Re: [PATCH v3] elf: Stop with zero entry point value [BZ #28453]

["H.J. Lu" <hjl.tools@gmail.com>]

On Thu, Dec 9, 2021 at 10:07 PM Florian Weimer <fweimer@redhat.com> wrote:
>
> * H. J. Lu:
>
> > diff --git a/elf/rtld.c b/elf/rtld.c
> > index 6ce1e07dc0..7d1801c51c 100644
> > --- a/elf/rtld.c
> > +++ b/elf/rtld.c
> > @@ -2491,6 +2491,12 @@ dl_main (const ElfW(Phdr) *phdr,
> >        rtld_timer_accum (&relocate_time, start);
> >      }
> >
> > +  /* Stop if there is no associated entry point.  */
> > +  if (rtld_is_main && main_map->l_entry == main_map->l_addr)
> > +    _dl_fatal_printf("%s: cannot execute shared object '%s' directly "
> > +                  "without entry point\n",
> > +                  ld_so_name, rtld_progname);
> > +
> >    /* Relocation is complete.  Perform early libc initialization.  This
> >       is the initial libc, even if audit modules have been loaded with
> >       other libcs.  */
>
> I think this is still too early for full backwards compatibility.
>
> However, the bug you are actually trying to fix occurs during
> relocation, so it looks to me as if using entry point addresses as an
> indicator will not actually work.  And entry point zero is generated by
> recent binutils only anyway.

True.  There is no easy way to identify a wrong entry point generated
by the old linkers.

> Thanks,
> Florian
>


-- 
H.J.