SM320F28335-EP: SHA-256 (ti implementaion) faulty hashes above message size 8192 bytes

this means that the first message length that generates a faulty sha256 hash is 8248 bytes.

Emil,

We will assign this to our security expert.  Please expect a reply early next week.

Thanks and regards,

Vamsi

Hi Emil,

This code was developed for the MSP430 class of products. As such, I'm not sure how compatible it is with C2000 MCUs. Do you have any additional information that could give some clues as to why this is failing? Did you get any compiler or linker warnings or errors during the port to C2000?

Follow up question, are you saying that for messages <8248 bytes the algo generates the correct result?

>Do you have any additional information that could give some clues as to why this is failing?

No i have no idea. but i guess that one cuurlprit could be that some counters use int keyword instead of fixed sized counters (e.g. uint16_t)? and i think the "i" variable should be a 32bit int? 

> Did you get any compiler or linker warnings or errors during the port to C2000?

no.

> Follow up question, are you saying that for messages <8248 bytes the algo generates the correct result?
yes according to my limited testing that seams to be the case. and after 8248 the error repeats every 64 bytes (the chunk size of sha256).

Hi

I will take over this issue from Emil

I don't know the libraries you are using, so I can't comment on whether there is a coding bug there or not, but it very well could be the cause of the problem.

Have you looked at the .map file generated by the linker to see if maybe this is a memory allocation issue? What optimization level are you using for the compiler? Just trying to rule out a compiler or linker problem.

Hello Gus

Thank you for your response. We will of course readily provide any information or files you might seem necessary, just let us know what you want to see and we'll find a minimal viable example for you to look at.

We are not really using any libraries, not that might affect the performance of the sha256 implementation at least. I am not really sure what to look for in the map file but I think we have allocated all the necessary working memory. No compiler optimization

Hi Johan,

Let me check with the author of this app note to see if they have any guidance. I may need a day or two before I can report back.

Regards,

James