CVE-2021-47187

What it actually does

Incorrect entry/exit latency and minimum residency values in ARM64 device-tree idle states for MSM8998 cause L2 cache power-collapse timings to be underestimated, leading to premature or unsafe cluster power-down when CPU frequency scaling enables frequent idle transitions.

Why it matters to the business

Systems using the affected SoC with CPU scaling enabled may experience unpredictable spontaneous reboots or hard lockups, reducing device availability and reliability in production or mobile deployments.

How an attack unfolds

An attacker with the ability to influence CPU load or frequency scaling (for example via a local process or workload) can drive the SoC into the misconfigured power-collapse idle states; the incorrect residency values then trigger hardware instability that manifests as a crash without any further attacker-controlled code execution.

How to fix it

Apply the upstream kernel commit that corrects the MSM8998 idle-state latency and residency values in the device tree; rebuild and deploy the updated kernel or device-tree blob on all affected platforms.

How to confirm the fix worked

Boot the patched kernel, enable CPU scaling, exercise idle states under load, and confirm the device remains stable without spontaneous resets; verify the corrected min-residency-us and entry/exit latency numbers appear in the running device tree.

Typical attack path

1. Obtain local execution or workload control on an MSM8998-based device
2. Enable or trigger CPU frequency scaling and idle transitions
3. Force repeated entry into the L2 power-collapse idle state
4. Observe hardware instability resulting in reboot or lockup

How to detect it

Review kernel logs and dmesg for unexplained sudden reboots or lockups that coincide with CPU idle or frequency scaling activity; inspect device-tree contents on running systems for the original incorrect latency and residency values.

What your security team should do

• Prioritize patching kernel/device-tree on any MSM8998 hardware that enables CPU frequency scaling
• Monitor system logs for sudden reboots without preceding kernel oops or panic messages
• Disable CPU scaling as a temporary mitigation only if patching is delayed