[Nottingham] Unkillable process?

[Johannes Kling]

Hello,

  On one of our machines two processes have cropped up that appear to
be completely unkillable; I was wondering if anyone here has come
across something like this and can suggest something other than a
reboot to get rid of it.

Some details:

It's RHEL running on an SMP enabled 2.4.21 and a Xeon that's
pretending to be two processors.

They were launched from a legitimate cron job and, to the best of my
knowledge, the machine is not compromised. It's a short shell script
that does a couple of checks and then invokes patch on a file; it's
run fine before and has been running fine since, so I don't think it's the
script itself.

"ps" shows them to be in state "R", so they're not in uninterruptible sleep:
----
# ps fax | grep -A1 ^31243
31243 ?        R    120409:20 /bin/sh /scripts/checkfpkey
31244 ?        RW     0:00  \_ [patch]
----
(Yes, it's been there for a while :-).

I've sent both a TERM, INT and a KILL (several times) but they
continue to merrily chug along as if nothing's happened. 

"top" claims that the processor the script's running on is working
flat out on "system" jobs, but an "strace" shows nothing at all (not
even a pending system call) and just hangs immediately after
attaching.

The entries in /proc for the script look OK, but the ones for patch
look quite zombie-esque (all links pointing nowhere); I'd expect to
see a Z in ps if that was the case, though, so it's confusing me a
bit.

There's no useful information in dmesg.

Google results for this didn't proove very encouraging; the gist of
the result was that it's probably it's stuck waiting for he kernel to
finish and, until that happens, it'll stay.

I'm beginning to think that the machine's just grown fond of the
process over time and doesn't want to give it up anymore...

Any suggestions would greatly appreciated :-).

Regards,
  Jo

-- 
"I shall press on valiantly in the face of apathy!" -- cmg