I am sorry about this. Since It's my first time to commit patch to the
kernel, I want to make sure my code and reply is correct,
so I actually paste my reply to AI before I send, let AI to rewrite my
words. I will fix my approach, and make sure V4 and afterward work
is worth to review.
Best regards,
Jason
Krzysztof Kozlowski [off-list ref] 於 2026年8月13日週四 下午3:59寫道：
On Thu, Aug 13, 2026 at 12:33:52PM +0800, 楊智成 wrote:
quoted
Hi Sebastian,
Thanks for the review.
quoted
Your new binding is lacking too. If you select <&mipidcphy0 PHY_TYPE_CSI>
you defined the direction of the PHY, but will it operate in C-PHY or in
D-PHY mode?
You are right, and I am sorry - this is a regression I introduced in v3.
v2 already described these as two separate cells: the first selected the
electrical layer, and the second selected the direction. Michael Riesch
suggested using the generic PHY_TYPE_DSI/PHY_TYPE_CSI constants for the
second cell instead of the Rockchip-private RK_DCPHY_DIR_* constants [1],
and Vinod Koul agreed [2].
In v3, while making that change, I also mistakenly collapsed the two cells
into one. That is what lost the electrical-layer information you pointed
out.
For v4, I will go back to the v2 layout and only rename the second cell.
This keeps the two dimensions separate, with each cell describing one
thing:
/* MIPI DSI host - the transmitter */
phys = <&mipidcphy0 PHY_TYPE_DPHY PHY_TYPE_DSI>;
/* MIPI CSI-2 host - the receiver */
phys = <&mipidcphy0 PHY_TYPE_DPHY PHY_TYPE_CSI>;
The same applies to C-PHY, so the binding can express all four combinations
without any further changes:
phys = <&mipidcphy0 PHY_TYPE_CPHY PHY_TYPE_DSI>;
phys = <&mipidcphy0 PHY_TYPE_CPHY PHY_TYPE_CSI>;
The driver accepts all four combinations at phy_get() time and returns
-EOPNOTSUPP from power_on() for the C-PHY combinations, as it did before
this series. Boards that only wire up the transmitter keep
'#phy-cells = <1>', so the existing in-tree device trees remain unaffected.
Before sending v4, I will go through the code and commit messages again,
so that the reasoning above is captured in the commits themselves rather
than only in this thread. I will also re-test the series on the board.
What is with this practice (third posting I see this week) of pasting
as reply prompts/tasks for AI?
All this is just wrong approach and your patches will be ignored if you
act as a proxy and keep playing with us that way.
Best regards,
Krzysztof