>>  > Shouldn't the FILE_UNUSE() also be moved down?  You don't want another
>>  > thread trying to use the socket while it's still being accepted.
>> FILE_UNUSE() prevents closef() from yanking the file out from under
>> you while you're still using it.
>s/FILE_UNUSE/FILE_USE/
>
>(I've been learning about this fd locking stuff whilst porting
>FreeBSD's kqueue stuff.  When the locking stuff isn't used correctly
>you either end up tsleeping in "closef" with the wait channel being
>the address of the refcount, or you panic. Fortunately, I haven't
>had any panics yet :)

	i have no idea about FILE_UNUSE part, could someone propose a patch
	if there needs to be any changes?

itojun