Re: [libvirt] [PATCH 2/5] conf: Add pSeries features
On Thu, Jan 25, 2018 at 05:12:49PM +0100, Andrea Bolognani wrote:
On Thu, 2018-01-25 at 15:03 +0000, Daniel P. Berrangé wrote:
> On Tue, Jan 23, 2018 at 04:45:01PM +0100, Andrea Bolognani wrote:
> > We're going to introduce a number of optional, pSeries-specific
> > features, so we want to have a dedicated sub-element to avoid
> > cluttering the <domain><features> element.
>
> I don't think we want todo this as all. Pretty much *every* single
> existing <feature> element is architecture specific. Adding a nested
> level just for pseries is making it different from existing practice
> and I don't see any benefit to that.
Well, there's going to be a bunch of them added pretty soon (six
or so) plus probably more down the line, so that's already quite
a bit of clutter. My reasoning was to try and organize them the
way we already do with <kvm> and <hyperv> capabilities - even
though pSeries is not a hypervisor per se, you can kinda see
sPAPR as a specification implemented by various hypervisors, like
PowerVM and in our case QEMU/KVM. But if you think this effort is
misguided and they belong to the top-level <features> element,
then I'm okay with that too.
I guess where the nesting makes sense is if there's a chance of having
namespace collision between features. eg if both kvm and hyperv
had a feature called "pvspinlocks", you might want to enable them
separately, so the nesting is important there.
If you do think the pSeries nesting is important in this respect, can
you still leave the existing "hpt" feature un-nested for sake of
full back-compat - just nest new features.
Regards,
Daniel
--
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org -o- https://fstop138.berrange.com :|
|: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|