Hi,

On 24/08/2020 15:32:22+0200, Bastian Krause wrote:

On 8/24/20 1:31 PM, Bastian Krause wrote:

quoted

On 4/15/20 8:56 PM, Alexandre Belloni wrote:

quoted

On 15/04/2020 18:37:00+0200, Bastian Krause wrote:

quoted

Signed-off-by: Bastian Krause <redacted>
---
 Documentation/devicetree/bindings/rtc/rtc-ds1307.txt | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/Documentation/devicetree/bindings/rtc/rtc-ds1307.txt b/Documentation/devicetree/bindings/rtc/rtc-ds1307.txt
index 66f0a31ae9ce..987a0c9e0cd7 100644
--- a/Documentation/devicetree/bindings/rtc/rtc-ds1307.txt
+++ b/Documentation/devicetree/bindings/rtc/rtc-ds1307.txt

@@ -34,6 +34,9 @@ Optional properties:
 - trickle-diode-disable : ds1339, ds1340 and ds 1388 only
 	Do not use internal trickle charger diode
 	Should be given if internal trickle charger diode should be disabled
+- aux-voltage-chargeable: rx8130 only
+	Epsons's rx8130 supports a backup battery/supercap.
+	This flag tells	whether the battery/supercap is chargeable or not.
 

I think we should make that a generic property and this should supersede
trickle-diode-disable which is a bit wonky as I would prefer the default
to be disabled instead of enabled with the current semantics.

Alright, I think I know how to transform the RTC drivers.

One question about the DTs though:

This means we should remove "trickle-diode-disable" from all upstream
DTs and add "aux-voltage-chargeable" to all upstream DTs that use a RTC
compatible whose driver care in their probe function for
"trickle-diode-disable", right?

Sorry, forget that.

Here's the situation:

Currently there is a switch to explicitly disable charging, so the
default is to charge. We cannot introduce another boolean switch to turn
that the other way around, because that would change the default and
break backwards compatibility.

The only way I can think of is to introduce "aux-voltage-chargeable" not
as a boolean switch but as an integer, without any default. If this
property is not available, the drivers should simply do what they did
prior to this change (look for the legacy trickle-diode-disable, use the
default they used before).

Are you okay with that?

I agree boolean should be avoided in RTC drivers because we need a way
to express "don't change this value".

Some more context:

I originally tried to add a chargeable flag for rx8130. Prior to this
patch, there was no need to set "trickle-diode-disable" for this,
because the driver did not pass the chargeable flag to the RTC. With the
patch the default would have been to charge as long as
"trickle-diode-disable" is not there. So there's a change in behavior.

Yes, IIRC, my point was simply to move the documentation for
aux-voltage-chargeable to the generice rtc binding documentation,
Documentation/devicetree/bindings/rtc/rtc.yaml

For now, you sould keep support for trickle-diode-disable but it has to be
superseded by aux-voltage-chargeable if present. Is that more clear?


-- 
Alexandre Belloni, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com