GPU VulnDB

Database/NVIDIA / GPU stack

Linux kernel drm/xe VF: NULL dereference detaching CCS buffers after the device is unbound

CVSS 7.8CVE-2026-68305NVIDIA / GPU stackcurated

Impact

In the Xe SR-IOV virtual-function path, CCS read/write buffers are freed during BO destruction, which can run after the device has been unbound while the DRM structure is still valid. xe_sriov_vf_ccs_rw_update_bb_addr() then dereferences device resources that are gone, producing a kernel NULL pointer oops - the reported trace comes from a process simply exiting with a mapped BO. On a virtualised GPU host this is a crash of the guest or host kernel at teardown, not a data-disclosure bug; the operator cost is an unplanned node outage rather than a tenancy break.

Who can reach it

Local, on a node running the Xe driver in SR-IOV VF mode. Requires the device-unbind window, so in practice an administrative action rather than something an unprivileged tenant drives at will.

What to do

Apply the stable fix adding drm_dev guards and reboot the node. Until patched, avoid unbinding or hot-removing the Xe device while GPU workloads hold mapped buffers.

References

Related entries

All NVIDIA / GPU stack 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.