RE: [RFC PATCH 3/6] fpga: dfl: add an API to get the base device for dfl device

[Wu, Hao]

> Subject: [RFC PATCH 3/6] fpga: dfl: add an API to get the base device for dfl
> device
> 
> This patch adds an API for dfl devices to find which physical device
> owns the DFL.
> 
> This patch makes preparation for supporting DFL Ether Group private
> feature driver. It uses this information to determine which retimer
> device physically connects to which ether group.
> 
> Signed-off-by: Xu Yilun <yilun.xu@xxxxxxxxx>
> ---
>  drivers/fpga/dfl.c  | 9 +++++++++
>  include/linux/dfl.h | 1 +
>  2 files changed, 10 insertions(+)
> 
> diff --git a/drivers/fpga/dfl.c b/drivers/fpga/dfl.c
> index ca3c678..52d18e6 100644
> --- a/drivers/fpga/dfl.c
> +++ b/drivers/fpga/dfl.c
> @@ -558,6 +558,15 @@ int dfl_dev_get_vendor_net_cfg(struct dfl_device
> *dfl_dev)
>  }
>  EXPORT_SYMBOL_GPL(dfl_dev_get_vendor_net_cfg);
> 
> +struct device *dfl_dev_get_base_dev(struct dfl_device *dfl_dev)
> +{
> +	if (!dfl_dev || !dfl_dev->cdev)
> +		return NULL;
> +
> +	return dfl_dev->cdev->parent;
> +}
> +EXPORT_SYMBOL_GPL(dfl_dev_get_base_dev);

It may confuse people that this get doesn't require a put operation on
the base device. could we avoid this by using a different name?

And why it needs to know the physical device here? DFL hides the
physical device for upper layer drivers, so this is not the same case?
or cdev can be used instead here?

Thanks
Hao

> +
>  #define is_header_feature(feature) ((feature)->id ==
> FEATURE_ID_FIU_HEADER)
> 
>  /**
> diff --git a/include/linux/dfl.h b/include/linux/dfl.h
> index 5ee2b1e..dd313f2 100644
> --- a/include/linux/dfl.h
> +++ b/include/linux/dfl.h
> @@ -68,6 +68,7 @@ struct dfl_driver {
>  #define to_dfl_drv(d) container_of(d, struct dfl_driver, drv)
> 
>  int dfl_dev_get_vendor_net_cfg(struct dfl_device *dfl_dev);
> +struct device *dfl_dev_get_base_dev(struct dfl_device *dfl_dev);
> 
>  /*
>   * use a macro to avoid include chaining to get THIS_MODULE.
> --
> 2.7.4