Hello! I have implemented a simple TCP server on lwIP 1.4.1 running on chibiOS (a realtime OS) with the netconn api on an STM32F746.
It goes like this:
TCP server is opened and waits for connections:
conn = netconn_new(NETCONN_TCP);
LWIP_ERROR("http_server: invalid conn", (conn != NULL),
chThdExit(MSG_RESET););
netconn_bind(conn, NULL, SERVER_THREAD_PORT);
netconn_listen(conn);
err_t err;
while (true) {
err = netconn_accept(conn, &newconn);
if (err != ERR_OK)
continue;
serve(newconn);
netconn_delete(newconn);
}
The serve function kicks in when a connection was accepted and receives some
data from the client:
void serve(struct netconn * conn)
{
struct netbuf *inbuf;
uint8_t *buf;
u16_t buflen;
err_t err;
/*
* Read the data from the port, blocking if nothing yet there.
* Assume that the request is in one netbuf.
*/
err = netconn_recv(conn, &inbuf);
if (err == ERR_OK) {
netbuf_data(inbuf, (void **)&buf, &buflen);
// evaluate the input!
}
/* Close the connection */
netconn_close(conn);
/*
* Delete the buffer (netconn_recv gives us ownership,
* so we have to make sure to deallocate the buffer)
*/
netbuf_delete(inbuf);
}
My TCP client (it is a Matlab PC) connects, sends some data, and disconnects.
The TCP server also closes the connection on its side and then deletes the
netconn structure.
The problem is that I get spurious ACK and FIN, ACK in Wireshark. I do not get
those, when I do not let Matlab close the TCP connection on client side.
However, I have to, since I want to open it again. The spurious retransmission
does not hurt my system but I wonder why it exists.
Is there something that is obviously wrong? Should I NOT close the connection
on the TCP server side? Find a capture attached.
Thanks in advance,
Adrian
dump.pcapng
Description: dump.pcapng
_______________________________________________ lwip-users mailing list lwip-users@nongnu.org https://lists.nongnu.org/mailman/listinfo/lwip-users
