Linux Kernel GPU Recovery Vulnerability in DRM/MSM Component

A vulnerability in the Linux kernel's DRM/MSM component affects the GPU recovery process. The issue arises because the recovery worker does not always trigger a recovery when the GPU hangs, instead relying on the GPU to automatically sleep and resume with new work. This can lead to the GPU remaining in a hung state, causing timeouts. The vulnerability is present in the stable Linux kernel.

Impact

The vulnerability can cause the GPU to remain in a hung state, leading to timeouts and potential performance issues.

Reproduction

The vulnerability can be reproduced by submitting a single job to a GPU ring. If this job causes the GPU to hang, the recovery worker will not trigger a recovery, allowing the GPU to remain in a hung state and eventually trigger a timeout.

Remediation

The vulnerability has been addressed in a patch available in the Linux kernel stable tree.