RE: [PATCH v6] pinctrl: add a pin config interface

[Stephen Warren <swarren@nvidia.com>]

Haojian Zhuang wrote at Tuesday, December 13, 2011 12:22 AM:
> On Tue, Dec 13, 2011 at 8:22 AM, Linus Walleij <linus.walleij@linaro.org> wrote:
> > On Mon, Dec 12, 2011 at 7:19 AM, Haojian Zhuang
> > <haojian.zhuang@gmail.com> wrote:
> >
> >> You mentioned that pin_config_set() is used in below.
> >> ret = pin_config_set(dev, "FOO_GPIO_PIN", PLATFORM_X_PULL_UP);
> >>
> >> struct pinctrl_dev is created while pinmux is registered. I think this
> >> structure is always internal structure. It can't be observed by
> >> platform driver or device driver.
> >
> > True. Was fixed today by a patch from Stephen Warren that
> > simply cuts out the middleman and use the original platform
> > device directly. Kudos to Stephen for fixing this!
> >
> > Linus Walleij
> 
> Excuse me that I didn't find that patch. Did you merge it into your
> git tree? The latest patch on pinconf.c is updating pdev->dev to
> &pdev->dev.

I just hit the same problem. This isn't what my patch addresses.

My patch prevents the pinctrl core from creating a struct device for the
pin controller, since it already has one.

The problem that Haojian mentions is regarding struct pinctrl_dev, not
plain device.

I'd suggest modifying all the pin_config_* APIs to take a device name
rather than a "struct pinctrl_dev *". I'll work on a patch to do this,
since I'm hitting the same problem.

An alternative may be to either:

* Add function pinctrl_get_dev_by_name(name)

Or:

* Use dev = bus_find_device_by_name(name) to get the plain device, and
add a pinctrl_find_dev_by_dev(dev).

Either of those sound more complex though, but I suppose do allow
direct operation if you somehow do already have the struct pinctrl_dev.

-- 
nvpublic