update synctrace

.github/ISSUE_TEMPLATE/config.yml

@@ -17,5 +17,5 @@ contact_links:
     about: 'The documentation is the best place to start if you are new to Nebula.'
 
   - name: 💁 Support/Chat
-    url: https://join.slack.com/t/nebulaoss/shared_invite/zt-39pk4xopc-CUKlGcb5Z39dQ0cK1v7ehA
+    url: https://join.slack.com/t/nebulaoss/shared_invite/zt-2xqe6e7vn-k_KGi8s13nsr7cvHVvHvuQ
     about: 'For faster support, join us on Slack for assistance!'

.github/workflows/smoke.yml

@@ -52,12 +52,4 @@ jobs:
       working-directory: ./.github/workflows/smoke
       run: NAME="smoke-p256" ./smoke.sh
 
-    - name: setup docker image for fips140
-      working-directory: ./.github/workflows/smoke
-      run: NAME="smoke-fips140" CURVE=P256 GOFIPS140=v1.0.0 LDFLAGS=-checklinkname=0 ./build.sh
-
-    - name: run smoke-fips140
-      working-directory: ./.github/workflows/smoke
-      run: NAME="smoke-fips140" ./smoke.sh
-
     timeout-minutes: 10

.github/workflows/test.yml

@@ -32,9 +32,9 @@ jobs:
       run: make vet
 
     - name: golangci-lint
-      uses: golangci/golangci-lint-action@v8
+      uses: golangci/golangci-lint-action@v7
       with:
-        version: v2.1
+        version: v2.0
 
     - name: Test
       run: make test
@@ -115,9 +115,9 @@ jobs:
       run: make vet
 
     - name: golangci-lint
-      uses: golangci/golangci-lint-action@v8
+      uses: golangci/golangci-lint-action@v7
       with:
-        version: v2.1
+        version: v2.0
 
     - name: Test
       run: make test

Makefile

@@ -61,7 +61,7 @@ ALL = $(ALL_LINUX) \
 	windows-arm64
 
 e2e:
-	$(TEST_ENV) go test -tags=e2e_testing -count=1 $(TEST_FLAGS) ./e2e
+	$(TEST_ENV) go test -tags=synctrace,e2e_testing -count=1 $(TEST_FLAGS) ./e2e
 
 e2ev: TEST_FLAGS += -v
 e2ev: e2e
@@ -121,12 +121,12 @@ bin-pkcs11: CGO_ENABLED = 1
 bin-pkcs11: bin
 
 bin:
-	$(GOENV) go build $(BUILD_ARGS) -ldflags "$(LDFLAGS)" -o ./nebula${NEBULA_CMD_SUFFIX} ${NEBULA_CMD_PATH}
+	go build $(BUILD_ARGS) -ldflags "$(LDFLAGS)" -o ./nebula${NEBULA_CMD_SUFFIX} ${NEBULA_CMD_PATH}
-	$(GOENV) go build $(BUILD_ARGS) -ldflags "$(LDFLAGS)" -o ./nebula-cert${NEBULA_CMD_SUFFIX} ./cmd/nebula-cert
+	go build $(BUILD_ARGS) -ldflags "$(LDFLAGS)" -o ./nebula-cert${NEBULA_CMD_SUFFIX} ./cmd/nebula-cert
 
 install:
-	$(GOENV) go install $(BUILD_ARGS) -ldflags "$(LDFLAGS)" ${NEBULA_CMD_PATH}
+	go install $(BUILD_ARGS) -ldflags "$(LDFLAGS)" ${NEBULA_CMD_PATH}
-	$(GOENV) go install $(BUILD_ARGS) -ldflags "$(LDFLAGS)" ./cmd/nebula-cert
+	go install $(BUILD_ARGS) -ldflags "$(LDFLAGS)" ./cmd/nebula-cert
 
 build/linux-arm-%: GOENV += GOARM=$(word 3, $(subst -, ,$*))
 build/linux-mips-%: GOENV += GOMIPS=$(word 3, $(subst -, ,$*))
@@ -215,14 +215,7 @@ ifeq ($(words $(MAKECMDGOALS)),1)
 	@$(MAKE) service ${.DEFAULT_GOAL} --no-print-directory
 endif
 
-fips140:
+bin-docker: BUILD_ARGS = -tags=synctrace
-	@echo > $(NULL_FILE)
-	$(eval GOENV += GOFIPS140=v1.0.0)
-	$(eval LDFLAGS += -checklinkname=0)
-ifeq ($(words $(MAKECMDGOALS)),1)
-	@$(MAKE) fips140 ${.DEFAULT_GOAL} --no-print-directory
-endif
-
 bin-docker: bin build/linux-amd64/nebula build/linux-amd64/nebula-cert
 
 smoke-docker: bin-docker
@@ -244,5 +237,5 @@ smoke-vagrant/%: bin-docker build/%/nebula
 	cd .github/workflows/smoke/ && ./smoke-vagrant.sh $*
 
 .FORCE:
-.PHONY: bench bench-cpu bench-cpu-long bin build-test-mobile e2e e2ev e2evv e2evvv e2evvvv fips140 proto release service smoke-docker smoke-docker-race test test-cov-html smoke-vagrant/%
+.PHONY: bench bench-cpu bench-cpu-long bin build-test-mobile e2e e2ev e2evv e2evvv e2evvvv proto release service smoke-docker smoke-docker-race test test-cov-html smoke-vagrant/%
 .DEFAULT_GOAL := bin

README.md

@@ -12,7 +12,7 @@ Further documentation can be found [here](https://nebula.defined.net/docs/).
 
 You can read more about Nebula [here](https://medium.com/p/884110a5579).
 
-You can also join the NebulaOSS Slack group [here](https://join.slack.com/t/nebulaoss/shared_invite/zt-39pk4xopc-CUKlGcb5Z39dQ0cK1v7ehA).
+You can also join the NebulaOSS Slack group [here](https://join.slack.com/t/nebulaoss/shared_invite/zt-2xqe6e7vn-k_KGi8s13nsr7cvHVvHvuQ).
 
 ## Supported Platforms
 
@@ -143,24 +143,17 @@ To build nebula for a specific platform (ex, Windows):
 
 See the [Makefile](Makefile) for more details on build targets
 
-## Curve P256, BoringCrypto and FIPS 140-3 mode
+## Curve P256 and BoringCrypto
 
 The default curve used for cryptographic handshakes and signatures is Curve25519. This is the recommended setting for most users. If your deployment has certain compliance requirements, you have the option of creating your CA using `nebula-cert ca -curve P256` to use NIST Curve P256. The CA will then sign certificates using ECDSA P256, and any hosts using these certificates will use P256 for ECDH handshakes.
 
-Nebula can be built using the [BoringCrypto GOEXPERIMENT](https://github.com/golang/go/blob/go1.20/src/crypto/internal/boring/README.md) by running either of the following make targets:
+In addition, Nebula can be built using the [BoringCrypto GOEXPERIMENT](https://github.com/golang/go/blob/go1.20/src/crypto/internal/boring/README.md) by running either of the following make targets:
 
 ```sh
 make bin-boringcrypto
 make release-boringcrypto
 ```
 
-Nebula can also be built using the [FIPS 140-3](https://go.dev/doc/security/fips140) mode of Go by running either of the following make targets:
-
-```sh
-make fips140
-make fips140 release
-```
-
 This is not the recommended default deployment, but may be useful based on your compliance requirements.
 
 ## Credits

config/config.go

@@ -11,12 +11,12 @@ import (
 	"sort"
 	"strconv"
 	"strings"
-	"sync"
 	"syscall"
 	"time"
 
 	"dario.cat/mergo"
 	"github.com/sirupsen/logrus"
+	"github.com/wadey/synctrace"
 	"gopkg.in/yaml.v3"
 )
 
@@ -27,13 +27,14 @@ type C struct {
 	oldSettings map[string]any
 	callbacks   []func(*C)
 	l           *logrus.Logger
-	reloadLock  sync.Mutex
+	reloadLock  synctrace.Mutex
 }
 
 func NewC(l *logrus.Logger) *C {
 	return &C{
-		Settings: make(map[string]any),
+		Settings:   make(map[string]any),
-		l:        l,
+		l:          l,
+		reloadLock: synctrace.NewMutex("config-reload"),
 	}
 }
 

connection_manager.go

@@ -4,17 +4,14 @@ 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"
+	"github.com/wadey/synctrace"
 )
 
 type trafficDecision int
@@ -30,124 +27,130 @@ const (
 )
 
 type connectionManager struct {
+	in     map[uint32]struct{}
+	inLock synctrace.RWMutex
+
+	out     map[uint32]struct{}
+	outLock synctrace.RWMutex
+
 	// relayUsed holds which relay localIndexs are in use
 	relayUsed     map[uint32]struct{}
-	relayUsedLock *sync.RWMutex
+	relayUsedLock synctrace.RWMutex
 
-	hostMap      *HostMap
+	hostMap                 *HostMap
-	trafficTimer *LockingTimerWheel[uint32]
+	trafficTimer            *LockingTimerWheel[uint32]
-	intf         *Interface
+	intf                    *Interface
-	punchy       *Punchy
+	pendingDeletion         map[uint32]struct{}
-
+	punchy                  *Punchy
-	// Configuration settings
 	checkInterval           time.Duration
 	pendingDeletionInterval time.Duration
-	inactivityTimeout       atomic.Int64
+	metricsTxPunchy         metrics.Counter
-	dropInactive            atomic.Bool
-
-	metricsTxPunchy metrics.Counter
 
 	l *logrus.Logger
 }
 
-func newConnectionManagerFromConfig(l *logrus.Logger, c *config.C, hm *HostMap, p *Punchy) *connectionManager {
+func newConnectionManager(ctx context.Context, l *logrus.Logger, intf *Interface, checkInterval, pendingDeletionInterval time.Duration, punchy *Punchy) *connectionManager {
-	cm := &connectionManager{
+	var max time.Duration
-		hostMap:         hm,
+	if checkInterval < pendingDeletionInterval {
-		l:               l,
+		max = pendingDeletionInterval
-		punchy:          p,
+	} else {
-		relayUsed:       make(map[uint32]struct{}),
+		max = checkInterval
-		relayUsedLock:   &sync.RWMutex{},
-		metricsTxPunchy: metrics.GetOrRegisterCounter("messages.tx.punchy", nil),
 	}
 
-	cm.reload(c, true)
+	nc := &connectionManager{
-	c.RegisterReloadCallback(func(c *config.C) {
+		hostMap:                 intf.hostMap,
-		cm.reload(c, false)
+		in:                      make(map[uint32]struct{}),
-	})
+		inLock:                  synctrace.NewRWMutex("connection-manager-in"),
-
+		out:                     make(map[uint32]struct{}),
-	return cm
+		outLock:                 synctrace.NewRWMutex("connection-manager-out"),
-}
+		relayUsed:               make(map[uint32]struct{}),
-
+		relayUsedLock:           synctrace.NewRWMutex("connection-manager-relay-used"),
-func (cm *connectionManager) reload(c *config.C, initial bool) {
+		trafficTimer:            NewLockingTimerWheel[uint32]("traffic-timer", time.Millisecond*500, max),
-	if initial {
+		intf:                    intf,
-		cm.checkInterval = time.Duration(c.GetInt("timers.connection_alive_interval", 5)) * time.Second
+		pendingDeletion:         make(map[uint32]struct{}),
-		cm.pendingDeletionInterval = time.Duration(c.GetInt("timers.pending_deletion_interval", 10)) * time.Second
+		checkInterval:           checkInterval,
-
+		pendingDeletionInterval: pendingDeletionInterval,
-		// We want at least a minimum resolution of 500ms per tick so that we can hit these intervals
+		punchy:                  punchy,
-		// pretty close to their configured duration.
+		metricsTxPunchy:         metrics.GetOrRegisterCounter("messages.tx.punchy", nil),
-		// The inactivity duration is checked each time a hostinfo ticks through so we don't need the wheel to contain it.
+		l:                       l,
-		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") {
+	nc.Start(ctx)
-		old := cm.getInactivityTimeout()
+	return nc
-		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")
-		}
-	}
 }
 
-func (cm *connectionManager) getInactivityTimeout() time.Duration {
+func (n *connectionManager) In(localIndex uint32) {
-	return (time.Duration)(cm.inactivityTimeout.Load())
+	n.inLock.RLock()
-}
-
-func (cm *connectionManager) In(h *HostInfo) {
-	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 := cm.relayUsed[localIndex]; ok {
+	if _, ok := n.in[localIndex]; ok {
-		cm.relayUsedLock.RUnlock()
+		n.inLock.RUnlock()
 		return
 	}
-	cm.relayUsedLock.RUnlock()
+	n.inLock.RUnlock()
-	cm.relayUsedLock.Lock()
+	n.inLock.Lock()
-	cm.relayUsed[localIndex] = struct{}{}
+	n.in[localIndex] = struct{}{}
-	cm.relayUsedLock.Unlock()
+	n.inLock.Unlock()
+}
+
+func (n *connectionManager) Out(localIndex uint32) {
+	n.outLock.RLock()
+	// 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) {
+	n.relayUsedLock.RLock()
+	// If this already exists, return
+	if _, ok := n.relayUsed[localIndex]; ok {
+		n.relayUsedLock.RUnlock()
+		return
+	}
+	n.relayUsedLock.RUnlock()
+	n.relayUsedLock.Lock()
+	n.relayUsed[localIndex] = struct{}{}
+	n.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 (cm *connectionManager) getAndResetTrafficCheck(h *HostInfo, now time.Time) (bool, bool) {
+func (n *connectionManager) getAndResetTrafficCheck(localIndex uint32) (bool, bool) {
-	in := h.in.Swap(false)
+	n.inLock.Lock()
-	out := h.out.Swap(false)
+	n.outLock.Lock()
-	if in || out {
+	_, in := n.in[localIndex]
-		h.lastUsed = now
+	_, out := n.out[localIndex]
-	}
+	delete(n.in, localIndex)
+	delete(n.out, localIndex)
+	n.inLock.Unlock()
+	n.outLock.Unlock()
 	return in, out
 }
 
-// AddTrafficWatch must be called for every new HostInfo.
+func (n *connectionManager) AddTrafficWatch(localIndex uint32) {
-// We will continue to monitor the HostInfo until the tunnel is dropped.
+	// Use a write lock directly because it should be incredibly rare that we are ever already tracking this index
-func (cm *connectionManager) AddTrafficWatch(h *HostInfo) {
+	n.outLock.Lock()
-	if h.out.Swap(true) == false {
+	if _, ok := n.out[localIndex]; ok {
-		cm.trafficTimer.Add(h.localIndexId, cm.checkInterval)
+		n.outLock.Unlock()
+		return
 	}
+	n.out[localIndex] = struct{}{}
+	n.trafficTimer.Add(localIndex, n.checkInterval)
+	n.outLock.Unlock()
 }
 
-func (cm *connectionManager) Start(ctx context.Context) {
+func (n *connectionManager) Start(ctx context.Context) {
-	clockSource := time.NewTicker(cm.trafficTimer.t.tickDuration)
+	go n.Run(ctx)
+}
+
+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("")
@@ -160,61 +163,61 @@ func (cm *connectionManager) Start(ctx context.Context) {
 			return
 
 		case now := <-clockSource.C:
-			cm.trafficTimer.Advance(now)
+			n.trafficTimer.Advance(now)
 			for {
-				localIndex, has := cm.trafficTimer.Purge()
+				localIndex, has := n.trafficTimer.Purge()
 				if !has {
 					break
 				}
 
-				cm.doTrafficCheck(localIndex, p, nb, out, now)
+				n.doTrafficCheck(localIndex, p, nb, out, now)
 			}
 		}
 	}
 }
 
-func (cm *connectionManager) doTrafficCheck(localIndex uint32, p, nb, out []byte, now time.Time) {
+func (n *connectionManager) doTrafficCheck(localIndex uint32, p, nb, out []byte, now time.Time) {
-	decision, hostinfo, primary := cm.makeTrafficDecision(localIndex, now)
+	decision, hostinfo, primary := n.makeTrafficDecision(localIndex, now)
 
 	switch decision {
 	case deleteTunnel:
-		if cm.hostMap.DeleteHostInfo(hostinfo) {
+		if n.hostMap.DeleteHostInfo(hostinfo) {
 			// Only clearing the lighthouse cache if this is the last hostinfo for this vpn ip in the hostmap
-			cm.intf.lightHouse.DeleteVpnAddrs(hostinfo.vpnAddrs)
+			n.intf.lightHouse.DeleteVpnAddrs(hostinfo.vpnAddrs)
 		}
 
 	case closeTunnel:
-		cm.intf.sendCloseTunnel(hostinfo)
+		n.intf.sendCloseTunnel(hostinfo)
-		cm.intf.closeTunnel(hostinfo)
+		n.intf.closeTunnel(hostinfo)
 
 	case swapPrimary:
-		cm.swapPrimary(hostinfo, primary)
+		n.swapPrimary(hostinfo, primary)
 
 	case migrateRelays:
-		cm.migrateRelayUsed(hostinfo, primary)
+		n.migrateRelayUsed(hostinfo, primary)
 
 	case tryRehandshake:
-		cm.tryRehandshake(hostinfo)
+		n.tryRehandshake(hostinfo)
 
 	case sendTestPacket:
-		cm.intf.SendMessageToHostInfo(header.Test, header.TestRequest, hostinfo, p, nb, out)
+		n.intf.SendMessageToHostInfo(header.Test, header.TestRequest, hostinfo, p, nb, out)
 	}
 
-	cm.resetRelayTrafficCheck(hostinfo)
+	n.resetRelayTrafficCheck(hostinfo)
 }
 
-func (cm *connectionManager) resetRelayTrafficCheck(hostinfo *HostInfo) {
+func (n *connectionManager) resetRelayTrafficCheck(hostinfo *HostInfo) {
 	if hostinfo != nil {
-		cm.relayUsedLock.Lock()
+		n.relayUsedLock.Lock()
-		defer cm.relayUsedLock.Unlock()
+		defer n.relayUsedLock.Unlock()
 		// No need to migrate any relays, delete usage info now.
 		for _, idx := range hostinfo.relayState.CopyRelayForIdxs() {
-			delete(cm.relayUsed, idx)
+			delete(n.relayUsed, idx)
 		}
 	}
 }
 
-func (cm *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo) {
+func (n *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo) {
 	relayFor := oldhostinfo.relayState.CopyAllRelayFor()
 
 	for _, r := range relayFor {
@@ -224,51 +227,46 @@ func (cm *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo
 		var relayFrom netip.Addr
 		var relayTo netip.Addr
 		switch {
-		case ok:
+		case ok && existing.State == Established:
-			switch existing.State {
+			// This relay already exists in newhostinfo, then do nothing.
-			case Established, PeerRequested, Disestablished:
+			continue
-				// This relay already exists in newhostinfo, then do nothing.
+		case ok && existing.State == Requested:
-				continue
+			// The relay exists in a Requested state; re-send the request
-			case Requested:
+			index = existing.LocalIndex
-				// The relay exists in a Requested state; re-send the request
+			switch r.Type {
-				index = existing.LocalIndex
+			case TerminalType:
-				switch r.Type {
+				relayFrom = n.intf.myVpnAddrs[0]
-				case TerminalType:
+				relayTo = existing.PeerAddr
-					relayFrom = cm.intf.myVpnAddrs[0]
+			case ForwardingType:
-					relayTo = existing.PeerAddr
+				relayFrom = existing.PeerAddr
-				case ForwardingType:
+				relayTo = newhostinfo.vpnAddrs[0]
-					relayFrom = existing.PeerAddr
+			default:
-					relayTo = newhostinfo.vpnAddrs[0]
+				// should never happen
-				default:
-					// should never happen
-					panic(fmt.Sprintf("Migrating unknown relay type: %v", r.Type))
-				}
 			}
 		case !ok:
-			cm.relayUsedLock.RLock()
+			n.relayUsedLock.RLock()
-			if _, relayUsed := cm.relayUsed[r.LocalIndex]; !relayUsed {
+			if _, relayUsed := n.relayUsed[r.LocalIndex]; !relayUsed {
 				// The relay hasn't been used; don't migrate it.
-				cm.relayUsedLock.RUnlock()
+				n.relayUsedLock.RUnlock()
 				continue
 			}
-			cm.relayUsedLock.RUnlock()
+			n.relayUsedLock.RUnlock()
 			// The relay doesn't exist at all; create some relay state and send the request.
 			var err error
-			index, err = AddRelay(cm.l, newhostinfo, cm.hostMap, r.PeerAddr, nil, r.Type, Requested)
+			index, err = AddRelay(n.l, newhostinfo, n.hostMap, r.PeerAddr, nil, r.Type, Requested)
 			if err != nil {
-				cm.l.WithError(err).Error("failed to migrate relay to new hostinfo")
+				n.l.WithError(err).Error("failed to migrate relay to new hostinfo")
 				continue
 			}
 			switch r.Type {
 			case TerminalType:
-				relayFrom = cm.intf.myVpnAddrs[0]
+				relayFrom = n.intf.myVpnAddrs[0]
 				relayTo = r.PeerAddr
 			case ForwardingType:
 				relayFrom = r.PeerAddr
 				relayTo = newhostinfo.vpnAddrs[0]
 			default:
 				// should never happen
-				panic(fmt.Sprintf("Migrating unknown relay type: %v", r.Type))
 			}
 		}
 
@@ -281,12 +279,12 @@ func (cm *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo
 		switch newhostinfo.GetCert().Certificate.Version() {
 		case cert.Version1:
 			if !relayFrom.Is4() {
-				cm.l.Error("can not migrate v1 relay with a v6 network because the relay is not running a current nebula version")
+				n.l.Error("can not migrate v1 relay with a v6 network because the relay is not running a current nebula version")
 				continue
 			}
 
 			if !relayTo.Is4() {
-				cm.l.Error("can not migrate v1 relay with a v6 remote network because the relay is not running a current nebula version")
+				n.l.Error("can not migrate v1 relay with a v6 remote network because the relay is not running a current nebula version")
 				continue
 			}
 
@@ -298,16 +296,16 @@ func (cm *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo
 			req.RelayFromAddr = netAddrToProtoAddr(relayFrom)
 			req.RelayToAddr = netAddrToProtoAddr(relayTo)
 		default:
-			newhostinfo.logger(cm.l).Error("Unknown certificate version found while attempting to migrate relay")
+			newhostinfo.logger(n.l).Error("Unknown certificate version found while attempting to migrate relay")
 			continue
 		}
 
 		msg, err := req.Marshal()
 		if err != nil {
-			cm.l.WithError(err).Error("failed to marshal Control message to migrate relay")
+			n.l.WithError(err).Error("failed to marshal Control message to migrate relay")
 		} else {
-			cm.intf.SendMessageToHostInfo(header.Control, 0, newhostinfo, msg, make([]byte, 12), make([]byte, mtu))
+			n.intf.SendMessageToHostInfo(header.Control, 0, newhostinfo, msg, make([]byte, 12), make([]byte, mtu))
-			cm.l.WithFields(logrus.Fields{
+			n.l.WithFields(logrus.Fields{
 				"relayFrom":           req.RelayFromAddr,
 				"relayTo":             req.RelayToAddr,
 				"initiatorRelayIndex": req.InitiatorRelayIndex,
@@ -318,45 +316,46 @@ func (cm *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo
 	}
 }
 
-func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Time) (trafficDecision, *HostInfo, *HostInfo) {
+func (n *connectionManager) makeTrafficDecision(localIndex uint32, now time.Time) (trafficDecision, *HostInfo, *HostInfo) {
-	// Read lock the main hostmap to order decisions based on tunnels being the primary tunnel
+	n.hostMap.RLock()
-	cm.hostMap.RLock()
+	defer n.hostMap.RUnlock()
-	defer cm.hostMap.RUnlock()
 
-	hostinfo := cm.hostMap.Indexes[localIndex]
+	hostinfo := n.hostMap.Indexes[localIndex]
 	if hostinfo == nil {
-		cm.l.WithField("localIndex", localIndex).Debugln("Not found in hostmap")
+		n.l.WithField("localIndex", localIndex).Debugf("Not found in hostmap")
+		delete(n.pendingDeletion, localIndex)
 		return doNothing, nil, nil
 	}
 
-	if cm.isInvalidCertificate(now, hostinfo) {
+	if n.isInvalidCertificate(now, hostinfo) {
+		delete(n.pendingDeletion, hostinfo.localIndexId)
 		return closeTunnel, hostinfo, nil
 	}
 
-	primary := cm.hostMap.Hosts[hostinfo.vpnAddrs[0]]
+	primary := n.hostMap.Hosts[hostinfo.vpnAddrs[0]]
 	mainHostInfo := true
 	if primary != nil && primary != hostinfo {
 		mainHostInfo = false
 	}
 
 	// Check for traffic on this hostinfo
-	inTraffic, outTraffic := cm.getAndResetTrafficCheck(hostinfo, now)
+	inTraffic, outTraffic := n.getAndResetTrafficCheck(localIndex)
 
 	// A hostinfo is determined alive if there is incoming traffic
 	if inTraffic {
 		decision := doNothing
-		if cm.l.Level >= logrus.DebugLevel {
+		if n.l.Level >= logrus.DebugLevel {
-			hostinfo.logger(cm.l).
+			hostinfo.logger(n.l).
 				WithField("tunnelCheck", m{"state": "alive", "method": "passive"}).
 				Debug("Tunnel status")
 		}
-		hostinfo.pendingDeletion.Store(false)
+		delete(n.pendingDeletion, hostinfo.localIndexId)
 
 		if mainHostInfo {
 			decision = tryRehandshake
 
 		} else {
-			if cm.shouldSwapPrimary(hostinfo, primary) {
+			if n.shouldSwapPrimary(hostinfo, primary) {
 				decision = swapPrimary
 			} else {
 				// migrate the relays to the primary, if in use.
@@ -364,55 +363,46 @@ func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Tim
 			}
 		}
 
-		cm.trafficTimer.Add(hostinfo.localIndexId, cm.checkInterval)
+		n.trafficTimer.Add(hostinfo.localIndexId, n.checkInterval)
 
 		if !outTraffic {
 			// Send a punch packet to keep the NAT state alive
-			cm.sendPunch(hostinfo)
+			n.sendPunch(hostinfo)
 		}
 
 		return decision, hostinfo, primary
 	}
 
-	if hostinfo.pendingDeletion.Load() {
+	if _, ok := n.pendingDeletion[hostinfo.localIndexId]; ok {
 		// We have already sent a test packet and nothing was returned, this hostinfo is dead
-		hostinfo.logger(cm.l).
+		hostinfo.logger(n.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.
-			cm.sendPunch(hostinfo)
+			n.sendPunch(hostinfo)
-			cm.trafficTimer.Add(hostinfo.localIndexId, cm.checkInterval)
+			n.trafficTimer.Add(hostinfo.localIndexId, n.checkInterval)
 			return doNothing, nil, nil
+
 		}
 
-		if cm.punchy.GetTargetEverything() {
+		if n.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
-			cm.sendPunch(hostinfo)
+			n.sendPunch(hostinfo)
 		}
 
-		if cm.l.Level >= logrus.DebugLevel {
+		if n.l.Level >= logrus.DebugLevel {
-			hostinfo.logger(cm.l).
+			hostinfo.logger(n.l).
 				WithField("tunnelCheck", m{"state": "testing", "method": "active"}).
 				Debug("Tunnel status")
 		}
@@ -421,33 +411,17 @@ func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Tim
 		decision = sendTestPacket
 
 	} else {
-		if cm.l.Level >= logrus.DebugLevel {
+		if n.l.Level >= logrus.DebugLevel {
-			hostinfo.logger(cm.l).Debugf("Hostinfo sadness")
+			hostinfo.logger(n.l).Debugf("Hostinfo sadness")
 		}
 	}
 
-	hostinfo.pendingDeletion.Store(true)
+	n.pendingDeletion[hostinfo.localIndexId] = struct{}{}
-	cm.trafficTimer.Add(hostinfo.localIndexId, cm.pendingDeletionInterval)
+	n.trafficTimer.Add(hostinfo.localIndexId, n.pendingDeletionInterval)
 	return decision, hostinfo, nil
 }
 
-func (cm *connectionManager) isInactive(hostinfo *HostInfo, now time.Time) (time.Duration, bool) {
+func (n *connectionManager) shouldSwapPrimary(current, primary *HostInfo) 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.
@@ -455,80 +429,73 @@ func (cm *connectionManager) shouldSwapPrimary(current, primary *HostInfo) bool
 	// Only one side should swap because if both swap then we may never resolve to a single tunnel.
 	// vpn addr is static across all tunnels for this host pair so lets
 	// use that to determine if we should consider swapping.
-	if current.vpnAddrs[0].Compare(cm.intf.myVpnAddrs[0]) < 0 {
+	if current.vpnAddrs[0].Compare(n.intf.myVpnAddrs[0]) < 0 {
 		// Their primary vpn addr is less than mine. Do not swap.
 		return false
 	}
 
-	crt := cm.intf.pki.getCertState().getCertificate(current.ConnectionState.myCert.Version())
+	crt := n.intf.pki.getCertState().getCertificate(current.ConnectionState.myCert.Version())
 	// If this tunnel is using the latest certificate then we should swap it to primary for a bit and see if things
 	// settle down.
 	return bytes.Equal(current.ConnectionState.myCert.Signature(), crt.Signature())
 }
 
-func (cm *connectionManager) swapPrimary(current, primary *HostInfo) {
+func (n *connectionManager) swapPrimary(current, primary *HostInfo) {
-	cm.hostMap.Lock()
+	n.hostMap.Lock()
 	// Make sure the primary is still the same after the write lock. This avoids a race with a rehandshake.
-	if cm.hostMap.Hosts[current.vpnAddrs[0]] == primary {
+	if n.hostMap.Hosts[current.vpnAddrs[0]] == primary {
-		cm.hostMap.unlockedMakePrimary(current)
+		n.hostMap.unlockedMakePrimary(current)
 	}
-	cm.hostMap.Unlock()
+	n.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 (cm *connectionManager) isInvalidCertificate(now time.Time, hostinfo *HostInfo) bool {
+func (n *connectionManager) isInvalidCertificate(now time.Time, hostinfo *HostInfo) bool {
 	remoteCert := hostinfo.GetCert()
 	if remoteCert == nil {
 		return false
 	}
 
-	caPool := cm.intf.pki.GetCAPool()
+	caPool := n.intf.pki.GetCAPool()
 	err := caPool.VerifyCachedCertificate(now, remoteCert)
 	if err == nil {
 		return false
 	}
 
-	if !cm.intf.disconnectInvalid.Load() && err != cert.ErrBlockListed {
+	if !n.intf.disconnectInvalid.Load() && err != cert.ErrBlockListed {
 		// Block listed certificates should always be disconnected
 		return false
 	}
 
-	hostinfo.logger(cm.l).WithError(err).
+	hostinfo.logger(n.l).WithError(err).
 		WithField("fingerprint", remoteCert.Fingerprint).
 		Info("Remote certificate is no longer valid, tearing down the tunnel")
 
 	return true
 }
 
-func (cm *connectionManager) sendPunch(hostinfo *HostInfo) {
+func (n *connectionManager) sendPunch(hostinfo *HostInfo) {
-	if !cm.punchy.GetPunch() {
+	if !n.punchy.GetPunch() {
 		// Punching is disabled
 		return
 	}
 
-	if cm.intf.lightHouse.IsAnyLighthouseAddr(hostinfo.vpnAddrs) {
+	if n.punchy.GetTargetEverything() {
-		// Do not punch to lighthouses, we assume our lighthouse update interval is good enough.
+		hostinfo.remotes.ForEach(n.hostMap.GetPreferredRanges(), func(addr netip.AddrPort, preferred bool) {
-		// In the event the update interval is not sufficient to maintain NAT state then a publicly available lighthouse
+			n.metricsTxPunchy.Inc(1)
-		// would lose the ability to notify us and punchy.respond would become unreliable.
+			n.intf.outside.WriteTo([]byte{1}, addr)
-		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() {
-		cm.metricsTxPunchy.Inc(1)
+		n.metricsTxPunchy.Inc(1)
-		cm.intf.outside.WriteTo([]byte{1}, hostinfo.remote)
+		n.intf.outside.WriteTo([]byte{1}, hostinfo.remote)
 	}
 }
 
-func (cm *connectionManager) tryRehandshake(hostinfo *HostInfo) {
+func (n *connectionManager) tryRehandshake(hostinfo *HostInfo) {
-	cs := cm.intf.pki.getCertState()
+	cs := n.intf.pki.getCertState()
 	curCrt := hostinfo.ConnectionState.myCert
 	myCrt := cs.getCertificate(curCrt.Version())
 	if curCrt.Version() >= cs.initiatingVersion && bytes.Equal(curCrt.Signature(), myCrt.Signature()) == true {
@@ -536,9 +503,9 @@ func (cm *connectionManager) tryRehandshake(hostinfo *HostInfo) {
 		return
 	}
 
-	cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
+	n.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
 		WithField("reason", "local certificate is not current").
 		Info("Re-handshaking with remote")
 
-	cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
+	n.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
 }

connection_manager_test.go

@@ -1,6 +1,7 @@
 package nebula
 
 import (
+	"context"
 	"crypto/ed25519"
 	"crypto/rand"
 	"net/netip"
@@ -63,10 +64,10 @@ func Test_NewConnectionManagerTest(t *testing.T) {
 	ifce.pki.cs.Store(cs)
 
 	// Create manager
-	conf := config.NewC(l)
+	ctx, cancel := context.WithCancel(context.Background())
-	punchy := NewPunchyFromConfig(l, conf)
+	defer cancel()
-	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
+	punchy := NewPunchyFromConfig(l, config.NewC(l))
-	nc.intf = ifce
+	nc := newConnectionManager(ctx, l, ifce, 5, 10, punchy)
 	p := []byte("")
 	nb := make([]byte, 12, 12)
 	out := make([]byte, mtu)
@@ -84,33 +85,32 @@ func Test_NewConnectionManagerTest(t *testing.T) {
 	nc.hostMap.unlockedAddHostInfo(hostinfo, ifce)
 
 	// We saw traffic out to vpnIp
-	nc.Out(hostinfo)
+	nc.Out(hostinfo.localIndexId)
-	nc.In(hostinfo)
+	nc.In(hostinfo.localIndexId)
-	assert.False(t, hostinfo.pendingDeletion.Load())
+	assert.NotContains(t, nc.pendingDeletion, hostinfo.localIndexId)
 	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs[0])
 	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
-	assert.True(t, hostinfo.out.Load())
+	assert.Contains(t, nc.out, hostinfo.localIndexId)
-	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.False(t, hostinfo.pendingDeletion.Load())
+	assert.NotContains(t, nc.pendingDeletion, hostinfo.localIndexId)
-	assert.False(t, hostinfo.out.Load())
+	assert.NotContains(t, nc.out, hostinfo.localIndexId)
-	assert.False(t, hostinfo.in.Load())
+	assert.NotContains(t, nc.in, hostinfo.localIndexId)
 
 	// Do another traffic check tick, this host should be pending deletion now
-	nc.Out(hostinfo)
+	nc.Out(hostinfo.localIndexId)
-	assert.True(t, hostinfo.out.Load())
 	nc.doTrafficCheck(hostinfo.localIndexId, p, nb, out, time.Now())
-	assert.True(t, hostinfo.pendingDeletion.Load())
+	assert.Contains(t, nc.pendingDeletion, hostinfo.localIndexId)
-	assert.False(t, hostinfo.out.Load())
+	assert.NotContains(t, nc.out, hostinfo.localIndexId)
-	assert.False(t, hostinfo.in.Load())
+	assert.NotContains(t, nc.in, hostinfo.localIndexId)
 	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
 	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs[0])
 
 	// 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.hostMap.Hosts, hostinfo.vpnAddrs)
+	assert.NotContains(t, nc.pendingDeletion, hostinfo.localIndexId)
+	assert.NotContains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs[0])
 	assert.NotContains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
 }
 
@@ -146,10 +146,10 @@ func Test_NewConnectionManagerTest2(t *testing.T) {
 	ifce.pki.cs.Store(cs)
 
 	// Create manager
-	conf := config.NewC(l)
+	ctx, cancel := context.WithCancel(context.Background())
-	punchy := NewPunchyFromConfig(l, conf)
+	defer cancel()
-	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
+	punchy := NewPunchyFromConfig(l, config.NewC(l))
-	nc.intf = ifce
+	nc := newConnectionManager(ctx, l, ifce, 5, 10, punchy)
 	p := []byte("")
 	nb := make([]byte, 12, 12)
 	out := make([]byte, mtu)
@@ -167,129 +167,33 @@ func Test_NewConnectionManagerTest2(t *testing.T) {
 	nc.hostMap.unlockedAddHostInfo(hostinfo, ifce)
 
 	// We saw traffic out to vpnIp
-	nc.Out(hostinfo)
+	nc.Out(hostinfo.localIndexId)
-	nc.In(hostinfo)
+	nc.In(hostinfo.localIndexId)
-	assert.True(t, hostinfo.in.Load())
+	assert.NotContains(t, nc.pendingDeletion, hostinfo.vpnAddrs[0])
-	assert.True(t, hostinfo.out.Load())
-	assert.False(t, hostinfo.pendingDeletion.Load())
 	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs[0])
 	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.False(t, hostinfo.pendingDeletion.Load())
+	assert.NotContains(t, nc.pendingDeletion, hostinfo.localIndexId)
-	assert.False(t, hostinfo.out.Load())
+	assert.NotContains(t, nc.out, hostinfo.localIndexId)
-	assert.False(t, hostinfo.in.Load())
+	assert.NotContains(t, nc.in, hostinfo.localIndexId)
 
 	// Do another traffic check tick, this host should be pending deletion now
-	nc.Out(hostinfo)
+	nc.Out(hostinfo.localIndexId)
 	nc.doTrafficCheck(hostinfo.localIndexId, p, nb, out, time.Now())
-	assert.True(t, hostinfo.pendingDeletion.Load())
+	assert.Contains(t, nc.pendingDeletion, hostinfo.localIndexId)
-	assert.False(t, hostinfo.out.Load())
+	assert.NotContains(t, nc.out, hostinfo.localIndexId)
-	assert.False(t, hostinfo.in.Load())
+	assert.NotContains(t, nc.in, hostinfo.localIndexId)
 	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
 	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs[0])
 
 	// We saw traffic, should no longer be pending deletion
-	nc.In(hostinfo)
+	nc.In(hostinfo.localIndexId)
 	nc.doTrafficCheck(hostinfo.localIndexId, p, nb, out, time.Now())
-	assert.False(t, hostinfo.pendingDeletion.Load())
+	assert.NotContains(t, nc.pendingDeletion, hostinfo.localIndexId)
-	assert.False(t, hostinfo.out.Load())
+	assert.NotContains(t, nc.out, hostinfo.localIndexId)
-	assert.False(t, hostinfo.in.Load())
+	assert.NotContains(t, nc.in, hostinfo.localIndexId)
-	assert.Contains(t, nc.hostMap.Indexes, hostinfo.localIndexId)
-	assert.Contains(t, nc.hostMap.Hosts, hostinfo.vpnAddrs[0])
-}
-
-func Test_NewConnectionManager_DisconnectInactive(t *testing.T) {
-	l := test.NewLogger()
-	localrange := netip.MustParsePrefix("10.1.1.1/24")
-	vpnAddrs := []netip.Addr{netip.MustParseAddr("172.1.1.2")}
-	preferredRanges := []netip.Prefix{localrange}
-
-	// Very incomplete mock objects
-	hostMap := newHostMap(l)
-	hostMap.preferredRanges.Store(&preferredRanges)
-
-	cs := &CertState{
-		initiatingVersion: cert.Version1,
-		privateKey:        []byte{},
-		v1Cert:            &dummyCert{version: cert.Version1},
-		v1HandshakeBytes:  []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[string]any{
-		"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{
-		vpnAddrs:      vpnAddrs,
-		localIndexId:  1099,
-		remoteIndexId: 9901,
-	}
-	hostinfo.ConnectionState = &ConnectionState{
-		myCert: &dummyCert{version: cert.Version1},
-		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.vpnAddrs[0])
-
-	// 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.vpnAddrs[0])
 }
@@ -360,10 +264,10 @@ func Test_NewConnectionManagerTest_DisconnectInvalid(t *testing.T) {
 	ifce.disconnectInvalid.Store(true)
 
 	// Create manager
-	conf := config.NewC(l)
+	ctx, cancel := context.WithCancel(context.Background())
-	punchy := NewPunchyFromConfig(l, conf)
+	defer cancel()
-	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
+	punchy := NewPunchyFromConfig(l, config.NewC(l))
-	nc.intf = ifce
+	nc := newConnectionManager(ctx, l, ifce, 5, 10, punchy)
 	ifce.connectionManager = nc
 
 	hostinfo := &HostInfo{

connection_state.go

@@ -4,13 +4,13 @@ import (
 	"crypto/rand"
 	"encoding/json"
 	"fmt"
-	"sync"
 	"sync/atomic"
 
 	"github.com/flynn/noise"
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/noiseutil"
+	"github.com/wadey/synctrace"
 )
 
 const ReplayWindow = 1024
@@ -24,7 +24,7 @@ type ConnectionState struct {
 	initiator      bool
 	messageCounter atomic.Uint64
 	window         *Bits
-	writeLock      sync.Mutex
+	writeLock      synctrace.Mutex
 }
 
 func NewConnectionState(l *logrus.Logger, cs *CertState, crt cert.Certificate, initiator bool, pattern noise.HandshakePattern) (*ConnectionState, error) {
@@ -76,6 +76,7 @@ func NewConnectionState(l *logrus.Logger, cs *CertState, crt cert.Certificate, i
 		initiator: initiator,
 		window:    b,
 		myCert:    crt,
+		writeLock: synctrace.NewMutex("connection-state"),
 	}
 	// always start the counter from 2, as packet 1 and packet 2 are handshake packets.
 	ci.messageCounter.Add(2)

control.go

@@ -26,15 +26,14 @@ 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 {
@@ -64,9 +63,6 @@ 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

@@ -53,7 +53,7 @@ func TestControl_GetHostInfoByVpnIp(t *testing.T) {
 		localIndexId:  201,
 		vpnAddrs:      []netip.Addr{vpnIp},
 		relayState: RelayState{
-			relays:         nil,
+			relays:         map[netip.Addr]struct{}{},
 			relayForByAddr: map[netip.Addr]*Relay{},
 			relayForByIdx:  map[uint32]*Relay{},
 		},
@@ -72,7 +72,7 @@ func TestControl_GetHostInfoByVpnIp(t *testing.T) {
 		localIndexId:  201,
 		vpnAddrs:      []netip.Addr{vpnIp2},
 		relayState: RelayState{
-			relays:         nil,
+			relays:         map[netip.Addr]struct{}{},
 			relayForByAddr: map[netip.Addr]*Relay{},
 			relayForByIdx:  map[uint32]*Relay{},
 		},

dns_server.go

@@ -6,12 +6,12 @@ import (
 	"net/netip"
 	"strconv"
 	"strings"
-	"sync"
 
 	"github.com/gaissmai/bart"
 	"github.com/miekg/dns"
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
+	"github.com/wadey/synctrace"
 )
 
 // This whole thing should be rewritten to use context
@@ -21,7 +21,7 @@ var dnsServer *dns.Server
 var dnsAddr string
 
 type dnsRecords struct {
-	sync.RWMutex
+	synctrace.RWMutex
 	l               *logrus.Logger
 	dnsMap4         map[string]netip.Addr
 	dnsMap6         map[string]netip.Addr
@@ -31,6 +31,7 @@ type dnsRecords struct {
 
 func newDnsRecords(l *logrus.Logger, cs *CertState, hostMap *HostMap) *dnsRecords {
 	return &dnsRecords{
+		RWMutex:         synctrace.NewRWMutex("dns-records"),
 		l:               l,
 		dnsMap4:         make(map[string]netip.Addr),
 		dnsMap6:         make(map[string]netip.Addr),

e2e/handshakes_test.go

@@ -506,7 +506,7 @@ 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, current=%v indexes", start, curIndexes)
+		r.Logf("Wait for the dead index to go away:start=%v indexes, currnet=%v indexes", start, curIndexes)
 		myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me should fail"))
 
 		r.RouteForAllExitFunc(func(p *udp.Packet, c *nebula.Control) router.ExitType {
@@ -1052,9 +1052,6 @@ func TestRehandshakingLoser(t *testing.T) {
 	t.Log("Stand up a tunnel between me and them")
 	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
 
-	myControl.GetHostInfoByVpnAddr(theirVpnIpNet[0].Addr(), false)
-	theirControl.GetHostInfoByVpnAddr(myVpnIpNet[0].Addr(), false)
-
 	r.RenderHostmaps("Starting hostmaps", myControl, theirControl)
 
 	r.Log("Renew their certificate and spin until mine sees it")

e2e/router/router.go

@@ -700,7 +700,6 @@ func (r *R) FlushAll() {
 			r.Unlock()
 			panic("Can't FlushAll for host: " + p.To.String())
 		}
-		receiver.InjectUDPPacket(p)
 		r.Unlock()
 	}
 }

e2e/tunnels_test.go

@@ -1,57 +0,0 @@
-//go:build e2e_testing
-// +build e2e_testing
-
-package e2e
-
-import (
-	"testing"
-	"time"
-
-	"github.com/slackhq/nebula/cert"
-	"github.com/slackhq/nebula/cert_test"
-	"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, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me", "10.128.0.1/24", m{"tunnels": m{"drop_inactive": true, "inactivity_timeout": "5s"}})
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them", "10.128.0.2/24", m{"tunnels": m{"drop_inactive": true, "inactivity_timeout": "10m"}})
-
-	// Share our underlay information
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-	theirControl.InjectLightHouseAddr(myVpnIpNet[0].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[0].Addr(), theirVpnIpNet[0].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

@@ -338,18 +338,6 @@ 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

@@ -10,7 +10,6 @@ import (
 	"reflect"
 	"strconv"
 	"strings"
-	"sync"
 	"time"
 
 	"github.com/gaissmai/bart"
@@ -19,6 +18,7 @@ import (
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/firewall"
+	"github.com/wadey/synctrace"
 )
 
 type FirewallInterface interface {
@@ -76,7 +76,7 @@ type firewallMetrics struct {
 }
 
 type FirewallConntrack struct {
-	sync.Mutex
+	synctrace.Mutex
 
 	Conns      map[firewall.Packet]*conn
 	TimerWheel *TimerWheel[firewall.Packet]
@@ -164,6 +164,7 @@ func NewFirewall(l *logrus.Logger, tcpTimeout, UDPTimeout, defaultTimeout time.D
 
 	return &Firewall{
 		Conntrack: &FirewallConntrack{
+			Mutex:      synctrace.NewMutex("firewall-conntrack"),
 			Conns:      make(map[firewall.Packet]*conn),
 			TimerWheel: NewTimerWheel[firewall.Packet](tmin, tmax),
 		},

go.mod

@@ -1,11 +1,11 @@
 module github.com/slackhq/nebula
 
-go 1.24.0
+go 1.23.0
 
 toolchain go1.24.1
 
 require (
-	dario.cat/mergo v1.0.2
+	dario.cat/mergo v1.0.1
 	github.com/anmitsu/go-shlex v0.0.0-20200514113438-38f4b401e2be
 	github.com/armon/go-radix v1.0.0
 	github.com/cyberdelia/go-metrics-graphite v0.0.0-20161219230853-39f87cc3b432
@@ -23,7 +23,8 @@ require (
 	github.com/skip2/go-qrcode v0.0.0-20200617195104-da1b6568686e
 	github.com/stefanberger/go-pkcs11uri v0.0.0-20230803200340-78284954bff6
 	github.com/stretchr/testify v1.10.0
-	github.com/vishvananda/netlink v1.3.1
+	github.com/vishvananda/netlink v1.3.0
+	github.com/wadey/synctrace v0.0.0-20250612192159-94547ef50dfe
 	golang.org/x/crypto v0.37.0
 	golang.org/x/exp v0.0.0-20230725093048-515e97ebf090
 	golang.org/x/net v0.39.0
@@ -42,13 +43,17 @@ require (
 	github.com/beorn7/perks v1.0.1 // indirect
 	github.com/cespare/xxhash/v2 v2.3.0 // indirect
 	github.com/davecgh/go-spew v1.1.1 // indirect
+	github.com/emirpasic/gods v1.18.1 // indirect
 	github.com/google/btree v1.1.2 // indirect
+	github.com/google/uuid v1.3.0 // indirect
+	github.com/heimdalr/dag v1.4.0 // 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 v0.6.1 // indirect
 	github.com/prometheus/common v0.62.0 // indirect
 	github.com/prometheus/procfs v0.15.1 // indirect
-	github.com/vishvananda/netns v0.0.5 // indirect
+	github.com/timandy/routine v1.1.5 // indirect
+	github.com/vishvananda/netns v0.0.4 // indirect
 	golang.org/x/mod v0.23.0 // indirect
 	golang.org/x/time v0.5.0 // indirect
 	golang.org/x/tools v0.30.0 // indirect

go.sum

@@ -1,6 +1,6 @@
 cloud.google.com/go v0.34.0/go.mod h1:aQUYkXzVsufM+DwF1aE+0xfcU+56JwCaLick0ClmMTw=
-dario.cat/mergo v1.0.2 h1:85+piFYR1tMbRrLcDwR18y4UKJ3aH1Tbzi24VRW1TK8=
+dario.cat/mergo v1.0.1 h1:Ra4+bf83h2ztPIQYNP99R6m+Y7KfnARDfID+a+vLl4s=
-dario.cat/mergo v1.0.2/go.mod h1:E/hbnu0NxMFBjpMIE34DRGLWqDy0g5FuKDhCb31ngxA=
+dario.cat/mergo v1.0.1/go.mod h1:uNxQE+84aUszobStD9th8a29P2fMDhsBdgRYvZOxGmk=
 github.com/alecthomas/template v0.0.0-20160405071501-a0175ee3bccc/go.mod h1:LOuyumcjzFXgccqObfd/Ljyb9UuFJ6TxHnclSeseNhc=
 github.com/alecthomas/template v0.0.0-20190718012654-fb15b899a751/go.mod h1:LOuyumcjzFXgccqObfd/Ljyb9UuFJ6TxHnclSeseNhc=
 github.com/alecthomas/units v0.0.0-20151022065526-2efee857e7cf/go.mod h1:ybxpYRFXyAe+OPACYpWeL0wqObRcbAqCMya13uyzqw0=
@@ -22,6 +22,8 @@ github.com/cyberdelia/go-metrics-graphite v0.0.0-20161219230853-39f87cc3b432/go.
 github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 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/emirpasic/gods v1.18.1 h1:FXtiHYKDGKCW2KzwZKx0iC0PQmdlorYgdFG9jPXJ1Bc=
+github.com/emirpasic/gods v1.18.1/go.mod h1:8tpGGwCnJ5H4r6BWwaV6OrWmMoPhUl5jm/FMNAnJvWQ=
 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 v0.20.4 h1:Ik47r1fy3jRVU+1eYzKSW3ho2UgBVTVnUS8O993584U=
@@ -33,6 +35,8 @@ github.com/go-logfmt/logfmt v0.3.0/go.mod h1:Qt1PoO58o5twSAckw1HlFXLmHsOX5/0LbT9
 github.com/go-logfmt/logfmt v0.4.0/go.mod h1:3RMwSq7FuexP4Kalkev3ejPJsZTpXXBr9+V4qmtdjCk=
 github.com/go-logfmt/logfmt v0.5.0/go.mod h1:wCYkCAKZfumFQihp8CzCvQ3paCTfi41vtzG1KdI/P7A=
 github.com/go-stack/stack v1.8.0/go.mod h1:v0f6uXyyMGvRgIKkXu+yp6POWl0qKG85gN/melR3HDY=
+github.com/go-test/deep v1.1.0 h1:WOcxcdHcvdgThNXjw0t76K42FXTU7HpNQWHpA2HHNlg=
+github.com/go-test/deep v1.1.0/go.mod h1:5C2ZWiW0ErCdrYzpqxLbTX7MG14M9iiw8DgHncVwcsE=
 github.com/gogo/protobuf v1.1.1/go.mod h1:r8qH/GZQm5c6nD/R0oafs1akxWv10x8SbQlK7atdtwQ=
 github.com/gogo/protobuf v1.3.2 h1:Ov1cvc58UF3b5XjBnZv7+opcTcQFZebYjWzi34vdm4Q=
 github.com/gogo/protobuf v1.3.2/go.mod h1:P1XiOD3dCwIKUDQYPy72D8LYyHL2YPYrpS2s69NZV8Q=
@@ -58,6 +62,10 @@ github.com/google/go-cmp v0.7.0/go.mod h1:pXiqmnSA92OHEEa9HXL2W4E7lf9JzCmGVUdgjX
 github.com/google/gofuzz v1.0.0/go.mod h1:dBl0BpW6vV/+mYPU4Po3pmUjxk6FQPldtuIdl/M65Eg=
 github.com/google/gopacket v1.1.19 h1:ves8RnFZPGiFnTS0uPQStjwru6uO6h+nlr9j6fL7kF8=
 github.com/google/gopacket v1.1.19/go.mod h1:iJ8V8n6KS+z2U1A8pUwu8bW5SyEMkXJB8Yo/Vo+TKTo=
+github.com/google/uuid v1.3.0 h1:t6JiXgmwXMjEs8VusXIJk2BXHsn+wx8BZdTaoZ5fu7I=
+github.com/google/uuid v1.3.0/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
+github.com/heimdalr/dag v1.4.0 h1:zG3JA4RDVLc55k3AXAgfwa+EgBNZ0TkfOO3C29Ucpmg=
+github.com/heimdalr/dag v1.4.0/go.mod h1:OCh6ghKmU0hPjtwMqWBoNxPmtRioKd1xSu7Zs4sbIqM=
 github.com/jpillora/backoff v1.0.0/go.mod h1:J/6gKK9jxlEcS3zixgDgUAsiuZ7yrSoa/FX5e0EB2j4=
 github.com/json-iterator/go v1.1.6/go.mod h1:+SdeFBvtyEkXs7REEP0seUULqWtbJapLOCVDaaPEHmU=
 github.com/json-iterator/go v1.1.10/go.mod h1:KdQUCv79m/52Kvf8AW2vK1V8akMuk1QjK/uOdHXbAo4=
@@ -145,10 +153,14 @@ github.com/stretchr/testify v1.4.0/go.mod h1:j7eGeouHqKxXV5pUuKE4zz7dFj8WfuZ+81P
 github.com/stretchr/testify v1.7.0/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
 github.com/stretchr/testify v1.10.0 h1:Xv5erBjTwe/5IxqUQTdXv5kgmIvbHo3QQyRwhJsOfJA=
 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/timandy/routine v1.1.5 h1:LSpm7Iijwb9imIPlucl4krpr2EeCeAUvifiQ9Uf5X+M=
-github.com/vishvananda/netlink v1.3.1/go.mod h1:ARtKouGSTGchR8aMwmkzC0qiNPrrWO5JS/XMVl45+b4=
+github.com/timandy/routine v1.1.5/go.mod h1:kXslgIosdY8LW0byTyPnenDgn4/azt2euufAq9rK51w=
-github.com/vishvananda/netns v0.0.5 h1:DfiHV+j8bA32MFM7bfEunvT8IAqQ/NzSJHtcmW5zdEY=
+github.com/vishvananda/netlink v1.3.0 h1:X7l42GfcV4S6E4vHTsw48qbrV+9PVojNfIhZcwQdrZk=
-github.com/vishvananda/netns v0.0.5/go.mod h1:SpkAiCQRtJ6TvvxPnOSyH3BMl6unz3xZlaprSwhNNJM=
+github.com/vishvananda/netlink v1.3.0/go.mod h1:i6NetklAujEcC6fK0JPjT8qSwWyO0HLn4UKG+hGqeJs=
+github.com/vishvananda/netns v0.0.4 h1:Oeaw1EM2JMxD51g9uhtC0D7erkIjgmj8+JZc26m1YX8=
+github.com/vishvananda/netns v0.0.4/go.mod h1:SpkAiCQRtJ6TvvxPnOSyH3BMl6unz3xZlaprSwhNNJM=
+github.com/wadey/synctrace v0.0.0-20250612192159-94547ef50dfe h1:dc8Q42VsX+ABr0drJw27f3smvGfcz7eB8rJx+IkVMAo=
+github.com/wadey/synctrace v0.0.0-20250612192159-94547ef50dfe/go.mod h1:F2VCml4UxGPgAAqqm9T0ZfnVRWITrQS1EMZM+KCAm/Q=
 github.com/yuin/goldmark v1.1.27/go.mod h1:3hX8gzYuyVAZsxl0MRgGTJEmQBFcNTphYh9decYSb74=
 github.com/yuin/goldmark v1.2.1/go.mod h1:3hX8gzYuyVAZsxl0MRgGTJEmQBFcNTphYh9decYSb74=
 golang.org/x/crypto v0.0.0-20180904163835-0709b304e793/go.mod h1:6SG95UA2DQfeDnfUPMdvaQW0Q7yPrPDi9nlGo2tz2b4=

handshake_ix.go

@@ -9,6 +9,7 @@ import (
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/header"
+	"github.com/wadey/synctrace"
 )
 
 // NOISE IX Handshakes
@@ -249,7 +250,8 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 		HandshakePacket:   make(map[uint8][]byte, 0),
 		lastHandshakeTime: hs.Details.Time,
 		relayState: RelayState{
-			relays:         nil,
+			RWMutex:        synctrace.NewRWMutex("relay-state"),
+			relays:         map[netip.Addr]struct{}{},
 			relayForByAddr: map[netip.Addr]*Relay{},
 			relayForByIdx:  map[uint32]*Relay{},
 		},
@@ -457,7 +459,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 			Info("Handshake message sent")
 	}
 
-	f.connectionManager.AddTrafficWatch(hostinfo)
+	f.connectionManager.AddTrafficWatch(hostinfo.localIndexId)
 
 	hostinfo.remotes.ResetBlockedRemotes()
 
@@ -652,7 +654,7 @@ func ixHandshakeStage2(f *Interface, addr netip.AddrPort, via *ViaSender, hh *Ha
 
 	// Complete our handshake and update metrics, this will replace any existing tunnels for the vpnAddrs here
 	f.handshakeManager.Complete(hostinfo, f)
-	f.connectionManager.AddTrafficWatch(hostinfo)
+	f.connectionManager.AddTrafficWatch(hostinfo.localIndexId)
 
 	if f.l.Level >= logrus.DebugLevel {
 		hostinfo.logger(f.l).Debugf("Sending %d stored packets", len(hh.packetStore))

handshake_manager.go

@@ -8,7 +8,6 @@ import (
 	"errors"
 	"net/netip"
 	"slices"
-	"sync"
 	"time"
 
 	"github.com/rcrowley/go-metrics"
@@ -16,6 +15,7 @@ import (
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/udp"
+	"github.com/wadey/synctrace"
 )
 
 const (
@@ -45,7 +45,7 @@ type HandshakeConfig struct {
 
 type HandshakeManager struct {
 	// Mutex for interacting with the vpnIps and indexes maps
-	sync.RWMutex
+	synctrace.RWMutex
 
 	vpnIps  map[netip.Addr]*HandshakeHostInfo
 	indexes map[uint32]*HandshakeHostInfo
@@ -66,7 +66,7 @@ type HandshakeManager struct {
 }
 
 type HandshakeHostInfo struct {
-	sync.Mutex
+	synctrace.Mutex
 
 	startTime   time.Time        // Time that we first started trying with this handshake
 	ready       bool             // Is the handshake ready
@@ -104,6 +104,7 @@ func (hh *HandshakeHostInfo) cachePacket(l *logrus.Logger, t header.MessageType,
 
 func NewHandshakeManager(l *logrus.Logger, mainHostMap *HostMap, lightHouse *LightHouse, outside udp.Conn, config HandshakeConfig) *HandshakeManager {
 	return &HandshakeManager{
+		RWMutex:                synctrace.NewRWMutex("handshake-manager"),
 		vpnIps:                 map[netip.Addr]*HandshakeHostInfo{},
 		indexes:                map[uint32]*HandshakeHostInfo{},
 		mainHostMap:            mainHostMap,
@@ -111,7 +112,7 @@ func NewHandshakeManager(l *logrus.Logger, mainHostMap *HostMap, lightHouse *Lig
 		outside:                outside,
 		config:                 config,
 		trigger:                make(chan netip.Addr, config.triggerBuffer),
-		OutboundHandshakeTimer: NewLockingTimerWheel[netip.Addr](config.tryInterval, hsTimeout(config.retries, config.tryInterval)),
+		OutboundHandshakeTimer: NewLockingTimerWheel[netip.Addr]("outbound-handshake-timer", config.tryInterval, hsTimeout(config.retries, config.tryInterval)),
 		messageMetrics:         config.messageMetrics,
 		metricInitiated:        metrics.GetOrRegisterCounter("handshake_manager.initiated", nil),
 		metricTimedOut:         metrics.GetOrRegisterCounter("handshake_manager.timed_out", nil),
@@ -450,13 +451,15 @@ func (hm *HandshakeManager) StartHandshake(vpnAddr netip.Addr, cacheCb func(*Han
 		vpnAddrs:        []netip.Addr{vpnAddr},
 		HandshakePacket: make(map[uint8][]byte, 0),
 		relayState: RelayState{
-			relays:         nil,
+			RWMutex:        synctrace.NewRWMutex("relay-state"),
+			relays:         map[netip.Addr]struct{}{},
 			relayForByAddr: map[netip.Addr]*Relay{},
 			relayForByIdx:  map[uint32]*Relay{},
 		},
 	}
 
 	hh := &HandshakeHostInfo{
+		Mutex:     synctrace.NewMutex("handshake-hostinfo"),
 		hostinfo:  hostinfo,
 		startTime: time.Now(),
 	}

hostmap.go

@@ -4,8 +4,6 @@ import (
 	"errors"
 	"net"
 	"net/netip"
-	"slices"
-	"sync"
 	"sync/atomic"
 	"time"
 
@@ -15,6 +13,7 @@ import (
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/header"
+	"github.com/wadey/synctrace"
 )
 
 // const ProbeLen = 100
@@ -54,22 +53,22 @@ type Relay struct {
 }
 
 type HostMap struct {
-	sync.RWMutex    //Because we concurrently read and write to our maps
+	synctrace.RWMutex //Because we concurrently read and write to our maps
-	Indexes         map[uint32]*HostInfo
+	Indexes           map[uint32]*HostInfo
-	Relays          map[uint32]*HostInfo // Maps a Relay IDX to a Relay HostInfo object
+	Relays            map[uint32]*HostInfo // Maps a Relay IDX to a Relay HostInfo object
-	RemoteIndexes   map[uint32]*HostInfo
+	RemoteIndexes     map[uint32]*HostInfo
-	Hosts           map[netip.Addr]*HostInfo
+	Hosts             map[netip.Addr]*HostInfo
-	preferredRanges atomic.Pointer[[]netip.Prefix]
+	preferredRanges   atomic.Pointer[[]netip.Prefix]
-	l               *logrus.Logger
+	l                 *logrus.Logger
 }
 
 // For synchronization, treat the pointed-to Relay struct as immutable. To edit the Relay
 // struct, make a copy of an existing value, edit the fileds in the copy, and
 // then store a pointer to the new copy in both realyForBy* maps.
 type RelayState struct {
-	sync.RWMutex
+	synctrace.RWMutex
 
-	relays []netip.Addr // Ordered set of VpnAddrs of Hosts to use as relays to access this peer
+	relays map[netip.Addr]struct{} // Set of vpnAddr's of Hosts to use as relays to access this peer
 	// For data race avoidance, the contents of a *Relay are treated immutably. To update a *Relay, copy the existing data,
 	// modify what needs to be updated, and store the new modified copy in the relayForByIp and relayForByIdx maps (with
 	// the RelayState Lock held)
@@ -80,12 +79,7 @@ type RelayState struct {
 func (rs *RelayState) DeleteRelay(ip netip.Addr) {
 	rs.Lock()
 	defer rs.Unlock()
-	for idx, val := range rs.relays {
+	delete(rs.relays, ip)
-		if val == ip {
-			rs.relays = append(rs.relays[:idx], rs.relays[idx+1:]...)
-			return
-		}
-	}
 }
 
 func (rs *RelayState) UpdateRelayForByIpState(vpnIp netip.Addr, state int) {
@@ -130,16 +124,16 @@ func (rs *RelayState) GetRelayForByAddr(addr netip.Addr) (*Relay, bool) {
 func (rs *RelayState) InsertRelayTo(ip netip.Addr) {
 	rs.Lock()
 	defer rs.Unlock()
-	if !slices.Contains(rs.relays, ip) {
+	rs.relays[ip] = struct{}{}
-		rs.relays = append(rs.relays, ip)
-	}
 }
 
 func (rs *RelayState) CopyRelayIps() []netip.Addr {
-	ret := make([]netip.Addr, len(rs.relays))
 	rs.RLock()
 	defer rs.RUnlock()
-	copy(ret, rs.relays)
+	ret := make([]netip.Addr, 0, len(rs.relays))
+	for ip := range rs.relays {
+		ret = append(ret, ip)
+	}
 	return ret
 }
 
@@ -256,14 +250,6 @@ 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 {
@@ -302,6 +288,7 @@ func NewHostMapFromConfig(l *logrus.Logger, c *config.C) *HostMap {
 
 func newHostMap(l *logrus.Logger) *HostMap {
 	return &HostMap{
+		RWMutex:       synctrace.NewRWMutex("hostmap"),
 		Indexes:       map[uint32]*HostInfo{},
 		Relays:        map[uint32]*HostInfo{},
 		RemoteIndexes: map[uint32]*HostInfo{},

hostmap_test.go

@@ -7,7 +7,6 @@ import (
 	"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) {
@@ -216,31 +215,3 @@ func TestHostMap_reload(t *testing.T) {
 	c.ReloadConfigString("preferred_ranges: [1.1.1.1/32]")
 	assert.Equal(t, []string{"1.1.1.1/32"}, toS(hm.GetPreferredRanges()))
 }
-
-func TestHostMap_RelayState(t *testing.T) {
-	h1 := &HostInfo{vpnAddrs: []netip.Addr{netip.MustParseAddr("0.0.0.1")}, localIndexId: 1}
-	a1 := netip.MustParseAddr("::1")
-	a2 := netip.MustParseAddr("2001::1")
-
-	h1.relayState.InsertRelayTo(a1)
-	assert.Equal(t, []netip.Addr{a1}, h1.relayState.relays)
-	h1.relayState.InsertRelayTo(a2)
-	assert.Equal(t, []netip.Addr{a1, a2}, h1.relayState.relays)
-	// 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, a1, currentRelays[0])
-
-	// Deleting the last one in the list works ok
-	h1.relayState.DeleteRelay(a2)
-	assert.Equal(t, []netip.Addr{a1}, h1.relayState.relays)
-
-	// Deleting an element not in the list works ok
-	h1.relayState.DeleteRelay(a2)
-	assert.Equal(t, []netip.Addr{a1}, h1.relayState.relays)
-
-	// Deleting the only element in the list works ok
-	h1.relayState.DeleteRelay(a1)
-	assert.Equal(t, []netip.Addr{}, h1.relayState.relays)
-
-}

inside.go

@@ -288,7 +288,7 @@ 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)
+	f.connectionManager.Out(via.localIndexId)
 
 	// 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.
@@ -356,7 +356,7 @@ 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)
+	f.connectionManager.Out(hostinfo.localIndexId)
 
 	// 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 addrs and enable a faster roaming.

interface.go

@@ -2,7 +2,6 @@ package nebula
 
 import (
 	"context"
-	"crypto/fips140"
 	"errors"
 	"fmt"
 	"io"
@@ -25,23 +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
+	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
@@ -158,9 +157,6 @@ 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")
-	}
 
 	cs := c.pki.getCertState()
 	ifce := &Interface{
@@ -185,7 +181,7 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 		myVpnAddrsTable:       cs.myVpnAddrsTable,
 		myBroadcastAddrsTable: cs.myVpnBroadcastAddrsTable,
 		relayManager:          c.relayManager,
-		connectionManager:     c.connectionManager,
+
 		conntrackCacheTimeout: c.ConntrackCacheTimeout,
 
 		metricHandshakes: metrics.GetOrRegisterHistogram("handshakes", nil, metrics.NewExpDecaySample(1028, 0.015)),
@@ -202,7 +198,7 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 	ifce.reQueryEvery.Store(c.reQueryEvery)
 	ifce.reQueryWait.Store(int64(c.reQueryWait))
 
-	ifce.connectionManager.intf = ifce
+	ifce.connectionManager = newConnectionManager(ctx, c.l, ifce, c.checkInterval, c.pendingDeletionInterval, c.punchy)
 
 	return ifce, nil
 }
@@ -221,7 +217,6 @@ func (f *Interface) activate() {
 	f.l.WithField("interface", f.inside.Name()).WithField("networks", f.myVpnNetworks).
 		WithField("build", f.version).WithField("udpAddr", addr).
 		WithField("boringcrypto", boringEnabled()).
-		WithField("fips140", fips140.Enabled()).
 		Info("Nebula interface is active")
 
 	metrics.GetOrRegisterGauge("routines", nil).Update(int64(f.routines))

lighthouse.go

@@ -9,7 +9,6 @@ import (
 	"net/netip"
 	"slices"
 	"strconv"
-	"sync"
 	"sync/atomic"
 	"time"
 
@@ -21,15 +20,16 @@ import (
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/udp"
 	"github.com/slackhq/nebula/util"
+	"github.com/wadey/synctrace"
 )
 
 var ErrHostNotKnown = errors.New("host not known")
 
 type LightHouse struct {
 	//TODO: We need a timer wheel to kick out vpnAddrs that haven't reported in a long time
-	sync.RWMutex //Because we concurrently read and write to our maps
+	synctrace.RWMutex //Because we concurrently read and write to our maps
-	ctx          context.Context
+	ctx               context.Context
-	amLighthouse bool
+	amLighthouse      bool
 
 	myVpnNetworks      []netip.Prefix
 	myVpnNetworksTable *bart.Lite
@@ -96,6 +96,7 @@ func NewLightHouseFromConfig(ctx context.Context, l *logrus.Logger, c *config.C,
 	}
 
 	h := LightHouse{
+		RWMutex:            synctrace.NewRWMutex("lighthouse"),
 		ctx:                ctx,
 		amLighthouse:       amLighthouse,
 		myVpnNetworks:      cs.myVpnNetworks,
@@ -472,6 +473,7 @@ func (lh *LightHouse) QueryServer(vpnAddr netip.Addr) {
 		return
 	}
 
+	synctrace.ChanDebugSend("lighthouse-querychan")
 	lh.queryChan <- vpnAddr
 }
 
@@ -725,9 +727,11 @@ func (lh *LightHouse) startQueryWorker() {
 		nb := make([]byte, 12, 12)
 		out := make([]byte, mtu)
 
+		synctrace.ChanDebugRecvLock("lighthouse-querychan")
 		for {
 			select {
 			case <-lh.ctx.Done():
+				synctrace.ChanDebugRecvUnlock("lighthouse-querychan")
 				return
 			case addr := <-lh.queryChan:
 				lh.innerQueryServer(addr, nb, out)

main.go

@@ -185,7 +185,6 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 
 	hostMap := NewHostMapFromConfig(l, c)
 	punchy := NewPunchyFromConfig(l, c)
-	connManager := newConnectionManagerFromConfig(l, c, hostMap, punchy)
 	lightHouse, err := NewLightHouseFromConfig(ctx, l, c, pki.getCertState(), udpConns[0], punchy)
 	if err != nil {
 		return nil, util.ContextualizeIfNeeded("Failed to initialize lighthouse handler", err)
@@ -221,26 +220,31 @@ 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,
-		Firewall:              fw,
+		Firewall:                fw,
-		ServeDns:              serveDns,
+		ServeDns:                serveDns,
-		HandshakeManager:      handshakeManager,
+		HandshakeManager:        handshakeManager,
-		connectionManager:     connManager,
+		lightHouse:              lightHouse,
-		lightHouse:            lightHouse,
+		checkInterval:           time.Second * time.Duration(checkInterval),
-		tryPromoteEvery:       c.GetUint32("counters.try_promote", defaultPromoteEvery),
+		pendingDeletionInterval: time.Second * time.Duration(pendingDeletionInterval),
-		reQueryEvery:          c.GetUint32("counters.requery_every_packets", defaultReQueryEvery),
+		tryPromoteEvery:         c.GetUint32("counters.try_promote", defaultPromoteEvery),
-		reQueryWait:           c.GetDuration("timers.requery_wait_duration", defaultReQueryWait),
+		reQueryEvery:            c.GetUint32("counters.requery_every_packets", defaultReQueryEvery),
-		DropLocalBroadcast:    c.GetBool("tun.drop_local_broadcast", false),
+		reQueryWait:             c.GetDuration("timers.requery_wait_duration", defaultReQueryWait),
-		DropMulticast:         c.GetBool("tun.drop_multicast", false),
+		DropLocalBroadcast:      c.GetBool("tun.drop_local_broadcast", false),
-		routines:              routines,
+		DropMulticast:           c.GetBool("tun.drop_multicast", false),
-		MessageMetrics:        messageMetrics,
+		routines:                routines,
-		version:               buildVersion,
+		MessageMetrics:          messageMetrics,
-		relayManager:          NewRelayManager(ctx, l, hostMap, c),
+		version:                 buildVersion,
-		punchy:                punchy,
+		relayManager:            NewRelayManager(ctx, l, hostMap, c),
+		punchy:                  punchy,
+
 		ConntrackCacheTimeout: conntrackCacheTimeout,
 		l:                     l,
 	}
@@ -292,6 +296,5 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		statsStart,
 		dnsStart,
 		lightHouse.StartUpdateWorker,
-		connManager.Start,
 	}, nil
 }

noise.go

@@ -25,11 +25,6 @@ func NewNebulaCipherState(s *noise.CipherState) *NebulaCipherState {
 
 }
 
-type cipherAEADDanger interface {
-	EncryptDanger(out, ad, plaintext []byte, n uint64, nb []byte) ([]byte, error)
-	DecryptDanger(out, ad, plaintext []byte, n uint64, nb []byte) ([]byte, error)
-}
-
 // EncryptDanger encrypts and authenticates a given payload.
 //
 // out is a destination slice to hold the output of the EncryptDanger operation.
@@ -40,25 +35,20 @@ type cipherAEADDanger interface {
 //     be re-used by callers to minimize garbage collection.
 func (s *NebulaCipherState) EncryptDanger(out, ad, plaintext []byte, n uint64, nb []byte) ([]byte, error) {
 	if s != nil {
-		switch ce := s.c.(type) {
+		// TODO: Is this okay now that we have made messageCounter atomic?
-		case cipherAEADDanger:
+		// Alternative may be to split the counter space into ranges
-			return ce.EncryptDanger(out, ad, plaintext, n, nb)
+		//if n <= s.n {
-		default:
+		//	return nil, errors.New("CRITICAL: a duplicate counter value was used")
-			// TODO: Is this okay now that we have made messageCounter atomic?
+		//}
-			// Alternative may be to split the counter space into ranges
+		//s.n = n
-			//if n <= s.n {
+		nb[0] = 0
-			//	return nil, errors.New("CRITICAL: a duplicate counter value was used")
+		nb[1] = 0
-			//}
+		nb[2] = 0
-			//s.n = n
+		nb[3] = 0
-			nb[0] = 0
+		noiseEndianness.PutUint64(nb[4:], n)
-			nb[1] = 0
+		out = s.c.(cipher.AEAD).Seal(out, nb, plaintext, ad)
-			nb[2] = 0
+		//l.Debugf("Encryption: outlen: %d, nonce: %d, ad: %s, plainlen %d", len(out), n, ad, len(plaintext))
-			nb[3] = 0
+		return out, nil
-			noiseEndianness.PutUint64(nb[4:], n)
-			out = s.c.(cipher.AEAD).Seal(out, nb, plaintext, ad)
-			//l.Debugf("Encryption: outlen: %d, nonce: %d, ad: %s, plainlen %d", len(out), n, ad, len(plaintext))
-			return out, nil
-		}
 	} else {
 		return nil, errors.New("no cipher state available to encrypt")
 	}
@@ -66,17 +56,12 @@ func (s *NebulaCipherState) EncryptDanger(out, ad, plaintext []byte, n uint64, n
 
 func (s *NebulaCipherState) DecryptDanger(out, ad, ciphertext []byte, n uint64, nb []byte) ([]byte, error) {
 	if s != nil {
-		switch ce := s.c.(type) {
+		nb[0] = 0
-		case cipherAEADDanger:
+		nb[1] = 0
-			return ce.DecryptDanger(out, ad, ciphertext, n, nb)
+		nb[2] = 0
-		default:
+		nb[3] = 0
-			nb[0] = 0
+		noiseEndianness.PutUint64(nb[4:], n)
-			nb[1] = 0
+		return s.c.(cipher.AEAD).Open(out, nb, ciphertext, ad)
-			nb[2] = 0
-			nb[3] = 0
-			noiseEndianness.PutUint64(nb[4:], n)
-			return s.c.(cipher.AEAD).Open(out, nb, ciphertext, ad)
-		}
 	} else {
 		return []byte{}, nil
 	}

noiseutil/fips140.go

@@ -1,78 +0,0 @@
-//go:build fips140v1.0
-// +build fips140v1.0
-
-package noiseutil
-
-import (
-	"crypto/cipher"
-	"encoding/binary"
-
-	// unsafe needed for go:linkname
-	_ "unsafe"
-
-	"github.com/flynn/noise"
-)
-
-// EncryptLockNeeded indicates if calls to Encrypt need a lock
-// This is true for fips140 because the Seal function verifies that the
-// nonce is strictly increasing.
-const EncryptLockNeeded = true
-
-// TODO: Use NewGCMWithCounterNonce once available:
-// - https://github.com/golang/go/issues/73110
-// Using tls.aeadAESGCM gives us the TLS 1.2 GCM, which also verifies
-// that the nonce is strictly increasing.
-//
-//go:linkname aeadAESGCM crypto/tls.aeadAESGCM
-func aeadAESGCM(key, noncePrefix []byte) cipher.AEAD
-
-type cipherFn struct {
-	fn   func([32]byte) noise.Cipher
-	name string
-}
-
-func (c cipherFn) Cipher(k [32]byte) noise.Cipher { return c.fn(k) }
-func (c cipherFn) CipherName() string             { return c.name }
-
-// CipherAESGCM is the AES256-GCM AEAD cipher (using aeadAESGCM when fips140 is enabled)
-var CipherAESGCM noise.CipherFunc = cipherFn{cipherAESGCM, "AESGCM"}
-
-// tls.aeadAESGCM uses a 4 byte static prefix and an 8 byte nonce
-var emptyPrefix = []byte{0, 0, 0, 0}
-
-func cipherAESGCM(k [32]byte) noise.Cipher {
-	gcm := aeadAESGCM(k[:], emptyPrefix)
-	return aeadCipher{
-		gcm,
-		func(n uint64) []byte {
-			// tls.aeadAESGCM uses a 4 byte static prefix and an 8 byte nonce
-			var nonce [8]byte
-			binary.BigEndian.PutUint64(nonce[:], n)
-			return nonce[:]
-		},
-	}
-}
-
-type aeadCipher struct {
-	cipher.AEAD
-	nonce func(uint64) []byte
-}
-
-func (c aeadCipher) Encrypt(out []byte, n uint64, ad, plaintext []byte) []byte {
-	return c.Seal(out, c.nonce(n), plaintext, ad)
-}
-
-func (c aeadCipher) Decrypt(out []byte, n uint64, ad, ciphertext []byte) ([]byte, error) {
-	return c.Open(out, c.nonce(n), ciphertext, ad)
-}
-
-func (c aeadCipher) EncryptDanger(out, ad, plaintext []byte, n uint64, nb []byte) ([]byte, error) {
-	binary.BigEndian.PutUint64(nb[4:], n)
-	out = c.Seal(out, nb[4:], plaintext, ad)
-	return out, nil
-}
-
-func (c aeadCipher) DecryptDanger(out, ad, ciphertext []byte, n uint64, nb []byte) ([]byte, error) {
-	binary.BigEndian.PutUint64(nb[4:], n)
-	return c.Open(out, nb[4:], ciphertext, ad)
-}

noiseutil/fips140_test.go

@@ -1,42 +0,0 @@
-//go:build fips140v1.0
-// +build fips140v1.0
-
-package noiseutil
-
-import (
-	"crypto/fips140"
-	"encoding/hex"
-	"log"
-	"testing"
-
-	"github.com/stretchr/testify/assert"
-)
-
-func TestEncryptLockNeeded(t *testing.T) {
-	assert.True(t, EncryptLockNeeded)
-}
-
-// Ensure NewAESGCM validates the nonce is non-repeating
-func TestNewAESGCM(t *testing.T) {
-	assert.True(t, fips140.Enabled())
-
-	key, _ := hex.DecodeString("feffe9928665731c6d6a8f9467308308feffe9928665731c6d6a8f9467308308")
-	iv, _ := hex.DecodeString("00000000facedbaddecaf888")
-	plaintext, _ := hex.DecodeString("d9313225f88406e5a55909c5aff5269a86a7a9531534f7da2e4c303d8a318a721c3c0c95956809532fcf0e2449a6b525b16aedf5aa0de657ba637b39")
-	aad, _ := hex.DecodeString("feedfacedeadbeeffeedfacedeadbeefabaddad2")
-	expected, _ := hex.DecodeString("72ce2ea385f88c20d856e9d1248c2ca08562bbe8a61459ffae06ec393540518e9b6b4c40a146053f26a3df83c5384a48d273148b15aba64d970107432b2892741359275676441c1572c3fa9e")
-
-	var keyArray [32]byte
-	copy(keyArray[:], key)
-	c := CipherAESGCM.Cipher(keyArray)
-	aead := c.(aeadCipher).AEAD
-
-	dst := aead.Seal([]byte{}, iv, plaintext, aad)
-	log.Printf("%x", dst)
-	assert.Equal(t, expected, dst)
-
-	// We expect this to fail since we are re-encrypting with a repeat IV
-	assert.PanicsWithValue(t, "crypto/cipher: counter decreased", func() {
-		dst = aead.Seal([]byte{}, iv, plaintext, aad)
-	})
-}

noiseutil/notboring.go

@@ -1,5 +1,5 @@
-//go:build !boringcrypto && !fips140v1.0
+//go:build !boringcrypto
-// +build !boringcrypto,!fips140v1.0
+// +build !boringcrypto
 
 package noiseutil
 

noiseutil/notboring_test.go

@@ -1,5 +1,5 @@
-//go:build !boringcrypto && !fips140v1.0
+//go:build !boringcrypto
-// +build !boringcrypto,!fips140v1.0
+// +build !boringcrypto
 
 package noiseutil
 

outside.go

@@ -81,7 +81,7 @@ 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)
+			f.connectionManager.In(hostinfo.localIndexId)
 			f.connectionManager.RelayUsed(h.RemoteIndex)
 
 			relay, ok := hostinfo.relayState.QueryRelayForByIdx(h.RemoteIndex)
@@ -213,7 +213,7 @@ func (f *Interface) readOutsidePackets(ip netip.AddrPort, via *ViaSender, out []
 
 	f.handleHostRoaming(hostinfo, ip)
 
-	f.connectionManager.In(hostinfo)
+	f.connectionManager.In(hostinfo.localIndexId)
 }
 
 // closeTunnel closes a tunnel locally, it does not send a closeTunnel packet to the remote
@@ -498,7 +498,7 @@ func (f *Interface) decryptToTun(hostinfo *HostInfo, messageCounter uint64, out
 		return false
 	}
 
-	f.connectionManager.In(hostinfo)
+	f.connectionManager.In(hostinfo.localIndexId)
 	_, err = f.readers[q].Write(out)
 	if err != nil {
 		f.l.WithError(err).Error("Failed to write to tun")

remote_list.go

@@ -7,11 +7,11 @@ import (
 	"slices"
 	"sort"
 	"strconv"
-	"sync"
 	"sync/atomic"
 	"time"
 
 	"github.com/sirupsen/logrus"
+	"github.com/wadey/synctrace"
 )
 
 // forEachFunc is used to benefit folks that want to do work inside the lock
@@ -185,7 +185,7 @@ func (hr *hostnamesResults) GetAddrs() []netip.AddrPort {
 // It serves as a local cache of query replies, host update notifications, and locally learned addresses
 type RemoteList struct {
 	// Every interaction with internals requires a lock!
-	sync.RWMutex
+	synctrace.RWMutex
 
 	// The full list of vpn addresses assigned to this host
 	vpnAddrs []netip.Addr
@@ -215,6 +215,7 @@ type RemoteList struct {
 // NewRemoteList creates a new empty RemoteList
 func NewRemoteList(vpnAddrs []netip.Addr, shouldAdd func(netip.Addr) bool) *RemoteList {
 	r := &RemoteList{
+		RWMutex:   synctrace.NewRWMutex("remote-list"),
 		vpnAddrs:  make([]netip.Addr, len(vpnAddrs)),
 		addrs:     make([]netip.AddrPort, 0),
 		relays:    make([]netip.Addr, 0),

sshd/server.go

@@ -5,10 +5,10 @@ import (
 	"errors"
 	"fmt"
 	"net"
-	"sync"
 
 	"github.com/armon/go-radix"
 	"github.com/sirupsen/logrus"
+	"github.com/wadey/synctrace"
 	"golang.org/x/crypto/ssh"
 )
 
@@ -28,7 +28,7 @@ type SSHServer struct {
 	listener    net.Listener
 
 	// Locks the conns/counter to avoid concurrent map access
-	connsLock sync.Mutex
+	connsLock synctrace.Mutex
 	conns     map[int]*session
 	counter   int
 }
@@ -41,6 +41,7 @@ func NewSSHServer(l *logrus.Entry) (*SSHServer, error) {
 		l:           l,
 		commands:    radix.New(),
 		conns:       make(map[int]*session),
+		connsLock:   synctrace.NewMutex("ssh-server-conns"),
 	}
 
 	cc := ssh.CertChecker{

timeout.go

@@ -1,8 +1,9 @@
 package nebula
 
 import (
-	"sync"
 	"time"
+
+	"github.com/wadey/synctrace"
 )
 
 // How many timer objects should be cached
@@ -34,7 +35,7 @@ type TimerWheel[T any] struct {
 }
 
 type LockingTimerWheel[T any] struct {
-	m sync.Mutex
+	m synctrace.Mutex
 	t *TimerWheel[T]
 }
 
@@ -81,8 +82,9 @@ func NewTimerWheel[T any](min, max time.Duration) *TimerWheel[T] {
 }
 
 // NewLockingTimerWheel is version of TimerWheel that is safe for concurrent use with a small performance penalty
-func NewLockingTimerWheel[T any](min, max time.Duration) *LockingTimerWheel[T] {
+func NewLockingTimerWheel[T any](name string, min, max time.Duration) *LockingTimerWheel[T] {
 	return &LockingTimerWheel[T]{
+		m: synctrace.NewMutex(name),
 		t: NewTimerWheel[T](min, max),
 	}
 }

udp/errors.go

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

udp/udp_darwin.go

@@ -3,62 +3,20 @@
 
 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"
 	"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) {
-	lc := NewListenConfig(multi)
+	return NewGenericListener(l, ip, port, multi, batch)
-	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 {
@@ -85,116 +43,16 @@ func NewListenConfig(multi bool) net.ListenConfig {
 	}
 }
 
-//go:linkname sendto golang.org/x/sys/unix.sendto
+func (u *GenericConn) Rebind() error {
-//go:noescape
+	rc, err := u.UDPConn.SyscallConn()
-func sendto(s int, buf []byte, flags int, to unsafe.Pointer, addrlen int32) (err error)
-
-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
-	}
-
-	// Golang stdlib doesn't handle EAGAIN correctly in some situations so we do writes ourselves
-	// See https://github.com/golang/go/issues/73919
-	for {
-		//_, _, 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) {
-	buffer := make([]byte, MTU)
-
-	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()), buffer[:n])
-	}
-}
-
-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 err
 	}
 
-	return nil
+	return rc.Control(func(fd uintptr) {
+		err := syscall.SetsockoptInt(int(fd), unix.IPPROTO_IPV6, unix.IPV6_BOUND_IF, 0)
+		if err != nil {
+			u.l.WithError(err).Error("Failed to rebind udp socket")
+		}
+	})
 }

udp/udp_generic.go

@@ -1,7 +1,6 @@
-//go:build (!linux || android) && !e2e_testing && !darwin
+//go:build (!linux || android) && !e2e_testing
 // +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.

udp/udp_linux.go

@@ -221,7 +221,7 @@ 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 ErrInvalidIPv6RemoteForSocket
+		return fmt.Errorf("Listener is IPv4, but writing to IPv6 remote")
 	}
 
 	var rsa unix.RawSockaddrInet4

udp/udp_rio_windows.go

@@ -11,13 +11,13 @@ import (
 	"io"
 	"net"
 	"net/netip"
-	"sync"
 	"sync/atomic"
 	"syscall"
 	"unsafe"
 
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
+	"github.com/wadey/synctrace"
 	"golang.org/x/sys/windows"
 	"golang.zx2c4.com/wireguard/conn/winrio"
 )
@@ -46,7 +46,7 @@ type ringBuffer struct {
 	iocp       windows.Handle
 	isFull     bool
 	cq         winrio.Cq
-	mu         sync.Mutex
+	mu         synctrace.Mutex
 	overlapped windows.Overlapped
 }
 
@@ -64,7 +64,11 @@ func NewRIOListener(l *logrus.Logger, addr netip.Addr, port int) (*RIOConn, erro
 		return nil, errors.New("could not initialize winrio")
 	}
 
-	u := &RIOConn{l: l}
+	u := &RIOConn{
+		l:  l,
+		rx: ringBuffer{mu: synctrace.NewMutex("rio-rx")},
+		tx: ringBuffer{mu: synctrace.NewMutex("rio-tx")},
+	}
 
 	err := u.bind(&windows.SockaddrInet6{Addr: addr.As16(), Port: port})
 	if err != nil {
@@ -92,25 +96,6 @@ func (u *RIOConn) bind(sa windows.Sockaddr) error {
 	// 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 {
-		return err
-	}
-	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 {
-		return err
-	}
-
 	err = u.rx.Open()
 	if err != nil {
 		return err
@@ -141,12 +126,8 @@ func (u *RIOConn) ListenOut(r EncReader) {
 		// Just read one packet at a time
 		n, rua, err := u.receive(buffer)
 		if err != nil {
-			if errors.Is(err, net.ErrClosed) {
+			u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
-				u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
+			return
-				return
-			}
-			u.l.WithError(err).Error("unexpected udp socket receive error")
-			continue
 		}
 
 		r(netip.AddrPortFrom(netip.AddrFrom16(rua.Addr).Unmap(), (rua.Port>>8)|((rua.Port&0xff)<<8)), buffer[:n])