Re: [PATCH] asus-laptop: correct error handling in asus_read_brightness()

[Darren Hart]

On Fri, Apr 22, 2016 at 10:43:30AM +0300, Giedrius StatkeviÄius wrote:
> On Fri, Apr 22, 2016 at 02:09:22AM +0300, Andy Shevchenko wrote:
> > On Sat, Apr 16, 2016 at 3:27 AM, Giedrius StatkeviÄius
> > <giedrius.statkevicius@xxxxxxxxx> wrote:
> > > It is possible that acpi_evaluate_integer might fail and value would not be
> > > set to any value so correct this defect by returning 0 in case of an
> > > error. This is also the correct thing to return because the backlight
> > > subsystem will print the old value of brightness in this case.
> > >
> > > Signed-off-by: Giedrius StatkeviÄius <giedrius.statkevicius@xxxxxxxxx>
> > > ---
> > >  drivers/platform/x86/asus-laptop.c | 4 +++-
> > >  1 file changed, 3 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/platform/x86/asus-laptop.c b/drivers/platform/x86/asus-laptop.c
> > > index 9a69734..15f1311 100644
> > > --- a/drivers/platform/x86/asus-laptop.c
> > > +++ b/drivers/platform/x86/asus-laptop.c
> > > @@ -775,8 +775,10 @@ static int asus_read_brightness(struct backlight_device *bd)
> > >
> > >         rv = acpi_evaluate_integer(asus->handle, METHOD_BRIGHTNESS_GET,
> > >                                    NULL, &value);
> > > -       if (ACPI_FAILURE(rv))
> > > +       if (ACPI_FAILURE(rv)) {
> > >                 pr_warn("Error reading brightness\n");
> > > +               return 0;
> > > +       }
> > 
> > This looks like a workaround.
> > I suppose the real fix is to return here an error code and fix all callers, like
> > drivers/video/backlight/backlight.c.
> > 
> 
> It just fixes the behaviour according to the current code in that file. I
> suppose that would be nice but I don't think it would make any difference
> because the backlight core code still prints out ->props.brightness in case
> ops->get_brightness fails. Just the difference would be that now actual error
> messages are printed in the drivers themselves instead of generic messages from
> the backlight core. Anyway, I think the current behaviour is more useful because
> the drivers know better about what has failed.
> 

As a matter of practice, based on feedback from the previous pdx86 maintainers,
I prefer to accept patches which fix an immediate issue and leave the rework or
broader work as a follow-on, or a TODO item.

We should look for a good place to document these TODOs.

Queued for 4.7.

Thanks,

-- 
Darren Hart
Intel Open Source Technology Center