1 | /* SPDX-License-Identifier: GPL-2.0 */ |
2 | #ifndef _LINUX_CGROUP_H |
3 | #define _LINUX_CGROUP_H |
4 | /* |
5 | * cgroup interface |
6 | * |
7 | * Copyright (C) 2003 BULL SA |
8 | * Copyright (C) 2004-2006 Silicon Graphics, Inc. |
9 | * |
10 | */ |
11 | |
12 | #include <linux/sched.h> |
13 | #include <linux/cpumask.h> |
14 | #include <linux/nodemask.h> |
15 | #include <linux/rculist.h> |
16 | #include <linux/cgroupstats.h> |
17 | #include <linux/fs.h> |
18 | #include <linux/seq_file.h> |
19 | #include <linux/kernfs.h> |
20 | #include <linux/jump_label.h> |
21 | #include <linux/types.h> |
22 | #include <linux/ns_common.h> |
23 | #include <linux/nsproxy.h> |
24 | #include <linux/user_namespace.h> |
25 | #include <linux/refcount.h> |
26 | #include <linux/kernel_stat.h> |
27 | |
28 | #include <linux/cgroup-defs.h> |
29 | |
30 | struct kernel_clone_args; |
31 | |
32 | #ifdef CONFIG_CGROUPS |
33 | |
34 | /* |
35 | * All weight knobs on the default hierarchy should use the following min, |
36 | * default and max values. The default value is the logarithmic center of |
37 | * MIN and MAX and allows 100x to be expressed in both directions. |
38 | */ |
39 | #define CGROUP_WEIGHT_MIN 1 |
40 | #define CGROUP_WEIGHT_DFL 100 |
41 | #define CGROUP_WEIGHT_MAX 10000 |
42 | |
43 | /* walk only threadgroup leaders */ |
44 | #define CSS_TASK_ITER_PROCS (1U << 0) |
45 | /* walk all threaded css_sets in the domain */ |
46 | #define CSS_TASK_ITER_THREADED (1U << 1) |
47 | |
48 | /* internal flags */ |
49 | #define CSS_TASK_ITER_SKIPPED (1U << 16) |
50 | |
51 | /* a css_task_iter should be treated as an opaque object */ |
52 | struct css_task_iter { |
53 | struct cgroup_subsys *ss; |
54 | unsigned int flags; |
55 | |
56 | struct list_head *cset_pos; |
57 | struct list_head *cset_head; |
58 | |
59 | struct list_head *tcset_pos; |
60 | struct list_head *tcset_head; |
61 | |
62 | struct list_head *task_pos; |
63 | |
64 | struct list_head *cur_tasks_head; |
65 | struct css_set *cur_cset; |
66 | struct css_set *cur_dcset; |
67 | struct task_struct *cur_task; |
68 | struct list_head iters_node; /* css_set->task_iters */ |
69 | }; |
70 | |
71 | extern struct cgroup_root cgrp_dfl_root; |
72 | extern struct css_set init_css_set; |
73 | |
74 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys; |
75 | #include <linux/cgroup_subsys.h> |
76 | #undef SUBSYS |
77 | |
78 | #define SUBSYS(_x) \ |
79 | extern struct static_key_true _x ## _cgrp_subsys_enabled_key; \ |
80 | extern struct static_key_true _x ## _cgrp_subsys_on_dfl_key; |
81 | #include <linux/cgroup_subsys.h> |
82 | #undef SUBSYS |
83 | |
84 | /** |
85 | * cgroup_subsys_enabled - fast test on whether a subsys is enabled |
86 | * @ss: subsystem in question |
87 | */ |
88 | #define cgroup_subsys_enabled(ss) \ |
89 | static_branch_likely(&ss ## _enabled_key) |
90 | |
91 | /** |
92 | * cgroup_subsys_on_dfl - fast test on whether a subsys is on default hierarchy |
93 | * @ss: subsystem in question |
94 | */ |
95 | #define cgroup_subsys_on_dfl(ss) \ |
96 | static_branch_likely(&ss ## _on_dfl_key) |
97 | |
98 | bool css_has_online_children(struct cgroup_subsys_state *css); |
99 | struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss); |
100 | struct cgroup_subsys_state *cgroup_e_css(struct cgroup *cgroup, |
101 | struct cgroup_subsys *ss); |
102 | struct cgroup_subsys_state *cgroup_get_e_css(struct cgroup *cgroup, |
103 | struct cgroup_subsys *ss); |
104 | struct cgroup_subsys_state *css_tryget_online_from_dir(struct dentry *dentry, |
105 | struct cgroup_subsys *ss); |
106 | |
107 | struct cgroup *cgroup_get_from_path(const char *path); |
108 | struct cgroup *cgroup_get_from_fd(int fd); |
109 | |
110 | int cgroup_attach_task_all(struct task_struct *from, struct task_struct *); |
111 | int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from); |
112 | |
113 | int cgroup_add_dfl_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); |
114 | int cgroup_add_legacy_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); |
115 | int cgroup_rm_cftypes(struct cftype *cfts); |
116 | void cgroup_file_notify(struct cgroup_file *cfile); |
117 | |
118 | int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen); |
119 | int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry); |
120 | int proc_cgroup_show(struct seq_file *m, struct pid_namespace *ns, |
121 | struct pid *pid, struct task_struct *tsk); |
122 | |
123 | void cgroup_fork(struct task_struct *p); |
124 | extern int cgroup_can_fork(struct task_struct *p, |
125 | struct kernel_clone_args *kargs); |
126 | extern void cgroup_cancel_fork(struct task_struct *p, |
127 | struct kernel_clone_args *kargs); |
128 | extern void cgroup_post_fork(struct task_struct *p, |
129 | struct kernel_clone_args *kargs); |
130 | void cgroup_exit(struct task_struct *p); |
131 | void cgroup_release(struct task_struct *p); |
132 | void cgroup_free(struct task_struct *p); |
133 | |
134 | int cgroup_init_early(void); |
135 | int cgroup_init(void); |
136 | |
137 | int cgroup_parse_float(const char *input, unsigned dec_shift, s64 *v); |
138 | |
139 | /* |
140 | * Iteration helpers and macros. |
141 | */ |
142 | |
143 | struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos, |
144 | struct cgroup_subsys_state *parent); |
145 | struct cgroup_subsys_state *css_next_descendant_pre(struct cgroup_subsys_state *pos, |
146 | struct cgroup_subsys_state *css); |
147 | struct cgroup_subsys_state *css_rightmost_descendant(struct cgroup_subsys_state *pos); |
148 | struct cgroup_subsys_state *css_next_descendant_post(struct cgroup_subsys_state *pos, |
149 | struct cgroup_subsys_state *css); |
150 | |
151 | struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset, |
152 | struct cgroup_subsys_state **dst_cssp); |
153 | struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset, |
154 | struct cgroup_subsys_state **dst_cssp); |
155 | |
156 | void css_task_iter_start(struct cgroup_subsys_state *css, unsigned int flags, |
157 | struct css_task_iter *it); |
158 | struct task_struct *css_task_iter_next(struct css_task_iter *it); |
159 | void css_task_iter_end(struct css_task_iter *it); |
160 | |
161 | /** |
162 | * css_for_each_child - iterate through children of a css |
163 | * @pos: the css * to use as the loop cursor |
164 | * @parent: css whose children to walk |
165 | * |
166 | * Walk @parent's children. Must be called under rcu_read_lock(). |
167 | * |
168 | * If a subsystem synchronizes ->css_online() and the start of iteration, a |
169 | * css which finished ->css_online() is guaranteed to be visible in the |
170 | * future iterations and will stay visible until the last reference is put. |
171 | * A css which hasn't finished ->css_online() or already finished |
172 | * ->css_offline() may show up during traversal. It's each subsystem's |
173 | * responsibility to synchronize against on/offlining. |
174 | * |
175 | * It is allowed to temporarily drop RCU read lock during iteration. The |
176 | * caller is responsible for ensuring that @pos remains accessible until |
177 | * the start of the next iteration by, for example, bumping the css refcnt. |
178 | */ |
179 | #define css_for_each_child(pos, parent) \ |
180 | for ((pos) = css_next_child(NULL, (parent)); (pos); \ |
181 | (pos) = css_next_child((pos), (parent))) |
182 | |
183 | /** |
184 | * css_for_each_descendant_pre - pre-order walk of a css's descendants |
185 | * @pos: the css * to use as the loop cursor |
186 | * @root: css whose descendants to walk |
187 | * |
188 | * Walk @root's descendants. @root is included in the iteration and the |
189 | * first node to be visited. Must be called under rcu_read_lock(). |
190 | * |
191 | * If a subsystem synchronizes ->css_online() and the start of iteration, a |
192 | * css which finished ->css_online() is guaranteed to be visible in the |
193 | * future iterations and will stay visible until the last reference is put. |
194 | * A css which hasn't finished ->css_online() or already finished |
195 | * ->css_offline() may show up during traversal. It's each subsystem's |
196 | * responsibility to synchronize against on/offlining. |
197 | * |
198 | * For example, the following guarantees that a descendant can't escape |
199 | * state updates of its ancestors. |
200 | * |
201 | * my_online(@css) |
202 | * { |
203 | * Lock @css's parent and @css; |
204 | * Inherit state from the parent; |
205 | * Unlock both. |
206 | * } |
207 | * |
208 | * my_update_state(@css) |
209 | * { |
210 | * css_for_each_descendant_pre(@pos, @css) { |
211 | * Lock @pos; |
212 | * if (@pos == @css) |
213 | * Update @css's state; |
214 | * else |
215 | * Verify @pos is alive and inherit state from its parent; |
216 | * Unlock @pos; |
217 | * } |
218 | * } |
219 | * |
220 | * As long as the inheriting step, including checking the parent state, is |
221 | * enclosed inside @pos locking, double-locking the parent isn't necessary |
222 | * while inheriting. The state update to the parent is guaranteed to be |
223 | * visible by walking order and, as long as inheriting operations to the |
224 | * same @pos are atomic to each other, multiple updates racing each other |
225 | * still result in the correct state. It's guaranateed that at least one |
226 | * inheritance happens for any css after the latest update to its parent. |
227 | * |
228 | * If checking parent's state requires locking the parent, each inheriting |
229 | * iteration should lock and unlock both @pos->parent and @pos. |
230 | * |
231 | * Alternatively, a subsystem may choose to use a single global lock to |
232 | * synchronize ->css_online() and ->css_offline() against tree-walking |
233 | * operations. |
234 | * |
235 | * It is allowed to temporarily drop RCU read lock during iteration. The |
236 | * caller is responsible for ensuring that @pos remains accessible until |
237 | * the start of the next iteration by, for example, bumping the css refcnt. |
238 | */ |
239 | #define css_for_each_descendant_pre(pos, css) \ |
240 | for ((pos) = css_next_descendant_pre(NULL, (css)); (pos); \ |
241 | (pos) = css_next_descendant_pre((pos), (css))) |
242 | |
243 | /** |
244 | * css_for_each_descendant_post - post-order walk of a css's descendants |
245 | * @pos: the css * to use as the loop cursor |
246 | * @css: css whose descendants to walk |
247 | * |
248 | * Similar to css_for_each_descendant_pre() but performs post-order |
249 | * traversal instead. @root is included in the iteration and the last |
250 | * node to be visited. |
251 | * |
252 | * If a subsystem synchronizes ->css_online() and the start of iteration, a |
253 | * css which finished ->css_online() is guaranteed to be visible in the |
254 | * future iterations and will stay visible until the last reference is put. |
255 | * A css which hasn't finished ->css_online() or already finished |
256 | * ->css_offline() may show up during traversal. It's each subsystem's |
257 | * responsibility to synchronize against on/offlining. |
258 | * |
259 | * Note that the walk visibility guarantee example described in pre-order |
260 | * walk doesn't apply the same to post-order walks. |
261 | */ |
262 | #define css_for_each_descendant_post(pos, css) \ |
263 | for ((pos) = css_next_descendant_post(NULL, (css)); (pos); \ |
264 | (pos) = css_next_descendant_post((pos), (css))) |
265 | |
266 | /** |
267 | * cgroup_taskset_for_each - iterate cgroup_taskset |
268 | * @task: the loop cursor |
269 | * @dst_css: the destination css |
270 | * @tset: taskset to iterate |
271 | * |
272 | * @tset may contain multiple tasks and they may belong to multiple |
273 | * processes. |
274 | * |
275 | * On the v2 hierarchy, there may be tasks from multiple processes and they |
276 | * may not share the source or destination csses. |
277 | * |
278 | * On traditional hierarchies, when there are multiple tasks in @tset, if a |
279 | * task of a process is in @tset, all tasks of the process are in @tset. |
280 | * Also, all are guaranteed to share the same source and destination csses. |
281 | * |
282 | * Iteration is not in any specific order. |
283 | */ |
284 | #define cgroup_taskset_for_each(task, dst_css, tset) \ |
285 | for ((task) = cgroup_taskset_first((tset), &(dst_css)); \ |
286 | (task); \ |
287 | (task) = cgroup_taskset_next((tset), &(dst_css))) |
288 | |
289 | /** |
290 | * cgroup_taskset_for_each_leader - iterate group leaders in a cgroup_taskset |
291 | * @leader: the loop cursor |
292 | * @dst_css: the destination css |
293 | * @tset: taskset to iterate |
294 | * |
295 | * Iterate threadgroup leaders of @tset. For single-task migrations, @tset |
296 | * may not contain any. |
297 | */ |
298 | #define cgroup_taskset_for_each_leader(leader, dst_css, tset) \ |
299 | for ((leader) = cgroup_taskset_first((tset), &(dst_css)); \ |
300 | (leader); \ |
301 | (leader) = cgroup_taskset_next((tset), &(dst_css))) \ |
302 | if ((leader) != (leader)->group_leader) \ |
303 | ; \ |
304 | else |
305 | |
306 | /* |
307 | * Inline functions. |
308 | */ |
309 | |
310 | static inline u64 cgroup_id(const struct cgroup *cgrp) |
311 | { |
312 | return cgrp->kn->id; |
313 | } |
314 | |
315 | /** |
316 | * css_get - obtain a reference on the specified css |
317 | * @css: target css |
318 | * |
319 | * The caller must already have a reference. |
320 | */ |
321 | static inline void css_get(struct cgroup_subsys_state *css) |
322 | { |
323 | if (!(css->flags & CSS_NO_REF)) |
324 | percpu_ref_get(&css->refcnt); |
325 | } |
326 | |
327 | /** |
328 | * css_get_many - obtain references on the specified css |
329 | * @css: target css |
330 | * @n: number of references to get |
331 | * |
332 | * The caller must already have a reference. |
333 | */ |
334 | static inline void css_get_many(struct cgroup_subsys_state *css, unsigned int n) |
335 | { |
336 | if (!(css->flags & CSS_NO_REF)) |
337 | percpu_ref_get_many(&css->refcnt, n); |
338 | } |
339 | |
340 | /** |
341 | * css_tryget - try to obtain a reference on the specified css |
342 | * @css: target css |
343 | * |
344 | * Obtain a reference on @css unless it already has reached zero and is |
345 | * being released. This function doesn't care whether @css is on or |
346 | * offline. The caller naturally needs to ensure that @css is accessible |
347 | * but doesn't have to be holding a reference on it - IOW, RCU protected |
348 | * access is good enough for this function. Returns %true if a reference |
349 | * count was successfully obtained; %false otherwise. |
350 | */ |
351 | static inline bool css_tryget(struct cgroup_subsys_state *css) |
352 | { |
353 | if (!(css->flags & CSS_NO_REF)) |
354 | return percpu_ref_tryget(&css->refcnt); |
355 | return true; |
356 | } |
357 | |
358 | /** |
359 | * css_tryget_online - try to obtain a reference on the specified css if online |
360 | * @css: target css |
361 | * |
362 | * Obtain a reference on @css if it's online. The caller naturally needs |
363 | * to ensure that @css is accessible but doesn't have to be holding a |
364 | * reference on it - IOW, RCU protected access is good enough for this |
365 | * function. Returns %true if a reference count was successfully obtained; |
366 | * %false otherwise. |
367 | */ |
368 | static inline bool css_tryget_online(struct cgroup_subsys_state *css) |
369 | { |
370 | if (!(css->flags & CSS_NO_REF)) |
371 | return percpu_ref_tryget_live(&css->refcnt); |
372 | return true; |
373 | } |
374 | |
375 | /** |
376 | * css_is_dying - test whether the specified css is dying |
377 | * @css: target css |
378 | * |
379 | * Test whether @css is in the process of offlining or already offline. In |
380 | * most cases, ->css_online() and ->css_offline() callbacks should be |
381 | * enough; however, the actual offline operations are RCU delayed and this |
382 | * test returns %true also when @css is scheduled to be offlined. |
383 | * |
384 | * This is useful, for example, when the use case requires synchronous |
385 | * behavior with respect to cgroup removal. cgroup removal schedules css |
386 | * offlining but the css can seem alive while the operation is being |
387 | * delayed. If the delay affects user visible semantics, this test can be |
388 | * used to resolve the situation. |
389 | */ |
390 | static inline bool css_is_dying(struct cgroup_subsys_state *css) |
391 | { |
392 | return !(css->flags & CSS_NO_REF) && percpu_ref_is_dying(&css->refcnt); |
393 | } |
394 | |
395 | /** |
396 | * css_put - put a css reference |
397 | * @css: target css |
398 | * |
399 | * Put a reference obtained via css_get() and css_tryget_online(). |
400 | */ |
401 | static inline void css_put(struct cgroup_subsys_state *css) |
402 | { |
403 | if (!(css->flags & CSS_NO_REF)) |
404 | percpu_ref_put(&css->refcnt); |
405 | } |
406 | |
407 | /** |
408 | * css_put_many - put css references |
409 | * @css: target css |
410 | * @n: number of references to put |
411 | * |
412 | * Put references obtained via css_get() and css_tryget_online(). |
413 | */ |
414 | static inline void css_put_many(struct cgroup_subsys_state *css, unsigned int n) |
415 | { |
416 | if (!(css->flags & CSS_NO_REF)) |
417 | percpu_ref_put_many(&css->refcnt, n); |
418 | } |
419 | |
420 | static inline void cgroup_get(struct cgroup *cgrp) |
421 | { |
422 | css_get(&cgrp->self); |
423 | } |
424 | |
425 | static inline bool cgroup_tryget(struct cgroup *cgrp) |
426 | { |
427 | return css_tryget(&cgrp->self); |
428 | } |
429 | |
430 | static inline void cgroup_put(struct cgroup *cgrp) |
431 | { |
432 | css_put(&cgrp->self); |
433 | } |
434 | |
435 | /** |
436 | * task_css_set_check - obtain a task's css_set with extra access conditions |
437 | * @task: the task to obtain css_set for |
438 | * @__c: extra condition expression to be passed to rcu_dereference_check() |
439 | * |
440 | * A task's css_set is RCU protected, initialized and exited while holding |
441 | * task_lock(), and can only be modified while holding both cgroup_mutex |
442 | * and task_lock() while the task is alive. This macro verifies that the |
443 | * caller is inside proper critical section and returns @task's css_set. |
444 | * |
445 | * The caller can also specify additional allowed conditions via @__c, such |
446 | * as locks used during the cgroup_subsys::attach() methods. |
447 | */ |
448 | #ifdef CONFIG_PROVE_RCU |
449 | extern struct mutex cgroup_mutex; |
450 | extern spinlock_t css_set_lock; |
451 | #define task_css_set_check(task, __c) \ |
452 | rcu_dereference_check((task)->cgroups, \ |
453 | rcu_read_lock_sched_held() || \ |
454 | lockdep_is_held(&cgroup_mutex) || \ |
455 | lockdep_is_held(&css_set_lock) || \ |
456 | ((task)->flags & PF_EXITING) || (__c)) |
457 | #else |
458 | #define task_css_set_check(task, __c) \ |
459 | rcu_dereference((task)->cgroups) |
460 | #endif |
461 | |
462 | /** |
463 | * task_css_check - obtain css for (task, subsys) w/ extra access conds |
464 | * @task: the target task |
465 | * @subsys_id: the target subsystem ID |
466 | * @__c: extra condition expression to be passed to rcu_dereference_check() |
467 | * |
468 | * Return the cgroup_subsys_state for the (@task, @subsys_id) pair. The |
469 | * synchronization rules are the same as task_css_set_check(). |
470 | */ |
471 | #define task_css_check(task, subsys_id, __c) \ |
472 | task_css_set_check((task), (__c))->subsys[(subsys_id)] |
473 | |
474 | /** |
475 | * task_css_set - obtain a task's css_set |
476 | * @task: the task to obtain css_set for |
477 | * |
478 | * See task_css_set_check(). |
479 | */ |
480 | static inline struct css_set *task_css_set(struct task_struct *task) |
481 | { |
482 | return task_css_set_check(task, false); |
483 | } |
484 | |
485 | /** |
486 | * task_css - obtain css for (task, subsys) |
487 | * @task: the target task |
488 | * @subsys_id: the target subsystem ID |
489 | * |
490 | * See task_css_check(). |
491 | */ |
492 | static inline struct cgroup_subsys_state *task_css(struct task_struct *task, |
493 | int subsys_id) |
494 | { |
495 | return task_css_check(task, subsys_id, false); |
496 | } |
497 | |
498 | /** |
499 | * task_get_css - find and get the css for (task, subsys) |
500 | * @task: the target task |
501 | * @subsys_id: the target subsystem ID |
502 | * |
503 | * Find the css for the (@task, @subsys_id) combination, increment a |
504 | * reference on and return it. This function is guaranteed to return a |
505 | * valid css. The returned css may already have been offlined. |
506 | */ |
507 | static inline struct cgroup_subsys_state * |
508 | task_get_css(struct task_struct *task, int subsys_id) |
509 | { |
510 | struct cgroup_subsys_state *css; |
511 | |
512 | rcu_read_lock(); |
513 | while (true) { |
514 | css = task_css(task, subsys_id); |
515 | /* |
516 | * Can't use css_tryget_online() here. A task which has |
517 | * PF_EXITING set may stay associated with an offline css. |
518 | * If such task calls this function, css_tryget_online() |
519 | * will keep failing. |
520 | */ |
521 | if (likely(css_tryget(css))) |
522 | break; |
523 | cpu_relax(); |
524 | } |
525 | rcu_read_unlock(); |
526 | return css; |
527 | } |
528 | |
529 | /** |
530 | * task_css_is_root - test whether a task belongs to the root css |
531 | * @task: the target task |
532 | * @subsys_id: the target subsystem ID |
533 | * |
534 | * Test whether @task belongs to the root css on the specified subsystem. |
535 | * May be invoked in any context. |
536 | */ |
537 | static inline bool task_css_is_root(struct task_struct *task, int subsys_id) |
538 | { |
539 | return task_css_check(task, subsys_id, true) == |
540 | init_css_set.subsys[subsys_id]; |
541 | } |
542 | |
543 | static inline struct cgroup *task_cgroup(struct task_struct *task, |
544 | int subsys_id) |
545 | { |
546 | return task_css(task, subsys_id)->cgroup; |
547 | } |
548 | |
549 | static inline struct cgroup *task_dfl_cgroup(struct task_struct *task) |
550 | { |
551 | return task_css_set(task)->dfl_cgrp; |
552 | } |
553 | |
554 | static inline struct cgroup *cgroup_parent(struct cgroup *cgrp) |
555 | { |
556 | struct cgroup_subsys_state *parent_css = cgrp->self.parent; |
557 | |
558 | if (parent_css) |
559 | return container_of(parent_css, struct cgroup, self); |
560 | return NULL; |
561 | } |
562 | |
563 | /** |
564 | * cgroup_is_descendant - test ancestry |
565 | * @cgrp: the cgroup to be tested |
566 | * @ancestor: possible ancestor of @cgrp |
567 | * |
568 | * Test whether @cgrp is a descendant of @ancestor. It also returns %true |
569 | * if @cgrp == @ancestor. This function is safe to call as long as @cgrp |
570 | * and @ancestor are accessible. |
571 | */ |
572 | static inline bool cgroup_is_descendant(struct cgroup *cgrp, |
573 | struct cgroup *ancestor) |
574 | { |
575 | if (cgrp->root != ancestor->root || cgrp->level < ancestor->level) |
576 | return false; |
577 | return cgrp->ancestor_ids[ancestor->level] == cgroup_id(ancestor); |
578 | } |
579 | |
580 | /** |
581 | * cgroup_ancestor - find ancestor of cgroup |
582 | * @cgrp: cgroup to find ancestor of |
583 | * @ancestor_level: level of ancestor to find starting from root |
584 | * |
585 | * Find ancestor of cgroup at specified level starting from root if it exists |
586 | * and return pointer to it. Return NULL if @cgrp doesn't have ancestor at |
587 | * @ancestor_level. |
588 | * |
589 | * This function is safe to call as long as @cgrp is accessible. |
590 | */ |
591 | static inline struct cgroup *cgroup_ancestor(struct cgroup *cgrp, |
592 | int ancestor_level) |
593 | { |
594 | if (cgrp->level < ancestor_level) |
595 | return NULL; |
596 | while (cgrp && cgrp->level > ancestor_level) |
597 | cgrp = cgroup_parent(cgrp); |
598 | return cgrp; |
599 | } |
600 | |
601 | /** |
602 | * task_under_cgroup_hierarchy - test task's membership of cgroup ancestry |
603 | * @task: the task to be tested |
604 | * @ancestor: possible ancestor of @task's cgroup |
605 | * |
606 | * Tests whether @task's default cgroup hierarchy is a descendant of @ancestor. |
607 | * It follows all the same rules as cgroup_is_descendant, and only applies |
608 | * to the default hierarchy. |
609 | */ |
610 | static inline bool task_under_cgroup_hierarchy(struct task_struct *task, |
611 | struct cgroup *ancestor) |
612 | { |
613 | struct css_set *cset = task_css_set(task); |
614 | |
615 | return cgroup_is_descendant(cset->dfl_cgrp, ancestor); |
616 | } |
617 | |
618 | /* no synchronization, the result can only be used as a hint */ |
619 | static inline bool cgroup_is_populated(struct cgroup *cgrp) |
620 | { |
621 | return cgrp->nr_populated_csets + cgrp->nr_populated_domain_children + |
622 | cgrp->nr_populated_threaded_children; |
623 | } |
624 | |
625 | /* returns ino associated with a cgroup */ |
626 | static inline ino_t cgroup_ino(struct cgroup *cgrp) |
627 | { |
628 | return kernfs_ino(cgrp->kn); |
629 | } |
630 | |
631 | /* cft/css accessors for cftype->write() operation */ |
632 | static inline struct cftype *of_cft(struct kernfs_open_file *of) |
633 | { |
634 | return of->kn->priv; |
635 | } |
636 | |
637 | struct cgroup_subsys_state *of_css(struct kernfs_open_file *of); |
638 | |
639 | /* cft/css accessors for cftype->seq_*() operations */ |
640 | static inline struct cftype *seq_cft(struct seq_file *seq) |
641 | { |
642 | return of_cft(seq->private); |
643 | } |
644 | |
645 | static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq) |
646 | { |
647 | return of_css(seq->private); |
648 | } |
649 | |
650 | /* |
651 | * Name / path handling functions. All are thin wrappers around the kernfs |
652 | * counterparts and can be called under any context. |
653 | */ |
654 | |
655 | static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen) |
656 | { |
657 | return kernfs_name(cgrp->kn, buf, buflen); |
658 | } |
659 | |
660 | static inline int cgroup_path(struct cgroup *cgrp, char *buf, size_t buflen) |
661 | { |
662 | return kernfs_path(cgrp->kn, buf, buflen); |
663 | } |
664 | |
665 | static inline void pr_cont_cgroup_name(struct cgroup *cgrp) |
666 | { |
667 | pr_cont_kernfs_name(cgrp->kn); |
668 | } |
669 | |
670 | static inline void pr_cont_cgroup_path(struct cgroup *cgrp) |
671 | { |
672 | pr_cont_kernfs_path(cgrp->kn); |
673 | } |
674 | |
675 | static inline struct psi_group *cgroup_psi(struct cgroup *cgrp) |
676 | { |
677 | return cgrp->psi; |
678 | } |
679 | |
680 | bool cgroup_psi_enabled(void); |
681 | |
682 | static inline void cgroup_init_kthreadd(void) |
683 | { |
684 | /* |
685 | * kthreadd is inherited by all kthreads, keep it in the root so |
686 | * that the new kthreads are guaranteed to stay in the root until |
687 | * initialization is finished. |
688 | */ |
689 | current->no_cgroup_migration = 1; |
690 | } |
691 | |
692 | static inline void cgroup_kthread_ready(void) |
693 | { |
694 | /* |
695 | * This kthread finished initialization. The creator should have |
696 | * set PF_NO_SETAFFINITY if this kthread should stay in the root. |
697 | */ |
698 | current->no_cgroup_migration = 0; |
699 | } |
700 | |
701 | void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen); |
702 | struct cgroup *cgroup_get_from_id(u64 id); |
703 | #else /* !CONFIG_CGROUPS */ |
704 | |
705 | struct cgroup_subsys_state; |
706 | struct cgroup; |
707 | |
708 | static inline u64 cgroup_id(const struct cgroup *cgrp) { return 1; } |
709 | static inline void css_get(struct cgroup_subsys_state *css) {} |
710 | static inline void css_put(struct cgroup_subsys_state *css) {} |
711 | static inline int cgroup_attach_task_all(struct task_struct *from, |
712 | struct task_struct *t) { return 0; } |
713 | static inline int cgroupstats_build(struct cgroupstats *stats, |
714 | struct dentry *dentry) { return -EINVAL; } |
715 | |
716 | static inline void cgroup_fork(struct task_struct *p) {} |
717 | static inline int cgroup_can_fork(struct task_struct *p, |
718 | struct kernel_clone_args *kargs) { return 0; } |
719 | static inline void cgroup_cancel_fork(struct task_struct *p, |
720 | struct kernel_clone_args *kargs) {} |
721 | static inline void cgroup_post_fork(struct task_struct *p, |
722 | struct kernel_clone_args *kargs) {} |
723 | static inline void cgroup_exit(struct task_struct *p) {} |
724 | static inline void cgroup_release(struct task_struct *p) {} |
725 | static inline void cgroup_free(struct task_struct *p) {} |
726 | |
727 | static inline int cgroup_init_early(void) { return 0; } |
728 | static inline int cgroup_init(void) { return 0; } |
729 | static inline void cgroup_init_kthreadd(void) {} |
730 | static inline void cgroup_kthread_ready(void) {} |
731 | |
732 | static inline struct cgroup *cgroup_parent(struct cgroup *cgrp) |
733 | { |
734 | return NULL; |
735 | } |
736 | |
737 | static inline struct psi_group *cgroup_psi(struct cgroup *cgrp) |
738 | { |
739 | return NULL; |
740 | } |
741 | |
742 | static inline bool cgroup_psi_enabled(void) |
743 | { |
744 | return false; |
745 | } |
746 | |
747 | static inline bool task_under_cgroup_hierarchy(struct task_struct *task, |
748 | struct cgroup *ancestor) |
749 | { |
750 | return true; |
751 | } |
752 | |
753 | static inline void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen) |
754 | {} |
755 | |
756 | static inline struct cgroup *cgroup_get_from_id(u64 id) |
757 | { |
758 | return NULL; |
759 | } |
760 | #endif /* !CONFIG_CGROUPS */ |
761 | |
762 | #ifdef CONFIG_CGROUPS |
763 | /* |
764 | * cgroup scalable recursive statistics. |
765 | */ |
766 | void cgroup_rstat_updated(struct cgroup *cgrp, int cpu); |
767 | void cgroup_rstat_flush(struct cgroup *cgrp); |
768 | void cgroup_rstat_flush_irqsafe(struct cgroup *cgrp); |
769 | void cgroup_rstat_flush_hold(struct cgroup *cgrp); |
770 | void cgroup_rstat_flush_release(void); |
771 | |
772 | /* |
773 | * Basic resource stats. |
774 | */ |
775 | #ifdef CONFIG_CGROUP_CPUACCT |
776 | void cpuacct_charge(struct task_struct *tsk, u64 cputime); |
777 | void cpuacct_account_field(struct task_struct *tsk, int index, u64 val); |
778 | #else |
779 | static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {} |
780 | static inline void cpuacct_account_field(struct task_struct *tsk, int index, |
781 | u64 val) {} |
782 | #endif |
783 | |
784 | void __cgroup_account_cputime(struct cgroup *cgrp, u64 delta_exec); |
785 | void __cgroup_account_cputime_field(struct cgroup *cgrp, |
786 | enum cpu_usage_stat index, u64 delta_exec); |
787 | |
788 | static inline void cgroup_account_cputime(struct task_struct *task, |
789 | u64 delta_exec) |
790 | { |
791 | struct cgroup *cgrp; |
792 | |
793 | cpuacct_charge(task, delta_exec); |
794 | |
795 | cgrp = task_dfl_cgroup(task); |
796 | if (cgroup_parent(cgrp)) |
797 | __cgroup_account_cputime(cgrp, delta_exec); |
798 | } |
799 | |
800 | static inline void cgroup_account_cputime_field(struct task_struct *task, |
801 | enum cpu_usage_stat index, |
802 | u64 delta_exec) |
803 | { |
804 | struct cgroup *cgrp; |
805 | |
806 | cpuacct_account_field(task, index, delta_exec); |
807 | |
808 | cgrp = task_dfl_cgroup(task); |
809 | if (cgroup_parent(cgrp)) |
810 | __cgroup_account_cputime_field(cgrp, index, delta_exec); |
811 | } |
812 | |
813 | #else /* CONFIG_CGROUPS */ |
814 | |
815 | static inline void cgroup_account_cputime(struct task_struct *task, |
816 | u64 delta_exec) {} |
817 | static inline void cgroup_account_cputime_field(struct task_struct *task, |
818 | enum cpu_usage_stat index, |
819 | u64 delta_exec) {} |
820 | |
821 | #endif /* CONFIG_CGROUPS */ |
822 | |
823 | /* |
824 | * sock->sk_cgrp_data handling. For more info, see sock_cgroup_data |
825 | * definition in cgroup-defs.h. |
826 | */ |
827 | #ifdef CONFIG_SOCK_CGROUP_DATA |
828 | |
829 | void cgroup_sk_alloc(struct sock_cgroup_data *skcd); |
830 | void cgroup_sk_clone(struct sock_cgroup_data *skcd); |
831 | void cgroup_sk_free(struct sock_cgroup_data *skcd); |
832 | |
833 | static inline struct cgroup *sock_cgroup_ptr(struct sock_cgroup_data *skcd) |
834 | { |
835 | return skcd->cgroup; |
836 | } |
837 | |
838 | #else /* CONFIG_CGROUP_DATA */ |
839 | |
840 | static inline void cgroup_sk_alloc(struct sock_cgroup_data *skcd) {} |
841 | static inline void cgroup_sk_clone(struct sock_cgroup_data *skcd) {} |
842 | static inline void cgroup_sk_free(struct sock_cgroup_data *skcd) {} |
843 | |
844 | #endif /* CONFIG_CGROUP_DATA */ |
845 | |
846 | struct cgroup_namespace { |
847 | struct ns_common ns; |
848 | struct user_namespace *user_ns; |
849 | struct ucounts *ucounts; |
850 | struct css_set *root_cset; |
851 | }; |
852 | |
853 | extern struct cgroup_namespace init_cgroup_ns; |
854 | |
855 | #ifdef CONFIG_CGROUPS |
856 | |
857 | void free_cgroup_ns(struct cgroup_namespace *ns); |
858 | |
859 | struct cgroup_namespace *copy_cgroup_ns(unsigned long flags, |
860 | struct user_namespace *user_ns, |
861 | struct cgroup_namespace *old_ns); |
862 | |
863 | int cgroup_path_ns(struct cgroup *cgrp, char *buf, size_t buflen, |
864 | struct cgroup_namespace *ns); |
865 | |
866 | #else /* !CONFIG_CGROUPS */ |
867 | |
868 | static inline void free_cgroup_ns(struct cgroup_namespace *ns) { } |
869 | static inline struct cgroup_namespace * |
870 | copy_cgroup_ns(unsigned long flags, struct user_namespace *user_ns, |
871 | struct cgroup_namespace *old_ns) |
872 | { |
873 | return old_ns; |
874 | } |
875 | |
876 | #endif /* !CONFIG_CGROUPS */ |
877 | |
878 | static inline void get_cgroup_ns(struct cgroup_namespace *ns) |
879 | { |
880 | if (ns) |
881 | refcount_inc(&ns->ns.count); |
882 | } |
883 | |
884 | static inline void put_cgroup_ns(struct cgroup_namespace *ns) |
885 | { |
886 | if (ns && refcount_dec_and_test(&ns->ns.count)) |
887 | free_cgroup_ns(ns); |
888 | } |
889 | |
890 | #ifdef CONFIG_CGROUPS |
891 | |
892 | void cgroup_enter_frozen(void); |
893 | void cgroup_leave_frozen(bool always_leave); |
894 | void cgroup_update_frozen(struct cgroup *cgrp); |
895 | void cgroup_freeze(struct cgroup *cgrp, bool freeze); |
896 | void cgroup_freezer_migrate_task(struct task_struct *task, struct cgroup *src, |
897 | struct cgroup *dst); |
898 | |
899 | static inline bool cgroup_task_frozen(struct task_struct *task) |
900 | { |
901 | return task->frozen; |
902 | } |
903 | |
904 | #else /* !CONFIG_CGROUPS */ |
905 | |
906 | static inline void cgroup_enter_frozen(void) { } |
907 | static inline void cgroup_leave_frozen(bool always_leave) { } |
908 | static inline bool cgroup_task_frozen(struct task_struct *task) |
909 | { |
910 | return false; |
911 | } |
912 | |
913 | #endif /* !CONFIG_CGROUPS */ |
914 | |
915 | #ifdef CONFIG_CGROUP_BPF |
916 | static inline void cgroup_bpf_get(struct cgroup *cgrp) |
917 | { |
918 | percpu_ref_get(&cgrp->bpf.refcnt); |
919 | } |
920 | |
921 | static inline void cgroup_bpf_put(struct cgroup *cgrp) |
922 | { |
923 | percpu_ref_put(&cgrp->bpf.refcnt); |
924 | } |
925 | |
926 | #else /* CONFIG_CGROUP_BPF */ |
927 | |
928 | static inline void cgroup_bpf_get(struct cgroup *cgrp) {} |
929 | static inline void cgroup_bpf_put(struct cgroup *cgrp) {} |
930 | |
931 | #endif /* CONFIG_CGROUP_BPF */ |
932 | |
933 | #endif /* _LINUX_CGROUP_H */ |
934 | |