[ewg] Re: pkey change handling patch (was Re: bugs to fix for OFED 1.2 RC1)

[Michael S. Tsirkin]

> 
> I looked at cache.c and you are right. Maybe we should either
> 1. report events after cache has been updated
> or
> 2. make cache queries error out (EBUSY?) if cache hs not updated yet.
> 
> Option 1 requires core changes, option 2 - ULP changes
> 
> I would be inclined to go for 2.
> Roland?

So, it seems option 2 is a lot of work.
Option 1 would mean making cache a special client, and moving
notification mechanism into cache itself.

This way a client gets an event after cache has already been updated.
So maybe that's the best way to handle this?

-- 
MST