RFR: 8255006: StringDedupTable lookup may encounter null oop with concurrent GC

[Roman Kennke]

On Mon, 19 Oct 2020 16:34:46 GMT, Zhengyu Gu <zgu at openjdk.org> wrote:

> Concurrent GC (e.g. Shenandoah) cleans string dedup table concurrently. It utilizes its load barrier to hide dead oop
> (returns null instead), so that, StringDedupTable::loopup() may encounter null oops during table lookup.
> - [x] tier1

src/hotspot/share/gc/shared/stringdedup/stringDedupTable.cpp line 281:

> 279:       oop* obj_addr = (oop*)entry->obj_addr();
> 280:       oop obj = NativeAccess<ON_PHANTOM_OOP_REF | AS_NO_KEEPALIVE>::oop_load(obj_addr);
> 281:       if (!CompressedOops::is_null(obj) && java_lang_String::value_equals(value, static_cast<typeArrayOop>(obj))) {

This cannot be a narrowOop, therefore there's no point in using CompressedOops::is_null(), you can use != NULL instead.
Also, I'd change the commit msg to reflect what you did/what the solution is like "Add NULL-check to..." and not what
the problem was.

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

PR: https://git.openjdk.java.net/jdk/pull/743