private field in led_classdev?

private field in led_classdev?

[Samuel Thibault]

Hello,

Would there be an objection to adding a void *private field to the
led_classdev struct?  It would be a lot less burden for the input
key board leds support for finding out the keyboard device from the
brightness_set callback, as each device may have a long series of leds,
not just a fixed number of them.

Samuel

Re: private field in led_classdev?

[Greg KH]

On Sun, Mar 07, 2010 at 03:31:12PM +0100, Samuel Thibault wrote:
> Hello,
> 
> Would there be an objection to adding a void *private field to the
> led_classdev struct?  It would be a lot less burden for the input
> key board leds support for finding out the keyboard device from the
> brightness_set callback, as each device may have a long series of leds,
> not just a fixed number of them.

What's wrong with using the private pointer in the struct device
instead?  That's what it is there for.

thanks,

greg k-h

Re: private field in led_classdev?

[Samuel Thibault]

Greg KH, le Tue 09 Mar 2010 18:54:11 -0800, a écrit :
> On Sun, Mar 07, 2010 at 03:31:12PM +0100, Samuel Thibault wrote:
> > Would there be an objection to adding a void *private field to the
> > led_classdev struct?  It would be a lot less burden for the input
> > key board leds support for finding out the keyboard device from the
> > brightness_set callback, as each device may have a long series of leds,
> > not just a fixed number of them.
> 
> What's wrong with using the private pointer in the struct device
> instead?

Ah, right. In my case there is a struct device corresponding to the
actual input device. I hadn't realized that since registering a led
creates a new device, I have another private pointer which I can use.

Samuel

Re: private field in led_classdev?

[Samuel Thibault]

Samuel Thibault, le Wed 10 Mar 2010 10:44:52 +0100, a écrit :
> Greg KH, le Tue 09 Mar 2010 18:54:11 -0800, a écrit :
> > On Sun, Mar 07, 2010 at 03:31:12PM +0100, Samuel Thibault wrote:
> > > Would there be an objection to adding a void *private field to the
> > > led_classdev struct?  It would be a lot less burden for the input
> > > key board leds support for finding out the keyboard device from the
> > > brightness_set callback, as each device may have a long series of leds,
> > > not just a fixed number of them.
> > 
> > What's wrong with using the private pointer in the struct device
> > instead?
> 
> Ah, right. In my case there is a struct device corresponding to the
> actual input device. I hadn't realized that since registering a led
> creates a new device, I have another private pointer which I can use.

One issue, however, is that led_classdev_register() calls
led_trigger_set_default(), which eventually calls the brightness_set
hook, but the private field (platform_data) still hasn't been
initialized, so the brightness_set hook has to abort setting the initial
state. In my case, after led_classdev_register and setting
platform_data, I can call the low level function by hand, it's just not
very pretty...

Samuel

Re: private field in led_classdev?

[Richard Purdie]

On Wed, 2010-03-10 at 12:01 +0100, Samuel Thibault wrote:
> Samuel Thibault, le Wed 10 Mar 2010 10:44:52 +0100, a écrit :
> > Greg KH, le Tue 09 Mar 2010 18:54:11 -0800, a écrit :
> > > What's wrong with using the private pointer in the struct device
> > > instead?
> > 
> > Ah, right. In my case there is a struct device corresponding to the
> > actual input device. I hadn't realized that since registering a led
> > creates a new device, I have another private pointer which I can use.
> 
> One issue, however, is that led_classdev_register() calls
> led_trigger_set_default(), which eventually calls the brightness_set
> hook, but the private field (platform_data) still hasn't been
> initialized, so the brightness_set hook has to abort setting the initial
> state. In my case, after led_classdev_register and setting
> platform_data, I can call the low level function by hand, it's just not
> very pretty...

That is a bit ugly :/.

It would probably make sense to pass that parameter to
led_classdev_register() but changing references to that function
everywhere will be a pain.

I'm tempted to make led_classdev_register() a wrapper around a
led_classdev_register_drvdata() function which takes an extra parameter.

I agree using the private pointer makes sense though.

Cheers,

Richard