Re: [PATCH 5/8] pinctrl-tz1090: add TZ1090 pinctrl driver

[James Hogan]

Hi Linus,

On 03/05/13 10:13, Linus Walleij wrote:
> On Fri, Apr 26, 2013 at 1:54 PM, James Hogan <james.hogan@xxxxxxxxxx> wrote:
>> On 25/04/13 23:39, Linus Walleij wrote:
>>> On Tue, Apr 23, 2013 at 4:33 PM, James Hogan <james.hogan@xxxxxxxxxx> wrote:
>>>> +static const struct cfg_param {
>>>> +       const char *property;
>>>> +       enum tz1090_pinconf_param param;
>>>> +} cfg_params[] = {
>>>> +       {"select",              TZ1090_PINCONF_PARAM_SELECT},
>>>> +       {"pull",                TZ1090_PINCONF_PARAM_PULL},
>>>> +       {"schmitt",             TZ1090_PINCONF_PARAM_SCHMITT},
>>>> +       {"slew-rate",           TZ1090_PINCONF_PARAM_SLEW_RATE},
>>>> +       {"drive-strength",      TZ1090_PINCONF_PARAM_DRIVE_STRENGTH},
>>>> +};
>>>
>>> Almost all exist in <linux/pinctrl/pinconf-generic.h>.
>>>
>>> What is "select"? If you need another config parameter
>>> we can just add it, but explain what it is and we can tell
>>> if it fits.
>>
>> select refers to the registers CR_PADS_GPIO_SELECT{0,1,2} with
>> descriptions like this:
>> Reset values: 0x3fffffc0, 0x3fffffff, 0x3fffffff
>> 29:0    CR_PADS_GPIO_SEL0       GPIO select (1 bit per GPIO pin)
>>                                 0 = serial interface
>>                                 1 = GPIO interface
>>                                 [29] SCB1_SCLK
>>                                 (etc)
>>
>> PARAM_GPIO may be a better name (select seems to have just stuck from
>> when the original gpio driver was written 3 years ago), although it
>> should be noted that the gpio system still has to enable it too, so it's
>> really about taking it out of the "serial interface" so that the
>> connected SoC peripheral cannot mess with it (1) by default (2) if it's
>> not connected to what the peripheral would expect, e.g. controlling
>> board power!
> 
> The GPIO select should not be visible to the outside like this,
> as it is a particular bit that should only be set on request from the
> GPIO framework.
> 
> If what you need is to set the pin into "GPIO mode" to drive it
> to some default state then from pinconf-generic.h you should use
> one of the existing defines like PIN_CONFIG_OUTPUT
> to actively drive it to high or low as default, or
> PIN_CONFIG_BIAS_HIGH_IMPEDANCE for some default
> GPIO input mode.
> 
> Read the new section named "GPIO mode pitfalls" in
> Documentation/pinctrl.txt

Thanks, that was interesting. I've had a think about this (and done some
experiments with a multimeter), and the problem is these generic
pinconfs already have meanings which don't match what the SELECT
register does. For example, having a pin be tristate and not controlled
by the peripheral, and having it tristate as far as the gpio hardware is
concerned (e.g. no pull-up) but still controlled by the peripheral, are
two very different things that need disambiguation.

I think what it comes down to as far as pinctrl is concerned is that the
SELECT registers enable/disable peripheral muxes on a per-pin basis.
Therefore perhaps it makes best sense to just have a custom/generic
pinconf PIN_CONFIG_PERIPHERAL_ENABLE/"peripheral=<1>;" to enable
peripheral muxing in the first place (sets SELECT=0), and require it to
be set on all individual pins that need it (i.e. don't automatically set
SELECT=0 on all pins in a group when the mux is enabled). What do you think?

Thanks
James

--
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/