Re: [PATCH 13/14] ib_srp: Implement transport layer ping

[David Dillow <dillowda@ornl.gov>]

On Mon, 2011-12-26 at 19:39 +0000, Bart Van Assche wrote:
> On Sat, Dec 24, 2011 at 8:07 PM, David Dillow <dillowda@ornl.gov> wrote:
> > This says to me that SRP should use the dev_loss_tmo semantics, though
> > the naming of fast_io_fail vs replacement_timeout is a bit more of a
> > question than I thought. I tend to think of SRP more in terms of FC than
> > iSCSI, so I still prefer the former, but perhaps not as strongly now.
> 
> Do we need dev_loss_tmo functionality ? Since multipathd switches over
> if the active path is in the blocked state, the posted patch set
> already provides a way to make multipathd switch over if communication
> is lost.

I think so. multipathd isn't the only use case, though I think we need
to reorganize the SRP device model to fully take advantage of
dev_loss_tmo -- but that can come later. I'd like to see dev_loss_tmo
take the place of max_reconnects to give admins an easy way to set a
maximum time.
-- 
Dave Dillow
National Center for Computational Science
Oak Ridge National Laboratory
(865) 241-6602 office