Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Mon, 12 May 2008 10:41:03 -0500

Michael Abbott wrote:

I guess there are two separate issues here:
1. Exiting with epicsThreadExitMain() appears to be a very bad thing, at
least on RHEL;

The whole point of the routine epicsThreadExitMain() is to arrange for
the main thread to complete _without_ shutting down the other threads om
the process. If you want to shut down an IOC completely, you have to
call epicsExit() instead. The softIoc implementation will do this if
you tell it to start an interactive shell (use the -s option in R3.14.9;
in R3.14.10 -s does nothing and the default is to start a shell unless
you explicitly say -S).
If you don't ask for an interactive shell there's nothing for the main
thread to do after loading databases so it calls epicsThreadExitMain().
At this point the process looks like a zombie, but it has many other
threads still running (ls /proc/<pid>/task). If you kill -9 one of
these other threads the whole process should shut down and the zombie
goes away.
If you want to be able to shut down a softIoc using Control-C, you have
to start an interactive shell. If you want to be able to shut down a
softIoc that has no interactive shell, load the softIocExit.db and do a
caput to its $(IOC):exit PV which causes epicsExit() to be called.
2. RHEL appears to have an unpleasant bug involving process cleanup -- how
else do I have persistent zombies owned by init? Or else I'm missing
something...

See above, there are still other threads running in the process.
The iocsh() call is necessary to reproduce the problem.

...

Curious. I would have expected the _exit(0) call to have done everything
that's necessary. As far as I can tell from the man page, _exit is pretty
well guaranteed to terminate the program.

The _exit() call does not trigger the atexit() callbacks which trigger
the shutting down of the other threads. In your example, iocsh() calls
errlogFlush() which calls errloginit() which starts the errlog thread.
This initialization is rather unfortunate, but it would take some work
to arrange for the errlog thread to only be started when there is
actually some work for it to do. I would love to see that done if
someone wants to take it on.
HTH,
- Andrew
--
Talk is cheap. Show me the code. -- Linus Torvalds

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System