Posts by chewitt

    Code
    rmmod r8712u
    insmod /usr/lib/kernel-overlays/base/lib/modules/6.1.0-rc8/kernel/drivers/staging/rtl8712/r8712u.ko debug=0x1416

    Mind the line wrap ^ above; remove and reinsert the driver with a debug option and then see if anything more was printed to dmesg? - If yes, run "dmesg | paste" and share the URL. If nothing.. I'm out of ideas, other than removing the staging driver and digging up an older vendor driver. The TODO file in staging points to https://github.com/chunkeey/rtl8192su as an alternative; you can probably cp -R an existing realtek driver package.mk and swap the URLs/GitHash details to build it; if it builds.

    rellla The plumbing for hardware deinterlace in FFMpeg under V4L2 has been figured out (for RPi, and similar works with other devices) but right now there is no support for the hardware deinterlace IP on Amlogic SoCs in Linux, so FFMpeg will fall back to software (Yadif or Bob, I forget which) sampling. It's watchable for normal media but anything sporting with faster panning shots isn't so great. I'd pay someone to write the driver code if I could find someone to pay..

    Your box is running an unofficial OS image created by someone called "surkovalex" under contract with Eminent. The sources for the OS are posted here https://github.com/Eminent-Online/ although it's years since I bothered to look at what changes have been made (and I don't plan to start now). In short; we know nothing about Eminent boxes and we don't provide updated images for them. That said, you may be able to use the dtech 9.2.8 release which still uses the old vendor kernel and fixes a few things. If you need a newer Kodi version and aren't too demanding with codecs there is also an LE11 nightly AMLGX image which runs reasonably well on an S905X device; there are quite a few other S905X box devices based on the Amlogic reference design so you can probably find a device-tree file to use and get things working.

    If you are trying to boot an LE 9.2.8 image on a recent RPi4 this will not boot/work as the (older) firmware in the LE 9.2.8 RPi4 image does not support the (newer) hardware in recent RPi4 models. You will need to use LE10.x or LE11.x nightlies.

    So use LE 10.0.4, connect HDMI and Ethernet only, test again. If you see problems still, remove "quiet" from boot params in cmdline.txt and then look for errors on-screen. The failure to get time hints at an issue with NTP in your network (or lack of Ethernet) but this should not be fatal, the NTP check should timeout and boot should continue.

    Staging means the driver is still in development (this one for many years) so the driver code is located in a specific separate area of the kernel tree (filesystem) but otherwise the process for enabling the module is the same as any other; uncomment and setting CONFIG_R8712U=m or =y should be the only thing required to build the driver. In LE buildystem, make the change to https://github.com/LibreELEC/Libr…ch64.conf#L4917 then build the image again; the kernel defconfig change will be detected and only the kernel (and any packages depending on it) will be rebuilt so it's normally quick to respin the image. I can see that the USB IDs for your card are present https://cateee.net/lkddb/web-lkddb/R8712U.html so in theory it should probe and be active automatically. If it doesn't, there's an issue to be investigated.

    _emanuel_ The W400 dtsi that GT-King dts consumes from sets the SDIO reset GPIO to GPIOX_6 (low):

    linux/meson-g12b-w400.dtsi at master · torvalds/linux
    Linux kernel source tree. Contribute to torvalds/linux development by creating an account on GitHub.
    github.com

    but the Dreambox One dts in your GitHub repo sets the SDIO reset to GPIOA_11 (high) and there are pwm things being defined below:

    linux-meson64/dreamone.dts at master · emanuel4you/linux-meson64
    linux meson64. Contribute to emanuel4you/linux-meson64 development by creating an account on GitHub.
    github.com

    From pics I've seen the Dreambox uses a custom PCB, so it's probably not a reference design clone like Android boxes, although so much of the core peripherals are on the SoC (and thus have common config) that a box dts can boot the board. If the reset pins for SDIO aren't being driven, this would explain why there's no attempt to probe the bus in dmesg. The BT part is physically on the same chip, but driven separately so working BT and non-working SDIO is normal (possible).

    I'm not completely sure how to read the pwm changes and bits in the vendor kernel dts, but there are differences.

    Code
    LABEL LibreELEC
      LINUX /KERNEL
      FDT /rk3399-rock-pi-4b.dtb
      APPEND boot=UUID=0503-4902 disk=UUID=8ff85d86-0886-4367-affc-e7e92877481a quiet console=uart8250,mmio32,0xff1a0000 console=tty0 coherent_pool=2M cec.debounce_ms=5000 drm.edid_firmware=HDMI-A-1:edid/edid-HDMI-A-1.bin video=HDMI-A-1:D

    ^ edit /flash/extlinux/extlinux.conf and append it to the APPEND line

    As a general rule support for ARM boards is best on the latest kernel so LibreComputer are probably bumping to something newer than the stock 5.15 one. It means the driver in the image is more likely to be the staging driver we're interested to prove, but there's only one way to really find out.

    https://wiki.libreelec.tv/development-1/build-basics has some instructions for building. There are build-in-container options in the wiki too if that's easier on your current setup. It's quite a simple process.

    The intiial issue with that add-on will be that LE only uses Pulse audio for streaming output to BlueTooth speakers, we use alsa for all other output. It's possible to reconfigure LE to use Pulse though. The wiki describes some other configurations, see https://wiki.libreelec.tv/configuration/pulseaudio - not specifically your requirements, but at least explains what the config files and required touch points are to make an alternative configuration. There are probably threads in this forum about using Pulse too.

    The kernel in the Ubuntu 22.04 VMs that I have running is Linux 5.15.x and although I haven't checked, I would expect the distro to package the Realtek vendor drivers. We avoid those because they are a pain in the arse to maintain over time; they break with every major kernel bump (something we do more often than most distros) and need constant patching. Ubuntu has more people contributing to their sources and puts up with crap like that.

    The kernel in LE master (nightlies) is currently Linux 6.0.x which has a staging (testing/development) driver for that NIC, but the module is not enabled in our kernel defconfig so the driver will not be built and thus the card should not work in our nightlies.

    See: https://github.com/LibreELEC/Libr…ch64.conf#L4917

    The kernel version and defconfig used in balbes150 images is unknown to me and these days I have no idea where Oleg publishes sources to check. It's possible his images have the module enabled and it's possible the staging-quality driver doesn't work (or work great). It's possible he has added the out-of-tree vendor driver /shrug

    If you're confident with Debian etc. perhaps self-build your own LE master nightly image with the staging driver enabled. It might then be missing firwmware, but dmesg will reveal which file you need to pick from the upstream linux-firmware repo to test/prove it. That file might be the same filename as older vendor firmware, but there's probably alignment between driver and specific firmware versions.

    The single log that you shared only shows a booting kernel where a driver loads without errors - it's not hinting there's a problem.

    NB: Please keep the convo in this thread instead of scatter posting over several.