pdp discovery coredump

[midding]

Is there an already existing issue for this?

• I have searched the existing issues

Expected behavior

PDPListener::onNewCacheChangeAdded
{
...
std::unique_lock<std::recursive_mutex> lock(*parent_pdp_->getMutex());
...
if (pdata == nullptr)
{
    ...
    parent_pdp_->assignRemoteEndpoints(pdata);
    ...
    lock.unlock();
}
else
{
    pdata->updateData(temp_participant_data_);
    ...
    lock.unlock();
} 

But if I change it this way, it is possible to encounter deadlock problems when communicating within a process. Hope to get a good solution, thanks

Current behavior

PDPListener::onNewCacheChangeAdded
{
...
std::unique_lock<std::recursive_mutex> lock(*parent_pdp_->getMutex());
...
if (pdata == nullptr)
{
   ...
   lock.unlock();
    ...
    parent_pdp_->assignRemoteEndpoints(pdata);  // use parent_pdp_ outside the lock
}
...
}
else
{
    pdata->updateData(temp_participant_data_);  // // update parent_pdp_ inside the lock
    ...
    lock.unlock();
} 

Steps to reproduce

uint32_t times = 1000;
while (times--) {
    DomainParticipant* participant_ =
        DomainParticipantFactory::get_instance()->create_participant_with_profile(0, "participant_profile51");
    if (participant_ == nullptr) {
        continue;
    }
    TypeSupport m_type(new HelloWorldPubSubType());
    m_type.register_type(participant_);
    evbs::edds::dds::Subscriber* subscriber_ = participant_->create_subscriber(SUBSCRIBER_QOS_DEFAULT, nullptr);
    if (subscriber_ == nullptr) {
        continue;
    }
    Topic* topic_(participant_->create_topic("HelloWorldTopic", "HelloWorld", TOPIC_QOS_DEFAULT));
    if (topic_ == nullptr) {
        continue;
    }
    DataReader* reader_(subscriber_->create_datareader(topic_, DATAREADER_QOS_DEFAULT));
    if (reader_ == nullptr) {
        continue;
    }
    participant_->delete_contained_entities();
    DomainParticipantFactory::get_instance()->delete_participant(participant_);
}

Using the above code, the PDP concurrency problem can be replicated by running it across multiple processes

Fast DDS version/commit

lastest master

Platform/Architecture

Ubuntu Focal 20.04 amd64

Transport layer

Default configuration, UDPv4 & SHM

Additional context

No response

XML configuration file

<?xml version='1.0' encoding='utf-8'?>
<dds>
  <profiles>
    <transport_descriptors>
      <transport_descriptor>
        <transport_id>Udp4LoTransport</transport_id>
        <type>UDPv4</type>
        <maxMessageSize>65000</maxMessageSize>
        <maxInitialPeersRange>43</maxInitialPeersRange>
        <sendBufferSize>65000</sendBufferSize>
        <receiveBufferSize>65000</receiveBufferSize>
        <TTL>10</TTL>
      </transport_descriptor>
    </transport_descriptors>
    <participant profile_name="participant_profile51">
      <domainId>0</domainId>
      <rtps>
        <useBuiltinTransports>false</useBuiltinTransports>
        <userTransports>
          <transport_id>Udp4LoTransport</transport_id>
        </userTransports>
        <builtin>
          <discovery_config>
            <discoveryProtocol>SIMPLE</discoveryProtocol>
            <EDP>STATIC</EDP>
            <static_edp_xml_config>file://participant_profile51WrapperDiscoverConfig.xml</static_edp_xml_config>
          </discovery_config>
        </builtin>
        <participantID>1</participantID>
        <name>participant_51</name>
      </rtps>
    </participant>
  </profiles>
</dds>

<?xml version='1.0' encoding='utf-8'?>
<staticdiscovery>
  <participant local_name="participant_51">
    <name>participant_51</name>
    <reader>
      <userId>110</userId>
      <entityID>10</entityID>
      <topicName>HelloWorldTopic</topicName>
      <topicDataType>HelloWorld</topicDataType>
      <topicKind>NO_KEY</topicKind>
    </reader>
  </participant>
</staticdiscovery>

Relevant log output

No response

Network traffic capture

No response

[Mario-DL]

Hi @midding
Thanks for using Fast DDS and for the report. We were already aware of the issue and we recently introduced several fixes in this regard. Please check if the issue persists with #4220.

I tested with the reproducer you provided and I can no longer reproduce the race condition across multiple processes.

[Mario-DL]

I'm afraid that according to our CONTRIBUTING.md guidelines, I am closing this issue due to inactivity. In addition, we recently merged the PR fixing a pdp data in this regard.
Please, feel free to reopen it if necessary.