IJ41693: GC CAN HAVE UNFLUSHED OWNABLE SYNCHRONIZER OBJECTS WHICH CAN EVENTUALLY LEAD TO A HEAP CORRUPTION AND FAILURE WHEN -XGC:CONCURRE

APAR status

• Closed as program error.

Error description

• Error Message: Assertion failure when -Xgc:concurrentScavenge is
  set.
  
                           ** ASSERTION FAILED ** at
  ./OwnableSynchronizerObjectBuffer.cpp:100: ((false))
                           or ** ASSERTION FAILED ** at
  ./Scavenger.cpp:1659: ((false))
  .
  Stack Trace:
  MM_MarkingSchemeRootClearer::scanOwnableSynchronizerObjects(MM_E
  nvironmentBase*)+0xea876010 (0x0000000000000000)
  
  MM_RootScanner::scanClearable(MM_EnvironmentBase*)+0xeab75328
  (0x0000000000000000)
  
  MM_MarkingDelegate::workerCompleteGC(MM_EnvironmentBase*)+0xea89
  5b08 (0x0000000000000000)
  
  MM_MarkingScheme::markLiveObjectsComplete(MM_EnvironmentBase*)+0
  xea8fd808 (0x0000000000000000)
  
  MM_ParallelMarkTask::run(MM_EnvironmentBase*)+0xea724418
  (0x0000000000000000)
  
  MM_ParallelDispatcher::workerEntryPoint(MM_EnvironmentBase*)+0xe
  a8d81d0 (0x0000000000000000)
  
  dispatcher_thread_proc2(OMRPortLibrary*,void*)+0xea8d8710
  (0x0000000000000000)
                           omrsig_protect+0xead07170
  (0x0000000000000000)
                           dispatcher_thread_proc+0xea8d8260
  (0x0000000000000000)
                           thread_wrapper+0xead9b0e8
  (0x0000000000000000)
  .
  Various crashes and assertion have been observed as a result of
  this issue.
  

Local fix

• This issue can be worked around by disabling concurrent
  scavenger. Removing concurrent scavenger related options is
  sufficient (e.g -Xgc:concurrentScavenge) as concurrent scavenger
  is disabled by default.
  

Problem summary

• Given specific thread scheduling conditions for concurrent
  scavenger tasks, GC can miss to flush thread local ownable
  synchronizer buffers to the global list when a concurrent
  scavenger cycle is completed as part of a global cycle.
  

Problem conclusion

• The solution is to force flush all ownabale synchronizer buffers
  when we complete a concurrent scavenger cycle as part of a
  global cycle.
          https://github.com/eclipse/omr/pull/6633
  
  .
  This APAR will be fixed in the following Releases:
  .
  IBM SDK, Java Technology Edition
     8    SR7 FP16  (8.0.7.16)
  .
  Contact your IBM Product's Service Team for these Service
  Refreshes and Fix Packs.
  For those running stand-alone, information about the available
  maintenance can be found at:
             https://www.ibm.com/support/pages/java-sdk
  

Temporary fix

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

• APAR is sysrouted TO one or more of the following:

Fix information

• Fixed component name

  J9 COMMON CODE

• Fixed component ID

  620700127

Applicable component levels