On Wed Mar 19, 2025 at 8:15 AM CET, Michael Walle wrote:
> Hi,
>
> > GPIO controller often have support for IRQ: allow to easily allocate
> > both gpio-regmap and regmap-irq in one operation.
> >
> > Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@xxxxxxxxxxx>
> > ---
> > drivers/gpio/gpio-regmap.c | 23 +++++++++++++++++++++--
> > include/linux/gpio/regmap.h | 15 +++++++++++++++
> > 2 files changed, 36 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/gpio/gpio-regmap.c b/drivers/gpio/gpio-regmap.c
> > index 05f8781b5204..61d5f48b445d 100644
> > --- a/drivers/gpio/gpio-regmap.c
> > +++ b/drivers/gpio/gpio-regmap.c
> > @@ -203,6 +203,7 @@ EXPORT_SYMBOL_GPL(gpio_regmap_get_drvdata);
> > */
> > struct gpio_regmap *gpio_regmap_register(const struct gpio_regmap_config *config)
> > {
> > + struct irq_domain *irq_domain;
> > struct gpio_regmap *gpio;
> > struct gpio_chip *chip;
> > int ret;
> > @@ -280,8 +281,26 @@ struct gpio_regmap *gpio_regmap_register(const struct gpio_regmap_config *config
> > if (ret < 0)
> > goto err_free_gpio;
> >
> > - if (config->irq_domain) {
> > - ret = gpiochip_irqchip_add_domain(chip, config->irq_domain);
> > + irq_domain = config->irq_domain;
> > +#ifdef CONFIG_GPIOLIB_IRQCHIP
>
> Why do we need this ifdef?
>
Hum yes, on second thought we probably need to depend on
CONFIG_REGMAP_IRQ here.
> > + if (config->regmap_irq_chip) {
> > + struct regmap_irq_chip_data *irq_chip_data;
> > +
> > + ret = devm_regmap_add_irq_chip_fwnode(config->parent, dev_fwnode(config->parent),
> > + config->regmap, config->regmap_irq_irqno,
> > + config->regmap_irq_flags, 0,
> > + config->regmap_irq_chip, &irq_chip_data);
> > + if (ret)
> > + goto err_free_gpio;
> > +
> > + irq_domain = regmap_irq_get_domain(irq_chip_data);
> > + if (config->regmap_irq_chip_data)
> > + *config->regmap_irq_chip_data = irq_chip_data;
>
> I'm not a fan of misusing the config to return any data. Can we have
> a normal gpio_regmap_get_...()? Usually, the config is on the stack
> of the caller, what if you need to get irq_chip_data afterwards?
> Then your caller has to save it somewhere.
>
Yes, makes sense. As suggested by Andy Shevchenko, I will remove this
parameter as there is no user today: a way to retrieve it can be added
later if needed.
> Also, what is the advantage of this? Your caller doesn't have to
> call devm_regmap_add_irq_chip_fwnode(), but on the flip side you
> have to cram all its parameters in the gpio_regmap config. I'd like
> to keep that small and simple (but still extensible!). IMHO just
> setting the irq_domain is enough to achieve that.
This was a request from Andy on my previous series.
>
> -michael
>
Thanks for your review.
Mathieu
--
Mathieu Dubois-Briand, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
