Hi all,

I have a weird problem with devg_i830 driver and the Intel 82Q35 Express Integrated Graphics Controller (on a Quad-Core 
PC) when doing some GF/OpenGLES grawing (under photon). 

If my program have a bug and receive a segmentation fault signal (SIGSEGV) after gf_dev_attach() is called, the system 
crash completely instead to generate a coredump. The mouse and keyboard stop responding, the computer is not accessible 
by network, nothing.  On another older computer that use the devg_extreme2 driver with chipset: 82865G Integrated 
Graphics Controller from Intel, I got the coredump as expected. Or if a SEGFAULT happens before calling gf_dev_attach(..
) or after calling gf_dev_detach(...), the coredump work and the system doesn't freeze.

I can reproduce the same thing with the gf-ph-3d opengl ES demo program, just by forcing a SEGFAULT after the call of 
init_gears(....) function on cbs.c. 

e.g. by adding
    strcpy(NULL, "CRASSSSHHHHHH");

It also crash with gf-ph-2d  (no opengles), so it doesn't need opengles code to got the system crash.

It doesn't crash the system when running under gdb, so it seems to crash when it attempt to write the coredump file...

It also does a system crash when I try to suspend the application with CTRL-Z (SIGSTOP) but not when quitting the 
application using SIGINT or SIGHUP signals.

I also notice that if I capture the SIGSEGV signal and does nothing (signal(SIGSEGV, empty_function)), it just hang the 
program but the system doesn't crash.

I know we normally don't have a SIGSEGV in normal condition, but this make program hard to debug if it crash on a 
production machine and we don't have access to the coredump. And anyway this might point to a real bug that can be 
solved....

Is there another driver that can be used for this chipset ? (en existing one by hacking the enum script or an 
experimental one).  Can somebody can reproduce the same problem (with similar system) ?

I've attach the output of pci -vvv for my system..

--
Martin Gagnon
NDT Technologies inc.