[PATCH v2] gas: Allow SHF_GNU_RETAIN on .note/.rodata.str sections

[Alan Modra]

On Mon, Feb 15, 2021 at 06:20:19PM -0800, H.J. Lu wrote:
> On Mon, Feb 15, 2021 at 3:28 PM Alan Modra <amodra@gmail.com> wrote:
> >
> > On Mon, Feb 15, 2021 at 07:43:39AM -0800, H.J. Lu wrote:
> > > On Sat, Feb 13, 2021 at 7:25 AM H.J. Lu <hjl.tools@gmail.com> wrote:
> > > >
> > > > Since SHF_GNU_RETAIN is allowed on SHT_NOTE section, strip SHF_GNU_RETAIN
> > > > when checking incorrect section attributes on SHT_NOTE section.
> > > >
> > > >         PR gas/27412
> > > >         * config/obj-elf.c (obj_elf_change_section): Strip SHF_GNU_RETAIN.
> > > >         * testsuite/gas/elf/elf.exp: Run section28.
> > > >         * testsuite/gas/elf/section28.d: New file.
> > > >         * testsuite/gas/elf/section28.s: Likewise.
> > > > ---
> > > >  gas/config/obj-elf.c              |  8 +++++++-
> > > >  gas/testsuite/gas/elf/elf.exp     |  1 +
> > > >  gas/testsuite/gas/elf/section28.d | 15 +++++++++++++++
> > > >  gas/testsuite/gas/elf/section28.s | 11 +++++++++++
> > > >  4 files changed, 34 insertions(+), 1 deletion(-)
> > > >  create mode 100644 gas/testsuite/gas/elf/section28.d
> > > >  create mode 100644 gas/testsuite/gas/elf/section28.s
> > > >
> > > > diff --git a/gas/config/obj-elf.c b/gas/config/obj-elf.c
> > > > index f52dc69f0c3..13f54b10857 100644
> > > > --- a/gas/config/obj-elf.c
> > > > +++ b/gas/config/obj-elf.c
> > > > @@ -667,12 +667,18 @@ obj_elf_change_section (const char *name,
> > > >                                         | SHF_MASKPROC))
> > > >                               & ~ssect->attr) != 0)
> > > >         {
> > > > +         /* Strip SHF_GNU_RETAIN.  */
> > > > +         bfd_vma generic_attr = attr;
> > > > +         if (elf_tdata (stdoutput)->has_gnu_osabi)
> > > > +           generic_attr &= ~SHF_GNU_RETAIN;
> > > > +
> > > >           /* As a GNU extension, we permit a .note section to be
> > > >              allocatable.  If the linker sees an allocatable .note
> > > >              section, it will create a PT_NOTE segment in the output
> > > >              file.  We also allow "x" for .note.GNU-stack.  */
> > > >           if (ssect->type == SHT_NOTE
> > > > -             && (attr == SHF_ALLOC || attr == SHF_EXECINSTR))
> > > > +             && (generic_attr == SHF_ALLOC
> > > > +                 || generic_attr == SHF_EXECINSTR))
> > > >             ;
> > > >           /* Allow different SHF_MERGE and SHF_STRINGS if we have
> > > >              something like .rodata.str.  */
> >
> > Should you be using generic_attr in later tests in this block?  If we
> > allow .rodata.str with a differing SHF_MERGE flag then it seems
> > reasonable to also allow SHF_GNU_RETAIN to differ.
> >
> 
> Like this?

Why not use generic_attr in *all* of the tests in this block?  I think
that would be better.  OK with that change.

-- 
Alan Modra
Australia Development Lab, IBM