How to handle CRC errors?

["Preetam Verma" <pverma@marantinetworks.com>]

I have a situation like this:

                                            CPU
                                               |
                                               |
<---------------->port<---------->TCP offload engine
<------------->Logic<----------------->Application processing

The logic is going to assemble PDUs and pass them on to the application. If
a PDU that came in has CRC errors, will the application need to signal the
initiator to terminate the session? Say a PDU header from the TCP offload
engine was corrupted, how would the "logic" need to handle the data that
comes later. How do I ensure that the application receives packets in order?
How can I ensure that in such an event the PDU processing logic does not go
out of sync?

Regards

Preetam