GPU VulnDB

Database/Container, Kubernetes & orchestration

Podman: auto-update trusts image labels to pick which systemd services to restart

CVSS 6.5CVE-2026-121885Container, Kubernetes & orchestrationcurated

Impact

Podman auto-update reads metadata labels from the container image to decide which system services to restart. Someone who controls an image the host runs can make auto-update restart arbitrary services with Podman's privileges. With rootful Podman that includes services that shut the host down. The result is denial of service on the node, not code execution. On a GPU node that means killing running jobs and possibly taking the machine offline.

Who can reach it

An attacker who can get a crafted image run on a host where podman auto-update is in use. That could be through a compromised or untrusted registry or image that an operator deploys. The attacker needs no credentials on the host, but someone has to run the image, so user interaction is required.

What to do

Apply the patched podman packages (RHSA-2026:98992 for RHEL; check your distribution for others). Running containers do not need a reboot, but any podman auto-update timers will use the new binary on their next run. Until patched, disable podman-auto-update for containers built from untrusted images, and avoid running auto-update rootful.

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.