The amdgpu driver does not complete KIQ ring fences when a graphics reset occurs, causing the first submission after a reset to poll indefinitely on a sequence number that is never updated. This leads to an infinite wait that can exhaust system resources and degrade or halt GPU responsiveness, effectively denying access to graphic services. The weakness arises from improper synchronization handling, making the GPU ring state stale across a reset.

Linux kernel implementations that include the amdgpu DRM driver are affected. No specific kernel versions are listed; any distribution that ships the unpatched kernel includes the flaw.

The flaw is local to systems running the affected kernel, as execution of GPU commands is required to trigger the conditions. The EPSS score of less than 1% suggests a low probability of active exploitation, and the vulnerability is not in the CISA KEV catalog. However, once triggered, the impact is a denial of service that can bring the entire GPU subsystem to a halt. Attackers would need the ability to submit graphics work or otherwise induce a reset, which typically requires local user privileges or compromised software.