Using SR-IOV functionality

Using SR-IOV functionality

The purpose of this page is to describe how to enable SR-IOV functionality available in OpenStack (using OpenStack Networking) as of the Juno release. This page serves as a how-to guide on configuring OpenStack Networking and OpenStack Compute to create neutron SR-IOV ports.

The basics

PCI-SIG Single Root I/O Virtualization and Sharing (SR-IOV) specification defines a standardized mechanism to virtualize PCIe devices. The mechanism can virtualize a single PCIe Ethernet controller to appear as multiple PCIe devices. You can directly assign each virtual PCIe device to a VM, bypassing the hypervisor and virtual switch layer. As a result, users are able to achieve low latency and near-line wire speed.

The following terms are used over the document:

Term Definition
PF Physical Function. This is the physical ethernet controller that supports SR-IOV.
VF Virtual Function. This is a virtual PCIe device created from a physical ethernet controller.

In order to enable SR-IOV, the following steps are required:

  1. Create Virtual Functions (Compute)
  2. Whitelist PCI devices in nova-compute (Compute)
  3. Configure neutron-server (Controller)
  4. Configure nova-scheduler (Controller)
  5. Enable neutron sriov-agent (Compute)

Neutron sriov-agent

There are 2 ways of configuring SR-IOV:

  1. Without the sriov-agent running on each compute node
  2. With the sriov-agent running on each compute node

The sriov-agent allows you to set the admin state of ports and starting from Liberty allows you to control port security (enable and disable spoofchecking) and QoS rate limit settings.

When would you decide to not use sriov-agent?

  • Hardware is not supported: Currently Intel cards are known not to work with sriov-agent.
  • Extra configuration burden: Using sriov-agent requires you to run an extra agent on each compute node.

Known limitations

  • No OpenStack Dashboard integration. Users need to use CLI or API to create neutron SR-IOV ports.

  • The following functionalities are not implemented for SR-IOV ports: security groups, QoS, and ARP spoofing filtering.

  • When using Intel SR-IOV cards the sriov-agent should be disabled.

  • Live migration is not supported for instances with SR-IOV ports.


    QoS and ARP spoofing filtering are supported since Liberty when using sriov-agent.

Environment example

We recommend using Open vSwitch with VLAN as segration. This way you can combine normal VMs without SR-IOV ports and instances with SR-IOV ports on a single neutron network.


Thoughout this guide, eth3 is used as the PF and physnet2 is used as the provider network configured as a VLAN range. You are expected to change this according to your actual environment.

Create Virtual Functions (Compute)

In this step, create the VFs for the network interface that will be used for SR-IOV. Use eth3 as PF, which is also used as the interface for Open vSwitch VLAN and has access to the private networks of all machines.

The step to create VFs differ between SR-IOV card ethernet controller manufacturers. Currently the following manufactuers are known to work:

  • Intel
  • Mellanox

For Melanox SR-IOV ethernet cards see: Mellanox: HowTo Configure SR-IOV VFs

To create the VFs on Ubuntu for Intel SR-IOV ethernet cards, do the following:

  1. Make sure SR-IOV is enabled in BIOS, check for VT-d and make sure it is enabled. After enabling VT-d, enable IOMMU on Linux by adding intel_iommu=on to kernel parameters. Edit the file /etc/default/grub:

    GRUB_CMDLINE_LINUX_DEFAULT="nomdmonddf nomdmonisw intel_iommu=on
  2. Run the following if you have added new parameters:

    # update-grub
    # reboot
  3. On each compute node, create the VFs via the PCI SYS interface:

    # echo '7' > /sys/class/net/eth3/device/sriov_numvfs

    Alternatively VFs can be created by passing the max_vfs to the kernel module of your network interface. The max_vfs parameter has been deprecated so the PCI SYS interface is the preferred method.

  4. Now verify that the VFs have been created (Should see Virtual Function device):

    # lspci | grep Ethernet
  5. Persist created VFs on reboot:

    # echo "echo '7' > /sys/class/net/eth3/device/sriov_numvfs" >> /etc/rc.local


    The suggested way of making PCI SYS settings persistent is through sysfs.conf but for unknown reason changing sysfs.conf does not have any effect on Ubuntu 14.04.

Whitelist PCI devices nova-compute (Compute)

Tell nova-compute which pci devices are allowed to be passed through. Edit the file /etc/nova/nova.conf:

pci_passthrough_whitelist = { "devname": "eth3", "physical_network": "physnet2"}

This tells nova that all VFs belonging to eth3 are allowed to be passed through to VMs and belong to the neutron provider network physnet2. Restart nova compute with service nova-compute restart to let the changes have effect.

Alternatively the pci_passthrough_whitelist parameter also supports whitelisting by:

  • PCI address: The address uses the same syntax as in lspci and an asterisk (*) can be used to match anything.

    pci_passthrough_whitelist = { "address": "[[[[<domain>]:]<bus>]:][<slot>][.[<function>]]", "physical_network": "physnet2" }
    # Example match any domain, bus 0a, slot 00, all function
    pci_passthrough_whitelist = { "address": "*:0a:00.*", "physical_network": "physnet2" }
  • PCI vendor_id and product_id as displayed by the Linux utility lspci.

    pci_passthrough_whitelist = { "vendor_id": "<id>", "product_id": "<id>",
                                  "physical_network": "physnet2"}

If the device defined by the PCI address or devname corresponds to a SR-IOV PF, all VFs under the PF will match the entry. Multiple pci_passhtrough_whitelist entries per host are supported.

Configure neutron-server (Controller)

  1. Add sriovnicswitch as mechanism driver. Edit the file /etc/neutron/plugins/ml2/ml2_conf.ini:

    mechanism_drivers = openvswitch,sriovnicswitch
  2. Find out the vendor_id and product_id of your VFs by logging in to your compute node with VFs previously created:

    # lspci -nn | grep -i ethernet
    87:00.0 Ethernet controller [0200]: Intel Corporation 82599 10 Gigabit Dual Port Backplane Connection [8086:10f8] (rev 01)
    87:10.1 Ethernet controller [0200]: Intel Corporation 82599 Ethernet Controller Virtual Function [8086:10ed] (rev 01)
    87:10.3 Ethernet controller [0200]: Intel Corporation 82599 Ethernet Controller Virtual Function [8086:10ed] (rev 01)
  3. Update the /etc/neutron/plugins/ml2/ml2_conf_sriov.ini on each controller. In our case the vendor_id is 8086 and the product_id is 10ed. Tell neutron the vendor_id and product_id of the VFs that are supported.

    supported_pci_vendor_devs = 8086:10ed
  4. Enable or disable the sriovagent. Please see the section, Neutron sriov-agent if you want to disable or enable the sriovagent. Edit the agent_required parameter under the ml2_sriov section in /etc/neutron/plugins/ml2/ml2_conf_sriov.ini:

    agent_required = True


    If you enabled agent_required=True make sure that you run the sriov-agent on each compute node.

  5. Add the newly configured ml2_conf_sriov.ini as parameter to the neutron-server daemon. Edit the file /etc/init/neutron-server.conf:

    --config-file /etc/neutron/neutron.conf --config-file /etc/neutron/plugin.ini
    --config-file /etc/neutron/plugins/ml2/ml2_conf_sriov.ini
  6. To make the changes have effect, restart the neutron-server service with the service neutron-server restart.

Configure nova-scheduler (Controller)

  1. On every controller node running nova-scheduler add PCIDeviceScheduler to the scheduler_default_filters parameter and add a new line for scheduler_available_filters parameter under the [default] section in /etc/nova/nova.conf:

    scheduler_default_filters = RetryFilter, AvailabilityZoneFilter, RamFilter, ComputeFilter, ComputeCapabilitiesFilter, ImagePropertiesFilter, ServerGroupAntiAffinityFilter, ServerGroupAffinityFilter, PciPassthroughFilter
    scheduler_available_filters = nova.scheduler.filters.all_filters
    scheduler_available_filters = nova.scheduler.filters.pci_passthrough_filter.PciPassthroughFilter
  2. Now restart the nova-scheduler service with service nova-scheduler restart.

Enable neutron sriov-agent (Compute)


You only need to enable the sriov-agent if you decided to set agent_required=True in the step Configure neutron-server (Controller). If you set agent_required=False, you can safely skip this step.

  1. On each compute node edit the file /etc/neutron/plugins/ml2/ml2_conf_sriov.ini:

    firewall_driver = neutron.agent.firewall.NoopFirewallDriver
    physical_device_mappings = physnet2:eth3
    exclude_devices =

    exclude_devices is empty so all the VFs associated with eth3 may be configured by the agent. If you want to exclude specific VFs, add them to the exclude_devices parameter as follows:

    exclude_devices = eth1:0000:07:00.2; 0000:07:00.3, eth2:0000:05:00.1; 0000:05:00.2
  2. Test whether the sriov-agent runs successfully:

    # neutron-sriov-nic-agent --config-file /etc/neutron/neutron.conf --config-file /etc/neutron/plugins/ml2/ml2_conf_sriov.ini
  3. Enable the neutron-sriov-agent to start automatically at system start. If your distribution does not come with a daemon file for your init system, create a daemon configuration file. For example on Ubuntu install the package:

    # apt-get install neutron-plugin-sriov-agent

Creating instances with SR-IOV ports

After the configuration is done, you can now launch Instances with neutron SR-IOV ports.

  1. Get the id of the neutron network where you want the SR-IOV port to be created:

    $ net_id=`neutron net-show net04 | grep "\ id\ " | awk '{ print $4 }'`
  2. Create the SR-IOV port. We specify vnic_type direct, but other options include macvtap:

    $ port_id=`neutron port-create $net_id --name sriov_port --binding:vnic_type direct | grep "\ id\ " | awk '{ print $4 }'`
  3. Create the VM. For the nic we specify the SR-IOV port created in step 2:

    $ nova boot --flavor m1.large --image ubuntu_14.04 --nic port-id=$port_id test-sriov
Creative Commons Attribution 3.0 License

Except where otherwise noted, this document is licensed under Creative Commons Attribution 3.0 License. See all OpenStack Legal Documents.