Comment 16 for bug 1770082

This is not a full solution. In Xenial this renaming works even if initramfs has *not* been updated and there is no 70-persistent-net.rules file in the initial ramdisk. I'm still figuring out why this is, but it means that even if my patch were applied, there would be a regression in bionic vs xenial.

Here's a snippet from dmesg showing the same device being renamed twice, this is a xenial guest. It shows the device going from the kernel name (eth0) to the udev slot based name (ens16), to the name specified in the udev rules file that is *not* present in initrd.

ubuntu@xt2:~$ dmesg|grep renamed
[ 2.428015] virtio_net virtio3 ens16: renamed from eth0
[ 5.317990] virtio_net virtio3 ens3: renamed from ens16