Re: [PATCH v7 2/9] dt-bindings: phy: rockchip,rk3588-mipi-dcphy: Select TX or RX, add phy-type
From: Quentin Schulz
Date: Wed Nov 18 2026 - 04:28:33 EST
Hi Krzysztof,
On 11/18/26 9:16 AM, Krzysztof Kozlowski wrote:
On 18/11/2026 06:55, Jason Yang wrote:
The Samsung MIPI D-/C-PHY block on RK3588 and RK3576 has a transmitter
for DSI and a receiver for CSI-2, and the DSI host and the CSI-2 host
can use the transmitter and the receiver at the same time. The single
cell of this binding selects the mode and refers to the transmitter, so
the receiver cannot be described.
Change the cell to select the transmitter or the receiver. The
transmitter and the receiver cannot operate in different modes [1], so
the optional phy-type property of the PHY node sets the mode for the
whole block, D-PHY by default:
/* the MIPI DSI host, driving the transmitter */
phys = <&mipidcphy0 PHY_DIRECTION_TX>;
/* the MIPI CSI-2 host, driving the receiver */
phys = <&mipidcphy0 PHY_DIRECTION_RX>;
For C-PHY the consumers stay the same and the PHY node gets:
phy-type = <PHY_TYPE_CPHY>;
Existing DTs on a newer kernel: PHY_TYPE_DPHY and PHY_TYPE_CPHY remain
valid as deprecated values. They select the transmitter and imply
phy-type = <PHY_TYPE_DPHY> or <PHY_TYPE_CPHY> respectively, so nothing
changes for them.
New DTs on an older kernel: an ABI break. The older kernel ignores
phy-type and takes any cell value as the mode of the block. A DSI node
with PHY_DIRECTION_TX therefore gets no display, and a CSI-2 node with
PHY_DIRECTION_RX gets no camera.
[1] RK3588 TRM: section 22.1 (overview): "The TX and RX only support
operate in the same PHY mode."
Suggested-by: Michael Riesch <michael.riesch@xxxxxxxxxxxxx>
Link: https://lore.kernel.org/r/94d47cab-10c6-4ec3-84a0-3bddc9d67f5f@xxxxxxxxxxxxx
Signed-off-by: Jason Yang <jason98166@xxxxxxxxx>
Reviewed-by: Rob Herring (Arm) <robh@xxxxxxxxxx>
Reviewed-by: Michael Riesch <michael.riesch@xxxxxxxxxxxxx>
Tested-by: Quentin Schulz <quentin.schulz@xxxxxxxxx> # RK3588 Tiger CSI1 2-lane IMX219, CSI0 2-lane OV5675+DSI
Please drop the tag, this did not happen. In the future, please add only
applicable tags. Otherwise share how this code was executed and tested
on the actual hardware (make process, like building code, is obviously
not testing).
What did not happen? I tested the v6, on the reported hardware, on the reported two DCPHYs, see https://lore.kernel.org/all/840285d6-a8a6-4777-8d38-6758ccbc7723@xxxxxxxxx/
If you're complaining about the Tested-by being given to a dt-binding patch, it was provided as an answer to the cover-letter, and the contributor is clearly using b4 considering the last few lines from the cover letter (base-commit/change-id lines). b4 trailers -u applies the tags given to the cover letter to all patches in the series, so if there's someone to blame here, it's me for adding it as an answer to the cover letter. So now the question is, do you want me to add a Tested-by: to each individual patches of long series when I technically built and tested only based on the last patch of the series, which is only a DTS patch adding the nodes to the RK3588 DTSI? In that case, I cannot provide Tested-by for patches before the last, as it can simply not work at all until the last (actually all of them, except the bindings) patch is applied.
Cheers,
Quentin