Re: [Qemu-devel] [PATCH 28/35] kvm: x86: Introduce kvmclock device to save/restore its state

[Anthony Liguori]

On 01/25/2011 05:06 AM, Avi Kivity wrote:
> On 01/19/2011 06:57 PM, Anthony Liguori wrote:
> > On 01/19/2011 07:15 AM, Markus Armbruster wrote:
> > > So they interact with KVM (need kvm_state), and they interact with the
> > > emulated PCI bus.  Could you elaborate on the fundamental difference
> > > between the two interactions that makes you choose the (hypothetical)
> > > KVM bus over the PCI bus as device parent?
> >
> > It's almost arbitrary, but I would say it's the direction that I/Os 
> > flow.
>
> In the case of kvm, things are somewhat misleading.  I/O still flows 
> through the (virtual) PCI bus, it's just short-circuited to a real device.

It doesn't.  If we have a PCI bus that transforms I/O or remaps I/O via 
an IOMMU, that device doesn't participate in it.

But this whole discussion is way off track.

We don't have to solve any of these problems today.  Just don't remove 
kvm_state and grab a global reference to it when we need to (which is 
*at best* one place in the code today) and let's move on with our lives.

Regards,

Anthony Liguori

>   Similarly when attaching an ioeventfd to a virtio kick register, 
> things still logically from the same way as without ioeventfd; we 
> simply add a fast path for the operation.  But it doesn't change the 
> logical view of things.