[PATCH -next] arm64: alternatives: Use vdso/bits.h instead of linux/bits.h

[Will Deacon]

On Mon, Oct 03, 2022 at 12:37:59PM -0700, Nathan Chancellor wrote:
> When building with CONFIG_LTO after commit ba00c2a04fa5 ("arm64: fix the
> build with binutils 2.27"), the following build error occurs:
> 
>   In file included from arch/arm64/kernel/module-plts.c:6:
>   In file included from include/linux/elf.h:6:
>   In file included from arch/arm64/include/asm/elf.h:8:
>   In file included from arch/arm64/include/asm/hwcap.h:9:
>   In file included from arch/arm64/include/asm/cpufeature.h:9:
>   In file included from arch/arm64/include/asm/alternative-macros.h:5:
>   In file included from include/linux/bits.h:22:
>   In file included from include/linux/build_bug.h:5:
>   In file included from include/linux/compiler.h:248:
>   In file included from arch/arm64/include/asm/rwonce.h:71:
>   include/asm-generic/rwonce.h:67:9: error: expected string literal in 'asm'
>           return __READ_ONCE(*(unsigned long *)addr);
>                 ^
>   arch/arm64/include/asm/rwonce.h:43:16: note: expanded from macro '__READ_ONCE'
>                   asm volatile(__LOAD_RCPC(b, %w0, %1)                    \
>                               ^
>   arch/arm64/include/asm/rwonce.h:17:2: note: expanded from macro '__LOAD_RCPC'
>           ALTERNATIVE(                                                    \
>           ^
> 
> Similar to the issue resolved by commit 0072dc1b53c3 ("arm64: avoid
> BUILD_BUG_ON() in alternative-macros"), there is a circular include
> dependency through <linux/bits.h> when CONFIG_LTO is enabled due to
> <asm/rwonce.h> appearing in the include chain before the contents of
> <asm/alternative-macros.h>, which results in ALTERNATIVE() not getting
> expanded properly because it has not been defined yet.
> 
> Avoid this issue by including <vdso/bits.h>, which includes the
> definition of the BIT() macro, instead of <linux/bits.h>, as BIT() is the
> only macro from bits.h that is relevant to this header.
> 
> Fixes: ba00c2a04fa5 ("arm64: fix the build with binutils 2.27")
> Link: https://github.com/ClangBuiltLinux/linux/issues/1728
> Signed-off-by: Nathan Chancellor <nathan at kernel.org>
> ---
>  arch/arm64/include/asm/alternative-macros.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/arm64/include/asm/alternative-macros.h b/arch/arm64/include/asm/alternative-macros.h
> index b5ac0b85c9bb..3622e9f4fb44 100644
> --- a/arch/arm64/include/asm/alternative-macros.h
> +++ b/arch/arm64/include/asm/alternative-macros.h
> @@ -2,8 +2,8 @@
>  #ifndef __ASM_ALTERNATIVE_MACROS_H
>  #define __ASM_ALTERNATIVE_MACROS_H
>  
> -#include <linux/bits.h>
>  #include <linux/const.h>
> +#include <vdso/bits.h>
>  
>  #include <asm/cpucaps.h>
>  #include <asm/insn-def.h>
> 
> base-commit: ba00c2a04fa5431d204a4183062b30372c062aa7

The 'vdso' namespace is a bit of a misnomer, but that's not your fault and I
can confirm that this fixes the build on for-next/core with defconfig + thin
LTO, so:

Tested-by: Will Deacon <will at kernel.org>

Thanks!

Will