Re: [PATCH] leds: gpio: Fix uninitialized gpio label for fwnode based probe

[Jacek Anaszewski]

On 12/21/19 7:49 PM, Pavel Machek wrote:
> Hi!
> 
>> When switching to using generic LED name composition mechanism via
>> devm_led_classdev_register_ext() API the part of code initializing
>> struct gpio_led's template name property was removed alongside.
>> It was however overlooked that the property was also passed to
>> devm_fwnode_get_gpiod_from_child() in place of "label" parameter,
>> which when set to NULL, results in gpio label being initialized to '?'.
>>
>> It could be observed in debugfs and failed to properly identify
>> gpio association with LED consumer.
>>
>> Fix this shortcoming by updating the GPIO label after the LED is
>> registered and its final name is known.
>>
>> Fixes: d7235f5feaa0 ("leds: gpio: Use generic support for composing LED names")
>> Cc: Linus Walleij <linus.walleij@xxxxxxxxxx>
>> Cc: Pavel Machek <pavel@xxxxxx>
>> Cc: Russell King <linux@xxxxxxxxxxxxxxx>
>> Signed-off-by: Jacek Anaszewski <jacek.anaszewski@xxxxxxxxx>
> 
> Patch looks good, except:
> 
>> @@ -151,9 +151,14 @@ static struct gpio_leds_priv *gpio_leds_create(struct platform_device *pdev)
>>  		struct gpio_led led = {};
>>  		const char *state = NULL;
>>  
>> +		/**
>> +		 * Acquire gpiod from DT with uninitialized label, which
>> +		 * will be updated after LED class device is registered,
>> +		 * Only then the final LED name is known.
>> +		 */
>>  		led.gpiod = devm_fwnode_get_gpiod_from_child(dev, NULL, child,
>>  							     GPIOD_ASIS,
>> -							     led.name);
>> +							     NULL);
> 
> This is not linuxdoc, so comment should beging with /* AFAICT.

Right.

> I'll probably hand-edit the patch.

Sure, thanks.

-- 
Best regards,
Jacek Anaszewski