Display, kernel, nvidia, X, drivers, HARD LOCKUP [SOLVED]

[HeXiLeD]

This is a long and painful
problem report (and thank you for using your time reading it)

I have been having a very complicated nvidia problem that for that past month i have been trying to
figure out where the problem is and fix it and i haven't been able to get to definitive conclusion and
possible solution.

For some reason my "display" freezes. With X and without X.

Since i have ran out of ideas here goes the report of everything that started to happen and how.

I had done emerge -e system and later -e world to try so solve other problems. However this started and i believe
after the first -e system ( i cannot be 100% sure )

The first time i stared to see this problem happen i was using 2 monitors.
One LCD with 1280x1024 res and a 14" CRT with 1024x768

Here is xorg config file: http://freethinker.ws/system/xorg-dual.txt (made by nvidia-settings)

If i opened a video file on the CRT and later, on the LCD, the video would fail and only show a green
image. The sound was ok.
Later this also happened on just one monitor and it also didn't matter the which monitor i used first.

Getting the video to display properly again would require X restart or some weird things like ( rm
/tmp/* ) or just wait until the next day ( i think do to a cron clean /tmp job )
(this /tmp idea only worked a few times)

Later on and after trying to troubleshoot this, the X lookups started to happen more frequently
I would either remote ssh to the box and reboot or use the reset button.

One day .. out of nothing the LCD just turned off as if it was DPMS that did it. However the CRT was still up.
I had to restart X to get both again.

In the middle of all this i decided to use just one monitor plugged and later on i re-did xorg to
use simple settings for just one monitor. The LCD.

Here is the final xorg config file: http://freethinker.ws/system/xorg.txt made by xorgconfig

The best i was able to get, was 3 days 'uptime' and this was after many other changes that i did and
that i will post bellow.

I also tried 2 Gentoo-sources.

2.6.23-r3 that i had for a while:
http://freethinker.ws/system/2.6.23-gentoo.kernel-r3.txt

and 2.6.23-r5 that i recompiled from fresh (not using oldconfig)
http://freethinker.ws/system/2.6.23-gentoo-r5.txt

My current Linux-headers: 2.6.23-r2

i also tried several nvidia-driver packages. The current stable one (100.14.19) some 9* series (
older) and one masked ( 100.14.23 )

The Xorg problem remained.

My nvidia is compiled this way:

[jcat]

Hi,



So what driver are you using for X now? "nv" or "vesa" or something?

Are you saying that without the nvidia driver you can run X and you don't suffer any freeze issues at all?

What are you using for framebuffer in the kernel config?

Also, can we see your

[HeXiLeD]

The emerge info is posted above after make.conf.

I do have framebuffer settings in the kernel but i dont run X without nvidia drivers.
I am just using the console.

emerge info:

[larand54]

[HeXiLeD]

What you are saying does make sense; but if the card is really broken for X, why does it freeze soon as i emerge the drivers, without even going to X.

It seems that it happens when the kernel 'feels' nvidia drivers being loaded.
_________________
443640
My UNSOLVED TOPICS
How to ask questions
Configs & Hardware SPECIFICATIONS

[larand54]

Hmm... I would suggest that you should remove all that nvidia-stuff that may freeze your system. Then re-emerge your tool-chain, your complete system using the emwrap-script.
Then you could try re-emerge your nvidia-stuff. This may take a while but it will give a more stable system.
If your problem still occure, well there must be a hardware problem or at best some misconfiguration of the kernel.
Why not try a compiled kernel from a working live-cd? If that's not work then it could not be the kernel or...?

Best wishes.

[yfh2]

Hexiled,

This quick one just to let you know you are not alone !
I've had exactly the same problem here. (Nvidia card is a 7300)

I was with kernel 2.6.17, and old NVidia drivers (possibly 7186, not sure now), and everything had been working fine for months ...

I had Alsa issue with the kernel, and for some obnoxious reason, it refused to compile. So I was forced to compile a newer kerner (tried, 2.6.19-r5, 2.6.22-r5, 2.6.23-r3), and installed the newer NVidia drivers.

And now, every now and then, I get a hard lockup. (and since I have no other machine, I can't ssh, and I screw my hard disk everytime because it gets unmounted the bad way).

No idea what to do now.

[HeXiLeD]

LONG POST BUT WITH POSITIVE CONCLUSIONS


First test
I decided to give a few more tries to fix this problem and emerged
vanilla-sources 2.6.23.9. It was compiled from fresh ( not from
oldconfig) and i removed most grafics support options .

These were all about vga vesa framebuffer and related. I only left a
couple options related to apm for the monitor and suport for monitor types.

after reboot i decided to leave these settings:

[HeXiLeD]

PROBLEM IS SOLVED!

So far the has been stable and i did push things to try to make xorg to hard lock.
It didnt happen !
This tells me that my main problem with all of this ... was due to xorg-server -r2 that some turkey thought to be amusing to "let them think it's stable"
just for fun. (lets just stick around this)

All of this makes lots of sense because i remember that the first time i started to see problems happening, was with mplayer and vlc with the image going green, out of nothing. after a while, X would lock.

New Test:

I decided to turn on both monitors to see if i had any lockups.

This is their xorg made with nvidia-settings:

[yfh2]

Hi,

As I said above, i had the exact same symtoms (green mplayer window, lockups, etc ...)
Since I am less skilled, I just played around with nvidia-drivers and kernels.

I ended up with the following working set :
kernel 2.6.22-gentoo-r5
xorg-server 1.3.0.0-r2
nvidia 96.43.01

4 days uptime now ...

[bpaddock]

[HeXiLeD]

After having solved the problem and ffor 15 days i had no issues, i wasnt happy with my results.
I did more testing and eventualy checked my hardware.
I found that my nvidia card coller broke and some plastic melted. I think i might have to do with the thermal gel there too.

At this time i was having display lookups in the console even without having xorg installed or nvidia.
However these lockups were now different. the system was not freezing. it was just the display.

I ended up by buying : http://www.nvidia.com/object/geforce_8800gt.html
x11-drivers/nvidia-drivers-100.14.19 and xorg-server-1.3.0.0-r4

other upgrades: bios to latest version; ram to 4 GB and cooling system
_________________
443640
My UNSOLVED TOPICS
How to ask questions
Configs & Hardware SPECIFICATIONS

[jcat]

[juantxorena]

Hello. I have a similar problem, but I'm using a CRT screen only. Like all of you, it works when booting from knoppix, and now I'm recompiling xorg with nv driver, so I can use the computer until I have time to fix it (not now). I'm writing this from a console using links, but the strange thing is that I'm using a framebuffer driver (uvesafb), and I can see the images and use the mouse without freezings or locks, something that sounds quite stressful for the graphic card.

But I have bad news (bad news for me): I have the same problem when booting windows in the same computer, so it might be a hardware problem. The other thing that annoys me is that I haven't done anything suspectful, I haven't updatde kernel, nvidia drivers, haven't open the case, haven't change any X configuration. The only thing is that I have encoded some long videos, but I don't think that is anything related.

So, I have read here that the OP have done some tests for the graphic card,, but I have not found them, and I don't know how can I test it. Any help before buying a new card?

Thanks in advance.
_________________
I cannot write English very well. Please, correct any mistake so that I can improve.

[bunder]

[pathfinder]

i had stability issues for a while

my card was finally the problem (hardware).

But, I learned a lot reading the linux stability guide part 2 from d. robbins.

And i also found out that if you install the game neverball, the system crashed almost immediately. This one STRESSED so much the system it was THE test.

cat /proc/acpi/--thermal.../temperature is also your friend: you check the temperature, you run neverball, if it crashes inmendiately, it is not temp related.

cat /proc/interrupts tells you whether you have IRQ conflicts, and sometimes plugging the other devices in other slots will help your system to be stable.
Another thing: I still use an old kernel (vulnerable to vmslice), but as no one is going to force my home to get my computer, and I don t think anyone would be able to use this vulnerability for MY computer: it s more probable a common thief would steal everything, without being able to crack the basci computer defenses, and then would just format the computer. The data is always easy to recover, unless you encrypt everything (I m not covinced about this at this point), and there are still guys able to do such a thing (and they will DEFINITELY not be interested in my computer's data)

I ll keep all the vesa and framebuffer things.
But I tried with other older kernels and same lockups.

I also upgraded the BIOS and turned on and of some AGP settings, modified the nvidia drivers to force an AGP speed.

Are your cards AGP or PCIe?
Because I read that with AGP there are also Voltage possible incompatibilities.


EDIT: for the mouse, in the gentoo handbook, there s something about cat /proc/mouse or cat /proc/mice or something like this, then if moving your mouse generates a lot of weird output, this is the correct output of the driver.