Utility Programs

Front-End Tools

• None documented yet.

Back-End Tools

• client_session.sh - Initializes the database and runs the network client. When the server closes the connection, it runs the post-processing scripts.
• list_all_dups.sh pid filename time - Gives you a list of all file descriptors pointing to filename at time time.
• get_active.sh start end - Gives you a list of all pid's that did anything between start and end.
• did_proc_write.sh pid filename start end - Prints "1" if pid wrote to filename between start and end, prints 0 otherwise. I would have just made it return 1 and 0 (which it also does), but I don't know how to check the returncode from a function in a bash script yet. Wu-chang will probably berate me for this.
• get_writers.sh file start end - Get a list of all pid's that wrote to file between start and end. Currently, the way this operates is very inefficient. It's just for demo purposes. Rewriting this is on my TODO list.
• get_io.sh pid - Takes a pid and outputs all of the writes that went to that pid's stdout, stdin and stderr. The format of the returned data is "[date]-tab-[count]-tab-[data]". The data is in binary-escaped format (i..e, you don't see a newline, you see "\n"). The data is not necessarily sorted by date.
• get_io pid rollcount - Takes a pid and a rollcount, and does what get_io.sh does. However, it does it much more efficiently, and it also takes into account the fcntl syscall. Does not handle dups that are interleaved between processes created with clone. But other than that it's pretty good.
• replay_shell [-f filename] [-r] [pid] - Replays a shell session. Or, in fact, this will replay the output written to stdout and stderr by any process. Parameters:
  • -f filename: a file in the format generated by get_io.sh.
  • -r: If present, this specifies that the session is to be replayed in realtime (default is to play as fast as possible).
  • pid Specifies a pid in the database. You do not need to specify a pid if you specify a filename. Note that retrieving the data from the database can be time-consuming. It is a better idea to use get_io.sh to retrieve the data and store it in a file, and then use replay_shell to simply access that file.
  Currently, this program is a bit buggy. It won't handle sessions larger than 32k in length. I'll fix that soon.
  Binary
  Source