Merge pull request #100482 from pohly/generic-ephemeral-volume-checks

generic ephemeral volume checks

pkg/kubelet/kubelet_pods.go

@@ -2018,21 +2018,31 @@ func hasHostNamespace(pod *v1.Pod) bool {
 // hasHostMountPVC returns true if a PVC is referencing a HostPath volume.
 func (kl *Kubelet) hasHostMountPVC(pod *v1.Pod) bool {
 	for _, volume := range pod.Spec.Volumes {
-		if volume.PersistentVolumeClaim != nil {
+		pvcName := ""
-			pvc, err := kl.kubeClient.CoreV1().PersistentVolumeClaims(pod.Namespace).Get(context.TODO(), volume.PersistentVolumeClaim.ClaimName, metav1.GetOptions{})
+		switch {
-			if err != nil {
+		case volume.PersistentVolumeClaim != nil:
-				klog.InfoS("Unable to retrieve pvc", "pvc", klog.KRef(pod.Namespace, volume.PersistentVolumeClaim.ClaimName), "err", err)
+			pvcName = volume.PersistentVolumeClaim.ClaimName
+		case volume.Ephemeral != nil:
+			if !utilfeature.DefaultFeatureGate.Enabled(features.GenericEphemeralVolume) {
 				continue
 			}
-			if pvc != nil {
+			pvcName = pod.Name + "-" + volume.Name
-				referencedVolume, err := kl.kubeClient.CoreV1().PersistentVolumes().Get(context.TODO(), pvc.Spec.VolumeName, metav1.GetOptions{})
+		default:
-				if err != nil {
+			continue
-					klog.InfoS("Unable to retrieve pv", "pvName", pvc.Spec.VolumeName, "err", err)
+		}
-					continue
+		pvc, err := kl.kubeClient.CoreV1().PersistentVolumeClaims(pod.Namespace).Get(context.TODO(), pvcName, metav1.GetOptions{})
-				}
+		if err != nil {
-				if referencedVolume != nil && referencedVolume.Spec.HostPath != nil {
+			klog.InfoS("Unable to retrieve pvc", "pvc", klog.KRef(pod.Namespace, pvcName), "err", err)
-					return true
+			continue
-				}
+		}
+		if pvc != nil {
+			referencedVolume, err := kl.kubeClient.CoreV1().PersistentVolumes().Get(context.TODO(), pvc.Spec.VolumeName, metav1.GetOptions{})
+			if err != nil {
+				klog.InfoS("Unable to retrieve pv", "pvName", pvc.Spec.VolumeName, "err", err)
+				continue
+			}
+			if referencedVolume != nil && referencedVolume.Spec.HostPath != nil {
+				return true
 			}
 		}
 	}

pkg/kubelet/kubelet_pods_test.go

@@ -38,8 +38,10 @@ import (
 	"k8s.io/apimachinery/pkg/runtime"
 	"k8s.io/apimachinery/pkg/types"
 	"k8s.io/apimachinery/pkg/util/diff"
+	utilfeature "k8s.io/apiserver/pkg/util/feature"
 	core "k8s.io/client-go/testing"
 	"k8s.io/client-go/tools/record"
+	featuregatetesting "k8s.io/component-base/featuregate/testing"
 	netutils "k8s.io/utils/net"
 
 	// TODO: remove this import if
@@ -47,6 +49,7 @@ import (
 	// to "v1"?
 
 	_ "k8s.io/kubernetes/pkg/apis/core/install"
+	"k8s.io/kubernetes/pkg/features"
 	kubecontainer "k8s.io/kubernetes/pkg/kubelet/container"
 	containertest "k8s.io/kubernetes/pkg/kubelet/container/testing"
 	"k8s.io/kubernetes/pkg/kubelet/cri/streaming/portforward"
@@ -2886,13 +2889,16 @@ func TestGetPortForward(t *testing.T) {
 }
 
 func TestHasHostMountPVC(t *testing.T) {
-	tests := map[string]struct {
+	type testcase struct {
-		pvError       error
+		pvError          error
-		pvcError      error
+		pvcError         error
-		expected      bool
+		expected         bool
-		podHasPVC     bool
+		podHasPVC        bool
-		pvcIsHostPath bool
+		pvcIsHostPath    bool
-	}{
+		podHasEphemeral  bool
+		ephemeralEnabled bool
+	}
+	tests := map[string]testcase{
 		"no pvc": {podHasPVC: false, expected: false},
 		"error fetching pvc": {
 			podHasPVC: true,
@@ -2909,6 +2915,18 @@ func TestHasHostMountPVC(t *testing.T) {
 			pvcIsHostPath: true,
 			expected:      true,
 		},
+		"enabled ephemeral host path": {
+			podHasEphemeral:  true,
+			pvcIsHostPath:    true,
+			ephemeralEnabled: true,
+			expected:         true,
+		},
+		"disabled ephemeral host path": {
+			podHasEphemeral:  true,
+			pvcIsHostPath:    true,
+			ephemeralEnabled: false,
+			expected:         false,
+		},
 		"non host path pvc": {
 			podHasPVC:     true,
 			pvcIsHostPath: false,
@@ -2916,7 +2934,8 @@ func TestHasHostMountPVC(t *testing.T) {
 		},
 	}
 
-	for k, v := range tests {
+	run := func(t *testing.T, v testcase) {
+		defer featuregatetesting.SetFeatureGateDuringTest(t, utilfeature.DefaultFeatureGate, features.GenericEphemeralVolume, v.ephemeralEnabled)()
 		testKubelet := newTestKubelet(t, false)
 		defer testKubelet.Cleanup()
 		pod := &v1.Pod{
@@ -2935,13 +2954,23 @@ func TestHasHostMountPVC(t *testing.T) {
 					},
 				},
 			}
+		}
 
-			if v.pvcIsHostPath {
+		if v.podHasEphemeral {
-				volumeToReturn.Spec.PersistentVolumeSource = v1.PersistentVolumeSource{
+			pod.Spec.Volumes = []v1.Volume{
-					HostPath: &v1.HostPathVolumeSource{},
+				{
-				}
+					Name: "xyz",
+					VolumeSource: v1.VolumeSource{
+						Ephemeral: &v1.EphemeralVolumeSource{},
+					},
+				},
 			}
+		}
 
+		if (v.podHasPVC || v.podHasEphemeral) && v.pvcIsHostPath {
+			volumeToReturn.Spec.PersistentVolumeSource = v1.PersistentVolumeSource{
+				HostPath: &v1.HostPathVolumeSource{},
+			}
 		}
 
 		testKubelet.fakeKubeClient.AddReactor("get", "persistentvolumeclaims", func(action core.Action) (bool, runtime.Object, error) {
@@ -2957,9 +2986,14 @@ func TestHasHostMountPVC(t *testing.T) {
 
 		actual := testKubelet.kubelet.hasHostMountPVC(pod)
 		if actual != v.expected {
-			t.Errorf("%s expected %t but got %t", k, v.expected, actual)
+			t.Errorf("expected %t but got %t", v.expected, actual)
 		}
+	}
 
+	for k, v := range tests {
+		t.Run(k, func(t *testing.T) {
+			run(t, v)
+		})
 	}
 }
 

pkg/scheduler/framework/plugins/nodevolumelimits/csi.go

@@ -22,6 +22,7 @@ import (
 
 	v1 "k8s.io/api/core/v1"
 	storagev1 "k8s.io/api/storage/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
 	"k8s.io/apimachinery/pkg/util/rand"
 	corelisters "k8s.io/client-go/listers/core/v1"
@@ -30,6 +31,7 @@ import (
 	csitrans "k8s.io/csi-translation-lib"
 	"k8s.io/klog/v2"
 	"k8s.io/kubernetes/pkg/scheduler/framework"
+	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/feature"
 	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/names"
 	volumeutil "k8s.io/kubernetes/pkg/volume/util"
 )
@@ -54,6 +56,8 @@ type CSILimits struct {
 	randomVolumeIDPrefix string
 
 	translator InTreeToCSITranslator
+
+	enableGenericEphemeralVolume bool
 }
 
 var _ framework.FilterPlugin = &CSILimits{}
@@ -96,7 +100,7 @@ func (pl *CSILimits) Filter(ctx context.Context, _ *framework.CycleState, pod *v
 	}
 
 	newVolumes := make(map[string]string)
-	if err := pl.filterAttachableVolumes(csiNode, pod.Spec.Volumes, pod.Namespace, newVolumes); err != nil {
+	if err := pl.filterAttachableVolumes(pod, csiNode, true /* new pod */, newVolumes); err != nil {
 		return framework.AsStatus(err)
 	}
 
@@ -113,7 +117,7 @@ func (pl *CSILimits) Filter(ctx context.Context, _ *framework.CycleState, pod *v
 
 	attachedVolumes := make(map[string]string)
 	for _, existingPod := range nodeInfo.Pods {
-		if err := pl.filterAttachableVolumes(csiNode, existingPod.Pod.Spec.Volumes, existingPod.Pod.Namespace, attachedVolumes); err != nil {
+		if err := pl.filterAttachableVolumes(existingPod.Pod, csiNode, false /* existing pod */, attachedVolumes); err != nil {
 			return framework.AsStatus(err)
 		}
 	}
@@ -144,25 +148,55 @@ func (pl *CSILimits) Filter(ctx context.Context, _ *framework.CycleState, pod *v
 }
 
 func (pl *CSILimits) filterAttachableVolumes(
-	csiNode *storagev1.CSINode, volumes []v1.Volume, namespace string, result map[string]string) error {
+	pod *v1.Pod, csiNode *storagev1.CSINode, newPod bool, result map[string]string) error {
-	for _, vol := range volumes {
+	for _, vol := range pod.Spec.Volumes {
-		// CSI volumes can only be used as persistent volumes
+		// CSI volumes can only be used through a PVC.
-		if vol.PersistentVolumeClaim == nil {
+		pvcName := ""
+		ephemeral := false
+		switch {
+		case vol.PersistentVolumeClaim != nil:
+			pvcName = vol.PersistentVolumeClaim.ClaimName
+		case vol.Ephemeral != nil:
+			if newPod && !pl.enableGenericEphemeralVolume {
+				return fmt.Errorf(
+					"volume %s is a generic ephemeral volume, but that feature is disabled in kube-scheduler",
+					vol.Name,
+				)
+			}
+			// Generic ephemeral inline volumes also use a PVC,
+			// just with a computed name and certain ownership.
+			// That is checked below once the pvc object is
+			// retrieved.
+			pvcName = pod.Name + "-" + vol.Name
+			ephemeral = true
+		default:
 			continue
 		}
-		pvcName := vol.PersistentVolumeClaim.ClaimName
 
 		if pvcName == "" {
 			return fmt.Errorf("PersistentVolumeClaim had no name")
 		}
 
-		pvc, err := pl.pvcLister.PersistentVolumeClaims(namespace).Get(pvcName)
+		pvc, err := pl.pvcLister.PersistentVolumeClaims(pod.Namespace).Get(pvcName)
 
 		if err != nil {
-			klog.V(5).InfoS("Unable to look up PVC info", "PVC", fmt.Sprintf("%s/%s", namespace, pvcName))
+			if newPod {
+				// The PVC is required to proceed with
+				// scheduling of a new pod because it cannot
+				// run without it. Bail out immediately.
+				return fmt.Errorf("looking up PVC %s/%s: %v", pod.Namespace, pvcName, err)
+			}
+			// If the PVC is invalid, we don't count the volume because
+			// there's no guarantee that it belongs to the running predicate.
+			klog.V(5).InfoS("Unable to look up PVC info", "PVC", fmt.Sprintf("%s/%s", pod.Namespace, pvcName))
 			continue
 		}
 
+		// The PVC for an ephemeral volume must be owned by the pod.
+		if ephemeral && !metav1.IsControlledBy(pvc, pod) {
+			return fmt.Errorf("PVC %s/%s is not owned by pod", pod.Namespace, pvcName)
+		}
+
 		driverName, volumeHandle := pl.getCSIDriverInfo(csiNode, pvc)
 		if driverName == "" || volumeHandle == "" {
 			klog.V(5).Info("Could not find a CSI driver name or volume handle, not counting volume")
@@ -276,7 +310,7 @@ func (pl *CSILimits) getCSIDriverInfoFromSC(csiNode *storagev1.CSINode, pvc *v1.
 }
 
 // NewCSI initializes a new plugin and returns it.
-func NewCSI(_ runtime.Object, handle framework.Handle) (framework.Plugin, error) {
+func NewCSI(_ runtime.Object, handle framework.Handle, fts feature.Features) (framework.Plugin, error) {
 	informerFactory := handle.SharedInformerFactory()
 	pvLister := informerFactory.Core().V1().PersistentVolumes().Lister()
 	pvcLister := informerFactory.Core().V1().PersistentVolumeClaims().Lister()
@@ -284,12 +318,13 @@ func NewCSI(_ runtime.Object, handle framework.Handle) (framework.Plugin, error)
 	scLister := informerFactory.Storage().V1().StorageClasses().Lister()
 
 	return &CSILimits{
-		csiNodeLister:        csiNodesLister,
+		csiNodeLister:                csiNodesLister,
-		pvLister:             pvLister,
+		pvLister:                     pvLister,
-		pvcLister:            pvcLister,
+		pvcLister:                    pvcLister,
-		scLister:             scLister,
+		scLister:                     scLister,
-		randomVolumeIDPrefix: rand.String(32),
+		randomVolumeIDPrefix:         rand.String(32),
-		translator:           csitrans.New(),
+		translator:                   csitrans.New(),
+		enableGenericEphemeralVolume: fts.EnableGenericEphemeralVolume,
 	}, nil
 }
 

pkg/scheduler/framework/plugins/nodevolumelimits/csi_test.go

@@ -47,6 +47,10 @@ const (
 	hostpathInTreePluginName = "kubernetes.io/hostpath"
 )
 
+var (
+	scName = "csi-sc"
+)
+
 // getVolumeLimitKey returns a ResourceName by filter type
 func getVolumeLimitKey(filterType string) v1.ResourceName {
 	switch filterType {
@@ -236,14 +240,112 @@ func TestCSILimits(t *testing.T) {
 		},
 	}
 
+	ephemeralVolumePod := &v1.Pod{
+		ObjectMeta: metav1.ObjectMeta{
+			Namespace: "test",
+			Name:      "abc",
+			UID:       "12345",
+		},
+		Spec: v1.PodSpec{
+			Volumes: []v1.Volume{
+				{
+					Name: "xyz",
+					VolumeSource: v1.VolumeSource{
+						Ephemeral: &v1.EphemeralVolumeSource{},
+					},
+				},
+			},
+		},
+	}
+	controller := true
+	ephemeralClaim := &v1.PersistentVolumeClaim{
+		ObjectMeta: metav1.ObjectMeta{
+			Namespace: ephemeralVolumePod.Namespace,
+			Name:      ephemeralVolumePod.Name + "-" + ephemeralVolumePod.Spec.Volumes[0].Name,
+			OwnerReferences: []metav1.OwnerReference{
+				{
+					Kind:       "Pod",
+					Name:       ephemeralVolumePod.Name,
+					UID:        ephemeralVolumePod.UID,
+					Controller: &controller,
+				},
+			},
+		},
+		Spec: v1.PersistentVolumeClaimSpec{
+			StorageClassName: &scName,
+		},
+	}
+	conflictingClaim := ephemeralClaim.DeepCopy()
+	conflictingClaim.OwnerReferences = nil
+
+	ephemeralTwoVolumePod := &v1.Pod{
+		ObjectMeta: metav1.ObjectMeta{
+			Namespace: "test",
+			Name:      "abc",
+			UID:       "12345II",
+		},
+		Spec: v1.PodSpec{
+			Volumes: []v1.Volume{
+				{
+					Name: "x",
+					VolumeSource: v1.VolumeSource{
+						Ephemeral: &v1.EphemeralVolumeSource{},
+					},
+				},
+				{
+					Name: "y",
+					VolumeSource: v1.VolumeSource{
+						Ephemeral: &v1.EphemeralVolumeSource{},
+					},
+				},
+			},
+		},
+	}
+	ephemeralClaimX := &v1.PersistentVolumeClaim{
+		ObjectMeta: metav1.ObjectMeta{
+			Namespace: ephemeralTwoVolumePod.Namespace,
+			Name:      ephemeralTwoVolumePod.Name + "-" + ephemeralTwoVolumePod.Spec.Volumes[0].Name,
+			OwnerReferences: []metav1.OwnerReference{
+				{
+					Kind:       "Pod",
+					Name:       ephemeralTwoVolumePod.Name,
+					UID:        ephemeralTwoVolumePod.UID,
+					Controller: &controller,
+				},
+			},
+		},
+		Spec: v1.PersistentVolumeClaimSpec{
+			StorageClassName: &scName,
+		},
+	}
+	ephemeralClaimY := &v1.PersistentVolumeClaim{
+		ObjectMeta: metav1.ObjectMeta{
+			Namespace: ephemeralTwoVolumePod.Namespace,
+			Name:      ephemeralTwoVolumePod.Name + "-" + ephemeralTwoVolumePod.Spec.Volumes[1].Name,
+			OwnerReferences: []metav1.OwnerReference{
+				{
+					Kind:       "Pod",
+					Name:       ephemeralTwoVolumePod.Name,
+					UID:        ephemeralTwoVolumePod.UID,
+					Controller: &controller,
+				},
+			},
+		},
+		Spec: v1.PersistentVolumeClaimSpec{
+			StorageClassName: &scName,
+		},
+	}
+
 	tests := []struct {
 		newPod           *v1.Pod
 		existingPods     []*v1.Pod
+		extraClaims      []v1.PersistentVolumeClaim
 		filterName       string
 		maxVols          int
 		driverNames      []string
 		test             string
 		migrationEnabled bool
+		ephemeralEnabled bool
 		limitSource      string
 		wantStatus       *framework.Status
 	}{
@@ -443,6 +545,97 @@ func TestCSILimits(t *testing.T) {
 			limitSource:      "csinode",
 			test:             "should not count in-tree and count csi volumes if migration is disabled (when scheduling in-tree volumes)",
 		},
+		// ephemeral volumes
+		{
+			newPod:      ephemeralVolumePod,
+			filterName:  "csi",
+			driverNames: []string{ebsCSIDriverName},
+			test:        "ephemeral volume feature disabled",
+			wantStatus:  framework.NewStatus(framework.Error, "volume xyz is a generic ephemeral volume, but that feature is disabled in kube-scheduler"),
+		},
+		{
+			newPod:           ephemeralVolumePod,
+			filterName:       "csi",
+			ephemeralEnabled: true,
+			driverNames:      []string{ebsCSIDriverName},
+			test:             "ephemeral volume missing",
+			wantStatus:       framework.NewStatus(framework.Error, `looking up PVC test/abc-xyz: persistentvolumeclaim "abc-xyz" not found`),
+		},
+		{
+			newPod:           ephemeralVolumePod,
+			filterName:       "csi",
+			ephemeralEnabled: true,
+			extraClaims:      []v1.PersistentVolumeClaim{*conflictingClaim},
+			driverNames:      []string{ebsCSIDriverName},
+			test:             "ephemeral volume not owned",
+			wantStatus:       framework.NewStatus(framework.Error, "PVC test/abc-xyz is not owned by pod"),
+		},
+		{
+			newPod:           ephemeralVolumePod,
+			filterName:       "csi",
+			ephemeralEnabled: true,
+			extraClaims:      []v1.PersistentVolumeClaim{*ephemeralClaim},
+			driverNames:      []string{ebsCSIDriverName},
+			test:             "ephemeral volume unbound",
+		},
+		{
+			newPod:           ephemeralVolumePod,
+			filterName:       "csi",
+			ephemeralEnabled: true,
+			extraClaims:      []v1.PersistentVolumeClaim{*ephemeralClaim},
+			driverNames:      []string{ebsCSIDriverName},
+			existingPods:     []*v1.Pod{runningPod, csiEBSTwoVolPod},
+			maxVols:          2,
+			limitSource:      "node",
+			test:             "ephemeral doesn't when node volume limit <= pods CSI volume",
+			wantStatus:       framework.NewStatus(framework.Unschedulable, ErrReasonMaxVolumeCountExceeded),
+		},
+		{
+			newPod:           csiEBSOneVolPod,
+			filterName:       "csi",
+			ephemeralEnabled: true,
+			extraClaims:      []v1.PersistentVolumeClaim{*ephemeralClaimX, *ephemeralClaimY},
+			driverNames:      []string{ebsCSIDriverName},
+			existingPods:     []*v1.Pod{runningPod, ephemeralTwoVolumePod},
+			maxVols:          2,
+			limitSource:      "node",
+			test:             "ephemeral doesn't when node volume limit <= pods ephemeral CSI volume",
+			wantStatus:       framework.NewStatus(framework.Unschedulable, ErrReasonMaxVolumeCountExceeded),
+		},
+		{
+			newPod:           csiEBSOneVolPod,
+			filterName:       "csi",
+			ephemeralEnabled: false,
+			extraClaims:      []v1.PersistentVolumeClaim{*ephemeralClaim},
+			driverNames:      []string{ebsCSIDriverName},
+			existingPods:     []*v1.Pod{runningPod, ephemeralVolumePod, csiEBSTwoVolPod},
+			maxVols:          3,
+			limitSource:      "node",
+			test:             "persistent doesn't when node volume limit <= pods ephemeral CSI volume + persistent volume, ephemeral disabled",
+			wantStatus:       framework.NewStatus(framework.Unschedulable, ErrReasonMaxVolumeCountExceeded),
+		},
+		{
+			newPod:           csiEBSOneVolPod,
+			filterName:       "csi",
+			ephemeralEnabled: true,
+			extraClaims:      []v1.PersistentVolumeClaim{*ephemeralClaim},
+			driverNames:      []string{ebsCSIDriverName},
+			existingPods:     []*v1.Pod{runningPod, ephemeralVolumePod, csiEBSTwoVolPod},
+			maxVols:          3,
+			limitSource:      "node",
+			test:             "persistent doesn't when node volume limit <= pods ephemeral CSI volume + persistent volume",
+			wantStatus:       framework.NewStatus(framework.Unschedulable, ErrReasonMaxVolumeCountExceeded),
+		},
+		{
+			newPod:           csiEBSOneVolPod,
+			filterName:       "csi",
+			ephemeralEnabled: true,
+			extraClaims:      []v1.PersistentVolumeClaim{*ephemeralClaim},
+			driverNames:      []string{ebsCSIDriverName},
+			existingPods:     []*v1.Pod{runningPod, ephemeralVolumePod, csiEBSTwoVolPod},
+			maxVols:          4,
+			test:             "persistent okay when node volume limit > pods ephemeral CSI volume + persistent volume",
+		},
 	}
 
 	// running attachable predicate tests with feature gate and limit present on nodes
@@ -457,14 +650,15 @@ func TestCSILimits(t *testing.T) {
 				defer featuregatetesting.SetFeatureGateDuringTest(t, utilfeature.DefaultFeatureGate, features.CSIMigration, false)()
 				defer featuregatetesting.SetFeatureGateDuringTest(t, utilfeature.DefaultFeatureGate, features.CSIMigrationAWS, false)()
 			}
-
 			p := &CSILimits{
 				csiNodeLister:        getFakeCSINodeLister(csiNode),
 				pvLister:             getFakeCSIPVLister(test.filterName, test.driverNames...),
-				pvcLister:            getFakeCSIPVCLister(test.filterName, "csi-sc", test.driverNames...),
+				pvcLister:            append(getFakeCSIPVCLister(test.filterName, scName, test.driverNames...), test.extraClaims...),
-				scLister:             getFakeCSIStorageClassLister("csi-sc", test.driverNames[0]),
+				scLister:             getFakeCSIStorageClassLister(scName, test.driverNames[0]),
 				randomVolumeIDPrefix: rand.String(32),
 				translator:           csitrans.New(),
+
+				enableGenericEphemeralVolume: test.ephemeralEnabled,
 			}
 			gotStatus := p.Filter(context.Background(), nil, test.newPod, node)
 			if !reflect.DeepEqual(gotStatus, test.wantStatus) {

pkg/scheduler/framework/plugins/nodevolumelimits/non_csi.go

@@ -26,6 +26,7 @@ import (
 
 	v1 "k8s.io/api/core/v1"
 	storage "k8s.io/api/storage/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
 	"k8s.io/apimachinery/pkg/util/rand"
 	"k8s.io/apimachinery/pkg/util/sets"
@@ -35,6 +36,7 @@ import (
 	csilibplugins "k8s.io/csi-translation-lib/plugins"
 	"k8s.io/klog/v2"
 	"k8s.io/kubernetes/pkg/scheduler/framework"
+	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/feature"
 	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/names"
 	volumeutil "k8s.io/kubernetes/pkg/volume/util"
 )
@@ -68,36 +70,36 @@ const (
 const AzureDiskName = names.AzureDiskLimits
 
 // NewAzureDisk returns function that initializes a new plugin and returns it.
-func NewAzureDisk(_ runtime.Object, handle framework.Handle) (framework.Plugin, error) {
+func NewAzureDisk(_ runtime.Object, handle framework.Handle, fts feature.Features) (framework.Plugin, error) {
 	informerFactory := handle.SharedInformerFactory()
-	return newNonCSILimitsWithInformerFactory(azureDiskVolumeFilterType, informerFactory), nil
+	return newNonCSILimitsWithInformerFactory(azureDiskVolumeFilterType, informerFactory, fts), nil
 }
 
 // CinderName is the name of the plugin used in the plugin registry and configurations.
 const CinderName = names.CinderLimits
 
 // NewCinder returns function that initializes a new plugin and returns it.
-func NewCinder(_ runtime.Object, handle framework.Handle) (framework.Plugin, error) {
+func NewCinder(_ runtime.Object, handle framework.Handle, fts feature.Features) (framework.Plugin, error) {
 	informerFactory := handle.SharedInformerFactory()
-	return newNonCSILimitsWithInformerFactory(cinderVolumeFilterType, informerFactory), nil
+	return newNonCSILimitsWithInformerFactory(cinderVolumeFilterType, informerFactory, fts), nil
 }
 
 // EBSName is the name of the plugin used in the plugin registry and configurations.
 const EBSName = names.EBSLimits
 
 // NewEBS returns function that initializes a new plugin and returns it.
-func NewEBS(_ runtime.Object, handle framework.Handle) (framework.Plugin, error) {
+func NewEBS(_ runtime.Object, handle framework.Handle, fts feature.Features) (framework.Plugin, error) {
 	informerFactory := handle.SharedInformerFactory()
-	return newNonCSILimitsWithInformerFactory(ebsVolumeFilterType, informerFactory), nil
+	return newNonCSILimitsWithInformerFactory(ebsVolumeFilterType, informerFactory, fts), nil
 }
 
 // GCEPDName is the name of the plugin used in the plugin registry and configurations.
 const GCEPDName = names.GCEPDLimits
 
 // NewGCEPD returns function that initializes a new plugin and returns it.
-func NewGCEPD(_ runtime.Object, handle framework.Handle) (framework.Plugin, error) {
+func NewGCEPD(_ runtime.Object, handle framework.Handle, fts feature.Features) (framework.Plugin, error) {
 	informerFactory := handle.SharedInformerFactory()
-	return newNonCSILimitsWithInformerFactory(gcePDVolumeFilterType, informerFactory), nil
+	return newNonCSILimitsWithInformerFactory(gcePDVolumeFilterType, informerFactory, fts), nil
 }
 
 // nonCSILimits contains information to check the max number of volumes for a plugin.
@@ -115,6 +117,8 @@ type nonCSILimits struct {
 	// It is used to prefix volumeID generated inside the predicate() method to
 	// avoid conflicts with any real volume.
 	randomVolumeIDPrefix string
+
+	enableGenericEphemeralVolume bool
 }
 
 var _ framework.FilterPlugin = &nonCSILimits{}
@@ -124,13 +128,14 @@ var _ framework.EnqueueExtensions = &nonCSILimits{}
 func newNonCSILimitsWithInformerFactory(
 	filterName string,
 	informerFactory informers.SharedInformerFactory,
+	fts feature.Features,
 ) framework.Plugin {
 	pvLister := informerFactory.Core().V1().PersistentVolumes().Lister()
 	pvcLister := informerFactory.Core().V1().PersistentVolumeClaims().Lister()
 	csiNodesLister := informerFactory.Storage().V1().CSINodes().Lister()
 	scLister := informerFactory.Storage().V1().StorageClasses().Lister()
 
-	return newNonCSILimits(filterName, csiNodesLister, scLister, pvLister, pvcLister)
+	return newNonCSILimits(filterName, csiNodesLister, scLister, pvLister, pvcLister, fts)
 }
 
 // newNonCSILimits creates a plugin which evaluates whether a pod can fit based on the
@@ -149,6 +154,7 @@ func newNonCSILimits(
 	scLister storagelisters.StorageClassLister,
 	pvLister corelisters.PersistentVolumeLister,
 	pvcLister corelisters.PersistentVolumeClaimLister,
+	fts feature.Features,
 ) framework.Plugin {
 	var filter VolumeFilter
 	var volumeLimitKey v1.ResourceName
@@ -185,6 +191,8 @@ func newNonCSILimits(
 		pvcLister:            pvcLister,
 		scLister:             scLister,
 		randomVolumeIDPrefix: rand.String(32),
+
+		enableGenericEphemeralVolume: fts.EnableGenericEphemeralVolume,
 	}
 
 	return pl
@@ -213,7 +221,7 @@ func (pl *nonCSILimits) Filter(ctx context.Context, _ *framework.CycleState, pod
 	}
 
 	newVolumes := make(sets.String)
-	if err := pl.filterVolumes(pod.Spec.Volumes, pod.Namespace, newVolumes); err != nil {
+	if err := pl.filterVolumes(pod, true /* new pod */, newVolumes); err != nil {
 		return framework.AsStatus(err)
 	}
 
@@ -246,7 +254,7 @@ func (pl *nonCSILimits) Filter(ctx context.Context, _ *framework.CycleState, pod
 	// count unique volumes
 	existingVolumes := make(sets.String)
 	for _, existingPod := range nodeInfo.Pods {
-		if err := pl.filterVolumes(existingPod.Pod.Spec.Volumes, existingPod.Pod.Namespace, existingVolumes); err != nil {
+		if err := pl.filterVolumes(existingPod.Pod, false /* existing pod */, existingVolumes); err != nil {
 			return framework.AsStatus(err)
 		}
 	}
@@ -270,57 +278,90 @@ func (pl *nonCSILimits) Filter(ctx context.Context, _ *framework.CycleState, pod
 	return nil
 }
 
-func (pl *nonCSILimits) filterVolumes(volumes []v1.Volume, namespace string, filteredVolumes sets.String) error {
+func (pl *nonCSILimits) filterVolumes(pod *v1.Pod, newPod bool, filteredVolumes sets.String) error {
+	volumes := pod.Spec.Volumes
 	for i := range volumes {
 		vol := &volumes[i]
 		if id, ok := pl.filter.FilterVolume(vol); ok {
 			filteredVolumes.Insert(id)
-		} else if vol.PersistentVolumeClaim != nil {
+			continue
-			pvcName := vol.PersistentVolumeClaim.ClaimName
+		}
-			if pvcName == "" {
-				return fmt.Errorf("PersistentVolumeClaim had no name")
-			}
 
-			// Until we know real ID of the volume use namespace/pvcName as substitute
+		pvcName := ""
-			// with a random prefix (calculated and stored inside 'c' during initialization)
+		ephemeral := false
-			// to avoid conflicts with existing volume IDs.
+		switch {
-			pvID := fmt.Sprintf("%s-%s/%s", pl.randomVolumeIDPrefix, namespace, pvcName)
+		case vol.PersistentVolumeClaim != nil:
-
+			pvcName = vol.PersistentVolumeClaim.ClaimName
-			pvc, err := pl.pvcLister.PersistentVolumeClaims(namespace).Get(pvcName)
+		case vol.Ephemeral != nil:
-			if err != nil {
+			if !pl.enableGenericEphemeralVolume {
-				// If the PVC is invalid, we don't count the volume because
+				return fmt.Errorf(
-				// there's no guarantee that it belongs to the running predicate.
+					"volume %s is a generic ephemeral volume, but that feature is disabled in kube-scheduler",
-				klog.V(4).InfoS("Unable to look up PVC info, assuming PVC doesn't match predicate when counting limits", "PVC", fmt.Sprintf("%s/%s", namespace, pvcName), "err", err)
+					vol.Name,
-				continue
+				)
 			}
+			// Generic ephemeral inline volumes also use a PVC,
+			// just with a computed name and certain ownership.
+			// That is checked below once the pvc object is
+			// retrieved.
+			pvcName = pod.Name + "-" + vol.Name
+			ephemeral = true
+		default:
+			continue
+		}
+		if pvcName == "" {
+			return fmt.Errorf("PersistentVolumeClaim had no name")
+		}
 
-			pvName := pvc.Spec.VolumeName
+		// Until we know real ID of the volume use namespace/pvcName as substitute
-			if pvName == "" {
+		// with a random prefix (calculated and stored inside 'c' during initialization)
-				// PVC is not bound. It was either deleted and created again or
+		// to avoid conflicts with existing volume IDs.
-				// it was forcefully unbound by admin. The pod can still use the
+		pvID := fmt.Sprintf("%s-%s/%s", pl.randomVolumeIDPrefix, pod.Namespace, pvcName)
-				// original PV where it was bound to, so we count the volume if
-				// it belongs to the running predicate.
-				if pl.matchProvisioner(pvc) {
-					klog.V(4).InfoS("PVC is not bound, assuming PVC matches predicate when counting limits", "PVC", fmt.Sprintf("%s/%s", namespace, pvcName))
-					filteredVolumes.Insert(pvID)
-				}
-				continue
-			}
 
-			pv, err := pl.pvLister.Get(pvName)
+		pvc, err := pl.pvcLister.PersistentVolumeClaims(pod.Namespace).Get(pvcName)
-			if err != nil {
+		if err != nil {
-				// If the PV is invalid and PVC belongs to the running predicate,
+			if newPod {
-				// log the error and count the PV towards the PV limit.
+				// The PVC is required to proceed with
-				if pl.matchProvisioner(pvc) {
+				// scheduling of a new pod because it cannot
-					klog.V(4).InfoS("Unable to look up PV, assuming PV matches predicate when counting limits", "PV", fmt.Sprintf("%s/%s/%s", namespace, pvcName, pvName), "err", err)
+				// run without it. Bail out immediately.
-					filteredVolumes.Insert(pvID)
+				return fmt.Errorf("looking up PVC %s/%s: %v", pod.Namespace, pvcName, err)
-				}
-				continue
 			}
+			// If the PVC is invalid, we don't count the volume because
+			// there's no guarantee that it belongs to the running predicate.
+			klog.V(4).InfoS("Unable to look up PVC info, assuming PVC doesn't match predicate when counting limits", "PVC", fmt.Sprintf("%s/%s", pod.Namespace, pvcName), "err", err)
+			continue
+		}
 
-			if id, ok := pl.filter.FilterPersistentVolume(pv); ok {
+		// The PVC for an ephemeral volume must be owned by the pod.
-				filteredVolumes.Insert(id)
+		if ephemeral && !metav1.IsControlledBy(pvc, pod) {
+			return fmt.Errorf("PVC %s/%s is not owned by pod", pod.Namespace, pvcName)
+		}
+
+		pvName := pvc.Spec.VolumeName
+		if pvName == "" {
+			// PVC is not bound. It was either deleted and created again or
+			// it was forcefully unbound by admin. The pod can still use the
+			// original PV where it was bound to, so we count the volume if
+			// it belongs to the running predicate.
+			if pl.matchProvisioner(pvc) {
+				klog.V(4).InfoS("PVC is not bound, assuming PVC matches predicate when counting limits", "PVC", fmt.Sprintf("%s/%s", pod.Namespace, pvcName))
+				filteredVolumes.Insert(pvID)
 			}
+			continue
+		}
+
+		pv, err := pl.pvLister.Get(pvName)
+		if err != nil {
+			// If the PV is invalid and PVC belongs to the running predicate,
+			// log the error and count the PV towards the PV limit.
+			if pl.matchProvisioner(pvc) {
+				klog.V(4).InfoS("Unable to look up PV, assuming PV matches predicate when counting limits", "PV", fmt.Sprintf("%s/%s/%s", pod.Namespace, pvcName, pvName), "err", err)
+				filteredVolumes.Insert(pvID)
+			}
+			continue
+		}
+
+		if id, ok := pl.filter.FilterPersistentVolume(pv); ok {
+			filteredVolumes.Insert(id)
 		}
 	}
 

pkg/scheduler/framework/plugins/nodevolumelimits/non_csi_test.go

@@ -28,9 +28,114 @@ import (
 	csilibplugins "k8s.io/csi-translation-lib/plugins"
 	"k8s.io/kubernetes/pkg/scheduler/framework"
 	fakeframework "k8s.io/kubernetes/pkg/scheduler/framework/fake"
+	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/feature"
 	utilpointer "k8s.io/utils/pointer"
 )
 
+func TestEphemeralLimits(t *testing.T) {
+	// We have to specify a valid filter and arbitrarily pick Cinder here.
+	// It doesn't matter for the test cases.
+	filterName := cinderVolumeFilterType
+	driverName := csilibplugins.CinderInTreePluginName
+
+	ephemeralVolumePod := &v1.Pod{
+		ObjectMeta: metav1.ObjectMeta{
+			Namespace: "test",
+			Name:      "abc",
+			UID:       "12345",
+		},
+		Spec: v1.PodSpec{
+			Volumes: []v1.Volume{
+				{
+					Name: "xyz",
+					VolumeSource: v1.VolumeSource{
+						Ephemeral: &v1.EphemeralVolumeSource{},
+					},
+				},
+			},
+		},
+	}
+	controller := true
+	ephemeralClaim := &v1.PersistentVolumeClaim{
+		ObjectMeta: metav1.ObjectMeta{
+			Namespace: ephemeralVolumePod.Namespace,
+			Name:      ephemeralVolumePod.Name + "-" + ephemeralVolumePod.Spec.Volumes[0].Name,
+			OwnerReferences: []metav1.OwnerReference{
+				{
+					Kind:       "Pod",
+					Name:       ephemeralVolumePod.Name,
+					UID:        ephemeralVolumePod.UID,
+					Controller: &controller,
+				},
+			},
+		},
+		Spec: v1.PersistentVolumeClaimSpec{
+			VolumeName:       "missing",
+			StorageClassName: &filterName,
+		},
+	}
+	conflictingClaim := ephemeralClaim.DeepCopy()
+	conflictingClaim.OwnerReferences = nil
+
+	tests := []struct {
+		newPod           *v1.Pod
+		existingPods     []*v1.Pod
+		extraClaims      []v1.PersistentVolumeClaim
+		ephemeralEnabled bool
+		maxVols          int
+		test             string
+		wantStatus       *framework.Status
+	}{
+		{
+			newPod:     ephemeralVolumePod,
+			test:       "volume feature disabled",
+			wantStatus: framework.NewStatus(framework.Error, "volume xyz is a generic ephemeral volume, but that feature is disabled in kube-scheduler"),
+		},
+		{
+			newPod:           ephemeralVolumePod,
+			ephemeralEnabled: true,
+			test:             "volume missing",
+			wantStatus:       framework.NewStatus(framework.Error, `looking up PVC test/abc-xyz: persistentvolumeclaim "abc-xyz" not found`),
+		},
+		{
+			newPod:           ephemeralVolumePod,
+			ephemeralEnabled: true,
+			extraClaims:      []v1.PersistentVolumeClaim{*conflictingClaim},
+			test:             "volume not owned",
+			wantStatus:       framework.NewStatus(framework.Error, "PVC test/abc-xyz is not owned by pod"),
+		},
+		{
+			newPod:           ephemeralVolumePod,
+			ephemeralEnabled: true,
+			extraClaims:      []v1.PersistentVolumeClaim{*ephemeralClaim},
+			maxVols:          1,
+			test:             "volume unbound, allowed",
+		},
+		{
+			newPod:           ephemeralVolumePod,
+			ephemeralEnabled: true,
+			extraClaims:      []v1.PersistentVolumeClaim{*ephemeralClaim},
+			maxVols:          0,
+			test:             "volume unbound, exceeds limit",
+			wantStatus:       framework.NewStatus(framework.Unschedulable, ErrReasonMaxVolumeCountExceeded),
+		},
+	}
+
+	for _, test := range tests {
+		t.Run(test.test, func(t *testing.T) {
+			fts := feature.Features{
+				EnableGenericEphemeralVolume: test.ephemeralEnabled,
+			}
+			node, csiNode := getNodeWithPodAndVolumeLimits("node", test.existingPods, int64(test.maxVols), filterName)
+			p := newNonCSILimits(filterName, getFakeCSINodeLister(csiNode), getFakeCSIStorageClassLister(filterName, driverName), getFakePVLister(filterName), append(getFakePVCLister(filterName), test.extraClaims...), fts).(framework.FilterPlugin)
+			gotStatus := p.Filter(context.Background(), nil, test.newPod, node)
+			if !reflect.DeepEqual(gotStatus, test.wantStatus) {
+				t.Errorf("status does not match: %v, want: %v", gotStatus, test.wantStatus)
+			}
+		})
+	}
+}
+
 func TestAzureDiskLimits(t *testing.T) {
 	oneVolPod := &v1.Pod{
 		Spec: v1.PodSpec{
@@ -360,7 +465,7 @@ func TestAzureDiskLimits(t *testing.T) {
 	for _, test := range tests {
 		t.Run(test.test, func(t *testing.T) {
 			node, csiNode := getNodeWithPodAndVolumeLimits("node", test.existingPods, int64(test.maxVols), test.filterName)
-			p := newNonCSILimits(test.filterName, getFakeCSINodeLister(csiNode), getFakeCSIStorageClassLister(test.filterName, test.driverName), getFakePVLister(test.filterName), getFakePVCLister(test.filterName)).(framework.FilterPlugin)
+			p := newNonCSILimits(test.filterName, getFakeCSINodeLister(csiNode), getFakeCSIStorageClassLister(test.filterName, test.driverName), getFakePVLister(test.filterName), getFakePVCLister(test.filterName), feature.Features{}).(framework.FilterPlugin)
 			gotStatus := p.Filter(context.Background(), nil, test.newPod, node)
 			if !reflect.DeepEqual(gotStatus, test.wantStatus) {
 				t.Errorf("status does not match: %v, want: %v", gotStatus, test.wantStatus)
@@ -427,7 +532,7 @@ func TestCinderLimits(t *testing.T) {
 	for _, test := range tests {
 		t.Run(test.test, func(t *testing.T) {
 			node, csiNode := getNodeWithPodAndVolumeLimits("node", test.existingPods, int64(test.maxVols), test.filterName)
-			p := newNonCSILimits(test.filterName, getFakeCSINodeLister(csiNode), getFakeCSIStorageClassLister(test.filterName, test.driverName), getFakePVLister(test.filterName), getFakePVCLister(test.filterName)).(framework.FilterPlugin)
+			p := newNonCSILimits(test.filterName, getFakeCSINodeLister(csiNode), getFakeCSIStorageClassLister(test.filterName, test.driverName), getFakePVLister(test.filterName), getFakePVCLister(test.filterName), feature.Features{}).(framework.FilterPlugin)
 			gotStatus := p.Filter(context.Background(), nil, test.newPod, node)
 			if !reflect.DeepEqual(gotStatus, test.wantStatus) {
 				t.Errorf("status does not match: %v, want: %v", gotStatus, test.wantStatus)
@@ -834,7 +939,7 @@ func TestEBSLimits(t *testing.T) {
 	for _, test := range tests {
 		t.Run(test.test, func(t *testing.T) {
 			node, csiNode := getNodeWithPodAndVolumeLimits("node", test.existingPods, int64(test.maxVols), test.filterName)
-			p := newNonCSILimits(test.filterName, getFakeCSINodeLister(csiNode), getFakeCSIStorageClassLister(test.filterName, test.driverName), getFakePVLister(test.filterName), getFakePVCLister(test.filterName)).(framework.FilterPlugin)
+			p := newNonCSILimits(test.filterName, getFakeCSINodeLister(csiNode), getFakeCSIStorageClassLister(test.filterName, test.driverName), getFakePVLister(test.filterName), getFakePVCLister(test.filterName), feature.Features{}).(framework.FilterPlugin)
 			gotStatus := p.Filter(context.Background(), nil, test.newPod, node)
 			if !reflect.DeepEqual(gotStatus, test.wantStatus) {
 				t.Errorf("status does not match: %v, want: %v", gotStatus, test.wantStatus)
@@ -1172,7 +1277,7 @@ func TestGCEPDLimits(t *testing.T) {
 	for _, test := range tests {
 		t.Run(test.test, func(t *testing.T) {
 			node, csiNode := getNodeWithPodAndVolumeLimits("node", test.existingPods, int64(test.maxVols), test.filterName)
-			p := newNonCSILimits(test.filterName, getFakeCSINodeLister(csiNode), getFakeCSIStorageClassLister(test.filterName, test.driverName), getFakePVLister(test.filterName), getFakePVCLister(test.filterName)).(framework.FilterPlugin)
+			p := newNonCSILimits(test.filterName, getFakeCSINodeLister(csiNode), getFakeCSIStorageClassLister(test.filterName, test.driverName), getFakePVLister(test.filterName), getFakePVCLister(test.filterName), feature.Features{}).(framework.FilterPlugin)
 			gotStatus := p.Filter(context.Background(), nil, test.newPod, node)
 			if !reflect.DeepEqual(gotStatus, test.wantStatus) {
 				t.Errorf("status does not match: %v, want: %v", gotStatus, test.wantStatus)

pkg/scheduler/framework/plugins/registry.go

@@ -75,11 +75,11 @@ func NewInTreeRegistry() runtime.Registry {
 		volumebinding.Name:                         runtime.FactoryAdapter(fts, volumebinding.New),
 		volumerestrictions.Name:                    runtime.FactoryAdapter(fts, volumerestrictions.New),
 		volumezone.Name:                            volumezone.New,
-		nodevolumelimits.CSIName:                   nodevolumelimits.NewCSI,
+		nodevolumelimits.CSIName:                   runtime.FactoryAdapter(fts, nodevolumelimits.NewCSI),
-		nodevolumelimits.EBSName:                   nodevolumelimits.NewEBS,
+		nodevolumelimits.EBSName:                   runtime.FactoryAdapter(fts, nodevolumelimits.NewEBS),
-		nodevolumelimits.GCEPDName:                 nodevolumelimits.NewGCEPD,
+		nodevolumelimits.GCEPDName:                 runtime.FactoryAdapter(fts, nodevolumelimits.NewGCEPD),
-		nodevolumelimits.AzureDiskName:             nodevolumelimits.NewAzureDisk,
+		nodevolumelimits.AzureDiskName:             runtime.FactoryAdapter(fts, nodevolumelimits.NewAzureDisk),
-		nodevolumelimits.CinderName:                nodevolumelimits.NewCinder,
+		nodevolumelimits.CinderName:                runtime.FactoryAdapter(fts, nodevolumelimits.NewCinder),
 		interpodaffinity.Name:                      runtime.FactoryAdapter(fts, interpodaffinity.New),
 		nodelabel.Name:                             nodelabel.New,
 		serviceaffinity.Name:                       serviceaffinity.New,

test/e2e/storage/csi_mock_volume.go

@@ -197,7 +197,11 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 		framework.ExpectNoError(err, "Failed to register CSIDriver %v", m.config.GetUniqueDriverName())
 	}
 
-	createPod := func(ephemeral bool) (class *storagev1.StorageClass, claim *v1.PersistentVolumeClaim, pod *v1.Pod) {
+	type volumeType string
+	csiEphemeral := volumeType("CSI")
+	genericEphemeral := volumeType("Ephemeral")
+	pvcReference := volumeType("PVC")
+	createPod := func(withVolume volumeType) (class *storagev1.StorageClass, claim *v1.PersistentVolumeClaim, pod *v1.Pod) {
 		ginkgo.By("Creating pod")
 		sc := m.driver.GetDynamicProvisionStorageClass(m.config, "")
 		scTest := testsuites.StorageClassTest{
@@ -213,12 +217,21 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 
 		// The mock driver only works when everything runs on a single node.
 		nodeSelection := m.config.ClientNodeSelection
-		if ephemeral {
+		switch withVolume {
+		case csiEphemeral:
 			pod = startPausePodInline(f.ClientSet, scTest, nodeSelection, f.Namespace.Name)
-			if pod != nil {
+		case genericEphemeral:
-				m.pods = append(m.pods, pod)
+			class, pod = startPausePodGenericEphemeral(f.ClientSet, scTest, nodeSelection, m.tp.scName, f.Namespace.Name)
+			if class != nil {
+				m.sc[class.Name] = class
 			}
-		} else {
+			claim = &v1.PersistentVolumeClaim{
+				ObjectMeta: metav1.ObjectMeta{
+					Name:      pod.Name + "-" + pod.Spec.Volumes[0].Name,
+					Namespace: f.Namespace.Name,
+				},
+			}
+		case pvcReference:
 			class, claim, pod = startPausePod(f.ClientSet, scTest, nodeSelection, m.tp.scName, f.Namespace.Name)
 			if class != nil {
 				m.sc[class.Name] = class
@@ -226,9 +239,9 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 			if claim != nil {
 				m.pvcs = append(m.pvcs, claim)
 			}
-			if pod != nil {
+		}
-				m.pods = append(m.pods, pod)
+		if pod != nil {
-			}
+			m.pods = append(m.pods, pod)
 		}
 		return // result variables set above
 	}
@@ -318,6 +331,7 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 			name                   string
 			disableAttach          bool
 			deployClusterRegistrar bool
+			volumeType             volumeType
 		}{
 			{
 				name:                   "should not require VolumeAttach for drivers without attachment",
@@ -328,6 +342,11 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 				name:                   "should require VolumeAttach for drivers with attachment",
 				deployClusterRegistrar: true,
 			},
+			{
+				name:                   "should require VolumeAttach for ephemermal volume and drivers with attachment",
+				deployClusterRegistrar: true,
+				volumeType:             genericEphemeral,
+			},
 			{
 				name:                   "should preserve attachment policy when no CSIDriver present",
 				deployClusterRegistrar: false,
@@ -340,7 +359,11 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 				init(testParameters{registerDriver: test.deployClusterRegistrar, disableAttach: test.disableAttach})
 				defer cleanup()
 
-				_, claim, pod := createPod(false)
+				volumeType := t.volumeType
+				if volumeType == "" {
+					volumeType = pvcReference
+				}
+				_, claim, pod := createPod(volumeType)
 				if pod == nil {
 					return
 				}
@@ -375,7 +398,7 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 			init(testParameters{registerDriver: false, disableAttach: true})
 			defer cleanup()
 
-			_, claim, pod := createPod(false /* persistent volume, late binding as specified above */)
+			_, claim, pod := createPod(pvcReference) // late binding as specified above
 			if pod == nil {
 				return
 			}
@@ -497,7 +520,11 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 
 				defer cleanup()
 
-				_, _, pod := createPod(test.expectEphemeral)
+				withVolume := pvcReference
+				if test.expectEphemeral {
+					withVolume = csiEphemeral
+				}
+				_, _, pod := createPod(withVolume)
 				if pod == nil {
 					return
 				}
@@ -539,23 +566,73 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 
 			gomega.Expect(csiNodeAttachLimit).To(gomega.BeNumerically("==", 2))
 
-			_, _, pod1 := createPod(false)
+			_, _, pod1 := createPod(pvcReference)
 			gomega.Expect(pod1).NotTo(gomega.BeNil(), "while creating first pod")
 
 			err = e2epod.WaitForPodNameRunningInNamespace(m.cs, pod1.Name, pod1.Namespace)
 			framework.ExpectNoError(err, "Failed to start pod1: %v", err)
 
-			_, _, pod2 := createPod(false)
+			_, _, pod2 := createPod(pvcReference)
 			gomega.Expect(pod2).NotTo(gomega.BeNil(), "while creating second pod")
 
 			err = e2epod.WaitForPodNameRunningInNamespace(m.cs, pod2.Name, pod2.Namespace)
 			framework.ExpectNoError(err, "Failed to start pod2: %v", err)
 
-			_, _, pod3 := createPod(false)
+			_, _, pod3 := createPod(pvcReference)
 			gomega.Expect(pod3).NotTo(gomega.BeNil(), "while creating third pod")
 			err = waitForMaxVolumeCondition(pod3, m.cs)
 			framework.ExpectNoError(err, "while waiting for max volume condition on pod : %+v", pod3)
 		})
+
+		ginkgo.It("should report attach limit for generic ephemeral volume when persistent volume is attached [Slow]", func() {
+			// define volume limit to be 2 for this test
+			var err error
+			init(testParameters{attachLimit: 1})
+			defer cleanup()
+			nodeName := m.config.ClientNodeSelection.Name
+			driverName := m.config.GetUniqueDriverName()
+
+			csiNodeAttachLimit, err := checkCSINodeForLimits(nodeName, driverName, m.cs)
+			framework.ExpectNoError(err, "while checking limits in CSINode: %v", err)
+
+			gomega.Expect(csiNodeAttachLimit).To(gomega.BeNumerically("==", 1))
+
+			_, _, pod1 := createPod(pvcReference)
+			gomega.Expect(pod1).NotTo(gomega.BeNil(), "while creating pod with persistent volume")
+
+			err = e2epod.WaitForPodNameRunningInNamespace(m.cs, pod1.Name, pod1.Namespace)
+			framework.ExpectNoError(err, "Failed to start pod1: %v", err)
+
+			_, _, pod2 := createPod(genericEphemeral)
+			gomega.Expect(pod2).NotTo(gomega.BeNil(), "while creating pod with ephemeral volume")
+			err = waitForMaxVolumeCondition(pod2, m.cs)
+			framework.ExpectNoError(err, "while waiting for max volume condition on pod : %+v", pod2)
+		})
+
+		ginkgo.It("should report attach limit for persistent volume when generic ephemeral volume is attached [Slow]", func() {
+			// define volume limit to be 2 for this test
+			var err error
+			init(testParameters{attachLimit: 1})
+			defer cleanup()
+			nodeName := m.config.ClientNodeSelection.Name
+			driverName := m.config.GetUniqueDriverName()
+
+			csiNodeAttachLimit, err := checkCSINodeForLimits(nodeName, driverName, m.cs)
+			framework.ExpectNoError(err, "while checking limits in CSINode: %v", err)
+
+			gomega.Expect(csiNodeAttachLimit).To(gomega.BeNumerically("==", 1))
+
+			_, _, pod1 := createPod(genericEphemeral)
+			gomega.Expect(pod1).NotTo(gomega.BeNil(), "while creating pod with persistent volume")
+
+			err = e2epod.WaitForPodNameRunningInNamespace(m.cs, pod1.Name, pod1.Namespace)
+			framework.ExpectNoError(err, "Failed to start pod1: %v", err)
+
+			_, _, pod2 := createPod(pvcReference)
+			gomega.Expect(pod2).NotTo(gomega.BeNil(), "while creating pod with ephemeral volume")
+			err = waitForMaxVolumeCondition(pod2, m.cs)
+			framework.ExpectNoError(err, "while waiting for max volume condition on pod : %+v", pod2)
+		})
 	})
 
 	ginkgo.Context("CSI Volume expansion", func() {
@@ -603,7 +680,7 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 				init(tp)
 				defer cleanup()
 
-				sc, pvc, pod := createPod(false)
+				sc, pvc, pod := createPod(pvcReference)
 				gomega.Expect(pod).NotTo(gomega.BeNil(), "while creating pod for resizing")
 
 				framework.ExpectEqual(*sc.AllowVolumeExpansion, true, "failed creating sc with allowed expansion")
@@ -696,7 +773,7 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 
 				defer cleanup()
 
-				sc, pvc, pod := createPod(false)
+				sc, pvc, pod := createPod(pvcReference)
 				gomega.Expect(pod).NotTo(gomega.BeNil(), "while creating pod for resizing")
 
 				framework.ExpectEqual(*sc.AllowVolumeExpansion, true, "failed creating sc with allowed expansion")
@@ -837,7 +914,7 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 				})
 				defer cleanup()
 
-				_, claim, pod := createPod(false)
+				_, claim, pod := createPod(pvcReference)
 				if pod == nil {
 					return
 				}
@@ -975,7 +1052,7 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 				})
 				defer cleanup()
 
-				_, claim, pod := createPod(false)
+				_, claim, pod := createPod(pvcReference)
 				if pod == nil {
 					return
 				}
@@ -1121,7 +1198,7 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 				framework.ExpectNoError(err, "create PVC watch")
 				defer pvcWatch.Stop()
 
-				sc, claim, pod := createPod(false)
+				sc, claim, pod := createPod(pvcReference)
 				gomega.Expect(pod).NotTo(gomega.BeNil(), "while creating pod")
 				bindingMode := storagev1.VolumeBindingImmediate
 				if test.lateBinding {
@@ -1331,7 +1408,7 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 				syncDelay := 5 * time.Second
 				time.Sleep(syncDelay)
 
-				sc, _, pod := createPod(false /* persistent volume, late binding as specified above */)
+				sc, _, pod := createPod(pvcReference) // late binding as specified above
 				framework.ExpectEqual(sc.Name, scName, "pre-selected storage class name not used")
 
 				waitCtx, cancel := context.WithTimeout(context.Background(), f.Timeouts.PodStart)
@@ -1530,7 +1607,7 @@ var _ = utils.SIGDescribe("CSI mock volume", func() {
 
 				defer cleanup()
 
-				_, _, pod := createPod(false)
+				_, _, pod := createPod(pvcReference)
 				if pod == nil {
 					return
 				}
@@ -1993,7 +2070,7 @@ func checkCSINodeForLimits(nodeName string, driverName string, cs clientset.Inte
 	return attachLimit, nil
 }
 
-func createClaim(cs clientset.Interface, t testsuites.StorageClassTest, node e2epod.NodeSelection, scName, ns string) (*storagev1.StorageClass, *v1.PersistentVolumeClaim) {
+func createSC(cs clientset.Interface, t testsuites.StorageClassTest, scName, ns string) *storagev1.StorageClass {
 	class := newStorageClass(t, ns, "")
 	if scName != "" {
 		class.Name = scName
@@ -2005,12 +2082,17 @@ func createClaim(cs clientset.Interface, t testsuites.StorageClassTest, node e2e
 		framework.ExpectNoError(err, "Failed to create class: %v", err)
 	}
 
+	return class
+}
+
+func createClaim(cs clientset.Interface, t testsuites.StorageClassTest, node e2epod.NodeSelection, scName, ns string) (*storagev1.StorageClass, *v1.PersistentVolumeClaim) {
+	class := createSC(cs, t, scName, ns)
 	claim := e2epv.MakePersistentVolumeClaim(e2epv.PersistentVolumeClaimConfig{
 		ClaimSize:        t.ClaimSize,
 		StorageClassName: &(class.Name),
 		VolumeMode:       &t.VolumeMode,
 	}, ns)
-	claim, err = cs.CoreV1().PersistentVolumeClaims(ns).Create(context.TODO(), claim, metav1.CreateOptions{})
+	claim, err := cs.CoreV1().PersistentVolumeClaims(ns).Create(context.TODO(), claim, metav1.CreateOptions{})
 	framework.ExpectNoError(err, "Failed to create claim: %v", err)
 
 	if !t.DelayBinding {
@@ -2046,6 +2128,21 @@ func startPausePodInline(cs clientset.Interface, t testsuites.StorageClassTest,
 	return pod
 }
 
+func startPausePodGenericEphemeral(cs clientset.Interface, t testsuites.StorageClassTest, node e2epod.NodeSelection, scName, ns string) (*storagev1.StorageClass, *v1.Pod) {
+	class := createSC(cs, t, scName, ns)
+	claim := e2epv.MakePersistentVolumeClaim(e2epv.PersistentVolumeClaimConfig{
+		ClaimSize:        t.ClaimSize,
+		StorageClassName: &(class.Name),
+		VolumeMode:       &t.VolumeMode,
+	}, ns)
+	pod, err := startPausePodWithVolumeSource(cs, v1.VolumeSource{
+		Ephemeral: &v1.EphemeralVolumeSource{
+			VolumeClaimTemplate: &v1.PersistentVolumeClaimTemplate{Spec: claim.Spec}},
+	}, node, ns)
+	framework.ExpectNoError(err, "Failed to create pod: %v", err)
+	return class, pod
+}
+
 func startPausePodWithClaim(cs clientset.Interface, pvc *v1.PersistentVolumeClaim, node e2epod.NodeSelection, ns string) (*v1.Pod, error) {
 	return startPausePodWithVolumeSource(cs,
 		v1.VolumeSource{