Attendees: Joel Schopp, IBM Chris Johnson, SUN Randall Loomis, Wind River Ganesh, Wind River Mary Edie Meredith, OSDL Cliff White, OSDL Martine Silbermann, HP Status of memory hotplug regression testing: - Update on hardware availability: Two systems have been shipped by IBM and arrived at OSDL's facility in Portland. AR: Mary will find out when the systems will be installed and made available in the lab. It is unclear if these systems will be used as project systems or integrated in the STP, since there are 2 of them we might integrate 1 into STP and keep the other for development. - Update on engineering support: Cliff White from OSDL has joined our group and has agreed to participate in the testing effort. He'll help to scope out what the testing would include, if/how it could be automated, how to modify currently used DB tests in conjunction w/ hotplug memory tests so that memory remove/add would be most meaningful. - Input on regression tests from hotplug list: We are still waiting from feedback from some memory hotplug active members on existing tests and scripts. We also discussed the fact that even though hotplug remove does not work at this point in time, there are enough intermediate results that should be tested. Martine has added Dave Hansen's latest patch to the PLM (Patch Lifecycle Manager). Cross compilations showed a lot of FAILs for 2 reasons: - the baseline this patch was applied to also failed on some architectures, - mostly there are many errors in the patch itself. However the way PLM is designed it only gives the number of errors and the name of the files the errors occurred in (there are no details like error messages or line number the error occurred on). We had a discussion leading to the conclusion that to take full advantage of PLM we would need to have a "clean" baseline and a somewhat clean reference patch. To achieve that we will need a few people (one per architecture of interest would be great) to help in the clean-up effort. AR: Martine will send out mail to the memory hotplug list to explain the goal, issues and ask for volunteers. Status of documentation of CPU hotplug: No new input on the doc status itself. However we discussed the possibility of doing CPU hotplug regression testing. While there aren't significant code changes in the CPU hotplug code itself, some other code changes in particular to the scheduler seem to often affect CPU hotplug. It was decided that regression testing of CPU hotplug release candidates would be added to this SIG's chart. AR: Mary will find out what scheduler benchmarks currently used in STP could be useful. Other topics: * Hotplug at the OpenHPI / OpenIPMI level. We have a few members that are particularly interested in hotplug for ATCA. We discussed the possibility of starting an initiative within this SIG to consider hotplug from an OpenHPI and OpenIPMI point of view. AR: Martine to send mail to the HPI, IPMI, ATCA, CGL, DCL communities to "poll" for interest (yes folks it's election season :-) * Joel Schopp proposed a BOF session at LinuxWorld Expo Boston this February. The OSDL hotplug SIG will endorse it and actively participate. * The DCL capability document doesn't currently have an entry for Node Hotplug. After discussion it was decided that it should be a P1 and therefore needs a description. AR: Martine to submit a draft for Node Hotplug entry in the DCL capability document. Next meeting: November 16th, 2004 at 11:00am Pacific Thanks for your participation. Martine J. Silbermann
