Arch Linux

chorlton

Member

Registered: 2015-04-07

Posts: 4

nvidia-xrun - card can't init - Possibly ACPI related

I have a 2019 Dell Inspiron 7000 (7580) with Optimus graphics (integrated Intel WhiskeyLake-U UHD Graphics 620 + dedicated NVIDIA GP108M [GeForce MX250 rev a1]). I'm running Arch 5.16.1-arch1-1 and i3 as my WM.

Relevant packages installed

aur/nvidia-xrun-git 0.3.79.gaf3b734-1 (+19 0.32) (Out-of-date: 2021-11-22) (Installed: 0.3.83.g270b6c0-1)
multilib/lib32-nvidia-utils 495.46-1 (24.9 MiB 116.2 MiB) (Installed)
extra/nvidia-utils 495.46-2 (171.8 MiB 436.9 MiB) (Installed)
extra/nvidia-settings 495.46-2 (775.9 KiB 1.6 MiB) (Installed)
extra/nvidia 495.46-10 (25.7 MiB 25.7 MiB) (Installed)
extra/libxnvctrl 495.46-2 (76.4 KiB 456.6 KiB) (Installed)
extra/libvdpau 1.4-1 (60.6 KiB 265.6 KiB) (Installed)
extra/egl-wayland 2:1.1.9+r3+g582b2d3-1 (29.7 KiB 81.2 KiB) (Installed)

To save power I want the NVIDIA graphics to be fully disabled most of the time and I'll run nvidia-xrun-git in a second TTY for gaming. Unfortunately the card is not able to be properly initialised after nvidia-xrun shuts it down.

The NVIDIA card appears to work; if I configure everything as per https://wiki.archlinux.org/title/NVIDIA … phics_only then I'm good. It's only after nvidia-xrun puts it to sleep that it doesn't come back. I'm following https://wiki.archlinux.org/title/NVIDIA … vidia-xrun

Xorg doesn't start because it can't find a compatible screen but I think this is due to the card not initialising when the nvidia driver loads. journalctl shows:

Jan 23 18:47:24 anya kernel: ACPI Warning: \_SB.PCI0.RP05.PEGP._DSM: Argument #4 type mismatch - Found [Buffer], ACP>
Jan 23 18:47:24 anya kernel: NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x24:0xffff:1433)
Jan 23 18:47:24 anya kernel: NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0
Jan 23 18:47:24 anya kernel: NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x24:0xffff:1433)
Jan 23 18:47:24 anya kernel: NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0

Xorg.1.log

nvidia-xrun provides a service which shut the card down on booting via various writes to sys/bus/pci/devices:

Removing Nvidia bus from the kernel
>>Dry run. Command: sudo tee /sys/bus/pci/devices/0000:01:00.0/remove <<<1
Enabling powersave for the PCIe controller
>>Dry run. Command: sudo tee /sys/bus/pci/devices/0000:00:1c.0/power/control <<<auto

where 0000:01:00.0 is the NVIDIA card and 000:00:1c.0 is the pci bus it's on.

When I want to use the card, nvidia-xrun attempts to wake it up and run startx.

>>Dry run. Command: sudo tee /sys/bus/pci/devices/0000:00:1c.0/power/control <<<on
>>Dry run. Command: sleep 1
>>Dry run. Command: sudo tee /sys/bus/pci/devices/0000:01:00.0/power/control <<<on
>>Dry run. Command: sudo modprobe nvidia
>>Dry run. Command: sudo modprobe nvidia_uvm
>>Dry run. Command: sudo modprobe nvidia_modeset
>>Dry run. Command: sudo modprobe nvidia_drm modeset=1
>>Dry run. Command: xinit /etc/X11/xinit/nvidia-xinitrc "/usr/bin/startx " -- :1 vt -nolisten tcp -br -config nvidia-xorg.conf -configdir nvidia-xorg.conf.d
>>Dry run. Command: sudo modprobe -r nvidia_drm
>>Dry run. Command: sudo modprobe -r nvidia_modeset
>>Dry run. Command: sudo modprobe -r nvidia_uvm
>>Dry run. Command: sudo modprobe -r nvidia
>>Dry run. Command: sudo tee /sys/bus/pci/devices/0000:01:00.0/remove <<<1
>>Dry run. Command: sudo tee /sys/bus/pci/devices/0000:00:1c.0/power/control <<<auto

I think the rm_init_adapter failed message in journalctl occurs on the call to xinit.

Much of my reading suggests there may be some kind of ACPI issue common with Dell laptops, but I've variously tried booting with kernel parameters including acpi_osi=!, acpi_osi='Windows 2009' (or 2017, 2019, 2011...), acpi_rev_override=1, acpi_rev_override=5, all to no avail.

Does anyone have experience of nvidia-xrun and/or this particular Dell laptop or similar? I've spent hours on this and have no hair left!

(Disclaimer - I had no hair when I started)