Re: Fwd: PXA27X and PINCTRL (gpio-pxa.c)

[Greg <greguu@xxxxxxxx>]

On Mon, 5 Nov 2018 at 10:31, Robert Jarzmik <robert.jarzmik@xxxxxxx> wrote:
>
> Greg <greguu@xxxxxxxx> writes:
>
> > Hi Robert,
> >
> > please be advised of the following issue I raised with Linus W. in regards to a
> > minor issue with GPIO and PINCTRL on PXA27x.
> > He advised to get in touch with you about it. Before I post this to wider
> > linux-gpio@xxxxxxxxxxxxxxx or LKML I thought I better ask you to please provide
> > your feedback on this issue and if we can add:
> >
> > case PXA27X_GPIO:
> > return false;
> >
> > as shown below to work around the issue on PXA27X until proper DTS support is
> > implemented.
> Well no, this is not the proper solution.
>
> In the Sharp's case, as most pxa27x platform_data pxa boards relying on MFP, the
> proper solution is to never select CONFIG_PINCTRL. A hybrid pinctrl +
> platform_data is still in pondering in my mind, even if so far I split the world
> into :
>  - platform_data => CONFIG_PINCTRL=n
>  - MACH_PXA27X_DT=y => CONFIG_PINCTRL=y (as stated in arch/arm/mach-pxa/Kconfig)
>
> This is why no board config (arch/arm/mach-pxa/Kconfig) selects PINCTRL.
>
I forgot to mention that the issue occurs when platform PINCTRL is
disabled in .config.

> >> Please let me know what would be the ideal solution. Eg, to implement DT
> >> and PINCTRL proper for this platform ?
> It is already done.
>
> > That is always the ideal solution. AFAIK there is some basic PXA
> > device tree support code but I have no idea how far it is advanced.
> > PXA is one of those platforms that have so very many users so
> > I do think we should put in effort to modernize it if we can.
> It is up to the point where a whole pxa27x platform (mioa701) was already
> ported.
>
> >> PXA3XX was exempt for a reason too I assume ?
> Yes, that's because it uses for device-tree pinctrl-single, and not a specific
> pinctrl driver in drivers/pinctrl/pxa/*.
>
As far as I know we do not have a .dts for spitz/borzoi or akita yet.
The workaround (not solution) was to add the exemption in gpio-pxa.c

static bool pxa_gpio_has_pinctrl(void)
{
switch (gpio_type) {
    case PXA27X_GPIO:
return false;

> > It is also possible to have pin control without device tree, but
> > that seems like an unnecessary stepping stone for PXA, though
> > I do not know the details for.
> That's not scheduled for the pxa platform.
I assume a device tree implementation for our sub platform (spitz) is
the future proof and desired way forward ?

Cheers,

Greg