Filename

Size

Checksum

8335_810.1603.20160814a_update.hpm

67109473

e08818a8ab92773a98592a735443719d

OP810.00
For Impact, Severity and other Firmware definitions, Please refer to the below 'Glossary of firmware terms' url: 
http://www14.software.ibm.com/webapp/set2/sas/f/power5cm/home.html#termdefs

8335_810.1603 / OP810.33

09/09/16

Impact:  New function      Severity:  NA

New features and functions

• • Support for the CAPI NVMe (Non-Volatible Memory express) Flash Accelerator Adapter with feature code #EJ1K. This feature provides a PCIe Gen3 adapter with an FPGA and 1.92 TB of low write latency, nonvolatile flash memory. The adapter physically is a half length x8 adapter, but requires a x16 PCIe CAPI-capable Gen3 slot in the system unit. The system connects to the FPGA using the CAPI interface. The FPGA connects to the flash memory using NVMe, which is a high performance software interface to read/write this flash memory. Use of the #EJ1K adapter requires one #EC2A CAPI activation feature per system. This CAPI Flash Accelerator Adapter does not run under PowerKVM but is a bare-metal install only for the following minimum Little Endian (LE) Linux distribution level: Ubuntu 16.04.1.

  The 1.92 TB of flash memory consists of two 960 GB M.2 NVMe modules mounted on the card. These modules are not removable. Each of the modules is treated as a separate LUN.

  The adapter card is designed for read-intensive workloads with light write activity. Approximately 1,370 TB of data can be written over the life of each of the 960 GB M.2 NVMe modules. There is a "fuel gauge" function provided as part of Linux utilities enabling users to view remaining write capacity on each of the modules. If the maximum write capability is achieved, the adapter's replacement or any NVMe M.2 module's replacement is not covered under IBM warranty or maintenance.

  This feature only pertains to the IBM Power System S822LC (8335-GCA) model.

  System firmware changes that affect all systems
  
  

  • • A problem was fixed for a system unrecoverable error that could occur if a CAPI adapter has an error exception signal and hangs when processing it. This is a rare problem that requires multiple error exceptions to the adapter in a short period of time, causing a deadlock in the adapter.

  • • A problem was fixed for CAPI adapter errors that caused a system processor to be called out and guarded instead of the CAPI adapter unit. The errors that cause this problem are the rare fatal adapter errors, so the problem should be infrequently seen. With the fix, the failing CAPI adapter is guarded after the checkstop instead of the system processor.

8335_810.1603 / OP810.32

07/28/16

Impact:  New function      Severity:  NA

New features and functions

• • Support for the CAPI Compression Accelerator Adapter with Feature Codes #EJ1A and #EJ1B and CCIN 2CF0. This CAPI FPGA (Field Programmable Gate Array) adapter acts as a co-processor for the POWER8 processor chip handling specialized, repetitive function extremely efficiently. The adapter is preloaded with a GZIP application and is intended to run as a gzip accelerator. The GZIP application maximum bandwidth is the PCIe3 interface bandwidth. Use of the #EJ1A or #EJ1B adapter requires one #EC2A CAPI activation feature per system. This CAPI gzip feature does not run under PowerKVM but as a bare-metal install only for the following minimum Little Endian (LE) Linux distributions levels:

  • •Red Hat Enterprise Linux 7.2

  • •Ubuntu 14.04.5

  • •Ubuntu 16.04.1

8335_810.1603 / OP810.31

06/10/16

Impact:  New      Severity:  PE

New features and functions

• •

  • • Support for the CAPI Compression Accelerator Adapter with Feature Codes #EJ1A and #EJ1B and CCIN 2CF0. This CAPI FPGA (Field Programmable Gate Array) adapter acts as a co-processor for the POWER8 processor chip handling specialized, repetitive function extremely efficiently. The adapter is preloaded with a GZIP application and is intended to run as a gzip accelerator. The GZIP application maximum bandwidth is the PCIe3 interface bandwidth. Use of the #EJ1A or #EJ1B adapter requires one #EC2A CAPI activation feature per system. This CAPI gzip feature does not run under PowerKVM but as a bare-metal install only for the following minimum Little Endian (LE) Linux distributions levels:

    • •Red Hat Enterprise Linux 7.2

    • •Ubuntu 14.04.5

    • •Ubuntu 16.04.1

8335_810.1603 / OP810.30

03/18/16

Impact:  New      Severity:  PE

New features and functions

• •Support was added for PowerKVM 3.1.0.1 with the following Linux Distributions as virtual guest machines:
  Little Endian: Red Hat Enterprise Linux 7.2 and Ubuntu 14.04.4.
  This feature does not apply to the IBM Power System S822LC(8335-GTA).
   

• •Support was added for an account lockout policy on the BMC service processor. Multiple bad password attempts from a single source will lock out any new or current sessions with that user account for the lockout time period and log a SEL to document the security violation.

• •Support for procedure enumeration in a SEL to be able to decode procedure callouts. This can be used along with the problem determination procedure documentation in the IBM Knowledge Center to produce a repair action for the specific machine type/model that is having the problem. Previously, without this feature, IBM support was required to collect extended data for the failure.

• •Support was added to the BMC Advanced System Management (ASM) to allow customization of the NTP service for setting the BMC time and date. The "Configuration/NTP" panel is used to configure the NTP server or view the current settings.

• •Support was added to the BMC Advanced System Manager (ASM) to be able to verify a Simple Mail Transport Protocol (SMTP) configuration. The BMC "Configuration/SMTP" panel has a "Verify" button that is used to verify the configuration.

• •Support was added to the BMC to light Fan LEDs based on SELs generated for the Fan faults. The LEDs are cleared when the system powers off and re-lit if a SEL occurs again for the fan fault.

• •Support was added to be able to view the Golden and Primary side firmware version information at Petitboot. 

• •Support was added to the BMC to allow Turbo Mode for the processors. Turbo Mode is the default for the systems.
  The Turbo Mode sensor is "green" when in Turbo Mode; "grey" when not in Turbo Mode; and "red" if the sensor has a fault. 

• •Support was added to mirror boot output to the VGA console. 

System firmware changes that affect all systems.

• •A problem was fixed in OPAL for intermittent errors in the module autoload function in the ibmpowernv driver. A compatible property "ibm.opal-sensor" was added to implement the fix for a smooth autoload in Linux. 

• •A problem was fixed in OPAL for lost console output for serial consoles during power downs and reboots. If a power down or reboot is detected, the console output buffer is now flushed before proceeding with the operation.

• •A performance problem was fixed in OPAL for the PCI Host Bridge (PHB) to prevent the PHB L3 cache from retrying defunct entries in the L3 after a MSI end of information (EOI) has been received. The cache line is now flushed after updating the P/Q bits in the priority queue. The situation is improved (and thus performance) by sending a DCBF (Data Cache Block Flush) to force a flush of PHB cache. This improves interrupt performance, reducing latency per interrupt. The improvement will vary by workload.

• •A problem was fixed in Hostboot that caused up to 4 SEL entries to be generated for each checkstop ESEL reported to the BMC service processor. The extra SEL entries would happen if the BMC was busy at the time the SEL was reported.

• •A problem was fixed that prevented multiple NVIDIA Tesla K80 GPUs from being attached to one PCIe adapter. This prevented using a PCIe attached GPU drawer. This fix increases the PCIe MMIO (memory-mapped I/O) space to 1 TB from a previous maximum of 64 GB per PHB/PCIe slot.

• •A problem was fixed in OPAL to identify the PCI Host Bridge (PHB) on CAPI adapter errors and not always assume PHB0.

• •A problem was fixed in the OPAL gard utility to remove gard records after guarded components have been replaced, Without the fix, Hostboot and the gard utility could be in disagreement on the replaced components, causing some components to still display as guarded after a repair. 

• •A problem was fixed for opal-prd service segmentation violations when trying to handle processor A bus errors. With the fix, spare lanes are available to allow fault recovery and not fail the system. 

• •A problem was fixed for the Golden Side of the Self-Boot Engine (SBE) pointing to the Primary PNOR. Without the fix, IBM support would not be able to fully change the system to the Golden Side. This problem originated in manufacturing for a small number of systems and does not affect normal operations of the system. 

• •A problem was fixed for an incorrect callout of DIMMs on a memory failure. The callout may not include the second DIMM in the DIMM pair when it should have. If the system continues to have memory issues and replacing a single DIMM does not fix the issue, replace the second DIMM in the DIMM pair. 

• •A problem was fixed for Nessus security report compliance for SSLv3 and CRIME vulnerabilities. With the fix, the SSLv2 and SSLv3 protocols are disabled. Nessus will still report a CRIME vulnerability due to a limitation in the check but there is no exploit possible for it on the BMC service processor. 

• •A problem was fixed for a SEL not being logged for a missing Power Supply Unit (PSU) and a failure to activate the power supply fault LED on the enclosure. 

• •A problem was fixed for false SEL callouts for DIMMs with high temperatures after OS failed with a checkstop. There were no problems with the DIMMs but the checkstop handling had inadvertently disrupted the monitoring of the DIMM temperatures by the On-Chip Controller (OCC), causing the false SELs. 

• •A problem was fixed for an inband code update error when updating to the same level. 

• •A problem was fixed for false SELs reported during the power off of the system for the power supplies. During the power off, the SEL of "Power Supply PSU Fault 2 | Power Supply AC lost" or "Power Supply PSU Fault 1 | Power Supply AC lost" is intermittently logged. There was no problem for the power supplies in this case. 

• •A problem was fixed for ASM security issues reported by the Open Web Application Security Project (OWASP) ZED Attack Proxy (ZAP) scan tool. These changes to ASM include fixes for X-Frame-options, cookies, cache-control, password auto-complete, XSS protection, and X-content-type-options. 

• •A problem was fixed for the ipmitool failing on the Write PNOR Partition Command and not writing the data expected, The command was not correctly erasing bits before writing, causing unexpected values to be written. 

• •A problem was fixed for a error when disabling or enabling power limits while the system is running. The ipmitool "dcmi power deactivate" and "dcmi power activate" commands failed with On-Chip Controller SRCs of BC8A2A00 and BC8A2A03 logged. 

• •A problem was fix for DHCP discovery on the BMC DHCP client when attached to a new network. The BMC failed to get a new dynamic IP address when attached to a new network unless an AC power cycle was done on the BMC. This also fixes the failure case where the networking is lost for a period of time and the BMC failed to get an IP address when the networking resumed.  

• •A problem was fixed for setting power limits on GPU devices using ipmitool. For the failure, a error message is given in the traces of " [685 CRITICAL][OEMSensorActionTable.c:320] [ GPU Device id: x ] Error in setting power limit to yyy return status : 8" where "x" is the device ID number and "yyy" is the needed power limit. 

• •A problem was fixed for the BMC FRU report to show the new data for a replaced power supply instead of the data for the original power supply. 

• •A problem was fixed for a processor core being disabled and guarded showing as "grey" instead of "red" in the ASM sensor gui. A core that is guarded because of a fault should be detected present and disabled, with a sensor color of "red". The sensor color of "grey" indicates the core is missing. 

• •A problem was fixed for the front and back "Power On" LEDs not being lit. With the fix, the LEDs will blink when the host is powered off. The LEDs will be solid green when the host is powered on. The system power status can also be checked using the ipmitool "chassis power status" command or by looking at the indicator on the power button in the front of the system and the indicators on the power supplies in the back. 

• •A problem was fixed for false SELs being generated for the Power Supplies during system power on with message "Power Supply PSU Fault N | AC out-of-range, but present |" where "N" is 1 or 2. These are asserted and de-asserted. The false SELs have been removed as there is no error with the power supplies. 

• •A problem was fixed for memory frequency not being throttled when required to keep power usage within limits. 

• •A problem was fixed for logging to the uartlog and a limit of 64k was imposed on a single log file to prevent a runaway log. 

• •A problem was fixed for missing SELs and FRU callouts associated with an ESEL for system checkstop errors. When ESELs occurred for fatal errors, there was a timing problem between the system shutdown and the delivery of the SELs to the BMC that caused the SELs to be lost. With the fix, the shutdown process is delayed until the SELs are processed. 

• •A problem was fixed for missing SELs and FRU callouts associated with SMP A-bus faults. Predictive error SELs have bee added that point to the endpoints of the failing A-bus to go along with the existing Unrecoverable SEL that points to an isolation procedure. 

• •A problem was fixed for On-Chip Controllers going into safe mode when workloads involve multiple GPUs. With the fix, there are GPU power capacity changes (lowering the power capacity for each GPU to 150W) and system power capacity changes such that the OCC is able to maintain the limit on the system power usage without failing into safe mode. 

• •A problem was fixed for SELs having incorrect entity instances so that, in the cases of multiple FRUs sharing the same type, the wrong FRU is called out. For example, the "ipmitool sel get" command returned a result that had a FRU description for "CPU 1" when it should have been "CPU 2" 

• •A problem was fixed for an ipmitool "PNOR Get Access Status" command failure getting overlaid with a good return code. The ipmi command did not return a status message nor an error message. With the fix, the error message is displayed for the failed command. 

• •A problem was fixed for the blue chassis Identify LED spuriously turning on during normal operations. This LED is used in service to identify the chassis and can be ignored if it has been turned on in error. With the fix, the chassis LED only turns on when running the service command to identify the chassis. 

System firmware changes that affect certain systems

• •A performance improvement was made for L3 cache prefetches in systems using Direct Memory Interface (DMI) adapters. Work load performances were improved by throttling back on L3 prefetches from the adapters when the cache became more than half way full with busy entries. Performance improvements vary with the type of work loads. 

• •A problem was fixed on a redundant power supply system for an ipmitool "dcmi power activate" success message of "Power limit successfully activated" where the ipmitool command has actually failed. The On-Chip Controller (OCC) correctly rejects the power limit changes that are beyond the maximum value for redundant power supply systems. With the fix, the ipmitool checks the return codes from the OCC to know if the power activate was successful and then puts out a proper message for the pass or fail of the command.

8335_810.1549 / OP810.21

12/14/15

Impact:  Security      Severity:  PE

System Firmware changes that affect all systems

• •A problem was fixed for a BMC service processor system warm reset with chassis power off that intermittently caused a SEL for a "Hard reset" as "| Watchdog 2 Watchdog | Hard reset | Asserted". 

• •A problem was fixed for the "CPU TEMP" processor temperature sensor reporting as "na" instead of providing the processor temperature as needed. For the failure, the report is shown this way: "CPU1 Temp | na | degrees C | na | 0.000 | 0.000 | 0.000 | 255.000 | 255.000 | 255.000".

• •A problem was fixed for the BMC becoming unresponsive and inaccessible from the network after system is running for 7 to 10 days. This was caused by a large memory leak on the BMC. When this failure occurs, the system administrator can allow the system to run until a power cycle (by unplugging the power cords after shutting down the system from the host OS) can be scheduled to clear the BMC of the error.

• •A problem was fixed in OPAL for NX unit checkstops not being reported in the system error log. These are rare NX faults that were forced by error injection methods in the IBM lab to detect the improper handling of the error condition by OPAL.

• •A problem was fixed in Hostboot that prevented the IPL of system when a DIMM error was not detected and memory was not guarded as needed to allow the IPL to continue. This error has SEL reference codes BC8A0D03 and BC8A1701 reported at the Hostboot istep 13.6 that does the memory initialization. Without the fix, the failing memory DIMM has to be manually replaced before the system will IPL. The location of the failed memory can be determined by diagnosing the SEL events.

• •A problem was fixed for a firmware update failure where the Hardware Package Manager (HPM) fails at the end of the image upload with "compcode=0x81: Unknown (0x81)". This error was caused by logging conflicts in the BMC. If the fix for this problem cannot be applied because the firmware update is failing with "0x81", the BMC needs to be switched to the golden side prior to the firmware update. For this service, contact IBM Support at "http://www.ibm.com/support" to have them change the BMC to the golden side and do the firmware update.

• •A problem was fixed to close a security exposure for the BMC.

8335_810.1543 / OP810.10

11/30/15

Impact:  Serviceability      Severity:  PE

System Firmware changes that affect all systems

• •A problem was fixed for the BMC service processor system error log (SEL) showing a non-relevant "Temperature CPU Diode Sensor" errror log. A change was made to filter the error logs since it is not a serviceable event. 

• •A problem was fixed for the BMC service processor system error log (SEL) showing an extra SEL ("OCC 1 Active" or "OCC 2 Active sensor disabled") during a power off of the system. A change was made to filter the error logs since they do not pertain to a real problem.

• •A problem was fixed for a Nessus network security scan detecting that SSLv2 and SSLv3 protocols were active on the BMC. The BMC has been changed to disable SSLv3 and only use TLS to improve the security of the communications. This may affect BMC connectivity for users that have browsers still using the legacy protocols and those users may need to upgrade their clients to use TLS to fix the connection issues.

• •A problem was fixed for the BMC not being able to recover from an intermittent power supply fault. With the fix, the BMC retries the power on so that the power on is able to succeed and also provides a SEL for the initial power on fault.

• •A problem was fixed for the ipmitool retrieval of Power Sequencer extended System Error Logs (eSELs) from the BMC. Without the fix, these eSELs cannot be retrieved but they are needed for problem diagnostics.

• •A problem was fixed for CPU numbers assigned out of order to the Ubuntu 14.04 NUMA nodes. An example of the error on a system with with two ten-core processor modules (for a total of 160 core threads), the lscpu output in Ubuntu showed the CPU numbering as "NUMA node0 CPU(s): 0-7,88-159" and "NUMA node8 CPU(s): 8-87". This was corrected so that the first NUMA node now has "0-79" and the second NUMA node now has "80-159" CPU numbers assigned.

• •A problem was fixed for a ssh service running unexpectedly in the petiboot environment. The ssh service is no longer run as it was not actively being used.

System Firmware changes that affect certain systems

• •On a system with a Ubuntu guest running on a PowerKVM 3.1 host, a problem was fixed for PCI link reset errors on a IBM Power Raid (IPR) adapter. The PCI links may fail to train on a reset. The fix hardens the PCI fundamental reset by retrying as needed to complete the reset.  

8335_810.1539 / OP810.00

10/30/15

Impact:  New      Severity:  New

New features and functions for MTM 8335-GCA and 8335-GTA:

GA Level

NOTE: 

• •This firmware release supports only the OPAL Hypervisor and the Ubuntu  14.04.03 operating system. 

• •GA restrictions on use of the BMC Advanced System Managment (ASM) web gui for doing firmware updates:

• •Use the Google Chrome browser at level 46.0.2490.71 m. Other browsers are failing for firmware updates.

• •ASM firmware update of single components do not work. Select all components for update (which is the default).

• •Back to back in-band update do not work and required reboot between two consecutive in-band updates.

• •Preserve configuration menu is not provided as an option during the ASM HPM update flow. The user has to go to the ASM Maintenance menu and set appropriate preserve flags before using ASM to firmware updates.