1CGROUPS(7) Linux Programmer's Manual CGROUPS(7)
2
6 cgroups - Linux control groups
7
9 Control groups, usually referred to as cgroups, are a Linux kernel fea‐
10 ture which allow processes to be organized into hierarchical groups
11 whose usage of various types of resources can then be limited and moni‐
12 tored. The kernel's cgroup interface is provided through a pseudo-
13 filesystem called cgroupfs. Grouping is implemented in the core cgroup
14 kernel code, while resource tracking and limits are implemented in a
15 set of per-resource-type subsystems (memory, CPU, and so on).
16 Terminology
18 A cgroup is a collection of processes that are bound to a set of limits
19 or parameters defined via the cgroup filesystem.
20 A subsystem is a kernel component that modifies the behavior of the
22 processes in a cgroup. Various subsystems have been implemented, mak‐
23 ing it possible to do things such as limiting the amount of CPU time
24 and memory available to a cgroup, accounting for the CPU time used by a
25 cgroup, and freezing and resuming execution of the processes in a
26 cgroup. Subsystems are sometimes also known as resource controllers
27 (or simply, controllers).
28 The cgroups for a controller are arranged in a hierarchy. This hierar‐
30 chy is defined by creating, removing, and renaming subdirectories
31 within the cgroup filesystem. At each level of the hierarchy,
32 attributes (e.g., limits) can be defined. The limits, control, and
33 accounting provided by cgroups generally have effect throughout the
34 subhierarchy underneath the cgroup where the attributes are defined.
35 Thus, for example, the limits placed on a cgroup at a higher level in
36 the hierarchy cannot be exceeded by descendant cgroups.
37 Cgroups version 1 and version 2
39 The initial release of the cgroups implementation was in Linux 2.6.24.
40 Over time, various cgroup controllers have been added to allow the man‐
41 agement of various types of resources. However, the development of
42 these controllers was largely uncoordinated, with the result that many
43 inconsistencies arose between controllers and management of the cgroup
44 hierarchies became rather complex. (A longer description of these
45 problems can be found in the kernel source file Documenta‐
46 tion/cgroup-v2.txt.)
47 Because of the problems with the initial cgroups implementation
49 (cgroups version 1), starting in Linux 3.10, work began on a new,
50 orthogonal implementation to remedy these problems. Initially marked
51 experimental, and hidden behind the -o __DEVEL__sane_behavior mount
52 option, the new version (cgroups version 2) was eventually made offi‐
53 cial with the release of Linux 4.5. Differences between the two ver‐
54 sions are described in the text below. The file cgroup.sane_behavior,
55 present in cgroups v1, is a relic of this mount option. The file always
56 reports "0" and is only retained for backward compatibility.
57 Although cgroups v2 is intended as a replacement for cgroups v1, the
59 older system continues to exist (and for compatibility reasons is
60 unlikely to be removed). Currently, cgroups v2 implements only a sub‐
61 set of the controllers available in cgroups v1. The two systems are
62 implemented so that both v1 controllers and v2 controllers can be
63 mounted on the same system. Thus, for example, it is possible to use
64 those controllers that are supported under version 2, while also using
65 version 1 controllers where version 2 does not yet support those con‐
66 trollers. The only restriction here is that a controller can't be
67 simultaneously employed in both a cgroups v1 hierarchy and in the
68 cgroups v2 hierarchy.
69
71 Under cgroups v1, each controller may be mounted against a separate
72 cgroup filesystem that provides its own hierarchical organization of
73 the processes on the system. It is also possible to comount multiple
74 (or even all) cgroups v1 controllers against the same cgroup filesys‐
75 tem, meaning that the comounted controllers manage the same hierarchi‐
76 cal organization of processes.
77 For each mounted hierarchy, the directory tree mirrors the control
79 group hierarchy. Each control group is represented by a directory,
80 with each of its child control cgroups represented as a child direc‐
81 tory. For instance, /user/joe/1.session represents control group
82 1.session, which is a child of cgroup joe, which is a child of /user.
83 Under each cgroup directory is a set of files which can be read or
84 written to, reflecting resource limits and a few general cgroup proper‐
85 ties.
86 Tasks (threads) versus processes
88 In cgroups v1, a distinction is drawn between processes and tasks. In
89 this view, a process can consist of multiple tasks (more commonly
90 called threads, from a user-space perspective, and called such in the
91 remainder of this man page). In cgroups v1, it is possible to indepen‐
92 dently manipulate the cgroup memberships of the threads in a process.
93 The cgroups v1 ability to split threads across different cgroups caused
95 problems in some cases. For example, it made no sense for the memory
96 controller, since all of the threads of a process share a single
97 address space. Because of these problems, the ability to independently
98 manipulate the cgroup memberships of the threads in a process was
99 removed in the initial cgroups v2 implementation, and subsequently
100 restored in a more limited form (see the discussion of "thread mode"
101 below).
102 Mounting v1 controllers
104 The use of cgroups requires a kernel built with the CONFIG_CGROUP
105 option. In addition, each of the v1 controllers has an associated con‐
106 figuration option that must be set in order to employ that controller.
107 In order to use a v1 controller, it must be mounted against a cgroup
109 filesystem. The usual place for such mounts is under a tmpfs(5)
110 filesystem mounted at /sys/fs/cgroup. Thus, one might mount the cpu
111 controller as follows:
112 mount -t cgroup -o cpu none /sys/fs/cgroup/cpu
114 It is possible to comount multiple controllers against the same hierar‐
116 chy. For example, here the cpu and cpuacct controllers are comounted
117 against a single hierarchy:
118 mount -t cgroup -o cpu,cpuacct none /sys/fs/cgroup/cpu,cpuacct
120 Comounting controllers has the effect that a process is in the same
122 cgroup for all of the comounted controllers. Separately mounting con‐
123 trollers allows a process to be in cgroup /foo1 for one controller
124 while being in /foo2/foo3 for another.
125 It is possible to comount all v1 controllers against the same hierar‐
127 chy:
128 mount -t cgroup -o all cgroup /sys/fs/cgroup
130 (One can achieve the same result by omitting -o all, since it is the
132 default if no controllers are explicitly specified.)
133 It is not possible to mount the same controller against multiple cgroup
135 hierarchies. For example, it is not possible to mount both the cpu and
136 cpuacct controllers against one hierarchy, and to mount the cpu con‐
137 troller alone against another hierarchy. It is possible to create mul‐
138 tiple mount points with exactly the same set of comounted controllers.
139 However, in this case all that results is multiple mount points provid‐
140 ing a view of the same hierarchy.
141 Note that on many systems, the v1 controllers are automatically mounted
143 under /sys/fs/cgroup; in particular, systemd(1) automatically creates
144 such mount points.
145 Unmounting v1 controllers
147 A mounted cgroup filesystem can be unmounted using the umount(8) com‐
148 mand, as in the following example:
149 umount /sys/fs/cgroup/pids
151 But note well: a cgroup filesystem is unmounted only if it is not busy,
153 that is, it has no child cgroups. If this is not the case, then the
154 only effect of the umount(8) is to make the mount invisible. Thus, to
155 ensure that the mount point is really removed, one must first remove
156 all child cgroups, which in turn can be done only after all member pro‐
157 cesses have been moved from those cgroups to the root cgroup.
158 Cgroups version 1 controllers
160 Each of the cgroups version 1 controllers is governed by a kernel con‐
161 figuration option (listed below). Additionally, the availability of
162 the cgroups feature is governed by the CONFIG_CGROUPS kernel configura‐
163 tion option.
164 cpu (since Linux 2.6.24; CONFIG_CGROUP_SCHED)
166 Cgroups can be guaranteed a minimum number of "CPU shares" when
167 a system is busy. This does not limit a cgroup's CPU usage if
168 the CPUs are not busy. For further information, see Documenta‐
169 tion/scheduler/sched-design-CFS.txt.
170 In Linux 3.2, this controller was extended to provide CPU "band‐
172 width" control. If the kernel is configured with CON‐
173 FIG_CFS_BANDWIDTH, then within each scheduling period (defined
174 via a file in the cgroup directory), it is possible to define an
175 upper limit on the CPU time allocated to the processes in a
176 cgroup. This upper limit applies even if there is no other com‐
177 petition for the CPU. Further information can be found in the
178 kernel source file Documentation/scheduler/sched-bwc.txt.
179 cpuacct (since Linux 2.6.24; CONFIG_CGROUP_CPUACCT)
181 This provides accounting for CPU usage by groups of processes.
182 Further information can be found in the kernel source file Docu‐
184 mentation/cgroup-v1/cpuacct.txt.
185 cpuset (since Linux 2.6.24; CONFIG_CPUSETS)
187 This cgroup can be used to bind the processes in a cgroup to a
188 specified set of CPUs and NUMA nodes.
189 Further information can be found in the kernel source file Docu‐
191 mentation/cgroup-v1/cpusets.txt.
192 memory (since Linux 2.6.25; CONFIG_MEMCG)
194 The memory controller supports reporting and limiting of process
195 memory, kernel memory, and swap used by cgroups.
196 Further information can be found in the kernel source file Docu‐
198 mentation/cgroup-v1/memory.txt.
199 devices (since Linux 2.6.26; CONFIG_CGROUP_DEVICE)
201 This supports controlling which processes may create (mknod)
202 devices as well as open them for reading or writing. The poli‐
203 cies may be specified as allow-lists and deny-lists. Hierarchy
204 is enforced, so new rules must not violate existing rules for
205 the target or ancestor cgroups.
206 Further information can be found in the kernel source file Docu‐
208 mentation/cgroup-v1/devices.txt.
209 freezer (since Linux 2.6.28; CONFIG_CGROUP_FREEZER)
211 The freezer cgroup can suspend and restore (resume) all pro‐
212 cesses in a cgroup. Freezing a cgroup /A also causes its chil‐
213 dren, for example, processes in /A/B, to be frozen.
214 Further information can be found in the kernel source file Docu‐
216 mentation/cgroup-v1/freezer-subsystem.txt.
217 net_cls (since Linux 2.6.29; CONFIG_CGROUP_NET_CLASSID)
219 This places a classid, specified for the cgroup, on network
220 packets created by a cgroup. These classids can then be used in
221 firewall rules, as well as used to shape traffic using tc(8).
222 This applies only to packets leaving the cgroup, not to traffic
223 arriving at the cgroup.
224 Further information can be found in the kernel source file Docu‐
226 mentation/cgroup-v1/net_cls.txt.
227 blkio (since Linux 2.6.33; CONFIG_BLK_CGROUP)
229 The blkio cgroup controls and limits access to specified block
230 devices by applying IO control in the form of throttling and
231 upper limits against leaf nodes and intermediate nodes in the
232 storage hierarchy.
233 Two policies are available. The first is a proportional-weight
235 time-based division of disk implemented with CFQ. This is in
236 effect for leaf nodes using CFQ. The second is a throttling
237 policy which specifies upper I/O rate limits on a device.
238 Further information can be found in the kernel source file Docu‐
240 mentation/cgroup-v1/blkio-controller.txt.
241 perf_event (since Linux 2.6.39; CONFIG_CGROUP_PERF)
243 This controller allows perf monitoring of the set of processes
244 grouped in a cgroup.
245 Further information can be found in the kernel source file
247 tools/perf/Documentation/perf-record.txt.
248 net_prio (since Linux 3.3; CONFIG_CGROUP_NET_PRIO)
250 This allows priorities to be specified, per network interface,
251 for cgroups.
252 Further information can be found in the kernel source file Docu‐
254 mentation/cgroup-v1/net_prio.txt.
255 hugetlb (since Linux 3.5; CONFIG_CGROUP_HUGETLB)
257 This supports limiting the use of huge pages by cgroups.
258 Further information can be found in the kernel source file Docu‐
260 mentation/cgroup-v1/hugetlb.txt.
261 pids (since Linux 4.3; CONFIG_CGROUP_PIDS)
263 This controller permits limiting the number of process that may
264 be created in a cgroup (and its descendants).
265 Further information can be found in the kernel source file Docu‐
267 mentation/cgroup-v1/pids.txt.
268 rdma (since Linux 4.11; CONFIG_CGROUP_RDMA)
270 The RDMA controller permits limiting the use of RDMA/IB-specific
271 resources per cgroup.
272 Further information can be found in the kernel source file Docu‐
274 mentation/cgroup-v1/rdma.txt.
275 Creating cgroups and moving processes
277 A cgroup filesystem initially contains a single root cgroup, '/', which
278 all processes belong to. A new cgroup is created by creating a direc‐
279 tory in the cgroup filesystem:
280 mkdir /sys/fs/cgroup/cpu/cg1
282 This creates a new empty cgroup.
284 A process may be moved to this cgroup by writing its PID into the
286 cgroup's cgroup.procs file:
287 echo $$ > /sys/fs/cgroup/cpu/cg1/cgroup.procs
289 Only one PID at a time should be written to this file.
291 Writing the value 0 to a cgroup.procs file causes the writing process
293 to be moved to the corresponding cgroup.
294 When writing a PID into the cgroup.procs, all threads in the process
296 are moved into the new cgroup at once.
297 Within a hierarchy, a process can be a member of exactly one cgroup.
299 Writing a process's PID to a cgroup.procs file automatically removes it
300 from the cgroup of which it was previously a member.
301 The cgroup.procs file can be read to obtain a list of the processes
303 that are members of a cgroup. The returned list of PIDs is not guaran‐
304 teed to be in order. Nor is it guaranteed to be free of duplicates.
305 (For example, a PID may be recycled while reading from the list.)
306 In cgroups v1, an individual thread can be moved to another cgroup by
308 writing its thread ID (i.e., the kernel thread ID returned by clone(2)
309 and gettid(2)) to the tasks file in a cgroup directory. This file can
310 be read to discover the set of threads that are members of the cgroup.
311 Removing cgroups
313 To remove a cgroup, it must first have no child cgroups and contain no
314 (nonzombie) processes. So long as that is the case, one can simply
315 remove the corresponding directory pathname. Note that files in a
316 cgroup directory cannot and need not be removed.
317 Cgroups v1 release notification
319 Two files can be used to determine whether the kernel provides notifi‐
320 cations when a cgroup becomes empty. A cgroup is considered to be
321 empty when it contains no child cgroups and no member processes.
322 A special file in the root directory of each cgroup hierarchy,
324 release_agent, can be used to register the pathname of a program that
325 may be invoked when a cgroup in the hierarchy becomes empty. The path‐
326 name of the newly empty cgroup (relative to the cgroup mount point) is
327 provided as the sole command-line argument when the release_agent pro‐
328 gram is invoked. The release_agent program might remove the cgroup
329 directory, or perhaps repopulate it with a process.
330 The default value of the release_agent file is empty, meaning that no
332 release agent is invoked.
333 The content of the release_agent file can also be specified via a mount
335 option when the cgroup filesystem is mounted:
336 mount -o release_agent=pathname ...
338 Whether or not the release_agent program is invoked when a particular
340 cgroup becomes empty is determined by the value in the
341 notify_on_release file in the corresponding cgroup directory. If this
342 file contains the value 0, then the release_agent program is not
343 invoked. If it contains the value 1, the release_agent program is
344 invoked. The default value for this file in the root cgroup is 0. At
345 the time when a new cgroup is created, the value in this file is inher‐
346 ited from the corresponding file in the parent cgroup.
347 Cgroup v1 named hierarchies
349 In cgroups v1, it is possible to mount a cgroup hierarchy that has no
350 attached controllers:
351 mount -t cgroup -o none,name=somename none /some/mount/point
353 Multiple instances of such hierarchies can be mounted; each hierarchy
355 must have a unique name. The only purpose of such hierarchies is to
356 track processes. (See the discussion of release notification below.)
357 An example of this is the name=systemd cgroup hierarchy that is used by
358 systemd(1) to track services and user sessions.
359 Since Linux 5.0, the cgroup_no_v1 kernel boot option (described below)
361 can be used to disable cgroup v1 named hierarchies, by specifying
362 cgroup_no_v1=named.
363
366 In cgroups v2, all mounted controllers reside in a single unified hier‐
367 archy. While (different) controllers may be simultaneously mounted
368 under the v1 and v2 hierarchies, it is not possible to mount the same
369 controller simultaneously under both the v1 and the v2 hierarchies.
370 The new behaviors in cgroups v2 are summarized here, and in some cases
372 elaborated in the following subsections.
373 1. Cgroups v2 provides a unified hierarchy against which all con‐
375 trollers are mounted.
376 2. "Internal" processes are not permitted. With the exception of the
378 root cgroup, processes may reside only in leaf nodes (cgroups that
379 do not themselves contain child cgroups). The details are somewhat
380 more subtle than this, and are described below.
381 3. Active cgroups must be specified via the files cgroup.controllers
383 and cgroup.subtree_control.
384 4. The tasks file has been removed. In addition, the
386 cgroup.clone_children file that is employed by the cpuset controller
387 has been removed.
388 5. An improved mechanism for notification of empty cgroups is provided
390 by the cgroup.events file.
391 For more changes, see the Documentation/cgroup-v2.txt file in the ker‐
393 nel source.
394 Some of the new behaviors listed above saw subsequent modification with
396 the addition in Linux 4.14 of "thread mode" (described below).
397 Cgroups v2 unified hierarchy
399 In cgroups v1, the ability to mount different controllers against dif‐
400 ferent hierarchies was intended to allow great flexibility for applica‐
401 tion design. In practice, though, the flexibility turned out to be
402 less useful than expected, and in many cases added complexity. There‐
403 fore, in cgroups v2, all available controllers are mounted against a
404 single hierarchy. The available controllers are automatically mounted,
405 meaning that it is not necessary (or possible) to specify the con‐
406 trollers when mounting the cgroup v2 filesystem using a command such as
407 the following:
408 mount -t cgroup2 none /mnt/cgroup2
410 A cgroup v2 controller is available only if it is not currently in use
412 via a mount against a cgroup v1 hierarchy. Or, to put things another
413 way, it is not possible to employ the same controller against both a v1
414 hierarchy and the unified v2 hierarchy. This means that it may be nec‐
415 essary first to unmount a v1 controller (as described above) before
416 that controller is available in v2. Since systemd(1) makes heavy use
417 of some v1 controllers by default, it can in some cases be simpler to
418 boot the system with selected v1 controllers disabled. To do this,
419 specify the cgroup_no_v1=list option on the kernel boot command line;
420 list is a comma-separated list of the names of the controllers to dis‐
421 able, or the word all to disable all v1 controllers. (This situation
422 is correctly handled by systemd(1), which falls back to operating with‐
423 out the specified controllers.)
424 Note that on many modern systems, systemd(1) automatically mounts the
426 cgroup2 filesystem at /sys/fs/cgroup/unified during the boot process.
427 Cgroups v2 mount options
429 The following options (mount -o) can be specified when mounting the
430 group v2 filesystem:
431 nsdelegate (since Linux 4.15)
433 Treat cgroup namespaces as delegation boundaries. For details,
434 see below.
435 memory_localevents (since Linux 5.2)
437 The memory.events should show statistics only for the cgroup
438 itself, and not for any descendant cgroups. This was the behav‐
439 ior before Linux 5.2. Starting in Linux 5.2, the default behav‐
440 ior is to include statistics for descendant cgroups in mem‐
441 ory.events, and this mount option can be used to revert to the
442 legacy behavior. This option is system wide and can be set on
443 mount or modified through remount only from the initial mount
444 namespace; it is silently ignored in noninitial namespaces.
445 Cgroups v2 controllers
447 The following controllers, documented in the kernel source file Docu‐
448 mentation/cgroup-v2.txt, are supported in cgroups version 2:
449 cpu (since Linux 4.15)
451 This is the successor to the version 1 cpu and cpuacct con‐
452 trollers.
453 cpuset (since Linux 5.0)
455 This is the successor of the version 1 cpuset controller.
456 freezer (since Linux 5.2)
458 This is the successor of the version 1 freezer controller.
459 hugetlb (since Linux 5.6)
461 This is the successor of the version 1 hugetlb controller.
462 io (since Linux 4.5)
464 This is the successor of the version 1 blkio controller.
465 memory (since Linux 4.5)
467 This is the successor of the version 1 memory controller.
468 perf_event (since Linux 4.11)
470 This is the same as the version 1 perf_event controller.
471 pids (since Linux 4.5)
473 This is the same as the version 1 pids controller.
474 rdma (since Linux 4.11)
476 This is the same as the version 1 rdma controller.
477 There is no direct equivalent of the net_cls and net_prio controllers
479 from cgroups version 1. Instead, support has been added to iptables(8)
480 to allow eBPF filters that hook on cgroup v2 pathnames to make deci‐
481 sions about network traffic on a per-cgroup basis.
482 The v2 devices controller provides no interface files; instead, device
484 control is gated by attaching an eBPF (BPF_CGROUP_DEVICE) program to a
485 v2 cgroup.
486 Cgroups v2 subtree control
488 Each cgroup in the v2 hierarchy contains the following two files:
489 cgroup.controllers
491 This read-only file exposes a list of the controllers that are
492 available in this cgroup. The contents of this file match the
493 contents of the cgroup.subtree_control file in the parent
494 cgroup.
495 cgroup.subtree_control
497 This is a list of controllers that are active (enabled) in the
498 cgroup. The set of controllers in this file is a subset of the
499 set in the cgroup.controllers of this cgroup. The set of active
500 controllers is modified by writing strings to this file contain‐
501 ing space-delimited controller names, each preceded by '+' (to
502 enable a controller) or '-' (to disable a controller), as in the
503 following example:
504 echo '+pids -memory' > x/y/cgroup.subtree_control
506 An attempt to enable a controller that is not present in
508 cgroup.controllers leads to an ENOENT error when writing to the
509 cgroup.subtree_control file.
510 Because the list of controllers in cgroup.subtree_control is a subset
512 of those cgroup.controllers, a controller that has been disabled in one
513 cgroup in the hierarchy can never be re-enabled in the subtree below
514 that cgroup.
515 A cgroup's cgroup.subtree_control file determines the set of con‐
517 trollers that are exercised in the child cgroups. When a controller
518 (e.g., pids) is present in the cgroup.subtree_control file of a parent
519 cgroup, then the corresponding controller-interface files (e.g.,
520 pids.max) are automatically created in the children of that cgroup and
521 can be used to exert resource control in the child cgroups.
522 Cgroups v2 "no internal processes" rule
524 Cgroups v2 enforces a so-called "no internal processes" rule. Roughly
525 speaking, this rule means that, with the exception of the root cgroup,
526 processes may reside only in leaf nodes (cgroups that do not themselves
527 contain child cgroups). This avoids the need to decide how to parti‐
528 tion resources between processes which are members of cgroup A and pro‐
529 cesses in child cgroups of A.
530 For instance, if cgroup /cg1/cg2 exists, then a process may reside in
532 /cg1/cg2, but not in /cg1. This is to avoid an ambiguity in cgroups v1
533 with respect to the delegation of resources between processes in /cg1
534 and its child cgroups. The recommended approach in cgroups v2 is to
535 create a subdirectory called leaf for any nonleaf cgroup which should
536 contain processes, but no child cgroups. Thus, processes which previ‐
537 ously would have gone into /cg1 would now go into /cg1/leaf. This has
538 the advantage of making explicit the relationship between processes in
539 /cg1/leaf and /cg1's other children.
540 The "no internal processes" rule is in fact more subtle than stated
542 above. More precisely, the rule is that a (nonroot) cgroup can't both
543 (1) have member processes, and (2) distribute resources into child
544 cgroups—that is, have a nonempty cgroup.subtree_control file. Thus, it
545 is possible for a cgroup to have both member processes and child
546 cgroups, but before controllers can be enabled for that cgroup, the
547 member processes must be moved out of the cgroup (e.g., perhaps into
548 the child cgroups).
549 With the Linux 4.14 addition of "thread mode" (described below), the
551 "no internal processes" rule has been relaxed in some cases.
552 Cgroups v2 cgroup.events file
554 Each nonroot cgroup in the v2 hierarchy contains a read-only file,
555 cgroup.events, whose contents are key-value pairs (delimited by newline
556 characters, with the key and value separated by spaces) providing state
557 information about the the cgroup:
558 $ cat mygrp/cgroup.events
560 populated 1
561 frozen 0
562 The following keys may appear in this file:
564 populated
566 The value of this key is either 1, if this cgroup or any of its
567 descendants has member processes, or otherwise 0.
568 frozen (since Linux 5.2)
570 The value of this key is 1 if this cgroup is currently frozen,
571 or 0 if it is not.
572 The cgroup.events file can be monitored, in order to receive notifica‐
574 tion when the value of one of its keys changes. Such monitoring can be
575 done using inotify(7), which notifies changes as IN_MODIFY events, or
576 poll(2), which notifies changes by returning the POLLPRI and POLLERR
577 bits in the revents field.
578 Cgroup v2 release notification
580 Cgroups v2 provides a new mechanism for obtaining notification when a
581 cgroup becomes empty. The cgroups v1 release_agent and
582 notify_on_release files are removed, and replaced by the populated key
583 in the cgroup.events file. This key either has the value 0, meaning
584 that the cgroup (and its descendants) contain no (nonzombie) member
585 processes, or 1, meaning that the cgroup (or one of its descendants)
586 contains member processes.
587 The cgroups v2 release-notification mechanism offers the following
589 advantages over the cgroups v1 release_agent mechanism:
590 * It allows for cheaper notification, since a single process can moni‐
592 tor multiple cgroup.events files (using the techniques described
593 earlier). By contrast, the cgroups v1 mechanism requires the
594 expense of creating a process for each notification.
595 * Notification for different cgroup subhierarchies can be delegated to
597 different processes. By contrast, the cgroups v1 mechanism allows
598 only one release agent for an entire hierarchy.
599 Cgroups v2 cgroup.stat file
601 Each cgroup in the v2 hierarchy contains a read-only cgroup.stat file
602 (first introduced in Linux 4.14) that consists of lines containing key-
603 value pairs. The following keys currently appear in this file:
604 nr_descendants
606 This is the total number of visible (i.e., living) descendant
607 cgroups underneath this cgroup.
608 nr_dying_descendants
610 This is the total number of dying descendant cgroups underneath
611 this cgroup. A cgroup enters the dying state after being
612 deleted. It remains in that state for an undefined period
613 (which will depend on system load) while resources are freed
614 before the cgroup is destroyed. Note that the presence of some
615 cgroups in the dying state is normal, and is not indicative of
616 any problem.
617 A process can't be made a member of a dying cgroup, and a dying
619 cgroup can't be brought back to life.
620 Limiting the number of descendant cgroups
622 Each cgroup in the v2 hierarchy contains the following files, which can
623 be used to view and set limits on the number of descendant cgroups
624 under that cgroup:
625 cgroup.max.depth (since Linux 4.14)
627 This file defines a limit on the depth of nesting of descendant
628 cgroups. A value of 0 in this file means that no descendant
629 cgroups can be created. An attempt to create a descendant whose
630 nesting level exceeds the limit fails (mkdir(2) fails with the
631 error EAGAIN).
632 Writing the string "max" to this file means that no limit is
634 imposed. The default value in this file is "max".
635 cgroup.max.descendants (since Linux 4.14)
637 This file defines a limit on the number of live descendant
638 cgroups that this cgroup may have. An attempt to create more
639 descendants than allowed by the limit fails (mkdir(2) fails with
640 the error EAGAIN).
641 Writing the string "max" to this file means that no limit is
643 imposed. The default value in this file is "max".
644
646 In the context of cgroups, delegation means passing management of some
647 subtree of the cgroup hierarchy to a nonprivileged user. Cgroups v1
648 provides support for delegation based on file permissions in the cgroup
649 hierarchy but with less strict containment rules than v2 (as noted
650 below). Cgroups v2 supports delegation with containment by explicit
651 design. The focus of the discussion in this section is on delegation
652 in cgroups v2, with some differences for cgroups v1 noted along the
653 way.
654 Some terminology is required in order to describe delegation. A dele‐
656 gater is a privileged user (i.e., root) who owns a parent cgroup. A
657 delegatee is a nonprivileged user who will be granted the permissions
658 needed to manage some subhierarchy under that parent cgroup, known as
659 the delegated subtree.
660 To perform delegation, the delegater makes certain directories and
662 files writable by the delegatee, typically by changing the ownership of
663 the objects to be the user ID of the delegatee. Assuming that we want
664 to delegate the hierarchy rooted at (say) /dlgt_grp and that there are
665 not yet any child cgroups under that cgroup, the ownership of the fol‐
666 lowing is changed to the user ID of the delegatee:
667 /dlgt_grp
669 Changing the ownership of the root of the subtree means that any
670 new cgroups created under the subtree (and the files they con‐
671 tain) will also be owned by the delegatee.
672 /dlgt_grp/cgroup.procs
674 Changing the ownership of this file means that the delegatee can
675 move processes into the root of the delegated subtree.
676 /dlgt_grp/cgroup.subtree_control (cgroups v2 only)
678 Changing the ownership of this file means that the delegatee can
679 enable controllers (that are present in /dlgt_grp/cgroup.con‐
680 trollers) in order to further redistribute resources at lower
681 levels in the subtree. (As an alternative to changing the own‐
682 ership of this file, the delegater might instead add selected
683 controllers to this file.)
684 /dlgt_grp/cgroup.threads (cgroups v2 only)
686 Changing the ownership of this file is necessary if a threaded
687 subtree is being delegated (see the description of "thread
688 mode", below). This permits the delegatee to write thread IDs
689 to the file. (The ownership of this file can also be changed
690 when delegating a domain subtree, but currently this serves no
691 purpose, since, as described below, it is not possible to move a
692 thread between domain cgroups by writing its thread ID to the
693 cgroup.threads file.)
694 In cgroups v1, the corresponding file that should instead be
696 delegated is the tasks file.
697 The delegater should not change the ownership of any of the controller
699 interfaces files (e.g., pids.max, memory.high) in dlgt_grp. Those
700 files are used from the next level above the delegated subtree in order
701 to distribute resources into the subtree, and the delegatee should not
702 have permission to change the resources that are distributed into the
703 delegated subtree.
704 See also the discussion of the /sys/kernel/cgroup/delegate file in
706 NOTES for information about further delegatable files in cgroups v2.
707 After the aforementioned steps have been performed, the delegatee can
709 create child cgroups within the delegated subtree (the cgroup subdirec‐
710 tories and the files they contain will be owned by the delegatee) and
711 move processes between cgroups in the subtree. If some controllers are
712 present in dlgt_grp/cgroup.subtree_control, or the ownership of that
713 file was passed to the delegatee, the delegatee can also control the
714 further redistribution of the corresponding resources into the dele‐
715 gated subtree.
716 Cgroups v2 delegation: nsdelegate and cgroup namespaces
718 Starting with Linux 4.13, there is a second way to perform cgroup dele‐
719 gation in the cgroups v2 hierarchy. This is done by mounting or
720 remounting the cgroup v2 filesystem with the nsdelegate mount option.
721 For example, if the cgroup v2 filesystem has already been mounted, we
722 can remount it with the nsdelegate option as follows:
723 mount -t cgroup2 -o remount,nsdelegate \
725 none /sys/fs/cgroup/unified
726 The effect of this mount option is to cause cgroup namespaces to auto‐
728 matically become delegation boundaries. More specifically, the follow‐
729 ing restrictions apply for processes inside the cgroup namespace:
730 * Writes to controller interface files in the root directory of the
732 namespace will fail with the error EPERM. Processes inside the
733 cgroup namespace can still write to delegatable files in the root
734 directory of the cgroup namespace such as cgroup.procs and
735 cgroup.subtree_control, and can create subhierarchy underneath the
736 root directory.
737 * Attempts to migrate processes across the namespace boundary are
739 denied (with the error ENOENT). Processes inside the cgroup names‐
740 pace can still (subject to the containment rules described below)
741 move processes between cgroups within the subhierarchy under the
742 namespace root.
743 The ability to define cgroup namespaces as delegation boundaries makes
745 cgroup namespaces more useful. To understand why, suppose that we
746 already have one cgroup hierarchy that has been delegated to a nonpriv‐
747 ileged user, cecilia, using the older delegation technique described
748 above. Suppose further that cecilia wanted to further delegate a sub‐
749 hierarchy under the existing delegated hierarchy. (For example, the
750 delegated hierarchy might be associated with an unprivileged container
751 run by cecilia.) Even if a cgroup namespace was employed, because both
752 hierarchies are owned by the unprivileged user cecilia, the following
753 illegitimate actions could be performed:
754 * A process in the inferior hierarchy could change the resource con‐
756 troller settings in the root directory of that hierarchy. (These
757 resource controller settings are intended to allow control to be
758 exercised from the parent cgroup; a process inside the child cgroup
759 should not be allowed to modify them.)
760 * A process inside the inferior hierarchy could move processes into
762 and out of the inferior hierarchy if the cgroups in the superior
763 hierarchy were somehow visible.
764 Employing the nsdelegate mount option prevents both of these possibili‐
766 ties.
767 The nsdelegate mount option only has an effect when performed in the
769 initial mount namespace; in other mount namespaces, the option is
770 silently ignored.
771 Note: On some systems, systemd(1) automatically mounts the cgroup v2
773 filesystem. In order to experiment with the nsdelegate operation, it
774 may be useful to boot the kernel with the following command-line
775 options:
776 cgroup_no_v1=all systemd.legacy_systemd_cgroup_controller
778 These options cause the kernel to boot with the cgroups v1 controllers
780 disabled (meaning that the controllers are available in the v2 hierar‐
781 chy), and tells systemd(1) not to mount and use the cgroup v2 hierar‐
782 chy, so that the v2 hierarchy can be manually mounted with the desired
783 options after boot-up.
784 Cgroup delegation containment rules
786 Some delegation containment rules ensure that the delegatee can move
787 processes between cgroups within the delegated subtree, but can't move
788 processes from outside the delegated subtree into the subtree or vice
789 versa. A nonprivileged process (i.e., the delegatee) can write the PID
790 of a "target" process into a cgroup.procs file only if all of the fol‐
791 lowing are true:
792 * The writer has write permission on the cgroup.procs file in the des‐
794 tination cgroup.
795 * The writer has write permission on the cgroup.procs file in the
797 nearest common ancestor of the source and destination cgroups. Note
798 that in some cases, the nearest common ancestor may be the source or
799 destination cgroup itself. This requirement is not enforced for
800 cgroups v1 hierarchies, with the consequence that containment in v1
801 is less strict than in v2. (For example, in cgroups v1 the user
802 that owns two distinct delegated subhierarchies can move a process
803 between the hierarchies.)
804 * If the cgroup v2 filesystem was mounted with the nsdelegate option,
806 the writer must be able to see the source and destination cgroups
807 from its cgroup namespace.
808 * In cgroups v1: the effective UID of the writer (i.e., the delegatee)
810 matches the real user ID or the saved set-user-ID of the target
811 process. Before Linux 4.11, this requirement also applied in
812 cgroups v2 (This was a historical requirement inherited from cgroups
813 v1 that was later deemed unnecessary, since the other rules suffice
814 for containment in cgroups v2.)
815 Note: one consequence of these delegation containment rules is that the
817 unprivileged delegatee can't place the first process into the delegated
818 subtree; instead, the delegater must place the first process (a process
819 owned by the delegatee) into the delegated subtree.
820
822 Among the restrictions imposed by cgroups v2 that were not present in
823 cgroups v1 are the following:
824 * No thread-granularity control: all of the threads of a process must
826 be in the same cgroup.
827 * No internal processes: a cgroup can't both have member processes and
829 exercise controllers on child cgroups.
830 Both of these restrictions were added because the lack of these
832 restrictions had caused problems in cgroups v1. In particular, the
833 cgroups v1 ability to allow thread-level granularity for cgroup member‐
834 ship made no sense for some controllers. (A notable example was the
835 memory controller: since threads share an address space, it made no
836 sense to split threads across different memory cgroups.)
837 Notwithstanding the initial design decision in cgroups v2, there were
839 use cases for certain controllers, notably the cpu controller, for
840 which thread-level granularity of control was meaningful and useful.
841 To accommodate such use cases, Linux 4.14 added thread mode for cgroups
842 v2.
843 Thread mode allows the following:
845 * The creation of threaded subtrees in which the threads of a process
847 may be spread across cgroups inside the tree. (A threaded subtree
848 may contain multiple multithreaded processes.)
849 * The concept of threaded controllers, which can distribute resources
851 across the cgroups in a threaded subtree.
852 * A relaxation of the "no internal processes rule", so that, within a
854 threaded subtree, a cgroup can both contain member threads and exer‐
855 cise resource control over child cgroups.
856 With the addition of thread mode, each nonroot cgroup now contains a
858 new file, cgroup.type, that exposes, and in some circumstances can be
859 used to change, the "type" of a cgroup. This file contains one of the
860 following type values:
861 domain This is a normal v2 cgroup that provides process-granularity
863 control. If a process is a member of this cgroup, then all
864 threads of the process are (by definition) in the same cgroup.
865 This is the default cgroup type, and provides the same behavior
866 that was provided for cgroups in the initial cgroups v2 imple‐
867 mentation.
868 threaded
870 This cgroup is a member of a threaded subtree. Threads can be
871 added to this cgroup, and controllers can be enabled for the
872 cgroup.
873 domain threaded
875 This is a domain cgroup that serves as the root of a threaded
876 subtree. This cgroup type is also known as "threaded root".
877 domain invalid
879 This is a cgroup inside a threaded subtree that is in an
880 "invalid" state. Processes can't be added to the cgroup, and
881 controllers can't be enabled for the cgroup. The only thing
882 that can be done with this cgroup (other than deleting it) is to
883 convert it to a threaded cgroup by writing the string "threaded"
884 to the cgroup.type file.
885 The rationale for the existence of this "interim" type during
887 the creation of a threaded subtree (rather than the kernel sim‐
888 ply immediately converting all cgroups under the threaded root
889 to the type threaded) is to allow for possible future extensions
890 to the thread mode model
891 Threaded versus domain controllers
893 With the addition of threads mode, cgroups v2 now distinguishes two
894 types of resource controllers:
895 * Threaded controllers: these controllers support thread-granularity
897 for resource control and can be enabled inside threaded subtrees,
898 with the result that the corresponding controller-interface files
899 appear inside the cgroups in the threaded subtree. As at Linux
900 4.19, the following controllers are threaded: cpu, perf_event, and
901 pids.
902 * Domain controllers: these controllers support only process granular‐
904 ity for resource control. From the perspective of a domain con‐
905 troller, all threads of a process are always in the same cgroup.
906 Domain controllers can't be enabled inside a threaded subtree.
907 Creating a threaded subtree
909 There are two pathways that lead to the creation of a threaded subtree.
910 The first pathway proceeds as follows:
911 1. We write the string "threaded" to the cgroup.type file of a cgroup
913 y/z that currently has the type domain. This has the following
914 effects:
915 * The type of the cgroup y/z becomes threaded.
917 * The type of the parent cgroup, y, becomes domain threaded. The
919 parent cgroup is the root of a threaded subtree (also known as
920 the "threaded root").
921 * All other cgroups under y that were not already of type threaded
923 (because they were inside already existing threaded subtrees
924 under the new threaded root) are converted to type domain
925 invalid. Any subsequently created cgroups under y will also have
926 the type domain invalid.
927 2. We write the string "threaded" to each of the domain invalid cgroups
929 under y, in order to convert them to the type threaded. As a conse‐
930 quence of this step, all threads under the threaded root now have
931 the type threaded and the threaded subtree is now fully usable. The
932 requirement to write "threaded" to each of these cgroups is somewhat
933 cumbersome, but allows for possible future extensions to the thread-
934 mode model.
935 The second way of creating a threaded subtree is as follows:
937 1. In an existing cgroup, z, that currently has the type domain, we (1)
939 enable one or more threaded controllers and (2) make a process a
940 member of z. (These two steps can be done in either order.) This
941 has the following consequences:
942 * The type of z becomes domain threaded.
944 * All of the descendant cgroups of x that were not already of type
946 threaded are converted to type domain invalid.
947 2. As before, we make the threaded subtree usable by writing the string
949 "threaded" to each of the domain invalid cgroups under y, in order
950 to convert them to the type threaded.
951 One of the consequences of the above pathways to creating a threaded
953 subtree is that the threaded root cgroup can be a parent only to
954 threaded (and domain invalid) cgroups. The threaded root cgroup can't
955 be a parent of a domain cgroups, and a threaded cgroup can't have a
956 sibling that is a domain cgroup.
957 Using a threaded subtree
959 Within a threaded subtree, threaded controllers can be enabled in each
960 subgroup whose type has been changed to threaded; upon doing so, the
961 corresponding controller interface files appear in the children of that
962 cgroup.
963 A process can be moved into a threaded subtree by writing its PID to
965 the cgroup.procs file in one of the cgroups inside the tree. This has
966 the effect of making all of the threads in the process members of the
967 corresponding cgroup and makes the process a member of the threaded
968 subtree. The threads of the process can then be spread across the
969 threaded subtree by writing their thread IDs (see gettid(2)) to the
970 cgroup.threads files in different cgroups inside the subtree. The
971 threads of a process must all reside in the same threaded subtree.
972 As with writing to cgroup.procs, some containment rules apply when
974 writing to the cgroup.threads file:
975 * The writer must have write permission on the cgroup.threads file in
977 the destination cgroup.
978 * The writer must have write permission on the cgroup.procs file in
980 the common ancestor of the source and destination cgroups. (In some
981 cases, the common ancestor may be the source or destination cgroup
982 itself.)
983 * The source and destination cgroups must be in the same threaded sub‐
985 tree. (Outside a threaded subtree, an attempt to move a thread by
986 writing its thread ID to the cgroup.threads file in a different
987 domain cgroup fails with the error EOPNOTSUPP.)
988 The cgroup.threads file is present in each cgroup (including domain
990 cgroups) and can be read in order to discover the set of threads that
991 is present in the cgroup. The set of thread IDs obtained when reading
992 this file is not guaranteed to be ordered or free of duplicates.
993 The cgroup.procs file in the threaded root shows the PIDs of all pro‐
995 cesses that are members of the threaded subtree. The cgroup.procs
996 files in the other cgroups in the subtree are not readable.
997 Domain controllers can't be enabled in a threaded subtree; no con‐
999 troller-interface files appear inside the cgroups underneath the
1000 threaded root. From the point of view of a domain controller, threaded
1001 subtrees are invisible: a multithreaded process inside a threaded sub‐
1002 tree appears to a domain controller as a process that resides in the
1003 threaded root cgroup.
1004 Within a threaded subtree, the "no internal processes" rule does not
1006 apply: a cgroup can both contain member processes (or thread) and exer‐
1007 cise controllers on child cgroups.
1008 Rules for writing to cgroup.type and creating threaded subtrees
1010 A number of rules apply when writing to the cgroup.type file:
1011 * Only the string "threaded" may be written. In other words, the only
1013 explicit transition that is possible is to convert a domain cgroup
1014 to type threaded.
1015 * The effect of writing "threaded" depends on the current value in
1017 cgroup.type, as follows:
1018 · domain or domain threaded: start the creation of a threaded sub‐
1020 tree (whose root is the parent of this cgroup) via the first of
1021 the pathways described above;
1022 · domain invalid: convert this cgroup (which is inside a threaded
1024 subtree) to a usable (i.e., threaded) state;
1025 · threaded: no effect (a "no-op").
1027 * We can't write to a cgroup.type file if the parent's type is domain
1029 invalid. In other words, the cgroups of a threaded subtree must be
1030 converted to the threaded state in a top-down manner.
1031 There are also some constraints that must be satisfied in order to cre‐
1033 ate a threaded subtree rooted at the cgroup x:
1034 * There can be no member processes in the descendant cgroups of x.
1036 (The cgroup x can itself have member processes.)
1037 * No domain controllers may be enabled in x's cgroup.subtree_control
1039 file.
1040 If any of the above constraints is violated, then an attempt to write
1042 "threaded" to a cgroup.type file fails with the error ENOTSUP.
1043 The "domain threaded" cgroup type
1045 According to the pathways described above, the type of a cgroup can
1046 change to domain threaded in either of the following cases:
1047 * The string "threaded" is written to a child cgroup.
1049 * A threaded controller is enabled inside the cgroup and a process is
1051 made a member of the cgroup.
1052 A domain threaded cgroup, x, can revert to the type domain if the above
1054 conditions no longer hold true—that is, if all threaded child cgroups
1055 of x are removed and either x no longer has threaded controllers
1056 enabled or no longer has member processes.
1057 When a domain threaded cgroup x reverts to the type domain:
1059 * All domain invalid descendants of x that are not in lower-level
1061 threaded subtrees revert to the type domain.
1062 * The root cgroups in any lower-level threaded subtrees revert to the
1064 type domain threaded.
1065 Exceptions for the root cgroup
1067 The root cgroup of the v2 hierarchy is treated exceptionally: it can be
1068 the parent of both domain and threaded cgroups. If the string
1069 "threaded" is written to the cgroup.type file of one of the children of
1070 the root cgroup, then
1071 * The type of that cgroup becomes threaded.
1073 * The type of any descendants of that cgroup that are not part of
1075 lower-level threaded subtrees changes to domain invalid.
1076 Note that in this case, there is no cgroup whose type becomes domain
1078 threaded. (Notionally, the root cgroup can be considered as the
1079 threaded root for the cgroup whose type was changed to threaded.)
1080 The aim of this exceptional treatment for the root cgroup is to allow a
1082 threaded cgroup that employs the cpu controller to be placed as high as
1083 possible in the hierarchy, so as to minimize the (small) cost of
1084 traversing the cgroup hierarchy.
1085 The cgroups v2 "cpu" controller and realtime threads
1087 As at Linux 4.19, the cgroups v2 cpu controller does not support con‐
1088 trol of realtime threads (specifically threads scheduled under any of
1089 the policies SCHED_FIFO, SCHED_RR, described SCHED_DEADLINE; see
1090 sched(7)). Therefore, the cpu controller can be enabled in the root
1091 cgroup only if all realtime threads are in the root cgroup. (If there
1092 are realtime threads in nonroot cgroups, then a write(2) of the string
1093 "+cpu" to the cgroup.subtree_control file fails with the error EINVAL.)
1094 On some systems, systemd(1) places certain realtime threads in nonroot
1096 cgroups in the v2 hierarchy. On such systems, these threads must first
1097 be moved to the root cgroup before the cpu controller can be enabled.
1098
1100 The following errors can occur for mount(2):
1101 EBUSY An attempt to mount a cgroup version 1 filesystem specified nei‐
1103 ther the name= option (to mount a named hierarchy) nor a con‐
1104 troller name (or all).
1105
1107 A child process created via fork(2) inherits its parent's cgroup mem‐
1108 berships. A process's cgroup memberships are preserved across
1109 execve(2).
1110 The clone3(2) CLONE_INTO_CGROUP flag can be used to create a child
1112 process that begins its life in a different version 2 cgroup from the
1113 parent process.
1114 /proc files
1116 /proc/cgroups (since Linux 2.6.24)
1117 This file contains information about the controllers that are
1118 compiled into the kernel. An example of the contents of this
1119 file (reformatted for readability) is the following:
1120 #subsys_name hierarchy num_cgroups enabled
1122 cpuset 4 1 1
1123 cpu 8 1 1
1124 cpuacct 8 1 1
1125 blkio 6 1 1
1126 memory 3 1 1
1127 devices 10 84 1
1128 freezer 7 1 1
1129 net_cls 9 1 1
1130 perf_event 5 1 1
1131 net_prio 9 1 1
1132 hugetlb 0 1 0
1133 pids 2 1 1
1134 The fields in this file are, from left to right:
1136 1. The name of the controller.
1138 2. The unique ID of the cgroup hierarchy on which this con‐
1140 troller is mounted. If multiple cgroups v1 controllers are
1141 bound to the same hierarchy, then each will show the same
1142 hierarchy ID in this field. The value in this field will be
1143 0 if:
1144 a) the controller is not mounted on a cgroups v1 hierarchy;
1146 b) the controller is bound to the cgroups v2 single unified
1148 hierarchy; or
1149 c) the controller is disabled (see below).
1151 3. The number of control groups in this hierarchy using this
1153 controller.
1154 4. This field contains the value 1 if this controller is
1156 enabled, or 0 if it has been disabled (via the cgroup_disable
1157 kernel command-line boot parameter).
1158 /proc/[pid]/cgroup (since Linux 2.6.24)
1160 This file describes control groups to which the process with the
1161 corresponding PID belongs. The displayed information differs
1162 for cgroups version 1 and version 2 hierarchies.
1163 For each cgroup hierarchy of which the process is a member,
1165 there is one entry containing three colon-separated fields:
1166 hierarchy-ID:controller-list:cgroup-path
1168 For example:
1170 5:cpuacct,cpu,cpuset:/daemons
1172 The colon-separated fields are, from left to right:
1174 1. For cgroups version 1 hierarchies, this field contains a
1176 unique hierarchy ID number that can be matched to a hierarchy
1177 ID in /proc/cgroups. For the cgroups version 2 hierarchy,
1178 this field contains the value 0.
1179 2. For cgroups version 1 hierarchies, this field contains a
1181 comma-separated list of the controllers bound to the hierar‐
1182 chy. For the cgroups version 2 hierarchy, this field is
1183 empty.
1184 3. This field contains the pathname of the control group in the
1186 hierarchy to which the process belongs. This pathname is
1187 relative to the mount point of the hierarchy.
1188 /sys/kernel/cgroup files
1190 /sys/kernel/cgroup/delegate (since Linux 4.15)
1191 This file exports a list of the cgroups v2 files (one per line)
1192 that are delegatable (i.e., whose ownership should be changed to
1193 the user ID of the delegatee). In the future, the set of dele‐
1194 gatable files may change or grow, and this file provides a way
1195 for the kernel to inform user-space applications of which files
1196 must be delegated. As at Linux 4.15, one sees the following
1197 when inspecting this file:
1198 $ cat /sys/kernel/cgroup/delegate
1200 cgroup.procs
1201 cgroup.subtree_control
1202 cgroup.threads
1203 /sys/kernel/cgroup/features (since Linux 4.15)
1205 Over time, the set of cgroups v2 features that are provided by
1206 the kernel may change or grow, or some features may not be
1207 enabled by default. This file provides a way for user-space
1208 applications to discover what features the running kernel sup‐
1209 ports and has enabled. Features are listed one per line:
1210 $ cat /sys/kernel/cgroup/features
1212 nsdelegate
1213 memory_localevents
1214 The entries that can appear in this file are:
1216 memory_localevents (since Linux 5.2)
1218 The kernel supports the memory_localevents mount option.
1219 nsdelegate (since Linux 4.15)
1221 The kernel supports the nsdelegate mount option.
1222
1224 prlimit(1), systemd(1), systemd-cgls(1), systemd-cgtop(1), clone(2),
1225 ioprio_set(2), perf_event_open(2), setrlimit(2), cgroup_namespaces(7),
1226 cpuset(7), namespaces(7), sched(7), user_namespaces(7)
1227 The kernel source file Documentation/admin-guide/cgroup-v2.rst.
1229
1231 This page is part of release 5.07 of the Linux man-pages project. A
1232 description of the project, information about reporting bugs, and the
1233 latest version of this page, can be found at
1234 https://www.kernel.org/doc/man-pages/.
1235Linux 2020-04-11 CGROUPS(7)