Re: [PATCH v3 1/3] platform: x86: int3472: Add MFD cell for tps68470 LED

[Lee Jones <lee@kernel.org>]

On Thu, 23 Mar 2023, Hans de Goede wrote:

> Hi,
>
> On 3/23/23 13:23, Lee Jones wrote:
> > On Tue, 21 Mar 2023, Kate Hsuan wrote:
> >
> >> Add MFD cell for tps68470-led.
> >>
> >> Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
> >> Signed-off-by: Kate Hsuan <hpa@redhat.com>
> >> ---
> >>  drivers/platform/x86/intel/int3472/tps68470.c | 5 +++--
> >>  1 file changed, 3 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/drivers/platform/x86/intel/int3472/tps68470.c b/drivers/platform/x86/intel/int3472/tps68470.c
> >> index 5b8d1a9620a5..82ef022f8916 100644
> >> --- a/drivers/platform/x86/intel/int3472/tps68470.c
> >> +++ b/drivers/platform/x86/intel/int3472/tps68470.c
> >> @@ -17,7 +17,7 @@
> >>  #define DESIGNED_FOR_CHROMEOS		1
> >>  #define DESIGNED_FOR_WINDOWS		2
> >>
> >> -#define TPS68470_WIN_MFD_CELL_COUNT	3
> >> +#define TPS68470_WIN_MFD_CELL_COUNT	4
> >>
> >>  static const struct mfd_cell tps68470_cros[] = {
> >>  	{ .name = "tps68470-gpio" },
> >> @@ -193,7 +193,8 @@ static int skl_int3472_tps68470_probe(struct i2c_client *client)
> >>  		cells[1].name = "tps68470-regulator";
> >>  		cells[1].platform_data = (void *)board_data->tps68470_regulator_pdata;
> >>  		cells[1].pdata_size = sizeof(struct tps68470_regulator_platform_data);
> >> -		cells[2].name = "tps68470-gpio";
> >> +		cells[2].name = "tps68470-led";
> >> +		cells[3].name = "tps68470-gpio";
> >
> > The question is, why is the MFD API being used out side of drivers/mfd?
>
> Because Intel made a big mess about how they describe camera sensors + the matching clks / regulators / GPIOs and the optional PMIC in ACPI.
>
> The drivers/platform/x86/intel/int3472/ code untangles this mess and in some cases it instantiates MFD cells (with a whole bunch of derived platform_data per cell) for a TPS68470 PMIC.
>
> And sometimes while binding to an INT3472 ACPI device-node it does not instantiate any MFD cells at all since the INT3472 ACPI device-node does not always describe such a PMIC.
>
> Oh and also depending on of the ACPI tables are targetting ChromeOS or Windows a different set of MFD cells needs to be instantiated. On ChromeOS most of the PMIC poking is done through ACPI through a ChomeOS specific custom ACPI OpRegion, so there there are only cells for GPIO and a driver providing the OpRegion are created.
>
> So lots of ugly x86 platform specific handling, ACPI parsing, etc. which is why this landed under drivers/platform/x86/ . IIRC you were even involved in the original merge since there once was a much simpler MFD driver under driver/mfd which only supported the ChromeOS setup.
>
> (but my memory may be deceiving me here).

Right, I guess we've both slept since then!

My normal request is that MFD handling should be in drivers/mfd.
Anything else can be farmed out to the various functional subsystems and
drivers/platform.

--
Lee Jones [李琼斯]