[PATCH] root-hopping for pre-2.3.41-3

[H. Peter Anvin]

Followup to:  <Pine.LNX.4.10.10001251214350.1212-100000 en penguin.transmeta.com>
By author:    Linus Torvalds <torvalds en transmeta.com>
In newsgroup: linux.dev.kernel
> 
> d_path() can easily be made to yield something else: in particular, we
> already have the "(deleted)" semantics in /proc/<pid>/fd to show that a
> name is not actually necessarily valid, and the same goes for things that
> are not actually reachable from the root.
> 
> I don't think device numbers are all that helpful. And in fact I think
> that if it shows any of the "unattached" filesystems at all, it's probably
> a bug in the first place - after the root hop, the old root should not
> show up.
> 
> Finally, as to the umount / new system call issue: I really think that the
> root hopping should be a special event, not a re-mount or a umount, but
> more of a "global chroot()" system call. I agree with making it possible
> to re-mount from user space, but I don't agree with overloading an
> existing call that does something completely different.
> 

I would like to suggest an alternate system call with, perhaps, more
useful semantics:

       pivot_root("new_root", "old_root");

This will detach the path "new_root" and turn it into the new root,
and *then* attach the *old* root, now detached, to the path
"old_root", which must be a valid mount point (i.e. directory) in the
new root.

For example, say that you have a ramdisk (initrd) as root, and you
just mounted your intended root at /newroot.  You could then do:

     mkdir("/newroot/initrd");
     pivot_root("/newroot", "/initrd");

     /* Now the initrd is mounted at /initrd, and the new root is / */

     /* Do stuff that involves the initrd */

     umount("/initrd");	       /* Done with it */

    -hpa
-- 
<hpa en transmeta.com> at work, <hpa en zytor.com> in private!
"Unix gives you enough rope to shoot yourself in the foot."

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo en vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/