Login | Register For Free | Help
Search for: (Advanced)

Mailing List Archive: exim: users

4.52 segfaults on x86_64

  

 
exim users RSS feed   Index | Next | Previous | View Threaded


Axel.Thimm at ATrpms

Jul 10, 2005, 4:06 PM

Post #1 of 10 (589 views)
Permalink
4.52 segfaults on x86_64
Hi,

upgrading from 4.51 to 4.52 and adding domainkeys support leads to
(seldom) segfaults:

http://bugzilla.atrpms.net/show_bug.cgi?id=563

On RHEL4/x86_64 the segfaults look like
segfault at 000000552acc8000 rip 0000002a9697cee9 rsp 0000007fbffbdfd8 error 6
on FC3/x86_64 like
segfault at 000000552acc8000 rip 0000002a968ae72d rsp 0000007fbffbe258 error 6

There is no further logs from exim, in fact the mails get properly
processed.

The i386 builds do not segfault, but perhps it is just not detected on
ia32 hardware.

The build logs for i386 vs x86_64 reveal this:

@@ -216,6 +216,8 @@
from ../exim_monitor/em_StripChart.c:43:
/usr/include/bits/string3.h:99:1: warning: this is the location of the previous definition
gcc exim_monitor/em_TextPop.c
+../exim_monitor/em_TextPop.c: In function `DoSearch':
+../exim_monitor/em_TextPop.c:475: warning: cast from pointer to integer of different size
gcc exim_monitor/em_globals.c
gcc exim_monitor/em_init.c
gcc exim_monitor/em_log.c
@@ -223,6 +225,10 @@
gcc exim_monitor/em_menu.c
gcc exim_monitor/em_queue.c
gcc exim_monitor/em_strip.c
+../exim_monitor/em_strip.c: In function `stripchartAction':
+../exim_monitor/em_strip.c:68: warning: cast from pointer to integer of different size
+../exim_monitor/em_strip.c: In function `create_stripchart':
+../exim_monitor/em_strip.c:254: warning: cast to pointer from integer of different size
gcc exim_monitor/em_text.c
gcc exim_monitor/em_xs.c
gcc exim_monitor/em_version.c

But this was true for 4.51, too, and 4.51 did not segfault.

Any hints/ideas? I will rebuild 4.52 w/o domainkeys and let it run for
a couple of days to see whether the segfaults are due to them.

Thanks!
--
Axel.Thimm at ATrpms.net


Axel.Thimm at ATrpms

Jul 11, 2005, 12:57 AM

Post #2 of 10 (560 views)
Permalink
Re: 4.52 segfaults on x86_64 [In reply to]
Hi,

On Mon, Jul 11, 2005 at 01:06:44AM +0200, Axel Thimm wrote:
> upgrading from 4.51 to 4.52 and adding domainkeys support leads to
> (seldom) segfaults:
>
> http://bugzilla.atrpms.net/show_bug.cgi?id=563
>
> On RHEL4/x86_64 the segfaults look like
> segfault at 000000552acc8000 rip 0000002a9697cee9 rsp 0000007fbffbdfd8 error 6
> on FC3/x86_64 like
> segfault at 000000552acc8000 rip 0000002a968ae72d rsp 0000007fbffbe258 error 6
>
> There is no further logs from exim, in fact the mails get properly
> processed.
>
> The i386 builds do not segfault, but perhps it is just not detected on
> ia32 hardware.
>
> The build logs for i386 vs x86_64 reveal this:
>
> @@ -216,6 +216,8 @@
> from ../exim_monitor/em_StripChart.c:43:
> /usr/include/bits/string3.h:99:1: warning: this is the location of the previous definition
> gcc exim_monitor/em_TextPop.c
> +../exim_monitor/em_TextPop.c: In function `DoSearch':
> +../exim_monitor/em_TextPop.c:475: warning: cast from pointer to integer of different size
> gcc exim_monitor/em_globals.c
> gcc exim_monitor/em_init.c
> gcc exim_monitor/em_log.c
> @@ -223,6 +225,10 @@
> gcc exim_monitor/em_menu.c
> gcc exim_monitor/em_queue.c
> gcc exim_monitor/em_strip.c
> +../exim_monitor/em_strip.c: In function `stripchartAction':
> +../exim_monitor/em_strip.c:68: warning: cast from pointer to integer of different size
> +../exim_monitor/em_strip.c: In function `create_stripchart':
> +../exim_monitor/em_strip.c:254: warning: cast to pointer from integer of different size
> gcc exim_monitor/em_text.c
> gcc exim_monitor/em_xs.c
> gcc exim_monitor/em_version.c
>
> But this was true for 4.51, too, and 4.51 did not segfault.
>
> Any hints/ideas? I will rebuild 4.52 w/o domainkeys and let it run for
> a couple of days to see whether the segfaults are due to them.

Update: The build w/o the domainkeys is segfaulting, too. So it is
just the upgrade from 4.51 to 4.52 that causes segfaults. :(

I downgraded to 4.51 for now. How can I help debug this?

Thanks.
--
Axel.Thimm at ATrpms.net


ph10 at cus

Jul 11, 2005, 2:08 AM

Post #3 of 10 (557 views)
Permalink
Re: Re: 4.52 segfaults on x86_64 [In reply to]
On Mon, 11 Jul 2005, Axel Thimm wrote:

> Update: The build w/o the domainkeys is segfaulting, too. So it is
> just the upgrade from 4.51 to 4.52 that causes segfaults. :(
>
> I downgraded to 4.51 for now. How can I help debug this?

Is it certain messages that provoke the fault? If the segfaults are
rare, it is going to be difficult to find a way of reproducing the
problem with debugging turned on.

I'm afraid I'm busy and away with Exim courses for most of the next 3
weeks, so I can't help much at this time.

--
Philip Hazel University of Cambridge Computing Service,
ph10[at]cus.cam.ac.uk Cambridge, England. Phone: +44 1223 334714.
Get the Exim 4 book: http://www.uit.co.uk/exim-book


ckm at vienenbox

Jul 11, 2005, 2:10 PM

Post #4 of 10 (558 views)
Permalink
Re: Re: 4.52 segfaults on x86_64 [In reply to]
Hello,

Am Montag, 11. Juli 2005 11:08 schrieb Philip Hazel:
> On Mon, 11 Jul 2005, Axel Thimm wrote:
> > Update: The build w/o the domainkeys is segfaulting, too. So it is
> > just the upgrade from 4.51 to 4.52 that causes segfaults. :(
> >
> > I downgraded to 4.51 for now. How can I help debug this?
>
> Is it certain messages that provoke the fault? If the segfaults are
> rare, it is going to be difficult to find a way of reproducing the
> problem with debugging turned on.
>

Same problem here. x86_64 System updated to 4.52 yesterday.
I have some logs appended. First logs are from kernel and second are from
exim. It seems the segfaults are related to ns lookups ?

Gruß
Carsten

=== Schnipp ===
Jul 11 01:21:01 master kernel: exim[8371]: segfault at 000000000060b000 rip
00002aaaab75fe6b rsp 00007ffffffbcba8 error 6

Jul 11 01:21:00 master exim[8370]: no host name found for IP address
220.117.248.203
Jul 11 01:21:00 master exim[8371]: no host name found for IP address
220.117.248.203
=== Schnapp ===

=== Schnipp ===
Jul 11 12:01:22 master kernel: exim[17419]: segfault at 000000000060b000 rip
00002aaaab75fe6b rsp 00007ffffffbcba8 error 6
Jul 11 12:02:35 master kernel: exim[17477]: segfault at 000000000060b000 rip
00002aaaab75fe6b rsp 00007ffffffbcba8 error 6

Jul 11 12:01:20 master exim[17419]: no host name found for IP address
218.150.88.135
Jul 11 12:01:23 master exim[17428]: no host name found for IP address
218.234.142.53
Jul 11 12:01:26 master exim[17428]: H=[218.234.142.53] rejected connection in
"connect" ACL: Your IP is blacklisted. Please us
e your ISP's SMTP to send mail for vienenbox.de . See also
sbl-xbl.spamhaus.org
Jul 11 12:01:26 master exim[17428]: H=[218.234.142.53] rejected connection in
"connect" ACL: Your IP is blacklisted. Please us
e your ISP's SMTP to send mail for vienenbox.de . See also
sbl-xbl.spamhaus.org
Jul 11 12:01:40 master exim[17437]: H=[219.155.95.200] rejected connection in
"connect" ACL: Your IP is blacklisted. Please us
e your ISP's SMTP to send mail for vienenbox.de
Jul 11 12:01:40 master exim[17437]: H=[219.155.95.200] rejected connection in
"connect" ACL: Your IP is blacklisted. Please us
e your ISP's SMTP to send mail for vienenbox.de
=== Schnapp ===


=== Schnipp ===
Jul 11 14:38:59 master kernel: exim[26569]: segfault at 000000000060b000 rip
00002aaaab75fe6b rsp 00007ffffffbcba8 error 6
Jul 11 14:39:01 master kernel: exim[26578]: segfault at 000000000060b000 rip
00002aaaab75fe6b rsp 00007ffffffbcba8 error 6
Jul 11 14:39:09 master kernel: exim[26585]: segfault at 000000000060b000 rip
00002aaaab75fe6b rsp 00007ffffffbcba8 error 6

Jul 11 14:38:53 master exim[26569]: no host name found for IP address
222.142.117.70
Jul 11 14:39:00 master exim[26578]: no host name found for IP address
84.55.140.38
=== Schnapp ===


dwmw2 at infradead

Jul 12, 2005, 3:37 AM

Post #5 of 10 (558 views)
Permalink
Re: Re: 4.52 segfaults on x86_64 [In reply to]
On Mon, 2005-07-11 at 23:10 +0200, Carsten Koch-Mauthe wrote:
> Same problem here. x86_64 System updated to 4.52 yesterday.
> I have some logs appended. First logs are from kernel and second are
> from exim. It seems the segfaults are related to ns lookups ?

What happens if you simulate an SMTP session from one of the offending
IP addresses ('exim -bh 220.117.248.203')?

It would be useful if you could get a tcpdump of a session which
triggers this, or a core dump -- or catch it in gdb.

--
dwmw2


Axel.Thimm at ATrpms

Oct 24, 2005, 11:05 AM

Post #6 of 10 (557 views)
Permalink
Re: 4.52 segfaults on x86_64 [In reply to]
Hello,

unfortunately the segfaults are still there with 4.54. :/

What can be done to help? There are no log messages from exim, I
dumped the traffic and it looks like exim bails out right after
getting the RCPT TO. Here is the scenario under 4.52, it is the same
with 4.54:

0.000000 <smtpclient> -> <exim> TCP 49883 > smtp [SYN] Seq=0 Ack=0 Win=5840 Len=0 MSS=1460 TSV=109455508 TSER=0 WS=0
0.000057 <exim> -> <smtpclient> TCP smtp > 49883 [SYN, ACK] Seq=0 Ack=1 Win=5792 Len=0 MSS=1460 TSV=2075899233 TSER=109455508 WS=2
0.016235 <smtpclient> -> <exim> TCP 49883 > smtp [ACK] Seq=1 Ack=1 Win=5840 Len=0 TSV=109455510 TSER=2075899233
3.929534 <exim> -> <smtpclient> SMTP Response: 220 mail.atrpms.net ESMTP Exim 4.52 Mon, 24 Oct 2005 19:15:19 +0200
3.945012 <smtpclient> -> <exim> TCP 49883 > smtp [ACK] Seq=1 Ack=70 Win=5840 Len=0 TSV=109455903 TSER=2075903163
3.945045 <smtpclient> -> <exim> SMTP Command: EHLO mx05.ispgateway.de
3.945357 <exim> -> <smtpclient> TCP smtp > 49883 [ACK] Seq=70 Ack=26 Win=5792 Len=0 TSV=2075903179 TSER=109455903
3.945898 <exim> -> <smtpclient> SMTP Response: 250-mail.atrpms.net Hello mx05.ispgateway.de [<smtpclient>]
3.963127 <smtpclient> -> <exim> SMTP Command: MAIL FROM:<dimitra.bekou[at]gmx.de> SIZE=1296
4.003421 <exim> -> <smtpclient> TCP smtp > 49883 [ACK] Seq=172 Ack=70 Win=5792 Len=0 TSV=2075903237 TSER=109455905
4.481178 <exim> -> <smtpclient> SMTP Response: 250 OK
4.497518 <smtpclient> -> <exim> SMTP Command: RCPT TO:<axel.thimm[at]atrpms.net>
4.497704 <exim> -> <smtpclient> TCP smtp > 49883 [ACK] Seq=180 Ack=103 Win=5792 Len=0 TSV=2075903731 TSER=109455958
4.500203 <exim> -> <smtpclient> TCP smtp > 49883 [FIN, ACK] Seq=180 Ack=103 Win=5792 Len=0 TSV=2075903734 TSER=109455958
4.516956 <smtpclient> -> <exim> TCP 49883 > smtp [FIN, ACK] Seq=103 Ack=181 Win=5840 Len=0 TSV=109455960 TSER=2075903734
4.516975 <exim> -> <smtpclient> TCP smtp > 49883 [ACK] Seq=181 Ack=104 Win=5792 Len=0 TSV=2075903751 TSER=109455960

On Mon, Jul 11, 2005 at 11:10:25PM +0200, Carsten Koch-Mauthe wrote:
> Hello,
>
> Am Montag, 11. Juli 2005 11:08 schrieb Philip Hazel:
> > On Mon, 11 Jul 2005, Axel Thimm wrote:
> > > Update: The build w/o the domainkeys is segfaulting, too. So it is
> > > just the upgrade from 4.51 to 4.52 that causes segfaults. :(
> > >
> > > I downgraded to 4.51 for now. How can I help debug this?
> >
> > Is it certain messages that provoke the fault? If the segfaults are
> > rare, it is going to be difficult to find a way of reproducing the
> > problem with debugging turned on.
> >
>
> Same problem here. x86_64 System updated to 4.52 yesterday.
> I have some logs appended. First logs are from kernel and second are from
> exim. It seems the segfaults are related to ns lookups ?
>
> Gruß
> Carsten
>
> === Schnipp ===
> Jul 11 01:21:01 master kernel: exim[8371]: segfault at 000000000060b000 rip
> 00002aaaab75fe6b rsp 00007ffffffbcba8 error 6
>
> Jul 11 01:21:00 master exim[8370]: no host name found for IP address
> 220.117.248.203
> Jul 11 01:21:00 master exim[8371]: no host name found for IP address
> 220.117.248.203
> === Schnapp ===
>
> === Schnipp ===
> Jul 11 12:01:22 master kernel: exim[17419]: segfault at 000000000060b000 rip
> 00002aaaab75fe6b rsp 00007ffffffbcba8 error 6
> Jul 11 12:02:35 master kernel: exim[17477]: segfault at 000000000060b000 rip
> 00002aaaab75fe6b rsp 00007ffffffbcba8 error 6
>
> Jul 11 12:01:20 master exim[17419]: no host name found for IP address
> 218.150.88.135
> Jul 11 12:01:23 master exim[17428]: no host name found for IP address
> 218.234.142.53
> Jul 11 12:01:26 master exim[17428]: H=[218.234.142.53] rejected connection in
> "connect" ACL: Your IP is blacklisted. Please us
> e your ISP's SMTP to send mail for vienenbox.de . See also
> sbl-xbl.spamhaus.org
> Jul 11 12:01:26 master exim[17428]: H=[218.234.142.53] rejected connection in
> "connect" ACL: Your IP is blacklisted. Please us
> e your ISP's SMTP to send mail for vienenbox.de . See also
> sbl-xbl.spamhaus.org
> Jul 11 12:01:40 master exim[17437]: H=[219.155.95.200] rejected connection in
> "connect" ACL: Your IP is blacklisted. Please us
> e your ISP's SMTP to send mail for vienenbox.de
> Jul 11 12:01:40 master exim[17437]: H=[219.155.95.200] rejected connection in
> "connect" ACL: Your IP is blacklisted. Please us
> e your ISP's SMTP to send mail for vienenbox.de
> === Schnapp ===
>
>
> === Schnipp ===
> Jul 11 14:38:59 master kernel: exim[26569]: segfault at 000000000060b000 rip
> 00002aaaab75fe6b rsp 00007ffffffbcba8 error 6
> Jul 11 14:39:01 master kernel: exim[26578]: segfault at 000000000060b000 rip
> 00002aaaab75fe6b rsp 00007ffffffbcba8 error 6
> Jul 11 14:39:09 master kernel: exim[26585]: segfault at 000000000060b000 rip
> 00002aaaab75fe6b rsp 00007ffffffbcba8 error 6
>
> Jul 11 14:38:53 master exim[26569]: no host name found for IP address
> 222.142.117.70
> Jul 11 14:39:00 master exim[26578]: no host name found for IP address
> 84.55.140.38
> === Schnapp ===
>
>

--
Axel.Thimm at ATrpms.net


jgh at wizmail

Oct 24, 2005, 11:59 AM

Post #7 of 10 (557 views)
Permalink
Re: Re: 4.52 segfaults on x86_64 [In reply to]
Axel Thimm wrote:
> unfortunately the segfaults are still there with 4.54. :/

And did you get a coredump, or a tcpdump, as advised?

- Jeremy

--
## List details at http://www.exim.org/mailman/listinfo/exim-users
## Exim details at http://www.exim.org/
## Please use the Wiki with this list - http://www.exim.org/eximwiki/


t.luettgert at pressestimmen

Oct 24, 2005, 12:36 PM

Post #8 of 10 (557 views)
Permalink
Re: Re: 4.52 segfaults on x86_64 [In reply to]
On Mon, 2005-10-24 at 20:05 +0200, Axel Thimm wrote:
> Hello,
>
> unfortunately the segfaults are still there with 4.54. :/
>
> What can be done to help? There are no log messages from exim, I
> dumped the traffic and it looks like exim bails out right after
> getting the RCPT TO. Here is the scenario under 4.52, it is the same
> with 4.54:

I had a similar phenomenon some months ago where exim would segfault on
certain (not all! Indeed, we only found one) recipients immediately
after RCPT TO:

The problem went away after I removed the hint DBs in
/var/spool/exim/db. One was obviously corrupted.

I don't remember how I found out that was the problem, though, perhaps
a simulated session with all debug switches on - like
"exim -d+all -bhc (some-IP)".

Hth,
Torsten


--
## List details at http://www.exim.org/mailman/listinfo/exim-users
## Exim details at http://www.exim.org/
## Please use the Wiki with this list - http://www.exim.org/eximwiki/


Axel.Thimm at ATrpms

Oct 24, 2005, 4:48 PM

Post #9 of 10 (558 views)
Permalink
Re: 4.52 segfaults on x86_64 [In reply to]
On Mon, Oct 24, 2005 at 07:59:41PM +0100, Jeremy Harris wrote:
> Axel Thimm wrote:
> >unfortunately the segfaults are still there with 4.54. :/
>
> And did you get a coredump, or a tcpdump, as advised?

No coredump, but a tcpdump (or tethereal dump FWIW) was in the post.

Looks like Carsten Koch-Mauthe found out that this is related to
libspf2, which isn't 64 bits safe. He has submitted a patch upstream
that fixes this. I'll try that patch and report back.
--
Axel.Thimm at ATrpms.net


Axel.Thimm at ATrpms

Oct 30, 2005, 9:12 AM

Post #10 of 10 (552 views)
Permalink
Re: 4.52 segfaults on x86_64 [In reply to]
On Tue, Oct 25, 2005 at 01:48:54AM +0200, Axel Thimm wrote:
> On Mon, Oct 24, 2005 at 07:59:41PM +0100, Jeremy Harris wrote:
> > Axel Thimm wrote:
> > >unfortunately the segfaults are still there with 4.54. :/
> >
> > And did you get a coredump, or a tcpdump, as advised?
>
> No coredump, but a tcpdump (or tethereal dump FWIW) was in the post.
>
> Looks like Carsten Koch-Mauthe found out that this is related to
> libspf2, which isn't 64 bits safe. He has submitted a patch upstream
> that fixes this. I'll try that patch and report back.

I had it now running for almost a week and didn't hit any more
segfaults with this patch. Perviously I would have a couple of dozen
per day.

So this is the fix, thanks to Carsten Koch-Mauthe for finding the bug
in libspf2.
--
Axel.Thimm at ATrpms.net

exim users RSS feed   Index | Next | Previous | View Threaded
 
 


Interested in having your list archived? Contact lists@gossamer-threads.com
 
  Web Applications & Managed Hosting Powered by Gossamer Threads Inc.