Official LE12 Test Images for Amlogic (Kodi-21)

    Hi chewitt

    If it's convenient can we clarify something ?

    In your commit https://github.com/chewitt/linux/…5dd8932dc86859c you wrote that

    - P271 omits VP9 support and uses Mali 450-MP3
    - P261 omits VP9 support and uses Mali 450-MP2

    And we are still not 100% sure what exactly my Amlogic SoC (S905L, L2, L3, L B-2, L B-3...) is on my box Q96 . From the UART U-boot log of my first post https://forum.libreelec.tv/core/attachmen…q96-u-boot-txt/ it appears that my board is p271. It is certain that my variant is Mali 450-MP2. I use device tree for p261 to run LE.

    If you think it will make sense I can remove the CPU heatsink and see the truth.

    Quote from chewitt

    I'm attempting to guess/fix the GXLX audio problem in the posts above. Nothing that needs you to update all the time.

    Ok. The problem is that the new releases do not have a build number on their names (unlike the regular nightlies) nor it can be found in any other way other than updating. I do not mind downloading the image, so if there is a way to learn the build number before updating to it, please tell me.

    The images in my test share exist solely to support me testing things with random users. I deliberately use a fixed version and rarely clean build and almost never document the changes inside. I have no plans to change that, so use them at your own peril :S

    If you want something that more directly tracks the upstream LE repo and with known content, it's time to learn how to build your own images. It's not rocket science: https://wiki.libreelec.tv/development/build-basics

    Quote from chewitt

    dmladenov Yes, it woud be good to confirm things. I think my general idea is correct but p261/p271 might be reversed. I also realised that I made the last changes in code but didn't sync them (as patches) to my build folder so the last image you tested didn't contain them. Please update using https://chewitt.libreelec.tv/testing/LibreE…h64-11.80.0.tar (again) as I definitely included them this time.

    I just updated - the sound is gone again !

    Quote from chewitt

    dmladenov Yes, it woud be good to confirm things. I think my general idea is correct but p261/p271 might be reversed. I also realised that I made the last changes in code but didn't sync them (as patches) to my build folder so the last image you tested didn't contain them. Please update using https://chewitt.libreelec.tv/testing/LibreE…h64-11.80.0.tar (again) as I definitely included them this time.

    And so, the SoC of my Q96 box is Amlogic S905L with marking just "2" (without "B") which is certainly Mali 450-MP2.

    And I found the device-tree from another p271 box and it is also Mali 450-MP2 variant.

    So can we confirm that all p271 boxes are Mali 450-MP2 and should use meson-gxl-s905l-p271.dtb from your images ?

    Quote from dmladenov

    So can we confirm that all p271 boxes are Mali 450-MP2 and should use meson-gxl-s905l-p271.dtb from your images ?

    The current assumption is that p271 is the "3" variant and p261 is the "2" variant, so I need to switch the numbering else p271 will try to use 3x Mali cores and everything will lock-up when the hardware probes. I'll let you know when that switch is done.

    What does the case on your box look like? .. Amazon UAE has some cheap S905L devices so I might get one to run experiments on the audio problem. That needs to be resolved before I think about trying to upstream anything.

    Quote from chewitt

    The current assumption is that p271 is the "3" variant and p261 is the "2" variant, so I need to switch the numbering else p271 will try to use 3x Mali cores and everything will lock-up when the hardware probes. I'll let you know when that switch is done.

    What does the case on your box look like? .. Amazon UAE has some cheap S905L devices so I might get one to run experiments on the audio problem. That needs to be resolved before I think about trying to upstream anything.

    I'm not sure if I'm allowed to put such links in the forum - if not I'll delete them

    My box looks just like these:

    ** links removed **

    I've removed the links. It's the same as the cheapest box I have available here :)

    EDIT: I've ordered one for $25, arriving this evening. It'll give me something to play with while resting legs next week (am running a marathon tomorrow morning).

    Great, you're a maniac :)

    This thing has no reset button and must use ADB.

    For your convenience, the easiest way to boot from USB is through the Android update menu, point to the zipped autoscript file and click on Update button.

    I wish you a medal in the marathon...

    Quote from chewitt

    The images in my test share exist solely to support me testing things with random users. I deliberately use a fixed version and rarely clean build and almost never document the changes inside. I have no plans to change that, so use them at your own peril :S

    If you want something that more directly tracks the upstream LE repo and with known content, it's time to learn how to build your own images. It's not rocket science: https://wiki.libreelec.tv/development/build-basics

    It may not be rocket science for you, because you know how everything works. Sadly, and leaving my old hw aside, it is for me. I won't say again why I use your images and not the ones from the repo. And no, wifi is not the only reason, wired connection does not work as well with the stock image.

    On the other hand, if the image creation is something that can be scripted, I would offer you some help on a 58000x system (+64gb of ram +m2 ssd). I damaged it ~2 months ago (psu failed) and I do not have the money and the time to fix it :(

    We have reached 6.7, which is now stable and is there since the previous build (the one before Jan 9th), yet the wifi issue remains :(

    https://paste.libreelec.tv/flexible-ghoul.log

    Honest question. Can I install this on a usb stick and free my sd card for something more important like openwrt? Will it still boot from the stick now that its boot method has changed and it no longer looks inside its emmc for an os?
    Ideally, I would rather install dtech's 9.2.x build on its emmc, since there is no real benefit of using the latest version of kodi or le just for iptv.

    Edited once, last by jim_p (January 20, 2024 at 10:47 AM).

    I've reported the warnings to the repo owner: https://github.com/jwrdegoede/rtl8189ES_linux/issues/106 .. but don't hold breath.

    You should be able to run the image from SD or USB cards. It still boots from eMMC, but the boot routine is modified to search for LE files on SD and then USB devices. In past testing I've found that some vendor u-boots struggle to init USB support and thus cannot find the device to boot from, but those seem to be the exception not the norm.

    chewitt Could you recommend any image and dtb file for my TV box? I haven't followed the developments for quite some years and would like to give it another try.

    My box is an A95x F3 air which has an S905x3 , LED clock display, RGB lights and MT7668 wifi chip.

    If the driver isn't included in upstream is it possible to manually compile it and how?
    I found the following links related to the driver:
    https://download.csdn.net/download/liush…i9pc3N1ZXMvMTU3

    https://github.com/ophub/amlogic-…7663.driver.zip

    https://lore.kernel.org/all/CAO3ALPytX…il.gmail.com/T/

    This guy figured out how to compile the driver for a recent kernel. Can anyone explain what he did?
    https://www.spinics.net/lists/linux-bluetooth/msg103048.html

    Edited once, last by em0ne (January 18, 2024 at 1:23 PM).

    The A95X-F3 has an upstream device-tree file (authored by me) so the current AMLGX image is usable; with caveats that accompany G12A/B and SM1 boards (read the LE11.x release notes for the info, nothing changed). Read the Amlogic page in the wiki for install instructions: https://wiki.libreelec.tv/hardware/amlogic.

    WiFi and BT won't work due to the MT7668 chip having no upstream drivers. The FireTV driver from the kernel mailing list Q&A is here: https://github.com/chewitt/MT7668 but I could never get it to compile in a usable form. Hence the repo is untouched and I archived it a while back to drop a hint to people that kept cloning the driver and asking how to compile it. I had a brief look at the ophub driver and that looks similarly bad to the one I found before. The guy that "figured it out" doesn't sound like he really figured it out either. I can't see an email for him, but his name is reasonably destinctive and Google shows the same name associated with some postings to kernel and freedesktop related things so he's probably traceable if you ask around.

    I never looked into LED lights on those boxes either. The clock display might be do-able but for the light ring .. I'd need to see the original Android dts file or decompiled dtb to comment further.

    I pushed an update to my test share which includes a fix for 4K VP9 @ 59.94/60 content on GXL/GXM (and newer) boards. Previously dmesg would show an invalid clock rate error and HDMI sync would be lost until playback was cancelled. That should help YouTube playback as the 4K default seems to be [email protected] for most media.

    jim_p there are also driver bumps for RTL8189ES/FS that might prevent the warning splats in dmesg? - please test.