1 systemd System and Service Manager
7 git@github.com:systemd/systemd.git
8 https://github.com/systemd/systemd
11 https://lists.freedesktop.org/mailman/listinfo/systemd-devel
14 #systemd on irc.libera.chat
17 https://github.com/systemd/systemd/issues
20 https://0pointer.de/blog/projects/systemd.html
21 https://www.freedesktop.org/wiki/Software/systemd
29 LGPL-2.1-or-later for all code, exceptions noted in LICENSES/README.md
33 ≥ 4.3 for ambient capabilities
34 ≥ 4.5 for pids controller in cgroup v2
35 ≥ 4.6 for cgroup namespaces
36 ≥ 4.9 for RENAME_NOREPLACE support in vfat
37 ≥ 4.10 for cgroup-bpf egress and ingress hooks
38 ≥ 4.15 for cgroup-bpf device hook and cpu controller in cgroup v2
39 ≥ 4.17 for cgroup-bpf socket address hooks
40 ≥ 4.20 for PSI (used by systemd-oomd)
41 ≥ 5.3 for bounded loops in BPF program
42 ≥ 5.4 for signed Verity images
43 ≥ 5.7 for BPF links and the BPF LSM hook
45 ⛔ Kernel versions below 3.15 ("minimum baseline") are not supported at
46 all, and are missing required functionality (e.g. CLOCK_BOOTTIME
47 support for timerfd_create()).
49 ⚠️ Kernel versions below 4.15 ("recommended baseline") have significant
50 gaps in functionality and are not recommended for use with this version
51 of systemd (e.g. lack sufficiently comprehensive and working cgroupv2
52 support). Taint flag 'old-kernel' will be set. systemd will most likely
53 still function, but upstream support and testing are limited.
55 Kernel Config Options:
57 CONFIG_CGROUPS (it is OK to disable all controllers)
62 CONFIG_UNIX (it requires CONFIG_NET, but every other flag in it is not necessary)
65 CONFIG_FHANDLE (libudev, mount and bind mount handling)
67 udev will fail to work with the legacy sysfs layout:
68 CONFIG_SYSFS_DEPRECATED=n
70 Legacy hotplug slows down the system and confuses udev:
71 CONFIG_UEVENT_HELPER_PATH=""
73 Userspace firmware loading is not supported and should be disabled in
75 CONFIG_FW_LOADER_USER_HELPER=n
77 Some udev rules and virtualization detection relies on it:
80 Support for some SCSI devices serial number retrieval, to create
81 additional symlinks in /dev/disk/ and /dev/tape:
84 Required for PrivateNetwork= in service units:
86 Note that systemd-localed.service and other systemd units use
87 PrivateNetwork so this is effectively required.
89 Required for PrivateUsers= in service units:
92 Optional but strongly recommended:
96 CONFIG_{TMPFS,EXT4_FS,XFS,BTRFS_FS,...}_POSIX_ACL
98 CONFIG_SECCOMP_FILTER (required for seccomp support)
99 CONFIG_KCMP (for the kcmp() syscall, used to be under
100 CONFIG_CHECKPOINT_RESTORE before ~5.12)
102 Required for CPUShares= in resource control unit settings:
104 CONFIG_FAIR_GROUP_SCHED
106 Required for CPUQuota= in resource control unit settings:
109 Required for IPAddressDeny=, IPAddressAllow=, IPIngressFilterPath=,
110 IPEgressFilterPath= in resource control unit settings unit settings:
117 Required for SocketBind{Allow|Deny}=, RestrictNetworkInterfaces= in
118 resource control unit settings:
129 Required for signed Verity images support:
130 CONFIG_DM_VERITY_VERIFY_ROOTHASH_SIG
131 Required to verify signed Verity images using keys enrolled in the MoK
132 (Machine-Owner Key) keyring:
133 CONFIG_DM_VERITY_VERIFY_ROOTHASH_SIG_SECONDARY_KEYRING
134 CONFIG_IMA_ARCH_POLICY
135 CONFIG_INTEGRITY_MACHINE_KEYRING
137 Required for RestrictFileSystems= in service units:
141 CONFIG_DEBUG_INFO_BTF
142 CONFIG_LSM="...,bpf" or kernel booted with lsm="...,bpf".
144 We recommend to turn off Real-Time group scheduling in the kernel when
145 using systemd. RT group scheduling effectively makes RT scheduling
146 unavailable for most userspace, since it requires explicit assignment of
147 RT budgets to each unit whose processes making use of RT. As there's no
148 sensible way to assign these budgets automatically this cannot really be
149 fixed, and it's best to disable group scheduling hence:
150 CONFIG_RT_GROUP_SCHED=n
152 It's a good idea to disable the implicit creation of networking bonding
153 devices by the kernel networking bonding module, so that the
154 automatically created "bond0" interface doesn't conflict with any such
155 device created by systemd-networkd (or other tools). Ideally there would
156 be a kernel compile-time option for this, but there currently isn't. The
157 next best thing is to make this change through a modprobe.d drop-in.
158 This is shipped by default, see modprobe.d/systemd.conf.
160 Required for systemd-nspawn:
161 CONFIG_DEVPTS_MULTIPLE_INSTANCES or Linux kernel >= 4.7
163 Required for systemd-oomd:
166 Note that kernel auditing is broken when used with systemd's container
167 code. When using systemd in conjunction with containers, please make
168 sure to either turn off auditing at runtime using the kernel command
169 line option "audit=0", or turn it off at kernel compile time using:
172 If systemd is compiled with libseccomp support on architectures which do
173 not use socketcall() and where seccomp is supported (this effectively
174 means x86-64 and ARM, but excludes 32-bit x86!), then nspawn will now
175 install a work-around seccomp filter that makes containers boot even
176 with audit being enabled. This works correctly only on kernels 3.14 and
177 newer though. TL;DR: turn audit off, still.
181 libmount >= 2.30 (from util-linux)
182 (util-linux *must* be built without --enable-libmount-support-mtab)
183 libseccomp >= 2.3.1 (optional)
184 libblkid >= 2.24 (from util-linux) (optional)
185 libkmod >= 15 (optional)
186 PAM >= 1.1.2 (optional)
187 libcryptsetup (optional), >= 2.3.0 required for signed Verity images support
190 libbpf >= 0.1.0 (optional)
191 libfdisk >= 2.32 (from util-linux) (optional)
192 libselinux (optional)
194 liblz4 >= 1.3.0 / 130 (optional)
195 libzstd >= 1.4.0 (optional)
197 libqrencode (optional)
198 libmicrohttpd (optional)
199 libidn2 or libidn (optional)
200 gnutls >= 3.1.4 (optional, >= 3.6.0 is required to support DNS-over-TLS with gnutls)
201 openssl >= 1.1.0 (optional, required to support DNS-over-TLS with openssl)
202 elfutils >= 158 (optional)
204 tzdata >= 2014f (optional)
207 docbook-xsl (optional, required for documentation)
208 xsltproc (optional, required for documentation)
209 python >= 3.7 (required by meson too, >= 3.9 is required for ukify)
211 python-pefile (optional, required for ukify)
212 python-lxml (optional, required to build the indices)
213 pyelftools (optional, required for systemd-boot)
217 awk, sed, grep, and similar tools
218 clang >= 10.0, llvm >= 10.0 (optional, required to build BPF programs
219 from source code in C)
221 During runtime, you need the following additional
224 util-linux >= v2.27.1 required (including but not limited to: mount,
225 umount, swapon, swapoff, sulogin,
227 dbus >= 1.4.0 (strictly speaking optional, but recommended)
228 NOTE: If using dbus < 1.9.18, you should override the default
229 policy directory (--with-dbuspolicydir=/etc/dbus-1/system.d).
233 To build in directory build/:
234 meson setup build/ && ninja -C build/
236 Any configuration options can be specified as -Darg=value... arguments
237 to meson. After the build directory is initially configured, meson will
238 refuse to run again, and options must be changed with:
239 meson configure -Darg=value build/
240 meson configure without any arguments will print out available options and
241 their current values.
244 ninja -C build -v some/target
246 sudo meson install -C build/ --no-rebuild
247 DESTDIR=... meson install -C build/
249 A tarball can be created with:
250 v=250 && git archive --prefix=systemd-$v/ v$v | zstd >systemd-$v.tar.zstd
252 When systemd-hostnamed is used, it is strongly recommended to install
253 nss-myhostname to ensure that, in a world of dynamically changing
254 hostnames, the hostname stays resolvable under all circumstances. In
255 fact, systemd-hostnamed will warn if nss-myhostname is not installed.
257 nss-systemd must be enabled on systemd systems, as that's required for
258 DynamicUser= to work. Note that we ship services out-of-the-box that
259 make use of DynamicUser= now, hence enabling nss-systemd is not
262 Note that the build prefix for systemd must be /usr. (Moreover, packages
263 systemd relies on — such as D-Bus — really should use the same prefix,
264 otherwise you are on your own.) -Dsplit-usr=false (which is the default
265 and does not need to be specified) is the recommended setting.
266 -Dsplit-usr=true can be used to give a semblance of support for systems
267 with programs installed split between / and /usr. Moving everything
268 under /usr is strongly encouraged.
270 Additional packages are necessary to run some tests:
271 - nc (used by test/TEST-12-ISSUE-3171)
272 - python (test-udev which is installed is in python)
274 - python-evdev (used by hwdb parsing tests)
275 - strace (used by test/test-functions)
276 - capsh (optional, used by test-execute)
278 POLICY FOR SUPPORT OF DISTRIBUTIONS AND ARCHITECTURES:
279 systemd main branch and latest major or stable releases are generally
280 expected to compile on current versions of popular distributions (at
281 least all non-EOL versions of Fedora, Debian unstable/testing/stable,
282 latest Ubuntu LTS and non-LTS releases, openSUSE Tumbleweed/Leap,
283 CentOS Stream 8 and 9, up-to-date Arch, etc.) We will generally
284 attempt to support also other non-EOL versions of various distros.
285 Features which would break compilation on slightly older distributions
286 will only be introduced if there are significant reasons for this
287 (i.e. supporting them interferes with development or requires too many
288 resources to support). In some cases backports of specific libraries or
289 tools might be required.
291 The policy is similar for architecture support. systemd is regularly
292 tested on popular architectures (currently amd64, i386, arm64, ppc64el,
293 and s390x), but should compile and work also on other architectures, for
294 which support has been added. systemd will emit warnings when
295 architecture-specific constants are not defined.
297 STATIC COMPILATION AND "STANDALONE" BINARIES:
298 systemd provides a public shared libraries libsystemd.so and
299 libudev.so. The latter is deprecated, and the sd-device APIs in
300 libsystemd should be used instead for new code. In addition, systemd is
301 built with a private shared library, libsystemd-shared-<suffix>.so,
302 that also includes the libsystemd code, and by default most systemd
303 binaries are linked to it. Using shared libraries saves disk space and
304 memory at runtime, because only one copy of the code is needed.
306 It is possible to build static versions of systemd public shared
307 libraries (via the configuration options '-Dstatic-libsystemd' and
308 '-Dstatic-libudev'). This allows the libsystemd and libudev code to be
309 linked statically into programs. Note that mixing & matching different
310 versions of libsystemd and systemd is generally not recommended, since
311 various of its APIs wrap internal state and protocols of systemd
312 (e.g. logind and udev databases), which are not considered
313 stable. Hence, using static libraries is not recommended since it
314 generally means that version of the static libsystemd linked into
315 applications and the host systemd are not in sync, and will thus create
316 compatibility problems.
318 In addition, it is possible to disable the use of
319 libsystemd-shared-<suffix>.so for various components (via the
320 configuration options '-Dlink-*-shared'). In this mode, the libsystemd
321 and libsystemd-shared code is linked statically into selected
322 binaries. This option is intended for systems where some of the
323 components are intended to be delivered independently of the main
324 systemd package. Finally, some binaries can be compiled in a second
325 version (via the configuration option '-Dstandalone-binaries'). The
326 version suffixed with ".standalone" has the libsystemd and
327 libsystemd-shared code linked statically. Those binaries are intended
328 as replacements to be used in limited installations where the full
329 systemd suite is not installed. Yet another option is to rebuild
330 systemd with a different '-Dshared-lib-tag' setting, allowing different
331 systemd binaries to be linked to instances of the private shared
332 library that can be installed in parallel.
334 Again: Using the default shared linking is recommended, building static
335 or "standalone" versions is not. Mixing versions of systemd components
336 that would normally be built and used together (in particular various
337 daemons and the manager) is not recommended: we do not test such
338 combinations upstream and cannot provide support. Distributors making
339 use of those options are responsible if things do not work as expected.
342 Default udev rules use the following standard system group names, which
343 need to be resolvable by getgrnam() at any time, even in the very early
344 boot stages, where no other databases and network are available:
346 audio, cdrom, dialout, disk, input, kmem, kvm, lp, render, tape, tty, video
348 During runtime, the journal daemon requires the "systemd-journal" system
349 group to exist. New journal files will be readable by this group (but
350 not writable), which may be used to grant specific users read access. In
351 addition, system groups "wheel" and "adm" will be given read-only access
352 to journal files using systemd-tmpfiles-setup.service.
354 The journal remote daemon requires the "systemd-journal-remote" system
355 user and group to exist. During execution this network facing service
356 will drop privileges and assume this uid/gid for security reasons.
358 Similarly, the network management daemon requires the "systemd-network"
359 system user and group to exist.
361 Similarly, the name resolution daemon requires the "systemd-resolve"
362 system user and group to exist.
364 Similarly, the coredump support requires the "systemd-coredump" system
365 user and group to exist.
368 systemd ships with four glibc NSS modules:
370 nss-myhostname resolves the local hostname to locally configured IP
371 addresses, as well as "localhost" to 127.0.0.1/::1.
373 nss-resolve enables DNS resolution via the systemd-resolved DNS/LLMNR
374 caching stub resolver "systemd-resolved".
376 nss-mymachines enables resolution of all local containers registered
377 with machined to their respective IP addresses.
379 nss-systemd enables resolution of users/group registered via the
380 User/Group Record Lookup API (https://systemd.io/USER_GROUP_API),
381 including all dynamically allocated service users. (See the
382 DynamicUser= setting in unit files.)
384 To make use of these NSS modules, please add them to the "hosts:",
385 "passwd:" and "group:" lines in /etc/nsswitch.conf. The "resolve" module
386 should replace the glibc "dns" module in this file (and don't worry, it
387 chain-loads the "dns" module if it can't talk to resolved).
389 The four modules should be used in the following order:
391 passwd: compat systemd
392 group: compat systemd
393 hosts: files mymachines resolve [!UNAVAIL=return] dns myhostname
396 When calling "systemctl enable/disable/is-enabled" on a unit which is a
397 SysV init.d script, it calls /usr/lib/systemd/systemd-sysv-install;
398 this needs to translate the action into the distribution specific
399 mechanism such as chkconfig or update-rc.d. Packagers need to provide
400 this script if you need this functionality (you don't if you disabled
403 Please see src/systemctl/systemd-sysv-install.SKELETON for how this
404 needs to look like, and provide an implementation at the marked places.
406 WARNINGS and TAINT FLAGS:
407 systemd will warn during early boot if /usr is not already mounted at
408 this point (that means: either located on the same file system as / or
409 already mounted in the initrd). While in systemd itself very little
410 will break if /usr is on a separate late-mounted partition, many of its
411 dependencies very likely will break sooner or later in one form or
412 another. For example, udev rules tend to refer to binaries in /usr,
413 binaries that link to libraries in /usr, or binaries that refer to data
414 files in /usr. Since these breakages are not always directly visible,
415 systemd will warn about this. Such setups are not really supported by
416 the basic set of Linux OS components. Taint flag 'split-usr' will be
417 set when this condition is detected.
419 For more information on this issue consult
420 https://www.freedesktop.org/wiki/Software/systemd/separate-usr-is-broken
422 systemd will warn if the filesystem is not usr-merged (i.e.: /bin, /sbin
423 and /lib* are not symlinks to their counterparts under /usr). Taint flag
424 'unmerged-usr' will be set when this condition is detected.
426 For more information on this issue consult
427 https://www.freedesktop.org/wiki/Software/systemd/TheCaseForTheUsrMerge
429 systemd requires that the /run mount point exists. systemd also
430 requires that /var/run is a symlink to /run. Taint flag 'var-run-bad'
431 will be set when this condition is detected.
433 Systemd will also warn when the cgroup support is unavailable in the
434 kernel (taint flag 'cgroups-missing'), the system is using the old
435 cgroup hierarchy (taint flag 'cgroupsv1'), the hardware clock is
436 running in non-UTC mode (taint flag 'local-hwclock'), the kernel
437 overflow UID or GID are not 65534 (taint flags 'overflowuid-not-65534'
438 and 'overflowgid-not-65534'), the UID or GID range assigned to the
439 running systemd instance covers less than 0…65534 (taint flags
440 'short-uid-range' and 'short-gid-range').
442 Taint conditions are logged during boot, but may also be checked at any
445 busctl get-property org.freedesktop.systemd1 /org/freedesktop/systemd1 org.freedesktop.systemd1.Manager Tainted
447 See org.freedesktop.systemd1(5) for more information.
450 To run systemd under valgrind, compile systemd with the valgrind
451 development headers available (i.e. valgrind-devel or equivalent).
452 Otherwise, false positives will be triggered by code which violates
453 some rules but is actually safe. Note that valgrind generates nice
454 output only on exit(), hence on shutdown we don't execve()
457 STABLE BRANCHES AND BACKPORTS:
458 Stable branches with backported patches are available in the
459 systemd-stable repo at https://github.com/systemd/systemd-stable.
461 Stable branches are started for certain releases of systemd and named
462 after them, e.g. v238-stable. Stable branches are managed by
463 distribution maintainers on an as needed basis. See
464 https://www.freedesktop.org/wiki/Software/systemd/Backports for some
465 more information and examples.