Merge remote-tracking branch 'origin/master' into fips140

cmd/nebula-service/main.go

@@ -53,7 +53,12 @@ func main() {
 	l := logging.NewLogger(os.Stdout)
 
 	if *serviceFlag != "" {
-		if err := doService(configPath, configTest, Build, serviceFlag); err != nil {
+		if *configTest {
+			fmt.Println("-test is not supported with -service, run the config test without -service")
+			os.Exit(1)
+		}
+
+		if err := doService(configPath, Build, serviceFlag); err != nil {
 			l.Error("Service command failed", "error", err)
 			os.Exit(1)
 		}
@@ -93,15 +98,14 @@ func main() {
 	}
 
 	if !*configTest {
-		wait, err := ctrl.Start()
-		if err != nil {
+		if err := ctrl.Start(); err != nil {
 			util.LogWithContextIfNeeded("Error while running", err, l)
 			os.Exit(1)
 		}
 
 		go ctrl.ShutdownBlock()
 
-		if err := wait(); err != nil {
+		if err := ctrl.Wait(); err != nil {
 			l.Error("Nebula stopped due to fatal error", "error", err)
 			os.Exit(2)
 		}

cmd/nebula-service/service.go

@@ -3,6 +3,7 @@ package main
 import (
 	"fmt"
 	"log"
+	"os"
 
 	"github.com/kardianos/service"
 	"github.com/slackhq/nebula"
@@ -14,7 +15,6 @@ var logger service.Logger
 
 type program struct {
 	configPath *string
-	configTest *bool
 	build      string
 	control    *nebula.Control
 }
@@ -40,22 +40,41 @@ func (p *program) Start(s service.Service) error {
 		}
 	})
 
-	p.control, err = nebula.Main(c, *p.configTest, Build, l, nil)
+	p.control, err = nebula.Main(c, false, Build, l, nil)
 	if err != nil {
 		return err
 	}
 
-	p.control.Start()
+	if err := p.control.Start(); err != nil {
+		return err
+	}
+
+	// Nebula can stop itself on a fatal packet reader error, make sure to log it if it happens.
+	go func() {
+		if err := p.control.Wait(); err != nil {
+			logger.Error(fmt.Sprintf("Nebula stopped due to fatal error: %v", err))
+			os.Exit(2)
+		}
+	}()
+
 	return nil
 }
 
 func (p *program) Stop(s service.Service) error {
 	logger.Info("Nebula service stopping.")
+	if p.control == nil {
+		return nil
+	}
+
 	p.control.Stop()
+
+	// block until nebula has fully drained before reporting stopped.
+	// error logging is handled by Start.
+	_ = p.control.Wait()
 	return nil
 }
 
-func doService(configPath *string, configTest *bool, build string, serviceFlag *string) error {
+func doService(configPath *string, build string, serviceFlag *string) error {
 	if *configPath == "" {
 		p, err := config.DefaultPath()
 		if err != nil {
@@ -73,7 +92,6 @@ func doService(configPath *string, configTest *bool, build string, serviceFlag *
 
 	prg := &program{
 		configPath: configPath,
-		configTest: configTest,
 		build:      build,
 	}
 
@@ -105,8 +123,9 @@ func doService(configPath *string, configTest *bool, build string, serviceFlag *
 	switch *serviceFlag {
 	case "run":
 		if err := s.Run(); err != nil {
-			// Route any errors to the system logger
+			// Route any errors to the system logger and report the failure
 			logger.Error(err)
+			return err
 		}
 	default:
 		if err := service.Control(s, *serviceFlag); err != nil {

cmd/nebula/close_on_timer_test.go

@@ -0,0 +1,96 @@
+//go:build linux && !android && !e2e_testing
+
+package main
+
+import (
+	"fmt"
+	"net/netip"
+	"os"
+	"path/filepath"
+	"runtime"
+	"testing"
+	"time"
+
+	"github.com/slackhq/nebula"
+	"github.com/slackhq/nebula/cert"
+	cert_test "github.com/slackhq/nebula/cert_test"
+	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/test"
+	"github.com/stretchr/testify/require"
+)
+
+// TestControlStopClosesOnTimer reproduces the dnclient lifecycle: nebula runs as
+// a library, and on a config update dnclient calls Stop() in-process to tear the
+// old instance down before starting a new one. This boots a real nebula (real
+// blocking UDP sockets, tun disabled), lets it run, then Stop()s it on a timer
+// and asserts it actually closes. If the reader goroutines parked in recvmmsg
+// don't wake on Close(), Wait() blocks forever and this fails with a goroutine
+// dump instead of relying on a process signal to unstick them.
+func TestControlStopClosesOnTimer(t *testing.T) {
+	l := test.NewLogger()
+	dir := t.TempDir()
+
+	before := time.Now().Add(-time.Hour)
+	after := time.Now().Add(time.Hour)
+	ca, _, caKey, caPEM := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, before, after, nil, nil, nil)
+	networks := []netip.Prefix{netip.MustParsePrefix("10.0.0.1/24")}
+	_, _, keyPEM, certPEM := cert_test.NewTestCert(cert.Version2, cert.Curve_CURVE25519, ca, caKey, "close-on-timer", before, after, networks, nil, nil)
+
+	caPath := filepath.Join(dir, "ca.pem")
+	certPath := filepath.Join(dir, "cert.pem")
+	keyPath := filepath.Join(dir, "key.pem")
+	require.NoError(t, os.WriteFile(caPath, caPEM, 0o600))
+	require.NoError(t, os.WriteFile(certPath, certPEM, 0o600))
+	require.NoError(t, os.WriteFile(keyPath, keyPEM, 0o600))
+
+	// tun disabled so no device/root is needed; routines: 2 so we exercise the
+	// multi-socket (SO_REUSEPORT) teardown, which is where dnclient runs.
+	configBody := fmt.Sprintf(`
+pki:
+  ca: %s
+  cert: %s
+  key: %s
+listen:
+  host: 127.0.0.1
+  port: 0
+tun:
+  disabled: true
+firewall:
+  outbound:
+    - port: any
+      proto: any
+      host: any
+  inbound:
+    - port: any
+      proto: any
+      host: any
+routines: 2
+`, caPath, certPath, keyPath)
+	require.NoError(t, os.WriteFile(filepath.Join(dir, "config.yml"), []byte(configBody), 0o600))
+
+	c := config.NewC(l)
+	require.NoError(t, c.Load(dir))
+
+	ctrl, err := nebula.Main(c, false, "close-on-timer", l, nil)
+	require.NoError(t, err)
+	require.NoError(t, ctrl.Start())
+
+	// Run like a live nebula, then close on a timer, exactly as dnclient does.
+	<-time.NewTimer(5 * time.Second).C
+
+	stopped := make(chan struct{})
+	go func() {
+		ctrl.Stop() // closes the udp sockets (shutdown(2)) and the tun
+		ctrl.Wait() // blocks until every reader goroutine has returned
+		close(stopped)
+	}()
+
+	select {
+	case <-stopped:
+		t.Log("nebula closed cleanly on timer")
+	case <-time.After(10 * time.Second):
+		buf := make([]byte, 1<<20)
+		n := runtime.Stack(buf, true)
+		t.Fatalf("nebula did NOT close within 10s of Stop(): a blocking reader never woke\n%s", buf[:n])
+	}
+}

cmd/nebula/main.go

@@ -84,8 +84,7 @@ func main() {
 	}
 
 	if !*configTest {
-		wait, err := ctrl.Start()
-		if err != nil {
+		if err := ctrl.Start(); err != nil {
 			util.LogWithContextIfNeeded("Error while running", err, l)
 			os.Exit(1)
 		}
@@ -93,7 +92,7 @@ func main() {
 		go ctrl.ShutdownBlock()
 		notifyReady(l)
 
-		if err := wait(); err != nil {
+		if err := ctrl.Wait(); err != nil {
 			l.Error("Nebula stopped due to fatal error", "error", err)
 			os.Exit(2)
 		}