Merge pull request #1 from p4lang/bmv2-samples

Bmv2 samples

examples/.gitignore

@@ -0,0 +1,10 @@
+# Python byte code
+*.pyc
+
+# Emacs
+*~
+
+# Compiled JSON
+*.json
+
+*.pcap

examples/README.md

@@ -0,0 +1,59 @@
+# P4 Code Samples
+
+## Introduction
+
+This directory includes P4 code samples, meant to be run on bmv2, which
+illustrates several more advanced features of P4. The following samples are
+included:
+
+- `copy_to_cpu`: how to use the `clone_ingress_to_egress primitive` to clone the
+  packet, encapsulate it and send it to a special port.
+- `meter`: how to use indirect meters in P4.
+- `TLV_parsing`: how to parse IPv4 options
+
+All examples are orgranized the same way, with a `p4src` directory containing
+the P4 source code, and a `README` file describing the P4 program and explaining
+how to run a quick demonstration.
+
+## Obtaining required software
+
+To complete the exercises, you will need to clone 2 p4lang Github repositories
+and install their dependencies. To clonde the repositories:
+
+- `git clone https://github.com/p4lang/behavioral-model.git bmv2`
+- `git clone https://github.com/p4lang/p4c-bm.git p4c-bmv2`
+
+The first repository ([bmv2](https://github.com/p4lang/behavioral-model)) is the
+second version of the behavioral model. It is a C++ software switch that will
+behave according to your P4 program. The second repository
+([p4c-bmv2](https://github.com/p4lang/p4c-bm)) is the compiler for the
+behavioral model: it takes P4 program and output a JSON file which can be loaded
+by the behavioral model.
+
+Each of these repositories come with dependencies. `p4c-bmv2` is a Python
+repository and installing the required Python dependencies is very easy to do
+using `pip`: `sudo pip install -r requirements.txt`.
+
+`bmv2` is a C++ repository and has more external dependencies. They are listed
+in the
+[README](https://github.com/p4lang/behavioral-model/blob/master/README.md). If
+you are running Ubuntu 14.04+, the dependencies should be easy to install (you
+can use the `install_deps.sh` script that comes with `bmv2`). Do not forget to
+build the code once all the dependencies have been installed:
+
+- `./autogen.sh`
+- `./configure`
+- `make`
+
+## Before starting the exercises
+
+You need to tell us where you cloned the `bmv2` and `p4c-bm` repositories
+:). Please update the values of the shell variables `BMV2_PATH` and
+`P4C_BM_PATH` in the `env.sh` file - located in this directory. Note that if you
+cloned both repositories in the same directory as this one (`tutorials`), you
+will not need to change the value of the variables.
+
+You will also need to run the `veth_setup.sh` script included in this directory
+as `sudo` to setup the veth interfaces needed by the switch.
+
+That's all :)

examples/TLV_parsing/README.md

@@ -0,0 +1,62 @@
+# TLV parsing of IPv4 options
+
+## Description
+
+This program illustrates how to parse IPv4 options with bmv2. There is a very
+easy way to parse IPv4 options in P4 using a single variable length
+field. However, this means that options are not parsed individually, but
+together, as one block. All the options are parsed to a single field, which is
+fine for many use cases but can be insufficient in some. In this example, we use
+TLV parsing to parse all options separately to their own header instance.
+
+The program is quite straightforward. The following IPv4 options are supported:
+
+- end of list
+- no-op
+- security
+- timestamp
+
+There is one important caveat: when compiling the P4 program, a strict ordering
+of all packet headers has to be known. This is usually done by inspecting the
+parse graph and running a topological sorting algorithm on it. However this
+algorithm will not work if there exists loops in the header graph, as is the
+case with TLV parsing. There is not yet an official way of enforcing your own
+header ordeing in the P4 program, so we had to bypass this restriction by using
+a `@pragma`, as you can see in the code:
+
+    @pragma header_ordering ethernet ipv4_base ipv4_option_security ipv4_option_NOP ipv4_option_timestamp ipv4_option_EOL
+
+This `@pragma` instruction will be interpreted by the P4 -> bmv2 compiler.
+
+This order is used by the deparser, when sending a packet out of the egress
+port, which means that the option layout for the outgoing packet may not be the
+same as for the incoming packet.
+
+The table `format_options` makes sure that the IPv4 header is formatted
+correctly in the outgoing packet.
+
+Note that the P4 program assumes the incoming packet is correctly formatted. We
+do not perform any sanity checking because *parser execptions* are not yet
+supported by bmv2.
+
+So in a nutshell, all this P4 program does is:
+
+1. parse the IPv4 options for the incoming packet
+2. re-serialize the packet again, with a potentially different order for options
+
+### Running the demo
+
+We provide a small demo to let you test the program. It consists of the
+following scripts:
+- [run_switch.sh] (run_switch.sh): compile the P4 program and starts the switch,
+  also configures the data plane by running the CLI [commands] (commands.txt).
+- [send_one.py] (send_one.py): send an IPv4 packet with options
+
+To run the demo:
+- start the switch and configure the tables: `sudo ./run_switch.sh`.
+- run the Python script: `sudo python send_one.py`.
+
+Then inspect the `pcap` file for port 0 of the switch (`veth0.pcap`) with
+Wireshark. You will observe that the order of the IPv4 options has changed but
+that the outgoing packet contains all the options of the incoming packet and is
+perfectly valid (with a correct checksum).

examples/TLV_parsing/commands.txt

@@ -0,0 +1,4 @@
+table_set_default format_options _nop
+table_add format_options format_options_security 1 0 =>
+table_add format_options format_options_timestamp 0 1 =>
+table_add format_options format_options_both 1 1 =>

examples/TLV_parsing/p4src/TLV_parsing.p4

@@ -0,0 +1,243 @@
+/*
+Copyright 2013-present Barefoot Networks, Inc. 
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+header_type ethernet_t {
+    fields {
+        dstAddr : 48;
+        srcAddr : 48;
+        etherType : 16;
+    }
+}
+
+header_type ipv4_base_t {
+    fields {
+        version : 4; 
+        ihl : 4;
+        diffserv : 8;
+        totalLen : 16;
+        identification : 16;
+        flags : 3;
+        fragOffset : 13;
+        ttl : 8;
+        protocol : 8;
+        hdrChecksum : 16;
+        srcAddr : 32;
+        dstAddr : 32;
+    }
+}
+
+// End of Option List
+#define IPV4_OPTION_EOL_VALUE 0x00
+header_type ipv4_option_EOL_t {
+    fields {
+        value : 8;
+    }
+}
+
+// No operation
+#define IPV4_OPTION_NOP_VALUE 0x01
+header_type ipv4_option_NOP_t {
+    fields {
+        value : 8;
+    }
+}
+
+#define IPV4_OPTION_SECURITY_VALUE 0x82
+header_type ipv4_option_security_t {
+    fields {
+        value : 8;
+        len : 8;
+        security : 72;
+    }
+}
+
+#define IPV4_OPTION_TIMESTAMP_VALUE 0x44
+header_type ipv4_option_timestamp_t {
+    fields {
+        value : 8;
+        len : 8;
+        data : *;
+    }
+    length : len;
+    max_length : 40;
+}
+
+header_type intrinsic_metadata_t {
+    fields {
+        mcast_grp : 4;
+        egress_rid : 4;
+        mcast_hash : 16;
+        lf_field_list: 32;
+    }
+}
+
+header_type my_metadata_t {
+    fields {
+        parse_ipv4_counter : 8;
+    }
+}
+
+header ethernet_t ethernet;
+header ipv4_base_t ipv4_base;
+header ipv4_option_EOL_t ipv4_option_EOL[3];
+header ipv4_option_NOP_t ipv4_option_NOP[3];
+header ipv4_option_security_t ipv4_option_security;
+header ipv4_option_timestamp_t ipv4_option_timestamp;
+metadata intrinsic_metadata_t intrinsic_metadata;
+metadata my_metadata_t my_metadata;
+
+@pragma header_ordering ethernet ipv4_base ipv4_option_security ipv4_option_NOP ipv4_option_timestamp ipv4_option_EOL
+parser start {
+    return parse_ethernet;
+}
+
+#define ETHERTYPE_IPV4 0x0800
+parser parse_ethernet {
+    extract(ethernet);
+    return select(ethernet.etherType) {
+        ETHERTYPE_IPV4 : parse_ipv4;
+        default: ingress;   
+    }
+}
+
+parser parse_ipv4 {
+    extract(ipv4_base);
+    set_metadata(my_metadata.parse_ipv4_counter, ipv4_base.ihl * 4 - 20);
+    return select(ipv4_base.ihl) {
+        0x05 : ingress;
+        default : parse_ipv4_options;
+    }
+}
+
+parser parse_ipv4_options {
+    // match on byte counter and option value
+    return select(my_metadata.parse_ipv4_counter, current(0, 8)) {
+        0x0000 mask 0xff00 : ingress;
+        0x0000 mask 0x00ff : parse_ipv4_option_EOL;
+        0x0001 mask 0x00ff : parse_ipv4_option_NOP;
+        0x0082 mask 0x00ff : parse_ipv4_option_security;
+        0x0044 mask 0x00ff : parse_ipv4_option_timestamp;
+    }
+}
+
+parser parse_ipv4_option_EOL {
+    extract(ipv4_option_EOL[next]);
+    set_metadata(my_metadata.parse_ipv4_counter,
+                 my_metadata.parse_ipv4_counter - 1);
+    return parse_ipv4_options;
+}
+
+parser parse_ipv4_option_NOP {
+    extract(ipv4_option_NOP[next]);
+    set_metadata(my_metadata.parse_ipv4_counter,
+                 my_metadata.parse_ipv4_counter - 1);
+    return parse_ipv4_options;
+}
+
+parser parse_ipv4_option_security {
+    extract(ipv4_option_security);
+    // security option must have length 11 bytes
+    set_metadata(my_metadata.parse_ipv4_counter,
+                 my_metadata.parse_ipv4_counter - 11);
+    return parse_ipv4_options;
+}
+
+parser parse_ipv4_option_timestamp {
+    extract(ipv4_option_timestamp);
+    set_metadata(my_metadata.parse_ipv4_counter,
+                 my_metadata.parse_ipv4_counter - ipv4_option_timestamp.len);
+    return parse_ipv4_options;
+}
+
+field_list ipv4_checksum_list {
+        ipv4_base.version;
+        ipv4_base.ihl;
+        ipv4_base.diffserv;
+        ipv4_base.totalLen;
+        ipv4_base.identification;
+        ipv4_base.flags;
+        ipv4_base.fragOffset;
+        ipv4_base.ttl;
+        ipv4_base.protocol;
+        ipv4_base.srcAddr;
+        ipv4_base.dstAddr;
+        ipv4_option_security;
+        ipv4_option_NOP[0];
+        ipv4_option_timestamp;        
+}
+
+field_list_calculation ipv4_checksum {
+    input {
+        ipv4_checksum_list;
+    }
+    algorithm : csum16;
+    output_width : 16;
+}
+
+calculated_field ipv4_base.hdrChecksum  {
+    update ipv4_checksum;
+}
+
+action _drop() {
+    drop();
+}
+
+action _nop() {
+}
+
+control ingress {
+
+}
+
+action format_options_security() {
+    pop(ipv4_option_NOP, 3);
+    pop(ipv4_option_EOL, 3);
+    push(ipv4_option_EOL, 1);
+    modify_field(ipv4_base.ihl, 8);
+}
+
+action format_options_timestamp() {
+    pop(ipv4_option_NOP, 3);
+    pop(ipv4_option_EOL, 3);
+    // timestamp option is word-aligned so no need for NOP or EOL
+    modify_field(ipv4_base.ihl, 5 + (ipv4_option_timestamp.len >> 3));
+}
+
+action format_options_both() {
+    pop(ipv4_option_NOP, 3);
+    pop(ipv4_option_EOL, 3);
+    push(ipv4_option_NOP, 1);
+    modify_field(ipv4_option_NOP[0].value, IPV4_OPTION_NOP_VALUE);
+    modify_field(ipv4_base.ihl, 8 + (ipv4_option_timestamp.len >> 2));
+}
+
+table format_options {
+    reads {
+        ipv4_option_security : valid;
+        ipv4_option_timestamp : valid;
+    }
+    actions {
+        format_options_security;
+        format_options_timestamp;
+        format_options_both;
+        _nop;
+    }
+    size : 4;
+}
+
+control egress {
+    apply(format_options);
+}

examples/TLV_parsing/run_switch.sh

@@ -0,0 +1,41 @@
+#!/bin/bash
+
+# Copyright 2013-present Barefoot Networks, Inc. 
+# 
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+# 
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+THIS_DIR=$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )
+
+source $THIS_DIR/../env.sh
+
+P4C_BM_SCRIPT=$P4C_BM_PATH/p4c_bm/__main__.py
+
+SWITCH_PATH=$BMV2_PATH/targets/simple_switch/simple_switch
+
+CLI_PATH=$BMV2_PATH/targets/simple_switch/sswitch_CLI
+
+# Probably not very elegant but it works nice here: we enable interactive mode
+# to be able to use fg. We start the switch in the background, sleep for 2
+# minutes to give it time to start, then add the entries and put the switch
+# process back in the foreground
+set -m
+$P4C_BM_SCRIPT p4src/TLV_parsing.p4 --json TLV_parsing.json
+sudo echo "sudo" > /dev/null
+sudo $BMV2_PATH/targets/simple_switch/simple_switch TLV_parsing.json \
+    -i 0@veth0 -i 1@veth2 -i 2@veth4 -i 3@veth6 -i 4@veth8 \
+    --nanolog ipc:///tmp/bm-0-log.ipc \
+    --pcap &
+sleep 2
+$CLI_PATH TLV_parsing.json < commands.txt
+echo "READY!!!"
+fg

examples/TLV_parsing/send_one.py

@@ -0,0 +1,6 @@
+from scapy.all import *
+
+p = Ether() / IP(options=IPOption('\x44\x0c\x05\x00\x01\x02\x03\x04\x05\x06\x07\x08') / IPOption('\x82\x0b\xa1\xa2\xa3\xa4\xa5\xa6\xa7\xa8\xa9')) / IPOption('\x00') / TCP() / "aaaaaaaaaaa"
+# p.show()
+hexdump(p)
+sendp(p, iface = "veth0")

examples/copy_to_cpu/README.md

@@ -0,0 +1,47 @@
+# Copy to CPU
+
+## Description
+
+This program illustrates as simply as possible how to *send packets to CPU*
+(e.g. to a controller).
+
+The P4 program does the following:
+- incoming packets are mirrored to the CPU port using the
+  `clone_ingress_pkt_to_egress` action primitive
+- packets mirrored to CPU are encapsulated with a custom `cpu_header` which
+  includes 2 fields: `device` (1 byte, set to `0`) and `reason` (one byte, set
+  to `0xab`)
+- the original packet is dropped in the egress pipeline
+
+Take a look at the [P4 code] (p4src/copy_to_cpu.p4). The program is very short
+and should be easy to understand.  You will notice that we use a mirror session
+id of `250` in the program. This number is not relevant in itself, but needs to
+be consistent between the P4 program and the runtime application.
+
+### Running the demo
+
+We provide a small demo to let you test the program. It consists of the
+following scripts:
+- [run_switch.sh] (run_switch.sh): compile the P4 program and starts the switch,
+  also configures the data plane by running the CLI [commands] (commands.txt)
+- [receive.py] (receive.py): sniff packets on port 3 (veth6) and print a hexdump
+  of them
+- [send_one.py] (send_one.py): send one simple IPv4 packet on port 0 (veth0)
+
+If you take a look at [commands.txt] (commands.txt), you'll notice the following
+command: `mirroring_add 250 3`. This means that all the cloned packets with
+mirror id `250` will be sent to port `3`, which is our de facto *CPU port*. This
+is the reason why [receive.py] (receive.py) listens for incoming packets on port
+`3`.
+
+To run the demo:
+- start the switch and configure the tables and the mirroring session: `sudo
+  ./run_switch.sh`
+- start the CPU port listener: `sudo python receive.py`
+- send packets with `sudo python send_one.py`. Every time you send one packet,
+  it should be displayed by the listener, encapsulated with our CPU header.
+
+This is a very simple example obviously. Feel free to build upon it. For
+example, instead of dropping the original packet, you could try to broadcast it
+out of every non-ingress port to have a working L2 switch. You could also build
+a L2 controller which receives CPU packets and modifies tables appropriately.

examples/copy_to_cpu/commands.txt

@@ -0,0 +1,4 @@
+table_set_default copy_to_cpu do_copy_to_cpu
+table_set_default redirect _drop
+table_add redirect do_cpu_encap 1 =>
+mirroring_add 250 3

examples/copy_to_cpu/p4src/copy_to_cpu.p4

@@ -0,0 +1,103 @@
+/*
+Copyright 2013-present Barefoot Networks, Inc. 
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+header_type ethernet_t {
+    fields {
+        dstAddr : 48;
+        srcAddr : 48;
+        etherType : 16;
+    }
+}
+
+header_type intrinsic_metadata_t {
+    fields {
+        mcast_grp : 4;
+        egress_rid : 4;
+        mcast_hash : 16;
+        lf_field_list: 32;
+    }
+}
+
+header_type cpu_header_t {
+    fields {
+        device: 8;
+        reason: 8;
+    }
+}
+
+header cpu_header_t cpu_header;
+
+parser start {
+    return select(current(0, 64)) {
+        0 : parse_cpu_header;
+        default: parse_ethernet;
+    }
+}
+
+header ethernet_t ethernet;
+metadata intrinsic_metadata_t intrinsic_metadata;
+
+parser parse_ethernet {
+    extract(ethernet);
+    return ingress;
+}
+
+parser parse_cpu_header {
+    extract(cpu_header);
+    return parse_ethernet;
+}
+
+action _drop() {
+    drop();
+}
+
+action _nop() {
+}
+
+#define CPU_MIRROR_SESSION_ID                  250
+
+field_list copy_to_cpu_fields {
+    standard_metadata;
+}
+
+action do_copy_to_cpu() {
+    clone_ingress_pkt_to_egress(CPU_MIRROR_SESSION_ID, copy_to_cpu_fields);
+}
+
+table copy_to_cpu {
+    actions {do_copy_to_cpu;}
+    size : 1;
+}
+
+control ingress {
+    apply(copy_to_cpu);
+}
+
+action do_cpu_encap() {
+    add_header(cpu_header);
+    modify_field(cpu_header.device, 0);
+    modify_field(cpu_header.reason, 0xab);
+}
+
+table redirect {
+    reads { standard_metadata.instance_type : exact; }
+    actions { _drop; do_cpu_encap; }
+    size : 16;
+}
+
+control egress {
+    apply(redirect);
+}

examples/copy_to_cpu/receive.py

@@ -0,0 +1,3 @@
+from scapy.all import *
+
+sniff(iface = "veth6", prn = lambda x: hexdump(x))

examples/copy_to_cpu/run_switch.sh

@@ -0,0 +1,41 @@
+#!/bin/bash
+
+# Copyright 2013-present Barefoot Networks, Inc. 
+# 
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+# 
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+THIS_DIR=$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )
+
+source $THIS_DIR/../env.sh
+
+P4C_BM_SCRIPT=$P4C_BM_PATH/p4c_bm/__main__.py
+
+SWITCH_PATH=$BMV2_PATH/targets/simple_switch/simple_switch
+
+CLI_PATH=$BMV2_PATH/targets/simple_switch/sswitch_CLI
+
+# Probably not very elegant but it works nice here: we enable interactive mode
+# to be able to use fg. We start the switch in the background, sleep for 2
+# minutes to give it time to start, then add the entries and put the switch
+# process back in the foreground
+set -m
+$P4C_BM_SCRIPT p4src/copy_to_cpu.p4 --json copy_to_cpu.json
+sudo echo "sudo" > /dev/null
+sudo $BMV2_PATH/targets/simple_switch/simple_switch copy_to_cpu.json \
+    -i 0@veth0 -i 1@veth2 -i 2@veth4 -i 3@veth6 -i 4@veth8 \
+    --nanolog ipc:///tmp/bm-0-log.ipc \
+    --pcap &
+sleep 2
+$CLI_PATH copy_to_cpu.json < commands.txt
+echo "READY!!!"
+fg

examples/copy_to_cpu/send_one.py

@@ -0,0 +1,6 @@
+from scapy.all import *
+
+p = Ether(dst="aa:bb:cc:dd:ee:ff") / IP(dst="10.0.1.10") / TCP() / "aaaaaaaaaaaaaaaaaaa"
+# p.show()
+hexdump(p)
+sendp(p, iface = "veth0")

examples/env.sh

@@ -0,0 +1,8 @@
+THIS_DIR=$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )
+
+# ---------------- EDIT THIS ------------------
+BMV2_PATH=$THIS_DIR/../../bmv2
+# e.g. BMV2_PATH=$THIS_DIR/../bmv2
+P4C_BM_PATH=$THIS_DIR/../../p4c-bmv2
+# e.g P4C_BM_PATH=$THIS_DIR/../p4c-bm
+# ---------------- END ------------------

examples/meter/README.md

@@ -0,0 +1,65 @@
+# Meter
+
+## Description
+
+This program illustrates as simply as possible how to use meters in P4 with
+bmv2. bmv2 uses two-rate three-color meters as described [here]
+(https://tools.ietf.org/html/rfc2698).
+
+For each incoming packet the `m_table` table is applied and the appropriate
+meter (based on the packet's source MAC address) is executed. Based on the
+observed traffic rate for this sender and the meter's configuration, executing
+the meter will yield one of 3 values: `0` (*GREEN*), `1` (*YELLOW*) or `2`
+(*RED*). This value will be copied to metadata field `meta.meter_tag`. Note that
+if no meter was associated to the sender's MAC address, the table will be a
+no-op. This table also redirects all packets - with a known source MAC address-
+to port 2 of the switch.
+
+After that, the packet will go through a second table, `m_filter`, which can
+either be a no-op or drop the packet based on how the packet was tagged by the
+meter. If you take a look at the [runtime commands] (commands.txt) we wrote for
+this example, you will see that we configure the table to drop all the packets
+for which the color is not *GREEN* (i.e. all packets for which `meta.meter_tag`
+is not `0`).
+
+The [commands.txt] (commands.txt) file also gives you the meter
+configuration. In this case, the first rate is 0.5 packets per second, with a
+burst size of 1, and the second rate is 10 packets per second, with a burst size
+of 1 also. Feel free to play with the numbers, but these play nicely with the
+demonstration below.
+
+Note that we use an `indirect` meter array, because `direct` ones are not
+supported yet by bmv2.
+
+### Running the demo
+
+We provide a small demo to let you test the program. It consists of the
+following scripts:
+- [run_switch.sh] (run_switch.sh): compile the P4 program and starts the switch,
+  also configures the data plane by running the CLI [commands] (commands.txt).
+- [send_and_receive.py] (send_and_receive.py): send packets periodically on port
+  0 and listen for packets on port 2.
+
+To run the demo:
+- start the switch and configure the tables and the meters: `sudo
+  ./run_switch.sh`.
+- run the Python script: `sudo python send_and_receive.py 1`. As you can see,
+  the script takes one argument, which is the time interval (in seconds) between
+  two consecutive packets.
+
+If you run the script with an interval of one second, you should observe the
+following output:
+
+    Received one
+    Sent one
+    Sent one
+    Received one
+    Sent one
+    Sent one
+    Received one
+    Sent one
+    ...
+
+This is because we send one packet every second, while the first rate of the
+meter is 0.5 packets per second. The P4 program therefore drops on average one
+packet out of two.

examples/meter/commands.txt

@@ -0,0 +1,5 @@
+table_set_default m_table _nop
+table_add m_table m_action aa:aa:aa:aa:aa:aa => 0
+table_set_default m_filter _drop
+table_add m_filter _nop 0 =>
+meter_set_rates my_meter 0.0000005:1 0.00001:1

examples/meter/p4src/meter.p4

@@ -0,0 +1,98 @@
+/*
+Copyright 2013-present Barefoot Networks, Inc. 
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+header_type ethernet_t {
+    fields {
+        dstAddr : 48;
+        srcAddr : 48;
+        etherType : 16;
+    }
+}
+
+header_type intrinsic_metadata_t {
+    fields {
+        mcast_grp : 4;
+        egress_rid : 4;
+        mcast_hash : 16;
+        lf_field_list: 32;
+    }
+}
+
+header_type meta_t {
+    fields {
+        meter_tag : 32;
+    }
+}
+
+metadata meta_t meta;
+
+parser start {
+    return parse_ethernet;
+}
+
+header ethernet_t ethernet;
+metadata intrinsic_metadata_t intrinsic_metadata;
+
+parser parse_ethernet {
+    extract(ethernet);
+    return ingress;
+}
+
+action _drop() {
+    drop();
+}
+
+action _nop() {
+}
+
+meter my_meter {
+    type: packets; // or bytes
+    static: m_table;
+    instance_count: 16384;
+}
+
+action m_action(meter_idx) {
+    execute_meter(my_meter, meter_idx, meta.meter_tag);
+    modify_field(standard_metadata.egress_spec, 1);
+}
+
+table m_table {
+    reads {
+        ethernet.srcAddr : exact;
+    }
+    actions {
+        m_action; _nop;
+    }
+    size : 16384;
+}
+
+table m_filter {
+    reads {
+        meta.meter_tag : exact;
+    }
+    actions {
+        _drop; _nop;
+    }
+    size: 16;
+}
+
+control ingress {
+    apply(m_table);
+    apply(m_filter);
+}
+
+control egress {
+}

examples/meter/run_switch.sh

@@ -0,0 +1,41 @@
+#!/bin/bash
+
+# Copyright 2013-present Barefoot Networks, Inc. 
+# 
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+# 
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+THIS_DIR=$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )
+
+source $THIS_DIR/../env.sh
+
+P4C_BM_SCRIPT=$P4C_BM_PATH/p4c_bm/__main__.py
+
+SWITCH_PATH=$BMV2_PATH/targets/simple_switch/simple_switch
+
+CLI_PATH=$BMV2_PATH/targets/simple_switch/sswitch_CLI
+
+# Probably not very elegant but it works nice here: we enable interactive mode
+# to be able to use fg. We start the switch in the background, sleep for 2
+# minutes to give it time to start, then add the entries and put the switch
+# process back in the foreground
+set -m
+$P4C_BM_SCRIPT p4src/meter.p4 --json meter.json
+sudo echo "sudo" > /dev/null
+sudo $BMV2_PATH/targets/simple_switch/simple_switch meter.json \
+    -i 0@veth0 -i 1@veth2 -i 2@veth4 -i 3@veth6 -i 4@veth8 \
+    --nanolog ipc:///tmp/bm-0-log.ipc \
+    --pcap &
+sleep 2
+$CLI_PATH meter.json < commands.txt
+echo "READY!!!"
+fg

examples/meter/send_and_receive.py

@@ -0,0 +1,44 @@
+from scapy.all import *
+import sys
+import threading
+
+
+big_lock = threading.Lock()
+
+
+class Receiver(threading.Thread):
+    def __init__(self):
+        threading.Thread.__init__(self)
+        self.daemon = True
+
+    def received(self, p):
+        big_lock.acquire()
+        print "Received one"
+        big_lock.release()
+
+    def run(self):
+        sniff(iface="veth2", prn=lambda x: self.received(x))
+
+
+def main():
+    try:
+        packet_int = int(sys.argv[1])
+        print "Sending packet with interval", packet_int
+    except:
+        print "Usage: sudo python send_and_receive.py <packet_int (seconds)>"
+        sys.exit(1)
+
+    Receiver().start()
+
+    p = Ether(src="aa:aa:aa:aa:aa:aa") / IP(dst="10.0.1.10") / TCP() / "aaaaaaaaaaaaaaaaaaa"
+
+    while True:
+        big_lock.acquire()
+        sendp(p, iface="veth0", verbose=0)
+        print "Sent one"
+        big_lock.release()
+        time.sleep(packet_int)
+
+
+if __name__ == '__main__':
+    main()

examples/veth_setup.sh

@@ -0,0 +1,26 @@
+#!/bin/bash
+noOfVeths=18
+if [ $# -eq 1 ]; then 
+    noOfVeths=$1
+fi
+echo "No of Veths is $noOfVeths"
+idx=0
+let "vethpairs=$noOfVeths/2"
+while [ $idx -lt $vethpairs ]
+do 
+    intf0="veth$(($idx*2))"
+    intf1="veth$(($idx*2+1))"
+    idx=$((idx + 1))
+    if ! ip link show $intf0 &> /dev/null; then
+        ip link add name $intf0 type veth peer name $intf1
+        ip link set dev $intf0 up
+        ip link set dev $intf1 up
+        TOE_OPTIONS="rx tx sg tso ufo gso gro lro rxvlan txvlan rxhash"
+        for TOE_OPTION in $TOE_OPTIONS; do
+           /sbin/ethtool --offload $intf0 "$TOE_OPTION" off
+           /sbin/ethtool --offload $intf1 "$TOE_OPTION" off
+        done
+    fi
+    sysctl net.ipv6.conf.$intf0.disable_ipv6=1
+    sysctl net.ipv6.conf.$intf1.disable_ipv6=1
+done