Question about MMC_PM_KEEP_POWER in MMCI driver

[Ulf Hansson]

On Wed, 25 Aug 2021 at 16:34, Yann Gautier <yann.gautier at foss.st.com> wrote:
>
> Hi,
>
> In drivers/mmc/host/mmci.c, MMC_PM_KEEP_POWER is unconditionally enabled.
> This prevents correct low-power sequence on STM32MP157C-DK2 board which
> embeds a Wifi chip brcm,bcm4329-fmac (this wifi part has not yet been
> sent upstream).

Exactly why doesn't this work with the STM32MP157C-DK2 board?

>
> This MMC_PM_KEEP_POWER can be taken from DT with the property
> keep-power-in-suspend. This is what is done for other MMC drivers.

The DT property is what should have been used for mmci as well.

>
> I wonder what should be the best solution for this.
>
> 1) Remove MMC_PM_KEEP_POWER from the driver, and modify all SoC device
> tree files embedding a arm,pl18x with adding keep-power-in-suspend;
> property (except stm32mp151.dtsi file).
> This can be easy to do (~10 files to modify). But that could be more
> board dependent, if an SDIO chip is plugged on this MMC IP.
> And the name keep-power-in-suspend can be misleading as it only applies
> to SDIO.
>
> 2) Remove MMC_PM_KEEP_POWER from the driver, and modify board DT files
> with the property. This could be a difficult task to find all those
> boards. And this should be applied only for SDIO configs.
>
> 3) Just modify the driver to apply this capability for all MMCI chips
> but STM32. This could be done in the dedicated file, in
> sdmmc_variant_init() function. But some boards based on STM32MP15 chip
> might want to keep this capability.

I would suggest option 3).

As a matter of fact, we also allow MMC_PM_KEEP_POWER to become set
when parsing the DTB via calling mmc_of_parse(). So just changing the
default value (don't set MMC_PM_KEEP_POWER) for the stm32 variant,
would do the trick I think.

Kind regards
Uffe