Re: [PATCH net-next 1/2] mlxsw: spectrum_router: Offload blackhole routes

[Ido Schimmel <idosch@mellanox.com>]

On Thu, Feb 07, 2019 at 04:40:11PM -0800, David Ahern wrote:
> On 2/6/19 11:42 AM, Ido Schimmel wrote:
> > Create a new FIB entry type for blackhole routes and set it in case the
> > type of the notified route is 'RTN_BLACKHOLE'.
> > 
> > Program such routes with a discard action and mark them as offloaded
> > since the device is dropping the packets instead of the kernel.
> > 
> > Signed-off-by: Ido Schimmel <idosch@mellanox.com>
> > Acked-by: Jiri Pirko <jiri@mellanox.com>
> > ---
> >  .../ethernet/mellanox/mlxsw/spectrum_router.c | 27 +++++++++++++++++--
> >  1 file changed, 25 insertions(+), 2 deletions(-)
> > 
> 
> One of the feature requests from the FRR team (and a feature I have
> implemented) is a blackhole nexthop. The idea is that prefixes are
> installed pointing to nexthop id N. That nexthop definition can be
> atomically updated to go between a device / gateway and a blackhole.
> 
> 
>  [ prefix ] --> [ nhid 1 ] --> [ dev1 / gateway1 ]
> 
> 
>  [ prefix ] --> [ nhid 1 ] --> [ blackhole ]
> 
> 
>  [ prefix ] --> [ nhid 1 ] --> [ dev2 / gateway2 ]
> 
> Do you see this working ok with mlxsw without having to update the
> prefix entries (which can be numerous) directly?

Yes. This patch configures the route itself to drop packets, but we can
instead configure it as a remote route and configure the adjacency entry
to drop packets.

If you later want to change X routes using this blackhole nexthop to a
different one, then create the new one and tell the hardware to do the
switch in a single operation. It will basically grep over all configured
routes and do:

s/blackhole_adjacency_index/new_adjacency_index/
s/black_ecmp_size/new_ecmp_size/

See RALEU in drivers/net/ethernet/mellanox/mlxsw/reg.h

I assume that user can't put blackhole and normal nexthops in the same
group?