GPU VulnDB

Database/NVIDIA / GPU stack

Linux kernel drm/xe GuC: steered register list sized from geometry DSS mask only, overflowing the buffer

CVSS 7.8CVE-2026-68274NVIDIA / GPU stackcurated

Impact

The allocation for the GuC steered register extarray is sized from the geometry DSS mask alone, while the fill loop iterates the union of geometry and compute DSS masks. On parts that have compute-only DSS bits - which is exactly the datacenter and compute-oriented Xe silicon - the driver writes past the allocation and corrupts neighbouring slab objects, showing up as list_del corruption and SLUB redzone overwrites during drmm_release on device unbind. Slab corruption on a GPU node is not contained to the GPU: it damages unrelated kernel allocations on the same host.

Who can reach it

Not remotely reachable. Triggered by driver initialisation on affected hardware; the observed corruption surfaces on device bind/unbind, which requires root or a driver-management path.

What to do

Apply the stable fix that sizes the array from the union of both DSS masks, then reboot the node. Until then, avoid unbind/rebind cycles of the xe driver on affected hosts.

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.