RFR: 8016304: ThreadMXBean.getDeadlockedThreads reports bogus deadlocks on JDK 8

Daniel D. Daugherty daniel.daugherty at oracle.com
Mon Jun 17 08:49:07 PDT 2013


 > webrev: http://cr.openjdk.java.net/~sla/8016304/webrev.00/

Thumbs up.

src/share/vm/services/threadService.cpp
     Can you add the following comment:

     // The setting of the current_pending_monitor field is speculative
     // and racy so we need to check to see if the object is actually owned.

     above these two lines:

     line 330 address currentOwner = (address)waitingToLockMonitor->owner();
     line 331 if (currentOwner != NULL) {

     Thanks! I couldn't think of a good comment until this AM. Or I'm
     delusional and this isn't a good comment. :-)

test/serviceability/threads/TestFalseDeadLock.java
     So the test does repro in 5 seconds sometimes. I'm good with that.


Dan


On 6/17/13 9:22 AM, Staffan Larsen wrote:
> Please review this fix. There is a long description in the bug report of the problem and the solution which I will not copy here. I have made the reproducer from the bug into a test which is also included.
>
> webrev: http://cr.openjdk.java.net/~sla/8016304/webrev.00/
> bug: http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=8016304
> jbs: https://jbs.oracle.com/bugs/browse/JDK-8016304
>
> Thanks,
> /Staffan



More information about the hotspot-runtime-dev mailing list