Hi,
> Also, I believe I was wrong insisting on requiring explicit
> ETHER_ALIGN alignment. It's not really necessary if we're going to
> pull-up anyways.
Without +-ETHER_ALIGN here is easier to understand for me.
> And finally I'm using "sizeof(struct ether_header)" instead
> of the define for
After some additional consideration I think I can improve
upon my previous diff. Pull-up of the Ethernet header can
and should be made conditional: since an Ethernet header
requires 2 byte alignment, any position within a real life
mbuf is a valid one so unless there's not enough data in
the first