[coreboot] GCC update broke AMD Fam10h boot

Aaron Durbin adurbin at chromium.org
Tue Mar 17 03:19:36 CET 2015


On Mon, Mar 16, 2015 at 4:33 PM, Stefan Reinauer
<stefan.reinauer at coreboot.org> wrote:
> * Aaron Durbin <adurbin at chromium.org> [150316 22:44]:
>> A quick hack is add ALIGN(32) to the linker script before
>> _bs_init_begin: src/arch/x86/ramstage.ld
>>
>> But I think we'll need to store pointers to the structures in order to
>> properly handle the situation where the compiler is effectively making
>> alignment/size decisions for some reason.
>
> I suggest trying to enforce alignment / size instead of adding another
> layer of indirection.

I'm not sure that's possible w/o marking the struct packed just for
the sake of it. The other issue is that this section is sitting in RO
while also being written to. The other thing, which has been known for
awhile, is that we can't take advantage of symbols being equal for an
empty set since C says no two symbols can be the same. I'm sure the
language lawyers will correct me where I got that wrong.



More information about the coreboot mailing list