[openib-general] CMA: port 2 loopback problems

[Or Gerlitz]

On 5/8/06, Michael S. Tsirkin <mst at mellanox.co.il> wrote:
> Quoting r. Sean Hefty <mshefty at ichips.intel.com>:

>>>>Is it possible to communicate between QPs on
>>>>the same device if that device is disconnected from the fabric?

>>>Yes.

Michael, can you educate me a little here, how does this loopback
works? is there a special case within the HCA saying "if QPX.DLID is
MY LID then move packets directly from QPX.TX queue to QPY.RX queue"
or is it something else?

> > I'm wondering if the correct solution to this issue isn't to create some
> > sort of "loopback" path record.  I'm not sure what changes would be
> > required of the IB CM.

> I thought about this too. People actually do expect loopback to work when link
> is down.  I guess we could create "loopback" path record, with parameters such
> as SL editable from sysfs.

> And on port event we could try to update it from the SM.

And trivially, if there's no such special purpose rule as i have
described above, this connection is broken when the port is UP since
the SM can change the port LID that was used  in the PATH set into the
QPs. So why is it intresting to deal with the IB infrastructure
working for non active ports?

Or.