debugging a segfault in libc-2.10.1.so ?

[aryan_]

Hi All,

I have several user-developed processes written in C++ which run as Linux daemons and communicate with each other using IPC.

At some point after starting the daemons, one of the them causes the following segfault:

kernel: [1040186.227966] my_app[16338]: segfault at 2f7265bc ip 003ef3d7 sp bfa10070 error 4 in libc-2.10.1.so[394000+13e000]

I am not sure if this is a very general message which does not carry any information to resolve it.

My questions are:
1. How should I read the error message above generated by the kernel ? What do the fields contained in the message mean? Do they carry any valuable information to resolve the segfault error?

2. What steps do you suggest to successfully isolate the problem and resolve it ? (in terms of debugging tools or anything else)

The system, where the daemons are deployed, has recently undergone an OS upgrade (upgrade to Ubuntu 9.10). (May it be the reason for the segfault ?)

Thanks.

[ngwasuma]

Which Diastro of Linux are you running /Developing them from?

[aryan_]

Which Diastro of Linux are you running /Developing them from?

Ubuntu 9.10

[Rubberman]

Assuming you also updated your development tools, then I would suggest that you recompile and relink your application code. My best guess is that there is an incompatibility between the version of libc that you compiled/linked with originally, and that provided by the new libc provided by 9.10.

[aryan_]

Assuming you also updated your development tools, then I would suggest that you recompile and relink your application code. My best guess is that there is an incompatibility between the version of libc that you compiled/linked with originally, and that provided by the new libc provided by 9.10.

I already got this segfault error after I recompiled and relinked my application daemons. Hence this might be more than a recompilation/relink problem.

Don't you think that this segfault may be stemming from my own application code itself instead of libc due to a memory leak, undefined behaviour, etc ?

[Rubberman]

I already got this segfault error after I recompiled and relinked my application daemons. Hence this might be more than a recompilation/relink problem.

Don't you think that this segfault may be stemming from my own application code itself instead of libc due to a memory leak, undefined behaviour, etc ?

It could be due to a bug in your code. Build it with the -g (debugable) option so you can do a post-mortem analysis of the crash.