Ubuntu 18.04.4 KVM PCI Passthrough Stopped Working

Above all: it had historically worked, hopefully someone out there knows which stones to squeeze for useful logs to find what broke.

The problem I now face is that if I use virt-manager to add a PCI device for passthrough the VM no longer starts, and the host has something hang which prevents the machine shutting down indefinitely. Each time I try something I now must "yank the cord" because it won't turn off which is infuriating. A dirty reboot then scans for errors during boot and before long each test takes 20+minutes. virsh and virt-manager become unresponsive after a failed VM start attempt.

System:
Intel e3-1245g
Ubuntu 18.04 originally installed
Uses ZFS for VM datastore

Using KVM, virt-manager and attempting to pass the network card to the guest (host has 4 identical Intel NICs)

Problems began approximately when Ubuntu 18.04.4 was released but I hadn't installed any updates yet, it had been working in early Feb 2020.

• started with 18.04.?3? and all was working
• guest performance dropped catastrophically for unknown reasons
• performed update/upgrade
• VM guests with NIC passthrough stopped working
• troubleshooted it for day 1, trying to find relevant logs
• gave up, formatted after the repeated cord yanks killed the original installation
• installed 18.04.4
• installed KVM with the 18.10 instructions thinking that would align with the newer 18.04.4 release: https://help.ubuntu.com/community/KVM/Installation
• still faced with the same problem, attempting to start the VM while with PCI passthrough of the desired NIC does not work, no apparent problems in libvirtd logs, and host machine won't turn off. If VM start is not attempted it will turn off.

I noticed that before this problem, the extra NIC would disappear gracefully from the host when starting the guest, I presume that what is hanging is that the guest can't get control of the NIC for passthrough like it used to which is why virsh then stops working and the system can't shut down.

I will try vfio to hijack the NIC from the host, but it did not need that extra effort before.

• journalctl has been used for libvirtd logs
• I looked at the ones mentioned here https://blogs.oracle.com/oda/kvm:-tr...with-log-files
• the client log doesn't exist, so its hanging before the client VM starts, the host doesn't crash completely, it remains responsive
• I was unable to manually unbind the host's driver from the desired NIC but I'm not familiar with that process, its using the igb driver, I wrote the address to its unbind file like I saw in other forums and no success

Where should I be looking for logs to determine what's hung when a guest VM attempts to start, doesn't and causes some unknown process to block indefinitely, and the NIC doesn't disappear on the host as expected?

Thank you for your assistance,