From cd777395f2cbb1d06161b37941182bde93bdacfa Mon Sep 17 00:00:00 2001 From: Bethuel Mmbaga Date: Mon, 15 Jun 2026 12:01:54 +0300 Subject: [PATCH 1/5] [management] Skip JWT group evaluation for embedded-IdP local users (#6422) When JWT group sync is enabled with a restrictive JWTAllowGroups list, the local owner of an embedded-IdP (Dex) deployment can get locked out. The allow-groups check runs account-wide but local password users do not receive external IdP group claims, so they can't satisfy the allowed list. This skips JWT group evaluation for local Dex users so the restriction and JWT group sync continue to apply to external-IdP users as intended. --- idp/dex/provider.go | 11 +++++++- idp/dex/provider_test.go | 20 ++++++++++++++ management/server/account.go | 6 +++- management/server/account_test.go | 23 ++++++++++++++++ management/server/auth/manager.go | 6 +++- management/server/auth/manager_test.go | 38 ++++++++++++++++++++++++++ 6 files changed, 101 insertions(+), 3 deletions(-) diff --git a/idp/dex/provider.go b/idp/dex/provider.go index 526d6a17a..67aeb995f 100644 --- a/idp/dex/provider.go +++ b/idp/dex/provider.go @@ -41,6 +41,8 @@ type Config struct { GRPCAddr string } +const localConnectorID = "local" + // Provider wraps a Dex server type Provider struct { config *Config @@ -544,7 +546,7 @@ func (p *Provider) CreateUser(ctx context.Context, email, username, password str // Encode the user ID in Dex's format: base64(protobuf{user_id, connector_id}) // This matches the format Dex uses in JWT tokens - encodedID := EncodeDexUserID(userID, "local") + encodedID := EncodeDexUserID(userID, localConnectorID) return encodedID, nil } @@ -619,6 +621,13 @@ func DecodeDexUserID(encodedID string) (userID, connectorID string, err error) { return userID, connectorID, nil } +// IsLocalUserID reports whether encodedID is a Dex subject for the built-in +// local password connector. +func IsLocalUserID(encodedID string) bool { + _, connectorID, err := DecodeDexUserID(encodedID) + return err == nil && connectorID == localConnectorID +} + // GetUser returns a user by email func (p *Provider) GetUser(ctx context.Context, email string) (storage.Password, error) { return p.storage.GetPassword(ctx, email) diff --git a/idp/dex/provider_test.go b/idp/dex/provider_test.go index 88828fbbb..3eb29db97 100644 --- a/idp/dex/provider_test.go +++ b/idp/dex/provider_test.go @@ -115,6 +115,26 @@ func TestDecodeDexUserID(t *testing.T) { } } +func TestIsLocalUserID(t *testing.T) { + tests := []struct { + name string + encodedID string + want bool + }{ + {name: "local connector", encodedID: EncodeDexUserID("7aad8c05-3287-473f-b42a-365504bf25e7", "local"), want: true}, + {name: "federated connector", encodedID: EncodeDexUserID("entra-user", "entra"), want: false}, + {name: "non-dex external IdP id", encodedID: "google-oauth2|1234567890", want: false}, + {name: "invalid base64", encodedID: "not-valid-base64!!!", want: false}, + {name: "empty", encodedID: "", want: false}, + } + + for _, tt := range tests { + t.Run(tt.name, func(t *testing.T) { + assert.Equal(t, tt.want, IsLocalUserID(tt.encodedID)) + }) + } +} + func TestEncodeDexUserID(t *testing.T) { userID := "7aad8c05-3287-473f-b42a-365504bf25e7" connectorID := "local" diff --git a/management/server/account.go b/management/server/account.go index f16717857..e7fcad9d1 100644 --- a/management/server/account.go +++ b/management/server/account.go @@ -28,6 +28,7 @@ import ( nbdns "github.com/netbirdio/netbird/dns" "github.com/netbirdio/netbird/formatter/hook" + "github.com/netbirdio/netbird/idp/dex" "github.com/netbirdio/netbird/management/internals/controllers/network_map" nbconfig "github.com/netbirdio/netbird/management/internals/server/config" "github.com/netbirdio/netbird/management/server/account" @@ -1588,7 +1589,10 @@ func (am *DefaultAccountManager) updateUserAuthWithSingleMode(ctx context.Contex // and propagates changes to peers if group propagation is enabled. // requires userAuth to have been ValidateAndParseToken and EnsureUserAccessByJWTGroups by the AuthManager func (am *DefaultAccountManager) SyncUserJWTGroups(ctx context.Context, userAuth auth.UserAuth) error { - if userAuth.IsChild || userAuth.IsPAT { + // Child accounts and PAT-authenticated requests do not sync JWT groups. + // Embedded-Dex local users also skip sync because local password authentication + // does not provide external IdP group claims. + if userAuth.IsChild || userAuth.IsPAT || dex.IsLocalUserID(userAuth.UserId) { return nil } diff --git a/management/server/account_test.go b/management/server/account_test.go index ba621030c..bb4779d85 100644 --- a/management/server/account_test.go +++ b/management/server/account_test.go @@ -26,6 +26,7 @@ import ( "github.com/netbirdio/netbird/shared/management/status" nbdns "github.com/netbirdio/netbird/dns" + "github.com/netbirdio/netbird/idp/dex" "github.com/netbirdio/netbird/management/internals/controllers/network_map" "github.com/netbirdio/netbird/management/internals/controllers/network_map/controller" "github.com/netbirdio/netbird/management/internals/controllers/network_map/update_channel" @@ -723,6 +724,28 @@ func TestDefaultAccountManager_SyncUserJWTGroups(t *testing.T) { require.Equal(t, g2.Name, "group2", "group2 name should match") require.Equal(t, g2.Issued, types.GroupIssuedJWT, "group2 issued should match") }) + t.Run("local embedded-Dex user is skipped", func(t *testing.T) { + initAccount.Settings.JWTGroupsEnabled = true + initAccount.Settings.JWTGroupsClaimName = "idp-groups" + err := manager.Store.SaveAccount(context.Background(), initAccount) + require.NoError(t, err, "save account failed") + + localClaims := auth.UserAuth{ + AccountId: accountID, + Domain: domain, + UserId: dex.EncodeDexUserID("local-owner", "local"), + Groups: []string{"group3", "group4"}, + } + err = manager.SyncUserJWTGroups(context.Background(), localClaims) + require.NoError(t, err, "sync should be a no-op for local users") + + account, err := manager.Store.GetAccount(context.Background(), accountID) + require.NoError(t, err, "get account failed") + for _, g := range account.Groups { + require.NotEqual(t, "group3", g.Name, "local user JWT groups must not be synced") + require.NotEqual(t, "group4", g.Name, "local user JWT groups must not be synced") + } + }) } func TestAccountManager_PrivateAccount(t *testing.T) { diff --git a/management/server/auth/manager.go b/management/server/auth/manager.go index 27346a604..9498789f2 100644 --- a/management/server/auth/manager.go +++ b/management/server/auth/manager.go @@ -12,6 +12,7 @@ import ( "github.com/netbirdio/netbird/shared/auth" "github.com/netbirdio/netbird/base62" + "github.com/netbirdio/netbird/idp/dex" "github.com/netbirdio/netbird/management/server/store" "github.com/netbirdio/netbird/management/server/types" nbjwt "github.com/netbirdio/netbird/shared/auth/jwt" @@ -74,7 +75,10 @@ func (m *manager) ValidateAndParseToken(ctx context.Context, value string) (auth } func (m *manager) EnsureUserAccessByJWTGroups(ctx context.Context, userAuth auth.UserAuth, token *jwt.Token) (auth.UserAuth, error) { - if userAuth.IsChild || userAuth.IsPAT { + // Child accounts and PAT-authenticated requests do not use JWT group access checks. + // Embedded-Dex local users also skip them because local password authentication + // does not provide external IdP group claims. + if userAuth.IsChild || userAuth.IsPAT || dex.IsLocalUserID(userAuth.UserId) { return userAuth, nil } diff --git a/management/server/auth/manager_test.go b/management/server/auth/manager_test.go index 469737f47..af8a30ef1 100644 --- a/management/server/auth/manager_test.go +++ b/management/server/auth/manager_test.go @@ -16,6 +16,7 @@ import ( "github.com/stretchr/testify/assert" "github.com/stretchr/testify/require" + "github.com/netbirdio/netbird/idp/dex" "github.com/netbirdio/netbird/management/server/auth" "github.com/netbirdio/netbird/management/server/store" "github.com/netbirdio/netbird/management/server/types" @@ -206,6 +207,43 @@ func TestAuthManager_EnsureUserAccessByJWTGroups(t *testing.T) { _, err = manager.EnsureUserAccessByJWTGroups(context.Background(), userAuth, token) require.Error(t, err, "ensure user access is not in allowed groups") }) + + t.Run("Local embedded-Dex user is exempt from JWT allow-groups", func(t *testing.T) { + account.Settings.JWTGroupsEnabled = true + account.Settings.JWTGroupsClaimName = "idp-groups" + account.Settings.JWTAllowGroups = []string{"not-a-group"} + err := store.SaveAccount(context.Background(), account) + require.NoError(t, err, "save account failed") + + // Local Dex users have a "local" connector encoded in their user ID. + localUserAuth := nbauth.UserAuth{ + AccountId: account.Id, + Domain: domain, + UserId: dex.EncodeDexUserID("local-owner", "local"), + } + + localUserAuth, err = manager.EnsureUserAccessByJWTGroups(context.Background(), localUserAuth, token) + require.NoError(t, err, "local user must not be locked out by JWT allow-groups (issue #5337)") + require.Len(t, localUserAuth.Groups, 0, "JWT groups must not be evaluated for local users") + }) + + t.Run("Federated embedded-Dex user is still subject to JWT allow-groups", func(t *testing.T) { + account.Settings.JWTGroupsEnabled = true + account.Settings.JWTGroupsClaimName = "idp-groups" + account.Settings.JWTAllowGroups = []string{"not-a-group"} + err := store.SaveAccount(context.Background(), account) + require.NoError(t, err, "save account failed") + + // A federated user (non-"local" connector) must remain restricted. + fedUserAuth := nbauth.UserAuth{ + AccountId: account.Id, + Domain: domain, + UserId: dex.EncodeDexUserID("entra-user", "entra"), + } + + _, err = manager.EnsureUserAccessByJWTGroups(context.Background(), fedUserAuth, token) + require.Error(t, err, "federated user must still be restricted by JWT allow-groups") + }) } func TestAuthManager_ValidateAndParseToken(t *testing.T) { From 60067619a1827d6aa574cf6d5f3e5055c452d4c5 Mon Sep 17 00:00:00 2001 From: Lee Sang Hoon Date: Mon, 15 Jun 2026 19:21:24 +0900 Subject: [PATCH 2/5] [proxy] Keep custom TCP listeners alive after mapping batches (#6415) --- proxy/server.go | 27 +++++++- proxy/server_test.go | 120 ++++++++++++++++++++++++++++++++++++ proxy/sync_mappings_test.go | 89 ++++++++++++++++++++++++++ 3 files changed, 234 insertions(+), 2 deletions(-) diff --git a/proxy/server.go b/proxy/server.go index cd90682b0..2d4767106 100644 --- a/proxy/server.go +++ b/proxy/server.go @@ -1105,7 +1105,7 @@ func (s *Server) getOrCreatePortRouter(ctx context.Context, port uint16) (*nbtcp router := nbtcp.NewPortRouter(s.Logger, s.resolveDialFunc) router.SetObserver(s.meter) router.SetAccessLogger(s.accessLog) - portCtx, cancel := context.WithCancel(ctx) + portCtx, cancel := context.WithCancel(s.portRouterContext(ctx)) s.portRouters[port] = &portRouter{ router: router, @@ -1121,10 +1121,26 @@ func (s *Server) getOrCreatePortRouter(ctx context.Context, port uint16) (*nbtcp } }() - s.Logger.Debugf("started per-port router on %s", listenAddr) + s.Logger.WithFields(log.Fields{ + "port": port, + "listen_addr": listenAddr, + "bound_addr": ln.Addr().String(), + "proxy_protocol": s.ProxyProtocol, + }).Info("custom TCP listener started") return router, nil } +// portRouterContext returns the server-lifetime context for custom TCP +// listeners. Mapping-batch contexts are cancelled after a batch is applied; a +// per-port listener must outlive that batch and only stop on service removal or +// server shutdown. +func (s *Server) portRouterContext(ctx context.Context) context.Context { + if s.ctx != nil { + return s.ctx + } + return ctx +} + // cleanupPortIfEmpty tears down a per-port router if it has no remaining // routes or fallback. The main port is never cleaned up. Active relay // connections are drained before the listener is closed. @@ -1718,6 +1734,13 @@ func (s *Server) setupTCPMapping(ctx context.Context, mapping *proto.ProxyMappin s.meter.L4ServiceAdded(types.ServiceModeTCP) s.sendStatusUpdate(ctx, accountID, svcID, proto.ProxyStatus_PROXY_STATUS_ACTIVE, nil) + + s.Logger.WithFields(log.Fields{ + "domain": mapping.GetDomain(), + "target": targetAddr, + "port": port, + "service": svcID, + }).Info("TCP mapping added") return nil } diff --git a/proxy/server_test.go b/proxy/server_test.go index aa4892201..f0c4765db 100644 --- a/proxy/server_test.go +++ b/proxy/server_test.go @@ -3,14 +3,20 @@ package proxy import ( "context" "errors" + "fmt" "io" + "net" "testing" "time" log "github.com/sirupsen/logrus" "github.com/stretchr/testify/assert" "github.com/stretchr/testify/require" + "go.opentelemetry.io/otel/metric/noop" + "google.golang.org/grpc" + proxymetrics "github.com/netbirdio/netbird/proxy/internal/metrics" + "github.com/netbirdio/netbird/proxy/internal/types" "github.com/netbirdio/netbird/shared/management/proto" ) @@ -202,3 +208,117 @@ func TestRedactMappingForLog_HandlesEmptyOrNilFields(t *testing.T) { assert.Nil(t, redacted.Auth, "nil Auth must remain nil") assert.Empty(t, redacted.Path, "empty Path must remain empty") } + +type statusUpdateOnlyClient struct { + proto.ProxyServiceClient +} + +func (statusUpdateOnlyClient) SendStatusUpdate(context.Context, *proto.SendStatusUpdateRequest, ...grpc.CallOption) (*proto.SendStatusUpdateResponse, error) { + return &proto.SendStatusUpdateResponse{}, nil +} + +func TestSetupTCPMappingBindsCustomListenPort(t *testing.T) { + ln, err := net.Listen("tcp", "127.0.0.1:0") + require.NoError(t, err) + port := uint16(ln.Addr().(*net.TCPAddr).Port) //nolint:gosec // test port allocated by the OS + require.NoError(t, ln.Close()) + + meter, err := proxymetrics.New(context.Background(), noop.Meter{}) + require.NoError(t, err) + + srv := &Server{ + Logger: quietLifecycleLogger(), + mgmtClient: statusUpdateOnlyClient{}, + meter: meter, + mainPort: 8443, + portRouters: make(map[uint16]*portRouter), + svcPorts: make(map[types.ServiceID][]uint16), + } + t.Cleanup(func() { + srv.portMu.Lock() + for p, pr := range srv.portRouters { + pr.cancel() + require.NoError(t, pr.listener.Close()) + delete(srv.portRouters, p) + } + srv.portMu.Unlock() + srv.portRouterWg.Wait() + }) + + mapping := &proto.ProxyMapping{ + Type: proto.ProxyMappingUpdateType_UPDATE_TYPE_CREATED, + Id: "svc-tcp", + AccountId: "acct-1", + Domain: "ssh.example.com", + Mode: "tcp", + ListenPort: int32(port), + Path: []*proto.PathMapping{ + {Target: "10.0.0.5:22"}, + }, + } + + require.NoError(t, srv.setupTCPMapping(context.Background(), mapping)) + + srv.portMu.RLock() + pr := srv.portRouters[port] + ports := append([]uint16(nil), srv.svcPorts[types.ServiceID("svc-tcp")]...) + srv.portMu.RUnlock() + + require.NotNil(t, pr, "custom TCP mapping must create a per-port router") + assert.Equal(t, []uint16{port}, ports, "service must track the custom listen port for cleanup") + + second, err := net.Listen("tcp", fmt.Sprintf(":%d", port)) + if err == nil { + _ = second.Close() + } + require.Error(t, err, "custom TCP listen port must be bound after setup") +} + +func TestCustomTCPPortRouterOutlivesMappingBatchContext(t *testing.T) { + ln, err := net.Listen("tcp", "127.0.0.1:0") + require.NoError(t, err) + port := uint16(ln.Addr().(*net.TCPAddr).Port) //nolint:gosec // test port allocated by the OS + require.NoError(t, ln.Close()) + + meter, err := proxymetrics.New(context.Background(), noop.Meter{}) + require.NoError(t, err) + + srvCtx, srvCancel := context.WithCancel(context.Background()) + t.Cleanup(srvCancel) + + srv := &Server{ + ctx: srvCtx, + Logger: quietLifecycleLogger(), + meter: meter, + mainPort: 8443, + portRouters: make(map[uint16]*portRouter), + svcPorts: make(map[types.ServiceID][]uint16), + } + t.Cleanup(func() { + srv.portMu.Lock() + for p, pr := range srv.portRouters { + pr.cancel() + if err := pr.listener.Close(); err != nil && !errors.Is(err, net.ErrClosed) { + require.NoError(t, err) + } + delete(srv.portRouters, p) + } + srv.portMu.Unlock() + srv.portRouterWg.Wait() + }) + + batchCtx, cancelBatch := context.WithCancel(context.Background()) + _, err = srv.getOrCreatePortRouter(batchCtx, port) + require.NoError(t, err) + + cancelBatch() + + assert.Never(t, func() bool { + second, err := net.Listen("tcp", fmt.Sprintf(":%d", port)) + if err == nil { + _ = second.Close() + return true + } + return false + }, 200*time.Millisecond, 10*time.Millisecond, "custom TCP listener must outlive mapping-batch context cancellation") +} diff --git a/proxy/sync_mappings_test.go b/proxy/sync_mappings_test.go index 801587e4c..c9c0dad03 100644 --- a/proxy/sync_mappings_test.go +++ b/proxy/sync_mappings_test.go @@ -81,6 +81,95 @@ func TestIntegration_SyncMappings_HappyPath(t *testing.T) { assert.Equal(t, "app2.test.proxy.io", rp2.GetDomain()) } +func TestIntegration_SyncMappings_CustomTCPMappingDeliveredWithCapabilities(t *testing.T) { + setup := setupIntegrationTest(t) + defer setup.cleanup() + + ctx := context.Background() + tcpSvc := &service.Service{ + ID: "tcp-custom", + AccountID: "test-account-1", + Name: "Custom TCP", + Domain: "ssh.test.proxy.io", + ProxyCluster: "test.proxy.io", + Mode: "tcp", + ListenPort: 10001, + Enabled: true, + Targets: []*service.Target{{ + Host: "10.0.0.5", + Port: 22, + Protocol: "tcp", + TargetId: "peer-ssh", + TargetType: "peer", + Enabled: true, + }}, + } + require.NoError(t, setup.store.CreateService(ctx, tcpSvc)) + + conn, err := grpc.NewClient(setup.grpcAddr, grpc.WithTransportCredentials(insecure.NewCredentials())) + require.NoError(t, err) + defer conn.Close() + + client := proto.NewProxyServiceClient(conn) + receiveSnapshot := func(proxyID string, caps *proto.ProxyCapabilities) map[string]*proto.ProxyMapping { + t.Helper() + + streamCtx, cancel := context.WithTimeout(context.Background(), 5*time.Second) + defer cancel() + + stream, err := client.SyncMappings(streamCtx) + require.NoError(t, err) + + err = stream.Send(&proto.SyncMappingsRequest{ + Msg: &proto.SyncMappingsRequest_Init{ + Init: &proto.SyncMappingsInit{ + ProxyId: proxyID, + Version: "test-v1", + Address: "test.proxy.io", + Capabilities: caps, + }, + }, + }) + require.NoError(t, err) + + mappingsByID := make(map[string]*proto.ProxyMapping) + for { + msg, err := stream.Recv() + require.NoError(t, err) + for _, m := range msg.GetMapping() { + mappingsByID[m.GetId()] = m + } + + err = stream.Send(&proto.SyncMappingsRequest{ + Msg: &proto.SyncMappingsRequest_Ack{Ack: &proto.SyncMappingsAck{}}, + }) + require.NoError(t, err) + + if msg.GetInitialSyncComplete() { + break + } + } + return mappingsByID + } + + legacyMappings := receiveSnapshot("sync-proxy-no-capabilities", nil) + assert.NotContains(t, legacyMappings, "tcp-custom", + "legacy proxies that do not report capabilities must not receive TCP custom-port mappings") + + supportsCustomPorts := true + modernMappings := receiveSnapshot("sync-proxy-custom-ports", &proto.ProxyCapabilities{ + SupportsCustomPorts: &supportsCustomPorts, + }) + + tcpMapping := modernMappings["tcp-custom"] + require.NotNil(t, tcpMapping, "capability-aware proxy must receive TCP custom-port mapping") + assert.Equal(t, "tcp", tcpMapping.GetMode()) + assert.Equal(t, int32(10001), tcpMapping.GetListenPort()) + require.Len(t, tcpMapping.GetPath(), 1) + assert.Equal(t, "10.0.0.5:22", tcpMapping.GetPath()[0].GetTarget()) + assert.NotEmpty(t, tcpMapping.GetAuthToken(), "snapshot mapping must include per-proxy auth token") +} + func TestIntegration_SyncMappings_BackPressure(t *testing.T) { setup := setupIntegrationTest(t) defer setup.cleanup() From f893abc41d32e02545089ebdb443f48037e62808 Mon Sep 17 00:00:00 2001 From: Viktor Liu <17948409+lixmal@users.noreply.github.com> Date: Mon, 15 Jun 2026 23:36:00 +0900 Subject: [PATCH 3/5] [client] Recover from tun device read/write panics and restart the client (#6419) --- client/iface/device/device_filter.go | 70 ++++++++++++++++++++--- client/iface/device/device_filter_test.go | 57 ++++++++++++++++++ client/internal/engine.go | 6 +- 3 files changed, 123 insertions(+), 10 deletions(-) diff --git a/client/iface/device/device_filter.go b/client/iface/device/device_filter.go index fc1c65efa..7d7493835 100644 --- a/client/iface/device/device_filter.go +++ b/client/iface/device/device_filter.go @@ -1,10 +1,13 @@ package device import ( + "fmt" "net/netip" + "runtime/debug" "sync" "sync/atomic" + log "github.com/sirupsen/logrus" "golang.zx2c4.com/wireguard/tun" ) @@ -41,10 +44,13 @@ type PacketCapture interface { type FilteredDevice struct { tun.Device - filter PacketFilter - capture atomic.Pointer[PacketCapture] - mutex sync.RWMutex - closeOnce sync.Once + filter PacketFilter + capture atomic.Pointer[PacketCapture] + // panicHandler is invoked after a panic in the underlying device is + // recovered in Read or Write. + panicHandler atomic.Pointer[func()] + mutex sync.RWMutex + closeOnce sync.Once } // newDeviceFilter constructor function @@ -70,7 +76,7 @@ func (d *FilteredDevice) Close() error { // Read wraps read method with filtering feature func (d *FilteredDevice) Read(bufs [][]byte, sizes []int, offset int) (n int, err error) { - if n, err = d.Device.Read(bufs, sizes, offset); err != nil { + if n, err = d.deviceRead(bufs, sizes, offset); err != nil { return 0, err } @@ -112,7 +118,7 @@ func (d *FilteredDevice) Write(bufs [][]byte, offset int) (int, error) { d.mutex.RUnlock() if filter == nil { - return d.Device.Write(bufs, offset) + return d.deviceWrite(bufs, offset) } filteredBufs := make([][]byte, 0, len(bufs)) @@ -125,9 +131,44 @@ func (d *FilteredDevice) Write(bufs [][]byte, offset int) (int, error) { } } - n, err := d.Device.Write(filteredBufs, offset) - n += dropped - return n, err + n, err := d.deviceWrite(filteredBufs, offset) + if err != nil { + return n, err + } + return n + dropped, nil +} + +// deviceRead calls the underlying device Read, recovering from panics in the +// wintun read path and converting them into errors. +func (d *FilteredDevice) deviceRead(bufs [][]byte, sizes []int, offset int) (n int, err error) { + defer d.recoverFromPanic("read", &n, &err) + return d.Device.Read(bufs, sizes, offset) +} + +// deviceWrite calls the underlying device Write, recovering from panics in the +// wintun write path and converting them into errors. +func (d *FilteredDevice) deviceWrite(bufs [][]byte, offset int) (n int, err error) { + defer d.recoverFromPanic("write", &n, &err) + return d.Device.Write(bufs, offset) +} + +// recoverFromPanic converts a panic in the underlying device into a regular +// error and invokes the registered panic handler. The wintun read path is +// known to panic on zero-length packets that third-party filter drivers can +// place in the ring. +func (d *FilteredDevice) recoverFromPanic(op string, n *int, err *error) { + r := recover() + if r == nil { + return + } + + log.Errorf("recovered panic in tun device %s: %v\n%s", op, r, debug.Stack()) + *n = 0 + *err = fmt.Errorf("tun device %s panic: %v", op, r) + + if handler := d.panicHandler.Load(); handler != nil { + (*handler)() + } } // SetFilter sets packet filter to device @@ -137,6 +178,17 @@ func (d *FilteredDevice) SetFilter(filter PacketFilter) { d.mutex.Unlock() } +// SetPanicHandler registers a handler invoked after a recovered panic in Read +// or Write. The device is unusable after such a panic; the handler should +// trigger recreation of the interface. Pass nil to remove. +func (d *FilteredDevice) SetPanicHandler(handler func()) { + if handler == nil { + d.panicHandler.Store(nil) + return + } + d.panicHandler.Store(&handler) +} + // SetCapture sets or clears the packet capture sink. Pass nil to disable. // Uses atomic store so the hot path (Read/Write) is a single pointer load // with no locking overhead when capture is off. diff --git a/client/iface/device/device_filter_test.go b/client/iface/device/device_filter_test.go index 8fb16ca8d..0d86c9323 100644 --- a/client/iface/device/device_filter_test.go +++ b/client/iface/device/device_filter_test.go @@ -221,3 +221,60 @@ func TestDeviceWrapperRead(t *testing.T) { } }) } + +func TestDeviceWrapperReadPanic(t *testing.T) { + ctrl := gomock.NewController(t) + defer ctrl.Finish() + + tun := mocks.NewMockDevice(ctrl) + tun.EXPECT().Read(gomock.Any(), gomock.Any(), gomock.Any()). + DoAndReturn(func(bufs [][]byte, sizes []int, offset int) (int, error) { + // Reproduce the wintun zero-length packet panic (index out of range). + packet := make([]byte, 0) + return int(packet[0]), nil + }) + + wrapped := newDeviceFilter(tun) + + handlerCalled := false + wrapped.SetPanicHandler(func() { handlerCalled = true }) + + n, err := wrapped.Read([][]byte{{}}, []int{0}, 0) + if err == nil { + t.Errorf("expected error from recovered panic, got nil") + } + if n != 0 { + t.Errorf("expected n=0, got %d", n) + } + if !handlerCalled { + t.Errorf("expected panic handler to be called") + } +} + +func TestDeviceWrapperWritePanic(t *testing.T) { + ctrl := gomock.NewController(t) + defer ctrl.Finish() + + tun := mocks.NewMockDevice(ctrl) + tun.EXPECT().Write(gomock.Any(), gomock.Any()). + DoAndReturn(func(bufs [][]byte, offset int) (int, error) { + packet := make([]byte, 0) + return int(packet[0]), nil + }) + + wrapped := newDeviceFilter(tun) + + handlerCalled := false + wrapped.SetPanicHandler(func() { handlerCalled = true }) + + n, err := wrapped.Write([][]byte{{0x45, 0x00}}, 0) + if err == nil { + t.Errorf("expected error from recovered panic, got nil") + } + if n != 0 { + t.Errorf("expected n=0, got %d", n) + } + if !handlerCalled { + t.Errorf("expected panic handler to be called") + } +} diff --git a/client/internal/engine.go b/client/internal/engine.go index 2b41d2015..09651196a 100644 --- a/client/internal/engine.go +++ b/client/internal/engine.go @@ -240,7 +240,7 @@ type Engine struct { syncStore syncstore.Store syncStoreDir string - flowManager nftypes.FlowManager + flowManager nftypes.FlowManager // auto-update updateManager *updater.Manager @@ -531,6 +531,10 @@ func (e *Engine) Start(netbirdConfig *mgmProto.NetbirdConfig, mgmtURL *url.URL) return fmt.Errorf("create wg interface: %w", err) } + if filteredDevice := e.wgInterface.GetDevice(); filteredDevice != nil { + filteredDevice.SetPanicHandler(e.triggerClientRestart) + } + if err := e.createFirewall(); err != nil { e.close() return err From b57f7143507b810910fa14705decac5b2f08a3c4 Mon Sep 17 00:00:00 2001 From: Viktor Liu <17948409+lixmal@users.noreply.github.com> Date: Mon, 15 Jun 2026 23:37:03 +0900 Subject: [PATCH 4/5] [client] Drop signaling-side ICE candidate filter, drop overlay STUN at mux read-side instead (#6142) --- client/iface/bind/ice_bind.go | 5 +- client/iface/bind/ice_bind_test.go | 2 +- client/iface/device/device_kernel_unix.go | 3 - client/iface/iface.go | 1 - client/iface/iface_new.go | 2 +- client/iface/iface_new_android.go | 2 +- client/iface/iface_new_ios.go | 2 +- client/iface/iface_new_linux.go | 4 +- client/iface/udpmux/universal.go | 87 ++++--------------- client/iface/wgproxy/proxy_linux_test.go | 2 +- client/iface/wgproxy/proxy_seed_test.go | 2 +- client/internal/engine.go | 17 ---- client/internal/peer/worker_ice.go | 33 ------- .../systemops/systemops_generic.go | 9 +- 14 files changed, 32 insertions(+), 139 deletions(-) diff --git a/client/iface/bind/ice_bind.go b/client/iface/bind/ice_bind.go index bf79ecd79..156450c61 100644 --- a/client/iface/bind/ice_bind.go +++ b/client/iface/bind/ice_bind.go @@ -41,7 +41,6 @@ type ICEBind struct { *wgConn.StdNetBind transportNet transport.Net - filterFn udpmux.FilterFn address wgaddr.Address mtu uint16 @@ -61,12 +60,11 @@ type ICEBind struct { ipv6Conn *net.UDPConn } -func NewICEBind(transportNet transport.Net, filterFn udpmux.FilterFn, address wgaddr.Address, mtu uint16) *ICEBind { +func NewICEBind(transportNet transport.Net, address wgaddr.Address, mtu uint16) *ICEBind { b, _ := wgConn.NewStdNetBind().(*wgConn.StdNetBind) ib := &ICEBind{ StdNetBind: b, transportNet: transportNet, - filterFn: filterFn, address: address, mtu: mtu, endpoints: make(map[netip.Addr]net.Conn), @@ -265,7 +263,6 @@ func (s *ICEBind) createOrUpdateMux() { udpmux.UniversalUDPMuxParams{ UDPConn: muxConn, Net: s.transportNet, - FilterFn: s.filterFn, WGAddress: s.address, MTU: s.mtu, }, diff --git a/client/iface/bind/ice_bind_test.go b/client/iface/bind/ice_bind_test.go index f49e68508..0b8db7640 100644 --- a/client/iface/bind/ice_bind_test.go +++ b/client/iface/bind/ice_bind_test.go @@ -289,7 +289,7 @@ func setupICEBind(t *testing.T) *ICEBind { IP: netip.MustParseAddr("100.64.0.1"), Network: netip.MustParsePrefix("100.64.0.0/10"), } - return NewICEBind(transportNet, nil, address, 1280) + return NewICEBind(transportNet, address, 1280) } func createDualStackConns(t *testing.T) (*net.UDPConn, *net.UDPConn) { diff --git a/client/iface/device/device_kernel_unix.go b/client/iface/device/device_kernel_unix.go index 25c4148a6..3c429fb96 100644 --- a/client/iface/device/device_kernel_unix.go +++ b/client/iface/device/device_kernel_unix.go @@ -32,8 +32,6 @@ type TunKernelDevice struct { link *wgLink udpMuxConn net.PacketConn udpMux *udpmux.UniversalUDPMuxDefault - - filterFn udpmux.FilterFn } func NewKernelDevice(name string, address wgaddr.Address, wgPort int, key string, mtu uint16, transportNet transport.Net) *TunKernelDevice { @@ -104,7 +102,6 @@ func (t *TunKernelDevice) Up() (*udpmux.UniversalUDPMuxDefault, error) { bindParams := udpmux.UniversalUDPMuxParams{ UDPConn: nbnet.WrapPacketConn(rawSock), Net: t.transportNet, - FilterFn: t.filterFn, WGAddress: t.address, MTU: t.mtu, } diff --git a/client/iface/iface.go b/client/iface/iface.go index 78c5080e7..247f421a2 100644 --- a/client/iface/iface.go +++ b/client/iface/iface.go @@ -63,7 +63,6 @@ type WGIFaceOpts struct { MTU uint16 MobileArgs *device.MobileIFaceArguments TransportNet transport.Net - FilterFn udpmux.FilterFn DisableDNS bool } diff --git a/client/iface/iface_new.go b/client/iface/iface_new.go index 28f350e3f..96a0e670f 100644 --- a/client/iface/iface_new.go +++ b/client/iface/iface_new.go @@ -11,7 +11,7 @@ import ( // NewWGIFace Creates a new WireGuard interface instance func NewWGIFace(opts WGIFaceOpts) (*WGIface, error) { - iceBind := bind.NewICEBind(opts.TransportNet, opts.FilterFn, opts.Address, opts.MTU) + iceBind := bind.NewICEBind(opts.TransportNet, opts.Address, opts.MTU) var tun WGTunDevice if netstack.IsEnabled() { diff --git a/client/iface/iface_new_android.go b/client/iface/iface_new_android.go index e28dcc0de..ce8b4da23 100644 --- a/client/iface/iface_new_android.go +++ b/client/iface/iface_new_android.go @@ -9,7 +9,7 @@ import ( // NewWGIFace Creates a new WireGuard interface instance func NewWGIFace(opts WGIFaceOpts) (*WGIface, error) { - iceBind := bind.NewICEBind(opts.TransportNet, opts.FilterFn, opts.Address, opts.MTU) + iceBind := bind.NewICEBind(opts.TransportNet, opts.Address, opts.MTU) if netstack.IsEnabled() { wgIFace := &WGIface{ diff --git a/client/iface/iface_new_ios.go b/client/iface/iface_new_ios.go index 41e0022b2..cedd55ce2 100644 --- a/client/iface/iface_new_ios.go +++ b/client/iface/iface_new_ios.go @@ -10,7 +10,7 @@ import ( // NewWGIFace Creates a new WireGuard interface instance func NewWGIFace(opts WGIFaceOpts) (*WGIface, error) { - iceBind := bind.NewICEBind(opts.TransportNet, opts.FilterFn, opts.Address, opts.MTU) + iceBind := bind.NewICEBind(opts.TransportNet, opts.Address, opts.MTU) wgIFace := &WGIface{ tun: device.NewTunDevice(opts.IFaceName, opts.Address, opts.WGPort, opts.WGPrivKey, opts.MTU, iceBind, opts.MobileArgs.TunFd), diff --git a/client/iface/iface_new_linux.go b/client/iface/iface_new_linux.go index 65ce67e88..2465130e6 100644 --- a/client/iface/iface_new_linux.go +++ b/client/iface/iface_new_linux.go @@ -14,7 +14,7 @@ import ( // NewWGIFace Creates a new WireGuard interface instance func NewWGIFace(opts WGIFaceOpts) (*WGIface, error) { if netstack.IsEnabled() { - iceBind := bind.NewICEBind(opts.TransportNet, opts.FilterFn, opts.Address, opts.MTU) + iceBind := bind.NewICEBind(opts.TransportNet, opts.Address, opts.MTU) return &WGIface{ tun: device.NewNetstackDevice(opts.IFaceName, opts.Address, opts.WGPort, opts.WGPrivKey, opts.MTU, iceBind, netstack.ListenAddr()), userspaceBind: true, @@ -30,7 +30,7 @@ func NewWGIFace(opts WGIFaceOpts) (*WGIface, error) { } if device.ModuleTunIsLoaded() { - iceBind := bind.NewICEBind(opts.TransportNet, opts.FilterFn, opts.Address, opts.MTU) + iceBind := bind.NewICEBind(opts.TransportNet, opts.Address, opts.MTU) return &WGIface{ tun: device.NewTunDevice(opts.IFaceName, opts.Address, opts.WGPort, opts.WGPrivKey, opts.MTU, iceBind), userspaceBind: true, diff --git a/client/iface/udpmux/universal.go b/client/iface/udpmux/universal.go index 89a7eefb9..77e1b1b35 100644 --- a/client/iface/udpmux/universal.go +++ b/client/iface/udpmux/universal.go @@ -8,8 +8,6 @@ import ( "context" "fmt" "net" - "net/netip" - "sync" "time" log "github.com/sirupsen/logrus" @@ -22,10 +20,6 @@ import ( "github.com/netbirdio/netbird/client/iface/wgaddr" ) -// FilterFn is a function that filters out candidates based on the address. -// If it returns true, the address is to be filtered. It also returns the prefix of matching route. -type FilterFn func(address netip.Addr) (bool, netip.Prefix, error) - // UniversalUDPMuxDefault handles STUN and TURN servers packets by wrapping the original UDPConn // It then passes packets to the UDPMux that does the actual connection muxing. type UniversalUDPMuxDefault struct { @@ -43,7 +37,6 @@ type UniversalUDPMuxParams struct { UDPConn net.PacketConn XORMappedAddrCacheTTL time.Duration Net transport.Net - FilterFn FilterFn WGAddress wgaddr.Address MTU uint16 } @@ -68,7 +61,6 @@ func NewUniversalUDPMuxDefault(params UniversalUDPMuxParams) *UniversalUDPMuxDef PacketConn: params.UDPConn, mux: m, logger: params.Logger, - filterFn: params.FilterFn, address: params.WGAddress, } @@ -115,15 +107,12 @@ func (m *UniversalUDPMuxDefault) ReadFromConn(ctx context.Context) { } } -// UDPConn is a wrapper around UDPMux conn that overrides ReadFrom and handles STUN/TURN packets +// UDPConn is a wrapper around UDPMux conn that overrides WriteTo to drop packets destined for the overlay subnet. type UDPConn struct { net.PacketConn - mux *UniversalUDPMuxDefault - logger logging.LeveledLogger - filterFn FilterFn - // TODO: reset cache on route changes - addrCache sync.Map - address wgaddr.Address + mux *UniversalUDPMuxDefault + logger logging.LeveledLogger + address wgaddr.Address } // GetPacketConn returns the underlying PacketConn @@ -132,65 +121,16 @@ func (u *UDPConn) GetPacketConn() net.PacketConn { } func (u *UDPConn) WriteTo(b []byte, addr net.Addr) (int, error) { - if u.filterFn == nil { + udpAddr, ok := addr.(*net.UDPAddr) + if !ok { return u.PacketConn.WriteTo(b, addr) } - - if isRouted, found := u.addrCache.Load(addr.String()); found { - return u.handleCachedAddress(isRouted.(bool), b, addr) - } - - return u.handleUncachedAddress(b, addr) -} - -func (u *UDPConn) handleCachedAddress(isRouted bool, b []byte, addr net.Addr) (int, error) { - if isRouted { - return 0, fmt.Errorf("address %s is part of a routed network, refusing to write", addr) - } - return u.PacketConn.WriteTo(b, addr) -} - -func (u *UDPConn) handleUncachedAddress(b []byte, addr net.Addr) (int, error) { - if err := u.performFilterCheck(addr); err != nil { - return 0, err - } - return u.PacketConn.WriteTo(b, addr) -} - -func (u *UDPConn) performFilterCheck(addr net.Addr) error { - host, err := getHostFromAddr(addr) - if err != nil { - log.Errorf("Failed to get host from address %s: %v", addr, err) - return nil - } - - a, err := netip.ParseAddr(host) - if err != nil { - log.Errorf("Failed to parse address %s: %v", addr, err) - return nil - } - - if u.address.Network.Contains(a) { + dst := udpAddr.AddrPort().Addr().Unmap() + if (u.address.Network.IsValid() && u.address.Network.Contains(dst)) || (u.address.IPv6Net.IsValid() && u.address.IPv6Net.Contains(dst)) { log.Warnf("address %s is part of the NetBird network %s, refusing to write", addr, u.address) - return fmt.Errorf("address %s is part of the NetBird network %s, refusing to write", addr, u.address) + return 0, fmt.Errorf("address %s is part of the NetBird network %s, refusing to write", addr, u.address) } - - if isRouted, prefix, err := u.filterFn(a); err != nil { - log.Errorf("Failed to check if address %s is routed: %v", addr, err) - } else { - u.addrCache.Store(addr.String(), isRouted) - if isRouted { - // Extra log, as the error only shows up with ICE logging enabled - log.Infof("address %s is part of routed network %s, refusing to write", addr, prefix) - return fmt.Errorf("address %s is part of routed network %s, refusing to write", addr, prefix) - } - } - return nil -} - -func getHostFromAddr(addr net.Addr) (string, error) { - host, _, err := net.SplitHostPort(addr.String()) - return host, err + return u.PacketConn.WriteTo(b, addr) } // GetSharedConn returns the shared udp conn @@ -225,6 +165,13 @@ func (m *UniversalUDPMuxDefault) HandleSTUNMessage(msg *stun.Message, addr net.A return nil } + src := udpAddr.AddrPort().Addr().Unmap() + wg := m.params.WGAddress + if (wg.Network.IsValid() && wg.Network.Contains(src)) || (wg.IPv6Net.IsValid() && wg.IPv6Net.Contains(src)) { + log.Debugf("dropping STUN message from overlay source %s", udpAddr) + return nil + } + if m.isXORMappedResponse(msg, udpAddr.String()) { err := m.handleXORMappedResponse(udpAddr, msg) if err != nil { diff --git a/client/iface/wgproxy/proxy_linux_test.go b/client/iface/wgproxy/proxy_linux_test.go index dd24d1cdc..7f7abcb4a 100644 --- a/client/iface/wgproxy/proxy_linux_test.go +++ b/client/iface/wgproxy/proxy_linux_test.go @@ -66,7 +66,7 @@ func seedProxyForProxyCloseByRemoteConn() ([]proxyInstance, error) { if err != nil { return nil, err } - iceBind := bind.NewICEBind(nil, nil, wgAddress, 1280) + iceBind := bind.NewICEBind(nil, wgAddress, 1280) endpointAddress := &net.UDPAddr{ IP: net.IPv4(10, 0, 0, 1), Port: 1234, diff --git a/client/iface/wgproxy/proxy_seed_test.go b/client/iface/wgproxy/proxy_seed_test.go index ad375ccde..9278029a5 100644 --- a/client/iface/wgproxy/proxy_seed_test.go +++ b/client/iface/wgproxy/proxy_seed_test.go @@ -22,7 +22,7 @@ func seedProxyForProxyCloseByRemoteConn() ([]proxyInstance, error) { if err != nil { return nil, err } - iceBind := bind.NewICEBind(nil, nil, wgAddress, 1280) + iceBind := bind.NewICEBind(nil, wgAddress, 1280) endpointAddress := &net.UDPAddr{ IP: net.IPv4(10, 0, 0, 1), Port: 1234, diff --git a/client/internal/engine.go b/client/internal/engine.go index 09651196a..cf40d8983 100644 --- a/client/internal/engine.go +++ b/client/internal/engine.go @@ -53,7 +53,6 @@ import ( "github.com/netbirdio/netbird/client/internal/relay" "github.com/netbirdio/netbird/client/internal/rosenpass" "github.com/netbirdio/netbird/client/internal/routemanager" - "github.com/netbirdio/netbird/client/internal/routemanager/systemops" "github.com/netbirdio/netbird/client/internal/statemanager" "github.com/netbirdio/netbird/client/internal/syncstore" "github.com/netbirdio/netbird/client/internal/updater" @@ -1913,7 +1912,6 @@ func (e *Engine) newWgIface() (*iface.WGIface, error) { WGPrivKey: e.config.WgPrivateKey.String(), MTU: e.config.MTU, TransportNet: transportNet, - FilterFn: e.addrViaRoutes, DisableDNS: e.config.DisableDNS, } @@ -2161,21 +2159,6 @@ func (e *Engine) startNetworkMonitor() { }() } -func (e *Engine) addrViaRoutes(addr netip.Addr) (bool, netip.Prefix, error) { - var vpnRoutes []netip.Prefix - for _, routes := range e.routeManager.GetClientRoutes() { - if len(routes) > 0 && routes[0] != nil { - vpnRoutes = append(vpnRoutes, routes[0].Network) - } - } - - if isVpn, prefix := systemops.IsAddrRouted(addr, vpnRoutes); isVpn { - return true, prefix, nil - } - - return false, netip.Prefix{}, nil -} - func (e *Engine) stopDNSServer() { if e.dnsServer == nil { return diff --git a/client/internal/peer/worker_ice.go b/client/internal/peer/worker_ice.go index 29bf5aaaa..b1aa3e0f9 100644 --- a/client/internal/peer/worker_ice.go +++ b/client/internal/peer/worker_ice.go @@ -4,7 +4,6 @@ import ( "context" "fmt" "net" - "net/netip" "strconv" "sync" "time" @@ -165,10 +164,6 @@ func (w *WorkerICE) OnRemoteCandidate(candidate ice.Candidate, haRoutes route.HA return } - if candidateViaRoutes(candidate, haRoutes) { - return - } - if err := w.agent.AddRemoteCandidate(candidate); err != nil { w.log.Errorf("error while handling remote candidate") return @@ -589,34 +584,6 @@ func extraSrflxCandidate(candidate ice.Candidate) (*ice.CandidateServerReflexive return ec, nil } -func candidateViaRoutes(candidate ice.Candidate, clientRoutes route.HAMap) bool { - addr, err := netip.ParseAddr(candidate.Address()) - if err != nil { - log.Errorf("Failed to parse IP address %s: %v", candidate.Address(), err) - return false - } - - var routePrefixes []netip.Prefix - for _, routes := range clientRoutes { - if len(routes) > 0 && routes[0] != nil { - routePrefixes = append(routePrefixes, routes[0].Network) - } - } - - for _, prefix := range routePrefixes { - // default route is handled by route exclusion / ip rules - if prefix.Bits() == 0 { - continue - } - - if prefix.Contains(addr) { - log.Debugf("Ignoring candidate [%s], its address is part of routed network %s", candidate.String(), prefix) - return true - } - } - return false -} - func isRelayCandidate(candidate ice.Candidate) bool { return candidate.Type() == ice.CandidateTypeRelay } diff --git a/client/internal/routemanager/systemops/systemops_generic.go b/client/internal/routemanager/systemops/systemops_generic.go index 2b96c14dc..bb9ac494d 100644 --- a/client/internal/routemanager/systemops/systemops_generic.go +++ b/client/internal/routemanager/systemops/systemops_generic.go @@ -121,9 +121,12 @@ func (r *SysOps) addRouteToNonVPNIntf(prefix netip.Prefix, vpnIntf wgIface, init return Nexthop{}, vars.ErrRouteNotAllowed } - // Check if the prefix is part of any local subnets - if isLocal, subnet := r.isPrefixInLocalSubnets(prefix); isLocal { - return Nexthop{}, fmt.Errorf("prefix %s is part of local subnet %s: %w", prefix, subnet, vars.ErrRouteNotAllowed) + // BSDs blackhole a /32 added inside a directly-connected subnet; Linux/Windows need it to beat the wt0 route. + switch runtime.GOOS { + case "darwin", "freebsd", "netbsd", "openbsd", "dragonfly": + if isLocal, subnet := r.isPrefixInLocalSubnets(prefix); isLocal { + return Nexthop{}, fmt.Errorf("prefix %s is part of local subnet %s: %w", prefix, subnet, vars.ErrRouteNotAllowed) + } } // Determine the exit interface and next hop for the prefix, so we can add a specific route From a44198fd7728d54da5cc47893c973a2fe655ddaa Mon Sep 17 00:00:00 2001 From: Viktor Liu <17948409+lixmal@users.noreply.github.com> Date: Mon, 15 Jun 2026 23:43:24 +0900 Subject: [PATCH 5/5] [client] Add dialWebSocket method to WASM client (#5980) --- client/wasm/cmd/main.go | 71 +++++ client/wasm/internal/websocket/websocket.go | 304 ++++++++++++++++++++ go.mod | 3 + go.sum | 7 + 4 files changed, 385 insertions(+) create mode 100644 client/wasm/internal/websocket/websocket.go diff --git a/client/wasm/cmd/main.go b/client/wasm/cmd/main.go index 066fe043b..4683f4033 100644 --- a/client/wasm/cmd/main.go +++ b/client/wasm/cmd/main.go @@ -21,6 +21,7 @@ import ( "github.com/netbirdio/netbird/client/wasm/internal/http" "github.com/netbirdio/netbird/client/wasm/internal/rdp" "github.com/netbirdio/netbird/client/wasm/internal/ssh" + nbwebsocket "github.com/netbirdio/netbird/client/wasm/internal/websocket" "github.com/netbirdio/netbird/util" ) @@ -30,6 +31,7 @@ const ( pingTimeout = 10 * time.Second defaultLogLevel = "warn" defaultSSHDetectionTimeout = 20 * time.Second + dialWebSocketTimeout = 30 * time.Second icmpEchoRequest = 8 icmpCodeEcho = 0 @@ -677,6 +679,7 @@ func createClientObject(client *netbird.Client) js.Value { obj["createSSHConnection"] = createSSHMethod(client) obj["proxyRequest"] = createProxyRequestMethod(client) obj["createRDPProxy"] = createRDPProxyMethod(client) + obj["dialWebSocket"] = createDialWebSocketMethod(client) obj["status"] = createStatusMethod(client) obj["statusSummary"] = createStatusSummaryMethod(client) obj["statusDetail"] = createStatusDetailMethod(client) @@ -691,6 +694,74 @@ func createClientObject(client *netbird.Client) js.Value { return js.ValueOf(obj) } +func createDialWebSocketMethod(client *netbird.Client) js.Func { + return js.FuncOf(func(_ js.Value, args []js.Value) any { + url, protocols, timeout, errVal := parseDialWebSocketArgs(args) + if !errVal.IsUndefined() { + return errVal + } + + return createPromise(func(resolve, reject js.Value) { + ctx, cancel := context.WithTimeout(context.Background(), timeout) + defer cancel() + + conn, err := nbwebsocket.Dial(ctx, client, url, protocols) + if err != nil { + reject.Invoke(js.ValueOf(fmt.Sprintf("dial websocket: %v", err))) + return + } + + resolve.Invoke(nbwebsocket.NewJSInterface(conn)) + }) + }) +} + +func parseDialWebSocketArgs(args []js.Value) (url string, protocols []string, timeout time.Duration, errVal js.Value) { + if len(args) < 1 || args[0].Type() != js.TypeString { + return "", nil, 0, js.ValueOf("error: dialWebSocket requires a URL string argument") + } + url = args[0].String() + + if len(args) >= 2 && !args[1].IsNull() && !args[1].IsUndefined() { + arr, err := jsStringArray(args[1]) + if err != nil { + return "", nil, 0, js.ValueOf(fmt.Sprintf("error: protocols: %v", err)) + } + protocols = arr + } + + timeout = dialWebSocketTimeout + if len(args) >= 3 && !args[2].IsNull() && !args[2].IsUndefined() { + if args[2].Type() != js.TypeNumber { + return "", nil, 0, js.ValueOf("error: timeoutMs must be a number") + } + timeoutMs := args[2].Int() + if timeoutMs <= 0 { + return "", nil, 0, js.ValueOf("error: timeout must be positive") + } + timeout = time.Duration(timeoutMs) * time.Millisecond + } + + return url, protocols, timeout, js.Undefined() +} + +// jsStringArray converts a JS array of strings to a Go []string. +func jsStringArray(v js.Value) ([]string, error) { + if !v.InstanceOf(js.Global().Get("Array")) { + return nil, fmt.Errorf("expected array") + } + n := v.Length() + out := make([]string, n) + for i := 0; i < n; i++ { + el := v.Index(i) + if el.Type() != js.TypeString { + return nil, fmt.Errorf("element %d is not a string", i) + } + out[i] = el.String() + } + return out, nil +} + // netBirdClientConstructor acts as a JavaScript constructor function func netBirdClientConstructor(_ js.Value, args []js.Value) any { return js.Global().Get("Promise").New(js.FuncOf(func(_ js.Value, promiseArgs []js.Value) any { diff --git a/client/wasm/internal/websocket/websocket.go b/client/wasm/internal/websocket/websocket.go new file mode 100644 index 000000000..19ddaa38c --- /dev/null +++ b/client/wasm/internal/websocket/websocket.go @@ -0,0 +1,304 @@ +//go:build js + +package websocket + +import ( + "context" + "encoding/binary" + "errors" + "fmt" + "io" + "net" + "sync" + "syscall/js" + + "github.com/gobwas/ws" + "github.com/gobwas/ws/wsutil" + netbird "github.com/netbirdio/netbird/client/embed" + log "github.com/sirupsen/logrus" +) + +type closeError struct { + code uint16 + reason string +} + +func (e *closeError) Error() string { + return fmt.Sprintf("websocket closed: %d %s", e.code, e.reason) +} + +// bufferedConn fronts a net.Conn with a reader that serves any bytes buffered +// during the WebSocket handshake before falling through to the raw conn. +type bufferedConn struct { + net.Conn + r io.Reader +} + +func (c *bufferedConn) Read(p []byte) (int, error) { return c.r.Read(p) } + +// Conn wraps a WebSocket connection over a NetBird TCP connection. +type Conn struct { + conn net.Conn + mu sync.Mutex + closed chan struct{} + closeOnce sync.Once + closeErr error +} + +// Dial establishes a WebSocket connection to the given URL through the NetBird network. +// Optional protocols are sent via the Sec-WebSocket-Protocol header. +func Dial(ctx context.Context, client *netbird.Client, rawURL string, protocols []string) (*Conn, error) { + d := ws.Dialer{ + NetDial: client.Dial, + Protocols: protocols, + } + + conn, br, _, err := d.Dial(ctx, rawURL) + if err != nil { + return nil, fmt.Errorf("websocket dial: %w", err) + } + + // br is non-nil when the server pushed frames alongside the handshake + // response; those bytes live in the bufio.Reader and must be drained + // before reading from conn, otherwise we'd skip the first frames. + if br != nil { + if br.Buffered() > 0 { + conn = &bufferedConn{Conn: conn, r: io.MultiReader(br, conn)} + } else { + ws.PutReader(br) + } + } + + return &Conn{ + conn: conn, + closed: make(chan struct{}), + }, nil +} + +// ReadMessage reads the next WebSocket message, handling control frames automatically. +func (c *Conn) ReadMessage() (ws.OpCode, []byte, error) { + for { + msgs, err := wsutil.ReadServerMessage(c.conn, nil) + if err != nil { + return 0, nil, err + } + + for _, msg := range msgs { + if msg.OpCode.IsControl() { + if err := c.handleControl(msg); err != nil { + return 0, nil, err + } + continue + } + return msg.OpCode, msg.Payload, nil + } + } +} + +func (c *Conn) handleControl(msg wsutil.Message) error { + switch msg.OpCode { + case ws.OpPing: + c.mu.Lock() + defer c.mu.Unlock() + return wsutil.WriteClientMessage(c.conn, ws.OpPong, msg.Payload) + case ws.OpClose: + code, reason := parseClosePayload(msg.Payload) + return &closeError{code: code, reason: reason} + default: + return nil + } +} + +// WriteText sends a text WebSocket message. +func (c *Conn) WriteText(data []byte) error { + c.mu.Lock() + defer c.mu.Unlock() + return wsutil.WriteClientMessage(c.conn, ws.OpText, data) +} + +// WriteBinary sends a binary WebSocket message. +func (c *Conn) WriteBinary(data []byte) error { + c.mu.Lock() + defer c.mu.Unlock() + return wsutil.WriteClientMessage(c.conn, ws.OpBinary, data) +} + +// Close sends a close frame with StatusNormalClosure and closes the underlying connection. +func (c *Conn) Close() error { + return c.closeWith(ws.StatusNormalClosure, "") +} + +// closeWith sends a close frame with the given code/reason and closes the underlying connection. +// Used to echo the server's code when responding to a server-initiated close per RFC 6455 ยง5.5.1. +func (c *Conn) closeWith(code ws.StatusCode, reason string) error { + var first bool + c.closeOnce.Do(func() { + first = true + close(c.closed) + + c.mu.Lock() + _ = wsutil.WriteClientMessage(c.conn, ws.OpClose, ws.NewCloseFrameBody(code, reason)) + c.mu.Unlock() + + c.closeErr = c.conn.Close() + }) + + if !first { + return net.ErrClosed + } + return c.closeErr +} + +// NewJSInterface creates a JavaScript object wrapping the WebSocket connection. +// It exposes: send(string|Uint8Array), close(), and callback properties +// onmessage, onclose, onerror. +// +// Callback properties may be set from the JS thread while the read loop +// goroutine reads them. In WASM this is safe because Go and JS share a +// single thread, but the design would need synchronization on +// multi-threaded runtimes. +func NewJSInterface(conn *Conn) js.Value { + obj := js.Global().Get("Object").Call("create", js.Null()) + + sendFunc := js.FuncOf(func(_ js.Value, args []js.Value) any { + if len(args) < 1 { + log.Errorf("websocket send requires a data argument") + return js.ValueOf(false) + } + + data := args[0] + switch data.Type() { + case js.TypeString: + if err := conn.WriteText([]byte(data.String())); err != nil { + log.Errorf("failed to send websocket text: %v", err) + return js.ValueOf(false) + } + default: + buf, err := jsToBytes(data) + if err != nil { + log.Errorf("failed to convert js value to bytes: %v", err) + return js.ValueOf(false) + } + if err := conn.WriteBinary(buf); err != nil { + log.Errorf("failed to send websocket binary: %v", err) + return js.ValueOf(false) + } + } + return js.ValueOf(true) + }) + obj.Set("send", sendFunc) + + closeFunc := js.FuncOf(func(_ js.Value, _ []js.Value) any { + if err := conn.Close(); err != nil { + log.Debugf("failed to close websocket: %v", err) + } + return js.Undefined() + }) + obj.Set("close", closeFunc) + + go func() { + defer func() { + if err := conn.Close(); err != nil && !errors.Is(err, net.ErrClosed) { + log.Debugf("close websocket on readLoop exit: %v", err) + } + }() + readLoop(conn, obj) + // Undefining before Release turns post-close JS calls into TypeError + // instead of a silent "call to released function". + obj.Set("send", js.Undefined()) + obj.Set("close", js.Undefined()) + sendFunc.Release() + closeFunc.Release() + }() + + return obj +} + +func jsToBytes(data js.Value) ([]byte, error) { + var uint8Array js.Value + switch { + case data.InstanceOf(js.Global().Get("Uint8Array")): + uint8Array = data + case data.InstanceOf(js.Global().Get("ArrayBuffer")): + uint8Array = js.Global().Get("Uint8Array").New(data) + default: + return nil, fmt.Errorf("send: unsupported data type, use string, Uint8Array, or ArrayBuffer") + } + + buf := make([]byte, uint8Array.Get("length").Int()) + js.CopyBytesToGo(buf, uint8Array) + return buf, nil +} + +func readLoop(conn *Conn, obj js.Value) { + var ce *closeError + defer func() { invokeOnClose(obj, ce) }() + + for { + select { + case <-conn.closed: + return + default: + } + + op, payload, err := conn.ReadMessage() + if err != nil { + ce = handleReadError(conn, obj, err) + return + } + + dispatchMessage(obj, op, payload) + } +} + +func handleReadError(conn *Conn, obj js.Value, err error) *closeError { + var ce *closeError + if errors.As(err, &ce) { + if cerr := conn.closeWith(ws.StatusCode(ce.code), ce.reason); cerr != nil { + log.Debugf("failed to close websocket after server close frame: %v", cerr) + } + return ce + } + if errors.Is(err, io.EOF) || errors.Is(err, net.ErrClosed) { + return nil + } + if onerror := obj.Get("onerror"); onerror.Truthy() { + onerror.Invoke(js.ValueOf(err.Error())) + } + return nil +} + +func invokeOnClose(obj js.Value, ce *closeError) { + onclose := obj.Get("onclose") + if !onclose.Truthy() { + return + } + if ce != nil { + onclose.Invoke(js.ValueOf(int(ce.code)), js.ValueOf(ce.reason)) + return + } + onclose.Invoke() +} + +func dispatchMessage(obj js.Value, op ws.OpCode, payload []byte) { + onmessage := obj.Get("onmessage") + if !onmessage.Truthy() { + return + } + switch op { + case ws.OpText: + onmessage.Invoke(js.ValueOf(string(payload))) + case ws.OpBinary: + uint8Array := js.Global().Get("Uint8Array").New(len(payload)) + js.CopyBytesToJS(uint8Array, payload) + onmessage.Invoke(uint8Array) + } +} + +func parseClosePayload(payload []byte) (uint16, string) { + if len(payload) < 2 { + return 1005, "" // RFC 6455: No Status Rcvd + } + code := binary.BigEndian.Uint16(payload[:2]) + return code, string(payload[2:]) +} diff --git a/go.mod b/go.mod index 0b9cc9f29..2858d2044 100644 --- a/go.mod +++ b/go.mod @@ -56,6 +56,7 @@ require ( github.com/fsnotify/fsnotify v1.9.0 github.com/gliderlabs/ssh v0.3.8 github.com/go-jose/go-jose/v4 v4.1.4 + github.com/gobwas/ws v1.4.0 github.com/goccy/go-yaml v1.18.0 github.com/godbus/dbus/v5 v5.1.0 github.com/golang-jwt/jwt/v5 v5.3.1 @@ -215,6 +216,8 @@ require ( github.com/go-viper/mapstructure/v2 v2.5.0 // indirect github.com/go-webauthn/webauthn v0.16.4 // indirect github.com/go-webauthn/x v0.2.3 // indirect + github.com/gobwas/httphead v0.1.0 // indirect + github.com/gobwas/pool v0.2.1 // indirect github.com/gogo/protobuf v1.3.2 // indirect github.com/golang-jwt/jwt/v4 v4.5.2 // indirect github.com/google/btree v1.1.3 // indirect diff --git a/go.sum b/go.sum index bc78d17e5..1768ee069 100644 --- a/go.sum +++ b/go.sum @@ -249,6 +249,12 @@ github.com/go-webauthn/webauthn v0.16.4 h1:R9jqR/cYZa7hRquFF7Za/8qoH/K/TIs1/Q/4C github.com/go-webauthn/webauthn v0.16.4/go.mod h1:SU2ljAgToTV/YLPI0C05QS4qn+e04WpB5g1RMfcZfS4= github.com/go-webauthn/x v0.2.3 h1:8oArS+Rc1SWFLXhE17KZNx258Z4kUSyaDgsSncCO5RA= github.com/go-webauthn/x v0.2.3/go.mod h1:tM04GF3V6VYq79AZMl7vbj4q6pz9r7L2criWRzbWhPk= +github.com/gobwas/httphead v0.1.0 h1:exrUm0f4YX0L7EBwZHuCF4GDp8aJfVeBrlLQrs6NqWU= +github.com/gobwas/httphead v0.1.0/go.mod h1:O/RXo79gxV8G+RqlR/otEwx4Q36zl9rqC5u12GKvMCM= +github.com/gobwas/pool v0.2.1 h1:xfeeEhW7pwmX8nuLVlqbzVc7udMDrwetjEv+TZIz1og= +github.com/gobwas/pool v0.2.1/go.mod h1:q8bcK0KcYlCgd9e7WYLm9LpyS+YeLd8JVDW6WezmKEw= +github.com/gobwas/ws v1.4.0 h1:CTaoG1tojrh4ucGPcoJFiAQUAsEWekEWvLy7GsVNqGs= +github.com/gobwas/ws v1.4.0/go.mod h1:G3gNqMNtPppf5XUz7O4shetPpcZ1VJ7zt18dlUeakrc= github.com/goccy/go-yaml v1.18.0 h1:8W7wMFS12Pcas7KU+VVkaiCng+kG8QiFeFwzFb+rwuw= github.com/goccy/go-yaml v1.18.0/go.mod h1:XBurs7gK8ATbW4ZPGKgcbrY1Br56PdM69F7LkFRi1kA= github.com/godbus/dbus/v5 v5.1.0 h1:4KLkAxT3aOY8Li4FRJe/KvhoNFFxo0m6fNuFUO8QJUk= @@ -845,6 +851,7 @@ golang.org/x/sys v0.0.0-20220908164124-27713097b956/go.mod h1:oPkhp1MJrh7nUepCBc golang.org/x/sys v0.1.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg= golang.org/x/sys v0.2.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg= golang.org/x/sys v0.5.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg= +golang.org/x/sys v0.6.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg= golang.org/x/sys v0.7.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg= golang.org/x/sys v0.8.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg= golang.org/x/sys v0.10.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=