GPU VulnDB

Database/Container, Kubernetes & orchestration

Portworx Operator: initial storage cluster provisioning can grant a namespaced user cluster-wide privileges

CVSS 7.7CVE-2026-15140Container, Kubernetes & orchestrationcurated

Impact

During the initial provisioning of a Portworx storage cluster on OpenShift, under conditions the vendor describes only as specific, the operator can grant broader RBAC than intended to a user who holds nothing more than namespace-scoped permissions. The result is elevated privileges inside the Kubernetes cluster - the tier that decides which pods land on which GPU nodes and which service accounts can read cluster secrets. On a shared GPU cluster that is a tenancy break: a tenant confined to their own namespace reaches control-plane objects belonging to everyone else. The vendor bulletin does not detail which conditions trigger it, so operators cannot easily tell from outside whether a given cluster was provisioned into the vulnerable state.

Who can reach it

An authenticated Kubernetes user with only namespace-scoped permissions in a cluster where a Portworx storage cluster is being provisioned for the first time. Not reachable anonymously, and the vendor scopes it to initial provisioning rather than steady-state operation.

What to do

Upgrade the Portworx Operator per the vendor security bulletin - the record does not name a fixed version. The upgrade replaces the operator deployment, so it is a control-plane pod restart, not a node drain: GPU workloads keep running. Because the excess grants persist once created, also audit ClusterRoleBindings and RoleBindings created around the time the storage cluster was first provisioned and remove any that exceed what the operator should have granted.

References

Related entries

All Container, Kubernetes & orchestration entries

This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.