510.54 - Black screen with newer Ubuntu kernels?

I’ve been testing the latest (510.54) drivers with Ubuntu Jammy and found that they give me black screens at boot on many newer kernels, when I should be seeing terminal output (plymouth uninstalled/deactivated). DKMS installation showed no issues. I thought the system just hung with no output, though it may have been waiting for disk encryption password at boot in the non-visible terminal. Anyone else seeing the same problem?

5.15.0-17 = good
5.15.0-18 = black screen
5.15.0-22 = black screen
5.16.0-051600 (mainline) = good
5.17-rc7 [mainline] = black screen

I’m not sure what to make of this. We’ll all want to upgrade soon due to dirty pipeline, so I worry that the next release will be black screened for terminals (if not completely broken) as well. Prior to 5.15.0-18, I’d seen no problems.

This was all tested with an RTX 2070 super.

This has been an ubuntu specific issue which hit several people. The real cause is unknown to me since no logs available from failed boots but I suspect this might be caused by a broken initrd since it started when ubuntu pushed out a new initramfs-tools package. But it’s also more or less random whether it hits or not.

1 Like

Thanks for the confirmation. Sounds like something I should take up in Ubuntu’s bug launchpad if it persists.

Thanks again!

Maybe try if recreating (not updating) the initrd on affected kernels helps.
sudo update-initramfs -c -k <kernel name>
and beforehand check if the nvidia driver is available for that kernel
dkms status

I just rebooted into 5.17 kernel and can confirm: it’s just the terminals that are FUBAR. I blind-typed my encryption password, and the system launched my desktop as expected. I ran update-initramfs prior to this and got no errors - the system simply doesn’t appear to acknowledge terminals in the newer kernel. I just now hit some ctrl+alt+fkeys… and got greeted with mostly black screens and garbage.

I havent experienced any such issues on Kubuntu 22.04.

I have no idea but at a guess,

It looks like a confluence of the failing on the Users part of installing the complete and correct set of post install configurations after OS install and Driver install as per the official release README or instruction, Grub’s known inadequacies,Uefi, efimanager, ubuntu-drivers-common gpu manager service, GBM, GDM, Wayland , Xorg and the Gnome Devs atittude of " sure we could resolve it, . . . but any decent distro should already have the built in contingencies to resolve it and if no one asks we wont tell…

The information you’ve provided is lacking.
Basic System information? SystemWide Native Linux, Dual Boot, nVidia w/ Intel iGPU?
Pure Uefi or Compatibility Mode?
Kernel information? OEM? Generic?

Manufacturers Boot Logo must be [ disbaled ] in bios permanently.
FastBoot must be [ disbaled ] in bios permanently.
Grub time out must be at least 5.
These 3 things will never change. I believe Systemd boot relsoves this.

Plymouth shouldnt be uninstalled.
Plymouth and Splash should be disabled.
“nosplash noplymouth”

Make sure “nogpuservice” service isnt installed and the gpumanager.service is running.

The " the nvidia-drm" kernel parameter MUST be set for early loading.
This isnt a choice anymore with wayland especially with GDM3.
" nvidia-drm.modeset=1" should be installed in /etc/default/grub

As per the Official nVidia-Driver-510_README

I also suggest adding:


options nvidia-drm modeset=1


to /etc/modprobe.d/nvidia.conf
and


MODULES=“nvidia nvidia_modeset nvidia_uvm nvidia_drm”
nvidia
nvidia_modeset
nvidia_uvm
nvidia_drm


to /etc/initramfs-tools/modules


MODULES=“nvidia nvidia_modeset nvidia_uvm nvidia_drm”
nvidia
nvidia_modeset
nvidia_uvm
nvidia_drm


to /etc/modules-load.d/modules.conf


update-initramfs -c -k all


I dont know if the above is right but I know its not wrong.

Is your system completely and correctly configured as per:
https://us.download.nvidia.com/XFree86/Linux-x86_64/510.54/README/index.html
https://wiki.archlinux.org/title/GRUB/Tips_and_tricks#Setting_the_framebuffer_resolution
https://wiki.archlinux.org/title/GRUB/Tips_and_tricks#Disable_framebuffer

I’ve be using Kubuntu 22.04 Jammy as a daily driver for the past 4 weeks with out issue.
As a sidenote it’s the best Linux Desktop OS experience I’ve have since using the Linux Desktop OS exclusively since 2014 and it’s in Alpha!

Source of kernel. Extenden Ubuntu /Canonical Repositories? Backports, Proposed? Canonical Proposed?
Type of kernel. OEM, Generic, Low-Latency,Generic, -hwe-20.04, -hwe-20.04-edge,

Re: 5.15.0-17 = good?
Where are you getting this from. There’s no 5.15.0-17 Kernel available that I can see?

Re:5.15.0-18 = black screen?
I have no issue at all with this kernel. Everything is functioning normally and optimally.

Re:5.15.0-22 = black screen?
I have no issue at all with this kernel. Everything is functioning normally and optimally.
What I have installed.
5.15.0-22 linux-generic
5.15.0-22 linux-generic-hwe-20.04
5.15.0-22 linux-headers-generic
5.15.0-22 linux-headers-generic-hwe-20.04
5.15.0-22 linux-image-generic
5.15.0-22 linux-image-generic-hwe-20.04
5.15.0-22 linux-source
5.15.0-22 linux-tools-generic
5.15.0-22 linux-tools-generic-hwe-20.04
5.15.0-22 nvidia-modprobe

The Ubuntu re-packaged nVidia-510-dkms-driver works perfectly regardless if the Linux-Objects,Modules,Sigs-nvidia blob is installed and configured or not.

I only have Linux-Objects,Modules,Sigs-nvidia blob installed aswell as a personal experiment.
I’ve have trouble finding any decent documentation from Ubuntu re: Linux-Objects,Modules,Sigs-nvidia blob
I havent found any conflict or benefit yet from late linking the nVidia blob from a system with no running desktop / OPenGl processes.

5.15.0-22 linux-modules-nvidia-510-5.15.0-22-generic
5.15.0-22 linux-modules-nvidia-510-generic-hwe-20.04
5.15.0-22 linux-objects-nvidia-510-5.15.0-22-generic
(linux-objects-nvidia isnt to be installed directly it is installed with linux-modules-nvidia)
5.15.0-22 linux-signatures-nvidia-5.15.0-22-generic

Installationh of Linux-Objects,Modules,Sigs-nvidia blob
apt-get install --reinstall --install-recomends linux-modules-nvidia-510-5.15.0-22-generic -y && #
apt-cache show -a linux-modules-nvidia-510-5.15.0-22-generic -y && #
How to invoke latelinking.
dpkg-reconfigure linux-modules-nvidia-510-5.15.0-22-generic -y && #
update-grub && update-initramfs -c -k all && update-pciids -y && #

Re:5.16.0-051600 (mainline) = good
Great I have no interest.I’ll take your word for it.

5.17-rc7 [mainline] = black screen
Well… It’s not to be unexpected along with a host of other issues.

Extended Sources . 22.04

https://launchpad.net/~albertomilone/+archive/ubuntu/u-d-c-release
Alberto is the Canonical nVidia Package Maintainer.

I usually will only use the Kisak PPAs. However KISAK hasnt provided 22.04 PPA yet.
I’ve had no issues using the oibaf ppa yet although I have in previous releases.

anyway thats my guess.

It’s not an Ubuntu specific issue. I met the same problem with kernel 5.16.14 in Arch Linux.

It may be related to a GNOME GDM issue

summary : - [nvidia][xorg] display hangs on boot LOGO

  • [nvidia][xorg] display hangs on boot LOGO due to race of gdm and nvidia
  • driver probe
    tags : added: gdm3
    If you read the last message in the Launchpad Bug tracker you will read the following.
  • " Hi Seb,
    FWIK, the fix from gdm is not mandatory.
    The gpu-manager (inside u-d-c) will handle it.

but if any flavor which doesn’t use u-d-c then gdm upstream also provide this commit

to fix this issue in GDM.

Says, if no one from a flavor (which doesn’t use u-d-c) reports this issue then we don’t hurry to SRU gdm patches."

The:
*local-display-factory: Stall startup until main graphics card is ready (ecbd9694) · Commits · GNOME / gdm · GitLab
" I don’t think that the “users’” point of view is interesting here. We have two issues:

  1. GDM needs to wait for the displays to be ready. Ideally, this should purely mean waiting for logind to tag the seat as graphical. But, we have a lot of other ways to achieve this using systemd unit ordering and such.
  2. GDM needs to adjust its configuration for nvidia integration.

These two issues are independent of each other. In the second case, we are dealing with something GDM specific to adjust to nvidia. In the first case, we are dealing with an nvidia peculiarity that is not GDM specific but affects all display managers."

Howdy, this lovely bug also seems to also be affecting Ubuntu Server. If I plug in directly to my server once I get past Grub I’m greeted with nothing but a black screen, no console. Since this is Ubuntu Server there is no gdm, wayland, x, etc.

I’m on 5.15.0-22-generic with the 510.54 drivers.