BI Container Runtime

Build

make build

Run

!Note: Make sure you have created rootfs first and ROOTFS env variable is set (see below).

Without any parameters bicr-run starts a shell inside the container

./build/bicr-run

you can run specific commands from the container:

./build/bicr-run /bin/env

You can test with the example web server

cp ./build/bicr-httpd-example ./build/rootfs/opt
PORT=9999 ./build/bicr-run /opt/bicr-httpd-example

Creating root filesystem

export ROOTFS="$(pwd)/build/rootfs"
mkdir -pv $ROOTFS

Depending on the flavour of Linux you like, you can use either minimalistic BusyBox environment or Debian, Arch, Fedora.

Here are some examples:

busybox

From source code

This is the default one.

You need to have installed gcc, g++, make and what Debian calls build-essentials

export MAKEFLAGS=-j`nproc` && make rootfs

when the config menu appears go to Settings -> Build Options and select Build static binary, then exit and save configuration.

Get busybox out of docker image

docker create busybox
CID=$(docker container ls -a --format json | grep busybox | jq -r '.ID')
docker export $CID | tar -xf - -C $ROOTFS

build alpine

pushd /tmp
    wget http://dl-cdn.alpinelinux.org/alpine/v3.19/main/x86_64/apk-tools-static-2.14.0-r5.apk
    mkdir -pv /tmp/apk-tools
    tar -xf apk-tools-static-2.14.0-r5.apk -C /tmp/apk-tools
popd

sudo /tmp/apk-tools/sbin/apk.static \
    -X http://dl-cdn.alpinelinux.org/alpine/v3.19/main -U \
    --arch x86_64 \
    --allow-untrusted --root $ROOTFS \
    --initdb add alpine-base gcompat

debootstrap

debootstrap bookworm $ROOTFS http://deb.debian.org/debian/

pacstrap

pacstrap -K $ROOTFS base vim

DNF on Fedora, RHEL, etc

sudo dnf -y --releasever=39 --installroot=$ROOTFS \
      --repo=fedora --repo=updates --setopt=install_weak_deps=False install \
      passwd dnf fedora-release vim-minimal 

Chroot in rootfs

If you want to start a shell in the new root filesystem you can:

sudo chroot $ROOTFS /bin/sh

Credits

this project started as a fork of teddyking/ns-process

Extending and reusing rootfs with Overlay FS

After you build $ROOTFS:

export ROOTFS_BASE="$ROOTFS-base"
export CONTAINERFS="$(pwd)/build/rootfs-container"

mv $ROOTFS $ROOTFS_BASE
mkdir -pv $ROOTFS $CONTAINERFS
mount -t tmpfs tmpfs $CONTAINERFS
mkdir -pv $CONTAINERFS/{up,work}

# note that we are copying the binary inside the 'up' directory
cp ./build/bicr-httpd-example $CONTAINERFS/up/bin/httpd

mount -t overlay overlay -o lowerdir=$ROOTFS_BASE,upperdir=$CONTAINERFS/up/,workdir=$CONTAINERFS/work/ $ROOTFS

# run container
./build/bicr-run httpd

Overlay FS applies everything inside $CONTAINERFS on top of $ROOTFS_BASE and makes accessible via $ROOTFS so you can chroot in it as per usual. Needless to say you can have many $CONTAINERFS on top of a single base rootfs.

If you change the base root filesystem or the container FS, you would need to remount so changes can be visible in $ROOTFS.