We have custom carrier board with an RGMII interface. We are using L4T 35.1. The RGMII interface is finally working. For what it is worth, the key piece is to remove the “interrupts” and “interrupt-parent” entry from the phy device tree node, as suggested by the patch in this forum post Orin rgmii cann't work. (We still haven’t figured out why that works).
Regardless, we have the RGMII port working with Orin from the main application kernel.
Our problem is that we still have a very slow boot process with the UEFI bootloader hanging here:
add-symbol-file /build/git/edkrepo/nvidia-uefi/Build/Jetson/DEBUG_GCC5/AARCH64/Silicon/NVIDIA/Drivers/EqosDeviceDxe/EqosDeviceDxe/DEBUG/EqosDeviceDxe.dll 0x8044D0000
Loading driver at 0x008044CF000 EntryPoint=0x008044E83A0 EqosDeviceDxe.efi
PROGRESS CODE: V03040002 I0
SNP:PHY: PhyDxeInitialization () Failed to reset Phy
DeviceDiscoveryBindingStart, driver returned Device Error to start notification
It takes about 50 seconds before timing out. This unacceptably lengthens our boot time.
Does anyone have any suggestions for how to make this step recognize the PHY or to disable the EQOS during this step but to still have it available later?