Awesome
Build RT_PREEMPT
kernel for Raspberry Pi
Introduction
This README describes necessary steps to build and install RT_PREEMPT
Linux kernel for the Raspberry Pi board. RT Kernel is a part of the ROS2 real-time system setup. Raspberry Pi is a reference board used by the ROS 2 real-time community for the development. RT Kernel is configured as described in Kernel configuration section. Kernel is built automatically by the Github action, and the artifacts are located under the build stable
. Please follow installation instructions to deploy a new kernel to the RPI board.
Raspberry Pi RT Linux kernel
Ubuntu raspi
kernel is modified to produce an RT Linux kernel. Ubuntu is a ROS 2 Tier 1 platform and Ubuntu kernel was selected to align to it. It is possible to build the raspi kernel for Ubuntu LTS release 24.04.
Download ready-to-use RT Kernel deb
packages
RT Kernel is configured using configuration parameters from the file. In the case you need to build your own kernel read the description below.
Using GUI
Go to the Action
tab, find the Build stable
, go inside the latest workflow run, download, and unzip artifacts called RPI4 RT Kernel deb packages
. This archive contains three debian packages. Follow instructions to deploy them on the RPI.
Using command line
Go to the Developer settings
and generate a token to access the repo via Github API. Use this token in conjunction with your Github name to retrieve build artifacts.
$ token=<my_token>
# rertieve all artifacts
$ curl -i -u <my github name>:$token -H "Accept: application/vnd.github.v3+json" https://api.github.com/repos/ros-realtime/linux-real-time-kernel-builder/actions/artifacts | grep archive_download_url
"archive_download_url": "https://api.github.com/repos/ros-realtime/linux-real-time-kernel-builder/actions/artifacts/91829081/zip",
"archive_download_url": "https://api.github.com/repos/ros-realtime/linux-real-time-kernel-builder/actions/artifacts/91534731/zip",
# download the latest one
$ curl -u <my github name>:$token -L -H "Accept: application/vnd.github.v3+json" https://api.github.com/repos/ros-realtime/linux-real-time-kernel-builder/actions/artifacts/91829081/zip --output rpi4_rt_kernel.zip
$ unzip rpi4_rt_kernel.zip
Raspberry Pi RT Linux kernel build
Ubuntu 24.04 x86_64
based Dockerfile
is developed to cross-compile a new kernel.
Build environment
Docker container comes with cross-compilation tools installed, and a ready-to-build RT Linux kernel:
- ARMv8 cross-compilation tools
- Linux source build dependencies
- Linux source
buildinfo
, from where kernel config is copied - Ubuntu
raspi
Linux source installed under~/linux_build
- RT kernel patch downloaded and applied - the nearest to the recent
raspi
Ubuntu kernel
It finds the latest raspi
linux-image
and the closest to it RT patch. If the build arguments specified it will build a corresponding kernel version instead.
Build and run docker container
For the local build:
git clone https://github.com/ros-realtime/linux-real-time-kernel-builder
cd linux-real-time-kernel-builder
docker build [--no-cache] [--build-arg UBUNTU_VERSION=<ubuntu name>] [--build-arg KERNEL_VERSION=<kernel version>] [--build-arg UNAME_R=<raspi release>] [--build-arg RT_PATCH=<RT patch>] [--build-arg LTTNG_VERSION=<LTTNG version>] -t rtwg-image .
where:
<ubuntu name>
isnoble
, default isnoble
<kernel version>
is6.8.0
, default is6.8.0
<raspi release>
is in a form of6.8.0-1005-raspi
, see Ubuntu raspi Linux kernels<RT patch>
is in a form of6.8.2-rt11
, see RT patches<LTTNG version>
is2.13
, default is2.13
docker run -t -i rtwg-image bash
Kernel configuration
There is a separate kernel configuration fragment.config-fragment
introduced to apply ROS2 real-time specific kernel settings. Below is an example:
$ cat .config-fragment
CONFIG_PREEMPT_RT=y
CONFIG_NO_HZ_FULL=y
CONFIG_HZ_1000=y
# CONFIG_AUFS_FS is not set
If you need to reconfigure it, run
make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- menuconfig
Alternatively, you can modify .config-fragment
and then merge your changes in the .config
by running
cd $HOME/linux_build/linux-raspi
ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- ./scripts/kconfig/merge_config.sh .config $HOME/linux_build/.config-fragment
Kernel build
cd $HOME/linux_build/linux-raspi
make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- LOCALVERSION=-raspi -j `nproc` bindeb-pkg
You need 16GB free disk space to build it, it takes a while, and the results are located:
raspi:~/linux_build/linux-raspi $ ls -la ../*.deb
-rw-r--r-- 1 user user 9355162 Jul 1 16:44 ../linux-headers-6.8.4-rt11-raspi_6.8.4-g75867ff0890f-4_arm64.deb
-rw-r--r-- 1 user user 70457678 Jul 1 16:44 ../linux-image-6.8.4-rt11-raspi_6.8.4-g75867ff0890f-4_arm64.deb
-rw-r--r-- 1 user user 1377154 Jul 1 16:44 ../linux-libc-dev_6.8.4-g75867ff0890f-4_arm64.deb
Deploy new kernel on Raspberry Pi
Download and install Ubuntu 24.04 server image
Follow these links to download and install Ubuntu 24.04 on your Raspberry Pi
- Install Ubuntu on a Raspberry Pi
- Download Ubuntu Raspberry Pi server image
- Create an Ubuntu image for a Raspberry Pi on Ubuntu
# initial username and password
ubuntu/ubuntu
Update your system
After that you need to connect to the Internet and update your system
$ sudo apt-get update && apt-get upgrade
Install Ubuntu Desktop (optional)
Optionally you can install a desktop version
$ sudo apt-get update && apt-get upgrade && apt-get install ubuntu-desktop
Copy a new kernel to your system and install it
Assumed you have already copied all *.deb
kernel packages to your $HOME
directory
cd $HOME
sudo dpkg -i linux-image-*.deb
sudo reboot
After reboot you should see a new RT kernel installed and real-time enabled
ubuntu@ubuntu:~$ uname -a
Linux ubuntu 6.8.4-rt11-raspi #1 SMP PREEMPT_RT Mon Jul 1 14:10:16 UTC 2024 aarch64 aarch64 aarch64 GNU/Linux
ubuntu@ubuntu:~$ cat /sys/kernel/realtime
1
Intel UP2 board RT kernel build
To build x86_64
Linux kernel, see Building Realtime rt_preempt kernel for ROS 2
Why is LTTng included in the kernel?
LTTng stands for Linux Trace Toolkit: next generation and is an open source toolkit that enables low-level kernel tracing which can be extremely useful when calculating callback times, memory usage and many other key characteristics.
As this repository is within the ros-realtime
organization it can be assumed that most users will install ROS 2 on the end system - which then they can use ros2_tracing
to trace various things. Since ros2_tracing
uses LTTng as its tracer, and since the lttng-modules
package is not easily available for the raspberry-pi RT linux kernel we build it into the kernel here as a work around.