On Tue, Sep 8, 2015 at 10:04 AM, Paul Osmialowski <pawelo@xxxxxxxxxxx> wrote: > On Tue, 14 Jul 2015, Linus Walleij wrote: > >> I want Shawn and Sascha to look at this as they worked with >> other Freescale pin controllers. Especially I want to know if this >> is a sibling to the other Freescale controllers or a separate hardware. >> >> If it is *not* a sibling I will *insist* that it use more generic pin >> control bindings and move away from the older Freescale-specific >> stuff. > > No one answered me about that. However, I looked at other Freescale > pinctrl drivers and realised that no one of them (IMX, IMX1, MXS) is > similar to what I need to do for Kinetis, also positions of configuration > bits differ significantly. OK I insist on using the generic bindings then. >> There exist generic pin config bindings, see >> Documentation/devicetree/bindings/pinctrl/pinctrl-bindings.txt >> >> I suggest to to function+group paring and then use generic pin config >> with this driver unless it is a very close sibling to the existing Freescale >> pin controllers. >> >> Hint: if it is a sibling, it should share code with them. >> >> There are several drivers doing generic pin control/pin config in the kernel >> tree. > > I tried to analyze few of the drivers (e.g. zynq family) and can't find > how can I assing clock gate (clock device) to each port (PORTA, PORTB, > PORTC,...) which is required for Kinetis. Is generic pin control capable > to express that requirement or is it a time to desing my own pinctrl > driver (maybe somewhat improved than the one I presented so far)? That has nothing to do with whether you use generic pinconf or not. I imagine if a pinctrl unit contains several blocks with individual clocks you can either: - Let each block/bank be a device node (this is common) and assign each a clocks = <&clock>; - Keep one node and assign an array of clocks, affected block indicated by the index. clocks = <&clk1>, <&clk2>, <&clk3>, ... <clkN>; This property is in pluralis for this reason I guess. > This pinctrl component is somwehat critical part of BSP. Until it is not > sorted, I don't see a point in releasing what was developed so far. True that, you need infrastructure first. The more important that it is as good as possible. Yours, Linus Walleij -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html