[RFC] gpiolib: introduce descriptor-based GPIO interface

[Grant Likely]

On Thu, 6 Dec 2012 16:45:18 +0900, Alexandre Courbot <acourbot at nvidia.com> wrote:
> With the current API, GPIOs are manipulated through an integer which
> represents their unique number across the system. This poses problems in
> terms of portability, scalability and flexibility: for instance, the
> number of valid GPIOs for a given system is fixed at system time, and a
> large array of that size is statically allocated to hold the GPIO
> descriptors. Worse, GPIOs can be used without being properly allocated.
> 
> In order to improve the situation, the integer namespace must first get
> away. This patch introduces an alternative GPIO API that uses opaque
> handlers and refactor gpiolib's internals to work with these handlers
> instead of GPIO numbers. The former integer-based API is still available
> as a light wrapper around this new API.
> 
> This first step will then us to build more improvements for gpiolib,
> like proper GPIO lookup functions per device and provider, and getting
> rid of the static GPIO array and the ARCH_NR_GPIO configuration option.
> 
> Signed-off-by: Alexandre Courbot <acourbot at nvidia.com>
> ---
>  drivers/gpio/gpiolib.c        | 302 ++++++++++++++++++++++--------------------
>  include/asm-generic/gpio.h    |  74 ++++++++---
>  include/linux/gpio/consumer.h |  45 +++++++
>  3 files changed, 261 insertions(+), 160 deletions(-)
>  create mode 100644 include/linux/gpio/consumer.h
> 
> diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
> index 1c8d9e3..bf32511 100644
> --- a/drivers/gpio/gpiolib.c
> +++ b/drivers/gpio/gpiolib.c
> @@ -83,6 +83,32 @@ static inline void desc_set_label(struct gpio_desc *d, const char *label)
>  #endif
>  }
>  
> +/**
> + * Convert a GPIO descriptor to the integer namespace.
> + * This should disappear in the future but is needed since we still
> + * use GPIO numbers for error messages and sysfs nodes
> + */
> +static inline int desc_to_gpio(struct gpio_desc *desc)
> +{
> +	return desc - &gpio_desc[0];
> +}
> +
> +/**
> + * Return the GPIO number of the passed descriptor relative to its chip
> + */
> +int gpio_chip_offset(struct gpio_desc *desc)
> +{
> +	return (desc - &gpio_desc[0]) - desc->chip->base;
> +}

how about "gpio_chip_hwnum()" to somewhat match irqdomain convention?

I've only lightly scanned this patch, but I like what I see. I would
keep going with it.

g.