[openib-general] [PATCH ] RFC IB/cm do not track remote QPN in timewait state

[Sean Hefty]

Michael S. Tsirkin wrote:
> I believe communication id should be checked to detect duplicates. Right?

Can you clarify this?  Check the remote comm id of an incoming REQ against a 
value in timewait?

> Remote QPN stale connection rule is only to avoid a case where we keep
> connection in established state forever if the remote side rebooted.

We can still end up keeping the connection state forever.  If the remote node 
reboots and re-uses the QP to connect to some other system, we'll never see the 
stale connection.  At some point, the user of the QP needs to notice that the 
connection isn't in use anymore.  Will an RC QP automatically go into the error 
state if this happens?  (I want to say no here...)

>>A stale connection check is explicitly 
>>listed under the established state.  This may help clarify stale connections.
> 
> So we agree stale connection rule only applies if connection is in established
> state?

I agree.  I'm trying to understand the full impact to the code (including users 
such as the CMA) given this interpretation, and why your patch could be failing.

- Sean