Raspberry PI Alpine Linux Image Builder

Create minimal Linux images based on Alpine Linux for the Raspberry PI.

Features

• Alpine base
  • Small image size
  • Fast image build
• A/B partition schema
  • Simple update of whole system
  • Fallback if update failed
• Single image for every Raspberry PI
• Read only root filesystem

Usage

Image Creation

A simple example for a go application can be found in the example directory.

To generate an empty image simply run:

docker run --rm -it -v $PWD/output:/output bboehmke/raspi-alpine-builder

This will create 2 image files in the directory $PWD/output/:

• alpine.img.gz: complete SD card image for the raspberry
• alpine_update.img.gz: image of root partition to update running raspberry

For each image a *.sha256 file will be generated to validate integrity.

To add custom modifications mount a script to /input/image.sh. The following variables can be useful for the for and image.sh:

Variable	Description
INPUT_PATH	Path to input directory
ROOTFS_PATH	Path to new root filesystem
BOOTFS_PATH	Path to new boot filesystem
DATAFS_PATH	Path to new data filesystem

There is also a function chroot_exec that can be used to run command inside the new root filesystem. To enable a service called example_daemon simple run:

chroot_exec rc-update add example_daemon default

Config Variables

The following variables can be used to modify the base behaviour of the image builder.

Variable	Default Value	Description
ALPINE_BRANCH	3.10	Alpine Branch to use for image
ALPINE_MIRROR	http://dl-cdn.alpinelinux.org/alpine	Mirror used for package download
CUSTOM_IMAGE_SCRIPT	image.sh	Name of script for image customizations (relative to input dir)
DEFAULT_HOSTNAME	alpine	Default hostname
DEFAULT_ROOT_PASSWORD	alpine	Default password for root user
DEFAULT_TIMEZONE	Etc/UTC	Default Timezone to use for image
IMG_NAME	sdcard	Base name of created image file
INPUT_PATH	/input	Input directory inside container
OUTPUT_PATH	/output	Output directory inside container
SIZE_BOOT	100M	Size of boot partition
SIZE_DATA	20M	Initial Size of data partition
SIZE_ROOT_FS	200M	Size of root file system
SIZE_ROOT_PART	500M	Size of root partition

Update running system

The system can be updated without a complete flash of the SD card from the running system with the following steps:

1. Transfer the update image to the running system
2. (Optional) Validate integrity of image with checksum file
3. Write the image to the actual inactive partition
4. Switch active partition
5. Reboot system

An example implementation can be found in the helper script ab_flash

Image structure

Partition Layout

The image contains 4 partitions:

1. Boot: (Size: SIZE_BOOT)
   Contains boot loader and boot configuration

2. Root A: (Size: SIZE_ROOT_PART)
   Contains complete root filesystem including all required kernels

3. Root B: (Size: SIZE_ROOT_PART)
   Same as Root A

4. Data: (Size: Initial SIZE_DATA & increases on first start)
   Contains persistent data for both root partitions

With the exception of the data partition every partition is mounted read only

The A/B root partitions enables an easy and reliable way to update the complete system. This is done by flashing the inactive partition with a new root filesystem and make this partition active. If this new filesystem does not boot the boot loader will fallback to the old partition.

The root file system should be as small as possible to reduce the update time later. To support future increase of the root file system the partition should contain some free space.

Boot loader

To support the A/B partition schema the U-Boot boot loader is used.

The configuration/script for the boot loader can be found in the boot.cmd. This will boot the active root partition and switch the active partition if the active one will not start.

This script also select the right kernel for old Raspberry PIs

The image contains a simple tool that resets the boot counter and switch the active partition from the running OS. The sources of the script can be found in the uboot.c.