Stability Problems amd64

[earoque]

Unfortunately, I've found that the lockups only occur when both screens are activated. Using only the one screen, the random lockups do not occur (unless, of course, beryl is started...) with the binary driver. Right now, I've been running a 2.6.21-rc6-mm1 kernel for over an hour without any problems on a completely rebuilt system.
I'll attach the log from the gentoo-sources kernel.

[energyman76b]

Does the problems also occur, if youcompile the agp driver into the kernel and use the kernel driver?

second, what is this:
atiixp 8272 0 [permanent], Live 0xffffffff8810c000

third, what is this:
00000000-00000000 : Crash kernel

do you have problems too, if you remove kexec or whatever that is from the kernel?

you are using PCIE, so why not get rid of the agp-stuff altogether?

btw, here is a nvidia-bugreport.log for a mighty fine working system:

[earoque]

atiixp: ATI IXP chipset IDE Controller. I'm not explicitly loading this module with any config files, so it is auto-magically loaded at boot.

Crash Kernel line is probably part of the debugging options I've enabled (I'll check on this).

I'm booting with "agp=off", the option is selected by "IOMMU" support in the kernel config. Other than the frame buffer driver for the card, what exactly do you mean with 'kernel driver'?!

What would eliminating "kexec" accomplish?

btw.: obviously there are working systems, but I'm having a heck of a time getting it stable. And: with one screen activated it crashes as well, it just takes at least a half an hour.... or can stay up for five....

[energyman76b]

Originally Posted by earoque

atiixp: ATI IXP chipset IDE Controller. I'm not explicitly loading this module with any config files, so it is auto-magically loaded at boot.

Crash Kernel line is probably part of the debugging options I've enabled (I'll check on this).

I'm booting with "agp=off", the option is selected by "IOMMU" support in the kernel config. Other than the frame buffer driver for the card, what exactly do you mean with 'kernel driver'?!

What would eliminating "kexec" accomplish?

btw.: obviously there are working systems, but I'm having a heck of a time getting it stable. And: with one screen activated it crashes as well, it just takes at least a half an hour.... or can stay up for five....

Well, at first I thought that you might have agp-support as a module - ignore that.

Removing kexec - and everything you don't need, reduces the risks and eleminates possible causes of your problems. Try building a kernel, where only stuff you really need is compiled in/built as module.

[earoque]

I've removed "kexec" and explicitly set "NvAGP" to "0", but my system is still very unstable.

I'm now using the latest stable kernel from www.kernel.org "2.6.20.7", but unfortunately, this doesn't make much difference.

Are there any conflicts with the preemption options in the kernel?

Most of the extraneous junk has been removed from the kernel config. I will streamline it further....

[energyman76b]

Originally Posted by earoque

Are there any conflicts with the preemption options in the kernel?

there shouldn't, but it won't hurt if you try different settings.

Btw, have you tried a different PSU? I solved a lot of stability problems with changing PSUs....

[earoque]

I plugged in the hard disk into another machine that has exactly the same hardware/setup as my own, and it crashed randomly, just like my own system, while the OS(s) that that machine usually runs has run without problems for months (WinXP/Vista), so I think that we can rule out the PSU issue.


I'll try to search for more experimental features and kill the preemption.

[ange70]

Originally Posted by earoque

When I run xorg with the xorg nv driver, I have no problems whatsoever, it runs for days on end.

The system is a HP dc5750 (ATI X-Press 1150 (w) Athlon X2 3800+) with 7300 LE card running an updated Gentoo Linux. I've had the same results with the latest Ubuntu and ArchLinux releases, so I assume it's not because of any interesting distribution-specific patches etc.

Any ideas?


Thanks,

Eamon Roque

Another victim of the rs480+7300GS/LE bug

[earoque]

Now I see.

So, does NVidia need more info or is a fix in the works?!?!

Have I not provided enough/reliable/sane info?

btw.: removing preemption did not help. nor removing quasi-expermental drivers, nor did trying the newest beta driver, froze even more quickly...