Re: [PATCH 04/22] ice: Register auxiliary device to provide RDMA

[Jason Gunthorpe <jgg@nvidia.com>]

On Fri, Jan 22, 2021 at 05:48:09PM -0600, Shiraz Saleem wrote:
> +static void ice_peer_adev_release(struct device *dev)
> +{
> +	struct iidc_auxiliary_object *abo;
> +	struct auxiliary_device *adev;
> +
> +	adev = container_of(dev, struct auxiliary_device, dev);
> +	abo = container_of(adev, struct iidc_auxiliary_object, adev);

This is just

 container_of(dev, struct iidc_auxiliary_object, adev.dev);

> @@ -1254,20 +1282,37 @@ int ice_init_peer_devices(struct ice_pf *pf)
>  		 * |--> iidc_peer_obj
>  		 * |--> *ice_peer_drv_int
>  		 *
> +		 * iidc_auxiliary_object (container_of parent for adev)
> +		 * |--> auxiliary_device
> +		 * |--> *iidc_peer_obj (pointer from internal struct)
> +		 *
>  		 * ice_peer_drv_int (internal only peer_drv struct)
>  		 */
>  		peer_obj_int = kzalloc(sizeof(*peer_obj_int), GFP_KERNEL);
> -		if (!peer_obj_int)
> +		if (!peer_obj_int) {
> +			ida_simple_remove(&ice_peer_ida, id);
>  			return -ENOMEM;
> +		}

Why is this allocated memory with a lifetime different from the aux
device?

This whole peer_dev/aux_dev split needs to go, why on earth does
peer_obj need an entire state machine for driver binding? This is what
the aux device and driver core or supposed to provide.

> +		abo = kzalloc(sizeof(*abo), GFP_KERNEL);
> +		if (!abo) {
> +			ida_simple_remove(&ice_peer_ida, id);
> +			kfree(peer_obj_int);
> +			return -ENOMEM;
> +		}

Put the auxiliary_device_init() directly after kzalloc.

Even better is to put everything up to the
kzalloc/auxiliary_device_init() into a function called
'alloc_aux_device'

Then all the error unwind here doesn't look so bad

Jason