Return UnknownDevice for assignments where device is unknown

[marmarta]

references QubesOS/qubes-issues#10409

[codecov]

Codecov Report

❌ Patch coverage is 0% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 76.10%. Comparing base (a5ea121) to head (33cc033).

Files with missing lines	Patch %	Lines
qubesadmin/device_protocol.py	0.00%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #400   +/-   ##
=======================================
  Coverage   76.10%   76.10%           
=======================================
  Files          53       53           
  Lines        9287     9287           
=======================================
  Hits         7068     7068           
  Misses       2219     2219           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[marmarek]

First, check carefully places where this attribute is used (in global config, VM settings, widgets, ansible etc). The behavior change may be unexpected in some, especially since the old behavior was explicitly documented in the docstring.

Then, update docstring.

And finally, I don't think attachment attribute should be set. Its docstring says "VM to which device is attached (frontend domain)", which cannot be true for not present device.

[marmarta]

I think you're implying this is a bad idea? I think it makes sense not to raise a ProtocolError if we have an UnknownDevice for those purposes, but I can try to think about tacking this problem in another way.

(and about the attachment attribute, it is incorrect to set here self, yes, but this is a property of a DeviceAssignment that knows its frontend domain, shouldn't we set it?)

[marmarek]

I think the change makes sense in general. But since it's behavior change, it needs at least checking if other places don't rely on the old behavior. For example attach_device in qubesadmin/tools/qvm_device.py seems to rely on this exception. I'm sure there are other places too (maybe also in other repositories...)

(and about the attachment attribute, it is incorrect to set here self, yes, but this is a property of a DeviceAssignment that knows its frontend domain, shouldn't we set it?)

"attachment" is about where the device is attached (like, currently connected), not assigned (potentially attached, automatically or not, may be even multiple assignments for the same device)

[marmarek]

Oh, and note this file is also copied to core-admin, so check if that side doesn't rely on this exception too.