Yes, we are running it with Valgrind. But must do it more. Will have
it working with Valgrind for long time and let's see.
On Wed, 23 Oct 2019 at 08:50, Santiago Gimeno <santiago.gim...@gmail.com> wrote:
>
> Hi,
>
> I haven't looked at the code, but as an idea, have you run the code with
> valgrind? It might provide some hints.
>
> Cheers,
>
> Santi
>
> El mié., 23 oct. 2019 a las 3:15, Iñaki Baz Castillo (<i...@aliax.net>)
> escribió:
>>
>> Hi,
>>
>> Although I've never been able to reproduce it (neither in Linux nor in
>> OSX) it seems that there is a bug in my libuv based C++ app related to
>> TCP. I'm a little desperate because I've spent a lot of days trying to
>> figure out where the issue is with no luck.
>>
>> A user of my software (who runs it on Linux) is experimenting TCP
>> related crashes. It's not even always the same coredump, but it seems
>> to always be related to TCP.
>>
>> Could anyone tell me if there is something wrong here?
>>
>> -------------------------------------------------
>> void TcpConnection::Close()
>> {
>> MS_TRACE();
>>
>> if (this->closed)
>> return;
>>
>> int err;
>>
>> this->closed = true;
>>
>> // Tell the UV handle that the TcpConnection has been closed.
>> this->uvHandle->data = nullptr;
>>
>> // Don't read more.
>> err = uv_read_stop(reinterpret_cast<uv_stream_t*>(this->uvHandle));
>>
>> if (err != 0)
>> MS_ABORT("uv_read_stop() failed: %s", uv_strerror(err));
>>
>> // If there is no error and the peer didn't close its connection
>> side then close gracefully.
>> if (!this->hasError && !this->isClosedByPeer)
>> {
>> // Use uv_shutdown() so pending data to be written will be sent to the
>> peer
>> // before closing.
>> auto req = new uv_shutdown_t;
>> req->data = static_cast<void*>(this);
>> err = uv_shutdown(
>> req, reinterpret_cast<uv_stream_t*>(this->uvHandle),
>> static_cast<uv_shutdown_cb>(onShutdown));
>>
>> if (err != 0)
>> MS_ABORT("uv_shutdown() failed: %s", uv_strerror(err));
>> }
>> // Otherwise directly close the socket.
>> else
>> {
>> uv_close(reinterpret_cast<uv_handle_t*>(this->uvHandle),
>> static_cast<uv_close_cb>(onClose));
>> }
>> }
>> -------------------------------------------------
>>
>> This is, I do uv_shutdown() if the connection is still open and within
>> the uv_shutdown_cb I call uv_close(). Otherwise, if the remote peer
>> closed its TCP side I directly call to uv_close():
>>
>> ---------------------------------------------------
>> inline static void onClose(uv_handle_t* handle)
>> {
>> delete handle;
>> }
>>
>> inline static void onShutdown(uv_shutdown_t* req, int /*status*/)
>> {
>> auto* handle = req->handle;
>>
>> delete req;
>>
>> // Now do close the handle.
>> uv_close(reinterpret_cast<uv_handle_t*>(handle),
>> static_cast<uv_close_cb>(onClose));
>> }
>> ---------------------------------------------------
>>
>> Both this->hasError and this->isClosedByPeer are booleans set to true
>> if there is an error in the uv_read_cb:
>>
>> ------------------------------------------------
>> inline void TcpConnection::OnUvRead(ssize_t nread, const uv_buf_t* /*buf*/)
>> {
>> MS_TRACE();
>>
>> if (nread == 0)
>> return;
>>
>> // Data received.
>> if (nread > 0)
>> {
>> // Update received bytes.
>> this->recvBytes += nread;
>>
>> // Update the buffer data length.
>> this->bufferDataLen += static_cast<size_t>(nread);
>>
>> // Notify the subclass.
>> UserOnTcpConnectionRead();
>> }
>> // Client disconneted.
>> else if (nread == UV_EOF || nread == UV_ECONNRESET)
>> {
>> MS_DEBUG_DEV("connection closed by peer, closing server side");
>>
>> this->isClosedByPeer = true;
>>
>> // Close server side of the connection.
>> Close();
>>
>> // Notify the listener.
>> this->listener->OnTcpConnectionClosed(this);
>> }
>> // Some error.
>> else
>> {
>> MS_WARN_DEV("read error, closing the connection: %s",
>> uv_strerror(nread));
>>
>> this->hasError = true;
>>
>> // Close server side of the connection.
>> Close();
>>
>> // Notify the listener.
>> this->listener->OnTcpConnectionClosed(this);
>> }
>> }
>> ------------------------------------------------
>>
>>
>> The whole code is here:
>>
>> -
>> https://github.com/versatica/mediasoup/blob/tcp-crash-wip/worker/src/handles/TcpConnection.cpp?ts=2
>> -
>> https://github.com/versatica/mediasoup/blob/tcp-crash-wip/worker/include/handles/TcpConnection.hpp?ts=2
>>
>> And an example coredump is here:
>>
>> https://mediasoup.discourse.group/t/error-worker-worker-process-died-unexpectedly/268/84?u=ibc
>>
>>
>> I'm pretty sure everything is ok in the code above. Could anyone
>> identify an issue into it?
>>
>> Really thanks a lot.
>>
>>
>> --
>> Iñaki Baz Castillo
>> <i...@aliax.net>
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "libuv" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to libuv+unsubscr...@googlegroups.com.
>> To view this discussion on the web visit
>> https://groups.google.com/d/msgid/libuv/CALiegf%3DhsERWve0k0O4Yut4SU9QPBXzbWuRS9s4une5xbyKwJw%40mail.gmail.com.
>
> --
> You received this message because you are subscribed to the Google Groups
> "libuv" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to libuv+unsubscr...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/libuv/CAAJY-XOOau8SJbWTArW8Y5EdAJCw5g8mGeyrZNTf_0RH%2Bf8ifQ%40mail.gmail.com.
--
Iñaki Baz Castillo
<i...@aliax.net>
--
You received this message because you are subscribed to the Google Groups
"libuv" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to libuv+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/libuv/CALiegfnhEO3aXiu8zHa2B6yhM%2BDfO99Oc80VJyCgTeB42k%2BVtA%40mail.gmail.com.
