• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 hierarhcy 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 
cgroup_id(struct cgroup * cgrp)310 static inline u64 cgroup_id(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  */
css_get(struct cgroup_subsys_state * css)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  */
css_get_many(struct cgroup_subsys_state * css,unsigned int n)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  */
css_tryget(struct cgroup_subsys_state * css)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  */
css_tryget_online(struct cgroup_subsys_state * css)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  */
css_is_dying(struct cgroup_subsys_state * css)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  */
css_put(struct cgroup_subsys_state * css)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  */
css_put_many(struct cgroup_subsys_state * css,unsigned int n)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 
cgroup_get(struct cgroup * cgrp)420 static inline void cgroup_get(struct cgroup *cgrp)
421 {
422 	css_get(&cgrp->self);
423 }
424 
cgroup_tryget(struct cgroup * cgrp)425 static inline bool cgroup_tryget(struct cgroup *cgrp)
426 {
427 	return css_tryget(&cgrp->self);
428 }
429 
cgroup_put(struct cgroup * cgrp)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 		lockdep_is_held(&cgroup_mutex) ||			\
454 		lockdep_is_held(&css_set_lock) ||			\
455 		((task)->flags & PF_EXITING) || (__c))
456 #else
457 #define task_css_set_check(task, __c)					\
458 	rcu_dereference((task)->cgroups)
459 #endif
460 
461 /**
462  * task_css_check - obtain css for (task, subsys) w/ extra access conds
463  * @task: the target task
464  * @subsys_id: the target subsystem ID
465  * @__c: extra condition expression to be passed to rcu_dereference_check()
466  *
467  * Return the cgroup_subsys_state for the (@task, @subsys_id) pair.  The
468  * synchronization rules are the same as task_css_set_check().
469  */
470 #define task_css_check(task, subsys_id, __c)				\
471 	task_css_set_check((task), (__c))->subsys[(subsys_id)]
472 
473 /**
474  * task_css_set - obtain a task's css_set
475  * @task: the task to obtain css_set for
476  *
477  * See task_css_set_check().
478  */
task_css_set(struct task_struct * task)479 static inline struct css_set *task_css_set(struct task_struct *task)
480 {
481 	return task_css_set_check(task, false);
482 }
483 
484 /**
485  * task_css - obtain css for (task, subsys)
486  * @task: the target task
487  * @subsys_id: the target subsystem ID
488  *
489  * See task_css_check().
490  */
task_css(struct task_struct * task,int subsys_id)491 static inline struct cgroup_subsys_state *task_css(struct task_struct *task,
492 						   int subsys_id)
493 {
494 	return task_css_check(task, subsys_id, false);
495 }
496 
497 /**
498  * task_get_css - find and get the css for (task, subsys)
499  * @task: the target task
500  * @subsys_id: the target subsystem ID
501  *
502  * Find the css for the (@task, @subsys_id) combination, increment a
503  * reference on and return it.  This function is guaranteed to return a
504  * valid css.  The returned css may already have been offlined.
505  */
506 static inline struct cgroup_subsys_state *
task_get_css(struct task_struct * task,int subsys_id)507 task_get_css(struct task_struct *task, int subsys_id)
508 {
509 	struct cgroup_subsys_state *css;
510 
511 	rcu_read_lock();
512 	while (true) {
513 		css = task_css(task, subsys_id);
514 		/*
515 		 * Can't use css_tryget_online() here.  A task which has
516 		 * PF_EXITING set may stay associated with an offline css.
517 		 * If such task calls this function, css_tryget_online()
518 		 * will keep failing.
519 		 */
520 		if (likely(css_tryget(css)))
521 			break;
522 		cpu_relax();
523 	}
524 	rcu_read_unlock();
525 	return css;
526 }
527 
528 /**
529  * task_css_is_root - test whether a task belongs to the root css
530  * @task: the target task
531  * @subsys_id: the target subsystem ID
532  *
533  * Test whether @task belongs to the root css on the specified subsystem.
534  * May be invoked in any context.
535  */
task_css_is_root(struct task_struct * task,int subsys_id)536 static inline bool task_css_is_root(struct task_struct *task, int subsys_id)
537 {
538 	return task_css_check(task, subsys_id, true) ==
539 		init_css_set.subsys[subsys_id];
540 }
541 
task_cgroup(struct task_struct * task,int subsys_id)542 static inline struct cgroup *task_cgroup(struct task_struct *task,
543 					 int subsys_id)
544 {
545 	return task_css(task, subsys_id)->cgroup;
546 }
547 
task_dfl_cgroup(struct task_struct * task)548 static inline struct cgroup *task_dfl_cgroup(struct task_struct *task)
549 {
550 	return task_css_set(task)->dfl_cgrp;
551 }
552 
cgroup_parent(struct cgroup * cgrp)553 static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
554 {
555 	struct cgroup_subsys_state *parent_css = cgrp->self.parent;
556 
557 	if (parent_css)
558 		return container_of(parent_css, struct cgroup, self);
559 	return NULL;
560 }
561 
562 /**
563  * cgroup_is_descendant - test ancestry
564  * @cgrp: the cgroup to be tested
565  * @ancestor: possible ancestor of @cgrp
566  *
567  * Test whether @cgrp is a descendant of @ancestor.  It also returns %true
568  * if @cgrp == @ancestor.  This function is safe to call as long as @cgrp
569  * and @ancestor are accessible.
570  */
cgroup_is_descendant(struct cgroup * cgrp,struct cgroup * ancestor)571 static inline bool cgroup_is_descendant(struct cgroup *cgrp,
572 					struct cgroup *ancestor)
573 {
574 	if (cgrp->root != ancestor->root || cgrp->level < ancestor->level)
575 		return false;
576 	return cgrp->ancestor_ids[ancestor->level] == cgroup_id(ancestor);
577 }
578 
579 /**
580  * cgroup_ancestor - find ancestor of cgroup
581  * @cgrp: cgroup to find ancestor of
582  * @ancestor_level: level of ancestor to find starting from root
583  *
584  * Find ancestor of cgroup at specified level starting from root if it exists
585  * and return pointer to it. Return NULL if @cgrp doesn't have ancestor at
586  * @ancestor_level.
587  *
588  * This function is safe to call as long as @cgrp is accessible.
589  */
cgroup_ancestor(struct cgroup * cgrp,int ancestor_level)590 static inline struct cgroup *cgroup_ancestor(struct cgroup *cgrp,
591 					     int ancestor_level)
592 {
593 	if (cgrp->level < ancestor_level)
594 		return NULL;
595 	while (cgrp && cgrp->level > ancestor_level)
596 		cgrp = cgroup_parent(cgrp);
597 	return cgrp;
598 }
599 
600 /**
601  * task_under_cgroup_hierarchy - test task's membership of cgroup ancestry
602  * @task: the task to be tested
603  * @ancestor: possible ancestor of @task's cgroup
604  *
605  * Tests whether @task's default cgroup hierarchy is a descendant of @ancestor.
606  * It follows all the same rules as cgroup_is_descendant, and only applies
607  * to the default hierarchy.
608  */
task_under_cgroup_hierarchy(struct task_struct * task,struct cgroup * ancestor)609 static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
610 					       struct cgroup *ancestor)
611 {
612 	struct css_set *cset = task_css_set(task);
613 
614 	return cgroup_is_descendant(cset->dfl_cgrp, ancestor);
615 }
616 
617 /* no synchronization, the result can only be used as a hint */
cgroup_is_populated(struct cgroup * cgrp)618 static inline bool cgroup_is_populated(struct cgroup *cgrp)
619 {
620 	return cgrp->nr_populated_csets + cgrp->nr_populated_domain_children +
621 		cgrp->nr_populated_threaded_children;
622 }
623 
624 /* returns ino associated with a cgroup */
cgroup_ino(struct cgroup * cgrp)625 static inline ino_t cgroup_ino(struct cgroup *cgrp)
626 {
627 	return kernfs_ino(cgrp->kn);
628 }
629 
630 /* cft/css accessors for cftype->write() operation */
of_cft(struct kernfs_open_file * of)631 static inline struct cftype *of_cft(struct kernfs_open_file *of)
632 {
633 	return of->kn->priv;
634 }
635 
636 struct cgroup_subsys_state *of_css(struct kernfs_open_file *of);
637 
638 /* cft/css accessors for cftype->seq_*() operations */
seq_cft(struct seq_file * seq)639 static inline struct cftype *seq_cft(struct seq_file *seq)
640 {
641 	return of_cft(seq->private);
642 }
643 
seq_css(struct seq_file * seq)644 static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq)
645 {
646 	return of_css(seq->private);
647 }
648 
649 /*
650  * Name / path handling functions.  All are thin wrappers around the kernfs
651  * counterparts and can be called under any context.
652  */
653 
cgroup_name(struct cgroup * cgrp,char * buf,size_t buflen)654 static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen)
655 {
656 	return kernfs_name(cgrp->kn, buf, buflen);
657 }
658 
cgroup_path(struct cgroup * cgrp,char * buf,size_t buflen)659 static inline int cgroup_path(struct cgroup *cgrp, char *buf, size_t buflen)
660 {
661 	return kernfs_path(cgrp->kn, buf, buflen);
662 }
663 
pr_cont_cgroup_name(struct cgroup * cgrp)664 static inline void pr_cont_cgroup_name(struct cgroup *cgrp)
665 {
666 	pr_cont_kernfs_name(cgrp->kn);
667 }
668 
pr_cont_cgroup_path(struct cgroup * cgrp)669 static inline void pr_cont_cgroup_path(struct cgroup *cgrp)
670 {
671 	pr_cont_kernfs_path(cgrp->kn);
672 }
673 
cgroup_psi(struct cgroup * cgrp)674 static inline struct psi_group *cgroup_psi(struct cgroup *cgrp)
675 {
676 	return &cgrp->psi;
677 }
678 
cgroup_init_kthreadd(void)679 static inline void cgroup_init_kthreadd(void)
680 {
681 	/*
682 	 * kthreadd is inherited by all kthreads, keep it in the root so
683 	 * that the new kthreads are guaranteed to stay in the root until
684 	 * initialization is finished.
685 	 */
686 	current->no_cgroup_migration = 1;
687 }
688 
cgroup_kthread_ready(void)689 static inline void cgroup_kthread_ready(void)
690 {
691 	/*
692 	 * This kthread finished initialization.  The creator should have
693 	 * set PF_NO_SETAFFINITY if this kthread should stay in the root.
694 	 */
695 	current->no_cgroup_migration = 0;
696 }
697 
698 void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen);
699 #else /* !CONFIG_CGROUPS */
700 
701 struct cgroup_subsys_state;
702 struct cgroup;
703 
cgroup_id(struct cgroup * cgrp)704 static inline u64 cgroup_id(struct cgroup *cgrp) { return 1; }
css_get(struct cgroup_subsys_state * css)705 static inline void css_get(struct cgroup_subsys_state *css) {}
css_put(struct cgroup_subsys_state * css)706 static inline void css_put(struct cgroup_subsys_state *css) {}
cgroup_attach_task_all(struct task_struct * from,struct task_struct * t)707 static inline int cgroup_attach_task_all(struct task_struct *from,
708 					 struct task_struct *t) { return 0; }
cgroupstats_build(struct cgroupstats * stats,struct dentry * dentry)709 static inline int cgroupstats_build(struct cgroupstats *stats,
710 				    struct dentry *dentry) { return -EINVAL; }
711 
cgroup_fork(struct task_struct * p)712 static inline void cgroup_fork(struct task_struct *p) {}
cgroup_can_fork(struct task_struct * p,struct kernel_clone_args * kargs)713 static inline int cgroup_can_fork(struct task_struct *p,
714 				  struct kernel_clone_args *kargs) { return 0; }
cgroup_cancel_fork(struct task_struct * p,struct kernel_clone_args * kargs)715 static inline void cgroup_cancel_fork(struct task_struct *p,
716 				      struct kernel_clone_args *kargs) {}
cgroup_post_fork(struct task_struct * p,struct kernel_clone_args * kargs)717 static inline void cgroup_post_fork(struct task_struct *p,
718 				    struct kernel_clone_args *kargs) {}
cgroup_exit(struct task_struct * p)719 static inline void cgroup_exit(struct task_struct *p) {}
cgroup_release(struct task_struct * p)720 static inline void cgroup_release(struct task_struct *p) {}
cgroup_free(struct task_struct * p)721 static inline void cgroup_free(struct task_struct *p) {}
722 
cgroup_init_early(void)723 static inline int cgroup_init_early(void) { return 0; }
cgroup_init(void)724 static inline int cgroup_init(void) { return 0; }
cgroup_init_kthreadd(void)725 static inline void cgroup_init_kthreadd(void) {}
cgroup_kthread_ready(void)726 static inline void cgroup_kthread_ready(void) {}
727 
cgroup_parent(struct cgroup * cgrp)728 static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
729 {
730 	return NULL;
731 }
732 
cgroup_psi(struct cgroup * cgrp)733 static inline struct psi_group *cgroup_psi(struct cgroup *cgrp)
734 {
735 	return NULL;
736 }
737 
task_under_cgroup_hierarchy(struct task_struct * task,struct cgroup * ancestor)738 static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
739 					       struct cgroup *ancestor)
740 {
741 	return true;
742 }
743 
cgroup_path_from_kernfs_id(u64 id,char * buf,size_t buflen)744 static inline void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen)
745 {}
746 #endif /* !CONFIG_CGROUPS */
747 
748 #ifdef CONFIG_CGROUPS
749 /*
750  * cgroup scalable recursive statistics.
751  */
752 void cgroup_rstat_updated(struct cgroup *cgrp, int cpu);
753 void cgroup_rstat_flush(struct cgroup *cgrp);
754 void cgroup_rstat_flush_irqsafe(struct cgroup *cgrp);
755 void cgroup_rstat_flush_hold(struct cgroup *cgrp);
756 void cgroup_rstat_flush_release(void);
757 
758 /*
759  * Basic resource stats.
760  */
761 #ifdef CONFIG_CGROUP_CPUACCT
762 void cpuacct_charge(struct task_struct *tsk, u64 cputime);
763 void cpuacct_account_field(struct task_struct *tsk, int index, u64 val);
764 #else
cpuacct_charge(struct task_struct * tsk,u64 cputime)765 static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {}
cpuacct_account_field(struct task_struct * tsk,int index,u64 val)766 static inline void cpuacct_account_field(struct task_struct *tsk, int index,
767 					 u64 val) {}
768 #endif
769 
770 void __cgroup_account_cputime(struct cgroup *cgrp, u64 delta_exec);
771 void __cgroup_account_cputime_field(struct cgroup *cgrp,
772 				    enum cpu_usage_stat index, u64 delta_exec);
773 
cgroup_account_cputime(struct task_struct * task,u64 delta_exec)774 static inline void cgroup_account_cputime(struct task_struct *task,
775 					  u64 delta_exec)
776 {
777 	struct cgroup *cgrp;
778 
779 	cpuacct_charge(task, delta_exec);
780 
781 	rcu_read_lock();
782 	cgrp = task_dfl_cgroup(task);
783 	if (cgroup_parent(cgrp))
784 		__cgroup_account_cputime(cgrp, delta_exec);
785 	rcu_read_unlock();
786 }
787 
cgroup_account_cputime_field(struct task_struct * task,enum cpu_usage_stat index,u64 delta_exec)788 static inline void cgroup_account_cputime_field(struct task_struct *task,
789 						enum cpu_usage_stat index,
790 						u64 delta_exec)
791 {
792 	struct cgroup *cgrp;
793 
794 	cpuacct_account_field(task, index, delta_exec);
795 
796 	rcu_read_lock();
797 	cgrp = task_dfl_cgroup(task);
798 	if (cgroup_parent(cgrp))
799 		__cgroup_account_cputime_field(cgrp, index, delta_exec);
800 	rcu_read_unlock();
801 }
802 
803 #else	/* CONFIG_CGROUPS */
804 
cgroup_account_cputime(struct task_struct * task,u64 delta_exec)805 static inline void cgroup_account_cputime(struct task_struct *task,
806 					  u64 delta_exec) {}
cgroup_account_cputime_field(struct task_struct * task,enum cpu_usage_stat index,u64 delta_exec)807 static inline void cgroup_account_cputime_field(struct task_struct *task,
808 						enum cpu_usage_stat index,
809 						u64 delta_exec) {}
810 
811 #endif	/* CONFIG_CGROUPS */
812 
813 /*
814  * sock->sk_cgrp_data handling.  For more info, see sock_cgroup_data
815  * definition in cgroup-defs.h.
816  */
817 #ifdef CONFIG_SOCK_CGROUP_DATA
818 
819 void cgroup_sk_alloc(struct sock_cgroup_data *skcd);
820 void cgroup_sk_clone(struct sock_cgroup_data *skcd);
821 void cgroup_sk_free(struct sock_cgroup_data *skcd);
822 
sock_cgroup_ptr(struct sock_cgroup_data * skcd)823 static inline struct cgroup *sock_cgroup_ptr(struct sock_cgroup_data *skcd)
824 {
825 	return skcd->cgroup;
826 }
827 
828 #else	/* CONFIG_CGROUP_DATA */
829 
cgroup_sk_alloc(struct sock_cgroup_data * skcd)830 static inline void cgroup_sk_alloc(struct sock_cgroup_data *skcd) {}
cgroup_sk_clone(struct sock_cgroup_data * skcd)831 static inline void cgroup_sk_clone(struct sock_cgroup_data *skcd) {}
cgroup_sk_free(struct sock_cgroup_data * skcd)832 static inline void cgroup_sk_free(struct sock_cgroup_data *skcd) {}
833 
834 #endif	/* CONFIG_CGROUP_DATA */
835 
836 struct cgroup_namespace {
837 	refcount_t		count;
838 	struct ns_common	ns;
839 	struct user_namespace	*user_ns;
840 	struct ucounts		*ucounts;
841 	struct css_set          *root_cset;
842 };
843 
844 extern struct cgroup_namespace init_cgroup_ns;
845 
846 #ifdef CONFIG_CGROUPS
847 
848 void free_cgroup_ns(struct cgroup_namespace *ns);
849 
850 struct cgroup_namespace *copy_cgroup_ns(unsigned long flags,
851 					struct user_namespace *user_ns,
852 					struct cgroup_namespace *old_ns);
853 
854 int cgroup_path_ns(struct cgroup *cgrp, char *buf, size_t buflen,
855 		   struct cgroup_namespace *ns);
856 
857 #else /* !CONFIG_CGROUPS */
858 
free_cgroup_ns(struct cgroup_namespace * ns)859 static inline void free_cgroup_ns(struct cgroup_namespace *ns) { }
860 static inline struct cgroup_namespace *
copy_cgroup_ns(unsigned long flags,struct user_namespace * user_ns,struct cgroup_namespace * old_ns)861 copy_cgroup_ns(unsigned long flags, struct user_namespace *user_ns,
862 	       struct cgroup_namespace *old_ns)
863 {
864 	return old_ns;
865 }
866 
867 #endif /* !CONFIG_CGROUPS */
868 
get_cgroup_ns(struct cgroup_namespace * ns)869 static inline void get_cgroup_ns(struct cgroup_namespace *ns)
870 {
871 	if (ns)
872 		refcount_inc(&ns->count);
873 }
874 
put_cgroup_ns(struct cgroup_namespace * ns)875 static inline void put_cgroup_ns(struct cgroup_namespace *ns)
876 {
877 	if (ns && refcount_dec_and_test(&ns->count))
878 		free_cgroup_ns(ns);
879 }
880 
881 #ifdef CONFIG_CGROUPS
882 
883 void cgroup_enter_frozen(void);
884 void cgroup_leave_frozen(bool always_leave);
885 void cgroup_update_frozen(struct cgroup *cgrp);
886 void cgroup_freeze(struct cgroup *cgrp, bool freeze);
887 void cgroup_freezer_migrate_task(struct task_struct *task, struct cgroup *src,
888 				 struct cgroup *dst);
889 
cgroup_task_freeze(struct task_struct * task)890 static inline bool cgroup_task_freeze(struct task_struct *task)
891 {
892 	bool ret;
893 
894 	if (task->flags & PF_KTHREAD)
895 		return false;
896 
897 	rcu_read_lock();
898 	ret = test_bit(CGRP_FREEZE, &task_dfl_cgroup(task)->flags);
899 	rcu_read_unlock();
900 
901 	return ret;
902 }
903 
cgroup_task_frozen(struct task_struct * task)904 static inline bool cgroup_task_frozen(struct task_struct *task)
905 {
906 	return task->frozen;
907 }
908 
909 #else /* !CONFIG_CGROUPS */
910 
cgroup_enter_frozen(void)911 static inline void cgroup_enter_frozen(void) { }
cgroup_leave_frozen(bool always_leave)912 static inline void cgroup_leave_frozen(bool always_leave) { }
cgroup_task_freeze(struct task_struct * task)913 static inline bool cgroup_task_freeze(struct task_struct *task)
914 {
915 	return false;
916 }
cgroup_task_frozen(struct task_struct * task)917 static inline bool cgroup_task_frozen(struct task_struct *task)
918 {
919 	return false;
920 }
921 
922 #endif /* !CONFIG_CGROUPS */
923 
924 #ifdef CONFIG_CGROUP_BPF
cgroup_bpf_get(struct cgroup * cgrp)925 static inline void cgroup_bpf_get(struct cgroup *cgrp)
926 {
927 	percpu_ref_get(&cgrp->bpf.refcnt);
928 }
929 
cgroup_bpf_put(struct cgroup * cgrp)930 static inline void cgroup_bpf_put(struct cgroup *cgrp)
931 {
932 	percpu_ref_put(&cgrp->bpf.refcnt);
933 }
934 
935 #else /* CONFIG_CGROUP_BPF */
936 
cgroup_bpf_get(struct cgroup * cgrp)937 static inline void cgroup_bpf_get(struct cgroup *cgrp) {}
cgroup_bpf_put(struct cgroup * cgrp)938 static inline void cgroup_bpf_put(struct cgroup *cgrp) {}
939 
940 #endif /* CONFIG_CGROUP_BPF */
941 
942 #endif /* _LINUX_CGROUP_H */
943