NDK is setting Frame Check Sequence to zero in transmitted TCP/IP packets

David,

I thought your question is why some MCSDK/PDK/NDK combination has TCP Frame Check Sequence field set to zero.

There is what I tested:

- on MCSDK 2.1.2.6, I imported client (not hello world) program without any code change

- I selected NDK 2.22.0.06 as you mentioned and re-build

- Load the application to C6678 DSP and run, I got an IP address

- From Windows, I run the C:\ti\ndk_2_22_00_06\packages\ti\ndk\winapps\send or recv program to do TCP with EVM and I captured packet with Wireshark

I also saw FCS zeroed out in this release combination. So, I don't know why in your case, FCS is non-zero and accepted by client.

Regards, Eric

Hi Eric

Thanks for testing this.  I think we are mis-communicating a little.  My point is that surely the Frame Check Sequence should not be zero.  I don't know why your winapp program worked in that situation but surely, in normal cases, the Frame Check Sequence should be non-zero?

Don't you see Wireshark mark the packets in red, indicating an error, because of the zero'd FCS?

Regards, David

David,

I tried several NDK releases, I always saw FCS is zero (red alerted in Wireshark). So, my point is that FCS zero issues has been there for quite sometimes, not a new NDK release introduce the behavior.

I don't know much about the history of NDK, maybe in TCP case, this is a reliable transimssion we don't need FCS for additional correction.

Regards, Eric

David,

Can you confirm a non-zero FCS for a similar TNyq application on previous NDK?

If so we know it's specific to new platform / new SDK components.

Regards,

Tom

Just for reference, this thread also discusses this topic.