Re: [PATCH V2 0/8] PM / OPP: Multiple regulator support

From: Dave Gerlach
Date: Sun Oct 23 2016 - 21:08:24 EST

On 10/21/2016 10:40 AM, Viresh Kumar wrote:
On 21 October 2016 at 19:09, Rafael J. Wysocki <rafael@xxxxxxxxxx> wrote:
On Thu, Oct 20, 2016 at 10:44 AM, Viresh Kumar <viresh.kumar@xxxxxxxxxx> wrote:

Some platforms (like TI) have complex DVFS configuration for CPU
devices, where multiple regulators are required to be configured to
change DVFS state of the device. This was explained well by Nishanth
earlier [1].

Some thoughts went into it few months back but then it all got lost. I
am trying to get that back on track with this thread.

One of the major complaints around multiple regulators case was that the
DT isn't responsible in any way to represent the ordering in which
multiple supplies need to be programmed, before or after frequency
change. It was considered in this patch and such information is left to
the platform specific OPP driver now, which can register its own
opp_set_rate() callback with the OPP core and the OPP core will then
call it during DVFS.

The patches are tested on Exynos5250 (Dual A15). I have hacked around DT
and code to pass values for multiple regulators and verified that they
are all properly read by the kernel (using debugfs interface).

Though more testing on real (TI) platforms would be useful.

This is rebased over: linux-next branch in the PM tree.

- Ack from Rob for 1st patch
- Moved the supplies structure to pm_opp.h (Dave)
- Fixed an compilation warning.

I need somebody from the OPP camp to review patches [2-8/8] for me.

Sure, I have already asked Stephen yesterday to do that.

Overall this series looks good to me apart from a few small things. Most importantly I was able to get a working implementation using two regulators on ti dra7xx platform with proper sequencing built on top of this series. We have cpu regulator and Adaptive body bias (abb) regulator that must be scaled in a certain order before or after clock scaling and I was able to implement a rough custom set_rate to perform this and ran some dvfs stress tests that all worked fine.

First comment, I think the platform specific set_rate is a good place to hook in for adaptive voltage scaling as well. I was able to implement TI Class0 AVS in the same code by using the requested transition voltage as a reference and programming AVS voltage using that, along with scaling the additional regulators in sequence (the original multi regulator functionality). I would think some people would want to use this even with single regulator platforms, no? cpufreq-dt works as is for that, we just swap out the regulators.

This raises some concerns about dependencies/probe sequencing. Right now we just need to make sure the cpufreq-dt driver probes after we have called _set_regulators, but if our platform code fails cpufreq-dt currently will treat this as no regulator needed for the platform and operate without one, which will likely hang the system. Is there a good way to to guarantee this doesn't happen? My main concern is that if we plan to provide a platform specific set-rate function, we should have a way to indicate this and prevent things from progressing if it isn't yet ready.

Again, overall I think it solves the multi regulator problem, and it works well for AVS as well. For the series:

Tested-by: Dave Gerlach <d-gerlach@xxxxxx>