Re: [PATCH v4 1/2] backlight: pwm_bl: Move the checks for initial power state to a separate function

[Philipp Zabel]

On Tue, 2017-03-21 at 19:48 +0100, Geert Uytterhoeven wrote:
> Hi Peter,
> 
> On Tue, Nov 22, 2016 at 2:41 PM, Peter Ujfalusi <peter.ujfalusi@xxxxxx> wrote:
> > Move the checks to select the initial state for the backlight to a new
> > function and document the checks we are doing.
> 
> This is far from a simple "move"...
> 
> > With the separate function it is going to be easier to fix or improve the
> > initial power state configuration later and it is easier to read the code.
> >
> > Signed-off-by: Peter Ujfalusi <peter.ujfalusi@xxxxxx>
> > Reviewed-by: Philipp Zabel <p.zabel@xxxxxxxxxxxxxx>
> > Reviewed-by: Thierry Reding <treding@xxxxxxxxxx>
> 
> This patch (commit 7613c922315e308a in v4.11-rc1) broke the display on
> r8a7740/armadillo.
> 
> > @@ -267,20 +292,16 @@ static int pwm_backlight_probe(struct platform_device *pdev)
> >                 pb->enable_gpio = gpio_to_desc(data->enable_gpio);
> >         }
> >
> > -       if (pb->enable_gpio) {
> > -               /*
> > -                * If the driver is probed from the device tree and there is a
> > -                * phandle link pointing to the backlight node, it is safe to
> > -                * assume that another driver will enable the backlight at the
> > -                * appropriate time. Therefore, if it is disabled, keep it so.
> > -                */
> > -               if (node && node->phandle &&
> > -                   gpiod_get_direction(pb->enable_gpio) == GPIOF_DIR_OUT &&
> > -                   gpiod_get_value(pb->enable_gpio) == 0)
> > -                       initial_blank = FB_BLANK_POWERDOWN;
> > -               else
> > -                       gpiod_direction_output(pb->enable_gpio, 1);
> 
> In my case, "node" points to the "/backlight" node, but phandle is NULL.
> Hence before, gpiod_direction_output() was called to enable the GPIO...
> 
> > -       }
> > +       /*
> > +        * If the GPIO is configured as input, change the direction to output
> > +        * and set the GPIO as active.
> > +        * Do not force the GPIO to active when it was already output as it
> > +        * could cause backlight flickering or we would enable the backlight too
> > +        * early. Leave the decision of the initial backlight state for later.
> > +        */
> > +       if (pb->enable_gpio &&
> > +           gpiod_get_direction(pb->enable_gpio) == GPIOF_DIR_IN)
> > +               gpiod_direction_output(pb->enable_gpio, 1);
> 
> ... while now it's no longer called, as gpiod_get_direction() returns
> -EINVAL.
> 
> Indeed, r8a7740_pfc does not implement the .get_direction() callback,
> so gpiod_get_direction() always returns -EINVAL, which is never equal
> to GPIOF_DIR_IN.

Oh, I didn't think about this at all, anymore. Though I believe to
remember that this was the reason that I checked for
(gpiod_get_direction(pb->enable_gpio) == GPIOF_DIR_OUT) before, so ...

> Restoring the old behavior by changing the above test to
> 
>         if (pb->enable_gpio &&
>             (!node || !node->phandle ||
>              gpiod_get_direction(pb->enable_gpio) == GPIOF_DIR_IN))
> 
> fixes the display for me, but leads to a more complex expression.
> 
> However, changing the test to
> 
>         if (pb->enable_gpio &&
>             gpiod_get_direction(pb->enable_gpio) != GPIOF_DIR_OUT)
> 
> also fixes the display, as an error is always different from GPIOF_DIR_OUT.
>
> Anyone with comments or suggestions to fix this for real?

... I'm in favor of the latter, as this is closer to the initial
intention. I'd also mention this in the comment:

       /*
        * If the GPIO is not known to be already configured as output, that is,
        * if gpiod_get_direction returns either GPIOF_DIR_IN or -EINVAL, change
        * the direction to output and set the GPIO as active.
        * Do not force the GPIO to active when it was already output as it
        * could cause backlight flickering or we would enable the backlight too
        * early. Leave the decision of the initial backlight state for later.
        */

regards
Philipp