Linux kernel drm/xe: user fence reference leaked when chain fence allocation fails
Impact
When dma_fence_chain_alloc() fails inside Xe sync handling, the user fence reference is never released. This is a separate leak from CVE-2026-43395 with its own fix commit: it triggers on allocation failure rather than on a parse error path, so a tenant that can drive the host into allocation pressure makes each failed submission leak a kernel object, compounding the memory shortage. The end state is availability loss on the GPU node rather than any data exposure (vector reports C:N/I:N/A:H).
Who can reach it
Local, authenticated: a user or container with access to an Xe DRM render node that can submit work with sync objects. Requires the allocation to actually fail, so it is realistic only under memory pressure the attacker or the workload mix creates.
What to do
Apply the stable kernel release containing the fix (commit a5d5634cde48 cherry-picked to stable) and reboot each affected node; drain GPU jobs before the reboot. No published workaround other than limiting who can open the Xe render nodes.
References
Related entries
- Linux kernel amdgpu user-mode queues (doorbell submission path) (drm/amdgpu): Missing or insufficient validation ofCVE-2026-43398 · Linux kernel amdgpu user-mode queues (doorbell submission path) (drm/amdgpu)Medium
- Linux kernel amdgpu user-mode queues (doorbell submission path) (drm/amdgpu/userq): A memory or reference-count leakCVE-2026-43399 · Linux kernel amdgpu user-mode queues (doorbell submission path) (drm/amdgpu/userq)Medium
- Linux kernel amdgpu user-mode queues (doorbell submission path) (drm/amdgpu): Missing or insufficient validation ofCVE-2026-43400 · Linux kernel amdgpu user-mode queues (doorbell submission path) (drm/amdgpu)Medium
- Linux kernel amdgpu firmware, ACPI and IP-block initialisation (drm/amdgpu): A memory or reference-count leakCVE-2026-45947 · Linux kernel amdgpu firmware, ACPI and IP-block initialisation (drm/amdgpu)Medium
- Linux kernel amdgpu RAS / GPU reset and recovery path (drm/amdgpu): A memory or reference-count leak in the amdgpu RASCVE-2026-45976 · Linux kernel amdgpu RAS / GPU reset and recovery path (drm/amdgpu)Medium
- Linux kernel amdgpu GEM/VM/command-submission ioctl surface (drm/amdgpu): A correctness defect in the amdgpuCVE-2026-45979 · Linux kernel amdgpu GEM/VM/command-submission ioctl surface (drm/amdgpu)Medium
This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.