vDPA: dont change vq irq after DRIVER_OK

IRQ of a vq is not expected to be changed in a DRIVER_OK ~ !DRIVER_OK
period for irq offloading purposes. Place this comment at the side of
bus ops get_vq_irq than in set_status in vhost_vdpa.

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Link: https://lore.kernel.org/r/20200804102123.69978-1-lingshan.zhu@intel.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

drivers/vhost/vdpa.c

@@ -172,7 +172,6 @@ static long vhost_vdpa_set_status(struct vhost_vdpa *v, u8 __user *statusp)
 
 	ops->set_status(vdpa, status);
 
-	/* vq irq is not expected to be changed once DRIVER_OK is set */
 	if ((status & VIRTIO_CONFIG_S_DRIVER_OK) && !(status_old & VIRTIO_CONFIG_S_DRIVER_OK))
 		for (i = 0; i < nvqs; i++)
 			vhost_vdpa_setup_vq_irq(v, i);

include/linux/vdpa.h

@@ -186,6 +186,7 @@ struct vdpa_config_ops {
 	u64 (*get_vq_state)(struct vdpa_device *vdev, u16 idx);
 	struct vdpa_notification_area
 	(*get_vq_notification)(struct vdpa_device *vdev, u16 idx);
+	/* vq irq is not expected to be changed once DRIVER_OK is set */
 	int (*get_vq_irq)(struct vdpa_device *vdv, u16 idx);
 
 	/* Device ops */