You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mina.apache.org by "Geoff Cadien (JIRA)" <ji...@apache.org> on 2008/06/05 23:56:45 UTC

[jira] Commented: (DIRMINA-584) Using apr-transport crashes jvm on Linux

    [ https://issues.apache.org/jira/browse/DIRMINA-584?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12602805#action_12602805 ] 

Geoff Cadien commented on DIRMINA-584:
--------------------------------------

Just for completeness here is a small patch that manually sets the file descriptor in AprSession to 0 when it is destroyed.  It's probably not the best solution but it does work.

 Index: src/main/java/org/apache/mina/transport/socket/apr/AprIoProcessor.java
===================================================================
--- src/main/java/org/apache/mina/transport/socket/apr/AprIoProcessor.java      (revision 656872)
+++ src/main/java/org/apache/mina/transport/socket/apr/AprIoProcessor.java      (working copy)
@@ -43,7 +43,6 @@
  * @author The Apache MINA Project (dev@mina.apache.org)
  * @version $Rev$, $Date$
  */
-
 public final class AprIoProcessor extends AbstractPollingIoProcessor<AprSession> {
     private static final int POLLSET_SIZE = 1024;

@@ -239,6 +238,11 @@
             }
         } finally {
             ret = Socket.close(session.getDescriptor());
+            // I think socket needs to be destroyed.  When a connection is accepted
+            // a pool is allocated for the socket and I don't see how else it will
+            // get destroyed.
+            Socket.destroy(session.getDescriptor());
+            session.setDescriptor(0);
             if (ret != Status.APR_SUCCESS) {
                 throwException(ret);
             }


Index: src/main/java/org/apache/mina/transport/socket/apr/AprSession.java
===================================================================
--- src/main/java/org/apache/mina/transport/socket/apr/AprSession.java	(revision 656872)
+++ src/main/java/org/apache/mina/transport/socket/apr/AprSession.java	(working copy)
@@ -38,7 +38,7 @@
  * @version $Rev$, $Date$
  */
 public abstract class AprSession extends AbstractIoSession {
-    private final long descriptor;
+    private  long descriptor;
 
     private final IoService service;
     private final IoProcessor<AprSession> processor;
@@ -89,6 +89,10 @@
         return descriptor;
     }
 
+    void setDescriptor(long desc) {
+        this.descriptor = desc;
+    }
+    
     @Override
     protected IoProcessor<AprSession> getProcessor() {
         return processor;





> Using apr-transport crashes jvm on Linux
> ----------------------------------------
>
>                 Key: DIRMINA-584
>                 URL: https://issues.apache.org/jira/browse/DIRMINA-584
>             Project: MINA
>          Issue Type: Bug
>          Components: Transport
>    Affects Versions: 2.0.0-M1
>         Environment: Linux jellikit 2.6.24-16-generic #1 SMP Thu Apr 10 13:23:42 UTC 2008 i686 GNU/Linux
> java version "1.6.0_06"
> Java(TM) SE Runtime Environment (build 1.6.0_06-b02)
> Java HotSpot(TM) Server VM (build 10.0-b22, mixed mode)
> tomcat-native-1.1.13
> apr-1.2.11
>            Reporter: Geoff Cadien
>            Assignee: Trustin Lee
>             Fix For: 2.0.0-M2
>
>
> To reproduce you can simply run the EchoServer from mina-example and use AprIoAcceptor instead of NioAcceptor.  I used the EchoClient from Julien Vermillard's sandbox here: https://svn.apache.org/repos/asf/mina/sandbox/jvermillard/ and used 20 or 30 concurrent connections.
> AprIoProcessor.destroy() is called multiple times when an exception is generated in AbstractPollingIoProcessor (RST, etc.) and either epoll or apr doesn't like to remove a closed fd from the pollset.  I believe that destroy is called twice because AprIoProcessor.state() returns SessionState.OPEN when the native fd is > 0.  I don't believe once open it will ever be 0 again.
> I was able to fix the problem by explicitly settion AprSession.descriptor to 0 after the socket was closed although I'm not sure this is the correct solution.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.