The vulnerability in the Linux kernel's amdkfd driver arises from a race condition between two processes: one executing kfd_process_wq_release (workqueue teardown) and another switching partitions via amdgpu_xcp_pre_partition_switch. The original check only verified if the kfd_processes_table was empty, but entries are deleted asynchronously, causing a race that can lead to accessing freed memory (kfd_node). The patch added an atomic kfd_processes_count per device to track active processes accurately, preventing the race and associated kernel crashes.

Exploitation of this race condition can lead to kernel crashes (divide errors) and potential denial of service due to invalid memory accesses in the AMD GPU kernel driver. The CVSS vector indicates high impact on confidentiality, integrity, and availability, but no known exploits are reported in the wild.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. The vulnerability description references a patch that introduces an atomic process counter to fix the race condition. Users should monitor official Linux kernel updates and apply relevant patches when available to mitigate this issue.