Re: PC bios recognizing PCI card

["Peter Marek" <peter.marek@marekmicro.de>]

Hi,

you're correct.

1) Chips have to accept config cycles after 2^25 clocks latest. They may
accept it earlier (as most chips do).
2) The BIOS mostly implicitly has such a delay in because it first
initializes the keyboard controller and other legacy devices before scanning
PCI which takes some time. However, some BIOSes provide a Quick Boot option
which vioaltes the 2^25 clock requirement. I.e. that some devices may not be
recognized correctly or at all. So, most BIOSes are strictly speaking not
PCI2.2 compliant.
Especially SRAM FPGA based PCI boards have problems with this.

regards,

Peter Marek
General Director
MarekMicro GmbH
tel.: 049-9621-9732-110
fax: 049-9621-9732-199
www.marekmicro.de

----- Original Message -----
From: "Tisani, Mohamad" <MTisani@pericom.com>
To: "'Richard Walter'" <rwalter@brocade.com>; "'Austin Franklin'"
<austin@darkroom.com>; "PCI SIG Mailing List" <pci-sig@znyx.com>
Sent: Tuesday, April 09, 2002 7:47 PM
Subject: RE: PC bios recognizing PCI card


>
> Thank you for your comments on this issue
>
> It seems to me that the BIOS in the system should be responsible of
Counting
> 2^^25 clocks before starting to configure the PCI Busses
>
> Therefore it should not be a PCI chip's responsibility to ignore all PCI
> Cycles before the 2 ^^25 clocks before accepting cycles
>
> for example I know that the latest 21150 and 21154 bridge from Intel they
> say that are 2.2 Compliant but they do accept Config cycles after 16
clocks
> .
>
> I am not a BIOS expert , is the above correct or not ?
>
>
>
> Regards,
> -----------------------------------------
> Mohamad Tisani                            Tel    :  (408) 435 0800   x347
>
> Director, Application Engineering   Fax  :  (408) 435 1100
> Pericom Semiconductor Corp.       Email: mtisani@pericom.com
> 2380 Bering Dr
> San Jose, Ca 95131                                   www.Pericom.com
>
>
> -----Original Message-----
> From: Richard Walter [mailto:rwalter@brocade.com]
> Sent: Friday, April 05, 2002 4:19 PM
> To: 'Austin Franklin'; PCI SIG Mailing List
> Subject: RE: PC bios recognizing PCI card
>
>
>
> > >
> > > > There is nothing in the PCI spec (2.1 at least) for
> > > > how long a card has to become "configured", per se...
> > >
> > > PCI 2.2, table 4-6, page 128, line 13:
> > > Trhfa - RST# High to First configuration Access - 2^25 clocks.
> > >
> > > PCI 2.2, section 4.3.2, page 134-135 talks about Reset & Trhfa.
> >
> > Right, and that's why I qualified it with 2.1 (I pushed for that to be
> > spec'd BTW, because of this very issue of FPGAs and FPGA
> > configuration)...since there are millions and millions of 2.1 spec'd
> > motherboards out there that most of these boards must be made to work
> > with...it's a good idea to pay attention to the fact it wasn't spec'd in
> 2.1
> > and earlier...and some "older" system board don't meet that spec.
> >
> > Austin
>
> Ooops.  I was writing this right before I had to run to a meeting and hit
> send without finishing my thoughts.  I had left Austin's "(2.1 at least)"
> part in the quote because he is correct that this is a new requirement of
> 2.2.
>
> My intent was not to be critical of Austin's comments but to add more
> details to it.  I appologize if my incomplete thought might have left any
> other impression.
>
> For the record, I agree with both of Austins' posts in this thread so far.
>
> Sincerely,
>
> -Richard Walter
> Note: I speak for myself, not for Brocade Communications.
>
>
>