[
https://issues.apache.org/jira/browse/ARTEMIS-4646?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17819726#comment-17819726
]
ASF subversion and git services commented on ARTEMIS-4646:
----------------------------------------------------------
Commit f4b59c9b25d10f5738518c7115febff37a32d0b2 in activemq-artemis's branch
refs/heads/main from Albertas Vyšniauskas
[ https://gitbox.apache.org/repos/asf?p=activemq-artemis.git;h=f4b59c9b25 ]
ARTEMIS-4646 Reuse only acknowledged MQTT IDs
Generate MQTT message IDs from full allowed range of 1-65535 and skip
currently used values. Do not use atomic integer for current ID, because
all accesses and modifications are performed in synchronized context.
> Unacknowledged MQTT message ID is reused after ID generator wraparound
> ----------------------------------------------------------------------
>
> Key: ARTEMIS-4646
> URL: https://issues.apache.org/jira/browse/ARTEMIS-4646
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Components: MQTT
> Reporter: Albertas Vyšniauskas
> Priority: Major
> Time Spent: 1h 10m
> Remaining Estimate: 0h
>
> Hello,
> MQTT message IDs are currently generated by incrementing *ids* variable in
> *{color:#000000}{color:#000000}OutboundStore{color}{color}* class and do not
> perform any additional checks before trusting that it is unique. This works
> great until one or more MQTT messages are not acknowledged in timely manner
> while receiving and acknowledging other messages. A different message with
> the same ID will be sent after 32767 additional messages are received and
> acknowledged.
> There are also more issues with MQTT message ID generation:
> # *AtomicInteger* is used for *ids* variable, but all accesses and
> modifications are done in synchronized context, so there is no need for it to
> be atomic.
> # Maximum allowed *ids* value before wraparound is {*}Short.MAX_VALUE{*},
> but MQTT protocol allows values in the range of 1 to 65535 (all unsigned
> short values except 0).
> # Wraparound is done to value 1 and is immediately incremented thus skipping
> message ID 1 and using 2 as first ID after wraparound.
> # There is a harmless but wrong attempt to remove from *mqttToServerIds*
> hash map by non-MQTT message ID in the following code:
> \{_}mqttToServerIds.remove(p.getA());{_}. It is harmless because _p.getA()_
> returns *Long* value and will never match any key in *mqttToServerIds* hash
> map which uses *Integer* keys.{*}{*}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
