NMI is generated while the kdump kernel boots on the system - IBM System x

Resolving The Problem

Source

RETAIN tip: H201743

Symptom

When enabling crash dumps in Red Hat Enterprise Linux 6 (RHEL6) or SUSE Linux Enterprise Server 11 (SLES11), the system will panic with an unknown Non-Maskable Interrupt (NMI) error. If the reset time on the system is delayed, then the kdump kernel cannot finish saving the dump file. The crash dump capture will fail.

Affected configurations

The system may be any of the following IBM servers:

• System x3400 M2, type 7836, any model
• System x3400 M2, type 7837, any model
• System x3400 M3, type 7378, any model
• System x3400 M3, type 7379, any model
• System x3500 M2, type 7839, any model
• System x3500 M3, type 7380, any model
• System x3550 M2, type 4198, any model
• System x3550 M2, type 7946, any model
• System x3550 M3, type 4254, any model
• System x3550 M3, type 7944, any model
• System x3650 M2, type 4199, any model
• System x3650 M2, type 7947, any model
• System x3650 M3, type 4255, any model
• System x3650 M3, type 7945, any model

The system is configured with at least one of the following:

• Red Hat Enterprise Linux 6, any update, any model
• SUSE Linux Enterprise Server 11, any service pack

This tip is not option specific.

Note: This does not imply that the network operating system will work under all combinations of hardware and software.

Solution

This behavior will be corrected in a future release of RHEL and SLES. The target date for this release is scheduled for second quarter 2011.

Workaround

In order to use the kdump utility in RHEL6 or SLES 11, users should disable the "Reboot system on NMI" setting in Unified Extensible Firmware Interface (UEFI). Complete the following steps:

1. Boot in to UEFI t he setup menu by pressing F1 during system boot.
   
2. Select System Settings.
   
3. Select Integrated Management Module.
   
4. Select "disable" on the "Reboot System on NMI" option line.
   
5. Save the settings and reboot the system.
   

Additional information

This is a technical issue on interrupt remapping between Linux and UEFI. The NMI is generated while the kdump kernel boots on the system. This NMI causes a system reset with the default Basic Input/Output System (BIOS) setting (enable reboot on NMI) and consequently leads to a failure of the crash dump capture. This issue is related to interrupt remapping. If interrupt remapping is disabled in the main kernel and the kdump kernel, then the issue will be resolved.