1SYSTEMD.RESOURCE-CONTROL(5)systemd.resource-controlSYSTEMD.RESOURCE-CONTROL(5)
2
6 systemd.resource-control - Resource control unit settings
7
9 slice.slice, scope.scope, service.service, socket.socket, mount.mount,
10 swap.swap
11
13 Unit configuration files for services, slices, scopes, sockets, mount
14 points, and swap devices share a subset of configuration options for
15 resource control of spawned processes. Internally, this relies on the
16 Linux Control Groups (cgroups) kernel concept for organizing processes
17 in a hierarchical tree of named groups for the purpose of resource
18 management.
19 This man page lists the configuration options shared by those six unit
21 types. See systemd.unit(5) for the common options of all unit
22 configuration files, and systemd.slice(5), systemd.scope(5),
23 systemd.service(5), systemd.socket(5), systemd.mount(5), and
24 systemd.swap(5) for more information on the specific unit configuration
25 files. The resource control configuration options are configured in the
26 [Slice], [Scope], [Service], [Socket], [Mount], or [Swap] sections,
27 depending on the unit type.
28 In addition, options which control resources available to programs
30 executed by systemd are listed in systemd.exec(5). Those options
31 complement options listed here.
32 See the New Control Group Interfaces[1] for an introduction on how to
34 make use of resource control APIs from programs.
35 Setting resource controls for a group of related units
37 As described in systemd.unit(5), the settings listed here may be set
38 through the main file of a unit and drop-in snippets in *.d/
39 directories. The list of directories searched for drop-ins includes
40 names formed by repeatedly truncating the unit name after all dashes.
41 This is particularly convenient to set resource limits for a group of
42 units with similar names.
43 For example, every user gets their own slice user-nnn.slice. Drop-ins
45 with local configuration that affect user 1000 may be placed in
46 /etc/systemd/system/user-1000.slice,
47 /etc/systemd/system/user-1000.slice.d/*.conf, but also
48 /etc/systemd/system/user-.slice.d/*.conf. This last directory applies
49 to all user slices.
50
52 The following dependencies are implicitly added:
53 • Units with the Slice= setting set automatically acquire Requires=
55 and After= dependencies on the specified slice unit.
56
58 The unified control group hierarchy is the new version of kernel
59 control group interface, see Control Groups v2[2]. Depending on the
60 resource type, there are differences in resource control capabilities.
61 Also, because of interface changes, some resource types have separate
62 set of options on the unified hierarchy.
63 CPU
65 CPUWeight= and StartupCPUWeight= replace CPUShares= and
66 StartupCPUShares=, respectively.
67 The "cpuacct" controller does not exist separately on the unified
69 hierarchy.
70 Memory
72 MemoryMax= replaces MemoryLimit=. MemoryLow= and MemoryHigh= are
73 effective only on unified hierarchy.
74 IO
76 "IO"-prefixed settings are a superset of and replace
77 "BlockIO"-prefixed ones. On unified hierarchy, IO resource control
78 also applies to buffered writes.
79 To ease the transition, there is best-effort translation between the
81 two versions of settings. For each controller, if any of the settings
82 for the unified hierarchy are present, all settings for the legacy
83 hierarchy are ignored. If the resulting settings are for the other type
84 of hierarchy, the configurations are translated before application.
85 Legacy control group hierarchy (see Control Groups version 1[3]), also
87 called cgroup-v1, doesn't allow safe delegation of controllers to
88 unprivileged processes. If the system uses the legacy control group
89 hierarchy, resource control is disabled for the systemd user instance,
90 see systemd(1).
91
93 Units of the types listed above can have settings for resource control
94 configuration:
95 CPUAccounting=
97 Turn on CPU usage accounting for this unit. Takes a boolean
98 argument. Note that turning on CPU accounting for one unit will
99 also implicitly turn it on for all units contained in the same
100 slice and for all its parent slices and the units contained
101 therein. The system default for this setting may be controlled with
102 DefaultCPUAccounting= in systemd-system.conf(5).
103 CPUWeight=weight, StartupCPUWeight=weight
105 Assign the specified CPU time weight to the processes executed, if
106 the unified control group hierarchy is used on the system. These
107 options take an integer value and control the "cpu.weight" control
108 group attribute. The allowed range is 1 to 10000. Defaults to 100.
109 For details about this control group attribute, see Control Groups
110 v2[2] and CFS Scheduler[4]. The available CPU time is split up
111 among all units within one slice relative to their CPU time weight.
112 A higher weight means more CPU time, a lower weight means less.
113 While StartupCPUWeight= applies to the startup and shutdown phases
115 of the system, CPUWeight= applies to normal runtime of the system,
116 and if the former is not set also to the startup and shutdown
117 phases. Using StartupCPUWeight= allows prioritizing specific
118 services at boot-up and shutdown differently than during normal
119 runtime.
120 These settings replace CPUShares= and StartupCPUShares=.
122 CPUQuota=
124 Assign the specified CPU time quota to the processes executed.
125 Takes a percentage value, suffixed with "%". The percentage
126 specifies how much CPU time the unit shall get at maximum, relative
127 to the total CPU time available on one CPU. Use values > 100% for
128 allotting CPU time on more than one CPU. This controls the
129 "cpu.max" attribute on the unified control group hierarchy and
130 "cpu.cfs_quota_us" on legacy. For details about these control group
131 attributes, see Control Groups v2[2] and sched-bwc.txt[5]. Setting
132 CPUQuota= to an empty value unsets the quota.
133 Example: CPUQuota=20% ensures that the executed processes will
135 never get more than 20% CPU time on one CPU.
136 CPUQuotaPeriodSec=
138 Assign the duration over which the CPU time quota specified by
139 CPUQuota= is measured. Takes a time duration value in seconds, with
140 an optional suffix such as "ms" for milliseconds (or "s" for
141 seconds.) The default setting is 100ms. The period is clamped to
142 the range supported by the kernel, which is [1ms, 1000ms].
143 Additionally, the period is adjusted up so that the quota interval
144 is also at least 1ms. Setting CPUQuotaPeriodSec= to an empty value
145 resets it to the default.
146 This controls the second field of "cpu.max" attribute on the
148 unified control group hierarchy and "cpu.cfs_period_us" on legacy.
149 For details about these control group attributes, see Control
150 Groups v2[2] and CFS Scheduler[4].
151 Example: CPUQuotaPeriodSec=10ms to request that the CPU quota is
153 measured in periods of 10ms.
154 AllowedCPUs=, StartupAllowedCPUs=
156 Restrict processes to be executed on specific CPUs. Takes a list of
157 CPU indices or ranges separated by either whitespace or commas. CPU
158 ranges are specified by the lower and upper CPU indices separated
159 by a dash.
160 Setting AllowedCPUs= or StartupAllowedCPUs= doesn't guarantee that
162 all of the CPUs will be used by the processes as it may be limited
163 by parent units. The effective configuration is reported as
164 EffectiveCPUs=.
165 While StartupAllowedCPUs= applies to the startup and shutdown
167 phases of the system, AllowedCPUs= applies to normal runtime of the
168 system, and if the former is not set also to the startup and
169 shutdown phases. Using StartupAllowedCPUs= allows prioritizing
170 specific services at boot-up and shutdown differently than during
171 normal runtime.
172 This setting is supported only with the unified control group
174 hierarchy.
175 AllowedMemoryNodes=, StartupAllowedMemoryNodes=
177 Restrict processes to be executed on specific memory NUMA nodes.
178 Takes a list of memory NUMA nodes indices or ranges separated by
179 either whitespace or commas. Memory NUMA nodes ranges are specified
180 by the lower and upper NUMA nodes indices separated by a dash.
181 Setting AllowedMemoryNodes= or StartupAllowedMemoryNodes= doesn't
183 guarantee that all of the memory NUMA nodes will be used by the
184 processes as it may be limited by parent units. The effective
185 configuration is reported as EffectiveMemoryNodes=.
186 While StartupAllowedMemoryNodes= applies to the startup and
188 shutdown phases of the system, AllowedMemoryNodes= applies to
189 normal runtime of the system, and if the former is not set also to
190 the startup and shutdown phases. Using StartupAllowedMemoryNodes=
191 allows prioritizing specific services at boot-up and shutdown
192 differently than during normal runtime.
193 This setting is supported only with the unified control group
195 hierarchy.
196 MemoryAccounting=
198 Turn on process and kernel memory accounting for this unit. Takes a
199 boolean argument. Note that turning on memory accounting for one
200 unit will also implicitly turn it on for all units contained in the
201 same slice and for all its parent slices and the units contained
202 therein. The system default for this setting may be controlled with
203 DefaultMemoryAccounting= in systemd-system.conf(5).
204 MemoryMin=bytes, MemoryLow=bytes
206 Specify the memory usage protection of the executed processes in
207 this unit. When reclaiming memory, the unit is treated as if it was
208 using less memory resulting in memory to be preferentially
209 reclaimed from unprotected units. Using MemoryLow= results in a
210 weaker protection where memory may still be reclaimed to avoid
211 invoking the OOM killer in case there is no other reclaimable
212 memory.
213 For a protection to be effective, it is generally required to set a
215 corresponding allocation on all ancestors, which is then
216 distributed between children (with the exception of the root
217 slice). Any MemoryMin= or MemoryLow= allocation that is not
218 explicitly distributed to specific children is used to create a
219 shared protection for all children. As this is a shared protection,
220 the children will freely compete for the memory.
221 Takes a memory size in bytes. If the value is suffixed with K, M, G
223 or T, the specified memory size is parsed as Kilobytes, Megabytes,
224 Gigabytes, or Terabytes (with the base 1024), respectively.
225 Alternatively, a percentage value may be specified, which is taken
226 relative to the installed physical memory on the system. If
227 assigned the special value "infinity", all available memory is
228 protected, which may be useful in order to always inherit all of
229 the protection afforded by ancestors. This controls the
230 "memory.min" or "memory.low" control group attribute. For details
231 about this control group attribute, see Memory Interface Files[6].
232 This setting is supported only if the unified control group
234 hierarchy is used and disables MemoryLimit=.
235 Units may have their children use a default "memory.min" or
237 "memory.low" value by specifying DefaultMemoryMin= or
238 DefaultMemoryLow=, which has the same semantics as MemoryMin= and
239 MemoryLow=. This setting does not affect "memory.min" or
240 "memory.low" in the unit itself. Using it to set a default child
241 allocation is only useful on kernels older than 5.7, which do not
242 support the "memory_recursiveprot" cgroup2 mount option.
243 MemoryHigh=bytes
245 Specify the throttling limit on memory usage of the executed
246 processes in this unit. Memory usage may go above the limit if
247 unavoidable, but the processes are heavily slowed down and memory
248 is taken away aggressively in such cases. This is the main
249 mechanism to control memory usage of a unit.
250 Takes a memory size in bytes. If the value is suffixed with K, M, G
252 or T, the specified memory size is parsed as Kilobytes, Megabytes,
253 Gigabytes, or Terabytes (with the base 1024), respectively.
254 Alternatively, a percentage value may be specified, which is taken
255 relative to the installed physical memory on the system. If
256 assigned the special value "infinity", no memory throttling is
257 applied. This controls the "memory.high" control group attribute.
258 For details about this control group attribute, see Memory
259 Interface Files[6].
260 This setting is supported only if the unified control group
262 hierarchy is used and disables MemoryLimit=.
263 MemoryMax=bytes
265 Specify the absolute limit on memory usage of the executed
266 processes in this unit. If memory usage cannot be contained under
267 the limit, out-of-memory killer is invoked inside the unit. It is
268 recommended to use MemoryHigh= as the main control mechanism and
269 use MemoryMax= as the last line of defense.
270 Takes a memory size in bytes. If the value is suffixed with K, M, G
272 or T, the specified memory size is parsed as Kilobytes, Megabytes,
273 Gigabytes, or Terabytes (with the base 1024), respectively.
274 Alternatively, a percentage value may be specified, which is taken
275 relative to the installed physical memory on the system. If
276 assigned the special value "infinity", no memory limit is applied.
277 This controls the "memory.max" control group attribute. For details
278 about this control group attribute, see Memory Interface Files[6].
279 This setting replaces MemoryLimit=.
281 MemorySwapMax=bytes
283 Specify the absolute limit on swap usage of the executed processes
284 in this unit.
285 Takes a swap size in bytes. If the value is suffixed with K, M, G
287 or T, the specified swap size is parsed as Kilobytes, Megabytes,
288 Gigabytes, or Terabytes (with the base 1024), respectively. If
289 assigned the special value "infinity", no swap limit is applied.
290 This controls the "memory.swap.max" control group attribute. For
291 details about this control group attribute, see Memory Interface
292 Files[6].
293 This setting is supported only if the unified control group
295 hierarchy is used and disables MemoryLimit=.
296 TasksAccounting=
298 Turn on task accounting for this unit. Takes a boolean argument. If
299 enabled, the system manager will keep track of the number of tasks
300 in the unit. The number of tasks accounted this way includes both
301 kernel threads and userspace processes, with each thread counting
302 individually. Note that turning on tasks accounting for one unit
303 will also implicitly turn it on for all units contained in the same
304 slice and for all its parent slices and the units contained
305 therein. The system default for this setting may be controlled with
306 DefaultTasksAccounting= in systemd-system.conf(5).
307 TasksMax=N
309 Specify the maximum number of tasks that may be created in the
310 unit. This ensures that the number of tasks accounted for the unit
311 (see above) stays below a specific limit. This either takes an
312 absolute number of tasks or a percentage value that is taken
313 relative to the configured maximum number of tasks on the system.
314 If assigned the special value "infinity", no tasks limit is
315 applied. This controls the "pids.max" control group attribute. For
316 details about this control group attribute, see Process Number
317 Controller[7].
318 The system default for this setting may be controlled with
320 DefaultTasksMax= in systemd-system.conf(5).
321 IOAccounting=
323 Turn on Block I/O accounting for this unit, if the unified control
324 group hierarchy is used on the system. Takes a boolean argument.
325 Note that turning on block I/O accounting for one unit will also
326 implicitly turn it on for all units contained in the same slice and
327 all for its parent slices and the units contained therein. The
328 system default for this setting may be controlled with
329 DefaultIOAccounting= in systemd-system.conf(5).
330 This setting replaces BlockIOAccounting= and disables settings
332 prefixed with BlockIO or StartupBlockIO.
333 IOWeight=weight, StartupIOWeight=weight
335 Set the default overall block I/O weight for the executed
336 processes, if the unified control group hierarchy is used on the
337 system. Takes a single weight value (between 1 and 10000) to set
338 the default block I/O weight. This controls the "io.weight" control
339 group attribute, which defaults to 100. For details about this
340 control group attribute, see IO Interface Files[8]. The available
341 I/O bandwidth is split up among all units within one slice relative
342 to their block I/O weight. A higher weight means more I/O
343 bandwidth, a lower weight means less.
344 While StartupIOWeight= applies to the startup and shutdown phases
346 of the system, IOWeight= applies to the later runtime of the
347 system, and if the former is not set also to the startup and
348 shutdown phases. This allows prioritizing specific services at
349 boot-up and shutdown differently than during runtime.
350 These settings replace BlockIOWeight= and StartupBlockIOWeight= and
352 disable settings prefixed with BlockIO or StartupBlockIO.
353 IODeviceWeight=device weight
355 Set the per-device overall block I/O weight for the executed
356 processes, if the unified control group hierarchy is used on the
357 system. Takes a space-separated pair of a file path and a weight
358 value to specify the device specific weight value, between 1 and
359 10000. (Example: "/dev/sda 1000"). The file path may be specified
360 as path to a block device node or as any other file, in which case
361 the backing block device of the file system of the file is
362 determined. This controls the "io.weight" control group attribute,
363 which defaults to 100. Use this option multiple times to set
364 weights for multiple devices. For details about this control group
365 attribute, see IO Interface Files[8].
366 This setting replaces BlockIODeviceWeight= and disables settings
368 prefixed with BlockIO or StartupBlockIO.
369 The specified device node should reference a block device that has
371 an I/O scheduler associated, i.e. should not refer to partition or
372 loopback block devices, but to the originating, physical device.
373 When a path to a regular file or directory is specified it is
374 attempted to discover the correct originating device backing the
375 file system of the specified path. This works correctly only for
376 simpler cases, where the file system is directly placed on a
377 partition or physical block device, or where simple 1:1 encryption
378 using dm-crypt/LUKS is used. This discovery does not cover complex
379 storage and in particular RAID and volume management storage
380 devices.
381 IOReadBandwidthMax=device bytes, IOWriteBandwidthMax=device bytes
383 Set the per-device overall block I/O bandwidth maximum limit for
384 the executed processes, if the unified control group hierarchy is
385 used on the system. This limit is not work-conserving and the
386 executed processes are not allowed to use more even if the device
387 has idle capacity. Takes a space-separated pair of a file path and
388 a bandwidth value (in bytes per second) to specify the device
389 specific bandwidth. The file path may be a path to a block device
390 node, or as any other file in which case the backing block device
391 of the file system of the file is used. If the bandwidth is
392 suffixed with K, M, G, or T, the specified bandwidth is parsed as
393 Kilobytes, Megabytes, Gigabytes, or Terabytes, respectively, to the
394 base of 1000. (Example:
395 "/dev/disk/by-path/pci-0000:00:1f.2-scsi-0:0:0:0 5M"). This
396 controls the "io.max" control group attributes. Use this option
397 multiple times to set bandwidth limits for multiple devices. For
398 details about this control group attribute, see IO Interface
399 Files[8].
400 These settings replace BlockIOReadBandwidth= and
402 BlockIOWriteBandwidth= and disable settings prefixed with BlockIO
403 or StartupBlockIO.
404 Similar restrictions on block device discovery as for
406 IODeviceWeight= apply, see above.
407 IOReadIOPSMax=device IOPS, IOWriteIOPSMax=device IOPS
409 Set the per-device overall block I/O IOs-Per-Second maximum limit
410 for the executed processes, if the unified control group hierarchy
411 is used on the system. This limit is not work-conserving and the
412 executed processes are not allowed to use more even if the device
413 has idle capacity. Takes a space-separated pair of a file path and
414 an IOPS value to specify the device specific IOPS. The file path
415 may be a path to a block device node, or as any other file in which
416 case the backing block device of the file system of the file is
417 used. If the IOPS is suffixed with K, M, G, or T, the specified
418 IOPS is parsed as KiloIOPS, MegaIOPS, GigaIOPS, or TeraIOPS,
419 respectively, to the base of 1000. (Example:
420 "/dev/disk/by-path/pci-0000:00:1f.2-scsi-0:0:0:0 1K"). This
421 controls the "io.max" control group attributes. Use this option
422 multiple times to set IOPS limits for multiple devices. For details
423 about this control group attribute, see IO Interface Files[8].
424 These settings are supported only if the unified control group
426 hierarchy is used and disable settings prefixed with BlockIO or
427 StartupBlockIO.
428 Similar restrictions on block device discovery as for
430 IODeviceWeight= apply, see above.
431 IODeviceLatencyTargetSec=device target
433 Set the per-device average target I/O latency for the executed
434 processes, if the unified control group hierarchy is used on the
435 system. Takes a file path and a timespan separated by a space to
436 specify the device specific latency target. (Example: "/dev/sda
437 25ms"). The file path may be specified as path to a block device
438 node or as any other file, in which case the backing block device
439 of the file system of the file is determined. This controls the
440 "io.latency" control group attribute. Use this option multiple
441 times to set latency target for multiple devices. For details about
442 this control group attribute, see IO Interface Files[8].
443 Implies "IOAccounting=yes".
445 These settings are supported only if the unified control group
447 hierarchy is used.
448 Similar restrictions on block device discovery as for
450 IODeviceWeight= apply, see above.
451 IPAccounting=
453 Takes a boolean argument. If true, turns on IPv4 and IPv6 network
454 traffic accounting for packets sent or received by the unit. When
455 this option is turned on, all IPv4 and IPv6 sockets created by any
456 process of the unit are accounted for.
457 When this option is used in socket units, it applies to all IPv4
459 and IPv6 sockets associated with it (including both listening and
460 connection sockets where this applies). Note that for
461 socket-activated services, this configuration setting and the
462 accounting data of the service unit and the socket unit are kept
463 separate, and displayed separately. No propagation of the setting
464 and the collected statistics is done, in either direction.
465 Moreover, any traffic sent or received on any of the socket unit's
466 sockets is accounted to the socket unit — and never to the service
467 unit it might have activated, even if the socket is used by it.
468 The system default for this setting may be controlled with
470 DefaultIPAccounting= in systemd-system.conf(5).
471 IPAddressAllow=ADDRESS[/PREFIXLENGTH]...,
473 IPAddressDeny=ADDRESS[/PREFIXLENGTH]...
474 Turn on network traffic filtering for IP packets sent and received
475 over AF_INET and AF_INET6 sockets. Both directives take a space
476 separated list of IPv4 or IPv6 addresses, each optionally suffixed
477 with an address prefix length in bits after a "/" character. If the
478 suffix is omitted, the address is considered a host address, i.e.
479 the filter covers the whole address (32 bits for IPv4, 128 bits for
480 IPv6).
481 The access lists configured with this option are applied to all
483 sockets created by processes of this unit (or in the case of socket
484 units, associated with it). The lists are implicitly combined with
485 any lists configured for any of the parent slice units this unit
486 might be a member of. By default both access lists are empty. Both
487 ingress and egress traffic is filtered by these settings. In case
488 of ingress traffic the source IP address is checked against these
489 access lists, in case of egress traffic the destination IP address
490 is checked. The following rules are applied in turn:
491 • Access is granted when the checked IP address matches an entry
493 in the IPAddressAllow= list.
494 • Otherwise, access is denied when the checked IP address matches
496 an entry in the IPAddressDeny= list.
497 • Otherwise, access is granted.
499 In order to implement an allow-listing IP firewall, it is
501 recommended to use a IPAddressDeny=any setting on an upper-level
502 slice unit (such as the root slice -.slice or the slice containing
503 all system services system.slice – see systemd.special(7) for
504 details on these slice units), plus individual per-service
505 IPAddressAllow= lines permitting network access to relevant
506 services, and only them.
507 Note that for socket-activated services, the IP access list
509 configured on the socket unit applies to all sockets associated
510 with it directly, but not to any sockets created by the ultimately
511 activated services for it. Conversely, the IP access list
512 configured for the service is not applied to any sockets passed
513 into the service via socket activation. Thus, it is usually a good
514 idea to replicate the IP access lists on both the socket and the
515 service unit. Nevertheless, it may make sense to maintain one list
516 more open and the other one more restricted, depending on the
517 usecase.
518 If these settings are used multiple times in the same unit the
520 specified lists are combined. If an empty string is assigned to
521 these settings the specific access list is reset and all previous
522 settings undone.
523 In place of explicit IPv4 or IPv6 address and prefix length
525 specifications a small set of symbolic names may be used. The
526 following names are defined:
527 Table 1. Special address/network names
529 ┌──────────────┬─────────────────────┬─────────────────────┐
530 │Symbolic Name │ Definition │ Meaning │
531 ├──────────────┼─────────────────────┼─────────────────────┤
532 │any │ 0.0.0.0/0 ::/0 │ Any host │
533 ├──────────────┼─────────────────────┼─────────────────────┤
534 │localhost │ 127.0.0.0/8 ::1/128 │ All addresses on │
535 │ │ │ the local loopback │
536 ├──────────────┼─────────────────────┼─────────────────────┤
537 │link-local │ 169.254.0.0/16 │ All link-local IP │
538 │ │ fe80::/64 │ addresses │
539 ├──────────────┼─────────────────────┼─────────────────────┤
540 │multicast │ 224.0.0.0/4 │ All IP multicasting │
541 │ │ ff00::/8 │ addresses │
542 └──────────────┴─────────────────────┴─────────────────────┘
543 Note that these settings might not be supported on some systems
544 (for example if eBPF control group support is not enabled in the
545 underlying kernel or container manager). These settings will have
546 no effect in that case. If compatibility with such systems is
547 desired it is hence recommended to not exclusively rely on them for
548 IP security.
549 IPIngressFilterPath=BPF_FS_PROGRAM_PATH,
551 IPEgressFilterPath=BPF_FS_PROGRAM_PATH
552 Add custom network traffic filters implemented as BPF programs,
553 applying to all IP packets sent and received over AF_INET and
554 AF_INET6 sockets. Takes an absolute path to a pinned BPF program in
555 the BPF virtual filesystem (/sys/fs/bpf/).
556 The filters configured with this option are applied to all sockets
558 created by processes of this unit (or in the case of socket units,
559 associated with it). The filters are loaded in addition to filters
560 any of the parent slice units this unit might be a member of as
561 well as any IPAddressAllow= and IPAddressDeny= filters in any of
562 these units. By default there are no filters specified.
563 If these settings are used multiple times in the same unit all the
565 specified programs are attached. If an empty string is assigned to
566 these settings the program list is reset and all previous specified
567 programs ignored.
568 If the path BPF_FS_PROGRAM_PATH in IPIngressFilterPath= assignment
570 is already being handled by BPFProgram= ingress hook, e.g.
571 BPFProgram=ingress:BPF_FS_PROGRAM_PATH, the assignment will be
572 still considered valid and the program will be attached to a
573 cgroup. Same for IPEgressFilterPath= path and egress hook.
574 Note that for socket-activated services, the IP filter programs
576 configured on the socket unit apply to all sockets associated with
577 it directly, but not to any sockets created by the ultimately
578 activated services for it. Conversely, the IP filter programs
579 configured for the service are not applied to any sockets passed
580 into the service via socket activation. Thus, it is usually a good
581 idea, to replicate the IP filter programs on both the socket and
582 the service unit, however it often makes sense to maintain one
583 configuration more open and the other one more restricted,
584 depending on the usecase.
585 Note that these settings might not be supported on some systems
587 (for example if eBPF control group support is not enabled in the
588 underlying kernel or container manager). These settings will fail
589 the service in that case. If compatibility with such systems is
590 desired it is hence recommended to attach your filter manually
591 (requires Delegate=yes) instead of using this setting.
592 BPFProgram=type:program-path
594 Add a custom cgroup BPF program.
595 BPFProgram= allows attaching BPF hooks to the cgroup of a systemd
597 unit. (This generalizes the functionality exposed via
598 IPEgressFilterPath= for egress and IPIngressFilterPath= for
599 ingress.) Cgroup-bpf hooks in the form of BPF programs loaded to
600 the BPF filesystem are attached with cgroup-bpf attach flags
601 determined by the unit. For details about attachment types and
602 flags see
603 https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/plain/include/uapi/linux/bpf.h.
604 For general BPF documentation please refer to
605 https://www.kernel.org/doc/html/latest/bpf/index.html.
606 The specification of BPF program consists of a type followed by a
608 program-path with ":" as the separator: type:program-path.
609 type is the string name of BPF attach type also used in bpftool.
611 type can be one of egress, ingress, sock_create, sock_ops, device,
612 bind4, bind6, connect4, connect6, post_bind4, post_bind6, sendmsg4,
613 sendmsg6, sysctl, recvmsg4, recvmsg6, getsockopt, setsockopt.
614 Setting BPFProgram= to an empty value makes previous assignments
616 ineffective.
617 Multiple assignments of the same type:program-path value have the
619 same effect as a single assignment: the program with the path
620 program-path will be attached to cgroup hook type just once.
621 If BPF egress pinned to program-path path is already being handled
623 by IPEgressFilterPath=, BPFProgram= assignment will be considered
624 valid and BPFProgram= will be attached to a cgroup. Similarly for
625 ingress hook and IPIngressFilterPath= assignment.
626 BPF programs passed with BPFProgram= are attached to the cgroup of
628 a unit with BPF attach flag multi, that allows further attachments
629 of the same type within cgroup hierarchy topped by the unit cgroup.
630 Examples:
632 BPFProgram=egress:/sys/fs/bpf/egress-hook
634 BPFProgram=bind6:/sys/fs/bpf/sock-addr-hook
635 SocketBindAllow=bind-rule, SocketBindDeny=bind-rule
637 Allow or deny binding a socket address to a socket by matching it
638 with the bind-rule and applying a corresponding action if there is
639 a match.
640 bind-rule describes socket properties such as address-family,
642 transport-protocol and ip-ports.
643 bind-rule := { [address-family:][transport-protocol:][ip-ports] |
645 any }
646 address-family := { ipv4 | ipv6 }
648 transport-protocol := { tcp | udp }
650 ip-ports := { ip-port | ip-port-range }
652 An optional address-family expects ipv4 or ipv6 values. If not
654 specified, a rule will be matched for both IPv4 and IPv6 addresses
655 and applied depending on other socket fields, e.g.
656 transport-protocol, ip-port.
657 An optional transport-protocol expects tcp or udp transport
659 protocol names. If not specified, a rule will be matched for any
660 transport protocol.
661 An optional ip-port value must lie within 1...65535 interval
663 inclusively, i.e. dynamic port 0 is not allowed. A range of
664 sequential ports is described by ip-port-range :=
665 ip-port-low-ip-port-high, where ip-port-low is smaller than or
666 equal to ip-port-high and both are within 1...65535 inclusively.
667 A special value any can be used to apply a rule to any address
669 family, transport protocol and any port with a positive value.
670 To allow multiple rules assign SocketBindAllow= or SocketBindDeny=
672 multiple times. To clear the existing assignments pass an empty
673 SocketBindAllow= or SocketBindDeny= assignment.
674 For each of SocketBindAllow= and SocketBindDeny=, maximum allowed
676 number of assignments is 128.
677 • Binding to a socket is allowed when a socket address matches an
679 entry in the SocketBindAllow= list.
680 • Otherwise, binding is denied when the socket address matches an
682 entry in the SocketBindDeny= list.
683 • Otherwise, binding is allowed.
685 The feature is implemented with cgroup/bind4 and cgroup/bind6
687 cgroup-bpf hooks.
688 Examples:
690 ...
692 # Allow binding IPv6 socket addresses with a port greater than or equal to 10000.
693 [Service]
694 SocketBindAllow=ipv6:10000-65535
695 SocketBindDeny=any
696 ...
697 # Allow binding IPv4 and IPv6 socket addresses with 1234 and 4321 ports.
698 [Service]
699 SocketBindAllow=1234
700 SocketBindAllow=4321
701 SocketBindDeny=any
702 ...
703 # Deny binding IPv6 socket addresses.
704 [Service]
705 SocketBindDeny=ipv6
706 ...
707 # Deny binding IPv4 and IPv6 socket addresses.
708 [Service]
709 SocketBindDeny=any
710 ...
711 # Allow binding only over TCP
712 [Service]
713 SocketBindAllow=tcp
714 SocketBindDeny=any
715 ...
716 # Allow binding only over IPv6/TCP
717 [Service]
718 SocketBindAllow=ipv6:tcp
719 SocketBindDeny=any
720 ...
721 # Allow binding ports within 10000-65535 range over IPv4/UDP.
722 [Service]
723 SocketBindAllow=ipv4:udp:10000-65535
724 SocketBindDeny=any
725 ...
726 RestrictNetworkInterfaces=
728 Takes a list of space-separated network interface names. This
729 option restricts the network interfaces that processes of this unit
730 can use. By default processes can only use the network interfaces
731 listed (allow-list). If the first character of the rule is "~", the
732 effect is inverted: the processes can only use network interfaces
733 not listed (deny-list).
734 This option can appear multiple times, in which case the network
736 interface names are merged. If the empty string is assigned the set
737 is reset, all prior assignments will have not effect.
738 If you specify both types of this option (i.e. allow-listing and
740 deny-listing), the first encountered will take precedence and will
741 dictate the default action (allow vs deny). Then the next
742 occurrences of this option will add or delete the listed network
743 interface names from the set, depending of its type and the default
744 action.
745 The loopback interface ("lo") is not treated in any special way,
747 you have to configure it explicitly in the unit file.
748 Example 1: allow-list
750 RestrictNetworkInterfaces=eth1
752 RestrictNetworkInterfaces=eth2
753 Programs in the unit will be only able to use the eth1 and eth2
755 network interfaces.
756 Example 2: deny-list
758 RestrictNetworkInterfaces=~eth1 eth2
760 Programs in the unit will be able to use any network interface but
762 eth1 and eth2.
763 Example 3: mixed
765 RestrictNetworkInterfaces=eth1 eth2
767 RestrictNetworkInterfaces=~eth1
768 Programs in the unit will be only able to use the eth2 network
770 interface.
771 DeviceAllow=
773 Control access to specific device nodes by the executed processes.
774 Takes two space-separated strings: a device node specifier followed
775 by a combination of r, w, m to control reading, writing, or
776 creation of the specific device node(s) by the unit (mknod),
777 respectively. On cgroup-v1 this controls the "devices.allow"
778 control group attribute. For details about this control group
779 attribute, see Device Whitelist Controller[9]. In the unified
780 cgroup hierarchy this functionality is implemented using eBPF
781 filtering.
782 When access to all physical devices should be disallowed,
784 PrivateDevices= may be used instead. See systemd.exec(5).
785 The device node specifier is either a path to a device node in the
787 file system, starting with /dev/, or a string starting with either
788 "char-" or "block-" followed by a device group name, as listed in
789 /proc/devices. The latter is useful to allow-list all current and
790 future devices belonging to a specific device group at once. The
791 device group is matched according to filename globbing rules, you
792 may hence use the "*" and "?" wildcards. (Note that such globbing
793 wildcards are not available for device node path specifications!)
794 In order to match device nodes by numeric major/minor, use device
795 node paths in the /dev/char/ and /dev/block/ directories. However,
796 matching devices by major/minor is generally not recommended as
797 assignments are neither stable nor portable between systems or
798 different kernel versions.
799 Examples: /dev/sda5 is a path to a device node, referring to an ATA
801 or SCSI block device. "char-pts" and "char-alsa" are specifiers
802 for all pseudo TTYs and all ALSA sound devices, respectively.
803 "char-cpu/*" is a specifier matching all CPU related device groups.
804 Note that allow lists defined this way should only reference device
806 groups which are resolvable at the time the unit is started. Any
807 device groups not resolvable then are not added to the device allow
808 list. In order to work around this limitation, consider extending
809 service units with a pair of After=modprobe@xyz.service and
810 Wants=modprobe@xyz.service lines that load the necessary kernel
811 module implementing the device group if missing. Example:
812 ...
814 [Unit]
815 Wants=modprobe@loop.service
816 After=modprobe@loop.service
817 [Service]
819 DeviceAllow=block-loop
820 DeviceAllow=/dev/loop-control
821 ...
822 DevicePolicy=auto|closed|strict
824 Control the policy for allowing device access:
825 strict
827 means to only allow types of access that are explicitly
828 specified.
829 closed
831 in addition, allows access to standard pseudo devices including
832 /dev/null, /dev/zero, /dev/full, /dev/random, and /dev/urandom.
833 auto
835 in addition, allows access to all devices if no explicit
836 DeviceAllow= is present. This is the default.
837 Slice=
839 The name of the slice unit to place the unit in. Defaults to
840 system.slice for all non-instantiated units of all unit types
841 (except for slice units themselves see below). Instance units are
842 by default placed in a subslice of system.slice that is named after
843 the template name.
844 This option may be used to arrange systemd units in a hierarchy of
846 slices each of which might have resource settings applied.
847 For units of type slice, the only accepted value for this setting
849 is the parent slice. Since the name of a slice unit implies the
850 parent slice, it is hence redundant to ever set this parameter
851 directly for slice units.
852 Special care should be taken when relying on the default slice
854 assignment in templated service units that have
855 DefaultDependencies=no set, see systemd.service(5), section
856 "Default Dependencies" for details.
857 Delegate=
859 Turns on delegation of further resource control partitioning to
860 processes of the unit. Units where this is enabled may create and
861 manage their own private subhierarchy of control groups below the
862 control group of the unit itself. For unprivileged services (i.e.
863 those using the User= setting) the unit's control group will be
864 made accessible to the relevant user. When enabled the service
865 manager will refrain from manipulating control groups or moving
866 processes below the unit's control group, so that a clear concept
867 of ownership is established: the control group tree above the
868 unit's control group (i.e. towards the root control group) is owned
869 and managed by the service manager of the host, while the control
870 group tree below the unit's control group is owned and managed by
871 the unit itself. Takes either a boolean argument or a list of
872 control group controller names. If true, delegation is turned on,
873 and all supported controllers are enabled for the unit, making them
874 available to the unit's processes for management. If false,
875 delegation is turned off entirely (and no additional controllers
876 are enabled). If set to a list of controllers, delegation is turned
877 on, and the specified controllers are enabled for the unit. Note
878 that additional controllers than the ones specified might be made
879 available as well, depending on configuration of the containing
880 slice unit or other units contained in it. Note that assigning the
881 empty string will enable delegation, but reset the list of
882 controllers, all assignments prior to this will have no effect.
883 Defaults to false.
884 Note that controller delegation to less privileged code is only
886 safe on the unified control group hierarchy. Accordingly, access to
887 the specified controllers will not be granted to unprivileged
888 services on the legacy hierarchy, even when requested.
889 The following controller names may be specified: cpu, cpuacct,
891 cpuset, io, blkio, memory, devices, pids, bpf-firewall, and
892 bpf-devices.
893 Not all of these controllers are available on all kernels however,
895 and some are specific to the unified hierarchy while others are
896 specific to the legacy hierarchy. Also note that the kernel might
897 support further controllers, which aren't covered here yet as
898 delegation is either not supported at all for them or not defined
899 cleanly.
900 For further details on the delegation model consult Control Group
902 APIs and Delegation[10].
903 DisableControllers=
905 Disables controllers from being enabled for a unit's children. If a
906 controller listed is already in use in its subtree, the controller
907 will be removed from the subtree. This can be used to avoid child
908 units being able to implicitly or explicitly enable a controller.
909 Defaults to not disabling any controllers.
910 It may not be possible to successfully disable a controller if the
912 unit or any child of the unit in question delegates controllers to
913 its children, as any delegated subtree of the cgroup hierarchy is
914 unmanaged by systemd.
915 Multiple controllers may be specified, separated by spaces. You may
917 also pass DisableControllers= multiple times, in which case each
918 new instance adds another controller to disable. Passing
919 DisableControllers= by itself with no controller name present
920 resets the disabled controller list.
921 The following controller names may be specified: cpu, cpuacct,
923 cpuset, io, blkio, memory, devices, pids, bpf-firewall, and
924 bpf-devices.
925 ManagedOOMSwap=auto|kill, ManagedOOMMemoryPressure=auto|kill
927 Specifies how systemd-oomd.service(8) will act on this unit's
928 cgroups. Defaults to auto.
929 When set to kill, systemd-oomd will actively monitor this unit's
931 cgroup metrics to decide whether it needs to act. If the cgroup
932 passes the limits set by oomd.conf(5) or its overrides,
933 systemd-oomd will send a SIGKILL to all of the processes under the
934 chosen candidate cgroup. Note that only descendant cgroups can be
935 eligible candidates for killing; the unit that set its property to
936 kill is not a candidate (unless one of its ancestors set their
937 property to kill). You can find more details on candidates and kill
938 behavior at systemd-oomd.service(8) and oomd.conf(5). Setting
939 either of these properties to kill will also automatically acquire
940 After= and Wants= dependencies on systemd-oomd.service unless
941 DefaultDependencies=no.
942 When set to auto, systemd-oomd will not actively use this cgroup's
944 data for monitoring and detection. However, if an ancestor cgroup
945 has one of these properties set to kill, a unit with auto can still
946 be an eligible candidate for systemd-oomd to act on.
947 ManagedOOMMemoryPressureLimit=
949 Overrides the default memory pressure limit set by oomd.conf(5) for
950 this unit (cgroup). Takes a percentage value between 0% and 100%,
951 inclusive. This property is ignored unless
952 ManagedOOMMemoryPressure=kill. Defaults to 0%, which means to use
953 the default set by oomd.conf(5).
954 ManagedOOMPreference=none|avoid|omit
956 Allows deprioritizing or omitting this unit's cgroup as a candidate
957 when systemd-oomd needs to act. Requires support for extended
958 attributes (see xattr(7)) in order to use avoid or omit.
959 Additionally, systemd-oomd will ignore these extended attributes if
960 the unit's cgroup is not owned by the root user.
961 If this property is set to avoid, the service manager will convey
963 this to systemd-oomd, which will only select this cgroup if there
964 are no other viable candidates.
965 If this property is set to omit, the service manager will convey
967 this to systemd-oomd, which will ignore this cgroup as a candidate
968 and will not perform any actions on it.
969 It is recommended to use avoid and omit sparingly, as it can
971 adversely affect systemd-oomd's kill behavior. Also note that these
972 extended attributes are not applied recursively to cgroups under
973 this unit's cgroup.
974 Defaults to none which means systemd-oomd will rank this unit's
976 cgroup as defined in systemd-oomd.service(8) and oomd.conf(5).
977
979 The following options are deprecated. Use the indicated superseding
980 options instead:
981 CPUShares=weight, StartupCPUShares=weight
983 Assign the specified CPU time share weight to the processes
984 executed. These options take an integer value and control the
985 "cpu.shares" control group attribute. The allowed range is 2 to
986 262144. Defaults to 1024. For details about this control group
987 attribute, see CFS Scheduler[4]. The available CPU time is split up
988 among all units within one slice relative to their CPU time share
989 weight.
990 While StartupCPUShares= applies to the startup and shutdown phases
992 of the system, CPUShares= applies to normal runtime of the system,
993 and if the former is not set also to the startup and shutdown
994 phases. Using StartupCPUShares= allows prioritizing specific
995 services at boot-up and shutdown differently than during normal
996 runtime.
997 Implies "CPUAccounting=yes".
999 These settings are deprecated. Use CPUWeight= and StartupCPUWeight=
1001 instead.
1002 MemoryLimit=bytes
1004 Specify the limit on maximum memory usage of the executed
1005 processes. The limit specifies how much process and kernel memory
1006 can be used by tasks in this unit. Takes a memory size in bytes. If
1007 the value is suffixed with K, M, G or T, the specified memory size
1008 is parsed as Kilobytes, Megabytes, Gigabytes, or Terabytes (with
1009 the base 1024), respectively. Alternatively, a percentage value may
1010 be specified, which is taken relative to the installed physical
1011 memory on the system. If assigned the special value "infinity", no
1012 memory limit is applied. This controls the "memory.limit_in_bytes"
1013 control group attribute. For details about this control group
1014 attribute, see Memory Resource Controller[11].
1015 Implies "MemoryAccounting=yes".
1017 This setting is deprecated. Use MemoryMax= instead.
1019 BlockIOAccounting=
1021 Turn on Block I/O accounting for this unit, if the legacy control
1022 group hierarchy is used on the system. Takes a boolean argument.
1023 Note that turning on block I/O accounting for one unit will also
1024 implicitly turn it on for all units contained in the same slice and
1025 all for its parent slices and the units contained therein. The
1026 system default for this setting may be controlled with
1027 DefaultBlockIOAccounting= in systemd-system.conf(5).
1028 This setting is deprecated. Use IOAccounting= instead.
1030 BlockIOWeight=weight, StartupBlockIOWeight=weight
1032 Set the default overall block I/O weight for the executed
1033 processes, if the legacy control group hierarchy is used on the
1034 system. Takes a single weight value (between 10 and 1000) to set
1035 the default block I/O weight. This controls the "blkio.weight"
1036 control group attribute, which defaults to 500. For details about
1037 this control group attribute, see Block IO Controller[12]. The
1038 available I/O bandwidth is split up among all units within one
1039 slice relative to their block I/O weight.
1040 While StartupBlockIOWeight= only applies to the startup and
1042 shutdown phases of the system, BlockIOWeight= applies to the later
1043 runtime of the system, and if the former is not set also to the
1044 startup and shutdown phases. This allows prioritizing specific
1045 services at boot-up and shutdown differently than during runtime.
1046 Implies "BlockIOAccounting=yes".
1048 These settings are deprecated. Use IOWeight= and StartupIOWeight=
1050 instead.
1051 BlockIODeviceWeight=device weight
1053 Set the per-device overall block I/O weight for the executed
1054 processes, if the legacy control group hierarchy is used on the
1055 system. Takes a space-separated pair of a file path and a weight
1056 value to specify the device specific weight value, between 10 and
1057 1000. (Example: "/dev/sda 500"). The file path may be specified as
1058 path to a block device node or as any other file, in which case the
1059 backing block device of the file system of the file is determined.
1060 This controls the "blkio.weight_device" control group attribute,
1061 which defaults to 1000. Use this option multiple times to set
1062 weights for multiple devices. For details about this control group
1063 attribute, see Block IO Controller[12].
1064 Implies "BlockIOAccounting=yes".
1066 This setting is deprecated. Use IODeviceWeight= instead.
1068 BlockIOReadBandwidth=device bytes, BlockIOWriteBandwidth=device bytes
1070 Set the per-device overall block I/O bandwidth limit for the
1071 executed processes, if the legacy control group hierarchy is used
1072 on the system. Takes a space-separated pair of a file path and a
1073 bandwidth value (in bytes per second) to specify the device
1074 specific bandwidth. The file path may be a path to a block device
1075 node, or as any other file in which case the backing block device
1076 of the file system of the file is used. If the bandwidth is
1077 suffixed with K, M, G, or T, the specified bandwidth is parsed as
1078 Kilobytes, Megabytes, Gigabytes, or Terabytes, respectively, to the
1079 base of 1000. (Example:
1080 "/dev/disk/by-path/pci-0000:00:1f.2-scsi-0:0:0:0 5M"). This
1081 controls the "blkio.throttle.read_bps_device" and
1082 "blkio.throttle.write_bps_device" control group attributes. Use
1083 this option multiple times to set bandwidth limits for multiple
1084 devices. For details about these control group attributes, see
1085 Block IO Controller[12].
1086 Implies "BlockIOAccounting=yes".
1088 These settings are deprecated. Use IOReadBandwidthMax= and
1090 IOWriteBandwidthMax= instead.
1091
1093 systemd(1), systemd-system.conf(5), systemd.unit(5),
1094 systemd.service(5), systemd.slice(5), systemd.scope(5),
1095 systemd.socket(5), systemd.mount(5), systemd.swap(5), systemd.exec(5),
1096 systemd.directives(7), systemd.special(7), systemd-oomd.service(8), The
1097 documentation for control groups and specific controllers in the Linux
1098 kernel: Control Groups v2[2].
1099
1101 1. New Control Group Interfaces
1102 https://www.freedesktop.org/wiki/Software/systemd/ControlGroupInterface/
1103 2. Control Groups v2
1105 https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v2.html
1106 3. Control Groups version 1
1108 https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v1/
1109 4. CFS Scheduler
1111 https://www.kernel.org/doc/html/latest/scheduler/sched-design-CFS.html
1112 5. sched-bwc.txt
1114 https://www.kernel.org/doc/Documentation/scheduler/sched-bwc.txt
1115 6. Memory Interface Files
1117 https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v2.html#memory-interface-files
1118 7. Process Number Controller
1120 https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v1/pids.html
1121 8. IO Interface Files
1123 https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v2.html#io-interface-files
1124 9. Device Whitelist Controller
1126 https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v1/devices.html
1127 10. Control Group APIs and Delegation
1129 https://systemd.io/CGROUP_DELEGATION
1130 11. Memory Resource Controller
1132 https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v1/memory.html
1133 12. Block IO Controller
1135 https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v1/blkio-controller.html
1136systemd 250 SYSTEMD.RESOURCE-CONTROL(5)