Hi Igor, On 9/2/26 6:28 PM, Igor Paunovic wrote: > Hi Cristian, > >> This has been tested on the following boards: > > One more, with a sink pair your boards cannot cover: > > Tested-by: Igor Paunovic <[email protected]> # RK3588 Orange Pi 5 Plus, > dual HDMI > > Setup: whole v11 on its stated base (drm-misc-next 47b823940e38), > both SoC HDMI ports driven at once: a Samsung with HF-VSDB (max > TMDS character rate 600 MHz, SCDC present) and a Sony limited to > 300 MHz TMDS. > > 1) HDMI 2.0 path: the Samsung now offers and runs 3840x2160@60 RGB > with a 594 MHz pixel clock (modetest-confirmed). Without the > series the same port tops out at 4K@30. A stable picture at > 594 MHz is only possible with scrambling and the 1/40 TMDS ratio > active; it stayed stable across boot, mode changes and hotplug. > > 2) Non-2.0 fallback: the Sony keeps its 297 MHz 4K@30 modes. > No regression. > > 3) The result I care most about: this Sony in standby keeps EDID/5V > up but drops the link, and on my 7.2 stack that generates a > storm of detect/modeset cycles - measured 119-194 per hour, > enough to make the desktop stutter. On v11, twenty minutes of > standby produced ZERO events, and wake restored both sinks by > itself. Whatever share of that the cached-status detect and the > SCDC-loss handling each take - thank you. > > 4) Hotplug on both ports: clean detach and reattach, no storms. > > 5) Audio on a dead link: this morning my 7.2 stack took an SError > panic (regmap read in dw_hdmi_qp_bridge_clear_audio_infoframe > when a PCM was closed on a link whose SCDC never came up - full > UART trace available if useful). On v11 I could not reproduce > the class: six open/close cycles against the standby sink plus > one natural prepare-failure-then-close during replug all > survived cleanly. Caveat: different base kernels, so this is > not a clean A/B on the series alone.
Thanks for the detailed testing report! > Scope: RGB output only (the YUV420 output series is separate), no > USB-C DP on this base, VC4 side not exercised. I will manually apply your T-b on the relevant patches before submitting a new revision. Cristian
