Flashing the NVIDIA Jetson (JetPack 6.x)

Refer to the Datasheets and install the Jetson Orin NX module, wiring and power to the Hadron carrier. Refer the Official Page For the download of the latest official BSP (Board Support Package) and release notes which contains the firmware install instructions.

Note: These steps target Jetson Orin NX/Nano on Connect Tech carriers (Hadron, etc.), Jetson Linux R36.3. Adjust versions as needed. Tested for CT Hadron NGX027 with the Jetpack 6.2.1 - L4T r36.4.4 (BSP Version: ORIN-NX-NANO-36.4.4 V002)

Prerequisities:

x86/x64 based host machine running Ubuntu 20.04 or 22.04
JetPack 6.2.1 / L4T 36.4.4 installed (Intstructions follows)
Orin™ NX or Orin™ NX module
Connect Tech Orin™ NX/Orin™ NX Carrier
USB A - USB A Cable for flashing

Install Wifi and SSD

Install the wifi and ssd cards to the PCI mini E and M slots respectively.

Powering up the carrier

The 6S Lipo can be used to power up the board directly. Install a 60V capacitor between the terminals of the main power. Alternatively a AC DC adapter with min 40 W (10-60V) can be used. Get the board ready to be powered up for he flashing of firmware.

Hadron Carrier

What you’ll do

Prepare a host machine and download JetPack BSP + sample rootfs
Install the CTI BSP on the host
Put the Jetson in recovery mode and flash (auto or manual)
Perform first boot steps and basic post-setup

1) Get JetPack (Driver Package + RootFS)

1) Create a working directory, referred to as BSP_ROOT below.

2) Open the Jetson Linux R36.4.4 downloads and notes:

Driver page: https://developer.nvidia.com/embedded/jetson-linux-r3644

3) Download both for Orin (LT4):

Driver Package (BSP): Jetson_Linux_R36.4.4_aarch64.tbz2
Sample Root Filesystem: Tegra_Linux_Sample-Root-Filesystem_R36.4.4_aarch64.tbz2

4) Place both files under BSP_ROOT, then extract them:

cd <BSP_ROOT>
sudo tar -jxf Jetson_Linux_R36.4.4_aarch64.tbz2
sudo tar -C Linux_for_Tegra/rootfs/ -xjf Tegra_Linux_Sample-Root-Filesystem_R36.4.4_aarch64.tbz2

After extraction, you should have <BSP_ROOT>/Linux_for_Tegra populated.

2) Install the CTI BSP

If you used NVIDIA SDK Manager, BSP_ROOT is typically:

~/nvidia/nvidia_sdk/<JetPack_Version>_Linux_JETSON_NX_ORIN_TARGETS/
or ~/nvidia/nvidia_sdk/<JetPack_Version>_Linux_JETSON_NANO_ORIN_TARGETS/

If you manually downloaded from the Embedded Download Center, use the folder you created as BSP_ROOT.

1) Copy the CTI package to Linux_for_Tegra:

cp CTI-L4T-ORIN-NX-NANO-36.4.4-V002.tgz BSP_ROOT/Linux_for_Tegra

2) Extract it and run the installer:

cd <BSP_ROOT>/Linux_for_Tegra
sudo tar -xzf CTI-L4T-ORIN-NX-NANO-36.4.4-V002.tgz
cd CTI-L4T
sudo ./install.sh
cd .. # back to Linux_for_Tegra

The CTI-L4T BSP is now installed on the host, and you can flash Orin NX/Nano.

3) Put the device in Recovery Mode

1) Connect the carrier to your host via USB per the carrier manual. 2) Ensure an NVMe M.2 SSD is installed. 3) Follow the carrier instructions to enter recovery (button/strap sequence).

Tip: You can verify via lsusb for an NVIDIA device in recovery.
Bus 001 Device 024: ID 0955:7323 NVIDIA Corp. APX

4) Flashing options

You can use either the CTI automated script or the manual NVMe flash script.

cd <BSP_ROOT>/Linux_for_Tegra
sudo ./cti-flash.sh

Follow the menu to select your module and board configuration. Flashing will begin.

Option B: Manual NVMe flash

Do not include the .conf extension in <config>.

cd <BSP_ROOT>/Linux_for_Tegra
sudo ./cti-nvme-flash.sh cti/<module>/<boardname>/<config>
# <module> is one of: orin-nx | orin-nano

Examples:

./cti-nvme-flash.sh cti/orin-nx/boson/base
./cti-nvme-flash.sh cti/orin-nano/boson/base
SUPER_MODE=1 ./cti-nvme-flash.sh cti/orin-nx/hadron/base

When flashing completes, the device will reboot.

Flash is successful
Reboot device
Cleaning up...
Log is saved to

5.1) Troubleshoot Flashing

ISSUE1 : A USB timeout error may occur during flashing. The following error indicates that your flash host’s USB port is not enabled:

[ 0.1172 ] Sending bct_br
[ 0.1603 ] ERROR: might be timeout in USB write.
Error: Return value

1. First try changing the USB port:

Move to a different USB port, if available.
Remove any USB-C docks if connected
Power cycle the board and retry flashing.

2. If that doesn’t work, try disabling autosuspend:

To disable autosuspend on your host’s USB ports, run the following command.

sudo bash -c 'echo -1 > /sys/module/usbcore/parameters/autosuspend'

Power cycle the board and retry flashing.

ISSUE2 : Cnnot mount the NFS server

12:52:45.297 - Error: Flash failure. 
12:52:45.300 - Error: Either the device cannot mount the NFS server on the host or a flash command has failed. Check your network setting (VPN, firewall,...) to make sure the device can mount NFS server. Debug log saved to /tmp/tmp.2HZma7rFSi. You can access the target's terminal through "sshpass -p root ssh root@fc00:1:1:0::2"

1. Disable VPN completely

CTI initrd flashing fails 100% if a VPN is on.

2. Restart NFS and RPC services

sudo systemctl stop rpcbind rpcbind.socket nfs-kernel-server
sudo systemctl start rpcbind rpcbind.socket nfs-kernel-server

Check status:

systemctl status nfs-kernel-server

3. Ensure firewall isn’t blocking NFS (especially IPv6)

Temporarily disable ufw (if running):

sudo ufw disable

5) First boot and basic setup

Connect via serial console (recommended)

Guide: https://jetsonhacks.com/2019/04/19/jetson-nano-serial-console/

Complete the initial setup prompts. You can skip network here and do it below.

System Configuration
         ┌────┤ License For Customer Use of NVIDIA Software ├─────┐
         │                                                        │
         │ Welcome to Jetson Initial Configuration                │
         │                                                        │
         │                         <Ok>                           │
         │                                                        │
         └────────────────────────────────────────────────────────┘

Join Wi‑Fi via nmcli

nmcli dev wifi rescan
nmcli dev wifi connect <SSID> password <PASSWORD>

SSH in

ssh <user>@<device-ip>

6) Useful post-setup

Setup Git

git config --global user.name "John Doe"
git config --global user.email johndoe@example.com

Increase Swap

You may need to increase swap depending on he RAM of your Jetson module Follow This Guide to upgrade swap memeory.

Install Jetpack extras and Jtop

sudo apt install nvidia-jetpack
sudo pip3 install -U jetson-stats
mkdir hacks && cd hacks
git clone https://github.com/jetsonhacks/jetson-jtop-patch.git
cd jetson-jtop-patch
chmod +x apply_jtop_fix.sh
./apply_jtop_fix.sh

Once pip installs the jetson-stats package, you will need to logout/login or reboot the Jetson.

After rebooting, to run jtop open a Terminal and execute: jtop

NoMachine (remote desktop)

To setup remote desktop to the headless system for debugging follow this guide

wget https://web9001.nomachine.com/download/9.2/Arm/
sudo dpkg -i nomachine_9.2.18_3_arm64.deb 
sudo systemctl disable gdm3 --now
sudo /usr/NX/bin/nxserver --restart

Change power mode

sudo nvpmodel -m <mode-number>

Reference: https://forums.developer.nvidia.com/t/power-mode-in-terminal/287224

Install VSCode

cd hacks
git clone https://github.com/JetsonHacksNano/installVSCode.git
cd installVSCode
./installVSCodeWithPython.sh

Cuda and cudnn

If cuda packages are not installed. ls /usr/local | grep cuda returns nothing

sudo apt install cuda-toolkit-12-6

if cudnn not installed. dpkg -l | grep cudnn returns nothing, Get Cudnn

wget https://developer.download.nvidia.com/compute/cudnn/9.16.0/local_installers/cudnn-local-tegra-repo-ubuntu2204-9.16.0_1.0-1_arm64.deb
sudo dpkg -i cudnn-local-tegra-repo-ubuntu2204-9.16.0_1.0-1_arm64.deb
sudo cp /var/cudnn-local-tegra-repo-ubuntu2204-9.16.0/cudnn-*-keyring.gpg /usr/share/keyrings/
sudo apt-get update
sudo apt-get -y install cudnn9-cuda-12

Setup `.bashrc`

Add following lines to your .bashrc to link important cuda libraries

export CUDA=12.6
export PATH=/usr/local/cuda-$CUDA/bin${PATH:+:${PATH}}
export CUDA_PATH=/usr/local/cuda-$CUDA
export CUDA_HOME=/usr/local/cuda-$CUDA
export LIBRARY_PATH=$CUDA_HOME/lib64:$LIBRARY_PATH
export LD_LIBRARY_PATH=/usr/local/cuda-$CUDA/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda/extras/CUPTI/lib64:$LD_LIBRARY_PATH
export NVCC=/usr/local/cuda-$CUDA/bin/nvcc
export CFLAGS="-I$CUDA_HOME/include $CFLAGS"
export LD_LIBRARY_PATH=/usr/local/cuda-$CUDA/targets/aarch64-linux/lib:$LD_LIBRARY_PATH
export CPATH=/usr/local/cuda-$CUDA/targets/aarch64-linux/include:$CPATH

Source your .bshrc

Verify Cuda and CUDNN

echo '#include <cuda_runtime.h>
#include <cudnn.h>
#include <iostream>
int main() {
    int cudaVersion;
    cudaRuntimeGetVersion(&cudaVersion);
    std::cout << "CUDA runtime version: " << cudaVersion/1000 << "." << (cudaVersion%1000)/10 << std::endl;
    std::cout << "cuDNN version: " << CUDNN_MAJOR << "." << CUDNN_MINOR << "." << CUDNN_PATCHLEVEL << std::endl;
    return 0;
}' > verify_cuda_cudnn.cpp && \
g++ verify_cuda_cudnn.cpp \
    -I/usr/local/cuda-12.6/targets/aarch64-linux/include \
    -L/usr/local/cuda-12.6/targets/aarch64-linux/lib \
    -lcudart -lcudnn -o verify_cuda_cudnn && \
./verify_cuda_cudnn

Fix Snap apps

If Snap apps like Chromium is not working, ref

snap download snapd --revision=24724
sudo snap ack snapd_24724.assert
sudo snap install snapd_24724.snap
sudo snap refresh --hold snapd

Mount a USB drive

https://linuxconfig.org/howto-mount-usb-drive-in-linux

References

1) Hadron carrier: https://connecttech.com/product/hadron-carrier-for-nvidia-jetson-orin-nx/ 2) CTI L4T release notes (36.3.0): https://connecttech.com/ftp/Drivers/L4T-Release-Notes/Jetson-Orin-NX-Orin-Nano/ORIN-NX-NANO-36.3.0.pdf 3) CTI knowledge base: https://connecttech.com/resource-center/kdb373/ 4) torchvision dev install: https://github.com/pytorch/vision/blob/main/CONTRIBUTING.md#development-installation

Continue

← Back: Hardware Assembly
Next → Software Setup

← Back to Home View on GitHub