Be careful when adding or removing PCIe hardware, the device addresses may change

General Discussion
1

So I removed my PCIe m.2 wifi card then went to sleep, when I woke up and tried booting into Qubes, it allowed me to enter disk pw but then crashed before reaching login. I forgot that I recently made an untested hardware change.

I retried a few times eventually recording the screen’s boot messages in slo-mo. I noticed the failure was shortly after auto-starting qubes. When I removed my wifi card it changed the PCIe order and was trying to autostart my net VM with some other piece of critical hardware.

I reinserted my wifi card and it now boots. I think disabling autostart in grub would also prevent the crash, but I did not test. I’m not exactly sure how PCIe bus ordering works, but it appears that if I assume the next device address was assigned it means I was attempting to passthrough the whole chipset!

2

Thanks for the warning.
This is a known issue, and as yet we have no solution.
(I think this is not a Qubes issue, but heated opinions differ.)

I never presume to speak for the Qubes team. When I comment in the Forum or in the mailing lists I speak for myself.
3

Perhaps some kind of early VM startup crash detection. If the machine makes it to VM start up twice but never to completion automatically disable autostart next boot and give user a suggestion.

Itd be nice is linux had some kindof symbolic filesystem like disks do that dont rely on POST (disk-by…xxx). I’m pretty sure that came to exist for the same reason PCIe is giving us problems today.