Hi,
I'm having issues with non-functional DSI output driving a 2-lane panel connected to DSI0 on MSM8x74 MDP5 v1.2 hardware. The code in drivers/gpu/drm/msm/dsi/dsi_host.c around line 703 enables lane 1 and 2 instead of lane 0 and 1 for performance reasons (and then enables lane swapping in the DSI PHY (LANE_SWAP_1230)). I cannot get this to work on the MSM8x74, no data observed on physical lane 0 (can't access lane1 easily unfortunately). Disabling this swap and enabling lane0 and 1 instead results in a fully functional panel.
Would it make sense to make this performance optimization a compile-time option that defaults to off (for certain problematic hardware or in general)?
/wj
Hi Werner,
I will prepare a change to make the lane swap configurable.
Thanks, Hai
-----Original Message----- From: dri-devel [mailto:dri-devel-bounces@lists.freedesktop.org] On Behalf Of Johansson, Werner Sent: Thursday, August 20, 2015 8:54 PM To: dri-devel@lists.freedesktop.org Subject: drm/msm/dsi Lane-swapping for 2-lane panels non-functional for msm8x74?
Hi,
I'm having issues with non-functional DSI output driving a 2-lane panel connected to DSI0 on MSM8x74 MDP5 v1.2 hardware. The code in drivers/gpu/drm/msm/dsi/dsi_host.c around line 703 enables lane 1 and 2 instead of lane 0 and 1 for performance reasons (and then enables lane swapping in the DSI PHY (LANE_SWAP_1230)). I cannot get this to work on the MSM8x74, no data observed on physical lane 0 (can't access lane1 easily unfortunately). Disabling this swap and enabling lane0 and 1 instead results in a fully functional panel.
Would it make sense to make this performance optimization a compile-time option that defaults to off (for certain problematic hardware or in general)?
/wj _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/dri-devel
From: Hai Li [mailto:hali@codeaurora.org] Sent: den 21 augusti 2015 13:14 Hi Werner,
I will prepare a change to make the lane swap configurable.
Thanks, Hai
That sounds really great Hai!
Any idea as to why this lane swap would misbehave? In theory the swap should be completely transparent to the rest of the code (and panel). This behavior is seen on a Dragonboard 800 (the only device with a display with low enough resolution to fit in two lanes).
Thanks! /wj
dri-devel@lists.freedesktop.org