Subject: Re: Supporting sector size != DEV_BSIZE on lfs
To: Trevin Beattie <trevin@xmission.com>
From: Bill Studenmund <wrstuden@netbsd.org>
List: tech-kern
Date: 06/20/2002 14:07:45
On Thu, 20 Jun 2002, Trevin Beattie wrote:

> Those patches are *way* outdated -- circa NetBSD-1.2!

Yes, they are. They however outline the three different ways we could
solve the general problem, and form a basis for everyone else to talk
about this change.

All the stuff you and I and Chuck have been talking about, Koji Imada had
thought about back around NetBSD-1.2. His thinking was beyond where
everyone else at the time was with respect to this issue, so it made it
hard to understand what he was talking about. That English was not his
native toungue didn't help either. After he talked about this for a while
on tech-kern, we all realized that he knew more about this than we did and
he also was discussing clean ideas. So I think we were going to just pick
one of the three approaches & go from there.

Basically the three options are: 1) block numbers (and i/o sizes) are in
sectors (wrstuden-devbsize), 2) block numbers are in DEV_BSIZE units but
have to be in full sector multiples (what you're doing), or 3) block
numbers in DEV_BSIZE units but i/os can be in DEV_BSIZE units. i.e. the
driver has to support possibe sub-sector reads/writes.

Unfortunately he died in a motorcycle accident, which is why something
that would have made 1.3 hasn't happened yet.

Take care,

Bill