[PATCH 2/2] efi_loader: Fix warnings on unaligned accesses

[Ilias Apalodimas]

Akashi-san

On Fri, 7 Apr 2023 at 10:33, Ilias Apalodimas
<ilias.apalodimas at linaro.org> wrote:
>
> On Fri, 7 Apr 2023 at 04:46, AKASHI Takahiro <takahiro.akashi at linaro.org> wrote:
> >
> > Hi Ilias,
> >
> > On Thu, Apr 06, 2023 at 10:37:07PM +0300, Ilias Apalodimas wrote:
> > > Tom reports that when building with clang we see this warning:
> > > field guid within 'struct efi_hii_keyboard_layout' is less aligned than 'efi_guid_t' and is usually due to 'struct efi_hii_keyboard_layout' being packed, which can lead to unaligned accesses [-Wunaligned-access]
> > >
> > > This happens because 'struct efi_hii_keyboard_layout' is defined as
> > > packed while efi_guid_t is 32bit aligned.
> >
> > There are a couple of 'struct' definitions which are *packed*
> > and contain an 'efi_guid_t' member in efi_api.h.
> > If 'efi_hii_keyboard_layout' is the only place that causes a clang warning,
> > we need a more specific explanation to clarify the problem.
>
> I assumed that all other definitions are aligned regardless of packed,
> i.e they are defined right after a u32, u64, 2xu16 etc, but I'll have
> a closer look

So I did look closer and my assumption was indeed correct.
IOW the warning is the only place in the struct definition where
efi_guid_t happens to be be aligned.

Tom would you like me to send a v2 on this?
I think what happens here is that struct efi_hii_keyboard_layout is
defined as packed, and efi_guid_t is aligned(4).
So clang is trying to tell us: I will generate safe code for unaligned
accesses, but one of the struct members requires specific alignment.

Regards
/Ilias
>
> >
> > > However the EFI spec describes the EFI_GUID as
> > > "128-bit buffer containing a unique identifier value.
> > > Unless otherwise specified aligned on a 64-bit boundary"
> >
> > That's right, but this text in this context may sound misleading.
> > (It doesn't explain why 'efi_guid_t' is 32-bit aligned.)
>
> commit 1dd705cf9903 ("efi: use 32-bit alignment for efi_guid_t")
> explains why, but it's a bit orthogonal to this commit.  In any case
> I'll include it in v2
>
> Thanks
> /Ilias
> >
> > -Takahiro Akashi
> >
> > >
> > > So convert the efi_guid_t -> u8 b[16] here and skip the alignment
> > > requirements.
> > >
> > > Reported-by: Tom Rini <trini at konsulko.com>
> > > Suggested-by: Heinrich Schuchardt <heinrich.schuchardt at canonical.com>
> > > Signed-off-by: Ilias Apalodimas <ilias.apalodimas at linaro.org>
> > > ---
> > >  include/efi_api.h | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/include/efi_api.h b/include/efi_api.h
> > > index 2fd0221c1c77..b84b577bd7b5 100644
> > > --- a/include/efi_api.h
> > > +++ b/include/efi_api.h
> > > @@ -1170,7 +1170,7 @@ struct efi_key_descriptor {
> > >
> > >  struct efi_hii_keyboard_layout {
> > >       u16 layout_length;
> > > -     efi_guid_t guid;
> > > +     u8 guid[16];
> > >       u32 layout_descriptor_string_offset;
> > >       u8 descriptor_count;
> > >       /* struct efi_key_descriptor descriptors[]; follows here */
> > > --
> > > 2.39.2
> > >