fly_kt

Thanks. It confirms the issue and that 100Mb is working without issues. The next post contains the link to github with details and some keywords I used in my search before posting. Do you know if there are any plans to provide that patch in a current kernel within the next months or is the only option to switch to edge for a longer time?

Here is the link to the armbianmonitor -u output https://paste.armbian.com/diyujadoju Please note, that I patched my dtb, therefore the output shows the settings that work for me and not the original values.

Hello, I've installed armbian with Debian Trixie on my Odroid-M1 to replace the Ubuntu installed via petitboot. Everything was running smoothly except the networking part which was very slow (with luck some 100kbit/s) and has shown very high packetloss, even while the switch and ethtool report absolutely no problems and replacing cables and switching ports didn't help. Switching to 100Mbit/s was working fine, no loss anymore, switching back to 1G caused massive packet loss. With some help from chatgpt and another M1 I own that was still running the hardkernel/petitboot installed kernel and dtb I could trace the issue down the RGMII timings provided in the armbian provided dtb compared to the hardkernel dtb. My odroid-m1 with 4G RAM running armbian used this (from dmesg) rk_gmac-dwmac fe2a0000.ethernet: init for rgmii with delay (tx: 0x4f, rx: 0x2d) The other one with 8G and hardkernel ubuntu (also from dmesg): rk_gmac-dwmac fe2a0000.ethernet: TX delay(0x35). rk_gmac-dwmac fe2a0000.ethernet: RX delay(0x2d). After installing busybux with devmem I could change the settings on the fly busybox devmem 0xfdc60380 32 0x007F0035 # restoring old value # busybox devmem 0xfdc60380 32 0x007F004F The previous values and new value can be check with this command busybox devmem 0xfdc60380 32 Before the change it had shown the old values (0x00002D4F), and afterwards 0x00002D35 (RX and TX). With that settings via devmem I could switch the packet loss on and off. Am I'm the only one with that issue or does it make sense to create bug to get it fixed? There must be a reason why armbian is using different values in the odroid-m1 dtb files than hardkernel itself. Thanks in advance for any replies. Karsten