[exim] Spurious segfaults with Exim 4.88

Pàgina inicial
Delete this message
Reply to this message
Autor: Christian Balzer
Data:  
A: exim-users
Assumpte: [exim] Spurious segfaults with Exim 4.88

Hello,

Since Phil is going on about 4.89 being a stabilizing release I thought
I'd better pipe up here and now.

Debian (Jessie, backports) packages here, but I doubt it's a distro thing
since 4.84 (the default Jessie version) doesn't exhibit this while being
linked against the same libraries.

Historically the only segfaults I've seen with Exim where a few EXPN/VRFY
attempts that got fixed with 4.87.

But since the first appearance of 4.88 I'm seeing a few (see below) of
these:
---
Jan 25 15:38:49 mx08 kernel: [3040142.375490] exim4[17474]: segfault at 0 ip 00007f41e8082062 sp 00007ffd509381d8 error 4 in libc-2.19.so[7f41e7ffd000+1a1000]
Jan 25 15:41:47 mx08 kernel: [3040319.783357] exim4[18070]: segfault at 0 ip 00007f41e8082062 sp 00007ffd509381d8 error 4 in libc-2.19.so[7f41e7ffd000+1a1000]
---

I've not been able to associate these to specific events or connections
with any certainty, but ylmf-pc scans and other spamware are very high on
my list of suspects.

Why?

Because we handle 1.6 million connections per day on our MXs with 3000 of
these segfaults happening. None at all on our MSAs, btw.

And 1000 are on the 6 primary MX servers while the other 2000 are on the 2
secondary ones, which of course see the typical extra attention by
spammers.

I'd be happy to collect cores if this can be turned on with the existing
binaries, but I have a feeling that the RLIMIT_CORE override is still in
there and not changeable w/o compiling stuff, right?

Christian
-- 
Christian Balzer        Network/Systems Engineer                
chibi@???       Global OnLine Japan/Rakuten Communications
http://www.gol.com/