Re: [PATCH 1/3] Adding Skyworks SKY81452 MFD driver

[Lee Jones]

On Wed, 27 Aug 2014, Gyungoh Yoo wrote:
> On Tue, Aug 26, 2014 at 09:22:58AM +0100, Lee Jones wrote:
> > On Mon, 25 Aug 2014, Gyungoh Yoo wrote:
> > > On Thu, Aug 21, 2014 at 10:45:02AM +0100, Lee Jones wrote:
> > > > When you send patch-sets, you should send them connected to one
> > > > another AKA threaded.  That way, when we're reviewing we can look at
> > > > the other patches in the set for reference.  See the man page for `git
> > > > send-email` for details.
> > > > 
> > > > <commit log>
> > > > 
> > > > > Signed-off-by: Gyungoh Yoo <jack.yoo@xxxxxxxxxxxxxxx>
> > > > > ---
> > 
> > [...]
> > 
> > > > > +static int sky81452_register_devices(struct device *dev,
> > > > > +		const struct sky81452_platform_data *pdata)
> > > > > +{
> > > > > +	struct mfd_cell cells[] = {
> > > > > +		{
> > > > > +			.name = "sky81452-bl",
> > > > > +			.platform_data = pdata->bl_pdata,
> > > > > +			.pdata_size = sizeof(*pdata->bl_pdata),
> > > > 
> > > > Have you tested this with DT?
> > > > 
> > > > You're not passing the compatible string and not using
> > > > of_platform_populate() so I'm struggling to see how it would work
> > > > properly.
> > > 
> > > sky81452-bl and regulator-sky81452 is parsing the information
> > > in regulator node of its parent node. So I thought these 2 drivers
> > > don't need compatible attribute. That is what it didn't have
> > > compatible string.
> > > Is is mandatory that all drivers should have compatible attribute?
> > 
> > How do they obtain their DT nodes?
> 
> The backlight driver which is one of the child driver is obtain its DT node like this
> 
> np = of_get_child_by_name(dev->parent->of_node, "backlight");

The MFD core provides infrastructure so you don't have to do this.

Just place the compatible string in 'struct mfd_cell cells[]' and the
core will match and populate dev->of_node for you.

> > [...]
> > 
> > > > > +	return mfd_add_devices(dev, -1, cells, ARRAY_SIZE(cells),
> > > > > +			NULL, 0, NULL);
> > > > 
> > > > This doesn't really need to be in a function of its own.  Please put
> > > > it in .probe().  Also check for the return value and present the user
> > > > with an error message if it fails.
> > > 
> > > I think this need to be, in case of !CONFIG_OF.
> > > Can you please explain more in details?
> > 
> > Then how to you obtain the shared register map you created?
> 
> regmap is stored in driver data in MFD.
> 
> i2c_set_clientdata(client, regmap);
> 
> The child drivers obain the regmap from the parent.
> 
> struct regmap *regmap = dev_get_drvdata(dev->parent);

Ah yes, of course you do.  Silly of me to miss this.

I also just noticed that you're also manually populating the
chlidren's platform data.  It's easier if you do this from the child
device drivers:

  const struct sky81452_platform_data ppdata = dev_get_platdata(dev->parent);
  const struct sky81452_bl_platform_data = pdata = ppdata->bl_pdata;

-- 
Lee Jones
Linaro STMicroelectronics Landing Team Lead
Linaro.org â Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/