[PATCH] gpio: rewrite U300 GPIO to use gpiolib

[Linus Walleij]

On Tue, Sep 6, 2011 at 11:51 AM, Arnaud Patard
<arnaud.patard at rtp-net.org> wrote:
> [Me]
>> -#ifdef U300_COH901571_3
>> +#if defined(CONFIG_MACH_U300_BS365) || defined(CONFIG_MACH_U300_BS335)
>>       {
>>               .name  = "gpio3",
>>               .start = IRQ_U300_GPIO_PORT3,
>> @@ -236,6 +237,7 @@ static struct resource gpio_resources[] = {
>>               .end   = IRQ_U300_GPIO_PORT4,
>>               .flags = IORESOURCE_IRQ,
>>       },
>> +#endif
>
> hmm.. silly question: Now that u300 gpios can/will be added through
> gpiochip_add, is it still required to used #ifdef instead of
> machine_is_foo() or use some different platform devices for that like it
> was done on mxc ?

Yes unfortunately all of these systems share a single
MACHINE_TYPE. I will surely try to fix that and maybe
just delete unsupported old hardware, or try to make this
dynamically decided from the platform atleast.

As far as I can tell this patch does not make the situation
worse atleast :-/

>> +struct u300_gpio {
>> +     struct gpio_chip chip;
>> +     struct list_head port_list;
>> +     struct clk *clk;
>> +     struct resource *memres;
>> +     void __iomem *base;
>> +     struct device *dev;
>> +     int irq_base;
>> +     int users;
>> +     u32 stride;
>> +     /* Register offsets */
>> +     u32 pcr;
>> +     u32 dor;
>> +     u32 dir;
>> +     u32 per;
>> +     u32 icr;
>> +     u32 ien;
>> +     u32 iev;
>> +};
>
> Do you really need all theses ? I'm thinking of the 'users' field but
> there are maybe others ?

You're absolutley right about users, I've deleted it.

The rest are needed I think.

The stride+register offsets are needed to handle different
HW variants at runtime, for example.

>>  /* Initial configuration */
>> -const struct u300_gpio_configuration_data
>> -u300_gpio_config[U300_GPIO_NUM_PORTS][U300_GPIO_PINS_PER_PORT] = {
>> -#ifdef CONFIG_MACH_U300_BS335
>> +static struct __initdata u300_gpio_confdata
>> +bs335_gpio_config[BS335_GPIO_NUM_PORTS][U300_GPIO_PINS_PER_PORT] = {
>
> looks like const has been lost ?

You're right, fixed it up by re-constifying and adding a few
alterations here and there to handle the const * pointers.

Thanks a lot Arnaud!
Linus Walleij