Andrew Doran <ad@netbsd.org> writes:

> On the other hand, maybe it's time we looked at pushing the locking back
> into the file systems and being more strict about the reference counting, so
> that these kinds of issues don't continue to appear. :-). What do people
> think of the idea? Some if the issues involved in doing that I know of are..
>
> o VOP_LOCK and VOP_UNLOCK would need to go away. To make VOP calls you'd
>   only need a reference to the vnode.

First, note that I don't really understand vnode locking as it is, just
mostly.

With lookup, there's a notion that a lock is obtained on a vnode, and
then a locked child is returned, enabling traversing down the tree.  I
am concerned about adverse consequences to lookup semantics under flux
with a rule that says each vop has to lock/munge/unlock internally with
no locks held across calls.  But I can't articulate what bad things
might happen, and I'm afraid I'm blurring refs and locks in my head.