1. 19 Feb, 2019 1 commit
  2. 15 Feb, 2019 6 commits
    • Leigh Stoller's avatar
      Fix the empty password on the user account we create on the dell · 69c75587
      Leigh Stoller authored
      switches, I must have forgot to deal with it.
      
      Basically, the FTOS manual says you can give it an MD5 or sha256 hash of
      the password. Well, FTOS must come from another dimension, cause the MD5
      and sha256 it generates, is different then what boss generates, for the
      same input string. So that makes it hard to set up an account with a
      password, unless I send over the cleartext. Which is okay, it stores it
      internally as the hash (I cannot seem to generate). So we send over the
      cleartext per-node root password that is regenerated for each time it is
      allocated to an experiment. Not a big deal, we load the user's ssh keys,
      they should not need to use passwords.
      69c75587
    • Mike Hibler's avatar
      Make sure iscsid picks up change of initiator name. · 1c2d994c
      Mike Hibler authored
      Otherwise you get a lot of this action:
      
      WARNING: 10.254.254.203 (iqn.1993-08.org.debian:01:5ad48b44316d): session reinstatement from different address 10.254.254.201
      
      The issue was that we would restart iscsid after changing the name, but if the
      iSCSI sessions were already open, the name change would not immediately take
      effect (til next reboot).
      
      This will happen if the open-iscsi and/or iscsid service is started at boot time
      prior to the Emulab blockstore config running. We explicitly turned these services
      off in Ubuntu 14, but not in 16 and above cuz that would involve interaction with
      systemd and some of us don't speak systemd. Anyway, this will work even if the
      services are enabled at boot.
      1c2d994c
    • Leigh Stoller's avatar
    • Leigh Stoller's avatar
      66e36ba7
    • Leigh Stoller's avatar
      5125ba82
    • Leigh Stoller's avatar
      Add missing library. · f360b627
      Leigh Stoller authored
      f360b627
  3. 14 Feb, 2019 2 commits
  4. 13 Feb, 2019 1 commit
  5. 12 Feb, 2019 2 commits
    • Leigh Stoller's avatar
      Fixes to previous revision. · 74e3d7c0
      Leigh Stoller authored
      74e3d7c0
    • Leigh Stoller's avatar
      Recovery mode: · bde6c94d
      Leigh Stoller authored
      * Add a new Portal context menu option to nodes, to boot into "recovery"
        mode, which will be a Linux MFS (rather then the FreeBSD MFS, which
        99% of user will not know what to do with).
      
      * Plumb all through to the Geni RPC interface, which invokes node_admin
        with a new option, to use the recovery mfs nodetype attribute.
      
      * recoverymfs_osid is a distinct osid from adminmfs_osid, we use that in
        the CM to add an Emulab name space attribute to the manifest, that
        tells the Portal that a node supports recovery mode (and thus gets a
        context menu option).
      
      * Add an inrecovery flag to the sliver status blob, which the Portal
        uses to determine that a node is currently in recovery mode, so that
        we can indicate that in the topology and list tabs.
      bde6c94d
  6. 11 Feb, 2019 3 commits
  7. 08 Feb, 2019 10 commits
  8. 07 Feb, 2019 2 commits
  9. 04 Feb, 2019 5 commits
  10. 01 Feb, 2019 5 commits
  11. 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
  12. 30 Jan, 2019 1 commit
  13. 29 Jan, 2019 1 commit
    • 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