[openib-general] How about ib_send_page() ?

[Libor Michalek]

On Tue, May 17, 2005 at 07:08:16PM -0700, Grant Grundler wrote:
> On Tue, May 17, 2005 at 06:32:38PM -0700, Jeff Carr wrote:
> > >>>But IPoIB can't really implement NAPI since it's sending work to
> > >>>a shared HCA. 
> > 
> > Hmm. I'm not knowledgeable to know why; I'll have to take your word for 
> > it. I'm not sure yet all the conditions that the HCA can generate 
> > interrupts.
> 
> Wellll..Looks like I'm wrong. Previous email on this thread
> suggested it's possible by people who know alot more about
> it than I do. But I'm still concerned it's going to affect
> latency.

  We do something very NAPI like in SDP. 

  Basically if userspace has the socket lock and an interrupt for one
of that sockets CQs is received we mark the socket as having received
a notification and do not reenable notifications. As part of userspace 
unlocking the socket the flags are checked to see if a CQ event has
been received, if so then the CQ is polled until it is empty and then
notifications are on it are reenabled. When notifications are not
enabled the HCA will not generate interrupts for events on that CQ.


-Libor