]>
Commit | Line | Data |
---|---|---|
ddbcc7e8 PM |
1 | CGROUPS |
2 | ------- | |
3 | ||
45ce80fb LZ |
4 | Written by Paul Menage <menage@google.com> based on |
5 | Documentation/cgroups/cpusets.txt | |
ddbcc7e8 PM |
6 | |
7 | Original copyright statements from cpusets.txt: | |
8 | Portions Copyright (C) 2004 BULL SA. | |
9 | Portions Copyright (c) 2004-2006 Silicon Graphics, Inc. | |
10 | Modified by Paul Jackson <pj@sgi.com> | |
11 | Modified by Christoph Lameter <clameter@sgi.com> | |
12 | ||
13 | CONTENTS: | |
14 | ========= | |
15 | ||
16 | 1. Control Groups | |
17 | 1.1 What are cgroups ? | |
18 | 1.2 Why are cgroups needed ? | |
19 | 1.3 How are cgroups implemented ? | |
20 | 1.4 What does notify_on_release do ? | |
21 | 1.5 How do I use cgroups ? | |
22 | 2. Usage Examples and Syntax | |
23 | 2.1 Basic Usage | |
24 | 2.2 Attaching processes | |
8ca712ea | 25 | 2.3 Mounting hierarchies by name |
0dea1168 | 26 | 2.4 Notification API |
ddbcc7e8 PM |
27 | 3. Kernel API |
28 | 3.1 Overview | |
29 | 3.2 Synchronization | |
30 | 3.3 Subsystem API | |
31 | 4. Questions | |
32 | ||
33 | 1. Control Groups | |
d19e0583 | 34 | ================= |
ddbcc7e8 PM |
35 | |
36 | 1.1 What are cgroups ? | |
37 | ---------------------- | |
38 | ||
39 | Control Groups provide a mechanism for aggregating/partitioning sets of | |
40 | tasks, and all their future children, into hierarchical groups with | |
41 | specialized behaviour. | |
42 | ||
43 | Definitions: | |
44 | ||
45 | A *cgroup* associates a set of tasks with a set of parameters for one | |
46 | or more subsystems. | |
47 | ||
48 | A *subsystem* is a module that makes use of the task grouping | |
49 | facilities provided by cgroups to treat groups of tasks in | |
50 | particular ways. A subsystem is typically a "resource controller" that | |
51 | schedules a resource or applies per-cgroup limits, but it may be | |
52 | anything that wants to act on a group of processes, e.g. a | |
53 | virtualization subsystem. | |
54 | ||
55 | A *hierarchy* is a set of cgroups arranged in a tree, such that | |
56 | every task in the system is in exactly one of the cgroups in the | |
57 | hierarchy, and a set of subsystems; each subsystem has system-specific | |
58 | state attached to each cgroup in the hierarchy. Each hierarchy has | |
59 | an instance of the cgroup virtual filesystem associated with it. | |
60 | ||
caa790ba | 61 | At any one time there may be multiple active hierarchies of task |
ddbcc7e8 PM |
62 | cgroups. Each hierarchy is a partition of all tasks in the system. |
63 | ||
64 | User level code may create and destroy cgroups by name in an | |
65 | instance of the cgroup virtual file system, specify and query to | |
66 | which cgroup a task is assigned, and list the task pids assigned to | |
67 | a cgroup. Those creations and assignments only affect the hierarchy | |
68 | associated with that instance of the cgroup file system. | |
69 | ||
70 | On their own, the only use for cgroups is for simple job | |
71 | tracking. The intention is that other subsystems hook into the generic | |
72 | cgroup support to provide new attributes for cgroups, such as | |
73 | accounting/limiting the resources which processes in a cgroup can | |
45ce80fb | 74 | access. For example, cpusets (see Documentation/cgroups/cpusets.txt) allows |
ddbcc7e8 PM |
75 | you to associate a set of CPUs and a set of memory nodes with the |
76 | tasks in each cgroup. | |
77 | ||
78 | 1.2 Why are cgroups needed ? | |
79 | ---------------------------- | |
80 | ||
81 | There are multiple efforts to provide process aggregations in the | |
82 | Linux kernel, mainly for resource tracking purposes. Such efforts | |
83 | include cpusets, CKRM/ResGroups, UserBeanCounters, and virtual server | |
84 | namespaces. These all require the basic notion of a | |
85 | grouping/partitioning of processes, with newly forked processes ending | |
86 | in the same group (cgroup) as their parent process. | |
87 | ||
88 | The kernel cgroup patch provides the minimum essential kernel | |
89 | mechanisms required to efficiently implement such groups. It has | |
90 | minimal impact on the system fast paths, and provides hooks for | |
91 | specific subsystems such as cpusets to provide additional behaviour as | |
92 | desired. | |
93 | ||
94 | Multiple hierarchy support is provided to allow for situations where | |
95 | the division of tasks into cgroups is distinctly different for | |
96 | different subsystems - having parallel hierarchies allows each | |
97 | hierarchy to be a natural division of tasks, without having to handle | |
98 | complex combinations of tasks that would be present if several | |
99 | unrelated subsystems needed to be forced into the same tree of | |
100 | cgroups. | |
101 | ||
102 | At one extreme, each resource controller or subsystem could be in a | |
103 | separate hierarchy; at the other extreme, all subsystems | |
104 | would be attached to the same hierarchy. | |
105 | ||
106 | As an example of a scenario (originally proposed by vatsa@in.ibm.com) | |
107 | that can benefit from multiple hierarchies, consider a large | |
108 | university server with various users - students, professors, system | |
109 | tasks etc. The resource planning for this server could be along the | |
110 | following lines: | |
111 | ||
112 | CPU : Top cpuset | |
113 | / \ | |
114 | CPUSet1 CPUSet2 | |
115 | | | | |
116 | (Profs) (Students) | |
117 | ||
118 | In addition (system tasks) are attached to topcpuset (so | |
119 | that they can run anywhere) with a limit of 20% | |
120 | ||
121 | Memory : Professors (50%), students (30%), system (20%) | |
122 | ||
123 | Disk : Prof (50%), students (30%), system (20%) | |
124 | ||
125 | Network : WWW browsing (20%), Network File System (60%), others (20%) | |
126 | / \ | |
127 | Prof (15%) students (5%) | |
128 | ||
caa790ba | 129 | Browsers like Firefox/Lynx go into the WWW network class, while (k)nfsd go |
ddbcc7e8 PM |
130 | into NFS network class. |
131 | ||
caa790ba | 132 | At the same time Firefox/Lynx will share an appropriate CPU/Memory class |
ddbcc7e8 PM |
133 | depending on who launched it (prof/student). |
134 | ||
135 | With the ability to classify tasks differently for different resources | |
136 | (by putting those resource subsystems in different hierarchies) then | |
137 | the admin can easily set up a script which receives exec notifications | |
138 | and depending on who is launching the browser he can | |
139 | ||
140 | # echo browser_pid > /mnt/<restype>/<userclass>/tasks | |
141 | ||
142 | With only a single hierarchy, he now would potentially have to create | |
143 | a separate cgroup for every browser launched and associate it with | |
144 | approp network and other resource class. This may lead to | |
145 | proliferation of such cgroups. | |
146 | ||
147 | Also lets say that the administrator would like to give enhanced network | |
148 | access temporarily to a student's browser (since it is night and the user | |
d19e0583 | 149 | wants to do online gaming :)) OR give one of the students simulation |
ddbcc7e8 PM |
150 | apps enhanced CPU power, |
151 | ||
d19e0583 | 152 | With ability to write pids directly to resource classes, it's just a |
ddbcc7e8 PM |
153 | matter of : |
154 | ||
155 | # echo pid > /mnt/network/<new_class>/tasks | |
156 | (after some time) | |
157 | # echo pid > /mnt/network/<orig_class>/tasks | |
158 | ||
159 | Without this ability, he would have to split the cgroup into | |
160 | multiple separate ones and then associate the new cgroups with the | |
161 | new resource classes. | |
162 | ||
163 | ||
164 | ||
165 | 1.3 How are cgroups implemented ? | |
166 | --------------------------------- | |
167 | ||
168 | Control Groups extends the kernel as follows: | |
169 | ||
170 | - Each task in the system has a reference-counted pointer to a | |
171 | css_set. | |
172 | ||
173 | - A css_set contains a set of reference-counted pointers to | |
174 | cgroup_subsys_state objects, one for each cgroup subsystem | |
175 | registered in the system. There is no direct link from a task to | |
176 | the cgroup of which it's a member in each hierarchy, but this | |
177 | can be determined by following pointers through the | |
178 | cgroup_subsys_state objects. This is because accessing the | |
179 | subsystem state is something that's expected to happen frequently | |
180 | and in performance-critical code, whereas operations that require a | |
181 | task's actual cgroup assignments (in particular, moving between | |
817929ec PM |
182 | cgroups) are less common. A linked list runs through the cg_list |
183 | field of each task_struct using the css_set, anchored at | |
184 | css_set->tasks. | |
ddbcc7e8 PM |
185 | |
186 | - A cgroup hierarchy filesystem can be mounted for browsing and | |
187 | manipulation from user space. | |
188 | ||
189 | - You can list all the tasks (by pid) attached to any cgroup. | |
190 | ||
191 | The implementation of cgroups requires a few, simple hooks | |
192 | into the rest of the kernel, none in performance critical paths: | |
193 | ||
194 | - in init/main.c, to initialize the root cgroups and initial | |
195 | css_set at system boot. | |
196 | ||
197 | - in fork and exit, to attach and detach a task from its css_set. | |
198 | ||
199 | In addition a new file system, of type "cgroup" may be mounted, to | |
200 | enable browsing and modifying the cgroups presently known to the | |
201 | kernel. When mounting a cgroup hierarchy, you may specify a | |
202 | comma-separated list of subsystems to mount as the filesystem mount | |
203 | options. By default, mounting the cgroup filesystem attempts to | |
204 | mount a hierarchy containing all registered subsystems. | |
205 | ||
206 | If an active hierarchy with exactly the same set of subsystems already | |
207 | exists, it will be reused for the new mount. If no existing hierarchy | |
208 | matches, and any of the requested subsystems are in use in an existing | |
209 | hierarchy, the mount will fail with -EBUSY. Otherwise, a new hierarchy | |
210 | is activated, associated with the requested subsystems. | |
211 | ||
212 | It's not currently possible to bind a new subsystem to an active | |
213 | cgroup hierarchy, or to unbind a subsystem from an active cgroup | |
214 | hierarchy. This may be possible in future, but is fraught with nasty | |
215 | error-recovery issues. | |
216 | ||
217 | When a cgroup filesystem is unmounted, if there are any | |
218 | child cgroups created below the top-level cgroup, that hierarchy | |
219 | will remain active even though unmounted; if there are no | |
220 | child cgroups then the hierarchy will be deactivated. | |
221 | ||
222 | No new system calls are added for cgroups - all support for | |
223 | querying and modifying cgroups is via this cgroup file system. | |
224 | ||
225 | Each task under /proc has an added file named 'cgroup' displaying, | |
226 | for each active hierarchy, the subsystem names and the cgroup name | |
227 | as the path relative to the root of the cgroup file system. | |
228 | ||
229 | Each cgroup is represented by a directory in the cgroup file system | |
230 | containing the following files describing that cgroup: | |
231 | ||
7823da36 PM |
232 | - tasks: list of tasks (by pid) attached to that cgroup. This list |
233 | is not guaranteed to be sorted. Writing a thread id into this file | |
234 | moves the thread into this cgroup. | |
235 | - cgroup.procs: list of tgids in the cgroup. This list is not | |
236 | guaranteed to be sorted or free of duplicate tgids, and userspace | |
237 | should sort/uniquify the list if this property is required. | |
7716fa66 | 238 | This is a read-only file, for now. |
d19e0583 LZ |
239 | - notify_on_release flag: run the release agent on exit? |
240 | - release_agent: the path to use for release notifications (this file | |
241 | exists in the top cgroup only) | |
ddbcc7e8 PM |
242 | |
243 | Other subsystems such as cpusets may add additional files in each | |
d19e0583 | 244 | cgroup dir. |
ddbcc7e8 PM |
245 | |
246 | New cgroups are created using the mkdir system call or shell | |
247 | command. The properties of a cgroup, such as its flags, are | |
248 | modified by writing to the appropriate file in that cgroups | |
249 | directory, as listed above. | |
250 | ||
251 | The named hierarchical structure of nested cgroups allows partitioning | |
252 | a large system into nested, dynamically changeable, "soft-partitions". | |
253 | ||
254 | The attachment of each task, automatically inherited at fork by any | |
255 | children of that task, to a cgroup allows organizing the work load | |
256 | on a system into related sets of tasks. A task may be re-attached to | |
257 | any other cgroup, if allowed by the permissions on the necessary | |
258 | cgroup file system directories. | |
259 | ||
260 | When a task is moved from one cgroup to another, it gets a new | |
261 | css_set pointer - if there's an already existing css_set with the | |
262 | desired collection of cgroups then that group is reused, else a new | |
b851ee79 LZ |
263 | css_set is allocated. The appropriate existing css_set is located by |
264 | looking into a hash table. | |
ddbcc7e8 | 265 | |
817929ec PM |
266 | To allow access from a cgroup to the css_sets (and hence tasks) |
267 | that comprise it, a set of cg_cgroup_link objects form a lattice; | |
268 | each cg_cgroup_link is linked into a list of cg_cgroup_links for | |
d19e0583 | 269 | a single cgroup on its cgrp_link_list field, and a list of |
817929ec PM |
270 | cg_cgroup_links for a single css_set on its cg_link_list. |
271 | ||
272 | Thus the set of tasks in a cgroup can be listed by iterating over | |
273 | each css_set that references the cgroup, and sub-iterating over | |
274 | each css_set's task set. | |
275 | ||
ddbcc7e8 PM |
276 | The use of a Linux virtual file system (vfs) to represent the |
277 | cgroup hierarchy provides for a familiar permission and name space | |
278 | for cgroups, with a minimum of additional kernel code. | |
279 | ||
280 | 1.4 What does notify_on_release do ? | |
281 | ------------------------------------ | |
282 | ||
ddbcc7e8 PM |
283 | If the notify_on_release flag is enabled (1) in a cgroup, then |
284 | whenever the last task in the cgroup leaves (exits or attaches to | |
285 | some other cgroup) and the last child cgroup of that cgroup | |
286 | is removed, then the kernel runs the command specified by the contents | |
287 | of the "release_agent" file in that hierarchy's root directory, | |
288 | supplying the pathname (relative to the mount point of the cgroup | |
289 | file system) of the abandoned cgroup. This enables automatic | |
290 | removal of abandoned cgroups. The default value of | |
291 | notify_on_release in the root cgroup at system boot is disabled | |
292 | (0). The default value of other cgroups at creation is the current | |
293 | value of their parents notify_on_release setting. The default value of | |
294 | a cgroup hierarchy's release_agent path is empty. | |
295 | ||
296 | 1.5 How do I use cgroups ? | |
297 | -------------------------- | |
298 | ||
299 | To start a new job that is to be contained within a cgroup, using | |
300 | the "cpuset" cgroup subsystem, the steps are something like: | |
301 | ||
302 | 1) mkdir /dev/cgroup | |
303 | 2) mount -t cgroup -ocpuset cpuset /dev/cgroup | |
304 | 3) Create the new cgroup by doing mkdir's and write's (or echo's) in | |
305 | the /dev/cgroup virtual file system. | |
306 | 4) Start a task that will be the "founding father" of the new job. | |
307 | 5) Attach that task to the new cgroup by writing its pid to the | |
308 | /dev/cgroup tasks file for that cgroup. | |
309 | 6) fork, exec or clone the job tasks from this founding father task. | |
310 | ||
311 | For example, the following sequence of commands will setup a cgroup | |
312 | named "Charlie", containing just CPUs 2 and 3, and Memory Node 1, | |
313 | and then start a subshell 'sh' in that cgroup: | |
314 | ||
315 | mount -t cgroup cpuset -ocpuset /dev/cgroup | |
316 | cd /dev/cgroup | |
317 | mkdir Charlie | |
318 | cd Charlie | |
0f146a76 DG |
319 | /bin/echo 2-3 > cpuset.cpus |
320 | /bin/echo 1 > cpuset.mems | |
ddbcc7e8 PM |
321 | /bin/echo $$ > tasks |
322 | sh | |
323 | # The subshell 'sh' is now running in cgroup Charlie | |
324 | # The next line should display '/Charlie' | |
325 | cat /proc/self/cgroup | |
326 | ||
327 | 2. Usage Examples and Syntax | |
328 | ============================ | |
329 | ||
330 | 2.1 Basic Usage | |
331 | --------------- | |
332 | ||
333 | Creating, modifying, using the cgroups can be done through the cgroup | |
334 | virtual filesystem. | |
335 | ||
caa790ba | 336 | To mount a cgroup hierarchy with all available subsystems, type: |
ddbcc7e8 PM |
337 | # mount -t cgroup xxx /dev/cgroup |
338 | ||
339 | The "xxx" is not interpreted by the cgroup code, but will appear in | |
340 | /proc/mounts so may be any useful identifying string that you like. | |
341 | ||
595f4b69 | 342 | To mount a cgroup hierarchy with just the cpuset and memory |
ddbcc7e8 | 343 | subsystems, type: |
b6719ec1 | 344 | # mount -t cgroup -o cpuset,memory hier1 /dev/cgroup |
ddbcc7e8 PM |
345 | |
346 | To change the set of subsystems bound to a mounted hierarchy, just | |
347 | remount with different options: | |
b6719ec1 | 348 | # mount -o remount,cpuset,ns hier1 /dev/cgroup |
ddbcc7e8 | 349 | |
b6719ec1 LZ |
350 | Now memory is removed from the hierarchy and ns is added. |
351 | ||
352 | Note this will add ns to the hierarchy but won't remove memory or | |
353 | cpuset, because the new options are appended to the old ones: | |
354 | # mount -o remount,ns /dev/cgroup | |
355 | ||
356 | To Specify a hierarchy's release_agent: | |
357 | # mount -t cgroup -o cpuset,release_agent="/sbin/cpuset_release_agent" \ | |
358 | xxx /dev/cgroup | |
359 | ||
360 | Note that specifying 'release_agent' more than once will return failure. | |
ddbcc7e8 PM |
361 | |
362 | Note that changing the set of subsystems is currently only supported | |
363 | when the hierarchy consists of a single (root) cgroup. Supporting | |
364 | the ability to arbitrarily bind/unbind subsystems from an existing | |
365 | cgroup hierarchy is intended to be implemented in the future. | |
366 | ||
367 | Then under /dev/cgroup you can find a tree that corresponds to the | |
368 | tree of the cgroups in the system. For instance, /dev/cgroup | |
369 | is the cgroup that holds the whole system. | |
370 | ||
b6719ec1 LZ |
371 | If you want to change the value of release_agent: |
372 | # echo "/sbin/new_release_agent" > /dev/cgroup/release_agent | |
373 | ||
374 | It can also be changed via remount. | |
375 | ||
ddbcc7e8 PM |
376 | If you want to create a new cgroup under /dev/cgroup: |
377 | # cd /dev/cgroup | |
378 | # mkdir my_cgroup | |
379 | ||
380 | Now you want to do something with this cgroup. | |
381 | # cd my_cgroup | |
382 | ||
383 | In this directory you can find several files: | |
384 | # ls | |
7823da36 | 385 | cgroup.procs notify_on_release tasks |
d19e0583 | 386 | (plus whatever files added by the attached subsystems) |
ddbcc7e8 PM |
387 | |
388 | Now attach your shell to this cgroup: | |
389 | # /bin/echo $$ > tasks | |
390 | ||
391 | You can also create cgroups inside your cgroup by using mkdir in this | |
392 | directory. | |
393 | # mkdir my_sub_cs | |
394 | ||
395 | To remove a cgroup, just use rmdir: | |
396 | # rmdir my_sub_cs | |
397 | ||
398 | This will fail if the cgroup is in use (has cgroups inside, or | |
399 | has processes attached, or is held alive by other subsystem-specific | |
400 | reference). | |
401 | ||
402 | 2.2 Attaching processes | |
403 | ----------------------- | |
404 | ||
405 | # /bin/echo PID > tasks | |
406 | ||
407 | Note that it is PID, not PIDs. You can only attach ONE task at a time. | |
408 | If you have several tasks to attach, you have to do it one after another: | |
409 | ||
410 | # /bin/echo PID1 > tasks | |
411 | # /bin/echo PID2 > tasks | |
412 | ... | |
413 | # /bin/echo PIDn > tasks | |
414 | ||
bef67c5a LZ |
415 | You can attach the current shell task by echoing 0: |
416 | ||
417 | # echo 0 > tasks | |
418 | ||
c6d57f33 PM |
419 | 2.3 Mounting hierarchies by name |
420 | -------------------------------- | |
421 | ||
422 | Passing the name=<x> option when mounting a cgroups hierarchy | |
423 | associates the given name with the hierarchy. This can be used when | |
424 | mounting a pre-existing hierarchy, in order to refer to it by name | |
425 | rather than by its set of active subsystems. Each hierarchy is either | |
426 | nameless, or has a unique name. | |
427 | ||
428 | The name should match [\w.-]+ | |
429 | ||
430 | When passing a name=<x> option for a new hierarchy, you need to | |
431 | specify subsystems manually; the legacy behaviour of mounting all | |
432 | subsystems when none are explicitly specified is not supported when | |
433 | you give a subsystem a name. | |
434 | ||
435 | The name of the subsystem appears as part of the hierarchy description | |
436 | in /proc/mounts and /proc/<pid>/cgroups. | |
437 | ||
0dea1168 KS |
438 | 2.4 Notification API |
439 | -------------------- | |
440 | ||
441 | There is mechanism which allows to get notifications about changing | |
442 | status of a cgroup. | |
443 | ||
444 | To register new notification handler you need: | |
445 | - create a file descriptor for event notification using eventfd(2); | |
446 | - open a control file to be monitored (e.g. memory.usage_in_bytes); | |
447 | - write "<event_fd> <control_fd> <args>" to cgroup.event_control. | |
448 | Interpretation of args is defined by control file implementation; | |
449 | ||
450 | eventfd will be woken up by control file implementation or when the | |
451 | cgroup is removed. | |
452 | ||
453 | To unregister notification handler just close eventfd. | |
454 | ||
455 | NOTE: Support of notifications should be implemented for the control | |
456 | file. See documentation for the subsystem. | |
c6d57f33 | 457 | |
ddbcc7e8 PM |
458 | 3. Kernel API |
459 | ============= | |
460 | ||
461 | 3.1 Overview | |
462 | ------------ | |
463 | ||
464 | Each kernel subsystem that wants to hook into the generic cgroup | |
465 | system needs to create a cgroup_subsys object. This contains | |
466 | various methods, which are callbacks from the cgroup system, along | |
467 | with a subsystem id which will be assigned by the cgroup system. | |
468 | ||
469 | Other fields in the cgroup_subsys object include: | |
470 | ||
471 | - subsys_id: a unique array index for the subsystem, indicating which | |
d19e0583 | 472 | entry in cgroup->subsys[] this subsystem should be managing. |
ddbcc7e8 | 473 | |
d19e0583 LZ |
474 | - name: should be initialized to a unique subsystem name. Should be |
475 | no longer than MAX_CGROUP_TYPE_NAMELEN. | |
ddbcc7e8 | 476 | |
d19e0583 LZ |
477 | - early_init: indicate if the subsystem needs early initialization |
478 | at system boot. | |
ddbcc7e8 PM |
479 | |
480 | Each cgroup object created by the system has an array of pointers, | |
481 | indexed by subsystem id; this pointer is entirely managed by the | |
482 | subsystem; the generic cgroup code will never touch this pointer. | |
483 | ||
484 | 3.2 Synchronization | |
485 | ------------------- | |
486 | ||
487 | There is a global mutex, cgroup_mutex, used by the cgroup | |
488 | system. This should be taken by anything that wants to modify a | |
489 | cgroup. It may also be taken to prevent cgroups from being | |
490 | modified, but more specific locks may be more appropriate in that | |
491 | situation. | |
492 | ||
493 | See kernel/cgroup.c for more details. | |
494 | ||
495 | Subsystems can take/release the cgroup_mutex via the functions | |
ddbcc7e8 PM |
496 | cgroup_lock()/cgroup_unlock(). |
497 | ||
498 | Accessing a task's cgroup pointer may be done in the following ways: | |
499 | - while holding cgroup_mutex | |
500 | - while holding the task's alloc_lock (via task_lock()) | |
501 | - inside an rcu_read_lock() section via rcu_dereference() | |
502 | ||
503 | 3.3 Subsystem API | |
d19e0583 | 504 | ----------------- |
ddbcc7e8 PM |
505 | |
506 | Each subsystem should: | |
507 | ||
508 | - add an entry in linux/cgroup_subsys.h | |
509 | - define a cgroup_subsys object called <name>_subsys | |
510 | ||
e6a1105b | 511 | If a subsystem can be compiled as a module, it should also have in its |
cf5d5941 BB |
512 | module initcall a call to cgroup_load_subsys(), and in its exitcall a |
513 | call to cgroup_unload_subsys(). It should also set its_subsys.module = | |
514 | THIS_MODULE in its .c file. | |
e6a1105b | 515 | |
ddbcc7e8 PM |
516 | Each subsystem may export the following methods. The only mandatory |
517 | methods are create/destroy. Any others that are null are presumed to | |
518 | be successful no-ops. | |
519 | ||
d19e0583 LZ |
520 | struct cgroup_subsys_state *create(struct cgroup_subsys *ss, |
521 | struct cgroup *cgrp) | |
8dc4f3e1 | 522 | (cgroup_mutex held by caller) |
ddbcc7e8 PM |
523 | |
524 | Called to create a subsystem state object for a cgroup. The | |
525 | subsystem should allocate its subsystem state object for the passed | |
526 | cgroup, returning a pointer to the new object on success or a | |
527 | negative error code. On success, the subsystem pointer should point to | |
528 | a structure of type cgroup_subsys_state (typically embedded in a | |
529 | larger subsystem-specific object), which will be initialized by the | |
530 | cgroup system. Note that this will be called at initialization to | |
531 | create the root subsystem state for this subsystem; this case can be | |
532 | identified by the passed cgroup object having a NULL parent (since | |
533 | it's the root of the hierarchy) and may be an appropriate place for | |
534 | initialization code. | |
535 | ||
d19e0583 | 536 | void destroy(struct cgroup_subsys *ss, struct cgroup *cgrp) |
8dc4f3e1 | 537 | (cgroup_mutex held by caller) |
ddbcc7e8 | 538 | |
8dc4f3e1 PM |
539 | The cgroup system is about to destroy the passed cgroup; the subsystem |
540 | should do any necessary cleanup and free its subsystem state | |
541 | object. By the time this method is called, the cgroup has already been | |
542 | unlinked from the file system and from the child list of its parent; | |
543 | cgroup->parent is still valid. (Note - can also be called for a | |
544 | newly-created cgroup if an error occurs after this subsystem's | |
545 | create() method has been called for the new cgroup). | |
ddbcc7e8 | 546 | |
ec64f515 | 547 | int pre_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp); |
d19e0583 LZ |
548 | |
549 | Called before checking the reference count on each subsystem. This may | |
550 | be useful for subsystems which have some extra references even if | |
ec64f515 KH |
551 | there are not tasks in the cgroup. If pre_destroy() returns error code, |
552 | rmdir() will fail with it. From this behavior, pre_destroy() can be | |
553 | called multiple times against a cgroup. | |
d19e0583 LZ |
554 | |
555 | int can_attach(struct cgroup_subsys *ss, struct cgroup *cgrp, | |
be367d09 | 556 | struct task_struct *task, bool threadgroup) |
8dc4f3e1 | 557 | (cgroup_mutex held by caller) |
ddbcc7e8 PM |
558 | |
559 | Called prior to moving a task into a cgroup; if the subsystem | |
560 | returns an error, this will abort the attach operation. If a NULL | |
561 | task is passed, then a successful result indicates that *any* | |
562 | unspecified task can be moved into the cgroup. Note that this isn't | |
563 | called on a fork. If this method returns 0 (success) then this should | |
2468c723 DN |
564 | remain valid while the caller holds cgroup_mutex and it is ensured that either |
565 | attach() or cancel_attach() will be called in future. If threadgroup is | |
be367d09 BB |
566 | true, then a successful result indicates that all threads in the given |
567 | thread's threadgroup can be moved together. | |
ddbcc7e8 | 568 | |
2468c723 DN |
569 | void cancel_attach(struct cgroup_subsys *ss, struct cgroup *cgrp, |
570 | struct task_struct *task, bool threadgroup) | |
571 | (cgroup_mutex held by caller) | |
572 | ||
573 | Called when a task attach operation has failed after can_attach() has succeeded. | |
574 | A subsystem whose can_attach() has some side-effects should provide this | |
88393161 | 575 | function, so that the subsystem can implement a rollback. If not, not necessary. |
2468c723 DN |
576 | This will be called only about subsystems whose can_attach() operation have |
577 | succeeded. | |
578 | ||
d19e0583 | 579 | void attach(struct cgroup_subsys *ss, struct cgroup *cgrp, |
be367d09 BB |
580 | struct cgroup *old_cgrp, struct task_struct *task, |
581 | bool threadgroup) | |
18e7f1f0 | 582 | (cgroup_mutex held by caller) |
ddbcc7e8 PM |
583 | |
584 | Called after the task has been attached to the cgroup, to allow any | |
585 | post-attachment activity that requires memory allocations or blocking. | |
be367d09 BB |
586 | If threadgroup is true, the subsystem should take care of all threads |
587 | in the specified thread's threadgroup. Currently does not support any | |
588 | subsystem that might need the old_cgrp for every thread in the group. | |
ddbcc7e8 PM |
589 | |
590 | void fork(struct cgroup_subsy *ss, struct task_struct *task) | |
ddbcc7e8 | 591 | |
e8d55fde | 592 | Called when a task is forked into a cgroup. |
ddbcc7e8 PM |
593 | |
594 | void exit(struct cgroup_subsys *ss, struct task_struct *task) | |
ddbcc7e8 | 595 | |
d19e0583 | 596 | Called during task exit. |
ddbcc7e8 | 597 | |
d19e0583 | 598 | int populate(struct cgroup_subsys *ss, struct cgroup *cgrp) |
18e7f1f0 | 599 | (cgroup_mutex held by caller) |
ddbcc7e8 PM |
600 | |
601 | Called after creation of a cgroup to allow a subsystem to populate | |
602 | the cgroup directory with file entries. The subsystem should make | |
603 | calls to cgroup_add_file() with objects of type cftype (see | |
604 | include/linux/cgroup.h for details). Note that although this | |
605 | method can return an error code, the error code is currently not | |
606 | always handled well. | |
607 | ||
d19e0583 | 608 | void post_clone(struct cgroup_subsys *ss, struct cgroup *cgrp) |
18e7f1f0 | 609 | (cgroup_mutex held by caller) |
697f4161 | 610 | |
caa790ba | 611 | Called at the end of cgroup_clone() to do any parameter |
697f4161 PM |
612 | initialization which might be required before a task could attach. For |
613 | example in cpusets, no task may attach before 'cpus' and 'mems' are set | |
614 | up. | |
615 | ||
ddbcc7e8 | 616 | void bind(struct cgroup_subsys *ss, struct cgroup *root) |
999cd8a4 | 617 | (cgroup_mutex and ss->hierarchy_mutex held by caller) |
ddbcc7e8 PM |
618 | |
619 | Called when a cgroup subsystem is rebound to a different hierarchy | |
620 | and root cgroup. Currently this will only involve movement between | |
621 | the default hierarchy (which never has sub-cgroups) and a hierarchy | |
622 | that is being created/destroyed (and hence has no sub-cgroups). | |
623 | ||
624 | 4. Questions | |
625 | ============ | |
626 | ||
627 | Q: what's up with this '/bin/echo' ? | |
628 | A: bash's builtin 'echo' command does not check calls to write() against | |
629 | errors. If you use it in the cgroup file system, you won't be | |
630 | able to tell whether a command succeeded or failed. | |
631 | ||
632 | Q: When I attach processes, only the first of the line gets really attached ! | |
633 | A: We can only return one error code per call to write(). So you should also | |
634 | put only ONE pid. | |
635 |