8 :Author: Tejun Heo <tj@kernel.org>
10 This is the authoritative documentation on the design, interface and
11 conventions of cgroup v2. It describes all userland-visible aspects
12 of cgroup including core and specific controller behaviors. All
13 future changes must be reflected in this document. Documentation for
14 v1 is available under :ref:`Documentation/admin-guide/cgroup-v1/index.rst <cgroup-v1>`.
23 2-2. Organizing Processes and Threads
26 2-3. [Un]populated Notification
27 2-4. Controlling Controllers
28 2-4-1. Enabling and Disabling
29 2-4-2. Top-down Constraint
30 2-4-3. No Internal Process Constraint
32 2-5-1. Model of Delegation
33 2-5-2. Delegation Containment
35 2-6-1. Organize Once and Control
36 2-6-2. Avoid Name Collisions
37 3. Resource Distribution Models
45 4-3. Core Interface Files
48 5-1-1. CPU Interface Files
50 5-2-1. Memory Interface Files
51 5-2-2. Usage Guidelines
52 5-2-3. Memory Ownership
54 5-3-1. IO Interface Files
57 5-3-3-1. How IO Latency Throttling Works
58 5-3-3-2. IO Latency Interface Files
61 5-4-1. PID Interface Files
63 5.5-1. Cpuset Interface Files
66 5-7-1. RDMA Interface Files
68 5.8-1. HugeTLB Interface Files
70 5.9-1 Miscellaneous cgroup Interface Files
71 5.9-2 Migration and Ownership
74 5-N. Non-normative information
75 5-N-1. CPU controller root cgroup process behaviour
76 5-N-2. IO controller root cgroup process behaviour
79 6-2. The Root and Views
80 6-3. Migration and setns(2)
81 6-4. Interaction with Other Namespaces
82 P. Information on Kernel Programming
83 P-1. Filesystem Support for Writeback
84 D. Deprecated v1 Core Features
85 R. Issues with v1 and Rationales for v2
86 R-1. Multiple Hierarchies
87 R-2. Thread Granularity
88 R-3. Competition Between Inner Nodes and Threads
89 R-4. Other Interface Issues
90 R-5. Controller Issues and Remedies
100 "cgroup" stands for "control group" and is never capitalized. The
101 singular form is used to designate the whole feature and also as a
102 qualifier as in "cgroup controllers". When explicitly referring to
103 multiple individual control groups, the plural form "cgroups" is used.
109 cgroup is a mechanism to organize processes hierarchically and
110 distribute system resources along the hierarchy in a controlled and
113 cgroup is largely composed of two parts - the core and controllers.
114 cgroup core is primarily responsible for hierarchically organizing
115 processes. A cgroup controller is usually responsible for
116 distributing a specific type of system resource along the hierarchy
117 although there are utility controllers which serve purposes other than
118 resource distribution.
120 cgroups form a tree structure and every process in the system belongs
121 to one and only one cgroup. All threads of a process belong to the
122 same cgroup. On creation, all processes are put in the cgroup that
123 the parent process belongs to at the time. A process can be migrated
124 to another cgroup. Migration of a process doesn't affect already
125 existing descendant processes.
127 Following certain structural constraints, controllers may be enabled or
128 disabled selectively on a cgroup. All controller behaviors are
129 hierarchical - if a controller is enabled on a cgroup, it affects all
130 processes which belong to the cgroups consisting the inclusive
131 sub-hierarchy of the cgroup. When a controller is enabled on a nested
132 cgroup, it always restricts the resource distribution further. The
133 restrictions set closer to the root in the hierarchy can not be
134 overridden from further away.
143 Unlike v1, cgroup v2 has only single hierarchy. The cgroup v2
144 hierarchy can be mounted with the following mount command::
146 # mount -t cgroup2 none $MOUNT_POINT
148 cgroup2 filesystem has the magic number 0x63677270 ("cgrp"). All
149 controllers which support v2 and are not bound to a v1 hierarchy are
150 automatically bound to the v2 hierarchy and show up at the root.
151 Controllers which are not in active use in the v2 hierarchy can be
152 bound to other hierarchies. This allows mixing v2 hierarchy with the
153 legacy v1 multiple hierarchies in a fully backward compatible way.
155 A controller can be moved across hierarchies only after the controller
156 is no longer referenced in its current hierarchy. Because per-cgroup
157 controller states are destroyed asynchronously and controllers may
158 have lingering references, a controller may not show up immediately on
159 the v2 hierarchy after the final umount of the previous hierarchy.
160 Similarly, a controller should be fully disabled to be moved out of
161 the unified hierarchy and it may take some time for the disabled
162 controller to become available for other hierarchies; furthermore, due
163 to inter-controller dependencies, other controllers may need to be
166 While useful for development and manual configurations, moving
167 controllers dynamically between the v2 and other hierarchies is
168 strongly discouraged for production use. It is recommended to decide
169 the hierarchies and controller associations before starting using the
170 controllers after system boot.
172 During transition to v2, system management software might still
173 automount the v1 cgroup filesystem and so hijack all controllers
174 during boot, before manual intervention is possible. To make testing
175 and experimenting easier, the kernel parameter cgroup_no_v1= allows
176 disabling controllers in v1 and make them always available in v2.
178 cgroup v2 currently supports the following mount options.
181 Consider cgroup namespaces as delegation boundaries. This
182 option is system wide and can only be set on mount or modified
183 through remount from the init namespace. The mount option is
184 ignored on non-init namespace mounts. Please refer to the
185 Delegation section for details.
188 Only populate memory.events with data for the current cgroup,
189 and not any subtrees. This is legacy behaviour, the default
190 behaviour without this option is to include subtree counts.
191 This option is system wide and can only be set on mount or
192 modified through remount from the init namespace. The mount
193 option is ignored on non-init namespace mounts.
196 Recursively apply memory.min and memory.low protection to
197 entire subtrees, without requiring explicit downward
198 propagation into leaf cgroups. This allows protecting entire
199 subtrees from one another, while retaining free competition
200 within those subtrees. This should have been the default
201 behavior but is a mount-option to avoid regressing setups
202 relying on the original semantics (e.g. specifying bogusly
203 high 'bypass' protection values at higher tree levels).
206 Organizing Processes and Threads
207 --------------------------------
212 Initially, only the root cgroup exists to which all processes belong.
213 A child cgroup can be created by creating a sub-directory::
217 A given cgroup may have multiple child cgroups forming a tree
218 structure. Each cgroup has a read-writable interface file
219 "cgroup.procs". When read, it lists the PIDs of all processes which
220 belong to the cgroup one-per-line. The PIDs are not ordered and the
221 same PID may show up more than once if the process got moved to
222 another cgroup and then back or the PID got recycled while reading.
224 A process can be migrated into a cgroup by writing its PID to the
225 target cgroup's "cgroup.procs" file. Only one process can be migrated
226 on a single write(2) call. If a process is composed of multiple
227 threads, writing the PID of any thread migrates all threads of the
230 When a process forks a child process, the new process is born into the
231 cgroup that the forking process belongs to at the time of the
232 operation. After exit, a process stays associated with the cgroup
233 that it belonged to at the time of exit until it's reaped; however, a
234 zombie process does not appear in "cgroup.procs" and thus can't be
235 moved to another cgroup.
237 A cgroup which doesn't have any children or live processes can be
238 destroyed by removing the directory. Note that a cgroup which doesn't
239 have any children and is associated only with zombie processes is
240 considered empty and can be removed::
244 "/proc/$PID/cgroup" lists a process's cgroup membership. If legacy
245 cgroup is in use in the system, this file may contain multiple lines,
246 one for each hierarchy. The entry for cgroup v2 is always in the
249 # cat /proc/842/cgroup
251 0::/test-cgroup/test-cgroup-nested
253 If the process becomes a zombie and the cgroup it was associated with
254 is removed subsequently, " (deleted)" is appended to the path::
256 # cat /proc/842/cgroup
258 0::/test-cgroup/test-cgroup-nested (deleted)
264 cgroup v2 supports thread granularity for a subset of controllers to
265 support use cases requiring hierarchical resource distribution across
266 the threads of a group of processes. By default, all threads of a
267 process belong to the same cgroup, which also serves as the resource
268 domain to host resource consumptions which are not specific to a
269 process or thread. The thread mode allows threads to be spread across
270 a subtree while still maintaining the common resource domain for them.
272 Controllers which support thread mode are called threaded controllers.
273 The ones which don't are called domain controllers.
275 Marking a cgroup threaded makes it join the resource domain of its
276 parent as a threaded cgroup. The parent may be another threaded
277 cgroup whose resource domain is further up in the hierarchy. The root
278 of a threaded subtree, that is, the nearest ancestor which is not
279 threaded, is called threaded domain or thread root interchangeably and
280 serves as the resource domain for the entire subtree.
282 Inside a threaded subtree, threads of a process can be put in
283 different cgroups and are not subject to the no internal process
284 constraint - threaded controllers can be enabled on non-leaf cgroups
285 whether they have threads in them or not.
287 As the threaded domain cgroup hosts all the domain resource
288 consumptions of the subtree, it is considered to have internal
289 resource consumptions whether there are processes in it or not and
290 can't have populated child cgroups which aren't threaded. Because the
291 root cgroup is not subject to no internal process constraint, it can
292 serve both as a threaded domain and a parent to domain cgroups.
294 The current operation mode or type of the cgroup is shown in the
295 "cgroup.type" file which indicates whether the cgroup is a normal
296 domain, a domain which is serving as the domain of a threaded subtree,
297 or a threaded cgroup.
299 On creation, a cgroup is always a domain cgroup and can be made
300 threaded by writing "threaded" to the "cgroup.type" file. The
301 operation is single direction::
303 # echo threaded > cgroup.type
305 Once threaded, the cgroup can't be made a domain again. To enable the
306 thread mode, the following conditions must be met.
308 - As the cgroup will join the parent's resource domain. The parent
309 must either be a valid (threaded) domain or a threaded cgroup.
311 - When the parent is an unthreaded domain, it must not have any domain
312 controllers enabled or populated domain children. The root is
313 exempt from this requirement.
315 Topology-wise, a cgroup can be in an invalid state. Please consider
316 the following topology::
318 A (threaded domain) - B (threaded) - C (domain, just created)
320 C is created as a domain but isn't connected to a parent which can
321 host child domains. C can't be used until it is turned into a
322 threaded cgroup. "cgroup.type" file will report "domain (invalid)" in
323 these cases. Operations which fail due to invalid topology use
324 EOPNOTSUPP as the errno.
326 A domain cgroup is turned into a threaded domain when one of its child
327 cgroup becomes threaded or threaded controllers are enabled in the
328 "cgroup.subtree_control" file while there are processes in the cgroup.
329 A threaded domain reverts to a normal domain when the conditions
332 When read, "cgroup.threads" contains the list of the thread IDs of all
333 threads in the cgroup. Except that the operations are per-thread
334 instead of per-process, "cgroup.threads" has the same format and
335 behaves the same way as "cgroup.procs". While "cgroup.threads" can be
336 written to in any cgroup, as it can only move threads inside the same
337 threaded domain, its operations are confined inside each threaded
340 The threaded domain cgroup serves as the resource domain for the whole
341 subtree, and, while the threads can be scattered across the subtree,
342 all the processes are considered to be in the threaded domain cgroup.
343 "cgroup.procs" in a threaded domain cgroup contains the PIDs of all
344 processes in the subtree and is not readable in the subtree proper.
345 However, "cgroup.procs" can be written to from anywhere in the subtree
346 to migrate all threads of the matching process to the cgroup.
348 Only threaded controllers can be enabled in a threaded subtree. When
349 a threaded controller is enabled inside a threaded subtree, it only
350 accounts for and controls resource consumptions associated with the
351 threads in the cgroup and its descendants. All consumptions which
352 aren't tied to a specific thread belong to the threaded domain cgroup.
354 Because a threaded subtree is exempt from no internal process
355 constraint, a threaded controller must be able to handle competition
356 between threads in a non-leaf cgroup and its child cgroups. Each
357 threaded controller defines how such competitions are handled.
360 [Un]populated Notification
361 --------------------------
363 Each non-root cgroup has a "cgroup.events" file which contains
364 "populated" field indicating whether the cgroup's sub-hierarchy has
365 live processes in it. Its value is 0 if there is no live process in
366 the cgroup and its descendants; otherwise, 1. poll and [id]notify
367 events are triggered when the value changes. This can be used, for
368 example, to start a clean-up operation after all processes of a given
369 sub-hierarchy have exited. The populated state updates and
370 notifications are recursive. Consider the following sub-hierarchy
371 where the numbers in the parentheses represent the numbers of processes
377 A, B and C's "populated" fields would be 1 while D's 0. After the one
378 process in C exits, B and C's "populated" fields would flip to "0" and
379 file modified events will be generated on the "cgroup.events" files of
383 Controlling Controllers
384 -----------------------
386 Enabling and Disabling
387 ~~~~~~~~~~~~~~~~~~~~~~
389 Each cgroup has a "cgroup.controllers" file which lists all
390 controllers available for the cgroup to enable::
392 # cat cgroup.controllers
395 No controller is enabled by default. Controllers can be enabled and
396 disabled by writing to the "cgroup.subtree_control" file::
398 # echo "+cpu +memory -io" > cgroup.subtree_control
400 Only controllers which are listed in "cgroup.controllers" can be
401 enabled. When multiple operations are specified as above, either they
402 all succeed or fail. If multiple operations on the same controller
403 are specified, the last one is effective.
405 Enabling a controller in a cgroup indicates that the distribution of
406 the target resource across its immediate children will be controlled.
407 Consider the following sub-hierarchy. The enabled controllers are
408 listed in parentheses::
410 A(cpu,memory) - B(memory) - C()
413 As A has "cpu" and "memory" enabled, A will control the distribution
414 of CPU cycles and memory to its children, in this case, B. As B has
415 "memory" enabled but not "CPU", C and D will compete freely on CPU
416 cycles but their division of memory available to B will be controlled.
418 As a controller regulates the distribution of the target resource to
419 the cgroup's children, enabling it creates the controller's interface
420 files in the child cgroups. In the above example, enabling "cpu" on B
421 would create the "cpu." prefixed controller interface files in C and
422 D. Likewise, disabling "memory" from B would remove the "memory."
423 prefixed controller interface files from C and D. This means that the
424 controller interface files - anything which doesn't start with
425 "cgroup." are owned by the parent rather than the cgroup itself.
431 Resources are distributed top-down and a cgroup can further distribute
432 a resource only if the resource has been distributed to it from the
433 parent. This means that all non-root "cgroup.subtree_control" files
434 can only contain controllers which are enabled in the parent's
435 "cgroup.subtree_control" file. A controller can be enabled only if
436 the parent has the controller enabled and a controller can't be
437 disabled if one or more children have it enabled.
440 No Internal Process Constraint
441 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
443 Non-root cgroups can distribute domain resources to their children
444 only when they don't have any processes of their own. In other words,
445 only domain cgroups which don't contain any processes can have domain
446 controllers enabled in their "cgroup.subtree_control" files.
448 This guarantees that, when a domain controller is looking at the part
449 of the hierarchy which has it enabled, processes are always only on
450 the leaves. This rules out situations where child cgroups compete
451 against internal processes of the parent.
453 The root cgroup is exempt from this restriction. Root contains
454 processes and anonymous resource consumption which can't be associated
455 with any other cgroups and requires special treatment from most
456 controllers. How resource consumption in the root cgroup is governed
457 is up to each controller (for more information on this topic please
458 refer to the Non-normative information section in the Controllers
461 Note that the restriction doesn't get in the way if there is no
462 enabled controller in the cgroup's "cgroup.subtree_control". This is
463 important as otherwise it wouldn't be possible to create children of a
464 populated cgroup. To control resource distribution of a cgroup, the
465 cgroup must create children and transfer all its processes to the
466 children before enabling controllers in its "cgroup.subtree_control"
476 A cgroup can be delegated in two ways. First, to a less privileged
477 user by granting write access of the directory and its "cgroup.procs",
478 "cgroup.threads" and "cgroup.subtree_control" files to the user.
479 Second, if the "nsdelegate" mount option is set, automatically to a
480 cgroup namespace on namespace creation.
482 Because the resource control interface files in a given directory
483 control the distribution of the parent's resources, the delegatee
484 shouldn't be allowed to write to them. For the first method, this is
485 achieved by not granting access to these files. For the second, the
486 kernel rejects writes to all files other than "cgroup.procs" and
487 "cgroup.subtree_control" on a namespace root from inside the
490 The end results are equivalent for both delegation types. Once
491 delegated, the user can build sub-hierarchy under the directory,
492 organize processes inside it as it sees fit and further distribute the
493 resources it received from the parent. The limits and other settings
494 of all resource controllers are hierarchical and regardless of what
495 happens in the delegated sub-hierarchy, nothing can escape the
496 resource restrictions imposed by the parent.
498 Currently, cgroup doesn't impose any restrictions on the number of
499 cgroups in or nesting depth of a delegated sub-hierarchy; however,
500 this may be limited explicitly in the future.
503 Delegation Containment
504 ~~~~~~~~~~~~~~~~~~~~~~
506 A delegated sub-hierarchy is contained in the sense that processes
507 can't be moved into or out of the sub-hierarchy by the delegatee.
509 For delegations to a less privileged user, this is achieved by
510 requiring the following conditions for a process with a non-root euid
511 to migrate a target process into a cgroup by writing its PID to the
514 - The writer must have write access to the "cgroup.procs" file.
516 - The writer must have write access to the "cgroup.procs" file of the
517 common ancestor of the source and destination cgroups.
519 The above two constraints ensure that while a delegatee may migrate
520 processes around freely in the delegated sub-hierarchy it can't pull
521 in from or push out to outside the sub-hierarchy.
523 For an example, let's assume cgroups C0 and C1 have been delegated to
524 user U0 who created C00, C01 under C0 and C10 under C1 as follows and
525 all processes under C0 and C1 belong to U0::
527 ~~~~~~~~~~~~~ - C0 - C00
530 ~~~~~~~~~~~~~ - C1 - C10
532 Let's also say U0 wants to write the PID of a process which is
533 currently in C10 into "C00/cgroup.procs". U0 has write access to the
534 file; however, the common ancestor of the source cgroup C10 and the
535 destination cgroup C00 is above the points of delegation and U0 would
536 not have write access to its "cgroup.procs" files and thus the write
537 will be denied with -EACCES.
539 For delegations to namespaces, containment is achieved by requiring
540 that both the source and destination cgroups are reachable from the
541 namespace of the process which is attempting the migration. If either
542 is not reachable, the migration is rejected with -ENOENT.
548 Organize Once and Control
549 ~~~~~~~~~~~~~~~~~~~~~~~~~
551 Migrating a process across cgroups is a relatively expensive operation
552 and stateful resources such as memory are not moved together with the
553 process. This is an explicit design decision as there often exist
554 inherent trade-offs between migration and various hot paths in terms
555 of synchronization cost.
557 As such, migrating processes across cgroups frequently as a means to
558 apply different resource restrictions is discouraged. A workload
559 should be assigned to a cgroup according to the system's logical and
560 resource structure once on start-up. Dynamic adjustments to resource
561 distribution can be made by changing controller configuration through
565 Avoid Name Collisions
566 ~~~~~~~~~~~~~~~~~~~~~
568 Interface files for a cgroup and its children cgroups occupy the same
569 directory and it is possible to create children cgroups which collide
570 with interface files.
572 All cgroup core interface files are prefixed with "cgroup." and each
573 controller's interface files are prefixed with the controller name and
574 a dot. A controller's name is composed of lower case alphabets and
575 '_'s but never begins with an '_' so it can be used as the prefix
576 character for collision avoidance. Also, interface file names won't
577 start or end with terms which are often used in categorizing workloads
578 such as job, service, slice, unit or workload.
580 cgroup doesn't do anything to prevent name collisions and it's the
581 user's responsibility to avoid them.
584 Resource Distribution Models
585 ============================
587 cgroup controllers implement several resource distribution schemes
588 depending on the resource type and expected use cases. This section
589 describes major schemes in use along with their expected behaviors.
595 A parent's resource is distributed by adding up the weights of all
596 active children and giving each the fraction matching the ratio of its
597 weight against the sum. As only children which can make use of the
598 resource at the moment participate in the distribution, this is
599 work-conserving. Due to the dynamic nature, this model is usually
600 used for stateless resources.
602 All weights are in the range [1, 10000] with the default at 100. This
603 allows symmetric multiplicative biases in both directions at fine
604 enough granularity while staying in the intuitive range.
606 As long as the weight is in range, all configuration combinations are
607 valid and there is no reason to reject configuration changes or
610 "cpu.weight" proportionally distributes CPU cycles to active children
611 and is an example of this type.
617 A child can only consume upto the configured amount of the resource.
618 Limits can be over-committed - the sum of the limits of children can
619 exceed the amount of resource available to the parent.
621 Limits are in the range [0, max] and defaults to "max", which is noop.
623 As limits can be over-committed, all configuration combinations are
624 valid and there is no reason to reject configuration changes or
627 "io.max" limits the maximum BPS and/or IOPS that a cgroup can consume
628 on an IO device and is an example of this type.
634 A cgroup is protected upto the configured amount of the resource
635 as long as the usages of all its ancestors are under their
636 protected levels. Protections can be hard guarantees or best effort
637 soft boundaries. Protections can also be over-committed in which case
638 only upto the amount available to the parent is protected among
641 Protections are in the range [0, max] and defaults to 0, which is
644 As protections can be over-committed, all configuration combinations
645 are valid and there is no reason to reject configuration changes or
648 "memory.low" implements best-effort memory protection and is an
649 example of this type.
655 A cgroup is exclusively allocated a certain amount of a finite
656 resource. Allocations can't be over-committed - the sum of the
657 allocations of children can not exceed the amount of resource
658 available to the parent.
660 Allocations are in the range [0, max] and defaults to 0, which is no
663 As allocations can't be over-committed, some configuration
664 combinations are invalid and should be rejected. Also, if the
665 resource is mandatory for execution of processes, process migrations
668 "cpu.rt.max" hard-allocates realtime slices and is an example of this
678 All interface files should be in one of the following formats whenever
681 New-line separated values
682 (when only one value can be written at once)
688 Space separated values
689 (when read-only or multiple values can be written at once)
701 KEY0 SUB_KEY0=VAL00 SUB_KEY1=VAL01...
702 KEY1 SUB_KEY0=VAL10 SUB_KEY1=VAL11...
705 For a writable file, the format for writing should generally match
706 reading; however, controllers may allow omitting later fields or
707 implement restricted shortcuts for most common use cases.
709 For both flat and nested keyed files, only the values for a single key
710 can be written at a time. For nested keyed files, the sub key pairs
711 may be specified in any order and not all pairs have to be specified.
717 - Settings for a single feature should be contained in a single file.
719 - The root cgroup should be exempt from resource control and thus
720 shouldn't have resource control interface files.
722 - The default time unit is microseconds. If a different unit is ever
723 used, an explicit unit suffix must be present.
725 - A parts-per quantity should use a percentage decimal with at least
726 two digit fractional part - e.g. 13.40.
728 - If a controller implements weight based resource distribution, its
729 interface file should be named "weight" and have the range [1,
730 10000] with 100 as the default. The values are chosen to allow
731 enough and symmetric bias in both directions while keeping it
732 intuitive (the default is 100%).
734 - If a controller implements an absolute resource guarantee and/or
735 limit, the interface files should be named "min" and "max"
736 respectively. If a controller implements best effort resource
737 guarantee and/or limit, the interface files should be named "low"
738 and "high" respectively.
740 In the above four control files, the special token "max" should be
741 used to represent upward infinity for both reading and writing.
743 - If a setting has a configurable default value and keyed specific
744 overrides, the default entry should be keyed with "default" and
745 appear as the first entry in the file.
747 The default value can be updated by writing either "default $VAL" or
750 When writing to update a specific override, "default" can be used as
751 the value to indicate removal of the override. Override entries
752 with "default" as the value must not appear when read.
754 For example, a setting which is keyed by major:minor device numbers
755 with integer values may look like the following::
757 # cat cgroup-example-interface-file
761 The default value can be updated by::
763 # echo 125 > cgroup-example-interface-file
767 # echo "default 125" > cgroup-example-interface-file
769 An override can be set by::
771 # echo "8:16 170" > cgroup-example-interface-file
775 # echo "8:0 default" > cgroup-example-interface-file
776 # cat cgroup-example-interface-file
780 - For events which are not very high frequency, an interface file
781 "events" should be created which lists event key value pairs.
782 Whenever a notifiable event happens, file modified event should be
783 generated on the file.
789 All cgroup core files are prefixed with "cgroup."
792 A read-write single value file which exists on non-root
795 When read, it indicates the current type of the cgroup, which
796 can be one of the following values.
798 - "domain" : A normal valid domain cgroup.
800 - "domain threaded" : A threaded domain cgroup which is
801 serving as the root of a threaded subtree.
803 - "domain invalid" : A cgroup which is in an invalid state.
804 It can't be populated or have controllers enabled. It may
805 be allowed to become a threaded cgroup.
807 - "threaded" : A threaded cgroup which is a member of a
810 A cgroup can be turned into a threaded cgroup by writing
811 "threaded" to this file.
814 A read-write new-line separated values file which exists on
817 When read, it lists the PIDs of all processes which belong to
818 the cgroup one-per-line. The PIDs are not ordered and the
819 same PID may show up more than once if the process got moved
820 to another cgroup and then back or the PID got recycled while
823 A PID can be written to migrate the process associated with
824 the PID to the cgroup. The writer should match all of the
825 following conditions.
827 - It must have write access to the "cgroup.procs" file.
829 - It must have write access to the "cgroup.procs" file of the
830 common ancestor of the source and destination cgroups.
832 When delegating a sub-hierarchy, write access to this file
833 should be granted along with the containing directory.
835 In a threaded cgroup, reading this file fails with EOPNOTSUPP
836 as all the processes belong to the thread root. Writing is
837 supported and moves every thread of the process to the cgroup.
840 A read-write new-line separated values file which exists on
843 When read, it lists the TIDs of all threads which belong to
844 the cgroup one-per-line. The TIDs are not ordered and the
845 same TID may show up more than once if the thread got moved to
846 another cgroup and then back or the TID got recycled while
849 A TID can be written to migrate the thread associated with the
850 TID to the cgroup. The writer should match all of the
851 following conditions.
853 - It must have write access to the "cgroup.threads" file.
855 - The cgroup that the thread is currently in must be in the
856 same resource domain as the destination cgroup.
858 - It must have write access to the "cgroup.procs" file of the
859 common ancestor of the source and destination cgroups.
861 When delegating a sub-hierarchy, write access to this file
862 should be granted along with the containing directory.
865 A read-only space separated values file which exists on all
868 It shows space separated list of all controllers available to
869 the cgroup. The controllers are not ordered.
871 cgroup.subtree_control
872 A read-write space separated values file which exists on all
873 cgroups. Starts out empty.
875 When read, it shows space separated list of the controllers
876 which are enabled to control resource distribution from the
877 cgroup to its children.
879 Space separated list of controllers prefixed with '+' or '-'
880 can be written to enable or disable controllers. A controller
881 name prefixed with '+' enables the controller and '-'
882 disables. If a controller appears more than once on the list,
883 the last one is effective. When multiple enable and disable
884 operations are specified, either all succeed or all fail.
887 A read-only flat-keyed file which exists on non-root cgroups.
888 The following entries are defined. Unless specified
889 otherwise, a value change in this file generates a file
893 1 if the cgroup or its descendants contains any live
894 processes; otherwise, 0.
896 1 if the cgroup is frozen; otherwise, 0.
898 cgroup.max.descendants
899 A read-write single value files. The default is "max".
901 Maximum allowed number of descent cgroups.
902 If the actual number of descendants is equal or larger,
903 an attempt to create a new cgroup in the hierarchy will fail.
906 A read-write single value files. The default is "max".
908 Maximum allowed descent depth below the current cgroup.
909 If the actual descent depth is equal or larger,
910 an attempt to create a new child cgroup will fail.
913 A read-only flat-keyed file with the following entries:
916 Total number of visible descendant cgroups.
919 Total number of dying descendant cgroups. A cgroup becomes
920 dying after being deleted by a user. The cgroup will remain
921 in dying state for some time undefined time (which can depend
922 on system load) before being completely destroyed.
924 A process can't enter a dying cgroup under any circumstances,
925 a dying cgroup can't revive.
927 A dying cgroup can consume system resources not exceeding
928 limits, which were active at the moment of cgroup deletion.
931 A read-write single value file which exists on non-root cgroups.
932 Allowed values are "0" and "1". The default is "0".
934 Writing "1" to the file causes freezing of the cgroup and all
935 descendant cgroups. This means that all belonging processes will
936 be stopped and will not run until the cgroup will be explicitly
937 unfrozen. Freezing of the cgroup may take some time; when this action
938 is completed, the "frozen" value in the cgroup.events control file
939 will be updated to "1" and the corresponding notification will be
942 A cgroup can be frozen either by its own settings, or by settings
943 of any ancestor cgroups. If any of ancestor cgroups is frozen, the
944 cgroup will remain frozen.
946 Processes in the frozen cgroup can be killed by a fatal signal.
947 They also can enter and leave a frozen cgroup: either by an explicit
948 move by a user, or if freezing of the cgroup races with fork().
949 If a process is moved to a frozen cgroup, it stops. If a process is
950 moved out of a frozen cgroup, it becomes running.
952 Frozen status of a cgroup doesn't affect any cgroup tree operations:
953 it's possible to delete a frozen (and empty) cgroup, as well as
954 create new sub-cgroups.
964 The "cpu" controllers regulates distribution of CPU cycles. This
965 controller implements weight and absolute bandwidth limit models for
966 normal scheduling policy and absolute bandwidth allocation model for
967 realtime scheduling policy.
969 In all the above models, cycles distribution is defined only on a temporal
970 base and it does not account for the frequency at which tasks are executed.
971 The (optional) utilization clamping support allows to hint the schedutil
972 cpufreq governor about the minimum desired frequency which should always be
973 provided by a CPU, as well as the maximum desired frequency, which should not
974 be exceeded by a CPU.
976 WARNING: cgroup2 doesn't yet support control of realtime processes and
977 the cpu controller can only be enabled when all RT processes are in
978 the root cgroup. Be aware that system management software may already
979 have placed RT processes into nonroot cgroups during the system boot
980 process, and these processes may need to be moved to the root cgroup
981 before the cpu controller can be enabled.
987 All time durations are in microseconds.
990 A read-only flat-keyed file.
991 This file exists whether the controller is enabled or not.
993 It always reports the following three stats:
999 and the following three when the controller is enabled:
1006 A read-write single value file which exists on non-root
1007 cgroups. The default is "100".
1009 The weight in the range [1, 10000].
1012 A read-write single value file which exists on non-root
1013 cgroups. The default is "0".
1015 The nice value is in the range [-20, 19].
1017 This interface file is an alternative interface for
1018 "cpu.weight" and allows reading and setting weight using the
1019 same values used by nice(2). Because the range is smaller and
1020 granularity is coarser for the nice values, the read value is
1021 the closest approximation of the current weight.
1024 A read-write two value file which exists on non-root cgroups.
1025 The default is "max 100000".
1027 The maximum bandwidth limit. It's in the following format::
1031 which indicates that the group may consume upto $MAX in each
1032 $PERIOD duration. "max" for $MAX indicates no limit. If only
1033 one number is written, $MAX is updated.
1036 A read-write nested-keyed file.
1038 Shows pressure stall information for CPU. See
1039 :ref:`Documentation/accounting/psi.rst <psi>` for details.
1042 A read-write single value file which exists on non-root cgroups.
1043 The default is "0", i.e. no utilization boosting.
1045 The requested minimum utilization (protection) as a percentage
1046 rational number, e.g. 12.34 for 12.34%.
1048 This interface allows reading and setting minimum utilization clamp
1049 values similar to the sched_setattr(2). This minimum utilization
1050 value is used to clamp the task specific minimum utilization clamp.
1052 The requested minimum utilization (protection) is always capped by
1053 the current value for the maximum utilization (limit), i.e.
1057 A read-write single value file which exists on non-root cgroups.
1058 The default is "max". i.e. no utilization capping
1060 The requested maximum utilization (limit) as a percentage rational
1061 number, e.g. 98.76 for 98.76%.
1063 This interface allows reading and setting maximum utilization clamp
1064 values similar to the sched_setattr(2). This maximum utilization
1065 value is used to clamp the task specific maximum utilization clamp.
1072 The "memory" controller regulates distribution of memory. Memory is
1073 stateful and implements both limit and protection models. Due to the
1074 intertwining between memory usage and reclaim pressure and the
1075 stateful nature of memory, the distribution model is relatively
1078 While not completely water-tight, all major memory usages by a given
1079 cgroup are tracked so that the total memory consumption can be
1080 accounted and controlled to a reasonable extent. Currently, the
1081 following types of memory usages are tracked.
1083 - Userland memory - page cache and anonymous memory.
1085 - Kernel data structures such as dentries and inodes.
1087 - TCP socket buffers.
1089 The above list may expand in the future for better coverage.
1092 Memory Interface Files
1093 ~~~~~~~~~~~~~~~~~~~~~~
1095 All memory amounts are in bytes. If a value which is not aligned to
1096 PAGE_SIZE is written, the value may be rounded up to the closest
1097 PAGE_SIZE multiple when read back.
1100 A read-only single value file which exists on non-root
1103 The total amount of memory currently being used by the cgroup
1104 and its descendants.
1107 A read-write single value file which exists on non-root
1108 cgroups. The default is "0".
1110 Hard memory protection. If the memory usage of a cgroup
1111 is within its effective min boundary, the cgroup's memory
1112 won't be reclaimed under any conditions. If there is no
1113 unprotected reclaimable memory available, OOM killer
1114 is invoked. Above the effective min boundary (or
1115 effective low boundary if it is higher), pages are reclaimed
1116 proportionally to the overage, reducing reclaim pressure for
1119 Effective min boundary is limited by memory.min values of
1120 all ancestor cgroups. If there is memory.min overcommitment
1121 (child cgroup or cgroups are requiring more protected memory
1122 than parent will allow), then each child cgroup will get
1123 the part of parent's protection proportional to its
1124 actual memory usage below memory.min.
1126 Putting more memory than generally available under this
1127 protection is discouraged and may lead to constant OOMs.
1129 If a memory cgroup is not populated with processes,
1130 its memory.min is ignored.
1133 A read-write single value file which exists on non-root
1134 cgroups. The default is "0".
1136 Best-effort memory protection. If the memory usage of a
1137 cgroup is within its effective low boundary, the cgroup's
1138 memory won't be reclaimed unless there is no reclaimable
1139 memory available in unprotected cgroups.
1140 Above the effective low boundary (or
1141 effective min boundary if it is higher), pages are reclaimed
1142 proportionally to the overage, reducing reclaim pressure for
1145 Effective low boundary is limited by memory.low values of
1146 all ancestor cgroups. If there is memory.low overcommitment
1147 (child cgroup or cgroups are requiring more protected memory
1148 than parent will allow), then each child cgroup will get
1149 the part of parent's protection proportional to its
1150 actual memory usage below memory.low.
1152 Putting more memory than generally available under this
1153 protection is discouraged.
1156 A read-write single value file which exists on non-root
1157 cgroups. The default is "max".
1159 Memory usage throttle limit. This is the main mechanism to
1160 control memory usage of a cgroup. If a cgroup's usage goes
1161 over the high boundary, the processes of the cgroup are
1162 throttled and put under heavy reclaim pressure.
1164 Going over the high limit never invokes the OOM killer and
1165 under extreme conditions the limit may be breached.
1168 A read-write single value file which exists on non-root
1169 cgroups. The default is "max".
1171 Memory usage hard limit. This is the final protection
1172 mechanism. If a cgroup's memory usage reaches this limit and
1173 can't be reduced, the OOM killer is invoked in the cgroup.
1174 Under certain circumstances, the usage may go over the limit
1177 In default configuration regular 0-order allocations always
1178 succeed unless OOM killer chooses current task as a victim.
1180 Some kinds of allocations don't invoke the OOM killer.
1181 Caller could retry them differently, return into userspace
1182 as -ENOMEM or silently ignore in cases like disk readahead.
1184 This is the ultimate protection mechanism. As long as the
1185 high limit is used and monitored properly, this limit's
1186 utility is limited to providing the final safety net.
1189 A read-write single value file which exists on non-root
1190 cgroups. The default value is "0".
1192 Determines whether the cgroup should be treated as
1193 an indivisible workload by the OOM killer. If set,
1194 all tasks belonging to the cgroup or to its descendants
1195 (if the memory cgroup is not a leaf cgroup) are killed
1196 together or not at all. This can be used to avoid
1197 partial kills to guarantee workload integrity.
1199 Tasks with the OOM protection (oom_score_adj set to -1000)
1200 are treated as an exception and are never killed.
1202 If the OOM killer is invoked in a cgroup, it's not going
1203 to kill any tasks outside of this cgroup, regardless
1204 memory.oom.group values of ancestor cgroups.
1207 A read-only flat-keyed file which exists on non-root cgroups.
1208 The following entries are defined. Unless specified
1209 otherwise, a value change in this file generates a file
1212 Note that all fields in this file are hierarchical and the
1213 file modified event can be generated due to an event down the
1214 hierarchy. For for the local events at the cgroup level see
1215 memory.events.local.
1218 The number of times the cgroup is reclaimed due to
1219 high memory pressure even though its usage is under
1220 the low boundary. This usually indicates that the low
1221 boundary is over-committed.
1224 The number of times processes of the cgroup are
1225 throttled and routed to perform direct memory reclaim
1226 because the high memory boundary was exceeded. For a
1227 cgroup whose memory usage is capped by the high limit
1228 rather than global memory pressure, this event's
1229 occurrences are expected.
1232 The number of times the cgroup's memory usage was
1233 about to go over the max boundary. If direct reclaim
1234 fails to bring it down, the cgroup goes to OOM state.
1237 The number of time the cgroup's memory usage was
1238 reached the limit and allocation was about to fail.
1240 This event is not raised if the OOM killer is not
1241 considered as an option, e.g. for failed high-order
1242 allocations or if caller asked to not retry attempts.
1245 The number of processes belonging to this cgroup
1246 killed by any kind of OOM killer.
1249 Similar to memory.events but the fields in the file are local
1250 to the cgroup i.e. not hierarchical. The file modified event
1251 generated on this file reflects only the local events.
1254 A read-only flat-keyed file which exists on non-root cgroups.
1256 This breaks down the cgroup's memory footprint into different
1257 types of memory, type-specific details, and other information
1258 on the state and past events of the memory management system.
1260 All memory amounts are in bytes.
1262 The entries are ordered to be human readable, and new entries
1263 can show up in the middle. Don't rely on items remaining in a
1264 fixed position; use the keys to look up specific values!
1266 If the entry has no per-node counter (or not show in the
1267 memory.numa_stat). We use 'npn' (non-per-node) as the tag
1268 to indicate that it will not show in the memory.numa_stat.
1271 Amount of memory used in anonymous mappings such as
1272 brk(), sbrk(), and mmap(MAP_ANONYMOUS)
1275 Amount of memory used to cache filesystem data,
1276 including tmpfs and shared memory.
1279 Amount of memory allocated to kernel stacks.
1282 Amount of memory allocated for page tables.
1285 Amount of memory used for storing per-cpu kernel
1289 Amount of memory used in network transmission buffers
1292 Amount of cached filesystem data that is swap-backed,
1293 such as tmpfs, shm segments, shared anonymous mmap()s
1296 Amount of cached filesystem data mapped with mmap()
1299 Amount of cached filesystem data that was modified but
1300 not yet written back to disk
1303 Amount of cached filesystem data that was modified and
1304 is currently being written back to disk
1307 Amount of swap cached in memory. The swapcache is accounted
1308 against both memory and swap usage.
1311 Amount of memory used in anonymous mappings backed by
1312 transparent hugepages
1315 Amount of cached filesystem data backed by transparent
1319 Amount of shm, tmpfs, shared anonymous mmap()s backed by
1320 transparent hugepages
1322 inactive_anon, active_anon, inactive_file, active_file, unevictable
1323 Amount of memory, swap-backed and filesystem-backed,
1324 on the internal memory management lists used by the
1325 page reclaim algorithm.
1327 As these represent internal list state (eg. shmem pages are on anon
1328 memory management lists), inactive_foo + active_foo may not be equal to
1329 the value for the foo counter, since the foo counter is type-based, not
1333 Part of "slab" that might be reclaimed, such as
1334 dentries and inodes.
1337 Part of "slab" that cannot be reclaimed on memory
1341 Amount of memory used for storing in-kernel data
1344 workingset_refault_anon
1345 Number of refaults of previously evicted anonymous pages.
1347 workingset_refault_file
1348 Number of refaults of previously evicted file pages.
1350 workingset_activate_anon
1351 Number of refaulted anonymous pages that were immediately
1354 workingset_activate_file
1355 Number of refaulted file pages that were immediately activated.
1357 workingset_restore_anon
1358 Number of restored anonymous pages which have been detected as
1359 an active workingset before they got reclaimed.
1361 workingset_restore_file
1362 Number of restored file pages which have been detected as an
1363 active workingset before they got reclaimed.
1365 workingset_nodereclaim
1366 Number of times a shadow node has been reclaimed
1369 Total number of page faults incurred
1372 Number of major page faults incurred
1375 Amount of scanned pages (in an active LRU list)
1378 Amount of scanned pages (in an inactive LRU list)
1381 Amount of reclaimed pages
1384 Amount of pages moved to the active LRU list
1387 Amount of pages moved to the inactive LRU list
1390 Amount of pages postponed to be freed under memory pressure
1393 Amount of reclaimed lazyfree pages
1395 thp_fault_alloc (npn)
1396 Number of transparent hugepages which were allocated to satisfy
1397 a page fault. This counter is not present when CONFIG_TRANSPARENT_HUGEPAGE
1400 thp_collapse_alloc (npn)
1401 Number of transparent hugepages which were allocated to allow
1402 collapsing an existing range of pages. This counter is not
1403 present when CONFIG_TRANSPARENT_HUGEPAGE is not set.
1406 A read-only nested-keyed file which exists on non-root cgroups.
1408 This breaks down the cgroup's memory footprint into different
1409 types of memory, type-specific details, and other information
1410 per node on the state of the memory management system.
1412 This is useful for providing visibility into the NUMA locality
1413 information within an memcg since the pages are allowed to be
1414 allocated from any physical node. One of the use case is evaluating
1415 application performance by combining this information with the
1416 application's CPU allocation.
1418 All memory amounts are in bytes.
1420 The output format of memory.numa_stat is::
1422 type N0=<bytes in node 0> N1=<bytes in node 1> ...
1424 The entries are ordered to be human readable, and new entries
1425 can show up in the middle. Don't rely on items remaining in a
1426 fixed position; use the keys to look up specific values!
1428 The entries can refer to the memory.stat.
1431 A read-only single value file which exists on non-root
1434 The total amount of swap currently being used by the cgroup
1435 and its descendants.
1438 A read-write single value file which exists on non-root
1439 cgroups. The default is "max".
1441 Swap usage throttle limit. If a cgroup's swap usage exceeds
1442 this limit, all its further allocations will be throttled to
1443 allow userspace to implement custom out-of-memory procedures.
1445 This limit marks a point of no return for the cgroup. It is NOT
1446 designed to manage the amount of swapping a workload does
1447 during regular operation. Compare to memory.swap.max, which
1448 prohibits swapping past a set amount, but lets the cgroup
1449 continue unimpeded as long as other memory can be reclaimed.
1451 Healthy workloads are not expected to reach this limit.
1454 A read-write single value file which exists on non-root
1455 cgroups. The default is "max".
1457 Swap usage hard limit. If a cgroup's swap usage reaches this
1458 limit, anonymous memory of the cgroup will not be swapped out.
1461 A read-only flat-keyed file which exists on non-root cgroups.
1462 The following entries are defined. Unless specified
1463 otherwise, a value change in this file generates a file
1467 The number of times the cgroup's swap usage was over
1471 The number of times the cgroup's swap usage was about
1472 to go over the max boundary and swap allocation
1476 The number of times swap allocation failed either
1477 because of running out of swap system-wide or max
1480 When reduced under the current usage, the existing swap
1481 entries are reclaimed gradually and the swap usage may stay
1482 higher than the limit for an extended period of time. This
1483 reduces the impact on the workload and memory management.
1486 A read-only nested-keyed file.
1488 Shows pressure stall information for memory. See
1489 :ref:`Documentation/accounting/psi.rst <psi>` for details.
1495 "memory.high" is the main mechanism to control memory usage.
1496 Over-committing on high limit (sum of high limits > available memory)
1497 and letting global memory pressure to distribute memory according to
1498 usage is a viable strategy.
1500 Because breach of the high limit doesn't trigger the OOM killer but
1501 throttles the offending cgroup, a management agent has ample
1502 opportunities to monitor and take appropriate actions such as granting
1503 more memory or terminating the workload.
1505 Determining whether a cgroup has enough memory is not trivial as
1506 memory usage doesn't indicate whether the workload can benefit from
1507 more memory. For example, a workload which writes data received from
1508 network to a file can use all available memory but can also operate as
1509 performant with a small amount of memory. A measure of memory
1510 pressure - how much the workload is being impacted due to lack of
1511 memory - is necessary to determine whether a workload needs more
1512 memory; unfortunately, memory pressure monitoring mechanism isn't
1519 A memory area is charged to the cgroup which instantiated it and stays
1520 charged to the cgroup until the area is released. Migrating a process
1521 to a different cgroup doesn't move the memory usages that it
1522 instantiated while in the previous cgroup to the new cgroup.
1524 A memory area may be used by processes belonging to different cgroups.
1525 To which cgroup the area will be charged is in-deterministic; however,
1526 over time, the memory area is likely to end up in a cgroup which has
1527 enough memory allowance to avoid high reclaim pressure.
1529 If a cgroup sweeps a considerable amount of memory which is expected
1530 to be accessed repeatedly by other cgroups, it may make sense to use
1531 POSIX_FADV_DONTNEED to relinquish the ownership of memory areas
1532 belonging to the affected files to ensure correct memory ownership.
1538 The "io" controller regulates the distribution of IO resources. This
1539 controller implements both weight based and absolute bandwidth or IOPS
1540 limit distribution; however, weight based distribution is available
1541 only if cfq-iosched is in use and neither scheme is available for
1549 A read-only nested-keyed file.
1551 Lines are keyed by $MAJ:$MIN device numbers and not ordered.
1552 The following nested keys are defined.
1554 ====== =====================
1556 wbytes Bytes written
1557 rios Number of read IOs
1558 wios Number of write IOs
1559 dbytes Bytes discarded
1560 dios Number of discard IOs
1561 ====== =====================
1563 An example read output follows::
1565 8:16 rbytes=1459200 wbytes=314773504 rios=192 wios=353 dbytes=0 dios=0
1566 8:0 rbytes=90430464 wbytes=299008000 rios=8950 wios=1252 dbytes=50331648 dios=3021
1569 A read-write nested-keyed file which exists only on the root
1572 This file configures the Quality of Service of the IO cost
1573 model based controller (CONFIG_BLK_CGROUP_IOCOST) which
1574 currently implements "io.weight" proportional control. Lines
1575 are keyed by $MAJ:$MIN device numbers and not ordered. The
1576 line for a given device is populated on the first write for
1577 the device on "io.cost.qos" or "io.cost.model". The following
1578 nested keys are defined.
1580 ====== =====================================
1581 enable Weight-based control enable
1582 ctrl "auto" or "user"
1583 rpct Read latency percentile [0, 100]
1584 rlat Read latency threshold
1585 wpct Write latency percentile [0, 100]
1586 wlat Write latency threshold
1587 min Minimum scaling percentage [1, 10000]
1588 max Maximum scaling percentage [1, 10000]
1589 ====== =====================================
1591 The controller is disabled by default and can be enabled by
1592 setting "enable" to 1. "rpct" and "wpct" parameters default
1593 to zero and the controller uses internal device saturation
1594 state to adjust the overall IO rate between "min" and "max".
1596 When a better control quality is needed, latency QoS
1597 parameters can be configured. For example::
1599 8:16 enable=1 ctrl=auto rpct=95.00 rlat=75000 wpct=95.00 wlat=150000 min=50.00 max=150.0
1601 shows that on sdb, the controller is enabled, will consider
1602 the device saturated if the 95th percentile of read completion
1603 latencies is above 75ms or write 150ms, and adjust the overall
1604 IO issue rate between 50% and 150% accordingly.
1606 The lower the saturation point, the better the latency QoS at
1607 the cost of aggregate bandwidth. The narrower the allowed
1608 adjustment range between "min" and "max", the more conformant
1609 to the cost model the IO behavior. Note that the IO issue
1610 base rate may be far off from 100% and setting "min" and "max"
1611 blindly can lead to a significant loss of device capacity or
1612 control quality. "min" and "max" are useful for regulating
1613 devices which show wide temporary behavior changes - e.g. a
1614 ssd which accepts writes at the line speed for a while and
1615 then completely stalls for multiple seconds.
1617 When "ctrl" is "auto", the parameters are controlled by the
1618 kernel and may change automatically. Setting "ctrl" to "user"
1619 or setting any of the percentile and latency parameters puts
1620 it into "user" mode and disables the automatic changes. The
1621 automatic mode can be restored by setting "ctrl" to "auto".
1624 A read-write nested-keyed file which exists only on the root
1627 This file configures the cost model of the IO cost model based
1628 controller (CONFIG_BLK_CGROUP_IOCOST) which currently
1629 implements "io.weight" proportional control. Lines are keyed
1630 by $MAJ:$MIN device numbers and not ordered. The line for a
1631 given device is populated on the first write for the device on
1632 "io.cost.qos" or "io.cost.model". The following nested keys
1635 ===== ================================
1636 ctrl "auto" or "user"
1637 model The cost model in use - "linear"
1638 ===== ================================
1640 When "ctrl" is "auto", the kernel may change all parameters
1641 dynamically. When "ctrl" is set to "user" or any other
1642 parameters are written to, "ctrl" become "user" and the
1643 automatic changes are disabled.
1645 When "model" is "linear", the following model parameters are
1648 ============= ========================================
1649 [r|w]bps The maximum sequential IO throughput
1650 [r|w]seqiops The maximum 4k sequential IOs per second
1651 [r|w]randiops The maximum 4k random IOs per second
1652 ============= ========================================
1654 From the above, the builtin linear model determines the base
1655 costs of a sequential and random IO and the cost coefficient
1656 for the IO size. While simple, this model can cover most
1657 common device classes acceptably.
1659 The IO cost model isn't expected to be accurate in absolute
1660 sense and is scaled to the device behavior dynamically.
1662 If needed, tools/cgroup/iocost_coef_gen.py can be used to
1663 generate device-specific coefficients.
1666 A read-write flat-keyed file which exists on non-root cgroups.
1667 The default is "default 100".
1669 The first line is the default weight applied to devices
1670 without specific override. The rest are overrides keyed by
1671 $MAJ:$MIN device numbers and not ordered. The weights are in
1672 the range [1, 10000] and specifies the relative amount IO time
1673 the cgroup can use in relation to its siblings.
1675 The default weight can be updated by writing either "default
1676 $WEIGHT" or simply "$WEIGHT". Overrides can be set by writing
1677 "$MAJ:$MIN $WEIGHT" and unset by writing "$MAJ:$MIN default".
1679 An example read output follows::
1686 A read-write nested-keyed file which exists on non-root
1689 BPS and IOPS based IO limit. Lines are keyed by $MAJ:$MIN
1690 device numbers and not ordered. The following nested keys are
1693 ===== ==================================
1694 rbps Max read bytes per second
1695 wbps Max write bytes per second
1696 riops Max read IO operations per second
1697 wiops Max write IO operations per second
1698 ===== ==================================
1700 When writing, any number of nested key-value pairs can be
1701 specified in any order. "max" can be specified as the value
1702 to remove a specific limit. If the same key is specified
1703 multiple times, the outcome is undefined.
1705 BPS and IOPS are measured in each IO direction and IOs are
1706 delayed if limit is reached. Temporary bursts are allowed.
1708 Setting read limit at 2M BPS and write at 120 IOPS for 8:16::
1710 echo "8:16 rbps=2097152 wiops=120" > io.max
1712 Reading returns the following::
1714 8:16 rbps=2097152 wbps=max riops=max wiops=120
1716 Write IOPS limit can be removed by writing the following::
1718 echo "8:16 wiops=max" > io.max
1720 Reading now returns the following::
1722 8:16 rbps=2097152 wbps=max riops=max wiops=max
1725 A read-only nested-keyed file.
1727 Shows pressure stall information for IO. See
1728 :ref:`Documentation/accounting/psi.rst <psi>` for details.
1734 Page cache is dirtied through buffered writes and shared mmaps and
1735 written asynchronously to the backing filesystem by the writeback
1736 mechanism. Writeback sits between the memory and IO domains and
1737 regulates the proportion of dirty memory by balancing dirtying and
1740 The io controller, in conjunction with the memory controller,
1741 implements control of page cache writeback IOs. The memory controller
1742 defines the memory domain that dirty memory ratio is calculated and
1743 maintained for and the io controller defines the io domain which
1744 writes out dirty pages for the memory domain. Both system-wide and
1745 per-cgroup dirty memory states are examined and the more restrictive
1746 of the two is enforced.
1748 cgroup writeback requires explicit support from the underlying
1749 filesystem. Currently, cgroup writeback is implemented on ext2, ext4,
1750 btrfs, f2fs, and xfs. On other filesystems, all writeback IOs are
1751 attributed to the root cgroup.
1753 There are inherent differences in memory and writeback management
1754 which affects how cgroup ownership is tracked. Memory is tracked per
1755 page while writeback per inode. For the purpose of writeback, an
1756 inode is assigned to a cgroup and all IO requests to write dirty pages
1757 from the inode are attributed to that cgroup.
1759 As cgroup ownership for memory is tracked per page, there can be pages
1760 which are associated with different cgroups than the one the inode is
1761 associated with. These are called foreign pages. The writeback
1762 constantly keeps track of foreign pages and, if a particular foreign
1763 cgroup becomes the majority over a certain period of time, switches
1764 the ownership of the inode to that cgroup.
1766 While this model is enough for most use cases where a given inode is
1767 mostly dirtied by a single cgroup even when the main writing cgroup
1768 changes over time, use cases where multiple cgroups write to a single
1769 inode simultaneously are not supported well. In such circumstances, a
1770 significant portion of IOs are likely to be attributed incorrectly.
1771 As memory controller assigns page ownership on the first use and
1772 doesn't update it until the page is released, even if writeback
1773 strictly follows page ownership, multiple cgroups dirtying overlapping
1774 areas wouldn't work as expected. It's recommended to avoid such usage
1777 The sysctl knobs which affect writeback behavior are applied to cgroup
1778 writeback as follows.
1780 vm.dirty_background_ratio, vm.dirty_ratio
1781 These ratios apply the same to cgroup writeback with the
1782 amount of available memory capped by limits imposed by the
1783 memory controller and system-wide clean memory.
1785 vm.dirty_background_bytes, vm.dirty_bytes
1786 For cgroup writeback, this is calculated into ratio against
1787 total available memory and applied the same way as
1788 vm.dirty[_background]_ratio.
1794 This is a cgroup v2 controller for IO workload protection. You provide a group
1795 with a latency target, and if the average latency exceeds that target the
1796 controller will throttle any peers that have a lower latency target than the
1799 The limits are only applied at the peer level in the hierarchy. This means that
1800 in the diagram below, only groups A, B, and C will influence each other, and
1801 groups D and F will influence each other. Group G will influence nobody::
1810 So the ideal way to configure this is to set io.latency in groups A, B, and C.
1811 Generally you do not want to set a value lower than the latency your device
1812 supports. Experiment to find the value that works best for your workload.
1813 Start at higher than the expected latency for your device and watch the
1814 avg_lat value in io.stat for your workload group to get an idea of the
1815 latency you see during normal operation. Use the avg_lat value as a basis for
1816 your real setting, setting at 10-15% higher than the value in io.stat.
1818 How IO Latency Throttling Works
1819 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1821 io.latency is work conserving; so as long as everybody is meeting their latency
1822 target the controller doesn't do anything. Once a group starts missing its
1823 target it begins throttling any peer group that has a higher target than itself.
1824 This throttling takes 2 forms:
1826 - Queue depth throttling. This is the number of outstanding IO's a group is
1827 allowed to have. We will clamp down relatively quickly, starting at no limit
1828 and going all the way down to 1 IO at a time.
1830 - Artificial delay induction. There are certain types of IO that cannot be
1831 throttled without possibly adversely affecting higher priority groups. This
1832 includes swapping and metadata IO. These types of IO are allowed to occur
1833 normally, however they are "charged" to the originating group. If the
1834 originating group is being throttled you will see the use_delay and delay
1835 fields in io.stat increase. The delay value is how many microseconds that are
1836 being added to any process that runs in this group. Because this number can
1837 grow quite large if there is a lot of swapping or metadata IO occurring we
1838 limit the individual delay events to 1 second at a time.
1840 Once the victimized group starts meeting its latency target again it will start
1841 unthrottling any peer groups that were throttled previously. If the victimized
1842 group simply stops doing IO the global counter will unthrottle appropriately.
1844 IO Latency Interface Files
1845 ~~~~~~~~~~~~~~~~~~~~~~~~~~
1848 This takes a similar format as the other controllers.
1850 "MAJOR:MINOR target=<target time in microseconds"
1853 If the controller is enabled you will see extra stats in io.stat in
1854 addition to the normal ones.
1857 This is the current queue depth for the group.
1860 This is an exponential moving average with a decay rate of 1/exp
1861 bound by the sampling interval. The decay rate interval can be
1862 calculated by multiplying the win value in io.stat by the
1863 corresponding number of samples based on the win value.
1866 The sampling window size in milliseconds. This is the minimum
1867 duration of time between evaluation events. Windows only elapse
1868 with IO activity. Idle periods extend the most recent window.
1873 A single attribute controls the behavior of the I/O priority cgroup policy,
1874 namely the blkio.prio.class attribute. The following values are accepted for
1878 Do not modify the I/O priority class.
1881 For requests that do not have an I/O priority class (NONE),
1882 change the I/O priority class into RT. Do not modify
1883 the I/O priority class of other requests.
1886 For requests that do not have an I/O priority class or that have I/O
1887 priority class RT, change it into BE. Do not modify the I/O priority
1888 class of requests that have priority class IDLE.
1891 Change the I/O priority class of all requests into IDLE, the lowest
1894 The following numerical values are associated with the I/O priority policies:
1906 The numerical value that corresponds to each I/O priority class is as follows:
1908 +-------------------------------+---+
1909 | IOPRIO_CLASS_NONE | 0 |
1910 +-------------------------------+---+
1911 | IOPRIO_CLASS_RT (real-time) | 1 |
1912 +-------------------------------+---+
1913 | IOPRIO_CLASS_BE (best effort) | 2 |
1914 +-------------------------------+---+
1915 | IOPRIO_CLASS_IDLE | 3 |
1916 +-------------------------------+---+
1918 The algorithm to set the I/O priority class for a request is as follows:
1920 - Translate the I/O priority class policy into a number.
1921 - Change the request I/O priority class into the maximum of the I/O priority
1922 class policy number and the numerical I/O priority class.
1927 The process number controller is used to allow a cgroup to stop any
1928 new tasks from being fork()'d or clone()'d after a specified limit is
1931 The number of tasks in a cgroup can be exhausted in ways which other
1932 controllers cannot prevent, thus warranting its own controller. For
1933 example, a fork bomb is likely to exhaust the number of tasks before
1934 hitting memory restrictions.
1936 Note that PIDs used in this controller refer to TIDs, process IDs as
1944 A read-write single value file which exists on non-root
1945 cgroups. The default is "max".
1947 Hard limit of number of processes.
1950 A read-only single value file which exists on all cgroups.
1952 The number of processes currently in the cgroup and its
1955 Organisational operations are not blocked by cgroup policies, so it is
1956 possible to have pids.current > pids.max. This can be done by either
1957 setting the limit to be smaller than pids.current, or attaching enough
1958 processes to the cgroup such that pids.current is larger than
1959 pids.max. However, it is not possible to violate a cgroup PID policy
1960 through fork() or clone(). These will return -EAGAIN if the creation
1961 of a new process would cause a cgroup policy to be violated.
1967 The "cpuset" controller provides a mechanism for constraining
1968 the CPU and memory node placement of tasks to only the resources
1969 specified in the cpuset interface files in a task's current cgroup.
1970 This is especially valuable on large NUMA systems where placing jobs
1971 on properly sized subsets of the systems with careful processor and
1972 memory placement to reduce cross-node memory access and contention
1973 can improve overall system performance.
1975 The "cpuset" controller is hierarchical. That means the controller
1976 cannot use CPUs or memory nodes not allowed in its parent.
1979 Cpuset Interface Files
1980 ~~~~~~~~~~~~~~~~~~~~~~
1983 A read-write multiple values file which exists on non-root
1984 cpuset-enabled cgroups.
1986 It lists the requested CPUs to be used by tasks within this
1987 cgroup. The actual list of CPUs to be granted, however, is
1988 subjected to constraints imposed by its parent and can differ
1989 from the requested CPUs.
1991 The CPU numbers are comma-separated numbers or ranges.
1997 An empty value indicates that the cgroup is using the same
1998 setting as the nearest cgroup ancestor with a non-empty
1999 "cpuset.cpus" or all the available CPUs if none is found.
2001 The value of "cpuset.cpus" stays constant until the next update
2002 and won't be affected by any CPU hotplug events.
2004 cpuset.cpus.effective
2005 A read-only multiple values file which exists on all
2006 cpuset-enabled cgroups.
2008 It lists the onlined CPUs that are actually granted to this
2009 cgroup by its parent. These CPUs are allowed to be used by
2010 tasks within the current cgroup.
2012 If "cpuset.cpus" is empty, the "cpuset.cpus.effective" file shows
2013 all the CPUs from the parent cgroup that can be available to
2014 be used by this cgroup. Otherwise, it should be a subset of
2015 "cpuset.cpus" unless none of the CPUs listed in "cpuset.cpus"
2016 can be granted. In this case, it will be treated just like an
2017 empty "cpuset.cpus".
2019 Its value will be affected by CPU hotplug events.
2022 A read-write multiple values file which exists on non-root
2023 cpuset-enabled cgroups.
2025 It lists the requested memory nodes to be used by tasks within
2026 this cgroup. The actual list of memory nodes granted, however,
2027 is subjected to constraints imposed by its parent and can differ
2028 from the requested memory nodes.
2030 The memory node numbers are comma-separated numbers or ranges.
2036 An empty value indicates that the cgroup is using the same
2037 setting as the nearest cgroup ancestor with a non-empty
2038 "cpuset.mems" or all the available memory nodes if none
2041 The value of "cpuset.mems" stays constant until the next update
2042 and won't be affected by any memory nodes hotplug events.
2044 cpuset.mems.effective
2045 A read-only multiple values file which exists on all
2046 cpuset-enabled cgroups.
2048 It lists the onlined memory nodes that are actually granted to
2049 this cgroup by its parent. These memory nodes are allowed to
2050 be used by tasks within the current cgroup.
2052 If "cpuset.mems" is empty, it shows all the memory nodes from the
2053 parent cgroup that will be available to be used by this cgroup.
2054 Otherwise, it should be a subset of "cpuset.mems" unless none of
2055 the memory nodes listed in "cpuset.mems" can be granted. In this
2056 case, it will be treated just like an empty "cpuset.mems".
2058 Its value will be affected by memory nodes hotplug events.
2060 cpuset.cpus.partition
2061 A read-write single value file which exists on non-root
2062 cpuset-enabled cgroups. This flag is owned by the parent cgroup
2063 and is not delegatable.
2065 It accepts only the following input values when written to.
2067 ======== ================================
2068 "root" a partition root
2069 "member" a non-root member of a partition
2070 ======== ================================
2072 When set to be a partition root, the current cgroup is the
2073 root of a new partition or scheduling domain that comprises
2074 itself and all its descendants except those that are separate
2075 partition roots themselves and their descendants. The root
2076 cgroup is always a partition root.
2078 There are constraints on where a partition root can be set.
2079 It can only be set in a cgroup if all the following conditions
2082 1) The "cpuset.cpus" is not empty and the list of CPUs are
2083 exclusive, i.e. they are not shared by any of its siblings.
2084 2) The parent cgroup is a partition root.
2085 3) The "cpuset.cpus" is also a proper subset of the parent's
2086 "cpuset.cpus.effective".
2087 4) There is no child cgroups with cpuset enabled. This is for
2088 eliminating corner cases that have to be handled if such a
2089 condition is allowed.
2091 Setting it to partition root will take the CPUs away from the
2092 effective CPUs of the parent cgroup. Once it is set, this
2093 file cannot be reverted back to "member" if there are any child
2094 cgroups with cpuset enabled.
2096 A parent partition cannot distribute all its CPUs to its
2097 child partitions. There must be at least one cpu left in the
2100 Once becoming a partition root, changes to "cpuset.cpus" is
2101 generally allowed as long as the first condition above is true,
2102 the change will not take away all the CPUs from the parent
2103 partition and the new "cpuset.cpus" value is a superset of its
2104 children's "cpuset.cpus" values.
2106 Sometimes, external factors like changes to ancestors'
2107 "cpuset.cpus" or cpu hotplug can cause the state of the partition
2108 root to change. On read, the "cpuset.sched.partition" file
2109 can show the following values.
2111 ============== ==============================
2112 "member" Non-root member of a partition
2113 "root" Partition root
2114 "root invalid" Invalid partition root
2115 ============== ==============================
2117 It is a partition root if the first 2 partition root conditions
2118 above are true and at least one CPU from "cpuset.cpus" is
2119 granted by the parent cgroup.
2121 A partition root can become invalid if none of CPUs requested
2122 in "cpuset.cpus" can be granted by the parent cgroup or the
2123 parent cgroup is no longer a partition root itself. In this
2124 case, it is not a real partition even though the restriction
2125 of the first partition root condition above will still apply.
2126 The cpu affinity of all the tasks in the cgroup will then be
2127 associated with CPUs in the nearest ancestor partition.
2129 An invalid partition root can be transitioned back to a
2130 real partition root if at least one of the requested CPUs
2131 can now be granted by its parent. In this case, the cpu
2132 affinity of all the tasks in the formerly invalid partition
2133 will be associated to the CPUs of the newly formed partition.
2134 Changing the partition state of an invalid partition root to
2135 "member" is always allowed even if child cpusets are present.
2141 Device controller manages access to device files. It includes both
2142 creation of new device files (using mknod), and access to the
2143 existing device files.
2145 Cgroup v2 device controller has no interface files and is implemented
2146 on top of cgroup BPF. To control access to device files, a user may
2147 create bpf programs of the BPF_CGROUP_DEVICE type and attach them
2148 to cgroups. On an attempt to access a device file, corresponding
2149 BPF programs will be executed, and depending on the return value
2150 the attempt will succeed or fail with -EPERM.
2152 A BPF_CGROUP_DEVICE program takes a pointer to the bpf_cgroup_dev_ctx
2153 structure, which describes the device access attempt: access type
2154 (mknod/read/write) and device (type, major and minor numbers).
2155 If the program returns 0, the attempt fails with -EPERM, otherwise
2158 An example of BPF_CGROUP_DEVICE program may be found in the kernel
2159 source tree in the tools/testing/selftests/bpf/progs/dev_cgroup.c file.
2165 The "rdma" controller regulates the distribution and accounting of
2168 RDMA Interface Files
2169 ~~~~~~~~~~~~~~~~~~~~
2172 A readwrite nested-keyed file that exists for all the cgroups
2173 except root that describes current configured resource limit
2174 for a RDMA/IB device.
2176 Lines are keyed by device name and are not ordered.
2177 Each line contains space separated resource name and its configured
2178 limit that can be distributed.
2180 The following nested keys are defined.
2182 ========== =============================
2183 hca_handle Maximum number of HCA Handles
2184 hca_object Maximum number of HCA Objects
2185 ========== =============================
2187 An example for mlx4 and ocrdma device follows::
2189 mlx4_0 hca_handle=2 hca_object=2000
2190 ocrdma1 hca_handle=3 hca_object=max
2193 A read-only file that describes current resource usage.
2194 It exists for all the cgroup except root.
2196 An example for mlx4 and ocrdma device follows::
2198 mlx4_0 hca_handle=1 hca_object=20
2199 ocrdma1 hca_handle=1 hca_object=23
2204 The HugeTLB controller allows to limit the HugeTLB usage per control group and
2205 enforces the controller limit during page fault.
2207 HugeTLB Interface Files
2208 ~~~~~~~~~~~~~~~~~~~~~~~
2210 hugetlb.<hugepagesize>.current
2211 Show current usage for "hugepagesize" hugetlb. It exists for all
2212 the cgroup except root.
2214 hugetlb.<hugepagesize>.max
2215 Set/show the hard limit of "hugepagesize" hugetlb usage.
2216 The default value is "max". It exists for all the cgroup except root.
2218 hugetlb.<hugepagesize>.events
2219 A read-only flat-keyed file which exists on non-root cgroups.
2222 The number of allocation failure due to HugeTLB limit
2224 hugetlb.<hugepagesize>.events.local
2225 Similar to hugetlb.<hugepagesize>.events but the fields in the file
2226 are local to the cgroup i.e. not hierarchical. The file modified event
2227 generated on this file reflects only the local events.
2232 The Miscellaneous cgroup provides the resource limiting and tracking
2233 mechanism for the scalar resources which cannot be abstracted like the other
2234 cgroup resources. Controller is enabled by the CONFIG_CGROUP_MISC config
2237 A resource can be added to the controller via enum misc_res_type{} in the
2238 include/linux/misc_cgroup.h file and the corresponding name via misc_res_name[]
2239 in the kernel/cgroup/misc.c file. Provider of the resource must set its
2240 capacity prior to using the resource by calling misc_cg_set_capacity().
2242 Once a capacity is set then the resource usage can be updated using charge and
2243 uncharge APIs. All of the APIs to interact with misc controller are in
2244 include/linux/misc_cgroup.h.
2246 Misc Interface Files
2247 ~~~~~~~~~~~~~~~~~~~~
2249 Miscellaneous controller provides 3 interface files. If two misc resources (res_a and res_b) are registered then:
2252 A read-only flat-keyed file shown only in the root cgroup. It shows
2253 miscellaneous scalar resources available on the platform along with
2261 A read-only flat-keyed file shown in the non-root cgroups. It shows
2262 the current usage of the resources in the cgroup and its children.::
2269 A read-write flat-keyed file shown in the non root cgroups. Allowed
2270 maximum usage of the resources in the cgroup and its children.::
2276 Limit can be set by::
2278 # echo res_a 1 > misc.max
2280 Limit can be set to max by::
2282 # echo res_a max > misc.max
2284 Limits can be set higher than the capacity value in the misc.capacity
2287 Migration and Ownership
2288 ~~~~~~~~~~~~~~~~~~~~~~~
2290 A miscellaneous scalar resource is charged to the cgroup in which it is used
2291 first, and stays charged to that cgroup until that resource is freed. Migrating
2292 a process to a different cgroup does not move the charge to the destination
2293 cgroup where the process has moved.
2301 perf_event controller, if not mounted on a legacy hierarchy, is
2302 automatically enabled on the v2 hierarchy so that perf events can
2303 always be filtered by cgroup v2 path. The controller can still be
2304 moved to a legacy hierarchy after v2 hierarchy is populated.
2307 Non-normative information
2308 -------------------------
2310 This section contains information that isn't considered to be a part of
2311 the stable kernel API and so is subject to change.
2314 CPU controller root cgroup process behaviour
2315 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2317 When distributing CPU cycles in the root cgroup each thread in this
2318 cgroup is treated as if it was hosted in a separate child cgroup of the
2319 root cgroup. This child cgroup weight is dependent on its thread nice
2322 For details of this mapping see sched_prio_to_weight array in
2323 kernel/sched/core.c file (values from this array should be scaled
2324 appropriately so the neutral - nice 0 - value is 100 instead of 1024).
2327 IO controller root cgroup process behaviour
2328 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2330 Root cgroup processes are hosted in an implicit leaf child node.
2331 When distributing IO resources this implicit child node is taken into
2332 account as if it was a normal child cgroup of the root cgroup with a
2333 weight value of 200.
2342 cgroup namespace provides a mechanism to virtualize the view of the
2343 "/proc/$PID/cgroup" file and cgroup mounts. The CLONE_NEWCGROUP clone
2344 flag can be used with clone(2) and unshare(2) to create a new cgroup
2345 namespace. The process running inside the cgroup namespace will have
2346 its "/proc/$PID/cgroup" output restricted to cgroupns root. The
2347 cgroupns root is the cgroup of the process at the time of creation of
2348 the cgroup namespace.
2350 Without cgroup namespace, the "/proc/$PID/cgroup" file shows the
2351 complete path of the cgroup of a process. In a container setup where
2352 a set of cgroups and namespaces are intended to isolate processes the
2353 "/proc/$PID/cgroup" file may leak potential system level information
2354 to the isolated processes. For example::
2356 # cat /proc/self/cgroup
2357 0::/batchjobs/container_id1
2359 The path '/batchjobs/container_id1' can be considered as system-data
2360 and undesirable to expose to the isolated processes. cgroup namespace
2361 can be used to restrict visibility of this path. For example, before
2362 creating a cgroup namespace, one would see::
2364 # ls -l /proc/self/ns/cgroup
2365 lrwxrwxrwx 1 root root 0 2014-07-15 10:37 /proc/self/ns/cgroup -> cgroup:[4026531835]
2366 # cat /proc/self/cgroup
2367 0::/batchjobs/container_id1
2369 After unsharing a new namespace, the view changes::
2371 # ls -l /proc/self/ns/cgroup
2372 lrwxrwxrwx 1 root root 0 2014-07-15 10:35 /proc/self/ns/cgroup -> cgroup:[4026532183]
2373 # cat /proc/self/cgroup
2376 When some thread from a multi-threaded process unshares its cgroup
2377 namespace, the new cgroupns gets applied to the entire process (all
2378 the threads). This is natural for the v2 hierarchy; however, for the
2379 legacy hierarchies, this may be unexpected.
2381 A cgroup namespace is alive as long as there are processes inside or
2382 mounts pinning it. When the last usage goes away, the cgroup
2383 namespace is destroyed. The cgroupns root and the actual cgroups
2390 The 'cgroupns root' for a cgroup namespace is the cgroup in which the
2391 process calling unshare(2) is running. For example, if a process in
2392 /batchjobs/container_id1 cgroup calls unshare, cgroup
2393 /batchjobs/container_id1 becomes the cgroupns root. For the
2394 init_cgroup_ns, this is the real root ('/') cgroup.
2396 The cgroupns root cgroup does not change even if the namespace creator
2397 process later moves to a different cgroup::
2399 # ~/unshare -c # unshare cgroupns in some cgroup
2400 # cat /proc/self/cgroup
2403 # echo 0 > sub_cgrp_1/cgroup.procs
2404 # cat /proc/self/cgroup
2407 Each process gets its namespace-specific view of "/proc/$PID/cgroup"
2409 Processes running inside the cgroup namespace will be able to see
2410 cgroup paths (in /proc/self/cgroup) only inside their root cgroup.
2411 From within an unshared cgroupns::
2415 # echo 7353 > sub_cgrp_1/cgroup.procs
2416 # cat /proc/7353/cgroup
2419 From the initial cgroup namespace, the real cgroup path will be
2422 $ cat /proc/7353/cgroup
2423 0::/batchjobs/container_id1/sub_cgrp_1
2425 From a sibling cgroup namespace (that is, a namespace rooted at a
2426 different cgroup), the cgroup path relative to its own cgroup
2427 namespace root will be shown. For instance, if PID 7353's cgroup
2428 namespace root is at '/batchjobs/container_id2', then it will see::
2430 # cat /proc/7353/cgroup
2431 0::/../container_id2/sub_cgrp_1
2433 Note that the relative path always starts with '/' to indicate that
2434 its relative to the cgroup namespace root of the caller.
2437 Migration and setns(2)
2438 ----------------------
2440 Processes inside a cgroup namespace can move into and out of the
2441 namespace root if they have proper access to external cgroups. For
2442 example, from inside a namespace with cgroupns root at
2443 /batchjobs/container_id1, and assuming that the global hierarchy is
2444 still accessible inside cgroupns::
2446 # cat /proc/7353/cgroup
2448 # echo 7353 > batchjobs/container_id2/cgroup.procs
2449 # cat /proc/7353/cgroup
2450 0::/../container_id2
2452 Note that this kind of setup is not encouraged. A task inside cgroup
2453 namespace should only be exposed to its own cgroupns hierarchy.
2455 setns(2) to another cgroup namespace is allowed when:
2457 (a) the process has CAP_SYS_ADMIN against its current user namespace
2458 (b) the process has CAP_SYS_ADMIN against the target cgroup
2461 No implicit cgroup changes happen with attaching to another cgroup
2462 namespace. It is expected that the someone moves the attaching
2463 process under the target cgroup namespace root.
2466 Interaction with Other Namespaces
2467 ---------------------------------
2469 Namespace specific cgroup hierarchy can be mounted by a process
2470 running inside a non-init cgroup namespace::
2472 # mount -t cgroup2 none $MOUNT_POINT
2474 This will mount the unified cgroup hierarchy with cgroupns root as the
2475 filesystem root. The process needs CAP_SYS_ADMIN against its user and
2478 The virtualization of /proc/self/cgroup file combined with restricting
2479 the view of cgroup hierarchy by namespace-private cgroupfs mount
2480 provides a properly isolated cgroup view inside the container.
2483 Information on Kernel Programming
2484 =================================
2486 This section contains kernel programming information in the areas
2487 where interacting with cgroup is necessary. cgroup core and
2488 controllers are not covered.
2491 Filesystem Support for Writeback
2492 --------------------------------
2494 A filesystem can support cgroup writeback by updating
2495 address_space_operations->writepage[s]() to annotate bio's using the
2496 following two functions.
2498 wbc_init_bio(@wbc, @bio)
2499 Should be called for each bio carrying writeback data and
2500 associates the bio with the inode's owner cgroup and the
2501 corresponding request queue. This must be called after
2502 a queue (device) has been associated with the bio and
2505 wbc_account_cgroup_owner(@wbc, @page, @bytes)
2506 Should be called for each data segment being written out.
2507 While this function doesn't care exactly when it's called
2508 during the writeback session, it's the easiest and most
2509 natural to call it as data segments are added to a bio.
2511 With writeback bio's annotated, cgroup support can be enabled per
2512 super_block by setting SB_I_CGROUPWB in ->s_iflags. This allows for
2513 selective disabling of cgroup writeback support which is helpful when
2514 certain filesystem features, e.g. journaled data mode, are
2517 wbc_init_bio() binds the specified bio to its cgroup. Depending on
2518 the configuration, the bio may be executed at a lower priority and if
2519 the writeback session is holding shared resources, e.g. a journal
2520 entry, may lead to priority inversion. There is no one easy solution
2521 for the problem. Filesystems can try to work around specific problem
2522 cases by skipping wbc_init_bio() and using bio_associate_blkg()
2526 Deprecated v1 Core Features
2527 ===========================
2529 - Multiple hierarchies including named ones are not supported.
2531 - All v1 mount options are not supported.
2533 - The "tasks" file is removed and "cgroup.procs" is not sorted.
2535 - "cgroup.clone_children" is removed.
2537 - /proc/cgroups is meaningless for v2. Use "cgroup.controllers" file
2538 at the root instead.
2541 Issues with v1 and Rationales for v2
2542 ====================================
2544 Multiple Hierarchies
2545 --------------------
2547 cgroup v1 allowed an arbitrary number of hierarchies and each
2548 hierarchy could host any number of controllers. While this seemed to
2549 provide a high level of flexibility, it wasn't useful in practice.
2551 For example, as there is only one instance of each controller, utility
2552 type controllers such as freezer which can be useful in all
2553 hierarchies could only be used in one. The issue is exacerbated by
2554 the fact that controllers couldn't be moved to another hierarchy once
2555 hierarchies were populated. Another issue was that all controllers
2556 bound to a hierarchy were forced to have exactly the same view of the
2557 hierarchy. It wasn't possible to vary the granularity depending on
2558 the specific controller.
2560 In practice, these issues heavily limited which controllers could be
2561 put on the same hierarchy and most configurations resorted to putting
2562 each controller on its own hierarchy. Only closely related ones, such
2563 as the cpu and cpuacct controllers, made sense to be put on the same
2564 hierarchy. This often meant that userland ended up managing multiple
2565 similar hierarchies repeating the same steps on each hierarchy
2566 whenever a hierarchy management operation was necessary.
2568 Furthermore, support for multiple hierarchies came at a steep cost.
2569 It greatly complicated cgroup core implementation but more importantly
2570 the support for multiple hierarchies restricted how cgroup could be
2571 used in general and what controllers was able to do.
2573 There was no limit on how many hierarchies there might be, which meant
2574 that a thread's cgroup membership couldn't be described in finite
2575 length. The key might contain any number of entries and was unlimited
2576 in length, which made it highly awkward to manipulate and led to
2577 addition of controllers which existed only to identify membership,
2578 which in turn exacerbated the original problem of proliferating number
2581 Also, as a controller couldn't have any expectation regarding the
2582 topologies of hierarchies other controllers might be on, each
2583 controller had to assume that all other controllers were attached to
2584 completely orthogonal hierarchies. This made it impossible, or at
2585 least very cumbersome, for controllers to cooperate with each other.
2587 In most use cases, putting controllers on hierarchies which are
2588 completely orthogonal to each other isn't necessary. What usually is
2589 called for is the ability to have differing levels of granularity
2590 depending on the specific controller. In other words, hierarchy may
2591 be collapsed from leaf towards root when viewed from specific
2592 controllers. For example, a given configuration might not care about
2593 how memory is distributed beyond a certain level while still wanting
2594 to control how CPU cycles are distributed.
2600 cgroup v1 allowed threads of a process to belong to different cgroups.
2601 This didn't make sense for some controllers and those controllers
2602 ended up implementing different ways to ignore such situations but
2603 much more importantly it blurred the line between API exposed to
2604 individual applications and system management interface.
2606 Generally, in-process knowledge is available only to the process
2607 itself; thus, unlike service-level organization of processes,
2608 categorizing threads of a process requires active participation from
2609 the application which owns the target process.
2611 cgroup v1 had an ambiguously defined delegation model which got abused
2612 in combination with thread granularity. cgroups were delegated to
2613 individual applications so that they can create and manage their own
2614 sub-hierarchies and control resource distributions along them. This
2615 effectively raised cgroup to the status of a syscall-like API exposed
2618 First of all, cgroup has a fundamentally inadequate interface to be
2619 exposed this way. For a process to access its own knobs, it has to
2620 extract the path on the target hierarchy from /proc/self/cgroup,
2621 construct the path by appending the name of the knob to the path, open
2622 and then read and/or write to it. This is not only extremely clunky
2623 and unusual but also inherently racy. There is no conventional way to
2624 define transaction across the required steps and nothing can guarantee
2625 that the process would actually be operating on its own sub-hierarchy.
2627 cgroup controllers implemented a number of knobs which would never be
2628 accepted as public APIs because they were just adding control knobs to
2629 system-management pseudo filesystem. cgroup ended up with interface
2630 knobs which were not properly abstracted or refined and directly
2631 revealed kernel internal details. These knobs got exposed to
2632 individual applications through the ill-defined delegation mechanism
2633 effectively abusing cgroup as a shortcut to implementing public APIs
2634 without going through the required scrutiny.
2636 This was painful for both userland and kernel. Userland ended up with
2637 misbehaving and poorly abstracted interfaces and kernel exposing and
2638 locked into constructs inadvertently.
2641 Competition Between Inner Nodes and Threads
2642 -------------------------------------------
2644 cgroup v1 allowed threads to be in any cgroups which created an
2645 interesting problem where threads belonging to a parent cgroup and its
2646 children cgroups competed for resources. This was nasty as two
2647 different types of entities competed and there was no obvious way to
2648 settle it. Different controllers did different things.
2650 The cpu controller considered threads and cgroups as equivalents and
2651 mapped nice levels to cgroup weights. This worked for some cases but
2652 fell flat when children wanted to be allocated specific ratios of CPU
2653 cycles and the number of internal threads fluctuated - the ratios
2654 constantly changed as the number of competing entities fluctuated.
2655 There also were other issues. The mapping from nice level to weight
2656 wasn't obvious or universal, and there were various other knobs which
2657 simply weren't available for threads.
2659 The io controller implicitly created a hidden leaf node for each
2660 cgroup to host the threads. The hidden leaf had its own copies of all
2661 the knobs with ``leaf_`` prefixed. While this allowed equivalent
2662 control over internal threads, it was with serious drawbacks. It
2663 always added an extra layer of nesting which wouldn't be necessary
2664 otherwise, made the interface messy and significantly complicated the
2667 The memory controller didn't have a way to control what happened
2668 between internal tasks and child cgroups and the behavior was not
2669 clearly defined. There were attempts to add ad-hoc behaviors and
2670 knobs to tailor the behavior to specific workloads which would have
2671 led to problems extremely difficult to resolve in the long term.
2673 Multiple controllers struggled with internal tasks and came up with
2674 different ways to deal with it; unfortunately, all the approaches were
2675 severely flawed and, furthermore, the widely different behaviors
2676 made cgroup as a whole highly inconsistent.
2678 This clearly is a problem which needs to be addressed from cgroup core
2682 Other Interface Issues
2683 ----------------------
2685 cgroup v1 grew without oversight and developed a large number of
2686 idiosyncrasies and inconsistencies. One issue on the cgroup core side
2687 was how an empty cgroup was notified - a userland helper binary was
2688 forked and executed for each event. The event delivery wasn't
2689 recursive or delegatable. The limitations of the mechanism also led
2690 to in-kernel event delivery filtering mechanism further complicating
2693 Controller interfaces were problematic too. An extreme example is
2694 controllers completely ignoring hierarchical organization and treating
2695 all cgroups as if they were all located directly under the root
2696 cgroup. Some controllers exposed a large amount of inconsistent
2697 implementation details to userland.
2699 There also was no consistency across controllers. When a new cgroup
2700 was created, some controllers defaulted to not imposing extra
2701 restrictions while others disallowed any resource usage until
2702 explicitly configured. Configuration knobs for the same type of
2703 control used widely differing naming schemes and formats. Statistics
2704 and information knobs were named arbitrarily and used different
2705 formats and units even in the same controller.
2707 cgroup v2 establishes common conventions where appropriate and updates
2708 controllers so that they expose minimal and consistent interfaces.
2711 Controller Issues and Remedies
2712 ------------------------------
2717 The original lower boundary, the soft limit, is defined as a limit
2718 that is per default unset. As a result, the set of cgroups that
2719 global reclaim prefers is opt-in, rather than opt-out. The costs for
2720 optimizing these mostly negative lookups are so high that the
2721 implementation, despite its enormous size, does not even provide the
2722 basic desirable behavior. First off, the soft limit has no
2723 hierarchical meaning. All configured groups are organized in a global
2724 rbtree and treated like equal peers, regardless where they are located
2725 in the hierarchy. This makes subtree delegation impossible. Second,
2726 the soft limit reclaim pass is so aggressive that it not just
2727 introduces high allocation latencies into the system, but also impacts
2728 system performance due to overreclaim, to the point where the feature
2729 becomes self-defeating.
2731 The memory.low boundary on the other hand is a top-down allocated
2732 reserve. A cgroup enjoys reclaim protection when it's within its
2733 effective low, which makes delegation of subtrees possible. It also
2734 enjoys having reclaim pressure proportional to its overage when
2735 above its effective low.
2737 The original high boundary, the hard limit, is defined as a strict
2738 limit that can not budge, even if the OOM killer has to be called.
2739 But this generally goes against the goal of making the most out of the
2740 available memory. The memory consumption of workloads varies during
2741 runtime, and that requires users to overcommit. But doing that with a
2742 strict upper limit requires either a fairly accurate prediction of the
2743 working set size or adding slack to the limit. Since working set size
2744 estimation is hard and error prone, and getting it wrong results in
2745 OOM kills, most users tend to err on the side of a looser limit and
2746 end up wasting precious resources.
2748 The memory.high boundary on the other hand can be set much more
2749 conservatively. When hit, it throttles allocations by forcing them
2750 into direct reclaim to work off the excess, but it never invokes the
2751 OOM killer. As a result, a high boundary that is chosen too
2752 aggressively will not terminate the processes, but instead it will
2753 lead to gradual performance degradation. The user can monitor this
2754 and make corrections until the minimal memory footprint that still
2755 gives acceptable performance is found.
2757 In extreme cases, with many concurrent allocations and a complete
2758 breakdown of reclaim progress within the group, the high boundary can
2759 be exceeded. But even then it's mostly better to satisfy the
2760 allocation from the slack available in other groups or the rest of the
2761 system than killing the group. Otherwise, memory.max is there to
2762 limit this type of spillover and ultimately contain buggy or even
2763 malicious applications.
2765 Setting the original memory.limit_in_bytes below the current usage was
2766 subject to a race condition, where concurrent charges could cause the
2767 limit setting to fail. memory.max on the other hand will first set the
2768 limit to prevent new charges, and then reclaim and OOM kill until the
2769 new limit is met - or the task writing to memory.max is killed.
2771 The combined memory+swap accounting and limiting is replaced by real
2772 control over swap space.
2774 The main argument for a combined memory+swap facility in the original
2775 cgroup design was that global or parental pressure would always be
2776 able to swap all anonymous memory of a child group, regardless of the
2777 child's own (possibly untrusted) configuration. However, untrusted
2778 groups can sabotage swapping by other means - such as referencing its
2779 anonymous memory in a tight loop - and an admin can not assume full
2780 swappability when overcommitting untrusted jobs.
2782 For trusted jobs, on the other hand, a combined counter is not an
2783 intuitive userspace interface, and it flies in the face of the idea
2784 that cgroup controllers should account and limit specific physical
2785 resources. Swap space is a resource like all others in the system,
2786 and that's why unified hierarchy allows distributing it separately.