The vulnerability in the Linux kernel’s AMD display driver introduces a race condition during the transition of the PHY finite state machine to the PLL_ON state for TMDS on DCN35. Because the transition is not performed atomically, an IOMMU watchdog timeout can occur, leaving the OTG subsystem stuck. This leads to a system hang where the GPU cannot acknowledge memory invalidations, ultimately causing overall kernel instability or service disruption. The weakness is a classic race condition that affects the integrity and availability of the system.

All Linux kernel installations that include the AMD display driver for DCN35 hardware configurations are affected. No specific kernel version or distribution was listed, so any distribution running a kernel that incorporates the unpatched AMD DCN35 display driver is potentially vulnerable.

The exploit requires local kernel privileges to trigger the improper PHY transition, such as disabling eDP output or manipulating the driver’s state. Because the vulnerability is limited to driver state changes and does not expose a remote entry point, the risk is primarily local. The EPSS score is not available, and the issue is not listed in CISA’s KEV catalog, implying no known active exploitation. However, an attacker with local access could force a system hang, causing denial of service. The CVSS score is not provided, but the effect is consistent with a high-severity issue that warrants immediate remediation.