On Thu, Oct 8, 2015 at 12:43 PM, Simon Glass sjg@chromium.org wrote:
Hi George,
On 7 October 2015 at 16:29, George McCollister george.mccollister@gmail.com wrote:
Instead of having x86-pinctrl work separately from ich6-gpio have it work underneath ich6-gpio. This removes redundant configuration and will allow the addition of shared bank settings in future commits.
Signed-off-by: George McCollister george.mccollister@gmail.com
arch/x86/dts/minnowmax.dts | 96 +++++++++--------- arch/x86/include/asm/gpio.h | 1 - board/intel/minnowmax/minnowmax.c | 9 +- doc/device-tree-bindings/gpio/intel,ich6-gpio.txt | 14 +++ .../gpio/intel,x86-pinctrl.txt | 17 ++-- drivers/gpio/intel_ich6_gpio.c | 107 ++++++++------------- 6 files changed, 118 insertions(+), 126 deletions(-) create mode 100644 doc/device-tree-bindings/gpio/intel,ich6-gpio.txt
Conceptually pinmux and GPIO are different concepts. What is the need to make pinctrl a subset of GPIO? In some ways the opposite would make more sense.
Perhaps instead we should have a pinctrl driver?
Yeah, I think you're right.
I think the only things you'd want the pinctrl driver to touch would be mode-func, pull-assign and pull-strength.
The gpio driver would unconditionally set do "ich6_gpio_set_function(dev, gpio_offset, 1);" if the gpio was setup and you'd just need a way of setting direction and output value in device tree.
Is there already sane structure defined we could copy or should I invent something? This is pretty low on my priority list but I might be able to get around to it in my spare time if you think it's the way to go.
Regards, Simon