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> |
03b1cde6 | 20 | #include <linux/xattr.h> |
ddbcc7e8 PM |
21 | |
22 | #ifdef CONFIG_CGROUPS | |
23 | ||
24 | struct cgroupfs_root; | |
25 | struct cgroup_subsys; | |
26 | struct inode; | |
84eea842 | 27 | struct cgroup; |
38460b48 | 28 | struct css_id; |
ddbcc7e8 PM |
29 | |
30 | extern int cgroup_init_early(void); | |
31 | extern int cgroup_init(void); | |
ddbcc7e8 | 32 | extern void cgroup_lock(void); |
d11c563d | 33 | extern int cgroup_lock_is_held(void); |
84eea842 | 34 | extern bool cgroup_lock_live_group(struct cgroup *cgrp); |
ddbcc7e8 | 35 | extern void cgroup_unlock(void); |
b4f48b63 PM |
36 | extern void cgroup_fork(struct task_struct *p); |
37 | extern void cgroup_fork_callbacks(struct task_struct *p); | |
817929ec | 38 | extern void cgroup_post_fork(struct task_struct *p); |
b4f48b63 | 39 | extern void cgroup_exit(struct task_struct *p, int run_callbacks); |
846c7bb0 BS |
40 | extern int cgroupstats_build(struct cgroupstats *stats, |
41 | struct dentry *dentry); | |
e6a1105b | 42 | extern int cgroup_load_subsys(struct cgroup_subsys *ss); |
cf5d5941 | 43 | extern void cgroup_unload_subsys(struct cgroup_subsys *ss); |
ddbcc7e8 | 44 | |
828c0950 | 45 | extern const struct file_operations proc_cgroup_operations; |
a424316c | 46 | |
aae8aab4 | 47 | /* Define the enumeration of all builtin cgroup subsystems */ |
817929ec PM |
48 | #define SUBSYS(_x) _x ## _subsys_id, |
49 | enum cgroup_subsys_id { | |
50 | #include <linux/cgroup_subsys.h> | |
aae8aab4 | 51 | CGROUP_BUILTIN_SUBSYS_COUNT |
817929ec PM |
52 | }; |
53 | #undef SUBSYS | |
aae8aab4 BB |
54 | /* |
55 | * This define indicates the maximum number of subsystems that can be loaded | |
56 | * at once. We limit to this many since cgroupfs_root has subsys_bits to keep | |
57 | * track of all of them. | |
58 | */ | |
59 | #define CGROUP_SUBSYS_COUNT (BITS_PER_BYTE*sizeof(unsigned long)) | |
817929ec | 60 | |
ddbcc7e8 PM |
61 | /* Per-subsystem/per-cgroup state maintained by the system. */ |
62 | struct cgroup_subsys_state { | |
d20a390a PM |
63 | /* |
64 | * The cgroup that this subsystem is attached to. Useful | |
ddbcc7e8 | 65 | * for subsystems that want to know about the cgroup |
d20a390a PM |
66 | * hierarchy structure |
67 | */ | |
ddbcc7e8 PM |
68 | struct cgroup *cgroup; |
69 | ||
d20a390a PM |
70 | /* |
71 | * State maintained by the cgroup system to allow subsystems | |
e7c5ec91 | 72 | * to be "busy". Should be accessed via css_get(), |
d20a390a PM |
73 | * css_tryget() and and css_put(). |
74 | */ | |
ddbcc7e8 PM |
75 | |
76 | atomic_t refcnt; | |
77 | ||
78 | unsigned long flags; | |
38460b48 | 79 | /* ID for this css, if possible */ |
2c392b8c | 80 | struct css_id __rcu *id; |
48ddbe19 TH |
81 | |
82 | /* Used to put @cgroup->dentry on the last css_put() */ | |
83 | struct work_struct dput_work; | |
ddbcc7e8 PM |
84 | }; |
85 | ||
86 | /* bits in struct cgroup_subsys_state flags field */ | |
87 | enum { | |
88 | CSS_ROOT, /* This CSS is the root of the subsystem */ | |
e7c5ec91 | 89 | CSS_REMOVED, /* This CSS is dead */ |
48ddbe19 | 90 | CSS_CLEAR_CSS_REFS, /* @ss->__DEPRECATED_clear_css_refs */ |
ddbcc7e8 PM |
91 | }; |
92 | ||
d7b9fff7 DN |
93 | /* Caller must verify that the css is not for root cgroup */ |
94 | static inline void __css_get(struct cgroup_subsys_state *css, int count) | |
95 | { | |
96 | atomic_add(count, &css->refcnt); | |
97 | } | |
98 | ||
ddbcc7e8 | 99 | /* |
e7c5ec91 PM |
100 | * Call css_get() to hold a reference on the css; it can be used |
101 | * for a reference obtained via: | |
102 | * - an existing ref-counted reference to the css | |
103 | * - task->cgroups for a locked task | |
ddbcc7e8 PM |
104 | */ |
105 | ||
106 | static inline void css_get(struct cgroup_subsys_state *css) | |
107 | { | |
108 | /* We don't need to reference count the root state */ | |
109 | if (!test_bit(CSS_ROOT, &css->flags)) | |
d7b9fff7 | 110 | __css_get(css, 1); |
ddbcc7e8 | 111 | } |
e7c5ec91 PM |
112 | |
113 | static inline bool css_is_removed(struct cgroup_subsys_state *css) | |
114 | { | |
115 | return test_bit(CSS_REMOVED, &css->flags); | |
116 | } | |
117 | ||
118 | /* | |
119 | * Call css_tryget() to take a reference on a css if your existing | |
120 | * (known-valid) reference isn't already ref-counted. Returns false if | |
121 | * the css has been destroyed. | |
122 | */ | |
123 | ||
28b4c27b | 124 | extern bool __css_tryget(struct cgroup_subsys_state *css); |
e7c5ec91 PM |
125 | static inline bool css_tryget(struct cgroup_subsys_state *css) |
126 | { | |
127 | if (test_bit(CSS_ROOT, &css->flags)) | |
128 | return true; | |
28b4c27b | 129 | return __css_tryget(css); |
e7c5ec91 PM |
130 | } |
131 | ||
ddbcc7e8 PM |
132 | /* |
133 | * css_put() should be called to release a reference taken by | |
e7c5ec91 | 134 | * css_get() or css_tryget() |
ddbcc7e8 PM |
135 | */ |
136 | ||
28b4c27b | 137 | extern void __css_put(struct cgroup_subsys_state *css); |
ddbcc7e8 PM |
138 | static inline void css_put(struct cgroup_subsys_state *css) |
139 | { | |
140 | if (!test_bit(CSS_ROOT, &css->flags)) | |
28b4c27b | 141 | __css_put(css); |
ddbcc7e8 PM |
142 | } |
143 | ||
3116f0e3 PM |
144 | /* bits in struct cgroup flags field */ |
145 | enum { | |
146 | /* Control Group is dead */ | |
147 | CGRP_REMOVED, | |
d20a390a PM |
148 | /* |
149 | * Control Group has previously had a child cgroup or a task, | |
150 | * but no longer (only if CGRP_NOTIFY_ON_RELEASE is set) | |
151 | */ | |
3116f0e3 PM |
152 | CGRP_RELEASABLE, |
153 | /* Control Group requires release notifications to userspace */ | |
154 | CGRP_NOTIFY_ON_RELEASE, | |
ec64f515 KH |
155 | /* |
156 | * A thread in rmdir() is wating for this cgroup. | |
157 | */ | |
158 | CGRP_WAIT_ON_RMDIR, | |
97978e6d DL |
159 | /* |
160 | * Clone cgroup values when creating a new child cgroup | |
161 | */ | |
162 | CGRP_CLONE_CHILDREN, | |
3116f0e3 PM |
163 | }; |
164 | ||
ddbcc7e8 PM |
165 | struct cgroup { |
166 | unsigned long flags; /* "unsigned long" so bitops work */ | |
167 | ||
d20a390a PM |
168 | /* |
169 | * count users of this cgroup. >0 means busy, but doesn't | |
170 | * necessarily indicate the number of tasks in the cgroup | |
171 | */ | |
ddbcc7e8 PM |
172 | atomic_t count; |
173 | ||
174 | /* | |
175 | * We link our 'sibling' struct into our parent's 'children'. | |
176 | * Our children link their 'sibling' into our 'children'. | |
177 | */ | |
178 | struct list_head sibling; /* my parent's children */ | |
179 | struct list_head children; /* my children */ | |
05ef1d7c | 180 | struct list_head files; /* my files */ |
ddbcc7e8 | 181 | |
d20a390a | 182 | struct cgroup *parent; /* my parent */ |
2c392b8c | 183 | struct dentry __rcu *dentry; /* cgroup fs entry, RCU protected */ |
ddbcc7e8 PM |
184 | |
185 | /* Private pointers for each registered subsystem */ | |
186 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
187 | ||
188 | struct cgroupfs_root *root; | |
189 | struct cgroup *top_cgroup; | |
817929ec PM |
190 | |
191 | /* | |
192 | * List of cg_cgroup_links pointing at css_sets with | |
193 | * tasks in this cgroup. Protected by css_set_lock | |
194 | */ | |
195 | struct list_head css_sets; | |
81a6a5cd | 196 | |
b0ca5a84 | 197 | struct list_head allcg_node; /* cgroupfs_root->allcg_list */ |
8e3f6541 | 198 | struct list_head cft_q_node; /* used during cftype add/rm */ |
b0ca5a84 | 199 | |
81a6a5cd PM |
200 | /* |
201 | * Linked list running through all cgroups that can | |
202 | * potentially be reaped by the release agent. Protected by | |
203 | * release_list_lock | |
204 | */ | |
205 | struct list_head release_list; | |
cc31edce | 206 | |
72a8cb30 BB |
207 | /* |
208 | * list of pidlists, up to two for each namespace (one for procs, one | |
209 | * for tasks); created on demand. | |
210 | */ | |
211 | struct list_head pidlists; | |
212 | struct mutex pidlist_mutex; | |
a47295e6 PM |
213 | |
214 | /* For RCU-protected deletion */ | |
215 | struct rcu_head rcu_head; | |
0dea1168 | 216 | |
25985edc | 217 | /* List of events which userspace want to receive */ |
0dea1168 KS |
218 | struct list_head event_list; |
219 | spinlock_t event_list_lock; | |
03b1cde6 AR |
220 | |
221 | /* directory xattrs */ | |
222 | struct simple_xattrs xattrs; | |
817929ec PM |
223 | }; |
224 | ||
d20a390a PM |
225 | /* |
226 | * A css_set is a structure holding pointers to a set of | |
817929ec PM |
227 | * cgroup_subsys_state objects. This saves space in the task struct |
228 | * object and speeds up fork()/exit(), since a single inc/dec and a | |
d20a390a PM |
229 | * list_add()/del() can bump the reference count on the entire cgroup |
230 | * set for a task. | |
817929ec PM |
231 | */ |
232 | ||
233 | struct css_set { | |
234 | ||
235 | /* Reference count */ | |
146aa1bd | 236 | atomic_t refcount; |
817929ec | 237 | |
472b1053 LZ |
238 | /* |
239 | * List running through all cgroup groups in the same hash | |
240 | * slot. Protected by css_set_lock | |
241 | */ | |
242 | struct hlist_node hlist; | |
243 | ||
817929ec PM |
244 | /* |
245 | * List running through all tasks using this cgroup | |
246 | * group. Protected by css_set_lock | |
247 | */ | |
248 | struct list_head tasks; | |
249 | ||
250 | /* | |
251 | * List of cg_cgroup_link objects on link chains from | |
252 | * cgroups referenced from this css_set. Protected by | |
253 | * css_set_lock | |
254 | */ | |
255 | struct list_head cg_links; | |
256 | ||
257 | /* | |
258 | * Set of subsystem states, one for each subsystem. This array | |
259 | * is immutable after creation apart from the init_css_set | |
cf5d5941 BB |
260 | * during subsystem registration (at boot time) and modular subsystem |
261 | * loading/unloading. | |
817929ec PM |
262 | */ |
263 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
c378369d BB |
264 | |
265 | /* For RCU-protected deletion */ | |
266 | struct rcu_head rcu_head; | |
ddbcc7e8 PM |
267 | }; |
268 | ||
91796569 PM |
269 | /* |
270 | * cgroup_map_cb is an abstract callback API for reporting map-valued | |
271 | * control files | |
272 | */ | |
273 | ||
274 | struct cgroup_map_cb { | |
275 | int (*fill)(struct cgroup_map_cb *cb, const char *key, u64 value); | |
276 | void *state; | |
277 | }; | |
278 | ||
d20a390a PM |
279 | /* |
280 | * struct cftype: handler definitions for cgroup control files | |
ddbcc7e8 PM |
281 | * |
282 | * When reading/writing to a file: | |
a043e3b2 | 283 | * - the cgroup to use is file->f_dentry->d_parent->d_fsdata |
ddbcc7e8 PM |
284 | * - the 'cftype' of the file is file->f_dentry->d_fsdata |
285 | */ | |
286 | ||
8e3f6541 TH |
287 | /* cftype->flags */ |
288 | #define CFTYPE_ONLY_ON_ROOT (1U << 0) /* only create on root cg */ | |
289 | #define CFTYPE_NOT_ON_ROOT (1U << 1) /* don't create onp root cg */ | |
290 | ||
291 | #define MAX_CFTYPE_NAME 64 | |
292 | ||
ddbcc7e8 | 293 | struct cftype { |
d20a390a PM |
294 | /* |
295 | * By convention, the name should begin with the name of the | |
8e3f6541 TH |
296 | * subsystem, followed by a period. Zero length string indicates |
297 | * end of cftype array. | |
d20a390a | 298 | */ |
ddbcc7e8 PM |
299 | char name[MAX_CFTYPE_NAME]; |
300 | int private; | |
099fca32 LZ |
301 | /* |
302 | * If not 0, file mode is set to this value, otherwise it will | |
303 | * be figured out automatically | |
304 | */ | |
a5e7ed32 | 305 | umode_t mode; |
db3b1497 PM |
306 | |
307 | /* | |
308 | * If non-zero, defines the maximum length of string that can | |
309 | * be passed to write_string; defaults to 64 | |
310 | */ | |
311 | size_t max_write_len; | |
312 | ||
8e3f6541 TH |
313 | /* CFTYPE_* flags */ |
314 | unsigned int flags; | |
315 | ||
03b1cde6 AR |
316 | /* file xattrs */ |
317 | struct simple_xattrs xattrs; | |
318 | ||
ce16b49d PM |
319 | int (*open)(struct inode *inode, struct file *file); |
320 | ssize_t (*read)(struct cgroup *cgrp, struct cftype *cft, | |
321 | struct file *file, | |
322 | char __user *buf, size_t nbytes, loff_t *ppos); | |
ddbcc7e8 | 323 | /* |
f4c753b7 | 324 | * read_u64() is a shortcut for the common case of returning a |
ddbcc7e8 PM |
325 | * single integer. Use it in place of read() |
326 | */ | |
ce16b49d | 327 | u64 (*read_u64)(struct cgroup *cgrp, struct cftype *cft); |
e73d2c61 PM |
328 | /* |
329 | * read_s64() is a signed version of read_u64() | |
330 | */ | |
ce16b49d | 331 | s64 (*read_s64)(struct cgroup *cgrp, struct cftype *cft); |
91796569 PM |
332 | /* |
333 | * read_map() is used for defining a map of key/value | |
334 | * pairs. It should call cb->fill(cb, key, value) for each | |
335 | * entry. The key/value pairs (and their ordering) should not | |
336 | * change between reboots. | |
337 | */ | |
ce16b49d PM |
338 | int (*read_map)(struct cgroup *cont, struct cftype *cft, |
339 | struct cgroup_map_cb *cb); | |
29486df3 SH |
340 | /* |
341 | * read_seq_string() is used for outputting a simple sequence | |
342 | * using seqfile. | |
343 | */ | |
ce16b49d PM |
344 | int (*read_seq_string)(struct cgroup *cont, struct cftype *cft, |
345 | struct seq_file *m); | |
91796569 | 346 | |
ce16b49d PM |
347 | ssize_t (*write)(struct cgroup *cgrp, struct cftype *cft, |
348 | struct file *file, | |
349 | const char __user *buf, size_t nbytes, loff_t *ppos); | |
355e0c48 PM |
350 | |
351 | /* | |
f4c753b7 | 352 | * write_u64() is a shortcut for the common case of accepting |
355e0c48 PM |
353 | * a single integer (as parsed by simple_strtoull) from |
354 | * userspace. Use in place of write(); return 0 or error. | |
355 | */ | |
ce16b49d | 356 | int (*write_u64)(struct cgroup *cgrp, struct cftype *cft, u64 val); |
e73d2c61 PM |
357 | /* |
358 | * write_s64() is a signed version of write_u64() | |
359 | */ | |
ce16b49d | 360 | int (*write_s64)(struct cgroup *cgrp, struct cftype *cft, s64 val); |
355e0c48 | 361 | |
db3b1497 PM |
362 | /* |
363 | * write_string() is passed a nul-terminated kernelspace | |
364 | * buffer of maximum length determined by max_write_len. | |
365 | * Returns 0 or -ve error code. | |
366 | */ | |
367 | int (*write_string)(struct cgroup *cgrp, struct cftype *cft, | |
368 | const char *buffer); | |
d447ea2f PE |
369 | /* |
370 | * trigger() callback can be used to get some kick from the | |
371 | * userspace, when the actual string written is not important | |
372 | * at all. The private field can be used to determine the | |
373 | * kick type for multiplexing. | |
374 | */ | |
375 | int (*trigger)(struct cgroup *cgrp, unsigned int event); | |
376 | ||
ce16b49d | 377 | int (*release)(struct inode *inode, struct file *file); |
0dea1168 KS |
378 | |
379 | /* | |
380 | * register_event() callback will be used to add new userspace | |
381 | * waiter for changes related to the cftype. Implement it if | |
382 | * you want to provide this functionality. Use eventfd_signal() | |
383 | * on eventfd to send notification to userspace. | |
384 | */ | |
385 | int (*register_event)(struct cgroup *cgrp, struct cftype *cft, | |
386 | struct eventfd_ctx *eventfd, const char *args); | |
387 | /* | |
388 | * unregister_event() callback will be called when userspace | |
389 | * closes the eventfd or on cgroup removing. | |
390 | * This callback must be implemented, if you want provide | |
391 | * notification functionality. | |
0dea1168 | 392 | */ |
907860ed | 393 | void (*unregister_event)(struct cgroup *cgrp, struct cftype *cft, |
0dea1168 | 394 | struct eventfd_ctx *eventfd); |
ddbcc7e8 PM |
395 | }; |
396 | ||
8e3f6541 TH |
397 | /* |
398 | * cftype_sets describe cftypes belonging to a subsystem and are chained at | |
399 | * cgroup_subsys->cftsets. Each cftset points to an array of cftypes | |
400 | * terminated by zero length name. | |
401 | */ | |
402 | struct cftype_set { | |
403 | struct list_head node; /* chained at subsys->cftsets */ | |
03b1cde6 | 404 | struct cftype *cfts; |
8e3f6541 TH |
405 | }; |
406 | ||
31a7df01 CW |
407 | struct cgroup_scanner { |
408 | struct cgroup *cg; | |
409 | int (*test_task)(struct task_struct *p, struct cgroup_scanner *scan); | |
410 | void (*process_task)(struct task_struct *p, | |
411 | struct cgroup_scanner *scan); | |
412 | struct ptr_heap *heap; | |
bd1a8ab7 | 413 | void *data; |
31a7df01 CW |
414 | }; |
415 | ||
03b1cde6 AR |
416 | int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); |
417 | int cgroup_rm_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); | |
8e3f6541 | 418 | |
ffd2d883 | 419 | int cgroup_is_removed(const struct cgroup *cgrp); |
ddbcc7e8 | 420 | |
ffd2d883 | 421 | int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen); |
ddbcc7e8 | 422 | |
ffd2d883 | 423 | int cgroup_task_count(const struct cgroup *cgrp); |
bbcb81d0 | 424 | |
313e924c GN |
425 | /* Return true if cgrp is a descendant of the task's cgroup */ |
426 | int cgroup_is_descendant(const struct cgroup *cgrp, struct task_struct *task); | |
ddbcc7e8 | 427 | |
88703267 KH |
428 | /* |
429 | * When the subsys has to access css and may add permanent refcnt to css, | |
430 | * it should take care of racy conditions with rmdir(). Following set of | |
431 | * functions, is for stop/restart rmdir if necessary. | |
432 | * Because these will call css_get/put, "css" should be alive css. | |
433 | * | |
434 | * cgroup_exclude_rmdir(); | |
435 | * ...do some jobs which may access arbitrary empty cgroup | |
436 | * cgroup_release_and_wakeup_rmdir(); | |
437 | * | |
438 | * When someone removes a cgroup while cgroup_exclude_rmdir() holds it, | |
439 | * it sleeps and cgroup_release_and_wakeup_rmdir() will wake him up. | |
440 | */ | |
441 | ||
442 | void cgroup_exclude_rmdir(struct cgroup_subsys_state *css); | |
443 | void cgroup_release_and_wakeup_rmdir(struct cgroup_subsys_state *css); | |
444 | ||
2f7ee569 TH |
445 | /* |
446 | * Control Group taskset, used to pass around set of tasks to cgroup_subsys | |
447 | * methods. | |
448 | */ | |
449 | struct cgroup_taskset; | |
450 | struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset); | |
451 | struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset); | |
452 | struct cgroup *cgroup_taskset_cur_cgroup(struct cgroup_taskset *tset); | |
453 | int cgroup_taskset_size(struct cgroup_taskset *tset); | |
454 | ||
455 | /** | |
456 | * cgroup_taskset_for_each - iterate cgroup_taskset | |
457 | * @task: the loop cursor | |
458 | * @skip_cgrp: skip if task's cgroup matches this, %NULL to iterate through all | |
459 | * @tset: taskset to iterate | |
460 | */ | |
461 | #define cgroup_taskset_for_each(task, skip_cgrp, tset) \ | |
462 | for ((task) = cgroup_taskset_first((tset)); (task); \ | |
463 | (task) = cgroup_taskset_next((tset))) \ | |
464 | if (!(skip_cgrp) || \ | |
465 | cgroup_taskset_cur_cgroup((tset)) != (skip_cgrp)) | |
466 | ||
21acb9ca TLSC |
467 | /* |
468 | * Control Group subsystem type. | |
469 | * See Documentation/cgroups/cgroups.txt for details | |
470 | */ | |
ddbcc7e8 PM |
471 | |
472 | struct cgroup_subsys { | |
761b3ef5 LZ |
473 | struct cgroup_subsys_state *(*create)(struct cgroup *cgrp); |
474 | int (*pre_destroy)(struct cgroup *cgrp); | |
475 | void (*destroy)(struct cgroup *cgrp); | |
476 | int (*can_attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); | |
477 | void (*cancel_attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); | |
478 | void (*attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); | |
479 | void (*fork)(struct task_struct *task); | |
480 | void (*exit)(struct cgroup *cgrp, struct cgroup *old_cgrp, | |
481 | struct task_struct *task); | |
761b3ef5 LZ |
482 | void (*post_clone)(struct cgroup *cgrp); |
483 | void (*bind)(struct cgroup *root); | |
e5991371 | 484 | |
ddbcc7e8 PM |
485 | int subsys_id; |
486 | int active; | |
8bab8dde | 487 | int disabled; |
ddbcc7e8 | 488 | int early_init; |
38460b48 KH |
489 | /* |
490 | * True if this subsys uses ID. ID is not available before cgroup_init() | |
491 | * (not available in early_init time.) | |
492 | */ | |
493 | bool use_id; | |
48ddbe19 TH |
494 | |
495 | /* | |
496 | * If %true, cgroup removal will try to clear css refs by retrying | |
497 | * ss->pre_destroy() until there's no css ref left. This behavior | |
498 | * is strictly for backward compatibility and will be removed as | |
499 | * soon as the current user (memcg) is updated. | |
500 | * | |
501 | * If %false, ss->pre_destroy() can't fail and cgroup removal won't | |
502 | * wait for css refs to drop to zero before proceeding. | |
503 | */ | |
504 | bool __DEPRECATED_clear_css_refs; | |
505 | ||
ddbcc7e8 PM |
506 | #define MAX_CGROUP_TYPE_NAMELEN 32 |
507 | const char *name; | |
508 | ||
999cd8a4 PM |
509 | /* |
510 | * Link to parent, and list entry in parent's children. | |
6be96a5c | 511 | * Protected by cgroup_lock() |
999cd8a4 PM |
512 | */ |
513 | struct cgroupfs_root *root; | |
ddbcc7e8 | 514 | struct list_head sibling; |
38460b48 KH |
515 | /* used when use_id == true */ |
516 | struct idr idr; | |
42aee6c4 | 517 | spinlock_t id_lock; |
e6a1105b | 518 | |
8e3f6541 TH |
519 | /* list of cftype_sets */ |
520 | struct list_head cftsets; | |
521 | ||
522 | /* base cftypes, automatically [de]registered with subsys itself */ | |
523 | struct cftype *base_cftypes; | |
524 | struct cftype_set base_cftset; | |
525 | ||
e6a1105b BB |
526 | /* should be defined only by modular subsystems */ |
527 | struct module *module; | |
ddbcc7e8 PM |
528 | }; |
529 | ||
530 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _subsys; | |
531 | #include <linux/cgroup_subsys.h> | |
532 | #undef SUBSYS | |
533 | ||
534 | static inline struct cgroup_subsys_state *cgroup_subsys_state( | |
ffd2d883 | 535 | struct cgroup *cgrp, int subsys_id) |
ddbcc7e8 | 536 | { |
ffd2d883 | 537 | return cgrp->subsys[subsys_id]; |
ddbcc7e8 PM |
538 | } |
539 | ||
dc61b1d6 PZ |
540 | /* |
541 | * function to get the cgroup_subsys_state which allows for extra | |
542 | * rcu_dereference_check() conditions, such as locks used during the | |
543 | * cgroup_subsys::attach() methods. | |
544 | */ | |
545 | #define task_subsys_state_check(task, subsys_id, __c) \ | |
546 | rcu_dereference_check(task->cgroups->subsys[subsys_id], \ | |
dc61b1d6 PZ |
547 | lockdep_is_held(&task->alloc_lock) || \ |
548 | cgroup_lock_is_held() || (__c)) | |
549 | ||
550 | static inline struct cgroup_subsys_state * | |
551 | task_subsys_state(struct task_struct *task, int subsys_id) | |
ddbcc7e8 | 552 | { |
dc61b1d6 | 553 | return task_subsys_state_check(task, subsys_id, false); |
ddbcc7e8 PM |
554 | } |
555 | ||
556 | static inline struct cgroup* task_cgroup(struct task_struct *task, | |
557 | int subsys_id) | |
558 | { | |
559 | return task_subsys_state(task, subsys_id)->cgroup; | |
560 | } | |
561 | ||
817929ec PM |
562 | /* A cgroup_iter should be treated as an opaque object */ |
563 | struct cgroup_iter { | |
564 | struct list_head *cg_link; | |
565 | struct list_head *task; | |
566 | }; | |
567 | ||
d20a390a PM |
568 | /* |
569 | * To iterate across the tasks in a cgroup: | |
817929ec | 570 | * |
b595076a | 571 | * 1) call cgroup_iter_start to initialize an iterator |
817929ec PM |
572 | * |
573 | * 2) call cgroup_iter_next() to retrieve member tasks until it | |
574 | * returns NULL or until you want to end the iteration | |
575 | * | |
576 | * 3) call cgroup_iter_end() to destroy the iterator. | |
31a7df01 | 577 | * |
d20a390a PM |
578 | * Or, call cgroup_scan_tasks() to iterate through every task in a |
579 | * cgroup - cgroup_scan_tasks() holds the css_set_lock when calling | |
580 | * the test_task() callback, but not while calling the process_task() | |
581 | * callback. | |
817929ec | 582 | */ |
ffd2d883 LZ |
583 | void cgroup_iter_start(struct cgroup *cgrp, struct cgroup_iter *it); |
584 | struct task_struct *cgroup_iter_next(struct cgroup *cgrp, | |
817929ec | 585 | struct cgroup_iter *it); |
ffd2d883 | 586 | void cgroup_iter_end(struct cgroup *cgrp, struct cgroup_iter *it); |
31a7df01 | 587 | int cgroup_scan_tasks(struct cgroup_scanner *scan); |
956db3ca | 588 | int cgroup_attach_task(struct cgroup *, struct task_struct *); |
31583bb0 MT |
589 | int cgroup_attach_task_all(struct task_struct *from, struct task_struct *); |
590 | ||
38460b48 KH |
591 | /* |
592 | * CSS ID is ID for cgroup_subsys_state structs under subsys. This only works | |
593 | * if cgroup_subsys.use_id == true. It can be used for looking up and scanning. | |
594 | * CSS ID is assigned at cgroup allocation (create) automatically | |
595 | * and removed when subsys calls free_css_id() function. This is because | |
596 | * the lifetime of cgroup_subsys_state is subsys's matter. | |
597 | * | |
598 | * Looking up and scanning function should be called under rcu_read_lock(). | |
6be96a5c | 599 | * Taking cgroup_mutex is not necessary for following calls. |
38460b48 KH |
600 | * But the css returned by this routine can be "not populated yet" or "being |
601 | * destroyed". The caller should check css and cgroup's status. | |
602 | */ | |
603 | ||
604 | /* | |
605 | * Typically Called at ->destroy(), or somewhere the subsys frees | |
606 | * cgroup_subsys_state. | |
607 | */ | |
608 | void free_css_id(struct cgroup_subsys *ss, struct cgroup_subsys_state *css); | |
609 | ||
610 | /* Find a cgroup_subsys_state which has given ID */ | |
611 | ||
612 | struct cgroup_subsys_state *css_lookup(struct cgroup_subsys *ss, int id); | |
613 | ||
614 | /* | |
615 | * Get a cgroup whose id is greater than or equal to id under tree of root. | |
616 | * Returning a cgroup_subsys_state or NULL. | |
617 | */ | |
618 | struct cgroup_subsys_state *css_get_next(struct cgroup_subsys *ss, int id, | |
619 | struct cgroup_subsys_state *root, int *foundid); | |
620 | ||
621 | /* Returns true if root is ancestor of cg */ | |
622 | bool css_is_ancestor(struct cgroup_subsys_state *cg, | |
0b7f569e | 623 | const struct cgroup_subsys_state *root); |
38460b48 KH |
624 | |
625 | /* Get id and depth of css */ | |
626 | unsigned short css_id(struct cgroup_subsys_state *css); | |
627 | unsigned short css_depth(struct cgroup_subsys_state *css); | |
e5d1367f | 628 | struct cgroup_subsys_state *cgroup_css_from_dir(struct file *f, int id); |
38460b48 | 629 | |
ddbcc7e8 PM |
630 | #else /* !CONFIG_CGROUPS */ |
631 | ||
632 | static inline int cgroup_init_early(void) { return 0; } | |
633 | static inline int cgroup_init(void) { return 0; } | |
b4f48b63 PM |
634 | static inline void cgroup_fork(struct task_struct *p) {} |
635 | static inline void cgroup_fork_callbacks(struct task_struct *p) {} | |
817929ec | 636 | static inline void cgroup_post_fork(struct task_struct *p) {} |
b4f48b63 | 637 | static inline void cgroup_exit(struct task_struct *p, int callbacks) {} |
ddbcc7e8 PM |
638 | |
639 | static inline void cgroup_lock(void) {} | |
640 | static inline void cgroup_unlock(void) {} | |
846c7bb0 BS |
641 | static inline int cgroupstats_build(struct cgroupstats *stats, |
642 | struct dentry *dentry) | |
643 | { | |
644 | return -EINVAL; | |
645 | } | |
ddbcc7e8 | 646 | |
d7926ee3 | 647 | /* No cgroups - nothing to do */ |
31583bb0 MT |
648 | static inline int cgroup_attach_task_all(struct task_struct *from, |
649 | struct task_struct *t) | |
650 | { | |
651 | return 0; | |
652 | } | |
d7926ee3 | 653 | |
ddbcc7e8 PM |
654 | #endif /* !CONFIG_CGROUPS */ |
655 | ||
656 | #endif /* _LINUX_CGROUP_H */ |