[PATCH v3] Fix FORTIFY_SOURCE false positive

[Florian Weimer]

* Volker Weißmann:

> Shall I implement it or is there anyone else who wants to do that?

Arjun has expressed in this.  He could help you along and review your
changes.

Our downstream issue is:

  glibc: Reduce likelihood of spurious %n failure in printf
  <https://issues.redhat.com/browse/RHEL-79064>

(It's worded from an older glibc perspective.  We might want upstream
issue for the SUID hardening regression, too.)

There, I proposed the following approach:

* Use _dl_find_object to get the link map
* Iterate over the LOAD segments to find a matching read-only segment.
* If it is read-write, also check the RELRO area.
* Do the link map access from a new function in the dynamic linker, not
  libc, to limit layout dependencies.
* Process termination should be handled from within libc.
* If we cannot find a link map, proceed as before with
  /proc/self/maps. But if opening or reading fails, write a different,
  new error message using __libc_fatal.

Thanks,
Florian