Re: [PATCH RESEND] media: rc: gpio-ir-recv: add support for device tree parsing

[Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>]

On 02/06/2013 02:48 PM, Sylwester Nawrocki wrote:
> On 02/06/2013 09:03 AM, Sebastian Hesselbarth wrote:
>> This patch adds device tree parsing for gpio_ir_recv platform_data and
>> the mandatory binding documentation. It basically follows what we already
>> have for e.g. gpio_keys. All required device tree properties are OS
>> independent but optional properties allow linux specific support for rc
>> protocols and maps.
>>
>> There was a similar patch sent by Matus Ujhelyi but that discussion
>> died after the first reviews.
>>
>> Signed-off-by: Sebastian Hesselbarth<sebastian.hesselbarth@gmail.com>
>> ---
> ...
>> diff --git a/Documentation/devicetree/bindings/media/gpio-ir-receiver.txt b/Documentation/devicetree/bindings/media/gpio-ir-receiver.txt
>> new file mode 100644
>> index 0000000..937760c
>> --- /dev/null
>> +++ b/Documentation/devicetree/bindings/media/gpio-ir-receiver.txt
>> @@ -0,0 +1,20 @@
>> +Device-Tree bindings for GPIO IR receiver
>> +
>> +Required properties:
>> +	- compatible = "gpio-ir-receiver";
>> +	- gpios: OF device-tree gpio specification.
>> +
>> +Optional properties:
>> +	- linux,allowed-rc-protocols: Linux specific u64 bitmask of allowed
>> +	    rc protocols.
>
> You likely need to specify in these bindings documentation which bit
> corresponds to which RC protocol.
>
> I'm not very familiar with the RC internals, but why it has to be
> specified statically in the device tree, when decoding seems to be
> mostly software defined ? I might be missing something though..

Sylwester,

I am not familiar with RC internals either. Maybe somebody with more
insight in media/rc can clarify the specific needs for the rc subsystem.
I was just transferring the DT support approach taken by gpio_keys to
gpio_ir_recv as I will be using it on mach-dove/cubox soon.

> Couldn't this be configured at run time, with all protocols allowed
> as the default ?

Actually, this is how the internal rc code works. If there is nothing
defined for allowed_protocols it assumes that all protocols are supported.
That is why above node properties are optional.

About the binding documentation of allowed_protocols, rc_map, or the
default behavior of current linux code, I don't think they will stay
in-sync for long. I'd rather completely remove those os-specific properties
from DT, but that hits the above statement about the needs of media/rc
subsystem.

>> +	- linux,rc-map-name: Linux specific remote control map name.
>> +
>> +Example node:
>> +
>> +	ir: ir-receiver {
>> +		compatible = "gpio-ir-receiver";
>> +		gpios =<&gpio0 19 1>;
>> +		/* allow rc protocols 1-4 */
>> +		linux,allowed-rc-protocols =<0x00000000 0x0000001e>;
>> +		linux,rc-map-name = "rc-rc6-mce";
>> +	};
>> diff --git a/drivers/media/rc/gpio-ir-recv.c b/drivers/media/rc/gpio-ir-recv.c
>> index 4f71a7d..25e09fa 100644
>> --- a/drivers/media/rc/gpio-ir-recv.c
>> +++ b/drivers/media/rc/gpio-ir-recv.c
>> @@ -16,6 +16,7 @@
>>   #include<linux/interrupt.h>
>>   #include<linux/gpio.h>
>>   #include<linux/slab.h>
>> +#include<linux/of_gpio.h>
>>   #include<linux/platform_device.h>
>>   #include<linux/irq.h>
>>   #include<media/rc-core.h>
>> @@ -30,6 +31,63 @@ struct gpio_rc_dev {
>>   	bool active_low;
>>   };
>>
>> +#ifdef CONFIG_OF
>> +/*
>> + * Translate OpenFirmware node properties into platform_data
>> + */
>> +static struct gpio_ir_recv_platform_data *
>> +gpio_ir_recv_get_devtree_pdata(struct device *dev)
>> +{
>> +	struct device_node *np = dev->of_node;
>> +	struct gpio_ir_recv_platform_data *pdata;
>> +	enum of_gpio_flags flags;
>> +	int gpio;
>> +
>> +	if (!np)
>> +		return ERR_PTR(-ENODEV);
>> +
>> +	pdata = devm_kzalloc(dev, sizeof(*pdata), GFP_KERNEL);
>> +	if (!pdata)
>> +		return ERR_PTR(-ENOMEM);
>> +
>> +	if (!of_find_property(np, "gpios", NULL)) {
>
> Why do you need this ? Isn't of_get_gpio_flags() sufficient ?

Ok. Now that you point at it, I agree that this check and the error
below is not needed. It is in gpio_keys, so that explains why it also
moved in here.

>> +		dev_err(dev, "Found gpio-ir-receiver without gpios\n");
>> +		return ERR_PTR(-EINVAL);
>> +	}
>> +
>> +	gpio = of_get_gpio_flags(np, 0,&flags);
>> +	if (gpio<  0) {
>> +		if (gpio != -EPROBE_DEFER)
>> +			dev_err(dev, "Failed to get gpio flags, error: %d\n",
>> +				gpio);
>> +		return ERR_PTR(gpio);
>> +	}
>> +
>> +	pdata->gpio_nr = gpio;
>> +	pdata->active_low = (flags&  OF_GPIO_ACTIVE_LOW) ? true : false;
>> +	pdata->map_name = of_get_property(np, "linux,rc-map-name", NULL);
>> +	of_property_read_u64(np, "linux,allowed-rc-protocols",
>> +			&pdata->allowed_protos);
>> +
>> +	return pdata;
>> +}
>> +
>> +static struct of_device_id gpio_ir_recv_of_match[] = {
>> +	{ .compatible = "gpio-ir-receiver", },
>> +	{ },
>> +};
>> +MODULE_DEVICE_TABLE(of, gpio_ir_recv_of_match);
>> +
>> +#else /* !CONFIG_OF */
>> +
>> +static inline struct gpio_ir_recv_platform_data *
>> +gpio_ir_recv_get_devtree_pdata(struct device *dev)
>> +{
>> +	return ERR_PTR(-ENODEV);
>> +}
>> +
>> +#endif
>> +
>>   static irqreturn_t gpio_ir_recv_irq(int irq, void *dev_id)
>>   {
>>   	struct gpio_rc_dev *gpio_dev = dev_id;
>> @@ -66,8 +124,11 @@ static int gpio_ir_recv_probe(struct platform_device *pdev)
>>   					pdev->dev.platform_data;
>>   	int rc;
>>
>> -	if (!pdata)
>> -		return -EINVAL;
>> +	if (!pdata) {
>> +		pdata = gpio_ir_recv_get_devtree_pdata(&pdev->dev);
>
> Could assigning to pdev->dev.platform_data be avoided here ?
>
> platform_data is only referenced in probe(), so maybe something like
> this would be better:
>
> 	const struct gpio_ir_recv_platform_data *pdata = NULL;
>
> 	if (pdev->dev.of_node) {
> 		ret = gpio_ir_recv_parse_dt(&pdev->dev,&pdata);
> 		if (ret<  0)
> 			return ret;
> 	} else {
> 		pdata = pdev->dev.platform_data;
> 	}
> 	if (!pdata)
> 		return -EINVAL;

Actually, I am not assigning the parsed gpio_ir_recv_platform_data to
pdev->dev.platform_data but pdata ptr instead. Either I don't see the
difference in pointer assignments between your code and mine or you
were mislead from struct gpio_ir_recv_platform_data above.

Anyway, I agree to test for pdev->dev.of_node and call gpio_ir_recv_parse_dt
if set.

>> +		if (IS_ERR(pdata))
>> +			return PTR_ERR(pdata);
>> +	}
>>
>>   	if (pdata->gpio_nr<  0)
>>   		return -EINVAL;
>> @@ -195,6 +256,9 @@ static struct platform_driver gpio_ir_recv_driver = {
>>   #ifdef CONFIG_PM
>>   		.pm	=&gpio_ir_recv_pm_ops,
>>   #endif
>> +#ifdef CONFIG_OF
>> +		.of_match_table = of_match_ptr(gpio_ir_recv_of_match),
>> +#endif
>
> There is not need for #ifdef here, of_match_ptr() macro was introduced
> just to allow to omit #ifdefs.

Ok, I will change that.

Thanks for the review!

Sebastian