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(), |
d0b2fdd2 | 69 | * css_tryget() and css_put(). |
d20a390a | 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 | 145 | /* |
2260e7fc TH |
146 | * Clone the parent's configuration when creating a new child |
147 | * cpuset cgroup. For historical reasons, this option can be | |
148 | * specified at mount time and thus is implemented here. | |
97978e6d | 149 | */ |
2260e7fc | 150 | CGRP_CPUSET_CLONE_CHILDREN, |
3116f0e3 PM |
151 | }; |
152 | ||
ddbcc7e8 PM |
153 | struct cgroup { |
154 | unsigned long flags; /* "unsigned long" so bitops work */ | |
155 | ||
d20a390a PM |
156 | /* |
157 | * count users of this cgroup. >0 means busy, but doesn't | |
158 | * necessarily indicate the number of tasks in the cgroup | |
159 | */ | |
ddbcc7e8 PM |
160 | atomic_t count; |
161 | ||
0a950f65 TH |
162 | int id; /* ida allocated in-hierarchy ID */ |
163 | ||
ddbcc7e8 PM |
164 | /* |
165 | * We link our 'sibling' struct into our parent's 'children'. | |
166 | * Our children link their 'sibling' into our 'children'. | |
167 | */ | |
168 | struct list_head sibling; /* my parent's children */ | |
169 | struct list_head children; /* my children */ | |
05ef1d7c | 170 | struct list_head files; /* my files */ |
ddbcc7e8 | 171 | |
d20a390a | 172 | struct cgroup *parent; /* my parent */ |
febfcef6 | 173 | struct dentry *dentry; /* cgroup fs entry, RCU protected */ |
ddbcc7e8 PM |
174 | |
175 | /* Private pointers for each registered subsystem */ | |
176 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
177 | ||
178 | struct cgroupfs_root *root; | |
179 | struct cgroup *top_cgroup; | |
817929ec PM |
180 | |
181 | /* | |
182 | * List of cg_cgroup_links pointing at css_sets with | |
183 | * tasks in this cgroup. Protected by css_set_lock | |
184 | */ | |
185 | struct list_head css_sets; | |
81a6a5cd | 186 | |
b0ca5a84 | 187 | struct list_head allcg_node; /* cgroupfs_root->allcg_list */ |
8e3f6541 | 188 | struct list_head cft_q_node; /* used during cftype add/rm */ |
b0ca5a84 | 189 | |
81a6a5cd PM |
190 | /* |
191 | * Linked list running through all cgroups that can | |
192 | * potentially be reaped by the release agent. Protected by | |
193 | * release_list_lock | |
194 | */ | |
195 | struct list_head release_list; | |
cc31edce | 196 | |
72a8cb30 BB |
197 | /* |
198 | * list of pidlists, up to two for each namespace (one for procs, one | |
199 | * for tasks); created on demand. | |
200 | */ | |
201 | struct list_head pidlists; | |
202 | struct mutex pidlist_mutex; | |
a47295e6 PM |
203 | |
204 | /* For RCU-protected deletion */ | |
205 | struct rcu_head rcu_head; | |
be445626 | 206 | struct work_struct free_work; |
0dea1168 | 207 | |
25985edc | 208 | /* List of events which userspace want to receive */ |
0dea1168 KS |
209 | struct list_head event_list; |
210 | spinlock_t event_list_lock; | |
03b1cde6 AR |
211 | |
212 | /* directory xattrs */ | |
213 | struct simple_xattrs xattrs; | |
817929ec PM |
214 | }; |
215 | ||
d20a390a PM |
216 | /* |
217 | * A css_set is a structure holding pointers to a set of | |
817929ec PM |
218 | * cgroup_subsys_state objects. This saves space in the task struct |
219 | * object and speeds up fork()/exit(), since a single inc/dec and a | |
d20a390a PM |
220 | * list_add()/del() can bump the reference count on the entire cgroup |
221 | * set for a task. | |
817929ec PM |
222 | */ |
223 | ||
224 | struct css_set { | |
225 | ||
226 | /* Reference count */ | |
146aa1bd | 227 | atomic_t refcount; |
817929ec | 228 | |
472b1053 LZ |
229 | /* |
230 | * List running through all cgroup groups in the same hash | |
231 | * slot. Protected by css_set_lock | |
232 | */ | |
233 | struct hlist_node hlist; | |
234 | ||
817929ec PM |
235 | /* |
236 | * List running through all tasks using this cgroup | |
237 | * group. Protected by css_set_lock | |
238 | */ | |
239 | struct list_head tasks; | |
240 | ||
241 | /* | |
242 | * List of cg_cgroup_link objects on link chains from | |
243 | * cgroups referenced from this css_set. Protected by | |
244 | * css_set_lock | |
245 | */ | |
246 | struct list_head cg_links; | |
247 | ||
248 | /* | |
249 | * Set of subsystem states, one for each subsystem. This array | |
250 | * is immutable after creation apart from the init_css_set | |
cf5d5941 BB |
251 | * during subsystem registration (at boot time) and modular subsystem |
252 | * loading/unloading. | |
817929ec PM |
253 | */ |
254 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
c378369d BB |
255 | |
256 | /* For RCU-protected deletion */ | |
257 | struct rcu_head rcu_head; | |
ddbcc7e8 PM |
258 | }; |
259 | ||
91796569 PM |
260 | /* |
261 | * cgroup_map_cb is an abstract callback API for reporting map-valued | |
262 | * control files | |
263 | */ | |
264 | ||
265 | struct cgroup_map_cb { | |
266 | int (*fill)(struct cgroup_map_cb *cb, const char *key, u64 value); | |
267 | void *state; | |
268 | }; | |
269 | ||
d20a390a PM |
270 | /* |
271 | * struct cftype: handler definitions for cgroup control files | |
ddbcc7e8 PM |
272 | * |
273 | * When reading/writing to a file: | |
a043e3b2 | 274 | * - the cgroup to use is file->f_dentry->d_parent->d_fsdata |
ddbcc7e8 PM |
275 | * - the 'cftype' of the file is file->f_dentry->d_fsdata |
276 | */ | |
277 | ||
8e3f6541 TH |
278 | /* cftype->flags */ |
279 | #define CFTYPE_ONLY_ON_ROOT (1U << 0) /* only create on root cg */ | |
d0b2fdd2 | 280 | #define CFTYPE_NOT_ON_ROOT (1U << 1) /* don't create on root cg */ |
8e3f6541 TH |
281 | |
282 | #define MAX_CFTYPE_NAME 64 | |
283 | ||
ddbcc7e8 | 284 | struct cftype { |
d20a390a PM |
285 | /* |
286 | * By convention, the name should begin with the name of the | |
8e3f6541 TH |
287 | * subsystem, followed by a period. Zero length string indicates |
288 | * end of cftype array. | |
d20a390a | 289 | */ |
ddbcc7e8 PM |
290 | char name[MAX_CFTYPE_NAME]; |
291 | int private; | |
099fca32 LZ |
292 | /* |
293 | * If not 0, file mode is set to this value, otherwise it will | |
294 | * be figured out automatically | |
295 | */ | |
a5e7ed32 | 296 | umode_t mode; |
db3b1497 PM |
297 | |
298 | /* | |
299 | * If non-zero, defines the maximum length of string that can | |
300 | * be passed to write_string; defaults to 64 | |
301 | */ | |
302 | size_t max_write_len; | |
303 | ||
8e3f6541 TH |
304 | /* CFTYPE_* flags */ |
305 | unsigned int flags; | |
306 | ||
03b1cde6 AR |
307 | /* file xattrs */ |
308 | struct simple_xattrs xattrs; | |
309 | ||
ce16b49d PM |
310 | int (*open)(struct inode *inode, struct file *file); |
311 | ssize_t (*read)(struct cgroup *cgrp, struct cftype *cft, | |
312 | struct file *file, | |
313 | char __user *buf, size_t nbytes, loff_t *ppos); | |
ddbcc7e8 | 314 | /* |
f4c753b7 | 315 | * read_u64() is a shortcut for the common case of returning a |
ddbcc7e8 PM |
316 | * single integer. Use it in place of read() |
317 | */ | |
ce16b49d | 318 | u64 (*read_u64)(struct cgroup *cgrp, struct cftype *cft); |
e73d2c61 PM |
319 | /* |
320 | * read_s64() is a signed version of read_u64() | |
321 | */ | |
ce16b49d | 322 | s64 (*read_s64)(struct cgroup *cgrp, struct cftype *cft); |
91796569 PM |
323 | /* |
324 | * read_map() is used for defining a map of key/value | |
325 | * pairs. It should call cb->fill(cb, key, value) for each | |
326 | * entry. The key/value pairs (and their ordering) should not | |
327 | * change between reboots. | |
328 | */ | |
ce16b49d PM |
329 | int (*read_map)(struct cgroup *cont, struct cftype *cft, |
330 | struct cgroup_map_cb *cb); | |
29486df3 SH |
331 | /* |
332 | * read_seq_string() is used for outputting a simple sequence | |
333 | * using seqfile. | |
334 | */ | |
ce16b49d PM |
335 | int (*read_seq_string)(struct cgroup *cont, struct cftype *cft, |
336 | struct seq_file *m); | |
91796569 | 337 | |
ce16b49d PM |
338 | ssize_t (*write)(struct cgroup *cgrp, struct cftype *cft, |
339 | struct file *file, | |
340 | const char __user *buf, size_t nbytes, loff_t *ppos); | |
355e0c48 PM |
341 | |
342 | /* | |
f4c753b7 | 343 | * write_u64() is a shortcut for the common case of accepting |
355e0c48 PM |
344 | * a single integer (as parsed by simple_strtoull) from |
345 | * userspace. Use in place of write(); return 0 or error. | |
346 | */ | |
ce16b49d | 347 | int (*write_u64)(struct cgroup *cgrp, struct cftype *cft, u64 val); |
e73d2c61 PM |
348 | /* |
349 | * write_s64() is a signed version of write_u64() | |
350 | */ | |
ce16b49d | 351 | int (*write_s64)(struct cgroup *cgrp, struct cftype *cft, s64 val); |
355e0c48 | 352 | |
db3b1497 PM |
353 | /* |
354 | * write_string() is passed a nul-terminated kernelspace | |
355 | * buffer of maximum length determined by max_write_len. | |
356 | * Returns 0 or -ve error code. | |
357 | */ | |
358 | int (*write_string)(struct cgroup *cgrp, struct cftype *cft, | |
359 | const char *buffer); | |
d447ea2f PE |
360 | /* |
361 | * trigger() callback can be used to get some kick from the | |
362 | * userspace, when the actual string written is not important | |
363 | * at all. The private field can be used to determine the | |
364 | * kick type for multiplexing. | |
365 | */ | |
366 | int (*trigger)(struct cgroup *cgrp, unsigned int event); | |
367 | ||
ce16b49d | 368 | int (*release)(struct inode *inode, struct file *file); |
0dea1168 KS |
369 | |
370 | /* | |
371 | * register_event() callback will be used to add new userspace | |
372 | * waiter for changes related to the cftype. Implement it if | |
373 | * you want to provide this functionality. Use eventfd_signal() | |
374 | * on eventfd to send notification to userspace. | |
375 | */ | |
376 | int (*register_event)(struct cgroup *cgrp, struct cftype *cft, | |
377 | struct eventfd_ctx *eventfd, const char *args); | |
378 | /* | |
379 | * unregister_event() callback will be called when userspace | |
380 | * closes the eventfd or on cgroup removing. | |
381 | * This callback must be implemented, if you want provide | |
382 | * notification functionality. | |
0dea1168 | 383 | */ |
907860ed | 384 | void (*unregister_event)(struct cgroup *cgrp, struct cftype *cft, |
0dea1168 | 385 | struct eventfd_ctx *eventfd); |
ddbcc7e8 PM |
386 | }; |
387 | ||
8e3f6541 TH |
388 | /* |
389 | * cftype_sets describe cftypes belonging to a subsystem and are chained at | |
390 | * cgroup_subsys->cftsets. Each cftset points to an array of cftypes | |
391 | * terminated by zero length name. | |
392 | */ | |
393 | struct cftype_set { | |
394 | struct list_head node; /* chained at subsys->cftsets */ | |
03b1cde6 | 395 | struct cftype *cfts; |
8e3f6541 TH |
396 | }; |
397 | ||
31a7df01 CW |
398 | struct cgroup_scanner { |
399 | struct cgroup *cg; | |
400 | int (*test_task)(struct task_struct *p, struct cgroup_scanner *scan); | |
401 | void (*process_task)(struct task_struct *p, | |
402 | struct cgroup_scanner *scan); | |
403 | struct ptr_heap *heap; | |
bd1a8ab7 | 404 | void *data; |
31a7df01 CW |
405 | }; |
406 | ||
03b1cde6 AR |
407 | int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); |
408 | int cgroup_rm_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); | |
8e3f6541 | 409 | |
ffd2d883 | 410 | int cgroup_is_removed(const struct cgroup *cgrp); |
ddbcc7e8 | 411 | |
ffd2d883 | 412 | int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen); |
ddbcc7e8 | 413 | |
ffd2d883 | 414 | int cgroup_task_count(const struct cgroup *cgrp); |
bbcb81d0 | 415 | |
313e924c GN |
416 | /* Return true if cgrp is a descendant of the task's cgroup */ |
417 | int cgroup_is_descendant(const struct cgroup *cgrp, struct task_struct *task); | |
ddbcc7e8 | 418 | |
2f7ee569 TH |
419 | /* |
420 | * Control Group taskset, used to pass around set of tasks to cgroup_subsys | |
421 | * methods. | |
422 | */ | |
423 | struct cgroup_taskset; | |
424 | struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset); | |
425 | struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset); | |
426 | struct cgroup *cgroup_taskset_cur_cgroup(struct cgroup_taskset *tset); | |
427 | int cgroup_taskset_size(struct cgroup_taskset *tset); | |
428 | ||
429 | /** | |
430 | * cgroup_taskset_for_each - iterate cgroup_taskset | |
431 | * @task: the loop cursor | |
432 | * @skip_cgrp: skip if task's cgroup matches this, %NULL to iterate through all | |
433 | * @tset: taskset to iterate | |
434 | */ | |
435 | #define cgroup_taskset_for_each(task, skip_cgrp, tset) \ | |
436 | for ((task) = cgroup_taskset_first((tset)); (task); \ | |
437 | (task) = cgroup_taskset_next((tset))) \ | |
438 | if (!(skip_cgrp) || \ | |
439 | cgroup_taskset_cur_cgroup((tset)) != (skip_cgrp)) | |
440 | ||
21acb9ca TLSC |
441 | /* |
442 | * Control Group subsystem type. | |
443 | * See Documentation/cgroups/cgroups.txt for details | |
444 | */ | |
ddbcc7e8 PM |
445 | |
446 | struct cgroup_subsys { | |
92fb9748 TH |
447 | struct cgroup_subsys_state *(*css_alloc)(struct cgroup *cgrp); |
448 | int (*css_online)(struct cgroup *cgrp); | |
449 | void (*css_offline)(struct cgroup *cgrp); | |
450 | void (*css_free)(struct cgroup *cgrp); | |
451 | ||
761b3ef5 LZ |
452 | int (*can_attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); |
453 | void (*cancel_attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); | |
454 | void (*attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); | |
455 | void (*fork)(struct task_struct *task); | |
456 | void (*exit)(struct cgroup *cgrp, struct cgroup *old_cgrp, | |
457 | struct task_struct *task); | |
761b3ef5 | 458 | void (*bind)(struct cgroup *root); |
e5991371 | 459 | |
ddbcc7e8 PM |
460 | int subsys_id; |
461 | int active; | |
8bab8dde | 462 | int disabled; |
ddbcc7e8 | 463 | int early_init; |
38460b48 KH |
464 | /* |
465 | * True if this subsys uses ID. ID is not available before cgroup_init() | |
466 | * (not available in early_init time.) | |
467 | */ | |
468 | bool use_id; | |
48ddbe19 | 469 | |
8c7f6edb TH |
470 | /* |
471 | * If %false, this subsystem is properly hierarchical - | |
472 | * configuration, resource accounting and restriction on a parent | |
473 | * cgroup cover those of its children. If %true, hierarchy support | |
474 | * is broken in some ways - some subsystems ignore hierarchy | |
475 | * completely while others are only implemented half-way. | |
476 | * | |
477 | * It's now disallowed to create nested cgroups if the subsystem is | |
478 | * broken and cgroup core will emit a warning message on such | |
479 | * cases. Eventually, all subsystems will be made properly | |
480 | * hierarchical and this will go away. | |
481 | */ | |
482 | bool broken_hierarchy; | |
483 | bool warned_broken_hierarchy; | |
484 | ||
ddbcc7e8 PM |
485 | #define MAX_CGROUP_TYPE_NAMELEN 32 |
486 | const char *name; | |
487 | ||
999cd8a4 PM |
488 | /* |
489 | * Link to parent, and list entry in parent's children. | |
6be96a5c | 490 | * Protected by cgroup_lock() |
999cd8a4 PM |
491 | */ |
492 | struct cgroupfs_root *root; | |
ddbcc7e8 | 493 | struct list_head sibling; |
38460b48 KH |
494 | /* used when use_id == true */ |
495 | struct idr idr; | |
42aee6c4 | 496 | spinlock_t id_lock; |
e6a1105b | 497 | |
8e3f6541 TH |
498 | /* list of cftype_sets */ |
499 | struct list_head cftsets; | |
500 | ||
501 | /* base cftypes, automatically [de]registered with subsys itself */ | |
502 | struct cftype *base_cftypes; | |
503 | struct cftype_set base_cftset; | |
504 | ||
e6a1105b BB |
505 | /* should be defined only by modular subsystems */ |
506 | struct module *module; | |
ddbcc7e8 PM |
507 | }; |
508 | ||
509 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _subsys; | |
5fc0b025 | 510 | #define IS_SUBSYS_ENABLED(option) IS_BUILTIN(option) |
ddbcc7e8 | 511 | #include <linux/cgroup_subsys.h> |
5fc0b025 | 512 | #undef IS_SUBSYS_ENABLED |
ddbcc7e8 PM |
513 | #undef SUBSYS |
514 | ||
515 | static inline struct cgroup_subsys_state *cgroup_subsys_state( | |
ffd2d883 | 516 | struct cgroup *cgrp, int subsys_id) |
ddbcc7e8 | 517 | { |
ffd2d883 | 518 | return cgrp->subsys[subsys_id]; |
ddbcc7e8 PM |
519 | } |
520 | ||
dc61b1d6 PZ |
521 | /* |
522 | * function to get the cgroup_subsys_state which allows for extra | |
523 | * rcu_dereference_check() conditions, such as locks used during the | |
524 | * cgroup_subsys::attach() methods. | |
525 | */ | |
526 | #define task_subsys_state_check(task, subsys_id, __c) \ | |
527 | rcu_dereference_check(task->cgroups->subsys[subsys_id], \ | |
dc61b1d6 PZ |
528 | lockdep_is_held(&task->alloc_lock) || \ |
529 | cgroup_lock_is_held() || (__c)) | |
530 | ||
531 | static inline struct cgroup_subsys_state * | |
532 | task_subsys_state(struct task_struct *task, int subsys_id) | |
ddbcc7e8 | 533 | { |
dc61b1d6 | 534 | return task_subsys_state_check(task, subsys_id, false); |
ddbcc7e8 PM |
535 | } |
536 | ||
537 | static inline struct cgroup* task_cgroup(struct task_struct *task, | |
538 | int subsys_id) | |
539 | { | |
540 | return task_subsys_state(task, subsys_id)->cgroup; | |
541 | } | |
542 | ||
574bd9f7 TH |
543 | /** |
544 | * cgroup_for_each_child - iterate through children of a cgroup | |
545 | * @pos: the cgroup * to use as the loop cursor | |
546 | * @cgroup: cgroup whose children to walk | |
547 | * | |
548 | * Walk @cgroup's children. Must be called under rcu_read_lock(). A child | |
92fb9748 TH |
549 | * cgroup which hasn't finished ->css_online() or already has finished |
550 | * ->css_offline() may show up during traversal and it's each subsystem's | |
574bd9f7 TH |
551 | * responsibility to verify that each @pos is alive. |
552 | * | |
92fb9748 TH |
553 | * If a subsystem synchronizes against the parent in its ->css_online() and |
554 | * before starting iterating, a cgroup which finished ->css_online() is | |
555 | * guaranteed to be visible in the future iterations. | |
574bd9f7 TH |
556 | */ |
557 | #define cgroup_for_each_child(pos, cgroup) \ | |
558 | list_for_each_entry_rcu(pos, &(cgroup)->children, sibling) | |
559 | ||
560 | struct cgroup *cgroup_next_descendant_pre(struct cgroup *pos, | |
561 | struct cgroup *cgroup); | |
12a9d2fe | 562 | struct cgroup *cgroup_rightmost_descendant(struct cgroup *pos); |
574bd9f7 TH |
563 | |
564 | /** | |
565 | * cgroup_for_each_descendant_pre - pre-order walk of a cgroup's descendants | |
566 | * @pos: the cgroup * to use as the loop cursor | |
567 | * @cgroup: cgroup whose descendants to walk | |
568 | * | |
569 | * Walk @cgroup's descendants. Must be called under rcu_read_lock(). A | |
92fb9748 TH |
570 | * descendant cgroup which hasn't finished ->css_online() or already has |
571 | * finished ->css_offline() may show up during traversal and it's each | |
574bd9f7 TH |
572 | * subsystem's responsibility to verify that each @pos is alive. |
573 | * | |
92fb9748 TH |
574 | * If a subsystem synchronizes against the parent in its ->css_online() and |
575 | * before starting iterating, and synchronizes against @pos on each | |
576 | * iteration, any descendant cgroup which finished ->css_offline() is | |
574bd9f7 TH |
577 | * guaranteed to be visible in the future iterations. |
578 | * | |
579 | * In other words, the following guarantees that a descendant can't escape | |
580 | * state updates of its ancestors. | |
581 | * | |
92fb9748 | 582 | * my_online(@cgrp) |
574bd9f7 TH |
583 | * { |
584 | * Lock @cgrp->parent and @cgrp; | |
585 | * Inherit state from @cgrp->parent; | |
586 | * Unlock both. | |
587 | * } | |
588 | * | |
589 | * my_update_state(@cgrp) | |
590 | * { | |
591 | * Lock @cgrp; | |
592 | * Update @cgrp's state; | |
593 | * Unlock @cgrp; | |
594 | * | |
595 | * cgroup_for_each_descendant_pre(@pos, @cgrp) { | |
596 | * Lock @pos; | |
597 | * Verify @pos is alive and inherit state from @pos->parent; | |
598 | * Unlock @pos; | |
599 | * } | |
600 | * } | |
601 | * | |
602 | * As long as the inheriting step, including checking the parent state, is | |
603 | * enclosed inside @pos locking, double-locking the parent isn't necessary | |
604 | * while inheriting. The state update to the parent is guaranteed to be | |
605 | * visible by walking order and, as long as inheriting operations to the | |
606 | * same @pos are atomic to each other, multiple updates racing each other | |
607 | * still result in the correct state. It's guaranateed that at least one | |
608 | * inheritance happens for any cgroup after the latest update to its | |
609 | * parent. | |
610 | * | |
611 | * If checking parent's state requires locking the parent, each inheriting | |
612 | * iteration should lock and unlock both @pos->parent and @pos. | |
613 | * | |
614 | * Alternatively, a subsystem may choose to use a single global lock to | |
92fb9748 | 615 | * synchronize ->css_online() and ->css_offline() against tree-walking |
574bd9f7 TH |
616 | * operations. |
617 | */ | |
618 | #define cgroup_for_each_descendant_pre(pos, cgroup) \ | |
619 | for (pos = cgroup_next_descendant_pre(NULL, (cgroup)); (pos); \ | |
620 | pos = cgroup_next_descendant_pre((pos), (cgroup))) | |
621 | ||
622 | struct cgroup *cgroup_next_descendant_post(struct cgroup *pos, | |
623 | struct cgroup *cgroup); | |
624 | ||
625 | /** | |
626 | * cgroup_for_each_descendant_post - post-order walk of a cgroup's descendants | |
627 | * @pos: the cgroup * to use as the loop cursor | |
628 | * @cgroup: cgroup whose descendants to walk | |
629 | * | |
630 | * Similar to cgroup_for_each_descendant_pre() but performs post-order | |
631 | * traversal instead. Note that the walk visibility guarantee described in | |
632 | * pre-order walk doesn't apply the same to post-order walks. | |
633 | */ | |
634 | #define cgroup_for_each_descendant_post(pos, cgroup) \ | |
635 | for (pos = cgroup_next_descendant_post(NULL, (cgroup)); (pos); \ | |
636 | pos = cgroup_next_descendant_post((pos), (cgroup))) | |
637 | ||
817929ec PM |
638 | /* A cgroup_iter should be treated as an opaque object */ |
639 | struct cgroup_iter { | |
640 | struct list_head *cg_link; | |
641 | struct list_head *task; | |
642 | }; | |
643 | ||
d20a390a PM |
644 | /* |
645 | * To iterate across the tasks in a cgroup: | |
817929ec | 646 | * |
b595076a | 647 | * 1) call cgroup_iter_start to initialize an iterator |
817929ec PM |
648 | * |
649 | * 2) call cgroup_iter_next() to retrieve member tasks until it | |
650 | * returns NULL or until you want to end the iteration | |
651 | * | |
652 | * 3) call cgroup_iter_end() to destroy the iterator. | |
31a7df01 | 653 | * |
d20a390a PM |
654 | * Or, call cgroup_scan_tasks() to iterate through every task in a |
655 | * cgroup - cgroup_scan_tasks() holds the css_set_lock when calling | |
656 | * the test_task() callback, but not while calling the process_task() | |
657 | * callback. | |
817929ec | 658 | */ |
ffd2d883 LZ |
659 | void cgroup_iter_start(struct cgroup *cgrp, struct cgroup_iter *it); |
660 | struct task_struct *cgroup_iter_next(struct cgroup *cgrp, | |
817929ec | 661 | struct cgroup_iter *it); |
ffd2d883 | 662 | void cgroup_iter_end(struct cgroup *cgrp, struct cgroup_iter *it); |
31a7df01 | 663 | int cgroup_scan_tasks(struct cgroup_scanner *scan); |
956db3ca | 664 | int cgroup_attach_task(struct cgroup *, struct task_struct *); |
31583bb0 MT |
665 | int cgroup_attach_task_all(struct task_struct *from, struct task_struct *); |
666 | ||
38460b48 KH |
667 | /* |
668 | * CSS ID is ID for cgroup_subsys_state structs under subsys. This only works | |
669 | * if cgroup_subsys.use_id == true. It can be used for looking up and scanning. | |
670 | * CSS ID is assigned at cgroup allocation (create) automatically | |
671 | * and removed when subsys calls free_css_id() function. This is because | |
672 | * the lifetime of cgroup_subsys_state is subsys's matter. | |
673 | * | |
674 | * Looking up and scanning function should be called under rcu_read_lock(). | |
6be96a5c | 675 | * Taking cgroup_mutex is not necessary for following calls. |
38460b48 KH |
676 | * But the css returned by this routine can be "not populated yet" or "being |
677 | * destroyed". The caller should check css and cgroup's status. | |
678 | */ | |
679 | ||
680 | /* | |
681 | * Typically Called at ->destroy(), or somewhere the subsys frees | |
682 | * cgroup_subsys_state. | |
683 | */ | |
684 | void free_css_id(struct cgroup_subsys *ss, struct cgroup_subsys_state *css); | |
685 | ||
686 | /* Find a cgroup_subsys_state which has given ID */ | |
687 | ||
688 | struct cgroup_subsys_state *css_lookup(struct cgroup_subsys *ss, int id); | |
689 | ||
690 | /* | |
691 | * Get a cgroup whose id is greater than or equal to id under tree of root. | |
692 | * Returning a cgroup_subsys_state or NULL. | |
693 | */ | |
694 | struct cgroup_subsys_state *css_get_next(struct cgroup_subsys *ss, int id, | |
695 | struct cgroup_subsys_state *root, int *foundid); | |
696 | ||
697 | /* Returns true if root is ancestor of cg */ | |
698 | bool css_is_ancestor(struct cgroup_subsys_state *cg, | |
0b7f569e | 699 | const struct cgroup_subsys_state *root); |
38460b48 KH |
700 | |
701 | /* Get id and depth of css */ | |
702 | unsigned short css_id(struct cgroup_subsys_state *css); | |
703 | unsigned short css_depth(struct cgroup_subsys_state *css); | |
e5d1367f | 704 | struct cgroup_subsys_state *cgroup_css_from_dir(struct file *f, int id); |
38460b48 | 705 | |
ddbcc7e8 PM |
706 | #else /* !CONFIG_CGROUPS */ |
707 | ||
708 | static inline int cgroup_init_early(void) { return 0; } | |
709 | static inline int cgroup_init(void) { return 0; } | |
b4f48b63 | 710 | static inline void cgroup_fork(struct task_struct *p) {} |
817929ec | 711 | static inline void cgroup_post_fork(struct task_struct *p) {} |
b4f48b63 | 712 | static inline void cgroup_exit(struct task_struct *p, int callbacks) {} |
ddbcc7e8 PM |
713 | |
714 | static inline void cgroup_lock(void) {} | |
715 | static inline void cgroup_unlock(void) {} | |
846c7bb0 BS |
716 | static inline int cgroupstats_build(struct cgroupstats *stats, |
717 | struct dentry *dentry) | |
718 | { | |
719 | return -EINVAL; | |
720 | } | |
ddbcc7e8 | 721 | |
d7926ee3 | 722 | /* No cgroups - nothing to do */ |
31583bb0 MT |
723 | static inline int cgroup_attach_task_all(struct task_struct *from, |
724 | struct task_struct *t) | |
725 | { | |
726 | return 0; | |
727 | } | |
d7926ee3 | 728 | |
ddbcc7e8 PM |
729 | #endif /* !CONFIG_CGROUPS */ |
730 | ||
731 | #endif /* _LINUX_CGROUP_H */ |