1. 08 Feb, 2019 2 commits
    • Leigh B Stoller's avatar
      749a51b9
    • Leigh B Stoller's avatar
      Fix to Jacks constraint stuff: · 3c5e82d6
      Leigh B Stoller authored
      Problem: A user with a reservation at one cluster, was trying to start
      an experiment at a different cluster. But after selecting the other
      cluster and clicking next, that selection was being reset to the cluster
      with the reservation, as part of our overly helpful cluster picker code,
      which is fired off in a callback from Jacks.
      
      Question: Why is that happening when clicking next, to go from the
      Finalize step to the Schedule step, thus silently resetting the user's
      choice.
      
      Reason: There is a call to SwitchJacks("small") to make sure that the
      small version of the topo was visible, in case the user goes back to the
      finalize step. Not sure why ... but in any event, every call to
      SwitchJacks(large|small) results in Jacks firing that callback, and
      resetting the user's cluster choice.
      
      Solution: Null out the callback when calling SwitchJacks().
      
      There is probably a better solution to this, perhaps as part of a code
      reorg on the entire instantiate page, which has grown many appendages
      over time.
      3c5e82d6
  2. 07 Feb, 2019 2 commits
  3. 04 Feb, 2019 5 commits
  4. 01 Feb, 2019 5 commits
  5. 31 Jan, 2019 1 commit
    • David Johnson's avatar
      Make Linux ipod module use multi-hook functions instead of single-hook wrappers. · 6d85b54d
      David Johnson authored
      Kernels built with a bare CONFIG_NETFILTER and the (relatively) new
      CONFIG_TRIM_UNUSED_KSYMS (like our initramfses) have a high probability
      of the nf_(un)register_net_hook functions being elided by the unused,
      exported symbol trim pass.
      
      So, move to using the base, non-wrapper functions (which can accept
      multiple hooks) instead, since those aren't going to be elided.
      6d85b54d
  6. 30 Jan, 2019 2 commits
    • David Johnson's avatar
      In Linux kernel ipod module, only define build targets if not inside kbuild. · 030c16e7
      David Johnson authored
      (Also bump the version, missed that in 4a801b22 .)
      030c16e7
    • David Johnson's avatar
      Move emergency_restart call out of bh in Linux ipod module on aarch64/powerpc64. · 4a801b22
      David Johnson authored
      Calling emergency_restart directly from the softirq/bh ipod icmp handler
      has a history of causing panics prior to the intended shutdown on both
      aarch64 and powerpc64.  The panics didn't inhibit the intended reboot,
      but on OPAL-booted ppc64le, were very noisy, because each "hyperthread"
      cpu attempted a dealloc of the irq handler from within the handler, so
      the console got a noisy stack trace for each thread.
      
      So now if IPOD_QUEUE_RESTART is defined (and it is by default on aarch64
      and powerpc64), we move the call to emergency_restart into a dedicated,
      preallocated workqueue thread, whose sole purpose is to (eventually)
      reboot the machine.  Wasteful, but we don't want to need to initialize
      anything, or use a shared workqueue, if an IPOD is really necessary, at
      IPOD time.
      4a801b22
  7. 29 Jan, 2019 7 commits
  8. 28 Jan, 2019 3 commits
  9. 25 Jan, 2019 1 commit
  10. 22 Jan, 2019 2 commits
  11. 19 Jan, 2019 1 commit
  12. 11 Jan, 2019 4 commits
  13. 10 Jan, 2019 1 commit
  14. 07 Jan, 2019 3 commits
  15. 04 Jan, 2019 1 commit