1 OpenFlow 1.1+ support in Open vSwitch
2 =====================================
4 Open vSwitch support for OpenFlow 1.1, 1.2, and 1.3 is a work in
5 progress. This file describes the work still to be done.
10 OpenFlow version support is not a build-time option. A single build
11 of Open vSwitch must be able to handle all supported versions of
12 OpenFlow. Ideally, even at runtime it should be able to support all
13 protocol versions at the same time on different OpenFlow bridges (and
14 perhaps even on the same bridge).
16 At the same time, it would be a shame to litter the core of the OVS
17 code with lots of ugly code concerned with the details of various
18 OpenFlow protocol versions.
20 The primary approach to compatibility is to abstract most of the
21 details of the differences from the core code, by adding a protocol
22 layer that translates between OF1.x and a slightly higher-level
23 abstract representation. The core of this approach is the many struct
24 ofputil_* structures in lib/ofp-util.h.
26 As a consequence of this approach, OVS cannot use OpenFlow protocol
27 definitions that closely resemble those in the OpenFlow specification,
28 because openflow.h in different versions of the OpenFlow specification
29 defines the same identifier with different values. Instead,
30 openflow-common.h contains definitions that are common to all the
31 specifications and separate protocol version-specific headers contain
32 protocol-specific definitions renamed so as not to conflict,
33 e.g. OFPAT10_ENQUEUE and OFPAT11_ENQUEUE for the OpenFlow 1.0 and 1.1
34 values for OFPAT_ENQUEUE. Generally, in cases of conflict, the
35 protocol layer will define a more abstract OFPUTIL_* or struct
38 Here are the current approaches in a few tricky areas:
40 * Port numbering. OpenFlow 1.0 has 16-bit port numbers and later
41 OpenFlow versions have 32-bit port numbers. For now, OVS
42 support for later protocol versions requires all port numbers to
43 fall into the 16-bit range, translating the reserved OFPP_* port
46 * Actions. OpenFlow 1.0 and later versions have very different
47 ideas of actions. OVS reconciles by translating all the
48 versions' actions (and instructions) to and from a common
49 internal representation.
54 The list of remaining work items for OpenFlow 1.1 is below. It is
57 * Implement Write-Actions instruction.
59 * The new in_phy_port field in OFPT_PACKET_IN needs some kind of
60 implementation. It has a sensible interpretation for tunnels
61 but in general the physical port is not in the datapath for OVS
62 so the value is not necessarily meaningful. We might have to
63 just fix it as the same as in_port.
65 * On OF1.1+ flow_mods, updates by MODIFY are now much better
66 specified. Check that OVS implements the new behavior, fix it
69 * OFPT_TABLE_MOD stats. This is new in OF1.1, so we need to
70 implement it. It should be implemented so that the default OVS
71 behavior does not change.
73 * Document how OVS does packet buffering.
75 * MPLS. Simon Horman maintains a patch series that adds this
76 feature. This is partially merged.
78 * SCTP. Joe Stringer maintains a patch series that adds this
79 feature. It has received review comments that need to be
80 addressed before it is merged.
82 * Match and set double-tagged VLANs (QinQ). This requires kernel
83 work for reasonable performance.
85 * VLANs tagged with 88a8 Ethertype. This requires kernel work for
86 reasonable performance.
93 OpenFlow 1.2 support requires OpenFlow 1.1 as a prerequisite, plus the
94 following additional work. (This is based on the change log at the
95 end of the OF1.2 spec. I didn't compare the specs carefully yet.)
97 * Use new OpenFlow extensible error infrastructure, on OF1.2+
98 only, instead of the OVS-specific extension used until now.
102 * MODIFY and MODIFY_STRICT commands now never insert new flows
103 in the table. We will still need variations that do,
104 though, both to support older OpenFlow protocols and to get
105 sensible behavior for the internal implementation of the
108 * New flag OFPFF_RESET_COUNTS.
110 * New cookie field behavior.
112 * Add ability to delete flow in all tables.
114 * Update DESIGN to describe OF1.2 behavior also.
116 * Add ability to turn off packet buffering with OFPCML_NO_BUFFER.
121 OpenFlow 1.3 support requires OpenFlow 1.2 as a prerequisite, plus the
122 following additional work. (This is based on the change log at the
123 end of the OF1.3 spec, reusing most of the section titles directly. I
124 didn't compare the specs carefully yet.)
126 * Add support for multipart requests.
128 * Add OFPMP_TABLE_FEATURES statistics.
130 * More flexible table miss support.
132 * IPv6 extension header handling support. Fully implementing this
133 requires kernel support. This likely will take some careful and
134 probably time-consuming design work. The actual coding, once
135 that is all done, is probably 2 or 3 days work.
137 * Per-flow meters. Similar to IPv6 extension headers in kernel
138 and design requirements. Might be politically difficult to add
139 directly to the kernel module, since its functionality overlaps
140 with tc. Ideally, therefore, we could implement these somehow
141 with tc, but I haven't investigated whether that makes sense.
143 * Per-connection event filtering. OF1.3 adopted Open vSwitch's
144 existing design for this feature so implementation should be
147 * Auxiliary connections. These are optional, so a minimal
148 implementation would not need them. An implementation in
149 generic code might be a week's worth of work. The value of an
150 implementation in generic code is questionable, though, since
151 much of the benefit of axuiliary connections is supposed to be
152 to take advantage of hardware support. (We could make the
153 kernel module somehow send packets across the auxiliary
154 connections directly, for some kind of "hardware" support, if we
155 judged it useful enough.)
157 * MPLS BoS matching. (Included in Simon's MPLS series?)
159 * Provider Backbone Bridge tagging. I don't plan to implement
160 this (but we'd accept an implementation).
162 * Rework tag order. I'm not sure whether we need to do anything
165 * Duration for stats.
167 * On-demand flow counters. I think this might be a real
168 optimization in some cases for the software switch.
173 If you plan to contribute code for a feature, please let everyone know
174 on ovs-dev before you start work. This will help avoid duplicating
177 Please consider the following:
179 * Testing. Please test your code.
181 * Unit tests. Please consider writing some. The tests directory
182 has many examples that you can use as a starting point.
184 * ovs-ofctl. If you add a feature that is useful for some
185 ovs-ofctl command then you should add support for it there.
187 * Documentation. If you add a user-visible feature, then you
188 should document it in the appropriate manpage and mention it in
191 * Coding style (see the CodingStyle file at the top of the source
194 * The patch submission guidelines (see SubmittingPatches). I
195 recommend using "git send-email", which automatically follows a
196 lot of those guidelines.
201 Please report problems to bugs@openvswitch.org.