[LinuxBIOS] node id and core id in internal tree

Lu, Yinghai yinghai.lu at amd.com
Thu Sep 8 01:46:24 CEST 2005


Eric,

 

It seems that cpu_path and cpu_bus_path in include/device/path.h are not
used.

 

So what do you prefer

 

struct apic_path

{

            unsigned apic_id;

            unsigned node id;

            unsigned core_id;

}

 

Or 

 

struct cpu_path

{

            unsigned id;

            unsigned apic_id;

            unsigned node_id;

            unsigned core_id;

}

 

 

YH

 

________________________________

From: linuxbios-bounces at openbios.org
[mailto:linuxbios-bounces at openbios.org] On Behalf Of Lu, Yinghai
Sent: Wednesday, September 07, 2005 9:09 AM
To: Stefan Reinauer; Eric W. Biederman; linuxbios at openbios.org
Subject: RE: [LinuxBIOS] Re: topics for linuxbios summit

 



eric> > Before we generate ACPI or anything else we need to move the
relevant
eric> > information
eric> > into the internal device tree.
eric>
> YH: Good point, But I worry that we make LinuxBIOS more complicated.

Stepan: No, the information has to be reorganized. Currently pirq and
mptables
Stepan: have to be hand crafted, and this is the complicated thing.
Instead the
Stepan: abstract information should be put to the configuration filess,
if it
Stepan: can not be probed.

YH: I suggest to add
1. one node id to ram_resource, so the index will be (idx<<8 | nodeid),
So I can re-use ram_resource in pci_domain for SRAT creating.
2. for apci_cluster_id, will add two fields core_id and node_id, (
instantiate it when cpu_init)...
3. pci device: add pci_irq_routing, and pci_possiable_resource ( for
apci _PRT and _PRS...)

how about ioapci in the internal tree?

YH



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.coreboot.org/pipermail/coreboot/attachments/20050907/e68f1412/attachment.html>


More information about the coreboot mailing list