[
https://issues.apache.org/jira/browse/HADOOP-12593?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15033446#comment-15033446
]
Steve Loughran commented on HADOOP-12593:
-----------------------------------------
I knew some were tagged, and others not. I think my main concern was whether
writes to a 64 bit value were atomic at all —but roman corrected me there.
Which leaves the issue of: are there any operations which cause problems, with
++, += and the equivalents being it.
I think we should just review them, and for all that we deem safe, add that in
big comments.
Life would be easier if java had an {{atomic}} type alongside {{volatile}},
atomic add/subtract operations are only a couple of cycles on a modern part,
plus the same impact on OoO scheduling that volatile r/w ops have.
> multiple "volatile long" field declarations exist in the Hadoop codebase
> ------------------------------------------------------------------------
>
> Key: HADOOP-12593
> URL: https://issues.apache.org/jira/browse/HADOOP-12593
> Project: Hadoop Common
> Issue Type: Bug
> Affects Versions: 2.8.0
> Reporter: Steve Loughran
> Priority: Minor
>
> If you get your IDE to scan for "volatile long", you find 20-30 entries.
> Volatile operations on `long` variables are not guaranteed to be atomic, so
> these usages can be vulnerable to race conditions generating invalid data.
> they need to be replaced by AtomicLong references, except in the specific
> case that you want performance values for statistics, and are prepared to
> take the risk
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
