Re: [PATCH 2/5] clk: qcom: Allow custom reset ops

[Dmitry Baryshkov <dmitry.baryshkov@xxxxxxxxxx>]

On 30/07/2022 12:17, Akhil P Oommen wrote:
> Add support to allow soc specific clk drivers to specify a custom reset
> operation. A consumer-driver of the reset framework can call
> "reset_control_reset()" api to trigger this.
>
> Signed-off-by: Akhil P Oommen <quic_akhilpo@xxxxxxxxxxx>
> ---
>
>   drivers/clk/qcom/reset.c | 6 ++++++
>   drivers/clk/qcom/reset.h | 2 ++
>   2 files changed, 8 insertions(+)
>
> diff --git a/drivers/clk/qcom/reset.c b/drivers/clk/qcom/reset.c
> index 819d194..4782bf1 100644
> --- a/drivers/clk/qcom/reset.c
> +++ b/drivers/clk/qcom/reset.c
> @@ -13,6 +13,12 @@
>   
>   static int qcom_reset(struct reset_controller_dev *rcdev, unsigned long id)
>   {
> +	struct qcom_reset_controller *rst = to_qcom_reset_controller(rcdev);
> +	const struct qcom_reset_map *map = &rst->reset_map[id];
> +
> +	if (map->op)
> +		return map->op(map);

This looks like a hack. For example, assert() and deassert() would still 
follow the usual pattern of updating the bits. Please at least make them 
return -EOPNOTSUP if map->op is defined.

A slightly better solution would be to make qcom_reset implementation 
optional (and depending on desc->num_resets being greater than 0). Then 
you can register your own reset controller implementation from the gpucc 
driver.


> +
>   	rcdev->ops->assert(rcdev, id);
>   	udelay(1);
>   	rcdev->ops->deassert(rcdev, id);
> diff --git a/drivers/clk/qcom/reset.h b/drivers/clk/qcom/reset.h
> index 2a08b5e..295deeb 100644
> --- a/drivers/clk/qcom/reset.h
> +++ b/drivers/clk/qcom/reset.h
> @@ -11,6 +11,8 @@
>   struct qcom_reset_map {
>   	unsigned int reg;
>   	u8 bit;
> +	int (*op)(const struct qcom_reset_map *map);
> +	void *priv;
>   };
>   
>   struct regmap;


--
With best wishes
Dmitry