[release/0.8] Remove blocking wait on container exit for every exec c…

…reated (microsoft#1605)

* Remove blocking wait on container exit for every exec created

Commit fixes the memory leak seen in the shim.
It removes creation of channel that waits on container exit
for every new exec. Instead, the container wait channel is exposed
through WaitChannel() function which callers can use to decide
if container has exited or not.

It also fixes CI to use the same version of golang and updates golangci-lint version to v1.48

Signed-off-by: Kirtana Ashok <[email protected]>
(cherry picked from commit 5fc00c5)
Signed-off-by: Kirtana Ashok <[email protected]>

* update golangci-lint version

Signed-off-by: Kirtana Ashok <[email protected]>

Signed-off-by: Kirtana Ashok <[email protected]>
Co-authored-by: Kirtana Ashok <[email protected]>

.github/workflows/ci.yml

@@ -6,6 +6,7 @@ on:
 env:
   GOPROXY: off
   GOFLAGS: -mod=vendor
+  GO_VERSION: '1.15.x'
 
 jobs:
   lint:
@@ -14,12 +15,12 @@ jobs:
       - uses: actions/checkout@v2
       - uses: actions/setup-go@v2
         with:
-          go-version: '^1.15.0'
+          go-version: ${{ env.GO_VERSION }}
 
       - name: golangci-lint
         uses: golangci/golangci-lint-action@v2
         with:
-          version: v1.38.0
+          version: v1.48
           args: --timeout=5m
           only-new-issues: true
 
@@ -29,7 +30,7 @@ jobs:
       - uses: actions/checkout@v2
       - uses: actions/setup-go@v2
         with:
-          go-version: '^1.15.0'
+          go-version: ${{ env.GO_VERSION }}
 
       - run: go test -gcflags=all=-d=checkptr -v ./... -tags admin
       - run: go test -gcflags=all=-d=checkptr -v ./internal -tags admin
@@ -61,7 +62,7 @@ jobs:
       - uses: actions/checkout@v2
       - uses: actions/setup-go@v2
         with:
-          go-version: '^1.15.0'
+          go-version: ${{ env.GO_VERSION }}
 
       - run: go build ./cmd/containerd-shim-runhcs-v1
       - run: go build ./cmd/runhcs

cmd/containerd-shim-runhcs-v1/exec_hcs.go

@@ -481,13 +481,9 @@ func (he *hcsExec) waitForContainerExit() {
 		trace.StringAttribute("tid", he.tid),
 		trace.StringAttribute("eid", he.id))
 
-	cexit := make(chan struct{})
-	go func() {
-		_ = he.c.Wait()
-		close(cexit)
-	}()
+	// wait for container or process to exit and ckean up resrources
 	select {
-	case <-cexit:
+	case <-he.c.WaitChannel():
 		// Container exited first. We need to force the process into the exited
 		// state and cleanup any resources
 		he.sl.Lock()

internal/cow/cow.go

@@ -86,6 +86,12 @@ type Container interface {
 	// container to be terminated by some error condition (including calling
 	// Close).
 	Wait() error
+	// WaitChannel returns the wait channel of the container
+	WaitChannel() <-chan struct{}
+	// WaitError returns the container termination error.
+	// This function should only be called after the channel in WaitChannel()
+	// is closed. Otherwise it is not thread safe.
+	WaitError() error
 	// Modify sends a request to modify container resources
 	Modify(ctx context.Context, config interface{}) error
 }

internal/gcs/container.go

@@ -26,6 +26,10 @@ type Container struct {
 	notifyCh  chan struct{}
 	closeCh   chan struct{}
 	closeOnce sync.Once
+	// waitBlock is the channel used to wait for container shutdown or termination
+	waitBlock chan struct{}
+	// waitError indicates the container termination error if any
+	waitError error
 }
 
 var _ cow.Container = &Container{}
@@ -39,10 +43,11 @@ func (gc *GuestConnection) CreateContainer(ctx context.Context, cid string, conf
 	span.AddAttributes(trace.StringAttribute("cid", cid))
 
 	c := &Container{
-		gc:       gc,
-		id:       cid,
-		notifyCh: make(chan struct{}),
-		closeCh:  make(chan struct{}),
+		gc:        gc,
+		id:        cid,
+		notifyCh:  make(chan struct{}),
+		closeCh:   make(chan struct{}),
+		waitBlock: make(chan struct{}),
 	}
 	err = gc.requestNotify(cid, c.notifyCh)
 	if err != nil {
@@ -65,10 +70,11 @@ func (gc *GuestConnection) CreateContainer(ctx context.Context, cid string, conf
 // container that is already running inside the UVM (after cloning).
 func (gc *GuestConnection) CloneContainer(ctx context.Context, cid string) (_ *Container, err error) {
 	c := &Container{
-		gc:       gc,
-		id:       cid,
-		notifyCh: make(chan struct{}),
-		closeCh:  make(chan struct{}),
+		gc:        gc,
+		id:        cid,
+		notifyCh:  make(chan struct{}),
+		closeCh:   make(chan struct{}),
+		waitBlock: make(chan struct{}),
 	}
 	err = gc.requestNotify(cid, c.notifyCh)
 	if err != nil {
@@ -95,6 +101,8 @@ func (c *Container) Close() error {
 		_, span := trace.StartSpan(context.Background(), "gcs::Container::Close")
 		defer span.End()
 		span.AddAttributes(trace.StringAttribute("cid", c.id))
+
+		close(c.closeCh)
 	})
 	return nil
 }
@@ -224,23 +232,33 @@ func (c *Container) Terminate(ctx context.Context) (err error) {
 	return c.shutdown(ctx, rpcShutdownForced)
 }
 
+func (c *Container) WaitChannel() <-chan struct{} {
+	return c.waitBlock
+}
+
+func (c *Container) WaitError() error {
+	return c.waitError
+}
+
 // Wait waits for the container to terminate (or Close to be called, or the
 // guest connection to terminate).
 func (c *Container) Wait() error {
-	select {
-	case <-c.notifyCh:
-		return nil
-	case <-c.closeCh:
-		return errors.New("container closed")
-	}
+	<-c.WaitChannel()
+	return c.WaitError()
 }
 
 func (c *Container) waitBackground() {
 	ctx, span := trace.StartSpan(context.Background(), "gcs::Container::waitBackground")
 	defer span.End()
 	span.AddAttributes(trace.StringAttribute("cid", c.id))
 
-	err := c.Wait()
+	select {
+	case <-c.notifyCh:
+	case <-c.closeCh:
+		c.waitError = errors.New("container closed")
+	}
+	close(c.waitBlock)
+
 	log.G(ctx).Debug("container exited")
-	oc.SetSpanStatus(span, err)
+	oc.SetSpanStatus(span, c.waitError)
 }

internal/hcs/system.go

@@ -275,11 +275,19 @@ func (computeSystem *System) waitBackground() {
 	oc.SetSpanStatus(span, err)
 }
 
+func (computeSystem *System) WaitChannel() <-chan struct{} {
+	return computeSystem.waitBlock
+}
+
+func (computeSystem *System) WaitError() error {
+	return computeSystem.waitError
+}
+
 // Wait synchronously waits for the compute system to shutdown or terminate. If
 // the compute system has already exited returns the previous error (if any).
 func (computeSystem *System) Wait() error {
-	<-computeSystem.waitBlock
-	return computeSystem.waitError
+	<-computeSystem.WaitChannel()
+	return computeSystem.WaitError()
 }
 
 // ExitError returns an error describing the reason the compute system terminated.

internal/jobcontainers/jobcontainer.go

@@ -469,11 +469,19 @@ func (c *JobContainer) Terminate(ctx context.Context) error {
 	return nil
 }
 
+func (c *JobContainer) WaitChannel() <-chan struct{} {
+	return c.waitBlock
+}
+
+func (c *JobContainer) WaitError() error {
+	return c.waitError
+}
+
 // Wait synchronously waits for the container to shutdown or terminate. If
 // the container has already exited returns the previous error (if any).
 func (c *JobContainer) Wait() error {
-	<-c.waitBlock
-	return c.waitError
+	<-c.WaitChannel()
+	return c.WaitError()
 }
 
 func (c *JobContainer) waitBackground(ctx context.Context) {