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 KH |
77 | /* ID for this css, if possible */ |
78 | struct css_id *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, | |
3116f0e3 PM |
157 | }; |
158 | ||
72a8cb30 BB |
159 | /* which pidlist file are we talking about? */ |
160 | enum cgroup_filetype { | |
161 | CGROUP_FILE_PROCS, | |
162 | CGROUP_FILE_TASKS, | |
163 | }; | |
164 | ||
165 | /* | |
166 | * A pidlist is a list of pids that virtually represents the contents of one | |
167 | * of the cgroup files ("procs" or "tasks"). We keep a list of such pidlists, | |
168 | * a pair (one each for procs, tasks) for each pid namespace that's relevant | |
169 | * to the cgroup. | |
170 | */ | |
102a775e | 171 | struct cgroup_pidlist { |
72a8cb30 BB |
172 | /* |
173 | * used to find which pidlist is wanted. doesn't change as long as | |
174 | * this particular list stays in the list. | |
175 | */ | |
176 | struct { enum cgroup_filetype type; struct pid_namespace *ns; } key; | |
102a775e BB |
177 | /* array of xids */ |
178 | pid_t *list; | |
179 | /* how many elements the above list has */ | |
180 | int length; | |
181 | /* how many files are using the current array */ | |
182 | int use_count; | |
72a8cb30 BB |
183 | /* each of these stored in a list by its cgroup */ |
184 | struct list_head links; | |
185 | /* pointer to the cgroup we belong to, for list removal purposes */ | |
186 | struct cgroup *owner; | |
187 | /* protects the other fields */ | |
188 | struct rw_semaphore mutex; | |
102a775e BB |
189 | }; |
190 | ||
ddbcc7e8 PM |
191 | struct cgroup { |
192 | unsigned long flags; /* "unsigned long" so bitops work */ | |
193 | ||
d20a390a PM |
194 | /* |
195 | * count users of this cgroup. >0 means busy, but doesn't | |
196 | * necessarily indicate the number of tasks in the cgroup | |
197 | */ | |
ddbcc7e8 PM |
198 | atomic_t count; |
199 | ||
200 | /* | |
201 | * We link our 'sibling' struct into our parent's 'children'. | |
202 | * Our children link their 'sibling' into our 'children'. | |
203 | */ | |
204 | struct list_head sibling; /* my parent's children */ | |
205 | struct list_head children; /* my children */ | |
206 | ||
d20a390a | 207 | struct cgroup *parent; /* my parent */ |
a47295e6 | 208 | struct dentry *dentry; /* cgroup fs entry, RCU protected */ |
ddbcc7e8 PM |
209 | |
210 | /* Private pointers for each registered subsystem */ | |
211 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
212 | ||
213 | struct cgroupfs_root *root; | |
214 | struct cgroup *top_cgroup; | |
817929ec PM |
215 | |
216 | /* | |
217 | * List of cg_cgroup_links pointing at css_sets with | |
218 | * tasks in this cgroup. Protected by css_set_lock | |
219 | */ | |
220 | struct list_head css_sets; | |
81a6a5cd PM |
221 | |
222 | /* | |
223 | * Linked list running through all cgroups that can | |
224 | * potentially be reaped by the release agent. Protected by | |
225 | * release_list_lock | |
226 | */ | |
227 | struct list_head release_list; | |
cc31edce | 228 | |
72a8cb30 BB |
229 | /* |
230 | * list of pidlists, up to two for each namespace (one for procs, one | |
231 | * for tasks); created on demand. | |
232 | */ | |
233 | struct list_head pidlists; | |
234 | struct mutex pidlist_mutex; | |
a47295e6 PM |
235 | |
236 | /* For RCU-protected deletion */ | |
237 | struct rcu_head rcu_head; | |
0dea1168 KS |
238 | |
239 | /* List of events which userspace want to recieve */ | |
240 | struct list_head event_list; | |
241 | spinlock_t event_list_lock; | |
817929ec PM |
242 | }; |
243 | ||
d20a390a PM |
244 | /* |
245 | * A css_set is a structure holding pointers to a set of | |
817929ec PM |
246 | * cgroup_subsys_state objects. This saves space in the task struct |
247 | * object and speeds up fork()/exit(), since a single inc/dec and a | |
d20a390a PM |
248 | * list_add()/del() can bump the reference count on the entire cgroup |
249 | * set for a task. | |
817929ec PM |
250 | */ |
251 | ||
252 | struct css_set { | |
253 | ||
254 | /* Reference count */ | |
146aa1bd | 255 | atomic_t refcount; |
817929ec | 256 | |
472b1053 LZ |
257 | /* |
258 | * List running through all cgroup groups in the same hash | |
259 | * slot. Protected by css_set_lock | |
260 | */ | |
261 | struct hlist_node hlist; | |
262 | ||
817929ec PM |
263 | /* |
264 | * List running through all tasks using this cgroup | |
265 | * group. Protected by css_set_lock | |
266 | */ | |
267 | struct list_head tasks; | |
268 | ||
269 | /* | |
270 | * List of cg_cgroup_link objects on link chains from | |
271 | * cgroups referenced from this css_set. Protected by | |
272 | * css_set_lock | |
273 | */ | |
274 | struct list_head cg_links; | |
275 | ||
276 | /* | |
277 | * Set of subsystem states, one for each subsystem. This array | |
278 | * is immutable after creation apart from the init_css_set | |
cf5d5941 BB |
279 | * during subsystem registration (at boot time) and modular subsystem |
280 | * loading/unloading. | |
817929ec PM |
281 | */ |
282 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
c378369d BB |
283 | |
284 | /* For RCU-protected deletion */ | |
285 | struct rcu_head rcu_head; | |
ddbcc7e8 PM |
286 | }; |
287 | ||
91796569 PM |
288 | /* |
289 | * cgroup_map_cb is an abstract callback API for reporting map-valued | |
290 | * control files | |
291 | */ | |
292 | ||
293 | struct cgroup_map_cb { | |
294 | int (*fill)(struct cgroup_map_cb *cb, const char *key, u64 value); | |
295 | void *state; | |
296 | }; | |
297 | ||
d20a390a PM |
298 | /* |
299 | * struct cftype: handler definitions for cgroup control files | |
ddbcc7e8 PM |
300 | * |
301 | * When reading/writing to a file: | |
a043e3b2 | 302 | * - the cgroup to use is file->f_dentry->d_parent->d_fsdata |
ddbcc7e8 PM |
303 | * - the 'cftype' of the file is file->f_dentry->d_fsdata |
304 | */ | |
305 | ||
306 | #define MAX_CFTYPE_NAME 64 | |
307 | struct cftype { | |
d20a390a PM |
308 | /* |
309 | * By convention, the name should begin with the name of the | |
310 | * subsystem, followed by a period | |
311 | */ | |
ddbcc7e8 PM |
312 | char name[MAX_CFTYPE_NAME]; |
313 | int private; | |
099fca32 LZ |
314 | /* |
315 | * If not 0, file mode is set to this value, otherwise it will | |
316 | * be figured out automatically | |
317 | */ | |
318 | mode_t mode; | |
db3b1497 PM |
319 | |
320 | /* | |
321 | * If non-zero, defines the maximum length of string that can | |
322 | * be passed to write_string; defaults to 64 | |
323 | */ | |
324 | size_t max_write_len; | |
325 | ||
ce16b49d PM |
326 | int (*open)(struct inode *inode, struct file *file); |
327 | ssize_t (*read)(struct cgroup *cgrp, struct cftype *cft, | |
328 | struct file *file, | |
329 | char __user *buf, size_t nbytes, loff_t *ppos); | |
ddbcc7e8 | 330 | /* |
f4c753b7 | 331 | * read_u64() is a shortcut for the common case of returning a |
ddbcc7e8 PM |
332 | * single integer. Use it in place of read() |
333 | */ | |
ce16b49d | 334 | u64 (*read_u64)(struct cgroup *cgrp, struct cftype *cft); |
e73d2c61 PM |
335 | /* |
336 | * read_s64() is a signed version of read_u64() | |
337 | */ | |
ce16b49d | 338 | s64 (*read_s64)(struct cgroup *cgrp, struct cftype *cft); |
91796569 PM |
339 | /* |
340 | * read_map() is used for defining a map of key/value | |
341 | * pairs. It should call cb->fill(cb, key, value) for each | |
342 | * entry. The key/value pairs (and their ordering) should not | |
343 | * change between reboots. | |
344 | */ | |
ce16b49d PM |
345 | int (*read_map)(struct cgroup *cont, struct cftype *cft, |
346 | struct cgroup_map_cb *cb); | |
29486df3 SH |
347 | /* |
348 | * read_seq_string() is used for outputting a simple sequence | |
349 | * using seqfile. | |
350 | */ | |
ce16b49d PM |
351 | int (*read_seq_string)(struct cgroup *cont, struct cftype *cft, |
352 | struct seq_file *m); | |
91796569 | 353 | |
ce16b49d PM |
354 | ssize_t (*write)(struct cgroup *cgrp, struct cftype *cft, |
355 | struct file *file, | |
356 | const char __user *buf, size_t nbytes, loff_t *ppos); | |
355e0c48 PM |
357 | |
358 | /* | |
f4c753b7 | 359 | * write_u64() is a shortcut for the common case of accepting |
355e0c48 PM |
360 | * a single integer (as parsed by simple_strtoull) from |
361 | * userspace. Use in place of write(); return 0 or error. | |
362 | */ | |
ce16b49d | 363 | int (*write_u64)(struct cgroup *cgrp, struct cftype *cft, u64 val); |
e73d2c61 PM |
364 | /* |
365 | * write_s64() is a signed version of write_u64() | |
366 | */ | |
ce16b49d | 367 | int (*write_s64)(struct cgroup *cgrp, struct cftype *cft, s64 val); |
355e0c48 | 368 | |
db3b1497 PM |
369 | /* |
370 | * write_string() is passed a nul-terminated kernelspace | |
371 | * buffer of maximum length determined by max_write_len. | |
372 | * Returns 0 or -ve error code. | |
373 | */ | |
374 | int (*write_string)(struct cgroup *cgrp, struct cftype *cft, | |
375 | const char *buffer); | |
d447ea2f PE |
376 | /* |
377 | * trigger() callback can be used to get some kick from the | |
378 | * userspace, when the actual string written is not important | |
379 | * at all. The private field can be used to determine the | |
380 | * kick type for multiplexing. | |
381 | */ | |
382 | int (*trigger)(struct cgroup *cgrp, unsigned int event); | |
383 | ||
ce16b49d | 384 | int (*release)(struct inode *inode, struct file *file); |
0dea1168 KS |
385 | |
386 | /* | |
387 | * register_event() callback will be used to add new userspace | |
388 | * waiter for changes related to the cftype. Implement it if | |
389 | * you want to provide this functionality. Use eventfd_signal() | |
390 | * on eventfd to send notification to userspace. | |
391 | */ | |
392 | int (*register_event)(struct cgroup *cgrp, struct cftype *cft, | |
393 | struct eventfd_ctx *eventfd, const char *args); | |
394 | /* | |
395 | * unregister_event() callback will be called when userspace | |
396 | * closes the eventfd or on cgroup removing. | |
397 | * This callback must be implemented, if you want provide | |
398 | * notification functionality. | |
0dea1168 | 399 | */ |
907860ed | 400 | void (*unregister_event)(struct cgroup *cgrp, struct cftype *cft, |
0dea1168 | 401 | struct eventfd_ctx *eventfd); |
ddbcc7e8 PM |
402 | }; |
403 | ||
31a7df01 CW |
404 | struct cgroup_scanner { |
405 | struct cgroup *cg; | |
406 | int (*test_task)(struct task_struct *p, struct cgroup_scanner *scan); | |
407 | void (*process_task)(struct task_struct *p, | |
408 | struct cgroup_scanner *scan); | |
409 | struct ptr_heap *heap; | |
bd1a8ab7 | 410 | void *data; |
31a7df01 CW |
411 | }; |
412 | ||
d20a390a PM |
413 | /* |
414 | * Add a new file to the given cgroup directory. Should only be | |
415 | * called by subsystems from within a populate() method | |
416 | */ | |
ffd2d883 | 417 | int cgroup_add_file(struct cgroup *cgrp, struct cgroup_subsys *subsys, |
ddbcc7e8 PM |
418 | const struct cftype *cft); |
419 | ||
d20a390a PM |
420 | /* |
421 | * Add a set of new files to the given cgroup directory. Should | |
422 | * only be called by subsystems from within a populate() method | |
423 | */ | |
ffd2d883 | 424 | int cgroup_add_files(struct cgroup *cgrp, |
ddbcc7e8 PM |
425 | struct cgroup_subsys *subsys, |
426 | const struct cftype cft[], | |
427 | int count); | |
428 | ||
ffd2d883 | 429 | int cgroup_is_removed(const struct cgroup *cgrp); |
ddbcc7e8 | 430 | |
ffd2d883 | 431 | int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen); |
ddbcc7e8 | 432 | |
ffd2d883 | 433 | int cgroup_task_count(const struct cgroup *cgrp); |
bbcb81d0 | 434 | |
313e924c GN |
435 | /* Return true if cgrp is a descendant of the task's cgroup */ |
436 | int cgroup_is_descendant(const struct cgroup *cgrp, struct task_struct *task); | |
ddbcc7e8 | 437 | |
88703267 KH |
438 | /* |
439 | * When the subsys has to access css and may add permanent refcnt to css, | |
440 | * it should take care of racy conditions with rmdir(). Following set of | |
441 | * functions, is for stop/restart rmdir if necessary. | |
442 | * Because these will call css_get/put, "css" should be alive css. | |
443 | * | |
444 | * cgroup_exclude_rmdir(); | |
445 | * ...do some jobs which may access arbitrary empty cgroup | |
446 | * cgroup_release_and_wakeup_rmdir(); | |
447 | * | |
448 | * When someone removes a cgroup while cgroup_exclude_rmdir() holds it, | |
449 | * it sleeps and cgroup_release_and_wakeup_rmdir() will wake him up. | |
450 | */ | |
451 | ||
452 | void cgroup_exclude_rmdir(struct cgroup_subsys_state *css); | |
453 | void cgroup_release_and_wakeup_rmdir(struct cgroup_subsys_state *css); | |
454 | ||
21acb9ca TLSC |
455 | /* |
456 | * Control Group subsystem type. | |
457 | * See Documentation/cgroups/cgroups.txt for details | |
458 | */ | |
ddbcc7e8 PM |
459 | |
460 | struct cgroup_subsys { | |
461 | struct cgroup_subsys_state *(*create)(struct cgroup_subsys *ss, | |
ffd2d883 | 462 | struct cgroup *cgrp); |
ec64f515 | 463 | int (*pre_destroy)(struct cgroup_subsys *ss, struct cgroup *cgrp); |
ffd2d883 | 464 | void (*destroy)(struct cgroup_subsys *ss, struct cgroup *cgrp); |
be367d09 BB |
465 | int (*can_attach)(struct cgroup_subsys *ss, struct cgroup *cgrp, |
466 | struct task_struct *tsk, bool threadgroup); | |
2468c723 DN |
467 | void (*cancel_attach)(struct cgroup_subsys *ss, struct cgroup *cgrp, |
468 | struct task_struct *tsk, bool threadgroup); | |
ffd2d883 | 469 | void (*attach)(struct cgroup_subsys *ss, struct cgroup *cgrp, |
be367d09 BB |
470 | struct cgroup *old_cgrp, struct task_struct *tsk, |
471 | bool threadgroup); | |
ddbcc7e8 PM |
472 | void (*fork)(struct cgroup_subsys *ss, struct task_struct *task); |
473 | void (*exit)(struct cgroup_subsys *ss, struct task_struct *task); | |
474 | int (*populate)(struct cgroup_subsys *ss, | |
ffd2d883 LZ |
475 | struct cgroup *cgrp); |
476 | void (*post_clone)(struct cgroup_subsys *ss, struct cgroup *cgrp); | |
ddbcc7e8 | 477 | void (*bind)(struct cgroup_subsys *ss, struct cgroup *root); |
e5991371 | 478 | |
ddbcc7e8 PM |
479 | int subsys_id; |
480 | int active; | |
8bab8dde | 481 | int disabled; |
ddbcc7e8 | 482 | int early_init; |
38460b48 KH |
483 | /* |
484 | * True if this subsys uses ID. ID is not available before cgroup_init() | |
485 | * (not available in early_init time.) | |
486 | */ | |
487 | bool use_id; | |
ddbcc7e8 PM |
488 | #define MAX_CGROUP_TYPE_NAMELEN 32 |
489 | const char *name; | |
490 | ||
999cd8a4 PM |
491 | /* |
492 | * Protects sibling/children links of cgroups in this | |
493 | * hierarchy, plus protects which hierarchy (or none) the | |
494 | * subsystem is a part of (i.e. root/sibling). To avoid | |
495 | * potential deadlocks, the following operations should not be | |
496 | * undertaken while holding any hierarchy_mutex: | |
497 | * | |
498 | * - allocating memory | |
499 | * - initiating hotplug events | |
500 | */ | |
501 | struct mutex hierarchy_mutex; | |
cfebe563 | 502 | struct lock_class_key subsys_key; |
ddbcc7e8 | 503 | |
999cd8a4 PM |
504 | /* |
505 | * Link to parent, and list entry in parent's children. | |
506 | * Protected by this->hierarchy_mutex and cgroup_lock() | |
507 | */ | |
508 | struct cgroupfs_root *root; | |
ddbcc7e8 | 509 | struct list_head sibling; |
38460b48 KH |
510 | /* used when use_id == true */ |
511 | struct idr idr; | |
512 | spinlock_t id_lock; | |
e6a1105b BB |
513 | |
514 | /* should be defined only by modular subsystems */ | |
515 | struct module *module; | |
ddbcc7e8 PM |
516 | }; |
517 | ||
518 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _subsys; | |
519 | #include <linux/cgroup_subsys.h> | |
520 | #undef SUBSYS | |
521 | ||
522 | static inline struct cgroup_subsys_state *cgroup_subsys_state( | |
ffd2d883 | 523 | struct cgroup *cgrp, int subsys_id) |
ddbcc7e8 | 524 | { |
ffd2d883 | 525 | return cgrp->subsys[subsys_id]; |
ddbcc7e8 PM |
526 | } |
527 | ||
dc61b1d6 PZ |
528 | /* |
529 | * function to get the cgroup_subsys_state which allows for extra | |
530 | * rcu_dereference_check() conditions, such as locks used during the | |
531 | * cgroup_subsys::attach() methods. | |
532 | */ | |
533 | #define task_subsys_state_check(task, subsys_id, __c) \ | |
534 | rcu_dereference_check(task->cgroups->subsys[subsys_id], \ | |
535 | rcu_read_lock_held() || \ | |
536 | lockdep_is_held(&task->alloc_lock) || \ | |
537 | cgroup_lock_is_held() || (__c)) | |
538 | ||
539 | static inline struct cgroup_subsys_state * | |
540 | task_subsys_state(struct task_struct *task, int subsys_id) | |
ddbcc7e8 | 541 | { |
dc61b1d6 | 542 | return task_subsys_state_check(task, subsys_id, false); |
ddbcc7e8 PM |
543 | } |
544 | ||
545 | static inline struct cgroup* task_cgroup(struct task_struct *task, | |
546 | int subsys_id) | |
547 | { | |
548 | return task_subsys_state(task, subsys_id)->cgroup; | |
549 | } | |
550 | ||
e885dcde SH |
551 | int cgroup_clone(struct task_struct *tsk, struct cgroup_subsys *ss, |
552 | char *nodename); | |
697f4161 | 553 | |
817929ec PM |
554 | /* A cgroup_iter should be treated as an opaque object */ |
555 | struct cgroup_iter { | |
556 | struct list_head *cg_link; | |
557 | struct list_head *task; | |
558 | }; | |
559 | ||
d20a390a PM |
560 | /* |
561 | * To iterate across the tasks in a cgroup: | |
817929ec PM |
562 | * |
563 | * 1) call cgroup_iter_start to intialize an iterator | |
564 | * | |
565 | * 2) call cgroup_iter_next() to retrieve member tasks until it | |
566 | * returns NULL or until you want to end the iteration | |
567 | * | |
568 | * 3) call cgroup_iter_end() to destroy the iterator. | |
31a7df01 | 569 | * |
d20a390a PM |
570 | * Or, call cgroup_scan_tasks() to iterate through every task in a |
571 | * cgroup - cgroup_scan_tasks() holds the css_set_lock when calling | |
572 | * the test_task() callback, but not while calling the process_task() | |
573 | * callback. | |
817929ec | 574 | */ |
ffd2d883 LZ |
575 | void cgroup_iter_start(struct cgroup *cgrp, struct cgroup_iter *it); |
576 | struct task_struct *cgroup_iter_next(struct cgroup *cgrp, | |
817929ec | 577 | struct cgroup_iter *it); |
ffd2d883 | 578 | void cgroup_iter_end(struct cgroup *cgrp, struct cgroup_iter *it); |
31a7df01 | 579 | int cgroup_scan_tasks(struct cgroup_scanner *scan); |
956db3ca | 580 | int cgroup_attach_task(struct cgroup *, struct task_struct *); |
d7926ee3 | 581 | int cgroup_attach_task_current_cg(struct task_struct *); |
817929ec | 582 | |
38460b48 KH |
583 | /* |
584 | * CSS ID is ID for cgroup_subsys_state structs under subsys. This only works | |
585 | * if cgroup_subsys.use_id == true. It can be used for looking up and scanning. | |
586 | * CSS ID is assigned at cgroup allocation (create) automatically | |
587 | * and removed when subsys calls free_css_id() function. This is because | |
588 | * the lifetime of cgroup_subsys_state is subsys's matter. | |
589 | * | |
590 | * Looking up and scanning function should be called under rcu_read_lock(). | |
591 | * Taking cgroup_mutex()/hierarchy_mutex() is not necessary for following calls. | |
592 | * But the css returned by this routine can be "not populated yet" or "being | |
593 | * destroyed". The caller should check css and cgroup's status. | |
594 | */ | |
595 | ||
596 | /* | |
597 | * Typically Called at ->destroy(), or somewhere the subsys frees | |
598 | * cgroup_subsys_state. | |
599 | */ | |
600 | void free_css_id(struct cgroup_subsys *ss, struct cgroup_subsys_state *css); | |
601 | ||
602 | /* Find a cgroup_subsys_state which has given ID */ | |
603 | ||
604 | struct cgroup_subsys_state *css_lookup(struct cgroup_subsys *ss, int id); | |
605 | ||
606 | /* | |
607 | * Get a cgroup whose id is greater than or equal to id under tree of root. | |
608 | * Returning a cgroup_subsys_state or NULL. | |
609 | */ | |
610 | struct cgroup_subsys_state *css_get_next(struct cgroup_subsys *ss, int id, | |
611 | struct cgroup_subsys_state *root, int *foundid); | |
612 | ||
613 | /* Returns true if root is ancestor of cg */ | |
614 | bool css_is_ancestor(struct cgroup_subsys_state *cg, | |
0b7f569e | 615 | const struct cgroup_subsys_state *root); |
38460b48 KH |
616 | |
617 | /* Get id and depth of css */ | |
618 | unsigned short css_id(struct cgroup_subsys_state *css); | |
619 | unsigned short css_depth(struct cgroup_subsys_state *css); | |
620 | ||
ddbcc7e8 PM |
621 | #else /* !CONFIG_CGROUPS */ |
622 | ||
623 | static inline int cgroup_init_early(void) { return 0; } | |
624 | static inline int cgroup_init(void) { return 0; } | |
b4f48b63 PM |
625 | static inline void cgroup_fork(struct task_struct *p) {} |
626 | static inline void cgroup_fork_callbacks(struct task_struct *p) {} | |
817929ec | 627 | static inline void cgroup_post_fork(struct task_struct *p) {} |
b4f48b63 | 628 | static inline void cgroup_exit(struct task_struct *p, int callbacks) {} |
ddbcc7e8 PM |
629 | |
630 | static inline void cgroup_lock(void) {} | |
631 | static inline void cgroup_unlock(void) {} | |
846c7bb0 BS |
632 | static inline int cgroupstats_build(struct cgroupstats *stats, |
633 | struct dentry *dentry) | |
634 | { | |
635 | return -EINVAL; | |
636 | } | |
ddbcc7e8 | 637 | |
d7926ee3 SS |
638 | /* No cgroups - nothing to do */ |
639 | static inline int cgroup_attach_task_current_cg(struct task_struct *t) | |
640 | { | |
641 | return 0; | |
642 | } | |
643 | ||
ddbcc7e8 PM |
644 | #endif /* !CONFIG_CGROUPS */ |
645 | ||
646 | #endif /* _LINUX_CGROUP_H */ |