20.40. Guest Virtual Machine CPU Model Configuration

20.40.1. Introduction

Every hypervisor has its own policy for what a guest virtual machine will see for its CPUs by default. Whereas some hypervisors decide which CPU host physical machine features will be available for the guest virtual machine, QEMU/KVM presents the guest virtual machine with a generic model named qemu32 or qemu64. These hypervisors perform more advanced filtering, classifying all physical CPUs into a handful of groups and have one baseline CPU model for each group that is presented to the guest virtual machine. Such behavior enables the safe migration of guest virtual machines between host physical machines, provided they all have physical CPUs that classify into the same group. libvirt does not typically enforce policy itself, rather it provides the mechanism on which the higher layers define their own required policy. Understanding how to obtain CPU model information and define a suitable guest virtual machine CPU model is critical to ensure guest virtual machine migration is successful between host physical machines. Note that a hypervisor can only emulate features that it is aware of and features that were created after the hypervisor was released may not be emulated.

20.40.2. Learning about the Host Physical Machine CPU Model

The virsh capabilities command displays an XML document describing the capabilities of the hypervisor connection and host physical machine. The XML schema displayed has been extended to provide information about the host physical machine CPU model. One of the big challenges in describing a CPU model is that every architecture has a different approach to exposing their capabilities. QEMU/KVM and libvirt use a scheme which combines a CPU model name string, with a set of named flags.
It is not practical to have a database listing all known CPU models, so libvirt has a small list of baseline CPU model names. It chooses the one that shares the greatest number of CPUID bits with the actual host physical machine CPU and then lists the remaining bits as named features. Notice that libvirt does not display which features the baseline CPU contains. This might seem like a flaw at first, but as will be explained in this section, it is not actually necessary to know this information.

20.40.3. Determining Support for VFIO IOMMU Devices

Use the virsh domcapabilities command to determine support for VFIO. See the following example output: 

# virsh domcapabilities

[...output truncated...]

<enum name='pciBackend'>
        <value>default</value>
        <value>vfio</value>

[...output truncated...]

Figure 20.3. Determining support for VFIO

20.40.4. Determining a Compatible CPU Model to Suit a Pool of Host Physical Machines

Now that it is possible to find out what CPU capabilities a single host physical machine has, the next step is to determine what CPU capabilities are best to expose to the guest virtual machine. If it is known that the guest virtual machine will never need to be migrated to another host physical machine, the host physical machine CPU model can be passed straight through unmodified. A virtualized data center may have a set of configurations that can guarantee all servers will have 100% identical CPUs. Again the host physical machine CPU model can be passed straight through unmodified. The more common case, though, is where there is variation in CPUs between host physical machines. In this mixed CPU environment, the lowest common denominator CPU must be determined. This is not entirely straightforward, so libvirt provides an API for exactly this task. If libvirt is provided a list of XML documents, each describing a CPU model for a host physical machine, libvirt will internally convert these to CPUID masks, calculate their intersection, and convert the CPUID mask result back into an XML CPU description.
Here is an example of what libvirt reports as the capabilities on a basic workstation, when the virsh capabilities is executed: 

<capabilities>
  <host>
    <cpu>
      <arch>i686</arch>
      <model>pentium3</model>
      <topology sockets='1' cores='2' threads='1'/>
      <feature name='lahf_lm'/>
      <feature name='lm'/>
      <feature name='xtpr'/>
      <feature name='cx16'/>
      <feature name='ssse3'/>
      <feature name='tm2'/>
      <feature name='est'/>
      <feature name='vmx'/>
      <feature name='ds_cpl'/>
      <feature name='monitor'/>
      <feature name='pni'/>
      <feature name='pbe'/>
      <feature name='tm'/>
      <feature name='ht'/>
      <feature name='ss'/>
      <feature name='sse2'/>
      <feature name='acpi'/>
      <feature name='ds'/>
      <feature name='clflush'/>
      <feature name='apic'/>
    </cpu>
 </host>
</capabilities>

Figure 20.4. Pulling host physical machine's CPU model information

Now compare that to a different server, with the same virsh capabilities command: 

<capabilities>
  <host>
    <cpu>
      <arch>x86_64</arch>
      <model>phenom</model>
      <topology sockets='2' cores='4' threads='1'/>
      <feature name='osvw'/>
      <feature name='3dnowprefetch'/>
      <feature name='misalignsse'/>
      <feature name='sse4a'/>
      <feature name='abm'/>
      <feature name='cr8legacy'/>
      <feature name='extapic'/>
      <feature name='cmp_legacy'/>
      <feature name='lahf_lm'/>
      <feature name='rdtscp'/>
      <feature name='pdpe1gb'/>
      <feature name='popcnt'/>
      <feature name='cx16'/>
      <feature name='ht'/>
      <feature name='vme'/>
    </cpu>
    ...snip...

Figure 20.5. Generate CPU description from a random server

To see if this CPU description is compatible with the previous workstation CPU description, use the virsh cpu-compare command.
The reduced content was stored in a file named virsh-caps-workstation-cpu-only.xml and the virsh cpu-compare command can be executed on this file: 
# virsh cpu-compare virsh-caps-workstation-cpu-only.xml
Host physical machine CPU is a superset of CPU described in virsh-caps-workstation-cpu-only.xml
As seen in this output, libvirt is correctly reporting that the CPUs are not strictly compatible. This is because there are several features in the server CPU that are missing in the client CPU. To be able to migrate between the client and the server, it will be necessary to open the XML file and comment out some features. To determine which features need to be removed, run the virsh cpu-baseline command, on the both-cpus.xml which contains the CPU information for both machines. Running # virsh cpu-baseline both-cpus.xml results in: 

<cpu match='exact'>
  <model>pentium3</model>
  <feature policy='require' name='lahf_lm'/>
  <feature policy='require' name='lm'/>
  <feature policy='require' name='cx16'/>
  <feature policy='require' name='monitor'/>
  <feature policy='require' name='pni'/>
  <feature policy='require' name='ht'/>
  <feature policy='require' name='sse2'/>
  <feature policy='require' name='clflush'/>
  <feature policy='require' name='apic'/>
</cpu>

Figure 20.6. Composite CPU baseline

This composite file shows which elements are in common. Everything that is not in common should be commented out.