System hangs: 2.6.6-1-k7 with driver 1.0-6106

[davidbalch]

Hi,

I've got an AMD Athlon based Debian testing/unstable system with a Geoforce FX 5600, and have installed the 6106 driver using the NVIDIA .run file.

Generally the system runs, and I get 3d acceleration in X, which is fine.

Unfortunately, the system hangs completely from time to time, requiring a hard reset. The hang seems to occur randomly, without any obvious pattern in timescale or use of 3d accelerated applications.

The usual recovery methods don't work (ALT+CTRL+BACKSPACE; SysRq r, s, e, i, u, b).

I have run nvidia-bug-report.sh, and attach the results. I have now disabled AGP (Option "NvAgp" "0"), to see if the hang still occurs without AGP.

Given the randomness of the hang, I think I'll need to run without a hangup for at least a week to be reasonably certain that disabling AGP has prevented the problem.

Any information or suggestions about the problem would be appreciated.

Cheers,
Dave.

[davidbalch]

Update: I've just had two freezes in quick succession with AGP disabled, so it's not triggered by using AGP.

[galiza_ceive]

I've get same experience... with kernel 2.6.7 (at a gentoo box, but kernel.org kernel not gentoo tweaked)...

The solution was uncheck EHCI usb option in the kernel and recompile it (verify that grep -i ehci /usr/src/linux/.config --> CONFIG_USB_EHCI_HCD is not set)... I see you have this option enabled (the module is loaded)... So try disabling it.

What mb are you using (maybe a nforce2 based one, like ASUS A7N8X-Deluxe --me--)?

hope this works (as you know there is no trouble with normal nv driver, switch to this at XF86Config to work)...

[davidbalch]

Hmm,

Wouldn't disabling EHCI stop my USB stuff from running at high speed? Although, my usb memory stick isn't working properly, so maybe it's worth a look for several reasons. Can you tell me where I should look to see how ehci + nvidia = crash (some error log maybe?)

My mb is a MSI KT6 Delta (MS-6590), using the VIA KT600 chipset.

Yeah, the nv driver works without problems - I'm not able to spend a lot of time tweaking my system, so I might have to stick with nv for a bit.

Cheers for your help,
Dave.

[galiza_ceive]

I haven't seen a crash log (and didn't read 'bout its existence, you can read /var/log/messages but i didn't found nothing related 2 nvidia module when crashing at startx --yes 'bout not could bring eth0 up at boot, you'll see below--, neither in xfree log)... You know Enhanced Host Controller Interface is set to use USB 2.0 so (till probably a new kernel release you have to choose between nv proprietary driver or USB2.0)... cfr http://forums.gentoo.org/viewtopic.p...87281&start=50
http://bugzilla.kernel.org/show_bug.cgi?id=2846

It's a kinda IRQ problem (yesterday when i found the solution, i couldn't use my internet connection, via a PCI net card --used Gentoo Live CD to try to get the solution from gentoo's forums, 'cuz it worked there---)...

[Anthaus]

It's an EHCI error. Some MoBos assign the same IRQ to USB 2.0 and VGA. When the ehci module crashes and restarts, the IRQ is disabled. If your VGA happens to use the same IRQ, X freezes.
I disabled ehci support while a bugfix is incorporated to the kernel.