Re: [PATCH v6 07/17] add minimal virtio support for devtree virtio-mmio

[Andrew Jones <drjones@redhat.com>]

On Fri, Jul 11, 2014 at 11:27:58AM +0200, Paolo Bonzini wrote:
> Il 11/07/2014 11:08, Andrew Jones ha scritto:
> >>> lib/arm/virtio.c
> >>>
> >>> where virtio_bind is in lib/arm/virtio.c.
> >>>
> >Well, virtio_bind will still need to be in lib/virtio.c, but just as
> >a wrapper to arch_virtio_bind.
> 
> Ok, that's just a naming thing.
> 
> >And, I'm inclined to keep virtio_bind_busses
> >in arm's arch_virtio_bind.
> 
> Why?  To support virtio-pci in the future?  It seems like a good thing to
> have (future support for virtio-pci) but even then you'd have only two tests
> and that's already the exception.  The common case would be just one.  You
> could write that as
> 
>     struct virtio_device *arch_virtio_bind(u32 devid)
>     {
>         struct virtio_device *vdev;
> 
> 	vdev = arch_virtio_mmio_bind(devid);
> 	if (!vdev)
> 	    vdev = arch_virtio_pci_bind(devid);
> 	return vdev;
>     }
> 
> (I don't see kvm-unit-tests using ACPI in the future.  Having DT+ACPI x
> mmio+pci would be a good reason to have the array, but even then it's

Yes, that was the reason.

> premature and these are unit tests not an OS...).

But, true and true. I guess I'll drop the table for now.

drew