proposal: etc/rc.d/initdiskless -> /etc/rc.initdiskless

[Brooks Davis]

On Sun, Feb 27, 2005 at 04:12:03PM +0300, Mike Makonnen wrote:
> On Thu, Feb 24, 2005 at 02:12:57PM -0800, Brooks Davis wrote:
> > I would like to propose that we move the initdiskless script out of rc.d
> > and back into /etc.  The reason for doing this is that initdiskless is
> > really about bootstrapping /etc from a minimial configuration.  Thus
> > running it via rcorder is really running it too early. 
>                                               ^^^^^^^^^
> I think you mean too late, right?

Err, yes. :-)

> > By moving it to
> > /etc and running it directly from /etc/rc (possiably conditionally based
> > on the existance of /conf) we fix this issue.  The main benefit of this
> > is that /etc/rc.d can now be modified or created entierly from entries
> > under /conf which could make it easier to maintain extra scripts in the
> > main flow of rc.d.
> > 
> > The change to /etc/rc would look something like:
> > 
> > diff -u -p -r1.335 rc
> > --- rc  8 Oct 2004 14:23:49 -0000       1.335
> > +++ rc  24 Feb 2005 21:02:18 -0000
> > @@ -50,6 +50,10 @@ HOME=/
> >  PATH=/sbin:/bin:/usr/sbin:/usr/bin
> >  export HOME PATH
> >  
> > +if [ -d /conf ]; then
> > +	/etc/rc.initdiskless
> > +fi
> > +
> 
> This is too generic. Isn't there some way of definitely knowing
> whether this is a diskless boot?

Looking at initdiskless again, this check is unnecessicary because the
first code in initdiskless is:

dlv=`/sbin/sysctl -n vfs.nfs.diskless_valid 2> /dev/null`
[ ${dlv:=0} -eq 0 ] && [ ! -f /etc/diskless ] && exit 0

There might be an argument for moving that test back in to /etc/rc to
avoid including the file at all in the normal case and to make it
entierly clear that initdiskless will not be run unless the syscall is
set or /etc/diskless exists.

> >  . /etc/rc.subr
> >  
> >  # Note: the system configuration files are loaded as part of
> > 
> > 
> > Does anyone have major objections to this change?  The goal is to run
> > initdiskless as early as possible so /etc can be as minimal as possible
> > at that point.
> 
> Sounds ok to me.

Great.

-- Brooks

-- 
Any statement of the form "X is the one, true Y" is FALSE.
PGP fingerprint 655D 519C 26A7 82E7 2529  9BF0 5D8E 8BE9 F238 1AD4
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-rc/attachments/20050228/4efdae02/attachment.bin