Re: [PATCH] [media] ov2640: make GPIOLIB an optional dependency

[Sakari Ailus <sakari.ailus@iki.fi>]

Hi Mauro,

On Mon, Apr 24, 2017 at 11:05:29PM -0300, Mauro Carvalho Chehab wrote:
> Em Mon, 24 Apr 2017 20:38:47 +0300
> Sakari Ailus <sakari.ailus@iki.fi> escreveu:
> 
> > Hi Mauro,
> > 
> > On Mon, Apr 24, 2017 at 12:50:36PM -0300, Mauro Carvalho Chehab wrote:
> > > Em Mon, 24 Apr 2017 17:44:02 +0300
> > > Sakari Ailus <sakari.ailus@iki.fi> escreveu:
> > >   
> > > > Hi Mauro and others,
> > > > 
> > > > On Fri, Apr 21, 2017 at 11:39:42AM -0300, Mauro Carvalho Chehab wrote:  
> > > > > Em Fri, 21 Apr 2017 08:33:12 +0200
> > > > > Pavel Machek <pavel@ucw.cz> escreveu:
> > > > >     
> > > > > > Hi!
> > > > > >     
> > > > > > > > Better solution would be for VIDEO_EM28XX_V4L2 to depend on GPIOLIB,
> > > > > > > > too, no? If not, should there be BUG_ON(priv->pwdn_gpio);
> > > > > > > > BUG_ON(priv->resetb_gpio);?      
> > > > > > > 
> > > > > > > Pavel,
> > > > > > > 
> > > > > > > The em28xx driver was added upstream several years the gpio driver. 
> > > > > > > It controls GPIO using a different logic. It makes no sense to make
> > > > > > > it dependent on GPIOLIB, except if someone converts it to use it.      
> > > > > > 
> > > > > > At least comment in the sourcecode...? Remove pwdn_gpio fields from
> > > > > > structure in !GPIOLIB case, because otherwise they are trap for the
> > > > > > programmer trying to understand what is going on?    
> > > > > 
> > > > > 
> > > > > Sorry, I answered to another e-mail thread related to it. I assumed
> > > > > that it was c/c to linux-media, but it is, in fact a private e-mail.    
> > > > 
> > > > I thought so, too! X-)
> > > >   
> > > > > 
> > > > > I can see two alternatives:
> > > > > 
> > > > > 1) Restore old behavior, assuming that all drivers that use OV2640 will
> > > > > have GPIOLIB enabled, with a patch like:
> > > > > 
> > > > > diff --git a/drivers/media/i2c/Kconfig b/drivers/media/i2c/Kconfig
> > > > > index fd181c99ce11..4e834c36f7da 100644
> > > > > --- a/drivers/media/i2c/Kconfig
> > > > > +++ b/drivers/media/i2c/Kconfig
> > > > > @@ -521,6 +521,7 @@ config VIDEO_OV2640
> > > > >         tristate "OmniVision OV2640 sensor support"
> > > > >         depends on VIDEO_V4L2 && I2C
> > > > >         depends on MEDIA_CAMERA_SUPPORT
> > > > > +       depends on GPIOLIB if OF
> > > > >         help
> > > > >           This is a Video4Linux2 sensor-level driver for the OmniVision
> > > > >           OV2640 camera.
> > > > > 
> > > > > However, I was told that some OF drivers don't actually define the GPIO
> > > > > pins.
> > > > > 
> > > > > So, the other option is:
> > > > > 
> > > > > 2) Make the logic smarter for OF, with this change:
> > > > > 
> > > > > 
> > > > > diff --git a/drivers/media/i2c/ov2640.c b/drivers/media/i2c/ov2640.c
> > > > > index 4a2ae24f8722..8855c81a9e1f 100644
> > > > > --- a/drivers/media/i2c/ov2640.c
> > > > > +++ b/drivers/media/i2c/ov2640.c
> > > > > @@ -1048,21 +1048,39 @@ static const struct v4l2_subdev_ops ov2640_subdev_ops = {
> > > > >  static int ov2640_probe_dt(struct i2c_client *client,
> > > > >  		struct ov2640_priv *priv)
> > > > >  {
> > > > > +	int ret;
> > > > > +
> > > > >  	/* Request the reset GPIO deasserted */
> > > > >  	priv->resetb_gpio = devm_gpiod_get_optional(&client->dev, "resetb",
> > > > >  			GPIOD_OUT_LOW);
> > > > > -	if (!priv->resetb_gpio)
> > > > > +	if (!priv->resetb_gpio) {
> > > > >  		dev_dbg(&client->dev, "resetb gpio is not assigned!\n");
> > > > > -	else if (IS_ERR(priv->resetb_gpio))
> > > > > -		return PTR_ERR(priv->resetb_gpio);
> > > > > +	} else {
> > > > > +		ret = PTR_ERR(priv->resetb_gpio);
> > > > > +
> > > > > +		if (ret && ret != -ENOSYS) {
> > > > > +			dev_dbg(&client->dev,
> > > > > +				"Error %d while getting resetb gpio\n",
> > > > > +				ret);
> > > > > +			return ret;
> > > > > +		}
> > > > > +	}    
> > > > 
> > > > This would work. I just wish it'd look nicer. :-)
> > > > 
> > > > How about something like:
> > > > 
> > > > ret = PTR_ERR(priv->reset_gpio);
> > > > if (!priv->reset_gpio) {
> > > > 	dev_dbg("reset gpio is not assigned");
> > > > } else if (ret != -ENOSYS) {
> > > > 	dev_dbg("error %d while getting reset gpio, ret);  
> > > 
> > > It still need to test if ret == 0, as otherwise it will do the wrong thing
> > > here.   
> > 
> > ret won't be zero here, that was checked above. You could check for just ret
> > though, it'd be easier to read that way.
> > 
> > > 
> > > 
> > > I guess the patch below does the trick and it is not ugly :-)
> > > 
> > >   
> > > > 	return ret;
> > > > }
> > > > 
> > > > I prefer this option as it's not dependent on the system firmware type.  
> > > 
> > > Thanks,
> > > Mauro
> > > 
> > > [PATCH] ov2640: print error if devm_*_optional*() fails
> > > 
> > > devm_gpiod_get_optional() can return -ENOSYS if GPIOLIB is
> > > disabled, causing probe to fail. Warn the user if this
> > > happens.
> > > 
> > > Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
> > > 
> > > diff --git a/drivers/media/i2c/ov2640.c b/drivers/media/i2c/ov2640.c
> > > index 4a2ae24f8722..8c87ed8306ea 100644
> > > --- a/drivers/media/i2c/ov2640.c
> > > +++ b/drivers/media/i2c/ov2640.c
> > > @@ -765,17 +765,17 @@ static int ov2640_s_register(struct v4l2_subdev *sd,
> > >  
> > >  static int ov2640_s_power(struct v4l2_subdev *sd, int on)
> > >  {
> > > +#ifdef CONFIG_GPIOLIB
> > >  	struct i2c_client *client = v4l2_get_subdevdata(sd);
> > >  	struct ov2640_priv *priv = to_ov2640(client);
> > >  
> > > -#ifdef CONFIG_GPIOLIB
> > >  	if (priv->pwdn_gpio)
> > >  		gpiod_direction_output(priv->pwdn_gpio, !on);
> > >  	if (on && priv->resetb_gpio) {
> > >  		/* Active the resetb pin to perform a reset pulse */
> > >  		gpiod_direction_output(priv->resetb_gpio, 1);
> > >  		usleep_range(3000, 5000);
> > > -		gpiod_direction_output(priv->resetb_gpio, 0);
> > > +		gpiod_set_value(priv->resetb_gpio, 0);
> > >  	}
> > >  #endif
> > >  	return 0;
> > > @@ -1048,21 +1048,35 @@ static const struct v4l2_subdev_ops ov2640_subdev_ops = {
> > >  static int ov2640_probe_dt(struct i2c_client *client,
> > >  		struct ov2640_priv *priv)
> > >  {
> > > +	int ret;
> > > +
> > >  	/* Request the reset GPIO deasserted */
> > >  	priv->resetb_gpio = devm_gpiod_get_optional(&client->dev, "resetb",
> > >  			GPIOD_OUT_LOW);
> > > +
> > >  	if (!priv->resetb_gpio)
> > >  		dev_dbg(&client->dev, "resetb gpio is not assigned!\n");
> > > -	else if (IS_ERR(priv->resetb_gpio))
> > > -		return PTR_ERR(priv->resetb_gpio);
> > > +
> > > +	ret = PTR_ERR_OR_ZERO(priv->pwdn_gpio);  
> > 
> > s/pwdn/resetb/
> 
> True.
> 
> > 
> > This works, but you're essentially checking the same thing twice. ret is
> > only zero iff priv->resetb_gpio is NULL.
> 
> No. That's the implementation for PTR_ERR_OR_ZERO:
> 
> static inline int __must_check PTR_ERR_OR_ZERO(__force const void *ptr)
> {
> 	if (IS_ERR(ptr))
> 		return PTR_ERR(ptr);
> 	else
> 		return 0;
> }
> 
> ret is zero if either priv->resetb_gpio is NULL or if it is a valid
> pointer.

Right. Indeed.

> 
> That is about the same as what you proposed on this code snippet:

Yeah, either should be good. With the above addressed,

Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>

> > How about:
> > 
> > ret = PTR_ERR(priv->reset_gpio);
> > if (!priv->reset_gpio) {
> > 	dev_dbg("reset gpio is not assigned\n");
> > } else if (IS_ERR(priv->reset_gpio) && ret != -ENOSYS) {
> > 	dev_dbg("error %d while getting reset gpio", ret);
> > 	return ret;
> > }
> > 
> 
> 

-- 
Regards,

Sakari Ailus
e-mail: sakari.ailus@iki.fi	XMPP: sailus@retiisi.org.uk