Re: [PATCH v2 00/12] arm64: Kconfig: Update ARCH_EXYNOS select configs

From: Krzysztof Kozlowski
Date: Thu Sep 30 2021 - 06:52:22 EST


On 30/09/2021 11:23, Lee Jones wrote:
> I've taken the liberty of cherry-picking some of the points you have
> reiteratted a few times. Hopefully I can help to address them
> adequently.
>
> On Thu, 30 Sep 2021, Krzysztof Kozlowski wrote:
>> Reminder: these are essential drivers and all Exynos platforms must have
>> them as built-in (at least till someone really tests this on multiple
>> setups).
>
>> Therefore I don't agree with calling it a "problem" that we select
>> *necessary* drivers for supported platforms. It's by design - supported
>> platforms should receive them without ability to remove.
>
>> The selected drivers are essential for supported platforms.
>
> SoC specific drivers are only essential/necessary/required in
> images designed to execute solely on a platform that requires them.
> For a kernel image which is designed to be generic i.e. one that has
> the ability to boot on vast array of platforms, the drivers simply
> have to be *available*.

By "essential", I meant drivers which are needed for supported platform
to boot properly. Also without significant performance penalty due to
probe deferrals.

Therefore if someone selects ARCH_EXYNOS in mainline kernel, it means
he/she wants such devices to be working fine with generic kernel.

This is the difference with term "drivers have to be available".

>
> Forcing all H/W drivers that are only *potentially* utilised on *some*
> platforms as core binary built-ins doesn't make any technical sense.
> The two most important issues this causes are image size and a lack of
> configurability/flexibility relating to real-world application i.e.
> the one issue we already agreed upon; H/W or features that are too
> new (pre-release).

Geert responded here. If drivers are essential for supported platforms
to boot, having them built-in has technical sense.
For other drivers not and we do not discuss such cases.

>
> Bloating a generic kernel with potentially hundreds of unnecessary
> drivers that will never be executed in the vast majority of instances
> doesn't achieve anything. If we have a kernel image that has the
> ability to boot on 10's of architectures which have 10's of platforms
> each, that's a whole host of unused/wasted executable space.

Geert responded here.

>
> In order for vendors to work more closely with upstream, they need the
> ability to over-ride a *few* drivers to supplement them with some
> functionality which they believe provides them with a competitive edge
> (I think you called this "value-add" before) prior to the release of a
> device. This is a requirement that cannot be worked around.
>
> By insisting on drivers (most of which will not be executed in the
> vast majority of cases) to be built-in, you are insisting on a
> downstream kernel fork, which all of us would like to avoid [0].

The same with all the DTS and hundreds of drivers you keep out of tree.

>
> [0] Full disclosure: part of my role at Linaro is to keep the Android
> kernel running as close to Mainline as possible and encourage/push the
> upstream-first mantra, hence my involvement with this and other sets.
> I assure you all intentions are good and honourable. If you haven't
> already seen it, please see Todd's most recent update on the goals and
> status of GKI:
>
> Article: https://tinyurl.com/saaen3sp
> Video: https://youtu.be/O_lCFGinFPM
>
>> We don't even know what are these unsupported, downstream platforms
>> you want customize kernel for. They cannot be audited, cannot be
>> compared. Affecting upstream platforms just because
>> vendor/downstream does not want to mainline some code is
>> unacceptable. Please upstream your drivers and DTS.
>
>> You also mentioned downstream devices but without actually ever defining
>> them. Please be more specific. What SoC, what hardware?
>
> Accepting changes based on the proviso that vendors upstream all of
> their work-in-progress solutions is an unfair position.

You twisted (or ignored) my words here. I said before - sacrificing any
mainline platform (e.g. reliable boot for distro) for an out-of-tree
vendor which does no want to upstream drivers is the unfair position.
One of the incentives of upstreaming is to be able to shape kernel and
be considered in kernel upstream decisions. That's the privileged for
upstreamed platforms - they have an impact.

Vendor decides to stay out - fine, vendor's choice. You loose impact.
Any sad words like "oh upstream does not want to change for me" should
receive a message: "please join the upstream :), so we will change
together".

> We already
> discussed why upstreaming support for bleeding edge H/W and
> functionality is unrealistic in terms of competitive advantage.

Nope, my last point was not responded to. You said that there is no
point for vendor to upstream bleeding edge HW. Then you said that there
is also little point to upstream older HW.

Basically following this approach you agree that vendor does not have to
do anything because it is inconvenient for him.

However upstream has to adapt to downstream vendor, right?

No, this is *unfair to all the platforms we upstreamed*.

>
> Besides, we might not be talking about new silicon at all (I don't
> believe anyone alluded to that). The flexibility in configuration
> simply allows for generic upstream drivers to be swapped out for ones
> which may have more or slightly different functionality (that can't be
> publicly shared until release).
>
>> Please explain why Exynos is special that it does not require essential
>> drivers to be selected as built-in. For example why aren't same changes
>> done for Renesas?
>
>> Everyone else are working like this. NXP, Renesas, Xilinx, TI, Rockchip,
>> AllWinner. Samsung or Google is not special to receive an exception for
>> this.
>
> Exynos isn't special in this regard. This applies to any vendor who
> releases Android images and wishes to be solve all of the issues the
> GKI project addresses (please read the article above for more about
> this).

We have then slightly different goals, because you are driven by Android
release images and this is why you are less interested in NXP and
Renesas. Only some vendors should receive such changes? No, in upstream
we are not focusing on any specific distro and there are other uses of
Exynos (and NXP and Renesas) platforms. Therefore the choice/policy and
overall design tries to match all vendors, not only some subset
convenient to Android.

If Android (or some vendor) wants to have exception for a specific
driver/platform, it must do it in upstream way, by contributing, not by
changing to match downstream needs while ignoring other users.

Best regards,
Krzysztof