SATA generates Machine Check Error (MCE) - IBM IntelliStation A Pro (Type 6224)

Resolving The Problem

RETAIN tip: H036254

Red Hat Enterprise Linux (RHEL) 3 Update 2 on SATA generates MCE with over 4GB of RAM.

The system has more than 4GB of RAM,and RHEL 3 for D64/EM64T. Update 2 is installed.
 
The MCE messages will appear on the console during system startup and shutdown, and will accumulate in the system log file at var/log/messages.
 
The system may be any of the following IBM IntelliStations:  

• an IntelliStation A Pro, type 6224, any model

The system is configured with 2 processors. The number of CPU's, type or speed matter do matters.  

The following NOS(es) are affected:  

• Red Hat Enterprise Linux, version 3 for AMD64/EM64T Update 2.

Note: This does not imply that the network operating system will work under all combinations of hardware and software. Please see the compatibility page for more information: http://www.ibm.com/servers/eserver/serverproven/compat/us/

Update the system to RHEL3 Update 3.

The Machine Check Exception (MCE) messages may be avoided by restricting the amount of RAM in the system to no more than 4GB, either literally or by adding the parameter "mem=4095M" to the kernel boot string in the boot loader configuration file at /boot/grub/grub.conf.
 
However, there is no indication that the error represents any threat to system stability or performance.
 
Also, the volume of additional output generated to the system log file is not large. Therefore, the recommendation would be to ignore the messages until the system can be updated to Update 3.

An interaction between the RHEL3 Update 2 kernel and the proprietary SATA driver in use results in the generation of a Machine Check Exception (MCE) message approximately every 30 seconds during system operation. The memory size is a trigger point since different means are used by the kernel to handle memory addressing with a 32-bit device when the available memory size is greater than the 32-bit address space (>4GB).
 
In this case, the generation of the messages was found to have no negative effect on the stability or performance of the system. Changes in the kernel and SATA driver in Update 3 result in the messages no longer being generated.