logrus.DebugLevel {
+		if cm.l.Level >= logrus.DebugLevel {
-			hostinfo.logger(n.l).
+			hostinfo.logger(cm.l).
 				WithField("tunnelCheck", m{"state": "testing", "method": "active"}).
 				Debug("Tunnel status")
 		}
@@ -392,95 +402,118 @@ func (n *connectionManager) makeTrafficDecision(localIndex uint32, now time.Time
 		decision = sendTestPacket
 
 	} else {
-		if n.l.Level >= logrus.DebugLevel {
+		if cm.l.Level >= logrus.DebugLevel {
-			hostinfo.logger(n.l).Debugf("Hostinfo sadness")
+			hostinfo.logger(cm.l).Debugf("Hostinfo sadness")
 		}
 	}
 
-	n.pendingDeletion[hostinfo.localIndexId] = struct{}{}
+	hostinfo.pendingDeletion.Store(true)
-	n.trafficTimer.Add(hostinfo.localIndexId, n.pendingDeletionInterval)
+	cm.trafficTimer.Add(hostinfo.localIndexId, cm.pendingDeletionInterval)
 	return decision, hostinfo, nil
 }
 
-func (n *connectionManager) shouldSwapPrimary(current, primary *HostInfo) bool {
+func (cm *connectionManager) isInactive(hostinfo *HostInfo, now time.Time) (time.Duration, bool) {
+	if cm.dropInactive.Load() == false {
+		// We aren't configured to drop inactive tunnels
+		return 0, false
+	}
+
+	inactiveDuration := now.Sub(hostinfo.lastUsed)
+	if inactiveDuration < cm.getInactivityTimeout() {
+		// It's not considered inactive
+		return inactiveDuration, false
+	}
+
+	// The tunnel is inactive
+	return inactiveDuration, true
+}
+
+func (cm *connectionManager) shouldSwapPrimary(current, primary *HostInfo) bool {
 	// The primary tunnel is the most recent handshake to complete locally and should work entirely fine.
 	// If we are here then we have multiple tunnels for a host pair and neither side believes the same tunnel is primary.
 	// Let's sort this out.
 
-	if current.vpnIp.Compare(n.intf.myVpnNet.Addr()) < 0 {
+	if current.vpnIp.Compare(cm.intf.myVpnNet.Addr()) < 0 {
 		// Only one side should flip primary because if both flip then we may never resolve to a single tunnel.
 		// vpn ip is static across all tunnels for this host pair so lets use that to determine who is flipping.
 		// The remotes vpn ip is lower than mine. I will not flip.
 		return false
 	}
 
-	certState := n.intf.pki.GetCertState()
+	certState := cm.intf.pki.GetCertState()
 	return bytes.Equal(current.ConnectionState.myCert.Signature, certState.Certificate.Signature)
 }
 
-func (n *connectionManager) swapPrimary(current, primary *HostInfo) {
+func (cm *connectionManager) swapPrimary(current, primary *HostInfo) {
-	n.hostMap.Lock()
+	cm.hostMap.Lock()
 	// Make sure the primary is still the same after the write lock. This avoids a race with a rehandshake.
-	if n.hostMap.Hosts[current.vpnIp] == primary {
+	if cm.hostMap.Hosts[current.vpnIp] == primary {
-		n.hostMap.unlockedMakePrimary(current)
+		cm.hostMap.unlockedMakePrimary(current)
 	}
-	n.hostMap.Unlock()
+	cm.hostMap.Unlock()
 }
 
 // isInvalidCertificate will check if we should destroy a tunnel if pki.disconnect_invalid is true and
 // the certificate is no longer valid. Block listed certificates will skip the pki.disconnect_invalid
 // check and return true.
-func (n *connectionManager) isInvalidCertificate(now time.Time, hostinfo *HostInfo) bool {
+func (cm *connectionManager) isInvalidCertificate(now time.Time, hostinfo *HostInfo) bool {
 	remoteCert := hostinfo.GetCert()
 	if remoteCert == nil {
 		return false
 	}
 
-	valid, err := remoteCert.VerifyWithCache(now, n.intf.pki.GetCAPool())
+	valid, err := remoteCert.VerifyWithCache(now, cm.intf.pki.GetCAPool())
 	if valid {
 		return false
 	}
 
-	if !n.intf.disconnectInvalid.Load() && err != cert.ErrBlockListed {
+	if !cm.intf.disconnectInvalid.Load() && err != cert.ErrBlockListed {
 		// Block listed certificates should always be disconnected
 		return false
 	}
 
 	fingerprint, _ := remoteCert.Sha256Sum()
-	hostinfo.logger(n.l).WithError(err).
+	hostinfo.logger(cm.l).WithError(err).
 		WithField("fingerprint", fingerprint).
 		Info("Remote certificate is no longer valid, tearing down the tunnel")
 
 	return true
 }
 
-func (n *connectionManager) sendPunch(hostinfo *HostInfo) {
+func (cm *connectionManager) sendPunch(hostinfo *HostInfo) {
-	if !n.punchy.GetPunch() {
+	if !cm.punchy.GetPunch() {
 		// Punching is disabled
 		return
 	}
 
-	if n.punchy.GetTargetEverything() {
+	if cm.intf.lightHouse.IsLighthouseIP(hostinfo.vpnIp) {
-		hostinfo.remotes.ForEach(n.hostMap.GetPreferredRanges(), func(addr netip.AddrPort, preferred bool) {
+		// Do not punch to lighthouses, we assume our lighthouse update interval is good enough.
-			n.metricsTxPunchy.Inc(1)
+		// In the event the update interval is not sufficient to maintain NAT state then a publicly available lighthouse
-			n.intf.outside.WriteTo([]byte{1}, addr)
+		// would lose the ability to notify us and punchy.respond would become unreliable.
+		return
+	}
+
+	if cm.punchy.GetTargetEverything() {
+		hostinfo.remotes.ForEach(cm.hostMap.GetPreferredRanges(), func(addr netip.AddrPort, preferred bool) {
+			cm.metricsTxPunchy.Inc(1)
+			cm.intf.outside.WriteTo([]byte{1}, addr)
 		})
 
 	} else if hostinfo.remote.IsValid() {
-		n.metricsTxPunchy.Inc(1)
+		cm.metricsTxPunchy.Inc(1)
-		n.intf.outside.WriteTo([]byte{1}, hostinfo.remote)
+		cm.intf.outside.WriteTo([]byte{1}, hostinfo.remote)
 	}
 }
 
-func (n *connectionManager) tryRehandshake(hostinfo *HostInfo) {
+func (cm *connectionManager) tryRehandshake(hostinfo *HostInfo) {
-	certState := n.intf.pki.GetCertState()
+	certState := cm.intf.pki.GetCertState()
 	if bytes.Equal(hostinfo.ConnectionState.myCert.Signature, certState.Certificate.Signature) {
 		return
 	}
 
-	n.l.WithField("vpnIp", hostinfo.vpnIp).
+	cm.l.WithField("vpnIp", hostinfo.vpnIp).
 		WithField("reason", "local certificate is not current").
 		Info("Re-handshaking with remote")
 
-	n.intf.handshakeManager.StartHandshake(hostinfo.vpnIp, nil)
+	cm.intf.handshakeManager.StartHandshake(hostinfo.vpnIp, nil)
 }

connection_manager_test.go

@@ -1,7 +1,6 @@
 package nebula
 
 import (
-	"context"
 	"crypto/ed25519"
 	"crypto/rand"
 	"net"
@@ -65,10 +64,10 @@ func Test_NewConnectionManagerTest(t *testing.T) {
 	ifce.pki.cs.Store(cs)
 
 	// Create manager
-	ctx, cancel := context.WithCancel(context.Background())
+	conf := config.NewC(l)
-	defer cancel()
+	punchy := NewPunchyFromConfig(l, conf)
-	punchy := NewPunchyFromConfig(l, config.NewC(l))
+	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
-	nc := newConnectionManager(ctx, l, ifce, 5, 10, punchy)
+	nc.intf = ifce
 	p := []byte("")
 	nb := make([]byte, 12, 12)
 	out := make([]byte, mtu)
@@ -86,31 +85,32 @@ func Test_NewConnectionManagerTest(t *testing.T) {
 	nc.hostMap.unlockedAddHostInfo(hostinfo, ifce)
 
 	// We saw traffic out to vpnIp
-	nc.Out(hostinfo.localIndexId)
+	nc.Out(hostinfo)
-	nc.In(hostinfo.localIndexId)
+	nc.In(hostinfo)
-	assert.NotContains(t, nc.pendingDeletion, hostinfo.localIndexId)
+	assert.False(t, hostinfo.pendingDeletion.Load())
 	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnIp)
 	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
-	assert.Contains(t, nc.out, hostinfo.localIndexId)
+	assert.True(t, hostinfo.out.Load())
+	assert.True(t, hostinfo.in.Load())
 
 	// Do a traffic check tick, should not be pending deletion but should not have any in/out packets recorded
 	nc.doTrafficCheck(hostinfo.localIndexId, p, nb, out, time.Now())
-	assert.NotContains(t, nc.pendingDeletion, hostinfo.localIndexId)
+	assert.False(t, hostinfo.pendingDeletion.Load())
-	assert.NotContains(t, nc.out, hostinfo.localIndexId)
+	assert.False(t, hostinfo.out.Load())
-	assert.NotContains(t, nc.in, hostinfo.localIndexId)
+	assert.False(t, hostinfo.in.Load())
 
 	// Do another traffic check tick, this host should be pending deletion now
-	nc.Out(hostinfo.localIndexId)
+	nc.Out(hostinfo)
+	assert.True(t, hostinfo.out.Load())
 	nc.doTrafficCheck(hostinfo.localIndexId, p, nb, out, time.Now())
-	assert.Contains(t, nc.pendingDeletion, hostinfo.localIndexId)
+	assert.True(t, hostinfo.pendingDeletion.Load())
-	assert.NotContains(t, nc.out, hostinfo.localIndexId)
+	assert.False(t, hostinfo.out.Load())
-	assert.NotContains(t, nc.in, hostinfo.localIndexId)
+	assert.False(t, hostinfo.in.Load())
 	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
 	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnIp)
 
 	// Do a final traffic check tick, the host should now be removed
 	nc.doTrafficCheck(hostinfo.localIndexId, p, nb, out, time.Now())
-	assert.NotContains(t, nc.pendingDeletion, hostinfo.localIndexId)
 	assert.NotContains(t, nc.hostMap.Hosts, hostinfo.vpnIp)
 	assert.NotContains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
 }
@@ -148,10 +148,10 @@ func Test_NewConnectionManagerTest2(t *testing.T) {
 	ifce.pki.cs.Store(cs)
 
 	// Create manager
-	ctx, cancel := context.WithCancel(context.Background())
+	conf := config.NewC(l)
-	defer cancel()
+	punchy := NewPunchyFromConfig(l, conf)
-	punchy := NewPunchyFromConfig(l, config.NewC(l))
+	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
-	nc := newConnectionManager(ctx, l, ifce, 5, 10, punchy)
+	nc.intf = ifce
 	p := []byte("")
 	nb := make([]byte, 12, 12)
 	out := make([]byte, mtu)
@@ -169,33 +169,130 @@ func Test_NewConnectionManagerTest2(t *testing.T) {
 	nc.hostMap.unlockedAddHostInfo(hostinfo, ifce)
 
 	// We saw traffic out to vpnIp
-	nc.Out(hostinfo.localIndexId)
+	nc.Out(hostinfo)
-	nc.In(hostinfo.localIndexId)
+	nc.In(hostinfo)
-	assert.NotContains(t, nc.pendingDeletion, hostinfo.vpnIp)
+	assert.True(t, hostinfo.in.Load())
+	assert.True(t, hostinfo.out.Load())
+	assert.False(t, hostinfo.pendingDeletion.Load())
 	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnIp)
 	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
 
 	// Do a traffic check tick, should not be pending deletion but should not have any in/out packets recorded
 	nc.doTrafficCheck(hostinfo.localIndexId, p, nb, out, time.Now())
-	assert.NotContains(t, nc.pendingDeletion, hostinfo.localIndexId)
+	assert.False(t, hostinfo.pendingDeletion.Load())
-	assert.NotContains(t, nc.out, hostinfo.localIndexId)
+	assert.False(t, hostinfo.out.Load())
-	assert.NotContains(t, nc.in, hostinfo.localIndexId)
+	assert.False(t, hostinfo.in.Load())
 
 	// Do another traffic check tick, this host should be pending deletion now
-	nc.Out(hostinfo.localIndexId)
+	nc.Out(hostinfo)
 	nc.doTrafficCheck(hostinfo.localIndexId, p, nb, out, time.Now())
-	assert.Contains(t, nc.pendingDeletion, hostinfo.localIndexId)
+	assert.True(t, hostinfo.pendingDeletion.Load())
-	assert.NotContains(t, nc.out, hostinfo.localIndexId)
+	assert.False(t, hostinfo.out.Load())
-	assert.NotContains(t, nc.in, hostinfo.localIndexId)
+	assert.False(t, hostinfo.in.Load())
 	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
 	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnIp)
 
 	// We saw traffic, should no longer be pending deletion
-	nc.In(hostinfo.localIndexId)
+	nc.In(hostinfo)
 	nc.doTrafficCheck(hostinfo.localIndexId, p, nb, out, time.Now())
-	assert.NotContains(t, nc.pendingDeletion, hostinfo.localIndexId)
+	assert.False(t, hostinfo.pendingDeletion.Load())
-	assert.NotContains(t, nc.out, hostinfo.localIndexId)
+	assert.False(t, hostinfo.out.Load())
-	assert.NotContains(t, nc.in, hostinfo.localIndexId)
+	assert.False(t, hostinfo.in.Load())
+	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
+	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnIp)
+}
+
+func Test_NewConnectionManager_DisconnectInactive(t *testing.T) {
+	l := test.NewLogger()
+	vpncidr := netip.MustParsePrefix("172.1.1.1/24")
+	localrange := netip.MustParsePrefix("10.1.1.1/24")
+	vpnIp := netip.MustParseAddr("172.1.1.2")
+	preferredRanges := []netip.Prefix{localrange}
+
+	// Very incomplete mock objects
+	hostMap := newHostMap(l, vpncidr)
+	hostMap.preferredRanges.Store(&preferredRanges)
+
+	cs := &CertState{
+		RawCertificate:      []byte{},
+		PrivateKey:          []byte{},
+		Certificate:         &cert.NebulaCertificate{},
+		RawCertificateNoKey: []byte{},
+	}
+
+	lh := newTestLighthouse()
+	ifce := &Interface{
+		hostMap:          hostMap,
+		inside:           &test.NoopTun{},
+		outside:          &udp.NoopConn{},
+		firewall:         &Firewall{},
+		lightHouse:       lh,
+		pki:              &PKI{},
+		handshakeManager: NewHandshakeManager(l, hostMap, lh, &udp.NoopConn{}, defaultHandshakeConfig),
+		l:                l,
+	}
+	ifce.pki.cs.Store(cs)
+
+	// Create manager
+	conf := config.NewC(l)
+	conf.Settings["tunnels"] = map[interface{}]interface{}{
+		"drop_inactive": true,
+	}
+	punchy := NewPunchyFromConfig(l, conf)
+	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
+	assert.True(t, nc.dropInactive.Load())
+	nc.intf = ifce
+
+	// Add an ip we have established a connection w/ to hostmap
+	hostinfo := &HostInfo{
+		vpnIp:         vpnIp,
+		localIndexId:  1099,
+		remoteIndexId: 9901,
+	}
+	hostinfo.ConnectionState = &ConnectionState{
+		myCert: &cert.NebulaCertificate{},
+		H:      &noise.HandshakeState{},
+	}
+	nc.hostMap.unlockedAddHostInfo(hostinfo, ifce)
+
+	// Do a traffic check tick, in and out should be cleared but should not be pending deletion
+	nc.Out(hostinfo)
+	nc.In(hostinfo)
+	assert.True(t, hostinfo.out.Load())
+	assert.True(t, hostinfo.in.Load())
+
+	now := time.Now()
+	decision, _, _ := nc.makeTrafficDecision(hostinfo.localIndexId, now)
+	assert.Equal(t, tryRehandshake, decision)
+	assert.Equal(t, now, hostinfo.lastUsed)
+	assert.False(t, hostinfo.pendingDeletion.Load())
+	assert.False(t, hostinfo.out.Load())
+	assert.False(t, hostinfo.in.Load())
+
+	decision, _, _ = nc.makeTrafficDecision(hostinfo.localIndexId, now.Add(time.Second*5))
+	assert.Equal(t, doNothing, decision)
+	assert.Equal(t, now, hostinfo.lastUsed)
+	assert.False(t, hostinfo.pendingDeletion.Load())
+	assert.False(t, hostinfo.out.Load())
+	assert.False(t, hostinfo.in.Load())
+
+	// Do another traffic check tick, should still not be pending deletion
+	decision, _, _ = nc.makeTrafficDecision(hostinfo.localIndexId, now.Add(time.Second*10))
+	assert.Equal(t, doNothing, decision)
+	assert.Equal(t, now, hostinfo.lastUsed)
+	assert.False(t, hostinfo.pendingDeletion.Load())
+	assert.False(t, hostinfo.out.Load())
+	assert.False(t, hostinfo.in.Load())
+	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
+	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnIp)
+
+	// Finally advance beyond the inactivity timeout
+	decision, _, _ = nc.makeTrafficDecision(hostinfo.localIndexId, now.Add(time.Minute*10))
+	assert.Equal(t, closeTunnel, decision)
+	assert.Equal(t, now, hostinfo.lastUsed)
+	assert.False(t, hostinfo.pendingDeletion.Load())
+	assert.False(t, hostinfo.out.Load())
+	assert.False(t, hostinfo.in.Load())
 	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
 	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnIp)
 }
@@ -273,10 +370,10 @@ func Test_NewConnectionManagerTest_DisconnectInvalid(t *testing.T) {
 	ifce.disconnectInvalid.Store(true)
 
 	// Create manager
-	ctx, cancel := context.WithCancel(context.Background())
+	conf := config.NewC(l)
-	defer cancel()
+	punchy := NewPunchyFromConfig(l, conf)
-	punchy := NewPunchyFromConfig(l, config.NewC(l))
+	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
-	nc := newConnectionManager(ctx, l, ifce, 5, 10, punchy)
+	nc.intf = ifce
 	ifce.connectionManager = nc
 
 	hostinfo := &HostInfo{

control.go

@@ -26,14 +26,15 @@ type controlHostLister interface {
 }
 
 type Control struct {
-	f               *Interface
+	f                      *Interface
-	l               *logrus.Logger
+	l                      *logrus.Logger
-	ctx             context.Context
+	ctx                    context.Context
-	cancel          context.CancelFunc
+	cancel                 context.CancelFunc
-	sshStart        func()
+	sshStart               func()
-	statsStart      func()
+	statsStart             func()
-	dnsStart        func()
+	dnsStart               func()
-	lighthouseStart func()
+	lighthouseStart        func()
+	connectionManagerStart func(context.Context)
 }
 
 type ControlHostInfo struct {
@@ -63,6 +64,9 @@ func (c *Control) Start() {
 	if c.dnsStart != nil {
 		go c.dnsStart()
 	}
+	if c.connectionManagerStart != nil {
+		go c.connectionManagerStart(c.ctx)
+	}
 	if c.lighthouseStart != nil {
 		c.lighthouseStart()
 	}

control_test.go

@@ -66,7 +66,7 @@ func TestControl_GetHostInfoByVpnIp(t *testing.T) {
 		localIndexId:  201,
 		vpnIp:         vpnIp,
 		relayState: RelayState{
-			relays:        map[netip.Addr]struct{}{},
+			relays:        nil,
 			relayForByIp:  map[netip.Addr]*Relay{},
 			relayForByIdx: map[uint32]*Relay{},
 		},
@@ -85,7 +85,7 @@ func TestControl_GetHostInfoByVpnIp(t *testing.T) {
 		localIndexId:  201,
 		vpnIp:         vpnIp2,
 		relayState: RelayState{
-			relays:        map[netip.Addr]struct{}{},
+			relays:        nil,
 			relayForByIp:  map[netip.Addr]*Relay{},
 			relayForByIdx: map[uint32]*Relay{},
 		},

e2e/handshakes_test.go

@@ -4,7 +4,6 @@
 package e2e
 
 import (
-	"fmt"
 	"net/netip"
 	"slices"
 	"testing"
@@ -414,7 +413,7 @@ func TestReestablishRelays(t *testing.T) {
 	curIndexes := len(myControl.GetHostmap().Indexes)
 	for curIndexes >= start {
 		curIndexes = len(myControl.GetHostmap().Indexes)
-		r.Logf("Wait for the dead index to go away:start=%v indexes, currnet=%v indexes", start, curIndexes)
+		r.Logf("Wait for the dead index to go away:start=%v indexes, current=%v indexes", start, curIndexes)
 		myControl.InjectTunUDPPacket(theirVpnIpNet.Addr(), 80, 80, []byte("Hi from me should fail"))
 
 		r.RouteForAllExitFunc(func(p *udp.Packet, c *nebula.Control) router.ExitType {
@@ -964,9 +963,8 @@ func TestRehandshakingLoser(t *testing.T) {
 	t.Log("Stand up a tunnel between me and them")
 	assertTunnel(t, myVpnIpNet.Addr(), theirVpnIpNet.Addr(), myControl, theirControl, r)
 
-	tt1 := myControl.GetHostInfoByVpnIp(theirVpnIpNet.Addr(), false)
+	myControl.GetHostInfoByVpnIp(theirVpnIpNet.Addr(), false)
-	tt2 := theirControl.GetHostInfoByVpnIp(myVpnIpNet.Addr(), false)
+	theirControl.GetHostInfoByVpnIp(myVpnIpNet.Addr(), false)
-	fmt.Println(tt1.LocalIndex, tt2.LocalIndex)
 
 	r.RenderHostmaps("Starting hostmaps", myControl, theirControl)
 

e2e/router/router.go

@@ -690,6 +690,7 @@ func (r *R) FlushAll() {
 			r.Unlock()
 			panic("Can't FlushAll for host: " + p.To.String())
 		}
+		receiver.InjectUDPPacket(p)
 		r.Unlock()
 	}
 }

e2e/tunnels_test.go

@@ -0,0 +1,55 @@
+//go:build e2e_testing
+// +build e2e_testing
+
+package e2e
+
+import (
+	"testing"
+	"time"
+
+	"github.com/slackhq/nebula/e2e/router"
+)
+
+func TestDropInactiveTunnels(t *testing.T) {
+	// The goal of this test is to ensure the shortest inactivity timeout will close the tunnel on both sides
+	// under ideal conditions
+	ca, _, caKey, _ := NewTestCaCert(time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
+	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(ca, caKey, "me", "10.128.0.1/24", m{"tunnels": m{"drop_inactive": true, "inactivity_timeout": "5s"}})
+	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(ca, caKey, "them", "10.128.0.2/24", m{"tunnels": m{"drop_inactive": true, "inactivity_timeout": "10m"}})
+
+	// Share our underlay information
+	myControl.InjectLightHouseAddr(theirVpnIpNet.Addr(), theirUdpAddr)
+	theirControl.InjectLightHouseAddr(myVpnIpNet.Addr(), myUdpAddr)
+
+	// Start the servers
+	myControl.Start()
+	theirControl.Start()
+
+	r := router.NewR(t, myControl, theirControl)
+
+	r.Log("Assert the tunnel between me and them works")
+	assertTunnel(t, myVpnIpNet.Addr(), theirVpnIpNet.Addr(), myControl, theirControl, r)
+
+	r.Log("Go inactive and wait for the tunnels to get dropped")
+	waitStart := time.Now()
+	for {
+		myIndexes := len(myControl.GetHostmap().Indexes)
+		theirIndexes := len(theirControl.GetHostmap().Indexes)
+		if myIndexes == 0 && theirIndexes == 0 {
+			break
+		}
+
+		since := time.Since(waitStart)
+		r.Logf("my tunnels: %v; their tunnels: %v; duration: %v", myIndexes, theirIndexes, since)
+		if since > time.Second*30 {
+			t.Fatal("Tunnel should have been declared inactive after 5 seconds and before 30 seconds")
+		}
+
+		time.Sleep(1 * time.Second)
+		r.FlushAll()
+	}
+
+	r.Logf("Inactive tunnels were dropped within %v", time.Since(waitStart))
+	myControl.Stop()
+	theirControl.Stop()
+}

examples/config.yml

@@ -303,6 +303,18 @@ logging:
   # after receiving the response for lighthouse queries
   #trigger_buffer: 64
 
+# Tunnel manager settings
+#tunnels:
+  # drop_inactive controls whether inactive tunnels are maintained or dropped after the inactive_timeout period has
+  # elapsed.
+  # In general, it is a good idea to enable this setting. It will be enabled by default in a future release.
+  # This setting is reloadable
+  #drop_inactive: false
+
+  # inactivity_timeout controls how long a tunnel MUST NOT see any inbound or outbound traffic before being considered
+  # inactive and eligible to be dropped.
+  # This setting is reloadable
+  #inactivity_timeout: 10m
 
 # Nebula security group configuration
 firewall:

firewall.go

@@ -862,16 +862,13 @@ func (fr *FirewallRule) match(p firewall.Packet, c *cert.NebulaCertificate) bool
 		}
 	}
 
-	matched := false
+	for _, v := range fr.CIDR.Supernets(netip.PrefixFrom(p.RemoteIP, p.RemoteIP.BitLen())) {
-	prefix := netip.PrefixFrom(p.RemoteIP, p.RemoteIP.BitLen())
+		if v.match(p, c) {
-	fr.CIDR.EachLookupPrefix(prefix, func(prefix netip.Prefix, val *firewallLocalCIDR) bool {
+			return true
-		if prefix.Contains(p.RemoteIP) && val.match(p, c) {
-			matched = true
-			return false
 		}
-		return true
+	}
-	})
+
-	return matched
+	return false
 }
 
 func (flc *firewallLocalCIDR) addRule(f *Firewall, localIp netip.Prefix) error {

go.mod

@@ -1,8 +1,6 @@
 module github.com/slackhq/nebula
 
-go 1.22.0
+go 1.24.0
-
-toolchain go1.22.2
 
 require (
 	dario.cat/mergo v1.0.0
@@ -10,7 +8,7 @@ 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 v0.11.1
+	github.com/gaissmai/bart v0.25.1
 	github.com/gogo/protobuf v1.3.2
 	github.com/google/gopacket v1.1.19
 	github.com/kardianos/service v1.2.2
@@ -23,12 +21,12 @@ require (
 	github.com/songgao/water v0.0.0-20200317203138-2b4b6d7c09d8
 	github.com/stretchr/testify v1.9.0
 	github.com/vishvananda/netlink v1.2.1-beta.2
-	golang.org/x/crypto v0.26.0
+	golang.org/x/crypto v0.43.0
 	golang.org/x/exp v0.0.0-20230725093048-515e97ebf090
-	golang.org/x/net v0.28.0
+	golang.org/x/net v0.45.0
 	golang.org/x/sync v0.8.0
-	golang.org/x/sys v0.24.0
+	golang.org/x/sys v0.37.0
-	golang.org/x/term v0.23.0
+	golang.org/x/term v0.36.0
 	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
@@ -39,7 +37,6 @@ require (
 
 require (
 	github.com/beorn7/perks v1.0.1 // indirect
-	github.com/bits-and-blooms/bitset v1.13.0 // indirect
 	github.com/cespare/xxhash/v2 v2.2.0 // indirect
 	github.com/davecgh/go-spew v1.1.1 // indirect
 	github.com/google/btree v1.1.2 // indirect

go.sum

@@ -14,8 +14,6 @@ github.com/beorn7/perks v0.0.0-20180321164747-3a771d992973/go.mod h1:Dwedo/Wpr24
 github.com/beorn7/perks v1.0.0/go.mod h1:KWe93zE9D1o94FZ5RNwFwVgaQK1VOXiVxmqh+CedLV8=
 github.com/beorn7/perks v1.0.1 h1:VlbKKnNfV8bJzeqoa4cOKqO6bYr3WgKZxO8Z16+hsOM=
 github.com/beorn7/perks v1.0.1/go.mod h1:G2ZrVWU2WbWT9wwq4/hrbKbnv/1ERSJQ0ibhJ6rlkpw=
-github.com/bits-and-blooms/bitset v1.13.0 h1:bAQ9OPNFYbGHV6Nez0tmNI0RiEu7/hxlYJRUA0wFAVE=
-github.com/bits-and-blooms/bitset v1.13.0/go.mod h1:7hO7Gc7Pp1vODcmWvKMRA9BNmbv6a/7QIWpPxHddWR8=
 github.com/cespare/xxhash/v2 v2.1.1/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs=
 github.com/cespare/xxhash/v2 v2.2.0 h1:DC2CZ1Ep5Y4k3ZQ899DldepgrayRUGE6BBZ/cd9Cj44=
 github.com/cespare/xxhash/v2 v2.2.0/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs=
@@ -26,8 +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 v0.11.1 h1:5Uv5XwsaFBRo4E5VBcb9TzY8B7zxFf+U7isDxqOrRfc=
+github.com/gaissmai/bart v0.25.1 h1:ctl4nH/za+trcbfmZz9uO9xGBZD684GWaDMsjMSo1l8=
-github.com/gaissmai/bart v0.11.1/go.mod h1:KHeYECXQiBjTzQz/om2tqn3sZF1J7hw9m6z41ftj3fg=
+github.com/gaissmai/bart v0.25.1/go.mod h1:GREWQfTLRWz/c5FTOsIw+KkscuFkIV5t8Rp7Nd1Td5c=
 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=
@@ -151,8 +149,8 @@ golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACk
 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 v0.26.0 h1:RrRspgV4mU+YwB4FYnuBoKsUapNIL5cohGAmSH3azsw=
+golang.org/x/crypto v0.43.0 h1:dduJYIi3A3KOfdGOHX8AVZ/jGiyPa3IbBozJ5kNuE04=
-golang.org/x/crypto v0.26.0/go.mod h1:GY7jblb9wI+FOo5y8/S2oY4zWP07AkOJ4+jxCqdqn54=
+golang.org/x/crypto v0.43.0/go.mod h1:BFbav4mRNlXJL4wNeejLpWxB7wMbc79PdRGhWKncxR0=
 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=
@@ -171,8 +169,8 @@ 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 v0.28.0 h1:a9JDOJc5GMUJ0+UDqmLT86WiEy7iWyIhz8gz8E4e5hE=
+golang.org/x/net v0.45.0 h1:RLBg5JKixCy82FtLJpeNlVM0nrSqpCRYzVU1n8kj0tM=
-golang.org/x/net v0.28.0/go.mod h1:yqtgsTWOOnlGLG9GFRrK3++bGOUEkNBoHZc8MEDWPNg=
+golang.org/x/net v0.45.0/go.mod h1:ECOoLqd5U3Lhyeyo/QDCEVQ4sNgYsqvCZ722XogGieY=
 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=
@@ -199,11 +197,11 @@ golang.org/x/sys v0.0.0-20201119102817-f84b799fce68/go.mod h1:h1NjWce9XRLGQEsW7w
 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 v0.24.0 h1:Twjiwq9dn6R1fQcyiK+wQyHWfaz/BJB+YIpzU/Cv3Xg=
+golang.org/x/sys v0.37.0 h1:fdNQudmxPjkdUTPnLn5mdQv7Zwvbvpaxqs831goi9kQ=
-golang.org/x/sys v0.24.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
+golang.org/x/sys v0.37.0/go.mod h1:OgkHotnGiDImocRcuBABYBEXf8A9a87e/uXjp9XT3ks=
 golang.org/x/term v0.0.0-20201126162022-7de9c90e9dd1/go.mod h1:bj7SfCRtBDWHUb9snDiAeCFNEtKQo2Wmx5Cou7ajbmo=
-golang.org/x/term v0.23.0 h1:F6D4vR+EHoL9/sWAWgAR1H2DcHr4PareCbAaCo1RpuU=
+golang.org/x/term v0.36.0 h1:zMPR+aF8gfksFprF/Nc/rd1wRS1EI6nDBGyWAvDzx2Q=
-golang.org/x/term v0.23.0/go.mod h1:DgV24QBUrK6jhZXl+20l6UWznPlwAHm1Q1mGHtydmSk=
+golang.org/x/term v0.36.0/go.mod h1:Qu394IJq6V6dCBRgwqshf3mPF85AqzYEzofzRdZkWss=
 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=

handshake_ix.go

@@ -151,7 +151,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 		HandshakePacket:   make(map[uint8][]byte, 0),
 		lastHandshakeTime: hs.Details.Time,
 		relayState: RelayState{
-			relays:        map[netip.Addr]struct{}{},
+			relays:        nil,
 			relayForByIp:  map[netip.Addr]*Relay{},
 			relayForByIdx: map[uint32]*Relay{},
 		},
@@ -335,7 +335,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 			Info("Handshake message sent")
 	}
 
-	f.connectionManager.AddTrafficWatch(hostinfo.localIndexId)
+	f.connectionManager.AddTrafficWatch(hostinfo)
 
 	hostinfo.remotes.ResetBlockedRemotes()
 
@@ -493,7 +493,7 @@ func ixHandshakeStage2(f *Interface, addr netip.AddrPort, via *ViaSender, hh *Ha
 
 	// Complete our handshake and update metrics, this will replace any existing tunnels for this vpnIp
 	f.handshakeManager.Complete(hostinfo, f)
-	f.connectionManager.AddTrafficWatch(hostinfo.localIndexId)
+	f.connectionManager.AddTrafficWatch(hostinfo)
 
 	if f.l.Level >= logrus.DebugLevel {
 		hostinfo.logger(f.l).Debugf("Sending %d stored packets", len(hh.packetStore))

handshake_manager.go

@@ -403,7 +403,7 @@ func (hm *HandshakeManager) StartHandshake(vpnIp netip.Addr, cacheCb func(*Hands
 		vpnIp:           vpnIp,
 		HandshakePacket: make(map[uint8][]byte, 0),
 		relayState: RelayState{
-			relays:        map[netip.Addr]struct{}{},
+			relays:        nil,
 			relayForByIp:  map[netip.Addr]*Relay{},
 			relayForByIdx: map[uint32]*Relay{},
 		},

hostmap.go

@@ -4,6 +4,7 @@ import (
 	"errors"
 	"net"
 	"net/netip"
+	"slices"
 	"sync"
 	"sync/atomic"
 	"time"
@@ -21,7 +22,6 @@ const defaultPromoteEvery = 1000       // Count of packets sent before we try mo
 const defaultReQueryEvery = 5000       // Count of packets sent before re-querying a hostinfo to the lighthouse
 const defaultReQueryWait = time.Minute // Minimum amount of seconds to wait before re-querying a hostinfo the lighthouse. Evaluated every ReQueryEvery
 const MaxRemotes = 10
-const maxRecvError = 4
 
 // MaxHostInfosPerVpnIp is the max number of hostinfos we will track for a given vpn ip
 // 5 allows for an initial handshake and each host pair re-handshaking twice
@@ -69,15 +69,20 @@ type HostMap struct {
 type RelayState struct {
 	sync.RWMutex
 
-	relays        map[netip.Addr]struct{} // Set of VpnIp's of Hosts to use as relays to access this peer
+	relays        []netip.Addr          // Ordered set of VpnIp's of Hosts to use as relays to access this peer
-	relayForByIp  map[netip.Addr]*Relay   // Maps VpnIps of peers for which this HostInfo is a relay to some Relay info
+	relayForByIp  map[netip.Addr]*Relay // Maps VpnIps of peers for which this HostInfo is a relay to some Relay info
-	relayForByIdx map[uint32]*Relay       // Maps a local index to some Relay info
+	relayForByIdx map[uint32]*Relay     // Maps a local index to some Relay info
 }
 
 func (rs *RelayState) DeleteRelay(ip netip.Addr) {
 	rs.Lock()
 	defer rs.Unlock()
-	delete(rs.relays, ip)
+	for idx, val := range rs.relays {
+		if val == ip {
+			rs.relays = append(rs.relays[:idx], rs.relays[idx+1:]...)
+			return
+		}
+	}
 }
 
 func (rs *RelayState) UpdateRelayForByIpState(vpnIp netip.Addr, state int) {
@@ -122,16 +127,16 @@ func (rs *RelayState) GetRelayForByIp(ip netip.Addr) (*Relay, bool) {
 func (rs *RelayState) InsertRelayTo(ip netip.Addr) {
 	rs.Lock()
 	defer rs.Unlock()
-	rs.relays[ip] = struct{}{}
+	if !slices.Contains(rs.relays, ip) {
+		rs.relays = append(rs.relays, ip)
+	}
 }
 
 func (rs *RelayState) CopyRelayIps() []netip.Addr {
+	ret := make([]netip.Addr, len(rs.relays))
 	rs.RLock()
 	defer rs.RUnlock()
-	ret := make([]netip.Addr, 0, len(rs.relays))
+	copy(ret, rs.relays)
-	for ip := range rs.relays {
-		ret = append(ret, ip)
-	}
 	return ret
 }
 
@@ -214,7 +219,6 @@ type HostInfo struct {
 	remoteIndexId   uint32
 	localIndexId    uint32
 	vpnIp           netip.Addr
-	recvError       atomic.Uint32
 	remoteCidr      *bart.Table[struct{}]
 	relayState      RelayState
 
@@ -242,6 +246,14 @@ type HostInfo struct {
 	// Used to track other hostinfos for this vpn ip since only 1 can be primary
 	// Synchronised via hostmap lock and not the hostinfo lock.
 	next, prev *HostInfo
+
+	//TODO: in, out, and others might benefit from being an atomic.Int32. We could collapse connectionManager pendingDeletion, relayUsed, and in/out into this 1 thing
+	in, out, pendingDeletion atomic.Bool
+
+	// lastUsed tracks the last time ConnectionManager checked the tunnel and it was in use.
+	// This value will be behind against actual tunnel utilization in the hot path.
+	// This should only be used by the ConnectionManagers ticker routine.
+	lastUsed time.Time
 }
 
 type ViaSender struct {
@@ -691,13 +703,6 @@ func (i *HostInfo) SetRemoteIfPreferred(hm *HostMap, newRemote netip.AddrPort) b
 	return false
 }
 
-func (i *HostInfo) RecvErrorExceeded() bool {
-	if i.recvError.Add(1) >= maxRecvError {
-		return true
-	}
-	return true
-}
-
 func (i *HostInfo) CreateRemoteCIDR(c *cert.NebulaCertificate) {
 	if len(c.Details.Ips) == 1 && len(c.Details.Subnets) == 0 {
 		// Simple case, no CIDRTree needed
@@ -709,8 +714,7 @@ func (i *HostInfo) CreateRemoteCIDR(c *cert.NebulaCertificate) {
 		//TODO: IPV6-WORK what to do when ip is invalid?
 		nip, _ := netip.AddrFromSlice(ip.IP)
 		nip = nip.Unmap()
-		bits, _ := ip.Mask.Size()
+		remoteCidr.Insert(netip.PrefixFrom(nip, nip.BitLen()), struct{}{})
-		remoteCidr.Insert(netip.PrefixFrom(nip, bits), struct{}{})
 	}
 
 	for _, n := range c.Details.Subnets {

hostmap_test.go

@@ -1,12 +1,15 @@
 package nebula
 
 import (
+	"net"
 	"net/netip"
 	"testing"
 
+	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/test"
 	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 )
 
 func TestHostMap_MakePrimary(t *testing.T) {
@@ -86,6 +89,40 @@ func TestHostMap_MakePrimary(t *testing.T) {
 	assert.Nil(t, h2.next)
 }
 
+func TestHostInfo_CreateRemoteCIDR(t *testing.T) {
+	h := HostInfo{}
+	c := &cert.NebulaCertificate{
+		Details: cert.NebulaCertificateDetails{
+			Ips: []*net.IPNet{
+				{
+					IP:   net.IPv4(1, 2, 3, 4),
+					Mask: net.IPv4Mask(255, 255, 255, 0),
+				},
+			},
+		},
+	}
+
+	// remoteCidr should be empty with only 1 ip address present in the certificate
+	h.CreateRemoteCIDR(c)
+	assert.Empty(t, h.remoteCidr)
+
+	// remoteCidr should be populated if there is also a subnet in the certificate
+	c.Details.Subnets = []*net.IPNet{
+		{
+			IP:   net.IPv4(9, 2, 3, 4),
+			Mask: net.IPv4Mask(255, 255, 255, 0),
+		},
+	}
+	h.CreateRemoteCIDR(c)
+	assert.NotEmpty(t, h.remoteCidr)
+	_, ok := h.remoteCidr.Lookup(netip.MustParseAddr("1.2.3.0"))
+	assert.False(t, ok, "An ip address within the certificates network should not be found")
+	_, ok = h.remoteCidr.Lookup(netip.MustParseAddr("1.2.3.4"))
+	assert.True(t, ok, "An exact ip address match should be found")
+	_, ok = h.remoteCidr.Lookup(netip.MustParseAddr("9.2.3.4"))
+	assert.True(t, ok, "An ip address within the subnets should be found")
+}
+
 func TestHostMap_DeleteHostInfo(t *testing.T) {
 	l := test.NewLogger()
 	hm := newHostMap(
@@ -225,3 +262,31 @@ func TestHostMap_reload(t *testing.T) {
 	c.ReloadConfigString("preferred_ranges: [1.1.1.1/32]")
 	assert.EqualValues(t, []string{"1.1.1.1/32"}, toS(hm.GetPreferredRanges()))
 }
+
+func TestHostMap_RelayState(t *testing.T) {
+	h1 := &HostInfo{vpnIp: netip.MustParseAddr("0.0.0.1"), localIndexId: 1}
+	a1 := netip.MustParseAddr("::1")
+	a2 := netip.MustParseAddr("2001::1")
+
+	h1.relayState.InsertRelayTo(a1)
+	assert.Equal(t, h1.relayState.relays, []netip.Addr{a1})
+	h1.relayState.InsertRelayTo(a2)
+	assert.Equal(t, h1.relayState.relays, []netip.Addr{a1, a2})
+	// Ensure that the first relay added is the first one returned in the copy
+	currentRelays := h1.relayState.CopyRelayIps()
+	require.Len(t, currentRelays, 2)
+	assert.Equal(t, currentRelays[0], a1)
+
+	// Deleting the last one in the list works ok
+	h1.relayState.DeleteRelay(a2)
+	assert.Equal(t, h1.relayState.relays, []netip.Addr{a1})
+
+	// Deleting an element not in the list works ok
+	h1.relayState.DeleteRelay(a2)
+	assert.Equal(t, h1.relayState.relays, []netip.Addr{a1})
+
+	// Deleting the only element in the list works ok
+	h1.relayState.DeleteRelay(a1)
+	assert.Equal(t, h1.relayState.relays, []netip.Addr{})
+
+}

inside.go

@@ -213,7 +213,7 @@ func (f *Interface) SendVia(via *HostInfo,
 	c := via.ConnectionState.messageCounter.Add(1)
 
 	out = header.Encode(out, header.Version, header.Message, header.MessageRelay, relay.RemoteIndex, c)
-	f.connectionManager.Out(via.localIndexId)
+	f.connectionManager.Out(via)
 
 	// Authenticate the header and payload, but do not encrypt for this message type.
 	// The payload consists of the inner, unencrypted Nebula header, as well as the end-to-end encrypted payload.
@@ -282,7 +282,7 @@ func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType
 
 	//l.WithField("trace", string(debug.Stack())).Error("out Header ", &Header{Version, t, st, 0, hostinfo.remoteIndexId, c}, p)
 	out = header.Encode(out, header.Version, t, st, hostinfo.remoteIndexId, c)
-	f.connectionManager.Out(hostinfo.localIndexId)
+	f.connectionManager.Out(hostinfo)
 
 	// Query our LH if we haven't since the last time we've been rebound, this will cause the remote to punch against
 	// all our IPs and enable a faster roaming.

interface.go

@@ -24,24 +24,23 @@ import (
 const mtu = 9001
 
 type InterfaceConfig struct {
-	HostMap                 *HostMap
+	HostMap            *HostMap
-	Outside                 udp.Conn
+	Outside            udp.Conn
-	Inside                  overlay.Device
+	Inside             overlay.Device
-	pki                     *PKI
+	pki                *PKI
-	Cipher                  string
+	Cipher             string
-	Firewall                *Firewall
+	Firewall           *Firewall
-	ServeDns                bool
+	ServeDns           bool
-	HandshakeManager        *HandshakeManager
+	HandshakeManager   *HandshakeManager
-	lightHouse              *LightHouse
+	lightHouse         *LightHouse
-	checkInterval           time.Duration
+	connectionManager  *connectionManager
-	pendingDeletionInterval time.Duration
+	DropLocalBroadcast bool
-	DropLocalBroadcast      bool
+	DropMulticast      bool
-	DropMulticast           bool
+	routines           int
-	routines                int
+	MessageMetrics     *MessageMetrics
-	MessageMetrics          *MessageMetrics
+	version            string
-	version                 string
+	relayManager       *relayManager
-	relayManager            *relayManager
+	punchy             *Punchy
-	punchy                  *Punchy
 
 	tryPromoteEvery uint32
 	reQueryEvery    uint32
@@ -154,6 +153,9 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 	if c.Firewall == nil {
 		return nil, errors.New("no firewall rules")
 	}
+	if c.connectionManager == nil {
+		return nil, errors.New("no connection manager")
+	}
 
 	certificate := c.pki.GetCertState().Certificate
 
@@ -196,6 +198,7 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 		readers:            make([]io.ReadWriteCloser, c.routines),
 		myVpnNet:           myVpnNet,
 		relayManager:       c.relayManager,
+		connectionManager:  c.connectionManager,
 
 		conntrackCacheTimeout: c.ConntrackCacheTimeout,
 
@@ -219,7 +222,7 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 	ifce.reQueryEvery.Store(c.reQueryEvery)
 	ifce.reQueryWait.Store(int64(c.reQueryWait))
 
-	ifce.connectionManager = newConnectionManager(ctx, c.l, ifce, c.checkInterval, c.pendingDeletionInterval, c.punchy)
+	ifce.connectionManager.intf = ifce
 
 	return ifce, nil
 }

main.go

@@ -199,6 +199,7 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 
 	hostMap := NewHostMapFromConfig(l, tunCidr, c)
 	punchy := NewPunchyFromConfig(l, c)
+	connManager := newConnectionManagerFromConfig(l, c, hostMap, punchy)
 	lightHouse, err := NewLightHouseFromConfig(ctx, l, c, tunCidr, udpConns[0], punchy)
 	if err != nil {
 		return nil, util.ContextualizeIfNeeded("Failed to initialize lighthouse handler", err)
@@ -234,31 +235,27 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		}
 	}
 
-	checkInterval := c.GetInt("timers.connection_alive_interval", 5)
-	pendingDeletionInterval := c.GetInt("timers.pending_deletion_interval", 10)
-
 	ifConfig := &InterfaceConfig{
-		HostMap:                 hostMap,
+		HostMap:            hostMap,
-		Inside:                  tun,
+		Inside:             tun,
-		Outside:                 udpConns[0],
+		Outside:            udpConns[0],
-		pki:                     pki,
+		pki:                pki,
-		Cipher:                  c.GetString("cipher", "aes"),
+		Cipher:             c.GetString("cipher", "aes"),
-		Firewall:                fw,
+		Firewall:           fw,
-		ServeDns:                serveDns,
+		ServeDns:           serveDns,
-		HandshakeManager:        handshakeManager,
+		HandshakeManager:   handshakeManager,
-		lightHouse:              lightHouse,
+		connectionManager:  connManager,
-		checkInterval:           time.Second * time.Duration(checkInterval),
+		lightHouse:         lightHouse,
-		pendingDeletionInterval: time.Second * time.Duration(pendingDeletionInterval),
+		tryPromoteEvery:    c.GetUint32("counters.try_promote", defaultPromoteEvery),
-		tryPromoteEvery:         c.GetUint32("counters.try_promote", defaultPromoteEvery),
+		reQueryEvery:       c.GetUint32("counters.requery_every_packets", defaultReQueryEvery),
-		reQueryEvery:            c.GetUint32("counters.requery_every_packets", defaultReQueryEvery),
+		reQueryWait:        c.GetDuration("timers.requery_wait_duration", defaultReQueryWait),
-		reQueryWait:             c.GetDuration("timers.requery_wait_duration", defaultReQueryWait),
+		DropLocalBroadcast: c.GetBool("tun.drop_local_broadcast", false),
-		DropLocalBroadcast:      c.GetBool("tun.drop_local_broadcast", false),
+		DropMulticast:      c.GetBool("tun.drop_multicast", false),
-		DropMulticast:           c.GetBool("tun.drop_multicast", false),
+		routines:           routines,
-		routines:                routines,
+		MessageMetrics:     messageMetrics,
-		MessageMetrics:          messageMetrics,
+		version:            buildVersion,
-		version:                 buildVersion,
+		relayManager:       NewRelayManager(ctx, l, hostMap, c),
-		relayManager:            NewRelayManager(ctx, l, hostMap, c),
+		punchy:             punchy,
-		punchy:                  punchy,
 
 		ConntrackCacheTimeout: conntrackCacheTimeout,
 		l:                     l,
@@ -325,5 +322,6 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		statsStart,
 		dnsStart,
 		lightHouse.StartUpdateWorker,
+		connManager.Start,
 	}, nil
 }

outside.go

@@ -102,7 +102,7 @@ func (f *Interface) readOutsidePackets(ip netip.AddrPort, via *ViaSender, out []
 			// Pull the Roaming parts up here, and return in all call paths.
 			f.handleHostRoaming(hostinfo, ip)
 			// Track usage of both the HostInfo and the Relay for the received & authenticated packet
-			f.connectionManager.In(hostinfo.localIndexId)
+			f.connectionManager.In(hostinfo)
 			f.connectionManager.RelayUsed(h.RemoteIndex)
 
 			relay, ok := hostinfo.relayState.QueryRelayForByIdx(h.RemoteIndex)
@@ -246,7 +246,7 @@ func (f *Interface) readOutsidePackets(ip netip.AddrPort, via *ViaSender, out []
 
 	f.handleHostRoaming(hostinfo, ip)
 
-	f.connectionManager.In(hostinfo.localIndexId)
+	f.connectionManager.In(hostinfo)
 }
 
 // closeTunnel closes a tunnel locally, it does not send a closeTunnel packet to the remote
@@ -286,16 +286,18 @@ func (f *Interface) handleHostRoaming(hostinfo *HostInfo, ip netip.AddrPort) {
 
 }
 
+// handleEncrypted returns true if a packet should be processed, false otherwise
 func (f *Interface) handleEncrypted(ci *ConnectionState, addr netip.AddrPort, h *header.H) bool {
-	// If connectionstate exists and the replay protector allows, process packet
+	// If connectionstate does not exist, send a recv error, if possible, to encourage a fast reconnect
-	// Else, send recv errors for 300 seconds after a restart to allow fast reconnection.
+	if ci == nil {
-	if ci == nil || !ci.window.Check(f.l, h.MessageCounter) {
 		if addr.IsValid() {
 			f.maybeSendRecvError(addr, h.RemoteIndex)
-			return false
-		} else {
-			return false
 		}
+		return false
+	}
+	// If the window check fails, refuse to process the packet, but don't send a recv error
+	if !ci.window.Check(f.l, h.MessageCounter) {
+		return false
 	}
 
 	return true
@@ -418,7 +420,7 @@ func (f *Interface) decryptToTun(hostinfo *HostInfo, messageCounter uint64, out
 		return false
 	}
 
-	f.connectionManager.In(hostinfo.localIndexId)
+	f.connectionManager.In(hostinfo)
 	_, err = f.readers[q].Write(out)
 	if err != nil {
 		f.l.WithError(err).Error("Failed to write to tun")
@@ -458,10 +460,6 @@ func (f *Interface) handleRecvError(addr netip.AddrPort, h *header.H) {
 		return
 	}
 
-	if !hostinfo.RecvErrorExceeded() {
-		return
-	}
-
 	if hostinfo.remote.IsValid() && hostinfo.remote != addr {
 		f.l.Infoln("Someone spoofing recv_errors? ", addr, hostinfo.remote)
 		return

udp/errors.go

@@ -0,0 +1,5 @@
+package udp
+
+import "errors"
+
+var ErrInvalidIPv6RemoteForSocket = errors.New("listener is IPv4, but writing to IPv6 remote")

udp/udp_darwin.go

@@ -6,17 +6,63 @@ package udp
 // Darwin support is primarily implemented in udp_generic, besides NewListenConfig
 
 import (
+	"context"
+	"encoding/binary"
+	"errors"
 	"fmt"
 	"net"
 	"net/netip"
 	"syscall"
+	"unsafe"
 
 	"github.com/sirupsen/logrus"
+	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/firewall"
+	"github.com/slackhq/nebula/header"
 	"golang.org/x/sys/unix"
 )
 
+type StdConn struct {
+	*net.UDPConn
+	isV4  bool
+	sysFd uintptr
+	l     *logrus.Logger
+}
+
+var _ Conn = &StdConn{}
+
 func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
-	return NewGenericListener(l, ip, port, multi, batch)
+	lc := NewListenConfig(multi)
+	pc, err := lc.ListenPacket(context.TODO(), "udp", net.JoinHostPort(ip.String(), fmt.Sprintf("%v", port)))
+	if err != nil {
+		return nil, err
+	}
+
+	if uc, ok := pc.(*net.UDPConn); ok {
+		c := &StdConn{UDPConn: uc, l: l}
+
+		rc, err := uc.SyscallConn()
+		if err != nil {
+			return nil, fmt.Errorf("failed to open udp socket: %w", err)
+		}
+
+		err = rc.Control(func(fd uintptr) {
+			c.sysFd = fd
+		})
+		if err != nil {
+			return nil, fmt.Errorf("failed to get udp fd: %w", err)
+		}
+
+		la, err := c.LocalAddr()
+		if err != nil {
+			return nil, err
+		}
+		c.isV4 = la.Addr().Is4()
+
+		return c, nil
+	}
+
+	return nil, fmt.Errorf("unexpected PacketConn: %T %#v", pc, pc)
 }
 
 func NewListenConfig(multi bool) net.ListenConfig {
@@ -43,16 +89,130 @@ func NewListenConfig(multi bool) net.ListenConfig {
 	}
 }
 
-func (u *GenericConn) Rebind() error {
+//go:linkname sendto golang.org/x/sys/unix.sendto
-	rc, err := u.UDPConn.SyscallConn()
+//go:noescape
-	if err != nil {
+func sendto(s int, buf []byte, flags int, to unsafe.Pointer, addrlen int32) (err error)
-		return err
+
+func (u *StdConn) WriteTo(b []byte, ap netip.AddrPort) error {
+	var sa unsafe.Pointer
+	var addrLen int32
+
+	if u.isV4 {
+		if ap.Addr().Is6() {
+			return ErrInvalidIPv6RemoteForSocket
+		}
+
+		var rsa unix.RawSockaddrInet6
+		rsa.Family = unix.AF_INET6
+		rsa.Addr = ap.Addr().As16()
+		binary.BigEndian.PutUint16((*[2]byte)(unsafe.Pointer(&rsa.Port))[:], ap.Port())
+		sa = unsafe.Pointer(&rsa)
+		addrLen = syscall.SizeofSockaddrInet4
+	} else {
+		var rsa unix.RawSockaddrInet6
+		rsa.Family = unix.AF_INET6
+		rsa.Addr = ap.Addr().As16()
+		binary.BigEndian.PutUint16((*[2]byte)(unsafe.Pointer(&rsa.Port))[:], ap.Port())
+		sa = unsafe.Pointer(&rsa)
+		addrLen = syscall.SizeofSockaddrInet6
 	}
 
-	return rc.Control(func(fd uintptr) {
+	// Golang stdlib doesn't handle EAGAIN correctly in some situations so we do writes ourselves
-		err := syscall.SetsockoptInt(int(fd), unix.IPPROTO_IPV6, unix.IPV6_BOUND_IF, 0)
+	// See https://github.com/golang/go/issues/73919
-		if err != nil {
+	for {
-			u.l.WithError(err).Error("Failed to rebind udp socket")
+		//_, _, err := unix.Syscall6(unix.SYS_SENDTO, u.sysFd, uintptr(unsafe.Pointer(&b[0])), uintptr(len(b)), 0, sa, addrLen)
+		err := sendto(int(u.sysFd), b, 0, sa, addrLen)
+		if err == nil {
+			// Written, get out before the error handling
+			return nil
 		}
-	})
+
+		if errors.Is(err, syscall.EINTR) {
+			// Write was interrupted, retry
+			continue
+		}
+
+		if errors.Is(err, syscall.EAGAIN) {
+			return &net.OpError{Op: "sendto", Err: unix.EWOULDBLOCK}
+		}
+
+		if errors.Is(err, syscall.EBADF) {
+			return net.ErrClosed
+		}
+
+		return &net.OpError{Op: "sendto", Err: err}
+	}
+}
+
+func (u *StdConn) LocalAddr() (netip.AddrPort, error) {
+	a := u.UDPConn.LocalAddr()
+
+	switch v := a.(type) {
+	case *net.UDPAddr:
+		addr, ok := netip.AddrFromSlice(v.IP)
+		if !ok {
+			return netip.AddrPort{}, fmt.Errorf("LocalAddr returned invalid IP address: %s", v.IP)
+		}
+		return netip.AddrPortFrom(addr, uint16(v.Port)), nil
+
+	default:
+		return netip.AddrPort{}, fmt.Errorf("LocalAddr returned: %#v", a)
+	}
+}
+
+func (u *StdConn) ReloadConfig(c *config.C) {
+	// TODO
+}
+
+func NewUDPStatsEmitter(udpConns []Conn) func() {
+	// No UDP stats for non-linux
+	return func() {}
+}
+
+func (u *StdConn) ListenOut(r EncReader, lhf LightHouseHandlerFunc, cache *firewall.ConntrackCacheTicker, q int) {
+	plaintext := make([]byte, MTU)
+	buffer := make([]byte, MTU)
+	h := &header.H{}
+	fwPacket := &firewall.Packet{}
+	nb := make([]byte, 12, 12)
+
+	for {
+		// Just read one packet at a time
+		n, rua, err := u.ReadFromUDPAddrPort(buffer)
+		if err != nil {
+			if errors.Is(err, net.ErrClosed) {
+				u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
+				return
+			}
+
+			u.l.WithError(err).Error("unexpected udp socket receive error")
+		}
+
+		r(
+			netip.AddrPortFrom(rua.Addr().Unmap(), rua.Port()),
+			plaintext[:0],
+			buffer[:n],
+			h,
+			fwPacket,
+			lhf,
+			nb,
+			q,
+			cache.Get(u.l),
+		)
+	}
+}
+
+func (u *StdConn) Rebind() error {
+	var err error
+	if u.isV4 {
+		err = syscall.SetsockoptInt(int(u.sysFd), syscall.IPPROTO_IP, syscall.IP_BOUND_IF, 0)
+	} else {
+		err = syscall.SetsockoptInt(int(u.sysFd), syscall.IPPROTO_IPV6, syscall.IPV6_BOUND_IF, 0)
+	}
+
+	if err != nil {
+		u.l.WithError(err).Error("Failed to rebind udp socket")
+	}
+
+	return nil
 }

udp/udp_generic.go

@@ -1,6 +1,7 @@
-//go:build (!linux || android) && !e2e_testing
+//go:build (!linux || android) && !e2e_testing && !darwin
 // +build !linux android
 // +build !e2e_testing
+// +build !darwin
 
 // udp_generic implements the nebula UDP interface in pure Go stdlib. This
 // means it can be used on platforms like Darwin and Windows.
@@ -9,9 +10,11 @@ package udp
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"net"
 	"net/netip"
+	"time"
 
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
@@ -79,12 +82,22 @@ func (u *GenericConn) ListenOut(r EncReader, lhf LightHouseHandlerFunc, cache *f
 	fwPacket := &firewall.Packet{}
 	nb := make([]byte, 12, 12)
 
+	var lastRecvErr time.Time
+
 	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")
+			if errors.Is(err, net.ErrClosed) {
-			return
+				u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
+				return
+			}
+			// Dampen unexpected message warns to once per minute
+			if lastRecvErr.IsZero() || time.Since(lastRecvErr) > time.Minute {
+				lastRecvErr = time.Now()
+				u.l.WithError(err).Warn("unexpected udp socket receive error")
+			}
+			continue
 		}
 
 		r(

udp/udp_linux.go

@@ -243,7 +243,7 @@ func (u *StdConn) writeTo6(b []byte, ip netip.AddrPort) error {
 
 func (u *StdConn) writeTo4(b []byte, ip netip.AddrPort) error {
 	if !ip.Addr().Is4() {
-		return fmt.Errorf("Listener is IPv4, but writing to IPv6 remote")
+		return ErrInvalidIPv6RemoteForSocket
 	}
 
 	var rsa unix.RawSockaddrInet4

udp/udp_rio_windows.go

@@ -14,6 +14,7 @@ import (
 	"sync"
 	"sync/atomic"
 	"syscall"
+	"time"
 	"unsafe"
 
 	"github.com/sirupsen/logrus"
@@ -69,7 +70,7 @@ func NewRIOListener(l *logrus.Logger, addr netip.Addr, port int) (*RIOConn, erro
 
 	u := &RIOConn{l: l}
 
-	err := u.bind(&windows.SockaddrInet6{Addr: addr.As16(), Port: port})
+	err := u.bind(l, &windows.SockaddrInet6{Addr: addr.As16(), Port: port})
 	if err != nil {
 		return nil, fmt.Errorf("bind: %w", err)
 	}
@@ -85,34 +86,58 @@ func NewRIOListener(l *logrus.Logger, addr netip.Addr, port int) (*RIOConn, erro
 	return u, nil
 }
 
-func (u *RIOConn) bind(sa windows.Sockaddr) error {
+func (u *RIOConn) bind(l *logrus.Logger, sa windows.Sockaddr) error {
 	var err error
 	u.sock, err = winrio.Socket(windows.AF_INET6, windows.SOCK_DGRAM, windows.IPPROTO_UDP)
 	if err != nil {
-		return err
+		return fmt.Errorf("winrio.Socket error: %w", err)
 	}
 
 	// Enable v4 for this socket
 	syscall.SetsockoptInt(syscall.Handle(u.sock), syscall.IPPROTO_IPV6, syscall.IPV6_V6ONLY, 0)
 
+	// Disable reporting of PORT_UNREACHABLE and NET_UNREACHABLE errors from the UDP socket receive call.
+	// These errors are returned on Windows during UDP receives based on the receipt of ICMP packets. Disable
+	// the UDP receive error returns with these ioctl calls.
+	ret := uint32(0)
+	flag := uint32(0)
+	size := uint32(unsafe.Sizeof(flag))
+	err = syscall.WSAIoctl(syscall.Handle(u.sock), syscall.SIO_UDP_CONNRESET, (*byte)(unsafe.Pointer(&flag)), size, nil, 0, &ret, nil, 0)
+	if err != nil {
+		// This is a best-effort to prevent errors from being returned by the udp recv operation.
+		// Quietly log a failure and continue.
+		l.WithError(err).Debug("failed to set UDP_CONNRESET ioctl")
+	}
+
+	ret = 0
+	flag = 0
+	size = uint32(unsafe.Sizeof(flag))
+	SIO_UDP_NETRESET := uint32(syscall.IOC_IN | syscall.IOC_VENDOR | 15)
+	err = syscall.WSAIoctl(syscall.Handle(u.sock), SIO_UDP_NETRESET, (*byte)(unsafe.Pointer(&flag)), size, nil, 0, &ret, nil, 0)
+	if err != nil {
+		// This is a best-effort to prevent errors from being returned by the udp recv operation.
+		// Quietly log a failure and continue.
+		l.WithError(err).Debug("failed to set UDP_NETRESET ioctl")
+	}
+
 	err = u.rx.Open()
 	if err != nil {
-		return err
+		return fmt.Errorf("error rx.Open(): %w", err)
 	}
 
 	err = u.tx.Open()
 	if err != nil {
-		return err
+		return fmt.Errorf("error tx.Open(): %w", err)
 	}
 
 	u.rq, err = winrio.CreateRequestQueue(u.sock, packetsPerRing, 1, packetsPerRing, 1, u.rx.cq, u.tx.cq, 0)
 	if err != nil {
-		return err
+		return fmt.Errorf("error CreateRequestQueue: %w", err)
 	}
 
 	err = windows.Bind(u.sock, sa)
 	if err != nil {
-		return err
+		return fmt.Errorf("error windows.Bind(): %w", err)
 	}
 
 	return nil
@@ -125,12 +150,23 @@ func (u *RIOConn) ListenOut(r EncReader, lhf LightHouseHandlerFunc, cache *firew
 	fwPacket := &firewall.Packet{}
 	nb := make([]byte, 12, 12)
 
+	var lastRecvErr time.Time
+
 	for {
 		// Just read one packet at a time
 		n, rua, err := u.receive(buffer)
+
 		if err != nil {
-			u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
+			if errors.Is(err, net.ErrClosed) {
-			return
+				u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
+				return
+			}
+			// Dampen unexpected message warns to once per minute
+			if lastRecvErr.IsZero() || time.Since(lastRecvErr) > time.Minute {
+				lastRecvErr = time.Now()
+				u.l.WithError(err).Warn("unexpected udp socket receive error")
+			}
+			continue
 		}
 
 		r(