CVE-2026-98178

A flaw was found in the Linux kernel's AMD GPU driver. When compute subsystem initialization fails, uninitialized device structures remain attached to the graphics driver. A local user capable of triggering a GPU reset can cause the system to dereference an unallocated memory pointer, resulting in a kernel crash and Denial of Service (DoS).

A low impact vulnerability was found in the Linux kernel's AMD GPU driver (drm/amdgpu). The amdgpu_amdkfd_clear_kfd_mapping() function assumed that an attached KFD (compute) device always had a fully populated node array. When KFD device initialization failed after GPU probe, for example on Polaris hardware without the required PCIe atomics support, the kfd_dev structure remained attached to the amdgpu device even though its nodes had not been allocated. A local user able to trigger a subsequent GPU reset could cause the kernel to dereference an unallocated pointer (nodes[0]), resulting in a kernel crash and denial of service. Red Hat Product Security rates this flaw as low impact (CVSS 3.1: 5.5, AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H) because exploitation requires local access with privileges and specific affected hardware configurations where KFD initialization fails. All maintained Red Hat Enterprise Linux kernel and kernel-rt update streams are not affected by this flaw because the vulnerable code is not present; only the Red Hat Enterprise Linux 6 Extended Lifecycle Support (rhel-6-els) stream, which is out of support scope, remains affected-status pending for this issue.

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.

Mitigation for this issue is either not available or the currently available options do not meet the Red Hat Product Security criteria comprising ease of use and deployment, applicability to widespread installation base, or stability.