none
Windows CE 6.0 RC3 - Socket communication does not work when adapter is configured for static IP address RRS feed

  • Question

  • Hi all,

    I'm trying to implement time synchronization in a local network. One device in this network is based on a Kontron ETX-DC (X86) module running Windows CE 6.0 RC3 (with the 20141231 roll-up installed). The network adapter is configured to use static IP address. The Windows PC (also with static IP) running the NTP server is set as Default Gateway.

    My first attempt was to use the timesvc service. It did not work and in the line of trouble-shooting I wrote a test application for communication with the local NTP server.

    The test application appeared to be working as it should when sending the request, socket send reported that the correct number of bytes had been sent, netstat indicated the an UDP message had been sent. The only problem was that no UDP/NTP message was sent over the network (I am sniffing the network with Wireshark). Ping works in both directions. I modified the test application to send an ICMP message (ping request) and that message was sent over the network!

    When I changed to DHCP assigned IP addresses for the time server and for the Windows CE device (an option only available in my test setup) the test application was able to communicate with the NTP server as intended. Wireshark did, of course, also catch the request and the response messages on the network.

    The routing table looks the same in both cases (except for the specific IP addresses).

    Has anyone seen this behavior? What is missing in the static IP case making some lower level (the TCP/IP stack?) silently discard messages?

    Thx in advance,

    /Anders

    Progress update April 27

    It was discovered that the Windows CE (or the hardware) silently drops packets sent to the 192.0.4.XX net! Is this an expected behavior? I think no but on the other hand I'm not an expert on TCP/IP! Everything works when addresses in the 192.168.1.XX is used!

    • Edited by Anders Th Monday, April 27, 2015 7:40 AM Some progress
    Tuesday, April 21, 2015 1:07 PM

All replies

  • I'm always suspicious when a subnet has a tuple of "0". For your network with addresses in range 192.0.4.XX, what's the subnet mask? It's quite possible that the masking/subnet identification code in Windows CE works a little differently than desktop Windows in that case...

    Paul T.

    Tuesday, April 28, 2015 8:19 PM
  • Hi Paul,

    Thanks for looking into my little problem.

    The subnet mask is 255.255.255.0.

    I haven't had time to study the behavior more in detail. The solution will (probably) be to change to a more conventional private subnet without a "0" tuple.

    I agree that refraining from the use of "0" in the subnet address is a very good idea :-)

    /Anders

    Tuesday, May 5, 2015 8:57 AM