Re: [PATCH v3 00/10] pmdomain: Partial refactor, support modem and RTFF

[Ulf Hansson <ulf.hansson@linaro.org>]

On Tue, 5 Aug 2025 at 09:48, AngeloGioacchino Del Regno
<angelogioacchino.delregno@collabora.com> wrote:
>
> Changes in v3:
>  - Dropped specified items for cells restriction as suggested by Rob
>  - Fixed an issue in patch 4 still referencing "mediatek,bus-protection"
>    as it is entirely replaced by "access-controllers"
>
> Changes in v2:
>  - Added #access-controller-cells allowance for MT8188/95 infracfg_ao
>
> This series is a subset of [1], leaving out the Hardware Voter specific
> bits for MT8196 until the discussion around it reaches a conclusion.
>
> Even though the proposed code was born as a preparation to support the
> MT8196/MT6991 SoCs power domain controllers, it is a necessary cleanup
> for all power domain controllers of all of the currently supported SoCs
> from MediaTek.
>
> You may also notice the addition of support for modem power sequences:
> this was brought up 6 months ago (or more) by community contributors
> (mainly Yassine Oudjana) that were trying to upstream the MediaTek
> MT6735 Smartphone SoC and needed support to provide power to the MD
> subsystem - so, even though in this specific series the code for the
> modem power sequence is not yet triggered by any SoC, please please
> please, let it in.
> Besides, "a bunch" of upstream supported SoCs do have the MD power
> domain even though it wasn't added to their drivers (because if there
> was no support in the driver, it would just crash the system); the
> addition is something that I plan to do at some point, but definitely
> not now as I have no bandwidth for that (bar MT8196, which will have
> this domain).
>
> Compared to v1 in [1]:
>  - Changed mediatek,bus-protection to access-controllers
>    as suggested by Rob (thanks!)
>  - Added commits to document #access-controller-cells on all of
>    the access control providers
>
> In the meanwhile.... relevant excerpt from the old series:
>
> This series refactors the bus protection regmaps retrieval to avoid
> searching in all power domain devicetree subnodes for vendor properties
> to get syscons for different busses, and adds a new property which is
> located in the power controller root node containing handles to the same.
>
> Retrocompatibility is retained and was tested on multiple SoCs in the
> Collabora lab - specifically, on Genio 350/510/700/1200, and manually
> on MT6795 Helio (Xperia M5 Smartphone), MT8186, MT8192 and MT8195
> Chromebooks.
>
> This was tested *three times*:
>  - Before the per-SoC conversion in drivers/pmdomain/mediatek
>  - With per-SoC conversion code but with *legacy* devicetree
>  - With per-SoC conversion code and with *new* devicetree conversion
>
> All of those tests were successful on all of the aforementioned SoCs.
>
> This also adds support for:
>  - Modem power domain for both old and new MediaTek SoCs, useful for
>    bringing up the GSM/3G/4G/5G modem for both laptop and smartphone use
>  - RTFF MCU HW, as found in MT8196 Chromebooks and MT6991 Dimensity 9400
>
> ...and prepares the pmdomain code to accomodate only the directly
> controlled power domains for MT8196 (HW Voter support was left out).
>
> [1] https://lore.kernel.org/all/20250623120154.109429-1-angelogioacchino.delregno@collabora.com
>
> AngeloGioacchino Del Regno (10):
>   dt-bindings: memory: mtk-smi: Document #access-controller-cells
>   dt-bindings: clock: mediatek: Document #access-controller-cells
>   dt-bindings: power: mediatek: Document access-controllers property
>   pmdomain: mediatek: Refactor bus protection regmaps retrieval
>   pmdomain: mediatek: Handle SoCs with inverted SRAM power-down bits
>   pmdomain: mediatek: Move ctl sequences out of power_on/off functions
>   pmdomain: mediatek: Add support for modem power sequences
>   pmdomain: mediatek: Add support for RTFF Hardware in MT8196/MT6991
>   pmdomain: mediatek: Convert all SoCs to new style regmap retrieval
>   arm64: dts: mediatek: Convert all SoCs to use access-controllers
>
>  .../bindings/clock/mediatek,infracfg.yaml     |   3 +
>  .../clock/mediatek,mt8186-sys-clock.yaml      |  15 +
>  .../clock/mediatek,mt8188-sys-clock.yaml      |  15 +
>  .../clock/mediatek,mt8192-sys-clock.yaml      |  15 +
>  .../clock/mediatek,mt8195-sys-clock.yaml      |  15 +
>  .../clock/mediatek,mt8365-sys-clock.yaml      |  15 +
>  .../mediatek,smi-common.yaml                  |  16 +
>  .../power/mediatek,power-controller.yaml      |  37 ++
>  arch/arm64/boot/dts/mediatek/mt6795.dtsi      |   5 +-
>  arch/arm64/boot/dts/mediatek/mt8167.dtsi      |   6 +-
>  arch/arm64/boot/dts/mediatek/mt8173.dtsi      |   4 +-
>  arch/arm64/boot/dts/mediatek/mt8183.dtsi      |  17 +-
>  arch/arm64/boot/dts/mediatek/mt8186.dtsi      |  12 +-
>  arch/arm64/boot/dts/mediatek/mt8188.dtsi      |  23 +-
>  arch/arm64/boot/dts/mediatek/mt8192.dtsi      |  13 +-
>  arch/arm64/boot/dts/mediatek/mt8195.dtsi      |  20 +-
>  arch/arm64/boot/dts/mediatek/mt8365.dtsi      |  16 +-
>  drivers/pmdomain/mediatek/mt6795-pm-domains.h |   5 +
>  drivers/pmdomain/mediatek/mt8167-pm-domains.h |   5 +
>  drivers/pmdomain/mediatek/mt8173-pm-domains.h |   5 +
>  drivers/pmdomain/mediatek/mt8183-pm-domains.h |   5 +
>  drivers/pmdomain/mediatek/mt8186-pm-domains.h |   5 +
>  drivers/pmdomain/mediatek/mt8188-pm-domains.h |   6 +
>  drivers/pmdomain/mediatek/mt8192-pm-domains.h |   5 +
>  drivers/pmdomain/mediatek/mt8195-pm-domains.h |   5 +
>  drivers/pmdomain/mediatek/mt8365-pm-domains.h |  14 +-
>  drivers/pmdomain/mediatek/mtk-pm-domains.c    | 399 +++++++++++++++---
>  drivers/pmdomain/mediatek/mtk-pm-domains.h    |  74 +++-
>  28 files changed, 594 insertions(+), 181 deletions(-)
>

Patch 3->9 applied for next, thanks!

Note, patch 3 is also available on the immutable dt branch for SoC
maintainers to pull in.

Kind regards
Uffe