[SOLVED] X freezes, NVIDIA(GPU-0): WAIT

[AndrewAmmerlaan]

X has been freezing a lot lately, almost everyday at least once. The screen just freezes completely, the cursor can move but the system does not respond to keyboard input (apart from SysRq, I can force it to reboot, but it won't switch tty even after SysRq+r).
Also the system does not seem to recover, because simply waiting doesn't seem to help.
Xorg.log shows me:

[dmpogo]

It does look like hardware/driver hang.

Can you ssh into the machine when it is hanged and reload nividia module ? (rmmod nvidia, modprobe nvidia)

[Frautoincnam]

I have a similar problem with a very similar machine, but I have no mouse cursor. X is totally dead.
This problem is recent (about 10 days). All worked nice with the same configuration during about 1 month.
I have nothing at all in logs.
I have 3 monitors : 2 (displayport & DVI) on Nvidia card, 1 (HDMI) on integrated CPU

I found that X totally freezes when (and only when, not each time but very often) I quit WoW (World of Warcraft started by wine).

I can ssh to the machine, but can't kill Wow.exe, nor X.
I always have to reboot.

BUT, if I "chvt 1" before leaving WoW, kill Wow.exe, then chvt 7, all works fine.

I've upgraded from stable version in portage to :
- app-emulation/wine-staging-4.0
- sys-kernel/gentoo-sources-4.19.19
- x11-drivers/nvidia-drivers-418.30
without any change.
_________________
OS: Gentoo KDE x86_64 (4.19.97-gentoo)
MB: Z370 GAMING PLUS (MS-7B61)
CPU: Coffeelake Intel(R) Core(TM) i5-9600K CPU @ 3.70GHz
GPU: NVIDIA MSI GeForce GTX 1050 TI 4GT LP & i915 CPU Integrated
SSD: Samsung SSD 860 250GB
RAM: G.Skill 32GB DDR4-2133

[AndrewAmmerlaan]

[AndrewAmmerlaan]

Today it happened again, while I was closing a full-screen video.

I ssh'd into it, and tried to rmmod nvidia, however nividia depends on nvidia_modeset which depends on nvidia_drm.
Rmmod nvidia_drm failed because 'module was in use', however when I forced it with the -f option it didn't complain. Nothing changed though, so I rmmod -f nivida_modeset as well, this caused the monitor connected to the nvidia gpu to turn black, however the other 2 monitors remained unresponsive (though I could still move the cursor). Next I did rmmod -f nvidia, this command refused to complete (it did not return the command prompt). In a new ssh session I tried to chvt 1 as Frautoincnam suggested, however that refused to complete as well. So instead I opened another ssh session and did shutdown -h now, however the system did not shutdown. The two remaining monitors kept their content, however the system stopped to respond to ssh at all, so it must have started the shutdown process anyway and froze somewhere in the middle.

The rc.log shows me that:

[Frautoincnam]

[AndrewAmmerlaan]

[Frautoincnam]

But it could be du to i915 too, which is in the kernel.

EDIT :
No, you're surely right, it's a nvidia module bug.
_________________
OS: Gentoo KDE x86_64 (4.19.97-gentoo)
MB: Z370 GAMING PLUS (MS-7B61)
CPU: Coffeelake Intel(R) Core(TM) i5-9600K CPU @ 3.70GHz
GPU: NVIDIA MSI GeForce GTX 1050 TI 4GT LP & i915 CPU Integrated
SSD: Samsung SSD 860 250GB
RAM: G.Skill 32GB DDR4-2133

[AndrewAmmerlaan]

As generix from the nvidia forums suggested I have switched the monitors around. The middle one (DVI-D) is now connected to intel instead of nvidia and the right one (DVI-D) is connected to nvidia instead of intel..
I am now testing this setup (so far so good, but who knows)

Also idk if you've been getting this annoying OSD popup asking you to select monitor configuration at every boot as well. But if you did, you can get rid of it by disabling the kscreen2 service in KDE settings.

[AndrewAmmerlaan]

X hasn't crashed for nearly four days, I guess the changes I made to the monitor configuration fixed this. I found this solution on the nvidia forums, I'll quote it here for future readers:

[Frautoincnam]