linux-kernel - Re: [PATCH] regulator: qcom-rpmh: Revert "regulator: qcom-rpmh: Use PROBE_FORCE

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 14 Jun 2023 12:35:08 -0700
From:   Doug Anderson <dianders@...omium.org>
To:     Amit Pundir <amit.pundir@...aro.org>
Cc:     Mark Brown <broonie@...nel.org>,
        Bjorn Andersson <andersson@...nel.org>,
        Saravana Kannan <saravanak@...gle.com>,
        Caleb Connolly <caleb.connolly@...aro.org>,
        Marek Szyprowski <m.szyprowski@...sung.com>,
        Andy Gross <agross@...nel.org>,
        Konrad Dybcio <konrad.dybcio@...aro.org>,
        Liam Girdwood <lgirdwood@...il.com>,
        linux-arm-msm@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] regulator: qcom-rpmh: Revert "regulator: qcom-rpmh: Use PROBE_FORCE_SYNCHRONOUS"

Hi,

On Wed, Jun 14, 2023 at 12:03 PM Amit Pundir <amit.pundir@...aro.org> wrote:
>
> On Wed, 7 Jun 2023 at 04:59, Doug Anderson <dianders@...omium.org> wrote:
> >
> > From a black box perspective, I guess the things I could think of
> > would be to keep poking around with things that you control. Best
> > ideas I have:
> >
> > 1. Use "bisect" style techniques to figure out how much you really
> > need to move the "lvs" regulators. Try moving it halfway up the list.
> > If that works, move it closer to the bottom. If that doesn't work,
> > move it closer to the top. Eventually you'd find out which regulator
> > it's important to be before.
>
> Hi, I tried this bisect style technique to move lvs regulators up in
> the list gradually and I found that they need to be enabled atleast
> before ldo12 and the ldo regulators which follow the ldo12 in the
> list.

Super weird. I was hoping that something would jump out, but nothing
does. :( I don't understand how lvs1 / lvs2 could have any impact on
ldo12. :(


> > 2. Try adding some delays to some of the regulators with
> > "regulator-enable-ramp-delay" and/or "regulator-settling-time-us".
> > Without a scope, it'll be tricky to figure out exactly which
> > regulators might need delays, but you could at least confirm if the
> > "overkill" approach of having all the regulators have some delay
> > helps... I guess you could also try putting a big delay for "ldo26".
> > If that works, you could try moving it up (again using a bisect style
> > approach) to see where the delay matters?
>
> I tried this approach as well earlier today but I don't know how big
> "the big" delay can be. The device fails to boot if I add a settling
> time of as much as 2sec per each ldo and lvs regulator too. I didn't
> try increasing the delay further.

Yeah, 2 seconds is plenty big. If that doesn't fix it then it's not a
timing issue.

I guess with the above results, I'm still super confused about why the
async probe has any impact at all on this. It sounds like the
_ordering_ of the rpmh-regulators init matters but not the timing, and
I'd expect the ordering to be the same between normal probe and async
probe. Specifically, I think:

a) There is exactly one rpmh-regulator driver instance in your system, right?

b) Regulator initialization happens in rpmh_regulator_probe().

c) The rpmh_regulator_probe() function is itself synchronous. That is,
it sets up one regulator at a time and, I believe, nothing about the
behavior of rpmh_regulator_probe() changes for async vs. sync probe.

...so I'm left baffled...