linux-kernel - Re: [PATCH] ARM: dts: imx6sx-sdb: Remove cpufreq OPP override

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]

Message-ID: <1493890943.11226.42.camel@nxp.com>
Date:   Thu, 4 May 2017 12:42:23 +0300
From:   Leonard Crestez <leonard.crestez@....com>
To:     Marek Vasut <marex@...x.de>
CC:     Peter Chen <Peter.Chen@....com>, Anson Huang <Anson.Huang@....com>,
        "linux-pm@...r.kernel.org" <linux-pm@...r.kernel.org>,
        Viresh Kumar <viresh.kumar@...aro.org>,
        "Rafael J. Wysocki" <rjw@...ysocki.net>,
        linux-kernel <linux-kernel@...r.kernel.org>,
        Christoph Fritz <chf.fritz@...glemail.com>,
        Sascha Hauer <kernel@...gutronix.de>,
        "Fabio Estevam" <fabio.estevam@....com>,
        Shawn Guo <shawnguo@...nel.org>,
        "Fabio Estevam" <festevam@...il.com>,
        "linux-arm-kernel@...ts.infradead.org" 
        <linux-arm-kernel@...ts.infradead.org>
Subject: Re: [PATCH] ARM: dts: imx6sx-sdb: Remove cpufreq OPP override

On Wed, 2017-05-03 at 21:33 +0200, Marek Vasut wrote:
> On 05/03/2017 07:58 PM, Leonard Crestez wrote:
> > On Wed, 2017-05-03 at 17:59 +0200, Marek Vasut wrote:
> > > On 05/03/2017 04:58 PM, Leonard Crestez wrote:
> > > > On Wed, 2017-05-03 at 16:26 +0200, Marek Vasut wrote:

> > > > > 2) It actually fixes a problem with the voltage rails such that the DVFS
> > > > >    works without leaving the system in unstable or dead state. You do
> > > > >    need the second part of my patch if you drop the OPP hackery, without
> > > > >    it the power framework cannot correctly configure the core voltages,
> > > > >    so the patch from Leonard makes things worse.

> > > > No, I think there is a misunderstanding here. The second part of your
> > > > patch will cause cpufreq poking at LDOs to indirectly adjust the input
> > > > from the PMIC to the minimum required (this is LDO target +
> > > > min_dropout_uv). Without it by default VDD_ARM_SOC_IN will remain fixed
> > > > as 1375mV from boot.

> > > Who sets / guarantees that default value for ARM and SOC rails ?

> > I think it's from the PMIC hardware itself (but maybe uboot plays with
> > it). VDD_ARM_SOC_IN on this board is tied to SW1AB from MMPF0200:
> > 
> > http://www.nxp.com/assets/documents/data/en/data-sheets/MMPF0200.pdf
> > 
> > It seems reasonable to rely on such voltages set externally.

> Isn't it an established rule that Linux should not depend on bootloader
> settings ? Or did that change ?

I don't actually know. Is there a hard and fast rule about this, even when it comes to voltages?

In theory it is possible for a bootloader to set a low cpu frequency and low voltage and then have the chip fail when the cpufreq driver attempts to go higher. Setting vin-supply on reg_arm/reg_soc would fix that.

> Well the regulator(s) cannot be correctly configured if the kernel
> doesn't have the correct power distribution described in the DT .

It depends on your definition of "correctness". It it certainly
possible to get a functional system while only partially describing
regulator relationships.

I think there is a further misunderstanding here. I have a problem
where imx6sx-sdb rev C boards crash on boot with upstream (but are
reported to work fine with rev B). Removing the OPP overrides fixes
this specific issue.

I don't object to the second part of your patch, setting correct supply links is a good thing for various reasons. It is just not necessary for fixing the concrete crash mentioned above (and I tested this). It should probably go in a separate patch.

It might seem a pedantic difference but it's good to accurately describe the effect of patches in commit messages. For example it might help somebody looking to backport various fixes.

--
Regards,
Leonard