1. 04 Apr, 2006 1 commit
  2. 03 Apr, 2006 1 commit
  3. 31 Mar, 2006 1 commit
    • Robert Ricci's avatar
      Some more robust error handling. Now, nothing in the main loop calls · 47a82473
      Robert Ricci authored
      croak(). Instead, the client is dropped, and an error is logged. If
      we get too many errors (10), we bail. But, if we go a while without
      getting any errors (60 seconds), we reset the error count. This is
      so that if we have any bugs (or the node is just totally fubared),
      we'll exit, but long-running daemons won't just accumulate the error
      counter forever.
      
      Also, handle a rare (but, on plab, who knows!) DOS-type condition with
      blocking accept() noted in Stevens.
      47a82473
  4. 30 Mar, 2006 1 commit
  5. 28 Mar, 2006 5 commits
  6. 27 Mar, 2006 2 commits
  7. 24 Mar, 2006 2 commits
  8. 22 Mar, 2006 1 commit
  9. 21 Mar, 2006 6 commits
  10. 17 Mar, 2006 1 commit
  11. 16 Mar, 2006 1 commit
    • Robert Ricci's avatar
      More work on the framework for running the monitor and stub · c8bbc432
      Robert Ricci authored
      automatically.
      
      Now automatically determines the number of peers in the experiment.
      
      run-monitor-libnetmon.sh automatically generates the IP mapping file
      now.
      
      Auto-create /local/logs if it doesn't exist (it doesn't on our
      devbox image)
      
      Add two new 'auto' scripts - these are intended to be run as part
      of the automatic experiments. They properly wait for the stubs
      to start before starting the monitors, etc. They log the outputs
      of the stub and monitor to /local/logs, where loghole can pick
      them up. They also kill their children if they are killed (to be
      used with the program agent)
      
      The auto-* scripts need a little more work in term of properly waiting
      for their children to exit.
      c8bbc432
  12. 15 Mar, 2006 3 commits
  13. 14 Mar, 2006 2 commits
  14. 13 Mar, 2006 1 commit
  15. 07 Mar, 2006 1 commit
  16. 02 Mar, 2006 1 commit
  17. 23 Feb, 2006 2 commits
  18. 20 Feb, 2006 1 commit
    • Robert Ricci's avatar
      Keep some state about the connections we're monitoring, so that we · 78db03dc
      Robert Ricci authored
      can report it once the connection is successful.
      
      Now decides whether or not to monitor an FD in socket(), rather than
      connect(), since we now have to start collecting data before connect()
      is called.
      
      Move code to report on connection state changes into inform*()
      functions.
      
      Warning: Tested, but not exhaustively.
      78db03dc
  19. 16 Feb, 2006 1 commit
    • Robert Ricci's avatar
      Add support for reporting if the application changes a few sockopts: · 4d565e26
      Robert Ricci authored
      SO_RCVSIZE
      SO_SNDSIZE
      TCP_NODELAY
      TCP_MAXSEG
      
      There is a problem, though. Right now, we identify a connection by a
      triple of (local port, remote IP, remote port). However, some apps
      (iperf, at least), set these options before calling connect(). Thus,
      we don't know these things yet. To fix this, we could either:
        Delay notification of sockopt changes until after connect() is
          called for the socket
        Use something else (ie. (pid,fd)) as the ID for connections.
      We should probably talk about this (at least Jon and I) at the
      next meeting.
      4d565e26
  20. 15 Feb, 2006 2 commits
    • Robert Ricci's avatar
      Make a new control channel between libnetmon and netmond. Right now, · b195dbd5
      Robert Ricci authored
      there are only a few control messages implemented:
      
      The client can ask the server for the current operating parameters
      
      The server can tell the client about maximum sockbuf sizes, or
        sockbuf sizes that should be force
      
      The server can tell the client which version of the output format to
        use.
      
      The plan is that there will be more in the future: ie., the client
        and server will be able to exchange messages about new socket
        buffer sizes set by client apps, and exchange information about
        sockopts, etc. the client app may have set.
      
      Add support to libnetmon for capping the socket buffer size.
      b195dbd5
    • Robert Ricci's avatar
      Make a few things work on FreeBSD. · 115a00ee
      Robert Ricci authored
      115a00ee
  21. 03 Feb, 2006 2 commits
  22. 02 Feb, 2006 2 commits