Re: [PATCH V12 10/23] mmc: sdhci-uhs2: add reset function and uhs2_mode function

From: Ulf Hansson
Date: Tue Oct 03 2023 - 06:30:58 EST


On Fri, 15 Sept 2023 at 11:44, Victor Shih <victorshihgli@xxxxxxxxx> wrote:
>
> From: Victor Shih <victor.shih@xxxxxxxxxxxxxxxxxxx>
>
> Sdhci_uhs2_reset() does a UHS-II specific reset operation.
>
> Signed-off-by: Ben Chuang <ben.chuang@xxxxxxxxxxxxxxxxxxx>
> Signed-off-by: AKASHI Takahiro <takahiro.akashi@xxxxxxxxxx>
> Signed-off-by: Victor Shih <victor.shih@xxxxxxxxxxxxxxxxxxx>
> Acked-by: Adrian Hunter <adrian.hunter@xxxxxxxxx>
> ---
>
> Updates in V8:
> - Adjust the position of matching brackets.
>
> Updates in V6:
> - Remove unnecessary functions and simplify code.
>
> ---
>
> drivers/mmc/host/sdhci-uhs2.c | 45 +++++++++++++++++++++++++++++++++++
> drivers/mmc/host/sdhci-uhs2.h | 2 ++
> 2 files changed, 47 insertions(+)
>
> diff --git a/drivers/mmc/host/sdhci-uhs2.c b/drivers/mmc/host/sdhci-uhs2.c
> index e339821d3504..dfc80a7f1bad 100644
> --- a/drivers/mmc/host/sdhci-uhs2.c
> +++ b/drivers/mmc/host/sdhci-uhs2.c
> @@ -10,7 +10,9 @@
> * Author: AKASHI Takahiro <takahiro.akashi@xxxxxxxxxx>
> */
>
> +#include <linux/delay.h>
> #include <linux/module.h>
> +#include <linux/iopoll.h>
>
> #include "sdhci.h"
> #include "sdhci-uhs2.h"
> @@ -49,6 +51,49 @@ void sdhci_uhs2_dump_regs(struct sdhci_host *host)
> }
> EXPORT_SYMBOL_GPL(sdhci_uhs2_dump_regs);
>
> +/*****************************************************************************\
> + * *
> + * Low level functions *
> + * *
> +\*****************************************************************************/
> +
> +bool sdhci_uhs2_mode(struct sdhci_host *host)
> +{
> + return host->mmc->flags & MMC_UHS2_SUPPORT;

The MMC_UHS2_SUPPORT bit looks redundant to me. Instead, I think we
should be using mmc->ios.timings, which already indicates whether we
are using UHS2 (MMC_TIMING_UHS2_SPEED_*). See patch2 where we added
this.

That said, I think we should drop the sdhci_uhs2_mode() function
altogether and instead use mmc_card_uhs2(), which means we should move
it to include/linux/mmc/host.h, so it becomes available for host
drivers.

> +}
> +
> +/**
> + * sdhci_uhs2_reset - invoke SW reset
> + * @host: SDHCI host
> + * @mask: Control mask
> + *
> + * Invoke SW reset, depending on a bit in @mask and wait for completion.
> + */
> +void sdhci_uhs2_reset(struct sdhci_host *host, u16 mask)
> +{
> + unsigned long timeout;
> + u32 val;
> +
> + sdhci_writew(host, mask, SDHCI_UHS2_SW_RESET);
> +
> + if (mask & SDHCI_UHS2_SW_RESET_FULL)
> + host->clock = 0;
> +
> + /* Wait max 100 ms */
> + timeout = 100000;

Please convert into using a define and use that directly, below instead.

> +
> + /* hw clears the bit when it's done */
> + if (read_poll_timeout_atomic(sdhci_readw, val, !(val & mask), 10,
> + timeout, true, host, SDHCI_UHS2_SW_RESET)) {
> + pr_err("%s: %s: Reset 0x%x never completed.\n", __func__,
> + mmc_hostname(host->mmc), (int)mask);
> + pr_err("%s: clean reset bit\n", mmc_hostname(host->mmc));

It looks silly to do two pr_err immediately after each other, please
combine them into one.

Moreover, I think we should probably convert into using a pr_warn() instead.

> + sdhci_writeb(host, 0, SDHCI_UHS2_SW_RESET);
> + return;
> + }
> +}
> +EXPORT_SYMBOL_GPL(sdhci_uhs2_reset);
> +
> /*****************************************************************************\
> * *
> * Driver init/exit *
> diff --git a/drivers/mmc/host/sdhci-uhs2.h b/drivers/mmc/host/sdhci-uhs2.h
> index 2bfe18d29bca..8253d50f7852 100644
> --- a/drivers/mmc/host/sdhci-uhs2.h
> +++ b/drivers/mmc/host/sdhci-uhs2.h
> @@ -177,5 +177,7 @@
> struct sdhci_host;
>
> void sdhci_uhs2_dump_regs(struct sdhci_host *host);
> +bool sdhci_uhs2_mode(struct sdhci_host *host);
> +void sdhci_uhs2_reset(struct sdhci_host *host, u16 mask);
>
> #endif /* __SDHCI_UHS2_H */
> --
> 2.25.1
>

Kind regards
Uffe