Re: [Qemu-devel] [PATCH 3/3] virtio-pci: don't poll masked vectors

[Stefan Hajnoczi]

On Tue, Dec 18, 2012 at 02:39:19PM +0200, Michael S. Tsirkin wrote:
> -static int virtio_pci_set_guest_notifier(DeviceState *d, int n, bool assign)
> +static void kvm_virtio_pci_vector_poll(PCIDevice *dev,
> +                                       unsigned int vector_start,
> +                                       unsigned int vector_end)
> +{
> +    VirtIOPCIProxy *proxy = container_of(dev, VirtIOPCIProxy, pci_dev);
> +    VirtIODevice *vdev = proxy->vdev;
> +    int queue_no;
> +    unsigned int vector;
> +    EventNotifier *notifier;
> +    VirtQueue *vq;
> +
> +    for (queue_no = 0; queue_no < VIRTIO_PCI_QUEUE_MAX; queue_no++) {
> +        if (!virtio_queue_get_num(vdev, queue_no)) {
> +            break;
> +        }
> +        vector = virtio_queue_vector(vdev, queue_no);
> +        if (vector < vector_start || vector >= vector_end ||
> +            !msix_is_masked(dev, vector)) {
> +            continue;
> +        }
> +        vq = virtio_get_queue(vdev, queue_no);
> +        notifier = virtio_queue_get_guest_notifier(vq);
> +        if (event_notifier_test_and_clear(notifier)) {
> +            msix_set_pending(dev, vector);
> +        }

Small difference to virtio_queue_guest_notifier_read() here: we do not
set vq->vdev->isr |= 0x01.  I guess no guest drivers use isr with MSIX
but should we still set it or even just call
virtio_queue_guest_notifier_read() instead of duplicating that event
notifier read here?

Stefan