Here is a patch to the asus_acpi driver that links the Asus laptop LED's
into the new LED subsystem.  It creates LED class devices named
asus:mail, asus:wireless, and asus:touchpad, depending on if the laptop
supports the mled, wled, and tled LED's.

Since it's so new, I added a config option to turn it on and off.  It's
worked for me, though I have an Asus M2N, which only has the mail and
wireless LED's.

Signed-off-by: Thomas Tuttle <thinkinginbinary@gmail.com>


I believe I've fixed everything you asked about, plus a few things
Richard Purdie (the LED subsystem guy) suggested.

On July 06 at 18:49 EDT, Andrew Morton hastily scribbled:
> Thomas Tuttle <thinkinginbinary@gmail.com> wrote:
> > +#ifdef CONFIG_ACPI_ASUS_NEW_LED
> > +#include <linux/leds.h>
> > +#endif
> The ifdef shouldn't be required.  If it is, ldes.h needs fixing.
Fixed.

> This mingles declarations with definitions.  It's more conventional to keep
> them together.
Fixed.  I moved all the declarations together, and initialized the
led_cdev structs if and when they are going to be registered.  (Just as
with the zero initializers, this should stop them from wasting space in
.vmlinux.)

> Please don' t initialise static storage to zero (the "= 0").  The C runtime
> environment does that already, and the above construct will place the
> values into .data and hence into .vmlinux.
Fixed.

> > +/* LED update work structs. */
> These all should be declared static.
Fixed.

> > +/* LED work functions. */
> > +static void led_update_mled(void *private) {
> The opening brace goes in column 1.
Fixed.

> It's usual to put a blank line at the end of the declaration of the
> automatic valriables.
Fixed.

> Add:
> 
> #else
> static inline int led_initialize(struct device *parent)
> {
> }
> 
> static inline void led_terminate(void)
> {
> }
> #endif
Fixed.

> > +#ifdef CONFIG_ACPI_ASUS_NEW_LED
> > +        result = led_initialize(acpi_get_physical_device(device->handle));
> > +#endif
> > +
> >  
> > +#ifdef CONFIG_ACPI_ASUS_NEW_LED
> > +        led_terminate();
> > +#endif          
> > +
> And remove these ifdefs.
Fixed.

I believe it's a little less ugly now.  The declarations are better
organized, error recovery from the initialization is better (sorry to
use goto, but if I didn't, I'd have to duplicate the abort code for each
possible point of failure).

This revision is essentially the same, functionally, and it appears to
work in quick testing.  I think it would be nice if someone else with an
Asus laptop could give this a test before it goes into mainline.  The
code that updates the LED's is basically copied from the /proc
interface, but testing is good.

-- Thomas Tuttle