[openib-general] ipoib mcast questions...

[Arthur Jones]

hi hal, ...

On Fri, Oct 06, 2006 at 04:09:05PM -0400, Hal Rosenstock wrote:
> On Fri, 2006-10-06 at 15:47, Arthur Jones wrote:
> > hi hal, ...
> > 
> > On Fri, Oct 06, 2006 at 11:26:26AM -0400, Hal Rosenstock wrote:
> > > > [...]
> > > > i'm imagining that all the "proprietary" eth
> > > > interfaces + ipoib need to do about the same
> > > > thing when it comes to registering with mcast
> > > > groups.  would you (all) be averse to pulling some
> > > > of the mcast group registration code out into
> > > > the core ib driver for all to use?
> > > 
> > > Isn't this already done with Sean's multicast work ?
> > 
> > after reading the code, iiuc, sean's work provides
> > nice infrastructure for ib_multicast group join/leave.
> > i was thinking about one more level up, i.e. generic
> > _net_ multicast join/leave infrastructure.  i'm not
> > sure exactly how it would go -- but i think all the ib
> > net_devices are going to need a way to associate a
> > multicast hw addr w/ a live mgid.
> 
> Don't IPmc addresses translate to MGIDs per the RFC ?

that's a different problem than the one i'm
trying to address.  i think you're talking
about mapping ip mcast addresses to "hardware"
addresses.  rfc4391 tells ipoib how to do that,
for the virtual ethernet devices, we'll need
to come up w/ something different...

> MGIDs are not hardware addresses (MLIDs are).

mgids are generated from the mc_list->dmi_addr.
this is a "hardware" address to the linux net
code.  i'm looking for commonality to reduce
duplicated code.  we all (ipoib + virtual eth)
need to associate mgids, however we got them,
with the mlids (i think).  i'm guessing we'll
do it in a very similar way...

arthur