gpu-manager Error: can't open /lib/modules/

[Sam_]

At least once a day there is an unexpected desktop freeze incl. mouse and keyboard. Even magic-sys-rq isn't possible, only hard switch-off. From smartmontools I can exclude a hardware failure.

I've found that gpu-manager and systemd fail to load kernel modules.
This is among the output of journalctl | grep modules
Jan 10 13:12:34 systemd[1]: systemd-modules-load.service: Unit entered failed state.
Jan 10 13:12:34 systemd[1]: systemd-modules-load.service: Failed with result 'exit-code'.
Jan 10 13:12:49 gpu-manager[791]: Error: can't open /lib/modules/4.4.0-57-generic/updates/dkms
Jan 10 13:12:49 gpu-manager[791]: Error: can't open /lib/modules/4.4.0-57-generic/updates/dkms
(terminal-output attached)

dito with latest kernel

Jan 11 07:53:04 gpu-manager[786]: Error: can't open /lib/modules/4.4.0-59-generic/updates/dkms
Jan 11 07:53:04 gpu-manager[786]: Error: can't open /lib/modules/4.4.0-59-generic/updates/dkms

Please note answer #7
Please provide log files and outputs from following commands:
sudo systemctl status gpu-manager.service
dkms status

ProblemType: Bug
DistroRelease: Ubuntu 16.04
Package: ubuntu-drivers-common 1:0.4.17.2
ProcVersionSignature: Ubuntu 4.4.0-57.78-generic 4.4.35
Uname: Linux 4.4.0-57-generic x86_64
ApportVersion: 2.20.1-0ubuntu2.4
Architecture: amd64
CurrentDesktop: Unity
Date: Tue Jan 10 14:37:00 2017
InstallationDate: Installed on 2012-02-05 (1801 days ago)
InstallationMedia: Ubuntu 12.04 LTS "Precise Pangolin" - Alpha amd64 (20120204)
SourcePackage: ubuntu-drivers-common
UpgradeStatus: Upgraded to xenial on 2016-08-01 (162 days ago)

[Sam_ (and-sam)]

Close it for now, reason: Since change of HDMI-cable no freeze.

[Sam_ (and-sam)]

Reopen.
Few minutes ago complete freeze again. Nothing works except hard power off.
Every time there is chromium open and firefox closed.
Firefox freezes only itself, but not the whole system.

Yesterday, I stressed with two chromium profiles, one had 20 tabs open, the other watching videos. Result, no freeze.

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Gábor Lipták (gliptak)]

Description similar to https://bugs.launchpad.net/ubuntu/+source/ubuntu-drivers-common/+bug/1443364

Here is the output of sudo systemctl status gpu-manager.service

● gpu-manager.service - Detect the available GPUs and deal with any system changes
   Loaded: loaded (/lib/systemd/system/gpu-manager.service; enabled; vendor preset: ena
   Active: inactive (dead) since Sat 2017-02-04 10:57:42 EST; 20min ago
 Main PID: 1025 (code=exited, status=0/SUCCESS)

Feb 04 10:57:40 laptop1 systemd[1]: Starting Detect the available GPUs and deal with an
Feb 04 10:57:40 laptop1 gpu-manager[1025]: /etc/modprobe.d is not a file
Feb 04 10:57:40 laptop1 gpu-manager[1025]: /etc/modprobe.d is not a file
Feb 04 10:57:40 laptop1 gpu-manager[1025]: /etc/modprobe.d is not a file
Feb 04 10:57:40 laptop1 gpu-manager[1025]: /etc/modprobe.d is not a file
Feb 04 10:57:40 laptop1 gpu-manager[1025]: /etc/modprobe.d is not a file
Feb 04 10:57:40 laptop1 gpu-manager[1025]: Error: can't open /lib/modules/4.8.0-34-gene
Feb 04 10:57:40 laptop1 gpu-manager[1025]: Error: can't open /lib/modules/4.8.0-34-gene
Feb 04 10:57:42 laptop1 gpu-manager[1025]: update-alternatives: error: no alternatives
Feb 04 10:57:42 laptop1 systemd[1]: Started Detect the available GPUs and deal with any

[spu (steven-puttemans)]

I just ran into this critical issue, which has not been solved yet ... Ubuntu16.04

[Alberto Milone (albertomilone)]

That error is not really something that could ever cause the system to freeze, as it simply means that we cannot find a dkms module (and therefore gpu-manager can't get information about it). It should be more of a debugging message than an actual error.

Something else is probably failing in the kernel.

Can you attach your /var/log/gpu-manager.log , please?

Also, please attach your /var/log/kern.0.log after reproducing the problem.

[Sam_ (and-sam)]

Thanks for the advise. There is no kern.0.log, but kern.log.1, maybe you meant xorg.0.log (all attached)

Just had two complete freezes with an updated kernel linux-image-4.8.0-56-generic 4.8.0-56.61~16.04.1 amd64. Switch to console and REISUB didn't work thus hard shutdown. I've booted with 4.8.0-54 and no freeze since one hour.

[Sam_ (and-sam)]

Now complete freeze with updated 4.8.0-58-generic, hard shutdown, boot 4.8.0-56-generic, running now 30 minutes without freeze.
It happens so sporadic which makes it almost impossible to find the trigger.

[Peter Brandon (slowtrain55)]

I may have a similar problem in 17.04 (to which I just upgraded). When I try to play a video game, the computer completely freezes anywhere from 30 seconds to 40 minutes after I start. It's done so repeatedly and only when playing this game. I see very little by way of errors in my logs, but I do get errors about the gpu-manager and dkms like those reported here.

Oddly, when I run the following: sudo lshw -c video | grep driver
I get: configuration: driver=i915 latency=0

Which is a driver for my intel CPU's graphics processor. This is odd because I have a separate graphics chip, a radeon, installed, and radeon drivers installed but these do not show.

When I run: sudo systemctl status gpu-manager.service
I get the following, which is similar to what I see in my system log:

● gpu-manager.service - Detect the available GPUs and deal with any system changes
   Loaded: loaded (/lib/systemd/system/gpu-manager.service; enabled; vendor preset: enabled)
   Active: inactive (dead) since Sat 2017-07-29 23:01:32 EDT; 25min ago
 Main PID: 1156 (code=exited, status=0/SUCCESS)

Jul 29 23:01:30 ghostwheelLapHP systemd[1]: Starting Detect the available GPUs and deal with any system changes...
Jul 29 23:01:30 ghostwheelLapHP gpu-manager[1156]: /etc/modprobe.d is not a file
Jul 29 23:01:30 ghostwheelLapHP gpu-manager[1156]: /etc/modprobe.d is not a file
Jul 29 23:01:30 ghostwheelLapHP gpu-manager[1156]: /etc/modprobe.d is not a file
Jul 29 23:01:30 ghostwheelLapHP gpu-manager[1156]: /etc/modprobe.d is not a file
Jul 29 23:01:30 ghostwheelLapHP gpu-manager[1156]: /etc/modprobe.d is not a file
Jul 29 23:01:30 ghostwheelLapHP gpu-manager[1156]: Error: can't open /lib/modules/4.10.0-28-generic/updates/dkms
Jul 29 23:01:30 ghostwheelLapHP gpu-manager[1156]: update-alternatives: error: no alternatives for x86_64-linux-gnu_gfxcore_conf
Jul 29 23:01:32 ghostwheelLapHP systemd[1]: Started Detect the available GPUs and deal with any system changes.

[Peter Brandon (slowtrain55)]

I did some more experimentation. I did not get any system freezes when I shifted down to kernel 4.8 from 4.10, where I got a lot of freezes. In kernel 4.8 I get the same output when I run sudo systemctl status gpu-manager.service as I reported for kernel 4.10 above. This might indicate that this output isn't pointing to a system-crashing bug, but I'm not certain. In 4.8, the game periodically freezes but after a few seconds the screen flashes and suddenly it unfreezes. My guess is that the same bug that causes the freezing occurs in 4.8 but the system has some code that unfreezes the computer.

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Ayoub Boudhar (aneutron)]

I have a similar bug with the same premise, freezes using Google Chrome. They tend to be very random, I only have them during playback.

Google Chrome: 63.0.3239.132 (64bits) (Although numerous versions before had the same problem)
Ubuntu 17.10 (4.13.0-21-generic)
Running on a Pentium N3530, no discrete GPU.

The logs are almost identical.

[lupa (lupa18)]

Same here

[ALinuxUser (buntulongername-new)]

Cf. #1731651.

[Enrique Lafuente Diaz-Ripoll (piri-lafuente)]

the same for me :(

[Sam_ (and-sam)]

This is on 18.04.2:
Issue solved from my side by not using chromium any longer, since then no system freeze.
The error message in title remains, but obviously causing not a complete system freeze.

[Willem Hobers (whobers)]

seeing this on 18.04.3 xubuntu:

Aug 18 17:25:32 LAPTOP gpu-manager[806]: Error: can't open /lib/modules/5.0.0-25-generic/updates/dkms

Any more info I can provide?

[dagoulis (giannisdag)]

for me it stopped happened after changing the PSU (Power Unit)

[Alberto Milone (albertomilone)]

@Willem: can you post the output of this command, please?

dkms status

[Fellipe Weno (felco)]

I'm experiencing those hangs too. I have two different machines running different versions of Ubuntu.
One is a Ryzen with a NVIDIA card, using the proprietary drivers. The other is a Intel Coffee Lake, no HT and no discrete GPU.

In the first machine(Ryzen + NVIDIA) this started to happen more often after a upgrade to 19.04. On 18.04 it happened, in the whole year it ran, no more than a couple of times. Also, different of what the OP said, in my case even that the machine freezes, it comes back to life after a couple of minutes. But most of the time I just SSH to the machine, and kill Chrome or WINE, and it recover faster.

In the second machine(Intel + iGPU), it started to happen after a reinstall of the 18.04, it didn't happened before. But the behavior is different, in this machine the XWindow freezes and dies miserably, ending all processes running on it, and that is the machine I use to work... so pretty bad.

In terms of LOG, I - like others in this thread; couldn't find anything useful, in fact is hard to pinpoint the origin of the problem.

I really willing to help with this, but I need guidance to gather information on this. I just can't see what may be causing this issue, as I pointed out above, this is not related to what GPU we are using, as I have two completely different setups. Though it seems to be related to the graphical system.

[Door Attila (attiladoor)]

I have the same issue on ubuntu 20.04 with Ryzen 5 3600 + Nvidia GTX 1050ti. It freezes several times per a day. (kernel: 5.4.0-47). I tried using lightdm instead of gdm3 but it is the same.

[Door Attila (attiladoor)]

I have looked at the problem a bit closer and things is, lightdm works fine for a while, but eventually it has some kernel package problems:

sep 16 09:09:47 weird-stuff lightdm[1193]: gkr-pam: stashed password to try later in open session
sep 16 09:09:47 weird-stuff lightdm[1193]: gkr-pam: stashed password to try later in open session
sep 16 09:09:47 weird-stuff lightdm[1193]: pam_unix(lightdm:session): session opened for user attila by (uid=0)
sep 16 09:09:47 weird-stuff lightdm[1193]: gkr-pam: gnome-keyring-daemon started properly and unlocked keyring
sep 16 09:20:35 weird-stuff systemd[1]: Stopping Light Display Manager...
sep 16 09:20:40 weird-stuff lightdm[4045]: Error: can't open /lib/modules/5.4.0-47-generic/updates/dkms
sep 16 09:20:40 weird-stuff lightdm[4045]: Error: can't open /lib/modules/5.4.0-47-generic/updates/dkms
sep 16 09:20:40 weird-stuff systemd[1]: lightdm.service: Succeeded.
sep 16 09:20:40 weird-stuff systemd[1]: Stopped Light Display Manager.
sep 16 09:20:40 weird-stuff systemd[1]: Starting Light Display Manager...
sep 16 09:20:40 weird-stuff lightdm[4095]: Seat type 'xlocal' is deprecated, use 'type=local' instead
sep 16 09:20:40 weird-stuff systemd[1]: Started Light Display Manager.
sep 16 09:20:41 weird-stuff lightdm[4118]: Error: can't open /lib/modules/5.4.0-47-generic/updates/dkms
sep 16 09:20:41 weird-stuff lightdm[4118]: Error: can't open /lib/modules/5.4.0-47-generic/updates/dkms
sep 16 09:20:41 weird-stuff systemd[1]: lightdm.service: Main process exited, code=exited, status=1/FAILURE
sep 16 09:20:41 weird-stuff systemd[1]: lightdm.service: Failed with result 'exit-code'.
sep 16 09:20:41 weird-stuff systemd[1]: lightdm.service: Scheduled restart job, restart counter is at 1.
sep 16 09:20:41 weird-stuff systemd[1]: Stopped Light Display Manager.

[Alberto Milone (albertomilone)]

Unfortunately "Error: can't open /lib/modules/" has nothing to do with any freezes. It is not even an actual error message. Feel free to file a bug report against the nvidia driver, if that is what you are using.