Re: [PATCH v3] mmc: mediatek: Add MMC_CAP_SDIO_IRQ support
From: Jjian Zhou
Date: Tue Dec 11 2018 - 07:11:48 EST
On Tue, 2018-12-11 at 17:29 +0800, Ulf Hansson wrote:
> On Mon, 10 Dec 2018 at 12:44, Jjian Zhou <jjian.zhou@xxxxxxxxxxxx> wrote:
> >
> > From: jjian zhou <jjian.zhou@xxxxxxxxxxxx>
> >
> > This patch enables support SDIO IRQs. It enables
> > MMC_CAP_SDIO_IRQ & MMC_CAP2_SDIO_IRQ_NOTHREAD
> > and implement the ->ack_sdio_irq callback.
> >
> > Signed-off-by: Jjian Zhou <jjian.zhou@xxxxxxxxxxxx>
> > Signed-off-by: Chaotian Jing <chaotian.jing@xxxxxxxxxxxx>
> > Signed-off-by: Yong Mao <yong.mao@xxxxxxxxxxxx>
> > ---
> > drivers/mmc/host/mtk-sd.c | 49 ++++++++++++++++++++++++++++++++++++++++++++---
> > 1 file changed, 46 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/mmc/host/mtk-sd.c b/drivers/mmc/host/mtk-sd.c
> > index 6334cc7..40a1848 100644
> > --- a/drivers/mmc/host/mtk-sd.c
> > +++ b/drivers/mmc/host/mtk-sd.c
> > @@ -1114,6 +1114,7 @@ static void msdc_start_command(struct msdc_host *host,
> > struct mmc_request *mrq, struct mmc_command *cmd)
> > {
> > u32 rawcmd;
> > + unsigned long flags;
> >
> > WARN_ON(host->cmd);
> > host->cmd = cmd;
> > @@ -1131,7 +1132,10 @@ static void msdc_start_command(struct msdc_host *host,
> > cmd->error = 0;
> > rawcmd = msdc_cmd_prepare_raw_cmd(host, mrq, cmd);
> >
> > + spin_lock_irqsave(&host->lock, flags);
> > sdr_set_bits(host->base + MSDC_INTEN, cmd_ints_mask);
> > + spin_unlock_irqrestore(&host->lock, flags);
> > +
>
> The above looks like a reasonable change, however does it really
> belong in $subject patch. If so, can you explain why?
>
> Maybe you should move this into a standalone separate patch?
>
Thanks for your suggestion.
The modification of MSDC_INTEN is non-atomic operation. If it enanbles
MMC_CAP_SDIO_IRQ, there is a race condition of access MSDC_INTEN in irq
and thread context. So need make modification of MSDC_INTEN by atomic.
> > writel(cmd->arg, host->base + SDC_ARG);
> > writel(rawcmd, host->base + SDC_CMD);
> > }
> > @@ -1351,6 +1355,27 @@ static void msdc_request_timeout(struct work_struct *work)
> > }
> > }
> >
> > +static void msdc_enable_sdio_irq(struct mmc_host *mmc, int enb)
> > +{
> > + unsigned long flags;
> > + struct msdc_host *host = mmc_priv(mmc);
> > +
> > + if (enb)
> > + pm_runtime_get_sync(host->dev);
>
> I would convert to pm_runtime_get_noresume() as the host is already
> been runtime resumed by the mmc core.
will do it in next version.
> > +
> > + spin_lock_irqsave(&host->lock, flags);
> > + if (enb)
> > + sdr_set_bits(host->base + MSDC_INTEN, MSDC_INTEN_SDIOIRQ);
> > + else
> > + sdr_clr_bits(host->base + MSDC_INTEN, MSDC_INTEN_SDIOIRQ);
> > + spin_unlock_irqrestore(&host->lock, flags);
>
> You should move the code above into a separate function. Along the lines of:
>
> static void __msdc_enable_sdio_irq(struct mmc_host *mmc, int enb)
>
> You need this, because you will otherwise mess up the runtime PM usage
> count when acking the SDIO irq. More comments about that below.
will do it in next version.
> > +
> > + if (!enb) {
> > + pm_runtime_mark_lastwill do it in next version._busy(host->dev);
> > + pm_runtime_put_autosuspend(host->dev);
>
> Convert to pm_runtime_put_noidle(), as the above is already managed by
> the mmc core.
>
will do it in next version.
> > + }
> > +}
> > +
> > static irqreturn_t msdc_irq(int irq, void *dev_id)
> > {
> > struct msdc_host *host = (struct msdc_host *) dev_id;
> > @@ -1373,7 +1398,12 @@ static irqreturn_t msdc_irq(int irq, void *dev_id)
> > data = host->data;
> > spin_unlock_irqrestore(&host->lock, flags);
> >
> > - if (!(events & event_mask))
> > + if ((events & event_mask) & MSDC_INT_SDIOIRQ) {
> > + msdc_enable_sdio_irq(host->mmc, 0);
>
> To not mess up the runtime PM reference count and to avoid invoking
> runtime PM callbacks from atomic context, I suggest you call
> __msdc_enable_sdio_irq(host->mmc, 0) (new proposed function according
> to earlier comment).
will do it in next version.
>
> > + sdio_signal_irq(host->mmc);
> > + }
> > +
> > + if (!(events & (event_mask & ~MSDC_INT_SDIOIRQ)))
> > break;
> >
> > if (!mrq) {
> > @@ -1493,8 +1523,11 @@ static void msdc_init_hw(struct msdc_host *host)
> > */
> > sdr_set_bits(host->base + SDC_CFG, SDC_CFG_SDIO);
> >
> > - /* disable detect SDIO device interrupt function */
> > - sdr_clr_bits(host->base + SDC_CFG, SDC_CFG_SDIOIDE);
> > + /* Config SDIO device detect interrupt function */
> > + if (host->mmc->caps & MMC_CAP_SDIO_IRQ)
> > + sdr_set_bits(host->base + SDC_CFG, SDC_CFG_SDIOIDE);
> > + else
> > + sdr_clr_bits(host->base + SDC_CFG, SDC_CFG_SDIOIDE);
> >
> > /* Configure to default data timeout */
> > sdr_set_field(host->base + SDC_CFG, SDC_CFG_DTOC, 3);
> > @@ -2013,6 +2046,11 @@ static void msdc_hw_reset(struct mmc_host *mmc)
> > sdr_clr_bits(host->base + EMMC_IOCON, 1);
> > }
> >
> > +static void msdc_ack_sdio_irq(struct mmc_host *mmc)
> > +{
> > + msdc_enable_sdio_irq(mmc, 1);
>
> To not mess up the runtime PM reference count, I suggest you call
> __msdc_enable_sdio_irq(host->mmc, 1) (new proposed function according
> to earlier comment).
>
will do it in next version.
> > +}
> > +
> > static const struct mmc_host_ops mt_msdc_ops = {
> > .post_req = msdc_post_req,
> > .pre_req = msdc_pre_req,
> > @@ -2020,6 +2058,8 @@ static void msdc_hw_reset(struct mmc_host *mmc)
> > .set_ios = msdc_ops_set_ios,
> > .get_ro = mmc_gpio_get_ro,
> > .get_cd = mmc_gpio_get_cd,
> > + .enable_sdio_irq = msdc_enable_sdio_irq,
> > + .ack_sdio_irq = msdc_ack_sdio_irq,
> > .start_signal_voltage_switch = msdc_ops_switch_volt,
> > .card_busy = msdc_card_busy,
> > .execute_tuning = msdc_execute_tuning,
> > @@ -2147,6 +2187,9 @@ static int msdc_drv_probe(struct platform_device *pdev)
> > else
> > mmc->f_min = DIV_ROUND_UP(host->src_clk_freq, 4 * 4095);
> >
> > + if (mmc->caps & MMC_CAP_SDIO_IRQ)
> > + mmc->caps2 |= MMC_CAP2_SDIO_IRQ_NOTHREAD;
> > +
> > mmc->caps |= MMC_CAP_ERASE | MMC_CAP_CMD23;
> > /* MMC core transfer sizes tunable parameters */
> > mmc->max_segs = MAX_BD_NUM;
> > --
> > 1.9.1
> >
>
> Kind regards
> Uffe