Re: [PATCH v5] gpio: driver for PrimeCell PL061 GPIO controller

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

On Monday 15 June 2009, Baruch Siach wrote:
>
> > > +static unsigned int pl061_irq_startup(unsigned irq)
> > > +{
> > > +       if (gpio_request(irq_to_gpio(irq), "IRQ") == 0)
> > > +               pr_warning("%s: warning: GPIO%d has not been requested\n",
> > > +                               __func__, irq_to_gpio(irq));
> > 
> > No, that's the responsibility of whoever is setting this up.
> 
> The gpio_request() here is only for warning. It is expected to fail. Should I 
> remove it? Shouldn't there be an is_gpio_requested() as part of gpiolib?

As I said elsewhere:  yes, remove it; and no, such a call
would be pointless since it's conceptually broken in threaded
environments, like inside the Linux kernel.


> > Normally, platform setup code would request the GPIO since it's
> > being dedicated to some task on the board ... and then pass the
> > gpio_to_irq() value as a device resource, or maybe the GPIO itself
> > (if the driver had to deal with GPIOs per se).
> > 
> > Also, irq_to_gpio() can be problematic in some systems; avoid it.
> 
> I use irq_to_gpio() extensively to implement the enable/disable etc. methods 
> of irq_chip. Is there a way to implement them without irq_to_gpio()?

If it works on your systems, so be it ... it's highly
system-specific whether it works.  But this driver is
not supposed to be system-specific, and shouldn't use it.
(And it doesn't need to use it, as noted.)
  

> > > +
> > > +       pl061_irq_enable(irq);
> > > +
> > > +       return 0;
> > > +}
> > 
> > Remove that gpio_request(), and the irq_to_gpio(), and you get:
> > 
> > Acked-by: David Brownell <dbrownell@users.sourceforge.net>
> 
> Thanks,
> 
> Baruch
> 
> -- 
>                                                      ~. .~   Tk Open Systems
> =}------------------------------------------------ooO--U--Ooo------------{=
>    - baruch@tkos.co.il - tel: +972.2.679.5364, http://www.tkos.co.il -
> 
>