[coreboot] domain vs device statictree order

Myles Watson mylesgw at gmail.com
Thu Jan 8 17:08:14 CET 2009


On Thu, Jan 8, 2009 at 8:29 AM, Mart Raudsepp
<mart.raudsepp at artecdesign.ee> wrote:
> Ühel kenal päeval, N, 2009-01-08 kell 02:11, kirjutas Peter Stuge:
>> Myles Watson wrote:
>> > This patch implements preorder traversal for Phase 2 and Phase 6, and prints a
>> > warning that Phase 1 is obsolete if anyone implements one.
>> >
>> > The only difference between the Phase 2 and Phase 6 implementations is
>> > that phase2_fixup is always called if it's defined, but phase6_init is
>> > only called if the device is enabled.
>> >
>> > Note that any devices not found in the tree will not have their init
>> > functions called with this patch.  I think that's a good thing, but it
>> > will require some dts fixes.
>
> What's the easy way to find out which devices could be missing from the
> tree?

If you look at the device tree that's printed "after phase 3", any
device that is dynamic wasn't found in the device tree.  PCI cards
that you plug in will never be in the dts, so they will always be
dynamic.

>> > Boot tested on Serengeti and qemu.
>> >
>> > Signed-off-by: Myles Watson <mylesgw at gmail.com>
>>
>> Acked-by: Peter Stuge <peter at stuge.se>
>
> This patch is necessary to get GeodeLX to boot again (DBE61). rev1103
> had broken it with the memory controller device separation being
> committed before, as stuff appears to not be set up right and it resets
> in the end parts of stage2. Peter suggested the memory controller isn't
> set up right, and something like that appears to be the case then, as
> this patch here fixed it up again with the execution order being
> changed.

I'm glad it fixes it.  Thanks for testing.

Myles




More information about the coreboot mailing list