if you have a device that exposes /sys/class/net/eno2/device/sriov_numvfs
without the "virt_functions" capablity or a device wehre that change based the number
of VFs allcoated you should file a driver bug with your nic vendor as both behaviors would be incorrect.
this predates queens but i dont think we will backport it further then queens so i have added the different releases between master and queens.
no we do not take account of /sys/class/ net/eno2/ device/ sriov_numvfs when determining the type
its based solely on the pci config space capablities
chaning /sys/class/ net/eno2/ device/ sriov_numvfs form 0 to another value should not affect the pcie capabilities.
that is a common misconception however. /github. com/openstack/ nova/blob/ master/ nova/virt/ libvirt/ driver. py#L7125- L7163
this is the logic that determins the device type
https:/
if you have a device that exposes /sys/class/ net/eno2/ device/ sriov_numvfs
without the "virt_functions" capablity or a device wehre that change based the number
of VFs allcoated you should file a driver bug with your nic vendor as both behaviors would be incorrect.
this predates queens but i dont think we will backport it further then queens so i have added the different releases between master and queens.