Start Intel (OK); Switch NVIDIA (OK); Switch Intel (OK); no more switching: modprobe: FATAL: Module nvidia is in use.

Describe the bug Thanks for this project! I’m switching from bumblebee.

I can switch from Intel to NVIDIA and back to Intel once more, but the nvidia module doesn’t get unloaded, and optimus-manager won’t do anything more, complaining that the Xorg pre-start hook failed (logs below).

System info

• Arch Linux
• KDE Plasma
• SDDM
• optimus-manager 1.3
• /etc/optimus-manager/optimus-manager.conf (I think this is still as default):

[intel]
DRI=3
accel=
driver=modesetting
modeset=yes
tearfree=
[nvidia]
DPI=96
PAT=yes
allow_external_gpus=no
ignore_abi=no
modeset=yes
options=overclocking
[optimus]
auto_logout=yes
pci_power_control=no
pci_remove=no
pci_reset=no
startup_auto_battery_mode=intel
startup_auto_extpower_mode=nvidia
startup_mode=intel
switching=none

Logs

$ optimus-manager --status
ERROR: the latest GPU setup attempt failed at Xorg pre-start hook.
Log at /var/log/optimus-manager/switch/switch-20200621T142717.log

Cannot execute command because of previous errors.

Best of /var/log/optimus-manager/switch/switch-20200621T142717.log:

subprocess.CalledProcessError: Command '['bash', '-c', 'modprobe -r nvidia_drm nvidia_modeset nvidia_uvm nvidia']' returned non-zero exit status 1.
optimus_manager.bash.BashError: Failed to execute 'modprobe -r nvidia_drm nvidia_modeset nvidia_uvm nvidia' :
modprobe: FATAL: Module nvidia is in use.
optimus_manager.kernel.KernelSetupError: Cannot unload modules ['nvidia_drm', 'nvidia_modeset', 'nvidia_uvm', 'nvidia'] : Failed to execute 'modprobe -r nvidia_drm nvidia_modeset nvidia_uvm nvidia' :
modprobe: FATAL: Module nvidia is in use.

Full log

[10] INFO: # Xorg pre-start hook
[10] INFO: Previous state was: {'type': 'pending_pre_xorg_start', 'requested_mode': 'intel', 'current_mode': 'nvidia'}
[10] INFO: Requested mode is: intel
[10] INFO: Checking for GDM display servers
[890] INFO: Available modules: ['nouveau', 'bbswitch', 'nvidia', 'nvidia_drm', 'nvidia_modeset', 'nvidia_uvm']
[890] INFO: Unloading modules ['nvidia_drm', 'nvidia_modeset', 'nvidia_uvm', 'nvidia'] (if loaded)
[916] ERROR: Xorg pre-start setup error
Traceback (most recent call last):
  File "/usr/lib/python3.8/site-packages/optimus_manager/bash.py", line 11, in exec_bash
    out = subprocess.check_output(
  File "/usr/lib/python3.8/subprocess.py", line 411, in check_output
    return run(*popenargs, stdout=PIPE, timeout=timeout, check=True,
  File "/usr/lib/python3.8/subprocess.py", line 512, in run
    raise CalledProcessError(retcode, process.args,
subprocess.CalledProcessError: Command '['bash', '-c', 'modprobe -r nvidia_drm nvidia_modeset nvidia_uvm nvidia']' returned non-zero exit status 1.

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/usr/lib/python3.8/site-packages/optimus_manager/kernel.py", line 196, in _unload_modules
    exec_bash("modprobe -r " + " ".join(modules_to_unload))
  File "/usr/lib/python3.8/site-packages/optimus_manager/bash.py", line 18, in exec_bash
    raise BashError(
optimus_manager.bash.BashError: Failed to execute 'modprobe -r nvidia_drm nvidia_modeset nvidia_uvm nvidia' :
modprobe: FATAL: Module nvidia is in use.


During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/usr/lib/python3.8/site-packages/optimus_manager/hooks/pre_xorg_start.py", line 45, in main
    setup_kernel_state(config, prev_state, requested_mode)
  File "/usr/lib/python3.8/site-packages/optimus_manager/kernel.py", line 24, in setup_kernel_state
    _nvidia_down(config)
  File "/usr/lib/python3.8/site-packages/optimus_manager/kernel.py", line 65, in _nvidia_down
    _unload_nvidia_modules(available_modules)
  File "/usr/lib/python3.8/site-packages/optimus_manager/kernel.py", line 148, in _unload_nvidia_modules
    _unload_modules(available_modules, ["nvidia_drm", "nvidia_modeset", "nvidia_uvm", "nvidia"])
  File "/usr/lib/python3.8/site-packages/optimus_manager/kernel.py", line 198, in _unload_modules
    raise KernelSetupError("Cannot unload modules %s : %s" % (str(modules_to_unload), str(e)))
optimus_manager.kernel.KernelSetupError: Cannot unload modules ['nvidia_drm', 'nvidia_modeset', 'nvidia_uvm', 'nvidia'] : Failed to execute 'modprobe -r nvidia_drm nvidia_modeset nvidia_uvm nvidia' :
modprobe: FATAL: Module nvidia is in use.

[917] INFO: Removing /etc/X11/xorg.conf.d/10-optimus-manager.conf (if present)
[918] INFO: Writing state {'type': 'pre_xorg_start_failed', 'switch_id': '20200621T142717', 'requested_mode': 'intel'}

In this second Intel session:

$ lsmod | grep -i nvid
nvidia              20520960  0
ipmi_msghandler        77824  2 ipmi_devintf,nvidia
$ sudo modprobe -r nvidia
$ lsmod | grep -i nvid
$ pacman -Qqs nvidia
egl-wayland
lib32-libvdpau
lib32-nvidia-utils
lib32-opencl-nvidia
libvdpau
nvidia-dkms
nvidia-utils
opencl-nvidia

SDDM Configs

/etc/sddm.conf.d/20-optimus-manager.conf:

[X11]
DisplayCommand=/sbin/prime-offload
DisplayStopCommand=/sbin/prime-switch

/etc/sddm.conf.d/kde_settings.conf:

[Autologin]
Relogin=false
Session=
User=andy

[General]
HaltCommand=/usr/bin/systemctl poweroff
Numlock=on
RebootCommand=/usr/bin/systemctl reboot

[Theme]
Current=breeze
CursorTheme=breeze_cursors

[Users]
MaximumUid=60000
MinimumUid=1000

[X11]
ServerArguments=-dpi 0

/etc/sddm.conf:

[General]
InputMethod=

[Theme]
DisableAvatarsThreshold=7
EnableAvatars=true
FacesDir=/usr/share/sddm/faces
ThemeDir=/usr/share/sddm/themes

[Users]
DefaultPath=/bin:/usr/bin:/usr/local/bin
HideShells=
HideUsers=
RememberLastSession=true
RememberLastUser=true
ReuseSession=false

[Wayland]
EnableHiDPI=false
SessionCommand=/usr/share/sddm/scripts/wayland-session
SessionDir=/usr/share/wayland-sessions
SessionLogFile=.local/share/sddm/wayland-session.log

[X11]
# DisplayCommand=/usr/share/sddm/scripts/Xsetup
# DisplayStopCommand=/usr/share/sddm/scripts/Xstop
EnableHiDPI=false
MinimumVT=1
ServerPath=/usr/bin/X
SessionCommand=/usr/share/sddm/scripts/Xsession
SessionDir=/usr/share/xsessions
SessionLogFile=.local/share/sddm/xorg-session.log
UserAuthFile=.Xauthority
XauthPath=/usr/bin/xauth
XephyrPath=/usr/bin/Xephyr

[Askannz]

Late reply, but yeah, I know module unloading is wonky. It’s an issue that I can sporadically reproduce on my machine too. There must be some kind of race condition where the optimus-manager hook is called before X is done with the nvidia driver. It would explain why adding a simple lsmod is enough to hide the issue, since you’re basically adding a small delay before the module unloading.

I’ll look into putting a synchronization point to loop over modprobe -r with a delay, and try a few times before giving up. It’s kind of a hack though, the proper way would be to check which programs are still hanging to the GPU, but I don’t know how to do that.

[leledumbo]

Having the same problem, let me join the party. As an additional info, even after manually unloading nvidia module, it still refuses to switch to nvidia referring that there was a problem in last switching. Restarting the daemon and wait for some time (not sure how long, just for sure cannot be done immediately) allows the switching. Weird, eh?