I think we should definitely be able to do a 'permanently attached disk' version of this. I took a quick look into the possibility of hotplugging. openQA (os-autoinst, really) mostly uses qemu's QMP for the task of 'doing stuff to the VM while it's running'. The code is in backend/qemu.pm. From a quick look, though, I don't see anything about hotplugging storage devices. You can use the misleadingly-named eject_cd in os-autoinst to send a QMP eject command to any device, though I'm not sure what happens if you try to 'eject' a USB stick - whether it works, or whether eject is strictly for "drives with separate media", like optical drives or floppy drives.

It looks like the device_add and device_del QMP commands can be used for hotplugging devices (at least that's what AI said, and a quick look at the doc seems to confirm that), but os-autoinst doesn't currently provide a wrapper for those AFAICT.

Possibly test code can get to the backend's generic handle_qmp_command function, I'm not sure. If so we could do something with that. If not we'd have to write a wrapper and submit it upstream.

For the permanently attached disk scenario, I think it would be good to
1. write the image
2. close FMW
3. verify the image
4. start FMW
5. restore the drive

The closing&reopening FMW would make it closer to a real-world scenario, when the user wants to restore their flash drive a few hours/days later after using it. It means FMW needs to be able to recognize the drive again after a fresh start, and offer the restore functionality. It's not as good as hotplugging the flash drive after FMW is started (and verifying that the hardware change is detected and option actually pops up dynamically), but it's close enough, and hotplugging can be tested by a human.

It would be 80% work done for 20% of time (compared to implementing missing features in handle_qmp_command), which sounds good to me.

I will try to make this happen.

A note on this: I only implemented a disconnect function upstream using QMP device_del so far, but QMP does also have a device_add which I guess could potentially be used to connect the USB device after system boot, simulating hotplug and making this test more accurate to likely real-world use. I'll try and find a bit of time to poke into whether I can figure out how to use that. (It would also let us automate https://fedoraproject.org/wiki/QA:Testcase_desktop_automount from the desktop matrix, which would be nice).

I am finishing this without the hot-plugging stuff. Incorporating this into the PR would another amount of time and added complexity to the ticket which I have not expected.
I suggest we close the original version first and have another ticket to add hot-plugging to this later, when we know how we should do it exactly. In between, we could have tests running with the original requirements.

https://pagure.io/fedora-qa/os-autoinst-distri-fedora/pull-request/462

This issue has been migrated to Fedora Forge:
https://forge.fedoraproject.org/quality/os-autoinst-distri-fedora/issues/450

Please continue any further discussion there.