Re: [PATCH] drm/phy: rockchip: fix RK3506 DSI low-power command handling
From: Heiko Stübner
Date: Sat Nov 21 2026 - 08:34:29 EST
Am Samstag, 21. November 2026, 09:05:55 Mitteleuropäische Normalzeit schrieb ZhuXiangyu:
> Two bugs keep the RK3506 MIPI-DSI link from initialising a panel: the
> DPHY programs the wrong high-speed timing entry, and the PHY data lanes
> are never enabled in the GRF.
>
> 1. phy-rockchip-inno-dsidphy picks the wrong HS timing entry
>
> The 1.5 GHz inno DSI PHY timing table stores its rates in Mbps while
> the other tables and the lookup use Hz. The mode enable code compares
> inno->pll.rate in Hz (e.g. 372000000) against timings[].rate in Mbps
> (e.g. 400), so the comparison never matches. The loop runs to the end
> and the last, fastest entry is used, which programs a TLPX/LPX
> counter that is far too large. The panel then cannot decode the
> low-power DCS commands and stays in sleep mode, so the display stays
> black even though the LP packets are visible on the data lane.
>
> Convert the 1.5 GHz timing table rates to Hz so that the lookup
> selects the correct entry.
>
> 2. The RK3506 DSI GRF lane enable bits are never written
>
> rk3506_chip_data only clears the turnaround / force-stop fields, so
> the PHY data lanes stay disabled unless the boot loader happens to
> enable them. Program RK3506_DSI_PHY_ENABLE_LANE0/1 as well.
>
> Tested on the RK3506B JerryTech board: the panel now receives the init
> sequence over LP and the framebuffer console is visible.
>
> Fixes: 785a9d5bb145 ("phy: rockchip: inno-dsidphy: Add support for rk3506")
> Fixes: cabeacc7eadc ("drm/rockchip: dsi: Add support for rk3506")
> Signed-off-by: ZhuXiangyu <1194781887@xxxxxx>
don't mix subsystems in patch submission please.
All the process issues, Krzysztof pointed out in another thread still
apply too.
Heiko