Skip to content
This repository has been archived by the owner on Oct 20, 2022. It is now read-only.
/ oko Public archive

Extend Open vSwitch with BPF programs at runtime

License

Notifications You must be signed in to change notification settings

Orange-OpenSource/oko

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Oko

The original Open vSwitch README is available at README-original.md.

What is Oko?

Oko is an extension of Open vSwitch-DPDK that provides runtime extension with BPF programs. BPF programs act as filters over packets: they are referenced as an additional match field in the OpenFlow tables and cannot write to packets. They can however read and write to persistent maps (array or hash table) to retain information on flows.

Oko was based on Open vSwitch v2.5 (commit b63bf24) and relies on a modified version of the ubpf project to execute BPF programs.

This version of Oko is a research prototype: it almost certainly contains serious bugs and should only be used for experimentation and research purposes.

How to install?

To install Oko, you can follow the usual guidelines to install Open vSwitch-DPDK. No additional dependencies are required.

How to use?

$ ovs-vsctl add-port br0 dpdk0 -- set Interface dpdk0 type=dpdk ofport_request=1
$ ovs-vsctl add-port br0 dpdk1 -- set Interface dpdk1 type=dpdk ofport_request=2
$ ovs-vsctl show
509b64f2-a893-490a-9fd5-7582c29e8b89
    Bridge "br0"
        Port "dpdk0"
            Interface "dpdk0"
                type: dpdk
        Port "dpdk1"
            Interface "dpdk1"
                type: dpdk
$ clang -O2 -target bpf -c examples/bpf/stateless-firewall.c -o /tmp/stateless-firewall.o
$ ovs-ofctl load-filter-prog br0 1 /tmp/stateless-firewall.o
$ ovs-ofctl add-flow br0 priority=1,in_port=1,filter_prog=1,actions=output:2
$ ovs-ofctl add-flow br0 priority=1,in_port=2,actions=output:1
$ ovs-ofctl dump-flows br0
NXST_FLOW reply (xid=0x4):
 cookie=0x0, duration=103.730s, table=0, n_packets=0, n_bytes=0, idle_age=103, priority=1,in_port=1,filter_prog=1 actions=output:2
 cookie=0x0, duration=103.842s, table=0, n_packets=0, n_bytes=0, idle_age=103, in_port=2,actions=output:1
# Drop (value=1) packets destined to IP 172.16.0.14 through map 0 of filter_prog 1.
$ ovs-ofctl update-map br0 1 0 key 14 0 16 172 value 1 0 0 0
# Show the content of map 0 of filter_prog 1. Use 'hex' flag to print in hexadecimal format.
$ ovs-ofctl dump-map br0 1 0 hex
NXT_DUMP_MAP_REPLY (xid=0x4):
The map contains 1 element(s)
Key: 
0e 00 10 ac
Value: 
01 00 00 00

License

Except for the lib/bpf/lookup3.c file in the public domain, all new files introduced by Oko compared to Open vSwitch are licensed under Apache 2.0. Modified files from both Open vSwitch and ubpf are also licensed under their original license, Apache 2.0.

Modifications to source codes:

For compliance with the Apache 2.0 license, the following lists our modifications to the source codes of ubpf and Open vSwitch.

ubpf

  • Support for maps allocation (ELF parsing, memory allocation, and map relocation).
  • Support for Array, Hash table, Bloom filter, and Count-Min sketch maps.
  • Increase the stack size of the virtual machine to 512.
  • Fix warnings related to pointer arithmetic.
  • Support for LDIND* and LDABS* bytecode instructions.
  • BPF helpers to compute a hash value and retrieve the current time.
  • BPF verifier for register types and variable-sized loops.

Open vSwitch

  • New filter_prog match field in OpenFlow table.
  • New LOAD_FILTER_PROG OpenFlow message to send a BPF program to load to the switch, as an ELF file.
  • New UPDATE_MAP OpenFlow message to write entry (key-value pair) to the BPF map of the given BPF program.
  • New DUMP_MAP OpenFlow message to dump the BPF map of the given BPF program.
  • New SEND_MAPS action and message to send the content of maps to the controller.
  • New filter program chain structure in the datapath to cache a succession of BPF programs.

Contacts

Paul Chaignon <[email protected]>

Tomasz Osiński <[email protected]>

Mateusz Kossakowski <[email protected]>