Re: patches

[Maarten Vanraes <maarten.vanraes-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>]

Op woensdag 21 juli 2010 17:18:28 schreef Maarten Vanraes:
> Op woensdag 21 juli 2010 13:33:05 schreef Harald Hoyer:
> > On 07/14/2010 02:05 AM, Maarten Vanraes wrote:
> > > Hi,
> > >
> > > Here are some patches, some that were used as patches in the mandriva
> > > builds, and a first one from me, to fix nfs hostonly on chrooted
> > > installs.
> > >
> > > Kind regards,
> > >
> > > Maarten Vanraes
> >
> > -[ "$1" = "-h" ] && ! egrep -q '/ nfs[34 ]' /proc/mounts && exit 1
> > +[ "$1" = "-h" ] && ! egrep -q '\s/\s+nfs[34]?\s' /etc/fstab && exit 1
> >
> > Hmm, I think "hostonly" should be really "hostonly", so the check is
> > valid.
> >
> > You can always force the addition of nfs, if you are building a special
> >  image for an NFS client. Or just boot with the generic (non-hostonly)
> > one, and build the hostonly one on the client.
> 
> i see what you mean, but i beg to differ.
> 
> this thing is not really about nfs, it's about building an initrd on a
>  chroot.
> 
> if you're building initrd on a chroot, it's always a hostonly, in fact,
>  only distros ship generic initrds. and even that's not always true.
> 
> this fix has no consequences whatsoever in any case, except that it checks
> fstab instead of /proc/mounts. which means if you adjust your fstab,
>  because you're going to run it elsewhere.
> 
> I was planning on checking everywhere and trying to reduce the dependencies
>  on /proc and /sys, IF it works the same everywhere without any
>  consequences (just like this here).
> 
> Also, if you look at it; not having / mounted as an nfs WILL disable the
>  use of this module entirely and thus all the root=nfs stuff will fail.
> 
> What do you think?
> 

imo,

Either we get hostonly to work for this, or we need a separate chroot option 
to make this work.

We can fiddle all we want, but in effect the purpose of this is to have an easy 
way to generate an initrd.