Enable QEMU Podman machine on Windows

[arixmkii]

Fixes #13006

This simply enables new machine provider as all the required changes were implemented in separate PRs.

Technically this could be merged, but this will not be really functional before #17473 is completed and merged.

No new tests, because this doesn't introduce new unit testable functionality, but for e2e tests this would require Windows VM with nested virtualization or Windows bare metal machine, where QEMU will run.

Does this PR introduce a user-facing change?

Enable QEMU Podman machine on Windows amd64 hosts

[NO NEW TESTS NEEDED]

[TomSweeneyRedHat]

@arixmkii is this ready to have the WIP label removed?

[cevich]

Windows bare metal machine, where QEMU will run.

We use bare-metal Windows instances for installer and WSL2 testing already. Possibly the only major part missing to enable addition of QEMU tests, is inclusion of it and any dependencies inside the CI VM images. I'm not a Windows expert, but I think the place to do that is https://github.com/containers/automation_images/blob/main/win_images/win_packaging.ps1

With updated images in hand, I can help with updating .cirrus.yml. But I'm afraid I wouldn't be useful in adding/updating the powershell scripts (under contrib/cirrus/). Any assistance there would be greatly appreciated.

[arixmkii]

With updated images in hand, I can help with updating .cirrus.yml. But I'm afraid I wouldn't be useful in adding/updating the powershell scripts (under contrib/cirrus/). Any assistance there would be greatly appreciated.

Sounds good. I will try to move forward with #17473 and then come back here to do my best to help with this.

[github-actions]

A friendly reminder that this PR had no activity for 30 days.

[rhatdan]

We now support hyperV, I don't think We should be supporting a third way.

[arixmkii]

Last time I checked Hyper-V required super user for starting up the machine. This one didn't. But I have to check the current state for Hyper-V. So, there is some reasoning for this to exist still, but I can understand this being dropped - this is fine if this is easy to add and compile from sources (if #17473 is kept and finished).

[github-actions]

A friendly reminder that this PR had no activity for 30 days.

[cevich]

A friendly reminder that this PR had no activity for 30 days.

Still unlikely a problem. There's lots of thought/planning/work going on with this topic. It's part of a bigger multi-os podman-machine testing initiative and design-work. Not sure we're far enough along to decide what to do with this PR yet.

[arixmkii]

I have rebased version of this one, which I use to test my ongoing work for #17473 I just delay the re-submit until the other part is finalized.

[openshift-ci]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: arixmkii
Once this PR has been reviewed and has the lgtm label, please assign umohnani8 for approval. For more information see the Kubernetes Code Review Process.

The full list of commands accepted by this bot can be found here.

Needs approval from an approver in each of these files:

• OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[arixmkii]

Build issues will be addressed in #17473 (as this one has no point going forward before the latter). The build issues are caused by recently added functionality, which had no Windows compatible alternatives and were added straight to machine.go instead of machine_windows.go.

[arixmkii]

Rebased and updated for latest codebase changes.
Still dependent on #17473

[baude]

hello @arixmkii We've been watching this PR with interest and frankly are struggling with it a bit. We already have two supported platforms on windows in WSL and soon to be HyperV. Could we have a broader conversation around why QEMU is more attractive that those two options?

One concern we have is that this broadens and deepens our provider support to three different providers on Windows and we are struggling to keep as it is.

[arixmkii]

Ran e2e tests with this provider. Only 3 failed tests, which are most probably some OS specifics in them, which has to be checked in details.

  [FAIL] podman machine rm [It] Remove running machine
  C:/msys64/home/username/gitrepos/podman/pkg/machine/e2e/rm_test.go:67
  [FAIL] podman machine start [It] start simple machine
  C:/msys64/home/username/gitrepos/podman/pkg/machine/e2e/start_test.go:40 
  [FAIL] podman machine stop [It] Stop running machine
  C:/msys64/home/username/gitrepos/podman/pkg/machine/e2e/stop_test.go:39 

Brief check is that in all cases successful (expected result) machine stop commands return non zero exit code (not able to reproduce manually outside of test)

[arixmkii]

@baude sure! First of all this was a thing way before Hyper-V full virt was even publicly debated. I started it in January 2022 #13006

What was beneficial for me:

• it worked exactly like the one I'm using on macOS (it will soon become irrelevant, when applevf is finalized)
• full virt - no shared resources with other WSL infra
• networking - no issues publishing ports for entire network, not only local host
• comparing to Hyper-V - no need to use elevated shell to start/stop machine
• verified to work on bare metal and with nested virtualization (with access to published ports from host machine as well)
• it exposes API via UNIX domain socket in addition to named pipe, so, it is possible to use same automation script written in bash/curl (running it with msys2 on Windows) written on macOS (very niche usecase, but still valid)

I can understand if this one never lands in main branch :) It was pretty interesting learning experience anyway.

[arixmkii]

Found a cross process file access/race condition like combination, which doesn't work pretty on Windows. Now it passes all e2e tests. Will update PRs after full machine e2e run passes (fixes were tested running tests isolated).

Test run results (time is irrelevant, it ran inside Hyper-V machine on power gapped laptop).

Ran 34 of 34 Specs in 1599.477 seconds
SUCCESS! -- 34 Passed | 0 Failed | 0 Pending | 0 Skipped
PASS

[arixmkii]

Will be continued in #21594