Bug#741412: exim4: process crashed with signal 11 while delivering

[Dominic Hargreaves]

On Thu, Mar 13, 2014 at 09:30:53AM +0100, Marc Haber wrote:
> tags #741412 unreproducible moreinfo
> thanks
> 
> On Wed, Mar 12, 2014 at 10:43:14AM +0100, Leszek Dubiel wrote:
> > After upgradading debian to woody I started to get exim crashed for some
> > emails (a few emails for few thousand deliveries). 
> 
> Woody has been out of support for, IIRC, ten years.
> 
> That being said, signal 11 errors usually indicate faulty hardware,
> most probably bad RAM or a bad CPU. Please check with memtest or other
> tools.

I'm seeing something similar just after upgrading to jessie on an i386
system:

2015-04-19 11:34:32 1YjmYa-0004me-9S <= dom at larted.org.uk U=dom P=local S=357
2015-04-19 11:34:33 1YjmYa-0004me-9S == dom at larted.org.uk <dom at thyone.larted.org.uk> R=smarthost T=remote_smtp_smarthost defer (-1): smtp transport process returned non-zero status 0x000b: terminated by signal 11
2015-04-19 11:34:33 1YjmYb-0004mk-2e <= <> R=1YjmYa-0004me-9S U=Debian-exim P=local S=668
2015-04-19 11:34:33 1YjmYa-0004me-9S Frozen
2015-04-19 11:34:33 1YjmYb-0004mk-2e == dom at larted.org.uk <postmaster at thyone.larted.org.uk> R=smarthost T=remote_smtp_smarthost defer (-1): smtp transport process returned non-zero status 0x000b: terminated by signal 11
2015-04-19 11:34:33 1YjmYb-0004mk-2e Frozen

What's interesting is that the messages are delivered anyway (okay, and
not logged or noticed as such because of the crash), however when the
queue is thawed, the messages aren't delivered as expected:

2015-04-19 11:43:53 Start queue run: pid=19664 -qff
2015-04-19 11:43:53 1YjmYa-0004me-9S Unfrozen by forced delivery
2015-04-19 11:43:53 1YjmYa-0004me-9S Completed
2015-04-19 11:43:53 1YjmYb-0004mk-2e Unfrozen by forced delivery
2015-04-19 11:43:53 1YjmYb-0004mk-2e Completed
2015-04-19 11:43:53 End queue run: pid=19664 -qff

So I wonder if there is a secondary bug here, since in the second set of
logs the messages which were on the queue simply disappear without trace.

There is a clue about the underlying cause of the crash here:

Apr 19 11:34:33 thyone kernel: [73206.710071] exim4[18397]: segfault at be50a32d ip b72ab614 sp bfa30430 error 6 in libgnutls-deb0.so.28.41.0[b71f7000+13a000]
Apr 19 11:34:33 thyone kernel: [73207.469822] exim4[18403]: segfault at 96f9278d ip b72ea614 sp bfdc50d0 error 6 in libgnutls-deb0.so.28.41.0[b7236000+13a000]

I will run memtest on this system overnight. In the meantime I've avoided
this problem by disabling TLS, which makes me sad.

Cheers,
Dominic.