[jboss-jira] [JBoss JIRA] (JGRP-1679) tryLock with (and without) timeout allows simultaneous concurrent access to not thread-safe data to more then one client

[Bela Ban (JIRA)]

    [ https://issues.jboss.org/browse/JGRP-1679?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12798414#comment-12798414 ] 

Bela Ban commented on JGRP-1679:
--------------------------------

Ok, I think I know what the problem is:
- Thread T1 calls tryLock(timeout) on lock L
- The coordinator gets the GRANT-LOCK request and sends back a LOCK-GRANTED response
- Before the LOCK-GRANTED response is received, T1 times out and removes the client lock
- T1 now creates a new client lock L (same name) and tries to acquire it again
- Say L is held by someone else, so a LOCK-DENIED response is sent back
- Before T1 gets the LOCK-DENIED response, the *old LOCK-GRANTED* response is received
- This sets L to 'acquired' although it should be denied !

SOLUTION:
* Mark client locks with unique lock IDs and send those IDs with lock requests
* When a lock response is received and its request ID doesn't match the client lock's ID, discard the response
* In the above example, the old LOCK-GRANTED response would be discarded by the 2nd client lock L created by T2
                
> tryLock with (and without) timeout allows simultaneous concurrent access to not thread-safe data to more then one client
> ------------------------------------------------------------------------------------------------------------------------
>
>                 Key: JGRP-1679
>                 URL: https://issues.jboss.org/browse/JGRP-1679
>             Project: JGroups
>          Issue Type: Bug
>    Affects Versions: 3.2.4, 3.3.4
>         Environment: Intel(R) Core(TM) i5-2310 CPU @ 2.90GHz 
> 8 GB RAM 
> Windows 7 
> JRE 1.6.31
>            Reporter: Architect SoftWeb.ISD
>            Assignee: Bela Ban
>             Fix For: 3.4
>
>         Attachments: lockservice_stability_testcase.jar
>
>
> It looks as it is possible to access a not thread-safe data simultaneously by several clients when tryLock with timeout is used.
> It is reproduced easily with timeouts from 1 to 3 seconds inclusive.
> It is very difficult to reproduce the issue with bigger values of the timeout.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira