vboxdrv & vboxnet services under debian based systems : incorrect S an K runlevel sequence numbers => Fixed in SVN

[Samuel Monsarrat]

The vboxdrv & vboxnet service scripts contain the following line:

# chkconfig: 35 30 60

Which suggests that the intent is to start these services at sequence 30 and stop them at 60.

First, it would seem more logical to follow basic formula K = 100 - S i.e. 30-70 or 40-60. Is there a special reason for not stopping the vbox services in the same sequence as they are started?

More importantly, under debian systems the chkconfig directive has no effect, chkconfig is a fedora/redhat directive. The start ans stop sequence numbers should be explicitly passed to update-rc.d command. This is no longer the case as with version 2.1.4 the start and stop sequence numbers are both 20 - the default for update-rc.d.

This causes serious problems on systems that run vbox guests as services as they must be stoped before any attempt to stop vboxdrv & vboxnet. On a Ubuntu 8.10 stopping vboxdrv & vboxnet with a guest running seems to result in a kernel crash, changing the sequence numbers to 20 80 solved the problem.

[Frank Mehnert]

First, we use the same script for all our Linux systems. So, if chkconfig is not supported by Debian that is fine. Second, the crash when removing the kernel modules while a VM is still active is fixed in SVN. I don't remember the ticket anymore, can you help?

Apart from this, thanks for this report, will check the sequence numbers.