systemd/man/machine-id.xml

204 lines
11 KiB
XML
Raw Normal View History

<?xml version='1.0'?> <!--*-nxml-*-->
<!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.5//EN"
"http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd">
<!-- SPDX-License-Identifier: LGPL-2.1-or-later -->
<refentry id="machine-id">
2015-02-04 10:14:13 +08:00
<refentryinfo>
<title>machine-id</title>
<productname>systemd</productname>
</refentryinfo>
<refmeta>
<refentrytitle>machine-id</refentrytitle>
<manvolnum>5</manvolnum>
</refmeta>
<refnamediv>
<refname>machine-id</refname>
<refpurpose>Local machine ID configuration file</refpurpose>
</refnamediv>
<refsynopsisdiv>
<para><filename>/etc/machine-id</filename></para>
</refsynopsisdiv>
<refsect1>
<title>Description</title>
<para>The <filename>/etc/machine-id</filename> file contains the unique machine ID of
the local system that is set during installation or boot. The machine ID is a single
newline-terminated, hexadecimal, 32-character, lowercase ID. When decoded from
hexadecimal, this corresponds to a 16-byte/128-bit value. This ID may not be all
zeros.</para>
<para>The machine ID is usually generated from a random source during system
installation or first boot and stays constant for all subsequent boots. Optionally,
for stateless systems, it is generated during runtime during early boot if necessary.
</para>
<para>The machine ID may be set, for example when network booting, with the
<varname>systemd.machine_id=</varname> kernel command line parameter or by passing the
option <option>--machine-id=</option> to systemd. An ID specified in this manner
has higher priority and will be used instead of the ID stored in
<filename>/etc/machine-id</filename>.</para>
<para>The machine ID does not change based on local or network configuration or when
hardware is replaced. Due to this and its greater length, it is a more useful
replacement for the
<citerefentry project='man-pages'><refentrytitle>gethostid</refentrytitle><manvolnum>3</manvolnum></citerefentry>
call that POSIX specifies.</para>
2015-02-04 10:14:13 +08:00
<para>This machine ID adheres to the same format and logic as the
D-Bus machine ID.</para>
<para>This ID uniquely identifies the host. It should be considered "confidential", and must not be exposed in
untrusted environments, in particular on the network. If a stable unique identifier that is tied to the machine is
needed for some application, the machine ID or any part of it must not be used directly. Instead the machine ID
should be hashed with a cryptographic, keyed hash function, using a fixed, application-specific key. That way the
ID will be properly unique, and derived in a constant way from the machine ID but there will be no way to retrieve
the original machine ID from the application-specific one. The
<citerefentry><refentrytitle>sd_id128_get_machine_app_specific</refentrytitle><manvolnum>3</manvolnum></citerefentry>
API provides an implementation of such an algorithm.</para>
</refsect1>
<refsect1>
<title>Initialization</title>
<para>Each machine should have a non-empty ID in normal operation. The ID of each
2018-06-12 22:19:21 +08:00
machine should be unique. To achieve those objectives,
<filename>/etc/machine-id</filename> can be initialized in a few different ways.
</para>
<para>For normal operating system installations, where a custom image is created for a
specific machine, <filename>/etc/machine-id</filename> should be populated during
installation.</para>
2015-02-04 10:14:13 +08:00
<para>
2015-02-04 10:14:13 +08:00
<citerefentry><refentrytitle>systemd-machine-id-setup</refentrytitle><manvolnum>1</manvolnum></citerefentry>
may be used by installer tools to initialize the machine ID at install time, but
<filename>/etc/machine-id</filename> may also be written using any other means.
</para>
<para>For operating system images which are created once and used on multiple machines, for example for
containers or in the cloud, <filename>/etc/machine-id</filename> should be either missing or an empty
file in the generic file system image (the difference between the two options is described under "First
Boot Semantics" below). An ID will be generated during boot and saved to this file if possible. Having an
empty file in place is useful because it allows a temporary file to be bind-mounted over the real file,
in case the image is used read-only. Also see <ulink url="https://systemd.io/BUILDING_IMAGES">Safely
Building Images</ulink>.</para>
<para><citerefentry><refentrytitle>systemd-firstboot</refentrytitle><manvolnum>1</manvolnum></citerefentry>
2018-06-12 22:19:21 +08:00
may be used to initialize <filename>/etc/machine-id</filename> on mounted (but not
booted) system images.</para>
<para>When a machine is booted with
<citerefentry><refentrytitle>systemd</refentrytitle><manvolnum>1</manvolnum></citerefentry>
the ID of the machine will be established. If <varname>systemd.machine_id=</varname>
or <option>--machine-id=</option> options (see first section) are specified, this
value will be used. Otherwise, the value in <filename>/etc/machine-id</filename> will
be used. If this file is empty or missing, <filename>systemd</filename> will attempt
to use the D-Bus machine ID from <filename>/var/lib/dbus/machine-id</filename>, the
value of the kernel command line option <varname>container_uuid</varname>, the KVM DMI
<filename>product_uuid</filename> or the devicetree <filename>vm,uuid</filename>
(on KVM systems), and finally a randomly generated UUID.</para>
<para>After the machine ID is established,
<citerefentry><refentrytitle>systemd</refentrytitle><manvolnum>1</manvolnum></citerefentry>
will attempt to save it to <filename>/etc/machine-id</filename>. If this fails, it
will attempt to bind-mount a temporary file over <filename>/etc/machine-id</filename>.
It is an error if the file system is read-only and does not contain a (possibly empty)
<filename>/etc/machine-id</filename> file.</para>
<para><citerefentry><refentrytitle>systemd-machine-id-commit.service</refentrytitle><manvolnum>8</manvolnum></citerefentry>
will attempt to write the machine ID to the file system if
<filename>/etc/machine-id</filename> or <filename>/etc/</filename> are read-only during
early boot but become writable later on.</para>
2015-02-04 10:14:13 +08:00
</refsect1>
<refsect1>
<title>First Boot Semantics</title>
manager: fix/change evaluation of ConditionFirstBoot The code to evaluate the kernel command line option was busted because it was doing 'return b == !!r' at a point where 'r > 0'. Thus we'd return "true" in both cases: $ SYSTEMD_PROC_CMDLINE=systemd.condition-first-boot build/systemd-analyze condition 'ConditionFirstBoot=true' test.service: ConditionFirstBoot=true succeeded. Conditions succeeded. $ SYSTEMD_PROC_CMDLINE=systemd.condition-first-boot build/systemd-analyze condition 'ConditionFirstBoot=false' test.service: ConditionFirstBoot=false succeeded. Conditions succeeded. We only use 'ConditionFirstBoot=true' in units, so this wasn't noticed. But I think the logic is broken in general: the condition should evaluate as true only during initial boot. If we rerun the units at later points, we should not consider ConditionFirstBoot to be true. Also, the first boot logic is also used in pid1 itself. AFAICT, for two things: in first boot machine-id is initialized transiently (this allows first-boot operations to be restarted if boot fails), and preset-all is executed. But this logic was different and separate from the logic to evaluate ConditionFirstBoot. The distinction is abolished, and the operations in pid1 now use the same logic as ConditionFirstBoot, which means that the kernel command line option is checked, and condition_test_first_boot() just tests whether pid1 thinks we're in first boot. This makes things easier to grok for the user: there's just one condition for "first boot" and it applies to both pid1 and units.
2022-09-30 18:50:40 +08:00
<para><filename>/etc/machine-id</filename> is used to decide whether a boot is the first one. The rules
are as follows:</para>
<orderedlist>
manager: fix/change evaluation of ConditionFirstBoot The code to evaluate the kernel command line option was busted because it was doing 'return b == !!r' at a point where 'r > 0'. Thus we'd return "true" in both cases: $ SYSTEMD_PROC_CMDLINE=systemd.condition-first-boot build/systemd-analyze condition 'ConditionFirstBoot=true' test.service: ConditionFirstBoot=true succeeded. Conditions succeeded. $ SYSTEMD_PROC_CMDLINE=systemd.condition-first-boot build/systemd-analyze condition 'ConditionFirstBoot=false' test.service: ConditionFirstBoot=false succeeded. Conditions succeeded. We only use 'ConditionFirstBoot=true' in units, so this wasn't noticed. But I think the logic is broken in general: the condition should evaluate as true only during initial boot. If we rerun the units at later points, we should not consider ConditionFirstBoot to be true. Also, the first boot logic is also used in pid1 itself. AFAICT, for two things: in first boot machine-id is initialized transiently (this allows first-boot operations to be restarted if boot fails), and preset-all is executed. But this logic was different and separate from the logic to evaluate ConditionFirstBoot. The distinction is abolished, and the operations in pid1 now use the same logic as ConditionFirstBoot, which means that the kernel command line option is checked, and condition_test_first_boot() just tests whether pid1 thinks we're in first boot. This makes things easier to grok for the user: there's just one condition for "first boot" and it applies to both pid1 and units.
2022-09-30 18:50:40 +08:00
<listitem><para>The kernel command argument <varname>systemd.condition-first-boot=</varname> may be
used to override the autodetection logic, see
<citerefentry><refentrytitle>kernel-command-line</refentrytitle><manvolnum>7</manvolnum></citerefentry>.
</para></listitem>
<listitem><para>Otherwise, if <filename>/etc/machine-id</filename> does not exist, this is a first
boot. During early boot, <command>systemd</command> will write <literal>uninitialized\n</literal> to
this file and overmount a temporary file which contains the actual machine ID. Later (after
<filename>first-boot-complete.target</filename> has been reached), the real machine ID will be written
to disk.</para></listitem>
<listitem><para>If <filename>/etc/machine-id</filename> contains the string <literal>uninitialized</literal>,
manager: fix/change evaluation of ConditionFirstBoot The code to evaluate the kernel command line option was busted because it was doing 'return b == !!r' at a point where 'r > 0'. Thus we'd return "true" in both cases: $ SYSTEMD_PROC_CMDLINE=systemd.condition-first-boot build/systemd-analyze condition 'ConditionFirstBoot=true' test.service: ConditionFirstBoot=true succeeded. Conditions succeeded. $ SYSTEMD_PROC_CMDLINE=systemd.condition-first-boot build/systemd-analyze condition 'ConditionFirstBoot=false' test.service: ConditionFirstBoot=false succeeded. Conditions succeeded. We only use 'ConditionFirstBoot=true' in units, so this wasn't noticed. But I think the logic is broken in general: the condition should evaluate as true only during initial boot. If we rerun the units at later points, we should not consider ConditionFirstBoot to be true. Also, the first boot logic is also used in pid1 itself. AFAICT, for two things: in first boot machine-id is initialized transiently (this allows first-boot operations to be restarted if boot fails), and preset-all is executed. But this logic was different and separate from the logic to evaluate ConditionFirstBoot. The distinction is abolished, and the operations in pid1 now use the same logic as ConditionFirstBoot, which means that the kernel command line option is checked, and condition_test_first_boot() just tests whether pid1 thinks we're in first boot. This makes things easier to grok for the user: there's just one condition for "first boot" and it applies to both pid1 and units.
2022-09-30 18:50:40 +08:00
a boot is also considered the first boot. The same mechanism as above applies.</para></listitem>
<listitem><para>If <filename>/etc/machine-id</filename> exists and is empty, a boot is
manager: fix/change evaluation of ConditionFirstBoot The code to evaluate the kernel command line option was busted because it was doing 'return b == !!r' at a point where 'r > 0'. Thus we'd return "true" in both cases: $ SYSTEMD_PROC_CMDLINE=systemd.condition-first-boot build/systemd-analyze condition 'ConditionFirstBoot=true' test.service: ConditionFirstBoot=true succeeded. Conditions succeeded. $ SYSTEMD_PROC_CMDLINE=systemd.condition-first-boot build/systemd-analyze condition 'ConditionFirstBoot=false' test.service: ConditionFirstBoot=false succeeded. Conditions succeeded. We only use 'ConditionFirstBoot=true' in units, so this wasn't noticed. But I think the logic is broken in general: the condition should evaluate as true only during initial boot. If we rerun the units at later points, we should not consider ConditionFirstBoot to be true. Also, the first boot logic is also used in pid1 itself. AFAICT, for two things: in first boot machine-id is initialized transiently (this allows first-boot operations to be restarted if boot fails), and preset-all is executed. But this logic was different and separate from the logic to evaluate ConditionFirstBoot. The distinction is abolished, and the operations in pid1 now use the same logic as ConditionFirstBoot, which means that the kernel command line option is checked, and condition_test_first_boot() just tests whether pid1 thinks we're in first boot. This makes things easier to grok for the user: there's just one condition for "first boot" and it applies to both pid1 and units.
2022-09-30 18:50:40 +08:00
<emphasis>not</emphasis> considered the first boot. <command>systemd</command> will still bind-mount a file
containing the actual machine-id over it and later try to commit it to disk (if <filename>/etc/</filename> is
writable).</para></listitem>
<listitem><para>If <filename>/etc/machine-id</filename> already contains a valid machine-id, this is
not a first boot.</para></listitem>
</orderedlist>
manager: fix/change evaluation of ConditionFirstBoot The code to evaluate the kernel command line option was busted because it was doing 'return b == !!r' at a point where 'r > 0'. Thus we'd return "true" in both cases: $ SYSTEMD_PROC_CMDLINE=systemd.condition-first-boot build/systemd-analyze condition 'ConditionFirstBoot=true' test.service: ConditionFirstBoot=true succeeded. Conditions succeeded. $ SYSTEMD_PROC_CMDLINE=systemd.condition-first-boot build/systemd-analyze condition 'ConditionFirstBoot=false' test.service: ConditionFirstBoot=false succeeded. Conditions succeeded. We only use 'ConditionFirstBoot=true' in units, so this wasn't noticed. But I think the logic is broken in general: the condition should evaluate as true only during initial boot. If we rerun the units at later points, we should not consider ConditionFirstBoot to be true. Also, the first boot logic is also used in pid1 itself. AFAICT, for two things: in first boot machine-id is initialized transiently (this allows first-boot operations to be restarted if boot fails), and preset-all is executed. But this logic was different and separate from the logic to evaluate ConditionFirstBoot. The distinction is abolished, and the operations in pid1 now use the same logic as ConditionFirstBoot, which means that the kernel command line option is checked, and condition_test_first_boot() just tests whether pid1 thinks we're in first boot. This makes things easier to grok for the user: there's just one condition for "first boot" and it applies to both pid1 and units.
2022-09-30 18:50:40 +08:00
<para>If according to the above rules a first boot is detected, units with
<varname>ConditionFirstBoot=yes</varname> will be run and <command>systemd</command> will perform
additional initialization steps, in particular presetting units.</para>
</refsect1>
2015-02-04 10:14:13 +08:00
<refsect1>
<title>Relation to OSF UUIDs</title>
<para>Note that the machine ID historically is not an OSF UUID as defined by <ulink
url="https://tools.ietf.org/html/rfc4122">RFC 4122</ulink>, nor a Microsoft GUID; however, starting with
systemd v30, newly generated machine IDs do qualify as Variant 1 Version 4 UUIDs, as per RFC 4122.</para>
<para>In order to maintain compatibility with existing installations, an application requiring a strictly
RFC 4122 compliant UUID should decode the machine ID, and then (non-reversibly) apply the following
operations to turn it into a valid RFC 4122 Variant 1 Version 4 UUID. With <literal>id</literal> being an
2015-02-04 10:14:13 +08:00
unsigned character array:</para>
<programlisting>/* Set UUID version to 4 --- truly random generation */
id[6] = (id[6] &amp; 0x0F) | 0x40;
/* Set the UUID variant to DCE */
id[8] = (id[8] &amp; 0x3F) | 0x80;</programlisting>
2015-02-04 10:14:13 +08:00
<para>(This code is inspired by
<literal>generate_random_uuid()</literal> of
<filename>drivers/char/random.c</filename> from the Linux kernel
sources.)</para>
</refsect1>
<refsect1>
<title>History</title>
<para>The simple configuration file format of
<filename>/etc/machine-id</filename> originates in the
<filename>/var/lib/dbus/machine-id</filename> file introduced by
D-Bus. In fact, this latter file might be a symlink to
<filename>/etc/machine-id</filename>.</para>
2015-02-04 10:14:13 +08:00
</refsect1>
<refsect1>
<title>See Also</title>
<para>
<citerefentry><refentrytitle>systemd</refentrytitle><manvolnum>1</manvolnum></citerefentry>,
<citerefentry><refentrytitle>systemd-machine-id-setup</refentrytitle><manvolnum>1</manvolnum></citerefentry>,
<citerefentry project='man-pages'><refentrytitle>gethostid</refentrytitle><manvolnum>3</manvolnum></citerefentry>,
2015-02-04 10:14:13 +08:00
<citerefentry><refentrytitle>hostname</refentrytitle><manvolnum>5</manvolnum></citerefentry>,
<citerefentry><refentrytitle>machine-info</refentrytitle><manvolnum>5</manvolnum></citerefentry>,
<citerefentry><refentrytitle>os-release</refentrytitle><manvolnum>5</manvolnum></citerefentry>,
<citerefentry><refentrytitle>sd-id128</refentrytitle><manvolnum>3</manvolnum></citerefentry>,
<citerefentry><refentrytitle>sd_id128_get_machine</refentrytitle><manvolnum>3</manvolnum></citerefentry>,
<citerefentry><refentrytitle>systemd-firstboot</refentrytitle><manvolnum>1</manvolnum></citerefentry>
</para>
</refsect1>
</refentry>