Bug 1145105

Created attachment 939978 [details]
Thread dump

Description of problem:
This issue has the same impact as bug 1110794. In both cases a code that is accessed by multiple threads uses thread-unsafe HashMap. When the race condition happens, threads accessing the HashMap concurrently will create a pointer loop in its internal table. The loop will than cause other threads getting trapped in HashMap.getEntry().

Version-Release number of selected component (if applicable):
kie-wb-distribution-wars-6.2.0-20140914.210952-116-eap-6_1.war

How reproducible:
Race condition.

Steps to Reproduce:
Not yet.

Actual results:
Threads trapped in HashMap.getEntry()

Expected results:
Should be avoided.

Additional info:
See attached thread dump. There is "http-localhost/127.0.0.1:8080-4" executing Kie module build through BuildServiceImpl. In its stack trace it is visible that the thread acquired lock at some point. On the other hand "http-localhost/127.0.0.1:8080-5" is loading data model without acquiring any lock (this may be the synchronization hole).