Il 17/12/2012 16:24, Michael S. Tsirkin ha scritto: > On Mon, Dec 17, 2012 at 04:14:00PM +0100, Paolo Bonzini wrote: >> Il 17/12/2012 11:40, Michael S. Tsirkin ha scritto: >>> How about the following? Then we can put reset >>> in generic code where it belongs. >>> It's untested - really kind of pseudo code - and >>> s390 is still to be updated. >>> >>> Posting to see what does everyone thinks. >> >> I'm not (yet) sure how that helps my problem, > > It makes it possible for virtio.c to get at the > device state through the binding pointer. > So you will be able to qdev_reset_all from virtio.c > where it belongs, instead of duplicating code > in all bindings.
Yes, but where does it belong? Do you want to move handling of the status register (and others) to hw/virtio.c? Also, you're proposing that I do qdev_reset_all(vdev->binding_opaque) but that would be a layering violation. Generic virtio code should not be able to reset the transport-specific setup (e.g. MSIs). Paolo >> but it is definitely a >> step in the right direction! >> >> Paolo
