linux-kernel - Re: [PATCH] soc: samsung: pm_domains: Convert to regular platform driver

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: <20210115151520.GA43049@kozik-lap>
Date:   Fri, 15 Jan 2021 16:15:20 +0100
From:   Krzysztof Kozlowski <krzk@...nel.org>
To:     Saravana Kannan <saravanak@...gle.com>
Cc:     Marek Szyprowski <m.szyprowski@...sung.com>,
        Linux Samsung SOC <linux-samsung-soc@...r.kernel.org>,
        Bartlomiej Zolnierkiewicz <b.zolnierkie@...sung.com>,
        LKML <linux-kernel@...r.kernel.org>,
        "Rafael J . Wysocki" <rafael@...nel.org>
Subject: Re: [PATCH] soc: samsung: pm_domains: Convert to regular platform
 driver

On Thu, Jan 14, 2021 at 11:07:30AM -0800, Saravana Kannan wrote:
> On Thu, Jan 14, 2021 at 11:03 AM Marek Szyprowski
> <m.szyprowski@...sung.com> wrote:
> >
> > Hi Saravana,
> >
> > On 13.01.2021 21:27, Saravana Kannan wrote:
> > > On Wed, Jan 13, 2021 at 3:03 AM Marek Szyprowski
> > > <m.szyprowski@...sung.com> wrote:
> > >> When Exynos power domain driver was introduced, the only way to ensure
> > >> that power domains will be instantiated before the devices which belongs
> > >> to them was to initialize them early enough, before the devices are
> > >> instantiated in the system. This in turn required not to use any platform
> > >> device infrastructure at all, as there have been no way to ensure proper
> > >> probe order between devices.
> > >>
> > >> This has been finally changed and patch e590474768f1 ("driver core: Set
> > >> fw_devlink=on by default") ensures that each device will be probbed only
> > >> when its resource providers are ready. This allows to convert Exynos
> > >> power domain driver to regular platform driver.
> > >>
> > >> This is also required by the mentioned commit to enable probing any
> > >> device which belongs to the Exynos power domains, as otherwise the core
> > >> won't notice that the power domains are in fact available.
> > >>
> > >> Signed-off-by: Marek Szyprowski <m.szyprowski@...sung.com>
> > >> ---
> > >> Some more comments are in the following thread:
> > >> https://protect2.fireeye.com/v1/url?k=8ac052ac-d55b6ba4-8ac1d9e3-0cc47a31c8b4-9068b559b0fd155d&q=1&e=b393c3ff-16ba-48a4-9d72-6805d02971d5&u=https%3A%2F%2Flore.kernel.org%2Flkml%2F2556a69b-5da5-bf80-e051-df2d02fbc40f%40samsung.com%2F
> > >> ---
> > >> ...
> > > Skimmed through this patch and at a high level, it looks good for what
> > > it's trying to do. Thanks for doing this!
> > >
> > > Btw, I assume that this won't work with fw_devlink=off/permissive
> > > (default since 5.10 or earlier)? My concern is that we might
> > > temporarily set fw_devlink=permissive by default if the other
> > > breakages aren't fixed in time for 5.12? How do you want to handle that?
> >
> > I've applied my patch on top of vanilla v5.10 and checked on my test
> > boards. Surprisingly everything works fine, so something must have been
> > changed during the last few years as the power domain driver in the
> > current form has been written long time ago. I remember that the moment
> > when platform devices are created from the of nodes has been change at
> > some point, so maybe this is somehow related. Anyway, the platform
> > driver for Exynos power domains registered from core_initcall works fine
> > with v5.10 kernel.
> >
> > I have no strong opinion on the way of merging this fix. It can go via
> > Samsung tree, so in the end the v5.12-rc1 will have both my fix and your
> > change, but won't be fully bisectable in-between. Krzysztof, what's your
> > opinion?
> 
> If it doesn't break anything without my changes, then let's try to get
> it merged independent of my series. This is a good change even without
> my changes.

I agree, I'll take it via Samsung SoC. It's not the perfect solution -
as Marek said, the tree won't be bisectable. Have in mind that some
other boards/architectures might be broken where no one reported it yet,
so fw_devlink=off/permissive might still be good choice for v5.12.

Best regards,
Krzysztof