varlock and varrun

[Christoph Pleger]

Hello,

On Thu, 12 Apr 2007 17:26:40 -0400
John Dangler <jdangler at atlantic.net> wrote:

> > I generated a minimal edgy system by debootstrap, copied that system
> > to a machine with fresh formatted partitions and installed a kernel
> > and a bootloader. After rebooting, the machine did not fully come up
> > and I found out that the reason for that was that /var/lock and
> > /var/run had not been mounted. After some investigation I found out
> > that /var/lock and /var/run should be mounted by S01mountkernfs.sh
> > and I wonder how that is possible when S35mountall.sh is run later,
> > because the mountpoints /var/lock and /var/run cannot be present
> > before mounting /var, what normally should be done by
> > S35mountall.sh.
> > 
> > In spite of that, after a normal CD-installation the machine boots
> > successfully. But I cannot find the significant difference between a
> > CD-installation and my debootstrap-installation.
> > 
> > Does anybody know why mounting /var/lock and /var/run works ok after
> > installing from CD?

> Hrmm - does it load at the same init level?  I'm not sure that
> /var/run is necessary for the boot of the machine... (or am I agreeing
> with you?)

It seems that having /var/run mounted for example is necessary for
bringing up the network interfaces because that action wants to create a
file /var/run/network/ifstate. 

On a machine where I used the CD-installation, I can see (by using df)
that /var/lock and /var/run were mounted and I cannot find a place where
there could happen other than in S01mountkernfs.sh.

Regards
  Christoph