[PATCH] pxa: extend gpio pins from 128 to 256

[Haojian Zhuang]

On Sun, Oct 18, 2009 at 2:03 AM, Eric Miao <eric.y.miao at gmail.com> wrote:
> On Sun, Oct 18, 2009 at 1:01 PM, Haojian Zhuang
> <haojian.zhuang at gmail.com> wrote:
>> On Thu, Oct 15, 2009 at 7:01 AM, Eric Miao <eric.y.miao at gmail.com> wrote:
>>> Haojian,
>>>
>>> Sorry for late reply. See my comments below:
>>>
>>>> Subject: [PATCH] pxa: extend gpio pins from 128 to 256
>>>>
>>>> GPIO pins exceeds 128 pins. So extend the maximum GPIO pins to 256.
>>>> And extend MFP pins also.
>>>>
>>>> Signed-off-by: Haojian Zhuang <haojian.zhuang at marvell.com>
>>>> ---
>>>> diff --git a/arch/arm/mach-pxa/mfp-pxa2xx.c b/arch/arm/mach-pxa/mfp-pxa2xx.c
>>>> index cf6b720..767db92 100644
>>>> --- a/arch/arm/mach-pxa/mfp-pxa2xx.c
>>>> +++ b/arch/arm/mach-pxa/mfp-pxa2xx.c
>>>> @@ -41,7 +41,7 @@ struct gpio_desc {
>>>>      unsigned long    config;
>>>>  };
>>>>
>>>> -static struct gpio_desc gpio_desc[MFP_PIN_GPIO127 + 1];
>>>> +static struct gpio_desc gpio_desc[MFP_PIN_GPIO255 + 1];
>>>>
>>>
>>> I'm not sure if this is really necessary since mfp-pxa2xx.c is pxa2xx specific,
>>> and extending the array by double means wasting of space?
>>>
>>>>  static unsigned long gpdr_lpm[4];
>>>>
>>>> @@ -117,7 +117,7 @@ static inline int __mfp_validate(int mfp)
>>>>  {
>>>>      int gpio = mfp_to_gpio(mfp);
>>>>
>>>> -    if ((mfp > MFP_PIN_GPIO127) || !gpio_desc[gpio].valid) {
>>>> +    if ((mfp > MFP_PIN_GPIO255) || !gpio_desc[gpio].valid) {
>>>>          pr_warning("%s: GPIO%d is invalid pin\n", __func__, gpio);
>>>>          return -1;
>>>>      }
>>>
>>> Same here, pxa2xx specific.
>>>
>>
>> Up to now, mfp driver is shared between pxa2xx, pxa3xx and so on. It
>> seems that extending mfp io is easier.
>>
>
> Actually, they are not. There are two separate mfp drivers for pxa2xx
> and pxa3xx if you look closely. And mfp-pxa2xx.c is used by pxa2xx
> solely, and the special binding of gpio_desc[] with pxa2xx MFP is
> very specific to pxa2xx only. Please help double check.
>

I think that mfp_to_gpio() macro is also a part of mfp driver. This
macro is used widely
in platform driver in order to allocate platform irq resource.
NR_BUILTIN_GPIO is similar
to this macro.

Because they're used in structure initialization, we can't use
detecting CPU method to
distinguish how much mfp/gpio pins can be applied on current CPU.
Since we couldn't
import variable or function into structure intialization code. Current
patch is the easiest
modification.

If there's a better solution, I'm glad to adopt it. :)

Thanks
Haojian