[
https://issues.apache.org/jira/browse/AMQ-9658?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17924109#comment-17924109
]
ASF subversion and git services commented on AMQ-9658:
------------------------------------------------------
Commit d9e89f4b5fdb21da54e47588988161accd0fc9d6 in activemq's branch
refs/heads/main from Christopher L. Shannon
[ https://gitbox.apache.org/repos/asf?p=activemq.git;h=d9e89f4b5f ]
AMQ-9658 - Properly increment transport receive counter (#1389)
Switch to using an AtomicInteger for tracking bytes received in a
TcpTransport. This makes incrementing the counter an atomic operation.
Previously a volatile int was used and incrementing volatiles is not
atomic because it's a 3 step process of read, update, set.
This also makes a small fix to ensure that the full initialization
buffer will always be entirely read and processed when using
the auto+nio+ssl transport. Previous the code assumed only the first
command was stored in the initialization buffer but technically more
bytes could exist for a future command (even if unlikely with the
current Java implementation).
> TcpTransport volatile receiveCounter is not incremented atomically
> ------------------------------------------------------------------
>
> Key: AMQ-9658
> URL: https://issues.apache.org/jira/browse/AMQ-9658
> Project: ActiveMQ Classic
> Issue Type: Bug
> Affects Versions: 5.18.6, 6.1.5
> Reporter: Christopher L. Shannon
> Assignee: Christopher L. Shannon
> Priority: Minor
> Fix For: 6.2.0, 5.19.0, 6.1.6
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> incrementing volatile integers is not an atomic operation because the
> previous value must first be read, then incremented, and reset so it is a
> multi step process. I noticed in the TCP transport code that the receive
> counter was a volatile integer and not being incremented properly. This
> counter is should only be incremented in the same thread anyways (it can be
> read by other threads like inactivity monitor) so it may not technically be
> an issue in practice but it still should be handled properly.
> The fix is to switch to an AtomicInteger. I also noticed a minor improvement
> to make to to the NIOSSLtransport init code that is only used for the
> auto+nio+ssl transport that I will fix as well. (ensure the full
> initialization buffer will always be entirely read and processed when using
> the auto+nio+ssl transport. )
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
For further information, visit: https://activemq.apache.org/contact
