Occasionally, my system gets into a state where some kernels are missing a module or two, because DKMS somehow forgot to compile those modules for that kernel. Rather than spend time diagnosing the problem, it would be nice if there was a single command I could run that woudl just rebuild every dkms-controlled module for every installed kernel. Is there such a command?

  • I always get the output Module broadcom-sta/5.100.82.112 already installed on kernel 2.6.38jon-64/x86_64 I really want a --force or a --rebuild --just-do-what-i-say option ;) – user140350 Mar 14 '13 at 21:14
up vote 39 down vote accepted

I figured out a shell one-liner to do it:

ls /var/lib/initramfs-tools | \
    sudo xargs -n1 /usr/lib/dkms/dkms_autoinstaller start

This works because the names of the directories in /var/lib/initramfs-tools are exactly the kernel version names that you need to pass to dkms_autoinstaller to tell it to rebuild all modules for those kernel versions. Note that if you have uninstalled some old kernels, their directories might still be lying around and cause some errors to be reported, but this isn't a problem because dkms_autoinstaller will just do nothing for those kernel versions that aren't installed.

  • 1
    it gave some errors because it came up headers-xxx and headers-xxx-generic but it seemed to rebuild the correct stuff despite the errors – frankster Feb 14 '12 at 19:09
  • @frankster After finding multiple "Error! Could not locate dkms.conf file." for a new kernel installation, I was able to install previous kernel modules listed by dkms status to the new kernel using the following per specific modules I wanted from "/usr/src" . Parameters need to be changed as needed for -c, -m, -v . Here is an example for the nvidia-384-384.90 module: ls /var/lib/initramfs-tools | \ sudo xargs -n1 /usr/sbin/dkms install -c /usr/src/nvidia-384-384.90/dkms.conf -m nvidia -v 384-384.90 -k – m1st0 Oct 24 '17 at 11:14

Doesn't look like the dkms command allows you to do that. I created a small Python script that should do what you want. You can put an alias in your ~/.bashrc like

alias dkms-buildall='sudo ./wherever/your/script/is'

Of course you'd need to make it executable first. Here's the code:

#!/bin/env python
#
# NOTE: This assumes that all modules and versions are built for at
#       least one kernel. If that's not the case, adapt parsing as needed.
import os
import subprocess

# Permission check.
if os.geteuid() != 0:
    print "You need to be root to run this script."
    exit(1)

# Get DKMS status output.
cmd = ['dkms', 'status']
process = subprocess.Popen(cmd, stdout=subprocess.PIPE)
dkms_status = process.communicate()[0].strip('\n').split('\n')
dkms_status = [x.split(', ') for x in dkms_status]

# Get kernel versions (probably crap).
cmd = ['ls', '/var/lib/initramfs-tools/']
# Alternative (for use with Arch Linux for example)
# cmd = ['ls', '/usr/lib/modules/']
process = subprocess.Popen(cmd, stdout=subprocess.PIPE)
kernels = process.communicate()[0].strip('\n').split('\n')

# Parse output, 'modules' will contain all modules pointing to a set
# of versions.
modules = {}

for entry in dkms_status:
   module = entry[0]
   version = entry[1].split(': ')[0]
   try:
      modules[module].add(version)
   except KeyError:
      # We don't have that module, add it.
      modules[module] = set([version])

# For each module, build all versions for all kernels.
for module in modules:
   for version in modules[module]:
      for kernel in kernels:
         cmd = ['dkms', 'build', '-m', module, '-v', version, '-k', kernel]
         ret = subprocess.call(cmd)

Tested it here, seems to work just fine:

$ dkms status
nvidia-current, 275.09.07, 3.0.0-5-generic, x86_64: installed
virtualbox, 4.0.10, 3.0.0-5-generic, x86_64: installed

$ sudo python dkms.py
...

$ dkms status
nvidia-current, 275.09.07, 3.0.0-5-generic, x86_64: installed
nvidia-current, 275.09.07, 3.0-2-generic, x86_64: built
nvidia-current, 275.09.07, 3.0-3-generic, x86_64: built
virtualbox, 4.0.10, 3.0.0-5-generic, x86_64: installed
virtualbox, 4.0.10, 3.0-2-generic, x86_64: built
virtualbox, 4.0.10, 3.0-3-generic, x86_64: built

If you also want to install the modules, replace build with install in the second last line.

Combining @htorque and @Ryan Thompson's answers, here's my (as root) one-liner:

dkms status | sed s/,//g | awk '{print "-m",$1,"-v",$2}' | while read line; do ls /var/lib/initramfs-tools | xargs -n 1 dkms install $line -k; done

The above don't work on all variants, this might be a bit more helpful in those cases...

$modulename="whatever"
$moduleversion=`modinfo $modulename | grep "^version:" | awk '{ print $2 }'`
dkms status | grep $modulename | tr -d ',' | awk '{ print $3 }' | xargs -n1 dkms build $modulename/$moduleversion -k
  • Could you elaborate on what this does that the other methods don't? – Ryan Thompson Feb 19 '16 at 23:00
  • It works on systems that don't have /usr/src/linux-headers-* and /var/lib/initramfs-tools – stu Feb 21 '16 at 12:43
  • consider the situation where you need it to run on many various linux distros that only have the fact that dkms (sort of) works, in common. – stu Feb 21 '16 at 12:43

dkms status and dkms_autoinstaller does not work in ubuntu 16.x. So some shell script will do. This script assumed you have proper *-dkms deb-package installed, and bash is your shell.

for k in $(ls /var/lib/initramfs-tools) ; do
 for d in $(cd /usr/src; ls -d *-*) ; do
  [[ -f /usr/src/${d}/dkms.conf ]] || continue
  m=$(echo $d | sed -r -e 's/-([0-9]).+//')
  v=$(echo $d | sed -r -e 's/[^0-9]+-([0-9])/\1/')
  sudo /usr/sbin/dkms install -c /usr/src/$d/dkms.conf -m $m -v $v -k $k
 done
done 

Your Answer

 

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.