On Mon, May 07, 2001 at 01:54:36AM -0400, Thor Lancelot Simon wrote:
> This leaves me wondering if the actual bug that's corrupting data
> here is not in the IDE controller at all, but perhaps in VIA's
> north bridge (memory controller, &c.).  Has anyone else had problems
> with a motherboard with a 686B but an AMD or other (e.g. Samsung)
> north bridge on it?

The problem is (as far as it has been diagnosed) in PCI DMA, not
in the IDE part of the chip. It's just that the IDE controller
is the most likely candidate to do busmastering DMA in the
most usual system configurations.

Also, some motherboards seem to exhibit the problem, while others
are very stable. I do not think that it's only the 686B, this is
just the revision that has been widely publicized lately. I have
seen reports of problems (like system lockups when trying to use
a tv card) on earlier revisions of the chipset.

Given the variety of systems that the problem has shown up on
(Linux, Windows, NetBSD), there is definitely a hardware problem
there, but it seems to be require specific circumstances to
show up, even specific boards. This makes it hard to diagnose
if in fact there is, say, a problem with VIA timings as set
by our driver, since the chipset is a variable with a seemingly
random value.

- Frank

-- 
Frank van der Linden                           fvdl@wasabisystems.com
======================================================================
Quality NetBSD CDs, Support & Service.   http://www.wasabisystems.com/