Re: [Qemu-devel] [RFC v2] libvirt vGPU QEMU integration

[Alex Williamson <alex.williamson@redhat.com>]

On Fri, 7 Oct 2016 10:46:22 +0530
Kirti Wankhede <kwankhede@nvidia.com> wrote:

> Ping..
> 
> Pulling the questions at the top.
> 
> >> Will libvirt report 'description' RO attribute, its output would be
> >> string, so that user could be able to see the configuration of that
> >> profile?
> >>  
> >
> > Daniel,
> > Waiting for your input on this.
> >  
> >> We can have 'class' as optional attribute. So Intel don't have to
> >> provide 'class' attribute and they don't have to specify mandatory
> >> attributes of that class. We would provide 'class' attribute and provide
> >> mandatory attributes.  

Hi Kirti,

I would proceed with your design and posting, this is not a
sufficiently significant issue to delay a new version.  The next
posting will likely not be the last, more feedback will come.

Personally I don't see an issue.  We should have a small number
of mandatory sysfs interfaces for mdev devices, essentially just
enough to manage the life cycle of a device, even if it requires
vendor specific knowledge to understand the exact composition of
that device.  Some sort of human readable description seems like
a reasonable optional feature for a userspace tool like libvirt
to take advantage of.  A class definition with an available set
of attributes per class would clearly enhance the ability of a
userspace tool to understand the device, but does not seem
strictly required for managing the device.  Thanks,

Alex