On 04/25/2013 01:03 AM, John Pilkington wrote:
> On 25/04/13 03:23, Joseph Areeda wrote:
>> Here I am again.
>>
>> I have my system working well with nouveau drivers but I need to install
>> the nvidia drivers because I do do some GPU work.
>>
>> I tried installing kmod-nvidia and nvidia-x11-drv from ELREPO but X
>> won't start.
>>
>> BTW - if this happens to you it is very helpful to set the boot level to
>> 3 so you can poke around.  Edit /etc/inittab and change the level in the
>> last line from 5 to 3.
>>
>> I reread the thread from a few weeks ago on this problem but didn't see
>> any resolution.
>>
>> The error messages that seem pertinent are:
>>
>>  From /var/log/Xorg.0.log.old (I'm now running nouveau drivers)
>>
>>     [    45.437] (II) LoadModule: "glx"
>>     [    45.449] (II) Loading
>>     /usr/lib64/xorg/modules/extensions/nvidia/libglx.so
>>     [    46.197] (II) Module glx: vendor="NVIDIA Corporation"
>>     [    46.197]    compiled for 4.0.2, module version = 1.0.0
>>     [    46.197]    Module class: X.Org Server Extension
>>     [    46.197] (II) NVIDIA GLX Module  310.44  Wed Mar 27 15:10:55 PDT
>>     2013
>>     [    46.197] Loading extension GLX
>>     [    46.197] (II) LoadModule: "nvidia"
>>     [    46.206] (II) Loading 
>> /usr/lib64/xorg/modules/drivers/nvidia_drv.so
>>     [    46.269] (II) Module nvidia: vendor="NVIDIA Corporation"
>>     [    46.269]    compiled for 4.0.2, module version = 1.0.0
>>     [    46.269]    Module class: X.Org Video Driver
>>     [    46.581] (EE) NVIDIA: Failed to load the NVIDIA kernel module.
>>     Please check your
>>     [    46.581] (EE) NVIDIA:     system's kernel log for additional
>>     error messages.
>>     [    46.581] (II) UnloadModule: "nvidia"
>>     [    46.581] (II) Unloading nvidia
>>     [    46.581] (EE) Failed to load module "nvidia" (module-specific
>>     error, 0)
>>     [    46.581] (EE) No drivers available.
>>
>> from /var/log/messages:
>>
>>     Apr 24 18:53:21 george kernel: nvidia: module license 'NVIDIA'
>>     taints kernel.
>>     Apr 24 18:53:21 george kernel: Disabling lock debugging due to
>>     kernel taint
>>     Apr 24 18:53:21 george kernel: NVRM: The NVIDIA probe routine was
>>     not called for 1 device(s).
>>     Apr 24 18:53:21 george kernel: NVRM: This can occur when a driver
>>     such as nouveau, rivafb,
>>     Apr 24 18:53:21 george kernel: NVRM: nvidiafb, or rivatv was loaded
>>     and obtained ownership of
>>     Apr 24 18:53:21 george kernel: NVRM: the NVIDIA device(s).
>>     Apr 24 18:53:21 george kernel: NVRM: Try unloading the conflicting
>>     kernel module (and/or
>>     Apr 24 18:53:21 george kernel: NVRM: reconfigure your kernel without
>>     the conflicting
>>     Apr 24 18:53:21 george kernel: NVRM: driver(s)), then try loading
>>     the NVIDIA kernel module
>>     Apr 24 18:53:21 george kernel: NVRM: again.
>>     Apr 24 18:53:21 george kernel: NVRM: No NVIDIA graphics adapter 
>> probed!
>>
>> Checking /etc/modprobe.d (with the nvidia drivers installed) showed
>> those modules were either in bloacklist.conf or blacklist-nouveau.conf.
>>
>> I also tried
>>
>>     dracut -f /boot/initramfs-$(uname -r).img $(uname -r)
>>
>> which has worked in the past when blacklist wasn't really blacklisting.
>>
>> I'm not sure I will be running 6.4 from now on.  It has not been tested
>> with the rpms from our group.
>>
>> I'd appreciate any suggestions.
>>
>> Joe
>>
> I've just had what looks like a similar problem on Fedora 17, solved 
> by editing the 'files' section of xorg.conf.  Modules sourced by X.org 
> were being loaded instead of those from nvidia. Here's an extract from 
> the thread (304.64) on rpmfusion-users.
>
>>> Section "Files"
>>> #    ModulePath   "/usr/lib64/xorg/modules/extensions/nvidia"
>>>           ModulePath   "/usr/lib64/nvidia/xorg"
>>> #    ModulePath   "/usr/lib64/xorg/modules/extensions"
>>>     ModulePath   "/usr/lib64/xorg/modules"
>>>     FontPath     "/usr/share/fonts/default/Type1"
>>> EndSection
>>>
>>> as suggested here:
>>>
>>> http://forums.fedoraforum.org/showthread.php?t=286084
>
> John P
Thanks for the suggestion John but that doesn't seem to be my problem at 
least not exactly.

Reading the thread and following the suggestions did not solve my 
problem.  I believe that refers to a packaging issue with 304.64 and the 
current ELREPO version is 310.44.  /etc/X11/xorg.conf was using 
different paths and changing them to the ones suggested gave the same 
errors.

I don't seem to be able to go back 304 because I need the 32 bit drivers 
also and those seem to only be available for the current version.  I 
wish I didn't need to install 32 bit drivers but that's not my program 
that uses them.

My next straw to grasp is to try the drivers from nvidia even though I 
hate to do it.  I had an issue a couple of years ago doing that on 
ubuntu.  Seems their packagers put files in a different location than 
nvidia.  So mixing the deb package and the run download ended up with 
mismatched dynamic libraries which was a whole lotta fun tracking down.  
I mention that to remind myself what a bad idea it is to mix the two.

I'll report back if that works.

Joe