[XMLRPC-162] WebServer worker thread is stuck on an incoming XML RPC call

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.1.1
• Fix Version/s: 3.1.2
• Environment: Linux on PPC - underpowered platform PPC8270
• Assignee: Jochen Wiedmann
• Reporter: Mark Gertsvolf
• Created: 28/Nov/08 04:31
• Updated: 29/Apr/09 20:03
• Resolved: 28/Nov/08 20:49

Description

I am running XML-RPC server on an underpowered Linux PPC box with JamVM or Cacao, both use GNU classlib. This environment exposes a race condition in the WebServer code.
Specificlly the issue is with how org.apache.xmlrpc.util.ThreadPool signals the worker threads.

Poolable.startTask creates a new Poolable object and calls Poolable.start, which sets the Poolable.task member and signals the thread encapsulated by Poolable object. However, there is a potential race condition where the signal can be lost if the encapsulated thread has just pased beyond the getTask check:
(line 55 trunk/common/src/main/java/org/apache/xmlrpc/util/ThreadPool.java - revision 720508)
The worker thread is suspended on wait and there is nobody to wake him up.

One way of fixing this issue is to add (getTask() == null) check inside the synchronized block.

diff:

Index: /var/opt/apachews/trunk/common/src/main/java/org/apache/xmlrpc/util/ThreadPool.java
===================================================================
— /var/opt/apachews/trunk/common/src/main/java/org/apache/xmlrpc/util/ThreadPool.java (revision 720508)
+++ /var/opt/apachews/trunk/common/src/main/java/org/apache/xmlrpc/util/ThreadPool.java (working copy)
@@ -56,7 +56,9 @@
if (t == null) {
try {
synchronized (this) {

• wait();
  + if (!isShuttingDown() && getTask() == null) { + wait(); + }

  }
  } catch (InterruptedException e) {
  // Do nothing