regulator: BD71837: possible regression

[Vaittinen, Matti]

Hello Angus,

I'll add the linux list and Mark to CC as this sounds like a regression
which may impact to other regulator drivers too. Mark, please let me
know if you don't feel adding you to discussions like this are
appropriate so I don't do it in the future.

On Mon, 2019-05-13 at 17:21 -0700, Angus Ainslie wrote:
> Hi Matti,
> 
> I've been doing some integration with our feature branches on linux-
> next 
> (next-20190510) and I hit a snag as soon as I enable the BD71837.
> 
> There's a couple of weird things I'm seeing
> 
> [    0.907025] bd718xx-pmic bd718xx-pmic.2.auto: no default pinctrl 
> state
> [    0.907651] bd718xx-pmic bd718xx-pmic.2.auto: Unlocked lock
> register 
> 0x2f
> [    0.932483] rohm-bd718x7 0-004b: Looking up buck6-supply from
> device 
> tree
> [    0.932499] rohm-bd718x7 0-004b: Looking up buck6-supply property
> in 
> node /soc@0/bus@30800000/i2c@30a20000/pmic@4b fa
> iled
> [    0.937861] rohm-bd718x7 0-004b: Looking up buck7-supply from
> device 
> tree
> [    0.937877] rohm-bd718x7 0-004b: Looking up buck7-supply property
> in 
> node /soc@0/bus@30800000/i2c@30a20000/pmic@4b fa
> iled

Does this mean we have a regression in linux-next? I guess you have not
seen these issues in older kernels?

> What should buck6/7 supply be set to ?

The buck6 and buck7 are not supplied by any other regulators. What is
special about buck6 and buck7 is that they are supplying LDOs 5 and 6.
So I assume the problem emerges when LDOs are trying to find their
suppliers. (I am not sure if supplier is correct word - but what I mean
is that bucks 6 and 7 are kind of parents for LDOs). This is reflected
in regulator_desc for LDOs. (The supply_name field is set).

I am not sure but perhaps the regulator core is changed so that this
parent/child relation must be modelled using <foo>-supply properties in
device-tree. Are you able to bisect the change which breaks this? There
may be other regulator drivers doing the same as bd718x7 is (which
means trusiting to setting the supply_name in desc to be enough - and
without deeper understanding I'd say it should be enough).

If this change is intentional and buck6-supply and buck7-supply are bow
required also in DT, then we should reflect this fact also in bindings
doc for BD71837 and BD71847.

> The rtc also seems to have lost it's mind
> 
> [  497.363621] rtc rtc0: Timeout waiting for LPSRT Counter to change
> 
> Am I missing linking the bd71837 clock to the rtc somehow ?

I don't think there is any SW actions required. The 30.72K clock can be
controlled using clk-bd718x7 driver - but I don't think there has been
changes there and this drivers should not touch the clk gate unless
asked by some user. Still, if the bucks6/7 get disabled for some reason
(or if LDOs are enabled before these bucks are enabled) - then, without
the correct parent child relation the LDOs can't get enabled - and
voltage monitoring may kick-in. Yet, assuming the SoC is powered by
bd71837 this should probably lead to reset-loop and not just to clock
errors...

I would definitely start looking at the changes in how suppliers are
looked up and maybe also tried setting the buck6-supply and buck7-
supply properties with phandles to buck6 and buck7 nodes. I can also
try checking the core if you don't have the possibility to do bisecting
- but I am not able to do it right now. It may be I can check this only
at next week.

> And finally the reset is hanging again even with 
> "rohm,reset-snvs-powered".

This could well be due to the missing parent/child relation between
bucks and LDOs.

> 
> Pointers would be appreciated.
> 
> Thanks
> Angus

Br,
	Matti Vaittinen