docs/BUILDING_IMAGES.md

   1 ---
   2 title: Safely Building Images
   3 category: Concepts
   4 layout: default
   5 SPDX-License-Identifier: LGPL-2.1-or-later
   6 ---
   7
   8 # Safely Building Images
   9
  10 In many scenarios OS installations are shipped as pre-built images, that
  11 require no further installation process beyond simple `dd`-ing the image to
  12 disk and booting it up. When building such "golden" OS images for
  13 `systemd`-based OSes a few points should be taken into account.
  14
  15 Most of the points described here are implemented by the
  16 [`mkosi`](https://github.com/systemd/mkosi) OS image builder developed and
  17 maintained by the systemd project. If you are using or working on another image
  18 builder it's recommended to keep the following concepts and recommendations in
  19 mind.
  20
  21 ## Resources to Reset
  22
  23 Typically the same OS image shall be deployable in multiple instances, and each
  24 instance should automatically acquire its own identifying credentials on first
  25 boot. For that it's essential to:
  26
  27 1. Remove the
  28    [`/etc/machine-id`](https://www.freedesktop.org/software/systemd/man/machine-id.html)
  29    file or write the string `uninitialized\n` into it. This file is supposed to
  30    carry a 128bit identifier unique to the system. Only when it is reset it
  31    will be auto-generated on first boot and thus be truly unique. If this file
  32    is not reset, and carries a valid ID every instance of the system will come
  33    up with the same ID and that will likely lead to problems sooner or later,
  34    as many network-visible identifiers are commonly derived from the machine
  35    ID, for example IPv6 addresses or transient MAC addresses.
  36
  37 2. Remove the `/var/lib/systemd/random-seed` file (see
  38    [`systemd-random-seed(8)`](https://www.freedesktop.org/software/systemd/man/systemd-random-seed.service.html),
  39    which is used to seed the kernel's random pool on boot. If this file is
  40    shipped pre-initialized, every instance will seed its random pool with the
  41    same random data that is included in the image, and thus possibly generate
  42    random data that is more similar to other instances booted off the same image
  43    than advisable.
  44
  45 3. Remove the `/loader/random-seed` file (see
  46    [`systemd-boot(7)`](https://www.freedesktop.org/software/systemd/man/systemd-boot.html)
  47    from the UEFI System Partition (ESP), in case the `systemd-boot` boot loader
  48    is used in the image.
  49
  50 4. It might also make sense to remove `/etc/hostname` and `/etc/machine-info`
  51    which carry additional identifying information about the OS image.
  52
  53 ## Boot Menu Entry Identifiers
  54
  55 The `kernel-install` logic used to generate [Boot Loader Specification Type
  56 1](https://systemd.io/BOOT_LOADER_SPECIFICATION) entries by default uses the
  57 machine ID as stored in `/etc/machine-id` for naming boot menu entries and the
  58 directories in the ESP to place kernel images in. This is done in order to
  59 allow multiple installations of the same OS on the same system without
  60 conflicts. However, this is problematic if the machine ID shall be generated
  61 automatically on first boot: if the ID is not known before the first boot it
  62 cannot be used to name the most basic resources required for the boot process
  63 to complete.
  64
  65 Thus, for images that shall acquire their identity on first boot only, it is
  66 required to use a different identifier for naming boot menu entries. To allow
  67 this the `kernel-install` logic knows the generalized *entry* *token* concept,
  68 which can be a freely chosen string to use for identifying the boot menu
  69 resources of the OS. If not configured explicitly it defaults to the machine
  70 ID. The file `/etc/kernel/entry-token` may be used to configure this string
  71 explicitly. Thus, golden image builders should write a suitable identifier into
  72 this file, for example the `IMAGE_ID=` or `ID=` field from
  73 `/etc/os-release`. It is recommended to do this before the `kernel-install`
  74 functionality is invoked (i.e. before the package manager is used to install
  75 packages into the OS tree being prepared), so that the selected string is
  76 automatically used for all entries to be generated.
  77
  78 ## Booting with Empty `/var/` and/or Empty Root File System
  79
  80 `systemd` is designed to be able to come up safely and robustly if the `/var/`
  81 file system or even the entire root file system (with exception of `/usr/`,
  82 i.e. the vendor OS resources) is empty (i.e. "unpopulated"). With this in mind
  83 it's relatively easy to build images that only ship a `/usr/` tree, and
  84 otherwise carry no other data, populating the rest of the directory hierarchy
  85 on first boot as needed.
  86
  87 Specifically, the following mechanisms are in place:
  88
  89 1. The `swich-root` logic in systemd, that is used to switch from the initrd
  90    phase to the host will create the basic OS hierarchy skeleton if missing. It
  91    will create a couple of directories strictly necessary to boot up
  92    successfully, plus essential symlinks (such as those necessary for the
  93    dynamic loader `ld.so` to function).
  94
  95 2. PID 1 will initialize `/etc/machine-id` automatically if not initialized yet
  96    (see above).
  97
  98 3. The `nss-systemd` glibc NSS module ensures the `root` and `nobody` users and
  99    groups remain resolvable, even without `/etc/passwd` and `/etc/group` around.
 100
 101 4. The
 102    [`systemd-sysusers`](https://www.freedesktop.org/software/systemd/man/systemd-sysusers.service.html)
 103    will component automatically populate `/etc/passwd` and `/etc/group` on
 104    first boot with further necessary system users.
 105
 106 5. The
 107    [`systemd-tmpfiles`](https://www.freedesktop.org/software/systemd/man/systemd-tmpfiles-setup.service.html)
 108    component ensures that various files and directories below `/etc/`, `/var/`
 109    and other places are created automatically at boot if missing. Unlike the
 110    directories/symlinks created by the `switch-root` logic above this logic is
 111    extensible by packages, and can adjust access modes, file ownership and
 112    more. Among others this will also link `/etc/os-release` →
 113    `/usr/lib/os-release`, ensuring that the OS release information is
 114    unconditionally accessible through `/etc/os-release`.
 115
 116 6. The `nss-myhostname` glibc NSS module will ensure the local host name as
 117    well as `localhost` remains resolvable, even without `/etc/hosts` around.
 118
 119 With these mechanisms the hierarchies below `/var/` and `/etc/` can be safely
 120 and robustly populated on first boot, so that the OS can safely boot up. Note
 121 that some auxiliary package are not prepared to operate correctly if their
 122 configuration data in `/etc/` or their state directories in `/var/` are
 123 missing. This can typically be addressed via `systemd-tmpfiles` lines that
 124 ensure the missing files and directories are created if missing. In particular,
 125 configuration files that are necessary for operation can be automatically
 126 copied or symlinked from the `/usr/share/factory/etc/` tree via the `C` or `L`
 127 line types. That said, we recommend that all packages safely fall back to
 128 internal defaults if their configuration is missing, making such additional
 129 steps unnecessary.
 130
 131 Note that while `systemd` itself explicitly supports booting up with entirely
 132 unpopulated images (`/usr/` being the only required directory to be populated)
 133 distributions might not be there yet: depending on your distribution further,
 134 manual work might be required to make this scenario work.
 135
 136 ## Adapting OS Images to Storage
 137
 138 Typically, if an image is `dd`-ed onto a target disk it will be minimal:
 139 i.e. only consist of necessary vendor data, and lack "payload" data, that shall
 140 be individual to the system, and dependent on host parameters. On first boot,
 141 the OS should take possession of the backing storage as necessary, dynamically
 142 using available space. Specifically:
 143
 144 1. Additional partitions should be created, that make no sense to ship
 145    pre-built in the image. For example `/tmp/` or `/home/` partitions, or even
 146    `/var/` or the root file system (see above).
 147
 148 2. Additional partitions should be created that shall function as A/B
 149    secondaries for partitions shipped in the original image. In other words: if
 150    the `/usr/` file system shall be updated in an A/B fashion it typically
 151    makes sense to ship the original A file system in the deployed image, but
 152    create the B partition on first boot.
 153
 154 3. Partitions covering only a part of the disk should be grown to the full
 155    extent of the disk.
 156
 157 4. File systems in uninitialized partitions should be formatted with a file
 158    system of choice.
 159
 160 5. File systems covering only a part of a partition should be grown to the full
 161    extent of the partition.
 162
 163 6. Partitions should be encrypted with cryptographic keys generated locally on
 164    the machine the system is first booted on, ensuring these keys remain local
 165    and are not shared with any other instance of the OS image.
 166
 167 Or any combination of the above: i.e. first create a partition, then encrypt
 168 it, then format it.
 169
 170 `systemd` provides multiple tools to implement the above logic:
 171
 172 1. The
 173    [`systemd-repart`](https://www.freedesktop.org/software/systemd/man/systemd-repart.service.html)
 174    component may manipulate GPT partition tables automatically on boot, growing
 175    partitions or adding in partitions taking the backing storage size into
 176    account. It can also encrypt partitions automatically it creates (even bind
 177    to TPM2, automatically) and populate partitions from various sources. It
 178    does this all in a robust fashion so that aborted invocations will not leave
 179    incompletely set up partitions around.
 180
 181 2. The
 182    [`systemd-makefs@(8).service`](https://www.freedesktop.org/software/systemd/man/systemd-growfs.html)
 183    tool can automatically grow a file system to the partition it is contained
 184    in. The `x-systemd.growfs` `/etc/fstab` mount option is sufficient to enable
 185    this logic for specific mounts. If the file system is already grown it
 186    executes no operation.
 187
 188 3. Similar, the `systemd-makefs@.service` and `systemd-makeswap@.service`
 189    services can format file systems and swap spaces before first use, if they
 190    carry no file system signature yet. The `x-systemd.makefs` mount option in
 191    `/etc/fstab` may be used to request this functionality.
 192
 193 ## Provisioning Image Settings
 194
 195 While a lot of work has gone into ensuring `systemd` systems can safely boot
 196 with unpopulated `/etc/` trees, it sometimes is desirable to set a couple of
 197 basic settings *after* `dd`-ing the image to disk, but *before* first boot. For
 198 this the tool
 199 [`systemd-firstboot`](https://www.freedesktop.org/software/systemd/man/systemd-firstboot.html)
 200 can be useful, with its `--image=` switch. It may be used to set very basic
 201 settings, such as the root password or hostname on an OS disk image or
 202 installed block device.
 203
 204 ## Distinguishing First Boot
 205
 206 For various purposes it's useful to be able to distinguish the first boot-up of
 207 the system from later boot-ups (for example, to set up TPM hardware
 208 specifically, or register a system somewhere). `systemd` provides mechanisms to
 209 implement that. Specifically, the `ConditionFirstBoot=` and `AssertFirstBoot=`
 210 settings may be used to conditionalize units to only run on first boot. See
 211 [`systemd.unit(5)`](https://www.freedesktop.org/software/systemd/man/systemd.unit.html#ConditionFirstBoot=)
 212 for details.
 213
 214 A special target unit `first-boot-complete.target` may be used as milestone to
 215 safely handle first boots where the system is powered off too early: if the
 216 first boot process is aborted before this target is reached, the following boot
 217 process will be considered a first boot, too. Once the target is reached,
 218 subsequent boots will not be considered first boots anymore, even if the boot
 219 process is aborted immediately after. Thus, services that must complete fully
 220 before a system shall be considered fully past the first boot should be ordered
 221 before this target unit.
 222
 223 Whether a system will come up in first boot state or not is derived from the
 224 initialization status of `/etc/machine-id`: if the file already carries a valid
 225 ID the system is already past the first boot. If it is not initialized yet it
 226 is still considered in the first boot state. For details see
 227 [`machine-id(5)`](https://www.freedesktop.org/software/systemd/man/machine-id.html).