stub: cancel Start() for early lost connection.

If our connection gets closed before we had a chance to get
Configure()'d by the runtime, cancel Start()'s wait for the
result by letting it know about the failure.

Signed-off-by: Krisztian Litkey <krisztian.litkey@intel.com>

Author: klihub

pkg/adaptation/adaptation_suite_test.go

@@ -18,6 +18,7 @@ package adaptation_test
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"os"
 	"path/filepath"
@@ -38,6 +39,7 @@ import (
 	"github.com/containerd/nri/pkg/api"
 	"github.com/containerd/nri/pkg/plugin"
 	validator "github.com/containerd/nri/plugins/default-validator/builtin"
+	"github.com/containerd/ttrpc"
 	rspec "github.com/opencontainers/runtime-spec/specs-go"
 )
 
@@ -94,6 +96,57 @@ var _ = Describe("Configuration", func() {
 			Expect(plugin.Start(s.dir)).ToNot(Succeed())
 		})
 	})
+
+	When("early connection loss during plugin startup", func() {
+		BeforeEach(func() {
+			nri.SetPluginRegistrationTimeout(1 * time.Second)
+			nri.SetPluginRequestTimeout(1 * time.Second)
+
+			s.Prepare(
+				&mockRuntime{},
+				&mockPlugin{
+					idx:  "00",
+					name: "test",
+					configure: func(_ *mockPlugin, _, _, _ string) (api.EventMask, error) {
+						time.Sleep(3 * time.Second)
+						return 0, nil
+					},
+				},
+			)
+		})
+
+		AfterEach(func() {
+			nri.SetPluginRegistrationTimeout(nri.DefaultPluginRegistrationTimeout)
+			nri.SetPluginRequestTimeout(nri.DefaultPluginRequestTimeout)
+		})
+
+		It("should not cause a plugin to get stuck", func() {
+			var (
+				runtime = s.runtime
+				plugin  = s.plugins[0]
+				errCh   = make(chan error, 1)
+				err     error
+			)
+
+			Expect(runtime.Start(s.dir)).To(Succeed())
+
+			go func() {
+				err := plugin.Start(s.dir)
+				errCh <- err
+			}()
+
+			ctx, cancel := context.WithTimeout(context.Background(), 3*time.Second)
+			defer cancel()
+
+			select {
+			case <-ctx.Done():
+				err = ctx.Err()
+			case err = <-errCh:
+			}
+
+			Expect(errors.Is(err, ttrpc.ErrClosed)).To(BeTrue())
+		})
+	})
 })
 
 var _ = Describe("Adaptation", func() {

pkg/adaptation/suite_test.go

@@ -378,6 +378,7 @@ type mockPlugin struct {
 	pods map[string]*api.PodSandbox
 	ctrs map[string]*api.Container
 
+	configure            func(*mockPlugin, string, string, string) (stub.EventMask, error)
 	runPodSandbox        func(*mockPlugin, *api.PodSandbox, *api.Container) error
 	updatePodSandbox     func(*mockPlugin, *api.PodSandbox, *api.LinuxResources, *api.LinuxResources) error
 	postUpdatePodSandbox func(*mockPlugin, *api.PodSandbox, *api.Container) error
@@ -559,18 +560,28 @@ func (m *mockPlugin) onClose() {
 	}
 }
 
-func (m *mockPlugin) Configure(_ context.Context, _, runtime, version string) (stub.EventMask, error) {
+func (m *mockPlugin) Configure(_ context.Context, cfg, runtime, version string) (stub.EventMask, error) {
+	var (
+		events stub.EventMask
+		err    error
+	)
+
 	m.q.Add(PluginConfigured)
 
 	m.runtime = runtime
 	m.version = version
 
-	events := m.mask
+	if m.configure != nil {
+		events, err = m.configure(m, cfg, runtime, version)
+	} else {
+		events, err = m.mask, nil
+	}
+
 	if m.validateAdjustment == nil {
 		events.Clear(api.Event_VALIDATE_CONTAINER_ADJUSTMENT)
 	}
 
-	return events, nil
+	return events, err
 }
 
 func (m *mockPlugin) Synchronize(_ context.Context, pods []*api.PodSandbox, ctrs []*api.Container) ([]*api.ContainerUpdate, error) {

pkg/stub/stub.go

@@ -584,6 +584,12 @@ func (stub *stub) register(ctx context.Context) error {
 
 // Handle a lost connection.
 func (stub *stub) connClosed() {
+	select {
+	// if our connection gets closed before we get Configure()'d, let Start() know
+	case stub.cfgErrC <- ttrpc.ErrClosed:
+	default:
+	}
+
 	stub.Lock()
 	stub.close()
 	stub.Unlock()
@@ -628,7 +634,10 @@ func (stub *stub) Configure(ctx context.Context, req *api.ConfigureRequest) (rpl
 	stub.requestTimeout = time.Duration(req.RequestTimeout * int64(time.Millisecond))
 
 	defer func() {
-		stub.cfgErrC <- retErr
+		select {
+		case stub.cfgErrC <- retErr:
+		default:
+		}
 	}()
 
 	if handler := stub.handlers.Configure; handler == nil {