System crash possibly due to edid errors

[videosmith]

Describe the bug

I plug an RPi 400 (single 720p HDMI-0 output) into a 1x4 hdmi splitter with the goal of duplicating the computer output on two or more monitors.

It works under both Bookworm (latest updates) and Trixie (latest updates) - both under labwc - either from boot up or adding monitors after system stabilized.

Steps to reproduce the behaviour

If I replace one of the monitors with a Decimator on the HDMI splitter, it works under Bookworm (either from boot up or after stabilized) however computer fails to boot under Trixie-or if already booted and stabilized the os appears to crash.

All attempts to add HDMI entries into the boot/firmware/config.txt are apparently ignored and commenting out #dtoverlay=vc4-kms-v3d prevented the system from booting.

In this configuration when Trixie booting- computer shows the rainbow image on decimator and other monitors on power up, but then no signal after.

However, if i connect the computer directly to the Decimator and then run its HDMI output to the HDMI splitter all monitors work under Trixie.

Device (s)

Raspberry Pi 400

System

Raspberry Pi Bookworm and Trixie, latest updates.

Logs

No response

Additional context

No response

[popcornmix]

If I replace one of the monitors with a Decimator on the HDMI splitter, it works under Bookworm (either from boot up or after stabilized) however computer fails to boot under Trixie-or if already booted and stabilized the os appears to crash.

Remove sdcard (or other boot device), and other connected peripherals. Power on the pi400, and report the "display:" line of the diagnostic display (on a Pi400 you may need to press ESC to switch to diagnostic display).

[6by9]

HDMI splitters typically end up rewriting the EDID based on the displays connected, and it often varies based on the order that they are detected/connected, generally advertising the subset of modes that are supported by all displays.
Without knowing the exact rewrite it has done, there is no way to debug it.

Even if labwc fails to start, you should still be able to get SSH access. Please post the output of raspinfo when the splitter with decimator connected is attached, as that will give us the combined EDID for analysis.

If you have one particular screen that has modes that all others support, the best approach is to copy the EDID (from /sys/class/drm/card1-HDMI-A-1/edid or /sys/class/drm/card1-HDMI-A-2/edid) to /lib/firmware as say edid.bin. Edit /boot/firmware/cmdline.txt and add drm.edid_firmware=HDMI-A-1:edid.bin. The system will then always use that EDID for the connector instead of trying to read it from the display.

[videosmith]

If I replace one of the monitors with a Decimator on the HDMI splitter, it works under Bookworm (either from boot up or after stabilized) however computer fails to boot under Trixie-or if already booted and stabilized the os appears to crash.

Remove sdcard (or other boot device), and other connected peripherals. Power on the pi400, and report the "display:" line of the diagnostic display (on a Pi400 you may need to press ESC to switch to diagnostic display).

[videosmith]

HDMI splitters typically end up rewriting the EDID based on the displays connected, and it often varies based on the order that they are detected/connected, generally advertising the subset of modes that are supported by all displays. Without knowing the exact rewrite it has done, there is no way to debug it.

Even if labwc fails to start, you should still be able to get SSH access. Please post the output of raspinfo when the splitter with decimator connected is attached, as that will give us the combined EDID for analysis.

If you have one particular screen that has modes that all others support, the best approach is to copy the EDID (from /sys/class/drm/card1-HDMI-A-1/edid or /sys/class/drm/card1-HDMI-A-2/edid) to /lib/firmware as say edid.bin. Edit /boot/firmware/cmdline.txt and add drm.edid_firmware=HDMI-A-1:edid.bin. The system will then always use that EDID for the connector instead of trying to read it from the display.

Thank you. I will attempt this after work. Still confused as to why it works under Bookworm but not Trixie...

[videosmith]

HDMI splitters typically end up rewriting the EDID based on the displays connected, and it often varies based on the order that they are detected/connected, generally advertising the subset of modes that are supported by all displays. Without knowing the exact rewrite it has done, there is no way to debug it.
Even if labwc fails to start, you should still be able to get SSH access. Please post the output of raspinfo when the splitter with decimator connected is attached, as that will give us the combined EDID for analysis.
If you have one particular screen that has modes that all others support, the best approach is to copy the EDID (from /sys/class/drm/card1-HDMI-A-1/edid or /sys/class/drm/card1-HDMI-A-2/edid) to /lib/firmware as say edid.bin. Edit /boot/firmware/cmdline.txt and add drm.edid_firmware=HDMI-A-1:edid.bin. The system will then always use that EDID for the connector instead of trying to read it from the display.

Thank you. I will attempt this after work. Still confused as to why it works under Bookworm but not Trixie...

Due to the file length, attached as a zip file

info.zip

[videosmith]

HDMI splitters typically end up rewriting the EDID based on the displays connected, and it often varies based on the order that they are detected/connected, generally advertising the subset of modes that are supported by all displays. Without knowing the exact rewrite it has done, there is no way to debug it.
Even if labwc fails to start, you should still be able to get SSH access. Please post the output of raspinfo when the splitter with decimator connected is attached, as that will give us the combined EDID for analysis.
If you have one particular screen that has modes that all others support, the best approach is to copy the EDID (from /sys/class/drm/card1-HDMI-A-1/edid or /sys/class/drm/card1-HDMI-A-2/edid) to /lib/firmware as say edid.bin. Edit /boot/firmware/cmdline.txt and add drm.edid_firmware=HDMI-A-1:edid.bin. The system will then always use that EDID for the connector instead of trying to read it from the display.

Thank you. I will attempt this after work. Still confused as to why it works under Bookworm but not Trixie...

Due to the file length, attached as a zip file

info.zip

Performed operations as suggested, no change. Still able to communicate via SSH however.

[popcornmix]

@6by9 the dmesg does contain the vc4_hdmi_write_infoframe warning.
Does that cause an actual issue? Otherwise, hotplug and edid are present, but not overlays present:

Using labwc with kms
Connector 0 (33) HDMI-A-1 (connected)
  Encoder 0 (32) TMDS
Connector 1 (42) HDMI-A-2 (disconnected)
  Encoder 1 (41) TMDS

[6by9]

the dmesg does contain the vc4_hdmi_write_infoframe warning.
Does that cause an actual issue?

No, it doesn't cause an issue.

The driver is throwing an error when trying to enable the mode due pixel clock being out of range.
Decode the EDID using edid-decode and you'll see that it includes the block

  Vendor-Specific Data Block (HDMI), OUI 00-0C-03:
    Source physical address: 1.1.0.0
    Maximum TMDS clock: 145 MHz

1920x1080@60Hz (or 50Hz) with 8bpp RGB requires a TMDS clock of 148.5MHz.
The full timing is as per the detailed timing in that EDID

    DTD 1:  1920x1080   60.000000 Hz  16:9     67.500 kHz    148.500000 MHz (1872 mm x 1053 mm)
                 Hfront   88 Hsync  44 Hback  148 Hpol P
                 Vfront    4 Vsync   5 Vback   36 Vpol P

The driver therefore fails to enable when 1920x1080@60 is requested.

I doubt that the Decimator advertises an EDID with that limitation, so it is presumably down to the splitter rewriting it.

I've done the quick hack to the EDID to increase that limit to 150MHz (it's in 5MHz increments). With that I get the display coming up correctly.
linux7199.zip

[videosmith]

the dmesg does contain the vc4_hdmi_write_infoframe warning.
Does that cause an actual issue?

No, it doesn't cause an issue.

The driver is throwing an error when trying to enable the mode due pixel clock being out of range. Decode the EDID using edid-decode and you'll see that it includes the block

  Vendor-Specific Data Block (HDMI), OUI 00-0C-03:
    Source physical address: 1.1.0.0
    Maximum TMDS clock: 145 MHz

1920x1080@60Hz (or 50Hz) with 8bpp RGB requires a TMDS clock of 148.5MHz. The full timing is as per the detailed timing in that EDID

    DTD 1:  1920x1080   60.000000 Hz  16:9     67.500 kHz    148.500000 MHz (1872 mm x 1053 mm)
                 Hfront   88 Hsync  44 Hback  148 Hpol P
                 Vfront    4 Vsync   5 Vback   36 Vpol P

The driver therefore fails to enable when 1920x1080@60 is requested.

I doubt that the Decimator advertises an EDID with that limitation, so it is presumably down to the splitter rewriting it.

I've done the quick hack to the EDID to increase that limit to 150MHz (it's in 5MHz increments). With that I get the display coming up correctly. linux7199.zip

Thank you for your assistance. On a hunch, I connected an Atomos Ninja instead of a Decimator to the splitter and same result. It appears your conclusions concerning the splitter may be correct (why other multiple monitors/converters work is a mystery to me).

I was able to get two monitors and then connect the Ninja to all display signal one time, but could not repeat it, so I consider that a fluke.

I assume the reason for these events in Trixie and not Bookworm is more rigid hdmi specifications were written into Trixie?

I attempted to incorporate your patch (thank you again), however I may have implemented it incorrectly - was not able to boot after modifications.

[popcornmix]

Is the cmdline.txt you posted all on one line? It's not clear to me if I'm seeing word-wrap or newlines (there should be no newlines).

Use D for digital rather than e. I've not used the , notation for edid_firmware. You could try with them separated. e.g.

video=HDMI-A-1:1280x720@60D video=HDMI-A-2:1280x720@60D drm.edid_firmware=HDMI-A-1:edid/linux-7199.bin drm.edid_firmware=HDMI-A-2:edid/linux-7199.bin 

The file should be in /lib/firmware/edid/linux-7199.bin.
initramfs makes this more complicated - if you don't need it to boot, try disabling it (remove auto_initramfs=1 from config.txt).

[videosmith]

Is the cmdline.txt you posted all on one line? It's not clear to me if I'm seeing word-wrap or newlines (there should be no newlines).

Use D for digital rather than e. I've not used the , notation for edid_firmware. You could try with them separated. e.g.

video=HDMI-A-1:1280x720@60D video=HDMI-A-2:1280x720@60D drm.edid_firmware=HDMI-A-1:edid/linux-7199.bin drm.edid_firmware=HDMI-A-2:edid/linux-7199.bin 

The file should be in /lib/firmware/edid/linux-7199.bin. initramfs makes this more complicated - if you don't need it to boot, try disabling it (remove auto_initramfs=1 from config.txt).

My apologies it is word wrapped.
Thank you for your suggestions, I will attempt shortly.

[videosmith]

Attempted your suggestions... Sadly, no improvement.
Boot or hot plug monitors in the splitter, all good, Decimator or Ninja either no boot to gui or if already loaded gui disappears - no video signal.
I appreciate the assistance :)