• 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 #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 ext_css_set init_ext_css_set;
75 #define init_css_set init_ext_css_set.cset
76 
77 #define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys;
78 #include <linux/cgroup_subsys.h>
79 #undef SUBSYS
80 
81 #define SUBSYS(_x)								\
82 	extern struct static_key_true _x ## _cgrp_subsys_enabled_key;		\
83 	extern struct static_key_true _x ## _cgrp_subsys_on_dfl_key;
84 #include <linux/cgroup_subsys.h>
85 #undef SUBSYS
86 
87 /**
88  * cgroup_subsys_enabled - fast test on whether a subsys is enabled
89  * @ss: subsystem in question
90  */
91 #define cgroup_subsys_enabled(ss)						\
92 	static_branch_likely(&ss ## _enabled_key)
93 
94 /**
95  * cgroup_subsys_on_dfl - fast test on whether a subsys is on default hierarchy
96  * @ss: subsystem in question
97  */
98 #define cgroup_subsys_on_dfl(ss)						\
99 	static_branch_likely(&ss ## _on_dfl_key)
100 
101 bool css_has_online_children(struct cgroup_subsys_state *css);
102 struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss);
103 struct cgroup_subsys_state *cgroup_e_css(struct cgroup *cgroup,
104 					 struct cgroup_subsys *ss);
105 struct cgroup_subsys_state *cgroup_get_e_css(struct cgroup *cgroup,
106 					     struct cgroup_subsys *ss);
107 struct cgroup_subsys_state *css_tryget_online_from_dir(struct dentry *dentry,
108 						       struct cgroup_subsys *ss);
109 
110 struct cgroup *cgroup_get_from_path(const char *path);
111 struct cgroup *cgroup_get_from_fd(int fd);
112 
113 int cgroup_attach_task_all(struct task_struct *from, struct task_struct *);
114 int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from);
115 
116 int cgroup_add_dfl_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
117 int cgroup_add_legacy_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
118 int cgroup_rm_cftypes(struct cftype *cfts);
119 void cgroup_file_notify(struct cgroup_file *cfile);
120 
121 int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen);
122 int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry);
123 int proc_cgroup_show(struct seq_file *m, struct pid_namespace *ns,
124 		     struct pid *pid, struct task_struct *tsk);
125 
126 void cgroup_fork(struct task_struct *p);
127 extern int cgroup_can_fork(struct task_struct *p);
128 extern void cgroup_cancel_fork(struct task_struct *p);
129 extern void cgroup_post_fork(struct task_struct *p);
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 cgroup_enable_task_cg_lists(void);
157 void css_task_iter_start(struct cgroup_subsys_state *css, unsigned int flags,
158 			 struct css_task_iter *it);
159 struct task_struct *css_task_iter_next(struct css_task_iter *it);
160 void css_task_iter_end(struct css_task_iter *it);
161 
162 /**
163  * css_for_each_child - iterate through children of a css
164  * @pos: the css * to use as the loop cursor
165  * @parent: css whose children to walk
166  *
167  * Walk @parent's children.  Must be called under rcu_read_lock().
168  *
169  * If a subsystem synchronizes ->css_online() and the start of iteration, a
170  * css which finished ->css_online() is guaranteed to be visible in the
171  * future iterations and will stay visible until the last reference is put.
172  * A css which hasn't finished ->css_online() or already finished
173  * ->css_offline() may show up during traversal.  It's each subsystem's
174  * responsibility to synchronize against on/offlining.
175  *
176  * It is allowed to temporarily drop RCU read lock during iteration.  The
177  * caller is responsible for ensuring that @pos remains accessible until
178  * the start of the next iteration by, for example, bumping the css refcnt.
179  */
180 #define css_for_each_child(pos, parent)					\
181 	for ((pos) = css_next_child(NULL, (parent)); (pos);		\
182 	     (pos) = css_next_child((pos), (parent)))
183 
184 /**
185  * css_for_each_descendant_pre - pre-order walk of a css's descendants
186  * @pos: the css * to use as the loop cursor
187  * @root: css whose descendants to walk
188  *
189  * Walk @root's descendants.  @root is included in the iteration and the
190  * first node to be visited.  Must be called under rcu_read_lock().
191  *
192  * If a subsystem synchronizes ->css_online() and the start of iteration, a
193  * css which finished ->css_online() is guaranteed to be visible in the
194  * future iterations and will stay visible until the last reference is put.
195  * A css which hasn't finished ->css_online() or already finished
196  * ->css_offline() may show up during traversal.  It's each subsystem's
197  * responsibility to synchronize against on/offlining.
198  *
199  * For example, the following guarantees that a descendant can't escape
200  * state updates of its ancestors.
201  *
202  * my_online(@css)
203  * {
204  *	Lock @css's parent and @css;
205  *	Inherit state from the parent;
206  *	Unlock both.
207  * }
208  *
209  * my_update_state(@css)
210  * {
211  *	css_for_each_descendant_pre(@pos, @css) {
212  *		Lock @pos;
213  *		if (@pos == @css)
214  *			Update @css's state;
215  *		else
216  *			Verify @pos is alive and inherit state from its parent;
217  *		Unlock @pos;
218  *	}
219  * }
220  *
221  * As long as the inheriting step, including checking the parent state, is
222  * enclosed inside @pos locking, double-locking the parent isn't necessary
223  * while inheriting.  The state update to the parent is guaranteed to be
224  * visible by walking order and, as long as inheriting operations to the
225  * same @pos are atomic to each other, multiple updates racing each other
226  * still result in the correct state.  It's guaranateed that at least one
227  * inheritance happens for any css after the latest update to its parent.
228  *
229  * If checking parent's state requires locking the parent, each inheriting
230  * iteration should lock and unlock both @pos->parent and @pos.
231  *
232  * Alternatively, a subsystem may choose to use a single global lock to
233  * synchronize ->css_online() and ->css_offline() against tree-walking
234  * operations.
235  *
236  * It is allowed to temporarily drop RCU read lock during iteration.  The
237  * caller is responsible for ensuring that @pos remains accessible until
238  * the start of the next iteration by, for example, bumping the css refcnt.
239  */
240 #define css_for_each_descendant_pre(pos, css)				\
241 	for ((pos) = css_next_descendant_pre(NULL, (css)); (pos);	\
242 	     (pos) = css_next_descendant_pre((pos), (css)))
243 
244 /**
245  * css_for_each_descendant_post - post-order walk of a css's descendants
246  * @pos: the css * to use as the loop cursor
247  * @css: css whose descendants to walk
248  *
249  * Similar to css_for_each_descendant_pre() but performs post-order
250  * traversal instead.  @root is included in the iteration and the last
251  * node to be visited.
252  *
253  * If a subsystem synchronizes ->css_online() and the start of iteration, a
254  * css which finished ->css_online() is guaranteed to be visible in the
255  * future iterations and will stay visible until the last reference is put.
256  * A css which hasn't finished ->css_online() or already finished
257  * ->css_offline() may show up during traversal.  It's each subsystem's
258  * responsibility to synchronize against on/offlining.
259  *
260  * Note that the walk visibility guarantee example described in pre-order
261  * walk doesn't apply the same to post-order walks.
262  */
263 #define css_for_each_descendant_post(pos, css)				\
264 	for ((pos) = css_next_descendant_post(NULL, (css)); (pos);	\
265 	     (pos) = css_next_descendant_post((pos), (css)))
266 
267 /**
268  * cgroup_taskset_for_each - iterate cgroup_taskset
269  * @task: the loop cursor
270  * @dst_css: the destination css
271  * @tset: taskset to iterate
272  *
273  * @tset may contain multiple tasks and they may belong to multiple
274  * processes.
275  *
276  * On the v2 hierarchy, there may be tasks from multiple processes and they
277  * may not share the source or destination csses.
278  *
279  * On traditional hierarchies, when there are multiple tasks in @tset, if a
280  * task of a process is in @tset, all tasks of the process are in @tset.
281  * Also, all are guaranteed to share the same source and destination csses.
282  *
283  * Iteration is not in any specific order.
284  */
285 #define cgroup_taskset_for_each(task, dst_css, tset)			\
286 	for ((task) = cgroup_taskset_first((tset), &(dst_css));		\
287 	     (task);							\
288 	     (task) = cgroup_taskset_next((tset), &(dst_css)))
289 
290 /**
291  * cgroup_taskset_for_each_leader - iterate group leaders in a cgroup_taskset
292  * @leader: the loop cursor
293  * @dst_css: the destination css
294  * @tset: taskset to iterate
295  *
296  * Iterate threadgroup leaders of @tset.  For single-task migrations, @tset
297  * may not contain any.
298  */
299 #define cgroup_taskset_for_each_leader(leader, dst_css, tset)		\
300 	for ((leader) = cgroup_taskset_first((tset), &(dst_css));	\
301 	     (leader);							\
302 	     (leader) = cgroup_taskset_next((tset), &(dst_css)))	\
303 		if ((leader) != (leader)->group_leader)			\
304 			;						\
305 		else
306 
307 /*
308  * Inline functions.
309  */
310 
311 /**
312  * css_get - obtain a reference on the specified css
313  * @css: target css
314  *
315  * The caller must already have a reference.
316  */
css_get(struct cgroup_subsys_state * css)317 static inline void css_get(struct cgroup_subsys_state *css)
318 {
319 	if (!(css->flags & CSS_NO_REF))
320 		percpu_ref_get(&css->refcnt);
321 }
322 
323 /**
324  * css_get_many - obtain references on the specified css
325  * @css: target css
326  * @n: number of references to get
327  *
328  * The caller must already have a reference.
329  */
css_get_many(struct cgroup_subsys_state * css,unsigned int n)330 static inline void css_get_many(struct cgroup_subsys_state *css, unsigned int n)
331 {
332 	if (!(css->flags & CSS_NO_REF))
333 		percpu_ref_get_many(&css->refcnt, n);
334 }
335 
336 /**
337  * css_tryget - try to obtain a reference on the specified css
338  * @css: target css
339  *
340  * Obtain a reference on @css unless it already has reached zero and is
341  * being released.  This function doesn't care whether @css is on or
342  * offline.  The caller naturally needs to ensure that @css is accessible
343  * but doesn't have to be holding a reference on it - IOW, RCU protected
344  * access is good enough for this function.  Returns %true if a reference
345  * count was successfully obtained; %false otherwise.
346  */
css_tryget(struct cgroup_subsys_state * css)347 static inline bool css_tryget(struct cgroup_subsys_state *css)
348 {
349 	if (!(css->flags & CSS_NO_REF))
350 		return percpu_ref_tryget(&css->refcnt);
351 	return true;
352 }
353 
354 /**
355  * css_tryget_online - try to obtain a reference on the specified css if online
356  * @css: target css
357  *
358  * Obtain a reference on @css if it's online.  The caller naturally needs
359  * to ensure that @css is accessible but doesn't have to be holding a
360  * reference on it - IOW, RCU protected access is good enough for this
361  * function.  Returns %true if a reference count was successfully obtained;
362  * %false otherwise.
363  */
css_tryget_online(struct cgroup_subsys_state * css)364 static inline bool css_tryget_online(struct cgroup_subsys_state *css)
365 {
366 	if (!(css->flags & CSS_NO_REF))
367 		return percpu_ref_tryget_live(&css->refcnt);
368 	return true;
369 }
370 
371 /**
372  * css_is_dying - test whether the specified css is dying
373  * @css: target css
374  *
375  * Test whether @css is in the process of offlining or already offline.  In
376  * most cases, ->css_online() and ->css_offline() callbacks should be
377  * enough; however, the actual offline operations are RCU delayed and this
378  * test returns %true also when @css is scheduled to be offlined.
379  *
380  * This is useful, for example, when the use case requires synchronous
381  * behavior with respect to cgroup removal.  cgroup removal schedules css
382  * offlining but the css can seem alive while the operation is being
383  * delayed.  If the delay affects user visible semantics, this test can be
384  * used to resolve the situation.
385  */
css_is_dying(struct cgroup_subsys_state * css)386 static inline bool css_is_dying(struct cgroup_subsys_state *css)
387 {
388 	return !(css->flags & CSS_NO_REF) && percpu_ref_is_dying(&css->refcnt);
389 }
390 
391 /**
392  * css_put - put a css reference
393  * @css: target css
394  *
395  * Put a reference obtained via css_get() and css_tryget_online().
396  */
css_put(struct cgroup_subsys_state * css)397 static inline void css_put(struct cgroup_subsys_state *css)
398 {
399 	if (!(css->flags & CSS_NO_REF))
400 		percpu_ref_put(&css->refcnt);
401 }
402 
403 /**
404  * css_put_many - put css references
405  * @css: target css
406  * @n: number of references to put
407  *
408  * Put references obtained via css_get() and css_tryget_online().
409  */
css_put_many(struct cgroup_subsys_state * css,unsigned int n)410 static inline void css_put_many(struct cgroup_subsys_state *css, unsigned int n)
411 {
412 	if (!(css->flags & CSS_NO_REF))
413 		percpu_ref_put_many(&css->refcnt, n);
414 }
415 
cgroup_get(struct cgroup * cgrp)416 static inline void cgroup_get(struct cgroup *cgrp)
417 {
418 	css_get(&cgrp->self);
419 }
420 
cgroup_tryget(struct cgroup * cgrp)421 static inline bool cgroup_tryget(struct cgroup *cgrp)
422 {
423 	return css_tryget(&cgrp->self);
424 }
425 
cgroup_put(struct cgroup * cgrp)426 static inline void cgroup_put(struct cgroup *cgrp)
427 {
428 	css_put(&cgrp->self);
429 }
430 
431 /**
432  * task_css_set_check - obtain a task's css_set with extra access conditions
433  * @task: the task to obtain css_set for
434  * @__c: extra condition expression to be passed to rcu_dereference_check()
435  *
436  * A task's css_set is RCU protected, initialized and exited while holding
437  * task_lock(), and can only be modified while holding both cgroup_mutex
438  * and task_lock() while the task is alive.  This macro verifies that the
439  * caller is inside proper critical section and returns @task's css_set.
440  *
441  * The caller can also specify additional allowed conditions via @__c, such
442  * as locks used during the cgroup_subsys::attach() methods.
443  */
444 #ifdef CONFIG_PROVE_RCU
445 extern struct mutex cgroup_mutex;
446 extern spinlock_t css_set_lock;
447 #define task_css_set_check(task, __c)					\
448 	rcu_dereference_check((task)->cgroups,				\
449 		lockdep_is_held(&cgroup_mutex) ||			\
450 		lockdep_is_held(&css_set_lock) ||			\
451 		((task)->flags & PF_EXITING) || (__c))
452 #else
453 #define task_css_set_check(task, __c)					\
454 	rcu_dereference((task)->cgroups)
455 #endif
456 
457 /**
458  * task_css_check - obtain css for (task, subsys) w/ extra access conds
459  * @task: the target task
460  * @subsys_id: the target subsystem ID
461  * @__c: extra condition expression to be passed to rcu_dereference_check()
462  *
463  * Return the cgroup_subsys_state for the (@task, @subsys_id) pair.  The
464  * synchronization rules are the same as task_css_set_check().
465  */
466 #define task_css_check(task, subsys_id, __c)				\
467 	task_css_set_check((task), (__c))->subsys[(subsys_id)]
468 
469 /**
470  * task_css_set - obtain a task's css_set
471  * @task: the task to obtain css_set for
472  *
473  * See task_css_set_check().
474  */
task_css_set(struct task_struct * task)475 static inline struct css_set *task_css_set(struct task_struct *task)
476 {
477 	return task_css_set_check(task, false);
478 }
479 
480 /**
481  * task_css - obtain css for (task, subsys)
482  * @task: the target task
483  * @subsys_id: the target subsystem ID
484  *
485  * See task_css_check().
486  */
task_css(struct task_struct * task,int subsys_id)487 static inline struct cgroup_subsys_state *task_css(struct task_struct *task,
488 						   int subsys_id)
489 {
490 	return task_css_check(task, subsys_id, false);
491 }
492 
493 /**
494  * task_get_css - find and get the css for (task, subsys)
495  * @task: the target task
496  * @subsys_id: the target subsystem ID
497  *
498  * Find the css for the (@task, @subsys_id) combination, increment a
499  * reference on and return it.  This function is guaranteed to return a
500  * valid css.  The returned css may already have been offlined.
501  */
502 static inline struct cgroup_subsys_state *
task_get_css(struct task_struct * task,int subsys_id)503 task_get_css(struct task_struct *task, int subsys_id)
504 {
505 	struct cgroup_subsys_state *css;
506 
507 	rcu_read_lock();
508 	while (true) {
509 		css = task_css(task, subsys_id);
510 		/*
511 		 * Can't use css_tryget_online() here.  A task which has
512 		 * PF_EXITING set may stay associated with an offline css.
513 		 * If such task calls this function, css_tryget_online()
514 		 * will keep failing.
515 		 */
516 		if (likely(css_tryget(css)))
517 			break;
518 		cpu_relax();
519 	}
520 	rcu_read_unlock();
521 	return css;
522 }
523 
524 /**
525  * task_css_is_root - test whether a task belongs to the root css
526  * @task: the target task
527  * @subsys_id: the target subsystem ID
528  *
529  * Test whether @task belongs to the root css on the specified subsystem.
530  * May be invoked in any context.
531  */
task_css_is_root(struct task_struct * task,int subsys_id)532 static inline bool task_css_is_root(struct task_struct *task, int subsys_id)
533 {
534 	return task_css_check(task, subsys_id, true) ==
535 		init_css_set.subsys[subsys_id];
536 }
537 
task_cgroup(struct task_struct * task,int subsys_id)538 static inline struct cgroup *task_cgroup(struct task_struct *task,
539 					 int subsys_id)
540 {
541 	return task_css(task, subsys_id)->cgroup;
542 }
543 
task_dfl_cgroup(struct task_struct * task)544 static inline struct cgroup *task_dfl_cgroup(struct task_struct *task)
545 {
546 	return task_css_set(task)->dfl_cgrp;
547 }
548 
cgroup_parent(struct cgroup * cgrp)549 static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
550 {
551 	struct cgroup_subsys_state *parent_css = cgrp->self.parent;
552 
553 	if (parent_css)
554 		return container_of(parent_css, struct cgroup, self);
555 	return NULL;
556 }
557 
558 /**
559  * cgroup_is_descendant - test ancestry
560  * @cgrp: the cgroup to be tested
561  * @ancestor: possible ancestor of @cgrp
562  *
563  * Test whether @cgrp is a descendant of @ancestor.  It also returns %true
564  * if @cgrp == @ancestor.  This function is safe to call as long as @cgrp
565  * and @ancestor are accessible.
566  */
cgroup_is_descendant(struct cgroup * cgrp,struct cgroup * ancestor)567 static inline bool cgroup_is_descendant(struct cgroup *cgrp,
568 					struct cgroup *ancestor)
569 {
570 	if (cgrp->root != ancestor->root || cgrp->level < ancestor->level)
571 		return false;
572 	return cgrp->ancestor_ids[ancestor->level] == ancestor->id;
573 }
574 
575 /**
576  * cgroup_ancestor - find ancestor of cgroup
577  * @cgrp: cgroup to find ancestor of
578  * @ancestor_level: level of ancestor to find starting from root
579  *
580  * Find ancestor of cgroup at specified level starting from root if it exists
581  * and return pointer to it. Return NULL if @cgrp doesn't have ancestor at
582  * @ancestor_level.
583  *
584  * This function is safe to call as long as @cgrp is accessible.
585  */
cgroup_ancestor(struct cgroup * cgrp,int ancestor_level)586 static inline struct cgroup *cgroup_ancestor(struct cgroup *cgrp,
587 					     int ancestor_level)
588 {
589 	if (cgrp->level < ancestor_level)
590 		return NULL;
591 	while (cgrp && cgrp->level > ancestor_level)
592 		cgrp = cgroup_parent(cgrp);
593 	return cgrp;
594 }
595 
596 /**
597  * task_under_cgroup_hierarchy - test task's membership of cgroup ancestry
598  * @task: the task to be tested
599  * @ancestor: possible ancestor of @task's cgroup
600  *
601  * Tests whether @task's default cgroup hierarchy is a descendant of @ancestor.
602  * It follows all the same rules as cgroup_is_descendant, and only applies
603  * to the default hierarchy.
604  */
task_under_cgroup_hierarchy(struct task_struct * task,struct cgroup * ancestor)605 static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
606 					       struct cgroup *ancestor)
607 {
608 	struct css_set *cset = task_css_set(task);
609 
610 	return cgroup_is_descendant(cset->dfl_cgrp, ancestor);
611 }
612 
613 /* no synchronization, the result can only be used as a hint */
cgroup_is_populated(struct cgroup * cgrp)614 static inline bool cgroup_is_populated(struct cgroup *cgrp)
615 {
616 	return cgrp->nr_populated_csets + cgrp->nr_populated_domain_children +
617 		cgrp->nr_populated_threaded_children;
618 }
619 
620 /* returns ino associated with a cgroup */
cgroup_ino(struct cgroup * cgrp)621 static inline ino_t cgroup_ino(struct cgroup *cgrp)
622 {
623 	return cgrp->kn->id.ino;
624 }
625 
626 /* cft/css accessors for cftype->write() operation */
of_cft(struct kernfs_open_file * of)627 static inline struct cftype *of_cft(struct kernfs_open_file *of)
628 {
629 	return of->kn->priv;
630 }
631 
632 struct cgroup_subsys_state *of_css(struct kernfs_open_file *of);
633 
634 /* cft/css accessors for cftype->seq_*() operations */
seq_cft(struct seq_file * seq)635 static inline struct cftype *seq_cft(struct seq_file *seq)
636 {
637 	return of_cft(seq->private);
638 }
639 
seq_css(struct seq_file * seq)640 static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq)
641 {
642 	return of_css(seq->private);
643 }
644 
645 /*
646  * Name / path handling functions.  All are thin wrappers around the kernfs
647  * counterparts and can be called under any context.
648  */
649 
cgroup_name(struct cgroup * cgrp,char * buf,size_t buflen)650 static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen)
651 {
652 	return kernfs_name(cgrp->kn, buf, buflen);
653 }
654 
cgroup_path(struct cgroup * cgrp,char * buf,size_t buflen)655 static inline int cgroup_path(struct cgroup *cgrp, char *buf, size_t buflen)
656 {
657 	return kernfs_path(cgrp->kn, buf, buflen);
658 }
659 
pr_cont_cgroup_name(struct cgroup * cgrp)660 static inline void pr_cont_cgroup_name(struct cgroup *cgrp)
661 {
662 	pr_cont_kernfs_name(cgrp->kn);
663 }
664 
pr_cont_cgroup_path(struct cgroup * cgrp)665 static inline void pr_cont_cgroup_path(struct cgroup *cgrp)
666 {
667 	pr_cont_kernfs_path(cgrp->kn);
668 }
669 
cgroup_psi(struct cgroup * cgrp)670 static inline struct psi_group *cgroup_psi(struct cgroup *cgrp)
671 {
672 	return &cgrp->psi;
673 }
674 
cgroup_init_kthreadd(void)675 static inline void cgroup_init_kthreadd(void)
676 {
677 	/*
678 	 * kthreadd is inherited by all kthreads, keep it in the root so
679 	 * that the new kthreads are guaranteed to stay in the root until
680 	 * initialization is finished.
681 	 */
682 	current->no_cgroup_migration = 1;
683 }
684 
cgroup_kthread_ready(void)685 static inline void cgroup_kthread_ready(void)
686 {
687 	/*
688 	 * This kthread finished initialization.  The creator should have
689 	 * set PF_NO_SETAFFINITY if this kthread should stay in the root.
690 	 */
691 	current->no_cgroup_migration = 0;
692 }
693 
cgroup_get_kernfs_id(struct cgroup * cgrp)694 static inline union kernfs_node_id *cgroup_get_kernfs_id(struct cgroup *cgrp)
695 {
696 	return &cgrp->kn->id;
697 }
698 
699 void cgroup_path_from_kernfs_id(const union kernfs_node_id *id,
700 					char *buf, size_t buflen);
701 #else /* !CONFIG_CGROUPS */
702 
703 struct cgroup_subsys_state;
704 struct cgroup;
705 
css_get(struct cgroup_subsys_state * css)706 static inline void css_get(struct cgroup_subsys_state *css) {}
css_put(struct cgroup_subsys_state * css)707 static inline void css_put(struct cgroup_subsys_state *css) {}
cgroup_attach_task_all(struct task_struct * from,struct task_struct * t)708 static inline int cgroup_attach_task_all(struct task_struct *from,
709 					 struct task_struct *t) { return 0; }
cgroupstats_build(struct cgroupstats * stats,struct dentry * dentry)710 static inline int cgroupstats_build(struct cgroupstats *stats,
711 				    struct dentry *dentry) { return -EINVAL; }
712 
cgroup_fork(struct task_struct * p)713 static inline void cgroup_fork(struct task_struct *p) {}
cgroup_can_fork(struct task_struct * p)714 static inline int cgroup_can_fork(struct task_struct *p) { return 0; }
cgroup_cancel_fork(struct task_struct * p)715 static inline void cgroup_cancel_fork(struct task_struct *p) {}
cgroup_post_fork(struct task_struct * p)716 static inline void cgroup_post_fork(struct task_struct *p) {}
cgroup_exit(struct task_struct * p)717 static inline void cgroup_exit(struct task_struct *p) {}
cgroup_release(struct task_struct * p)718 static inline void cgroup_release(struct task_struct *p) {}
cgroup_free(struct task_struct * p)719 static inline void cgroup_free(struct task_struct *p) {}
720 
cgroup_init_early(void)721 static inline int cgroup_init_early(void) { return 0; }
cgroup_init(void)722 static inline int cgroup_init(void) { return 0; }
cgroup_init_kthreadd(void)723 static inline void cgroup_init_kthreadd(void) {}
cgroup_kthread_ready(void)724 static inline void cgroup_kthread_ready(void) {}
cgroup_get_kernfs_id(struct cgroup * cgrp)725 static inline union kernfs_node_id *cgroup_get_kernfs_id(struct cgroup *cgrp)
726 {
727 	return NULL;
728 }
729 
cgroup_parent(struct cgroup * cgrp)730 static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
731 {
732 	return NULL;
733 }
734 
cgroup_psi(struct cgroup * cgrp)735 static inline struct psi_group *cgroup_psi(struct cgroup *cgrp)
736 {
737 	return NULL;
738 }
739 
task_under_cgroup_hierarchy(struct task_struct * task,struct cgroup * ancestor)740 static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
741 					       struct cgroup *ancestor)
742 {
743 	return true;
744 }
745 
cgroup_path_from_kernfs_id(const union kernfs_node_id * id,char * buf,size_t buflen)746 static inline void cgroup_path_from_kernfs_id(const union kernfs_node_id *id,
747 	char *buf, size_t buflen) {}
748 #endif /* !CONFIG_CGROUPS */
749 
750 #ifdef CONFIG_CGROUPS
751 /*
752  * cgroup scalable recursive statistics.
753  */
754 void cgroup_rstat_updated(struct cgroup *cgrp, int cpu);
755 void cgroup_rstat_flush(struct cgroup *cgrp);
756 void cgroup_rstat_flush_irqsafe(struct cgroup *cgrp);
757 void cgroup_rstat_flush_hold(struct cgroup *cgrp);
758 void cgroup_rstat_flush_release(void);
759 
760 /*
761  * Basic resource stats.
762  */
763 #ifdef CONFIG_CGROUP_CPUACCT
764 void cpuacct_charge(struct task_struct *tsk, u64 cputime);
765 void cpuacct_account_field(struct task_struct *tsk, int index, u64 val);
766 #else
cpuacct_charge(struct task_struct * tsk,u64 cputime)767 static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {}
cpuacct_account_field(struct task_struct * tsk,int index,u64 val)768 static inline void cpuacct_account_field(struct task_struct *tsk, int index,
769 					 u64 val) {}
770 #endif
771 
772 void __cgroup_account_cputime(struct cgroup *cgrp, u64 delta_exec);
773 void __cgroup_account_cputime_field(struct cgroup *cgrp,
774 				    enum cpu_usage_stat index, u64 delta_exec);
775 
cgroup_account_cputime(struct task_struct * task,u64 delta_exec)776 static inline void cgroup_account_cputime(struct task_struct *task,
777 					  u64 delta_exec)
778 {
779 	struct cgroup *cgrp;
780 
781 	cpuacct_charge(task, delta_exec);
782 
783 	rcu_read_lock();
784 	cgrp = task_dfl_cgroup(task);
785 	if (cgroup_parent(cgrp))
786 		__cgroup_account_cputime(cgrp, delta_exec);
787 	rcu_read_unlock();
788 }
789 
cgroup_account_cputime_field(struct task_struct * task,enum cpu_usage_stat index,u64 delta_exec)790 static inline void cgroup_account_cputime_field(struct task_struct *task,
791 						enum cpu_usage_stat index,
792 						u64 delta_exec)
793 {
794 	struct cgroup *cgrp;
795 
796 	cpuacct_account_field(task, index, delta_exec);
797 
798 	rcu_read_lock();
799 	cgrp = task_dfl_cgroup(task);
800 	if (cgroup_parent(cgrp))
801 		__cgroup_account_cputime_field(cgrp, index, delta_exec);
802 	rcu_read_unlock();
803 }
804 
805 #else	/* CONFIG_CGROUPS */
806 
cgroup_account_cputime(struct task_struct * task,u64 delta_exec)807 static inline void cgroup_account_cputime(struct task_struct *task,
808 					  u64 delta_exec) {}
cgroup_account_cputime_field(struct task_struct * task,enum cpu_usage_stat index,u64 delta_exec)809 static inline void cgroup_account_cputime_field(struct task_struct *task,
810 						enum cpu_usage_stat index,
811 						u64 delta_exec) {}
812 
813 #endif	/* CONFIG_CGROUPS */
814 
815 /*
816  * sock->sk_cgrp_data handling.  For more info, see sock_cgroup_data
817  * definition in cgroup-defs.h.
818  */
819 #ifdef CONFIG_SOCK_CGROUP_DATA
820 
821 #if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
822 extern spinlock_t cgroup_sk_update_lock;
823 #endif
824 
825 void cgroup_sk_alloc_disable(void);
826 void cgroup_sk_alloc(struct sock_cgroup_data *skcd);
827 void cgroup_sk_clone(struct sock_cgroup_data *skcd);
828 void cgroup_sk_free(struct sock_cgroup_data *skcd);
829 
sock_cgroup_ptr(struct sock_cgroup_data * skcd)830 static inline struct cgroup *sock_cgroup_ptr(struct sock_cgroup_data *skcd)
831 {
832 #if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
833 	unsigned long v;
834 
835 	/*
836 	 * @skcd->val is 64bit but the following is safe on 32bit too as we
837 	 * just need the lower ulong to be written and read atomically.
838 	 */
839 	v = READ_ONCE(skcd->val);
840 
841 	if (v & 3)
842 		return &cgrp_dfl_root.cgrp;
843 
844 	return (struct cgroup *)(unsigned long)v ?: &cgrp_dfl_root.cgrp;
845 #else
846 	return (struct cgroup *)(unsigned long)skcd->val;
847 #endif
848 }
849 
850 #else	/* CONFIG_CGROUP_DATA */
851 
cgroup_sk_alloc(struct sock_cgroup_data * skcd)852 static inline void cgroup_sk_alloc(struct sock_cgroup_data *skcd) {}
cgroup_sk_clone(struct sock_cgroup_data * skcd)853 static inline void cgroup_sk_clone(struct sock_cgroup_data *skcd) {}
cgroup_sk_free(struct sock_cgroup_data * skcd)854 static inline void cgroup_sk_free(struct sock_cgroup_data *skcd) {}
855 
856 #endif	/* CONFIG_CGROUP_DATA */
857 
858 struct cgroup_namespace {
859 	refcount_t		count;
860 	struct ns_common	ns;
861 	struct user_namespace	*user_ns;
862 	struct ucounts		*ucounts;
863 	struct css_set          *root_cset;
864 };
865 
866 extern struct cgroup_namespace init_cgroup_ns;
867 
868 #ifdef CONFIG_CGROUPS
869 
870 void free_cgroup_ns(struct cgroup_namespace *ns);
871 
872 struct cgroup_namespace *copy_cgroup_ns(unsigned long flags,
873 					struct user_namespace *user_ns,
874 					struct cgroup_namespace *old_ns);
875 
876 int cgroup_path_ns(struct cgroup *cgrp, char *buf, size_t buflen,
877 		   struct cgroup_namespace *ns);
878 
879 #else /* !CONFIG_CGROUPS */
880 
free_cgroup_ns(struct cgroup_namespace * ns)881 static inline void free_cgroup_ns(struct cgroup_namespace *ns) { }
882 static inline struct cgroup_namespace *
copy_cgroup_ns(unsigned long flags,struct user_namespace * user_ns,struct cgroup_namespace * old_ns)883 copy_cgroup_ns(unsigned long flags, struct user_namespace *user_ns,
884 	       struct cgroup_namespace *old_ns)
885 {
886 	return old_ns;
887 }
888 
889 #endif /* !CONFIG_CGROUPS */
890 
get_cgroup_ns(struct cgroup_namespace * ns)891 static inline void get_cgroup_ns(struct cgroup_namespace *ns)
892 {
893 	if (ns)
894 		refcount_inc(&ns->count);
895 }
896 
put_cgroup_ns(struct cgroup_namespace * ns)897 static inline void put_cgroup_ns(struct cgroup_namespace *ns)
898 {
899 	if (ns && refcount_dec_and_test(&ns->count))
900 		free_cgroup_ns(ns);
901 }
902 
903 #ifdef CONFIG_CGROUPS
904 
905 void cgroup_enter_frozen(void);
906 void cgroup_leave_frozen(bool always_leave);
907 void cgroup_update_frozen(struct cgroup *cgrp);
908 void cgroup_freeze(struct cgroup *cgrp, bool freeze);
909 void cgroup_freezer_migrate_task(struct task_struct *task, struct cgroup *src,
910 				 struct cgroup *dst);
911 
cgroup_task_freeze(struct task_struct * task)912 static inline bool cgroup_task_freeze(struct task_struct *task)
913 {
914 	bool ret;
915 
916 	if (task->flags & PF_KTHREAD)
917 		return false;
918 
919 	rcu_read_lock();
920 	ret = test_bit(CGRP_FREEZE, &task_dfl_cgroup(task)->flags);
921 	rcu_read_unlock();
922 
923 	return ret;
924 }
925 
cgroup_task_frozen(struct task_struct * task)926 static inline bool cgroup_task_frozen(struct task_struct *task)
927 {
928 	return task->frozen;
929 }
930 
931 #else /* !CONFIG_CGROUPS */
932 
cgroup_enter_frozen(void)933 static inline void cgroup_enter_frozen(void) { }
cgroup_leave_frozen(bool always_leave)934 static inline void cgroup_leave_frozen(bool always_leave) { }
cgroup_task_freeze(struct task_struct * task)935 static inline bool cgroup_task_freeze(struct task_struct *task)
936 {
937 	return false;
938 }
cgroup_task_frozen(struct task_struct * task)939 static inline bool cgroup_task_frozen(struct task_struct *task)
940 {
941 	return false;
942 }
943 
944 #endif /* !CONFIG_CGROUPS */
945 
946 #ifdef CONFIG_CGROUP_BPF
cgroup_bpf_get(struct cgroup * cgrp)947 static inline void cgroup_bpf_get(struct cgroup *cgrp)
948 {
949 	percpu_ref_get(&cgrp->bpf.refcnt);
950 }
951 
cgroup_bpf_put(struct cgroup * cgrp)952 static inline void cgroup_bpf_put(struct cgroup *cgrp)
953 {
954 	percpu_ref_put(&cgrp->bpf.refcnt);
955 }
956 
957 #else /* CONFIG_CGROUP_BPF */
958 
cgroup_bpf_get(struct cgroup * cgrp)959 static inline void cgroup_bpf_get(struct cgroup *cgrp) {}
cgroup_bpf_put(struct cgroup * cgrp)960 static inline void cgroup_bpf_put(struct cgroup *cgrp) {}
961 
962 #endif /* CONFIG_CGROUP_BPF */
963 
964 #endif /* _LINUX_CGROUP_H */
965