FAQ.md

   1 Frequently Asked Questions
   2 ==========================
   3
   4 Open vSwitch <http://openvswitch.org>
   5
   6 General
   7 -------
   8
   9 ### Q: What is Open vSwitch?
  10
  11 A: Open vSwitch is a production quality open source software switch
  12    designed to be used as a vswitch in virtualized server
  13    environments.  A vswitch forwards traffic between different VMs on
  14    the same physical host and also forwards traffic between VMs and
  15    the physical network.  Open vSwitch supports standard management
  16    interfaces (e.g. sFlow, NetFlow, IPFIX, RSPAN, CLI), and is open to
  17    programmatic extension and control using OpenFlow and the OVSDB
  18    management protocol.
  19
  20    Open vSwitch as designed to be compatible with modern switching
  21    chipsets.  This means that it can be ported to existing high-fanout
  22    switches allowing the same flexible control of the physical
  23    infrastructure as the virtual infrastructure.  It also means that
  24    Open vSwitch will be able to take advantage of on-NIC switching
  25    chipsets as their functionality matures.
  26
  27 ### Q: What virtualization platforms can use Open vSwitch?
  28
  29 A: Open vSwitch can currently run on any Linux-based virtualization
  30    platform (kernel 2.6.32 and newer), including: KVM, VirtualBox, Xen,
  31    Xen Cloud Platform, XenServer. As of Linux 3.3 it is part of the
  32    mainline kernel.  The bulk of the code is written in platform-
  33    independent C and is easily ported to other environments.  We welcome
  34    inquires about integrating Open vSwitch with other virtualization
  35    platforms.
  36
  37 ### Q: How can I try Open vSwitch?
  38
  39 A: The Open vSwitch source code can be built on a Linux system.  You can
  40    build and experiment with Open vSwitch on any Linux machine.
  41    Packages for various Linux distributions are available on many
  42    platforms, including: Debian, Ubuntu, Fedora.
  43
  44    You may also download and run a virtualization platform that already
  45    has Open vSwitch integrated.  For example, download a recent ISO for
  46    XenServer or Xen Cloud Platform.  Be aware that the version
  47    integrated with a particular platform may not be the most recent Open
  48    vSwitch release.
  49
  50 ### Q: Does Open vSwitch only work on Linux?
  51
  52 A: No, Open vSwitch has been ported to a number of different operating
  53    systems and hardware platforms.  Most of the development work occurs
  54    on Linux, but the code should be portable to any POSIX system.  We've
  55    seen Open vSwitch ported to a number of different platforms,
  56    including FreeBSD, Windows, and even non-POSIX embedded systems.
  57
  58    By definition, the Open vSwitch Linux kernel module only works on
  59    Linux and will provide the highest performance.  However, a userspace
  60    datapath is available that should be very portable.
  61
  62 ### Q: What's involved with porting Open vSwitch to a new platform or switching ASIC?
  63
  64 A: The [PORTING.md] document describes how one would go about
  65    porting Open vSwitch to a new operating system or hardware platform.
  66
  67 ### Q: Why would I use Open vSwitch instead of the Linux bridge?
  68
  69 A: Open vSwitch is specially designed to make it easier to manage VM
  70    network configuration and monitor state spread across many physical
  71    hosts in dynamic virtualized environments.  Please see
  72    [WHY-OVS.md] for a more detailed description of how Open vSwitch
  73    relates to the Linux Bridge.
  74
  75 ### Q: How is Open vSwitch related to distributed virtual switches like the VMware vNetwork distributed switch or the Cisco Nexus 1000V?
  76
  77 A: Distributed vswitch applications (e.g., VMware vNetwork distributed
  78    switch, Cisco Nexus 1000V) provide a centralized way to configure and
  79    monitor the network state of VMs that are spread across many physical
  80    hosts.  Open vSwitch is not a distributed vswitch itself, rather it
  81    runs on each physical host and supports remote management in a way
  82    that makes it easier for developers of virtualization/cloud
  83    management platforms to offer distributed vswitch capabilities.
  84
  85    To aid in distribution, Open vSwitch provides two open protocols that
  86    are specially designed for remote management in virtualized network
  87    environments: OpenFlow, which exposes flow-based forwarding state,
  88    and the OVSDB management protocol, which exposes switch port state.
  89    In addition to the switch implementation itself, Open vSwitch
  90    includes tools (ovs-ofctl, ovs-vsctl) that developers can script and
  91    extend to provide distributed vswitch capabilities that are closely
  92    integrated with their virtualization management platform.
  93
  94 ### Q: Why doesn't Open vSwitch support distribution?
  95
  96 A: Open vSwitch is intended to be a useful component for building
  97    flexible network infrastructure. There are many different approaches
  98    to distribution which balance trade-offs between simplicity,
  99    scalability, hardware compatibility, convergence times, logical
 100    forwarding model, etc. The goal of Open vSwitch is to be able to
 101    support all as a primitive building block rather than choose a
 102    particular point in the distributed design space.
 103
 104 ### Q: How can I contribute to the Open vSwitch Community?
 105
 106 A: You can start by joining the mailing lists and helping to answer
 107    questions.  You can also suggest improvements to documentation.  If
 108    you have a feature or bug you would like to work on, send a mail to
 109    one of the mailing lists:
 110
 111    http://openvswitch.org/mlists/
 112
 113
 114 Releases
 115 --------
 116
 117 ### Q: What does it mean for an Open vSwitch release to be LTS (long-term support)?
 118
 119 A: All official releases have been through a comprehensive testing
 120    process and are suitable for production use.  Planned releases will
 121    occur several times a year.  If a significant bug is identified in an
 122    LTS release, we will provide an updated release that includes the
 123    fix.  Releases that are not LTS may not be fixed and may just be
 124    supplanted by the next major release.  The current LTS release is
 125    1.9.x.
 126
 127 ### Q: What Linux kernel versions does each Open vSwitch release work with?
 128
 129 A: The following table lists the Linux kernel versions against which the
 130    given versions of the Open vSwitch kernel module will successfully
 131    build.  The Linux kernel versions are upstream kernel versions, so
 132    Linux kernels modified from the upstream sources may not build in
 133    some cases even if they are based on a supported version.  This is
 134    most notably true of Red Hat Enterprise Linux (RHEL) kernels, which
 135    are extensively modified from upstream.
 136
 137 | Open vSwitch | Linux kernel
 138 |:------------:|:-------------:
 139 |    1.4.x     | 2.6.18 to 3.2
 140 |    1.5.x     | 2.6.18 to 3.2
 141 |    1.6.x     | 2.6.18 to 3.2
 142 |    1.7.x     | 2.6.18 to 3.3
 143 |    1.8.x     | 2.6.18 to 3.4
 144 |    1.9.x     | 2.6.18 to 3.8
 145 |    1.10.x    | 2.6.18 to 3.8
 146 |    1.11.x    | 2.6.18 to 3.8
 147 |    2.0.x     | 2.6.32 to 3.10
 148 |    2.1.x     | 2.6.32 to 3.11
 149 |    2.3.x     | 2.6.32 to 3.14
 150
 151    Open vSwitch userspace should also work with the Linux kernel module
 152    built into Linux 3.3 and later.
 153
 154    Open vSwitch userspace is not sensitive to the Linux kernel version.
 155    It should build against almost any kernel, certainly against 2.6.32
 156    and later.
 157
 158 ### Q: I get an error like this when I configure Open vSwitch:
 159
 160        configure: error: Linux kernel in <dir> is version <x>, but
 161        version newer than <y> is not supported (please refer to the
 162        FAQ for advice)
 163
 164    What should I do?
 165
 166 A: If there is a newer version of Open vSwitch, consider building that
 167    one, because it may support the kernel that you are building
 168    against.  (To find out, consult the table in the previous answer.)
 169
 170    Otherwise, use the Linux kernel module supplied with the kernel
 171    that you are using.  All versions of Open vSwitch userspace are
 172    compatible with all versions of the Open vSwitch kernel module, so
 173    this will also work.  See also the following question.
 174
 175 ### Q: What features are not available in the Open vSwitch kernel datapath that ships as part of the upstream Linux kernel?
 176
 177 A: The kernel module in upstream Linux does not include support for
 178    LISP. Work is in progress to add support for LISP to the upstream
 179    Linux version of the Open vSwitch kernel module. For now, if you
 180    need this feature, use the kernel module from the Open vSwitch
 181    distribution instead of the upstream Linux kernel module.
 182
 183    Certain features require kernel support to function or to have
 184    reasonable performance. If the ovs-vswitchd log file indicates that
 185    a feature is not supported, consider upgrading to a newer upstream
 186    Linux release or using the kernel module paired with the userspace
 187    distribution.
 188
 189 ### Q: Why do tunnels not work when using a kernel module other than the one packaged with Open vSwitch?
 190
 191 A: Support for tunnels was added to the upstream Linux kernel module
 192    after the rest of Open vSwitch. As a result, some kernels may contain
 193    support for Open vSwitch but not tunnels. The minimum kernel version
 194    that supports each tunnel protocol is:
 195
 196 | Protocol |  Linux Kernel
 197 |:--------:|:-------------:
 198 | GRE      |    3.11
 199 | VXLAN    |    3.12
 200 | LISP     | <not upstream>
 201
 202    If you are using a version of the kernel that is older than the one
 203    listed above, it is still possible to use that tunnel protocol. However,
 204    you must compile and install the kernel module included with the Open
 205    vSwitch distribution rather than the one on your machine. If problems
 206    persist after doing this, check to make sure that the module that is
 207    loaded is the one you expect.
 208
 209 ### Q: What features are not available when using the userspace datapath?
 210
 211 A: Tunnel virtual ports are not supported, as described in the
 212    previous answer.  It is also not possible to use queue-related
 213    actions.  On Linux kernels before 2.6.39, maximum-sized VLAN packets
 214    may not be transmitted.
 215
 216 ### Q: What Linux kernel versions does IPFIX flow monitoring work with?
 217
 218 A: IPFIX flow monitoring requires the Linux kernel module from Open
 219    vSwitch version 1.10.90 or later.
 220
 221 ### Q: Should userspace or kernel be upgraded first to minimize downtime?
 222
 223    In general, the Open vSwitch userspace should be used with the
 224    kernel version included in the same release or with the version
 225    from upstream Linux.  However, when upgrading between two releases
 226    of Open vSwitch it is best to migrate userspace first to reduce
 227    the possibility of incompatibilities.
 228
 229 ### Q: What happened to the bridge compatibility feature?
 230
 231 A: Bridge compatibility was a feature of Open vSwitch 1.9 and earlier.
 232    When it was enabled, Open vSwitch imitated the interface of the
 233    Linux kernel "bridge" module.  This allowed users to drop Open
 234    vSwitch into environments designed to use the Linux kernel bridge
 235    module without adapting the environment to use Open vSwitch.
 236
 237    Open vSwitch 1.10 and later do not support bridge compatibility.
 238    The feature was dropped because version 1.10 adopted a new internal
 239    architecture that made bridge compatibility difficult to maintain.
 240    Now that many environments use OVS directly, it would be rarely
 241    useful in any case.
 242
 243    To use bridge compatibility, install OVS 1.9 or earlier, including
 244    the accompanying kernel modules (both the main and bridge
 245    compatibility modules), following the instructions that come with
 246    the release.  Be sure to start the ovs-brcompatd daemon.
 247
 248
 249 Terminology
 250 -----------
 251
 252 ### Q: I thought Open vSwitch was a virtual Ethernet switch, but the documentation keeps talking about bridges.  What's a bridge?
 253
 254 A: In networking, the terms "bridge" and "switch" are synonyms.  Open
 255    vSwitch implements an Ethernet switch, which means that it is also
 256    an Ethernet bridge.
 257
 258 ### Q: What's a VLAN?
 259
 260 A: See the "VLAN" section below.
 261
 262
 263 Basic Configuration
 264 -------------------
 265
 266 ### Q: How do I configure a port as an access port?
 267
 268 A: Add "tag=VLAN" to your "ovs-vsctl add-port" command.  For example,
 269    the following commands configure br0 with eth0 as a trunk port (the
 270    default) and tap0 as an access port for VLAN 9:
 271
 272        ovs-vsctl add-br br0
 273        ovs-vsctl add-port br0 eth0
 274        ovs-vsctl add-port br0 tap0 tag=9
 275
 276    If you want to configure an already added port as an access port,
 277    use "ovs-vsctl set", e.g.:
 278
 279        ovs-vsctl set port tap0 tag=9
 280
 281 ### Q: How do I configure a port as a SPAN port, that is, enable mirroring of all traffic to that port?
 282
 283 A: The following commands configure br0 with eth0 and tap0 as trunk
 284    ports.  All traffic coming in or going out on eth0 or tap0 is also
 285    mirrored to tap1; any traffic arriving on tap1 is dropped:
 286
 287        ovs-vsctl add-br br0
 288        ovs-vsctl add-port br0 eth0
 289        ovs-vsctl add-port br0 tap0
 290        ovs-vsctl add-port br0 tap1 \
 291            -- --id=@p get port tap1 \
 292            -- --id=@m create mirror name=m0 select-all=true output-port=@p \
 293            -- set bridge br0 mirrors=@m
 294
 295    To later disable mirroring, run:
 296
 297        ovs-vsctl clear bridge br0 mirrors
 298
 299 ### Q: Does Open vSwitch support configuring a port in promiscuous mode?
 300
 301 A: Yes.  How you configure it depends on what you mean by "promiscuous
 302    mode":
 303
 304   - Conventionally, "promiscuous mode" is a feature of a network
 305     interface card.  Ordinarily, a NIC passes to the CPU only the
 306     packets actually destined to its host machine.  It discards
 307     the rest to avoid wasting memory and CPU cycles.  When
 308     promiscuous mode is enabled, however, it passes every packet
 309     to the CPU.  On an old-style shared-media or hub-based
 310     network, this allows the host to spy on all packets on the
 311     network.  But in the switched networks that are almost
 312     everywhere these days, promiscuous mode doesn't have much
 313     effect, because few packets not destined to a host are
 314     delivered to the host's NIC.
 315
 316     This form of promiscuous mode is configured in the guest OS of
 317     the VMs on your bridge, e.g. with "ifconfig".
 318
 319   - The VMware vSwitch uses a different definition of "promiscuous
 320     mode".  When you configure promiscuous mode on a VMware vNIC,
 321     the vSwitch sends a copy of every packet received by the
 322     vSwitch to that vNIC.  That has a much bigger effect than just
 323     enabling promiscuous mode in a guest OS.  Rather than getting
 324     a few stray packets for which the switch does not yet know the
 325     correct destination, the vNIC gets every packet.  The effect
 326     is similar to replacing the vSwitch by a virtual hub.
 327
 328     This "promiscuous mode" is what switches normally call "port
 329     mirroring" or "SPAN".  For information on how to configure
 330     SPAN, see "How do I configure a port as a SPAN port, that is,
 331     enable mirroring of all traffic to that port?"
 332
 333 ### Q: How do I configure a VLAN as an RSPAN VLAN, that is, enable mirroring of all traffic to that VLAN?
 334
 335 A: The following commands configure br0 with eth0 as a trunk port and
 336    tap0 as an access port for VLAN 10.  All traffic coming in or going
 337    out on tap0, as well as traffic coming in or going out on eth0 in
 338    VLAN 10, is also mirrored to VLAN 15 on eth0.  The original tag for
 339    VLAN 10, in cases where one is present, is dropped as part of
 340    mirroring:
 341
 342        ovs-vsctl add-br br0
 343        ovs-vsctl add-port br0 eth0
 344        ovs-vsctl add-port br0 tap0 tag=10
 345        ovs-vsctl \
 346            -- --id=@m create mirror name=m0 select-all=true select-vlan=10 \
 347                                     output-vlan=15 \
 348            -- set bridge br0 mirrors=@m
 349
 350    To later disable mirroring, run:
 351
 352        ovs-vsctl clear bridge br0 mirrors
 353
 354    Mirroring to a VLAN can disrupt a network that contains unmanaged
 355    switches.  See ovs-vswitchd.conf.db(5) for details.  Mirroring to a
 356    GRE tunnel has fewer caveats than mirroring to a VLAN and should
 357    generally be preferred.
 358
 359 ### Q: Can I mirror more than one input VLAN to an RSPAN VLAN?
 360
 361 A: Yes, but mirroring to a VLAN strips the original VLAN tag in favor
 362    of the specified output-vlan.  This loss of information may make
 363    the mirrored traffic too hard to interpret.
 364
 365    To mirror multiple VLANs, use the commands above, but specify a
 366    comma-separated list of VLANs as the value for select-vlan.  To
 367    mirror every VLAN, use the commands above, but omit select-vlan and
 368    its value entirely.
 369
 370    When a packet arrives on a VLAN that is used as a mirror output
 371    VLAN, the mirror is disregarded.  Instead, in standalone mode, OVS
 372    floods the packet across all the ports for which the mirror output
 373    VLAN is configured.  (If an OpenFlow controller is in use, then it
 374    can override this behavior through the flow table.)  If OVS is used
 375    as an intermediate switch, rather than an edge switch, this ensures
 376    that the RSPAN traffic is distributed through the network.
 377
 378    Mirroring to a VLAN can disrupt a network that contains unmanaged
 379    switches.  See ovs-vswitchd.conf.db(5) for details.  Mirroring to a
 380    GRE tunnel has fewer caveats than mirroring to a VLAN and should
 381    generally be preferred.
 382
 383 ### Q: How do I configure mirroring of all traffic to a GRE tunnel?
 384
 385 A: The following commands configure br0 with eth0 and tap0 as trunk
 386    ports.  All traffic coming in or going out on eth0 or tap0 is also
 387    mirrored to gre0, a GRE tunnel to the remote host 192.168.1.10; any
 388    traffic arriving on gre0 is dropped:
 389
 390        ovs-vsctl add-br br0
 391        ovs-vsctl add-port br0 eth0
 392        ovs-vsctl add-port br0 tap0
 393        ovs-vsctl add-port br0 gre0 \
 394            -- set interface gre0 type=gre options:remote_ip=192.168.1.10 \
 395            -- --id=@p get port gre0 \
 396            -- --id=@m create mirror name=m0 select-all=true output-port=@p \
 397            -- set bridge br0 mirrors=@m
 398
 399    To later disable mirroring and destroy the GRE tunnel:
 400
 401        ovs-vsctl clear bridge br0 mirrors
 402        ovs-vcstl del-port br0 gre0
 403
 404 ### Q: Does Open vSwitch support ERSPAN?
 405
 406 A: No.  ERSPAN is an undocumented proprietary protocol.  As an
 407    alternative, Open vSwitch supports mirroring to a GRE tunnel (see
 408    above).
 409
 410 ### Q: How do I connect two bridges?
 411
 412 A: First, why do you want to do this?  Two connected bridges are not
 413    much different from a single bridge, so you might as well just have
 414    a single bridge with all your ports on it.
 415
 416    If you still want to connect two bridges, you can use a pair of
 417    patch ports.  The following example creates bridges br0 and br1,
 418    adds eth0 and tap0 to br0, adds tap1 to br1, and then connects br0
 419    and br1 with a pair of patch ports.
 420
 421        ovs-vsctl add-br br0
 422        ovs-vsctl add-port br0 eth0
 423        ovs-vsctl add-port br0 tap0
 424        ovs-vsctl add-br br1
 425        ovs-vsctl add-port br1 tap1
 426        ovs-vsctl \
 427            -- add-port br0 patch0 \
 428            -- set interface patch0 type=patch options:peer=patch1 \
 429            -- add-port br1 patch1 \
 430            -- set interface patch1 type=patch options:peer=patch0
 431
 432    Bridges connected with patch ports are much like a single bridge.
 433    For instance, if the example above also added eth1 to br1, and both
 434    eth0 and eth1 happened to be connected to the same next-hop switch,
 435    then you could loop your network just as you would if you added
 436    eth0 and eth1 to the same bridge (see the "Configuration Problems"
 437    section below for more information).
 438
 439    If you are using Open vSwitch 1.9 or an earlier version, then you
 440    need to be using the kernel module bundled with Open vSwitch rather
 441    than the one that is integrated into Linux 3.3 and later, because
 442    Open vSwitch 1.9 and earlier versions need kernel support for patch
 443    ports.  This also means that in Open vSwitch 1.9 and earlier, patch
 444    ports will not work with the userspace datapath, only with the
 445    kernel module.
 446
 447 ### Q: How do I configure a bridge without an OpenFlow local port?  (Local port in the sense of OFPP_LOCAL)
 448
 449 A: Open vSwitch does not support such a configuration.
 450    Bridges always have their local ports.
 451
 452
 453 Implementation Details
 454 ----------------------
 455
 456 ### Q: I hear OVS has a couple of kinds of flows.  Can you tell me about them?
 457
 458 A: Open vSwitch uses different kinds of flows for different purposes:
 459
 460   - OpenFlow flows are the most important kind of flow.  OpenFlow
 461     controllers use these flows to define a switch's policy.
 462     OpenFlow flows support wildcards, priorities, and multiple
 463     tables.
 464
 465     When in-band control is in use, Open vSwitch sets up a few
 466     "hidden" flows, with priority higher than a controller or the
 467     user can configure, that are not visible via OpenFlow.  (See
 468     the "Controller" section of the FAQ for more information
 469     about hidden flows.)
 470
 471   - The Open vSwitch software switch implementation uses a second
 472     kind of flow internally.  These flows, called "datapath" or
 473     "kernel" flows, do not support priorities and comprise only a
 474     single table, which makes them suitable for caching.  (Like
 475     OpenFlow flows, datapath flows do support wildcarding, in Open
 476     vSwitch 1.11 and later.)  OpenFlow flows and datapath flows
 477     also support different actions and number ports differently.
 478
 479     Datapath flows are an implementation detail that is subject to
 480     change in future versions of Open vSwitch.  Even with the
 481     current version of Open vSwitch, hardware switch
 482     implementations do not necessarily use this architecture.
 483
 484    Users and controllers directly control only the OpenFlow flow
 485    table.  Open vSwitch manages the datapath flow table itself, so
 486    users should not normally be concerned with it.
 487
 488 ### Q: Why are there so many different ways to dump flows?
 489
 490 A: Open vSwitch has two kinds of flows (see the previous question), so
 491    it has commands with different purposes for dumping each kind of
 492    flow:
 493
 494   - `ovs-ofctl dump-flows <br>` dumps OpenFlow flows, excluding
 495     hidden flows.  This is the most commonly useful form of flow
 496     dump.  (Unlike the other commands, this should work with any
 497     OpenFlow switch, not just Open vSwitch.)
 498
 499   - `ovs-appctl bridge/dump-flows <br>` dumps OpenFlow flows,
 500     including hidden flows.  This is occasionally useful for
 501     troubleshooting suspected issues with in-band control.
 502
 503   - `ovs-dpctl dump-flows [dp]` dumps the datapath flow table
 504     entries for a Linux kernel-based datapath.  In Open vSwitch
 505     1.10 and later, ovs-vswitchd merges multiple switches into a
 506     single datapath, so it will show all the flows on all your
 507     kernel-based switches.  This command can occasionally be
 508     useful for debugging.
 509
 510   - `ovs-appctl dpif/dump-flows <br>`, new in Open vSwitch 1.10,
 511     dumps datapath flows for only the specified bridge, regardless
 512     of the type.
 513
 514 ### Q: How does multicast snooping works with VLANs?
 515
 516 A: Open vSwitch maintains snooping tables for each VLAN.
 517
 518
 519 Performance
 520 -----------
 521
 522 ### Q: I just upgraded and I see a performance drop.  Why?
 523
 524 A: The OVS kernel datapath may have been updated to a newer version than
 525    the OVS userspace components.  Sometimes new versions of OVS kernel
 526    module add functionality that is backwards compatible with older
 527    userspace components but may cause a drop in performance with them.
 528    Especially, if a kernel module from OVS 2.1 or newer is paired with
 529    OVS userspace 1.10 or older, there will be a performance drop for
 530    TCP traffic.
 531
 532    Updating the OVS userspace components to the latest released
 533    version should fix the performance degradation.
 534
 535    To get the best possible performance and functionality, it is
 536    recommended to pair the same versions of the kernel module and OVS
 537    userspace.
 538
 539
 540 Configuration Problems
 541 ----------------------
 542
 543 ### Q: I created a bridge and added my Ethernet port to it, using commands
 544    like these:
 545
 546        ovs-vsctl add-br br0
 547        ovs-vsctl add-port br0 eth0
 548
 549    and as soon as I ran the "add-port" command I lost all connectivity
 550    through eth0.  Help!
 551
 552 A: A physical Ethernet device that is part of an Open vSwitch bridge
 553    should not have an IP address.  If one does, then that IP address
 554    will not be fully functional.
 555
 556    You can restore functionality by moving the IP address to an Open
 557    vSwitch "internal" device, such as the network device named after
 558    the bridge itself.  For example, assuming that eth0's IP address is
 559    192.168.128.5, you could run the commands below to fix up the
 560    situation:
 561
 562        ifconfig eth0 0.0.0.0
 563        ifconfig br0 192.168.128.5
 564
 565    (If your only connection to the machine running OVS is through the
 566    IP address in question, then you would want to run all of these
 567    commands on a single command line, or put them into a script.)  If
 568    there were any additional routes assigned to eth0, then you would
 569    also want to use commands to adjust these routes to go through br0.
 570
 571    If you use DHCP to obtain an IP address, then you should kill the
 572    DHCP client that was listening on the physical Ethernet interface
 573    (e.g. eth0) and start one listening on the internal interface
 574    (e.g. br0).  You might still need to manually clear the IP address
 575    from the physical interface (e.g. with "ifconfig eth0 0.0.0.0").
 576
 577    There is no compelling reason why Open vSwitch must work this way.
 578    However, this is the way that the Linux kernel bridge module has
 579    always worked, so it's a model that those accustomed to Linux
 580    bridging are already used to.  Also, the model that most people
 581    expect is not implementable without kernel changes on all the
 582    versions of Linux that Open vSwitch supports.
 583
 584    By the way, this issue is not specific to physical Ethernet
 585    devices.  It applies to all network devices except Open vSwitch
 586    "internal" devices.
 587
 588 ### Q: I created a bridge and added a couple of Ethernet ports to it,
 589 ### using commands like these:
 590
 591        ovs-vsctl add-br br0
 592        ovs-vsctl add-port br0 eth0
 593        ovs-vsctl add-port br0 eth1
 594
 595    and now my network seems to have melted: connectivity is unreliable
 596    (even connectivity that doesn't go through Open vSwitch), all the
 597    LEDs on my physical switches are blinking, wireshark shows
 598    duplicated packets, and CPU usage is very high.
 599
 600 A: More than likely, you've looped your network.  Probably, eth0 and
 601    eth1 are connected to the same physical Ethernet switch.  This
 602    yields a scenario where OVS receives a broadcast packet on eth0 and
 603    sends it out on eth1, then the physical switch connected to eth1
 604    sends the packet back on eth0, and so on forever.  More complicated
 605    scenarios, involving a loop through multiple switches, are possible
 606    too.
 607
 608    The solution depends on what you are trying to do:
 609
 610    - If you added eth0 and eth1 to get higher bandwidth or higher
 611      reliability between OVS and your physical Ethernet switch,
 612      use a bond.  The following commands create br0 and then add
 613      eth0 and eth1 as a bond:
 614
 615          ovs-vsctl add-br br0
 616          ovs-vsctl add-bond br0 bond0 eth0 eth1
 617
 618      Bonds have tons of configuration options.  Please read the
 619      documentation on the Port table in ovs-vswitchd.conf.db(5)
 620      for all the details.
 621
 622    - Perhaps you don't actually need eth0 and eth1 to be on the
 623      same bridge.  For example, if you simply want to be able to
 624      connect each of them to virtual machines, then you can put
 625      each of them on a bridge of its own:
 626
 627          ovs-vsctl add-br br0
 628          ovs-vsctl add-port br0 eth0
 629
 630          ovs-vsctl add-br br1
 631          ovs-vsctl add-port br1 eth1
 632
 633      and then connect VMs to br0 and br1.  (A potential
 634      disadvantage is that traffic cannot directly pass between br0
 635      and br1.  Instead, it will go out eth0 and come back in eth1,
 636      or vice versa.)
 637
 638    - If you have a redundant or complex network topology and you
 639      want to prevent loops, turn on spanning tree protocol (STP).
 640      The following commands create br0, enable STP, and add eth0
 641      and eth1 to the bridge.  The order is important because you
 642      don't want have to have a loop in your network even
 643      transiently:
 644
 645          ovs-vsctl add-br br0
 646          ovs-vsctl set bridge br0 stp_enable=true
 647          ovs-vsctl add-port br0 eth0
 648          ovs-vsctl add-port br0 eth1
 649
 650      The Open vSwitch implementation of STP is not well tested.
 651      Please report any bugs you observe, but if you'd rather avoid
 652      acting as a beta tester then another option might be your
 653      best shot.
 654
 655 ### Q: I can't seem to use Open vSwitch in a wireless network.
 656
 657 A: Wireless base stations generally only allow packets with the source
 658    MAC address of NIC that completed the initial handshake.
 659    Therefore, without MAC rewriting, only a single device can
 660    communicate over a single wireless link.
 661
 662    This isn't specific to Open vSwitch, it's enforced by the access
 663    point, so the same problems will show up with the Linux bridge or
 664    any other way to do bridging.
 665
 666 ### Q: I can't seem to add my PPP interface to an Open vSwitch bridge.
 667
 668 A: PPP most commonly carries IP packets, but Open vSwitch works only
 669    with Ethernet frames.  The correct way to interface PPP to an
 670    Ethernet network is usually to use routing instead of switching.
 671
 672 ### Q: Is there any documentation on the database tables and fields?
 673
 674 A: Yes.  ovs-vswitchd.conf.db(5) is a comprehensive reference.
 675
 676 ### Q: When I run ovs-dpctl I no longer see the bridges I created.  Instead,
 677    I only see a datapath called "ovs-system".  How can I see datapath
 678    information about a particular bridge?
 679
 680 A: In version 1.9.0, OVS switched to using a single datapath that is
 681    shared by all bridges of that type.  The "ovs-appctl dpif/*"
 682    commands provide similar functionality that is scoped by the bridge.
 683
 684 ### Q: I created a GRE port using ovs-vsctl so why can't I send traffic or
 685    see the port in the datapath?
 686
 687 A: On Linux kernels before 3.11, the OVS GRE module and Linux GRE module
 688    cannot be loaded at the same time. It is likely that on your system the
 689    Linux GRE module is already loaded and blocking OVS (to confirm, check
 690    dmesg for errors regarding GRE registration). To fix this, unload all
 691    GRE modules that appear in lsmod as well as the OVS kernel module. You
 692    can then reload the OVS module following the directions in
 693    [INSTALL.md], which will ensure that dependencies are satisfied.
 694
 695 ### Q: Open vSwitch does not seem to obey my packet filter rules.
 696
 697 A: It depends on mechanisms and configurations you want to use.
 698
 699    You cannot usefully use typical packet filters, like iptables, on
 700    physical Ethernet ports that you add to an Open vSwitch bridge.
 701    This is because Open vSwitch captures packets from the interface at
 702    a layer lower below where typical packet-filter implementations
 703    install their hooks.  (This actually applies to any interface of
 704    type "system" that you might add to an Open vSwitch bridge.)
 705
 706    You can usefully use typical packet filters on Open vSwitch
 707    internal ports as they are mostly ordinary interfaces from the point
 708    of view of packet filters.
 709
 710    For example, suppose you create a bridge br0 and add Ethernet port
 711    eth0 to it.  Then you can usefully add iptables rules to affect the
 712    internal interface br0, but not the physical interface eth0.  (br0
 713    is also where you would add an IP address, as discussed elsewhere
 714    in the FAQ.)
 715
 716    For simple filtering rules, it might be possible to achieve similar
 717    results by installing appropriate OpenFlow flows instead.
 718
 719    If the use of a particular packet filter setup is essential, Open
 720    vSwitch might not be the best choice for you.  On Linux, you might
 721    want to consider using the Linux Bridge.  (This is the only choice if
 722    you want to use ebtables rules.)  On NetBSD, you might want to
 723    consider using the bridge(4) with BRIDGE_IPF option.
 724
 725 ### Q: It seems that Open vSwitch does nothing when I removed a port and
 726    then immediately put it back.  For example, consider that p1 is
 727    a port of type=internal:
 728
 729        ovs-vsctl del-port br0 p1 -- \
 730            add-port br0 p1 -- \
 731            set interface p1 type=internal
 732
 733 A: It's an expected behaviour.
 734
 735    If del-port and add-port happen in a single OVSDB transaction as
 736    your example, Open vSwitch always "skips" the intermediate steps.
 737    Even if they are done in multiple transactions, it's still allowed
 738    for Open vSwitch to skip the intermediate steps and just implement
 739    the overall effect.  In both cases, your example would be turned
 740    into a no-op.
 741
 742    If you want to make Open vSwitch actually destroy and then re-create
 743    the port for some side effects like resetting kernel setting for the
 744    corresponding interface, you need to separate operations into multiple
 745    OVSDB transactions and ensure that at least the first one does not have
 746    --no-wait.  In the following example, the first ovs-vsctl will block
 747    until Open vSwitch reloads the new configuration and removes the port:
 748
 749        ovs-vsctl del-port br0 p1
 750        ovs-vsctl add-port br0 p1 -- \
 751            set interface p1 type=internal
 752
 753 ### Q: I want to add thousands of ports to an Open vSwitch bridge, but
 754    it takes too long (minutes or hours) to do it with ovs-vsctl.  How
 755    can I do it faster?
 756
 757 A: If you add them one at a time with ovs-vsctl, it can take a long
 758    time to add thousands of ports to an Open vSwitch bridge.  This is
 759    because every invocation of ovs-vsctl first reads the current
 760    configuration from OVSDB.  As the number of ports grows, this
 761    starts to take an appreciable amount of time, and when it is
 762    repeated thousands of times the total time becomes significant.
 763
 764    The solution is to add the ports in one invocation of ovs-vsctl (or
 765    a small number of them).  For example, using bash:
 766
 767        ovs-vsctl add-br br0
 768        cmds=; for i in {1..5000}; do cmds+=" -- add-port br0 p$i"; done
 769        ovs-vsctl $cmds
 770
 771    takes seconds, not minutes or hours, in the OVS sandbox environment.
 772
 773 Quality of Service (QoS)
 774 ------------------------
 775
 776 ### Q: How do I configure Quality of Service (QoS)?
 777
 778 A: Suppose that you want to set up bridge br0 connected to physical
 779    Ethernet port eth0 (a 1 Gbps device) and virtual machine interfaces
 780    vif1.0 and vif2.0, and that you want to limit traffic from vif1.0
 781    to eth0 to 10 Mbps and from vif2.0 to eth0 to 20 Mbps.  Then, you
 782    could configure the bridge this way:
 783
 784        ovs-vsctl -- \
 785            add-br br0 -- \
 786            add-port br0 eth0 -- \
 787            add-port br0 vif1.0 -- set interface vif1.0 ofport_request=5 -- \
 788            add-port br0 vif2.0 -- set interface vif2.0 ofport_request=6 -- \
 789            set port eth0 qos=@newqos -- \
 790            --id=@newqos create qos type=linux-htb \
 791                other-config:max-rate=1000000000 \
 792                queues:123=@vif10queue \
 793                queues:234=@vif20queue -- \
 794            --id=@vif10queue create queue other-config:max-rate=10000000 -- \
 795            --id=@vif20queue create queue other-config:max-rate=20000000
 796
 797    At this point, bridge br0 is configured with the ports and eth0 is
 798    configured with the queues that you need for QoS, but nothing is
 799    actually directing packets from vif1.0 or vif2.0 to the queues that
 800    we have set up for them.  That means that all of the packets to
 801    eth0 are going to the "default queue", which is not what we want.
 802
 803    We use OpenFlow to direct packets from vif1.0 and vif2.0 to the
 804    queues reserved for them:
 805
 806        ovs-ofctl add-flow br0 in_port=5,actions=set_queue:123,normal
 807        ovs-ofctl add-flow br0 in_port=6,actions=set_queue:234,normal
 808
 809    Each of the above flows matches on the input port, sets up the
 810    appropriate queue (123 for vif1.0, 234 for vif2.0), and then
 811    executes the "normal" action, which performs the same switching
 812    that Open vSwitch would have done without any OpenFlow flows being
 813    present.  (We know that vif1.0 and vif2.0 have OpenFlow port
 814    numbers 5 and 6, respectively, because we set their ofport_request
 815    columns above.  If we had not done that, then we would have needed
 816    to find out their port numbers before setting up these flows.)
 817
 818    Now traffic going from vif1.0 or vif2.0 to eth0 should be
 819    rate-limited.
 820
 821    By the way, if you delete the bridge created by the above commands,
 822    with:
 823
 824        ovs-vsctl del-br br0
 825
 826    then that will leave one unreferenced QoS record and two
 827    unreferenced Queue records in the Open vSwich database.  One way to
 828    clear them out, assuming you don't have other QoS or Queue records
 829    that you want to keep, is:
 830
 831        ovs-vsctl -- --all destroy QoS -- --all destroy Queue
 832
 833    If you do want to keep some QoS or Queue records, or the Open
 834    vSwitch you are using is older than version 1.8 (which added the
 835    --all option), then you will have to destroy QoS and Queue records
 836    individually.
 837
 838 ### Q: I configured Quality of Service (QoS) in my OpenFlow network by
 839    adding records to the QoS and Queue table, but the results aren't
 840    what I expect.
 841
 842 A: Did you install OpenFlow flows that use your queues?  This is the
 843    primary way to tell Open vSwitch which queues you want to use.  If
 844    you don't do this, then the default queue will be used, which will
 845    probably not have the effect you want.
 846
 847    Refer to the previous question for an example.
 848
 849 ### Q: I'd like to take advantage of some QoS feature that Open vSwitch
 850    doesn't yet support.  How do I do that?
 851
 852 A: Open vSwitch does not implement QoS itself.  Instead, it can
 853    configure some, but not all, of the QoS features built into the
 854    Linux kernel.  If you need some QoS feature that OVS cannot
 855    configure itself, then the first step is to figure out whether
 856    Linux QoS supports that feature.  If it does, then you can submit a
 857    patch to support Open vSwitch configuration for that feature, or
 858    you can use "tc" directly to configure the feature in Linux.  (If
 859    Linux QoS doesn't support the feature you want, then first you have
 860    to add that support to Linux.)
 861
 862 ### Q: I configured QoS, correctly, but my measurements show that it isn't
 863    working as well as I expect.
 864
 865 A: With the Linux kernel, the Open vSwitch implementation of QoS has
 866    two aspects:
 867
 868    - Open vSwitch configures a subset of Linux kernel QoS
 869      features, according to what is in OVSDB.  It is possible that
 870      this code has bugs.  If you believe that this is so, then you
 871      can configure the Linux traffic control (QoS) stack directly
 872      with the "tc" program.  If you get better results that way,
 873      you can send a detailed bug report to bugs@openvswitch.org.
 874
 875      It is certain that Open vSwitch cannot configure every Linux
 876      kernel QoS feature.  If you need some feature that OVS cannot
 877      configure, then you can also use "tc" directly (or add that
 878      feature to OVS).
 879
 880    - The Open vSwitch implementation of OpenFlow allows flows to
 881      be directed to particular queues.  This is pretty simple and
 882      unlikely to have serious bugs at this point.
 883
 884    However, most problems with QoS on Linux are not bugs in Open
 885    vSwitch at all.  They tend to be either configuration errors
 886    (please see the earlier questions in this section) or issues with
 887    the traffic control (QoS) stack in Linux.  The Open vSwitch
 888    developers are not experts on Linux traffic control.  We suggest
 889    that, if you believe you are encountering a problem with Linux
 890    traffic control, that you consult the tc manpages (e.g. tc(8),
 891    tc-htb(8), tc-hfsc(8)), web resources (e.g. http://lartc.org/), or
 892    mailing lists (e.g. http://vger.kernel.org/vger-lists.html#netdev).
 893
 894 ### Q: Does Open vSwitch support OpenFlow meters?
 895
 896 A: Since version 2.0, Open vSwitch has OpenFlow protocol support for
 897    OpenFlow meters.  There is no implementation of meters in the Open
 898    vSwitch software switch (neither the kernel-based nor userspace
 899    switches).
 900
 901
 902 VLANs
 903 -----
 904
 905 ### Q: What's a VLAN?
 906
 907 A: At the simplest level, a VLAN (short for "virtual LAN") is a way to
 908    partition a single switch into multiple switches.  Suppose, for
 909    example, that you have two groups of machines, group A and group B.
 910    You want the machines in group A to be able to talk to each other,
 911    and you want the machine in group B to be able to talk to each
 912    other, but you don't want the machines in group A to be able to
 913    talk to the machines in group B.  You can do this with two
 914    switches, by plugging the machines in group A into one switch and
 915    the machines in group B into the other switch.
 916
 917    If you only have one switch, then you can use VLANs to do the same
 918    thing, by configuring the ports for machines in group A as VLAN
 919    "access ports" for one VLAN and the ports for group B as "access
 920    ports" for a different VLAN.  The switch will only forward packets
 921    between ports that are assigned to the same VLAN, so this
 922    effectively subdivides your single switch into two independent
 923    switches, one for each group of machines.
 924
 925    So far we haven't said anything about VLAN headers.  With access
 926    ports, like we've described so far, no VLAN header is present in
 927    the Ethernet frame.  This means that the machines (or switches)
 928    connected to access ports need not be aware that VLANs are
 929    involved, just like in the case where we use two different physical
 930    switches.
 931
 932    Now suppose that you have a whole bunch of switches in your
 933    network, instead of just one, and that some machines in group A are
 934    connected directly to both switches 1 and 2.  To allow these
 935    machines to talk to each other, you could add an access port for
 936    group A's VLAN to switch 1 and another to switch 2, and then
 937    connect an Ethernet cable between those ports.  That works fine,
 938    but it doesn't scale well as the number of switches and the number
 939    of VLANs increases, because you use up a lot of valuable switch
 940    ports just connecting together your VLANs.
 941
 942    This is where VLAN headers come in.  Instead of using one cable and
 943    two ports per VLAN to connect a pair of switches, we configure a
 944    port on each switch as a VLAN "trunk port".  Packets sent and
 945    received on a trunk port carry a VLAN header that says what VLAN
 946    the packet belongs to, so that only two ports total are required to
 947    connect the switches, regardless of the number of VLANs in use.
 948    Normally, only switches (either physical or virtual) are connected
 949    to a trunk port, not individual hosts, because individual hosts
 950    don't expect to see a VLAN header in the traffic that they receive.
 951
 952    None of the above discussion says anything about particular VLAN
 953    numbers.  This is because VLAN numbers are completely arbitrary.
 954    One must only ensure that a given VLAN is numbered consistently
 955    throughout a network and that different VLANs are given different
 956    numbers.  (That said, VLAN 0 is usually synonymous with a packet
 957    that has no VLAN header, and VLAN 4095 is reserved.)
 958
 959 ### Q: VLANs don't work.
 960
 961 A: Many drivers in Linux kernels before version 3.3 had VLAN-related
 962    bugs.  If you are having problems with VLANs that you suspect to be
 963    driver related, then you have several options:
 964
 965    - Upgrade to Linux 3.3 or later.
 966
 967    - Build and install a fixed version of the particular driver
 968      that is causing trouble, if one is available.
 969
 970    - Use a NIC whose driver does not have VLAN problems.
 971
 972    - Use "VLAN splinters", a feature in Open vSwitch 1.4 and later
 973      that works around bugs in kernel drivers.  To enable VLAN
 974      splinters on interface eth0, use the command:
 975
 976        ovs-vsctl set interface eth0 other-config:enable-vlan-splinters=true
 977
 978      For VLAN splinters to be effective, Open vSwitch must know
 979      which VLANs are in use.  See the "VLAN splinters" section in
 980      the Interface table in ovs-vswitchd.conf.db(5) for details on
 981      how Open vSwitch infers in-use VLANs.
 982
 983      VLAN splinters increase memory use and reduce performance, so
 984      use them only if needed.
 985
 986    - Apply the "vlan workaround" patch from the XenServer kernel
 987      patch queue, build Open vSwitch against this patched kernel,
 988      and then use ovs-vlan-bug-workaround(8) to enable the VLAN
 989      workaround for each interface whose driver is buggy.
 990
 991      (This is a nontrivial exercise, so this option is included
 992      only for completeness.)
 993
 994    It is not always easy to tell whether a Linux kernel driver has
 995    buggy VLAN support.  The ovs-vlan-test(8) and ovs-test(8) utilities
 996    can help you test.  See their manpages for details.  Of the two
 997    utilities, ovs-test(8) is newer and more thorough, but
 998    ovs-vlan-test(8) may be easier to use.
 999
1000 ### Q: VLANs still don't work.  I've tested the driver so I know that it's OK.
1001
1002 A: Do you have VLANs enabled on the physical switch that OVS is
1003    attached to?  Make sure that the port is configured to trunk the
1004    VLAN or VLANs that you are using with OVS.
1005
1006 ### Q: Outgoing VLAN-tagged traffic goes through OVS to my physical switch
1007    and to its destination host, but OVS seems to drop incoming return
1008    traffic.
1009
1010 A: It's possible that you have the VLAN configured on your physical
1011    switch as the "native" VLAN.  In this mode, the switch treats
1012    incoming packets either tagged with the native VLAN or untagged as
1013    part of the native VLAN.  It may also send outgoing packets in the
1014    native VLAN without a VLAN tag.
1015
1016    If this is the case, you have two choices:
1017
1018    - Change the physical switch port configuration to tag packets
1019      it forwards to OVS with the native VLAN instead of forwarding
1020      them untagged.
1021
1022    - Change the OVS configuration for the physical port to a
1023      native VLAN mode.  For example, the following sets up a
1024      bridge with port eth0 in "native-tagged" mode in VLAN 9:
1025
1026          ovs-vsctl add-br br0
1027          ovs-vsctl add-port br0 eth0 tag=9 vlan_mode=native-tagged
1028
1029      In this situation, "native-untagged" mode will probably work
1030      equally well.  Refer to the documentation for the Port table
1031      in ovs-vswitchd.conf.db(5) for more information.
1032
1033 ### Q: I added a pair of VMs on different VLANs, like this:
1034
1035        ovs-vsctl add-br br0
1036        ovs-vsctl add-port br0 eth0
1037        ovs-vsctl add-port br0 tap0 tag=9
1038        ovs-vsctl add-port br0 tap1 tag=10
1039
1040     but the VMs can't access each other, the external network, or the
1041     Internet.
1042
1043 A: It is to be expected that the VMs can't access each other.  VLANs
1044    are a means to partition a network.  When you configured tap0 and
1045    tap1 as access ports for different VLANs, you indicated that they
1046    should be isolated from each other.
1047
1048    As for the external network and the Internet, it seems likely that
1049    the machines you are trying to access are not on VLAN 9 (or 10) and
1050    that the Internet is not available on VLAN 9 (or 10).
1051
1052 ### Q: I added a pair of VMs on the same VLAN, like this:
1053
1054        ovs-vsctl add-br br0
1055        ovs-vsctl add-port br0 eth0
1056        ovs-vsctl add-port br0 tap0 tag=9
1057        ovs-vsctl add-port br0 tap1 tag=9
1058
1059     The VMs can access each other, but not the external network or the
1060     Internet.
1061
1062 A: It seems likely that the machines you are trying to access in the
1063    external network are not on VLAN 9 and that the Internet is not
1064    available on VLAN 9.  Also, ensure VLAN 9 is set up as an allowed
1065    trunk VLAN on the upstream switch port to which eth0 is connected.
1066
1067 ### Q: Can I configure an IP address on a VLAN?
1068
1069 A: Yes.  Use an "internal port" configured as an access port.  For
1070    example, the following configures IP address 192.168.0.7 on VLAN 9.
1071    That is, OVS will forward packets from eth0 to 192.168.0.7 only if
1072    they have an 802.1Q header with VLAN 9.  Conversely, traffic
1073    forwarded from 192.168.0.7 to eth0 will be tagged with an 802.1Q
1074    header with VLAN 9:
1075
1076        ovs-vsctl add-br br0
1077        ovs-vsctl add-port br0 eth0
1078        ovs-vsctl add-port br0 vlan9 tag=9 -- set interface vlan9 type=internal
1079        ifconfig vlan9 192.168.0.7
1080
1081    See also the following question.
1082
1083 ### Q: I configured one IP address on VLAN 0 and another on VLAN 9, like
1084    this:
1085
1086        ovs-vsctl add-br br0
1087        ovs-vsctl add-port br0 eth0
1088        ifconfig br0 192.168.0.5
1089        ovs-vsctl add-port br0 vlan9 tag=9 -- set interface vlan9 type=internal
1090        ifconfig vlan9 192.168.0.9
1091
1092    but other hosts that are only on VLAN 0 can reach the IP address
1093    configured on VLAN 9.  What's going on?
1094
1095 A: RFC 1122 section 3.3.4.2 "Multihoming Requirements" describes two
1096    approaches to IP address handling in Internet hosts:
1097
1098    - In the "Strong ES Model", where an ES is a host ("End
1099      System"), an IP address is primarily associated with a
1100      particular interface.  The host discards packets that arrive
1101      on interface A if they are destined for an IP address that is
1102      configured on interface B.  The host never sends packets from
1103      interface A using a source address configured on interface B.
1104
1105    - In the "Weak ES Model", an IP address is primarily associated
1106      with a host.  The host accepts packets that arrive on any
1107      interface if they are destined for any of the host's IP
1108      addresses, even if the address is configured on some
1109      interface other than the one on which it arrived.  The host
1110      does not restrict itself to sending packets from an IP
1111      address associated with the originating interface.
1112
1113    Linux uses the weak ES model.  That means that when packets
1114    destined to the VLAN 9 IP address arrive on eth0 and are bridged to
1115    br0, the kernel IP stack accepts them there for the VLAN 9 IP
1116    address, even though they were not received on vlan9, the network
1117    device for vlan9.
1118
1119    To simulate the strong ES model on Linux, one may add iptables rule
1120    to filter packets based on source and destination address and
1121    adjust ARP configuration with sysctls.
1122
1123    BSD uses the strong ES model.
1124
1125 ### Q: My OpenFlow controller doesn't see the VLANs that I expect.
1126
1127 A: The configuration for VLANs in the Open vSwitch database (e.g. via
1128    ovs-vsctl) only affects traffic that goes through Open vSwitch's
1129    implementation of the OpenFlow "normal switching" action.  By
1130    default, when Open vSwitch isn't connected to a controller and
1131    nothing has been manually configured in the flow table, all traffic
1132    goes through the "normal switching" action.  But, if you set up
1133    OpenFlow flows on your own, through a controller or using ovs-ofctl
1134    or through other means, then you have to implement VLAN handling
1135    yourself.
1136
1137    You can use "normal switching" as a component of your OpenFlow
1138    actions, e.g. by putting "normal" into the lists of actions on
1139    ovs-ofctl or by outputting to OFPP_NORMAL from an OpenFlow
1140    controller.  In situations where this is not suitable, you can
1141    implement VLAN handling yourself, e.g.:
1142
1143    - If a packet comes in on an access port, and the flow table
1144      needs to send it out on a trunk port, then the flow can add
1145      the appropriate VLAN tag with the "mod_vlan_vid" action.
1146
1147    - If a packet comes in on a trunk port, and the flow table
1148      needs to send it out on an access port, then the flow can
1149      strip the VLAN tag with the "strip_vlan" action.
1150
1151 ### Q: I configured ports on a bridge as access ports with different VLAN
1152    tags, like this:
1153
1154        ovs-vsctl add-br br0
1155        ovs-vsctl set-controller br0 tcp:192.168.0.10:6633
1156        ovs-vsctl add-port br0 eth0
1157        ovs-vsctl add-port br0 tap0 tag=9
1158        ovs-vsctl add-port br0 tap1 tag=10
1159
1160    but the VMs running behind tap0 and tap1 can still communicate,
1161    that is, they are not isolated from each other even though they are
1162    on different VLANs.
1163
1164 A: Do you have a controller configured on br0 (as the commands above
1165    do)?  If so, then this is a variant on the previous question, "My
1166    OpenFlow controller doesn't see the VLANs that I expect," and you
1167    can refer to the answer there for more information.
1168
1169 ### Q: How MAC learning works with VLANs?
1170
1171 A: Open vSwitch implements Independent VLAN Learning (IVL) for
1172    OFPP_NORMAL action.  I.e. it logically has separate learning tables
1173    for each VLANs.
1174
1175
1176 VXLANs
1177 -----
1178
1179 ### Q: What's a VXLAN?
1180
1181 A: VXLAN stands for Virtual eXtensible Local Area Network, and is a means
1182    to solve the scaling challenges of VLAN networks in a multi-tenant
1183    environment. VXLAN is an overlay network which transports an L2 network
1184    over an existing L3 network. For more information on VXLAN, please see
1185    the IETF draft available here:
1186
1187    http://tools.ietf.org/html/draft-mahalingam-dutt-dcops-vxlan-03
1188
1189 ### Q: How much of the VXLAN protocol does Open vSwitch currently support?
1190
1191 A: Open vSwitch currently supports the framing format for packets on the
1192    wire. There is currently no support for the multicast aspects of VXLAN.
1193    To get around the lack of multicast support, it is possible to
1194    pre-provision MAC to IP address mappings either manually or from a
1195    controller.
1196
1197 ### Q: What destination UDP port does the VXLAN implementation in Open vSwitch
1198    use?
1199
1200 A: By default, Open vSwitch will use the assigned IANA port for VXLAN, which
1201    is 4789. However, it is possible to configure the destination UDP port
1202    manually on a per-VXLAN tunnel basis. An example of this configuration is
1203    provided below.
1204
1205        ovs-vsctl add-br br0
1206        ovs-vsctl add-port br0 vxlan1 -- set interface vxlan1
1207        type=vxlan options:remote_ip=192.168.1.2 options:key=flow
1208        options:dst_port=8472
1209
1210
1211 Using OpenFlow (Manually or Via Controller)
1212 -------------------------------------------
1213
1214 ### Q: What versions of OpenFlow does Open vSwitch support?
1215
1216 A: The following table lists the versions of OpenFlow supported by
1217    each version of Open vSwitch:
1218
1219        Open vSwitch      OF1.0  OF1.1  OF1.2  OF1.3  OF1.4  OF1.5
1220        ###============   =====  =====  =====  =====  =====  =====
1221        1.9 and earlier    yes    ---    ---    ---    ---    ---
1222        1.10               yes    ---    [*]    [*]    ---    ---
1223        1.11               yes    ---    [*]    [*]    ---    ---
1224        2.0                yes    [*]    [*]    [*]    ---    ---
1225        2.1                yes    [*]    [*]    [*]    ---    ---
1226        2.2                yes    [*]    [*]    [*]    [%]    [*]
1227        2.3                yes    yes    yes    yes    [*]    [*]
1228
1229        [*] Supported, with one or more missing features.
1230        [%] Experimental, unsafe implementation.
1231
1232    Open vSwitch 2.3 enables OpenFlow 1.0, 1.1, 1.2, and 1.3 by default
1233    in ovs-vswitchd.  In Open vSwitch 1.10 through 2.2, OpenFlow 1.1,
1234    1.2, and 1.3 must be enabled manually in ovs-vswitchd.  OpenFlow
1235    1.4 and 1.5 are also supported, with missing features, in Open
1236    vSwitch 2.3 and later, but not enabled by default.  In any case,
1237    the user may override the default:
1238
1239    - To enable OpenFlow 1.0, 1.1, 1.2, and 1.3 on bridge br0:
1240
1241      ovs-vsctl set bridge br0 protocols=OpenFlow10,OpenFlow11,OpenFlow12,OpenFlow13
1242
1243    - To enable OpenFlow 1.0, 1.1, 1.2, 1.3, 1.4, and 1.5 on bridge br0:
1244
1245      ovs-vsctl set bridge br0 protocols=OpenFlow10,OpenFlow11,OpenFlow12,OpenFlow13,OpenFlow14,OpenFlow15
1246
1247    - To enable only OpenFlow 1.0 on bridge br0:
1248
1249      ovs-vsctl set bridge br0 protocols=OpenFlow10
1250
1251    All current versions of ovs-ofctl enable only OpenFlow 1.0 by
1252    default.  Use the -O option to enable support for later versions of
1253    OpenFlow in ovs-ofctl.  For example:
1254
1255        ovs-ofctl -O OpenFlow13 dump-flows br0
1256
1257    (Open vSwitch 2.2 had an experimental implementation of OpenFlow
1258    1.4 that could cause crashes.  We don't recommend enabling it.)
1259
1260    [OPENFLOW-1.1+.md] in the Open vSwitch source tree tracks support for
1261    OpenFlow 1.1 and later features.  When support for OpenFlow 1.4 and
1262    1.5 is solidly implemented, Open vSwitch will enable those version
1263    by default.  Also, the OpenFlow 1.5 specification is still under
1264    development and thus subject to change.
1265
1266 ### Q: Does Open vSwitch support MPLS?
1267
1268 A: Before version 1.11, Open vSwitch did not support MPLS.  That is,
1269    these versions can match on MPLS Ethernet types, but they cannot
1270    match, push, or pop MPLS labels, nor can they look past MPLS labels
1271    into the encapsulated packet.
1272
1273    Open vSwitch versions 1.11, 2.0, and 2.1 have very minimal support
1274    for MPLS.  With the userspace datapath only, these versions can
1275    match, push, or pop a single MPLS label, but they still cannot look
1276    past MPLS labels (even after popping them) into the encapsulated
1277    packet.  Kernel datapath support is unchanged from earlier
1278    versions.
1279
1280    Open vSwitch version 2.3 can match, push, or pop a single MPLS
1281    label and look past the MPLS label into the encapsulated packet.
1282    Both userspace and kernel datapaths will be supported, but MPLS
1283    processing always happens in userspace either way, so kernel
1284    datapath performance will be disappointing.
1285
1286    Open vSwitch version 2.4 can match, push, or pop up to 3 MPLS
1287    labels and look past the MPLS label into the encapsulated packet.
1288    It will have kernel support for MPLS, yielding improved
1289    performance.
1290
1291 ### Q: I'm getting "error type 45250 code 0".  What's that?
1292
1293 A: This is a Open vSwitch extension to OpenFlow error codes.  Open
1294    vSwitch uses this extension when it must report an error to an
1295    OpenFlow controller but no standard OpenFlow error code is
1296    suitable.
1297
1298    Open vSwitch logs the errors that it sends to controllers, so the
1299    easiest thing to do is probably to look at the ovs-vswitchd log to
1300    find out what the error was.
1301
1302    If you want to dissect the extended error message yourself, the
1303    format is documented in include/openflow/nicira-ext.h in the Open
1304    vSwitch source distribution.  The extended error codes are
1305    documented in lib/ofp-errors.h.
1306
1307 Q1: Some of the traffic that I'd expect my OpenFlow controller to see
1308     doesn't actually appear through the OpenFlow connection, even
1309     though I know that it's going through.
1310 Q2: Some of the OpenFlow flows that my controller sets up don't seem
1311     to apply to certain traffic, especially traffic between OVS and
1312     the controller itself.
1313
1314 A: By default, Open vSwitch assumes that OpenFlow controllers are
1315    connected "in-band", that is, that the controllers are actually
1316    part of the network that is being controlled.  In in-band mode,
1317    Open vSwitch sets up special "hidden" flows to make sure that
1318    traffic can make it back and forth between OVS and the controllers.
1319    These hidden flows are higher priority than any flows that can be
1320    set up through OpenFlow, and they are not visible through normal
1321    OpenFlow flow table dumps.
1322
1323    Usually, the hidden flows are desirable and helpful, but
1324    occasionally they can cause unexpected behavior.  You can view the
1325    full OpenFlow flow table, including hidden flows, on bridge br0
1326    with the command:
1327
1328        ovs-appctl bridge/dump-flows br0
1329
1330    to help you debug.  The hidden flows are those with priorities
1331    greater than 65535 (the maximum priority that can be set with
1332    OpenFlow).
1333
1334    The DESIGN file at the top level of the Open vSwitch source
1335    distribution describes the in-band model in detail.
1336
1337    If your controllers are not actually in-band (e.g. they are on
1338    localhost via 127.0.0.1, or on a separate network), then you should
1339    configure your controllers in "out-of-band" mode.  If you have one
1340    controller on bridge br0, then you can configure out-of-band mode
1341    on it with:
1342
1343        ovs-vsctl set controller br0 connection-mode=out-of-band
1344
1345 ### Q: I configured all my controllers for out-of-band control mode but
1346    "ovs-appctl bridge/dump-flows" still shows some hidden flows.
1347
1348 A: You probably have a remote manager configured (e.g. with "ovs-vsctl
1349    set-manager").  By default, Open vSwitch assumes that managers need
1350    in-band rules set up on every bridge.  You can disable these rules
1351    on bridge br0 with:
1352
1353        ovs-vsctl set bridge br0 other-config:disable-in-band=true
1354
1355    This actually disables in-band control entirely for the bridge, as
1356    if all the bridge's controllers were configured for out-of-band
1357    control.
1358
1359 ### Q: My OpenFlow controller doesn't see the VLANs that I expect.
1360
1361 A: See answer under "VLANs", above.
1362
1363 ### Q: I ran "ovs-ofctl add-flow br0 nw_dst=192.168.0.1,actions=drop"
1364    but I got a funny message like this:
1365
1366        ofp_util|INFO|normalization changed ofp_match, details:
1367        ofp_util|INFO| pre: nw_dst=192.168.0.1
1368        ofp_util|INFO|post:
1369
1370    and when I ran "ovs-ofctl dump-flows br0" I saw that my nw_dst
1371    match had disappeared, so that the flow ends up matching every
1372    packet.
1373
1374 A: The term "normalization" in the log message means that a flow
1375    cannot match on an L3 field without saying what L3 protocol is in
1376    use.  The "ovs-ofctl" command above didn't specify an L3 protocol,
1377    so the L3 field match was dropped.
1378
1379    In this case, the L3 protocol could be IP or ARP.  A correct
1380    command for each possibility is, respectively:
1381
1382        ovs-ofctl add-flow br0 ip,nw_dst=192.168.0.1,actions=drop
1383
1384    and
1385
1386        ovs-ofctl add-flow br0 arp,nw_dst=192.168.0.1,actions=drop
1387
1388    Similarly, a flow cannot match on an L4 field without saying what
1389    L4 protocol is in use.  For example, the flow match "tp_src=1234"
1390    is, by itself, meaningless and will be ignored.  Instead, to match
1391    TCP source port 1234, write "tcp,tp_src=1234", or to match UDP
1392    source port 1234, write "udp,tp_src=1234".
1393
1394 ### Q: How can I figure out the OpenFlow port number for a given port?
1395
1396 A: The OFPT_FEATURES_REQUEST message requests an OpenFlow switch to
1397    respond with an OFPT_FEATURES_REPLY that, among other information,
1398    includes a mapping between OpenFlow port names and numbers.  From a
1399    command prompt, "ovs-ofctl show br0" makes such a request and
1400    prints the response for switch br0.
1401
1402    The Interface table in the Open vSwitch database also maps OpenFlow
1403    port names to numbers.  To print the OpenFlow port number
1404    associated with interface eth0, run:
1405
1406        ovs-vsctl get Interface eth0 ofport
1407
1408    You can print the entire mapping with:
1409
1410        ovs-vsctl -- --columns=name,ofport list Interface
1411
1412    but the output mixes together interfaces from all bridges in the
1413    database, so it may be confusing if more than one bridge exists.
1414
1415    In the Open vSwitch database, ofport value -1 means that the
1416    interface could not be created due to an error.  (The Open vSwitch
1417    log should indicate the reason.)  ofport value [] (the empty set)
1418    means that the interface hasn't been created yet.  The latter is
1419    normally an intermittent condition (unless ovs-vswitchd is not
1420    running).
1421
1422 ### Q: I added some flows with my controller or with ovs-ofctl, but when I
1423    run "ovs-dpctl dump-flows" I don't see them.
1424
1425 A: ovs-dpctl queries a kernel datapath, not an OpenFlow switch.  It
1426    won't display the information that you want.  You want to use
1427    "ovs-ofctl dump-flows" instead.
1428
1429 ### Q: It looks like each of the interfaces in my bonded port shows up
1430    as an individual OpenFlow port.  Is that right?
1431
1432 A: Yes, Open vSwitch makes individual bond interfaces visible as
1433    OpenFlow ports, rather than the bond as a whole.  The interfaces
1434    are treated together as a bond for only a few purposes:
1435
1436    - Sending a packet to the OFPP_NORMAL port.  (When an OpenFlow
1437      controller is not configured, this happens implicitly to
1438      every packet.)
1439
1440    - Mirrors configured for output to a bonded port.
1441
1442    It would make a lot of sense for Open vSwitch to present a bond as
1443    a single OpenFlow port.  If you want to contribute an
1444    implementation of such a feature, please bring it up on the Open
1445    vSwitch development mailing list at dev@openvswitch.org.
1446
1447 ### Q: I have a sophisticated network setup involving Open vSwitch, VMs or
1448    multiple hosts, and other components.  The behavior isn't what I
1449    expect.  Help!
1450
1451 A: To debug network behavior problems, trace the path of a packet,
1452    hop-by-hop, from its origin in one host to a remote host.  If
1453    that's correct, then trace the path of the response packet back to
1454    the origin.
1455
1456    Usually a simple ICMP echo request and reply ("ping") packet is
1457    good enough.  Start by initiating an ongoing "ping" from the origin
1458    host to a remote host.  If you are tracking down a connectivity
1459    problem, the "ping" will not display any successful output, but
1460    packets are still being sent.  (In this case the packets being sent
1461    are likely ARP rather than ICMP.)
1462
1463    Tools available for tracing include the following:
1464
1465    - "tcpdump" and "wireshark" for observing hops across network
1466      devices, such as Open vSwitch internal devices and physical
1467      wires.
1468
1469    - "ovs-appctl dpif/dump-flows <br>" in Open vSwitch 1.10 and
1470      later or "ovs-dpctl dump-flows <br>" in earlier versions.
1471      These tools allow one to observe the actions being taken on
1472      packets in ongoing flows.
1473
1474      See ovs-vswitchd(8) for "ovs-appctl dpif/dump-flows"
1475      documentation, ovs-dpctl(8) for "ovs-dpctl dump-flows"
1476      documentation, and "Why are there so many different ways to
1477      dump flows?" above for some background.
1478
1479    - "ovs-appctl ofproto/trace" to observe the logic behind how
1480      ovs-vswitchd treats packets.  See ovs-vswitchd(8) for
1481      documentation.  You can out more details about a given flow
1482      that "ovs-dpctl dump-flows" displays, by cutting and pasting
1483      a flow from the output into an "ovs-appctl ofproto/trace"
1484      command.
1485
1486    - SPAN, RSPAN, and ERSPAN features of physical switches, to
1487      observe what goes on at these physical hops.
1488
1489    Starting at the origin of a given packet, observe the packet at
1490    each hop in turn.  For example, in one plausible scenario, you
1491    might:
1492
1493    1. "tcpdump" the "eth" interface through which an ARP egresses
1494       a VM, from inside the VM.
1495
1496    2. "tcpdump" the "vif" or "tap" interface through which the ARP
1497       ingresses the host machine.
1498
1499    3. Use "ovs-dpctl dump-flows" to spot the ARP flow and observe
1500       the host interface through which the ARP egresses the
1501       physical machine.  You may need to use "ovs-dpctl show" to
1502       interpret the port numbers.  If the output seems surprising,
1503       you can use "ovs-appctl ofproto/trace" to observe details of
1504       how ovs-vswitchd determined the actions in the "ovs-dpctl
1505       dump-flows" output.
1506
1507    4. "tcpdump" the "eth" interface through which the ARP egresses
1508       the physical machine.
1509
1510    5. "tcpdump" the "eth" interface through which the ARP
1511       ingresses the physical machine, at the remote host that
1512       receives the ARP.
1513
1514    6. Use "ovs-dpctl dump-flows" to spot the ARP flow on the
1515       remote host that receives the ARP and observe the VM "vif"
1516       or "tap" interface to which the flow is directed.  Again,
1517       "ovs-dpctl show" and "ovs-appctl ofproto/trace" might help.
1518
1519    7. "tcpdump" the "vif" or "tap" interface to which the ARP is
1520       directed.
1521
1522    8. "tcpdump" the "eth" interface through which the ARP
1523       ingresses a VM, from inside the VM.
1524
1525    It is likely that during one of these steps you will figure out the
1526    problem.  If not, then follow the ARP reply back to the origin, in
1527    reverse.
1528
1529 ### Q: How do I make a flow drop packets?
1530
1531 A: To drop a packet is to receive it without forwarding it.  OpenFlow
1532    explicitly specifies forwarding actions.  Thus, a flow with an
1533    empty set of actions does not forward packets anywhere, causing
1534    them to be dropped.  You can specify an empty set of actions with
1535    "actions=" on the ovs-ofctl command line.  For example:
1536
1537        ovs-ofctl add-flow br0 priority=65535,actions=
1538
1539    would cause every packet entering switch br0 to be dropped.
1540
1541    You can write "drop" explicitly if you like.  The effect is the
1542    same.  Thus, the following command also causes every packet
1543    entering switch br0 to be dropped:
1544
1545        ovs-ofctl add-flow br0 priority=65535,actions=drop
1546
1547    "drop" is not an action, either in OpenFlow or Open vSwitch.
1548    Rather, it is only a way to say that there are no actions.
1549
1550 ### Q: I added a flow to send packets out the ingress port, like this:
1551
1552        ovs-ofctl add-flow br0 in_port=2,actions=2
1553
1554    but OVS drops the packets instead.
1555
1556 A: Yes, OpenFlow requires a switch to ignore attempts to send a packet
1557    out its ingress port.  The rationale is that dropping these packets
1558    makes it harder to loop the network.  Sometimes this behavior can
1559    even be convenient, e.g. it is often the desired behavior in a flow
1560    that forwards a packet to several ports ("floods" the packet).
1561
1562    Sometimes one really needs to send a packet out its ingress port
1563    ("hairpin"). In this case, output to OFPP_IN_PORT, which in
1564    ovs-ofctl syntax is expressed as just "in_port", e.g.:
1565
1566        ovs-ofctl add-flow br0 in_port=2,actions=in_port
1567
1568    This also works in some circumstances where the flow doesn't match
1569    on the input port.  For example, if you know that your switch has
1570    five ports numbered 2 through 6, then the following will send every
1571    received packet out every port, even its ingress port:
1572
1573        ovs-ofctl add-flow br0 actions=2,3,4,5,6,in_port
1574
1575    or, equivalently:
1576
1577        ovs-ofctl add-flow br0 actions=all,in_port
1578
1579    Sometimes, in complicated flow tables with multiple levels of
1580    "resubmit" actions, a flow needs to output to a particular port
1581    that may or may not be the ingress port.  It's difficult to take
1582    advantage of OFPP_IN_PORT in this situation.  To help, Open vSwitch
1583    provides, as an OpenFlow extension, the ability to modify the
1584    in_port field.  Whatever value is currently in the in_port field is
1585    the port to which outputs will be dropped, as well as the
1586    destination for OFPP_IN_PORT.  This means that the following will
1587    reliably output to port 2 or to ports 2 through 6, respectively:
1588
1589        ovs-ofctl add-flow br0 in_port=2,actions=load:0->NXM_OF_IN_PORT[],2
1590        ovs-ofctl add-flow br0 actions=load:0->NXM_OF_IN_PORT[],2,3,4,5,6
1591
1592    If the input port is important, then one may save and restore it on
1593    the stack:
1594
1595         ovs-ofctl add-flow br0 actions=push:NXM_OF_IN_PORT[],\
1596                                        load:0->NXM_OF_IN_PORT[],\
1597                                        2,3,4,5,6,\
1598                                        pop:NXM_OF_IN_PORT[]
1599
1600 ### Q: My bridge br0 has host 192.168.0.1 on port 1 and host 192.168.0.2
1601    on port 2.  I set up flows to forward only traffic destined to the
1602    other host and drop other traffic, like this:
1603
1604       priority=5,in_port=1,ip,nw_dst=192.168.0.2,actions=2
1605       priority=5,in_port=2,ip,nw_dst=192.168.0.1,actions=1
1606       priority=0,actions=drop
1607
1608    But it doesn't work--I don't get any connectivity when I do this.
1609    Why?
1610
1611 A: These flows drop the ARP packets that IP hosts use to establish IP
1612    connectivity over Ethernet.  To solve the problem, add flows to
1613    allow ARP to pass between the hosts:
1614
1615       priority=5,in_port=1,arp,actions=2
1616       priority=5,in_port=2,arp,actions=1
1617
1618    This issue can manifest other ways, too.  The following flows that
1619    match on Ethernet addresses instead of IP addresses will also drop
1620    ARP packets, because ARP requests are broadcast instead of being
1621    directed to a specific host:
1622
1623       priority=5,in_port=1,dl_dst=54:00:00:00:00:02,actions=2
1624       priority=5,in_port=2,dl_dst=54:00:00:00:00:01,actions=1
1625       priority=0,actions=drop
1626
1627    The solution already described above will also work in this case.
1628    It may be better to add flows to allow all multicast and broadcast
1629    traffic:
1630
1631       priority=5,in_port=1,dl_dst=01:00:00:00:00:00/01:00:00:00:00:00,actions=2
1632       priority=5,in_port=2,dl_dst=01:00:00:00:00:00/01:00:00:00:00:00,actions=1
1633
1634 ### Q: My bridge disconnects from my controller on add-port/del-port.
1635
1636 A: Reconfiguring your bridge can change your bridge's datapath-id because
1637    Open vSwitch generates datapath-id from the MAC address of one of its ports.
1638    In that case, Open vSwitch disconnects from controllers because there's
1639    no graceful way to notify controllers about the change of datapath-id.
1640
1641    To avoid the behaviour, you can configure datapath-id manually.
1642
1643       ovs-vsctl set bridge br0 other-config:datapath-id=0123456789abcdef
1644
1645 ### Q: My controller is getting errors about "buffers".  What's going on?
1646
1647 A: When a switch sends a packet to an OpenFlow controller using a
1648    "packet-in" message, it can also keep a copy of that packet in a
1649    "buffer", identified by a 32-bit integer "buffer_id".  There are
1650    two advantages to buffering.  First, when the controller wants to
1651    tell the switch to do something with the buffered packet (with a
1652    "packet-out" OpenFlow request), it does not need to send another
1653    copy of the packet back across the OpenFlow connection, which
1654    reduces the bandwidth cost of the connection and improves latency.
1655    This enables the second advantage: the switch can optionally send
1656    only the first part of the packet to the controller (assuming that
1657    the switch only needs to look at the first few bytes of the
1658    packet), further reducing bandwidth and improving latency.
1659
1660    However, buffering introduces some issues of its own.  First, any
1661    switch has limited resources, so if the controller does not use a
1662    buffered packet, the switch has to decide how long to keep it
1663    buffered.  When many packets are sent to a controller and buffered,
1664    Open vSwitch can discard buffered packets that the controller has
1665    not used after as little as 5 seconds.  This means that
1666    controllers, if they make use of packet buffering, should use the
1667    buffered packets promptly.  (This includes sending a "packet-out"
1668    with no actions if the controller does not want to do anything with
1669    a buffered packet, to clear the packet buffer and effectively
1670    "drop" its packet.)
1671
1672    Second, packet buffers are one-time-use, meaning that a controller
1673    cannot use a single packet buffer in two or more "packet-out"
1674    commands.  Open vSwitch will respond with an error to the second
1675    and subsequent "packet-out"s in such a case.
1676
1677    Finally, a common error early in controller development is to try
1678    to use buffer_id 0 in a "packet-out" message as if 0 represented
1679    "no buffered packet".  This is incorrect usage: the buffer_id with
1680    this meaning is actually 0xffffffff.
1681
1682    ovs-vswitchd(8) describes some details of Open vSwitch packet
1683    buffering that the OpenFlow specification requires implementations
1684    to document.
1685
1686
1687 Development
1688 -----------
1689
1690 ### Q: How do I implement a new OpenFlow message?
1691
1692 A: Add your new message to "enum ofpraw" and "enum ofptype" in
1693    lib/ofp-msgs.h, following the existing pattern.  Then recompile and
1694    fix all of the new warnings, implementing new functionality for the
1695    new message as needed.  (If you configure with --enable-Werror, as
1696    described in [INSTALL.md], then it is impossible to miss any warnings.)
1697
1698    If you need to add an OpenFlow vendor extension message for a
1699    vendor that doesn't yet have any extension messages, then you will
1700    also need to edit build-aux/extract-ofp-msgs.
1701
1702 ### Q: How do I add support for a new field or header?
1703
1704 A: Add new members for your field to "struct flow" in lib/flow.h, and
1705    add new enumerations for your new field to "enum mf_field_id" in
1706    lib/meta-flow.h, following the existing pattern.  Then recompile
1707    and fix all of the new warnings, implementing new functionality for
1708    the new field or header as needed.  (If you configure with
1709    --enable-Werror, as described in [INSTALL.md], then it is
1710    impossible to miss any warnings.)
1711
1712 ### Q: How do I add support for a new OpenFlow action?
1713
1714 A: Add your new action to "enum ofp_raw_action_type" in
1715    lib/ofp-actions.c, following the existing pattern.  Then recompile
1716    and fix all of the new warnings, implementing new functionality for
1717    the new action as needed.  (If you configure with --enable-Werror,
1718    as described in [INSTALL.md], then it is impossible to miss any
1719    warnings.)
1720
1721    If you need to add an OpenFlow vendor extension action for a vendor
1722    that doesn't yet have any extension actions, then you will also
1723    need to edit build-aux/extract-ofp-actions.
1724
1725
1726 Contact
1727 -------
1728
1729 bugs@openvswitch.org
1730 http://openvswitch.org/
1731
1732 [PORTING.md]:PORTING.md
1733 [WHY-OVS.md]:WHY-OVS.md
1734 [INSTALL.md]:INSTALL.md
1735 [OPENFLOW-1.1+.md]:OPENFLOW-1.1+.md