Re: [PATCH 1/3] net: rfkill-gpio: add device tree support

[Marc Dietrich <marvin24-Mmb7MZpHnFY@public.gmane.org>]

Am 13.02.2012 20:36, schrieb Olof Johansson:
> On Mon, Feb 13, 2012 at 11:25 AM, Rhyland Klein<rklein-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>  wrote:
>> On Sun, 2012-02-12 at 11:13 -0800, Marc Dietrich wrote:
>>> This adds device tree support for rfkill-gpio. The optional platform
>>> paramters gpio_runtime_close and gpio_runtime_setup are not implemented.
>>>
>>> Cc: linux-wireless-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
>>> Cc: "John W. Linville"<linville-2XuSBdqkA4R54TAoqtyWWQ@public.gmane.org>
>>> Cc: Johannes Berg<johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org>
>>> Cc: Rhyland Klein<rklein-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>
>>> Signed-off-by: Marc Dietrich<marvin24-Mmb7MZpHnFY@public.gmane.org>
>>> +
>>>   static int rfkill_gpio_probe(struct platform_device *pdev)
>>>   {
>>>        struct rfkill_gpio_data *rfkill;
>>>        struct rfkill_gpio_platform_data *pdata = pdev->dev.platform_data;
>>> +     struct device_node *np = pdev->dev.of_node;
>>>        int ret = 0;
>>>        int len = 0;
>>>
>>> +     if (np)
>>> +             pdata = rfkill_gpio_parse_pdata(pdev);
>>> +
>> The only concern I have is the precedence of devicetree settings vs
>> platform data settings? If there is pdata passed in from board file
>> initialization, and there is a device tree (a corner case but I think a
>> valid one) then I believe the order would be that defined pdata would
>> override the devicetree settings. That way if someone wanted to make a
>> quick update, they wouldn't need to change the boot loader as well.
> Yes, that is how other drivers tend to be coded -- only of pdata is
> null will the driver try to fill in from the DT.

If the device tree from the bootloader does not contain the rfkill 
information the resources from platform_data will be used (so no need to 
update the bootloader). If the bootloader contains rfkill information, 
why shouldn't it be used?

Marc