[openib-general] [RFC] determining which changes in svn to merge upstream or remove

[Michael S. Tsirkin]

Quoting r. Sean Hefty <sean.hefty at intel.com>:
> Subject: RE: [RFC] determining which changes in svn to merge upstream or remove
> 
> >Connections taking 60 sec to create is an issue.
> >Can you please explain how the fact that some connections are used affect
> >the time it takes to send the response?
> 
> This is in userspace, and IMO, an application issue.  Threads using established
> connections simply begin consuming all processor time.  This is while running
> under heavy load and trying to scale up the application.
>
> >Why would sending MRA be faster than sending the response?
> 
> An MRA could be sent directly by the RDMA CM in the kernel in a REQ callback,
> whereas the response requires the userspace application to poll the REQ and
> generate a REP.

I see. So it actually does look like for userspace clients, CMA should send MRA
immediately and then let userspace send REP in its own good time.

-- 
MST