The flaw lies in the AMDGPU JPEG ring of the Linux kernel. When a 64‑bit user fence is written, the ring does not support such fences and therefore rejects the command submission. The denial of the submission prevents JPEG decoding tasks from completing. This behavior can lead to a loss of service for applications relying on GPU‑accelerated JPEG processing, though it does not crash the kernel or allow arbitrary code execution.

All Linux kernel builds that have not yet integrated commit 4d7d774f100efb5089c86a1fb8c5bf47c63fc9ef are affected. This includes every distribution package shipping a kernel version prior to that commit, regardless of the vendor.

The CVSS score of 7.8 indicates high severity, and the EPSS score of less than 1% suggests automated exploitation is unlikely. The description does not confirm a remote or public exploitation path; based on the details, the likely attack vector is a local user or a compromised component that initiates GPU decoding, which could supply the problematic user fence. Because the flaw resides solely in kernel‑space JPEG ring handling, the main risk is a disruptive denial of service rather than privilege escalation or memory corruption. The vulnerability is not listed in the CISA KEV catalog.