Date: Sun, 21 Mar 1999 20:58:21 +0100 From: Stefan Esser <se@mi.uni-koeln.de> To: David Dawes <dawes@rf900.physics.usyd.edu.au> Cc: hackers@freebsd.org, Stefan Esser <se@freebsd.org> Subject: Re: !! Emergency !! Help FreeBSD 3.0 with IBM Netfinity 5000 Message-ID: <19990321205821.B557@dialup124.mi.uni-koeln.de> In-Reply-To: <19990321132356.M11159@rf900.physics.usyd.edu.au>; from David Dawes on Sun, Mar 21, 1999 at 01:23:56PM %2B1100 References: <4825673A.00577C58.00@tw.ibm.com> <19990321132356.M11159@rf900.physics.usyd.edu.au>
next in thread | previous in thread | raw e-mail | index | archive | help
On 1999-03-21 13:23 +1100, David Dawes <dawes@rf900.physics.usyd.edu.au> wrote: > >Probing for devices on PCI bus 0: > >chip0: <Host to PCI bridge (vendor=1166 device=0007)> rev 0x04 on pci0.0.0 > >chip1: <Host to PCI bridge (vendor=1166 device=0005)> rev 0x02 on pci0.0.1 > > I guess the problem is multiple host-PCI bridges. This is something we > need to deal with at XFree86 too. Is it correct that there is no > generic way of handling these? While PCI to PCI bridges are well defined, there is no standard for the position of the secondary bus /subordinate bus registers for host to PCI bridges that connect to multiple PCI buses. The standard asks for one bus with busno=0, where all the other buses are connected to through PCI to PCI bridges. But that is not what (good) systems with many PCI slots have, since each level of PCI to PCI bridges adds latency. Now there is the problem, that the information about the bus range behind each of the CPU to PCI bridges is encoded in the device private range of configuration space, and differently in each device. (If only one CPU to PCI bridge exists, then buses 0 to 255 are reached through that bridge, and there is no need to specify that range. But even the Intel Saturn chip set back in 1994 had secondary/subordinate bus register in order to support multiple host bridges. Too bad these registers have not been declared mandatory for host bridges, and are hardwired to appropriate values (secondary bus=1, subordinate bus=255) in chip sets that do not support multiple directly attached busese ... Regards, STefan To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19990321205821.B557>