CEC works on OPI6-plus

[royka1]

With this patch and CONFIG_DRM_DISPLAY_DP_AUX_CEC=y CEC works on the Orange Pi 6 and I was wondering if this also works on the RADXA Orion o6 since that dp to HDMI adapter also seems to support it. It also fixes that the screen stays black after a disconnect on the opi6

https://github.com/user-attachments/files/25153594/dptx-fix-plus-cec.patch

[coffnix]

I tested this patch using DisplayPort on my Orange Pi 6 Plus and it worked perfectly. I no longer need to boot with the DisplayPort cable connected. Now I can power on the board first and connect the cable later and the video comes up automatically every time. I also tested several connect and disconnect cycles and did not find any issues, video always came back correctly over DisplayPort.

I will test HDMI next and if I notice any problems I will report back.

I am currently running kernel 6.19 with the following GRUB configuration:

opi6plus ~ # cat /boot/efi/GRUB/GRUB.CFG
set debug=loader,mm
set term=vt100
set default=0
set timeout=2

menuentry '1 OrangePi 6 Plus (Device Tree)' {
    devicetree /sky1-orangepi-6-plus-6.19.dtb
    linux /vmlinuz-6.19 \
        loglevel=0 drm.debug=0 \
        console=ttyAMA2,115200 console=tty1 fbcon=map:1 \
        efi=noruntime \
        earlycon=pl011,0x040d0000 \
        arm-smmu-v3.disable_bypass=0 \
        clk_ignore_unused \
        linlon_dp.enable_fb=1 \
        nowatchdog watchdog_dev.handle_boot_enabled=0 \
        acpi=off \
        root=PARTUUID=8112a51c-af33-4033-9676-c142ea88d79d rootwait rw
    initrd /initramfs-6.19.img
}

[coffnix]

Yo, is this patch needed on kernel 6.19.4?

opi6plus /usr/src/linux # patch -p1 < /root/dptx-fix-plus-cec.patch
patching file drivers/gpu/drm/cix/dptx/trilin_dptx.c
Hunk #1 succeeded at 32 with fuzz 2 (offset 4 lines).
Hunk #2 succeeded at 60 (offset 6 lines).
Hunk #3 succeeded at 2551 (offset 42 lines).
Hunk #4 succeeded at 2630 (offset 42 lines).
Hunk #5 FAILED at 2628.
Hunk #6 succeeded at 2711 (offset 63 lines).
Hunk #7 succeeded at 2769 (offset 63 lines).
Hunk #8 succeeded at 2840 (offset 63 lines).
Hunk #9 succeeded at 3257 (offset 139 lines).
1 out of 9 hunks FAILED -- saving rejects to file drivers/gpu/drm/cix/dptx/trilin_dptx.c.rej
patching file drivers/gpu/drm/cix/dptx/trilin_dptx.h
Hunk #1 succeeded at 446 (offset 1 line).
patching file drivers/gpu/drm/cix/dptx/trilin_drm.c
Hunk #3 succeeded at 1246 (offset 26 lines).

[Entrpi]

Thanks for the patch and testing! We've incorporated the CEC support and the HPD bounce fix, though with a different approach for each.

What we took

CEC-over-DP-AUX (commit e925bb8)

We wired up the kernel's drm_dp_cec helpers — 5 hook points following the i915/nouveau/amdgpu pattern:

• drm_dp_cec_register_connector() at init
• drm_dp_cec_set_edid() when EDID is read in get_modes
• drm_dp_cec_unset_edid() on HPD disconnect
• drm_dp_cec_irq() in the HPD IRQ handler
• drm_dp_cec_unregister_connector() at deinit

The kernel framework handles all the debouncing, retry logic, and adapter lifecycle internally, so the custom delayed work and retry counters from your patch weren't needed. Requires CONFIG_DRM_DISPLAY_DP_AUX_CEC=y. All functions compile to no-ops when disabled.

HPD bounce recovery (commit 361e862)

Your patch's "hard restart" (stop streams, reset core, restart streams) catches the real problem — fast cable unplug/replug where the disconnect event is never seen by the driver. Instead of directly manipulating streams from the work queue, we check live DPCD link status (clock recovery + channel equalization) when an HPD event arrives on an already-connected port. If the link has degraded, we force a disconnect/reconnect cycle through the normal code path. This lets the DRM atomic modeset framework handle the stream teardown and restart.

DP core teardown on disconnect

We already had this from an earlier patch — trilin_dp_core_off() in handle_disconnect() with active_stream_cnt reset.

Not taken

The stream stop/restart logic in hpd_event_work_func — it bypasses the atomic commit path and has ordering concerns. The DPCD link status check achieves the same recovery without directly touching streams.

Testing needed

• Hot-plug with display connected at boot vs. connected after boot
• Fast unplug/replug cycles (the bounce scenario)
• CEC passthrough on DP-to-HDMI bridges (PS186 on Orion O6, whatever bridge the OPi6+ uses)
• cec-ctl for CEC message send/receive

Would be great to hear if this works on the Orange Pi 6 Plus — especially the HPD bounce recovery.

[royka1]

@Entrpi Great I'll test it now and let you know. There's another patch coming for multi-channel audio/ channel mapping correction which needs to be adapted and cleaned up first. Do you want me to place it as a patch in another issue or pull request? HBR audio is still not working.

[coffnix]

If I boot with the DisplayPort disconnected, when I plug it in later no image appears and this shows up in dmesg:

[  855.280488] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_aux_transfer][ERROR]failed to do aux transfer (-110)
[  855.300032] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_hpd_event_work_func][info]dp hpd event received: Plugged
[  855.339922] linlondp 14160000.disp-controller: [drm] fb0: linlondpdrmfb frame buffer device

If I boot with the DisplayPort connected, the video appears normally. If I unplug it and then plug it back in, the image appears normally and this shows up in dmesg:

[   38.791852] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_hpd_event_work_func][info]dp hpd event received: Unplugged
[   38.792120] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_disable][info]enter
[   38.792138] linlondp 14160000.disp-controller: linlondp_crtc_atomic_disable enter
[   38.819656] linlondp 14160000.disp-controller: linlondp_crtc_atomic_disable exit
[   38.819678] linlondp 14160000.disp-controller: linlondp_dev_suspend
[   38.819689] [drm] Disconnect mmu without internal TBU!
[   43.125959] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_hpd_event_work_func][info]dp hpd event received: Plugged
[   43.161045] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_compute_config][info]mode 3840x2160 clock=533250 bpc=10 bpp=30 link_rate=540000 (HBR2) lanes=4 max_bpc=10
[   43.161159] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_atomic_mode_set][info]set mode: 3840x2160 533250
[   43.161194] linlondp 14160000.disp-controller: linlondp_crtc_atomic_enable enter
[   43.161220] linlondp 14160000.disp-controller: linlondp_crtc_atomic_enable, succeed to enable aclk.
[   43.161303] linlondp 14160000.disp-controller: linlondp_dev_resume
[   43.161331] [drm] Connect mmu without internal TBU!
[   43.161594] linlondp 14160000.disp-controller: linlondp_crtc_atomic_enable exit
[   43.161619] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_enable][info]enter
[   43.163962] cix-usbdp-phy 9110000.usb-phy: sky1_udphy_init: typec dir= 0x0
[   43.164090] cix-usbdp-phy 9110000.usb-phy: sky1_udphy_init: sky1_udphy_dp_conf
[   43.164450] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_core_on][info]Successly initialize DP phy
[   43.200834] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_core_on][info]main link training done! rate:540000 lanes:4
[   43.200896] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_ctrl_stream_on][info]trilin_dp_encoder_enable stream_id=0, active_stream_cnt:1

My grub.cfg:

opi6plus ~ # cat /boot/efi/GRUB/GRUB.CFG
set debug=loader,mm
set term=vt100
set default=0
set timeout=2

menuentry '1 OrangePi 6 Plus (6.19.4)' {
    devicetree /sky1-orangepi-6-plus-6.19.4.dtb
    linux /vmlinuz-6.19.4 \
    loglevel=7 ignore_loglevel printk.devkmsg=on drm.debug=0 \
        console=ttyAMA2,115200 console=tty1 \
        efi=noruntime \
        earlycon=pl011,0x040d0000 \
        arm-smmu-v3.disable_bypass=0 \
        clk_ignore_unused \
        linlon_dp.enable_fb=1 \
        nowatchdog watchdog_dev.handle_boot_enabled=0 \
        acpi=off \
        root=PARTUUID=8112a51c-af33-4033-9676-c142ea88d79d rootwait rw
    initrd /initramfs-6.19.4.img
}

I had to keep console=ttyAMA2,115200 console=tty1 for the framebuffer to work. Without setting console=tty1, no image appears at all before the X login screen starts.

Another test. When I boot without X starting, and then start X, this shows up in dmesg:

[Fri Feb 27 13:09:27 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_compute_config][info]mode  clock=241500 bpc=10 bpp=30 link_rate=270000 (HBR) lanes=4 max_bpc=10
[Fri Feb 27 13:09:27 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_disable][info]enter
[Fri Feb 27 13:09:27 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_ctrl_stream_off][info]trilin_dp_encoder_disable stream_id=0, active_stream_cnt=0
[Fri Feb 27 13:09:27 2026] linlondp 14160000.disp-controller: linlondp_crtc_atomic_disable enter
[Fri Feb 27 13:09:27 2026] linlondp 14160000.disp-controller: linlondp_crtc_atomic_disable exit
[Fri Feb 27 13:09:27 2026] linlondp 14160000.disp-controller: linlondp_dev_suspend
[Fri Feb 27 13:09:27 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_atomic_mode_set][info]set mode:  241500
[Fri Feb 27 13:09:27 2026] linlondp 14160000.disp-controller: linlondp_crtc_atomic_enable enter
[Fri Feb 27 13:09:27 2026] [drm] Disconnect mmu without internal TBU!
[Fri Feb 27 13:09:27 2026] linlondp 14160000.disp-controller: linlondp_crtc_atomic_enable, succeed to enable aclk.
[Fri Feb 27 13:09:27 2026] linlondp 14160000.disp-controller: linlondp_dev_resume
[Fri Feb 27 13:09:27 2026] [drm] Connect mmu without internal TBU!
[Fri Feb 27 13:09:27 2026] linlondp 14160000.disp-controller: linlondp_crtc_atomic_enable exit
[Fri Feb 27 13:09:27 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_enable][info]enter
[Fri Feb 27 13:09:27 2026] cix-usbdp-phy 9110000.usb-phy: sky1_udphy_init: typec dir= 0x0
[Fri Feb 27 13:09:27 2026] cix-usbdp-phy 9110000.usb-phy: sky1_udphy_init: sky1_udphy_dp_conf
[Fri Feb 27 13:09:27 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_core_on][info]Successly initialize DP phy
[Fri Feb 27 13:09:27 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_core_on][info]main link training done! rate:270000 lanes:4
[Fri Feb 27 13:09:27 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_ctrl_stream_on][info]trilin_dp_encoder_enable stream_id=0, active_stream_cnt:1

After stopping X, the image comes back again, and this shows up in dmesg:

[Fri Feb 27 13:09:35 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_compute_config][info]mode 3840x2160 clock=533250 bpc=10 bpp=30 link_rate=540000 (HBR2) lanes=4 max_bpc=10
[Fri Feb 27 13:09:35 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_disable][info]enter
[Fri Feb 27 13:09:35 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_ctrl_stream_off][info]trilin_dp_encoder_disable stream_id=0, active_stream_cnt=0
[Fri Feb 27 13:09:35 2026] linlondp 14160000.disp-controller: linlondp_crtc_atomic_disable enter
[Fri Feb 27 13:09:35 2026] linlondp 14160000.disp-controller: linlondp_crtc_atomic_disable exit
[Fri Feb 27 13:09:35 2026] linlondp 14160000.disp-controller: linlondp_dev_suspend
[Fri Feb 27 13:09:35 2026] [drm] Disconnect mmu without internal TBU!
[Fri Feb 27 13:09:35 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_atomic_mode_set][info]set mode: 3840x2160 533250
[Fri Feb 27 13:09:35 2026] linlondp 14160000.disp-controller: linlondp_crtc_atomic_enable enter
[Fri Feb 27 13:09:35 2026] linlondp 14160000.disp-controller: linlondp_crtc_atomic_enable, succeed to enable aclk.
[Fri Feb 27 13:09:35 2026] linlondp 14160000.disp-controller: linlondp_dev_resume
[Fri Feb 27 13:09:35 2026] [drm] Connect mmu without internal TBU!
[Fri Feb 27 13:09:35 2026] linlondp 14160000.disp-controller: linlondp_crtc_atomic_enable exit
[Fri Feb 27 13:09:35 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_enable][info]enter
[Fri Feb 27 13:09:35 2026] cix-usbdp-phy 9110000.usb-phy: sky1_udphy_init: typec dir= 0x0
[Fri Feb 27 13:09:35 2026] cix-usbdp-phy 9110000.usb-phy: sky1_udphy_init: sky1_udphy_dp_conf
[Fri Feb 27 13:09:35 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_core_on][info]Successly initialize DP phy
[Fri Feb 27 13:09:36 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_core_on][info]main link training done! rate:540000 lanes:4
[Fri Feb 27 13:09:36 2026] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_ctrl_stream_on][info]trilin_dp_encoder_enable stream_id=0, active_stream_cnt:1

[royka1]

Great work again! Under acpi CEC works, hotplugging hdmi works, audio gets stuck on deferred probing because the Orange Pi 6 doesn't have dptx_audio2. This patch fixes the problem.

fix-deferred-probe.patch

[Entrpi]

New round of fixes pushed to all tracks. Here's what landed:

Cold-plug fix

@coffnix's cold-plug failure (boot without display → plug in later → AUX timeout) is now fixed. The root cause is that handle_connect() attempts AUX reads before the sink has powered up. We added retry logic — up to 5 attempts with 100ms delays when read_sink_caps() returns -ETIMEDOUT. The vendor code had a comment at that exact spot: /*Fixme: may need sleep for ready...*/.

Audio deferred probe fix

@royka1's fix-deferred-probe.patch is incorporated. On boards like the OPi6+ that don't have all I2S instances, the sound card would defer-probe indefinitely. We now check whether each I2S platform device exists (via bus_find_device_by_name()) before creating dai_links, skipping any that are missing.

Audio channel mapping

Added three things the DPTX audio driver was missing:

• get_eld() callback — copies the connector's EDID-Like Data to hdmi-codec so it can read the sink's audio capabilities (supported formats, channel counts, speaker allocation)
• Audio InfoFrame SDP — packs and transmits a CEA Audio InfoFrame via hdmi_audio_infoframe_pack_for_dp(). Without this, the sink defaults to stereo regardless of how many channels are being transmitted.
• CEA channel allocation forwarded from hdmi-codec's hw_params to the InfoFrame

This should fix basic multi-channel audio routing. @royka1 — still interested in seeing your channel mapping correction patch, these fixes might complement it.

Extended sample rates (88.2–192kHz)

The DPTX audio channel status registers only had IEC60958 values for 32/44.1/48kHz — anything else silently fell through to the 48kHz default. Added proper IEC60958 channel status encoding (sampling frequency + original sampling frequency) for 88.2, 96, 176.4, and 192kHz.

Audio shutdown & disconnect robustness

• audio_shutdown() now actually writes AUDIO_ENABLE=0 to stop audio packets in hardware (previously it only set a flag)
• core_off() (called on HPD disconnect) disables audio before tearing down the PHY — stops audio streaming into a dead link when cable is pulled during playback
• Removed the !plugin early-return from shutdown() so the running state is always cleaned up, even if the cable was already pulled

HBR audio investigation

Investigated whether we can enable HBR (High Bit Rate) audio. Based on the CIX P1 TRM:

• No dedicated HBR mode in the DPTX hardware
• S/PDIF is "future enhancement" — not available in current silicon, so compressed audio passthrough isn't possible
• MAUD/NAUD are auto-generated in async clock mode (which is what we use) — the get_audio_clk_rate() returning 0 is correct, not a bug
• I2S MC supports up to 192kHz — the I2S interface is not the bottleneck

The extended sample rate support above is the practical fix — 96/192kHz PCM should now work correctly with the proper channel status metadata. True HBR compressed audio would require S/PDIF support that the hardware doesn't currently provide.

Patch counts: main=138, latest=139, rc=42, next=42

[coffnix]

Thanks for the fix @Entrpi , I tested it with kernel 6.19.5 and it worked perfectly. I booted without the DisplayPort cable connected, and when I plugged it in after the system had finished booting, the image came up perfectly. Look:

[   42.635797] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_aux_transfer][ERROR]failed to do aux transfer (-110)
[   42.655363] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_hpd_event_work_func][info]dp hpd event received: Plugged
[   42.695814] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_compute_config][info]mode 3840x2160 clock=533250 bpc=10 bpp=30 link_rate=540000 (HBR2) lanes=4 max_bpc=10
[   42.695891] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_atomic_mode_set][info]set mode: 3840x2160 533250
[   42.695900] linlondp 14160000.disp-controller: linlondp_crtc_atomic_enable enter
[   42.695908] linlondp 14160000.disp-controller: linlondp_crtc_atomic_enable, succeed to enable aclk.
[   42.695970] linlondp 14160000.disp-controller: linlondp_dev_resume
[   42.695983] [drm] Connect mmu without internal TBU!
[   42.696704] linlondp 14160000.disp-controller: linlondp_crtc_atomic_enable exit
[   42.696711] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_enable][info]enter
[   42.698959] cix-usbdp-phy 9110000.usb-phy: sky1_udphy_init: typec dir= 0x0
[   42.699054] cix-usbdp-phy 9110000.usb-phy: sky1_udphy_init: sky1_udphy_dp_conf
[   42.699393] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_core_on][info]Successly initialize DP phy
[   42.734152] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_core_on][info]main link training done! rate:540000 lanes:4
[   42.734254] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_ctrl_stream_on][info]trilin_dp_encoder_enable stream_id=0, active_stream_cnt:1
[   42.735351] Console: switching to colour frame buffer device 480x135
[   42.756210] linlondp 14160000.disp-controller: [drm] fb0: linlondpdrmfb frame buffer device

I tested unplugging and plugging it back in, and the image worked perfectly on the screen through the DisplayPort connector, see below.

[  140.254593] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_hpd_event_work_func][info]dp hpd event received: Unplugged
[  140.254806] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_disable][info]enter
[  140.254825] linlondp 14160000.disp-controller: linlondp_crtc_atomic_disable enter
[  140.286341] linlondp 14160000.disp-controller: linlondp_crtc_atomic_disable exit
[  140.286354] linlondp 14160000.disp-controller: linlondp_dev_suspend
[  140.286429] [drm] Disconnect mmu without internal TBU!
[  161.912289] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_hpd_event_work_func][info]dp hpd event received: Plugged
[  161.947401] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_compute_config][info]mode 3840x2160 clock=533250 bpc=10 bpp=30 link_rate=540000 (HBR2) lanes=4 max_bpc=10
[  161.947511] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_atomic_mode_set][info]set mode: 3840x2160 533250
[  161.947546] linlondp 14160000.disp-controller: linlondp_crtc_atomic_enable enter
[  161.947572] linlondp 14160000.disp-controller: linlondp_crtc_atomic_enable, succeed to enable aclk.
[  161.947654] linlondp 14160000.disp-controller: linlondp_dev_resume
[  161.947683] [drm] Connect mmu without internal TBU!
[  161.947953] linlondp 14160000.disp-controller: linlondp_crtc_atomic_enable exit
[  161.947979] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_encoder_enable][info]enter
[  161.950274] cix-usbdp-phy 9110000.usb-phy: sky1_udphy_init: typec dir= 0x0
[  161.950383] cix-usbdp-phy 9110000.usb-phy: sky1_udphy_init: sky1_udphy_dp_conf
[  161.950794] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_core_on][info]Successly initialize DP phy
[  161.987572] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_core_on][info]main link training done! rate:540000 lanes:4
[  161.987638] trilin-dptx-cix 141b4000.dp: [drm:trilin_dp_ctrl_stream_on][info]trilin_dp_encoder_enable stream_id=0, active_stream_cnt:1

[royka1]

spdif you can just put a 1 in trilin_dptx.c and passthrough works with ac3/dts/dd+ except hbr.
I'm still experimenting to make hbr work so I didn't have time to clean up anything. This is the dirty patch for 6.19.0+ that fixes the mapping and all changes I made in the dptx driver. I'll clean it up once I've got it fully working.

channel-mapping.patch

[RadxaYuntian]

From what I understand our DP to HDMI adapter does not support CEC.