Subject: Re: scsi-disks hardsectored to 1024 byte/sec
To: Guenther Grau <Guenther.Grau@bk.bosch.de>
From: Leo Weppelman <email@example.com>
Date: 07/14/1998 16:34:11
On Tue 14 Jul 1998, Guenther Grau wrote:
> Leo Weppelman wrote:
> > On Wed 08 Jul 1998, Kurt Schreiner wrote:
> > > is there some way to make this disks really usable for netbsd? I'm
> > > willing to do some kernel-hacking, if someone would tell me where to
> > > begin.
> > On Wed 08 Jul 1998, Manuel Bouyer added:
> > >
> > > See PR kern/3790, kern/3791 and kern/3792.
> > I just send my patches to apply kern/3792 on -current to gnats in an effort
> > to get them appended to the pr. These patches are not complete, I omitted
> > the MD-parts and the vnd patches. The MD part should not be to difficult,
> > it is only disksubr.c - with luck the the original diff 'just works'. The
> > last time I really tried using non 512-byte sectored disks was 2 or 3 months
> > ago.
> anybody knows why these patches haven't been commited? Lack of time?
> Or are there any `structural` problems with it?
Take a look at the patches ;-) Each of the 3 above mentioned pr's provides
a different method of fixing the problem... Apart from the heavy testing
that needs to be done, there has to be a discussion of which method should
be choosen. Part of that discussion has already taken place with no apparent
outcome [ I have assembled some docs on my system somewhere ]. Just for the
fun of it, I took the patches to play with them a bit until I bought a new
house that drains a lot of my time :-( I currently just can't find enough
time to start such a project while being a reasonable portmaster...
What I have done until now:
- gathered some mail of the discussions preluding the pr's.
- kept a large piece of the kern/3792 method patches up-to-date
- did some performance tests to measure the overhead added as requested
by Eric Fair(?). At first sight the overhead was nearly measurable - not