RFR: ShenandoahBarrierSet::AB::cmpxchg must call non-oop super implementation

[Aleksey Shipilev]

On 07/05/2018 02:03 PM, Roman Kennke wrote:
> This is a really bad snafu: the primitive cmpxchg in SBS::AB is calling
> the oop-cmpxchg super implementation. This will not go well, because
> super may have different width and even worse, it will try to
> (un-)compress the value as oop.
> 
> Fix:
> 
> # HG changeset patch
> # User rkennke
> # Date 1530792124 -7200
> #      Thu Jul 05 14:02:04 2018 +0200
> # Node ID c7ee69eee5073cb21a6fee90e0332b0cd3141069
> # Parent  38239dd666a162451120df9075b257c6b22d3db1
> ShenandoahBarrierSet::AB::cmpxchg must call non-oop super implementation
> 
> diff --git a/src/hotspot/share/gc/shenandoah/shenandoahBarrierSet.hpp
> b/src/hotspot/share/gc/shenandoah/shenandoahBarrierSet.hpp
> --- a/src/hotspot/share/gc/shenandoah/shenandoahBarrierSet.hpp
> +++ b/src/hotspot/share/gc/shenandoah/shenandoahBarrierSet.hpp
> @@ -190,7 +190,7 @@
>      template <typename T>
>      static T atomic_cmpxchg_in_heap_at(T new_value, oop base, ptrdiff_t
> offset, T compare_value) {
>        base = ShenandoahBarrierSet::barrier_set()->write_barrier(base);
> -      return Raw::oop_atomic_cmpxchg_at(new_value, base, offset,
> compare_value);
> +      return Raw::atomic_cmpxchg_at(new_value, base, offset,
> compare_value);
>      }
> 
>      template <typename T>

Yes, please.

Thanks,
-Aleksey