On 14.06.2023 18:36, Shawn Anastasio wrote: > On Wed Jun 14, 2023 at 10:51 AM CDT, Jan Beulich wrote: >> On 13.06.2023 16:50, Shawn Anastasio wrote: >>> --- /dev/null >>> +++ b/xen/arch/ppc/arch.mk >>> @@ -0,0 +1,11 @@ >>> +######################################## >>> +# Power-specific definitions >>> + >>> +ppc-march-$(CONFIG_POWER_ISA_2_07B) := power8 >>> +ppc-march-$(CONFIG_POWER_ISA_3_00) := power9 >>> + >>> +CFLAGS += -mcpu=$(ppc-march-y) -mstrict-align -mcmodel=large -mabi=elfv2 >>> -mno-altivec -mno-vsx >> >> Wouldn't it make sense to also pass -mlittle here, such that a tool >> chain defaulting to big-endian can still be used? > > Good call. On this topic, I suppose I'll also add -m64 to allow 32-bit > toolchains to be used as well.
Turns out this isn't quite enough. When trying to test my little bit of re-basing of Anthony's series, I ran into ld complaining about little endian input when the target format is big endian (like for the compiler I'm using a default configured, i.e. big-endian, binutils build). Looks like we need to pass "-m elf64lppc" to the linker; I'll see if that helps (and where exactly to put it). Jan
