RFR: 8290025: Remove the Sweeper [v11]

Martin Doerr mdoerr at openjdk.org
Mon Aug 22 17:05:42 UTC 2022


On Mon, 22 Aug 2022 16:04:18 GMT, Erik Österlund <eosterlund at openjdk.org> wrote:

> > Can you please add a null check? C1 patching stubs contain null Oops on PPC64.
> > ```diff
> > --- a/src/hotspot/share/gc/shared/barrierSetNMethod.cpp
> > +++ b/src/hotspot/share/gc/shared/barrierSetNMethod.cpp
> > @@ -72,7 +72,9 @@ bool BarrierSetNMethod::nmethod_entry_barrier(nmethod* nm) {
> >        // conversion that performs the load barriers. This is too subtle, so we instead
> >        // perform an explicit keep alive call.
> >        oop obj = NativeAccess<ON_PHANTOM_OOP_REF | AS_NO_KEEPALIVE>::oop_load(p);
> > -      Universe::heap()->keep_alive(obj);
> > +      if (obj != nullptr) {
> > +        Universe::heap()->keep_alive(obj);
> > +      }
> >      }
> >  
> >      virtual void do_oop(narrowOop* p) { ShouldNotReachHere(); }
> > ```
> > 
> > 
> >     
> >       
> >     
> > 
> >       
> >     
> > 
> >     
> >   
> > Or should we fix that separately? Seems to be related to [JDK-8292077](https://bugs.openjdk.org/browse/JDK-8292077).
> 
> Ah. Good find. I can add the null check.

I just noticed that there is already a bug open: https://bugs.openjdk.org/browse/JDK-8292368
I think I should better fix it separately in case it will be needed by backports. Thanks for taking a look!

-------------

PR: https://git.openjdk.org/jdk/pull/9741


More information about the hotspot-dev mailing list