On 09/06/2018 09:04 AM, Martin Kletzander wrote:
> On Wed, Sep 05, 2018 at 03:37:22PM +0100, Daniel P. Berrangé wrote:
>> On Wed, Sep 05, 2018 at 02:01:42PM +0200, Martin Kletzander wrote:
>>> On Wed, Sep 05, 2018 at 09:28:52AM +0100, Daniel P. Berrangé wrote:
>>> > On Tue, Sep 04, 2018 at 03:44:12PM -0400, Cole Robinson wrote:
>>> > > Right now in virt-manager we only track a VM's OS name (win10,
>>> fedora28,
>>> > > etc.) during the VM install phase. This piece of data is important
>>> > > post-install though: if the user adds a new disk to the VM later,
>>> we want to
>>> > > be able to ask libosinfo about what devices the installed OS
>>> supports, so we
>>> > > can set optimal defaults, like enabling virtio.
>>> > >
>>> > > There isn't any standard libvirt XML field to track this kind
of
>>> info
>>> > > though, so apps have to invent their own schema. nova and rhev do
it
>>> > > indirectly AFAICT. gnome-boxes does it directly with XML like
this:
>>> > >
>>> > > <metadata>
>>> > > <boxes:gnome-boxes
>>>
xmlns:boxes="https://wiki.gnome.org/Apps/Boxes">
>>> > >
<
os-id>http://fedoraproject.org/fedora/28</os-id>
>>> > > ....
>>> > > </boxes:gnome-boxes>
>>> > > </metadata>
>>> > >
>>> > > I want to add something similar to virt-manager but it seems a
>>> shame to
>>> > > invent our own private schema for something that most non-trivial
>>> virt apps
>>> > > will want to know about. I was thinking a schema we could
>>> document with
>>> > > libosinfo, something like
>>> > >
>>> > > <metadata>
>>> > > <libosinfo
>>> > >
xmlns:libosinfo="http://libosinfo.org/xmlns/libvirt/domain/1.0"...
>>> > > <
os-id>http://fedoraproject.org/fedora/28</os-id>
>>> > > </libosinfo>
>>> > > </metadata>
>>> >
>>> > Yes, I would like to see this standardized under <matadata>.
>>> >
>>>
>>> Me too and what Cole suggested looks fine.
>>
>> It occurs to me that we actually need more than just the os-id value.
>>
>> When you query devices for a given OS, you'll often be told that multiple
>> devices are compatible, and the mgmt app can decide which of them to then
>> use.
>>
>> So if we want consistency when later hotplugging, we should make a record
>> of which devices we decided to use too, so if the mgmt app changes its
>> preference, we still know what we originally picked.
>>
>> eg to express that we use virtio-net and virtio-blk (even if virtio-scsi
>> was supported by the OS):
>>
>> <metadata>
>> <libosinfo
>>
xmlns:libosinfo="http://libosinfo.org/xmlns/libvirt/domain/1.0"...
>> <os
id="http://fedoraproject.org/fedora/28"/>
>> <device
id="http://pcisig.com/pci/1af4/1000"/>
>> <device
id="http://pcisig.com/pci/1af4/1001"/>
>> </libosinfo>
>> </metadata>
>>
>> Note, I'm suggesting using an 'id' attribute, rather than naming the
>> element 'os-id', to be more closely aligned with osinfo schema.
>>
>
> I'm not against that <device id =''/> but it is going to take some
> effort to
> properly specify what is really meant by that. The fact that some
> device model
> was chosen for a particular device does not necessarily mean that it is
> requested as the default. It only means what is actually encoded in the
> XML
> already, that is a particular model for a particular device.
>
Yeah I'm a bit confused by this as well, it's not exactly clear to me
how we would use or set XML like that for virt-manager, and how other
apps would be expected to consume it.
That's what I though of when trying to say we need to define the meaning of
that. What might be meaningful is if the user selects a particular *default*
model for new devices (e.g. disks should be IDE by default) then that option
could be honoured when adding a new device of that type (unless requested
otherwise). I'm not sure if that's what Daniel meant by that.