Date: Fri, 3 Jan 1997 12:03:42 -0500
From: cbreece@telogy.com (Cris Breece)
Subject: Item #399: PCI Binding; map-in/out clarification.


P1275 Openboot Working Group Proposal -- Proposal #:399 Ver 1

Title:          PCI binding, v2.0, map-in & map-out clarification.

Author:         Cris Breece

Date:           1997.01.03

Ed/Tech:        Technical

Synopsis:       Undefined behavior of map-in & map-out in FCode
Evaluation Semantics sec.

Doc & Version:  None

Problem:

A line in document ``PCI Bus Binding to: IEEE Std 1275-1994 Standard for
Boot (Initialization Configuration) Firmware revision 2.0, Section 2.5
FCode Evaluation Semantics, page 8 line 29, states:

Enter probe state, thus affecting subsequent behavior of the map-in and
map-out methods.

The affect to the behavior of these methods is not defined in said
document.  This seems to imply that the behavior of map-in and map-out
change in an unknown, undefined manner.  Implying that the methods should
not be used in probe state.

After reviewing other documents, and examining examples of drivers, the
conclusion is that this is simply a ill-phrased statement that should be
reworded.


Proposal:

ASSUMING:  That what is being talked about, is that an associated CPU
binding, as with the PowerPC CPU, might dictate a different operational or
addressing mode while in probe state.  Which would affect the map-in and
map-out methods.  Change the phrasing to:

Enter probe state, possibly affecting subsequent behavior of the map-in
and map-out methods.  Reference appropriate CPU binding document for
details.

Possibly even giving the ``Memory Management'' section, if it exists in all
CPU binding documents, as the area in which to look for clarification.




[ P1275 Item #399 -- Received: Fri Jan  3 13:06:07 PST 1997 ]