[openib-general] [PATCH] for 2-6-19 rdma/addr: use client registration to fix module unload race

[Or Gerlitz]

Roland Dreier wrote:
> Unfortunately I don't think this solves the module unloading race at
> all: there is still a window where code in the client module callback
> is running, but the callback has dropped all references etc. so the
> client module will happily proceed to unload.
> 
>  > At the bottom line, users must call xxx_destory_id() explicitly the
>  > xxx module would be able to handle in_callback situations.
> 
> I think this is actually a good point for the CM case at least.
> Clients already have something registered with the CM (namely the CM
> ID itself), so if we required all consumers to destroy their IDs
> explicitly, then there's no reason to add additional client
> registration.

I agree. This applies also to the rdma cm. I think that as others 
pointed, the case of new id's generated by the cm / rdma cm for incoming 
connection request might be an exception, but lets first decide this is 
the only case we need to solve, and when time comes, discuss how to do 
that.

As for client registration with the ib_mad ib_sa and ib_addr modules, i 
understand the first two where already implemented... and now Sean wants 
to add it also for the ib_addr module.

Now, this module does not have ID's, so we can either add them or 
implement the registration... let it be what ever Sean prefers, i just 
think we should not take it to the cm and rdma cm level.

Or.