Stefan Reinauer wrote:
Why is this one needed? Which of the defines actually break the tree? Can't you #define GX2_FOO and LX_FOO instead of FOO and use it based on some compile time or runtime test?
diff -u -r -b -B LinuxBIOSv2/src/southbridge/amd/cs5536/cs5536.c LinuxBIOSv2_cp1/src/southbridge/amd/cs5536/cs5536.c --- LinuxBIOSv2/src/southbridge/amd/cs5536/cs5536.c 2006-07-22 02:21:01.000000000 +0300 +++ LinuxBIOSv2_cp1/src/southbridge/amd/cs5536/cs5536.c 2006-07-31 17:53:04.000000000 +0300 @@ -5,7 +5,7 @@ #include <device/pci_ops.h> #include <device/pci_ids.h> #include <console/console.h> -#include <cpu/amd/gx2def.h> +#include <cpu/amd/lxdef.h> #include <cpu/x86/msr.h> #include "chip.h"
Yes - and the question was what should be that compile time test.
the next one is simple. it is included from here: src/mainboard/amd/rumba/auto.c src/mainboard/olpc/rev_a/auto.c src/mainboard/artecgroup/dbe61/auto.c
so these 3 should be patched to include gx2def.h or lxdef.h accordingly before including cs5536_early_setup.c
This is OK for me.
Question about OLPC: if memtest is set as payload then is it works? We have got dbe61 up and running after some ram_resource() hacking. With ehterbooted Linux the thing is still totally quiet.
thanks, Indrek