The Linux kernel AMD XDNA accelerator driver (amdxdna) improperly handles user command submissions when the command buffer (BO) handle is AMDXDNA_INVALID_BO_HANDLE (0). The function amdxdna_drm_submit_execbuf() passes this invalid handle without a proper check if drv_cmd is NULL, skipping the retrieval of job->cmd_bo and leaving it NULL. This leads to a NULL pointer dereference in the drm_sched worker when aie2_sched_job_run() calls amdxdna_cmd_set_state(job->cmd_bo), ultimately causing a kernel oops. Legitimate internal driver commands use this invalid handle but always set drv_cmd, so the fix is to reject invalid handles on user submissions where drv_cmd is NULL. This vulnerability can be triggered by any process with access to the accel node on systems with a probed AMD NPU.

A local user with access to the AMD XDNA accelerator device node can cause a kernel NULL pointer dereference, resulting in a kernel oops and denial of service. This disrupts normal kernel operation and may require a system reboot. There is no indication of privilege escalation or code execution from the provided data.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until a fix is applied, restrict access to the AMD XDNA accelerator device node to trusted users only to prevent exploitation. Monitor vendor channels for an official patch or update addressing this issue.