shimv2: handle sigint/sigterm

This causes sigint/sigterm to trigger a shutdown of the shim.
It is needed because otherwise the v2 shim hangs system shutdown.

Signed-off-by: Brian Goff <cpuguy83@gmail.com>
(cherry picked from commit 3ffb6a6)
Signed-off-by: Varsha Teratipally <teratipally@google.com>

runtime/v2/shim/shim.go

@@ -43,6 +43,7 @@ type Client struct {
 	service shimapi.TaskService
 	context context.Context
 	signals chan os.Signal
+	shutdown context.CancelFunc
 }
 
 // Publisher for events
@@ -214,7 +215,7 @@ func run(id string, initFunc Init, config Config) error {
 			"pid":       os.Getpid(),
 			"namespace": namespaceFlag,
 		})
-		go handleSignals(ctx, logger, signals)
+		go reap(ctx, logger, signals)
 		response, err := service.Cleanup(ctx)
 		if err != nil {
 			return err
@@ -248,7 +249,7 @@ func run(id string, initFunc Init, config Config) error {
 				return err
 			}
 		}
-		client := NewShimClient(ctx, service, signals)
+		client := NewShimClient(ctx, service, signals, cancel)
 		if err := client.Serve(); err != nil {
 			if err != context.Canceled {
 				return err
@@ -271,11 +272,12 @@ func run(id string, initFunc Init, config Config) error {
 }
 
 // NewShimClient creates a new shim server client
-func NewShimClient(ctx context.Context, svc shimapi.TaskService, signals chan os.Signal) *Client {
+func NewShimClient(ctx context.Context, svc shimapi.TaskService, signals chan os.Signal, shutdown context.CancelFunc) *Client {
 	s := &Client{
 		service: svc,
 		context: ctx,
 		signals: signals,
+		shutdown: shutdown,
 	}
 	return s
 }
@@ -310,7 +312,8 @@ func (s *Client) Serve() error {
 			dumpStacks(logger)
 		}
 	}()
-	return handleSignals(s.context, logger, s.signals)
+	go handleExitSignals(s.context, logger, s.shutdown)
+	return reap(s.context, logger, s.signals)
 }
 
 // serve serves the ttrpc API over a unix socket at the provided path
@@ -323,7 +326,7 @@ func serve(ctx context.Context, server *ttrpc.Server, path string) error {
 	go func() {
 		defer l.Close()
 		if err := server.Serve(ctx, l); err != nil &&
-			!strings.Contains(err.Error(), "use of closed network connection") {
+				!strings.Contains(err.Error(), "use of closed network connection") {
 			logrus.WithError(err).Fatal("containerd-shim: ttrpc server failure")
 		}
 	}()
@@ -343,4 +346,4 @@ func dumpStacks(logger *logrus.Entry) {
 	}
 	buf = buf[:stackSize]
 	logger.Infof("=== BEGIN goroutine stack dump ===\n%s\n=== END goroutine stack dump ===", buf)
-}
+}

runtime/v2/shim/shim_unix.go

@@ -70,14 +70,16 @@ func serveListener(path string) (net.Listener, error) {
 	return l, nil
 }
 
-func handleSignals(ctx context.Context, logger *logrus.Entry, signals chan os.Signal) error {
+func reap(ctx context.Context, logger *logrus.Entry, signals chan os.Signal) error {
 	logger.Info("starting signal loop")
 
 	for {
 		select {
 		case <-ctx.Done():
 			return ctx.Err()
 		case s := <-signals:
+			// Exit signals are handled separately from this loop
+			// They get registered with this channel so that we can ignore such signals for short-running actions (e.g. `delete`)
 			switch s {
 			case unix.SIGCHLD:
 				if err := reaper.Reap(); err != nil {
@@ -89,6 +91,22 @@ func handleSignals(ctx context.Context, logger *logrus.Entry, signals chan os.Si
 	}
 }
 
+func handleExitSignals(ctx context.Context, logger *logrus.Entry, cancel context.CancelFunc) {
+	ch := make(chan os.Signal, 32)
+	signal.Notify(ch, syscall.SIGINT, syscall.SIGTERM)
+
+	for {
+		select {
+		case s := <-ch:
+			logger.WithField("signal", s).Debugf("Caught exit signal")
+			cancel()
+			return
+		case <-ctx.Done():
+			return
+		}
+	}
+}
+
 func openLog(ctx context.Context, _ string) (io.Writer, error) {
 	return fifo.OpenFifoDup2(ctx, "log", unix.O_WRONLY, 0700, int(os.Stderr.Fd()))
 }

runtime/v2/shim/shim_windows.go

@@ -48,10 +48,13 @@ func serveListener(path string) (net.Listener, error) {
 	return nil, errors.New("not supported")
 }
 
-func handleSignals(ctx context.Context, logger *logrus.Entry, signals chan os.Signal) error {
+func reap(ctx context.Context, logger *logrus.Entry, signals chan os.Signal) error {
 	return errors.New("not supported")
 }
 
+func handleExitSignals(ctx context.Context, logger *logrus.Entry, cancel context.CancelFunc) {
+}
+
 func openLog(ctx context.Context, _ string) (io.Writer, error) {
 	return nil, errors.New("not supported")
 }