[ewg] Re: [PATCH] [RFC] IB/cache: Add ib_cache report for cache in process

[Roland Dreier]

 > That does not seem trivial for all the ULPs. Can't we just assume that
 > -ESTALE would fail the ULPs instead of misleading them ? That way we
 > are not making anything behave
 > worse.

It is entirely possible that this change makes the P_Key lookup return
-ESTALE when it would have returned perfectly correct information (eg
if a P_Key is being added to the end of the table without affecting
existing P_Keys).  So this change as it stands introduces a window
where spurious failures might occur.