Hello, I have a working FreeRTOS + TCP/IP application which manages to receive and send responses to pings. In WireShark can see that FreeRTOS sends DHCP discover messages as both unicast and broadcast, yet there is no response received from the server (I suppose I should look for a packet directed to my MAC address, shall enter the first query in eConsiderFrameForProcessing function). What could be the reason for that? Attached are shapshot of FreeRTOS discover request and a DHCP discover of some service (which I couldn't turn off) for comparison.

https://s29.postimg.org/49b73hwxj/DiscoverPacketFreeRTOS.png https://s24.postimg.org/5b788fbv9/DiscoverPacketNotFree_RTOS.png

For some reason, setting ipconfigDRIVERINCLUDEDTXIPCHECKSUM and ipconfigDRIVERINCLUDEDRXIPCHECKSUM to 0 has solved the problem.

Hi Orit,

When developing DHCP it can be handy to install a temporary DHCP-server on your own laptop. In that way you make sure to see all messages in WireShark.

Here is an example of a DHCP server: https://sourceforge.net/projects/dhcp-dns-server I chose the option "Run stand alone", it runs from a command-prompt.

The DHCP client will use both 'unicast' and 'broadcast', it starts with a unicast (dhcpBROADCAST flag cleared).

When the answer comes in, it is directed to UDP port 68, and xPortHasUDPSocket() should return pdTRUE for that port.

Make sure that your EMAC accepts the Ethernet broadcast address of all 1's.

Our posts were crossing each other.

enabling LLMNR and setting ipconfigDHCPREGISTERHOSTNAME to 1 have solved the problem.

Good to hear. I'm not sure why changing those features would make DHCP working.

I tested the DHCP client with LLMNR, with NBNS, and also without any name service at all.

Does LLMNR cause different settings in your NetworkInterface.c ?

Hi Hein, thank you for your answer. As mentioned above, apparently setting ipconfigDRIVERINCLUDEDTXIPCHECKSUM and ipconfigDRIVERINCLUDEDRXIPCHECKSUM to 0 has solved the problem. I'll enable the Checksum Offload Engine in the ethernet controller (LAN9220) in order to set them again to 1.

All good, you're getting closer and closer.

Note that the "Checksum Offload Engine" must be studie well: test it with TCP, UDP, and ICMP messages. Some EMAC's want you to set the checksum field(s) to zero first, others don't care about the value.

Please feel free to publish you NetworkInterface.c or send it to us directly, so I can have a look.

Thanks, Hein