[
https://issues.apache.org/jira/browse/HDFS-5195?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13766833#comment-13766833
]
Colin Patrick McCabe commented on HDFS-5195:
--------------------------------------------
mlock and munlock are able to handle {{NULL}} pointers on their own, without
any assistance from us. They will simply fail with {{ENOMEM}}.
I do agree that it's helpful to print out a special error message in this case
to let the user know that JNI access to direct buffers is not available.
However, I would rather not put the check in like you have it here, since it's
an extra "if" statement we have to execute all the time, even though 99.9999%
of the time it does nothing. Let's add the check after mlock or munlock fails.
> Prevent passing null pointer to mlock and munlock.
> --------------------------------------------------
>
> Key: HDFS-5195
> URL: https://issues.apache.org/jira/browse/HDFS-5195
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode
> Affects Versions: HDFS-4949
> Reporter: Chris Nauroth
> Assignee: Chris Nauroth
> Attachments: HDFS-5195.1.patch
>
>
> According to JNI documentation, it is optional for the JVM to support the
> {{GetDirectBufferAddress}} function. If unsupported, then the function will
> return null. This is probably a very rare thing, but let's be defensive by
> checking the return value for null and throwing an exception instead of
> passing null down to {{mlock}} and {{munlock}}.
--
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
