libvirt/docs/formatnetwork.rst
Jiri Denemark 43c0325b10 network: Make virtual domains resolvable from the host
This patch adds a new attribute "register" to the <domain> element. If
set to "yes", the DNS server created for the virtual network is
registered with systemd-resolved as a name server for the associated
domain. The names known to the dnsmasq process serving DNS and DHCP
requests for the virtual network will then be resolvable from the host
by appending the domain name to them.

Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
2024-02-07 14:19:42 +01:00

55 KiB

Network XML format

This page provides an introduction to the network XML format. For background information on the concepts referred to here, consult the relevant wiki page.

Element and attribute overview

The root element required for all virtual networks is named network and has no configurable attributes (although since 0.10.0 there is one optional read-only attribute - when examining the live configuration of a network, the attribute connections, if present, specifies the number of guest interfaces currently connected via this network). The network XML format is available since 0.3.0

General metadata

The first elements provide basic metadata about the virtual network.

<network ipv6='yes' trustGuestRxFilters='no'>
  <name>default</name>
  <uuid>3e3fce45-4f53-4fa7-bb32-11f34168b82b</uuid>
  <title>A short description - title - of the network</title>
  <description>Some human readable description</description>
  <metadata>
    <app1:foo xmlns:app1="http://app1.org/app1/">..</app1:foo>
    <app2:bar xmlns:app2="http://app1.org/app2/">..</app2:bar>
  </metadata>
  ...
name

The content of the name element provides a short name for the virtual network. This name should consist only of alphanumeric characters and is required to be unique within the scope of a single host. It is used to form the filename for storing the persistent configuration file. Since 0.3.0

uuid

The content of the uuid element provides a globally unique identifier for the virtual network. The format must be RFC 4122 compliant, eg 3e3fce45-4f53-4fa7-bb32-11f34168b82b. If omitted when defining/creating a new network, a random UUID is generated. Since 0.3.0

metadata

The metadata node can be used by applications to store custom metadata in the form of XML nodes/trees. Applications must use custom namespaces on their XML nodes/trees, with only one top-level element per namespace (if the application needs structure, they should have sub-elements to their namespace element). Since 2.1.0

ipv6

When set to yes, the optional parameter ipv6 enables a network definition with no IPv6 gateway addresses specified to have guest-to-guest communications. For further information, see the example below for the example with no gateway addresses. Since 1.0.1

trustGuestRxFilters

The optional parameter trustGuestRxFilters can be used to set that attribute of the same name for each domain interface connected to this network ( since 1.2.10 ). See the Network interfaces section of the domain XML documentation for more details. Note that an explicit setting of this attribute in a portgroup or the individual domain interface will override the setting in the network.

title

The optional element title provides space for a short description of the network. The title should not contain any newlines. Since 9.7.0 .

description

The content of the description element provides a human readable description of the network. This data is not used by libvirt in any way, it can contain any information the user wants. Since 9.7.0

Connectivity

The next set of elements control how a virtual network is provided connectivity to the physical LAN (if at all).

...
<bridge name="virbr0" stp="on" delay="5" macTableManager="libvirt"/>
<mtu size="9000"/>
<domain name="example.com" localOnly="no" register="no"/>
<forward mode="nat" dev="eth0"/>
...
bridge

The name attribute on the bridge element defines the name of a bridge device which will be used to construct the virtual network. The virtual machines will be connected to this bridge device allowing them to talk to each other. The bridge device may also be connected to the LAN. When defining a new network with a <forward> mode of "nat", "route", or "open" (or an isolated network with no <forward> element), libvirt will automatically generate a unique name for the bridge device if none is given, and this name will be permanently stored in the network configuration so that that the same name will be used every time the network is started. For these types of networks (nat, route, open, and isolated), a bridge name beginning with the prefix "virbr" is recommended (and that is what is auto-generated), but not enforced. Attribute stp specifies if Spanning Tree Protocol is 'on' or 'off' (default is 'on'). Attribute delay sets the bridge's forward delay value in seconds (default is 0. As the kernel has a minimum delay, values below it may not be counted). Since 0.3.0

The macTableManager attribute of the bridge element is used to tell libvirt how the bridge's MAC address table (used to determine the correct egress port for packets based on destination MAC address) will be managed. In the default kernel setting, the kernel automatically adds and removes entries, typically using learning, flooding, and promiscuous mode on the bridge's ports in order to determine the proper egress port for packets. When macTableManager is set to libvirt, libvirt disables kernel management of the MAC table (in the case of the Linux host bridge, this means enabling vlan_filtering on the bridge, and disabling learning and unicast_filter for all bridge ports), and explicitly adds/removes entries to the table according to the MAC addresses in the domain interface configurations. Allowing libvirt to manage the MAC table can improve performance - with a Linux host bridge, for example, turning off learning and unicast_flood on ports has its own performance advantage, and can also lead to an additional boost by permitting the kernel to automatically turn off promiscuous mode on some ports of the bridge (in particular, the port attaching the bridge to the physical network). However, it can also cause some networking setups to stop working (e.g. vlan tagging, multicast, guest-initiated changes to MAC address) and is not supported by older kernels. Since 1.2.11, requires kernel 3.17 or newer

The optional zone attribute of the bridge element is used to specify the firewalld zone for the bridge of a network with forward mode of "nat", "route", "open", or one with no forward specified. By default, the bridges of all virtual networks with these forward modes are placed in the firewalld zone named "libvirt", which permits incoming DNS, DHCP, TFTP, and SSH to the host from guests on the network. This behavior can be changed either by modifying the libvirt zone (using firewalld management tools), or by placing the network in a different zone (which will also be managed using firewalld tools). Since 5.1.0

mtu

The size attribute of the <mtu> element specifies the Maximum Transmission Unit (MTU) for the network. Since 3.1.0 . In the case of a libvirt-managed network (one with forward mode of nat, route, open, or no forward element (i.e. an isolated network), this will be the MTU assigned to the bridge device when libvirt creates it, and thereafter also assigned to all tap devices created to connect guest interfaces. Network types not specifically mentioned here don't support having an MTU set in the libvirt network config. If mtu size is unspecified, the default setting for the type of device being used is assumed (usually 1500).

domain

The name attribute on the domain element defines the DNS domain of the DHCP server. This element is optional, and is only used for those networks with a <forward> mode of "nat" or "route" (or an isolated network with no <forward> element). Since 0.4.5

If the optional localOnly attribute on the domain element is "yes", then DNS requests under this domain will only be resolved by the virtual network's own DNS server - they will not be forwarded to the host's upstream DNS server. If localOnly is "no", and by default, unresolved requests will be forwarded. Since 1.2.12

Since 10.1.0 the optional register attribute can be used to request registering the DNS server for resolving this domain with the host's DNS resolver. When set to "yes", the host resolver will forward all requests for domain names from this domain to the DNS server created for this virtual network. To avoid DNS loops localOnly has to be set to "yes" as well. This feature requires systemd-resolved to be running on the host.

forward

Inclusion of the forward element indicates that the virtual network is to be connected to the physical LAN. Since 0.3.0. The mode attribute determines the method of forwarding. If there is no forward element, the network will be isolated from any other network (unless a guest connected to that network is acting as a router, of course). The following are valid settings for mode (if there is a forward element but mode is not specified, mode='nat' is assumed):

nat

All traffic between guests connected to this network and the physical network will be forwarded to the physical network via the host's IP routing stack, after the guest's IP address is translated to appear as the host machine's public IP address (a.k.a. Network Address Translation, or "NAT"). This allows multiple guests, all having access to the physical network, on a host that is only allowed a single public IP address. If a network has any IPv6 addresses defined, the IPv6 traffic will be forwarded using plain routing, since IPv6 has no concept of NAT. Firewall rules will allow outbound connections to any other network device whether ethernet, wireless, dialup, or VPN. If the dev attribute is set, the firewall rules will restrict forwarding to the named device only. Inbound connections from other networks are all prohibited; all connections between guests on the same network, and to/from the host to the guests, are unrestricted and not NATed. Since 0.4.2

Since 1.0.3 it is possible to specify a public IPv4 address and port range to be used for the NAT by using the <nat> subelement. Note that all addresses from the range are used, not just those that are in use on the host. The address range is set with the <address> subelements and start and stop attributes:

...
  <forward mode='nat'>
    <nat>
      <address start='1.2.3.4' end='1.2.3.10'/>
    </nat>
  </forward>
...

A single IPv4 address can be set by setting start and end attributes to the same value.

The port range to be used for the <nat> can be set via the subelement <port>:

...
  <forward mode='nat'>
    <nat>
      <port start='500' end='1000'/>
    </nat>
  </forward>
...

Since 6.5.0 it is possible to enable NAT with IPv6 networking. As noted above, IPv6 has historically done plain forwarding and thus to avoid breaking historical compatibility, IPv6 NAT must be explicitly requested.

...
  <forward mode='nat'>
    <nat ipv6='yes'/>
  </forward>
...
route

Guest network traffic will be forwarded to the physical network via the host's IP routing stack, but without having NAT applied. Again, if the dev attribute is set, firewall rules will restrict forwarding to the named device only. This presumes that the local LAN router has suitable routing table entries to return traffic to this host. All incoming and outgoing sessions to guest on these networks are unrestricted. (To restrict incoming traffic to a guest on a routed network, you can configure nwfilter rules on the guest's interfaces.) Since 0.4.2

open

As with mode='route', guest network traffic will be forwarded to the physical network via the host's IP routing stack, but there will be no firewall rules added to either enable or prevent any of this traffic. When forward='open' is set, the dev attribute cannot be set (because the forward dev is enforced with firewall rules, and the purpose of forward='open' is to have a forwarding mode where libvirt doesn't add any firewall rules). This mode presumes that the local LAN router has suitable routing table entries to return traffic to this host, and that some other management system has been used to put in place any necessary firewall rules. Although no firewall rules will be added for the network, it is of course still possible to add restrictions for specific guests using nwfilter rules on the guests' interfaces.) Since 2.2.0

bridge

This network describes either 1) an existing host bridge that was configured outside of libvirt (if a <bridge name='xyz'/> element has been specified, Since 0.9.4 ), 2) an existing Open vSwitch bridge that was configured outside of libvirt (if both a <bridge name='xyz'/> element and a <virtualport type='openvswitch'/> have been specified Since 0.10.0 ) 3) an interface or group of interfaces to be used for a "direct" connection via macvtap using macvtap's "bridge" mode (if the forward element has one or more <interface> subelements, Since 0.9.4 ) (see Direct attachment to physical interface for descriptions of the various macvtap modes). libvirt doesn't attempt to manage the bridge interface at all, thus the <bridge> element's stp and delay attributes are not allowed; no iptables rules, IP addresses, or DHCP/DNS services are added; at the IP level, the guest interface appears to be directly connected to the physical interface. Since 0.9.4

private

This network uses a macvtap "direct" connection in "private" mode to connect each guest to the network. The physical interface to be used will be picked from among those listed in <interface> subelements of the <forward> element; when using 802.1Qbh mode (as indicated by the <virtualport> type attribute - note that this requires an 802.1Qbh-capable hardware switch), each physical interface can only be in use by a single guest interface at a time; in modes other than 802.1Qbh, multiple guest interfaces can share each physical interface (libvirt will attempt to balance usage between all available interfaces). Since 0.9.4

vepa

This network uses a macvtap "direct" connection in "vepa" mode to connect each guest to the network (this requires that the physical interfaces used be connected to a vepa-capable hardware switch. The physical interface to be used will be picked from among those listed in <interface> subelements of the <forward> element; multiple guest interfaces can share each physical interface (libvirt will attempt to balance usage between all available interfaces). Since 0.9.4

passthrough

This network uses a macvtap "direct" connection in "passthrough" mode to connect each guest to the network (note that this is not the same thing as "PCI passthrough"). The physical interface to be used will be picked from among those listed in <interface> subelements of the <forward> element. Each physical interface can only be in use by a single guest interface at a time, so libvirt will keep track of which interfaces are currently in use, and only assign unused interfaces (if there are no available physical interfaces when a domain interface is being attached, an error will be logged, and the operation causing the attach will fail (usually either a domain start, or a hotplug interface attach to a domain). Since 0.9.4

hostdev

This network facilitates PCI Passthrough of a network device. A network device is chosen from the interface pool and directly assigned to the guest using generic device passthrough, after first optionally setting the device's MAC address and vlan tag to the configured value, and optionally associating the device with an 802.1Qbh capable switch using a <virtualport> element. Note that - due to limitations in standard single-port PCI ethernet card driver design - only SR-IOV (Single Root I/O Virtualization) virtual function (VF) devices can be assigned in this manner; to assign a standard single-port PCI or PCIe ethernet card to a guest, use the traditional <hostdev> device definition. Since 0.10.0

To force use of a particular device-specific VFIO driver when assigning the devices to a guest, a <forward type='hostdev'> interface can have an optional driver sub-element with a model attribute set to the name of the driver to use Since 10.0.0 (QEMU only). When not specified, libvirt will attempt to find a suitable VFIO variant driver for the device, and if not found it will use the generic driver "vfio-pci".

Note that this "intelligent passthrough" of network devices is very similar to the functionality of a standard <hostdev> device, the difference being that this method allows specifying a MAC address, vlan tag, and <virtualport> for the passed-through device. If these capabilities are not required, if you have a standard single-port PCI, PCIe, or USB network card that doesn't support SR-IOV (and hence would anyway lose the configured MAC address during reset after being assigned to the guest domain), or if you are using a version of libvirt older than 0.10.0, you should use a standard <hostdev> device definition in the domain's configuration to assign the device to the guest instead of defining an <interface type='network'> pointing to a network with <forward mode='hostdev'/>.

As mentioned above, a <forward> element can have multiple <interface> subelements, each one giving the name of a physical interface that can be used for this network Since 0.9.4 :

...
  <forward mode='passthrough'>
    <interface dev='eth10'/>
    <interface dev='eth11'/>
    <interface dev='eth12'/>
    <interface dev='eth13'/>
    <interface dev='eth14'/>
  </forward>
...

since 0.10.0 , <interface> also has an optional read-only attribute - when examining the live configuration of a network, the attribute connections, if present, specifies the number of guest interfaces currently connected via this physical interface.

Additionally, since 0.9.10 , libvirt allows a shorthand for specifying all virtual interfaces associated with a single physical function, by using the <pf> subelement to call out the corresponding physical interface associated with multiple virtual interfaces:

...
  <forward mode='passthrough'>
    <pf dev='eth0'/>
  </forward>
...

When a guest interface is being constructed, libvirt will pick an interface from this list to use for the connection. In modes where physical interfaces can be shared by multiple guest interfaces, libvirt will choose the interface that currently has the least number of connections. For those modes that do not allow sharing of the physical device (in particular, 'passthrough' mode, and 'private' mode when using 802.1Qbh), libvirt will choose an unused physical interface or, if it can't find an unused interface, fail the operation.

since 0.10.0 When using forward mode 'hostdev', the interface pool is specified with a list of <address> elements, each of which has <type> (must always be 'pci'), <domain>, <bus>, <slot>and <function> attributes.

...
  <forward mode='hostdev' managed='yes'>
    <address type='pci' domain='0' bus='4' slot='0' function='1'/>
    <address type='pci' domain='0' bus='4' slot='0' function='2'/>
    <address type='pci' domain='0' bus='4' slot='0' function='3'/>
  </forward>
...

Alternatively the interface pool can also be defined using a single physical function <pf> subelement to call out the corresponding physical interface associated with multiple virtual interfaces (similar to passthrough mode):

...
  <forward mode='hostdev' managed='yes'>
    <pf dev='eth0'/>
  </forward>
...

Quality of service

...
  <forward mode='nat' dev='eth0'/>
  <bandwidth>
    <inbound average='1000' peak='5000' burst='5120'/>
    <outbound average='128' peak='256' burst='256'/>
  </bandwidth>
...

The <bandwidth> element allows setting quality of service for a particular network ( since 0.9.4 ). Setting bandwidth for a network is supported only for networks with a <forward> mode of route, nat, bridge, or no mode at all (i.e. an "isolated" network). Setting bandwidth is not supported for forward modes passthrough, private, or hostdev. Attempts to do this will lead to a failure to define the network or to create a transient network.

The <bandwidth> element can only be a subelement of a domain's <interface>, a subelement of a <network>, or a subelement of a <portgroup> in a <network>.

As a subelement of a domain's <interface>, the bandwidth only applies to that one interface of the domain. As a subelement of a <network>, the bandwidth is a total aggregate bandwidth to/from all guest interfaces attached to that network, not to each guest interface individually. If a domain's <interface> has <bandwidth> element values higher than the aggregate for the entire network, then the aggregate bandwidth for the <network> takes precedence. This is because the two choke points are independent of each other where the domain's <interface> bandwidth control is applied on the interface's tap device, while the <network> bandwidth control is applied on the interface part of the bridge device created for that network.

As a subelement of a <portgroup> in a <network>, if a domain's <interface> has a portgroup attribute in its <source> element and if the <interface> itself has no <bandwidth> element, then the <bandwidth> element of the portgroup will be applied individually to each guest interface defined to be a member of that portgroup. Any <bandwidth> element in the domain's <interface> definition will override the setting in the portgroup ( since 1.0.1 ).

Incoming and outgoing traffic can be shaped independently. The bandwidth element can have at most one inbound and at most one outbound child element. Leaving either of these children elements out results in no QoS applied for that traffic direction. So, when you want to shape only incoming traffic, use inbound only, and vice versa. Each of these elements have one mandatory attribute - average (or floor as described below). The attributes are as follows, where accepted values for each attribute is an integer number.

average

Specifies the desired average bit rate for the interface being shaped (in kilobytes/second).

peak

Optional attribute which specifies the maximum rate at which the bridge can send data (in kilobytes/second). Note the limitation of implementation: this attribute in the outbound element is ignored (as Linux ingress filters don't know it yet).

burst

Optional attribute which specifies the amount of kibibytes that can be transmitted in a single burst at peak speed.

floor

Optional attribute available only for the inbound element. This attribute guarantees minimal throughput for shaped interfaces. This, however, requires that all traffic goes through one point where QoS decisions can take place, hence why this attribute works only for virtual networks for now (that is <interface type='network'/> with a forward type of route, nat, open or no forward at all). Moreover, the virtual network the interface is connected to is required to have at least inbound QoS set (average at least). If using the floor attribute users don't need to specify average. However, peak and burst attributes still require average. Currently, the Linux kernel doesn't allow ingress qdiscs to have any classes therefore floor can be applied only on inbound and not outbound.

Attributes average, peak, and burst are available since 0.9.4 , while the floor attribute is available since 1.0.1 .

Setting VLAN tag (on supported network types only)

<network>
  <name>ovs-net</name>
  <forward mode='bridge'/>
  <bridge name='ovsbr0'/>
  <virtualport type='openvswitch'>
    <parameters interfaceid='09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f'/>
  </virtualport>
  <vlan trunk='yes'>
    <tag id='42' nativeMode='untagged'/>
    <tag id='47'/>
  </vlan>
  <portgroup name='dontpanic'>
    <vlan>
      <tag id='42'/>
    </vlan>
  </portgroup>
</network>

If (and only if) the network connection used by the guest supports VLAN tagging transparent to the guest, an optional <vlan> element can specify one or more VLAN tags to apply to the guest's network traffic Since 0.10.0 . Network connections that support guest-transparent VLAN tagging include 1) type='bridge' interfaces connected to an Open vSwitch bridge Since 0.10.0 , 2) SRIOV Virtual Functions (VF) used via type='hostdev' (direct device assignment) Since 0.10.0 , and 3) SRIOV VFs used via type='direct' with mode='passthrough' (macvtap "passthru" mode) Since 1.3.5 . All other connection types, including standard linux bridges and libvirt's own virtual networks, do not support it. 802.1Qbh (vn-link) and 802.1Qbg (VEPA) switches provide their own way (outside of libvirt) to tag guest traffic onto a specific VLAN. Each tag is given in a separate <tag> subelement of <vlan> (for example: <tag id='42'/>). For VLAN trunking of multiple tags (which is supported only on Open vSwitch connections), multiple <tag> subelements can be specified, which implies that the user wants to do VLAN trunking on the interface for all the specified tags. In the case that VLAN trunking of a single tag is desired, the optional attribute trunk='yes' can be added to the toplevel <vlan> element to differentiate trunking of a single tag from normal tagging.

For network connections using Open vSwitch it is also possible to configure 'native-tagged' and 'native-untagged' VLAN modes Since 1.1.0. This is done with the optional nativeMode attribute on the <tag> subelement: nativeMode may be set to 'tagged' or 'untagged'. The id attribute of the <tag> subelement containing nativeMode sets which VLAN is considered to be the "native" VLAN for this interface, and the nativeMode attribute determines whether or not traffic for that VLAN will be tagged.

<vlan> elements can also be specified in a <portgroup> element, as well as directly in a domain's <interface> element. In the case that a vlan tag is specified in multiple locations, the setting in <interface> takes precedence, followed by the setting in the <portgroup> selected by the interface config. The <vlan> in <network> will be selected only if none is given in <portgroup> or <interface>.

Isolating ports from one another

<network>
  <name>isolated-ports</name>
  <forward mode='bridge'/>
  <bridge name='br0'/>
  <port isolated='yes'/>
</network>

Since 6.1.0. The port element property isolated, when set to yes (default setting is no) is used to isolate the network traffic of each guest on the network from all other guests connected to the network; it does not have an effect on communication between the guests and the host, or between the guests and destinations beyond this network. This setting is only supported for networks that use a Linux host bridge to connect guest interfaces via a standard tap device (i.e. those with a forward mode of nat, route, open, bridge, or no forward mode).

Portgroups

...
  <forward mode='private'/>
    <interface dev="eth20"/>
    <interface dev="eth21"/>
    <interface dev="eth22"/>
    <interface dev="eth23"/>
    <interface dev="eth24"/>
  </forward>
  <portgroup name='engineering' default='yes'>
    <virtualport type='802.1Qbh'>
      <parameters profileid='test'/>
    </virtualport>
    <bandwidth>
      <inbound average='1000' peak='5000' burst='5120'/>
      <outbound average='1000' peak='5000' burst='5120'/>
    </bandwidth>
  </portgroup>
  <portgroup name='sales' trustGuestRxFilters='no'>
    <virtualport type='802.1Qbh'>
      <parameters profileid='salestest'/>
    </virtualport>
    <bandwidth>
      <inbound average='500' peak='2000' burst='2560'/>
      <outbound average='128' peak='256' burst='256'/>
    </bandwidth>
  </portgroup>
...

Since 0.9.4 A portgroup provides a method of easily putting guest connections to the network into different classes, with each class potentially having a different level/type of service. Since 0.9.4 Each network can have multiple portgroup elements (and one of those can optionally be designated as the 'default' portgroup for the network), and each portgroup has a name, as well as various attributes and subelements associated with it. The currently supported subelements are <bandwidth> (described in Quality of service) and <virtualport> (documented here). If a domain interface definition specifies a portgroup (by adding a portgroup attribute to the <source> subelement), that portgroup's info will be merged into the interface's configuration. If no portgroup is given in the interface definition, and one of the network's portgroups has default='yes', that default portgroup will be used. If no portgroup is given in the interface definition, and there is no default portgroup, then none will be used. Any <bandwidth> specified directly in the domain XML will take precedence over any setting in the chosen portgroup. if a <virtualport> is specified in the portgroup (and/or directly in the network definition), the multiple virtualports will be merged, and any parameter that is specified in more than one virtualport, and is not identical, will be considered an error, and will prevent the interface from starting.

portgroups also support the optional parameter trustGuestRxFilters which can be used to set that attribute of the same name for each domain interface using this portgroup ( since 1.2.10 ). See the Network interfaces section of the domain XML documentation for more details. Note that an explicit setting of this attribute in the portgroup overrides the network-wide setting, and an explicit setting in the individual domain interface will override the setting in the portgroup.

Static Routes

Static route definitions are used to provide routing information to the virtualization host for networks which are not directly reachable from the virtualization host, but *are* reachable from a guest domain that is itself reachable from the host since 1.0.6 .

As shown in Network config with no gateway addresses example, it is possible to define a virtual network interface with no IPv4 or IPv6 addresses. Such networks are useful to provide host connectivity to networks which are only reachable via a guest. A guest with connectivity both to the guest-only network and to another network that is directly reachable from the host can act as a gateway between the networks. A static route added to the "host-visible" network definition provides the routing information so that IP packets can be sent from the virtualization host to guests on the hidden network.

Here is a fragment of a definition which shows the static route specification as well as the IPv4 and IPv6 definitions for network addresses which are referred to in the gateway gateway address specifications. Note that the third static route specification includes the metric attribute specification with a value of 2. This particular route would *not* be preferred if there was another existing rout on the system with the same address and prefix but with a lower value for the metric. If there is a route in the host system configuration that should be overridden by a route in a virtual network whenever the virtual network is running, the configuration for the system-defined route should be modified to have a higher metric, and the route on the virtual network given a lower metric (for example, the default metric of "1").

...
  <ip address="192.168.122.1" netmask="255.255.255.0">
    <dhcp>
      <range start="192.168.122.128" end="192.168.122.254"/>
    </dhcp>
  </ip>
  <route address="192.168.222.0" prefix="24" gateway="192.168.122.2"/>
  <ip family="ipv6" address="2001:db8:ca2:2::1" prefix="64"/>
  <route family="ipv6" address="2001:db8:ca2:3::" prefix="64" gateway="2001:db8:ca2:2::2"/>
  <route family="ipv6" address="2001:db9:4:1::" prefix="64" gateway="2001:db8:ca2:2::3" metric='2'/>
...

Addressing

The final set of elements define the addresses (IPv4 and/or IPv6, as well as MAC) to be assigned to the bridge device associated with the virtual network, and optionally enable DHCP services. These elements are only valid for isolated networks (no forward element specified), and for those with a forward mode of 'route' or 'nat'.

...
<mac address='00:16:3E:5D:C7:9E'/>
<domain name="example.com"/>
<dns>
  <txt name="example" value="example value"/>
  <forwarder addr="8.8.8.8"/>
  <forwarder domain='example.com' addr="8.8.4.4"/>
  <forwarder domain='www.example.com'/>
  <srv service='name' protocol='tcp' domain='test-domain-name' target='.'
    port='1024' priority='10' weight='10'/>
  <host ip='192.168.122.2'>
    <hostname>myhost</hostname>
    <hostname>myhostalias</hostname>
  </host>
</dns>
<ip address="192.168.122.1" netmask="255.255.255.0" localPtr="yes">
  <dhcp>
    <range start="192.168.122.100" end="192.168.122.254">
      <lease expiry='1' unit='hours'/>
    </range>
    <host mac="00:16:3e:77:e2:ed" name="foo.example.com" ip="192.168.122.10">
      <lease expiry='30' unit='minutes'/>
    </host>
    <host mac="00:16:3e:3e:a9:1a" name="bar.example.com" ip="192.168.122.11"/>
  </dhcp>
</ip>
<ip family="ipv6" address="2001:db8:ca2:2::1" prefix="64" localPtr="yes"/>
<route family="ipv6" address="2001:db9:ca1:1::" prefix="64" gateway="2001:db8:ca2:2::2"/>
mac

The address attribute defines a MAC (hardware) address formatted as 6 groups of 2-digit hexadecimal numbers, the groups separated by colons (eg, "52:54:00:1C:DA:2F"). This MAC address is assigned to the bridge device when it is created. Generally it is best to not specify a MAC address when creating a network - in this case, if a defined MAC address is needed for proper operation, libvirt will automatically generate a random MAC address and save it in the config. Allowing libvirt to generate the MAC address will assure that it is compatible with the idiosyncrasies of the platform where libvirt is running. Since 0.8.8

dns

The dns element of a network contains configuration information for the virtual network's DNS server Since 0.9.3 .

The dns element can have an optional enable attribute Since 2.2.0 . If enable is "no", then no DNS server will be setup by libvirt for this network (and any other configuration in <dns> will be ignored). If enable is "yes" or unspecified (including the complete absence of any <dns> element) then a DNS server will be setup by libvirt to listen on all IP addresses specified in the network's configuration.

The dns element can have an optional forwardPlainNames attribute Since 1.1.2 . If forwardPlainNames is "no", then DNS resolution requests for names that are not qualified with a domain (i.e. names with no "." character) will not be forwarded to the host's upstream DNS server - they will only be resolved if they are known locally within the virtual network's own DNS server. If forwardPlainNames is "yes", unqualified names will be forwarded to the upstream DNS server if they can't be resolved by the virtual network's own DNS server.

Currently supported sub-elements of <dns> are:

forwarder

The dns element can have 0 or more <forwarder> elements. Each forwarder element defines an alternate DNS server to use for some, or all, DNS requests sent to this network's DNS server. There are two attributes -domain, and addr; at least one of these must be specified in any <forwarder> element. If both domain and addr are specified, then all requests that match the given domain will be forwarded to the DNS server at addr. If only domain is specified, then all matching domains will be resolved locally (or via the host's standard DNS forwarding if they can't be resolved locally). If an addr is specified by itself, then all DNS requests to the network's DNS server will be forwarded to the DNS server at that address with no exceptions. addr Since 1.1.3 , domain Since 2.2.0 .

txt

A dns element can have 0 or more txt elements. Each txt element defines a DNS TXT record and has two attributes, both required: a name that can be queried via dns, and a value that will be returned when that name is queried. names cannot contain embedded spaces or commas. value is a single string that can contain multiple values separated by commas. Since 0.9.3

host

The host element within dns is the definition of DNS hosts to be passed to the DNS service. The IP address is identified by the ip attribute and the names for that IP address are identified in the hostname sub-elements of the host element. Since 0.9.3

srv

The dns element can have also 0 or more srv record elements. Each srv record element defines a DNS SRV record and has 2 mandatory and 5 optional attributes. The mandatory attributes are service and protocol (tcp, udp) and the optional attributes are target, port, priority, weight and domain as defined in DNS server SRV RFC (RFC 2782). Since 0.9.9

ip

The address attribute defines an IPv4 address in dotted-decimal format, or an IPv6 address in standard colon-separated hexadecimal format, that will be configured on the bridge device associated with the virtual network. To the guests this IPv4 address will be their IPv4 default route. For IPv6, the default route is established via Router Advertisement. For IPv4 addresses, the netmask attribute defines the significant bits of the network address, again specified in dotted-decimal format. For IPv6 addresses, and as an alternate method for IPv4 addresses, the significant bits of the network address can be specified with the prefix attribute, which is an integer (for example, netmask='255.255.255.0' could also be given as prefix='24'). The family attribute is used to specify the type of address - ipv4 or ipv6; if no family is given, ipv4 is assumed. More than one address of each family can be defined for a network. The optional localPtr attribute ( since 3.0.0 ) configures the DNS server to not forward any reverse DNS requests for IP addresses from the network configured by the address and netmask/prefix attributes. For some unusual network prefixes (not divisible by 8 for IPv4 or not divisible by 4 for IPv6) libvirt may be unable to compute the PTR domain automatically. The ip element is supported since 0.3.0 . IPv6, multiple addresses on a single network, family, and prefix are supported since 0.8.7 . The ip element may contain the following elements:

tftp

The optional tftp element and its mandatory root attribute enable TFTP services. The attribute specifies the path to the root directory served via TFTP. The tftp element is not supported for IPv6 addresses, and can only be specified on a single IPv4 address per network. Since 0.7.1

dhcp

The presence of this element enables DHCP services on the virtual network. The dhcp element is supported for both IPv4 ( since 0.3.0 ) and IPv6 ( since 1.0.1 ), but only for one IP address of each type per network. The following sub-elements are supported:

range

The start and end attributes on the range element specify the boundaries of a pool of addresses to be provided to DHCP clients. These two addresses must lie within the scope of the network defined on the parent ip element. There may be zero or more range elements specified. Since 0.3.0

host

Within the dhcp element there may be zero or more host elements. These specify hosts which will be given names and predefined IP addresses by the built-in DHCP server. Any IPv4 host element must specify the MAC address of the host to be assigned a given name (via the mac attribute), the IP to be assigned to that host (via the ip attribute), and the name itself (the name attribute). The IPv6 host element differs slightly from that for IPv4: there is no mac attribute since a MAC address has no defined meaning in IPv6. Instead, the name attribute is used to identify the host to be assigned the IPv6 address. For DHCPv6, the name is the plain name of the client host sent by the client to the server. Note that this method of assigning a specific IP address can also be used for IPv4 instead of the mac attribute. Since 0.4.5

bootp

The optional bootp element specifies BOOTP options to be provided by the DHCP server for IPv4 only. Two attributes are supported: file is mandatory and gives the file to be used for the boot image; server is optional and gives the address of the TFTP server from which the boot image will be fetched. server defaults to the same host that runs the DHCP server, as is the case when the tftp element is used. The BOOTP options currently have to be the same for all address ranges and statically assigned addresses. Since 0.7.1 (server since 0.7.3 )

Optionally, range and host elements can have lease child element which specifies the lease time through it's attributes expiry and unit (which accepts seconds, minutes and hours and defaults to minutes if omitted). The minimal lease time is 2 minutes, except when setting an infinite lease time (expiry='0'). Since 6.3.0

Network namespaces

A special XML namespace is available for passing options directly to the underlying dnsmasq configuration file since 5.6.0 . Usage of XML namespaces comes with no support guarantees, so use at your own risk.

This example XML will pass the option strings foo=bar and cname=*.foo.example.com,master.example.com directly to the underlying dnsmasq instance.

<network xmlns:dnsmasq='http://libvirt.org/schemas/network/dnsmasq/1.0'>
  ...
  <dnsmasq:options>
    <dnsmasq:option value="foo=bar"/>
    <dnsmasq:option value="cname=*.foo.example.com,master.example.com"/>
  </dnsmasq:options>
</network>

Example configuration

NAT based network

This example is the so called "default" virtual network. It is provided and enabled out-of-the-box for all libvirt installations. This is a configuration that allows guest OS to get outbound connectivity regardless of whether the host uses ethernet, wireless, dialup, or VPN networking without requiring any specific admin configuration. In the absence of host networking, it at least allows guests to talk directly to each other.

<network>
  <name>default</name>
  <bridge name="virbr0"/>
  <forward mode="nat"/>
  <ip address="192.168.122.1" netmask="255.255.255.0">
    <dhcp>
      <range start="192.168.122.2" end="192.168.122.254"/>
    </dhcp>
  </ip>
  <ip family="ipv6" address="2001:db8:ca2:2::1" prefix="64"/>
</network>

Below is a variation of the above example which adds an IPv6 dhcp range definition.

<network>
  <name>default6</name>
  <bridge name="virbr0"/>
  <forward mode="nat"/>
  <ip address="192.168.122.1" netmask="255.255.255.0">
    <dhcp>
      <range start="192.168.122.2" end="192.168.122.254"/>
    </dhcp>
  </ip>
  <ip family="ipv6" address="2001:db8:ca2:2::1" prefix="64">
    <dhcp>
      <range start="2001:db8:ca2:2:1::10" end="2001:db8:ca2:2:1::ff"/>
    </dhcp>
  </ip>
</network>

IPv6 NAT based network

Below is a variation for also providing IPv6 NAT. This can be especially useful when using multiple interfaces where some, such as WiFi cards, can not be bridged (usually on a laptop), making it difficult to provide end-to-end IPv6 routing.

<network>
  <name>default6</name>
  <bridge name="virbr0"/>
  <forward mode="nat">
    <nat ipv6='yes'>
      <port start='1024' end='65535'/>
    </nat>
  </forward>
  <ip address="192.168.122.1" netmask="255.255.255.0">
    <dhcp>
      <range start="192.168.122.2" end="192.168.122.254"/>
    </dhcp>
  </ip>
  <ip family="ipv6" address="fdXX:XXXX:XXXX:NNNN::" prefix="64"/>
</network>

IPv6 NAT addressing has some caveats over the more straight forward IPv4 case. RFC 4193 defines the address range fd00::/8 for /48 IPv6 private networks. It should be concatenated with a random 40-bit string (i.e. 10 random hexadecimal digits replacing the X values above, RFC 4193 provides an algorithm if you do not have a source of sufficient randomness). This leaves 0 through ffff for subnets (N above) which you can use at will.

Many operating systems will not consider these addresses as preferential to IPv4, due to some practical history of these addresses being present but unroutable and causing networking issues. On many Linux distributions, you may need to override /etc/gai.conf with values from RFC 3484 to have your IPv6 NAT network correctly preferenced over IPv4.

Routed network config

This is a variant on the default network which routes traffic from the virtual network to the LAN without applying any NAT. It requires that the IP address range be pre-configured in the routing tables of the router on the host network. This example further specifies that guest traffic may only go out via the eth1 host network device.

<network>
  <name>local</name>
  <bridge name="virbr1"/>
  <forward mode="route" dev="eth1"/>
  <ip address="192.168.122.1" netmask="255.255.255.0">
    <dhcp>
      <range start="192.168.122.2" end="192.168.122.254"/>
    </dhcp>
  </ip>
  <ip family="ipv6" address="2001:db8:ca2:2::1" prefix="64"/>
</network>

Below is another IPv6 variation. Instead of a dhcp range being specified, this example has a couple of IPv6 host definitions. Note that most of the dhcp host definitions use an "id" (client id or DUID) since this has proven to be a more reliable way of specifying the interface and its association with an IPv6 address. The first is a DUID-LLT, the second a DUID-LL, and the third a DUID-UUID. Since 1.0.3

<network>
  <name>local6</name>
  <bridge name="virbr1"/>
  <forward mode="route" dev="eth1"/>
  <ip address="192.168.122.1" netmask="255.255.255.0">
    <dhcp>
      <range start="192.168.122.2" end="192.168.122.254"/>
    </dhcp>
  </ip>
  <ip family="ipv6" address="2001:db8:ca2:2::1" prefix="64">
    <dhcp>
      <host name="paul" ip="2001:db8:ca2:2:3::1"/>
      <host id="0:1:0:1:18:aa:62:fe:0:16:3e:44:55:66" ip="2001:db8:ca2:2:3::2"/>
      <host id="0:3:0:1:0:16:3e:11:22:33" name="ralph" ip="2001:db8:ca2:2:3::3"/>
      <host id="0:4:7e:7d:f0:7d:a8:bc:c5:d2:13:32:11:ed:16:ea:84:63"
        name="badbob" ip="2001:db8:ca2:2:3::4"/>
    </dhcp>
  </ip>
</network>

Below is yet another IPv6 variation. This variation has only IPv6 defined with DHCPv6 on the primary IPv6 network. A static link if defined for a second IPv6 network which will not be directly visible on the bridge interface but there will be a static route defined for this network via the specified gateway. Note that the gateway address must be directly reachable via (on the same subnet as) one of the <ip> addresses defined for this <network>. Since 1.0.6

<network>
  <name>net7</name>
  <bridge name="virbr7"/>
  <forward mode="route"/>
  <ip family="ipv6" address="2001:db8:ca2:7::1" prefix="64">
    <dhcp>
      <range start="2001:db8:ca2:7::100" end="2001:db8:ca2::1ff"/>
      <host id="0:4:7e:7d:f0:7d:a8:bc:c5:d2:13:32:11:ed:16:ea:84:63"
        name="lucas" ip="2001:db8:ca2:2:3::4"/>
    </dhcp>
  </ip>
  <route family="ipv6" address="2001:db8:ca2:8::" prefix="64" gateway="2001:db8:ca2:7::4"/>
</network>

Isolated network config

This variant provides a completely isolated private network for guests. The guests can talk to each other, and the host OS, but cannot reach any other machines on the LAN, due to the omission of the forward element in the XML description.

<network>
  <name>private</name>
  <bridge name="virbr2"/>
  <ip address="192.168.152.1" netmask="255.255.255.0">
    <dhcp>
      <range start="192.168.152.2" end="192.168.152.254"/>
    </dhcp>
  </ip>
  <ip family="ipv6" address="2001:db8:ca2:3::1" prefix="64"/>
</network>

Isolated IPv6 network config

This variation of an isolated network defines only IPv6. Note that most of the dhcp host definitions use an "id" (client id or DUID) since this has proven to be a more reliable way of specifying the interface and its association with an IPv6 address. The first is a DUID-LLT, the second a DUID-LL, and the third a DUID-UUID. Since 1.0.3

<network>
  <name>sixnet</name>
  <bridge name="virbr6"/>
  <ip family="ipv6" address="2001:db8:ca2:6::1" prefix="64">
    <dhcp>
      <host name="peter" ip="2001:db8:ca2:6:6::1"/>
      <host id="0:1:0:1:18:aa:62:fe:0:16:3e:44:55:66" ip="2001:db8:ca2:6:6::2"/>
      <host id="0:3:0:1:0:16:3e:11:22:33" name="dariusz" ip="2001:db8:ca2:6:6::3"/>
      <host id="0:4:7e:7d:f0:7d:a8:bc:c5:d2:13:32:11:ed:16:ea:84:63"
        name="anita" ip="2001:db8:ca2:6:6::4"/>
    </dhcp>
  </ip>
</network>

Using an existing host bridge

Since 0.9.4 This shows how to use a pre-existing host bridge "br0". The guests will effectively be directly connected to the physical network (i.e. their IP addresses will all be on the subnet of the physical network, and there will be no restrictions on inbound or outbound connections).

<network>
  <name>host-bridge</name>
  <forward mode="bridge"/>
  <bridge name="br0"/>
</network>

Using a macvtap "direct" connection

Since 0.9.4, QEMU and KVM only, requires Linux kernel 2.6.34 or newer This shows how to use macvtap to connect to the physical network directly through one of a group of physical devices (without using a host bridge device). As with the host bridge network, the guests will effectively be directly connected to the physical network so their IP addresses will all be on the subnet of the physical network, and there will be no restrictions on inbound or outbound connections. Note that, due to a limitation in the implementation of macvtap, these connections do not allow communication directly between the host and the guests - if you require this you will either need the attached physical switch to be operating in a mirroring mode (so that all traffic coming to the switch is reflected back to the host's interface), or provide alternate means for this communication (e.g. a second interface on each guest that is connected to an isolated network). The other forward modes that use macvtap (private, vepa, and passthrough) would be used in a similar fashion.

<network>
  <name>direct-macvtap</name>
  <forward mode="bridge">
    <interface dev="eth20"/>
    <interface dev="eth21"/>
    <interface dev="eth22"/>
    <interface dev="eth23"/>
    <interface dev="eth24"/>
  </forward>
</network>

Network config with no gateway addresses

A valid network definition can contain no IPv4 or IPv6 addresses. Such a definition can be used for a "very private" or "very isolated" network since it will not be possible to communicate with the virtualization host via this network. However, this virtual network interface can be used for communication between virtual guest systems. This works for IPv4 and (Since 1.0.1) IPv6. However, the new ipv6='yes' must be added for guest-to-guest IPv6 communication.

<network ipv6='yes'>
  <name>nogw</name>
  <uuid>7a3b7497-1ec7-8aef-6d5c-38dff9109e93</uuid>
  <bridge name="virbr2" stp="on" delay="0"/>
  <mac address='00:16:3E:5D:C7:9E'/>
</network>