EFS CSI Driver fixes mount option injection in Kubernetes
TL;DR — A mount option injection in the Amazon EFS CSI Driver lets Kubernetes principals who can create PersistentVolumes inject unauthorized mount flags via unsanitized volumeHandle/mounttargetip fields.
What happened
The Amazon EFS CSI Driver is a Kubernetes Container Storage Interface (CSI) driver that mounts Amazon Elastic File System (EFS) volumes into pods.
AWS reported CVE-2026-6437: under certain circumstances, unsanitized values from two user-controlled inputs are passed to the mount command, enabling injection of arbitrary mount options:
- Access Point ID inside the
volumeHandle mounttargetipvolumeAttribute
In both cases, an attacker can append comma-separated values, and the mount utility parses them as separate mount options.
Why it matters: storage drivers run with elevated permissions and sit on a high-trust boundary between Kubernetes control plane intent and node-level filesystem mounts. In multi-tenant clusters, “who can create PVs/StorageClasses” is often broader than intended, making this class of injection a practical platform risk.
Who is impacted
- Kubernetes clusters using the Amazon EFS CSI Driver at versions
<= v3.0.0. - Environments where non-admin users (or compromised workloads/identities) have privileges to create
PersistentVolumes (and potentiallyStorageClassresources).
| Component | Affected versions (per AWS) | Patched version (per AWS) |
|---|---|---|
| Amazon EFS CSI Driver | <= v3.0.0 | v3.0.1 |
What to do now
- Follow vendor remediation guidance and apply the fixed release.
"This issue has been addressed in EFS CSI Driver version v3.0.1"
- Audit RBAC for who can create or modify
PersistentVolumeandStorageClassresources; AWS explicitly recommends tightening these permissions."Restrict PersistentVolume and StorageClass creation to cluster administrators using Kubernetes RBAC"
- Inventory clusters (including forks/derivative builds) to ensure the remediation is carried into internal images/charts and GitOps-managed deployments.
Content is AI-assisted and reviewed by our team, but issues may be missed and best practices evolve rapidly, send corrections to [email protected]. Always consult official documentation and validate key implementation decisions before making design or security choices.
