Build Kernel from Scratch (JetPack 4.6 on NRU-50)

A. Prepare Kernel Building Environment

In a nutshell, you will build the kernel image on your host machine, an x86 computer with Ubuntu, NVIDIA SDKManager, x86 to aarch64 cross-compiler installed. You can download the kernel source from NVIDIA, modify, compile it, then flash the compiled result to NRU.

1. Download NVIDIA SDKManager

https://developer.nvidia.com/nvidia-sdk-manager

2. Install JetPack 4.6 for Xavier NX NVIDIA SDKManager

After installation, you will see a L4T folder.

# The default L4T folder path for JetPack 4.6 of Xavier NX is 
$HOME/nvidia/nvidia_sdk/JetPack_4.6_Linux_JETSON_XAVIER_NX_TARGETS/Linux_for_Tegra

3. Download Device Tree and Kernel Source from NVIDIA

export L4T=$HOME/nvidia/nvidia_sdk/JetPack_4.6_Linux_JETSON_XAVIER_NX_TARGETS/Linux_for_Tegra

cd $L4T

# "tegra-l4t-r32.6.1" menas JetPack 4.6, more JetPack and L4T mapping can be found on <https://developer.nvidia.com/embedded/jetpack-archive>
./source_sync.sh -k tegra-l4t-r32.6.1

Then you will find sources under L4T . The sourecs folder includes the kernel and device tree source code from NVIDIA.

root@nvidia-POC-500:~/nvidia/nvidia_sdk/JetPack_4.6_Linux_JETSON_XAVIER_NX_TARGETS/Linux_for_Tegra/sources# ll
total 16
drwxr-xr-x  4 root   root   4096 十一  5 10:46 ./
drwxrwxr-x 11 nvidia nvidia 4096 十一  5 10:40 ../
drwxr-xr-x  3 root   root   4096 十一  5 10:46 hardware/
drwxr-xr-x  5 root   root   4096 十一  5 10:46 kernel/

4. Download Toolchain to Compile aarch64 on x86 Computers

cd $HOME
wget http://releases.linaro.org/components/toolchain/binaries/7.3-2018.05/aarch64-linux-gnu/gcc-linaro-7.3.1-2018.05-x86_64_aarch64-linux-gnu.tar.xz

mkdir $HOME/l4t-gcc
cd $HOME/l4t-gcc

tar xf gcc-linaro-7.3.1-2018.05-x86_64_aarch64-linux-gnu.tar.xz

5. Create a Default Makefile Config

sudo -s

L4T=$HOME/nvidia/nvidia_sdk/JetPack_4.6_Linux_JETSON_XAVIER_NX_TARGETS/Linux_for_Tegra
TOOLCHAIN_PREFIX=$HOME/l4t-gcc/gcc-linaro-7.3.1-2018.05-x86_64_aarch64-linux-gnu/bin/aarch64-linux-gnu-
TEGRA_KERNEL_OUT=$L4T/build
KERNEL_MODULES_OUT=$L4T/sources/modules

cd $L4T

sudo make -C sources/kernel/kernel-4.9/ ARCH=arm64 O=$TEGRA_KERNEL_OUT tegra_defconfig

## Make kernel
### 1. change config if necessary
# vim $TEGRA_KERNEL_OUT/.config
### 2. make kernel
# make ARCH=arm64 O=$TEGRA_KERNEL_OUT LOCALVERSION=-tegra CROSS_COMPILE=${TOOLCHAIN_PREFIX} -j$(nproc)
### 3. the made kernel is under $TEGRA_KERNEL_OUT/arch/arm64/boot/Image

B. Modify Default Kernel Config

The default config file is under the following files.

$TEGRA_KERNEL_OUT/.config

Each configuration can be set as

• =y -> Put a specific feature in the built-in kernel image, which can't be removed or inserted dynamically

• =m -> Build a specific feature as a kernel module, which can be removed or inserted dynamically

• =n -> Not include a specific feature in the kernel image

For example, to enable GPS PPS service, we will put the following config as builtin

CONFIG_PPS=y
CONFIG_PPS_CLIENT_LDISC=y
CONFIG_PPS_CLIENT_GPIO=y

C. Compile Kernel

L4T=$HOME/nvidia/nvidia_sdk/JetPack_4.6_Linux_JETSON_XAVIER_NX_TARGETS/Linux_for_Tegra
TOOLCHAIN_PREFIX=$HOME/l4t-gcc/gcc-linaro-7.3.1-2018.05-x86_64_aarch64-linux-gnu/bin/aarch64-linux-gnu-
TEGRA_KERNEL_OUT=$L4T/build
KERNEL_MODULES_OUT=$L4T/build/modules
NRU50_dtb_name=nru50

cd $L4T

# in case of make clean
# make -C sources/kernel/kernel-4.9 clean

make -C sources/kernel/kernel-4.9/ ARCH=arm64 O=$TEGRA_KERNEL_OUT LOCALVERSION=-tegra CROSS_COMPILE=${TOOLCHAIN_PREFIX} -j$(nproc) 
# The compiled kernel image will be locate under 
# $TEGRA_KERNEL_OUT/arch/arm64/boot

D. Replace the Kernel Image

Based on personal experience, the kernel image might be damaged once I uploaded on to the cloud. So I would suggest a USB driver copy might be the safest way to move the compiled kernel image to NRU-50.

The location to put the compiled kernel image is:

/boot/Image