nvmf/rdma host crash during heavy load and keep alive recovery

['Christoph Hellwig']

On Fri, Sep 23, 2016 at 04:57:21PM -0700, Sagi Grimberg wrote:
> I'm still trying to understand how it is possible to
> get to a point where the request queue is stopped while
> the hardware context is not...

Not sure either, I need to deep dive into the code.  But just a week
or two ago Bart posted some fixes in this area for SRP, so that's where
I will start my deep dive.