xref: /linux/Documentation/admin-guide/cgroup-v2.rst (revision 0678df8271820bcf8fb4f877129f05d68a237de4)
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. HugeTLB
68       5.8-1. HugeTLB Interface Files
69     5-9. Misc
70       5.9-1 Miscellaneous cgroup Interface Files
71       5.9-2 Migration and Ownership
72     5-10. Others
73       5-10-1. perf_event
74     5-N. Non-normative information
75       5-N-1. CPU controller root cgroup process behaviour
76       5-N-2. IO controller root cgroup process behaviour
77   6. Namespace
78     6-1. Basics
79     6-2. The Root and Views
80     6-3. Migration and setns(2)
81     6-4. Interaction with Other Namespaces
82   P. Information on Kernel Programming
83     P-1. Filesystem Support for Writeback
84   D. Deprecated v1 Core Features
85   R. Issues with v1 and Rationales for v2
86     R-1. Multiple Hierarchies
87     R-2. Thread Granularity
88     R-3. Competition Between Inner Nodes and Threads
89     R-4. Other Interface Issues
90     R-5. Controller Issues and Remedies
91       R-5-1. Memory
92
93
94Introduction
95============
96
97Terminology
98-----------
99
100"cgroup" stands for "control group" and is never capitalized.  The
101singular form is used to designate the whole feature and also as a
102qualifier as in "cgroup controllers".  When explicitly referring to
103multiple individual control groups, the plural form "cgroups" is used.
104
105
106What is cgroup?
107---------------
108
109cgroup is a mechanism to organize processes hierarchically and
110distribute system resources along the hierarchy in a controlled and
111configurable manner.
112
113cgroup is largely composed of two parts - the core and controllers.
114cgroup core is primarily responsible for hierarchically organizing
115processes.  A cgroup controller is usually responsible for
116distributing a specific type of system resource along the hierarchy
117although there are utility controllers which serve purposes other than
118resource distribution.
119
120cgroups form a tree structure and every process in the system belongs
121to one and only one cgroup.  All threads of a process belong to the
122same cgroup.  On creation, all processes are put in the cgroup that
123the parent process belongs to at the time.  A process can be migrated
124to another cgroup.  Migration of a process doesn't affect already
125existing descendant processes.
126
127Following certain structural constraints, controllers may be enabled or
128disabled selectively on a cgroup.  All controller behaviors are
129hierarchical - if a controller is enabled on a cgroup, it affects all
130processes which belong to the cgroups consisting the inclusive
131sub-hierarchy of the cgroup.  When a controller is enabled on a nested
132cgroup, it always restricts the resource distribution further.  The
133restrictions set closer to the root in the hierarchy can not be
134overridden from further away.
135
136
137Basic Operations
138================
139
140Mounting
141--------
142
143Unlike v1, cgroup v2 has only single hierarchy.  The cgroup v2
144hierarchy can be mounted with the following mount command::
145
146  # mount -t cgroup2 none $MOUNT_POINT
147
148cgroup2 filesystem has the magic number 0x63677270 ("cgrp").  All
149controllers which support v2 and are not bound to a v1 hierarchy are
150automatically bound to the v2 hierarchy and show up at the root.
151Controllers which are not in active use in the v2 hierarchy can be
152bound to other hierarchies.  This allows mixing v2 hierarchy with the
153legacy v1 multiple hierarchies in a fully backward compatible way.
154
155A controller can be moved across hierarchies only after the controller
156is no longer referenced in its current hierarchy.  Because per-cgroup
157controller states are destroyed asynchronously and controllers may
158have lingering references, a controller may not show up immediately on
159the v2 hierarchy after the final umount of the previous hierarchy.
160Similarly, a controller should be fully disabled to be moved out of
161the unified hierarchy and it may take some time for the disabled
162controller to become available for other hierarchies; furthermore, due
163to inter-controller dependencies, other controllers may need to be
164disabled too.
165
166While useful for development and manual configurations, moving
167controllers dynamically between the v2 and other hierarchies is
168strongly discouraged for production use.  It is recommended to decide
169the hierarchies and controller associations before starting using the
170controllers after system boot.
171
172During transition to v2, system management software might still
173automount the v1 cgroup filesystem and so hijack all controllers
174during boot, before manual intervention is possible. To make testing
175and experimenting easier, the kernel parameter cgroup_no_v1= allows
176disabling controllers in v1 and make them always available in v2.
177
178cgroup v2 currently supports the following mount options.
179
180  nsdelegate
181	Consider cgroup namespaces as delegation boundaries.  This
182	option is system wide and can only be set on mount or modified
183	through remount from the init namespace.  The mount option is
184	ignored on non-init namespace mounts.  Please refer to the
185	Delegation section for details.
186
187  favordynmods
188        Reduce the latencies of dynamic cgroup modifications such as
189        task migrations and controller on/offs at the cost of making
190        hot path operations such as forks and exits more expensive.
191        The static usage pattern of creating a cgroup, enabling
192        controllers, and then seeding it with CLONE_INTO_CGROUP is
193        not affected by this option.
194
195  memory_localevents
196        Only populate memory.events with data for the current cgroup,
197        and not any subtrees. This is legacy behaviour, the default
198        behaviour without this option is to include subtree counts.
199        This option is system wide and can only be set on mount or
200        modified through remount from the init namespace. The mount
201        option is ignored on non-init namespace mounts.
202
203  memory_recursiveprot
204        Recursively apply memory.min and memory.low protection to
205        entire subtrees, without requiring explicit downward
206        propagation into leaf cgroups.  This allows protecting entire
207        subtrees from one another, while retaining free competition
208        within those subtrees.  This should have been the default
209        behavior but is a mount-option to avoid regressing setups
210        relying on the original semantics (e.g. specifying bogusly
211        high 'bypass' protection values at higher tree levels).
212
213  memory_hugetlb_accounting
214        Count HugeTLB memory usage towards the cgroup's overall
215        memory usage for the memory controller (for the purpose of
216        statistics reporting and memory protetion). This is a new
217        behavior that could regress existing setups, so it must be
218        explicitly opted in with this mount option.
219
220        A few caveats to keep in mind:
221
222        * There is no HugeTLB pool management involved in the memory
223          controller. The pre-allocated pool does not belong to anyone.
224          Specifically, when a new HugeTLB folio is allocated to
225          the pool, it is not accounted for from the perspective of the
226          memory controller. It is only charged to a cgroup when it is
227          actually used (for e.g at page fault time). Host memory
228          overcommit management has to consider this when configuring
229          hard limits. In general, HugeTLB pool management should be
230          done via other mechanisms (such as the HugeTLB controller).
231        * Failure to charge a HugeTLB folio to the memory controller
232          results in SIGBUS. This could happen even if the HugeTLB pool
233          still has pages available (but the cgroup limit is hit and
234          reclaim attempt fails).
235        * Charging HugeTLB memory towards the memory controller affects
236          memory protection and reclaim dynamics. Any userspace tuning
237          (of low, min limits for e.g) needs to take this into account.
238        * HugeTLB pages utilized while this option is not selected
239          will not be tracked by the memory controller (even if cgroup
240          v2 is remounted later on).
241
242
243Organizing Processes and Threads
244--------------------------------
245
246Processes
247~~~~~~~~~
248
249Initially, only the root cgroup exists to which all processes belong.
250A child cgroup can be created by creating a sub-directory::
251
252  # mkdir $CGROUP_NAME
253
254A given cgroup may have multiple child cgroups forming a tree
255structure.  Each cgroup has a read-writable interface file
256"cgroup.procs".  When read, it lists the PIDs of all processes which
257belong to the cgroup one-per-line.  The PIDs are not ordered and the
258same PID may show up more than once if the process got moved to
259another cgroup and then back or the PID got recycled while reading.
260
261A process can be migrated into a cgroup by writing its PID to the
262target cgroup's "cgroup.procs" file.  Only one process can be migrated
263on a single write(2) call.  If a process is composed of multiple
264threads, writing the PID of any thread migrates all threads of the
265process.
266
267When a process forks a child process, the new process is born into the
268cgroup that the forking process belongs to at the time of the
269operation.  After exit, a process stays associated with the cgroup
270that it belonged to at the time of exit until it's reaped; however, a
271zombie process does not appear in "cgroup.procs" and thus can't be
272moved to another cgroup.
273
274A cgroup which doesn't have any children or live processes can be
275destroyed by removing the directory.  Note that a cgroup which doesn't
276have any children and is associated only with zombie processes is
277considered empty and can be removed::
278
279  # rmdir $CGROUP_NAME
280
281"/proc/$PID/cgroup" lists a process's cgroup membership.  If legacy
282cgroup is in use in the system, this file may contain multiple lines,
283one for each hierarchy.  The entry for cgroup v2 is always in the
284format "0::$PATH"::
285
286  # cat /proc/842/cgroup
287  ...
288  0::/test-cgroup/test-cgroup-nested
289
290If the process becomes a zombie and the cgroup it was associated with
291is removed subsequently, " (deleted)" is appended to the path::
292
293  # cat /proc/842/cgroup
294  ...
295  0::/test-cgroup/test-cgroup-nested (deleted)
296
297
298Threads
299~~~~~~~
300
301cgroup v2 supports thread granularity for a subset of controllers to
302support use cases requiring hierarchical resource distribution across
303the threads of a group of processes.  By default, all threads of a
304process belong to the same cgroup, which also serves as the resource
305domain to host resource consumptions which are not specific to a
306process or thread.  The thread mode allows threads to be spread across
307a subtree while still maintaining the common resource domain for them.
308
309Controllers which support thread mode are called threaded controllers.
310The ones which don't are called domain controllers.
311
312Marking a cgroup threaded makes it join the resource domain of its
313parent as a threaded cgroup.  The parent may be another threaded
314cgroup whose resource domain is further up in the hierarchy.  The root
315of a threaded subtree, that is, the nearest ancestor which is not
316threaded, is called threaded domain or thread root interchangeably and
317serves as the resource domain for the entire subtree.
318
319Inside a threaded subtree, threads of a process can be put in
320different cgroups and are not subject to the no internal process
321constraint - threaded controllers can be enabled on non-leaf cgroups
322whether they have threads in them or not.
323
324As the threaded domain cgroup hosts all the domain resource
325consumptions of the subtree, it is considered to have internal
326resource consumptions whether there are processes in it or not and
327can't have populated child cgroups which aren't threaded.  Because the
328root cgroup is not subject to no internal process constraint, it can
329serve both as a threaded domain and a parent to domain cgroups.
330
331The current operation mode or type of the cgroup is shown in the
332"cgroup.type" file which indicates whether the cgroup is a normal
333domain, a domain which is serving as the domain of a threaded subtree,
334or a threaded cgroup.
335
336On creation, a cgroup is always a domain cgroup and can be made
337threaded by writing "threaded" to the "cgroup.type" file.  The
338operation is single direction::
339
340  # echo threaded > cgroup.type
341
342Once threaded, the cgroup can't be made a domain again.  To enable the
343thread mode, the following conditions must be met.
344
345- As the cgroup will join the parent's resource domain.  The parent
346  must either be a valid (threaded) domain or a threaded cgroup.
347
348- When the parent is an unthreaded domain, it must not have any domain
349  controllers enabled or populated domain children.  The root is
350  exempt from this requirement.
351
352Topology-wise, a cgroup can be in an invalid state.  Please consider
353the following topology::
354
355  A (threaded domain) - B (threaded) - C (domain, just created)
356
357C is created as a domain but isn't connected to a parent which can
358host child domains.  C can't be used until it is turned into a
359threaded cgroup.  "cgroup.type" file will report "domain (invalid)" in
360these cases.  Operations which fail due to invalid topology use
361EOPNOTSUPP as the errno.
362
363A domain cgroup is turned into a threaded domain when one of its child
364cgroup becomes threaded or threaded controllers are enabled in the
365"cgroup.subtree_control" file while there are processes in the cgroup.
366A threaded domain reverts to a normal domain when the conditions
367clear.
368
369When read, "cgroup.threads" contains the list of the thread IDs of all
370threads in the cgroup.  Except that the operations are per-thread
371instead of per-process, "cgroup.threads" has the same format and
372behaves the same way as "cgroup.procs".  While "cgroup.threads" can be
373written to in any cgroup, as it can only move threads inside the same
374threaded domain, its operations are confined inside each threaded
375subtree.
376
377The threaded domain cgroup serves as the resource domain for the whole
378subtree, and, while the threads can be scattered across the subtree,
379all the processes are considered to be in the threaded domain cgroup.
380"cgroup.procs" in a threaded domain cgroup contains the PIDs of all
381processes in the subtree and is not readable in the subtree proper.
382However, "cgroup.procs" can be written to from anywhere in the subtree
383to migrate all threads of the matching process to the cgroup.
384
385Only threaded controllers can be enabled in a threaded subtree.  When
386a threaded controller is enabled inside a threaded subtree, it only
387accounts for and controls resource consumptions associated with the
388threads in the cgroup and its descendants.  All consumptions which
389aren't tied to a specific thread belong to the threaded domain cgroup.
390
391Because a threaded subtree is exempt from no internal process
392constraint, a threaded controller must be able to handle competition
393between threads in a non-leaf cgroup and its child cgroups.  Each
394threaded controller defines how such competitions are handled.
395
396Currently, the following controllers are threaded and can be enabled
397in a threaded cgroup::
398
399- cpu
400- cpuset
401- perf_event
402- pids
403
404[Un]populated Notification
405--------------------------
406
407Each non-root cgroup has a "cgroup.events" file which contains
408"populated" field indicating whether the cgroup's sub-hierarchy has
409live processes in it.  Its value is 0 if there is no live process in
410the cgroup and its descendants; otherwise, 1.  poll and [id]notify
411events are triggered when the value changes.  This can be used, for
412example, to start a clean-up operation after all processes of a given
413sub-hierarchy have exited.  The populated state updates and
414notifications are recursive.  Consider the following sub-hierarchy
415where the numbers in the parentheses represent the numbers of processes
416in each cgroup::
417
418  A(4) - B(0) - C(1)
419              \ D(0)
420
421A, B and C's "populated" fields would be 1 while D's 0.  After the one
422process in C exits, B and C's "populated" fields would flip to "0" and
423file modified events will be generated on the "cgroup.events" files of
424both cgroups.
425
426
427Controlling Controllers
428-----------------------
429
430Enabling and Disabling
431~~~~~~~~~~~~~~~~~~~~~~
432
433Each cgroup has a "cgroup.controllers" file which lists all
434controllers available for the cgroup to enable::
435
436  # cat cgroup.controllers
437  cpu io memory
438
439No controller is enabled by default.  Controllers can be enabled and
440disabled by writing to the "cgroup.subtree_control" file::
441
442  # echo "+cpu +memory -io" > cgroup.subtree_control
443
444Only controllers which are listed in "cgroup.controllers" can be
445enabled.  When multiple operations are specified as above, either they
446all succeed or fail.  If multiple operations on the same controller
447are specified, the last one is effective.
448
449Enabling a controller in a cgroup indicates that the distribution of
450the target resource across its immediate children will be controlled.
451Consider the following sub-hierarchy.  The enabled controllers are
452listed in parentheses::
453
454  A(cpu,memory) - B(memory) - C()
455                            \ D()
456
457As A has "cpu" and "memory" enabled, A will control the distribution
458of CPU cycles and memory to its children, in this case, B.  As B has
459"memory" enabled but not "CPU", C and D will compete freely on CPU
460cycles but their division of memory available to B will be controlled.
461
462As a controller regulates the distribution of the target resource to
463the cgroup's children, enabling it creates the controller's interface
464files in the child cgroups.  In the above example, enabling "cpu" on B
465would create the "cpu." prefixed controller interface files in C and
466D.  Likewise, disabling "memory" from B would remove the "memory."
467prefixed controller interface files from C and D.  This means that the
468controller interface files - anything which doesn't start with
469"cgroup." are owned by the parent rather than the cgroup itself.
470
471
472Top-down Constraint
473~~~~~~~~~~~~~~~~~~~
474
475Resources are distributed top-down and a cgroup can further distribute
476a resource only if the resource has been distributed to it from the
477parent.  This means that all non-root "cgroup.subtree_control" files
478can only contain controllers which are enabled in the parent's
479"cgroup.subtree_control" file.  A controller can be enabled only if
480the parent has the controller enabled and a controller can't be
481disabled if one or more children have it enabled.
482
483
484No Internal Process Constraint
485~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
486
487Non-root cgroups can distribute domain resources to their children
488only when they don't have any processes of their own.  In other words,
489only domain cgroups which don't contain any processes can have domain
490controllers enabled in their "cgroup.subtree_control" files.
491
492This guarantees that, when a domain controller is looking at the part
493of the hierarchy which has it enabled, processes are always only on
494the leaves.  This rules out situations where child cgroups compete
495against internal processes of the parent.
496
497The root cgroup is exempt from this restriction.  Root contains
498processes and anonymous resource consumption which can't be associated
499with any other cgroups and requires special treatment from most
500controllers.  How resource consumption in the root cgroup is governed
501is up to each controller (for more information on this topic please
502refer to the Non-normative information section in the Controllers
503chapter).
504
505Note that the restriction doesn't get in the way if there is no
506enabled controller in the cgroup's "cgroup.subtree_control".  This is
507important as otherwise it wouldn't be possible to create children of a
508populated cgroup.  To control resource distribution of a cgroup, the
509cgroup must create children and transfer all its processes to the
510children before enabling controllers in its "cgroup.subtree_control"
511file.
512
513
514Delegation
515----------
516
517Model of Delegation
518~~~~~~~~~~~~~~~~~~~
519
520A cgroup can be delegated in two ways.  First, to a less privileged
521user by granting write access of the directory and its "cgroup.procs",
522"cgroup.threads" and "cgroup.subtree_control" files to the user.
523Second, if the "nsdelegate" mount option is set, automatically to a
524cgroup namespace on namespace creation.
525
526Because the resource control interface files in a given directory
527control the distribution of the parent's resources, the delegatee
528shouldn't be allowed to write to them.  For the first method, this is
529achieved by not granting access to these files.  For the second, the
530kernel rejects writes to all files other than "cgroup.procs" and
531"cgroup.subtree_control" on a namespace root from inside the
532namespace.
533
534The end results are equivalent for both delegation types.  Once
535delegated, the user can build sub-hierarchy under the directory,
536organize processes inside it as it sees fit and further distribute the
537resources it received from the parent.  The limits and other settings
538of all resource controllers are hierarchical and regardless of what
539happens in the delegated sub-hierarchy, nothing can escape the
540resource restrictions imposed by the parent.
541
542Currently, cgroup doesn't impose any restrictions on the number of
543cgroups in or nesting depth of a delegated sub-hierarchy; however,
544this may be limited explicitly in the future.
545
546
547Delegation Containment
548~~~~~~~~~~~~~~~~~~~~~~
549
550A delegated sub-hierarchy is contained in the sense that processes
551can't be moved into or out of the sub-hierarchy by the delegatee.
552
553For delegations to a less privileged user, this is achieved by
554requiring the following conditions for a process with a non-root euid
555to migrate a target process into a cgroup by writing its PID to the
556"cgroup.procs" file.
557
558- The writer must have write access to the "cgroup.procs" file.
559
560- The writer must have write access to the "cgroup.procs" file of the
561  common ancestor of the source and destination cgroups.
562
563The above two constraints ensure that while a delegatee may migrate
564processes around freely in the delegated sub-hierarchy it can't pull
565in from or push out to outside the sub-hierarchy.
566
567For an example, let's assume cgroups C0 and C1 have been delegated to
568user U0 who created C00, C01 under C0 and C10 under C1 as follows and
569all processes under C0 and C1 belong to U0::
570
571  ~~~~~~~~~~~~~ - C0 - C00
572  ~ cgroup    ~      \ C01
573  ~ hierarchy ~
574  ~~~~~~~~~~~~~ - C1 - C10
575
576Let's also say U0 wants to write the PID of a process which is
577currently in C10 into "C00/cgroup.procs".  U0 has write access to the
578file; however, the common ancestor of the source cgroup C10 and the
579destination cgroup C00 is above the points of delegation and U0 would
580not have write access to its "cgroup.procs" files and thus the write
581will be denied with -EACCES.
582
583For delegations to namespaces, containment is achieved by requiring
584that both the source and destination cgroups are reachable from the
585namespace of the process which is attempting the migration.  If either
586is not reachable, the migration is rejected with -ENOENT.
587
588
589Guidelines
590----------
591
592Organize Once and Control
593~~~~~~~~~~~~~~~~~~~~~~~~~
594
595Migrating a process across cgroups is a relatively expensive operation
596and stateful resources such as memory are not moved together with the
597process.  This is an explicit design decision as there often exist
598inherent trade-offs between migration and various hot paths in terms
599of synchronization cost.
600
601As such, migrating processes across cgroups frequently as a means to
602apply different resource restrictions is discouraged.  A workload
603should be assigned to a cgroup according to the system's logical and
604resource structure once on start-up.  Dynamic adjustments to resource
605distribution can be made by changing controller configuration through
606the interface files.
607
608
609Avoid Name Collisions
610~~~~~~~~~~~~~~~~~~~~~
611
612Interface files for a cgroup and its children cgroups occupy the same
613directory and it is possible to create children cgroups which collide
614with interface files.
615
616All cgroup core interface files are prefixed with "cgroup." and each
617controller's interface files are prefixed with the controller name and
618a dot.  A controller's name is composed of lower case alphabets and
619'_'s but never begins with an '_' so it can be used as the prefix
620character for collision avoidance.  Also, interface file names won't
621start or end with terms which are often used in categorizing workloads
622such as job, service, slice, unit or workload.
623
624cgroup doesn't do anything to prevent name collisions and it's the
625user's responsibility to avoid them.
626
627
628Resource Distribution Models
629============================
630
631cgroup controllers implement several resource distribution schemes
632depending on the resource type and expected use cases.  This section
633describes major schemes in use along with their expected behaviors.
634
635
636Weights
637-------
638
639A parent's resource is distributed by adding up the weights of all
640active children and giving each the fraction matching the ratio of its
641weight against the sum.  As only children which can make use of the
642resource at the moment participate in the distribution, this is
643work-conserving.  Due to the dynamic nature, this model is usually
644used for stateless resources.
645
646All weights are in the range [1, 10000] with the default at 100.  This
647allows symmetric multiplicative biases in both directions at fine
648enough granularity while staying in the intuitive range.
649
650As long as the weight is in range, all configuration combinations are
651valid and there is no reason to reject configuration changes or
652process migrations.
653
654"cpu.weight" proportionally distributes CPU cycles to active children
655and is an example of this type.
656
657
658.. _cgroupv2-limits-distributor:
659
660Limits
661------
662
663A child can only consume up to the configured amount of the resource.
664Limits can be over-committed - the sum of the limits of children can
665exceed the amount of resource available to the parent.
666
667Limits are in the range [0, max] and defaults to "max", which is noop.
668
669As limits can be over-committed, all configuration combinations are
670valid and there is no reason to reject configuration changes or
671process migrations.
672
673"io.max" limits the maximum BPS and/or IOPS that a cgroup can consume
674on an IO device and is an example of this type.
675
676.. _cgroupv2-protections-distributor:
677
678Protections
679-----------
680
681A cgroup is protected up to the configured amount of the resource
682as long as the usages of all its ancestors are under their
683protected levels.  Protections can be hard guarantees or best effort
684soft boundaries.  Protections can also be over-committed in which case
685only up to the amount available to the parent is protected among
686children.
687
688Protections are in the range [0, max] and defaults to 0, which is
689noop.
690
691As protections can be over-committed, all configuration combinations
692are valid and there is no reason to reject configuration changes or
693process migrations.
694
695"memory.low" implements best-effort memory protection and is an
696example of this type.
697
698
699Allocations
700-----------
701
702A cgroup is exclusively allocated a certain amount of a finite
703resource.  Allocations can't be over-committed - the sum of the
704allocations of children can not exceed the amount of resource
705available to the parent.
706
707Allocations are in the range [0, max] and defaults to 0, which is no
708resource.
709
710As allocations can't be over-committed, some configuration
711combinations are invalid and should be rejected.  Also, if the
712resource is mandatory for execution of processes, process migrations
713may be rejected.
714
715"cpu.rt.max" hard-allocates realtime slices and is an example of this
716type.
717
718
719Interface Files
720===============
721
722Format
723------
724
725All interface files should be in one of the following formats whenever
726possible::
727
728  New-line separated values
729  (when only one value can be written at once)
730
731	VAL0\n
732	VAL1\n
733	...
734
735  Space separated values
736  (when read-only or multiple values can be written at once)
737
738	VAL0 VAL1 ...\n
739
740  Flat keyed
741
742	KEY0 VAL0\n
743	KEY1 VAL1\n
744	...
745
746  Nested keyed
747
748	KEY0 SUB_KEY0=VAL00 SUB_KEY1=VAL01...
749	KEY1 SUB_KEY0=VAL10 SUB_KEY1=VAL11...
750	...
751
752For a writable file, the format for writing should generally match
753reading; however, controllers may allow omitting later fields or
754implement restricted shortcuts for most common use cases.
755
756For both flat and nested keyed files, only the values for a single key
757can be written at a time.  For nested keyed files, the sub key pairs
758may be specified in any order and not all pairs have to be specified.
759
760
761Conventions
762-----------
763
764- Settings for a single feature should be contained in a single file.
765
766- The root cgroup should be exempt from resource control and thus
767  shouldn't have resource control interface files.
768
769- The default time unit is microseconds.  If a different unit is ever
770  used, an explicit unit suffix must be present.
771
772- A parts-per quantity should use a percentage decimal with at least
773  two digit fractional part - e.g. 13.40.
774
775- If a controller implements weight based resource distribution, its
776  interface file should be named "weight" and have the range [1,
777  10000] with 100 as the default.  The values are chosen to allow
778  enough and symmetric bias in both directions while keeping it
779  intuitive (the default is 100%).
780
781- If a controller implements an absolute resource guarantee and/or
782  limit, the interface files should be named "min" and "max"
783  respectively.  If a controller implements best effort resource
784  guarantee and/or limit, the interface files should be named "low"
785  and "high" respectively.
786
787  In the above four control files, the special token "max" should be
788  used to represent upward infinity for both reading and writing.
789
790- If a setting has a configurable default value and keyed specific
791  overrides, the default entry should be keyed with "default" and
792  appear as the first entry in the file.
793
794  The default value can be updated by writing either "default $VAL" or
795  "$VAL".
796
797  When writing to update a specific override, "default" can be used as
798  the value to indicate removal of the override.  Override entries
799  with "default" as the value must not appear when read.
800
801  For example, a setting which is keyed by major:minor device numbers
802  with integer values may look like the following::
803
804    # cat cgroup-example-interface-file
805    default 150
806    8:0 300
807
808  The default value can be updated by::
809
810    # echo 125 > cgroup-example-interface-file
811
812  or::
813
814    # echo "default 125" > cgroup-example-interface-file
815
816  An override can be set by::
817
818    # echo "8:16 170" > cgroup-example-interface-file
819
820  and cleared by::
821
822    # echo "8:0 default" > cgroup-example-interface-file
823    # cat cgroup-example-interface-file
824    default 125
825    8:16 170
826
827- For events which are not very high frequency, an interface file
828  "events" should be created which lists event key value pairs.
829  Whenever a notifiable event happens, file modified event should be
830  generated on the file.
831
832
833Core Interface Files
834--------------------
835
836All cgroup core files are prefixed with "cgroup."
837
838  cgroup.type
839	A read-write single value file which exists on non-root
840	cgroups.
841
842	When read, it indicates the current type of the cgroup, which
843	can be one of the following values.
844
845	- "domain" : A normal valid domain cgroup.
846
847	- "domain threaded" : A threaded domain cgroup which is
848          serving as the root of a threaded subtree.
849
850	- "domain invalid" : A cgroup which is in an invalid state.
851	  It can't be populated or have controllers enabled.  It may
852	  be allowed to become a threaded cgroup.
853
854	- "threaded" : A threaded cgroup which is a member of a
855          threaded subtree.
856
857	A cgroup can be turned into a threaded cgroup by writing
858	"threaded" to this file.
859
860  cgroup.procs
861	A read-write new-line separated values file which exists on
862	all cgroups.
863
864	When read, it lists the PIDs of all processes which belong to
865	the cgroup one-per-line.  The PIDs are not ordered and the
866	same PID may show up more than once if the process got moved
867	to another cgroup and then back or the PID got recycled while
868	reading.
869
870	A PID can be written to migrate the process associated with
871	the PID to the cgroup.  The writer should match all of the
872	following conditions.
873
874	- It must have write access to the "cgroup.procs" file.
875
876	- It must have write access to the "cgroup.procs" file of the
877	  common ancestor of the source and destination cgroups.
878
879	When delegating a sub-hierarchy, write access to this file
880	should be granted along with the containing directory.
881
882	In a threaded cgroup, reading this file fails with EOPNOTSUPP
883	as all the processes belong to the thread root.  Writing is
884	supported and moves every thread of the process to the cgroup.
885
886  cgroup.threads
887	A read-write new-line separated values file which exists on
888	all cgroups.
889
890	When read, it lists the TIDs of all threads which belong to
891	the cgroup one-per-line.  The TIDs are not ordered and the
892	same TID may show up more than once if the thread got moved to
893	another cgroup and then back or the TID got recycled while
894	reading.
895
896	A TID can be written to migrate the thread associated with the
897	TID to the cgroup.  The writer should match all of the
898	following conditions.
899
900	- It must have write access to the "cgroup.threads" file.
901
902	- The cgroup that the thread is currently in must be in the
903          same resource domain as the destination cgroup.
904
905	- It must have write access to the "cgroup.procs" file of the
906	  common ancestor of the source and destination cgroups.
907
908	When delegating a sub-hierarchy, write access to this file
909	should be granted along with the containing directory.
910
911  cgroup.controllers
912	A read-only space separated values file which exists on all
913	cgroups.
914
915	It shows space separated list of all controllers available to
916	the cgroup.  The controllers are not ordered.
917
918  cgroup.subtree_control
919	A read-write space separated values file which exists on all
920	cgroups.  Starts out empty.
921
922	When read, it shows space separated list of the controllers
923	which are enabled to control resource distribution from the
924	cgroup to its children.
925
926	Space separated list of controllers prefixed with '+' or '-'
927	can be written to enable or disable controllers.  A controller
928	name prefixed with '+' enables the controller and '-'
929	disables.  If a controller appears more than once on the list,
930	the last one is effective.  When multiple enable and disable
931	operations are specified, either all succeed or all fail.
932
933  cgroup.events
934	A read-only flat-keyed file which exists on non-root cgroups.
935	The following entries are defined.  Unless specified
936	otherwise, a value change in this file generates a file
937	modified event.
938
939	  populated
940		1 if the cgroup or its descendants contains any live
941		processes; otherwise, 0.
942	  frozen
943		1 if the cgroup is frozen; otherwise, 0.
944
945  cgroup.max.descendants
946	A read-write single value files.  The default is "max".
947
948	Maximum allowed number of descent cgroups.
949	If the actual number of descendants is equal or larger,
950	an attempt to create a new cgroup in the hierarchy will fail.
951
952  cgroup.max.depth
953	A read-write single value files.  The default is "max".
954
955	Maximum allowed descent depth below the current cgroup.
956	If the actual descent depth is equal or larger,
957	an attempt to create a new child cgroup will fail.
958
959  cgroup.stat
960	A read-only flat-keyed file with the following entries:
961
962	  nr_descendants
963		Total number of visible descendant cgroups.
964
965	  nr_dying_descendants
966		Total number of dying descendant cgroups. A cgroup becomes
967		dying after being deleted by a user. The cgroup will remain
968		in dying state for some time undefined time (which can depend
969		on system load) before being completely destroyed.
970
971		A process can't enter a dying cgroup under any circumstances,
972		a dying cgroup can't revive.
973
974		A dying cgroup can consume system resources not exceeding
975		limits, which were active at the moment of cgroup deletion.
976
977  cgroup.freeze
978	A read-write single value file which exists on non-root cgroups.
979	Allowed values are "0" and "1". The default is "0".
980
981	Writing "1" to the file causes freezing of the cgroup and all
982	descendant cgroups. This means that all belonging processes will
983	be stopped and will not run until the cgroup will be explicitly
984	unfrozen. Freezing of the cgroup may take some time; when this action
985	is completed, the "frozen" value in the cgroup.events control file
986	will be updated to "1" and the corresponding notification will be
987	issued.
988
989	A cgroup can be frozen either by its own settings, or by settings
990	of any ancestor cgroups. If any of ancestor cgroups is frozen, the
991	cgroup will remain frozen.
992
993	Processes in the frozen cgroup can be killed by a fatal signal.
994	They also can enter and leave a frozen cgroup: either by an explicit
995	move by a user, or if freezing of the cgroup races with fork().
996	If a process is moved to a frozen cgroup, it stops. If a process is
997	moved out of a frozen cgroup, it becomes running.
998
999	Frozen status of a cgroup doesn't affect any cgroup tree operations:
1000	it's possible to delete a frozen (and empty) cgroup, as well as
1001	create new sub-cgroups.
1002
1003  cgroup.kill
1004	A write-only single value file which exists in non-root cgroups.
1005	The only allowed value is "1".
1006
1007	Writing "1" to the file causes the cgroup and all descendant cgroups to
1008	be killed. This means that all processes located in the affected cgroup
1009	tree will be killed via SIGKILL.
1010
1011	Killing a cgroup tree will deal with concurrent forks appropriately and
1012	is protected against migrations.
1013
1014	In a threaded cgroup, writing this file fails with EOPNOTSUPP as
1015	killing cgroups is a process directed operation, i.e. it affects
1016	the whole thread-group.
1017
1018  cgroup.pressure
1019	A read-write single value file that allowed values are "0" and "1".
1020	The default is "1".
1021
1022	Writing "0" to the file will disable the cgroup PSI accounting.
1023	Writing "1" to the file will re-enable the cgroup PSI accounting.
1024
1025	This control attribute is not hierarchical, so disable or enable PSI
1026	accounting in a cgroup does not affect PSI accounting in descendants
1027	and doesn't need pass enablement via ancestors from root.
1028
1029	The reason this control attribute exists is that PSI accounts stalls for
1030	each cgroup separately and aggregates it at each level of the hierarchy.
1031	This may cause non-negligible overhead for some workloads when under
1032	deep level of the hierarchy, in which case this control attribute can
1033	be used to disable PSI accounting in the non-leaf cgroups.
1034
1035  irq.pressure
1036	A read-write nested-keyed file.
1037
1038	Shows pressure stall information for IRQ/SOFTIRQ. See
1039	:ref:`Documentation/accounting/psi.rst <psi>` for details.
1040
1041Controllers
1042===========
1043
1044.. _cgroup-v2-cpu:
1045
1046CPU
1047---
1048
1049The "cpu" controllers regulates distribution of CPU cycles.  This
1050controller implements weight and absolute bandwidth limit models for
1051normal scheduling policy and absolute bandwidth allocation model for
1052realtime scheduling policy.
1053
1054In all the above models, cycles distribution is defined only on a temporal
1055base and it does not account for the frequency at which tasks are executed.
1056The (optional) utilization clamping support allows to hint the schedutil
1057cpufreq governor about the minimum desired frequency which should always be
1058provided by a CPU, as well as the maximum desired frequency, which should not
1059be exceeded by a CPU.
1060
1061WARNING: cgroup2 doesn't yet support control of realtime processes and
1062the cpu controller can only be enabled when all RT processes are in
1063the root cgroup.  Be aware that system management software may already
1064have placed RT processes into nonroot cgroups during the system boot
1065process, and these processes may need to be moved to the root cgroup
1066before the cpu controller can be enabled.
1067
1068
1069CPU Interface Files
1070~~~~~~~~~~~~~~~~~~~
1071
1072All time durations are in microseconds.
1073
1074  cpu.stat
1075	A read-only flat-keyed file.
1076	This file exists whether the controller is enabled or not.
1077
1078	It always reports the following three stats:
1079
1080	- usage_usec
1081	- user_usec
1082	- system_usec
1083
1084	and the following five when the controller is enabled:
1085
1086	- nr_periods
1087	- nr_throttled
1088	- throttled_usec
1089	- nr_bursts
1090	- burst_usec
1091
1092  cpu.weight
1093	A read-write single value file which exists on non-root
1094	cgroups.  The default is "100".
1095
1096	The weight in the range [1, 10000].
1097
1098  cpu.weight.nice
1099	A read-write single value file which exists on non-root
1100	cgroups.  The default is "0".
1101
1102	The nice value is in the range [-20, 19].
1103
1104	This interface file is an alternative interface for
1105	"cpu.weight" and allows reading and setting weight using the
1106	same values used by nice(2).  Because the range is smaller and
1107	granularity is coarser for the nice values, the read value is
1108	the closest approximation of the current weight.
1109
1110  cpu.max
1111	A read-write two value file which exists on non-root cgroups.
1112	The default is "max 100000".
1113
1114	The maximum bandwidth limit.  It's in the following format::
1115
1116	  $MAX $PERIOD
1117
1118	which indicates that the group may consume up to $MAX in each
1119	$PERIOD duration.  "max" for $MAX indicates no limit.  If only
1120	one number is written, $MAX is updated.
1121
1122  cpu.max.burst
1123	A read-write single value file which exists on non-root
1124	cgroups.  The default is "0".
1125
1126	The burst in the range [0, $MAX].
1127
1128  cpu.pressure
1129	A read-write nested-keyed file.
1130
1131	Shows pressure stall information for CPU. See
1132	:ref:`Documentation/accounting/psi.rst <psi>` for details.
1133
1134  cpu.uclamp.min
1135        A read-write single value file which exists on non-root cgroups.
1136        The default is "0", i.e. no utilization boosting.
1137
1138        The requested minimum utilization (protection) as a percentage
1139        rational number, e.g. 12.34 for 12.34%.
1140
1141        This interface allows reading and setting minimum utilization clamp
1142        values similar to the sched_setattr(2). This minimum utilization
1143        value is used to clamp the task specific minimum utilization clamp.
1144
1145        The requested minimum utilization (protection) is always capped by
1146        the current value for the maximum utilization (limit), i.e.
1147        `cpu.uclamp.max`.
1148
1149  cpu.uclamp.max
1150        A read-write single value file which exists on non-root cgroups.
1151        The default is "max". i.e. no utilization capping
1152
1153        The requested maximum utilization (limit) as a percentage rational
1154        number, e.g. 98.76 for 98.76%.
1155
1156        This interface allows reading and setting maximum utilization clamp
1157        values similar to the sched_setattr(2). This maximum utilization
1158        value is used to clamp the task specific maximum utilization clamp.
1159
1160
1161
1162Memory
1163------
1164
1165The "memory" controller regulates distribution of memory.  Memory is
1166stateful and implements both limit and protection models.  Due to the
1167intertwining between memory usage and reclaim pressure and the
1168stateful nature of memory, the distribution model is relatively
1169complex.
1170
1171While not completely water-tight, all major memory usages by a given
1172cgroup are tracked so that the total memory consumption can be
1173accounted and controlled to a reasonable extent.  Currently, the
1174following types of memory usages are tracked.
1175
1176- Userland memory - page cache and anonymous memory.
1177
1178- Kernel data structures such as dentries and inodes.
1179
1180- TCP socket buffers.
1181
1182The above list may expand in the future for better coverage.
1183
1184
1185Memory Interface Files
1186~~~~~~~~~~~~~~~~~~~~~~
1187
1188All memory amounts are in bytes.  If a value which is not aligned to
1189PAGE_SIZE is written, the value may be rounded up to the closest
1190PAGE_SIZE multiple when read back.
1191
1192  memory.current
1193	A read-only single value file which exists on non-root
1194	cgroups.
1195
1196	The total amount of memory currently being used by the cgroup
1197	and its descendants.
1198
1199  memory.min
1200	A read-write single value file which exists on non-root
1201	cgroups.  The default is "0".
1202
1203	Hard memory protection.  If the memory usage of a cgroup
1204	is within its effective min boundary, the cgroup's memory
1205	won't be reclaimed under any conditions. If there is no
1206	unprotected reclaimable memory available, OOM killer
1207	is invoked. Above the effective min boundary (or
1208	effective low boundary if it is higher), pages are reclaimed
1209	proportionally to the overage, reducing reclaim pressure for
1210	smaller overages.
1211
1212	Effective min boundary is limited by memory.min values of
1213	all ancestor cgroups. If there is memory.min overcommitment
1214	(child cgroup or cgroups are requiring more protected memory
1215	than parent will allow), then each child cgroup will get
1216	the part of parent's protection proportional to its
1217	actual memory usage below memory.min.
1218
1219	Putting more memory than generally available under this
1220	protection is discouraged and may lead to constant OOMs.
1221
1222	If a memory cgroup is not populated with processes,
1223	its memory.min is ignored.
1224
1225  memory.low
1226	A read-write single value file which exists on non-root
1227	cgroups.  The default is "0".
1228
1229	Best-effort memory protection.  If the memory usage of a
1230	cgroup is within its effective low boundary, the cgroup's
1231	memory won't be reclaimed unless there is no reclaimable
1232	memory available in unprotected cgroups.
1233	Above the effective low	boundary (or
1234	effective min boundary if it is higher), pages are reclaimed
1235	proportionally to the overage, reducing reclaim pressure for
1236	smaller overages.
1237
1238	Effective low boundary is limited by memory.low values of
1239	all ancestor cgroups. If there is memory.low overcommitment
1240	(child cgroup or cgroups are requiring more protected memory
1241	than parent will allow), then each child cgroup will get
1242	the part of parent's protection proportional to its
1243	actual memory usage below memory.low.
1244
1245	Putting more memory than generally available under this
1246	protection is discouraged.
1247
1248  memory.high
1249	A read-write single value file which exists on non-root
1250	cgroups.  The default is "max".
1251
1252	Memory usage throttle limit.  If a cgroup's usage goes
1253	over the high boundary, the processes of the cgroup are
1254	throttled and put under heavy reclaim pressure.
1255
1256	Going over the high limit never invokes the OOM killer and
1257	under extreme conditions the limit may be breached. The high
1258	limit should be used in scenarios where an external process
1259	monitors the limited cgroup to alleviate heavy reclaim
1260	pressure.
1261
1262  memory.max
1263	A read-write single value file which exists on non-root
1264	cgroups.  The default is "max".
1265
1266	Memory usage hard limit.  This is the main mechanism to limit
1267	memory usage of a cgroup.  If a cgroup's memory usage reaches
1268	this limit and can't be reduced, the OOM killer is invoked in
1269	the cgroup. Under certain circumstances, the usage may go
1270	over the limit temporarily.
1271
1272	In default configuration regular 0-order allocations always
1273	succeed unless OOM killer chooses current task as a victim.
1274
1275	Some kinds of allocations don't invoke the OOM killer.
1276	Caller could retry them differently, return into userspace
1277	as -ENOMEM or silently ignore in cases like disk readahead.
1278
1279  memory.reclaim
1280	A write-only nested-keyed file which exists for all cgroups.
1281
1282	This is a simple interface to trigger memory reclaim in the
1283	target cgroup.
1284
1285	This file accepts a single key, the number of bytes to reclaim.
1286	No nested keys are currently supported.
1287
1288	Example::
1289
1290	  echo "1G" > memory.reclaim
1291
1292	The interface can be later extended with nested keys to
1293	configure the reclaim behavior. For example, specify the
1294	type of memory to reclaim from (anon, file, ..).
1295
1296	Please note that the kernel can over or under reclaim from
1297	the target cgroup. If less bytes are reclaimed than the
1298	specified amount, -EAGAIN is returned.
1299
1300	Please note that the proactive reclaim (triggered by this
1301	interface) is not meant to indicate memory pressure on the
1302	memory cgroup. Therefore socket memory balancing triggered by
1303	the memory reclaim normally is not exercised in this case.
1304	This means that the networking layer will not adapt based on
1305	reclaim induced by memory.reclaim.
1306
1307  memory.peak
1308	A read-only single value file which exists on non-root
1309	cgroups.
1310
1311	The max memory usage recorded for the cgroup and its
1312	descendants since the creation of the cgroup.
1313
1314  memory.oom.group
1315	A read-write single value file which exists on non-root
1316	cgroups.  The default value is "0".
1317
1318	Determines whether the cgroup should be treated as
1319	an indivisible workload by the OOM killer. If set,
1320	all tasks belonging to the cgroup or to its descendants
1321	(if the memory cgroup is not a leaf cgroup) are killed
1322	together or not at all. This can be used to avoid
1323	partial kills to guarantee workload integrity.
1324
1325	Tasks with the OOM protection (oom_score_adj set to -1000)
1326	are treated as an exception and are never killed.
1327
1328	If the OOM killer is invoked in a cgroup, it's not going
1329	to kill any tasks outside of this cgroup, regardless
1330	memory.oom.group values of ancestor cgroups.
1331
1332  memory.events
1333	A read-only flat-keyed file which exists on non-root cgroups.
1334	The following entries are defined.  Unless specified
1335	otherwise, a value change in this file generates a file
1336	modified event.
1337
1338	Note that all fields in this file are hierarchical and the
1339	file modified event can be generated due to an event down the
1340	hierarchy. For the local events at the cgroup level see
1341	memory.events.local.
1342
1343	  low
1344		The number of times the cgroup is reclaimed due to
1345		high memory pressure even though its usage is under
1346		the low boundary.  This usually indicates that the low
1347		boundary is over-committed.
1348
1349	  high
1350		The number of times processes of the cgroup are
1351		throttled and routed to perform direct memory reclaim
1352		because the high memory boundary was exceeded.  For a
1353		cgroup whose memory usage is capped by the high limit
1354		rather than global memory pressure, this event's
1355		occurrences are expected.
1356
1357	  max
1358		The number of times the cgroup's memory usage was
1359		about to go over the max boundary.  If direct reclaim
1360		fails to bring it down, the cgroup goes to OOM state.
1361
1362	  oom
1363		The number of time the cgroup's memory usage was
1364		reached the limit and allocation was about to fail.
1365
1366		This event is not raised if the OOM killer is not
1367		considered as an option, e.g. for failed high-order
1368		allocations or if caller asked to not retry attempts.
1369
1370	  oom_kill
1371		The number of processes belonging to this cgroup
1372		killed by any kind of OOM killer.
1373
1374          oom_group_kill
1375                The number of times a group OOM has occurred.
1376
1377  memory.events.local
1378	Similar to memory.events but the fields in the file are local
1379	to the cgroup i.e. not hierarchical. The file modified event
1380	generated on this file reflects only the local events.
1381
1382  memory.stat
1383	A read-only flat-keyed file which exists on non-root cgroups.
1384
1385	This breaks down the cgroup's memory footprint into different
1386	types of memory, type-specific details, and other information
1387	on the state and past events of the memory management system.
1388
1389	All memory amounts are in bytes.
1390
1391	The entries are ordered to be human readable, and new entries
1392	can show up in the middle. Don't rely on items remaining in a
1393	fixed position; use the keys to look up specific values!
1394
1395	If the entry has no per-node counter (or not show in the
1396	memory.numa_stat). We use 'npn' (non-per-node) as the tag
1397	to indicate that it will not show in the memory.numa_stat.
1398
1399	  anon
1400		Amount of memory used in anonymous mappings such as
1401		brk(), sbrk(), and mmap(MAP_ANONYMOUS)
1402
1403	  file
1404		Amount of memory used to cache filesystem data,
1405		including tmpfs and shared memory.
1406
1407	  kernel (npn)
1408		Amount of total kernel memory, including
1409		(kernel_stack, pagetables, percpu, vmalloc, slab) in
1410		addition to other kernel memory use cases.
1411
1412	  kernel_stack
1413		Amount of memory allocated to kernel stacks.
1414
1415	  pagetables
1416                Amount of memory allocated for page tables.
1417
1418	  sec_pagetables
1419		Amount of memory allocated for secondary page tables,
1420		this currently includes KVM mmu allocations on x86
1421		and arm64.
1422
1423	  percpu (npn)
1424		Amount of memory used for storing per-cpu kernel
1425		data structures.
1426
1427	  sock (npn)
1428		Amount of memory used in network transmission buffers
1429
1430	  vmalloc (npn)
1431		Amount of memory used for vmap backed memory.
1432
1433	  shmem
1434		Amount of cached filesystem data that is swap-backed,
1435		such as tmpfs, shm segments, shared anonymous mmap()s
1436
1437	  zswap
1438		Amount of memory consumed by the zswap compression backend.
1439
1440	  zswapped
1441		Amount of application memory swapped out to zswap.
1442
1443	  file_mapped
1444		Amount of cached filesystem data mapped with mmap()
1445
1446	  file_dirty
1447		Amount of cached filesystem data that was modified but
1448		not yet written back to disk
1449
1450	  file_writeback
1451		Amount of cached filesystem data that was modified and
1452		is currently being written back to disk
1453
1454	  swapcached
1455		Amount of swap cached in memory. The swapcache is accounted
1456		against both memory and swap usage.
1457
1458	  anon_thp
1459		Amount of memory used in anonymous mappings backed by
1460		transparent hugepages
1461
1462	  file_thp
1463		Amount of cached filesystem data backed by transparent
1464		hugepages
1465
1466	  shmem_thp
1467		Amount of shm, tmpfs, shared anonymous mmap()s backed by
1468		transparent hugepages
1469
1470	  inactive_anon, active_anon, inactive_file, active_file, unevictable
1471		Amount of memory, swap-backed and filesystem-backed,
1472		on the internal memory management lists used by the
1473		page reclaim algorithm.
1474
1475		As these represent internal list state (eg. shmem pages are on anon
1476		memory management lists), inactive_foo + active_foo may not be equal to
1477		the value for the foo counter, since the foo counter is type-based, not
1478		list-based.
1479
1480	  slab_reclaimable
1481		Part of "slab" that might be reclaimed, such as
1482		dentries and inodes.
1483
1484	  slab_unreclaimable
1485		Part of "slab" that cannot be reclaimed on memory
1486		pressure.
1487
1488	  slab (npn)
1489		Amount of memory used for storing in-kernel data
1490		structures.
1491
1492	  workingset_refault_anon
1493		Number of refaults of previously evicted anonymous pages.
1494
1495	  workingset_refault_file
1496		Number of refaults of previously evicted file pages.
1497
1498	  workingset_activate_anon
1499		Number of refaulted anonymous pages that were immediately
1500		activated.
1501
1502	  workingset_activate_file
1503		Number of refaulted file pages that were immediately activated.
1504
1505	  workingset_restore_anon
1506		Number of restored anonymous pages which have been detected as
1507		an active workingset before they got reclaimed.
1508
1509	  workingset_restore_file
1510		Number of restored file pages which have been detected as an
1511		active workingset before they got reclaimed.
1512
1513	  workingset_nodereclaim
1514		Number of times a shadow node has been reclaimed
1515
1516	  pgscan (npn)
1517		Amount of scanned pages (in an inactive LRU list)
1518
1519	  pgsteal (npn)
1520		Amount of reclaimed pages
1521
1522	  pgscan_kswapd (npn)
1523		Amount of scanned pages by kswapd (in an inactive LRU list)
1524
1525	  pgscan_direct (npn)
1526		Amount of scanned pages directly  (in an inactive LRU list)
1527
1528	  pgscan_khugepaged (npn)
1529		Amount of scanned pages by khugepaged  (in an inactive LRU list)
1530
1531	  pgsteal_kswapd (npn)
1532		Amount of reclaimed pages by kswapd
1533
1534	  pgsteal_direct (npn)
1535		Amount of reclaimed pages directly
1536
1537	  pgsteal_khugepaged (npn)
1538		Amount of reclaimed pages by khugepaged
1539
1540	  pgfault (npn)
1541		Total number of page faults incurred
1542
1543	  pgmajfault (npn)
1544		Number of major page faults incurred
1545
1546	  pgrefill (npn)
1547		Amount of scanned pages (in an active LRU list)
1548
1549	  pgactivate (npn)
1550		Amount of pages moved to the active LRU list
1551
1552	  pgdeactivate (npn)
1553		Amount of pages moved to the inactive LRU list
1554
1555	  pglazyfree (npn)
1556		Amount of pages postponed to be freed under memory pressure
1557
1558	  pglazyfreed (npn)
1559		Amount of reclaimed lazyfree pages
1560
1561	  thp_fault_alloc (npn)
1562		Number of transparent hugepages which were allocated to satisfy
1563		a page fault. This counter is not present when CONFIG_TRANSPARENT_HUGEPAGE
1564                is not set.
1565
1566	  thp_collapse_alloc (npn)
1567		Number of transparent hugepages which were allocated to allow
1568		collapsing an existing range of pages. This counter is not
1569		present when CONFIG_TRANSPARENT_HUGEPAGE is not set.
1570
1571	  thp_swpout (npn)
1572		Number of transparent hugepages which are swapout in one piece
1573		without splitting.
1574
1575	  thp_swpout_fallback (npn)
1576		Number of transparent hugepages which were split before swapout.
1577		Usually because failed to allocate some continuous swap space
1578		for the huge page.
1579
1580  memory.numa_stat
1581	A read-only nested-keyed file which exists on non-root cgroups.
1582
1583	This breaks down the cgroup's memory footprint into different
1584	types of memory, type-specific details, and other information
1585	per node on the state of the memory management system.
1586
1587	This is useful for providing visibility into the NUMA locality
1588	information within an memcg since the pages are allowed to be
1589	allocated from any physical node. One of the use case is evaluating
1590	application performance by combining this information with the
1591	application's CPU allocation.
1592
1593	All memory amounts are in bytes.
1594
1595	The output format of memory.numa_stat is::
1596
1597	  type N0=<bytes in node 0> N1=<bytes in node 1> ...
1598
1599	The entries are ordered to be human readable, and new entries
1600	can show up in the middle. Don't rely on items remaining in a
1601	fixed position; use the keys to look up specific values!
1602
1603	The entries can refer to the memory.stat.
1604
1605  memory.swap.current
1606	A read-only single value file which exists on non-root
1607	cgroups.
1608
1609	The total amount of swap currently being used by the cgroup
1610	and its descendants.
1611
1612  memory.swap.high
1613	A read-write single value file which exists on non-root
1614	cgroups.  The default is "max".
1615
1616	Swap usage throttle limit.  If a cgroup's swap usage exceeds
1617	this limit, all its further allocations will be throttled to
1618	allow userspace to implement custom out-of-memory procedures.
1619
1620	This limit marks a point of no return for the cgroup. It is NOT
1621	designed to manage the amount of swapping a workload does
1622	during regular operation. Compare to memory.swap.max, which
1623	prohibits swapping past a set amount, but lets the cgroup
1624	continue unimpeded as long as other memory can be reclaimed.
1625
1626	Healthy workloads are not expected to reach this limit.
1627
1628  memory.swap.peak
1629	A read-only single value file which exists on non-root
1630	cgroups.
1631
1632	The max swap usage recorded for the cgroup and its
1633	descendants since the creation of the cgroup.
1634
1635  memory.swap.max
1636	A read-write single value file which exists on non-root
1637	cgroups.  The default is "max".
1638
1639	Swap usage hard limit.  If a cgroup's swap usage reaches this
1640	limit, anonymous memory of the cgroup will not be swapped out.
1641
1642  memory.swap.events
1643	A read-only flat-keyed file which exists on non-root cgroups.
1644	The following entries are defined.  Unless specified
1645	otherwise, a value change in this file generates a file
1646	modified event.
1647
1648	  high
1649		The number of times the cgroup's swap usage was over
1650		the high threshold.
1651
1652	  max
1653		The number of times the cgroup's swap usage was about
1654		to go over the max boundary and swap allocation
1655		failed.
1656
1657	  fail
1658		The number of times swap allocation failed either
1659		because of running out of swap system-wide or max
1660		limit.
1661
1662	When reduced under the current usage, the existing swap
1663	entries are reclaimed gradually and the swap usage may stay
1664	higher than the limit for an extended period of time.  This
1665	reduces the impact on the workload and memory management.
1666
1667  memory.zswap.current
1668	A read-only single value file which exists on non-root
1669	cgroups.
1670
1671	The total amount of memory consumed by the zswap compression
1672	backend.
1673
1674  memory.zswap.max
1675	A read-write single value file which exists on non-root
1676	cgroups.  The default is "max".
1677
1678	Zswap usage hard limit. If a cgroup's zswap pool reaches this
1679	limit, it will refuse to take any more stores before existing
1680	entries fault back in or are written out to disk.
1681
1682  memory.pressure
1683	A read-only nested-keyed file.
1684
1685	Shows pressure stall information for memory. See
1686	:ref:`Documentation/accounting/psi.rst <psi>` for details.
1687
1688
1689Usage Guidelines
1690~~~~~~~~~~~~~~~~
1691
1692"memory.high" is the main mechanism to control memory usage.
1693Over-committing on high limit (sum of high limits > available memory)
1694and letting global memory pressure to distribute memory according to
1695usage is a viable strategy.
1696
1697Because breach of the high limit doesn't trigger the OOM killer but
1698throttles the offending cgroup, a management agent has ample
1699opportunities to monitor and take appropriate actions such as granting
1700more memory or terminating the workload.
1701
1702Determining whether a cgroup has enough memory is not trivial as
1703memory usage doesn't indicate whether the workload can benefit from
1704more memory.  For example, a workload which writes data received from
1705network to a file can use all available memory but can also operate as
1706performant with a small amount of memory.  A measure of memory
1707pressure - how much the workload is being impacted due to lack of
1708memory - is necessary to determine whether a workload needs more
1709memory; unfortunately, memory pressure monitoring mechanism isn't
1710implemented yet.
1711
1712
1713Memory Ownership
1714~~~~~~~~~~~~~~~~
1715
1716A memory area is charged to the cgroup which instantiated it and stays
1717charged to the cgroup until the area is released.  Migrating a process
1718to a different cgroup doesn't move the memory usages that it
1719instantiated while in the previous cgroup to the new cgroup.
1720
1721A memory area may be used by processes belonging to different cgroups.
1722To which cgroup the area will be charged is in-deterministic; however,
1723over time, the memory area is likely to end up in a cgroup which has
1724enough memory allowance to avoid high reclaim pressure.
1725
1726If a cgroup sweeps a considerable amount of memory which is expected
1727to be accessed repeatedly by other cgroups, it may make sense to use
1728POSIX_FADV_DONTNEED to relinquish the ownership of memory areas
1729belonging to the affected files to ensure correct memory ownership.
1730
1731
1732IO
1733--
1734
1735The "io" controller regulates the distribution of IO resources.  This
1736controller implements both weight based and absolute bandwidth or IOPS
1737limit distribution; however, weight based distribution is available
1738only if cfq-iosched is in use and neither scheme is available for
1739blk-mq devices.
1740
1741
1742IO Interface Files
1743~~~~~~~~~~~~~~~~~~
1744
1745  io.stat
1746	A read-only nested-keyed file.
1747
1748	Lines are keyed by $MAJ:$MIN device numbers and not ordered.
1749	The following nested keys are defined.
1750
1751	  ======	=====================
1752	  rbytes	Bytes read
1753	  wbytes	Bytes written
1754	  rios		Number of read IOs
1755	  wios		Number of write IOs
1756	  dbytes	Bytes discarded
1757	  dios		Number of discard IOs
1758	  ======	=====================
1759
1760	An example read output follows::
1761
1762	  8:16 rbytes=1459200 wbytes=314773504 rios=192 wios=353 dbytes=0 dios=0
1763	  8:0 rbytes=90430464 wbytes=299008000 rios=8950 wios=1252 dbytes=50331648 dios=3021
1764
1765  io.cost.qos
1766	A read-write nested-keyed file which exists only on the root
1767	cgroup.
1768
1769	This file configures the Quality of Service of the IO cost
1770	model based controller (CONFIG_BLK_CGROUP_IOCOST) which
1771	currently implements "io.weight" proportional control.  Lines
1772	are keyed by $MAJ:$MIN device numbers and not ordered.  The
1773	line for a given device is populated on the first write for
1774	the device on "io.cost.qos" or "io.cost.model".  The following
1775	nested keys are defined.
1776
1777	  ======	=====================================
1778	  enable	Weight-based control enable
1779	  ctrl		"auto" or "user"
1780	  rpct		Read latency percentile    [0, 100]
1781	  rlat		Read latency threshold
1782	  wpct		Write latency percentile   [0, 100]
1783	  wlat		Write latency threshold
1784	  min		Minimum scaling percentage [1, 10000]
1785	  max		Maximum scaling percentage [1, 10000]
1786	  ======	=====================================
1787
1788	The controller is disabled by default and can be enabled by
1789	setting "enable" to 1.  "rpct" and "wpct" parameters default
1790	to zero and the controller uses internal device saturation
1791	state to adjust the overall IO rate between "min" and "max".
1792
1793	When a better control quality is needed, latency QoS
1794	parameters can be configured.  For example::
1795
1796	  8:16 enable=1 ctrl=auto rpct=95.00 rlat=75000 wpct=95.00 wlat=150000 min=50.00 max=150.0
1797
1798	shows that on sdb, the controller is enabled, will consider
1799	the device saturated if the 95th percentile of read completion
1800	latencies is above 75ms or write 150ms, and adjust the overall
1801	IO issue rate between 50% and 150% accordingly.
1802
1803	The lower the saturation point, the better the latency QoS at
1804	the cost of aggregate bandwidth.  The narrower the allowed
1805	adjustment range between "min" and "max", the more conformant
1806	to the cost model the IO behavior.  Note that the IO issue
1807	base rate may be far off from 100% and setting "min" and "max"
1808	blindly can lead to a significant loss of device capacity or
1809	control quality.  "min" and "max" are useful for regulating
1810	devices which show wide temporary behavior changes - e.g. a
1811	ssd which accepts writes at the line speed for a while and
1812	then completely stalls for multiple seconds.
1813
1814	When "ctrl" is "auto", the parameters are controlled by the
1815	kernel and may change automatically.  Setting "ctrl" to "user"
1816	or setting any of the percentile and latency parameters puts
1817	it into "user" mode and disables the automatic changes.  The
1818	automatic mode can be restored by setting "ctrl" to "auto".
1819
1820  io.cost.model
1821	A read-write nested-keyed file which exists only on the root
1822	cgroup.
1823
1824	This file configures the cost model of the IO cost model based
1825	controller (CONFIG_BLK_CGROUP_IOCOST) which currently
1826	implements "io.weight" proportional control.  Lines are keyed
1827	by $MAJ:$MIN device numbers and not ordered.  The line for a
1828	given device is populated on the first write for the device on
1829	"io.cost.qos" or "io.cost.model".  The following nested keys
1830	are defined.
1831
1832	  =====		================================
1833	  ctrl		"auto" or "user"
1834	  model		The cost model in use - "linear"
1835	  =====		================================
1836
1837	When "ctrl" is "auto", the kernel may change all parameters
1838	dynamically.  When "ctrl" is set to "user" or any other
1839	parameters are written to, "ctrl" become "user" and the
1840	automatic changes are disabled.
1841
1842	When "model" is "linear", the following model parameters are
1843	defined.
1844
1845	  =============	========================================
1846	  [r|w]bps	The maximum sequential IO throughput
1847	  [r|w]seqiops	The maximum 4k sequential IOs per second
1848	  [r|w]randiops	The maximum 4k random IOs per second
1849	  =============	========================================
1850
1851	From the above, the builtin linear model determines the base
1852	costs of a sequential and random IO and the cost coefficient
1853	for the IO size.  While simple, this model can cover most
1854	common device classes acceptably.
1855
1856	The IO cost model isn't expected to be accurate in absolute
1857	sense and is scaled to the device behavior dynamically.
1858
1859	If needed, tools/cgroup/iocost_coef_gen.py can be used to
1860	generate device-specific coefficients.
1861
1862  io.weight
1863	A read-write flat-keyed file which exists on non-root cgroups.
1864	The default is "default 100".
1865
1866	The first line is the default weight applied to devices
1867	without specific override.  The rest are overrides keyed by
1868	$MAJ:$MIN device numbers and not ordered.  The weights are in
1869	the range [1, 10000] and specifies the relative amount IO time
1870	the cgroup can use in relation to its siblings.
1871
1872	The default weight can be updated by writing either "default
1873	$WEIGHT" or simply "$WEIGHT".  Overrides can be set by writing
1874	"$MAJ:$MIN $WEIGHT" and unset by writing "$MAJ:$MIN default".
1875
1876	An example read output follows::
1877
1878	  default 100
1879	  8:16 200
1880	  8:0 50
1881
1882  io.max
1883	A read-write nested-keyed file which exists on non-root
1884	cgroups.
1885
1886	BPS and IOPS based IO limit.  Lines are keyed by $MAJ:$MIN
1887	device numbers and not ordered.  The following nested keys are
1888	defined.
1889
1890	  =====		==================================
1891	  rbps		Max read bytes per second
1892	  wbps		Max write bytes per second
1893	  riops		Max read IO operations per second
1894	  wiops		Max write IO operations per second
1895	  =====		==================================
1896
1897	When writing, any number of nested key-value pairs can be
1898	specified in any order.  "max" can be specified as the value
1899	to remove a specific limit.  If the same key is specified
1900	multiple times, the outcome is undefined.
1901
1902	BPS and IOPS are measured in each IO direction and IOs are
1903	delayed if limit is reached.  Temporary bursts are allowed.
1904
1905	Setting read limit at 2M BPS and write at 120 IOPS for 8:16::
1906
1907	  echo "8:16 rbps=2097152 wiops=120" > io.max
1908
1909	Reading returns the following::
1910
1911	  8:16 rbps=2097152 wbps=max riops=max wiops=120
1912
1913	Write IOPS limit can be removed by writing the following::
1914
1915	  echo "8:16 wiops=max" > io.max
1916
1917	Reading now returns the following::
1918
1919	  8:16 rbps=2097152 wbps=max riops=max wiops=max
1920
1921  io.pressure
1922	A read-only nested-keyed file.
1923
1924	Shows pressure stall information for IO. See
1925	:ref:`Documentation/accounting/psi.rst <psi>` for details.
1926
1927
1928Writeback
1929~~~~~~~~~
1930
1931Page cache is dirtied through buffered writes and shared mmaps and
1932written asynchronously to the backing filesystem by the writeback
1933mechanism.  Writeback sits between the memory and IO domains and
1934regulates the proportion of dirty memory by balancing dirtying and
1935write IOs.
1936
1937The io controller, in conjunction with the memory controller,
1938implements control of page cache writeback IOs.  The memory controller
1939defines the memory domain that dirty memory ratio is calculated and
1940maintained for and the io controller defines the io domain which
1941writes out dirty pages for the memory domain.  Both system-wide and
1942per-cgroup dirty memory states are examined and the more restrictive
1943of the two is enforced.
1944
1945cgroup writeback requires explicit support from the underlying
1946filesystem.  Currently, cgroup writeback is implemented on ext2, ext4,
1947btrfs, f2fs, and xfs.  On other filesystems, all writeback IOs are
1948attributed to the root cgroup.
1949
1950There are inherent differences in memory and writeback management
1951which affects how cgroup ownership is tracked.  Memory is tracked per
1952page while writeback per inode.  For the purpose of writeback, an
1953inode is assigned to a cgroup and all IO requests to write dirty pages
1954from the inode are attributed to that cgroup.
1955
1956As cgroup ownership for memory is tracked per page, there can be pages
1957which are associated with different cgroups than the one the inode is
1958associated with.  These are called foreign pages.  The writeback
1959constantly keeps track of foreign pages and, if a particular foreign
1960cgroup becomes the majority over a certain period of time, switches
1961the ownership of the inode to that cgroup.
1962
1963While this model is enough for most use cases where a given inode is
1964mostly dirtied by a single cgroup even when the main writing cgroup
1965changes over time, use cases where multiple cgroups write to a single
1966inode simultaneously are not supported well.  In such circumstances, a
1967significant portion of IOs are likely to be attributed incorrectly.
1968As memory controller assigns page ownership on the first use and
1969doesn't update it until the page is released, even if writeback
1970strictly follows page ownership, multiple cgroups dirtying overlapping
1971areas wouldn't work as expected.  It's recommended to avoid such usage
1972patterns.
1973
1974The sysctl knobs which affect writeback behavior are applied to cgroup
1975writeback as follows.
1976
1977  vm.dirty_background_ratio, vm.dirty_ratio
1978	These ratios apply the same to cgroup writeback with the
1979	amount of available memory capped by limits imposed by the
1980	memory controller and system-wide clean memory.
1981
1982  vm.dirty_background_bytes, vm.dirty_bytes
1983	For cgroup writeback, this is calculated into ratio against
1984	total available memory and applied the same way as
1985	vm.dirty[_background]_ratio.
1986
1987
1988IO Latency
1989~~~~~~~~~~
1990
1991This is a cgroup v2 controller for IO workload protection.  You provide a group
1992with a latency target, and if the average latency exceeds that target the
1993controller will throttle any peers that have a lower latency target than the
1994protected workload.
1995
1996The limits are only applied at the peer level in the hierarchy.  This means that
1997in the diagram below, only groups A, B, and C will influence each other, and
1998groups D and F will influence each other.  Group G will influence nobody::
1999
2000			[root]
2001		/	   |		\
2002		A	   B		C
2003	       /  \        |
2004	      D    F	   G
2005
2006
2007So the ideal way to configure this is to set io.latency in groups A, B, and C.
2008Generally you do not want to set a value lower than the latency your device
2009supports.  Experiment to find the value that works best for your workload.
2010Start at higher than the expected latency for your device and watch the
2011avg_lat value in io.stat for your workload group to get an idea of the
2012latency you see during normal operation.  Use the avg_lat value as a basis for
2013your real setting, setting at 10-15% higher than the value in io.stat.
2014
2015How IO Latency Throttling Works
2016~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2017
2018io.latency is work conserving; so as long as everybody is meeting their latency
2019target the controller doesn't do anything.  Once a group starts missing its
2020target it begins throttling any peer group that has a higher target than itself.
2021This throttling takes 2 forms:
2022
2023- Queue depth throttling.  This is the number of outstanding IO's a group is
2024  allowed to have.  We will clamp down relatively quickly, starting at no limit
2025  and going all the way down to 1 IO at a time.
2026
2027- Artificial delay induction.  There are certain types of IO that cannot be
2028  throttled without possibly adversely affecting higher priority groups.  This
2029  includes swapping and metadata IO.  These types of IO are allowed to occur
2030  normally, however they are "charged" to the originating group.  If the
2031  originating group is being throttled you will see the use_delay and delay
2032  fields in io.stat increase.  The delay value is how many microseconds that are
2033  being added to any process that runs in this group.  Because this number can
2034  grow quite large if there is a lot of swapping or metadata IO occurring we
2035  limit the individual delay events to 1 second at a time.
2036
2037Once the victimized group starts meeting its latency target again it will start
2038unthrottling any peer groups that were throttled previously.  If the victimized
2039group simply stops doing IO the global counter will unthrottle appropriately.
2040
2041IO Latency Interface Files
2042~~~~~~~~~~~~~~~~~~~~~~~~~~
2043
2044  io.latency
2045	This takes a similar format as the other controllers.
2046
2047		"MAJOR:MINOR target=<target time in microseconds>"
2048
2049  io.stat
2050	If the controller is enabled you will see extra stats in io.stat in
2051	addition to the normal ones.
2052
2053	  depth
2054		This is the current queue depth for the group.
2055
2056	  avg_lat
2057		This is an exponential moving average with a decay rate of 1/exp
2058		bound by the sampling interval.  The decay rate interval can be
2059		calculated by multiplying the win value in io.stat by the
2060		corresponding number of samples based on the win value.
2061
2062	  win
2063		The sampling window size in milliseconds.  This is the minimum
2064		duration of time between evaluation events.  Windows only elapse
2065		with IO activity.  Idle periods extend the most recent window.
2066
2067IO Priority
2068~~~~~~~~~~~
2069
2070A single attribute controls the behavior of the I/O priority cgroup policy,
2071namely the io.prio.class attribute. The following values are accepted for
2072that attribute:
2073
2074  no-change
2075	Do not modify the I/O priority class.
2076
2077  promote-to-rt
2078	For requests that have a non-RT I/O priority class, change it into RT.
2079	Also change the priority level of these requests to 4. Do not modify
2080	the I/O priority of requests that have priority class RT.
2081
2082  restrict-to-be
2083	For requests that do not have an I/O priority class or that have I/O
2084	priority class RT, change it into BE. Also change the priority level
2085	of these requests to 0. Do not modify the I/O priority class of
2086	requests that have priority class IDLE.
2087
2088  idle
2089	Change the I/O priority class of all requests into IDLE, the lowest
2090	I/O priority class.
2091
2092  none-to-rt
2093	Deprecated. Just an alias for promote-to-rt.
2094
2095The following numerical values are associated with the I/O priority policies:
2096
2097+----------------+---+
2098| no-change      | 0 |
2099+----------------+---+
2100| promote-to-rt  | 1 |
2101+----------------+---+
2102| restrict-to-be | 2 |
2103+----------------+---+
2104| idle           | 3 |
2105+----------------+---+
2106
2107The numerical value that corresponds to each I/O priority class is as follows:
2108
2109+-------------------------------+---+
2110| IOPRIO_CLASS_NONE             | 0 |
2111+-------------------------------+---+
2112| IOPRIO_CLASS_RT (real-time)   | 1 |
2113+-------------------------------+---+
2114| IOPRIO_CLASS_BE (best effort) | 2 |
2115+-------------------------------+---+
2116| IOPRIO_CLASS_IDLE             | 3 |
2117+-------------------------------+---+
2118
2119The algorithm to set the I/O priority class for a request is as follows:
2120
2121- If I/O priority class policy is promote-to-rt, change the request I/O
2122  priority class to IOPRIO_CLASS_RT and change the request I/O priority
2123  level to 4.
2124- If I/O priority class policy is not promote-to-rt, translate the I/O priority
2125  class policy into a number, then change the request I/O priority class
2126  into the maximum of the I/O priority class policy number and the numerical
2127  I/O priority class.
2128
2129PID
2130---
2131
2132The process number controller is used to allow a cgroup to stop any
2133new tasks from being fork()'d or clone()'d after a specified limit is
2134reached.
2135
2136The number of tasks in a cgroup can be exhausted in ways which other
2137controllers cannot prevent, thus warranting its own controller.  For
2138example, a fork bomb is likely to exhaust the number of tasks before
2139hitting memory restrictions.
2140
2141Note that PIDs used in this controller refer to TIDs, process IDs as
2142used by the kernel.
2143
2144
2145PID Interface Files
2146~~~~~~~~~~~~~~~~~~~
2147
2148  pids.max
2149	A read-write single value file which exists on non-root
2150	cgroups.  The default is "max".
2151
2152	Hard limit of number of processes.
2153
2154  pids.current
2155	A read-only single value file which exists on all cgroups.
2156
2157	The number of processes currently in the cgroup and its
2158	descendants.
2159
2160Organisational operations are not blocked by cgroup policies, so it is
2161possible to have pids.current > pids.max.  This can be done by either
2162setting the limit to be smaller than pids.current, or attaching enough
2163processes to the cgroup such that pids.current is larger than
2164pids.max.  However, it is not possible to violate a cgroup PID policy
2165through fork() or clone(). These will return -EAGAIN if the creation
2166of a new process would cause a cgroup policy to be violated.
2167
2168
2169Cpuset
2170------
2171
2172The "cpuset" controller provides a mechanism for constraining
2173the CPU and memory node placement of tasks to only the resources
2174specified in the cpuset interface files in a task's current cgroup.
2175This is especially valuable on large NUMA systems where placing jobs
2176on properly sized subsets of the systems with careful processor and
2177memory placement to reduce cross-node memory access and contention
2178can improve overall system performance.
2179
2180The "cpuset" controller is hierarchical.  That means the controller
2181cannot use CPUs or memory nodes not allowed in its parent.
2182
2183
2184Cpuset Interface Files
2185~~~~~~~~~~~~~~~~~~~~~~
2186
2187  cpuset.cpus
2188	A read-write multiple values file which exists on non-root
2189	cpuset-enabled cgroups.
2190
2191	It lists the requested CPUs to be used by tasks within this
2192	cgroup.  The actual list of CPUs to be granted, however, is
2193	subjected to constraints imposed by its parent and can differ
2194	from the requested CPUs.
2195
2196	The CPU numbers are comma-separated numbers or ranges.
2197	For example::
2198
2199	  # cat cpuset.cpus
2200	  0-4,6,8-10
2201
2202	An empty value indicates that the cgroup is using the same
2203	setting as the nearest cgroup ancestor with a non-empty
2204	"cpuset.cpus" or all the available CPUs if none is found.
2205
2206	The value of "cpuset.cpus" stays constant until the next update
2207	and won't be affected by any CPU hotplug events.
2208
2209  cpuset.cpus.effective
2210	A read-only multiple values file which exists on all
2211	cpuset-enabled cgroups.
2212
2213	It lists the onlined CPUs that are actually granted to this
2214	cgroup by its parent.  These CPUs are allowed to be used by
2215	tasks within the current cgroup.
2216
2217	If "cpuset.cpus" is empty, the "cpuset.cpus.effective" file shows
2218	all the CPUs from the parent cgroup that can be available to
2219	be used by this cgroup.  Otherwise, it should be a subset of
2220	"cpuset.cpus" unless none of the CPUs listed in "cpuset.cpus"
2221	can be granted.  In this case, it will be treated just like an
2222	empty "cpuset.cpus".
2223
2224	Its value will be affected by CPU hotplug events.
2225
2226  cpuset.mems
2227	A read-write multiple values file which exists on non-root
2228	cpuset-enabled cgroups.
2229
2230	It lists the requested memory nodes to be used by tasks within
2231	this cgroup.  The actual list of memory nodes granted, however,
2232	is subjected to constraints imposed by its parent and can differ
2233	from the requested memory nodes.
2234
2235	The memory node numbers are comma-separated numbers or ranges.
2236	For example::
2237
2238	  # cat cpuset.mems
2239	  0-1,3
2240
2241	An empty value indicates that the cgroup is using the same
2242	setting as the nearest cgroup ancestor with a non-empty
2243	"cpuset.mems" or all the available memory nodes if none
2244	is found.
2245
2246	The value of "cpuset.mems" stays constant until the next update
2247	and won't be affected by any memory nodes hotplug events.
2248
2249	Setting a non-empty value to "cpuset.mems" causes memory of
2250	tasks within the cgroup to be migrated to the designated nodes if
2251	they are currently using memory outside of the designated nodes.
2252
2253	There is a cost for this memory migration.  The migration
2254	may not be complete and some memory pages may be left behind.
2255	So it is recommended that "cpuset.mems" should be set properly
2256	before spawning new tasks into the cpuset.  Even if there is
2257	a need to change "cpuset.mems" with active tasks, it shouldn't
2258	be done frequently.
2259
2260  cpuset.mems.effective
2261	A read-only multiple values file which exists on all
2262	cpuset-enabled cgroups.
2263
2264	It lists the onlined memory nodes that are actually granted to
2265	this cgroup by its parent. These memory nodes are allowed to
2266	be used by tasks within the current cgroup.
2267
2268	If "cpuset.mems" is empty, it shows all the memory nodes from the
2269	parent cgroup that will be available to be used by this cgroup.
2270	Otherwise, it should be a subset of "cpuset.mems" unless none of
2271	the memory nodes listed in "cpuset.mems" can be granted.  In this
2272	case, it will be treated just like an empty "cpuset.mems".
2273
2274	Its value will be affected by memory nodes hotplug events.
2275
2276  cpuset.cpus.exclusive
2277	A read-write multiple values file which exists on non-root
2278	cpuset-enabled cgroups.
2279
2280	It lists all the exclusive CPUs that are allowed to be used
2281	to create a new cpuset partition.  Its value is not used
2282	unless the cgroup becomes a valid partition root.  See the
2283	"cpuset.cpus.partition" section below for a description of what
2284	a cpuset partition is.
2285
2286	When the cgroup becomes a partition root, the actual exclusive
2287	CPUs that are allocated to that partition are listed in
2288	"cpuset.cpus.exclusive.effective" which may be different
2289	from "cpuset.cpus.exclusive".  If "cpuset.cpus.exclusive"
2290	has previously been set, "cpuset.cpus.exclusive.effective"
2291	is always a subset of it.
2292
2293	Users can manually set it to a value that is different from
2294	"cpuset.cpus".	The only constraint in setting it is that the
2295	list of CPUs must be exclusive with respect to its sibling.
2296
2297	For a parent cgroup, any one of its exclusive CPUs can only
2298	be distributed to at most one of its child cgroups.  Having an
2299	exclusive CPU appearing in two or more of its child cgroups is
2300	not allowed (the exclusivity rule).  A value that violates the
2301	exclusivity rule will be rejected with a write error.
2302
2303	The root cgroup is a partition root and all its available CPUs
2304	are in its exclusive CPU set.
2305
2306  cpuset.cpus.exclusive.effective
2307	A read-only multiple values file which exists on all non-root
2308	cpuset-enabled cgroups.
2309
2310	This file shows the effective set of exclusive CPUs that
2311	can be used to create a partition root.  The content of this
2312	file will always be a subset of "cpuset.cpus" and its parent's
2313	"cpuset.cpus.exclusive.effective" if its parent is not the root
2314	cgroup.  It will also be a subset of "cpuset.cpus.exclusive"
2315	if it is set.  If "cpuset.cpus.exclusive" is not set, it is
2316	treated to have an implicit value of "cpuset.cpus" in the
2317	formation of local partition.
2318
2319  cpuset.cpus.partition
2320	A read-write single value file which exists on non-root
2321	cpuset-enabled cgroups.  This flag is owned by the parent cgroup
2322	and is not delegatable.
2323
2324	It accepts only the following input values when written to.
2325
2326	  ==========	=====================================
2327	  "member"	Non-root member of a partition
2328	  "root"	Partition root
2329	  "isolated"	Partition root without load balancing
2330	  ==========	=====================================
2331
2332	A cpuset partition is a collection of cpuset-enabled cgroups with
2333	a partition root at the top of the hierarchy and its descendants
2334	except those that are separate partition roots themselves and
2335	their descendants.  A partition has exclusive access to the
2336	set of exclusive CPUs allocated to it.	Other cgroups outside
2337	of that partition cannot use any CPUs in that set.
2338
2339	There are two types of partitions - local and remote.  A local
2340	partition is one whose parent cgroup is also a valid partition
2341	root.  A remote partition is one whose parent cgroup is not a
2342	valid partition root itself.  Writing to "cpuset.cpus.exclusive"
2343	is optional for the creation of a local partition as its
2344	"cpuset.cpus.exclusive" file will assume an implicit value that
2345	is the same as "cpuset.cpus" if it is not set.	Writing the
2346	proper "cpuset.cpus.exclusive" values down the cgroup hierarchy
2347	before the target partition root is mandatory for the creation
2348	of a remote partition.
2349
2350	Currently, a remote partition cannot be created under a local
2351	partition.  All the ancestors of a remote partition root except
2352	the root cgroup cannot be a partition root.
2353
2354	The root cgroup is always a partition root and its state cannot
2355	be changed.  All other non-root cgroups start out as "member".
2356
2357	When set to "root", the current cgroup is the root of a new
2358	partition or scheduling domain.  The set of exclusive CPUs is
2359	determined by the value of its "cpuset.cpus.exclusive.effective".
2360
2361	When set to "isolated", the CPUs in that partition will
2362	be in an isolated state without any load balancing from the
2363	scheduler.  Tasks placed in such a partition with multiple
2364	CPUs should be carefully distributed and bound to each of the
2365	individual CPUs for optimal performance.
2366
2367	A partition root ("root" or "isolated") can be in one of the
2368	two possible states - valid or invalid.  An invalid partition
2369	root is in a degraded state where some state information may
2370	be retained, but behaves more like a "member".
2371
2372	All possible state transitions among "member", "root" and
2373	"isolated" are allowed.
2374
2375	On read, the "cpuset.cpus.partition" file can show the following
2376	values.
2377
2378	  =============================	=====================================
2379	  "member"			Non-root member of a partition
2380	  "root"			Partition root
2381	  "isolated"			Partition root without load balancing
2382	  "root invalid (<reason>)"	Invalid partition root
2383	  "isolated invalid (<reason>)"	Invalid isolated partition root
2384	  =============================	=====================================
2385
2386	In the case of an invalid partition root, a descriptive string on
2387	why the partition is invalid is included within parentheses.
2388
2389	For a local partition root to be valid, the following conditions
2390	must be met.
2391
2392	1) The parent cgroup is a valid partition root.
2393	2) The "cpuset.cpus.exclusive.effective" file cannot be empty,
2394	   though it may contain offline CPUs.
2395	3) The "cpuset.cpus.effective" cannot be empty unless there is
2396	   no task associated with this partition.
2397
2398	For a remote partition root to be valid, all the above conditions
2399	except the first one must be met.
2400
2401	External events like hotplug or changes to "cpuset.cpus" or
2402	"cpuset.cpus.exclusive" can cause a valid partition root to
2403	become invalid and vice versa.	Note that a task cannot be
2404	moved to a cgroup with empty "cpuset.cpus.effective".
2405
2406	A valid non-root parent partition may distribute out all its CPUs
2407	to its child local partitions when there is no task associated
2408	with it.
2409
2410	Care must be taken to change a valid partition root to "member"
2411	as all its child local partitions, if present, will become
2412	invalid causing disruption to tasks running in those child
2413	partitions. These inactivated partitions could be recovered if
2414	their parent is switched back to a partition root with a proper
2415	value in "cpuset.cpus" or "cpuset.cpus.exclusive".
2416
2417	Poll and inotify events are triggered whenever the state of
2418	"cpuset.cpus.partition" changes.  That includes changes caused
2419	by write to "cpuset.cpus.partition", cpu hotplug or other
2420	changes that modify the validity status of the partition.
2421	This will allow user space agents to monitor unexpected changes
2422	to "cpuset.cpus.partition" without the need to do continuous
2423	polling.
2424
2425	A user can pre-configure certain CPUs to an isolated state
2426	with load balancing disabled at boot time with the "isolcpus"
2427	kernel boot command line option.  If those CPUs are to be put
2428	into a partition, they have to be used in an isolated partition.
2429
2430
2431Device controller
2432-----------------
2433
2434Device controller manages access to device files. It includes both
2435creation of new device files (using mknod), and access to the
2436existing device files.
2437
2438Cgroup v2 device controller has no interface files and is implemented
2439on top of cgroup BPF. To control access to device files, a user may
2440create bpf programs of type BPF_PROG_TYPE_CGROUP_DEVICE and attach
2441them to cgroups with BPF_CGROUP_DEVICE flag. On an attempt to access a
2442device file, corresponding BPF programs will be executed, and depending
2443on the return value the attempt will succeed or fail with -EPERM.
2444
2445A BPF_PROG_TYPE_CGROUP_DEVICE program takes a pointer to the
2446bpf_cgroup_dev_ctx structure, which describes the device access attempt:
2447access type (mknod/read/write) and device (type, major and minor numbers).
2448If the program returns 0, the attempt fails with -EPERM, otherwise it
2449succeeds.
2450
2451An example of BPF_PROG_TYPE_CGROUP_DEVICE program may be found in
2452tools/testing/selftests/bpf/progs/dev_cgroup.c in the kernel source tree.
2453
2454
2455RDMA
2456----
2457
2458The "rdma" controller regulates the distribution and accounting of
2459RDMA resources.
2460
2461RDMA Interface Files
2462~~~~~~~~~~~~~~~~~~~~
2463
2464  rdma.max
2465	A readwrite nested-keyed file that exists for all the cgroups
2466	except root that describes current configured resource limit
2467	for a RDMA/IB device.
2468
2469	Lines are keyed by device name and are not ordered.
2470	Each line contains space separated resource name and its configured
2471	limit that can be distributed.
2472
2473	The following nested keys are defined.
2474
2475	  ==========	=============================
2476	  hca_handle	Maximum number of HCA Handles
2477	  hca_object 	Maximum number of HCA Objects
2478	  ==========	=============================
2479
2480	An example for mlx4 and ocrdma device follows::
2481
2482	  mlx4_0 hca_handle=2 hca_object=2000
2483	  ocrdma1 hca_handle=3 hca_object=max
2484
2485  rdma.current
2486	A read-only file that describes current resource usage.
2487	It exists for all the cgroup except root.
2488
2489	An example for mlx4 and ocrdma device follows::
2490
2491	  mlx4_0 hca_handle=1 hca_object=20
2492	  ocrdma1 hca_handle=1 hca_object=23
2493
2494HugeTLB
2495-------
2496
2497The HugeTLB controller allows to limit the HugeTLB usage per control group and
2498enforces the controller limit during page fault.
2499
2500HugeTLB Interface Files
2501~~~~~~~~~~~~~~~~~~~~~~~
2502
2503  hugetlb.<hugepagesize>.current
2504	Show current usage for "hugepagesize" hugetlb.  It exists for all
2505	the cgroup except root.
2506
2507  hugetlb.<hugepagesize>.max
2508	Set/show the hard limit of "hugepagesize" hugetlb usage.
2509	The default value is "max".  It exists for all the cgroup except root.
2510
2511  hugetlb.<hugepagesize>.events
2512	A read-only flat-keyed file which exists on non-root cgroups.
2513
2514	  max
2515		The number of allocation failure due to HugeTLB limit
2516
2517  hugetlb.<hugepagesize>.events.local
2518	Similar to hugetlb.<hugepagesize>.events but the fields in the file
2519	are local to the cgroup i.e. not hierarchical. The file modified event
2520	generated on this file reflects only the local events.
2521
2522  hugetlb.<hugepagesize>.numa_stat
2523	Similar to memory.numa_stat, it shows the numa information of the
2524        hugetlb pages of <hugepagesize> in this cgroup.  Only active in
2525        use hugetlb pages are included.  The per-node values are in bytes.
2526
2527Misc
2528----
2529
2530The Miscellaneous cgroup provides the resource limiting and tracking
2531mechanism for the scalar resources which cannot be abstracted like the other
2532cgroup resources. Controller is enabled by the CONFIG_CGROUP_MISC config
2533option.
2534
2535A resource can be added to the controller via enum misc_res_type{} in the
2536include/linux/misc_cgroup.h file and the corresponding name via misc_res_name[]
2537in the kernel/cgroup/misc.c file. Provider of the resource must set its
2538capacity prior to using the resource by calling misc_cg_set_capacity().
2539
2540Once a capacity is set then the resource usage can be updated using charge and
2541uncharge APIs. All of the APIs to interact with misc controller are in
2542include/linux/misc_cgroup.h.
2543
2544Misc Interface Files
2545~~~~~~~~~~~~~~~~~~~~
2546
2547Miscellaneous controller provides 3 interface files. If two misc resources (res_a and res_b) are registered then:
2548
2549  misc.capacity
2550        A read-only flat-keyed file shown only in the root cgroup.  It shows
2551        miscellaneous scalar resources available on the platform along with
2552        their quantities::
2553
2554	  $ cat misc.capacity
2555	  res_a 50
2556	  res_b 10
2557
2558  misc.current
2559        A read-only flat-keyed file shown in the all cgroups.  It shows
2560        the current usage of the resources in the cgroup and its children.::
2561
2562	  $ cat misc.current
2563	  res_a 3
2564	  res_b 0
2565
2566  misc.max
2567        A read-write flat-keyed file shown in the non root cgroups. Allowed
2568        maximum usage of the resources in the cgroup and its children.::
2569
2570	  $ cat misc.max
2571	  res_a max
2572	  res_b 4
2573
2574	Limit can be set by::
2575
2576	  # echo res_a 1 > misc.max
2577
2578	Limit can be set to max by::
2579
2580	  # echo res_a max > misc.max
2581
2582        Limits can be set higher than the capacity value in the misc.capacity
2583        file.
2584
2585  misc.events
2586	A read-only flat-keyed file which exists on non-root cgroups. The
2587	following entries are defined. Unless specified otherwise, a value
2588	change in this file generates a file modified event. All fields in
2589	this file are hierarchical.
2590
2591	  max
2592		The number of times the cgroup's resource usage was
2593		about to go over the max boundary.
2594
2595Migration and Ownership
2596~~~~~~~~~~~~~~~~~~~~~~~
2597
2598A miscellaneous scalar resource is charged to the cgroup in which it is used
2599first, and stays charged to that cgroup until that resource is freed. Migrating
2600a process to a different cgroup does not move the charge to the destination
2601cgroup where the process has moved.
2602
2603Others
2604------
2605
2606perf_event
2607~~~~~~~~~~
2608
2609perf_event controller, if not mounted on a legacy hierarchy, is
2610automatically enabled on the v2 hierarchy so that perf events can
2611always be filtered by cgroup v2 path.  The controller can still be
2612moved to a legacy hierarchy after v2 hierarchy is populated.
2613
2614
2615Non-normative information
2616-------------------------
2617
2618This section contains information that isn't considered to be a part of
2619the stable kernel API and so is subject to change.
2620
2621
2622CPU controller root cgroup process behaviour
2623~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2624
2625When distributing CPU cycles in the root cgroup each thread in this
2626cgroup is treated as if it was hosted in a separate child cgroup of the
2627root cgroup. This child cgroup weight is dependent on its thread nice
2628level.
2629
2630For details of this mapping see sched_prio_to_weight array in
2631kernel/sched/core.c file (values from this array should be scaled
2632appropriately so the neutral - nice 0 - value is 100 instead of 1024).
2633
2634
2635IO controller root cgroup process behaviour
2636~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2637
2638Root cgroup processes are hosted in an implicit leaf child node.
2639When distributing IO resources this implicit child node is taken into
2640account as if it was a normal child cgroup of the root cgroup with a
2641weight value of 200.
2642
2643
2644Namespace
2645=========
2646
2647Basics
2648------
2649
2650cgroup namespace provides a mechanism to virtualize the view of the
2651"/proc/$PID/cgroup" file and cgroup mounts.  The CLONE_NEWCGROUP clone
2652flag can be used with clone(2) and unshare(2) to create a new cgroup
2653namespace.  The process running inside the cgroup namespace will have
2654its "/proc/$PID/cgroup" output restricted to cgroupns root.  The
2655cgroupns root is the cgroup of the process at the time of creation of
2656the cgroup namespace.
2657
2658Without cgroup namespace, the "/proc/$PID/cgroup" file shows the
2659complete path of the cgroup of a process.  In a container setup where
2660a set of cgroups and namespaces are intended to isolate processes the
2661"/proc/$PID/cgroup" file may leak potential system level information
2662to the isolated processes.  For example::
2663
2664  # cat /proc/self/cgroup
2665  0::/batchjobs/container_id1
2666
2667The path '/batchjobs/container_id1' can be considered as system-data
2668and undesirable to expose to the isolated processes.  cgroup namespace
2669can be used to restrict visibility of this path.  For example, before
2670creating a cgroup namespace, one would see::
2671
2672  # ls -l /proc/self/ns/cgroup
2673  lrwxrwxrwx 1 root root 0 2014-07-15 10:37 /proc/self/ns/cgroup -> cgroup:[4026531835]
2674  # cat /proc/self/cgroup
2675  0::/batchjobs/container_id1
2676
2677After unsharing a new namespace, the view changes::
2678
2679  # ls -l /proc/self/ns/cgroup
2680  lrwxrwxrwx 1 root root 0 2014-07-15 10:35 /proc/self/ns/cgroup -> cgroup:[4026532183]
2681  # cat /proc/self/cgroup
2682  0::/
2683
2684When some thread from a multi-threaded process unshares its cgroup
2685namespace, the new cgroupns gets applied to the entire process (all
2686the threads).  This is natural for the v2 hierarchy; however, for the
2687legacy hierarchies, this may be unexpected.
2688
2689A cgroup namespace is alive as long as there are processes inside or
2690mounts pinning it.  When the last usage goes away, the cgroup
2691namespace is destroyed.  The cgroupns root and the actual cgroups
2692remain.
2693
2694
2695The Root and Views
2696------------------
2697
2698The 'cgroupns root' for a cgroup namespace is the cgroup in which the
2699process calling unshare(2) is running.  For example, if a process in
2700/batchjobs/container_id1 cgroup calls unshare, cgroup
2701/batchjobs/container_id1 becomes the cgroupns root.  For the
2702init_cgroup_ns, this is the real root ('/') cgroup.
2703
2704The cgroupns root cgroup does not change even if the namespace creator
2705process later moves to a different cgroup::
2706
2707  # ~/unshare -c # unshare cgroupns in some cgroup
2708  # cat /proc/self/cgroup
2709  0::/
2710  # mkdir sub_cgrp_1
2711  # echo 0 > sub_cgrp_1/cgroup.procs
2712  # cat /proc/self/cgroup
2713  0::/sub_cgrp_1
2714
2715Each process gets its namespace-specific view of "/proc/$PID/cgroup"
2716
2717Processes running inside the cgroup namespace will be able to see
2718cgroup paths (in /proc/self/cgroup) only inside their root cgroup.
2719From within an unshared cgroupns::
2720
2721  # sleep 100000 &
2722  [1] 7353
2723  # echo 7353 > sub_cgrp_1/cgroup.procs
2724  # cat /proc/7353/cgroup
2725  0::/sub_cgrp_1
2726
2727From the initial cgroup namespace, the real cgroup path will be
2728visible::
2729
2730  $ cat /proc/7353/cgroup
2731  0::/batchjobs/container_id1/sub_cgrp_1
2732
2733From a sibling cgroup namespace (that is, a namespace rooted at a
2734different cgroup), the cgroup path relative to its own cgroup
2735namespace root will be shown.  For instance, if PID 7353's cgroup
2736namespace root is at '/batchjobs/container_id2', then it will see::
2737
2738  # cat /proc/7353/cgroup
2739  0::/../container_id2/sub_cgrp_1
2740
2741Note that the relative path always starts with '/' to indicate that
2742its relative to the cgroup namespace root of the caller.
2743
2744
2745Migration and setns(2)
2746----------------------
2747
2748Processes inside a cgroup namespace can move into and out of the
2749namespace root if they have proper access to external cgroups.  For
2750example, from inside a namespace with cgroupns root at
2751/batchjobs/container_id1, and assuming that the global hierarchy is
2752still accessible inside cgroupns::
2753
2754  # cat /proc/7353/cgroup
2755  0::/sub_cgrp_1
2756  # echo 7353 > batchjobs/container_id2/cgroup.procs
2757  # cat /proc/7353/cgroup
2758  0::/../container_id2
2759
2760Note that this kind of setup is not encouraged.  A task inside cgroup
2761namespace should only be exposed to its own cgroupns hierarchy.
2762
2763setns(2) to another cgroup namespace is allowed when:
2764
2765(a) the process has CAP_SYS_ADMIN against its current user namespace
2766(b) the process has CAP_SYS_ADMIN against the target cgroup
2767    namespace's userns
2768
2769No implicit cgroup changes happen with attaching to another cgroup
2770namespace.  It is expected that the someone moves the attaching
2771process under the target cgroup namespace root.
2772
2773
2774Interaction with Other Namespaces
2775---------------------------------
2776
2777Namespace specific cgroup hierarchy can be mounted by a process
2778running inside a non-init cgroup namespace::
2779
2780  # mount -t cgroup2 none $MOUNT_POINT
2781
2782This will mount the unified cgroup hierarchy with cgroupns root as the
2783filesystem root.  The process needs CAP_SYS_ADMIN against its user and
2784mount namespaces.
2785
2786The virtualization of /proc/self/cgroup file combined with restricting
2787the view of cgroup hierarchy by namespace-private cgroupfs mount
2788provides a properly isolated cgroup view inside the container.
2789
2790
2791Information on Kernel Programming
2792=================================
2793
2794This section contains kernel programming information in the areas
2795where interacting with cgroup is necessary.  cgroup core and
2796controllers are not covered.
2797
2798
2799Filesystem Support for Writeback
2800--------------------------------
2801
2802A filesystem can support cgroup writeback by updating
2803address_space_operations->writepage[s]() to annotate bio's using the
2804following two functions.
2805
2806  wbc_init_bio(@wbc, @bio)
2807	Should be called for each bio carrying writeback data and
2808	associates the bio with the inode's owner cgroup and the
2809	corresponding request queue.  This must be called after
2810	a queue (device) has been associated with the bio and
2811	before submission.
2812
2813  wbc_account_cgroup_owner(@wbc, @page, @bytes)
2814	Should be called for each data segment being written out.
2815	While this function doesn't care exactly when it's called
2816	during the writeback session, it's the easiest and most
2817	natural to call it as data segments are added to a bio.
2818
2819With writeback bio's annotated, cgroup support can be enabled per
2820super_block by setting SB_I_CGROUPWB in ->s_iflags.  This allows for
2821selective disabling of cgroup writeback support which is helpful when
2822certain filesystem features, e.g. journaled data mode, are
2823incompatible.
2824
2825wbc_init_bio() binds the specified bio to its cgroup.  Depending on
2826the configuration, the bio may be executed at a lower priority and if
2827the writeback session is holding shared resources, e.g. a journal
2828entry, may lead to priority inversion.  There is no one easy solution
2829for the problem.  Filesystems can try to work around specific problem
2830cases by skipping wbc_init_bio() and using bio_associate_blkg()
2831directly.
2832
2833
2834Deprecated v1 Core Features
2835===========================
2836
2837- Multiple hierarchies including named ones are not supported.
2838
2839- All v1 mount options are not supported.
2840
2841- The "tasks" file is removed and "cgroup.procs" is not sorted.
2842
2843- "cgroup.clone_children" is removed.
2844
2845- /proc/cgroups is meaningless for v2.  Use "cgroup.controllers" file
2846  at the root instead.
2847
2848
2849Issues with v1 and Rationales for v2
2850====================================
2851
2852Multiple Hierarchies
2853--------------------
2854
2855cgroup v1 allowed an arbitrary number of hierarchies and each
2856hierarchy could host any number of controllers.  While this seemed to
2857provide a high level of flexibility, it wasn't useful in practice.
2858
2859For example, as there is only one instance of each controller, utility
2860type controllers such as freezer which can be useful in all
2861hierarchies could only be used in one.  The issue is exacerbated by
2862the fact that controllers couldn't be moved to another hierarchy once
2863hierarchies were populated.  Another issue was that all controllers
2864bound to a hierarchy were forced to have exactly the same view of the
2865hierarchy.  It wasn't possible to vary the granularity depending on
2866the specific controller.
2867
2868In practice, these issues heavily limited which controllers could be
2869put on the same hierarchy and most configurations resorted to putting
2870each controller on its own hierarchy.  Only closely related ones, such
2871as the cpu and cpuacct controllers, made sense to be put on the same
2872hierarchy.  This often meant that userland ended up managing multiple
2873similar hierarchies repeating the same steps on each hierarchy
2874whenever a hierarchy management operation was necessary.
2875
2876Furthermore, support for multiple hierarchies came at a steep cost.
2877It greatly complicated cgroup core implementation but more importantly
2878the support for multiple hierarchies restricted how cgroup could be
2879used in general and what controllers was able to do.
2880
2881There was no limit on how many hierarchies there might be, which meant
2882that a thread's cgroup membership couldn't be described in finite
2883length.  The key might contain any number of entries and was unlimited
2884in length, which made it highly awkward to manipulate and led to
2885addition of controllers which existed only to identify membership,
2886which in turn exacerbated the original problem of proliferating number
2887of hierarchies.
2888
2889Also, as a controller couldn't have any expectation regarding the
2890topologies of hierarchies other controllers might be on, each
2891controller had to assume that all other controllers were attached to
2892completely orthogonal hierarchies.  This made it impossible, or at
2893least very cumbersome, for controllers to cooperate with each other.
2894
2895In most use cases, putting controllers on hierarchies which are
2896completely orthogonal to each other isn't necessary.  What usually is
2897called for is the ability to have differing levels of granularity
2898depending on the specific controller.  In other words, hierarchy may
2899be collapsed from leaf towards root when viewed from specific
2900controllers.  For example, a given configuration might not care about
2901how memory is distributed beyond a certain level while still wanting
2902to control how CPU cycles are distributed.
2903
2904
2905Thread Granularity
2906------------------
2907
2908cgroup v1 allowed threads of a process to belong to different cgroups.
2909This didn't make sense for some controllers and those controllers
2910ended up implementing different ways to ignore such situations but
2911much more importantly it blurred the line between API exposed to
2912individual applications and system management interface.
2913
2914Generally, in-process knowledge is available only to the process
2915itself; thus, unlike service-level organization of processes,
2916categorizing threads of a process requires active participation from
2917the application which owns the target process.
2918
2919cgroup v1 had an ambiguously defined delegation model which got abused
2920in combination with thread granularity.  cgroups were delegated to
2921individual applications so that they can create and manage their own
2922sub-hierarchies and control resource distributions along them.  This
2923effectively raised cgroup to the status of a syscall-like API exposed
2924to lay programs.
2925
2926First of all, cgroup has a fundamentally inadequate interface to be
2927exposed this way.  For a process to access its own knobs, it has to
2928extract the path on the target hierarchy from /proc/self/cgroup,
2929construct the path by appending the name of the knob to the path, open
2930and then read and/or write to it.  This is not only extremely clunky
2931and unusual but also inherently racy.  There is no conventional way to
2932define transaction across the required steps and nothing can guarantee
2933that the process would actually be operating on its own sub-hierarchy.
2934
2935cgroup controllers implemented a number of knobs which would never be
2936accepted as public APIs because they were just adding control knobs to
2937system-management pseudo filesystem.  cgroup ended up with interface
2938knobs which were not properly abstracted or refined and directly
2939revealed kernel internal details.  These knobs got exposed to
2940individual applications through the ill-defined delegation mechanism
2941effectively abusing cgroup as a shortcut to implementing public APIs
2942without going through the required scrutiny.
2943
2944This was painful for both userland and kernel.  Userland ended up with
2945misbehaving and poorly abstracted interfaces and kernel exposing and
2946locked into constructs inadvertently.
2947
2948
2949Competition Between Inner Nodes and Threads
2950-------------------------------------------
2951
2952cgroup v1 allowed threads to be in any cgroups which created an
2953interesting problem where threads belonging to a parent cgroup and its
2954children cgroups competed for resources.  This was nasty as two
2955different types of entities competed and there was no obvious way to
2956settle it.  Different controllers did different things.
2957
2958The cpu controller considered threads and cgroups as equivalents and
2959mapped nice levels to cgroup weights.  This worked for some cases but
2960fell flat when children wanted to be allocated specific ratios of CPU
2961cycles and the number of internal threads fluctuated - the ratios
2962constantly changed as the number of competing entities fluctuated.
2963There also were other issues.  The mapping from nice level to weight
2964wasn't obvious or universal, and there were various other knobs which
2965simply weren't available for threads.
2966
2967The io controller implicitly created a hidden leaf node for each
2968cgroup to host the threads.  The hidden leaf had its own copies of all
2969the knobs with ``leaf_`` prefixed.  While this allowed equivalent
2970control over internal threads, it was with serious drawbacks.  It
2971always added an extra layer of nesting which wouldn't be necessary
2972otherwise, made the interface messy and significantly complicated the
2973implementation.
2974
2975The memory controller didn't have a way to control what happened
2976between internal tasks and child cgroups and the behavior was not
2977clearly defined.  There were attempts to add ad-hoc behaviors and
2978knobs to tailor the behavior to specific workloads which would have
2979led to problems extremely difficult to resolve in the long term.
2980
2981Multiple controllers struggled with internal tasks and came up with
2982different ways to deal with it; unfortunately, all the approaches were
2983severely flawed and, furthermore, the widely different behaviors
2984made cgroup as a whole highly inconsistent.
2985
2986This clearly is a problem which needs to be addressed from cgroup core
2987in a uniform way.
2988
2989
2990Other Interface Issues
2991----------------------
2992
2993cgroup v1 grew without oversight and developed a large number of
2994idiosyncrasies and inconsistencies.  One issue on the cgroup core side
2995was how an empty cgroup was notified - a userland helper binary was
2996forked and executed for each event.  The event delivery wasn't
2997recursive or delegatable.  The limitations of the mechanism also led
2998to in-kernel event delivery filtering mechanism further complicating
2999the interface.
3000
3001Controller interfaces were problematic too.  An extreme example is
3002controllers completely ignoring hierarchical organization and treating
3003all cgroups as if they were all located directly under the root
3004cgroup.  Some controllers exposed a large amount of inconsistent
3005implementation details to userland.
3006
3007There also was no consistency across controllers.  When a new cgroup
3008was created, some controllers defaulted to not imposing extra
3009restrictions while others disallowed any resource usage until
3010explicitly configured.  Configuration knobs for the same type of
3011control used widely differing naming schemes and formats.  Statistics
3012and information knobs were named arbitrarily and used different
3013formats and units even in the same controller.
3014
3015cgroup v2 establishes common conventions where appropriate and updates
3016controllers so that they expose minimal and consistent interfaces.
3017
3018
3019Controller Issues and Remedies
3020------------------------------
3021
3022Memory
3023~~~~~~
3024
3025The original lower boundary, the soft limit, is defined as a limit
3026that is per default unset.  As a result, the set of cgroups that
3027global reclaim prefers is opt-in, rather than opt-out.  The costs for
3028optimizing these mostly negative lookups are so high that the
3029implementation, despite its enormous size, does not even provide the
3030basic desirable behavior.  First off, the soft limit has no
3031hierarchical meaning.  All configured groups are organized in a global
3032rbtree and treated like equal peers, regardless where they are located
3033in the hierarchy.  This makes subtree delegation impossible.  Second,
3034the soft limit reclaim pass is so aggressive that it not just
3035introduces high allocation latencies into the system, but also impacts
3036system performance due to overreclaim, to the point where the feature
3037becomes self-defeating.
3038
3039The memory.low boundary on the other hand is a top-down allocated
3040reserve.  A cgroup enjoys reclaim protection when it's within its
3041effective low, which makes delegation of subtrees possible. It also
3042enjoys having reclaim pressure proportional to its overage when
3043above its effective low.
3044
3045The original high boundary, the hard limit, is defined as a strict
3046limit that can not budge, even if the OOM killer has to be called.
3047But this generally goes against the goal of making the most out of the
3048available memory.  The memory consumption of workloads varies during
3049runtime, and that requires users to overcommit.  But doing that with a
3050strict upper limit requires either a fairly accurate prediction of the
3051working set size or adding slack to the limit.  Since working set size
3052estimation is hard and error prone, and getting it wrong results in
3053OOM kills, most users tend to err on the side of a looser limit and
3054end up wasting precious resources.
3055
3056The memory.high boundary on the other hand can be set much more
3057conservatively.  When hit, it throttles allocations by forcing them
3058into direct reclaim to work off the excess, but it never invokes the
3059OOM killer.  As a result, a high boundary that is chosen too
3060aggressively will not terminate the processes, but instead it will
3061lead to gradual performance degradation.  The user can monitor this
3062and make corrections until the minimal memory footprint that still
3063gives acceptable performance is found.
3064
3065In extreme cases, with many concurrent allocations and a complete
3066breakdown of reclaim progress within the group, the high boundary can
3067be exceeded.  But even then it's mostly better to satisfy the
3068allocation from the slack available in other groups or the rest of the
3069system than killing the group.  Otherwise, memory.max is there to
3070limit this type of spillover and ultimately contain buggy or even
3071malicious applications.
3072
3073Setting the original memory.limit_in_bytes below the current usage was
3074subject to a race condition, where concurrent charges could cause the
3075limit setting to fail. memory.max on the other hand will first set the
3076limit to prevent new charges, and then reclaim and OOM kill until the
3077new limit is met - or the task writing to memory.max is killed.
3078
3079The combined memory+swap accounting and limiting is replaced by real
3080control over swap space.
3081
3082The main argument for a combined memory+swap facility in the original
3083cgroup design was that global or parental pressure would always be
3084able to swap all anonymous memory of a child group, regardless of the
3085child's own (possibly untrusted) configuration.  However, untrusted
3086groups can sabotage swapping by other means - such as referencing its
3087anonymous memory in a tight loop - and an admin can not assume full
3088swappability when overcommitting untrusted jobs.
3089
3090For trusted jobs, on the other hand, a combined counter is not an
3091intuitive userspace interface, and it flies in the face of the idea
3092that cgroup controllers should account and limit specific physical
3093resources.  Swap space is a resource like all others in the system,
3094and that's why unified hierarchy allows distributing it separately.
3095