Ambiguities in the specification of PC/PCI

["Ali Najafi" <alinajafi@aztech.com.sg>]

There are a few points in the specification of PC/PCI that are not
described
properly and I have doubts how to interpret them. The specification I am
talking about is the document "Mobile PC/PCI DMA Arbitration and Protocols,
April 22, 1996, Revision 2.2", file DMA22.DOC. In this mail, I explain my
understanding and my questions. I appreciate any comments.

1- The serial request and serial grant signals (serial REQ# and GNT#) are
all PC/PCI signals which are not defined by PCI spec. In other words, they
are sideband signals. Is this correct?

2- If item "1" above is correct, by assertion of a serial REQ#, a PCI REQ#
should be generated. Who generates it?

3- As it seems from the spec, PBArbi, the Primary Bus Arbiter of Figure 1
of the spec, both arbitrates for DMA request signals (serial REQ# and
GNT#),
and for PCI request signals (PCI REQ# and GNT#). These two arbitrations,
however, are two separate functions of PBArbi. Is this correct?

4- In section 2.2 of the spec, it is mentioned that serial REQ# should be
de-asserted at least 7 PCICLKs before the final TRDY# assertion, otherwise
an additional I/O transfer may occur. What is the reason?

5- In Figure 21 of the spec, there are two REQ# signals, one is REQ#-Serial
and the other one is REQ#-SIOC, and two GNT# signals, one is GNT#-Serial
and
the other one is GNT#-SIOC. The serial ones are the ones generated by DMA
devices. The SIOC ones are PCI REQ# and GNT# signals. Is this correct? If
so, who generates the REQ#-SIOC and GNT#-SIOC signals? PBArbi?

I would appreciate any clues and comments.

Regards,
Ali Najafi

D1