In the Linux kernel the Qualcomm Iris VPU driver used the non‑synchronizing function disable_irq_nosync() during power‑off. Because the IRQ is threaded, the handler could still run after the function returned, allowing code to access VPU registers that had already been powered down. This race can cause undefined hardware behavior, crashes or denial of service. The weakness is a race condition that violates correct shutdown sequencing.

Affected systems are Linux kernel builds that include the Qualcomm iris VPU driver. The change applies to the media subsystem and any kernel version where the iris driver registers a threaded IRQ. No specific kernel releases are listed, so all kernels that ship the legacy iris driver are potentially impacted.

The EPSS score is reported as less than 1 % and the vulnerability is not listed in CISA’s KEV catalog, indicating a low probability of public exploitation at present. The lack of a publicly disclosed CVSS score limits a formal severity assessment, but the code path described involves kernel privilege and hardware access, implying that an attacker with kernel‑level authority could potentially trigger the race condition. The risk of exploitation depends on whether the Iris VPU driver is loaded and active; if the driver is present, the unreconciled threaded IRQ handler could run after the power‑off path, potentially accessing powered‑down registers and causing instability or denial of service.