Board: HINLINK H68K (community-maintained) Image: Armbian_community_26.11.0-trunk.108_Hinlink-h68k_trixie_current_6.18.57_minimal (kernel 6.18.57-current-rockchip64) What happened? 1. The port labelled eth0 never works. At boot its link LED blinks once and stays dark. mii-tool end1 reports "No MII transceiver present!" and bringing it up logs "Failed to reset the dma". I observed the problem on two separate H68K units. Once that was resolved, I further noticed that... 2. With a good cable, eth0 and eth1 link at 1000 Mbit/s but can't receive. Received frames land in mmc_rx_crc_error (on eth1, all of them), so DHCP never completes. With a cable that negotiates 100 Mbit/s, both ports work. Expected: both 1G ports work at 1000 Mbit/s, as the 2.5G ports (eth2/eth3) do. Root Causes: Both are in patch/kernel/archive/rockchip64-*/dt/rk3568-hinlink-h68k.dts: 1. &i2s1_8ch (RK809 audio) has no pinctrl-0, so it takes the full 8-channel i2s1m0 pin set. That set includes GPIO1_B0, which is gmac1's snps,reset-gpio. Once the i2s driver probes, the eth0 PHY is held in reset: pin 40 (gpio1-8): fe410000.i2s gpio1:40 function i2s1 group i2s1m0-sdo1 2. Both gmacs are rgmii-id (the PHY adds the delays) but also set tx_delay = <0x3c>; rx_delay = <0x2f>;. The rockchip64-6.18 tree carries the WIP patch "dwmac-rk: Allow use of rx/tx delayline for rgmii-id modes", which applies those MAC delays as well, so the delay is applied twice: init for rgmii-id with delay (tx: 0x3c, rx: 0x2f) How to reproduce? Boot the image above on an H68K. Plug a cable into the port labelled eth0: no link. Plug a known-good cable into eth1 with a gigabit partner: it links at 1000 Mbit/s but gets no DHCP lease, and ethtool -S end0 | grep rx_crc climbs. Fix: PR: https://github.com/armbian/build/pull/11251. It restricts i2s1 to the 4 pins the codec uses (same as mainline's rk3568-rock-3a.dts) and drops tx_delay/rx_delay. I tested it on 6.18.57 with an equivalent overlay: both 1G ports pass traffic at 1000 Mbit/s with no CRC errors, and analog audio still plays.