Mailing List Archive: exim 3.35 crashes with sig11

exim 3.35 crashes with sig11

Jun 20, 2002, 9:54 AM

Post #1 of 8 (1115 views)

Hi,

my exim (debian 3.35-1) crashes with signal 11 since I set the flag
queue_only=true. I checked my RAM, but there is nothing. I compiled the
debian package new with debugging symbols and the error never occured
since then.

I had run ltrace on the old package and it seems, there is a bug.

#v+
17654 strchr("hlL", 's') = NULL
17654 sprintf("alea.gnuu.de", "%*.*s", 12, 12, "alea.gnuu.de") = 12
17654 strcpy(0x080c3bc8, "T:|/usr/lib/news/bin/mailpost -a"...) = 0x080c3bc8
17654 _setjmp(0x080ba880, 0x080c96d0, 0x080c96a8, 0x40000000, 0x080c39c8) = 0
17654 signal(11, 0x08050e10 <unfinished ...>
17654 SYS_rt_sigaction(11, 0xbfffeeb4, 0xbfffee28, 8, 11) = 0
17654 <... signal resumed> ) = NULL

17695 strchr("hlL", 's') = NULL
17695 sprintf("alea.gnuu.de", "%*.*s", 12, 12, "alea.gnuu.de") = 12
17695 strcpy(0x080c3ab8, "T:|/usr/lib/news/bin/mailpost -a"...) = 0x080c3ab8
17695 _setjmp(0x080ba880, 0x080c96d0, 0x080c96a8, 0x40000000, 0x080c38b8) = 0
17695 signal(11, 0x08050e10 <unfinished ...>
17695 SYS_rt_sigaction(11, 0xbfffeeb4, 0xbfffee28, 8, 11) = 0
17695 <... signal resumed> ) = NULL

17712 strchr("hlL", 's') = NULL
17712 sprintf("alea.gnuu.de", "%*.*s", 12, 12, "alea.gnuu.de") = 12
17712 strcpy(0x080c3a60, "T:|/usr/lib/news/bin/mailpost -a"...) = 0x080c3a60
17712 _setjmp(0x080ba880, 0x080c96d0, 0x080c96a8, 0x40000000, 0x080c3860) = 0
17712 signal(11, 0x08050e10 <unfinished ...>
17712 SYS_rt_sigaction(11, 0xbfffeeb4, 0xbfffee28, 8, 11) = 0
17712 <... signal resumed> ) = NULL

17721 strchr("hlL", 's') = NULL
17721 sprintf("alea.gnuu.de", "%*.*s", 12, 12, "alea.gnuu.de") = 12
17721 strcpy(0x080c3ad8, "T:|/usr/lib/news/bin/mailpost -a"...) = 0x080c3ad8
17721 _setjmp(0x080ba880, 0x080c96d0, 0x080c96a8, 0x40000000, 0x080c38d8) = 0
17721 signal(11, 0x08050e10 <unfinished ...>
17721 SYS_rt_sigaction(11, 0xbfffeeb4, 0xbfffee28, 8, 11) = 0
17721 <... signal resumed> ) = NULL
#v-

So is it possible, that there is a bug?

Joerg.

Re: exim 3.35 crashes with sig11 [ In reply to ]

Jun 21, 2002, 9:13 AM

Post #2 of 8 (1101 views)

On Thu, 20 Jun 2002, Joerg Sommer wrote:

> my exim (debian 3.35-1) crashes with signal 11 since I set the flag
> queue_only=true.

I don't know what Debian did to Exim to make their version. Certainly
lots of people use queue_only=true.

> I checked my RAM, but there is nothing. I compiled the
> debian package new with debugging symbols and the error never occured
> since then.

Oh dear. That kind of bug is always nasty.

> 17721 strchr("hlL", 's') = NULL
> 17721 sprintf("alea.gnuu.de", "%*.*s", 12, 12, "alea.gnuu.de") = 12
> 17721 strcpy(0x080c3ad8, "T:|/usr/lib/news/bin/mailpost -a"...) = 0x080c3ad8
> 17721 _setjmp(0x080ba880, 0x080c96d0, 0x080c96a8, 0x40000000, 0x080c38d8) = 0
> 17721 signal(11, 0x08050e10 <unfinished ...>
> 17721 SYS_rt_sigaction(11, 0xbfffeeb4, 0xbfffee28, 8, 11) = 0
> 17721 <... signal resumed> ) = NULL
> #v-
>
> So is it possible, that there is a bug?

It's *always* possible there is a bug. Usually more than one. :-)

The only use of setjmp() within Exim 3 is when it is trying to catch
errors opening DBM files. This dates back to early days when the locking
was done differently. In Exim 4, there are no instances of setjmp.

This looks like a crash inside a DBM library. These are usually caused
by mismatches between different library versions. Maybe re-compiling has
sorted out the problem. (Might as well be optimistic.)

--
Philip Hazel University of Cambridge Computing Service,
ph10@cus.cam.ac.uk Cambridge, England. Phone: +44 1223 334714.

Re: exim 3.35 crashes with sig11 [ In reply to ]

Jun 21, 2002, 1:26 PM

Post #3 of 8 (1099 views)

On Fri, 21 Jun 2002, Philip Hazel wrote:

> > my exim (debian 3.35-1) crashes with signal 11 since I set the flag
> > queue_only=true.
>
> I don't know what Debian did to Exim to make their version. Certainly
> lots of people use queue_only=true.

nothing. all the differences to upstream are things like configuration
location, spool location and the like. there's no code change at all.

--
[-]

Re: exim 3.35 crashes with sig11 [ In reply to ]

sar at pressenter

Jun 21, 2002, 5:06 PM

Post #4 of 8 (1105 views)

On Fri, Jun 21, 2002 at 05:13:03PM +0100, Philip Hazel wrote:
> On Thu, 20 Jun 2002, Joerg Sommer wrote:

> > my exim (debian 3.35-1) crashes with signal 11 since I set the flag
> > queue_only=true.

> I don't know what Debian did to Exim to make their version. Certainly
> lots of people use queue_only=true.

> > I checked my RAM, but there is nothing. I compiled the
> > debian package new with debugging symbols and the error never occured
> > since then.

> Oh dear. That kind of bug is always nasty.

Linux crashing with a SIG 11 can be an indication of dodgy hardware.

Use Google and search on: linux sig 11

Steve

----------------------------------------------------------------------
Steven A. Reisman <sar@pressenter.com> P.O. Box 409
PressEnter LLP 421 N 2nd Street
715-426-2100 or 651-436-5254 River Falls, WI 54022
----------------------------------------------------------------------

Re: exim 3.35 crashes with sig11 [ In reply to ]

Jun 22, 2002, 5:26 AM

Post #5 of 8 (1103 views)

"Steven A. Reisman" schrieb :
> On Fri, Jun 21, 2002 at 05:13:03PM +0100, Philip Hazel wrote:
>> On Thu, 20 Jun 2002, Joerg Sommer wrote:
>
>> > my exim (debian 3.35-1) crashes with signal 11 since I set the flag
>> > queue_only=true.
>
>> I don't know what Debian did to Exim to make their version. Certainly
>> lots of people use queue_only=true.
>
>> > I checked my RAM, but there is nothing. I compiled the
>> > debian package new with debugging symbols and the error never occured
>> > since then.
>
>> Oh dear. That kind of bug is always nasty.
>
>
> Linux crashing with a SIG 11 can be an indication of dodgy hardware.
>
> Use Google and search on: linux sig 11

This case I can exclude, I hope. I ran memory tests and other programs
haven't any problems and ltrace shows the sig11 always at the same point.

Jörg.

Re: exim 3.35 crashes with sig11 [ In reply to ]

Jun 23, 2002, 9:24 AM

Post #6 of 8 (1103 views)

On Sat, Jun 22, 2002 at 12:26:12PM +0000, Joerg Sommer wrote:

> > Linux crashing with a SIG 11 can be an indication of dodgy hardware.
> >
> > Use Google and search on: linux sig 11
>
> This case I can exclude, I hope. I ran memory tests and other programs
> haven't any problems

It's unlikely that there would be memory problems that memtest86 doesn't
pick up, but not completely impossible.

> and ltrace shows the sig11 always at the same point.

That's not unlikely if it's memory problems, actually. If it's a few
particular locations in memory that are faulty, then what you often find is
that it always accesses those memory locations at the same point in the
program.

Re: exim 3.35 crashes with sig11 [ In reply to ]

Jun 23, 2002, 9:27 AM

Post #7 of 8 (1101 views)

On Fri, Jun 21, 2002 at 05:13:03PM +0100, Philip Hazel wrote:

> This looks like a crash inside a DBM library. These are usually caused
> by mismatches between different library versions. Maybe re-compiling has
> sorted out the problem. (Might as well be optimistic.)

Actually I think that's quite likely to be the case.

I've seen similar problems before caused by DBM library mismatches, and the
fact that it crashes consistently for some people but works consistently for
most people, using exactly the same binary, would tend to point to a problem
with a library.

Re: exim 3.35 crashes with sig11 [ In reply to ]

David.Watson at team17

Jun 24, 2002, 5:07 AM

Post #8 of 8 (1104 views)

I've started getting a sig 11 crash with exim 3.31 (RedHat 7.2) and a core
dump on running the sendmail startup script (I relink sendmail to the exim
binary), I recently upgraded apache which required an upgrade of db3 to 3.3.

After a recompile exim is passing mail on, I deleted the db files in
/etc/mail and exims db files.

_____________________________________________________________
David Watson, Network Manager, Team17 Software Ltd.
Phone: +44-1924-267776 Fax: +44-1924-267658
_____________________________________________________________