On Thu, Oct 01, 2020 at 09:09:07PM +0930, Alan Modra via Binutils wrote:
> On Thu, Oct 01, 2020 at 11:50:33AM +0100, Jozef Lawrynowicz wrote:
> > --- a/bfd/elflink.c
> > +++ b/bfd/elflink.c
> > @@ -14102,7 +14102,8 @@ bfd_elf_gc_sections (bfd *abfd, struct bfd_link_info *info)
> >  			|| (elf_section_data (o)->this_hdr.sh_type
> >  			    == SHT_FINI_ARRAY)))
> >  		|| (elf_section_data (o)->this_hdr.sh_type == SHT_NOTE
> > -		    && elf_next_in_group (o) == NULL )))
> > +		    && elf_next_in_group (o) == NULL)
> > +		|| (elf_section_flags (o) & SHF_GNU_RETAIN)))
> 
> Flag bits in SHF_MASKOS depend on OS, so this needs a test of OSABI.
> That can be done by checking elf_tdata (sub)->has_gnu_osabi for the
> appropriate bit.
> 

Fixed the two OSABI issues in the attached patch.

I also added handling of numeric flag values in the SHF_MASKOS range for
the .section directive for some OSABIs.
There are new GAS tests to test this behavior for targets that either do
or do not support the GNU OSABI.

> >  	  {
> >  	    if (!_bfd_elf_gc_mark (info, o, gc_mark_hook))
> >  	      return FALSE;
> > diff --git a/binutils/NEWS b/binutils/NEWS
> > index c0dc73d7d8..6c7d3f3953 100644
> > --- a/binutils/NEWS
> > +++ b/binutils/NEWS
> > @@ -4,6 +4,10 @@
> >    symbol names.  In addition the --demangle=<style>, --no-demangle,
> >    --recurse-limit and --no-recurse-limit options are also now availale.
> >  
> > +* Add support for the SHF_GNU_RETAIN ELF section flag.
> > +  This flag specifies that the section should not be garbage collected by the
> > +  linker if it is unused.
> 
> I would drop "if it is unused".  The phrase doesn't really add
> anything to a user's understanding.
> 

I don't really have a strong opinion either way, but I thought "if it is
unused" might be helpful to add to the description since "garbage
collection" is not precisely defined in ELF.

However, since the flag is only part of the GNU gABI, and the GNU linker
supports "garbage collection", it probably does not need the extra
clarification, so I'm fine with removing it.

Since there was consensus for that explicit wording on the gABI thread
(https://sourceware.org/pipermail/gnu-gabi/2020q3/000449.htm), I'll wait
until next week to see if there are any objections before adjusting it.

Thanks,
Jozef

> > --- a/binutils/readelf.c
> > +++ b/binutils/readelf.c
> > @@ -6010,6 +6012,7 @@ get_elf_section_flags (Filedata * filedata, bfd_vma sh_flags)
> >  	    case SHF_EXCLUDE:		sindex = 18; break;
> >  	    case SHF_COMPRESSED:	sindex = 20; break;
> >  	    case SHF_GNU_MBIND:		sindex = 24; break;
> > +	    case SHF_GNU_RETAIN:	sindex = 26; break;
> 
> Again, interpretation of the bit depends on e_ident[EI_OSABI].
> SHF_GNU_RETAIN should only be printed for ELFOSABI_GNU and
> ELFOSABI_FREEBSD.
> 
> -- 
> Alan Modra
> Australia Development Lab, IBM