Re: [PATCH v3 2/5] efi: Export arch_tables variable

[Daniel Kiper <daniel.kiper@oracle.com>]

On Wed, Mar 26, 2014 at 01:21:19PM +0000, Jan Beulich wrote:
> >>> On 25.03.14 at 21:57, <daniel.kiper@oracle.com> wrote:
> > Export arch_tables variable. Xen init function calls efi_config_init()
> > which takes it as an argument.
> >
> > Additionally, put __initdata in place suggested by include/linux/init.h.
>
> Which isn't necessarily the most appropriate place.

Why? If comments in include/linux/init.h are not valid they should be changed.

> > --- a/arch/x86/platform/efi/efi.c
> > +++ b/arch/x86/platform/efi/efi.c
> > @@ -70,7 +70,7 @@ static efi_system_table_t efi_systab __initdata;
> >
> >  unsigned long x86_efi_facility;
> >
> > -static __initdata efi_config_table_type_t arch_tables[] = {
> > +efi_config_table_type_t arch_tables[] __initdata = {
>
> efi_config_table_type_t __initdata arch_tables[] = {
>
> would be what I'd recommend.
>
> > --- a/include/linux/efi.h
> > +++ b/include/linux/efi.h
> > @@ -583,6 +583,8 @@ extern struct efi {
> >  	struct efi_memory_map *memmap;
> >  } efi;
> >
> > +extern efi_config_table_type_t arch_tables[] __initdata;
>
> And section placement annotations are bogus on declarations.

Hmmm... I am not sure which approach is better. I saw that
in many places declarations have annotations. Could you
point me some docs which states (and explains) that this
is wrong idea.

Daniel