podman machine rm not locking correctly

[Luap99]

podman machine uses a lockfile to synchronize access for the VM.

However if we look at the podman machine rm logic it does lock in this function but only to get the current file paths. Then it returns the paths in a function that is returned. The cli remove code then waits for the user to confirm the removal before deleting the files.
This means the actually removal of the files is not locked and it could lead to another machine command run in parallel and then having files removed while being used.

This problems seem to effect all providers.

podman/pkg/machine/qemu/machine.go

Lines 1011 to 1051 in 0a9914f

	func (v *MachineVM) Remove(_ string, opts machine.RemoveOptions) (string, func() error, error) {
	var (
	files []string
	)
	v.lock.Lock()
	defer v.lock.Unlock()
	// cannot remove a running vm unless --force is used
	state, err := v.State(false)
	if err != nil {
	return "", nil, err
	}
	if state == machine.Running {
	if !opts.Force {
	return "", nil, &machine.ErrVMRunningCannotDestroyed{Name: v.Name}
	}
	err := v.stopLocked()
	if err != nil {
	return "", nil, err
	}
	}
	files, err = v.collectFilesToDestroy(opts)
	if err != nil {
	return "", nil, err
	}
	confirmationMessage := "\nThe following files will be deleted:\n\n"
	for _, msg := range files {
	confirmationMessage += msg + "\n"
	}
	v.removeQMPMonitorSocketAndVMPidFile()
	confirmationMessage += "\n"
	return confirmationMessage, func() error {
	machine.RemoveFilesAndConnections(files, v.Name, v.Name+"-root")
	return nil
	}, nil
	}

[github-actions]

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

[Luap99]

was fixed in 35133c8