nvidia works after build but not after reboot

[brokndodge]

ok... this is strange to me. I have successfully built and installed nvidia 7167 drivers. modified xorg.conf. startx. everything works. glxgrears tells me i'm getting 285fps (it's an old riva tnt2 pro card). "glxinfo | grep direct" says "direct rendering: Yes"

so everything is installed and working correctly. now I reboot to login through gdm but x fails saying:"(EE) NVIDIA(0): Failed to load the NVIDIA kernal Module!"

I type startx.. same thing
modprobe nvidia no errors
lsmod nvidia is listed
startx same thing

so i reran the installer, let it rebuild the kernal module, when it was finished I typed startx and everything is werking again... when I reboot it's back to the same thing. I've been through this every way I can think of and a few I've found on here but cannot figure it out.

help please.

[antidrugue]

Try the latest drivers instead:
http://www.nvidia.com/object/linux_d..._1.0-8756.html

And follow techieMoe how-to:
http://www.linuxforums.org/forum/lin...d-drivers.html

[brokndodge]

followed the howto exactly. the driver link you provided does not support my card (riva tnt2 pro) have to use an older one. i did find a slightly newer driver (7174) gonna try it.

[antidrugue]

the driver link you provided does not support my card (riva tnt2 pro) have to use an older one.

Oh, sorry didn't knew that.

Here are a few different installation methods. If you use the Debian default kernel, you can use Debian packages (I think Sarge uses driver version 1.0-7174):
http://home.comcast.net/~andrex/Debi...tallation.html

You can find every possible NVIDIA drivers version here:
http://www.nvidia.com/object/linux_display_archive.html

I see that 1.0-7174 is the latest to support the TNT2:
ftp://download.nvidia.com/XFree86/Li...174/README.txt

Starting with 1.0-7664, the TNT2 is considered "legacy":
ftp://download.nvidia.com/XFree86/Li...664/README.txt

[antidrugue]

Plus, don't forget that old drivers like the 1.0-7174 may not work very well with very recent kernel (2.6.15 and up).

[brokndodge]

Here are a few different installation methods. If you use the Debian default kernel, you can use Debian packages (I think Sarge uses driver version 1.0-7174):
http://home.comcast.net/~andrex/Debi...tallation.html

The above link and corresponding howto trashed my system beyond my ability to repair. I believe they are using some experimental code there, not sure. Xorg failed and wouldn't start back. Yes, I did follow the howto exactly. Eventually lead to a complete reinstall. Thats life.

The other links you provided I am very familiar with and have them burned onto my bookmark backup cd.

Well, now I'm working with a fresh install. Gonna try all over again.

[antidrugue]

The above link and corresponding howto trashed my system beyond my ability to repair. I believe they are using some experimental code there, not sure. Xorg failed and wouldn't start back. Yes, I did follow the howto exactly. Eventually lead to a complete reinstall. Thats life.

Those instructions have always been good to me. Never had a problem with them (didn't try the 4 different ways, but still).

Are you using Debian Sarge? Or Etch? Or Sid?

[brokndodge]

Are you using Debian Sarge? Or Etch? Or Sid?

etch

the driver is installed... just every time i have to reboot, i have to recompile the driver. so i won't reboot. it's linux right? won't have to reboot till the next upgrade in maybe 6 months.

thx for your help

[brokndodge]

ok, i was searching nvidia's linux forums and found a fellow that had the identical problem to mine on fc5. someone posted a fix, but the fix is for fc5. would someone please translate this into the debian equivalant for me?

I have solved the problem below, the problem is that at reboot the /dev/nvidia* links are non recreated because most probably udev daemon delete it, so the solution that i have founded is:

1) Create a file called /etc/udev/makedev.d/51-nvida.nodes
2) Write in the file following lines:
nvidiactl
nvidia0
nvidia1
nvidia2
nvidia3
nvidia4
nvidia5
nvidia6
nvidia7

once i am booted into the system, everything is fine, but maybe the kernal module is trying to load before the links are created?

here is the original thread

[brokndodge]

Got it fixed... udev wasn't creating the nvidia nodes in /dev. Apparently in latter driver versions nvidia uses an init.d script to create these nodes after udev runs. I couldn't find a copy of nvidias script so I wrote one to take it's place. This script is written per the Debian init.d quidelines found here.

Solution

#!/bin/sh
#
# nvidia.sh
# Script to create the nvidia nodes in /dev
#
# These nodes need to be present before xorg can load the kernal module
# Created by Mark C. West, 14 may 2006
# e-mail: brokndodge (at) yahoo.com
# License: BSD
#
#
echo -n "Setting up nVidia nodes 0..."
mknod /dev/nvidia0 c 195 0
echo -n "1..."
mknod /dev/nvidia1 c 195 1
echo -n "2..."
mknod /dev/nvidia2 c 195 2
echo -n "3..."
mknod /dev/nvidia3 c 195 3
echo -n "4..."
mknod /dev/nvidia4 c 195 4
echo -n "5..."
mknod /dev/nvidia5 c 195 5
echo -n "6..."
mknod /dev/nvidia6 c 195 6
echo -n "7..."
mknod /dev/nvidia7 c 195 7
echo -n "ctl..."
mknod /dev/nvidiactl c 195 255
echo "Done."

please tell me what you think and if there are any errors as you see it.

Those of you having the same problem I described in earlier posts may use this script at your discretion.

Login as root
Save the script as

/etc/init.d/nvidia.sh

then do

chmod 755 /etc/init.d/nvidia.sh
update-rc.d nvidia.sh start 21 2 3 4 5 .
reboot