Re: [PATCH v2 0/5] Fix ELF / FDPIC ELF core dumping, and use mmap_sem properly in there

[Greg Ungerer]

On 30/4/20 9:03 am, Linus Torvalds wrote:
> On Wed, Apr 29, 2020 at 2:57 PM Russell King - ARM Linux admin
> <linux@xxxxxxxxxxxxxxx> wrote:
> > I've never had any reason to use FDPIC, and I don't have any binaries
> > that would use it.  Nicolas Pitre added ARM support, so I guess he
> > would be the one to talk to about it.  (Added Nicolas.)
>
> While we're at it, is there anybody who knows binfmt_flat?
>
> It might be Nicolas too.
>
> binfmt_flat doesn't do core-dumping, but it has some other oddities.
> In particular, I'd like to bring sanity to the installation of the new
> creds, and all the _normal_ binfmt cases do it largely close together
> with setup_new_exec().
>
> binfmt_flat is doing odd things. It's doing this:
>
>          /* Flush all traces of the currently running executable */
>          if (id == 0) {
>                  ret = flush_old_exec(bprm);
>                  if (ret)
>                          goto err;
>
>                  /* OK, This is the point of no return */
>                  set_personality(PER_LINUX_32BIT);
>                  setup_new_exec(bprm);
>          }
>
> in load_flat_file() - which is also used to loading _libraries_. Where
> it makes no sense at all.

I haven't looked at the shared lib support in there for a long time,
but I thought that "id" is only 0 for the actual final program.
Libraries have a slot or id number associated with them.

> It does the
>
>          install_exec_creds(bprm);
>
> in load_flat_binary() (which makes more sense: that is only for actual
> binary loading, no library case).
>
> I would _like_ for every binfmt loader to do
>
>          /* Flush all traces of the currently running executable */
>          retval = flush_old_exec(bprm);
>          if (retval)
>                  return retval;
>
>     .. possibly set up personalities here ..
>
>          setup_new_exec(bprm);
>          install_exec_creds(bprm);
>
> all together, and at least merge 'setup_new_exec()' with 'install_exec_creds()'.
>
> And I think all the binfmt handlers would be ok with that, but the
> flat one in particular is really oddly set up.
>
> *Particularly* with that flush_old_exec/setup_new_exec() being done by
> the same routine that is also loading libraries (and called from
> 'calc_reloc()' from binary loading too).
>
> Adding Greg Ungerer for m68knommu. Can somebody sort out why that
> flush_old_exec/setup_new_exec() isn't in load_flat_binary() like
> install_exec_creds() is?
>
> Most of that file goes back to pre-git days. And most of the commits
> since are not so much about binfmt_flat, as they are about cleanups or
> changes elsewhere where binfmt_flat was just a victim.

I'll have a look at this.

Quick hack test shows moving setup_new_exec(bprm) to be just before
install_exec_creds(bprm) works fine for the static binaries case.
Doing the flush_old_exec(bprm) there too crashed out - I'll need to
dig into that to see why.

Regards
Greg