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

[Ilias Apalodimas]

On Thu, 20 Apr 2023 at 10:16, AKASHI Takahiro
<takahiro.akashi at linaro.org> wrote:
>
> On Thu, Apr 20, 2023 at 09:35:42AM +0300, Ilias Apalodimas wrote:
> > 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.
>
> My concern is that we use char[] in one place and efi_guid_t elsewhere.
> It sounds arbitrary without any clear explanation.

I can send a v2 adding a comment, but I changed my mind as well.  I
think explicitly disabling the warning in such places (as Tom did on
his original patch) is a better solution.
We still have to add a comment about why, but I'd prefer keeping a
consistent efi_guid_t as well


Regards
/Ilias
>
> -Takahiro Akashi
>
>
> > 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
> > > > >