[PATCH 04/11] ep93xx: Configure GPIO ports in core code

[H Hartley Sweeten]

On Friday, January 13, 2012 12:01 AM, Ryan Mallon wrote:
> On 13/01/12 17:27, Mika Westerberg wrote:
>> On Wed, Jan 11, 2012 at 02:14:29PM +1100, Ryan Mallon wrote:
>>> Move the pinmux setting of the EP93xx GPIOs to the core code. This 
>>> removes the need for the GPIO driver to have access to the system 
>>> controller registers.
>>>
>>> Cc: Hartley Sweeten <hsweeten at visionengravers.com>
>>> Cc: Mika Westerberg <mika.westerberg at iki.fi>
>>> Cc: Grant Likely <grant.likely at secretlab.ca>
>>> Signed-off-by: Ryan Mallon <rmallon at gmail.com>
>>> ---
>>>  arch/arm/mach-ep93xx/core.c |    7 +++++++
>>>  drivers/gpio/gpio-ep93xx.c  |    7 -------
>>>  2 files changed, 7 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/arch/arm/mach-ep93xx/core.c 
>>> b/arch/arm/mach-ep93xx/core.c index dcb10c2..978c5a6 100644
>>> --- a/arch/arm/mach-ep93xx/core.c
>>> +++ b/arch/arm/mach-ep93xx/core.c
>>> @@ -896,6 +896,13 @@ void __init ep93xx_init_devices(void)
>>>  	/* Disallow access to MaverickCrunch initially */
>>>  	ep93xx_devcfg_clear_bits(EP93XX_SYSCON_DEVCFG_CPENA);
>>>  
>>> +	/* Default all ports to GPIO */
>>> +	ep93xx_devcfg_set_bits(EP93XX_SYSCON_DEVCFG_KEYS |
>>> +			       EP93XX_SYSCON_DEVCFG_GONK |
>>> +			       EP93XX_SYSCON_DEVCFG_EONIDE |
>>> +			       EP93XX_SYSCON_DEVCFG_GONIDE |
>>> +			       EP93XX_SYSCON_DEVCFG_HONIDE);
>>> +
>> Doesn't this now change the behaviour? Now you unconditionally mux the 
>> pins whereas before they were only muxed if we had GPIO driver enabled.
>
> True, but I think the previous behaviour was broken anyway. If the gpio driver
> was installed after any driver which has muxed gpio pins for other purposes
> then it will override those settings. All of the ep93xx drivers will mux the pins
> they need at probe time. IIRC, before the gpio driver was moved out of
> arch/arm/mach-ep93xx we muxed all the pins unconditionally anyway.
>
> I can update the changelog to reflect the changed behaviour, but I don't think
> it will cause problems for anyone. Will see if Hartley has any comments.

Actually, there shouldn't be any change in behavior.

The gpio-ep93xx driver is always enabled in drivers/gpio/Kconfig when ARCH_EP93XX
Is enabled:

config GPIO_EP93XX
	def_bool y
	depends on ARCH_EP93XX
	select GPIO_GENERIC

Because it's a hidden Kconfig option it cannot be disabled.

Since the gpio-ep93xx driver is the first driver registered it also ends up being the
first one probed. So the gpiomux ends up getting set before any other driver is
loaded anyway.

The other drivers then setup the gpiomux as required as they load.

Regards,
Hartley