[client] Fix race between WG watcher initial handshake read and endpoint creation (#6626)

* [client] Fix race between WG watcher initial handshake read and endpoint config

The watcher's initial handshake read ran in a separate goroutine with no
ordering guarantee relative to the WireGuard endpoint configuration, so it
would sometimes race with the peer being added to the interface. Split
enabling into a synchronous PrepareInitialHandshake, called before the
endpoint is configured, and an EnableWgWatcher that only runs the
monitoring loop, making the baseline read deterministic and keeping it
correct for reconnects where the peer's WireGuard entry survives.

* [client] Skip WG watcher disconnect callback when context is cancelled

A superseded or cancelled watcher whose handshake-check timer fires before
it observes ctx.Done() would still invoke onDisconnectedFn, tearing down a
now-healthy connection. Re-check ctx before firing the disconnect and
handshake-success callbacks and stand down silently if it was cancelled.

client/internal/peer/conn.go

@@ -803,15 +803,17 @@ func (conn *Conn) isConnectedOnAllWay() (status guard.ConnStatus) {
 }
 
 func (conn *Conn) enableWgWatcherIfNeeded(enabledTime time.Time) {
-	if !conn.wgWatcher.IsEnabled() {
+	if !conn.wgWatcher.PrepareInitialHandshake() {
-		wgWatcherCtx, wgWatcherCancel := context.WithCancel(conn.ctx)
+		return
-		conn.wgWatcherCancel = wgWatcherCancel
-		conn.wgWatcherWg.Add(1)
-		go func() {
-			defer conn.wgWatcherWg.Done()
-			conn.wgWatcher.EnableWgWatcher(wgWatcherCtx, enabledTime, conn.onWGDisconnected, conn.onWGHandshakeSuccess)
-		}()
 	}
+
+	wgWatcherCtx, wgWatcherCancel := context.WithCancel(conn.ctx)
+	conn.wgWatcherCancel = wgWatcherCancel
+	conn.wgWatcherWg.Add(1)
+	go func() {
+		defer conn.wgWatcherWg.Done()
+		conn.wgWatcher.EnableWgWatcher(wgWatcherCtx, enabledTime, conn.onWGDisconnected, conn.onWGHandshakeSuccess)
+	}()
 }
 
 func (conn *Conn) disableWgWatcherIfNeeded() {

client/internal/peer/wg_watcher.go

@@ -31,7 +31,9 @@ type WGWatcher struct {
 	stateDump     *stateDump
 
 	enabled   bool
-	muEnabled sync.RWMutex
+	muEnabled sync.Mutex
+	// initialHandshake is not thread-safe; never call PrepareInitialHandshake and EnableWgWatcher concurrently.
+	initialHandshake time.Time
 
 	resetCh chan struct{}
 }
@@ -46,38 +48,38 @@ func NewWGWatcher(log *log.Entry, wgIfaceStater WGInterfaceStater, peerKey strin
 	}
 }
 
-// EnableWgWatcher starts the WireGuard watcher. If it is already enabled, it will return immediately and do nothing.
+// PrepareInitialHandshake reserves the watcher and reads the peer's current WireGuard
-// The watcher runs until ctx is cancelled. Caller is responsible for context lifecycle management.
+// handshake time. It must be called before the peer is (re)configured on the WireGuard
-func (w *WGWatcher) EnableWgWatcher(ctx context.Context, enabledTime time.Time, onDisconnectedFn func(), onHandshakeSuccessFn func(when time.Time)) {
+// interface, so the captured baseline reflects the state prior to this connection attempt
+// instead of racing with that configuration. Returns ok=false if the watcher is already
+// running, in which case EnableWgWatcher must not be called.
+func (w *WGWatcher) PrepareInitialHandshake() (ok bool) {
 	w.muEnabled.Lock()
 	if w.enabled {
 		w.muEnabled.Unlock()
-		return
+		return false
 	}
 
 	w.log.Debugf("enable WireGuard watcher")
 	w.enabled = true
 	w.muEnabled.Unlock()
 
-	initialHandshake, err := w.wgState()
+	handshake, _ := w.wgState()
-	if err != nil {
+	w.initialHandshake = handshake
-		w.log.Warnf("failed to read initial wg stats: %v", err)
+	return true
-	}
+}
 
-	w.periodicHandshakeCheck(ctx, onDisconnectedFn, onHandshakeSuccessFn, enabledTime, initialHandshake)
+// EnableWgWatcher runs the WireGuard watcher loop using the handshake baseline captured by
+// PrepareInitialHandshake. The watcher runs until ctx is cancelled. Caller is responsible
+// for context lifecycle management.
+func (w *WGWatcher) EnableWgWatcher(ctx context.Context, enabledTime time.Time, onDisconnectedFn func(), onHandshakeSuccessFn func(when time.Time)) {
+	w.periodicHandshakeCheck(ctx, onDisconnectedFn, onHandshakeSuccessFn, enabledTime, w.initialHandshake)
 
 	w.muEnabled.Lock()
 	w.enabled = false
 	w.muEnabled.Unlock()
 }
 
-// IsEnabled returns true if the WireGuard watcher is currently enabled
-func (w *WGWatcher) IsEnabled() bool {
-	w.muEnabled.RLock()
-	defer w.muEnabled.RUnlock()
-	return w.enabled
-}
-
 // Reset signals the watcher that the WireGuard peer has been reset and a new
 // handshake is expected. This restarts the handshake timeout from scratch.
 func (w *WGWatcher) Reset() {
@@ -101,13 +103,16 @@ func (w *WGWatcher) periodicHandshakeCheck(ctx context.Context, onDisconnectedFn
 		case <-timer.C:
 			handshake, ok := w.handshakeCheck(lastHandshake)
 			if !ok {
+				if ctx.Err() != nil {
+					return
+				}
 				onDisconnectedFn()
 				return
 			}
 			if lastHandshake.IsZero() {
 				elapsed := calcElapsed(enabledTime, *handshake)
 				w.log.Infof("first wg handshake detected within: %.2fsec, (%s)", elapsed, handshake)
-				if onHandshakeSuccessFn != nil {
+				if onHandshakeSuccessFn != nil && ctx.Err() == nil {
 					onHandshakeSuccessFn(*handshake)
 				}
 			}

client/internal/peer/wg_watcher_test.go

@@ -7,6 +7,7 @@ import (
 	"time"
 
 	log "github.com/sirupsen/logrus"
+	"github.com/stretchr/testify/require"
 
 	"github.com/netbirdio/netbird/client/iface/configurer"
 )
@@ -34,6 +35,9 @@ func TestWGWatcher_EnableWgWatcher(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 
+	ok := watcher.PrepareInitialHandshake()
+	require.True(t, ok, "watcher should not be enabled yet")
+
 	onDisconnected := make(chan struct{}, 1)
 	go watcher.EnableWgWatcher(ctx, time.Now(), func() {
 		mlog.Infof("onDisconnectedFn")
@@ -62,6 +66,9 @@ func TestWGWatcher_ReEnable(t *testing.T) {
 	watcher := NewWGWatcher(mlog, mocWgIface, "", newStateDump("peer", mlog, &Status{}))
 
 	ctx, cancel := context.WithCancel(context.Background())
+	ok := watcher.PrepareInitialHandshake()
+	require.True(t, ok, "watcher should not be enabled yet")
+
 	wg := &sync.WaitGroup{}
 	wg.Add(1)
 	go func() {
@@ -76,6 +83,9 @@ func TestWGWatcher_ReEnable(t *testing.T) {
 	ctx, cancel = context.WithCancel(context.Background())
 	defer cancel()
 
+	ok = watcher.PrepareInitialHandshake()
+	require.True(t, ok, "watcher should be re-enabled after the previous run stopped")
+
 	onDisconnected := make(chan struct{}, 1)
 	go watcher.EnableWgWatcher(ctx, time.Now(), func() {
 		onDisconnected <- struct{}{}