Re: [PATCH 0/8] fix gcc9 build with -Werror=address-of-packed-member

[Michael Chang <mchang@suse.com>]

On Tue, Apr 09, 2019 at 10:15:33PM +1000, Vladimir 'phcoder' Serbinenko wrote:
> On Tue, Apr 9, 2019 at 9:40 PM Michael Chang <mchang@suse.com> wrote:
> >
> > On Tue, Apr 09, 2019 at 08:57:25PM +1000, Vladimir 'phcoder' Serbinenko wrote:
> > > On Tue, Apr 9, 2019 at 8:48 PM Michael Chang <mchang@suse.com> wrote:
> > > >
> > > > This patch set attempts to resolve the build failure in openSUSE build
> > > > service equipped with new gcc 9 compiler, which has added a new warning
> > > > flag -Waddress-of-packed-member.
> > > >
> > > > The new warning performs the check for taking the address of packed
> > > > member directly to a pointer variable with higher alignment requirement
> > > > and the outcome is risky to memory alignment fault on some architecture
> > > > when deferencing it.
> > > The problem here is that there is no way to tell the compiler to
> > > create a structure without padding but a pointer to which is aligned.
> > > This is very common in GRUB. Consider a struct like following:
> > > grub_uint16_t a;
> > > grub_uint32_t b;
> > > grub_uint16_t c[40];
> > >
> > > This needs packed attribute as otherwise compiler will insert a
> > > padding before b. Yet c is guaranteed to be aligned if the pointer to
> > > the structure is aligned.
> >
> > But sometimes it is not easy to know as the packed structure could be
> > embedded to another and also it could map to a specific portion in
> > memory read from a file, with no guarantee where will be started (for eg
> > via searching occurrence of specific magic number ..)
> Yes. It's not easy. But that's why just following warnings isn't
> always the good way to go [without going into specific details for
> particular patches].

OK. Point taken. I'll go through the pathes again and ignore the
warnings if the diagnostic is found to be false positive.

Thanks,
Michael

> >
> > Thanks,
> > Michael
> >
> > > >
> > > > Please help to review.
> > > >
> > > > Thanks.
> > > >
> > > > Michael Chang (8):
> > > >   cpio: fix gcc9 error address-of-packed-member
> > > >   jfs: fix gcc9 error address-of-packed-member
> > > >   hfs: fix gcc9 error address-of-packed-member
> > > >   hfsplus: fix gcc9 error address-of-packed-member
> > > >   acpi: fix gcc9 error address-of-packed-member
> > > >   usbtest: fix gcc9 error address-of-packed-member
> > > >   chainloader: fix gcc9 error address-of-packed-member
> > > >   efi: fix gcc9 error address-of-packed-member
> > > >
> > > >  grub-core/commands/usbtest.c       | 13 ++++++++-
> > > >  grub-core/fs/cpio.c                |  5 ++--
> > > >  grub-core/fs/cpio_be.c             |  5 ++--
> > > >  grub-core/fs/hfsplus.c             | 57 ++++++++++++++++++++++++++------------
> > > >  grub-core/fs/jfs.c                 |  5 ++--
> > > >  grub-core/kern/efi/efi.c           | 27 ++++++++++++++++--
> > > >  grub-core/loader/efi/chainloader.c | 12 ++++++--
> > > >  include/grub/acpi.h                |  2 +-
> > > >  include/grub/hfs.h                 |  2 +-
> > > >  9 files changed, 96 insertions(+), 32 deletions(-)
> > > >
> > > > --
> > > > 2.16.4
> > > >
> > > >
> > > > _______________________________________________
> > > > Grub-devel mailing list
> > > > Grub-devel@gnu.org
> > > > https://lists.gnu.org/mailman/listinfo/grub-devel
> > >
> > >
> > >
> > > --
> > > Regards
> > > Vladimir 'phcoder' Serbinenko
> > >
> > > _______________________________________________
> > > Grub-devel mailing list
> > > Grub-devel@gnu.org
> > > https://lists.gnu.org/mailman/listinfo/grub-devel
> > >
> >
> > _______________________________________________
> > Grub-devel mailing list
> > Grub-devel@gnu.org
> > https://lists.gnu.org/mailman/listinfo/grub-devel
> 
> 
> 
> -- 
> Regards
> Vladimir 'phcoder' Serbinenko
> 
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel