Re: [PATCH v4 1/2] mfd: rohm-bd71828: Use software nodes for gpio-keys

[Lee Jones <lee@kernel.org>]

On Thu, 07 May 2026, Dmitry Torokhov wrote:

> On Thu, May 07, 2026 at 03:42:47PM +0100, Lee Jones wrote:
> > On Mon, 27 Apr 2026, Dmitry Torokhov wrote:
> > 
> > > Refactor the rohm-bd71828 MFD driver to use software nodes for
> > > instantiating the gpio-keys child device, replacing the old
> > > platform_data mechanism.
> > > 
> > > The power key's properties are now defined using software nodes and
> > > property entries. The IRQ is passed as a resource attached to the
> > > platform device.
> > > 
> > > This will allow dropping support for using platform data for configuring
> > > gpio-keys in the future.
> > > 
> > > Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
> > > ---
> > >  drivers/mfd/rohm-bd71828.c | 122 +++++++++++++++++++++++++++++++++------------
> > >  1 file changed, 90 insertions(+), 32 deletions(-)
> > > 
> > > diff --git a/drivers/mfd/rohm-bd71828.c b/drivers/mfd/rohm-bd71828.c
> > > index a79f354bf5cb..a8bdb9c955a4 100644
> > > --- a/drivers/mfd/rohm-bd71828.c
> > > +++ b/drivers/mfd/rohm-bd71828.c
> > > @@ -5,7 +5,8 @@
> > >   * ROHM BD718[15/28/79] and BD72720 PMIC driver
> > >   */
> > >  
> > > -#include <linux/gpio_keys.h>
> > > +#include <linux/device/devres.h>
> > > +#include <linux/gfp_types.h>
> > >  #include <linux/i2c.h>
> > >  #include <linux/input.h>
> > >  #include <linux/interrupt.h>
> > > @@ -18,6 +19,7 @@
> > >  #include <linux/mfd/rohm-generic.h>
> > >  #include <linux/module.h>
> > >  #include <linux/of.h>
> > > +#include <linux/property.h>
> > >  #include <linux/regmap.h>
> > >  #include <linux/types.h>
> > >  
> > > @@ -37,19 +39,6 @@
> > >  		},							   \
> > >  	}
> > >  
> > > -static struct gpio_keys_button button = {
> > > -	.code = KEY_POWER,
> > > -	.gpio = -1,
> > > -	.type = EV_KEY,
> > > -	.wakeup = 1,
> > > -};
> > > -
> > > -static const struct gpio_keys_platform_data bd71828_powerkey_data = {
> > > -	.buttons = &button,
> > > -	.nbuttons = 1,
> > > -	.name = "bd71828-pwrkey",
> > > -};
> > > -
> > >  static const struct resource bd71815_rtc_irqs[] = {
> > >  	DEFINE_RES_IRQ_NAMED(BD71815_INT_RTC0, "bd70528-rtc-alm-0"),
> > >  	DEFINE_RES_IRQ_NAMED(BD71815_INT_RTC1, "bd70528-rtc-alm-1"),
> > > @@ -174,11 +163,8 @@ static struct mfd_cell bd71828_mfd_cells[] = {
> > >  		.name = "bd71828-rtc",
> > >  		.resources = bd71828_rtc_irqs,
> > >  		.num_resources = ARRAY_SIZE(bd71828_rtc_irqs),
> > > -	}, {
> > > -		.name = "gpio-keys",
> > > -		.platform_data = &bd71828_powerkey_data,
> > > -		.pdata_size = sizeof(bd71828_powerkey_data),
> > >  	},
> > > +	/* Power button is registered separately */
> > 
> > This happens a lot in MFD - no need to call it out.
> 
> OK.
> 
> > 
> > >  };
> > >  
> > >  static const struct resource bd72720_power_irqs[] = {
> > > @@ -242,11 +228,8 @@ static const struct mfd_cell bd72720_mfd_cells[] = {
> > >  		.name = "bd72720-rtc",
> > >  		.resources = bd72720_rtc_irqs,
> > >  		.num_resources = ARRAY_SIZE(bd72720_rtc_irqs),
> > > -	}, {
> > > -		.name = "gpio-keys",
> > > -		.platform_data = &bd71828_powerkey_data,
> > > -		.pdata_size = sizeof(bd71828_powerkey_data),
> > >  	},
> > > +	/* Power button is registered separately */
> > >  };
> > >  
> > >  static const struct regmap_range bd71815_volatile_ranges[] = {
> > > @@ -877,6 +860,80 @@ static int set_clk_mode(struct device *dev, struct regmap *regmap,
> > >  				  OUT32K_MODE_CMOS);
> > >  }
> > >  
> > > +static int bd71828_i2c_register_swnodes(const struct software_node *nodes)
> > > +{
> > > +	const struct software_node * const node_group[] = {
> > > +		&nodes[0], &nodes[1], NULL
> > > +	};
> > 
> > I only see handling like this in the testing infra.
> > 
> > This is all very opaque and fiddly.
> > 
> > Are you sure we can't do better with statically declared arrays?
> 
> The nodes represent per-device data, so they can't be static/shared if
> we want to continue using the non-singleton approach in the driver.

If the handling has to stay the same, then we need to at least present
it in a nicer way.  Manually playing with nameless array indexes looks
super hacky.

> > > +	return software_node_register_node_group(node_group);
> > > +}
> > > +
> > > +static void bd71828_i2c_unregister_swnodes(void *data)
> > > +{
> > > +	const struct software_node *nodes = data;
> > > +	const struct software_node * const node_group[] = {
> > > +		&nodes[0], &nodes[1], NULL
> > > +	};
> > > +
> > > +	software_node_unregister_node_group(node_group);
> > > +}
> > > +
> > > +static int bd71828_i2c_register_pwrbutton(struct device *dev, int button_irq,
> > > +					  struct irq_domain *irq_domain)
> > > +{
> > > +	static const struct property_entry bd71828_powerkey_parent_props[] = {
> > > +		PROPERTY_ENTRY_STRING("label", "bd71828-pwrkey"),
> > > +		{ }
> > > +	};
> > > +	static const struct property_entry bd71828_powerkey_props[] = {
> > > +		PROPERTY_ENTRY_U32("linux,code", KEY_POWER),
> > > +		PROPERTY_ENTRY_BOOL("wakeup-source"),
> > > +		{ }
> > > +	};
> > > +	const struct resource res[] = {
> > > +		DEFINE_RES_IRQ_NAMED(button_irq, "bd71828-pwrkey"),
> > > +	};
> > > +	struct mfd_cell gpio_keys_cell = {
> > > +		.name = "gpio-keys",
> > > +		.resources = res,
> > > +		.num_resources = ARRAY_SIZE(res),
> > > +	};
> > > +
> > +
> > +Please break all 3 of these out of function context.
> > +
> > +We nearly always declare these externally unless they contain dynamic
> > +values and even then we try and avoid it.
> 
> "button_irq" is not a constant, so we need to have "res[]" and therefore
> gpio_keys_cell as locals. I can move out the properties, but I believe
> there is a value in grouping them together.

They don't need to be locals, they can be !const and you can fill them in.

Again, this is common practice.

-- 
Lee Jones