Messages in this thread Patch in this message | | | Date | Fri, 24 May 2013 13:27:51 -0700 | From | Andrew Morton <> | Subject | Re: [PATCH 1/2] backlight: add devm_backlight_device_{register,unregister}() |
| |
On Thu, 16 May 2013 05:28:26 +0000 (GMT) Jingoo Han <jg1.han@samsung.com> wrote:
> These functions allow the driver core to automatically clean up any > allocation made by backlight drivers. Thus it simplifies the error > paths. > > ... > > --- a/drivers/video/backlight/backlight.c > +++ b/drivers/video/backlight/backlight.c > @@ -370,6 +370,81 @@ void backlight_device_unregister(struct backlight_device *bd) > } > EXPORT_SYMBOL(backlight_device_unregister); > > +static void devm_backlight_device_release(struct device *dev, void *res) > +{ > + struct backlight_device *backlight = *(struct backlight_device **)res; > + > + backlight_device_unregister(backlight); > +} > > ... > > +struct backlight_device *devm_backlight_device_register(struct device *dev, > + const char *name, struct device *parent, void *devdata, > + const struct backlight_ops *ops, > + const struct backlight_properties *props) > +{ > + struct backlight_device **ptr, *backlight; > + > + ptr = devres_alloc(devm_backlight_device_release, sizeof(*ptr), > + GFP_KERNEL);
Well this is an awkward-looking thing. We allocate a silly little 4- or 8-byte object just to track the backlight_device* via the devres system.
I assume this was done so that devres could be used as a wrapper around the existing code?
Really it would be better if things like backlight_device_register() were themselves directly converted to use devm_foo(), yes?
| |