second thoughts on using dl_iterate_phdr() for cache-validation

[Roland McGrath]

No new ABI is going to go in very soon.  We have frozen the GLIBC_2.3.4 ABI
already.  (Your patch adds symbols to a version set that has been fixed for
a long time, which is never a kind of patch that would be accepted.)

unsigned long long int is not an atomic type on all platforms, so your code
as written is not acceptable without adding locking for machines where it's
not.  It seems preferable to just used unsigned long int instead, which it
is safe to increment and examine simultaneously on all the platforms we have.

An interface to return the address of such variables is highly dubious,
since it constrains the implementation and does so permanently for the future.