Re: [PATCH v3] input/keyboard: new OpenCores Keyboard Controller driver

[Dmitry Torokhov <dmitry.torokhov@gmail.com>]

On Tue, Sep 15, 2009 at 07:16:25AM -0400, Mike Frysinger wrote:
> On Tue, Sep 15, 2009 at 01:52, Dmitry Torokhov wrote:
> > On Mon, Sep 14, 2009 at 08:18:48PM +0200, Javier Herrero wrote:
> >> It has a bit long since last time I touched the driver, so I should also
> >> try to refresh my memory about it :). I suppose that you're right in the
> >> double allocation issue (I took another keyboard driver as a starting
> >> point and probably the double allocation was already there...), so feel
> >> free to introduce the change and I will test it as soon as I can.
> >>
> >> About the exact scancode - key mapping, the reason is that since the
> >> FPGA opencores device already implements a translation table, I found
> >> that another translation table sounded a bit redundant.
> >
> > OK, below is what I have now... One concern though - don't we need to do
> > request_mem_region/ioremap for the addr_res?
> 
> i think so ... these operations are nops on a Blackfin CPU which is
> probably why it "just works".
> 

Surely request_mem_region is not a nop? I think even if uoremap is a nop
we need to convert the driver since it does not have to be on a
Blackfin, does it?

> > +struct opencores_kbd {
> > +       struct input_dev *input;
> > +       void __iomem *addr;
> > +       int irq;
> > +       struct resource *irq_res;
> 
> the irq_res member is no longer needed
> 
> +#define NUM_KEYS 128
> > +       unsigned short keycodes[NUM_KEYS];
> 
> since this is the only usage of NUM_KEYS, could just inline it now ...
> 
> > +       error = request_irq(irq, &opencores_kbd_isr,
> > +                           IRQF_TRIGGER_RISING, pdev->name, opencores_kbd);
> > +       if (error) {
> > +               dev_err(&pdev->dev, "unable to claim irq %d\n", irq);
> > +               goto err_free_mem;
> > +       }
> > +
> > +       error = input_register_device(input);
> > +       if (error) {
> > +               dev_err(&pdev->dev, "unable to register input device\n");
> > +               goto err_free_irq;
> > +       }
> 
> the input layer can handle input even if it's not registered ?

The device can survive input events as soon as it was allocated with
input_allocate_device() but of course the event will not get anywhere.
This is a property that I intend to keep since it is very convenient.

-- 
Dmitry