On Fri, Jul 02, 2004 at 01:13:40PM -0500, Donald Lee wrote:
> I was hoping that this support article would have information useful
> to fixing the driver (which I still suspect is the problem - endian problem),
> but it calls out cable problems and drive utilities.
> 
> The same drives that are flaky on my Beige with the Sonnet temp card
> (and drop back to Ultra/33) run fine - without hiccup - on the G4/AGP
> on the internal busses.
> 
> I still think it's a driver problem with the wdc and/or wd driver.

It's clearly a problem with the driver. But I don't think it's an
endian issue, because it has the exact same behavior on alpha, if I remember
properly of what tests I did.

I suspect that the issue is that the driver fails to perform some
initialisation test. It's not an issue on i386, because the BIOS does it
for us.
I dumped the PCI registers value on i386 and sparc64, and they are identical.
So it's possible that the missing initialisation isn't a static value in a
register, but a sequence of writes that would form a command.
Without the technical docs about this chip, I'm stuck here.

> 
> I could try using the cables from the G4 on the Sonnet card, but the cables
> I'm using with the Sonnnet card are the cables that came with the card.
> It would be pretty silly if Sonnet sent me inadequate cables.......
> (I should try it....)

I'm almost sure it's not the cables: the driver can detect the cable type,
and set the appropriate mode automatically. So if you didn't use the
right cable, the driver wouldn't attach the drive at Ultra/100

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la difference
--