I doubt that there is a measurable performance impact here but I think the resulting code is clearer and the scope of the object lock is further reduced.
We are very close to being able to eliminate explicit object locking from kern/*. It would be much easier to reason about the system if it was contained in vm/.