I’m running Tubleweed on my laptop which has Nvidia 1650M graphics card and just installed all available updates which also installed new kernel 6.11.0-1. I’m also running the Nvidia driver version 550.107.02-26.1.
The issue is after updating and rebooting, it does not boot to graphical mode. I don’t have the specific error from the Xorg log since I had to reboot and use a different kernel which overwrote the log. However I was just curious if anyone has seen the same?
So I was able to copy the log and get the error messages. Here’s the pertinent section of the log:
49.312] (II) NVIDIA dlloader X Driver 550.107.02 Wed Jul 24 23:27:47 UTC 2024
[ 49.312] (II) NVIDIA Unified Driver for all Supported NVIDIA GPUs
[ 49.312] (II) modesetting: Driver for Modesetting Kernel Drivers: kms
[ 49.312] (II) Loading sub module "fb"
[ 49.312] (II) LoadModule: "fb"
[ 49.312] (II) Module "fb" already built-in
[ 49.312] (II) Loading sub module "wfb"
[ 49.312] (II) LoadModule: "wfb"
[ 49.312] (II) Loading /usr/lib64/xorg/modules/libwfb.so
[ 49.312] (II) Module wfb: vendor="X.Org Foundation"
[ 49.312] compiled for 1.21.1.12, module version = 1.0.0
[ 49.312] ABI class: X.Org ANSI C Emulation, version 0.4
[ 49.324] (EE) NVIDIA: Failed to initialize the NVIDIA kernel module. Please see the
[ 49.324] (EE) NVIDIA: system's kernel log for additional error messages and
[ 49.324] (EE) NVIDIA: consult the NVIDIA README for details.
[ 49.324] (II) modeset(G0): using drv /dev/dri/card1
[ 49.326] (EE) NVIDIA: Failed to initialize the NVIDIA kernel module. Please see the
[ 49.326] (EE) NVIDIA: system's kernel log for additional error messages and
[ 49.326] (EE) NVIDIA: consult the NVIDIA README for details.
[ 49.326] (WW) Falling back to old probe method for modesetting
[ 49.327] (EE) open /dev/dri/card0: No such file or directory
[ 49.327] (EE) No devices detected.
[ 49.327] (EE)
Fatal server error:
[ 49.327] (EE) no screens found(EE)
[ 49.327] (EE)
Please consult the The X.Org Foundation support
at http://wiki.x.org
for help.
[ 49.327] (EE) Please also check the log file at "/home/david/.local/share/xorg/Xorg.0.log" for additional information.
[ 49.327] (EE)
[ 49.332] (EE) Server terminated with error (1). Closing log file.
You’re not alone. RTX3090 here, same thing.
The 560 drivers on other repos also have serious problems, you will get a graphical UI which does not function properly. I tried all of the available drivers from the nvidia and cuda repos, none worked. Snapper to the rescue, I’ve rolled back.
I’d recommend anyone with an nvidia GPU hold off on updates until things are sorted.
RTX 3060 Laptop here. After updating to kernel 6.11.0, KDE Plasma failed to show the desktop panel and app windows, while sddm showed random glitches. It’s a shame that only after something went wrong did it come to my mind to check the forum.
@ramdomPTM @pallaswept I use the run file here with the open driver, now I only use the Nvidia GPU as Prime Render Offload it’s working fine with the 6.11.0 kernel and 560.35.03 driver along with GNOME…
Thanks Malcolm. I’m guessing this is a packaging issue, since we haven’t heard widespread complaints of this kernel being incompatible with nvidia drivers.
I know there were problems with the automatic rebuilding of the kmp packages this week, which I believe are ongoing (none of my modules built, they had to be manually triggered). Perhaps the newly built packages aren’t even there yet. We don’t have visibility of these things, on nvidia’s repo.
It seems like a specific problem for only some/few users. No issues here so far with Nvidia-550.107.02 (G06) and kernel-default-6.11.0-1.1
Different machines with different GPU generations including Optimus setup with suse-prime installed fine and working after upgrade. The Optimus setup uses nearly the same GPU type like from the TO (NVIDIA TU117M [GeForce GTX 1650 Ti Mobile]) and works flawlessly.
Thanks Hui. Is that installed via package or the run file/hard way/etc? I couldn’t find a G06 package for this.
I thought I should try it, and collect logs, but I wasn’t game to install this given the wrong kernel version, seems doomed to failure:
nvidia-driver-G06-kmp-default 550.107.02_k6.10.5_1-26.1
None of the modules on the nvidia repo have been updated for 6.11
The open driver on the oss repo appears to have been updated though, but that’s not .107:
nvidia-open-driver-G06-signed-kmp-default 550.120_k6.11.0_1-1.2
That last one is probably the most recent released driver available, and also for the new kernel, but I’ve not tried these ones. Wish me luck…
Why should they have been updated? These packages contain sources that are compiled on your own system. There is no need to update the packages themselves as long as sources compile successfully for new kernel. Do you have any evidences that compilation fails?
Of course. This package contains the binary kernel module that has to match the new kernel version and has to be rebuilt for each new kernel.
Thanks for clearing up that there are two different types of kernel module package here. How might I distinguish one variety from the other, aside from prior knowledge?
Regardless, having tried that module, I had no luck, the .120 kernel package is there, but package dependencies were not resolvable with that.
I installed the 550.107 driver, the normal one from nvidia repos, cuda repos disabled. Just to cut down on variables. Nice and vanilla. Booted fine in 6.10.11.
Name : nvidia-driver-G06-kmp-default
Version : 550.107.02_k6.10.5_1-26.1
Vendor : obs://build.suse.de/Proprietary:X11:Drivers
zypper dup to 6.11.0 and got a bizarro-world distorted and semi-unresponsive plasma, and a desktop image, which is strange, as it is usually black.
I grabbed the full boot in journalctl, I didn’t see any obvious problem until kwin (wayland) tried to talk to the card and it’s just thousands of the same error:
Failed to create framebuffer: Invalid argument
Sep 28 17:01:42 Pallas kwin_wayland[14017]: No backend specified, automatically choosing drm
Sep 28 17:01:42 Pallas kwin_wayland_wrapper[14017]: MESA-LOADER: failed to open simpledrm: /usr/lib64/dri/simpledrm_dri.so: cannot open shared object file: No such file or directory (search paths /usr/lib64/dri, suffix _dri)
....
Sep 28 17:01:52 Pallas kactivitymanagerd[19258]: libEGL warning: MESA-LOADER: failed to open simpledrm: /usr/lib64/dri/simpledrm_dri.so: cannot open shared object file: No such file or directory (search paths /usr/lib64/dri, suffix _dri)
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_wayland_drm: Failed to create framebuffer: Invalid argument
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_wayland_drm: Failed to create framebuffer: Invalid argument
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_wayland_drm: Failed to create framebuffer: Invalid argument
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_wayland_drm: Failed to create framebuffer: Invalid argument
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_wayland_drm: Failed to create framebuffer: Invalid argument
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_wayland_drm: Failed to create framebuffer: Invalid argument
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_wayland_drm: Failed to create framebuffer: Invalid argument
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_wayland_drm: Failed to create framebuffer: Invalid argument
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_wayland_drm: Failed to create framebuffer: Invalid argument
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_wayland_drm: Failed to create framebuffer: Invalid argument
Sep 28 17:01:52 Pallas kactivitymanagerd[19258]: kf.windowsystem: virtual void KX11Extras::connectNotify(const QMetaMethod&) may only be used on X11
Sep 28 17:01:52 Pallas kactivitymanagerd[19258]: kf.windowsystem: virtual void KX11Extras::connectNotify(const QMetaMethod&) may only be used on X11
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_wayland_drm: Failed to create framebuffer: Invalid argument
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_scene_opengl: 0x500: GL_INVALID_ENUM error generated. Invalid <face>.
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_scene_opengl: Invalid framebuffer status: "GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT"
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_scene_opengl: 0x502: GL_INVALID_OPERATION error generated. <image> and <target> are incompatible
Sep 28 17:01:52 Pallas kernel: [drm] [nvidia-drm] [GPU ID 0x00000700] Framebuffer memory not appropriate for scanout
Sep 28 17:01:52 Pallas kernel: [drm] [nvidia-drm] [GPU ID 0x00000700] Framebuffer memory not appropriate for scanout
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_scene_opengl: Invalid framebuffer status: "GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT"
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_wayland_drm: Failed to create framebuffer: Invalid argument
Sep 28 17:01:52 Pallas kwin_wayland[14017]: kwin_wayland_drm: Failed to create framebuffer: Invalid argument
Sep 28 17:01:52 Pallas org_kde_powerdevil[19263]: libEGL warning: MESA-LOADER: failed to open simpledrm: /usr/lib64/dri/simpledrm_dri.so: cannot open shared object file: No such file or directory (search paths /usr/lib64/dri, suffix _dri)
Just a tiny slice of it 
My intention was to warn others so they might not find out the hard way, but now I am unsure if this is a bug, or a configuration problem on our end, or what. I feel like I should go file a bugzilla. Sorry if I’m in the wrong spot now.
I have the same issue (with 6.11 & 550.107 & RTX4080 & plasma & wayland). Instead of rolling back I booted with 6.10 now.
I have the following drivers installed:
I didn’t have any luck when I tried this earlier, sorry… but apparently it behaves differently for different machines, so, perhaps you get lucky 
I tried reinstall nvidia-drivers-G06 and nvidia-open from cuda repo, but no luck. Booting into kernel 6.10.11 and reinstalled nvidia-driver-G06-kmp-default from nvidia repo for Tumbleweed, and this time dracut showed that:
dracut-install: Failed to find module 'nvidia_drm'
dracut[E]: FAILED: /usr/lib/dracut/dracut-install -D /var/tmp/dracut.OZA8qz/initramfs -N ^i2o_scsi$ --kerneldir /lib/modules/6.10.11-1-default/ -m nvidia nvidia_drm nvidia-modeset nvidia-uvm
nvidia-drm.ko can be fount ad /lib/modules/6.11.0-1-default/updates but not in 6.10.11 directory. I suppose that kmp-default packges would only target the latest kernel. I wonder how can I re-compile nvidia kernel modules targeting kernel 6.10.11, or can I just copy those .ko files from the 6.11.0 directory?
The latter one does not work, which was expected. As for the previous one, I figured out that only kernel-default would save the previous version, not kernel-devel which is used to compile the driver. Sooo… out of luck here, I guess. 
No, I have 4 versions of both currently, so check that in /etc/zypp/zypp.conf there is a line reading:
multiversion = provides:multiversion(kernel)
otherwise there is something else wrong in your system.
But AFAIK the rebuild happens only for the running kernel, so to rebuild for 6.10.x you have to boot to 6.10.x and then force reinstall of the kmp package.
No issues here too, with Gnome, switcherooctl, prime render offload on Optimus setup.
Trying to form some kind of pattern between the failures and successes…Is it KDE-specific perhaps? Any successes with KDE? Any failures that aren’t KDE?
KDE Plasma user here.
I believe the pattern seems to be the mixing of packages from different repos: open kernel modules from openSUSE, Nvidia repo and CUDA repo.
I solely use the proprietary drivers as described in the SDB, and there are no issues for different GPUs and machines here.
Above, I show logs of the failure when using exclusively the stock setup as per that link.
@pallaswept can you show the output from inxi -GSaz