In this exercise, we will be using P4Runtime to send flow entries to the
switch instead of using the switch's CLI. We will be building on the same P4
program that you used in the basic_tunnel exercise. The
P4 program has been renamed to advanced_tunnel.p4
and has been augmented
with two counters (ingressTunnelCounter
, egressTunnelCounter
) and
two new actions (myTunnel_ingress
, myTunnel_egress
).
You will use the starter program, mycontroller.py
, and a few helper
libraries in the p4runtime_lib
directory to create the table entries
necessary to tunnel traffic between host 1 and 2.
Spoiler alert: There is a reference solution in the
solution
sub-directory. Feel free to compare your implementation to the reference.
The starter code for this assignment is in a file called mycontroller.py
,
and it will install only some of the rules that you need to tunnel traffic between
two hosts.
Let's first compile the new P4 program, start the network, use mycontroller.py
to install a few rules, and look at the ingressTunnelCounter
to see that things
are working as expected.
-
In your shell, run:
make
This will:
- compile
advanced_tunnel.p4
, - start a Mininet instance with three switches (
s1
,s2
,s3
) configured in a triangle, each connected to one host (h1
,h2
,h3
), and - assign IPs of
10.0.1.1
,10.0.2.2
,10.0.3.3
to the respective hosts.
- compile
-
You should now see a Mininet command prompt. Start a ping between h1 and h2:
mininet> h1 ping h2
Because there are no rules on the switches, you should not receive any replies yet. You should leave the ping running in this shell.
-
Open another shell and run the starter code:
cd ~/tutorials/exercises/p4runtime ./mycontroller.py
This will install the
advanced_tunnel.p4
program on the switches and push the tunnel ingress rules. The program prints the tunnel ingress and egress counters every 2 seconds. You should see the ingress tunnel counter for s1 increasing:s1 ingressTunnelCounter 100: 2 packets
The other counters should remain at zero.
-
Press
Ctrl-C
to the second shell to stopmycontroller.py
Each switch is currently mapping traffic into tunnels based on the destination IP address. Your job is to write the rules that forward the traffic between the switches based on the tunnel ID.
If you see the following error message when running mycontroller.py
, then
the gRPC server is not running on one or more switches.
p4@p4:~/tutorials/exercises/p4runtime$ ./mycontroller.py
...
grpc._channel._Rendezvous: <_Rendezvous of RPC that terminated with (StatusCode.UNAVAILABLE, Connect Failed)>
You can check to see which of gRPC ports are listening on the machine by running:
sudo netstat -lpnt
The easiest solution is to enter Ctrl-D
or exit
in the mininet>
prompt,
and re-run make
.
A P4 program defines a packet-processing pipeline, but the rules
within each table are inserted by the control plane. In this case,
mycontroller.py
implements our control plane, instead of installing static
table entries like we have in the previous exercises.
Important: A P4 program also defines the interface between the
switch pipeline and control plane. This interface is defined in the
advanced_tunnel.p4info
file. The table entries that you build in mycontroller.py
refer to specific tables, keys, and actions by name, and we use a P4Info helper
to convert the names into the IDs that are required for P4Runtime. Any changes
in the P4 program that add or rename tables, keys, or actions will need to be
reflected in your table entries.
The mycontroller.py
file is a basic controller plane that does the following:
- Establishes a gRPC connection to the switches for the P4Runtime service.
- Pushes the P4 program to each switch.
- Writes tunnel ingress and tunnel egress rules for two tunnels between h1 and h2.
- Reads tunnel ingress and egress counters every 2 seconds.
It also contains comments marked with TODO
which indicate the functionality
that you need to implement.
Your job will be to write the tunnel transit rule in the writeTunnelRules
function
that will match on tunnel ID and forward packets to the next hop.
In this exercise, you will be interacting with some of the classes and methods in
the p4runtime_lib
directory. Here is a summary of each of the files in the directory:
helper.py
- Contains the
P4InfoHelper
class which is used to parse thep4info
files. - Provides translation methods from entity name to and from ID number.
- Builds P4 program-dependent sections of P4Runtime table entries.
- Contains the
switch.py
- Contains the
SwitchConnection
class which grabs the gRPC client stub, and establishes connections to the switches. - Provides helper methods that construct the P4Runtime protocol buffer messages and makes the P4Runtime gRPC service calls.
- Contains the
bmv2.py
- Contains
Bmv2SwitchConnection
which extendsSwitchConnections
and provides the BMv2-specific device payload to load the P4 program.
- Contains
convert.py
- Provides convenience methods to encode and decode from friendly strings and numbers to the byte strings required for the protocol buffer messages.
- Used by
helper.py
Follow the instructions from Step 1. If your Mininet network is still running, you will just need to run the following in your second shell:
./my_controller.py
You should start to see ICMP replies in your Mininet prompt, and you should start to see the values for all counters start to increment.
You might notice that the rules that are printed by mycontroller.py
contain the entity
IDs rather than the table names. You can use the P4Info helper to translate these IDs
into entry names.
Also, you may want to think about the following:
-
What assumptions about the topology are baked into your implementation? How would you need to change it for a more realistic network?
-
Why are the byte counters different between the ingress and egress counters?
-
What is the TTL in the ICMP replies? Why is it the value that it is? Hint: The default TTL is 64 for packets sent by the hosts.
If you are interested, you can find the protocol buffer and gRPC definitions here:
If the Mininet shell crashes, it may leave a Mininet instance running in the background. Use the following command to clean up:
make clean
To run the reference solution, you should run the following command from the
~/tutorials/P4D2_2017_Fall/exercises/p4runtime
directory:
solution/my_controller.py
Congratulations, your implementation works! Move onto the next assignment mri!
The documentation for P4_16 and P4Runtime is available here
All excercises in this repository use the v1model architecture, the documentation for which is available at: