Hardware

NVIDIA Jetson

What is NVIDIA Jetson series?

The NVIDIA Jetson series is a range of embedded computing boards that are designed for edge computing applications. They are powered by NVIDIA GPUs and are capable of running deep learning models and AI algorithms at the edge. The Jetson series includes the Jetson Nano, Jetson Xavier NX, and Jetson AGX Xavier.

Serial

To observe the U-boot console, you need to connect a TTL to USB cable. The connection method is as shown in the following:

  • Jetson (UART RXD) -> TXD

  • Jetson (UART TXD) -> RXD

  • Jetson (GND) -> GND

  • The baud rate is 115200.

Flash Jetson Nano

Download SDK Manager here

Jetson SDK Manager

Boot from SD Card

Set the Jetson Nano enter recovery mode then connect it to the host PC. Flash an Jetson OS, do not check the SDK option.

This should be done on a Ubuntu Host PC:

  1. Install dtc tool:
> sudo apt-get install device-tree-compiler
  1. Decompile the dts file (use your path according to the Jetpack 4.x)
> cd ~/nvidia/nvidia_sdk/JetPack_4.6.5_Linux_JETSON_NANO_TARGETS/Linux_for_Tegra/kernel/dtb/
> dtc -I dtb -O dts -o tegra210-p3448-0002-p3449-0000-b00.dts tegra210-p3448-0002-p3449-0000-b00.dtb
  1. Modify the tegra210-p3448-0002-p3449-0000-b00.dts file, you should modify the codes as below which are marked in red.
sdhci@700b0400 {
   compatible = "nvidia,tegra210-sdhci";
   reg = <0x0 0x700b0400 0x0 0x200>;
   interrupts = <0x0 0x13 0x4>;
   aux-device-name = "sdhci-tegra.2";
   iommus = <0x30 0x1b>;
   nvidia,runtime-pm-type = <0x0>;
   clocks = <0x26 0x45 0x26 0xf3 0x26 0x136 0x26 0xc1>;
   clock-names = "sdmmc", "pll_p", "pll_c4_out2", "sdmmc_legacy_tm";
   resets = <0x26 0x45>;
   reset-names = "sdhci";
   status = "okay"; #change this to okay
   tap-delay = <0x3>;
   trim-delay = <0x3>;
   mmc-ocr-mask = <0x3>;
   max-clk-limit = <0xc28cb00>;
   ddr-clk-limit = <0x2dc6c00>;
   bus-width = <0x4>;
   calib-3v3-offsets = <0x7d>;
   calib-1v8-offsets = <0x7b7b>;
   compad-vref-3v3 = <0x7>;
   compad-vref-1v8 = <0x7>;
   pll_source = "pll_p", "pll_c4_out2";
   ignore-pm-notify;
   cap-mmc-highspeed;
   cap-sd-highspeed;
   nvidia,en-io-trim-volt;
   nvidia,en-periodic-calib;
   cd-inverted;
   wp-inverted;
   pwrdet-support;
   nvidia,min-tap-delay = <0x6a>;
   nvidia,max-tap-delay = <0xb9>;
   pinctrl-names = "sdmmc_schmitt_enable", "sdmmc_schmitt_disable", "sdmmc_clk_schmitt_enable", "sdmmc_clk_schmitt_disable", "sdmmc_drv_code", "sdmmc_default_drv_code", "sdmmc_e_33v_enable", "sdmmc_e_33v_disable";
   pinctrl-0 = <0x8c>;
   pinctrl-1 = <0x8d>;
   pinctrl-2 = <0x8e>;
   pinctrl-3 = <0x8f>;
   pinctrl-4 = <0x90>;
   pinctrl-5 = <0x91>;
   pinctrl-6 = <0x92>;
   pinctrl-7 = <0x93>;
   vqmmc-supply = <0x3b>;
   vmmc-supply = <0x4c>;
   
   # Add these lines
   cd-gpios = <0x5b 0xc2 0x0>;
   sd-uhs-sdr104;
   sd-uhs-sdr50;
   sd-uhs-sdr25;
   sd-uhs-sdr12;
   # end of added lines
   mmc-ddr-1_8v;
   
   # Add these two lines
   no-mmc;
   uhs-mask = <0xc>;
   # end of added lines
   
   linux,phandle = <0xba>;
   phandle = <0xba>;
  1. Compile the dtb file
> dtc -I dts -O dtb -o tegra210-p3448-0002-p3449-0000-b00.dtb tegra210-p3448-0002-p3449-0000-b00.dts
  1. Flash the OS to Jetson Nano emmc again with the following command:
> cd ~/nvidia/nvidia_sdk/JetPack_4.6.5_Linux_JETSON_NANO_TARGETS/Linux_for_Tegra/
> sudo ./flash.sh jetson-nano-emmc mmcblk0p1

The flash.sh will flash the os to emmc again.

  1. When the flash finished, Jetson Nano will reboot shortly. Serial to the Jetson Nano and finish the basic Ubuntu configuration:
> sudo screen /dev/ttyACM0 115200

The following steps should be done on Jetson Nano.

  1. After second reboot, ssh to Jetson Nano via network. Check the SD card device:
> ls /dev/mmc*

You will see a new device like mmcblk1

  1. Create partition and format mmcblk1:
> sudo fdisk /dev/mmcblk1

In the fdisk interface:

  • Press p to print the current partition table (optional but useful to see existing partitions).
  • Press n to create a new partition.
  • Choose p for a primary partition.
  • Specify the partition number (1 for the first partition if it’s empty).
  • Choose the starting and ending sectors, or press Enter to use the default values, which will use all available space.

Write the partition table:

  • Press w to write the changes and exit fdisk.
  1. Format mmcblk1p1
> sudo mkfs.ext4 /dev/mmcblk1p1
  1. Mount the SD Card:
> sudo mkdir /mnt/sdcard
> sudo mount /dev/mmcblk1p1 /mnt/sdcard

You first need to flash the eMMC normally with SDK manager. This process will install bootloader and firmware to the module.

  1. Copy Root Filesystem to SD Card
> sudo rsync -axHAWX --numeric-ids --info=progress2 --exclude=/proc / /mnt/sdcard
  1. Edit fstab file from the SD card to make it mount automatically after reboot:
> sudo nano /mnt/sdcard/etc/fstab

For example:

/dev/mmcblk1p1  /  ext4  defaults  0  1
  1. Modify Boot Sequence

What you need to do is to add a new boot entry, change root to the desired device (e.g. "/dev/mmcblk1p1") and update the default entry to be "microsd". The following is an example of the boot configuration file:

> sudo nano /boot/extlinux/extlinux.conf

For example:

TIMEOUT 30
DEFAULT microsd

MENU TITLE L4T boot options

LABEL microsd
      MENU LABEL microsd kernel
      LINUX /boot/Image
      INITRD /boot/initrd
      APPEND ${cbootargs} quiet root=/dev/mmcblk1p1 rw rootwait rootfstype=ext4 console=ttyS0,115200n8 console=tty0 OS=l4t fbcon=map:0 net.ifnames=0

LABEL primary
      MENU LABEL primary kernel
      LINUX /boot/Image
      INITRD /boot/initrd
      APPEND ${cbootargs} quiet root=/dev/mmcblk0p1 rw rootwait rootfstype=ext4 console=ttyS0,115200n8 console=tty0 OS=l4t fbcon=map:0 net.ifnames=0 

# When testing a custom kernel, it is recommended that you create a backup of
# ...
# ...
  1. Now reboot and verify your rootfilesystem is on the SD card.
> df -h
  1. Use emmc as a data storage
> sudo mkfs.ext4 /dev/mmcblk0p1
> sudo mkdir /mnt/emmc
> sudo mount /dev/mmcblk0p1 /mnt/emmc

Edit /etc/fstab and add this line:

/dev/mmcblk0p1  /mnt/emmc  ext4  defaults  0  2

Install CUDA

Verify the Jetson device is supported by CUDA:

> sudo lshw -C system

Get version:

> dpkg-query --show nvidia-l4t-core

> wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/sbsa/cuda-ubuntu1804.pin
> sudo mv cuda-ubuntu1804.pin /etc/apt/preferences.d/cuda-repository-pin-600
> wget https://developer.download.nvidia.com/compute/cuda/11.3.1/local_installers/cuda-repo-ubuntu1804-11-3-local_11.3.1-465.19.01-1_arm64.deb
> sudo dpkg -i cuda-repo-ubuntu1804-11-3-local_11.3.1-465.19.01-1_arm64.deb
> sudo apt-key add /var/cuda-repo-ubuntu1804-11-3-local/7fa2af80.pub
> sudo apt-get update
> sudo apt-get -y install cuda

In case you face any dependency issue while executing the last command, run the below fix:

> sudo apt-get -o dpkg::Options::="--force-overwrite" install --fix-broken

Jetson Nano Install CUDA

Setting up PATH variable:

> cd /usr/local/cuda
> export PATH=/usr/local/cuda-11.3/bin${PATH:+:${PATH}}

Try to run nvcc:

> nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2021 NVIDIA Corporation
Built on Mon_May__3_19:15:10_PDT_2021
Cuda compilation tools, release 11.3, V11.3.109
Build cuda_11.3.r11.3/compiler.29920130_0
Previous
ESP32
Next
KiCAD