在 2026-11-11三的 11:28 +0100,Krzysztof Kozlowski写道: > On 11/11/2026 11:26, Icenowy Zheng wrote: > > 在 2026-11-10二的 13:00 +0100,Krzysztof Kozlowski写道: > > > On Mon, Nov 09, 2026 at 05:07:04PM +0100, Romain TISSERAND wrote: > > > > The panel is a MIPI-DSI peripheral, a child node of its DSI > > > > host, > > > > and the > > > > driver does not use a graph port. A port is only needed when > > > > the > > > > host > > > > links its panel through the graph; other DSI hosts, such as the > > > > Allwinner > > > > sun6i-a31-mipi-dsi one, link the child node and have no output > > > > port > > > > to > > > > point at. The PinePhone panel is described that way on the same > > > > host, > > > > without a port. > > > > > > > > Keep the port allowed (panel-common) but no longer required, as > > > > for > > > > other > > > > DSI panels such as ilitek,ili9881c. > > > > > > ilitek,ili7807s and a few others have the same and it is > > > confusing. > > > The > > > port should be required when this is connected over SPI or I2C, > > > but > > > not > > > for MIPI DSI. I think that for MIPI DSI port is meaningless so > > > actually > > > incorrect. > > > > I found that the dw-mipi-dsi common code (e.g. on Rockchip SoCs > > with > > dw_mipi_dsi-rockchip) would expect the panel to be connected via a > > port > > even if it's the child node of the DSI controller. > > > > Such an example is at arch/arm64/boot/dts/rockchip/rk3399- > > pinephone- > > pro.dts . > > > > Yes, the same for Qualcomm. So I actually wonder who needs the change > here.
I wonder too because I just created a MIPI DSI driver based on the dw- mipi-dsi common code. I think the different requirement between sun6i-mipi-dsi and dw-mipi- dsi is exposing the implementation details of the corresponding drivers. Maybe this should be eliminated, but this would be some big work. Thanks, Icenowy > > Best regards, > Krzysztof
