[LinuxBIOS] PCI: Left over static devices. Check your Config.lb

ron minnich rminnich at gmail.com
Mon Dec 10 22:19:58 CET 2007

On Dec 10, 2007 10:08 AM, Steve Isaacs <yasteve at gmail.com> wrote:

> This is part of my confusion. I installed a Phoenix BIOS to boot Linux
> and be able to run lspci. I have no visibility into what it's (Phoenix
> BIOS) design is for this board. I suspect there are several devices it
> hides (a feature of the chipset) because the assumption is they won't be
> needed for the reference board.

once again. You probably don't need a lot of those "pci 6.0 on" bits.
yank them out.

Your board is not working, and the best thing you can do is shrink
down Config.lb until it works.

You need the ioapic, the 18. devices, and the superio. try starting there.

> >       device pci 18.0 on end                    # Link 1
> >       device pci 18.0 on end

each 18. device connects to three HT channels. There's no real way to
express this in PCI, so we got stuck with the three instances.

You need 3 19. devices since the 19. bits (CPU1) are configured from
CPU0 and they each have 3 HT links.

> I couldn't find an explanation for this in the PDF you mentioned. There
> is an example showing 19.0 listed three times but no comment as to why.
> Is this really part of hyper-transport configuration? The first for link
> 0, the second for link 1 and the third for link 2 (listed as LDT0, LDT1,
> and LDT2 in the PDF)?

that's it!

> One thing that keeps tripping me is it appears that some device numbers
> are 0 based and others are 1 based. For example 18.0 agrees with a PCI
> bus scan as well as 19.0 but 6.0 in the Config shows up as 7.0 in the
> scan and 0.0 as 1.0. What's up with that? Is there a rule I need to
> understand? On the surface this seems very inconsistent.

The rule is that difference bios'es will order busses differently.
hence the weirdness.
Shrink the config.lb to start.


More information about the coreboot mailing list