Red Hat Enterprise Linux (RHEL) 6.1 PCI-Express Adapters may fail configuration on October 2011 GA IBM Power 7 systems

Environment

• Red Hat Enterprise Linux (RHEL) 6.1

IBM Systems Affected

• 8202 E4C
• 8205 E6C
• 8231 E1C
• 8231 E2C
• 8246-L2C
• 9117-MMC
• 9179-MHC

Symptom

• PCI-Express adapters internal to the system unit enclosure  may fail to configure within RHEL 6.1 Linux at adapter configuration or adapter boot time due to PCI-Express platform errors.
• In some cases a PCI-Express adapter internal to the system unit enclosure is the RHEL 6.1 boot source and Linux fails to boot.
• This problem affects RHEL 6.1 Linux in system configurations in which:  
  1. The system is configured with less than 8GB of physical mainstore (RAM) installed in the system and the system is s not HMC/SDMC managed; or  
  2. The system is HMC or SDMC managed and a RHEL 6.1 Linux  partition profile specifies a maximum mainstore (RAM) configuration for that partition of less than 8GB.  

Workaround

• Systems with 8GB or more of physical mainstore (RAM) that ARE NOT partitioned require no workaround;  the problem does not manifest on systems configured in this way.
• For RHEL6.1 users with systems having less than 8GB of mainstore and that are not HMC/SDMC managed:  
  • The user must connect SDMC or HMC and create a partition profile for the RHEL6.1 Linux OS and set the profile maximum memory to 8GB or more.  The partition profile desired memory can be set to less than 8GB provided the profile maximum memory is 8GB or greater.
• For RHEL6.1 users with systems having 8GB or more of mainstore  that are HMC or SDMC managed and ARE partitioned:  
  • Users should set the maximum memory setting in the partition profile for all RHEL 6.1 logical partitions to 8GB or larger.  The desired memory setting in the partition profile can be set to less than 8GB.
• It is not necessary to actually assign 8GB or more of memory to the partition to circumvent this problem. Linux partitions that failed to boot or configure all assigned system unit enclosure PCI-Express IO should be then rebooted with this mainstore to prevent the problem.