Hi Linus, Geert,
On 12/16/2018 12:04 PM, Geert Uytterhoeven wrote:
> Hi Linus,
>
> On Sun, Dec 16, 2018 at 1:19 AM Linus Walleij <linus.walleij@xxxxxxxxxx> wrote:
>> On Mon, Dec 10, 2018 at 3:22 PM Vladimir Zapolskiy
>> <vladimir_zapolskiy@xxxxxxxxxx> wrote:
>>> R-Car GPIO controller provides two interfaces to set GPIO line output
>>> signal state, and for a particular GPIO line the selected interface is
>>> determined by OUTDTSEL bit value.
>>>
>>> At the moment the driver supports only one of two interfaces, namely
>>> OUTDT General Output Register is used to control the output signal.
>>>
>>> While this selection is the default one on reset, it is not explicitly
>>> configured on probe, thus it might be possible that kernel and userspace
>>> consumers of a GPIO won't be able to set the wanted GPIO output signal.
>>>
>>> Below is a simple test case to reproduce the described problem and
>>> verify this fix in the kernel on H3 ULCB by setting non-default OUTDTSEL
>>> configuration from a bootloader:
>>>
>>> u-boot > mw.l 0xe6055440 0x3000 1
>>> ...
>>> userspace > echo -n default-on > /sys/devices/platform/leds/leds/led5/trigger
>>> userspace > echo -n default-on > /sys/devices/platform/leds/leds/led6/trigger
>>>
>>> Fixes: 119f5e448d32c ("gpio: Renesas R-Car GPIO driver V3")
>>> Signed-off-by: Vladimir Zapolskiy <vladimir_zapolskiy@xxxxxxxxxx>
>>
>> Patch applied for devel as "non-urgent fix" as I can't see any
>> immediate regressions.
>
> Unfortunately the driver now writes unconditionally to a register that does
> not exist on R-Car Gen1. Hence please revert the patch.
yes, for sure revert the change then.
I'll prepare v2 which does not touch R-Car Gen1 platforms.
--
Best wishes,
Vladimir
