On Thu, Jul 21, 2022 at 10:03 AM Nuno Sá <nuno.sa@xxxxxxxxxx> wrote:
>
> This change replaces the support for GPIs as key event generators.
> Instead of reporting the events directly, we add a gpio based irqchip
> so that these events can be consumed by keys defined in the gpio-keys
> driver (as it's goal is indeed for keys on GPIOs capable of generating
> interrupts). With this, the gpio-adp5588 driver can also be dropped.
>
> The basic idea is that all the pins that are not being used as part of
> the keymap matrix can be possibly requested as GPIOs by gpio-keys
> (it's also fine to use these pins as plain interrupts though that's not
> really the point).
>
> Since the gpiochip now also has irqchip capabilities, we should only
> remove it after we free the device interrupt (otherwise we could, in
> theory, be handling GPIs interrupts while the gpiochip is concurrently
> removed). Thus the call 'adp5588_gpio_add()' is moved and since the
> setup phase also needs to come before making the gpios visible, we also
> need to move 'adp5588_setup()'.
>
> While at it, always select GPIOLIB so that we don't need to use #ifdef
> guards.
...
> +static void adp5588_irq_mask(struct irq_data *d)
> +{
> + struct gpio_chip *gc = irq_data_get_irq_chip_data(d);
> + struct adp5588_kpad *kpad = gpiochip_get_data(gc);
Can it be
irq_hw_number_t hwirq = irqd_to_hwirq(d);
?
> + unsigned long real_irq = kpad->gpiomap[irqd_to_hwirq(d)];
> +
> + kpad->irq_mask[ADP5588_BANK(real_irq)] &= ~ADP5588_BIT(real_irq);
> + gpiochip_disable_irq(gc, irqd_to_hwirq(d));
> +}
> +
> +static void adp5588_irq_unmask(struct irq_data *d)
> +{
> + struct gpio_chip *gc = irq_data_get_irq_chip_data(d);
> + struct adp5588_kpad *kpad = gpiochip_get_data(gc);
> + unsigned long real_irq = kpad->gpiomap[irqd_to_hwirq(d)];
> +
> + gpiochip_enable_irq(gc, irqd_to_hwirq(d));
> + kpad->irq_mask[ADP5588_BANK(real_irq)] |= ADP5588_BIT(real_irq);
Ditto.
> +}
...
> + /* gpio line used as IRQ source */
GPIO
--
With Best Regards,
Andy Shevchenko
