jackrabbit-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jukka Zitting (JIRA)" <j...@apache.org>
Subject [jira] Updated: (JCR-2753) Deadlock in DefaultISMLocking
Date Wed, 22 Sep 2010 14:40:35 GMT

     [ https://issues.apache.org/jira/browse/JCR-2753?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Jukka Zitting updated JCR-2753:
-------------------------------

    Attachment: JCR-2753.patch

We discussed this with Thomas, and it turns out that synchronous observation listeners do
need to reacquire ISM read locks after the SISM.Update.end() has downgraded the write lock
it was holding. One way to solve this problem without having to maintain a set of current
holders of the read lock is to simply remove the writer preference in the locking strategy,
i.e. allow readers to acquire the lock even when there are pending writers waiting for the
lock. Another solution, implemented in the attached patch (against the "new" DefaultISMLocking
implementation of revision 99541), is to retain the writer thread identifier in a downgraded
lock, so a downgraded write lock would still allow related threads to reacquire the lock even
when there are other writers waiting.

> Deadlock in DefaultISMLocking
> -----------------------------
>
>                 Key: JCR-2753
>                 URL: https://issues.apache.org/jira/browse/JCR-2753
>             Project: Jackrabbit Content Repository
>          Issue Type: Bug
>          Components: jackrabbit-core
>            Reporter: Thomas Mueller
>            Assignee: Thomas Mueller
>             Fix For: 2.2.0
>
>         Attachments: JCR-2753.patch
>
>
> There seems to be a bug in DefaultISMLocking which was detected as part of JCR-2746.
> 1) The main thread gets a read lock.
> 2) The ObservationManager thread tries to lock for writing, which is blocked because
there is still a read lock.
> 3) Then the main thread tries to get a second read lock, which is blocked because there
is a waiting write lock.
> The bug was introduced as part of JCR-2089 (Use java.util.concurrent), revisions 995411
and 995412. I think the safe solution is to revert those to commits, and add a test case.
If the DefaultISMLocking is changed later on, more test cases are required. An efficient solution
is relatively complicated.

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


Mime
View raw message