On Thu, 19 Dec 2002, Bill Studenmund wrote:

# I think that part of the dispart and the wedges proposals is that the
# kernel would get out of writing disklabels; only userland tools would do
# it. Writing disklabels is too much of a mess. :-)

Not if you can provide it with a flexible data spec (but that's no better
than writing directly to the disklabel area on the raw disk...).

And I hope you're not proposing that the kernel would get out of *reading*
disklabels.  The kernel kind of needs to have that information present.
But you knew that.

Someone was objecting to writing disklabels when securelevel > 0.
I don't think you can even do that *now*, since you have to be able
to open the raw device R/W with securelevel > 0.

Another point to ponder:  Whatever userspace thing writes disklabels
will need to be aware AND CORRECT in what and where it writes its
trash, or else you will end up zorching maybe a cluster of inodes
(and hence the data associated therewith (yes, I know there are
ways to sort of recover this data but they're not easy)).

# Take care,
#
# Bill


				--*greywolf;
--
NetBSD: Stable and strong!