Am 01.08.2014 01:26, schrieb Tomasz Figa:

On 01.08.2014 01:17, Andreas F?rber wrote:

quoted

Am 31.07.2014 21:40, schrieb Tomasz Figa:

quoted

On 31.07.2014 21:20, Andreas F?rber wrote:

quoted

Am 31.07.2014 21:05, schrieb Tomasz Figa:

quoted

On 31.07.2014 18:08, Andreas F?rber wrote:

[snip]

quoted

quoted

quoted

quoted

quoted

+&dp_hpd {
+	samsung,pins = "gpc3-0";
+	samsung,pin-function = <0>;
+	samsung,pin-pud = <3>;
+	samsung,pin-drv = <0>;
+};

Hmm, what node is this referencing? I believe this should rather
reference the pin controller and add a new board-specific pinconf/pinmux
group instead....

It's a -pinctrl node. See v3->v4 change log and discussion on v3.

Well, this is clearly a board specific node anyway, because it does not
refer to a special function, but simply an input/interrupt GPIO. If it
somehow has landed in generic pinctrl dtsi then it should be removed
from there and this patch should simply introduce its own instance of
dp_hpd node, so you did the right thing in v3.

Well, my point was that the 3.8 tree contains only one dp-hpd node, not
two as we would get by adding a new node here.

Apart from Spring, it's used in Snow and SMDK5250, so moving it there
seems feasible and the cleanest solution to me.

What I mean is that in exynos5250-pinctrl.dtsi only generic SoC pin
groups should be defined and those more or less correspond to groups
with samsung,pin-function set to something other than 0 (input) or 1
(output). Now here hpd_gpio is just a normal GPIO input used as
interrupt source to detect when a cable is plugged or unplugged. This is
by no means generic to the SoC, because any GPIO with interrupt
capability can be used for this purpose. This means that the whole
pin{conf,mux} group should be defined on board level.

Exactly what I meant! :)

Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 N?rnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imend?rffer; HRB 16746 AG N?rnberg