CVE-2021-47551

What it actually does

The flaw permits a kernel panic in the AMD KFD driver under SRIOV when an ASIC reset fails after stop_cpsch has already executed; a subsequent reset attempt triggers repeated uninitialization without a matching start_cpsch or resume, crashing the kernel.

Why it matters to the business

Virtualized environments using AMD GPUs may suffer unexpected system crashes and downtime during reset operations; the issue was resolved in kernel updates nearly two years ago and shows negligible active exploitation risk.

How an attack unfolds

An attacker with local access on an affected system forces a failed SRIOV ASIC reset that invokes stop_cpsch without resume, then triggers a second reset; the driver attempts uninitialization again and panics the kernel.

How to fix it

Upgrade the Linux kernel to version 5.10.205-2 or newer on Debian 11 systems; apply the upstream patch that guards against repeated uninitialization after a failed reset.

How to confirm the fix worked

Confirm the running kernel is at least 5.10.205-2, then exercise SRIOV reset sequences and verify the system remains stable without panic.

Typical attack path

1. Obtain local access to a system running the vulnerable kernel with AMD SRIOV GPU configuration
2. Induce a failed ASIC reset that calls stop_cpsch
3. Trigger a subsequent reset to force repeated uninitialization and kernel panic

How to detect it

Inspect kernel panic logs and dmesg output for references to amdkfd, cpsch, or SRIOV reset failures on systems that have not yet been updated.

What your security team should do

• Prioritize kernel upgrade to 5.10.205-2 on all Debian 11 hosts using AMD SRIOV GPUs
• Review and test GPU reset handling in virtualized workloads
• Monitor kernel logs for amdkfd-related panics until patching is complete