new packages

Richard Lightman richard at reika.demon.co.uk
Fri Mar 23 00:50:18 PST 2001


Misquoted from SamBC on 2001/03/23 at 02:24 +0000:
> ----- Original Message -----
> From: "Jesse Tie Ten Quee" <highos at highos.com>
> 
> 
> > On Wed, Mar 21, 2001 at 02:32:36PM -0000, SamBC wrote:
> > > <snip /> If only one could mount the same fs twice...
> 
> >        The type bind is not a filesystem type but allows remount
> >        ing  a part of the file hierarchy somewhere else. The call
> >        is `mount -t bind olddir newdir' or `mount -t bind oldfile
> >        newfile',  or  `mount  --bind  olddir  newdir'  with later
> >        mount/kernel versions.
> 
> For this to work in chroot, would one not have to mount the boot
> partition at /lfs/boot (as I do it) and bind it to /boot? And would this
> then work?
> 
> I say this as it seems questionable if the bind will work outof the
> chroot - and if it does, that is a worry... chroot is often uesd to stop
> linking out to the rest of the fs for things like ftpd
> 
Type the mount command before entering chroot. It will not work inside
chroot because you cannot get to the directories you want from inside.
The /proc/1/root/ hole has been fixed in kernel 2.4.x, so it is possible to
to mount the proc file system in chroot without giving access to the real
root.

I currently use the following binds without problems:

/mnt/build       -  One partition to build all the packages so my two LFS
                    partitions can be smaller.
/usr/share/ralfs -  so my install script can run itself during chroot
/proc            -  Some packages need it to install properly
/dev             -  I get the user/group ids right by grepping from the hosts
                    /etc/passwd, /etc/shadow, ...
Richard


-- 
Unsubscribe: send email to lfs-discuss-request at linuxfromscratch.org
and put unsubscribe in the subject header of the message




More information about the lfs-dev mailing list