How to Build ARM Docker Images on Intel host

Hiya everyone!

Ever wondered how to build ARM Docker Images directly on you PC? I had to struggle finding out how to do so, in order to build up the ARM version for my Docker-based M.Sc. thesis work architecture, NTBD (trilogy post ‘bout that here).

The following is a quick guide to how to setup your Intel host to build Docker Images that can be run on an ARM-processor-based host, exploiting QEMU.

Remark: in this tutorial I will assume that the reader has knowledge about Docker. For a quick overview about Docker see my post, “Docker, this stranger!”. This guide will give advices for building a Docker Image for a Raspberry Pi board on a Ubuntu host.

Index:

• 1. What is QEMU and QEMU installation
• 2. QEMU and Docker Images
• 3. Register QEMU in the build agent
• 4. Build the Image

1. What is QEMU and QEMU installation

QEMU (Quick EMUlator) is an Open-Source hosted hypervisor, i.e. an hypervisor running on a OS just as other computer programs, which performs hardware virtualization. QEMU emulates CPUs of several architectures, e.g. x86, PPC, ARM and SPARC. It allows the execution of non-native target executables emulating the native execution and, as we require in this case, the cross-building process.

Since I am using an Ubuntu host these are the commands to install qemu, qemu-user-static and binfmt-support packages from the command line:

 sudo apt update
 sudo apt install -y qemu qemu-user-static qemu-user binfmt-support

Note that package qemu-user-static provides statically built emulation QEMU binaries, which do not have dependencies.

2. QEMU and Docker Images

So as to obtain a Docker Image which can be correctly built and run on an ARM host, you need to have a base Image with the target qemu statically linked executable, qemu-arm-static in this case. There are plenty ready Images complete of the qemu static binary:

• Hypriot rpi-alpine Image
• Resin rpi-raspbian Image
• Resin raspberry-pi-alpine-node:slim Image.

However you can also use “simple” Raspberry Images and then copy into the container the binary you can find at /usr/bin/ after having downloaded the qemu packages (see step 1).

Suppose you are starting from the “official” Ubuntu Docker images for the ARMv7(armhf), available here, and that you have made a copy of the required binary, qemu-arm-static, in your build context (folder containing Dockerfile, i.e. “.”). The first lines of your Dockerfile file will be:

FROM armv7/armhf-ubuntu:16.04

COPY ./qemu-arm-static /usr/bin/qemu-arm-static

In my case, for NTBD Base Image, I have used as base Image the HotBlack Robotics hbrobotics/ros-base:rpi3 image, which is based on the “Official” ARM Image I mentioned before, with ROS Kinetic installed on it. I then copied on it the QEMU ARM static binary, getting:

FROM  hbrobotics/ros-base:rpi3

COPY ./qemu-arm-static /usr/bin/qemu-arm-static

Then complete your image Dockerfile with all the needed layers.

3. Register QEMU in the build agent

To add QEMU in the build agent there is a specific Docker Image performing what we need, so just run in your command line:

docker run --rm --privileged multiarch/qemu-user-static:register --reset

4. Build the Image

Now you are ready to build your image. Just build with the usual docker build command. In my case, having a specific Dockerfile name, the command is:

docker build -f ./Dockerfile.rpi3 -t ntbd/base:rpi3 .

That’s it! Your Docker Image should be ready to be pushed on your Docker Hub repository and eventually downloaded and deployed on your Raspberry Pi board.

Thanks, for most the informations, to Hypriot.
See ya!