Re: Deadlock problems with 'kill PID' on CURRENT

From: Peter Wemm <peter_at_wemm.org>
Date: Mon, 15 Nov 2004 08:57:42 -0800
On Monday 15 November 2004 05:35 am, Robert Watson wrote:
> On Mon, 15 Nov 2004, Martin Blapp wrote:
> > > syscons is typically a sign that some thread holding Giant is
> > > spinning, or that Giant has been leaked.  Alternatively, some
> > > other lock is leaked, and a thread with Giant waits on that lock.
> > >  You may be able to get into DDB using the serial console as it
> > > doesn't require Giant (or you might not).
> >
> > I was not able to get into DDB over serial console. Sorry for not
> > saying this exactly. Hardlock.
>
> Any chance that any of the boxes has an NMI button?

I had this problem too BTW, but it didn't prevent DDB.  Once named or 
rndc got stuck, trying to kill them was Really Bad(TM).  Both use 
libpthread.  The programs were stuck in "runnable" state according to
CTRL-T (siginfo).

No, I haven't looked closer.  That system was built about when the last 
round of libpthread turmoil was happening.
-- 
Peter Wemm - peter_at_wemm.org; peter_at_FreeBSD.org; peter_at_yahoo-inc.com
"All of this is for nothing if we don't go to the stars" - JMS/B5
Received on Mon Nov 15 2004 - 15:57:44 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:22 UTC