App crash calling QTcpSocket::waitForConnected

[Kiyov09]

Hi, I have the following error:
the next function is called multiple times from diferent threads and after some time, say 20 minutes, crash in socket->waitForConnected(1000). Anybody? Thanks.

bool Utils::testConnectionToServer(QString host, int port)
{
    QTcpSocket * socket = new QTcpSocket();
    socket->connectToHost(host, port, QIODevice::ReadOnly);
    if (socket->waitForConnected(1000)) {
        socket->abort();
        socket->close();
        delete socket;
        socket = 0;
        return true;
    } else {
        qDebug() << "### " << socket->errorString() << " ###";
        delete socket;
        socket = 0;
        return false;
    }
}

[mrjj]

Hi and welcome
Have you checked that new can allocate always ?
You might fragment memory if calls happens really fast.

Also , any chance that the threads call it at the SAME time ?
so following situation could arise ? (when times perfectly)
QTcpSocket * socket = new QTcpSocket(); << Thread 1 here
socket->connectToHost(host, port, QIODevice::ReadOnly) ; << Thread 2 here

[Kiyov09]

Previously he had a version of the function that used QMutexLocker to avoid just that but also crash in waitForConnected. Specifically, the error displayed is as follows:

*** buffer overflow detected ***: /home/kiyov/RepoSVN/tss-streamingserver/trunk/StreamingServer/StreamingServer/Server terminated
======= Backtrace: =========
/lib/x86_64-linux-gnu/libc.so.6(+0x731ff)[0x7ffff493b1ff]
/lib/x86_64-linux-gnu/libc.so.6(__fortify_fail+0x37)[0x7ffff49be4c7]
/lib/x86_64-linux-gnu/libc.so.6(+0xf46e0)[0x7ffff49bc6e0]
/lib/x86_64-linux-gnu/libc.so.6(+0xf6437)[0x7ffff49be437]
/usr/lib/x86_64-linux-gnu/libQtNetwork.so.4(+0xd4e25)[0x7ffff5968e25]
/usr/lib/x86_64-linux-gnu/libQtNetwork.so.4(+0xd2a8d)[0x7ffff5966a8d]
/usr/lib/x86_64-linux-gnu/libQtNetwork.so.4(_ZN15QAbstractSocket16waitForConnectedEi+0x14a)[0x7ffff595f27a]
/home/kiyov/RepoSVN/tss-streamingserver/trunk/StreamingServer/StreamingServer/Server[0x439eda]
/home/kiyov/RepoSVN/tss-streamingserver/trunk/StreamingServer/StreamingServer/Server[0x4360c2]
/home/kiyov/RepoSVN/tss-streamingserver/trunk/StreamingServer/StreamingServer/Server[0x439a9e]
/home/kiyov/RepoSVN/tss-streamingserver/trunk/StreamingServer/StreamingServer/Server[0x4151ab]
/home/kiyov/RepoSVN/tss-streamingserver/trunk/StreamingServer/StreamingServer/Server[0x458e9b]
/usr/lib/x86_64-linux-gnu/libQtCore.so.4(_ZN7QObject5eventEP6QEvent+0x1e1)[0x7ffff5544e11]
/usr/lib/x86_64-linux-gnu/libQtCore.so.4(_ZN16QCoreApplication14notifyInternalEP7QObjectP6QEvent+0x8d)[0x7ffff552b71d]
/usr/lib/x86_64-linux-gnu/libQtCore.so.4(_ZN23QCoreApplicationPrivate16sendPostedEventsEP7QObjectiP11QThreadData+0x231)[0x7ffff552efc1]
/usr/lib/x86_64-linux-gnu/libQtCore.so.4(+0x1b5bd3)[0x7ffff555abd3]
/lib/x86_64-linux-gnu/libglib-2.0.so.0(g_main_context_dispatch+0x24d)[0x7ffff641dc5d]
/lib/x86_64-linux-gnu/libglib-2.0.so.0(+0x49f48)[0x7ffff641df48]
/lib/x86_64-linux-gnu/libglib-2.0.so.0(g_main_context_iteration+0x2c)[0x7ffff641dffc]
/usr/lib/x86_64-linux-gnu/libQtCore.so.4(_ZN20QEventDispatcherGlib13processEventsE6QFlagsIN10QEventLoop17ProcessEventsFlagEE+0x5d)[0x7ffff555ad1d]
/usr/lib/x86_64-linux-gnu/libQtCore.so.4(_ZN10QEventLoop13processEventsE6QFlagsINS_17ProcessEventsFlagEE+0x41)[0x7ffff552a271]
/usr/lib/x86_64-linux-gnu/libQtCore.so.4(_ZN10QEventLoop4execE6QFlagsINS_17ProcessEventsFlagEE+0x1a5)[0x7ffff552a5d5]
/usr/lib/x86_64-linux-gnu/libQtCore.so.4(_ZN16QCoreApplication4execEv+0x99)[0x7ffff5530059]
/home/kiyov/RepoSVN/tss-streamingserver/trunk/StreamingServer/StreamingServer/Server[0x40cf12]
/lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xf5)[0x7ffff48e9b45]
/home/kiyov/RepoSVN/tss-streamingserver/trunk/StreamingServer/StreamingServer/Server[0x40ce09]

[mrjj]

Hi
it looks like buffer overrun happens inside Qt.
Have you searched the bug database for issues sounding like this`?

[Kiyov09]

@mrjj no, where I can search that bug?

[mrjj]

@Kiyov09
https://bugreports.qt.io/secure/Dashboard.jspa

https://bugreports.qt.io/browse/QTBUG-7753?jql=text ~ "waitForConnected"

[Kiyov09]

@mrjj Thanks

[mrjj]

@Kiyov09
Hope you find something.

Is it always after the same amount of calls?

[Kiyov09]

@mrjj I do not know, I have done this test.

[SGaist]

Hi and welcome to devnet,

I'd change the delete statement by socket->deleteLater() to ensure that you are not deleting the sockets while there's still an event going on the socket.

[Kiyov09]

@mrjj No, but it made approximately 1348 calls.

@SGaist Even changing the delete statement by socket->deleteLeter() the problem persist.

Apparently all starts in the call to QNativeSocketEnginePrivate::nativeSelect.

This is the stack trace
0 __GI_raise raise.c 56 0x7ffff48fd107
1 __GI_abort abort.c 89 0x7ffff48fe4e8
2 __libc_message libc_fatal.c 175 0x7ffff493b204
3 __GI___fortify_fail fortify_fail.c 31 0x7ffff49be4c7
4 __GI___chk_fail chk_fail.c 28 0x7ffff49bc6e0
5 __fdelt_chk fdelt_chk.c 25 0x7ffff49be437
6 QNativeSocketEnginePrivate::nativeSelect qnativesocketengine_unix.cpp 1073 0x7ffff5968e25
7 QNativeSocketEngine::waitForWrite qnativesocketengine.cpp 947 0x7ffff5966a8d
8 QAbstractSocket::waitForConnected qabstractsocket.cpp 1814 0x7ffff595f27a
9 Utils::testConnectionToServer Utils.cpp 106 0x43a099
10 MyAmqpClient::publishMessage myamqpclient.cpp 206 0x4361d2
11 Utils::sendMessageToAmqp Utils.cpp 73 0x439bae
12 StreamingServer::checkSystemResources streamingserver.cpp 1083 0x4152bb
13 StreamingServer::qt_static_metacall moc_streamingserver.cpp 107 0x459099
14 QObject::event qobject.cpp 1222 0x7ffff5544e11
15 QCoreApplication::notifyInternal qcoreapplication.cpp 955 0x7ffff552b71d
16 sendEvent qcoreapplication.h 231 0x7ffff552efc1
17 QCoreApplicationPrivate::sendPostedEvents qcoreapplication.cpp 1579 0x7ffff552efc1
18 QCoreApplication::sendPostedEvents qcoreapplication.cpp 1472 0x7ffff552f453
19 sendPostedEvents qcoreapplication.h 236 0x7ffff555abd3
20 postEventSourceDispatch qeventdispatcher_glib.cpp 280 0x7ffff555abd3
... <More>

[Jodavi]

Hello everyone,

sorry for reviving that old thread, but it looks nearly the same as my problem.

I am using QT 5.6.2 on a headless gentoo. My (QCore) application is using sockets with multiple threads.

The threads are in a threadpool. Every thread sends its result to the same target after finishing its work (~0.5 to 10 seconds). Each "sends its result" involves a code like this:

int WorkerThread::sendMessageOverTCPSocket(const QString & result, const QString &ip, const quint16 &port)
{
  try
  {
	QTcpSocket socket;
	socket.connectToHost(ip,port);
	if(socket.waitForConnected(10203))
	{
		socket.write(result.toStdString().c_str(),result.length());
		if(!socket.waitForBytesWritten(5134))
		{
			ss << m_id <<":timeout while writing data to socket";
			LOG(ss.str().c_str()); ss.str("");
			return Constants::TCPSOCKETERROR::TIMEOUT_WAIT_WRITTEN;
		}
	}
	else
	{
		ss << m_id <<":could not connect to socket";
		LOG(ss.str().c_str()); ss.str("");
		return Constants::TCPSOCKETERROR::TIMEOUT_CONNECTED;
	}
	socket.close();
  }
  catch (std::exception &e)
  {
	qFatal("Error(%s) sending results ",e.what());
  }
  catch (...)
  {
	qFatal("Unknown error occured while trying to send results");
  }

  return 0;
}

After about 1100 -1300 threads, the application always terminates. This may take up to a couple of days until it happens "naturally". The one I am currently investigating terminates with SIGABRT.

No immediate output or log can be found.

This is a bt from a sample coredump

#0  0x00007faa868442b8 in raise () from /lib64/libc.so.6
#1  0x00007faa868456da in abort () from /lib64/libc.so.6
#2  0x00007faa868803f1 in __libc_message () from /lib64/libc.so.6
#3  0x00007faa86908b27 in __fortify_fail () from /lib64/libc.so.6
#4  0x00007faa86906bb0 in __chk_fail () from /lib64/libc.so.6
#5  0x00007faa86908a9a in __fdelt_warn () from /lib64/libc.so.6
#6  0x00007faa8a0737a4 in ?? () from /usr/lib64/libQt5Network.so.5
#7  0x00007faa8a070ce2 in ?? () from /usr/lib64/libQt5Network.so.5
#8  0x00007faa8a06a5dc in QAbstractSocket::waitForConnected(int) () from /usr/lib64/libQt5Network.so.5

The line number points to the if(socket.waitForConnected(10203))

Can someone hint me to the right direction? There seems to be no solution posted to the problem of OP

Edit: Updated code

[SGaist]

Hi and welcome to devnet,

Are you sure you are cleaning up properly your resources when you're done ?

[Jodavi]

@SGaist
Thanks for the welcome!

The resources in use are all local and should get cleaned up, once they are out of scope.
See the update in my post above.

I am not sure whether this really is "properly cleaning up". Actually, I assumed the DTOR of the QTCPSocket is enough?
From the docs:

[virtual] QTcpSocket::~QTcpSocket()
Destroys the socket, closing the connection if necessary.
See also close().

My current vague suspicion:
Is the line socket.close(); positioned correctly?

[SGaist]

Looks like it.

How are you handling your thread pool ?