[PATCH] media: rc: gpio-ir-recv.c: change platform_data to DT binding

Sylwester Nawrocki sylvester.nawrocki at gmail.com
Wed Jan 2 09:23:16 EST 2013


Hi,

Cc: devicetree-discuss at lists.ozlabs.org

Please Cc this list when introducing any device tree bindings.

And you need documentation for the bindings, so when someone wants
to use the driver they don't have to extract the bindings' semantics
from the code.

On 12/18/2012 02:41 PM, ujhelyi.m at gmail.com wrote:
> From: Matus Ujhelyi<ujhelyi.m at gmail.com>

Please provide proper commit text so it is more clear why the patch is
needed.

Also, your patch summary line is very misleading, you're not replacing
platform_data based driver's initialization with device tree based.
This patch is simply _adding_ another initialization method, without
breaking support for platform data, isn't it ?

> Signed-off-by: Matus Ujhelyi<ujhelyi.m at gmail.com>
> ---
>   drivers/media/rc/gpio-ir-recv.c |   84 ++++++++++++++++++++++++++++++---------
>   1 file changed, 66 insertions(+), 18 deletions(-)
>
> diff --git a/drivers/media/rc/gpio-ir-recv.c b/drivers/media/rc/gpio-ir-recv.c
> index ba1a1eb..cfba079 100644
> --- a/drivers/media/rc/gpio-ir-recv.c
> +++ b/drivers/media/rc/gpio-ir-recv.c
> @@ -58,19 +58,49 @@ err_get_value:
>   	return IRQ_HANDLED;
>   }
>
> +static const struct of_device_id gpio_ir_of_match[] = {
> +	{
> +		.compatible = "gpio-ir-recv",

Probably better to make it "gpio-ir-receiver".

> +	},
> +};
> +MODULE_DEVICE_TABLE(of, gpio_ir_of_matchof_match);

I think you didn't try to build this driver as a module, otherwise you 
would
have noticed the above typo...

> +static int __devinit gpio_ir_recv_dt_probe (struct gpio_rc_dev *gpio_dev,
> +		struct rc_dev *rcdev, struct platform_device *pdev) {
> +
> +	struct device_node *node = pdev->dev.of_node;
> +
> +	if (!node) {
> +		pr_err("Missing gpio_nr in the DT\n");

Really ? :)

> +		return -EINVAL;
> +	}
> +
> +	if (of_property_read_u32(node, "gpio_nr",&gpio_dev->gpio_nr)) {

If I'm not mistaken, generic "gpios" property would be more appropriate 
here...

> +		pr_err("Missing gpio_nr in the DT\n");
> +		return -EINVAL;
> +	}
> +
> +	if (of_property_read_u64(node, "allowed_protos",&rcdev->allowed_protos))
> +		rcdev->allowed_protos = RC_BIT_ALL;
> +
> +	if (of_property_read_bool(node, "active_low"))

...and this one would not be needed, since the GPIO's active state could be
passed through generic OF gpio flags, e.g.

	enum of_gpio_flags flags;

	gpio_dev->gpio_nr = of_get_gpio_flags(node, 0, &flags);
	gpio_dev->active_low = flags & OF_GPIO_ACTIVE_LOW;

> +		gpio_dev->active_low = true;
> +	else
> +		gpio_dev->active_low = false;
> +
> +	if (of_property_read_string(node, "map_name",&rcdev->map_name))
> +		rcdev->map_name = NULL;

IIRC underscores are discouraged in the property names. What map_name
signifies here, anyway ?

> +
> +	return 0;
> +}
> +
>   static int __devinit gpio_ir_recv_probe(struct platform_device *pdev)
>   {
>   	struct gpio_rc_dev *gpio_dev;
>   	struct rc_dev *rcdev;
>   	const struct gpio_ir_recv_platform_data *pdata =
>   					pdev->dev.platform_data;
> -	int rc;
> -
> -	if (!pdata)
> -		return -EINVAL;
> -
> -	if (pdata->gpio_nr<  0)
> -		return -EINVAL;
> +	int rc = -EINVAL;
>
>   	gpio_dev = kzalloc(sizeof(struct gpio_rc_dev), GFP_KERNEL);
>   	if (!gpio_dev)
> @@ -82,6 +112,28 @@ static int __devinit gpio_ir_recv_probe(struct platform_device *pdev)
>   		goto err_allocate_device;
>   	}
>
> +	if (!pdata) {
> +

nit: Unnecessary empty line.

> +		if (gpio_ir_recv_dt_probe(gpio_dev, rcdev, pdev))
> +			goto err_dt_create_of;
> +

Ditto.

> +	} else {
> +

Ditto.
> +		if (pdata->allowed_protos)
> +			rcdev->allowed_protos = pdata->allowed_protos;
> +		else
> +			rcdev->allowed_protos = RC_BIT_ALL;
> +		rcdev->map_name = pdata->map_name ?: RC_MAP_EMPTY;
> +
> +		gpio_dev->gpio_nr = pdata->gpio_nr;
> +		gpio_dev->active_low = pdata->active_low;
> +

Ditto.

> +	}
> +
> +	if (gpio_dev->gpio_nr<  0) {

if (gpio_is_valid(gpio_dev->gpio_nr))

> +		goto err_gpio_nr;
> +	}
> +
>   	rcdev->priv = gpio_dev;
>   	rcdev->driver_type = RC_DRIVER_IR_RAW;
>   	rcdev->input_name = GPIO_IR_DEVICE_NAME;
> @@ -92,20 +144,13 @@ static int __devinit gpio_ir_recv_probe(struct platform_device *pdev)
>   	rcdev->input_id.version = 0x0100;
>   	rcdev->dev.parent =&pdev->dev;
>   	rcdev->driver_name = GPIO_IR_DRIVER_NAME;
> -	if (pdata->allowed_protos)
> -		rcdev->allowed_protos = pdata->allowed_protos;
> -	else
> -		rcdev->allowed_protos = RC_BIT_ALL;
> -	rcdev->map_name = pdata->map_name ?: RC_MAP_EMPTY;
>
>   	gpio_dev->rcdev = rcdev;
> -	gpio_dev->gpio_nr = pdata->gpio_nr;
> -	gpio_dev->active_low = pdata->active_low;
>
> -	rc = gpio_request(pdata->gpio_nr, "gpio-ir-recv");
> +	rc = gpio_request(gpio_dev->gpio_nr, "gpio-ir-recv");
>   	if (rc<  0)
>   		goto err_gpio_request;
> -	rc  = gpio_direction_input(pdata->gpio_nr);
> +	rc  = gpio_direction_input(gpio_dev->gpio_nr);
>   	if (rc<  0)
>   		goto err_gpio_direction_input;
>
> @@ -117,7 +162,7 @@ static int __devinit gpio_ir_recv_probe(struct platform_device *pdev)
>
>   	platform_set_drvdata(pdev, gpio_dev);
>
> -	rc = request_any_context_irq(gpio_to_irq(pdata->gpio_nr),
> +	rc = request_any_context_irq(gpio_to_irq(gpio_dev->gpio_nr),
>   				gpio_ir_recv_irq,
>   			IRQF_TRIGGER_FALLING | IRQF_TRIGGER_RISING,
>   					"gpio-ir-recv-irq", gpio_dev);
> @@ -131,8 +176,10 @@ err_request_irq:
>   	rc_unregister_device(rcdev);
>   err_register_rc_device:
>   err_gpio_direction_input:
> -	gpio_free(pdata->gpio_nr);
> +	gpio_free(gpio_dev->gpio_nr);
>   err_gpio_request:
> +err_dt_create_of:
> +err_gpio_nr:
>   	rc_free_device(rcdev);
>   	rcdev = NULL;
>   err_allocate_device:
> @@ -195,6 +242,7 @@ static struct platform_driver gpio_ir_recv_driver = {
>   #ifdef CONFIG_PM
>   		.pm	=&gpio_ir_recv_pm_ops,
>   #endif
> +		.of_match_table = of_match_ptr(gpio_ir_of_match),
>   	},
>   };
>   module_platform_driver(gpio_ir_recv_driver);

Thanks,
Sylwester


More information about the devicetree-discuss mailing list