[pci] Re: Config Space

["Tilen" <tilen@opencores.org>]

Hi !

If the issue of your question is, that you want all the core registers to be
also
accessed with configuration cycles, it can be changed (usually, there is no
need
for that). But you probably want to gain all 6 base address registers? It is
nothing wrong with that, but PCI specification recommend, that all PCI
devices
map control functions (what core registers are) to memory space. This is
described in PCI 2.1 spec in the last sentence on the page 197 and
in PCI 2.2 spec in the last sentence before chapter 6.2.5.2 on the page 204.
One reason for this is (e.g. you use yours PC PCI card), that your driver
simply configure your card through your memory mapped registers, once
the bios configured yours PCI configuration header.

Regards, Tadej


----- Original Message -----
From: <oliver.amft@ch.abb.com>
To: <pci@opencores.org>
Cc: <tadej@opencores.org>
Sent: Wednesday, May 23, 2001 2:04 PM
Subject: Config Space


>
>
>
> Tadej
>
> Sorry, I should have read the papers more precisely.
> Has there been any public discussion about this extended config space - I
walked
> through the mailing list history but couldn't find anything. So I ask my
> question right away: Why are those configuration registers not in the PCI
> configuration space? This would lead to the maximum config flexibility on
PCI
> side (1-6 possible base address registers).
>
> Cheers,
> Oliver
>
>
>
>
> ----- Original Message -----
>
>
> Subject:  Re: [pci]  PCI IP core
>
> Hi!
>
> Yes, I actually mentioned the extended Config Space (I'm sorry, because
> there is no clear reference to Device Specific Configuration registers -
my
> mistake). Registers in this space are all registers of the core (Device).
> Yes, a part of this registers are also for image mapping, but there are
also
> registers for interrupts, error control and two special registers, which
are
> used for generating the configuration cycles on the PCI bus, but only when
> core is configured as PCI host.
>
> At this point, I must say, that there is SHORTER PDF specification (still
> the
> same revision) on the web, about 660 kB.
>
> Regards, Tadej
>
>
>
>
>