Commit | Line | Data |
---|---|---|
99c8b231 MCC |
1 | ============== |
2 | Control Groups | |
3 | ============== | |
ddbcc7e8 | 4 | |
45ce80fb | 5 | Written by Paul Menage <menage@google.com> based on |
da82c92f | 6 | Documentation/admin-guide/cgroup-v1/cpusets.rst |
ddbcc7e8 PM |
7 | |
8 | Original copyright statements from cpusets.txt: | |
99c8b231 | 9 | |
ddbcc7e8 | 10 | Portions Copyright (C) 2004 BULL SA. |
99c8b231 | 11 | |
ddbcc7e8 | 12 | Portions Copyright (c) 2004-2006 Silicon Graphics, Inc. |
99c8b231 | 13 | |
ddbcc7e8 | 14 | Modified by Paul Jackson <pj@sgi.com> |
99c8b231 | 15 | |
93e205a7 | 16 | Modified by Christoph Lameter <cl@linux.com> |
ddbcc7e8 | 17 | |
99c8b231 MCC |
18 | .. CONTENTS: |
19 | ||
20 | 1. Control Groups | |
21 | 1.1 What are cgroups ? | |
22 | 1.2 Why are cgroups needed ? | |
23 | 1.3 How are cgroups implemented ? | |
24 | 1.4 What does notify_on_release do ? | |
25 | 1.5 What does clone_children do ? | |
26 | 1.6 How do I use cgroups ? | |
27 | 2. Usage Examples and Syntax | |
28 | 2.1 Basic Usage | |
29 | 2.2 Attaching processes | |
30 | 2.3 Mounting hierarchies by name | |
31 | 3. Kernel API | |
32 | 3.1 Overview | |
33 | 3.2 Synchronization | |
34 | 3.3 Subsystem API | |
35 | 4. Extended attributes usage | |
36 | 5. Questions | |
ddbcc7e8 PM |
37 | |
38 | 1. Control Groups | |
d19e0583 | 39 | ================= |
ddbcc7e8 PM |
40 | |
41 | 1.1 What are cgroups ? | |
42 | ---------------------- | |
43 | ||
44 | Control Groups provide a mechanism for aggregating/partitioning sets of | |
45 | tasks, and all their future children, into hierarchical groups with | |
46 | specialized behaviour. | |
47 | ||
48 | Definitions: | |
49 | ||
50 | A *cgroup* associates a set of tasks with a set of parameters for one | |
51 | or more subsystems. | |
52 | ||
53 | A *subsystem* is a module that makes use of the task grouping | |
54 | facilities provided by cgroups to treat groups of tasks in | |
55 | particular ways. A subsystem is typically a "resource controller" that | |
56 | schedules a resource or applies per-cgroup limits, but it may be | |
57 | anything that wants to act on a group of processes, e.g. a | |
58 | virtualization subsystem. | |
59 | ||
60 | A *hierarchy* is a set of cgroups arranged in a tree, such that | |
61 | every task in the system is in exactly one of the cgroups in the | |
62 | hierarchy, and a set of subsystems; each subsystem has system-specific | |
63 | state attached to each cgroup in the hierarchy. Each hierarchy has | |
64 | an instance of the cgroup virtual filesystem associated with it. | |
65 | ||
caa790ba | 66 | At any one time there may be multiple active hierarchies of task |
ddbcc7e8 PM |
67 | cgroups. Each hierarchy is a partition of all tasks in the system. |
68 | ||
83b061fc | 69 | User-level code may create and destroy cgroups by name in an |
ddbcc7e8 | 70 | instance of the cgroup virtual file system, specify and query to |
83b061fc | 71 | which cgroup a task is assigned, and list the task PIDs assigned to |
ddbcc7e8 PM |
72 | a cgroup. Those creations and assignments only affect the hierarchy |
73 | associated with that instance of the cgroup file system. | |
74 | ||
75 | On their own, the only use for cgroups is for simple job | |
76 | tracking. The intention is that other subsystems hook into the generic | |
77 | cgroup support to provide new attributes for cgroups, such as | |
78 | accounting/limiting the resources which processes in a cgroup can | |
da82c92f | 79 | access. For example, cpusets (see Documentation/admin-guide/cgroup-v1/cpusets.rst) allow |
ddbcc7e8 PM |
80 | you to associate a set of CPUs and a set of memory nodes with the |
81 | tasks in each cgroup. | |
82 | ||
83 | 1.2 Why are cgroups needed ? | |
84 | ---------------------------- | |
85 | ||
86 | There are multiple efforts to provide process aggregations in the | |
83b061fc | 87 | Linux kernel, mainly for resource-tracking purposes. Such efforts |
ddbcc7e8 PM |
88 | include cpusets, CKRM/ResGroups, UserBeanCounters, and virtual server |
89 | namespaces. These all require the basic notion of a | |
90 | grouping/partitioning of processes, with newly forked processes ending | |
83b061fc | 91 | up in the same group (cgroup) as their parent process. |
ddbcc7e8 PM |
92 | |
93 | The kernel cgroup patch provides the minimum essential kernel | |
94 | mechanisms required to efficiently implement such groups. It has | |
95 | minimal impact on the system fast paths, and provides hooks for | |
96 | specific subsystems such as cpusets to provide additional behaviour as | |
97 | desired. | |
98 | ||
99 | Multiple hierarchy support is provided to allow for situations where | |
100 | the division of tasks into cgroups is distinctly different for | |
101 | different subsystems - having parallel hierarchies allows each | |
102 | hierarchy to be a natural division of tasks, without having to handle | |
103 | complex combinations of tasks that would be present if several | |
104 | unrelated subsystems needed to be forced into the same tree of | |
105 | cgroups. | |
106 | ||
107 | At one extreme, each resource controller or subsystem could be in a | |
108 | separate hierarchy; at the other extreme, all subsystems | |
109 | would be attached to the same hierarchy. | |
110 | ||
111 | As an example of a scenario (originally proposed by vatsa@in.ibm.com) | |
112 | that can benefit from multiple hierarchies, consider a large | |
113 | university server with various users - students, professors, system | |
114 | tasks etc. The resource planning for this server could be along the | |
99c8b231 | 115 | following lines:: |
ddbcc7e8 | 116 | |
6ad85239 | 117 | CPU : "Top cpuset" |
ddbcc7e8 PM |
118 | / \ |
119 | CPUSet1 CPUSet2 | |
6ad85239 GL |
120 | | | |
121 | (Professors) (Students) | |
ddbcc7e8 PM |
122 | |
123 | In addition (system tasks) are attached to topcpuset (so | |
124 | that they can run anywhere) with a limit of 20% | |
125 | ||
6ad85239 | 126 | Memory : Professors (50%), Students (30%), system (20%) |
ddbcc7e8 | 127 | |
6ad85239 | 128 | Disk : Professors (50%), Students (30%), system (20%) |
ddbcc7e8 PM |
129 | |
130 | Network : WWW browsing (20%), Network File System (60%), others (20%) | |
131 | / \ | |
6ad85239 | 132 | Professors (15%) students (5%) |
ddbcc7e8 | 133 | |
83b061fc MK |
134 | Browsers like Firefox/Lynx go into the WWW network class, while (k)nfsd goes |
135 | into the NFS network class. | |
ddbcc7e8 | 136 | |
caa790ba | 137 | At the same time Firefox/Lynx will share an appropriate CPU/Memory class |
ddbcc7e8 PM |
138 | depending on who launched it (prof/student). |
139 | ||
140 | With the ability to classify tasks differently for different resources | |
83b061fc | 141 | (by putting those resource subsystems in different hierarchies), |
ddbcc7e8 | 142 | the admin can easily set up a script which receives exec notifications |
99c8b231 | 143 | and depending on who is launching the browser he can:: |
ddbcc7e8 | 144 | |
f6e07d38 | 145 | # echo browser_pid > /sys/fs/cgroup/<restype>/<userclass>/tasks |
ddbcc7e8 PM |
146 | |
147 | With only a single hierarchy, he now would potentially have to create | |
148 | a separate cgroup for every browser launched and associate it with | |
67de0162 | 149 | appropriate network and other resource class. This may lead to |
ddbcc7e8 PM |
150 | proliferation of such cgroups. |
151 | ||
83b061fc | 152 | Also let's say that the administrator would like to give enhanced network |
ddbcc7e8 | 153 | access temporarily to a student's browser (since it is night and the user |
83b061fc MK |
154 | wants to do online gaming :)) OR give one of the student's simulation |
155 | apps enhanced CPU power. | |
ddbcc7e8 | 156 | |
83b061fc | 157 | With ability to write PIDs directly to resource classes, it's just a |
99c8b231 | 158 | matter of:: |
ddbcc7e8 | 159 | |
f6e07d38 | 160 | # echo pid > /sys/fs/cgroup/network/<new_class>/tasks |
ddbcc7e8 | 161 | (after some time) |
f6e07d38 | 162 | # echo pid > /sys/fs/cgroup/network/<orig_class>/tasks |
ddbcc7e8 | 163 | |
83b061fc | 164 | Without this ability, the administrator would have to split the cgroup into |
ddbcc7e8 PM |
165 | multiple separate ones and then associate the new cgroups with the |
166 | new resource classes. | |
167 | ||
168 | ||
169 | ||
170 | 1.3 How are cgroups implemented ? | |
171 | --------------------------------- | |
172 | ||
173 | Control Groups extends the kernel as follows: | |
174 | ||
175 | - Each task in the system has a reference-counted pointer to a | |
176 | css_set. | |
177 | ||
178 | - A css_set contains a set of reference-counted pointers to | |
179 | cgroup_subsys_state objects, one for each cgroup subsystem | |
180 | registered in the system. There is no direct link from a task to | |
181 | the cgroup of which it's a member in each hierarchy, but this | |
182 | can be determined by following pointers through the | |
183 | cgroup_subsys_state objects. This is because accessing the | |
184 | subsystem state is something that's expected to happen frequently | |
185 | and in performance-critical code, whereas operations that require a | |
186 | task's actual cgroup assignments (in particular, moving between | |
817929ec PM |
187 | cgroups) are less common. A linked list runs through the cg_list |
188 | field of each task_struct using the css_set, anchored at | |
189 | css_set->tasks. | |
ddbcc7e8 | 190 | |
83b061fc | 191 | - A cgroup hierarchy filesystem can be mounted for browsing and |
ddbcc7e8 PM |
192 | manipulation from user space. |
193 | ||
83b061fc | 194 | - You can list all the tasks (by PID) attached to any cgroup. |
ddbcc7e8 PM |
195 | |
196 | The implementation of cgroups requires a few, simple hooks | |
83b061fc | 197 | into the rest of the kernel, none in performance-critical paths: |
ddbcc7e8 PM |
198 | |
199 | - in init/main.c, to initialize the root cgroups and initial | |
200 | css_set at system boot. | |
201 | ||
202 | - in fork and exit, to attach and detach a task from its css_set. | |
203 | ||
83b061fc | 204 | In addition, a new file system of type "cgroup" may be mounted, to |
ddbcc7e8 PM |
205 | enable browsing and modifying the cgroups presently known to the |
206 | kernel. When mounting a cgroup hierarchy, you may specify a | |
207 | comma-separated list of subsystems to mount as the filesystem mount | |
208 | options. By default, mounting the cgroup filesystem attempts to | |
209 | mount a hierarchy containing all registered subsystems. | |
210 | ||
211 | If an active hierarchy with exactly the same set of subsystems already | |
212 | exists, it will be reused for the new mount. If no existing hierarchy | |
213 | matches, and any of the requested subsystems are in use in an existing | |
214 | hierarchy, the mount will fail with -EBUSY. Otherwise, a new hierarchy | |
215 | is activated, associated with the requested subsystems. | |
216 | ||
26d5bbe5 TH |
217 | It's not currently possible to bind a new subsystem to an active |
218 | cgroup hierarchy, or to unbind a subsystem from an active cgroup | |
219 | hierarchy. This may be possible in future, but is fraught with nasty | |
220 | error-recovery issues. | |
ddbcc7e8 PM |
221 | |
222 | When a cgroup filesystem is unmounted, if there are any | |
223 | child cgroups created below the top-level cgroup, that hierarchy | |
224 | will remain active even though unmounted; if there are no | |
225 | child cgroups then the hierarchy will be deactivated. | |
226 | ||
227 | No new system calls are added for cgroups - all support for | |
228 | querying and modifying cgroups is via this cgroup file system. | |
229 | ||
230 | Each task under /proc has an added file named 'cgroup' displaying, | |
231 | for each active hierarchy, the subsystem names and the cgroup name | |
232 | as the path relative to the root of the cgroup file system. | |
233 | ||
234 | Each cgroup is represented by a directory in the cgroup file system | |
235 | containing the following files describing that cgroup: | |
236 | ||
83b061fc MK |
237 | - tasks: list of tasks (by PID) attached to that cgroup. This list |
238 | is not guaranteed to be sorted. Writing a thread ID into this file | |
7823da36 | 239 | moves the thread into this cgroup. |
83b061fc MK |
240 | - cgroup.procs: list of thread group IDs in the cgroup. This list is |
241 | not guaranteed to be sorted or free of duplicate TGIDs, and userspace | |
7823da36 | 242 | should sort/uniquify the list if this property is required. |
83b061fc | 243 | Writing a thread group ID into this file moves all threads in that |
74a1166d | 244 | group into this cgroup. |
d19e0583 LZ |
245 | - notify_on_release flag: run the release agent on exit? |
246 | - release_agent: the path to use for release notifications (this file | |
247 | exists in the top cgroup only) | |
ddbcc7e8 PM |
248 | |
249 | Other subsystems such as cpusets may add additional files in each | |
d19e0583 | 250 | cgroup dir. |
ddbcc7e8 PM |
251 | |
252 | New cgroups are created using the mkdir system call or shell | |
253 | command. The properties of a cgroup, such as its flags, are | |
254 | modified by writing to the appropriate file in that cgroups | |
255 | directory, as listed above. | |
256 | ||
257 | The named hierarchical structure of nested cgroups allows partitioning | |
258 | a large system into nested, dynamically changeable, "soft-partitions". | |
259 | ||
260 | The attachment of each task, automatically inherited at fork by any | |
261 | children of that task, to a cgroup allows organizing the work load | |
262 | on a system into related sets of tasks. A task may be re-attached to | |
263 | any other cgroup, if allowed by the permissions on the necessary | |
264 | cgroup file system directories. | |
265 | ||
266 | When a task is moved from one cgroup to another, it gets a new | |
267 | css_set pointer - if there's an already existing css_set with the | |
83b061fc | 268 | desired collection of cgroups then that group is reused, otherwise a new |
b851ee79 LZ |
269 | css_set is allocated. The appropriate existing css_set is located by |
270 | looking into a hash table. | |
ddbcc7e8 | 271 | |
817929ec PM |
272 | To allow access from a cgroup to the css_sets (and hence tasks) |
273 | that comprise it, a set of cg_cgroup_link objects form a lattice; | |
274 | each cg_cgroup_link is linked into a list of cg_cgroup_links for | |
d19e0583 | 275 | a single cgroup on its cgrp_link_list field, and a list of |
817929ec PM |
276 | cg_cgroup_links for a single css_set on its cg_link_list. |
277 | ||
278 | Thus the set of tasks in a cgroup can be listed by iterating over | |
279 | each css_set that references the cgroup, and sub-iterating over | |
280 | each css_set's task set. | |
281 | ||
ddbcc7e8 PM |
282 | The use of a Linux virtual file system (vfs) to represent the |
283 | cgroup hierarchy provides for a familiar permission and name space | |
284 | for cgroups, with a minimum of additional kernel code. | |
285 | ||
286 | 1.4 What does notify_on_release do ? | |
287 | ------------------------------------ | |
288 | ||
ddbcc7e8 PM |
289 | If the notify_on_release flag is enabled (1) in a cgroup, then |
290 | whenever the last task in the cgroup leaves (exits or attaches to | |
291 | some other cgroup) and the last child cgroup of that cgroup | |
292 | is removed, then the kernel runs the command specified by the contents | |
293 | of the "release_agent" file in that hierarchy's root directory, | |
294 | supplying the pathname (relative to the mount point of the cgroup | |
295 | file system) of the abandoned cgroup. This enables automatic | |
296 | removal of abandoned cgroups. The default value of | |
297 | notify_on_release in the root cgroup at system boot is disabled | |
298 | (0). The default value of other cgroups at creation is the current | |
83b061fc | 299 | value of their parents' notify_on_release settings. The default value of |
ddbcc7e8 PM |
300 | a cgroup hierarchy's release_agent path is empty. |
301 | ||
97978e6d DL |
302 | 1.5 What does clone_children do ? |
303 | --------------------------------- | |
304 | ||
2260e7fc TH |
305 | This flag only affects the cpuset controller. If the clone_children |
306 | flag is enabled (1) in a cgroup, a new cpuset cgroup will copy its | |
307 | configuration from the parent during initialization. | |
97978e6d DL |
308 | |
309 | 1.6 How do I use cgroups ? | |
ddbcc7e8 PM |
310 | -------------------------- |
311 | ||
312 | To start a new job that is to be contained within a cgroup, using | |
99c8b231 | 313 | the "cpuset" cgroup subsystem, the steps are something like:: |
ddbcc7e8 | 314 | |
f6e07d38 JS |
315 | 1) mount -t tmpfs cgroup_root /sys/fs/cgroup |
316 | 2) mkdir /sys/fs/cgroup/cpuset | |
317 | 3) mount -t cgroup -ocpuset cpuset /sys/fs/cgroup/cpuset | |
318 | 4) Create the new cgroup by doing mkdir's and write's (or echo's) in | |
845502d2 | 319 | the /sys/fs/cgroup/cpuset virtual file system. |
f6e07d38 | 320 | 5) Start a task that will be the "founding father" of the new job. |
83b061fc | 321 | 6) Attach that task to the new cgroup by writing its PID to the |
845502d2 | 322 | /sys/fs/cgroup/cpuset tasks file for that cgroup. |
f6e07d38 | 323 | 7) fork, exec or clone the job tasks from this founding father task. |
ddbcc7e8 PM |
324 | |
325 | For example, the following sequence of commands will setup a cgroup | |
326 | named "Charlie", containing just CPUs 2 and 3, and Memory Node 1, | |
99c8b231 | 327 | and then start a subshell 'sh' in that cgroup:: |
ddbcc7e8 | 328 | |
f6e07d38 JS |
329 | mount -t tmpfs cgroup_root /sys/fs/cgroup |
330 | mkdir /sys/fs/cgroup/cpuset | |
331 | mount -t cgroup cpuset -ocpuset /sys/fs/cgroup/cpuset | |
332 | cd /sys/fs/cgroup/cpuset | |
ddbcc7e8 PM |
333 | mkdir Charlie |
334 | cd Charlie | |
0f146a76 DG |
335 | /bin/echo 2-3 > cpuset.cpus |
336 | /bin/echo 1 > cpuset.mems | |
ddbcc7e8 PM |
337 | /bin/echo $$ > tasks |
338 | sh | |
339 | # The subshell 'sh' is now running in cgroup Charlie | |
340 | # The next line should display '/Charlie' | |
341 | cat /proc/self/cgroup | |
342 | ||
343 | 2. Usage Examples and Syntax | |
344 | ============================ | |
345 | ||
346 | 2.1 Basic Usage | |
347 | --------------- | |
348 | ||
83b061fc | 349 | Creating, modifying, using cgroups can be done through the cgroup |
ddbcc7e8 PM |
350 | virtual filesystem. |
351 | ||
99c8b231 MCC |
352 | To mount a cgroup hierarchy with all available subsystems, type:: |
353 | ||
354 | # mount -t cgroup xxx /sys/fs/cgroup | |
ddbcc7e8 PM |
355 | |
356 | The "xxx" is not interpreted by the cgroup code, but will appear in | |
357 | /proc/mounts so may be any useful identifying string that you like. | |
358 | ||
bb6405ea EM |
359 | Note: Some subsystems do not work without some user input first. For instance, |
360 | if cpusets are enabled the user will have to populate the cpus and mems files | |
361 | for each new cgroup created before that group can be used. | |
362 | ||
99c8b231 | 363 | As explained in section `1.2 Why are cgroups needed?` you should create |
f6e07d38 JS |
364 | different hierarchies of cgroups for each single resource or group of |
365 | resources you want to control. Therefore, you should mount a tmpfs on | |
366 | /sys/fs/cgroup and create directories for each cgroup resource or resource | |
99c8b231 | 367 | group:: |
f6e07d38 | 368 | |
99c8b231 MCC |
369 | # mount -t tmpfs cgroup_root /sys/fs/cgroup |
370 | # mkdir /sys/fs/cgroup/rg1 | |
f6e07d38 | 371 | |
595f4b69 | 372 | To mount a cgroup hierarchy with just the cpuset and memory |
99c8b231 MCC |
373 | subsystems, type:: |
374 | ||
375 | # mount -t cgroup -o cpuset,memory hier1 /sys/fs/cgroup/rg1 | |
ddbcc7e8 | 376 | |
9a8054aa DW |
377 | While remounting cgroups is currently supported, it is not recommend |
378 | to use it. Remounting allows changing bound subsystems and | |
379 | release_agent. Rebinding is hardly useful as it only works when the | |
380 | hierarchy is empty and release_agent itself should be replaced with | |
381 | conventional fsnotify. The support for remounting will be removed in | |
382 | the future. | |
b6719ec1 | 383 | |
99c8b231 MCC |
384 | To Specify a hierarchy's release_agent:: |
385 | ||
386 | # mount -t cgroup -o cpuset,release_agent="/sbin/cpuset_release_agent" \ | |
387 | xxx /sys/fs/cgroup/rg1 | |
b6719ec1 LZ |
388 | |
389 | Note that specifying 'release_agent' more than once will return failure. | |
ddbcc7e8 | 390 | |
26d5bbe5 TH |
391 | Note that changing the set of subsystems is currently only supported |
392 | when the hierarchy consists of a single (root) cgroup. Supporting | |
393 | the ability to arbitrarily bind/unbind subsystems from an existing | |
394 | cgroup hierarchy is intended to be implemented in the future. | |
ddbcc7e8 | 395 | |
f6e07d38 JS |
396 | Then under /sys/fs/cgroup/rg1 you can find a tree that corresponds to the |
397 | tree of the cgroups in the system. For instance, /sys/fs/cgroup/rg1 | |
ddbcc7e8 PM |
398 | is the cgroup that holds the whole system. |
399 | ||
99c8b231 MCC |
400 | If you want to change the value of release_agent:: |
401 | ||
402 | # echo "/sbin/new_release_agent" > /sys/fs/cgroup/rg1/release_agent | |
b6719ec1 LZ |
403 | |
404 | It can also be changed via remount. | |
405 | ||
99c8b231 MCC |
406 | If you want to create a new cgroup under /sys/fs/cgroup/rg1:: |
407 | ||
408 | # cd /sys/fs/cgroup/rg1 | |
409 | # mkdir my_cgroup | |
410 | ||
411 | Now you want to do something with this cgroup: | |
412 | ||
413 | # cd my_cgroup | |
ddbcc7e8 | 414 | |
99c8b231 | 415 | In this directory you can find several files:: |
ddbcc7e8 | 416 | |
99c8b231 MCC |
417 | # ls |
418 | cgroup.procs notify_on_release tasks | |
419 | (plus whatever files added by the attached subsystems) | |
ddbcc7e8 | 420 | |
99c8b231 MCC |
421 | Now attach your shell to this cgroup:: |
422 | ||
423 | # /bin/echo $$ > tasks | |
ddbcc7e8 PM |
424 | |
425 | You can also create cgroups inside your cgroup by using mkdir in this | |
99c8b231 MCC |
426 | directory:: |
427 | ||
428 | # mkdir my_sub_cs | |
429 | ||
430 | To remove a cgroup, just use rmdir:: | |
ddbcc7e8 | 431 | |
99c8b231 | 432 | # rmdir my_sub_cs |
ddbcc7e8 PM |
433 | |
434 | This will fail if the cgroup is in use (has cgroups inside, or | |
435 | has processes attached, or is held alive by other subsystem-specific | |
436 | reference). | |
437 | ||
438 | 2.2 Attaching processes | |
439 | ----------------------- | |
440 | ||
99c8b231 MCC |
441 | :: |
442 | ||
443 | # /bin/echo PID > tasks | |
ddbcc7e8 PM |
444 | |
445 | Note that it is PID, not PIDs. You can only attach ONE task at a time. | |
99c8b231 | 446 | If you have several tasks to attach, you have to do it one after another:: |
ddbcc7e8 | 447 | |
99c8b231 MCC |
448 | # /bin/echo PID1 > tasks |
449 | # /bin/echo PID2 > tasks | |
450 | ... | |
451 | # /bin/echo PIDn > tasks | |
ddbcc7e8 | 452 | |
99c8b231 | 453 | You can attach the current shell task by echoing 0:: |
bef67c5a | 454 | |
99c8b231 | 455 | # echo 0 > tasks |
bef67c5a | 456 | |
74a1166d | 457 | You can use the cgroup.procs file instead of the tasks file to move all |
83b061fc | 458 | threads in a threadgroup at once. Echoing the PID of any task in a |
74a1166d | 459 | threadgroup to cgroup.procs causes all tasks in that threadgroup to be |
1ae65ae9 | 460 | attached to the cgroup. Writing 0 to cgroup.procs moves all tasks |
74a1166d BB |
461 | in the writing task's threadgroup. |
462 | ||
bb6405ea EM |
463 | Note: Since every task is always a member of exactly one cgroup in each |
464 | mounted hierarchy, to remove a task from its current cgroup you must | |
465 | move it into a new cgroup (possibly the root cgroup) by writing to the | |
466 | new cgroup's tasks file. | |
467 | ||
5fe69d7e LZ |
468 | Note: Due to some restrictions enforced by some cgroup subsystems, moving |
469 | a process to another cgroup can fail. | |
bb6405ea | 470 | |
c6d57f33 PM |
471 | 2.3 Mounting hierarchies by name |
472 | -------------------------------- | |
473 | ||
474 | Passing the name=<x> option when mounting a cgroups hierarchy | |
475 | associates the given name with the hierarchy. This can be used when | |
476 | mounting a pre-existing hierarchy, in order to refer to it by name | |
477 | rather than by its set of active subsystems. Each hierarchy is either | |
478 | nameless, or has a unique name. | |
479 | ||
480 | The name should match [\w.-]+ | |
481 | ||
482 | When passing a name=<x> option for a new hierarchy, you need to | |
483 | specify subsystems manually; the legacy behaviour of mounting all | |
484 | subsystems when none are explicitly specified is not supported when | |
485 | you give a subsystem a name. | |
486 | ||
487 | The name of the subsystem appears as part of the hierarchy description | |
488 | in /proc/mounts and /proc/<pid>/cgroups. | |
489 | ||
490 | ||
ddbcc7e8 PM |
491 | 3. Kernel API |
492 | ============= | |
493 | ||
494 | 3.1 Overview | |
495 | ------------ | |
496 | ||
497 | Each kernel subsystem that wants to hook into the generic cgroup | |
498 | system needs to create a cgroup_subsys object. This contains | |
499 | various methods, which are callbacks from the cgroup system, along | |
83b061fc | 500 | with a subsystem ID which will be assigned by the cgroup system. |
ddbcc7e8 PM |
501 | |
502 | Other fields in the cgroup_subsys object include: | |
503 | ||
504 | - subsys_id: a unique array index for the subsystem, indicating which | |
d19e0583 | 505 | entry in cgroup->subsys[] this subsystem should be managing. |
ddbcc7e8 | 506 | |
d19e0583 LZ |
507 | - name: should be initialized to a unique subsystem name. Should be |
508 | no longer than MAX_CGROUP_TYPE_NAMELEN. | |
ddbcc7e8 | 509 | |
d19e0583 LZ |
510 | - early_init: indicate if the subsystem needs early initialization |
511 | at system boot. | |
ddbcc7e8 PM |
512 | |
513 | Each cgroup object created by the system has an array of pointers, | |
83b061fc | 514 | indexed by subsystem ID; this pointer is entirely managed by the |
ddbcc7e8 PM |
515 | subsystem; the generic cgroup code will never touch this pointer. |
516 | ||
517 | 3.2 Synchronization | |
518 | ------------------- | |
519 | ||
520 | There is a global mutex, cgroup_mutex, used by the cgroup | |
521 | system. This should be taken by anything that wants to modify a | |
522 | cgroup. It may also be taken to prevent cgroups from being | |
523 | modified, but more specific locks may be more appropriate in that | |
524 | situation. | |
525 | ||
526 | See kernel/cgroup.c for more details. | |
527 | ||
528 | Subsystems can take/release the cgroup_mutex via the functions | |
ddbcc7e8 PM |
529 | cgroup_lock()/cgroup_unlock(). |
530 | ||
531 | Accessing a task's cgroup pointer may be done in the following ways: | |
532 | - while holding cgroup_mutex | |
533 | - while holding the task's alloc_lock (via task_lock()) | |
534 | - inside an rcu_read_lock() section via rcu_dereference() | |
535 | ||
536 | 3.3 Subsystem API | |
d19e0583 | 537 | ----------------- |
ddbcc7e8 PM |
538 | |
539 | Each subsystem should: | |
540 | ||
541 | - add an entry in linux/cgroup_subsys.h | |
d51b9dae | 542 | - define a cgroup_subsys object called <name>_cgrp_subsys |
e6a1105b | 543 | |
ddbcc7e8 | 544 | Each subsystem may export the following methods. The only mandatory |
92fb9748 | 545 | methods are css_alloc/free. Any others that are null are presumed to |
ddbcc7e8 PM |
546 | be successful no-ops. |
547 | ||
99c8b231 | 548 | ``struct cgroup_subsys_state *css_alloc(struct cgroup *cgrp)`` |
8dc4f3e1 | 549 | (cgroup_mutex held by caller) |
ddbcc7e8 | 550 | |
92fb9748 | 551 | Called to allocate a subsystem state object for a cgroup. The |
ddbcc7e8 PM |
552 | subsystem should allocate its subsystem state object for the passed |
553 | cgroup, returning a pointer to the new object on success or a | |
92fb9748 | 554 | ERR_PTR() value. On success, the subsystem pointer should point to |
ddbcc7e8 PM |
555 | a structure of type cgroup_subsys_state (typically embedded in a |
556 | larger subsystem-specific object), which will be initialized by the | |
557 | cgroup system. Note that this will be called at initialization to | |
558 | create the root subsystem state for this subsystem; this case can be | |
559 | identified by the passed cgroup object having a NULL parent (since | |
560 | it's the root of the hierarchy) and may be an appropriate place for | |
561 | initialization code. | |
562 | ||
99c8b231 | 563 | ``int css_online(struct cgroup *cgrp)`` |
8dc4f3e1 | 564 | (cgroup_mutex held by caller) |
ddbcc7e8 | 565 | |
92fb9748 TH |
566 | Called after @cgrp successfully completed all allocations and made |
567 | visible to cgroup_for_each_child/descendant_*() iterators. The | |
568 | subsystem may choose to fail creation by returning -errno. This | |
569 | callback can be used to implement reliable state sharing and | |
570 | propagation along the hierarchy. See the comment on | |
571 | cgroup_for_each_descendant_pre() for details. | |
572 | ||
99c8b231 | 573 | ``void css_offline(struct cgroup *cgrp);`` |
d7eeac19 | 574 | (cgroup_mutex held by caller) |
92fb9748 TH |
575 | |
576 | This is the counterpart of css_online() and called iff css_online() | |
577 | has succeeded on @cgrp. This signifies the beginning of the end of | |
578 | @cgrp. @cgrp is being removed and the subsystem should start dropping | |
579 | all references it's holding on @cgrp. When all references are dropped, | |
580 | cgroup removal will proceed to the next step - css_free(). After this | |
581 | callback, @cgrp should be considered dead to the subsystem. | |
582 | ||
99c8b231 | 583 | ``void css_free(struct cgroup *cgrp)`` |
92fb9748 TH |
584 | (cgroup_mutex held by caller) |
585 | ||
586 | The cgroup system is about to free @cgrp; the subsystem should free | |
587 | its subsystem state object. By the time this method is called, @cgrp | |
588 | is completely unused; @cgrp->parent is still valid. (Note - can also | |
589 | be called for a newly-created cgroup if an error occurs after this | |
590 | subsystem's create() method has been called for the new cgroup). | |
d19e0583 | 591 | |
99c8b231 | 592 | ``int can_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)`` |
8dc4f3e1 | 593 | (cgroup_mutex held by caller) |
ddbcc7e8 | 594 | |
2f7ee569 TH |
595 | Called prior to moving one or more tasks into a cgroup; if the |
596 | subsystem returns an error, this will abort the attach operation. | |
597 | @tset contains the tasks to be attached and is guaranteed to have at | |
598 | least one task in it. | |
599 | ||
600 | If there are multiple tasks in the taskset, then: | |
601 | - it's guaranteed that all are from the same thread group | |
602 | - @tset contains all tasks from the thread group whether or not | |
603 | they're switching cgroups | |
604 | - the first task is the leader | |
605 | ||
606 | Each @tset entry also contains the task's old cgroup and tasks which | |
607 | aren't switching cgroup can be skipped easily using the | |
608 | cgroup_taskset_for_each() iterator. Note that this isn't called on a | |
609 | fork. If this method returns 0 (success) then this should remain valid | |
610 | while the caller holds cgroup_mutex and it is ensured that either | |
f780bdb7 BB |
611 | attach() or cancel_attach() will be called in future. |
612 | ||
99c8b231 | 613 | ``void css_reset(struct cgroup_subsys_state *css)`` |
b4536f0c TH |
614 | (cgroup_mutex held by caller) |
615 | ||
616 | An optional operation which should restore @css's configuration to the | |
617 | initial state. This is currently only used on the unified hierarchy | |
618 | when a subsystem is disabled on a cgroup through | |
619 | "cgroup.subtree_control" but should remain enabled because other | |
620 | subsystems depend on it. cgroup core makes such a css invisible by | |
621 | removing the associated interface files and invokes this callback so | |
622 | that the hidden subsystem can return to the initial neutral state. | |
623 | This prevents unexpected resource control from a hidden css and | |
624 | ensures that the configuration is in the initial state when it is made | |
625 | visible again later. | |
626 | ||
99c8b231 | 627 | ``void cancel_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)`` |
2468c723 DN |
628 | (cgroup_mutex held by caller) |
629 | ||
630 | Called when a task attach operation has failed after can_attach() has succeeded. | |
631 | A subsystem whose can_attach() has some side-effects should provide this | |
88393161 | 632 | function, so that the subsystem can implement a rollback. If not, not necessary. |
2468c723 | 633 | This will be called only about subsystems whose can_attach() operation have |
2f7ee569 | 634 | succeeded. The parameters are identical to can_attach(). |
2468c723 | 635 | |
99c8b231 | 636 | ``void attach(struct cgroup *cgrp, struct cgroup_taskset *tset)`` |
18e7f1f0 | 637 | (cgroup_mutex held by caller) |
ddbcc7e8 PM |
638 | |
639 | Called after the task has been attached to the cgroup, to allow any | |
640 | post-attachment activity that requires memory allocations or blocking. | |
2f7ee569 | 641 | The parameters are identical to can_attach(). |
f780bdb7 | 642 | |
99c8b231 | 643 | ``void fork(struct task_struct *task)`` |
ddbcc7e8 | 644 | |
e8d55fde | 645 | Called when a task is forked into a cgroup. |
ddbcc7e8 | 646 | |
99c8b231 | 647 | ``void exit(struct task_struct *task)`` |
ddbcc7e8 | 648 | |
d19e0583 | 649 | Called during task exit. |
ddbcc7e8 | 650 | |
99c8b231 | 651 | ``void free(struct task_struct *task)`` |
afcf6c8b TH |
652 | |
653 | Called when the task_struct is freed. | |
654 | ||
99c8b231 | 655 | ``void bind(struct cgroup *root)`` |
26d5bbe5 TH |
656 | (cgroup_mutex held by caller) |
657 | ||
658 | Called when a cgroup subsystem is rebound to a different hierarchy | |
659 | and root cgroup. Currently this will only involve movement between | |
660 | the default hierarchy (which never has sub-cgroups) and a hierarchy | |
661 | that is being created/destroyed (and hence has no sub-cgroups). | |
ddbcc7e8 | 662 | |
19ec2567 AR |
663 | 4. Extended attribute usage |
664 | =========================== | |
665 | ||
666 | cgroup filesystem supports certain types of extended attributes in its | |
667 | directories and files. The current supported types are: | |
99c8b231 | 668 | |
19ec2567 AR |
669 | - Trusted (XATTR_TRUSTED) |
670 | - Security (XATTR_SECURITY) | |
671 | ||
672 | Both require CAP_SYS_ADMIN capability to set. | |
673 | ||
674 | Like in tmpfs, the extended attributes in cgroup filesystem are stored | |
675 | using kernel memory and it's advised to keep the usage at minimum. This | |
676 | is the reason why user defined extended attributes are not supported, since | |
677 | any user can do it and there's no limit in the value size. | |
678 | ||
679 | The current known users for this feature are SELinux to limit cgroup usage | |
680 | in containers and systemd for assorted meta data like main PID in a cgroup | |
681 | (systemd creates a cgroup per service). | |
682 | ||
683 | 5. Questions | |
ddbcc7e8 PM |
684 | ============ |
685 | ||
99c8b231 | 686 | :: |
ddbcc7e8 | 687 | |
99c8b231 MCC |
688 | Q: what's up with this '/bin/echo' ? |
689 | A: bash's builtin 'echo' command does not check calls to write() against | |
690 | errors. If you use it in the cgroup file system, you won't be | |
691 | able to tell whether a command succeeded or failed. | |
ddbcc7e8 | 692 | |
99c8b231 MCC |
693 | Q: When I attach processes, only the first of the line gets really attached ! |
694 | A: We can only return one error code per call to write(). So you should also | |
695 | put only ONE PID. |