[2/3] leds: pwm: convert to atomic PWM API

pwm_config(), pwm_enable() and pwm_disable() should get removed in the
long run. So update the driver to use the atomic API that is here to
stay.

A few side effects:

 - led_pwm_set() now returns an error when setting the PWM fails.
 - During .probe() the PWM isn't disabled implicitly by pwm_apply_args()
   any more.

Signed-off-by: Uwe Kleine-König <uwe@kleine-koenig.org>
---
 drivers/leds/leds-pwm.c | 41 +++++++++--------------------------------
 1 file changed, 9 insertions(+), 32 deletions(-)

Hi Uwe,

On Fri, Jan 24, 2020 at 05:54:08PM +0100, Uwe Kleine-König wrote:
> pwm_config(), pwm_enable() and pwm_disable() should get removed in the
> long run. So update the driver to use the atomic API that is here to
> stay.
> 
> A few side effects:
> 
>  - led_pwm_set() now returns an error when setting the PWM fails.
>  - During .probe() the PWM isn't disabled implicitly by pwm_apply_args()
>    any more.
> 
> Signed-off-by: Uwe Kleine-König <uwe@kleine-koenig.org>
> ---
>  drivers/leds/leds-pwm.c | 41 +++++++++--------------------------------
>  1 file changed, 9 insertions(+), 32 deletions(-)
> 
> diff --git a/drivers/leds/leds-pwm.c b/drivers/leds/leds-pwm.c
> index b72fd89ff390..9111cdede0ee 100644
> --- a/drivers/leds/leds-pwm.c
> +++ b/drivers/leds/leds-pwm.c
> @@ -22,9 +22,8 @@
>  struct led_pwm_data {
>  	struct led_classdev	cdev;
>  	struct pwm_device	*pwm;
> +	struct pwm_state	pwmstate;
>  	unsigned int		active_low;
> -	unsigned int		period;
> -	int			duty;
>  };
>  
>  struct led_pwm_priv {
> @@ -32,44 +31,29 @@ struct led_pwm_priv {
>  	struct led_pwm_data leds[0];
>  };
>  
> -static void __led_pwm_set(struct led_pwm_data *led_dat)
> -{
> -	int new_duty = led_dat->duty;
> -
> -	pwm_config(led_dat->pwm, new_duty, led_dat->period);
> -
> -	if (new_duty == 0)
> -		pwm_disable(led_dat->pwm);
> -	else
> -		pwm_enable(led_dat->pwm);
> -}
> -
>  static int led_pwm_set(struct led_classdev *led_cdev,
>  		       enum led_brightness brightness)
>  {
>  	struct led_pwm_data *led_dat =
>  		container_of(led_cdev, struct led_pwm_data, cdev);
>  	unsigned int max = led_dat->cdev.max_brightness;
> -	unsigned long long duty =  led_dat->period;
> +	unsigned long long duty = led_dat->pwmstate.period;
>  
>  	duty *= brightness;
>  	do_div(duty, max);
>  
>  	if (led_dat->active_low)
> -		duty = led_dat->period - duty;
> -
> -	led_dat->duty = duty;
> -
> -	__led_pwm_set(led_dat);
> +		duty = led_dat->pwmstate.period - duty;
>  
> -	return 0;
> +	led_dat->pwmstate.duty_cycle = duty;
> +	led_dat->pwmstate.enabled = duty > 0;
> +	return pwm_apply_state(led_dat->pwm, &led_dat->pwmstate);
>  }
>  
>  static int led_pwm_add(struct device *dev, struct led_pwm_priv *priv,
>  		       struct led_pwm *led, struct fwnode_handle *fwnode)
>  {
>  	struct led_pwm_data *led_data = &priv->leds[priv->num_leds];
> -	struct pwm_args pargs;
>  	int ret;
>  
>  	led_data->active_low = led->active_low;
> @@ -93,17 +77,10 @@ static int led_pwm_add(struct device *dev, struct led_pwm_priv *priv,
>  
>  	led_data->cdev.brightness_set_blocking = led_pwm_set;
>  
> -	/*
> -	 * FIXME: pwm_apply_args() should be removed when switching to the
> -	 * atomic PWM API.
> -	 */
> -	pwm_apply_args(led_data->pwm);
> -
> -	pwm_get_args(led_data->pwm, &pargs);
> +	pwm_init_state(led_data->pwm, &led_data->pwmstate);
>  
> -	led_data->period = pargs.period;
> -	if (!led_data->period)
> -		led_data->period = led->pwm_period_ns;
> +	if (!led_data->pwmstate.period)
> +		led_data->pwmstate.period = led->pwm_period_ns;
>  
>  	ret = devm_led_classdev_register(dev, &led_data->cdev);
>  	if (ret == 0) {
> -- 
> 2.24.0
> 

Using the same setup described in my response to patch [1/3], I checked
the following:

* LED appears as expected with brightness continuously swept between 0
  and 255 at fixed intervals
* I2C traffic appears as expected for random cases of brightness being
  changed between 0 and !0 and back to 0

And so:

Tested-by: Jeff LaBundy <jeff@labundy.com>

Kind regards,
Jeff LaBundy

Hi!

> On Fri, Jan 24, 2020 at 05:54:08PM +0100, Uwe Kleine-König wrote:
> > pwm_config(), pwm_enable() and pwm_disable() should get removed in the
> > long run. So update the driver to use the atomic API that is here to
> > stay.
> > 
> > A few side effects:
> > 
> >  - led_pwm_set() now returns an error when setting the PWM fails.
> >  - During .probe() the PWM isn't disabled implicitly by pwm_apply_args()
> >    any more.
> > 
> > Signed-off-by: Uwe Kleine-König <uwe@kleine-koenig.org>

> And so:
> 
> Tested-by: Jeff LaBundy <jeff@labundy.com>

Thanks for patches, thanks for testing.

1,2 applied to -for-next.

AFAICT 3/3 should really get the default-state: keep support before
being merged?

Best regards,
									Pavel

Hi Pavel,

On 2/26/20 3:35 PM, Pavel Machek wrote:
>> On Fri, Jan 24, 2020 at 05:54:08PM +0100, Uwe Kleine-König wrote:
>>> pwm_config(), pwm_enable() and pwm_disable() should get removed in the
>>> long run. So update the driver to use the atomic API that is here to
>>> stay.
>>>
>>> A few side effects:
>>>
>>>  - led_pwm_set() now returns an error when setting the PWM fails.
>>>  - During .probe() the PWM isn't disabled implicitly by pwm_apply_args()
>>>    any more.
>>>
>>> Signed-off-by: Uwe Kleine-König <uwe@kleine-koenig.org>
> 
>> And so:
>>
>> Tested-by: Jeff LaBundy <jeff@labundy.com>
> 
> Thanks for patches, thanks for testing.
> 
> 1,2 applied to -for-next.
> 
> AFAICT 3/3 should really get the default-state: keep support before
> being merged?

Maybe I miss something, but I wonder why default-state: keep support
isn't implemented in the core but in each led driver instead. (And as
soon as someone implements it in the core, patch 3 is the right thing to
do.)

Best regards
Uwe