| From: Vidiot <[EMAIL PROTECTED]>
| Subject: Re: [LINUX-ABIT] a cautionary tale
| >My message was warning about a design "feature" of the board that
| >limits the total number of bus mastering boards in PCI slots 4 and 5
| >to one. This was not explained at all in my manual, and needs to be
| >made very clear to all who might try to configure a BP6 system.
| >
| >Hugh Redelmeier
|
| It is plain as day on page 1-4 of the BP6 manual. My version is Rev 1.01.
| It is the 2nd bulleted item on the page.
It is plain as day on page 1-4 of my manual too BECAUSE I WROTE IT IN
WITH A RED PEN (a substitute for blood). Also on 3-30. I bought one
of the first batch of BP6 boards. My manual is MN-159-2A0-11 Rev
1.00.
I found it when I downloaded the PDF of the new manual and explicitly
searched for "master". This kind of question is only asked when the
answer is essentially known. Not fun. Hence my posting. To make it
more accessible, I'll change the subject on this message.
I think that sharing IRQs is reasonable, and it is explained in the
manual. Software can deal with this. True, drivers written for ISA
hardware don't take this into account -- ISA didn't allow IRQ sharing
between cards.
Sharing a Bus Master Control Signal seems like a big mistake. As I
understand it, there is no reasonable software workaround. And I
think that the results can to be spectacularly bad (perhaps
interleaved data from different devices). I admit that this is
speculation since I don't know the actual hardware details. The
failures I experienced were partial -- worse than a total failure
because they are harder to debug.
It would be a lot less error prone if they just eliminated the Bus
Master Signal from one of the slots. Then any device placed
there that required bus mastering would solidly fail. Now, failure
might only occur when the two devices happen to both be active (I
infer).
Hugh Redelmeier
[EMAIL PROTECTED] voice: +1 416 482-8253
--
=- To unsubscribe, email [EMAIL PROTECTED] with the -=
=- body of "unsubscribe linux-abit". -=
