Why does the setting of the kernel parameter kernel.unknown_nmi_panic cause an abnormal restart of a Linux instance? -

A Linux ECS instance may restart abnormally when the kernel parameter kernel.unknown_nmi_panic is set to 1. This article explains the cause and provides steps to resolve the issue.

Problem description

After you set the kernel parameter kernel.unknown_nmi_panic on a Linux ECS instance, the instance restarts abnormally. After the system recovers, the kernel log shows the following call stack:

[5912557.130943] Uhhuh. NMI received for unknown reason 20 on CPU 0.

[5912557.131115] Do you have a strange power saving mode enabled?

[5912557.131287] Kernel panic - not syncing: NMI: Not continuing

Cause

A Non-Maskable Interrupt (NMI) is a hardware interrupt that the CPU cannot ignore or defer. The Linux kernel sometimes receives NMIs from unknown sources. These are typically harmless and do not indicate actual hardware failure.

The kernel parameter kernel.unknown_nmi_panic controls how the kernel responds to an unknown NMI. When set to 1, the kernel treats any unknown NMI as a system error, triggers a kernel panic, and restarts the instance. Most Linux distributions set this parameter to 0 by default, which means the kernel ignores unknown NMIs.

Value	Behavior
`1`	The kernel triggers a kernel panic when it receives an unknown NMI. This causes unexpected instance restarts or service interruptions.
`0`	The kernel ignores the unknown NMI. The instance continues running.

Solution

Set kernel.unknown_nmi_panic to 0 to prevent the kernel from panicking on unknown NMIs.

Log on to the ECS instance.

For more information, see Connect to a Linux instance by using a password or key.
Check whether any existing configurations set the kernel.unknown_nmi_panic parameter.
```
grep 'kernel.unknown_nmi_panic' /etc/sysctl.conf /etc/sysctl.d/*.conf
```
- If configurations exist, delete the redundant entries and keep only one entry in /etc/sysctl.conf. Set the value to 0:
```
kernel.unknown_nmi_panic = 0
```
- If no configuration exists, add kernel.unknown_nmi_panic = 0 to the end of /etc/sysctl.conf:
```
sudo sh -c "echo >> /etc/sysctl.conf"
sudo sh -c "echo 'kernel.unknown_nmi_panic = 0' >> /etc/sysctl.conf"
```
Apply the configuration.
```
sudo sysctl -p
```
Note
To change the parameter value without reloading the configuration file, run sysctl -w kernel.unknown_nmi_panic=0 instead.
Verify that the parameter is set correctly.
```
sysctl kernel.unknown_nmi_panic
```
Expected output:
```
kernel.unknown_nmi_panic = 0
```
If the output shows kernel.unknown_nmi_panic = 0, the fix is applied. The instance will no longer restart when it receives an unknown NMI.