Re: [PATCH 3/5] gpio: ath79: Make the driver removable

From: Alexandre Courbot
Date: Sun Feb 07 2016 - 22:04:57 EST


On Fri, Jan 29, 2016 at 4:44 AM, Alban Bedel <albeu@xxxxxxx> wrote:
> As we now allow the driver to be built as a module it should be
> removable.
>
> Signed-off-by: Alban Bedel <albeu@xxxxxxx>
> ---
> drivers/gpio/gpio-ath79.c | 10 ++++++++++
> 1 file changed, 10 insertions(+)
>
> diff --git a/drivers/gpio/gpio-ath79.c b/drivers/gpio/gpio-ath79.c
> index afb535e..6b15792 100644
> --- a/drivers/gpio/gpio-ath79.c
> +++ b/drivers/gpio/gpio-ath79.c
> @@ -46,6 +46,7 @@ static int ath79_gpio_probe(struct platform_device *pdev)
> ctrl = devm_kzalloc(&pdev->dev, sizeof(*ctrl), GFP_KERNEL);
> if (!ctrl)
> return -ENOMEM;
> + platform_set_drvdata(pdev, ctrl);
>
> if (np) {
> err = of_property_read_u32(np, "ngpios", &ath79_gpio_count);
> @@ -97,12 +98,21 @@ static int ath79_gpio_probe(struct platform_device *pdev)
> return 0;
> }
>
> +static int ath79_gpio_remove(struct platform_device *pdev)
> +{
> + struct ath79_gpio_ctrl *ctrl = platform_get_drvdata(pdev);

platform_get_drvdata will return a gpio_chip *. I agree the address
will be the same, but for correctness you should use the expected
type. Especially since you will not use a member of ath79_gpio_ctrl in
this function anyway.

> +
> + gpiochip_remove(&ctrl->gc);
> + return 0;
> +}

I suspect this removal pattern to be quite common, maybe we should
just export bgpio_pdev_remove() to allow other drivers to use it
instead of rewriting their own version?