Tom,
On So, 05 Aug 2007, Tom Kistner wrote:
> > I see a very strange behaviour with exim on one x86_64
> > SLES10 Linux system. At the moment a daemon (exim -q15m -bd
> > or for tests exim -q15s) does not start the periodic queue runners.
>
> Bug in SuSE's SSH packages.
>
> See
>
quoting:
> http://www.exim.org/lurker/message/20061204.090958.aa0c4616.en.html
"What you are seeing is probably a bug in Suse's OpenSSH package which
incorrectly masks signals. When you restart Exim on a SSH shell, it
won't get SIGALRM any more.
To confirm, restart Exim on a regular console. It should then work."
Thanks for pointing into the right direction. It seems that it's no
SUSE openssh bug anymore (they do include a patch which blocks
SIGALRM while logging for a short moment to avoid a race condition,
but that looks correct).
On the problematic system openssh had been restarted from a
process with SIGALRM blocked though. Don't know how that happened
initially, but now (through long running screen sessions etc.)
the system is tainted with many processes that have
0000000000002000 set in the SigBlk: field of /proc/PID/status.
Strange, that no other problems resulted (or at least haven't
been recognized).
Anyway - no Exim topic.
Lutz
Lutz Preßler <Lutz.Pressler@???> http://www.SerNet.DE/
SerNet Service Network GmbH, Bahnhofsallee 1b, D-37081 Göttingen
Tel.: +49-551-370000-2, FAX: +49-551-370000-9
AG Göttingen, HRB 2816, GF: Dr. Johannes Loxen
