Re: [PATCH net-next RFC v2] switchdev: bridge: drop hardware forwarded packets

[Simon Horman <simon.horman@netronome.com>]

On Thu, Mar 26, 2015 at 03:49:22PM +0100, Jiri Pirko wrote:
> Thu, Mar 26, 2015 at 03:28:28PM CET, sfeldma@gmail.com wrote:
> >On Thu, Mar 26, 2015 at 1:20 AM, Jiri Pirko <jiri@resnulli.us> wrote:
> >> Thu, Mar 26, 2015 at 08:44:27AM CET, sfeldma@gmail.com wrote:
> >>>On Wed, Mar 25, 2015 at 10:01 AM, roopa <roopa@cumulusnetworks.com> wrote:
> >>>
> >>>[cut]
> >>>
> >>>So just to keep the discussion alive (because we really need to solve
> >>>this problem), my current thinking is back to Roopa's RFC patch to
> >>>mark the skb to avoid fwding in bridge driver.  One idea (sorry if
> >>>this was already suggested, thread is long) is to use
> >>>swdev_parent_id_get op in the following way:
> >>>
> >>>1) when port interface is added to bridge, bridge calls
> >>>swdev_parent_id_get() on port to get switch id.
> >>>swdev_parent_id_get() needs to be modified to work on stacked drivers.
> >>>For example, if a bond is the new bridge port, swdev_parent_id_get()
> >>>on the bond interface should get switch_id for bond member.  We stash
> >>>the switch_id in the bridge port private structure for later
> >>>comparison.
> >>
> >> Nope, that cannot work. You can bond 2 ports each belonging to a
> >> different switch.
> >
> >Are you thinking about two switch ASICs in the same box, and bonding
> >ports from each?  Or are you thinking about bonding ports from
> >different boxes, ala MLAG?
> 
> One machine, 2 switches.
> 
> >
> >In the first case the bond would report NULL switch_id if the member
> >ports don't all have the same switch_id.  If bond switch_id is NULL,
> >the bridge driver would fwd pkts to bond and now bond would make same
> >check as bridge: if dst port switch_id is same as skb switch_id, then
> >drop pkt.  In bridge, if bond switch_id is non-NULL and matches skb
> >switch_id, then drop pkt.  So it works as desired for this case.  It
> >requires the bonding/teaming driver to modify the default behavior for
> >swdev_parent_id_get() to only return switch_id if all ports agree on
> >switch_id.
> >
> >For second case using MLAG, I suspect bond member port switch_ids
> >would likely be different, and so with same logic in bonding/bridge
> >drivers as above in first case, the pkt would be fwded down.
> >
> >Is there another case to consider?  I think converting
> >swdev_parent_id_get() to use same algo we have for stp, allowing for
> >any layer to override like in my bonding example, will have benefits
> >down the road.
> >
> >What is the argument for not allowing stacked version of swdev_parent_id_get()?
> 
> That was suppose wo identify a switch port. "ip link" will show you that
> and you see right away what is going on. If bond implements that, that
> brigs a mess. I don't like that.

I'm not sure that I follow how this messes things up from a bridging point
of view. Would it help if bonds consistently returned a NULL parent id
even if all its slaves have the same parent id?