Merge branch 'main' into file-share

# Conflicts:
#	client/android/client.go
#	client/android/login.go
#	client/android/profile_prefs.go
#	client/internal/connect.go
#	client/internal/engine.go
#	client/mobile/profile_state.go
#	client/ui/i18n/locales/de/common.json
#	client/ui/i18n/locales/en/common.json
#	client/ui/i18n/locales/es/common.json
#	client/ui/i18n/locales/fr/common.json
#	client/ui/i18n/locales/hu/common.json
#	client/ui/i18n/locales/it/common.json
#	client/ui/i18n/locales/ja/common.json
#	client/ui/i18n/locales/pt/common.json
#	client/ui/i18n/locales/ru/common.json
#	client/ui/i18n/locales/zh-CN/common.json
#	client/ui/main.go

client/internal/acl/manager.go

@@ -116,11 +116,11 @@ func (d *DefaultManager) ApplyFiltering(networkMap *mgmProto.NetworkMap, dnsRout
 // firewall state, so an identical hash means an identical resulting ruleset.
 func (d *DefaultManager) firewallConfigHash(networkMap *mgmProto.NetworkMap, dnsRouteFeatureFlag bool) (uint64, error) {
 	return hashstructure.Hash(struct {
-		PeerRules            []*mgmProto.FirewallRule
-		PeerRulesIsEmpty     bool
-		RouteRules           []*mgmProto.RouteFirewallRule
-		RouteRulesIsEmpty    bool
-		DNSRouteFeatureFlag  bool
+		PeerRules           []*mgmProto.FirewallRule
+		PeerRulesIsEmpty    bool
+		RouteRules          []*mgmProto.RouteFirewallRule
+		RouteRulesIsEmpty   bool
+		DNSRouteFeatureFlag bool
 	}{
 		PeerRules:           networkMap.GetFirewallRules(),
 		PeerRulesIsEmpty:    networkMap.GetFirewallRulesIsEmpty(),
@@ -144,13 +144,13 @@ func (d *DefaultManager) applyPeerACLs(networkMap *mgmProto.NetworkMap) {
 		log.Warn("this peer is connected to a NetBird Management service with an older version. Allowing all traffic from connected peers")
 		rules = append(rules,
 			&mgmProto.FirewallRule{
-				PeerIP:    "0.0.0.0",
+				PeerIP:    "0.0.0.0", //nolint:staticcheck
 				Direction: mgmProto.RuleDirection_IN,
 				Action:    mgmProto.RuleAction_ACCEPT,
 				Protocol:  mgmProto.RuleProtocol_ALL,
 			},
 			&mgmProto.FirewallRule{
-				PeerIP:    "0.0.0.0",
+				PeerIP:    "0.0.0.0", //nolint:staticcheck
 				Direction: mgmProto.RuleDirection_OUT,
 				Action:    mgmProto.RuleAction_ACCEPT,
 				Protocol:  mgmProto.RuleProtocol_ALL,
@@ -407,7 +407,6 @@ func (d *DefaultManager) getRuleGroupingSelector(rule *mgmProto.FirewallRule) st
 	return fmt.Sprintf("%v:%v:%v:%s:%v", strconv.Itoa(int(rule.Direction)), rule.Action, rule.Protocol, rule.Port, rule.PortInfo)
 }
 
-
 // extractRuleIP extracts the peer IP from a firewall rule.
 // If sourcePrefixes is populated (new management), decode the first entry and use its address.
 // Otherwise fall back to the deprecated PeerIP string field (old management).

client/internal/acl/manager_test.go

@@ -5,9 +5,9 @@ import (
 	"net/netip"
 	"testing"
 
-	"github.com/golang/mock/gomock"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
+	"go.uber.org/mock/gomock"
 
 	"github.com/netbirdio/netbird/client/firewall"
 	"github.com/netbirdio/netbird/client/iface"
@@ -87,7 +87,7 @@ func TestDefaultManager(t *testing.T) {
 		networkMap.FirewallRules = append(
 			networkMap.FirewallRules,
 			&mgmProto.FirewallRule{
-				PeerIP:    "10.93.0.3",
+				PeerIP:    "10.93.0.3", //nolint:staticcheck
 				Direction: mgmProto.RuleDirection_IN,
 				Action:    mgmProto.RuleAction_DROP,
 				Protocol:  mgmProto.RuleProtocol_ICMP,
@@ -556,12 +556,12 @@ func TestApplyFilteringSkipsUnchangedConfig(t *testing.T) {
 
 func buildNetworkMap(peerRules, routeRules int) *mgmProto.NetworkMap {
 	nm := &mgmProto.NetworkMap{
-		FirewallRulesIsEmpty:      peerRules == 0,
+		FirewallRulesIsEmpty:       peerRules == 0,
 		RoutesFirewallRulesIsEmpty: routeRules == 0,
 	}
 	for i := range peerRules {
 		nm.FirewallRules = append(nm.FirewallRules, &mgmProto.FirewallRule{
-			PeerIP:    fmt.Sprintf("10.%d.%d.%d", i>>16&0xff, i>>8&0xff, i&0xff),
+			PeerIP:    fmt.Sprintf("10.%d.%d.%d", i>>16&0xff, i>>8&0xff, i&0xff), //nolint:staticcheck
 			Direction: mgmProto.RuleDirection_IN,
 			Action:    mgmProto.RuleAction_ACCEPT,
 			Protocol:  mgmProto.RuleProtocol_TCP,

client/internal/acl/mocks/iface_mapper.go

@@ -7,7 +7,7 @@ package mocks
 import (
 	reflect "reflect"
 
-	gomock "github.com/golang/mock/gomock"
+	gomock "go.uber.org/mock/gomock"
 	wgdevice "golang.zx2c4.com/wireguard/device"
 
 	"github.com/netbirdio/netbird/client/iface/device"

client/internal/conn_mgr.go

@@ -2,6 +2,7 @@ package internal
 
 import (
 	"context"
+	"maps"
 	"os"
 	"strconv"
 	"sync"
@@ -14,6 +15,7 @@ import (
 	"github.com/netbirdio/netbird/client/internal/peer"
 	"github.com/netbirdio/netbird/client/internal/peerstore"
 	"github.com/netbirdio/netbird/route"
+	mgmProto "github.com/netbirdio/netbird/shared/management/proto"
 )
 
 // lazyForce is the resolved local decision for lazy connections, layered above the
@@ -37,11 +39,13 @@ const (
 // The only exception is ActivatePeer, which is safe for concurrent use so the
 // DNS warm-up path can call it without contending on the engine mutex.
 type ConnMgr struct {
-	peerStore        *peerstore.Store
-	statusRecorder   *peer.Status
-	iface            lazyconn.WGIface
-	force            lazyForce
-	rosenpassEnabled bool
+	peerStore      *peerstore.Store
+	statusRecorder *peer.Status
+	iface          lazyconn.WGIface
+	force          lazyForce
+	// remoteLazyEnabled caches the account-wide lazy feature flag from management.
+	// It is the default for peers that do not carry a per-peer lazy hint.
+	remoteLazyEnabled bool
 
 	lazyConnMgr *manager.Manager
 	// lazyConnMgrMu guards the lazyConnMgr pointer for readers outside the
@@ -53,6 +57,10 @@ type ConnMgr struct {
 	// (re)armed (Mode A at arm time). Injected by the engine; nil disables the reconcile.
 	reconcileRoutedIPs func(peerKey string) error
 
+	// appliedExcludeList is the exclude set last handed to the lazy manager, kept so an
+	// unchanged set on the next sync skips the O(n) reconciliation.
+	appliedExcludeList map[string]bool
+
 	wg            sync.WaitGroup
 	lazyCtx       context.Context
 	lazyCtxCancel context.CancelFunc
@@ -66,78 +74,59 @@ func (e *ConnMgr) SetRoutedIPsReconciler(fn func(peerKey string) error) {
 
 func NewConnMgr(engineConfig *EngineConfig, statusRecorder *peer.Status, peerStore *peerstore.Store, iface lazyconn.WGIface) *ConnMgr {
 	e := &ConnMgr{
-		peerStore:        peerStore,
-		statusRecorder:   statusRecorder,
-		iface:            iface,
-		force:            resolveLazyForce(engineConfig.LazyConnection),
-		rosenpassEnabled: engineConfig.RosenpassEnabled,
+		peerStore:      peerStore,
+		statusRecorder: statusRecorder,
+		iface:          iface,
+		force:          resolveLazyForce(engineConfig.LazyConnection),
 	}
 	return e
 }
 
-// Start initializes the connection manager. It starts the lazy connection manager when a
-// local override forces it on; with no local override it waits for the management feature flag.
+// Start initializes the connection manager. The lazy connection manager always runs so that
+// per-peer lazy defaults (e.g. proxy peers) work even when the account flag is off; the
+// account flag and the local override decide the default lazy state per peer (see
+// PeerLazyDefault). Rosenpass peers stay lazy-capable too: their connections just never idle
+// on their own, since rosenpass rekey traffic keeps them active.
 func (e *ConnMgr) Start(ctx context.Context) {
 	if e.lazyConnMgr != nil {
 		log.Errorf("lazy connection manager is already started")
 		return
 	}
 
-	switch e.force {
-	case lazyForceOff:
-		log.Infof("lazy connection manager is disabled by local override (%s or MDM policy)", lazyconn.EnvLazyConn)
-		e.statusRecorder.UpdateLazyConnection(false)
-		return
-	case lazyForceNone:
-		log.Infof("lazy connection manager is managed by the management feature flag")
-		e.statusRecorder.UpdateLazyConnection(false)
-		return
-	}
-
-	if e.rosenpassEnabled {
-		log.Warnf("rosenpass connection manager is enabled, lazy connection manager will not be started")
-		e.statusRecorder.UpdateLazyConnection(false)
-		return
-	}
-
 	e.initLazyManager(ctx)
-	e.statusRecorder.UpdateLazyConnection(true)
+	e.statusRecorder.UpdateLazyConnection(e.PeerLazyDefault(mgmProto.LazyState_LazyStateDefault))
 }
 
-// UpdatedRemoteFeatureFlag is called when the remote feature flag is updated.
-// If enabled, it initializes the lazy connection manager and start it. Do not need to call Start() again.
-// If disabled, then it closes the lazy connection manager and open the connections to all peers.
-func (e *ConnMgr) UpdatedRemoteFeatureFlag(ctx context.Context, enabled bool) error {
-	// a local override (NB_LAZY_CONN or local config) takes precedence over management
-	if e.force != lazyForceNone {
-		return nil
+// UpdatedRemoteFeatureFlag caches the account-wide lazy feature flag. The manager itself is
+// not started or stopped here; the per-sync exclude-list reconciliation moves normal peers
+// between the lazy and always-active sets when the flag flips.
+func (e *ConnMgr) UpdatedRemoteFeatureFlag(_ context.Context, enabled bool) error {
+	e.remoteLazyEnabled = enabled
+	if e.isStartedWithLazyMgr() {
+		e.statusRecorder.UpdateLazyConnection(e.PeerLazyDefault(mgmProto.LazyState_LazyStateDefault))
+	}
+	return nil
+}
+
+// PeerLazyDefault reports whether a peer should be lazy. The local override
+// (NB_LAZY_CONN/MDM) wins over everything; without a local override the
+// management per-peer state applies (LazyStateLazy/Eager force the decision),
+// and LazyStateDefault follows the account-wide flag.
+func (e *ConnMgr) PeerLazyDefault(state mgmProto.LazyState) bool {
+	switch e.force {
+	case lazyForceOn:
+		return true
+	case lazyForceOff:
+		return false
 	}
 
-	if enabled {
-		// if the lazy connection manager is already started, do not start it again
-		if e.lazyConnMgr != nil {
-			return nil
-		}
-
-		if e.rosenpassEnabled {
-			log.Infof("rosenpass connection manager is enabled, lazy connection manager will not be started")
-			e.statusRecorder.UpdateLazyConnection(false)
-			return nil
-		}
-
-		log.Infof("lazy connection manager is enabled by the management feature flag")
-		e.initLazyManager(ctx)
-		e.statusRecorder.UpdateLazyConnection(true)
-		return e.addPeersToLazyConnManager()
-	} else {
-		if e.lazyConnMgr == nil {
-			e.statusRecorder.UpdateLazyConnection(false)
-			return nil
-		}
-		log.Infof("lazy connection manager is disabled by management feature flag")
-		e.closeManager(ctx)
-		e.statusRecorder.UpdateLazyConnection(false)
-		return nil
+	switch state {
+	case mgmProto.LazyState_LazyStateLazy:
+		return true
+	case mgmProto.LazyState_LazyStateEager:
+		return false
+	default:
+		return e.remoteLazyEnabled
 	}
 }
 
@@ -157,6 +146,13 @@ func (e *ConnMgr) SetExcludeList(ctx context.Context, peerIDs map[string]bool) {
 		return
 	}
 
+	// The exclude set is recomputed every sync but rarely changes; skip the O(n)
+	// store lookups and reconciliation when it matches what was already applied.
+	if maps.Equal(peerIDs, e.appliedExcludeList) {
+		return
+	}
+	e.appliedExcludeList = maps.Clone(peerIDs)
+
 	excludedPeers := make([]lazyconn.PeerConfig, 0, len(peerIDs))
 
 	for peerID := range peerIDs {
@@ -192,12 +188,16 @@ func (e *ConnMgr) SetExcludeList(ctx context.Context, peerIDs map[string]bool) {
 	}
 }
 
-func (e *ConnMgr) AddPeerConn(ctx context.Context, peerKey string, conn *peer.Conn) (exists bool) {
+// AddPeerConn registers a peer connection. permanent requests an always-active connection
+// (the peer belongs to the exclude set: a forwarder, or a peer that is not lazy by policy).
+// Non-permanent peers are handed to the lazy manager. The subsequent SetExcludeList call
+// reconciles membership for existing peers across flag flips.
+func (e *ConnMgr) AddPeerConn(ctx context.Context, peerKey string, conn *peer.Conn, permanent bool) (exists bool) {
 	if success := e.peerStore.AddPeerConn(peerKey, conn); !success {
 		return true
 	}
 
-	if !e.isStartedWithLazyMgr() {
+	if !e.isStartedWithLazyMgr() || permanent {
 		if err := conn.Open(ctx); err != nil {
 			conn.Log.Errorf("failed to open connection: %v", err)
 		}
@@ -296,6 +296,8 @@ func (e *ConnMgr) Close() {
 	e.lazyConnMgrMu.Lock()
 	e.lazyConnMgr = nil
 	e.lazyConnMgrMu.Unlock()
+
+	e.appliedExcludeList = nil
 }
 
 func (e *ConnMgr) initLazyManager(engineCtx context.Context) {
@@ -309,6 +311,8 @@ func (e *ConnMgr) initLazyManager(engineCtx context.Context) {
 	e.lazyCtx, e.lazyCtxCancel = context.WithCancel(engineCtx)
 	e.lazyConnMgrMu.Unlock()
 
+	e.appliedExcludeList = nil
+
 	e.wg.Add(1)
 	go func() {
 		defer e.wg.Done()
@@ -316,46 +320,6 @@ func (e *ConnMgr) initLazyManager(engineCtx context.Context) {
 	}()
 }
 
-func (e *ConnMgr) addPeersToLazyConnManager() error {
-	peers := e.peerStore.PeersPubKey()
-	lazyPeerCfgs := make([]lazyconn.PeerConfig, 0, len(peers))
-	for _, peerID := range peers {
-		var peerConn *peer.Conn
-		var exists bool
-		if peerConn, exists = e.peerStore.PeerConn(peerID); !exists {
-			log.Warnf("failed to find peer conn for peerID: %s", peerID)
-			continue
-		}
-
-		lazyPeerCfg := lazyconn.PeerConfig{
-			PublicKey:  peerID,
-			AllowedIPs: peerConn.WgConfig().AllowedIps,
-			PeerConnID: peerConn.ConnID(),
-			Log:        peerConn.Log,
-		}
-		lazyPeerCfgs = append(lazyPeerCfgs, lazyPeerCfg)
-	}
-
-	return e.lazyConnMgr.AddActivePeers(lazyPeerCfgs)
-}
-
-func (e *ConnMgr) closeManager(ctx context.Context) {
-	if e.lazyConnMgr == nil {
-		return
-	}
-
-	e.lazyCtxCancel()
-	e.wg.Wait()
-
-	e.lazyConnMgrMu.Lock()
-	e.lazyConnMgr = nil
-	e.lazyConnMgrMu.Unlock()
-
-	for _, peerID := range e.peerStore.PeersPubKey() {
-		e.peerStore.PeerConnOpen(ctx, peerID)
-	}
-}
-
 func (e *ConnMgr) isStartedWithLazyMgr() bool {
 	return e.lazyConnMgr != nil && e.lazyCtxCancel != nil
 }

client/internal/conn_mgr_test.go

@@ -16,6 +16,7 @@ import (
 	"github.com/netbirdio/netbird/client/internal/peer"
 	"github.com/netbirdio/netbird/client/internal/peerstore"
 	"github.com/netbirdio/netbird/monotime"
+	mgmProto "github.com/netbirdio/netbird/shared/management/proto"
 )
 
 func TestResolveLazyForce(t *testing.T) {
@@ -138,4 +139,91 @@ func TestInactivityThresholdEnv(t *testing.T) {
 	}
 }
 
+func TestPeerLazyDefault(t *testing.T) {
+	tests := []struct {
+		name          string
+		force         lazyForce
+		remoteEnabled bool
+		state         mgmProto.LazyState
+		want          bool
+	}{
+		{name: "force on wins over eager state", force: lazyForceOn, state: mgmProto.LazyState_LazyStateEager, want: true},
+		{name: "force off wins over lazy state", force: lazyForceOff, remoteEnabled: true, state: mgmProto.LazyState_LazyStateLazy, want: false},
+		{name: "none, default, account off -> active", force: lazyForceNone, state: mgmProto.LazyState_LazyStateDefault, want: false},
+		{name: "none, default, account on -> lazy", force: lazyForceNone, remoteEnabled: true, state: mgmProto.LazyState_LazyStateDefault, want: true},
+		{name: "none, lazy state, account off -> lazy", force: lazyForceNone, state: mgmProto.LazyState_LazyStateLazy, want: true},
+		{name: "none, eager state, account on -> active", force: lazyForceNone, remoteEnabled: true, state: mgmProto.LazyState_LazyStateEager, want: false},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			e := &ConnMgr{force: tt.force, remoteLazyEnabled: tt.remoteEnabled}
+			if got := e.PeerLazyDefault(tt.state); got != tt.want {
+				t.Fatalf("PeerLazyDefault(%v) = %v, want %v", tt.state, got, tt.want)
+			}
+		})
+	}
+}
+
 func durPtr(d time.Duration) *time.Duration { return &d }
+
+// TestToExcludedLazyPeers covers the per-peer lazy classification (proxy vs
+// normal, across the force/account-flag matrix). Forwarder-target exclusion is
+// covered by TestToExcludedLazyPeers_ForwardTarget.
+func TestToExcludedLazyPeers(t *testing.T) {
+	const (
+		normalKey = "normal"
+		lazyKey   = "lazy-state"
+		eagerKey  = "eager-state"
+	)
+
+	peers := []*mgmProto.RemotePeerConfig{
+		{WgPubKey: normalKey, AllowedIps: []string{"100.64.0.1/32"}},
+		{WgPubKey: lazyKey, AllowedIps: []string{"100.64.0.2/32"}, LazyState: mgmProto.LazyState_LazyStateLazy},
+		{WgPubKey: eagerKey, AllowedIps: []string{"100.64.0.3/32"}, LazyState: mgmProto.LazyState_LazyStateEager},
+	}
+
+	tests := []struct {
+		name          string
+		force         lazyForce
+		remoteEnabled bool
+		want          map[string]bool
+	}{
+		{
+			name:  "account off: lazy-state peer lazy, normal + eager active",
+			force: lazyForceNone, remoteEnabled: false,
+			want: map[string]bool{normalKey: true, eagerKey: true},
+		},
+		{
+			name:  "account on: only eager-state peer active",
+			force: lazyForceNone, remoteEnabled: true,
+			want: map[string]bool{eagerKey: true},
+		},
+		{
+			name:  "force off: everything active",
+			force: lazyForceOff, remoteEnabled: true,
+			want: map[string]bool{normalKey: true, lazyKey: true, eagerKey: true},
+		},
+		{
+			name:  "force on: nothing active",
+			force: lazyForceOn, remoteEnabled: false,
+			want: map[string]bool{},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			e := &Engine{connMgr: &ConnMgr{force: tt.force, remoteLazyEnabled: tt.remoteEnabled}}
+			got := e.toExcludedLazyPeers(peers)
+
+			if len(got) != len(tt.want) {
+				t.Fatalf("toExcludedLazyPeers() = %v, want %v", got, tt.want)
+			}
+			for k := range tt.want {
+				if !got[k] {
+					t.Fatalf("expected peer %s excluded, got %v", k, got)
+				}
+			}
+		})
+	}
+}

client/internal/connect.go

@@ -39,6 +39,7 @@ import (
 	"github.com/netbirdio/netbird/client/internal/updater"
 	"github.com/netbirdio/netbird/client/internal/updater/installer"
 	nbnet "github.com/netbirdio/netbird/client/net"
+	"github.com/netbirdio/netbird/client/netevents"
 	cProto "github.com/netbirdio/netbird/client/proto"
 	"github.com/netbirdio/netbird/client/ssh"
 	sshconfig "github.com/netbirdio/netbird/client/ssh/config"
@@ -72,18 +73,31 @@ type ConnectClient struct {
 	fileDropManager *filedrop.Manager
 
 	persistSyncResponse bool
+
+	// netMgr gates every reconnection loop on OS-reported network
+	// availability and sweeps connections on network change.
+	netMgr *netevents.Manager
+}
+
+// ConnectClientOption configures optional ConnectClient behavior.
+type ConnectClientOption func(*ConnectClient)
+
+// WithNetEvents injects the OS network event handling.
+func WithNetEvents(events *netevents.Manager) ConnectClientOption {
+	return func(c *ConnectClient) { c.netMgr = events }
 }
 
 func NewConnectClient(
 	ctx context.Context,
 	config *profilemanager.Config,
 	statusRecorder *peer.Status,
+	opts ...ConnectClientOption,
 ) *ConnectClient {
 	// Derive the run context here so Stop owns the cancel that unblocks the run
 	// loop. runCancel is set once at construction, so Stop can call it without
 	// racing the run loop's startup. Callers therefore need not cancel before Stop.
 	runCtx, runCancel := context.WithCancel(ctx)
-	return &ConnectClient{
+	c := &ConnectClient{
 		ctx:            runCtx,
 		runCancel:      runCancel,
 		runExited:      make(chan struct{}),
@@ -91,6 +105,10 @@ func NewConnectClient(
 		statusRecorder: statusRecorder,
 		engineMutex:    sync.Mutex{},
 	}
+	for _, opt := range opts {
+		opt(c)
+	}
+	return c
 }
 
 func (c *ConnectClient) SetUpdateManager(um *updater.Manager) {
@@ -282,6 +300,13 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan
 			return nil
 		}
 
+		// suspend connection attempts while the OS reports no usable network
+		if waited, err := c.netMgr.Wait(c.ctx); err != nil {
+			return nil
+		} else if waited {
+			backOff.Reset()
+		}
+
 		state.Set(StatusConnecting)
 
 		engineCtx, cancel := context.WithCancel(c.ctx)
@@ -293,7 +318,8 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan
 		}()
 
 		log.Debugf("connecting to the Management service %s", c.config.ManagementURL.Host)
-		mgmClient, err := mgm.NewClient(engineCtx, c.config.ManagementURL.Host, myPrivateKey, mgmTlsEnabled)
+		mgmClient, err := mgm.NewClient(engineCtx, c.config.ManagementURL.Host, myPrivateKey, mgmTlsEnabled,
+			mgm.WithNetEvents(c.netMgr))
 		if err != nil {
 			// On daemon shutdown / Down() the parent context is cancelled
 			// and the dial fails with "context canceled". Wrapping that
@@ -368,7 +394,7 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan
 		}()
 
 		// with the global Netbird config in hand connect (just a connection, no stream yet) Signal
-		signalClient, err := connectToSignal(engineCtx, loginResp.GetNetbirdConfig(), myPrivateKey)
+		signalClient, err := connectToSignal(engineCtx, loginResp.GetNetbirdConfig(), myPrivateKey, c.netMgr)
 		if err != nil {
 			log.Error(err)
 			return wrapErr(err)
@@ -404,7 +430,8 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan
 			engineConfig.StateDir = filepath.Dir(path)
 		}
 
-		relayManager := relayClient.NewManager(engineCtx, relayURLs, myPrivateKey.PublicKey().String(), engineConfig.MTU)
+		relayManager := relayClient.NewManager(engineCtx, relayURLs, myPrivateKey.PublicKey().String(), engineConfig.MTU,
+			relayClient.WithNetEvents(c.netMgr))
 		c.statusRecorder.SetRelayMgr(relayManager)
 		if len(relayURLs) > 0 {
 			if token != nil {
@@ -433,6 +460,7 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan
 			ClientMetrics:  c.clientMetrics,
 			MetricsCtx:     c.ctx,
 			FileDrop:       c.fileDropManager,
+			NetMgr:         c.netMgr,
 		}, mobileDependency)
 		engine.SetSyncResponsePersistence(c.persistSyncResponse)
 		c.engine = engine
@@ -489,6 +517,16 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan
 	// status stream stuck at Connecting.
 	err = backoff.Retry(operation, backoff.WithContext(backOff, c.ctx))
 	if err != nil {
+		// Once the client context is cancelled backoff.WithContext surfaces the
+		// bare context error, and any attempt torn down mid-flight reports the
+		// same. That cancellation is the caller asking us to stop (Stop, Down or
+		// an engine restart), so exit cleanly instead of handing back a failure
+		// the caller would have to distinguish from a real one.
+		if c.ctx.Err() != nil && errors.Is(err, context.Canceled) {
+			log.Info("exiting client retry loop, context cancelled")
+			return nil
+		}
+
 		log.Debugf("exiting client retry loop due to unrecoverable error: %s", err)
 		if s, ok := gstatus.FromError(err); ok && (s.Code() == codes.PermissionDenied) {
 			state.Set(StatusNeedsLogin)
@@ -682,7 +720,7 @@ func selectMTU(localMTU uint16, peerMTU int32) uint16 {
 }
 
 // connectToSignal creates Signal Service client and established a connection
-func connectToSignal(ctx context.Context, wtConfig *mgmProto.NetbirdConfig, ourPrivateKey wgtypes.Key) (*signal.GrpcClient, error) {
+func connectToSignal(ctx context.Context, wtConfig *mgmProto.NetbirdConfig, ourPrivateKey wgtypes.Key, netMgr *netevents.Manager) (*signal.GrpcClient, error) {
 	var sigTLSEnabled bool
 	if wtConfig.Signal.Protocol == mgmProto.HostConfig_HTTPS {
 		sigTLSEnabled = true
@@ -690,7 +728,8 @@ func connectToSignal(ctx context.Context, wtConfig *mgmProto.NetbirdConfig, ourP
 		sigTLSEnabled = false
 	}
 
-	signalClient, err := signal.NewClient(ctx, wtConfig.Signal.Uri, ourPrivateKey, sigTLSEnabled)
+	signalClient, err := signal.NewClient(ctx, wtConfig.Signal.Uri, ourPrivateKey, sigTLSEnabled,
+		signal.WithNetEvents(netMgr))
 	if err != nil {
 		log.Errorf("error while connecting to the Signal Exchange Service %s: %s", wtConfig.Signal.Uri, err)
 		return nil, gstatus.Errorf(codes.FailedPrecondition, "failed connecting to Signal Service : %s", err)

client/internal/daemonaddr/identity.go

@@ -0,0 +1,17 @@
+package daemonaddr
+
+import "strings"
+
+// CarriesIdentity reports whether the control channel at addr conveys the
+// connecting process's identity to the daemon. A Unix socket carries peer
+// credentials and a named pipe carries the client's token. Nothing else does, TCP
+// included, and there the daemon can authorize a privileged operation for nobody
+// at all: see ResolveDaemonAddr, which says as much to anyone still reaching the
+// Windows daemon on the address it served before it had a pipe.
+//
+// A client uses this to tell whether becoming privileged would get it anywhere.
+// It answers from the scheme and nothing else, so an address it does not
+// recognise counts as carrying no identity.
+func CarriesIdentity(addr string) bool {
+	return strings.HasPrefix(addr, "unix://") || strings.HasPrefix(addr, pipeScheme)
+}

client/internal/daemonaddr/identity_test.go

@@ -0,0 +1,29 @@
+package daemonaddr
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestCarriesIdentity(t *testing.T) {
+	tests := []struct {
+		addr string
+		want bool
+	}{
+		{"unix:///var/run/netbird.sock", true},
+		{"unix:///var/run/netbird/default.sock", true},
+		{"npipe://netbird", true},
+		{`npipe://\\.\pipe\ProtectedPrefix\Administrators\netbird`, true},
+		{"tcp://127.0.0.1:41731", false},
+		{"tcp://localhost:41731", false},
+		{"", false},
+		{"/var/run/netbird.sock", false},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.addr, func(t *testing.T) {
+			assert.Equal(t, tt.want, CarriesIdentity(tt.addr), "address %q", tt.addr)
+		})
+	}
+}

client/internal/dns/host_windows.go

@@ -35,6 +35,8 @@ var (
 // exported so a diagnostic reader reports the same locations that are written.
 const (
 	// NRPTKeyPrefix starts the name of every NRPT rule key this client creates.
+	// Older versions used different layouts under the same prefix: a single
+	// unsuffixed key, then one key per domain, now one key per batch of domains.
 	NRPTKeyPrefix = "NetBird-Match"
 
 	// DNSPolicyConfigRoot holds the NRPT rules of the local policy store.
@@ -89,7 +91,6 @@ type registryConfigurator struct {
 	guid            string
 	routingAll      bool
 	gpo             bool
-	nrptEntryCount  int
 	origNameservers []netip.Addr
 }
 
@@ -322,14 +323,9 @@ func (r *registryConfigurator) applyDNSConfig(config HostDNSConfig, stateManager
 	}
 
 	if len(matchDomains) != 0 {
-		count, err := r.addDNSMatchPolicy(matchDomains, config.ServerIP)
-		// Update count even on error to ensure cleanup covers partially created rules
-		r.nrptEntryCount = count
-		if err != nil {
+		if err := r.addDNSMatchPolicy(matchDomains, config.ServerIP); err != nil {
 			return fmt.Errorf("add dns match policy: %w", err)
 		}
-	} else {
-		r.nrptEntryCount = 0
 	}
 
 	r.updateState(stateManager)
@@ -345,9 +341,8 @@ func (r *registryConfigurator) applyDNSConfig(config HostDNSConfig, stateManager
 
 func (r *registryConfigurator) updateState(stateManager *statemanager.Manager) {
 	if err := stateManager.UpdateState(&ShutdownState{
-		Guid:           r.guid,
-		GPO:            r.gpo,
-		NRPTEntryCount: r.nrptEntryCount,
+		Guid: r.guid,
+		GPO:  r.gpo,
 	}); err != nil {
 		log.Errorf("failed to update shutdown state: %s", err)
 	}
@@ -362,7 +357,7 @@ func (r *registryConfigurator) addDNSSetupForAll(ip netip.Addr) error {
 	return nil
 }
 
-func (r *registryConfigurator) addDNSMatchPolicy(domains []string, ip netip.Addr) (int, error) {
+func (r *registryConfigurator) addDNSMatchPolicy(domains []string, ip netip.Addr) error {
 	// if the gpo key is present, we need to put our DNS settings there, otherwise our config might be ignored
 	// see https://learn.microsoft.com/en-us/openspecs/windows_protocols/ms-gpnrpt/8cc31cb9-20cb-4140-9e85-3e08703b4745
 
@@ -379,19 +374,17 @@ func (r *registryConfigurator) addDNSMatchPolicy(domains []string, ip netip.Addr
 		gpoPath := fmt.Sprintf("%s-%d", gpoDnsPolicyConfigMatchPath, ruleIndex)
 
 		if err := r.configureDNSPolicy(localPath, batchDomains, ip); err != nil {
-			return ruleIndex, fmt.Errorf("configure DNS Local policy for rule %d: %w", ruleIndex, err)
+			return fmt.Errorf("configure DNS Local policy for rule %d: %w", ruleIndex, err)
 		}
 
-		// Increment immediately so the caller's cleanup path knows about this rule
-		ruleIndex++
-
 		if r.gpo {
 			if err := r.configureDNSPolicy(gpoPath, batchDomains, ip); err != nil {
-				return ruleIndex, fmt.Errorf("configure gpo DNS policy for rule %d: %w", ruleIndex-1, err)
+				return fmt.Errorf("configure gpo DNS policy for rule %d: %w", ruleIndex, err)
 			}
 		}
 
-		log.Debugf("added NRPT rule %d with %d domains", ruleIndex-1, len(batchDomains))
+		log.Debugf("added NRPT rule %d with %d domains", ruleIndex, len(batchDomains))
+		ruleIndex++
 	}
 
 	if r.gpo {
@@ -401,7 +394,7 @@ func (r *registryConfigurator) addDNSMatchPolicy(domains []string, ip netip.Addr
 	}
 
 	log.Infof("added %d NRPT rules for %d domains", ruleIndex, len(domains))
-	return ruleIndex, nil
+	return nil
 }
 
 func (r *registryConfigurator) configureDNSPolicy(policyPath string, domains []string, ip netip.Addr) error {
@@ -466,7 +459,7 @@ func (r *registryConfigurator) flushDNSCache() {
 
 	ret, _, err := dnsFlushResolverCacheFn.Call()
 	if ret == 0 {
-		if err != nil && !errors.Is(err, syscall.Errno(0)) {
+		if !errors.Is(err, syscall.Errno(0)) {
 			log.Errorf("DnsFlushResolverCache failed: %v", err)
 			return
 		}
@@ -534,28 +527,28 @@ func (r *registryConfigurator) restoreHostDNS() error {
 	return nil
 }
 
+// removeDNSMatchPolicies deletes every NRPT rule this client may have created,
+// from the local and the GPO policy store. The rules are found by enumerating
+// the registry, the only authoritative record of what was written. Cleanup must
+// not depend on a rule count: the in-memory one is scoped to a single
+// registryConfigurator and the persisted one is deleted on every clean
+// disconnect, and a rule left behind keeps resolving names over an interface
+// that is gone, until reboot discards the volatile key.
 func (r *registryConfigurator) removeDNSMatchPolicies() error {
 	var merr *multierror.Error
 
-	// Try to remove the base entries (for backward compatibility)
-	if err := removeRegistryKeyFromDNSPolicyConfig(dnsPolicyConfigMatchPath); err != nil {
-		merr = multierror.Append(merr, fmt.Errorf("remove local base entry: %w", err))
-	}
-
-	if err := removeRegistryKeyFromDNSPolicyConfig(gpoDnsPolicyConfigMatchPath); err != nil {
-		merr = multierror.Append(merr, fmt.Errorf("remove GPO base entry: %w", err))
-	}
-
-	for i := 0; i < r.nrptEntryCount; i++ {
-		localPath := fmt.Sprintf("%s-%d", dnsPolicyConfigMatchPath, i)
-		gpoPath := fmt.Sprintf("%s-%d", gpoDnsPolicyConfigMatchPath, i)
-
-		if err := removeRegistryKeyFromDNSPolicyConfig(localPath); err != nil {
-			merr = multierror.Append(merr, fmt.Errorf("remove local entry %d: %w", i, err))
+	for _, root := range []string{DNSPolicyConfigRoot, GPODNSPolicyConfigRoot} {
+		names, err := listNRPTRuleKeys(root)
+		if err != nil {
+			merr = multierror.Append(merr, fmt.Errorf("list rule keys under %s: %w", root, err))
+			continue
 		}
 
-		if err := removeRegistryKeyFromDNSPolicyConfig(gpoPath); err != nil {
-			merr = multierror.Append(merr, fmt.Errorf("remove GPO entry %d: %w", i, err))
+		for _, name := range names {
+			path := root + `\` + name
+			if err := removeRegistryKeyFromDNSPolicyConfig(path); err != nil {
+				merr = multierror.Append(merr, fmt.Errorf("remove entry %s: %w", path, err))
+			}
 		}
 	}
 
@@ -570,6 +563,39 @@ func (r *registryConfigurator) restoreUncleanShutdownDNS() error {
 	return r.restoreHostDNS()
 }
 
+// listNRPTRuleKeys returns the names of our NRPT rule keys under a policy store
+// root. An absent root holds nothing to clean up, which is the normal state of
+// the GPO store on a machine without DNS Client policy.
+func listNRPTRuleKeys(root string) ([]string, error) {
+	k, err := registry.OpenKey(registry.LOCAL_MACHINE, root, registry.ENUMERATE_SUB_KEYS)
+	switch {
+	case errors.Is(err, registry.ErrNotExist), errors.Is(err, syscall.ERROR_PATH_NOT_FOUND):
+		// the GPO store is absent on a machine without DNS client policy
+		log.Debugf("HKEY_LOCAL_MACHINE\\%s does not exist", root)
+		return nil, nil
+	case err != nil:
+		// any other failure has to reach the caller: reporting no rules would
+		// report a successful cleanup while leaving the rules in place
+		return nil, fmt.Errorf("open HKEY_LOCAL_MACHINE\\%s: %w", root, err)
+	}
+	defer closer(k)
+
+	names, err := k.ReadSubKeyNames(-1)
+	if err != nil {
+		return nil, fmt.Errorf("read subkey names: %w", err)
+	}
+
+	var ruleKeys []string
+	for _, name := range names {
+		// registry key names are case insensitive
+		if strings.HasPrefix(strings.ToLower(name), strings.ToLower(NRPTKeyPrefix)) {
+			ruleKeys = append(ruleKeys, name)
+		}
+	}
+
+	return ruleKeys, nil
+}
+
 func removeRegistryKeyFromDNSPolicyConfig(regKeyPath string) error {
 	k, err := registry.OpenKey(registry.LOCAL_MACHINE, regKeyPath, registry.QUERY_VALUE)
 	if err != nil {
@@ -601,7 +627,7 @@ func refreshGroupPolicy() error {
 	)
 
 	if ret == 0 {
-		if err != nil && !errors.Is(err, syscall.Errno(0)) {
+		if !errors.Is(err, syscall.Errno(0)) {
 			return fmt.Errorf("RefreshPolicyEx failed: %w", err)
 		}
 		return fmt.Errorf("RefreshPolicyEx failed")

client/internal/dns/host_windows_test.go

@@ -25,7 +25,7 @@ func TestNRPTEntriesCleanupOnConfigChange(t *testing.T) {
 
 	// Create a test interface registry key so updateSearchDomains doesn't fail
 	testGUID := "{12345678-1234-1234-1234-123456789ABC}"
-	interfacePath := `SYSTEM\CurrentControlSet\Services\Tcpip\Parameters\Interfaces\` + testGUID
+	interfacePath := InterfaceConfigPath + `\` + testGUID
 	testKey, _, err := registry.CreateKey(registry.LOCAL_MACHINE, interfacePath, registry.SET_VALUE)
 	require.NoError(t, err, "Should create test interface registry key")
 	testKey.Close()
@@ -56,7 +56,7 @@ func TestNRPTEntriesCleanupOnConfigChange(t *testing.T) {
 	require.NoError(t, err)
 
 	// Verify 3 NRPT rules exist
-	assert.Equal(t, 3, cfg.nrptEntryCount, "Should create 3 NRPT rules for 125 domains")
+	assert.Equal(t, 3, countNRPTRuleKeys(t), "Should create 3 NRPT rules for 125 domains")
 	for i := 0; i < 3; i++ {
 		exists, err := registryKeyExists(fmt.Sprintf("%s-%d", dnsPolicyConfigMatchPath, i))
 		require.NoError(t, err)
@@ -81,7 +81,7 @@ func TestNRPTEntriesCleanupOnConfigChange(t *testing.T) {
 	require.NoError(t, err)
 
 	// Verify first 2 NRPT rules exist
-	assert.Equal(t, 2, cfg.nrptEntryCount, "Should create 2 NRPT rules for 75 domains")
+	assert.Equal(t, 2, countNRPTRuleKeys(t), "Should create 2 NRPT rules for 75 domains")
 	for i := 0; i < 2; i++ {
 		exists, err := registryKeyExists(fmt.Sprintf("%s-%d", dnsPolicyConfigMatchPath, i))
 		require.NoError(t, err)
@@ -106,9 +106,65 @@ func registryKeyExists(path string) (bool, error) {
 	return true, nil
 }
 
+// TestNRPTCleanupWithoutRuleCount verifies that rules written by a previous run
+// are removed by a configurator that has no record of how many there are: an
+// unclean exit loses the in-memory count and a clean disconnect deletes the
+// persisted one, so cleanup cannot depend on either.
+func TestNRPTCleanupWithoutRuleCount(t *testing.T) {
+	if testing.Short() {
+		t.Skip("skipping registry integration test in short mode")
+	}
+
+	defer cleanupRegistryKeys(t)
+	cleanupRegistryKeys(t)
+
+	testIP := netip.MustParseAddr("100.64.0.1")
+
+	// 75 domains produce two indexed rules, as the current layout does
+	domains := make([]string, 75)
+	for i := range domains {
+		domains[i] = fmt.Sprintf(".domain%d.com", i+1)
+	}
+
+	previousRun := &registryConfigurator{}
+	require.NoError(t, previousRun.addDNSMatchPolicy(domains, testIP))
+
+	// the unsuffixed key an older version would have written
+	require.NoError(t, previousRun.configureDNSPolicy(dnsPolicyConfigMatchPath, []string{".legacy.example.com"}, testIP))
+
+	// a policy owned by someone else, which cleanup must not touch
+	foreignPath := DNSPolicyConfigRoot + `\DnsPolicyConfigTestForeign`
+	foreignKey, _, err := registry.CreateKey(registry.LOCAL_MACHINE, foreignPath, registry.SET_VALUE)
+	require.NoError(t, err, "Should create foreign policy key")
+	foreignKey.Close()
+	defer func() {
+		_ = registry.DeleteKey(registry.LOCAL_MACHINE, foreignPath)
+	}()
+
+	require.Equal(t, 3, countNRPTRuleKeys(t), "Should have two indexed rules and the legacy one")
+
+	// a configurator that never applied a DNS config, as one built after a
+	// restart or from a shutdown state without a count is
+	freshRun := &registryConfigurator{}
+	require.NoError(t, freshRun.removeDNSMatchPolicies())
+
+	assert.Equal(t, 0, countNRPTRuleKeys(t), "Should remove every rule left by the previous run")
+
+	exists, err := registryKeyExists(foreignPath)
+	require.NoError(t, err)
+	assert.True(t, exists, "Should not remove a policy that is not ours")
+}
+
+func countNRPTRuleKeys(t *testing.T) int {
+	t.Helper()
+
+	names, err := listNRPTRuleKeys(DNSPolicyConfigRoot)
+	require.NoError(t, err, "Should list NRPT rule keys")
+	return len(names)
+}
+
 func cleanupRegistryKeys(*testing.T) {
-	// Clean up more entries to account for batching tests with many domains
-	cfg := &registryConfigurator{nrptEntryCount: 20}
+	cfg := &registryConfigurator{}
 	_ = cfg.removeDNSMatchPolicies()
 }
 
@@ -125,7 +181,7 @@ func TestNRPTDomainBatching(t *testing.T) {
 
 	// Create a test interface registry key so updateSearchDomains doesn't fail
 	testGUID := "{12345678-1234-1234-1234-123456789ABC}"
-	interfacePath := `SYSTEM\CurrentControlSet\Services\Tcpip\Parameters\Interfaces\` + testGUID
+	interfacePath := InterfaceConfigPath + `\` + testGUID
 	testKey, _, err := registry.CreateKey(registry.LOCAL_MACHINE, interfacePath, registry.SET_VALUE)
 	require.NoError(t, err, "Should create test interface registry key")
 	testKey.Close()
@@ -193,7 +249,7 @@ func TestNRPTDomainBatching(t *testing.T) {
 			require.NoError(t, err)
 
 			// Verify that exactly expectedRuleCount rules were created
-			assert.Equal(t, tc.expectedRuleCount, cfg.nrptEntryCount,
+			assert.Equal(t, tc.expectedRuleCount, countNRPTRuleKeys(t),
 				"Should create %d NRPT rules for %d domains", tc.expectedRuleCount, tc.domainCount)
 
 			// Verify all expected rules exist

client/internal/dns/mgmt/mgmt_refresh_test.go

@@ -224,6 +224,7 @@ func TestResolver_StaleTriggersAsyncRefresh(t *testing.T) {
 }
 
 func TestResolver_ConcurrentStaleHitsCollapseRefresh(t *testing.T) {
+	semaphore := make(chan struct{})
 	r := NewResolver()
 	chain := newFakeChain()
 	chain.setAnswer("mgmt.example.com.", dns.TypeA, "10.0.0.2")
@@ -239,7 +240,7 @@ func TestResolver_ConcurrentStaleHitsCollapseRefresh(t *testing.T) {
 				break
 			}
 		}
-		time.Sleep(50 * time.Millisecond) // hold inflight long enough to collide
+		<-semaphore // block the call to force request collision
 	}
 
 	r.SetChainResolver(chain, 50)
@@ -255,17 +256,17 @@ func TestResolver_ConcurrentStaleHitsCollapseRefresh(t *testing.T) {
 
 	var wg sync.WaitGroup
 	for i := 0; i < 50; i++ {
-		wg.Add(1)
-		go func() {
-			defer wg.Done()
+		wg.Go(func() {
 			queryA(t, r, "mgmt.example.com.")
-		}()
+		})
 	}
+
+	assert.Eventually(t, func() bool { return inflight.Load() >= 1 }, 2*time.Second, 100*time.Millisecond)
+
+	close(semaphore)
 	wg.Wait()
 
-	waitFor(t, 2*time.Second, func() bool {
-		return inflight.Load() == 0
-	})
+	assert.Eventually(t, func() bool { return inflight.Load() == 0 }, 2*time.Second, 100*time.Millisecond)
 
 	calls := chain.callCount("mgmt.example.com.", dns.TypeA)
 	assert.LessOrEqual(t, calls, 2, "singleflight must collapse concurrent refreshes (got %d)", calls)

client/internal/dns/response_writer_test.go

@@ -4,7 +4,7 @@ import (
 	"net"
 	"testing"
 
-	"github.com/golang/mock/gomock"
+	"go.uber.org/mock/gomock"
 	"github.com/google/gopacket"
 	"github.com/google/gopacket/layers"
 	"github.com/miekg/dns"

client/internal/dns/server_privileged_test.go

@@ -9,7 +9,7 @@ import (
 	"os"
 	"testing"
 
-	"github.com/golang/mock/gomock"
+	"go.uber.org/mock/gomock"
 	"github.com/miekg/dns"
 	"github.com/stretchr/testify/assert"
 	"golang.zx2c4.com/wireguard/wgctrl/wgtypes"

client/internal/dns/unclean_shutdown_windows.go

@@ -5,9 +5,8 @@ import (
 )
 
 type ShutdownState struct {
-	Guid           string
-	GPO            bool
-	NRPTEntryCount int
+	Guid string
+	GPO  bool
 }
 
 func (s *ShutdownState) Name() string {
@@ -16,9 +15,8 @@ func (s *ShutdownState) Name() string {
 
 func (s *ShutdownState) Cleanup() error {
 	manager := &registryConfigurator{
-		guid:           s.Guid,
-		gpo:            s.GPO,
-		nrptEntryCount: s.NRPTEntryCount,
+		guid: s.Guid,
+		gpo:  s.GPO,
 	}
 
 	if err := manager.restoreUncleanShutdownDNS(); err != nil {

client/internal/dnsfwd/manager.go

@@ -101,7 +101,7 @@ func (m *Manager) Start(fwdEntries []*ForwarderEntry) error {
 	m.dnsForwarder = NewDNSForwarder(listenAddress, dnsTTL, m.firewall, m.statusRecorder, m.wgIface)
 
 	go func() {
-		if err := m.dnsForwarder.Listen(fwdEntries); err != nil {
+		if err := m.dnsForwarder.Listen(fwdEntries); err != nil { //nolint:staticcheck
 			// todo handle close error if it is exists
 			log.Errorf("failed to start DNS forwarder, err: %v", err)
 		}

client/internal/elevate/elevate.go

@@ -0,0 +1,74 @@
+// Package elevate re-runs this very executable under the operating system's own
+// privilege-elevation mechanism and waits for it to finish.
+//
+// It exists so that a change the daemon restricts to root/administrator can be
+// authorized from the GUI, by the user, at the moment they ask for it: Windows
+// shows the UAC consent dialog, macOS the system authentication dialog, and
+// Linux/FreeBSD the session's polkit agent. The credentials, where any are
+// asked for, are collected by the operating system and never pass through
+// NetBird.
+//
+// What the elevated process then does is the caller's business: it is the same
+// binary, in a one-shot mode, and it is authorized by the daemon exactly like
+// any other privileged caller, from the identity the kernel reports on the
+// control channel. Nothing here grants privilege, and the daemon gains no new
+// way to be talked into something: elevation only changes who is calling it.
+package elevate
+
+import (
+	"context"
+	"errors"
+
+	log "github.com/sirupsen/logrus"
+)
+
+// AppliedMarker is what the elevated process prints on standard output once it has
+// done what it was run for.
+//
+// macOS's AuthorizationExecuteWithPrivileges reports no exit status and does not
+// say which process it started, so there this line is the only evidence that the
+// change was applied. The other platforms have an exit code and ignore it.
+const AppliedMarker = "netbird-elevated: applied"
+
+var (
+	// ErrDeclined reports that the user dismissed the prompt or did not
+	// authenticate. Nothing happened and nothing is wrong: a caller undoes its
+	// optimistic update and stays quiet.
+	ErrDeclined = errors.New("authorization declined")
+
+	// ErrUnavailable reports that this host has no elevation mechanism we can
+	// drive: no polkit on a Unix desktop, or an executable we decline to run as
+	// root. A caller falls back to telling the user which command to run.
+	ErrUnavailable = errors.New("no privilege elevation mechanism available")
+)
+
+// Run runs this executable with args under the platform's elevation mechanism
+// and waits for it to exit. A non-zero exit is returned as an error, so the
+// caller can treat a completed Run as the operation having succeeded.
+//
+// The args are the caller's own command line, so they cross no privilege
+// boundary: only a user who has just authenticated as an administrator can get
+// them run at all.
+func Run(ctx context.Context, args ...string) error {
+	self, err := trustedSelf()
+	if err != nil {
+		return err
+	}
+	return run(ctx, self, args)
+}
+
+// Available reports whether Run has a mechanism to use on this host, so a caller
+// can offer the prompt only when there is one and otherwise fall back to
+// guidance the user can act on. It answers from what is installed, not from what
+// the user is allowed to do: an administrator's password may still be required
+// and may still not be given, which is ErrDeclined from Run.
+func Available() bool {
+	if _, err := trustedSelf(); err != nil {
+		// Worth a line: this is also what a build run from a group-writable
+		// directory hits, and there is nothing in the UI to say why the offer is
+		// missing.
+		log.Debugf("not offering privilege elevation: %v", err)
+		return false
+	}
+	return mechanismAvailable()
+}

client/internal/elevate/output.go

@@ -0,0 +1,18 @@
+package elevate
+
+import "strings"
+
+// noOutput stands in for a process that said nothing, so that a report of what it
+// said still reads as a sentence.
+const noOutput = "no output"
+
+func firstLine(s string) string {
+	s = strings.TrimSpace(s)
+	if s == "" {
+		return noOutput
+	}
+	if i := strings.IndexByte(s, '\n'); i >= 0 {
+		return s[:i]
+	}
+	return s
+}

client/internal/elevate/output_test.go

@@ -0,0 +1,21 @@
+package elevate
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestFirstLine(t *testing.T) {
+	tests := []struct{ in, want string }{
+		{in: "", want: noOutput},
+		{in: "  \n ", want: noOutput},
+		{in: "one line", want: "one line"},
+		{in: "first\nsecond", want: "first"},
+		{in: "\nsecond\n", want: "second"},
+	}
+
+	for _, tt := range tests {
+		assert.Equal(t, tt.want, firstLine(tt.in), "input %q", tt.in)
+	}
+}

client/internal/elevate/run_darwin.go

@@ -0,0 +1,359 @@
+package elevate
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"os"
+	"runtime"
+	"strings"
+	"sync"
+	"syscall"
+	"unsafe"
+
+	"github.com/ebitengine/purego"
+	log "github.com/sirupsen/logrus"
+)
+
+// Authorization Services, reached through purego rather than cgo so the released
+// binaries keep building with CGO_ENABLED=0.
+//
+// The prompt belongs to this process, which is what makes it carry the
+// application's name and our own explanation. Going through osascript instead puts
+// the very same trampoline behind a dialog attributed to osascript, and means
+// handing a shell a command line to re-parse.
+//
+// # On AuthorizationExecuteWithPrivileges
+//
+// It is deprecated, and Apple's guidance (Quinn, "BSD Privilege Escalation on
+// macOS", developer.apple.com/forums/thread/708765) is "while it still works, it's
+// been deprecated for many years. Do not use it in a widely distributed product."
+// It is used here anyway, knowingly, because the alternatives Apple offers are for
+// *obtaining* ongoing privileges — an installer package, SMAppService, SMJobBless —
+// and NetBird already has what they would install: a launchd daemon running as
+// root. What is missing is only a way for an unprivileged client to ask it to act.
+//
+// The way to that without a deprecated call is to authorize the client instead of
+// elevating one: the app takes the right with AuthorizationCreate, passes the
+// AuthorizationExternalForm to the daemon, and the daemon checks it with
+// AuthorizationCopyRights before acting — none of which is deprecated. It is the
+// better design and it is where this should end up. It also means the daemon
+// accepting an authorization over its control socket, which is a new way to be
+// asked for privileged work and wants reviewing as such, so it is deliberately not
+// bundled in with the rest of this.
+//
+// Until then, three things keep the deprecation from being a trap. Every symbol is
+// resolved with an error rather than a panic, so a macOS that has dropped this
+// function leaves the app offering the user a command instead of crashing on the
+// way to a prompt. A failure to run the tool is reported as ErrUnavailable, so the
+// fallback is the same one an agent-less Linux session gets. And the whole path
+// runs under guard, which turns a panic out of the FFI layer into that same
+// fallback.
+//
+// The trampoline passes on the environment it was given, so what it starts as root
+// must be an executable this user's peers cannot influence: that is what
+// trustedSelf refuses, and what signing the binary settles for the loader.
+
+const (
+	securityFramework = "/System/Library/Frameworks/Security.framework/Security"
+	libSystem         = "/usr/lib/libSystem.B.dylib"
+
+	// trampoline is what the framework hands the tool to. Present on every macOS,
+	// and worth confirming before offering a prompt rather than mid-prompt.
+	trampoline = "/usr/libexec/security_authtrampoline"
+)
+
+// rightExecute is the right an administrator holds, and what
+// AuthorizationExecuteWithPrivileges requires of us.
+const rightExecute = "system.privilege.admin"
+
+// promptKey is kAuthorizationEnvironmentPrompt, which puts a sentence of ours above
+// the system's in the dialog. It is about the change rather than the mechanism.
+const (
+	promptKey  = "prompt"
+	promptText = "NetBird needs to change a setting that grants SSH access to this computer."
+)
+
+// OSStatus values from SecBase.h that mean something to us; anything else is
+// reported as it comes.
+const (
+	errAuthorizationSuccess               = 0
+	errAuthorizationDenied                = -60005
+	errAuthorizationCanceled              = -60006
+	errAuthorizationInteractionNotAllowed = -60007
+	errAuthorizationToolExecuteFailure    = -60031
+	errAuthorizationToolEnvironmentError  = -60032
+)
+
+// AuthorizationFlags from Authorization.h.
+const (
+	flagDefaults           = 0
+	flagInteractionAllowed = 1 << 0
+	flagExtendRights       = 1 << 1
+	flagDestroyRights      = 1 << 3
+	flagPreAuthorize       = 1 << 4
+)
+
+// authorizationItem mirrors AuthorizationItem: a name, and a value the name gives
+// meaning to. 32 bytes on both amd64 and arm64.
+type authorizationItem struct {
+	name        *byte
+	valueLength uintptr
+	value       unsafe.Pointer
+	// flags is reserved by the API and always zero. Declared because the layout
+	// is the contract: without it the struct is 24 bytes where C reads 32.
+	flags uint32 //nolint:unused // part of the C layout
+}
+
+// authorizationItemSet mirrors AuthorizationItemSet, which serves as both an
+// AuthorizationRights and an AuthorizationEnvironment.
+type authorizationItemSet struct {
+	count uint32
+	items *authorizationItem
+}
+
+var (
+	authorizationCreate                func(rights, environment *authorizationItemSet, flags uint32, authorization *uintptr) int32
+	authorizationExecuteWithPrivileges func(authorization uintptr, pathToTool string, options uint32, arguments *uintptr, communicationsPipe *uintptr) int32
+	authorizationFree                  func(authorization uintptr, flags uint32) int32
+	fileno                             func(stream uintptr) int32
+	fclose                             func(stream uintptr) int32
+
+	loadOnce sync.Once
+	loadErr  error
+)
+
+// load resolves the functions once. A framework that cannot be opened, or a symbol
+// that is no longer there, leaves the host without a mechanism rather than taking
+// the process down with it: see the note on deprecation above.
+func load() error {
+	loadOnce.Do(func() { loadErr = guard("loading Security.framework", resolve) })
+	return loadErr
+}
+
+// guard turns a panic out of the FFI layer into an error, so an API that has
+// changed under us costs the user a prompt rather than the window they were
+// clicking in. purego panics on a signature it cannot map, and this is the one
+// place in the client that calls a deprecated system function.
+//
+// It catches Go panics, which is what purego raises. A fault inside the framework
+// itself is not a panic and not recoverable; the layout the tests pin down is what
+// stands between us and that.
+func guard(what string, fn func() error) (err error) {
+	defer func() {
+		r := recover()
+		if r == nil {
+			return
+		}
+		log.Errorf("%s panicked: %v", what, r)
+		err = fmt.Errorf("%w: %s: %v", ErrUnavailable, what, r)
+	}()
+	return fn()
+}
+
+func resolve() error {
+	security, err := purego.Dlopen(securityFramework, purego.RTLD_LAZY|purego.RTLD_GLOBAL)
+	if err != nil {
+		return fmt.Errorf("open %s: %w", securityFramework, err)
+	}
+	system, err := purego.Dlopen(libSystem, purego.RTLD_LAZY|purego.RTLD_GLOBAL)
+	if err != nil {
+		return fmt.Errorf("open %s: %w", libSystem, err)
+	}
+
+	// purego.RegisterLibFunc panics on a symbol it cannot find, which is not how a
+	// deprecated function's disappearance should reach the user.
+	for _, fn := range []struct {
+		ptr    any
+		handle uintptr
+		name   string
+	}{
+		{&authorizationCreate, security, "AuthorizationCreate"},
+		{&authorizationExecuteWithPrivileges, security, "AuthorizationExecuteWithPrivileges"},
+		{&authorizationFree, security, "AuthorizationFree"},
+		{&fileno, system, "fileno"},
+		{&fclose, system, "fclose"},
+	} {
+		symbol, err := purego.Dlsym(fn.handle, fn.name)
+		if err != nil {
+			return fmt.Errorf("resolve %s: %w", fn.name, err)
+		}
+		if symbol == 0 {
+			return fmt.Errorf("resolve %s: not present on this system", fn.name)
+		}
+		purego.RegisterFunc(fn.ptr, symbol)
+	}
+	return nil
+}
+
+// run asks the system to run self as root: first for the right, which is what puts
+// up the authentication dialog and collects the password or takes the Touch ID,
+// then for the tool. The credentials go to the system's authorization trampoline
+// and never to us.
+//
+// The context bounds only our own waiting; the dialog belongs to the system and
+// closes when the user answers it.
+func run(ctx context.Context, self string, args []string) error {
+	if err := load(); err != nil {
+		return fmt.Errorf("%w: %v", ErrUnavailable, err)
+	}
+
+	return guard("asking for privileges", func() error {
+		authorization, err := authorize()
+		if err != nil {
+			return err
+		}
+		defer authorizationFree(authorization, flagDestroyRights)
+
+		return execute(ctx, authorization, self, args)
+	})
+}
+
+func mechanismAvailable() bool {
+	if err := load(); err != nil {
+		return false
+	}
+	info, err := os.Stat(trampoline)
+	return err == nil && !info.IsDir()
+}
+
+// authorize obtains the right, prompting for it. A dismissed dialog comes back as
+// errAuthorizationCanceled and a password given up on as errAuthorizationDenied;
+// both are the user's answer rather than a failure.
+func authorize() (uintptr, error) {
+	var pinner runtime.Pinner
+	defer pinner.Unpin()
+
+	rights := itemSet(&pinner, authorizationItem{name: cString(&pinner, rightExecute)})
+	environment := itemSet(&pinner, promptItem(&pinner))
+
+	var authorization uintptr
+	status := authorizationCreate(rights, environment,
+		flagDefaults|flagInteractionAllowed|flagPreAuthorize|flagExtendRights, &authorization)
+
+	switch status {
+	case errAuthorizationSuccess:
+		return authorization, nil
+	case errAuthorizationCanceled, errAuthorizationDenied:
+		return 0, ErrDeclined
+	case errAuthorizationInteractionNotAllowed:
+		// Nowhere to put a dialog, so there is nobody to ask: a launch daemon, or
+		// a session with no window server.
+		return 0, fmt.Errorf("%w: this session cannot show an authorization prompt", ErrUnavailable)
+	default:
+		return 0, fmt.Errorf("request %s: OSStatus %d", rightExecute, status)
+	}
+}
+
+// execute runs the tool with the right in hand and waits for it by reading the pipe
+// it is given until the tool closes it.
+//
+// AuthorizationExecuteWithPrivileges reports no exit status and does not say what
+// process it started, which is why the one-shot says so itself: what it prints is
+// the only evidence that the change was applied.
+func execute(ctx context.Context, authorization uintptr, self string, args []string) error {
+	var pinner runtime.Pinner
+	defer pinner.Unpin()
+
+	argv := make([]uintptr, 0, len(args)+1)
+	for _, arg := range args {
+		argv = append(argv, uintptr(unsafe.Pointer(cString(&pinner, arg))))
+	}
+	argv = append(argv, 0)
+	pinner.Pin(&argv[0])
+
+	var pipe uintptr
+	status := authorizationExecuteWithPrivileges(authorization, self, flagDefaults, &argv[0], &pipe)
+	switch status {
+	case errAuthorizationSuccess:
+	case errAuthorizationCanceled:
+		return ErrDeclined
+	case errAuthorizationToolExecuteFailure, errAuthorizationToolEnvironmentError:
+		// The right was granted and the tool still did not start. Nothing the user
+		// can do about it from here, so point them at the command instead.
+		return fmt.Errorf("%w: the system would not run %s elevated (OSStatus %d)", ErrUnavailable, self, status)
+	default:
+		return fmt.Errorf("run %s elevated: OSStatus %d", self, status)
+	}
+
+	out, err := readPipe(ctx, pipe)
+	if err != nil {
+		return err
+	}
+	return checkApplied(out)
+}
+
+// checkApplied reads the one-shot's report, which stands in for the exit status
+// there is no way to ask for here. A run that said nothing did not apply the
+// change, whatever else went on.
+func checkApplied(out string) error {
+	if !strings.Contains(out, AppliedMarker) {
+		return fmt.Errorf("elevated netbird did not report the change as applied: %s", firstLine(out))
+	}
+	return nil
+}
+
+// readPipe drains the tool's output, which ends when the tool exits and is
+// therefore also how we wait for it.
+func readPipe(ctx context.Context, pipe uintptr) (string, error) {
+	if pipe == 0 {
+		return "", nil
+	}
+	defer fclose(pipe)
+
+	fd := int(fileno(pipe))
+	if fd < 0 {
+		return "", nil
+	}
+
+	var out strings.Builder
+	buf := make([]byte, 4096)
+	for {
+		if err := ctx.Err(); err != nil {
+			return out.String(), err
+		}
+		n, err := syscall.Read(fd, buf)
+		if n > 0 {
+			out.Write(buf[:n])
+		}
+		switch {
+		case errors.Is(err, syscall.EINTR):
+			// A signal landed mid-read, which says nothing about the tool.
+			continue
+		case err != nil:
+			log.Debugf("read the elevated process's output: %v", err)
+			return out.String(), nil
+		case n <= 0:
+			// End of file: the tool closed the pipe, which is how it exiting
+			// reaches us.
+			return out.String(), nil
+		}
+	}
+}
+
+// itemSet builds an AuthorizationItemSet over items, pinned for the call.
+func itemSet(pinner *runtime.Pinner, items ...authorizationItem) *authorizationItemSet {
+	pinner.Pin(&items[0])
+	set := &authorizationItemSet{count: uint32(len(items)), items: &items[0]}
+	pinner.Pin(set)
+	return set
+}
+
+// promptItem is the environment entry carrying our sentence for the dialog.
+func promptItem(pinner *runtime.Pinner) authorizationItem {
+	value := []byte(promptText)
+	pinner.Pin(&value[0])
+	return authorizationItem{
+		name:        cString(pinner, promptKey),
+		valueLength: uintptr(len(value)),
+		value:       unsafe.Pointer(&value[0]),
+	}
+}
+
+// cString returns a NUL-terminated copy of s, pinned so the C side may hold it for
+// the duration of the call.
+func cString(pinner *runtime.Pinner, s string) *byte {
+	b := append([]byte(s), 0)
+	pinner.Pin(&b[0])
+	return &b[0]
+}

client/internal/elevate/run_darwin_test.go

@@ -0,0 +1,111 @@
+package elevate
+
+import (
+	"errors"
+	"runtime"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// The framework has to load and the symbols have to resolve, or nothing else here
+// means anything.
+func TestSecurityFrameworkLoads(t *testing.T) {
+	require.NoError(t, load(), "Security.framework must open")
+
+	for name, fn := range map[string]any{
+		"AuthorizationCreate":                authorizationCreate,
+		"AuthorizationExecuteWithPrivileges": authorizationExecuteWithPrivileges,
+		"AuthorizationFree":                  authorizationFree,
+		"fileno":                             fileno,
+		"fclose":                             fclose,
+	} {
+		assert.NotNil(t, fn, "%s must resolve", name)
+	}
+}
+
+// A request with no interaction allowed exercises the whole call — the rights and
+// environment structs, and the OSStatus that comes back — without a dialog anybody
+// has to answer. What the system decides is its business; that it decides at all is
+// what this asserts.
+func TestAuthorizationCreateWithoutInteraction(t *testing.T) {
+	if err := load(); err != nil {
+		t.Skipf("Security.framework did not open: %v", err)
+	}
+
+	var pinner runtime.Pinner
+	defer pinner.Unpin()
+
+	rights := itemSet(&pinner, authorizationItem{name: cString(&pinner, rightExecute)})
+	environment := itemSet(&pinner, promptItem(&pinner))
+	require.EqualValues(t, 1, rights.count, "the rights struct layout must match the C one")
+
+	var authorization uintptr
+	status := authorizationCreate(rights, environment, flagDefaults|flagExtendRights, &authorization)
+
+	switch status {
+	case errAuthorizationSuccess:
+		// Credentials were already cached for this session.
+		authorizationFree(authorization, flagDestroyRights)
+	case errAuthorizationDenied, errAuthorizationInteractionNotAllowed:
+		// The expected answers when nobody may be asked.
+	default:
+		require.Failf(t, "unknown OSStatus", "AuthorizationCreate returned %d, want a status we recognise", status)
+	}
+}
+
+// Asking with a right nobody has must not be mistaken for a declined prompt: the
+// caller would report nothing at all.
+func TestAuthorizeUnknownRightIsNotDeclined(t *testing.T) {
+	if err := load(); err != nil {
+		t.Skipf("Security.framework did not open: %v", err)
+	}
+
+	var pinner runtime.Pinner
+	defer pinner.Unpin()
+
+	rights := itemSet(&pinner, authorizationItem{name: cString(&pinner, "io.netbird.right.that.does.not.exist")})
+
+	var authorization uintptr
+	status := authorizationCreate(rights, nil, flagDefaults|flagExtendRights, &authorization)
+	if status == errAuthorizationSuccess {
+		authorizationFree(authorization, flagDestroyRights)
+	}
+	assert.NotEqual(t, int32(errAuthorizationSuccess), status, "a right that does not exist must not be granted")
+}
+
+func TestMechanismAvailable(t *testing.T) {
+	assert.True(t, mechanismAvailable(), "the trampoline exists on every macOS")
+}
+
+// The one-shot's report is what stands in for an exit status here, so a run that
+// says nothing must not read as success.
+func TestCheckApplied(t *testing.T) {
+	require.NoError(t, checkApplied(AppliedMarker+"\n"), "the report the one-shot prints")
+	require.NoError(t, checkApplied("some warning\n"+AppliedMarker+"\n"), "the report after other output")
+
+	assert.Error(t, checkApplied(""), "a run that printed nothing did not apply the change")
+	assert.Error(t, checkApplied("dyld: library not loaded\n"), "output that is not the report")
+}
+
+// A panic out of the FFI layer has to reach the caller as "no mechanism", which is
+// the outcome that offers the user the command instead of taking the window down.
+func TestGuardTurnsAPanicIntoUnavailable(t *testing.T) {
+	err := guard("pretending to call something", func() error {
+		panic("purego: signature it cannot map")
+	})
+
+	require.ErrorIs(t, err, ErrUnavailable, "a panic must read as a missing mechanism")
+	assert.Contains(t, err.Error(), "pretending to call something", "what panicked")
+}
+
+// guard wraps every darwin path, so what a caller switches on has to survive it.
+func TestGuardPassesErrorsThrough(t *testing.T) {
+	sentinel := errors.New("the call itself failed")
+	assert.ErrorIs(t, guard("calling", func() error { return sentinel }), sentinel,
+		"the error it was given")
+	assert.ErrorIs(t, guard("calling", func() error { return ErrDeclined }), ErrDeclined,
+		"a declined prompt stays declined")
+	assert.NoError(t, guard("calling", func() error { return nil }), "a call that worked")
+}

client/internal/elevate/run_unix.go

@@ -0,0 +1,117 @@
+//go:build linux
+
+package elevate
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"io"
+	"os"
+	"os/exec"
+	"strings"
+)
+
+// pkexec exit codes that are about the authorization rather than about the program
+// we asked it to run. The manual page reserves both.
+const (
+	// exitDismissed is returned when the user dismissed the authentication
+	// dialog.
+	exitDismissed = 126
+	// exitNotAuthorized is returned when the authorization was not obtained. That
+	// covers the user saying no as well as pkexec having had nobody to ask: see
+	// noAgentMarkers.
+	exitNotAuthorized = 127
+)
+
+// exitNotAuthorized covers three different endings that only pkexec's own words
+// tell apart, so they are matched here. Read with LC_ALL=C so the words are the
+// ones written below.
+//
+// refusedMarker is a refusal: the user said no, gave up on the password, or holds
+// an account that may not elevate at all.
+const refusedMarker = "Not authorized"
+
+// noAgentMarkers say pkexec had no way to ask: no agent registered for the
+// session, and no controlling terminal for the textual agent it falls back to.
+var noAgentMarkers = []string{"authentication agent", "controlling terminal"}
+
+// run asks polkit to run self as root. pkexec hands the request to the session's
+// polkit agent, which is what prompts and what collects any password; we see only
+// its verdict.
+//
+// The environment is otherwise deliberately not passed through: pkexec clears it
+// bar a small allowlist, and the one-shot needs nothing from it.
+func run(ctx context.Context, self string, args []string) error {
+	pkexec, err := exec.LookPath("pkexec")
+	if err != nil {
+		return fmt.Errorf("%w: pkexec is not installed", ErrUnavailable)
+	}
+
+	cmd := exec.CommandContext(ctx, pkexec, append([]string{self}, args...)...)
+	// C locale so pkexec's own diagnostics are the ones noAgentMarkers knows.
+	cmd.Env = append(os.Environ(), "LC_ALL=C")
+	var stderr strings.Builder
+	cmd.Stderr = &stderr
+	// The one-shot reports itself on stdout for macOS's sake, where there is no
+	// exit status to read. Here there is one, so that line is noise.
+	cmd.Stdout = io.Discard
+
+	err = cmd.Run()
+	if err == nil {
+		return nil
+	}
+
+	var exitErr *exec.ExitError
+	if !errors.As(err, &exitErr) {
+		return fmt.Errorf("run pkexec: %w", err)
+	}
+
+	// Matched against everything pkexec said, reported as one line: a complaint
+	// that is not the first thing printed still has to be recognised, and reading
+	// it as a refusal would swallow it.
+	full := stderr.String()
+	out := firstLine(full)
+
+	switch exitErr.ExitCode() {
+	case exitDismissed:
+		return ErrDeclined
+	case exitNotAuthorized:
+		return notAuthorized(full, out)
+	default:
+		return fmt.Errorf("elevated netbird exited with %d: %s", exitErr.ExitCode(), out)
+	}
+}
+
+// notAuthorized sorts out the three endings pkexec reports as exitNotAuthorized.
+//
+// It also returns that code when the authorization succeeded and it then could
+// not run the program, so a refusal has to be recognised rather than assumed:
+// reading every one of these as "the user said no" would revert the control in
+// silence on a host where elevation is broken.
+func notAuthorized(full, out string) error {
+	switch {
+	case hasAny(full, noAgentMarkers):
+		return fmt.Errorf("%w: polkit had no way to ask: %s", ErrUnavailable, out)
+	case out == noOutput, strings.Contains(full, refusedMarker):
+		// The user said no, which needs no message; that an account barred from
+		// elevating altogether lands here too is why the reason is kept.
+		return fmt.Errorf("%w: %s", ErrDeclined, out)
+	default:
+		return fmt.Errorf("pkexec could not run elevated netbird: %s", out)
+	}
+}
+
+func hasAny(s string, markers []string) bool {
+	for _, marker := range markers {
+		if strings.Contains(s, marker) {
+			return true
+		}
+	}
+	return false
+}
+
+func mechanismAvailable() bool {
+	_, err := exec.LookPath("pkexec")
+	return err == nil
+}

client/internal/elevate/run_unix_test.go

@@ -0,0 +1,110 @@
+//go:build linux
+
+package elevate
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// fakePkexec puts a pkexec on PATH that exits with the given code, so the
+// mapping from polkit's exit codes onto our errors can be exercised without a
+// polkit agent.
+func fakePkexec(t *testing.T, exitCode int, stderr string) {
+	t.Helper()
+
+	dir := t.TempDir()
+	script := fmt.Sprintf("#!/bin/sh\necho %s >&2\nexit %d\n", shellQuote(stderr), exitCode)
+	require.NoError(t, os.WriteFile(filepath.Join(dir, "pkexec"), []byte(script), 0o700), "write the fake pkexec")
+	t.Setenv("PATH", dir)
+}
+
+func shellQuote(s string) string {
+	return "'" + strings.ReplaceAll(s, "'", `'\''`) + "'"
+}
+
+func TestRunMapsPkexecExitCodes(t *testing.T) {
+	tests := []struct {
+		name     string
+		exitCode int
+		stderr   string
+		wantErr  error
+	}{
+		{name: "applied", exitCode: 0},
+		{
+			name:     "dialog dismissed",
+			exitCode: exitDismissed,
+			stderr:   "Error executing command as another user: Request dismissed",
+			wantErr:  ErrDeclined,
+		},
+		{
+			// What a graphical agent reports for a cancelled prompt. Not a
+			// failure: the user was asked and answered.
+			name:     "prompt cancelled",
+			exitCode: exitNotAuthorized,
+			stderr:   "Error executing command as another user: Not authorized",
+			wantErr:  ErrDeclined,
+		},
+		{
+			// The same status, but pkexec never got to ask anybody.
+			name:     "no agent and no terminal to fall back on",
+			exitCode: exitNotAuthorized,
+			stderr:   "Error creating textual authentication agent: Error opening current controlling terminal for the process (`/dev/tty'): No such device or address",
+			wantErr:  ErrUnavailable,
+		},
+		{
+			// And the same status again once the authorization succeeded and
+			// pkexec could not run what it had been authorized to run. Reading
+			// that as a refusal would revert the control in silence on a host
+			// where elevation is broken.
+			name:     "authorized but not runnable",
+			exitCode: exitNotAuthorized,
+			stderr:   "Error executing command as another user: No such file or directory",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			fakePkexec(t, tt.exitCode, tt.stderr)
+
+			err := run(context.Background(), "/nonexistent/netbird-ui", []string{"--flag"})
+			switch {
+			case tt.wantErr != nil:
+				require.ErrorIs(t, err, tt.wantErr, "exit %d said %q", tt.exitCode, tt.stderr)
+			case tt.exitCode == 0:
+				require.NoError(t, err, "a pkexec that exited cleanly applied the change")
+			default:
+				require.Error(t, err, "exit %d said %q", tt.exitCode, tt.stderr)
+				assert.NotErrorIs(t, err, ErrDeclined, "not the user's answer")
+				assert.NotErrorIs(t, err, ErrUnavailable, "not a missing mechanism")
+			}
+		})
+	}
+}
+
+// An exit code that is not polkit's is the one-shot's own failure, and has to
+// stay distinguishable from a declined prompt: the caller reports it.
+func TestRunReportsOneShotFailure(t *testing.T) {
+	fakePkexec(t, 3, "the one-shot said no")
+
+	err := run(context.Background(), "/nonexistent/netbird-ui", nil)
+
+	require.Error(t, err, "a one-shot that failed is not a prompt that was answered")
+	assert.NotErrorIs(t, err, ErrDeclined, "not the user's answer")
+	assert.NotErrorIs(t, err, ErrUnavailable, "not a missing mechanism")
+}
+
+func TestRunWithoutPkexecIsUnavailable(t *testing.T) {
+	t.Setenv("PATH", t.TempDir())
+
+	err := run(context.Background(), "/nonexistent/netbird-ui", nil)
+	require.ErrorIs(t, err, ErrUnavailable, "no pkexec means no mechanism")
+	assert.False(t, mechanismAvailable(), "mechanismAvailable without pkexec on PATH")
+}

client/internal/elevate/run_unsupported.go

@@ -0,0 +1,19 @@
+//go:build !windows && !darwin && !linux
+
+package elevate
+
+import "context"
+
+// run reports that this platform has no elevation prompt to drive.
+//
+// The desktop app is the only caller and is not built for any of these: mobile
+// and WASM have no local user to ask, and the FreeBSD client ships without a UI.
+// pkexec would be the mechanism there, and run_unix.go is what to widen if that
+// changes.
+func run(context.Context, string, []string) error {
+	return ErrUnavailable
+}
+
+func mechanismAvailable() bool {
+	return false
+}

client/internal/elevate/run_windows.go

@@ -0,0 +1,187 @@
+package elevate
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"runtime"
+	"unsafe"
+
+	log "github.com/sirupsen/logrus"
+	"golang.org/x/sys/windows"
+)
+
+const (
+	// seeMaskNoCloseProcess keeps the started process's handle open in
+	// hProcess so we can wait for it.
+	seeMaskNoCloseProcess = 0x00000040
+	// seeMaskNoAsync makes ShellExecuteExW finish its work before returning,
+	// which it must when the calling thread does not pump messages.
+	seeMaskNoAsync = 0x00000100
+	// seeMaskFlagNoUI suppresses the shell's own error dialogs; the UAC consent
+	// dialog is not one of them and still appears.
+	seeMaskFlagNoUI = 0x00000400
+
+	// swHide: the one-shot has no window to show.
+	swHide = 0
+)
+
+// shellExecuteInfoW mirrors SHELLEXECUTEINFOW. The field order and Go's own
+// padding match the C layout on both 386 and amd64.
+type shellExecuteInfoW struct {
+	cbSize         uint32
+	fMask          uint32
+	hwnd           windows.HWND
+	lpVerb         *uint16
+	lpFile         *uint16
+	lpParameters   *uint16
+	lpDirectory    *uint16
+	nShow          int32
+	hInstApp       windows.Handle
+	lpIDList       uintptr
+	lpClass        *uint16
+	hkeyClass      windows.Handle
+	dwHotKey       uint32
+	hIconOrMonitor windows.Handle
+	hProcess       windows.Handle
+}
+
+var (
+	shell32            = windows.NewLazySystemDLL("shell32.dll")
+	procShellExecuteEx = shell32.NewProc("ShellExecuteExW")
+)
+
+// run starts self elevated with the "runas" verb, which is what raises the UAC
+// consent dialog, and waits for it to finish. Windows decides whether consent is
+// enough or an administrator's credentials are needed, and collects them itself.
+func run(ctx context.Context, self string, args []string) error {
+	verb, err := windows.UTF16PtrFromString("runas")
+	if err != nil {
+		return fmt.Errorf("encode verb: %w", err)
+	}
+	file, err := windows.UTF16PtrFromString(self)
+	if err != nil {
+		return fmt.Errorf("encode %s: %w", self, err)
+	}
+	params, err := windows.UTF16PtrFromString(windows.ComposeCommandLine(args))
+	if err != nil {
+		return fmt.Errorf("encode arguments: %w", err)
+	}
+
+	info := shellExecuteInfoW{
+		fMask:        seeMaskNoCloseProcess | seeMaskNoAsync | seeMaskFlagNoUI,
+		hwnd:         ownerWindow(),
+		lpVerb:       verb,
+		lpFile:       file,
+		lpParameters: params,
+		nShow:        swHide,
+	}
+	info.cbSize = uint32(unsafe.Sizeof(info))
+
+	process, err := shellExecute(&info)
+	if err != nil {
+		return err
+	}
+	defer func() {
+		if err := windows.CloseHandle(process); err != nil {
+			log.Debugf("close elevated process handle: %v", err)
+		}
+	}()
+
+	return waitForProcess(ctx, process)
+}
+
+// shellExecute performs the call itself. ShellExecuteExW wants COM initialised on
+// the calling thread, so the goroutine is pinned to one for the duration and COM
+// is set up on it; an "already initialised, different mode" answer is fine,
+// because then somebody else has done it for us.
+func shellExecute(info *shellExecuteInfoW) (windows.Handle, error) {
+	runtime.LockOSThread()
+	defer runtime.UnlockOSThread()
+
+	switch err := windows.CoInitializeEx(0, windows.COINIT_APARTMENTTHREADED); {
+	case err == nil, isHResult(err, windows.S_FALSE):
+		// Ours, or already initialised in the same mode: either way this call
+		// counts and has to be balanced.
+		defer windows.CoUninitialize()
+	case isHResult(err, windows.RPC_E_CHANGED_MODE):
+		// The thread is already in the other apartment model. ShellExecuteExW
+		// works there too, and there is nothing of ours to balance.
+	default:
+		return 0, fmt.Errorf("initialise COM: %w", err)
+	}
+
+	ret, _, lastErr := procShellExecuteEx.Call(uintptr(unsafe.Pointer(info)))
+	if ret != 0 {
+		return info.hProcess, nil
+	}
+
+	if errors.Is(lastErr, windows.ERROR_CANCELLED) {
+		return 0, ErrDeclined
+	}
+	return 0, fmt.Errorf("run elevated: %w", lastErr)
+}
+
+// ownerWindow returns this process's foreground window, and 0 when the window in
+// front belongs to somebody else or cannot be attributed. ShellExecuteExW takes it
+// as the parent for the UI it raises, which is what keeps the consent dialog in
+// front of the window the user was just clicking in instead of behind it. It is
+// also what a remote-desktop session needs to place the dialog at all when the
+// secure desktop is switched off.
+func ownerWindow() windows.HWND {
+	hwnd := windows.GetForegroundWindow()
+	if hwnd == 0 {
+		return 0
+	}
+
+	var pid uint32
+	if _, err := windows.GetWindowThreadProcessId(hwnd, &pid); err != nil {
+		log.Debugf("cannot attribute the foreground window, raising the prompt without an owner: %v", err)
+		return 0
+	}
+	if pid != windows.GetCurrentProcessId() {
+		return 0
+	}
+	return hwnd
+}
+
+// isHResult reports whether err carries the given HRESULT. CoInitializeEx
+// returns its HRESULT as an Errno, so the comparison is on the raw value.
+func isHResult(err error, hresult windows.Handle) bool {
+	var errno windows.Errno
+	return errors.As(err, &errno) && uintptr(errno) == uintptr(hresult)
+}
+
+func waitForProcess(ctx context.Context, process windows.Handle) error {
+	// The wait is interruptible so a cancelled context stops us waiting on a
+	// consent dialog nobody is going to answer. The elevated process is not
+	// ours to kill, and it either applies the change or does not.
+	for {
+		event, err := windows.WaitForSingleObject(process, 250)
+		if err != nil {
+			return fmt.Errorf("wait for the elevated process: %w", err)
+		}
+		if event == uint32(windows.WAIT_OBJECT_0) {
+			break
+		}
+		if err := ctx.Err(); err != nil {
+			return err
+		}
+	}
+
+	var code uint32
+	if err := windows.GetExitCodeProcess(process, &code); err != nil {
+		return fmt.Errorf("read the elevated process's exit code: %w", err)
+	}
+	if code != 0 {
+		return fmt.Errorf("elevated netbird exited with %d", code)
+	}
+	return nil
+}
+
+// mechanismAvailable is true on Windows: UAC prompts for consent when the user
+// is an administrator and for an administrator's credentials when they are not,
+// so there is always something to ask.
+func mechanismAvailable() bool {
+	return true
+}

client/internal/elevate/trusted.go

@@ -0,0 +1,40 @@
+package elevate
+
+import (
+	"fmt"
+	"os"
+	"path/filepath"
+)
+
+// trustedSelf returns the path of this executable, provided it is one we are
+// willing to have run as root.
+//
+// The check is what keeps elevation from becoming a way to launder someone
+// else's code into a root process: the user consents to NetBird being elevated,
+// having been shown NetBird's name, so what runs must be the file NetBird was
+// installed as and not something a third party could have swapped for it. An
+// executable only its owner can write is that; anything wider is refused, and
+// the caller falls back to showing the command instead.
+//
+// The owner writing to their own executable is not part of that threat: code
+// running as the user can already prompt them for anything, and could just as
+// well ask them to run the command by hand. What matters is that no *other*
+// unprivileged account can reach it.
+func trustedSelf() (string, error) {
+	exe, err := os.Executable()
+	if err != nil {
+		return "", fmt.Errorf("locate this executable: %w", err)
+	}
+
+	// Resolve symlinks so the checks below apply to the file that would actually
+	// be executed, not to a link somebody else may control.
+	resolved, err := filepath.EvalSymlinks(exe)
+	if err != nil {
+		return "", fmt.Errorf("resolve %s: %w", exe, err)
+	}
+
+	if err := checkOnlyOwnerWritable(resolved); err != nil {
+		return "", fmt.Errorf("%w: %s cannot be trusted to run as root: %w", ErrUnavailable, resolved, err)
+	}
+	return resolved, nil
+}

client/internal/elevate/trusted_group_darwin.go

@@ -0,0 +1,10 @@
+package elevate
+
+// adminWriteGIDs are the groups whose write access to an executable does not
+// widen who could authorize elevating it.
+//
+// macOS installs applications as root:admin, mode 0775, /Applications included,
+// so requiring owner-only write would reject every normal install. Group admin
+// (gid 80) is exactly the set of accounts that can answer the authentication
+// dialog, so its write access grants nothing the prompt would not.
+var adminWriteGIDs = []uint32{0, 80}

client/internal/elevate/trusted_group_unix.go

@@ -0,0 +1,9 @@
+//go:build !windows && !darwin
+
+package elevate
+
+// adminWriteGIDs are the groups whose write access to an executable does not
+// widen who could authorize elevating it. Only root's own group qualifies here:
+// a distribution installs into root-owned directories, and there is no
+// system-wide administrators group that both writes them and answers polkit.
+var adminWriteGIDs = []uint32{0}

client/internal/elevate/trusted_unix.go

@@ -0,0 +1,119 @@
+//go:build !windows
+
+package elevate
+
+import (
+	"errors"
+	"fmt"
+	"os"
+	"path/filepath"
+	"slices"
+	"strconv"
+	"syscall"
+
+	log "github.com/sirupsen/logrus"
+
+	"github.com/netbirdio/netbird/client/internal/getent"
+)
+
+// checkOnlyOwnerWritable reports an error unless path, and every directory leading
+// to it, is owned by either root or this user and writable by nobody who could not
+// already act as its owner. A writable directory is as good as a writable file,
+// since anything in it can be replaced, so the whole chain is checked.
+func checkOnlyOwnerWritable(path string) error {
+	self := uint32(os.Getuid())
+
+	for dir := path; ; dir = filepath.Dir(dir) {
+		info, err := os.Lstat(dir)
+		if err != nil {
+			return fmt.Errorf("stat %s: %w", dir, err)
+		}
+
+		stat, ok := info.Sys().(*syscall.Stat_t)
+		if !ok {
+			return errors.New("file ownership is unavailable on this platform")
+		}
+		if stat.Uid != 0 && stat.Uid != self {
+			return fmt.Errorf("%s is owned by uid %d, neither root nor this user", dir, stat.Uid)
+		}
+
+		if err := checkWriteBits(dir, info, stat.Uid, stat.Gid); err != nil {
+			return err
+		}
+
+		if parent := filepath.Dir(dir); parent == dir {
+			return nil
+		}
+	}
+}
+
+func checkWriteBits(path string, info os.FileInfo, uid, gid uint32) error {
+	// On a directory the sticky bit stands in for the write bits: whoever may
+	// write there still cannot replace an entry they do not own, which is the
+	// only thing that would matter to us. /tmp is the usual example.
+	sticky := info.IsDir() && info.Mode()&os.ModeSticky != 0
+
+	return writeBitsAllow(path, info.Mode().Perm(), sticky, groupWriteAllowed(uid, gid))
+}
+
+// writeBitsAllow decides on the permission bits alone, given whether the group's
+// write access has been vouched for.
+func writeBitsAllow(path string, perm os.FileMode, sticky, groupAllowed bool) error {
+	if sticky {
+		return nil
+	}
+	if perm&0o020 != 0 && !groupAllowed {
+		return fmt.Errorf("%s is writable by a group with members other than its owner (%v)", path, perm)
+	}
+	if perm&0o002 != 0 {
+		return fmt.Errorf("%s is world-writable (%v)", path, perm)
+	}
+	return nil
+}
+
+// groupWriteAllowed reports whether a group's write access to a file owned by uid
+// puts it in reach of anyone who could not already act as that owner.
+//
+// Two ways it does not. A group in adminWriteGIDs holds the accounts that can
+// answer the elevation prompt anyway. And a user private group is how Debian,
+// Ubuntu and Fedora ship: their umask of 002 makes a home directory and
+// everything built in it group-writable, so refusing that would refuse every
+// build not installed from a package.
+func groupWriteAllowed(uid, gid uint32) bool {
+	if slices.Contains(adminWriteGIDs, gid) {
+		return true
+	}
+
+	group, err := getent.LookupGroupID(strconv.FormatUint(uint64(gid), 10))
+	if err != nil {
+		log.Debugf("cannot look up group %d, treating it as shared: %v", gid, err)
+		return false
+	}
+	owner, err := getent.LookupUserID(strconv.FormatUint(uint64(uid), 10))
+	if err != nil {
+		log.Debugf("cannot look up uid %d, treating its group as shared: %v", uid, err)
+		return false
+	}
+
+	if group.Name != owner.Username {
+		return false
+	}
+	return !groupHasOtherMembers(group.Name, owner.Username)
+}
+
+// groupHasOtherMembers reports whether the group lists a member besides owner.
+//
+// Sharing the owner's name is what a user private group is recognised by, and it
+// says nothing about who is in it: a group that has since gained a member is
+// still named that way, and that member can write whatever the group can. So the
+// membership is read rather than assumed. A group whose members cannot be
+// listed, because no source on this host describes it, is treated as shared:
+// the name alone cannot vouch for who writes through it.
+func groupHasOtherMembers(name, owner string) bool {
+	members, err := getent.GroupMembers(name)
+	if err != nil {
+		log.Debugf("cannot list the members of group %q, treating it as shared: %v", name, err)
+		return true
+	}
+	return slices.ContainsFunc(members, func(member string) bool { return member != owner })
+}

client/internal/elevate/trusted_unix_test.go

@@ -0,0 +1,148 @@
+//go:build !windows
+
+package elevate
+
+import (
+	"os"
+	"os/user"
+	"path/filepath"
+	"strconv"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// ownerOnlyDir is t.TempDir() with the write bits tightened. testing creates its
+// numbered directory with 0777 minus the umask, so under the common 002 umask it
+// is group-writable and would fail the check under test on its own.
+func ownerOnlyDir(t *testing.T) string {
+	t.Helper()
+	dir := t.TempDir()
+	require.NoError(t, os.Chmod(dir, 0o755), "tighten the temporary directory")
+	return dir
+}
+
+// writeExecutable creates a plain executable file, the shape trustedSelf checks.
+func writeExecutable(t *testing.T, dir string) string {
+	t.Helper()
+	path := filepath.Join(dir, "netbird-ui")
+	require.NoError(t, os.WriteFile(path, []byte("#!/bin/sh\n"), 0o755), "write the executable")
+	require.NoError(t, os.Chmod(path, 0o755), "set the executable's mode")
+	return path
+}
+
+func TestCheckOnlyOwnerWritableAcceptsOwnerOnly(t *testing.T) {
+	err := checkOnlyOwnerWritable(writeExecutable(t, ownerOnlyDir(t)))
+	assert.NoError(t, err, "an owner-only writable executable is trustworthy")
+}
+
+func TestCheckOnlyOwnerWritableRejectsWorldWritableFile(t *testing.T) {
+	path := writeExecutable(t, ownerOnlyDir(t))
+	require.NoError(t, os.Chmod(path, 0o777), "make the executable world-writable")
+
+	assert.Error(t, checkOnlyOwnerWritable(path), "a world-writable executable must be refused")
+}
+
+// The permission policy on its own, without a filesystem to arrange: whether the
+// group has been vouched for is the only thing that makes group write acceptable.
+func TestWriteBitsAllow(t *testing.T) {
+	tests := []struct {
+		name         string
+		perm         os.FileMode
+		sticky       bool
+		groupAllowed bool
+		wantErr      bool
+	}{
+		{name: "owner only", perm: 0o755},
+		{name: "group write in a private group", perm: 0o775, groupAllowed: true},
+		{name: "group write in a shared group", perm: 0o775, wantErr: true},
+		{name: "world write", perm: 0o777, groupAllowed: true, wantErr: true},
+		{name: "world write on a sticky directory", perm: 0o777, sticky: true},
+		{name: "group write on a sticky directory", perm: 0o775, sticky: true},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			err := writeBitsAllow("/path", tt.perm, tt.sticky, tt.groupAllowed)
+			if tt.wantErr {
+				assert.Error(t, err, "perm %v, sticky %v, group allowed %v", tt.perm, tt.sticky, tt.groupAllowed)
+				return
+			}
+			assert.NoError(t, err, "perm %v, sticky %v, group allowed %v", tt.perm, tt.sticky, tt.groupAllowed)
+		})
+	}
+}
+
+// A build under a home directory on a distribution with a 002 umask, which is what
+// a locally built or tarball-installed binary looks like. Its group has no members
+// but its owner, so it is as good as owner-only.
+//
+// Whether this host is such a distribution is read from the environment rather than
+// from groupWriteAllowed: asking the function under test whether to run would let
+// it skip its own coverage away if it regressed to refusing everything.
+func TestCheckOnlyOwnerWritableAcceptsOwnPrivateGroup(t *testing.T) {
+	requirePrivatePrimaryGroup(t)
+
+	dir := ownerOnlyDir(t)
+	path := writeExecutable(t, dir)
+	require.NoError(t, os.Chmod(dir, 0o775), "make the directory group-writable")
+	require.NoError(t, os.Chmod(path, 0o775), "make the executable group-writable")
+
+	err := checkOnlyOwnerWritable(path)
+	assert.NoError(t, err, "group write in the owner's own private group reaches nobody else")
+}
+
+// A group whose membership no source can answer for is treated as shared: the
+// private-group allowance must not stand on a name nobody can vouch for. The
+// membership listing itself lives in the getent package and is tested there.
+func TestGroupHasOtherMembersRejectsAnUnknownGroup(t *testing.T) {
+	assert.True(t, groupHasOtherMembers("nonexistent_group_xyzzy_12345", "vma"),
+		"a group no source describes")
+}
+
+// A writable directory is as good as a writable file: whoever can write the
+// directory can put a different binary at the same path.
+func TestCheckOnlyOwnerWritableRejectsWritableDirectory(t *testing.T) {
+	dir := filepath.Join(ownerOnlyDir(t), "bin")
+	require.NoError(t, os.Mkdir(dir, 0o755), "create the directory")
+	path := writeExecutable(t, dir)
+	require.NoError(t, os.Chmod(dir, 0o777), "make the directory world-writable")
+
+	assert.Error(t, checkOnlyOwnerWritable(path), "an executable in a world-writable directory must be refused")
+}
+
+// A sticky world-writable directory is exempt: the sticky bit is what stops one
+// user replacing another's entries. /tmp is why this matters.
+func TestCheckOnlyOwnerWritableAcceptsStickyDirectory(t *testing.T) {
+	dir := filepath.Join(ownerOnlyDir(t), "sticky")
+	require.NoError(t, os.Mkdir(dir, 0o755), "create the directory")
+	path := writeExecutable(t, dir)
+	require.NoError(t, os.Chmod(dir, 0o777|os.ModeSticky), "make the directory sticky and world-writable")
+
+	err := checkOnlyOwnerWritable(path)
+	assert.NoError(t, err, "the sticky bit stops another user replacing the executable")
+}
+
+func TestCheckOnlyOwnerWritableRejectsMissingFile(t *testing.T) {
+	err := checkOnlyOwnerWritable(filepath.Join(ownerOnlyDir(t), "absent"))
+	assert.Error(t, err, "an executable that is not there must be refused")
+}
+
+// requirePrivatePrimaryGroup skips unless this user's primary group is their own,
+// which is what the user-private-group allowance is about.
+func requirePrivatePrimaryGroup(t *testing.T) {
+	t.Helper()
+
+	self, err := user.Current()
+	require.NoError(t, err, "look up the test user")
+	group, err := user.LookupGroupId(strconv.Itoa(os.Getgid()))
+	require.NoError(t, err, "look up the test user's primary group")
+
+	if group.Name != self.Username {
+		t.Skipf("the test user's primary group is %q, not their own, so there is nothing to assert here", group.Name)
+	}
+	if groupHasOtherMembers(group.Name, self.Username) {
+		t.Skipf("group %q has other members, so it is not a private group", group.Name)
+	}
+}

client/internal/elevate/trusted_windows.go

@@ -0,0 +1,215 @@
+package elevate
+
+import (
+	"errors"
+	"fmt"
+	"path/filepath"
+	"slices"
+	"unsafe"
+
+	"golang.org/x/sys/windows"
+)
+
+const (
+	// fileDeleteChild is FILE_DELETE_CHILD, which x/sys does not define: the
+	// right to delete an entry of a directory without holding DELETE on it.
+	fileDeleteChild = 0x00000040
+
+	// accessAllowedCallbackACEType is an allow ACE with a condition appended to
+	// the ACCESS_ALLOWED_ACE layout, so its trustee is still at SidStart.
+	accessAllowedCallbackACEType = 0x9
+
+	// The allow ACE types that carry object GUIDs ahead of the trustee, so the
+	// SID is not at SidStart. They occur on directory-service objects rather
+	// than files, and are refused rather than skipped: see aceTrustee.
+	accessAllowedObjectACEType         = 0x5
+	accessAllowedCallbackObjectACEType = 0xB
+)
+
+// fileWriteAccess are the rights that let a trustee rewrite or replace a file,
+// or take it over and then do so.
+const fileWriteAccess = windows.FILE_WRITE_DATA | windows.FILE_APPEND_DATA |
+	windows.DELETE | windows.WRITE_DAC | windows.WRITE_OWNER |
+	windows.GENERIC_WRITE | windows.GENERIC_ALL
+
+// dirWriteAccess are the rights over a directory that let a trustee replace an
+// entry somebody else owns. Creating a new entry is not one of them, which is
+// what the Unix sticky bit says in one bit: the root of every volume grants
+// BUILTIN\Users the right to add directories under it, and that reaches nothing
+// already there.
+const dirWriteAccess = fileDeleteChild | windows.DELETE |
+	windows.WRITE_DAC | windows.WRITE_OWNER | windows.GENERIC_ALL
+
+// trustedInstallerSID owns much of what Windows itself installs. x/sys has no
+// well-known constant for it.
+const trustedInstallerSID = "S-1-5-80-956008885-3418522649-1831038044-1853292631-2271478464"
+
+// checkOnlyOwnerWritable reports an error unless path, and every directory
+// leading to it, is owned by an account that can elevate (or by this user) and
+// grants write access to nobody else. A writable directory is as good as a
+// writable file, since an entry in it can be replaced, so the whole chain is
+// checked.
+func checkOnlyOwnerWritable(path string) error {
+	owners, err := trustedOwners()
+	if err != nil {
+		return err
+	}
+	writers, err := trustedWriters(owners)
+	if err != nil {
+		return err
+	}
+
+	writeAccess := windows.ACCESS_MASK(fileWriteAccess)
+	for target := path; ; target = filepath.Dir(target) {
+		if err := checkSecurity(target, writeAccess, owners, writers); err != nil {
+			return err
+		}
+		if parent := filepath.Dir(target); parent == target {
+			return nil
+		}
+		writeAccess = dirWriteAccess
+	}
+}
+
+// trustedOwners are the accounts we accept as the owner of the executable and of
+// the directories above it: the ones that can already answer the UAC prompt,
+// plus this user, whose own executable is theirs to write. Code running as the
+// user could prompt them for anything anyway; what matters is that no *other*
+// unprivileged account can reach it.
+func trustedOwners() ([]*windows.SID, error) {
+	self, err := currentUserSID()
+	if err != nil {
+		return nil, err
+	}
+
+	owners := []*windows.SID{self}
+	for _, wellKnown := range []windows.WELL_KNOWN_SID_TYPE{
+		windows.WinLocalSystemSid,
+		windows.WinBuiltinAdministratorsSid,
+	} {
+		sid, err := windows.CreateWellKnownSid(wellKnown)
+		if err != nil {
+			return nil, fmt.Errorf("build well-known SID %d: %w", wellKnown, err)
+		}
+		owners = append(owners, sid)
+	}
+
+	installer, err := windows.StringToSid(trustedInstallerSID)
+	if err != nil {
+		return nil, fmt.Errorf("parse TrustedInstaller SID: %w", err)
+	}
+	return append(owners, installer), nil
+}
+
+// trustedWriters are the trustees whose write access does not widen who could
+// decide what runs behind the prompt. The owners, and CREATOR OWNER, which
+// resolves to the object's owner and is therefore already vetted.
+func trustedWriters(owners []*windows.SID) ([]*windows.SID, error) {
+	creatorOwner, err := windows.CreateWellKnownSid(windows.WinCreatorOwnerSid)
+	if err != nil {
+		return nil, fmt.Errorf("build the CREATOR OWNER SID: %w", err)
+	}
+	return append(slices.Clone(owners), creatorOwner), nil
+}
+
+func checkSecurity(path string, writeAccess windows.ACCESS_MASK, owners, writers []*windows.SID) error {
+	sd, err := windows.GetNamedSecurityInfo(path, windows.SE_FILE_OBJECT,
+		windows.OWNER_SECURITY_INFORMATION|windows.DACL_SECURITY_INFORMATION)
+	if err != nil {
+		return fmt.Errorf("read security descriptor of %s: %w", path, err)
+	}
+
+	owner, _, err := sd.Owner()
+	if err != nil {
+		return fmt.Errorf("read owner of %s: %w", path, err)
+	}
+	if !containsSID(owners, owner) {
+		return fmt.Errorf("%s is owned by %s, which is neither this user nor an account that can elevate", path, owner)
+	}
+
+	dacl, _, err := sd.DACL()
+	if err != nil {
+		return fmt.Errorf("read DACL of %s: %w", path, err)
+	}
+	// A NULL DACL grants everyone everything; only an absent security
+	// descriptor would have got us here without one, and neither is trustworthy.
+	if dacl == nil {
+		return fmt.Errorf("%s has no DACL, so it grants write access to everyone", path)
+	}
+
+	return checkDACL(path, dacl, writeAccess, writers)
+}
+
+// checkDACL refuses an ACL that grants write access to a trustee outside
+// writers.
+//
+// An allowlist, because the trustees that must not have it cannot be listed: an
+// ACE naming an ordinary user account hands that account the same power as one
+// naming Everyone, and only the accounts that may hold it are knowable.
+func checkDACL(path string, dacl *windows.ACL, writeAccess windows.ACCESS_MASK, writers []*windows.SID) error {
+	for i := uint32(0); i < uint32(dacl.AceCount); i++ {
+		var ace *windows.ACCESS_ALLOWED_ACE
+		if err := windows.GetAce(dacl, i, &ace); err != nil {
+			return fmt.Errorf("read ACE %d of %s: %w", i, path, err)
+		}
+		// An inherit-only ACE says what children of this object get, not what
+		// this object grants.
+		if ace.Header.AceFlags&windows.INHERIT_ONLY_ACE != 0 {
+			continue
+		}
+		if ace.Mask&writeAccess == 0 {
+			continue
+		}
+		// Only an allow ACE grants anything; a deny ACE narrows what one gave.
+		if !isAllowACE(ace.Header.AceType) {
+			continue
+		}
+
+		trustee, err := aceTrustee(ace)
+		if err != nil {
+			return fmt.Errorf("read the trustee of ACE %d of %s: %w", i, path, err)
+		}
+		if !containsSID(writers, trustee) {
+			return fmt.Errorf("%s grants write access to %s", path, trustee)
+		}
+	}
+	return nil
+}
+
+// isAllowACE reports whether an ACE type grants rights, rather than denying,
+// auditing or labelling them.
+func isAllowACE(aceType uint8) bool {
+	switch aceType {
+	case windows.ACCESS_ALLOWED_ACE_TYPE, accessAllowedCallbackACEType,
+		accessAllowedObjectACEType, accessAllowedCallbackObjectACEType:
+		return true
+	default:
+		return false
+	}
+}
+
+// aceTrustee returns who an allow ACE grants its rights to. An ACE whose trustee
+// cannot be located is an error rather than something to skip past: being unable
+// to read who is being given write access is a refusal.
+func aceTrustee(ace *windows.ACCESS_ALLOWED_ACE) (*windows.SID, error) {
+	switch ace.Header.AceType {
+	case windows.ACCESS_ALLOWED_ACE_TYPE, accessAllowedCallbackACEType:
+		//nolint:gosec // SidStart is the first uint32 of the variable-length SID that follows the ACE header.
+		return (*windows.SID)(unsafe.Pointer(&ace.SidStart)), nil
+	default:
+		return nil, errors.New("an object-type allow ACE does not carry its trustee where we can read it")
+	}
+}
+
+func containsSID(sids []*windows.SID, sid *windows.SID) bool {
+	return slices.ContainsFunc(sids, sid.Equals)
+}
+
+func currentUserSID() (*windows.SID, error) {
+	token := windows.GetCurrentProcessToken()
+	user, err := token.GetTokenUser()
+	if err != nil {
+		return nil, fmt.Errorf("read this process's user: %w", err)
+	}
+	return user.User.Sid, nil
+}

client/internal/elevate/trusted_windows_test.go

@@ -0,0 +1,126 @@
+package elevate
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"golang.org/x/sys/windows"
+)
+
+// A file the test user created under their own profile, which is what a per-user
+// install looks like. The whole chain up to the volume root is walked, so this is
+// also what says the walk does not refuse an ordinary Windows installation: the
+// root of every volume grants BUILTIN\Users rights that are not ours to worry
+// about.
+func TestCheckOnlyOwnerWritableAcceptsOwnFile(t *testing.T) {
+	err := checkOnlyOwnerWritable(writeExecutable(t))
+	assert.NoError(t, err, "a file the test user owns, under directories only administrators can write")
+}
+
+// Write access held by an account that cannot answer the UAC prompt means that
+// account decides what runs behind it, whoever the ACE names. The trustees that
+// must not have it cannot be listed, so the check names the ones that may.
+func TestCheckOnlyOwnerWritableRejectsUntrustedWriters(t *testing.T) {
+	tests := []struct {
+		name      string
+		wellKnown windows.WELL_KNOWN_SID_TYPE
+	}{
+		{name: "everyone", wellKnown: windows.WinWorldSid},
+		{name: "authenticated users", wellKnown: windows.WinAuthenticatedUserSid},
+		{name: "builtin users", wellKnown: windows.WinBuiltinUsersSid},
+		// A service account, which no denylist of the obvious groups would name
+		// and which cannot elevate any more than Everyone can.
+		{name: "local service", wellKnown: windows.WinLocalServiceSid},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			path := writeExecutable(t)
+			grantWrite(t, path, tt.wellKnown)
+
+			assert.Error(t, checkOnlyOwnerWritable(path),
+				"write access for %s must be refused", tt.name)
+		})
+	}
+}
+
+// The masks are the policy: on a file any write reaches its contents, while on a
+// directory only deleting or taking over an entry reaches something already
+// there. Adding an entry does not, which is why the walk survives a volume root.
+func TestWriteAccessMasks(t *testing.T) {
+	assert.NotZero(t, fileWriteAccess&windows.FILE_WRITE_DATA, "writing a file's data reaches its contents")
+	assert.NotZero(t, fileWriteAccess&windows.FILE_APPEND_DATA, "appending to a file reaches its contents")
+
+	assert.Zero(t, dirWriteAccess&windows.FILE_WRITE_DATA, "adding a file to a directory replaces nothing")
+	assert.Zero(t, dirWriteAccess&windows.FILE_APPEND_DATA, "adding a subdirectory replaces nothing")
+	assert.NotZero(t, dirWriteAccess&fileDeleteChild, "deleting an entry replaces it")
+	assert.NotZero(t, dirWriteAccess&windows.DELETE, "deleting the directory takes its entries with it")
+}
+
+func TestIsAllowACE(t *testing.T) {
+	tests := []struct {
+		name    string
+		aceType uint8
+		want    bool
+	}{
+		{name: "allowed", aceType: windows.ACCESS_ALLOWED_ACE_TYPE, want: true},
+		{name: "allowed callback", aceType: accessAllowedCallbackACEType, want: true},
+		{name: "allowed object", aceType: accessAllowedObjectACEType, want: true},
+		{name: "allowed callback object", aceType: accessAllowedCallbackObjectACEType, want: true},
+		{name: "denied", aceType: windows.ACCESS_DENIED_ACE_TYPE},
+		// SYSTEM_AUDIT_ACE_TYPE, which x/sys does not define: an ACE that records
+		// access rather than granting it.
+		{name: "audit", aceType: 0x2},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			assert.Equal(t, tt.want, isAllowACE(tt.aceType), "ACE type %#x", tt.aceType)
+		})
+	}
+}
+
+// writeExecutable creates a plain file under the test's own directory, the shape
+// trustedSelf checks.
+func writeExecutable(t *testing.T) string {
+	t.Helper()
+	path := filepath.Join(t.TempDir(), "netbird-ui.exe")
+	require.NoError(t, os.WriteFile(path, []byte("MZ"), 0o755), "write the executable")
+	return path
+}
+
+// grantWrite replaces the file's DACL with one that grants a well-known trustee
+// everything, keeping the test user's own access so the file stays deletable.
+func grantWrite(t *testing.T, path string, wellKnown windows.WELL_KNOWN_SID_TYPE) {
+	t.Helper()
+
+	trustee, err := windows.CreateWellKnownSid(wellKnown)
+	require.NoError(t, err, "build the trustee SID")
+	self, err := currentUserSID()
+	require.NoError(t, err, "read the test user's SID")
+
+	acl, err := windows.ACLFromEntries([]windows.EXPLICIT_ACCESS{
+		fullControl(self, windows.TRUSTEE_IS_USER),
+		fullControl(trustee, windows.TRUSTEE_IS_WELL_KNOWN_GROUP),
+	}, nil)
+	require.NoError(t, err, "build the ACL")
+
+	require.NoError(t, windows.SetNamedSecurityInfo(path, windows.SE_FILE_OBJECT,
+		windows.DACL_SECURITY_INFORMATION|windows.PROTECTED_DACL_SECURITY_INFORMATION,
+		nil, nil, acl, nil), "set the DACL")
+}
+
+func fullControl(sid *windows.SID, trusteeType uint32) windows.EXPLICIT_ACCESS {
+	return windows.EXPLICIT_ACCESS{
+		AccessPermissions: windows.GENERIC_ALL,
+		AccessMode:        windows.GRANT_ACCESS,
+		Trustee: windows.TRUSTEE{
+			TrusteeForm:  windows.TRUSTEE_IS_SID,
+			TrusteeType:  windows.TRUSTEE_TYPE(trusteeType),
+			TrusteeValue: windows.TrusteeValueFromSID(sid),
+		},
+	}
+}

client/internal/engine.go

@@ -60,6 +60,7 @@ import (
 	"github.com/netbirdio/netbird/client/internal/syncstore"
 	"github.com/netbirdio/netbird/client/internal/updater"
 	"github.com/netbirdio/netbird/client/jobexec"
+	"github.com/netbirdio/netbird/client/netevents"
 	cProto "github.com/netbirdio/netbird/client/proto"
 	"github.com/netbirdio/netbird/client/system"
 	nbdns "github.com/netbirdio/netbird/dns"
@@ -183,6 +184,9 @@ type EngineServices struct {
 	ClientMetrics  *metrics.ClientMetrics
 	MetricsCtx     context.Context
 	FileDrop       *filedrop.Manager
+	// NetMgr gates the reconnection loops on OS-reported network
+	// availability; nil disables gating.
+	NetMgr *netevents.Manager
 }
 
 // Engine is a mechanism responsible for reacting on Signal and Management stream events and managing connections to the remote peers.
@@ -206,6 +210,10 @@ type Engine struct {
 	config    *EngineConfig
 	mobileDep MobileDependency
 
+	// netMgr gates the peer reconnection guards on OS-reported network
+	// availability; nil disables gating.
+	netMgr *netevents.Manager
+
 	// STUNs is a list of STUN servers used by ICE
 	STUNs []*stun.URI
 	// TURNs is a list of STUN servers used by ICE
@@ -343,6 +351,7 @@ func NewEngine(
 		syncMsgMux:         &sync.Mutex{},
 		config:             config,
 		mobileDep:          mobileDep,
+		netMgr:             services.NetMgr,
 		STUNs:              []*stun.URI{},
 		TURNs:              []*stun.URI{},
 		networkSerial:      0,
@@ -872,8 +881,7 @@ func (e *Engine) modifyPeers(peersUpdate []*mgmProto.RemotePeerConfig) error {
 	}
 	// third, add the peer connections again
 	for _, p := range modified {
-		err := e.addNewPeer(p)
-		if err != nil {
+		if err := e.addNewPeer(p); err != nil {
 			return err
 		}
 	}
@@ -1497,8 +1505,12 @@ func (e *Engine) updateNetworkMap(networkMap *mgmProto.NetworkMap) error {
 		return nil
 	}
 
-	if err := e.connMgr.UpdatedRemoteFeatureFlag(e.ctx, networkMap.GetPeerConfig().GetLazyConnectionEnabled()); err != nil {
-		log.Errorf("failed to update lazy connection feature flag: %v", err)
+	// Only update the flag when the sync carries a peer config; a nil peer config
+	// (e.g. a partial update) must not reset the cached flag to false.
+	if peerConfig := networkMap.GetPeerConfig(); peerConfig != nil {
+		if err := e.connMgr.UpdatedRemoteFeatureFlag(e.ctx, peerConfig.GetLazyConnectionEnabled()); err != nil {
+			log.Errorf("failed to update lazy connection feature flag: %v", err)
+		}
 	}
 
 	if e.firewall != nil {
@@ -1564,8 +1576,7 @@ func (e *Engine) updateNetworkMap(networkMap *mgmProto.NetworkMap) error {
 
 	// Ingress forward rules
 	done = e.phase("forward_rules")
-	forwardingRules, err := e.updateForwardRules(networkMap.GetForwardingRules())
-	if err != nil {
+	if _, err := e.updateForwardRules(networkMap.GetForwardingRules()); err != nil {
 		log.Errorf("failed to update forward rules, err: %v", err)
 	}
 	done()
@@ -1583,8 +1594,7 @@ func (e *Engine) updateNetworkMap(networkMap *mgmProto.NetworkMap) error {
 
 	// must set the exclude list after the peers are added. Without it the manager can not figure out the peers parameters from the store
 	done = e.phase("lazy_exclude")
-	excludedLazyPeers := e.toExcludedLazyPeers(forwardingRules, remotePeers)
-	e.connMgr.SetExcludeList(e.ctx, excludedLazyPeers)
+	e.connMgr.SetExcludeList(e.ctx, e.toExcludedLazyPeers(remotePeers))
 	done()
 
 	e.networkSerial = serial
@@ -1828,15 +1838,15 @@ func addrToString(addr netip.Addr) string {
 // addNewPeers adds peers that were not know before but arrived from the Management service with the update
 func (e *Engine) addNewPeers(peersUpdate []*mgmProto.RemotePeerConfig) error {
 	for _, p := range peersUpdate {
-		err := e.addNewPeer(p)
-		if err != nil {
+		if err := e.addNewPeer(p); err != nil {
 			return err
 		}
 	}
 	return nil
 }
 
-// addNewPeer add peer if connection doesn't exist
+// addNewPeer add peer if connection doesn't exist. A peer that is not lazy by
+// policy gets an always-active connection instead.
 func (e *Engine) addNewPeer(peerConfig *mgmProto.RemotePeerConfig) error {
 	peerKey := peerConfig.GetWgPubKey()
 	peerIPs := make([]netip.Prefix, 0, len(peerConfig.GetAllowedIps()))
@@ -1871,7 +1881,8 @@ func (e *Engine) addNewPeer(peerConfig *mgmProto.RemotePeerConfig) error {
 		log.Warnf("error adding peer %s to status recorder, got error: %v", peerKey, err)
 	}
 
-	if exists := e.connMgr.AddPeerConn(e.ctx, peerKey, conn); exists {
+	permanent := !e.connMgr.PeerLazyDefault(peerConfig.GetLazyState())
+	if exists := e.connMgr.AddPeerConn(e.ctx, peerKey, conn, permanent); exists {
 		conn.Close(false)
 		return fmt.Errorf("peer already exists: %s", peerKey)
 	}
@@ -1905,6 +1916,7 @@ func (e *Engine) createPeerConn(pubKey string, allowedIPs []netip.Prefix, agentV
 			PermissiveMode: e.config.RosenpassPermissive,
 		},
 		ICEConfig: e.createICEConfig(),
+		NetMgr:    e.netMgr,
 	}
 
 	serviceDependencies := peer.ServiceDependencies{
@@ -2580,7 +2592,7 @@ func (e *Engine) SetCapture(pc device.PacketCapture) error {
 	}
 
 	afc := capture.NewAFPacketCapture(intf.Name(), sess)
-	if err := afc.Start(); err != nil {
+	if err := afc.Start(); err != nil { //nolint:staticcheck // always errors on non-Linux builds
 		return fmt.Errorf("start AF_PACKET capture on %s: %w", intf.Name(), err)
 	}
 	e.afpacketCapture = afc
@@ -2669,46 +2681,19 @@ func (e *Engine) updateForwardRules(rules []*mgmProto.ForwardingRule) ([]firewal
 	return forwardingRules, nberrors.FormatErrorOrNil(merr)
 }
 
-func (e *Engine) toExcludedLazyPeers(rules []firewallManager.ForwardRule, peers []*mgmProto.RemotePeerConfig) map[string]bool {
+// toExcludedLazyPeers returns the peers that must have an always-active
+// connection: those that are not lazy by policy (the per-peer lazy state or the
+// account flag, subject to the local override).
+func (e *Engine) toExcludedLazyPeers(peers []*mgmProto.RemotePeerConfig) map[string]bool {
 	excludedPeers := make(map[string]bool)
-
-	// Ingress forward targets: inbound forwarded traffic is initiated remotely and
-	// cannot wake a lazy connection, so the peer routing the target must stay
-	// permanently connected. AllowedIPs are already parsed on the peer conn, so
-	// reuse those typed prefixes instead of re-parsing the network map strings.
-	for _, r := range rules {
-		for _, p := range peers {
-			if e.peerRoutesAddr(p, r.TranslatedAddress) {
-				log.Infof("exclude forwarder peer from lazy connection: %s", p.GetWgPubKey())
-				excludedPeers[p.GetWgPubKey()] = true
-			}
+	for _, p := range peers {
+		if !e.connMgr.PeerLazyDefault(p.GetLazyState()) {
+			excludedPeers[p.GetWgPubKey()] = true
 		}
 	}
-
 	return excludedPeers
 }
 
-// peerRoutesAddr reports whether the peer is a router for addr, matched against
-// the peer's already-parsed AllowedIPs from the store (the same typed value the
-// lazy manager consumes) rather than re-parsing the network map strings.
-func (e *Engine) peerRoutesAddr(p *mgmProto.RemotePeerConfig, addr netip.Addr) bool {
-	prefixes, ok := e.peerStore.AllowedIPs(p.GetWgPubKey())
-	if !ok {
-		return false
-	}
-	return prefixesContain(prefixes, addr)
-}
-
-// prefixesContain reports whether addr falls within any of the prefixes.
-func prefixesContain(prefixes []netip.Prefix, addr netip.Addr) bool {
-	for _, prefix := range prefixes {
-		if prefix.Contains(addr) {
-			return true
-		}
-	}
-	return false
-}
-
 // isChecksEqual checks if two slices of checks are equal.
 func isChecksEqual(checks1, checks2 []*mgmProto.Checks) bool {
 	normalize := func(checks []*mgmProto.Checks) []string {

client/internal/engine_lazy_exclude_test.go

@@ -1,87 +0,0 @@
-package internal
-
-import (
-	"net/netip"
-	"testing"
-
-	"github.com/stretchr/testify/require"
-
-	firewallManager "github.com/netbirdio/netbird/client/firewall/manager"
-	"github.com/netbirdio/netbird/client/internal/peer"
-	"github.com/netbirdio/netbird/client/internal/peerstore"
-	mgmProto "github.com/netbirdio/netbird/shared/management/proto"
-)
-
-func TestPrefixesContain(t *testing.T) {
-	tests := []struct {
-		name     string
-		prefixes []string
-		addr     string
-		want     bool
-	}{
-		{name: "own overlay /32 matches", prefixes: []string{"100.110.8.145/32"}, addr: "100.110.8.145", want: true},
-		{name: "addr inside routed subnet", prefixes: []string{"10.121.0.0/16"}, addr: "10.121.208.4", want: true},
-		{name: "addr outside subnet", prefixes: []string{"10.121.0.0/16"}, addr: "10.122.0.1", want: false},
-		{name: "different /32", prefixes: []string{"100.110.8.145/32"}, addr: "100.110.8.146", want: false},
-		{name: "ipv6 /128 matches", prefixes: []string{"fd00::1/128"}, addr: "fd00::1", want: true},
-		{name: "no prefixes", prefixes: nil, addr: "10.121.208.4", want: false},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			prefixes := make([]netip.Prefix, 0, len(tt.prefixes))
-			for _, p := range tt.prefixes {
-				prefixes = append(prefixes, netip.MustParsePrefix(p))
-			}
-			require.Equal(t, tt.want, prefixesContain(prefixes, netip.MustParseAddr(tt.addr)))
-		})
-	}
-}
-
-// TestToExcludedLazyPeers_ForwardTarget guards a regression: the forward-target
-// peer (the peer routing a ForwardRule.TranslatedAddress) must be excluded from
-// lazy connections, matched via the peer's already-parsed AllowedIPs.
-func TestToExcludedLazyPeers_ForwardTarget(t *testing.T) {
-	const targetPeerKey = "cccccccccccccccccccccccccccccccccccccccccc0="
-	const otherPeerKey = "dddddddddddddddddddddddddddddddddddddddddd0="
-
-	store := peerstore.NewConnStore()
-	store.AddPeerConn(targetPeerKey, newTestConn(t, targetPeerKey, "100.110.8.145/32"))
-	store.AddPeerConn(otherPeerKey, newTestConn(t, otherPeerKey, "100.110.9.10/32"))
-
-	e := &Engine{peerStore: store}
-
-	peers := []*mgmProto.RemotePeerConfig{
-		{WgPubKey: targetPeerKey, AllowedIps: []string{"100.110.8.145/32"}},
-		{WgPubKey: otherPeerKey, AllowedIps: []string{"100.110.9.10/32"}},
-	}
-	rules := []firewallManager.ForwardRule{
-		{TranslatedAddress: netip.MustParseAddr("100.110.8.145")},
-	}
-
-	excluded := e.toExcludedLazyPeers(rules, peers)
-
-	require.True(t, excluded[targetPeerKey], "forward-target peer must be excluded from lazy connections")
-	require.False(t, excluded[otherPeerKey], "non-target peer must not be excluded")
-	require.Len(t, excluded, 1)
-}
-
-func TestToExcludedLazyPeers_NoRules(t *testing.T) {
-	e := &Engine{peerStore: peerstore.NewConnStore()}
-
-	peers := []*mgmProto.RemotePeerConfig{
-		{WgPubKey: "peer-a", AllowedIps: []string{"100.110.8.145/32"}},
-	}
-
-	require.Empty(t, e.toExcludedLazyPeers(nil, peers))
-}
-
-func newTestConn(t *testing.T, key, allowedIP string) *peer.Conn {
-	t.Helper()
-	conn, err := peer.NewConn(peer.ConnConfig{
-		Key:      key,
-		WgConfig: peer.WgConfig{AllowedIps: []netip.Prefix{netip.MustParsePrefix(allowedIP)}},
-	}, peer.ServiceDependencies{})
-	require.NoError(t, err)
-	return conn
-}

client/internal/engine_privileged_test.go

@@ -12,7 +12,7 @@ import (
 	"testing"
 	"time"
 
-	"github.com/golang/mock/gomock"
+	"go.uber.org/mock/gomock"
 	"github.com/google/uuid"
 	log "github.com/sirupsen/logrus"
 	"github.com/stretchr/testify/assert"

client/internal/engine_test.go

@@ -279,7 +279,8 @@ func TestEngine_UpdateNetworkMap(t *testing.T) {
 	}, MobileDependency{})
 
 	wgIface := &MockWGIface{
-		NameFunc: func() string { return "utun102" },
+		NameFunc:            func() string { return "utun102" },
+		IsUserspaceBindFunc: func() bool { return true },
 		RemovePeerFunc: func(peerKey string) error {
 			return nil
 		},

client/internal/getent/cgo_unix.go

@@ -0,0 +1,36 @@
+//go:build cgo && !osusergo && !windows
+
+package getent
+
+import "os/user"
+
+// Built with cgo, os/user resolves through libc (getpwnam_r and friends),
+// which goes through the host's NSS stack natively. Whatever it fails to
+// find, the getent command would not find either, so there is nothing to
+// fall back to.
+
+// LookupUser looks up a user by name.
+func LookupUser(username string) (*user.User, error) {
+	return user.Lookup(username)
+}
+
+// LookupUserID looks up a user by UID.
+func LookupUserID(uid string) (*user.User, error) {
+	return user.LookupId(uid)
+}
+
+// CurrentUser returns the user this process runs as.
+func CurrentUser() (*user.User, error) {
+	return user.Current()
+}
+
+// LookupGroupID looks up a group by GID.
+func LookupGroupID(gid string) (*user.Group, error) {
+	return user.LookupGroupId(gid)
+}
+
+// GroupIDs returns the IDs of the groups the user is a member of; libc's
+// getgrouplist handles NSS groups natively.
+func GroupIDs(u *user.User) ([]string, error) {
+	return u.GroupIds()
+}

client/internal/getent/getent.go

@@ -0,0 +1,6 @@
+// Package getent resolves users and groups through the host's NSS stack.
+// Built without cgo, os/user reads /etc/passwd and /etc/group alone and misses
+// anything LDAP, SSSD or winbind provide; the getent and id commands resolve
+// through NSS whatever the build. The lookups here try the standard library
+// first, which needs no subprocess, and fall back to those commands.
+package getent

client/internal/getent/getent_test.go

@@ -0,0 +1,155 @@
+package getent
+
+import (
+	"os/user"
+	"runtime"
+	"strconv"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestLookupUser_CurrentUser(t *testing.T) {
+	// The current user should always be resolvable on any platform
+	current, err := user.Current()
+	require.NoError(t, err)
+
+	u, err := LookupUser(current.Username)
+	require.NoError(t, err)
+	assert.Equal(t, current.Username, u.Username)
+	assert.Equal(t, current.Uid, u.Uid)
+	assert.Equal(t, current.Gid, u.Gid)
+}
+
+func TestLookupUser_NonexistentUser(t *testing.T) {
+	_, err := LookupUser("nonexistent_user_xyzzy_12345")
+	require.Error(t, err, "should fail for nonexistent user")
+}
+
+func TestLookupUserID_CurrentUser(t *testing.T) {
+	current, err := user.Current()
+	require.NoError(t, err)
+
+	u, err := LookupUserID(current.Uid)
+	require.NoError(t, err)
+	assert.Equal(t, current.Username, u.Username)
+	assert.Equal(t, current.Uid, u.Uid)
+}
+
+func TestCurrentUser(t *testing.T) {
+	stdUser, err := user.Current()
+	require.NoError(t, err)
+
+	u, err := CurrentUser()
+	require.NoError(t, err)
+	assert.Equal(t, stdUser.Uid, u.Uid)
+	assert.Equal(t, stdUser.Username, u.Username)
+}
+
+func TestGroupIDs_CurrentUser(t *testing.T) {
+	current, err := user.Current()
+	require.NoError(t, err)
+
+	groups, err := GroupIDs(current)
+	require.NoError(t, err)
+	require.NotEmpty(t, groups, "current user should have at least one group")
+
+	if runtime.GOOS != "windows" {
+		for _, gid := range groups {
+			_, err := strconv.ParseUint(gid, 10, 32)
+			assert.NoError(t, err, "group ID %q should be a valid uint32", gid)
+		}
+	}
+}
+
+func TestUserShell_CurrentUser(t *testing.T) {
+	current, err := user.Current()
+	require.NoError(t, err)
+
+	// getent may not be available on all systems (e.g., macOS without
+	// Homebrew getent), and Windows has no login shells at all.
+	shell, err := UserShell(current.Uid)
+	if err != nil {
+		t.Logf("UserShell failed, getent may not be available: %v", err)
+		return
+	}
+	if shell == "" {
+		t.Log("UserShell returned empty, the user has no shell set")
+		return
+	}
+	assert.True(t, shell[0] == '/', "shell should be an absolute path, got %q", shell)
+}
+
+func TestLookupUser_RootUser(t *testing.T) {
+	if runtime.GOOS == "windows" {
+		t.Skip("no root user on Windows")
+	}
+
+	u, err := LookupUser("root")
+	if err != nil {
+		t.Skip("root user not available on this system")
+	}
+	assert.Equal(t, "0", u.Uid, "root should have UID 0")
+}
+
+// TestIntegration_FullLookupChain exercises the complete user lookup chain
+// against the real system, testing that all wrappers (LookupUser,
+// CurrentUser, GroupIDs, UserShell) produce consistent and correct results
+// when composed together.
+func TestIntegration_FullLookupChain(t *testing.T) {
+	// Step 1: CurrentUser must resolve the running user.
+	current, err := CurrentUser()
+	require.NoError(t, err, "CurrentUser must resolve the running user")
+	require.NotEmpty(t, current.Uid)
+	require.NotEmpty(t, current.Username)
+
+	// Step 2: LookupUser by the same username must return matching identity.
+	byName, err := LookupUser(current.Username)
+	require.NoError(t, err)
+	assert.Equal(t, current.Uid, byName.Uid, "lookup by name should return same UID")
+	assert.Equal(t, current.Gid, byName.Gid, "lookup by name should return same GID")
+	assert.Equal(t, current.HomeDir, byName.HomeDir, "lookup by name should return same home")
+
+	// Step 3: GroupIDs must return at least the primary GID.
+	groups, err := GroupIDs(current)
+	require.NoError(t, err)
+	require.NotEmpty(t, groups, "user must have at least one group")
+
+	foundPrimary := false
+	for _, gid := range groups {
+		if runtime.GOOS != "windows" {
+			_, err := strconv.ParseUint(gid, 10, 32)
+			require.NoError(t, err, "group ID %q must be a valid uint32", gid)
+		}
+		if gid == current.Gid {
+			foundPrimary = true
+		}
+	}
+	assert.True(t, foundPrimary, "primary GID %s should appear in supplementary groups", current.Gid)
+}
+
+// TestIntegration_LookupAndGroupsConsistency verifies that a user resolved via
+// LookupUser can have their groups resolved via GroupIDs, testing the handoff
+// between the two functions as used by the SSH server.
+func TestIntegration_LookupAndGroupsConsistency(t *testing.T) {
+	current, err := user.Current()
+	require.NoError(t, err)
+
+	// Simulate the SSH server flow: lookup user, then get their groups.
+	resolved, err := LookupUser(current.Username)
+	require.NoError(t, err)
+
+	groups, err := GroupIDs(resolved)
+	require.NoError(t, err)
+	require.NotEmpty(t, groups, "resolved user must have groups")
+
+	// On Unix, all returned GIDs must be valid numeric values.
+	// On Windows, group IDs are SIDs (e.g., "S-1-5-32-544").
+	if runtime.GOOS != "windows" {
+		for _, gid := range groups {
+			_, err := strconv.ParseUint(gid, 10, 32)
+			assert.NoError(t, err, "group ID %q should be numeric", gid)
+		}
+	}
+}

client/internal/getent/nocgo_unix.go

@@ -0,0 +1,110 @@
+//go:build (!cgo || osusergo) && !windows
+
+package getent
+
+import (
+	"os"
+	"os/user"
+	"strconv"
+
+	log "github.com/sirupsen/logrus"
+)
+
+// Without cgo, os/user only reads /etc/passwd and /etc/group and misses
+// NSS-provided users and groups; the getent and id commands go through the
+// host's NSS stack.
+
+// LookupUser looks up a user by name, falling back to getent if os/user fails.
+func LookupUser(username string) (*user.User, error) {
+	u, err := user.Lookup(username)
+	if err == nil {
+		return u, nil
+	}
+
+	stdErr := err
+	log.Debugf("os/user.Lookup(%q) failed, trying getent: %v", username, err)
+
+	u, _, getentErr := passwdLookup(username)
+	if getentErr != nil {
+		log.Debugf("getent fallback for %q also failed: %v", username, getentErr)
+		return nil, stdErr
+	}
+	return u, nil
+}
+
+// LookupUserID looks up a user by UID, falling back to getent if os/user fails.
+func LookupUserID(uid string) (*user.User, error) {
+	u, err := user.LookupId(uid)
+	if err == nil {
+		return u, nil
+	}
+
+	stdErr := err
+	log.Debugf("os/user.LookupId(%q) failed, trying getent: %v", uid, err)
+
+	u, _, getentErr := passwdLookup(uid)
+	if getentErr != nil {
+		log.Debugf("getent fallback for uid %s also failed: %v", uid, getentErr)
+		return nil, stdErr
+	}
+	return u, nil
+}
+
+// CurrentUser returns the user this process runs as, falling back to getent
+// if os/user fails.
+func CurrentUser() (*user.User, error) {
+	u, err := user.Current()
+	if err == nil {
+		return u, nil
+	}
+
+	stdErr := err
+	uid := strconv.Itoa(os.Getuid())
+	log.Debugf("os/user.Current() failed, trying getent with UID %s: %v", uid, err)
+
+	u, _, getentErr := passwdLookup(uid)
+	if getentErr != nil {
+		return nil, stdErr
+	}
+	return u, nil
+}
+
+// LookupGroupID looks up a group by GID, falling back to getent if os/user
+// fails.
+func LookupGroupID(gid string) (*user.Group, error) {
+	g, err := user.LookupGroupId(gid)
+	if err == nil {
+		return g, nil
+	}
+
+	stdErr := err
+	log.Debugf("os/user.LookupGroupId(%q) failed, trying getent: %v", gid, err)
+
+	g, _, getentErr := groupLookup(gid)
+	if getentErr != nil {
+		log.Debugf("getent fallback for gid %s also failed: %v", gid, getentErr)
+		return nil, stdErr
+	}
+	return g, nil
+}
+
+// GroupIDs returns the IDs of the groups the user is a member of.
+// NOTE: unlike the lookups above, which try the standard library first, this
+// intentionally tries `id -G` first because without cgo, user.GroupIds only
+// reads /etc/group and silently returns incomplete results for NSS users
+// (no error, just missing groups). The id command goes through NSS and
+// returns the full set.
+func GroupIDs(u *user.User) ([]string, error) {
+	ids, err := idGroups(u.Username)
+	if err == nil {
+		return ids, nil
+	}
+
+	log.Debugf("id -G %q failed, falling back to user.GroupIds(): %v", u.Username, err)
+
+	ids, stdErr := u.GroupIds()
+	if stdErr != nil {
+		return nil, stdErr
+	}
+	return ids, nil
+}

client/internal/getent/unix.go

@@ -0,0 +1,224 @@
+//go:build !windows
+
+package getent
+
+import (
+	"bufio"
+	"context"
+	"fmt"
+	"os"
+	"os/exec"
+	"os/user"
+	"runtime"
+	"strings"
+	"time"
+
+	log "github.com/sirupsen/logrus"
+)
+
+const commandTimeout = 5 * time.Second
+
+// groupFile lists which accounts are in which group, for hosts where the
+// getent command is not available (macOS ships without it).
+const groupFile = "/etc/group"
+
+// UserShell returns the login shell getent reports for the user with this UID.
+// It reaches shells that /etc/passwd does not list, because getent resolves
+// through the host's NSS stack.
+func UserShell(uid string) (string, error) {
+	_, shell, err := passwdLookup(uid)
+	if err != nil {
+		return "", err
+	}
+	return shell, nil
+}
+
+// GroupMembers returns the names of the group's members: from getent, which
+// resolves through NSS, or from /etc/group where getent is not available. A
+// group neither source describes is an error; an empty member list is not,
+// since accounts with the group as their primary one are not listed in it.
+func GroupMembers(name string) ([]string, error) {
+	_, members, err := groupLookup(name)
+	if err == nil {
+		return members, nil
+	}
+	log.Debugf("getent cannot list group %q, reading %s: %v", name, groupFile, err)
+	return groupMembersFromFile(groupFile, name)
+}
+
+// passwdLookup executes `getent passwd <query>`, where query is a username or
+// UID, and returns the user and login shell.
+func passwdLookup(query string) (*user.User, string, error) {
+	out, err := run("passwd", query)
+	if err != nil {
+		return nil, "", err
+	}
+	return parsePasswd(string(out))
+}
+
+// groupLookup executes `getent group <query>`, where query is a group name or
+// GID, and returns the group and its member names.
+func groupLookup(query string) (*user.Group, []string, error) {
+	out, err := run("group", query)
+	if err != nil {
+		return nil, nil, err
+	}
+	return parseGroup(string(out))
+}
+
+// run executes `getent <database> <key>` with a timeout.
+func run(database, key string) ([]byte, error) {
+	if !validateInput(key) {
+		return nil, fmt.Errorf("invalid getent input: %q", key)
+	}
+
+	ctx, cancel := context.WithTimeout(context.Background(), commandTimeout)
+	defer cancel()
+
+	out, err := exec.CommandContext(ctx, "getent", database, key).Output()
+	if err != nil {
+		return nil, fmt.Errorf("getent %s %s: %w", database, key, err)
+	}
+	return out, nil
+}
+
+// parsePasswd parses getent passwd output: "name:x:uid:gid:gecos:home:shell"
+func parsePasswd(output string) (*user.User, string, error) {
+	fields := strings.SplitN(strings.TrimSpace(output), ":", 8)
+	if len(fields) < 6 {
+		return nil, "", fmt.Errorf("unexpected getent output (need 6+ fields): %q", output)
+	}
+
+	if fields[0] == "" || fields[2] == "" || fields[3] == "" {
+		return nil, "", fmt.Errorf("missing required fields in getent output: %q", output)
+	}
+
+	var shell string
+	if len(fields) >= 7 {
+		shell = fields[6]
+	}
+
+	return &user.User{
+		Username: fields[0],
+		Uid:      fields[2],
+		Gid:      fields[3],
+		Name:     fields[4],
+		HomeDir:  fields[5],
+	}, shell, nil
+}
+
+// parseGroup parses getent group output: "name:x:gid:member,member"
+func parseGroup(output string) (*user.Group, []string, error) {
+	fields := strings.SplitN(strings.TrimSpace(output), ":", 4)
+	if len(fields) < 3 {
+		return nil, nil, fmt.Errorf("unexpected getent output (need 3+ fields): %q", output)
+	}
+
+	if fields[0] == "" || fields[2] == "" {
+		return nil, nil, fmt.Errorf("missing required fields in getent output: %q", output)
+	}
+
+	var members []string
+	if len(fields) >= 4 {
+		members = splitMembers(fields[3])
+	}
+	return &user.Group{Name: fields[0], Gid: fields[2]}, members, nil
+}
+
+func splitMembers(list string) []string {
+	var members []string
+	for member := range strings.SplitSeq(list, ",") {
+		if member != "" {
+			members = append(members, member)
+		}
+	}
+	return members
+}
+
+// groupMembersFromFile finds the group's member list in a file of /etc/group's
+// format. A group the file does not describe, because it comes from LDAP or
+// another NSS source, is an error rather than an empty list.
+func groupMembersFromFile(path, name string) ([]string, error) {
+	file, err := os.Open(path)
+	if err != nil {
+		return nil, fmt.Errorf("open %s: %w", path, err)
+	}
+	defer func() {
+		if err := file.Close(); err != nil {
+			log.Debugf("close %s: %v", path, err)
+		}
+	}()
+
+	scanner := bufio.NewScanner(file)
+	for scanner.Scan() {
+		// name:password:gid:member,member
+		fields := strings.Split(scanner.Text(), ":")
+		if len(fields) < 4 || fields[0] != name {
+			continue
+		}
+		return splitMembers(fields[3]), nil
+	}
+	if err := scanner.Err(); err != nil {
+		return nil, fmt.Errorf("read %s: %w", path, err)
+	}
+	return nil, fmt.Errorf("%s does not describe group %q", path, name)
+}
+
+// validateInput checks that the input is safe to pass to getent or id.
+// Allows POSIX usernames, numeric IDs, and common NSS extensions
+// (@ for Kerberos, $ for Samba, + for NIS compat). A leading hyphen is
+// rejected so the input can never be parsed as a command-line flag.
+func validateInput(input string) bool {
+	maxLen := 32
+	if runtime.GOOS == "linux" {
+		maxLen = 256
+	}
+
+	if len(input) == 0 || len(input) > maxLen {
+		return false
+	}
+
+	if input[0] == '-' {
+		return false
+	}
+
+	for _, r := range input {
+		if isAllowedChar(r) {
+			continue
+		}
+		return false
+	}
+	return true
+}
+
+func isAllowedChar(r rune) bool {
+	if r >= 'a' && r <= 'z' || r >= 'A' && r <= 'Z' || r >= '0' && r <= '9' {
+		return true
+	}
+	switch r {
+	case '.', '_', '-', '@', '+', '$':
+		return true
+	}
+	return false
+}
+
+// idGroups runs `id -G <username>` and returns the space-separated group IDs.
+func idGroups(username string) ([]string, error) {
+	if !validateInput(username) {
+		return nil, fmt.Errorf("invalid username for id command: %q", username)
+	}
+
+	ctx, cancel := context.WithTimeout(context.Background(), commandTimeout)
+	defer cancel()
+
+	out, err := exec.CommandContext(ctx, "id", "-G", username).Output()
+	if err != nil {
+		return nil, fmt.Errorf("id -G %s: %w", username, err)
+	}
+
+	trimmed := strings.TrimSpace(string(out))
+	if trimmed == "" {
+		return nil, fmt.Errorf("id -G %s: empty output", username)
+	}
+	return strings.Fields(trimmed), nil
+}

client/internal/getent/unix_test.go

@@ -0,0 +1,471 @@
+//go:build !windows
+
+package getent
+
+import (
+	"os"
+	"os/exec"
+	"os/user"
+	"path/filepath"
+	"runtime"
+	"strconv"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestParsePasswd(t *testing.T) {
+	tests := []struct {
+		name        string
+		input       string
+		wantUser    *user.User
+		wantShell   string
+		wantErr     bool
+		errContains string
+	}{
+		{
+			name:  "standard entry",
+			input: "alice:x:1001:1001:Alice Smith:/home/alice:/bin/bash\n",
+			wantUser: &user.User{
+				Username: "alice",
+				Uid:      "1001",
+				Gid:      "1001",
+				Name:     "Alice Smith",
+				HomeDir:  "/home/alice",
+			},
+			wantShell: "/bin/bash",
+		},
+		{
+			name:  "root entry",
+			input: "root:x:0:0:root:/root:/bin/bash",
+			wantUser: &user.User{
+				Username: "root",
+				Uid:      "0",
+				Gid:      "0",
+				Name:     "root",
+				HomeDir:  "/root",
+			},
+			wantShell: "/bin/bash",
+		},
+		{
+			name:  "empty gecos field",
+			input: "svc:x:999:999::/var/lib/svc:/usr/sbin/nologin",
+			wantUser: &user.User{
+				Username: "svc",
+				Uid:      "999",
+				Gid:      "999",
+				Name:     "",
+				HomeDir:  "/var/lib/svc",
+			},
+			wantShell: "/usr/sbin/nologin",
+		},
+		{
+			name:  "gecos with commas",
+			input: "john:x:1002:1002:John Doe,Room 101,555-1234,555-4321:/home/john:/bin/zsh",
+			wantUser: &user.User{
+				Username: "john",
+				Uid:      "1002",
+				Gid:      "1002",
+				Name:     "John Doe,Room 101,555-1234,555-4321",
+				HomeDir:  "/home/john",
+			},
+			wantShell: "/bin/zsh",
+		},
+		{
+			name:  "remote user with large UID",
+			input: "remoteuser:*:50001:50001:Remote User:/home/remoteuser:/bin/bash\n",
+			wantUser: &user.User{
+				Username: "remoteuser",
+				Uid:      "50001",
+				Gid:      "50001",
+				Name:     "Remote User",
+				HomeDir:  "/home/remoteuser",
+			},
+			wantShell: "/bin/bash",
+		},
+		{
+			name:  "no shell field (only 6 fields)",
+			input: "minimal:x:1000:1000::/home/minimal",
+			wantUser: &user.User{
+				Username: "minimal",
+				Uid:      "1000",
+				Gid:      "1000",
+				Name:     "",
+				HomeDir:  "/home/minimal",
+			},
+			wantShell: "",
+		},
+		{
+			name:        "too few fields",
+			input:       "bad:x:1000",
+			wantErr:     true,
+			errContains: "need 6+ fields",
+		},
+		{
+			name:        "empty username",
+			input:       ":x:1000:1000::/home/test:/bin/bash",
+			wantErr:     true,
+			errContains: "missing required fields",
+		},
+		{
+			name:        "empty UID",
+			input:       "test:x::1000::/home/test:/bin/bash",
+			wantErr:     true,
+			errContains: "missing required fields",
+		},
+		{
+			name:        "empty GID",
+			input:       "test:x:1000:::/home/test:/bin/bash",
+			wantErr:     true,
+			errContains: "missing required fields",
+		},
+		{
+			name:        "empty input",
+			input:       "",
+			wantErr:     true,
+			errContains: "need 6+ fields",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			u, shell, err := parsePasswd(tt.input)
+			if tt.wantErr {
+				require.Error(t, err)
+				if tt.errContains != "" {
+					assert.Contains(t, err.Error(), tt.errContains)
+				}
+				return
+			}
+			require.NoError(t, err)
+			assert.Equal(t, tt.wantUser.Username, u.Username, "username")
+			assert.Equal(t, tt.wantUser.Uid, u.Uid, "UID")
+			assert.Equal(t, tt.wantUser.Gid, u.Gid, "GID")
+			assert.Equal(t, tt.wantUser.Name, u.Name, "name/gecos")
+			assert.Equal(t, tt.wantUser.HomeDir, u.HomeDir, "home directory")
+			assert.Equal(t, tt.wantShell, shell, "shell")
+		})
+	}
+}
+
+func TestParseGroup(t *testing.T) {
+	tests := []struct {
+		name        string
+		input       string
+		wantGroup   *user.Group
+		wantMembers []string
+		wantErr     bool
+	}{
+		{
+			name:      "no members",
+			input:     "vma:x:1000:\n",
+			wantGroup: &user.Group{Name: "vma", Gid: "1000"},
+		},
+		{
+			name:        "one member",
+			input:       "sudo:x:27:alice",
+			wantGroup:   &user.Group{Name: "sudo", Gid: "27"},
+			wantMembers: []string{"alice"},
+		},
+		{
+			name:        "several members",
+			input:       "docker:x:998:alice,bob\n",
+			wantGroup:   &user.Group{Name: "docker", Gid: "998"},
+			wantMembers: []string{"alice", "bob"},
+		},
+		{
+			name:    "too few fields",
+			input:   "bad:x",
+			wantErr: true,
+		},
+		{
+			name:    "empty group name",
+			input:   ":x:1000:alice",
+			wantErr: true,
+		},
+		{
+			name:    "empty GID",
+			input:   "vma:x::alice",
+			wantErr: true,
+		},
+		{
+			name:    "empty input",
+			input:   "",
+			wantErr: true,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			g, members, err := parseGroup(tt.input)
+			if tt.wantErr {
+				require.Error(t, err)
+				return
+			}
+			require.NoError(t, err)
+			assert.Equal(t, tt.wantGroup.Name, g.Name, "group name")
+			assert.Equal(t, tt.wantGroup.Gid, g.Gid, "GID")
+			assert.Equal(t, tt.wantMembers, members, "members")
+		})
+	}
+}
+
+func TestGroupMembersFromFile(t *testing.T) {
+	tests := []struct {
+		name  string
+		entry string
+		want  []string
+	}{
+		{name: "no members", entry: "vma:x:1000:"},
+		{name: "only the owner", entry: "vma:x:1000:vma", want: []string{"vma"}},
+		{name: "two members", entry: "vma:x:1000:vma,bob", want: []string{"vma", "bob"}},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			path := filepath.Join(t.TempDir(), "group")
+			body := "root:x:0:\n" + tt.entry + "\nsudo:x:27:vma\n"
+			require.NoError(t, os.WriteFile(path, []byte(body), 0o644), "write the group file")
+
+			members, err := groupMembersFromFile(path, "vma")
+			require.NoError(t, err, "entry %q", tt.entry)
+			assert.Equal(t, tt.want, members, "entry %q", tt.entry)
+		})
+	}
+}
+
+// A group the file does not describe, because it comes from LDAP or another
+// NSS source, is an error rather than an empty member list: the caller must
+// be able to tell "no members" from "no answer".
+func TestGroupMembersFromFileUnknownGroup(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "group")
+	require.NoError(t, os.WriteFile(path, []byte("root:x:0:\n"), 0o644), "write the group file")
+
+	_, err := groupMembersFromFile(path, "vma")
+	assert.Error(t, err, "a group the file does not describe")
+
+	_, err = groupMembersFromFile(filepath.Join(t.TempDir(), "absent"), "vma")
+	assert.Error(t, err, "no group file at all")
+}
+
+// GroupMembers on the root group, which every Unix has, whichever source
+// answers for it.
+func TestGroupMembers_RootGroup(t *testing.T) {
+	rootGroup := "root"
+	switch runtime.GOOS {
+	case "darwin", "dragonfly", "freebsd", "netbsd", "openbsd":
+		rootGroup = "wheel"
+	}
+
+	_, err := GroupMembers(rootGroup)
+	assert.NoError(t, err, "the %s group must be describable", rootGroup)
+}
+
+func TestValidateInput(t *testing.T) {
+	tests := []struct {
+		name  string
+		input string
+		want  bool
+	}{
+		{"normal username", "alice", true},
+		{"numeric UID", "1001", true},
+		{"dots and underscores", "alice.bob_test", true},
+		{"hyphen", "alice-bob", true},
+		{"leading hyphen rejected", "-i", false},
+		{"leading double hyphen rejected", "--no-idn", false},
+		{"lone hyphen rejected", "-", false},
+		{"kerberos principal", "user@REALM", true},
+		{"samba machine account", "MACHINE$", true},
+		{"NIS compat", "+user", true},
+		{"empty", "", false},
+		{"null byte", "alice\x00bob", false},
+		{"newline", "alice\nbob", false},
+		{"tab", "alice\tbob", false},
+		{"control char", "alice\x01bob", false},
+		{"DEL char", "alice\x7fbob", false},
+		{"space rejected", "alice bob", false},
+		{"semicolon rejected", "alice;bob", false},
+		{"backtick rejected", "alice`bob", false},
+		{"pipe rejected", "alice|bob", false},
+		{"33 chars exceeds non-linux max", makeLongString(33), runtime.GOOS == "linux"},
+		{"256 chars at linux max", makeLongString(256), runtime.GOOS == "linux"},
+		{"257 chars exceeds all limits", makeLongString(257), false},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			assert.Equal(t, tt.want, validateInput(tt.input))
+		})
+	}
+}
+
+func makeLongString(n int) string {
+	b := make([]byte, n)
+	for i := range b {
+		b[i] = 'a'
+	}
+	return string(b)
+}
+
+func TestPasswdLookup_RootUser(t *testing.T) {
+	if _, err := exec.LookPath("getent"); err != nil {
+		t.Skip("getent not available on this system")
+	}
+
+	u, shell, err := passwdLookup("root")
+	require.NoError(t, err)
+	assert.Equal(t, "root", u.Username)
+	assert.Equal(t, "0", u.Uid)
+	assert.Equal(t, "0", u.Gid)
+	assert.NotEmpty(t, shell, "root should have a shell")
+}
+
+func TestPasswdLookup_ByUID(t *testing.T) {
+	if _, err := exec.LookPath("getent"); err != nil {
+		t.Skip("getent not available on this system")
+	}
+
+	u, _, err := passwdLookup("0")
+	require.NoError(t, err)
+	assert.Equal(t, "root", u.Username)
+	assert.Equal(t, "0", u.Uid)
+}
+
+func TestPasswdLookup_NonexistentUser(t *testing.T) {
+	if _, err := exec.LookPath("getent"); err != nil {
+		t.Skip("getent not available on this system")
+	}
+
+	_, _, err := passwdLookup("nonexistent_user_xyzzy_12345")
+	assert.Error(t, err)
+}
+
+func TestPasswdLookup_InvalidInput(t *testing.T) {
+	_, _, err := passwdLookup("")
+	assert.Error(t, err)
+
+	_, _, err = passwdLookup("user\x00name")
+	assert.Error(t, err)
+}
+
+func TestPasswdLookup_NotAvailable(t *testing.T) {
+	if _, err := exec.LookPath("getent"); err == nil {
+		t.Skip("getent is available, can't test missing case")
+	}
+
+	_, _, err := passwdLookup("root")
+	assert.Error(t, err, "should fail when getent is not installed")
+}
+
+func TestGroupLookup_RootGroup(t *testing.T) {
+	if _, err := exec.LookPath("getent"); err != nil {
+		t.Skip("getent not available on this system")
+	}
+
+	g, _, err := groupLookup("0")
+	require.NoError(t, err)
+	assert.Equal(t, "0", g.Gid, "GID 0 resolves to the root group")
+	assert.NotEmpty(t, g.Name, "the root group has a name")
+}
+
+func TestIdGroups_CurrentUser(t *testing.T) {
+	if _, err := exec.LookPath("id"); err != nil {
+		t.Skip("id not available on this system")
+	}
+
+	current, err := user.Current()
+	require.NoError(t, err)
+
+	groups, err := idGroups(current.Username)
+	require.NoError(t, err)
+	require.NotEmpty(t, groups, "current user should have at least one group")
+
+	for _, gid := range groups {
+		_, err := strconv.ParseUint(gid, 10, 32)
+		assert.NoError(t, err, "group ID %q should be a valid uint32", gid)
+	}
+}
+
+func TestIdGroups_NonexistentUser(t *testing.T) {
+	if _, err := exec.LookPath("id"); err != nil {
+		t.Skip("id not available on this system")
+	}
+
+	_, err := idGroups("nonexistent_user_xyzzy_12345")
+	assert.Error(t, err)
+}
+
+func TestIdGroups_InvalidInput(t *testing.T) {
+	_, err := idGroups("")
+	assert.Error(t, err)
+
+	_, err = idGroups("user\x00name")
+	assert.Error(t, err)
+}
+
+func TestGetentResultsMatchStdlib(t *testing.T) {
+	if _, err := exec.LookPath("getent"); err != nil {
+		t.Skip("getent not available on this system")
+	}
+
+	current, err := user.Current()
+	require.NoError(t, err)
+
+	getentUser, _, err := passwdLookup(current.Username)
+	require.NoError(t, err)
+
+	assert.Equal(t, current.Username, getentUser.Username, "username should match")
+	assert.Equal(t, current.Uid, getentUser.Uid, "UID should match")
+	assert.Equal(t, current.Gid, getentUser.Gid, "GID should match")
+	assert.Equal(t, current.HomeDir, getentUser.HomeDir, "home directory should match")
+}
+
+func TestGetentResultsMatchStdlib_ByUID(t *testing.T) {
+	if _, err := exec.LookPath("getent"); err != nil {
+		t.Skip("getent not available on this system")
+	}
+
+	current, err := user.Current()
+	require.NoError(t, err)
+
+	getentUser, _, err := passwdLookup(current.Uid)
+	require.NoError(t, err)
+
+	assert.Equal(t, current.Username, getentUser.Username, "username should match when looked up by UID")
+	assert.Equal(t, current.Uid, getentUser.Uid, "UID should match")
+}
+
+func TestIdGroupsMatchStdlib(t *testing.T) {
+	if _, err := exec.LookPath("id"); err != nil {
+		t.Skip("id not available on this system")
+	}
+
+	current, err := user.Current()
+	require.NoError(t, err)
+
+	stdGroups, err := current.GroupIds()
+	if err != nil {
+		t.Skip("os/user.GroupIds() not working, likely CGO_ENABLED=0")
+	}
+
+	idGroupIDs, err := idGroups(current.Username)
+	require.NoError(t, err)
+
+	// Deduplicate both lists: id -G can return duplicates (e.g., root in Docker)
+	// and ElementsMatch treats duplicates as distinct.
+	assert.ElementsMatch(t, uniqueStrings(stdGroups), uniqueStrings(idGroupIDs), "id -G should return same groups as os/user")
+}
+
+func uniqueStrings(ss []string) []string {
+	seen := make(map[string]struct{}, len(ss))
+	out := make([]string, 0, len(ss))
+	for _, s := range ss {
+		if _, ok := seen[s]; ok {
+			continue
+		}
+		seen[s] = struct{}{}
+		out = append(out, s)
+	}
+	return out
+}

client/internal/getent/windows.go

@@ -0,0 +1,36 @@
+//go:build windows
+
+package getent
+
+import (
+	"errors"
+	"os/user"
+)
+
+// Windows does not use NSS or getent; os/user resolves accounts there
+// without cgo, so everything delegates to it.
+
+// LookupUser looks up a user by name.
+func LookupUser(username string) (*user.User, error) {
+	return user.Lookup(username)
+}
+
+// LookupUserID looks up a user by UID.
+func LookupUserID(uid string) (*user.User, error) {
+	return user.LookupId(uid)
+}
+
+// CurrentUser returns the user this process runs as.
+func CurrentUser() (*user.User, error) {
+	return user.Current()
+}
+
+// GroupIDs returns the IDs of the groups the user is a member of.
+func GroupIDs(u *user.User) ([]string, error) {
+	return u.GroupIds()
+}
+
+// UserShell is unanswerable on Windows, which has no login-shell database.
+func UserShell(string) (string, error) {
+	return "", errors.ErrUnsupported
+}

client/internal/ipcauth/privileged.go

@@ -91,6 +91,12 @@ func SelfDelegatesTo() (Identity, bool) {
 	return selfIdentity, true
 }
 
+// The values PrivilegedActorKey returns.
+const (
+	ActorKeyAdministrator = "administrator"
+	ActorKeyRoot          = "root"
+)
+
 // PrivilegedActor names the principal a privileged operation requires, for use
 // in messages shown to the user.
 func PrivilegedActor() string {
@@ -100,6 +106,16 @@ func PrivilegedActor() string {
 	return "root"
 }
 
+// PrivilegedActorKey identifies that principal without wording it, for a client
+// that writes its own message in the user's language. The words PrivilegedActor
+// returns are English, and a translated sentence cannot borrow them.
+func PrivilegedActorKey() string {
+	if runtime.GOOS == "windows" {
+		return ActorKeyAdministrator
+	}
+	return ActorKeyRoot
+}
+
 // ElevatedCommand renders a command so that running it grants the privileges the
 // operation needs. Windows has no in-line equivalent of sudo, so the command is
 // returned unchanged and the user is expected to run it from an elevated

client/internal/peer/conn.go

@@ -26,6 +26,7 @@ import (
 	"github.com/netbirdio/netbird/client/internal/portforward"
 	"github.com/netbirdio/netbird/client/internal/rosenpass"
 	"github.com/netbirdio/netbird/client/internal/stdnet"
+	"github.com/netbirdio/netbird/client/netevents"
 	"github.com/netbirdio/netbird/route"
 	relayClient "github.com/netbirdio/netbird/shared/relay/client"
 )
@@ -93,6 +94,10 @@ type ConnConfig struct {
 
 	// ICEConfig ICE protocol configuration
 	ICEConfig icemaker.Config
+
+	// NetMgr gates the reconnection guard on OS-reported network
+	// availability; nil disables gating.
+	NetMgr *netevents.Manager
 }
 
 type Conn struct {
@@ -254,7 +259,7 @@ func (conn *Conn) open(engineCtx context.Context, firstPacket []byte) error {
 		conn.handshaker.AddICEListener(conn.workerICE.OnNewOffer)
 	}
 
-	conn.guard = guard.NewGuard(conn.Log, conn.isConnectedOnAllWay, conn.config.Timeout, conn.srWatcher)
+	conn.guard = guard.NewGuard(conn.Log, conn.isConnectedOnAllWay, conn.config.Timeout, conn.srWatcher, conn.config.NetMgr)
 
 	conn.wg.Add(1)
 	go func() {
@@ -440,7 +445,7 @@ func (conn *Conn) onICEConnectionIsReady(priority conntype.ConnPriority, iceConn
 		conn.dumpState.NewLocalProxy()
 		wgProxy, err = conn.newProxy(iceConnInfo.RemoteConn)
 		if err != nil {
-			conn.Log.Errorf("failed to add turn net.Conn to local proxy: %v", err)
+			conn.Log.Errorf("failed to add relayed net.Conn to local proxy: %v", err)
 			return
 		}
 		ep = wgProxy.EndpointAddr()
@@ -878,9 +883,8 @@ func (conn *Conn) newProxy(remoteConn net.Conn) (wgproxy.Proxy, error) {
 	}
 
 	wgProxy := conn.config.WgConfig.WgInterface.GetProxy()
-	if err := wgProxy.AddTurnConn(conn.ctx, udpAddr, remoteConn); err != nil {
-		conn.Log.Errorf("failed to add turn net.Conn to local proxy: %v", err)
-		return nil, err
+	if err := wgProxy.AddRelayedConn(conn.ctx, udpAddr, remoteConn); err != nil {
+		return nil, fmt.Errorf("add relayed conn to proxy: %w", err)
 	}
 	return wgProxy, nil
 }

client/internal/peer/guard/guard.go

@@ -22,6 +22,12 @@ const (
 
 type connStatusFunc func() ConnStatus
 
+// NetworkWatcher is the availability view the guard gates reconnects on.
+type NetworkWatcher interface {
+	IsOnline() bool
+	Changed() <-chan struct{}
+}
+
 // Guard is responsible for the reconnection logic.
 // It will trigger to send an offer to the peer then has connection issues.
 // Watch these events:
@@ -31,20 +37,26 @@ type connStatusFunc func() ConnStatus
 // - Relayed connection disconnected
 // - ICE candidate changes
 type Guard struct {
-	log                     *log.Entry
-	isConnectedOnAllWay     connStatusFunc
-	timeout                 time.Duration
-	srWatcher               *SRWatcher
+	log                 *log.Entry
+	isConnectedOnAllWay connStatusFunc
+	timeout             time.Duration
+	srWatcher           *SRWatcher
+	// netWatcher gates reconnect attempts on OS-reported network availability;
+	// nil disables gating.
+	netWatcher              NetworkWatcher
 	relayedConnDisconnected chan struct{}
 	iCEConnDisconnected     chan struct{}
 }
 
-func NewGuard(log *log.Entry, isConnectedFn connStatusFunc, timeout time.Duration, srWatcher *SRWatcher) *Guard {
+// NewGuard creates a reconnection guard for a peer connection. A nil netWatcher
+// disables network availability gating.
+func NewGuard(log *log.Entry, isConnectedFn connStatusFunc, timeout time.Duration, srWatcher *SRWatcher, netWatcher NetworkWatcher) *Guard {
 	return &Guard{
 		log:                     log,
 		isConnectedOnAllWay:     isConnectedFn,
 		timeout:                 timeout,
 		srWatcher:               srWatcher,
+		netWatcher:              netWatcher,
 		relayedConnDisconnected: make(chan struct{}, 1),
 		iCEConnDisconnected:     make(chan struct{}, 1),
 	}
@@ -96,9 +108,19 @@ func (g *Guard) reconnectLoopWithRetry(ctx context.Context, callback func()) {
 	iceState := &iceRetryState{log: g.log}
 	defer iceState.reset()
 
+	var netChanged <-chan struct{}
+	if g.netWatcher != nil {
+		netChanged = g.netWatcher.Changed()
+	}
+
 	for {
 		select {
 		case <-tickerChannel:
+			// skip attempts while the OS reports no usable network; the
+			// netChanged case below resumes the loop once it returns
+			if g.netWatcher != nil && !g.netWatcher.IsOnline() {
+				continue
+			}
 			switch g.isConnectedOnAllWay() {
 			case ConnStatusConnected:
 				// all good, nothing to do
@@ -135,6 +157,23 @@ func (g *Guard) reconnectLoopWithRetry(ctx context.Context, callback func()) {
 			tickerChannel = ticker.C
 			iceState.reset()
 
+		case <-netChanged:
+			// Re-arm for the next transition before acting on this one.
+			netChanged = g.netWatcher.Changed()
+			if !g.netWatcher.IsOnline() {
+				continue
+			}
+			// Ticks skipped while offline drove the backoff towards its
+			// maximum without ever attempting, and left the ICE budget
+			// frozen — possibly in hourly mode. Recover on our own so the
+			// peer does not depend on a signal or relay event that never
+			// comes when both stayed up across the outage.
+			g.log.Debugf("network is back, reset reconnection ticker")
+			ticker.Stop()
+			ticker = g.newReconnectTicker(ctx)
+			tickerChannel = ticker.C
+			iceState.reset()
+
 		case <-ctx.Done():
 			g.log.Debugf("context is done, stop reconnect loop")
 			return

client/internal/peer/guard/guard_leak_test.go

@@ -15,7 +15,7 @@ import (
 
 func newTestGuard(status connStatusFunc) *Guard {
 	srw := NewSRWatcher(nil, nil, nil, ice.Config{})
-	return NewGuard(log.WithField("test", "guard"), status, 50*time.Millisecond, srw)
+	return NewGuard(log.WithField("test", "guard"), status, 50*time.Millisecond, srw, nil)
 }
 
 // countBackoffTickerGoroutines returns how many goroutines are currently sitting

client/internal/peer/guard/guard_netstate_test.go

@@ -0,0 +1,107 @@
+package guard
+
+import (
+	"context"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	log "github.com/sirupsen/logrus"
+
+	"github.com/netbirdio/netbird/client/internal/peer/ice"
+	"github.com/netbirdio/netbird/client/netevents/netstate"
+)
+
+// newTestGuardWithNetState builds a guard with a realistic MaxInterval: the
+// backoff must be able to grow well past the outage, as it does in production
+// where the timeout is seconds to minutes.
+func newTestGuardWithNetState(status connStatusFunc, netState *netstate.State) *Guard {
+	srw := NewSRWatcher(nil, nil, nil, ice.Config{})
+	return NewGuard(log.WithField("test", "guard"), status, 30*time.Second, srw, netState)
+}
+
+// TestGuard_RecoversAfterOfflineToOnline covers a peer that stays disconnected
+// across a network outage while neither signal nor relay reports an event —
+// both stayed up, as on a short airplane mode toggle over Wi-Fi.
+//
+// Every tick taken while offline is skipped, but it still advances the
+// exponential backoff, so by the time the network returns the next tick can be
+// tens of seconds away. Without an explicit reaction to the transition the
+// peer waits out that interval for a recovery that could start immediately.
+func TestGuard_RecoversAfterOfflineToOnline(t *testing.T) {
+	netState := netstate.New()
+
+	var attempts atomic.Int32
+	g := newTestGuardWithNetState(func() ConnStatus { return ConnStatusDisconnected }, netState)
+
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	// Start from the reconnect ticker (800ms initial interval), the state a
+	// peer is in after it loses its connection.
+	go g.Start(ctx, func() { attempts.Add(1) })
+	g.SetRelayedConnDisconnected()
+
+	// Let the backoff climb: 0.8s, 1.6s, 3.2s, 6.4s ... every tick is skipped
+	// while offline, but each one doubles the wait for the next.
+	netState.Set(false)
+	time.Sleep(8 * time.Second)
+
+	offlineAttempts := attempts.Load()
+	if offlineAttempts != 0 {
+		t.Fatalf("callback ran %d times while offline, want 0", offlineAttempts)
+	}
+
+	netState.Set(true)
+
+	// The next organic tick is now several seconds out, so anything within
+	// this window can only come from reacting to the transition itself.
+	pollCtx, stopPolling := context.WithTimeout(ctx, 2*time.Second)
+	defer stopPolling()
+
+	select {
+	case <-pollCtx.Done():
+		t.Fatal("peer was not retried within 2s of the network coming back, " +
+			"with neither a signal nor a relay event to fall back on")
+	case <-pollUntil(pollCtx, func() bool { return attempts.Load() > 0 }):
+	}
+}
+
+// TestGuard_OfflineTransitionDoesNotRetry checks the other direction: going
+// offline must not itself trigger an attempt.
+func TestGuard_OfflineTransitionDoesNotRetry(t *testing.T) {
+	netState := netstate.New()
+
+	var attempts atomic.Int32
+	g := newTestGuardWithNetState(func() ConnStatus { return ConnStatusDisconnected }, netState)
+
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	go g.Start(ctx, func() { attempts.Add(1) })
+
+	netState.Set(false)
+	time.Sleep(5 * time.Second)
+
+	if got := attempts.Load(); got != 0 {
+		t.Fatalf("callback ran %d times after going offline, want 0", got)
+	}
+}
+
+// pollUntil closes the returned channel once cond holds. It gives up when ctx
+// is done, so the polling goroutine never outlives the test that started it.
+func pollUntil(ctx context.Context, cond func() bool) <-chan struct{} {
+	done := make(chan struct{})
+	go func() {
+		for {
+			if cond() {
+				close(done)
+				return
+			}
+			select {
+			case <-ctx.Done():
+				return
+			case <-time.After(10 * time.Millisecond):
+			}
+		}
+	}()
+	return done
+}

client/internal/peer/handshaker.go

@@ -81,14 +81,19 @@ type Handshaker struct {
 
 func NewHandshaker(log *log.Entry, config ConnConfig, signaler *Signaler, ice *WorkerICE, relay *WorkerRelay, metricsStages *MetricsStages) *Handshaker {
 	h := &Handshaker{
-		log:            log,
-		config:         config,
-		signaler:       signaler,
-		ice:            ice,
-		relay:          relay,
-		metricsStages:  metricsStages,
-		remoteOffersCh: make(chan OfferAnswer),
-		remoteAnswerCh: make(chan OfferAnswer),
+		log:           log,
+		config:        config,
+		signaler:      signaler,
+		ice:           ice,
+		relay:         relay,
+		metricsStages: metricsStages,
+		// Buffered by one so an offer or answer that arrives between Open launching
+		// the Listen goroutine and it reaching its receive is held rather than
+		// dropped. A peer activated by an incoming signal receives the remote's
+		// message in that window; an unbuffered channel skips it as "receiver not
+		// ready", and the connection cannot proceed until the remote re-sends.
+		remoteOffersCh: make(chan OfferAnswer, 1),
+		remoteAnswerCh: make(chan OfferAnswer, 1),
 	}
 	// assume remote supports ICE until we learn otherwise from received offers
 	h.remoteICESupported.Store(ice != nil)
@@ -162,29 +167,38 @@ func (h *Handshaker) SendOffer() error {
 	return h.sendOffer()
 }
 
-// OnRemoteOffer handles an offer from the remote peer and returns true if the message was accepted, false otherwise
-// doesn't block, discards the message if connection wasn't ready
+// OnRemoteOffer hands an offer to Listen without blocking, keeping only the most
+// recent one if several arrive before Listen reads them.
 func (h *Handshaker) OnRemoteOffer(offer OfferAnswer) {
-	select {
-	case h.remoteOffersCh <- offer:
-		return
-	default:
-		h.log.Warnf("skipping remote offer message because receiver not ready")
-		// connection might not be ready yet to receive so we ignore the message
-		return
-	}
+	enqueueLatest(h.remoteOffersCh, offer)
 }
 
-// OnRemoteAnswer handles an offer from the remote peer and returns true if the message was accepted, false otherwise
-// doesn't block, discards the message if connection wasn't ready
+// OnRemoteAnswer hands an answer to Listen without blocking, keeping only the most
+// recent one if several arrive before Listen reads them.
 func (h *Handshaker) OnRemoteAnswer(answer OfferAnswer) {
+	enqueueLatest(h.remoteAnswerCh, answer)
+}
+
+// enqueueLatest delivers msg on a one-slot channel without blocking. When the slot
+// already holds an unread message the older one is discarded in favor of msg, so a
+// message arriving before Listen starts reading is held rather than dropped, and
+// the newest wins if several arrive first. Safe because there is a single producer
+// (the engine loop): after draining the stale value the send always has room.
+func enqueueLatest(ch chan OfferAnswer, msg OfferAnswer) {
 	select {
-	case h.remoteAnswerCh <- answer:
+	case ch <- msg:
 		return
 	default:
-		// connection might not be ready yet to receive so we ignore the message
-		h.log.Warnf("skipping remote answer message because receiver not ready")
-		return
+	}
+
+	select {
+	case <-ch:
+	default:
+	}
+
+	select {
+	case ch <- msg:
+	default:
 	}
 }
 

client/internal/peer/handshaker_test.go

@@ -0,0 +1,63 @@
+package peer
+
+import (
+	"testing"
+	"time"
+
+	log "github.com/sirupsen/logrus"
+	"github.com/stretchr/testify/assert"
+)
+
+func newTestHandshaker(t *testing.T) *Handshaker {
+	t.Helper()
+	// The tests exercise the answer path, whose Listen branch dispatches to the
+	// relay listener without sending an answer, so no signaler/ICE/relay is needed.
+	return NewHandshaker(log.WithField("test", t.Name()), ConnConfig{}, nil, nil, nil, nil)
+}
+
+// TestHandshakerHoldsSignalArrivingBeforeListen covers the case where a peer is
+// activated by an incoming signal: the remote's offer/answer arrives in the same
+// step that opens the connection, before the Listen loop starts reading. The
+// message must be held rather than dropped, or the connection cannot proceed until
+// the remote re-sends. This is the path taken when an eager peer connects to a
+// lazily-managed one.
+func TestHandshakerHoldsSignalArrivingBeforeListen(t *testing.T) {
+	h := newTestHandshaker(t)
+
+	processed := make(chan *OfferAnswer, 4)
+	h.AddRelayListener(func(o *OfferAnswer) { processed <- o })
+
+	// Delivered before Listen is reading, as when the peer is woken by the remote's
+	// signal and the message is delivered right after Open.
+	h.OnRemoteAnswer(OfferAnswer{WgListenPort: 51820})
+
+	go h.Listen(t.Context())
+
+	select {
+	case <-processed:
+	case <-time.After(2 * time.Second):
+		assert.Fail(t, "remote-answer dispatch: signal delivered before Listen was ready was dropped")
+	}
+}
+
+// TestHandshakerKeepsLatestSignalBeforeListen covers several signals arriving
+// before Listen reads: the newest must win (matching the latest-offer contract),
+// rather than the first being kept and later ones discarded.
+func TestHandshakerKeepsLatestSignalBeforeListen(t *testing.T) {
+	h := newTestHandshaker(t)
+
+	processed := make(chan *OfferAnswer, 4)
+	h.AddRelayListener(func(o *OfferAnswer) { processed <- o })
+
+	h.OnRemoteAnswer(OfferAnswer{WgListenPort: 1111})
+	h.OnRemoteAnswer(OfferAnswer{WgListenPort: 2222})
+
+	go h.Listen(t.Context())
+
+	select {
+	case got := <-processed:
+		assert.Equal(t, 2222, got.WgListenPort, "remote-answer dispatch: the latest queued signal should be processed")
+	case <-time.After(2 * time.Second):
+		assert.Fail(t, "remote-answer dispatch: queued signal was dropped")
+	}
+}

client/internal/peer/listener.go

@@ -1,11 +1,40 @@
 package peer
 
+// ClientState identifies the client connection state delivered via
+// Listener.OnStateChanged.
+type ClientState int
+
+// Client states. The numeric values cross the gomobile boundary (the mobile
+// bindings re-export them as integer constants), so they are a wire format:
+// append new states at the end, never reorder or insert.
+const (
+	ClientStateDisconnected ClientState = iota
+	ClientStateConnected
+	ClientStateConnecting
+	ClientStateDisconnecting
+	// ClientStateNoNetwork is an overlay state: it is never stored as the
+	// last notification, only derived from ClientStateConnecting while the
+	// OS reports no usable network (see notifier.effectiveState).
+	ClientStateNoNetwork
+)
+
 // Listener is a callback type about the NetBird network connection state
 type Listener interface {
+	// OnStateChanged reports every client state transition. New states are
+	// delivered only through this callback; the per-state callbacks below
+	// are kept for compatibility and will be removed once all consumers
+	// have migrated.
+	OnStateChanged(state ClientState)
+
+	// Deprecated: consume OnStateChanged instead.
 	OnConnected()
+	// Deprecated: consume OnStateChanged instead.
 	OnDisconnected()
+	// Deprecated: consume OnStateChanged instead.
 	OnConnecting()
+	// Deprecated: consume OnStateChanged instead.
 	OnDisconnecting()
+
 	OnAddressChanged(string, string)
 	OnPeersListChanged(int)
 }

client/internal/peer/notifier.go

@@ -4,31 +4,64 @@ import (
 	"sync"
 )
 
-const (
-	stateDisconnected = iota
-	stateConnected
-	stateConnecting
-	stateDisconnecting
-)
-
 type notifier struct {
+	// publishLock orders state publication: it is held across computing the
+	// effective state and handing it to the listener, so a transition cannot
+	// overtake a newer one and leave the listener on a stale state.
+	publishLock        sync.Mutex
 	serverStateLock    sync.Mutex
 	listenersLock      sync.Mutex
 	listener           Listener
 	currentClientState bool
-	lastNotification   int
+	lastNotification   ClientState
 	lastNumberOfPeers  int
 	lastFqdnAddress    string
 	lastIPAddress      string
+	networkAvailable   bool
 }
 
 func newNotifier() *notifier {
-	return &notifier{}
+	return &notifier{
+		networkAvailable: true,
+	}
+}
+
+// effectiveState maps the computed state to what listeners should see:
+// while the OS reports no usable network, "Connecting" would be a lie —
+// connection attempts are suspended — so it is reported as NoNetwork.
+// Caller must hold serverStateLock.
+func (n *notifier) effectiveState(state ClientState) ClientState {
+	if !n.networkAvailable && state == ClientStateConnecting {
+		return ClientStateNoNetwork
+	}
+	return state
+}
+
+// setNetworkAvailable records the OS network availability and re-notifies
+// the listener when the flag flips the effective state (Connecting <->
+// NoNetwork).
+func (n *notifier) setNetworkAvailable(available bool) {
+	n.publishLock.Lock()
+	defer n.publishLock.Unlock()
+
+	n.serverStateLock.Lock()
+	if n.networkAvailable == available {
+		n.serverStateLock.Unlock()
+		return
+	}
+	previous := n.effectiveState(n.lastNotification)
+	n.networkAvailable = available
+	current := n.effectiveState(n.lastNotification)
+	n.serverStateLock.Unlock()
+
+	if previous != current {
+		n.notify(current)
+	}
 }
 
 func (n *notifier) setListener(listener Listener) {
 	n.serverStateLock.Lock()
-	lastNotification := n.lastNotification
+	lastNotification := n.effectiveState(n.lastNotification)
 	numOfPeers := n.lastNumberOfPeers
 	fqdnAddress := n.lastFqdnAddress
 	address := n.lastIPAddress
@@ -52,6 +85,9 @@ func (n *notifier) removeListener() {
 }
 
 func (n *notifier) updateServerStates(mgmState bool, signalState bool) {
+	n.publishLock.Lock()
+	defer n.publishLock.Unlock()
+
 	n.serverStateLock.Lock()
 	calculatedState := n.calculateState(mgmState, signalState)
 
@@ -61,43 +97,54 @@ func (n *notifier) updateServerStates(mgmState bool, signalState bool) {
 	}
 
 	n.lastNotification = calculatedState
+	effective := n.effectiveState(calculatedState)
 	n.serverStateLock.Unlock()
 
-	n.notify(calculatedState)
+	n.notify(effective)
 }
 
 func (n *notifier) clientStart() {
+	n.publishLock.Lock()
+	defer n.publishLock.Unlock()
+
 	n.serverStateLock.Lock()
 	n.currentClientState = true
-	n.lastNotification = stateConnecting
+	n.lastNotification = ClientStateConnecting
+	effective := n.effectiveState(ClientStateConnecting)
 	n.serverStateLock.Unlock()
 
-	n.notify(stateConnecting)
+	n.notify(effective)
 }
 
 func (n *notifier) clientStop() {
+	n.publishLock.Lock()
+	defer n.publishLock.Unlock()
+
 	n.serverStateLock.Lock()
 	n.currentClientState = false
-	n.lastNotification = stateDisconnected
+	n.lastNotification = ClientStateDisconnected
 	n.serverStateLock.Unlock()
 
-	n.notify(stateDisconnected)
+	n.notify(ClientStateDisconnected)
 }
 
 func (n *notifier) clientTearDown() {
+	n.publishLock.Lock()
+	defer n.publishLock.Unlock()
+
 	n.serverStateLock.Lock()
 	n.currentClientState = false
-	n.lastNotification = stateDisconnecting
+	n.lastNotification = ClientStateDisconnecting
 	n.serverStateLock.Unlock()
 
-	n.notify(stateDisconnecting)
+	n.notify(ClientStateDisconnecting)
 }
 
-func (n *notifier) isServerStateChanged(newState int) bool {
+func (n *notifier) isServerStateChanged(newState ClientState) bool {
 	return n.lastNotification != newState
 }
 
-func (n *notifier) notify(state int) {
+func (n *notifier) notify(state ClientState) {
 	n.listenersLock.Lock()
 	listener := n.listener
 	n.listenersLock.Unlock()
@@ -109,20 +156,20 @@ func (n *notifier) notify(state int) {
 	notifyListener(listener, state)
 }
 
-func (n *notifier) calculateState(managementConn, signalConn bool) int {
+func (n *notifier) calculateState(managementConn, signalConn bool) ClientState {
 	if managementConn && signalConn {
-		return stateConnected
+		return ClientStateConnected
 	}
 
 	if !managementConn && !signalConn && !n.currentClientState {
-		return stateDisconnected
+		return ClientStateDisconnected
 	}
 
-	if n.lastNotification == stateDisconnecting {
-		return stateDisconnecting
+	if n.lastNotification == ClientStateDisconnecting {
+		return ClientStateDisconnecting
 	}
 
-	return stateConnecting
+	return ClientStateConnecting
 }
 
 func (n *notifier) peerListChanged(numOfPeers int) {
@@ -159,15 +206,19 @@ func (n *notifier) localAddressChanged(fqdn, address string) {
 	listener.OnAddressChanged(fqdn, address)
 }
 
-func notifyListener(l Listener, state int) {
+func notifyListener(l Listener, state ClientState) {
+	// legacy per-state callbacks; NoNetwork is delivered only via
+	// OnStateChanged below
 	switch state {
-	case stateDisconnected:
+	case ClientStateDisconnected:
 		l.OnDisconnected()
-	case stateConnected:
+	case ClientStateConnected:
 		l.OnConnected()
-	case stateConnecting:
+	case ClientStateConnecting:
 		l.OnConnecting()
-	case stateDisconnecting:
+	case ClientStateDisconnecting:
 		l.OnDisconnecting()
 	}
+
+	l.OnStateChanged(state)
 }

client/internal/peer/notifier_concurrent_test.go

@@ -0,0 +1,108 @@
+package peer
+
+import (
+	"sync"
+	"testing"
+	"time"
+)
+
+type recordingListener struct {
+	mu      sync.Mutex
+	states  []ClientState
+	onState func(ClientState)
+}
+
+func (l *recordingListener) OnStateChanged(state ClientState) {
+	l.mu.Lock()
+	l.states = append(l.states, state)
+	hook := l.onState
+	l.mu.Unlock()
+
+	if hook != nil {
+		hook(state)
+	}
+}
+
+func (l *recordingListener) last() (ClientState, bool) {
+	l.mu.Lock()
+	defer l.mu.Unlock()
+	if len(l.states) == 0 {
+		return 0, false
+	}
+	return l.states[len(l.states)-1], true
+}
+
+func (l *recordingListener) snapshot() []ClientState {
+	l.mu.Lock()
+	defer l.mu.Unlock()
+	return append([]ClientState(nil), l.states...)
+}
+
+func (l *recordingListener) OnConnected()                    {}
+func (l *recordingListener) OnDisconnected()                 {}
+func (l *recordingListener) OnConnecting()                   {}
+func (l *recordingListener) OnDisconnecting()                {}
+func (l *recordingListener) OnAddressChanged(string, string) {}
+func (l *recordingListener) OnPeersListChanged(int)          {}
+
+// TestNotifier_ConcurrentAvailabilityFlipOrdersPublication holds the first
+// transition inside the listener callback and flips availability again from
+// another goroutine while it is parked. The second flip must not publish
+// ahead of the one in flight, otherwise the listener ends up on a state the
+// notifier already superseded.
+func TestNotifier_ConcurrentAvailabilityFlipOrdersPublication(t *testing.T) {
+	n := newNotifier()
+	n.currentClientState = true
+	n.lastNotification = ClientStateConnecting
+
+	entered := make(chan struct{})
+	release := make(chan struct{})
+
+	l := &recordingListener{}
+	l.onState = func(state ClientState) {
+		if state != ClientStateNoNetwork {
+			return
+		}
+		l.mu.Lock()
+		l.onState = nil
+		l.mu.Unlock()
+		close(entered)
+		<-release
+	}
+	n.listener = l
+
+	var wg sync.WaitGroup
+	wg.Add(1)
+	go func() {
+		defer wg.Done()
+		n.setNetworkAvailable(false)
+	}()
+
+	<-entered
+
+	flipped := make(chan struct{})
+	go func() {
+		defer close(flipped)
+		n.setNetworkAvailable(true)
+	}()
+
+	select {
+	case <-flipped:
+		t.Fatal("the online transition published while the offline one was " +
+			"still in flight; publication is not serialized")
+	case <-time.After(200 * time.Millisecond):
+	}
+
+	close(release)
+	<-flipped
+	wg.Wait()
+
+	got, ok := l.last()
+	if !ok {
+		t.Fatal("listener never observed a state")
+	}
+	if got != ClientStateConnecting {
+		t.Fatalf("listener holds %v after the network came back, want Connecting; sequence: %v",
+			got, l.snapshot())
+	}
+}

client/internal/peer/notifier_test.go

@@ -6,29 +6,32 @@ import (
 )
 
 type mocListener struct {
-	lastState int
+	lastState ClientState
 	wg        sync.WaitGroup
 	peersWg   sync.WaitGroup
 	peers     int
 }
 
 func (l *mocListener) OnConnected() {
-	l.lastState = stateConnected
+	l.lastState = ClientStateConnected
 	l.wg.Done()
 }
 func (l *mocListener) OnDisconnected() {
-	l.lastState = stateDisconnected
+	l.lastState = ClientStateDisconnected
 	l.wg.Done()
 }
 func (l *mocListener) OnConnecting() {
-	l.lastState = stateConnecting
+	l.lastState = ClientStateConnecting
 	l.wg.Done()
 }
 func (l *mocListener) OnDisconnecting() {
-	l.lastState = stateDisconnecting
+	l.lastState = ClientStateDisconnecting
 	l.wg.Done()
 }
 
+func (l *mocListener) OnStateChanged(state ClientState) {
+
+}
 func (l *mocListener) OnAddressChanged(host, addr string) {
 
 }
@@ -57,15 +60,15 @@ func Test_notifier_serverState(t *testing.T) {
 
 	type scenario struct {
 		name        string
-		expected    int
+		expected    ClientState
 		mgmState    bool
 		signalState bool
 	}
 	scenarios := []scenario{
-		{"connected", stateConnected, true, true},
-		{"mgm down", stateConnecting, false, true},
-		{"signal down", stateConnecting, true, false},
-		{"disconnected", stateDisconnected, false, false},
+		{"connected", ClientStateConnected, true, true},
+		{"mgm down", ClientStateConnecting, false, true},
+		{"signal down", ClientStateConnecting, true, false},
+		{"disconnected", ClientStateDisconnected, false, false},
 	}
 
 	for _, tt := range scenarios {
@@ -85,7 +88,7 @@ func Test_notifier_SetListener(t *testing.T) {
 	listener.setPeersWaiter()
 
 	n := newNotifier()
-	n.lastNotification = stateConnecting
+	n.lastNotification = ClientStateConnecting
 	n.setListener(listener)
 	listener.wait()
 	listener.waitPeers()
@@ -99,7 +102,7 @@ func Test_notifier_RemoveListener(t *testing.T) {
 	listener.setWaiter()
 	listener.setPeersWaiter()
 	n := newNotifier()
-	n.lastNotification = stateConnecting
+	n.lastNotification = ClientStateConnecting
 	n.setListener(listener)
 	// setListener replays cached state on a goroutine; wait for both the state
 	// and peers callbacks to finish so we don't race on listener.peers.

client/internal/peer/status.go

@@ -1211,6 +1211,12 @@ func (d *Status) ClientTeardown() {
 	d.notifyStateChange()
 }
 
+// SetNetworkAvailable records the OS-reported network availability; while
+// unavailable, listeners see NoNetwork instead of Connecting.
+func (d *Status) SetNetworkAvailable(available bool) {
+	d.notifier.setNetworkAvailable(available)
+}
+
 // SetConnectionListener set a listener to the notifier
 func (d *Status) SetConnectionListener(listener Listener) {
 	d.notifier.setListener(listener)

client/internal/peer/worker_ice.go

@@ -255,8 +255,8 @@ func (w *WorkerICE) connect(ctx context.Context, agent *icemaker.ThreadSafeAgent
 		return
 	}
 
-	w.log.Debugf("turn agent dial")
-	remoteConn, err := w.turnAgentDial(ctx, agent, remoteOfferAnswer)
+	w.log.Debugf("agent dial")
+	remoteConn, err := w.agentDial(ctx, agent, remoteOfferAnswer)
 	if err != nil {
 		w.log.Debugf("failed to dial the remote peer: %s", err)
 		w.closeAgent(agent, w.agentDialerCancel)
@@ -389,6 +389,17 @@ func (w *WorkerICE) injectPortForwardedCandidate(srflxCandidate ice.Candidate) {
 		return
 	}
 
+	// A forwarded candidate only makes sense for an IPv4 mapping, which
+	// translates a port on the gateway's address. An IPv6 pinhole translates
+	// nothing: it unblocks the address ICE already gathers as a host candidate,
+	// so there is no second address to advertise. Injecting one here would also
+	// paste an IPv6 address onto whichever server-reflexive candidate arrived
+	// first, which is usually IPv4.
+	if mapping.ExternalIP != nil && mapping.ExternalIP.To4() == nil {
+		w.log.Debugf("skipping port-forwarded candidate: %s mapping is IPv6-only", mapping.NATType)
+		return
+	}
+
 	w.muxAgent.Lock()
 	if w.portForwardAttempted {
 		w.muxAgent.Unlock()
@@ -517,8 +528,8 @@ func (w *WorkerICE) onConnectionStateChange(agent *icemaker.ThreadSafeAgent, dia
 			w.logSuccessfulPaths(agent)
 			return
 		case ice.ConnectionStateFailed, ice.ConnectionStateDisconnected, ice.ConnectionStateClosed:
-			// ice.ConnectionStateClosed happens when we recreate the agent. For the P2P to TURN switch important to
-			// notify the conn.onICEStateDisconnected changes to update the current used priority
+			// ice.ConnectionStateClosed happens when we recreate the agent. The P2P to relay switch requires
+			// notifying conn.onICEStateDisconnected so it can update the currently used priority.
 
 			sessionChanged := w.closeAgent(agent, dialerCancel)
 
@@ -532,7 +543,7 @@ func (w *WorkerICE) onConnectionStateChange(agent *icemaker.ThreadSafeAgent, dia
 	}
 }
 
-func (w *WorkerICE) turnAgentDial(ctx context.Context, agent *icemaker.ThreadSafeAgent, remoteOfferAnswer *OfferAnswer) (*ice.Conn, error) {
+func (w *WorkerICE) agentDial(ctx context.Context, agent *icemaker.ThreadSafeAgent, remoteOfferAnswer *OfferAnswer) (*ice.Conn, error) {
 	if isController(w.config) {
 		return agent.Dial(ctx, remoteOfferAnswer.IceCredentials.UFrag, remoteOfferAnswer.IceCredentials.Pwd)
 	} else {

client/internal/portforward/manager.go

@@ -10,10 +10,8 @@ import (
 	"sync"
 	"time"
 
-	"github.com/libp2p/go-nat"
+	"github.com/netbirdio/go-nat"
 	log "github.com/sirupsen/logrus"
-
-	"github.com/netbirdio/netbird/client/internal/portforward/pcp"
 )
 
 const (
@@ -168,6 +166,11 @@ func (m *Manager) setup(ctx context.Context) (nat.NAT, *Mapping, error) {
 	if err != nil {
 		return nil, nil, fmt.Errorf("create port mapping: %w", err)
 	}
+
+	// Only meaningful once a mapping has been attempted: that is what opens the
+	// pinhole and records its outcome.
+	logIPv6Pinhole(gateway)
+
 	return gateway, mapping, nil
 }
 
@@ -265,7 +268,9 @@ func (m *Manager) checkHealthAndRecreate(ctx context.Context, gateway nat.NAT) b
 		return false
 	}
 
-	pcpNAT, ok := gateway.(*pcp.NAT)
+	// Assert on the interface, not on a concrete type: a dual-stack gateway is
+	// a wrapper around the IPv4 NAT, so a type assertion misses it.
+	checker, ok := gateway.(nat.HealthChecker)
 	if !ok {
 		return false
 	}
@@ -273,7 +278,7 @@ func (m *Manager) checkHealthAndRecreate(ctx context.Context, gateway nat.NAT) b
 	ctx, cancel := context.WithTimeout(ctx, 10*time.Second)
 	defer cancel()
 
-	epoch, serverRestarted, err := pcpNAT.CheckServerHealth(ctx)
+	epoch, serverRestarted, err := checker.CheckServerHealth(ctx)
 	if err != nil {
 		log.Debugf("PCP health check failed: %v", err)
 		return false
@@ -340,3 +345,18 @@ func (m *Manager) startTearDown(ctx context.Context) {
 func isPermanentLeaseRequired(err error) bool {
 	return err != nil && upnpErrPermanentLeaseOnly.MatchString(err.Error())
 }
+
+// logIPv6Pinhole reports the outcome of the IPv6 pinhole. Pinholes are best
+// effort and never fail a mapping on their own, so this is the only way to see
+// whether one was actually opened.
+func logIPv6Pinhole(gateway nat.NAT) {
+	reporter, ok := gateway.(nat.IPv6PinholeReporter)
+	if !ok {
+		return
+	}
+	if err := reporter.IPv6PinholeError(); err != nil {
+		log.Warnf("IPv6 pinhole: %v", err)
+		return
+	}
+	log.Infof("IPv6 pinhole open")
+}

client/internal/portforward/pcp/client.go

@@ -1,408 +0,0 @@
-package pcp
-
-import (
-	"context"
-	"crypto/rand"
-	"errors"
-	"fmt"
-	"net"
-	"net/netip"
-	"sync"
-	"time"
-
-	log "github.com/sirupsen/logrus"
-)
-
-const (
-	defaultTimeout     = 3 * time.Second
-	responseBufferSize = 128
-
-	// RFC 6887 Section 8.1.1 retry timing
-	initialRetryDelay = 3 * time.Second
-	maxRetryDelay     = 1024 * time.Second
-	maxRetries        = 4 // 3s + 6s + 12s + 24s = 45s total worst case
-)
-
-// Client is a PCP protocol client.
-// All methods are safe for concurrent use.
-type Client struct {
-	gateway netip.Addr
-	timeout time.Duration
-
-	mu sync.Mutex
-	// localIP caches the resolved local IP address.
-	localIP netip.Addr
-	// lastEpoch is the last observed server epoch value.
-	lastEpoch uint32
-	// epochTime tracks when lastEpoch was received for state loss detection.
-	epochTime time.Time
-	// externalIP caches the external IP from the last successful MAP response.
-	externalIP netip.Addr
-	// epochStateLost is set when epoch indicates server restart.
-	epochStateLost bool
-}
-
-// NewClient creates a new PCP client for the gateway at the given IP.
-func NewClient(gateway net.IP) *Client {
-	addr, ok := netip.AddrFromSlice(gateway)
-	if !ok {
-		log.Debugf("invalid gateway IP: %v", gateway)
-	}
-	return &Client{
-		gateway: addr.Unmap(),
-		timeout: defaultTimeout,
-	}
-}
-
-// NewClientWithTimeout creates a new PCP client with a custom timeout.
-func NewClientWithTimeout(gateway net.IP, timeout time.Duration) *Client {
-	addr, ok := netip.AddrFromSlice(gateway)
-	if !ok {
-		log.Debugf("invalid gateway IP: %v", gateway)
-	}
-	return &Client{
-		gateway: addr.Unmap(),
-		timeout: timeout,
-	}
-}
-
-// SetLocalIP sets the local IP address to use in PCP requests.
-func (c *Client) SetLocalIP(ip net.IP) {
-	addr, ok := netip.AddrFromSlice(ip)
-	if !ok {
-		log.Debugf("invalid local IP: %v", ip)
-	}
-	c.mu.Lock()
-	c.localIP = addr.Unmap()
-	c.mu.Unlock()
-}
-
-// Gateway returns the gateway IP address.
-func (c *Client) Gateway() net.IP {
-	return c.gateway.AsSlice()
-}
-
-// Announce sends a PCP ANNOUNCE request to discover PCP support.
-// Returns the server's epoch time on success.
-func (c *Client) Announce(ctx context.Context) (epoch uint32, err error) {
-	localIP, err := c.getLocalIP()
-	if err != nil {
-		return 0, fmt.Errorf("get local IP: %w", err)
-	}
-
-	req := buildAnnounceRequest(localIP)
-	resp, err := c.sendRequest(ctx, req)
-	if err != nil {
-		return 0, fmt.Errorf("send announce: %w", err)
-	}
-
-	parsed, err := parseResponse(resp)
-	if err != nil {
-		return 0, fmt.Errorf("parse announce response: %w", err)
-	}
-
-	if parsed.ResultCode != ResultSuccess {
-		return 0, fmt.Errorf("PCP ANNOUNCE failed: %s", ResultCodeString(parsed.ResultCode))
-	}
-
-	c.mu.Lock()
-	if c.updateEpochLocked(parsed.Epoch) {
-		log.Warnf("PCP server epoch indicates state loss - mappings may need refresh")
-	}
-	c.mu.Unlock()
-	return parsed.Epoch, nil
-}
-
-// AddPortMapping requests a port mapping from the PCP server.
-func (c *Client) AddPortMapping(ctx context.Context, protocol string, internalPort int, lifetime time.Duration) (*MapResponse, error) {
-	return c.addPortMappingWithHint(ctx, protocol, internalPort, internalPort, netip.Addr{}, lifetime)
-}
-
-// AddPortMappingWithHint requests a port mapping with suggested external port and IP.
-// Use lifetime <= 0 to delete a mapping.
-func (c *Client) AddPortMappingWithHint(ctx context.Context, protocol string, internalPort, suggestedExtPort int, suggestedExtIP net.IP, lifetime time.Duration) (*MapResponse, error) {
-	var extIP netip.Addr
-	if suggestedExtIP != nil {
-		var ok bool
-		extIP, ok = netip.AddrFromSlice(suggestedExtIP)
-		if !ok {
-			log.Debugf("invalid suggested external IP: %v", suggestedExtIP)
-		}
-		extIP = extIP.Unmap()
-	}
-	return c.addPortMappingWithHint(ctx, protocol, internalPort, suggestedExtPort, extIP, lifetime)
-}
-
-func (c *Client) addPortMappingWithHint(ctx context.Context, protocol string, internalPort, suggestedExtPort int, suggestedExtIP netip.Addr, lifetime time.Duration) (*MapResponse, error) {
-	localIP, err := c.getLocalIP()
-	if err != nil {
-		return nil, fmt.Errorf("get local IP: %w", err)
-	}
-
-	proto, err := protocolNumber(protocol)
-	if err != nil {
-		return nil, fmt.Errorf("parse protocol: %w", err)
-	}
-
-	var nonce [12]byte
-	if _, err := rand.Read(nonce[:]); err != nil {
-		return nil, fmt.Errorf("generate nonce: %w", err)
-	}
-
-	// Convert lifetime to seconds. Lifetime 0 means delete, so only apply
-	// default for positive durations that round to 0 seconds.
-	var lifetimeSec uint32
-	if lifetime > 0 {
-		lifetimeSec = uint32(lifetime.Seconds())
-		if lifetimeSec == 0 {
-			lifetimeSec = DefaultLifetime
-		}
-	}
-
-	req := buildMapRequest(localIP, nonce, proto, uint16(internalPort), uint16(suggestedExtPort), suggestedExtIP, lifetimeSec)
-
-	resp, err := c.sendRequest(ctx, req)
-	if err != nil {
-		return nil, fmt.Errorf("send map request: %w", err)
-	}
-
-	mapResp, err := parseMapResponse(resp)
-	if err != nil {
-		return nil, fmt.Errorf("parse map response: %w", err)
-	}
-
-	if mapResp.Nonce != nonce {
-		return nil, fmt.Errorf("nonce mismatch in response")
-	}
-
-	if mapResp.Protocol != proto {
-		return nil, fmt.Errorf("protocol mismatch: requested %d, got %d", proto, mapResp.Protocol)
-	}
-	if mapResp.InternalPort != uint16(internalPort) {
-		return nil, fmt.Errorf("internal port mismatch: requested %d, got %d", internalPort, mapResp.InternalPort)
-	}
-
-	if mapResp.ResultCode != ResultSuccess {
-		return nil, &Error{
-			Code:    mapResp.ResultCode,
-			Message: ResultCodeString(mapResp.ResultCode),
-		}
-	}
-
-	c.mu.Lock()
-	if c.updateEpochLocked(mapResp.Epoch) {
-		log.Warnf("PCP server epoch indicates state loss - mappings may need refresh")
-	}
-	c.cacheExternalIPLocked(mapResp.ExternalIP)
-	c.mu.Unlock()
-	return mapResp, nil
-}
-
-// DeletePortMapping removes a port mapping by requesting zero lifetime.
-func (c *Client) DeletePortMapping(ctx context.Context, protocol string, internalPort int) error {
-	if _, err := c.addPortMappingWithHint(ctx, protocol, internalPort, 0, netip.Addr{}, 0); err != nil {
-		var pcpErr *Error
-		if errors.As(err, &pcpErr) && pcpErr.Code == ResultNotAuthorized {
-			return nil
-		}
-		return fmt.Errorf("delete mapping: %w", err)
-	}
-	return nil
-}
-
-// GetExternalAddress returns the external IP address.
-// First checks for a cached value from previous MAP responses.
-// If not cached, creates a short-lived mapping to discover the external IP.
-func (c *Client) GetExternalAddress(ctx context.Context) (net.IP, error) {
-	c.mu.Lock()
-	if c.externalIP.IsValid() {
-		ip := c.externalIP.AsSlice()
-		c.mu.Unlock()
-		return ip, nil
-	}
-	c.mu.Unlock()
-
-	// Use an ephemeral port in the dynamic range (49152-65535).
-	// Port 0 is not valid with UDP/TCP protocols per RFC 6887.
-	ephemeralPort := 49152 + int(uint16(time.Now().UnixNano()))%(65535-49152)
-
-	// Use minimal lifetime (1 second) for discovery.
-	resp, err := c.AddPortMapping(ctx, "udp", ephemeralPort, time.Second)
-	if err != nil {
-		return nil, fmt.Errorf("create temporary mapping: %w", err)
-	}
-
-	if err := c.DeletePortMapping(ctx, "udp", ephemeralPort); err != nil {
-		log.Debugf("cleanup temporary PCP mapping: %v", err)
-	}
-
-	return resp.ExternalIP.AsSlice(), nil
-}
-
-// LastEpoch returns the last observed server epoch value.
-// A decrease in epoch indicates the server may have restarted and mappings may be lost.
-func (c *Client) LastEpoch() uint32 {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	return c.lastEpoch
-}
-
-// EpochStateLost returns true if epoch state loss was detected and clears the flag.
-func (c *Client) EpochStateLost() bool {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	lost := c.epochStateLost
-	c.epochStateLost = false
-	return lost
-}
-
-// updateEpoch updates the epoch tracking and detects potential state loss.
-// Returns true if state loss was detected (server likely restarted).
-// Caller must hold c.mu.
-func (c *Client) updateEpochLocked(newEpoch uint32) bool {
-	now := time.Now()
-	stateLost := false
-
-	// RFC 6887 Section 8.5: Detect invalid epoch indicating server state loss.
-	// client_delta = time since last response
-	// server_delta = epoch change since last response
-	// Invalid if: client_delta+2 < server_delta - server_delta/16
-	//         OR: server_delta+2 < client_delta - client_delta/16
-	// The +2 handles quantization, /16 (6.25%) handles clock drift.
-	if !c.epochTime.IsZero() && c.lastEpoch > 0 {
-		clientDelta := uint32(now.Sub(c.epochTime).Seconds())
-		serverDelta := newEpoch - c.lastEpoch
-
-		// Check for epoch going backwards or jumping unexpectedly.
-		// Subtraction is safe: serverDelta/16 is always <= serverDelta.
-		if clientDelta+2 < serverDelta-(serverDelta/16) ||
-			serverDelta+2 < clientDelta-(clientDelta/16) {
-			stateLost = true
-			c.epochStateLost = true
-		}
-	}
-
-	c.lastEpoch = newEpoch
-	c.epochTime = now
-	return stateLost
-}
-
-// cacheExternalIP stores the external IP from a successful MAP response.
-// Caller must hold c.mu.
-func (c *Client) cacheExternalIPLocked(ip netip.Addr) {
-	if ip.IsValid() && !ip.IsUnspecified() {
-		c.externalIP = ip
-	}
-}
-
-// sendRequest sends a PCP request with retries per RFC 6887 Section 8.1.1.
-func (c *Client) sendRequest(ctx context.Context, req []byte) ([]byte, error) {
-	addr := &net.UDPAddr{IP: c.gateway.AsSlice(), Port: Port}
-
-	var lastErr error
-	delay := initialRetryDelay
-
-	for range maxRetries {
-		resp, err := c.sendOnce(ctx, addr, req)
-		if err == nil {
-			return resp, nil
-		}
-		lastErr = err
-
-		if ctx.Err() != nil {
-			return nil, ctx.Err()
-		}
-
-		// RFC 6887 Section 8.1.1: RT = (1 + RAND) * MIN(2 * RTprev, MRT)
-		// RAND is random between -0.1 and +0.1
-		select {
-		case <-ctx.Done():
-			return nil, ctx.Err()
-		case <-time.After(retryDelayWithJitter(delay)):
-		}
-		delay = min(delay*2, maxRetryDelay)
-	}
-
-	return nil, fmt.Errorf("PCP request failed after %d retries: %w", maxRetries, lastErr)
-}
-
-// retryDelayWithJitter applies RFC 6887 jitter: multiply by (1 + RAND) where RAND is [-0.1, +0.1].
-func retryDelayWithJitter(d time.Duration) time.Duration {
-	var b [1]byte
-	_, _ = rand.Read(b[:])
-	// Convert byte to range [-0.1, +0.1]: (b/255 * 0.2) - 0.1
-	jitter := (float64(b[0])/255.0)*0.2 - 0.1
-	return time.Duration(float64(d) * (1 + jitter))
-}
-
-func (c *Client) sendOnce(ctx context.Context, addr *net.UDPAddr, req []byte) ([]byte, error) {
-	// Use ListenUDP instead of DialUDP to validate response source address per RFC 6887 §8.3.
-	conn, err := net.ListenUDP("udp", nil)
-	if err != nil {
-		return nil, fmt.Errorf("listen: %w", err)
-	}
-	defer func() {
-		if err := conn.Close(); err != nil {
-			log.Debugf("close UDP connection: %v", err)
-		}
-	}()
-
-	timeout := c.timeout
-	if deadline, ok := ctx.Deadline(); ok {
-		if remaining := time.Until(deadline); remaining < timeout {
-			timeout = remaining
-		}
-	}
-
-	if err := conn.SetDeadline(time.Now().Add(timeout)); err != nil {
-		return nil, fmt.Errorf("set deadline: %w", err)
-	}
-
-	if _, err := conn.WriteToUDP(req, addr); err != nil {
-		return nil, fmt.Errorf("write: %w", err)
-	}
-
-	resp := make([]byte, responseBufferSize)
-	n, from, err := conn.ReadFromUDP(resp)
-	if err != nil {
-		return nil, fmt.Errorf("read: %w", err)
-	}
-
-	// RFC 6887 §8.3: Validate response came from expected PCP server.
-	if !from.IP.Equal(addr.IP) {
-		return nil, fmt.Errorf("response from unexpected source %s (expected %s)", from.IP, addr.IP)
-	}
-
-	return resp[:n], nil
-}
-
-func (c *Client) getLocalIP() (netip.Addr, error) {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-
-	if !c.localIP.IsValid() {
-		return netip.Addr{}, fmt.Errorf("local IP not set for gateway %s", c.gateway)
-	}
-	return c.localIP, nil
-}
-
-func protocolNumber(protocol string) (uint8, error) {
-	switch protocol {
-	case "udp", "UDP":
-		return ProtoUDP, nil
-	case "tcp", "TCP":
-		return ProtoTCP, nil
-	default:
-		return 0, fmt.Errorf("unsupported protocol: %s", protocol)
-	}
-}
-
-// Error represents a PCP error response.
-type Error struct {
-	Code    uint8
-	Message string
-}
-
-func (e *Error) Error() string {
-	return fmt.Sprintf("PCP error: %s (%d)", e.Message, e.Code)
-}

client/internal/portforward/pcp/client_test.go

@@ -1,187 +0,0 @@
-package pcp
-
-import (
-	"context"
-	"net"
-	"net/netip"
-	"testing"
-	"time"
-
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestAddrConversion(t *testing.T) {
-	tests := []struct {
-		name string
-		addr netip.Addr
-	}{
-		{"IPv4", netip.MustParseAddr("192.168.1.100")},
-		{"IPv4 loopback", netip.MustParseAddr("127.0.0.1")},
-		{"IPv6", netip.MustParseAddr("2001:db8::1")},
-		{"IPv6 loopback", netip.MustParseAddr("::1")},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			b16 := addrTo16(tt.addr)
-
-			recovered := addrFrom16(b16)
-			assert.Equal(t, tt.addr, recovered, "address should round-trip")
-		})
-	}
-}
-
-func TestBuildAnnounceRequest(t *testing.T) {
-	clientIP := netip.MustParseAddr("192.168.1.100")
-	req := buildAnnounceRequest(clientIP)
-
-	require.Len(t, req, headerSize)
-	assert.Equal(t, byte(Version), req[0], "version")
-	assert.Equal(t, byte(OpAnnounce), req[1], "opcode")
-
-	// Check client IP is properly encoded as IPv4-mapped IPv6
-	assert.Equal(t, byte(0xff), req[18], "IPv4-mapped prefix byte 10")
-	assert.Equal(t, byte(0xff), req[19], "IPv4-mapped prefix byte 11")
-	assert.Equal(t, byte(192), req[20], "IP octet 1")
-	assert.Equal(t, byte(168), req[21], "IP octet 2")
-	assert.Equal(t, byte(1), req[22], "IP octet 3")
-	assert.Equal(t, byte(100), req[23], "IP octet 4")
-}
-
-func TestBuildMapRequest(t *testing.T) {
-	clientIP := netip.MustParseAddr("192.168.1.100")
-	nonce := [12]byte{1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12}
-	req := buildMapRequest(clientIP, nonce, ProtoUDP, 51820, 51820, netip.Addr{}, 3600)
-
-	require.Len(t, req, mapRequestSize)
-	assert.Equal(t, byte(Version), req[0], "version")
-	assert.Equal(t, byte(OpMap), req[1], "opcode")
-
-	// Lifetime at bytes 4-7
-	assert.Equal(t, uint32(3600), (uint32(req[4])<<24)|(uint32(req[5])<<16)|(uint32(req[6])<<8)|uint32(req[7]), "lifetime")
-
-	// Nonce at bytes 24-35
-	assert.Equal(t, nonce[:], req[24:36], "nonce")
-
-	// Protocol at byte 36
-	assert.Equal(t, byte(ProtoUDP), req[36], "protocol")
-
-	// Internal port at bytes 40-41
-	assert.Equal(t, uint16(51820), (uint16(req[40])<<8)|uint16(req[41]), "internal port")
-
-	// External port at bytes 42-43
-	assert.Equal(t, uint16(51820), (uint16(req[42])<<8)|uint16(req[43]), "external port")
-}
-
-func TestParseResponse(t *testing.T) {
-	// Construct a valid ANNOUNCE response
-	resp := make([]byte, headerSize)
-	resp[0] = Version
-	resp[1] = OpAnnounce | OpReply
-	// Result code = 0 (success)
-	// Lifetime = 0
-	// Epoch = 12345
-	resp[8] = 0
-	resp[9] = 0
-	resp[10] = 0x30
-	resp[11] = 0x39
-
-	parsed, err := parseResponse(resp)
-	require.NoError(t, err)
-	assert.Equal(t, uint8(Version), parsed.Version)
-	assert.Equal(t, uint8(OpAnnounce|OpReply), parsed.Opcode)
-	assert.Equal(t, uint8(ResultSuccess), parsed.ResultCode)
-	assert.Equal(t, uint32(12345), parsed.Epoch)
-}
-
-func TestParseResponseErrors(t *testing.T) {
-	t.Run("too short", func(t *testing.T) {
-		_, err := parseResponse([]byte{1, 2, 3})
-		assert.Error(t, err)
-	})
-
-	t.Run("wrong version", func(t *testing.T) {
-		resp := make([]byte, headerSize)
-		resp[0] = 1 // Wrong version
-		resp[1] = OpReply
-		_, err := parseResponse(resp)
-		assert.Error(t, err)
-	})
-
-	t.Run("missing reply bit", func(t *testing.T) {
-		resp := make([]byte, headerSize)
-		resp[0] = Version
-		resp[1] = OpAnnounce // Missing OpReply bit
-		_, err := parseResponse(resp)
-		assert.Error(t, err)
-	})
-}
-
-func TestResultCodeString(t *testing.T) {
-	assert.Equal(t, "SUCCESS", ResultCodeString(ResultSuccess))
-	assert.Equal(t, "NOT_AUTHORIZED", ResultCodeString(ResultNotAuthorized))
-	assert.Equal(t, "ADDRESS_MISMATCH", ResultCodeString(ResultAddressMismatch))
-	assert.Contains(t, ResultCodeString(255), "UNKNOWN")
-}
-
-func TestProtocolNumber(t *testing.T) {
-	proto, err := protocolNumber("udp")
-	require.NoError(t, err)
-	assert.Equal(t, uint8(ProtoUDP), proto)
-
-	proto, err = protocolNumber("tcp")
-	require.NoError(t, err)
-	assert.Equal(t, uint8(ProtoTCP), proto)
-
-	proto, err = protocolNumber("UDP")
-	require.NoError(t, err)
-	assert.Equal(t, uint8(ProtoUDP), proto)
-
-	_, err = protocolNumber("icmp")
-	assert.Error(t, err)
-}
-
-func TestClientCreation(t *testing.T) {
-	gateway := netip.MustParseAddr("192.168.1.1").AsSlice()
-
-	client := NewClient(gateway)
-	assert.Equal(t, net.IP(gateway), client.Gateway())
-	assert.Equal(t, defaultTimeout, client.timeout)
-
-	clientWithTimeout := NewClientWithTimeout(gateway, 5*time.Second)
-	assert.Equal(t, 5*time.Second, clientWithTimeout.timeout)
-}
-
-func TestNATType(t *testing.T) {
-	n := NewNAT(netip.MustParseAddr("192.168.1.1").AsSlice(), netip.MustParseAddr("192.168.1.100").AsSlice())
-	assert.Equal(t, "PCP", n.Type())
-}
-
-// Integration test - skipped unless PCP_TEST_GATEWAY env is set
-func TestClientIntegration(t *testing.T) {
-	t.Skip("Integration test - run manually with PCP_TEST_GATEWAY=<gateway-ip>")
-
-	gateway := netip.MustParseAddr("10.0.1.1").AsSlice()   // Change to your test gateway
-	localIP := netip.MustParseAddr("10.0.1.100").AsSlice() // Change to your local IP
-
-	client := NewClient(gateway)
-	client.SetLocalIP(localIP)
-	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
-	defer cancel()
-
-	// Test ANNOUNCE
-	epoch, err := client.Announce(ctx)
-	require.NoError(t, err)
-	t.Logf("Server epoch: %d", epoch)
-
-	// Test MAP
-	resp, err := client.AddPortMapping(ctx, "udp", 51820, 1*time.Hour)
-	require.NoError(t, err)
-	t.Logf("Mapping: internal=%d external=%d externalIP=%s",
-		resp.InternalPort, resp.ExternalPort, resp.ExternalIP)
-
-	// Cleanup
-	err = client.DeletePortMapping(ctx, "udp", 51820)
-	require.NoError(t, err)
-}

client/internal/portforward/pcp/nat.go

@@ -1,222 +0,0 @@
-package pcp
-
-import (
-	"context"
-	"fmt"
-	"net"
-	"net/netip"
-	"runtime"
-	"sync"
-	"time"
-
-	log "github.com/sirupsen/logrus"
-
-	"github.com/libp2p/go-nat"
-	"github.com/libp2p/go-netroute"
-)
-
-var _ nat.NAT = (*NAT)(nil)
-
-// NAT implements the go-nat NAT interface using PCP.
-// Supports dual-stack (IPv4 and IPv6) when available.
-// All methods are safe for concurrent use.
-//
-// TODO: IPv6 pinholes use the local IPv6 address. If the address changes
-// (e.g., due to SLAAC rotation or network change), the pinhole becomes stale
-// and needs to be recreated with the new address.
-type NAT struct {
-	client *Client
-
-	mu sync.RWMutex
-	// client6 is the IPv6 PCP client, nil if IPv6 is unavailable.
-	client6 *Client
-	// localIP6 caches the local IPv6 address used for PCP requests.
-	localIP6 netip.Addr
-}
-
-// NewNAT creates a new NAT instance backed by PCP.
-func NewNAT(gateway, localIP net.IP) *NAT {
-	client := NewClient(gateway)
-	client.SetLocalIP(localIP)
-	return &NAT{
-		client: client,
-	}
-}
-
-// Type returns "PCP" as the NAT type.
-func (n *NAT) Type() string {
-	return "PCP"
-}
-
-// GetDeviceAddress returns the gateway IP address.
-func (n *NAT) GetDeviceAddress() (net.IP, error) {
-	return n.client.Gateway(), nil
-}
-
-// GetExternalAddress returns the external IP address.
-func (n *NAT) GetExternalAddress() (net.IP, error) {
-	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
-	defer cancel()
-	return n.client.GetExternalAddress(ctx)
-}
-
-// GetInternalAddress returns the local IP address used to communicate with the gateway.
-func (n *NAT) GetInternalAddress() (net.IP, error) {
-	addr, err := n.client.getLocalIP()
-	if err != nil {
-		return nil, err
-	}
-	return addr.AsSlice(), nil
-}
-
-// AddPortMapping creates a port mapping on both IPv4 and IPv6 (if available).
-func (n *NAT) AddPortMapping(ctx context.Context, protocol string, internalPort int, _ string, timeout time.Duration) (int, error) {
-	resp, err := n.client.AddPortMapping(ctx, protocol, internalPort, timeout)
-	if err != nil {
-		return 0, fmt.Errorf("add mapping: %w", err)
-	}
-
-	n.mu.RLock()
-	client6 := n.client6
-	localIP6 := n.localIP6
-	n.mu.RUnlock()
-
-	if client6 == nil {
-		return int(resp.ExternalPort), nil
-	}
-
-	if _, err := client6.AddPortMapping(ctx, protocol, internalPort, timeout); err != nil {
-		log.Warnf("IPv6 PCP mapping failed (continuing with IPv4): %v", err)
-		return int(resp.ExternalPort), nil
-	}
-
-	log.Infof("created IPv6 PCP pinhole: %s:%d", localIP6, internalPort)
-	return int(resp.ExternalPort), nil
-}
-
-// DeletePortMapping removes a port mapping from both IPv4 and IPv6.
-func (n *NAT) DeletePortMapping(ctx context.Context, protocol string, internalPort int) error {
-	err := n.client.DeletePortMapping(ctx, protocol, internalPort)
-
-	n.mu.RLock()
-	client6 := n.client6
-	n.mu.RUnlock()
-
-	if client6 != nil {
-		if err6 := client6.DeletePortMapping(ctx, protocol, internalPort); err6 != nil {
-			log.Warnf("IPv6 PCP delete mapping failed: %v", err6)
-		}
-	}
-
-	if err != nil {
-		return fmt.Errorf("delete mapping: %w", err)
-	}
-	return nil
-}
-
-// CheckServerHealth sends an ANNOUNCE to verify the server is still responsive.
-// Returns the current epoch and whether the server may have restarted (epoch state loss detected).
-func (n *NAT) CheckServerHealth(ctx context.Context) (epoch uint32, serverRestarted bool, err error) {
-	epoch, err = n.client.Announce(ctx)
-	if err != nil {
-		return 0, false, fmt.Errorf("announce: %w", err)
-	}
-	return epoch, n.client.EpochStateLost(), nil
-}
-
-// DiscoverPCP attempts to discover a PCP-capable gateway.
-// Returns a NAT interface if PCP is supported, or an error otherwise.
-// Discovers both IPv4 and IPv6 gateways when available.
-func DiscoverPCP(ctx context.Context) (nat.NAT, error) {
-	gateway, localIP, err := getDefaultGateway()
-	if err != nil {
-		return nil, fmt.Errorf("get default gateway: %w", err)
-	}
-
-	client := NewClient(gateway)
-	client.SetLocalIP(localIP)
-	if _, err := client.Announce(ctx); err != nil {
-		return nil, fmt.Errorf("PCP announce: %w", err)
-	}
-
-	result := &NAT{client: client}
-	discoverIPv6(ctx, result)
-
-	return result, nil
-}
-
-func discoverIPv6(ctx context.Context, result *NAT) {
-	gateway6, localIP6, err := getDefaultGateway6()
-	if err != nil {
-		log.Debugf("IPv6 gateway discovery failed: %v", err)
-		return
-	}
-
-	client6 := NewClient(gateway6)
-	client6.SetLocalIP(localIP6)
-	if _, err := client6.Announce(ctx); err != nil {
-		log.Debugf("PCP IPv6 announce failed: %v", err)
-		return
-	}
-
-	addr, ok := netip.AddrFromSlice(localIP6)
-	if !ok {
-		log.Debugf("invalid IPv6 local IP: %v", localIP6)
-		return
-	}
-	result.mu.Lock()
-	result.client6 = client6
-	result.localIP6 = addr
-	result.mu.Unlock()
-	log.Debugf("PCP IPv6 gateway discovered: %s (local: %s)", gateway6, localIP6)
-}
-
-// getDefaultGateway returns the default IPv4 gateway and local IP using the system routing table.
-func getDefaultGateway() (gateway net.IP, localIP net.IP, err error) {
-	router, err := netroute.New()
-	if err != nil {
-		return nil, nil, err
-	}
-
-	dst := net.IPv4zero
-	if runtime.GOOS == "linux" || runtime.GOOS == "android" {
-		// go-netroute v0.4.0 rejects unspecified destinations client-side on Linux/Android.
-		// TODO: on android/ios, use platform APIs (ConnectivityManager.getLinkProperties /
-		// NWPathMonitor) when netlink-based lookup is restricted or unavailable.
-		dst = net.IPv4(0, 0, 0, 1)
-	}
-	_, gateway, localIP, err = router.Route(dst)
-	if err != nil {
-		return nil, nil, err
-	}
-
-	if gateway == nil {
-		return nil, nil, nat.ErrNoNATFound
-	}
-
-	return gateway, localIP, nil
-}
-
-// getDefaultGateway6 returns the default IPv6 gateway IP address using the system routing table.
-func getDefaultGateway6() (gateway net.IP, localIP net.IP, err error) {
-	router, err := netroute.New()
-	if err != nil {
-		return nil, nil, err
-	}
-
-	dst := net.IPv6zero
-	if runtime.GOOS == "linux" || runtime.GOOS == "android" {
-		// ::2
-		dst = net.IP{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2}
-	}
-	_, gateway, localIP, err = router.Route(dst)
-	if err != nil {
-		return nil, nil, err
-	}
-
-	if gateway == nil {
-		return nil, nil, nat.ErrNoNATFound
-	}
-
-	return gateway, localIP, nil
-}

client/internal/portforward/pcp/protocol.go

@@ -1,225 +0,0 @@
-// Package pcp implements the Port Control Protocol (RFC 6887).
-//
-// # Implemented Features
-//
-//   - ANNOUNCE opcode: Discovers PCP server support
-//   - MAP opcode: Creates/deletes port mappings (IPv4 NAT) and firewall pinholes (IPv6)
-//   - Dual-stack: Simultaneous IPv4 and IPv6 support via separate clients
-//   - Nonce validation: Prevents response spoofing
-//   - Epoch tracking: Detects server restarts per Section 8.5
-//   - RFC-compliant retry timing: 3s initial, exponential backoff to 1024s max (Section 8.1.1)
-//
-// # Not Implemented
-//
-//   - PEER opcode: For outbound peer connections (not needed for inbound NAT traversal)
-//   - THIRD_PARTY option: For managing mappings on behalf of other devices
-//   - PREFER_FAILURE option: Requires exact external port or fail (IPv4 NAT only, not needed for IPv6 pinholing)
-//   - FILTER option: To restrict remote peer addresses
-//
-// These optional features are omitted because the primary use case is simple
-// port forwarding for WireGuard, which only requires MAP with default behavior.
-package pcp
-
-import (
-	"encoding/binary"
-	"fmt"
-	"net/netip"
-)
-
-const (
-	// Version is the PCP protocol version (RFC 6887).
-	Version = 2
-
-	// Port is the standard PCP server port.
-	Port = 5351
-
-	// DefaultLifetime is the default requested mapping lifetime in seconds.
-	DefaultLifetime = 7200 // 2 hours
-
-	// Header sizes
-	headerSize     = 24
-	mapPayloadSize = 36
-	mapRequestSize = headerSize + mapPayloadSize // 60 bytes
-)
-
-// Opcodes
-const (
-	OpAnnounce = 0
-	OpMap      = 1
-	OpPeer     = 2
-	OpReply    = 0x80 // OR'd with opcode in responses
-)
-
-// Protocol numbers for MAP requests
-const (
-	ProtoUDP = 17
-	ProtoTCP = 6
-)
-
-// Result codes (RFC 6887 Section 7.4)
-const (
-	ResultSuccess              = 0
-	ResultUnsuppVersion        = 1
-	ResultNotAuthorized        = 2
-	ResultMalformedRequest     = 3
-	ResultUnsuppOpcode         = 4
-	ResultUnsuppOption         = 5
-	ResultMalformedOption      = 6
-	ResultNetworkFailure       = 7
-	ResultNoResources          = 8
-	ResultUnsuppProtocol       = 9
-	ResultUserExQuota          = 10
-	ResultCannotProvideExt     = 11
-	ResultAddressMismatch      = 12
-	ResultExcessiveRemotePeers = 13
-)
-
-// ResultCodeString returns a human-readable string for a result code.
-func ResultCodeString(code uint8) string {
-	switch code {
-	case ResultSuccess:
-		return "SUCCESS"
-	case ResultUnsuppVersion:
-		return "UNSUPP_VERSION"
-	case ResultNotAuthorized:
-		return "NOT_AUTHORIZED"
-	case ResultMalformedRequest:
-		return "MALFORMED_REQUEST"
-	case ResultUnsuppOpcode:
-		return "UNSUPP_OPCODE"
-	case ResultUnsuppOption:
-		return "UNSUPP_OPTION"
-	case ResultMalformedOption:
-		return "MALFORMED_OPTION"
-	case ResultNetworkFailure:
-		return "NETWORK_FAILURE"
-	case ResultNoResources:
-		return "NO_RESOURCES"
-	case ResultUnsuppProtocol:
-		return "UNSUPP_PROTOCOL"
-	case ResultUserExQuota:
-		return "USER_EX_QUOTA"
-	case ResultCannotProvideExt:
-		return "CANNOT_PROVIDE_EXTERNAL"
-	case ResultAddressMismatch:
-		return "ADDRESS_MISMATCH"
-	case ResultExcessiveRemotePeers:
-		return "EXCESSIVE_REMOTE_PEERS"
-	default:
-		return fmt.Sprintf("UNKNOWN(%d)", code)
-	}
-}
-
-// Response represents a parsed PCP response header.
-type Response struct {
-	Version    uint8
-	Opcode     uint8
-	ResultCode uint8
-	Lifetime   uint32
-	Epoch      uint32
-}
-
-// MapResponse contains the full response to a MAP request.
-type MapResponse struct {
-	Response
-	Nonce        [12]byte
-	Protocol     uint8
-	InternalPort uint16
-	ExternalPort uint16
-	ExternalIP   netip.Addr
-}
-
-// addrTo16 converts an address to its 16-byte IPv4-mapped IPv6 representation.
-func addrTo16(addr netip.Addr) [16]byte {
-	if addr.Is4() {
-		return netip.AddrFrom4(addr.As4()).As16()
-	}
-	return addr.As16()
-}
-
-// addrFrom16 extracts an address from a 16-byte representation, unmapping IPv4.
-func addrFrom16(b [16]byte) netip.Addr {
-	return netip.AddrFrom16(b).Unmap()
-}
-
-// buildAnnounceRequest creates a PCP ANNOUNCE request packet.
-func buildAnnounceRequest(clientIP netip.Addr) []byte {
-	req := make([]byte, headerSize)
-	req[0] = Version
-	req[1] = OpAnnounce
-	mapped := addrTo16(clientIP)
-	copy(req[8:24], mapped[:])
-	return req
-}
-
-// buildMapRequest creates a PCP MAP request packet.
-func buildMapRequest(clientIP netip.Addr, nonce [12]byte, protocol uint8, internalPort, suggestedExtPort uint16, suggestedExtIP netip.Addr, lifetime uint32) []byte {
-	req := make([]byte, mapRequestSize)
-
-	// Header
-	req[0] = Version
-	req[1] = OpMap
-	binary.BigEndian.PutUint32(req[4:8], lifetime)
-	mapped := addrTo16(clientIP)
-	copy(req[8:24], mapped[:])
-
-	// MAP payload
-	copy(req[24:36], nonce[:])
-	req[36] = protocol
-	binary.BigEndian.PutUint16(req[40:42], internalPort)
-	binary.BigEndian.PutUint16(req[42:44], suggestedExtPort)
-	if suggestedExtIP.IsValid() {
-		extMapped := addrTo16(suggestedExtIP)
-		copy(req[44:60], extMapped[:])
-	}
-
-	return req
-}
-
-// parseResponse parses the common PCP response header.
-func parseResponse(data []byte) (*Response, error) {
-	if len(data) < headerSize {
-		return nil, fmt.Errorf("response too short: %d bytes", len(data))
-	}
-
-	resp := &Response{
-		Version:    data[0],
-		Opcode:     data[1],
-		ResultCode: data[3], // Byte 2 is reserved, byte 3 is result code (RFC 6887 §7.2)
-		Lifetime:   binary.BigEndian.Uint32(data[4:8]),
-		Epoch:      binary.BigEndian.Uint32(data[8:12]),
-	}
-
-	if resp.Version != Version {
-		return nil, fmt.Errorf("unsupported PCP version: %d", resp.Version)
-	}
-
-	if resp.Opcode&OpReply == 0 {
-		return nil, fmt.Errorf("response missing reply bit: opcode=0x%02x", resp.Opcode)
-	}
-
-	return resp, nil
-}
-
-// parseMapResponse parses a complete MAP response.
-func parseMapResponse(data []byte) (*MapResponse, error) {
-	if len(data) < mapRequestSize {
-		return nil, fmt.Errorf("MAP response too short: %d bytes", len(data))
-	}
-
-	resp, err := parseResponse(data)
-	if err != nil {
-		return nil, fmt.Errorf("parse header: %w", err)
-	}
-
-	mapResp := &MapResponse{
-		Response:     *resp,
-		Protocol:     data[36],
-		InternalPort: binary.BigEndian.Uint16(data[40:42]),
-		ExternalPort: binary.BigEndian.Uint16(data[42:44]),
-		ExternalIP:   addrFrom16([16]byte(data[44:60])),
-	}
-	copy(mapResp.Nonce[:], data[24:36])
-
-	return mapResp, nil
-}

client/internal/portforward/pinhole_test.go

@@ -0,0 +1,116 @@
+//go:build !js
+
+package portforward
+
+import (
+	"context"
+	"errors"
+	"strings"
+	"testing"
+
+	"github.com/netbirdio/go-nat"
+	log "github.com/sirupsen/logrus"
+	"github.com/sirupsen/logrus/hooks/test"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// mockPinholeNAT is a gateway that also reports an IPv6 pinhole outcome, the
+// shape a dual-stack gateway has.
+type mockPinholeNAT struct {
+	*mockNAT
+	pinholeErr error
+}
+
+func (m *mockPinholeNAT) IPv6PinholeError() error {
+	return m.pinholeErr
+}
+
+func TestSetupLogsPinholeOutcome(t *testing.T) {
+	pinholeErr := errors.New("pcp ipv6: NOT_AUTHORIZED")
+
+	tests := []struct {
+		name       string
+		pinholeErr error
+		mappingErr error
+		wantLevel  log.Level
+		wantText   string
+	}{
+		{
+			name:      "an open pinhole is reported",
+			wantLevel: log.InfoLevel,
+			wantText:  "IPv6 pinhole open",
+		},
+		{
+			name: "a failed pinhole is reported without failing the mapping",
+			// The IPv4 mapping is what the caller asked for, so the pinhole
+			// failure surfaces only in the log.
+			pinholeErr: pinholeErr,
+			wantLevel:  log.WarnLevel,
+			wantText:   pinholeErr.Error(),
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			gateway := &mockPinholeNAT{mockNAT: newMockNAT(), pinholeErr: tt.pinholeErr}
+			hook := stubGatewayDiscovery(t, gateway)
+
+			m := NewManager()
+			m.wgPort = 51820
+
+			_, mapping, err := m.setup(context.Background())
+
+			require.NoError(t, err)
+			require.NotNil(t, mapping)
+
+			entry := findEntry(hook, tt.wantText)
+			require.NotNil(t, entry, "no log entry mentioning %q", tt.wantText)
+			assert.Equal(t, tt.wantLevel, entry.Level)
+		})
+	}
+
+	t.Run("a failed mapping reports no pinhole outcome", func(t *testing.T) {
+		// Nothing opened the pinhole, so whatever it currently reports says
+		// nothing about this attempt.
+		gateway := &mockPinholeNAT{mockNAT: newMockNAT()}
+		gateway.addMappingErr = errors.New("gateway refused")
+		hook := stubGatewayDiscovery(t, gateway)
+
+		m := NewManager()
+		m.wgPort = 51820
+
+		_, _, err := m.setup(context.Background())
+
+		require.Error(t, err)
+		assert.Nil(t, findEntry(hook, "IPv6 pinhole"))
+	})
+}
+
+// stubGatewayDiscovery makes discovery return gateway and captures log output.
+func stubGatewayDiscovery(t *testing.T, gateway nat.NAT) *test.Hook {
+	t.Helper()
+
+	orig := discoverGateway
+	discoverGateway = func(context.Context) (nat.NAT, error) { return gateway, nil }
+	t.Cleanup(func() { discoverGateway = orig })
+
+	hook := test.NewGlobal()
+	origLevel := log.GetLevel()
+	log.SetLevel(log.DebugLevel)
+	t.Cleanup(func() {
+		hook.Reset()
+		log.SetLevel(origLevel)
+	})
+
+	return hook
+}
+
+func findEntry(hook *test.Hook, substr string) *log.Entry {
+	for _, entry := range hook.AllEntries() {
+		if strings.Contains(entry.Message, substr) {
+			return entry
+		}
+	}
+	return nil
+}

client/internal/portforward/state.go

@@ -4,27 +4,94 @@ package portforward
 
 import (
 	"context"
+	"errors"
 	"fmt"
+	"time"
 
-	"github.com/libp2p/go-nat"
+	"github.com/netbirdio/go-nat"
+	"github.com/netbirdio/go-nat/pcp"
 	log "github.com/sirupsen/logrus"
-
-	"github.com/netbirdio/netbird/client/internal/portforward/pcp"
 )
 
 // discoverGateway is the function used for NAT gateway discovery.
 // It can be replaced in tests to avoid real network operations.
-// Tries PCP first, then falls back to NAT-PMP/UPnP.
 var discoverGateway = defaultDiscoverGateway
 
-func defaultDiscoverGateway(ctx context.Context) (nat.NAT, error) {
-	pcpGateway, err := pcp.DiscoverPCP(ctx)
-	if err == nil {
-		return pcpGateway, nil
-	}
-	log.Debugf("PCP discovery failed: %v, trying NAT-PMP/UPnP", err)
+// pinholeDiscoveryTimeout is the slice of the discovery budget held back for
+// the IPv6 pinhole probe.
+//
+// Sizing it is coarser than it looks: PCP retransmits on a 3s socket timeout
+// and a 3s first backoff, so a second attempt needs about 9s. Anything from
+// roughly 1s to 8s therefore buys exactly one attempt, and this only sets how
+// long that attempt waits. A PCP server sits on the local link and answers in
+// milliseconds, so 3s is margin rather than need, and the rest is left to
+// gateway discovery, whose multicast SSDP search alone takes 5s. A probe lost
+// to a dropped packet is retried by the next discovery round.
+//
+// It is a variable so tests can shorten it.
+var pinholeDiscoveryTimeout = 3 * time.Second
 
-	return nat.DiscoverGateway(ctx)
+// Discovery entry points, as variables so tests can drive the fallback without
+// touching the network.
+var (
+	discoverNATGateway = nat.DiscoverGateway
+
+	discoverPCPPinhole = func(ctx context.Context) (nat.NAT, error) {
+		pinhole, err := pcp.DiscoverPCP(ctx)
+		if err != nil {
+			return nil, err
+		}
+		return pinhole, nil
+	}
+)
+
+// defaultDiscoverGateway finds a gateway that can make the WireGuard port
+// reachable. DiscoverGateway prefers PCP for IPv4, races UPnP and NAT-PMP
+// behind it, and attaches an IPv6 pinhole independently of which IPv4 protocol
+// wins.
+//
+// It reports no gateway on a network offering only IPv6, having no IPv4 mapping
+// to attach a pinhole to. Such a network still needs one: there is no
+// translation to traverse, but the router drops inbound IPv6 until something
+// opens it. Fall back to PCP alone, which yields a gateway holding just the
+// pinhole.
+func defaultDiscoverGateway(ctx context.Context) (nat.NAT, error) {
+	gatewayCtx, cancel := reserveForPinhole(ctx)
+	defer cancel()
+
+	gateway, err := discoverNATGateway(gatewayCtx)
+	if err == nil {
+		return gateway, nil
+	}
+	if !errors.Is(err, nat.ErrNoNATFound) {
+		return nil, err
+	}
+
+	pinhole, pinholeErr := discoverPCPPinhole(ctx)
+	if pinholeErr != nil {
+		log.Debugf("no IPv6 pinhole after %v: %v", err, pinholeErr)
+		return nil, err
+	}
+
+	log.Infof("no IPv4 gateway, continuing with an IPv6 pinhole only")
+	return pinhole, nil
+}
+
+// reserveForPinhole shortens ctx so that a pinhole probe still has time to run
+// afterwards. Finding nothing takes gateway discovery everything it is given,
+// so on the unshortened context the probe would start already expired. A budget
+// too small to divide is left to gateway discovery, which is the likelier win.
+func reserveForPinhole(ctx context.Context) (context.Context, context.CancelFunc) {
+	deadline, ok := ctx.Deadline()
+	if !ok {
+		return context.WithCancel(ctx)
+	}
+
+	remaining := time.Until(deadline)
+	if remaining <= pinholeDiscoveryTimeout {
+		return context.WithCancel(ctx)
+	}
+	return context.WithTimeout(ctx, remaining-pinholeDiscoveryTimeout)
 }
 
 // State is persisted only for crash recovery cleanup

client/internal/portforward/state_test.go

@@ -0,0 +1,140 @@
+//go:build !js
+
+package portforward
+
+import (
+	"context"
+	"errors"
+	"testing"
+	"time"
+
+	"github.com/netbirdio/go-nat"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// stubDiscovery replaces both discovery entry points for the duration of a
+// test. gatewayDelay simulates gateway discovery spending everything it is
+// given before reporting that it found nothing.
+func stubDiscovery(t *testing.T, gateway nat.NAT, gatewayErr error, gatewayDelay time.Duration, pinhole nat.NAT, pinholeErr error) {
+	t.Helper()
+
+	origGateway, origPinhole := discoverNATGateway, discoverPCPPinhole
+	discoverNATGateway = func(ctx context.Context) (nat.NAT, error) {
+		if gatewayDelay > 0 {
+			select {
+			case <-time.After(gatewayDelay):
+			case <-ctx.Done():
+			}
+		}
+		return gateway, gatewayErr
+	}
+	discoverPCPPinhole = func(ctx context.Context) (nat.NAT, error) {
+		if err := ctx.Err(); err != nil {
+			return nil, err
+		}
+		return pinhole, pinholeErr
+	}
+
+	t.Cleanup(func() { discoverNATGateway, discoverPCPPinhole = origGateway, origPinhole })
+}
+
+func TestDefaultDiscoverGateway(t *testing.T) {
+	ipv4Gateway := &mockNAT{natType: "PCP+PCPv6"}
+	ipv6Pinhole := &mockNAT{natType: "PCP"}
+	otherErr := errors.New("routing table unavailable")
+
+	t.Run("an IPv4 gateway is used as is", func(t *testing.T) {
+		stubDiscovery(t, ipv4Gateway, nil, 0, ipv6Pinhole, nil)
+
+		got, err := defaultDiscoverGateway(context.Background())
+
+		require.NoError(t, err)
+		assert.Same(t, ipv4Gateway, got)
+	})
+
+	t.Run("no IPv4 gateway still opens an IPv6 pinhole", func(t *testing.T) {
+		stubDiscovery(t, nil, nat.ErrNoNATFound, 0, ipv6Pinhole, nil)
+
+		got, err := defaultDiscoverGateway(context.Background())
+
+		require.NoError(t, err)
+		assert.Same(t, ipv6Pinhole, got)
+	})
+
+	t.Run("no gateway and no pinhole reports the original failure", func(t *testing.T) {
+		stubDiscovery(t, nil, nat.ErrNoNATFound, 0, nil, errors.New("no IPv6 route"))
+
+		got, err := defaultDiscoverGateway(context.Background())
+
+		assert.Nil(t, got)
+		assert.ErrorIs(t, err, nat.ErrNoNATFound, "the pinhole failure must not mask why no gateway was found")
+	})
+
+	t.Run("a failure other than no-gateway is reported as is", func(t *testing.T) {
+		stubDiscovery(t, nil, otherErr, 0, ipv6Pinhole, nil)
+
+		got, err := defaultDiscoverGateway(context.Background())
+
+		assert.Nil(t, got)
+		assert.ErrorIs(t, err, otherErr)
+	})
+
+	t.Run("the pinhole survives gateway discovery using its whole budget", func(t *testing.T) {
+		// On one shared context the probe would start already expired, which is
+		// how this failed against a real gateway.
+		reserve := 50 * time.Millisecond
+		origReserve := pinholeDiscoveryTimeout
+		pinholeDiscoveryTimeout = reserve
+		t.Cleanup(func() { pinholeDiscoveryTimeout = origReserve })
+
+		budget := 4 * reserve
+		ctx, cancel := context.WithTimeout(context.Background(), budget)
+		defer cancel()
+
+		stubDiscovery(t, nil, nat.ErrNoNATFound, budget, ipv6Pinhole, nil)
+
+		got, err := defaultDiscoverGateway(ctx)
+
+		require.NoError(t, err)
+		assert.Same(t, ipv6Pinhole, got)
+	})
+}
+
+func TestReserveForPinhole(t *testing.T) {
+	origReserve := pinholeDiscoveryTimeout
+	pinholeDiscoveryTimeout = time.Second
+	t.Cleanup(func() { pinholeDiscoveryTimeout = origReserve })
+
+	t.Run("a budget is divided", func(t *testing.T) {
+		ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+		defer cancel()
+
+		gatewayCtx, cancelGateway := reserveForPinhole(ctx)
+		defer cancelGateway()
+
+		deadline, ok := gatewayCtx.Deadline()
+		require.True(t, ok)
+		assert.InDelta(t, 9*time.Second, time.Until(deadline), float64(500*time.Millisecond))
+	})
+
+	t.Run("a budget too small to divide is left whole", func(t *testing.T) {
+		ctx, cancel := context.WithTimeout(context.Background(), 500*time.Millisecond)
+		defer cancel()
+
+		gatewayCtx, cancelGateway := reserveForPinhole(ctx)
+		defer cancelGateway()
+
+		deadline, ok := gatewayCtx.Deadline()
+		require.True(t, ok)
+		assert.InDelta(t, 500*time.Millisecond, time.Until(deadline), float64(100*time.Millisecond))
+	})
+
+	t.Run("no deadline stays unbounded", func(t *testing.T) {
+		gatewayCtx, cancelGateway := reserveForPinhole(context.Background())
+		defer cancelGateway()
+
+		_, ok := gatewayCtx.Deadline()
+		assert.False(t, ok)
+	})
+}

client/internal/profilemanager/state.go

@@ -87,9 +87,10 @@ func (pm *ProfileManager) SetActiveProfileState(state *ProfileState) error {
 
 // RemoveProfileState deletes the per-profile state file (which holds the
 // account email used for the SSO login hint and the UI display). Called after
-// a successful logout so a logged-out profile no longer shows a stale account
-// email. The state file only stores the email, so deleting it is equivalent to
-// clearing it; the next SSO login recreates it. A missing file is not an error.
+// profile removal; logout keeps the file so the next login can pass the email
+// as the login_hint. The state file only stores the email, so deleting it is
+// equivalent to clearing it; the next SSO login recreates it. A missing file
+// is not an error.
 func (pm *ProfileManager) RemoveProfileState(profileName string) error {
 	configDir, err := getConfigDir()
 	if err != nil {

client/internal/routemanager/systemops/routeselection_windows_test.go

@@ -0,0 +1,82 @@
+//go:build windows
+
+package systemops
+
+import (
+	"math"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestSortRouteCandidates(t *testing.T) {
+	tests := []struct {
+		name       string
+		candidates []candidateRoute
+		wantOrder  []uint32
+	}{
+		{
+			name: "longest prefix wins over metrics",
+			candidates: []candidateRoute{
+				{interfaceIndex: 1, prefixLength: 0, routeMetric: 0, interfaceMetric: 5},
+				{interfaceIndex: 2, prefixLength: 24, routeMetric: 100, interfaceMetric: 50},
+			},
+			wantOrder: []uint32{2, 1},
+		},
+		{
+			// Windows ranks equal-length prefixes by route metric + interface metric,
+			// so a higher route metric on a low metric interface can still win.
+			name: "combined metric beats route metric alone",
+			candidates: []candidateRoute{
+				{interfaceIndex: 8, prefixLength: 0, routeMetric: 0, interfaceMetric: 100},
+				{interfaceIndex: 5, prefixLength: 0, routeMetric: 10, interfaceMetric: 5},
+			},
+			wantOrder: []uint32{5, 8},
+		},
+		{
+			name: "lower combined metric wins",
+			candidates: []candidateRoute{
+				{interfaceIndex: 5, prefixLength: 0, routeMetric: 300, interfaceMetric: 5},
+				{interfaceIndex: 8, prefixLength: 0, routeMetric: 0, interfaceMetric: 100},
+			},
+			wantOrder: []uint32{8, 5},
+		},
+		{
+			name: "equal combined metric falls back to route metric",
+			candidates: []candidateRoute{
+				{interfaceIndex: 1, prefixLength: 0, routeMetric: 20, interfaceMetric: 10},
+				{interfaceIndex: 2, prefixLength: 0, routeMetric: 5, interfaceMetric: 25},
+			},
+			wantOrder: []uint32{2, 1},
+		},
+		{
+			// The metrics are uint32 on the Windows side, so the sum must not wrap.
+			name: "combined metric beyond the uint32 range",
+			candidates: []candidateRoute{
+				{interfaceIndex: 1, prefixLength: 0, routeMetric: math.MaxUint32, interfaceMetric: 5},
+				{interfaceIndex: 2, prefixLength: 0, routeMetric: math.MaxUint32 - 10, interfaceMetric: 5},
+			},
+			wantOrder: []uint32{2, 1},
+		},
+		{
+			name: "unknown interface metric ranks on route metric only",
+			candidates: []candidateRoute{
+				{interfaceIndex: 1, prefixLength: 0, routeMetric: 30, interfaceMetric: -1},
+				{interfaceIndex: 2, prefixLength: 0, routeMetric: 5, interfaceMetric: 10},
+			},
+			wantOrder: []uint32{2, 1},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			sortRouteCandidates(tt.candidates)
+
+			got := make([]uint32, 0, len(tt.candidates))
+			for _, c := range tt.candidates {
+				got = append(got, c.interfaceIndex)
+			}
+			assert.Equal(t, tt.wantOrder, got)
+		})
+	}
+}

client/internal/routemanager/systemops/systemops_windows.go

@@ -882,26 +882,40 @@ func getInterfaceMetric(interfaceIndex uint32, family int16) int {
 	return int(ipInterfaceRow.Metric)
 }
 
-// sortRouteCandidates sorts route candidates by priority: prefix length -> route metric -> interface metric
+// sortRouteCandidates sorts route candidates by priority: prefix length -> combined metric -> route metric.
+// Windows prefers the longest matching prefix and, among prefixes of the same length, the lowest metric, see
+// https://learn.microsoft.com/en-us/windows-hardware/customize/desktop/unattend/microsoft-windows-tcpip-interfaces-interface-routes-route-metric
 func sortRouteCandidates(candidates []candidateRoute) {
 	sort.Slice(candidates, func(i, j int) bool {
 		if candidates[i].prefixLength != candidates[j].prefixLength {
 			return candidates[i].prefixLength > candidates[j].prefixLength
 		}
-		if candidates[i].routeMetric != candidates[j].routeMetric {
-			return candidates[i].routeMetric < candidates[j].routeMetric
+		mi, mj := combinedMetric(candidates[i]), combinedMetric(candidates[j])
+		if mi != mj {
+			return mi < mj
 		}
-		return candidates[i].interfaceMetric < candidates[j].interfaceMetric
+		return candidates[i].routeMetric < candidates[j].routeMetric
 	})
 }
 
+// combinedMetric returns the effective metric Windows uses to rank routes with an equal prefix length:
+// the sum of the route metric and the metric of the interface the route is on, see
+// https://learn.microsoft.com/en-us/windows-server/networking/technologies/network-subsystem/net-sub-interface-metric
+// An unknown interface metric contributes nothing.
+func combinedMetric(candidate candidateRoute) uint64 {
+	if candidate.interfaceMetric < 0 {
+		return uint64(candidate.routeMetric)
+	}
+	return uint64(candidate.routeMetric) + uint64(candidate.interfaceMetric)
+}
+
 // GetBestInterface finds the best interface for reaching a destination,
 // excluding the VPN interface to avoid routing loops.
 //
 // Route selection priority:
 // 1. Longest prefix match (most specific route)
-// 2. Lowest route metric
-// 3. Lowest interface metric
+// 2. Lowest combined metric (route metric + interface metric)
+// 3. Lowest route metric.
 func GetBestInterface(dest netip.Addr, vpnIntf string) (*net.Interface, error) {
 	var skipInterfaceIndex int
 	if vpnIntf != "" {
@@ -925,7 +939,6 @@ func GetBestInterface(dest netip.Addr, vpnIntf string) (*net.Interface, error) {
 		return nil, fmt.Errorf("no route to %s", dest)
 	}
 
-	// Sort routes: prefix length -> route metric -> interface metric
 	sortRouteCandidates(candidates)
 
 	for _, candidate := range candidates {

client/internal/routemanager/systemops/v6route_linux_test.go

@@ -5,6 +5,7 @@ package systemops
 import (
 	"errors"
 	"net"
+	"net/netip"
 	"syscall"
 	"testing"
 
@@ -29,6 +30,7 @@ func ensureIPv6DefaultRoute(t *testing.T) {
 	}
 	if err := netlink.RouteAdd(route); err != nil {
 		if errors.Is(err, syscall.EEXIST) {
+			requireUsableIPv6Nexthop(t)
 			return
 		}
 		t.Skipf("install IPv6 fallback default route: %v", err)
@@ -38,4 +40,36 @@ func ensureIPv6DefaultRoute(t *testing.T) {
 			t.Logf("delete IPv6 fallback default route: %v", err)
 		}
 	})
+
+	requireUsableIPv6Nexthop(t)
+}
+
+// requireUsableIPv6Nexthop skips the test unless the resolved IPv6 default
+// nexthop can actually carry a route. Installing the default route succeeding
+// does not imply the kernel accepts it as a nexthop for a concrete prefix.
+func requireUsableIPv6Nexthop(t *testing.T) {
+	t.Helper()
+
+	nexthop, err := GetNextHop(netip.IPv6Unspecified())
+	if err != nil {
+		t.Skipf("resolve IPv6 default nexthop: %v", err)
+	}
+
+	probe := &netlink.Route{
+		Scope:  netlink.SCOPE_UNIVERSE,
+		Table:  syscall.RT_TABLE_MAIN,
+		Family: netlink.FAMILY_V6,
+		Dst:    &net.IPNet{IP: net.ParseIP("100::64"), Mask: net.CIDRMask(128, 128)},
+	}
+	require.NoError(t, addNextHop(nexthop, probe), "build IPv6 probe route")
+
+	switch err := netlink.RouteAdd(probe); {
+	case err == nil:
+		if err := netlink.RouteDel(probe); err != nil && !errors.Is(err, syscall.ESRCH) {
+			t.Logf("delete IPv6 probe route: %v", err)
+		}
+	case errors.Is(err, syscall.EEXIST):
+	default:
+		t.Skipf("IPv6 nexthop %s unusable for route installation: %v", nexthop, err)
+	}
 }

client/internal/sleep/service.go

@@ -18,8 +18,8 @@ type Service struct {
 }
 
 func New() (*Service, error) {
-	d, err := NewDetector()
-	if err != nil {
+	d, err := NewDetector() //nolint:staticcheck
+	if err != nil {         //nolint:staticcheck // always errors on platforms without a sleep detector
 		return nil, err
 	}
 

client/internal/updater/installer/doc.go

@@ -37,23 +37,32 @@
 // Updater Process (Setup):
 //
 //  1. Receives parameters from service via command-line arguments
-//  2. Runs installer with appropriate silent/quiet flags:
+//  2. Terminates the UI so the installer does not have to replace a locked image
+//     file, which would otherwise leave the install needing a reboot
+//  3. Runs installer with appropriate silent/quiet flags:
 //     - Windows EXE: installer.exe /S
-//     - Windows MSI: msiexec.exe /i installer.msi /quiet /qn /l*v msi.log
+//     - Windows MSI: msiexec.exe /i installer.msi /qn /norestart REBOOT=ReallySuppress /l*v msi.log
 //     - macOS PKG: installer -pkg installer.pkg -target /
 //     - macOS Homebrew: brew upgrade netbirdio/tap/netbird
-//  3. Installer terminates daemon and UI processes
-//  4. Installer replaces binaries with new version
-//  5. Updater waits for installer to complete
-//  6. Updater restarts daemon:
+//  4. Installer terminates the daemon
+//  5. Installer replaces binaries with new version
+//  6. Updater waits for installer to complete. On Windows, MSI exit codes 3010
+//     (ERROR_SUCCESS_REBOOT_REQUIRED) and 1641 (ERROR_SUCCESS_REBOOT_INITIATED)
+//     are a pending-reboot outcome, not a failure: the install succeeded, but
+//     some files are only replaced on the next restart (the reboot itself is
+//     suppressed via /norestart and REBOOT=ReallySuppress), and the flow
+//     continues as on success
+//  7. Updater restarts daemon:
 //     - Windows: netbird.exe service start
 //     - macOS/Linux: netbird service start
-//  7. Updater restarts UI:
-//     - Windows: Launches netbird-ui.exe as active console user using CreateProcessAsUser
+//  8. Updater restarts UI:
+//     - Windows: Launches netbird-ui.exe using CreateProcessAsUser in every
+//     session it was terminated in, falling back to the active console session
 //     - macOS: Uses launchctl asuser to launch NetBird.app for console user
 //     - Linux: Not implemented (UI typically auto-starts)
-//  8. Updater writes result.json with success/error status
-//  9. Updater process exits
+//  9. Updater writes result.json with success/error status (a pending reboot is
+//     recorded as success)
+//  10. Updater process exits
 //
 // # Result Communication
 //
@@ -100,6 +109,10 @@
 //   - Does NOT remove result.json (cleaned by ResultHandler after read)
 //   - Does NOT remove msi.log (kept for debugging)
 //
+// On Windows the updater copy is often still locked when the daemon it restarted
+// runs cleanup, so removing it is retried briefly and otherwise left in place for
+// the next update to overwrite rather than reported as a failure.
+//
 // # Dry-Run Mode
 //
 // Dry-run mode allows testing the update process without actually installing:

client/internal/updater/installer/installer_cleanup_windows_test.go

@@ -0,0 +1,67 @@
+package installer
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"golang.org/x/sys/windows"
+)
+
+// lockFile opens path without FILE_SHARE_DELETE, so os.Remove fails the way it does
+// while the updater process still holds its own image.
+func lockFile(t *testing.T, path string) windows.Handle {
+	t.Helper()
+
+	p, err := windows.UTF16PtrFromString(path)
+	if err != nil {
+		t.Fatalf("convert path: %v", err)
+	}
+
+	handle, err := windows.CreateFile(p, windows.GENERIC_READ, windows.FILE_SHARE_READ, nil, windows.OPEN_EXISTING, windows.FILE_ATTRIBUTE_NORMAL, 0)
+	if err != nil {
+		t.Fatalf("lock %s: %v", path, err)
+	}
+	return handle
+}
+
+// releaseAfter closes the handle once the delay has passed, standing in for the
+// updater process finally exiting.
+func releaseAfter(t *testing.T, handle windows.Handle, delay time.Duration) {
+	t.Helper()
+
+	released := make(chan struct{})
+	t.Cleanup(func() { <-released })
+
+	go func() {
+		defer close(released)
+		time.Sleep(delay)
+		if err := windows.CloseHandle(handle); err != nil {
+			t.Errorf("close handle: %v", err)
+		}
+	}()
+}
+
+// TestCleanUpInstallerFilesLockedUpdater covers the post-update cleanup race: the
+// daemon cleans up at startup while the updater that restarted it is still exiting,
+// so the updater image is locked and Windows refuses the delete. Cleanup must wait
+// the lock out instead of reporting a failure and leaving the binary behind.
+func TestCleanUpInstallerFilesLockedUpdater(t *testing.T) {
+	tempDir := t.TempDir()
+	path := filepath.Join(tempDir, updaterBinary)
+	if err := os.WriteFile(path, []byte("x"), 0o600); err != nil {
+		t.Fatalf("write updater: %v", err)
+	}
+
+	releaseAfter(t, lockFile(t, path), 300*time.Millisecond)
+
+	u := NewWithDir(tempDir)
+	if err := u.CleanUpInstallerFiles(); err != nil {
+		t.Fatalf("cleanup must tolerate a still-locked updater: %v", err)
+	}
+
+	if _, err := os.Stat(path); !os.IsNotExist(err) {
+		t.Errorf("updater binary still present (stat err: %v)", err)
+	}
+}

client/internal/updater/installer/installer_common.go

@@ -42,6 +42,9 @@ func NewWithDir(tempDir string) *Installer {
 // This will run by the original service process
 func (u *Installer) RunInstallation(ctx context.Context, targetVersion string) (err error) {
 	resultHandler := NewResultHandler(u.tempDir)
+	if err := resultHandler.ClearStaleResult(); err != nil {
+		log.Warnf("clear stale installer result: %v", err)
+	}
 
 	defer func() {
 		if err != nil {
@@ -149,8 +152,8 @@ func (u *Installer) CleanUpInstallerFiles() error {
 
 	var merr *multierror.Error
 
-	if err := os.Remove(filepath.Join(u.tempDir, updaterBinary)); err != nil && !os.IsNotExist(err) {
-		merr = multierror.Append(merr, fmt.Errorf("failed to remove updater binary: %w", err))
+	if err := removeUpdaterBinary(filepath.Join(u.tempDir, updaterBinary)); err != nil {
+		merr = multierror.Append(merr, fmt.Errorf("remove updater binary: %w", err))
 	}
 
 	entries, err := os.ReadDir(u.tempDir)
@@ -164,10 +167,16 @@ func (u *Installer) CleanUpInstallerFiles() error {
 		}
 
 		name := entry.Name()
+		// The updater copy is handled above; on Windows its name also matches the
+		// extension sweep, which would report the same file twice.
+		if strings.EqualFold(name, updaterBinary) {
+			continue
+		}
+
 		for _, ext := range binaryExtensions {
 			if strings.HasSuffix(strings.ToLower(name), strings.ToLower(ext)) {
 				if err := os.Remove(filepath.Join(u.tempDir, name)); err != nil {
-					merr = multierror.Append(merr, fmt.Errorf("failed to remove %s: %w", name, err))
+					merr = multierror.Append(merr, fmt.Errorf("remove %s: %w", name, err))
 				}
 				break
 			}

client/internal/updater/installer/installer_common_test.go

@@ -0,0 +1,52 @@
+//go:build windows || darwin
+
+package installer
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+)
+
+// TestCleanUpInstallerFiles checks that cleanup removes the updater copy and the
+// downloaded installer while leaving the logs and the result file for the daemon.
+func TestCleanUpInstallerFiles(t *testing.T) {
+	tempDir := t.TempDir()
+
+	installers := make([]string, 0, len(binaryExtensions))
+	for _, ext := range binaryExtensions {
+		installers = append(installers, "netbird_installer."+ext)
+	}
+
+	kept := []string{"installer.log", "result.json"}
+
+	for _, name := range append(append([]string{updaterBinary}, installers...), kept...) {
+		if err := os.WriteFile(filepath.Join(tempDir, name), []byte("x"), 0o600); err != nil {
+			t.Fatalf("write %s: %v", name, err)
+		}
+	}
+
+	u := NewWithDir(tempDir)
+	if err := u.CleanUpInstallerFiles(); err != nil {
+		t.Fatalf("CleanUpInstallerFiles: %v", err)
+	}
+
+	for _, name := range append([]string{updaterBinary}, installers...) {
+		if _, err := os.Stat(filepath.Join(tempDir, name)); !os.IsNotExist(err) {
+			t.Errorf("%s was not removed (stat err: %v)", name, err)
+		}
+	}
+
+	for _, name := range kept {
+		if _, err := os.Stat(filepath.Join(tempDir, name)); err != nil {
+			t.Errorf("%s should have been kept: %v", name, err)
+		}
+	}
+}
+
+func TestCleanUpInstallerFilesMissingTempDir(t *testing.T) {
+	u := NewWithDir(filepath.Join(t.TempDir(), "does-not-exist"))
+	if err := u.CleanUpInstallerFiles(); err != nil {
+		t.Errorf("a missing temp dir is not a cleanup failure, got: %v", err)
+	}
+}

client/internal/updater/installer/installer_run_windows.go

@@ -2,6 +2,7 @@ package installer
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"os"
 	"os/exec"
@@ -22,6 +23,12 @@ const (
 
 	msiLogFile = "msi.log"
 
+	// ERROR_SUCCESS_REBOOT_REQUIRED and ERROR_SUCCESS_REBOOT_INITIATED
+	msiRebootRequired  = 3010
+	msiRebootInitiated = 1641
+
+	processExitWait = 10 * time.Second
+
 	msiDownloadURL = "https://github.com/netbirdio/netbird/releases/download/v%version/netbird_installer_%version_windows_%arch.msi"
 	exeDownloadURL = "https://github.com/netbirdio/netbird/releases/download/v%version/netbird_installer_%version_windows_%arch.exe"
 )
@@ -38,6 +45,8 @@ var (
 func (u *Installer) Setup(ctx context.Context, dryRun bool, installerFile string, daemonFolder string) (resultErr error) {
 	resultHandler := NewResultHandler(u.tempDir)
 
+	var uiSessions []uint32
+
 	// Always ensure daemon and UI are restarted after setup
 	defer func() {
 		log.Infof("starting daemon back")
@@ -46,7 +55,7 @@ func (u *Installer) Setup(ctx context.Context, dryRun bool, installerFile string
 		}
 
 		log.Infof("starting UI back")
-		if err := u.startUIAsUser(daemonFolder); err != nil {
+		if err := u.startUI(daemonFolder, uiSessions); err != nil {
 			log.Errorf("failed to start UI: %v", err)
 		}
 
@@ -75,6 +84,14 @@ func (u *Installer) Setup(ctx context.Context, dryRun bool, installerFile string
 		return
 	}
 
+	// The UI holds an open handle on its own image. Left running, Restart Manager
+	// cannot shut it down (msiexec runs as LocalSystem here, the UI as the
+	// interactive user), so the MSI falls back to replacing the file on reboot and
+	// marks the install as restart-required. The deferred close-application action
+	// in the package runs too late to prevent that, it happens after
+	// InstallValidate has already registered the file as in use.
+	uiSessions = killUI()
+
 	var cmd *exec.Cmd
 	switch installerType {
 	case TypeExe:
@@ -84,7 +101,9 @@ func (u *Installer) Setup(ctx context.Context, dryRun bool, installerFile string
 		installerDir := filepath.Dir(installerFile)
 		logPath := filepath.Join(installerDir, msiLogFile)
 		log.Infof("run msi installer: %s", installerFile)
-		cmd = exec.CommandContext(ctx, "msiexec.exe", "/i", filepath.Base(installerFile), "/quiet", "/qn", "/l*v", logPath)
+		// REBOOT=ReallySuppress: a silent install has no way to ask, so without it
+		// msiexec reboots the machine on its own if it decides one is needed.
+		cmd = exec.CommandContext(ctx, "msiexec.exe", "/i", filepath.Base(installerFile), "/qn", "/norestart", "REBOOT=ReallySuppress", "/l*v", logPath)
 	}
 
 	cmd.Dir = filepath.Dir(installerFile)
@@ -95,9 +114,13 @@ func (u *Installer) Setup(ctx context.Context, dryRun bool, installerFile string
 	}
 
 	log.Infof("installer started with PID %d", cmd.Process.Pid)
-	if resultErr = cmd.Wait(); resultErr != nil {
-		log.Errorf("installer process finished with error: %v", resultErr)
-		return
+	if err := cmd.Wait(); err != nil {
+		if !isRebootPending(err) {
+			resultErr = err
+			log.Errorf("installer process finished with error: %v", err)
+			return
+		}
+		log.Warnf("installer completed but reported a pending reboot, some files will be replaced on the next restart")
 	}
 
 	return nil
@@ -117,16 +140,142 @@ func (u *Installer) startDaemon(daemonFolder string) error {
 	return nil
 }
 
-func (u *Installer) startUIAsUser(daemonFolder string) error {
+func (u *Installer) startUI(daemonFolder string, sessionIDs []uint32) error {
 	uiPath := filepath.Join(daemonFolder, uiName)
 	log.Infof("starting netbird-ui: %s", uiPath)
 
-	// Get the active console session ID
-	sessionID := windows.WTSGetActiveConsoleSessionId()
-	if sessionID == 0xFFFFFFFF {
-		return fmt.Errorf("no active user session found")
+	if len(sessionIDs) == 0 {
+		sessionID := windows.WTSGetActiveConsoleSessionId()
+		if sessionID == 0xFFFFFFFF {
+			return fmt.Errorf("no active user session found")
+		}
+		sessionIDs = []uint32{sessionID}
 	}
 
+	var errs []error
+	for _, sessionID := range sessionIDs {
+		if err := startUIInSession(uiPath, sessionID); err != nil {
+			errs = append(errs, fmt.Errorf("session %d: %w", sessionID, err))
+			continue
+		}
+		log.Infof("netbird-ui started successfully in session %d", sessionID)
+	}
+	return errors.Join(errs...)
+}
+
+// isRebootPending reports whether the installer exit code means it succeeded but
+// left work for the next restart. The reboot itself is suppressed, so this is not
+// a failure.
+func isRebootPending(err error) bool {
+	var exitErr *exec.ExitError
+	if !errors.As(err, &exitErr) {
+		return false
+	}
+
+	switch exitErr.ExitCode() {
+	case msiRebootRequired, msiRebootInitiated:
+		return true
+	default:
+		return false
+	}
+}
+
+// killUI terminates any running netbird-ui process and returns the IDs of the
+// interactive sessions the terminated processes belonged to. Setup starts the
+// UI again in those sessions once the installer is done.
+func killUI() []uint32 {
+	pids, err := processIDsByName(uiName)
+	if err != nil {
+		log.Warnf("failed to look up %s processes: %v", uiName, err)
+		return nil
+	}
+
+	sessions := make(map[uint32]struct{})
+	for _, pid := range pids {
+		var sessionID uint32
+		if err := windows.ProcessIdToSessionId(pid, &sessionID); err != nil {
+			log.Warnf("failed to look up session of %s (PID %d): %v", uiName, pid, err)
+		}
+
+		if err := terminateProcess(pid); err != nil {
+			log.Warnf("failed to terminate %s (PID %d): %v", uiName, pid, err)
+			continue
+		}
+		log.Infof("terminated %s (PID %d) in session %d", uiName, pid, sessionID)
+
+		if sessionID != 0 {
+			sessions[sessionID] = struct{}{}
+		}
+	}
+
+	sessionIDs := make([]uint32, 0, len(sessions))
+	for sessionID := range sessions {
+		sessionIDs = append(sessionIDs, sessionID)
+	}
+	return sessionIDs
+}
+
+func processIDsByName(name string) ([]uint32, error) {
+	snapshot, err := windows.CreateToolhelp32Snapshot(windows.TH32CS_SNAPPROCESS, 0)
+	if err != nil {
+		return nil, fmt.Errorf("create process snapshot: %w", err)
+	}
+	defer func() {
+		if err := windows.CloseHandle(snapshot); err != nil {
+			log.Warnf("failed to close process snapshot: %v", err)
+		}
+	}()
+
+	var entry windows.ProcessEntry32
+	entry.Size = uint32(unsafe.Sizeof(entry))
+
+	var pids []uint32
+	for err = windows.Process32First(snapshot, &entry); err == nil; err = windows.Process32Next(snapshot, &entry) {
+		if strings.EqualFold(windows.UTF16ToString(entry.ExeFile[:]), name) {
+			pids = append(pids, entry.ProcessID)
+		}
+	}
+	if !errors.Is(err, windows.ERROR_NO_MORE_FILES) {
+		return nil, fmt.Errorf("enumerate processes: %w", err)
+	}
+
+	return pids, nil
+}
+
+func terminateProcess(pid uint32) error {
+	handle, err := windows.OpenProcess(windows.PROCESS_TERMINATE|windows.SYNCHRONIZE, false, pid)
+	if err != nil {
+		// The process may have exited between enumeration and now.
+		if errors.Is(err, windows.ERROR_INVALID_PARAMETER) {
+			return nil
+		}
+		return fmt.Errorf("open process: %w", err)
+	}
+	defer func() {
+		if err := windows.CloseHandle(handle); err != nil {
+			log.Warnf("failed to close process handle: %v", err)
+		}
+	}()
+
+	if err := windows.TerminateProcess(handle, 0); err != nil {
+		return fmt.Errorf("terminate process: %w", err)
+	}
+
+	// Wait for the handle to signal so the image file is released before the
+	// installer tries to overwrite it. A timeout is reported through the returned
+	// event, not through err, which stays nil unless the wait itself failed.
+	event, err := windows.WaitForSingleObject(handle, uint32(processExitWait.Milliseconds()))
+	if err != nil {
+		return fmt.Errorf("wait for process exit: %w", err)
+	}
+	if event != windows.WAIT_OBJECT_0 {
+		return fmt.Errorf("wait for process exit: unexpected wait result %#x", event)
+	}
+
+	return nil
+}
+
+func startUIInSession(uiPath string, sessionID uint32) error {
 	// Get the user token for that session
 	var userToken windows.Token
 	err := windows.WTSQueryUserToken(sessionID, &userToken)
@@ -158,6 +307,16 @@ func (u *Installer) startUIAsUser(daemonFolder string) error {
 		}
 	}()
 
+	var env *uint16
+	if err := windows.CreateEnvironmentBlock(&env, primaryToken, false); err != nil {
+		return fmt.Errorf("create environment block: %w", err)
+	}
+	defer func() {
+		if err := windows.DestroyEnvironmentBlock(env); err != nil {
+			log.Warnf("failed to destroy environment block: %v", err)
+		}
+	}()
+
 	// Prepare startup info
 	var si windows.StartupInfo
 	si.Cb = uint32(unsafe.Sizeof(si))
@@ -180,7 +339,7 @@ func (u *Installer) startUIAsUser(daemonFolder string) error {
 		nil,
 		false,
 		creationFlags,
-		nil,
+		env,
 		nil,
 		&si,
 		&pi,
@@ -197,7 +356,6 @@ func (u *Installer) startUIAsUser(daemonFolder string) error {
 		log.Warnf("failed to close thread handle: %v", err)
 	}
 
-	log.Infof("netbird-ui started successfully in session %d", sessionID)
 	return nil
 }
 

client/internal/updater/installer/installer_run_windows_test.go

@@ -0,0 +1,108 @@
+package installer
+
+import (
+	"errors"
+	"os/exec"
+	"slices"
+	"strconv"
+	"testing"
+)
+
+// exitErrorWithCode returns a real *exec.ExitError carrying the given exit code.
+func exitErrorWithCode(t *testing.T, code int) error {
+	t.Helper()
+
+	err := exec.Command("cmd.exe", "/c", "exit "+strconv.Itoa(code)).Run()
+	if err == nil {
+		t.Fatalf("expected a non-zero exit for code %d", code)
+	}
+	return err
+}
+
+func TestIsRebootPending(t *testing.T) {
+	tests := []struct {
+		name string
+		code int
+		want bool
+	}{
+		{name: "reboot required", code: msiRebootRequired, want: true},
+		{name: "reboot initiated", code: msiRebootInitiated, want: true},
+		{name: "generic failure", code: 1603, want: false},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			if got := isRebootPending(exitErrorWithCode(t, tt.code)); got != tt.want {
+				t.Errorf("isRebootPending(exit %d) = %v, want %v", tt.code, got, tt.want)
+			}
+		})
+	}
+}
+
+// TestProcessIDsByNameAndTerminate spawns a long-running system process, finds it
+// by name and terminates it, covering the path the updater uses to release the UI
+// image file before the installer replaces it.
+func TestProcessIDsByNameAndTerminate(t *testing.T) {
+	cmd := exec.Command("ping.exe", "-n", "60", "127.0.0.1")
+	if err := cmd.Start(); err != nil {
+		t.Fatalf("start ping: %v", err)
+	}
+
+	pid := uint32(cmd.Process.Pid)
+	killed := false
+	t.Cleanup(func() {
+		if !killed {
+			_ = cmd.Process.Kill()
+		}
+		_ = cmd.Wait()
+	})
+
+	// Name matching must be case-insensitive: the snapshot reports PING.EXE.
+	pids, err := processIDsByName("ping.exe")
+	if err != nil {
+		t.Fatalf("processIDsByName: %v", err)
+	}
+
+	if !slices.Contains(pids, pid) {
+		t.Fatalf("PID %d not among the ping.exe processes found: %v", pid, pids)
+	}
+
+	if err := terminateProcess(pid); err != nil {
+		t.Fatalf("terminateProcess: %v", err)
+	}
+	killed = true
+
+	// terminateProcess only returns once the handle has signalled, so the process
+	// is already gone and Wait must not block. It exits with the code passed to
+	// TerminateProcess, which is 0, so Wait reports no error.
+	if err := cmd.Wait(); err != nil {
+		t.Fatalf("wait for terminated ping: %v", err)
+	}
+	if !cmd.ProcessState.Exited() {
+		t.Error("process did not exit after terminateProcess")
+	}
+
+	remaining, err := processIDsByName("ping.exe")
+	if err != nil {
+		t.Fatalf("processIDsByName after terminate: %v", err)
+	}
+	if slices.Contains(remaining, pid) {
+		t.Errorf("PID %d still listed after terminateProcess", pid)
+	}
+}
+
+func TestProcessIDsByNameNoMatch(t *testing.T) {
+	pids, err := processIDsByName("netbird-nonexistent-process.exe")
+	if err != nil {
+		t.Fatalf("processIDsByName: %v", err)
+	}
+	if len(pids) != 0 {
+		t.Errorf("expected no matches, got %v", pids)
+	}
+}
+
+func TestIsRebootPendingNonExitError(t *testing.T) {
+	if isRebootPending(errors.New("start installer: file not found")) {
+		t.Error("a non-exit error must not be treated as a pending reboot")
+	}
+}

client/internal/updater/installer/remove_updater_darwin.go

@@ -0,0 +1,12 @@
+package installer
+
+import "os"
+
+// removeUpdaterBinary deletes the updater copy left in the temp dir. On darwin a
+// running binary can be unlinked, so no retry is needed.
+func removeUpdaterBinary(path string) error {
+	if err := os.Remove(path); err != nil && !os.IsNotExist(err) {
+		return err
+	}
+	return nil
+}

client/internal/updater/installer/remove_updater_windows.go

@@ -0,0 +1,45 @@
+package installer
+
+import (
+	"errors"
+	"os"
+	"time"
+
+	log "github.com/sirupsen/logrus"
+	"golang.org/x/sys/windows"
+)
+
+const (
+	// The updater is the process that restarted the daemon, so when the daemon
+	// cleans up at startup the updater is often still exiting and Windows refuses
+	// to delete its locked image. These bound how long cleanup waits for it.
+	updaterRemoveAttempts = 5
+	updaterRemoveDelay    = 200 * time.Millisecond
+)
+
+// removeUpdaterBinary deletes the updater copy left in the temp dir, retrying
+// while the still-exiting updater process holds its image. A binary that stays
+// locked for the whole window is left in place and reported at info level: the
+// next update overwrites it, so it is not worth failing cleanup over.
+func removeUpdaterBinary(path string) error {
+	for attempt := 0; attempt < updaterRemoveAttempts; attempt++ {
+		if attempt > 0 {
+			time.Sleep(updaterRemoveDelay)
+		}
+
+		err := os.Remove(path)
+		if err == nil || os.IsNotExist(err) {
+			return nil
+		}
+		if !isFileLocked(err) {
+			return err
+		}
+	}
+
+	log.Infof("updater binary %s is still locked, leaving it for the next update to overwrite", path)
+	return nil
+}
+
+func isFileLocked(err error) bool {
+	return errors.Is(err, windows.ERROR_ACCESS_DENIED) || errors.Is(err, windows.ERROR_SHARING_VIOLATION)
+}

client/internal/updater/installer/remove_updater_windows_test.go

@@ -0,0 +1,59 @@
+package installer
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"golang.org/x/sys/windows"
+)
+
+func TestRemoveUpdaterBinaryRetriesWhileLocked(t *testing.T) {
+	path := filepath.Join(t.TempDir(), updaterBinary)
+	if err := os.WriteFile(path, []byte("x"), 0o600); err != nil {
+		t.Fatalf("write updater: %v", err)
+	}
+
+	releaseAfter(t, lockFile(t, path), updaterRemoveDelay+50*time.Millisecond)
+
+	if err := removeUpdaterBinary(path); err != nil {
+		t.Fatalf("removeUpdaterBinary: %v", err)
+	}
+
+	if _, err := os.Stat(path); !os.IsNotExist(err) {
+		t.Errorf("updater binary still present (stat err: %v)", err)
+	}
+}
+
+// TestRemoveUpdaterBinaryStaysLocked covers an updater that never releases its
+// image within the retry window. Cleanup gives up quietly and leaves the file
+// behind rather than reporting a failure.
+func TestRemoveUpdaterBinaryStaysLocked(t *testing.T) {
+	path := filepath.Join(t.TempDir(), updaterBinary)
+	if err := os.WriteFile(path, []byte("x"), 0o600); err != nil {
+		t.Fatalf("write updater: %v", err)
+	}
+
+	handle := lockFile(t, path)
+	t.Cleanup(func() {
+		if err := windows.CloseHandle(handle); err != nil {
+			t.Errorf("close handle: %v", err)
+		}
+	})
+
+	if err := removeUpdaterBinary(path); err != nil {
+		t.Fatalf("a permanently locked updater is not a cleanup failure, got: %v", err)
+	}
+
+	if _, err := os.Stat(path); err != nil {
+		t.Errorf("locked updater binary should be left in place, stat: %v", err)
+	}
+}
+
+func TestRemoveUpdaterBinaryMissingFile(t *testing.T) {
+	path := filepath.Join(t.TempDir(), updaterBinary)
+	if err := removeUpdaterBinary(path); err != nil {
+		t.Errorf("a missing updater binary is not a failure, got: %v", err)
+	}
+}

client/internal/updater/installer/result.go

@@ -54,6 +54,12 @@ func (rh *ResultHandler) GetErrorResultReason() string {
 	return ""
 }
 
+// ClearStaleResult removes a result file left over from a previous installation
+// attempt so result watchers cannot read an outdated outcome for the current attempt.
+func (rh *ResultHandler) ClearStaleResult() error {
+	return rh.cleanup()
+}
+
 func (rh *ResultHandler) WriteSuccess() error {
 	result := Result{
 		Success:    true,

client/internal/updater/manager.go

@@ -435,7 +435,7 @@ func (m *Manager) install(ctx context.Context, pendingVersion *v.Version) error
 	}
 
 	inst := installer.New()
-	if err := inst.RunInstallation(ctx, pendingVersion.String()); err != nil {
+	if err := inst.RunInstallation(ctx, pendingVersion.String()); err != nil { //nolint:staticcheck // always errors on platforms without an installer
 		log.Errorf("error triggering update: %v", err)
 		m.statusRecorder.PublishEvent(
 			cProto.SystemEvent_ERROR,