RE: EISA/PCI bridge is uni-directional?

["John R Pierce" <pierce@scruznet.com>]

I see one possible problem with this concept right off...
PCI bios... The EISA system bios obviously isn't going to
perform the PCI Bios functionality required to support
PCI peripherals, including PCI board initialization.

If you tried to do this with existing PCI -> EISA bridges, you are 
going to have problems initializing the bridge, its expecting
the host to be on the PCI side to setup the PCI config space
of the bridge...  A 'inverse' bridge (EISA->PCI) will need some
way of generating PCI config cycles, probably via IO ports on
the EISA side...  Hmm.  PCI IRQ routing is another potential
thorny issue.

As you point out, timing is going to be pretty ugly too...

-jrp

----------
From: 	Daniele Beccari
Sent: 	Tuesday, July 23, 1996 5:00 AM
To: 	Mailing List Recipients
Subject: 	EISA/PCI bridge is uni-directional?



Hi,

I have a question about bridging EISA and PCI, (and I probably already know
the answer).

Here is the point:

- An EISA/PCI bridge can be used to add an EISA extension to a "system" PCI
  bus:

  Host bus -- Host/PCI bridge -- PCI Bus -- EISA/PCI bridge -- EISA bus


- What if I wanted to have a PCI bus beyond a "system" EISA bus? 

  Host-bus -- Host/EISA bridge -- EISA bus -- EISA/PCI bridge -- PCI bus

This is ignoring all performance issues (it would make no sense), but could
be 
needed to plug NEW PCI devices on OLD EISA systems.

The question is: Do exisiting EISA/PCI bridges allow to do that?
As far as my knowledge of bridges goes, the answer is: no.

I would appreciate a confirmation from the list, before closing this issue.

Thank you in advance,

		
					Daniele



--
daniele@petrus.grenoble.hp.com



l