Re: [PATCH 06/13] gpio: add AD24xx GPIO driver

[Andy Shevchenko]

Fri, May 17, 2024 at 02:58:04PM +0200, Alvin Šipraga kirjoitti:
> From: Alvin Šipraga <alsi@xxxxxxxxxxxxxxx>
> 
> This driver adds GPIO function support for AD24xx A2B transceiver chips.

"Add GPIO..."

> When a GPIO is requested, the relevant pin is automatically muxed to
> GPIO mode. The device tree property gpio-reserved-ranges can be used to
> protect certain pins which are reserved for other functionality such as
> I2S/TDM data.

Why this doesn't use gpio-regmap?

..

> +config GPIO_AD24XX
> +	tristate "Analog Devies Inc. AD24xx GPIO support"
> +	depends on A2B_AD24XX_NODE
> +	help
> +	  Say Y here to enable GPIO support for AD24xx A2B transceivers.

checkpatch probably complain about too short help text. You may extend it by
explaining how module will be called.

..

> +#include <linux/a2b/a2b.h>
> +#include <linux/a2b/ad24xx.h>

This seems to me not so generic as below...

+ bits.h
+ device.h
+ err.h

> +#include <linux/gpio/driver.h>

> +#include <linux/interrupt.h>

+ mod_devicetable.h

> +#include <linux/module.h>

+ mutex.h

> +#include <linux/of_irq.h>

Please, can we avoid OF in a new code?

> +#include <linux/regmap.h>

..hence move that group here and put a blank line before.

..

> +struct ad24xx_gpio {
> +	struct device *dev;
> +	struct a2b_func *func;
> +	struct a2b_node *node;
> +	struct regmap *regmap;
> +	int irqs[AD24XX_MAX_GPIOS];

> +	struct gpio_chip gpio_chip;

If you move this to be the first member, you might get less code being
generated at compile time.

> +	struct irq_chip irq_chip;

Should not be here, but static.

> +	struct mutex mutex;
> +	unsigned int irq_invert : AD24XX_MAX_GPIOS;
> +	unsigned int irq_enable : AD24XX_MAX_GPIOS;
> +};

..

> +	if (ret)
> +		dev_err(adg->dev,
> +			"failed to update interrupt configuration: %d\n", ret);

Why and how is this useful?

..

> +	struct fwnode_handle *fwnode = of_node_to_fwnode(dev->of_node);

First of all it uses a wrong API (custom to IRQ core), second why do you need
this?

..

> +	struct device_node *np;

> +	np = of_irq_find_parent(dev->of_node);
> +	if (!np)
> +		return -ENOENT;
> +
> +	parent_domain = irq_find_host(np);
> +	of_node_put(np);
> +	if (!parent_domain)
> +		return -ENOENT;

Why is this magic needed?

..

> +	ret = devm_gpiochip_add_data(dev, gpio_chip, adg);
> +	if (ret)
> +		return ret;
> +
> +	return 0;

	return devm_gpiochip_add_data(...);

-- 
With Best Regards,
Andy Shevchenko