Real-time kernel and ipvlan support

[nvasudevan]

...
I raised this issue (#21) few weeks back, unfortunately I didn't respond in time. Apologies. But we have found a set of instructions that will enable real-time and ipvlan on seed studio boards. We just wanted to share it here for other users.

@alanmillard was able to enable real-time kernel and ipvlan kernel module for seeed studio board. Below are the details.

My setup:

Jetson Orin Nano 8GB - reComputer J3011 (J401 carrier board)
Host PC: Ubuntu 22.
Jetpack 6.1 (R36.4.0)

Please find below the instructions to enable RT and IPVLAN:

The instructions to prepare the host computer is what is outlined in the README (steps 1 - 6).

<Linux_for_Tegra>$ export ARCH=arm64
# set up cross compile path
<Linux_for_Tegra>$ export CROSS_COMPILE=~/l4t-gcc/aarch64--glibc--stable-2022.08-1/bin/aarch64-buildroot-linux-gnu-
cd <install-path>/Linux_for_Tegra/source

Enable real-time kernel

<Linux_for_Tegra>$ cd source/
<Linux_for_Tegra/source>$ ./generic_rt_build.sh "enable"
Enable RT kernel
PREEMPT RT config is set successfully!

Enable IPVLAN

<Linux_for_Tegra/source>$ cd kernel/kernel-jammy-src/
<Linux_for_Tegra/source/kernel/kernel-jammy-src>$ ./scripts/config --file "arch/arm64/configs/defconfig" --module CONFIG_IPVLAN

Build kernel

<Linux_for_Tegra/source>$ make -C kernel
<Linux_for_Tegra/source>$ export INSTALL_MOD_PATH=`realpath ../rootfs/`
<Linux_for_Tegra/source>$ sudo -E make install -C kernel
<Linux_for_Tegra/source>$ cp kernel/kernel-jammy-src/arch/arm64/boot/Image ../kernel/Image

Build OOT modules

<Linux_for_Tegra/source>$ export IGNORE_PREEMPT_RT_PRESENCE=1
<Linux_for_Tegra/source>$ export KERNEL_HEADERS=$PWD/kernel/kernel-jammy-src/
<Linux_for_Tegra/source>$ make modules
<Linux_for_Tegra/source>$ sudo -E make modules_install

Update initrd and build device tree binaries

<Linux_for_Tegra/source>$ cd ..
<Linux_for_Tegra>$ sudo ./tools/l4t_update_initrd.sh
<Linux_for_Tegra>$ cd source/
<Linux_for_Tegra/source>$ make dtbs
<Linux_for_Tegra/source>$ cp kernel-devicetree/generic-dts/dtbs/* ../kernel/dtb/

Now we are good to flash, we run step 9 from the README (https://github.com/Seeed-Studio/Linux_for_Tegra/blob/r36.4.0/README.md)

We were also able to get this working on Ubuntu 20 as our host PC, however we
found that post flashing jetson wouldn't come up clean, and we get a blank
screen. Using UART and serial console, we worked out that the default user was
missing. So here are the additional steps to resolve that:

<Linux_for_Tegra>$ sudo ./tools/l4t_create_default_user.sh 
Creating: Username - nvidia, Password - ..., Autologin - false
...

Now re-run the flash command.

---

N.B. We tried to follow the nvidia script and use ./nvbuild.sh -r to build real-time kernel but it didn't work for us (i.e. created an image that wouldn't boot). Perhaps there is some difference between nvbuild.sh and the official NVIDIA documentation steps.

[bigbearishappy]

Thank you for sharing this detail steps.
I believe this can help someone who is in trouble with it.