Re: [RFC] rfkill - Add support for input key to control wireless radio

[Ivo van Doorn <ivdoorn@gmail.com>]

Hi,

> > +	/*
> > +	 * Pointer to rfkill structure
> > +	 * that was filled in by key driver.
> > +	 */
> > +	struct rfkill *rfkill;
> 
> Since rfkill is basically a function pointer table,
> can it be made const?

Sounds good to me.

> > +	/*
> > +	 * Once key status change has been detected, the toggled
> > +	 * field should be set to indicate a notification to
> > +	 * user or driver should be performed.
> > +	 */
> > +	int toggled;
> > +
> > +	/*
> > +	 * Current state of the device radio, this state will
> > +	 * change after the radio has actually been toggled since
> > +	 * receiving the radio key event.
> > +	 */
> > +	int radio_status;
> > +
> > +	/*
> > +	 * Current status of the key which controls the radio,
> > +	 * this value will change after the key state has changed
> > +	 * after polling, or the key driver has send the new state
> > +	 * manually.
> > +	 */
> > +	int key_status;
> 
> 
> Maybe turn these bits into a bit values (set_bit/clear_bit) in an unsigned long.

Will do.

> > +	/*
> > +	 * Input device for this key,
> > +	 * we also keep track of the number of
> > +	 * times this input device is open. This
> > +	 * is important for determining to whom we
> > +	 * should report key events.
> > +	 */
> > +	struct input_dev *input;
> > +	unsigned int open_count;
> 
> atomic on open_count?

There seems to have gone something wrong with the patch,
latest version should have had this field removed.

> > +	/*
> > +	 * Name of this radio type.
> > +	 */
> > +	char *name;
> 
> const?

Will do.

> > +	/*
> > +	 * All access to the master structure
> > +	 * and its children (the keys) are protected
> > +	 * by this key lock.
> > +	 */
> > +	struct semaphore key_sem;
> 
> mutex instead of semaphort

Strange, this should have already be fixed. :S

> > +	/*
> > +	 * Work structures for periodic polling,
> > +	 * as well as the scheduled radio toggling.
> > +	 */
> > +	struct work_struct toggle_work;
> > +	struct work_struct poll_work;
> 
> delayed_rearming_work instead?

Same here, rfkill should already have the new workqueue api...

I'll resubmit this within a few moments.

Thanks

Ivo