On Wed, Jan 6, 2016 at 1:53 PM, Nishanth Menon [off-list ref] wrote:

On 01/06/2016 01:34 PM, Rob Herring wrote:

quoted

On Wed, Jan 6, 2016 at 8:36 AM, Nishanth Menon [off-list ref] wrote:

quoted

On 01/06/2016 02:13 AM, Laxman Dewangan wrote:

quoted

On Wednesday 06 January 2016 01:12 PM, H. Nikolaus Schaller wrote:

quoted

Hi,

Am 06.01.2016 um 00:40 schrieb Nishanth Menon [off-list ref]:

quoted

On 01/05/2016 06:01 AM, H. Nikolaus Schaller wrote:

quoted

+        rtc {
+            compatible = "ti,palmas-rtc";
+            interrupt-parent = <&palmas>;
+            interrupts = <8 IRQ_TYPE_NONE>;

IRQ_TYPE_NONE is not correct here -> it should have some polarity - if
it had none, there'd be no interrupt, right?

Well, it just translates IRQ_TYPE_NONE through

Linux/include/dt-bindings/interrupt-controller/irq.h

to

interrupts = <8 0>;

which is given as an example in

Documentation//devicetree/bindings/rtc/rtc-palmas.txt

Since I don't know anything about the rtc driver beyond the bindings
documentation I assume it is correct...
I have added Laxman Dewangan because he introduced this interrupts =
<8 0>;

As this is for palmas interrupt controller, it does not use the second
field for interrupt from RTC.
So there is no really any polarity. It can be set to 0.

The second argument will be used for GPIOs mainly. However, support need
to be added on GPIO driver for rising/falling configuration.

Device tree represents the hardware - not to reflect how the driver
works. if the driver is wrong, fix it. the interrupt polarity needs to
be described in DT. based on palmas like designs, you should probably
use IRQ_TYPE_EDGE_FALLING because that is the default signaling inside
the SoC as it reaches Secondary interrupt handlers(SIH) registers.

If the trigger type is not programmable, then not setting the trigger
type in the DT is fine. Internal connections are often not documented.

Weird, that is not what I got feedback when I send
https://patchwork.ozlabs.org/patch/381125/

If this is the new norm, I retract my objection.

I'm not saying that necessarily makes sense here. Given the above, you do know the polarity.

To put it another way, if the trigger type doesn't matter, why do you have a second cell?

Rob