You can do whatever you want in the fpga. The critical part for you is
to identify the phy and mac layer components that are latency sensitive
and implement those on the fpga. You don’t necessarily have to slap the
entire mac and phy onto the fpga.
Tim