My two cents:

disklabel(8) should not be allowing modifications of mounted filesystems
(by checking at write time).  At least not without sufficient warnings
that will drive off someone with a shadow of a doubt as to what they
are doing.

On Wed, 16 Jan 2002, Alfred Perlstein wrote:

# * David Laight <david@nohope.l8s.co.uk> [020116 11:13] wrote:

# > When a disk write is done, the offset of the partition on the disk is
# > taken from the CURRENT disklabel, not the label that existed at the time
# > the partition was mounted.
# >
# > There are no checks to stop you changing the partition offsets of a
# > partition that has a filesystem mounted on it.  However changing these
# > values will be rather devastating on the disk area that the label now
# > points to.
# >
# > IMHO the mount should grab the current offset and stick with it.
# >
[...]
#
# This would seem to be a major layering violation.  The FS is only
# supposed to know about the underlying disk, not the boundry of
# the partition.  It sees 0..endofpartition, not start..end. :)

				--*greywolf;
--
Microsoft:  Living proof that Borg screw Ferengi.