Re: [PATCH 2/2] gpio: bcm-kona: use platform_irq_count

[Uwe Kleine-König <u.kleine-koenig@xxxxxxxxxxxxxx>]

On Tue, Dec 03, 2019 at 07:23:14AM +0000, Peng Fan wrote:
> > Subject: Re: [PATCH 2/2] gpio: bcm-kona: use platform_irq_count
> > 
> > On Tue, Dec 03, 2019 at 07:00:40AM +0000, Peng Fan wrote:
> > > > Subject: Re: [PATCH 2/2] gpio: bcm-kona: use platform_irq_count
> > > >
> > > > On Tue, Dec 03, 2019 at 06:04:27AM +0000, Peng Fan wrote:
> > > > > From: Peng Fan <peng.fan@xxxxxxx>
> > > > >
> > > > > Use platform_irq_count to replace of_irq_count
> > > > >
> > > > > Signed-off-by: Peng Fan <peng.fan@xxxxxxx>
> > > > > ---
> > > > >
> > > > > V1:
> > > > >  Code inspection, not tested
> > > > >
> > > > >  drivers/gpio/gpio-bcm-kona.c | 3 +--
> > > > >  1 file changed, 1 insertion(+), 2 deletions(-)
> > > > >
> > > > > diff --git a/drivers/gpio/gpio-bcm-kona.c
> > > > > b/drivers/gpio/gpio-bcm-kona.c index 4122683eb1f9..c50721980a7c
> > > > 100644
> > > > > --- a/drivers/gpio/gpio-bcm-kona.c
> > > > > +++ b/drivers/gpio/gpio-bcm-kona.c
> > > > > @@ -19,7 +19,6 @@
> > > > >  #include <linux/io.h>
> > > > >  #include <linux/gpio/driver.h>
> > > > >  #include <linux/of_device.h>
> > > > > -#include <linux/of_irq.h>
> > > > >  #include <linux/init.h>
> > > > >  #include <linux/irqdomain.h>
> > > > >  #include <linux/irqchip/chained_irq.h> @@ -586,7 +585,7 @@ static
> > > > > int bcm_kona_gpio_probe(struct platform_device *pdev)
> > > > >
> > > > >  	kona_gpio->gpio_chip = template_chip;
> > > > >  	chip = &kona_gpio->gpio_chip;
> > > > > -	kona_gpio->num_bank = of_irq_count(dev->of_node);
> > > > > +	kona_gpio->num_bank = platform_irq_count(pdev);
> > > >
> > > > of_irq_count returns 0 or a positive int while platform_irq_count
> > > > might return a negative error code. This needs handling. Also I
> > > > wonder why
> > > > platform_irq_count() is better than of_irq_count() which would be
> > > > good to describe in the commit log.
> > >
> > > Inspired from
> > > https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flkml
> > > .org%2Flkml%2F2015%2F11%2F18%2F466&amp;data=02%7C01%7Cpeng.f
> > an%40nxp.c
> > >
> > om%7C46364bf12479463df7d308d777c09b39%7C686ea1d3bc2b4c6fa92cd9
> > 9c5c3016
> > >
> > 35%7C0%7C0%7C637109541440912984&amp;sdata=awTvwwmo4692Hx7IJ
> > H%2BllEVJH7
> > > ngINufoMH8UsosU%2BA%3D&amp;reserved=0
> > > From Rob:
> > > "
> > > So I started looking at why you are using of_irq_count which drivers
> > > shouldn't need to. In patch 5 you use it to allocate memory to store
> > > the irq names, then use them here...
> > > "
> > >
> > > Is this ok?
> > 
> > I would say something like:
> > 
> > 	platform_irq_count() is the more generic way (independent of
> > 	device trees) to determine the count of available interrupts. So
> > 	use this instead.
> > 
> > 	As platform_irq_count() might return an error code (which
> > 	of_irq_count doesn't) some additional handling is necessary.
> 
> Thanks, how about this change?
> @@ -586,11 +585,15 @@ static int bcm_kona_gpio_probe(struct platform_device *pdev)
> 
>         kona_gpio->gpio_chip = template_chip;
>         chip = &kona_gpio->gpio_chip;
> -       kona_gpio->num_bank = of_irq_count(dev->of_node);
> -       if (kona_gpio->num_bank == 0) {
> +       ret = platform_irq_count(pdev);
> +       if (!ret) {
>                 dev_err(dev, "Couldn't determine # GPIO banks\n");
>                 return -ENOENT;
> +       } else if (ret < 0) {
> +               return ret;

This is inconsitent. In the ret==0 case you emit an error message, in
the ret < 0 case you don't. I think the sensible approach would be to
do:

	if (ret == 0) {
		dev_err(dev, "Couldn't determine # GPIO banks\n");
		return -ENOENT;
	} else if (ret < 0) {
		if (ret != -EPROBE_DEFER)
			dev_err(dev, "Failed to determine count of GPIO banks (%pe)\n", ERR_PTR(ret));
		return ret;
	}

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | https://www.pengutronix.de/ |