On Tue, Jun 21, 2005 at 12:06:11AM -0400, Allen Briggs wrote:
> OK.  Comparing the 'd_secperunit' field in the default label to that
> in the actual label read from the disk does work for me.

After sleeping on it, this seems like a hack.  It's testing for a
side effect of a condition, not for the condition itself.

In general, if a disk is a member of an array, I really only want
to see that disk at autoconfig time and for access to S.M.A.R.T.
data or some such.

The system you describe seems broken, and I really wonder if
that's a model that we want to support at the expense of what
seems to me a more correct solution.

You can easily continue to operate in your configuration if
you simply wire the root device in your kernel.

-allen

-- 
                  Use NetBSD!  http://www.NetBSD.org/