Linux kernel amdgpu: register BAR mapping leaks on every driver unload or hot-unplug

Impact

amdgpu_pci_remove() calls drm_dev_unplug() before fini_sw(), so drm_dev_enter() is already false by the time the guarded iounmap() runs and the register BAR ioremap mapping is never released. The .remove path covers both PCIe hot-unplug and a plain rmmod, so one mapping leaks per unload. On AMD accelerator nodes where driver reload is part of routine maintenance - reset sequences, driver version swaps, SR-IOV teardown - kernel mappings accumulate until the node is rebooted. This is a resource leak, not memory corruption, and there is no attacker-controlled input: severity is low and the record carries no score or CWE.

Who can reach it

Local root, or anything in the host's control plane that unloads or rebinds the amdgpu module, plus PCIe hot-unplug events. No unprivileged tenant path and nothing remote.

What to do

Pick up a stable kernel containing the fix (commits linked in the record) and reboot the node at your next maintenance window. Until then, rebooting rather than repeatedly reloading amdgpu clears the accumulated mappings. No vendor advisory or fixed distribution version is present in the record.

References

Related entries

• amdgpu nbio_v7_9: NULL dereference in hard IRQ when a RAS interrupt arrives before RAS late_initCVE-2026-98270 · Linux kernel amdgpu nbio_v7_9 (RAS controller interrupt handler)Unscored
• GPU / accelerator firmware (VBIOS, GSP, NVSwitch): GPU-resident firmware sits below the host OS and is not coveredNCVD-0000-012-gpu-accelerator-firmware-vbios-g · GPU / accelerator firmware (VBIOS, GSP, NVSwitch)Unscored
• NVIDIA Multi-Instance GPU (MIG) partitioning: MIG gives each instance its own SM slice, L2 slice, memory slice andNCVD-2020-001-nvidia-multi-instance-gpu-mig-pa · NVIDIA Multi-Instance GPU (MIG) partitioningUnscored
• NVIDIA Multi-Instance GPU (MIG) partitioning: MIG gives each instance its own SM slice, L2 slice, memory slice andNCVD-2020-003-nvidia-multi-instance-gpu-mig-pa · NVIDIA Multi-Instance GPU (MIG) partitioningUnscored
• Integrated GPU graphics data compression (Intel, AMD, Apple, Arm, Qualcomm, NVIDIA): GPUs apply data-dependent losslessNCVD-2023-003-integrated-gpu-graphics-data-com · Integrated GPU graphics data compression (Intel, AMD, Apple, Arm, Qualcomm, NVIDIA)Unscored
• NVIDIA Confidential Computing (H100/H200/B100/B200/GB200) - CC-DevTools operating mode: NVIDIA GPU confidentialNCVD-2023-004-nvidia-confidential-computing-h1 · NVIDIA Confidential Computing (H100/H200/B100/B200/GB200) - CC-DevTools operating modeUnscored