RFC - Bugs 688 & 732

Greg Schafer gschafer at zip.com.au
Sun Jan 25 19:31:58 PST 2004

On Thu, Jan 22, 2004 at 11:00:23PM +0100, Alex Groenewoud wrote:
> Anderson Lizardo wrote:
> > On Wednesday 21 January 2004 04:15, Greg Schafer wrote:
> > > Sure, /proc/mounts and the real /etc/mtab can get out of sync, but
> > > the consequences of this aren't too worrisome when it's only some virtual
> > > file systems we're dealing with.
> > 
> > Except for the fact that, if you shutdown the host system without umount'ing 
> > /proc and /dev/pts (from inside chroot), the "umount -a" command issued by 
> > the shutdown script from the host system will fail with something like
> > 
> > umount: /mnt/lfs: device is busy
> Yes, that's the whole reason why mounts should be done outside of
> chroot: so that the halt scripts on the host know what to unmount.
> I've seen those red "failed" messages far too often.

Ok, well how about this then.. instead of moving the mounts to Ch 5 like I
suggested in the original report, we instead still do it in Ch 6, but before
entering the chroot? In this way, disruption to current flow of the book is
minimal. We just move "Mounting the proc and devpts file systems" to
immediately before "Entering the chroot environment" and Bob's yer uncle! Of
course, some minor adjustments to the text would be needed and we'd have to:

mkdir $LFS/proc
mkdir $LFS/dev/pts

before the actual mounts.

Ch 5 is still done as non-root. In Ch 6, instead of saying "become root" in
the "Entering the chroot environment" section we just say it one section
earlier. No more red "failed" messages :-)

Anyone see a problem with this approach?

More information about the lfs-dev mailing list