Re: [PATCH 9/9] OMAP:GPIO: Implement GPIO as a platform device

[Kevin Hilman <khilman@deeprootsystems.com>]

"Varadarajan, Charulatha" <charu@ti.com> writes:

>> 
>> >> -----Original Message-----
>> >> From: Kevin Hilman [mailto:khilman@deeprootsystems.com]
>> >> Sent: Saturday, May 01, 2010 5:34 AM
>> >> To: Varadarajan, Charulatha
>> >> Cc: linux-omap@vger.kernel.org; Nayak, Rajendra; paul@pwsan.com;
>> tony@atomide.com
>> >> Subject: Re: [PATCH 9/9] OMAP:GPIO: Implement GPIO as a platform device
>> >>
> [...]
>> >> >
>> >> > +static int init_gpio_info(void)
>> >> > +{
>> >> > +   gpio_bank_bits = 32;
>> >> > +
>> >> > +   if (cpu_is_omap15xx()) {
>> >> > +           gpio_bank_count = 2;
>> >> > +           gpio_bank_bits = 16;
>> >> > +   } else if (cpu_is_omap16xx()) {
>> >> > +           gpio_bank_count = 5;
>> >> > +           gpio_bank_bits = 16;
>> >> > +   } else if (cpu_is_omap7xx())
>> >> > +           gpio_bank_count = 7;
>> >> > +   else if (cpu_is_omap242x())
>> >> > +           gpio_bank_count = 4;
>> >> > +   else if (cpu_is_omap243x())
>> >> > +           gpio_bank_count = 5;
>> >> > +   else if (cpu_is_omap34xx() || cpu_is_omap44xx())
>> >> > +           gpio_bank_count = OMAP34XX_NR_GPIOS;
>> >>
>> >> Both the bank count and bank bits could be part of platform_data
>> >> and set in the SoC specific init.  This is the GPIO driver part
>> >> and we're trying to make this as SoC independent as possible.  Anytime
>> >> you need to add a cpu_is* or #ifdef in this code indicates something
>> >> that should be part of SoC specific init and passed in.
>> >
>> > bank count and bank bits are not specific to each device, but SoC specific.
>> > Hence I did not consider passing it as part of platform_data, because this
>> information would be duplicated across all devices in that SoC.
>> >
>> > It would be good if we have a way to pass the SoC specific data which is
>> > common for all the devices rather than duplicating them by sending them via
>> > platform_data.
>> >
>> > Anyways, I can move it to platform_data if there is no other way.
>> 
>> Probably the right place for the SoC specifics is attached to the SoC
>> specific hwmod using the dev_attr pointer.  That struct can then
>> be passed in via platform_data.
>> 
>> You can see how Thara did this for SmartReflex as an example.
>> 
>
> Using dev_attr for OMAP2PLUS is fine. How about OMAP1? 
> 1. To be implemented in a uniform way, may I get this info as 
> part of platform_data? 
> 2. Use dev_attr for OMAP2PLUS and use platform_data for OMAP1? 

Not sure I understand the question.

You should be able to use a dev_attr struct in OMAP1 and OMAP2+, as
there is technically nothing hwmod specific in the dev_attr.  It
is attached to the hwmod as a 'void *' and should be passed into the driver
as the platform_data.

Kevin