Power7 High-End System Firmware

08/29/17

New features and functions

• DEFERRED:   Support for concurrent replacement of the DCCA on a dual DCCA system.
• Support was added to increase the power capacity limit of the system by 30%, up to 25,000 watts, to handle workloads for drawers with high processor and memory utilization.  Highly-active workloads were driving the power capacity to the limit, resulting in system throttling that reduced performance.  These heavier workloads can now run at normal performance levels.
  
• Support was added to the Advanced System Management Interface (ASMI) to be able to add an IPv4 static route definition for each ethernet interface on the service processor.  Using a static route definition,  a Hardware Management Console (HMC) configured on a private subnet that is different from the service processor subnet is now able to connect to the service processor and manage the CEC.  A static route persists until it is deleted or until the service processor settings are restored to manufacturing defaults.  The static route is managed with the ASMI panel "Network Services/Network Configuration/Static Route Configuration" IPv4 radio button.  The "Add" button is used to add a static route (only one is allowed for each ethernet interface) and the "Delete" button is used to delete the static route.
• Support was added for a concurrent replacement of a DCCA that restores full redundancy of the service processor for the affected drawer.  The DCCA replacement is done concurrently, with the affected drawer powered up and running.

• DEFERRED:   A problem was fixed for filtering Local Network Manager Controller (LNMC) errors for a Host Fabric Interface (HFI) that has failed and gone to a "not ready" state.  Without the fix, the failed HFI continues to log errors (such as "Multicast HW Internal error")  and can flood the Central Network Manager (CNM) error log file.  The HFI error conditions that can cause the extra message logging are a rare occurrence.
  
• A problem was fixed for PCI adapters locking up when powered on.  The problem is rare but frequency varies with the specific adapter models.  A system power down and power up is required to get the adapter out of the locked state.
• A problem was fixed for a Network boot/install failure using bootp in a network with switches using the Spanning Tree Protocol (STP).  A Network boot/install using lpar_netboot on the management console was enhanced to allow the number of retries to be increased.  If the user is not using lpar_netboot, the number of bootp retries can be increased using the SMS menus.  If the SMS menus are not an option, the STP in the switch can be set up to allow packets to pass through while the switch is learning the network configuration.
• A problem was fixed that prevented a second management console from being added to the CEC.  In some cases, network outages caused defunct management console connection entries to remain in the service processor connection table,  making connection slots unavailable for new management consoles  A reset of the service processor could be used to remove the defunct entries.
• A problem was fixed for NIM installs using the Host Fabric Interface (HFI) that failed or other times appear to hang but could complete after many hours of delay.  When the NIM install operation fails,  recover by doing a retry of the operation.  This infrequent problem is triggered by hardware instructions in the HFI Fcode not executing in the required order because of missing synchronization instructions.
• A problem was fixed for a Host Fabric Interface (HFI)  FCode driver error that caused Red Hat Enterprise 7.3 boot failures using the HFI interface.
  The problem has been seen with certain diskless boot images.  The problem is not very frequent, but once encountered, cannot be remedied without a rebuild of the Linux boot image.  The image is gzipped so simply rebuilding the image can cause gzip to compress the image differently due to the new timestamp.  This can be done several times and that may correct the issue.
• A problem was fixed for the DCCA replacement procedure in the HMC R&V (Repair and Verify) to prevent a firmware synchronization error during the DCCA replacement.  The error would also have a connection lost between the HMC and the service processor as the service processor is reset.  The fix involved a change to the error recovery of the ncfgMultSetup application on the service processor to support the DCCA replacement process.  Without the fix, the connection between the HMC and the service processor can be lost during the R&V DCCA replacement procedure, resulting in a failure of the firmware synchronization step.  With the fix, the recovery policy of the ncfgMultSetup daemon was changed so that it would restart itself to handle the setup timing windows for the new DCCA configuration instead of forcing a reset of the service processor, allowing the DCCA replacement process to complete successfully.  The error only occurred infrequently during DCCA replacements on some systems.
• A problem was fixed for incorrect error messages from the Advanced System Management Interface (ASMI) functions when the system is powered on but in the  "Incomplete State".  For this condition, ASMI was assuming the system was powered off because it could not communicate to the PowerVM hypervisor.  With the fix, the ASMI error messages will indicate that ASMI functions have failed because of the bad hypervisor connection instead of falsely stating that the system is powered off.
  

• On systems in IPv6 networks, a  problem was fixed for a network boot/install failing with SRC B2004158 and IP address resolution failing using neighbor solicitation to the partition firmware client.
• For systems with an invalid P-side or T-side in the firmware, a problem was fixed in the partition firmware Real-Time Abstraction System (RTAS) so that system Vital Product Data (VPD) is returned at least from the valid side instead of returning no VPD data.   This allows AIX host commands such as lsmcode, lsvpd, and lsattr that rely on the VPD data to work to some extent even if there is one bad code side.  Without the fix,  all the VPD data is blocked from the OS until the invalid code side is recovered by either rejecting the firmware update or attempting to update the system firmware again.
• For systems with a IBM i load source disk attached to an Emulex-based fibre channel adapter such as F/C #5735, a problem was fixed that caused an IBM i load source boot to fail with SRC B2006110 logged and a message to the boot console of  "SPLIT-MEM Out of Room".  This problem occurred for load source disks that needed extra disk scans to be found, such as those attached to a port other than the first port of a fibre channel adapter (first port requires fewest disk scans).
• A problem was fixed for systems in networks using the Juniper 1GBe and 10GBe switches (F/Cs #1108, #1145, and #1151) to prevent network ping errors and boot from network (bootp) failures.  The Address Resolution Protocol (ARP) table information on the Juniper aggregated switches is not being shared between the switches and that causes problems for address resolution in certain network configurations.  Therefore, the CEC network stack code has been enhanced to add three gratuitous ARPs (ARP replies sent without a request received) before each ping and bootp request to ensure that all the network switches have the latest network information for the system.
• On systems with a PowerVM Active Memory Sharing (AMS) partition with AIX  Level 7.2.0.0 or later with Firmware Assisted Dump enabled, a problem was fixed for a Restart Dump operation failing into KDB mode.  If "q" is entered to exit from KDB mode, the partition fails to start.  The AIX partition must be powered off and back on to recover.  The problem can be circumvented by disabling Firmware Assisted Dump (default is enabled in AIX 7.2).
  
• On systems with dedicated processor partitions,  a problem was fixed for the dedicated processor partition becoming intermittently unresponsive. The problem can be circumvented by changing the partition to use shared processors.
  

07/27/17

Changes:

• No system firmware changes. Refreshing code only to coincide with the BPC update.

04/01/16

System firmware changes that affect all systems

• A problem was fixed for logical partitions not booting after replacement of both DCCAs and service processors in the service drawer.  If the service processors contained incorrect topology data, it is not recalculated, causing bad route information and a hang when booting the partitions.  With the fix, the Local Network Management Controller (LNMC) does a recalculation for the topology when both service processors are replaced, allowing the partitions to boot successfully.
• A problem was fixed for the Integrated Switch Network Manager (ISNM) performance counter output having an incorrect Global Counter timestamp value.  Without the fix, the global counter value is filled with the local GC ID.
• A security problem was fixed in the lighttpd server on the service processor, where a remote attacker, while attempting authentication, could insert strings into the lighttpd server log file.  Under normal operations on the service processor, this does not impact anything because the log is disabled by default.  The Common Vulnerabilities and Exposures issue number is CVE-2015-3200.
• A problem was fixed for reporting all optical link UE errors through the Local Network Management Controller (LNMC).  Without the fix, some of the errors are hidden from the LNMC reports because a threshold count for the error must be exceeded before it is reported to the LNMC.  Even though some of the errors are hidden from the LNMC, they are all visible in the service processor error log.
• On the BPC, a problem was fixed for the remote hardware vitals (rvitals) command returning an incorrect input voltage when there are failed Bulk Power Regulators (BPRs) on the line cord.  With the fix, the BPC reports the highest valid value from BPRs, instead of averaging the voltages.
• On the BPC, a problem was fixed for the remote hardware vitals (rvitals) command returning old (stale) power usage numbers for CECs that are deactivated when the power usage should be zero.  With the fix, the deactivated CECs show zero power usage.
• A security problem was fixed in OpenSSL for a possible service processor reset on a null pointer de-reference during RSA PPS signature verification. The Common Vulnerabilities and Exposures issue number is CVE-2015-3194.
  

10/25/15

09/15/15

06/26/15

10/17/14

09/08/14

08/21/14

System firmware changes that affect all systems

• HIPER/Pervasive:  A security problem was fixed in the OpenSSL (Secure Socket Layer) protocol that allowed clients and servers, via a specially crafted handshake packet, to use weak keying material for communication.  A man-in-the-middle attacker could use this flaw to decrypt and modify traffic between the management console and the service processor.  The Common Vulnerabilities and Exposures issue number for this problem is CVE-2014-0224.
• HIPER/Pervasive:  A security problem was fixed in OpenSSL for a buffer overflow in the Datagram Transport Layer Security (DTLS) when handling invalid DTLS packet fragments.  This could be used to execute arbitrary code on the service processor.  The Common Vulnerabilities and Exposures issue number for this problem is CVE-2014-0195.
• HIPER/Pervasive:  Multiple security problems were fixed in the way that OpenSSL handled read and write buffers when the SSL_MODE_RELEASE_BUFFERS mode was enabled to prevent denial of service.  These could cause the service processor to reset or unexpectedly drop connections to the management console when processing certain SSL commands.  The Common Vulnerabilities and Exposures issue numbers for these problems are CVE-2010-5298 and CVE-2014-0198.
• HIPER/Pervasive:  A security problem was fixed in OpenSSL to prevent a denial of service when handling certain Datagram Transport Layer Security (DTLS) ServerHello requests. A specially crafted DTLS handshake packet could cause the service processor to reset.  The Common Vulnerabilities and Exposures issue number for this problem is CVE-2014-0221.
• HIPER/Pervasive:  A security problem was fixed in OpenSSL to prevent a denial of service by using an exploit of a null pointer de-reference during anonymous Elliptic Curve Diffie Hellman (ECDH) key exchange.  A specially crafted handshake packet could cause the service processor to reset.  The Common Vulnerabilities and Exposures issue number for this problem is CVE-2014-3470.

05/09/14

System firmware changes that affect all systems

• DEFERRED: A problem was fixed that caused a system checkstop during hypervisor time keeping services. This deferred fix addresses a problem that has a very low probability of occurrence.  As such customers may wait for the next planned service window to activate the deferred fix via a system reboot.
  
• DEFERRED: A problem was fixed that caused a system checkstop with SRC B113E504 for a recoverable hardware fault.  This deferred fix addresses a problem that has a very low probability of occurrence.  As such customers may wait for the next planned service window to activate the deferred fix via a system reboot.

10/25/13

03/11/13

11/02/12

System firmware changes that affect all systems

• DEFERRED:  A problem was fixed that could cause a live lock on the power bus resulting in a system crash.
  

06/27/12

06/13/12

System firmware changes that affect all systems

• DEFERRED:  The firmware was enhanced to fix a potential performance degradation on systems utilizing the stride-N stream prefetch instructions dcbt (with TH=1011) or dcbtst (with TH=1011).  Typical applications executing these algorithms include High Performance Computing, data intensive applications exploiting streaming instruction prefetchs, and applications utilizing the Engineering and Scientific Subroutine Library (ESSL) 5.1.

04/12/12