Date

  1. I compiled a list of patches that James was talking about here:
    1. https://wiki.hpdd.intel.com/display/LNet/MOFED+4.2+Support
    2. As I noted none of these are to support MOFED 4.2 except maybe “LU-9810 lnet: fix build with M-OFED 4.1” which is not on 2.10. Maybe this can be ported to b2.10
  2. The other o2iblnd patches that I and Alex worked on are fixes and I don’t think we need to necessarily port them back.
  3. https://jira.hpdd.intel.com/browse/LU-10707
    1. I think this is the reason that whole “should we support MOFED 4.2 in b2.10” came up. MOFED 4.2 mention is wrong. I don’t think there are any issues with MOFED 4.2 support in b2_10. The real issue is LU-10707. James feels that porting a whole bunch of time patches back to b2_10 would fix it. But it’s much safer to revert the patches I identified on that ticket to solve the problem. I believe James agrees on that approach as well.
  4. https://jira.hpdd.intel.com/browse/LU-10775
    1. This issue still needs to be investigated in more details. There is a mishmash of discussions and I don’t think we really nailed down what the issue here is.
    2. I’ll spend some time on this to ensure we do not have any backwards compatibility issues in the future.
  5. https://review.whamcloud.com/#/c/31559
    1. This also came up regarding ARM support, but it’s really about > 4K page support  as on PPC and ARM
    2. This needs some thorough review and testing before it lands. We need to make sure it doesn’t break any backwards compatibility
  6. LNet selftest discussion
    1. This is something that came up during a “complaining session” we were having. I was suggesting that it’ll be nice to have a cleaner lnet_selftest interface and as Andreas mentioned a more robust implementation.
    2. But this should be a low priority work and I don’t think it should make it into the 2.12 list, unless there is a Nice-to-have section.