Re: atmel_lcdfb: max pixclock check

[Nicolas Ferre <nicolas.ferre@atmel.com>]

Daniel Glöckner :
> Hi,
> in the current Atmel LCD framebuffer driver there is the following check in the
> fb_check_var callback:
> 
> if ((PICOS2KHZ(var->pixclock) * var->bits_per_pixel / 8) > clk_value_khz) {
> 	dev_err(dev, "%lu KHz pixel clock is too fast\n",
> 		PICOS2KHZ(var->pixclock));
> 	return -EINVAL;
> }
> 
> I can't find any constraint like this in the data sheets and application note.
> What I can find is a minimum for clk_value_khz/PICOS2KHZ(var->pixclock)
> depending on the display type, scan mode, and interface width.

Indeed, I have just acked a patch from Ben Nizette removing this constrain.
http://lkml.org/lkml/2009/5/12/189

> Is the quoted if-statement correct or should it be changed to a minimum clock
> divider check?

We can imagine such a check depending on display type. Patches welcome ;-)

> And while we're at it, is it correct to return -EINVAL here instead of changing
> var->pixclock to the closest supported value?

I do not know... Maybe someone on linux-fb-devel can answer ?

Bye,
-- 
Nicolas Ferre