[ https://issues.apache.org/jira/browse/BEANUTILS-373?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Andrew Sunde updated BEANUTILS-373: ----------------------------------- Description: The problem lies in the class org.apache.commons.beanutils.MethodUtils. This class is keeping a global cache in a non synchronized WeakFastHashMap (which claims to be syncronized). When operating in Fast mode (which is what the WeakFastHashMap is set to after it is created), it then access the underlining WeakHashMap without synchronization. Therefore this can lead to a corruption in the underlining HashMap. See the description of the WeakHashMap javadoc: >From Java Sun: " Like most collection classes, this class is not synchronized. A synchronized WeakHashMap may be constructed using the Collections.synchronizedMap method. .... The behavior of the WeakHashMap class depends in part upon the actions of the garbage collector, so several familiar (though not required) Map invariants do not hold for this class. Because the garbage collector may discard keys at any time, a WeakHashMap may behave as though an unknown thread is silently removing entries. " Therefore even if you are doing a "get" from the hashmap, there is a chance that you could get a corrupted hashmap because the garbage collector could be changing the hashmap on the fly. This could lead to threads getting stuck in an infinite loop. The modification to make the MethodUtils cache thread safe is a one line change. Before: private static WeakFastHashMap cache = new WeakFastHashMap(); After: private static Map cache = Collections.synchronizedMap(new WeakHashMap()); was: The problem lies in the class org.apache.commons.beanutils.MethodUtils. This class is keeping a global cache in a non synchronized WeakHashMap. WeakHashMap is not thread safe and required external synchronization. The lack of synchronization can cause corruption and the WeakHashMap.get() method to go into an infinite loop. Googling "WeakHashMap infinite loop" returns many cases of similar problem with WeakHashMap. The solution is to decorate the WeakHashMap in a synchronized Map as described in this thread: http://dev.eclipse.org/mhonarc/lists/aspectj-users/msg08824.html The modification to make the MethodUtils cache thread safe is a one line change. Before: private static WeakHashMap cache = new WeakHashMap(); After: private static Map cache = Collections.synchronizedMap(new WeakHashMap()); Example of thread dump "ExecuteThread: '0' for queue: 'weblogic.kernel.Default'" id=13 idx=0x3c tid=5905 prio=5 alive, daemon at org/apache/commons/beanutils/MethodUtils$MethodDescriptor.equals(MethodUtils.java:828)[optimized] at java/util/WeakHashMap.eq(WeakHashMap.java:254)[inlined] at java/util/WeakHashMap.get(WeakHashMap.java:345)[optimized] at org/apache/commons/beanutils/MethodUtils.getMatchingAccessibleMethod(MethodUtils.java:530)[optimized] at org/apache/commons/beanutils/MethodUtils.invokeMethod(MethodUtils.java:209)[inlined] :::: "ExecuteThread: '1' for queue: 'weblogic.kernel.Default'" id=14 idx=0x40 tid=5906 prio=5 alive, daemon at org/apache/commons/beanutils/MethodUtils$MethodDescriptor.equals(MethodUtils.java:833)[optimized] at java/util/WeakHashMap.eq(WeakHashMap.java:254)[inlined] at java/util/WeakHashMap.get(WeakHashMap.java:345)[optimized] at org/apache/commons/beanutils/MethodUtils.getMatchingAccessibleMethod(MethodUtils.java:530)[optimized] at org/apache/commons/beanutils/MethodUtils.invokeMethod(MethodUtils.java:209)[inlined] ::: "ExecuteThread: '2' for queue: 'weblogic.kernel.Default'" id=15 idx=0x44 tid=5907 prio=5 alive, daemon at org/apache/commons/beanutils/MethodUtils$MethodDescriptor.equals(MethodUtils.java:833)[optimized] at java/util/WeakHashMap.eq(WeakHashMap.java:254)[inlined] at java/util/WeakHashMap.get(WeakHashMap.java:345)[optimized] at org/apache/commons/beanutils/MethodUtils.getMatchingAccessibleMethod(MethodUtils.java:530)[optimized] at org/apache/commons/beanutils/MethodUtils.invokeMethod(MethodUtils.java:209)[inlined] > CLONE - MethodUtils is not thread safe because WeakFastHashMap which uses > WeakHashMap is not thread-safe. > --------------------------------------------------------------------------------------------------------- > > Key: BEANUTILS-373 > URL: https://issues.apache.org/jira/browse/BEANUTILS-373 > Project: Commons BeanUtils > Issue Type: Bug > Components: Bean / Property Utils > Affects Versions: 1.8.2 > Environment: Weblogic 9.2 on Linux, but should not be relevant > Reporter: Andrew Sunde > Assignee: Niall Pemberton > > The problem lies in the class org.apache.commons.beanutils.MethodUtils. This > class is keeping a global cache in a non synchronized WeakFastHashMap (which > claims to be syncronized). When operating in Fast mode (which is what the > WeakFastHashMap is set to after it is created), it then access the > underlining WeakHashMap without synchronization. Therefore this can lead to a > corruption in the underlining HashMap. > See the description of the WeakHashMap javadoc: > From Java Sun: " > Like most collection classes, this class is not synchronized. A synchronized > WeakHashMap may be constructed using the Collections.synchronizedMap method. > .... > The behavior of the WeakHashMap class depends in part upon the actions of the > garbage collector, so several familiar (though not required) Map invariants > do not hold for this class. Because the garbage collector may discard keys at > any time, a WeakHashMap may behave as though an unknown thread is silently > removing entries. > " > Therefore even if you are doing a "get" from the hashmap, there is a chance > that you could get a corrupted hashmap because the garbage collector could be > changing the hashmap on the fly. This could lead to threads getting stuck in > an infinite loop. > The modification to make the MethodUtils cache thread safe is a one line > change. > Before: > private static WeakFastHashMap cache = new WeakFastHashMap(); > After: > private static Map cache = Collections.synchronizedMap(new WeakHashMap()); -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.