kernel page fault on modprobe of 9746's nvidia.ko

[netllama]

You originally reported that just modprobing the nvidia kernel module resulted in an immediate crash. You now seem to be stating that the system is hanging after some period of time, so that sounds like a different problem altogether.

Does this new problem persist if you're not using the Composite extension and/or set RenderAccel to false in xorg.conf ?
If you set NvAGP to 0 in xorg.conf does that have any impact?

Thanks,
Lonni

[baswesterbaan]

Originally Posted by netllama

You originally reported that just modprobing the nvidia kernel module resulted in an immediate crash.

Indeed. And I tracked that to the corruption of the module, which I now have solved (it was the raid array causing it).

You now seem to be stating that the system is hanging after some period of time, so that sounds like a different problem altogether.

I experienced these problems with the old drivers too. Therefore I wanted to upgrade. I ran into the corruption problem with the newer version. Now I have solved the corruption bug, it seems that this random crash bug that I originally experienced persists.

Does this new problem persist if you're not using the Composite extension and/or set RenderAccel to false in xorg.conf ?
If you set NvAGP to 0 in xorg.conf does that have any impact?

I'll try it out this evening.

By the way, it doesn't seem to freeze when I'm playing ut2004. (However, it does crash on the desktop with and without compiz)

[baswesterbaan]

Does this new problem persist if you're not using the Composite extension and/or set RenderAccel to false in xorg.conf ?

It still seems to crash when composite is disabled. A lot less often though, which is the reason for this slow reply.

I have not set RenderAccel in Xorg.conf. Is the default true or false?

If you set NvAGP to 0 in xorg.conf does that have any impact?

Doesn't seem so.

I attached the bugreport.

[netllama]

RenderAccel defaults to true, you need to set it to false for this test.

Thanks,
Lonni

[baswesterbaan]

Originally Posted by netllama

RenderAccel defaults to true, you need to set it to false for this test.

Ok, it still crashes with RenderAccel on false.

The bug report is included.

[netllama]

What steps are required to reproduce the crash?

[baswesterbaan]

Because the crashes are relatively rare (once in the four hours), it's hard to reproduce.

When under load (bittorent, compiling) it doesn't seem to crash if there is no user interaction.

When there is user interaction (moving with mouse, doing things, opening windows, the usual desktop behaviour) it tends to crash randomly. (It still takes some time).

It tends to crash more likely when doing a rsync (gentoo's emerge --sync). (but that can be coincidence).

I'll try to find a way to reproduce it. But I'm afraid it's too random.