Re: [Qemu-devel] [Xen-devel] [PATCH] Citrix PV Bus device

[Ian Campbell]

On Tue, 2013-07-02 at 13:51 +0100, Paul Durrant wrote:
> Yes, that makes sense *but* I would still like to avoid carrying a
> private patch to QEMU (and potentially have to keep rebasing it),

It's small and pretty self contained I think.

Someone (Anthony?) recommended making it subclass the existing one,
which ought to reduce the size of the patch to be maintained even
further I think.

>  hence my posting the patch the qemu-devel. Having the code in QEMU
> does no harm, clearly reserves the device id,

FWIW I don't think QEMU should be the registry of these device IDs. Can
you create a file somewhere in the Xen source base to serve as the
registry, probably somewhere under xen/include/public.

We should probably have some sort of scheme. How about we declare the
topmost available bit of the device id to be the "vendor specific" bit?
We could split the dev id into N-bits of "vendor" and M-bits of device,
or add a "locally administered" bit, but that might be overkill?

>  and any VM provider (with a suitable version of QEMU on their host)
> can then enable the device should they wish.

I imagine they would be worried about you pushing new drivers which
depend on new XenServer features. So unless they also intend to
implement (and track) your version compatibility xenstore keys (or
whatever mechanism) I'm not sure why they would want to do such a thing.
Or are you proposing to maintain forward and backward compatibility?
i.e. based on feature flags and negotiation with backends rather than
versioned platform devices or other XenServer specific mechanisms?

In any case they could also apply the patch. If it turns out that lots
of people are doing so then maybe that is the time to consider it for
upstream.

But in the meantime this avoids anyone other than XenServer having to
think about policy and/or mechanism WRT multiple platform devices.

>  Am I missing something?
> 
>   Paul