staprun segmentation fault with module `fglrx' in use

Martin Hunt hunt@redhat.com
Mon Mar 19 14:10:00 GMT 2007


On Mon, 2007-03-19 at 02:50 +0000, Guanqun Lu wrote:

> It seems that it fixes my problem too. Thanks, Martin.
> 
> Maybe I should diff the two files to see what's actually the problem.

I think the problem is that fglrx is enormous and has 167 sections. It
was overflowing buffers and  the code in staprun wasn't checking all
buffer writes before they happened so the overflow sometimes didn't get
caught. I checked in changes this week that ignore the  sections that we
obviously don't care about. I also made the temporary buffers much
larger and made sure every buffer write was first checked to be sure it
would fit, so at least we should get an error message and clean exit.

Martin




More information about the Systemtap mailing list