NVidia - Kernel - X.org Troubleshooting - Please Help.

riverty · n00b Joined: 23 Sep 2004 Posts: 22

Getting Started:

I have a machine that's about 1 year old. On this machine, I have had a dual-boot setup with Windows XP Professional and some flavor of Linux ever since it came home. I take good care of the hardware and this machine is solid.

I've been running Linux for about 7-8 years now (yea, I started with Slackware) and am fairly seasoned in the ways of the Linux world.

Now, for the past, say 6 months, I've been running Fedora Core 2. Built on the 2.6 kernel with X.org and NVidia driver version 1.0-6111 all working great. But it's Fedora. All bloated, chunk full of crap I'll never use.

Enter Gentoo:

I went for it. Yea man! Bootstrapped the system and all. Took days to get where I am now but seemingly worth it?

Ok. I'm running 2.6.8 with the .config based off of "genkernel," X.org version 6.7.0 and KDE Desktop. No errors to report on system compilation. Everything smooth except running the NVidia driver.

Now this is killing me. I've compiled a LOT of kernels and installed a lot of drivers and editied a lot of configs. If I Run NVidia Drivers, This Machine Will Fail Within An Hour! If I don't run NVidia drivers, this machine will run until I shut it off.

Now, I'm not one to write a post if my answer can be found somewhere. I've spent a lot of time here in Gentoo Forums, NVidia Forums, and the Internet in general. No solid answer. No decent fix. Other people have written posts about this but nothing solid came out.

What Happens:

The system seems to continue to run but KDE / X is froze solid. Mouse still moves around the screen though. Can't click on anything, nothing responds, but .mp3's in XMMS keep playing! Harddrive keeps kerchunkin' away at whatever it's doing. Can't Alt/Tab, Ctrl/Alt/Backspace, or Alt/F1-6. There's really nothing I can do but kill the power! Unacceptable!

I could use some help with this one. Use my machine to troubleshoot anything right now. I'm sure others could use an answer to this as well. BTW, I have emerged 3 different versions of the NVidia driver and 2 kernel versions to no avail. Everything compiles, installs, and seems to want to run OK.

Here are these. If anyone wants anything else, just ask! And, thank you for reading.

Linux Kernel 2.6.8 - current .config:

Archangel1 · Veteran Joined: 21 Apr 2004 Posts: 1212 Location: Work

You're meant to comment out the Load "dri" line in xorg.conf with nVidia drivers.
I've had a similar issue with a couple of random lockups, which I've blamed on dodgy kernels. What kernel sources are you using?

c0bblers · Guru Joined: 28 Mar 2003 Posts: 403

Hi,

How are you loading the nvidia kernel module? If you're using fastwrites or sba try turning them off. Also, you might want to fiddle with AGP, like trying it with no agp, with kernel agp and then with the nvAGP (check the docs that come with nvidia-glx to find out how). I'm assuming that said card is stable in other situations...but it could be hardware. BTW If you install acpid (if your system has acpi support) you should be able to shut your system down safely by using the power button...failing that running an ssh server is always handy for remote logging in and shutting down, your data will thank you :-)

Cheers,
James

riverty · n00b Joined: 23 Sep 2004 Posts: 22

Archangel1,

First, thanks for the reply! Second, I am fully aware that the NVidia driver suggests commenting out "Load DRI" when using their driver and when testing, I DO have this commented out.

c0bblers,

I am NOT using fastwrites or sba. I wouldn't engage these options without having a stable driver in the first place. As for AGP. The card I'm running is PCI. It's a 128MB NVidia GeForce FX 5200. I have tried "noagp" to no avail.

tkoster · n00b Joined: 26 May 2003 Posts: 46

I've got little to offer in advice at this stage (the drivers have always worked great for me, except in suspending) but a couple more questions:

1) Which version of the driver are you using?
2) After things have hung up, do the logs give you any hints?
3) Have you recompiled the kernel, xorg or anything else major since you created the driver. One thing I have found with the nvidia driver is that it likes to be recompiled everytime you mess with the kernel.

riverty · n00b Joined: 23 Sep 2004 Posts: 22

tkoster,

Driver versions I have tried:

43?? - I forgot!
5335
6111

I have checked the logs and found nothing.

I am aware that one has to recompile the NVidia driver after a new kernel installation. I have not tried replacing/recompiling KDE or X as I see no reason to. Maybe I'm missing something there. Both X and KDE seem to run fine on the 'nv' driver.

Again, I appreciate your responses. Keep it up! :wink:

tkoster · n00b Joined: 26 May 2003 Posts: 46

Hmmm. I guess it couldn't be anything simple, could it? The only time I've run into trouble with nvidia was when I recompiled my kernel to add a module and forgot to recompile nvidia as well. Otherwise it has been rock solid.

The only other thing I can think of (and you've probably tried this too) is to search the forums on nvidia's site.

lucid · n00b Joined: 02 Jan 2003 Posts: 12 Location: Norway

Just wanted to say that i've got the same problem with the same grapics card, except that i've got 256mb ddr memory.
It works fine with the 'nv' driver, but when i switch to 'nvidia' it hangs after a short time, although i can ssh in and shut it down.

I use:

riverty · n00b Joined: 23 Sep 2004 Posts: 22

lucid,

Well, you and I and a host of others seem to have this problem. It's funny that I have found many individual posts all saying about the same thing. Would be nice to get all these posts into one thread. I think everyone would see that there are MANY people having Gentoo / NVidia / X.org lockups.

Anyway, no - this is not an easy fix, that's why I'm here ranting about it! My same hardware will game for hours without even a hiccup in Windows XP. Gaming! Yea, running the system hard for hours at a time without problems. I test this daily. It's not hardware. It's a setting in my BIOS, it's a setting in my kernel, or maybe even it's compiling my whole system with:

CFLAGS="-march=athlon-xp -O3 -pipe"

I dunno. I read a few posts about people with patches applied to the NVidia driver that claim it fixed their problems and it's now stable. If you keep reading these threads you'll find that the patch only works for a few people, while some still have their problems. Not a good enough (or definitive) fix for me.

I wonder if I re-installed the NVidia driver, then tried to run some programs from JUST X-windows, NOT KDE, if it would lockup. I'm thinking it would answer if it's KDE locking the system or not. Wadda you think? We've got to narrow this down a bit. Has anyone else tried running the NVidia driver in just plain X?

lucid · n00b Joined: 02 Jan 2003 Posts: 12 Location: Norway

What patch are you speaking of?

Ryle · Apprentice Joined: 27 Aug 2004 Posts: 204

I am going to post my 2 cents because I have been having a very similar problem.

In my case it seems to have to do with the AGP. At first I was running with agpgart, any games/apps using glx (or any of the 3d type functions) seemed to either hang completely like you described (xscreensaver did this, had to power down each time) OR it would hang but I could ctrl-alt-f1 to the console then ctrl-alt-f7 back to the program and it would be ok again until it happened again.

Now, I took agpgart out of the kernel which disabled AGP altogether. This totally fixed the problem for me... BUT, without AGP the games that ran fine when I had AGP going were now really slow (playable, but choppy). I next tried NvAGP and I run into the same problems I had with agpgart.

If I could only come up with a solution to this problem Gentoo would run flawless.

riverty · n00b Joined: 23 Sep 2004 Posts: 22

Ryle,

Interesting. I have read this in other posts as well. Some have good luck, some don't. I have AGP compiled into my kernel (as modules) for future use (my mobo has a 4X-8X AGP slot) as I plan to replace my 5200FX sometime in the future and go with AGP.

I will recompile my kernel without AGP support and try. I'll report back what happens...

riverty · n00b Joined: 23 Sep 2004 Posts: 22

Recompiled my kernel without AGP support at all. X and/or KDE locked up in less than 5 minutes!

I found out that there is a new BIOS upgrade for my mobo. I downloaded, flashed, and reset my BIOS. I am writing this on that reboot now and so far, so good! Been putting the machine through hell too. Got XMMS playing .mp3's, checked out some OpenGL screen savers, copiled a few proggies too. Been up about 35 minutes now!

Keeping my fingers crossed...

tkoster · n00b Joined: 26 May 2003 Posts: 46

Just out of curiosity, what brand of computer/mobo are those of you with the problem using?

riverty · n00b Joined: 23 Sep 2004 Posts: 22

tkoster,

I'm using a Soyo SY-K7VME Mobo
AMD Athlon 2400 XP
512 MB PC2700
GeForce FX 5200 PCI 128MB
Sound Blaster Live 5:1 PCI
Onboard VIA Ethernet
40GB Seagate UDMA-100
60GB Western Digital UDMA-100
48X Sony CD-RW

Well, I updated the BIOS and ran X. Wrote the last post and locked up about 1 hour into my session. I'm back to the 'nv' driver for now. What's next??

Ryle · Apprentice Joined: 27 Aug 2004 Posts: 204

Wow, I'm using a soyo sy-kt600 Dragon plus 1.0 myself. One thing I've disliked so far I cannot get it to work with 8x agp. I've read a lot about others not being able to do so either. I've updated my BIOS, messed with all the BIOS settings, tried both agpgart and nvagp, enabled/disabled fast writes and sba, and even turned off eth0 and all my usb to ensure nvidia driver was on its on IRQ.

The only thing that seems to have prevented lockups for me was turning off agp.

tkoster · n00b Joined: 26 May 2003 Posts: 46

The reason I ask was that I was looking over the nvidia forums and found you were not alone in dealing with this problem. They had no real solutions, but seemed to attribute this particular combinations of motherboards and nvidia cards. Asus motherboards seemed the most common, though it was not exclusive to them (I take that one back the mobos involved are all over the place). GeForce 2 and 4 cards as well but again not exclusive. It seems to be distro independent.

Reliable solutions are unavailable. Some of the solutions that worked for some people but not all include:
1) Putting in a different nvidia card
2) Blocking all frame buffer stuff from kernel as well as xorg.conf (is dbe a framebuffer or some other type of buffer)
3) doing something different with agp - sometimes in the xorg.conf and sometimes in the kernel
4) going back to a 2.4 kernel
5) one guy tied it to using a vga console with color as he booted instead of a plain text console

So you may try any of those ideas, but if they work for you, you'll be one of the fortunate ones.

tkoster · n00b Joined: 26 May 2003 Posts: 46

riverty,

If it was all working in fedora, you didn't happen to save your kernel settings or xorg.conf file from there did you. It would be interesting to do a compare.

ribx · Posted: Thu Sep 30, 2004 6:14 pm Post subject:

are you sure that you are always running the correct nvidia-glx version?

did you try to disable ALL modules in your xorg.conf?

did you read the newest README.nvidia? i doesnt so far...

hope i could help a bit

electrofreak · l33t Joined: 30 Jun 2004 Posts: 713 Location: Ohio, USA

Perhaps it isn't X, but KDE for some reason. Try 'emerge fluxbox' and use fluxbox for a while. Its light weight and fast as hell. But I guess I don't know about Ctrl+Alt+Bksp and such not working... But Since the music is still playing, your system is still responsive. You could probably actually save it if you ssh'd into it or something.
_________________
Desktop: ABit AN8, Athlon64 X2 4400+ 939 2.75GHz, 2x1GB Corsair XMS DDR400, 2x160GB SATA RAID-0, 2x20"W, Vista Ultimate x64
Laptop: 15.4" MacBook Pro 2.4Ghz, 2x1GB RAM, 160GB, Mac OS X 10.5.1
Server: PIII 550Mhz, 3x128MB RAM, 160GB, Ubuntu Server 7.10

riverty · n00b Joined: 23 Sep 2004 Posts: 22

Ryle said:

"The only thing that seems to have prevented lockups for me was turning off agp."

I have read others turning off AGP and getting mixed results. My card, being PCI, shouldn't have this problem but I have tried turning off AGP in BIOS, Kernel, and X to no avail.

tkoster:

Yes, everything was working in Fedora. No, I didn't save any configs from that installation. Hell, I've had Fedora Core 2, Mandrake, a few other Linux Distros and Windows XP Pro and Home on this very box without problems at all. With this machines track record, I guess I wasn't expecting to have video problems. Would be nice to have those configs though! :wink:

No tkoster, I think I'm gonna go back to a 2.4 kernel and try from there. With this there are many options. Do you think it would be from using CFLAGS="-march=athlon-xp -O3 -pipe"? I think the 'normal' optimizations setting is -02. Maybe I'm pushing it a bit?

If the -03 optimizations flag could be the culprit, then I should format and build a whole new system based on the 2.4 kernel. If the -03 is OK, then it would save me from bootstrapping the system.

What do you all think?

riverty · n00b Joined: 23 Sep 2004 Posts: 22

No replies on this one? OK. Well, I decided to format and reinstall basing this installation on the 2.4 kernel. To save time, I took the Athlon-XP, stage 2 tarball and built the system from there. BTW - I DID backup my configs this time! At this time I have a clean boot into 2.4 with all needed modules loaded and ready. X and KDE come next. Will report more in a few days.

RtX...

moocha · Watchman Joined: 21 Oct 2003 Posts: 5722

It's NOT AGP, it's NOT a specific NVidia card, it's NOT ACPI, it's NOT the APIC controller, it's NOT a specific BIOS, it's NOT a specific motherboard, it's NOT AGP voltage, it's NOT a specific X.Org or XFree86 version, it's NOT a specific NVidia or ATi driver version, it's NOT even necessary to use the proprietary drivers (I get the lockeups even when using the nv driver), it's NOT the web browser (I get lockups in Opera and Firefox and *Dillo* out of all

), it's NOT the window manager (I get them in Fluxbox and KDE).

The only thing that these lock-ups seem to have in common is using a web browser (which employ a lot of widgets), using a 2.6 kernel, getting the "mouse moves, keys are locked up, I can ssh in" symptoms, and... using Gentoo.

By now I'm almost convinced it's a Gentoo specific patch to the compiler(s). There's no other probable scenario spanning so much different hardware.

For reference, here's the setup I get the infamous lockups with:

Dual P3 (Katmai 450) SMP system
Soyo D6IBA motherboard (Intel 440BX chipset)
384 MB SDRAM
Riva TNT (yes, TNT one, first NV core card, and it still locks up

)
AGP turned *off* in BIOS, xorg.conf, and everywhere I could get my hands into (even no agpgart support in kernel)
Anything related to APM (duh, SMP disables APM), ACPI, and IO-APIC is turned *off*, additionally I'm booting with noacpi noapic on the kernel command line
Any kernel from 2.6.1/SMP from the 2004.0 LiveCD to 2.6.9-rc3 (I tested at the very least 35 different kernels)
Gentoo (fresh install, bootstrapped w/gcc 3.4.2-r2)
X.Org 6.8.1-r1
_________________
Military Commissions Act of 2006: http://tinyurl.com/jrcto

"Those who would give up essential liberty to purchase a little temporary safety deserve neither liberty nor safety."
-- attributed to Benjamin Franklin

tkoster · n00b Joined: 26 May 2003 Posts: 46

No, its not gentoo either. If you look at the nvidia forums, they have a large number of the same complaints and the distros are all over the place. Fedora, Mandrake, SUSE, Slackware, Debian, etc.

The frustrating thing about this bug is that there are all sorts of things that it isn't but it can't get nailed down to what it is.

moocha · Watchman Joined: 21 Oct 2003 Posts: 5722

Hum... In this case it *is* Gentoo however, because this exact hardware worked just fine 1 month ago (with the proprietary drivers even, hardware OpeGL acceleration and all). Now it doesn't. Must be the compiler, the bootstrap procedure I used was my standard one and I've installed over 40 Gentoo systems.
_________________
Military Commissions Act of 2006: http://tinyurl.com/jrcto

"Those who would give up essential liberty to purchase a little temporary safety deserve neither liberty nor safety."
-- attributed to Benjamin Franklin