Re: [PATCH v3 02/15] Driver core: Unified device properties interface for platform firmware

[Arnd Bergmann <arnd@arndb.de>]

On Thursday 02 October 2014 00:09:44 Rafael J. Wysocki wrote:
> On Wednesday, October 01, 2014 09:47:40 AM Arnd Bergmann wrote:
> > On Wednesday 01 October 2014 04:10:03 Rafael J. Wysocki wrote:
> > I still have my reservations against the child accessors, and would
> > like to hear what other people think. Passing a void pointer rather
> > than struct fw_dev_node has both advantages and disadvantages, and
> > I won't complain about either one if enough other people on the DT
> > side would like to see the addition of the child functions.
> 
> I actually would rather like to know if the people on the DT side have any
> problems with the child functions.

Sure, any kind of feedback would be helpful really. 

> Because, suppose that they wouldn't like those functions at all.  What are we
> supposed to do, then, honestly?  Add the whole DT vs ACPI logic to the leds-gpio
> and gpio_keys_polled drivers?  But these drivers have no reason whatsoever
> to include that.  Zero.
> 
> So suggestions welcome.
> 
> [BTW, In principle we also could use something like
> 
> typedef union dev_node {
> 	struct acpi_device *acpi_node;
> 	struct device_node *of_node;
> } dev_node_t;
> 
> instead of the (void *) for more type safety.  It still is useful to pass the
> parent pointer along with that, though.]

Yes, I'm not worried about the implementation details.

> > > Finally, device_for_each_child_node() is added for iterating over
> > > the children of the device description object associated with a
> > > given device.
> > > 
> > > The interface covers both ACPI and Device Trees.
> > > 
> > > This change set includes material from Mika Westerberg and Aaron Lu.
> > > 
> > 
> > Regarding device_for_each_child_node(), the syntax is inconsistent
> > with what we normally use, which can probably be changed. All of the
> > DT for_each_* helpers are macros that are used like
> > 
> > 	struct device *dev = ...;
> > 	void *child; /* iterator */
> > 
> > 	device_for_each_child_node(dev, child) {
> > 		u32 something;
> > 		device_child_property_read_u32(dev, child, "propname", &something);
> > 
> > 		do_something(dev, something);
> > 	}
> > 
> > If we get a consensus on having the child interfaces, I'd rather see
> > them done this way than with a callback pointer, for consistency
> > reasons.
> 
> That certainly is doable, although the resulting macro would generate a rather
> large chunk of code each time it is used.


#define device_for_each_child_node(dev, child) \
	for (child = device_get_next_child_node(dev, NULL), child, \
	     child = device_get_next_child_node(dev, child))

void *device_get_next_child_node(struct device *dev, void *child)
{
	if (IS_ENABLED(CONFIG_OF) && dev->of_node)
		return of_get_next_child(dev->of_node, child);
	else if (IS_ENABLED(CONFIG_ACPI) && ...)
		return acpi_get_next_child(dev, child);
	return NULL;
}

Not any more code than what we have today for the DT-only case, and it's
really just a function call in a loop.

	Arnd