> Hmmm... I'll have to think about how to handle this. Probably the best way
> to do it is have the layer inactivates look to see what the count on the
> overall node is. If it's zero, fire off an update to flush that out.

It is possible that there may be direct reference to lower vnode when
layer_inactive is called and the reference may be released while layer
node is cached, isn't it?

enami.