Summary list of ARs from this meeting: AR for Joel: Find out if IBM is willing to share their user memory remove tests and in what time frame. AR for Mary: 1.Find out the details of what was wrong w/ the data when the problem first came to light, add that info to the wiki. 2. Ask Bryce to contact community about this fact that stats data are not being re-initialized when CPU is onlined. AR for Martine: 1. Add appeal to have fragmentation patches reviewed in the SIG report at the Beijing Face to face. 2. Contact Christoph Lameter directly to ask about tests/scripts for memory fragmentation. Attendees: Mary Edie Meredith, OSDL Bruce Vessey, Unisys Natalie Protasevich, Unisys George Mann, Indivual Contributor Joel Schopp, IBM * Progress report on automation of the testing process Bryce is still trying to determine how to automate on ppc64 falling back on a well known good kernel if the default kernel doesn't boot. Joel suggested that the script might need to be done in an interactive way. * Memory hotplug common regressions - testing plan. Mary will post the memory migration questions on the memory hotplug mailing list and integrate Dave Hansen's suggestions on fragmentation testing in the testing plan. The memory regression testing plan doesn't currently have any input for either user memory remove or kernel memory remove. Joel suggested that IBM has some tests for removing user memory and that they might be willing to share them w/ OSDL. AR for Joel: Find out if IBM is willing to share their user memory remove tests and in what time frame. There seems to be an issue with some of the data in /proc that is used by the statistics tools. This implies that we might need to do a sanity check on those data. AR for Mary: Find out the details of what was wrong w/ the data when the problem first came to light, add that info to the wiki. It was also noted that the data in /proc was the same for a CPU that was offline and online again. The CPU data goes away when the CPU is offlined but when it is online the data is not reinitialized. AR for Mary: Ask Bryce to contact community about this issue. * Progress report on i386 hotplug CPU patches. Natalie reported that her patches to support hotplug CPU for ia32 are in mm and will be pushed into mainline soon. She has tested them w/ intensive I/O for 48h uninterrupted on an ia32 system. * Other topics: Mary expressed concerns for the limited amount of resources that we have to address the memory regression testing. Bryce is only on the project on a part time basis and there is currently no intern to help along. In order to help us prioritize the work Joel share the following info. The basics to make memory remove work are: * fragmentation avoidance, there are currently a set of patches that have been submitted for review but have gotten very little feedback. Please encourage your local memory people to review those patches. * memory migration, in development * interfaces required to have all the remove code work on the various architecture. We need to accelerate the development for fragmentation avoidance, the goal is to get this into mainline before the end of the year. AR for Martine: Add appeal to have fragmentation patches reviewed by community and member companies in the SIG report at the Beijing Face to Face. AR for Martine: Contact Christoph Lameter directly to ask about tests/scripts for memory fragmentation. Next meeting is scheduled for October 25th at 11:00am -12:00pm PST, 2:00pm - 3:00pm EST. Thanks for your participation. Martine J. Silbermann
