Re: [PATCH v2] mmc: mmci: manage MMC_PM_KEEP_POWER per variant config

[Philipp Zabel]

Hi Yann,

On Mo, 2022-03-14 at 13:55 +0100, Yann Gautier wrote:
> Add a disable_keep_power field in variant_data struct. The
> MMC_PM_KEEP_POWER flag will be enabled if disable_keep_power is not
> set.
> It is only set to true for stm32_sdmmc variants.
> 
> The issue was seen on STM32MP157C-DK2 board, which embeds a wifi
> chip.
> It doesn't correctly support low power on this board. The Wifi chip
> awaits an always-on regulator, but it was connected to v3v3 which is
> off
> in low-power sequence. MMC_PM_KEEP_POWER should then be disabled.
> 
> The flag can still be enabled through DT property:
> keep-power-in-suspend.
> 
> Signed-off-by: Yann Gautier <yann.gautier@xxxxxxxxxxx>
> ---
> Update in v2:
> Reword commit message to better explain the issue.
> 
> Resend the patch alone. It was previoulsy in a series [1] for which
> the
> other patches will be reworked.
> 
> [1] 
> https://lore.kernel.org/lkml/20220304135134.47827-1-yann.gautier@xxxxxxxxxxx/
> 
>  drivers/mmc/host/mmci.c | 5 ++++-
>  drivers/mmc/host/mmci.h | 1 +
>  2 files changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/mmc/host/mmci.c b/drivers/mmc/host/mmci.c
> index 45b8608c935c..0e2f2f5d6a52 100644
> --- a/drivers/mmc/host/mmci.c
> +++ b/drivers/mmc/host/mmci.c
> @@ -274,6 +274,7 @@ static struct variant_data variant_stm32_sdmmc =
> {
>         .busy_detect            = true,
>         .busy_detect_flag       = MCI_STM32_BUSYD0,
>         .busy_detect_mask       = MCI_STM32_BUSYD0ENDMASK,
> +       .disable_keep_power     = true,
>         .init                   = sdmmc_variant_init,
>  };
>  
> @@ -301,6 +302,7 @@ static struct variant_data variant_stm32_sdmmcv2
> = {
>         .busy_detect            = true,
>         .busy_detect_flag       = MCI_STM32_BUSYD0,
>         .busy_detect_mask       = MCI_STM32_BUSYD0ENDMASK,
> +       .disable_keep_power     = true,
>         .init                   = sdmmc_variant_init,
>  };
>  
> @@ -2172,7 +2174,8 @@ static int mmci_probe(struct amba_device *dev,
>         host->stop_abort.flags = MMC_RSP_R1B | MMC_CMD_AC;
>  
>         /* We support these PM capabilities. */
> -       mmc->pm_caps |= MMC_PM_KEEP_POWER;
> +       if (!variant->disable_keep_power)
> +               mmc->pm_caps |= MMC_PM_KEEP_POWER;
>  
>         /*
>          * We can do SGIO
> diff --git a/drivers/mmc/host/mmci.h b/drivers/mmc/host/mmci.h
> index e1a9b96a3396..2cad1ef9766a 100644
> --- a/drivers/mmc/host/mmci.h
> +++ b/drivers/mmc/host/mmci.h
> @@ -361,6 +361,7 @@ struct variant_data {
>         u32                     opendrain;
>         u8                      dma_lli:1;
>         u32                     stm32_idmabsize_mask;
> +       u8                      disable_keep_power:1;

There are already four separate bitfields in struct variant_data, why
not move this up into one of them?

regards
Philipp