Re: [PATCH v2 0/9] media: camss: add SM8750 support

[Qianhui Xu]

Hello Vladimir,

understood, and accepted - the CSIPHY must not grow another per-SoC register
table inside camss. Our 2/9 does exactly that (a 103-entry lane table added to
camss-csiphy-3ph-1-0.c for the 2.4.0 two-phase PHY), so that patch is the one
that has to go, not be reworked in place.

What I am going to do before proposing a shape:

  - read the SM8450 series (20261026224957.1115370-1-vz@xxxxxxxxxx) as the
    reference, in particular "media: qcom: camss: Allow CSI2 PHY devices
    without CSIPHY resources" - that is the part that defines how camss is
    expected to consume a generic PHY instead of owning the CSIPHY;
  - read phy-qcom-mipi-csi2-core.c and phy-qcom-mipi-csi2.h in tree, and the
    Kaanapali v22 phy patches, to see where the 2.4.0 two-phase register map
    and the common-status offset are supposed to live;
  - then restate the SM8750 support with the CSIPHY as a qcom,x1e80100-csi2-phy
    style PHY device, camss referencing it through phys/phy-names, and the
    binding describing the PHY nodes rather than CSIPHY registers.

Two things I want to be honest about up front:

  - our binding currently has no phys/phy-names at all; it lists the CSIPHY
    register windows and clocks as camss resources. That is the part that has
    to be rewritten, together with the DTS.
  - the 2.4.0 PHY data we have is from the vendor camera.ko and covers both a
    two-phase and a three-phase variant of the same block. Moving it into the
    PHY driver means deciding which variant the SoC actually wires, which is
    the kind of thing that belongs in the PHY driver's per-SoC data, not in
    camss. I would like to get that right rather than port the table as-is.

Separately: Hangxiang asked us to hold the SM8750 series until the Kaanapali
work is rebased, and I have agreed to that. So this will not appear as a v3
soon; when it does, the CSIPHY rework will be part of it from the start rather
than a follow-up.

Thanks for the review - it saved us from pushing the wrong shape twice.

Best regards,
Qianhui