Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20190128144119.10092-1-martink@posteo.de> <CAPDyKFo7949rcxEX5U2QUS0c2xhC_z+JGeOiF2HC6+QukVpEaQ@mail.gmail.com>
 <20190131092006.363d1dd4@erd987>
In-Reply-To: <20190131092006.363d1dd4@erd987>
From:   Ulf Hansson <ulf.hansson@linaro.org>
Date:   Thu, 31 Jan 2019 13:17:23 +0100
Message-ID: <CAPDyKFr=kcoGK0C2A8JH-nhfgUgoRVa-E=6KPRjFO51QQxjJLA@mail.gmail.com>
Subject: Re: [PATCH v4] mmc: mxs-mmc: Introduce regulator support
To:     Robin van der Gracht <robin@protonic.nl>
Cc:     Martin Kepplinger <martink@posteo.de>,
        "linux-mmc@vger.kernel.org" <linux-mmc@vger.kernel.org>,
        Linux ARM <linux-arm-kernel@lists.infradead.org>,
        Shawn Guo <shawnguo@kernel.org>,
        Sascha Hauer <s.hauer@pengutronix.de>,
        dl-linux-imx <linux-imx@nxp.com>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Martin Kepplinger <martin.kepplinger@ginzinger.com>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, 31 Jan 2019 at 09:20, Robin van der Gracht <robin@protonic.nl> wrote:
>
> On Mon, 28 Jan 2019 22:15:23 +0100
> Ulf Hansson <ulf.hansson@linaro.org> wrote:
>
> > On Mon, 28 Jan 2019 at 15:41, Martin Kepplinger <martink@posteo.de> wrote:
> > >
> > > From: Martin Kepplinger <martin.kepplinger@ginzinger.com>
> > >
> > > This adds support for explicitly switching the mmc's power on and off
> > > which is needed for example for WL1837 WL1271 wifi controllers on imx28.
> > >
> > > While the wifi's vmmc-supply regulator can be configured in devicetree,
> > > "ip link set wlan0 down" doesn't turn off the VMMC regulator which leads
> > > to hangs when loading firmware, for example.
> > >
> > > Signed-off-by: Martin Kepplinger <martin.kepplinger@ginzinger.com>
> > > ---
> > >
> > >
> > > revision history
> > > ----------------
> > > v4: re-added forgotten regulator_enable() during probe
> > > v3: improve API usage as suggested by Ulf
> > > v2: tested patch with changes suggested by Robin
> > > v1: question, why https://patchwork.kernel.org/patch/4365751/ didn't get in
> > >
> > >
> > >  drivers/mmc/host/mxs-mmc.c | 34 +++++++++++++++++++++++++---------
> > >  1 file changed, 25 insertions(+), 9 deletions(-)
> > >
> > > diff --git a/drivers/mmc/host/mxs-mmc.c b/drivers/mmc/host/mxs-mmc.c
> > > index add1e70195ea..23d275269d61 100644
> > > --- a/drivers/mmc/host/mxs-mmc.c
> > > +++ b/drivers/mmc/host/mxs-mmc.c
> > > @@ -517,6 +517,22 @@ static void mxs_mmc_set_ios(struct mmc_host *mmc, struct mmc_ios *ios)
> > >         else
> > >                 host->bus_width = 0;
> > >
> > > +       switch (ios->power_mode) {
> > > +       case MMC_POWER_OFF:
> > > +               if (!IS_ERR(host->mmc->supply.vmmc))
> > > +                       mmc_regulator_set_ocr(host->mmc,
> > > +                                             host->mmc->supply.vmmc, 0);
> > > +               break;
> > > +       case MMC_POWER_UP:
> > > +               if (!IS_ERR(host->mmc->supply.vmmc))
> > > +                       mmc_regulator_set_ocr(host->mmc,
> > > +                                             host->mmc->supply.vmmc,
> > > +                                             ios->vdd);
> > > +               break;
> > > +       default:
> > > +               break;
> > > +       }
> > > +
> > >         if (ios->clock)
> > >                 mxs_ssp_set_clk_rate(&host->ssp, ios->clock);
> > >  }
> > > @@ -588,7 +604,6 @@ static int mxs_mmc_probe(struct platform_device *pdev)
> > >         struct mmc_host *mmc;
> > >         struct resource *iores;
> > >         int ret = 0, irq_err;
> > > -       struct regulator *reg_vmmc;
> > >         struct mxs_ssp *ssp;
> > >
> > >         irq_err = platform_get_irq(pdev, 0);
> > > @@ -614,14 +629,15 @@ static int mxs_mmc_probe(struct platform_device *pdev)
> > >         host->mmc = mmc;
> > >         host->sdio_irq_en = 0;
> > >
> > > -       reg_vmmc = devm_regulator_get(&pdev->dev, "vmmc");
> > > -       if (!IS_ERR(reg_vmmc)) {
> > > -               ret = regulator_enable(reg_vmmc);
> > > -               if (ret) {
> > > -                       dev_err(&pdev->dev,
> > > -                               "Failed to enable vmmc regulator: %d\n", ret);
> > > -                       goto out_mmc_free;
> > > -               }
> > > +       ret = mmc_regulator_get_supply(mmc);
> > > +       if (ret == -EPROBE_DEFER)
> > > +               goto out_mmc_free;
> > > +
> > > +       ret = regulator_enable(mmc->supply.vmmc);
> >
> > This is wrong, as it may cause the regulator usage count to become
> > wrongly balanced.
> >
> > Instead, via ->set_ios() when calling mmc_regulator_set_ocr(), it will
> > take care of enabling and disabling the regulator depending of the
> > requested vdd voltage level.
> >
> > > +       if (ret) {
> > > +               dev_err(&pdev->dev,
> > > +                       "Failed to enable vmmc regulator: %d\n", ret);
> > > +               goto out_mmc_free;
> > >         }
> > >
> > >         ssp->clk = devm_clk_get(&pdev->dev, NULL);
> > > --
> > > 2.20.1
> > >
> >
> > BTW, you didn't really answer my earlier question about the TI WiFi
> > chip. Doesn't you need a special clock for WiFi chip as well? How do
> > you intend to manage that?
>
> I used an external 32K oscillator (SLOW_CLK) for my wl1271. Other
> clocks ware generated on the module.

Right. How do you control that clock? Did you model it as clock via
the common clock framework?

>
> I had to supply a 'vmmc-supply' in your wl1271 devicetree node,
> which will be used to power on/off the wlan module. The supply should
> be a (delayed) GPIO controlled 'fixed-regulator' attached to the
> wlan_en pin on the module.

Right, thanks for explaining.

>
> 1: Documentation/devicetree/bindings/net/wireless/ti,wlcore.txt
>

This sounds like a good fit for mmc pwrseq simple. There are already
similar users for it.

Have a look at: /drivers/mmc/core/pwrseq*
If the mmc host driver calls mmc_of_parse() during ->probe(), a pwrseq
instance will be hooked up to it. Once the mmc core tries to power up
the card it will make use of the attached pwrseq for the mmc host in
question.

In this way, you can control the clock and GPIO line, in more exact
ways that is needed by the WiFi chip.

Here is a DT example (look for "mmc-pwrseq-simple"):
arch/arm/boot/dts/imx6qdl-sr-som-ti.dtsi

This should do the trick for you. On the other hand, I don't mind that
you still add regulator support to the driver, along the lines of what
$subject patch does, however it may not be exactly what you need for
the WiFi case.

Kind regards
Uffe