Segfault in 2.1.10 backports version advice
Any advice on a segfault situation...? Jan 23 13:29:17 LX800476 kernel: [1366692.780725] freeradius[23459]: segfault at 8 ip b7461326 sp b5105988 error 4 in libc-2.7.so[b7403000+155000] Running a backports verison of freeradius on Debian Lenny: 2.1.10+dfsg-2~bpo50+1, 2.6.26-2-686 on Vmware ESX cluster. I cannot reproduce it on a test server and it only happens in production. Probably a load thing...? I could upgrade to current stable version in git, I could upgrade the OS (Lenny to Squeeze). Debugging from this backports version seems an impossible road? Or I could install the -dbg version and perhaps run the server in a screen session? However I have experienced it won't crash if run in debug mode (-X). I reckon in -X it is run in single threaded mode? Rg, Arnaud
Arnaud Loonstra wrote:
Any advice on a segfault situation...?
Upgrade.
I cannot reproduce it on a test server and it only happens in production. Probably a load thing...?
Possibly.
I could upgrade to current stable version in git,
Upgrade to the v2.1.x branch in git.
I could upgrade the OS (Lenny to Squeeze). Debugging from this backports version seems an impossible road? Or I could install the -dbg version and perhaps run the server in a screen session? However I have experienced it won't crash if run in debug mode (-X). I reckon in -X it is run in single threaded mode?
Yes. Alan DeKok.
On 2012/01/23 03:20 PM, Alan DeKok wrote:
I could upgrade the OS (Lenny to Squeeze). Debugging from this backports version seems an impossible road? Or I could install the -dbg version and perhaps run the server in a screen session? However I have experienced it won't crash if run in debug mode (-X). I reckon in -X it is run in single threaded mode?
Yes.
Hi, I can confirm the same problem. Version is freeradius-git downloaded about 4 days before 2.1.12 was released. Running with -X it runs forever. (About two months now) Without, it crashes about once a week. Have not had the time to collect debug info. -- Johan Meiring Cape PC Services CC Tel: (021) 883-8271 Fax: (021) 886-7782 -------------------- Before acting on this email or opening any attachments you should read Cape PC Service's email disclaimer at: http://www.pcservices.co.za/disclaimer.html
Johan Meiring wrote:
I can confirm the same problem.
Version is freeradius-git downloaded about 4 days before 2.1.12 was released.
Running with -X it runs forever. (About two months now) Without, it crashes about once a week.
Well, the only thing I can see which could be it is the changes to the rlm_passwd module. If you're using that, set "hashsize=0". Otherwise it *will* crash. If you're not using rlm_passwd, see doc/bugs. Backtraces will help a lot. Alan DeKok.
On 01/23/2012 02:44 PM, Alan DeKok wrote:
Johan Meiring wrote:
I can confirm the same problem.
Version is freeradius-git downloaded about 4 days before 2.1.12 was released.
Running with -X it runs forever. (About two months now) Without, it crashes about once a week.
Well, the only thing I can see which could be it is the changes to the rlm_passwd module. If you're using that, set "hashsize=0". Otherwise it *will* crash.
If you're not using rlm_passwd, see doc/bugs. Backtraces will help a lot.
Alan DeKok. - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html
Well, I am using an rlm_passwd module with hashsize=0. So that shouldn't be the case in my case? I'll see if I can upgrade and see if it differs. Rg, Arnaud
Hi, On Mon, Jan 23, 2012 at 02:13:55PM +0100, Arnaud Loonstra wrote:
Jan 23 13:29:17 LX800476 kernel: [1366692.780725] freeradius[23459]: segfault at 8 ip b7461326 sp b5105988 error 4 in libc-2.7.so[b7403000+155000]
Running a backports verison of freeradius on Debian Lenny: 2.1.10+dfsg-2~bpo50+1, 2.6.26-2-686 on Vmware ESX cluster.
I cannot reproduce it on a test server and it only happens in production. Probably a load thing...?
Had similar segfaults here on 2.1.10. Was very quick to debug - enabled core dumps, and waited for it to crash. Then started up gdb on the corefile and looked at the backtrace. For me, it was a feof bug in the detail listener (copy-acct-to-home-server) - so if you're using that, then you may have hit the same. Fixed in 2.1.11. The backtrace is what you really need to debug it. Upgrade first, though. Compiling the latest from git (or 2.1.12) is trivial as all the Debian stuff is there to build your package for you :-). Matthew -- Matthew Newton, Ph.D. <mcn4@le.ac.uk> Systems Architect (UNIX and Networks), Network Services, I.T. Services, University of Leicester, Leicester LE1 7RH, United Kingdom For IT help contact helpdesk extn. 2253, <ithelp@le.ac.uk>
participants (5)
-
Alan Buxey -
Alan DeKok -
Arnaud Loonstra -
Johan Meiring -
Matthew Newton