1SYSTEMD.EXEC(5) systemd.exec SYSTEMD.EXEC(5)
2
6 systemd.exec - Execution environment configuration
7
9 service.service, socket.socket, mount.mount, swap.swap
10
12 Unit configuration files for services, sockets, mount points, and swap
13 devices share a subset of configuration options which define the
14 execution environment of spawned processes.
15 This man page lists the configuration options shared by these four unit
17 types. See systemd.unit(5) for the common options of all unit
18 configuration files, and systemd.service(5), systemd.socket(5),
19 systemd.swap(5), and systemd.mount(5) for more information on the
20 specific unit configuration files. The execution specific configuration
21 options are configured in the [Service], [Socket], [Mount], or [Swap]
22 sections, depending on the unit type.
23 In addition, options which control resources through Linux Control
25 Groups (cgroups) are listed in systemd.resource-control(5). Those
26 options complement options listed here.
27
29 A few execution parameters result in additional, automatic dependencies
30 to be added:
31 • Units with WorkingDirectory=, RootDirectory=, RootImage=,
33 RuntimeDirectory=, StateDirectory=, CacheDirectory=, LogsDirectory=
34 or ConfigurationDirectory= set automatically gain dependencies of
35 type Requires= and After= on all mount units required to access the
36 specified paths. This is equivalent to having them listed
37 explicitly in RequiresMountsFor=.
38 • Similarly, units with PrivateTmp= enabled automatically get mount
40 unit dependencies for all mounts required to access /tmp/ and
41 /var/tmp/. They will also gain an automatic After= dependency on
42 systemd-tmpfiles-setup.service(8).
43 • Units whose standard output or error output is connected to journal
45 or kmsg (or their combinations with console output, see below)
46 automatically acquire dependencies of type After= on
47 systemd-journald.socket.
48 • Units using LogNamespace= will automatically gain ordering and
50 requirement dependencies on the two socket units associated with
51 systemd-journald@.service instances.
52
54 The following settings may be used to change a service's view of the
55 filesystem. Please note that the paths must be absolute and must not
56 contain a ".." path component.
57 ExecSearchPath=
59 Takes a colon separated list of absolute paths relative to which
60 the executable used by the Exec*= (e.g. ExecStart=, ExecStop=,
61 etc.) properties can be found. ExecSearchPath= overrides $PATH if
62 $PATH is not supplied by the user through Environment=,
63 EnvironmentFile= or PassEnvironment=. Assigning an empty string
64 removes previous assignments and setting ExecSearchPath= to a value
65 multiple times will append to the previous setting.
66 WorkingDirectory=
68 Takes a directory path relative to the service's root directory
69 specified by RootDirectory=, or the special value "~". Sets the
70 working directory for executed processes. If set to "~", the home
71 directory of the user specified in User= is used. If not set,
72 defaults to the root directory when systemd is running as a system
73 instance and the respective user's home directory if run as user.
74 If the setting is prefixed with the "-" character, a missing
75 working directory is not considered fatal. If
76 RootDirectory=/RootImage= is not set, then WorkingDirectory= is
77 relative to the root of the system running the service manager.
78 Note that setting this parameter might result in additional
79 dependencies to be added to the unit (see above).
80 RootDirectory=
82 Takes a directory path relative to the host's root directory (i.e.
83 the root of the system running the service manager). Sets the root
84 directory for executed processes, with the chroot(2) system call.
85 If this is used, it must be ensured that the process binary and all
86 its auxiliary files are available in the chroot() jail. Note that
87 setting this parameter might result in additional dependencies to
88 be added to the unit (see above).
89 The MountAPIVFS= and PrivateUsers= settings are particularly useful
91 in conjunction with RootDirectory=. For details, see below.
92 If RootDirectory=/RootImage= are used together with NotifyAccess=
94 the notification socket is automatically mounted from the host into
95 the root environment, to ensure the notification interface can work
96 correctly.
97 Note that services using RootDirectory=/RootImage= will not be able
99 to log via the syslog or journal protocols to the host logging
100 infrastructure, unless the relevant sockets are mounted from the
101 host, specifically:
102 Example 1. Mounting logging sockets into root environment
104 BindReadOnlyPaths=/dev/log /run/systemd/journal/socket /run/systemd/journal/stdout
106 This option is only available for system services and is not
108 supported for services running in per-user instances of the service
109 manager.
110 RootImage=
112 Takes a path to a block device node or regular file as argument.
113 This call is similar to RootDirectory= however mounts a file system
114 hierarchy from a block device node or loopback file instead of a
115 directory. The device node or file system image file needs to
116 contain a file system without a partition table, or a file system
117 within an MBR/MS-DOS or GPT partition table with only a single
118 Linux-compatible partition, or a set of file systems within a GPT
119 partition table that follows the Discoverable Partitions
120 Specification[1].
121 When DevicePolicy= is set to "closed" or "strict", or set to "auto"
123 and DeviceAllow= is set, then this setting adds /dev/loop-control
124 with rw mode, "block-loop" and "block-blkext" with rwm mode to
125 DeviceAllow=. See systemd.resource-control(5) for the details about
126 DevicePolicy= or DeviceAllow=. Also, see PrivateDevices= below, as
127 it may change the setting of DevicePolicy=.
128 Units making use of RootImage= automatically gain an After=
130 dependency on systemd-udevd.service.
131 This option is only available for system services and is not
133 supported for services running in per-user instances of the service
134 manager.
135 RootImageOptions=
137 Takes a comma-separated list of mount options that will be used on
138 disk images specified by RootImage=. Optionally a partition name
139 can be prefixed, followed by colon, in case the image has multiple
140 partitions, otherwise partition name "root" is implied. Options for
141 multiple partitions can be specified in a single line with space
142 separators. Assigning an empty string removes previous assignments.
143 Duplicated options are ignored. For a list of valid mount options,
144 please refer to mount(8).
145 Valid partition names follow the Discoverable Partitions
147 Specification[1]: root, usr, home, srv, esp, xbootldr, tmp, var.
148 This option is only available for system services and is not
150 supported for services running in per-user instances of the service
151 manager.
152 RootHash=
154 Takes a data integrity (dm-verity) root hash specified in
155 hexadecimal, or the path to a file containing a root hash in ASCII
156 hexadecimal format. This option enables data integrity checks using
157 dm-verity, if the used image contains the appropriate integrity
158 data (see above) or if RootVerity= is used. The specified hash must
159 match the root hash of integrity data, and is usually at least 256
160 bits (and hence 64 formatted hexadecimal characters) long (in case
161 of SHA256 for example). If this option is not specified, but the
162 image file carries the "user.verity.roothash" extended file
163 attribute (see xattr(7)), then the root hash is read from it, also
164 as formatted hexadecimal characters. If the extended file attribute
165 is not found (or is not supported by the underlying file system),
166 but a file with the .roothash suffix is found next to the image
167 file, bearing otherwise the same name (except if the image has the
168 .raw suffix, in which case the root hash file must not have it in
169 its name), the root hash is read from it and automatically used,
170 also as formatted hexadecimal characters.
171 If the disk image contains a separate /usr/ partition it may also
173 be Verity protected, in which case the root hash may configured via
174 an extended attribute "user.verity.usrhash" or a .usrhash file
175 adjacent to the disk image. There's currently no option to
176 configure the root hash for the /usr/ file system via the unit file
177 directly.
178 This option is only available for system services and is not
180 supported for services running in per-user instances of the service
181 manager.
182 RootHashSignature=
184 Takes a PKCS7 signature of the RootHash= option as a path to a
185 DER-encoded signature file, or as an ASCII base64 string encoding
186 of a DER-encoded signature prefixed by "base64:". The dm-verity
187 volume will only be opened if the signature of the root hash is
188 valid and signed by a public key present in the kernel keyring. If
189 this option is not specified, but a file with the .roothash.p7s
190 suffix is found next to the image file, bearing otherwise the same
191 name (except if the image has the .raw suffix, in which case the
192 signature file must not have it in its name), the signature is read
193 from it and automatically used.
194 If the disk image contains a separate /usr/ partition it may also
196 be Verity protected, in which case the signature for the root hash
197 may configured via a .usrhash.p7s file adjacent to the disk image.
198 There's currently no option to configure the root hash signature
199 for the /usr/ via the unit file directly.
200 This option is only available for system services and is not
202 supported for services running in per-user instances of the service
203 manager.
204 RootVerity=
206 Takes the path to a data integrity (dm-verity) file. This option
207 enables data integrity checks using dm-verity, if RootImage= is
208 used and a root-hash is passed and if the used image itself does
209 not contains the integrity data. The integrity data must be matched
210 by the root hash. If this option is not specified, but a file with
211 the .verity suffix is found next to the image file, bearing
212 otherwise the same name (except if the image has the .raw suffix,
213 in which case the verity data file must not have it in its name),
214 the verity data is read from it and automatically used.
215 This option is supported only for disk images that contain a single
217 file system, without an enveloping partition table. Images that
218 contain a GPT partition table should instead include both root file
219 system and matching Verity data in the same image, implementing the
220 Discoverable Partitions Specification[1].
221 This option is only available for system services and is not
223 supported for services running in per-user instances of the service
224 manager.
225 MountAPIVFS=
227 Takes a boolean argument. If on, a private mount namespace for the
228 unit's processes is created and the API file systems /proc/, /sys/,
229 /dev/ and /run/ (as an empty "tmpfs") are mounted inside of it,
230 unless they are already mounted. Note that this option has no
231 effect unless used in conjunction with RootDirectory=/RootImage= as
232 these four mounts are generally mounted in the host anyway, and
233 unless the root directory is changed, the private mount namespace
234 will be a 1:1 copy of the host's, and include these four mounts.
235 Note that the /dev/ file system of the host is bind mounted if this
236 option is used without PrivateDevices=. To run the service with a
237 private, minimal version of /dev/, combine this option with
238 PrivateDevices=.
239 In order to allow propagating mounts at runtime in a safe manner,
241 /run/systemd/propagate on the host will be used to set up new
242 mounts, and /run/host/incoming/ in the private namespace will be
243 used as an intermediate step to store them before being moved to
244 the final mount point.
245 This option is only available for system services and is not
247 supported for services running in per-user instances of the service
248 manager.
249 ProtectProc=
251 Takes one of "noaccess", "invisible", "ptraceable" or "default"
252 (which it defaults to). When set, this controls the "hidepid="
253 mount option of the "procfs" instance for the unit that controls
254 which directories with process metainformation (/proc/PID) are
255 visible and accessible: when set to "noaccess" the ability to
256 access most of other users' process metadata in /proc/ is taken
257 away for processes of the service. When set to "invisible"
258 processes owned by other users are hidden from /proc/. If
259 "ptraceable" all processes that cannot be ptrace()'ed by a process
260 are hidden to it. If "default" no restrictions on /proc/ access or
261 visibility are made. For further details see The /proc
262 Filesystem[2]. It is generally recommended to run most system
263 services with this option set to "invisible". This option is
264 implemented via file system namespacing, and thus cannot be used
265 with services that shall be able to install mount points in the
266 host file system hierarchy. Note that the root user is unaffected
267 by this option, so to be effective it has to be used together with
268 User= or DynamicUser=yes, and also without the "CAP_SYS_PTRACE"
269 capability, which also allows a process to bypass this feature. It
270 cannot be used for services that need to access metainformation
271 about other users' processes. This option implies MountAPIVFS=.
272 If the kernel doesn't support per-mount point hidepid= mount
274 options this setting remains without effect, and the unit's
275 processes will be able to access and see other process as if the
276 option was not used.
277 This option is only available for system services and is not
279 supported for services running in per-user instances of the service
280 manager.
281 ProcSubset=
283 Takes one of "all" (the default) and "pid". If "pid", all files and
284 directories not directly associated with process management and
285 introspection are made invisible in the /proc/ file system
286 configured for the unit's processes. This controls the "subset="
287 mount option of the "procfs" instance for the unit. For further
288 details see The /proc Filesystem[2]. Note that Linux exposes
289 various kernel APIs via /proc/, which are made unavailable with
290 this setting. Since these APIs are used frequently this option is
291 useful only in a few, specific cases, and is not suitable for most
292 non-trivial programs.
293 Much like ProtectProc= above, this is implemented via file system
295 mount namespacing, and hence the same restrictions apply: it is
296 only available to system services, it disables mount propagation to
297 the host mount table, and it implies MountAPIVFS=. Also, like
298 ProtectProc= this setting is gracefully disabled if the used kernel
299 does not support the "subset=" mount option of "procfs".
300 BindPaths=, BindReadOnlyPaths=
302 Configures unit-specific bind mounts. A bind mount makes a
303 particular file or directory available at an additional place in
304 the unit's view of the file system. Any bind mounts created with
305 this option are specific to the unit, and are not visible in the
306 host's mount table. This option expects a whitespace separated list
307 of bind mount definitions. Each definition consists of a
308 colon-separated triple of source path, destination path and option
309 string, where the latter two are optional. If only a source path is
310 specified the source and destination is taken to be the same. The
311 option string may be either "rbind" or "norbind" for configuring a
312 recursive or non-recursive bind mount. If the destination path is
313 omitted, the option string must be omitted too. Each bind mount
314 definition may be prefixed with "-", in which case it will be
315 ignored when its source path does not exist.
316 BindPaths= creates regular writable bind mounts (unless the source
318 file system mount is already marked read-only), while
319 BindReadOnlyPaths= creates read-only bind mounts. These settings
320 may be used more than once, each usage appends to the unit's list
321 of bind mounts. If the empty string is assigned to either of these
322 two options the entire list of bind mounts defined prior to this is
323 reset. Note that in this case both read-only and regular bind
324 mounts are reset, regardless which of the two settings is used.
325 This option is particularly useful when RootDirectory=/RootImage=
327 is used. In this case the source path refers to a path on the host
328 file system, while the destination path refers to a path below the
329 root directory of the unit.
330 Note that the destination directory must exist or systemd must be
332 able to create it. Thus, it is not possible to use those options
333 for mount points nested underneath paths specified in
334 InaccessiblePaths=, or under /home/ and other protected directories
335 if ProtectHome=yes is specified. TemporaryFileSystem= with ":ro"
336 or ProtectHome=tmpfs should be used instead.
337 This option is only available for system services and is not
339 supported for services running in per-user instances of the service
340 manager.
341 MountImages=
343 This setting is similar to RootImage= in that it mounts a file
344 system hierarchy from a block device node or loopback file, but the
345 destination directory can be specified as well as mount options.
346 This option expects a whitespace separated list of mount
347 definitions. Each definition consists of a colon-separated tuple of
348 source path and destination definitions, optionally followed by
349 another colon and a list of mount options.
350 Mount options may be defined as a single comma-separated list of
352 options, in which case they will be implicitly applied to the root
353 partition on the image, or a series of colon-separated tuples of
354 partition name and mount options. Valid partition names and mount
355 options are the same as for RootImageOptions= setting described
356 above.
357 Each mount definition may be prefixed with "-", in which case it
359 will be ignored when its source path does not exist. The source
360 argument is a path to a block device node or regular file. If
361 source or destination contain a ":", it needs to be escaped as
362 "\:". The device node or file system image file needs to follow the
363 same rules as specified for RootImage=. Any mounts created with
364 this option are specific to the unit, and are not visible in the
365 host's mount table.
366 These settings may be used more than once, each usage appends to
368 the unit's list of mount paths. If the empty string is assigned,
369 the entire list of mount paths defined prior to this is reset.
370 Note that the destination directory must exist or systemd must be
372 able to create it. Thus, it is not possible to use those options
373 for mount points nested underneath paths specified in
374 InaccessiblePaths=, or under /home/ and other protected directories
375 if ProtectHome=yes is specified.
376 When DevicePolicy= is set to "closed" or "strict", or set to "auto"
378 and DeviceAllow= is set, then this setting adds /dev/loop-control
379 with rw mode, "block-loop" and "block-blkext" with rwm mode to
380 DeviceAllow=. See systemd.resource-control(5) for the details about
381 DevicePolicy= or DeviceAllow=. Also, see PrivateDevices= below, as
382 it may change the setting of DevicePolicy=.
383 This option is only available for system services and is not
385 supported for services running in per-user instances of the service
386 manager.
387 ExtensionImages=
389 This setting is similar to MountImages= in that it mounts a file
390 system hierarchy from a block device node or loopback file, but
391 instead of providing a destination path, an overlay will be set up.
392 This option expects a whitespace separated list of mount
393 definitions. Each definition consists of a source path, optionally
394 followed by a colon and a list of mount options.
395 A read-only OverlayFS will be set up on top of /usr/ and /opt/
397 hierarchies. The order in which the images are listed will
398 determine the order in which the overlay is laid down: images
399 specified first to last will result in overlayfs layers bottom to
400 top.
401 Mount options may be defined as a single comma-separated list of
403 options, in which case they will be implicitly applied to the root
404 partition on the image, or a series of colon-separated tuples of
405 partition name and mount options. Valid partition names and mount
406 options are the same as for RootImageOptions= setting described
407 above.
408 Each mount definition may be prefixed with "-", in which case it
410 will be ignored when its source path does not exist. The source
411 argument is a path to a block device node or regular file. If the
412 source path contains a ":", it needs to be escaped as "\:". The
413 device node or file system image file needs to follow the same
414 rules as specified for RootImage=. Any mounts created with this
415 option are specific to the unit, and are not visible in the host's
416 mount table.
417 These settings may be used more than once, each usage appends to
419 the unit's list of image paths. If the empty string is assigned,
420 the entire list of mount paths defined prior to this is reset.
421 Each image must carry a
423 /usr/lib/extension-release.d/extension-release.IMAGE file, with the
424 appropriate metadata which matches RootImage=/RootDirectory= or the
425 host. See: os-release(5).
426 When DevicePolicy= is set to "closed" or "strict", or set to "auto"
428 and DeviceAllow= is set, then this setting adds /dev/loop-control
429 with rw mode, "block-loop" and "block-blkext" with rwm mode to
430 DeviceAllow=. See systemd.resource-control(5) for the details about
431 DevicePolicy= or DeviceAllow=. Also, see PrivateDevices= below, as
432 it may change the setting of DevicePolicy=.
433 This option is only available for system services and is not
435 supported for services running in per-user instances of the service
436 manager.
437
439 These options are only available for system services and are not
440 supported for services running in per-user instances of the service
441 manager.
442 User=, Group=
444 Set the UNIX user or group that the processes are executed as,
445 respectively. Takes a single user or group name, or a numeric ID as
446 argument. For system services (services run by the system service
447 manager, i.e. managed by PID 1) and for user services of the root
448 user (services managed by root's instance of systemd --user), the
449 default is "root", but User= may be used to specify a different
450 user. For user services of any other user, switching user identity
451 is not permitted, hence the only valid setting is the same user the
452 user's service manager is running as. If no group is set, the
453 default group of the user is used. This setting does not affect
454 commands whose command line is prefixed with "+".
455 Note that this enforces only weak restrictions on the user/group
457 name syntax, but will generate warnings in many cases where
458 user/group names do not adhere to the following rules: the
459 specified name should consist only of the characters a-z, A-Z, 0-9,
460 "_" and "-", except for the first character which must be one of
461 a-z, A-Z and "_" (i.e. digits and "-" are not permitted as first
462 character). The user/group name must have at least one character,
463 and at most 31. These restrictions are made in order to avoid
464 ambiguities and to ensure user/group names and unit files remain
465 portable among Linux systems. For further details on the names
466 accepted and the names warned about see User/Group Name Syntax[3].
467 When used in conjunction with DynamicUser= the user/group name
469 specified is dynamically allocated at the time the service is
470 started, and released at the time the service is stopped — unless
471 it is already allocated statically (see below). If DynamicUser= is
472 not used the specified user and group must have been created
473 statically in the user database no later than the moment the
474 service is started, for example using the sysusers.d(5) facility,
475 which is applied at boot or package install time. If the user does
476 not exist by then program invocation will fail.
477 If the User= setting is used the supplementary group list is
479 initialized from the specified user's default group list, as
480 defined in the system's user and group database. Additional groups
481 may be configured through the SupplementaryGroups= setting (see
482 below).
483 DynamicUser=
485 Takes a boolean parameter. If set, a UNIX user and group pair is
486 allocated dynamically when the unit is started, and released as
487 soon as it is stopped. The user and group will not be added to
488 /etc/passwd or /etc/group, but are managed transiently during
489 runtime. The nss-systemd(8) glibc NSS module provides integration
490 of these dynamic users/groups into the system's user and group
491 databases. The user and group name to use may be configured via
492 User= and Group= (see above). If these options are not used and
493 dynamic user/group allocation is enabled for a unit, the name of
494 the dynamic user/group is implicitly derived from the unit name. If
495 the unit name without the type suffix qualifies as valid user name
496 it is used directly, otherwise a name incorporating a hash of it is
497 used. If a statically allocated user or group of the configured
498 name already exists, it is used and no dynamic user/group is
499 allocated. Note that if User= is specified and the static group
500 with the name exists, then it is required that the static user with
501 the name already exists. Similarly, if Group= is specified and the
502 static user with the name exists, then it is required that the
503 static group with the name already exists. Dynamic users/groups are
504 allocated from the UID/GID range 61184...65519. It is recommended
505 to avoid this range for regular system or login users. At any point
506 in time each UID/GID from this range is only assigned to zero or
507 one dynamically allocated users/groups in use. However, UID/GIDs
508 are recycled after a unit is terminated. Care should be taken that
509 any processes running as part of a unit for which dynamic
510 users/groups are enabled do not leave files or directories owned by
511 these users/groups around, as a different unit might get the same
512 UID/GID assigned later on, and thus gain access to these files or
513 directories. If DynamicUser= is enabled, RemoveIPC= and PrivateTmp=
514 are implied (and cannot be turned off). This ensures that the
515 lifetime of IPC objects and temporary files created by the executed
516 processes is bound to the runtime of the service, and hence the
517 lifetime of the dynamic user/group. Since /tmp/ and /var/tmp/ are
518 usually the only world-writable directories on a system this
519 ensures that a unit making use of dynamic user/group allocation
520 cannot leave files around after unit termination. Furthermore
521 NoNewPrivileges= and RestrictSUIDSGID= are implicitly enabled (and
522 cannot be disabled), to ensure that processes invoked cannot take
523 benefit or create SUID/SGID files or directories. Moreover
524 ProtectSystem=strict and ProtectHome=read-only are implied, thus
525 prohibiting the service to write to arbitrary file system
526 locations. In order to allow the service to write to certain
527 directories, they have to be allow-listed using ReadWritePaths=,
528 but care must be taken so that UID/GID recycling doesn't create
529 security issues involving files created by the service. Use
530 RuntimeDirectory= (see below) in order to assign a writable runtime
531 directory to a service, owned by the dynamic user/group and removed
532 automatically when the unit is terminated. Use StateDirectory=,
533 CacheDirectory= and LogsDirectory= in order to assign a set of
534 writable directories for specific purposes to the service in a way
535 that they are protected from vulnerabilities due to UID reuse (see
536 below). If this option is enabled, care should be taken that the
537 unit's processes do not get access to directories outside of these
538 explicitly configured and managed ones. Specifically, do not use
539 BindPaths= and be careful with AF_UNIX file descriptor passing for
540 directory file descriptors, as this would permit processes to
541 create files or directories owned by the dynamic user/group that
542 are not subject to the lifecycle and access guarantees of the
543 service. Defaults to off.
544 SupplementaryGroups=
546 Sets the supplementary Unix groups the processes are executed as.
547 This takes a space-separated list of group names or IDs. This
548 option may be specified more than once, in which case all listed
549 groups are set as supplementary groups. When the empty string is
550 assigned, the list of supplementary groups is reset, and all
551 assignments prior to this one will have no effect. In any way, this
552 option does not override, but extends the list of supplementary
553 groups configured in the system group database for the user. This
554 does not affect commands prefixed with "+".
555 PAMName=
557 Sets the PAM service name to set up a session as. If set, the
558 executed process will be registered as a PAM session under the
559 specified service name. This is only useful in conjunction with the
560 User= setting, and is otherwise ignored. If not set, no PAM session
561 will be opened for the executed processes. See pam(8) for details.
562 Note that for each unit making use of this option a PAM session
564 handler process will be maintained as part of the unit and stays
565 around as long as the unit is active, to ensure that appropriate
566 actions can be taken when the unit and hence the PAM session
567 terminates. This process is named "(sd-pam)" and is an immediate
568 child process of the unit's main process.
569 Note that when this option is used for a unit it is very likely
571 (depending on PAM configuration) that the main unit process will be
572 migrated to its own session scope unit when it is activated. This
573 process will hence be associated with two units: the unit it was
574 originally started from (and for which PAMName= was configured),
575 and the session scope unit. Any child processes of that process
576 will however be associated with the session scope unit only. This
577 has implications when used in combination with NotifyAccess=all, as
578 these child processes will not be able to affect changes in the
579 original unit through notification messages. These messages will be
580 considered belonging to the session scope unit and not the original
581 unit. It is hence not recommended to use PAMName= in combination
582 with NotifyAccess=all.
583
585 These options are only available for system services and are not
586 supported for services running in per-user instances of the service
587 manager.
588 CapabilityBoundingSet=
590 Controls which capabilities to include in the capability bounding
591 set for the executed process. See capabilities(7) for details.
592 Takes a whitespace-separated list of capability names, e.g.
593 CAP_SYS_ADMIN, CAP_DAC_OVERRIDE, CAP_SYS_PTRACE. Capabilities
594 listed will be included in the bounding set, all others are
595 removed. If the list of capabilities is prefixed with "~", all but
596 the listed capabilities will be included, the effect of the
597 assignment inverted. Note that this option also affects the
598 respective capabilities in the effective, permitted and inheritable
599 capability sets. If this option is not used, the capability
600 bounding set is not modified on process execution, hence no limits
601 on the capabilities of the process are enforced. This option may
602 appear more than once, in which case the bounding sets are merged
603 by OR, or by AND if the lines are prefixed with "~" (see below). If
604 the empty string is assigned to this option, the bounding set is
605 reset to the empty capability set, and all prior settings have no
606 effect. If set to "~" (without any further argument), the bounding
607 set is reset to the full set of available capabilities, also
608 undoing any previous settings. This does not affect commands
609 prefixed with "+".
610 Use systemd-analyze(1)'s capability command to retrieve a list of
612 capabilities defined on the local system.
613 Example: if a unit has the following,
615 CapabilityBoundingSet=CAP_A CAP_B
617 CapabilityBoundingSet=CAP_B CAP_C
618 then CAP_A, CAP_B, and CAP_C are set. If the second line is
620 prefixed with "~", e.g.,
621 CapabilityBoundingSet=CAP_A CAP_B
623 CapabilityBoundingSet=~CAP_B CAP_C
624 then, only CAP_A is set.
626 AmbientCapabilities=
628 Controls which capabilities to include in the ambient capability
629 set for the executed process. Takes a whitespace-separated list of
630 capability names, e.g. CAP_SYS_ADMIN, CAP_DAC_OVERRIDE,
631 CAP_SYS_PTRACE. This option may appear more than once in which case
632 the ambient capability sets are merged (see the above examples in
633 CapabilityBoundingSet=). If the list of capabilities is prefixed
634 with "~", all but the listed capabilities will be included, the
635 effect of the assignment inverted. If the empty string is assigned
636 to this option, the ambient capability set is reset to the empty
637 capability set, and all prior settings have no effect. If set to
638 "~" (without any further argument), the ambient capability set is
639 reset to the full set of available capabilities, also undoing any
640 previous settings. Note that adding capabilities to ambient
641 capability set adds them to the process's inherited capability set.
642 Ambient capability sets are useful if you want to execute a process
644 as a non-privileged user but still want to give it some
645 capabilities. Note that in this case option keep-caps is
646 automatically added to SecureBits= to retain the capabilities over
647 the user change. AmbientCapabilities= does not affect commands
648 prefixed with "+".
649
651 NoNewPrivileges=
652 Takes a boolean argument. If true, ensures that the service process
653 and all its children can never gain new privileges through execve()
654 (e.g. via setuid or setgid bits, or filesystem capabilities). This
655 is the simplest and most effective way to ensure that a process and
656 its children can never elevate privileges again. Defaults to false,
657 but certain settings override this and ignore the value of this
658 setting. This is the case when DynamicUser=, LockPersonality=,
659 MemoryDenyWriteExecute=, PrivateDevices=, ProtectClock=,
660 ProtectHostname=, ProtectKernelLogs=, ProtectKernelModules=,
661 ProtectKernelTunables=, RestrictAddressFamilies=,
662 RestrictNamespaces=, RestrictRealtime=, RestrictSUIDSGID=,
663 SystemCallArchitectures=, SystemCallFilter=, or SystemCallLog= are
664 specified. Note that even if this setting is overridden by them,
665 systemctl show shows the original value of this setting. In case
666 the service will be run in a new mount namespace anyway and SELinux
667 is disabled, all file systems are mounted with MS_NOSUID flag. Also
668 see No New Privileges Flag[4].
669 SecureBits=
671 Controls the secure bits set for the executed process. Takes a
672 space-separated combination of options from the following list:
673 keep-caps, keep-caps-locked, no-setuid-fixup,
674 no-setuid-fixup-locked, noroot, and noroot-locked. This option may
675 appear more than once, in which case the secure bits are ORed. If
676 the empty string is assigned to this option, the bits are reset to
677 0. This does not affect commands prefixed with "+". See
678 capabilities(7) for details.
679
681 These options are only available for system services and are not
682 supported for services running in per-user instances of the service
683 manager.
684 SELinuxContext=
686 Set the SELinux security context of the executed process. If set,
687 this will override the automated domain transition. However, the
688 policy still needs to authorize the transition. This directive is
689 ignored if SELinux is disabled. If prefixed by "-", failing to set
690 the SELinux security context will be ignored, but it's still
691 possible that the subsequent execve() may fail if the policy
692 doesn't allow the transition for the non-overridden context. This
693 does not affect commands prefixed with "+". See setexeccon(3) for
694 details.
695 AppArmorProfile=
697 Takes a profile name as argument. The process executed by the unit
698 will switch to this profile when started. Profiles must already be
699 loaded in the kernel, or the unit will fail. If prefixed by "-",
700 all errors will be ignored. This setting has no effect if AppArmor
701 is not enabled. This setting does not affect commands prefixed with
702 "+".
703 SmackProcessLabel=
705 Takes a SMACK64 security label as argument. The process executed by
706 the unit will be started under this label and SMACK will decide
707 whether the process is allowed to run or not, based on it. The
708 process will continue to run under the label specified here unless
709 the executable has its own SMACK64EXEC label, in which case the
710 process will transition to run under that label. When not
711 specified, the label that systemd is running under is used. This
712 directive is ignored if SMACK is disabled.
713 The value may be prefixed by "-", in which case all errors will be
715 ignored. An empty value may be specified to unset previous
716 assignments. This does not affect commands prefixed with "+".
717
719 LimitCPU=, LimitFSIZE=, LimitDATA=, LimitSTACK=, LimitCORE=, LimitRSS=,
720 LimitNOFILE=, LimitAS=, LimitNPROC=, LimitMEMLOCK=, LimitLOCKS=,
721 LimitSIGPENDING=, LimitMSGQUEUE=, LimitNICE=, LimitRTPRIO=,
722 LimitRTTIME=
723 Set soft and hard limits on various resources for executed
724 processes. See setrlimit(2) for details on the resource limit
725 concept. Resource limits may be specified in two formats: either as
726 single value to set a specific soft and hard limit to the same
727 value, or as colon-separated pair soft:hard to set both limits
728 individually (e.g. "LimitAS=4G:16G"). Use the string infinity to
729 configure no limit on a specific resource. The multiplicative
730 suffixes K, M, G, T, P and E (to the base 1024) may be used for
731 resource limits measured in bytes (e.g. "LimitAS=16G"). For the
732 limits referring to time values, the usual time units ms, s, min, h
733 and so on may be used (see systemd.time(7) for details). Note that
734 if no time unit is specified for LimitCPU= the default unit of
735 seconds is implied, while for LimitRTTIME= the default unit of
736 microseconds is implied. Also, note that the effective granularity
737 of the limits might influence their enforcement. For example, time
738 limits specified for LimitCPU= will be rounded up implicitly to
739 multiples of 1s. For LimitNICE= the value may be specified in two
740 syntaxes: if prefixed with "+" or "-", the value is understood as
741 regular Linux nice value in the range -20...19. If not prefixed
742 like this the value is understood as raw resource limit parameter
743 in the range 0...40 (with 0 being equivalent to 1).
744 Note that most process resource limits configured with these
746 options are per-process, and processes may fork in order to acquire
747 a new set of resources that are accounted independently of the
748 original process, and may thus escape limits set. Also note that
749 LimitRSS= is not implemented on Linux, and setting it has no
750 effect. Often it is advisable to prefer the resource controls
751 listed in systemd.resource-control(5) over these per-process
752 limits, as they apply to services as a whole, may be altered
753 dynamically at runtime, and are generally more expressive. For
754 example, MemoryMax= is a more powerful (and working) replacement
755 for LimitRSS=.
756 Resource limits not configured explicitly for a unit default to the
758 value configured in the various DefaultLimitCPU=,
759 DefaultLimitFSIZE=, ... options available in systemd-
760 system.conf(5), and – if not configured there – the kernel or
761 per-user defaults, as defined by the OS (the latter only for user
762 services, see below).
763 For system units these resource limits may be chosen freely. When
765 these settings are configured in a user service (i.e. a service run
766 by the per-user instance of the service manager) they cannot be
767 used to raise the limits above those set for the user manager
768 itself when it was first invoked, as the user's service manager
769 generally lacks the privileges to do so. In user context these
770 configuration options are hence only useful to lower the limits
771 passed in or to raise the soft limit to the maximum of the hard
772 limit as configured for the user. To raise the user's limits
773 further, the available configuration mechanisms differ between
774 operating systems, but typically require privileges. In most cases
775 it is possible to configure higher per-user resource limits via PAM
776 or by setting limits on the system service encapsulating the user's
777 service manager, i.e. the user's instance of user@.service. After
778 making such changes, make sure to restart the user's service
779 manager.
780 Table 1. Resource limit directives, their equivalent ulimit shell
782 commands and the unit used
783 ┌─────────────────┬───────────────────┬─────────────────────┐
784 │Directive │ ulimit equivalent │ Unit │
785 ├─────────────────┼───────────────────┼─────────────────────┤
786 │LimitCPU= │ ulimit -t │ Seconds │
787 ├─────────────────┼───────────────────┼─────────────────────┤
788 │LimitFSIZE= │ ulimit -f │ Bytes │
789 ├─────────────────┼───────────────────┼─────────────────────┤
790 │LimitDATA= │ ulimit -d │ Bytes │
791 ├─────────────────┼───────────────────┼─────────────────────┤
792 │LimitSTACK= │ ulimit -s │ Bytes │
793 ├─────────────────┼───────────────────┼─────────────────────┤
794 │LimitCORE= │ ulimit -c │ Bytes │
795 ├─────────────────┼───────────────────┼─────────────────────┤
796 │LimitRSS= │ ulimit -m │ Bytes │
797 ├─────────────────┼───────────────────┼─────────────────────┤
798 │LimitNOFILE= │ ulimit -n │ Number of File │
799 │ │ │ Descriptors │
800 ├─────────────────┼───────────────────┼─────────────────────┤
801 │LimitAS= │ ulimit -v │ Bytes │
802 ├─────────────────┼───────────────────┼─────────────────────┤
803 │LimitNPROC= │ ulimit -u │ Number of Processes │
804 ├─────────────────┼───────────────────┼─────────────────────┤
805 │LimitMEMLOCK= │ ulimit -l │ Bytes │
806 ├─────────────────┼───────────────────┼─────────────────────┤
807 │LimitLOCKS= │ ulimit -x │ Number of Locks │
808 ├─────────────────┼───────────────────┼─────────────────────┤
809 │LimitSIGPENDING= │ ulimit -i │ Number of Queued │
810 │ │ │ Signals │
811 ├─────────────────┼───────────────────┼─────────────────────┤
812 │LimitMSGQUEUE= │ ulimit -q │ Bytes │
813 ├─────────────────┼───────────────────┼─────────────────────┤
814 │LimitNICE= │ ulimit -e │ Nice Level │
815 ├─────────────────┼───────────────────┼─────────────────────┤
816 │LimitRTPRIO= │ ulimit -r │ Realtime Priority │
817 ├─────────────────┼───────────────────┼─────────────────────┤
818 │LimitRTTIME= │ No equivalent │ Microseconds │
819 └─────────────────┴───────────────────┴─────────────────────┘
820 UMask=
822 Controls the file mode creation mask. Takes an access mode in octal
823 notation. See umask(2) for details. Defaults to 0022 for system
824 units. For user units the default value is inherited from the
825 per-user service manager (whose default is in turn inherited from
826 the system service manager, and thus typically also is 0022 —
827 unless overridden by a PAM module). In order to change the per-user
828 mask for all user services, consider setting the UMask= setting of
829 the user's user@.service system service instance. The per-user
830 umask may also be set via the umask field of a user's JSON User
831 Record[5] (for users managed by systemd-homed.service(8) this field
832 may be controlled via homectl --umask=). It may also be set via a
833 PAM module, such as pam_umask(8).
834 CoredumpFilter=
836 Controls which types of memory mappings will be saved if the
837 process dumps core (using the /proc/pid/coredump_filter file).
838 Takes a whitespace-separated combination of mapping type names or
839 numbers (with the default base 16). Mapping type names are
840 private-anonymous, shared-anonymous, private-file-backed,
841 shared-file-backed, elf-headers, private-huge, shared-huge,
842 private-dax, shared-dax, and the special values all (all types) and
843 default (the kernel default of "private-anonymous shared-anonymous
844 elf-headers private-huge"). See core(5) for the meaning of the
845 mapping types. When specified multiple times, all specified masks
846 are ORed. When not set, or if the empty value is assigned, the
847 inherited value is not changed.
848 Example 2. Add DAX pages to the dump filter
850 CoredumpFilter=default private-dax shared-dax
852 KeyringMode=
854 Controls how the kernel session keyring is set up for the service
855 (see session-keyring(7) for details on the session keyring). Takes
856 one of inherit, private, shared. If set to inherit no special
857 keyring setup is done, and the kernel's default behaviour is
858 applied. If private is used a new session keyring is allocated when
859 a service process is invoked, and it is not linked up with any user
860 keyring. This is the recommended setting for system services, as
861 this ensures that multiple services running under the same system
862 user ID (in particular the root user) do not share their key
863 material among each other. If shared is used a new session keyring
864 is allocated as for private, but the user keyring of the user
865 configured with User= is linked into it, so that keys assigned to
866 the user may be requested by the unit's processes. In this modes
867 multiple units running processes under the same user ID may share
868 key material. Unless inherit is selected the unique invocation ID
869 for the unit (see below) is added as a protected key by the name
870 "invocation_id" to the newly created session keyring. Defaults to
871 private for services of the system service manager and to inherit
872 for non-service units and for services of the user service manager.
873 OOMScoreAdjust=
875 Sets the adjustment value for the Linux kernel's Out-Of-Memory
876 (OOM) killer score for executed processes. Takes an integer between
877 -1000 (to disable OOM killing of processes of this unit) and 1000
878 (to make killing of processes of this unit under memory pressure
879 very likely). See proc.txt[6] for details. If not specified
880 defaults to the OOM score adjustment level of the service manager
881 itself, which is normally at 0.
882 Use the OOMPolicy= setting of service units to configure how the
884 service manager shall react to the kernel OOM killer terminating a
885 process of the service. See systemd.service(5) for details.
886 TimerSlackNSec=
888 Sets the timer slack in nanoseconds for the executed processes. The
889 timer slack controls the accuracy of wake-ups triggered by timers.
890 See prctl(2) for more information. Note that in contrast to most
891 other time span definitions this parameter takes an integer value
892 in nano-seconds if no unit is specified. The usual time units are
893 understood too.
894 Personality=
896 Controls which kernel architecture uname(2) shall report, when
897 invoked by unit processes. Takes one of the architecture
898 identifiers x86, x86-64, ppc, ppc-le, ppc64, ppc64-le, s390 or
899 s390x. Which personality architectures are supported depends on the
900 system architecture. Usually the 64bit versions of the various
901 system architectures support their immediate 32bit personality
902 architecture counterpart, but no others. For example, x86-64
903 systems support the x86-64 and x86 personalities but no others. The
904 personality feature is useful when running 32-bit services on a
905 64-bit host system. If not specified, the personality is left
906 unmodified and thus reflects the personality of the host system's
907 kernel.
908 IgnoreSIGPIPE=
910 Takes a boolean argument. If true, causes SIGPIPE to be ignored in
911 the executed process. Defaults to true because SIGPIPE generally is
912 useful only in shell pipelines.
913
915 Nice=
916 Sets the default nice level (scheduling priority) for executed
917 processes. Takes an integer between -20 (highest priority) and 19
918 (lowest priority). In case of resource contention, smaller values
919 mean more resources will be made available to the unit's processes,
920 larger values mean less resources will be made available. See
921 setpriority(2) for details.
922 CPUSchedulingPolicy=
924 Sets the CPU scheduling policy for executed processes. Takes one of
925 other, batch, idle, fifo or rr. See sched_setscheduler(2) for
926 details.
927 CPUSchedulingPriority=
929 Sets the CPU scheduling priority for executed processes. The
930 available priority range depends on the selected CPU scheduling
931 policy (see above). For real-time scheduling policies an integer
932 between 1 (lowest priority) and 99 (highest priority) can be used.
933 In case of CPU resource contention, smaller values mean less CPU
934 time is made available to the service, larger values mean more. See
935 sched_setscheduler(2) for details.
936 CPUSchedulingResetOnFork=
938 Takes a boolean argument. If true, elevated CPU scheduling
939 priorities and policies will be reset when the executed processes
940 call fork(2), and can hence not leak into child processes. See
941 sched_setscheduler(2) for details. Defaults to false.
942 CPUAffinity=
944 Controls the CPU affinity of the executed processes. Takes a list
945 of CPU indices or ranges separated by either whitespace or commas.
946 Alternatively, takes a special "numa" value in which case systemd
947 automatically derives allowed CPU range based on the value of
948 NUMAMask= option. CPU ranges are specified by the lower and upper
949 CPU indices separated by a dash. This option may be specified more
950 than once, in which case the specified CPU affinity masks are
951 merged. If the empty string is assigned, the mask is reset, all
952 assignments prior to this will have no effect. See
953 sched_setaffinity(2) for details.
954 NUMAPolicy=
956 Controls the NUMA memory policy of the executed processes. Takes a
957 policy type, one of: default, preferred, bind, interleave and
958 local. A list of NUMA nodes that should be associated with the
959 policy must be specified in NUMAMask=. For more details on each
960 policy please see, set_mempolicy(2). For overall overview of NUMA
961 support in Linux see, numa(7).
962 NUMAMask=
964 Controls the NUMA node list which will be applied alongside with
965 selected NUMA policy. Takes a list of NUMA nodes and has the same
966 syntax as a list of CPUs for CPUAffinity= option or special "all"
967 value which will include all available NUMA nodes in the mask. Note
968 that the list of NUMA nodes is not required for default and local
969 policies and for preferred policy we expect a single NUMA node.
970 IOSchedulingClass=
972 Sets the I/O scheduling class for executed processes. Takes one of
973 the strings realtime, best-effort or idle. The kernel's default
974 scheduling class is best-effort at a priority of 4. If the empty
975 string is assigned to this option, all prior assignments to both
976 IOSchedulingClass= and IOSchedulingPriority= have no effect. See
977 ioprio_set(2) for details.
978 IOSchedulingPriority=
980 Sets the I/O scheduling priority for executed processes. Takes an
981 integer between 0 (highest priority) and 7 (lowest priority). In
982 case of I/O contention, smaller values mean more I/O bandwidth is
983 made available to the unit's processes, larger values mean less
984 bandwidth. The available priorities depend on the selected I/O
985 scheduling class (see above). If the empty string is assigned to
986 this option, all prior assignments to both IOSchedulingClass= and
987 IOSchedulingPriority= have no effect. For the kernel's default
988 scheduling class (best-effort) this defaults to 4. See
989 ioprio_set(2) for details.
990
992 The following sandboxing options are an effective way to limit the
993 exposure of the system towards the unit's processes. It is recommended
994 to turn on as many of these options for each unit as is possible
995 without negatively affecting the process' ability to operate. Note that
996 many of these sandboxing features are gracefully turned off on systems
997 where the underlying security mechanism is not available. For example,
998 ProtectSystem= has no effect if the kernel is built without file system
999 namespacing or if the service manager runs in a container manager that
1000 makes file system namespacing unavailable to its payload. Similar,
1001 RestrictRealtime= has no effect on systems that lack support for
1002 SECCOMP system call filtering, or in containers where support for this
1003 is turned off.
1004 Also note that some sandboxing functionality is generally not available
1006 in user services (i.e. services run by the per-user service manager).
1007 Specifically, the various settings requiring file system namespacing
1008 support (such as ProtectSystem=) are not available, as the underlying
1009 kernel functionality is only accessible to privileged processes.
1010 However, most namespacing settings, that will not work on their own in
1011 user services, will work when used in conjunction with
1012 PrivateUsers=true.
1013 ProtectSystem=
1015 Takes a boolean argument or the special values "full" or "strict".
1016 If true, mounts the /usr/ and the boot loader directories (/boot
1017 and /efi) read-only for processes invoked by this unit. If set to
1018 "full", the /etc/ directory is mounted read-only, too. If set to
1019 "strict" the entire file system hierarchy is mounted read-only,
1020 except for the API file system subtrees /dev/, /proc/ and /sys/
1021 (protect these directories using PrivateDevices=,
1022 ProtectKernelTunables=, ProtectControlGroups=). This setting
1023 ensures that any modification of the vendor-supplied operating
1024 system (and optionally its configuration, and local mounts) is
1025 prohibited for the service. It is recommended to enable this
1026 setting for all long-running services, unless they are involved
1027 with system updates or need to modify the operating system in other
1028 ways. If this option is used, ReadWritePaths= may be used to
1029 exclude specific directories from being made read-only. This
1030 setting is implied if DynamicUser= is set. This setting cannot
1031 ensure protection in all cases. In general it has the same
1032 limitations as ReadOnlyPaths=, see below. Defaults to off.
1033 ProtectHome=
1035 Takes a boolean argument or the special values "read-only" or
1036 "tmpfs". If true, the directories /home/, /root, and /run/user are
1037 made inaccessible and empty for processes invoked by this unit. If
1038 set to "read-only", the three directories are made read-only
1039 instead. If set to "tmpfs", temporary file systems are mounted on
1040 the three directories in read-only mode. The value "tmpfs" is
1041 useful to hide home directories not relevant to the processes
1042 invoked by the unit, while still allowing necessary directories to
1043 be made visible when listed in BindPaths= or BindReadOnlyPaths=.
1044 Setting this to "yes" is mostly equivalent to set the three
1046 directories in InaccessiblePaths=. Similarly, "read-only" is mostly
1047 equivalent to ReadOnlyPaths=, and "tmpfs" is mostly equivalent to
1048 TemporaryFileSystem= with ":ro".
1049 It is recommended to enable this setting for all long-running
1051 services (in particular network-facing ones), to ensure they cannot
1052 get access to private user data, unless the services actually
1053 require access to the user's private data. This setting is implied
1054 if DynamicUser= is set. This setting cannot ensure protection in
1055 all cases. In general it has the same limitations as
1056 ReadOnlyPaths=, see below.
1057 This option is only available for system services and is not
1059 supported for services running in per-user instances of the service
1060 manager.
1061 RuntimeDirectory=, StateDirectory=, CacheDirectory=, LogsDirectory=,
1063 ConfigurationDirectory=
1064 These options take a whitespace-separated list of directory names.
1065 The specified directory names must be relative, and may not include
1066 "..". If set, when the unit is started, one or more directories by
1067 the specified names will be created (including their parents) below
1068 the locations defined in the following table. Also, the
1069 corresponding environment variable will be defined with the full
1070 paths of the directories. If multiple directories are set, then in
1071 the environment variable the paths are concatenated with colon
1072 (":").
1073 Table 2. Automatic directory creation and environment variables
1075 ┌────────────────────────┬────────────────┬───────────────────────┬──────────────────────────┐
1076 │Directory │ Below path for │ Below path for │ Environment │
1077 │ │ system units │ user units │ variable set │
1078 ├────────────────────────┼────────────────┼───────────────────────┼──────────────────────────┤
1079 │RuntimeDirectory= │ /run/ │ $XDG_RUNTIME_DIR │ $RUNTIME_DIRECTORY │
1080 ├────────────────────────┼────────────────┼───────────────────────┼──────────────────────────┤
1081 │StateDirectory= │ /var/lib/ │ $XDG_CONFIG_HOME │ $STATE_DIRECTORY │
1082 ├────────────────────────┼────────────────┼───────────────────────┼──────────────────────────┤
1083 │CacheDirectory= │ /var/cache/ │ $XDG_CACHE_HOME │ $CACHE_DIRECTORY │
1084 ├────────────────────────┼────────────────┼───────────────────────┼──────────────────────────┤
1085 │LogsDirectory= │ /var/log/ │ $XDG_CONFIG_HOME/log/ │ $LOGS_DIRECTORY │
1086 ├────────────────────────┼────────────────┼───────────────────────┼──────────────────────────┤
1087 │ConfigurationDirectory= │ /etc/ │ $XDG_CONFIG_HOME │ $CONFIGURATION_DIRECTORY │
1088 └────────────────────────┴────────────────┴───────────────────────┴──────────────────────────┘
1089 In case of RuntimeDirectory= the innermost subdirectories are
1090 removed when the unit is stopped. It is possible to preserve the
1091 specified directories in this case if RuntimeDirectoryPreserve= is
1092 configured to restart or yes (see below). The directories specified
1093 with StateDirectory=, CacheDirectory=, LogsDirectory=,
1094 ConfigurationDirectory= are not removed when the unit is stopped.
1095 Except in case of ConfigurationDirectory=, the innermost specified
1097 directories will be owned by the user and group specified in User=
1098 and Group=. If the specified directories already exist and their
1099 owning user or group do not match the configured ones, all files
1100 and directories below the specified directories as well as the
1101 directories themselves will have their file ownership recursively
1102 changed to match what is configured. As an optimization, if the
1103 specified directories are already owned by the right user and
1104 group, files and directories below of them are left as-is, even if
1105 they do not match what is requested. The innermost specified
1106 directories will have their access mode adjusted to the what is
1107 specified in RuntimeDirectoryMode=, StateDirectoryMode=,
1108 CacheDirectoryMode=, LogsDirectoryMode= and
1109 ConfigurationDirectoryMode=.
1110 These options imply BindPaths= for the specified paths. When
1112 combined with RootDirectory= or RootImage= these paths always
1113 reside on the host and are mounted from there into the unit's file
1114 system namespace.
1115 If DynamicUser= is used, the logic for CacheDirectory=,
1117 LogsDirectory= and StateDirectory= is slightly altered: the
1118 directories are created below /var/cache/private, /var/log/private
1119 and /var/lib/private, respectively, which are host directories made
1120 inaccessible to unprivileged users, which ensures that access to
1121 these directories cannot be gained through dynamic user ID
1122 recycling. Symbolic links are created to hide this difference in
1123 behaviour. Both from perspective of the host and from inside the
1124 unit, the relevant directories hence always appear directly below
1125 /var/cache, /var/log and /var/lib.
1126 Use RuntimeDirectory= to manage one or more runtime directories for
1128 the unit and bind their lifetime to the daemon runtime. This is
1129 particularly useful for unprivileged daemons that cannot create
1130 runtime directories in /run/ due to lack of privileges, and to make
1131 sure the runtime directory is cleaned up automatically after use.
1132 For runtime directories that require more complex or different
1133 configuration or lifetime guarantees, please consider using
1134 tmpfiles.d(5).
1135 RuntimeDirectory=, StateDirectory=, CacheDirectory= and
1137 LogsDirectory= optionally support a second parameter, separated by
1138 ":". The second parameter will be interpreted as a destination path
1139 that will be created as a symlink to the directory. The symlinks
1140 will be created after any BindPaths= or TemporaryFileSystem=
1141 options have been set up, to make ephemeral symlinking possible.
1142 The same source can have multiple symlinks, by using the same first
1143 parameter, but a different second parameter.
1144 The directories defined by these options are always created under
1146 the standard paths used by systemd (/var/, /run/, /etc/, ...). If
1147 the service needs directories in a different location, a different
1148 mechanism has to be used to create them.
1149 tmpfiles.d(5) provides functionality that overlaps with these
1151 options. Using these options is recommended, because the lifetime
1152 of the directories is tied directly to the lifetime of the unit,
1153 and it is not necessary to ensure that the tmpfiles.d configuration
1154 is executed before the unit is started.
1155 To remove any of the directories created by these settings, use the
1157 systemctl clean ... command on the relevant units, see
1158 systemctl(1) for details.
1159 Example: if a system service unit has the following,
1161 RuntimeDirectory=foo/bar baz
1163 the service manager creates /run/foo (if it does not exist),
1165 /run/foo/bar, and /run/baz. The directories /run/foo/bar and
1166 /run/baz except /run/foo are owned by the user and group specified
1167 in User= and Group=, and removed when the service is stopped.
1168 Example: if a system service unit has the following,
1170 RuntimeDirectory=foo/bar
1172 StateDirectory=aaa/bbb ccc
1173 then the environment variable "RUNTIME_DIRECTORY" is set with
1175 "/run/foo/bar", and "STATE_DIRECTORY" is set with
1176 "/var/lib/aaa/bbb:/var/lib/ccc".
1177 Example: if a system service unit has the following,
1179 RuntimeDirectory=foo:bar foo:baz
1181 the service manager creates /run/foo (if it does not exist), and
1183 /run/bar plus /run/baz as symlinks to /run/foo.
1184 RuntimeDirectoryMode=, StateDirectoryMode=, CacheDirectoryMode=,
1186 LogsDirectoryMode=, ConfigurationDirectoryMode=
1187 Specifies the access mode of the directories specified in
1188 RuntimeDirectory=, StateDirectory=, CacheDirectory=,
1189 LogsDirectory=, or ConfigurationDirectory=, respectively, as an
1190 octal number. Defaults to 0755. See "Permissions" in
1191 path_resolution(7) for a discussion of the meaning of permission
1192 bits.
1193 RuntimeDirectoryPreserve=
1195 Takes a boolean argument or restart. If set to no (the default),
1196 the directories specified in RuntimeDirectory= are always removed
1197 when the service stops. If set to restart the directories are
1198 preserved when the service is both automatically and manually
1199 restarted. Here, the automatic restart means the operation
1200 specified in Restart=, and manual restart means the one triggered
1201 by systemctl restart foo.service. If set to yes, then the
1202 directories are not removed when the service is stopped. Note that
1203 since the runtime directory /run/ is a mount point of "tmpfs", then
1204 for system services the directories specified in RuntimeDirectory=
1205 are removed when the system is rebooted.
1206 TimeoutCleanSec=
1208 Configures a timeout on the clean-up operation requested through
1209 systemctl clean ..., see systemctl(1) for details. Takes the usual
1210 time values and defaults to infinity, i.e. by default no timeout is
1211 applied. If a timeout is configured the clean operation will be
1212 aborted forcibly when the timeout is reached, potentially leaving
1213 resources on disk.
1214 ReadWritePaths=, ReadOnlyPaths=, InaccessiblePaths=, ExecPaths=,
1216 NoExecPaths=
1217 Sets up a new file system namespace for executed processes. These
1218 options may be used to limit access a process has to the file
1219 system. Each setting takes a space-separated list of paths relative
1220 to the host's root directory (i.e. the system running the service
1221 manager). Note that if paths contain symlinks, they are resolved
1222 relative to the root directory set with RootDirectory=/RootImage=.
1223 Paths listed in ReadWritePaths= are accessible from within the
1225 namespace with the same access modes as from outside of it. Paths
1226 listed in ReadOnlyPaths= are accessible for reading only, writing
1227 will be refused even if the usual file access controls would permit
1228 this. Nest ReadWritePaths= inside of ReadOnlyPaths= in order to
1229 provide writable subdirectories within read-only directories. Use
1230 ReadWritePaths= in order to allow-list specific paths for write
1231 access if ProtectSystem=strict is used.
1232 Paths listed in InaccessiblePaths= will be made inaccessible for
1234 processes inside the namespace along with everything below them in
1235 the file system hierarchy. This may be more restrictive than
1236 desired, because it is not possible to nest ReadWritePaths=,
1237 ReadOnlyPaths=, BindPaths=, or BindReadOnlyPaths= inside it. For a
1238 more flexible option, see TemporaryFileSystem=.
1239 Content in paths listed in NoExecPaths= are not executable even if
1241 the usual file access controls would permit this. Nest ExecPaths=
1242 inside of NoExecPaths= in order to provide executable content
1243 within non-executable directories.
1244 Non-directory paths may be specified as well. These options may be
1246 specified more than once, in which case all paths listed will have
1247 limited access from within the namespace. If the empty string is
1248 assigned to this option, the specific list is reset, and all prior
1249 assignments have no effect.
1250 Paths in ReadWritePaths=, ReadOnlyPaths=, InaccessiblePaths=,
1252 ExecPaths= and NoExecPaths= may be prefixed with "-", in which case
1253 they will be ignored when they do not exist. If prefixed with "+"
1254 the paths are taken relative to the root directory of the unit, as
1255 configured with RootDirectory=/RootImage=, instead of relative to
1256 the root directory of the host (see above). When combining "-" and
1257 "+" on the same path make sure to specify "-" first, and "+"
1258 second.
1259 Note that these settings will disconnect propagation of mounts from
1261 the unit's processes to the host. This means that this setting may
1262 not be used for services which shall be able to install mount
1263 points in the main mount namespace. For ReadWritePaths= and
1264 ReadOnlyPaths= propagation in the other direction is not affected,
1265 i.e. mounts created on the host generally appear in the unit
1266 processes' namespace, and mounts removed on the host also disappear
1267 there too. In particular, note that mount propagation from host to
1268 unit will result in unmodified mounts to be created in the unit's
1269 namespace, i.e. writable mounts appearing on the host will be
1270 writable in the unit's namespace too, even when propagated below a
1271 path marked with ReadOnlyPaths=! Restricting access with these
1272 options hence does not extend to submounts of a directory that are
1273 created later on. This means the lock-down offered by that setting
1274 is not complete, and does not offer full protection.
1275 Note that the effect of these settings may be undone by privileged
1277 processes. In order to set up an effective sandboxed environment
1278 for a unit it is thus recommended to combine these settings with
1279 either CapabilityBoundingSet=~CAP_SYS_ADMIN or
1280 SystemCallFilter=~@mount.
1281 Simple allow-list example using these directives:
1283 [Service]
1285 ReadOnlyPaths=/
1286 ReadWritePaths=/var /run
1287 InaccessiblePaths=-/lost+found
1288 NoExecPaths=/
1289 ExecPaths=/usr/sbin/my_daemon /usr/lib /usr/lib64
1290 These options are only available for system services and are not
1292 supported for services running in per-user instances of the service
1293 manager.
1294 TemporaryFileSystem=
1296 Takes a space-separated list of mount points for temporary file
1297 systems (tmpfs). If set, a new file system namespace is set up for
1298 executed processes, and a temporary file system is mounted on each
1299 mount point. This option may be specified more than once, in which
1300 case temporary file systems are mounted on all listed mount points.
1301 If the empty string is assigned to this option, the list is reset,
1302 and all prior assignments have no effect. Each mount point may
1303 optionally be suffixed with a colon (":") and mount options such as
1304 "size=10%" or "ro". By default, each temporary file system is
1305 mounted with "nodev,strictatime,mode=0755". These can be disabled
1306 by explicitly specifying the corresponding mount options, e.g.,
1307 "dev" or "nostrictatime".
1308 This is useful to hide files or directories not relevant to the
1310 processes invoked by the unit, while necessary files or directories
1311 can be still accessed by combining with BindPaths= or
1312 BindReadOnlyPaths=:
1313 Example: if a unit has the following,
1315 TemporaryFileSystem=/var:ro
1317 BindReadOnlyPaths=/var/lib/systemd
1318 then the invoked processes by the unit cannot see any files or
1320 directories under /var/ except for /var/lib/systemd or its
1321 contents.
1322 This option is only available for system services and is not
1324 supported for services running in per-user instances of the service
1325 manager.
1326 PrivateTmp=
1328 Takes a boolean argument. If true, sets up a new file system
1329 namespace for the executed processes and mounts private /tmp/ and
1330 /var/tmp/ directories inside it that are not shared by processes
1331 outside of the namespace. This is useful to secure access to
1332 temporary files of the process, but makes sharing between processes
1333 via /tmp/ or /var/tmp/ impossible. If true, all temporary files
1334 created by a service in these directories will be removed after the
1335 service is stopped. Defaults to false. It is possible to run two or
1336 more units within the same private /tmp/ and /var/tmp/ namespace by
1337 using the JoinsNamespaceOf= directive, see systemd.unit(5) for
1338 details. This setting is implied if DynamicUser= is set. For this
1339 setting the same restrictions regarding mount propagation and
1340 privileges apply as for ReadOnlyPaths= and related calls, see
1341 above. Enabling this setting has the side effect of adding
1342 Requires= and After= dependencies on all mount units necessary to
1343 access /tmp/ and /var/tmp/. Moreover an implicitly After= ordering
1344 on systemd-tmpfiles-setup.service(8) is added.
1345 Note that the implementation of this setting might be impossible
1347 (for example if mount namespaces are not available), and the unit
1348 should be written in a way that does not solely rely on this
1349 setting for security.
1350 This option is only available for system services and is not
1352 supported for services running in per-user instances of the service
1353 manager.
1354 PrivateDevices=
1356 Takes a boolean argument. If true, sets up a new /dev/ mount for
1357 the executed processes and only adds API pseudo devices such as
1358 /dev/null, /dev/zero or /dev/random (as well as the pseudo TTY
1359 subsystem) to it, but no physical devices such as /dev/sda, system
1360 memory /dev/mem, system ports /dev/port and others. This is useful
1361 to turn off physical device access by the executed process.
1362 Defaults to false.
1363 Enabling this option will install a system call filter to block
1365 low-level I/O system calls that are grouped in the @raw-io set,
1366 remove CAP_MKNOD and CAP_SYS_RAWIO from the capability bounding set
1367 for the unit, and set DevicePolicy=closed (see systemd.resource-
1368 control(5) for details). Note that using this setting will
1369 disconnect propagation of mounts from the service to the host
1370 (propagation in the opposite direction continues to work). This
1371 means that this setting may not be used for services which shall be
1372 able to install mount points in the main mount namespace. The new
1373 /dev/ will be mounted read-only and 'noexec'. The latter may break
1374 old programs which try to set up executable memory by using mmap(2)
1375 of /dev/zero instead of using MAP_ANON. For this setting the same
1376 restrictions regarding mount propagation and privileges apply as
1377 for ReadOnlyPaths= and related calls, see above. If turned on and
1378 if running in user mode, or in system mode, but without the
1379 CAP_SYS_ADMIN capability (e.g. setting User=), NoNewPrivileges=yes
1380 is implied.
1381 Note that the implementation of this setting might be impossible
1383 (for example if mount namespaces are not available), and the unit
1384 should be written in a way that does not solely rely on this
1385 setting for security.
1386 This option is only available for system services and is not
1388 supported for services running in per-user instances of the service
1389 manager.
1390 When access to some but not all devices must be possible, the
1392 DeviceAllow= setting might be used instead. See systemd.resource-
1393 control(5).
1394 PrivateNetwork=
1396 Takes a boolean argument. If true, sets up a new network namespace
1397 for the executed processes and configures only the loopback network
1398 device "lo" inside it. No other network devices will be available
1399 to the executed process. This is useful to turn off network access
1400 by the executed process. Defaults to false. It is possible to run
1401 two or more units within the same private network namespace by
1402 using the JoinsNamespaceOf= directive, see systemd.unit(5) for
1403 details. Note that this option will disconnect all socket families
1404 from the host, including AF_NETLINK and AF_UNIX. Effectively, for
1405 AF_NETLINK this means that device configuration events received
1406 from systemd-udevd.service(8) are not delivered to the unit's
1407 processes. And for AF_UNIX this has the effect that AF_UNIX sockets
1408 in the abstract socket namespace of the host will become
1409 unavailable to the unit's processes (however, those located in the
1410 file system will continue to be accessible).
1411 Note that the implementation of this setting might be impossible
1413 (for example if network namespaces are not available), and the unit
1414 should be written in a way that does not solely rely on this
1415 setting for security.
1416 When this option is used on a socket unit any sockets bound on
1418 behalf of this unit will be bound within a private network
1419 namespace. This may be combined with JoinsNamespaceOf= to listen on
1420 sockets inside of network namespaces of other services.
1421 This option is only available for system services and is not
1423 supported for services running in per-user instances of the service
1424 manager.
1425 NetworkNamespacePath=
1427 Takes an absolute file system path refererring to a Linux network
1428 namespace pseudo-file (i.e. a file like /proc/$PID/ns/net or a bind
1429 mount or symlink to one). When set the invoked processes are added
1430 to the network namespace referenced by that path. The path has to
1431 point to a valid namespace file at the moment the processes are
1432 forked off. If this option is used PrivateNetwork= has no effect.
1433 If this option is used together with JoinsNamespaceOf= then it only
1434 has an effect if this unit is started before any of the listed
1435 units that have PrivateNetwork= or NetworkNamespacePath=
1436 configured, as otherwise the network namespace of those units is
1437 reused.
1438 When this option is used on a socket unit any sockets bound on
1440 behalf of this unit will be bound within the specified network
1441 namespace.
1442 This option is only available for system services and is not
1444 supported for services running in per-user instances of the service
1445 manager.
1446 PrivateIPC=
1448 Takes a boolean argument. If true, sets up a new IPC namespace for
1449 the executed processes. Each IPC namespace has its own set of
1450 System V IPC identifiers and its own POSIX message queue file
1451 system. This is useful to avoid name clash of IPC identifiers.
1452 Defaults to false. It is possible to run two or more units within
1453 the same private IPC namespace by using the JoinsNamespaceOf=
1454 directive, see systemd.unit(5) for details.
1455 Note that IPC namespacing does not have an effect on AF_UNIX
1457 sockets, which are the most common form of IPC used on Linux.
1458 Instead, AF_UNIX sockets in the file system are subject to mount
1459 namespacing, and those in the abstract namespace are subject to
1460 network namespacing. IPC namespacing only has an effect on SysV IPC
1461 (which is mostly legacy) as well as POSIX message queues (for which
1462 AF_UNIX/SOCK_SEQPACKET sockets are typically a better replacement).
1463 IPC namespacing also has no effect on POSIX shared memory (which is
1464 subject to mount namespacing) either. See ipc_namespaces(7) for the
1465 details.
1466 Note that the implementation of this setting might be impossible
1468 (for example if IPC namespaces are not available), and the unit
1469 should be written in a way that does not solely rely on this
1470 setting for security.
1471 This option is only available for system services and is not
1473 supported for services running in per-user instances of the service
1474 manager.
1475 IPCNamespacePath=
1477 Takes an absolute file system path refererring to a Linux IPC
1478 namespace pseudo-file (i.e. a file like /proc/$PID/ns/ipc or a bind
1479 mount or symlink to one). When set the invoked processes are added
1480 to the network namespace referenced by that path. The path has to
1481 point to a valid namespace file at the moment the processes are
1482 forked off. If this option is used PrivateIPC= has no effect. If
1483 this option is used together with JoinsNamespaceOf= then it only
1484 has an effect if this unit is started before any of the listed
1485 units that have PrivateIPC= or IPCNamespacePath= configured, as
1486 otherwise the network namespace of those units is reused.
1487 This option is only available for system services and is not
1489 supported for services running in per-user instances of the service
1490 manager.
1491 PrivateUsers=
1493 Takes a boolean argument. If true, sets up a new user namespace for
1494 the executed processes and configures a minimal user and group
1495 mapping, that maps the "root" user and group as well as the unit's
1496 own user and group to themselves and everything else to the
1497 "nobody" user and group. This is useful to securely detach the user
1498 and group databases used by the unit from the rest of the system,
1499 and thus to create an effective sandbox environment. All files,
1500 directories, processes, IPC objects and other resources owned by
1501 users/groups not equaling "root" or the unit's own will stay
1502 visible from within the unit but appear owned by the "nobody" user
1503 and group. If this mode is enabled, all unit processes are run
1504 without privileges in the host user namespace (regardless if the
1505 unit's own user/group is "root" or not). Specifically this means
1506 that the process will have zero process capabilities on the host's
1507 user namespace, but full capabilities within the service's user
1508 namespace. Settings such as CapabilityBoundingSet= will affect only
1509 the latter, and there's no way to acquire additional capabilities
1510 in the host's user namespace. Defaults to off.
1511 When this setting is set up by a per-user instance of the service
1513 manager, the mapping of the "root" user and group to itself is
1514 omitted (unless the user manager is root). Additionally, in the
1515 per-user instance manager case, the user namespace will be set up
1516 before most other namespaces. This means that combining
1517 PrivateUsers=true with other namespaces will enable use of features
1518 not normally supported by the per-user instances of the service
1519 manager.
1520 This setting is particularly useful in conjunction with
1522 RootDirectory=/RootImage=, as the need to synchronize the user and
1523 group databases in the root directory and on the host is reduced,
1524 as the only users and groups who need to be matched are "root",
1525 "nobody" and the unit's own user and group.
1526 Note that the implementation of this setting might be impossible
1528 (for example if user namespaces are not available), and the unit
1529 should be written in a way that does not solely rely on this
1530 setting for security.
1531 ProtectHostname=
1533 Takes a boolean argument. When set, sets up a new UTS namespace for
1534 the executed processes. In addition, changing hostname or
1535 domainname is prevented. Defaults to off.
1536 Note that the implementation of this setting might be impossible
1538 (for example if UTS namespaces are not available), and the unit
1539 should be written in a way that does not solely rely on this
1540 setting for security.
1541 Note that when this option is enabled for a service hostname
1543 changes no longer propagate from the system into the service, it is
1544 hence not suitable for services that need to take notice of system
1545 hostname changes dynamically.
1546 If this setting is on, but the unit doesn't have the CAP_SYS_ADMIN
1548 capability (e.g. services for which User= is set),
1549 NoNewPrivileges=yes is implied.
1550 This option is only available for system services and is not
1552 supported for services running in per-user instances of the service
1553 manager.
1554 ProtectClock=
1556 Takes a boolean argument. If set, writes to the hardware clock or
1557 system clock will be denied. It is recommended to turn this on for
1558 most services that do not need modify the clock. Defaults to off.
1559 Enabling this option removes CAP_SYS_TIME and CAP_WAKE_ALARM from
1560 the capability bounding set for this unit, installs a system call
1561 filter to block calls that can set the clock, and
1562 DeviceAllow=char-rtc r is implied. This ensures /dev/rtc0,
1563 /dev/rtc1, etc. are made read-only to the service. See
1564 systemd.resource-control(5) for the details about DeviceAllow=. If
1565 this setting is on, but the unit doesn't have the CAP_SYS_ADMIN
1566 capability (e.g. services for which User= is set),
1567 NoNewPrivileges=yes is implied.
1568 This option is only available for system services and is not
1570 supported for services running in per-user instances of the service
1571 manager.
1572 ProtectKernelTunables=
1574 Takes a boolean argument. If true, kernel variables accessible
1575 through /proc/sys/, /sys/, /proc/sysrq-trigger,
1576 /proc/latency_stats, /proc/acpi, /proc/timer_stats, /proc/fs and
1577 /proc/irq will be made read-only to all processes of the unit.
1578 Usually, tunable kernel variables should be initialized only at
1579 boot-time, for example with the sysctl.d(5) mechanism. Few services
1580 need to write to these at runtime; it is hence recommended to turn
1581 this on for most services. For this setting the same restrictions
1582 regarding mount propagation and privileges apply as for
1583 ReadOnlyPaths= and related calls, see above. Defaults to off. If
1584 this setting is on, but the unit doesn't have the CAP_SYS_ADMIN
1585 capability (e.g. services for which User= is set),
1586 NoNewPrivileges=yes is implied. Note that this option does not
1587 prevent indirect changes to kernel tunables effected by IPC calls
1588 to other processes. However, InaccessiblePaths= may be used to make
1589 relevant IPC file system objects inaccessible. If
1590 ProtectKernelTunables= is set, MountAPIVFS=yes is implied.
1591 This option is only available for system services and is not
1593 supported for services running in per-user instances of the service
1594 manager.
1595 ProtectKernelModules=
1597 Takes a boolean argument. If true, explicit module loading will be
1598 denied. This allows module load and unload operations to be turned
1599 off on modular kernels. It is recommended to turn this on for most
1600 services that do not need special file systems or extra kernel
1601 modules to work. Defaults to off. Enabling this option removes
1602 CAP_SYS_MODULE from the capability bounding set for the unit, and
1603 installs a system call filter to block module system calls, also
1604 /usr/lib/modules is made inaccessible. For this setting the same
1605 restrictions regarding mount propagation and privileges apply as
1606 for ReadOnlyPaths= and related calls, see above. Note that limited
1607 automatic module loading due to user configuration or kernel
1608 mapping tables might still happen as side effect of requested user
1609 operations, both privileged and unprivileged. To disable module
1610 auto-load feature please see sysctl.d(5) kernel.modules_disabled
1611 mechanism and /proc/sys/kernel/modules_disabled documentation. If
1612 this setting is on, but the unit doesn't have the CAP_SYS_ADMIN
1613 capability (e.g. services for which User= is set),
1614 NoNewPrivileges=yes is implied.
1615 This option is only available for system services and is not
1617 supported for services running in per-user instances of the service
1618 manager.
1619 ProtectKernelLogs=
1621 Takes a boolean argument. If true, access to the kernel log ring
1622 buffer will be denied. It is recommended to turn this on for most
1623 services that do not need to read from or write to the kernel log
1624 ring buffer. Enabling this option removes CAP_SYSLOG from the
1625 capability bounding set for this unit, and installs a system call
1626 filter to block the syslog(2) system call (not to be confused with
1627 the libc API syslog(3) for userspace logging). The kernel exposes
1628 its log buffer to userspace via /dev/kmsg and /proc/kmsg. If
1629 enabled, these are made inaccessible to all the processes in the
1630 unit. If this setting is on, but the unit doesn't have the
1631 CAP_SYS_ADMIN capability (e.g. services for which User= is set),
1632 NoNewPrivileges=yes is implied.
1633 This option is only available for system services and is not
1635 supported for services running in per-user instances of the service
1636 manager.
1637 ProtectControlGroups=
1639 Takes a boolean argument. If true, the Linux Control Groups
1640 (cgroups(7)) hierarchies accessible through /sys/fs/cgroup/ will be
1641 made read-only to all processes of the unit. Except for container
1642 managers no services should require write access to the control
1643 groups hierarchies; it is hence recommended to turn this on for
1644 most services. For this setting the same restrictions regarding
1645 mount propagation and privileges apply as for ReadOnlyPaths= and
1646 related calls, see above. Defaults to off. If ProtectControlGroups=
1647 is set, MountAPIVFS=yes is implied.
1648 This option is only available for system services and is not
1650 supported for services running in per-user instances of the service
1651 manager.
1652 RestrictAddressFamilies=
1654 Restricts the set of socket address families accessible to the
1655 processes of this unit. Takes "none", or a space-separated list of
1656 address family names to allow-list, such as AF_UNIX, AF_INET or
1657 AF_INET6. When "none" is specified, then all address families will
1658 be denied. When prefixed with "~" the listed address families will
1659 be applied as deny list, otherwise as allow list. Note that this
1660 restricts access to the socket(2) system call only. Sockets passed
1661 into the process by other means (for example, by using socket
1662 activation with socket units, see systemd.socket(5)) are
1663 unaffected. Also, sockets created with socketpair() (which creates
1664 connected AF_UNIX sockets only) are unaffected. Note that this
1665 option has no effect on 32-bit x86, s390, s390x, mips, mips-le,
1666 ppc, ppc-le, ppc64, ppc64-le and is ignored (but works correctly on
1667 other ABIs, including x86-64). Note that on systems supporting
1668 multiple ABIs (such as x86/x86-64) it is recommended to turn off
1669 alternative ABIs for services, so that they cannot be used to
1670 circumvent the restrictions of this option. Specifically, it is
1671 recommended to combine this option with
1672 SystemCallArchitectures=native or similar. If running in user mode,
1673 or in system mode, but without the CAP_SYS_ADMIN capability (e.g.
1674 setting User=), NoNewPrivileges=yes is implied. By default, no
1675 restrictions apply, all address families are accessible to
1676 processes. If assigned the empty string, any previous address
1677 family restriction changes are undone. This setting does not affect
1678 commands prefixed with "+".
1679 Use this option to limit exposure of processes to remote access, in
1681 particular via exotic and sensitive network protocols, such as
1682 AF_PACKET. Note that in most cases, the local AF_UNIX address
1683 family should be included in the configured allow list as it is
1684 frequently used for local communication, including for syslog(2)
1685 logging.
1686 RestrictFileSystems=
1688 Restricts the set of filesystems processes of this unit can open
1689 files on. Takes a space-separated list of filesystem names. Any
1690 filesystem listed is made accessible to the unit's processes,
1691 access to filesystem types not listed is prohibited
1692 (allow-listing). If the first character of the list is "~", the
1693 effect is inverted: access to the filesystems listed is prohibited
1694 (deny-listing). If the empty string is assigned, access to
1695 filesystems is not restricted.
1696 If you specify both types of this option (i.e. allow-listing and
1698 deny-listing), the first encountered will take precedence and will
1699 dictate the default action (allow access to the filesystem or deny
1700 it). Then the next occurrences of this option will add or delete
1701 the listed filesystems from the set of the restricted filesystems,
1702 depending on its type and the default action.
1703 Example: if a unit has the following,
1705 RestrictFileSystems=ext4 tmpfs
1707 RestrictFileSystems=ext2 ext4
1708 then access to ext4, tmpfs, and ext2 is allowed and access to other
1710 filesystems is denied.
1711 Example: if a unit has the following,
1713 RestrictFileSystems=ext4 tmpfs
1715 RestrictFileSystems=~ext4
1716 then only access tmpfs is allowed.
1718 Example: if a unit has the following,
1720 RestrictFileSystems=~ext4 tmpfs
1722 RestrictFileSystems=ext4
1723 then only access to tmpfs is denied.
1725 As the number of possible filesystems is large, predefined sets of
1727 filesystems are provided. A set starts with "@" character, followed
1728 by name of the set.
1729 Table 3. Currently predefined filesystem sets
1731 ┌──────────────────┬────────────────────────────┐
1732 │Set │ Description │
1733 ├──────────────────┼────────────────────────────┤
1734 │@basic-api │ Basic filesystem API. │
1735 ├──────────────────┼────────────────────────────┤
1736 │@auxiliary-api │ Auxiliary filesystem API. │
1737 ├──────────────────┼────────────────────────────┤
1738 │@common-block │ Common block device │
1739 │ │ filesystems. │
1740 ├──────────────────┼────────────────────────────┤
1741 │@historical-block │ Historical block device │
1742 │ │ filesystems. │
1743 ├──────────────────┼────────────────────────────┤
1744 │@network │ Well-known network │
1745 │ │ filesystems. │
1746 ├──────────────────┼────────────────────────────┤
1747 │@privileged-api │ Privileged filesystem API. │
1748 ├──────────────────┼────────────────────────────┤
1749 │@temporary │ Temporary filesystems: │
1750 │ │ tmpfs, ramfs. │
1751 ├──────────────────┼────────────────────────────┤
1752 │@known │ All known filesystems │
1753 │ │ defined by the kernel. │
1754 │ │ This list is defined │
1755 │ │ statically in systemd │
1756 │ │ based on a kernel version │
1757 │ │ that was available when │
1758 │ │ this systemd version was │
1759 │ │ released. It will become │
1760 │ │ progressively more │
1761 │ │ out-of-date as the kernel │
1762 │ │ is updated. │
1763 └──────────────────┴────────────────────────────┘
1764 Use systemd-analyze(1)'s filesystems command to retrieve a list of
1765 filesystems defined on the local system.
1766 Note that this setting might not be supported on some systems (for
1768 example if the LSM eBPF hook is not enabled in the underlying
1769 kernel or if not using the unified control group hierarchy). In
1770 that case this setting has no effect.
1771 RestrictNamespaces=
1773 Restricts access to Linux namespace functionality for the processes
1774 of this unit. For details about Linux namespaces, see
1775 namespaces(7). Either takes a boolean argument, or a
1776 space-separated list of namespace type identifiers. If false (the
1777 default), no restrictions on namespace creation and switching are
1778 made. If true, access to any kind of namespacing is prohibited.
1779 Otherwise, a space-separated list of namespace type identifiers
1780 must be specified, consisting of any combination of: cgroup, ipc,
1781 net, mnt, pid, user and uts. Any namespace type listed is made
1782 accessible to the unit's processes, access to namespace types not
1783 listed is prohibited (allow-listing). By prepending the list with a
1784 single tilde character ("~") the effect may be inverted: only the
1785 listed namespace types will be made inaccessible, all unlisted ones
1786 are permitted (deny-listing). If the empty string is assigned, the
1787 default namespace restrictions are applied, which is equivalent to
1788 false. This option may appear more than once, in which case the
1789 namespace types are merged by OR, or by AND if the lines are
1790 prefixed with "~" (see examples below). Internally, this setting
1791 limits access to the unshare(2), clone(2) and setns(2) system
1792 calls, taking the specified flags parameters into account. Note
1793 that — if this option is used — in addition to restricting creation
1794 and switching of the specified types of namespaces (or all of them,
1795 if true) access to the setns() system call with a zero flags
1796 parameter is prohibited. This setting is only supported on x86,
1797 x86-64, mips, mips-le, mips64, mips64-le, mips64-n32,
1798 mips64-le-n32, ppc64, ppc64-le, s390 and s390x, and enforces no
1799 restrictions on other architectures. If running in user mode, or in
1800 system mode, but without the CAP_SYS_ADMIN capability (e.g. setting
1801 User=), NoNewPrivileges=yes is implied.
1802 Example: if a unit has the following,
1804 RestrictNamespaces=cgroup ipc
1806 RestrictNamespaces=cgroup net
1807 then cgroup, ipc, and net are set. If the second line is prefixed
1809 with "~", e.g.,
1810 RestrictNamespaces=cgroup ipc
1812 RestrictNamespaces=~cgroup net
1813 then, only ipc is set.
1815 LockPersonality=
1817 Takes a boolean argument. If set, locks down the personality(2)
1818 system call so that the kernel execution domain may not be changed
1819 from the default or the personality selected with Personality=
1820 directive. This may be useful to improve security, because odd
1821 personality emulations may be poorly tested and source of
1822 vulnerabilities. If running in user mode, or in system mode, but
1823 without the CAP_SYS_ADMIN capability (e.g. setting User=),
1824 NoNewPrivileges=yes is implied.
1825 MemoryDenyWriteExecute=
1827 Takes a boolean argument. If set, attempts to create memory
1828 mappings that are writable and executable at the same time, or to
1829 change existing memory mappings to become executable, or mapping
1830 shared memory segments as executable are prohibited. Specifically,
1831 a system call filter is added that rejects mmap(2) system calls
1832 with both PROT_EXEC and PROT_WRITE set, mprotect(2) or
1833 pkey_mprotect(2) system calls with PROT_EXEC set and shmat(2)
1834 system calls with SHM_EXEC set. Note that this option is
1835 incompatible with programs and libraries that generate program code
1836 dynamically at runtime, including JIT execution engines, executable
1837 stacks, and code "trampoline" feature of various C compilers. This
1838 option improves service security, as it makes harder for software
1839 exploits to change running code dynamically. However, the
1840 protection can be circumvented, if the service can write to a
1841 filesystem, which is not mounted with noexec (such as /dev/shm), or
1842 it can use memfd_create(). This can be prevented by making such
1843 file systems inaccessible to the service (e.g.
1844 InaccessiblePaths=/dev/shm) and installing further system call
1845 filters (SystemCallFilter=~memfd_create). Note that this feature is
1846 fully available on x86-64, and partially on x86. Specifically, the
1847 shmat() protection is not available on x86. Note that on systems
1848 supporting multiple ABIs (such as x86/x86-64) it is recommended to
1849 turn off alternative ABIs for services, so that they cannot be used
1850 to circumvent the restrictions of this option. Specifically, it is
1851 recommended to combine this option with
1852 SystemCallArchitectures=native or similar. If running in user mode,
1853 or in system mode, but without the CAP_SYS_ADMIN capability (e.g.
1854 setting User=), NoNewPrivileges=yes is implied.
1855 RestrictRealtime=
1857 Takes a boolean argument. If set, any attempts to enable realtime
1858 scheduling in a process of the unit are refused. This restricts
1859 access to realtime task scheduling policies such as SCHED_FIFO,
1860 SCHED_RR or SCHED_DEADLINE. See sched(7) for details about these
1861 scheduling policies. If running in user mode, or in system mode,
1862 but without the CAP_SYS_ADMIN capability (e.g. setting User=),
1863 NoNewPrivileges=yes is implied. Realtime scheduling policies may be
1864 used to monopolize CPU time for longer periods of time, and may
1865 hence be used to lock up or otherwise trigger Denial-of-Service
1866 situations on the system. It is hence recommended to restrict
1867 access to realtime scheduling to the few programs that actually
1868 require them. Defaults to off.
1869 RestrictSUIDSGID=
1871 Takes a boolean argument. If set, any attempts to set the
1872 set-user-ID (SUID) or set-group-ID (SGID) bits on files or
1873 directories will be denied (for details on these bits see
1874 inode(7)). If running in user mode, or in system mode, but without
1875 the CAP_SYS_ADMIN capability (e.g. setting User=),
1876 NoNewPrivileges=yes is implied. As the SUID/SGID bits are
1877 mechanisms to elevate privileges, and allows users to acquire the
1878 identity of other users, it is recommended to restrict creation of
1879 SUID/SGID files to the few programs that actually require them.
1880 Note that this restricts marking of any type of file system object
1881 with these bits, including both regular files and directories
1882 (where the SGID is a different meaning than for files, see
1883 documentation). This option is implied if DynamicUser= is enabled.
1884 Defaults to off.
1885 RemoveIPC=
1887 Takes a boolean parameter. If set, all System V and POSIX IPC
1888 objects owned by the user and group the processes of this unit are
1889 run as are removed when the unit is stopped. This setting only has
1890 an effect if at least one of User=, Group= and DynamicUser= are
1891 used. It has no effect on IPC objects owned by the root user.
1892 Specifically, this removes System V semaphores, as well as System V
1893 and POSIX shared memory segments and message queues. If multiple
1894 units use the same user or group the IPC objects are removed when
1895 the last of these units is stopped. This setting is implied if
1896 DynamicUser= is set.
1897 This option is only available for system services and is not
1899 supported for services running in per-user instances of the service
1900 manager.
1901 PrivateMounts=
1903 Takes a boolean parameter. If set, the processes of this unit will
1904 be run in their own private file system (mount) namespace with all
1905 mount propagation from the processes towards the host's main file
1906 system namespace turned off. This means any file system mount
1907 points established or removed by the unit's processes will be
1908 private to them and not be visible to the host. However, file
1909 system mount points established or removed on the host will be
1910 propagated to the unit's processes. See mount_namespaces(7) for
1911 details on file system namespaces. Defaults to off.
1912 When turned on, this executes three operations for each invoked
1914 process: a new CLONE_NEWNS namespace is created, after which all
1915 existing mounts are remounted to MS_SLAVE to disable propagation
1916 from the unit's processes to the host (but leaving propagation in
1917 the opposite direction in effect). Finally, the mounts are
1918 remounted again to the propagation mode configured with
1919 MountFlags=, see below.
1920 File system namespaces are set up individually for each process
1922 forked off by the service manager. Mounts established in the
1923 namespace of the process created by ExecStartPre= will hence be
1924 cleaned up automatically as soon as that process exits and will not
1925 be available to subsequent processes forked off for ExecStart= (and
1926 similar applies to the various other commands configured for
1927 units). Similarly, JoinsNamespaceOf= does not permit sharing kernel
1928 mount namespaces between units, it only enables sharing of the
1929 /tmp/ and /var/tmp/ directories.
1930 Other file system namespace unit settings — PrivateMounts=,
1932 PrivateTmp=, PrivateDevices=, ProtectSystem=, ProtectHome=,
1933 ReadOnlyPaths=, InaccessiblePaths=, ReadWritePaths=, ... — also
1934 enable file system namespacing in a fashion equivalent to this
1935 option. Hence it is primarily useful to explicitly request this
1936 behaviour if none of the other settings are used.
1937 This option is only available for system services and is not
1939 supported for services running in per-user instances of the service
1940 manager.
1941 MountFlags=
1943 Takes a mount propagation setting: shared, slave or private, which
1944 controls whether file system mount points in the file system
1945 namespaces set up for this unit's processes will receive or
1946 propagate mounts and unmounts from other file system namespaces.
1947 See mount(2) for details on mount propagation, and the three
1948 propagation flags in particular.
1949 This setting only controls the final propagation setting in effect
1951 on all mount points of the file system namespace created for each
1952 process of this unit. Other file system namespacing unit settings
1953 (see the discussion in PrivateMounts= above) will implicitly
1954 disable mount and unmount propagation from the unit's processes
1955 towards the host by changing the propagation setting of all mount
1956 points in the unit's file system namespace to slave first. Setting
1957 this option to shared does not reestablish propagation in that
1958 case.
1959 If not set – but file system namespaces are enabled through another
1961 file system namespace unit setting – shared mount propagation is
1962 used, but — as mentioned — as slave is applied first, propagation
1963 from the unit's processes to the host is still turned off.
1964 It is not recommended to use private mount propagation for units,
1966 as this means temporary mounts (such as removable media) of the
1967 host will stay mounted and thus indefinitely busy in forked off
1968 processes, as unmount propagation events won't be received by the
1969 file system namespace of the unit.
1970 Usually, it is best to leave this setting unmodified, and use
1972 higher level file system namespacing options instead, in particular
1973 PrivateMounts=, see above.
1974 This option is only available for system services and is not
1976 supported for services running in per-user instances of the service
1977 manager.
1978
1980 SystemCallFilter=
1981 Takes a space-separated list of system call names. If this setting
1982 is used, all system calls executed by the unit processes except for
1983 the listed ones will result in immediate process termination with
1984 the SIGSYS signal (allow-listing). (See SystemCallErrorNumber=
1985 below for changing the default action). If the first character of
1986 the list is "~", the effect is inverted: only the listed system
1987 calls will result in immediate process termination (deny-listing).
1988 Deny-listed system calls and system call groups may optionally be
1989 suffixed with a colon (":") and "errno" error number (between 0 and
1990 4095) or errno name such as EPERM, EACCES or EUCLEAN (see errno(3)
1991 for a full list). This value will be returned when a deny-listed
1992 system call is triggered, instead of terminating the processes
1993 immediately. Special setting "kill" can be used to explicitly
1994 specify killing. This value takes precedence over the one given in
1995 SystemCallErrorNumber=, see below. If running in user mode, or in
1996 system mode, but without the CAP_SYS_ADMIN capability (e.g. setting
1997 User=), NoNewPrivileges=yes is implied. This feature makes use of
1998 the Secure Computing Mode 2 interfaces of the kernel ('seccomp
1999 filtering') and is useful for enforcing a minimal sandboxing
2000 environment. Note that the execve(), exit(), exit_group(),
2001 getrlimit(), rt_sigreturn(), sigreturn() system calls and the
2002 system calls for querying time and sleeping are implicitly
2003 allow-listed and do not need to be listed explicitly. This option
2004 may be specified more than once, in which case the filter masks are
2005 merged. If the empty string is assigned, the filter is reset, all
2006 prior assignments will have no effect. This does not affect
2007 commands prefixed with "+".
2008 Note that on systems supporting multiple ABIs (such as x86/x86-64)
2010 it is recommended to turn off alternative ABIs for services, so
2011 that they cannot be used to circumvent the restrictions of this
2012 option. Specifically, it is recommended to combine this option with
2013 SystemCallArchitectures=native or similar.
2014 Note that strict system call filters may impact execution and error
2016 handling code paths of the service invocation. Specifically, access
2017 to the execve() system call is required for the execution of the
2018 service binary — if it is blocked service invocation will
2019 necessarily fail. Also, if execution of the service binary fails
2020 for some reason (for example: missing service executable), the
2021 error handling logic might require access to an additional set of
2022 system calls in order to process and log this failure correctly. It
2023 might be necessary to temporarily disable system call filters in
2024 order to simplify debugging of such failures.
2025 If you specify both types of this option (i.e. allow-listing and
2027 deny-listing), the first encountered will take precedence and will
2028 dictate the default action (termination or approval of a system
2029 call). Then the next occurrences of this option will add or delete
2030 the listed system calls from the set of the filtered system calls,
2031 depending of its type and the default action. (For example, if you
2032 have started with an allow list rule for read() and write(), and
2033 right after it add a deny list rule for write(), then write() will
2034 be removed from the set.)
2035 As the number of possible system calls is large, predefined sets of
2037 system calls are provided. A set starts with "@" character,
2038 followed by name of the set.
2039 Table 4. Currently predefined system call sets
2041 ┌────────────────┬────────────────────────────┐
2042 │Set │ Description │
2043 ├────────────────┼────────────────────────────┤
2044 │@aio │ Asynchronous I/O │
2045 │ │ (io_setup(2), │
2046 │ │ io_submit(2), and related │
2047 │ │ calls) │
2048 ├────────────────┼────────────────────────────┤
2049 │@basic-io │ System calls for basic │
2050 │ │ I/O: reading, writing, │
2051 │ │ seeking, file descriptor │
2052 │ │ duplication and closing │
2053 │ │ (read(2), write(2), and │
2054 │ │ related calls) │
2055 ├────────────────┼────────────────────────────┤
2056 │@chown │ Changing file ownership │
2057 │ │ (chown(2), fchownat(2), │
2058 │ │ and related calls) │
2059 ├────────────────┼────────────────────────────┤
2060 │@clock │ System calls for changing │
2061 │ │ the system clock │
2062 │ │ (adjtimex(2), │
2063 │ │ settimeofday(2), and │
2064 │ │ related calls) │
2065 ├────────────────┼────────────────────────────┤
2066 │@cpu-emulation │ System calls for CPU │
2067 │ │ emulation functionality │
2068 │ │ (vm86(2) and related │
2069 │ │ calls) │
2070 ├────────────────┼────────────────────────────┤
2071 │@debug │ Debugging, performance │
2072 │ │ monitoring and tracing │
2073 │ │ functionality (ptrace(2), │
2074 │ │ perf_event_open(2) and │
2075 │ │ related calls) │
2076 ├────────────────┼────────────────────────────┤
2077 │@file-system │ File system operations: │
2078 │ │ opening, creating files │
2079 │ │ and directories for read │
2080 │ │ and write, renaming and │
2081 │ │ removing them, reading │
2082 │ │ file properties, or │
2083 │ │ creating hard and symbolic │
2084 │ │ links │
2085 ├────────────────┼────────────────────────────┤
2086 │@io-event │ Event loop system calls │
2087 │ │ (poll(2), select(2), │
2088 │ │ epoll(7), eventfd(2) and │
2089 │ │ related calls) │
2090 ├────────────────┼────────────────────────────┤
2091 │@ipc │ Pipes, SysV IPC, POSIX │
2092 │ │ Message Queues and other │
2093 │ │ IPC (mq_overview(7), │
2094 │ │ svipc(7)) │
2095 ├────────────────┼────────────────────────────┤
2096 │@keyring │ Kernel keyring access │
2097 │ │ (keyctl(2) and related │
2098 │ │ calls) │
2099 ├────────────────┼────────────────────────────┤
2100 │@memlock │ Locking of memory in RAM │
2101 │ │ (mlock(2), mlockall(2) and │
2102 │ │ related calls) │
2103 ├────────────────┼────────────────────────────┤
2104 │@module │ Loading and unloading of │
2105 │ │ kernel modules │
2106 │ │ (init_module(2), │
2107 │ │ delete_module(2) and │
2108 │ │ related calls) │
2109 ├────────────────┼────────────────────────────┤
2110 │@mount │ Mounting and unmounting of │
2111 │ │ file systems (mount(2), │
2112 │ │ chroot(2), and related │
2113 │ │ calls) │
2114 ├────────────────┼────────────────────────────┤
2115 │@network-io │ Socket I/O (including │
2116 │ │ local AF_UNIX): socket(7), │
2117 │ │ unix(7) │
2118 ├────────────────┼────────────────────────────┤
2119 │@obsolete │ Unusual, obsolete or │
2120 │ │ unimplemented │
2121 │ │ (create_module(2), │
2122 │ │ gtty(2), ...) │
2123 ├────────────────┼────────────────────────────┤
2124 │@privileged │ All system calls which │
2125 │ │ need super-user │
2126 │ │ capabilities │
2127 │ │ (capabilities(7)) │
2128 ├────────────────┼────────────────────────────┤
2129 │@process │ Process control, │
2130 │ │ execution, namespacing │
2131 │ │ operations (clone(2), │
2132 │ │ kill(2), namespaces(7), │
2133 │ │ ...) │
2134 ├────────────────┼────────────────────────────┤
2135 │@raw-io │ Raw I/O port access │
2136 │ │ (ioperm(2), iopl(2), │
2137 │ │ pciconfig_read(), ...) │
2138 ├────────────────┼────────────────────────────┤
2139 │@reboot │ System calls for rebooting │
2140 │ │ and reboot preparation │
2141 │ │ (reboot(2), kexec(), ...) │
2142 ├────────────────┼────────────────────────────┤
2143 │@resources │ System calls for changing │
2144 │ │ resource limits, memory │
2145 │ │ and scheduling parameters │
2146 │ │ (setrlimit(2), │
2147 │ │ setpriority(2), ...) │
2148 ├────────────────┼────────────────────────────┤
2149 │@setuid │ System calls for changing │
2150 │ │ user ID and group ID │
2151 │ │ credentials, (setuid(2), │
2152 │ │ setgid(2), setresuid(2), │
2153 │ │ ...) │
2154 ├────────────────┼────────────────────────────┤
2155 │@signal │ System calls for │
2156 │ │ manipulating and handling │
2157 │ │ process signals │
2158 │ │ (signal(2), │
2159 │ │ sigprocmask(2), ...) │
2160 ├────────────────┼────────────────────────────┤
2161 │@swap │ System calls for │
2162 │ │ enabling/disabling swap │
2163 │ │ devices (swapon(2), │
2164 │ │ swapoff(2)) │
2165 ├────────────────┼────────────────────────────┤
2166 │@sync │ Synchronizing files and │
2167 │ │ memory to disk (fsync(2), │
2168 │ │ msync(2), and related │
2169 │ │ calls) │
2170 ├────────────────┼────────────────────────────┤
2171 │@system-service │ A reasonable set of system │
2172 │ │ calls used by common │
2173 │ │ system services, excluding │
2174 │ │ any special purpose calls. │
2175 │ │ This is the recommended │
2176 │ │ starting point for │
2177 │ │ allow-listing system calls │
2178 │ │ for system services, as it │
2179 │ │ contains what is typically │
2180 │ │ needed by system services, │
2181 │ │ but excludes overly │
2182 │ │ specific interfaces. For │
2183 │ │ example, the following │
2184 │ │ APIs are excluded: │
2185 │ │ "@clock", "@mount", │
2186 │ │ "@swap", "@reboot". │
2187 ├────────────────┼────────────────────────────┤
2188 │@timer │ System calls for │
2189 │ │ scheduling operations by │
2190 │ │ time (alarm(2), │
2191 │ │ timer_create(2), ...) │
2192 ├────────────────┼────────────────────────────┤
2193 │@known │ All system calls defined │
2194 │ │ by the kernel. This list │
2195 │ │ is defined statically in │
2196 │ │ systemd based on a kernel │
2197 │ │ version that was available │
2198 │ │ when this systemd version │
2199 │ │ was released. It will │
2200 │ │ become progressively more │
2201 │ │ out-of-date as the kernel │
2202 │ │ is updated. │
2203 └────────────────┴────────────────────────────┘
2204 Note, that as new system calls are added to the kernel, additional
2205 system calls might be added to the groups above. Contents of the
2206 sets may also change between systemd versions. In addition, the
2207 list of system calls depends on the kernel version and architecture
2208 for which systemd was compiled. Use systemd-analyze syscall-filter
2209 to list the actual list of system calls in each filter.
2210 Generally, allow-listing system calls (rather than deny-listing) is
2212 the safer mode of operation. It is recommended to enforce system
2213 call allow lists for all long-running system services.
2214 Specifically, the following lines are a relatively safe basic
2215 choice for the majority of system services:
2216 [Service]
2218 SystemCallFilter=@system-service
2219 SystemCallErrorNumber=EPERM
2220 Note that various kernel system calls are defined redundantly:
2222 there are multiple system calls for executing the same operation.
2223 For example, the pidfd_send_signal() system call may be used to
2224 execute operations similar to what can be done with the older
2225 kill() system call, hence blocking the latter without the former
2226 only provides weak protection. Since new system calls are added
2227 regularly to the kernel as development progresses, keeping system
2228 call deny lists comprehensive requires constant work. It is thus
2229 recommended to use allow-listing instead, which offers the benefit
2230 that new system calls are by default implicitly blocked until the
2231 allow list is updated.
2232 Also note that a number of system calls are required to be
2234 accessible for the dynamic linker to work. The dynamic linker is
2235 required for running most regular programs (specifically: all
2236 dynamic ELF binaries, which is how most distributions build
2237 packaged programs). This means that blocking these system calls
2238 (which include open(), openat() or mmap()) will make most programs
2239 typically shipped with generic distributions unusable.
2240 It is recommended to combine the file system namespacing related
2242 options with SystemCallFilter=~@mount, in order to prohibit the
2243 unit's processes to undo the mappings. Specifically these are the
2244 options PrivateTmp=, PrivateDevices=, ProtectSystem=, ProtectHome=,
2245 ProtectKernelTunables=, ProtectControlGroups=, ProtectKernelLogs=,
2246 ProtectClock=, ReadOnlyPaths=, InaccessiblePaths= and
2247 ReadWritePaths=.
2248 SystemCallErrorNumber=
2250 Takes an "errno" error number (between 1 and 4095) or errno name
2251 such as EPERM, EACCES or EUCLEAN, to return when the system call
2252 filter configured with SystemCallFilter= is triggered, instead of
2253 terminating the process immediately. See errno(3) for a full list
2254 of error codes. When this setting is not used, or when the empty
2255 string or the special setting "kill" is assigned, the process will
2256 be terminated immediately when the filter is triggered.
2257 SystemCallArchitectures=
2259 Takes a space-separated list of architecture identifiers to include
2260 in the system call filter. The known architecture identifiers are
2261 the same as for ConditionArchitecture= described in
2262 systemd.unit(5), as well as x32, mips64-n32, mips64-le-n32, and the
2263 special identifier native. The special identifier native implicitly
2264 maps to the native architecture of the system (or more precisely:
2265 to the architecture the system manager is compiled for). If running
2266 in user mode, or in system mode, but without the CAP_SYS_ADMIN
2267 capability (e.g. setting User=), NoNewPrivileges=yes is implied. By
2268 default, this option is set to the empty list, i.e. no filtering is
2269 applied.
2270 If this setting is used, processes of this unit will only be
2272 permitted to call native system calls, and system calls of the
2273 specified architectures. For the purposes of this option, the x32
2274 architecture is treated as including x86-64 system calls. However,
2275 this setting still fulfills its purpose, as explained below, on
2276 x32.
2277 System call filtering is not equally effective on all
2279 architectures. For example, on x86 filtering of network
2280 socket-related calls is not possible, due to ABI limitations — a
2281 limitation that x86-64 does not have, however. On systems
2282 supporting multiple ABIs at the same time — such as x86/x86-64 — it
2283 is hence recommended to limit the set of permitted system call
2284 architectures so that secondary ABIs may not be used to circumvent
2285 the restrictions applied to the native ABI of the system. In
2286 particular, setting SystemCallArchitectures=native is a good choice
2287 for disabling non-native ABIs.
2288 System call architectures may also be restricted system-wide via
2290 the SystemCallArchitectures= option in the global configuration.
2291 See systemd-system.conf(5) for details.
2292 SystemCallLog=
2294 Takes a space-separated list of system call names. If this setting
2295 is used, all system calls executed by the unit processes for the
2296 listed ones will be logged. If the first character of the list is
2297 "~", the effect is inverted: all system calls except the listed
2298 system calls will be logged. If running in user mode, or in system
2299 mode, but without the CAP_SYS_ADMIN capability (e.g. setting
2300 User=), NoNewPrivileges=yes is implied. This feature makes use of
2301 the Secure Computing Mode 2 interfaces of the kernel ('seccomp
2302 filtering') and is useful for auditing or setting up a minimal
2303 sandboxing environment. This option may be specified more than
2304 once, in which case the filter masks are merged. If the empty
2305 string is assigned, the filter is reset, all prior assignments will
2306 have no effect. This does not affect commands prefixed with "+".
2307
2309 Environment=
2310 Sets environment variables for executed processes. Each line is
2311 unquoted using the rules described in "Quoting" section in
2312 systemd.syntax(7) and becomes a list of variable assignments. If
2313 you need to assign a value containing spaces or the equals sign to
2314 a variable, put quotes around the whole assignment. Variable
2315 expansion is not performed inside the strings and the "$" character
2316 has no special meaning. Specifier expansion is performed, see the
2317 "Specifiers" section in systemd.unit(5).
2318 This option may be specified more than once, in which case all
2320 listed variables will be set. If the same variable is listed twice,
2321 the later setting will override the earlier setting. If the empty
2322 string is assigned to this option, the list of environment
2323 variables is reset, all prior assignments have no effect.
2324 The names of the variables can contain ASCII letters, digits, and
2326 the underscore character. Variable names cannot be empty or start
2327 with a digit. In variable values, most characters are allowed, but
2328 non-printable characters are currently rejected.
2329 Example:
2331 Environment="VAR1=word1 word2" VAR2=word3 "VAR3=$word 5 6"
2333 gives three variables "VAR1", "VAR2", "VAR3" with the values "word1
2335 word2", "word3", "$word 5 6".
2336 See environ(7) for details about environment variables.
2338 Note that environment variables are not suitable for passing
2340 secrets (such as passwords, key material, ...) to service
2341 processes. Environment variables set for a unit are exposed to
2342 unprivileged clients via D-Bus IPC, and generally not understood as
2343 being data that requires protection. Moreover, environment
2344 variables are propagated down the process tree, including across
2345 security boundaries (such as setuid/setgid executables), and hence
2346 might leak to processes that should not have access to the secret
2347 data. Use LoadCredential=, LoadCredentialEncrypted= or
2348 SetCredentialEncrypted= (see below) to pass data to unit processes
2349 securely.
2350 EnvironmentFile=
2352 Similar to Environment= but reads the environment variables from a
2353 text file. The text file should contain new-line-separated variable
2354 assignments. Empty lines, lines without an "=" separator, or lines
2355 starting with ; or # will be ignored, which may be used for
2356 commenting. A line ending with a backslash will be concatenated
2357 with the following one, allowing multiline variable definitions.
2358 The parser strips leading and trailing whitespace from the values
2359 of assignments, unless you use double quotes (").
2360 C escapes[7] are supported, but not most control characters[8].
2362 "\t" and "\n" can be used to insert tabs and newlines within
2363 EnvironmentFile=.
2364 The argument passed should be an absolute filename or wildcard
2366 expression, optionally prefixed with "-", which indicates that if
2367 the file does not exist, it will not be read and no error or
2368 warning message is logged. This option may be specified more than
2369 once in which case all specified files are read. If the empty
2370 string is assigned to this option, the list of file to read is
2371 reset, all prior assignments have no effect.
2372 The files listed with this directive will be read shortly before
2374 the process is executed (more specifically, after all processes
2375 from a previous unit state terminated. This means you can generate
2376 these files in one unit state, and read it with this option in the
2377 next. The files are read from the file system of the service
2378 manager, before any file system changes like bind mounts take
2379 place).
2380 Settings from these files override settings made with Environment=.
2382 If the same variable is set twice from these files, the files will
2383 be read in the order they are specified and the later setting will
2384 override the earlier setting.
2385 PassEnvironment=
2387 Pass environment variables set for the system service manager to
2388 executed processes. Takes a space-separated list of variable names.
2389 This option may be specified more than once, in which case all
2390 listed variables will be passed. If the empty string is assigned to
2391 this option, the list of environment variables to pass is reset,
2392 all prior assignments have no effect. Variables specified that are
2393 not set for the system manager will not be passed and will be
2394 silently ignored. Note that this option is only relevant for the
2395 system service manager, as system services by default do not
2396 automatically inherit any environment variables set for the service
2397 manager itself. However, in case of the user service manager all
2398 environment variables are passed to the executed processes anyway,
2399 hence this option is without effect for the user service manager.
2400 Variables set for invoked processes due to this setting are subject
2402 to being overridden by those configured with Environment= or
2403 EnvironmentFile=.
2404 C escapes[7] are supported, but not most control characters[8].
2406 "\t" and "\n" can be used to insert tabs and newlines within
2407 EnvironmentFile=.
2408 Example:
2410 PassEnvironment=VAR1 VAR2 VAR3
2412 passes three variables "VAR1", "VAR2", "VAR3" with the values set
2414 for those variables in PID1.
2415 See environ(7) for details about environment variables.
2417 UnsetEnvironment=
2419 Explicitly unset environment variable assignments that would
2420 normally be passed from the service manager to invoked processes of
2421 this unit. Takes a space-separated list of variable names or
2422 variable assignments. This option may be specified more than once,
2423 in which case all listed variables/assignments will be unset. If
2424 the empty string is assigned to this option, the list of
2425 environment variables/assignments to unset is reset. If a variable
2426 assignment is specified (that is: a variable name, followed by "=",
2427 followed by its value), then any environment variable matching this
2428 precise assignment is removed. If a variable name is specified
2429 (that is a variable name without any following "=" or value), then
2430 any assignment matching the variable name, regardless of its value
2431 is removed. Note that the effect of UnsetEnvironment= is applied as
2432 final step when the environment list passed to executed processes
2433 is compiled. That means it may undo assignments from any
2434 configuration source, including assignments made through
2435 Environment= or EnvironmentFile=, inherited from the system
2436 manager's global set of environment variables, inherited via
2437 PassEnvironment=, set by the service manager itself (such as
2438 $NOTIFY_SOCKET and such), or set by a PAM module (in case PAMName=
2439 is used).
2440 See "Environment Variables in Spawned Processes" below for a
2442 description of how those settings combine to form the inherited
2443 environment. See environ(7) for general information about
2444 environment variables.
2445
2447 StandardInput=
2448 Controls where file descriptor 0 (STDIN) of the executed processes
2449 is connected to. Takes one of null, tty, tty-force, tty-fail, data,
2450 file:path, socket or fd:name.
2451 If null is selected, standard input will be connected to /dev/null,
2453 i.e. all read attempts by the process will result in immediate EOF.
2454 If tty is selected, standard input is connected to a TTY (as
2456 configured by TTYPath=, see below) and the executed process becomes
2457 the controlling process of the terminal. If the terminal is already
2458 being controlled by another process, the executed process waits
2459 until the current controlling process releases the terminal.
2460 tty-force is similar to tty, but the executed process is forcefully
2462 and immediately made the controlling process of the terminal,
2463 potentially removing previous controlling processes from the
2464 terminal.
2465 tty-fail is similar to tty, but if the terminal already has a
2467 controlling process start-up of the executed process fails.
2468 The data option may be used to configure arbitrary textual or
2470 binary data to pass via standard input to the executed process. The
2471 data to pass is configured via
2472 StandardInputText=/StandardInputData= (see below). Note that the
2473 actual file descriptor type passed (memory file, regular file, UNIX
2474 pipe, ...) might depend on the kernel and available privileges. In
2475 any case, the file descriptor is read-only, and when read returns
2476 the specified data followed by EOF.
2477 The file:path option may be used to connect a specific file system
2479 object to standard input. An absolute path following the ":"
2480 character is expected, which may refer to a regular file, a FIFO or
2481 special file. If an AF_UNIX socket in the file system is specified,
2482 a stream socket is connected to it. The latter is useful for
2483 connecting standard input of processes to arbitrary system
2484 services.
2485 The socket option is valid in socket-activated services only, and
2487 requires the relevant socket unit file (see systemd.socket(5) for
2488 details) to have Accept=yes set, or to specify a single socket
2489 only. If this option is set, standard input will be connected to
2490 the socket the service was activated from, which is primarily
2491 useful for compatibility with daemons designed for use with the
2492 traditional inetd(8) socket activation daemon.
2493 The fd:name option connects standard input to a specific, named
2495 file descriptor provided by a socket unit. The name may be
2496 specified as part of this option, following a ":" character (e.g.
2497 "fd:foobar"). If no name is specified, the name "stdin" is implied
2498 (i.e. "fd" is equivalent to "fd:stdin"). At least one socket unit
2499 defining the specified name must be provided via the Sockets=
2500 option, and the file descriptor name may differ from the name of
2501 its containing socket unit. If multiple matches are found, the
2502 first one will be used. See FileDescriptorName= in
2503 systemd.socket(5) for more details about named file descriptors and
2504 their ordering.
2505 This setting defaults to null, unless
2507 StandardInputText=/StandardInputData= are set, in which case it
2508 defaults to data.
2509 StandardOutput=
2511 Controls where file descriptor 1 (stdout) of the executed processes
2512 is connected to. Takes one of inherit, null, tty, journal, kmsg,
2513 journal+console, kmsg+console, file:path, append:path,
2514 truncate:path, socket or fd:name.
2515 inherit duplicates the file descriptor of standard input for
2517 standard output.
2518 null connects standard output to /dev/null, i.e. everything written
2520 to it will be lost.
2521 tty connects standard output to a tty (as configured via TTYPath=,
2523 see below). If the TTY is used for output only, the executed
2524 process will not become the controlling process of the terminal,
2525 and will not fail or wait for other processes to release the
2526 terminal.
2527 journal connects standard output with the journal, which is
2529 accessible via journalctl(1). Note that everything that is written
2530 to kmsg (see below) is implicitly stored in the journal as well,
2531 the specific option listed below is hence a superset of this one.
2532 (Also note that any external, additional syslog daemons receive
2533 their log data from the journal, too, hence this is the option to
2534 use when logging shall be processed with such a daemon.)
2535 kmsg connects standard output with the kernel log buffer which is
2537 accessible via dmesg(1), in addition to the journal. The journal
2538 daemon might be configured to send all logs to kmsg anyway, in
2539 which case this option is no different from journal.
2540 journal+console and kmsg+console work in a similar way as the two
2542 options above but copy the output to the system console as well.
2543 The file:path option may be used to connect a specific file system
2545 object to standard output. The semantics are similar to the same
2546 option of StandardInput=, see above. If path refers to a regular
2547 file on the filesystem, it is opened (created if it doesn't exist
2548 yet) for writing at the beginning of the file, but without
2549 truncating it. If standard input and output are directed to the
2550 same file path, it is opened only once, for reading as well as
2551 writing and duplicated. This is particularly useful when the
2552 specified path refers to an AF_UNIX socket in the file system, as
2553 in that case only a single stream connection is created for both
2554 input and output.
2555 append:path is similar to file:path above, but it opens the file in
2557 append mode.
2558 truncate:path is similar to file:path above, but it truncates the
2560 file when opening it. For units with multiple command lines, e.g.
2561 Type=oneshot services with multiple ExecStart=, or services with
2562 ExecCondition=, ExecStartPre= or ExecStartPost=, the output file is
2563 reopened and therefore re-truncated for each command line. If the
2564 output file is truncated while another process still has the file
2565 open, e.g. by an ExecReload= running concurrently with an
2566 ExecStart=, and the other process continues writing to the file
2567 without adjusting its offset, then the space between the file
2568 pointers of the two processes may be filled with NUL bytes,
2569 producing a sparse file. Thus, truncate:path is typically only
2570 useful for units where only one process runs at a time, such as
2571 services with a single ExecStart= and no ExecStartPost=,
2572 ExecReload=, ExecStop= or similar.
2573 socket connects standard output to a socket acquired via socket
2575 activation. The semantics are similar to the same option of
2576 StandardInput=, see above.
2577 The fd:name option connects standard output to a specific, named
2579 file descriptor provided by a socket unit. A name may be specified
2580 as part of this option, following a ":" character (e.g.
2581 "fd:foobar"). If no name is specified, the name "stdout" is implied
2582 (i.e. "fd" is equivalent to "fd:stdout"). At least one socket unit
2583 defining the specified name must be provided via the Sockets=
2584 option, and the file descriptor name may differ from the name of
2585 its containing socket unit. If multiple matches are found, the
2586 first one will be used. See FileDescriptorName= in
2587 systemd.socket(5) for more details about named descriptors and
2588 their ordering.
2589 If the standard output (or error output, see below) of a unit is
2591 connected to the journal or the kernel log buffer, the unit will
2592 implicitly gain a dependency of type After= on
2593 systemd-journald.socket (also see the "Implicit Dependencies"
2594 section above). Also note that in this case stdout (or stderr, see
2595 below) will be an AF_UNIX stream socket, and not a pipe or FIFO
2596 that can be re-opened. This means when executing shell scripts the
2597 construct echo "hello" > /dev/stderr for writing text to stderr
2598 will not work. To mitigate this use the construct echo "hello" >&2
2599 instead, which is mostly equivalent and avoids this pitfall.
2600 This setting defaults to the value set with DefaultStandardOutput=
2602 in systemd-system.conf(5), which defaults to journal. Note that
2603 setting this parameter might result in additional dependencies to
2604 be added to the unit (see above).
2605 StandardError=
2607 Controls where file descriptor 2 (stderr) of the executed processes
2608 is connected to. The available options are identical to those of
2609 StandardOutput=, with some exceptions: if set to inherit the file
2610 descriptor used for standard output is duplicated for standard
2611 error, while fd:name will use a default file descriptor name of
2612 "stderr".
2613 This setting defaults to the value set with DefaultStandardError=
2615 in systemd-system.conf(5), which defaults to inherit. Note that
2616 setting this parameter might result in additional dependencies to
2617 be added to the unit (see above).
2618 StandardInputText=, StandardInputData=
2620 Configures arbitrary textual or binary data to pass via file
2621 descriptor 0 (STDIN) to the executed processes. These settings have
2622 no effect unless StandardInput= is set to data (which is the
2623 default if StandardInput= is not set otherwise, but
2624 StandardInputText=/StandardInputData= is). Use this option to embed
2625 process input data directly in the unit file.
2626 StandardInputText= accepts arbitrary textual data. C-style escapes
2628 for special characters as well as the usual "%"-specifiers are
2629 resolved. Each time this setting is used the specified text is
2630 appended to the per-unit data buffer, followed by a newline
2631 character (thus every use appends a new line to the end of the
2632 buffer). Note that leading and trailing whitespace of lines
2633 configured with this option is removed. If an empty line is
2634 specified the buffer is cleared (hence, in order to insert an empty
2635 line, add an additional "\n" to the end or beginning of a line).
2636 StandardInputData= accepts arbitrary binary data, encoded in
2638 Base64[9]. No escape sequences or specifiers are resolved. Any
2639 whitespace in the encoded version is ignored during decoding.
2640 Note that StandardInputText= and StandardInputData= operate on the
2642 same data buffer, and may be mixed in order to configure both
2643 binary and textual data for the same input stream. The textual or
2644 binary data is joined strictly in the order the settings appear in
2645 the unit file. Assigning an empty string to either will reset the
2646 data buffer.
2647 Please keep in mind that in order to maintain readability long unit
2649 file settings may be split into multiple lines, by suffixing each
2650 line (except for the last) with a "\" character (see
2651 systemd.unit(5) for details). This is particularly useful for large
2652 data configured with these two options. Example:
2653 ...
2655 StandardInput=data
2656 StandardInputData=SWNrIHNpdHplIGRhIHVuJyBlc3NlIEtsb3BzLAp1ZmYgZWVtYWwga2xvcHAncy4KSWNrIGtpZWtl \
2657 LCBzdGF1bmUsIHd1bmRyZSBtaXIsCnVmZiBlZW1hbCBqZWh0IHNlIHVmZiBkaWUgVMO8ci4KTmFu \
2658 dSwgZGVuayBpY2ssIGljayBkZW5rIG5hbnUhCkpldHogaXNzZSB1ZmYsIGVyc2NodCB3YXIgc2Ug \
2659 enUhCkljayBqZWhlIHJhdXMgdW5kIGJsaWNrZSDigJQKdW5kIHdlciBzdGVodCBkcmF1w59lbj8g \
2660 SWNrZSEK
2661 ...
2662 LogLevelMax=
2664 Configures filtering by log level of log messages generated by this
2665 unit. Takes a syslog log level, one of emerg (lowest log level,
2666 only highest priority messages), alert, crit, err, warning, notice,
2667 info, debug (highest log level, also lowest priority messages). See
2668 syslog(3) for details. By default no filtering is applied (i.e. the
2669 default maximum log level is debug). Use this option to configure
2670 the logging system to drop log messages of a specific service above
2671 the specified level. For example, set LogLevelMax=info in order to
2672 turn off debug logging of a particularly chatty unit. Note that the
2673 configured level is applied to any log messages written by any of
2674 the processes belonging to this unit, as well as any log messages
2675 written by the system manager process (PID 1) in reference to this
2676 unit, sent via any supported logging protocol. The filtering is
2677 applied early in the logging pipeline, before any kind of further
2678 processing is done. Moreover, messages which pass through this
2679 filter successfully might still be dropped by filters applied at a
2680 later stage in the logging subsystem. For example, MaxLevelStore=
2681 configured in journald.conf(5) might prohibit messages of higher
2682 log levels to be stored on disk, even though the per-unit
2683 LogLevelMax= permitted it to be processed.
2684 LogExtraFields=
2686 Configures additional log metadata fields to include in all log
2687 records generated by processes associated with this unit. This
2688 setting takes one or more journal field assignments in the format
2689 "FIELD=VALUE" separated by whitespace. See systemd.journal-
2690 fields(7) for details on the journal field concept. Even though the
2691 underlying journal implementation permits binary field values, this
2692 setting accepts only valid UTF-8 values. To include space
2693 characters in a journal field value, enclose the assignment in
2694 double quotes ("). The usual specifiers are expanded in all
2695 assignments (see below). Note that this setting is not only useful
2696 for attaching additional metadata to log records of a unit, but
2697 given that all fields and values are indexed may also be used to
2698 implement cross-unit log record matching. Assign an empty string to
2699 reset the list.
2700 LogRateLimitIntervalSec=, LogRateLimitBurst=
2702 Configures the rate limiting that is applied to messages generated
2703 by this unit. If, in the time interval defined by
2704 LogRateLimitIntervalSec=, more messages than specified in
2705 LogRateLimitBurst= are logged by a service, all further messages
2706 within the interval are dropped until the interval is over. A
2707 message about the number of dropped messages is generated. The time
2708 specification for LogRateLimitIntervalSec= may be specified in the
2709 following units: "s", "min", "h", "ms", "us" (see systemd.time(7)
2710 for details). The default settings are set by RateLimitIntervalSec=
2711 and RateLimitBurst= configured in journald.conf(5).
2712 LogNamespace=
2714 Run the unit's processes in the specified journal namespace.
2715 Expects a short user-defined string identifying the namespace. If
2716 not used the processes of the service are run in the default
2717 journal namespace, i.e. their log stream is collected and processed
2718 by systemd-journald.service. If this option is used any log data
2719 generated by processes of this unit (regardless if via the
2720 syslog(), journal native logging or stdout/stderr logging) is
2721 collected and processed by an instance of the
2722 systemd-journald@.service template unit, which manages the
2723 specified namespace. The log data is stored in a data store
2724 independent from the default log namespace's data store. See
2725 systemd-journald.service(8) for details about journal namespaces.
2726 Internally, journal namespaces are implemented through Linux mount
2728 namespacing and over-mounting the directory that contains the
2729 relevant AF_UNIX sockets used for logging in the unit's mount
2730 namespace. Since mount namespaces are used this setting disconnects
2731 propagation of mounts from the unit's processes to the host,
2732 similar to how ReadOnlyPaths= and similar settings (see above)
2733 work. Journal namespaces may hence not be used for services that
2734 need to establish mount points on the host.
2735 When this option is used the unit will automatically gain ordering
2737 and requirement dependencies on the two socket units associated
2738 with the systemd-journald@.service instance so that they are
2739 automatically established prior to the unit starting up. Note that
2740 when this option is used log output of this service does not appear
2741 in the regular journalctl(1) output, unless the --namespace= option
2742 is used.
2743 This option is only available for system services and is not
2745 supported for services running in per-user instances of the service
2746 manager.
2747 SyslogIdentifier=
2749 Sets the process name ("syslog tag") to prefix log lines sent to
2750 the logging system or the kernel log buffer with. If not set,
2751 defaults to the process name of the executed process. This option
2752 is only useful when StandardOutput= or StandardError= are set to
2753 journal or kmsg (or to the same settings in combination with
2754 +console) and only applies to log messages written to stdout or
2755 stderr.
2756 SyslogFacility=
2758 Sets the syslog facility identifier to use when logging. One of
2759 kern, user, mail, daemon, auth, syslog, lpr, news, uucp, cron,
2760 authpriv, ftp, local0, local1, local2, local3, local4, local5,
2761 local6 or local7. See syslog(3) for details. This option is only
2762 useful when StandardOutput= or StandardError= are set to journal or
2763 kmsg (or to the same settings in combination with +console), and
2764 only applies to log messages written to stdout or stderr. Defaults
2765 to daemon.
2766 SyslogLevel=
2768 The default syslog log level to use when logging to the logging
2769 system or the kernel log buffer. One of emerg, alert, crit, err,
2770 warning, notice, info, debug. See syslog(3) for details. This
2771 option is only useful when StandardOutput= or StandardError= are
2772 set to journal or kmsg (or to the same settings in combination with
2773 +console), and only applies to log messages written to stdout or
2774 stderr. Note that individual lines output by executed processes may
2775 be prefixed with a different log level which can be used to
2776 override the default log level specified here. The interpretation
2777 of these prefixes may be disabled with SyslogLevelPrefix=, see
2778 below. For details, see sd-daemon(3). Defaults to info.
2779 SyslogLevelPrefix=
2781 Takes a boolean argument. If true and StandardOutput= or
2782 StandardError= are set to journal or kmsg (or to the same settings
2783 in combination with +console), log lines written by the executed
2784 process that are prefixed with a log level will be processed with
2785 this log level set but the prefix removed. If set to false, the
2786 interpretation of these prefixes is disabled and the logged lines
2787 are passed on as-is. This only applies to log messages written to
2788 stdout or stderr. For details about this prefixing see sd-
2789 daemon(3). Defaults to true.
2790 TTYPath=
2792 Sets the terminal device node to use if standard input, output, or
2793 error are connected to a TTY (see above). Defaults to /dev/console.
2794 TTYReset=
2796 Reset the terminal device specified with TTYPath= before and after
2797 execution. Defaults to "no".
2798 TTYVHangup=
2800 Disconnect all clients which have opened the terminal device
2801 specified with TTYPath= before and after execution. Defaults to
2802 "no".
2803 TTYRows=, TTYColumns=
2805 Configure the size of the TTY specified with TTYPath=. If unset or
2806 set to the empty string, the kernel default is used.
2807 TTYVTDisallocate=
2809 If the terminal device specified with TTYPath= is a virtual console
2810 terminal, try to deallocate the TTY before and after execution.
2811 This ensures that the screen and scrollback buffer is cleared.
2812 Defaults to "no".
2813
2815 LoadCredential=ID[:PATH], LoadCredentialEncrypted=ID[:PATH]
2816 Pass a credential to the unit. Credentials are limited-size binary
2817 or textual objects that may be passed to unit processes. They are
2818 primarily used for passing cryptographic keys (both public and
2819 private) or certificates, user account information or identity
2820 information from host to services. The data is accessible from the
2821 unit's processes via the file system, at a read-only location that
2822 (if possible and permitted) is backed by non-swappable memory. The
2823 data is only accessible to the user associated with the unit, via
2824 the User=/DynamicUser= settings (as well as the superuser). When
2825 available, the location of credentials is exported as the
2826 $CREDENTIALS_DIRECTORY environment variable to the unit's
2827 processes.
2828 The LoadCredential= setting takes a textual ID to use as name for a
2830 credential plus a file system path, separated by a colon. The ID
2831 must be a short ASCII string suitable as filename in the
2832 filesystem, and may be chosen freely by the user. If the specified
2833 path is absolute it is opened as regular file and the credential
2834 data is read from it. If the absolute path refers to an AF_UNIX
2835 stream socket in the file system a connection is made to it (only
2836 once at unit start-up) and the credential data read from the
2837 connection, providing an easy IPC integration point for dynamically
2838 providing credentials from other services. If the specified path is
2839 not absolute and itself qualifies as valid credential identifier it
2840 is understood to refer to a credential that the service manager
2841 itself received via the $CREDENTIALS_DIRECTORY environment
2842 variable, which may be used to propagate credentials from an
2843 invoking environment (e.g. a container manager that invoked the
2844 service manager) into a service. The contents of the file/socket
2845 may be arbitrary binary or textual data, including newline
2846 characters and NUL bytes. If the file system path is omitted it is
2847 chosen identical to the credential name, i.e. this is a terse way
2848 do declare credentials to inherit from the service manager into a
2849 service. This option may be used multiple times, each time defining
2850 an additional credential to pass to the unit.
2851 The LoadCredentialEncrypted= setting is identical to
2853 LoadCredential=, except that the credential data is decrypted
2854 before being passed on to the executed processes. Specifically, the
2855 referenced path should refer to a file or socket with an encrypted
2856 credential, as implemented by systemd-creds(1). This credential is
2857 loaded, decrypted and then passed to the application in decrypted
2858 plaintext form, in the same way a regular credential specified via
2859 LoadCredential= would be. A credential configured this way may
2860 encrypted with a secret key derived from the system's TPM2 security
2861 chip, or with a secret key stored in
2862 /var/lib/systemd/credentials.secret, or with both. Using encrypted
2863 credentials improves security as credentials are not stored in
2864 plaintext and only decrypted into plaintext the moment a service
2865 requiring them is started. Moreover, credentials may be bound to
2866 the local hardware and installations, so that they cannot easily be
2867 analyzed offline.
2868 The credential files/IPC sockets must be accessible to the service
2870 manager, but don't have to be directly accessible to the unit's
2871 processes: the credential data is read and copied into separate,
2872 read-only copies for the unit that are accessible to appropriately
2873 privileged processes. This is particularly useful in combination
2874 with DynamicUser= as this way privileged data can be made available
2875 to processes running under a dynamic UID (i.e. not a previously
2876 known one) without having to open up access to all users.
2877 In order to reference the path a credential may be read from within
2879 a ExecStart= command line use "${CREDENTIALS_DIRECTORY}/mycred",
2880 e.g. "ExecStart=cat ${CREDENTIALS_DIRECTORY}/mycred".
2881 Currently, an accumulated credential size limit of 1 MB per unit is
2883 enforced.
2884 If referencing an AF_UNIX stream socket to connect to, the
2886 connection will originate from an abstract namespace socket, that
2887 includes information about the unit and the credential ID in its
2888 socket name. Use getpeername(2) to query this information. The
2889 returned socket name is formatted as NUL RANDOM "/unit/" UNIT "/"
2890 ID, i.e. a NUL byte (as required for abstract namespace socket
2891 names), followed by a random string (consisting of alphadecimal
2892 characters), followed by the literal string "/unit/", followed by
2893 the requesting unit name, followed by the literal character "/",
2894 followed by the textual credential ID requested. Example:
2895 "\0adf9d86b6eda275e/unit/foobar.service/credx" in case the
2896 credential "credx" is requested for a unit "foobar.service". This
2897 functionality is useful for using a single listening socket to
2898 serve credentials to multiple consumers.
2899 SetCredential=ID:VALUE, SetCredentialEncrypted=ID:VALUE
2901 The SetCredential= setting is similar to LoadCredential= but
2902 accepts a literal value to use as data for the credential, instead
2903 of a file system path to read the data from. Do not use this option
2904 for data that is supposed to be secret, as it is accessible to
2905 unprivileged processes via IPC. It's only safe to use this for user
2906 IDs, public key material and similar non-sensitive data. For
2907 everything else use LoadCredential=. In order to embed binary data
2908 into the credential data use C-style escaping (i.e. "\n" to embed
2909 a newline, or "\x00" to embed a NUL byte).
2910 The SetCredentialEncrypted= setting is identical to SetCredential=
2912 but expects an encrypted credential in literal form as value. This
2913 allows embedding confidential credentials securely directly in unit
2914 files. Use systemd-creds(1)' -p switch to generate suitable
2915 SetCredentialEncrypted= lines directly from plaintext credentials.
2916 For further details see LoadCredentialEncrypted= above.
2917 If a credential of the same ID is listed in both LoadCredential=
2919 and SetCredential=, the latter will act as default if the former
2920 cannot be retrieved. In this case not being able to retrieve the
2921 credential from the path specified in LoadCredential= is not
2922 considered fatal.
2923
2925 UtmpIdentifier=
2926 Takes a four character identifier string for an utmp(5) and wtmp
2927 entry for this service. This should only be set for services such
2928 as getty implementations (such as agetty(8)) where utmp/wtmp
2929 entries must be created and cleared before and after execution, or
2930 for services that shall be executed as if they were run by a getty
2931 process (see below). If the configured string is longer than four
2932 characters, it is truncated and the terminal four characters are
2933 used. This setting interprets %I style string replacements. This
2934 setting is unset by default, i.e. no utmp/wtmp entries are created
2935 or cleaned up for this service.
2936 UtmpMode=
2938 Takes one of "init", "login" or "user". If UtmpIdentifier= is set,
2939 controls which type of utmp(5)/wtmp entries for this service are
2940 generated. This setting has no effect unless UtmpIdentifier= is set
2941 too. If "init" is set, only an INIT_PROCESS entry is generated and
2942 the invoked process must implement a getty-compatible utmp/wtmp
2943 logic. If "login" is set, first an INIT_PROCESS entry, followed by
2944 a LOGIN_PROCESS entry is generated. In this case, the invoked
2945 process must implement a login(1)-compatible utmp/wtmp logic. If
2946 "user" is set, first an INIT_PROCESS entry, then a LOGIN_PROCESS
2947 entry and finally a USER_PROCESS entry is generated. In this case,
2948 the invoked process may be any process that is suitable to be run
2949 as session leader. Defaults to "init".
2950
2952 Processes started by the service manager are executed with an
2953 environment variable block assembled from multiple sources. Processes
2954 started by the system service manager generally do not inherit
2955 environment variables set for the service manager itself (but this may
2956 be altered via PassEnvironment=), but processes started by the user
2957 service manager instances generally do inherit all environment
2958 variables set for the service manager itself.
2959 For each invoked process the list of environment variables set is
2961 compiled from the following sources:
2962 • Variables globally configured for the service manager, using the
2964 DefaultEnvironment= setting in systemd-system.conf(5), the kernel
2965 command line option systemd.setenv= understood by systemd(1), or
2966 via systemctl(1) set-environment verb.
2967 • Variables defined by the service manager itself (see the list
2969 below).
2970 • Variables set in the service manager's own environment variable
2972 block (subject to PassEnvironment= for the system service manager).
2973 • Variables set via Environment= in the unit file.
2975 • Variables read from files specified via EnvironmentFile= in the
2977 unit file.
2978 • Variables set by any PAM modules in case PAMName= is in effect,
2980 cf. pam_env(8).
2981 If the same environment variable is set by multiple of these sources,
2983 the later source — according to the order of the list above — wins.
2984 Note that as the final step all variables listed in UnsetEnvironment=
2985 are removed from the compiled environment variable list, immediately
2986 before it is passed to the executed process.
2987 The general philosophy is to expose a small curated list of environment
2989 variables to processes. Services started by the system manager (PID 1)
2990 will be started, without additional service-specific configuration,
2991 with just a few environment variables. The user manager inherits
2992 environment variables as any other system service, but in addition may
2993 receive additional environment variables from PAM, and, typically,
2994 additional imported variables when the user starts a graphical session.
2995 It is recommended to keep the environment blocks in both the system and
2996 user managers lean. Importing all variables inherited by the graphical
2997 session or by one of the user shells is strongly discouraged.
2998 Hint: systemd-run -P env and systemd-run --user -P env print the
3000 effective system and user service environment blocks.
3001 Environment Variables Set or Propagated by the Service Manager
3003 The following environment variables are propagated by the service
3004 manager or generated internally for each invoked process:
3005 $PATH
3007 Colon-separated list of directories to use when launching
3008 executables. systemd uses a fixed value of
3009 "/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin" in the system
3010 manager. When compiled for systems with "unmerged /usr/" (/bin is
3011 not a symlink to /usr/bin), ":/sbin:/bin" is appended. In case of
3012 the user manager, a different path may be configured by the
3013 distribution. It is recommended to not rely on the order of
3014 entries, and have only one program with a given name in $PATH.
3015 $LANG
3017 Locale. Can be set in locale.conf(5) or on the kernel command line
3018 (see systemd(1) and kernel-command-line(7)).
3019 $USER, $LOGNAME, $HOME, $SHELL
3021 User name (twice), home directory, and the login shell. The
3022 variables are set for the units that have User= set, which includes
3023 user systemd instances. See passwd(5).
3024 $INVOCATION_ID
3026 Contains a randomized, unique 128bit ID identifying each runtime
3027 cycle of the unit, formatted as 32 character hexadecimal string. A
3028 new ID is assigned each time the unit changes from an inactive
3029 state into an activating or active state, and may be used to
3030 identify this specific runtime cycle, in particular in data stored
3031 offline, such as the journal. The same ID is passed to all
3032 processes run as part of the unit.
3033 $XDG_RUNTIME_DIR
3035 The directory to use for runtime objects (such as IPC objects) and
3036 volatile state. Set for all services run by the user systemd
3037 instance, as well as any system services that use PAMName= with a
3038 PAM stack that includes pam_systemd. See below and pam_systemd(8)
3039 for more information.
3040 $RUNTIME_DIRECTORY, $STATE_DIRECTORY, $CACHE_DIRECTORY,
3042 $LOGS_DIRECTORY, $CONFIGURATION_DIRECTORY
3043 Absolute paths to the directories defined with RuntimeDirectory=,
3044 StateDirectory=, CacheDirectory=, LogsDirectory=, and
3045 ConfigurationDirectory= when those settings are used.
3046 $CREDENTIALS_DIRECTORY
3048 An absolute path to the per-unit directory with credentials
3049 configured via LoadCredential=/SetCredential=. The directory is
3050 marked read-only and is placed in unswappable memory (if supported
3051 and permitted), and is only accessible to the UID associated with
3052 the unit via User= or DynamicUser= (and the superuser).
3053 $MAINPID
3055 The PID of the unit's main process if it is known. This is only set
3056 for control processes as invoked by ExecReload= and similar.
3057 $MANAGERPID
3059 The PID of the user systemd instance, set for processes spawned by
3060 it.
3061 $LISTEN_FDS, $LISTEN_PID, $LISTEN_FDNAMES
3063 Information about file descriptors passed to a service for socket
3064 activation. See sd_listen_fds(3).
3065 $NOTIFY_SOCKET
3067 The socket sd_notify() talks to. See sd_notify(3).
3068 $WATCHDOG_PID, $WATCHDOG_USEC
3070 Information about watchdog keep-alive notifications. See
3071 sd_watchdog_enabled(3).
3072 $SYSTEMD_EXEC_PID
3074 The PID of the unit process (e.g. process invoked by ExecStart=).
3075 The child process can use this information to determine whether the
3076 process is directly invoked by the service manager or indirectly as
3077 a child of another process by comparing this value with the current
3078 PID (as similar to the scheme used in sd_listen_fds(3) with
3079 $LISTEN_PID and $LISTEN_FDS).
3080 $TERM
3082 Terminal type, set only for units connected to a terminal
3083 (StandardInput=tty, StandardOutput=tty, or StandardError=tty). See
3084 termcap(5).
3085 $LOG_NAMESPACE
3087 Contains the name of the selected logging namespace when the
3088 LogNamespace= service setting is used.
3089 $JOURNAL_STREAM
3091 If the standard output or standard error output of the executed
3092 processes are connected to the journal (for example, by setting
3093 StandardError=journal) $JOURNAL_STREAM contains the device and
3094 inode numbers of the connection file descriptor, formatted in
3095 decimal, separated by a colon (":"). This permits invoked processes
3096 to safely detect whether their standard output or standard error
3097 output are connected to the journal. The device and inode numbers
3098 of the file descriptors should be compared with the values set in
3099 the environment variable to determine whether the process output is
3100 still connected to the journal. Note that it is generally not
3101 sufficient to only check whether $JOURNAL_STREAM is set at all as
3102 services might invoke external processes replacing their standard
3103 output or standard error output, without unsetting the environment
3104 variable.
3105 If both standard output and standard error of the executed
3107 processes are connected to the journal via a stream socket, this
3108 environment variable will contain information about the standard
3109 error stream, as that's usually the preferred destination for log
3110 data. (Note that typically the same stream is used for both
3111 standard output and standard error, hence very likely the
3112 environment variable contains device and inode information matching
3113 both stream file descriptors.)
3114 This environment variable is primarily useful to allow services to
3116 optionally upgrade their used log protocol to the native journal
3117 protocol (using sd_journal_print(3) and other functions) if their
3118 standard output or standard error output is connected to the
3119 journal anyway, thus enabling delivery of structured metadata along
3120 with logged messages.
3121 $SERVICE_RESULT
3123 Only defined for the service unit type, this environment variable
3124 is passed to all ExecStop= and ExecStopPost= processes, and encodes
3125 the service "result". Currently, the following values are defined:
3126 Table 5. Defined $SERVICE_RESULT values
3128 ┌──────────────────┬────────────────────────────┐
3129 │Value │ Meaning │
3130 ├──────────────────┼────────────────────────────┤
3131 │"success" │ The service ran │
3132 │ │ successfully and exited │
3133 │ │ cleanly. │
3134 ├──────────────────┼────────────────────────────┤
3135 │"protocol" │ A protocol violation │
3136 │ │ occurred: the service did │
3137 │ │ not take the steps │
3138 │ │ required by its unit │
3139 │ │ configuration │
3140 │ │ (specifically what is │
3141 │ │ configured in its Type= │
3142 │ │ setting). │
3143 ├──────────────────┼────────────────────────────┤
3144 │"timeout" │ One of the steps timed │
3145 │ │ out. │
3146 ├──────────────────┼────────────────────────────┤
3147 │"exit-code" │ Service process exited │
3148 │ │ with a non-zero exit code; │
3149 │ │ see $EXIT_CODE below for │
3150 │ │ the actual exit code │
3151 │ │ returned. │
3152 ├──────────────────┼────────────────────────────┤
3153 │"signal" │ A service process was │
3154 │ │ terminated abnormally by a │
3155 │ │ signal, without dumping │
3156 │ │ core. See $EXIT_CODE below │
3157 │ │ for the actual signal │
3158 │ │ causing the termination. │
3159 ├──────────────────┼────────────────────────────┤
3160 │"core-dump" │ A service process │
3161 │ │ terminated abnormally with │
3162 │ │ a signal and dumped core. │
3163 │ │ See $EXIT_CODE below for │
3164 │ │ the signal causing the │
3165 │ │ termination. │
3166 ├──────────────────┼────────────────────────────┤
3167 │"watchdog" │ Watchdog keep-alive ping │
3168 │ │ was enabled for the │
3169 │ │ service, but the deadline │
3170 │ │ was missed. │
3171 ├──────────────────┼────────────────────────────┤
3172 │"start-limit-hit" │ A start limit was defined │
3173 │ │ for the unit and it was │
3174 │ │ hit, causing the unit to │
3175 │ │ fail to start. See │
3176 │ │ systemd.unit(5)'s │
3177 │ │ StartLimitIntervalSec= and │
3178 │ │ StartLimitBurst= for │
3179 │ │ details. │
3180 ├──────────────────┼────────────────────────────┤
3181 │"resources" │ A catch-all condition in │
3182 │ │ case a system operation │
3183 │ │ failed. │
3184 └──────────────────┴────────────────────────────┘
3185 This environment variable is useful to monitor failure or
3186 successful termination of a service. Even though this variable is
3187 available in both ExecStop= and ExecStopPost=, it is usually a
3188 better choice to place monitoring tools in the latter, as the
3189 former is only invoked for services that managed to start up
3190 correctly, and the latter covers both services that failed during
3191 their start-up and those which failed during their runtime.
3192 $EXIT_CODE, $EXIT_STATUS
3194 Only defined for the service unit type, these environment variables
3195 are passed to all ExecStop=, ExecStopPost= processes and contain
3196 exit status/code information of the main process of the service.
3197 For the precise definition of the exit code and status, see
3198 wait(2). $EXIT_CODE is one of "exited", "killed", "dumped".
3199 $EXIT_STATUS contains the numeric exit code formatted as string if
3200 $EXIT_CODE is "exited", and the signal name in all other cases.
3201 Note that these environment variables are only set if the service
3202 manager succeeded to start and identify the main process of the
3203 service.
3204 Table 6. Summary of possible service result variable values
3206 ┌──────────────────┬──────────────────┬─────────────────────┐
3207 │$SERVICE_RESULT │ $EXIT_CODE │ $EXIT_STATUS │
3208 ├──────────────────┼──────────────────┼─────────────────────┤
3209 │"success" │ "killed" │ "HUP", "INT", │
3210 │ │ │ "TERM", "PIPE" │
3211 │ ├──────────────────┼─────────────────────┤
3212 │ │ "exited" │ "0" │
3213 ├──────────────────┼──────────────────┼─────────────────────┤
3214 │"protocol" │ not set │ not set │
3215 │ ├──────────────────┼─────────────────────┤
3216 │ │ "exited" │ "0" │
3217 ├──────────────────┼──────────────────┼─────────────────────┤
3218 │"timeout" │ "killed" │ "TERM", "KILL" │
3219 │ ├──────────────────┼─────────────────────┤
3220 │ │ "exited" │ "0", "1", "2", "3", │
3221 │ │ │ ..., "255" │
3222 ├──────────────────┼──────────────────┼─────────────────────┤
3223 │"exit-code" │ "exited" │ "1", "2", "3", ..., │
3224 │ │ │ "255" │
3225 ├──────────────────┼──────────────────┼─────────────────────┤
3226 │"signal" │ "killed" │ "HUP", "INT", │
3227 │ │ │ "KILL", ... │
3228 ├──────────────────┼──────────────────┼─────────────────────┤
3229 │"core-dump" │ "dumped" │ "ABRT", "SEGV", │
3230 │ │ │ "QUIT", ... │
3231 ├──────────────────┼──────────────────┼─────────────────────┤
3232 │"watchdog" │ "dumped" │ "ABRT" │
3233 │ ├──────────────────┼─────────────────────┤
3234 │ │ "killed" │ "TERM", "KILL" │
3235 │ ├──────────────────┼─────────────────────┤
3236 │ │ "exited" │ "0", "1", "2", "3", │
3237 │ │ │ ..., "255" │
3238 ├──────────────────┼──────────────────┼─────────────────────┤
3239 │"exec-condition" │ "exited" │ "1", "2", "3", "4", │
3240 │ │ │ ..., "254" │
3241 ├──────────────────┼──────────────────┼─────────────────────┤
3242 │"oom-kill" │ "killed" │ "TERM", "KILL" │
3243 ├──────────────────┼──────────────────┼─────────────────────┤
3244 │"start-limit-hit" │ not set │ not set │
3245 ├──────────────────┼──────────────────┼─────────────────────┤
3246 │"resources" │ any of the above │ any of the above │
3247 ├──────────────────┴──────────────────┴─────────────────────┤
3248 │Note: the process may be also terminated by a signal not │
3249 │sent by systemd. In particular the process may send an │
3250 │arbitrary signal to itself in a handler for any of the │
3251 │non-maskable signals. Nevertheless, in the "timeout" and │
3252 │"watchdog" rows above only the signals that systemd sends │
3253 │have been included. Moreover, using SuccessExitStatus= │
3254 │additional exit statuses may be declared to indicate clean │
3255 │termination, which is not reflected by this table. │
3256 └───────────────────────────────────────────────────────────┘
3257 $PIDFILE
3259 The path to the configured PID file, in case the process is forked
3260 off on behalf of a service that uses the PIDFile= setting, see
3261 systemd.service(5) for details. Service code may use this
3262 environment variable to automatically generate a PID file at the
3263 location configured in the unit file. This field is set to an
3264 absolute path in the file system.
3265 For system services, when PAMName= is enabled and pam_systemd is part
3267 of the selected PAM stack, additional environment variables defined by
3268 systemd may be set for services. Specifically, these are $XDG_SEAT,
3269 $XDG_VTNR, see pam_systemd(8) for details.
3270
3272 When invoking a unit process the service manager possibly fails to
3273 apply the execution parameters configured with the settings above. In
3274 that case the already created service process will exit with a non-zero
3275 exit code before the configured command line is executed. (Or in other
3276 words, the child process possibly exits with these error codes, after
3277 having been created by the fork(2) system call, but before the matching
3278 execve(2) system call is called.) Specifically, exit codes defined by
3279 the C library, by the LSB specification and by the systemd service
3280 manager itself are used.
3281 The following basic service exit codes are defined by the C library.
3283 Table 7. Basic C library exit codes
3285 ┌──────────┬───────────────┬────────────────────┐
3286 │Exit Code │ Symbolic Name │ Description │
3287 ├──────────┼───────────────┼────────────────────┤
3288 │0 │ EXIT_SUCCESS │ Generic success │
3289 │ │ │ code. │
3290 ├──────────┼───────────────┼────────────────────┤
3291 │1 │ EXIT_FAILURE │ Generic failure or │
3292 │ │ │ unspecified error. │
3293 └──────────┴───────────────┴────────────────────┘
3294 The following service exit codes are defined by the LSB
3296 specification[10].
3297 Table 8. LSB service exit codes
3299 ┌──────────┬──────────────────────┬────────────────────┐
3300 │Exit Code │ Symbolic Name │ Description │
3301 ├──────────┼──────────────────────┼────────────────────┤
3302 │2 │ EXIT_INVALIDARGUMENT │ Invalid or excess │
3303 │ │ │ arguments. │
3304 ├──────────┼──────────────────────┼────────────────────┤
3305 │3 │ EXIT_NOTIMPLEMENTED │ Unimplemented │
3306 │ │ │ feature. │
3307 ├──────────┼──────────────────────┼────────────────────┤
3308 │4 │ EXIT_NOPERMISSION │ The user has │
3309 │ │ │ insufficient │
3310 │ │ │ privileges. │
3311 ├──────────┼──────────────────────┼────────────────────┤
3312 │5 │ EXIT_NOTINSTALLED │ The program is not │
3313 │ │ │ installed. │
3314 ├──────────┼──────────────────────┼────────────────────┤
3315 │6 │ EXIT_NOTCONFIGURED │ The program is not │
3316 │ │ │ configured. │
3317 ├──────────┼──────────────────────┼────────────────────┤
3318 │7 │ EXIT_NOTRUNNING │ The program is not │
3319 │ │ │ running. │
3320 └──────────┴──────────────────────┴────────────────────┘
3321 The LSB specification suggests that error codes 200 and above are
3323 reserved for implementations. Some of them are used by the service
3324 manager to indicate problems during process invocation:
3325 Table 9. systemd-specific exit codes
3327 ┌──────────┬──────────────────────────────┬─────────────────────────────────────────────┐
3328 │Exit Code │ Symbolic Name │ Description │
3329 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3330 │200 │ EXIT_CHDIR │ Changing to the │
3331 │ │ │ requested working │
3332 │ │ │ directory failed. │
3333 │ │ │ See │
3334 │ │ │ WorkingDirectory= │
3335 │ │ │ above. │
3336 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3337 │201 │ EXIT_NICE │ Failed to set up │
3338 │ │ │ process scheduling │
3339 │ │ │ priority (nice │
3340 │ │ │ level). See Nice= │
3341 │ │ │ above. │
3342 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3343 │202 │ EXIT_FDS │ Failed to close │
3344 │ │ │ unwanted file │
3345 │ │ │ descriptors, or to │
3346 │ │ │ adjust passed file │
3347 │ │ │ descriptors. │
3348 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3349 │203 │ EXIT_EXEC │ The actual process │
3350 │ │ │ execution failed │
3351 │ │ │ (specifically, the │
3352 │ │ │ execve(2) system │
3353 │ │ │ call). Most likely │
3354 │ │ │ this is caused by a │
3355 │ │ │ missing or │
3356 │ │ │ non-accessible │
3357 │ │ │ executable file. │
3358 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3359 │204 │ EXIT_MEMORY │ Failed to perform │
3360 │ │ │ an action due to │
3361 │ │ │ memory shortage. │
3362 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3363 │205 │ EXIT_LIMITS │ Failed to adjust │
3364 │ │ │ resource limits. │
3365 │ │ │ See LimitCPU= and │
3366 │ │ │ related settings │
3367 │ │ │ above. │
3368 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3369 │206 │ EXIT_OOM_ADJUST │ Failed to adjust │
3370 │ │ │ the OOM setting. │
3371 │ │ │ See OOMScoreAdjust= │
3372 │ │ │ above. │
3373 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3374 │207 │ EXIT_SIGNAL_MASK │ Failed to set │
3375 │ │ │ process signal │
3376 │ │ │ mask. │
3377 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3378 │208 │ EXIT_STDIN │ Failed to set up │
3379 │ │ │ standard input. See │
3380 │ │ │ StandardInput= │
3381 │ │ │ above. │
3382 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3383 │209 │ EXIT_STDOUT │ Failed to set up │
3384 │ │ │ standard output. │
3385 │ │ │ See StandardOutput= │
3386 │ │ │ above. │
3387 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3388 │210 │ EXIT_CHROOT │ Failed to change │
3389 │ │ │ root directory │
3390 │ │ │ (chroot(2)). See │
3391 │ │ │ RootDirectory=/RootImage= │
3392 │ │ │ above. │
3393 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3394 │211 │ EXIT_IOPRIO │ Failed to set up IO │
3395 │ │ │ scheduling priority. See │
3396 │ │ │ IOSchedulingClass=/IOSchedulingPriority= │
3397 │ │ │ above. │
3398 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3399 │212 │ EXIT_TIMERSLACK │ Failed to set up timer slack. See │
3400 │ │ │ TimerSlackNSec= above. │
3401 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3402 │213 │ EXIT_SECUREBITS │ Failed to set process secure bits. See │
3403 │ │ │ SecureBits= above. │
3404 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3405 │214 │ EXIT_SETSCHEDULER │ Failed to set up CPU scheduling. See │
3406 │ │ │ CPUSchedulingPolicy=/CPUSchedulingPriority= │
3407 │ │ │ above. │
3408 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3409 │215 │ EXIT_CPUAFFINITY │ Failed to set up CPU affinity. See │
3410 │ │ │ CPUAffinity= above. │
3411 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3412 │216 │ EXIT_GROUP │ Failed to determine or change group │
3413 │ │ │ credentials. See │
3414 │ │ │ Group=/SupplementaryGroups= above. │
3415 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3416 │217 │ EXIT_USER │ Failed to determine or change user │
3417 │ │ │ credentials, or to set up user namespacing. │
3418 │ │ │ See User=/PrivateUsers= above. │
3419 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3420 │218 │ EXIT_CAPABILITIES │ Failed to drop capabilities, or apply │
3421 │ │ │ ambient capabilities. See │
3422 │ │ │ CapabilityBoundingSet=/AmbientCapabilities= │
3423 │ │ │ above. │
3424 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3425 │219 │ EXIT_CGROUP │ Setting up the service control group │
3426 │ │ │ failed. │
3427 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3428 │220 │ EXIT_SETSID │ Failed to create new process session. │
3429 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3430 │221 │ EXIT_CONFIRM │ Execution has been cancelled by the user. │
3431 │ │ │ See the systemd.confirm_spawn= kernel │
3432 │ │ │ command line setting on kernel-command- │
3433 │ │ │ line(7) for details. │
3434 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3435 │222 │ EXIT_STDERR │ Failed to set up standard error output. See │
3436 │ │ │ StandardError= above. │
3437 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3438 │224 │ EXIT_PAM │ Failed to set up PAM session. See PAMName= │
3439 │ │ │ above. │
3440 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3441 │225 │ EXIT_NETWORK │ Failed to set up network namespacing. See │
3442 │ │ │ PrivateNetwork= above. │
3443 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3444 │226 │ EXIT_NAMESPACE │ Failed to set up mount, UTS, or IPC │
3445 │ │ │ namespacing. See ReadOnlyPaths=, │
3446 │ │ │ ProtectHostname=, PrivateIPC=, and related │
3447 │ │ │ settings above. │
3448 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3449 │227 │ EXIT_NO_NEW_PRIVILEGES │ Failed to disable new privileges. See │
3450 │ │ │ NoNewPrivileges=yes above. │
3451 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3452 │228 │ EXIT_SECCOMP │ Failed to apply system call filters. See │
3453 │ │ │ SystemCallFilter= and related settings │
3454 │ │ │ above. │
3455 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3456 │229 │ EXIT_SELINUX_CONTEXT │ Determining or changing SELinux context │
3457 │ │ │ failed. See SELinuxContext= above. │
3458 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3459 │230 │ EXIT_PERSONALITY │ Failed to set up an execution domain │
3460 │ │ │ (personality). See Personality= above. │
3461 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3462 │231 │ EXIT_APPARMOR_PROFILE │ Failed to prepare changing AppArmor │
3463 │ │ │ profile. See AppArmorProfile= above. │
3464 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3465 │232 │ EXIT_ADDRESS_FAMILIES │ Failed to restrict address families. See │
3466 │ │ │ RestrictAddressFamilies= above. │
3467 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3468 │233 │ EXIT_RUNTIME_DIRECTORY │ Setting up runtime directory failed. See │
3469 │ │ │ RuntimeDirectory= and related settings │
3470 │ │ │ above. │
3471 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3472 │235 │ EXIT_CHOWN │ Failed to adjust socket ownership. Used for │
3473 │ │ │ socket units only. │
3474 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3475 │236 │ EXIT_SMACK_PROCESS_LABEL │ Failed to set SMACK label. See │
3476 │ │ │ SmackProcessLabel= above. │
3477 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3478 │237 │ EXIT_KEYRING │ Failed to set up kernel keyring. │
3479 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3480 │238 │ EXIT_STATE_DIRECTORY │ Failed to set up unit's state directory. │
3481 │ │ │ See StateDirectory= above. │
3482 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3483 │239 │ EXIT_CACHE_DIRECTORY │ Failed to set up unit's cache directory. │
3484 │ │ │ See CacheDirectory= above. │
3485 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3486 │240 │ EXIT_LOGS_DIRECTORY │ Failed to set up unit's logging directory. │
3487 │ │ │ See LogsDirectory= above. │
3488 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3489 │241 │ EXIT_CONFIGURATION_DIRECTORY │ Failed to set up unit's configuration │
3490 │ │ │ directory. See ConfigurationDirectory= │
3491 │ │ │ above. │
3492 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3493 │242 │ EXIT_NUMA_POLICY │ Failed to set up unit's NUMA memory policy. │
3494 │ │ │ See NUMAPolicy= and NUMAMask= above. │
3495 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3496 │243 │ EXIT_CREDENTIALS │ Failed to set up unit's credentials. See │
3497 │ │ │ LoadCredential= and SetCredential= above. │
3498 ├──────────┼──────────────────────────────┼─────────────────────────────────────────────┤
3499 │245 │ EXIT_BPF │ Failed to apply BPF restrictions. See │
3500 │ │ │ RestrictFileSystems= above. │
3501 └──────────┴──────────────────────────────┴─────────────────────────────────────────────┘
3502 Finally, the BSD operating systems define a set of exit codes,
3504 typically defined on Linux systems too:
3505 Table 10. BSD exit codes
3507 ┌──────────┬────────────────┬─────────────────────┐
3508 │Exit Code │ Symbolic Name │ Description │
3509 ├──────────┼────────────────┼─────────────────────┤
3510 │64 │ EX_USAGE │ Command line usage │
3511 │ │ │ error │
3512 ├──────────┼────────────────┼─────────────────────┤
3513 │65 │ EX_DATAERR │ Data format error │
3514 ├──────────┼────────────────┼─────────────────────┤
3515 │66 │ EX_NOINPUT │ Cannot open input │
3516 ├──────────┼────────────────┼─────────────────────┤
3517 │67 │ EX_NOUSER │ Addressee unknown │
3518 ├──────────┼────────────────┼─────────────────────┤
3519 │68 │ EX_NOHOST │ Host name unknown │
3520 ├──────────┼────────────────┼─────────────────────┤
3521 │69 │ EX_UNAVAILABLE │ Service unavailable │
3522 ├──────────┼────────────────┼─────────────────────┤
3523 │70 │ EX_SOFTWARE │ internal software │
3524 │ │ │ error │
3525 ├──────────┼────────────────┼─────────────────────┤
3526 │71 │ EX_OSERR │ System error (e.g., │
3527 │ │ │ can't fork) │
3528 ├──────────┼────────────────┼─────────────────────┤
3529 │72 │ EX_OSFILE │ Critical OS file │
3530 │ │ │ missing │
3531 ├──────────┼────────────────┼─────────────────────┤
3532 │73 │ EX_CANTCREAT │ Can't create (user) │
3533 │ │ │ output file │
3534 ├──────────┼────────────────┼─────────────────────┤
3535 │74 │ EX_IOERR │ Input/output error │
3536 ├──────────┼────────────────┼─────────────────────┤
3537 │75 │ EX_TEMPFAIL │ Temporary failure; │
3538 │ │ │ user is invited to │
3539 │ │ │ retry │
3540 ├──────────┼────────────────┼─────────────────────┤
3541 │76 │ EX_PROTOCOL │ Remote error in │
3542 │ │ │ protocol │
3543 ├──────────┼────────────────┼─────────────────────┤
3544 │77 │ EX_NOPERM │ Permission denied │
3545 ├──────────┼────────────────┼─────────────────────┤
3546 │78 │ EX_CONFIG │ Configuration error │
3547 └──────────┴────────────────┴─────────────────────┘
3548
3550 systemd(1), systemctl(1), systemd-analyze(1), journalctl(1), systemd-
3551 system.conf(5), systemd.unit(5), systemd.service(5), systemd.socket(5),
3552 systemd.swap(5), systemd.mount(5), systemd.kill(5), systemd.resource-
3553 control(5), systemd.time(7), systemd.directives(7), tmpfiles.d(5),
3554 exec(3), fork(2)
3555
3557 1. Discoverable Partitions Specification
3558 https://systemd.io/DISCOVERABLE_PARTITIONS
3559 2. The /proc Filesystem
3561 https://www.kernel.org/doc/html/latest/filesystems/proc.html#mount-options
3562 3. User/Group Name Syntax
3564 https://systemd.io/USER_NAMES
3565 4. No New Privileges Flag
3567 https://www.kernel.org/doc/html/latest/userspace-api/no_new_privs.html
3568 5. JSON User Record
3570 https://systemd.io/USER_RECORD
3571 6. proc.txt
3573 https://www.kernel.org/doc/Documentation/filesystems/proc.txt
3574 7. C escapes
3576 https://en.wikipedia.org/wiki/Escape_sequences_in_C#Table_of_escape_sequences
3577 8. most control characters
3579 https://en.wikipedia.org/wiki/Control_character#In_ASCII
3580 9. Base64
3582 https://tools.ietf.org/html/rfc2045#section-6.8
3583 10. LSB specification
3585 https://refspecs.linuxbase.org/LSB_5.0.0/LSB-Core-generic/LSB-Core-generic/iniscrptact.html
3586systemd 250 SYSTEMD.EXEC(5)