Re: [PATCH v3] leds: Add MAX6956 driver

["Uwe Kleine-König" <u.kleine-koenig@pengutronix.de>]

On Mon, May 21, 2012 at 11:30:28PM +0200, Linus Walleij wrote:
> On Mon, May 21, 2012 at 9:33 PM, Uwe Kleine-König
> <u.kleine-koenig@pengutronix.de> wrote:
> 
> > +config LEDS_MAX6956
> > +       tristate "LED support for MAX6956 LED Display Driver and I/O Expander"
> > +       depends on LEDS_CLASS
> > +       depends on GPIOLIB
> 
> Shouldn't this be select GPIOLIB?
> 
> You seem to require it when using this driver.
> 
> Better than hiding it if not selecting GPIOLIB somewhere else?
I don't care much, but:

$ git ls-files | grep Kconfig | xargs grep '\<GPIOLIB\>' | grep -c select
7
$ git ls-files | grep Kconfig | xargs grep '\<GPIOLIB\>' | grep -c depends
39

Is it save to select GPIOLIB on a machine that provides its own
gpio-API?

> > +struct max6956_ddata {
> > +       struct device *dev;
> > +
> > +       struct mutex lock;
> > +
> > +       struct regmap *regmap;
> > +
> > +       struct gpio_chip gpio_chip;
> > +
> > +       struct max6956_pdata pdata;
> > +
> > +       struct max6956_led_ddata leds[32];
> > +
> > +       const char *gpio_names[32];
> > +};
> 
> You can never have enough whitespace? ;-)
> 
> Anyway, so this thing has a gpio_chip and leds.
> 
> The archaic way is to create mfd/max-6956.c and have this
> MFD device spawn two cells, one for GPIO landing in
> driver/gpio/gpio-max6956.c and one for LED landing in
> leds/leds-max6956.c, then mediate register read/writes and
> regmap in the MFD driver.
> 
> The MFD driver decide using platform data whether each
> line should be used for a LED or GPIO.
> 
> Is there some problem with this design pattern?
I thought about that, too, but I think it's overkill to create an mfd
driver. The mfd driver would have essentially the same functions as the
driver I posted. Then add all the oneline wrappers for these added in
drivers/gpio/ and drivers/leds/. I'd expect the SLOC to double even if I
remove all the whitespace above. Moreover it increases complexity for a
driver that is quite simple otherwise.

There are two other drivers that handle gpios below drivers/leds
(leds-pca9532 and leds-tca6507). Are these bad examples? The chip can
only do leds and gpio so the argument won't change in the future.

Best regards
Uwe

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