Commit | Line | Data |
---|---|---|
ddbcc7e8 PM |
1 | #ifndef _LINUX_CGROUP_H |
2 | #define _LINUX_CGROUP_H | |
3 | /* | |
4 | * cgroup interface | |
5 | * | |
6 | * Copyright (C) 2003 BULL SA | |
7 | * Copyright (C) 2004-2006 Silicon Graphics, Inc. | |
8 | * | |
9 | */ | |
10 | ||
11 | #include <linux/sched.h> | |
ddbcc7e8 PM |
12 | #include <linux/cpumask.h> |
13 | #include <linux/nodemask.h> | |
14 | #include <linux/rcupdate.h> | |
846c7bb0 | 15 | #include <linux/cgroupstats.h> |
31a7df01 | 16 | #include <linux/prio_heap.h> |
cc31edce | 17 | #include <linux/rwsem.h> |
38460b48 | 18 | #include <linux/idr.h> |
ddbcc7e8 PM |
19 | |
20 | #ifdef CONFIG_CGROUPS | |
21 | ||
22 | struct cgroupfs_root; | |
23 | struct cgroup_subsys; | |
24 | struct inode; | |
84eea842 | 25 | struct cgroup; |
38460b48 | 26 | struct css_id; |
ddbcc7e8 PM |
27 | |
28 | extern int cgroup_init_early(void); | |
29 | extern int cgroup_init(void); | |
ddbcc7e8 | 30 | extern void cgroup_lock(void); |
d11c563d | 31 | extern int cgroup_lock_is_held(void); |
84eea842 | 32 | extern bool cgroup_lock_live_group(struct cgroup *cgrp); |
ddbcc7e8 | 33 | extern void cgroup_unlock(void); |
b4f48b63 PM |
34 | extern void cgroup_fork(struct task_struct *p); |
35 | extern void cgroup_fork_callbacks(struct task_struct *p); | |
817929ec | 36 | extern void cgroup_post_fork(struct task_struct *p); |
b4f48b63 | 37 | extern void cgroup_exit(struct task_struct *p, int run_callbacks); |
846c7bb0 BS |
38 | extern int cgroupstats_build(struct cgroupstats *stats, |
39 | struct dentry *dentry); | |
e6a1105b | 40 | extern int cgroup_load_subsys(struct cgroup_subsys *ss); |
cf5d5941 | 41 | extern void cgroup_unload_subsys(struct cgroup_subsys *ss); |
ddbcc7e8 | 42 | |
828c0950 | 43 | extern const struct file_operations proc_cgroup_operations; |
a424316c | 44 | |
aae8aab4 | 45 | /* Define the enumeration of all builtin cgroup subsystems */ |
817929ec PM |
46 | #define SUBSYS(_x) _x ## _subsys_id, |
47 | enum cgroup_subsys_id { | |
48 | #include <linux/cgroup_subsys.h> | |
aae8aab4 | 49 | CGROUP_BUILTIN_SUBSYS_COUNT |
817929ec PM |
50 | }; |
51 | #undef SUBSYS | |
aae8aab4 BB |
52 | /* |
53 | * This define indicates the maximum number of subsystems that can be loaded | |
54 | * at once. We limit to this many since cgroupfs_root has subsys_bits to keep | |
55 | * track of all of them. | |
56 | */ | |
57 | #define CGROUP_SUBSYS_COUNT (BITS_PER_BYTE*sizeof(unsigned long)) | |
817929ec | 58 | |
ddbcc7e8 PM |
59 | /* Per-subsystem/per-cgroup state maintained by the system. */ |
60 | struct cgroup_subsys_state { | |
d20a390a PM |
61 | /* |
62 | * The cgroup that this subsystem is attached to. Useful | |
ddbcc7e8 | 63 | * for subsystems that want to know about the cgroup |
d20a390a PM |
64 | * hierarchy structure |
65 | */ | |
ddbcc7e8 PM |
66 | struct cgroup *cgroup; |
67 | ||
d20a390a PM |
68 | /* |
69 | * State maintained by the cgroup system to allow subsystems | |
e7c5ec91 | 70 | * to be "busy". Should be accessed via css_get(), |
d20a390a PM |
71 | * css_tryget() and and css_put(). |
72 | */ | |
ddbcc7e8 PM |
73 | |
74 | atomic_t refcnt; | |
75 | ||
76 | unsigned long flags; | |
38460b48 | 77 | /* ID for this css, if possible */ |
2c392b8c | 78 | struct css_id __rcu *id; |
ddbcc7e8 PM |
79 | }; |
80 | ||
81 | /* bits in struct cgroup_subsys_state flags field */ | |
82 | enum { | |
83 | CSS_ROOT, /* This CSS is the root of the subsystem */ | |
e7c5ec91 | 84 | CSS_REMOVED, /* This CSS is dead */ |
ddbcc7e8 PM |
85 | }; |
86 | ||
d7b9fff7 DN |
87 | /* Caller must verify that the css is not for root cgroup */ |
88 | static inline void __css_get(struct cgroup_subsys_state *css, int count) | |
89 | { | |
90 | atomic_add(count, &css->refcnt); | |
91 | } | |
92 | ||
ddbcc7e8 | 93 | /* |
e7c5ec91 PM |
94 | * Call css_get() to hold a reference on the css; it can be used |
95 | * for a reference obtained via: | |
96 | * - an existing ref-counted reference to the css | |
97 | * - task->cgroups for a locked task | |
ddbcc7e8 PM |
98 | */ |
99 | ||
100 | static inline void css_get(struct cgroup_subsys_state *css) | |
101 | { | |
102 | /* We don't need to reference count the root state */ | |
103 | if (!test_bit(CSS_ROOT, &css->flags)) | |
d7b9fff7 | 104 | __css_get(css, 1); |
ddbcc7e8 | 105 | } |
e7c5ec91 PM |
106 | |
107 | static inline bool css_is_removed(struct cgroup_subsys_state *css) | |
108 | { | |
109 | return test_bit(CSS_REMOVED, &css->flags); | |
110 | } | |
111 | ||
112 | /* | |
113 | * Call css_tryget() to take a reference on a css if your existing | |
114 | * (known-valid) reference isn't already ref-counted. Returns false if | |
115 | * the css has been destroyed. | |
116 | */ | |
117 | ||
118 | static inline bool css_tryget(struct cgroup_subsys_state *css) | |
119 | { | |
120 | if (test_bit(CSS_ROOT, &css->flags)) | |
121 | return true; | |
122 | while (!atomic_inc_not_zero(&css->refcnt)) { | |
123 | if (test_bit(CSS_REMOVED, &css->flags)) | |
124 | return false; | |
804b3c28 | 125 | cpu_relax(); |
e7c5ec91 PM |
126 | } |
127 | return true; | |
128 | } | |
129 | ||
ddbcc7e8 PM |
130 | /* |
131 | * css_put() should be called to release a reference taken by | |
e7c5ec91 | 132 | * css_get() or css_tryget() |
ddbcc7e8 PM |
133 | */ |
134 | ||
d7b9fff7 | 135 | extern void __css_put(struct cgroup_subsys_state *css, int count); |
ddbcc7e8 PM |
136 | static inline void css_put(struct cgroup_subsys_state *css) |
137 | { | |
138 | if (!test_bit(CSS_ROOT, &css->flags)) | |
d7b9fff7 | 139 | __css_put(css, 1); |
ddbcc7e8 PM |
140 | } |
141 | ||
3116f0e3 PM |
142 | /* bits in struct cgroup flags field */ |
143 | enum { | |
144 | /* Control Group is dead */ | |
145 | CGRP_REMOVED, | |
d20a390a PM |
146 | /* |
147 | * Control Group has previously had a child cgroup or a task, | |
148 | * but no longer (only if CGRP_NOTIFY_ON_RELEASE is set) | |
149 | */ | |
3116f0e3 PM |
150 | CGRP_RELEASABLE, |
151 | /* Control Group requires release notifications to userspace */ | |
152 | CGRP_NOTIFY_ON_RELEASE, | |
ec64f515 KH |
153 | /* |
154 | * A thread in rmdir() is wating for this cgroup. | |
155 | */ | |
156 | CGRP_WAIT_ON_RMDIR, | |
97978e6d DL |
157 | /* |
158 | * Clone cgroup values when creating a new child cgroup | |
159 | */ | |
160 | CGRP_CLONE_CHILDREN, | |
3116f0e3 PM |
161 | }; |
162 | ||
72a8cb30 BB |
163 | /* which pidlist file are we talking about? */ |
164 | enum cgroup_filetype { | |
165 | CGROUP_FILE_PROCS, | |
166 | CGROUP_FILE_TASKS, | |
167 | }; | |
168 | ||
169 | /* | |
170 | * A pidlist is a list of pids that virtually represents the contents of one | |
171 | * of the cgroup files ("procs" or "tasks"). We keep a list of such pidlists, | |
172 | * a pair (one each for procs, tasks) for each pid namespace that's relevant | |
173 | * to the cgroup. | |
174 | */ | |
102a775e | 175 | struct cgroup_pidlist { |
72a8cb30 BB |
176 | /* |
177 | * used to find which pidlist is wanted. doesn't change as long as | |
178 | * this particular list stays in the list. | |
179 | */ | |
180 | struct { enum cgroup_filetype type; struct pid_namespace *ns; } key; | |
102a775e BB |
181 | /* array of xids */ |
182 | pid_t *list; | |
183 | /* how many elements the above list has */ | |
184 | int length; | |
185 | /* how many files are using the current array */ | |
186 | int use_count; | |
72a8cb30 BB |
187 | /* each of these stored in a list by its cgroup */ |
188 | struct list_head links; | |
189 | /* pointer to the cgroup we belong to, for list removal purposes */ | |
190 | struct cgroup *owner; | |
191 | /* protects the other fields */ | |
192 | struct rw_semaphore mutex; | |
102a775e BB |
193 | }; |
194 | ||
ddbcc7e8 PM |
195 | struct cgroup { |
196 | unsigned long flags; /* "unsigned long" so bitops work */ | |
197 | ||
d20a390a PM |
198 | /* |
199 | * count users of this cgroup. >0 means busy, but doesn't | |
200 | * necessarily indicate the number of tasks in the cgroup | |
201 | */ | |
ddbcc7e8 PM |
202 | atomic_t count; |
203 | ||
204 | /* | |
205 | * We link our 'sibling' struct into our parent's 'children'. | |
206 | * Our children link their 'sibling' into our 'children'. | |
207 | */ | |
208 | struct list_head sibling; /* my parent's children */ | |
209 | struct list_head children; /* my children */ | |
210 | ||
d20a390a | 211 | struct cgroup *parent; /* my parent */ |
2c392b8c | 212 | struct dentry __rcu *dentry; /* cgroup fs entry, RCU protected */ |
ddbcc7e8 PM |
213 | |
214 | /* Private pointers for each registered subsystem */ | |
215 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
216 | ||
217 | struct cgroupfs_root *root; | |
218 | struct cgroup *top_cgroup; | |
817929ec PM |
219 | |
220 | /* | |
221 | * List of cg_cgroup_links pointing at css_sets with | |
222 | * tasks in this cgroup. Protected by css_set_lock | |
223 | */ | |
224 | struct list_head css_sets; | |
81a6a5cd PM |
225 | |
226 | /* | |
227 | * Linked list running through all cgroups that can | |
228 | * potentially be reaped by the release agent. Protected by | |
229 | * release_list_lock | |
230 | */ | |
231 | struct list_head release_list; | |
cc31edce | 232 | |
72a8cb30 BB |
233 | /* |
234 | * list of pidlists, up to two for each namespace (one for procs, one | |
235 | * for tasks); created on demand. | |
236 | */ | |
237 | struct list_head pidlists; | |
238 | struct mutex pidlist_mutex; | |
a47295e6 PM |
239 | |
240 | /* For RCU-protected deletion */ | |
241 | struct rcu_head rcu_head; | |
0dea1168 | 242 | |
25985edc | 243 | /* List of events which userspace want to receive */ |
0dea1168 KS |
244 | struct list_head event_list; |
245 | spinlock_t event_list_lock; | |
817929ec PM |
246 | }; |
247 | ||
d20a390a PM |
248 | /* |
249 | * A css_set is a structure holding pointers to a set of | |
817929ec PM |
250 | * cgroup_subsys_state objects. This saves space in the task struct |
251 | * object and speeds up fork()/exit(), since a single inc/dec and a | |
d20a390a PM |
252 | * list_add()/del() can bump the reference count on the entire cgroup |
253 | * set for a task. | |
817929ec PM |
254 | */ |
255 | ||
256 | struct css_set { | |
257 | ||
258 | /* Reference count */ | |
146aa1bd | 259 | atomic_t refcount; |
817929ec | 260 | |
472b1053 LZ |
261 | /* |
262 | * List running through all cgroup groups in the same hash | |
263 | * slot. Protected by css_set_lock | |
264 | */ | |
265 | struct hlist_node hlist; | |
266 | ||
817929ec PM |
267 | /* |
268 | * List running through all tasks using this cgroup | |
269 | * group. Protected by css_set_lock | |
270 | */ | |
271 | struct list_head tasks; | |
272 | ||
273 | /* | |
274 | * List of cg_cgroup_link objects on link chains from | |
275 | * cgroups referenced from this css_set. Protected by | |
276 | * css_set_lock | |
277 | */ | |
278 | struct list_head cg_links; | |
279 | ||
280 | /* | |
281 | * Set of subsystem states, one for each subsystem. This array | |
282 | * is immutable after creation apart from the init_css_set | |
cf5d5941 BB |
283 | * during subsystem registration (at boot time) and modular subsystem |
284 | * loading/unloading. | |
817929ec PM |
285 | */ |
286 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
c378369d BB |
287 | |
288 | /* For RCU-protected deletion */ | |
289 | struct rcu_head rcu_head; | |
ddbcc7e8 PM |
290 | }; |
291 | ||
91796569 PM |
292 | /* |
293 | * cgroup_map_cb is an abstract callback API for reporting map-valued | |
294 | * control files | |
295 | */ | |
296 | ||
297 | struct cgroup_map_cb { | |
298 | int (*fill)(struct cgroup_map_cb *cb, const char *key, u64 value); | |
299 | void *state; | |
300 | }; | |
301 | ||
d20a390a PM |
302 | /* |
303 | * struct cftype: handler definitions for cgroup control files | |
ddbcc7e8 PM |
304 | * |
305 | * When reading/writing to a file: | |
a043e3b2 | 306 | * - the cgroup to use is file->f_dentry->d_parent->d_fsdata |
ddbcc7e8 PM |
307 | * - the 'cftype' of the file is file->f_dentry->d_fsdata |
308 | */ | |
309 | ||
310 | #define MAX_CFTYPE_NAME 64 | |
311 | struct cftype { | |
d20a390a PM |
312 | /* |
313 | * By convention, the name should begin with the name of the | |
314 | * subsystem, followed by a period | |
315 | */ | |
ddbcc7e8 PM |
316 | char name[MAX_CFTYPE_NAME]; |
317 | int private; | |
099fca32 LZ |
318 | /* |
319 | * If not 0, file mode is set to this value, otherwise it will | |
320 | * be figured out automatically | |
321 | */ | |
a5e7ed32 | 322 | umode_t mode; |
db3b1497 PM |
323 | |
324 | /* | |
325 | * If non-zero, defines the maximum length of string that can | |
326 | * be passed to write_string; defaults to 64 | |
327 | */ | |
328 | size_t max_write_len; | |
329 | ||
ce16b49d PM |
330 | int (*open)(struct inode *inode, struct file *file); |
331 | ssize_t (*read)(struct cgroup *cgrp, struct cftype *cft, | |
332 | struct file *file, | |
333 | char __user *buf, size_t nbytes, loff_t *ppos); | |
ddbcc7e8 | 334 | /* |
f4c753b7 | 335 | * read_u64() is a shortcut for the common case of returning a |
ddbcc7e8 PM |
336 | * single integer. Use it in place of read() |
337 | */ | |
ce16b49d | 338 | u64 (*read_u64)(struct cgroup *cgrp, struct cftype *cft); |
e73d2c61 PM |
339 | /* |
340 | * read_s64() is a signed version of read_u64() | |
341 | */ | |
ce16b49d | 342 | s64 (*read_s64)(struct cgroup *cgrp, struct cftype *cft); |
91796569 PM |
343 | /* |
344 | * read_map() is used for defining a map of key/value | |
345 | * pairs. It should call cb->fill(cb, key, value) for each | |
346 | * entry. The key/value pairs (and their ordering) should not | |
347 | * change between reboots. | |
348 | */ | |
ce16b49d PM |
349 | int (*read_map)(struct cgroup *cont, struct cftype *cft, |
350 | struct cgroup_map_cb *cb); | |
29486df3 SH |
351 | /* |
352 | * read_seq_string() is used for outputting a simple sequence | |
353 | * using seqfile. | |
354 | */ | |
ce16b49d PM |
355 | int (*read_seq_string)(struct cgroup *cont, struct cftype *cft, |
356 | struct seq_file *m); | |
91796569 | 357 | |
ce16b49d PM |
358 | ssize_t (*write)(struct cgroup *cgrp, struct cftype *cft, |
359 | struct file *file, | |
360 | const char __user *buf, size_t nbytes, loff_t *ppos); | |
355e0c48 PM |
361 | |
362 | /* | |
f4c753b7 | 363 | * write_u64() is a shortcut for the common case of accepting |
355e0c48 PM |
364 | * a single integer (as parsed by simple_strtoull) from |
365 | * userspace. Use in place of write(); return 0 or error. | |
366 | */ | |
ce16b49d | 367 | int (*write_u64)(struct cgroup *cgrp, struct cftype *cft, u64 val); |
e73d2c61 PM |
368 | /* |
369 | * write_s64() is a signed version of write_u64() | |
370 | */ | |
ce16b49d | 371 | int (*write_s64)(struct cgroup *cgrp, struct cftype *cft, s64 val); |
355e0c48 | 372 | |
db3b1497 PM |
373 | /* |
374 | * write_string() is passed a nul-terminated kernelspace | |
375 | * buffer of maximum length determined by max_write_len. | |
376 | * Returns 0 or -ve error code. | |
377 | */ | |
378 | int (*write_string)(struct cgroup *cgrp, struct cftype *cft, | |
379 | const char *buffer); | |
d447ea2f PE |
380 | /* |
381 | * trigger() callback can be used to get some kick from the | |
382 | * userspace, when the actual string written is not important | |
383 | * at all. The private field can be used to determine the | |
384 | * kick type for multiplexing. | |
385 | */ | |
386 | int (*trigger)(struct cgroup *cgrp, unsigned int event); | |
387 | ||
ce16b49d | 388 | int (*release)(struct inode *inode, struct file *file); |
0dea1168 KS |
389 | |
390 | /* | |
391 | * register_event() callback will be used to add new userspace | |
392 | * waiter for changes related to the cftype. Implement it if | |
393 | * you want to provide this functionality. Use eventfd_signal() | |
394 | * on eventfd to send notification to userspace. | |
395 | */ | |
396 | int (*register_event)(struct cgroup *cgrp, struct cftype *cft, | |
397 | struct eventfd_ctx *eventfd, const char *args); | |
398 | /* | |
399 | * unregister_event() callback will be called when userspace | |
400 | * closes the eventfd or on cgroup removing. | |
401 | * This callback must be implemented, if you want provide | |
402 | * notification functionality. | |
0dea1168 | 403 | */ |
907860ed | 404 | void (*unregister_event)(struct cgroup *cgrp, struct cftype *cft, |
0dea1168 | 405 | struct eventfd_ctx *eventfd); |
ddbcc7e8 PM |
406 | }; |
407 | ||
31a7df01 CW |
408 | struct cgroup_scanner { |
409 | struct cgroup *cg; | |
410 | int (*test_task)(struct task_struct *p, struct cgroup_scanner *scan); | |
411 | void (*process_task)(struct task_struct *p, | |
412 | struct cgroup_scanner *scan); | |
413 | struct ptr_heap *heap; | |
bd1a8ab7 | 414 | void *data; |
31a7df01 CW |
415 | }; |
416 | ||
d20a390a PM |
417 | /* |
418 | * Add a new file to the given cgroup directory. Should only be | |
419 | * called by subsystems from within a populate() method | |
420 | */ | |
ffd2d883 | 421 | int cgroup_add_file(struct cgroup *cgrp, struct cgroup_subsys *subsys, |
ddbcc7e8 PM |
422 | const struct cftype *cft); |
423 | ||
d20a390a PM |
424 | /* |
425 | * Add a set of new files to the given cgroup directory. Should | |
426 | * only be called by subsystems from within a populate() method | |
427 | */ | |
ffd2d883 | 428 | int cgroup_add_files(struct cgroup *cgrp, |
ddbcc7e8 PM |
429 | struct cgroup_subsys *subsys, |
430 | const struct cftype cft[], | |
431 | int count); | |
432 | ||
ffd2d883 | 433 | int cgroup_is_removed(const struct cgroup *cgrp); |
ddbcc7e8 | 434 | |
ffd2d883 | 435 | int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen); |
ddbcc7e8 | 436 | |
ffd2d883 | 437 | int cgroup_task_count(const struct cgroup *cgrp); |
bbcb81d0 | 438 | |
313e924c GN |
439 | /* Return true if cgrp is a descendant of the task's cgroup */ |
440 | int cgroup_is_descendant(const struct cgroup *cgrp, struct task_struct *task); | |
ddbcc7e8 | 441 | |
88703267 KH |
442 | /* |
443 | * When the subsys has to access css and may add permanent refcnt to css, | |
444 | * it should take care of racy conditions with rmdir(). Following set of | |
445 | * functions, is for stop/restart rmdir if necessary. | |
446 | * Because these will call css_get/put, "css" should be alive css. | |
447 | * | |
448 | * cgroup_exclude_rmdir(); | |
449 | * ...do some jobs which may access arbitrary empty cgroup | |
450 | * cgroup_release_and_wakeup_rmdir(); | |
451 | * | |
452 | * When someone removes a cgroup while cgroup_exclude_rmdir() holds it, | |
453 | * it sleeps and cgroup_release_and_wakeup_rmdir() will wake him up. | |
454 | */ | |
455 | ||
456 | void cgroup_exclude_rmdir(struct cgroup_subsys_state *css); | |
457 | void cgroup_release_and_wakeup_rmdir(struct cgroup_subsys_state *css); | |
458 | ||
2f7ee569 TH |
459 | /* |
460 | * Control Group taskset, used to pass around set of tasks to cgroup_subsys | |
461 | * methods. | |
462 | */ | |
463 | struct cgroup_taskset; | |
464 | struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset); | |
465 | struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset); | |
466 | struct cgroup *cgroup_taskset_cur_cgroup(struct cgroup_taskset *tset); | |
467 | int cgroup_taskset_size(struct cgroup_taskset *tset); | |
468 | ||
469 | /** | |
470 | * cgroup_taskset_for_each - iterate cgroup_taskset | |
471 | * @task: the loop cursor | |
472 | * @skip_cgrp: skip if task's cgroup matches this, %NULL to iterate through all | |
473 | * @tset: taskset to iterate | |
474 | */ | |
475 | #define cgroup_taskset_for_each(task, skip_cgrp, tset) \ | |
476 | for ((task) = cgroup_taskset_first((tset)); (task); \ | |
477 | (task) = cgroup_taskset_next((tset))) \ | |
478 | if (!(skip_cgrp) || \ | |
479 | cgroup_taskset_cur_cgroup((tset)) != (skip_cgrp)) | |
480 | ||
21acb9ca TLSC |
481 | /* |
482 | * Control Group subsystem type. | |
483 | * See Documentation/cgroups/cgroups.txt for details | |
484 | */ | |
ddbcc7e8 PM |
485 | |
486 | struct cgroup_subsys { | |
487 | struct cgroup_subsys_state *(*create)(struct cgroup_subsys *ss, | |
ffd2d883 | 488 | struct cgroup *cgrp); |
ec64f515 | 489 | int (*pre_destroy)(struct cgroup_subsys *ss, struct cgroup *cgrp); |
ffd2d883 | 490 | void (*destroy)(struct cgroup_subsys *ss, struct cgroup *cgrp); |
be367d09 | 491 | int (*can_attach)(struct cgroup_subsys *ss, struct cgroup *cgrp, |
2f7ee569 | 492 | struct cgroup_taskset *tset); |
2468c723 | 493 | void (*cancel_attach)(struct cgroup_subsys *ss, struct cgroup *cgrp, |
2f7ee569 | 494 | struct cgroup_taskset *tset); |
ffd2d883 | 495 | void (*attach)(struct cgroup_subsys *ss, struct cgroup *cgrp, |
2f7ee569 | 496 | struct cgroup_taskset *tset); |
ddbcc7e8 | 497 | void (*fork)(struct cgroup_subsys *ss, struct task_struct *task); |
d41d5a01 PZ |
498 | void (*exit)(struct cgroup_subsys *ss, struct cgroup *cgrp, |
499 | struct cgroup *old_cgrp, struct task_struct *task); | |
ddbcc7e8 | 500 | int (*populate)(struct cgroup_subsys *ss, |
ffd2d883 LZ |
501 | struct cgroup *cgrp); |
502 | void (*post_clone)(struct cgroup_subsys *ss, struct cgroup *cgrp); | |
ddbcc7e8 | 503 | void (*bind)(struct cgroup_subsys *ss, struct cgroup *root); |
e5991371 | 504 | |
ddbcc7e8 PM |
505 | int subsys_id; |
506 | int active; | |
8bab8dde | 507 | int disabled; |
ddbcc7e8 | 508 | int early_init; |
38460b48 KH |
509 | /* |
510 | * True if this subsys uses ID. ID is not available before cgroup_init() | |
511 | * (not available in early_init time.) | |
512 | */ | |
513 | bool use_id; | |
ddbcc7e8 PM |
514 | #define MAX_CGROUP_TYPE_NAMELEN 32 |
515 | const char *name; | |
516 | ||
999cd8a4 PM |
517 | /* |
518 | * Protects sibling/children links of cgroups in this | |
519 | * hierarchy, plus protects which hierarchy (or none) the | |
520 | * subsystem is a part of (i.e. root/sibling). To avoid | |
521 | * potential deadlocks, the following operations should not be | |
522 | * undertaken while holding any hierarchy_mutex: | |
523 | * | |
524 | * - allocating memory | |
525 | * - initiating hotplug events | |
526 | */ | |
527 | struct mutex hierarchy_mutex; | |
cfebe563 | 528 | struct lock_class_key subsys_key; |
ddbcc7e8 | 529 | |
999cd8a4 PM |
530 | /* |
531 | * Link to parent, and list entry in parent's children. | |
532 | * Protected by this->hierarchy_mutex and cgroup_lock() | |
533 | */ | |
534 | struct cgroupfs_root *root; | |
ddbcc7e8 | 535 | struct list_head sibling; |
38460b48 KH |
536 | /* used when use_id == true */ |
537 | struct idr idr; | |
c1e2ee2d | 538 | rwlock_t id_lock; |
e6a1105b BB |
539 | |
540 | /* should be defined only by modular subsystems */ | |
541 | struct module *module; | |
ddbcc7e8 PM |
542 | }; |
543 | ||
544 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _subsys; | |
545 | #include <linux/cgroup_subsys.h> | |
546 | #undef SUBSYS | |
547 | ||
548 | static inline struct cgroup_subsys_state *cgroup_subsys_state( | |
ffd2d883 | 549 | struct cgroup *cgrp, int subsys_id) |
ddbcc7e8 | 550 | { |
ffd2d883 | 551 | return cgrp->subsys[subsys_id]; |
ddbcc7e8 PM |
552 | } |
553 | ||
dc61b1d6 PZ |
554 | /* |
555 | * function to get the cgroup_subsys_state which allows for extra | |
556 | * rcu_dereference_check() conditions, such as locks used during the | |
557 | * cgroup_subsys::attach() methods. | |
558 | */ | |
559 | #define task_subsys_state_check(task, subsys_id, __c) \ | |
560 | rcu_dereference_check(task->cgroups->subsys[subsys_id], \ | |
dc61b1d6 PZ |
561 | lockdep_is_held(&task->alloc_lock) || \ |
562 | cgroup_lock_is_held() || (__c)) | |
563 | ||
564 | static inline struct cgroup_subsys_state * | |
565 | task_subsys_state(struct task_struct *task, int subsys_id) | |
ddbcc7e8 | 566 | { |
dc61b1d6 | 567 | return task_subsys_state_check(task, subsys_id, false); |
ddbcc7e8 PM |
568 | } |
569 | ||
570 | static inline struct cgroup* task_cgroup(struct task_struct *task, | |
571 | int subsys_id) | |
572 | { | |
573 | return task_subsys_state(task, subsys_id)->cgroup; | |
574 | } | |
575 | ||
817929ec PM |
576 | /* A cgroup_iter should be treated as an opaque object */ |
577 | struct cgroup_iter { | |
578 | struct list_head *cg_link; | |
579 | struct list_head *task; | |
580 | }; | |
581 | ||
d20a390a PM |
582 | /* |
583 | * To iterate across the tasks in a cgroup: | |
817929ec | 584 | * |
b595076a | 585 | * 1) call cgroup_iter_start to initialize an iterator |
817929ec PM |
586 | * |
587 | * 2) call cgroup_iter_next() to retrieve member tasks until it | |
588 | * returns NULL or until you want to end the iteration | |
589 | * | |
590 | * 3) call cgroup_iter_end() to destroy the iterator. | |
31a7df01 | 591 | * |
d20a390a PM |
592 | * Or, call cgroup_scan_tasks() to iterate through every task in a |
593 | * cgroup - cgroup_scan_tasks() holds the css_set_lock when calling | |
594 | * the test_task() callback, but not while calling the process_task() | |
595 | * callback. | |
817929ec | 596 | */ |
ffd2d883 LZ |
597 | void cgroup_iter_start(struct cgroup *cgrp, struct cgroup_iter *it); |
598 | struct task_struct *cgroup_iter_next(struct cgroup *cgrp, | |
817929ec | 599 | struct cgroup_iter *it); |
ffd2d883 | 600 | void cgroup_iter_end(struct cgroup *cgrp, struct cgroup_iter *it); |
31a7df01 | 601 | int cgroup_scan_tasks(struct cgroup_scanner *scan); |
956db3ca | 602 | int cgroup_attach_task(struct cgroup *, struct task_struct *); |
31583bb0 MT |
603 | int cgroup_attach_task_all(struct task_struct *from, struct task_struct *); |
604 | ||
605 | static inline int cgroup_attach_task_current_cg(struct task_struct *tsk) | |
606 | { | |
607 | return cgroup_attach_task_all(current, tsk); | |
608 | } | |
817929ec | 609 | |
38460b48 KH |
610 | /* |
611 | * CSS ID is ID for cgroup_subsys_state structs under subsys. This only works | |
612 | * if cgroup_subsys.use_id == true. It can be used for looking up and scanning. | |
613 | * CSS ID is assigned at cgroup allocation (create) automatically | |
614 | * and removed when subsys calls free_css_id() function. This is because | |
615 | * the lifetime of cgroup_subsys_state is subsys's matter. | |
616 | * | |
617 | * Looking up and scanning function should be called under rcu_read_lock(). | |
618 | * Taking cgroup_mutex()/hierarchy_mutex() is not necessary for following calls. | |
619 | * But the css returned by this routine can be "not populated yet" or "being | |
620 | * destroyed". The caller should check css and cgroup's status. | |
621 | */ | |
622 | ||
623 | /* | |
624 | * Typically Called at ->destroy(), or somewhere the subsys frees | |
625 | * cgroup_subsys_state. | |
626 | */ | |
627 | void free_css_id(struct cgroup_subsys *ss, struct cgroup_subsys_state *css); | |
628 | ||
629 | /* Find a cgroup_subsys_state which has given ID */ | |
630 | ||
631 | struct cgroup_subsys_state *css_lookup(struct cgroup_subsys *ss, int id); | |
632 | ||
633 | /* | |
634 | * Get a cgroup whose id is greater than or equal to id under tree of root. | |
635 | * Returning a cgroup_subsys_state or NULL. | |
636 | */ | |
637 | struct cgroup_subsys_state *css_get_next(struct cgroup_subsys *ss, int id, | |
638 | struct cgroup_subsys_state *root, int *foundid); | |
639 | ||
640 | /* Returns true if root is ancestor of cg */ | |
641 | bool css_is_ancestor(struct cgroup_subsys_state *cg, | |
0b7f569e | 642 | const struct cgroup_subsys_state *root); |
38460b48 KH |
643 | |
644 | /* Get id and depth of css */ | |
645 | unsigned short css_id(struct cgroup_subsys_state *css); | |
646 | unsigned short css_depth(struct cgroup_subsys_state *css); | |
e5d1367f | 647 | struct cgroup_subsys_state *cgroup_css_from_dir(struct file *f, int id); |
38460b48 | 648 | |
ddbcc7e8 PM |
649 | #else /* !CONFIG_CGROUPS */ |
650 | ||
651 | static inline int cgroup_init_early(void) { return 0; } | |
652 | static inline int cgroup_init(void) { return 0; } | |
b4f48b63 PM |
653 | static inline void cgroup_fork(struct task_struct *p) {} |
654 | static inline void cgroup_fork_callbacks(struct task_struct *p) {} | |
817929ec | 655 | static inline void cgroup_post_fork(struct task_struct *p) {} |
b4f48b63 | 656 | static inline void cgroup_exit(struct task_struct *p, int callbacks) {} |
ddbcc7e8 PM |
657 | |
658 | static inline void cgroup_lock(void) {} | |
659 | static inline void cgroup_unlock(void) {} | |
846c7bb0 BS |
660 | static inline int cgroupstats_build(struct cgroupstats *stats, |
661 | struct dentry *dentry) | |
662 | { | |
663 | return -EINVAL; | |
664 | } | |
ddbcc7e8 | 665 | |
d7926ee3 | 666 | /* No cgroups - nothing to do */ |
31583bb0 MT |
667 | static inline int cgroup_attach_task_all(struct task_struct *from, |
668 | struct task_struct *t) | |
669 | { | |
670 | return 0; | |
671 | } | |
d7926ee3 SS |
672 | static inline int cgroup_attach_task_current_cg(struct task_struct *t) |
673 | { | |
674 | return 0; | |
675 | } | |
676 | ||
ddbcc7e8 PM |
677 | #endif /* !CONFIG_CGROUPS */ |
678 | ||
679 | #endif /* _LINUX_CGROUP_H */ |