struggeling with nvidia drivers [solved]

esromneb · n00b Joined: 20 Apr 2005 Posts: 37

So I've been searching around the net inorder to make glxgears run. I have a nvidia nx6800, and a k8n neo4 platinum mobo with a 64bit 3200. Anyways, I was getting the error:

Xlib: extension "GLX" missing on display ":0.0".
glxgears: Error: couldn't get an RGB, Double-buffered visual.

And I was able to fix it by running opengl-update xorg-x11. But my problem is that x won't start when I use
Driver "nvidia"
It will only run with
Driver "nv"

When I run glxgears I get about 700 fps, which is WAY below what I should be getting. How can I make x work with driver nvidia? Will this fix my low fps problem?
xorg.conf:

yaneurabeya · Veteran Joined: 13 May 2004 Posts: 1754 Location: Seattle

Sounds like nvidia-glx is missing. You might want to merge that if you don't already have it emerged.

esromneb · n00b Joined: 20 Apr 2005 Posts: 37

nvidia-glx is definatly emerged. What else can I do? Maybe I should check some logs or something?
-ben

tkrysak · n00b Joined: 23 Apr 2005 Posts: 11

How can I make x work with driver nvidia?

I should be able to help you here.. I'm running a Quadro 4 on an ~amd64
with the following nvidia packages installed, nvidia-glx-1.0.7174-r1/ nvidia-kernel-1.0.7174/ nvidia-settings-1.0.7167/

Will this fix my low fps problem? Absoloutely.

I take it your running an amd64 installation with the k8n, am I correct? I think what you've probably got is a couple of whacked
symbolic links (which is how the opengl-update selects which GL libraries you're using), several of the switching scripts like
opengl-update and motif-config, leave some things in a non-operational state sometimes, and you have to be careful of not
getting 32-bit and 64-bit stuff mixed up.

I suggest getting the 'nvidia' driver up 1st, then tackling the GL symbolic link issue after the nvidia driver is up and going.

1st you might try commenting out the "Load "glx"" line, and starting 'nvidia', if X comes up, you've just got whacked GL links,
easy to fix, but will take a post or two to iron it out

if not we need to look at /var/log/Xorg.0.log from a failed 'nvidia' run to see what's up,
it's a bit lengthy, but will usually tell you just why it died.

esromneb · n00b Joined: 20 Apr 2005 Posts: 37

One thing I didn't mention is that I have a 64 bit[code] processor, but
I'm running in 32 bit mode. Sorry!!!!

So within the last hour, I've managed to get the nvidia driver to work. I'm not sure what I changed, but now I can X starts with driver=nvidia. The problem is that when I run glxgears, it segfaults, and so does glxinfo. When X starts, it has an nvidia logo now. On top of that, I am pretty sure that my symlinks are whacky, I have used
opengl-update several times.

Should I copy your emerge'd packeges? I tried but:
[code]
emerge media-video/nvidia-glx-1.0.7174-r1
Calculating dependencies

!!! Problem in media-video/nvidia-glx-1.0.7174-r1 dependencies.
!!! "Specific key requires an operator
(media-video/nvidia-glx-1.0.7174-r1) (try adding an '=')" exceptions
[/code]
What do I do!? I'm in a worse situation that I was in before.
-ben

tkrysak · n00b Joined: 23 Apr 2005 Posts: 11

nvidia running = good,

when you emerge, don't use the version numbers, just 'emerge nvidia-glx' is what you want to get the latest stable
(or testing if you ask in /etc/make.conf)

even though your running 32-bit, most everything is the same, let's see what drivers/libraries you're using 1st

run the following command and post the output... ldd /usr/bin/glxgears

this will tell us what libaries glxgears wants to run, output should look close to this...

libGL.so.1 => /usr/lib/libGL.so.1 (0x00002aaaaabc2000)
libXp.so.6 => /usr/lib/libXp.so.6 (0x00002aaaaad62000)
libXext.so.6 => /usr/lib/libXext.so.6 (0x00002aaaaae6c000)
libX11.so.6 => /usr/lib/libX11.so.6 (0x00002aaaaaf7d000)
libpthread.so.0 => /lib/libpthread.so.0 (0x00002aaaab15e000)
libm.so.6 => /lib/libm.so.6 (0x00002aaaab2f4000)
libc.so.6 => /lib/libc.so.6 (0x00002aaaab47a000)
libGLcore.so.1 => /usr/lib/libGLcore.so.1 (0x00002aaaab69b000)
libnvidia-tls.so.1 => /usr/lib/tls/libnvidia-tls.so.1 (0x00002aaaabe80000)
libdl.so.2 => /lib/libdl.so.2 (0x00002aaaabf81000)
/lib64/ld-linux-x86-64.so.2 (0x00002aaaaaaab000)

then we want to look at all the GL files in whatever directory your libGL stuff is in (may be different than mine)
I would run "ls -l /usr/lib/*GL*" and my output looks like this:

-rw-r--r-- 1 root root 655 Apr 14 10:57 libGL.la
lrwxrwxrwx 1 root root 10 Apr 14 10:57 libGL.so -> libGL.so.1*
lrwxrwxrwx 1 root root 17 Apr 14 10:57 libGL.so.1 -> libGL.so.1.0.7174*
-rwxr-xr-x 1 root root 648488 Apr 14 10:57 libGL.so.1.0.7174*
-rw-r--r-- 1 root root 906774 Mar 19 23:19 libGLU.a
-rw-r--r-- 1 root root 754 Mar 19 23:19 libGLU.la
lrwxrwxrwx 1 root root 13 Mar 19 23:19 libGLU.so -> libGLU.so.1.3*
lrwxrwxrwx 1 root root 13 Mar 19 23:19 libGLU.so.1 -> libGLU.so.1.3*
-rwxr-xr-x 1 root root 624809 Mar 19 23:19 libGLU.so.1.3*
lrwxrwxrwx 1 root root 21 Apr 14 10:57 libGLcore.so.1 -> libGLcore.so.1.0.7174*
-rwxr-xr-x 1 root root 7093936 Apr 14 10:57 libGLcore.so.1.0.7174*
-rw-r--r-- 1 root root 46094 Mar 19 23:19 libGLw.a
lrwxrwxrwx 1 root root 13 Mar 19 23:19 libGLw.so -> libGLw.so.1.0*
lrwxrwxrwx 1 root root 13 Mar 19 23:19 libGLw.so.1 -> libGLw.so.1.0*
-rwxr-xr-x 1 root root 39859 Mar 19 23:19 libGLw.so.1.0*
lrwxrwxrwx 1 root root 13 Feb 4 13:03 libMesaGLU.so -> libGLU.so.1.3*

esromneb · n00b Joined: 20 Apr 2005 Posts: 37

ldd /usr/bin/glxgears:

tkrysak · n00b Joined: 23 Apr 2005 Posts: 11

need one more "ls-l /usr/lib32/opengl/nvidia/lib"

I want to compare sizes, dates, versions, etc.. against those in /usr/lib

I bet if we get your box using the GL libraries in /usr/lib (which at 1st blush look ok to me) your GL will be in business

/usr/lib32/opengl/nvidia/lib is where the libraries that are crashing are, (you can see that from the ldd listing), both libGL.so.1 and
libGLcore.so.1 are coming from there

I also might need to see the /var/log/Xorg.0.log file (I'm looking for the version number of the nvidia module, I'm looking for 7174,
same as what I see in /usr/lib...

it's getting late here, I'll check back tomorrow, and I bet we get you up and running...

and by the way, your welcome, hopefully a few others will learn a thing or two from the discussion

esromneb · n00b Joined: 20 Apr 2005 Posts: 37

here u go:

esromneb · n00b Joined: 20 Apr 2005 Posts: 37

And here is my Xorg.0.log:

yaneurabeya · Veteran Joined: 13 May 2004 Posts: 1754 Location: Seattle

From my experience dealing with the nvidia drivers (not the 64-bit ones... sorry just the 32-bit ones), symlinks to the glx related stuff is one of the main causes for pains with the driver overall. Most of the time the symlink is goofed up somehow...

After you get past the initial stuff it's usually smooth sailing though...

HydroSan · Posted: Mon Apr 25, 2005 7:55 am Post subject:

EDIT: Nevermind. I should read the post first!
_________________
I was a Gangster for Capitalism, by Major General Smedley Butler.

Server status: Currently down, being replaced with fresh install - 20% completed.

esromneb · n00b Joined: 20 Apr 2005 Posts: 37

So what do you advise me do to with the sym links? Delete them all? I wonder if this problem is affecing a large ammount of people?
-ben

esromneb · n00b Joined: 20 Apr 2005 Posts: 37

So I rebuilt my kernel, inorder to get my 2nd ethernet card to work, and x stopped booting. I then tried to boot with the kernel that WAS working before, and that didn't work either. This is from /var/log/xdm.log

tkrysak · n00b Joined: 23 Apr 2005 Posts: 11

your problem is that you kernel module is version 1.0.7174, I get this from your xorg log file...

(II) Loading /usr/lib/modules/drivers/nvidia_drv.so
(II) Module nvidia: vendor="NVIDIA Corporation"
compiled for 4.0.2, module version = 1.0.7174
Module class: XFree86 Video Driver

the libraries you have in /usr/lib32/opengl/nvidia/lib are version 1.0.6629 (from the ls -l)

nvidia wants the same versions, (like the ones in you /usr/lib directory), I'm not sure how you got here, but there are many ways
out of this. The 'correct way' is to 'emerge nvidia-glx', and it should get the 7174 version (assuming you haven't modified
/etc/portage/package.mask) After that /usr/lib32/opengl/nvidia/lib should have the 7174 version, and you should be off to the races.

But there is a quick and simpler method (but you're cheating portage here, so be careful) since you already have what appear
to be the correct libs in /usr/lib.

So if you like taking chances, you can copy all those nvidia libs to /usr/lib32/opengl/nvidia/lib using a clever 'tar' command...
do the following, as root of course..

cd /usr/lib
tar -cvf - libGL* libnvidia* | (cd /usr/lib32/opengl/nvidia/lib; tar -xvf -)
ldconfig

First you're switching to the directory that has the files
the tar command will create a tar file (-c), display the files it's taring on the console (-v) to stdout (-f -) of libGL* and libnvidia*,
this all gets piped to a shell (that's what the parentesis does), in that shell it chages to your target directory, and then exectues
a tar exctract (-x) to stdout (-f -)...whew...the above tar sequence can be a lifesaver when trying to copy directories, links, etc...

the ldconfig rebuilds the library hash table...and as long as you haven't whacked the 7174 module you were running,
(and you have the Load "glx" in the xorg.conf file), GL should be operational....

good luck

tkrysak · n00b Joined: 23 Apr 2005 Posts: 11

I just read that last post in more detail, when using 'nvidia' it requires a kernel module. EVERY time you rebuild your kernel, you MUST

emerge nvidia-kernel

X wouldn't boot because you changed your kernel and not your nvidia kernel module

esromneb · n00b Joined: 20 Apr 2005 Posts: 37

Ok, your tar command didn't work

. Any other ideas? I can post output from commands or whatever you need. And when must I emerge the kernel? After I recompile and boot? Or before or when? Thanks.
-ben

esromneb · n00b Joined: 20 Apr 2005 Posts: 37

P.S:

esromneb · n00b Joined: 20 Apr 2005 Posts: 37

I've come to realize that my x11 might be really fucked over. Should I just remerge x11 and nvidia stuff? How should I go about this? Thanks.
-ben

yaneurabeya · Veteran Joined: 13 May 2004 Posts: 1754 Location: Seattle

Unmerge nvidia stuff, delete the symlinks, reemerge the previously unmerged programs and then reboot for good measure or modprobe the nvidia kernel module.

esromneb · n00b Joined: 20 Apr 2005 Posts: 37

My filesystem went bad and I'm on another install. Same problem of course. Right now I can't even get the nvidia module to boot. Is it possible that emerging emul-linux-x86-nvidia would help? The problem is that I don't remember how I did it, because I was just messing around till I got it. Do I even need the emul libraries if I'm running in 32bit mode?
-ben

esromneb · n00b Joined: 20 Apr 2005 Posts: 37

yaneurabeya · Veteran Joined: 13 May 2004 Posts: 1754 Location: Seattle

*shrugs* Dunno cause I've never used the drivers since I have a 32-bit system.

esromneb · n00b Joined: 20 Apr 2005 Posts: 37

woot, so I got it to work. All I had to do was switch my kernel back to gentoo-sources. After I booted with the new kernel, I unemerged everything nvidia, then did:
ACCEPT_KEYWORDS="*" emerge emul-linux-x86-nvidia
nvidia-kernel and nvidia-glx are dependencies so they get automerged. I'm not sure if I acually needed the emu-linux, but I'm starting to think so. So basically mm sources don't work with nvidia. Heh, I guess I learned that the hard way. Thanks for all of you who posted here and tried to help me.
-ben