[jboss-jira] [JBoss JIRA] (JGRP-1464) TCPConnectionMap: message from different JGroups version may cause OOME

[Bela Ban (JIRA)]

     [ https://issues.jboss.org/browse/JGRP-1464?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Bela Ban resolved JGRP-1464.
----------------------------

    Resolution: Done

    
> TCPConnectionMap: message from different JGroups version may cause OOME
> -----------------------------------------------------------------------
>
>                 Key: JGRP-1464
>                 URL: https://issues.jboss.org/browse/JGRP-1464
>             Project: JGroups
>          Issue Type: Bug
>            Reporter: Bela Ban
>            Assignee: Bela Ban
>             Fix For: 3.0.10, 3.1
>
>
> If discard_incompatible_packets is enabled at the transport level, we discard packets from other JGroups versions. However, this is not done when TCP is used, at the TCPConnectionMap level.
> When we read a packet from a different version, we first read the version, then the length. If the length is garbage, and we interpret it as a long, it can be huge, leading to an OOME.
> [email from http://old.nabble.com/ArrayIndexOutOfBoundsException-in-3.09-ts33781725.html#a33852559]
> I'd like to add some additional information to this post since I am seeing this crash the entire JVM. I've added some debugging statements to the TCPConnectionMap - ConnectionPeerReceiver.run method to try an understand the issue and recompiled the 3.0.9 jar (I am printing the len of the DataInputStream).  It appears that when an incompatible message is received JG will discard some messages appropriately but not others.  When some of these messages make it through they end up being of enormous size 1192331780 (or ~1.1 GB) which is enough to crash the JVM due to OOM (catching the OOMError in not sufficient in our configuration).  Some are of size 0 as well...  curious if they is some signed/unsigned conversion going on here?  At any rate this is prohibiting us from moving to the latest version of JGroups in hopes of receiving better support.
> 2012-05-14 09:27:05,623 [Connection.Receiver [135.9.96.63:59953 - 135.9.128.31:7800],135.9.148.15_InterCluster-2.12,asmblade23-47979] jgroups.blocks.TCPConnectionMap$TCPConnection WARN - TG: DEBUG len=70
> 2012-05-14 09:27:05,624 [Connection.Receiver [135.9.96.63:59953 - 135.9.128.31:7800],135.9.148.15_InterCluster-2.12,asmblade23-47979] jgroups.blocks.TCPConnectionMap$TCPConnection WARN - TG: DEBUG len=70
> 2012-05-14 09:27:05,624 [OOB-1,135.9.148.15_InterCluster-2.12,asmblade23-47979] jgroups.protocols.TCP WARN - packet from 135.9.128.31:7800 has different version (2.6.10) from ours (3.0.10). Packet is discarded
> 2012-05-14 09:27:05,625 [OOB-2,135.9.148.15_InterCluster-2.12,asmblade23-47979] jgroups.protocols.TCP WARN - packet from 135.9.128.31:7800 has different version (2.6.10) from ours (3.0.10). Packet is discarded
> 2012-05-14 09:27:12,028 [ConnectionMap.Acceptor,null,null] jgroups.blocks.TCPConnectionMap$TCPConnection WARN - packet from /135.9.96.59:56077 has different version (2.6.10) from ours (3.0.10). This may cause problems
> 2012-05-14 09:27:12,030 [Connection.Receiver [135.9.96.63:7800 - 135.9.96.59:56077],135.9.148.15_InterCluster-2.12,asmblade23-47979] jgroups.blocks.TCPConnectionMap$TCPConnection WARN - TG: DEBUG len=0
> 2012-05-14 09:27:12,032 [Connection.Receiver [135.9.96.63:7800 - 135.9.96.59:56077],135.9.148.15_InterCluster-2.12,asmblade23-47979] jgroups.protocols.TCP ERROR - failed handling data from 135.9.96.59:7800
> java.lang.ArrayIndexOutOfBoundsException: Array index out of range: 2
>         at org.jgroups.protocols.TP.receive(TP.java:1200)
>         at org.jgroups.protocols.BasicTCP.receive(BasicTCP.java:104)
>         at org.jgroups.blocks.TCPConnectionMap$TCPConnection$ConnectionPeerReceiver.run(TCPConnectionMap.java:603)
>         at java.lang.Thread.run(Thread.java:769)
> 2012-05-14 09:27:12,034 [Connection.Receiver [135.9.96.63:7800 - 135.9.96.59:56077],135.9.148.15_InterCluster-2.12,asmblade23-47979] jgroups.blocks.TCPConnectionMap$TCPConnection WARN - TG: DEBUG len=1192331780 
> SOLUTION: discard packets from different JGroups versions in TCPConnectionMap.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.jboss.org/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira