Applies to SUSE Linux Enterprise Server 15 SP4

15 Configuring virtual machines with `virsh` #

You can use virsh to configure virtual machines (VM) on the command line as an alternative to using the Virtual Machine Manager. With virsh, you can control the state of a VM, edit the configuration of a VM or even migrate a VM to another host. The following sections describe how to manage VMs by using virsh.

15.1 Editing the VM configuration #

The configuration of a VM is stored in an XML file in /etc/libvirt/qemu/ and looks like this:

Example 15.1: Example XML configuration file #

<domain type='kvm'>
  <name>sles15</name>
  <uuid>ab953e2f-9d16-4955-bb43-1178230ee625</uuid>
  <memory unit='KiB'>2097152</memory>
  <currentMemory unit='KiB'>2097152</currentMemory>
  <vcpu placement='static'>2</vcpu>
  <os>
    <type arch='x86_64' machine='pc-q35-2.0'>hvm</type>
  </os>
  <features>...</features>
  <cpu mode='custom' match='exact' check='partial'>
    <model fallback='allow'>Skylake-Client-IBRS</model>
  </cpu>
  <clock>...</clock>
  <on_poweroff>destroy</on_poweroff>
  <on_reboot>restart</on_reboot>
  <on_crash>destroy</on_crash>
  <pm>
    <suspend-to-mem enabled='no'/>
    <suspend-to-disk enabled='no'/>
  </pm>
  <devices>
    <emulator>/usr/bin/qemu-system-x86_64</emulator>
    <disk type='file' device='disk'>...</disk>
  </devices>
  ...
</domain>

If you want to edit the configuration of a VM Guest, check if it is offline:

> sudo virsh list --inactive

If your VM Guest is in this list, you can safely edit its configuration:

> sudo virsh edit NAME_OF_VM_GUEST

Before saving the changes, virsh validates your input against a RelaxNG schema.

15.2 Changing the machine type #

When installing with the virt-install tool, the machine type for a VM Guest is pc-q35 by default. The machine type is stored in the VM Guest's configuration file in the type element:

<type arch='x86_64' machine='pc-q35-2.3'>hvm</type>

As an example, the following procedure shows how to change this value to the machine type q35. The value q35 is an Intel* chipset and includes PCIe, supports up to 12 USB ports, and has support for SATA and IOMMU.

Procedure 15.1: Changing machine type #

Check whether your VM Guest is inactive:

> sudo virsh list --inactive
Id    Name                           State
----------------------------------------------------
-     sles15                         shut off

Edit the configuration for this VM Guest:
```
> sudo virsh edit sles15
```
Replace the value of the machine attribute with pc-q35-2.0 :
```
<type arch='x86_64' machine='pc-q35-2.0'>hvm</type>
```
Restart the VM Guest:
```
> sudo virsh start sles15
```
Check if the machine type has changed. Log in to the VM Guest and run the following command:
```
> sudo dmidecode | grep Product
Product Name: Standard PC (Q35 + ICH9, 2009)
```

Tip: Machine type update recommendations

Whenever the QEMU version on the host system is upgraded (for example, when upgrading the VM Host Server to a new service pack), upgrade the machine type of the VM Guests to the latest available version. To check, use the command qemu-system-x86_64 -M help on the VM Host Server.

The default machine type pc-i440fx, for example, is regularly updated. If your VM Guest still runs with a machine type of pc-i440fx-1.X, we strongly recommend an update to pc-i440fx-2.X. This allows taking advantage of the most recent updates and corrections in machine definitions, and ensures better future compatibility.

15.3 Configuring hypervisor features #

libvirt automatically enables a default set of hypervisor features that are sufficient in most circumstances, but also allows enabling and disabling features as needed. As an example, Xen does not support enabling PCI pass-through by default. It must be enabled with the passthrough setting. Hypervisor features can be configured with virsh. Look for the <features> element in the VM Guest's configuration file and adjust the various features as required. Continuing with the Xen pass-through example:

> sudo virsh edit sle15sp1
 <features>
    <xen>
      <passthrough/>
    </xen>
 </features>

Save your changes and restart the VM Guest.

See the Hypervisor features section of the libvirt Domain XML format manual at https://libvirt.org/formatdomain.html#elementsFeatures for more information.

15.4 Configuring CPU #

Many aspects of the virtual CPUs presented to VM Guests are configurable with virsh. The number of current and maximum CPUs allocated to a VM Guest can be changed, as well as the model of the CPU and its feature set. The following subsections describe how to change the common CPU settings of a VM Guest.

15.4.1 Configuring the number of CPUs #

The number of allocated CPUs is stored in the VM Guest's XML configuration file in /etc/libvirt/qemu/ in the vcpu element:

<vcpu placement='static'>1</vcpu>

In this example, the VM Guest has only one allocated CPU. The following procedure shows how to change the number of allocated CPUs for the VM Guest:

Check whether your VM Guest is inactive:

> sudo virsh list --inactive
Id    Name                           State
----------------------------------------------------
-     sles15                         shut off

Edit the configuration for an existing VM Guest:
```
> sudo virsh edit sles15
```
Change the number of allocated CPUs:
```
<vcpu placement='static'>2</vcpu>
```
Restart the VM Guest:
```
> sudo virsh start sles15
```

Check if the number of CPUs in the VM has changed.

> sudo virsh vcpuinfo sled15
VCPU:           0
CPU:            N/A
State:          N/A
CPU time        N/A
CPU Affinity:   yy

VCPU:           1
CPU:            N/A
State:          N/A
CPU time        N/A
CPU Affinity:   yy

You can also change the number of CPUs while the VM Guest is running. CPUs can be hotplugged until the maximum number configured at VM Guest start is reached. Likewise they can be hot-unplugged until the lower limit of 1 is reached. The following example demonstrates changing the active CPU count from 2 to pre-defined maximum of 4.

Check the current live vcpu count:

> sudo virsh vcpucount sles15 | grep live
maximum      live           4
current      live           2

Change the current, or active, number of CPUs to 4:
```
> sudo virsh setvcpus sles15 --count 4 --live
```

Check that the current live vcpu count is now 4:

> sudo virsh vcpucount sles15 | grep live
maximum      live           4
current      live           4

Important: Exceeding 255 CPUs

With KVM it is possible to define a VM Guest with more than 255 CPUs, however additional configuration is necessary to start and run the VM Guest. The ioapic feature needs to be tuned and an IOMMU device needs to be added to the VM Guest. Below is an example configuration for 288 CPUs.

<domain>
 <vcpu placement='static'>288</vcpu>
 <features>
  <ioapic driver='qemu'/>
 </features>
 <devices>
  <iommu model='intel'>
   <driver intremap='on' eim='on'/>
  </iommu>
 </devices>
</domain>

15.4.2 Configuring the CPU model #

The CPU model exposed to a VM Guest can often influence the workload running within it. The default CPU model is derived from a CPU mode known as host-model.

<cpu mode='host-model'/>

When starting a VM Guest with CPU mode host-model, libvirt will copy its model of the host CPU into the VM Guest definition. The host CPU model and features copied to the VM Guest definition can be observed in the output of the virsh capabilities.

Another interesting CPU mode is host-passthrough.

<cpu mode='host-passthrough'/>

When starting a VM Guest with CPU mode host-passthrough, it is presented with a CPU that is exactly the same as the VM Host Server CPU. This can be useful when the VM Guest workload requires CPU features not available in libvirt's simplified host-model CPU. The host-passthrough CPU mode comes with the disadvantage of reduced migration flexibility. A VM Guest with host-passthrough CPU mode can only be migrated to a VM Host Server with identical hardware.

When using the host-passthrough CPU mode, it is still possible to disable undesirable features. The following configuration will present the VM Guest with a CPU that is exactly the same as the host CPU but with the vmx feature disabled.

<cpu mode='host-passthrough'>
  <feature policy='disable' name='vmx'/>
  </cpu>

The custom CPU mode is another common mode used to define a normalized CPU that can be migrated throughout dissimilar hosts in a cluster. For example, in a cluster with hosts containing Nehalem, IvyBridge, and SandyBridge CPUs, the VM Guest can be configured with a custom CPU mode that contains a Nehalem CPU model.

<cpu mode='custom' match='exact'>
  <model fallback='allow'>Nehalem</model>
  <feature policy='require' name='vme'/>
  <feature policy='require' name='ds'/>
  <feature policy='require' name='acpi'/>
  <feature policy='require' name='ss'/>
  <feature policy='require' name='ht'/>
  <feature policy='require' name='tm'/>
  <feature policy='require' name='pbe'/>
  <feature policy='require' name='dtes64'/>
  <feature policy='require' name='monitor'/>
  <feature policy='require' name='ds_cpl'/>
  <feature policy='require' name='vmx'/>
  <feature policy='require' name='est'/>
  <feature policy='require' name='tm2'/>
  <feature policy='require' name='xtpr'/>
  <feature policy='require' name='pdcm'/>
  <feature policy='require' name='dca'/>
  <feature policy='require' name='rdtscp'/>
  <feature policy='require' name='invtsc'/>
  </cpu>

For more information on libvirt's CPU model and topology options, see the CPU model and topology documentation at https://libvirt.org/formatdomain.html#cpu-model-and-topology.

15.5 Changing boot options #

The boot menu of the VM Guest can be found in the os element and usually looks like this:

<os>
  <type>hvm</type>
  <loader>readonly='yes' secure='no' type='rom'/>/usr/lib/xen/boot/hvmloader</loader>
  <nvram template='/usr/share/OVMF/OVMF_VARS.fd'/>/var/lib/libvirt/nvram/guest_VARS.fd</nvram>
  <boot dev='hd'/>
  <boot dev='cdrom'/>
  <bootmenu enable='yes' timeout='3000'/>
  <smbios mode='sysinfo'/>
  <bios useserial='yes' rebootTimeout='0'/>
  </os>

In this example, two devices are available, hd and cdrom . The configuration also reflects the actual boot order, so the hd comes before the cdrom .

15.5.1 Changing boot order #

The VM Guest's boot order is represented through the order of devices in the XML configuration file. As the devices are interchangeable, it is possible to change the boot order of the VM Guest.

Open the VM Guest's XML configuration.
```
> sudo virsh edit sles15
```

Change the sequence of the bootable devices.

...
<boot dev='cdrom'/>
<boot dev='hd'/>
...

Check if the boot order was changed successfully by looking at the boot menu in the BIOS of the VM Guest.

15.5.2 Using direct kernel boot #

Direct Kernel Boot allows you to boot from a kernel and initrd stored on the host. Set the path to both files in the kernel and initrd elements:

<os>
    ...
  <kernel>/root/f8-i386-vmlinuz</kernel>
  <initrd>/root/f8-i386-initrd</initrd>
    ...
<os>

To enable Direct Kernel Boot:

Open the VM Guest's XML configuration:
```
> sudo virsh edit sles15
```
Inside the os element, add a kernel element and the path to the kernel file on the host:
```
...
<kernel>/root/f8-i386-vmlinuz</kernel>
...
```
Add an initrd element and the path to the initrd file on the host:
```
...
<initrd>/root/f8-i386-initrd</initrd>
...
```
Start your VM to boot from the new kernel:
```
> sudo virsh start sles15
```

15.6 Configuring memory allocation #

The amount of memory allocated for the VM Guest can also be configured with virsh. It is stored in the memory element and defines the maximum allocation of memory for the VM Guest at boot time. The optional currentMemory element defines the actual memory allocated to the VM Guest. currentMemory can be less than memory, allowing for increasing (or ballooning) the memory while the VM Guest is running. If currentMemory is omitted, it defaults to the same value as the memory element.

You can adjust memory settings by editing the VM Guest configuration, but be aware that changes do not take place until the next boot. The following steps demonstrate changing a VM Guest to boot with 4G of memory, but allow later expansion to 8G:

Open the VM Guest's XML configuration:
```
> sudo virsh edit sles15
```

Search for the memory element and set to 8G:

...
<memory unit='KiB'>8388608</memory>
...

If the currentMemory element does not exist, add it below the memory element, or change its value to 4G:
```
[...]
<memory unit='KiB'>8388608</memory>
<currentMemory unit='KiB'>4194304</currentMemory>
[...]
```

Changing the memory allocation while the VM Guest is running can be done with the setmem subcommand. The following example shows increasing the memory allocation to 8G:

Check VM Guest existing memory settings:

> sudo virsh dominfo sles15 | grep memory
Max memory:     8388608 KiB
Used memory:    4194608 KiB

Change the used memory to 8G:
```
> sudo virsh setmem sles15 8388608
```

Check the updated memory settings:

> sudo virsh dominfo sles15 | grep memory
Max memory:     8388608 KiB
Used memory:    8388608 KiB

Important: Large memory VM Guests

VM Guests with memory requirements of 4TB or more must either use the host-passthrough CPU mode, or explicitly specify the virtual CPU address size when using host-model or custom CPU modes. The default virtual CPU address size may not be sufficient for memory configurations of 4TB or more. The following example shows how to use the VM Host Server's physical CPU address size when using the host-model CPU mode.

[...]
<cpu mode='host-model' check='partial'>
<maxphysaddr mode='passthrough'>
</cpu>
[...]

For more information on specifying virtual CPU address size, see the maxphysaddr option in the CPU model and topology documentation at https://libvirt.org/formatdomain.html#cpu-model-and-topology.

15.7 Adding a PCI device #

To assign a PCI device to VM Guest with virsh, follow these steps:

Identify the host PCI device to assign to the VM Guest. In the following example, we are assigning a DEC network card to the guest:
```
> sudo lspci -nn
[...]
03:07.0 Ethernet controller [0200]: Digital Equipment Corporation DECchip \
21140 [FasterNet] [1011:0009] (rev 22)
[...]
```
Write down the device ID (03:07.0 in this case).

Gather detailed information about the device using virsh nodedev-dumpxml ID. To get the ID, replace the colon and the period in the device ID (03:07.0) with underscores. Prefix the result with “pci_0000_”: pci_0000_03_07_0.

> sudo virsh nodedev-dumpxml pci_0000_03_07_0
<device>
  <name>pci_0000_03_07_0</name>
  <path>/sys/devices/pci0000:00/0000:00:14.4/0000:03:07.0</path>
  <parent>pci_0000_00_14_4</parent>
  <driver>
    <name>tulip</name>
  </driver>
  <capability type='pci'>
    <domain>0</domain>
    <bus>3</bus>
    <slot>7</slot>
    <function>0</function>
    <product id='0x0009'>DECchip 21140 [FasterNet]</product>
    <vendor id='0x1011'>Digital Equipment Corporation</vendor>
    <numa node='0'/>
  </capability>
</device>

Write down the values for domain, bus, and function (see the previous XML code printed in bold).

Detach the device from the host system prior to attaching it to the VM Guest:
```
> sudo virsh nodedev-detach pci_0000_03_07_0
  Device pci_0000_03_07_0 detached
```
Tip: Multi-function PCI devices
When using a multi-function PCI device that does not support FLR (function level reset) or PM (power management) reset, you need to detach all its functions from the VM Host Server. The whole device must be reset for security reasons. libvirt will refuse to assign the device if one of its functions is still in use by the VM Host Server or another VM Guest.
Convert the domain, bus, slot, and function value from decimal to hexadecimal. In our example, domain = 0, bus = 3, slot = 7, and function = 0. Ensure that the values are inserted in the right order:
```
> printf "<address domain='0x%x' bus='0x%x' slot='0x%x' function='0x%x'/>\n" 0 3 7 0
```
This results in:
```
<address domain='0x0' bus='0x3' slot='0x7' function='0x0'/>
```
Run virsh edit on your domain, and add the following device entry in the <devices> section using the result from the previous step:
```
<hostdev mode='subsystem' type='pci' managed='yes'>
  <source>
    <address domain='0x0' bus='0x03' slot='0x07' function='0x0'/>
  </source>
</hostdev>
```
Tip: managed compared to unmanaged
libvirt recognizes two modes for handling PCI devices: they can be either managed or unmanaged. In the managed case, libvirt handles all details of unbinding the device from the existing driver if needed, resetting the device, binding it to vfio-pci before starting the domain, etc. When the domain is terminated or the device is removed from the domain, libvirt unbinds from vfio-pci and rebinds to the original driver in the case of a managed device. If the device is unmanaged, the user must ensure all of these management aspects of the device are done before assigning it to a domain, and after the device is no longer used by the domain.
In the example above, the managed='yes' option means that the device is managed. To switch the device mode to unmanaged, set managed='no' in the listing above. If you do so, you need to take care of the related driver with the virsh nodedev-detach and virsh nodedev-reattach commands. Prior to starting the VM Guest you need to detach the device from the host by running virsh nodedev-detach pci_0000_03_07_0. In case the VM Guest is not running, you can make the device available for the host by running virsh nodedev-reattach pci_0000_03_07_0.
Shut down the VM Guest and disable SELinux if it is running on the host.
```
> sudo setsebool -P virt_use_sysfs 1
```
Start your VM Guest to make the assigned PCI device available:
```
> sudo virsh start sles15
```

Important: SLES11 SP4 KVM guests

On a newer QEMU machine type (pc-i440fx-2.0 or higher) with SLES11 SP4 KVM guests, the acpiphp module is not loaded by default in the guest. This module must be loaded to enable hotplugging of disk and network devices. To load the module manually, use the command modprobe acpiphp. It is also possible to autoload the module by adding install acpiphp /bin/true to the /etc/modprobe.conf.local file.

Important: KVM guests using QEMU Q35 machine type

KVM guests using the QEMU Q35 machine type have a PCI topology that includes a pcie-root controller and seven pcie-root-port controllers. The pcie-root controller does not support hotplugging. Each pcie-root-port controller supports hotplugging a single PCIe device. PCI controllers cannot be hotplugged, so plan accordingly and add more pcie-root-ports if more than seven PCIe devices will be hotplugged. A pcie-to-pci-bridge controller can be added to support hotplugging legacy PCI devices. See https://libvirt.org/pci-hotplug.html for more information about PCI topology between QEMU machine types.

15.7.1 PCI Pass-Through for IBM Z #

In order to support IBM Z, QEMU extended PCI representation by allowing to configure extra attributes. Two more attributes—uid and fid—were added to the <zpci/> libvirt specification. uid represents user-defined identifier, while fid represents PCI function identifier. These attributes are optional and if you do not specify them, they are automatically generated with non-conflicting values.

To include zPCI attribute in your domain specification, use the following example definition:

<controller type='pci' index='0' model='pci-root'/>
<controller type='pci' index='1' model='pci-bridge'>
  <model name='pci-bridge'/>
  <target chassisNr='1'/>
  <address type='pci' domain='0x0000' bus='0x00' slot='0x01' function='0x0'>
    <zpci uid='0x0001' fid='0x00000000'/>
  </address>
</controller>
<interface type='bridge'>
  <source bridge='virbr0'/>
  <model type='virtio'/>
  <address type='pci' domain='0x0000' bus='0x01' slot='0x01' function='0x0'>
    <zpci uid='0x0007' fid='0x00000003'/>
  </address>
</interface>

15.8 Adding a USB device #

To assign a USB device to VM Guest using virsh, follow these steps:

Identify the host USB device to assign to the VM Guest:
```
> sudo lsusb
[...]
Bus 001 Device 003: ID 0557:2221 ATEN International Co., Ltd Winbond Hermon
[...]
```
Write down the vendor and product IDs. In our example, the vendor ID is 0557 and the product ID is 2221.
Run virsh edit on your domain, and add the following device entry in the <devices> section using the values from the previous step:
```
<hostdev mode='subsystem' type='usb'>
  <source startupPolicy='optional'>
   <vendor id='0557'/>
   <product id='2221'/>
  </source>
</hostdev>
```
Tip: Vendor/product or device's address
Instead of defining the host device with <vendor/> and <product/> IDs, you can use the <address/> element as described for host PCI devices in Section 15.7, “Adding a PCI device”.
Shut down the VM Guest and disable SELinux if it is running on the host:
```
> sudo setsebool -P virt_use_sysfs 1
```
Start your VM Guest to make the assigned PCI device available:
```
> sudo virsh start sles15
```

15.9 Adding SR-IOV devices #

Single Root I/O Virtualization (SR-IOV) capable PCIe devices can replicate their resources, so they appear to be multiple devices. Each of these “pseudo-devices” can be assigned to a VM Guest.

SR-IOV is an industry specification that was created by the Peripheral Component Interconnect Special Interest Group (PCI-SIG) consortium. It introduces physical functions (PF) and virtual functions (VF). PFs are full PCIe functions used to manage and configure the device. PFs also can move data. VFs lack the configuration and management part—they only can move data and a reduced set of configuration functions. As VFs do not have all PCIe functions, the host operating system or the Hypervisor must support SR-IOV to be able to access and initialize VFs. The theoretical maximum for VFs is 256 per device (consequently the maximum for a dual-port Ethernet card would be 512). In practice this maximum is much lower, since each VF consumes resources.

15.9.1 Requirements #

The following requirements must be met to use SR-IOV:

An SR-IOV-capable network card (as of SUSE Linux Enterprise Server 15, only network cards support SR-IOV)
An AMD64/Intel 64 host supporting hardware virtualization (AMD-V or Intel VT-x), see Section 7.1.1, “KVM hardware requirements” for more information
A chipset that supports device assignment (AMD-Vi or Intel VT-d)
libvirt 0.9.10 or better
SR-IOV drivers must be loaded and configured on the host system
A host configuration that meets the requirements listed at Important: Requirements for VFIO and SR-IOV
A list of the PCI addresses of the VF(s) that will be assigned to VM Guests

Tip: Checking if a device is SR-IOV-capable

The information whether a device is SR-IOV-capable can be obtained from its PCI descriptor by running lspci. A device that supports SR-IOV reports a capability similar to the following:

Capabilities: [160 v1] Single Root I/O Virtualization (SR-IOV)

Note: Adding an SR-IOV device at VM Guest creation

Before adding an SR-IOV device to a VM Guest when initially setting it up, the VM Host Server already needs to be configured as described in Section 15.9.2, “Loading and configuring the SR-IOV host drivers”.

15.9.2 Loading and configuring the SR-IOV host drivers #

To access and initialize VFs, an SR-IOV-capable driver needs to be loaded on the host system.

Before loading the driver, make sure the card is properly detected by running lspci. The following example shows the lspci output for the dual-port Intel 82576NS network card:
```
> sudo /sbin/lspci | grep 82576
01:00.0 Ethernet controller: Intel Corporation 82576NS Gigabit Network Connection (rev 01)
01:00.1 Ethernet controller: Intel Corporation 82576NS Gigabit Network Connection (rev 01)
04:00.0 Ethernet controller: Intel Corporation 82576NS Gigabit Network Connection (rev 01)
04:00.1 Ethernet controller: Intel Corporation 82576NS Gigabit Network Connection (rev 01)
```
In case the card is not detected, it is likely that the hardware virtualization support in the BIOS/EFI has not been enabled. To check if hardware virtualization support is enabled, look at the settings in the host's BIOS.
Check whether the SR-IOV driver is already loaded by running lsmod. In the following example a check for the igb driver (for the Intel 82576NS network card) returns a result. That means the driver is already loaded. If the command returns nothing, the driver is not loaded.
```
> sudo /sbin/lsmod | egrep "^igb "
igb                   185649  0
```
Skip the following step if the driver is already loaded. If the SR-IOV driver is not yet loaded, the non-SR-IOV driver needs to be removed first, before loading the new driver. Use rmmod to unload a driver. The following example unloads the non-SR-IOV driver for the Intel 82576NS network card:
```
> sudo /sbin/rmmod igbvf
```
Load the SR-IOV driver subsequently using the modprobe command—the VF parameter (max_vfs) is mandatory:
```
> sudo /sbin/modprobe igb max_vfs=8
```

As an alternative, you can also load the driver via SYSFS:

Find the PCI ID of the physical NIC by listing Ethernet devices:

> sudo lspci | grep Eth
06:00.0 Ethernet controller: Emulex Corporation OneConnect NIC (Skyhawk) (rev 10)
06:00.1 Ethernet controller: Emulex Corporation OneConnect NIC (Skyhawk) (rev 10)

To enable VFs, echo the number of desired VFs to load to the sriov_numvfs parameter:
```
> sudo echo 1 > /sys/bus/pci/devices/0000:06:00.1/sriov_numvfs
```

Verify that the VF NIC was loaded:

> sudo lspci | grep Eth
06:00.0 Ethernet controller: Emulex Corporation OneConnect NIC (Skyhawk) (rev 10)
06:00.1 Ethernet controller: Emulex Corporation OneConnect NIC (Skyhawk) (rev 10)
06:08.0 Ethernet controller: Emulex Corporation OneConnect NIC (Skyhawk) (rev 10)

Obtain the maximum number of VFs available:

> sudo lspci -vvv -s 06:00.1 | grep 'Initial VFs'
                       Initial VFs: 32, Total VFs: 32, Number of VFs: 0,
Function Dependency Link: 01

Create a /etc/systemd/system/before.service file which loads VF via SYSFS on boot:

[Unit]
Before=
[Service]
Type=oneshot
RemainAfterExit=true
ExecStart=/bin/bash -c "echo 1 > /sys/bus/pci/devices/0000:06:00.1/sriov_numvfs"
# beware, executable is run directly, not through a shell, check the man pages
# systemd.service and systemd.unit for full syntax
[Install]
# target in which to start the service
WantedBy=multi-user.target
#WantedBy=graphical.target

Prior to starting the VM, it is required to create another service file (after-local.service) pointing to the /etc/init.d/after.local script that detaches the NIC. Otherwise the VM would fail to start:

[Unit]
Description=/etc/init.d/after.local Compatibility
After=libvirtd.service
Requires=libvirtd.service
[Service]
Type=oneshot
ExecStart=/etc/init.d/after.local
RemainAfterExit=true

[Install]
WantedBy=multi-user.target

Copy it to /etc/systemd/system.
```
#! /bin/sh
# ...
virsh nodedev-detach pci_0000_06_08_0
```
Save it as /etc/init.d/after.local.

Reboot the machine and check if the SR-IOV driver is loaded by re-running the lspci command from the first step of this procedure. If the SR-IOV driver was loaded successfully you should see additional lines for the VFs:

01:00.0 Ethernet controller: Intel Corporation 82576NS Gigabit Network Connection (rev 01)
01:00.1 Ethernet controller: Intel Corporation 82576NS Gigabit Network Connection (rev 01)
01:10.0 Ethernet controller: Intel Corporation 82576 Virtual Function (rev 01)
01:10.1 Ethernet controller: Intel Corporation 82576 Virtual Function (rev 01)
01:10.2 Ethernet controller: Intel Corporation 82576 Virtual Function (rev 01)
[...]
04:00.0 Ethernet controller: Intel Corporation 82576NS Gigabit Network Connection (rev 01)
04:00.1 Ethernet controller: Intel Corporation 82576NS Gigabit Network Connection (rev 01)
04:10.0 Ethernet controller: Intel Corporation 82576 Virtual Function (rev 01)
04:10.1 Ethernet controller: Intel Corporation 82576 Virtual Function (rev 01)
04:10.2 Ethernet controller: Intel Corporation 82576 Virtual Function (rev 01)
[...]

15.9.3 Adding a VF network device to a VM Guest #

When the SR-IOV hardware is properly set up on the VM Host Server, you can add VFs to VM Guests. To do so, you need to collect some data first.

Procedure 15.2: Adding a VF network device to an existing VM Guest #

The following procedure uses example data. Make sure to replace it by appropriate data from your setup.

Use the virsh nodedev-list command to get the PCI address of the VF you want to assign and its corresponding PF. Numerical values from the lspci output shown in Section 15.9.2, “Loading and configuring the SR-IOV host drivers” (for example 01:00.0 or 04:00.1) are transformed by adding the prefix "pci_0000_" and by replacing colons and dots with underscores. So a PCI ID listed as "04:00.0" by lspci is listed as "pci_0000_04_00_0" by virsh. The following example lists the PCI IDs for the second port of the Intel 82576NS network card:
```
> sudo virsh nodedev-list | grep 0000_04_
pci_0000_04_00_0
pci_0000_04_00_1
pci_0000_04_10_0
pci_0000_04_10_1
pci_0000_04_10_2
pci_0000_04_10_3
pci_0000_04_10_4
pci_0000_04_10_5
pci_0000_04_10_6
pci_0000_04_10_7
pci_0000_04_11_0
pci_0000_04_11_1
pci_0000_04_11_2
pci_0000_04_11_3
pci_0000_04_11_4
pci_0000_04_11_5
```
The first two entries represent the PFs, whereas the other entries represent the VFs.

Run the following virsh nodedev-dumpxml command on the PCI ID of the VF you want to add:

> sudo virsh nodedev-dumpxml pci_0000_04_10_0
<device>
  <name>pci_0000_04_10_0</name>
  <parent>pci_0000_00_02_0</parent>
  <capability type='pci'>
    <domain>0</domain>
    <bus>4</bus>
    <slot>16</slot>
    <function>0</function>
    <product id='0x10ca'>82576 Virtual Function</product>
    <vendor id='0x8086'>Intel Corporation</vendor>
    <capability type='phys_function'>
      <address domain='0x0000' bus='0x04' slot='0x00' function='0x0'/>
    </capability>
  </capability>
</device>

The following data is needed for the next step:

<domain>0</domain>
<bus>4</bus>
<slot>16</slot>
<function>0</function>

Create a temporary XML file (for example /tmp/vf-interface.xml containing the data necessary to add a VF network device to an existing VM Guest. The minimal content of the file needs to look like the following:

<interface type='hostdev'>1
 <source>
  <address type='pci' domain='0' bus='11' slot='16' function='0'2/>2
 </source>
</interface>

VFs do not get a fixed MAC address; it changes every time the host reboots. When adding network devices the “traditional” way with hostdev, it would require to reconfigure the VM Guest's network device after each reboot of the host, because of the MAC address change. To avoid this kind of problem, libvirt introduced the hostdev value, which sets up network-specific data before assigning the device.

Specify the data you acquired in the previous step here.

In case a device is already attached to the host, it cannot be attached to a VM Guest. To make it available for guests, detach it from the host first:
```
> sudo virsh nodedev-detach pci_0000_04_10_0
```
Add the VF interface to an existing VM Guest:
```
> sudo virsh attach-device GUEST /tmp/vf-interface.xml --OPTION
```
GUEST needs to be replaced by the domain name, ID or UUID of the VM Guest. --OPTION can be one of the following:
--persistent
This option will always add the device to the domain's persistent XML. In addition, if the domain is running, it will be hotplugged.
--config
This option will only affect the persistent XML, even if the domain is running. The device will only show up in the VM Guest on next boot.
--live
This option will only affect a running domain. If the domain is inactive, the operation will fail. The device is not persisted in the XML and will not be available in the VM Guest on next boot.
--current
This option affects the current state of the domain. If the domain is inactive, the device is added to the persistent XML and will be available on next boot. If the domain is active, the device is hotplugged but not added to the persistent XML.
To detach a VF interface, use the virsh detach-device command, which also takes the options listed above.

15.9.4 Dynamic allocation of VFs from a pool #

If you define the PCI address of a VF into a VM Guest's configuration statically as described in Section 15.9.3, “Adding a VF network device to a VM Guest”, it is hard to migrate such guest to another host. The host must have identical hardware in the same location on the PCI bus, or the VM Guest configuration must be modified prior to each start.

Another approach is to create a libvirt network with a device pool that contains all the VFs of an SR-IOV device. The VM Guest then references this network, and each time it is started, a single VF is dynamically allocated to it. When the VM Guest is stopped, the VF is returned to the pool, available for another guest.

15.9.4.1 Defining network with pool of VFs on VM Host Server #

The following example of network definition creates a pool of all VFs for the SR-IOV device with its physical function (PF) at the network interface eth0 on the host:

<network>
  <name>passthrough</name>
    <forward mode='hostdev' managed='yes'>
      <pf dev='eth0'/>
    </forward>
  </network>

To use this network on the host, save the above code to a file, for example /tmp/passthrough.xml, and execute the following commands. Remember to replace eth0 with the real network interface name of your SR-IOV device's PF:

> sudo virsh net-define /tmp/passthrough.xml
> sudo virsh net-autostart passthrough
> sudo virsh net-start passthrough

15.9.4.2 Configuring VM Guests to use VF from the pool #

The following example of VM Guest device interface definition uses a VF of the SR-IOV device from the pool created in Section 15.9.4.1, “Defining network with pool of VFs on VM Host Server”. libvirt automatically derives the list of all VFs associated with that PF the first time the guest is started.

<interface type='network'>
  <source network='passthrough'>
</interface>

After the first VM Guest starts that uses the network with the pool of VFs, verify the list of associated VFs. Do so by running virsh net-dumpxml passthrough on the host.

<network connections='1'>
  <name>passthrough</name>
  <uuid>a6a26429-d483-d4ed-3465-4436ac786437</uuid>
  <forward mode='hostdev' managed='yes'>
    <pf dev='eth0'/>
    <address type='pci' domain='0x0000' bus='0x02' slot='0x10' function='0x1'/>
    <address type='pci' domain='0x0000' bus='0x02' slot='0x10' function='0x3'/>
    <address type='pci' domain='0x0000' bus='0x02' slot='0x10' function='0x5'/>
    <address type='pci' domain='0x0000' bus='0x02' slot='0x10' function='0x7'/>
    <address type='pci' domain='0x0000' bus='0x02' slot='0x11' function='0x1'/>
    <address type='pci' domain='0x0000' bus='0x02' slot='0x11' function='0x3'/>
    <address type='pci' domain='0x0000' bus='0x02' slot='0x11' function='0x5'/>
  </forward>
  </network>

15.10 Listing attached devices #

Although there is no mechanism in virsh to list all VM Host Server's devices that have already been attached to its VM Guests, you can list all devices attached to a specific VM Guest by running the following command:

virsh dumpxml VMGUEST_NAME | xpath -e /domain/devices/hostdev

For example:

> sudo virsh dumpxml sles12 | -e xpath /domain/devices/hostdev
Found 2 nodes:
-- NODE --
<hostdev mode="subsystem" type="pci" managed="yes">
  <driver name="xen" />
  <source>
    <address domain="0x0000" bus="0x0a" slot="0x10" function="0x1" />
  </source>
  <address type="pci" domain="0x0000" bus="0x00" slot="0x0a" function="0x0" />
  </hostdev>
-- NODE --
<hostdev mode="subsystem" type="pci" managed="yes">
  <driver name="xen" />
  <source>
    <address domain="0x0000" bus="0x0a" slot="0x10" function="0x2" />
  </source>
  <address type="pci" domain="0x0000" bus="0x00" slot="0x0b" function="0x0" />
</hostdev>

Tip: Listing SR-IOV devices attached via <interface type='hostdev'>

For SR-IOV devices that are attached to the VM Host Server by means of <interface type='hostdev'>, you need to use a different XPath query:

virsh dumpxml VMGUEST_NAME | xpath -e /domain/devices/interface/@type

15.11 Configuring storage devices #

Storage devices are defined within the disk element. The usual disk element supports several attributes. The following two attributes are the most important:

The type attribute describes the source of the virtual disk device. Valid values are file , block , dir , network , or volume .
The device attribute indicates how the disk is exposed to the VM Guest OS. As an example, possible values can include floppy , disk , cdrom , and others.

The following child elements are the most important:

driver contains the driver and the bus. These are used by the VM Guest to work with the new disk device.
The target element contains the device name under which the new disk is shown in the VM Guest. It also contains the optional bus attribute, which defines the type of bus on which the new disk should operate.

The following procedure shows how to add storage devices to the VM Guest:

Edit the configuration for an existing VM Guest:
```
> sudo virsh edit sles15
```
Add a disk element inside the disk element together with the attributes type and device:
```
<disk type='file' device='disk'>
```
Specify a driver element and use the default values:
```
<driver name='qemu' type='qcow2'/>
```
Create a disk image, which will be used as a source for the new virtual disk device:
```
> sudo qemu-img create -f qcow2 /var/lib/libvirt/images/sles15.qcow2 32G
```

Add the path for the disk source:

<source file='/var/lib/libvirt/images/sles15.qcow2'/>

Define the target device name in the VM Guest and the bus on which the disk should work:
```
<target dev='vda' bus='virtio'/>
```
Restart your VM:
```
> sudo virsh start sles15
```

Your new storage device should be available in the VM Guest OS.

15.12 Configuring controller devices #

libvirt generally manages controllers automatically based on the type of virtual devices used by the VM Guest. If the VM Guest contains PCI and SCSI devices, PCI and SCSI controllers will be created and managed automatically. libvirt will also model controllers that are hypervisor-specific, for example, a virtio-serial controller for KVM VM Guests or a xenbus controller for Xen VM Guests. Although the default controllers and their configuration are generally fine, there may be use cases where controllers or their attributes need to be adjusted manually. For example, a virtio-serial controller may need more ports, or a xenbus controller may need more memory or more virtual interrupts.

The xenbus controller is unique in that it serves as the controller for all Xen paravirtual devices. If a VM Guest has many disk and/or network devices, the controller may need more memory. Xen's max_grant_frames attribute sets how many grant frames, or blocks of shared memory, are allocated to the xenbus controller for each VM Guest.

The default of 32 is enough in most circumstances, but a VM Guest with a large number of I/O devices and an I/O-intensive workload may experience performance issues because of grant frame exhaustion. The xen-diag can be used to check the current and maximum max_grant_frames values for dom0 and your VM Guests. The VM Guests must be running:

> sudo virsh list
 Id   Name             State
--------------------------------
 0    Domain-0         running
 3    sle15sp1         running

 > sudo xen-diag gnttab_query_size 0
domid=0: nr_frames=1, max_nr_frames=256

> sudo xen-diag gnttab_query_size 3
domid=3: nr_frames=3, max_nr_frames=32

The sle15sp1 guest is using only three frames out of 32. If you are seeing performance issues, and log entries that point to insufficient frames, increase the value with virsh. Look for the <controller type='xenbus'> line in the guest's configuration file, and add the maxGrantFrames control element:

> sudo virsh edit sle15sp1
 <controller type='xenbus' index='0' maxGrantFrames='40'/>

Save your changes and restart the guest. Now it should show your change:

> sudo xen-diag gnttab_query_size 3
domid=3: nr_frames=3, max_nr_frames=40

Similar to maxGrantFrames, the xenbus controller also supports maxEventChannels. Event channels are like paravirtual interrupts, and in conjunction with grant frames, form a data transfer mechanism for paravirtual drivers. They are also used for inter-processor interrupts. VM Guests with a large number of vCPUs and/or many paravirtual devices may need to increase the maximum default value of 1023. maxEventChannels can be changed similarly to maxGrantFrames:

> sudo virsh edit sle15sp1
 <controller type='xenbus' index='0' maxGrantFrames='128' maxEventChannels='2047'/>

See the Controllers section of the libvirt Domain XML format manual at https://libvirt.org/formatdomain.html#elementsControllers for more information.

15.13 Configuring video devices #

When using the Virtual Machine Manager, only the Video device model can be defined. The amount of allocated VRAM or 2D/3D acceleration can only be changed in the XML configuration.

15.13.1 Changing the amount of allocated VRAM #

Edit the configuration for an existing VM Guest:
```
> sudo virsh edit sles15
```

Change the size of the allocated VRAM:

<video>
<model type='vga' vram='65535' heads='1'>
...
</model>
</video>

Check if the amount of VRAM in the VM has changed by looking at the amount in the Virtual Machine Manager.

15.13.2 Changing the state of 2D/3D acceleration #

Edit the configuration for an existing VM Guest:
```
> sudo virsh edit sles15
```
To enable/disable 2D/3D acceleration, change the value of accel3d and accel2d accordingly:
```
<video>
 <model>
  <acceleration accel3d='yes' accel2d='no'>
 </model>
</video>
```

Tip: Enabling 2D/3D acceleration

Only virtio and vbox video devices are capable of 2D/3D acceleration. You cannot enable it on other video devices.

15.14 Configuring network devices #

This section describes how to configure specific aspects of virtual network devices by using virsh.

Find more details about libvirt network interface specification in https://libvirt.org/formatdomain.html#elementsDriverBackendOptions.

15.14.1 Scaling network performance with multiqueue virtio-net #

The multiqueue virtio-net feature scales the network performance by allowing the VM Guest's virtual CPUs to transfer packets in parallel. Refer to Section 33.3.3, “Scaling network performance with multiqueue virtio-net” for more general information.

To enable multiqueue virtio-net for a specific VM Guest, edit its XML configuration as described in Section 15.1, “Editing the VM configuration” and modify its network interface as follows:

<interface type='network'>
 [...]
 <model type='virtio'/>
 <driver name='vhost' queues='NUMBER_OF_QUEUES'/>
</interface>

15.15 Using macvtap to share VM Host Server network interfaces #

Macvtap provides direct attachment of a VM Guest virtual interface to a host network interface. The macvtap-based interface extends the VM Host Server network interface and has its own MAC address on the same Ethernet segment. Typically, this is used to make both the VM Guest and the VM Host Server show up directly on the switch that the VM Host Server is connected to.

Note: Macvtap cannot be used with a Linux bridge

Macvtap cannot be used with network interfaces already connected to a Linux bridge. Before attempting to create the macvtap interface, remove the interface from the bridge.

Note: VM Guest to VM Host Server communication with macvtap

When using macvtap, a VM Guest can communicate with other VM Guests, and with other external hosts on the network. But it cannot communicate with the VM Host Server on which the VM Guest runs. This is the defined behavior of macvtap, because of the way the VM Host Server's physical Ethernet is attached to the macvtap bridge. Traffic from the VM Guest into that bridge that is forwarded to the physical interface cannot be bounced back up to the VM Host Server's IP stack. Similarly, traffic from the VM Host Server's IP stack that is sent to the physical interface cannot be bounced back up to the macvtap bridge for forwarding to the VM Guest.

Virtual network interfaces based on macvtap are supported by libvirt by specifying an interface type of direct. For example:

<interface type='direct'>
   <mac address='aa:bb:cc:dd:ee:ff'/>
   <source dev='eth0' mode='bridge'/>
   <model type='virtio'/>
   </interface>

The operation mode of the macvtap device can be controlled with the mode attribute. The following list show its possible values and a description for each:

vepa: All VM Guest packets are sent to an external bridge. Packets whose destination is a VM Guest on the same VM Host Server as where the packet originates from are sent back to the VM Host Server by the VEPA capable bridge (today's bridges are typically not VEPA capable).
bridge: Packets whose destination is on the same VM Host Server as where they originate from are directly delivered to the target macvtap device. Both origin and destination devices need to be in bridge mode for direct delivery. If either one of them is in vepa mode, a VEPA capable bridge is required.
private: All packets are sent to the external bridge and will only be delivered to a target VM Guest on the same VM Host Server if they are sent through an external router or gateway and that device sends them back to the VM Host Server. This procedure is followed if either the source or destination device is in private mode.
passthrough: A special mode that gives more power to the network interface. All packets will be forwarded to the interface, allowing virtio VM Guests to change the MAC address or set promiscuous mode to bridge the interface or create VLAN interfaces on top of it. Note that a network interface is not shareable in passthrough mode. Assigning an interface to a VM Guest will disconnect it from the VM Host Server. For this reason SR-IOV virtual functions are often assigned to the VM Guest in passthrough mode.

15.16 Disabling a memory balloon device #

Memory Balloon has become a default option for KVM. The device will be added to the VM Guest explicitly, so you do not need to add this element in the VM Guest's XML configuration. However, if you want to disable Memory Balloon in the VM Guest for any reason, you need to set model='none' as shown below:

<devices>
   <memballoon model='none'/>
</device>

15.17 Configuring multiple monitors (dual head) #

libvirt supports a dual head configuration to display the video output of the VM Guest on multiple monitors.

Important: No support for Xen

The Xen hypervisor does not support dual head configuration.

Procedure 15.3: Configuring dual head #

While the virtual machine is running, verify that the xf86-video-qxl package is installed in the VM Guest:
```
> rpm -q xf86-video-qxl
```
Shut down the VM Guest and start editing its configuration XML as described in Section 15.1, “Editing the VM configuration”.
Verify that the model of the virtual graphics card is 'qxl':
```
<video>
 <model type='qxl' ... />
```

Increase the heads parameter in the graphics card model specification from the default 1 to 2, for example:

<video>
 <model type='qxl' ram='65536' vram='65536' vgamem='16384' heads='2' primary='yes'/>
 <alias name='video0'/>
 <address type='pci' domain='0x0000' bus='0x00' slot='0x01' function='0x0'/>
</video>

Configure the virtual machine to use the Spice display instead of VNC:

<graphics type='spice' port='5916' autoport='yes' listen='0.0.0.0'>
 <listen type='address' address='0.0.0.0'/>
</graphics>

Start the virtual machine and connect to its display with virt-viewer, for example:
```
> virt-viewer --connect qemu+ssh://USER@VM_HOST/system
```
From the list of VMs, select the one whose configuration you have modified and confirm with Connect.
After the graphical subsystem (Xorg) loads in the VM Guest, select View › Displays › Display 2 to open a new window with the second monitor's output.

15.18 Crypto adapter pass-through to KVM guests on IBM Z #

15.18.1 Introduction #

IBM Z machines include cryptographic hardware with useful functions such as random number generation, digital signature generation, or encryption. KVM allows dedicating these crypto adapters to guests as pass-through devices. The means that the hypervisor cannot observe communications between the guest and the device.

15.18.2 What is covered #

You will learn how to dedicate a crypto adapter and domains on an IBM Z host to a KVM guest. The procedure includes the following basic steps:

Mask the crypto adapter and domains from the default driver on the host.
Load the vfio-ap driver.
Assign the crypto adapter and domains to the vfio-ap driver.
Configure the guest to use the crypto adapter.

15.18.3 Requirements #

You need to have the QEMU / libvirt virtualization environment correctly installed and functional.
The vfio_ap and vfio_mdev modules for the running kernel need to be available on the host operating system.

15.18.4 Dedicate a crypto adapter to a KVM host #

Verify that the vfio_ap and vfio_mdev kernel modules are loaded on the host:
```
> lsmod | grep vfio_
```
If any of them is not listed, load it manually, for example:
```
> sudo modprobe vfio_mdev
```

Create a new MDEV device on the host and verify that it was added:

uuid=$(uuidgen)
$ echo ${uuid} | sudo tee /sys/devices/vfio_ap/matrix/mdev_supported_types/vfio_ap-passthrough/create
dmesg | tail
[...]
[272197.818811] iommu: Adding device 24f952b3-03d1-4df2-9967-0d5f7d63d5f2 to group 0
[272197.818815] vfio_mdev 24f952b3-03d1-4df2-9967-0d5f7d63d5f2: MDEV: group_id = 0

Identify the device on the host's logical partition that you intend to dedicate to a KVM guest:
```
> ls -l /sys/bus/ap/devices/
[...]
lrwxrwxrwx 1 root root 0 Nov 23 03:29 00.0016 -> ../../../devices/ap/card00/00.0016/
lrwxrwxrwx 1 root root 0 Nov 23 03:29 card00 -> ../../../devices/ap/card00/
```
In this example, it is card 0 queue 16. To match the Hardware Management Console (HMC) configuration, you need to convert from 16 hexadecimal to 22 decimal.

Mask the adapter from the zcrypt use:

> lszcrypt
CARD.DOMAIN TYPE MODE STATUS REQUEST_CNT
-------------------------------------------------
00 CEX5C CCA-Coproc online 5
00.0016 CEX5C CCA-Coproc online 5

Mask the adapter:

> cat /sys/bus/ap/apmask
0xffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff
echo -0x0 | sudo tee /sys/bus/ap/apmask
0x7fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff

Mask the domain:

> cat /sys/bus/ap/aqmask
0xffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff
echo -0x0 | sudo tee /sys/bus/ap/aqmask
0xfffffdffffffffffffffffffffffffffffffffffffffffffffffffffffffffff

Assign adapter 0 and domain 16 (22 decimal) to vfio-ap:

> sudo echo +0x0 > /sys/devices/vfio_ap/matrix/${uuid}/assign_adapter
> echo +0x16 | sudo tee /sys/devices/vfio_ap/matrix/${uuid}/assign_domain
> echo +0x16 | sudo tee /sys/devices/vfio_ap/matrix/${uuid}/assign_control_domain

Verify the matrix that you have configured:

> cat /sys/devices/vfio_ap/matrix/${uuid}/matrix
00.0016

Either create a new VM (refer to Chapter 10, Guest installation) and wait until it is initialized, or use an existing VM. In both cases, make sure the VM is shut down.

Change its configuration to use the MDEV device:

> sudo virsh edit VM_NAME
[...]
<hostdev mode='subsystem' type='mdev' model='vfio-ap'>
 <source>
  <address uuid='24f952b3-03d1-4df2-9967-0d5f7d63d5f2'/>
 </source>
</hostdev>
[...]

Restart the VM:
```
> sudo virsh reboot VM_NAME
```

> lszcrypt
CARD.DOMAIN TYPE MODE STATUS REQUEST_CNT
-------------------------------------------------
00 CEX5C CCA-Coproc online 1
00.0016 CEX5C CCA-Coproc online 1

15.18.5 Further reading #

The installation of virtualization components is detailed in Chapter 6, Installation of virtualization components.
The vfio_ap architecture is detailed in https://www.kernel.org/doc/Documentation/s390/vfio-ap.txt.
A general outline together with a detailed procedure is described in https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1787405.
The architecture of VFIO Mediated devices (MDEVs) is detailed in https://www.kernel.org/doc/html/latest/driver-api/vfio-mediated-device.html.