CVE-2025-21784

What it actually does

Failure to check firmware load result in psp_init_cap_microcode() leads to subsequent invalid memory access in the AMD GPU PSP driver.

Why it matters to the business

Linux systems with AMD GPUs may experience kernel crashes or memory corruption; organizations running affected kernels face unplanned downtime and must prioritize kernel updates across fleets.

How an attack unfolds

An attacker with local access triggers the AMD PSP firmware initialization path under conditions where firmware loading fails, causing the driver to dereference uninitialized or invalid memory and resulting in kernel-level memory corruption.

How to fix it

Update the Linux kernel to a version containing commit 3f40a7ff39d9f1d283d5aa9b13e2fb16200aff5f or later; rebuild and deploy the updated kernel across all affected systems.

How to confirm the fix worked

Verify the running kernel contains commit 3f40a7ff39d9f1d283d5aa9b13e2fb16200aff5f; inspect dmesg or kernel logs for the added bail-out path in psp_init_cap_microcode during driver initialization.

Typical attack path

1. Obtain local code execution on a Linux system with an AMD GPU
2. Trigger or wait for AMD PSP microcode initialization during driver load or reset
3. Cause firmware load to fail (e.g., via missing or corrupted firmware files)
4. Trigger invalid memory access leading to crash or corruption

How to detect it

Review kernel logs and dmesg for amdgpu PSP initialization failures or invalid memory access oops around driver load; correlate with boot-time firmware load attempts on AMD hardware.

What your security team should do

• Immediately identify all systems running vulnerable kernels and schedule kernel updates
• Prioritize patching on multi-tenant or high-value AMD GPU hosts
• Monitor for kernel oops or panic events related to amdgpu/psp during boot or reset