[coreboot] how to delete symbol link created at compile time

Marc Jones marcj303 at gmail.com
Tue Oct 18 04:38:31 CEST 2011


On Mon, Oct 17, 2011 at 4:44 PM, Stefan Reinauer
<stefan.reinauer at coreboot.org> wrote:
> * Marc Jones <marcj303 at gmail.com> [111016 10:10]:
>> >> > I have created 2 devicetree file :
>> >> >
>> >> > devicetree_f15.cb for platform with family 15 CPU
>> >> >
>> >> > devicetree_f10.cb  for platform with family 10 CPU
>> >> >
>> >> >
>> >> >
>> >> > I changed the makefile to create a symbol link “devicetree.cb” link to
>> >> > devicetree_f10.cb or devicetree_f15.cb at compile time.
>> >> >
>> >> > The problem is that I can’t delete the symbol link when make
>> >> > clean/distclean.
>> >
>> > Please fix the problem by using one device tree for both platforms.
>
>> Stefan,
>>
>> Can you explain your thoughts on how that would work? Can we put a #if
>> in the devicetree.cb? It uses the c precompiler? It requires different
>> CPU files/device locations.  We can try it next week.
>
> Usually the way this is handled in coreboot is that there is one socket
> that binds together all CPU types. Then in the device tree only the
> socket type is specified, and code for both CPUs is pulled in.
> Maybe we need something like a socket for northbridge code, since the
> northbridge now lives in the CPU?
>
> It seems like a bad idea to have to recompile your BIOS because you
> change the CPU. We did a lot of nastyness with K8 and Fam10, but we
> should find a better way to do this for future chipsets/CPUs.


Yes, we are discussing how the AGESA code would work. The socket
decision is rather complicated as we need a way to handle multiple
calls with the same names (function point tables etc). I think that
there may be a solution within AGESA, but the device tree may still be
a problem as the CPUs have different HT link numbering. This makes it
hard to have the same device tree layout for the same socket.


Marc


-- 
http://se-eng.com




More information about the coreboot mailing list