RFR(XS): JDK-8129440 G1 crash during concurrent root region scan.

[Siebenborn, Axel]

Hi,
we have seen crashes with G1 during concurrent root region scan.
It turned out, that the reason for the crashes was reloading an oop after a null check.
The compiler may legally do this, as the pointer is not declared as volatile.
The code runs concurrently to mutator threads.

template <class T>
inline void G1RootRegionScanClosure::do_oop_nv(T* p) {
  T heap_oop = oopDesc::load_heap_oop(p);                                           // 1. Load oop
  if (!oopDesc::is_null(heap_oop)) {  
    oop obj = oopDesc::decode_heap_oop_not_null(heap_oop);    // 2. Compiler decides to re-load the oop
    HeapRegion* hr = _g1h->heap_region_containing((HeapWord*) obj);
    _cm->grayRoot(obj, obj->size(), _worker_id, hr);
  }
}


We have seen the problem on AIX with the xlC compiler. However, we have seen similar optimizations on other platforms and other platforms.

As this code pattern is used quite often, I would suggest a global fix in  oopDesc::load_heap_oop(p).

Bug: https://bugs.openjdk.java.net/browse/JDK-8129440
Webrev: http://cr.openjdk.java.net/~asiebenborn/8129440/webrev.00/

Thanks,
Axel