Message-ID: <8d72c895ece6dce7d8badb241eebcbe076a03f81.camel@puri.sm>
Subject: Re: [PATCH v2] media: i2c: dw9714: add optional regulator support
From:   Martin Kepplinger <martin.kepplinger@puri.sm>
To:     Sakari Ailus <sakari.ailus@linux.intel.com>
Cc:     mchehab@kernel.org, broonie@kernel.org, kernel@puri.sm,
        linux-kernel@vger.kernel.org, linux-media@vger.kernel.org,
        linux-pm@vger.kernel.org, Angus Ainslie <angus@akkea.ca>
Date:   Fri, 26 Nov 2021 12:06:03 +0100
In-Reply-To: <YaC6nZIQOsrpBY8V@paasikivi.fi.intel.com>
References: <20211126090107.1243558-1-martin.kepplinger@puri.sm>
         <YaC6nZIQOsrpBY8V@paasikivi.fi.intel.com>
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.38.3-1 
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

Am Freitag, dem 26.11.2021 um 12:44 +0200 schrieb Sakari Ailus:
> Hi Martin,
> 
> On Fri, Nov 26, 2021 at 10:01:07AM +0100, Martin Kepplinger wrote:
> > From: Angus Ainslie <angus@akkea.ca>
> > 
> > Allow the dw9714 to control a regulator and adjust suspend() and
> > resume()
> > to support both runtime and system pm.
> > 
> > Signed-off-by: Angus Ainslie <angus@akkea.ca>
> > Signed-off-by: Martin Kepplinger <martin.kepplinger@puri.sm>
> > ---
> > 
> > revision history
> > ----------------
> > 
> > v2: (thank you Mark)
> >  * simplify the regulator_get_optional() error path
> >  * fix regulator usage during probe()
> > 
> > v1:
> > https://lore.kernel.org/linux-media/20211125080922.978583-1-martin.kepplinger@puri.sm/
> > 
> > 
> > 
> >  drivers/media/i2c/dw9714.c | 39
> > ++++++++++++++++++++++++++++++++++++++
> >  1 file changed, 39 insertions(+)
> > 
> > diff --git a/drivers/media/i2c/dw9714.c
> > b/drivers/media/i2c/dw9714.c
> > index 3863dfeb8293..e8cc19b89861 100644
> > --- a/drivers/media/i2c/dw9714.c
> > +++ b/drivers/media/i2c/dw9714.c
> > @@ -5,6 +5,7 @@
> >  #include <linux/i2c.h>
> >  #include <linux/module.h>
> >  #include <linux/pm_runtime.h>
> > +#include <linux/regulator/consumer.h>
> >  #include <media/v4l2-ctrls.h>
> >  #include <media/v4l2-device.h>
> >  #include <media/v4l2-event.h>
> > @@ -36,6 +37,7 @@ struct dw9714_device {
> >         struct v4l2_ctrl_handler ctrls_vcm;
> >         struct v4l2_subdev sd;
> >         u16 current_val;
> > +       struct regulator *vcc;
> >  };
> >  
> >  static inline struct dw9714_device *to_dw9714_vcm(struct v4l2_ctrl
> > *ctrl)
> > @@ -145,6 +147,21 @@ static int dw9714_probe(struct i2c_client
> > *client)
> >         if (dw9714_dev == NULL)
> >                 return -ENOMEM;
> >  
> > +       dw9714_dev->vcc = devm_regulator_get_optional(&client->dev,
> > "vcc");
> 
> You you used regular devm_regulator_get(), you could remove the error
> handling below. If there's no regulator, you'll simply get a dummy
> one.

ok thanks


> 
> > +       if (IS_ERR(dw9714_dev->vcc)) {
> > +               dev_dbg(&client->dev, "No vcc regulator found:
> > %ld\n",
> > +                       PTR_ERR(dw9714_dev->vcc));
> > +               dw9714_dev->vcc = NULL;
> > +       }
> > +
> > +       if (dw9714_dev->vcc) {
> 
> With (dummy) regulators, these checks become unnecessary.
> 
> > +               rval = regulator_enable(dw9714_dev->vcc);
> > +               if (rval < 0) {
> > +                       dev_err(&client->dev, "failed to enable
> > vcc: %d\n", rval);
> > +                       return rval;
> > +               }
> > +       }
> > +
> >         v4l2_i2c_subdev_init(&dw9714_dev->sd, client, &dw9714_ops);
> >         dw9714_dev->sd.flags |= V4L2_SUBDEV_FL_HAS_DEVNODE |
> >                                 V4L2_SUBDEV_FL_HAS_EVENTS;
> > @@ -200,6 +217,9 @@ static int __maybe_unused
> > dw9714_vcm_suspend(struct device *dev)
> >         struct dw9714_device *dw9714_dev = sd_to_dw9714_vcm(sd);
> >         int ret, val;
> >  
> > +       if (pm_runtime_suspended(&client->dev))
> > +               return 0;
> 
> This can't take place in a runtime PM suspend callback. You'll need
> to add
> system suspend callback for this.

but this function is both the system and runtime suspend callback.
doesn't splitting up the callbacks just add lines of code
unnecessarily?

> 
> > +
> >         for (val = dw9714_dev->current_val & ~(DW9714_CTRL_STEPS -
> > 1);
> >              val >= 0; val -= DW9714_CTRL_STEPS) {
> >                 ret = dw9714_i2c_write(client,
> > @@ -208,6 +228,13 @@ static int __maybe_unused
> > dw9714_vcm_suspend(struct device *dev)
> >                         dev_err_once(dev, "%s I2C failure: %d",
> > __func__, ret);
> >                 usleep_range(DW9714_CTRL_DELAY_US,
> > DW9714_CTRL_DELAY_US + 10);
> >         }
> > +
> > +       if (dw9714_dev->vcc) {
> > +               ret = regulator_disable(dw9714_dev->vcc);
> > +               if (ret)
> > +                       dev_err(dev, "Failed to disable vcc: %d\n",
> > ret);
> > +       }
> > +
> >         return 0;
> >  }
> >  
> > @@ -224,6 +251,18 @@ static int  __maybe_unused
> > dw9714_vcm_resume(struct device *dev)
> >         struct dw9714_device *dw9714_dev = sd_to_dw9714_vcm(sd);
> >         int ret, val;
> >  
> > +       if (pm_runtime_suspended(&client->dev))
> 
> Same for this one.
> 
> > +               return 0;
> > +
> > +       if (dw9714_dev->vcc) {
> > +               ret = regulator_enable(dw9714_dev->vcc);
> > +               if (ret) {
> > +                       dev_err(dev, "Failed to enable vcc: %d\n",
> > ret);
> > +                       return ret;
> > +               }
> > +               usleep_range(1000, 2000);
> > +       }
> > +
> >         for (val = dw9714_dev->current_val % DW9714_CTRL_STEPS;
> >              val < dw9714_dev->current_val + DW9714_CTRL_STEPS - 1;
> >              val += DW9714_CTRL_STEPS) {
>