[ewg] Re: pkey change handling patch

[Roland Dreier]

 > > Let's do it over query_pkey/query_port for now.
 > > Long term providers will just optimize these I think.
 > 
 > How ? Caching at device driver level ?

Yes... for the most part, it should be much easier to do within the
driver.  For example mthca, mlx4 and ipath at least know exactly when
the P_Key table is being changed and can just snoop the operation
without needing to worry about deferring things to a workqueue, etc.

ehca seems to have a hypercall that returns the whole P_Key table in
one go.

I think it would be fine to change the interface to something like

	query_pkey(struct ib_device *dev, u8 port, u16 start_index,
		   u16 num_pkeys, u16 *pkey)

that returns a block of P_Keys in one go, but I don't see it as that critical.