[Exynos] LE 13.0-nightly for Odroid XU4 no longer boots

    I guess I'm one of the few who occasionally test new nightly images for the xu4, so starting with LibreELEC-Exynos.arm-13.0-nightly-20250730-ecfa284-odroid-xu4 the testserver images no longer boot. The blue led is blinking, but the display stays blank.

    Can somebody with an xu4 uart cable check what is happening? This seems to be present in all images I checked since then.

    So last one working is LibreELEC-Exynos.arm-13.0-nightly-20250725-f0482de-odroid-xu4. I noticed 4 consecutive missed daily builds in that timespan, could this be related?

    no, can't even ping the device, so the router did not hand out an ip to conman. Can I just use the previous kernel from 07.25 with it's dtb file and check against a current system squash file?

    At the moment I'm just able to boot and run with kernel 6.16. With kernels 6.16.1 I can boot but kodi crashes right after.

    In the end I could add the driver I wanted in kernel 6.16 and it's working like a charm booting from SD. However, booting from EMMC doesn't work, where as I'm able to boot an ubuntu image from EMMC :/

    tetienne unsure what state you have your Samsung in? What changes were required? Are you saying that an image with 6.16.0 works but 6.16.1 fails? What are the logs? Does 6.18 work? SD is the place to start. You would need to share the uboot uart logs of the emmc boot? What uboot is being used?

    Hi heitbaum

    so... Question of boot on EMMC is solved, I used another uboot. I have a working 6.16.0 build on my EMMC module (build made by myself as I will explain here after).

    Regarding linux kernel modifs, I've tested all of them on SD cards.

    - kernel 6.16.0: I started a branch from commit 46be10c (which was not booting at all) and modified kernel config as described in attached patch. This built boots fine (SD and now EMMC as well) and kodi works fine.

    For other kernels, I realized that I did not branch from relevant commits, so I did other tests.

    - kernel 6.16.9: I built a booting and working image using the same additions in kernel config as 6.16.0

    - kernel 6.17.1: I created a branch from commit d712d796. No way to get something booting even with kernel additions I was using for kernel 6.16*

    - kernel 6.18: same story after fetching last commits today, even with kernel additions from 6.16*

    I don't really know how to get logs, I don't have any UART cable :/

    lack of the odroid xu4 uart kit is my main issue as well, otherwise i'd like to help test and debug this. The kit is 10$ + a whopping 20$ of shipping fees when you order it in Europe from the hardkernel shop. My budet stops at 20$ for sth like this, shipping fees included :)

    chewitt: can you compare your item with the one from hardkernel [link] ?

    I spot 2 differences:

    1. Molex5268-04 connector goes to the xu4 board [I would have to solder it or find the bare connector plug and crimp it.]
    2. CP2102N: USB to Serial UART bridge IC with 1.8~3.3V interface [lower voltage than what amazon offers]

    The old uart kit might work with the xu4, but it is discontinued.

    The XU4 board has UART pins. The connector(s) on the USB UART can be pushed onto said pins. There is 1000% guaranteed no need for a special molex plug. Just connect GND/TX/RX (do not connect +3.3v, it's not needed).

    I tried this twice and it failed both times. The uart adapters I used are cheap ones like I use for connecting to rpi uarts, they work fine there. I might test one more time over the holidays, maybe it does work with just rx+tx+gnd. If so I would post uart logs in here of a recent testserver image.

    I've been using el cheapo things from Amazon for years. HK shipped me one of their UART cables alongside some samples at some point a decade ago but it's long lost and not required. I don't know of any ARM SoC hardware that needs the +3.3v pin wiring up.

    chewitt: you're right it works :) curiosity made me test it right now, so here are 2 uart logs from 07.25 (the last one booting) and 12.19 from today. Interesting enough you can see 12.19 loading the kernel without a panic or trace, but it is stuck later and is not getting an ip. Blue led keeps blinking, no display.

    tetienne: if you want me to test some of your sdhc images please post a link to your raw image. I'll then send you the uart log as well.

    The console log ends when init is executed and userspace is loaded from the SYSTEM file. I would assume boot continues though, so add "ssh" to kernel boot params in extlinux.conf and boot up. The SSH daemon will be forced to start on boot, allowing you to SSH in and run "journalctl | paste" then share a URL which will hopefully give clues on what's not working from later in the boot process.

    "ssh" parameter does not work. I can't ping the host, because the image is not asking for an ip so it is not getting one. Can I set additional debug switches in extlinux.conf? The ones I use are "systemd.debug_shell=ttySAC2 console=ttySAC2,115200n8 console=tty0"