Merge branch 'main' of github.com:netbirdio/netbird into client-json-socket

client/cmd/debug.go

@@ -130,7 +130,7 @@ func debugConfigDump(cmd *cobra.Command, _ []string) error {
 
 	client := proto.NewDaemonServiceClient(conn)
 	resp, err := client.GetConfig(cmd.Context(), &proto.GetConfigRequest{
-		ProfileName: activeProf.Name,
+		ProfileName: string(activeProf.ID),
 		Username:    currUser.Username,
 	})
 	if err != nil {

client/cmd/service.go

@@ -5,6 +5,7 @@ package cmd
 import (
 	"context"
 	"fmt"
+	"net/http"
 	"runtime"
 	"strings"
 	"sync"
@@ -22,15 +23,21 @@ var serviceCmd = &cobra.Command{
 	Short: "Manage the NetBird daemon service",
 }
 
+const defaultJSONSocket = "unix:///var/run/netbird-http.sock"
+
 var (
-	serviceName    string
-	serviceEnvVars []string
+	serviceName        string
+	serviceEnvVars     []string
+	jsonSocket         string
+	jsonSocketDisabled bool
 )
 
 type program struct {
 	ctx              context.Context
 	cancel           context.CancelFunc
 	serv             *grpc.Server
+	jsonServ         *http.Server
+	jsonServMu       sync.Mutex
 	serverInstance   *server.Server
 	serverInstanceMu sync.Mutex
 }
@@ -46,6 +53,8 @@ func init() {
 	serviceCmd.PersistentFlags().BoolVar(&updateSettingsDisabled, "disable-update-settings", false, "Disables update settings feature. If enabled, the client will not be able to change or edit any settings. To persist this setting, use: netbird service install --disable-update-settings")
 	serviceCmd.PersistentFlags().BoolVar(&captureEnabled, "enable-capture", false, "Enables packet capture via 'netbird debug capture'. To persist, use: netbird service install --enable-capture")
 	serviceCmd.PersistentFlags().BoolVar(&networksDisabled, "disable-networks", false, "Disables network selection. If enabled, the client will not allow listing, selecting, or deselecting networks. To persist, use: netbird service install --disable-networks")
+	serviceCmd.PersistentFlags().StringVar(&jsonSocket, "json-socket", defaultJSONSocket, "HTTP/JSON API socket address served by grpc-gateway [unix|tcp]://[path|host:port]. To persist, use: netbird service install --json-socket")
+	serviceCmd.PersistentFlags().BoolVar(&jsonSocketDisabled, "disable-json-socket", false, "Disables the HTTP/JSON API socket. To persist, use: netbird service install --disable-json-socket")
 
 	rootCmd.PersistentFlags().StringVarP(&serviceName, "service", "s", defaultServiceName, "Netbird system service name")
 	serviceEnvDesc := `Sets extra environment variables for the service. ` +

client/cmd/service_controller.go

@@ -5,9 +5,6 @@ package cmd
 import (
 	"context"
 	"fmt"
-	"net"
-	"os"
-	"strings"
 	"time"
 
 	"github.com/kardianos/service"
@@ -32,31 +29,35 @@ func (p *program) Start(svc service.Service) error {
 	// in any case, even if configuration does not exists we run daemon to serve CLI gRPC API.
 	p.serv = grpc.NewServer()
 
-	split := strings.Split(daemonAddr, "://")
-	switch split[0] {
-	case "unix":
-		// cleanup failed close
-		stat, err := os.Stat(split[1])
-		if err == nil && !stat.IsDir() {
-			if err := os.Remove(split[1]); err != nil {
-				log.Debugf("remove socket file: %v", err)
-			}
-		}
-	case "tcp":
-	default:
-		return fmt.Errorf("unsupported daemon address protocol: %v", split[0])
-	}
-
-	listen, err := net.Listen(split[0], split[1])
+	daemonListener, err := listenOnAddress(daemonAddr)
 	if err != nil {
 		return fmt.Errorf("listen daemon interface: %w", err)
 	}
-	go func() {
-		defer listen.Close()
 
-		if split[0] == "unix" {
-			if err := os.Chmod(split[1], 0666); err != nil {
-				log.Errorf("failed setting daemon permissions: %v", split[1])
+	var jsonListener *socketListener
+	if !jsonSocketDisabled {
+		jsonListener, err = listenOnAddress(jsonSocket)
+		if err != nil {
+			_ = daemonListener.Close()
+			return fmt.Errorf("listen daemon JSON interface: %w", err)
+		}
+	} else {
+		removeStaleUnixSocketForAddress(jsonSocket)
+	}
+
+	go func() {
+		defer daemonListener.Close()
+		if jsonListener != nil {
+			defer jsonListener.Close()
+		}
+
+		if err := daemonListener.chmodUnixSocket("daemon"); err != nil {
+			log.Error(err)
+			return
+		}
+		if jsonListener != nil {
+			if err := jsonListener.chmodUnixSocket("daemon JSON"); err != nil {
+				log.Error(err)
 				return
 			}
 		}
@@ -71,8 +72,16 @@ func (p *program) Start(svc service.Service) error {
 		p.serverInstance = serverInstance
 		p.serverInstanceMu.Unlock()
 
-		log.Printf("started daemon server: %v", split[1])
-		if err := p.serv.Serve(listen); err != nil {
+		if jsonListener != nil {
+			if err := p.startJSONGateway(jsonListener, daemonAddr); err != nil {
+				log.Fatalf("failed to start daemon JSON server: %v", err)
+			}
+		} else {
+			log.Debug("daemon JSON socket disabled")
+		}
+
+		log.Printf("started daemon server: %v", daemonListener.address)
+		if err := p.serv.Serve(daemonListener.Listener); err != nil {
 			log.Errorf("failed to serve daemon requests: %v", err)
 		}
 	}()
@@ -92,6 +101,20 @@ func (p *program) Stop(srv service.Service) error {
 
 	p.cancel()
 
+	p.jsonServMu.Lock()
+	jsonServ := p.jsonServ
+	p.jsonServMu.Unlock()
+	if jsonServ != nil {
+		shutdownCtx, shutdownCancel := context.WithTimeout(context.Background(), 2*time.Second)
+		if err := jsonServ.Shutdown(shutdownCtx); err != nil {
+			log.Errorf("failed to stop daemon JSON server gracefully: %v", err)
+			if err := jsonServ.Close(); err != nil {
+				log.Errorf("failed to close daemon JSON server: %v", err)
+			}
+		}
+		shutdownCancel()
+	}
+
 	if p.serv != nil {
 		p.serv.Stop()
 	}

client/cmd/service_installer.go

@@ -67,6 +67,11 @@ func buildServiceArguments() []string {
 		args = append(args, "--disable-networks")
 	}
 
+	args = append(args, "--json-socket", jsonSocket)
+	if jsonSocketDisabled {
+		args = append(args, "--disable-json-socket")
+	}
+
 	return args
 }
 

client/cmd/service_json_gateway.go

@@ -0,0 +1,52 @@
+//go:build !ios && !android
+
+package cmd
+
+import (
+	"context"
+	"errors"
+	"net"
+	"net/http"
+	"strings"
+	"time"
+
+	"github.com/grpc-ecosystem/grpc-gateway/v2/runtime"
+	log "github.com/sirupsen/logrus"
+	"google.golang.org/grpc"
+	"google.golang.org/grpc/credentials/insecure"
+
+	"github.com/netbirdio/netbird/client/proto"
+)
+
+func grpcGatewayEndpoint(addr string) string {
+	return strings.TrimPrefix(addr, "tcp://")
+}
+
+func (p *program) startJSONGateway(jsonListener *socketListener, daemonEndpoint string) error {
+	mux := runtime.NewServeMux()
+	opts := []grpc.DialOption{grpc.WithTransportCredentials(insecure.NewCredentials())}
+	if err := proto.RegisterDaemonServiceHandlerFromEndpoint(p.ctx, mux, grpcGatewayEndpoint(daemonEndpoint), opts); err != nil {
+		return err
+	}
+
+	jsonServer := &http.Server{
+		Handler:           mux,
+		ReadHeaderTimeout: 5 * time.Second,
+		BaseContext: func(net.Listener) context.Context {
+			return p.ctx
+		},
+	}
+
+	p.jsonServMu.Lock()
+	p.jsonServ = jsonServer
+	p.jsonServMu.Unlock()
+
+	go func() {
+		log.Printf("started daemon JSON server: %v", jsonListener.address)
+		if err := jsonServer.Serve(jsonListener.Listener); err != nil && !errors.Is(err, http.ErrServerClosed) {
+			log.Errorf("failed to serve daemon JSON requests: %v", err)
+		}
+	}()
+
+	return nil
+}

client/cmd/service_params.go

@@ -23,6 +23,7 @@ const serviceParamsFile = "service.json"
 type serviceParams struct {
 	LogLevel              string            `json:"log_level"`
 	DaemonAddr            string            `json:"daemon_addr"`
+	JSONSocket            string            `json:"json_socket"`
 	ManagementURL         string            `json:"management_url,omitempty"`
 	ConfigPath            string            `json:"config_path,omitempty"`
 	LogFiles              []string          `json:"log_files,omitempty"`
@@ -30,6 +31,7 @@ type serviceParams struct {
 	DisableUpdateSettings bool              `json:"disable_update_settings,omitempty"`
 	EnableCapture         bool              `json:"enable_capture,omitempty"`
 	DisableNetworks       bool              `json:"disable_networks,omitempty"`
+	DisableJSONSocket     bool              `json:"disable_json_socket,omitempty"`
 	ServiceEnvVars        map[string]string `json:"service_env_vars,omitempty"`
 }
 
@@ -75,6 +77,7 @@ func currentServiceParams() *serviceParams {
 	params := &serviceParams{
 		LogLevel:              logLevel,
 		DaemonAddr:            daemonAddr,
+		JSONSocket:            jsonSocket,
 		ManagementURL:         managementURL,
 		ConfigPath:            configPath,
 		LogFiles:              logFiles,
@@ -82,6 +85,7 @@ func currentServiceParams() *serviceParams {
 		DisableUpdateSettings: updateSettingsDisabled,
 		EnableCapture:         captureEnabled,
 		DisableNetworks:       networksDisabled,
+		DisableJSONSocket:     jsonSocketDisabled,
 	}
 
 	if len(serviceEnvVars) > 0 {
@@ -113,9 +117,8 @@ func applyServiceParams(cmd *cobra.Command, params *serviceParams) {
 		return
 	}
 
-	// For fields with non-empty defaults (log-level, daemon-addr), keep the
-	// != "" guard so that an older service.json missing the field doesn't
-	// clobber the default with an empty string.
+	// For fields with non-empty defaults, keep the != "" guard so that an older
+	// service.json missing the field doesn't clobber the default with an empty string.
 	if !rootCmd.PersistentFlags().Changed("log-level") && params.LogLevel != "" {
 		logLevel = params.LogLevel
 	}
@@ -124,6 +127,20 @@ func applyServiceParams(cmd *cobra.Command, params *serviceParams) {
 		daemonAddr = params.DaemonAddr
 	}
 
+	jsonSocketChanged := serviceCmd.PersistentFlags().Changed("json-socket")
+	if !jsonSocketChanged && params.JSONSocket != "" {
+		jsonSocket = params.JSONSocket
+	}
+
+	if !serviceCmd.PersistentFlags().Changed("disable-json-socket") {
+		jsonSocketDisabled = params.DisableJSONSocket
+		// Passing --json-socket should re-enable the JSON gateway unless
+		// --disable-json-socket was explicitly provided too.
+		if jsonSocketChanged {
+			jsonSocketDisabled = false
+		}
+	}
+
 	// For optional fields where empty means "use default", always apply so
 	// that an explicit clear (--management-url "") persists across reinstalls.
 	if !rootCmd.PersistentFlags().Changed("management-url") {

client/cmd/service_params_test.go

@@ -530,6 +530,7 @@ func fieldToGlobalVar(field string) string {
 	m := map[string]string{
 		"LogLevel":              "logLevel",
 		"DaemonAddr":            "daemonAddr",
+		"JSONSocket":            "jsonSocket",
 		"ManagementURL":         "managementURL",
 		"ConfigPath":            "configPath",
 		"LogFiles":              "logFiles",
@@ -537,6 +538,7 @@ func fieldToGlobalVar(field string) string {
 		"DisableUpdateSettings": "updateSettingsDisabled",
 		"EnableCapture":         "captureEnabled",
 		"DisableNetworks":       "networksDisabled",
+		"DisableJSONSocket":     "jsonSocketDisabled",
 		"ServiceEnvVars":        "serviceEnvVars",
 	}
 	if v, ok := m[field]; ok {

client/cmd/service_socket.go

@@ -0,0 +1,83 @@
+//go:build !ios && !android
+
+package cmd
+
+import (
+	"fmt"
+	"net"
+	"os"
+	"strings"
+
+	log "github.com/sirupsen/logrus"
+)
+
+type socketListener struct {
+	net.Listener
+	network string
+	address string
+}
+
+func listenOnAddress(addr string) (*socketListener, error) {
+	network, address, err := parseListenAddress(addr)
+	if err != nil {
+		return nil, err
+	}
+
+	if network == "unix" {
+		removeStaleUnixSocket(address)
+	}
+
+	listener, err := net.Listen(network, address)
+	if err != nil {
+		return nil, err
+	}
+
+	return &socketListener{Listener: listener, network: network, address: address}, nil
+}
+
+func parseListenAddress(addr string) (string, string, error) {
+	network, address, ok := strings.Cut(addr, "://")
+	if !ok || network == "" || address == "" {
+		return "", "", fmt.Errorf("address must be in [unix|tcp]://[path|host:port] format: %q", addr)
+	}
+
+	switch network {
+	case "unix", "tcp":
+		return network, address, nil
+	default:
+		return "", "", fmt.Errorf("unsupported daemon address protocol: %v", network)
+	}
+}
+
+func removeStaleUnixSocket(path string) {
+	stat, err := os.Stat(path)
+	if err == nil && !stat.IsDir() {
+		if err := os.Remove(path); err != nil {
+			log.Debugf("remove socket file: %v", err)
+		}
+		return
+	}
+
+	if err != nil && !os.IsNotExist(err) {
+		log.Debugf("stat socket file: %v", err)
+	}
+}
+
+func removeStaleUnixSocketForAddress(addr string) {
+	network, address, err := parseListenAddress(addr)
+	if err != nil || network != "unix" {
+		return
+	}
+	removeStaleUnixSocket(address)
+}
+
+func (l *socketListener) chmodUnixSocket(description string) error {
+	if l == nil || l.network != "unix" {
+		return nil
+	}
+
+	if err := os.Chmod(l.address, 0666); err != nil {
+		return fmt.Errorf("failed setting %s permissions for %s: %w", description, l.address, err)
+	}
+	return nil
+}

client/internal/dns/mgmt/mgmt.go

@@ -51,13 +51,20 @@ type cachedRecord struct {
 }
 
 // Resolver caches critical NetBird infrastructure domains.
-// records, refreshing, mgmtDomain and serverDomains are all guarded by mutex.
+// records, refreshing, failedResolves, mgmtDomain and serverDomains are all
+// guarded by mutex.
 type Resolver struct {
 	records       map[dns.Question]*cachedRecord
 	mgmtDomain    *domain.Domain
 	serverDomains *dnsconfig.ServerDomains
 	mutex         sync.RWMutex
 
+	// failedResolves records the last failed initial resolve per domain so a
+	// domain that never resolves isn't retried on every server-domains update
+	// until refreshBackoff elapses. Entries are cleared on success and pruned
+	// to the current server-domains set.
+	failedResolves map[domain.Domain]time.Time
+
 	chain            ChainResolver
 	chainMaxPriority int
 	refreshGroup     singleflight.Group
@@ -76,9 +83,10 @@ type Resolver struct {
 // NewResolver creates a new management domains cache resolver.
 func NewResolver() *Resolver {
 	return &Resolver{
-		records:    make(map[dns.Question]*cachedRecord),
-		refreshing: make(map[dns.Question]*atomic.Bool),
-		cacheTTL:   resolveCacheTTL(),
+		records:        make(map[dns.Question]*cachedRecord),
+		refreshing:     make(map[dns.Question]*atomic.Bool),
+		failedResolves: make(map[domain.Domain]time.Time),
+		cacheTTL:       resolveCacheTTL(),
 	}
 }
 
@@ -173,7 +181,9 @@ func (m *Resolver) continueToNext(w dns.ResponseWriter, r *dns.Msg) {
 
 // AddDomain resolves a domain and stores its A/AAAA records in the cache.
 // A family that resolves NODATA (nil err, zero records) evicts any stale
-// entry for that qtype.
+// entry for that qtype. When one family hard-errors while the other succeeds,
+// the resolved family is still cached but AddDomain returns an error so the
+// caller retries the incomplete resolve rather than treating it as complete.
 func (m *Resolver) AddDomain(ctx context.Context, d domain.Domain) error {
 	dnsName := strings.ToLower(dns.Fqdn(d.PunycodeString()))
 
@@ -203,6 +213,10 @@ func (m *Resolver) AddDomain(ctx context.Context, d domain.Domain) error {
 	log.Debugf("added/updated domain=%s with %d A records and %d AAAA records",
 		d.SafeString(), len(aRecords), len(aaaaRecords))
 
+	if errA != nil || errAAAA != nil {
+		return fmt.Errorf("resolve %s: incomplete, a family failed: %w", d.SafeString(), errors.Join(errA, errAAAA))
+	}
+
 	return nil
 }
 
@@ -462,6 +476,7 @@ func (m *Resolver) RemoveDomain(d domain.Domain) error {
 	delete(m.records, qAAAA)
 	delete(m.refreshing, qA)
 	delete(m.refreshing, qAAAA)
+	delete(m.failedResolves, d)
 
 	log.Debugf("removed domain=%s from cache", d.SafeString())
 	return nil
@@ -505,6 +520,7 @@ func (m *Resolver) UpdateFromServerDomains(ctx context.Context, serverDomains dn
 		allDomains := m.extractDomainsFromServerDomains(updatedServerDomains)
 		currentDomains := m.GetCachedDomains()
 		removedDomains = m.removeStaleDomains(currentDomains, allDomains)
+		m.pruneFailedResolves(allDomains)
 	}
 
 	m.addNewDomains(ctx, newDomains)
@@ -577,13 +593,85 @@ func (m *Resolver) isManagementDomain(domain domain.Domain) bool {
 	return m.mgmtDomain != nil && domain == *m.mgmtDomain
 }
 
-// addNewDomains resolves and caches all domains from the update
+// addNewDomains resolves and caches domains that are not yet in the cache,
+// running the lookups concurrently. Domains already cached are skipped and left
+// to the stale-while-revalidate refresh path, so a sync never re-resolves them
+// synchronously: once NetBird owns the OS resolver the resolve runs through the
+// handler chain and would otherwise dial the managed upstreams under the engine
+// sync lock on every update.
 func (m *Resolver) addNewDomains(ctx context.Context, newDomains domain.List) {
+	var wg sync.WaitGroup
+	seen := make(map[domain.Domain]struct{}, len(newDomains))
 	for _, newDomain := range newDomains {
-		if err := m.AddDomain(ctx, newDomain); err != nil {
-			log.Warnf("failed to add/update domain=%s: %v", newDomain.SafeString(), err)
-		} else {
-			log.Debugf("added/updated management cache domain=%s", newDomain.SafeString())
+		if _, dup := seen[newDomain]; dup {
+			continue
+		}
+		seen[newDomain] = struct{}{}
+
+		if !m.needsResolve(newDomain) {
+			continue
+		}
+
+		wg.Add(1)
+		go func(d domain.Domain) {
+			defer wg.Done()
+			if err := m.AddDomain(ctx, d); err != nil {
+				m.markResolveFailed(d)
+				log.Warnf("failed to add/update domain=%s: %v", d.SafeString(), err)
+				return
+			}
+			m.clearResolveFailed(d)
+			log.Debugf("added/updated management cache domain=%s", d.SafeString())
+		}(newDomain)
+	}
+	wg.Wait()
+}
+
+// needsResolve reports whether d should be resolved now. A recent failed or
+// incomplete resolve gates retries on the backoff even when one family is
+// already cached, so a transiently-failed family is retried instead of being
+// treated as fully resolved. Otherwise a domain with any cached record is left
+// to the stale-while-revalidate refresh path.
+func (m *Resolver) needsResolve(d domain.Domain) bool {
+	dnsName := strings.ToLower(dns.Fqdn(d.PunycodeString()))
+
+	m.mutex.RLock()
+	defer m.mutex.RUnlock()
+
+	if failedAt, ok := m.failedResolves[d]; ok {
+		return time.Since(failedAt) >= refreshBackoff
+	}
+
+	for _, qtype := range []uint16{dns.TypeA, dns.TypeAAAA} {
+		q := dns.Question{Name: dnsName, Qtype: qtype, Qclass: dns.ClassINET}
+		if _, ok := m.records[q]; ok {
+			return false
+		}
+	}
+	return true
+}
+
+func (m *Resolver) markResolveFailed(d domain.Domain) {
+	m.mutex.Lock()
+	m.failedResolves[d] = time.Now()
+	m.mutex.Unlock()
+}
+
+func (m *Resolver) clearResolveFailed(d domain.Domain) {
+	m.mutex.Lock()
+	delete(m.failedResolves, d)
+	m.mutex.Unlock()
+}
+
+// pruneFailedResolves drops failure markers for domains no longer present in
+// the server-domains set, keeping the map bounded to the current set (a
+// failed-only domain has no cached record, so RemoveDomain never sees it).
+func (m *Resolver) pruneFailedResolves(domains domain.List) {
+	m.mutex.Lock()
+	defer m.mutex.Unlock()
+	for d := range m.failedResolves {
+		if !slices.Contains(domains, d) {
+			delete(m.failedResolves, d)
 		}
 	}
 }

client/internal/dns/mgmt/mgmt_refresh_test.go

@@ -21,6 +21,7 @@ type fakeChain struct {
 	mu       sync.Mutex
 	calls    map[string]int
 	answers  map[string][]dns.RR
+	qErr     map[string]error
 	err      error
 	hasRoot  bool
 	onLookup func()
@@ -30,6 +31,7 @@ func newFakeChain() *fakeChain {
 	return &fakeChain{
 		calls:   map[string]int{},
 		answers: map[string][]dns.RR{},
+		qErr:    map[string]error{},
 		hasRoot: true,
 	}
 }
@@ -47,6 +49,9 @@ func (f *fakeChain) ResolveInternal(ctx context.Context, msg *dns.Msg, maxPriori
 	f.calls[key]++
 	answers := f.answers[key]
 	err := f.err
+	if err == nil {
+		err = f.qErr[key]
+	}
 	onLookup := f.onLookup
 	f.mu.Unlock()
 
@@ -75,6 +80,12 @@ func (f *fakeChain) setAnswer(name string, qtype uint16, ip string) {
 	}
 }
 
+func (f *fakeChain) setErr(name string, qtype uint16, err error) {
+	f.mu.Lock()
+	defer f.mu.Unlock()
+	f.qErr[name+"|"+dns.TypeToString[qtype]] = err
+}
+
 func (f *fakeChain) callCount(name string, qtype uint16) int {
 	f.mu.Lock()
 	defer f.mu.Unlock()

client/internal/dns/mgmt/mgmt_resolve_test.go

@@ -0,0 +1,183 @@
+package mgmt
+
+import (
+	"context"
+	"errors"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"github.com/miekg/dns"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	dnsconfig "github.com/netbirdio/netbird/client/internal/dns/config"
+	"github.com/netbirdio/netbird/shared/management/domain"
+)
+
+// A domain already in the cache must not be re-resolved on a subsequent server
+// domains update; it is left to the stale-while-revalidate refresh path.
+func TestResolver_UpdateFromServerDomains_SkipsCached(t *testing.T) {
+	r := NewResolver()
+	chain := newFakeChain()
+	chain.setAnswer("signal.example.com.", dns.TypeA, "10.0.0.2")
+	r.SetChainResolver(chain, 50)
+
+	sd := dnsconfig.ServerDomains{Signal: domain.Domain("signal.example.com")}
+
+	_, err := r.UpdateFromServerDomains(context.Background(), sd)
+	require.NoError(t, err)
+	require.Equal(t, 1, chain.callCount("signal.example.com.", dns.TypeA),
+		"first update must resolve the domain")
+
+	_, err = r.UpdateFromServerDomains(context.Background(), sd)
+	require.NoError(t, err)
+	assert.Equal(t, 1, chain.callCount("signal.example.com.", dns.TypeA),
+		"cached domain must not be re-resolved on a subsequent update")
+}
+
+// New domains in a single update must resolve concurrently rather than serially.
+func TestResolver_AddNewDomains_ResolvesConcurrently(t *testing.T) {
+	r := NewResolver()
+	chain := newFakeChain()
+
+	var inflight, maxInflight atomic.Int32
+	chain.onLookup = func() {
+		n := inflight.Add(1)
+		for {
+			old := maxInflight.Load()
+			if n <= old || maxInflight.CompareAndSwap(old, n) {
+				break
+			}
+		}
+		time.Sleep(50 * time.Millisecond)
+		inflight.Add(-1)
+	}
+
+	relays := []domain.Domain{"a.example.com", "b.example.com", "c.example.com", "d.example.com"}
+	for _, d := range relays {
+		chain.setAnswer(dns.Fqdn(string(d)), dns.TypeA, "10.0.0.2")
+	}
+	r.SetChainResolver(chain, 50)
+
+	start := time.Now()
+	_, err := r.UpdateFromServerDomains(context.Background(), dnsconfig.ServerDomains{Relay: relays})
+	require.NoError(t, err)
+	elapsed := time.Since(start)
+
+	assert.GreaterOrEqual(t, int(maxInflight.Load()), 2, "domains must resolve concurrently")
+	// Serial resolution of 4 domains would take at least 4*50ms; concurrent is far less.
+	assert.Less(t, elapsed, 300*time.Millisecond, "resolution should not be serial")
+}
+
+// A domain that fails to resolve must not be retried on every update; the
+// failure backoff suppresses re-resolution until it expires.
+func TestResolver_UpdateFromServerDomains_BacksOffFailures(t *testing.T) {
+	r := NewResolver()
+	chain := newFakeChain()
+	chain.err = errors.New("resolve boom")
+	r.SetChainResolver(chain, 50)
+
+	sd := dnsconfig.ServerDomains{Signal: domain.Domain("signal.example.com")}
+
+	_, err := r.UpdateFromServerDomains(context.Background(), sd)
+	require.NoError(t, err)
+	require.Equal(t, 1, chain.callCount("signal.example.com.", dns.TypeA),
+		"first update must attempt the resolve")
+
+	_, err = r.UpdateFromServerDomains(context.Background(), sd)
+	require.NoError(t, err)
+	assert.Equal(t, 1, chain.callCount("signal.example.com.", dns.TypeA),
+		"failed resolve must back off and not retry on the next update")
+}
+
+// A domain listed under more than one server-domain type (e.g. STUN and TURN on
+// the same host) must be resolved once per update, not once per occurrence.
+func TestResolver_AddNewDomains_DedupesDuplicateDomains(t *testing.T) {
+	r := NewResolver()
+	chain := newFakeChain()
+	chain.setAnswer("dup.example.com.", dns.TypeA, "10.0.0.9")
+	r.SetChainResolver(chain, 50)
+
+	sd := dnsconfig.ServerDomains{
+		Stuns: []domain.Domain{"dup.example.com"},
+		Turns: []domain.Domain{"dup.example.com"},
+	}
+
+	_, err := r.UpdateFromServerDomains(context.Background(), sd)
+	require.NoError(t, err)
+	assert.Equal(t, 1, chain.callCount("dup.example.com.", dns.TypeA),
+		"a domain appearing under multiple server-domain types must resolve once")
+}
+
+// A failure marker must be dropped once its domain leaves the server-domains set
+// so the map stays bounded to the current set.
+func TestResolver_UpdateFromServerDomains_PrunesFailedResolves(t *testing.T) {
+	r := NewResolver()
+	chain := newFakeChain()
+	chain.err = errors.New("resolve boom")
+	r.SetChainResolver(chain, 50)
+
+	_, err := r.UpdateFromServerDomains(context.Background(), dnsconfig.ServerDomains{Signal: domain.Domain("gone.example.com")})
+	require.NoError(t, err)
+	r.mutex.RLock()
+	_, marked := r.failedResolves[domain.Domain("gone.example.com")]
+	r.mutex.RUnlock()
+	require.True(t, marked, "failed resolve must be recorded")
+
+	_, err = r.UpdateFromServerDomains(context.Background(), dnsconfig.ServerDomains{Signal: domain.Domain("other.example.com")})
+	require.NoError(t, err)
+	r.mutex.RLock()
+	_, stillMarked := r.failedResolves[domain.Domain("gone.example.com")]
+	r.mutex.RUnlock()
+	assert.False(t, stillMarked, "failure marker for a domain no longer in the set must be pruned")
+}
+
+// When one family hard-errors while the other resolves, the domain is cached
+// for the working family but recorded as incomplete so the failed family is
+// retried under backoff instead of being treated as fully resolved forever.
+func TestResolver_AddNewDomains_RetriesPartialFamilyFailure(t *testing.T) {
+	d := domain.Domain("relay.example.com")
+	r := NewResolver()
+	chain := newFakeChain()
+	chain.setAnswer("relay.example.com.", dns.TypeA, "10.0.0.2")
+	chain.setErr("relay.example.com.", dns.TypeAAAA, errors.New("servfail"))
+	r.SetChainResolver(chain, 50)
+
+	_, err := r.UpdateFromServerDomains(context.Background(), dnsconfig.ServerDomains{Relay: []domain.Domain{d}})
+	require.NoError(t, err)
+
+	r.mutex.RLock()
+	_, aCached := r.records[dns.Question{Name: "relay.example.com.", Qtype: dns.TypeA, Qclass: dns.ClassINET}]
+	_, marked := r.failedResolves[d]
+	r.mutex.RUnlock()
+	require.True(t, aCached, "the working family must still be cached")
+	require.True(t, marked, "a partial failure must be recorded so the failed family is retried")
+
+	assert.False(t, r.needsResolve(d), "within the backoff window the domain is not retried")
+
+	r.mutex.Lock()
+	r.failedResolves[d] = time.Now().Add(-2 * refreshBackoff)
+	r.mutex.Unlock()
+	assert.True(t, r.needsResolve(d), "after the backoff elapses the domain is retried to pick up the missing family")
+}
+
+// A family that returns NODATA (legitimately absent, e.g. an IPv4-only host) is
+// not a failure: the domain must not be marked for retry, otherwise it would be
+// re-resolved on every sync.
+func TestResolver_AddNewDomains_NodataIsNotFailure(t *testing.T) {
+	d := domain.Domain("v4only.example.com")
+	r := NewResolver()
+	chain := newFakeChain()
+	chain.setAnswer("v4only.example.com.", dns.TypeA, "10.0.0.2")
+	r.SetChainResolver(chain, 50)
+
+	_, err := r.UpdateFromServerDomains(context.Background(), dnsconfig.ServerDomains{Relay: []domain.Domain{d}})
+	require.NoError(t, err)
+
+	r.mutex.RLock()
+	_, marked := r.failedResolves[d]
+	r.mutex.RUnlock()
+	assert.False(t, marked, "a NODATA family must not be recorded as a failure")
+	assert.False(t, r.needsResolve(d), "an IPv4-only host must not be re-resolved on later syncs")
+}

client/proto/generate.sh

@@ -12,5 +12,11 @@ script_path=$(dirname "$(realpath "$0")")
 cd "$script_path"
 go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.36.6
 go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@v1.6.1
-protoc -I ./ ./daemon.proto --go_out=../ --go-grpc_out=../ --experimental_allow_proto3_optional
+go install github.com/grpc-ecosystem/grpc-gateway/v2/protoc-gen-grpc-gateway@v2.26.3
+protoc -I ./ ./daemon.proto \
+  --go_out=../ \
+  --go-grpc_out=../ \
+  --grpc-gateway_out=../ \
+  --grpc-gateway_opt=generate_unbound_methods=true \
+  --experimental_allow_proto3_optional
 cd "$old_pwd"

go.mod

@@ -66,6 +66,7 @@ require (
 	github.com/google/nftables v0.3.0
 	github.com/gopacket/gopacket v1.4.0
 	github.com/grpc-ecosystem/go-grpc-middleware/v2 v2.0.2-0.20240212192251-757544f21357
+	github.com/grpc-ecosystem/grpc-gateway/v2 v2.26.3
 	github.com/hashicorp/go-multierror v1.1.1
 	github.com/hashicorp/go-secure-stdlib/base62 v0.1.2
 	github.com/hashicorp/go-version v1.7.0
@@ -330,6 +331,7 @@ require (
 	golang.org/x/text v0.36.0 // indirect
 	golang.org/x/tools v0.43.0 // indirect
 	golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2 // indirect
+	google.golang.org/genproto/googleapis/api v0.0.0-20260319201613-d00831a3d3e7 // indirect
 	google.golang.org/genproto/googleapis/rpc v0.0.0-20260401024825-9d38bb4040a9 // indirect
 	gopkg.in/square/go-jose.v2 v2.6.0 // indirect
 	gopkg.in/tomb.v1 v1.0.0-20141024135613-dd632973f1e7 // indirect

management/server/account_test.go

@@ -1916,6 +1916,117 @@ func TestDefaultAccountManager_MarkPeerConnected_PeerLoginExpiration(t *testing.
 	}
 }
 
+func TestDefaultAccountManager_MarkPeerDisconnected_SchedulesInactivityExpiration(t *testing.T) {
+	manager, _, err := createManager(t)
+	require.NoError(t, err, "unable to create account manager")
+
+	accountID, err := manager.GetAccountIDByUserID(context.Background(), auth.UserAuth{UserId: userID})
+	require.NoError(t, err, "unable to create an account")
+
+	key, err := wgtypes.GenerateKey()
+	require.NoError(t, err, "unable to generate WireGuard key")
+	peerPubKey := key.PublicKey().String()
+
+	_, _, _, _, err = manager.AddPeer(context.Background(), "", "", userID, &nbpeer.Peer{
+		Key:                         peerPubKey,
+		Meta:                        nbpeer.PeerSystemMeta{Hostname: "test-peer"},
+		InactivityExpirationEnabled: true,
+	}, false)
+	require.NoError(t, err, "unable to add peer")
+
+	_, err = manager.UpdateAccountSettings(context.Background(), accountID, userID, &types.Settings{
+		PeerLoginExpiration:             time.Hour,
+		PeerLoginExpirationEnabled:      true,
+		PeerInactivityExpiration:        time.Hour,
+		PeerInactivityExpirationEnabled: true,
+		Extra:                           &types.ExtraSettings{},
+	})
+	require.NoError(t, err, "expecting to update account settings successfully but got error")
+
+	// Establish a session so the matching-token disconnect is actually applied.
+	streamStartTime := time.Now().UTC()
+	err = manager.MarkPeerConnected(context.Background(), peerPubKey, accountID, streamStartTime.UnixNano(), nil)
+	require.NoError(t, err, "unable to mark peer connected")
+
+	// Install the mock only now, so the assertion observes the disconnect, not
+	// the earlier connect.
+	scheduled := make(chan struct{}, 1)
+	manager.peerInactivityExpiry = &MockScheduler{
+		CancelFunc: func(ctx context.Context, IDs []string) {},
+		ScheduleFunc: func(ctx context.Context, in time.Duration, ID string, job func() (nextRunIn time.Duration, reschedule bool)) {
+			select {
+			case scheduled <- struct{}{}:
+			default:
+			}
+		},
+	}
+
+	err = manager.MarkPeerDisconnected(context.Background(), peerPubKey, accountID, streamStartTime.UnixNano())
+	require.NoError(t, err, "unable to mark peer disconnected")
+
+	select {
+	case <-scheduled:
+		// expected: disconnect re-armed the inactivity expiry timer
+	case <-time.After(time.Second):
+		t.Fatal("expected inactivity expiration to be rescheduled when an eligible peer disconnects")
+	}
+}
+
+func TestDefaultAccountManager_MarkPeerDisconnected_SkipsInactivityExpirationWhenDisabled(t *testing.T) {
+	manager, _, err := createManager(t)
+	require.NoError(t, err, "unable to create account manager")
+
+	accountID, err := manager.GetAccountIDByUserID(context.Background(), auth.UserAuth{UserId: userID})
+	require.NoError(t, err, "unable to create an account")
+
+	key, err := wgtypes.GenerateKey()
+	require.NoError(t, err, "unable to generate WireGuard key")
+	peerPubKey := key.PublicKey().String()
+
+	_, _, _, _, err = manager.AddPeer(context.Background(), "", "", userID, &nbpeer.Peer{
+		Key:                         peerPubKey,
+		Meta:                        nbpeer.PeerSystemMeta{Hostname: "test-peer"},
+		InactivityExpirationEnabled: true,
+	}, false)
+	require.NoError(t, err, "unable to add peer")
+
+	// Peer is eligible (SSO + inactivity enabled) but the account-level setting
+	// stays disabled, so disconnect must not schedule anything.
+	_, err = manager.UpdateAccountSettings(context.Background(), accountID, userID, &types.Settings{
+		PeerLoginExpiration:             time.Hour,
+		PeerLoginExpirationEnabled:      true,
+		PeerInactivityExpiration:        time.Hour,
+		PeerInactivityExpirationEnabled: false,
+		Extra:                           &types.ExtraSettings{},
+	})
+	require.NoError(t, err, "expecting to update account settings successfully but got error")
+
+	streamStartTime := time.Now().UTC()
+	err = manager.MarkPeerConnected(context.Background(), peerPubKey, accountID, streamStartTime.UnixNano(), nil)
+	require.NoError(t, err, "unable to mark peer connected")
+
+	scheduled := make(chan struct{}, 1)
+	manager.peerInactivityExpiry = &MockScheduler{
+		CancelFunc: func(ctx context.Context, IDs []string) {},
+		ScheduleFunc: func(ctx context.Context, in time.Duration, ID string, job func() (nextRunIn time.Duration, reschedule bool)) {
+			select {
+			case scheduled <- struct{}{}:
+			default:
+			}
+		},
+	}
+
+	err = manager.MarkPeerDisconnected(context.Background(), peerPubKey, accountID, streamStartTime.UnixNano())
+	require.NoError(t, err, "unable to mark peer disconnected")
+
+	select {
+	case <-scheduled:
+		t.Fatal("inactivity expiration must not be scheduled while the account-level setting is disabled")
+	case <-time.After(200 * time.Millisecond):
+		// expected: nothing scheduled
+	}
+}
+
 func TestDefaultAccountManager_OnPeerDisconnected_LastSeenCheck(t *testing.T) {
 	manager, _, err := createManager(t)
 	require.NoError(t, err, "unable to create account manager")

management/server/peer.go

@@ -188,6 +188,15 @@ func (am *DefaultAccountManager) MarkPeerDisconnected(ctx context.Context, peerP
 		}
 	}
 
+	if peer.AddedWithSSOLogin() && peer.InactivityExpirationEnabled {
+		settings, err := am.Store.GetAccountSettings(ctx, store.LockingStrengthNone, accountID)
+		if err != nil {
+			log.WithContext(ctx).Warnf("failed getting account settings to schedule inactivity expiration for peer %s: %v", peer.ID, err)
+		} else if settings.PeerInactivityExpirationEnabled {
+			am.checkAndSchedulePeerInactivityExpiration(ctx, accountID)
+		}
+	}
+
 	return nil
 }