Re: probe process.function on libc not working

[Adrien Kunysz <adrien@kunysz.be>]

[-- Attachment #1: Type: text/plain, Size: 1766 bytes --]

On Wed, Jan 04, 2012 at 09:09:51AM -0500, Laris Benkis wrote:
> >> Warning: child process exited with signal 11 (Segmentation fault)
> > What does the backtrace, the EIP and the code around the EIP in that
> > core look like?
> I've attached the core file.  Here are the registers and backtrace.  Thanks
> Laris
> 
> (gdb) info registers
> eax            0xfffff000       -4096
> ecx            0x2400   9216
> edx            0x800000 8388608
> ebx            0x44fdeff4       1157492724
> esp            0xbf84c85c       0xbf84c85c
> ebp            0x44e16000       0x44e16000
> esi            0x8      8
> edi            0x1      1
> eip            0xbf850070       0xbf850070
> eflags         0x10206  [ PF IF RF ]
> cs             0x73     115
> ss             0x7b     123
> ds             0x7b     123
> es             0x7b     123
> fs             0x0      0
> gs             0x33     51
> (gdb) bt
> #0  0xbf850070 in ?? ()
> #1  0x0000000b in ?? ()
> #2  0x44e16a54 in ?? ()
> #3  0x44e0ee78 in ?? () from /lib/ld-2.14.90.so
> Backtrace stopped: Not enough registers or memory available to unwind
> further

That core file is pretty much useless without de debug symbols and
I don't have a Fedora machine. What do you see if you disassemble
the code around the instruction pointer? (0xbf850070).

Another idea to try to understand this would be to see whether probing
only part of the libc functions cause the problem. Is this just with
one execve()-related function? All the functions? Or does this happen
only when probing many functions? If you are familiar with Python
an adaptation of the script Timo posted last month might be helpful
to explore this: http://sourceware.org/ml/systemtap/2011-q4/msg00402.html

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]