Merge pull request kubernetes#116739 from moshe010/clone-cdi-devices

kubelet dra: lock before getting claimInfo CDIDevices and annotations fields
oppegard · Jul 7, 2023 · 7581ae8 · 7581ae8
2 parents ddb2013 + 04ad946
commit 7581ae8
Show file tree

Hide file tree

Showing 2 changed files with 6 additions and 2 deletions.
diff --git a/pkg/kubelet/cm/container_manager_linux.go b/pkg/kubelet/cm/container_manager_linux.go
@@ -978,6 +978,7 @@ func (cm *containerManagerImpl) GetDynamicResources(pod *v1.Pod, container *v1.C
 	}
 	for _, containerClaimInfo := range containerClaimInfos {
 		var claimResources []*podresourcesapi.ClaimResource
+		containerClaimInfo.RLock()
 		// TODO: Currently  we maintain a list of ClaimResources, each of which contains
 		// a set of CDIDevices from a different kubelet plugin. In the future we may want to
 		// include the name of the kubelet plugin and/or other types of resources that are
@@ -989,6 +990,7 @@ func (cm *containerManagerImpl) GetDynamicResources(pod *v1.Pod, container *v1.C
 			}
 			claimResources = append(claimResources, &podresourcesapi.ClaimResource{CDIDevices: cdiDevices})
 		}
+		containerClaimInfo.RUnlock()
 		containerDynamicResource := podresourcesapi.DynamicResource{
 			ClassName:      containerClaimInfo.ClassName,
 			ClaimName:      containerClaimInfo.ClaimName,

diff --git a/pkg/kubelet/cm/dra/manager.go b/pkg/kubelet/cm/dra/manager.go
@@ -233,13 +233,15 @@ func (m *ManagerImpl) GetResources(pod *v1.Pod, container *v1.Container) (*Conta
 				return nil, fmt.Errorf("unable to get resource for namespace: %s, claim: %s", pod.Namespace, claimName)
 			}
 
+			claimInfo.RLock()
 			klog.V(3).InfoS("Add resource annotations", "claim", claimName, "annotations", claimInfo.annotations)
 			annotations = append(annotations, claimInfo.annotations...)
 			for _, devices := range claimInfo.CDIDevices {
 				for _, device := range devices {
 					cdiDevices = append(cdiDevices, kubecontainer.CDIDevice{Name: device})
 				}
 			}
+			claimInfo.RUnlock()
 		}
 	}
 
@@ -313,8 +315,8 @@ func (m *ManagerImpl) UnprepareResources(pod *v1.Pod) error {
 				resourceHandle.Data)
 			if err != nil {
 				return fmt.Errorf(
-					"NodeUnprepareResource failed, pod: %s, claim UID: %s, claim name: %s, CDI devices: %s, err: %+v",
-					pod.Name, claimInfo.ClaimUID, claimInfo.ClaimName, claimInfo.CDIDevices, err)
+					"NodeUnprepareResource failed, pod: %s, claim UID: %s, claim name: %s, resource handle: %s, err: %+v",
+					pod.Name, claimInfo.ClaimUID, claimInfo.ClaimName, resourceHandle.Data, err)
 			}
 			klog.V(3).InfoS("NodeUnprepareResource succeeded", "response", response)
 		}