Copy relocation support for _r_debug

[Florian Weimer]

* H. J. Lu:

> On Thu, Sep 23, 2021 at 10:24 AM Florian Weimer <fweimer@redhat.com> wrote:
>>
>> I think we lost copy relocation support for _r_debug.  We only update
>> _r_debug_extended, so we have effectively created a rtld_hidden_proto
>> alias here.
>>
>> Is this a problem?  I somehow missed this during review. 8-(
>
> We have discussed it during Monday's patch review. I don't believe it is
> a problem.  The _r_debug data used for copy relocation has been set up
> before copy relocation is invoked.  The users of _r_debug only use the
> value in the r_map field, not the address of the r_map field as shown in
> the _r_debug tests in glibc.  The value of the r_map field in the _r_debug
> copy is always the same as the internal _r_debug_extended.

Okay, this is nice.  We got rid of one symbol-based relocation.

So only __stack_chk_guard and __libc_stack_end are left in generic code
(after my exception handling clean up goes in).

Thanks,
Florian