You are not logged in.
This is a multi-layered issue, so it will be a bit long.
I am currently trying to setup a VM with virt manager/QEMU, and I want to use GPU passthrough with my old NVIDIA GeForce 9800 GT. I installed nvidia-340xx-lts-dkms and nvidia-340xx-utils through AUR. After completing the installation, I rebooted the PC. X11 didn't open after I booted into the machine. I tried to manually open the X11 through TTY with `Xorg`, it gave me an error about that my GPU is too old for the current version of the X11. I didn't bothered with this error, as I will be using my integrated graphics card for the host machine. I setup the /etc/modprobe.d/vfio.conf as following:
options vfio-pci ids=10de:0614
softdep nvdia pre: vfio-pciAfter setting this up and running sudo mkinitcpio -p linux-lts and then rebooting, the end result is a X11 that runs, KDE Plasma with black screen on its login with a visible cursor (and can't login with it), a Wayland session that works but buggy beyond use, and a NVIDIA card that uses vfio-pci as its driver.
Specs:
CPU: AMD Ryzen 7 PRO 4750G with Radeon Graphics
GPU: NVIDIA GeForce 9800 GT
Motherboard: Asus PRIME A520M-K
BIOS Version: American Megatrends Inc. 3002, 02/24/2023
RAM: GeIL 8GB Evo Spear AMD Edition 3600mhz CL18 DDR4 Ram (GASB48GB3600C18BSC) x2
PSU: Corsair CV550 — 550 Watt 80 Plus Bronze
Case: Corsair Spec-05
Monitor: LG 24GL600F-B
Last edited by brogolem35 (2023-11-11 14:52:32)
Offline
For the problematic setup:
Please post your complete system journal for the boot:
sudo journalctl -b | curl -F 'file=@-' 0x0.stPlease post your Xorg log, https://wiki.archlinux.org/title/Xorg#General
Offline
For the problematic setup:
Please post your complete system journal for the boot:sudo journalctl -b | curl -F 'file=@-' 0x0.stPlease post your Xorg log, https://wiki.archlinux.org/title/Xorg#General
Journal:
http://0x0.st/Hthy.txt
/var/log/Xorg.0.log:
http://0x0.st/Hthw.txt
Offline
[ 10.823] (EE) modeset(0): eglGetDisplay() failed
[ 10.835] (II) AIGLX: Screen 0 is not DRI2 capable
[ 10.840] (II) IGLX: Loaded and initialized swrast
[ 10.840] (II) GLX: Initialized DRISWRAST GL provider for screen 0But that might or not be from an xwayland session? That runs into the same problem?
Kas 11 13:45:37 bronzgolem kwin_wayland[14038]: No backend specified, automatically choosing drm
Kas 11 13:45:37 bronzgolem kwin_wayland[14038]: kf.globalaccel.kglobalacceld: No desktop file found for service "flameshot.desktop"
Kas 11 13:45:37 bronzgolem kwin_wayland[14038]: kwin_scene_opengl: Creating the OpenGL rendering failed: "Missing one or more extensions between EGL_EXT_platform_base, EGL_MESA_platform_gbm, EGL_KHR_platform_gbm"
Kas 11 13:45:37 bronzgolem kwin_wayland[14038]: kwin_scene_opengl: Creating the OpenGL rendering failed: "Could not initialize egl"Kas 11 09:32:00 bronzgolem systemd-coredump[1485]: Process 1480 (plasmashell) of user 1000 dumped core.
Module libnvidia-glcore.so.340.108 without build-id.
Module libnvidia-tls.so.340.108 without build-id.
Module libGL.so.1 without build-id.
It's most likely because you've the nvidia-340xx stuff installed that iirc didn't have any proper libglvnd support, so it's getting in the way, mesa tries to operate on the wrong libGL and everything blows up.
=> remove the 340xx packages, check that "pacman -Qkk mesa" doesn't have broken links/files and try again.
Offline
[ 10.823] (EE) modeset(0): eglGetDisplay() failed [ 10.835] (II) AIGLX: Screen 0 is not DRI2 capable [ 10.840] (II) IGLX: Loaded and initialized swrast [ 10.840] (II) GLX: Initialized DRISWRAST GL provider for screen 0But that might or not be from an xwayland session? That runs into the same problem?
Kas 11 13:45:37 bronzgolem kwin_wayland[14038]: No backend specified, automatically choosing drm Kas 11 13:45:37 bronzgolem kwin_wayland[14038]: kf.globalaccel.kglobalacceld: No desktop file found for service "flameshot.desktop" Kas 11 13:45:37 bronzgolem kwin_wayland[14038]: kwin_scene_opengl: Creating the OpenGL rendering failed: "Missing one or more extensions between EGL_EXT_platform_base, EGL_MESA_platform_gbm, EGL_KHR_platform_gbm" Kas 11 13:45:37 bronzgolem kwin_wayland[14038]: kwin_scene_opengl: Creating the OpenGL rendering failed: "Could not initialize egl"Kas 11 09:32:00 bronzgolem systemd-coredump[1485]: Process 1480 (plasmashell) of user 1000 dumped core.
Module libnvidia-glcore.so.340.108 without build-id.
Module libnvidia-tls.so.340.108 without build-id.
Module libGL.so.1 without build-id.It's most likely because you've the nvidia-340xx stuff installed that iirc didn't have any proper libglvnd support, so it's getting in the way, mesa tries to operate on the wrong libGL and everything blows up.
=> remove the 340xx packages, check that "pacman -Qkk mesa" doesn't have broken links/files and try again.
I uninstalled both nvidia-340xx-lts-dkms and nvidia-340xx-utils, then ran `pacman -Qkk mesa`. I unfotunatelly don't have the output right now, as I rebooted the machine, but there were no warning or errors in the output. The only thing that was interesting in the output was that it said that 0 files were altered.
After rebooting, I reinstalled both nvidia-340xx-lts-dkms and nvidia-340xx-utils. While installing the nvidia-340xx-lts-dkms, there was an interesting warning.
(1/2) Arming ConditionNeedsUpdate...
(2/2) Install DKMS modules
==> dkms install --no-depmod nvidia/340.108 -k 6.1.62-1-lts
==> dkms install --no-depmod nvidia/340.108 -k 6.6.1-arch1-1
Error! Bad return status for module build on kernel: 6.6.1-arch1-1 (x86_64)
Consult /var/lib/dkms/nvidia/340.108/build/make.log for more information.
==> WARNING: `dkms install --no-depmod nvidia/340.108 -k 6.6.1-arch1-1' exited 10
==> depmod 6.1.62-1-lts/var/lib/dkms/nvidia/340.108/build/make.log:
http://0x0.st/HtFb.txt
I currently use the lts kernel but both the lts and the normal kernel is installed. May that be the cause of the issue?
Last edited by brogolem35 (2023-11-11 13:56:13)
Offline
After rebooting, I reinstalled both nvidia-340xx-lts-dkms and nvidia-340xx-utils.
Why?
You're going to forward the device, the drivers are no use but will screw your libGL installation.
Offline
After rebooting, I reinstalled both nvidia-340xx-lts-dkms and nvidia-340xx-utils.
Why?
You're going to forward the device, the drivers are no use but will screw your libGL installation.
Because without installing the drivers, I couldn't get the card to use vfio-pci as its driver.
Offline
That's nonsense - the entire point of vfio is to take away the device before the host start to operate it.
If you get the device under control of the nvidia kernel modules, you'd rather have to jump through some hoops to get it back for vfio.
Remove the nvidia drivers, keep them away, post an updated system journal (and xorg log)
Offline
I really didn't need the drivers after all. Swapping the
softdep nvdia pre: vfio-pciwith
softdep drm pre: vfio-pcidid the trick without having to install the drivers. My wholehearted thanks for your efforts! Now I need to go back to setting up my VM.
Offline
\o/
Please always remember to mark resolved threads by editing your initial posts subject - so others will know that there's no task left, but maybe a solution to find.
Thanks.
Offline