Trusted Boot Installations

“Trusted Boot” is a combination of technologies that allows us to guarantee that a system was not tampered with, and the user-data is protected by cold attacks, it is composed by FDE, Secure Boot and Measured Boot.

If you want to learn more on what Trusted Boot is and how it works, see the Trusted Boot Architecture page. This page describes how to enable Trusted Boot support in Kairos.

Kairos supports Trusted boot by generating specific installable medium. This feature is optional and works alongside how Kairos works.

Requirements

The Hardware that will run Kairos needs to have the following requirements:

  • Secure boot available in the system
  • The Hardware should have a TPM chip or fTPM enabled
  • The Hardware should be capable of booting large EFI files (>32MB)
  • Base image of the OS needs to have at least systemd 252 or newer ( for example ubuntu >=23.10 or fedora >=38 )

To build the installable medium you need the following installed in the system you use to build the installable medium:

  • Docker
  • Git
  • A Linux machine with KVM (for testing the images locally)

Usage

In order to boot into UKI mode, you need to build a special ISO file with the UKI files. To build this medium you have to generate a set of keypairs first: one for the Secure boot and one for the PCR policies required to encrypt the user-data.

Any change, or upgrade of the node to a new version of the OS requires those assets to be regenerated with these keypairs, including the installer ISO, and the EFI files used for upgrading. The keys are used to sign and verify the EFI files, and the PCR policies are used to encrypt and decrypt the user-data, and thus are required to be the same for the whole lifecycle of the node.

The steps below will guide you into generating the installable assets, and how to re-generate the assets to upgrade the node to a new version of the OS.

Key generation

Keys can be generated from scratch, or the Microsoft certificates can be used, alternatively, if you can export the keys from your BIOS/UEFI you can use the same PK keys.

By default keys are generated including the Microsoft ones, but you can skip them if you know what you are doing (see below).

To generate the Secure boot certificates and keys along with the Microsoft keys run the following commands:

MY_ORG="Acme Corp"
# Generate the keys
docker run -v $PWD/keys:/work/keys -ti --rm quay.io/kairos/osbuilder-tools:latest genkey "$MY_ORG" --expiration-in-days 365 -o /work/keys

Custom certificates by skipping Microsoft certificate keys

If your hardware supports booting with custom Secure Boot keys, you can optionally create keys from scratch and skip the Microsoft certificate keys. To do so, run the following command:

MY_ORG="Acme Corp"
# Generate the keys
docker run -v $PWD/keys:/work/keys -ti --rm quay.io/kairos/osbuilder-tools:latest genkey "$MY_ORG" --skip-microsoft-certs-I-KNOW-WHAT-IM-DOING --expiration-in-days 365 -o /work/keys

Exporting keys from BIOS/UEFI and using them

This is useful if you want to handle mass-installations where you don’t want to enroll the generated keys manually in your bios/uefi.

Some hardware might require additional vendor keys, that could be e.g. used to sign additional firmware. In order to re-use the same set of keys in the machine, you can export the keys from the BIOS/UEFI and use them to generate the UKI files.

Some BIOS/UEFI allows to export the keys to USB stick directly from the BIOS/UEFI setup menu (for example HPE). If you have the keys in a USB stick, you can copy in a directory and use them to generate the UKI files.

MY_ORG="Acme Corp"
MACHINE_CERTS="$PWD/path/to/machine-certs"
# ~$ tree $MACHINE_CERTS
# /path/to/machine-certs/
# ├── db
# ├── dbx
# ├── KEK
# └── PK

# Generate the keys
docker run -v $MACHINE_CERTS:/work/machine-keys -v $PWD/keys:/work/keys -ti --rm quay.io/kairos/osbuilder-tools:latest genkey "$MY_ORG" --custom-cert-dir /work/machine-keys --expiration-in-days 365 -o /work/keys

Building installable medium

To build the installable medium you need to run the following commands:

CONTAINER_IMAGE=quay.io/kairos/fedora:38-core-amd64-generic-v3.0.11-uki
docker run -ti --rm -v $PWD/build:/result -v $PWD/keys/:/keys quay.io/kairos/osbuilder-tools:latest build-uki $CONTAINER_IMAGE -t iso -d /result/ -k /keys
# to build an EFI file only
docker run -ti --rm -v $PWD/build:/result -v $PWD/keys/:/keys quay.io/kairos/osbuilder-tools:latest build-uki $CONTAINER_IMAGE -t uki -d /result/ -k /keys
# Assuming you have a "rootfs" directory with the content of the OS
# If the image is in a directory ($PWD/rootfs) you can use the following command
docker run -ti --rm -v $PWD/build:/result -v $PWD/rootfs:/rootfs -v $PWD/keys/:/keys quay.io/kairos/osbuilder-tools:latest build-uki dir:/rootfs/ -t iso -d /result/ -k /keys

Config files

Included in the artifact will be a configuration file. On the installation/upgrade media, it is called norole.conf but once it has been installed it will be named active.conf, passive.conf or recovery.conf which represents the configuration for the “cos”, “fallback” and “recovery” images respectively.

This configuration file includes the “title” of the artifact in the boot menu and the location of the “efi” file to boot. Here’s an example:

# cat /efi/loader/entries/active.conf
efi /EFI/kairos/active.efi
title Kairos

Branding

You can overwrite the default “Kairos” title if you pass the --boot-branding flag to enki.

CONTAINER_IMAGE=quay.io/kairos/ubuntu:24.04-core-amd64-generic-v3.0.11-uki
docker run -ti --rm -v $PWD/build:/result -v $PWD/keys/:/keys enki build-uki $CONTAINER_IMAGE -t iso -d /result/ -k /keys --boot-branding "My Awesome OS"

Your config file should look now like this:

# cat /efi/loader/entries/active.conf
efi /EFI/kairos/active.efi
title My Awesome OS

Version and cmdline in the config files

In addition to having the “title” and “efi” attributes, you can include in the config files the OS “version” and “cmdline”. To do so, pass --include-version-in-config or --include-cmdline-in-config flags to enki.

cmdline awesome=true
title Awesome OS
efi /EFI/kairos/active.efi
version v3.0.11

Installation

The installation process is performed as usual and the Installation instructions can be followed, however the difference is that user-data will be automatically encrypted (both the OEM and the persistent partition) by using the TPM chip and the Trusted Boot mechanism.

Enroll the keys in Secure Boot

If your machine is in UEFI setup mode Secure Boot keys will be automatically enrolled. To enter UEFI Setup mode you need to clear the Secure Boot keys (PKs) from the BIOS/UEFI.

If UEFI setup mode is not available, you need to enroll the keys manually in the BIOS/UEFI.

This process can vary depending on the vendor, but in general you need to enter the BIOS/UEFI setup during early boot and import the keys, for an example outline you can check the steps for HPE Hardware.

A video of the process of importing keys in QEMU is available here.

Upgrades

See the Trusted Boot Upgrade page.

Testing the images locally

To test the ISO file locally QEMU can be used. In order to test Secure Boot components you need an ed2k firmware with secureboot in QEMU. If you don’t have QEMU locally and/or you don’t have the correct dependencies you can follow the steps below that build a container image with QEMU and the needed dependencies and use that container to run the ISO file in a VM with Docker.

  1. Build the container image with the QEMU/Secure Boot dependencies (note to replace disk size/cpu/ram not needed):
docker build -t fedora-qemu -<<DOCKER
FROM fedora
RUN dnf install -y dnf-plugins-core
RUN dnf config-manager --add-repo http://www.kraxel.org/repos/firmware.repo
RUN dnf install -y edk2.git-ovmf-x64 qemu
RUN dnf install -y swtpm wget
WORKDIR /
RUN wget -q https://github.com/mudler/cos-demo-labs/raw/tests/efivars.fd -O /efivars.fd
WORKDIR /work

RUN <<EOF
echo "#!/bin/bash -ex" >> /entrypoint.sh
echo '[ ! -e /work/disk.img ] && qemu-img create -f qcow2 "/work/disk.img" 35G' >> /entrypoint.sh
echo '/usr/bin/qemu-system-x86_64 -drive if=pflash,format=raw,unit=0,file="/usr/share/edk2/ovmf/OVMF_CODE.secboot.fd",readonly=on -drive if=pflash,unit=1,format=raw,file="/efivars.fd" -accel kvm -cpu host -m 8096 -drive file=/work/disk.img,if=none,index=0,media=disk,format=qcow2,id=disk1 -device virtio-blk-pci,drive=disk1,bootindex=0 -boot order=dc -vga virtio -cpu host -smp cores=4,threads=1 -machine q35,smm=on -chardev socket,id=chrtpm,path=/tmp/swtpm-sock -tpmdev emulator,id=tpm0,chardev=chrtpm -device tpm-tis,tpmdev=tpm0 \$@' >> /entrypoint.sh
EOF
RUN chmod +x /entrypoint.sh
ENTRYPOINT [ "/entrypoint.sh" ]
DOCKER
  1. Start a TPM socket:
docker run --privileged --entrypoint swtpm -v $PWD/tpmstate:/tmp --rm -ti fedora-qemu socket --tpmstate dir=/tmp/ --ctrl type=unixio,path=/tmp/swtpm-sock --log level=20 --tpm2

Note: you need to keep the TPM container up and running for the VM to boot. Run the commands below in another terminal window.

  1. Run the container image with the ISO file (replace the iso file name with yours):
# console only
docker run --privileged -v $PWD/tpmstate:/tmp -v $PWD:/work -v /dev/kvm:/dev/kvm --rm -ti fedora-qemu -nographic -cdrom kairos-fedora-38-core-amd64-generic-v3.0.11.uki.iso

Note: To stop the QEMU container you can use Ctrl-a x or Ctrl-a c to enter the QEMU console and then quit to exit.

  1. After installation, you can run the container image by booting only with the disk
# console only
docker run --privileged -v $PWD/tpmstate:/tmp -v $PWD:/work -v /dev/kvm:/dev/kvm --rm -ti fedora-qemu -nographic

Note: you need to keep the TPM container up and running for the VM to boot.

Data Encryption

The user-data will be automatically encrypted during installation, along with the OEM and the persistent partition by using the TPM chip and the Trusted Boot mechanism.

Additional partitions

Additional partitions can be encrypted and specified as part of the cloud-config used during the installation process, for example:

install:
  extra-partitions:
    - name: second_partition
      size: 100
      fs: ext2
      label: PARTITION_TWO
  encrypted_partitions:
  - PARTITION_TWO

A full example can be:

install:
  device: "auto" # Install to the biggest drive
  auto: true    # Enables auto installation
  partitions:
    persistent:
      size: 500 # Set persistent partition to 500MB (otherwise takes the whole disk)
  extra-partitions:
    - name: second_partition
      size: 100
      fs: ext2
      label: PARTITION_TWO
  encrypted_partitions:
  - PARTITION_TWO

Notes

Install additional files from the Installer ISO to the encrypted portions

The ISO installer images can be used to install additional content in the encrypted portion of the disk.

The osbuilder image can overlay additional files into the iso. For example specify --overlay-iso /additional/path to have added the files in the folder inside the ISO. The content can be accessed during installation in /run/initramfs/live. For example, to install content of the ISO inside the OEM partition, and to restore those after a reset we can use the following cloud config:

#cloud-config

install:
  device: "/dev/vda"
  auto: true
  partitions:
    oem:
      size: 4000
      fs: ext4

users:
- name: "kairos"
  passwd: "kairos"

stages:
  after-install:
      - commands:
        - echo "Copying files to oem and persistent"
        - /usr/lib/systemd/systemd-cryptsetup attach persistent $(findfs PARTLABEL=persistent) - tpm2-device=auto
        - /usr/lib/systemd/systemd-cryptsetup attach oem $(findfs PARTLABEL=oem) - tpm2-device=auto
        - mount /dev/mapper/persistent /usr/local
        - mount /dev/mapper/oem /oem
        - mkdir -p /usr/local/.state/opt.bind
        - mkdir -p /oem/opt.bind
        - cp -rfv /run/initramfs/live/data/* /usr/local/.state/opt.bind
        - cp -rfv /run/initramfs/live/data/* /oem/opt.bind
        - umount /dev/mapper/persistent
        - umount /dev/mapper/oem
        - cryptsetup close /dev/mapper/persistent
        - cryptsetup close /dev/mapper/oem

  after-reset:
      - commands:
        - |
          /bin/bash <<'EOF'
          #!/bin/bash

          set -e
          echo "Copying files from oem to persistent"
          # /oem was mounted in my tests. Let's umount it to be sure.
          umount /oem || true
          # Close all encrypted partitions
          for p in $(ls /dev/mapper/vda*); do cryptsetup close $p; done
          /usr/lib/systemd/systemd-cryptsetup attach persistent $(findfs PARTLABEL=persistent) - tpm2-device=auto
          /usr/lib/systemd/systemd-cryptsetup attach oem $(findfs PARTLABEL=oem) - tpm2-device=auto
          mount /dev/mapper/persistent /usr/local
          mount /dev/mapper/oem /oem
          mkdir -p /usr/local/.state/opt.bind
          cp -rfv /oem/opt.bind/* /usr/local/.state/opt.bind
          umount /dev/mapper/persistent
          umount /dev/mapper/oem
          cryptsetup close /dev/mapper/persistent
          cryptsetup close /dev/mapper/oem

          EOF          

Mount partitions after install

/oem and /usr/local can be mounted after installation to prepare content before first-boot.

# Note: replace /dev/vda2 with the oem partition location (see with `blkid`)
# [root@ ~]# blkid 
# /dev/sr0: BLOCK_SIZE="2048" UUID="2024-02-05-17-00-05-00" LABEL="UKI_ISO_INSTALL" TYPE="iso9660"
# /dev/vda2: UUID="8bfa06f9-ca4f-56dc-90c9-49cf20f4f45e" TYPE="crypto_LUKS" PARTLABEL="oem" PARTUUID="63deb673-ec99-46f6-9cb6-8399315e4f19"
# /dev/vda3: UUID="85c39d0f-4867-5227-8334-f5eec606d9eb" TYPE="crypto_LUKS" PARTLABEL="persistent" PARTUUID="d01d9b51-d61a-4b7e-bb1a-8af5c212a213"
# /dev/vda1: LABEL_FATBOOT="COS_GRUB" LABEL="COS_GRUB" UUID="1C4C-97AA" BLOCK_SIZE="512" TYPE="vfat" PARTLABEL="efi" PARTUUID="6e42d80e-d67a-462b-b99c-2c1b5dda91cf"
# /dev/mapper/oem: LABEL="COS_OEM" UUID="d10fc63d-9387-442c-9db8-a00e081858ec" BLOCK_SIZE="1024" TYPE="ext4"

# Mount OEM
/usr/lib/systemd/systemd-cryptsetup attach oem $(findfs PARTLABEL=oem) - tpm2-device=auto
mount /dev/mapper/oem /oem

# Mount persistent
/usr/lib/systemd/systemd-cryptsetup attach persistent $(findfs PARTLABEL=persistent) - tpm2-device=auto
mount /dev/mapper/persistent /usr/local

To mount /oem and /usr/local after install you can also manually call kcrypt unlock-all. However this isn’t supported yet.

Force certificates auto-enrollments

If you want to force the auto-enrollment of the certificates in the BIOS/UEFI, you can use the --secure-boot-enroll flag in the build-uki command.

CONTAINER_IMAGE=quay.io/kairos/fedora:38-core-amd64-generic-v3.0.11-uki
docker run -ti --rm -v $PWD/build:/result -v $PWD/keys/:/keys quay.io/kairos/osbuilder-tools:latest build-uki $CONTAINER_IMAGE --secure-boot-enroll force -t iso -d /result/ -k /keys

Additional efi entries

When booting in UKI mode, the cmdline is part of the signed artifact. This means that in order to pass additional options through the cmdline, this has to be done at build time. For that reason, enki build-uki supports the following arguments:

  • --extend-cmdline: Use this option to add cmdline parameters to the default entries (active/passive/recovery/auto-reset). This option will not generate additional entries in the systemd-boot menu.
  • --extra-cmdline: Use this option to generate additional entries in the systemd-boot menu. For each one of the default entries, a new entry will be created with the specified cmdline (added to the default cmdline).
  • --single-efi-cmdline: Use this option to generate entries that are not related to default ones. This option will generate a single entry with the provided cmdline (added to the default cmdline). This argument can be used multiple times.

Examples:

Building with a command like this:

enki build-uki ...more options... --single-efi-cmdline "Lets debug: rd.debug rd.immucore.debug"

Will create a boot menu with the default entries, plus one more: “Lets debug” which will have the debug parameters set. Notice how the title of the entry and the cmdline to be used are split by a colon.

Building with a command like this:

enki build-uki ...more options... --extra-cmdline "rd.debug rd.immucore.debug"

Will create a boot menu with the default entries, plus one more for each of the active/passive/recovery entries. Each of these entries will have the debug parameters set.

Building with a command like this:

enki build-uki ...more options... --extend-cmdline "rd.debug rd.immucore.debug"

Will create a boot menu with just the default entries but this time they will have the debug parameters set.

NOTE: --boot-branding is applied to --single-efi-cmdline too. For example, this command:

enki build-uki ...more options... --boot-branding "My awesome OS" --single-efi-cmdline "Debug logs: rd.debug rd.immucore.debug"

will create this entry:

My awesome OS (Debug logs)

with the additional cmdline parameters.