On Monday, July 29, 2013 11:36:05 PM Felipe Balbi wrote:
> Hi,
>
> On Mon, Jul 29, 2013 at 12:59:23PM -0700, Dmitry Torokhov wrote:
> > > > > > @@ -439,12 +444,50 @@ static const struct of_device_id
> > > > > > omap_keypad_dt_match[] = {>
> > > > > >
> > > > > > MODULE_DEVICE_TABLE(of, omap_keypad_dt_match);
> > > > > > #endif
> > > > > >
> > > > > > +#ifdef CONFIG_PM_SLEEP
> > > > > > +static int omap4_keypad_suspend(struct device *dev)
> > > > > > +{
> > > > > > + struct platform_device *pdev = to_platform_device(dev);
> > > > >
> > > > > you don't need to access the platform_device...
> > > > >
> > > > > > + struct omap4_keypad *keypad_data = platform_get_drvdata(pdev);
> > > > >
> > > > > ... since this can become:
> > > > > struct omap4_keypad *keypad_data = dev_get_drvdata(dev);
> > > >
> > > > No, please use correct accessors for the objects. Platform drivers
> > > > deal
> > > > with platform devices and I prefer using platform_get_drvdata() on
> > > > them.
> > >
> > > The argument to this function is a struct device, you prefer to do some
> > > pointer math to find the containing pdev, then deref that back to dev,
> > > then to struct device_private and further to driver_data ?
> > >
> > > Sounds like a waste of time IMHO. You already have the device pointer
> > > anyway, why would you go through the trouble of calculating the
> > > offsets for the containing struct platform_device ?
> >
> > This assumes knowledge of dev_get_drvdata() implementation and assumption
> > that it will stay the same. Unless I hear from device core guys that
> > <bus>_{get|set}_drvdata() methods are obsolete and will be eventually
> > removed I will require proper accessors being used.
>
> they're not obsolete and will never be removed. They're nothing but
> helpers though. Instead of calling:
>
> dev_set_drvdata(&pdev->dev);
>
> you call:
>
> platform_set_drvdata(pdev);
>
> same is valid for every single bus, but in the end they all just wrap a
> call dev_{set,get}_drvdata() internally. If you already have a struct
> device pointer as argument, why waste cycles doing pointer math just to
> go back to the same struct device pointer on the next call ?
Because I do not want to rely on the fact that what my driver set up
with platform_set_drvdata(pdev, XXX) is the same as what dev_get_drvdata()
will return *in the current implementation*. Software layers and all
that...
--
Dmitry
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
