The vulnerability arises from a race condition in the AMD amdgpu driver that manages the GLOBAL_SYNC_STATUS register. The driver performs non‑atomic read/modify/write operations to enable or clear VUPDATE_NO_LOCK, resulting in a scenario where an IRQ handler can inadvertently clear the IRQ enable bit while another routine simultaneously acknowledges the interrupt. This flaw, indexed by CWE-362, can allow a privileged attacker to cause the driver to lose a vertical blank interrupt, potentially leading to a display timeout or loss of frame synchronization. While the flaw does not directly expose a remote code execution vector, it can degrade system stability and user experience.

All Linux kernels that include the amdgpu display driver before the commit that atomizes these register accesses. This includes any Linux distributions running a kernel that has not been patched with commit 70de0a0216583a53c946155f8c8adedfdca6b4e7 or newer. The issue is specific to AMD GPUs that use the Global Sync registers for vblank handling; non‑AMD GPUs or kernels lacking the affected code are not impacted.

The CVSS score is not provided, and EPSS is not available. The flaw is listed as not being part of the CISA KEV catalog. Exploiting the race requires code execution in the kernel context, which typically limits attackers to local or privileged users rather than remote actors. Because the bug manifests as a timing race and the mitigation directly applies a spinlock, the likelihood of successful exploitation in the wild is considered low, but the impact of a successful exploit—loss of interrupt handling and potential display hang—can be significant for affected systems.