Attaching physical PCI devices to guests

Attaching physical PCI devices to guests

The PCI passthrough feature in OpenStack allows full access and direct control of a physical PCI device in guests. This mechanism is generic for any kind of PCI device, and runs with a Network Interface Card (NIC), Graphics Processing Unit (GPU), or any other devices that can be attached to a PCI bus. Correct driver installation is the only requirement for the guest to properly use the devices.

Some PCI devices provide Single Root I/O Virtualization and Sharing (SR-IOV) capabilities. When SR-IOV is used, a physical device is virtualized and appears as multiple PCI devices. Virtual PCI devices are assigned to the same or different guests. In the case of PCI passthrough, the full physical device is assigned to only one guest and cannot be shared.

Note

For information on creating servers with virtual SR-IOV devices, refer to the Networking Guide.

Limitations

  • Attaching SR-IOV ports to existing servers is not currently supported, see bug 1708433 for details.
  • Cold migration (resize) of servers with SR-IOV devices attached was not supported until the 14.0.0 Newton release, see bug 1512800 for details.

To enable PCI passthrough, follow the steps below:

  1. Configure nova-scheduler (Controller)
  2. Configure nova-api (Controller)**
  3. Configure a flavor (Controller)
  4. Enable PCI passthrough (Compute)
  5. Configure PCI devices in nova-compute (Compute)

Note

The PCI device with address 0000:41:00.0 is used as an example. This will differ between environments.

Configure nova-scheduler (Controller)

  1. Configure nova-scheduler as specified in Configure nova-scheduler.
  2. Restart the nova-scheduler service.

Configure nova-api (Controller)

  1. Specify the PCI alias for the device.

    Configure a PCI alias a1 to request a PCI device with a vendor_id of 0x8086 and a product_id of 0x154d. The vendor_id and product_id correspond the PCI device with address 0000:41:00.0.

    Edit /etc/nova/nova.conf:

    [pci]
    alias = { "vendor_id":"8086", "product_id":"154d", "device_type":"type-PF", "name":"a1" }
    

    Refer to pci.alias for syntax information.

  2. Restart the nova-api service.

Configure a flavor (Controller)

Configure a flavor to request two PCI devices, each with vendor_id of 0x8086 and product_id of 0x154d:

# openstack flavor set m1.large --property "pci_passthrough:alias"="a1:2"

For more information about the syntax for pci_passthrough:alias, refer to Flavors.

Enable PCI passthrough (Compute)

Enable VT-d and IOMMU. For more information, refer to steps one and two in Create Virtual Functions.

For Hyper-V compute nodes, the requirements are as follows:

  • Windows 10 or Windows / Hyper-V Server 2016 or newer.
  • VT-d and SR-IOV enabled on the host.
  • Assignable PCI devices.

In order to check the requirements above and if there are any assignable PCI devices, run the following Powershell commands:

Start-BitsTransfer https://raw.githubusercontent.com/Microsoft/Virtualization-Documentation/master/hyperv-samples/benarm-powershell/DDA/survey-dda.ps1
 .\survey-dda.ps1

If the compute node passes all the requirements, the desired assignable PCI devices to be disabled and unmounted from the host, in order to be assignable by Hyper-V. The following can be read for more details: Hyper-V PCI passthrough.

Configure PCI devices (Compute)

  1. Configure nova-compute to allow the PCI device to pass through to VMs. Edit /etc/nova/nova.conf:

    [pci]
    passthrough_whitelist = { "address": "0000:41:00.0" }
    

    Alternatively specify multiple PCI devices using whitelisting:

    [pci]
    passthrough_whitelist = { "vendor_id": "8086", "product_id": "10fb" }
    

    All PCI devices matching the vendor_id and product_id are added to the pool of PCI devices available for passthrough to VMs.

    Refer to pci.passthrough_whitelist for syntax information.

  2. Specify the PCI alias for the device.

    From the Newton release, to resize guest with PCI device, configure the PCI alias on the compute node as well.

    Configure a PCI alias a1 to request a PCI device with a vendor_id of 0x8086 and a product_id of 0x154d. The vendor_id and product_id correspond the PCI device with address 0000:41:00.0.

    Edit /etc/nova/nova.conf:

    [pci]
    alias = { "vendor_id":"8086", "product_id":"154d", "device_type":"type-PF", "name":"a1" }
    

    Refer to pci.alias for syntax information.

  3. Restart the nova-compute service.

Create instances with PCI passthrough devices

The nova-scheduler selects a destination host that has PCI devices available with the specified vendor_id and product_id that matches the alias from the flavor.

# openstack server create --flavor m1.large --image cirros-0.3.5-x86_64-uec --wait test-pci
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.