>> 	if the only difference is the lack of target hardware address, does the
>> 	following diff look sane?  i really need an environment to test it...
>Trailing junk will be sent unless we fixup the length of packet in
>arprequest(), won't it?

	correct.  i've corrected it locally, as well as input side
	(when ar_hrd is not set by the sender, as well as length validation).

itojun