Nvidia module not loaded; falling back to Nouveau

[Tim Evans]

Linux osprey 4.19.10-300.fc29.x86_64 #1 SMP Mon Dec 17 15:34:44 UTC 2018

Noticed this message while booting this weekend, although I had thought 
the Nvidia driver was being used.

# rpm -qa | grep -i nvidia
nvidia-settings-410.78-1.fc29.x86_64
xorg-x11-drv-nvidia-410.78-1.fc29.x86_64
xorg-x11-drv-nvidia-libs-410.78-1.fc29.x86_64
nvidia-persistenced-410.78-1.fc29.x86_64
xorg-x11-drv-nvidia-cuda-libs-410.78-1.fc29.x86_64
xorg-x11-drv-nvidia-cuda-410.78-1.fc29.x86_64
akmod-nvidia-410.78-1.fc29.x86_64
xorg-x11-drv-nvidia-kmodsrc-410.78-1.fc29.x86_64

journalctl says, in part:

Dec 22 22:21:22 osprey /usr/libexec/gdm-x-session[3409]: Kernel command 
line: BOOT_IMAGE=/vmlinuz-4.19.10-300.fc29.x86_64 
root=/dev/mapper/fedora_newosprey-root ro 
rd.lvm.lv=fedora_newosprey/root rd.lvm.lv=fedora_newosprey/swap 
pci=noaer rhgb quiet LANG=en_US.UTF-8 rd.driver.blacklist=nouveau 
modprobe.blacklist=nouveau nvidia-drm.modeset=1
Dec 22 22:21:24 osprey /usr/libexec/gdm-x-session[3409]: (II) 
config/udev: Adding input device HDA NVidia HDMI/DP,pcm=3 
(/dev/input/event16)
Dec 22 22:21:24 osprey /usr/libexec/gdm-x-session[3409]: (II) 
config/udev: Adding input device HDA NVidia HDMI/DP,pcm=7 
(/dev/input/event17)
Dec 22 22:21:38 osprey nvidia-settings-user.desktop[3812]: ERROR: NVIDIA 
driver is not loaded

Is something missing? What else to look at?  Thanks.

[Samuel Sieb]

On 12/23/18 1:51 PM, Tim Evans wrote:
> Dec 22 22:21:38 osprey nvidia-settings-user.desktop[3812]: ERROR: NVIDIA 
> driver is not loaded

I don't see why you have the "falling back to Nouveau" in the subject 
since I don't see that in the logs you pasted.  All that log line says 
is that the nvidia-settings application can't talk to the NVidia driver 
for some reason.
Are there any other messages in the log about either nouveau or nvidia?
If you run "lsmod", is the nouveau or nvidia driver loaded?
If you run "lspci -v", what driver is managing the VGA controller?

[Tim Evans]

On 12/23/18 5:21 PM, Samuel Sieb wrote:
> On 12/23/18 1:51 PM, Tim Evans wrote:
> > Dec 22 22:21:38 osprey nvidia-settings-user.desktop[3812]: ERROR: 
> > NVIDIA driver is not loaded
>
> I don't see why you have the "falling back to Nouveau" in the subject 
> since I don't see that in the logs you pasted.  All that log line says 
> is that the nvidia-settings application can't talk to the NVidia driver 
> for some reason.
> Are there any other messages in the log about either nouveau or nvidia?
> If you run "lsmod", is the nouveau or nvidia driver loaded?
> If you run "lspci -v", what driver is managing the VGA controller?

Thanks for your reply.

Nouveau module is loaded, according to both 'lsmod' and 'lspci'.

Found this (different wording than I recall seeing at boot time):

Dec 22 11:32:21 osprey systemd[1]: Started Fallback to nouveau as nvidia 
did not load.
Dec 22 11:32:38 osprey /usr/libexec/gdm-x-session[1349]: Kernel command 
line: BOOT_IMAGE=/vmlinuz-4.19.10-300.fc29.x86_64 
root=/dev/mapper/fedora_newosprey-root ro 
rd.lvm.lv=fedora_newosprey/root rd.lvm.lv=fedora_newosprey/swap 
pci=noaer rhgb quiet LANG=en_US.UTF-8 rd.driver.blacklist=nouveau 
modprobe.blacklist=nouveau nvidia-drm.modeset=1

[Samuel Sieb]

On 12/23/18 2:40 PM, Tim Evans wrote:
> Nouveau module is loaded, according to both 'lsmod' and 'lspci'.
>
> Found this (different wording than I recall seeing at boot time):
>
> Dec 22 11:32:21 osprey systemd[1]: Started Fallback to nouveau as nvidia 
> did not load.
> Dec 22 11:32:38 osprey /usr/libexec/gdm-x-session[1349]: Kernel command 
> line: BOOT_IMAGE=/vmlinuz-4.19.10-300.fc29.x86_64 
> root=/dev/mapper/fedora_newosprey-root ro 
> rd.lvm.lv=fedora_newosprey/root rd.lvm.lv=fedora_newosprey/swap 
> pci=noaer rhgb quiet LANG=en_US.UTF-8 rd.driver.blacklist=nouveau 
> modprobe.blacklist=nouveau nvidia-drm.modeset=1

Ok, that's clear then.  Edit the boot command from grub and try adding 
"nomodeset single" at the end of the kernel line.  I'm not sure if you 
need to have a root password set to get in.  If you can't get to a 
command prompt, then use "3" instead of "single".  Then you can login as 
a user and sudo to root.  Once you have a root prompt, try "modprobe 
nvidia" and see what happens.

[Tim Evans]

On 12/23/18 6:25 PM, Samuel Sieb wrote:
> modprobe nvidia

# modprobe nvidia
modprobe: FATAL: Module nvidia not found in directory 
/lib/modules/4.19.10-300.fc29.x86_64

In fact, there are no files anywhere in /lib/modules/ named (or 
containing) "nvidia".

Yet, as noted previously:

# rpm -qa | grep -i nvidia
nvidia-settings-410.78-1.fc29.x86_64
xorg-x11-drv-nvidia-410.78-1.fc29.x86_64
xorg-x11-drv-nvidia-libs-410.78-1.fc29.x86_64
nvidia-persistenced-410.78-1.fc29.x86_64
xorg-x11-drv-nvidia-cuda-libs-410.78-1.fc29.x86_64
xorg-x11-drv-nvidia-cuda-410.78-1.fc29.x86_64
akmod-nvidia-410.78-1.fc29.x86_64
xorg-x11-drv-nvidia-kmodsrc-410.78-1.fc29.x86_64

Have I not installed all the required packages?

[John Pilkington]

On 24/12/2018 14:26, Tim Evans wrote:
> On 12/23/18 6:25 PM, Samuel Sieb wrote:
> > modprobe nvidia
>
> # modprobe nvidia
> modprobe: FATAL: Module nvidia not found in directory 
> /lib/modules/4.19.10-300.fc29.x86_64
>
> In fact, there are no files anywhere in /lib/modules/ named (or 
> containing) "nvidia".
>
> Yet, as noted previously:
>
> # rpm -qa | grep -i nvidia
> nvidia-settings-410.78-1.fc29.x86_64
> xorg-x11-drv-nvidia-410.78-1.fc29.x86_64
> xorg-x11-drv-nvidia-libs-410.78-1.fc29.x86_64
> nvidia-persistenced-410.78-1.fc29.x86_64
> xorg-x11-drv-nvidia-cuda-libs-410.78-1.fc29.x86_64
> xorg-x11-drv-nvidia-cuda-410.78-1.fc29.x86_64
> akmod-nvidia-410.78-1.fc29.x86_64
> xorg-x11-drv-nvidia-kmodsrc-410.78-1.fc29.x86_64
>
> Have I not installed all the required packages?

I don't know what is happening here, but something very like it happened 
to you around 11 Nov.  Then you said it the problem went away on 
reinstalling the nvidia drivers from rpmfusion, so that whatever nouveau 
had put in place was overwritten.  At that time I thought you were using 
a current nvidia card, and could use the current driver, but I don't 
recollect seeing an lspci from you:  I think I had confused your thread 
with one on the mythtv list.

https://rpmfusion.org/Howto/NVIDIA

If this *is* the problem, you need to allow time for the kmodule to be 
built.  And I have found that sometimes the normal graphical 
shutdown/reboot will hang when a new kernel has been installed.

HTH

John P

Hi.

On Mon, 24 Dec 2018 09:26:27 -0500 Tim Evans wrote:
> # modprobe nvidia
> modprobe: FATAL: Module nvidia not found in directory 
> /lib/modules/4.19.10-300.fc29.x86_64
> Have I not installed all the required packages?

No, except perhaps akmods. The main one for the nvidia kernel module
is akmod-nvidia that you have.

Install the akmods RPM if you don't have it, then try to rebuild the
kernel module with:

/usr/sbin/akmods --kernels $(uname -r)

or, for all the kernels:

/usr/sbin/akmods-shutdown

If this succeeds, the nidia kernel modules will be installed under
/lib/modules/ and as member of a RPM named:

kmod-nvidia-KERNEL_VERSION-NVIDIA_VERSION

Ex: kmod-nvidia-4.19.3-200.fc28.x86_64-410.78-1.fc28.x86_64

[Andrey Ponomarenko]

You are probably need an older version of drivers rather than 410: 304, 340 or 390. See https://rpmfusion.org/Howto/NVIDIA#Current_GeForce.2FQuadro.2FTesla or https://launchpad.net/~graphics-drivers/+archive/ubuntu/ppa.

Please provide more logs for debugging by:

sudo dnf install hw-probe
    sudo hw-probe -all -upload

I'll take a look.

24.12.2018, 17:27, "Tim Evans" :
> On 12/23/18 6:25 PM, Samuel Sieb wrote:
> > modprobe nvidia
>
> # modprobe nvidia
> modprobe: FATAL: Module nvidia not found in directory
> /lib/modules/4.19.10-300.fc29.x86_64
>
> In fact, there are no files anywhere in /lib/modules/ named (or
> containing) "nvidia".
>
> Yet, as noted previously:
>
> # rpm -qa | grep -i nvidia
> nvidia-settings-410.78-1.fc29.x86_64
> xorg-x11-drv-nvidia-410.78-1.fc29.x86_64
> xorg-x11-drv-nvidia-libs-410.78-1.fc29.x86_64
> nvidia-persistenced-410.78-1.fc29.x86_64
> xorg-x11-drv-nvidia-cuda-libs-410.78-1.fc29.x86_64
> xorg-x11-drv-nvidia-cuda-410.78-1.fc29.x86_64
> akmod-nvidia-410.78-1.fc29.x86_64
> xorg-x11-drv-nvidia-kmodsrc-410.78-1.fc29.x86_64
>
> Have I not installed all the required packages?

[Mr Brian Domenick]

Hi All,

I had a similar problem and it had to do with the kernel sources/headers
in my case. Fedora was running a later kernel and sources than the
latest version of stuff from rpmfusion for nvidia, once I put the
correct older version of sources/headers in place, I could force the
module to build and it all worked great from there.

Brian Domenick

On 12/24/18 10:05 PM, Andrey Ponomarenko wrote:
> You are probably need an older version of drivers rather than 410: 304, 340 or 390. See https://rpmfusion.org/Howto/NVIDIA#Current_GeForce.2FQuadro.2FTesla or https://launchpad.net/~graphics-drivers/+archive/ubuntu/ppa.
>
> Please provide more logs for debugging by:
>  sudo dnf install hw-probe
>  sudo hw-probe -all -upload
>
> I'll take a look.
>
> 24.12.2018, 17:27, "Tim Evans" :
> > On 12/23/18 6:25 PM, Samuel Sieb wrote:
> > > modprobe nvidia
> >
> > # modprobe nvidia
> > modprobe: FATAL: Module nvidia not found in directory
> > /lib/modules/4.19.10-300.fc29.x86_64
> >
> > In fact, there are no files anywhere in /lib/modules/ named (or
> > containing) "nvidia".
> >
> > Yet, as noted previously:
> >
> > # rpm -qa | grep -i nvidia
> > nvidia-settings-410.78-1.fc29.x86_64
> > xorg-x11-drv-nvidia-410.78-1.fc29.x86_64
> > xorg-x11-drv-nvidia-libs-410.78-1.fc29.x86_64
> > nvidia-persistenced-410.78-1.fc29.x86_64
> > xorg-x11-drv-nvidia-cuda-libs-410.78-1.fc29.x86_64
> > xorg-x11-drv-nvidia-cuda-410.78-1.fc29.x86_64
> > akmod-nvidia-410.78-1.fc29.x86_64
> > xorg-x11-drv-nvidia-kmodsrc-410.78-1.fc29.x86_64
> >
> > Have I not installed all the required packages?
>
> users mailing list -- [e-mail]
> To unsubscribe send an email to [e-mail]
> Fedora Code of Conduct: https://getfedora.org/code-of-conduct.html
> List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
> List Archives: https://lists.fedoraproject.org/archives/list/[e-mail]

[Tim Evans]

On 12/25/18 10:59 AM, Mr Brian Domenick wrote:
> Hi All,
>
> I had a similar problem and it had to do with the kernel sources/headers
> in my case. Fedora was running a later kernel and sources than the
> latest version of stuff from rpmfusion for nvidia, once I put the
> correct older version of sources/headers in place, I could force the
> module to build and it all worked great from there.

It appears my "Solved" message from yesterday morning did not make it to 
the list.  Here's what I wrote in response to [e-mail]
> Hi.
>
> On Mon, 24 Dec 2018 09:26:27 -0500 Tim Evans wrote:
> > # modprobe nvidia
> > modprobe: FATAL: Module nvidia not found in directory
> > /lib/modules/4.19.10-300.fc29.x86_64
> > Have I not installed all the required packages?
>
> No, except perhaps akmods. The main one for the nvidia kernel module
> is akmod-nvidia that you have.
>
> Install the akmods RPM if you don't have it, then try to rebuild the
> kernel module with:
> /usr/sbin/akmods --kernels $(uname -r)

Aha!  'akmods' was installed, but fails to run, reporting a missing 
dependency--'kernel-devel'! So, every time the kernel is updated, the 
build of the nvidia module fails (silently, as far as I can see).

Having installed that, akmods now happily builds the nvidia module and 
the system boots and uses it.

Thank you, Francis.

Hi.

On Tue, 25 Dec 2018 12:34:49 -0500 Tim Evans wrote:
> > No, except perhaps akmods. The main one for the nvidia kernel module
> > is akmod-nvidia that you have.
> Aha!  'akmods' was installed, but fails to run, reporting a missing 
> dependency--'kernel-devel'!

Old story :-( See: https://bugzilla.rpmfusion.org/show_bug.cgi?id=2569
> So, every time the kernel is updated, the build of the nvidia module
> fails (silently, as far as I can see).

There is perhaps a trace in the journal for [e-mail] or
in the logs under /var/cache/akmods
> Having installed that, akmods now happily builds the nvidia module and 
> the system boots and uses it.

Good.

I have also validated the akmods-shutdown service that recompiles (if
needed) all the kmods for all the kernels at shutdown. I don't care to
loose a few minutes when rebooting since I do that only every two or 3
weeks after a full update.

[Samuel Sieb]

On 12/24/18 6:26 AM, Tim Evans wrote:
> On 12/23/18 6:25 PM, Samuel Sieb wrote:
> > modprobe nvidia
>
> # modprobe nvidia
> modprobe: FATAL: Module nvidia not found in directory 
> /lib/modules/4.19.10-300.fc29.x86_64
>
> In fact, there are no files anywhere in /lib/modules/ named (or 
> containing) "nvidia".
>
> Yet, as noted previously:
>
> # rpm -qa | grep -i nvidia
> nvidia-settings-410.78-1.fc29.x86_64
> xorg-x11-drv-nvidia-410.78-1.fc29.x86_64
> xorg-x11-drv-nvidia-libs-410.78-1.fc29.x86_64
> nvidia-persistenced-410.78-1.fc29.x86_64
> xorg-x11-drv-nvidia-cuda-libs-410.78-1.fc29.x86_64
> xorg-x11-drv-nvidia-cuda-410.78-1.fc29.x86_64
> akmod-nvidia-410.78-1.fc29.x86_64
> xorg-x11-drv-nvidia-kmodsrc-410.78-1.fc29.x86_64

Try running "sudo akmods --force" and see what happens.

[Richard Shaw]

Better to ask on the RPM Fusion mailing list but the time I had that
problem my video card was no longer supported by the latest Nvidia drivers
so I had to use one of the compatibility packages, nvidia-390xx.

Thanks,
Richard