jock

@ROOD your board (YX_RK3328) should be already supported by rk3318-box image. The "marketing name" of the TV Box means nothing, the only important discriminant is the board name printed on the PCB. Anyway you made two mistakes: took the image from an unkown source burnt a non-working image in eMMC Now your only chance to restore board functionality is to go in maskrom mode: eMMC is always first priority, so you don't have any chance to boot from SDcard if a image that does not boot is stored in eMMC.

Not to hijack the thread, but there is a much better alternative driver rather than OpenVFD, you can take a look here

Well, not really. SOLID does not apply just to object oriented languages like C# or C++. They are principles, so by definition they always apply. Also note that the linux kernel is written in C (a non object oriented language) but most of the Linux kernel is indeed object oriented. It's a common mistake to think that OO software can be written with OO languages, but it is absolutely not true. OO languages makes easier and safer to handle OO principles, but you can apply the same OO principles even with C language. Getting back from philosophycal, I also agree that a module parameter is the right choice to fulfill what you need and keep the separation of concerns. Perhaps you may also go with an udev rule. I was wondering though if you want the led panel to show the string at the very power up of the board because you mentioned the stock firmware. In that case (the stock firmware) the boot loader (u-boot) is turning on the led panel chip and instructing it to show the boot string (via a driver, or even just tinkering directly with the I²C bus). Later the kernel boots and loads its own driver, and then changes the boot string with the clock or with the moving segments or whatever...

I would say, and I bet @Jean-Francois Lessard agrees, that is not a matter of "messing with my code" thing, but rather a matter of principles. A kernel module is supposed to do a kernel module or, In simple words, it interfaces the kernel with the hardware. The kernel means are to give the userspace a way to access the hardware. The module should not run any user business (eg: display something user-specific at boot/shutdown/whatever), but should in turn provide a service for the userspace to control the hardware. That's a simple principle in software engineering that's called "separation of responsibility", and is generally considered a good habit to agree with that principle. By the way, if your board has one or more colored leds, they are generally set up in the device tree and access basic GPIO system, so they are turned on as soon as the kernel starts.

@dfahren tm16xx driver from @Jean-Francois Lessard is already integrated in the armbian rockchip64 family patches, this is a typical device tree overlay configuration you can take as an example that is, by the way, basically similar to my (discontinued) driver

@paradigman try looking here for latest rockchip64 kernel headers, but I'm not sure the version will match. Usually I install the kernel headers via regular apt, so don't know if armbian-config is ok or is broken.

@Constantin Gatej added on first page, thanks!

@Christopher Ruehl thanks for looking into this issue, I will incorporate the fixes into mainline armbian as soon as possible!

@Christopher Ruehl congratulations, great finding! That could explain some of the compatibility issues people is still having. I have a couple of questions: 1) what is the purpose of enabling the internal gpio pull-up when there already is the external pull-up? 2) do you have any suggestions to apply to existing device tree to enhance compatibility for other users? Exchanging SMD components on the board is out of the league for most people Thanks!

@A t sorry but I have no idea, no logs, no board info and photos couldn't help understand what the problem may be

Find a possible IP address for the box and try to access via SSH, or use the UART serial to debug the issue

There is a driver for ssv6051p in the armbian images, but if the kernel does not see the chip it is probably turned off. You need to set the right led-conf using rk322x-config script that enables the GPIO to turn on the wifi chip. MX10 is the tv box commercial name, but it tells nothing about the board which is inside: you have to look for the signatures on the board itself and see if you have a match with the led-conf proposed by rk322x-config

Yes, the same. Technically the SoC enters in maskrom mode when all the boot devices fails. Gating the eMMC will make the SoC boot from sdcard, if there's no sdcard, the SoC will attempt other boot options programmed in its ROM. Finally, if nothing is available, will sit and wait in maskrom mode.

Yes, ffmpeg needs some patches (borrowed from LibreELEC project) to add support for v4l2request, which is in turn the interface for the kernel to use the hardware decoders on allwinner and rockchip platforms.

exactly; shorting the clock pin will make the eMMC disappear from the system, hence the SoC will boot from sdcard

Just a note: there is no modified mpv. Ubuntu Jammy just requires an updated version which my repository provides. Debian bookworm works fine with the packaged version it comes with.

As said, you can either remove the eMMC or short the clock pin to ground. Otherwise learn how the rockchip vendor boot happens and use the multitool binaries to hack the armbian boot

@ToShuk it depends upon a tree of factors. mmc is totally invisible to the SoC The the SoC will automatically look for the sdcard - just plug armbian in the sdcard and it will work the above won't work if the sdcard slot is attached to the sdmmc_ext controller of the SoC; this depends on how the manufactured designed the board and, if this is your case, your only solution is to replace the onboard eMMC mmc is stuck in read-only mode the SoC will indeed execute the bootloader, if any was present in the eMMC when it went read-only If the existing bootloader is the stock one, then you're stuck with the possibilities of the stock bootloader (eg: see the rockchip boot process wiki page): multitool will boot, but not plain armbian images; tinkering with armbian images bootloader will let them boot though. You can either short permanently the eMMC clock pin to ground, so the eMMC is clock gated and actually excluded from the system, or phisically remove the eMMC chip, so you can reconduct to the "totally invisible" case above if the existing bootloader is the one provided with armbian, the you're lucky and it should be possibile to boot from sdcard and USB

@Francisco Hasuky it is a known problem: some boards are sent to "suspension" after one minute by the proprietary Trust OS; you need to build u-boot with an opensource Trust to overcome the problem. At the moment I'm unable to build an image or provide the binary, but will do in the forthcoming hours/days @Jota Ce if you got ssv6051p chip, then you should be ok using any recent armbian image with kernel 6.12 (you could upgrade to edge kernel via armbian-config, or use the beta.armbian.com APT repository). Also kernel 6.6 should work with ssv6051p because the (horrible) driver has been ported from 4.4 kernel.

@Riyadh Zaman from the first page:

@Fcn3 this what you can expect from such hardware. There is a chromium fork around with patches to allow v4l2request accelerated video decoding, but has never been tested on rk3318 and probably you have to built it yourself.

Hello @Francisco Hasuky, does the HDMI turns off but the board is still responsive or the board just drops dead?

Hello, exactly the same instruction, yes

@Budi Pekerti boot via sdcard with multitool, mount the block device where you installed armbian and then remove the overlays=... line from /boot/armbianEnv.txt If you don't know how to do this, you have to reinstall from scratch

Do you mean the led-conf9 device tree overlay? Does the wifi works even without this overlay applied?