Re: [PATCH] ARM: mark prepare_page_table as __init

[Ard Biesheuvel]

On Tue, 11 May 2021 at 11:04, Arnd Bergmann <arnd@xxxxxxxxxx> wrote:
>
> From: Arnd Bergmann <arnd@xxxxxxxx>
>
> In some configurations when building with gcc-11, prepare_page_table
> does not get inline, which causes a build time warning for a section
> mismatch:
>
> WARNING: modpost: vmlinux.o(.text.unlikely+0xce8): Section mismatch in reference from the function prepare_page_table() to the (unknown reference) .init.data:(unknown)
> The function prepare_page_table() references
> the (unknown reference) __initdata (unknown).
> This is often because prepare_page_table lacks a __initdata
> annotation or the annotation of (unknown) is wrong.
>
> Mark the function as __init to avoid the warning regardless of the
> inlining.
>
> Signed-off-by: Arnd Bergmann <arnd@xxxxxxxx>

I think the inline should be dropped here. Either inlining is
absolutely required for correct code generation (which is rare), or
the inline is optional, and better left up to the compiler, especially
given that this is __init code so the time/space tradeoff is moot
anyway.

With or without that change,

Acked-by: Ard Biesheuvel <ardb@xxxxxxxxxx>

> ---
>  arch/arm/mm/mmu.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/arch/arm/mm/mmu.c b/arch/arm/mm/mmu.c
> index 051f4f82414b..5220b8147f40 100644
> --- a/arch/arm/mm/mmu.c
> +++ b/arch/arm/mm/mmu.c
> @@ -1246,7 +1246,7 @@ void __init adjust_lowmem_bounds(void)
>         memblock_set_current_limit(memblock_limit);
>  }
>
> -static inline void prepare_page_table(void)
> +static inline __init void prepare_page_table(void)
>  {
>         unsigned long addr;
>         phys_addr_t end;
> --
> 2.29.2
>