Re: [PATCH 03/13] clk: implement clk_hw_set_rate()

[Maxime Ripard]

Hi,

On Mon, Sep 18, 2023 at 12:39:59AM +0200, Benjamin Bara wrote:
> From: Benjamin Bara <benjamin.bara@xxxxxxxxxxx>
> 
> This function can be used to set the rate of a clock hardware from a
> driver, e.g. to adapt the rate to a clock change coming from the parent.
> 
> Signed-off-by: Benjamin Bara <benjamin.bara@xxxxxxxxxxx>
> ---
>  drivers/clk/clk.c            | 15 +++++++++++++++
>  include/linux/clk-provider.h |  1 +
>  2 files changed, 16 insertions(+)
> 
> diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c
> index c249f9791ae8..3e222802b712 100644
> --- a/drivers/clk/clk.c
> +++ b/drivers/clk/clk.c
> @@ -2462,6 +2462,21 @@ static int clk_core_set_rate_nolock(struct clk_core *core,
>  	return ret;
>  }
>  
> +int clk_hw_set_rate(struct clk_hw *hw, unsigned long req_rate)
> +{
> +	/* A rate change is ongoing, so just target the required rate.
> +	 * Note: this does not work if one clock along the line has
> +	 * CLK_RECALC_NEW_RATES active, as this overwrites the new_rate again.
> +	 */
> +	if (hw->core->new_rate != hw->core->rate) {
> +		hw->core->new_rate = req_rate;
> +		return 0;
> +	}
> +
> +	return clk_core_set_rate_nolock(hw->core, req_rate);
> +}
> +EXPORT_SYMBOL_GPL(clk_hw_set_rate);

So, we discussed it recently, and it's non-obvious to see how it would
work if you're in a set_rate change that modifies the tree and the new
call to clk_hw_set_rate modifies the tree too. Some explanation on how
it's handled, and some unit tests are required here imo.

Maxime

Attachment: signature.asc
Description: PGP signature