Spurious machine check errors occur on some IBM systems with specific Intel CPUs

SUSE Linux Enterprise Server 11 Service Pack 1

Spurious machine check errors can occur on IBM server that use the Intel Xeon E7, Family 6 Model 47 (Westmere EX) Processors:

Bladecenter HX5, machine type 7873
System x3850 X5, machine types 7143 and 7191
System x3950 X5, machine types 7143 and 7191

Note that these are errors on the hardware that are correctable by the hardware, so although the errors are reported by SLES, they are not causing any problem. 

When the error occurs the message "Machine errors logged" is displayed on the console and saved in /var/log/messages. A record of the spurious machine check error is recorded in /var/log/mcelog.

Here is an example record:
Hardware event. This is not a software error.
CPU 0 BANK 6
TIME 1305137281 Wed May 11 14:08:01 2011
MCG status:
MCi status:
Machine check not valid
Corrected error
MCA: No Error
STATUS 0 MCGSTATUS 0
TIME 1305137281 Wed May 11 14:08:01 2011
MCG status:
MCi status:
MCi_MISC register valid
MCA: BUS Level-3 Generic Generic Other-transaction Request-timeout Error
STATUS 8800004020000e0f MCGSTATUS 0
MCGCAP 1000c18 APICID 0 SOCKETID 0

Some fields may vary, but the spurious errors are always reported on BANK 6, and the STATUS is 20000e0f in the lower 32 bits.

Upcoming UEFI updates for these systems plan to include a solution for these spurious machine check errors. 

Currently the problem can be worked around by disabling the C3/C6 states in the system UEFI settings.