Re: [RFC PATCH v3 2/8] leds: add function to configure hardware controlled LED

["Marek Behún" <kabel@kernel.org>]

On Tue, 9 Nov 2021 15:22:53 +0100
Ansuel Smith <ansuelsmth@gmail.com> wrote:

> On Tue, Nov 09, 2021 at 04:01:03AM +0100, Marek Behún wrote:
> > Hello Ansuel,
> > 
> > On Tue,  9 Nov 2021 03:26:02 +0100
> > Ansuel Smith <ansuelsmth@gmail.com> wrote:
> >   
> > > Add hw_control_configure helper to configure how the LED should work in
> > > hardware mode. The function require to support the particular trigger and
> > > will use the passed flag to elaborate the data and apply the
> > > correct configuration. This function will then be used by the trigger to
> > > request and update hardware configuration.
> > > 
> > > Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
> > > ---
> > >  Documentation/leds/leds-class.rst | 25 ++++++++++++++++++++
> > >  include/linux/leds.h              | 39 +++++++++++++++++++++++++++++++
> > >  2 files changed, 64 insertions(+)
> > > 
> > > diff --git a/Documentation/leds/leds-class.rst b/Documentation/leds/leds-class.rst
> > > index 645940b78d81..efd2f68c46a7 100644
> > > --- a/Documentation/leds/leds-class.rst
> > > +++ b/Documentation/leds/leds-class.rst
> > > @@ -198,6 +198,31 @@ With HARDWARE_CONTROLLED blink_mode hw_control_status/start/stop is optional
> > >  and any software only trigger will reject activation as the LED supports only
> > >  hardware mode.
> > >  
> > > +A trigger once he declared support for hardware controlled blinks, will use the function
> > > +hw_control_configure() provided by the driver to check support for a particular blink mode.
> > > +This function passes as the first argument (flag) a u32 flag.
> > > +The second argument (cmd) of hw_control_configure() method can be used to do various
> > > +operations for the specific blink mode. We currently support ENABLE, DISABLE, READ, ZERO
> > > +and SUPPORTED to enable, disable, read the state of the blink mode, ask the LED
> > > +driver if it does supports the specific blink mode and to reset any blink mode active.
> > > +
> > > +In ENABLE/DISABLE hw_control_configure() should configure the LED to enable/disable the
> > > +requested blink mode (flag).
> > > +In READ hw_control_configure() should return 0 or 1 based on the status of the requested
> > > +blink mode (flag).
> > > +In SUPPORTED hw_control_configure() should return 0 or 1 if the LED driver supports the
> > > +requested blink mode (flags) or not.
> > > +In ZERO hw_control_configure() should return 0 with success operation or error.
> > > +
> > > +The unsigned long flag is specific to the trigger and change across them. It's in the LED
> > > +driver interest know how to elaborate this flag and to declare support for a
> > > +particular trigger. For this exact reason explicit support for the specific
> > > +trigger is mandatory or the driver returns -EOPNOTSUPP if asked to enter offload mode
> > > +with a not supported trigger.
> > > +If the driver returns -EOPNOTSUPP on hw_control_configure(), the trigger activation will
> > > +fail as the driver doesn't support that specific offload trigger or doesn't know
> > > +how to handle the provided flags.
> > > +
> > >  Known Issues
> > >  ============
> > >  
> > > diff --git a/include/linux/leds.h b/include/linux/leds.h
> > > index cf0c6005c297..00bc4d6ed7ca 100644
> > > --- a/include/linux/leds.h
> > > +++ b/include/linux/leds.h
> > > @@ -73,6 +73,16 @@ enum led_blink_modes {
> > >  	SOFTWARE_HARDWARE_CONTROLLED,
> > >  };
> > >  
> > > +#ifdef CONFIG_LEDS_HARDWARE_CONTROL
> > > +enum blink_mode_cmd {
> > > +	BLINK_MODE_ENABLE, /* Enable the hardware blink mode */
> > > +	BLINK_MODE_DISABLE, /* Disable the hardware blink mode */
> > > +	BLINK_MODE_READ, /* Read the status of the hardware blink mode */
> > > +	BLINK_MODE_SUPPORTED, /* Ask the driver if the hardware blink mode is supported */
> > > +	BLINK_MODE_ZERO, /* Disable any hardware blink active */
> > > +};
> > > +#endif  
> > 
> > this is a strange proposal for the API.
> > 
> > Anyway, led_classdev already has the blink_set() method, which is documented as
> > 	/*
> > 	  * Activate hardware accelerated blink, delays are in milliseconds
> > 	  * and if both are zero then a sensible default should be chosen.
> > 	  * The call should adjust the timings in that case and if it can't
> > 	  * match the values specified exactly.
> > 	  * Deactivate blinking again when the brightness is set to LED_OFF
> > 	  * via the brightness_set() callback.
> > 	  */
> > 	int		(*blink_set)(struct led_classdev *led_cdev,
> > 				     unsigned long *delay_on,
> > 				     unsigned long *delay_off);
> > 
> > So we already have a method to set hardware blkinking, we don't need
> > another one.
> > 
> > Marek  
> 
> But that is about hardware blink, not a LED controlled by hardware based
> on some rules/modes.
> Doesn't really match the use for the hardware control.

I think there is a miscommunication, because I don't quite understand
what you are trying to say here.

How is "hardware blink" different from "a LED controlled by hardware
based on some rules/modes", when it is used for blinking ?

If the hardware, any hardware, supports blinking with different
frequencies, it should implement the blink_set() method.

> Blink_set makes the LED blink contantly at the declared delay.
> The blink_mode_cmd are used to request stuff to a LED in hardware mode.
> 
> Doesn't seem correct to change/enhance the blink_set function with
> something that would do something completely different.
>