]>
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> | |
eb6fd504 | 15 | #include <linux/rculist.h> |
846c7bb0 | 16 | #include <linux/cgroupstats.h> |
31a7df01 | 17 | #include <linux/prio_heap.h> |
cc31edce | 18 | #include <linux/rwsem.h> |
38460b48 | 19 | #include <linux/idr.h> |
48ddbe19 | 20 | #include <linux/workqueue.h> |
03b1cde6 | 21 | #include <linux/xattr.h> |
25a7e684 | 22 | #include <linux/fs.h> |
ddbcc7e8 PM |
23 | |
24 | #ifdef CONFIG_CGROUPS | |
25 | ||
26 | struct cgroupfs_root; | |
27 | struct cgroup_subsys; | |
28 | struct inode; | |
84eea842 | 29 | struct cgroup; |
38460b48 | 30 | struct css_id; |
a27bb332 | 31 | struct eventfd_ctx; |
ddbcc7e8 PM |
32 | |
33 | extern int cgroup_init_early(void); | |
34 | extern int cgroup_init(void); | |
b4f48b63 | 35 | extern void cgroup_fork(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 | |
8d8b97ba | 43 | extern int proc_cgroup_show(struct seq_file *, void *); |
a424316c | 44 | |
7d8e0bf5 LZ |
45 | /* |
46 | * Define the enumeration of all cgroup subsystems. | |
47 | * | |
48 | * We define ids for builtin subsystems and then modular ones. | |
49 | */ | |
817929ec PM |
50 | #define SUBSYS(_x) _x ## _subsys_id, |
51 | enum cgroup_subsys_id { | |
7d8e0bf5 | 52 | #define IS_SUBSYS_ENABLED(option) IS_BUILTIN(option) |
817929ec | 53 | #include <linux/cgroup_subsys.h> |
7d8e0bf5 LZ |
54 | #undef IS_SUBSYS_ENABLED |
55 | CGROUP_BUILTIN_SUBSYS_COUNT, | |
56 | ||
57 | __CGROUP_SUBSYS_TEMP_PLACEHOLDER = CGROUP_BUILTIN_SUBSYS_COUNT - 1, | |
58 | ||
59 | #define IS_SUBSYS_ENABLED(option) IS_MODULE(option) | |
817929ec | 60 | #include <linux/cgroup_subsys.h> |
7d8e0bf5 | 61 | #undef IS_SUBSYS_ENABLED |
a6f00298 | 62 | CGROUP_SUBSYS_COUNT, |
817929ec PM |
63 | }; |
64 | #undef SUBSYS | |
65 | ||
ddbcc7e8 PM |
66 | /* Per-subsystem/per-cgroup state maintained by the system. */ |
67 | struct cgroup_subsys_state { | |
d20a390a PM |
68 | /* |
69 | * The cgroup that this subsystem is attached to. Useful | |
ddbcc7e8 | 70 | * for subsystems that want to know about the cgroup |
d20a390a PM |
71 | * hierarchy structure |
72 | */ | |
ddbcc7e8 PM |
73 | struct cgroup *cgroup; |
74 | ||
d20a390a PM |
75 | /* |
76 | * State maintained by the cgroup system to allow subsystems | |
e7c5ec91 | 77 | * to be "busy". Should be accessed via css_get(), |
d0b2fdd2 | 78 | * css_tryget() and css_put(). |
d20a390a | 79 | */ |
ddbcc7e8 PM |
80 | |
81 | atomic_t refcnt; | |
82 | ||
83 | unsigned long flags; | |
38460b48 | 84 | /* ID for this css, if possible */ |
2c392b8c | 85 | struct css_id __rcu *id; |
48ddbe19 TH |
86 | |
87 | /* Used to put @cgroup->dentry on the last css_put() */ | |
88 | struct work_struct dput_work; | |
ddbcc7e8 PM |
89 | }; |
90 | ||
91 | /* bits in struct cgroup_subsys_state flags field */ | |
92 | enum { | |
38b53aba | 93 | CSS_ROOT = (1 << 0), /* this CSS is the root of the subsystem */ |
92fb9748 | 94 | CSS_ONLINE = (1 << 1), /* between ->css_online() and ->css_offline() */ |
ddbcc7e8 PM |
95 | }; |
96 | ||
d7b9fff7 DN |
97 | /* Caller must verify that the css is not for root cgroup */ |
98 | static inline void __css_get(struct cgroup_subsys_state *css, int count) | |
99 | { | |
100 | atomic_add(count, &css->refcnt); | |
101 | } | |
102 | ||
ddbcc7e8 | 103 | /* |
e7c5ec91 PM |
104 | * Call css_get() to hold a reference on the css; it can be used |
105 | * for a reference obtained via: | |
106 | * - an existing ref-counted reference to the css | |
107 | * - task->cgroups for a locked task | |
ddbcc7e8 PM |
108 | */ |
109 | ||
110 | static inline void css_get(struct cgroup_subsys_state *css) | |
111 | { | |
112 | /* We don't need to reference count the root state */ | |
38b53aba | 113 | if (!(css->flags & CSS_ROOT)) |
d7b9fff7 | 114 | __css_get(css, 1); |
ddbcc7e8 | 115 | } |
e7c5ec91 | 116 | |
e7c5ec91 PM |
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 | { | |
38b53aba | 126 | if (css->flags & CSS_ROOT) |
e7c5ec91 | 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 | { | |
38b53aba | 139 | if (!(css->flags & CSS_ROOT)) |
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, | |
97978e6d | 154 | /* |
2260e7fc TH |
155 | * Clone the parent's configuration when creating a new child |
156 | * cpuset cgroup. For historical reasons, this option can be | |
157 | * specified at mount time and thus is implemented here. | |
97978e6d | 158 | */ |
2260e7fc | 159 | CGRP_CPUSET_CLONE_CHILDREN, |
873fe09e TH |
160 | /* see the comment above CGRP_ROOT_SANE_BEHAVIOR for details */ |
161 | CGRP_SANE_BEHAVIOR, | |
3116f0e3 PM |
162 | }; |
163 | ||
65dff759 LZ |
164 | struct cgroup_name { |
165 | struct rcu_head rcu_head; | |
166 | char name[]; | |
3116f0e3 PM |
167 | }; |
168 | ||
ddbcc7e8 PM |
169 | struct cgroup { |
170 | unsigned long flags; /* "unsigned long" so bitops work */ | |
171 | ||
d20a390a PM |
172 | /* |
173 | * count users of this cgroup. >0 means busy, but doesn't | |
174 | * necessarily indicate the number of tasks in the cgroup | |
175 | */ | |
ddbcc7e8 PM |
176 | atomic_t count; |
177 | ||
0a950f65 TH |
178 | int id; /* ida allocated in-hierarchy ID */ |
179 | ||
ddbcc7e8 PM |
180 | /* |
181 | * We link our 'sibling' struct into our parent's 'children'. | |
182 | * Our children link their 'sibling' into our 'children'. | |
183 | */ | |
184 | struct list_head sibling; /* my parent's children */ | |
185 | struct list_head children; /* my children */ | |
05ef1d7c | 186 | struct list_head files; /* my files */ |
ddbcc7e8 | 187 | |
d20a390a | 188 | struct cgroup *parent; /* my parent */ |
febfcef6 | 189 | struct dentry *dentry; /* cgroup fs entry, RCU protected */ |
ddbcc7e8 | 190 | |
53fa5261 TH |
191 | /* |
192 | * Monotonically increasing unique serial number which defines a | |
193 | * uniform order among all cgroups. It's guaranteed that all | |
194 | * ->children lists are in the ascending order of ->serial_nr. | |
195 | * It's used to allow interrupting and resuming iterations. | |
196 | */ | |
197 | u64 serial_nr; | |
198 | ||
65dff759 LZ |
199 | /* |
200 | * This is a copy of dentry->d_name, and it's needed because | |
201 | * we can't use dentry->d_name in cgroup_path(). | |
202 | * | |
203 | * You must acquire rcu_read_lock() to access cgrp->name, and | |
204 | * the only place that can change it is rename(), which is | |
205 | * protected by parent dir's i_mutex. | |
206 | * | |
207 | * Normally you should use cgroup_name() wrapper rather than | |
208 | * access it directly. | |
209 | */ | |
210 | struct cgroup_name __rcu *name; | |
211 | ||
ddbcc7e8 PM |
212 | /* Private pointers for each registered subsystem */ |
213 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
214 | ||
215 | struct cgroupfs_root *root; | |
817929ec PM |
216 | |
217 | /* | |
69d0206c TH |
218 | * List of cgrp_cset_links pointing at css_sets with tasks in this |
219 | * cgroup. Protected by css_set_lock. | |
817929ec | 220 | */ |
69d0206c | 221 | struct list_head cset_links; |
81a6a5cd | 222 | |
b0ca5a84 | 223 | struct list_head allcg_node; /* cgroupfs_root->allcg_list */ |
8e3f6541 | 224 | struct list_head cft_q_node; /* used during cftype add/rm */ |
b0ca5a84 | 225 | |
81a6a5cd PM |
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; | |
be445626 | 242 | struct work_struct free_work; |
0dea1168 | 243 | |
25985edc | 244 | /* List of events which userspace want to receive */ |
0dea1168 KS |
245 | struct list_head event_list; |
246 | spinlock_t event_list_lock; | |
03b1cde6 AR |
247 | |
248 | /* directory xattrs */ | |
249 | struct simple_xattrs xattrs; | |
817929ec PM |
250 | }; |
251 | ||
25a7e684 TH |
252 | #define MAX_CGROUP_ROOT_NAMELEN 64 |
253 | ||
254 | /* cgroupfs_root->flags */ | |
255 | enum { | |
873fe09e TH |
256 | /* |
257 | * Unfortunately, cgroup core and various controllers are riddled | |
258 | * with idiosyncrasies and pointless options. The following flag, | |
259 | * when set, will force sane behavior - some options are forced on, | |
260 | * others are disallowed, and some controllers will change their | |
261 | * hierarchical or other behaviors. | |
262 | * | |
263 | * The set of behaviors affected by this flag are still being | |
264 | * determined and developed and the mount option for this flag is | |
265 | * prefixed with __DEVEL__. The prefix will be dropped once we | |
266 | * reach the point where all behaviors are compatible with the | |
267 | * planned unified hierarchy, which will automatically turn on this | |
268 | * flag. | |
269 | * | |
270 | * The followings are the behaviors currently affected this flag. | |
271 | * | |
272 | * - Mount options "noprefix" and "clone_children" are disallowed. | |
273 | * Also, cgroupfs file cgroup.clone_children is not created. | |
274 | * | |
275 | * - When mounting an existing superblock, mount options should | |
276 | * match. | |
277 | * | |
278 | * - Remount is disallowed. | |
279 | * | |
f00baae7 TH |
280 | * - memcg: use_hierarchy is on by default and the cgroup file for |
281 | * the flag is not created. | |
282 | * | |
873fe09e TH |
283 | * The followings are planned changes. |
284 | * | |
285 | * - release_agent will be disallowed once replacement notification | |
286 | * mechanism is implemented. | |
287 | */ | |
288 | CGRP_ROOT_SANE_BEHAVIOR = (1 << 0), | |
289 | ||
25a7e684 TH |
290 | CGRP_ROOT_NOPREFIX = (1 << 1), /* mounted subsystems have no named prefix */ |
291 | CGRP_ROOT_XATTR = (1 << 2), /* supports extended attributes */ | |
292 | }; | |
293 | ||
294 | /* | |
295 | * A cgroupfs_root represents the root of a cgroup hierarchy, and may be | |
296 | * associated with a superblock to form an active hierarchy. This is | |
297 | * internal to cgroup core. Don't access directly from controllers. | |
298 | */ | |
299 | struct cgroupfs_root { | |
300 | struct super_block *sb; | |
301 | ||
302 | /* | |
303 | * The bitmask of subsystems intended to be attached to this | |
304 | * hierarchy | |
305 | */ | |
306 | unsigned long subsys_mask; | |
307 | ||
308 | /* Unique id for this hierarchy. */ | |
309 | int hierarchy_id; | |
310 | ||
311 | /* The bitmask of subsystems currently attached to this hierarchy */ | |
312 | unsigned long actual_subsys_mask; | |
313 | ||
314 | /* A list running through the attached subsystems */ | |
315 | struct list_head subsys_list; | |
316 | ||
317 | /* The root cgroup for this hierarchy */ | |
318 | struct cgroup top_cgroup; | |
319 | ||
320 | /* Tracks how many cgroups are currently defined in hierarchy.*/ | |
321 | int number_of_cgroups; | |
322 | ||
323 | /* A list running through the active hierarchies */ | |
324 | struct list_head root_list; | |
325 | ||
326 | /* All cgroups on this root, cgroup_mutex protected */ | |
327 | struct list_head allcg_list; | |
328 | ||
329 | /* Hierarchy-specific flags */ | |
330 | unsigned long flags; | |
331 | ||
332 | /* IDs for cgroups in this hierarchy */ | |
333 | struct ida cgroup_ida; | |
334 | ||
335 | /* The path to use for release notifications. */ | |
336 | char release_agent_path[PATH_MAX]; | |
337 | ||
338 | /* The name for this hierarchy - may be empty */ | |
339 | char name[MAX_CGROUP_ROOT_NAMELEN]; | |
340 | }; | |
341 | ||
d20a390a PM |
342 | /* |
343 | * A css_set is a structure holding pointers to a set of | |
817929ec PM |
344 | * cgroup_subsys_state objects. This saves space in the task struct |
345 | * object and speeds up fork()/exit(), since a single inc/dec and a | |
d20a390a PM |
346 | * list_add()/del() can bump the reference count on the entire cgroup |
347 | * set for a task. | |
817929ec PM |
348 | */ |
349 | ||
350 | struct css_set { | |
351 | ||
352 | /* Reference count */ | |
146aa1bd | 353 | atomic_t refcount; |
817929ec | 354 | |
472b1053 LZ |
355 | /* |
356 | * List running through all cgroup groups in the same hash | |
357 | * slot. Protected by css_set_lock | |
358 | */ | |
359 | struct hlist_node hlist; | |
360 | ||
817929ec PM |
361 | /* |
362 | * List running through all tasks using this cgroup | |
363 | * group. Protected by css_set_lock | |
364 | */ | |
365 | struct list_head tasks; | |
366 | ||
367 | /* | |
69d0206c TH |
368 | * List of cgrp_cset_links pointing at cgroups referenced from this |
369 | * css_set. Protected by css_set_lock. | |
817929ec | 370 | */ |
69d0206c | 371 | struct list_head cgrp_links; |
817929ec PM |
372 | |
373 | /* | |
374 | * Set of subsystem states, one for each subsystem. This array | |
375 | * is immutable after creation apart from the init_css_set | |
cf5d5941 BB |
376 | * during subsystem registration (at boot time) and modular subsystem |
377 | * loading/unloading. | |
817929ec PM |
378 | */ |
379 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
c378369d BB |
380 | |
381 | /* For RCU-protected deletion */ | |
382 | struct rcu_head rcu_head; | |
ddbcc7e8 PM |
383 | }; |
384 | ||
91796569 PM |
385 | /* |
386 | * cgroup_map_cb is an abstract callback API for reporting map-valued | |
387 | * control files | |
388 | */ | |
389 | ||
390 | struct cgroup_map_cb { | |
391 | int (*fill)(struct cgroup_map_cb *cb, const char *key, u64 value); | |
392 | void *state; | |
393 | }; | |
394 | ||
d20a390a PM |
395 | /* |
396 | * struct cftype: handler definitions for cgroup control files | |
ddbcc7e8 PM |
397 | * |
398 | * When reading/writing to a file: | |
a043e3b2 | 399 | * - the cgroup to use is file->f_dentry->d_parent->d_fsdata |
ddbcc7e8 PM |
400 | * - the 'cftype' of the file is file->f_dentry->d_fsdata |
401 | */ | |
402 | ||
8e3f6541 TH |
403 | /* cftype->flags */ |
404 | #define CFTYPE_ONLY_ON_ROOT (1U << 0) /* only create on root cg */ | |
d0b2fdd2 | 405 | #define CFTYPE_NOT_ON_ROOT (1U << 1) /* don't create on root cg */ |
873fe09e | 406 | #define CFTYPE_INSANE (1U << 2) /* don't create if sane_behavior */ |
8e3f6541 TH |
407 | |
408 | #define MAX_CFTYPE_NAME 64 | |
409 | ||
ddbcc7e8 | 410 | struct cftype { |
d20a390a PM |
411 | /* |
412 | * By convention, the name should begin with the name of the | |
8e3f6541 TH |
413 | * subsystem, followed by a period. Zero length string indicates |
414 | * end of cftype array. | |
d20a390a | 415 | */ |
ddbcc7e8 PM |
416 | char name[MAX_CFTYPE_NAME]; |
417 | int private; | |
099fca32 LZ |
418 | /* |
419 | * If not 0, file mode is set to this value, otherwise it will | |
420 | * be figured out automatically | |
421 | */ | |
a5e7ed32 | 422 | umode_t mode; |
db3b1497 PM |
423 | |
424 | /* | |
425 | * If non-zero, defines the maximum length of string that can | |
426 | * be passed to write_string; defaults to 64 | |
427 | */ | |
428 | size_t max_write_len; | |
429 | ||
8e3f6541 TH |
430 | /* CFTYPE_* flags */ |
431 | unsigned int flags; | |
432 | ||
ce16b49d PM |
433 | int (*open)(struct inode *inode, struct file *file); |
434 | ssize_t (*read)(struct cgroup *cgrp, struct cftype *cft, | |
435 | struct file *file, | |
436 | char __user *buf, size_t nbytes, loff_t *ppos); | |
ddbcc7e8 | 437 | /* |
f4c753b7 | 438 | * read_u64() is a shortcut for the common case of returning a |
ddbcc7e8 PM |
439 | * single integer. Use it in place of read() |
440 | */ | |
ce16b49d | 441 | u64 (*read_u64)(struct cgroup *cgrp, struct cftype *cft); |
e73d2c61 PM |
442 | /* |
443 | * read_s64() is a signed version of read_u64() | |
444 | */ | |
ce16b49d | 445 | s64 (*read_s64)(struct cgroup *cgrp, struct cftype *cft); |
91796569 PM |
446 | /* |
447 | * read_map() is used for defining a map of key/value | |
448 | * pairs. It should call cb->fill(cb, key, value) for each | |
449 | * entry. The key/value pairs (and their ordering) should not | |
450 | * change between reboots. | |
451 | */ | |
ce16b49d PM |
452 | int (*read_map)(struct cgroup *cont, struct cftype *cft, |
453 | struct cgroup_map_cb *cb); | |
29486df3 SH |
454 | /* |
455 | * read_seq_string() is used for outputting a simple sequence | |
456 | * using seqfile. | |
457 | */ | |
ce16b49d PM |
458 | int (*read_seq_string)(struct cgroup *cont, struct cftype *cft, |
459 | struct seq_file *m); | |
91796569 | 460 | |
ce16b49d PM |
461 | ssize_t (*write)(struct cgroup *cgrp, struct cftype *cft, |
462 | struct file *file, | |
463 | const char __user *buf, size_t nbytes, loff_t *ppos); | |
355e0c48 PM |
464 | |
465 | /* | |
f4c753b7 | 466 | * write_u64() is a shortcut for the common case of accepting |
355e0c48 PM |
467 | * a single integer (as parsed by simple_strtoull) from |
468 | * userspace. Use in place of write(); return 0 or error. | |
469 | */ | |
ce16b49d | 470 | int (*write_u64)(struct cgroup *cgrp, struct cftype *cft, u64 val); |
e73d2c61 PM |
471 | /* |
472 | * write_s64() is a signed version of write_u64() | |
473 | */ | |
ce16b49d | 474 | int (*write_s64)(struct cgroup *cgrp, struct cftype *cft, s64 val); |
355e0c48 | 475 | |
db3b1497 PM |
476 | /* |
477 | * write_string() is passed a nul-terminated kernelspace | |
478 | * buffer of maximum length determined by max_write_len. | |
479 | * Returns 0 or -ve error code. | |
480 | */ | |
481 | int (*write_string)(struct cgroup *cgrp, struct cftype *cft, | |
482 | const char *buffer); | |
d447ea2f PE |
483 | /* |
484 | * trigger() callback can be used to get some kick from the | |
485 | * userspace, when the actual string written is not important | |
486 | * at all. The private field can be used to determine the | |
487 | * kick type for multiplexing. | |
488 | */ | |
489 | int (*trigger)(struct cgroup *cgrp, unsigned int event); | |
490 | ||
ce16b49d | 491 | int (*release)(struct inode *inode, struct file *file); |
0dea1168 KS |
492 | |
493 | /* | |
494 | * register_event() callback will be used to add new userspace | |
495 | * waiter for changes related to the cftype. Implement it if | |
496 | * you want to provide this functionality. Use eventfd_signal() | |
497 | * on eventfd to send notification to userspace. | |
498 | */ | |
499 | int (*register_event)(struct cgroup *cgrp, struct cftype *cft, | |
500 | struct eventfd_ctx *eventfd, const char *args); | |
501 | /* | |
502 | * unregister_event() callback will be called when userspace | |
503 | * closes the eventfd or on cgroup removing. | |
504 | * This callback must be implemented, if you want provide | |
505 | * notification functionality. | |
0dea1168 | 506 | */ |
907860ed | 507 | void (*unregister_event)(struct cgroup *cgrp, struct cftype *cft, |
0dea1168 | 508 | struct eventfd_ctx *eventfd); |
ddbcc7e8 PM |
509 | }; |
510 | ||
8e3f6541 TH |
511 | /* |
512 | * cftype_sets describe cftypes belonging to a subsystem and are chained at | |
513 | * cgroup_subsys->cftsets. Each cftset points to an array of cftypes | |
514 | * terminated by zero length name. | |
515 | */ | |
516 | struct cftype_set { | |
517 | struct list_head node; /* chained at subsys->cftsets */ | |
03b1cde6 | 518 | struct cftype *cfts; |
8e3f6541 TH |
519 | }; |
520 | ||
31a7df01 CW |
521 | struct cgroup_scanner { |
522 | struct cgroup *cg; | |
523 | int (*test_task)(struct task_struct *p, struct cgroup_scanner *scan); | |
524 | void (*process_task)(struct task_struct *p, | |
525 | struct cgroup_scanner *scan); | |
526 | struct ptr_heap *heap; | |
bd1a8ab7 | 527 | void *data; |
31a7df01 CW |
528 | }; |
529 | ||
873fe09e TH |
530 | /* |
531 | * See the comment above CGRP_ROOT_SANE_BEHAVIOR for details. This | |
532 | * function can be called as long as @cgrp is accessible. | |
533 | */ | |
534 | static inline bool cgroup_sane_behavior(const struct cgroup *cgrp) | |
535 | { | |
536 | return cgrp->root->flags & CGRP_ROOT_SANE_BEHAVIOR; | |
537 | } | |
538 | ||
65dff759 LZ |
539 | /* Caller should hold rcu_read_lock() */ |
540 | static inline const char *cgroup_name(const struct cgroup *cgrp) | |
541 | { | |
542 | return rcu_dereference(cgrp->name)->name; | |
543 | } | |
544 | ||
03b1cde6 AR |
545 | int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); |
546 | int cgroup_rm_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); | |
8e3f6541 | 547 | |
78574cf9 | 548 | bool cgroup_is_descendant(struct cgroup *cgrp, struct cgroup *ancestor); |
ddbcc7e8 | 549 | |
ffd2d883 | 550 | int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen); |
857a2beb TH |
551 | int task_cgroup_path_from_hierarchy(struct task_struct *task, int hierarchy_id, |
552 | char *buf, size_t buflen); | |
ddbcc7e8 | 553 | |
ffd2d883 | 554 | int cgroup_task_count(const struct cgroup *cgrp); |
bbcb81d0 | 555 | |
2f7ee569 TH |
556 | /* |
557 | * Control Group taskset, used to pass around set of tasks to cgroup_subsys | |
558 | * methods. | |
559 | */ | |
560 | struct cgroup_taskset; | |
561 | struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset); | |
562 | struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset); | |
563 | struct cgroup *cgroup_taskset_cur_cgroup(struct cgroup_taskset *tset); | |
564 | int cgroup_taskset_size(struct cgroup_taskset *tset); | |
565 | ||
566 | /** | |
567 | * cgroup_taskset_for_each - iterate cgroup_taskset | |
568 | * @task: the loop cursor | |
569 | * @skip_cgrp: skip if task's cgroup matches this, %NULL to iterate through all | |
570 | * @tset: taskset to iterate | |
571 | */ | |
572 | #define cgroup_taskset_for_each(task, skip_cgrp, tset) \ | |
573 | for ((task) = cgroup_taskset_first((tset)); (task); \ | |
574 | (task) = cgroup_taskset_next((tset))) \ | |
575 | if (!(skip_cgrp) || \ | |
576 | cgroup_taskset_cur_cgroup((tset)) != (skip_cgrp)) | |
577 | ||
21acb9ca TLSC |
578 | /* |
579 | * Control Group subsystem type. | |
580 | * See Documentation/cgroups/cgroups.txt for details | |
581 | */ | |
ddbcc7e8 PM |
582 | |
583 | struct cgroup_subsys { | |
92fb9748 TH |
584 | struct cgroup_subsys_state *(*css_alloc)(struct cgroup *cgrp); |
585 | int (*css_online)(struct cgroup *cgrp); | |
586 | void (*css_offline)(struct cgroup *cgrp); | |
587 | void (*css_free)(struct cgroup *cgrp); | |
588 | ||
761b3ef5 LZ |
589 | int (*can_attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); |
590 | void (*cancel_attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); | |
591 | void (*attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); | |
592 | void (*fork)(struct task_struct *task); | |
593 | void (*exit)(struct cgroup *cgrp, struct cgroup *old_cgrp, | |
594 | struct task_struct *task); | |
761b3ef5 | 595 | void (*bind)(struct cgroup *root); |
e5991371 | 596 | |
ddbcc7e8 | 597 | int subsys_id; |
8bab8dde | 598 | int disabled; |
ddbcc7e8 | 599 | int early_init; |
38460b48 KH |
600 | /* |
601 | * True if this subsys uses ID. ID is not available before cgroup_init() | |
602 | * (not available in early_init time.) | |
603 | */ | |
604 | bool use_id; | |
48ddbe19 | 605 | |
8c7f6edb TH |
606 | /* |
607 | * If %false, this subsystem is properly hierarchical - | |
608 | * configuration, resource accounting and restriction on a parent | |
609 | * cgroup cover those of its children. If %true, hierarchy support | |
610 | * is broken in some ways - some subsystems ignore hierarchy | |
611 | * completely while others are only implemented half-way. | |
612 | * | |
613 | * It's now disallowed to create nested cgroups if the subsystem is | |
614 | * broken and cgroup core will emit a warning message on such | |
615 | * cases. Eventually, all subsystems will be made properly | |
616 | * hierarchical and this will go away. | |
617 | */ | |
618 | bool broken_hierarchy; | |
619 | bool warned_broken_hierarchy; | |
620 | ||
ddbcc7e8 PM |
621 | #define MAX_CGROUP_TYPE_NAMELEN 32 |
622 | const char *name; | |
623 | ||
999cd8a4 PM |
624 | /* |
625 | * Link to parent, and list entry in parent's children. | |
6be96a5c | 626 | * Protected by cgroup_lock() |
999cd8a4 PM |
627 | */ |
628 | struct cgroupfs_root *root; | |
ddbcc7e8 | 629 | struct list_head sibling; |
38460b48 KH |
630 | /* used when use_id == true */ |
631 | struct idr idr; | |
42aee6c4 | 632 | spinlock_t id_lock; |
e6a1105b | 633 | |
8e3f6541 TH |
634 | /* list of cftype_sets */ |
635 | struct list_head cftsets; | |
636 | ||
637 | /* base cftypes, automatically [de]registered with subsys itself */ | |
638 | struct cftype *base_cftypes; | |
639 | struct cftype_set base_cftset; | |
640 | ||
e6a1105b BB |
641 | /* should be defined only by modular subsystems */ |
642 | struct module *module; | |
ddbcc7e8 PM |
643 | }; |
644 | ||
645 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _subsys; | |
5fc0b025 | 646 | #define IS_SUBSYS_ENABLED(option) IS_BUILTIN(option) |
ddbcc7e8 | 647 | #include <linux/cgroup_subsys.h> |
5fc0b025 | 648 | #undef IS_SUBSYS_ENABLED |
ddbcc7e8 PM |
649 | #undef SUBSYS |
650 | ||
651 | static inline struct cgroup_subsys_state *cgroup_subsys_state( | |
ffd2d883 | 652 | struct cgroup *cgrp, int subsys_id) |
ddbcc7e8 | 653 | { |
ffd2d883 | 654 | return cgrp->subsys[subsys_id]; |
ddbcc7e8 PM |
655 | } |
656 | ||
dc61b1d6 PZ |
657 | /* |
658 | * function to get the cgroup_subsys_state which allows for extra | |
659 | * rcu_dereference_check() conditions, such as locks used during the | |
660 | * cgroup_subsys::attach() methods. | |
661 | */ | |
2219449a TH |
662 | #ifdef CONFIG_PROVE_RCU |
663 | extern struct mutex cgroup_mutex; | |
dc61b1d6 | 664 | #define task_subsys_state_check(task, subsys_id, __c) \ |
2219449a TH |
665 | rcu_dereference_check((task)->cgroups->subsys[(subsys_id)], \ |
666 | lockdep_is_held(&(task)->alloc_lock) || \ | |
667 | lockdep_is_held(&cgroup_mutex) || (__c)) | |
668 | #else | |
dc61b1d6 | 669 | #define task_subsys_state_check(task, subsys_id, __c) \ |
2219449a TH |
670 | rcu_dereference((task)->cgroups->subsys[(subsys_id)]) |
671 | #endif | |
dc61b1d6 PZ |
672 | |
673 | static inline struct cgroup_subsys_state * | |
674 | task_subsys_state(struct task_struct *task, int subsys_id) | |
ddbcc7e8 | 675 | { |
dc61b1d6 | 676 | return task_subsys_state_check(task, subsys_id, false); |
ddbcc7e8 PM |
677 | } |
678 | ||
679 | static inline struct cgroup* task_cgroup(struct task_struct *task, | |
680 | int subsys_id) | |
681 | { | |
682 | return task_subsys_state(task, subsys_id)->cgroup; | |
683 | } | |
684 | ||
53fa5261 TH |
685 | struct cgroup *cgroup_next_sibling(struct cgroup *pos); |
686 | ||
574bd9f7 TH |
687 | /** |
688 | * cgroup_for_each_child - iterate through children of a cgroup | |
689 | * @pos: the cgroup * to use as the loop cursor | |
75501a6d | 690 | * @cgrp: cgroup whose children to walk |
574bd9f7 | 691 | * |
75501a6d | 692 | * Walk @cgrp's children. Must be called under rcu_read_lock(). A child |
92fb9748 TH |
693 | * cgroup which hasn't finished ->css_online() or already has finished |
694 | * ->css_offline() may show up during traversal and it's each subsystem's | |
574bd9f7 TH |
695 | * responsibility to verify that each @pos is alive. |
696 | * | |
92fb9748 TH |
697 | * If a subsystem synchronizes against the parent in its ->css_online() and |
698 | * before starting iterating, a cgroup which finished ->css_online() is | |
699 | * guaranteed to be visible in the future iterations. | |
75501a6d TH |
700 | * |
701 | * It is allowed to temporarily drop RCU read lock during iteration. The | |
702 | * caller is responsible for ensuring that @pos remains accessible until | |
703 | * the start of the next iteration by, for example, bumping the css refcnt. | |
574bd9f7 | 704 | */ |
75501a6d TH |
705 | #define cgroup_for_each_child(pos, cgrp) \ |
706 | for ((pos) = list_first_or_null_rcu(&(cgrp)->children, \ | |
707 | struct cgroup, sibling); \ | |
708 | (pos); (pos) = cgroup_next_sibling((pos))) | |
574bd9f7 TH |
709 | |
710 | struct cgroup *cgroup_next_descendant_pre(struct cgroup *pos, | |
711 | struct cgroup *cgroup); | |
12a9d2fe | 712 | struct cgroup *cgroup_rightmost_descendant(struct cgroup *pos); |
574bd9f7 TH |
713 | |
714 | /** | |
715 | * cgroup_for_each_descendant_pre - pre-order walk of a cgroup's descendants | |
716 | * @pos: the cgroup * to use as the loop cursor | |
717 | * @cgroup: cgroup whose descendants to walk | |
718 | * | |
719 | * Walk @cgroup's descendants. Must be called under rcu_read_lock(). A | |
92fb9748 TH |
720 | * descendant cgroup which hasn't finished ->css_online() or already has |
721 | * finished ->css_offline() may show up during traversal and it's each | |
574bd9f7 TH |
722 | * subsystem's responsibility to verify that each @pos is alive. |
723 | * | |
92fb9748 TH |
724 | * If a subsystem synchronizes against the parent in its ->css_online() and |
725 | * before starting iterating, and synchronizes against @pos on each | |
7805d000 | 726 | * iteration, any descendant cgroup which finished ->css_online() is |
574bd9f7 TH |
727 | * guaranteed to be visible in the future iterations. |
728 | * | |
729 | * In other words, the following guarantees that a descendant can't escape | |
730 | * state updates of its ancestors. | |
731 | * | |
92fb9748 | 732 | * my_online(@cgrp) |
574bd9f7 TH |
733 | * { |
734 | * Lock @cgrp->parent and @cgrp; | |
735 | * Inherit state from @cgrp->parent; | |
736 | * Unlock both. | |
737 | * } | |
738 | * | |
739 | * my_update_state(@cgrp) | |
740 | * { | |
741 | * Lock @cgrp; | |
742 | * Update @cgrp's state; | |
743 | * Unlock @cgrp; | |
744 | * | |
745 | * cgroup_for_each_descendant_pre(@pos, @cgrp) { | |
746 | * Lock @pos; | |
747 | * Verify @pos is alive and inherit state from @pos->parent; | |
748 | * Unlock @pos; | |
749 | * } | |
750 | * } | |
751 | * | |
752 | * As long as the inheriting step, including checking the parent state, is | |
753 | * enclosed inside @pos locking, double-locking the parent isn't necessary | |
754 | * while inheriting. The state update to the parent is guaranteed to be | |
755 | * visible by walking order and, as long as inheriting operations to the | |
756 | * same @pos are atomic to each other, multiple updates racing each other | |
757 | * still result in the correct state. It's guaranateed that at least one | |
758 | * inheritance happens for any cgroup after the latest update to its | |
759 | * parent. | |
760 | * | |
761 | * If checking parent's state requires locking the parent, each inheriting | |
762 | * iteration should lock and unlock both @pos->parent and @pos. | |
763 | * | |
764 | * Alternatively, a subsystem may choose to use a single global lock to | |
92fb9748 | 765 | * synchronize ->css_online() and ->css_offline() against tree-walking |
574bd9f7 | 766 | * operations. |
75501a6d TH |
767 | * |
768 | * It is allowed to temporarily drop RCU read lock during iteration. The | |
769 | * caller is responsible for ensuring that @pos remains accessible until | |
770 | * the start of the next iteration by, for example, bumping the css refcnt. | |
574bd9f7 TH |
771 | */ |
772 | #define cgroup_for_each_descendant_pre(pos, cgroup) \ | |
773 | for (pos = cgroup_next_descendant_pre(NULL, (cgroup)); (pos); \ | |
774 | pos = cgroup_next_descendant_pre((pos), (cgroup))) | |
775 | ||
776 | struct cgroup *cgroup_next_descendant_post(struct cgroup *pos, | |
777 | struct cgroup *cgroup); | |
778 | ||
779 | /** | |
780 | * cgroup_for_each_descendant_post - post-order walk of a cgroup's descendants | |
781 | * @pos: the cgroup * to use as the loop cursor | |
782 | * @cgroup: cgroup whose descendants to walk | |
783 | * | |
784 | * Similar to cgroup_for_each_descendant_pre() but performs post-order | |
785 | * traversal instead. Note that the walk visibility guarantee described in | |
786 | * pre-order walk doesn't apply the same to post-order walks. | |
787 | */ | |
788 | #define cgroup_for_each_descendant_post(pos, cgroup) \ | |
789 | for (pos = cgroup_next_descendant_post(NULL, (cgroup)); (pos); \ | |
790 | pos = cgroup_next_descendant_post((pos), (cgroup))) | |
791 | ||
817929ec PM |
792 | /* A cgroup_iter should be treated as an opaque object */ |
793 | struct cgroup_iter { | |
69d0206c | 794 | struct list_head *cset_link; |
817929ec PM |
795 | struct list_head *task; |
796 | }; | |
797 | ||
d20a390a PM |
798 | /* |
799 | * To iterate across the tasks in a cgroup: | |
817929ec | 800 | * |
b595076a | 801 | * 1) call cgroup_iter_start to initialize an iterator |
817929ec PM |
802 | * |
803 | * 2) call cgroup_iter_next() to retrieve member tasks until it | |
804 | * returns NULL or until you want to end the iteration | |
805 | * | |
806 | * 3) call cgroup_iter_end() to destroy the iterator. | |
31a7df01 | 807 | * |
d20a390a PM |
808 | * Or, call cgroup_scan_tasks() to iterate through every task in a |
809 | * cgroup - cgroup_scan_tasks() holds the css_set_lock when calling | |
810 | * the test_task() callback, but not while calling the process_task() | |
811 | * callback. | |
817929ec | 812 | */ |
ffd2d883 LZ |
813 | void cgroup_iter_start(struct cgroup *cgrp, struct cgroup_iter *it); |
814 | struct task_struct *cgroup_iter_next(struct cgroup *cgrp, | |
817929ec | 815 | struct cgroup_iter *it); |
ffd2d883 | 816 | void cgroup_iter_end(struct cgroup *cgrp, struct cgroup_iter *it); |
31a7df01 | 817 | int cgroup_scan_tasks(struct cgroup_scanner *scan); |
31583bb0 | 818 | int cgroup_attach_task_all(struct task_struct *from, struct task_struct *); |
8cc99345 | 819 | int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from); |
31583bb0 | 820 | |
38460b48 KH |
821 | /* |
822 | * CSS ID is ID for cgroup_subsys_state structs under subsys. This only works | |
823 | * if cgroup_subsys.use_id == true. It can be used for looking up and scanning. | |
824 | * CSS ID is assigned at cgroup allocation (create) automatically | |
825 | * and removed when subsys calls free_css_id() function. This is because | |
826 | * the lifetime of cgroup_subsys_state is subsys's matter. | |
827 | * | |
828 | * Looking up and scanning function should be called under rcu_read_lock(). | |
6be96a5c | 829 | * Taking cgroup_mutex is not necessary for following calls. |
38460b48 KH |
830 | * But the css returned by this routine can be "not populated yet" or "being |
831 | * destroyed". The caller should check css and cgroup's status. | |
832 | */ | |
833 | ||
834 | /* | |
835 | * Typically Called at ->destroy(), or somewhere the subsys frees | |
836 | * cgroup_subsys_state. | |
837 | */ | |
838 | void free_css_id(struct cgroup_subsys *ss, struct cgroup_subsys_state *css); | |
839 | ||
840 | /* Find a cgroup_subsys_state which has given ID */ | |
841 | ||
842 | struct cgroup_subsys_state *css_lookup(struct cgroup_subsys *ss, int id); | |
843 | ||
38460b48 KH |
844 | /* Returns true if root is ancestor of cg */ |
845 | bool css_is_ancestor(struct cgroup_subsys_state *cg, | |
0b7f569e | 846 | const struct cgroup_subsys_state *root); |
38460b48 KH |
847 | |
848 | /* Get id and depth of css */ | |
849 | unsigned short css_id(struct cgroup_subsys_state *css); | |
e5d1367f | 850 | struct cgroup_subsys_state *cgroup_css_from_dir(struct file *f, int id); |
38460b48 | 851 | |
ddbcc7e8 PM |
852 | #else /* !CONFIG_CGROUPS */ |
853 | ||
854 | static inline int cgroup_init_early(void) { return 0; } | |
855 | static inline int cgroup_init(void) { return 0; } | |
b4f48b63 | 856 | static inline void cgroup_fork(struct task_struct *p) {} |
817929ec | 857 | static inline void cgroup_post_fork(struct task_struct *p) {} |
b4f48b63 | 858 | static inline void cgroup_exit(struct task_struct *p, int callbacks) {} |
ddbcc7e8 | 859 | |
846c7bb0 BS |
860 | static inline int cgroupstats_build(struct cgroupstats *stats, |
861 | struct dentry *dentry) | |
862 | { | |
863 | return -EINVAL; | |
864 | } | |
ddbcc7e8 | 865 | |
d7926ee3 | 866 | /* No cgroups - nothing to do */ |
31583bb0 MT |
867 | static inline int cgroup_attach_task_all(struct task_struct *from, |
868 | struct task_struct *t) | |
869 | { | |
870 | return 0; | |
871 | } | |
d7926ee3 | 872 | |
ddbcc7e8 PM |
873 | #endif /* !CONFIG_CGROUPS */ |
874 | ||
875 | #endif /* _LINUX_CGROUP_H */ |