Re: MIPS multigot fixes for Linux

[Daniel Jacobowitz <drow at mvista dot com>]

On Thu, Nov 13, 2003 at 09:35:40AM +0000, Richard Sandiford wrote:
> Daniel Jacobowitz <drow@mvista.com> writes:
> > 	* elfxx-mips.c (mips_elf_set_global_got_offset): Don't set
> > 	no_fn_stub.
> > 	(mips_elf_set_no_stub): New function.
> > 	(mips_elf_multi_got): Call it.
> > 	(_bfd_mips_elf_finish_dynamic_symbol): Fill relocated GOT entries
> > 	with zero for ! SGI_COMPAT.  Simplify calculation of value.
> 
> Looks good to me FWIW.  One thing though:
> 
> > +      /* We can't do lazy update of GOT entries for non-primary GOTs since
> > +        the PLT entries don't use the right offsets, so punt at it for now.
> > +        We set this here because we are called via mips_elf_multi_got
> > +        before _bfd_mips_elf_adjust_dynamic_symbol reads the no_fn_stub
> > +        flag; this only matters for the global case, but
> > +        _bfd_mips_elf_size_dynamic_sections is too late.  */
> > +      entry->d.h->no_fn_stub = TRUE;
> 
> I'm not sure whether it's useful to keep this comment, especially given
> the new ones you've added above the function itself, and above the call
> in mips_elf_multi_got.  "Punt at it for now" makes it sound like a FIXME.
> But surely it's just Plain Wrong to allow lazy binding of functions that
> are pointed to by a secondary GOT?  From the dynamic linker's point of
> view, it's no different from a function pointer in .data or whatever.
> 
> And setting no_fn_stub in its new position seems natural enough to me.
> I'm not sure you need to excuse it.  If you do still want to mention
> _bfd_mips_elf_size_dynamic_sections, it might be better to do it above
> the call in mips_elf_multi_got.  But IMO it could be dropped altogether.

Yes, that part of the comment could be dropped.  To be honest, I
haven't even read it since I wrote it two revisions ago - it was just
following the assignment around and doesn't make much sense where it
is.

I'm not so sure about it making no sense, btw.  It would certainly have
to be done _differently_.  It would involve fiddling with the
_dl_runtime_resolve ABI.  But creating lazy stubs for secondary GOTs is
definitely possible.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer