-
Bug
-
Resolution: Fixed
-
P2
-
8, 11, 16, 17
-
b35
Issue | Fix Version | Assignee | Priority | Status | Resolution | Resolved In Build |
---|---|---|---|---|---|---|
JDK-8261120 | 17 | Kim Barrett | P2 | Resolved | Fixed | b09 |
JDK-8261386 | 16.0.2 | Kim Barrett | P2 | Resolved | Fixed | b01 |
JDK-8261373 | 16.0.1 | Kim Barrett | P2 | Resolved | Fixed | b05 |
JDK-8273898 | 15.0.5 | Olga Mikhaltcova | P2 | Resolved | Fixed | b02 |
JDK-8273897 | 13.0.9 | Olga Mikhaltcova | P2 | Resolved | Fixed | b02 |
JDK-8261370 | 11.0.12-oracle | Dukebot | P2 | Closed | Fixed | b01 |
JDK-8261875 | 11.0.11 | Kim Barrett | P2 | Resolved | Fixed | b04 |
JDK-8262794 | openjdk8u302 | Aleksey Shipilev | P2 | Resolved | Fixed | b01 |
JDK-8261756 | 8u301 | Fairoz Matte | P2 | Closed | Fixed | b01 |
Unfortunately, that's the wrong place for the ordering barrier. It ensures the stores done by MutableSpace::initialize (including the update of _end) follow the other data structure updates in PSOldGen, but there are also writes within MS::initialize that must preceed the update of _end. In particular, when AlwaysPreTouch, the writes done for pretouching are presently unordered wrto the write of _end. Another thread could see the new _end value and allocate and start using that newly available space, only to have pretouch writes scribble on it.
A barrier is needed to ensure the write to _end is really last.
- backported by
-
JDK-8261120 ParallelGC: oldgen expansion needs release-store for _end
- Resolved
-
JDK-8261373 ParallelGC: oldgen expansion needs release-store for _end
- Resolved
-
JDK-8261386 ParallelGC: oldgen expansion needs release-store for _end
- Resolved
-
JDK-8261875 ParallelGC: oldgen expansion needs release-store for _end
- Resolved
-
JDK-8262794 ParallelGC: oldgen expansion needs release-store for _end
- Resolved
-
JDK-8273897 ParallelGC: oldgen expansion needs release-store for _end
- Resolved
-
JDK-8273898 ParallelGC: oldgen expansion needs release-store for _end
- Resolved
-
JDK-8261370 ParallelGC: oldgen expansion needs release-store for _end
- Closed
-
JDK-8261756 ParallelGC: oldgen expansion needs release-store for _end
- Closed
- relates to
-
JDK-8257999 Parallel GC crash in gc/parallel/TestDynShrinkHeap.java: new region is not in covered_region
- Closed
- links to
-
Commit openjdk/jdk13u-dev/56d491cb
-
Commit openjdk/jdk15u-dev/d5d9b458
-
Commit openjdk/jdk16/afd5eefd
-
Review openjdk/jdk13u-dev/262
-
Review openjdk/jdk15u-dev/100
-
Review openjdk/jdk16/141