I believe you misread my post or are mistaking it for the other recent post that is similar. I am not migrating from an older Pi to the Pi 5, rather, I wrote a bespoke driver specifically targeting the Pi 5 using the exact timings provided by the LCD manufacturer.As has been covered numerous times, and I believe is in the white paper on interfacing to DSI displays, Pi0-4 all only have an integer divider from a core clock for generating the DSI clock. The core clock is 2GHz on Pi0-3, and 3GHz on Pi4. The vc4 driver on Pi4/CM4 will have fixed your requested mode up to 375MHz.
Whilst Pi5 has more flexibility in generating the DSI clock, you can always ask for the same thing as generated by vc4. The easiest way is via "kmstest -r" as I've already given as an example.
I am aware of the hardware differences between the platforms and even noted in my original post that this is not related to vc4_dsi_bridge_mode_fixup. The CM4 having a fixed divider actually further emphasizes the problem as it shows the LCD still works with some flexibility in its timings, while it remains unfunctional on the Pi 5 despite trying multiple different timings. The most recent results are using your suggested timings, and I even tried using the exact pixel clock required to maintain the same refresh rate to rule that out as being a potential problem.
I unfortunately do not have the flexibility to use an older version of the Pi as this is for a commercial product that requires the Pi 5 due to the number of I2S channels that are needed.
This post has the results of using that PR. I would be happy to perform additional tests but am unsure what else would be beneficial. I also tested the DSI_TO_CNT_CFG change, which also resulted in a non-functional display, though I cannot remember specifically what the waveform looked like.If it's now returning to LP11 state, then the "HS timeout" bug will no longer apply, and the PR changes are not needed.
It would however be quite helpful if you could test the PR just to verify there's no regression.
Good catch. I edited the post with the correct values for the Raspberry Pi 5 (the previous was the CM4 values).BTW in "measurements" above you seem to have quoted the same numbers twice - should they be different?
Statistics: Posted by colerd — Tue Jul 09, 2024 2:28 pm
