Re: [RFC 10/11] iio: Add OF support

On Fri, Feb 01, 2013 at 03:59:17PM +0100, Lars-Peter Clausen wrote:

On 02/01/2013 03:33 PM, Guenter Roeck wrote:

quoted

On Fri, Feb 01, 2013 at 12:58:02PM +0100, Lars-Peter Clausen wrote:

quoted

013 10:43 PM, Guenter Roeck wrote:

quoted

Provide bindings, new API access functions, and parse OF data
during initialization.

Hi Guenter,

Thanks for taking care of this.

I'd prefer to have only one iio_get_channel which handles both the dt and the
non-dt case. Otherwise we'll soon have constructs like

if (dev->of_node)
	chan = of_iio_get_channel(dev->of_node, 1);
else
	chan = iio_get_channel(dev_name(dev), "vcc");

Clock code has of_clk_get_by_name(node *, name), clk_get(dev, name), and
of_clk_get(node *, index). Right now of_iio_get_channel matches of_clk_get,
and iio_get_channel matches clk_get.

Question is really how we want to API to look like. I am open to suggestions.

quoted

quoted

appearing all over the place in drivers code. I'd keep the actual
implementation pretty close to what the clk framework does. E.g. take a look at
of_clk_get_by_name() and clk_get() in clkdev.c. And don't take a detour via the
iio_map lookup for the devicetree case, just loop over all IIO devices and
match by of_node. This should allow us to simplify the code quite a bit.

Yes, I just could not figure out how to loop over the list of iio devices. That
is why I hijacked iio_map which does provide a list.

Any hints ?

Yes, use bus_find_device on iio_bus_type. A nice example how to use this to
lookup device by of node is of_find_i2c_device_by_node. For IIO you also need
to make sure that dev->type is iio_dev_type, since both devices and triggers
are registered on the same bus.

Great, that works nicely, and, yes, the code is now much simpler.

quoted

quoted

quoted

+static struct iio_map *iio_map_of_init(struct iio_dev *idev)
+{

To be honest I find this whole function a bit confusing. If we are using
devicetree, we shouldn't need to register iio_map. Maybe just skip over
io-channel-output-names for now, until we figure out if we really need this and
what for.

Correct, I don't need this anymore after implementing the above.

Thanks,
Guenter