Re: [PATCH V6 8/8] virtio: Intel IFC VF driver for VDPA

[Jason Gunthorpe]

On Wed, Mar 18, 2020 at 04:03:27PM +0800, Jason Wang wrote:
> From: Zhu Lingshan <lingshan.zhu@xxxxxxxxx>
> +
> +static int ifcvf_vdpa_attach(struct ifcvf_adapter *adapter)
> +{
> +	int ret;
> +
> +	adapter->vdpa_dev  = vdpa_alloc_device(adapter->dev, adapter->dev,
> +					       &ifc_vdpa_ops);
> +	if (IS_ERR(adapter->vdpa_dev)) {
> +		IFCVF_ERR(adapter->dev, "Failed to init ifcvf on vdpa bus");
> +		put_device(&adapter->vdpa_dev->dev);
> +		return -ENODEV;
> +	}

The point of having an alloc call is so that the drivers
ifcvf_adaptor memory could be placed in the same struct - eg use
container_of to flip between them, and have a kref for both memories.

It seem really weird to have an alloc followed immediately by
register.

> diff --git a/drivers/virtio/virtio_vdpa.c b/drivers/virtio/virtio_vdpa.c
> index c30eb55030be..de64b88ee7e4 100644
> +++ b/drivers/virtio/virtio_vdpa.c
> @@ -362,6 +362,7 @@ static int virtio_vdpa_probe(struct vdpa_device *vdpa)
>  		goto err;
>  
>  	vdpa_set_drvdata(vdpa, vd_dev);
> +	dev_info(vd_dev->vdev.dev.parent, "device attached to VDPA bus\n");
>  
>  	return 0;

This hunk seems out of place

Jason