CVE-2021-47410

What it actually does

Redundant devm_memunmap_pages and devm_release_mem_region calls in svm_migrate_fini trigger a WARNING in devm_release_action during amdgpu driver detach and device release.

Why it matters to the business

Kernel log spam and potential monitoring alerts on AMD GPU systems during normal driver unload or shutdown; low operational risk but noisy in production logging.

How an attack unfolds

No attacker action required; the warning occurs automatically when the amdgpu kernel module is unloaded or the device is detached, as part of normal kgd2kfd_device_exit and amdgpu_device_fini_sw flow.

How to fix it

Update to a Linux kernel containing the fix that removes the svm_migrate_fini function; apply the patch referenced in the upstream commit resolving the redundant devm calls.

How to confirm the fix worked

Confirm absence of the devm_release_action WARNING trace on amdgpu module unload after applying the fix; verify svm_migrate_fini symbol is no longer present in the amdgpu module.

Typical attack path

1. Trigger amdgpu driver unload or PCI device removal
2. kgd2kfd_device_exit calls svm_migrate_fini
3. Redundant devres release executes and emits WARNING

How to detect it

Search kernel logs for the exact WARNING string containing devm_release_action and svm_migrate_fini during amdgpu module removal; correlate with kgd2kfd_device_exit and amdgpu_device_fini_sw call traces.

• WARNING: CPU:.*PID:.*at drivers/base/devres.c:795 devm_release_action
• svm_migrate_fini+0x2d/0x60 [amdgpu]
• kgd2kfd_device_exit+0x23/0xa0 [amdgpu]

What your security team should do

• Apply the kernel fix that removes svm_migrate_fini
• Monitor dmesg for the specific WARNING during driver detach on affected AMD systems
• Prioritize updates on GPU-heavy workloads where frequent module reloads occur