The vulnerability stems from a Linux kernel commit that removed a schedule toggle when the GPU execution queue becomes idle during system suspend. This optimization bypassed the GuC suspend, preventing the context switch that would normally flush TLB entries for invalidated userptr virtual memory areas. In LR/preempt‑fence VM mode, the missing flush can lead to stale TLB entries and trigger page faults when userptr pages are invalidated, resulting in a kernel fault that may crash the system, effectively creating a denial‑of‑service condition.

The affected product is the Linux kernel. No specific version ranges are provided by the CNA; therefore any kernel release that contains the commit 8533051ce92015e9cc6f75e0d52119b9d91610b6, its revert, or the associated keyword ‘drm/xe: Skip exec queue schedule toggle’ is relevant. Administrators should verify relevant commit in their running kernel and check for the identical keyword in the changelog. Systems that enable LR/preempt‑fence VM mode are at higher risk because the flaw is exposed only in that execution mode.

The CVSS score is not available, and the EPSS score is unknown, so official quantification of likelihood is unavailable. The flaw is not listed in the CISA KEV catalog, suggesting no known exploitation campaigns yet. However, the defect can lead to kernel panics when userptr pages are invalidated, presenting a local or elevated‑privilege denial‑of‑service vector. A remote attacker would need to gain sufficient privileges to trigger the fault conditions, making the attack less likely in a purely remote scenario. The primary concern remains availability, as an affected kernel can crash and disrupt system operations.