[PATCH] RFC: name the Raspberry Pi GPIO lines

[Linus Walleij]

On Thu, Oct 6, 2016 at 8:26 PM, Eric Anholt <eric at anholt.net> wrote:
> Linus Walleij <linus.walleij at linaro.org> writes:

> Nice.  I really like this idea.  I'll spend some time reviewing these
> against the schematics I have and send some updates.

Please just squash in your changes and apply as one patch in the
end. I don't care about credits for the patch, the important part is that
it happens.

>         line 45: "[PWM1_OUT]" unused
>         line 46: "HDMI_HPD_P" unused
>         line 47: "SD_CARD_DET" "?" [kernel output]
>         line 48: "[SD_CLK_R]" unused
>         line 49: "[SD_CMD_R]" unused
>         line 50: "[SD_DATA0_R]" unused
>         line 51: "[SD_DATA1_R]" unused
>         line 52: "[SD_DATA2_R]" unused
>         line 53: "[SD_DATA3_R]" unused
>
> So we've got nice names now, but everything is showing up as unused.

Not this one:

>         line 47: "SD_CARD_DET" "?" [kernel output]

It is used as a GPIO by the kernel driver for SD card detection.

> I know the HDMI HPD has been grabbed by the DRM, and my SD lines are
> obviously in use by my SD card.  What does "unused" mean here?  Are we
> doing something wrong in our platform?

>From a GPIO point of view the lines used by other peripherals are
"unused" as they are not used for GPIO. I don't have a mechanism
for letting the pin control subsystem tell the GPIO subsystem that it
can't use a certain pin, but I guess it is possible. However it would
require the pin controller to set the .strict flag in struct pinmux_ops
and the BCM2835 currently does not, i.e. it says it is OK to use
pins for GPIO and other devices at the same time.

> The thing I want most when trying to debug gpio/pinctrl issues is
> something that shows me what the current pinmux is and what the state of
> the line is.  /debug/pinctrl/3f200000.gpio/pins is promising, but it
> doesn't get the new names:
>
> ...
> pin 44 (gpio44) function alt0 in lo; irq 138 (none)
> pin 45 (gpio45) function alt0 in lo; irq 139 (none)
> pin 46 (gpio46) function gpio_in in lo; irq 140 (none)
> pin 47 (gpio47) function gpio_out in lo; irq 141 (none)
> pin 48 (gpio48) function alt3 in hi; irq 142 (none)

Don't you have
/debug/pinctrl/3f200000.gpio/pinmux-pins
?

Typically this shows the device, function and group using the pins.

> /debug/gpio is also promising, but it doesn't have the mux or the state
> of HDMI HPD:
>
> ...
>  gpio-503 ([PWM1_OUT]          )
>  gpio-504 (HDMI_HPD_P          )
>  gpio-505 (SD_CARD_DET         |?                   ) out lo
>  gpio-506 ([SD_CLK_R]          )
>  gpio-507 ([SD_CMD_R]          )


I guess the HDMI is a device, not GPIO, so it should be named
"[HDMI_HPD_P]" with []. As it is not used by the GPIO subsystem
GPIO doesn't know this. Currently. Ideas for cross-talk between
GPIO and pin control for this are welcome.

> What I'm looking for is something where I can ask a user "please plug
> and unplug your HDMI monitor while 'watch grep HDMI
> /debug/gpio/whatever' is running" and see what happens.

Hm... Not familiar with the use case.

If you want to monitor the value of the line and it is possible to
use the lines as GPIO *at the same time* as it is used by the
HDMI (with or without hacks) this is achieveable. There are platforms
that do this, but I don't know about the BCM2835 pin control and
GPIO hardware.

Yours,
Linus Walleij