> I like the idea, but I'd suggest we do it differently. I'd suggest we add
> a new lock manager call, lock_haslock(), which takes a lock and a proc
> (pointer), and indicates if that proc is the lock holder. Then we add
> vn_haslock which takes a vnode and a proc and calls lock_islocked on the
> lock.

Doing this in a way which preserves abstraction boundaries would
require adding (at a minimum) vn_haslock() and VOP_HASLOCK() as well
as lock_haslock.  Doing it via lock flags does not require as many
interface changes.

					- Bill