On Tue, Apr 12, 2011 at 11:59:10PM -0400, Mike Frysinger wrote:
> On Tuesday, April 12, 2011 23:41:23 Phillip Susi wrote:
> > mkdir /mnt/dev
> > mount --rbind /dev /mnt/dev
> > umount /mnt/dev
> >
> > The umount fails because of the sub mounts that the recursive bind also
> > bind mounted under /mnt/dev, which you have to remove manually first.
> > When you mount with --rbind, umount should be able to reverse the
> > process, unmounting all recursively bound mounts.
>
> yes and no ... i dont think the fact you used "--rbind" is tracked anywhere.
Yep, it's important to understand that "bind" mounts are nothing
special from kernel point of view.
# mount /dev/foo /mntA
# mount /dev/foo /mntB
is exactly the same thing like:
# mount/dev/foo /mntA
# mount --bind /mntA /mntB
For Linux kernel there is not any extra relation between the mountpoints.
The filesystem is connected to the two places in the VFS. That's all.
For more details see /proc/self/mountinfo. It's mount(8) disadvantage
that mtab entries use directories as a source (1st field) for bind mounts.
For example:
/mntA /mntB fstype bind,options 0 0
in mtab is nonsense (e.g. /mntA could be umounted independently on /mntB).
Unfortunately many users still believe that "bind" mounts are something
special. It's probably because mount(8) maintaining the "bind" flag in
the mtab file and because there is a special semantic for
MS_REMOUNT|MS_RDONLY|MS_BIND in kernel.
Note that MS_BIND operation allows to mount subdirectory (non-root
directory) from the filesystem, then
mount --bind /mntA/subdir /mntB
/subdir is the root for /mntB -- this is also tracked in
/proc/self/mouninfo (4th field). It seems that
mount -t btrfs -o subvol=subdir /dev/foo /mntB
has the same effect on btrfs.
Karel
--
Karel Zak <kzak@xxxxxxxxxx>
http://karelzak.blogspot.com
--
To unsubscribe from this list: send the line "unsubscribe util-linux" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
