Addresses CI fixes

client/internal/peer/conn.go

@@ -84,14 +84,14 @@ type RosenpassConfig struct {
 type PQHandshaker interface {
 	// OfferPayload returns the KEM offer to embed in an outgoing offer (nil if this
 	// peer is not the KEM initiator) and the local PQ data-path port to announce.
-	OfferPayload(remoteKey string) (payload []byte, port int)
+	OfferPayload(remoteKey string) (payload []byte, port uint16)
 	// ShouldSendBootstrapOffer reports whether, as the controller, we should reply to a
 	// received responder offer with our own KEM offer (true only when no exchange is
 	// already in flight — so we kick the KEM once and ignore further offers).
 	ShouldSendBootstrapOffer(remoteKey string) bool
 	// AnswerPayload processes a received KEM offer (nil if absent) and returns the KEM
 	// answer to embed in the outgoing answer (nil if none) and the local PQ port.
-	AnswerPayload(remoteKey string, recvOffer []byte) (payload []byte, port int)
+	AnswerPayload(remoteKey string, recvOffer []byte) (payload []byte, port uint16)
 	// OnAnswer feeds a received KEM answer (nil if absent).
 	OnAnswer(remoteKey string, recvAnswer []byte)
 	// PSK returns the peer's latest derived post-quantum PSK to program at WG
@@ -468,6 +468,9 @@ func (conn *Conn) ConnID() id.ConnID {
 
 // configureConnection starts proxying traffic from/to local Wireguard and sets connection status to StatusConnected
 func (conn *Conn) onICEConnectionIsReady(priority conntype.ConnPriority, iceConnInfo ICEConnInfo) {
+	// Read the PQ PSK before conn.mu to keep the lock order conn.mu -> manager.
+	pqPSK, pqOK := conn.pqPSK()
+
 	conn.mu.Lock()
 	defer conn.mu.Unlock()
 
@@ -485,7 +488,7 @@ func (conn *Conn) onICEConnectionIsReady(priority conntype.ConnPriority, iceConn
 	if conn.currentConnPriority > priority {
 		conn.Log.Infof("current connection priority (%s) is higher than the new one (%s), do not upgrade connection", conn.currentConnPriority, priority)
 		conn.statusICE.SetConnected()
-		conn.updateIceState(iceConnInfo, time.Now())
+		conn.updateIceState(iceConnInfo, pqOK, time.Now())
 		return
 	}
 
@@ -528,7 +531,7 @@ func (conn *Conn) onICEConnectionIsReady(priority conntype.ConnPriority, iceConn
 	updateTime := time.Now()
 	conn.enableWgWatcherIfNeeded(updateTime)
 
-	presharedKey := conn.presharedKey(iceConnInfo.RosenpassPubKey)
+	presharedKey := conn.presharedKey(iceConnInfo.RosenpassPubKey, pqPSK)
 	if err = conn.endpointUpdater.ConfigureWGEndpoint(ep, presharedKey); err != nil {
 		conn.handleConfigurationFailure(err, wgProxy)
 		return
@@ -544,11 +547,14 @@ func (conn *Conn) onICEConnectionIsReady(priority conntype.ConnPriority, iceConn
 
 	conn.currentConnPriority = priority
 	conn.statusICE.SetConnected()
-	conn.updateIceState(iceConnInfo, updateTime)
+	conn.updateIceState(iceConnInfo, pqOK, updateTime)
 	conn.doOnConnected(iceConnInfo.RosenpassPubKey, iceConnInfo.RosenpassAddr, updateTime)
 }
 
 func (conn *Conn) onICEStateDisconnected(sessionChanged bool) {
+	// Read the PQ PSK before conn.mu to keep the lock order conn.mu -> manager.
+	pqPSK, _ := conn.pqPSK()
+
 	conn.mu.Lock()
 	defer conn.mu.Unlock()
 
@@ -575,7 +581,7 @@ func (conn *Conn) onICEStateDisconnected(sessionChanged bool) {
 		// todo consider to move after the ConfigureWGEndpoint
 		conn.wgProxyRelay.Work()
 
-		presharedKey := conn.presharedKey(conn.rosenpassRemoteKey)
+		presharedKey := conn.presharedKey(conn.rosenpassRemoteKey, pqPSK)
 		if err := conn.endpointUpdater.SwitchWGEndpoint(conn.wgProxyRelay.EndpointAddr(), presharedKey); err != nil {
 			conn.Log.Errorf("failed to switch to relay conn: %v", err)
 		}
@@ -613,6 +619,9 @@ func (conn *Conn) onICEStateDisconnected(sessionChanged bool) {
 }
 
 func (conn *Conn) onRelayConnectionIsReady(rci RelayConnInfo) {
+	// Read the PQ PSK before conn.mu to keep the lock order conn.mu -> manager.
+	pqPSK, pqOK := conn.pqPSK()
+
 	conn.mu.Lock()
 	defer conn.mu.Unlock()
 
@@ -641,7 +650,7 @@ func (conn *Conn) onRelayConnectionIsReady(rci RelayConnInfo) {
 		conn.Log.Debugf("do not switch to relay because current priority is: %s", conn.currentConnPriority.String())
 		conn.setRelayedProxy(wgProxy)
 		conn.statusRelay.SetConnected()
-		conn.updateRelayStatus(rci.relayedConn.RemoteAddr().String(), rci.rosenpassPubKey, time.Now())
+		conn.updateRelayStatus(rci.relayedConn.RemoteAddr().String(), rci.rosenpassPubKey, pqOK, time.Now())
 		return
 	}
 
@@ -652,7 +661,7 @@ func (conn *Conn) onRelayConnectionIsReady(rci RelayConnInfo) {
 	}
 	updateTime := time.Now()
 	conn.enableWgWatcherIfNeeded(updateTime)
-	if err := conn.endpointUpdater.ConfigureWGEndpoint(wgProxy.EndpointAddr(), conn.presharedKey(rci.rosenpassPubKey)); err != nil {
+	if err := conn.endpointUpdater.ConfigureWGEndpoint(wgProxy.EndpointAddr(), conn.presharedKey(rci.rosenpassPubKey, pqPSK)); err != nil {
 		if err := wgProxy.CloseConn(); err != nil {
 			conn.Log.Warnf("Failed to close relay connection: %v", err)
 		}
@@ -671,7 +680,7 @@ func (conn *Conn) onRelayConnectionIsReady(rci RelayConnInfo) {
 	conn.currentConnPriority = conntype.Relay
 	conn.statusRelay.SetConnected()
 	conn.setRelayedProxy(wgProxy)
-	conn.updateRelayStatus(rci.relayedConn.RemoteAddr().String(), rci.rosenpassPubKey, updateTime)
+	conn.updateRelayStatus(rci.relayedConn.RemoteAddr().String(), rci.rosenpassPubKey, pqOK, updateTime)
 	conn.Log.Infof("start to communicate with peer via relay")
 	conn.doOnConnected(rci.rosenpassPubKey, rci.rosenpassAddr, updateTime)
 }
@@ -751,19 +760,15 @@ func (conn *Conn) RequestReoffer() {
 
 func (conn *Conn) onWGDisconnected(watcherCtx context.Context) {
 	conn.mu.Lock()
-	defer conn.mu.Unlock()
 
 	// watcherCtx guards against a stale watcher tearing down a connection that already superseded it.
 	if conn.ctx.Err() != nil || watcherCtx.Err() != nil {
+		conn.mu.Unlock()
 		return
 	}
 
 	conn.Log.Warnf("WireGuard handshake timeout detected, closing current connection")
 
-	if conn.config.PQ != nil {
-		conn.config.PQ.OnDataPathDown(conn.config.Key)
-	}
-
 	// Close the active connection based on current priority
 	switch conn.currentConnPriority {
 	case conntype.Relay:
@@ -776,6 +781,15 @@ func (conn *Conn) onWGDisconnected(watcherCtx context.Context) {
 	}
 
 	conn.escalateWGTimeoutLocked()
+	pq := conn.config.PQ
+	key := conn.config.Key
+	conn.mu.Unlock()
+
+	// Signal the PQ manager outside conn.mu: it may re-enter Conn (reoffer) under
+	// conn.mu, so calling it while holding the lock would invert the lock order.
+	if pq != nil {
+		pq.OnDataPathDown(key)
+	}
 }
 
 // escalateWGTimeoutLocked resets the peer's rosenpass state after repeated
@@ -799,14 +813,14 @@ func (conn *Conn) escalateWGTimeoutLocked() {
 	conn.onDisconnected(conn.config.WgConfig.RemoteKey)
 }
 
-func (conn *Conn) updateRelayStatus(relayServerAddr string, rosenpassPubKey []byte, updateTime time.Time) {
+func (conn *Conn) updateRelayStatus(relayServerAddr string, rosenpassPubKey []byte, pqEstablished bool, updateTime time.Time) {
 	peerState := State{
 		PubKey:             conn.config.Key,
 		ConnStatusUpdate:   updateTime,
 		ConnStatus:         conn.evalStatus(),
 		Relayed:            conn.isRelayed(),
 		RelayServerAddress: relayServerAddr,
-		RosenpassEnabled:   conn.quantumResistant(rosenpassPubKey),
+		RosenpassEnabled:   conn.quantumResistant(rosenpassPubKey, pqEstablished),
 	}
 
 	err := conn.statusRecorder.UpdatePeerRelayedState(peerState)
@@ -815,7 +829,7 @@ func (conn *Conn) updateRelayStatus(relayServerAddr string, rosenpassPubKey []by
 	}
 }
 
-func (conn *Conn) updateIceState(iceConnInfo ICEConnInfo, updateTime time.Time) {
+func (conn *Conn) updateIceState(iceConnInfo ICEConnInfo, pqEstablished bool, updateTime time.Time) {
 	peerState := State{
 		PubKey:                     conn.config.Key,
 		ConnStatusUpdate:           updateTime,
@@ -825,7 +839,7 @@ func (conn *Conn) updateIceState(iceConnInfo ICEConnInfo, updateTime time.Time)
 		RemoteIceCandidateType:     iceConnInfo.RemoteIceCandidateType,
 		LocalIceCandidateEndpoint:  iceConnInfo.LocalIceCandidateEndpoint,
 		RemoteIceCandidateEndpoint: iceConnInfo.RemoteIceCandidateEndpoint,
-		RosenpassEnabled:           conn.quantumResistant(iceConnInfo.RosenpassPubKey),
+		RosenpassEnabled:           conn.quantumResistant(iceConnInfo.RosenpassPubKey, pqEstablished),
 	}
 
 	err := conn.statusRecorder.UpdatePeerICEState(peerState)
@@ -1094,13 +1108,31 @@ func (conn *Conn) AgentVersionString() string {
 	return conn.config.AgentVersion
 }
 
-func (conn *Conn) presharedKey(remoteRosenpassKey []byte) *wgtypes.Key {
+// pqPSK returns the post-quantum PSK derived for this peer, if the ML-KEM exchange has
+// produced one. It reads the manager WITHOUT conn.mu, so callers fetch it before taking
+// conn.mu: the lock order is always conn.mu -> manager, never the reverse (the manager's
+// reoffer callback re-enters Conn under conn.mu).
+func (conn *Conn) pqPSK() (*wgtypes.Key, bool) {
+	if conn.config.PQ == nil {
+		return nil, false
+	}
+	psk, ok := conn.config.PQ.PSK(conn.config.Key)
+	if !ok {
+		return nil, false
+	}
+	return &psk, true
+}
+
+// presharedKey resolves the WireGuard preshared key for the peer. pqPSK is the
+// post-quantum PSK looked up out of band via pqPSK (nil when none is derived yet), passed
+// in so the manager lock is never taken under conn.mu.
+func (conn *Conn) presharedKey(remoteRosenpassKey []byte, pqPSK *wgtypes.Key) *wgtypes.Key {
 	// Post-quantum: once the ML-KEM exchange has derived a PSK for this peer, program
 	// it here so the peer's next WireGuard handshake adopts it. Applied at peer-config
 	// time (bootstrap / reconnect); steady-state rotation is pushed separately.
 	if conn.config.PQ != nil {
-		if psk, ok := conn.config.PQ.PSK(conn.config.Key); ok {
-			return &psk
+		if pqPSK != nil {
+			return pqPSK
 		}
 		if conn.config.PQStrict && conn.pqStrictSentinelKey != nil {
 			// Fail closed: program a non-matching sentinel so no session forms on a
@@ -1154,16 +1186,8 @@ func isRosenpassEnabled(remoteRosenpassPubKey []byte) bool {
 // quantumResistant reports whether the peer's tunnel is post-quantum protected, for
 // the status "Quantum resistance" field: either Rosenpass (the remote advertised a
 // Rosenpass key) or the ML-KEM exchange (a PQ PSK has been derived for this peer).
-func (conn *Conn) quantumResistant(remoteRosenpassPubKey []byte) bool {
-	if isRosenpassEnabled(remoteRosenpassPubKey) {
-		return true
-	}
-	if conn.config.PQ != nil {
-		if _, ok := conn.config.PQ.PSK(conn.config.Key); ok {
-			return true
-		}
-	}
-	return false
+func (conn *Conn) quantumResistant(remoteRosenpassPubKey []byte, pqEstablished bool) bool {
+	return isRosenpassEnabled(remoteRosenpassPubKey) || pqEstablished
 }
 
 func evalConnStatus(in connStatusInputs) guard.ConnStatus {

client/internal/peer/conn_pq_test.go

@@ -53,10 +53,12 @@ func TestConn_presharedKey_PQ(t *testing.T) {
 			c.config.PQ = fakePQ{psk: derivedPSK, ok: true}
 			c.config.PQStrict = strict
 			if strict {
-				sentinel, _ := wgtypes.GenerateKey()
+				sentinel, err := wgtypes.GenerateKey()
+				require.NoError(t, err)
 				c.pqStrictSentinelKey = &sentinel
 			}
-			got := c.presharedKey(nil)
+			pqPSK, _ := c.pqPSK()
+			got := c.presharedKey(nil, pqPSK)
 			require.NotNil(t, got)
 			require.Equal(t, derivedPSK, *got, "the derived PQ PSK must win (strict=%v)", strict)
 		}
@@ -66,7 +68,8 @@ func TestConn_presharedKey_PQ(t *testing.T) {
 		c := newConn()
 		c.config.PQ = fakePQ{ok: false}
 		c.config.PQStrict = false
-		got := c.presharedKey(nil)
+		pqPSK, _ := c.pqPSK()
+		got := c.presharedKey(nil, pqPSK)
 		require.NotNil(t, got, "non-strict must not block")
 		require.Equal(t, nbPSK, *got, "non-strict falls through to the NetBird PSK, not a sentinel")
 	})
@@ -78,7 +81,8 @@ func TestConn_presharedKey_PQ(t *testing.T) {
 		c.config.PQ = fakePQ{ok: false}
 		c.config.PQStrict = true
 		c.pqStrictSentinelKey = &sentinel
-		got := c.presharedKey(nil)
+		pqPSK, _ := c.pqPSK()
+		got := c.presharedKey(nil, pqPSK)
 		require.NotNil(t, got)
 		require.Equal(t, sentinel, *got, "strict must return the blocking sentinel")
 		require.NotEqual(t, nbPSK, *got, "the sentinel must not be the ordinary key")

client/internal/peer/conn_test.go

@@ -257,8 +257,8 @@ func TestConn_presharedKey(t *testing.T) {
 			}
 			conn2.config.RosenpassConfig.PermissiveMode = test.conn2Permissive
 
-			conn1PresharedKey := conn1.presharedKey(conn2.config.RosenpassConfig.PubKey)
-			conn2PresharedKey := conn2.presharedKey(conn1.config.RosenpassConfig.PubKey)
+			conn1PresharedKey := conn1.presharedKey(conn2.config.RosenpassConfig.PubKey, nil)
+			conn2PresharedKey := conn2.presharedKey(conn1.config.RosenpassConfig.PubKey, nil)
 
 			if test.conn1ExpectedInitialKey {
 				if conn1PresharedKey == nil {
@@ -296,14 +296,14 @@ func TestConn_presharedKey_RosenpassManaged(t *testing.T) {
 	// When Rosenpass has already initialized the PSK for this peer,
 	// presharedKey must return nil to avoid UpdatePeer overwriting it.
 	conn.rosenpassInitializedPresharedKeyValidator = func(peerKey string) bool { return true }
-	if k := conn.presharedKey([]byte("remote")); k != nil {
+	if k := conn.presharedKey([]byte("remote"), nil); k != nil {
 		t.Fatalf("expected nil presharedKey when Rosenpass manages PSK, got %v", k)
 	}
 
 	// When Rosenpass hasn't taken over yet, presharedKey should provide
 	// a non-nil initial key (deterministic or from NetBird PSK).
 	conn.rosenpassInitializedPresharedKeyValidator = func(peerKey string) bool { return false }
-	if k := conn.presharedKey([]byte("remote")); k == nil {
+	if k := conn.presharedKey([]byte("remote"), nil); k == nil {
 		t.Fatalf("expected non-nil presharedKey before Rosenpass manages PSK")
 	}
 }

client/internal/peer/handshaker.go

@@ -47,7 +47,7 @@ type OfferAnswer struct {
 
 	// MlkemPort is the peer's ML-KEM PQ service UDP port (bound on its WG overlay
 	// IP) where data-path rekey messages are sent. Zero when not running the exchange.
-	MlkemPort int
+	MlkemPort uint16
 
 	// relay server address
 	RelaySrvAddress string
@@ -221,14 +221,38 @@ func (h *Handshaker) pqControllerReoffer() bool {
 
 // pqRegisterEndpoint feeds the post-quantum handshaker the peer's data-path endpoint
 // (its WG overlay IP plus the advertised pq UDP port) learned from a remote offer/answer.
-func (h *Handshaker) pqRegisterEndpoint(remotePort int) {
-	if h.config.PQ == nil || remotePort < 0 || remotePort > 65535 || len(h.config.WgConfig.AllowedIps) == 0 {
+func (h *Handshaker) pqRegisterEndpoint(remotePort uint16) {
+	if h.config.PQ == nil {
+		return
+	}
+	overlay, ok := h.pqPeerOverlayAddr()
+	if !ok {
 		return
 	}
 	// remotePort may be 0 (the peer omitted it, meaning the default port); the adapter
 	// resolves 0 to DefaultPort.
-	addr := netip.AddrPortFrom(h.config.WgConfig.AllowedIps[0].Addr(), uint16(remotePort))
-	h.config.PQ.SetRemoteAddr(h.config.Key, addr)
+	h.config.PQ.SetRemoteAddr(h.config.Key, netip.AddrPortFrom(overlay, remotePort))
+}
+
+// pqPeerOverlayAddr picks the peer's overlay address for the pq data path. The pq
+// transport binds on the local WG overlay IPv4, so an IPv4 AllowedIP is preferred; a v6
+// prefix is used only when this interface actually has a v6 overlay, and never in place
+// of a usable v4. Returns false when no suitable address exists.
+func (h *Handshaker) pqPeerOverlayAddr() (netip.Addr, bool) {
+	var v6 netip.Addr
+	for _, p := range h.config.WgConfig.AllowedIps {
+		a := p.Addr().Unmap()
+		if a.Is4() {
+			return a, true
+		}
+		if a.Is6() && !v6.IsValid() {
+			v6 = a
+		}
+	}
+	if v6.IsValid() && h.config.WgConfig.WgInterface.Address().HasIPv6() {
+		return v6, true
+	}
+	return netip.Addr{}, false
 }
 
 func (h *Handshaker) SendOffer() error {

client/internal/peer/signaler.go

@@ -64,7 +64,7 @@ func (s *Signaler) signalOfferAnswer(offerAnswer OfferAnswer, remoteKey string,
 		RosenpassPubKey: offerAnswer.RosenpassPubKey,
 		RosenpassAddr:   offerAnswer.RosenpassAddr,
 		MlkemPayload:    offerAnswer.MlkemPayload,
-		MlkemPort:       offerAnswer.MlkemPort,
+		MlkemPort:       int(offerAnswer.MlkemPort),
 		RelaySrvAddress: offerAnswer.RelaySrvAddress,
 		RelaySrvIP:      offerAnswer.RelaySrvIP,
 		SessionID:       sessionIDBytes,