GPU VulnDB

Database/NVIDIA / GPU stack

Linux kernel nouveau/uvmm: use-after-free reading resource type when BO sits in TTM_PL_SYSTEM

CVSS 7.8CVE-2026-98517NVIDIA / GPU stackcurated

Impact

When a buffer object is left in TTM_PL_SYSTEM, its bo->resource is a plain struct ttm_resource, but the uvmm bind path casts it to struct nouveau_mem and reads mem->mem.type past the end of the allocation. A local user submitting GPU VA bind/unbind jobs gets a slab use-after-free read in a kernel worker, which is an information-disclosure and memory-corruption primitive on the host kernel. Exposure is limited to nodes actually running nouveau with the userspace VMM uAPI; fleets on the proprietary NVIDIA driver are not affected. Found by fuzzing; no exploitation reported in the record.

Who can reach it

Local user able to open the nouveau DRM device and submit VM_BIND jobs - in a container that is any tenant handed a DRI device node. No further authentication required.

What to do

Upgrade to a stable kernel carrying the fix for your series (five stable commits listed) and reboot the node. Where nouveau is unused, keeping the module unloaded or blacklisted removes the exposure without a maintenance window. The record names no distro fixed versions.

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.