View previous topic :: View next topic |
Author |
Message |
dasPaul Apprentice
Joined: 14 Feb 2012 Posts: 243 Location: Dresden
|
Posted: Fri Aug 17, 2018 2:57 pm Post subject: remove nvidia module without reboot? [solved] |
|
|
Hi
My normal system configuration is:
Primary Intel GPU for desktop/X11
Secondary Nividia GPU for pci-passthrough/nvenc-encoding/OpenCl tasks
Per default I boot with vfio that grabs the nvidia card just before anything else would...
Then I could just normally start qemu with the nvidia card as pci-passthrough.
After I finished my gaming session in qemu I would like to use the nvidia card for
ffmpeg encoding tasks, so I unload vfio module and do my stuff with ffmpeg.
That somehow automatically loaded all the nvidia modules.
Now if I want to start another gaming-qemu session, how can I unload all the nvidia modules
and re-enable vfio?
I tried to unload nvidia_drm/nvidia_modeset/nvidia_uvm in any combination but it simply says:
Code: | modprobe -r nvidia
modprobe: FATAL: Module nvidia_drm is in use.
modprobe: FATAL: Error running remove command for nvidia |
I have no idea what else prevents the modules from unloading them...
Any ideas? _________________ -=human without Windows®=-
sorry for my bad english!
Last edited by dasPaul on Fri Aug 17, 2018 3:39 pm; edited 1 time in total |
|
Back to top |
|
|
Naib Watchman
Joined: 21 May 2004 Posts: 6051 Location: Removed by Neddy
|
Posted: Fri Aug 17, 2018 3:10 pm Post subject: |
|
|
Either there is still an Xorg or other session still running or not all the modules are unloaded
Quote: | $ lsmod
Module Size Used by
nct6775 53248 0
hwmon_vid 16384 1 nct6775
nvidia_drm 40960 6
nvidia_modeset 1060864 15 nvidia_drm
nvidia 13549568 737 nvidia_modeset
i2c_piix4 24576 0
k10temp 16384 0
input_leds 16384 0
|
from my output you can see there is a dependancy to the modules but nvidia_drm is the
Code: | alias nvidia="/etc/init.d/xdm stop && rmmod nvidia_drm && rmmod nvidia_modeset && rmmod nvidia && modprobe nvidia && /etc/init.d/xdm start" |
From a kernel module point of view, drm does need to be removed 1st (which you are doing).
So there must be an application still using the driver:
Code: | lsof /dev/nvidia0 | grep mem
X 3505 root mem CHR 195,0 1204 /dev/nvidia0
Discord 3934 Naib mem CHR 195,0 1204 /dev/nvidia0
steam 4948 Naib mem CHR 195,0 1204 /dev/nvidia0
steam 4952 Naib mem CHR 195,0 1204 /dev/nvidia0
steamwebh 4956 Naib mem CHR 195,0 1204 /dev/nvidia0
hon-x86_6 13930 Naib mem CHR 195,0 1204 /dev/nvidia0
hon-x86_6 13945 Naib mem CHR 195,0 1204 /dev/nvidia0
chrome 14154 Naib mem CHR 195,0 1204 /dev/nvidia0
chrome 14233 Naib mem CHR 195,0 1204 /dev/nvidia0
|
_________________
Quote: | Removed by Chiitoo |
|
|
Back to top |
|
|
dasPaul Apprentice
Joined: 14 Feb 2012 Posts: 243 Location: Dresden
|
Posted: Fri Aug 17, 2018 3:29 pm Post subject: |
|
|
Problem solved.
I just noticed that I accidentally uncommented the modules in my "/etc/modprobe.d/blacklist.conf"
So the nvidia modules got loaded from boot by whatever (even with enabled vfio that should prevent this if my understanding is correct) and they somehow dig deeper into the system, especially nvidia_drm...
If I do now my cuda/nvenc stuff it automatically loads only "nvidia" and nvidia_uvm, no drm... and those two I can simply unload. _________________ -=human without Windows®=-
sorry for my bad english! |
|
Back to top |
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|