55

During a recent apt-get dist-upgrade of 12.10 I received the following error:

Setting up linux-headers-3.5.0-19-generic (3.5.0-19.30) ...
Examining /etc/kernel/header_postinst.d.
run-parts: executing /etc/kernel/header_postinst.d/dkms 3.5.0-19-generic /boot/vmlinuz-3.5.0-19-generic
Error! Could not locate dkms.conf file.
File:  does not exist.

Any ideas?

muru
  • 197,895
  • 55
  • 485
  • 740
Andrew Tomazos
  • 1,604
  • I know this is a bad answer, but I always have that error, and things work fine. I guess dkms just uses default settings. – MiJyn Dec 09 '12 at 20:28

5 Answers5

47

This is usually caused by modules in /var/lib/dkms that don't have a dkms.conf file within their source subdirectories - dkms expects this file, so will report an error if it's missing, and then may miss out on compiling some valid modules.

To find the offending module, run this short script (thanks to Lekensteyn):

for i in /var/lib/dkms/*/[^k]*/source; do [ -e "$i" ] || echo "$i";done

That will output any dkms module directories; you can then decide whether to uninstall the package that created them, or if they've been manually installed or renamed, to move them somewhere else or remove them.

David Fraser
  • 1,717
  • 3
    Perfect. The other answers may apply to specific instances, but this very quickly demonstrated where the issue was (and it was really a non-issue, as certain software was copying old installations to a /var/lib/dkms/*.old directory). – Auspex Dec 29 '15 at 17:31
  • In case nvidia is the offender e.g. /var/lib/dkms/nvidia/396.24/source is the output, check there are already newer versions in /var/lib/dkms/nvidia and remove the offending one. – Flatron Jul 17 '18 at 07:31
  • Thanks! In my case the offender was open-vm-tools packages, and after purging I had to remember to run apt auto-remove, as a dependency was causing this problem. – Elliptical view Nov 12 '19 at 19:50
  • for me the package was evdi – Native Coder Mar 04 '24 at 21:31
36

I've been getting that on kernel upgrades for a while, with the result that not all of my dkms drivers get updated on kernel-upgrades.

Firstly I was able to workaround the problem by reconfiguring any packages that used dkms to force them to be recompiled for the current kernel - eg it was my AMD video drivers that failed (package == fglrx):

sudo dpkg-reconfigure fglrx

That would at least get the drivers configured for the current version.

Finally I found this bug, which includes some steps you can use to diagnose: https://bugs.launchpad.net/ubuntu/+source/dkms/+bug/830915

dkms status
ls -R /var/lib/dkms

Basically what they're doing is looking around for anything that is surprising, or unexpected - eg packages you have uninstalled - or software you have manually installed on an earlier ubuntu version and may not work with the newer version. Particularly check the date-stamp on the directories which might show you particularly old packages could be from manually installed packages.

ls -l /var/lib/dkms

In my case I had an old version of the fglrx install I had used to diagnose some problems a year ago and had simply re-named it. Deleting this old cruft made the problem go away.

If there's junk delete it (or move it out of that directory) - if there's a manually installed package there, consider updating it, or uninstalling it and using the maintained version.

Greg
  • 1,413
  • 1
    Thank you for this, the problem I had was with the nvidia-340 driver... – ionreflex Oct 21 '15 at 13:52
  • 1
    I had the same issue with anbox – Airfield20 Jun 09 '18 at 12:44
  • 1
    mine problem was in vboxhost. Thanks! – Drey Dec 08 '19 at 16:45
  • Great answer, explains how to diagnose the issue. In my case it was that in Bookworm (as Debian Testing codename) the Nvidia module started living under /var/lib/dkms/nvidia-current, and older /var/lib/dkms/nvidia/current/xxx modules werent removed - which sudo apt-get autoremove from time to time would have prevented. I always keep an old kernel for emerg, and hesitate to autoremove, but you can mark older kernel versions (or any package) as manually installed to prevent auto-removal. See: https://askubuntu.com/questions/779266/stop-apt-get-autoremove-from-removing-old-kernel – Rik Nov 16 '21 at 01:57
16

I have had this problem with VirtualBox from Oracle's ppa, rather than the one packaged with 12.04 LTS :

Error! Could not locate dkms.conf file.
File:  does not exist.

I copied the contents of /var/lib/dkms to another directory as a backup, then removed the vboxhost directory in /var/lib/dkms;

mkdir ~/backup_dkms
cp -r /var/lib/dkms/* ~/backup_dkms 
rm -rf /var/lib/dkms/vboxhost

I then reconfigured the virtualbox package like this:

sudo dpkg-reconfigure virtualbox-4.2

and it succeeded in doing this:

   * Trying to register the VirtualBox kernel modules using DKMS
Braiam
  • 67,791
  • 32
  • 179
  • 269
Ian Ohr
  • 161
13

I was also suffering from this problem for a long time. I need to reinstall the NVIDIA driver each time after kernel update and restart. Recently, I started to look into this problem. Actually, my problem is that there are multiple module folders with the same prefix nvidia- under /usr/src/ and multiple folders with different version numbers under /var/lib/dkms/nvidia. After removing the older versions, both following commands

    dkms status
    dkms autoinstall

worked.

ionreflex
  • 629
antiquity
  • 141
  • 1
  • 3
-1

This happened to me once on a Red Hat 7.5 Workstation with an Nvidia driver. I know it's not Ubuntu but this may help someone...

Remove the Nvidia files from dkms and then reinstall dkms:

rm -rf /var/lib/dkms/nvidia
yum reinstall dkms

Then reinstall the Nvidia driver

./NVIDIA-installer.bin --dkms
reboot

WARNING, this worked for me but may not be the correct way of doing things.

Stephen Rauch
  • 1,156
Jake
  • 1