[pve-devel] [PATCH qemu-server/manger/docs v6] cluster mapping

Markus Frank m.frank at proxmox.com
Wed Jun 14 14:01:31 CEST 2023

Tested with my viommu patches installed on my host so that I can use iommu inside
a VM and pass though virtual devices to Level-2 VMs.

I created a pve8 vm and a pve8 2-node vm cluster.

I used the Web UI for this testing but made a quick test with pvesh on the
API (/cluster/mapping/pci).

* Passed through single PCI devices and groups (Pass through all functions as
one device). works fine

* made a mapping for 2 nodes and started a vm using that mapping on each node.
works fine

provoke errors:
* "PCI device map for 'mapping1' not found" is displayed when starting a VM
if a map was added that has no configuration for this node.
Also displays a warning when adding a map that has no configuration.
* could only migrate if PCI/USB Mapping is preset on the target node
* logically could not live migrate with PCI/USB Devices assigned.

Error handling also works fine

There is only one thing I noticed:
* When assigning multiple devices to a node, not all of them are passed through,
just one. There is a hint field (multiple_warning) which is not visible
right now.

Tested-by: Markus Frank <m.frank at proxmox.com>

On 6/14/23 10:46, Dominik Csapak wrote:
> this series is the remaining part to add a cluster-wide device mapping for pci and usb
> devices. so that an admin can configure a device to be availble for
> migration and configuring for uses that are non-root
> (the existing pattern can be copied easily for other types, e.g.
> markus upcoming folder sharing)

More information about the pve-devel mailing list