On 11/23/2015 01:00 PM, Mark Brown wrote:

On Mon, Nov 23, 2015 at 11:40:55AM -0600, Andrew F. Davis wrote:

quoted

But which of_node?

quoted

regulator_config->of_node
regulator_config->dev->of_node

quoted

The second is the only one I see getting used, the first is only
used when drivers provide their own init_data and automatic init
data getting fails.

The configuration of_node is there to override the device one if there
were some reason to do it.  This should only happen in a situation where
we weren't able to use the core parsing, with modern drivers it
indicates a problematic binding so the code deliberately doesn't handle
it.  Anything with a problematic binding will have generated the
init_data in driver code anyway.  If we come up with a reason to extend
the interface we can do that but for now there is no need.

Right, so this is the kind of description that would be nice with the
declaration.

quoted

The same issue is present in GPIO (gpiolib.c:612), where the of_node

Line number references are complately unhelpful if you don't say what
you're looking at (for me that's a call to irq_find_mapping() which I'm
guessing isn't what you were talking about).

My bad, that is for v4.4-rc1 line 694. The lines are:

	of_node = gpiochip->dev->of_node;
#ifdef CONFIG_OF_GPIO
	/*
	 * If the gpiochip has an assigned OF node this takes precedence
	 * FIXME: get rid of this and use gpiochip->dev->of_node everywhere
	 */
	if (gpiochip->of_node)
		of_node = gpiochip->of_node;
#endif

So, if we have a config->of_node it is used over config->dev->of_node.

quoted

in the config takes precedence over the one in config->dev, the
opposite is true for regulators, this is very confusing and should be
standardized.

No, they both do the same thing.

I don't see that, config->dev->of_node is checked for the init data in
regulator_of_get_init_data, then config->of_node is then ignored if that
succeeds.