New vmctl module to interface with the OpenBSD VMM hypervisor

[jasperla]

What does this PR do?

Provide a new module that allows Salt to interact with the OpenBSD VMM hypervisor.

Previous Behavior

NA

New Behavior

This new module implements the following functionality:

• create a new disk image
• load a new configuration file
• reload the default configuration
• stop VMs
• reset VMM or one of the subsystems
• define new VMs and start existing VMs
• request status from all or one VM(s)

Tests written?

No

Commits signed with GPG?

Yes

[rallytime]

@jasperla Thanks again for these additions! Keep 'em coming!

I have a couple of comments here. This will also need some unit tests added as well before we can get it in.

[rallytime]

A versionadded for Fluorine will be needed for this new file.

[jasperla]

Check, I'll it.

[rallytime]

We're moving away from bare salt.utils imports. All of the functions there have been moved over into their own files. Can you update this import to import the individual files needed import salt.utils.x? (And of course update the references in the function calls below.)

The salt.utils.__init__.py file should have the up-to-date paths in the deprecation warnings for each function for reference.

[jasperla]

There was actually only one (salt.utils.path). I'll adjust it.

[jasperla]

I've now added tests too.

[eradman]

If no argument is passed to vmctl.status then this function mistakenly filters VMs by the name None

[INFO    ] Executing command ['vmctl', 'status', None] in directory '/home/eradman'
[TRACE   ] stdout:    ID   PID VCPUS  MAXMEM  CURMEM     TTY        OWNER NAME
[DEBUG   ] LazyLoaded nested.output
[TRACE   ] data = {'local': {}}

[eradman]

Should be if id:

[jasperla]

Good point, I had accidentally changed that. It's fixed now.

[eradman]

There is an edge case not handled here when vmctl reports that the VM is in the state "stopping"

[jasperla]

@eradman I've fixed the edge case too now. Turns out it's a bug in vmctl where it doesn't output the status line for a VM that is stopping when the status is requested for the VM by name (works by id). i.e. vmctl stop foo && vmctl status foo emits the header but nothing else.

[eradman]

@jasperla: previous fixes work. There is another flaw in the way vmctl reports status that causes some confusion. If I start a VM with a name that is already used the output is unclear:

$ doas salt-call --local vmctl.start name=web1 bootpath='/bsd.rd' nics=2 memory=512M disk='/tmp/disk.img'
local:
    ----------
    changes:
        True
    console:
        None

The output from vmctl is not much better

$ doas vmctl start web1 -i 2 -b /bsd.rd -m 512M -d /tmp/disk.img
vmctl: start vm command failed: Operation not permitted
$ echo $?
0

Not sure what this Salt module should do--perhaps we need to parse the status line for this VM?

(saltenv) $ doas vmctl status web1
   ID   PID VCPUS  MAXMEM  CURMEM     TTY        OWNER NAME
   12 54520     1    512M   93.3M   ttypj         root web1
 VCPU:  0 STATE: STOPPED

[eradman]

s/boothpath/bootpath/

[jasperla]

@eradman thanks for pointing out that bug in vmctl, I'll look into that.

I've now changed start() to return:

local:
    ----------
    changes:
        False
    comment:
        VM already exists and cannot be redefined
    console:
        None

So changes is now False and a comment is populated in case the VM that one tries to define already exists.

EDIT: tests now pass again too.

[eradman]

Functionally this change seems to be good!

[rallytime]

re-run py

[rallytime]

re-run py

[jasperla]

Looking better now :)