Can't boot after upgrading to 510 driver on Ubuntu 20.04

Graphics / Linux Linux
, , ,
1

Can’t boot into ubuntu 20.04 after apt-get installed nvidia-driver-510

syslog shows no matching GPU found (I have 1080ti):

Feb 06 08:47:21 shiy-desktop systemd[1]: Starting nvidia-powerd service…
Feb 06 08:47:21 shiy-desktop nvidia-persistenced[657]: Verbose syslog connection opened
Feb 06 08:47:21 shiy-desktop nvidia-persistenced[657]: Started (657)
Feb 06 08:47:21 shiy-desktop nvidia-persistenced[657]: device 0000:01:00.0 - registered
Feb 06 08:47:21 shiy-desktop /usr/bin/nvidia-powerd[656]: nvidia-powerd version:1.0(build 1)
Feb 06 08:47:22 shiy-desktop /usr/bin/nvidia-powerd[656]: No matching GPU found
Feb 06 08:47:22 shiy-desktop /usr/bin/nvidia-powerd[656]: Failed to initialize RM Client
Feb 06 08:47:22 shiy-desktop systemd[1]: nvidia-powerd.service: Main process exited, code=exited, status=1/FAILURE
Feb 06 08:47:22 shiy-desktop systemd[1]: nvidia-powerd.service: Failed with result ‘exit-code’.
Feb 06 08:47:22 shiy-desktop systemd[1]: Failed to start nvidia-powerd service.
Feb 06 08:47:22 shiy-desktop nvidia-persistenced[657]: device 0000:01:00.0 - persistence mode enabled.
Feb 06 08:47:22 shiy-desktop nvidia-persistenced[657]: device 0000:01:00.0 - NUMA memory onlined.
Feb 06 08:47:22 shiy-desktop nvidia-persistenced[657]: Local RPC services initialized
Feb 06 08:49:02 shiy-desktop systemd[1]: Starting nvidia-powerd service…
Feb 06 08:49:02 shiy-desktop /usr/bin/nvidia-powerd[1340]: nvidia-powerd version:1.0(build 1)
Feb 06 08:49:02 shiy-desktop /usr/bin/nvidia-powerd[1340]: No matching GPU found
Feb 06 08:49:02 shiy-desktop /usr/bin/nvidia-powerd[1340]: Failed to initialize RM Client
Feb 06 08:49:02 shiy-desktop systemd[1]: nvidia-powerd.service: Main process exited, code=exited, status=1/FAILURE
Feb 06 08:49:02 shiy-desktop systemd[1]: nvidia-powerd.service: Failed with result ‘exit-code’.
Feb 06 08:49:02 shiy-desktop systemd[1]: Failed to start nvidia-powerd service.
nvidia-bug-report.log.gz (362.5 KB)

2 Likes
2

journalctl -b -0:
Feb 06 08:47:20 shiy-desktop kernel: nvidia-nvlink: Nvlink Core is being initialized, major device number 509
Feb 06 08:47:20 shiy-desktop kernel: NVRM: loading NVIDIA UNIX x86_64 Kernel Module 510.39.01 Fri Dec 31 11:03:22 UTC 2021
Feb 06 08:47:20 shiy-desktop kernel: nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms 510.39.01 Fri Dec 31 10:52:52 UTC 2021
Feb 06 08:47:20 shiy-desktop kernel: [drm] [nvidia-drm] [GPU ID 0x00000100] Loading driver
Feb 06 08:47:20 shiy-desktop kernel: [drm] Initialized nvidia-drm 0.0.0 20160202 for 0000:01:00.0 on minor 0
Feb 06 08:47:20 shiy-desktop kernel: nvidia-uvm: Loaded the UVM driver, major device number 507.
Feb 06 08:47:21 shiy-desktop systemd[1]: Starting nvidia-powerd service…
Feb 06 08:47:21 shiy-desktop nvidia-persistenced[657]: Verbose syslog connection opened
Feb 06 08:47:21 shiy-desktop nvidia-persistenced[657]: Started (657)
Feb 06 08:47:21 shiy-desktop nvidia-persistenced[657]: device 0000:01:00.0 - registered
Feb 06 08:47:21 shiy-desktop /usr/bin/nvidia-powerd[656]: nvidia-powerd version:1.0(build 1)
Feb 06 08:47:22 shiy-desktop /usr/bin/nvidia-powerd[656]: No matching GPU found
Feb 06 08:47:22 shiy-desktop /usr/bin/nvidia-powerd[656]: Failed to initialize RM Client

All the logs show the driver loading fine.

This is just powerd (for power management) not finding a matching GPU, because it is meant for notebooks.
You can simply disable it - sudo systemctl disable nvidia-powerd.

So what exactly is the problem, as you say you can’t boot into the OS?

3

I don’t know what the problem is. I have included the full log.

when I boot, it gets stuck at the motherboard’s splash screen.

If I enable grub’s terminal, I could see a message NVIDIA Error something. but it was too quick to capture.

then after that error message, I could boot into the text terminal , then I had to manually start the desktop with https://linuxconfig.org/start-gui-from-command-line-on-ubuntu-20-04-focal-fossa

4

Ok, just to make sure… Does

ls -lah /etc/systemd/system/default.target
point to graphical.target?

Please post the output of:
dkpg -l | grep nvidia

Please also boot into the console and run
journalctl -b0 >journal.txt
and upload journal.txt here.

5

I have the same Problem.
Ubuntu_Linux Mint 20.2

Everything worked perfectly.
But, after the update the system does not start. Thank you.

Lenovo Legion 5 Pro 16ITH6H
Nvidia 3070.

6

I have the same problem, starting Friday evening on 2 of my workstations. And i did not install anything. This is weird
Both run Ubuntu20.04 (5.13 kernel).
One has a Nvidia rtx3090 and one has a Nvidia rtx3070. The driver version was 460.91.
What is weird is that my notebook has a rtx3050, the same OS and the same driver and has no problems up until now.

7

Looking at the logs @billconan provided, it seems the previous boots Xorg wasn’t started at all. Maybe it was trying to start a wayland session without having nvidia-drm.modeset=1 set.
Two options to try

  1. set kernel parameter/module option
  2. disable wayland https://linuxconfig.org/how-to-enable-disable-wayland-on-ubuntu-20-04-desktop
8

I had same problem this morning and fixed it by doing

sudo apt-get remove --purge '^nvidia-.*'
sudo apt install --install-recommends linux-generic
sudo apt install --install-recommends linux-modules-nvidia-510-generic
sudo ubuntu-drivers autoinstall

I was on 5.4.0-97 kernel and install-recommends linux-generic brought me to 5.4.0-99, maybe that was it.

4 Likes
10

@generix - I thought so too, but then I saw the nvidia-smi output:

    Process ID                        : 1411
        Type                          : G
        Name                          : /usr/lib/xorg/Xorg

If OP started the GUI with systemctl isolate graphical.target, it would have gone the regular way and use wayland, no?
Or am I missing something?

11

I’ve been getting the same issue.

I updated to the latest kernel on Ubuntu and suddenly I’m unable to boot. I’m getting stuck at

A job process is running with no time limit.

Interestingly enough my Manjaro machine is currently having the exact same issue with the latest update.

Before the update I was running 460 perfectly fine. Today I tried manually installing 460 and it seemed to work (minus the 32 bit packages being broken)

12

I have the same Problem. I’ll explain it a little better.
Lenovo Legion 5 Pro 16ITH6H
Nvidia 3070
Ubuntu_Linux Mint 20.2 and now 20.3
Kernel 5.13.0-28-generic

Everything worked perfectly whit the nvidia-driver-495.
But, after the update to nvidia-driver-510 the system does not start.
Why did you removed the nvidia-driver-495 from the repositories?
It worked fine with the 3070 card.
Scaling with nvidia-driver-470 doesn’t work. Nor Hi-DPI resolution.
nvidia-driver-470 does not have good support for the 3070 card.

I think the problem is that the new driver breaks something in the kernel, because if you log in with an older kernel from grub and if you reinstall the kernel then you can log in to your user. But I think something is lost along the way.
Also something is wrong with the MOK. It does not work correctly.

I think that if we pay for a card like the 3070 we are entitled to good support.
Thank you.

13

@Mart That was the current boot where billconan manually started the desktop. I was looking at the journalctl output from boot -1 and -2. There weren’t any Xorg log traces to be found. Kind of “educated guess” based on what I didn’t see…

14

@user166790:

If by “same problem” you mean that you couldn’t log into your system after upgrading from nvidia-driver-460 to nvidia-driver-510, and that this is the problem you fixed by entering those 4 lines of commands,

I ran those exact commands and it did not fix that problem for me. I’ve tried to do this upgrade a few times now, and I was again put into a login-screen loop.

I am running Ubuntu-MATE 20.04.3, kernel 5.4.0-99, Nvidia GPU Gigabyte RTX 3060, driver version 470.103.01.

I now know what to do when this upgrade attempt fails (in case it helps anyone else): from the login screen, enter CTRL-ALT-F2 to get into the console, enter username & password at the console commandline promt, and enter:
sudo apt install nvidia-driver-470 (or whichever driver-version you were using before trying to upgrade to 510), and reboot.

For now I’m sticking with driver 470. I’ve wasted enough time. I think solrak may be right, and that 510 is breaking something.

2 Likes
15

I have another feeling, is it possible that the problem is due to using an encrypted installation?

16

I have the same issue.
Cannot boot after installing driver 510.

System:

System:    Kernel: 5.13.0-28-generic x86_64 bits: 64 Desktop: MATE 1.26.0 Distro: Ubuntu 21.10 (Impish Indri) 
Graphics:  Device-1: NVIDIA GP104 [GeForce GTX 1070] driver: nvidia v: 470.103.01 
           Display: x11 server: X.Org 1.20.13 driver: loaded: nvidia resolution: 1: 1680x1050~60Hz 2: 2560x1440 
           OpenGL: renderer: NVIDIA GeForce GTX 1070/PCIe/SSE2 v: 4.6.0 NVIDIA 470.103.01
17

Same issue for me. I even tried to uninstall the 510 driver and install the 470 but it still doesn’t work.

Ubuntu 20.04.3 LTS
Kernel 5.13.0-28-generic
GPU Nvidia GeForce GTX 1650

Here my logs:
nvidia-bug-report.log.gz (258.5 KB)

18

Still fails to boot even after installing the xanmod kernel 16.8.1 first and then the nvidia 510 driver.

I could revert back to 470 without much trouble.
But I got a Houdini and 470 :/

19

I can say that my installation is not and has never been an encrypted one, so I don’t think that’s the problem with upgrading the nvidia driver to 510.

I think it’s 510 that has the problem - not that I know what the problem is.

But enough people can’t upgrade that you’d think by now nvidia would take some notice of this and bring out a corrected update to 510.

As was said earlier, we deserve better support, having paid (and waited, inmy case half-a-year) to get a decent GPU.

20

There is an interesting Linux Mint fourm thread about this issue.

It appears that there may be a conflict between driver 510 and the MATE desktop environment (used by both the Linux Mint OS and the Ubuntu MATE OS – at least I think it’s the same DE).

The upshot seems to be that if (window manager) Marco is set in Window Preferences to enable compositing, then installation of nvidia-driver-510 will not allow successful reboot.

Also, I notice that when I run (in Ubuntu-MATE) either of the ubuntu-drivers list or ubuntu-drivers devices commands, the first line returned is:

WARNING:root:_pkg_get_support nvidia-driver-510: package has invalid Support PBheader, cannot determine support level

Would anyone here be able to decipher this warning message, or know what it means?

UPDATE: The root cause of this bug has been patched. See this post on the Ubuntu-MATE forum, and indicate if it affects you.

1 Like
21

This problem is not for MATE.
This bug is happening with several desktops, mate, cinnamon, gnome and KDE. And in ubuntu and derivatives. Doesn’t seem like a problem with Mate.