When the AMD GPU (AMDGPU) driver attempts to reset the machine execution scheduler (MES) ring, the force‑completion loop skips MES because it has no scheduler. The polling fence mechanism leaves the hardware fence value unsynced with the expected sequence number. When a reset is triggered—either by MES timing out or after a system suspend/resume—the stale sequence causes an endless poll, preventing the GPU from resuming and ultimately forcing another reset. This results in a system hang, as evidenced by "MES ring buffer is full" and resume failure messages. The flaw causes a denial of service in that normal operation is interrupted until the machine is rebooted.

All Linux kernel variants that include the AMGPU driver and have not yet applied the kernel patch that forces completion of MES ring fences. The specific affected kernel releases are not enumerated within the data provided, so any kernel prior to the inclusion of the fix is potentially vulnerable.

The vulnerability is a denial of service affecting GPU operation and system stability when a reset is triggered due to MES. The EPSS score of <1% indicates a very low likelihood of exploitation presently, and the vulnerability is not in CISA’s KEV catalog. An attacker would need to trigger a GPU reset or a system suspend/resume path, implying local or privileged access. The practical risk is that on affected kernels, an unpatched system could become unresponsive requiring reboot.