Linux kernel amdgpu: unvalidated UVD decode message fields overflow the DPB size calculation
Impact
The UVD decode path took frame dimensions and H.265 reference counts straight from the guest-supplied decode message and fed them into the minimum DPB size arithmetic without bounds checks, so an oversized value overflows that calculation and yields a buffer smaller than the decoder will actually use. The kernel CNA split this across two ids covering the same missing-validation flaw in the same code path - CVE-2026-80908 rejects dimensions above 4096, CVE-2026-80909 rejects invalid H.265 reference counts - and the fix and the operator action are identical for both. Reach is limited: it needs a process that already holds an AMD render node and it only matters on ASICs that carry the UVD decode block, which compute-only Instinct fleets generally are not running decode workloads on. The upstream commits describe the overflow and the added rejection; they do not claim a demonstrated escalation, so treat this as driver hardening on shared AMD nodes rather than a break-glass event.
Who can reach it
Local user on the node with access to an AMD DRM render node - any tenant whose container has /dev/dri passed through for AMD GPU work. Host authentication required; no network path.
What to do
Move to a stable kernel carrying both commits and reboot the affected nodes; the validation is in the driver, so a drain-and-reboot per node is the only route and there is no runtime toggle. Nodes that are not using the UVD decode engine can be scheduled into a normal maintenance window rather than an emergency one. No AMD advisory or fixed distro version appears in the record - only the kernel stable commits.
Also covers 1 CVE
The vendor assigned a separate id to each affected code path. They share this advisory, this score and this fix, so they are one entry here.
References
Related entries
- GPU / accelerator firmware (VBIOS, GSP, NVSwitch): GPU-resident firmware sits below the host OS and is not coveredNCVD-0000-012-gpu-accelerator-firmware-vbios-g · GPU / accelerator firmware (VBIOS, GSP, NVSwitch)Unscored
- NVIDIA Multi-Instance GPU (MIG) partitioning: MIG gives each instance its own SM slice, L2 slice, memory slice andNCVD-2020-001-nvidia-multi-instance-gpu-mig-pa · NVIDIA Multi-Instance GPU (MIG) partitioningUnscored
- NVIDIA Multi-Instance GPU (MIG) partitioning: MIG gives each instance its own SM slice, L2 slice, memory slice andNCVD-2020-003-nvidia-multi-instance-gpu-mig-pa · NVIDIA Multi-Instance GPU (MIG) partitioningUnscored
- Integrated GPU graphics data compression (Intel, AMD, Apple, Arm, Qualcomm, NVIDIA): GPUs apply data-dependent losslessNCVD-2023-003-integrated-gpu-graphics-data-com · Integrated GPU graphics data compression (Intel, AMD, Apple, Arm, Qualcomm, NVIDIA)Unscored
- NVIDIA Confidential Computing (H100/H200/B100/B200/GB200) - CC-DevTools operating mode: NVIDIA GPU confidentialNCVD-2023-004-nvidia-confidential-computing-h1 · NVIDIA Confidential Computing (H100/H200/B100/B200/GB200) - CC-DevTools operating modeUnscored
- Integrated GPU graphics data compression (Intel, AMD, Apple, Arm, Qualcomm, NVIDIA): GPUs apply data-dependent losslessNCVD-2023-005-integrated-gpu-graphics-data-com · Integrated GPU graphics data compression (Intel, AMD, Apple, Arm, Qualcomm, NVIDIA)Unscored
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.