Filename

Size

Checksum

8335GTB_820.1642.20170405p_update.hpm

67109473

efe7b0029da1fa3ffc5a4506db986a88

OP820.00
For Impact, Severity and other Firmware definitions, Please refer to the below 'Glossary of firmware terms' url: 
https://www14.software.ibm.com/webapp/set2/sas/f/power5cm/home.html#termdefs

 8335GTB_820.1642 / OP820.10

04/28/17

New features and functions for MTM 8335-GTB:

• • Support was added for HTTP(S) proxies when downloading resources during the Petitboot. For example, this support allows the user to set HTTP(S) proxies for use when loading configuration or boot files.

• • Support for expanded time out options for the BMC web gui/KVM sessions. The old time out range (300 - 1800 seconds) has been changed as follows: 1) Increased the maximum timeout from 1800 seconds to 1296000 seconds, i.e., basically increased the maximum timeout to 15 days. 2) Decreased the minimum timeout from 300 seconds to 0. If the timeout value is 0, then it is considered an infinite timeout. KVM will not timeout and that in-turn will also keep the BMC web gui running too.

• • Support for 2200w redundant power supply.

• • Support was added for booting from the PCIe3 2-port 100GB Ethernet ConnectX-4 EN (NIC and RoCE) QSFP28 adapter with feature code #EC3L. The #EC3L feature adapter can be obtained using RPQ #8A2399 for the 8335-GTB system.

• • Support was added to allow booting from a PCIe Non-Volatile Memory express (NVMe) adapter. The adapters affected are Feature code #EC54 and #EC56 NVMe flash adapters with CCIN 58CB and 58CC respectively.

• • Support was added for detecting and logging a SEL for power supply fan faults and turning on the system attention LED.

• • Support was enhanced in the BMC LDAP configuration: 1) Added support for allowing the hostname in any of LDAP fields that currently only allow IP addresses. 2) Added support for BIND names greater than 64 characters.

• • Support was enhanced for the configuration of the System and Audit Log settings: 1) Log settings were enhanced to support "local Logs" and "remote Logs" at the same time. 2) The "Server Address" field was enhanced to have a custom port number in addition to the default 514 port. 3) The Syslog System log was enhanced to support a TCP configuration in addition to the UDP configuration that was already supported.

• • Support was changed to for the BMC SMTP configuration to remove the "machine name" field since a SMTP relay server cannot be configured.

• • Support was enhanced for the BMC SMTP configuration to allow the "Server Address" field to have a symbolic host name with a domain name or an IP address as was supported.

• • Support was enhanced for the BMC DNS configuration to allow host names to have a "." included in the name and to allow the host names to be greater than 15 characters.

System firmware changes that affect all systems

• • A problem was fixed for error handling in complete resets for the PCI Host Bridge (PHB). During a complete reset, there can be a timeout waiting for a pending transaction, resulting in the PHB being marked as broken and the reset is not completed, leaving the adapters in an error state. With the fix, the PHB is fenced and the Linux kernel can retry the complete reset.

• • A problem was fixed for a missing device discovery message and overly verbose output messages during the boot. The output is now less verbose during the boot - only error-level messages are printed during Petitboot bootloader initialization. This means that there will be fewer messages printed as the system boots. Additionally, the Petitboot user interface is started earlier in the boot process. This means that the user will be presented with the user interface sooner, but it may still take time, potentially up to 30 seconds, for the user interface to be populated with boot options as storage and network hardware is being initialized. During this time, Petitboot will show the status message "Info: Waiting for device discovery". When Petitboot device discovery is completed, the following status message will be shown "Info: Connected to pb-discover!".

• • A problem was fixed for Java error messages being displayed when logging into the BMC web gui. The Remote Console Preview window is no longer displayed on the dashboard which was causing all the extra Java error messages.

• • A problem has been fixed for the system MAC address being cleared with zeroes on an AC power cycle.

• • A problem was fixed for a GPU hang where NVLink (#EC4c, #EC4D, or #EC4F) is being used.

• • A problem was fixed for SEL "Progress | Unknown Progress |Asserted" during the firmware boot. This error can occur during any boot and can be ignored.

• • A problem was fixed for "Preserve All Configuration" not working for the BMC web gui HPM update.

• • A problem was fixed for not being able to change the IP address settings via Petitboot. Fixed a timing issue with the BCM5421 Controller. Without this fix, the timing issue would lead to BMC not getting any network even though all the network parameters/configuration are proper.

• • A problem was fixed for PCIe Non-Volatile Memory express (NVMe) adapters not showing up after a power on of the system. The adapters affected are Feature code #EC54 and #EC56 NVMe flash adapters with CCIN 58CB and 58CC respectively. These flash adapters are not supported on the IBM Power System S822LC (8335-GTA).

• • A problem has been fixed for Fault LEDs returning to the off state after an AC cycle when a FRU had failed and was guarded. With the fix, after an AC power cycle and next power on, the Fault LED will turn on again if the FRUs are guarded.

• • A problem was fixed for a missing SEL log for a fault on the second power supply that caused OCC 1 and 2 "Active Device Disabled" and Quick Pwr Drop 1 and 2 "State Asserted" console messages.

• • A problem has been fixed for systems losing performance and going into safe mode because of On-Chip Controller (OCC) timeouts collecting Analog Power Subsystem Sweep (APSS) data. This data is used by OCC to tune the processor frequency. This problem occurs more frequently on systems with large configurations that are running heavy workloads.

• • A problem has been fixed for processor checkstops that occur after a processor has gone into sleep mode. The fix for this problem required nap mode be substituted for sleep mode. While this significantly reduces the power savings that might be achieved on idle systems, it is not expected to make much of a power consumption difference on system running HPC workloads.

• • A problem was fixed for a security issue on the BMC login.

• • A problem was fixed for GPU power faults preventing the IPL of the system. With the fix, the system will IPL with GPUs that have power faults. From the OS, the unavailable GPUs can be identified for repair.

 8335GTB_820.1636 / OP820.01

11/02/16

New features and functions for MTM 8335-GTB:

• • Support for the Red Hat Enterprise Linux 7.3 OS as a OPAL bare-metal install. For more information on the features delivered with RHEL7.3, see the Red Hat information portal: https://access.redhat.com/documentation/en/red-hat-enterprise-linux/.

• • Support was added to allow booting from a PCIe Non-Volatile Memory express (NVMe) adapter. The adapters affected are Feature code #EC54 and #EC56 NVMe flash adapters with CCIN 58CB and 58CC respectively.

• •

  Support for a OPAL raw console to receive output from the PowerPC boot EPAPR (Embedded Power Architecture Platform Requirements) wrapper. This allows decompression failures inside the wrapper caused by data corruption to be reported to the user.
  
  

System firmware changes that affect all systems

• • A problem was fixed for ipmitool "mc reset cold" failing to reset the the BMC service processor. The ipmitool "mc reset cold" will stop working after the user issues power on when system is already on or after user tries to do power off when system is already powered off. The problem circumvention is to run the ipmitool "mc reset warm" command to the BMC which will restart the service processor IPMI process and clear internal flags that are preventing "mc reset cold" from working.

• • A problem was fixed for slow IPMI Serial Over LAN (SOL) console connection to the server on the BMC. The problem was triggered by an incorrect handling of the definition bits for VGA and serial console output.

• • A problem was fixed for the IPMI Serial Over LAN (SOL) console to the Petitboot user interface for the left and right arrow movements. When editing the command line for the kernel, the user could not go to the start of the line and then go forward one character at a time.

• • A problem was fixed for the BMC integrated ethernet adapter BCM5421 connection speed being downgraded to 100 Mb/s instead of running at the expected 1000 Mb/s for the petitboot and the Linux OS. After the fix is applied, an A/C power cycle of the system is needed to activate the fix.

• • A problem was fixed for a kexec-hardboot reboot of the system that caused USB devices to be lost. A system power cycle is needed to recover the USB devices when this error occurs.

• • A problem was fixed for the shutdown of PCI devices that was causing spurious reboots of the system for a power off. The logical PCI devices are now removed during the shutdown.

• • A problem was fixed for failures that happen when multiple Hypervisor Virtual Console (HVC) are active at the same time. On machines with more than one HVC console, any console after the first failed to register an interrupt handler since all consoles shared the same IRQ number.

• • A problem was fixed for fundamental PCI resets at boot time causing the PCI adapters to not be usable in the Linux OS. No errors occur in the skiboot but the adapters are configurable once the OS is reached.

• • A problem was fixed for time-out errors during the power off of PCI slots with " Timeout powering off slot ... FIRENZE-PCI: Wrong state 00000000 on slot" error message during a power off of the system.

• • A problem was fixed for the system remaining in "safe" mode after an On-Chip Controller (OCC) reset. In "safe" mode, the system is running at reduced processor frequencies, affecting system performance. The OCC reset is an error recovery command that can be requested by the BMC or OPAL for certain OCC errors.

• • A problem was fixed for a false kernel error message during the boot when configuring the Nvidia GPUs. This message can be ignored:

  " [ 0.000000] pnv_ioda_reserve_pe: Invalid PE 4 on PHB#4 [ 0.000000] pnv_ioda_reserve_pe: Invalid PE 4 on PHB#5".

 8335GTB_820.1634 / OP820.00

09/26/16

Impact:  New      Severity:  New

New features and functions for MTM 8335-GTB:

GA Level

NOTE: 

This firmware release only supports OPAL firmware. It does not support running the PowerVM hypervisor.

System firmware changes that affect all systems

• • A problem was fixed for looping error processing for some hardware failures where OPAL-PRD becomes unresponsive. The loop has been fixed to prevent repeated error messages and system slow-down. The error message "0xdeadbeefdeadbeef" was added so it is known when the error handling in OPAL-PRD has failed.

• • An error message was changed for a problem where the SLW (Sleep Winkle) timer gets stuck and the firmware falls back to OPAL pollers. The previous error message was "Stuck with odd generation !". The new message to the SOL console is "SLW : timer stuck, falling back to OPAL pollers. You will likely have slower I2C and may have experienced increased jitter." These messages can be safely ignored at this time until a future firmware release resolves the issue of the stuck timer. The error only occurs when running test procedures that stress the hardware.