Re: [LinuxBIOS] bad value in HT UnitID Registers

[Lu, Yinghai]

You need to modify pci1234x to make it have three entries.

And second entry need to mark the 8131 link on K8

And third is for your device with node 0 and link ...

0x00000ff0, 
0x00000200,  // for amd8131 and it will be bus 0x80 (with your device)
or 0x40 (without your HT device)
0x00000100,  // for your HT deive on socket 1 it will be bus 0x40

Hcdnx need three entries too.

Later get ht info for 8131 chain from sysconf.pci1234[1]
And ht info for your ht device will on sysconf.pci1234[2] always.

And you can add some code to handle sysconf.pci1234[2].

YH

-----Original Message-----
From: [e-mail]
[mailto:[e-mail]] On Behalf Of Roman Kononov
Sent: Tuesday, October 10, 2006 12:21 PM
To: LinuxBIOS
Subject: Re: [LinuxBIOS] bad value in HT UnitID Registers

On 10/10/2006 01:46 PM, Lu, Yinghai wrote:
> The option are used to put HT device on specific bus num and device
> logic.

How?
> Are you testing new HT device on socket?

Yes, I am. And the code in get_bus_conf.c does not allow me to do
so unless I modify it to expect the PCI-X bridge on another logical
bus. Essentially, the modified code searches the bridge on the bus
number pci1234[1], if not found, it searches it on the bus number
pci1234[2]. And if the bridge is not there, it is a catastrophe,
unless Tyan makes boards without a PCI-X bridge.

Roman

[LinuxBIOS] bad value in HT UnitID Registers

[Roman Kononov]

On 10/10/2006 03:34 PM, Lu, Yinghai wrote:
> 0x00000ff0,
> 0x00000200,  // for amd8131 and it will be bus 0x80 (with your device)
> or 0x40 (without your HT device)
> 0x00000100,  // for your HT deive on socket 1 it will be bus 0x40

This is tricky. Thank you.
I still do not understand the point of "default HT_CHAIN_UNITID_BASE=0".

On 10/10/2006 03:51 PM, Lu, Yinghai wrote:
> Check more on s2891.., it already support DRC coprocessor.

I wish this were implemented in a more generic way which works in all N-way
motherboards.

Roman