Merge remote-tracking branch 'origin/main' into mdm_integration

client/cmd/kubernetes.go

@@ -0,0 +1,301 @@
+package cmd
+
+import (
+	"context"
+	"crypto/tls"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"io"
+	"net"
+	"net/http"
+	"net/url"
+	"os"
+	"path/filepath"
+	"slices"
+	"strings"
+
+	"github.com/goccy/go-yaml"
+	log "github.com/sirupsen/logrus"
+	"github.com/spf13/cobra"
+
+	"github.com/netbirdio/netbird/client/proto"
+)
+
+const (
+	KubernetesDNSSuffix = "netbird-kubeapi-proxy"
+)
+
+var kubernetesCmd = &cobra.Command{
+	Use:   "kubernetes",
+	Short: "Kubernetes cluster commands.",
+	Long:  "Kubernetes cluster commands.",
+}
+
+var kubernetesListCmd = &cobra.Command{
+	Use:   "list",
+	RunE:  kubernetesList,
+	Short: "List Kubernetes clusters.",
+	Long:  "List Kubernetes clusters by discovering NetBird peers running netbird-kubeapi-proxy.",
+}
+
+var kubernetesWriteKubeconfigCmd = &cobra.Command{
+	Use:   "write-kubeconfig",
+	RunE:  kubernetesWriteKubeconfig,
+	Args:  cobra.ExactArgs(1),
+	Short: "Write kubeconfig for a Kubernetes cluster.",
+	Long:  "Updates kubeconfig in place to allow token-less access to the Kubernetes cluster through NetBird.",
+}
+
+func init() {
+	kubernetesWriteKubeconfigCmd.Flags().String("kubeconfig", "", "path to kubeconfig file")
+}
+
+func kubernetesList(cmd *cobra.Command, _ []string) error {
+	conn, err := getClient(cmd)
+	if err != nil {
+		return err
+	}
+	defer conn.Close()
+	client := proto.NewDaemonServiceClient(conn)
+	statusResp, err := client.Status(cmd.Context(), &proto.StatusRequest{GetFullPeerStatus: true})
+	if err != nil {
+		return err
+	}
+
+	kcs, err := getKubernetesClusters(cmd.Context(), statusResp.FullStatus.Peers, "")
+	if err != nil {
+		return err
+	}
+	if len(kcs) == 0 {
+		cmd.Println("No Kubernetes clusters available.")
+		return nil
+	}
+	cmd.Println("Available Kubernetes clusters:")
+	for _, k := range kcs {
+		cmd.Printf("\n  - Name: %s\n    FQDN: %s\n    Version: %s\n", k.name, k.url.Host, k.version)
+	}
+	return nil
+}
+
+func kubernetesWriteKubeconfig(cmd *cobra.Command, args []string) error {
+	kubeconfigPath, err := resolveKubeconfigPath(cmd)
+	if err != nil {
+		return err
+	}
+
+	conn, err := getClient(cmd)
+	if err != nil {
+		return err
+	}
+	defer conn.Close()
+	client := proto.NewDaemonServiceClient(conn)
+	statusResp, err := client.Status(cmd.Context(), &proto.StatusRequest{GetFullPeerStatus: true})
+	if err != nil {
+		return err
+	}
+
+	clusterName := args[0]
+	kcs, err := getKubernetesClusters(cmd.Context(), statusResp.FullStatus.Peers, clusterName)
+	if err != nil {
+		return err
+	}
+	if len(kcs) == 0 {
+		return fmt.Errorf("kubernetes cluster named %s not found", clusterName)
+	}
+	if len(kcs) > 1 {
+		return fmt.Errorf("too many Kubernetes clusters returned")
+	}
+	err = writeKubeconfig(kubeconfigPath, kcs[0])
+	if err != nil {
+		return err
+	}
+	return nil
+}
+
+type kubernetesCluster struct {
+	name    string
+	url     *url.URL
+	version string
+}
+
+func getKubernetesClusters(ctx context.Context, peers []*proto.PeerState, nameFilter string) ([]kubernetesCluster, error) {
+	transport := http.DefaultTransport.(*http.Transport).Clone()
+	transport.TLSClientConfig = &tls.Config{
+		InsecureSkipVerify: true,
+	}
+	httpClient := &http.Client{
+		Transport: transport,
+	}
+	resolver := net.Resolver{
+		// Required so both DNS records are returned.
+		// https://github.com/golang/go/issues/17093
+		PreferGo: true,
+	}
+
+	kcs := []kubernetesCluster{}
+	attempted := map[string]struct{}{}
+	for _, peer := range peers {
+		fqdns, err := resolver.LookupAddr(ctx, peer.IP)
+		if err != nil {
+			return nil, err
+		}
+		for _, fqdn := range fqdns {
+			if _, ok := attempted[fqdn]; ok {
+				continue
+			}
+			attempted[fqdn] = struct{}{}
+			comps := strings.Split(fqdn, ".")
+			if len(comps) < 2 {
+				continue
+			}
+			if comps[1] != KubernetesDNSSuffix {
+				continue
+			}
+			if nameFilter != "" && nameFilter != comps[0] {
+				continue
+			}
+			clusterURL, clusterVersion, err := fingerprintClusters(ctx, httpClient, fqdn)
+			if err != nil {
+				log.Debugf("could not fingerprint Kubernetes cluster %s %q", fqdn, err)
+				continue
+			}
+			kc := kubernetesCluster{
+				name:    comps[0],
+				url:     clusterURL,
+				version: clusterVersion,
+			}
+			if nameFilter != "" {
+				return []kubernetesCluster{kc}, nil
+			}
+			kcs = append(kcs, kc)
+		}
+	}
+	return kcs, nil
+}
+
+func fingerprintClusters(ctx context.Context, httpClient *http.Client, fqdn string) (*url.URL, string, error) {
+	clusterURL, err := url.Parse("https://" + fqdn)
+	if err != nil {
+		return nil, "", err
+	}
+	versionURL, err := clusterURL.Parse("/version")
+	if err != nil {
+		return nil, "", err
+	}
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, versionURL.String(), nil)
+	if err != nil {
+		return nil, "", err
+	}
+	resp, err := httpClient.Do(req)
+	if err != nil {
+		return nil, "", err
+	}
+	defer resp.Body.Close()
+	if resp.StatusCode != http.StatusOK {
+		return nil, "", fmt.Errorf("expected %d response but got %s", http.StatusOK, resp.Status)
+	}
+	b, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return nil, "", err
+	}
+	versionData := map[string]string{}
+	err = json.Unmarshal(b, &versionData)
+	if err != nil {
+		return nil, "", err
+	}
+	version, ok := versionData["gitVersion"]
+	if !ok {
+		return nil, "", errors.New("no version found in response")
+	}
+	return clusterURL, version, nil
+}
+
+func resolveKubeconfigPath(cmd *cobra.Command) (string, error) {
+	if cmd.Flags().Changed("kubeconfig") {
+		path, err := cmd.Flags().GetString("kubeconfig")
+		if err != nil {
+			return "", err
+		}
+		return path, nil
+	}
+	if env := os.Getenv("KUBECONFIG"); env != "" {
+		return env, nil
+	}
+	home, err := os.UserHomeDir()
+	if err != nil {
+		return "", fmt.Errorf("could not determine home directory: %w", err)
+	}
+	return filepath.Join(home, ".kube", "config"), nil
+}
+
+func writeKubeconfig(kubeconfigPath string, kc kubernetesCluster) error {
+	b, err := os.ReadFile(kubeconfigPath)
+	if err != nil && !errors.Is(err, os.ErrNotExist) {
+		return err
+	}
+	var cfg map[string]any
+	if err := yaml.Unmarshal(b, &cfg); err != nil {
+		return err
+	}
+	if cfg == nil {
+		cfg = map[string]any{
+			"apiVersion": "v1",
+			"kind":       "Config",
+		}
+	}
+
+	cfg["clusters"] = appendWithName(cfg["clusters"], map[string]any{
+		"name": kc.name,
+		"cluster": map[string]any{
+			"server":                   kc.url.String(),
+			"insecure-skip-tls-verify": true,
+		},
+	})
+	cfg["users"] = appendWithName(cfg["users"], map[string]any{
+		"name": "netbird",
+		"user": map[string]any{
+			"token": "none",
+		},
+	})
+	cfg["contexts"] = appendWithName(cfg["contexts"], map[string]any{
+		"name": kc.name,
+		"context": map[string]any{
+			"cluster":   kc.name,
+			"user":      "netbird",
+			"namespace": "default",
+		},
+	})
+	cfg["current-context"] = kc.name
+
+	out, err := yaml.Marshal(cfg)
+	if err != nil {
+		return err
+	}
+	if err := os.WriteFile(kubeconfigPath, out, 0o600); err != nil {
+		return err
+	}
+	return nil
+}
+
+func appendWithName(data any, add map[string]any) any {
+	if data == nil {
+		return []any{add}
+	}
+	v, ok := data.([]any)
+	if !ok {
+		return []any{add}
+	}
+	i := slices.IndexFunc(v, func(item any) bool {
+		m, ok := item.(map[string]any)
+		if !ok {
+			return false
+		}
+		return m["name"] == add["name"]
+	})
+	if i == -1 {
+		return append(v, add)
+	}
+	v[i] = add
+	return v
+}

client/cmd/kubernetes_test.go

@@ -0,0 +1,120 @@
+package cmd
+
+import (
+	"net/http"
+	"net/http/httptest"
+	"net/url"
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/spf13/cobra"
+	"github.com/stretchr/testify/require"
+)
+
+func TestFingerprintClusters(t *testing.T) {
+	t.Parallel()
+
+	srv := httptest.NewTLSServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		//nolint: errcheck
+		w.Write([]byte(`{"gitVersion": "foobar"}`))
+	}))
+	defer srv.Close()
+
+	clusterURL, clusterVersion, err := fingerprintClusters(t.Context(), srv.Client(), srv.Listener.Addr().String())
+	require.NoError(t, err)
+	require.Equal(t, srv.URL, clusterURL.String())
+	require.Equal(t, "foobar", clusterVersion)
+}
+
+func TestResolveKubeconfigPath(t *testing.T) {
+	home, err := os.UserHomeDir()
+	if err != nil {
+		t.Fatalf("could not determine home directory: %v", err)
+	}
+	defaultPath := filepath.Join(home, ".kube", "config")
+	path, err := resolveKubeconfigPath(&cobra.Command{})
+	require.NoError(t, err)
+	require.Equal(t, defaultPath, path)
+
+	flagPath := "flag-path"
+	cmd := &cobra.Command{}
+	cmd.Flags().String("kubeconfig", "", "")
+	err = cmd.Flags().Set("kubeconfig", flagPath)
+	require.NoError(t, err)
+	path, err = resolveKubeconfigPath(cmd)
+	require.NoError(t, err)
+	require.Equal(t, flagPath, path)
+
+	envPath := "env-path"
+	t.Setenv("KUBECONFIG", envPath)
+	path, err = resolveKubeconfigPath(&cobra.Command{})
+	require.NoError(t, err)
+	require.Equal(t, envPath, path)
+}
+
+func TestWriteKubeconfig(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name     string
+		existing string
+	}{
+		{
+			name: "empty file",
+		},
+		{
+			name: "existing content",
+			existing: `apiVersion: v1
+clusters:
+- cluster:
+    insecure-skip-tls-verify: true
+    server: https://foobar.com
+  name: foo
+current-context: test
+kind: Config
+users: []
+`,
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			kubeconfigPath := filepath.Join(t.TempDir(), "config")
+			err := os.WriteFile(kubeconfigPath, []byte(tt.existing), 0o644)
+			require.NoError(t, err)
+
+			kc := kubernetesCluster{
+				name: "foo",
+				url:  &url.URL{Scheme: "https", Host: "example.com"},
+			}
+			err = writeKubeconfig(kubeconfigPath, kc)
+			require.NoError(t, err)
+
+			b, err := os.ReadFile(kubeconfigPath)
+			require.NoError(t, err)
+			expected := `apiVersion: v1
+clusters:
+- cluster:
+    insecure-skip-tls-verify: true
+    server: https://example.com
+  name: foo
+contexts:
+- context:
+    cluster: foo
+    namespace: default
+    user: netbird
+  name: foo
+current-context: foo
+kind: Config
+users:
+- name: netbird
+  user:
+    token: none
+`
+			require.Equal(t, expected, string(b))
+		})
+	}
+
+}

client/cmd/root.go

@@ -182,6 +182,11 @@ func init() {
 	debugCmd.AddCommand(persistenceCmd)
 	debugCmd.AddCommand(debugConfigCmd)
 
+	// kubernetes commands
+	rootCmd.AddCommand(kubernetesCmd)
+	kubernetesCmd.AddCommand(kubernetesListCmd)
+	kubernetesCmd.AddCommand(kubernetesWriteKubeconfigCmd)
+
 	// profile commands
 	profileCmd.AddCommand(profileListCmd)
 	profileCmd.AddCommand(profileAddCmd)

client/embed/embed.go

@@ -279,6 +279,10 @@ func (c *Client) Start(startCtx context.Context) error {
 
 	select {
 	case <-startCtx.Done():
+		// Cancel the client context before stopping: Engine.Start blocks on the
+		// signal stream while holding the engine mutex and only unblocks on
+		// cancellation. Stopping first would deadlock on that mutex.
+		cancel()
 		if stopErr := client.Stop(); stopErr != nil {
 			return fmt.Errorf("stop error after context done. Stop error: %w. Context done: %w", stopErr, startCtx.Err())
 		}

client/embed/embed_test.go

@@ -0,0 +1,168 @@
+package embed
+
+import (
+	"context"
+	"net"
+	"testing"
+	"time"
+
+	"github.com/golang/mock/gomock"
+	"github.com/stretchr/testify/require"
+	"google.golang.org/grpc"
+
+	"github.com/netbirdio/netbird/management/internals/controllers/network_map/controller"
+	"github.com/netbirdio/netbird/management/internals/controllers/network_map/update_channel"
+	"github.com/netbirdio/netbird/management/internals/modules/peers"
+	"github.com/netbirdio/netbird/management/internals/modules/peers/ephemeral/manager"
+	"github.com/netbirdio/netbird/management/internals/server/config"
+	nbgrpc "github.com/netbirdio/netbird/management/internals/shared/grpc"
+	mgmt "github.com/netbirdio/netbird/management/server"
+	"github.com/netbirdio/netbird/management/server/activity"
+	nbcache "github.com/netbirdio/netbird/management/server/cache"
+	"github.com/netbirdio/netbird/management/server/groups"
+	"github.com/netbirdio/netbird/management/server/integrations/integrated_validator/validator"
+	"github.com/netbirdio/netbird/management/server/integrations/port_forwarding"
+	"github.com/netbirdio/netbird/management/server/job"
+	"github.com/netbirdio/netbird/management/server/permissions"
+	"github.com/netbirdio/netbird/management/server/settings"
+	"github.com/netbirdio/netbird/management/server/store"
+	"github.com/netbirdio/netbird/management/server/telemetry"
+	"github.com/netbirdio/netbird/management/server/types"
+	mgmtProto "github.com/netbirdio/netbird/shared/management/proto"
+	"github.com/netbirdio/netbird/util"
+)
+
+const testSetupKey = "A2C8E62B-38F5-4553-B31E-DD66C696CEBB"
+
+// TestClientStartTimeoutRollback reproduces a deadlock between Engine.Start and
+// Engine.Stop. The signal endpoint accepts gRPC connections but never serves the
+// SignalExchange service, so Engine.Start parks in WaitStreamConnected while
+// holding the engine mutex. When the Start context expires, the rollback path
+// calls ConnectClient.Stop, which must not block forever acquiring that mutex.
+func TestClientStartTimeoutRollback(t *testing.T) {
+	signalAddr := startBlackholeSignal(t)
+	mgmAddr := startManagement(t, signalAddr)
+
+	wgPort := 0
+	client, err := New(Options{
+		DeviceName:    "embed-rollback-test",
+		SetupKey:      testSetupKey,
+		ManagementURL: "http://" + mgmAddr,
+		WireguardPort: &wgPort,
+	})
+	require.NoError(t, err, "embed client creation must succeed")
+
+	startCtx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+
+	startErr := make(chan error, 1)
+	go func() {
+		startErr <- client.Start(startCtx)
+	}()
+
+	select {
+	case err := <-startErr:
+		require.ErrorIs(t, err, context.DeadlineExceeded)
+	case <-time.After(60 * time.Second):
+		t.Fatal("client.Start did not return after its context expired: Engine.Stop deadlocked against Engine.Start waiting for the signal stream")
+	}
+}
+
+// startBlackholeSignal starts a gRPC server without the SignalExchange service
+// registered. Connections succeed, but the signal stream can never be
+// established, which keeps Engine.Start parked in WaitStreamConnected.
+func startBlackholeSignal(t *testing.T) string {
+	t.Helper()
+
+	lis, err := net.Listen("tcp", "localhost:0")
+	require.NoError(t, err)
+
+	s := grpc.NewServer()
+	go func() {
+		if err := s.Serve(lis); err != nil {
+			t.Error(err)
+		}
+	}()
+	t.Cleanup(s.Stop)
+
+	return lis.Addr().String()
+}
+
+func startManagement(t *testing.T, signalAddr string) string {
+	t.Helper()
+
+	cfg := &config.Config{
+		Stuns:      []*config.Host{},
+		TURNConfig: &config.TURNConfig{},
+		Relay: &config.Relay{
+			Addresses:      []string{"127.0.0.1:1234"},
+			CredentialsTTL: util.Duration{Duration: time.Hour},
+			Secret:         "222222222222222222",
+		},
+		Signal: &config.Host{
+			Proto: "http",
+			URI:   signalAddr,
+		},
+		Datadir:    t.TempDir(),
+		HttpConfig: nil,
+	}
+
+	lis, err := net.Listen("tcp", "localhost:0")
+	require.NoError(t, err)
+
+	s := grpc.NewServer()
+
+	testStore, cleanUp, err := store.NewTestStoreFromSQL(context.Background(), "../testdata/store.sql", cfg.Datadir)
+	require.NoError(t, err)
+	t.Cleanup(cleanUp)
+
+	eventStore := &activity.InMemoryEventStore{}
+
+	permissionsManager := permissions.NewManager(testStore)
+	peersManager := peers.NewManager(testStore, permissionsManager)
+	jobManager := job.NewJobManager(nil, testStore, peersManager)
+
+	cacheStore, err := nbcache.NewStore(context.Background(), 100*time.Millisecond, 300*time.Millisecond, 100)
+	require.NoError(t, err)
+
+	iv, err := validator.NewIntegratedValidator(context.Background(), peersManager, nil, eventStore, cacheStore)
+	require.NoError(t, err)
+	metrics, err := telemetry.NewDefaultAppMetrics(context.Background())
+	require.NoError(t, err)
+
+	ctrl := gomock.NewController(t)
+	t.Cleanup(ctrl.Finish)
+	settingsMockManager := settings.NewMockManager(ctrl)
+	settingsMockManager.EXPECT().
+		GetSettings(gomock.Any(), gomock.Any(), gomock.Any()).
+		Return(&types.Settings{}, nil).
+		AnyTimes()
+	settingsMockManager.EXPECT().
+		GetExtraSettings(gomock.Any(), gomock.Any()).
+		Return(&types.ExtraSettings{}, nil).
+		AnyTimes()
+
+	groupsManager := groups.NewManagerMock()
+
+	updateManager := update_channel.NewPeersUpdateManager(metrics)
+	requestBuffer := mgmt.NewAccountRequestBuffer(context.Background(), testStore)
+	networkMapController := controller.NewController(context.Background(), testStore, metrics, updateManager, requestBuffer, mgmt.MockIntegratedValidator{}, settingsMockManager, "netbird.selfhosted", port_forwarding.NewControllerMock(), manager.NewEphemeralManager(testStore, peersManager), cfg)
+	accountManager, err := mgmt.BuildManager(context.Background(), cfg, testStore, networkMapController, jobManager, nil, "", eventStore, nil, false, iv, metrics, port_forwarding.NewControllerMock(), settingsMockManager, permissionsManager, false, cacheStore)
+	require.NoError(t, err)
+
+	secretsManager, err := nbgrpc.NewTimeBasedAuthSecretsManager(updateManager, cfg.TURNConfig, cfg.Relay, settingsMockManager, groupsManager)
+	require.NoError(t, err)
+
+	mgmtServer, err := nbgrpc.NewServer(cfg, accountManager, settingsMockManager, jobManager, secretsManager, nil, nil, &mgmt.MockIntegratedValidator{}, networkMapController, nil, nil)
+	require.NoError(t, err)
+	mgmtProto.RegisterManagementServiceServer(s, mgmtServer)
+
+	go func() {
+		if err := s.Serve(lis); err != nil {
+			t.Error(err)
+		}
+	}()
+	t.Cleanup(s.Stop)
+
+	return lis.Addr().String()
+}

client/firewall/iptables/acl_linux.go

@@ -3,6 +3,7 @@ package iptables
 import (
 	"errors"
 	"fmt"
+	"maps"
 	"net"
 	"slices"
 
@@ -421,12 +422,17 @@ func (m *aclManager) updateState() {
 	currentState.Lock()
 	defer currentState.Unlock()
 
+	// Clone the maps so the persisted state holds a private snapshot. The
+	// live maps keep being mutated by subsequent rule operations while the
+	// state manager marshals the state from its periodic-save goroutine.
+	// Sharing them by reference races the two and aborts the process with a
+	// concurrent map iteration and write.
 	if m.v6 {
-		currentState.ACLEntries6 = m.entries
-		currentState.ACLIPsetStore6 = m.ipsetStore
+		currentState.ACLEntries6 = maps.Clone(m.entries)
+		currentState.ACLIPsetStore6 = m.ipsetStore.clone()
 	} else {
-		currentState.ACLEntries = m.entries
-		currentState.ACLIPsetStore = m.ipsetStore
+		currentState.ACLEntries = maps.Clone(m.entries)
+		currentState.ACLIPsetStore = m.ipsetStore.clone()
 	}
 
 	if err := m.stateManager.UpdateState(currentState); err != nil {

client/firewall/iptables/router_linux.go

@@ -4,6 +4,7 @@ package iptables
 
 import (
 	"fmt"
+	"maps"
 	"net/netip"
 	"strconv"
 	"strings"
@@ -749,11 +750,17 @@ func (r *router) updateState() {
 	currentState.Lock()
 	defer currentState.Unlock()
 
+	// Clone the rule map so the persisted state holds a private snapshot. The
+	// live map keeps being mutated by subsequent rule operations while the
+	// state manager marshals the state from its periodic-save goroutine.
+	// Sharing it by reference races the two and aborts the process with a
+	// concurrent map iteration and write. The ipset counter guards itself
+	// during marshaling, so it can be shared directly.
 	if r.v6 {
-		currentState.RouteRules6 = r.rules
+		currentState.RouteRules6 = maps.Clone(r.rules)
 		currentState.RouteIPsetCounter6 = r.ipsetCounter
 	} else {
-		currentState.RouteRules = r.rules
+		currentState.RouteRules = maps.Clone(r.rules)
 		currentState.RouteIPsetCounter = r.ipsetCounter
 	}
 

client/firewall/iptables/rulestore_linux.go

@@ -1,6 +1,9 @@
 package iptables
 
-import "encoding/json"
+import (
+	"encoding/json"
+	"maps"
+)
 
 type ipList struct {
 	ips map[string]struct{}
@@ -19,6 +22,14 @@ func (s *ipList) addIP(ip string) {
 	s.ips[ip] = struct{}{}
 }
 
+// clone returns a deep copy of the ipList with its own ips map.
+func (s *ipList) clone() *ipList {
+	if s == nil {
+		return nil
+	}
+	return &ipList{ips: maps.Clone(s.ips)}
+}
+
 // MarshalJSON implements json.Marshaler
 func (s *ipList) MarshalJSON() ([]byte, error) {
 	return json.Marshal(struct {
@@ -55,6 +66,19 @@ func newIpsetStore() *ipsetStore {
 	}
 }
 
+// clone returns a deep copy of the ipsetStore with its own ipsets map and
+// independent ipList entries.
+func (s *ipsetStore) clone() *ipsetStore {
+	if s == nil {
+		return nil
+	}
+	cloned := &ipsetStore{ipsets: make(map[string]*ipList, len(s.ipsets))}
+	for name, list := range s.ipsets {
+		cloned.ipsets[name] = list.clone()
+	}
+	return cloned
+}
+
 func (s *ipsetStore) ipset(ipsetName string) (*ipList, bool) {
 	r, ok := s.ipsets[ipsetName]
 	return r, ok

client/internal/dns/server.go

@@ -777,13 +777,24 @@ func (s *DefaultServer) applyHostConfig() {
 // context is released rather than leaked until GC.
 func (s *DefaultServer) registerFallback() {
 	originalNameservers := s.hostManager.getOriginalNameservers()
-	if len(originalNameservers) == 0 {
+
+	serverIP := s.service.RuntimeIP()
+	var servers []netip.AddrPort
+	for _, ns := range originalNameservers {
+		if ns == serverIP {
+			log.Debugf("skipping original nameserver %s as it is the same as the server IP %s", ns, serverIP)
+			continue
+		}
+		servers = append(servers, netip.AddrPortFrom(ns, DefaultPort))
+	}
+
+	if len(servers) == 0 {
 		log.Debugf("no fallback upstreams to register; clearing PriorityFallback handler")
 		s.clearFallback()
 		return
 	}
 
-	log.Infof("registering original nameservers %v as upstream handlers with priority %d", originalNameservers, PriorityFallback)
+	log.Infof("registering original nameservers %v as upstream handlers with priority %d", servers, PriorityFallback)
 
 	handler, err := newUpstreamResolver(
 		s.ctx,
@@ -797,11 +808,6 @@ func (s *DefaultServer) registerFallback() {
 		return
 	}
 	handler.selectedRoutes = s.selectedRoutes
-
-	var servers []netip.AddrPort
-	for _, ns := range originalNameservers {
-		servers = append(servers, netip.AddrPortFrom(ns, DefaultPort))
-	}
 	handler.addRace(servers)
 
 	prev := s.fallbackHandler

client/internal/engine.go

@@ -880,62 +880,25 @@ func (e *Engine) handleSync(update *mgmProto.SyncResponse) error {
 		e.handleAutoUpdateVersion(update.NetworkMap.PeerConfig.AutoUpdate)
 	}
 
-	if update.GetNetbirdConfig() != nil {
-		wCfg := update.GetNetbirdConfig()
-		err := e.updateTURNs(wCfg.GetTurns())
-		if err != nil {
-			return fmt.Errorf("update TURNs: %w", err)
-		}
+	if err := e.updateNetbirdConfig(update.GetNetbirdConfig()); err != nil {
+		return err
+	}
 
-		err = e.updateSTUNs(wCfg.GetStuns())
-		if err != nil {
-			return fmt.Errorf("update STUNs: %w", err)
-		}
-
-		var stunTurn []*stun.URI
-		stunTurn = append(stunTurn, e.STUNs...)
-		stunTurn = append(stunTurn, e.TURNs...)
-		e.stunTurn.Store(stunTurn)
-
-		err = e.handleRelayUpdate(wCfg.GetRelay())
-		if err != nil {
-			return err
-		}
-
-		err = e.handleFlowUpdate(wCfg.GetFlow())
-		if err != nil {
-			return fmt.Errorf("handle the flow configuration: %w", err)
-		}
-
-		if err := e.PopulateNetbirdConfig(wCfg, nil); err != nil {
-			log.Warnf("Failed to update DNS server config: %v", err)
-		}
-
-		// todo update signal
+	// Posture checks are bound to the network map presence:
+	//   NetworkMap != nil, checks present -> apply the received checks
+	//   NetworkMap != nil, checks nil     -> posture checks were removed, clear them
+	//   NetworkMap == nil                 -> config-only update (e.g. relay token rotation),
+	//                                        leave the previously applied checks untouched
+	nm := update.GetNetworkMap()
+	if nm == nil {
+		return nil
 	}
 
 	if err := e.updateChecksIfNew(update.Checks); err != nil {
 		return err
 	}
 
-	nm := update.GetNetworkMap()
-	if nm == nil {
-		return nil
-	}
-
-	// Persist sync response under the dedicated lock (syncRespMux), not under syncMsgMux.
-	// A non-nil syncStore is what marks persistence as enabled. Hold the lock for
-	// the whole Set so the store cannot be cleared (disabled / engine close)
-	// mid-call and have this write resurrect a file that was just removed.
-	e.syncRespMux.RLock()
-	if e.syncStore != nil {
-		if err := e.syncStore.Set(update); err != nil {
-			log.Errorf("failed to persist sync response: %v", err)
-		} else {
-			log.Debugf("sync response persisted with serial %d", nm.GetSerial())
-		}
-	}
-	e.syncRespMux.RUnlock()
+	e.persistSyncResponse(update)
 
 	// only apply new changes and ignore old ones
 	if err := e.updateNetworkMap(nm); err != nil {
@@ -947,6 +910,64 @@ func (e *Engine) handleSync(update *mgmProto.SyncResponse) error {
 	return nil
 }
 
+// updateNetbirdConfig applies the management-provided NetBird configuration:
+// STUN/TURN and relay servers, flow logging and DNS settings. A nil config is a no-op,
+// which is the case for sync updates carrying only a network map.
+func (e *Engine) updateNetbirdConfig(wCfg *mgmProto.NetbirdConfig) error {
+	if wCfg == nil {
+		return nil
+	}
+
+	if err := e.updateTURNs(wCfg.GetTurns()); err != nil {
+		return fmt.Errorf("update TURNs: %w", err)
+	}
+
+	if err := e.updateSTUNs(wCfg.GetStuns()); err != nil {
+		return fmt.Errorf("update STUNs: %w", err)
+	}
+
+	var stunTurn []*stun.URI
+	stunTurn = append(stunTurn, e.STUNs...)
+	stunTurn = append(stunTurn, e.TURNs...)
+	e.stunTurn.Store(stunTurn)
+
+	if err := e.handleRelayUpdate(wCfg.GetRelay()); err != nil {
+		return err
+	}
+
+	if err := e.handleFlowUpdate(wCfg.GetFlow()); err != nil {
+		return fmt.Errorf("handle the flow configuration: %w", err)
+	}
+
+	if err := e.PopulateNetbirdConfig(wCfg, nil); err != nil {
+		log.Warnf("Failed to update DNS server config: %v", err)
+	}
+
+	// todo update signal
+
+	return nil
+}
+
+// persistSyncResponse stores the full sync response so it can be restored on the next
+// startup. Persistence is enabled only when syncStore is set. The dedicated syncRespMux
+// (not syncMsgMux) is held for the whole Set so the store cannot be cleared (disabled /
+// engine close) mid-call and have this write resurrect a file that was just removed.
+func (e *Engine) persistSyncResponse(update *mgmProto.SyncResponse) {
+	e.syncRespMux.RLock()
+	defer e.syncRespMux.RUnlock()
+
+	if e.syncStore == nil {
+		return
+	}
+
+	if err := e.syncStore.Set(update); err != nil {
+		log.Errorf("failed to persist sync response: %v", err)
+		return
+	}
+
+	log.Debugf("sync response persisted with serial %d", update.GetNetworkMap().GetSerial())
+}
+
 func (e *Engine) handleRelayUpdate(update *mgmProto.RelayConfig) error {
 	if update != nil {
 		// when we receive token we expect valid address list too

client/internal/routemanager/manager.go

@@ -700,6 +700,13 @@ func resolveURLsToIPs(urls []string) []net.IP {
 
 // updateRouteSelectorFromManagement updates the route selector based on the isSelected status from the management server
 func (m *DefaultManager) updateRouteSelectorFromManagement(clientRoutes route.HAMap) {
+	// An explicit user "deselect all" must not be overridden by management auto-apply.
+	// Auto-applying an exit node here would call SelectRoutes, which clears the
+	// deselect-all flag and re-enables every route the user turned off.
+	if m.routeSelector.IsDeselectAll() {
+		return
+	}
+
 	exitNodeInfo := m.collectExitNodeInfo(clientRoutes)
 	if len(exitNodeInfo.allIDs) == 0 {
 		return

client/internal/routemanager/selector_management_test.go

@@ -0,0 +1,71 @@
+package routemanager
+
+import (
+	"net/netip"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/client/internal/routeselector"
+	"github.com/netbirdio/netbird/route"
+)
+
+func exitNodeRoutes(netID route.NetID, skipAutoApply bool) route.HAMap {
+	haID := route.HAUniqueID(string(netID) + "|0.0.0.0/0")
+	return route.HAMap{
+		haID: []*route.Route{
+			{
+				ID:            "r-" + route.ID(netID),
+				NetID:         netID,
+				Network:       netip.MustParsePrefix("0.0.0.0/0"),
+				NetworkType:   route.IPv4Network,
+				Enabled:       true,
+				SkipAutoApply: skipAutoApply,
+			},
+		},
+	}
+}
+
+func TestUpdateRouteSelectorFromManagement(t *testing.T) {
+	t.Run("management auto-apply selects exit node without user selection", func(t *testing.T) {
+		m := &DefaultManager{routeSelector: routeselector.NewRouteSelector()}
+		routes := exitNodeRoutes("exit1", false)
+
+		m.updateRouteSelectorFromManagement(routes)
+
+		require.True(t, m.routeSelector.IsSelected("exit1"), "auto-apply exit node should be selected")
+		require.Len(t, m.routeSelector.FilterSelectedExitNodes(routes), 1, "selected exit node should pass the filter")
+	})
+
+	t.Run("management SkipAutoApply leaves exit node deselected", func(t *testing.T) {
+		m := &DefaultManager{routeSelector: routeselector.NewRouteSelector()}
+		routes := exitNodeRoutes("exit1", true)
+
+		m.updateRouteSelectorFromManagement(routes)
+
+		require.False(t, m.routeSelector.IsSelected("exit1"), "SkipAutoApply exit node should not be selected")
+		require.Empty(t, m.routeSelector.FilterSelectedExitNodes(routes), "deselected exit node should be filtered out")
+	})
+
+	t.Run("user selection is not overridden by management", func(t *testing.T) {
+		m := &DefaultManager{routeSelector: routeselector.NewRouteSelector()}
+		require.NoError(t, m.routeSelector.SelectRoutes([]route.NetID{"exit1"}, true, []route.NetID{"exit1"}))
+		routes := exitNodeRoutes("exit1", true)
+
+		m.updateRouteSelectorFromManagement(routes)
+
+		require.True(t, m.routeSelector.IsSelected("exit1"), "explicit user selection must survive a management sync that wants to skip auto-apply")
+		require.Len(t, m.routeSelector.FilterSelectedExitNodes(routes), 1, "user-selected exit node should pass the filter")
+	})
+
+	t.Run("deselect-all is preserved across a management sync", func(t *testing.T) {
+		m := &DefaultManager{routeSelector: routeselector.NewRouteSelector()}
+		m.routeSelector.DeselectAllRoutes()
+		routes := exitNodeRoutes("exit1", false)
+
+		m.updateRouteSelectorFromManagement(routes)
+
+		require.True(t, m.routeSelector.IsDeselectAll(), "an explicit deselect-all must not be cleared by management auto-apply")
+		require.Empty(t, m.routeSelector.FilterSelectedExitNodes(routes), "no routes should be selected while deselect-all is set")
+	})
+}

client/internal/routeselector/routeselector.go

@@ -116,6 +116,14 @@ func (rs *RouteSelector) DeselectAllRoutes() {
 	clear(rs.selectedRoutes)
 }
 
+// IsDeselectAll reports whether the user has explicitly deselected all routes.
+func (rs *RouteSelector) IsDeselectAll() bool {
+	rs.mu.RLock()
+	defer rs.mu.RUnlock()
+
+	return rs.deselectAll
+}
+
 // IsSelected checks if a specific route is selected.
 func (rs *RouteSelector) IsSelected(routeID route.NetID) bool {
 	rs.mu.RLock()