On Wed, 2022-06-08 at 16:13 +0200, Ulf Hansson wrote:
> On Wed, 25 May 2022 at 03:51, Axe Yang <axe.yang@xxxxxxxxxxxx> wrote:
> >
> > Add support for eint IRQ when MSDC is used as an SDIO host. This
> > feature requires SDIO device support async IRQ function. With this
> > feature, SDIO host can be awakened by SDIO card in suspend state,
> > without additional pin.
> >
> > MSDC driver will time-share the SDIO DAT1 pin. During suspend, MSDC
> > turn off clock and switch SDIO DAT1 pin to GPIO mode. And during
> > resume, switch GPIO function back to DAT1 mode then turn on clock.
> >
> > Some device tree property should be added or modified in MSDC node
> > to support SDIO eint IRQ. Pinctrls "state_eint" is mandatory. Since
> > this feature depends on asynchronous interrupts, "wakeup-source",
> > "keep-power-in-suspend" and "cap-sdio-irq" flags are necessary, and
> > the interrupts list should be extended(the interrupt named with
> > sdio_wakeup):
> > &mmcX {
> > ...
> > interrupt-names = "msdc", "sdio_wakeup";
> > interrupts-extended = <...>,
> > <&pio xxx
> > IRQ_TYPE_LEVEL_LOW>;
> > ...
> > pinctrl-names = "default", "state_uhs",
> > "state_eint";
> > ...
> > pinctrl-2 = <&mmc2_pins_eint>;
> > ...
> > cap-sdio-irq;
> > keep-power-in-suspend;
> > wakeup-source;
> > ...
> > };
> > Reviewed-by: AngeloGioacchino Del Regno <
> > angelogioacchino.delregno@xxxxxxxxxxxxx>
> > Co-developed-by: Yong Mao <yong.mao@xxxxxxxxxxxx>
> > Signed-off-by: Yong Mao <yong.mao@xxxxxxxxxxxx>
> > Signed-off-by: Axe Yang <axe.yang@xxxxxxxxxxxx>
> > ---
> > drivers/mmc/host/mtk-sd.c | 80
> > ++++++++++++++++++++++++++++++++++++---
> > 1 file changed, 74 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/mmc/host/mtk-sd.c b/drivers/mmc/host/mtk-sd.c
> > index 195dc897188b..2d5b23616df6 100644
> > --- a/drivers/mmc/host/mtk-sd.c
> > +++ b/drivers/mmc/host/mtk-sd.c
> > @@ -1,6 +1,6 @@
> > // SPDX-License-Identifier: GPL-2.0-only
> > /*
> > - * Copyright (c) 2014-2015 MediaTek Inc.
> > + * Copyright (c) 2014-2015, 2022 MediaTek Inc.
> > * Author: Chaotian.Jing <chaotian.jing@xxxxxxxxxxxx>
> > */
> >
> > @@ -20,6 +20,7 @@
> > #include <linux/platform_device.h>
> > #include <linux/pm.h>
> > #include <linux/pm_runtime.h>
> > +#include <linux/pm_wakeirq.h>
> > #include <linux/regulator/consumer.h>
> > #include <linux/slab.h>
> > #include <linux/spinlock.h>
> > @@ -440,8 +441,10 @@ struct msdc_host {
> > struct pinctrl *pinctrl;
> > struct pinctrl_state *pins_default;
> > struct pinctrl_state *pins_uhs;
> > + struct pinctrl_state *pins_eint;
> > struct delayed_work req_timeout;
> > int irq; /* host interrupt */
> > + int eint_irq; /* interrupt from sdio device for
> > waking up system */
> > struct reset_control *reset;
> >
> > struct clk *src_clk; /* msdc source clock */
> > @@ -1520,17 +1523,41 @@ static void __msdc_enable_sdio_irq(struct
> > msdc_host *host, int enb)
> >
> > static void msdc_enable_sdio_irq(struct mmc_host *mmc, int enb)
> > {
> > - unsigned long flags;
> > struct msdc_host *host = mmc_priv(mmc);
> > + unsigned long flags;
> > + int ret;
> >
> > spin_lock_irqsave(&host->lock, flags);
> > __msdc_enable_sdio_irq(host, enb);
> > spin_unlock_irqrestore(&host->lock, flags);
> >
> > - if (enb)
> > - pm_runtime_get_noresume(host->dev);
> > - else
> > - pm_runtime_put_noidle(host->dev);
> > + if (mmc_card_enable_async_irq(mmc->card) && host-
> > >pins_eint) {
> > + if (enb) {
> > + pinctrl_select_state(host->pinctrl, host-
> > >pins_eint);
>
> This looks a bit odd to me. The pins are not supposed to be
> configured
> for wakeirq at this point, right?
>
> As I understand it, the pin state for wakeirq should be set from the
> ->runtime_suspend() callback, no?
Yes, it is odd, but necessary.
In dev_pm_set_dedicated_wake_irq_reverse() -> ... ->
request_threaded_irq() -> __setup_irq() -> irq_request_resources() ->
mtk_eint_irq_request_resources(), the SDIO DAT1 pin will be force reset
to GPIO mode.
I have to call pinctrl_select_state() in pairs to restore DAT1 pin
state to MSDC mode.
>
> > + ret =
> > dev_pm_set_dedicated_wake_irq_reverse(host->dev, host->eint_irq);
> > +
> > + if (ret) {
> > + dev_err(host->dev, "Failed to
> > register SDIO wakeup irq!\n");
> > + host->pins_eint = NULL;
> > + pm_runtime_get_noresume(host->dev);
> > + } else {
> > + dev_info(host->dev, "SDIO eint irq:
> > %d!\n", host->eint_irq);
>
> If you want to log a message, please use a dev_dbg for this instead.
Sure, will change to dev_dbg in next version.
>
> > + device_init_wakeup(host->dev,
> > true);
>
> To me, it looks like this is better called from ->probe(), once.
Will move it to probe() in next version.
>
> > + }
> > +
> > + pinctrl_select_state(host->pinctrl, host-
> > >pins_uhs);
>
> Assuming that we can drop the earlier call to pinctrl_select_state()
> to set "host->pins_eint", this call can be dropped too.
Can not drop this call, it is for restore DAT1 pinmux to MSDC mode.
>
> > + } else {
> > + dev_pm_clear_wake_irq(host->dev);
> > + }
> > + } else {
> > + if (enb) {
> > + /* Ensure host->pins_eint is NULL */
> > + host->pins_eint = NULL;
> > + pm_runtime_get_noresume(host->dev);
> > + } else {
> > + pm_runtime_put_noidle(host->dev);
> > + }
> > + }
> > }
> >
> > static irqreturn_t msdc_cmdq_irq(struct msdc_host *host, u32
> > intsts)
> > @@ -2631,6 +2658,19 @@ static int msdc_drv_probe(struct
> > platform_device *pdev)
> > goto host_free;
> > }
> >
> > + /* Support for SDIO eint irq ? */
> > + if ((mmc->pm_caps & MMC_PM_WAKE_SDIO_IRQ) && (mmc->pm_caps
> > & MMC_PM_KEEP_POWER)) {
> > + host->eint_irq = platform_get_irq_byname(pdev,
> > "sdio_wakeup");
> > + if (host->eint_irq > 0) {
> > + host->pins_eint =
> > pinctrl_lookup_state(host->pinctrl, "state_eint");
> > + if (IS_ERR(host->pins_eint)) {
> > + dev_err_probe(&pdev->dev,
> > PTR_ERR(host->pins_eint),
> > + "Cannot find pinctrl
> > eint!\n");
>
> We can probably use dev_err() instead of dev_err_probe() as
> pinctrl_lookup_state() should never return -EPROBE_DEFER, I think.
Yes. will update that in next version.
>
> > + host->pins_eint = NULL;
> > + }
> > + }
> > + }
> > +
> > msdc_of_property_parse(pdev, host);
> >
> > host->dev = &pdev->dev;
> > @@ -2845,6 +2885,12 @@ static int __maybe_unused
> > msdc_runtime_suspend(struct device *dev)
> > struct msdc_host *host = mmc_priv(mmc);
> >
> > msdc_save_reg(host);
> > +
> > + if (sdio_irq_claimed(mmc) && host->pins_eint) {
> > + disable_irq(host->irq);
> > + pinctrl_select_state(host->pinctrl, host-
> > >pins_eint);
> > + sdr_clr_bits(host->base + SDC_CFG,
> > SDC_CFG_SDIOIDE);
>
> __msdc_enable_sdio_irq() also calls "sdr_clr_bits(host->base +
> MSDC_INTEN, MSDC_INTEN_SDIOIRQ);"
>
> Perhaps we should call __msdc_enable_sdio_irq() here instead? To be
> consistent.
Yes, we can call __msdc_enable_sdio_irq() instead. Will update that in
next verison.
>
> > + }
> > msdc_gate_clock(host);
> > return 0;
> > }
> > @@ -2860,12 +2906,19 @@ static int __maybe_unused
> > msdc_runtime_resume(struct device *dev)
> > return ret;
> >
> > msdc_restore_reg(host);
> > +
> > + if (sdio_irq_claimed(mmc) && host->pins_eint) {
>
> msdc_restore_reg() already calls __msdc_enable_sdio_irq(), but based
> only upon whether sdio_irq_claimed() returns true.
>
Will remove redundancy MSDC_INTEN_SDIOIRQ control in
msdc_runtion_resume() in next version.
> It looks like we should align the code in
> msdc_runtime_resume|suspend(). Perhaps sdio_irq_claimed() should
> indicate in both cases that __msdc_enable_sdio_irq() needs to be
> called, while "host->pins_eint" means that we have also additional
> wakeup configurations (pins and irqs) to handle.
sdr_set_bits(..., SDC_CFG_SDIOIDE) is only needed when wake up irq is
enabled. So, maybe I can refactor this part as below:
if (host->pins_eint) {
...
if (sdio_irq_claimed())
__msdc_enable_sdio_irq(host, 0);
...
}
in msdc_runtime_suspend() ?
in msdc_runtime_resume(), __msdc_enable_sdio_irq() is called in
msdc_restore_reg().
While in msdc_runtime_suspend(), it is called only if host->pins_eint
return true.
In this case, msdc_runtime_resume|suspend() are not prefectly aligned.
Any better way to do that?
>
> > + sdr_set_bits(host->base + SDC_CFG,
> > SDC_CFG_SDIOIDE);
> > + pinctrl_select_state(host->pinctrl, host-
> > >pins_uhs);
> > + enable_irq(host->irq);
> > + }
> > return 0;
> > }
> >
> > static int __maybe_unused msdc_suspend(struct device *dev)
> > {
> > struct mmc_host *mmc = dev_get_drvdata(dev);
> > + struct msdc_host *host = mmc_priv(mmc);
> > int ret;
> >
> > if (mmc->caps2 & MMC_CAP2_CQE) {
> > @@ -2874,11 +2927,26 @@ static int __maybe_unused
> > msdc_suspend(struct device *dev)
> > return ret;
> > }
> >
> > + if (sdio_irq_claimed(mmc) && host->pins_eint) {
> > + pm_runtime_put_sync_suspend(dev);
> > +
> > + return 0;
> > + }
> > +
>
> I assume the point with the above is to trigger rpm_suspend() to be
> called for the device, so that the wakeirq can be enabled, correctly?
Yes, the intention is to trigger rpm_suspend|resume(), with these two
functions all operations on wakeirq can be performed completely.
But with pm_rumtime_force_resume(), dev->power.needs_force_resume is
false, the function goto out directly everytime.
>
> However, this isn't the correct way to do it (for various reasons I
> can explain, if you want). Instead I think there are two options
> going
> forward:
> 1. Deal with the wakeirq from the system suspend/resume callbacks,
> locally in the driver.
> 2. Extend pm_runtime_force_suspend|resume() to let it deal with the
> wakeirq for us. Similar to what rpm_suspend|resume() do.
>
> I am inclined to try with option 2) first, as this would prevent the
> boilerplate code that otherwise gets introduced by option 1). To help
> out, I have prepared a patch that I am about to send, I will keep you
> posted.
Thanks for the patch. And I still need to use pm_runtime_get_noresume()
to bump up runtime PM usage counter to ensure pm_rumtime_force_resume()
can be fully executed.
>
> > return pm_runtime_force_suspend(dev);
> > }
> >
> > static int __maybe_unused msdc_resume(struct device *dev)
> > {
> > + struct mmc_host *mmc = dev_get_drvdata(dev);
> > + struct msdc_host *host = mmc_priv(mmc);
> > +
> > + if (sdio_irq_claimed(mmc) && host->pins_eint) {
> > + pm_runtime_get_sync(dev);
>
> Similar comments apply to this as for msdc_suspend().
pm_runtime_put_noidle() will be added here to decrement runtime PM
usage counter. Do you have any comments on this?
>
> > +
> > + return 0;
> > + }
> > +
> > return pm_runtime_force_resume(dev);
> > }
> >
>
> Kind regards
> Uffe
Regards,
Axe
