CVE-2023-53628

A logic error was found in the Linux kernel amdgpu graphics driver's interrupt handling for GFX version 11 hardware. A local user with access to an AMD GPU can trigger a GPU device reset or suspend operation, causing the driver to attempt cleanup of a retired interrupt handler that no longer exists, which results in a kernel call trace and denial of service through system crash or GPU malfunction.

The issue arises because the gfx_v11_0_cp_ecc_error_irq interrupt handler was retired in GFX11 hardware, but the cleanup path in gfx_v11_0_hw_fini still attempts to disable this non-existent interrupt via amdgpu_irq_put. When GPU recovery or device suspend operations execute on affected hardware, the driver follows the standard teardown sequence and hits this stale reference, producing a call trace. Any workload that triggers GPU reset (such as through debugfs, recovery from GPU hang, or system suspend) will encounter this condition. The impact is limited to availability; there is no evidence of memory corruption or information disclosure, only disruption of GPU functionality.

The CVSS score(s) listed for this vulnerability do not reflect the associated product's status, and are included for informational purposes to better understand the severity of this vulnerability.

To mitigate this issue, prevent the amdgpu module from loading. See https://access.redhat.com/solutions/41278 for instructions on blacklisting kernel modules.