[En-Nut-Discussion] Missing [SYN, ACK] packets when NutTcpConnect() attempts to connect to a .NET TcpListener socket server
Erik Lindstein
erik at ledutveckling.com
Fri Nov 9 15:33:09 CET 2007
I have a Ethernut 2.0 (ethernut-4.2.1) board connected to a PC thru ethernet.
In the ethernut i have client thread that handles some dataexchange
with the PC(running a pretty simple .NET TcpListener TCP socket server
on WinXP)
The ethernut thread is supposed to connect, exchange some data(a
couple of xml formatted lines), read back a command from the PC,
disconnect and wait 300ms then start all over again.
For most of the time this works fine but sometimes(for some reason)
the systems just don't want to start exchange data.
If i analyze the trafic with wireshark i can se that the ethernut
board sends [SYN] packets but the PC doesn't respond with anything.
I don´t have any knowledge about how the low level TCP communication
work's but i can se that when the communication works the PC sends
[SYN, ACK]
after ethernut sends [SYN].
When the communication isn't working the ethernut sends a couple of
[SYN] then after a while the PC starts responding with [SYN, ACK] and
the communication
starts working again.
While this ethernut client dossnt get any response from the PC the PC
might still at the same time communicate with 3-4 other clients
successfully.
So its usually not all the connecting clients that get the silent
treatment just a couple of them.
The best way to get this problem is to have the PC software running
with connected clients and restart the clients.
If i let the ethernut clients bootup then start trying to connect a
couple of times and then start the PC software the problems usully
dossnt show.
But when the communication is up and running it work's fine(for days)
Its just the first connecting moment that sometimes doesn't come thru.
I'm not sure who to blame, ethernut or the PC.
I wrote a simple tcpclient in .NET that connects to the server and it
looks like that one work's fine but still this problem doesn't show it
self all the time.
Perhaps someone in the list have some ideas on how to fix this or at
least sort out how to blame ;-)
Bellow is some data from wireshark ( Ethernut IP = 192.168.0.60, PC =
192.168.0.115 )
------------------------------------------------------------------------
No. Time Source Destination Protocol Info
1 0.000000 192.168.0.60 192.168.0.115 TCP
4149 > 9050 [SYN] Seq=0 Len=0 MSS=536
No. Time Source Destination Protocol Info
2 0.000018 192.168.0.115 192.168.0.60 TCP
9050 > 4149 [RST, ACK] Seq=0 Ack=1 Win=0 Len=0
No. Time Source Destination Protocol Info
3 8.491477 192.168.0.60 192.168.0.115 TCP
4097 > 9050 [SYN] Seq=0 Len=0 MSS=536
No. Time Source Destination Protocol Info
4 8.491514 192.168.0.115 192.168.0.60 TCP
9050 > 4097 [SYN, ACK] Seq=0 Ack=1 Win=65535 Len=0 MSS=1460
No. Time Source Destination Protocol Info
5 8.492851 192.168.0.60 192.168.0.115 TCP
4097 > 9050 [RST] Seq=1 Len=0
( Here i reset the ethernut client )
( Here the communication isn't working )
No. Time Source Destination Protocol Info
6 13.408382 192.168.0.60 192.168.0.115 TCP
4098 > 9050 [SYN] Seq=0 Len=0 MSS=536
No. Time Source Destination Protocol Info
7 14.580253 192.168.0.60 192.168.0.115 TCP
4098 > 9050 [SYN] Seq=0 Len=0 MSS=536
No. Time Source Destination Protocol Info
8 15.758478 192.168.0.60 192.168.0.115 TCP
4098 > 9050 [SYN] Seq=0 Len=0 MSS=536
No. Time Source Destination Protocol Info
9 16.936782 192.168.0.60 192.168.0.115 TCP
4098 > 9050 [SYN] Seq=0 Len=0 MSS=536
No. Time Source Destination Protocol Info
10 18.115065 192.168.0.60 192.168.0.115 TCP
4098 > 9050 [SYN] Seq=0 Len=0 MSS=536
No. Time Source Destination Protocol Info
11 18.115084 192.168.0.115 192.168.0.60 TCP
9050 > 4098 [SYN, ACK] Seq=0 Ack=1 Win=65535 Len=0 MSS=1460
No. Time Source Destination Protocol Info
12 18.117012 192.168.0.60 192.168.0.115 TCP
4098 > 9050 [ACK] Seq=1 Ack=1 Win=3216 Len=0
( Here its up and running again )
No. Time Source Destination Protocol Info
13 18.121677 192.168.0.60 192.168.0.115 TCP
4098 > 9050 [PSH, ACK] Seq=1 Ack=1 Win=3216 Len=401
No. Time Source Destination Protocol Info
14 18.121805 192.168.0.115 192.168.0.60 TCP
9050 > 4098 [PSH, ACK] Seq=1 Ack=402 Win=65134 [TCP CHECKSUM
INCORRECT] Len=14
------------------------------------------------------------------------
The ethernut tcpclient thread ( Some code arent included but i think
it should be enough )
------------------------------------------------------------------------
THREAD(InetdThread, arg)
{
TCPSOCKET *sock;
FILE *stream;
u_long rip = inet_addr(TCPSERVERADDR);
uint8_t *start;
uint8_t cmd[20], value[20];
u_long tmo = TCP_TIMEOUT;
uint32_t exchange_timeout = TCP_CLIENT_NO_CONNECTION_TIMEOUT;
uint8_t buffer[40];
int socket_error = 0;
for (;;)
{
if ((sock = NutTcpCreateSocket()) != 0)
{
NutTcpSetSockOpt(sock, SO_RCVTIMEO, &tmo, sizeof(tmo));
NutTcpSetSockOpt(sock, SO_SNDTIMEO, &tmo, sizeof(tmo));
if(NutTcpConnect(sock, rip, TCPSERVERPORT) == 0)
{
stream = _fdopen((int) ((uptr_t) sock), "r+b");
if(stream != 0)
{
fprintf_P(stream, info_P, INFO_P_ARGS);
fflush(stream);
fgets(exchange_data_buff,
sizeof(exchange_data_buff), stream);
if((start = strchr(exchange_data_buff, '\r')) !=
0) *start = 0;
if((start = strchr(exchange_data_buff, '\n')) != 0) *start = 0;
if((start =
strchr(exchange_data_buff, ':')) != 0)
{
strlcpy(cmd,exchange_data_buff, (start+1-exchange_data_buff));
start++;
strcpy(value,start);
//
if(strncmp(cmd, "cmd",
sizeof(cmd)) == 0)
{
if(strncmp(value,
"wdtreset", sizeof(cmd)) == 0)
{
if(status & (1UL<<WDT_ENABLE))
{
wdt_reset();
}
else
{
wdt_reset();
wdt_enable(WDTO_2S);
status |= (1UL<<WDT_ENABLE);
}
}
}
exchange_timeout =
TCP_CLIENT_CONNECTION_RATE;
}
else
{
socket_error = NutTcpError(sock);
sprintf(buffer,"Connect
error\nCode: %d", socket_error);
lcd_menu_write_text_to_display(buffer, 3000);
exchange_timeout =
TCP_CLIENT_NO_CONNECTION_TIMEOUT;
}
}
fclose(stream);
NutTcpCloseSocket(sock);
}
}
else
{
exchange_timeout = TCP_CLIENT_NO_CONNECTION_TIMEOUT;
}
NutSleep(exchange_timeout);
}
}
---------------------------------------------------
/Erik
More information about the En-Nut-Discussion
mailing list