Implement support for multiple sizes huge pages

This implementation allows Pod to request multiple hugepage resources
of different size and mount hugepage volumes using storage medium
HugePage-<size>, e.g.

spec:
  containers:
    resources:
      requests:
        hugepages-2Mi: 2Mi
        hugepages-1Gi: 2Gi
    volumeMounts:
      - mountPath: /hugepages-2Mi
        name: hugepage-2mi
      - mountPath: /hugepages-1Gi
        name: hugepage-1gi
    ...
  volumes:
    - name: hugepage-2mi
      emptyDir:
        medium: HugePages-2Mi
    - name: hugepage-1gi
      emptyDir:
        medium: HugePages-1Gi

NOTE: This is an alpha feature.
      Feature gate HugePageStorageMediumSize must be enabled for it to work.

pkg/apis/core/types.go

@@ -583,9 +583,10 @@ type StorageMedium string
 
 // These are the valid value for StorageMedium
 const (
-	StorageMediumDefault   StorageMedium = ""          // use whatever the default is for the node
-	StorageMediumMemory    StorageMedium = "Memory"    // use memory (tmpfs)
-	StorageMediumHugePages StorageMedium = "HugePages" // use hugepages
+	StorageMediumDefault         StorageMedium = ""           // use whatever the default is for the node
+	StorageMediumMemory          StorageMedium = "Memory"     // use memory (tmpfs)
+	StorageMediumHugePages       StorageMedium = "HugePages"  // use hugepages
+	StorageMediumHugePagesPrefix StorageMedium = "HugePages-" // prefix for full medium notation HugePages-<size>
 )
 
 // Protocol defines network protocols supported for things like container ports.

pkg/apis/core/v1/helper/helpers.go

@@ -103,6 +103,27 @@ func HugePageUnitSizeFromByteSize(size int64) (string, error) {
 	return fmt.Sprintf("%d%s", size, hugePageSizeUnitList[idx]), nil
 }
 
+// IsHugePageMedium returns true if the volume medium is in 'HugePages[-size]' format
+func IsHugePageMedium(medium v1.StorageMedium) bool {
+	if medium == v1.StorageMediumHugePages {
+		return true
+	}
+	return strings.HasPrefix(string(medium), string(v1.StorageMediumHugePagesPrefix))
+}
+
+// HugePageSizeFromMedium returns the page size for the specified huge page medium.
+// If the specified input is not a valid huge page medium an error is returned.
+func HugePageSizeFromMedium(medium v1.StorageMedium) (resource.Quantity, error) {
+	if !IsHugePageMedium(medium) {
+		return resource.Quantity{}, fmt.Errorf("medium: %s is not a hugepage medium", medium)
+	}
+	if medium == v1.StorageMediumHugePages {
+		return resource.Quantity{}, fmt.Errorf("medium: %s doesn't have size information", medium)
+	}
+	pageSize := strings.TrimPrefix(string(medium), string(v1.StorageMediumHugePagesPrefix))
+	return resource.ParseQuantity(pageSize)
+}
+
 // IsOvercommitAllowed returns true if the resource is in the default
 // namespace and is not hugepages.
 func IsOvercommitAllowed(name v1.ResourceName) bool {

pkg/apis/core/validation/validation.go

@@ -3083,8 +3083,33 @@ func validateContainersOnlyForPod(containers []core.Container, fldPath *field.Pa
 	return allErrs
 }
 
+// PodValidationOptions contains the different settings for pod validation
+type PodValidationOptions struct {
+	// Allow pod spec to have more than one huge page resource (with different sizes)
+	AllowMultipleHugePageResources bool
+}
+
+// ValidatePodSingleHugePageResources checks if there are multiple huge
+// pages resources in the pod object.
+func ValidatePodSingleHugePageResources(pod *core.Pod, specPath *field.Path) field.ErrorList {
+	allErrs := field.ErrorList{}
+	hugePageResources := sets.NewString()
+	for i := range pod.Spec.Containers {
+		resourceSet := toContainerResourcesSet(&pod.Spec.Containers[i])
+		for resourceStr := range resourceSet {
+			if v1helper.IsHugePageResourceName(v1.ResourceName(resourceStr)) {
+				hugePageResources.Insert(resourceStr)
+			}
+		}
+	}
+	if len(hugePageResources) > 1 {
+		allErrs = append(allErrs, field.Invalid(specPath, hugePageResources.List(), "must use a single hugepage size in a pod spec"))
+	}
+	return allErrs
+}
+
 // ValidatePod tests if required fields in the pod are set.
-func ValidatePod(pod *core.Pod) field.ErrorList {
+func ValidatePod(pod *core.Pod, opts PodValidationOptions) field.ErrorList {
 	fldPath := field.NewPath("metadata")
 	allErrs := ValidateObjectMeta(&pod.ObjectMeta, true, ValidatePodName, fldPath)
 	allErrs = append(allErrs, ValidatePodSpecificAnnotations(pod.ObjectMeta.Annotations, &pod.Spec, fldPath.Child("annotations"))...)
@@ -3111,17 +3136,8 @@ func ValidatePod(pod *core.Pod) field.ErrorList {
 	allErrs = append(allErrs, validateContainersOnlyForPod(pod.Spec.Containers, specPath.Child("containers"))...)
 	allErrs = append(allErrs, validateContainersOnlyForPod(pod.Spec.InitContainers, specPath.Child("initContainers"))...)
 
-	hugePageResources := sets.NewString()
-	for i := range pod.Spec.Containers {
-		resourceSet := toContainerResourcesSet(&pod.Spec.Containers[i])
-		for resourceStr := range resourceSet {
-			if v1helper.IsHugePageResourceName(v1.ResourceName(resourceStr)) {
-				hugePageResources.Insert(resourceStr)
-			}
-		}
-	}
-	if len(hugePageResources) > 1 {
-		allErrs = append(allErrs, field.Invalid(specPath, hugePageResources, "must use a single hugepage size in a pod spec"))
+	if !opts.AllowMultipleHugePageResources {
+		allErrs = append(allErrs, ValidatePodSingleHugePageResources(pod, specPath)...)
 	}
 
 	podIPsField := field.NewPath("status", "podIPs")
@@ -3679,8 +3695,8 @@ func ValidateContainerUpdates(newContainers, oldContainers []core.Container, fld
 }
 
 // ValidatePodCreate validates a pod in the context of its initial create
-func ValidatePodCreate(pod *core.Pod) field.ErrorList {
-	allErrs := ValidatePod(pod)
+func ValidatePodCreate(pod *core.Pod, opts PodValidationOptions) field.ErrorList {
+	allErrs := ValidatePod(pod, opts)
 
 	fldPath := field.NewPath("spec")
 	// EphemeralContainers can only be set on update using the ephemeralcontainers subresource
@@ -3693,12 +3709,16 @@ func ValidatePodCreate(pod *core.Pod) field.ErrorList {
 
 // ValidatePodUpdate tests to see if the update is legal for an end user to make. newPod is updated with fields
 // that cannot be changed.
-func ValidatePodUpdate(newPod, oldPod *core.Pod) field.ErrorList {
+func ValidatePodUpdate(newPod, oldPod *core.Pod, opts PodValidationOptions) field.ErrorList {
 	fldPath := field.NewPath("metadata")
 	allErrs := ValidateObjectMetaUpdate(&newPod.ObjectMeta, &oldPod.ObjectMeta, fldPath)
 	allErrs = append(allErrs, ValidatePodSpecificAnnotationUpdates(newPod, oldPod, fldPath.Child("annotations"))...)
 	specPath := field.NewPath("spec")
 
+	if !opts.AllowMultipleHugePageResources {
+		allErrs = append(allErrs, ValidatePodSingleHugePageResources(newPod, specPath)...)
+	}
+
 	// validate updateable fields:
 	// 1.  spec.containers[*].image
 	// 2.  spec.initContainers[*].image