JDK-8178542 : G1: VM hangs during shutdown due to mark stack overflow
  • Type: Bug
  • Component: hotspot
  • Sub-Component: gc
  • Affected Version: 9,10
  • Priority: P3
  • Status: Resolved
  • Resolution: Fixed
  • Submitted: 2017-04-12
  • Updated: 2018-06-21
  • Resolved: 2017-05-01
The Version table provides details related to the release that this issue/RFE will be addressed.

Unresolved : Release in which this issue/RFE will be addressed.
Resolved: Release in which this issue/RFE has been resolved.
Fixed : Release in which this issue/RFE has been fixed. The release containing this fix may be available for download as an Early Access Release or a General Availability Release.

To download the current JDK release, click here.
JDK 10
10 b21Fixed
Related Reports
Relates :  
JDK-8040245 - G1: VM hangs during shutdown
Relates :  
JDK-8065402 - G1 does not expand marking stack when mark stack overflow happens during concurrent marking
Description
Some GC threads are waiting in the first sync barrier and others marking threads cannot get there because flag _has_overflown was cleaned
Comments
Problem related to race condition at line g1ConcurrentMark.cpp:2961 near the end of G1CMTask::do_marking_step() when it was called from G1CMConcurrentMarkingTask::work() : 2961 if (_cm->has_overflown()) { ... Let assume that some thread passed this check with _cm->has_overflown() = false and overflow happened in some other thread before first thread returned from G1CMTask::do_marking_step() to the G1CMConcurrentMarkingTask::work() Some other thread may enter now first_sync_barrier First thread then returns to the G1CMConcurrentMarkingTask::work() 927 the_task->do_marking_step(mark_step_duration_ms, 928 true /* do_termination */, 929 false /* is_serial*/); 930 931 double end_vtime_sec = os::elapsedVTime(); 932 double elapsed_vtime_sec = end_vtime_sec - start_vtime_sec; 933 _cm->clear_has_overflown(); and at line 933 clears _has_overflown flag As result this thread cannot reach first_sync_barrier. Marking threads will loop till Full GC abort first_sync_barrier. And if VM will exit without Full GC we will get hang in shutdown. This problem might be fixed by removing wrong line 933 - no need to clear _has_overflown flag. This issue was reproduced with b8065402.java (attached to the JDK-8065402)
14-04-2017