On Sat 20 Nov 1999 at 09:17:20 -0800, Jason Thorpe wrote:
> On Sat, 20 Nov 1999 15:41:49 +0100 
>  Olaf Seibert <rhialto@polder.ubc.kun.nl> wrote:
> 
>  > In the AlphaBIOS setup I selected the SRM for Digital Unix (that seemed
>  > a better choice than SRM for OpenVMS). When I next powered down and up,
>  > the SRM came up, but it told me there was an unsuppored card in slot 4,
>  > and would I please remove it before it would do anything else.
>  > 
>  > The card in slot 4 was the video card (S3 864).
> 
> Earlier revisions of Miatas have a horrible core logic bug in them which
> manifests itself when a device attempts to prefetch across a page boundary.
> The SRM worked around this by refusing to enable cards that it didn't "know
> about" which were plugged into the primary PCI bus.  These cards that it
> "knew about" were known by SRM to not do the thing that caused the bug to
> surface.

Does it mean that I have such a revision, or is the SRM simply still
being very cautious? The firmware revision date seemed fairly recent (4
april 1999). I haven't updated it yet because my cable ISP has been
messing up for the mast 30 hours or so.

> Plugging such devices into the secondary PCI bus (the 32-bit slots, which
> are behind a PCI-PCI bridge, which prevents the bug from manifesting itself)
> is allowed by SRM.
> 
> In the case of a video card, you're probably okay, since by default, that
> card isn't going to be doing DMA.
> 
> However, if the card has a DMA engine on it for e.g. copying pixmaps from
> host memory to display memory, there could be a problem when that DMA engine
> is used.

Apparently it works with NT (because there was one hopefully identical
machine on display running that). And the AlphaBIOS didn't care.

> I don't remember what card DEC generally OEM'd in those machines... I know
> that AlphaServer 4100s often have S3 ... Trio64s maybe?  Or maybe 968s.
> 
> Anyhow, SRM is obviously allowing it to function, tho it probably didn't
> set the Bus Master Enable bit in the PCI Command register on the card.
> 
>  > But my main question is: how can I use the SRM with the video card that
>  > I have? It works well enough to show the countdown and the error
>  > message,I don't see any reason why the SRM would have any trouble with
>  > it. Besides, I obviously want to use the machine with wscons and X, not
>  > headless.
> 
> So, maybe I'm misunderstanding... after it spews the error, it disables
> the card?  Well, now, that's annoying.

It does not disable it, it simply halts (doesn't do anything visible,
in any case). What I see on my screen (using that same video card) is
the countdown (how nicely like my VAXen ;-) then the message about it
not knowing that card, and that I should power down and remove it.
After that, nothing. The very same thing happened on the serial console,
so I was a bit nervous until I found how to revert to AlphaBIOS ;-)

> I dunno what the score with PCI VGA on secondary busses is, but maybe if
> you could try putting it in one of the 32-bit slots, it'll work okay.

Thanks, I'll do that. When I get this working, maybe I'll have a nice
starting point to install NetBSD on the VAXen and Sun 3/50 I was given
today.

>         -- Jason R. Thorpe <thorpej@nas.nasa.gov>
-Olaf.
--
___ Olaf 'Rhialto' Seibert - rhialto@polder.ubc.     -- If one tells the truth,
\X/ .kun.nl     -- one is sure, sooner or later, to be found out. (Oscar Wilde)