CVE-2025-38038

What it actually does

The amd-pstate cpufreq driver acquires a driver-wide lock in set_boost that can deadlock with the lock taken in status_store during mode switches on per-policy operations.

Why it matters to the business

On affected AMD Linux systems, CPU frequency scaling operations may cause kernel deadlocks leading to system unresponsiveness and operational disruption until reboot.

How an attack unfolds

An attacker with local access triggers concurrent calls to set_boost and status_store to induce the mutex deadlock in the cpufreq subsystem.

How to fix it

Update the Linux kernel to commit 61e931ee145eeab8196e585ff4334870b130b744 or later versions containing the fix that removes the unnecessary driver_lock.

How to confirm the fix worked

Verify the patch is applied by inspecting the amd-pstate source for removal of the driver_lock acquisition in the set_boost function and test mode-switch and boost paths for absence of deadlock.

Typical attack path

1. Obtain local access to an affected AMD Linux system
2. Trigger concurrent cpufreq boost and mode switch operations via user-space tools or scripts

How to detect it

Review kernel logs for deadlock or mutex contention messages involving amd-pstate set_boost and status_store functions; check for repeated cpufreq-related hangs without other causes.

What your security team should do

• Apply the kernel patch immediately on all AMD systems using the amd-pstate driver
• Prioritize updates for systems with high uptime requirements or frequent cpufreq changes
• Monitor kernel logs for mutex-related warnings in cpufreq paths post-update