Re: [PATCH] mfd: ab8500: update kconfig for ab8500 core driver

[Randy Dunlap <randy.dunlap@oracle.com>]

On 09/15/10 11:21, Samuel Ortiz wrote:
> Hi Randy,
> 
> On Wed, Sep 15, 2010 at 10:41:05AM -0700, Randy Dunlap wrote:
>> I don't know what the __crc_* symbols are (I can't find them anywhere).
>>
>> The other functions (pwm_config, pwm_free, pwm_request, pwm_disable,
>> pwm_enable) exist in multiple places.  This is not good.
>> They are very generically named.  The instances of these that are
>> provided by platform code are OK (these):
>>
>> ./include/linux/pwm.h:19:int pwm_config(struct pwm_device *pwm, int duty_ns, int period_ns);
>> ./arch/arm/plat-pxa/pwm.c:64:int pwm_config(struct pwm_device *pwm, int duty_ns, int period_ns)
>> ./arch/arm/plat-pxa/pwm.c:101:EXPORT_SYMBOL(pwm_config);
>> ./arch/arm/plat-samsung/pwm.c:194:int pwm_config(struct pwm_device *pwm, int duty_ns, int period_ns)
>> ./arch/arm/plat-samsung/pwm.c:281:EXPORT_SYMBOL(pwm_config);
>> ./arch/arm/plat-mxc/pwm.c:55:int pwm_config(struct pwm_device *pwm, int duty_ns, int period_ns)
>> ./arch/arm/plat-mxc/pwm.c:114:EXPORT_SYMBOL(pwm_config);
>> ./arch/mips/jz4740/pwm.c:94:int pwm_config(struct pwm_device *pwm, int duty_ns, int period_ns)
>>
>> but the instances of these that are defined in
>> drivers/mfd/twl6030-pwm.c should not be named so generically.
>>
>> Changing (Fixing) the function names in twl6030-pwm.c should fix the build problem
>> that you reported, I think.  
> The pwm "API" relies on someone providing the pwm_* symbols, and then you have
> leds_pwm or backlight_pwm calling those symbols out of the blue.
> If we change the twl6030-pwm.c function names, no pwm users (backlight, led or
> input) will actually be able to use the twl6030 PWM driver.
> The pwm API is very poorly designed in my opinion and should provide a way for
> pwm drivers to register against it. pwm users will then call into the pwm
> framework who would select which driver to use.

So twl6030-pwm is a provider, just like some of the arch/ providers?
OK, I couldn't see that.

>> And it will still allow the ab8500 driver to be
>> built on other platforms, which is what we prefer when that is possible.
> I agree with that, but couldnt see any other fix with the current pwm
> situation.

OK, thanks.  Do whatever you have to do.

-- 
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***