'modprobe nvidia' fails

jtgd · Posted: Sun Apr 17, 2005 3:39 am Post subject: 'modprobe nvidia' fails

Could there be a connection between 'modprobe nvidia' failing and the xserver saying 'Failed to initialize the NVIDIA kernel module'?
Can something be done about it?

CyD · Tux's lil' helper Joined: 23 Jul 2004 Posts: 117

yes, the modprobe failure is why x won't work. try remerging the driver, if you haven't already:

jtgd · Posted: Sun Apr 17, 2005 4:40 am Post subject:

CyD · Tux's lil' helper Joined: 23 Jul 2004 Posts: 117

so, you don't get any errors when compiling? and 2.6.11-gentoo-r3 is your current kernel? i mean, it's basically tellin you that the module you have isn't built for your kernel. when emerging nvidia-kernel, does it say it's compiling for the same kernel? 'uname -r' will give you the currently running kernel. doing a 'ls -l /usr/src' will show you if usr/src/linux is properly linked.

jtgd · Posted: Sun Apr 17, 2005 6:00 am Post subject:

jtgd · Posted: Sun Apr 17, 2005 6:28 am Post subject:

Maybe this will help...

Riftwing · Apprentice Joined: 06 Oct 2002 Posts: 293

The kernel and nvidia-kernel have to compiled with the SAME version of gcc. I'm guessing you updated gcc, re-emerged nvidia-kernel but didn't recompile your kernel.
_________________
Good, bad, I'm the guy with the gun. - Ash, Army of Darkness

i92guboj · Posted: Sun Apr 17, 2005 11:56 am Post subject:

xevz · Posted: Sun Apr 17, 2005 1:09 pm Post subject:

I got the same problem before, with any module...

The problem was that I compiled my kernel with SMP support (which I don't have) by mistake. After I discovered this, removed it and tried again it worked!

gerard27 · Posted: Sun Apr 17, 2005 1:46 pm Post subject:

Look in 'bugs' and search for nvidia.
You'll see that there are unresolved bugs for amd64

CyD · Tux's lil' helper Joined: 23 Jul 2004 Posts: 117

as everyone said, recompiling your kernel and booting to the new one will fix those unresolved symbols. do you use genkernel? it can be helpful for first timers. following the gentoo install docs will help as well.

jtgd · Posted: Sun Apr 17, 2005 4:50 pm Post subject:

Thanks to everyone for the advice.

Riftwing: This is a fresh install of 2005.0. If the install process changes the gcc version then OK, but I did nothing to change it.

6thpink: I will try recompiling the kernel.

xevz: I have dual Opteron so I must have SMP in there. I'm sure it was there for the kernel compile and nvidia.

Gerard: When you say "look in bugs", what do you mean? If there are bugs for nvidia this sounds rather ominous. You mean I can't run a Linux/amd64/nvidia combination?

CyD: I went by the install docs and I used 'make menuconfig'. Is it too late to run genkernel now? Will it see all the settings that I generated with menuconfig?

CyD · Tux's lil' helper Joined: 23 Jul 2004 Posts: 117

from the install docs:

jtgd · Posted: Sun Apr 17, 2005 7:06 pm Post subject:

I don't know what the coldplug/hotplug thing has to do with my issue, but I did it anyway.

I recomplied the kernel, setting SMP and everything else in menuconfig (it seems to have forgotten everything from my original build), then emerged nvidia-kernel nvidia-glx.

It made no difference. Same modprobe error, same startx error.

Here is the emerge output:

i92guboj · Posted: Sun Apr 17, 2005 9:58 pm Post subject:

You sure that you are using the correct kernel image? I've seen so many similar issues in the past related to this thing. Check that the /boot partition is mounted before you install the kernel with 'make && make install modules_install'.

Just to discard any posibility of this try 'uname -a' and check the compilation time shown, so you can make sure that the kernel image that is currently running is the last that you compiled.

You can try also 'dmesg' just after you do the modprobe, and see if it reports something related to the nvidia driver and the bottom of the output.

jtgd · Posted: Mon Apr 18, 2005 1:33 am Post subject: SOLVED

Indeed you were right, 6thpink. Thanks much for the clue.

The handbook seemed to suggest that one should not automount the /boot partition, and sure enough my kernel compiles were pointless, going into the dummy /boot mountpoint directory. Once I mounted /boot, correct behaviour followed.

Your make && make install modules_install line seemed to just make a /boot/vmlinuz file, which does not seem to be involved in booting. But I did the usual genkernel and that worked. At first it generated an unbootable kernel, and I resorted to the install CD, but it was a problem I had previously -- you have to enable low level SCSI for SATA to work.

So I have a KDE screen (after nvidia logo) and I am one happy boy. Thanks to all of you who gave advice.

i92guboj · Posted: Mon Apr 18, 2005 3:17 am Post subject: Re: SOLVED

Maedhros · Posted: Tue Sep 20, 2005 5:18 pm Post subject:

Some consolidation from all the threads I've just seen about this:

Make sure that your gcc version hasn't changed between compiling the kernel and nvidia-kernel
Make sure the kernel you're compiling against (/usr/src/linux) is the kernel you're actually running
If you already have an nvidia module loaded, rmmod it before modprobing the new one
When upgrading, explicitly specify nvidia-glx and nvidia-kernel as nvidia-glx won't necessarily update nvidia-kernel if they're both already installed.

_________________
No-one's more important than the earthworm.

i92guboj · Posted: Tue Sep 20, 2005 5:24 pm Post subject:

Add this: if you have to recompile the kernel due to a gcc version change between the kernel release that you are using and the new nvidia drivers, make sure that /boot is mounted when you install the new kernel. And then reboot and check the kernel date of compilation with uname -a before complaining. This is a frequent issue. :wink:

brekehan · n00b Joined: 20 Sep 2005 Posts: 3

fatshady · Posted: Mon Oct 10, 2005 8:42 am Post subject:

I had to remove nvidia framebuffer support from the kernel before I could get the nvidia module loaded.

Dunno if this will help?
_________________
How come monosyllabic has five syllables?

H4wk · n00b Joined: 09 Oct 2005 Posts: 16

i don't know why everyone gives weird solutions.
Looks like he has a non-working config to me.

Please post your xorg.conf (/etc/x11/) and your Xorg.0.log (/var/log/)

The error its giving is that nvidia can't find supporting screen resolutions (probably)

blice · Guru Joined: 12 Feb 2005 Posts: 598 Location: D- 26188

I wonder why nobody did a gentoo-nvidia wiki so far ?

I had some trouble too, searched through dozends of threads and found an answer by browsing the log&man -files ...

So I would do an wiki / howto next weekend (german/english), if there is interesst and maybe someone can help me translating to "correct" english, as my school-english is quite bad :lol:

Greets blice
_________________
Gentoo seit 2004.3
Skype: bodokhan
SecondLife: Bodokhan Fall
Email: blice@gmx.de

i92guboj · Posted: Mon Oct 10, 2005 4:15 pm Post subject:

H4wk · n00b Joined: 09 Oct 2005 Posts: 16

last time i had the nvidia module error, the "failed to initialize" came first, and the reason for failing directly after - exactly as it is now.

Let me try and simulate an nvidia crash and post a part of the log

edit: can't simulate a crash (very frustrating - i could always do this in windows)
What i CAN tell you, is that 'cannot initialize' does NOT mean 'cannot find'

The error says: "Hey, i cannot start this up because there's nothing for me to do. I don't have a compatible screen to render on"