Skip to content

fix webhook to deny privileged containers #2139

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
hami-robot merged 3 commits into from
Jul 28, 2026
Merged

fix webhook to deny privileged containers #2139

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
9f4313d
fix webhook to deny privileged containers
Jay2006sawant Jul 27, 2026
db52c72
check privileged init containers before scheduler bypass
Jay2006sawant Jul 27, 2026
d9bb61e
deny privileged containers only when gpu resources are requested
Jay2006sawant Jul 27, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
33 changes: 26 additions & 7 deletions pkg/scheduler/webhook.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,7 @@ package scheduler
import (
"context"
"encoding/json"
"fmt"
"net/http"

corev1 "k8s.io/api/core/v1"
Expand Down Expand Up @@ -61,6 +62,10 @@ func (h *webhook) Handle(_ context.Context, req admission.Request) admission.Res
klog.Warningf(template+" - Denying admission as pod has no containers", pod.Namespace, pod.Name, pod.UID)
return admission.Denied("pod has no containers")
}
if name, privileged := privilegedContainerName(pod); privileged {
Comment thread
archlitchi marked this conversation as resolved.
Outdated
klog.Warningf(template+" - Denying admission as container %s is privileged", pod.Namespace, pod.Name, pod.UID, name)
return admission.Denied(fmt.Sprintf("container %s is privileged", name))
}
if pod.Spec.SchedulerName != "" &&
(pod.Spec.SchedulerName != corev1.DefaultSchedulerName || !config.ForceOverwriteDefaultScheduler) &&
(len(config.SchedulerName) == 0 || pod.Spec.SchedulerName != config.SchedulerName) {
Expand All @@ -69,14 +74,8 @@ func (h *webhook) Handle(_ context.Context, req admission.Request) admission.Res
}
klog.V(5).Infof(template, pod.Namespace, pod.Name, pod.UID)
hasResource := false
for idx, ctr := range pod.Spec.Containers {
for idx := range pod.Spec.Containers {
c := &pod.Spec.Containers[idx]
Comment thread
archlitchi marked this conversation as resolved.
if ctr.SecurityContext != nil {
if ctr.SecurityContext.Privileged != nil && *ctr.SecurityContext.Privileged {
klog.Warningf(template+" - Denying admission as container %s is privileged", pod.Namespace, pod.Name, pod.UID, c.Name)
continue
}
}
for _, val := range device.GetDevices() {
found, err := val.MutateAdmission(c, pod)
if err != nil {
Expand Down Expand Up @@ -108,6 +107,26 @@ func (h *webhook) Handle(_ context.Context, req admission.Request) admission.Res
return admission.PatchResponseFromRaw(req.Object.Raw, marshaledPod)
}

func privilegedContainerName(pod *corev1.Pod) (string, bool) {
for _, ctr := range pod.Spec.InitContainers {
if isPrivilegedContainer(&ctr) {
return ctr.Name, true
}
}
for _, ctr := range pod.Spec.Containers {
if isPrivilegedContainer(&ctr) {
return ctr.Name, true
}
}
return "", false
}

func isPrivilegedContainer(ctr *corev1.Container) bool {
return ctr.SecurityContext != nil &&
ctr.SecurityContext.Privileged != nil &&
*ctr.SecurityContext.Privileged
}

func fitResourceQuota(pod *corev1.Pod) bool {
for deviceName, dev := range device.GetDevices() {
// Only supports NVIDIA
Expand Down
154 changes: 154 additions & 0 deletions pkg/scheduler/webhook_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -642,3 +642,157 @@ func TestSchedulerNameEmptyNoOverwrite(t *testing.T) {
t.Fatalf("Expected schedulerName patch to %q, got patches: %+v", config.SchedulerName, resp.Patches)
}
}

func TestPrivilegedContainerDenied(t *testing.T) {
prevSchedulerName := config.SchedulerName
prevDevicesMap := device.DevicesMap
prevDevicesToHandle := device.DevicesToHandle
t.Cleanup(func() {
config.SchedulerName = prevSchedulerName
device.DevicesMap = prevDevicesMap
device.DevicesToHandle = prevDevicesToHandle
})

config.SchedulerName = "hami-scheduler"
sConfig := &config.Config{
NvidiaConfig: nvidia.NvidiaConfig{
ResourceCountName: "hami.io/gpu",
ResourceMemoryName: "hami.io/gpumem",
ResourceMemoryPercentageName: "hami.io/gpumem-percentage",
ResourceCoreName: "hami.io/gpucores",
DefaultMemory: 0,
DefaultCores: 0,
DefaultGPUNum: 1,
},
}
if err := config.InitDevicesWithConfig(sConfig); err != nil {
t.Fatalf("Failed to initialize devices with config: %v", err)
}
Comment thread
coderabbitai[bot] marked this conversation as resolved.

privileged := true
testCases := []struct {
name string
pod *corev1.Pod
}{
{
name: "privileged container only",
pod: &corev1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "privileged-pod", Namespace: "default"},
Spec: corev1.PodSpec{
Containers: []corev1.Container{
{
Name: "privileged",
SecurityContext: &corev1.SecurityContext{
Privileged: &privileged,
},
},
},
},
},
},
{
name: "privileged sidecar with gpu workload",
pod: &corev1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "mixed-pod", Namespace: "default"},
Spec: corev1.PodSpec{
Containers: []corev1.Container{
{
Name: "privileged-sidecar",
SecurityContext: &corev1.SecurityContext{
Privileged: &privileged,
},
},
{
Name: "gpu-workload",
Resources: corev1.ResourceRequirements{
Limits: corev1.ResourceList{
"hami.io/gpu": resource.MustParse("1"),
},
},
},
},
},
},
},
{
name: "privileged init container with gpu workload",
pod: &corev1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "init-privileged-pod", Namespace: "default"},
Spec: corev1.PodSpec{
InitContainers: []corev1.Container{
{
Name: "privileged-init",
SecurityContext: &corev1.SecurityContext{
Privileged: &privileged,
},
},
},
Containers: []corev1.Container{
{
Name: "gpu-workload",
Resources: corev1.ResourceRequirements{
Limits: corev1.ResourceList{
"hami.io/gpu": resource.MustParse("1"),
},
},
},
},
},
},
},
{
name: "privileged pod with different scheduler",
pod: &corev1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "other-scheduler-pod", Namespace: "default"},
Spec: corev1.PodSpec{
SchedulerName: "other-scheduler",
Containers: []corev1.Container{
{
Name: "privileged",
SecurityContext: &corev1.SecurityContext{
Privileged: &privileged,
},
},
},
},
},
},
}

wh, err := NewWebHook()
if err != nil {
t.Fatalf("Error creating WebHook: %v", err)
}

scheme := runtime.NewScheme()
corev1.AddToScheme(scheme)
codec := serializer.NewCodecFactory(scheme).LegacyCodec(corev1.SchemeGroupVersion)

for _, tc := range testCases {
t.Run(tc.name, func(t *testing.T) {
podBytes, err := runtime.Encode(codec, tc.pod)
if err != nil {
t.Fatalf("Error encoding pod: %v", err)
}

req := admission.Request{
AdmissionRequest: admissionv1.AdmissionRequest{
UID: "test-uid",
Namespace: tc.pod.Namespace,
Name: tc.pod.Name,
Object: runtime.RawExtension{
Raw: podBytes,
},
},
}

resp := wh.Handle(context.Background(), req)
if resp.Allowed {
t.Fatalf("Expected denied response for privileged pod, but got allowed with %d patches", len(resp.Patches))
}
if resp.Result == nil || resp.Result.Message == "" {
t.Fatalf("Expected denial message, got: %+v", resp.Result)
}
})
}
}
Loading