Merge "fix sr-iov support on Cavium ThunderX hosts."

nova/virt/libvirt/host.py

@@ -1269,11 +1269,24 @@ class Host(object):
                     }
                     parent_ifname = None
                     # NOTE(sean-k-mooney): if the VF is a parent of a netdev
-                    # the PF should also have a netdev.
+                    # the PF should also have a netdev, however on some exotic
+                    # hardware such as Cavium ThunderX this may not be the case
+                    # see bug #1915255 for details. As such we wrap this in a
+                    # try except block.
                     if device.name() in net_dev_parents:
-                        parent_ifname = pci_utils.get_ifname_by_pci_address(
-                            pci_address, pf_interface=True)
-                        result['parent_ifname'] = parent_ifname
+                        try:
+                            parent_ifname = (
+                                pci_utils.get_ifname_by_pci_address(
+                                    pci_address, pf_interface=True))
+                            result['parent_ifname'] = parent_ifname
+                        except exception.PciDeviceNotFoundById:
+                            # NOTE(sean-k-mooney): we ignore this error as it
+                            # is expected when the virtual function is not a
+                            # NIC or the VF does not have a parent PF with a
+                            # netdev. We do not log here as this is called
+                            # in a periodic task and that would be noisy at
+                            # debug level.
+                            pass
                     if device.name() in vdpa_parents:
                         result['dev_type'] = fields.PciDeviceType.VDPA
                     return result

releasenotes/notes/fix-pci-passthrough-for-cavium-thunderx-8fbd1c40718569e2.yaml

@@ -0,0 +1,12 @@
+---
+fixes:
+  - |
+    On some hardware platforms, an SR-IOV virtual function for a NIC port may
+    exist without being associated with a parent physical function that has
+    an assocatied netdev. In such a case the the PF interface name lookup
+    will fail. As the ``PciDeviceNotFoundById`` exception was not handled
+    this would prevent the nova compute agent from starting on affected
+    hardware. See: https://bugs.launchpad.net/nova/+bug/1915255 for more
+    details. This edgecase has now been addressed, however, features
+    that depend on the PF name such as minimum bandwidth based QoS cannot
+    be supported on these platforms.