CVE-2021-47362

What it actually does

NULL pointer dereference in AMD GPU power management (drm/amd/pm) during DPM initialization when set_power_state references an unpopulated current state before boot state is set, exposed on platforms with ACPI ATCS link speed queries.

Why it matters to the business

AMD GPU systems may suffer kernel crashes or instability during boot or power transitions, causing downtime, lost work, and requiring reboots or driver reloads on affected Linux deployments.

How an attack unfolds

An attacker with local access boots or triggers power state changes on systems with the latent uninitialized state; the bug surfaces after ATCS logic fixes allow the dereference path to execute during link speed queries.

How to fix it

Update to a Linux kernel containing the drm/amd/pm fix that sets the current state to boot state during DPM initialization; apply the change that ensures set_power_state has a valid state before ACPI ATCS transitions.

How to confirm the fix worked

Verify the kernel commit updating intermediate power state for SI is present, confirm no NULL dereference occurs during amdgpu DPM init on affected hardware, and test power state transitions post-upgrade.

Typical attack path

1. Obtain local access to a Linux system with an affected AMD GPU
2. Trigger boot or DPM initialization sequence
3. Invoke set_power_state path that queries current state before initialization completes
4. Cause NULL pointer dereference leading to kernel panic

How to detect it

Search kernel logs, dmesg, and crash dumps for NULL pointer dereference messages referencing drm/amd/pm, set_power_state, or ATCS during amdgpu initialization.

• drm/amd/pm
• set_power_state
• NULL pointer dereference
• amdgpu.*ATCS

What your security team should do

• Prioritize kernel upgrades for AMD GPU hosts within standard patching windows
• Monitor kernel logs for amdgpu-related NULL pointer oops during boot
• Test power management initialization in staging environments before broad rollout