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> |
48ddbe19 | 19 | #include <linux/workqueue.h> |
ddbcc7e8 PM |
20 | |
21 | #ifdef CONFIG_CGROUPS | |
22 | ||
23 | struct cgroupfs_root; | |
24 | struct cgroup_subsys; | |
25 | struct inode; | |
84eea842 | 26 | struct cgroup; |
38460b48 | 27 | struct css_id; |
ddbcc7e8 PM |
28 | |
29 | extern int cgroup_init_early(void); | |
30 | extern int cgroup_init(void); | |
ddbcc7e8 | 31 | extern void cgroup_lock(void); |
d11c563d | 32 | extern int cgroup_lock_is_held(void); |
84eea842 | 33 | extern bool cgroup_lock_live_group(struct cgroup *cgrp); |
ddbcc7e8 | 34 | extern void cgroup_unlock(void); |
b4f48b63 PM |
35 | extern void cgroup_fork(struct task_struct *p); |
36 | extern void cgroup_fork_callbacks(struct task_struct *p); | |
817929ec | 37 | extern void cgroup_post_fork(struct task_struct *p); |
b4f48b63 | 38 | extern void cgroup_exit(struct task_struct *p, int run_callbacks); |
846c7bb0 BS |
39 | extern int cgroupstats_build(struct cgroupstats *stats, |
40 | struct dentry *dentry); | |
e6a1105b | 41 | extern int cgroup_load_subsys(struct cgroup_subsys *ss); |
cf5d5941 | 42 | extern void cgroup_unload_subsys(struct cgroup_subsys *ss); |
ddbcc7e8 | 43 | |
828c0950 | 44 | extern const struct file_operations proc_cgroup_operations; |
a424316c | 45 | |
aae8aab4 | 46 | /* Define the enumeration of all builtin cgroup subsystems */ |
817929ec PM |
47 | #define SUBSYS(_x) _x ## _subsys_id, |
48 | enum cgroup_subsys_id { | |
49 | #include <linux/cgroup_subsys.h> | |
aae8aab4 | 50 | CGROUP_BUILTIN_SUBSYS_COUNT |
817929ec PM |
51 | }; |
52 | #undef SUBSYS | |
aae8aab4 BB |
53 | /* |
54 | * This define indicates the maximum number of subsystems that can be loaded | |
55 | * at once. We limit to this many since cgroupfs_root has subsys_bits to keep | |
56 | * track of all of them. | |
57 | */ | |
58 | #define CGROUP_SUBSYS_COUNT (BITS_PER_BYTE*sizeof(unsigned long)) | |
817929ec | 59 | |
ddbcc7e8 PM |
60 | /* Per-subsystem/per-cgroup state maintained by the system. */ |
61 | struct cgroup_subsys_state { | |
d20a390a PM |
62 | /* |
63 | * The cgroup that this subsystem is attached to. Useful | |
ddbcc7e8 | 64 | * for subsystems that want to know about the cgroup |
d20a390a PM |
65 | * hierarchy structure |
66 | */ | |
ddbcc7e8 PM |
67 | struct cgroup *cgroup; |
68 | ||
d20a390a PM |
69 | /* |
70 | * State maintained by the cgroup system to allow subsystems | |
e7c5ec91 | 71 | * to be "busy". Should be accessed via css_get(), |
d20a390a PM |
72 | * css_tryget() and and css_put(). |
73 | */ | |
ddbcc7e8 PM |
74 | |
75 | atomic_t refcnt; | |
76 | ||
77 | unsigned long flags; | |
38460b48 | 78 | /* ID for this css, if possible */ |
2c392b8c | 79 | struct css_id __rcu *id; |
48ddbe19 TH |
80 | |
81 | /* Used to put @cgroup->dentry on the last css_put() */ | |
82 | struct work_struct dput_work; | |
ddbcc7e8 PM |
83 | }; |
84 | ||
85 | /* bits in struct cgroup_subsys_state flags field */ | |
86 | enum { | |
87 | CSS_ROOT, /* This CSS is the root of the subsystem */ | |
e7c5ec91 | 88 | CSS_REMOVED, /* This CSS is dead */ |
48ddbe19 | 89 | CSS_CLEAR_CSS_REFS, /* @ss->__DEPRECATED_clear_css_refs */ |
ddbcc7e8 PM |
90 | }; |
91 | ||
d7b9fff7 DN |
92 | /* Caller must verify that the css is not for root cgroup */ |
93 | static inline void __css_get(struct cgroup_subsys_state *css, int count) | |
94 | { | |
95 | atomic_add(count, &css->refcnt); | |
96 | } | |
97 | ||
ddbcc7e8 | 98 | /* |
e7c5ec91 PM |
99 | * Call css_get() to hold a reference on the css; it can be used |
100 | * for a reference obtained via: | |
101 | * - an existing ref-counted reference to the css | |
102 | * - task->cgroups for a locked task | |
ddbcc7e8 PM |
103 | */ |
104 | ||
105 | static inline void css_get(struct cgroup_subsys_state *css) | |
106 | { | |
107 | /* We don't need to reference count the root state */ | |
108 | if (!test_bit(CSS_ROOT, &css->flags)) | |
d7b9fff7 | 109 | __css_get(css, 1); |
ddbcc7e8 | 110 | } |
e7c5ec91 PM |
111 | |
112 | static inline bool css_is_removed(struct cgroup_subsys_state *css) | |
113 | { | |
114 | return test_bit(CSS_REMOVED, &css->flags); | |
115 | } | |
116 | ||
117 | /* | |
118 | * Call css_tryget() to take a reference on a css if your existing | |
119 | * (known-valid) reference isn't already ref-counted. Returns false if | |
120 | * the css has been destroyed. | |
121 | */ | |
122 | ||
28b4c27b | 123 | extern bool __css_tryget(struct cgroup_subsys_state *css); |
e7c5ec91 PM |
124 | static inline bool css_tryget(struct cgroup_subsys_state *css) |
125 | { | |
126 | if (test_bit(CSS_ROOT, &css->flags)) | |
127 | return true; | |
28b4c27b | 128 | return __css_tryget(css); |
e7c5ec91 PM |
129 | } |
130 | ||
ddbcc7e8 PM |
131 | /* |
132 | * css_put() should be called to release a reference taken by | |
e7c5ec91 | 133 | * css_get() or css_tryget() |
ddbcc7e8 PM |
134 | */ |
135 | ||
28b4c27b | 136 | extern void __css_put(struct cgroup_subsys_state *css); |
ddbcc7e8 PM |
137 | static inline void css_put(struct cgroup_subsys_state *css) |
138 | { | |
139 | if (!test_bit(CSS_ROOT, &css->flags)) | |
28b4c27b | 140 | __css_put(css); |
ddbcc7e8 PM |
141 | } |
142 | ||
3116f0e3 PM |
143 | /* bits in struct cgroup flags field */ |
144 | enum { | |
145 | /* Control Group is dead */ | |
146 | CGRP_REMOVED, | |
d20a390a PM |
147 | /* |
148 | * Control Group has previously had a child cgroup or a task, | |
149 | * but no longer (only if CGRP_NOTIFY_ON_RELEASE is set) | |
150 | */ | |
3116f0e3 PM |
151 | CGRP_RELEASABLE, |
152 | /* Control Group requires release notifications to userspace */ | |
153 | CGRP_NOTIFY_ON_RELEASE, | |
ec64f515 KH |
154 | /* |
155 | * A thread in rmdir() is wating for this cgroup. | |
156 | */ | |
157 | CGRP_WAIT_ON_RMDIR, | |
97978e6d DL |
158 | /* |
159 | * Clone cgroup values when creating a new child cgroup | |
160 | */ | |
161 | CGRP_CLONE_CHILDREN, | |
3116f0e3 PM |
162 | }; |
163 | ||
ddbcc7e8 PM |
164 | struct cgroup { |
165 | unsigned long flags; /* "unsigned long" so bitops work */ | |
166 | ||
d20a390a PM |
167 | /* |
168 | * count users of this cgroup. >0 means busy, but doesn't | |
169 | * necessarily indicate the number of tasks in the cgroup | |
170 | */ | |
ddbcc7e8 PM |
171 | atomic_t count; |
172 | ||
173 | /* | |
174 | * We link our 'sibling' struct into our parent's 'children'. | |
175 | * Our children link their 'sibling' into our 'children'. | |
176 | */ | |
177 | struct list_head sibling; /* my parent's children */ | |
178 | struct list_head children; /* my children */ | |
05ef1d7c | 179 | struct list_head files; /* my files */ |
ddbcc7e8 | 180 | |
d20a390a | 181 | struct cgroup *parent; /* my parent */ |
2c392b8c | 182 | struct dentry __rcu *dentry; /* cgroup fs entry, RCU protected */ |
ddbcc7e8 PM |
183 | |
184 | /* Private pointers for each registered subsystem */ | |
185 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
186 | ||
187 | struct cgroupfs_root *root; | |
188 | struct cgroup *top_cgroup; | |
817929ec PM |
189 | |
190 | /* | |
191 | * List of cg_cgroup_links pointing at css_sets with | |
192 | * tasks in this cgroup. Protected by css_set_lock | |
193 | */ | |
194 | struct list_head css_sets; | |
81a6a5cd | 195 | |
b0ca5a84 | 196 | struct list_head allcg_node; /* cgroupfs_root->allcg_list */ |
8e3f6541 | 197 | struct list_head cft_q_node; /* used during cftype add/rm */ |
b0ca5a84 | 198 | |
81a6a5cd PM |
199 | /* |
200 | * Linked list running through all cgroups that can | |
201 | * potentially be reaped by the release agent. Protected by | |
202 | * release_list_lock | |
203 | */ | |
204 | struct list_head release_list; | |
cc31edce | 205 | |
72a8cb30 BB |
206 | /* |
207 | * list of pidlists, up to two for each namespace (one for procs, one | |
208 | * for tasks); created on demand. | |
209 | */ | |
210 | struct list_head pidlists; | |
211 | struct mutex pidlist_mutex; | |
a47295e6 PM |
212 | |
213 | /* For RCU-protected deletion */ | |
214 | struct rcu_head rcu_head; | |
0dea1168 | 215 | |
25985edc | 216 | /* List of events which userspace want to receive */ |
0dea1168 KS |
217 | struct list_head event_list; |
218 | spinlock_t event_list_lock; | |
817929ec PM |
219 | }; |
220 | ||
d20a390a PM |
221 | /* |
222 | * A css_set is a structure holding pointers to a set of | |
817929ec PM |
223 | * cgroup_subsys_state objects. This saves space in the task struct |
224 | * object and speeds up fork()/exit(), since a single inc/dec and a | |
d20a390a PM |
225 | * list_add()/del() can bump the reference count on the entire cgroup |
226 | * set for a task. | |
817929ec PM |
227 | */ |
228 | ||
229 | struct css_set { | |
230 | ||
231 | /* Reference count */ | |
146aa1bd | 232 | atomic_t refcount; |
817929ec | 233 | |
472b1053 LZ |
234 | /* |
235 | * List running through all cgroup groups in the same hash | |
236 | * slot. Protected by css_set_lock | |
237 | */ | |
238 | struct hlist_node hlist; | |
239 | ||
817929ec PM |
240 | /* |
241 | * List running through all tasks using this cgroup | |
242 | * group. Protected by css_set_lock | |
243 | */ | |
244 | struct list_head tasks; | |
245 | ||
246 | /* | |
247 | * List of cg_cgroup_link objects on link chains from | |
248 | * cgroups referenced from this css_set. Protected by | |
249 | * css_set_lock | |
250 | */ | |
251 | struct list_head cg_links; | |
252 | ||
253 | /* | |
254 | * Set of subsystem states, one for each subsystem. This array | |
255 | * is immutable after creation apart from the init_css_set | |
cf5d5941 BB |
256 | * during subsystem registration (at boot time) and modular subsystem |
257 | * loading/unloading. | |
817929ec PM |
258 | */ |
259 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
c378369d BB |
260 | |
261 | /* For RCU-protected deletion */ | |
262 | struct rcu_head rcu_head; | |
ddbcc7e8 PM |
263 | }; |
264 | ||
91796569 PM |
265 | /* |
266 | * cgroup_map_cb is an abstract callback API for reporting map-valued | |
267 | * control files | |
268 | */ | |
269 | ||
270 | struct cgroup_map_cb { | |
271 | int (*fill)(struct cgroup_map_cb *cb, const char *key, u64 value); | |
272 | void *state; | |
273 | }; | |
274 | ||
d20a390a PM |
275 | /* |
276 | * struct cftype: handler definitions for cgroup control files | |
ddbcc7e8 PM |
277 | * |
278 | * When reading/writing to a file: | |
a043e3b2 | 279 | * - the cgroup to use is file->f_dentry->d_parent->d_fsdata |
ddbcc7e8 PM |
280 | * - the 'cftype' of the file is file->f_dentry->d_fsdata |
281 | */ | |
282 | ||
8e3f6541 TH |
283 | /* cftype->flags */ |
284 | #define CFTYPE_ONLY_ON_ROOT (1U << 0) /* only create on root cg */ | |
285 | #define CFTYPE_NOT_ON_ROOT (1U << 1) /* don't create onp root cg */ | |
286 | ||
287 | #define MAX_CFTYPE_NAME 64 | |
288 | ||
ddbcc7e8 | 289 | struct cftype { |
d20a390a PM |
290 | /* |
291 | * By convention, the name should begin with the name of the | |
8e3f6541 TH |
292 | * subsystem, followed by a period. Zero length string indicates |
293 | * end of cftype array. | |
d20a390a | 294 | */ |
ddbcc7e8 PM |
295 | char name[MAX_CFTYPE_NAME]; |
296 | int private; | |
099fca32 LZ |
297 | /* |
298 | * If not 0, file mode is set to this value, otherwise it will | |
299 | * be figured out automatically | |
300 | */ | |
a5e7ed32 | 301 | umode_t mode; |
db3b1497 PM |
302 | |
303 | /* | |
304 | * If non-zero, defines the maximum length of string that can | |
305 | * be passed to write_string; defaults to 64 | |
306 | */ | |
307 | size_t max_write_len; | |
308 | ||
8e3f6541 TH |
309 | /* CFTYPE_* flags */ |
310 | unsigned int flags; | |
311 | ||
ce16b49d PM |
312 | int (*open)(struct inode *inode, struct file *file); |
313 | ssize_t (*read)(struct cgroup *cgrp, struct cftype *cft, | |
314 | struct file *file, | |
315 | char __user *buf, size_t nbytes, loff_t *ppos); | |
ddbcc7e8 | 316 | /* |
f4c753b7 | 317 | * read_u64() is a shortcut for the common case of returning a |
ddbcc7e8 PM |
318 | * single integer. Use it in place of read() |
319 | */ | |
ce16b49d | 320 | u64 (*read_u64)(struct cgroup *cgrp, struct cftype *cft); |
e73d2c61 PM |
321 | /* |
322 | * read_s64() is a signed version of read_u64() | |
323 | */ | |
ce16b49d | 324 | s64 (*read_s64)(struct cgroup *cgrp, struct cftype *cft); |
91796569 PM |
325 | /* |
326 | * read_map() is used for defining a map of key/value | |
327 | * pairs. It should call cb->fill(cb, key, value) for each | |
328 | * entry. The key/value pairs (and their ordering) should not | |
329 | * change between reboots. | |
330 | */ | |
ce16b49d PM |
331 | int (*read_map)(struct cgroup *cont, struct cftype *cft, |
332 | struct cgroup_map_cb *cb); | |
29486df3 SH |
333 | /* |
334 | * read_seq_string() is used for outputting a simple sequence | |
335 | * using seqfile. | |
336 | */ | |
ce16b49d PM |
337 | int (*read_seq_string)(struct cgroup *cont, struct cftype *cft, |
338 | struct seq_file *m); | |
91796569 | 339 | |
ce16b49d PM |
340 | ssize_t (*write)(struct cgroup *cgrp, struct cftype *cft, |
341 | struct file *file, | |
342 | const char __user *buf, size_t nbytes, loff_t *ppos); | |
355e0c48 PM |
343 | |
344 | /* | |
f4c753b7 | 345 | * write_u64() is a shortcut for the common case of accepting |
355e0c48 PM |
346 | * a single integer (as parsed by simple_strtoull) from |
347 | * userspace. Use in place of write(); return 0 or error. | |
348 | */ | |
ce16b49d | 349 | int (*write_u64)(struct cgroup *cgrp, struct cftype *cft, u64 val); |
e73d2c61 PM |
350 | /* |
351 | * write_s64() is a signed version of write_u64() | |
352 | */ | |
ce16b49d | 353 | int (*write_s64)(struct cgroup *cgrp, struct cftype *cft, s64 val); |
355e0c48 | 354 | |
db3b1497 PM |
355 | /* |
356 | * write_string() is passed a nul-terminated kernelspace | |
357 | * buffer of maximum length determined by max_write_len. | |
358 | * Returns 0 or -ve error code. | |
359 | */ | |
360 | int (*write_string)(struct cgroup *cgrp, struct cftype *cft, | |
361 | const char *buffer); | |
d447ea2f PE |
362 | /* |
363 | * trigger() callback can be used to get some kick from the | |
364 | * userspace, when the actual string written is not important | |
365 | * at all. The private field can be used to determine the | |
366 | * kick type for multiplexing. | |
367 | */ | |
368 | int (*trigger)(struct cgroup *cgrp, unsigned int event); | |
369 | ||
ce16b49d | 370 | int (*release)(struct inode *inode, struct file *file); |
0dea1168 KS |
371 | |
372 | /* | |
373 | * register_event() callback will be used to add new userspace | |
374 | * waiter for changes related to the cftype. Implement it if | |
375 | * you want to provide this functionality. Use eventfd_signal() | |
376 | * on eventfd to send notification to userspace. | |
377 | */ | |
378 | int (*register_event)(struct cgroup *cgrp, struct cftype *cft, | |
379 | struct eventfd_ctx *eventfd, const char *args); | |
380 | /* | |
381 | * unregister_event() callback will be called when userspace | |
382 | * closes the eventfd or on cgroup removing. | |
383 | * This callback must be implemented, if you want provide | |
384 | * notification functionality. | |
0dea1168 | 385 | */ |
907860ed | 386 | void (*unregister_event)(struct cgroup *cgrp, struct cftype *cft, |
0dea1168 | 387 | struct eventfd_ctx *eventfd); |
ddbcc7e8 PM |
388 | }; |
389 | ||
8e3f6541 TH |
390 | /* |
391 | * cftype_sets describe cftypes belonging to a subsystem and are chained at | |
392 | * cgroup_subsys->cftsets. Each cftset points to an array of cftypes | |
393 | * terminated by zero length name. | |
394 | */ | |
395 | struct cftype_set { | |
396 | struct list_head node; /* chained at subsys->cftsets */ | |
397 | const struct cftype *cfts; | |
398 | }; | |
399 | ||
31a7df01 CW |
400 | struct cgroup_scanner { |
401 | struct cgroup *cg; | |
402 | int (*test_task)(struct task_struct *p, struct cgroup_scanner *scan); | |
403 | void (*process_task)(struct task_struct *p, | |
404 | struct cgroup_scanner *scan); | |
405 | struct ptr_heap *heap; | |
bd1a8ab7 | 406 | void *data; |
31a7df01 CW |
407 | }; |
408 | ||
8e3f6541 | 409 | int cgroup_add_cftypes(struct cgroup_subsys *ss, const struct cftype *cfts); |
79578621 | 410 | int cgroup_rm_cftypes(struct cgroup_subsys *ss, const struct cftype *cfts); |
8e3f6541 | 411 | |
ffd2d883 | 412 | int cgroup_is_removed(const struct cgroup *cgrp); |
ddbcc7e8 | 413 | |
ffd2d883 | 414 | int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen); |
ddbcc7e8 | 415 | |
ffd2d883 | 416 | int cgroup_task_count(const struct cgroup *cgrp); |
bbcb81d0 | 417 | |
313e924c GN |
418 | /* Return true if cgrp is a descendant of the task's cgroup */ |
419 | int cgroup_is_descendant(const struct cgroup *cgrp, struct task_struct *task); | |
ddbcc7e8 | 420 | |
88703267 KH |
421 | /* |
422 | * When the subsys has to access css and may add permanent refcnt to css, | |
423 | * it should take care of racy conditions with rmdir(). Following set of | |
424 | * functions, is for stop/restart rmdir if necessary. | |
425 | * Because these will call css_get/put, "css" should be alive css. | |
426 | * | |
427 | * cgroup_exclude_rmdir(); | |
428 | * ...do some jobs which may access arbitrary empty cgroup | |
429 | * cgroup_release_and_wakeup_rmdir(); | |
430 | * | |
431 | * When someone removes a cgroup while cgroup_exclude_rmdir() holds it, | |
432 | * it sleeps and cgroup_release_and_wakeup_rmdir() will wake him up. | |
433 | */ | |
434 | ||
435 | void cgroup_exclude_rmdir(struct cgroup_subsys_state *css); | |
436 | void cgroup_release_and_wakeup_rmdir(struct cgroup_subsys_state *css); | |
437 | ||
2f7ee569 TH |
438 | /* |
439 | * Control Group taskset, used to pass around set of tasks to cgroup_subsys | |
440 | * methods. | |
441 | */ | |
442 | struct cgroup_taskset; | |
443 | struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset); | |
444 | struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset); | |
445 | struct cgroup *cgroup_taskset_cur_cgroup(struct cgroup_taskset *tset); | |
446 | int cgroup_taskset_size(struct cgroup_taskset *tset); | |
447 | ||
448 | /** | |
449 | * cgroup_taskset_for_each - iterate cgroup_taskset | |
450 | * @task: the loop cursor | |
451 | * @skip_cgrp: skip if task's cgroup matches this, %NULL to iterate through all | |
452 | * @tset: taskset to iterate | |
453 | */ | |
454 | #define cgroup_taskset_for_each(task, skip_cgrp, tset) \ | |
455 | for ((task) = cgroup_taskset_first((tset)); (task); \ | |
456 | (task) = cgroup_taskset_next((tset))) \ | |
457 | if (!(skip_cgrp) || \ | |
458 | cgroup_taskset_cur_cgroup((tset)) != (skip_cgrp)) | |
459 | ||
21acb9ca TLSC |
460 | /* |
461 | * Control Group subsystem type. | |
462 | * See Documentation/cgroups/cgroups.txt for details | |
463 | */ | |
ddbcc7e8 PM |
464 | |
465 | struct cgroup_subsys { | |
761b3ef5 LZ |
466 | struct cgroup_subsys_state *(*create)(struct cgroup *cgrp); |
467 | int (*pre_destroy)(struct cgroup *cgrp); | |
468 | void (*destroy)(struct cgroup *cgrp); | |
469 | int (*can_attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); | |
470 | void (*cancel_attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); | |
471 | void (*attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); | |
472 | void (*fork)(struct task_struct *task); | |
473 | void (*exit)(struct cgroup *cgrp, struct cgroup *old_cgrp, | |
474 | struct task_struct *task); | |
761b3ef5 LZ |
475 | void (*post_clone)(struct cgroup *cgrp); |
476 | void (*bind)(struct cgroup *root); | |
e5991371 | 477 | |
ddbcc7e8 PM |
478 | int subsys_id; |
479 | int active; | |
8bab8dde | 480 | int disabled; |
ddbcc7e8 | 481 | int early_init; |
38460b48 KH |
482 | /* |
483 | * True if this subsys uses ID. ID is not available before cgroup_init() | |
484 | * (not available in early_init time.) | |
485 | */ | |
486 | bool use_id; | |
48ddbe19 TH |
487 | |
488 | /* | |
489 | * If %true, cgroup removal will try to clear css refs by retrying | |
490 | * ss->pre_destroy() until there's no css ref left. This behavior | |
491 | * is strictly for backward compatibility and will be removed as | |
492 | * soon as the current user (memcg) is updated. | |
493 | * | |
494 | * If %false, ss->pre_destroy() can't fail and cgroup removal won't | |
495 | * wait for css refs to drop to zero before proceeding. | |
496 | */ | |
497 | bool __DEPRECATED_clear_css_refs; | |
498 | ||
ddbcc7e8 PM |
499 | #define MAX_CGROUP_TYPE_NAMELEN 32 |
500 | const char *name; | |
501 | ||
999cd8a4 PM |
502 | /* |
503 | * Protects sibling/children links of cgroups in this | |
504 | * hierarchy, plus protects which hierarchy (or none) the | |
505 | * subsystem is a part of (i.e. root/sibling). To avoid | |
506 | * potential deadlocks, the following operations should not be | |
507 | * undertaken while holding any hierarchy_mutex: | |
508 | * | |
509 | * - allocating memory | |
510 | * - initiating hotplug events | |
511 | */ | |
512 | struct mutex hierarchy_mutex; | |
cfebe563 | 513 | struct lock_class_key subsys_key; |
ddbcc7e8 | 514 | |
999cd8a4 PM |
515 | /* |
516 | * Link to parent, and list entry in parent's children. | |
517 | * Protected by this->hierarchy_mutex and cgroup_lock() | |
518 | */ | |
519 | struct cgroupfs_root *root; | |
ddbcc7e8 | 520 | struct list_head sibling; |
38460b48 KH |
521 | /* used when use_id == true */ |
522 | struct idr idr; | |
42aee6c4 | 523 | spinlock_t id_lock; |
e6a1105b | 524 | |
8e3f6541 TH |
525 | /* list of cftype_sets */ |
526 | struct list_head cftsets; | |
527 | ||
528 | /* base cftypes, automatically [de]registered with subsys itself */ | |
529 | struct cftype *base_cftypes; | |
530 | struct cftype_set base_cftset; | |
531 | ||
e6a1105b BB |
532 | /* should be defined only by modular subsystems */ |
533 | struct module *module; | |
ddbcc7e8 PM |
534 | }; |
535 | ||
536 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _subsys; | |
537 | #include <linux/cgroup_subsys.h> | |
538 | #undef SUBSYS | |
539 | ||
540 | static inline struct cgroup_subsys_state *cgroup_subsys_state( | |
ffd2d883 | 541 | struct cgroup *cgrp, int subsys_id) |
ddbcc7e8 | 542 | { |
ffd2d883 | 543 | return cgrp->subsys[subsys_id]; |
ddbcc7e8 PM |
544 | } |
545 | ||
dc61b1d6 PZ |
546 | /* |
547 | * function to get the cgroup_subsys_state which allows for extra | |
548 | * rcu_dereference_check() conditions, such as locks used during the | |
549 | * cgroup_subsys::attach() methods. | |
550 | */ | |
551 | #define task_subsys_state_check(task, subsys_id, __c) \ | |
552 | rcu_dereference_check(task->cgroups->subsys[subsys_id], \ | |
dc61b1d6 PZ |
553 | lockdep_is_held(&task->alloc_lock) || \ |
554 | cgroup_lock_is_held() || (__c)) | |
555 | ||
556 | static inline struct cgroup_subsys_state * | |
557 | task_subsys_state(struct task_struct *task, int subsys_id) | |
ddbcc7e8 | 558 | { |
dc61b1d6 | 559 | return task_subsys_state_check(task, subsys_id, false); |
ddbcc7e8 PM |
560 | } |
561 | ||
562 | static inline struct cgroup* task_cgroup(struct task_struct *task, | |
563 | int subsys_id) | |
564 | { | |
565 | return task_subsys_state(task, subsys_id)->cgroup; | |
566 | } | |
567 | ||
817929ec PM |
568 | /* A cgroup_iter should be treated as an opaque object */ |
569 | struct cgroup_iter { | |
570 | struct list_head *cg_link; | |
571 | struct list_head *task; | |
572 | }; | |
573 | ||
d20a390a PM |
574 | /* |
575 | * To iterate across the tasks in a cgroup: | |
817929ec | 576 | * |
b595076a | 577 | * 1) call cgroup_iter_start to initialize an iterator |
817929ec PM |
578 | * |
579 | * 2) call cgroup_iter_next() to retrieve member tasks until it | |
580 | * returns NULL or until you want to end the iteration | |
581 | * | |
582 | * 3) call cgroup_iter_end() to destroy the iterator. | |
31a7df01 | 583 | * |
d20a390a PM |
584 | * Or, call cgroup_scan_tasks() to iterate through every task in a |
585 | * cgroup - cgroup_scan_tasks() holds the css_set_lock when calling | |
586 | * the test_task() callback, but not while calling the process_task() | |
587 | * callback. | |
817929ec | 588 | */ |
ffd2d883 LZ |
589 | void cgroup_iter_start(struct cgroup *cgrp, struct cgroup_iter *it); |
590 | struct task_struct *cgroup_iter_next(struct cgroup *cgrp, | |
817929ec | 591 | struct cgroup_iter *it); |
ffd2d883 | 592 | void cgroup_iter_end(struct cgroup *cgrp, struct cgroup_iter *it); |
31a7df01 | 593 | int cgroup_scan_tasks(struct cgroup_scanner *scan); |
956db3ca | 594 | int cgroup_attach_task(struct cgroup *, struct task_struct *); |
31583bb0 MT |
595 | int cgroup_attach_task_all(struct task_struct *from, struct task_struct *); |
596 | ||
38460b48 KH |
597 | /* |
598 | * CSS ID is ID for cgroup_subsys_state structs under subsys. This only works | |
599 | * if cgroup_subsys.use_id == true. It can be used for looking up and scanning. | |
600 | * CSS ID is assigned at cgroup allocation (create) automatically | |
601 | * and removed when subsys calls free_css_id() function. This is because | |
602 | * the lifetime of cgroup_subsys_state is subsys's matter. | |
603 | * | |
604 | * Looking up and scanning function should be called under rcu_read_lock(). | |
605 | * Taking cgroup_mutex()/hierarchy_mutex() is not necessary for following calls. | |
606 | * But the css returned by this routine can be "not populated yet" or "being | |
607 | * destroyed". The caller should check css and cgroup's status. | |
608 | */ | |
609 | ||
610 | /* | |
611 | * Typically Called at ->destroy(), or somewhere the subsys frees | |
612 | * cgroup_subsys_state. | |
613 | */ | |
614 | void free_css_id(struct cgroup_subsys *ss, struct cgroup_subsys_state *css); | |
615 | ||
616 | /* Find a cgroup_subsys_state which has given ID */ | |
617 | ||
618 | struct cgroup_subsys_state *css_lookup(struct cgroup_subsys *ss, int id); | |
619 | ||
620 | /* | |
621 | * Get a cgroup whose id is greater than or equal to id under tree of root. | |
622 | * Returning a cgroup_subsys_state or NULL. | |
623 | */ | |
624 | struct cgroup_subsys_state *css_get_next(struct cgroup_subsys *ss, int id, | |
625 | struct cgroup_subsys_state *root, int *foundid); | |
626 | ||
627 | /* Returns true if root is ancestor of cg */ | |
628 | bool css_is_ancestor(struct cgroup_subsys_state *cg, | |
0b7f569e | 629 | const struct cgroup_subsys_state *root); |
38460b48 KH |
630 | |
631 | /* Get id and depth of css */ | |
632 | unsigned short css_id(struct cgroup_subsys_state *css); | |
633 | unsigned short css_depth(struct cgroup_subsys_state *css); | |
e5d1367f | 634 | struct cgroup_subsys_state *cgroup_css_from_dir(struct file *f, int id); |
38460b48 | 635 | |
ddbcc7e8 PM |
636 | #else /* !CONFIG_CGROUPS */ |
637 | ||
638 | static inline int cgroup_init_early(void) { return 0; } | |
639 | static inline int cgroup_init(void) { return 0; } | |
b4f48b63 PM |
640 | static inline void cgroup_fork(struct task_struct *p) {} |
641 | static inline void cgroup_fork_callbacks(struct task_struct *p) {} | |
817929ec | 642 | static inline void cgroup_post_fork(struct task_struct *p) {} |
b4f48b63 | 643 | static inline void cgroup_exit(struct task_struct *p, int callbacks) {} |
ddbcc7e8 PM |
644 | |
645 | static inline void cgroup_lock(void) {} | |
646 | static inline void cgroup_unlock(void) {} | |
846c7bb0 BS |
647 | static inline int cgroupstats_build(struct cgroupstats *stats, |
648 | struct dentry *dentry) | |
649 | { | |
650 | return -EINVAL; | |
651 | } | |
ddbcc7e8 | 652 | |
d7926ee3 | 653 | /* No cgroups - nothing to do */ |
31583bb0 MT |
654 | static inline int cgroup_attach_task_all(struct task_struct *from, |
655 | struct task_struct *t) | |
656 | { | |
657 | return 0; | |
658 | } | |
d7926ee3 | 659 | |
ddbcc7e8 PM |
660 | #endif /* !CONFIG_CGROUPS */ |
661 | ||
662 | #endif /* _LINUX_CGROUP_H */ |