blob: 98a26f81fa75062b104edb796f3b9a0449f6f28c [file] [log] [blame]
Paul Menageddbcc7e2007-10-18 23:39:30 -07001 CGROUPS
2 -------
3
4Written by Paul Menage <menage@google.com> based on Documentation/cpusets.txt
5
6Original copyright statements from cpusets.txt:
7Portions Copyright (C) 2004 BULL SA.
8Portions Copyright (c) 2004-2006 Silicon Graphics, Inc.
9Modified by Paul Jackson <pj@sgi.com>
10Modified by Christoph Lameter <clameter@sgi.com>
11
12CONTENTS:
13=========
14
151. Control Groups
16 1.1 What are cgroups ?
17 1.2 Why are cgroups needed ?
18 1.3 How are cgroups implemented ?
19 1.4 What does notify_on_release do ?
20 1.5 How do I use cgroups ?
212. Usage Examples and Syntax
22 2.1 Basic Usage
23 2.2 Attaching processes
243. Kernel API
25 3.1 Overview
26 3.2 Synchronization
27 3.3 Subsystem API
284. Questions
29
301. Control Groups
31==========
32
331.1 What are cgroups ?
34----------------------
35
36Control Groups provide a mechanism for aggregating/partitioning sets of
37tasks, and all their future children, into hierarchical groups with
38specialized behaviour.
39
40Definitions:
41
42A *cgroup* associates a set of tasks with a set of parameters for one
43or more subsystems.
44
45A *subsystem* is a module that makes use of the task grouping
46facilities provided by cgroups to treat groups of tasks in
47particular ways. A subsystem is typically a "resource controller" that
48schedules a resource or applies per-cgroup limits, but it may be
49anything that wants to act on a group of processes, e.g. a
50virtualization subsystem.
51
52A *hierarchy* is a set of cgroups arranged in a tree, such that
53every task in the system is in exactly one of the cgroups in the
54hierarchy, and a set of subsystems; each subsystem has system-specific
55state attached to each cgroup in the hierarchy. Each hierarchy has
56an instance of the cgroup virtual filesystem associated with it.
57
58At any one time there may be multiple active hierachies of task
59cgroups. Each hierarchy is a partition of all tasks in the system.
60
61User level code may create and destroy cgroups by name in an
62instance of the cgroup virtual file system, specify and query to
63which cgroup a task is assigned, and list the task pids assigned to
64a cgroup. Those creations and assignments only affect the hierarchy
65associated with that instance of the cgroup file system.
66
67On their own, the only use for cgroups is for simple job
68tracking. The intention is that other subsystems hook into the generic
69cgroup support to provide new attributes for cgroups, such as
70accounting/limiting the resources which processes in a cgroup can
71access. For example, cpusets (see Documentation/cpusets.txt) allows
72you to associate a set of CPUs and a set of memory nodes with the
73tasks in each cgroup.
74
751.2 Why are cgroups needed ?
76----------------------------
77
78There are multiple efforts to provide process aggregations in the
79Linux kernel, mainly for resource tracking purposes. Such efforts
80include cpusets, CKRM/ResGroups, UserBeanCounters, and virtual server
81namespaces. These all require the basic notion of a
82grouping/partitioning of processes, with newly forked processes ending
83in the same group (cgroup) as their parent process.
84
85The kernel cgroup patch provides the minimum essential kernel
86mechanisms required to efficiently implement such groups. It has
87minimal impact on the system fast paths, and provides hooks for
88specific subsystems such as cpusets to provide additional behaviour as
89desired.
90
91Multiple hierarchy support is provided to allow for situations where
92the division of tasks into cgroups is distinctly different for
93different subsystems - having parallel hierarchies allows each
94hierarchy to be a natural division of tasks, without having to handle
95complex combinations of tasks that would be present if several
96unrelated subsystems needed to be forced into the same tree of
97cgroups.
98
99At one extreme, each resource controller or subsystem could be in a
100separate hierarchy; at the other extreme, all subsystems
101would be attached to the same hierarchy.
102
103As an example of a scenario (originally proposed by vatsa@in.ibm.com)
104that can benefit from multiple hierarchies, consider a large
105university server with various users - students, professors, system
106tasks etc. The resource planning for this server could be along the
107following lines:
108
109 CPU : Top cpuset
110 / \
111 CPUSet1 CPUSet2
112 | |
113 (Profs) (Students)
114
115 In addition (system tasks) are attached to topcpuset (so
116 that they can run anywhere) with a limit of 20%
117
118 Memory : Professors (50%), students (30%), system (20%)
119
120 Disk : Prof (50%), students (30%), system (20%)
121
122 Network : WWW browsing (20%), Network File System (60%), others (20%)
123 / \
124 Prof (15%) students (5%)
125
126Browsers like firefox/lynx go into the WWW network class, while (k)nfsd go
127into NFS network class.
128
129At the same time firefox/lynx will share an appropriate CPU/Memory class
130depending on who launched it (prof/student).
131
132With the ability to classify tasks differently for different resources
133(by putting those resource subsystems in different hierarchies) then
134the admin can easily set up a script which receives exec notifications
135and depending on who is launching the browser he can
136
137 # echo browser_pid > /mnt/<restype>/<userclass>/tasks
138
139With only a single hierarchy, he now would potentially have to create
140a separate cgroup for every browser launched and associate it with
141approp network and other resource class. This may lead to
142proliferation of such cgroups.
143
144Also lets say that the administrator would like to give enhanced network
145access temporarily to a student's browser (since it is night and the user
146wants to do online gaming :) OR give one of the students simulation
147apps enhanced CPU power,
148
149With ability to write pids directly to resource classes, its just a
150matter of :
151
152 # echo pid > /mnt/network/<new_class>/tasks
153 (after some time)
154 # echo pid > /mnt/network/<orig_class>/tasks
155
156Without this ability, he would have to split the cgroup into
157multiple separate ones and then associate the new cgroups with the
158new resource classes.
159
160
161
1621.3 How are cgroups implemented ?
163---------------------------------
164
165Control Groups extends the kernel as follows:
166
167 - Each task in the system has a reference-counted pointer to a
168 css_set.
169
170 - A css_set contains a set of reference-counted pointers to
171 cgroup_subsys_state objects, one for each cgroup subsystem
172 registered in the system. There is no direct link from a task to
173 the cgroup of which it's a member in each hierarchy, but this
174 can be determined by following pointers through the
175 cgroup_subsys_state objects. This is because accessing the
176 subsystem state is something that's expected to happen frequently
177 and in performance-critical code, whereas operations that require a
178 task's actual cgroup assignments (in particular, moving between
Paul Menage817929e2007-10-18 23:39:36 -0700179 cgroups) are less common. A linked list runs through the cg_list
180 field of each task_struct using the css_set, anchored at
181 css_set->tasks.
Paul Menageddbcc7e2007-10-18 23:39:30 -0700182
183 - A cgroup hierarchy filesystem can be mounted for browsing and
184 manipulation from user space.
185
186 - You can list all the tasks (by pid) attached to any cgroup.
187
188The implementation of cgroups requires a few, simple hooks
189into the rest of the kernel, none in performance critical paths:
190
191 - in init/main.c, to initialize the root cgroups and initial
192 css_set at system boot.
193
194 - in fork and exit, to attach and detach a task from its css_set.
195
196In addition a new file system, of type "cgroup" may be mounted, to
197enable browsing and modifying the cgroups presently known to the
198kernel. When mounting a cgroup hierarchy, you may specify a
199comma-separated list of subsystems to mount as the filesystem mount
200options. By default, mounting the cgroup filesystem attempts to
201mount a hierarchy containing all registered subsystems.
202
203If an active hierarchy with exactly the same set of subsystems already
204exists, it will be reused for the new mount. If no existing hierarchy
205matches, and any of the requested subsystems are in use in an existing
206hierarchy, the mount will fail with -EBUSY. Otherwise, a new hierarchy
207is activated, associated with the requested subsystems.
208
209It's not currently possible to bind a new subsystem to an active
210cgroup hierarchy, or to unbind a subsystem from an active cgroup
211hierarchy. This may be possible in future, but is fraught with nasty
212error-recovery issues.
213
214When a cgroup filesystem is unmounted, if there are any
215child cgroups created below the top-level cgroup, that hierarchy
216will remain active even though unmounted; if there are no
217child cgroups then the hierarchy will be deactivated.
218
219No new system calls are added for cgroups - all support for
220querying and modifying cgroups is via this cgroup file system.
221
222Each task under /proc has an added file named 'cgroup' displaying,
223for each active hierarchy, the subsystem names and the cgroup name
224as the path relative to the root of the cgroup file system.
225
226Each cgroup is represented by a directory in the cgroup file system
227containing the following files describing that cgroup:
228
229 - tasks: list of tasks (by pid) attached to that cgroup
230 - notify_on_release flag: run /sbin/cgroup_release_agent on exit?
231
232Other subsystems such as cpusets may add additional files in each
233cgroup dir
234
235New cgroups are created using the mkdir system call or shell
236command. The properties of a cgroup, such as its flags, are
237modified by writing to the appropriate file in that cgroups
238directory, as listed above.
239
240The named hierarchical structure of nested cgroups allows partitioning
241a large system into nested, dynamically changeable, "soft-partitions".
242
243The attachment of each task, automatically inherited at fork by any
244children of that task, to a cgroup allows organizing the work load
245on a system into related sets of tasks. A task may be re-attached to
246any other cgroup, if allowed by the permissions on the necessary
247cgroup file system directories.
248
249When a task is moved from one cgroup to another, it gets a new
250css_set pointer - if there's an already existing css_set with the
251desired collection of cgroups then that group is reused, else a new
252css_set is allocated. Note that the current implementation uses a
253linear search to locate an appropriate existing css_set, so isn't
254very efficient. A future version will use a hash table for better
255performance.
256
Paul Menage817929e2007-10-18 23:39:36 -0700257To allow access from a cgroup to the css_sets (and hence tasks)
258that comprise it, a set of cg_cgroup_link objects form a lattice;
259each cg_cgroup_link is linked into a list of cg_cgroup_links for
260a single cgroup on its cont_link_list field, and a list of
261cg_cgroup_links for a single css_set on its cg_link_list.
262
263Thus the set of tasks in a cgroup can be listed by iterating over
264each css_set that references the cgroup, and sub-iterating over
265each css_set's task set.
266
Paul Menageddbcc7e2007-10-18 23:39:30 -0700267The use of a Linux virtual file system (vfs) to represent the
268cgroup hierarchy provides for a familiar permission and name space
269for cgroups, with a minimum of additional kernel code.
270
2711.4 What does notify_on_release do ?
272------------------------------------
273
274*** notify_on_release is disabled in the current patch set. It will be
275*** reactivated in a future patch in a less-intrusive manner
276
277If the notify_on_release flag is enabled (1) in a cgroup, then
278whenever the last task in the cgroup leaves (exits or attaches to
279some other cgroup) and the last child cgroup of that cgroup
280is removed, then the kernel runs the command specified by the contents
281of the "release_agent" file in that hierarchy's root directory,
282supplying the pathname (relative to the mount point of the cgroup
283file system) of the abandoned cgroup. This enables automatic
284removal of abandoned cgroups. The default value of
285notify_on_release in the root cgroup at system boot is disabled
286(0). The default value of other cgroups at creation is the current
287value of their parents notify_on_release setting. The default value of
288a cgroup hierarchy's release_agent path is empty.
289
2901.5 How do I use cgroups ?
291--------------------------
292
293To start a new job that is to be contained within a cgroup, using
294the "cpuset" cgroup subsystem, the steps are something like:
295
296 1) mkdir /dev/cgroup
297 2) mount -t cgroup -ocpuset cpuset /dev/cgroup
298 3) Create the new cgroup by doing mkdir's and write's (or echo's) in
299 the /dev/cgroup virtual file system.
300 4) Start a task that will be the "founding father" of the new job.
301 5) Attach that task to the new cgroup by writing its pid to the
302 /dev/cgroup tasks file for that cgroup.
303 6) fork, exec or clone the job tasks from this founding father task.
304
305For example, the following sequence of commands will setup a cgroup
306named "Charlie", containing just CPUs 2 and 3, and Memory Node 1,
307and then start a subshell 'sh' in that cgroup:
308
309 mount -t cgroup cpuset -ocpuset /dev/cgroup
310 cd /dev/cgroup
311 mkdir Charlie
312 cd Charlie
313 /bin/echo 2-3 > cpus
314 /bin/echo 1 > mems
315 /bin/echo $$ > tasks
316 sh
317 # The subshell 'sh' is now running in cgroup Charlie
318 # The next line should display '/Charlie'
319 cat /proc/self/cgroup
320
3212. Usage Examples and Syntax
322============================
323
3242.1 Basic Usage
325---------------
326
327Creating, modifying, using the cgroups can be done through the cgroup
328virtual filesystem.
329
330To mount a cgroup hierarchy will all available subsystems, type:
331# mount -t cgroup xxx /dev/cgroup
332
333The "xxx" is not interpreted by the cgroup code, but will appear in
334/proc/mounts so may be any useful identifying string that you like.
335
336To mount a cgroup hierarchy with just the cpuset and numtasks
337subsystems, type:
338# mount -t cgroup -o cpuset,numtasks hier1 /dev/cgroup
339
340To change the set of subsystems bound to a mounted hierarchy, just
341remount with different options:
342
343# mount -o remount,cpuset,ns /dev/cgroup
344
345Note that changing the set of subsystems is currently only supported
346when the hierarchy consists of a single (root) cgroup. Supporting
347the ability to arbitrarily bind/unbind subsystems from an existing
348cgroup hierarchy is intended to be implemented in the future.
349
350Then under /dev/cgroup you can find a tree that corresponds to the
351tree of the cgroups in the system. For instance, /dev/cgroup
352is the cgroup that holds the whole system.
353
354If you want to create a new cgroup under /dev/cgroup:
355# cd /dev/cgroup
356# mkdir my_cgroup
357
358Now you want to do something with this cgroup.
359# cd my_cgroup
360
361In this directory you can find several files:
362# ls
363notify_on_release release_agent tasks
364(plus whatever files are added by the attached subsystems)
365
366Now attach your shell to this cgroup:
367# /bin/echo $$ > tasks
368
369You can also create cgroups inside your cgroup by using mkdir in this
370directory.
371# mkdir my_sub_cs
372
373To remove a cgroup, just use rmdir:
374# rmdir my_sub_cs
375
376This will fail if the cgroup is in use (has cgroups inside, or
377has processes attached, or is held alive by other subsystem-specific
378reference).
379
3802.2 Attaching processes
381-----------------------
382
383# /bin/echo PID > tasks
384
385Note that it is PID, not PIDs. You can only attach ONE task at a time.
386If you have several tasks to attach, you have to do it one after another:
387
388# /bin/echo PID1 > tasks
389# /bin/echo PID2 > tasks
390 ...
391# /bin/echo PIDn > tasks
392
3933. Kernel API
394=============
395
3963.1 Overview
397------------
398
399Each kernel subsystem that wants to hook into the generic cgroup
400system needs to create a cgroup_subsys object. This contains
401various methods, which are callbacks from the cgroup system, along
402with a subsystem id which will be assigned by the cgroup system.
403
404Other fields in the cgroup_subsys object include:
405
406- subsys_id: a unique array index for the subsystem, indicating which
407 entry in cgroup->subsys[] this subsystem should be
408 managing. Initialized by cgroup_register_subsys(); prior to this
409 it should be initialized to -1
410
411- hierarchy: an index indicating which hierarchy, if any, this
412 subsystem is currently attached to. If this is -1, then the
413 subsystem is not attached to any hierarchy, and all tasks should be
414 considered to be members of the subsystem's top_cgroup. It should
415 be initialized to -1.
416
417- name: should be initialized to a unique subsystem name prior to
418 calling cgroup_register_subsystem. Should be no longer than
419 MAX_CGROUP_TYPE_NAMELEN
420
421Each cgroup object created by the system has an array of pointers,
422indexed by subsystem id; this pointer is entirely managed by the
423subsystem; the generic cgroup code will never touch this pointer.
424
4253.2 Synchronization
426-------------------
427
428There is a global mutex, cgroup_mutex, used by the cgroup
429system. This should be taken by anything that wants to modify a
430cgroup. It may also be taken to prevent cgroups from being
431modified, but more specific locks may be more appropriate in that
432situation.
433
434See kernel/cgroup.c for more details.
435
436Subsystems can take/release the cgroup_mutex via the functions
437cgroup_lock()/cgroup_unlock(), and can
438take/release the callback_mutex via the functions
439cgroup_lock()/cgroup_unlock().
440
441Accessing a task's cgroup pointer may be done in the following ways:
442- while holding cgroup_mutex
443- while holding the task's alloc_lock (via task_lock())
444- inside an rcu_read_lock() section via rcu_dereference()
445
4463.3 Subsystem API
447--------------------------
448
449Each subsystem should:
450
451- add an entry in linux/cgroup_subsys.h
452- define a cgroup_subsys object called <name>_subsys
453
454Each subsystem may export the following methods. The only mandatory
455methods are create/destroy. Any others that are null are presumed to
456be successful no-ops.
457
458struct cgroup_subsys_state *create(struct cgroup *cont)
459LL=cgroup_mutex
460
461Called to create a subsystem state object for a cgroup. The
462subsystem should allocate its subsystem state object for the passed
463cgroup, returning a pointer to the new object on success or a
464negative error code. On success, the subsystem pointer should point to
465a structure of type cgroup_subsys_state (typically embedded in a
466larger subsystem-specific object), which will be initialized by the
467cgroup system. Note that this will be called at initialization to
468create the root subsystem state for this subsystem; this case can be
469identified by the passed cgroup object having a NULL parent (since
470it's the root of the hierarchy) and may be an appropriate place for
471initialization code.
472
473void destroy(struct cgroup *cont)
474LL=cgroup_mutex
475
476The cgroup system is about to destroy the passed cgroup; the
477subsystem should do any necessary cleanup
478
479int can_attach(struct cgroup_subsys *ss, struct cgroup *cont,
480 struct task_struct *task)
481LL=cgroup_mutex
482
483Called prior to moving a task into a cgroup; if the subsystem
484returns an error, this will abort the attach operation. If a NULL
485task is passed, then a successful result indicates that *any*
486unspecified task can be moved into the cgroup. Note that this isn't
487called on a fork. If this method returns 0 (success) then this should
488remain valid while the caller holds cgroup_mutex.
489
490void attach(struct cgroup_subsys *ss, struct cgroup *cont,
491 struct cgroup *old_cont, struct task_struct *task)
492LL=cgroup_mutex
493
494
495Called after the task has been attached to the cgroup, to allow any
496post-attachment activity that requires memory allocations or blocking.
497
498void fork(struct cgroup_subsy *ss, struct task_struct *task)
499LL=callback_mutex, maybe read_lock(tasklist_lock)
500
501Called when a task is forked into a cgroup. Also called during
502registration for all existing tasks.
503
504void exit(struct cgroup_subsys *ss, struct task_struct *task)
505LL=callback_mutex
506
507Called during task exit
508
509int populate(struct cgroup_subsys *ss, struct cgroup *cont)
510LL=none
511
512Called after creation of a cgroup to allow a subsystem to populate
513the cgroup directory with file entries. The subsystem should make
514calls to cgroup_add_file() with objects of type cftype (see
515include/linux/cgroup.h for details). Note that although this
516method can return an error code, the error code is currently not
517always handled well.
518
Paul Menage697f4162007-10-18 23:39:34 -0700519void post_clone(struct cgroup_subsys *ss, struct cgroup *cont)
520
521Called at the end of cgroup_clone() to do any paramater
522initialization which might be required before a task could attach. For
523example in cpusets, no task may attach before 'cpus' and 'mems' are set
524up.
525
Paul Menageddbcc7e2007-10-18 23:39:30 -0700526void bind(struct cgroup_subsys *ss, struct cgroup *root)
527LL=callback_mutex
528
529Called when a cgroup subsystem is rebound to a different hierarchy
530and root cgroup. Currently this will only involve movement between
531the default hierarchy (which never has sub-cgroups) and a hierarchy
532that is being created/destroyed (and hence has no sub-cgroups).
533
5344. Questions
535============
536
537Q: what's up with this '/bin/echo' ?
538A: bash's builtin 'echo' command does not check calls to write() against
539 errors. If you use it in the cgroup file system, you won't be
540 able to tell whether a command succeeded or failed.
541
542Q: When I attach processes, only the first of the line gets really attached !
543A: We can only return one error code per call to write(). So you should also
544 put only ONE pid.
545