Re: [PATCH 2/2] mfd: twl4030: add twl4030-pwrbutton as our child

[David Brownell <david-b@pacbell.net>]

On Friday 27 February 2009, Andrew Morton wrote:
> > --- a/drivers/mfd/twl4030-core.c
> > +++ b/drivers/mfd/twl4030-core.c
> > @@ -101,6 +101,12 @@
> >  #define twl_has_usb()        false
> >  #endif
> >  
> > +#if defined(CONFIG_INPUT_TWL4030_PWRBUTTON) \
> > +     || defined(CONFIG_INPUT_TWL4030_PWBUTTON_MODULE)
> 
> OK, this is "wrong".  The core shouldn't need to know about specific
> clients.

This is a pretty standard idiom:  only create the device
nodes a system actually uses.  Applied comprehensively,
the kernel footprint shrinks ... supporting a device can
require a lot of ancillary infrastructure, which may not
need to be compiled in.


> > +#define twl_has_pwrbutton()  true
> > +#else
> > +#define twl_has_pwrbutton()  false
> > +#endif
> >  
> >  /* Triton Core internal information (BEGIN) */
> >  
> > @@ -526,6 +532,13 @@ add_children(struct twl4030_platform_data *pdata, unsigned long features)
> >               usb_transceiver = child;
> >       }
> >  
> > +     if (twl_has_pwrbutton()) {
> > +             child = add_child(1, "twl4030_pwrbutton",
> > +                             NULL, 0, true, pdata->irq_base + 8 + 0, 0);
> > +             if (IS_ERR(child))
> > +                     return PTR_ERR(child);
> > +     }
> > +
> >       if (twl_has_regulator()) {
> >               /*
> >               child = add_regulator(TWL4030_REG_VPLL1, pdata->vpll1);
> 
> The client module should register itself with the core, rather than the core
> registering the client.
> 
> What has gone wrong here?

Not much I can see.  It's registering a platform_device,
but only if it could be used on this system.  Quite a lot
of OMAP3 boards don't hook up this power button.

Maybe it should also verify that *this* board supports
a power button.  A boolean flag in the twl4030 platform
data would suffice, for multi-board kernels.

- Dave