[PATCH] ARM: dts: am335x-bone* enable pmic-shutdown-controller

[Matthijs van Duin]

On 18 May 2015 at 17:21, Tony Lindgren <tony at atomide.com> wrote:
> If some of these depend on the SoC revision and cannot be detected
> based on the RTC driver revision register, that information should
> be be passed to the RTC driver in platform data.

The SoC revision is easy enough to check via the control module as
usual, but is not really the biggest issue. Even on SoC rev 1.0
RTC-only sleep still has some functionality (the RTC freezes if PMIC
enters sleep-state, but that may still be preferred compared to
complete loss of date and time), so I'm not even sure I'd bother with
this check.

The serious problems however depend on the PCB: Entering RTC-only
sleep is only properly supported on some early prototype series
(pre-A6) of the BeagleBone Black. Since rev A6A (which includes all
production versions) it is not supported at all.

On rev A6 of the Black, as well as (afaik) all versions of the White
(if anyone cares, since they normally have SoC rev 1.0), the impact of
entering RTC-only sleep is heavily dependent on external connections
and hardware, most of which not really detectable. It is possible to
go from "hmm, annoying standby current" to a recipe for stir-fried I/O
pin simply by connecting a console cable while the device is in sleep.
This is not something the kernel can foresee or prevent.

I'm pretty sure the sane thing to do here is disable RTC-only sleep by
default. People who care about it can still enable it after they
verified the particular combination of pcb revision, external
hardware, and usage scenario is safe (or patched the hardware to make
it safe, or decided they're willing to take the risk.)

It may be useful to leave some references for those who seek guidance:

1. The issues with AM335x rev 1.0 SoCs are documented in the errata.

2. Unfortunately I still have no idea why TI altered the connection
diagram w.r.t. "VDDS" (and as a consequence officially defeatured
RTC-only sleep) when the TPS65217C/D PMIC for AM335x + DDR3/3L memory
is used. Maybe I'll still get an answer from TI on E2E, but I'm also
planning to do more thorough testing this week especially w.r.t.
current consumption patterns during boot and shutdown.

3. The BeagleBone regulator problem is particularly notorious among
people who (try to) use the PMIC's battery power support, since the
problem then also surfaces during full shutdown. It has a long
discussion thread here:
https://groups.google.com/d/topic/beagleboard/7sxPePT7wkM/discussion
which includes two posts by me with fairly detailed analysis along
with scope pics:
https://groups.google.com/d/msg/beagleboard/7sxPePT7wkM/3vFMPydR20IJ
https://groups.google.com/d/msg/beagleboard/7sxPePT7wkM/V1Ft-xxh0agJ