Add mocked disk device to test pci passthrough

[kwu83tw]

Add mocked disk device to test pci passthrough

We need q35 as the machine type in order to support IOMMU and
tweaking pcie controllers to place our virtio device properly.

1. Modify node config to accommodate iommu related section
2. Modify libvirt modification for pci passthru
3. Add pci passthru related node template
4. Expose option want_pci_passthrough

[kwu83tw]

This is the card for such task. https://jira.suse.com/browse/SOC-9304

[JanZerebecki]

Please fix the failing checks.

[jgrassler]

Looks like there's trouble with the CI jobs (unretrievable zypper repositories - looks unrelated to this pull request). I restarted suse/mkcloud/stable normally and suse/mkcloud with a bit of a twist: I put an export want_pci_passthrough=1 in the custom_settings field for the suse/mkcloud job. This should show us (a) whether the code causes no regressions when want_pci_passthrough is not being passed and (b) will actually exercise this code in the CI job.

[aspiers]

Do I correctly understand that this corresponds to this part of the Jira acceptance criteria?

A sensible, Crowbar based test automation is available to minimally test the functionality in a virtualized environment. A virtio based device we pass through to a nested VM is enough: we just need to make sure PCI pass-through in general works - no need to pass a GPU through.

IOW, your compute node is a VM set up via mkcloud, and this PR tweaks mkcloud to do PCI pass-through from the physical host to that VM - right?

Based on a quick scan through, it all looks fine to me, although I'm not familiar enough with the exact details of what libvirt config changes are required for PCI pass-through in order to be able to review this thoroughly. (Although on a related note, I am getting gradually more familiar with that area of the nova code.)

But I have one remaining question: if this only passes through from the physical host to the compute node, don't you also need to test pass-through from the compute node to the (nested) guest? IOW, double pass-through by modifying the OpenStack functional tests to boot a nested VM with pass-through enabled?

[kwu83tw]

IOW, your compute node is a VM set up via mkcloud, and this PR tweaks mkcloud to do PCI pass-through from the physical host to that VM - right?

This PR allows user to do prior preparation work of configuring node VM with proper machine type and additional virtio disk. The virtio disk on compute node (L0) will be passing through later on into its nested VM (L1). We're not going to pass anything from host to VM.

As for the testing part (to test pci-passthru from L0 VM to L1 VM), it will be another PR. 3543