The Linux kernel DRM amdkfd driver contained an error‑handling flaw where the resume_all routine was invoked immediately after suspend_all, even if queue restoration had failed. If suspend_all left the GPU in an inconsistent state, the subsequent resume_all could trigger an unintended GPU reset or cause the driver to hang, potentially affecting all applications that rely on the GPU.

All Linux kernel releases that compile the amdkfd DRM driver for AMD GPUs are impacted, as the vulnerability was not tied to a specific kernel version in the public data. Any kernel version older than the commit that restores proper resume_all ordering is potentially vulnerable, while kernels containing the upstream fix are no longer affected.

The CVSS score of 5.5 indicates a moderate impact. The EPSS score of less than 1 % shows a very low probability of exploitation. The vulnerability is not listed in CISA’s KEV catalog. Based on the description, it is inferred that the faulty code path could be triggered only by a privileged attacker or one with kernel exploitation capabilities. A successful exploit would allow a local attacker to force a GPU reset or hang the driver, resulting in a denial of service for GPU‑dependent workloads.