[cascardo/ovs.git] / ovn / ovn-sb.xml

<?xml version="1.0" encoding="utf-8"?>
<database name="ovn-sb" title="OVN Southbound Database">
  <p>
    This database holds logical and physical configuration and state for the
    Open Virtual Network (OVN) system to support virtual network abstraction.
    For an introduction to OVN, please see <code>ovn-architecture</code>(7).
  </p>

  <p>
    The OVN Southbound database sits at the center of the OVN
    architecture.  It is the one component that speaks both southbound
    directly to all the hypervisors and gateways, via
    <code>ovn-controller</code>/<code>ovn-controller-vtep</code>, and
    northbound to the Cloud Management System, via <code>ovn-northd</code>:
  </p>

  <h2>Database Structure</h2>

  <p>
    The OVN Southbound database contains classes of data with
    different properties, as described in the sections below.
  </p>

  <h3>Physical Network (PN) data</h3>

  <p>
    PN tables contain information about the chassis nodes in the system.  This
    contains all the information necessary to wire the overlay, such as IP
    addresses, supported tunnel types, and security keys.
  </p>

  <p>
    The amount of PN data is small (O(n) in the number of chassis) and it
    changes infrequently, so it can be replicated to every chassis.
  </p>

  <p>
    The <ref table="Chassis"/> table comprises the PN tables.
  </p>

  <h3>Logical Network (LN) data</h3>

  <p>
    LN tables contain the topology of logical switches and routers, ACLs,
    firewall rules, and everything needed to describe how packets traverse a
    logical network, represented as logical datapath flows (see Logical
    Datapath Flows, below).
  </p>

  <p>
    LN data may be large (O(n) in the number of logical ports, ACL rules,
    etc.).  Thus, to improve scaling, each chassis should receive only data
    related to logical networks in which that chassis participates.  Past
    experience shows that in the presence of large logical networks, even
    finer-grained partitioning of data, e.g. designing logical flows so that
    only the chassis hosting a logical port needs related flows, pays off
    scale-wise.  (This is not necessary initially but it is worth bearing in
    mind in the design.)
  </p>

  <p>
    The LN is a slave of the cloud management system running northbound of OVN.
    That CMS determines the entire OVN logical configuration and therefore the
    LN's content at any given time is a deterministic function of the CMS's
    configuration, although that happens indirectly via the
    <ref db="OVN_Northbound"/> database and <code>ovn-northd</code>.
  </p>

  <p>
    LN data is likely to change more quickly than PN data.  This is especially
    true in a container environment where VMs are created and destroyed (and
    therefore added to and deleted from logical switches) quickly.
  </p>

  <p>
    <ref table="Logical_Flow"/> and <ref table="Multicast_Group"/> contain LN
    data.
  </p>

  <h3>Logical-physical bindings</h3>

  <p>
    These tables link logical and physical components.  They show the current
    placement of logical components (such as VMs and VIFs) onto chassis, and
    map logical entities to the values that represent them in tunnel
    encapsulations.
  </p>

  <p>
    These tables change frequently, at least every time a VM powers up or down
    or migrates, and especially quickly in a container environment.  The
    amount of data per VM (or VIF) is small.
  </p>

  <p>
    Each chassis is authoritative about the VMs and VIFs that it hosts at any
    given time and can efficiently flood that state to a central location, so
    the consistency needs are minimal.
  </p>

  <p>
    The <ref table="Port_Binding"/> and <ref table="Datapath_Binding"/> tables
    contain binding data.
  </p>

  <h3>MAC bindings</h3>

  <p>
    The <ref table="MAC_Binding"/> table tracks the bindings from IP addresses
    to Ethernet addresses that are dynamically discovered using ARP (for IPv4)
    and neighbor discovery (for IPv6).  Usually, IP-to-MAC bindings for virtual
    machines are statically populated into the <ref table="Port_Binding"/>
    table, so <ref table="MAC_Binding"/> is primarily used to discover bindings
    on physical networks.
  </p>

  <h2>Common Columns</h2>

  <p>
    Some tables contain a special column named <code>external_ids</code>.  This
    column has the same form and purpose each place that it appears, so we
    describe it here to save space later.
  </p>

  <dl>
    <dt><code>external_ids</code>: map of string-string pairs</dt>
    <dd>
      Key-value pairs for use by the software that manages the OVN Southbound
      database rather than by
      <code>ovn-controller</code>/<code>ovn-controller-vtep</code>.  In
      particular, <code>ovn-northd</code> can use key-value pairs in this
      column to relate entities in the southbound database to higher-level
      entities (such as entities in the OVN Northbound database).  Individual
      key-value pairs in this column may be documented in some cases to aid
      in understanding and troubleshooting, but the reader should not mistake
      such documentation as comprehensive.
    </dd>
  </dl>

  <table name="Chassis" title="Physical Network Hypervisor and Gateway Information">
    <p>
      Each row in this table represents a hypervisor or gateway (a chassis) in
      the physical network (PN).  Each chassis, via
      <code>ovn-controller</code>/<code>ovn-controller-vtep</code>, adds
      and updates its own row, and keeps a copy of the remaining rows to
      determine how to reach other hypervisors.
    </p>

    <p>
      When a chassis shuts down gracefully, it should remove its own row.
      (This is not critical because resources hosted on the chassis are equally
      unreachable regardless of whether the row is present.)  If a chassis
      shuts down permanently without removing its row, some kind of manual or
      automatic cleanup is eventually needed; we can devise a process for that
      as necessary.
    </p>

    <column name="name">
      OVN does not prescribe a particular format for chassis names.
      ovn-controller populates this column using <ref key="system-id"
      table="Open_vSwitch" column="external_ids" db="Open_vSwitch"/>
      in the Open_vSwitch database's <ref table="Open_vSwitch"
      db="Open_vSwitch"/> table.  ovn-controller-vtep populates this
      column with <ref table="Physical_Switch" column="name"
      db="hardware_vtep"/> in the hardware_vtep database's
      <ref table="Physical_Switch" db="hardware_vtep"/> table.
    </column>

    <column name="hostname">
      The hostname of the chassis, if applicable.  ovn-controller will populate
      this column with the hostname of the host it is running on.
      ovn-controller-vtep will leave this column empty.
    </column>

    <column name="external_ids" key="ovn-bridge-mappings">
      <code>ovn-controller</code> populates this key with the set of bridge
      mappings it has been configured to use.  Other applications should treat
      this key as read-only.  See <code>ovn-controller</code>(8) for more
      information.
    </column>

    <group title="Common Columns">
      The overall purpose of these columns is described under <code>Common
      Columns</code> at the beginning of this document.

      <column name="external_ids"/>
    </group>

    <group title="Encapsulation Configuration">
      <p>
        OVN uses encapsulation to transmit logical dataplane packets
        between chassis.
      </p>

      <column name="encaps">
        Points to supported encapsulation configurations to transmit
        logical dataplane packets to this chassis.  Each entry is a <ref
        table="Encap"/> record that describes the configuration.
      </column>
    </group>

     <group title="Gateway Configuration">
       <p>
        A <dfn>gateway</dfn> is a chassis that forwards traffic between the
        OVN-managed part of a logical network and a physical VLAN, extending a
        tunnel-based logical network into a physical network.  Gateways are
        typically dedicated nodes that do not host VMs and will be controlled
        by <code>ovn-controller-vtep</code>.
      </p>

      <column name="vtep_logical_switches">
        Stores all VTEP logical switch names connected by this gateway
        chassis.  The <ref table="Port_Binding"/> table entry with
        <ref column="options" table="Port_Binding"/>:<code>vtep-physical-switch</code>
        equal <ref table="Chassis"/> <ref column="name" table="Chassis"/>, and
        <ref column="options" table="Port_Binding"/>:<code>vtep-logical-switch</code>
        value in <ref table="Chassis"/>
        <ref column="vtep_logical_switches" table="Chassis"/>, will be
        associated with this <ref table="Chassis"/>.
      </column>
     </group>
  </table>

  <table name="Encap" title="Encapsulation Types">
    <p>
      The <ref column="encaps" table="Chassis"/> column in the <ref
      table="Chassis"/> table refers to rows in this table to identify
      how OVN may transmit logical dataplane packets to this chassis.
      Each chassis, via <code>ovn-controller</code>(8) or
      <code>ovn-controller-vtep</code>(8), adds and updates its own rows
      and keeps a copy of the remaining rows to determine how to reach
      other chassis.
    </p>

    <column name="type">
      The encapsulation to use to transmit packets to this chassis.
      Hypervisors must use either <code>geneve</code> or
      <code>stt</code>.  Gateways may use <code>vxlan</code>,
      <code>geneve</code>, or <code>stt</code>.
    </column>

    <column name="options">
      Options for configuring the encapsulation, e.g. IPsec parameters when
      IPsec support is introduced.  No options are currently defined.
    </column>

    <column name="ip">
      The IPv4 address of the encapsulation tunnel endpoint.
    </column>
  </table>

  <table name="Address_Set" title="Address Sets">
    <p>
      See the documentation for the <ref table="Address_Set"
      db="OVN_Northbound"/> table in the <ref db="OVN_Northbound"/> database
      for details.
    </p>

    <column name="name"/>
    <column name="addresses"/>
  </table>

  <table name="Logical_Flow" title="Logical Network Flows">
    <p>
      Each row in this table represents one logical flow.
      <code>ovn-northd</code> populates this table with logical flows
      that implement the L2 and L3 topologies specified in the
      <ref db="OVN_Northbound"/> database.  Each hypervisor, via
      <code>ovn-controller</code>, translates the logical flows into
      OpenFlow flows specific to its hypervisor and installs them into
      Open vSwitch.
    </p>

    <p>
      Logical flows are expressed in an OVN-specific format, described here.  A
      logical datapath flow is much like an OpenFlow flow, except that the
      flows are written in terms of logical ports and logical datapaths instead
      of physical ports and physical datapaths.  Translation between logical
      and physical flows helps to ensure isolation between logical datapaths.
      (The logical flow abstraction also allows the OVN centralized
      components to do less work, since they do not have to separately
      compute and push out physical flows to each chassis.)
    </p>

    <p>
      The default action when no flow matches is to drop packets.
    </p>

    <p><em>Architectural Logical Life Cycle of a Packet</em></p>

    <p>
      This following description focuses on the life cycle of a packet through
      a logical datapath, ignoring physical details of the implementation.
      Please refer to <em>Architectural Physical Life Cycle of a Packet</em> in
      <code>ovn-architecture</code>(7) for the physical information.
    </p>

    <p>
      The description here is written as if OVN itself executes these steps,
      but in fact OVN (that is, <code>ovn-controller</code>) programs Open
      vSwitch, via OpenFlow and OVSDB, to execute them on its behalf.
    </p>

    <p>
      At a high level, OVN passes each packet through the logical datapath's
      logical ingress pipeline, which may output the packet to one or more
      logical port or logical multicast groups.  For each such logical output
      port, OVN passes the packet through the datapath's logical egress
      pipeline, which may either drop the packet or deliver it to the
      destination.  Between the two pipelines, outputs to logical multicast
      groups are expanded into logical ports, so that the egress pipeline only
      processes a single logical output port at a time.  Between the two
      pipelines is also where, when necessary, OVN encapsulates a packet in a
      tunnel (or tunnels) to transmit to remote hypervisors.
    </p>

    <p>
      In more detail, to start, OVN searches the <ref table="Logical_Flow"/>
      table for a row with correct <ref column="logical_datapath"/>, a <ref
      column="pipeline"/> of <code>ingress</code>, a <ref column="table_id"/>
      of 0, and a <ref column="match"/> that is true for the packet.  If none
      is found, OVN drops the packet.  If OVN finds more than one, it chooses
      the match with the highest <ref column="priority"/>.  Then OVN executes
      each of the actions specified in the row's <ref table="actions"/> column,
      in the order specified.  Some actions, such as those to modify packet
      headers, require no further details.  The <code>next</code> and
      <code>output</code> actions are special.
    </p>

    <p>
      The <code>next</code> action causes the above process to be repeated
      recursively, except that OVN searches for <ref column="table_id"/> of 1
      instead of 0.  Similarly, any <code>next</code> action in a row found in
      that table would cause a further search for a <ref column="table_id"/> of
      2, and so on.  When recursive processing completes, flow control returns
      to the action following <code>next</code>.
    </p>

    <p>
      The <code>output</code> action also introduces recursion.  Its effect
      depends on the current value of the <code>outport</code> field.  Suppose
      <code>outport</code> designates a logical port.  First, OVN compares
      <code>inport</code> to <code>outport</code>; if they are equal, it treats
      the <code>output</code> as a no-op.  In the common case, where they are
      different, the packet enters the egress pipeline.  This transition to the
      egress pipeline discards register data, e.g. <code>reg0</code> ...
      <code>reg9</code> and connection tracking state, to achieve
      uniform behavior regardless of whether the egress pipeline is on a
      different hypervisor (because registers aren't preserve across
      tunnel encapsulation).
    </p>

    <p>
      To execute the egress pipeline, OVN again searches the <ref
      table="Logical_Flow"/> table for a row with correct <ref
      column="logical_datapath"/>, a <ref column="table_id"/> of 0, a <ref
      column="match"/> that is true for the packet, but now looking for a <ref
      column="pipeline"/> of <code>egress</code>.  If no matching row is found,
      the output becomes a no-op.  Otherwise, OVN executes the actions for the
      matching flow (which is chosen from multiple, if necessary, as already
      described).
    </p>

    <p>
      In the <code>egress</code> pipeline, the <code>next</code> action acts as
      already described, except that it, of course, searches for
      <code>egress</code> flows.  The <code>output</code> action, however, now
      directly outputs the packet to the output port (which is now fixed,
      because <code>outport</code> is read-only within the egress pipeline).
    </p>

    <p>
      The description earlier assumed that <code>outport</code> referred to a
      logical port.  If it instead designates a logical multicast group, then
      the description above still applies, with the addition of fan-out from
      the logical multicast group to each logical port in the group.  For each
      member of the group, OVN executes the logical pipeline as described, with
      the logical output port replaced by the group member.
    </p>

    <p><em>Pipeline Stages</em></p>

    <p>
      <code>ovn-northd</code> is responsible for populating the
      <ref table="Logical_Flow"/> table, so the stages are an
      implementation detail and subject to change.  This section
      describes the current logical flow table.
    </p>

    <p>
      The ingress pipeline consists of the following stages:
    </p>
    <ul>
      <li>
        Port Security (Table 0): Validates the source address, drops
        packets with a VLAN tag, and, if configured, verifies that the
        logical port is allowed to send with the source address.
      </li>

      <li>
        L2 Destination Lookup (Table 1): Forwards known unicast
        addresses to the appropriate logical port.  Unicast packets to
        unknown hosts are forwarded to logical ports configured with the
        special <code>unknown</code> mac address.  Broadcast, and
        multicast are flooded to all ports in the logical switch.
      </li>
    </ul>

    <p>
      The egress pipeline consists of the following stages:
    </p>
    <ul>
      <li>
        ACL (Table 0): Applies any specified access control lists.
      </li>

      <li>
        Port Security (Table 1): If configured, verifies that the
        logical port is allowed to receive packets with the destination
        address.
      </li>
    </ul>

    <column name="logical_datapath">
      The logical datapath to which the logical flow belongs.
    </column>

    <column name="pipeline">
      <p>
        The primary flows used for deciding on a packet's destination are the
        <code>ingress</code> flows.  The <code>egress</code> flows implement
        ACLs.  See <em>Logical Life Cycle of a Packet</em>, above, for details.
      </p>
    </column>

    <column name="table_id">
      The stage in the logical pipeline, analogous to an OpenFlow table number.
    </column>

    <column name="priority">
      The flow's priority.  Flows with numerically higher priority take
      precedence over those with lower.  If two logical datapath flows with the
      same priority both match, then the one actually applied to the packet is
      undefined.
    </column>

    <column name="match">
      <p>
        A matching expression.  OVN provides a superset of OpenFlow matching
        capabilities, using a syntax similar to Boolean expressions in a
        programming language.
      </p>

      <p>
        The most important components of match expression are
        <dfn>comparisons</dfn> between <dfn>symbols</dfn> and
        <dfn>constants</dfn>, e.g. <code>ip4.dst == 192.168.0.1</code>,
        <code>ip.proto == 6</code>, <code>arp.op == 1</code>, <code>eth.type ==
        0x800</code>.  The logical AND operator <code>&amp;&amp;</code> and
        logical OR operator <code>||</code> can combine comparisons into a
        larger expression.
      </p>

      <p>
        Matching expressions also support parentheses for grouping, the logical
        NOT prefix operator <code>!</code>, and literals <code>0</code> and
        <code>1</code> to express ``false'' or ``true,'' respectively.  The
        latter is useful by itself as a catch-all expression that matches every
        packet.
      </p>

      <p><em>Symbols</em></p>

      <p>
        <em>Type</em>.  Symbols have <dfn>integer</dfn> or <dfn>string</dfn>
        type.  Integer symbols have a <dfn>width</dfn> in bits.
      </p>

      <p>
        <em>Kinds</em>.  There are three kinds of symbols:
      </p>

      <ul>
        <li>
          <p>
            <dfn>Fields</dfn>.  A field symbol represents a packet header or
            metadata field.  For example, a field
            named <code>vlan.tci</code> might represent the VLAN TCI field in a
            packet.
          </p>

          <p>
            A field symbol can have integer or string type.  Integer fields can
            be nominal or ordinal (see <em>Level of Measurement</em>,
            below).
          </p>
        </li>

        <li>
          <p>
            <dfn>Subfields</dfn>.  A subfield represents a subset of bits from
            a larger field.  For example, a field <code>vlan.vid</code> might
            be defined as an alias for <code>vlan.tci[0..11]</code>.  Subfields
            are provided for syntactic convenience, because it is always
            possible to instead refer to a subset of bits from a field
            directly.
          </p>

          <p>
            Only ordinal fields (see <em>Level of Measurement</em>,
            below) may have subfields.  Subfields are always ordinal.
          </p>
        </li>

        <li>
          <p>
            <dfn>Predicates</dfn>.  A predicate is shorthand for a Boolean
            expression.  Predicates may be used much like 1-bit fields.  For
            example, <code>ip4</code> might expand to <code>eth.type ==
            0x800</code>.  Predicates are provided for syntactic convenience,
            because it is always possible to instead specify the underlying
            expression directly.
          </p>

          <p>
            A predicate whose expansion refers to any nominal field or
            predicate (see <em>Level of Measurement</em>, below) is nominal;
            other predicates have Boolean level of measurement.
          </p>
        </li>
      </ul>

      <p>
        <em>Level of Measurement</em>.  See
        http://en.wikipedia.org/wiki/Level_of_measurement for the statistical
        concept on which this classification is based.  There are three
        levels:
      </p>

      <ul>
        <li>
          <p>
            <dfn>Ordinal</dfn>.  In statistics, ordinal values can be ordered
            on a scale.  OVN considers a field (or subfield) to be ordinal if
            its bits can be examined individually.  This is true for the
            OpenFlow fields that OpenFlow or Open vSwitch makes ``maskable.''
          </p>

          <p>
            Any use of a nominal field may specify a single bit or a range of
            bits, e.g. <code>vlan.tci[13..15]</code> refers to the PCP field
            within the VLAN TCI, and <code>eth.dst[40]</code> refers to the
            multicast bit in the Ethernet destination address.
          </p>

          <p>
            OVN supports all the usual arithmetic relations (<code>==</code>,
            <code>!=</code>, <code>&lt;</code>, <code>&lt;=</code>,
            <code>&gt;</code>, and <code>&gt;=</code>) on ordinal fields and
            their subfields, because OVN can implement these in OpenFlow and
            Open vSwitch as collections of bitwise tests.
          </p>
        </li>

        <li>
          <p>
            <dfn>Nominal</dfn>.  In statistics, nominal values cannot be
            usefully compared except for equality.  This is true of OpenFlow
            port numbers, Ethernet types, and IP protocols are examples: all of
            these are just identifiers assigned arbitrarily with no deeper
            meaning.  In OpenFlow and Open vSwitch, bits in these fields
            generally aren't individually addressable.
          </p>

          <p>
            OVN only supports arithmetic tests for equality on nominal fields,
            because OpenFlow and Open vSwitch provide no way for a flow to
            efficiently implement other comparisons on them.  (A test for
            inequality can be sort of built out of two flows with different
            priorities, but OVN matching expressions always generate flows with
            a single priority.)
          </p>

          <p>
            String fields are always nominal.
          </p>
        </li>

        <li>
          <p>
            <dfn>Boolean</dfn>.  A nominal field that has only two values, 0
            and 1, is somewhat exceptional, since it is easy to support both
            equality and inequality tests on such a field: either one can be
            implemented as a test for 0 or 1.
          </p>

          <p>
            Only predicates (see above) have a Boolean level of measurement.
          </p>

          <p>
            This isn't a standard level of measurement.
          </p>
        </li>
      </ul>

      <p>
        <em>Prerequisites</em>.  Any symbol can have prerequisites, which are
        additional condition implied by the use of the symbol.  For example,
        For example, <code>icmp4.type</code> symbol might have prerequisite
        <code>icmp4</code>, which would cause an expression <code>icmp4.type ==
        0</code> to be interpreted as <code>icmp4.type == 0 &amp;&amp;
        icmp4</code>, which would in turn expand to <code>icmp4.type == 0
        &amp;&amp; eth.type == 0x800 &amp;&amp; ip4.proto == 1</code> (assuming
        <code>icmp4</code> is a predicate defined as suggested under
        <em>Types</em> above).
      </p>

      <p><em>Relational operators</em></p>

      <p>
        All of the standard relational operators <code>==</code>,
        <code>!=</code>, <code>&lt;</code>, <code>&lt;=</code>,
        <code>&gt;</code>, and <code>&gt;=</code> are supported.  Nominal
        fields support only <code>==</code> and <code>!=</code>, and only in a
        positive sense when outer <code>!</code> are taken into account,
        e.g. given string field <code>inport</code>, <code>inport ==
        "eth0"</code> and <code>!(inport != "eth0")</code> are acceptable, but
        not <code>inport != "eth0"</code>.
      </p>

      <p>
        The implementation of <code>==</code> (or <code>!=</code> when it is
        negated), is more efficient than that of the other relational
        operators.
      </p>

      <p><em>Constants</em></p>

      <p>
        Integer constants may be expressed in decimal, hexadecimal prefixed by
        <code>0x</code>, or as dotted-quad IPv4 addresses, IPv6 addresses in
        their standard forms, or Ethernet addresses as colon-separated hex
        digits.  A constant in any of these forms may be followed by a slash
        and a second constant (the mask) in the same form, to form a masked
        constant.  IPv4 and IPv6 masks may be given as integers, to express
        CIDR prefixes.
      </p>

      <p>
        String constants have the same syntax as quoted strings in JSON (thus,
        they are Unicode strings).
      </p>

      <p>
        Some operators support sets of constants written inside curly braces
        <code>{</code> ... <code>}</code>.  Commas between elements of a set,
        and after the last elements, are optional.  With <code>==</code>,
        ``<code><var>field</var> == { <var>constant1</var>,
        <var>constant2</var>,</code> ... <code>}</code>'' is syntactic sugar
        for ``<code><var>field</var> == <var>constant1</var> ||
        <var>field</var> == <var>constant2</var> || </code>...<code></code>.
        Similarly, ``<code><var>field</var> != { <var>constant1</var>,
        <var>constant2</var>, </code>...<code> }</code>'' is equivalent to
        ``<code><var>field</var> != <var>constant1</var> &amp;&amp;
        <var>field</var> != <var>constant2</var> &amp;&amp;
        </code>...<code></code>''.
      </p>

      <p>
        You may refer to a set of IPv4, IPv6, or MAC addresses stored in the
        <ref table="Address_Set"/> table by its <ref column="name"
        table="Address_Set"/>.  An <ref table="Address_Set"/> with a name
        of <code>set1</code> can be referred to as
        <code>$set1</code>.
      </p>

      <p><em>Miscellaneous</em></p>

      <p>
        Comparisons may name the symbol or the constant first,
        e.g. <code>tcp.src == 80</code> and <code>80 == tcp.src</code> are both
        acceptable.
      </p>

      <p>
        Tests for a range may be expressed using a syntax like <code>1024 &lt;=
        tcp.src &lt;= 49151</code>, which is equivalent to <code>1024 &lt;=
        tcp.src &amp;&amp; tcp.src &lt;= 49151</code>.
      </p>

      <p>
        For a one-bit field or predicate, a mention of its name is equivalent
        to <code><var>symobl</var> == 1</code>, e.g. <code>vlan.present</code>
        is equivalent to <code>vlan.present == 1</code>.  The same is true for
        one-bit subfields, e.g. <code>vlan.tci[12]</code>.  There is no
        technical limitation to implementing the same for ordinal fields of all
        widths, but the implementation is expensive enough that the syntax
        parser requires writing an explicit comparison against zero to make
        mistakes less likely, e.g. in <code>tcp.src != 0</code> the comparison
        against 0 is required.
      </p>

      <p>
        <em>Operator precedence</em> is as shown below, from highest to lowest.
        There are two exceptions where parentheses are required even though the
        table would suggest that they are not: <code>&amp;&amp;</code> and
        <code>||</code> require parentheses when used together, and
        <code>!</code> requires parentheses when applied to a relational
        expression.  Thus, in <code>(eth.type == 0x800 || eth.type == 0x86dd)
        &amp;&amp; ip.proto == 6</code> or <code>!(arp.op == 1)</code>, the
        parentheses are mandatory.
      </p>

      <ul>
        <li><code>()</code></li>
        <li><code>==   !=   &lt;   &lt;=   &gt;   &gt;=</code></li>
        <li><code>!</code></li>
        <li><code>&amp;&amp;   ||</code></li>
      </ul>

      <p>
        <em>Comments</em> may be introduced by <code>//</code>, which extends
        to the next new-line.  Comments within a line may be bracketed by
        <code>/*</code> and <code>*/</code>.  Multiline comments are not
        supported.
      </p>

      <p><em>Symbols</em></p>

      <p>
        Most of the symbols below have integer type.  Only <code>inport</code>
        and <code>outport</code> have string type.  <code>inport</code> names a
        logical port.  Thus, its value is a <ref column="logical_port"/> name
        from the <ref table="Port_Binding"/> table.  <code>outport</code> may
        name a logical port, as <code>inport</code>, or a logical multicast
        group defined in the <ref table="Multicast_Group"/> table.  For both
        symbols, only names within the flow's logical datapath may be used.
      </p>

      <p>
        The <code>reg</code><var>X</var> symbols are 32-bit integers.
        The <code>xxreg</code><var>X</var> symbols are 128-bit integers,
        which overlay four of the 32-bit registers: <code>xxreg0</code>
        overlays <code>reg0</code> through <code>reg3</code>, with
        <code>reg0</code> supplying the most-significant bits of
        <code>xxreg0</code> and <code>reg3</code> the least-signficant.
        <code>xxreg1</code> similarly overlays <code>reg4</code> through
        <code>reg7</code>.
      </p>

      <ul>
        <li><code>reg0</code>...<code>reg9</code></li>
        <li><code>xxreg0</code> <code>xxreg1</code></li>
        <li><code>inport</code> <code>outport</code></li>
        <li><code>eth.src</code> <code>eth.dst</code> <code>eth.type</code></li>
        <li><code>vlan.tci</code> <code>vlan.vid</code> <code>vlan.pcp</code> <code>vlan.present</code></li>
        <li><code>ip.proto</code> <code>ip.dscp</code> <code>ip.ecn</code> <code>ip.ttl</code> <code>ip.frag</code></li>
        <li><code>ip4.src</code> <code>ip4.dst</code></li>
        <li><code>ip6.src</code> <code>ip6.dst</code> <code>ip6.label</code></li>
        <li><code>arp.op</code> <code>arp.spa</code> <code>arp.tpa</code> <code>arp.sha</code> <code>arp.tha</code></li>
        <li><code>tcp.src</code> <code>tcp.dst</code> <code>tcp.flags</code></li>
        <li><code>udp.src</code> <code>udp.dst</code></li>
        <li><code>sctp.src</code> <code>sctp.dst</code></li>
        <li><code>icmp4.type</code> <code>icmp4.code</code></li>
        <li><code>icmp6.type</code> <code>icmp6.code</code></li>
        <li><code>nd.target</code> <code>nd.sll</code> <code>nd.tll</code></li>
        <li><code>ct_mark</code> <code>ct_label</code></li>
        <li>
          <p>
            <code>ct_state</code>, which has the following Boolean subfields:
          </p>
          <ul>
            <li><code>ct.new</code>: True for a new flow</li>
            <li><code>ct.est</code>: True for an established flow</li>
            <li><code>ct.rel</code>: True for a related flow</li>
            <li><code>ct.rpl</code>: True for a reply flow</li>
            <li><code>ct.inv</code>: True for a connection entry in a bad state</li>
          </ul>
          <p>
            <code>ct_state</code> and its subfields are initialized by the
            <code>ct_next</code> action, described below.
          </p>
        </li>
      </ul>

      <p>
        The following predicates are supported:
      </p>

      <ul>
        <li><code>eth.bcast</code> expands to <code>eth.dst == ff:ff:ff:ff:ff:ff</code></li>
        <li><code>eth.mcast</code> expands to <code>eth.dst[40]</code></li>
        <li><code>vlan.present</code> expands to <code>vlan.tci[12]</code></li>
        <li><code>ip4</code> expands to <code>eth.type == 0x800</code></li>
        <li><code>ip4.mcast</code> expands to <code>ip4.dst[28..31] == 0xe</code></li>
        <li><code>ip6</code> expands to <code>eth.type == 0x86dd</code></li>
        <li><code>ip</code> expands to <code>ip4 || ip6</code></li>
        <li><code>icmp4</code> expands to <code>ip4 &amp;&amp; ip.proto == 1</code></li>
        <li><code>icmp6</code> expands to <code>ip6 &amp;&amp; ip.proto == 58</code></li>
        <li><code>icmp</code> expands to <code>icmp4 || icmp6</code></li>
        <li><code>ip.is_frag</code> expands to <code>ip.frag[0]</code></li>
        <li><code>ip.later_frag</code> expands to <code>ip.frag[1]</code></li>
        <li><code>ip.first_frag</code> expands to <code>ip.is_frag &amp;&amp; !ip.later_frag</code></li>
        <li><code>arp</code> expands to <code>eth.type == 0x806</code></li>
        <li><code>nd</code> expands to <code>icmp6.type == {135, 136} &amp;&amp; icmp6.code == 0</code></li>
        <li><code>tcp</code> expands to <code>ip.proto == 6</code></li>
        <li><code>udp</code> expands to <code>ip.proto == 17</code></li>
        <li><code>sctp</code> expands to <code>ip.proto == 132</code></li>
      </ul>
    </column>

    <column name="actions">
      <p>
        Logical datapath actions, to be executed when the logical flow
        represented by this row is the highest-priority match.
      </p>

      <p>
        Actions share lexical syntax with the <ref column="match"/> column.  An
        empty set of actions (or one that contains just white space or
        comments), or a set of actions that consists of just
        <code>drop;</code>, causes the matched packets to be dropped.
        Otherwise, the column should contain a sequence of actions, each
        terminated by a semicolon.
      </p>

      <p>
        The following actions are defined:
      </p>

      <dl>
        <dt><code>output;</code></dt>
        <dd>
          <p>
            In the ingress pipeline, this action executes the
            <code>egress</code> pipeline as a subroutine.  If
            <code>outport</code> names a logical port, the egress pipeline
            executes once; if it is a multicast group, the egress pipeline runs
            once for each logical port in the group.
          </p>

          <p>
            In the egress pipeline, this action performs the actual
            output to the <code>outport</code> logical port.  (In the egress
            pipeline, <code>outport</code> never names a multicast group.)
          </p>

          <p>
            Output to the input port is implicitly dropped, that is,
            <code>output</code> becomes a no-op if <code>outport</code> ==
            <code>inport</code>.  Occasionally it may be useful to override
            this behavior, e.g. to send an ARP reply to an ARP request; to do
            so, use <code>inport = "";</code> to set the logical input port to
            an empty string (which should not be used as the name of any
            logical port).
          </p>
        </dd>

        <dt><code>next;</code></dt>
        <dt><code>next(<var>table</var>);</code></dt>
        <dd>
          Executes another logical datapath table as a subroutine.  By default,
          the table after the current one is executed.  Specify
          <var>table</var> to jump to a specific table in the same pipeline.
        </dd>

        <dt><code><var>field</var> = <var>constant</var>;</code></dt>
        <dd>
          <p>
            Sets data or metadata field <var>field</var> to constant value
            <var>constant</var>, e.g. <code>outport = "vif0";</code> to set the
            logical output port.  To set only a subset of bits in a field,
            specify a subfield for <var>field</var> or a masked
            <var>constant</var>, e.g. one may use <code>vlan.pcp[2] = 1;</code>
            or <code>vlan.pcp = 4/4;</code> to set the most sigificant bit of
            the VLAN PCP.
          </p>

          <p>
            Assigning to a field with prerequisites implicitly adds those
            prerequisites to <ref column="match"/>; thus, for example, a flow
            that sets <code>tcp.dst</code> applies only to TCP flows,
            regardless of whether its <ref column="match"/> mentions any TCP
            field.
          </p>

          <p>
            Not all fields are modifiable (e.g. <code>eth.type</code> and
            <code>ip.proto</code> are read-only), and not all modifiable fields
            may be partially modified (e.g. <code>ip.ttl</code> must assigned
            as a whole).  The <code>outport</code> field is modifiable in the
            <code>ingress</code> pipeline but not in the <code>egress</code>
            pipeline.
          </p>
        </dd>

        <dt><code><var>field1</var> = <var>field2</var>;</code></dt>
        <dd>
          <p>
            Sets data or metadata field <var>field1</var> to the value of data
            or metadata field <var>field2</var>, e.g. <code>reg0 =
            ip4.src;</code> copies <code>ip4.src</code> into <code>reg0</code>.
            To modify only a subset of a field's bits, specify a subfield for
            <var>field1</var> or <var>field2</var> or both, e.g. <code>vlan.pcp
            = reg0[0..2];</code> copies the least-significant bits of
            <code>reg0</code> into the VLAN PCP.
          </p>

          <p>
            <var>field1</var> and <var>field2</var> must be the same type,
            either both string or both integer fields.  If they are both
            integer fields, they must have the same width.
          </p>

          <p>
            If <var>field1</var> or <var>field2</var> has prerequisites, they
            are added implicitly to <ref column="match"/>.  It is possible to
            write an assignment with contradictory prerequisites, such as
            <code>ip4.src = ip6.src[0..31];</code>, but the contradiction means
            that a logical flow with such an assignment will never be matched.
          </p>
        </dd>

        <dt><code><var>field1</var> &lt;-&gt; <var>field2</var>;</code></dt>
        <dd>
          <p>
            Similar to <code><var>field1</var> = <var>field2</var>;</code>
            except that the two values are exchanged instead of copied.  Both
            <var>field1</var> and <var>field2</var> must modifiable.
          </p>
        </dd>

        <dt><code>ip.ttl--;</code></dt>
        <dd>
          <p>
            Decrements the IPv4 or IPv6 TTL.  If this would make the TTL zero
            or negative, then processing of the packet halts; no further
            actions are processed.  (To properly handle such cases, a
            higher-priority flow should match on
            <code>ip.ttl == {0, 1};</code>.)
          </p>

          <p><b>Prerequisite:</b> <code>ip</code></p>
        </dd>

        <dt><code>ct_next;</code></dt>
        <dd>
          <p>
            Apply connection tracking to the flow, initializing
            <code>ct_state</code> for matching in later tables.
            Automatically moves on to the next table, as if followed by
            <code>next</code>.
          </p>

          <p>
            As a side effect, IP fragments will be reassembled for matching.
            If a fragmented packet is output, then it will be sent with any
            overlapping fragments squashed.  The connection tracking state is
            scoped by the logical port, so overlapping addresses may be used.
            To allow traffic related to the matched flow, execute
            <code>ct_commit</code>.
          </p>

          <p>
            It is possible to have actions follow <code>ct_next</code>,
            but they will not have access to any of its side-effects and
            is not generally useful.
          </p>
        </dd>

        <dt><code>ct_commit;</code></dt>
        <dt><code>ct_commit(ct_mark=<var>value[/mask]</var>);</code></dt>
        <dt><code>ct_commit(ct_label=<var>value[/mask]</var>);</code></dt>
        <dt><code>ct_commit(ct_mark=<var>value[/mask]</var>, ct_label=<var>value[/mask]</var>);</code></dt>
        <dd>
          <p>
            Commit the flow to the connection tracking entry associated with it
            by a previous call to <code>ct_next</code>.  When
            <code>ct_mark=<var>value[/mask]</var></code> and/or
            <code>ct_label=<var>value[/mask]</var></code> are supplied,
            <code>ct_mark</code> and/or <code>ct_label</code> will be set to the
            values indicated by <var>value[/mask]</var> on the connection
            tracking entry. <code>ct_mark</code> is a 32-bit field.
            <code>ct_label</code> is a 128-bit field. The <var>value[/mask]</var>
            should be specified in hex string if more than 64bits are to be used.
          </p>

          <p>
            Note that if you want processing to continue in the next table,
            you must execute the <code>next</code> action after
            <code>ct_commit</code>.  You may also leave out <code>next</code>
            which will commit connection tracking state, and then drop the
            packet.  This could be useful for setting <code>ct_mark</code>
            on a connection tracking entry before dropping a packet,
            for example.
          </p>
        </dd>

        <dt><code>ct_dnat;</code></dt>
        <dt><code>ct_dnat(<var>IP</var>);</code></dt>
        <dd>
          <p>
            <code>ct_dnat</code> sends the packet through the DNAT zone in
            connection tracking table to unDNAT any packet that was DNATed in
            the opposite direction.  The packet is then automatically sent to
            to the next tables as if followed by <code>next;</code> action.
            The next tables will see the changes in the packet caused by
            the connection tracker.
          </p>
          <p>
            <code>ct_dnat(<var>IP</var>)</code> sends the packet through the
            DNAT zone to change the destination IP address of the packet to
            the one provided inside the parentheses and commits the connection.
            The packet is then automatically sent to the next tables as if
            followed by <code>next;</code> action.  The next tables will see
            the changes in the packet caused by the connection tracker.
          </p>
        </dd>

        <dt><code>ct_snat;</code></dt>
        <dt><code>ct_snat(<var>IP</var>);</code></dt>
        <dd>
          <p>
            <code>ct_snat</code> sends the packet through the SNAT zone to
            unSNAT any packet that was SNATed in the opposite direction.  If
            the packet needs to be sent to the next tables, then it should be
            followed by a <code>next;</code> action.  The next tables will not
            see the changes in the packet caused by the connection tracker.
          </p>
          <p>
            <code>ct_snat(<var>IP</var>)</code> sends the packet through the
            SNAT zone to change the source IP address of the packet to
            the one provided inside the parenthesis and commits the connection.
            The packet is then automatically sent to the next tables as if
            followed by <code>next;</code> action.  The next tables will see the
            changes in the packet caused by the connection tracker.
          </p>
        </dd>

        <dt><code>arp { <var>action</var>; </code>...<code> };</code></dt>
        <dd>
          <p>
            Temporarily replaces the IPv4 packet being processed by an ARP
            packet and executes each nested <var>action</var> on the ARP
            packet.  Actions following the <var>arp</var> action, if any, apply
            to the original, unmodified packet.
          </p>

          <p>
            The ARP packet that this action operates on is initialized based on
            the IPv4 packet being processed, as follows.  These are default
            values that the nested actions will probably want to change:
          </p>

          <ul>
            <li><code>eth.src</code> unchanged</li>
            <li><code>eth.dst</code> unchanged</li>
            <li><code>eth.type = 0x0806</code></li>
            <li><code>arp.op = 1</code> (ARP request)</li>
            <li><code>arp.sha</code> copied from <code>eth.src</code></li>
            <li><code>arp.spa</code> copied from <code>ip4.src</code></li>
            <li><code>arp.tha = 00:00:00:00:00:00</code></li>
            <li><code>arp.tpa</code> copied from <code>ip4.dst</code></li>
          </ul>

          <p>
            The ARP packet has the same VLAN header, if any, as the IP packet
            it replaces.
          </p>

          <p><b>Prerequisite:</b> <code>ip4</code></p>
        </dd>

        <dt>
          <code>na { <var>action</var>; </code>...<code> };</code>
        </dt>

        <dd>
          <p>
            Temporarily replaces the IPv6 packet being processed by an IPv6
            neighbor advertisement (NA) packet and executes each nested
            <var>action</var> on the NA packet.  Actions following the
            <var>na</var> action, if any, apply to the original, unmodified
            packet.
          </p>

          <p>
            The NA packet that this action operates on is initialized based on
            the IPv6 packet being processed, as follows. These are default
            values that the nested actions will probably want to change:
          </p>

          <ul>
            <li><code>eth.dst</code> exchanged with <code>eth.src</code></li>
            <li><code>eth.type = 0x86dd</code></li>
            <li><code>ip6.dst</code> copied from <code>ip6.src</code></li>
            <li><code>ip6.src</code> copied from <code>nd.target</code></li>
            <li><code>icmp6.type = 136</code> (Neighbor Advertisement)</li>
            <li><code>nd.target</code> unchanged</li>
            <li><code>nd.sll = 00:00:00:00:00:00</code></li>
            <li><code>nd.tll</code> copied from <code>eth.dst</code></li>
          </ul>

          <p>
            The ND packet has the same VLAN header, if any, as the IPv6 packet
            it replaces.
          </p>

          <p>
            <b>Prerequisite:</b> <code>nd</code>
          </p>
        </dd>

        <dt><code>get_arp(<var>P</var>, <var>A</var>);</code></dt>

        <dd>
          <p>
            <b>Parameters</b>: logical port string field <var>P</var>, 32-bit
            IP address field <var>A</var>.
          </p>

          <p>
            Looks up <var>A</var> in <var>P</var>'s ARP table.  If an entry is
            found, stores its Ethernet address in <code>eth.dst</code>,
            otherwise stores <code>00:00:00:00:00:00</code> in
            <code>eth.dst</code>.
          </p>

          <p><b>Example:</b> <code>get_arp(outport, ip4.dst);</code></p>
        </dd>

        <dt>
          <code>put_arp(<var>P</var>, <var>A</var>, <var>E</var>);</code>
        </dt>

        <dd>
          <p>
            <b>Parameters</b>: logical port string field <var>P</var>, 32-bit
            IP address field <var>A</var>, 48-bit Ethernet address field
            <var>E</var>.
          </p>

          <p>
            Adds or updates the entry for IP address <var>A</var> in logical
            port <var>P</var>'s ARP table, setting its Ethernet address to
            <var>E</var>.
          </p>

          <p><b>Example:</b> <code>put_arp(inport, arp.spa, arp.sha);</code></p>
        </dd>

        <dt>
          <code><var>R</var> = put_dhcp_opts(<code>offerip</code> = <var>IP</var>, <var>D1</var> = <var>V1</var>, <var>D2</var> = <var>V2</var>, ..., <var>Dn</var> = <var>Vn</var>);</code>
        </dt>

        <dd>
          <p>
            <b>Parameters</b>: one or more DHCP option/value pairs, the first
            of which must set a value for the offered IP, <code>offerip</code>.
          </p>

          <p>
            <b>Result</b>: stored to a 1-bit subfield <var>R</var>.
          </p>

          <p>
            Valid only in the ingress pipeline.
          </p>

          <p>
            When this action is applied to a DHCP request packet (DHCPDISCOVER
            or DHCPREQUEST), it changes the packet into a DHCP reply (DHCPOFFER
            or DHCPACK, respectively), replaces the options by those specified
            as parameters, and stores 1 in <var>R</var>.
          </p>

          <p>
            When this action is applied to a non-DHCP packet or a DHCP packet
            that is not DHCPDISCOVER or DHCPREQUEST, it leaves the packet
            unchanged and stores 0 in <var>R</var>.
          </p>

          <p>
            The contents of the <ref table="DHCP_Option"/> table control the
            DHCP option names and values that this action supports.
          </p>

          <p>
            <b>Example:</b>
            <code>
              reg0[0] = put_dhcp_opts(offerip = 10.0.0.2, router = 10.0.0.1,
              netmask = 255.255.255.0, dns_server = {8.8.8.8, 7.7.7.7});
            </code>
          </p>
        </dd>

        <dt><code>ct_lb;</code></dt>
        <dt><code>ct_lb(</code><var>ip</var>[<code>:</code><var>port</var>]...<code>);</code></dt>
        <dd>
          <p>
            With one or more arguments, <code>ct_lb</code> commits the packet
            to the connection tracking table and DNATs the packet's destination
            IP address (and port) to the IP address or addresses (and optional
            ports) specified in the string.  If multiple comma-separated IP
            addresses are specified, each is given equal weight for picking the
            DNAT address.  Processing automatically moves on to the next table,
            as if <code>next;</code> were specified, and later tables act on
            the packet as modified by the connection tracker.  Connection
            tracking state is scoped by the logical port, so overlapping
            addresses may be used.
          </p>
          <p>
            Without arguments, <code>ct_lb</code> sends the packet to the
            connection tracking table to NAT the packets.  If the packet is
            part of an established connection that was previously committed to
            the connection tracker via <code>ct_lb(</code>...<code>)</code>, it
            will automatically get DNATed to the same IP address as the first
            packet in that connection.
          </p>
        </dd>
      </dl>

      <p>
        The following actions will likely be useful later, but they have not
        been thought out carefully.
      </p>

      <dl>
        <dt><code>icmp4 { <var>action</var>; </code>...<code> };</code></dt>
        <dd>
          <p>
            Temporarily replaces the IPv4 packet being processed by an ICMPv4
            packet and executes each nested <var>action</var> on the ICMPv4
            packet.  Actions following the <var>icmp4</var> action, if any,
            apply to the original, unmodified packet.
          </p>

          <p>
            The ICMPv4 packet that this action operates on is initialized based
            on the IPv4 packet being processed, as follows.  These are default
            values that the nested actions will probably want to change.
            Ethernet and IPv4 fields not listed here are not changed:
          </p>

          <ul>
            <li><code>ip.proto = 1</code> (ICMPv4)</li>
            <li><code>ip.frag = 0</code> (not a fragment)</li>
            <li><code>icmp4.type = 3</code> (destination unreachable)</li>
            <li><code>icmp4.code = 1</code> (host unreachable)</li>
          </ul>

          <p>
            Details TBD.
          </p>

          <p><b>Prerequisite:</b> <code>ip4</code></p>
        </dd>

        <dt><code>tcp_reset;</code></dt>
        <dd>
          <p>
            This action transforms the current TCP packet according to the
            following pseudocode:
          </p>

          <pre>
if (tcp.ack) {
        tcp.seq = tcp.ack;
} else {
        tcp.ack = tcp.seq + length(tcp.payload);
        tcp.seq = 0;
}
tcp.flags = RST;
</pre>

          <p>
            Then, the action drops all TCP options and payload data, and
            updates the TCP checksum.
          </p>

          <p>
            Details TBD.
          </p>

          <p><b>Prerequisite:</b> <code>tcp</code></p>
        </dd>
      </dl>
    </column>

    <column name="external_ids" key="stage-name">
      Human-readable name for this flow's stage in the pipeline.
    </column>

    <group title="Common Columns">
      The overall purpose of these columns is described under <code>Common
      Columns</code> at the beginning of this document.

      <column name="external_ids"/>
    </group>
  </table>

  <table name="Multicast_Group" title="Logical Port Multicast Groups">
    <p>
      The rows in this table define multicast groups of logical ports.
      Multicast groups allow a single packet transmitted over a tunnel to a
      hypervisor to be delivered to multiple VMs on that hypervisor, which
      uses bandwidth more efficiently.
    </p>

    <p>
      Each row in this table defines a logical multicast group numbered <ref
      column="tunnel_key"/> within <ref column="datapath"/>, whose logical
      ports are listed in the <ref column="ports"/> column.
    </p>

    <column name="datapath">
      The logical datapath in which the multicast group resides.
    </column>

    <column name="tunnel_key">
      The value used to designate this logical egress port in tunnel
      encapsulations.  An index forces the key to be unique within the <ref
      column="datapath"/>.  The unusual range ensures that multicast group IDs
      do not overlap with logical port IDs.
    </column>

    <column name="name">
      <p>
        The logical multicast group's name.  An index forces the name to be
        unique within the <ref column="datapath"/>.  Logical flows in the
        ingress pipeline may output to the group just as for individual logical
        ports, by assigning the group's name to <code>outport</code> and
        executing an <code>output</code> action.
      </p>

      <p>
        Multicast group names and logical port names share a single namespace
        and thus should not overlap (but the database schema cannot enforce
        this).  To try to avoid conflicts, <code>ovn-northd</code> uses names
        that begin with <code>_MC_</code>.
      </p>
    </column>

    <column name="ports">
      The logical ports included in the multicast group.  All of these ports
      must be in the <ref column="datapath"/> logical datapath (but the
      database schema cannot enforce this).
    </column>
  </table>

  <table name="Datapath_Binding" title="Physical-Logical Datapath Bindings">
    <p>
      Each row in this table identifies physical bindings of a logical
      datapath.  A logical datapath implements a logical pipeline among the
      ports in the <ref table="Port_Binding"/> table associated with it.  In
      practice, the pipeline in a given logical datapath implements either a
      logical switch or a logical router.
    </p>

    <column name="tunnel_key">
      The tunnel key value to which the logical datapath is bound.
      The <code>Tunnel Encapsulation</code> section in
      <code>ovn-architecture</code>(7) describes how tunnel keys are
      constructed for each supported encapsulation.
    </column>

    <group title="OVN_Northbound Relationship">
      <p>
        Each row in <ref table="Datapath_Binding"/> is associated with some
        logical datapath.  <code>ovn-northd</code> uses these keys to track the
        association of a logical datapath with concepts in the <ref
        db="OVN_Northbound"/> database.
      </p>

      <column name="external_ids" key="logical-switch" type='{"type": "uuid"}'>
        For a logical datapath that represents a logical switch,
        <code>ovn-northd</code> stores in this key the UUID of the
        corresponding <ref table="Logical_Switch" db="OVN_Northbound"/> row in
        the <ref db="OVN_Northbound"/> database.
      </column>

      <column name="external_ids" key="logical-router" type='{"type": "uuid"}'>
        For a logical datapath that represents a logical router,
        <code>ovn-northd</code> stores in this key the UUID of the
        corresponding <ref table="Logical_Router" db="OVN_Northbound"/> row in
        the <ref db="OVN_Northbound"/> database.
      </column>
    </group>

    <group title="Common Columns">
      The overall purpose of these columns is described under <code>Common
      Columns</code> at the beginning of this document.

      <column name="external_ids"/>
    </group>
  </table>

  <table name="Port_Binding" title="Physical-Logical Port Bindings">
    <p>
      Most rows in this table identify the physical location of a logical port.
      (The exceptions are logical patch ports, which do not have any physical
      location.)
    </p>

    <p>
      For every <code>Logical_Switch_Port</code> record in
      <code>OVN_Northbound</code> database, <code>ovn-northd</code>
      creates a record in this table.  <code>ovn-northd</code> populates
      and maintains every column except the <code>chassis</code> column,
      which it leaves empty in new records.
    </p>

    <p>
      <code>ovn-controller</code>/<code>ovn-controller-vtep</code>
      populates the <code>chassis</code> column for the records that
      identify the logical ports that are located on its hypervisor/gateway,
      which <code>ovn-controller</code>/<code>ovn-controller-vtep</code> in
      turn finds out by monitoring the local hypervisor's Open_vSwitch
      database, which identifies logical ports via the conventions described
      in <code>IntegrationGuide.md</code>.  (The exceptions are for
      <code>Port_Binding</code> records with <code>type</code> of
      <code>gateway</code>, whose locations are identified by
      <code>ovn-northd</code> via the <code>options:gateway-chassis</code>
      column in this table.  <code>ovn-controller</code> is still responsible
      to populate the <code>chassis</code> column.)
    </p>

    <p>
      When a chassis shuts down gracefully, it should clean up the
      <code>chassis</code> column that it previously had populated.
      (This is not critical because resources hosted on the chassis are equally
      unreachable regardless of whether their rows are present.)  To handle the
      case where a VM is shut down abruptly on one chassis, then brought up
      again on a different one,
      <code>ovn-controller</code>/<code>ovn-controller-vtep</code> must
      overwrite the <code>chassis</code> column with new information.
    </p>

    <group title="Core Features">
      <column name="datapath">
        The logical datapath to which the logical port belongs.
      </column>

      <column name="logical_port">
        A logical port, taken from <ref table="Logical_Switch_Port"
        column="name" db="OVN_Northbound"/> in the OVN_Northbound
        database's <ref table="Logical_Switch_Port" db="OVN_Northbound"/>
        table.  OVN does not prescribe a particular format for the
        logical port ID.
      </column>

      <column name="chassis">
        The meaning of this column depends on the value of the <ref column="type"/>
        column.  This is the meaning for each <ref column="type"/>

        <dl>
          <dt>(empty string)</dt>
          <dd>
            The physical location of the logical port.  To successfully identify a
            chassis, this column must be a <ref table="Chassis"/> record.  This is
            populated by <code>ovn-controller</code>.
          </dd>

          <dt>vtep</dt>
          <dd>
            The physical location of the hardware_vtep gateway.  To successfully
            identify a chassis, this column must be a <ref table="Chassis"/> record.
            This is populated by <code>ovn-controller-vtep</code>.
          </dd>

          <dt>localnet</dt>
          <dd>
            Always empty.  A localnet port is realized on every chassis that has
            connectivity to the corresponding physical network.
          </dd>

          <dt>gateway</dt>
          <dd>
            The physical location of the L3 gateway.  To successfully identify a
            chassis, this column must be a <ref table="Chassis"/> record.  This is
            populated by <code>ovn-controller</code> based on the value of
            the <code>options:gateway-chassis</code> column in this table.
          </dd>

          <dt>l2gateway</dt>
          <dd>
            The physical location of this L2 gateway.  To successfully identify a
            chassis, this column must be a <ref table="Chassis"/> record.
            This is populated by <code>ovn-controller</code> based on the value
            of the <code>options:l2gateway-chassis</code> column in this table.
          </dd>
        </dl>

      </column>

      <column name="tunnel_key">
        <p>
          A number that represents the logical port in the key (e.g. STT key or
          Geneve TLV) field carried within tunnel protocol packets.
        </p>

        <p>
          The tunnel ID must be unique within the scope of a logical datapath.
        </p>
      </column>

      <column name="mac">
        <p>
          The Ethernet address or addresses used as a source address on the
          logical port, each in the form
          <var>xx</var>:<var>xx</var>:<var>xx</var>:<var>xx</var>:<var>xx</var>:<var>xx</var>.
          The string <code>unknown</code> is also allowed to indicate that the
          logical port has an unknown set of (additional) source addresses.
        </p>

        <p>
          A VM interface would ordinarily have a single Ethernet address.  A
          gateway port might initially only have <code>unknown</code>, and then
          add MAC addresses to the set as it learns new source addresses.
        </p>
      </column>

      <column name="type">
        <p>
          A type for this logical port.  Logical ports can be used to model other
          types of connectivity into an OVN logical switch.  The following types
          are defined:
        </p>

        <dl>
          <dt>(empty string)</dt>
          <dd>VM (or VIF) interface.</dd>

          <dt><code>patch</code></dt>
          <dd>
            One of a pair of logical ports that act as if connected by a patch
            cable.  Useful for connecting two logical datapaths, e.g. to connect
            a logical router to a logical switch or to another logical router.
          </dd>

          <dt><code>gateway</code></dt>
          <dd>
            One of a pair of logical ports that act as if connected by a patch
            cable across multiple chassis.  Useful for connecting a logical
            switch with a Gateway router (which is only resident on a
            particular chassis).
          </dd>

          <dt><code>localnet</code></dt>
          <dd>
            A connection to a locally accessible network from each
            <code>ovn-controller</code> instance.  A logical switch can only
            have a single <code>localnet</code> port attached.  This is used
            to model direct connectivity to an existing network.
          </dd>

          <dt><code>l2gateway</code></dt>
          <dd>
            An L2 connection to a physical network.  The chassis this
            <ref table="Port_Binding"/> is bound to will serve as
            an L2 gateway to the network named by
            <ref column="options" table="Port_Binding"/>:<code>network_name</code>.
          </dd>

          <dt><code>vtep</code></dt>
          <dd>
            A port to a logical switch on a VTEP gateway chassis.  In order to
            get this port correctly recognized by the OVN controller, the <ref
            column="options"
            table="Port_Binding"/>:<code>vtep-physical-switch</code> and <ref
            column="options"
            table="Port_Binding"/>:<code>vtep-logical-switch</code> must also
            be defined.
          </dd>
        </dl>
      </column>
    </group>

    <group title="Patch Options">
      <p>
        These options apply to logical ports with <ref column="type"/> of
        <code>patch</code>.
      </p>

      <column name="options" key="peer">
        The <ref column="logical_port"/> in the <ref table="Port_Binding"/>
        record for the other side of the patch.  The named <ref
        column="logical_port"/> must specify this <ref column="logical_port"/>
        in its own <code>peer</code> option.  That is, the two patch logical
        ports must have reversed <ref column="logical_port"/> and
        <code>peer</code> values.
      </column>
    </group>

    <group title="L3 Gateway Options">
      <p>
        These options apply to logical ports with <ref column="type"/> of
        <code>gateway</code>.
      </p>

      <column name="options" key="peer">
        The <ref column="logical_port"/> in the <ref table="Port_Binding"/>
        record for the other side of the 'gateway' port.  The named <ref
        column="logical_port"/> must specify this <ref column="logical_port"/>
        in its own <code>peer</code> option.  That is, the two 'gateway'
        logical ports must have reversed <ref column="logical_port"/> and
        <code>peer</code> values.
      </column>

      <column name="options" key="gateway-chassis">
        The <code>chassis</code> in which the port resides.
      </column>
    </group>

    <group title="Localnet Options">
      <p>
        These options apply to logical ports with <ref column="type"/> of
        <code>localnet</code>.
      </p>

      <column name="options" key="network_name">
        Required.  <code>ovn-controller</code> uses the configuration entry
        <code>ovn-bridge-mappings</code> to determine how to connect to this
        network.  <code>ovn-bridge-mappings</code> is a list of network names
        mapped to a local OVS bridge that provides access to that network.  An
        example of configuring <code>ovn-bridge-mappings</code> would be:

        <pre>$ ovs-vsctl set open . external-ids:ovn-bridge-mappings=physnet1:br-eth0,physnet2:br-eth1</pre>

        <p>
          When a logical switch has a <code>localnet</code> port attached,
          every chassis that may have a local vif attached to that logical
          switch must have a bridge mapping configured to reach that
          <code>localnet</code>.  Traffic that arrives on a
          <code>localnet</code> port is never forwarded over a tunnel to
          another chassis.
        </p>
      </column>

      <column name="tag">
        If set, indicates that the port represents a connection to a specific
        VLAN on a locally accessible network. The VLAN ID is used to match
        incoming traffic and is also added to outgoing traffic.
      </column>
    </group>

    <group title="L2 Gateway Options">
      <p>
        These options apply to logical ports with <ref column="type"/> of
        <code>l2gateway</code>.
      </p>

      <column name="options" key="network_name">
        Required.  <code>ovn-controller</code> uses the configuration entry
        <code>ovn-bridge-mappings</code> to determine how to connect to this
        network.  <code>ovn-bridge-mappings</code> is a list of network names
        mapped to a local OVS bridge that provides access to that network.  An
        example of configuring <code>ovn-bridge-mappings</code> would be:

        <pre>$ ovs-vsctl set open . external-ids:ovn-bridge-mappings=physnet1:br-eth0,physnet2:br-eth1</pre>

        <p>
          When a logical switch has a <code>l2gateway</code> port attached,
          the chassis that the <code>l2gateway</code> port is bound to
          must have a bridge mapping configured to reach the network
          identified by <code>network_name</code>.
        </p>
      </column>

      <column name="options" key="l2gateway-chassis">
        Required. The <code>chassis</code> in which the port resides.
      </column>

      <column name="tag">
        If set, indicates that the gateway is connected to a specific
        VLAN on the physical network. The VLAN ID is used to match
        incoming traffic and is also added to outgoing traffic.
      </column>
    </group>

    <group title="VTEP Options">
      <p>
        These options apply to logical ports with <ref column="type"/> of
        <code>vtep</code>.
      </p>

      <column name="options" key="vtep-physical-switch">
        Required. The name of the VTEP gateway.
      </column>

      <column name="options" key="vtep-logical-switch">
        Required.  A logical switch name connected by the VTEP gateway.  Must
        be set when <ref column="type"/> is <code>vtep</code>.
      </column>
    </group>

    <group title="VMI (or VIF) Options">
      <p>
        These options apply to logical ports with <ref column="type"/> having
        (empty string)
      </p>

      <column name="options" key="policing_rate">
        If set, indicates the maximum rate for data sent from this interface,
        in kbps. Data exceeding this rate is dropped.
      </column>

      <column name="options" key="policing_burst">
        If set, indicates the maximum burst size for data sent from this
        interface, in kb.
      </column>
    </group>

    <group title="Nested Containers">
      <p>
        These columns support containers nested within a VM.  Specifically,
        they are used when <ref column="type"/> is empty and <ref
        column="logical_port"/> identifies the interface of a container spawned
        inside a VM.  They are empty for containers or VMs that run directly on
        a hypervisor.
      </p>

      <column name="parent_port">
        This is taken from
        <ref table="Logical_Switch_Port" column="parent_name"
        db="OVN_Northbound"/> in the OVN_Northbound database's
        <ref table="Logical_Switch_Port" db="OVN_Northbound"/> table.
      </column>

      <column name="tag">
        <p>
          Identifies the VLAN tag in the network traffic associated with that
          container's network interface.
        </p>

        <p>
          This column is used for a different purpose when <ref column="type"/>
          is <code>localnet</code> (see <code>Localnet Options</code>, above)
          or <code>l2gateway</code> (see <code>L2 Gateway Options</code>, above).
        </p>
      </column>
    </group>
  </table>

  <table name="MAC_Binding" title="IP to MAC bindings">
    <p>
      Each row in this table specifies a binding from an IP address to an
      Ethernet address that has been discovered through ARP (for IPv4) or
      neighbor discovery (for IPv6).  This table is primarily used to discover
      bindings on physical networks, because IP-to-MAC bindings for virtual
      machines are usually populated statically into the <ref
      table="Port_Binding"/> table.
    </p>

    <p>
      This table expresses a functional relationship: <ref
      table="MAC_Binding"/>(<ref column="logical_port"/>, <ref column="ip"/>) =
      <ref column="mac"/>.
    </p>

    <p>
      In outline, the lifetime of a logical router's MAC binding looks like
      this:
    </p>

    <ol>
      <li>
        On hypervisor 1, a logical router determines that a packet should be
        forwarded to IP address <var>A</var> on one of its router ports.  It
        uses its logical flow table to determine that <var>A</var> lacks a
        static IP-to-MAC binding and the <code>get_arp</code> action to
        determine that it lacks a dynamic IP-to-MAC binding.
      </li>

      <li>
        Using an OVN logical <code>arp</code> action, the logical router
        generates and sends a broadcast ARP request to the router port.  It
        drops the IP packet.
      </li>

      <li>
        The logical switch attached to the router port delivers the ARP request
        to all of its ports.  (It might make sense to deliver it only to ports
        that have no static IP-to-MAC bindings, but this could also be
        surprising behavior.)
      </li>

      <li>
        A host or VM on hypervisor 2 (which might be the same as hypervisor 1)
        attached to the logical switch owns the IP address in question.  It
        composes an ARP reply and unicasts it to the logical router port's
        Ethernet address.
      </li>

      <li>
        The logical switch delivers the ARP reply to the logical router port.
      </li>

      <li>
        The logical router flow table executes a <code>put_arp</code> action.
        To record the IP-to-MAC binding, <code>ovn-controller</code> adds a row
        to the <ref table="MAC_Binding"/> table.
      </li>

      <li>
        On hypervisor 1, <code>ovn-controller</code> receives the updated <ref
        table="MAC_Binding"/> table from the OVN southbound database.  The next
        packet destined to <var>A</var> through the logical router is sent
        directly to the bound Ethernet address.
      </li>
    </ol>

    <column name="logical_port">
      The logical port on which the binding was discovered.
    </column>

    <column name="ip">
      The bound IP address.
    </column>

    <column name="mac">
      The Ethernet address to which the IP is bound.
    </column>
    <column name="datapath">
      The logical datapath to which the logical port belongs.
    </column>
  </table>

  <table name="DHCP_Options" title="DHCP Options supported by native OVN DHCP">
    <p>
      Each row in this table stores the DHCP Options supported by native OVN
      DHCP. <code>ovn-northd</code> populates this table with the supported
      DHCP options. <code>ovn-controller</code> looks up this table to get the
      DHCP codes of the DHCP options defined in the "put_dhcp_opts" action.
      Please refer to the RFC 2132 <code>"https://tools.ietf.org/html/rfc2132"</code>
      for the possible list of DHCP options that can be defined here.
    </p>

    <column name="name">
      <p>
        Name of the DHCP option.
      </p>

      <p>
        Example. name="router"
      </p>
    </column>

    <column name="code">
      <p>
        DHCP option code for the DHCP option as defined in the RFC 2132.
      </p>

      <p>
        Example. code=3
      </p>
    </column>

    <column name="type">
      <p>
        Data type of the DHCP option code.
      </p>

      <dl>
        <dt><code>value: bool</code></dt>
        <dd>
          <p>
            This indicates that the value of the DHCP option is a bool.
          </p>

          <p>
            Example. "name=ip_forward_enable", "code=19", "type=bool".
          </p>

          <p>
            put_dhcp_opts(..., ip_forward_enable = 1,...)
          </p>
        </dd>

        <dt><code>value: uint8</code></dt>
        <dd>
          <p>
            This indicates that the value of the DHCP option is an unsigned
            int8 (8 bits)
          </p>

          <p>
            Example. "name=default_ttl", "code=23", "type=uint8".
          </p>

          <p>
            put_dhcp_opts(..., default_ttl = 50,...)
          </p>
        </dd>

        <dt><code>value: uint16</code></dt>
        <dd>
          <p>
            This indicates that the value of the DHCP option is an unsigned
            int16 (16 bits).
          </p>

          <p>
            Example. "name=mtu", "code=26", "type=uint16".
          </p>

          <p>
            put_dhcp_opts(..., mtu = 1450,...)
          </p>
        </dd>

        <dt><code>value: uint32</code></dt>
        <dd>
          <p>
            This indicates that the value of the DHCP option is an unsigned
            int32 (32 bits).
          </p>

          <p>
            Example. "name=lease_time", "code=51", "type=uint32".
          </p>

          <p>
            put_dhcp_opts(..., lease_time = 86400,...)
          </p>
        </dd>

        <dt><code>value: ipv4</code></dt>
        <dd>
          <p>
            This indicates that the value of the DHCP option is an IPv4
            address or addresses.
          </p>

          <p>
            Example. "name=router", "code=3", "type=ipv4".
          </p>

          <p>
            put_dhcp_opts(..., router = 10.0.0.1,...)
          </p>

          <p>
            Example. "name=dns_server", "code=6", "type=ipv4".
          </p>

          <p>
            put_dhcp_opts(..., dns_server = {8.8.8.8 7.7.7.7},...)
          </p>
        </dd>

        <dt><code>value: static_routes</code></dt>
        <dd>
          <p>
            This indicates that the value of the DHCP option contains a pair of
            IPv4 route and next hop addresses.
          </p>

          <p>
            Example. "name=classless_static_route", "code=121", "type=static_routes".
          </p>

          <p>
            put_dhcp_opts(..., classless_static_route = {30.0.0.0/24,10.0.0.4,0.0.0.0/0,10.0.0.1}...)
          </p>
        </dd>

        <dt><code>value: str</code></dt>
        <dd>
          <p>
            This indicates that the value of the DHCP option is a string.
          </p>

          <p>
            Example. "name=host_name", "code=12", "type=str".
          </p>
        </dd>
      </dl>
    </column>
  </table>
</database>