On Mon, 2016-08-22 at 15:45 +0200, Linus Walleij wrote:

On Fri, Aug 19, 2016 at 2:44 PM, Andrew Jeffery [off-list ref] wrote:

quoted

+++ b/drivers/pinctrl/aspeed/Kconfig

@@ -0,0 +1,8 @@
+config PINCTRL_ASPEED
+       bool
+       depends on (ARCH_ASPEED || COMPILE_TEST) && OF
+       select PINMUX
+       select PINCONF
+       select GENERIC_PINCONF
+       select MFD_SYSCON
+       select REGMAP_MMIO

Since this device is spawn from the syscon, should it not be
"depends on MFD_SYSCON"?

(No big deal, if you think this is right then go with it.)

I think that's a fair point, I will look at rearranging it.

quoted

+#include 

What is this include for?

Cruft from past iterations. Will remove.

quoted

+#include 
+#include 
+#include 
+#include 
+#include "../core.h"
+#include "pinctrl-aspeed.h"

No #include ?

Maybe some #includes are centralized to pinctrl-aspeed.h
I don't know, just make sure you don't have implicit includes.

linux/regmap.h is included in pinctrl-aspeed.h

quoted

+               if (regmap_read(map, desc->reg, &val) < 0)
+                       return false;
+
+               val &= ~desc->mask;
+               val |= pattern << __ffs(desc->mask);
+
+               if (regmap_write(map, desc->reg, val) < 0)
+                       return false;

Use:

regmap_update_bits(map,
                              desc->reg,
                              desc->mask,
                              pattern << __ffs->desc->mask);

Or something like that instead of reimplementing
mask-and-set. Regmap already knows how to do the
business.

(Applied everywhere in the driver where you have a
mask-and-set like this).

Right, I will clean that up.

The expression core engine is still a complete mystery
for me, I will just trust you that it works as intended.

Gah! However, thanks!

I will clean up the issues above and re-send.

Cheers,

Andrew