Pinmux bindings proposal V2

[tony@atomide.com (Tony Lindgren)]

* Dong Aisheng <dongas86@gmail.com> [120203 09:42]:
> On Sat, Feb 4, 2012 at 1:32 AM, Tony Lindgren <tony@atomide.com> wrote:
> > * Dong Aisheng <dongas86@gmail.com> [120202 11:36]:
> >>
> >> Actually i think i'd rather do not use config property, then i could
> >> be more compact:
> >> (anyway it's another issue and is flexible to be controlled by #pinmux-cells)
> >> pinctrl_usdhc4: pinconfig-usdhc4 {
> >> ? ? ? ? /* 0: pin 1: group */
> >> ? ? ? ? mux-entity = <0>;
> >> ? ? ? ? func-name = "usdhc4func";
> >> ? ? ? ? grp-name = "usdhc4grp";
> >
> > The func-name and grp-name should be optional here.
> > This mux entry is already the group, and can be used as
> > the group name.
> For the case i discussed here, the mux entry is PIN.
> (the mux-entity value is 0).
> we introduce this value here for treating all pins is one group.
> When do map parsing, only one pinmux map will be created.
> So we need a grp-name.
> And we also need a func-name here for construct pinmux map.

Sounds like a similar setup I have, I do not need any func-name
or grp-name though.. The group name is already unique with
pinconfig-usdhc4 in your example?
 
> > And the function name can be generated
> > dynamically in most cases. I'm currently using np->full_name
> > of the driver claiming these pins as the function name.
>
> Why i did not use np->name as function name is because the np->name
> can be different
> while actually these nodes may represent the same function but just
> different pins, so the function name should be same.

I used a function name based on the group name initially
(pinconfig-usdhc4 in your example), then renamed it to the
np->full_name of the device requesting the mux when the
mux was found ;)

> >> ? ? ? ? mux =
> >> ? ? ? ? ? ? ? ? <MX6Q_SD4_CMD ?0 MX6Q_USDHC_PAD_CTRL>
> >> ? ? ? ? ? ? ? ? <MX6Q_SD4_CLK ?0 MX6Q_USDHC_PAD_CTRL>
> >> ? ? ? ? ? ? ? ? <MX6Q_SD4_DAT0 1 MX6Q_USDHC_PAD_CTRL>
> >> ? ? ? ? ? ? ? ? <MX6Q_SD4_DAT1 1 MX6Q_USDHC_PAD_CTRL>
> >> ? ? ? ? ? ? ? ? <MX6Q_SD4_DAT2 1 MX6Q_USDHC_PAD_CTRL>
> >> ? ? ? ? ? ? ? ? <MX6Q_SD4_DAT3 1 MX6Q_USDHC_PAD_CTRL>
> >> ? ? ? ? ? ? ? ? <MX6Q_SD4_DAT4 1 MX6Q_USDHC_PAD_CTRL>
> >> ? ? ? ? ? ? ? ? <MX6Q_SD4_DAT5 1 MX6Q_USDHC_PAD_CTRL>
> >> ? ? ? ? ? ? ? ? <MX6Q_SD4_DAT6 1 MX6Q_USDHC_PAD_CTRL>
> >> ? ? ? ? ? ? ? ? <MX6Q_SD4_DAT7 1 MX6Q_USDHC_PAD_CTRL>;
> >> };
> >
> > For listing basic pins this format works fine for me. It seems
> > to have low overhead for parsing. And the width of the array
> > can be driver specific.
> >
> > Looks like it's the binding for altenative states that's still a
> > bit open..
> >
> Yes, it does not have states support.
> 
> > So how about let's first standardize on the mux format above?
> >
> I'm afraid it may be hard for us to standardize the mux format for a
> standard binding in pinctrl core due to  hw difference.

Yes the width would have to be hardare specific for the array.

> I'm think the new way which i sent in this thread after the mail you replied.
> You can refer to them to see if it's reasonable for you too.

Hmm, sorry now I'm confused. Got a link for that mail as so
many have been posted?

Regards,

Tony