[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAMRc=Mejnr8UzN93X=CWcV5jDTt9-U+Nxcm3qb=6uVV0PMiZVQ@mail.gmail.com>
Date: Wed, 23 Jul 2025 10:39:32 +0200
From: Bartosz Golaszewski <brgl@...ev.pl>
To: Arnd Bergmann <arnd@...nel.org>
Cc: Linus Walleij <linus.walleij@...aro.org>, Arnd Bergmann <arnd@...db.de>, Peng Fan <peng.fan@....com>,
Lee Jones <lee@...nel.org>, Koichiro Den <koichiro.den@...onical.com>,
Geert Uytterhoeven <geert+renesas@...der.be>,
Andy Shevchenko <andriy.shevchenko@...ux.intel.com>,
Alexander Sverdlin <alexander.sverdlin@...il.com>, Uwe Kleine-König <ukleinek@...nel.org>,
linux-gpio@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] gpiolib: make legacy interfaces optional
On Tue, Jul 22, 2025 at 5:36 PM Arnd Bergmann <arnd@...nel.org> wrote:
>
> From: Arnd Bergmann <arnd@...db.de>
>
> The traditional interfaces are only used on a small number of ancient
> boards. Make these optional now so they can be disabled by default.
>
> Signed-off-by: Arnd Bergmann <arnd@...db.de>
> ---
> This is the first patch of a series to turn off the legacy interfaces
> by default. If we can still have this one in linux-6.17, we can more
> easily merge the other patches for 6.18.
>
> See for the longer series:
> https://git.kernel.org/pub/scm/linux/kernel/git/arnd/playground.git/log/?h=config-gpio-legacy
>
> I'm sure there are still problems in the other patches, but it
> does pass my randconfig build tests on the three architectures
> I'm testing on. I plan to post them after some more testing
> once -rc1 is out.
>
> drivers/gpio/Kconfig | 3 +++
> drivers/gpio/Makefile | 2 +-
> include/linux/gpio.h | 10 ++++++----
> 3 files changed, 10 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpio/Kconfig b/drivers/gpio/Kconfig
> index 12bdf6e965f1..8bda3c9d47b4 100644
> --- a/drivers/gpio/Kconfig
> +++ b/drivers/gpio/Kconfig
> @@ -14,6 +14,9 @@ menuconfig GPIOLIB
>
> if GPIOLIB
>
> +config GPIOLIB_LEGACY
> + def_bool y
> +
> config GPIOLIB_FASTPATH_LIMIT
> int "Maximum number of GPIOs for fast path"
> range 32 512
> diff --git a/drivers/gpio/Makefile b/drivers/gpio/Makefile
> index 88dedd298256..b01ff2b68bf6 100644
> --- a/drivers/gpio/Makefile
> +++ b/drivers/gpio/Makefile
> @@ -5,7 +5,7 @@ ccflags-$(CONFIG_DEBUG_GPIO) += -DDEBUG
>
> obj-$(CONFIG_GPIOLIB) += gpiolib.o
> obj-$(CONFIG_GPIOLIB) += gpiolib-devres.o
> -obj-$(CONFIG_GPIOLIB) += gpiolib-legacy.o
> +obj-$(CONFIG_GPIOLIB_LEGACY) += gpiolib-legacy.o
> obj-$(CONFIG_OF_GPIO) += gpiolib-of.o
> obj-$(CONFIG_GPIO_CDEV) += gpiolib-cdev.o
> obj-$(CONFIG_GPIO_SYSFS) += gpiolib-sysfs.o
> diff --git a/include/linux/gpio.h b/include/linux/gpio.h
> index ff99ed76fdc3..8f85ddb26429 100644
> --- a/include/linux/gpio.h
> +++ b/include/linux/gpio.h
> @@ -13,6 +13,11 @@
> #define __LINUX_GPIO_H
>
> #include <linux/types.h>
> +#ifdef CONFIG_GPIOLIB
> +#include <linux/gpio/consumer.h>
I want to queue this ASAP but do we really need this guard here?
consumer.h already guards against !CONFIG_GPIOLIB internally, right?
Bart
> +#endif
> +
> +#ifdef CONFIG_GPIOLIB_LEGACY
>
> struct device;
>
> @@ -22,9 +27,6 @@ struct device;
> #define GPIOF_OUT_INIT_HIGH ((0 << 0) | (1 << 1))
>
> #ifdef CONFIG_GPIOLIB
> -
> -#include <linux/gpio/consumer.h>
> -
> /*
> * "valid" GPIO numbers are nonnegative and may be passed to
> * setup routines like gpio_request(). Only some valid numbers
> @@ -170,5 +172,5 @@ static inline int devm_gpio_request_one(struct device *dev, unsigned gpio,
> }
>
> #endif /* ! CONFIG_GPIOLIB */
> -
> +#endif /* CONFIG_GPIOLIB_LEGACY */
> #endif /* __LINUX_GPIO_H */
> --
> 2.39.5
>
Powered by blists - more mailing lists