[Java] fix ClientConductor concurrent bug

[WorkingChen]

java.lang.IndexOutOfBoundsException: Index 1 out of bounds for length 1 at java.base/jdk.internal.util.Preconditions.outOfBounds(Preconditions.java:100) ~[?:?] at java.base/jdk.internal.util.Preconditions.outOfBoundsCheckIndex(Preconditions.java:106) ~[?:?] at java.base/jdk.internal.util.Preconditions.checkIndex(Preconditions.java:302) ~[?:?] at java.base/java.util.Objects.checkIndex(Objects.java:385) ~[?:?] at java.base/java.util.ArrayList.get(ArrayList.java:427) ~[?:?] at io.aeron.ClientConductor.checkLingeringResources(ClientConductor.java:1655) ~[aeron-client-1.46.8.jar!/:1.46.8] at io.aeron.ClientConductor.checkTimeouts(ClientConductor.java:1565) ~[aeron-client-1.46.8.jar!/:1.46.8] at io.aeron.ClientConductor.service(ClientConductor.java:1470) ~[aeron-client-1.46.8.jar!/:1.46.8] at io.aeron.ClientConductor.awaitResponse(ClientConductor.java:1529) ~[aeron-client-1.46.8.jar!/:1.46.8] at io.aeron.ClientConductor.addExclusivePublication(ClientConductor.java:473) ~[aeron-client-1.46.8.jar!/:1.46.8] at io.aeron.Aeron.addExclusivePublication(Aeron.java:294) ~[aeron-client-1.46.8.jar!/:1.46.8] at io.aeron.archive.client.AeronArchive.connect(AeronArchive.java:245) ~[aeron-archive-1.46.8.jar!/:1.46.8] at io.aeron.cluster.service.ClusteredServiceAgent.loadSnapshot(ClusteredServiceAgent.java:892) ~[aeron-cluster-1.46.8.jar!/:1.46.8] at io.aeron.cluster.service.ClusteredServiceAgent.recoverState(ClusteredServiceAgent.java:754) ~[aeron-cluster-1.46.8.jar!/:1.46.8] at io.aeron.cluster.service.ClusteredServiceAgent.onStart(ClusteredServiceAgent.java:182) ~[aeron-cluster-1.46.8.jar!/:1.46.8] at org.agrona.concurrent.AgentRunner.run(AgentRunner.java:150) ~[agrona-1.23.1.jar!/:1.23.1] at java.base/java.lang.Thread.run(Thread.java:1584) [?:?]

[WorkingChen]

We need to protect the io.aeron.ClientConductor#releaseLogBuffers method with clientLock and avoid manipulating lingeringLogBuffers objects concurrently

[TomatoCream]

We might need to lock this part too. When closing Subscription, this method will be called by a non ClientConductor thread.

    void closeImages(final Image[] images, final UnavailableImageHandler unavailableImageHandler, final long lingerNs)
    {
        for (final Image image : images)
        {
            image.close();
        }

        for (final Image image : images)
        {
            releaseLogBuffers(image.logBuffers(), image.correlationId(), lingerNs);
        }

        if (null != unavailableImageHandler)
        {
            for (final Image image : images)
            {
                notifyImageUnavailable(unavailableImageHandler, image);
            }
        }
    }

[WorkingChen]

We might need to lock this part too. When closing Subscription, this method will be called by a non ClientConductor thread.

    void closeImages(final Image[] images, final UnavailableImageHandler unavailableImageHandler, final long lingerNs)
    {
        for (final Image image : images)
        {
            image.close();
        }

        for (final Image image : images)
        {
            releaseLogBuffers(image.logBuffers(), image.correlationId(), lingerNs);
        }

        if (null != unavailableImageHandler)
        {
            for (final Image image : images)
            {
                notifyImageUnavailable(unavailableImageHandler, image);
            }
        }
    }

yeah, the closeImages method is called by internalClose(), which has already been protected by clientLock in current branch.

[vyazelenko]

False positive.

[vyazelenko]

For posterity: ConsensusModule and ClusteredServiceContainer were configured using a shared AeronArchive.Context instance which lead to ConsensusModule's Aeron instance (which has no lock) being used by multiple threads.