>>This change looks wrong to me.  While the libc bcopy() implementation
>>is indeed documented as allowing overlapping memory objects, this is
>>not the case for its kernel pendant, which led (historically) to the
>>alternate version ovbcopy() being used for possibly overlapping
>>objects.

	but there's no ovbcopy() supplied in NetBSD kernel tree,
	it seems to me (it has been a big pain for me dealing with KAME tree)

itojun