xref: /linux/Documentation/admin-guide/cgroup-v2.rst (revision 47cf96fbe393839b125a9b694a8cfdd3f4216baa)
1.. _cgroup-v2:
2
3================
4Control Group v2
5================
6
7:Date: October, 2015
8:Author: Tejun Heo <tj@kernel.org>
9
10This is the authoritative documentation on the design, interface and
11conventions of cgroup v2.  It describes all userland-visible aspects
12of cgroup including core and specific controller behaviors.  All
13future changes must be reflected in this document.  Documentation for
14v1 is available under :ref:`Documentation/admin-guide/cgroup-v1/index.rst <cgroup-v1>`.
15
16.. CONTENTS
17
18   1. Introduction
19     1-1. Terminology
20     1-2. What is cgroup?
21   2. Basic Operations
22     2-1. Mounting
23     2-2. Organizing Processes and Threads
24       2-2-1. Processes
25       2-2-2. 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
31     2-5. Delegation
32       2-5-1. Model of Delegation
33       2-5-2. Delegation Containment
34     2-6. Guidelines
35       2-6-1. Organize Once and Control
36       2-6-2. Avoid Name Collisions
37   3. Resource Distribution Models
38     3-1. Weights
39     3-2. Limits
40     3-3. Protections
41     3-4. Allocations
42   4. Interface Files
43     4-1. Format
44     4-2. Conventions
45     4-3. Core Interface Files
46   5. Controllers
47     5-1. CPU
48       5-1-1. CPU Interface Files
49     5-2. Memory
50       5-2-1. Memory Interface Files
51       5-2-2. Usage Guidelines
52       5-2-3. Memory Ownership
53     5-3. IO
54       5-3-1. IO Interface Files
55       5-3-2. Writeback
56       5-3-3. IO Latency
57         5-3-3-1. How IO Latency Throttling Works
58         5-3-3-2. IO Latency Interface Files
59       5-3-4. IO Priority
60     5-4. PID
61       5-4-1. PID Interface Files
62     5-5. Cpuset
63       5.5-1. Cpuset Interface Files
64     5-6. Device
65     5-7. RDMA
66       5-7-1. RDMA Interface Files
67     5-8. DMEM
68     5-9. HugeTLB
69       5.9-1. HugeTLB Interface Files
70     5-10. Misc
71       5.10-1 Miscellaneous cgroup Interface Files
72       5.10-2 Migration and Ownership
73     5-11. Others
74       5-11-1. perf_event
75     5-N. Non-normative information
76       5-N-1. CPU controller root cgroup process behaviour
77       5-N-2. IO controller root cgroup process behaviour
78   6. Namespace
79     6-1. Basics
80     6-2. The Root and Views
81     6-3. Migration and setns(2)
82     6-4. Interaction with Other Namespaces
83   P. Information on Kernel Programming
84     P-1. Filesystem Support for Writeback
85   D. Deprecated v1 Core Features
86   R. Issues with v1 and Rationales for v2
87     R-1. Multiple Hierarchies
88     R-2. Thread Granularity
89     R-3. Competition Between Inner Nodes and Threads
90     R-4. Other Interface Issues
91     R-5. Controller Issues and Remedies
92       R-5-1. Memory
93
94
95Introduction
96============
97
98Terminology
99-----------
100
101"cgroup" stands for "control group" and is never capitalized.  The
102singular form is used to designate the whole feature and also as a
103qualifier as in "cgroup controllers".  When explicitly referring to
104multiple individual control groups, the plural form "cgroups" is used.
105
106
107What is cgroup?
108---------------
109
110cgroup is a mechanism to organize processes hierarchically and
111distribute system resources along the hierarchy in a controlled and
112configurable manner.
113
114cgroup is largely composed of two parts - the core and controllers.
115cgroup core is primarily responsible for hierarchically organizing
116processes.  A cgroup controller is usually responsible for
117distributing a specific type of system resource along the hierarchy
118although there are utility controllers which serve purposes other than
119resource distribution.
120
121cgroups form a tree structure and every process in the system belongs
122to one and only one cgroup.  All threads of a process belong to the
123same cgroup.  On creation, all processes are put in the cgroup that
124the parent process belongs to at the time.  A process can be migrated
125to another cgroup.  Migration of a process doesn't affect already
126existing descendant processes.
127
128Following certain structural constraints, controllers may be enabled or
129disabled selectively on a cgroup.  All controller behaviors are
130hierarchical - if a controller is enabled on a cgroup, it affects all
131processes which belong to the cgroups consisting the inclusive
132sub-hierarchy of the cgroup.  When a controller is enabled on a nested
133cgroup, it always restricts the resource distribution further.  The
134restrictions set closer to the root in the hierarchy can not be
135overridden from further away.
136
137
138Basic Operations
139================
140
141Mounting
142--------
143
144Unlike v1, cgroup v2 has only single hierarchy.  The cgroup v2
145hierarchy can be mounted with the following mount command::
146
147  # mount -t cgroup2 none $MOUNT_POINT
148
149cgroup2 filesystem has the magic number 0x63677270 ("cgrp").  All
150controllers which support v2 and are not bound to a v1 hierarchy are
151automatically bound to the v2 hierarchy and show up at the root.
152Controllers which are not in active use in the v2 hierarchy can be
153bound to other hierarchies.  This allows mixing v2 hierarchy with the
154legacy v1 multiple hierarchies in a fully backward compatible way.
155
156A controller can be moved across hierarchies only after the controller
157is no longer referenced in its current hierarchy.  Because per-cgroup
158controller states are destroyed asynchronously and controllers may
159have lingering references, a controller may not show up immediately on
160the v2 hierarchy after the final umount of the previous hierarchy.
161Similarly, a controller should be fully disabled to be moved out of
162the unified hierarchy and it may take some time for the disabled
163controller to become available for other hierarchies; furthermore, due
164to inter-controller dependencies, other controllers may need to be
165disabled too.
166
167While useful for development and manual configurations, moving
168controllers dynamically between the v2 and other hierarchies is
169strongly discouraged for production use.  It is recommended to decide
170the hierarchies and controller associations before starting using the
171controllers after system boot.
172
173During transition to v2, system management software might still
174automount the v1 cgroup filesystem and so hijack all controllers
175during boot, before manual intervention is possible. To make testing
176and experimenting easier, the kernel parameter cgroup_no_v1= allows
177disabling controllers in v1 and make them always available in v2.
178
179cgroup v2 currently supports the following mount options.
180
181  nsdelegate
182	Consider cgroup namespaces as delegation boundaries.  This
183	option is system wide and can only be set on mount or modified
184	through remount from the init namespace.  The mount option is
185	ignored on non-init namespace mounts.  Please refer to the
186	Delegation section for details.
187
188  favordynmods
189        Reduce the latencies of dynamic cgroup modifications such as
190        task migrations and controller on/offs at the cost of making
191        hot path operations such as forks and exits more expensive.
192        The static usage pattern of creating a cgroup, enabling
193        controllers, and then seeding it with CLONE_INTO_CGROUP is
194        not affected by this option.
195
196  memory_localevents
197        Only populate memory.events with data for the current cgroup,
198        and not any subtrees. This is legacy behaviour, the default
199        behaviour without this option is to include subtree counts.
200        This option is system wide and can only be set on mount or
201        modified through remount from the init namespace. The mount
202        option is ignored on non-init namespace mounts.
203
204  memory_recursiveprot
205        Recursively apply memory.min and memory.low protection to
206        entire subtrees, without requiring explicit downward
207        propagation into leaf cgroups.  This allows protecting entire
208        subtrees from one another, while retaining free competition
209        within those subtrees.  This should have been the default
210        behavior but is a mount-option to avoid regressing setups
211        relying on the original semantics (e.g. specifying bogusly
212        high 'bypass' protection values at higher tree levels).
213
214  memory_hugetlb_accounting
215        Count HugeTLB memory usage towards the cgroup's overall
216        memory usage for the memory controller (for the purpose of
217        statistics reporting and memory protetion). This is a new
218        behavior that could regress existing setups, so it must be
219        explicitly opted in with this mount option.
220
221        A few caveats to keep in mind:
222
223        * There is no HugeTLB pool management involved in the memory
224          controller. The pre-allocated pool does not belong to anyone.
225          Specifically, when a new HugeTLB folio is allocated to
226          the pool, it is not accounted for from the perspective of the
227          memory controller. It is only charged to a cgroup when it is
228          actually used (for e.g at page fault time). Host memory
229          overcommit management has to consider this when configuring
230          hard limits. In general, HugeTLB pool management should be
231          done via other mechanisms (such as the HugeTLB controller).
232        * Failure to charge a HugeTLB folio to the memory controller
233          results in SIGBUS. This could happen even if the HugeTLB pool
234          still has pages available (but the cgroup limit is hit and
235          reclaim attempt fails).
236        * Charging HugeTLB memory towards the memory controller affects
237          memory protection and reclaim dynamics. Any userspace tuning
238          (of low, min limits for e.g) needs to take this into account.
239        * HugeTLB pages utilized while this option is not selected
240          will not be tracked by the memory controller (even if cgroup
241          v2 is remounted later on).
242
243  pids_localevents
244        The option restores v1-like behavior of pids.events:max, that is only
245        local (inside cgroup proper) fork failures are counted. Without this
246        option pids.events.max represents any pids.max enforcemnt across
247        cgroup's subtree.
248
249
250
251Organizing Processes and Threads
252--------------------------------
253
254Processes
255~~~~~~~~~
256
257Initially, only the root cgroup exists to which all processes belong.
258A child cgroup can be created by creating a sub-directory::
259
260  # mkdir $CGROUP_NAME
261
262A given cgroup may have multiple child cgroups forming a tree
263structure.  Each cgroup has a read-writable interface file
264"cgroup.procs".  When read, it lists the PIDs of all processes which
265belong to the cgroup one-per-line.  The PIDs are not ordered and the
266same PID may show up more than once if the process got moved to
267another cgroup and then back or the PID got recycled while reading.
268
269A process can be migrated into a cgroup by writing its PID to the
270target cgroup's "cgroup.procs" file.  Only one process can be migrated
271on a single write(2) call.  If a process is composed of multiple
272threads, writing the PID of any thread migrates all threads of the
273process.
274
275When a process forks a child process, the new process is born into the
276cgroup that the forking process belongs to at the time of the
277operation.  After exit, a process stays associated with the cgroup
278that it belonged to at the time of exit until it's reaped; however, a
279zombie process does not appear in "cgroup.procs" and thus can't be
280moved to another cgroup.
281
282A cgroup which doesn't have any children or live processes can be
283destroyed by removing the directory.  Note that a cgroup which doesn't
284have any children and is associated only with zombie processes is
285considered empty and can be removed::
286
287  # rmdir $CGROUP_NAME
288
289"/proc/$PID/cgroup" lists a process's cgroup membership.  If legacy
290cgroup is in use in the system, this file may contain multiple lines,
291one for each hierarchy.  The entry for cgroup v2 is always in the
292format "0::$PATH"::
293
294  # cat /proc/842/cgroup
295  ...
296  0::/test-cgroup/test-cgroup-nested
297
298If the process becomes a zombie and the cgroup it was associated with
299is removed subsequently, " (deleted)" is appended to the path::
300
301  # cat /proc/842/cgroup
302  ...
303  0::/test-cgroup/test-cgroup-nested (deleted)
304
305
306Threads
307~~~~~~~
308
309cgroup v2 supports thread granularity for a subset of controllers to
310support use cases requiring hierarchical resource distribution across
311the threads of a group of processes.  By default, all threads of a
312process belong to the same cgroup, which also serves as the resource
313domain to host resource consumptions which are not specific to a
314process or thread.  The thread mode allows threads to be spread across
315a subtree while still maintaining the common resource domain for them.
316
317Controllers which support thread mode are called threaded controllers.
318The ones which don't are called domain controllers.
319
320Marking a cgroup threaded makes it join the resource domain of its
321parent as a threaded cgroup.  The parent may be another threaded
322cgroup whose resource domain is further up in the hierarchy.  The root
323of a threaded subtree, that is, the nearest ancestor which is not
324threaded, is called threaded domain or thread root interchangeably and
325serves as the resource domain for the entire subtree.
326
327Inside a threaded subtree, threads of a process can be put in
328different cgroups and are not subject to the no internal process
329constraint - threaded controllers can be enabled on non-leaf cgroups
330whether they have threads in them or not.
331
332As the threaded domain cgroup hosts all the domain resource
333consumptions of the subtree, it is considered to have internal
334resource consumptions whether there are processes in it or not and
335can't have populated child cgroups which aren't threaded.  Because the
336root cgroup is not subject to no internal process constraint, it can
337serve both as a threaded domain and a parent to domain cgroups.
338
339The current operation mode or type of the cgroup is shown in the
340"cgroup.type" file which indicates whether the cgroup is a normal
341domain, a domain which is serving as the domain of a threaded subtree,
342or a threaded cgroup.
343
344On creation, a cgroup is always a domain cgroup and can be made
345threaded by writing "threaded" to the "cgroup.type" file.  The
346operation is single direction::
347
348  # echo threaded > cgroup.type
349
350Once threaded, the cgroup can't be made a domain again.  To enable the
351thread mode, the following conditions must be met.
352
353- As the cgroup will join the parent's resource domain.  The parent
354  must either be a valid (threaded) domain or a threaded cgroup.
355
356- When the parent is an unthreaded domain, it must not have any domain
357  controllers enabled or populated domain children.  The root is
358  exempt from this requirement.
359
360Topology-wise, a cgroup can be in an invalid state.  Please consider
361the following topology::
362
363  A (threaded domain) - B (threaded) - C (domain, just created)
364
365C is created as a domain but isn't connected to a parent which can
366host child domains.  C can't be used until it is turned into a
367threaded cgroup.  "cgroup.type" file will report "domain (invalid)" in
368these cases.  Operations which fail due to invalid topology use
369EOPNOTSUPP as the errno.
370
371A domain cgroup is turned into a threaded domain when one of its child
372cgroup becomes threaded or threaded controllers are enabled in the
373"cgroup.subtree_control" file while there are processes in the cgroup.
374A threaded domain reverts to a normal domain when the conditions
375clear.
376
377When read, "cgroup.threads" contains the list of the thread IDs of all
378threads in the cgroup.  Except that the operations are per-thread
379instead of per-process, "cgroup.threads" has the same format and
380behaves the same way as "cgroup.procs".  While "cgroup.threads" can be
381written to in any cgroup, as it can only move threads inside the same
382threaded domain, its operations are confined inside each threaded
383subtree.
384
385The threaded domain cgroup serves as the resource domain for the whole
386subtree, and, while the threads can be scattered across the subtree,
387all the processes are considered to be in the threaded domain cgroup.
388"cgroup.procs" in a threaded domain cgroup contains the PIDs of all
389processes in the subtree and is not readable in the subtree proper.
390However, "cgroup.procs" can be written to from anywhere in the subtree
391to migrate all threads of the matching process to the cgroup.
392
393Only threaded controllers can be enabled in a threaded subtree.  When
394a threaded controller is enabled inside a threaded subtree, it only
395accounts for and controls resource consumptions associated with the
396threads in the cgroup and its descendants.  All consumptions which
397aren't tied to a specific thread belong to the threaded domain cgroup.
398
399Because a threaded subtree is exempt from no internal process
400constraint, a threaded controller must be able to handle competition
401between threads in a non-leaf cgroup and its child cgroups.  Each
402threaded controller defines how such competitions are handled.
403
404Currently, the following controllers are threaded and can be enabled
405in a threaded cgroup::
406
407- cpu
408- cpuset
409- perf_event
410- pids
411
412[Un]populated Notification
413--------------------------
414
415Each non-root cgroup has a "cgroup.events" file which contains
416"populated" field indicating whether the cgroup's sub-hierarchy has
417live processes in it.  Its value is 0 if there is no live process in
418the cgroup and its descendants; otherwise, 1.  poll and [id]notify
419events are triggered when the value changes.  This can be used, for
420example, to start a clean-up operation after all processes of a given
421sub-hierarchy have exited.  The populated state updates and
422notifications are recursive.  Consider the following sub-hierarchy
423where the numbers in the parentheses represent the numbers of processes
424in each cgroup::
425
426  A(4) - B(0) - C(1)
427              \ D(0)
428
429A, B and C's "populated" fields would be 1 while D's 0.  After the one
430process in C exits, B and C's "populated" fields would flip to "0" and
431file modified events will be generated on the "cgroup.events" files of
432both cgroups.
433
434
435Controlling Controllers
436-----------------------
437
438Enabling and Disabling
439~~~~~~~~~~~~~~~~~~~~~~
440
441Each cgroup has a "cgroup.controllers" file which lists all
442controllers available for the cgroup to enable::
443
444  # cat cgroup.controllers
445  cpu io memory
446
447No controller is enabled by default.  Controllers can be enabled and
448disabled by writing to the "cgroup.subtree_control" file::
449
450  # echo "+cpu +memory -io" > cgroup.subtree_control
451
452Only controllers which are listed in "cgroup.controllers" can be
453enabled.  When multiple operations are specified as above, either they
454all succeed or fail.  If multiple operations on the same controller
455are specified, the last one is effective.
456
457Enabling a controller in a cgroup indicates that the distribution of
458the target resource across its immediate children will be controlled.
459Consider the following sub-hierarchy.  The enabled controllers are
460listed in parentheses::
461
462  A(cpu,memory) - B(memory) - C()
463                            \ D()
464
465As A has "cpu" and "memory" enabled, A will control the distribution
466of CPU cycles and memory to its children, in this case, B.  As B has
467"memory" enabled but not "CPU", C and D will compete freely on CPU
468cycles but their division of memory available to B will be controlled.
469
470As a controller regulates the distribution of the target resource to
471the cgroup's children, enabling it creates the controller's interface
472files in the child cgroups.  In the above example, enabling "cpu" on B
473would create the "cpu." prefixed controller interface files in C and
474D.  Likewise, disabling "memory" from B would remove the "memory."
475prefixed controller interface files from C and D.  This means that the
476controller interface files - anything which doesn't start with
477"cgroup." are owned by the parent rather than the cgroup itself.
478
479
480Top-down Constraint
481~~~~~~~~~~~~~~~~~~~
482
483Resources are distributed top-down and a cgroup can further distribute
484a resource only if the resource has been distributed to it from the
485parent.  This means that all non-root "cgroup.subtree_control" files
486can only contain controllers which are enabled in the parent's
487"cgroup.subtree_control" file.  A controller can be enabled only if
488the parent has the controller enabled and a controller can't be
489disabled if one or more children have it enabled.
490
491
492No Internal Process Constraint
493~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
494
495Non-root cgroups can distribute domain resources to their children
496only when they don't have any processes of their own.  In other words,
497only domain cgroups which don't contain any processes can have domain
498controllers enabled in their "cgroup.subtree_control" files.
499
500This guarantees that, when a domain controller is looking at the part
501of the hierarchy which has it enabled, processes are always only on
502the leaves.  This rules out situations where child cgroups compete
503against internal processes of the parent.
504
505The root cgroup is exempt from this restriction.  Root contains
506processes and anonymous resource consumption which can't be associated
507with any other cgroups and requires special treatment from most
508controllers.  How resource consumption in the root cgroup is governed
509is up to each controller (for more information on this topic please
510refer to the Non-normative information section in the Controllers
511chapter).
512
513Note that the restriction doesn't get in the way if there is no
514enabled controller in the cgroup's "cgroup.subtree_control".  This is
515important as otherwise it wouldn't be possible to create children of a
516populated cgroup.  To control resource distribution of a cgroup, the
517cgroup must create children and transfer all its processes to the
518children before enabling controllers in its "cgroup.subtree_control"
519file.
520
521
522Delegation
523----------
524
525Model of Delegation
526~~~~~~~~~~~~~~~~~~~
527
528A cgroup can be delegated in two ways.  First, to a less privileged
529user by granting write access of the directory and its "cgroup.procs",
530"cgroup.threads" and "cgroup.subtree_control" files to the user.
531Second, if the "nsdelegate" mount option is set, automatically to a
532cgroup namespace on namespace creation.
533
534Because the resource control interface files in a given directory
535control the distribution of the parent's resources, the delegatee
536shouldn't be allowed to write to them.  For the first method, this is
537achieved by not granting access to these files.  For the second, files
538outside the namespace should be hidden from the delegatee by the means
539of at least mount namespacing, and the kernel rejects writes to all
540files on a namespace root from inside the cgroup namespace, except for
541those files listed in "/sys/kernel/cgroup/delegate" (including
542"cgroup.procs", "cgroup.threads", "cgroup.subtree_control", etc.).
543
544The end results are equivalent for both delegation types.  Once
545delegated, the user can build sub-hierarchy under the directory,
546organize processes inside it as it sees fit and further distribute the
547resources it received from the parent.  The limits and other settings
548of all resource controllers are hierarchical and regardless of what
549happens in the delegated sub-hierarchy, nothing can escape the
550resource restrictions imposed by the parent.
551
552Currently, cgroup doesn't impose any restrictions on the number of
553cgroups in or nesting depth of a delegated sub-hierarchy; however,
554this may be limited explicitly in the future.
555
556
557Delegation Containment
558~~~~~~~~~~~~~~~~~~~~~~
559
560A delegated sub-hierarchy is contained in the sense that processes
561can't be moved into or out of the sub-hierarchy by the delegatee.
562
563For delegations to a less privileged user, this is achieved by
564requiring the following conditions for a process with a non-root euid
565to migrate a target process into a cgroup by writing its PID to the
566"cgroup.procs" file.
567
568- The writer must have write access to the "cgroup.procs" file.
569
570- The writer must have write access to the "cgroup.procs" file of the
571  common ancestor of the source and destination cgroups.
572
573The above two constraints ensure that while a delegatee may migrate
574processes around freely in the delegated sub-hierarchy it can't pull
575in from or push out to outside the sub-hierarchy.
576
577For an example, let's assume cgroups C0 and C1 have been delegated to
578user U0 who created C00, C01 under C0 and C10 under C1 as follows and
579all processes under C0 and C1 belong to U0::
580
581  ~~~~~~~~~~~~~ - C0 - C00
582  ~ cgroup    ~      \ C01
583  ~ hierarchy ~
584  ~~~~~~~~~~~~~ - C1 - C10
585
586Let's also say U0 wants to write the PID of a process which is
587currently in C10 into "C00/cgroup.procs".  U0 has write access to the
588file; however, the common ancestor of the source cgroup C10 and the
589destination cgroup C00 is above the points of delegation and U0 would
590not have write access to its "cgroup.procs" files and thus the write
591will be denied with -EACCES.
592
593For delegations to namespaces, containment is achieved by requiring
594that both the source and destination cgroups are reachable from the
595namespace of the process which is attempting the migration.  If either
596is not reachable, the migration is rejected with -ENOENT.
597
598
599Guidelines
600----------
601
602Organize Once and Control
603~~~~~~~~~~~~~~~~~~~~~~~~~
604
605Migrating a process across cgroups is a relatively expensive operation
606and stateful resources such as memory are not moved together with the
607process.  This is an explicit design decision as there often exist
608inherent trade-offs between migration and various hot paths in terms
609of synchronization cost.
610
611As such, migrating processes across cgroups frequently as a means to
612apply different resource restrictions is discouraged.  A workload
613should be assigned to a cgroup according to the system's logical and
614resource structure once on start-up.  Dynamic adjustments to resource
615distribution can be made by changing controller configuration through
616the interface files.
617
618
619Avoid Name Collisions
620~~~~~~~~~~~~~~~~~~~~~
621
622Interface files for a cgroup and its children cgroups occupy the same
623directory and it is possible to create children cgroups which collide
624with interface files.
625
626All cgroup core interface files are prefixed with "cgroup." and each
627controller's interface files are prefixed with the controller name and
628a dot.  A controller's name is composed of lower case alphabets and
629'_'s but never begins with an '_' so it can be used as the prefix
630character for collision avoidance.  Also, interface file names won't
631start or end with terms which are often used in categorizing workloads
632such as job, service, slice, unit or workload.
633
634cgroup doesn't do anything to prevent name collisions and it's the
635user's responsibility to avoid them.
636
637
638Resource Distribution Models
639============================
640
641cgroup controllers implement several resource distribution schemes
642depending on the resource type and expected use cases.  This section
643describes major schemes in use along with their expected behaviors.
644
645
646Weights
647-------
648
649A parent's resource is distributed by adding up the weights of all
650active children and giving each the fraction matching the ratio of its
651weight against the sum.  As only children which can make use of the
652resource at the moment participate in the distribution, this is
653work-conserving.  Due to the dynamic nature, this model is usually
654used for stateless resources.
655
656All weights are in the range [1, 10000] with the default at 100.  This
657allows symmetric multiplicative biases in both directions at fine
658enough granularity while staying in the intuitive range.
659
660As long as the weight is in range, all configuration combinations are
661valid and there is no reason to reject configuration changes or
662process migrations.
663
664"cpu.weight" proportionally distributes CPU cycles to active children
665and is an example of this type.
666
667
668.. _cgroupv2-limits-distributor:
669
670Limits
671------
672
673A child can only consume up to the configured amount of the resource.
674Limits can be over-committed - the sum of the limits of children can
675exceed the amount of resource available to the parent.
676
677Limits are in the range [0, max] and defaults to "max", which is noop.
678
679As limits can be over-committed, all configuration combinations are
680valid and there is no reason to reject configuration changes or
681process migrations.
682
683"io.max" limits the maximum BPS and/or IOPS that a cgroup can consume
684on an IO device and is an example of this type.
685
686.. _cgroupv2-protections-distributor:
687
688Protections
689-----------
690
691A cgroup is protected up to the configured amount of the resource
692as long as the usages of all its ancestors are under their
693protected levels.  Protections can be hard guarantees or best effort
694soft boundaries.  Protections can also be over-committed in which case
695only up to the amount available to the parent is protected among
696children.
697
698Protections are in the range [0, max] and defaults to 0, which is
699noop.
700
701As protections can be over-committed, all configuration combinations
702are valid and there is no reason to reject configuration changes or
703process migrations.
704
705"memory.low" implements best-effort memory protection and is an
706example of this type.
707
708
709Allocations
710-----------
711
712A cgroup is exclusively allocated a certain amount of a finite
713resource.  Allocations can't be over-committed - the sum of the
714allocations of children can not exceed the amount of resource
715available to the parent.
716
717Allocations are in the range [0, max] and defaults to 0, which is no
718resource.
719
720As allocations can't be over-committed, some configuration
721combinations are invalid and should be rejected.  Also, if the
722resource is mandatory for execution of processes, process migrations
723may be rejected.
724
725"cpu.rt.max" hard-allocates realtime slices and is an example of this
726type.
727
728
729Interface Files
730===============
731
732Format
733------
734
735All interface files should be in one of the following formats whenever
736possible::
737
738  New-line separated values
739  (when only one value can be written at once)
740
741	VAL0\n
742	VAL1\n
743	...
744
745  Space separated values
746  (when read-only or multiple values can be written at once)
747
748	VAL0 VAL1 ...\n
749
750  Flat keyed
751
752	KEY0 VAL0\n
753	KEY1 VAL1\n
754	...
755
756  Nested keyed
757
758	KEY0 SUB_KEY0=VAL00 SUB_KEY1=VAL01...
759	KEY1 SUB_KEY0=VAL10 SUB_KEY1=VAL11...
760	...
761
762For a writable file, the format for writing should generally match
763reading; however, controllers may allow omitting later fields or
764implement restricted shortcuts for most common use cases.
765
766For both flat and nested keyed files, only the values for a single key
767can be written at a time.  For nested keyed files, the sub key pairs
768may be specified in any order and not all pairs have to be specified.
769
770
771Conventions
772-----------
773
774- Settings for a single feature should be contained in a single file.
775
776- The root cgroup should be exempt from resource control and thus
777  shouldn't have resource control interface files.
778
779- The default time unit is microseconds.  If a different unit is ever
780  used, an explicit unit suffix must be present.
781
782- A parts-per quantity should use a percentage decimal with at least
783  two digit fractional part - e.g. 13.40.
784
785- If a controller implements weight based resource distribution, its
786  interface file should be named "weight" and have the range [1,
787  10000] with 100 as the default.  The values are chosen to allow
788  enough and symmetric bias in both directions while keeping it
789  intuitive (the default is 100%).
790
791- If a controller implements an absolute resource guarantee and/or
792  limit, the interface files should be named "min" and "max"
793  respectively.  If a controller implements best effort resource
794  guarantee and/or limit, the interface files should be named "low"
795  and "high" respectively.
796
797  In the above four control files, the special token "max" should be
798  used to represent upward infinity for both reading and writing.
799
800- If a setting has a configurable default value and keyed specific
801  overrides, the default entry should be keyed with "default" and
802  appear as the first entry in the file.
803
804  The default value can be updated by writing either "default $VAL" or
805  "$VAL".
806
807  When writing to update a specific override, "default" can be used as
808  the value to indicate removal of the override.  Override entries
809  with "default" as the value must not appear when read.
810
811  For example, a setting which is keyed by major:minor device numbers
812  with integer values may look like the following::
813
814    # cat cgroup-example-interface-file
815    default 150
816    8:0 300
817
818  The default value can be updated by::
819
820    # echo 125 > cgroup-example-interface-file
821
822  or::
823
824    # echo "default 125" > cgroup-example-interface-file
825
826  An override can be set by::
827
828    # echo "8:16 170" > cgroup-example-interface-file
829
830  and cleared by::
831
832    # echo "8:0 default" > cgroup-example-interface-file
833    # cat cgroup-example-interface-file
834    default 125
835    8:16 170
836
837- For events which are not very high frequency, an interface file
838  "events" should be created which lists event key value pairs.
839  Whenever a notifiable event happens, file modified event should be
840  generated on the file.
841
842
843Core Interface Files
844--------------------
845
846All cgroup core files are prefixed with "cgroup."
847
848  cgroup.type
849	A read-write single value file which exists on non-root
850	cgroups.
851
852	When read, it indicates the current type of the cgroup, which
853	can be one of the following values.
854
855	- "domain" : A normal valid domain cgroup.
856
857	- "domain threaded" : A threaded domain cgroup which is
858          serving as the root of a threaded subtree.
859
860	- "domain invalid" : A cgroup which is in an invalid state.
861	  It can't be populated or have controllers enabled.  It may
862	  be allowed to become a threaded cgroup.
863
864	- "threaded" : A threaded cgroup which is a member of a
865          threaded subtree.
866
867	A cgroup can be turned into a threaded cgroup by writing
868	"threaded" to this file.
869
870  cgroup.procs
871	A read-write new-line separated values file which exists on
872	all cgroups.
873
874	When read, it lists the PIDs of all processes which belong to
875	the cgroup one-per-line.  The PIDs are not ordered and the
876	same PID may show up more than once if the process got moved
877	to another cgroup and then back or the PID got recycled while
878	reading.
879
880	A PID can be written to migrate the process associated with
881	the PID to the cgroup.  The writer should match all of the
882	following conditions.
883
884	- It must have write access to the "cgroup.procs" file.
885
886	- It must have write access to the "cgroup.procs" file of the
887	  common ancestor of the source and destination cgroups.
888
889	When delegating a sub-hierarchy, write access to this file
890	should be granted along with the containing directory.
891
892	In a threaded cgroup, reading this file fails with EOPNOTSUPP
893	as all the processes belong to the thread root.  Writing is
894	supported and moves every thread of the process to the cgroup.
895
896  cgroup.threads
897	A read-write new-line separated values file which exists on
898	all cgroups.
899
900	When read, it lists the TIDs of all threads which belong to
901	the cgroup one-per-line.  The TIDs are not ordered and the
902	same TID may show up more than once if the thread got moved to
903	another cgroup and then back or the TID got recycled while
904	reading.
905
906	A TID can be written to migrate the thread associated with the
907	TID to the cgroup.  The writer should match all of the
908	following conditions.
909
910	- It must have write access to the "cgroup.threads" file.
911
912	- The cgroup that the thread is currently in must be in the
913          same resource domain as the destination cgroup.
914
915	- It must have write access to the "cgroup.procs" file of the
916	  common ancestor of the source and destination cgroups.
917
918	When delegating a sub-hierarchy, write access to this file
919	should be granted along with the containing directory.
920
921  cgroup.controllers
922	A read-only space separated values file which exists on all
923	cgroups.
924
925	It shows space separated list of all controllers available to
926	the cgroup.  The controllers are not ordered.
927
928  cgroup.subtree_control
929	A read-write space separated values file which exists on all
930	cgroups.  Starts out empty.
931
932	When read, it shows space separated list of the controllers
933	which are enabled to control resource distribution from the
934	cgroup to its children.
935
936	Space separated list of controllers prefixed with '+' or '-'
937	can be written to enable or disable controllers.  A controller
938	name prefixed with '+' enables the controller and '-'
939	disables.  If a controller appears more than once on the list,
940	the last one is effective.  When multiple enable and disable
941	operations are specified, either all succeed or all fail.
942
943  cgroup.events
944	A read-only flat-keyed file which exists on non-root cgroups.
945	The following entries are defined.  Unless specified
946	otherwise, a value change in this file generates a file
947	modified event.
948
949	  populated
950		1 if the cgroup or its descendants contains any live
951		processes; otherwise, 0.
952	  frozen
953		1 if the cgroup is frozen; otherwise, 0.
954
955  cgroup.max.descendants
956	A read-write single value files.  The default is "max".
957
958	Maximum allowed number of descent cgroups.
959	If the actual number of descendants is equal or larger,
960	an attempt to create a new cgroup in the hierarchy will fail.
961
962  cgroup.max.depth
963	A read-write single value files.  The default is "max".
964
965	Maximum allowed descent depth below the current cgroup.
966	If the actual descent depth is equal or larger,
967	an attempt to create a new child cgroup will fail.
968
969  cgroup.stat
970	A read-only flat-keyed file with the following entries:
971
972	  nr_descendants
973		Total number of visible descendant cgroups.
974
975	  nr_dying_descendants
976		Total number of dying descendant cgroups. A cgroup becomes
977		dying after being deleted by a user. The cgroup will remain
978		in dying state for some time undefined time (which can depend
979		on system load) before being completely destroyed.
980
981		A process can't enter a dying cgroup under any circumstances,
982		a dying cgroup can't revive.
983
984		A dying cgroup can consume system resources not exceeding
985		limits, which were active at the moment of cgroup deletion.
986
987	  nr_subsys_<cgroup_subsys>
988		Total number of live cgroup subsystems (e.g memory
989		cgroup) at and beneath the current cgroup.
990
991	  nr_dying_subsys_<cgroup_subsys>
992		Total number of dying cgroup subsystems (e.g. memory
993		cgroup) at and beneath the current cgroup.
994
995  cgroup.freeze
996	A read-write single value file which exists on non-root cgroups.
997	Allowed values are "0" and "1". The default is "0".
998
999	Writing "1" to the file causes freezing of the cgroup and all
1000	descendant cgroups. This means that all belonging processes will
1001	be stopped and will not run until the cgroup will be explicitly
1002	unfrozen. Freezing of the cgroup may take some time; when this action
1003	is completed, the "frozen" value in the cgroup.events control file
1004	will be updated to "1" and the corresponding notification will be
1005	issued.
1006
1007	A cgroup can be frozen either by its own settings, or by settings
1008	of any ancestor cgroups. If any of ancestor cgroups is frozen, the
1009	cgroup will remain frozen.
1010
1011	Processes in the frozen cgroup can be killed by a fatal signal.
1012	They also can enter and leave a frozen cgroup: either by an explicit
1013	move by a user, or if freezing of the cgroup races with fork().
1014	If a process is moved to a frozen cgroup, it stops. If a process is
1015	moved out of a frozen cgroup, it becomes running.
1016
1017	Frozen status of a cgroup doesn't affect any cgroup tree operations:
1018	it's possible to delete a frozen (and empty) cgroup, as well as
1019	create new sub-cgroups.
1020
1021  cgroup.kill
1022	A write-only single value file which exists in non-root cgroups.
1023	The only allowed value is "1".
1024
1025	Writing "1" to the file causes the cgroup and all descendant cgroups to
1026	be killed. This means that all processes located in the affected cgroup
1027	tree will be killed via SIGKILL.
1028
1029	Killing a cgroup tree will deal with concurrent forks appropriately and
1030	is protected against migrations.
1031
1032	In a threaded cgroup, writing this file fails with EOPNOTSUPP as
1033	killing cgroups is a process directed operation, i.e. it affects
1034	the whole thread-group.
1035
1036  cgroup.pressure
1037	A read-write single value file that allowed values are "0" and "1".
1038	The default is "1".
1039
1040	Writing "0" to the file will disable the cgroup PSI accounting.
1041	Writing "1" to the file will re-enable the cgroup PSI accounting.
1042
1043	This control attribute is not hierarchical, so disable or enable PSI
1044	accounting in a cgroup does not affect PSI accounting in descendants
1045	and doesn't need pass enablement via ancestors from root.
1046
1047	The reason this control attribute exists is that PSI accounts stalls for
1048	each cgroup separately and aggregates it at each level of the hierarchy.
1049	This may cause non-negligible overhead for some workloads when under
1050	deep level of the hierarchy, in which case this control attribute can
1051	be used to disable PSI accounting in the non-leaf cgroups.
1052
1053  irq.pressure
1054	A read-write nested-keyed file.
1055
1056	Shows pressure stall information for IRQ/SOFTIRQ. See
1057	:ref:`Documentation/accounting/psi.rst <psi>` for details.
1058
1059Controllers
1060===========
1061
1062.. _cgroup-v2-cpu:
1063
1064CPU
1065---
1066
1067The "cpu" controllers regulates distribution of CPU cycles.  This
1068controller implements weight and absolute bandwidth limit models for
1069normal scheduling policy and absolute bandwidth allocation model for
1070realtime scheduling policy.
1071
1072In all the above models, cycles distribution is defined only on a temporal
1073base and it does not account for the frequency at which tasks are executed.
1074The (optional) utilization clamping support allows to hint the schedutil
1075cpufreq governor about the minimum desired frequency which should always be
1076provided by a CPU, as well as the maximum desired frequency, which should not
1077be exceeded by a CPU.
1078
1079WARNING: cgroup2 cpu controller doesn't yet support the (bandwidth) control of
1080realtime processes. For a kernel built with the CONFIG_RT_GROUP_SCHED option
1081enabled for group scheduling of realtime processes, the cpu controller can only
1082be enabled when all RT processes are in the root cgroup. Be aware that system
1083management software may already have placed RT processes into non-root cgroups
1084during the system boot process, and these processes may need to be moved to the
1085root cgroup before the cpu controller can be enabled with a
1086CONFIG_RT_GROUP_SCHED enabled kernel.
1087
1088With CONFIG_RT_GROUP_SCHED disabled, this limitation does not apply and some of
1089the interface files either affect realtime processes or account for them. See
1090the following section for details. Only the cpu controller is affected by
1091CONFIG_RT_GROUP_SCHED. Other controllers can be used for the resource control of
1092realtime processes irrespective of CONFIG_RT_GROUP_SCHED.
1093
1094
1095CPU Interface Files
1096~~~~~~~~~~~~~~~~~~~
1097
1098The interaction of a process with the cpu controller depends on its scheduling
1099policy and the underlying scheduler. From the point of view of the cpu controller,
1100processes can be categorized as follows:
1101
1102* Processes under the fair-class scheduler
1103* Processes under a BPF scheduler with the ``cgroup_set_weight`` callback
1104* Everything else: ``SCHED_{FIFO,RR,DEADLINE}`` and processes under a BPF scheduler
1105  without the ``cgroup_set_weight`` callback
1106
1107For details on when a process is under the fair-class scheduler or a BPF scheduler,
1108check out :ref:`Documentation/scheduler/sched-ext.rst <sched-ext>`.
1109
1110For each of the following interface files, the above categories
1111will be referred to. All time durations are in microseconds.
1112
1113  cpu.stat
1114	A read-only flat-keyed file.
1115	This file exists whether the controller is enabled or not.
1116
1117	It always reports the following three stats, which account for all the
1118	processes in the cgroup:
1119
1120	- usage_usec
1121	- user_usec
1122	- system_usec
1123
1124	and the following five when the controller is enabled, which account for
1125	only the processes under the fair-class scheduler:
1126
1127	- nr_periods
1128	- nr_throttled
1129	- throttled_usec
1130	- nr_bursts
1131	- burst_usec
1132
1133  cpu.weight
1134	A read-write single value file which exists on non-root
1135	cgroups.  The default is "100".
1136
1137	For non idle groups (cpu.idle = 0), the weight is in the
1138	range [1, 10000].
1139
1140	If the cgroup has been configured to be SCHED_IDLE (cpu.idle = 1),
1141	then the weight will show as a 0.
1142
1143	This file affects only processes under the fair-class scheduler and a BPF
1144	scheduler with the ``cgroup_set_weight`` callback depending on what the
1145	callback actually does.
1146
1147  cpu.weight.nice
1148	A read-write single value file which exists on non-root
1149	cgroups.  The default is "0".
1150
1151	The nice value is in the range [-20, 19].
1152
1153	This interface file is an alternative interface for
1154	"cpu.weight" and allows reading and setting weight using the
1155	same values used by nice(2).  Because the range is smaller and
1156	granularity is coarser for the nice values, the read value is
1157	the closest approximation of the current weight.
1158
1159	This file affects only processes under the fair-class scheduler and a BPF
1160	scheduler with the ``cgroup_set_weight`` callback depending on what the
1161	callback actually does.
1162
1163  cpu.max
1164	A read-write two value file which exists on non-root cgroups.
1165	The default is "max 100000".
1166
1167	The maximum bandwidth limit.  It's in the following format::
1168
1169	  $MAX $PERIOD
1170
1171	which indicates that the group may consume up to $MAX in each
1172	$PERIOD duration.  "max" for $MAX indicates no limit.  If only
1173	one number is written, $MAX is updated.
1174
1175	This file affects only processes under the fair-class scheduler.
1176
1177  cpu.max.burst
1178	A read-write single value file which exists on non-root
1179	cgroups.  The default is "0".
1180
1181	The burst in the range [0, $MAX].
1182
1183	This file affects only processes under the fair-class scheduler.
1184
1185  cpu.pressure
1186	A read-write nested-keyed file.
1187
1188	Shows pressure stall information for CPU. See
1189	:ref:`Documentation/accounting/psi.rst <psi>` for details.
1190
1191	This file accounts for all the processes in the cgroup.
1192
1193  cpu.uclamp.min
1194	A read-write single value file which exists on non-root cgroups.
1195	The default is "0", i.e. no utilization boosting.
1196
1197	The requested minimum utilization (protection) as a percentage
1198	rational number, e.g. 12.34 for 12.34%.
1199
1200	This interface allows reading and setting minimum utilization clamp
1201	values similar to the sched_setattr(2). This minimum utilization
1202	value is used to clamp the task specific minimum utilization clamp,
1203	including those of realtime processes.
1204
1205	The requested minimum utilization (protection) is always capped by
1206	the current value for the maximum utilization (limit), i.e.
1207	`cpu.uclamp.max`.
1208
1209	This file affects all the processes in the cgroup.
1210
1211  cpu.uclamp.max
1212	A read-write single value file which exists on non-root cgroups.
1213	The default is "max". i.e. no utilization capping
1214
1215	The requested maximum utilization (limit) as a percentage rational
1216	number, e.g. 98.76 for 98.76%.
1217
1218	This interface allows reading and setting maximum utilization clamp
1219	values similar to the sched_setattr(2). This maximum utilization
1220	value is used to clamp the task specific maximum utilization clamp,
1221	including those of realtime processes.
1222
1223	This file affects all the processes in the cgroup.
1224
1225  cpu.idle
1226	A read-write single value file which exists on non-root cgroups.
1227	The default is 0.
1228
1229	This is the cgroup analog of the per-task SCHED_IDLE sched policy.
1230	Setting this value to a 1 will make the scheduling policy of the
1231	cgroup SCHED_IDLE. The threads inside the cgroup will retain their
1232	own relative priorities, but the cgroup itself will be treated as
1233	very low priority relative to its peers.
1234
1235	This file affects only processes under the fair-class scheduler.
1236
1237Memory
1238------
1239
1240The "memory" controller regulates distribution of memory.  Memory is
1241stateful and implements both limit and protection models.  Due to the
1242intertwining between memory usage and reclaim pressure and the
1243stateful nature of memory, the distribution model is relatively
1244complex.
1245
1246While not completely water-tight, all major memory usages by a given
1247cgroup are tracked so that the total memory consumption can be
1248accounted and controlled to a reasonable extent.  Currently, the
1249following types of memory usages are tracked.
1250
1251- Userland memory - page cache and anonymous memory.
1252
1253- Kernel data structures such as dentries and inodes.
1254
1255- TCP socket buffers.
1256
1257The above list may expand in the future for better coverage.
1258
1259
1260Memory Interface Files
1261~~~~~~~~~~~~~~~~~~~~~~
1262
1263All memory amounts are in bytes.  If a value which is not aligned to
1264PAGE_SIZE is written, the value may be rounded up to the closest
1265PAGE_SIZE multiple when read back.
1266
1267  memory.current
1268	A read-only single value file which exists on non-root
1269	cgroups.
1270
1271	The total amount of memory currently being used by the cgroup
1272	and its descendants.
1273
1274  memory.min
1275	A read-write single value file which exists on non-root
1276	cgroups.  The default is "0".
1277
1278	Hard memory protection.  If the memory usage of a cgroup
1279	is within its effective min boundary, the cgroup's memory
1280	won't be reclaimed under any conditions. If there is no
1281	unprotected reclaimable memory available, OOM killer
1282	is invoked. Above the effective min boundary (or
1283	effective low boundary if it is higher), pages are reclaimed
1284	proportionally to the overage, reducing reclaim pressure for
1285	smaller overages.
1286
1287	Effective min boundary is limited by memory.min values of
1288	all ancestor cgroups. If there is memory.min overcommitment
1289	(child cgroup or cgroups are requiring more protected memory
1290	than parent will allow), then each child cgroup will get
1291	the part of parent's protection proportional to its
1292	actual memory usage below memory.min.
1293
1294	Putting more memory than generally available under this
1295	protection is discouraged and may lead to constant OOMs.
1296
1297	If a memory cgroup is not populated with processes,
1298	its memory.min is ignored.
1299
1300  memory.low
1301	A read-write single value file which exists on non-root
1302	cgroups.  The default is "0".
1303
1304	Best-effort memory protection.  If the memory usage of a
1305	cgroup is within its effective low boundary, the cgroup's
1306	memory won't be reclaimed unless there is no reclaimable
1307	memory available in unprotected cgroups.
1308	Above the effective low	boundary (or
1309	effective min boundary if it is higher), pages are reclaimed
1310	proportionally to the overage, reducing reclaim pressure for
1311	smaller overages.
1312
1313	Effective low boundary is limited by memory.low values of
1314	all ancestor cgroups. If there is memory.low overcommitment
1315	(child cgroup or cgroups are requiring more protected memory
1316	than parent will allow), then each child cgroup will get
1317	the part of parent's protection proportional to its
1318	actual memory usage below memory.low.
1319
1320	Putting more memory than generally available under this
1321	protection is discouraged.
1322
1323  memory.high
1324	A read-write single value file which exists on non-root
1325	cgroups.  The default is "max".
1326
1327	Memory usage throttle limit.  If a cgroup's usage goes
1328	over the high boundary, the processes of the cgroup are
1329	throttled and put under heavy reclaim pressure.
1330
1331	Going over the high limit never invokes the OOM killer and
1332	under extreme conditions the limit may be breached. The high
1333	limit should be used in scenarios where an external process
1334	monitors the limited cgroup to alleviate heavy reclaim
1335	pressure.
1336
1337  memory.max
1338	A read-write single value file which exists on non-root
1339	cgroups.  The default is "max".
1340
1341	Memory usage hard limit.  This is the main mechanism to limit
1342	memory usage of a cgroup.  If a cgroup's memory usage reaches
1343	this limit and can't be reduced, the OOM killer is invoked in
1344	the cgroup. Under certain circumstances, the usage may go
1345	over the limit temporarily.
1346
1347	In default configuration regular 0-order allocations always
1348	succeed unless OOM killer chooses current task as a victim.
1349
1350	Some kinds of allocations don't invoke the OOM killer.
1351	Caller could retry them differently, return into userspace
1352	as -ENOMEM or silently ignore in cases like disk readahead.
1353
1354  memory.reclaim
1355	A write-only nested-keyed file which exists for all cgroups.
1356
1357	This is a simple interface to trigger memory reclaim in the
1358	target cgroup.
1359
1360	Example::
1361
1362	  echo "1G" > memory.reclaim
1363
1364	Please note that the kernel can over or under reclaim from
1365	the target cgroup. If less bytes are reclaimed than the
1366	specified amount, -EAGAIN is returned.
1367
1368	Please note that the proactive reclaim (triggered by this
1369	interface) is not meant to indicate memory pressure on the
1370	memory cgroup. Therefore socket memory balancing triggered by
1371	the memory reclaim normally is not exercised in this case.
1372	This means that the networking layer will not adapt based on
1373	reclaim induced by memory.reclaim.
1374
1375The following nested keys are defined.
1376
1377	  ==========            ================================
1378	  swappiness            Swappiness value to reclaim with
1379	  ==========            ================================
1380
1381	Specifying a swappiness value instructs the kernel to perform
1382	the reclaim with that swappiness value. Note that this has the
1383	same semantics as vm.swappiness applied to memcg reclaim with
1384	all the existing limitations and potential future extensions.
1385
1386  memory.peak
1387	A read-write single value file which exists on non-root cgroups.
1388
1389	The max memory usage recorded for the cgroup and its descendants since
1390	either the creation of the cgroup or the most recent reset for that FD.
1391
1392	A write of any non-empty string to this file resets it to the
1393	current memory usage for subsequent reads through the same
1394	file descriptor.
1395
1396  memory.oom.group
1397	A read-write single value file which exists on non-root
1398	cgroups.  The default value is "0".
1399
1400	Determines whether the cgroup should be treated as
1401	an indivisible workload by the OOM killer. If set,
1402	all tasks belonging to the cgroup or to its descendants
1403	(if the memory cgroup is not a leaf cgroup) are killed
1404	together or not at all. This can be used to avoid
1405	partial kills to guarantee workload integrity.
1406
1407	Tasks with the OOM protection (oom_score_adj set to -1000)
1408	are treated as an exception and are never killed.
1409
1410	If the OOM killer is invoked in a cgroup, it's not going
1411	to kill any tasks outside of this cgroup, regardless
1412	memory.oom.group values of ancestor cgroups.
1413
1414  memory.events
1415	A read-only flat-keyed file which exists on non-root cgroups.
1416	The following entries are defined.  Unless specified
1417	otherwise, a value change in this file generates a file
1418	modified event.
1419
1420	Note that all fields in this file are hierarchical and the
1421	file modified event can be generated due to an event down the
1422	hierarchy. For the local events at the cgroup level see
1423	memory.events.local.
1424
1425	  low
1426		The number of times the cgroup is reclaimed due to
1427		high memory pressure even though its usage is under
1428		the low boundary.  This usually indicates that the low
1429		boundary is over-committed.
1430
1431	  high
1432		The number of times processes of the cgroup are
1433		throttled and routed to perform direct memory reclaim
1434		because the high memory boundary was exceeded.  For a
1435		cgroup whose memory usage is capped by the high limit
1436		rather than global memory pressure, this event's
1437		occurrences are expected.
1438
1439	  max
1440		The number of times the cgroup's memory usage was
1441		about to go over the max boundary.  If direct reclaim
1442		fails to bring it down, the cgroup goes to OOM state.
1443
1444	  oom
1445		The number of time the cgroup's memory usage was
1446		reached the limit and allocation was about to fail.
1447
1448		This event is not raised if the OOM killer is not
1449		considered as an option, e.g. for failed high-order
1450		allocations or if caller asked to not retry attempts.
1451
1452	  oom_kill
1453		The number of processes belonging to this cgroup
1454		killed by any kind of OOM killer.
1455
1456          oom_group_kill
1457                The number of times a group OOM has occurred.
1458
1459  memory.events.local
1460	Similar to memory.events but the fields in the file are local
1461	to the cgroup i.e. not hierarchical. The file modified event
1462	generated on this file reflects only the local events.
1463
1464  memory.stat
1465	A read-only flat-keyed file which exists on non-root cgroups.
1466
1467	This breaks down the cgroup's memory footprint into different
1468	types of memory, type-specific details, and other information
1469	on the state and past events of the memory management system.
1470
1471	All memory amounts are in bytes.
1472
1473	The entries are ordered to be human readable, and new entries
1474	can show up in the middle. Don't rely on items remaining in a
1475	fixed position; use the keys to look up specific values!
1476
1477	If the entry has no per-node counter (or not show in the
1478	memory.numa_stat). We use 'npn' (non-per-node) as the tag
1479	to indicate that it will not show in the memory.numa_stat.
1480
1481	  anon
1482		Amount of memory used in anonymous mappings such as
1483		brk(), sbrk(), and mmap(MAP_ANONYMOUS). Note that
1484		some kernel configurations might account complete larger
1485		allocations (e.g., THP) if only some, but not all the
1486		memory of such an allocation is mapped anymore.
1487
1488	  file
1489		Amount of memory used to cache filesystem data,
1490		including tmpfs and shared memory.
1491
1492	  kernel (npn)
1493		Amount of total kernel memory, including
1494		(kernel_stack, pagetables, percpu, vmalloc, slab) in
1495		addition to other kernel memory use cases.
1496
1497	  kernel_stack
1498		Amount of memory allocated to kernel stacks.
1499
1500	  pagetables
1501                Amount of memory allocated for page tables.
1502
1503	  sec_pagetables
1504		Amount of memory allocated for secondary page tables,
1505		this currently includes KVM mmu allocations on x86
1506		and arm64 and IOMMU page tables.
1507
1508	  percpu (npn)
1509		Amount of memory used for storing per-cpu kernel
1510		data structures.
1511
1512	  sock (npn)
1513		Amount of memory used in network transmission buffers
1514
1515	  vmalloc (npn)
1516		Amount of memory used for vmap backed memory.
1517
1518	  shmem
1519		Amount of cached filesystem data that is swap-backed,
1520		such as tmpfs, shm segments, shared anonymous mmap()s
1521
1522	  zswap
1523		Amount of memory consumed by the zswap compression backend.
1524
1525	  zswapped
1526		Amount of application memory swapped out to zswap.
1527
1528	  file_mapped
1529		Amount of cached filesystem data mapped with mmap(). Note
1530		that some kernel configurations might account complete
1531		larger allocations (e.g., THP) if only some, but not
1532		not all the memory of such an allocation is mapped.
1533
1534	  file_dirty
1535		Amount of cached filesystem data that was modified but
1536		not yet written back to disk
1537
1538	  file_writeback
1539		Amount of cached filesystem data that was modified and
1540		is currently being written back to disk
1541
1542	  swapcached
1543		Amount of swap cached in memory. The swapcache is accounted
1544		against both memory and swap usage.
1545
1546	  anon_thp
1547		Amount of memory used in anonymous mappings backed by
1548		transparent hugepages
1549
1550	  file_thp
1551		Amount of cached filesystem data backed by transparent
1552		hugepages
1553
1554	  shmem_thp
1555		Amount of shm, tmpfs, shared anonymous mmap()s backed by
1556		transparent hugepages
1557
1558	  inactive_anon, active_anon, inactive_file, active_file, unevictable
1559		Amount of memory, swap-backed and filesystem-backed,
1560		on the internal memory management lists used by the
1561		page reclaim algorithm.
1562
1563		As these represent internal list state (eg. shmem pages are on anon
1564		memory management lists), inactive_foo + active_foo may not be equal to
1565		the value for the foo counter, since the foo counter is type-based, not
1566		list-based.
1567
1568	  slab_reclaimable
1569		Part of "slab" that might be reclaimed, such as
1570		dentries and inodes.
1571
1572	  slab_unreclaimable
1573		Part of "slab" that cannot be reclaimed on memory
1574		pressure.
1575
1576	  slab (npn)
1577		Amount of memory used for storing in-kernel data
1578		structures.
1579
1580	  workingset_refault_anon
1581		Number of refaults of previously evicted anonymous pages.
1582
1583	  workingset_refault_file
1584		Number of refaults of previously evicted file pages.
1585
1586	  workingset_activate_anon
1587		Number of refaulted anonymous pages that were immediately
1588		activated.
1589
1590	  workingset_activate_file
1591		Number of refaulted file pages that were immediately activated.
1592
1593	  workingset_restore_anon
1594		Number of restored anonymous pages which have been detected as
1595		an active workingset before they got reclaimed.
1596
1597	  workingset_restore_file
1598		Number of restored file pages which have been detected as an
1599		active workingset before they got reclaimed.
1600
1601	  workingset_nodereclaim
1602		Number of times a shadow node has been reclaimed
1603
1604	  pswpin (npn)
1605		Number of pages swapped into memory
1606
1607	  pswpout (npn)
1608		Number of pages swapped out of memory
1609
1610	  pgscan (npn)
1611		Amount of scanned pages (in an inactive LRU list)
1612
1613	  pgsteal (npn)
1614		Amount of reclaimed pages
1615
1616	  pgscan_kswapd (npn)
1617		Amount of scanned pages by kswapd (in an inactive LRU list)
1618
1619	  pgscan_direct (npn)
1620		Amount of scanned pages directly  (in an inactive LRU list)
1621
1622	  pgscan_khugepaged (npn)
1623		Amount of scanned pages by khugepaged  (in an inactive LRU list)
1624
1625	  pgscan_proactive (npn)
1626		Amount of scanned pages proactively (in an inactive LRU list)
1627
1628	  pgsteal_kswapd (npn)
1629		Amount of reclaimed pages by kswapd
1630
1631	  pgsteal_direct (npn)
1632		Amount of reclaimed pages directly
1633
1634	  pgsteal_khugepaged (npn)
1635		Amount of reclaimed pages by khugepaged
1636
1637	  pgsteal_proactive (npn)
1638		Amount of reclaimed pages proactively
1639
1640	  pgfault (npn)
1641		Total number of page faults incurred
1642
1643	  pgmajfault (npn)
1644		Number of major page faults incurred
1645
1646	  pgrefill (npn)
1647		Amount of scanned pages (in an active LRU list)
1648
1649	  pgactivate (npn)
1650		Amount of pages moved to the active LRU list
1651
1652	  pgdeactivate (npn)
1653		Amount of pages moved to the inactive LRU list
1654
1655	  pglazyfree (npn)
1656		Amount of pages postponed to be freed under memory pressure
1657
1658	  pglazyfreed (npn)
1659		Amount of reclaimed lazyfree pages
1660
1661	  swpin_zero
1662		Number of pages swapped into memory and filled with zero, where I/O
1663		was optimized out because the page content was detected to be zero
1664		during swapout.
1665
1666	  swpout_zero
1667		Number of zero-filled pages swapped out with I/O skipped due to the
1668		content being detected as zero.
1669
1670	  zswpin
1671		Number of pages moved in to memory from zswap.
1672
1673	  zswpout
1674		Number of pages moved out of memory to zswap.
1675
1676	  zswpwb
1677		Number of pages written from zswap to swap.
1678
1679	  thp_fault_alloc (npn)
1680		Number of transparent hugepages which were allocated to satisfy
1681		a page fault. This counter is not present when CONFIG_TRANSPARENT_HUGEPAGE
1682                is not set.
1683
1684	  thp_collapse_alloc (npn)
1685		Number of transparent hugepages which were allocated to allow
1686		collapsing an existing range of pages. This counter is not
1687		present when CONFIG_TRANSPARENT_HUGEPAGE is not set.
1688
1689	  thp_swpout (npn)
1690		Number of transparent hugepages which are swapout in one piece
1691		without splitting.
1692
1693	  thp_swpout_fallback (npn)
1694		Number of transparent hugepages which were split before swapout.
1695		Usually because failed to allocate some continuous swap space
1696		for the huge page.
1697
1698	  numa_pages_migrated (npn)
1699		Number of pages migrated by NUMA balancing.
1700
1701	  numa_pte_updates (npn)
1702		Number of pages whose page table entries are modified by
1703		NUMA balancing to produce NUMA hinting faults on access.
1704
1705	  numa_hint_faults (npn)
1706		Number of NUMA hinting faults.
1707
1708	  pgdemote_kswapd
1709		Number of pages demoted by kswapd.
1710
1711	  pgdemote_direct
1712		Number of pages demoted directly.
1713
1714	  pgdemote_khugepaged
1715		Number of pages demoted by khugepaged.
1716
1717	  pgdemote_proactive
1718		Number of pages demoted by proactively.
1719
1720	  hugetlb
1721		Amount of memory used by hugetlb pages. This metric only shows
1722		up if hugetlb usage is accounted for in memory.current (i.e.
1723		cgroup is mounted with the memory_hugetlb_accounting option).
1724
1725  memory.numa_stat
1726	A read-only nested-keyed file which exists on non-root cgroups.
1727
1728	This breaks down the cgroup's memory footprint into different
1729	types of memory, type-specific details, and other information
1730	per node on the state of the memory management system.
1731
1732	This is useful for providing visibility into the NUMA locality
1733	information within an memcg since the pages are allowed to be
1734	allocated from any physical node. One of the use case is evaluating
1735	application performance by combining this information with the
1736	application's CPU allocation.
1737
1738	All memory amounts are in bytes.
1739
1740	The output format of memory.numa_stat is::
1741
1742	  type N0=<bytes in node 0> N1=<bytes in node 1> ...
1743
1744	The entries are ordered to be human readable, and new entries
1745	can show up in the middle. Don't rely on items remaining in a
1746	fixed position; use the keys to look up specific values!
1747
1748	The entries can refer to the memory.stat.
1749
1750  memory.swap.current
1751	A read-only single value file which exists on non-root
1752	cgroups.
1753
1754	The total amount of swap currently being used by the cgroup
1755	and its descendants.
1756
1757  memory.swap.high
1758	A read-write single value file which exists on non-root
1759	cgroups.  The default is "max".
1760
1761	Swap usage throttle limit.  If a cgroup's swap usage exceeds
1762	this limit, all its further allocations will be throttled to
1763	allow userspace to implement custom out-of-memory procedures.
1764
1765	This limit marks a point of no return for the cgroup. It is NOT
1766	designed to manage the amount of swapping a workload does
1767	during regular operation. Compare to memory.swap.max, which
1768	prohibits swapping past a set amount, but lets the cgroup
1769	continue unimpeded as long as other memory can be reclaimed.
1770
1771	Healthy workloads are not expected to reach this limit.
1772
1773  memory.swap.peak
1774	A read-write single value file which exists on non-root cgroups.
1775
1776	The max swap usage recorded for the cgroup and its descendants since
1777	the creation of the cgroup or the most recent reset for that FD.
1778
1779	A write of any non-empty string to this file resets it to the
1780	current memory usage for subsequent reads through the same
1781	file descriptor.
1782
1783  memory.swap.max
1784	A read-write single value file which exists on non-root
1785	cgroups.  The default is "max".
1786
1787	Swap usage hard limit.  If a cgroup's swap usage reaches this
1788	limit, anonymous memory of the cgroup will not be swapped out.
1789
1790  memory.swap.events
1791	A read-only flat-keyed file which exists on non-root cgroups.
1792	The following entries are defined.  Unless specified
1793	otherwise, a value change in this file generates a file
1794	modified event.
1795
1796	  high
1797		The number of times the cgroup's swap usage was over
1798		the high threshold.
1799
1800	  max
1801		The number of times the cgroup's swap usage was about
1802		to go over the max boundary and swap allocation
1803		failed.
1804
1805	  fail
1806		The number of times swap allocation failed either
1807		because of running out of swap system-wide or max
1808		limit.
1809
1810	When reduced under the current usage, the existing swap
1811	entries are reclaimed gradually and the swap usage may stay
1812	higher than the limit for an extended period of time.  This
1813	reduces the impact on the workload and memory management.
1814
1815  memory.zswap.current
1816	A read-only single value file which exists on non-root
1817	cgroups.
1818
1819	The total amount of memory consumed by the zswap compression
1820	backend.
1821
1822  memory.zswap.max
1823	A read-write single value file which exists on non-root
1824	cgroups.  The default is "max".
1825
1826	Zswap usage hard limit. If a cgroup's zswap pool reaches this
1827	limit, it will refuse to take any more stores before existing
1828	entries fault back in or are written out to disk.
1829
1830  memory.zswap.writeback
1831	A read-write single value file. The default value is "1".
1832	Note that this setting is hierarchical, i.e. the writeback would be
1833	implicitly disabled for child cgroups if the upper hierarchy
1834	does so.
1835
1836	When this is set to 0, all swapping attempts to swapping devices
1837	are disabled. This included both zswap writebacks, and swapping due
1838	to zswap store failures. If the zswap store failures are recurring
1839	(for e.g if the pages are incompressible), users can observe
1840	reclaim inefficiency after disabling writeback (because the same
1841	pages might be rejected again and again).
1842
1843	Note that this is subtly different from setting memory.swap.max to
1844	0, as it still allows for pages to be written to the zswap pool.
1845	This setting has no effect if zswap is disabled, and swapping
1846	is allowed unless memory.swap.max is set to 0.
1847
1848  memory.pressure
1849	A read-only nested-keyed file.
1850
1851	Shows pressure stall information for memory. See
1852	:ref:`Documentation/accounting/psi.rst <psi>` for details.
1853
1854
1855Usage Guidelines
1856~~~~~~~~~~~~~~~~
1857
1858"memory.high" is the main mechanism to control memory usage.
1859Over-committing on high limit (sum of high limits > available memory)
1860and letting global memory pressure to distribute memory according to
1861usage is a viable strategy.
1862
1863Because breach of the high limit doesn't trigger the OOM killer but
1864throttles the offending cgroup, a management agent has ample
1865opportunities to monitor and take appropriate actions such as granting
1866more memory or terminating the workload.
1867
1868Determining whether a cgroup has enough memory is not trivial as
1869memory usage doesn't indicate whether the workload can benefit from
1870more memory.  For example, a workload which writes data received from
1871network to a file can use all available memory but can also operate as
1872performant with a small amount of memory.  A measure of memory
1873pressure - how much the workload is being impacted due to lack of
1874memory - is necessary to determine whether a workload needs more
1875memory; unfortunately, memory pressure monitoring mechanism isn't
1876implemented yet.
1877
1878
1879Memory Ownership
1880~~~~~~~~~~~~~~~~
1881
1882A memory area is charged to the cgroup which instantiated it and stays
1883charged to the cgroup until the area is released.  Migrating a process
1884to a different cgroup doesn't move the memory usages that it
1885instantiated while in the previous cgroup to the new cgroup.
1886
1887A memory area may be used by processes belonging to different cgroups.
1888To which cgroup the area will be charged is in-deterministic; however,
1889over time, the memory area is likely to end up in a cgroup which has
1890enough memory allowance to avoid high reclaim pressure.
1891
1892If a cgroup sweeps a considerable amount of memory which is expected
1893to be accessed repeatedly by other cgroups, it may make sense to use
1894POSIX_FADV_DONTNEED to relinquish the ownership of memory areas
1895belonging to the affected files to ensure correct memory ownership.
1896
1897
1898IO
1899--
1900
1901The "io" controller regulates the distribution of IO resources.  This
1902controller implements both weight based and absolute bandwidth or IOPS
1903limit distribution; however, weight based distribution is available
1904only if cfq-iosched is in use and neither scheme is available for
1905blk-mq devices.
1906
1907
1908IO Interface Files
1909~~~~~~~~~~~~~~~~~~
1910
1911  io.stat
1912	A read-only nested-keyed file.
1913
1914	Lines are keyed by $MAJ:$MIN device numbers and not ordered.
1915	The following nested keys are defined.
1916
1917	  ======	=====================
1918	  rbytes	Bytes read
1919	  wbytes	Bytes written
1920	  rios		Number of read IOs
1921	  wios		Number of write IOs
1922	  dbytes	Bytes discarded
1923	  dios		Number of discard IOs
1924	  ======	=====================
1925
1926	An example read output follows::
1927
1928	  8:16 rbytes=1459200 wbytes=314773504 rios=192 wios=353 dbytes=0 dios=0
1929	  8:0 rbytes=90430464 wbytes=299008000 rios=8950 wios=1252 dbytes=50331648 dios=3021
1930
1931  io.cost.qos
1932	A read-write nested-keyed file which exists only on the root
1933	cgroup.
1934
1935	This file configures the Quality of Service of the IO cost
1936	model based controller (CONFIG_BLK_CGROUP_IOCOST) which
1937	currently implements "io.weight" proportional control.  Lines
1938	are keyed by $MAJ:$MIN device numbers and not ordered.  The
1939	line for a given device is populated on the first write for
1940	the device on "io.cost.qos" or "io.cost.model".  The following
1941	nested keys are defined.
1942
1943	  ======	=====================================
1944	  enable	Weight-based control enable
1945	  ctrl		"auto" or "user"
1946	  rpct		Read latency percentile    [0, 100]
1947	  rlat		Read latency threshold
1948	  wpct		Write latency percentile   [0, 100]
1949	  wlat		Write latency threshold
1950	  min		Minimum scaling percentage [1, 10000]
1951	  max		Maximum scaling percentage [1, 10000]
1952	  ======	=====================================
1953
1954	The controller is disabled by default and can be enabled by
1955	setting "enable" to 1.  "rpct" and "wpct" parameters default
1956	to zero and the controller uses internal device saturation
1957	state to adjust the overall IO rate between "min" and "max".
1958
1959	When a better control quality is needed, latency QoS
1960	parameters can be configured.  For example::
1961
1962	  8:16 enable=1 ctrl=auto rpct=95.00 rlat=75000 wpct=95.00 wlat=150000 min=50.00 max=150.0
1963
1964	shows that on sdb, the controller is enabled, will consider
1965	the device saturated if the 95th percentile of read completion
1966	latencies is above 75ms or write 150ms, and adjust the overall
1967	IO issue rate between 50% and 150% accordingly.
1968
1969	The lower the saturation point, the better the latency QoS at
1970	the cost of aggregate bandwidth.  The narrower the allowed
1971	adjustment range between "min" and "max", the more conformant
1972	to the cost model the IO behavior.  Note that the IO issue
1973	base rate may be far off from 100% and setting "min" and "max"
1974	blindly can lead to a significant loss of device capacity or
1975	control quality.  "min" and "max" are useful for regulating
1976	devices which show wide temporary behavior changes - e.g. a
1977	ssd which accepts writes at the line speed for a while and
1978	then completely stalls for multiple seconds.
1979
1980	When "ctrl" is "auto", the parameters are controlled by the
1981	kernel and may change automatically.  Setting "ctrl" to "user"
1982	or setting any of the percentile and latency parameters puts
1983	it into "user" mode and disables the automatic changes.  The
1984	automatic mode can be restored by setting "ctrl" to "auto".
1985
1986  io.cost.model
1987	A read-write nested-keyed file which exists only on the root
1988	cgroup.
1989
1990	This file configures the cost model of the IO cost model based
1991	controller (CONFIG_BLK_CGROUP_IOCOST) which currently
1992	implements "io.weight" proportional control.  Lines are keyed
1993	by $MAJ:$MIN device numbers and not ordered.  The line for a
1994	given device is populated on the first write for the device on
1995	"io.cost.qos" or "io.cost.model".  The following nested keys
1996	are defined.
1997
1998	  =====		================================
1999	  ctrl		"auto" or "user"
2000	  model		The cost model in use - "linear"
2001	  =====		================================
2002
2003	When "ctrl" is "auto", the kernel may change all parameters
2004	dynamically.  When "ctrl" is set to "user" or any other
2005	parameters are written to, "ctrl" become "user" and the
2006	automatic changes are disabled.
2007
2008	When "model" is "linear", the following model parameters are
2009	defined.
2010
2011	  =============	========================================
2012	  [r|w]bps	The maximum sequential IO throughput
2013	  [r|w]seqiops	The maximum 4k sequential IOs per second
2014	  [r|w]randiops	The maximum 4k random IOs per second
2015	  =============	========================================
2016
2017	From the above, the builtin linear model determines the base
2018	costs of a sequential and random IO and the cost coefficient
2019	for the IO size.  While simple, this model can cover most
2020	common device classes acceptably.
2021
2022	The IO cost model isn't expected to be accurate in absolute
2023	sense and is scaled to the device behavior dynamically.
2024
2025	If needed, tools/cgroup/iocost_coef_gen.py can be used to
2026	generate device-specific coefficients.
2027
2028  io.weight
2029	A read-write flat-keyed file which exists on non-root cgroups.
2030	The default is "default 100".
2031
2032	The first line is the default weight applied to devices
2033	without specific override.  The rest are overrides keyed by
2034	$MAJ:$MIN device numbers and not ordered.  The weights are in
2035	the range [1, 10000] and specifies the relative amount IO time
2036	the cgroup can use in relation to its siblings.
2037
2038	The default weight can be updated by writing either "default
2039	$WEIGHT" or simply "$WEIGHT".  Overrides can be set by writing
2040	"$MAJ:$MIN $WEIGHT" and unset by writing "$MAJ:$MIN default".
2041
2042	An example read output follows::
2043
2044	  default 100
2045	  8:16 200
2046	  8:0 50
2047
2048  io.max
2049	A read-write nested-keyed file which exists on non-root
2050	cgroups.
2051
2052	BPS and IOPS based IO limit.  Lines are keyed by $MAJ:$MIN
2053	device numbers and not ordered.  The following nested keys are
2054	defined.
2055
2056	  =====		==================================
2057	  rbps		Max read bytes per second
2058	  wbps		Max write bytes per second
2059	  riops		Max read IO operations per second
2060	  wiops		Max write IO operations per second
2061	  =====		==================================
2062
2063	When writing, any number of nested key-value pairs can be
2064	specified in any order.  "max" can be specified as the value
2065	to remove a specific limit.  If the same key is specified
2066	multiple times, the outcome is undefined.
2067
2068	BPS and IOPS are measured in each IO direction and IOs are
2069	delayed if limit is reached.  Temporary bursts are allowed.
2070
2071	Setting read limit at 2M BPS and write at 120 IOPS for 8:16::
2072
2073	  echo "8:16 rbps=2097152 wiops=120" > io.max
2074
2075	Reading returns the following::
2076
2077	  8:16 rbps=2097152 wbps=max riops=max wiops=120
2078
2079	Write IOPS limit can be removed by writing the following::
2080
2081	  echo "8:16 wiops=max" > io.max
2082
2083	Reading now returns the following::
2084
2085	  8:16 rbps=2097152 wbps=max riops=max wiops=max
2086
2087  io.pressure
2088	A read-only nested-keyed file.
2089
2090	Shows pressure stall information for IO. See
2091	:ref:`Documentation/accounting/psi.rst <psi>` for details.
2092
2093
2094Writeback
2095~~~~~~~~~
2096
2097Page cache is dirtied through buffered writes and shared mmaps and
2098written asynchronously to the backing filesystem by the writeback
2099mechanism.  Writeback sits between the memory and IO domains and
2100regulates the proportion of dirty memory by balancing dirtying and
2101write IOs.
2102
2103The io controller, in conjunction with the memory controller,
2104implements control of page cache writeback IOs.  The memory controller
2105defines the memory domain that dirty memory ratio is calculated and
2106maintained for and the io controller defines the io domain which
2107writes out dirty pages for the memory domain.  Both system-wide and
2108per-cgroup dirty memory states are examined and the more restrictive
2109of the two is enforced.
2110
2111cgroup writeback requires explicit support from the underlying
2112filesystem.  Currently, cgroup writeback is implemented on ext2, ext4,
2113btrfs, f2fs, and xfs.  On other filesystems, all writeback IOs are
2114attributed to the root cgroup.
2115
2116There are inherent differences in memory and writeback management
2117which affects how cgroup ownership is tracked.  Memory is tracked per
2118page while writeback per inode.  For the purpose of writeback, an
2119inode is assigned to a cgroup and all IO requests to write dirty pages
2120from the inode are attributed to that cgroup.
2121
2122As cgroup ownership for memory is tracked per page, there can be pages
2123which are associated with different cgroups than the one the inode is
2124associated with.  These are called foreign pages.  The writeback
2125constantly keeps track of foreign pages and, if a particular foreign
2126cgroup becomes the majority over a certain period of time, switches
2127the ownership of the inode to that cgroup.
2128
2129While this model is enough for most use cases where a given inode is
2130mostly dirtied by a single cgroup even when the main writing cgroup
2131changes over time, use cases where multiple cgroups write to a single
2132inode simultaneously are not supported well.  In such circumstances, a
2133significant portion of IOs are likely to be attributed incorrectly.
2134As memory controller assigns page ownership on the first use and
2135doesn't update it until the page is released, even if writeback
2136strictly follows page ownership, multiple cgroups dirtying overlapping
2137areas wouldn't work as expected.  It's recommended to avoid such usage
2138patterns.
2139
2140The sysctl knobs which affect writeback behavior are applied to cgroup
2141writeback as follows.
2142
2143  vm.dirty_background_ratio, vm.dirty_ratio
2144	These ratios apply the same to cgroup writeback with the
2145	amount of available memory capped by limits imposed by the
2146	memory controller and system-wide clean memory.
2147
2148  vm.dirty_background_bytes, vm.dirty_bytes
2149	For cgroup writeback, this is calculated into ratio against
2150	total available memory and applied the same way as
2151	vm.dirty[_background]_ratio.
2152
2153
2154IO Latency
2155~~~~~~~~~~
2156
2157This is a cgroup v2 controller for IO workload protection.  You provide a group
2158with a latency target, and if the average latency exceeds that target the
2159controller will throttle any peers that have a lower latency target than the
2160protected workload.
2161
2162The limits are only applied at the peer level in the hierarchy.  This means that
2163in the diagram below, only groups A, B, and C will influence each other, and
2164groups D and F will influence each other.  Group G will influence nobody::
2165
2166			[root]
2167		/	   |		\
2168		A	   B		C
2169	       /  \        |
2170	      D    F	   G
2171
2172
2173So the ideal way to configure this is to set io.latency in groups A, B, and C.
2174Generally you do not want to set a value lower than the latency your device
2175supports.  Experiment to find the value that works best for your workload.
2176Start at higher than the expected latency for your device and watch the
2177avg_lat value in io.stat for your workload group to get an idea of the
2178latency you see during normal operation.  Use the avg_lat value as a basis for
2179your real setting, setting at 10-15% higher than the value in io.stat.
2180
2181How IO Latency Throttling Works
2182~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2183
2184io.latency is work conserving; so as long as everybody is meeting their latency
2185target the controller doesn't do anything.  Once a group starts missing its
2186target it begins throttling any peer group that has a higher target than itself.
2187This throttling takes 2 forms:
2188
2189- Queue depth throttling.  This is the number of outstanding IO's a group is
2190  allowed to have.  We will clamp down relatively quickly, starting at no limit
2191  and going all the way down to 1 IO at a time.
2192
2193- Artificial delay induction.  There are certain types of IO that cannot be
2194  throttled without possibly adversely affecting higher priority groups.  This
2195  includes swapping and metadata IO.  These types of IO are allowed to occur
2196  normally, however they are "charged" to the originating group.  If the
2197  originating group is being throttled you will see the use_delay and delay
2198  fields in io.stat increase.  The delay value is how many microseconds that are
2199  being added to any process that runs in this group.  Because this number can
2200  grow quite large if there is a lot of swapping or metadata IO occurring we
2201  limit the individual delay events to 1 second at a time.
2202
2203Once the victimized group starts meeting its latency target again it will start
2204unthrottling any peer groups that were throttled previously.  If the victimized
2205group simply stops doing IO the global counter will unthrottle appropriately.
2206
2207IO Latency Interface Files
2208~~~~~~~~~~~~~~~~~~~~~~~~~~
2209
2210  io.latency
2211	This takes a similar format as the other controllers.
2212
2213		"MAJOR:MINOR target=<target time in microseconds>"
2214
2215  io.stat
2216	If the controller is enabled you will see extra stats in io.stat in
2217	addition to the normal ones.
2218
2219	  depth
2220		This is the current queue depth for the group.
2221
2222	  avg_lat
2223		This is an exponential moving average with a decay rate of 1/exp
2224		bound by the sampling interval.  The decay rate interval can be
2225		calculated by multiplying the win value in io.stat by the
2226		corresponding number of samples based on the win value.
2227
2228	  win
2229		The sampling window size in milliseconds.  This is the minimum
2230		duration of time between evaluation events.  Windows only elapse
2231		with IO activity.  Idle periods extend the most recent window.
2232
2233IO Priority
2234~~~~~~~~~~~
2235
2236A single attribute controls the behavior of the I/O priority cgroup policy,
2237namely the io.prio.class attribute. The following values are accepted for
2238that attribute:
2239
2240  no-change
2241	Do not modify the I/O priority class.
2242
2243  promote-to-rt
2244	For requests that have a non-RT I/O priority class, change it into RT.
2245	Also change the priority level of these requests to 4. Do not modify
2246	the I/O priority of requests that have priority class RT.
2247
2248  restrict-to-be
2249	For requests that do not have an I/O priority class or that have I/O
2250	priority class RT, change it into BE. Also change the priority level
2251	of these requests to 0. Do not modify the I/O priority class of
2252	requests that have priority class IDLE.
2253
2254  idle
2255	Change the I/O priority class of all requests into IDLE, the lowest
2256	I/O priority class.
2257
2258  none-to-rt
2259	Deprecated. Just an alias for promote-to-rt.
2260
2261The following numerical values are associated with the I/O priority policies:
2262
2263+----------------+---+
2264| no-change      | 0 |
2265+----------------+---+
2266| promote-to-rt  | 1 |
2267+----------------+---+
2268| restrict-to-be | 2 |
2269+----------------+---+
2270| idle           | 3 |
2271+----------------+---+
2272
2273The numerical value that corresponds to each I/O priority class is as follows:
2274
2275+-------------------------------+---+
2276| IOPRIO_CLASS_NONE             | 0 |
2277+-------------------------------+---+
2278| IOPRIO_CLASS_RT (real-time)   | 1 |
2279+-------------------------------+---+
2280| IOPRIO_CLASS_BE (best effort) | 2 |
2281+-------------------------------+---+
2282| IOPRIO_CLASS_IDLE             | 3 |
2283+-------------------------------+---+
2284
2285The algorithm to set the I/O priority class for a request is as follows:
2286
2287- If I/O priority class policy is promote-to-rt, change the request I/O
2288  priority class to IOPRIO_CLASS_RT and change the request I/O priority
2289  level to 4.
2290- If I/O priority class policy is not promote-to-rt, translate the I/O priority
2291  class policy into a number, then change the request I/O priority class
2292  into the maximum of the I/O priority class policy number and the numerical
2293  I/O priority class.
2294
2295PID
2296---
2297
2298The process number controller is used to allow a cgroup to stop any
2299new tasks from being fork()'d or clone()'d after a specified limit is
2300reached.
2301
2302The number of tasks in a cgroup can be exhausted in ways which other
2303controllers cannot prevent, thus warranting its own controller.  For
2304example, a fork bomb is likely to exhaust the number of tasks before
2305hitting memory restrictions.
2306
2307Note that PIDs used in this controller refer to TIDs, process IDs as
2308used by the kernel.
2309
2310
2311PID Interface Files
2312~~~~~~~~~~~~~~~~~~~
2313
2314  pids.max
2315	A read-write single value file which exists on non-root
2316	cgroups.  The default is "max".
2317
2318	Hard limit of number of processes.
2319
2320  pids.current
2321	A read-only single value file which exists on non-root cgroups.
2322
2323	The number of processes currently in the cgroup and its
2324	descendants.
2325
2326  pids.peak
2327	A read-only single value file which exists on non-root cgroups.
2328
2329	The maximum value that the number of processes in the cgroup and its
2330	descendants has ever reached.
2331
2332  pids.events
2333	A read-only flat-keyed file which exists on non-root cgroups. Unless
2334	specified otherwise, a value change in this file generates a file
2335	modified event. The following entries are defined.
2336
2337	  max
2338		The number of times the cgroup's total number of processes hit the pids.max
2339		limit (see also pids_localevents).
2340
2341  pids.events.local
2342	Similar to pids.events but the fields in the file are local
2343	to the cgroup i.e. not hierarchical. The file modified event
2344	generated on this file reflects only the local events.
2345
2346Organisational operations are not blocked by cgroup policies, so it is
2347possible to have pids.current > pids.max.  This can be done by either
2348setting the limit to be smaller than pids.current, or attaching enough
2349processes to the cgroup such that pids.current is larger than
2350pids.max.  However, it is not possible to violate a cgroup PID policy
2351through fork() or clone(). These will return -EAGAIN if the creation
2352of a new process would cause a cgroup policy to be violated.
2353
2354
2355Cpuset
2356------
2357
2358The "cpuset" controller provides a mechanism for constraining
2359the CPU and memory node placement of tasks to only the resources
2360specified in the cpuset interface files in a task's current cgroup.
2361This is especially valuable on large NUMA systems where placing jobs
2362on properly sized subsets of the systems with careful processor and
2363memory placement to reduce cross-node memory access and contention
2364can improve overall system performance.
2365
2366The "cpuset" controller is hierarchical.  That means the controller
2367cannot use CPUs or memory nodes not allowed in its parent.
2368
2369
2370Cpuset Interface Files
2371~~~~~~~~~~~~~~~~~~~~~~
2372
2373  cpuset.cpus
2374	A read-write multiple values file which exists on non-root
2375	cpuset-enabled cgroups.
2376
2377	It lists the requested CPUs to be used by tasks within this
2378	cgroup.  The actual list of CPUs to be granted, however, is
2379	subjected to constraints imposed by its parent and can differ
2380	from the requested CPUs.
2381
2382	The CPU numbers are comma-separated numbers or ranges.
2383	For example::
2384
2385	  # cat cpuset.cpus
2386	  0-4,6,8-10
2387
2388	An empty value indicates that the cgroup is using the same
2389	setting as the nearest cgroup ancestor with a non-empty
2390	"cpuset.cpus" or all the available CPUs if none is found.
2391
2392	The value of "cpuset.cpus" stays constant until the next update
2393	and won't be affected by any CPU hotplug events.
2394
2395  cpuset.cpus.effective
2396	A read-only multiple values file which exists on all
2397	cpuset-enabled cgroups.
2398
2399	It lists the onlined CPUs that are actually granted to this
2400	cgroup by its parent.  These CPUs are allowed to be used by
2401	tasks within the current cgroup.
2402
2403	If "cpuset.cpus" is empty, the "cpuset.cpus.effective" file shows
2404	all the CPUs from the parent cgroup that can be available to
2405	be used by this cgroup.  Otherwise, it should be a subset of
2406	"cpuset.cpus" unless none of the CPUs listed in "cpuset.cpus"
2407	can be granted.  In this case, it will be treated just like an
2408	empty "cpuset.cpus".
2409
2410	Its value will be affected by CPU hotplug events.
2411
2412  cpuset.mems
2413	A read-write multiple values file which exists on non-root
2414	cpuset-enabled cgroups.
2415
2416	It lists the requested memory nodes to be used by tasks within
2417	this cgroup.  The actual list of memory nodes granted, however,
2418	is subjected to constraints imposed by its parent and can differ
2419	from the requested memory nodes.
2420
2421	The memory node numbers are comma-separated numbers or ranges.
2422	For example::
2423
2424	  # cat cpuset.mems
2425	  0-1,3
2426
2427	An empty value indicates that the cgroup is using the same
2428	setting as the nearest cgroup ancestor with a non-empty
2429	"cpuset.mems" or all the available memory nodes if none
2430	is found.
2431
2432	The value of "cpuset.mems" stays constant until the next update
2433	and won't be affected by any memory nodes hotplug events.
2434
2435	Setting a non-empty value to "cpuset.mems" causes memory of
2436	tasks within the cgroup to be migrated to the designated nodes if
2437	they are currently using memory outside of the designated nodes.
2438
2439	There is a cost for this memory migration.  The migration
2440	may not be complete and some memory pages may be left behind.
2441	So it is recommended that "cpuset.mems" should be set properly
2442	before spawning new tasks into the cpuset.  Even if there is
2443	a need to change "cpuset.mems" with active tasks, it shouldn't
2444	be done frequently.
2445
2446  cpuset.mems.effective
2447	A read-only multiple values file which exists on all
2448	cpuset-enabled cgroups.
2449
2450	It lists the onlined memory nodes that are actually granted to
2451	this cgroup by its parent. These memory nodes are allowed to
2452	be used by tasks within the current cgroup.
2453
2454	If "cpuset.mems" is empty, it shows all the memory nodes from the
2455	parent cgroup that will be available to be used by this cgroup.
2456	Otherwise, it should be a subset of "cpuset.mems" unless none of
2457	the memory nodes listed in "cpuset.mems" can be granted.  In this
2458	case, it will be treated just like an empty "cpuset.mems".
2459
2460	Its value will be affected by memory nodes hotplug events.
2461
2462  cpuset.cpus.exclusive
2463	A read-write multiple values file which exists on non-root
2464	cpuset-enabled cgroups.
2465
2466	It lists all the exclusive CPUs that are allowed to be used
2467	to create a new cpuset partition.  Its value is not used
2468	unless the cgroup becomes a valid partition root.  See the
2469	"cpuset.cpus.partition" section below for a description of what
2470	a cpuset partition is.
2471
2472	When the cgroup becomes a partition root, the actual exclusive
2473	CPUs that are allocated to that partition are listed in
2474	"cpuset.cpus.exclusive.effective" which may be different
2475	from "cpuset.cpus.exclusive".  If "cpuset.cpus.exclusive"
2476	has previously been set, "cpuset.cpus.exclusive.effective"
2477	is always a subset of it.
2478
2479	Users can manually set it to a value that is different from
2480	"cpuset.cpus".	One constraint in setting it is that the list of
2481	CPUs must be exclusive with respect to "cpuset.cpus.exclusive"
2482	of its sibling.  If "cpuset.cpus.exclusive" of a sibling cgroup
2483	isn't set, its "cpuset.cpus" value, if set, cannot be a subset
2484	of it to leave at least one CPU available when the exclusive
2485	CPUs are taken away.
2486
2487	For a parent cgroup, any one of its exclusive CPUs can only
2488	be distributed to at most one of its child cgroups.  Having an
2489	exclusive CPU appearing in two or more of its child cgroups is
2490	not allowed (the exclusivity rule).  A value that violates the
2491	exclusivity rule will be rejected with a write error.
2492
2493	The root cgroup is a partition root and all its available CPUs
2494	are in its exclusive CPU set.
2495
2496  cpuset.cpus.exclusive.effective
2497	A read-only multiple values file which exists on all non-root
2498	cpuset-enabled cgroups.
2499
2500	This file shows the effective set of exclusive CPUs that
2501	can be used to create a partition root.  The content
2502	of this file will always be a subset of its parent's
2503	"cpuset.cpus.exclusive.effective" if its parent is not the root
2504	cgroup.  It will also be a subset of "cpuset.cpus.exclusive"
2505	if it is set.  If "cpuset.cpus.exclusive" is not set, it is
2506	treated to have an implicit value of "cpuset.cpus" in the
2507	formation of local partition.
2508
2509  cpuset.cpus.isolated
2510	A read-only and root cgroup only multiple values file.
2511
2512	This file shows the set of all isolated CPUs used in existing
2513	isolated partitions. It will be empty if no isolated partition
2514	is created.
2515
2516  cpuset.cpus.partition
2517	A read-write single value file which exists on non-root
2518	cpuset-enabled cgroups.  This flag is owned by the parent cgroup
2519	and is not delegatable.
2520
2521	It accepts only the following input values when written to.
2522
2523	  ==========	=====================================
2524	  "member"	Non-root member of a partition
2525	  "root"	Partition root
2526	  "isolated"	Partition root without load balancing
2527	  ==========	=====================================
2528
2529	A cpuset partition is a collection of cpuset-enabled cgroups with
2530	a partition root at the top of the hierarchy and its descendants
2531	except those that are separate partition roots themselves and
2532	their descendants.  A partition has exclusive access to the
2533	set of exclusive CPUs allocated to it.	Other cgroups outside
2534	of that partition cannot use any CPUs in that set.
2535
2536	There are two types of partitions - local and remote.  A local
2537	partition is one whose parent cgroup is also a valid partition
2538	root.  A remote partition is one whose parent cgroup is not a
2539	valid partition root itself.  Writing to "cpuset.cpus.exclusive"
2540	is optional for the creation of a local partition as its
2541	"cpuset.cpus.exclusive" file will assume an implicit value that
2542	is the same as "cpuset.cpus" if it is not set.	Writing the
2543	proper "cpuset.cpus.exclusive" values down the cgroup hierarchy
2544	before the target partition root is mandatory for the creation
2545	of a remote partition.
2546
2547	Currently, a remote partition cannot be created under a local
2548	partition.  All the ancestors of a remote partition root except
2549	the root cgroup cannot be a partition root.
2550
2551	The root cgroup is always a partition root and its state cannot
2552	be changed.  All other non-root cgroups start out as "member".
2553
2554	When set to "root", the current cgroup is the root of a new
2555	partition or scheduling domain.  The set of exclusive CPUs is
2556	determined by the value of its "cpuset.cpus.exclusive.effective".
2557
2558	When set to "isolated", the CPUs in that partition will be in
2559	an isolated state without any load balancing from the scheduler
2560	and excluded from the unbound workqueues.  Tasks placed in such
2561	a partition with multiple CPUs should be carefully distributed
2562	and bound to each of the individual CPUs for optimal performance.
2563
2564	A partition root ("root" or "isolated") can be in one of the
2565	two possible states - valid or invalid.  An invalid partition
2566	root is in a degraded state where some state information may
2567	be retained, but behaves more like a "member".
2568
2569	All possible state transitions among "member", "root" and
2570	"isolated" are allowed.
2571
2572	On read, the "cpuset.cpus.partition" file can show the following
2573	values.
2574
2575	  =============================	=====================================
2576	  "member"			Non-root member of a partition
2577	  "root"			Partition root
2578	  "isolated"			Partition root without load balancing
2579	  "root invalid (<reason>)"	Invalid partition root
2580	  "isolated invalid (<reason>)"	Invalid isolated partition root
2581	  =============================	=====================================
2582
2583	In the case of an invalid partition root, a descriptive string on
2584	why the partition is invalid is included within parentheses.
2585
2586	For a local partition root to be valid, the following conditions
2587	must be met.
2588
2589	1) The parent cgroup is a valid partition root.
2590	2) The "cpuset.cpus.exclusive.effective" file cannot be empty,
2591	   though it may contain offline CPUs.
2592	3) The "cpuset.cpus.effective" cannot be empty unless there is
2593	   no task associated with this partition.
2594
2595	For a remote partition root to be valid, all the above conditions
2596	except the first one must be met.
2597
2598	External events like hotplug or changes to "cpuset.cpus" or
2599	"cpuset.cpus.exclusive" can cause a valid partition root to
2600	become invalid and vice versa.	Note that a task cannot be
2601	moved to a cgroup with empty "cpuset.cpus.effective".
2602
2603	A valid non-root parent partition may distribute out all its CPUs
2604	to its child local partitions when there is no task associated
2605	with it.
2606
2607	Care must be taken to change a valid partition root to "member"
2608	as all its child local partitions, if present, will become
2609	invalid causing disruption to tasks running in those child
2610	partitions. These inactivated partitions could be recovered if
2611	their parent is switched back to a partition root with a proper
2612	value in "cpuset.cpus" or "cpuset.cpus.exclusive".
2613
2614	Poll and inotify events are triggered whenever the state of
2615	"cpuset.cpus.partition" changes.  That includes changes caused
2616	by write to "cpuset.cpus.partition", cpu hotplug or other
2617	changes that modify the validity status of the partition.
2618	This will allow user space agents to monitor unexpected changes
2619	to "cpuset.cpus.partition" without the need to do continuous
2620	polling.
2621
2622	A user can pre-configure certain CPUs to an isolated state
2623	with load balancing disabled at boot time with the "isolcpus"
2624	kernel boot command line option.  If those CPUs are to be put
2625	into a partition, they have to be used in an isolated partition.
2626
2627
2628Device controller
2629-----------------
2630
2631Device controller manages access to device files. It includes both
2632creation of new device files (using mknod), and access to the
2633existing device files.
2634
2635Cgroup v2 device controller has no interface files and is implemented
2636on top of cgroup BPF. To control access to device files, a user may
2637create bpf programs of type BPF_PROG_TYPE_CGROUP_DEVICE and attach
2638them to cgroups with BPF_CGROUP_DEVICE flag. On an attempt to access a
2639device file, corresponding BPF programs will be executed, and depending
2640on the return value the attempt will succeed or fail with -EPERM.
2641
2642A BPF_PROG_TYPE_CGROUP_DEVICE program takes a pointer to the
2643bpf_cgroup_dev_ctx structure, which describes the device access attempt:
2644access type (mknod/read/write) and device (type, major and minor numbers).
2645If the program returns 0, the attempt fails with -EPERM, otherwise it
2646succeeds.
2647
2648An example of BPF_PROG_TYPE_CGROUP_DEVICE program may be found in
2649tools/testing/selftests/bpf/progs/dev_cgroup.c in the kernel source tree.
2650
2651
2652RDMA
2653----
2654
2655The "rdma" controller regulates the distribution and accounting of
2656RDMA resources.
2657
2658RDMA Interface Files
2659~~~~~~~~~~~~~~~~~~~~
2660
2661  rdma.max
2662	A readwrite nested-keyed file that exists for all the cgroups
2663	except root that describes current configured resource limit
2664	for a RDMA/IB device.
2665
2666	Lines are keyed by device name and are not ordered.
2667	Each line contains space separated resource name and its configured
2668	limit that can be distributed.
2669
2670	The following nested keys are defined.
2671
2672	  ==========	=============================
2673	  hca_handle	Maximum number of HCA Handles
2674	  hca_object 	Maximum number of HCA Objects
2675	  ==========	=============================
2676
2677	An example for mlx4 and ocrdma device follows::
2678
2679	  mlx4_0 hca_handle=2 hca_object=2000
2680	  ocrdma1 hca_handle=3 hca_object=max
2681
2682  rdma.current
2683	A read-only file that describes current resource usage.
2684	It exists for all the cgroup except root.
2685
2686	An example for mlx4 and ocrdma device follows::
2687
2688	  mlx4_0 hca_handle=1 hca_object=20
2689	  ocrdma1 hca_handle=1 hca_object=23
2690
2691DMEM
2692----
2693
2694The "dmem" controller regulates the distribution and accounting of
2695device memory regions. Because each memory region may have its own page size,
2696which does not have to be equal to the system page size, the units are always bytes.
2697
2698DMEM Interface Files
2699~~~~~~~~~~~~~~~~~~~~
2700
2701  dmem.max, dmem.min, dmem.low
2702	A readwrite nested-keyed file that exists for all the cgroups
2703	except root that describes current configured resource limit
2704	for a region.
2705
2706	An example for xe follows::
2707
2708	  drm/0000:03:00.0/vram0 1073741824
2709	  drm/0000:03:00.0/stolen max
2710
2711	The semantics are the same as for the memory cgroup controller, and are
2712	calculated in the same way.
2713
2714  dmem.capacity
2715	A read-only file that describes maximum region capacity.
2716	It only exists on the root cgroup. Not all memory can be
2717	allocated by cgroups, as the kernel reserves some for
2718	internal use.
2719
2720	An example for xe follows::
2721
2722	  drm/0000:03:00.0/vram0 8514437120
2723	  drm/0000:03:00.0/stolen 67108864
2724
2725  dmem.current
2726	A read-only file that describes current resource usage.
2727	It exists for all the cgroup except root.
2728
2729	An example for xe follows::
2730
2731	  drm/0000:03:00.0/vram0 12550144
2732	  drm/0000:03:00.0/stolen 8650752
2733
2734HugeTLB
2735-------
2736
2737The HugeTLB controller allows to limit the HugeTLB usage per control group and
2738enforces the controller limit during page fault.
2739
2740HugeTLB Interface Files
2741~~~~~~~~~~~~~~~~~~~~~~~
2742
2743  hugetlb.<hugepagesize>.current
2744	Show current usage for "hugepagesize" hugetlb.  It exists for all
2745	the cgroup except root.
2746
2747  hugetlb.<hugepagesize>.max
2748	Set/show the hard limit of "hugepagesize" hugetlb usage.
2749	The default value is "max".  It exists for all the cgroup except root.
2750
2751  hugetlb.<hugepagesize>.events
2752	A read-only flat-keyed file which exists on non-root cgroups.
2753
2754	  max
2755		The number of allocation failure due to HugeTLB limit
2756
2757  hugetlb.<hugepagesize>.events.local
2758	Similar to hugetlb.<hugepagesize>.events but the fields in the file
2759	are local to the cgroup i.e. not hierarchical. The file modified event
2760	generated on this file reflects only the local events.
2761
2762  hugetlb.<hugepagesize>.numa_stat
2763	Similar to memory.numa_stat, it shows the numa information of the
2764        hugetlb pages of <hugepagesize> in this cgroup.  Only active in
2765        use hugetlb pages are included.  The per-node values are in bytes.
2766
2767Misc
2768----
2769
2770The Miscellaneous cgroup provides the resource limiting and tracking
2771mechanism for the scalar resources which cannot be abstracted like the other
2772cgroup resources. Controller is enabled by the CONFIG_CGROUP_MISC config
2773option.
2774
2775A resource can be added to the controller via enum misc_res_type{} in the
2776include/linux/misc_cgroup.h file and the corresponding name via misc_res_name[]
2777in the kernel/cgroup/misc.c file. Provider of the resource must set its
2778capacity prior to using the resource by calling misc_cg_set_capacity().
2779
2780Once a capacity is set then the resource usage can be updated using charge and
2781uncharge APIs. All of the APIs to interact with misc controller are in
2782include/linux/misc_cgroup.h.
2783
2784Misc Interface Files
2785~~~~~~~~~~~~~~~~~~~~
2786
2787Miscellaneous controller provides 3 interface files. If two misc resources (res_a and res_b) are registered then:
2788
2789  misc.capacity
2790        A read-only flat-keyed file shown only in the root cgroup.  It shows
2791        miscellaneous scalar resources available on the platform along with
2792        their quantities::
2793
2794	  $ cat misc.capacity
2795	  res_a 50
2796	  res_b 10
2797
2798  misc.current
2799        A read-only flat-keyed file shown in the all cgroups.  It shows
2800        the current usage of the resources in the cgroup and its children.::
2801
2802	  $ cat misc.current
2803	  res_a 3
2804	  res_b 0
2805
2806  misc.peak
2807        A read-only flat-keyed file shown in all cgroups.  It shows the
2808        historical maximum usage of the resources in the cgroup and its
2809        children.::
2810
2811	  $ cat misc.peak
2812	  res_a 10
2813	  res_b 8
2814
2815  misc.max
2816        A read-write flat-keyed file shown in the non root cgroups. Allowed
2817        maximum usage of the resources in the cgroup and its children.::
2818
2819	  $ cat misc.max
2820	  res_a max
2821	  res_b 4
2822
2823	Limit can be set by::
2824
2825	  # echo res_a 1 > misc.max
2826
2827	Limit can be set to max by::
2828
2829	  # echo res_a max > misc.max
2830
2831        Limits can be set higher than the capacity value in the misc.capacity
2832        file.
2833
2834  misc.events
2835	A read-only flat-keyed file which exists on non-root cgroups. The
2836	following entries are defined. Unless specified otherwise, a value
2837	change in this file generates a file modified event. All fields in
2838	this file are hierarchical.
2839
2840	  max
2841		The number of times the cgroup's resource usage was
2842		about to go over the max boundary.
2843
2844  misc.events.local
2845        Similar to misc.events but the fields in the file are local to the
2846        cgroup i.e. not hierarchical. The file modified event generated on
2847        this file reflects only the local events.
2848
2849Migration and Ownership
2850~~~~~~~~~~~~~~~~~~~~~~~
2851
2852A miscellaneous scalar resource is charged to the cgroup in which it is used
2853first, and stays charged to that cgroup until that resource is freed. Migrating
2854a process to a different cgroup does not move the charge to the destination
2855cgroup where the process has moved.
2856
2857Others
2858------
2859
2860perf_event
2861~~~~~~~~~~
2862
2863perf_event controller, if not mounted on a legacy hierarchy, is
2864automatically enabled on the v2 hierarchy so that perf events can
2865always be filtered by cgroup v2 path.  The controller can still be
2866moved to a legacy hierarchy after v2 hierarchy is populated.
2867
2868
2869Non-normative information
2870-------------------------
2871
2872This section contains information that isn't considered to be a part of
2873the stable kernel API and so is subject to change.
2874
2875
2876CPU controller root cgroup process behaviour
2877~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2878
2879When distributing CPU cycles in the root cgroup each thread in this
2880cgroup is treated as if it was hosted in a separate child cgroup of the
2881root cgroup. This child cgroup weight is dependent on its thread nice
2882level.
2883
2884For details of this mapping see sched_prio_to_weight array in
2885kernel/sched/core.c file (values from this array should be scaled
2886appropriately so the neutral - nice 0 - value is 100 instead of 1024).
2887
2888
2889IO controller root cgroup process behaviour
2890~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2891
2892Root cgroup processes are hosted in an implicit leaf child node.
2893When distributing IO resources this implicit child node is taken into
2894account as if it was a normal child cgroup of the root cgroup with a
2895weight value of 200.
2896
2897
2898Namespace
2899=========
2900
2901Basics
2902------
2903
2904cgroup namespace provides a mechanism to virtualize the view of the
2905"/proc/$PID/cgroup" file and cgroup mounts.  The CLONE_NEWCGROUP clone
2906flag can be used with clone(2) and unshare(2) to create a new cgroup
2907namespace.  The process running inside the cgroup namespace will have
2908its "/proc/$PID/cgroup" output restricted to cgroupns root.  The
2909cgroupns root is the cgroup of the process at the time of creation of
2910the cgroup namespace.
2911
2912Without cgroup namespace, the "/proc/$PID/cgroup" file shows the
2913complete path of the cgroup of a process.  In a container setup where
2914a set of cgroups and namespaces are intended to isolate processes the
2915"/proc/$PID/cgroup" file may leak potential system level information
2916to the isolated processes.  For example::
2917
2918  # cat /proc/self/cgroup
2919  0::/batchjobs/container_id1
2920
2921The path '/batchjobs/container_id1' can be considered as system-data
2922and undesirable to expose to the isolated processes.  cgroup namespace
2923can be used to restrict visibility of this path.  For example, before
2924creating a cgroup namespace, one would see::
2925
2926  # ls -l /proc/self/ns/cgroup
2927  lrwxrwxrwx 1 root root 0 2014-07-15 10:37 /proc/self/ns/cgroup -> cgroup:[4026531835]
2928  # cat /proc/self/cgroup
2929  0::/batchjobs/container_id1
2930
2931After unsharing a new namespace, the view changes::
2932
2933  # ls -l /proc/self/ns/cgroup
2934  lrwxrwxrwx 1 root root 0 2014-07-15 10:35 /proc/self/ns/cgroup -> cgroup:[4026532183]
2935  # cat /proc/self/cgroup
2936  0::/
2937
2938When some thread from a multi-threaded process unshares its cgroup
2939namespace, the new cgroupns gets applied to the entire process (all
2940the threads).  This is natural for the v2 hierarchy; however, for the
2941legacy hierarchies, this may be unexpected.
2942
2943A cgroup namespace is alive as long as there are processes inside or
2944mounts pinning it.  When the last usage goes away, the cgroup
2945namespace is destroyed.  The cgroupns root and the actual cgroups
2946remain.
2947
2948
2949The Root and Views
2950------------------
2951
2952The 'cgroupns root' for a cgroup namespace is the cgroup in which the
2953process calling unshare(2) is running.  For example, if a process in
2954/batchjobs/container_id1 cgroup calls unshare, cgroup
2955/batchjobs/container_id1 becomes the cgroupns root.  For the
2956init_cgroup_ns, this is the real root ('/') cgroup.
2957
2958The cgroupns root cgroup does not change even if the namespace creator
2959process later moves to a different cgroup::
2960
2961  # ~/unshare -c # unshare cgroupns in some cgroup
2962  # cat /proc/self/cgroup
2963  0::/
2964  # mkdir sub_cgrp_1
2965  # echo 0 > sub_cgrp_1/cgroup.procs
2966  # cat /proc/self/cgroup
2967  0::/sub_cgrp_1
2968
2969Each process gets its namespace-specific view of "/proc/$PID/cgroup"
2970
2971Processes running inside the cgroup namespace will be able to see
2972cgroup paths (in /proc/self/cgroup) only inside their root cgroup.
2973From within an unshared cgroupns::
2974
2975  # sleep 100000 &
2976  [1] 7353
2977  # echo 7353 > sub_cgrp_1/cgroup.procs
2978  # cat /proc/7353/cgroup
2979  0::/sub_cgrp_1
2980
2981From the initial cgroup namespace, the real cgroup path will be
2982visible::
2983
2984  $ cat /proc/7353/cgroup
2985  0::/batchjobs/container_id1/sub_cgrp_1
2986
2987From a sibling cgroup namespace (that is, a namespace rooted at a
2988different cgroup), the cgroup path relative to its own cgroup
2989namespace root will be shown.  For instance, if PID 7353's cgroup
2990namespace root is at '/batchjobs/container_id2', then it will see::
2991
2992  # cat /proc/7353/cgroup
2993  0::/../container_id2/sub_cgrp_1
2994
2995Note that the relative path always starts with '/' to indicate that
2996its relative to the cgroup namespace root of the caller.
2997
2998
2999Migration and setns(2)
3000----------------------
3001
3002Processes inside a cgroup namespace can move into and out of the
3003namespace root if they have proper access to external cgroups.  For
3004example, from inside a namespace with cgroupns root at
3005/batchjobs/container_id1, and assuming that the global hierarchy is
3006still accessible inside cgroupns::
3007
3008  # cat /proc/7353/cgroup
3009  0::/sub_cgrp_1
3010  # echo 7353 > batchjobs/container_id2/cgroup.procs
3011  # cat /proc/7353/cgroup
3012  0::/../container_id2
3013
3014Note that this kind of setup is not encouraged.  A task inside cgroup
3015namespace should only be exposed to its own cgroupns hierarchy.
3016
3017setns(2) to another cgroup namespace is allowed when:
3018
3019(a) the process has CAP_SYS_ADMIN against its current user namespace
3020(b) the process has CAP_SYS_ADMIN against the target cgroup
3021    namespace's userns
3022
3023No implicit cgroup changes happen with attaching to another cgroup
3024namespace.  It is expected that the someone moves the attaching
3025process under the target cgroup namespace root.
3026
3027
3028Interaction with Other Namespaces
3029---------------------------------
3030
3031Namespace specific cgroup hierarchy can be mounted by a process
3032running inside a non-init cgroup namespace::
3033
3034  # mount -t cgroup2 none $MOUNT_POINT
3035
3036This will mount the unified cgroup hierarchy with cgroupns root as the
3037filesystem root.  The process needs CAP_SYS_ADMIN against its user and
3038mount namespaces.
3039
3040The virtualization of /proc/self/cgroup file combined with restricting
3041the view of cgroup hierarchy by namespace-private cgroupfs mount
3042provides a properly isolated cgroup view inside the container.
3043
3044
3045Information on Kernel Programming
3046=================================
3047
3048This section contains kernel programming information in the areas
3049where interacting with cgroup is necessary.  cgroup core and
3050controllers are not covered.
3051
3052
3053Filesystem Support for Writeback
3054--------------------------------
3055
3056A filesystem can support cgroup writeback by updating
3057address_space_operations->writepages() to annotate bio's using the
3058following two functions.
3059
3060  wbc_init_bio(@wbc, @bio)
3061	Should be called for each bio carrying writeback data and
3062	associates the bio with the inode's owner cgroup and the
3063	corresponding request queue.  This must be called after
3064	a queue (device) has been associated with the bio and
3065	before submission.
3066
3067  wbc_account_cgroup_owner(@wbc, @folio, @bytes)
3068	Should be called for each data segment being written out.
3069	While this function doesn't care exactly when it's called
3070	during the writeback session, it's the easiest and most
3071	natural to call it as data segments are added to a bio.
3072
3073With writeback bio's annotated, cgroup support can be enabled per
3074super_block by setting SB_I_CGROUPWB in ->s_iflags.  This allows for
3075selective disabling of cgroup writeback support which is helpful when
3076certain filesystem features, e.g. journaled data mode, are
3077incompatible.
3078
3079wbc_init_bio() binds the specified bio to its cgroup.  Depending on
3080the configuration, the bio may be executed at a lower priority and if
3081the writeback session is holding shared resources, e.g. a journal
3082entry, may lead to priority inversion.  There is no one easy solution
3083for the problem.  Filesystems can try to work around specific problem
3084cases by skipping wbc_init_bio() and using bio_associate_blkg()
3085directly.
3086
3087
3088Deprecated v1 Core Features
3089===========================
3090
3091- Multiple hierarchies including named ones are not supported.
3092
3093- All v1 mount options are not supported.
3094
3095- The "tasks" file is removed and "cgroup.procs" is not sorted.
3096
3097- "cgroup.clone_children" is removed.
3098
3099- /proc/cgroups is meaningless for v2.  Use "cgroup.controllers" or
3100  "cgroup.stat" files at the root instead.
3101
3102
3103Issues with v1 and Rationales for v2
3104====================================
3105
3106Multiple Hierarchies
3107--------------------
3108
3109cgroup v1 allowed an arbitrary number of hierarchies and each
3110hierarchy could host any number of controllers.  While this seemed to
3111provide a high level of flexibility, it wasn't useful in practice.
3112
3113For example, as there is only one instance of each controller, utility
3114type controllers such as freezer which can be useful in all
3115hierarchies could only be used in one.  The issue is exacerbated by
3116the fact that controllers couldn't be moved to another hierarchy once
3117hierarchies were populated.  Another issue was that all controllers
3118bound to a hierarchy were forced to have exactly the same view of the
3119hierarchy.  It wasn't possible to vary the granularity depending on
3120the specific controller.
3121
3122In practice, these issues heavily limited which controllers could be
3123put on the same hierarchy and most configurations resorted to putting
3124each controller on its own hierarchy.  Only closely related ones, such
3125as the cpu and cpuacct controllers, made sense to be put on the same
3126hierarchy.  This often meant that userland ended up managing multiple
3127similar hierarchies repeating the same steps on each hierarchy
3128whenever a hierarchy management operation was necessary.
3129
3130Furthermore, support for multiple hierarchies came at a steep cost.
3131It greatly complicated cgroup core implementation but more importantly
3132the support for multiple hierarchies restricted how cgroup could be
3133used in general and what controllers was able to do.
3134
3135There was no limit on how many hierarchies there might be, which meant
3136that a thread's cgroup membership couldn't be described in finite
3137length.  The key might contain any number of entries and was unlimited
3138in length, which made it highly awkward to manipulate and led to
3139addition of controllers which existed only to identify membership,
3140which in turn exacerbated the original problem of proliferating number
3141of hierarchies.
3142
3143Also, as a controller couldn't have any expectation regarding the
3144topologies of hierarchies other controllers might be on, each
3145controller had to assume that all other controllers were attached to
3146completely orthogonal hierarchies.  This made it impossible, or at
3147least very cumbersome, for controllers to cooperate with each other.
3148
3149In most use cases, putting controllers on hierarchies which are
3150completely orthogonal to each other isn't necessary.  What usually is
3151called for is the ability to have differing levels of granularity
3152depending on the specific controller.  In other words, hierarchy may
3153be collapsed from leaf towards root when viewed from specific
3154controllers.  For example, a given configuration might not care about
3155how memory is distributed beyond a certain level while still wanting
3156to control how CPU cycles are distributed.
3157
3158
3159Thread Granularity
3160------------------
3161
3162cgroup v1 allowed threads of a process to belong to different cgroups.
3163This didn't make sense for some controllers and those controllers
3164ended up implementing different ways to ignore such situations but
3165much more importantly it blurred the line between API exposed to
3166individual applications and system management interface.
3167
3168Generally, in-process knowledge is available only to the process
3169itself; thus, unlike service-level organization of processes,
3170categorizing threads of a process requires active participation from
3171the application which owns the target process.
3172
3173cgroup v1 had an ambiguously defined delegation model which got abused
3174in combination with thread granularity.  cgroups were delegated to
3175individual applications so that they can create and manage their own
3176sub-hierarchies and control resource distributions along them.  This
3177effectively raised cgroup to the status of a syscall-like API exposed
3178to lay programs.
3179
3180First of all, cgroup has a fundamentally inadequate interface to be
3181exposed this way.  For a process to access its own knobs, it has to
3182extract the path on the target hierarchy from /proc/self/cgroup,
3183construct the path by appending the name of the knob to the path, open
3184and then read and/or write to it.  This is not only extremely clunky
3185and unusual but also inherently racy.  There is no conventional way to
3186define transaction across the required steps and nothing can guarantee
3187that the process would actually be operating on its own sub-hierarchy.
3188
3189cgroup controllers implemented a number of knobs which would never be
3190accepted as public APIs because they were just adding control knobs to
3191system-management pseudo filesystem.  cgroup ended up with interface
3192knobs which were not properly abstracted or refined and directly
3193revealed kernel internal details.  These knobs got exposed to
3194individual applications through the ill-defined delegation mechanism
3195effectively abusing cgroup as a shortcut to implementing public APIs
3196without going through the required scrutiny.
3197
3198This was painful for both userland and kernel.  Userland ended up with
3199misbehaving and poorly abstracted interfaces and kernel exposing and
3200locked into constructs inadvertently.
3201
3202
3203Competition Between Inner Nodes and Threads
3204-------------------------------------------
3205
3206cgroup v1 allowed threads to be in any cgroups which created an
3207interesting problem where threads belonging to a parent cgroup and its
3208children cgroups competed for resources.  This was nasty as two
3209different types of entities competed and there was no obvious way to
3210settle it.  Different controllers did different things.
3211
3212The cpu controller considered threads and cgroups as equivalents and
3213mapped nice levels to cgroup weights.  This worked for some cases but
3214fell flat when children wanted to be allocated specific ratios of CPU
3215cycles and the number of internal threads fluctuated - the ratios
3216constantly changed as the number of competing entities fluctuated.
3217There also were other issues.  The mapping from nice level to weight
3218wasn't obvious or universal, and there were various other knobs which
3219simply weren't available for threads.
3220
3221The io controller implicitly created a hidden leaf node for each
3222cgroup to host the threads.  The hidden leaf had its own copies of all
3223the knobs with ``leaf_`` prefixed.  While this allowed equivalent
3224control over internal threads, it was with serious drawbacks.  It
3225always added an extra layer of nesting which wouldn't be necessary
3226otherwise, made the interface messy and significantly complicated the
3227implementation.
3228
3229The memory controller didn't have a way to control what happened
3230between internal tasks and child cgroups and the behavior was not
3231clearly defined.  There were attempts to add ad-hoc behaviors and
3232knobs to tailor the behavior to specific workloads which would have
3233led to problems extremely difficult to resolve in the long term.
3234
3235Multiple controllers struggled with internal tasks and came up with
3236different ways to deal with it; unfortunately, all the approaches were
3237severely flawed and, furthermore, the widely different behaviors
3238made cgroup as a whole highly inconsistent.
3239
3240This clearly is a problem which needs to be addressed from cgroup core
3241in a uniform way.
3242
3243
3244Other Interface Issues
3245----------------------
3246
3247cgroup v1 grew without oversight and developed a large number of
3248idiosyncrasies and inconsistencies.  One issue on the cgroup core side
3249was how an empty cgroup was notified - a userland helper binary was
3250forked and executed for each event.  The event delivery wasn't
3251recursive or delegatable.  The limitations of the mechanism also led
3252to in-kernel event delivery filtering mechanism further complicating
3253the interface.
3254
3255Controller interfaces were problematic too.  An extreme example is
3256controllers completely ignoring hierarchical organization and treating
3257all cgroups as if they were all located directly under the root
3258cgroup.  Some controllers exposed a large amount of inconsistent
3259implementation details to userland.
3260
3261There also was no consistency across controllers.  When a new cgroup
3262was created, some controllers defaulted to not imposing extra
3263restrictions while others disallowed any resource usage until
3264explicitly configured.  Configuration knobs for the same type of
3265control used widely differing naming schemes and formats.  Statistics
3266and information knobs were named arbitrarily and used different
3267formats and units even in the same controller.
3268
3269cgroup v2 establishes common conventions where appropriate and updates
3270controllers so that they expose minimal and consistent interfaces.
3271
3272
3273Controller Issues and Remedies
3274------------------------------
3275
3276Memory
3277~~~~~~
3278
3279The original lower boundary, the soft limit, is defined as a limit
3280that is per default unset.  As a result, the set of cgroups that
3281global reclaim prefers is opt-in, rather than opt-out.  The costs for
3282optimizing these mostly negative lookups are so high that the
3283implementation, despite its enormous size, does not even provide the
3284basic desirable behavior.  First off, the soft limit has no
3285hierarchical meaning.  All configured groups are organized in a global
3286rbtree and treated like equal peers, regardless where they are located
3287in the hierarchy.  This makes subtree delegation impossible.  Second,
3288the soft limit reclaim pass is so aggressive that it not just
3289introduces high allocation latencies into the system, but also impacts
3290system performance due to overreclaim, to the point where the feature
3291becomes self-defeating.
3292
3293The memory.low boundary on the other hand is a top-down allocated
3294reserve.  A cgroup enjoys reclaim protection when it's within its
3295effective low, which makes delegation of subtrees possible. It also
3296enjoys having reclaim pressure proportional to its overage when
3297above its effective low.
3298
3299The original high boundary, the hard limit, is defined as a strict
3300limit that can not budge, even if the OOM killer has to be called.
3301But this generally goes against the goal of making the most out of the
3302available memory.  The memory consumption of workloads varies during
3303runtime, and that requires users to overcommit.  But doing that with a
3304strict upper limit requires either a fairly accurate prediction of the
3305working set size or adding slack to the limit.  Since working set size
3306estimation is hard and error prone, and getting it wrong results in
3307OOM kills, most users tend to err on the side of a looser limit and
3308end up wasting precious resources.
3309
3310The memory.high boundary on the other hand can be set much more
3311conservatively.  When hit, it throttles allocations by forcing them
3312into direct reclaim to work off the excess, but it never invokes the
3313OOM killer.  As a result, a high boundary that is chosen too
3314aggressively will not terminate the processes, but instead it will
3315lead to gradual performance degradation.  The user can monitor this
3316and make corrections until the minimal memory footprint that still
3317gives acceptable performance is found.
3318
3319In extreme cases, with many concurrent allocations and a complete
3320breakdown of reclaim progress within the group, the high boundary can
3321be exceeded.  But even then it's mostly better to satisfy the
3322allocation from the slack available in other groups or the rest of the
3323system than killing the group.  Otherwise, memory.max is there to
3324limit this type of spillover and ultimately contain buggy or even
3325malicious applications.
3326
3327Setting the original memory.limit_in_bytes below the current usage was
3328subject to a race condition, where concurrent charges could cause the
3329limit setting to fail. memory.max on the other hand will first set the
3330limit to prevent new charges, and then reclaim and OOM kill until the
3331new limit is met - or the task writing to memory.max is killed.
3332
3333The combined memory+swap accounting and limiting is replaced by real
3334control over swap space.
3335
3336The main argument for a combined memory+swap facility in the original
3337cgroup design was that global or parental pressure would always be
3338able to swap all anonymous memory of a child group, regardless of the
3339child's own (possibly untrusted) configuration.  However, untrusted
3340groups can sabotage swapping by other means - such as referencing its
3341anonymous memory in a tight loop - and an admin can not assume full
3342swappability when overcommitting untrusted jobs.
3343
3344For trusted jobs, on the other hand, a combined counter is not an
3345intuitive userspace interface, and it flies in the face of the idea
3346that cgroup controllers should account and limit specific physical
3347resources.  Swap space is a resource like all others in the system,
3348and that's why unified hierarchy allows distributing it separately.
3349