Re: [PATCH v8 2/3] Input: adp5588-keys - add support for pure gpio

["Nuno Sá" <noname.nuno@gmail.com>]

On Fri, 2024-08-02 at 17:37 -0700, Dmitry Torokhov wrote:
> Hi Utsav,
> 
> On Thu, Jul 04, 2024 at 03:26:31PM +0100, Utsav Agarwal via B4 Relay wrote:
> > From: Utsav Agarwal <utsav.agarwal@analog.com>
> > 
> > Keypad specific setup is relaxed if no keypad rows/columns are specified,
> > enabling a purely gpio operation.
> > 
> > Signed-off-by: Utsav Agarwal <utsav.agarwal@analog.com>
> > ---
> >  drivers/input/keyboard/adp5588-keys.c | 37 +++++++++++++++++++++++++++++++-
> > ---
> >  1 file changed, 33 insertions(+), 4 deletions(-)
> > 
> > diff --git a/drivers/input/keyboard/adp5588-keys.c
> > b/drivers/input/keyboard/adp5588-keys.c
> > index 09bcfc6b9408..7c32f8b69a3e 100644
> > --- a/drivers/input/keyboard/adp5588-keys.c
> > +++ b/drivers/input/keyboard/adp5588-keys.c
> > @@ -188,6 +188,7 @@ struct adp5588_kpad {
> >  	u32 cols;
> >  	u32 unlock_keys[2];
> >  	int nkeys_unlock;
> > +	bool gpio_only;
> >  	unsigned short keycode[ADP5588_KEYMAPSIZE];
> >  	unsigned char gpiomap[ADP5588_MAXGPIO];
> >  	struct gpio_chip gc;
> > @@ -431,10 +432,12 @@ static int adp5588_gpio_add(struct adp5588_kpad *kpad)
> >  	kpad->gc.label = kpad->client->name;
> >  	kpad->gc.owner = THIS_MODULE;
> >  
> > -	girq = &kpad->gc.irq;
> > -	gpio_irq_chip_set_chip(girq, &adp5588_irq_chip);
> > -	girq->handler = handle_bad_irq;
> > -	girq->threaded = true;
> > +	if (kpad->client->irq) {
> > +		girq = &kpad->gc.irq;
> > +		gpio_irq_chip_set_chip(girq, &adp5588_irq_chip);
> > +		girq->handler = handle_bad_irq;
> > +		girq->threaded = true;
> > +	}
> >  
> >  	mutex_init(&kpad->gpio_lock);
> >  
> > @@ -632,6 +635,21 @@ static int adp5588_fw_parse(struct adp5588_kpad *kpad)
> >  	struct i2c_client *client = kpad->client;
> >  	int ret, i;
> >  
> > +	/*
> > +	 * Check if the device is to be operated purely in GPIO mode. To do
> > +	 * so, check that no keypad rows or columns have been specified,
> > +	 * since all GPINS should be configured as GPIO.
> > +	 */
> > +	ret = device_property_present(&client->dev,
> > +			"keypad,num-rows");
> > +	ret |= device_property_present(&client->dev,
> > +			"keypad,num-columns");
> > +	/* If purely GPIO, skip keypad setup */
> > +	if (!ret) {
> > +		kpad->gpio_only = true;
> > +		return 0;
> > +	}
> > +
> >  	ret = matrix_keypad_parse_properties(&client->dev, &kpad->rows,
> >  					     &kpad->cols);
> >  	if (ret)
> > @@ -775,6 +793,11 @@ static int adp5588_probe(struct i2c_client *client)
> >  	if (error)
> >  		return error;
> >  
> > +	if (kpad->gpio_only && !client->irq) {
> > +		dev_info(&client->dev, "Rev.%d, started as GPIO only\n",
> > revid);
> > +		return 0;
> 
> I think we need more elaborate handling here (and probably more
> elaborate binding yaml file): now that you are making interrupt optional
> you should check if interrupt-controller functionality of the GPIO
> block/gpiochip is requested. If it was, then we should not allow missing
> interrupt. If only GPIO controller is needed, without interrupt
> capabilities, tnen running without interrupt is fine.
> 
Hi Dmitry,

I need to double check but I don't think we can act as an interrupt-controller
without the interrupt line connected. So, I think the only thing we could likely
add/improve is to express that dependency in the bindings.

- Nuno Sá