1. 29 Feb, 2016 1 commit
  2. 22 Feb, 2016 1 commit
  3. 15 Feb, 2016 1 commit
  4. 10 Feb, 2016 1 commit
  5. 09 Feb, 2016 1 commit
    • Kirk Webb's avatar
      Extend the horrible "non-modular" hack in the Cisco snmpit module. · 0981023d
      Kirk Webb authored
      First, a "bug fix": By coincidence the snmpit Cisco module was able to
      build up the mod/port-to-ifindex mapping on newer switches where
      interface descriptions include a submodule ID (e.g.
      "GigabitEthernet1/0/1" vs. "GigabitEthernet1/1"). However, it was
      actually just grabbing the submodule ID in place of the module ID.  This
      is OK (in some loose sense) for non-modular switches where everything is
      effectivley on the same module. Not good for IOS/NX-OS switches with
      actual modules and submodule identifiers. Things would not have worked
      for these.  As a fix that retains backward compatibility, the Cisco
      snmpit module now correctly extracts the module ID, but subtracts '1'
      from it.  This allows existing installations that have zero-based module
      numbering in their database for non-modular IOS/NX-OS switches with
      submodule IDs to continue to work as is.
      
      The above nonsense is not what I set out to do, however, and the commit
      just gets worse. I extended the hack for non-modular switches with a mix
      of gigabit and ten gigabit to bump any ten gigabit interface port's
      module ID to "1". The existing hack already did this for non-modular
      switches with a mix of fast Ethernet and gigabit.  Absolutely horrific.
      I need a shower.
      0981023d
  6. 29 Jan, 2016 1 commit
  7. 28 Jan, 2016 4 commits
  8. 22 Jan, 2016 2 commits
  9. 21 Jan, 2016 3 commits
  10. 11 Jan, 2016 1 commit
  11. 01 Jan, 2016 1 commit
  12. 17 Dec, 2015 1 commit
  13. 15 Dec, 2015 1 commit
  14. 08 Dec, 2015 4 commits
  15. 07 Dec, 2015 1 commit
  16. 04 Dec, 2015 1 commit
  17. 02 Dec, 2015 1 commit
  18. 30 Nov, 2015 1 commit
  19. 16 Nov, 2015 4 commits
  20. 10 Nov, 2015 2 commits
  21. 09 Nov, 2015 1 commit
    • Mike Hibler's avatar
      Two semi-related exports_setup changes. · 622e2187
      Mike Hibler authored
      1. Don't stat all exported mountpoints from boss to see if they exist.
         Not only is this wrong from boss (a mount may be valid, but not yet
         exported) it also makes the automounter work its butt off as we might
         stat hundreds or thousands of directories across NFS. The original
         purpose of this was to keep mountd from freaking out when presented
         with a non-existent directory. Since mountd freaking out is about as
         rare as the sun rising in the east, we keep this check but do it on
         the proxy (ops) side instead.
      
      2. Have exports_setup (boss) generate a list of directories that it should
         be able to mount. This goes in /usr/testbed/etc/validmounts.txt and will
         be used by autofs (I hope) to narrow down the wildcarded "/users/*"
         and "/proj/*" mounts to those that might actually work. Right now you
         can "ls /users/WtF" and it will try to mount it from ops and take 4
         seconds to timeout and fail.
      622e2187
  22. 05 Nov, 2015 1 commit
  23. 02 Nov, 2015 2 commits
    • Kirk Webb's avatar
      Add horrible hack to snmpit_h3c module to cope with submodules · 5a2b215f
      Kirk Webb authored
      The Emulab DB and codebase cannot currently cope with switch submodules.
      Hack around this in the Comware module by using the ifindex as the
      port number when the submodule ID is non-zero.  So gross, but no less
      arbitrary than some other translation from submodule and port to unique
      identifier.
      5a2b215f
    • Kirk Webb's avatar
      Incorporate Comware v5 support from Keith Sklower · 8881dd48
      Kirk Webb authored
      Also includes cutover to ifDescr parsing instead of using static module
      offsets (which are unreliable in Comware 7).  This allowed me to get rid
      of the awful per-switch module offset hack I added.
      
      Includes some cleanup and minor bug fixes too.
      8881dd48
  24. 29 Oct, 2015 2 commits
    • Kirk Webb's avatar
      snmpit_force10: Enable various OF goodies on created instances. · 4e6f1b2c
      Kirk Webb authored
      L2 and L3 flow maps (hardware-assisted matching and forwarding),
      learning bridge assisted flows (hardware also, I think), and multiple
      forwarding tables.
      
      Default flows go into the "acl" table, which is carved up between various
      switch functions, and I believe is software-based (docs are vague).
      4e6f1b2c
    • Kirk Webb's avatar
      Add support for OpenFlow to Force10 snmpit module. · 84c3cfb2
      Kirk Webb authored
      Adding this support required a change to the createVlan() API call,
      which is why this commit touches so many modules. Optional/supplementary
      arguments to createVlan() are now passed as a fourth hash reference
      argument.
      
      This supplementary hash is used to inform device modules that a vlan to
      be created is to be associated with an OpenFlow instance.  Only the
      snmpit_force10 makes use of this information at this time. FTOS requires
      that openflow-enabled vlans be declared as such at creation time. However,
      snmpit's default control flow creates vlans first, then associates them
      with OF instances (if requested).  This is why the supplementary argument
      (OF hint) to createVlan() is needed here.
      
      OpenFlow configuration actions must all be performed via the
      existing Expect-based CLI interface to FTOS. Yuck.
      84c3cfb2
  25. 22 Oct, 2015 1 commit