From owner-freebsd-current Thu Aug 13 22:21:19 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id WAA22247 for freebsd-current-outgoing; Thu, 13 Aug 1998 22:21:19 -0700 (PDT) (envelope-from owner-freebsd-current@FreeBSD.ORG) Received: from vader.cs.berkeley.edu (vader.CS.Berkeley.EDU [128.32.38.234]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id WAA22242 for ; Thu, 13 Aug 1998 22:21:17 -0700 (PDT) (envelope-from asami@vader.cs.berkeley.edu) Received: from silvia.hip.berkeley.edu (sji-ca1-94.ix.netcom.com [209.109.232.94]) by vader.cs.berkeley.edu (8.8.7/8.7.3) with ESMTP id WAA29838; Thu, 13 Aug 1998 22:20:41 -0700 (PDT) Received: (from asami@localhost) by silvia.hip.berkeley.edu (8.8.8/8.6.9) id WAA12851; Thu, 13 Aug 1998 22:20:38 -0700 (PDT) Date: Thu, 13 Aug 1998 22:20:38 -0700 (PDT) Message-Id: <199808140520.WAA12851@silvia.hip.berkeley.edu> To: tom@uniserve.com CC: scrappy@hub.org, freebsd-current@FreeBSD.ORG In-reply-to: (message from Tom on Thu, 13 Aug 1998 10:57:25 -0700 (PDT)) Subject: Re: More PCI Slots... (fwd) From: asami@FreeBSD.ORG (Satoshi Asami) Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG * > > PCI extender boxes have been available for a long time. Not long ago * > > there was a mention of someone who hooked several of these up to a FreeBSD * > > system. That was probably me. We daisy-chained couple of the Bit3 boxes (7 slots each) and put a twin-channel SCSI adapter at the far end. It worked on some of our machines. * > The OS doesn't need to do anything special to recognize them? Are * > there any limitations? The BIOS needs to set up the bridge chips correctly. Note that there can be a full tree of these things hooked up to each other, so it has to recurse into the buses correctly to set up the PCI bus ID registers and such. Obviously there is some point it will stop working if you keep daisy-chaining the boxes because there is only a limited amount of memory the BIOS can use. At the time we tested (more than 3 years ago), Award BIOS didn't work with more than one level of bridging. All of them worked with twin-channel adapters (one bridge in there too) in the host PCI bus though, so they did know about bridge chips. * It is part of the PCI standard. PCI-PCI bridging. I think the chip has to be recognized correctly. Stefan added ours (the IBM chip) to the list. * > At work, we are looking at a Sparc Enterprise 450, which has 10PCI * > slots spread over something like 3-4 PCI busses...do the extende do * > something like that, or is it all one big bus? * * I belive each extender is considered a separate bus, but of course is * linked back to whatever bus it is plugged into. However, some x86 * motherboards have their PCI slots slit over two buses too. There are usually 4 or 5 slots per bus. And the host-expansion cable is also a separate bus. So, for instance, a 7-slot expansion box (which actually only adds 6 slots because one of the host PCI slots is taken) plugged into a 4-slot host will look like this (`*'s are the bridge chips): host | bus 0 bus 2 bus 3 +-+-+-+ +-+-+-+ * +-+-+-+ | | | | | | | | | | | | * * +---------+ bus 1 ^ | this slot is for the host interface card only Satoshi To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message