The AMDGPU driver returns a virtual memory context pointer that is only safe while a lock is held. After the lock is released, the driver continues to dereference vm->task_info, resulting in a use‑after‑unlock that can corrupt kernel memory and, if exploited, grant local privilege escalation. Based on the description, it is inferred that this condition could lead to local privilege escalation. This flaw exemplifies a race condition (CWE‑825) and concurrent execution issue (CWE‑362).

All Linux kernel installations that contain the AMDGPU DRM module before commit 9d01579f3f868b333acc901815972685989092c7 are vulnerable. This includes every distribution kernel that ships an unpatched AMDGPU driver, regardless of vendor, and affects any systems utilizing AMD GPUs or GPU pass‑through setups.

The CVSS score of 7.8 classifies the flaw as high severity, while the EPSS score of less than 1% indicates a low probability of exploitation in the wild. Based on the description, it is inferred that the vulnerability requires local or compromise‑level access to trigger the race condition, and that local privilege escalation is a potential outcome. The vulnerability is not listed in CISA KEV. Nevertheless, the potential for kernel memory corruption and privilege escalation warrants immediate attention.