File _patchinfo of Package patchinfo.21079
<patchinfo incident="21079">
<issue tracker="bnc" id="1189015">SLES 15 SP3 Snapshot16 - [P10] HST:ITC1010:Den:SLES15SP3:HNV:/usr/sbin/hcnmgr: line 608: [: bond61d06af4-: integer expression expected:den11fp1:den11p3 (powerpc-utils)</issue>
<issue tracker="bnc" id="1189571">RPT:SLES15SP3:LPM is failing at the 1st iteration with latest VIOS build,also VASI error seen in lpar too (powerpc-utils)</issue>
<packager>michals</packager>
<rating>important</rating>
<category>recommended</category>
<summary>Recommended update for powerpc-utils</summary>
<description>This update for powerpc-utils fixes the following issues:
- Update from version 1.3.8 to version 1.3.9 (bsc#1189015)
- Raise the log rotation threshold to 1MB so that more history is preserved.
- Fix checking HCNID array size at boot time
- Set autoconnect-slaves on HNV connections.
- When trying to detect HNV bond list all connections, not only active ones.
- Use hexdump from util-linux rather than xxd from vim.
- hcn-init.service: Start together with NetworkManager.
- ofpathname: Fix OF to logical FC lookup for multipath and partitions
- bootlist: Fix for multipath devices with > 5 paths
- Add the missing substring extraction of the partition number from the logical device path so that the partition
number is appended to OFPATH when the logical device is a partition.
- Introducing 'lparnumascore' that is is computing the LPAR NUMA score for CPU and memory.
- This new command detects CPU and memory resources the kernel keep binded to the NUMA node topology of the
departure node. It computes a ratio, and print it to the standard output. The ratio is between 0 (worst) and
100 (best). The command is **not** locking '/var/lock/dr_config_lock' so user should ensure that there is not
ongoing command impacting the computed scores. This allows the command to be run without special privileges.
- Introduce 'of_associativity_to_node'
- This device tree related function reads the ibm associativity property of a node and, using the specified
'min_common_deth', returns the NUMA node id
- Rename 'is_lsslot_cmd' into 'read_dynamic_memory_v2'
- Add manpage for the drmgr utility
- Add a check against 'usr_drc_index' which is set when a remove by index operation is done to ensure the NUMA
removal code is not triggered in that case.
- Add support for the plug ID of a SCSI/SATA host
- 'lpartstat': add '-x' option for the security flavor.
- The output number means:
- 0 = Speculative execution fully enabled
- 1 = Speculative execution controls to mitigate user-to-kernel side-channel attacks
- 2 = Speculative execution controls to mitigate user-to-kernel and user-to-user side-channel attacks
- \- = The running kernel is not exposing the security flavor in '/proc/powerpc/lparcfg'
- Introduce NUMA based LMB removal
- When the NUMA topology can be read, all the LMBs found in the Device Tree are linked the corresponding node.
LMB not associated to node are considered as not used. LMB associated to CPU less node are accounted separately
because they will be targeted first to be remove. The LMB are removed from the CPU less nodes to reach an average
number LMBs per CPU less node. The removal is done through the remove by DRC index API, allowing to remove a LMB
at a time. When the requested amount of LMB could not be removed a partial status is reported.
If the NUMA topology can't be read, we fallback using the legacy remove way.
- Fix ofpathname race with udev rename
- ofpathname: Use NVMe controller physical nsid
- Use a loop to check for all related bonding connections and remove them explicitly one by one instead of using
xargs
- Avoid using 'ifcfg' file for checking bonding interface status
- Wait for sysfs device ready when looking up device name
- Avoid cleanup of bond interface at boot time when no HNV exists
- Adds the option 'ppc64_cpu --version in 'ppc64_cpu --help' in the usage information of ppc64-cpu command.
- Clean up dead network config interface after inactive migration
- Disable 'vnic' as backup vdevice for migratable 'SR_IOV'
- Fix qrydev checking for active interface error
- Skip collecting lsdevinfo right before migration. When the LPARs has large number of devices, this can take quite
a long time and flood the log message.
- Wait for OS ready to lookup device name for configure HNV device.
- The hybrid network virtualization is only supported on PowerVM LPAR. Exit the 'hcnmgr' gracefully on other power
platform instead of return error.
- Optimize 'lsdevinfo' filtering to prevent LPM timeouts (bsc#1189571)
</description>
</patchinfo>
