[PATCH v2 1/2] pinctrl: meson: meson8b: fix requesting GPIOs greater than GPIOZ_3

[Linus Walleij]

On Sun, Feb 25, 2018 at 12:38 PM, Martin Blumenstingl
<martin.blumenstingl at googlemail.com> wrote:

> Meson8b is a cost reduced variant of the Meson8 SoC. It's package size
> is smaller than Meson8.
> Unfortunately there are a few key differences which cannot be seen
> without close inspection of the code and the public S805 datasheet:
> - the GPIOX bank is missing the GPIOX_12, GPIOX_13, GPIOX_14 and
>   GPIOX_15 GPIOs
> - the GPIOY bank is missing the GPIOY_2, GPIOY_4, GPIOY_5, GPIOY_15 and
>   GPIOY_16 GPIOs
> - the GPIODV bank is missing all GPIOs except GPIODV_9, GPIODV_24,
>   GPIODV_25, GPIODV_26, GPIODV_27, GPIODV_28 and GPIODV_29
> - the GPIOZ bank is missing completely
> - there is a new GPIO bank called "DIF"
>
> This means that Meson8b only has 83 actual GPIO lines. Without any holes
> there would be 130 GPIO lines in total (120 are inherited from Meson8
> plus 10 new from the DIF bank).
>
> GPIOs greater GPIOZ_3 (whose ID is 83 - as a reminder: this is exactly
> the number of actual GPIO lines on Meson8b and also the value of
> meson8b_cbus_pinctrl_data.num_pins) cannot berequested. Using CARD_6
> (which used ID 100 prior to this patch, "base of the GPIO controller was
> 382) as an example:
> $ echo 482 > /sys/class/gpio/export
> export_store: invalid GPIO 482
>
> This removes all non-existing pins from to dt-bindings header file
> (include/dt-bindings/gpio/meson8b-gpio.h). This allows us to have a
> consecutive numbering for the GPIO #defines (GPIOY_2 doesn't exist for
> example, so previously the GPIOY_3 ID was "GPIOY_1 + 2", after this
> patch it is "GPIOY_1 + 1"). As a nice side-effect this means that we get
> compile-time (instead of runtime) errors if Meson8b .dts uses a pin that
> only exists on Meson8.
>
> Additionally the pinctrl-meson8b driver has to be updated to handle this
> new GPIO numbering. By default a struct meson_bank only handles GPIO
> banks where the pins are numbered consecutively because it calculates
> the bit offsets based on the GPIO IDs.
> This is solved by  taking the original BANK() definition and splitting it
> into consecutive subsets (X0..11 and X16..21). The bit offsets for each
> new bank includes the skipped GPIOs (the definition of the "X0..11" bank
> is identical to the old "X" bank apart from the "last IRQ" field, the
> definition of the new, split "X16..21" bank takes the original "X" bank
> and adds 16 - the start of the new split bank - to the "first IRQ",
> pullen bit, pull bit, dir bit, out bit and in bit).
>
> Commit 984cffdeaeb7ea ("pinctrl: Fix gpio/pin mapping for Meson8b")
> fixed the same issue by setting "ngpio" (of the gpio_chip) to 130.
> Unfortunately this broke in db80f0e158e621 ("pinctrl: meson: get rid of
> unneeded domain structures").
> The solution from this patch was considered to be better than the
> previous attempt at fixing this because it provides compile-time error
> checking for the GPIOs that exist on Meson8 but don't exist on Meson8b.
>
> The following pins were tested on an Odroid-C1 using the sysfs GPIO
> interface checking that their value (high or low) could be read:
> - GPIOX_0, GPIOX_1, GPIOX_2, GPIOX_3, GPIOX_4, GPIOX_5, GPIOX_6,
>   GPIOX_7, GPIOX_8, GPIOX_9, GPIOX_10, GPIOX_11, GPIOX_18, GPIOX_19,
>   GPIOX_20, GPIOX_21
> - GPIOY_3, GPIOY_7, GPIOY_8
> (some of these had to be pulled up because they were low by default,
> others were high by default so these had to be pulled down)
>
> Reported-by: Linus Lüssing <linus.luessing at c0d3.blue>
> Suggested-by: Jerome Brunet <jbrunet at baylibre.com>
> Signed-off-by: Martin Blumenstingl <martin.blumenstingl at googlemail.com>

Patch applied with Jerome's review tag.

Do I need to queue this for v4.16 fixes or is v4.17 good
enough?

Should it even be tagged for stable?

Yours,
Linus Walleij