Issues with Jetson Nano Orin Super Dev Kit

Robotics & Edge Computing Jetson Systems Jetson Orin Nano
,
1

Cannot flash JetPack 6.2/6.2.2 to NVMe via SDK Manager or l4t_initrd_flash.sh — SSH always refused in initrd

Hardware

  • Board: Jetson Orin Nano Super Developer Kit 8GB (official NVIDIA kit)
  • Firmware: 36.4.3-gcid-38968081 (date 2025-01-08)
  • NVMe SSD: BIWIN CE430T5D100-512G (512GB, M.2)
  • microSD: SanDisk 128GB
  • Host PC: HP Victus laptop, Intel i5, Ubuntu 22.04 LTS (x86_64)
  • Cable: USB-C to USB-C, connected directly (no hub)

What I tried

Attempts 1–8 — SDK Manager JetPack 6.2.2 and 6.2, NVMe target, after wiping SSD

Every single attempt fails at the same point:

Device should have booted into initrd kernel now. However, the host cannot connect to its ssh server
Command failed: sshpass -p root ssh root@fc00:1:1:0::2

Additional error appearing consistently after SSH failure:

rpcbind: another rpcbind is already running. Aborting

I disabled and masked rpcbind completely:

sudo systemctl mask rpcbind
sudo systemctl mask rpcbind.socket
sudo killall rpcbind

But the SSH failure persists regardless.

Network interface analysis

dmesg shows RNDIS negotiates correctly:

usb 3-4: New USB device found, idVendor=0955, idProduct=7035
usb 3-4: Product: Linux for Tegra
rndis_host 3-4:1.0 usb0: register 'rndis_host', RNDIS device
rndis_host 3-4:1.0 enxXXXXXXXXXXXX: renamed from usb0

The interface gets an IPv6 address (fc00:1:1::1/64) and ping6 to fc00:1:1::2 responds correctly with 0% packet loss. However:

sshpass -p root ssh -6 -o StrictHostKeyChecking=no root@fc00:1:1::2

nmap shows all 1000 ports closed on fc00:1:1::2 — the initrd boots but SSH never starts.

Attempt 9 — l4t_initrd_flash.sh directly (bypassing SDK Manager)

Same result — RNDIS connects, ping responds, SSH refused.

Attempt 10 — SD card method with Balena Etcher (JetPack 6.2)

Flashed SanDisk 128GB microSD 3 times with Balena Etcher on Windows. The UEFI Boot Manager does not detect the SD card as a bootable device. Only the BIWIN NVMe appears in the boot list. Boot from file in UEFI shell shows the SD content (fs0) but there is no EFI/BOOT directory — only raw partition files.

Attempt 11 — SDK Manager JetPack 6.2, target_image mode

Fails at a different point now:

Errors were encountered while processing:
/opt/nvidia/l4t-packages/kernel/nvidia-l4t-kernel-headers_5.15.148-tegra-36.4.3-20250107174145_arm64.deb
command error code: 11

This happens consistently even after clearing ~/.nvsdkm and re-downloading.

I have followed all of the official and non-official tutorials out there, I bought my board from RS Eleectronics, and is an official carrier board too. I’m really perplexed since I’ve been at it for a couple of days to no avail, it really shouldn’t be this hard.

Is there a known issue with the initrd SSH not starting on certain host configurations? Is there a way to flash the NVMe without relying on SSH in the initrd (e.g. a direct write mode)? Any help would be appreciated.

5

Hi,

Please run the following commands under the Linux_for_Tegra directory.

sudo ./tools/kernel_flash/l4t_initrd_flash.sh --external-device nvme0n1p1 \
  -c tools/kernel_flash/flash_l4t_t234_nvme.xml -p "-c bootloader/generic/cfg/flash_t234_qspi.xml" \
  --showlogs --network usb0 jetson-orin-nano-devkit internal

And also record and upload the serial console log during flashing.

Thanks

6

I also tried running it though CLI, but i got the same result as with the SDK Manager.

After writing this post I noticed there was a new version for Windows for the SDK Manager, and even though I knew it would be WSL2 which would probably work even worse, it flashed it perfectly from my Host Win11 machine, absurd after trying all the other “correct and better tested” methods. I have it running perfectly now.

Thanks anyway for the quick help!