All of lore.kernel.org
 help / color / mirror / Atom feed
From: Ulf Hansson <ulf.hansson@linaro.org>
To: Ludovic Barre <ludovic.Barre@st.com>
Cc: Rob Herring <robh+dt@kernel.org>,
	Srinivas Kandagatla <srinivas.kandagatla@linaro.org>,
	Maxime Coquelin <mcoquelin.stm32@gmail.com>,
	Alexandre Torgue <alexandre.torgue@st.com>,
	Linux ARM <linux-arm-kernel@lists.infradead.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	DTML <devicetree@vger.kernel.org>,
	"linux-mmc@vger.kernel.org" <linux-mmc@vger.kernel.org>,
	linux-stm32@st-md-mailman.stormreply.com
Subject: Re: [PATCH V2 2/2] mmc: mmci: add variant property to send stop cmd if a command fail
Date: Wed, 21 Nov 2018 18:56:12 +0100	[thread overview]
Message-ID: <CAPDyKFpZyYJbDDjYaFPfGEfKme6QedxcD715M2v7kAf7tpUf5A@mail.gmail.com> (raw)
In-Reply-To: <1541583041-17461-3-git-send-email-ludovic.Barre@st.com>

On 7 November 2018 at 10:30, Ludovic Barre <ludovic.Barre@st.com> wrote:
> From: Ludovic Barre <ludovic.barre@st.com>
>
> The mmc framework follows the requirement of SD_Specification:
> the STOP_TRANSMISSION is sent on multiple write/read commands
> and the stop command (alone), not needed on other ADTC commands.
>
> But, if an error happens on command or data transmission, some
> variants require a stop command "STOP_TRANSMISION" to clear the DPSM
> "Data Path State Machine". If it's not done the next data
> command freezes hardware block.
> Needed to support the STM32 sdmmc variant.

May I suggest some re-wording of this changelog, as to make it more
clear of why this is needed. Something along the lines of:

"The current approach with sending a CMD12 (STOP_TRANSMISSION) to
complete a data transfer request, either because of using the open
ended transmission type or because of receiving an error during a data
transfer, isn't sufficient for the STM32 sdmmc variant.

More precisely, for STM32 sdmmc the DPSM ("Data Path State Machine" )
needs to be cleared by sending a CMD12, also for the so called ADTC
commands. For this reason, add a new mmci variant property and let the
driver send a CMD12 to complete ADTC commands, in case it's set."

>
> Signed-off-by: Ludovic Barre <ludovic.barre@st.com>
> ---
>  drivers/mmc/host/mmci.c | 33 +++++++++++++++++++++++++++++++++
>  drivers/mmc/host/mmci.h |  4 ++++
>  2 files changed, 37 insertions(+)
>
> diff --git a/drivers/mmc/host/mmci.c b/drivers/mmc/host/mmci.c
> index 13fa640..47b865d 100644
> --- a/drivers/mmc/host/mmci.c
> +++ b/drivers/mmc/host/mmci.c
> @@ -21,6 +21,7 @@
>  #include <linux/err.h>
>  #include <linux/highmem.h>
>  #include <linux/log2.h>
> +#include <linux/mmc/mmc.h>
>  #include <linux/mmc/pm.h>
>  #include <linux/mmc/host.h>
>  #include <linux/mmc/card.h>
> @@ -57,6 +58,8 @@ void sdmmc_variant_init(struct mmci_host *host);
>  #else
>  static inline void sdmmc_variant_init(struct mmci_host *host) {}
>  #endif
> +static void
> +mmci_start_command(struct mmci_host *host, struct mmc_command *cmd, u32 c);
>
>  static unsigned int fmax = 515633;
>
> @@ -274,6 +277,7 @@ static struct variant_data variant_stm32_sdmmc = {
>         .cmdreg_lrsp_crc        = MCI_CPSM_STM32_LRSP_CRC,
>         .cmdreg_srsp_crc        = MCI_CPSM_STM32_SRSP_CRC,
>         .cmdreg_srsp            = MCI_CPSM_STM32_SRSP,
> +       .cmdreg_stop            = MCI_CPSM_STM32_CMDSTOP,
>         .data_cmd_enable        = MCI_CPSM_STM32_CMDTRANS,
>         .irq_pio_mask           = MCI_IRQ_PIO_STM32_MASK,
>         .datactrl_first         = true,
> @@ -573,6 +577,24 @@ void mmci_dma_error(struct mmci_host *host)
>  static void
>  mmci_request_end(struct mmci_host *host, struct mmc_request *mrq)
>  {
> +       /*
> +        * If an error happens on command or data transmission, some variants
> +        * require a stop command to reinit the DPSM.
> +        * If it's not done the next data command freeze hardware block.
> +        */
> +       if (host->variant->cmdreg_stop) {
> +               u32 dpsm;
> +
> +               dpsm = readl_relaxed(host->base + MMCISTATUS);
> +               dpsm &= MCI_STM32_DPSMACTIVE;
> +
> +               if (dpsm && ((mrq->cmd && mrq->cmd->error) ||
> +                            (mrq->data && mrq->data->error))) {
> +                       mmci_start_command(host, &host->stop_abort, 0);
> +                       return;
> +               }

I would rather move this code to a separate function. Also, I think
you need something else (or additional) than polling the MMCISTATUS
register, as in principle you could end up sending CMD12 several times
for the same request, which isn't correct.

To me the best solution would probably be to make use of the
host->data pointer, as it becomes set when DPSM has been enabled.
However, host->data is also cleared in mmci_stop_data() which is being
called prior mmci_request_end(). In other words, we need to figure
under what conditions the new code above should be triggered/called
and then also change the conditions for when mmci_request_end() shall
be called.

In principle look at callers of mmci_request_end() and
mmci_stop_data() and update those paths.

> +       }
> +
>         writel(0, host->base + MMCICOMMAND);
>
>         BUG_ON(host->data);
> @@ -1100,6 +1122,10 @@ mmci_start_command(struct mmci_host *host, struct mmc_command *cmd, u32 c)
>                 mmci_reg_delay(host);
>         }
>
> +       if (host->variant->cmdreg_stop &&
> +           cmd->opcode == MMC_STOP_TRANSMISSION)
> +               c |= host->variant->cmdreg_stop;
> +

Hmm.

It looks like the above changes, together with the introduction of the
variant property, belongs in a separate patch, preceding $subject
patch in the series.

The reason is, that to me, it makes sense to add the special treatment
of the ADTC commands in a separate patch. Can you please split it up
and then of course update the change log accordingly!?

>         c |= cmd->opcode | host->variant->cmdreg_cpsm_enable;
>         if (cmd->flags & MMC_RSP_PRESENT) {
>                 if (cmd->flags & MMC_RSP_136)
> @@ -1950,6 +1976,13 @@ static int mmci_probe(struct amba_device *dev,
>                 mmc->max_busy_timeout = 0;
>         }
>
> +       /* prepare the stop command, used to abort and reinitialized the DPSM */
> +       if (variant->cmdreg_stop) {
> +               host->stop_abort.opcode = MMC_STOP_TRANSMISSION;
> +               host->stop_abort.arg = 0;
> +               host->stop_abort.flags = MMC_RSP_R1B | MMC_CMD_AC;
> +       }
> +
>         mmc->ops = &mmci_ops;
>
>         /* We support these PM capabilities. */
> diff --git a/drivers/mmc/host/mmci.h b/drivers/mmc/host/mmci.h
> index 550dd39..35372cd 100644
> --- a/drivers/mmc/host/mmci.h
> +++ b/drivers/mmc/host/mmci.h
> @@ -161,6 +161,7 @@
>  #define MCI_ST_CEATAEND                (1 << 23)
>  #define MCI_ST_CARDBUSY                (1 << 24)
>  /* Extended status bits for the STM32 variants */
> +#define MCI_STM32_DPSMACTIVE   BIT(12)
>  #define MCI_STM32_BUSYD0       BIT(20)
>
>  #define MMCICLEAR              0x038
> @@ -264,6 +265,7 @@ struct mmci_host;
>   * @cmdreg_lrsp_crc: enable value for long response with crc
>   * @cmdreg_srsp_crc: enable value for short response with crc
>   * @cmdreg_srsp: enable value for short response without crc
> + * @cmdreg_stop: enable value for stop and abort transmission
>   * @datalength_bits: number of bits in the MMCIDATALENGTH register
>   * @fifosize: number of bytes that can be written when MMCI_TXFIFOEMPTY
>   *           is asserted (likewise for RX)
> @@ -316,6 +318,7 @@ struct variant_data {
>         unsigned int            cmdreg_lrsp_crc;
>         unsigned int            cmdreg_srsp_crc;
>         unsigned int            cmdreg_srsp;
> +       unsigned int            cmdreg_stop;
>         unsigned int            datalength_bits;
>         unsigned int            fifosize;
>         unsigned int            fifohalfsize;
> @@ -375,6 +378,7 @@ struct mmci_host {
>         void __iomem            *base;
>         struct mmc_request      *mrq;
>         struct mmc_command      *cmd;
> +       struct mmc_command      stop_abort;
>         struct mmc_data         *data;
>         struct mmc_host         *mmc;
>         struct clk              *clk;
> --
> 2.7.4
>

Kind regards
Uffe

WARNING: multiple messages have this Message-ID (diff)
From: ulf.hansson@linaro.org (Ulf Hansson)
To: linux-arm-kernel@lists.infradead.org
Subject: [PATCH V2 2/2] mmc: mmci: add variant property to send stop cmd if a command fail
Date: Wed, 21 Nov 2018 18:56:12 +0100	[thread overview]
Message-ID: <CAPDyKFpZyYJbDDjYaFPfGEfKme6QedxcD715M2v7kAf7tpUf5A@mail.gmail.com> (raw)
In-Reply-To: <1541583041-17461-3-git-send-email-ludovic.Barre@st.com>

On 7 November 2018 at 10:30, Ludovic Barre <ludovic.Barre@st.com> wrote:
> From: Ludovic Barre <ludovic.barre@st.com>
>
> The mmc framework follows the requirement of SD_Specification:
> the STOP_TRANSMISSION is sent on multiple write/read commands
> and the stop command (alone), not needed on other ADTC commands.
>
> But, if an error happens on command or data transmission, some
> variants require a stop command "STOP_TRANSMISION" to clear the DPSM
> "Data Path State Machine". If it's not done the next data
> command freezes hardware block.
> Needed to support the STM32 sdmmc variant.

May I suggest some re-wording of this changelog, as to make it more
clear of why this is needed. Something along the lines of:

"The current approach with sending a CMD12 (STOP_TRANSMISSION) to
complete a data transfer request, either because of using the open
ended transmission type or because of receiving an error during a data
transfer, isn't sufficient for the STM32 sdmmc variant.

More precisely, for STM32 sdmmc the DPSM ("Data Path State Machine" )
needs to be cleared by sending a CMD12, also for the so called ADTC
commands. For this reason, add a new mmci variant property and let the
driver send a CMD12 to complete ADTC commands, in case it's set."

>
> Signed-off-by: Ludovic Barre <ludovic.barre@st.com>
> ---
>  drivers/mmc/host/mmci.c | 33 +++++++++++++++++++++++++++++++++
>  drivers/mmc/host/mmci.h |  4 ++++
>  2 files changed, 37 insertions(+)
>
> diff --git a/drivers/mmc/host/mmci.c b/drivers/mmc/host/mmci.c
> index 13fa640..47b865d 100644
> --- a/drivers/mmc/host/mmci.c
> +++ b/drivers/mmc/host/mmci.c
> @@ -21,6 +21,7 @@
>  #include <linux/err.h>
>  #include <linux/highmem.h>
>  #include <linux/log2.h>
> +#include <linux/mmc/mmc.h>
>  #include <linux/mmc/pm.h>
>  #include <linux/mmc/host.h>
>  #include <linux/mmc/card.h>
> @@ -57,6 +58,8 @@ void sdmmc_variant_init(struct mmci_host *host);
>  #else
>  static inline void sdmmc_variant_init(struct mmci_host *host) {}
>  #endif
> +static void
> +mmci_start_command(struct mmci_host *host, struct mmc_command *cmd, u32 c);
>
>  static unsigned int fmax = 515633;
>
> @@ -274,6 +277,7 @@ static struct variant_data variant_stm32_sdmmc = {
>         .cmdreg_lrsp_crc        = MCI_CPSM_STM32_LRSP_CRC,
>         .cmdreg_srsp_crc        = MCI_CPSM_STM32_SRSP_CRC,
>         .cmdreg_srsp            = MCI_CPSM_STM32_SRSP,
> +       .cmdreg_stop            = MCI_CPSM_STM32_CMDSTOP,
>         .data_cmd_enable        = MCI_CPSM_STM32_CMDTRANS,
>         .irq_pio_mask           = MCI_IRQ_PIO_STM32_MASK,
>         .datactrl_first         = true,
> @@ -573,6 +577,24 @@ void mmci_dma_error(struct mmci_host *host)
>  static void
>  mmci_request_end(struct mmci_host *host, struct mmc_request *mrq)
>  {
> +       /*
> +        * If an error happens on command or data transmission, some variants
> +        * require a stop command to reinit the DPSM.
> +        * If it's not done the next data command freeze hardware block.
> +        */
> +       if (host->variant->cmdreg_stop) {
> +               u32 dpsm;
> +
> +               dpsm = readl_relaxed(host->base + MMCISTATUS);
> +               dpsm &= MCI_STM32_DPSMACTIVE;
> +
> +               if (dpsm && ((mrq->cmd && mrq->cmd->error) ||
> +                            (mrq->data && mrq->data->error))) {
> +                       mmci_start_command(host, &host->stop_abort, 0);
> +                       return;
> +               }

I would rather move this code to a separate function. Also, I think
you need something else (or additional) than polling the MMCISTATUS
register, as in principle you could end up sending CMD12 several times
for the same request, which isn't correct.

To me the best solution would probably be to make use of the
host->data pointer, as it becomes set when DPSM has been enabled.
However, host->data is also cleared in mmci_stop_data() which is being
called prior mmci_request_end(). In other words, we need to figure
under what conditions the new code above should be triggered/called
and then also change the conditions for when mmci_request_end() shall
be called.

In principle look at callers of mmci_request_end() and
mmci_stop_data() and update those paths.

> +       }
> +
>         writel(0, host->base + MMCICOMMAND);
>
>         BUG_ON(host->data);
> @@ -1100,6 +1122,10 @@ mmci_start_command(struct mmci_host *host, struct mmc_command *cmd, u32 c)
>                 mmci_reg_delay(host);
>         }
>
> +       if (host->variant->cmdreg_stop &&
> +           cmd->opcode == MMC_STOP_TRANSMISSION)
> +               c |= host->variant->cmdreg_stop;
> +

Hmm.

It looks like the above changes, together with the introduction of the
variant property, belongs in a separate patch, preceding $subject
patch in the series.

The reason is, that to me, it makes sense to add the special treatment
of the ADTC commands in a separate patch. Can you please split it up
and then of course update the change log accordingly!?

>         c |= cmd->opcode | host->variant->cmdreg_cpsm_enable;
>         if (cmd->flags & MMC_RSP_PRESENT) {
>                 if (cmd->flags & MMC_RSP_136)
> @@ -1950,6 +1976,13 @@ static int mmci_probe(struct amba_device *dev,
>                 mmc->max_busy_timeout = 0;
>         }
>
> +       /* prepare the stop command, used to abort and reinitialized the DPSM */
> +       if (variant->cmdreg_stop) {
> +               host->stop_abort.opcode = MMC_STOP_TRANSMISSION;
> +               host->stop_abort.arg = 0;
> +               host->stop_abort.flags = MMC_RSP_R1B | MMC_CMD_AC;
> +       }
> +
>         mmc->ops = &mmci_ops;
>
>         /* We support these PM capabilities. */
> diff --git a/drivers/mmc/host/mmci.h b/drivers/mmc/host/mmci.h
> index 550dd39..35372cd 100644
> --- a/drivers/mmc/host/mmci.h
> +++ b/drivers/mmc/host/mmci.h
> @@ -161,6 +161,7 @@
>  #define MCI_ST_CEATAEND                (1 << 23)
>  #define MCI_ST_CARDBUSY                (1 << 24)
>  /* Extended status bits for the STM32 variants */
> +#define MCI_STM32_DPSMACTIVE   BIT(12)
>  #define MCI_STM32_BUSYD0       BIT(20)
>
>  #define MMCICLEAR              0x038
> @@ -264,6 +265,7 @@ struct mmci_host;
>   * @cmdreg_lrsp_crc: enable value for long response with crc
>   * @cmdreg_srsp_crc: enable value for short response with crc
>   * @cmdreg_srsp: enable value for short response without crc
> + * @cmdreg_stop: enable value for stop and abort transmission
>   * @datalength_bits: number of bits in the MMCIDATALENGTH register
>   * @fifosize: number of bytes that can be written when MMCI_TXFIFOEMPTY
>   *           is asserted (likewise for RX)
> @@ -316,6 +318,7 @@ struct variant_data {
>         unsigned int            cmdreg_lrsp_crc;
>         unsigned int            cmdreg_srsp_crc;
>         unsigned int            cmdreg_srsp;
> +       unsigned int            cmdreg_stop;
>         unsigned int            datalength_bits;
>         unsigned int            fifosize;
>         unsigned int            fifohalfsize;
> @@ -375,6 +378,7 @@ struct mmci_host {
>         void __iomem            *base;
>         struct mmc_request      *mrq;
>         struct mmc_command      *cmd;
> +       struct mmc_command      stop_abort;
>         struct mmc_data         *data;
>         struct mmc_host         *mmc;
>         struct clk              *clk;
> --
> 2.7.4
>

Kind regards
Uffe

  reply	other threads:[~2018-11-21 17:56 UTC|newest]

Thread overview: 24+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-11-07  9:30 [PATCH V2 0/2] mmc: mmci: add stop command Ludovic Barre
2018-11-07  9:30 ` Ludovic Barre
2018-11-07  9:30 ` Ludovic Barre
2018-11-07  9:30 ` [PATCH V2 1/2] mmc: mmci: send stop command if sbc error issue Ludovic Barre
2018-11-07  9:30   ` Ludovic Barre
2018-11-07  9:30   ` Ludovic Barre
2018-11-20  9:42   ` Ulf Hansson
2018-11-20  9:42     ` Ulf Hansson
2018-12-05 14:23     ` Ulf Hansson
2018-12-05 14:23       ` Ulf Hansson
2018-12-05 15:49       ` Ludovic BARRE
2018-12-05 15:49         ` Ludovic BARRE
2018-12-05 15:49         ` Ludovic BARRE
2018-12-05 16:00       ` Ludovic BARRE
2018-12-05 16:00         ` Ludovic BARRE
2018-12-05 16:00         ` Ludovic BARRE
2018-11-07  9:30 ` [PATCH V2 2/2] mmc: mmci: add variant property to send stop cmd if a command fail Ludovic Barre
2018-11-07  9:30   ` Ludovic Barre
2018-11-07  9:30   ` Ludovic Barre
2018-11-21 17:56   ` Ulf Hansson [this message]
2018-11-21 17:56     ` Ulf Hansson
2018-11-22 10:14     ` Ludovic BARRE
2018-11-22 10:14       ` Ludovic BARRE
2018-11-22 10:14       ` Ludovic BARRE

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=CAPDyKFpZyYJbDDjYaFPfGEfKme6QedxcD715M2v7kAf7tpUf5A@mail.gmail.com \
    --to=ulf.hansson@linaro.org \
    --cc=alexandre.torgue@st.com \
    --cc=devicetree@vger.kernel.org \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mmc@vger.kernel.org \
    --cc=linux-stm32@st-md-mailman.stormreply.com \
    --cc=ludovic.Barre@st.com \
    --cc=mcoquelin.stm32@gmail.com \
    --cc=robh+dt@kernel.org \
    --cc=srinivas.kandagatla@linaro.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.