[PATCH v4 06/19] soc: mediatek: Add runtime PM and top clocks and async controls for MMSYS

[CK Hu (胡俊光)]

On Thu, 2025-08-28 at 16:07 +0800, Paul Chen wrote:
> From: Nancy Lin <nancy.lin at mediatek.com>
> 
> - Add initialization of top clocks and async clocks for each MMSYS.
> - Add PM runtime control and new functions to manage these clocks.
> - Add functions to set these clocks according to the default
>   configuration of the corresponding MMSYS.

This patch include 3 things:
1. Add runtime PM
2. top clock control
3. async control

I would like you to separate these to different patch.

> 
> Signed-off-by: Nancy Lin <nancy.lin at mediatek.com>
> Signed-off-by: Paul-pl Chen <paul-pl.chen at mediatek.com>
> ---

[snip]

> +
> +int mtk_mmsys_ddp_clk_enable(struct device *dev, enum mtk_ddp_comp_id comp_id)

This is to control async clock, so change the function name to

mtk_mmsys_async_clk_enable()

> +{
> +	struct mtk_mmsys *mmsys = dev_get_drvdata(dev);
> +	const struct mtk_mmsys_async_info *async = mmsys->data->async_info;
> +
> +	int i;
> +
> +	if (!mmsys->data->num_async_info)

for-loop would check this, so drop this.

> +		return 0;
> +
> +	for (i = 0; i < mmsys->data->num_async_info; i++)
> +		if (comp_id == async[i].comp_id)
> +			return clk_prepare_enable(mmsys->async_clk[async[i].index]);
> +	return 0;
> +}
> +EXPORT_SYMBOL_GPL(mtk_mmsys_ddp_clk_enable);
> +
> +void mtk_mmsys_ddp_clk_disable(struct device *dev, enum mtk_ddp_comp_id comp_id)
> +{
> +	struct mtk_mmsys *mmsys = dev_get_drvdata(dev);
> +	const struct mtk_mmsys_async_info *async = mmsys->data->async_info;
> +	int i;
> +
> +	if (!mmsys->data->num_async_info)

for-loop would check this, so drop this.

> +		return;
> +
> +	for (i = 0; i < mmsys->data->num_async_info; i++)
> +		if (comp_id == async[i].comp_id)
> +			clk_disable_unprepare(mmsys->async_clk[async[i].index]);
> +}
> +EXPORT_SYMBOL_GPL(mtk_mmsys_ddp_clk_disable);
> +
> +void mtk_mmsys_ddp_config(struct device *dev, enum mtk_ddp_comp_id comp_id,
> +			  int width, int height, struct cmdq_pkt *cmdq_pkt)

This is to control async. So change the function name to

mtk_mmsys_async_config()

> +{
> +	struct mtk_mmsys *mmsys = dev_get_drvdata(dev);
> +	const struct mtk_mmsys_async_info *async = mmsys->data->async_info;
> +	int i;
> +	u32 val;
> +
> +	if (!mmsys->data->num_async_info)

If mmsys->data->num_async_info = 0, then i = 0, and it would return below.
So this checking is redundant.

> +		return;
> +
> +	for (i = 0; i < mmsys->data->num_async_info; i++)
> +		if (comp_id == async[i].comp_id)
> +			break;
> +
> +	if (i == mmsys->data->num_async_info)
> +		return;
> +
> +	val = FIELD_PREP(GENMASK(31, 16), height);
> +	val |= FIELD_PREP(GENMASK(15, 0), width);
> +	mtk_mmsys_update_bits(mmsys, async[i].offset, async[i].mask, val, cmdq_pkt);
> +}
> +EXPORT_SYMBOL_GPL(mtk_mmsys_ddp_config);
> +
> +void mtk_mmsys_default_config(struct device *dev)

Why not do this when mmsys probe?

Regards,
CK

> +{
> +	struct mtk_mmsys *mmsys = dev_get_drvdata(dev);
> +	const struct mtk_mmsys_default *def_config = mmsys->data->def_config;
> +	int i;
> +
> +	if (!mmsys->data->num_def_config)
> +		return;
> +
> +	for (i = 0; i < mmsys->data->num_def_config; i++)
> +		mtk_mmsys_update_bits(mmsys, def_config[i].offset, def_config[i].mask,
> +				      def_config[i].val, NULL);
> +}
> +EXPORT_SYMBOL_GPL(mtk_mmsys_default_config);
> +