X or nvidia_drv segmentation fault [Solved]

[vaettchen]

Not sure whether this is the right forum but I didn't find a more suitable one:

After the most recent upgrade, where kernel and nvidia-drivers were bumped from 2.6.37 to 2.6.39 on an x86_64 machine, I can't start the X server anymore. It gets a segmentation fault at address 0x240 after backtrace showing numerous lines with nvida_drv.so and the last before the segfault showing usr/bin/X.

Unfortunately I have no network access to the machine right now, so that I cannot copy over the output or the Xorg.0.log file (which does not show any EE entries). I'm on a trip and need my machine desparately (most of my work I can't do on the Windows PC I'm writing this here with) - any ideas how I can tackle the situation?

I have uninstalled nvidia-drivers already and reinstalled with the --empty --deep options, and I did a revdep-rebuild.

Thanks in advance for any help,
Rainer

[anonimo]

I have an error similar at yours and I have resolve it (thanks to Micia)
viewtopic.php?f=24&t=24536&p=136863#p136863
For me it was a problem to nvidia driver, I had to downgrade it with portage

1) Add sabayon overlay (as root)

Code: Select all

equo install layman git
layman -a sabayon

2)update your packages list of portage

Code: Select all

emerge --sync && layman -S

3)install a previous version of nvidia-driver

Code: Select all

equo install linux-sources
emerge '=nvidia-drivers-270.41.19::sabayon' '=nvidia-userspace-270.41.19::sabayon' '=nvidia-settings-270.41.19::sabayon'
equo rescue spmsync

4)Sure that selected drivers are nvidia

Code: Select all

eselect opengl set nvidia

and then reboot
if you resolve, don't forget to thank Micia!

[vaettchen]

Thanks a lot, anonimo!

I ran through all the steps and no problems during the process but now I get this error message

Code: Select all

FATAL: Error inserting nvidia (/lib/modules/2.6.29-sabayon/video/nvidia.ko): No such device

although the file does exist.

Would be grateful for advice how to overcome that situation!

Thanks,
Rainer

PS: I also read from the post anonimo was linking to that it is not a good idea to re-install Sabayon with the most recent image since the problem would not disappear. My Italian is far from perfect - is my understanding correct?

[micia]

In that thread the problem was solved with a downgrade of the package by using portage,
anonimo reported that the problem was there even after a complete reinstall, since he updated again the system, so your understanding is correct

To be sure about the consistency of your drivers, could you post the output of:

Code: Select all

uname -a
eselect opengl list
equo query installed nvidia-drivers


[gallienus]

I'm having the same problem with the nvidia 275 drivers. I followed the directions but when I got to emerge '=nvidia-drivers-270.41.19::sabayon' '=nvidia-userspace-270.41.19::sabayon' '=nvidia-settings-270.41.19::sabayon'


It gives me the error there are no ebuilds to satisfy "nvidia-drivers-270.41.19::sabayon". I checked for typos and I don't think I have any.



I was using kernel 2.6.35 and the 270.41.19 drivers worked well with it. Isn't there just a simple way I can roll back my nvidia drivers and nvidia settings to 270.41.19 to use with the 2.6.35 kernel?

Edit:

When I type the command emerge -search nvidia-drivers the only thing that comes up are the 275 drivers

[genfool]

Most nvidia issues can be solved with a few simple steps

Code: Select all

jokertu genfool # eselect opengl list
Available OpenGL implementations:
  [1]   ati
  [2]   nvidia *
  [3]   xorg-x11


If nvidia does not have the *

Code: Select all

eselect opengl set nvidia

If a kernel update is involved

Code: Select all

jokertu genfool # eselect kernel list
Available kernel symlink targets:
  [1]   linux-2.6.39-sabayon *

Code: Select all

jokertu genfool # uname -r
2.6.39-sabayon

we want to see that opengl is set to nvidia, and that we do not have a kernel mismatch between the booted kernel shown by uname -r, and the working kernel shown by the * in eselect kenel list.
Assuming these are all correct, lets look at the xorg.conf file.

Code: Select all

nano /etc/X11/xorg.conf

Here we are looking for this section, we want to see that nvidia is selected

Code: Select all

Section "Device"
    Identifier     "Device0"
    Driver         "nvidia"
    VendorName     "NVIDIA Corporation"
    BoardName      "GeForce 8400 GS"
EndSection

If all is set correctly, we now need to troubleshoot the problem.
First we need to know what model of nvidia card we have installed, so we know what drivers we need.

Code: Select all

jokertu genfool # lspci |grep VGA
02:00.0 VGA compatible controller: nVidia Corporation G86 [GeForce 8400 GS] (rev a1)

This tells me that I have a 8k series card, I know that a 6k series card and up, should work fine with the newest drivers.
If I had a 5k series card, a geforce 5900 or lower, I will need the legacy drivers for the kernel I am using, .39 kernel listed.
x11-drivers/nvidia-drivers-173.14.30#2.6.39-sabayon

Now that we know we have everything setup correctly, we know which driver we need, this should work.
Sometimes a driver may not get installed correctly, I sometimes in this case, prefer to purge the system of nvidia-drivers.

Code: Select all

equo remove nvidia-drivers

Code: Select all

equo install nvidia-drivers

If you need legacy drivers, equo install nvidia-drivers will not work, instead you would need for a .39 kernel in my example below, adjust your kernel as needed

Code: Select all

equo install x11-drivers/nvidia-drivers-173.14.30#2.6.39-sabayon

If after all of this, your nvidia still does not work, you can simply switch it to a nv driver,

Code: Select all

eselect opengl set xorg-x11

nano /etc/X11/xorg.conf

Change nvidia to nv or vesa, either will work and give you a working generic driver and x
At this point will be easier to get logs and more professional help then I can offer.

[gallienus]

My bad after posting that I did a search on rolling back nvidia drivers and found this link on the sabayon forum http://forum.sabayon.org/viewtopic.php?f=86&t=24532&p=136926&hilit=roll+back+nvidia+drivers#p136926. I rolled the drivers back to 173 and everything is working fine now. I'll just wait a while for different drivers for the 2.6.39.

I should have known better. Search first ask second.

[genfool]

Maybe the older drivers will work for you, I am also looking at that post, just concerned about it.
Nothing wrong with using older drivers if they work.
If you do not have nvidia-settings installed,

equo install nvidia-settings

At the very least, watch the gpu temp settings, if they are fine, use the older drivers.
If they rise to a unsafe level, some older drivers will not work with newer cards, you risk overheating and hosing a new card.
I should go back and see if you posted which model card you have. I did not see it when I glanced through the thread.

But running older drivers on newer cards, you are taking a risk, I dunno, maybe you need older drivers to begin with.

[gallienus]

No I didn't post the type of card I had but since you asked it's a ASUS EN7600GS SILENT/HTD/256M GeForce 7600GS 256MB 128-bit GDDR2 PCI Express x16 SLI Support Video Card(newegg order history is a wonderful thing). I don't know the original release date of the drivers but my newegg invoice is dated 3/18/2007. I don't think it was a cutting edge card when I got it so if I were a betting man I'd say the card is probably older than the drivers lol. Cost 96 bucks 6 years ago it's been running without a problem since I put it in. I'd say I got my money's worth out of it. Oh and it runs Compiz/xfce fairly well too I might add.

[vaettchen]

Thanks for all the advice, but I'm still stuck. After another equo update / upgrade yesterday, I have made sure that

- opengl is set to nvidia
- the right kernel is active
- my hardware (GeForce Go 7950 GTX) is certified for the driver I have
- the driver is actually the one it should be (275-09-02#2.6.39-sabayon)
- xorg.conf is set to nvidia
- nouveau is blacklisted

but still I get the segmentation fault at address 0x240.

I'm also not successful with other drivers: vesa gives me

Code: Select all

module ABI major version (8) doesn't match the server's version (10)

I have made sure that I have the latest fx86-video-vesa installed (2.3.0)

Trying nv after blacklisting nvidia and un-blacklisting nouveau, startx gives me this message (re-typing, still writing this from Windows):

Code: Select all

NV: The PCI device (my GeForce card) has a kernel module claiming it
NV: This driver cannot operate until it has been unloaded

lsmod shows that the nouveau driver is loaded - that is what should be, isn't it? With my limited knowledge, I don't find any other suspect.

I tried blacklisting both, nvidia and nouveau - than X starts, but the login screen doen't even know my usernames and passwords.

I would be so grateful for help, this is my main working machine!

Thanks,
Rainer