Re: [exim] Exim Segfaults

Page principale
Supprimer ce message
Répondre à ce message
Auteur: Chris Lightfoot
Date:  
À: exim users
Sujet: Re: [exim] Exim Segfaults
On Fri, Oct 13, 2006 at 03:34:34PM +0100, Graeme Fowler wrote:
> On 13/10/2006 14:25, Marc Perkel wrote:
> > Been getting a fair number of these. Maybe one every 20 minutes or so.
> > Running 64 bit FC5. Anyone seen this? What is Error 4?
> >
> > Oct 13 00:24:57 pascal kernel: exim[19299]: segfault at 0000000000000000
> > rip 0000003845f73590 rsp 00007fff428aad68 error 4


Further to Graeme's suggestions:

  - Is your Exim built with debug symbols? If not rebuild
    it with debug symbols (gcc option -g), otherwise
    debugging will be difficult.


  - Is exim dumping a core file? If not, figure out why
    not and fix it, so that you have something to examine.


- Are the fault addresses reported always the same?

  - What does addr2line say about the instruction pointer
    address at which the fault is reported?


  - What was that exim process doing at the time it
    crashed (if your logs don't tell you add more
    logging)? Was it an SMTP listener, invoked by another
    process, a delivery process, a queue-runner etc.?


  - Against what external libraries has exim been linked?
    (E.g. RDBMS client libraries.) Do you need all of
    them?  If not, rebuild it without the unnecessary ones
    linked and see if the problem persists.


On the plus side if you're seeing crashes every twenty
minutes this shouldn't take long to diagnose. It's the
crashes that occur every twenty days that are a pain.

--
The meek may inherit the earth - but not its mineral rights. (Getty)