VOL-1180: Enable ponsim to send packet-outs on NNI port.
This required adding the out_port metatdata to the message that
VOLTHA uses to send packets to ponsim, so that VOLTHA can tell
ponsim which port to send the packet out.
Also refactored packet-outs in ponsim so that they don't get run
through the forwarding pipeline, they just get sent out the port.
This is what happens in a normal openflow device.
Change-Id: Ic40ea730f061d82d2ecee33fb1a1f371d0ba73ef
diff --git a/voltha/adapters/ponsim_olt/ponsim_olt.py b/voltha/adapters/ponsim_olt/ponsim_olt.py
index b84209c..5ef3fc3 100644
--- a/voltha/adapters/ponsim_olt/ponsim_olt.py
+++ b/voltha/adapters/ponsim_olt/ponsim_olt.py
@@ -427,7 +427,7 @@
self.alarms = AdapterAlarms(self.adapter, device)
nni_port = Port(
- port_no=2,
+ port_no=info.nni_port,
label='NNI facing Ethernet port',
type=Port.ETHERNET_NNI,
admin_state=AdminState.ENABLED,
@@ -450,7 +450,7 @@
hw_desc='simualted pon',
sw_desc='simualted pon',
# serial_num=uuid4().hex,
- serial_num="9b7cfd85441b407c87ee3261df7d4818",
+ serial_num=device.serial_number,
dp_desc='n/a'
),
switch_features=ofp_switch_features(
@@ -703,17 +703,22 @@
self.log.info('sending-packet-out', egress_port=egress_port,
msg=hexify(msg))
pkt = Ether(msg)
- out_pkt = (
- Ether(src=pkt.src, dst=pkt.dst) /
- Dot1Q(vlan=4000) /
- Dot1Q(vlan=egress_port, type=pkt.type) /
- pkt.payload
- )
+ out_pkt = pkt
+ if egress_port != self.nni_port.port_no:
+ # don't do the vlan manipulation for the NNI port, vlans are already correct
+ out_pkt = (
+ Ether(src=pkt.src, dst=pkt.dst) /
+ Dot1Q(vlan=egress_port, type=pkt.type) /
+ pkt.payload
+ )
+
+ # TODO need better way of mapping logical ports to PON ports
+ out_port = self.nni_port.port_no if egress_port == self.nni_port.port_no else 1
if self.ponsim_comm == 'grpc':
# send over grpc stream
stub = ponsim_pb2.PonSimStub(self.get_channel())
- frame = PonSimFrame(id=self.device_id, payload=str(out_pkt))
+ frame = PonSimFrame(id=self.device_id, payload=str(out_pkt), out_port=out_port)
stub.SendFrame(frame)
else:
# send over frameio