add stats.pprof flag to expose pprof on stats listener

* specifically report if nebula start fails because of tun device name

* close all routines when closing the tun

examples/config.yml

@@ -318,6 +318,11 @@ logging:
   #subsystem: nebula
   #interval: 10s
 
+  # enables pprof endpoints on the stats listener
+  # endpoints: /debug/pprof/, /debug/pprof/cmdline, /debug/pprof/profile,
+  #            /debug/pprof/symbol, /debug/pprof/trace
+  #pprof: false
+
   # enables counter metrics for meta packets
   #   e.g.: `messages.tx.handshake`
   # NOTE: `message.{tx,rx}.recv_error` is always emitted

interface.go

@@ -490,6 +490,14 @@ func (f *Interface) Close() error {
 			f.l.WithError(err).Error("Error while closing udp socket")
 		}
 	}
+	for i, r := range f.readers {
+		if i == 0 {
+			continue // f.readers[0] is f.inside, which we want to save for last
+		}
+		if err := r.Close(); err != nil {
+			f.l.WithError(err).Error("Error while closing tun reader")
+		}
+	}
 
 	// Release the tun device
 	return f.inside.Close()

lighthouse_test.go

@@ -1,7 +1,6 @@
 package nebula
 
 import (
-	"context"
 	"encoding/binary"
 	"fmt"
 	"net/netip"
@@ -42,14 +41,14 @@ func Test_lhStaticMapping(t *testing.T) {
 	c := config.NewC(l)
 	c.Settings["lighthouse"] = map[string]any{"hosts": []any{lh1}}
 	c.Settings["static_host_map"] = map[string]any{lh1: []any{"1.1.1.1:4242"}}
-	_, err := NewLightHouseFromConfig(context.Background(), l, c, cs, nil, nil)
+	_, err := NewLightHouseFromConfig(t.Context(), l, c, cs, nil, nil)
 	require.NoError(t, err)
 
 	lh2 := "10.128.0.3"
 	c = config.NewC(l)
 	c.Settings["lighthouse"] = map[string]any{"hosts": []any{lh1, lh2}}
 	c.Settings["static_host_map"] = map[string]any{lh1: []any{"100.1.1.1:4242"}}
-	_, err = NewLightHouseFromConfig(context.Background(), l, c, cs, nil, nil)
+	_, err = NewLightHouseFromConfig(t.Context(), l, c, cs, nil, nil)
 	require.EqualError(t, err, "lighthouse 10.128.0.3 does not have a static_host_map entry")
 }
 
@@ -71,7 +70,7 @@ func TestReloadLighthouseInterval(t *testing.T) {
 	}
 
 	c.Settings["static_host_map"] = map[string]any{lh1: []any{"1.1.1.1:4242"}}
-	lh, err := NewLightHouseFromConfig(context.Background(), l, c, cs, nil, nil)
+	lh, err := NewLightHouseFromConfig(t.Context(), l, c, cs, nil, nil)
 	require.NoError(t, err)
 	lh.ifce = &mockEncWriter{}
 
@@ -99,7 +98,7 @@ func BenchmarkLighthouseHandleRequest(b *testing.B) {
 	}
 
 	c := config.NewC(l)
-	lh, err := NewLightHouseFromConfig(context.Background(), l, c, cs, nil, nil)
+	lh, err := NewLightHouseFromConfig(b.Context(), l, c, cs, nil, nil)
 	require.NoError(b, err)
 
 	hAddr := netip.MustParseAddrPort("4.5.6.7:12345")
@@ -202,7 +201,7 @@ func TestLighthouse_Memory(t *testing.T) {
 		myVpnNetworks:      []netip.Prefix{myVpnNet},
 		myVpnNetworksTable: nt,
 	}
-	lh, err := NewLightHouseFromConfig(context.Background(), l, c, cs, nil, nil)
+	lh, err := NewLightHouseFromConfig(t.Context(), l, c, cs, nil, nil)
 	lh.ifce = &mockEncWriter{}
 	require.NoError(t, err)
 	lhh := lh.NewRequestHandler()
@@ -288,7 +287,7 @@ func TestLighthouse_reload(t *testing.T) {
 		myVpnNetworksTable: nt,
 	}
 
-	lh, err := NewLightHouseFromConfig(context.Background(), l, c, cs, nil, nil)
+	lh, err := NewLightHouseFromConfig(t.Context(), l, c, cs, nil, nil)
 	require.NoError(t, err)
 
 	nc := map[string]any{
@@ -523,7 +522,7 @@ func TestLighthouse_Dont_Delete_Static_Hosts(t *testing.T) {
 		myVpnNetworks:      []netip.Prefix{myVpnNet},
 		myVpnNetworksTable: nt,
 	}
-	lh, err := NewLightHouseFromConfig(context.Background(), l, c, cs, nil, nil)
+	lh, err := NewLightHouseFromConfig(t.Context(), l, c, cs, nil, nil)
 	require.NoError(t, err)
 	lh.ifce = &mockEncWriter{}
 
@@ -589,7 +588,7 @@ func TestLighthouse_DeletesWork(t *testing.T) {
 		myVpnNetworks:      []netip.Prefix{myVpnNet},
 		myVpnNetworksTable: nt,
 	}
-	lh, err := NewLightHouseFromConfig(context.Background(), l, c, cs, nil, nil)
+	lh, err := NewLightHouseFromConfig(t.Context(), l, c, cs, nil, nil)
 	require.NoError(t, err)
 	lh.ifce = &mockEncWriter{}
 

overlay/tun.go

@@ -12,6 +12,15 @@ import (
 
 const DefaultMTU = 1300
 
+type NameError struct {
+	Name       string
+	Underlying error
+}
+
+func (e *NameError) Error() string {
+	return fmt.Sprintf("could not set tun device name: %s because %s", e.Name, e.Underlying)
+}
+
 // TODO: We may be able to remove routines
 type DeviceFactory func(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error)
 

overlay/tun_freebsd.go

@@ -266,7 +266,7 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 		}
 
 		// Set the device name
-		ioctl(fd, syscall.SIOCSIFNAME, uintptr(unsafe.Pointer(&ifrr)))
+		_ = ioctl(fd, syscall.SIOCSIFNAME, uintptr(unsafe.Pointer(&ifrr)))
 	}
 
 	t := &tun{

overlay/tun_linux.go

@@ -112,9 +112,13 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueu
 	if multiqueue {
 		req.Flags |= unix.IFF_MULTI_QUEUE
 	}
-	copy(req.Name[:], c.GetString("tun.dev", ""))
+	nameStr := c.GetString("tun.dev", "")
+	copy(req.Name[:], nameStr)
 	if err = ioctl(uintptr(fd), uintptr(unix.TUNSETIFF), uintptr(unsafe.Pointer(&req))); err != nil {
-		return nil, err
+		return nil, &NameError{
+			Name:       nameStr,
+			Underlying: err,
+		}
 	}
 	name := strings.Trim(string(req.Name[:]), "\x00")
 
@@ -713,6 +717,7 @@ func (t *tun) Close() error {
 
 	if t.ioctlFd > 0 {
 		_ = os.NewFile(t.ioctlFd, "ioctlFd").Close()
+		t.ioctlFd = 0
 	}
 
 	return nil

overlay/tun_windows.go

@@ -74,7 +74,10 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 		l.WithError(err).Debug("Failed to create wintun device, retrying")
 		tunDevice, err = wintun.CreateTUNWithRequestedGUID(deviceName, guid, t.MTU)
 		if err != nil {
-			return nil, fmt.Errorf("create TUN device failed: %w", err)
+			return nil, &NameError{
+				Name:       deviceName,
+				Underlying: fmt.Errorf("create TUN device failed: %w", err),
+			}
 		}
 	}
 	t.tun = tunDevice.(*wintun.NativeTun)

stats.go

@@ -6,6 +6,7 @@ import (
 	"log"
 	"net"
 	"net/http"
+	"net/http/pprof"
 	"runtime"
 	"strconv"
 	"time"
@@ -93,6 +94,8 @@ func startPrometheusStats(l *logrus.Logger, i time.Duration, c *config.C, buildV
 		return nil, fmt.Errorf("stats.path should not be empty")
 	}
 
+	enablePprof := c.GetBool("stats.pprof", false)
+
 	pr := prometheus.NewRegistry()
 	pClient := mp.NewPrometheusProvider(metrics.DefaultRegistry, namespace, subsystem, pr, i)
 	if !configTest {
@@ -119,6 +122,14 @@ func startPrometheusStats(l *logrus.Logger, i time.Duration, c *config.C, buildV
 		startFn = func() {
 			l.Infof("Prometheus stats listening on %s at %s", listen, path)
 			http.Handle(path, promhttp.HandlerFor(pr, promhttp.HandlerOpts{ErrorLog: l}))
+			if enablePprof {
+				l.Info("pprof endpoints enabled on stats listener")
+				http.HandleFunc("GET /debug/pprof/", pprof.Index)
+				http.HandleFunc("GET /debug/pprof/cmdline", pprof.Cmdline)
+				http.HandleFunc("GET /debug/pprof/profile", pprof.Profile)
+				http.HandleFunc("GET /debug/pprof/symbol", pprof.Symbol)
+				http.HandleFunc("GET /debug/pprof/trace", pprof.Trace)
+			}
 			log.Fatal(http.ListenAndServe(listen, nil))
 		}
 	}