KVM supports passing through PCI devices to its guests. A typical libvirt xml statement looks like
<hostdev mode='subsystem' type='pci' managed='yes'>
<driver name='vfio'/>
<source>
<address domain='0x0012' bus='0x00' slot='0x00' function='0x0'/>
</source>
<address type='pci' domain='0x0000' bus='0x00' slot='0x01' function='0x0'>
<zpci uid='0x0001' fid='0x00000000'/>
</address>
</hostdev>
<driver name='vfio'/>
<source>
<address domain='0x0012' bus='0x00' slot='0x00' function='0x0'/>
</source>
<address type='pci' domain='0x0000' bus='0x00' slot='0x01' function='0x0'>
<zpci uid='0x0001' fid='0x00000000'/>
</address>
</hostdev>
See KVM Virtual Server Management SC34-2752 for details.
With newer kernel and QEMU versions (e.g. those found in RHEL 10.2, SLES 16.1 and Ubuntu 25.10) it is now possible to improve performance and reduce overhead when doing so.
You have to specify the parameter iommu.passthrough=1 on the guest kernel command line via zipl or /boot/loader/entries.
This will pin the guest in memory but gets rid of the RPCIT instructions for IO that are used by the guest to notify the host about memory used for IO to control host paging.
As KVM on Z does pin the guest in memory anyway as of today when a PCI device is passed through, this has no immediate downside.
By checking in the guest
cat /sys/bus/pci/devices/<device number>/iommu_group/type
it is possible to check if everything worked. The file should say "identity",
