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