[openib-general] APM support in openib stack

[Michael S. Tsirkin]

> >Hmm, I think we agreed the sensible thing for stale connections is to prevent
> >QPN reuse for timewait period below the verb interface?
> >
> >I didn't get to this yet, but you didn't change your mind, did you?
> 
> I didn't change my mind about QP reuse.
> 
> I was referring to the case where: node A crashes, comes back up, and tries to
> reconnect to node B using the same QP.

But it will get a new QP in node B, won't it, even if sockets
are used for CM?



-- 
MST