On Wed, May 15, 2002 at 02:04:10AM -0700, Matt Thomas wrote:
> At 01:46 AM 5/15/2002, Manuel Bouyer wrote:
> >On Wed, May 15, 2002 at 01:30:05AM -0700, Matt Thomas wrote:
> > > That "fixed" the problem but looking throught the scsipi code, I don't
> > > think DT is handled properly.  DT implys WIDE16/SYNC which the code clears.
> >
> >Yes, and it's because DT implys WIDE16/SYNC that WIDE16/SYNC is used for ST
> >only. We need a way to tell a drive supports DT clocking but not ST, as
> >it's allowed by the specs.
> >
> > > That seems wrong.  Also the periph_cap code doesn't about DT and so doesn't
> > > print correct drive information.
> >
> >We could use DT as WIDE16+SYNC here. It's easy to add.
> >
> > >  Also, I don't any negotiation is done for
> > > DT and that's wrong as well.
> >
> >Because your SCSI adapter doesn't support DT clocking.
> 
> Not true.  Using esiop which does support DT I had the exact same problem.
> Or does the 896 not support DT clocking?

No, the 896 doesn't. Only the 1010 does (the ones that are Ultra160 capable).

DT clocking latches the data bus on both edges of the clock signal (this is
required to do more than 40 Mtrasfers/s), where ST latches data bus on one
(I don't remember which one) edge of the clock signal.

--
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
--