]>
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> |
ddbcc7e8 PM |
22 | |
23 | #ifdef CONFIG_CGROUPS | |
24 | ||
25 | struct cgroupfs_root; | |
26 | struct cgroup_subsys; | |
27 | struct inode; | |
84eea842 | 28 | struct cgroup; |
38460b48 | 29 | struct css_id; |
ddbcc7e8 PM |
30 | |
31 | extern int cgroup_init_early(void); | |
32 | extern int cgroup_init(void); | |
ddbcc7e8 | 33 | extern void cgroup_lock(void); |
d11c563d | 34 | extern int cgroup_lock_is_held(void); |
84eea842 | 35 | extern bool cgroup_lock_live_group(struct cgroup *cgrp); |
ddbcc7e8 | 36 | extern void cgroup_unlock(void); |
b4f48b63 | 37 | extern void cgroup_fork(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 | 48 | #define SUBSYS(_x) _x ## _subsys_id, |
8a8e04df | 49 | #define IS_SUBSYS_ENABLED(option) IS_ENABLED(option) |
817929ec PM |
50 | enum cgroup_subsys_id { |
51 | #include <linux/cgroup_subsys.h> | |
a6f00298 | 52 | CGROUP_SUBSYS_COUNT, |
817929ec | 53 | }; |
5fc0b025 | 54 | #undef IS_SUBSYS_ENABLED |
817929ec PM |
55 | #undef SUBSYS |
56 | ||
ddbcc7e8 PM |
57 | /* Per-subsystem/per-cgroup state maintained by the system. */ |
58 | struct cgroup_subsys_state { | |
d20a390a PM |
59 | /* |
60 | * The cgroup that this subsystem is attached to. Useful | |
ddbcc7e8 | 61 | * for subsystems that want to know about the cgroup |
d20a390a PM |
62 | * hierarchy structure |
63 | */ | |
ddbcc7e8 PM |
64 | struct cgroup *cgroup; |
65 | ||
d20a390a PM |
66 | /* |
67 | * State maintained by the cgroup system to allow subsystems | |
e7c5ec91 | 68 | * to be "busy". Should be accessed via css_get(), |
d20a390a PM |
69 | * css_tryget() and and css_put(). |
70 | */ | |
ddbcc7e8 PM |
71 | |
72 | atomic_t refcnt; | |
73 | ||
74 | unsigned long flags; | |
38460b48 | 75 | /* ID for this css, if possible */ |
2c392b8c | 76 | struct css_id __rcu *id; |
48ddbe19 TH |
77 | |
78 | /* Used to put @cgroup->dentry on the last css_put() */ | |
79 | struct work_struct dput_work; | |
ddbcc7e8 PM |
80 | }; |
81 | ||
82 | /* bits in struct cgroup_subsys_state flags field */ | |
83 | enum { | |
38b53aba | 84 | CSS_ROOT = (1 << 0), /* this CSS is the root of the subsystem */ |
92fb9748 | 85 | CSS_ONLINE = (1 << 1), /* between ->css_online() and ->css_offline() */ |
ddbcc7e8 PM |
86 | }; |
87 | ||
d7b9fff7 DN |
88 | /* Caller must verify that the css is not for root cgroup */ |
89 | static inline void __css_get(struct cgroup_subsys_state *css, int count) | |
90 | { | |
91 | atomic_add(count, &css->refcnt); | |
92 | } | |
93 | ||
ddbcc7e8 | 94 | /* |
e7c5ec91 PM |
95 | * Call css_get() to hold a reference on the css; it can be used |
96 | * for a reference obtained via: | |
97 | * - an existing ref-counted reference to the css | |
98 | * - task->cgroups for a locked task | |
ddbcc7e8 PM |
99 | */ |
100 | ||
101 | static inline void css_get(struct cgroup_subsys_state *css) | |
102 | { | |
103 | /* We don't need to reference count the root state */ | |
38b53aba | 104 | if (!(css->flags & CSS_ROOT)) |
d7b9fff7 | 105 | __css_get(css, 1); |
ddbcc7e8 | 106 | } |
e7c5ec91 | 107 | |
e7c5ec91 PM |
108 | /* |
109 | * Call css_tryget() to take a reference on a css if your existing | |
110 | * (known-valid) reference isn't already ref-counted. Returns false if | |
111 | * the css has been destroyed. | |
112 | */ | |
113 | ||
28b4c27b | 114 | extern bool __css_tryget(struct cgroup_subsys_state *css); |
e7c5ec91 PM |
115 | static inline bool css_tryget(struct cgroup_subsys_state *css) |
116 | { | |
38b53aba | 117 | if (css->flags & CSS_ROOT) |
e7c5ec91 | 118 | return true; |
28b4c27b | 119 | return __css_tryget(css); |
e7c5ec91 PM |
120 | } |
121 | ||
ddbcc7e8 PM |
122 | /* |
123 | * css_put() should be called to release a reference taken by | |
e7c5ec91 | 124 | * css_get() or css_tryget() |
ddbcc7e8 PM |
125 | */ |
126 | ||
28b4c27b | 127 | extern void __css_put(struct cgroup_subsys_state *css); |
ddbcc7e8 PM |
128 | static inline void css_put(struct cgroup_subsys_state *css) |
129 | { | |
38b53aba | 130 | if (!(css->flags & CSS_ROOT)) |
28b4c27b | 131 | __css_put(css); |
ddbcc7e8 PM |
132 | } |
133 | ||
3116f0e3 PM |
134 | /* bits in struct cgroup flags field */ |
135 | enum { | |
136 | /* Control Group is dead */ | |
137 | CGRP_REMOVED, | |
d20a390a PM |
138 | /* |
139 | * Control Group has previously had a child cgroup or a task, | |
140 | * but no longer (only if CGRP_NOTIFY_ON_RELEASE is set) | |
141 | */ | |
3116f0e3 PM |
142 | CGRP_RELEASABLE, |
143 | /* Control Group requires release notifications to userspace */ | |
144 | CGRP_NOTIFY_ON_RELEASE, | |
97978e6d DL |
145 | /* |
146 | * Clone cgroup values when creating a new child cgroup | |
147 | */ | |
148 | CGRP_CLONE_CHILDREN, | |
3116f0e3 PM |
149 | }; |
150 | ||
ddbcc7e8 PM |
151 | struct cgroup { |
152 | unsigned long flags; /* "unsigned long" so bitops work */ | |
153 | ||
d20a390a PM |
154 | /* |
155 | * count users of this cgroup. >0 means busy, but doesn't | |
156 | * necessarily indicate the number of tasks in the cgroup | |
157 | */ | |
ddbcc7e8 PM |
158 | atomic_t count; |
159 | ||
160 | /* | |
161 | * We link our 'sibling' struct into our parent's 'children'. | |
162 | * Our children link their 'sibling' into our 'children'. | |
163 | */ | |
164 | struct list_head sibling; /* my parent's children */ | |
165 | struct list_head children; /* my children */ | |
05ef1d7c | 166 | struct list_head files; /* my files */ |
ddbcc7e8 | 167 | |
d20a390a | 168 | struct cgroup *parent; /* my parent */ |
febfcef6 | 169 | struct dentry *dentry; /* cgroup fs entry, RCU protected */ |
ddbcc7e8 PM |
170 | |
171 | /* Private pointers for each registered subsystem */ | |
172 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
173 | ||
174 | struct cgroupfs_root *root; | |
175 | struct cgroup *top_cgroup; | |
817929ec PM |
176 | |
177 | /* | |
178 | * List of cg_cgroup_links pointing at css_sets with | |
179 | * tasks in this cgroup. Protected by css_set_lock | |
180 | */ | |
181 | struct list_head css_sets; | |
81a6a5cd | 182 | |
b0ca5a84 | 183 | struct list_head allcg_node; /* cgroupfs_root->allcg_list */ |
8e3f6541 | 184 | struct list_head cft_q_node; /* used during cftype add/rm */ |
b0ca5a84 | 185 | |
81a6a5cd PM |
186 | /* |
187 | * Linked list running through all cgroups that can | |
188 | * potentially be reaped by the release agent. Protected by | |
189 | * release_list_lock | |
190 | */ | |
191 | struct list_head release_list; | |
cc31edce | 192 | |
72a8cb30 BB |
193 | /* |
194 | * list of pidlists, up to two for each namespace (one for procs, one | |
195 | * for tasks); created on demand. | |
196 | */ | |
197 | struct list_head pidlists; | |
198 | struct mutex pidlist_mutex; | |
a47295e6 PM |
199 | |
200 | /* For RCU-protected deletion */ | |
201 | struct rcu_head rcu_head; | |
0dea1168 | 202 | |
25985edc | 203 | /* List of events which userspace want to receive */ |
0dea1168 KS |
204 | struct list_head event_list; |
205 | spinlock_t event_list_lock; | |
03b1cde6 AR |
206 | |
207 | /* directory xattrs */ | |
208 | struct simple_xattrs xattrs; | |
817929ec PM |
209 | }; |
210 | ||
d20a390a PM |
211 | /* |
212 | * A css_set is a structure holding pointers to a set of | |
817929ec PM |
213 | * cgroup_subsys_state objects. This saves space in the task struct |
214 | * object and speeds up fork()/exit(), since a single inc/dec and a | |
d20a390a PM |
215 | * list_add()/del() can bump the reference count on the entire cgroup |
216 | * set for a task. | |
817929ec PM |
217 | */ |
218 | ||
219 | struct css_set { | |
220 | ||
221 | /* Reference count */ | |
146aa1bd | 222 | atomic_t refcount; |
817929ec | 223 | |
472b1053 LZ |
224 | /* |
225 | * List running through all cgroup groups in the same hash | |
226 | * slot. Protected by css_set_lock | |
227 | */ | |
228 | struct hlist_node hlist; | |
229 | ||
817929ec PM |
230 | /* |
231 | * List running through all tasks using this cgroup | |
232 | * group. Protected by css_set_lock | |
233 | */ | |
234 | struct list_head tasks; | |
235 | ||
236 | /* | |
237 | * List of cg_cgroup_link objects on link chains from | |
238 | * cgroups referenced from this css_set. Protected by | |
239 | * css_set_lock | |
240 | */ | |
241 | struct list_head cg_links; | |
242 | ||
243 | /* | |
244 | * Set of subsystem states, one for each subsystem. This array | |
245 | * is immutable after creation apart from the init_css_set | |
cf5d5941 BB |
246 | * during subsystem registration (at boot time) and modular subsystem |
247 | * loading/unloading. | |
817929ec PM |
248 | */ |
249 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
c378369d BB |
250 | |
251 | /* For RCU-protected deletion */ | |
252 | struct rcu_head rcu_head; | |
ddbcc7e8 PM |
253 | }; |
254 | ||
91796569 PM |
255 | /* |
256 | * cgroup_map_cb is an abstract callback API for reporting map-valued | |
257 | * control files | |
258 | */ | |
259 | ||
260 | struct cgroup_map_cb { | |
261 | int (*fill)(struct cgroup_map_cb *cb, const char *key, u64 value); | |
262 | void *state; | |
263 | }; | |
264 | ||
d20a390a PM |
265 | /* |
266 | * struct cftype: handler definitions for cgroup control files | |
ddbcc7e8 PM |
267 | * |
268 | * When reading/writing to a file: | |
a043e3b2 | 269 | * - the cgroup to use is file->f_dentry->d_parent->d_fsdata |
ddbcc7e8 PM |
270 | * - the 'cftype' of the file is file->f_dentry->d_fsdata |
271 | */ | |
272 | ||
8e3f6541 TH |
273 | /* cftype->flags */ |
274 | #define CFTYPE_ONLY_ON_ROOT (1U << 0) /* only create on root cg */ | |
275 | #define CFTYPE_NOT_ON_ROOT (1U << 1) /* don't create onp root cg */ | |
276 | ||
277 | #define MAX_CFTYPE_NAME 64 | |
278 | ||
ddbcc7e8 | 279 | struct cftype { |
d20a390a PM |
280 | /* |
281 | * By convention, the name should begin with the name of the | |
8e3f6541 TH |
282 | * subsystem, followed by a period. Zero length string indicates |
283 | * end of cftype array. | |
d20a390a | 284 | */ |
ddbcc7e8 PM |
285 | char name[MAX_CFTYPE_NAME]; |
286 | int private; | |
099fca32 LZ |
287 | /* |
288 | * If not 0, file mode is set to this value, otherwise it will | |
289 | * be figured out automatically | |
290 | */ | |
a5e7ed32 | 291 | umode_t mode; |
db3b1497 PM |
292 | |
293 | /* | |
294 | * If non-zero, defines the maximum length of string that can | |
295 | * be passed to write_string; defaults to 64 | |
296 | */ | |
297 | size_t max_write_len; | |
298 | ||
8e3f6541 TH |
299 | /* CFTYPE_* flags */ |
300 | unsigned int flags; | |
301 | ||
03b1cde6 AR |
302 | /* file xattrs */ |
303 | struct simple_xattrs xattrs; | |
304 | ||
ce16b49d PM |
305 | int (*open)(struct inode *inode, struct file *file); |
306 | ssize_t (*read)(struct cgroup *cgrp, struct cftype *cft, | |
307 | struct file *file, | |
308 | char __user *buf, size_t nbytes, loff_t *ppos); | |
ddbcc7e8 | 309 | /* |
f4c753b7 | 310 | * read_u64() is a shortcut for the common case of returning a |
ddbcc7e8 PM |
311 | * single integer. Use it in place of read() |
312 | */ | |
ce16b49d | 313 | u64 (*read_u64)(struct cgroup *cgrp, struct cftype *cft); |
e73d2c61 PM |
314 | /* |
315 | * read_s64() is a signed version of read_u64() | |
316 | */ | |
ce16b49d | 317 | s64 (*read_s64)(struct cgroup *cgrp, struct cftype *cft); |
91796569 PM |
318 | /* |
319 | * read_map() is used for defining a map of key/value | |
320 | * pairs. It should call cb->fill(cb, key, value) for each | |
321 | * entry. The key/value pairs (and their ordering) should not | |
322 | * change between reboots. | |
323 | */ | |
ce16b49d PM |
324 | int (*read_map)(struct cgroup *cont, struct cftype *cft, |
325 | struct cgroup_map_cb *cb); | |
29486df3 SH |
326 | /* |
327 | * read_seq_string() is used for outputting a simple sequence | |
328 | * using seqfile. | |
329 | */ | |
ce16b49d PM |
330 | int (*read_seq_string)(struct cgroup *cont, struct cftype *cft, |
331 | struct seq_file *m); | |
91796569 | 332 | |
ce16b49d PM |
333 | ssize_t (*write)(struct cgroup *cgrp, struct cftype *cft, |
334 | struct file *file, | |
335 | const char __user *buf, size_t nbytes, loff_t *ppos); | |
355e0c48 PM |
336 | |
337 | /* | |
f4c753b7 | 338 | * write_u64() is a shortcut for the common case of accepting |
355e0c48 PM |
339 | * a single integer (as parsed by simple_strtoull) from |
340 | * userspace. Use in place of write(); return 0 or error. | |
341 | */ | |
ce16b49d | 342 | int (*write_u64)(struct cgroup *cgrp, struct cftype *cft, u64 val); |
e73d2c61 PM |
343 | /* |
344 | * write_s64() is a signed version of write_u64() | |
345 | */ | |
ce16b49d | 346 | int (*write_s64)(struct cgroup *cgrp, struct cftype *cft, s64 val); |
355e0c48 | 347 | |
db3b1497 PM |
348 | /* |
349 | * write_string() is passed a nul-terminated kernelspace | |
350 | * buffer of maximum length determined by max_write_len. | |
351 | * Returns 0 or -ve error code. | |
352 | */ | |
353 | int (*write_string)(struct cgroup *cgrp, struct cftype *cft, | |
354 | const char *buffer); | |
d447ea2f PE |
355 | /* |
356 | * trigger() callback can be used to get some kick from the | |
357 | * userspace, when the actual string written is not important | |
358 | * at all. The private field can be used to determine the | |
359 | * kick type for multiplexing. | |
360 | */ | |
361 | int (*trigger)(struct cgroup *cgrp, unsigned int event); | |
362 | ||
ce16b49d | 363 | int (*release)(struct inode *inode, struct file *file); |
0dea1168 KS |
364 | |
365 | /* | |
366 | * register_event() callback will be used to add new userspace | |
367 | * waiter for changes related to the cftype. Implement it if | |
368 | * you want to provide this functionality. Use eventfd_signal() | |
369 | * on eventfd to send notification to userspace. | |
370 | */ | |
371 | int (*register_event)(struct cgroup *cgrp, struct cftype *cft, | |
372 | struct eventfd_ctx *eventfd, const char *args); | |
373 | /* | |
374 | * unregister_event() callback will be called when userspace | |
375 | * closes the eventfd or on cgroup removing. | |
376 | * This callback must be implemented, if you want provide | |
377 | * notification functionality. | |
0dea1168 | 378 | */ |
907860ed | 379 | void (*unregister_event)(struct cgroup *cgrp, struct cftype *cft, |
0dea1168 | 380 | struct eventfd_ctx *eventfd); |
ddbcc7e8 PM |
381 | }; |
382 | ||
8e3f6541 TH |
383 | /* |
384 | * cftype_sets describe cftypes belonging to a subsystem and are chained at | |
385 | * cgroup_subsys->cftsets. Each cftset points to an array of cftypes | |
386 | * terminated by zero length name. | |
387 | */ | |
388 | struct cftype_set { | |
389 | struct list_head node; /* chained at subsys->cftsets */ | |
03b1cde6 | 390 | struct cftype *cfts; |
8e3f6541 TH |
391 | }; |
392 | ||
31a7df01 CW |
393 | struct cgroup_scanner { |
394 | struct cgroup *cg; | |
395 | int (*test_task)(struct task_struct *p, struct cgroup_scanner *scan); | |
396 | void (*process_task)(struct task_struct *p, | |
397 | struct cgroup_scanner *scan); | |
398 | struct ptr_heap *heap; | |
bd1a8ab7 | 399 | void *data; |
31a7df01 CW |
400 | }; |
401 | ||
03b1cde6 AR |
402 | int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); |
403 | int cgroup_rm_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); | |
8e3f6541 | 404 | |
ffd2d883 | 405 | int cgroup_is_removed(const struct cgroup *cgrp); |
ddbcc7e8 | 406 | |
ffd2d883 | 407 | int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen); |
ddbcc7e8 | 408 | |
ffd2d883 | 409 | int cgroup_task_count(const struct cgroup *cgrp); |
bbcb81d0 | 410 | |
313e924c GN |
411 | /* Return true if cgrp is a descendant of the task's cgroup */ |
412 | int cgroup_is_descendant(const struct cgroup *cgrp, struct task_struct *task); | |
ddbcc7e8 | 413 | |
2f7ee569 TH |
414 | /* |
415 | * Control Group taskset, used to pass around set of tasks to cgroup_subsys | |
416 | * methods. | |
417 | */ | |
418 | struct cgroup_taskset; | |
419 | struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset); | |
420 | struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset); | |
421 | struct cgroup *cgroup_taskset_cur_cgroup(struct cgroup_taskset *tset); | |
422 | int cgroup_taskset_size(struct cgroup_taskset *tset); | |
423 | ||
424 | /** | |
425 | * cgroup_taskset_for_each - iterate cgroup_taskset | |
426 | * @task: the loop cursor | |
427 | * @skip_cgrp: skip if task's cgroup matches this, %NULL to iterate through all | |
428 | * @tset: taskset to iterate | |
429 | */ | |
430 | #define cgroup_taskset_for_each(task, skip_cgrp, tset) \ | |
431 | for ((task) = cgroup_taskset_first((tset)); (task); \ | |
432 | (task) = cgroup_taskset_next((tset))) \ | |
433 | if (!(skip_cgrp) || \ | |
434 | cgroup_taskset_cur_cgroup((tset)) != (skip_cgrp)) | |
435 | ||
21acb9ca TLSC |
436 | /* |
437 | * Control Group subsystem type. | |
438 | * See Documentation/cgroups/cgroups.txt for details | |
439 | */ | |
ddbcc7e8 PM |
440 | |
441 | struct cgroup_subsys { | |
92fb9748 TH |
442 | struct cgroup_subsys_state *(*css_alloc)(struct cgroup *cgrp); |
443 | int (*css_online)(struct cgroup *cgrp); | |
444 | void (*css_offline)(struct cgroup *cgrp); | |
445 | void (*css_free)(struct cgroup *cgrp); | |
446 | ||
761b3ef5 LZ |
447 | int (*can_attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); |
448 | void (*cancel_attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); | |
449 | void (*attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); | |
450 | void (*fork)(struct task_struct *task); | |
451 | void (*exit)(struct cgroup *cgrp, struct cgroup *old_cgrp, | |
452 | struct task_struct *task); | |
761b3ef5 LZ |
453 | void (*post_clone)(struct cgroup *cgrp); |
454 | void (*bind)(struct cgroup *root); | |
e5991371 | 455 | |
ddbcc7e8 PM |
456 | int subsys_id; |
457 | int active; | |
8bab8dde | 458 | int disabled; |
ddbcc7e8 | 459 | int early_init; |
38460b48 KH |
460 | /* |
461 | * True if this subsys uses ID. ID is not available before cgroup_init() | |
462 | * (not available in early_init time.) | |
463 | */ | |
464 | bool use_id; | |
48ddbe19 | 465 | |
8c7f6edb TH |
466 | /* |
467 | * If %false, this subsystem is properly hierarchical - | |
468 | * configuration, resource accounting and restriction on a parent | |
469 | * cgroup cover those of its children. If %true, hierarchy support | |
470 | * is broken in some ways - some subsystems ignore hierarchy | |
471 | * completely while others are only implemented half-way. | |
472 | * | |
473 | * It's now disallowed to create nested cgroups if the subsystem is | |
474 | * broken and cgroup core will emit a warning message on such | |
475 | * cases. Eventually, all subsystems will be made properly | |
476 | * hierarchical and this will go away. | |
477 | */ | |
478 | bool broken_hierarchy; | |
479 | bool warned_broken_hierarchy; | |
480 | ||
ddbcc7e8 PM |
481 | #define MAX_CGROUP_TYPE_NAMELEN 32 |
482 | const char *name; | |
483 | ||
999cd8a4 PM |
484 | /* |
485 | * Link to parent, and list entry in parent's children. | |
6be96a5c | 486 | * Protected by cgroup_lock() |
999cd8a4 PM |
487 | */ |
488 | struct cgroupfs_root *root; | |
ddbcc7e8 | 489 | struct list_head sibling; |
38460b48 KH |
490 | /* used when use_id == true */ |
491 | struct idr idr; | |
42aee6c4 | 492 | spinlock_t id_lock; |
e6a1105b | 493 | |
8e3f6541 TH |
494 | /* list of cftype_sets */ |
495 | struct list_head cftsets; | |
496 | ||
497 | /* base cftypes, automatically [de]registered with subsys itself */ | |
498 | struct cftype *base_cftypes; | |
499 | struct cftype_set base_cftset; | |
500 | ||
e6a1105b BB |
501 | /* should be defined only by modular subsystems */ |
502 | struct module *module; | |
ddbcc7e8 PM |
503 | }; |
504 | ||
505 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _subsys; | |
5fc0b025 | 506 | #define IS_SUBSYS_ENABLED(option) IS_BUILTIN(option) |
ddbcc7e8 | 507 | #include <linux/cgroup_subsys.h> |
5fc0b025 | 508 | #undef IS_SUBSYS_ENABLED |
ddbcc7e8 PM |
509 | #undef SUBSYS |
510 | ||
511 | static inline struct cgroup_subsys_state *cgroup_subsys_state( | |
ffd2d883 | 512 | struct cgroup *cgrp, int subsys_id) |
ddbcc7e8 | 513 | { |
ffd2d883 | 514 | return cgrp->subsys[subsys_id]; |
ddbcc7e8 PM |
515 | } |
516 | ||
dc61b1d6 PZ |
517 | /* |
518 | * function to get the cgroup_subsys_state which allows for extra | |
519 | * rcu_dereference_check() conditions, such as locks used during the | |
520 | * cgroup_subsys::attach() methods. | |
521 | */ | |
522 | #define task_subsys_state_check(task, subsys_id, __c) \ | |
523 | rcu_dereference_check(task->cgroups->subsys[subsys_id], \ | |
dc61b1d6 PZ |
524 | lockdep_is_held(&task->alloc_lock) || \ |
525 | cgroup_lock_is_held() || (__c)) | |
526 | ||
527 | static inline struct cgroup_subsys_state * | |
528 | task_subsys_state(struct task_struct *task, int subsys_id) | |
ddbcc7e8 | 529 | { |
dc61b1d6 | 530 | return task_subsys_state_check(task, subsys_id, false); |
ddbcc7e8 PM |
531 | } |
532 | ||
533 | static inline struct cgroup* task_cgroup(struct task_struct *task, | |
534 | int subsys_id) | |
535 | { | |
536 | return task_subsys_state(task, subsys_id)->cgroup; | |
537 | } | |
538 | ||
574bd9f7 TH |
539 | /** |
540 | * cgroup_for_each_child - iterate through children of a cgroup | |
541 | * @pos: the cgroup * to use as the loop cursor | |
542 | * @cgroup: cgroup whose children to walk | |
543 | * | |
544 | * Walk @cgroup's children. Must be called under rcu_read_lock(). A child | |
92fb9748 TH |
545 | * cgroup which hasn't finished ->css_online() or already has finished |
546 | * ->css_offline() may show up during traversal and it's each subsystem's | |
574bd9f7 TH |
547 | * responsibility to verify that each @pos is alive. |
548 | * | |
92fb9748 TH |
549 | * If a subsystem synchronizes against the parent in its ->css_online() and |
550 | * before starting iterating, a cgroup which finished ->css_online() is | |
551 | * guaranteed to be visible in the future iterations. | |
574bd9f7 TH |
552 | */ |
553 | #define cgroup_for_each_child(pos, cgroup) \ | |
554 | list_for_each_entry_rcu(pos, &(cgroup)->children, sibling) | |
555 | ||
556 | struct cgroup *cgroup_next_descendant_pre(struct cgroup *pos, | |
557 | struct cgroup *cgroup); | |
558 | ||
559 | /** | |
560 | * cgroup_for_each_descendant_pre - pre-order walk of a cgroup's descendants | |
561 | * @pos: the cgroup * to use as the loop cursor | |
562 | * @cgroup: cgroup whose descendants to walk | |
563 | * | |
564 | * Walk @cgroup's descendants. Must be called under rcu_read_lock(). A | |
92fb9748 TH |
565 | * descendant cgroup which hasn't finished ->css_online() or already has |
566 | * finished ->css_offline() may show up during traversal and it's each | |
574bd9f7 TH |
567 | * subsystem's responsibility to verify that each @pos is alive. |
568 | * | |
92fb9748 TH |
569 | * If a subsystem synchronizes against the parent in its ->css_online() and |
570 | * before starting iterating, and synchronizes against @pos on each | |
571 | * iteration, any descendant cgroup which finished ->css_offline() is | |
574bd9f7 TH |
572 | * guaranteed to be visible in the future iterations. |
573 | * | |
574 | * In other words, the following guarantees that a descendant can't escape | |
575 | * state updates of its ancestors. | |
576 | * | |
92fb9748 | 577 | * my_online(@cgrp) |
574bd9f7 TH |
578 | * { |
579 | * Lock @cgrp->parent and @cgrp; | |
580 | * Inherit state from @cgrp->parent; | |
581 | * Unlock both. | |
582 | * } | |
583 | * | |
584 | * my_update_state(@cgrp) | |
585 | * { | |
586 | * Lock @cgrp; | |
587 | * Update @cgrp's state; | |
588 | * Unlock @cgrp; | |
589 | * | |
590 | * cgroup_for_each_descendant_pre(@pos, @cgrp) { | |
591 | * Lock @pos; | |
592 | * Verify @pos is alive and inherit state from @pos->parent; | |
593 | * Unlock @pos; | |
594 | * } | |
595 | * } | |
596 | * | |
597 | * As long as the inheriting step, including checking the parent state, is | |
598 | * enclosed inside @pos locking, double-locking the parent isn't necessary | |
599 | * while inheriting. The state update to the parent is guaranteed to be | |
600 | * visible by walking order and, as long as inheriting operations to the | |
601 | * same @pos are atomic to each other, multiple updates racing each other | |
602 | * still result in the correct state. It's guaranateed that at least one | |
603 | * inheritance happens for any cgroup after the latest update to its | |
604 | * parent. | |
605 | * | |
606 | * If checking parent's state requires locking the parent, each inheriting | |
607 | * iteration should lock and unlock both @pos->parent and @pos. | |
608 | * | |
609 | * Alternatively, a subsystem may choose to use a single global lock to | |
92fb9748 | 610 | * synchronize ->css_online() and ->css_offline() against tree-walking |
574bd9f7 TH |
611 | * operations. |
612 | */ | |
613 | #define cgroup_for_each_descendant_pre(pos, cgroup) \ | |
614 | for (pos = cgroup_next_descendant_pre(NULL, (cgroup)); (pos); \ | |
615 | pos = cgroup_next_descendant_pre((pos), (cgroup))) | |
616 | ||
617 | struct cgroup *cgroup_next_descendant_post(struct cgroup *pos, | |
618 | struct cgroup *cgroup); | |
619 | ||
620 | /** | |
621 | * cgroup_for_each_descendant_post - post-order walk of a cgroup's descendants | |
622 | * @pos: the cgroup * to use as the loop cursor | |
623 | * @cgroup: cgroup whose descendants to walk | |
624 | * | |
625 | * Similar to cgroup_for_each_descendant_pre() but performs post-order | |
626 | * traversal instead. Note that the walk visibility guarantee described in | |
627 | * pre-order walk doesn't apply the same to post-order walks. | |
628 | */ | |
629 | #define cgroup_for_each_descendant_post(pos, cgroup) \ | |
630 | for (pos = cgroup_next_descendant_post(NULL, (cgroup)); (pos); \ | |
631 | pos = cgroup_next_descendant_post((pos), (cgroup))) | |
632 | ||
817929ec PM |
633 | /* A cgroup_iter should be treated as an opaque object */ |
634 | struct cgroup_iter { | |
635 | struct list_head *cg_link; | |
636 | struct list_head *task; | |
637 | }; | |
638 | ||
d20a390a PM |
639 | /* |
640 | * To iterate across the tasks in a cgroup: | |
817929ec | 641 | * |
b595076a | 642 | * 1) call cgroup_iter_start to initialize an iterator |
817929ec PM |
643 | * |
644 | * 2) call cgroup_iter_next() to retrieve member tasks until it | |
645 | * returns NULL or until you want to end the iteration | |
646 | * | |
647 | * 3) call cgroup_iter_end() to destroy the iterator. | |
31a7df01 | 648 | * |
d20a390a PM |
649 | * Or, call cgroup_scan_tasks() to iterate through every task in a |
650 | * cgroup - cgroup_scan_tasks() holds the css_set_lock when calling | |
651 | * the test_task() callback, but not while calling the process_task() | |
652 | * callback. | |
817929ec | 653 | */ |
ffd2d883 LZ |
654 | void cgroup_iter_start(struct cgroup *cgrp, struct cgroup_iter *it); |
655 | struct task_struct *cgroup_iter_next(struct cgroup *cgrp, | |
817929ec | 656 | struct cgroup_iter *it); |
ffd2d883 | 657 | void cgroup_iter_end(struct cgroup *cgrp, struct cgroup_iter *it); |
31a7df01 | 658 | int cgroup_scan_tasks(struct cgroup_scanner *scan); |
956db3ca | 659 | int cgroup_attach_task(struct cgroup *, struct task_struct *); |
31583bb0 MT |
660 | int cgroup_attach_task_all(struct task_struct *from, struct task_struct *); |
661 | ||
38460b48 KH |
662 | /* |
663 | * CSS ID is ID for cgroup_subsys_state structs under subsys. This only works | |
664 | * if cgroup_subsys.use_id == true. It can be used for looking up and scanning. | |
665 | * CSS ID is assigned at cgroup allocation (create) automatically | |
666 | * and removed when subsys calls free_css_id() function. This is because | |
667 | * the lifetime of cgroup_subsys_state is subsys's matter. | |
668 | * | |
669 | * Looking up and scanning function should be called under rcu_read_lock(). | |
6be96a5c | 670 | * Taking cgroup_mutex is not necessary for following calls. |
38460b48 KH |
671 | * But the css returned by this routine can be "not populated yet" or "being |
672 | * destroyed". The caller should check css and cgroup's status. | |
673 | */ | |
674 | ||
675 | /* | |
676 | * Typically Called at ->destroy(), or somewhere the subsys frees | |
677 | * cgroup_subsys_state. | |
678 | */ | |
679 | void free_css_id(struct cgroup_subsys *ss, struct cgroup_subsys_state *css); | |
680 | ||
681 | /* Find a cgroup_subsys_state which has given ID */ | |
682 | ||
683 | struct cgroup_subsys_state *css_lookup(struct cgroup_subsys *ss, int id); | |
684 | ||
685 | /* | |
686 | * Get a cgroup whose id is greater than or equal to id under tree of root. | |
687 | * Returning a cgroup_subsys_state or NULL. | |
688 | */ | |
689 | struct cgroup_subsys_state *css_get_next(struct cgroup_subsys *ss, int id, | |
690 | struct cgroup_subsys_state *root, int *foundid); | |
691 | ||
692 | /* Returns true if root is ancestor of cg */ | |
693 | bool css_is_ancestor(struct cgroup_subsys_state *cg, | |
0b7f569e | 694 | const struct cgroup_subsys_state *root); |
38460b48 KH |
695 | |
696 | /* Get id and depth of css */ | |
697 | unsigned short css_id(struct cgroup_subsys_state *css); | |
698 | unsigned short css_depth(struct cgroup_subsys_state *css); | |
e5d1367f | 699 | struct cgroup_subsys_state *cgroup_css_from_dir(struct file *f, int id); |
38460b48 | 700 | |
ddbcc7e8 PM |
701 | #else /* !CONFIG_CGROUPS */ |
702 | ||
703 | static inline int cgroup_init_early(void) { return 0; } | |
704 | static inline int cgroup_init(void) { return 0; } | |
b4f48b63 PM |
705 | static inline void cgroup_fork(struct task_struct *p) {} |
706 | static inline void cgroup_fork_callbacks(struct task_struct *p) {} | |
817929ec | 707 | static inline void cgroup_post_fork(struct task_struct *p) {} |
b4f48b63 | 708 | static inline void cgroup_exit(struct task_struct *p, int callbacks) {} |
ddbcc7e8 PM |
709 | |
710 | static inline void cgroup_lock(void) {} | |
711 | static inline void cgroup_unlock(void) {} | |
846c7bb0 BS |
712 | static inline int cgroupstats_build(struct cgroupstats *stats, |
713 | struct dentry *dentry) | |
714 | { | |
715 | return -EINVAL; | |
716 | } | |
ddbcc7e8 | 717 | |
d7926ee3 | 718 | /* No cgroups - nothing to do */ |
31583bb0 MT |
719 | static inline int cgroup_attach_task_all(struct task_struct *from, |
720 | struct task_struct *t) | |
721 | { | |
722 | return 0; | |
723 | } | |
d7926ee3 | 724 | |
ddbcc7e8 PM |
725 | #endif /* !CONFIG_CGROUPS */ |
726 | ||
727 | #endif /* _LINUX_CGROUP_H */ |