[coreboot] Change in coreboot[master]: console: Skip non-BSP printk() earlier

[Kyösti Mälkki]

On Sun, May 6, 2018 at 4:48 PM, Jonathan A. Kollasch (Code Review)
<gerrit at coreboot.org> wrote:
> I'm skeptical this will address my issue on msi/ms7135, as it's a
> Socket 754 board that only single-core CPU packages are available
> for. Nevertheless, I'll give this change a try.
>
> Indeed, it does not fix the issue I'm having.
>

Apologies for the spam/duplicate some may have received, I did not
check recipients of the previous post.

Right.. I ack the no-go.

Try if DEBUG_CAR=y (could be in menuconfig) works before/after the
regression commit.

My guess is it fails in cpu/amd/car/post_cache_as_ram.c:101, that's
about the first place where car_get_var() goes wrong on systems with
LATE_CBMEM_INIT. IMHO it should not even have built when there is any
CAR_GLOBAL symbol present.

I would rather not see any EARLY_CBMEM_INIT or LATE_CBMEM_INIT tests
added to the tree today. Those K8 were meant to be left behind last
October. Once they are all removed from master, fixing those "*.c"
includes and CAR migration can bring your board back...