Re: [PATCH v1 2/5] powerpc: Pass correct CPU reference to assembler

["Pali Rohár" <pali@kernel.org>]

On Monday 19 December 2022 19:45:57 Christophe Leroy wrote:
> Jan-Benedict reported issue with building ppc64e_defconfig
> with mainline GCC work:
> 
>   powerpc64-linux-gcc -Wp,-MMD,arch/powerpc/kernel/vdso/.gettimeofday-64.o.d -nostdinc -I./arch/powerpc/include -I./arch/powerpc/include/generated  -I./include -I./arch/powerpc/include/uapi -I./arch/powerpc/include/generated/uapi -I./include/uapi -I./include/generated/uapi -include ./include/linux/compiler-version.h -include ./include/linux/kconfig.h -D__KERNEL__ -I ./arch/powerpc -DHAVE_AS_ATHIGH=1 -fmacro-prefix-map=./= -D__ASSEMBLY__ -fno-PIE -m64 -Wl,-a64 -mabi=elfv1 -Wa,-me500 -Wa,-me500mc -mabi=elfv1 -mbig-endian    -Wl,-soname=linux-vdso64.so.1 -D__VDSO64__ -s -c -o arch/powerpc/kernel/vdso/gettimeofday-64.o arch/powerpc/kernel/vdso/gettimeofday.S
> 	arch/powerpc/kernel/vdso/gettimeofday.S: Assembler messages:
> 	arch/powerpc/kernel/vdso/gettimeofday.S:72: Error: unrecognized opcode: `stdu'
> 	arch/powerpc/kernel/vdso/gettimeofday.S:72: Error: unrecognized opcode: `stdu'
> 	arch/powerpc/kernel/vdso/gettimeofday.S:72: Error: unrecognized opcode: `std'
> 	arch/powerpc/kernel/vdso/gettimeofday.S:72: Error: unrecognized opcode: `std'
> 	arch/powerpc/kernel/vdso/gettimeofday.S:72: Error: unrecognized opcode: `ld'
> 	arch/powerpc/kernel/vdso/gettimeofday.S:72: Error: unrecognized opcode: `ld'
> 	...
> 	make[1]: *** [arch/powerpc/kernel/vdso/Makefile:76: arch/powerpc/kernel/vdso/gettimeofday-64.o] Error 1
> 	make: *** [arch/powerpc/Makefile:387: vdso_prepare] Error 2
> 
> This is due to assembler being called with -me500mc which is
> a 32 bits target.
> 
> The problem comes from the fact that CONFIG_PPC_E500MC is selected for
> both the e500mc (32 bits) and the e5500 (64 bits), and therefore the
> following makefile rule is wrong:
> 
>   cpu-as-$(CONFIG_PPC_E500MC)    += $(call as-option,-Wa$(comma)-me500mc)
> 
> Today we have CONFIG_TARGET_CPU which provides the identification of the
> expected CPU, it is used for GCC. Once GCC knows the target CPU, it adds
> the correct CPU option to assembler, no need to add it explicitely.
> 
> With that change (And also commit 825eada7717c ("powerpc/64: Set
> default CPU in Kconfig")), it now is:
> 
>   powerpc64-linux-gcc -Wp,-MMD,arch/powerpc/kernel/vdso/.gettimeofday-64.o.d -nostdinc -I./arch/powerpc/include -I./arch/powerpc/include/generated  -I./include -I./arch/powerpc/include/uapi -I./arch/powerpc/include/generated/uapi -I./include/uapi -I./include/generated/uapi -include ./include/linux/compiler-version.h -include ./include/linux/kconfig.h -D__KERNEL__ -I ./arch/powerpc -DHAVE_AS_ATHIGH=1 -fmacro-prefix-map=./= -D__ASSEMBLY__ -fno-PIE -m64 -Wl,-a64 -mabi=elfv1 -mcpu=e500mc64 -mabi=elfv1 -mbig-endian    -Wl,-soname=linux-vdso64.so.1 -D__VDSO64__ -s -c -o arch/powerpc/kernel/vdso/gettimeofday-64.o arch/powerpc/kernel/vdso/gettimeofday.S
> 
> Reported-by: Jan-Benedict Glaw <jbglaw@lug-owl.de>
> Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>

Acked-by: Pali Rohár <pali@kernel.org>

> ---
> v3: Remove -Wa,-mXXX completery as there are additional mismatches between -mcpu and -Wa-m (eg: 8540 vs e500). When given the proper -mcpu=, GCC gives the correct option to assembler.
> 
> v2: When target cpu is powerpc, the option to be used is -mppc
> 
> Commit 825eada7717c is in powerpc/next-test branch. Make sure the SHA doesn't change when it goes into powerpc/next
> ---
>  arch/powerpc/Makefile | 6 +-----
>  1 file changed, 1 insertion(+), 5 deletions(-)
> 
> diff --git a/arch/powerpc/Makefile b/arch/powerpc/Makefile
> index bf5f0a998273..0f9f291895cb 100644
> --- a/arch/powerpc/Makefile
> +++ b/arch/powerpc/Makefile
> @@ -201,18 +201,14 @@ KBUILD_CFLAGS += -fno-asynchronous-unwind-tables
>  # often slow when they are implemented at all
>  KBUILD_CFLAGS		+= $(call cc-option,-mno-string)
>  
> -cpu-as-$(CONFIG_40x)		+= -Wa,-m405
> -cpu-as-$(CONFIG_44x)		+= -Wa,-m440
>  cpu-as-$(CONFIG_ALTIVEC)	+= $(call as-option,-Wa$(comma)-maltivec)
> -cpu-as-$(CONFIG_PPC_E500)		+= -Wa,-me500
>  
>  # When using '-many -mpower4' gas will first try and find a matching power4
>  # mnemonic and failing that it will allow any valid mnemonic that GAS knows
>  # about. GCC will pass -many to GAS when assembling, clang does not.
>  # LLVM IAS doesn't understand either flag: https://github.com/ClangBuiltLinux/linux/issues/675
>  # but LLVM IAS only supports ISA >= 2.06 for Book3S 64 anyway...
> -cpu-as-$(CONFIG_PPC_BOOK3S_64)	+= $(call as-option,-Wa$(comma)-mpower4) $(call as-option,-Wa$(comma)-many)
> -cpu-as-$(CONFIG_PPC_E500MC)	+= $(call as-option,-Wa$(comma)-me500mc)
> +cpu-as-$(CONFIG_PPC_BOOK3S_64)	+= $(call as-option,-Wa$(comma)-many)
>  
>  KBUILD_AFLAGS += $(cpu-as-y)
>  KBUILD_CFLAGS += $(cpu-as-y)
> -- 
> 2.38.1
>