Re: [RFC PATCH 13/17] sunxi: add early Allwinner R528/T113 SoC support

[Andre Przywara <andre.przywara@arm.com>]

On Mon, 15 May 2023 20:52:23 -0600
Sam Edwards <cfsworks@gmail.com> wrote:

Hi Sam,

> On 12/5/22 17:45, Andre Przywara wrote:
> > diff --git a/arch/arm/mach-sunxi/board.c b/arch/arm/mach-sunxi/board.c
> > index 3763ec3d2e4..33331cda5e2 100644
> > --- a/arch/arm/mach-sunxi/board.c
> > +++ b/arch/arm/mach-sunxi/board.c
> > @@ -148,6 +148,10 @@ static int gpio_init(void)
> >   	sunxi_gpio_set_cfgpin(SUNXI_GPH(12), SUN9I_GPH_UART0);
> >   	sunxi_gpio_set_cfgpin(SUNXI_GPH(13), SUN9I_GPH_UART0);
> >   	sunxi_gpio_set_pull(SUNXI_GPH(13), SUNXI_GPIO_PULL_UP);
> > +#elif CONFIG_CONS_INDEX == 1 && defined(CONFIG_MACH_SUN8I_R528)
> > +	sunxi_gpio_set_cfgpin(SUNXI_GPE(2), 6);
> > +	sunxi_gpio_set_cfgpin(SUNXI_GPE(3), 6);
> > +	sunxi_gpio_set_pull(SUNXI_GPE(3), SUNXI_GPIO_PULL_UP);
> >   #elif CONFIG_CONS_INDEX == 2 && defined(CONFIG_MACH_SUNIV)
> >   	sunxi_gpio_set_cfgpin(SUNXI_GPA(2), SUNIV_GPE_UART0);
> >   	sunxi_gpio_set_cfgpin(SUNXI_GPA(3), SUNIV_GPE_UART0);  
> 
> I'm a relative outsider to sunxi, but it does look like the NCAT2 
> generation's MUX is a lot more flexible in its UART assignments: e.g. 
> while H3's UART1-3 only had one set of pins each UART could use, the 
> T113 (my target) supports anywhere from 2-6 (typically 4) different 
> options for each of its UARTs. It is not as simple as configuring which 
> UART is the serial console anymore, because different boards might wire 
> the console header to different pins. My particular target uses UART3 on 
> pins PB6/PB7, for example.

Well, theoretically there are indeed quite some options, but in reality
we find that most boards follow some reference design, and the *debug*
UART is only ever using a very few of the possible settings. For the
later (ARMv8) SoCs this is mostly UART0, but for the T113 it's
apparently indeed UART3. I also changed that in the Github version, to
use the very same PB6/7 UART3 mux as you.

> Since this patchset is updating the PIO driver anyway, maybe it's worth 
> updating the (preprocessor, not runtime) logic in gpio_init to support 
> getting the UART TX/RX pins from Kconfig, validating that the 
> CONFIG_CONS_INDEX and pin selection are compatible, and setting the 
> correct MUX function for those pins with a value pulled from a table?

Mmmh, that's an idea one could follow, as this function is indeed
insane. I actually rewrote gpio_init() already, to replace the hard to
read #ifdef's with actual C if statements, though that's still
following the same idea as the current code.
Let me have a look how this would look like.

Cheers,
Andre