[PATCH] nscd: bump GC cycle during cache pruning (bug 26130)

[Florian Weimer]

* Andreas Schwab:

> On Jun 29 2020, Florian Weimer via Libc-alpha wrote:
>
>> * Andreas Schwab:
>>
>>> On Jun 29 2020, Florian Weimer wrote:
>>>
>>>> I think this needs barriers after and before the increments.
>>>
>>> Why doesn't gc need those barriers?
>>
>> I think it needs them as well.  I thought we had them there. 8-(
>
> Should they use atomic_increment?

Yes, that as well.  I don't know if the legacy atomic_increment function
implies a barrier.  I think we need the equivalent of __atomic_fetch_add
(or __atomic_add_fetch) with __ATOMIC_ACQ_REL, and our <atomic.h> does
not have this, and we are not supposed to use the compiler built-ins.

atomic_fetch_add_relaxed together with atomic_full_barrier should work,
though.

Thanks,
Florian