[SOLVED - sort of] nvidia module causing problems with udev?

[Napalm Llama]

Hello! I'm returning to Gentoo after a long absence. New hardware, fresh install.

I'm getting a weird, intermittent problem, which appears to be related to the nvidia module somehow. It has three main symptoms:

• X doesn't start
  
• When restarting the system, "Stopping udev" takes an unusually long time
  
• The system hangs at "remounting filesystems readonly" and never actually reboots

Either all three of these things happen, or none of them happen and the system functions as expected. Sometimes I get an additional error at shutdown that the root filesystem can't be remounted readonly because it's in use.

There's nothing different in /var/log/Xorg.0.log when the problem occurs. The reason I think this is related to nvidia is that blacklisting the nvidia module from the kernel cmdline in grub seems to fix the problem (except still no X, of course). I think there must be some kind of race condition at boot, because it seems arbitrary whether the problem happens or not. I'm hoping someone can recognise the symptoms and point me in a useful direction...
_________________
Ryzen 5600x; Asus TUF Gaming B550-Plus; Geforce 1660 Super
Registered Linux User #381314
# killall humans

[alamahant]

Maybe an

[Napalm Llama]

I do apologise, like I said it's been a while!

[alamahant]

Do you have this as a kernel paeameter

[Napalm Llama]

I don't have that kernel parameter. I'll try it and see if it helps - thanks.
I've emerged everything, and have a fully functional system when it works. The problem started appearing after I set everything up.
My kernel is very well thankyou, how's yours? Not sure what you're asking me there!

The one thing that changed along with the problem appearing is I moved the system partition from my HDD to my NVMe drive. I updated fstab of course, and I can't think how it could be relevant - but as the problem appeared almost straight after doing it I thought I should mention it.
_________________
Ryzen 5600x; Asus TUF Gaming B550-Plus; Geforce 1660 Super
Registered Linux User #381314
# killall humans

[alamahant]

I meant is your kernel properly configured?
If you moved your partition
you should update
/etc/default/grub
rebuild your initrd if any
and run update grub
in addition to fstab which you already did.
Also is your kernel supprorting nvme?
Is your kernel supporting all file systems
?
This is what i meant "how is your kernel"

for starters plz use the above kernel parameter and blacklist nouveau.
I think the nvidia issue is different from your other issues.
_________________

[Napalm Llama]

Hmm, the nvidia-drm.modeset=1 parameter doesn't seem to help. As far as I know the kernel is correctly configured, although it's obviously in question. It has booted just fine before, though - like I said the problem is intermittent. I just configured and booted the latest 5.12 kernel, but it didn't help.

Regards nvme support, grub, filesystems etc - yes, all present and correct.

The only interesting thing from dmesg is this:

[alamahant]

[Napalm Llama]

[alamahant]

[Napalm Llama]

Rebuilding eudev appeared to help, ("udevadm settle" settled immediately), but only because I restarted the udev service after rebuilding. Rebooted and it was back to how it was.
linux-firmware is installed and up to date. I don't need to reinstall it for each new kernel, do I?
lspci | grep -i 3d finds nothing.
modprobe nvidia_drm just hangs...

[edit]
Kernel config - https://pastebin.com/sH1ycuwr
lspci -v - https://pastebin.com/1YeRvxSH

[Napalm Llama]

Hmm, the recent closeout comments on this long running bug suggest the nvidia driver no longer has anything to do with udev.
https://bugs.gentoo.org/454740

Curious then that when udev hangs, the nvidia-drm module won't load, but when udev works correctly nvidia-drm does load... I don't know udev well enough to investigate the connection, but maybe there's some leftover bad config sending udev astray?

Maybe one time in five I boot, and everything just works. Seems to be usually after I've changed something. But then when I reboot, I'm always back where I was...

[edit]
Just noticed one of my udevd threads is pegging one of my CPU cores at 100%! lsof says it has several kernel modules open, including nvidia.ko. Connection confirmed?

[more edit]
The CPU-eating udev process is unkillable, even with -9. This would explain:

• Why udev-related commands hang
  
• Why I can't rmmod nvidia (the module is in use by the udev process)
  
• Why the system can't remount filesystems readonly on shutdown (the unkillable udev process is still running)

If I boot with the nvidia module blacklisted in the kernel command line, I don't get this behaviour, but of course I can't have X then either so it's not really a solution.

[even more edit]
This person seems to have a related problem, which is also intermittent - ie. usually present, but sometimes not.
This thread might be a dupe of this one.
_________________
Ryzen 5600x; Asus TUF Gaming B550-Plus; Geforce 1660 Super
Registered Linux User #381314
# killall humans

[hotstoast]

I posted in the other thread last week. Wondering if anyone has found a solution for this yet.

I have all the same symptoms. The system will only shutdown if the uptime is ~ a day.

If you get to the bottom of any of this I'd greatly appreciate the solution you found

[iandoug]

Also struggling to set up new box.

Where is this setting?

nvidia-drm.modeset=1

nvidia-drivers compiled OK until I set things as per https://wiki.gentoo.org/wiki/NVIDIA/nvidia-drivers

and now it complains

[Ionen]

^ This is kind of off topic, this thread is about a udev+nvidia issue. It's better to make a new thread than hijack an old one, but I'll answer anyway.

[iandoug]

[Ionen]

[Napalm Llama]

For the sake of completeness - I updated a few things, and (fingers crossed) since then the problem has vanished.

Things I updated that might be relevant (giving the new versions that are now working):

x11-drivers/nvidia-drivers-470.63.01
gentoo-sources-5.13.13
sys-fs/eudev-3.2.10-r1

Still no idea what caused the problem, or why it went away unfortunately. I'll mark this thread as solved though.
_________________
Ryzen 5600x; Asus TUF Gaming B550-Plus; Geforce 1660 Super
Registered Linux User #381314
# killall humans