Re: [PATCH 2/2] pwm: meson: support meson A1 SoC family

[George Stark]

Hello Jerome

Thanks for the review


On 4/23/24 20:35, Jerome Brunet wrote:
> On Tue 23 Apr 2024 at 19:10, George Stark <gnstark@xxxxxxxxxxxxxxxxx> wrote:
>
> > From: George Stark <gnstark@xxxxxxxxxxxxxx>
> >
> > Add a compatible string and configuration for the meson A1 SoC family
> > PWM. Additionally, provide an external clock initialization helper
> > specifically designed for these PWM IPs.
> >
> > Signed-off-by: George Stark <gnstark@xxxxxxxxxxxxxx>
> > Signed-off-by: Dmitry Rokosov <ddrokosov@xxxxxxxxxxxxxxxxx>
> > ---
> >   drivers/pwm/pwm-meson.c | 35 +++++++++++++++++++++++++++++++++++
> >   1 file changed, 35 insertions(+)
> >
> > diff --git a/drivers/pwm/pwm-meson.c b/drivers/pwm/pwm-meson.c
> > index ea96c5973488..529a541ba7b6 100644
> > --- a/drivers/pwm/pwm-meson.c
> > +++ b/drivers/pwm/pwm-meson.c
> > @@ -462,6 +462,33 @@ static int meson_pwm_init_channels_meson8b_v2(struct pwm_chip *chip)
> >   	return meson_pwm_init_clocks_meson8b(chip, mux_parent_data);
> >   }
> >   
> > +static int meson_pwm_init_channels_ext_clock(struct pwm_chip *chip)
>
> That kind on naming (ext) is almost sure to clash with whatever comes next.
> Just use the name of the first SoC using the method, a1 for instance.

It's true that pwm core in a1 s4, t7 etc is the same AFAWK.
I just want to clarify your proposal:
I add a1 compatible to the dt-bindings with s4 as fallback,
t7 compatible will be added later in the same way.

Here in the driver I don't mention a1 at all and use s4-centric naming e.g.:

{
	.compatible = "amlogic,meson-s4-pwm",
	.data = &pwm_meson_s4_data
},
static const struct meson_pwm_data pwm_meson_s4_data = {
	.channels_init = meson_pwm_init_channels_s4,
};

right?

> > +{
> > +	struct device *dev = pwmchip_parent(chip);
> > +	struct meson_pwm *meson = to_meson_pwm(chip);
> > +	struct meson_pwm_channel *channels = meson->channels;
> > +	struct clk_bulk_data *clks = NULL;
> > +	unsigned int i;
> > +	int res;
> > +
> > +	res = devm_clk_bulk_get_all(dev, &clks);
> > +	if (res < 0) {
> > +		dev_err(dev, "can't get device clocks\n");
> > +		return res;
> > +	}
>
> I don't think allocating the 'clk_bulk_data *clks' is necessary or safe.
> We know exactly how many clocks we expect, there is no need for a get all.
>
> > +
> > +	if (res != MESON_NUM_PWMS) {
> > +		dev_err(dev, "clock count must be %d, got %d\n",
> > +			MESON_NUM_PWMS, res);
> > +		return -EINVAL;
> > +	}
>
> ... and this only catches the problem after the fact.
>
> It is probably convinient but not necessary.
>
> > +
> > +	for (i = 0; i < MESON_NUM_PWMS; i++)
> > +		channels[i].clk = clks[i].clk;
>
> channels[i].clk could be assigned directly of_clk_get() using clock
> indexes. No extra allocation needed.

if we use of_clk_get then we'll have to free the clock objects in the
end. Could we use devm_clk_bulk_get instead?

> > +
> > +	return 0;
> > +}
> > +
> >   static const struct meson_pwm_data pwm_meson8b_data = {
> >   	.parent_names = { "xtal", NULL, "fclk_div4", "fclk_div3" },
> >   	.channels_init = meson_pwm_init_channels_meson8b_legacy,
> > @@ -500,11 +527,19 @@ static const struct meson_pwm_data pwm_meson8_v2_data = {
> >   	.channels_init = meson_pwm_init_channels_meson8b_v2,
> >   };
> >   
> > +static const struct meson_pwm_data pwm_meson_ext_clock_data = {
> > +	.channels_init = meson_pwm_init_channels_ext_clock,
> > +};
> > +
> >   static const struct of_device_id meson_pwm_matches[] = {
> >   	{
> >   		.compatible = "amlogic,meson8-pwm-v2",
> >   		.data = &pwm_meson8_v2_data
> >   	},
> > +	{
> > +		.compatible = "amlogic,meson-a1-pwm",
> > +		.data = &pwm_meson_ext_clock_data
> > +	},
> >   	/* The following compatibles are obsolete */
> >   	{
> >   		.compatible = "amlogic,meson8b-pwm",

--
Best regards
George