ARP problem when running UDP server and client in the same host

Libraries
1

Hello all,

We have ported the FreeRTOS TCP stack to our platform. In order to test it we copied the UDP example from the demos in Github.

It’s nothing fancy, very simple server and client.

The problem

  • If I run the client task by itself, all good no problems
  • If I run the server task by itself, all good no problems

The problem happens when I try to run both tasks at the same time. I have a termshark terminal connected to my network. When I run the client by itself the 1st message out of my host is a ARP request for the correct address of the server (192.168.0.253).

But when I run both tasks at the same time, the same ARP request is now for 0.0.0.0.

However, when I talk to the server from another host (the one of IP addr 192.168.0.253), the ARP tables in the TCP stack are updated and now the client starts sending data to the correct IP address.

Why is this? Any explanations would be really helpful

2

Which buffer allocation scheme are you using? Is it possible you run out of memory of buffers?

Is your driver thread safe? Which device did you port to?

3

I am using BufferAllocation_2.c, with the corresponding heap_4.c.

I know I’m not running out of buffers because I have a UDP server listening for messages, and in a LAN I can send about 5000 messages per second with 0 packets dropped and all data being valid, from another host.

I can also send the same amount of data from the MCU to the same host, again, on a LAN, with 0 packets dropped and no data corruption.

The device is the S32K344 automotive System on Chip from NXP semiconductors.

The driver itself is not “thread safe”. However, I was convinced that since the TCP stack is thread safe there was no need for my low level interface output to also be thread safe.

Could you explain a little why?

Thank you

4

Hi Mateo (and welcome btw),

what happens if your client attempts to connect to localhost (127.0.0.1) instead of the public address?

Does your device obtain a DHCP address, or is it static?

5

If I connect the client to local host (127.0.0.1), the server does not receive any messages.

I’m using static IP addresses

6

Update

I am now trying to run a TCP server on my microcontroller. However, here are the error messages I’m getting

image

And here’s the wireshark capture from the host where the client is running

image
image1253×142 11.9 KB

Any idea what the issue may be in this instance?

7

Maybe a corrupted packet (eg wrong checksum)? Can we see one of the packets tha appears to be dropped?

8

It looks like the device on 192.168.0.253 doesn’t not receive the reply from 192.168.0.217.
The strange thing is that x.217 is able to send a UDP packet, which is received by x.253
Who is calculating the checksum? The EMAC, the Network Interface module, or the IP-task?
Are checksums offloaded?

EDIT

x.217 is able to send a UDP packet

that should be an ARP packet, and ARP packets do not have their own checksum.