1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_CGROUP_H
3 #define _LINUX_CGROUP_H
4 /*
5  *  cgroup interface
6  *
7  *  Copyright (C) 2003 BULL SA
8  *  Copyright (C) 2004-2006 Silicon Graphics, Inc.
9  *
10  */
11 
12 #include <linux/sched.h>
13 #include <linux/cpumask.h>
14 #include <linux/nodemask.h>
15 #include <linux/rculist.h>
16 #include <linux/cgroupstats.h>
17 #include <linux/fs.h>
18 #include <linux/seq_file.h>
19 #include <linux/kernfs.h>
20 #include <linux/jump_label.h>
21 #include <linux/types.h>
22 #include <linux/ns_common.h>
23 #include <linux/nsproxy.h>
24 #include <linux/user_namespace.h>
25 #include <linux/refcount.h>
26 #include <linux/kernel_stat.h>
27 
28 #include <linux/cgroup-defs.h>
29 
30 struct kernel_clone_args;
31 
32 #ifdef CONFIG_CGROUPS
33 
34 /*
35  * All weight knobs on the default hierarchy should use the following min,
36  * default and max values.  The default value is the logarithmic center of
37  * MIN and MAX and allows 100x to be expressed in both directions.
38  */
39 #define CGROUP_WEIGHT_MIN		1
40 #define CGROUP_WEIGHT_DFL		100
41 #define CGROUP_WEIGHT_MAX		10000
42 
43 /* walk only threadgroup leaders */
44 #define CSS_TASK_ITER_PROCS		(1U << 0)
45 /* walk all threaded css_sets in the domain */
46 #define CSS_TASK_ITER_THREADED		(1U << 1)
47 
48 /* internal flags */
49 #define CSS_TASK_ITER_SKIPPED		(1U << 16)
50 
51 /* a css_task_iter should be treated as an opaque object */
52 struct css_task_iter {
53 	struct cgroup_subsys		*ss;
54 	unsigned int			flags;
55 
56 	struct list_head		*cset_pos;
57 	struct list_head		*cset_head;
58 
59 	struct list_head		*tcset_pos;
60 	struct list_head		*tcset_head;
61 
62 	struct list_head		*task_pos;
63 
64 	struct list_head		*cur_tasks_head;
65 	struct css_set			*cur_cset;
66 	struct css_set			*cur_dcset;
67 	struct task_struct		*cur_task;
68 	struct list_head		iters_node;	/* css_set->task_iters */
69 };
70 
71 extern struct file_system_type cgroup_fs_type;
72 extern struct cgroup_root cgrp_dfl_root;
73 extern struct css_set init_css_set;
74 
75 #define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys;
76 #include <linux/cgroup_subsys.h>
77 #undef SUBSYS
78 
79 #define SUBSYS(_x)								\
80 	extern struct static_key_true _x ## _cgrp_subsys_enabled_key;		\
81 	extern struct static_key_true _x ## _cgrp_subsys_on_dfl_key;
82 #include <linux/cgroup_subsys.h>
83 #undef SUBSYS
84 
85 /**
86  * cgroup_subsys_enabled - fast test on whether a subsys is enabled
87  * @ss: subsystem in question
88  */
89 #define cgroup_subsys_enabled(ss)						\
90 	static_branch_likely(&ss ## _enabled_key)
91 
92 /**
93  * cgroup_subsys_on_dfl - fast test on whether a subsys is on default hierarchy
94  * @ss: subsystem in question
95  */
96 #define cgroup_subsys_on_dfl(ss)						\
97 	static_branch_likely(&ss ## _on_dfl_key)
98 
99 bool css_has_online_children(struct cgroup_subsys_state *css);
100 struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss);
101 struct cgroup_subsys_state *cgroup_e_css(struct cgroup *cgroup,
102 					 struct cgroup_subsys *ss);
103 struct cgroup_subsys_state *cgroup_get_e_css(struct cgroup *cgroup,
104 					     struct cgroup_subsys *ss);
105 struct cgroup_subsys_state *css_tryget_online_from_dir(struct dentry *dentry,
106 						       struct cgroup_subsys *ss);
107 
108 struct cgroup *cgroup_get_from_path(const char *path);
109 struct cgroup *cgroup_get_from_fd(int fd);
110 struct cgroup *cgroup_v1v2_get_from_fd(int fd);
111 
112 int cgroup_attach_task_all(struct task_struct *from, struct task_struct *);
113 int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from);
114 
115 int cgroup_add_dfl_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
116 int cgroup_add_legacy_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
117 int cgroup_rm_cftypes(struct cftype *cfts);
118 void cgroup_file_notify(struct cgroup_file *cfile);
119 void cgroup_file_show(struct cgroup_file *cfile, bool show);
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 			   struct kernel_clone_args *kargs);
129 extern void cgroup_cancel_fork(struct task_struct *p,
130 			       struct kernel_clone_args *kargs);
131 extern void cgroup_post_fork(struct task_struct *p,
132 			     struct kernel_clone_args *kargs);
133 void cgroup_exit(struct task_struct *p);
134 void cgroup_release(struct task_struct *p);
135 void cgroup_free(struct task_struct *p);
136 
137 int cgroup_init_early(void);
138 int cgroup_init(void);
139 
140 int cgroup_parse_float(const char *input, unsigned dec_shift, s64 *v);
141 
142 /*
143  * Iteration helpers and macros.
144  */
145 
146 struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos,
147 					   struct cgroup_subsys_state *parent);
148 struct cgroup_subsys_state *css_next_descendant_pre(struct cgroup_subsys_state *pos,
149 						    struct cgroup_subsys_state *css);
150 struct cgroup_subsys_state *css_rightmost_descendant(struct cgroup_subsys_state *pos);
151 struct cgroup_subsys_state *css_next_descendant_post(struct cgroup_subsys_state *pos,
152 						     struct cgroup_subsys_state *css);
153 
154 struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset,
155 					 struct cgroup_subsys_state **dst_cssp);
156 struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset,
157 					struct cgroup_subsys_state **dst_cssp);
158 
159 void css_task_iter_start(struct cgroup_subsys_state *css, unsigned int flags,
160 			 struct css_task_iter *it);
161 struct task_struct *css_task_iter_next(struct css_task_iter *it);
162 void css_task_iter_end(struct css_task_iter *it);
163 
164 /**
165  * css_for_each_child - iterate through children of a css
166  * @pos: the css * to use as the loop cursor
167  * @parent: css whose children to walk
168  *
169  * Walk @parent's children.  Must be called under rcu_read_lock().
170  *
171  * If a subsystem synchronizes ->css_online() and the start of iteration, a
172  * css which finished ->css_online() is guaranteed to be visible in the
173  * future iterations and will stay visible until the last reference is put.
174  * A css which hasn't finished ->css_online() or already finished
175  * ->css_offline() may show up during traversal.  It's each subsystem's
176  * responsibility to synchronize against on/offlining.
177  *
178  * It is allowed to temporarily drop RCU read lock during iteration.  The
179  * caller is responsible for ensuring that @pos remains accessible until
180  * the start of the next iteration by, for example, bumping the css refcnt.
181  */
182 #define css_for_each_child(pos, parent)					\
183 	for ((pos) = css_next_child(NULL, (parent)); (pos);		\
184 	     (pos) = css_next_child((pos), (parent)))
185 
186 /**
187  * css_for_each_descendant_pre - pre-order walk of a css's descendants
188  * @pos: the css * to use as the loop cursor
189  * @root: css whose descendants to walk
190  *
191  * Walk @root's descendants.  @root is included in the iteration and the
192  * first node to be visited.  Must be called under rcu_read_lock().
193  *
194  * If a subsystem synchronizes ->css_online() and the start of iteration, a
195  * css which finished ->css_online() is guaranteed to be visible in the
196  * future iterations and will stay visible until the last reference is put.
197  * A css which hasn't finished ->css_online() or already finished
198  * ->css_offline() may show up during traversal.  It's each subsystem's
199  * responsibility to synchronize against on/offlining.
200  *
201  * For example, the following guarantees that a descendant can't escape
202  * state updates of its ancestors.
203  *
204  * my_online(@css)
205  * {
206  *	Lock @css's parent and @css;
207  *	Inherit state from the parent;
208  *	Unlock both.
209  * }
210  *
211  * my_update_state(@css)
212  * {
213  *	css_for_each_descendant_pre(@pos, @css) {
214  *		Lock @pos;
215  *		if (@pos == @css)
216  *			Update @css's state;
217  *		else
218  *			Verify @pos is alive and inherit state from its parent;
219  *		Unlock @pos;
220  *	}
221  * }
222  *
223  * As long as the inheriting step, including checking the parent state, is
224  * enclosed inside @pos locking, double-locking the parent isn't necessary
225  * while inheriting.  The state update to the parent is guaranteed to be
226  * visible by walking order and, as long as inheriting operations to the
227  * same @pos are atomic to each other, multiple updates racing each other
228  * still result in the correct state.  It's guaranateed that at least one
229  * inheritance happens for any css after the latest update to its parent.
230  *
231  * If checking parent's state requires locking the parent, each inheriting
232  * iteration should lock and unlock both @pos->parent and @pos.
233  *
234  * Alternatively, a subsystem may choose to use a single global lock to
235  * synchronize ->css_online() and ->css_offline() against tree-walking
236  * operations.
237  *
238  * It is allowed to temporarily drop RCU read lock during iteration.  The
239  * caller is responsible for ensuring that @pos remains accessible until
240  * the start of the next iteration by, for example, bumping the css refcnt.
241  */
242 #define css_for_each_descendant_pre(pos, css)				\
243 	for ((pos) = css_next_descendant_pre(NULL, (css)); (pos);	\
244 	     (pos) = css_next_descendant_pre((pos), (css)))
245 
246 /**
247  * css_for_each_descendant_post - post-order walk of a css's descendants
248  * @pos: the css * to use as the loop cursor
249  * @css: css whose descendants to walk
250  *
251  * Similar to css_for_each_descendant_pre() but performs post-order
252  * traversal instead.  @root is included in the iteration and the last
253  * node to be visited.
254  *
255  * If a subsystem synchronizes ->css_online() and the start of iteration, a
256  * css which finished ->css_online() is guaranteed to be visible in the
257  * future iterations and will stay visible until the last reference is put.
258  * A css which hasn't finished ->css_online() or already finished
259  * ->css_offline() may show up during traversal.  It's each subsystem's
260  * responsibility to synchronize against on/offlining.
261  *
262  * Note that the walk visibility guarantee example described in pre-order
263  * walk doesn't apply the same to post-order walks.
264  */
265 #define css_for_each_descendant_post(pos, css)				\
266 	for ((pos) = css_next_descendant_post(NULL, (css)); (pos);	\
267 	     (pos) = css_next_descendant_post((pos), (css)))
268 
269 /**
270  * cgroup_taskset_for_each - iterate cgroup_taskset
271  * @task: the loop cursor
272  * @dst_css: the destination css
273  * @tset: taskset to iterate
274  *
275  * @tset may contain multiple tasks and they may belong to multiple
276  * processes.
277  *
278  * On the v2 hierarchy, there may be tasks from multiple processes and they
279  * may not share the source or destination csses.
280  *
281  * On traditional hierarchies, when there are multiple tasks in @tset, if a
282  * task of a process is in @tset, all tasks of the process are in @tset.
283  * Also, all are guaranteed to share the same source and destination csses.
284  *
285  * Iteration is not in any specific order.
286  */
287 #define cgroup_taskset_for_each(task, dst_css, tset)			\
288 	for ((task) = cgroup_taskset_first((tset), &(dst_css));		\
289 	     (task);							\
290 	     (task) = cgroup_taskset_next((tset), &(dst_css)))
291 
292 /**
293  * cgroup_taskset_for_each_leader - iterate group leaders in a cgroup_taskset
294  * @leader: the loop cursor
295  * @dst_css: the destination css
296  * @tset: taskset to iterate
297  *
298  * Iterate threadgroup leaders of @tset.  For single-task migrations, @tset
299  * may not contain any.
300  */
301 #define cgroup_taskset_for_each_leader(leader, dst_css, tset)		\
302 	for ((leader) = cgroup_taskset_first((tset), &(dst_css));	\
303 	     (leader);							\
304 	     (leader) = cgroup_taskset_next((tset), &(dst_css)))	\
305 		if ((leader) != (leader)->group_leader)			\
306 			;						\
307 		else
308 
309 /*
310  * Inline functions.
311  */
312 
cgroup_id(const struct cgroup * cgrp)313 static inline u64 cgroup_id(const struct cgroup *cgrp)
314 {
315 	return cgrp->kn->id;
316 }
317 
318 /**
319  * css_get - obtain a reference on the specified css
320  * @css: target css
321  *
322  * The caller must already have a reference.
323  */
css_get(struct cgroup_subsys_state * css)324 static inline void css_get(struct cgroup_subsys_state *css)
325 {
326 	if (!(css->flags & CSS_NO_REF))
327 		percpu_ref_get(&css->refcnt);
328 }
329 
330 /**
331  * css_get_many - obtain references on the specified css
332  * @css: target css
333  * @n: number of references to get
334  *
335  * The caller must already have a reference.
336  */
css_get_many(struct cgroup_subsys_state * css,unsigned int n)337 static inline void css_get_many(struct cgroup_subsys_state *css, unsigned int n)
338 {
339 	if (!(css->flags & CSS_NO_REF))
340 		percpu_ref_get_many(&css->refcnt, n);
341 }
342 
343 /**
344  * css_tryget - try to obtain a reference on the specified css
345  * @css: target css
346  *
347  * Obtain a reference on @css unless it already has reached zero and is
348  * being released.  This function doesn't care whether @css is on or
349  * offline.  The caller naturally needs to ensure that @css is accessible
350  * but doesn't have to be holding a reference on it - IOW, RCU protected
351  * access is good enough for this function.  Returns %true if a reference
352  * count was successfully obtained; %false otherwise.
353  */
css_tryget(struct cgroup_subsys_state * css)354 static inline bool css_tryget(struct cgroup_subsys_state *css)
355 {
356 	if (!(css->flags & CSS_NO_REF))
357 		return percpu_ref_tryget(&css->refcnt);
358 	return true;
359 }
360 
361 /**
362  * css_tryget_online - try to obtain a reference on the specified css if online
363  * @css: target css
364  *
365  * Obtain a reference on @css if it's online.  The caller naturally needs
366  * to ensure that @css is accessible but doesn't have to be holding a
367  * reference on it - IOW, RCU protected access is good enough for this
368  * function.  Returns %true if a reference count was successfully obtained;
369  * %false otherwise.
370  */
css_tryget_online(struct cgroup_subsys_state * css)371 static inline bool css_tryget_online(struct cgroup_subsys_state *css)
372 {
373 	if (!(css->flags & CSS_NO_REF))
374 		return percpu_ref_tryget_live(&css->refcnt);
375 	return true;
376 }
377 
378 /**
379  * css_is_dying - test whether the specified css is dying
380  * @css: target css
381  *
382  * Test whether @css is in the process of offlining or already offline.  In
383  * most cases, ->css_online() and ->css_offline() callbacks should be
384  * enough; however, the actual offline operations are RCU delayed and this
385  * test returns %true also when @css is scheduled to be offlined.
386  *
387  * This is useful, for example, when the use case requires synchronous
388  * behavior with respect to cgroup removal.  cgroup removal schedules css
389  * offlining but the css can seem alive while the operation is being
390  * delayed.  If the delay affects user visible semantics, this test can be
391  * used to resolve the situation.
392  */
css_is_dying(struct cgroup_subsys_state * css)393 static inline bool css_is_dying(struct cgroup_subsys_state *css)
394 {
395 	return !(css->flags & CSS_NO_REF) && percpu_ref_is_dying(&css->refcnt);
396 }
397 
398 /**
399  * css_put - put a css reference
400  * @css: target css
401  *
402  * Put a reference obtained via css_get() and css_tryget_online().
403  */
css_put(struct cgroup_subsys_state * css)404 static inline void css_put(struct cgroup_subsys_state *css)
405 {
406 	if (!(css->flags & CSS_NO_REF))
407 		percpu_ref_put(&css->refcnt);
408 }
409 
410 /**
411  * css_put_many - put css references
412  * @css: target css
413  * @n: number of references to put
414  *
415  * Put references obtained via css_get() and css_tryget_online().
416  */
css_put_many(struct cgroup_subsys_state * css,unsigned int n)417 static inline void css_put_many(struct cgroup_subsys_state *css, unsigned int n)
418 {
419 	if (!(css->flags & CSS_NO_REF))
420 		percpu_ref_put_many(&css->refcnt, n);
421 }
422 
cgroup_get(struct cgroup * cgrp)423 static inline void cgroup_get(struct cgroup *cgrp)
424 {
425 	css_get(&cgrp->self);
426 }
427 
cgroup_tryget(struct cgroup * cgrp)428 static inline bool cgroup_tryget(struct cgroup *cgrp)
429 {
430 	return css_tryget(&cgrp->self);
431 }
432 
cgroup_put(struct cgroup * cgrp)433 static inline void cgroup_put(struct cgroup *cgrp)
434 {
435 	css_put(&cgrp->self);
436 }
437 
438 extern struct mutex cgroup_mutex;
439 
cgroup_lock(void)440 static inline void cgroup_lock(void)
441 {
442 	mutex_lock(&cgroup_mutex);
443 }
444 
cgroup_unlock(void)445 static inline void cgroup_unlock(void)
446 {
447 	mutex_unlock(&cgroup_mutex);
448 }
449 
450 /**
451  * task_css_set_check - obtain a task's css_set with extra access conditions
452  * @task: the task to obtain css_set for
453  * @__c: extra condition expression to be passed to rcu_dereference_check()
454  *
455  * A task's css_set is RCU protected, initialized and exited while holding
456  * task_lock(), and can only be modified while holding both cgroup_mutex
457  * and task_lock() while the task is alive.  This macro verifies that the
458  * caller is inside proper critical section and returns @task's css_set.
459  *
460  * The caller can also specify additional allowed conditions via @__c, such
461  * as locks used during the cgroup_subsys::attach() methods.
462  */
463 #ifdef CONFIG_PROVE_RCU
464 extern spinlock_t css_set_lock;
465 #define task_css_set_check(task, __c)					\
466 	rcu_dereference_check((task)->cgroups,				\
467 		rcu_read_lock_sched_held() ||				\
468 		lockdep_is_held(&cgroup_mutex) ||			\
469 		lockdep_is_held(&css_set_lock) ||			\
470 		((task)->flags & PF_EXITING) || (__c))
471 #else
472 #define task_css_set_check(task, __c)					\
473 	rcu_dereference((task)->cgroups)
474 #endif
475 
476 /**
477  * task_css_check - obtain css for (task, subsys) w/ extra access conds
478  * @task: the target task
479  * @subsys_id: the target subsystem ID
480  * @__c: extra condition expression to be passed to rcu_dereference_check()
481  *
482  * Return the cgroup_subsys_state for the (@task, @subsys_id) pair.  The
483  * synchronization rules are the same as task_css_set_check().
484  */
485 #define task_css_check(task, subsys_id, __c)				\
486 	task_css_set_check((task), (__c))->subsys[(subsys_id)]
487 
488 /**
489  * task_css_set - obtain a task's css_set
490  * @task: the task to obtain css_set for
491  *
492  * See task_css_set_check().
493  */
task_css_set(struct task_struct * task)494 static inline struct css_set *task_css_set(struct task_struct *task)
495 {
496 	return task_css_set_check(task, false);
497 }
498 
499 /**
500  * task_css - obtain css for (task, subsys)
501  * @task: the target task
502  * @subsys_id: the target subsystem ID
503  *
504  * See task_css_check().
505  */
task_css(struct task_struct * task,int subsys_id)506 static inline struct cgroup_subsys_state *task_css(struct task_struct *task,
507 						   int subsys_id)
508 {
509 	return task_css_check(task, subsys_id, false);
510 }
511 
512 /**
513  * task_get_css - find and get the css for (task, subsys)
514  * @task: the target task
515  * @subsys_id: the target subsystem ID
516  *
517  * Find the css for the (@task, @subsys_id) combination, increment a
518  * reference on and return it.  This function is guaranteed to return a
519  * valid css.  The returned css may already have been offlined.
520  */
521 static inline struct cgroup_subsys_state *
task_get_css(struct task_struct * task,int subsys_id)522 task_get_css(struct task_struct *task, int subsys_id)
523 {
524 	struct cgroup_subsys_state *css;
525 
526 	rcu_read_lock();
527 	while (true) {
528 		css = task_css(task, subsys_id);
529 		/*
530 		 * Can't use css_tryget_online() here.  A task which has
531 		 * PF_EXITING set may stay associated with an offline css.
532 		 * If such task calls this function, css_tryget_online()
533 		 * will keep failing.
534 		 */
535 		if (likely(css_tryget(css)))
536 			break;
537 		cpu_relax();
538 	}
539 	rcu_read_unlock();
540 	return css;
541 }
542 
543 /**
544  * task_css_is_root - test whether a task belongs to the root css
545  * @task: the target task
546  * @subsys_id: the target subsystem ID
547  *
548  * Test whether @task belongs to the root css on the specified subsystem.
549  * May be invoked in any context.
550  */
task_css_is_root(struct task_struct * task,int subsys_id)551 static inline bool task_css_is_root(struct task_struct *task, int subsys_id)
552 {
553 	return task_css_check(task, subsys_id, true) ==
554 		init_css_set.subsys[subsys_id];
555 }
556 
task_cgroup(struct task_struct * task,int subsys_id)557 static inline struct cgroup *task_cgroup(struct task_struct *task,
558 					 int subsys_id)
559 {
560 	return task_css(task, subsys_id)->cgroup;
561 }
562 
task_dfl_cgroup(struct task_struct * task)563 static inline struct cgroup *task_dfl_cgroup(struct task_struct *task)
564 {
565 	return task_css_set(task)->dfl_cgrp;
566 }
567 
cgroup_parent(struct cgroup * cgrp)568 static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
569 {
570 	struct cgroup_subsys_state *parent_css = cgrp->self.parent;
571 
572 	if (parent_css)
573 		return container_of(parent_css, struct cgroup, self);
574 	return NULL;
575 }
576 
577 /**
578  * cgroup_is_descendant - test ancestry
579  * @cgrp: the cgroup to be tested
580  * @ancestor: possible ancestor of @cgrp
581  *
582  * Test whether @cgrp is a descendant of @ancestor.  It also returns %true
583  * if @cgrp == @ancestor.  This function is safe to call as long as @cgrp
584  * and @ancestor are accessible.
585  */
cgroup_is_descendant(struct cgroup * cgrp,struct cgroup * ancestor)586 static inline bool cgroup_is_descendant(struct cgroup *cgrp,
587 					struct cgroup *ancestor)
588 {
589 	if (cgrp->root != ancestor->root || cgrp->level < ancestor->level)
590 		return false;
591 	return cgrp->ancestors[ancestor->level] == ancestor;
592 }
593 
594 /**
595  * cgroup_ancestor - find ancestor of cgroup
596  * @cgrp: cgroup to find ancestor of
597  * @ancestor_level: level of ancestor to find starting from root
598  *
599  * Find ancestor of cgroup at specified level starting from root if it exists
600  * and return pointer to it. Return NULL if @cgrp doesn't have ancestor at
601  * @ancestor_level.
602  *
603  * This function is safe to call as long as @cgrp is accessible.
604  */
cgroup_ancestor(struct cgroup * cgrp,int ancestor_level)605 static inline struct cgroup *cgroup_ancestor(struct cgroup *cgrp,
606 					     int ancestor_level)
607 {
608 	if (ancestor_level < 0 || ancestor_level > cgrp->level)
609 		return NULL;
610 	return cgrp->ancestors[ancestor_level];
611 }
612 
613 /**
614  * task_under_cgroup_hierarchy - test task's membership of cgroup ancestry
615  * @task: the task to be tested
616  * @ancestor: possible ancestor of @task's cgroup
617  *
618  * Tests whether @task's default cgroup hierarchy is a descendant of @ancestor.
619  * It follows all the same rules as cgroup_is_descendant, and only applies
620  * to the default hierarchy.
621  */
task_under_cgroup_hierarchy(struct task_struct * task,struct cgroup * ancestor)622 static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
623 					       struct cgroup *ancestor)
624 {
625 	struct css_set *cset = task_css_set(task);
626 
627 	return cgroup_is_descendant(cset->dfl_cgrp, ancestor);
628 }
629 
630 /* no synchronization, the result can only be used as a hint */
cgroup_is_populated(struct cgroup * cgrp)631 static inline bool cgroup_is_populated(struct cgroup *cgrp)
632 {
633 	return cgrp->nr_populated_csets + cgrp->nr_populated_domain_children +
634 		cgrp->nr_populated_threaded_children;
635 }
636 
637 /* returns ino associated with a cgroup */
cgroup_ino(struct cgroup * cgrp)638 static inline ino_t cgroup_ino(struct cgroup *cgrp)
639 {
640 	return kernfs_ino(cgrp->kn);
641 }
642 
643 /* cft/css accessors for cftype->write() operation */
of_cft(struct kernfs_open_file * of)644 static inline struct cftype *of_cft(struct kernfs_open_file *of)
645 {
646 	return of->kn->priv;
647 }
648 
649 struct cgroup_subsys_state *of_css(struct kernfs_open_file *of);
650 
651 /* cft/css accessors for cftype->seq_*() operations */
seq_cft(struct seq_file * seq)652 static inline struct cftype *seq_cft(struct seq_file *seq)
653 {
654 	return of_cft(seq->private);
655 }
656 
seq_css(struct seq_file * seq)657 static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq)
658 {
659 	return of_css(seq->private);
660 }
661 
662 /*
663  * Name / path handling functions.  All are thin wrappers around the kernfs
664  * counterparts and can be called under any context.
665  */
666 
cgroup_name(struct cgroup * cgrp,char * buf,size_t buflen)667 static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen)
668 {
669 	return kernfs_name(cgrp->kn, buf, buflen);
670 }
671 
cgroup_path(struct cgroup * cgrp,char * buf,size_t buflen)672 static inline int cgroup_path(struct cgroup *cgrp, char *buf, size_t buflen)
673 {
674 	return kernfs_path(cgrp->kn, buf, buflen);
675 }
676 
pr_cont_cgroup_name(struct cgroup * cgrp)677 static inline void pr_cont_cgroup_name(struct cgroup *cgrp)
678 {
679 	pr_cont_kernfs_name(cgrp->kn);
680 }
681 
pr_cont_cgroup_path(struct cgroup * cgrp)682 static inline void pr_cont_cgroup_path(struct cgroup *cgrp)
683 {
684 	pr_cont_kernfs_path(cgrp->kn);
685 }
686 
687 bool cgroup_psi_enabled(void);
688 
cgroup_init_kthreadd(void)689 static inline void cgroup_init_kthreadd(void)
690 {
691 	/*
692 	 * kthreadd is inherited by all kthreads, keep it in the root so
693 	 * that the new kthreads are guaranteed to stay in the root until
694 	 * initialization is finished.
695 	 */
696 	current->no_cgroup_migration = 1;
697 }
698 
cgroup_kthread_ready(void)699 static inline void cgroup_kthread_ready(void)
700 {
701 	/*
702 	 * This kthread finished initialization.  The creator should have
703 	 * set PF_NO_SETAFFINITY if this kthread should stay in the root.
704 	 */
705 	current->no_cgroup_migration = 0;
706 }
707 
708 void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen);
709 struct cgroup *cgroup_get_from_id(u64 id);
710 #else /* !CONFIG_CGROUPS */
711 
712 struct cgroup_subsys_state;
713 struct cgroup;
714 
cgroup_id(const struct cgroup * cgrp)715 static inline u64 cgroup_id(const struct cgroup *cgrp) { return 1; }
css_get(struct cgroup_subsys_state * css)716 static inline void css_get(struct cgroup_subsys_state *css) {}
css_put(struct cgroup_subsys_state * css)717 static inline void css_put(struct cgroup_subsys_state *css) {}
cgroup_lock(void)718 static inline void cgroup_lock(void) {}
cgroup_unlock(void)719 static inline void cgroup_unlock(void) {}
cgroup_attach_task_all(struct task_struct * from,struct task_struct * t)720 static inline int cgroup_attach_task_all(struct task_struct *from,
721 					 struct task_struct *t) { return 0; }
cgroupstats_build(struct cgroupstats * stats,struct dentry * dentry)722 static inline int cgroupstats_build(struct cgroupstats *stats,
723 				    struct dentry *dentry) { return -EINVAL; }
724 
cgroup_fork(struct task_struct * p)725 static inline void cgroup_fork(struct task_struct *p) {}
cgroup_can_fork(struct task_struct * p,struct kernel_clone_args * kargs)726 static inline int cgroup_can_fork(struct task_struct *p,
727 				  struct kernel_clone_args *kargs) { return 0; }
cgroup_cancel_fork(struct task_struct * p,struct kernel_clone_args * kargs)728 static inline void cgroup_cancel_fork(struct task_struct *p,
729 				      struct kernel_clone_args *kargs) {}
cgroup_post_fork(struct task_struct * p,struct kernel_clone_args * kargs)730 static inline void cgroup_post_fork(struct task_struct *p,
731 				    struct kernel_clone_args *kargs) {}
cgroup_exit(struct task_struct * p)732 static inline void cgroup_exit(struct task_struct *p) {}
cgroup_release(struct task_struct * p)733 static inline void cgroup_release(struct task_struct *p) {}
cgroup_free(struct task_struct * p)734 static inline void cgroup_free(struct task_struct *p) {}
735 
cgroup_init_early(void)736 static inline int cgroup_init_early(void) { return 0; }
cgroup_init(void)737 static inline int cgroup_init(void) { return 0; }
cgroup_init_kthreadd(void)738 static inline void cgroup_init_kthreadd(void) {}
cgroup_kthread_ready(void)739 static inline void cgroup_kthread_ready(void) {}
740 
cgroup_parent(struct cgroup * cgrp)741 static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
742 {
743 	return NULL;
744 }
745 
cgroup_psi_enabled(void)746 static inline bool cgroup_psi_enabled(void)
747 {
748 	return false;
749 }
750 
task_under_cgroup_hierarchy(struct task_struct * task,struct cgroup * ancestor)751 static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
752 					       struct cgroup *ancestor)
753 {
754 	return true;
755 }
756 
cgroup_path_from_kernfs_id(u64 id,char * buf,size_t buflen)757 static inline void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen)
758 {}
759 #endif /* !CONFIG_CGROUPS */
760 
761 #ifdef CONFIG_CGROUPS
762 /*
763  * cgroup scalable recursive statistics.
764  */
765 void cgroup_rstat_updated(struct cgroup *cgrp, int cpu);
766 void cgroup_rstat_flush(struct cgroup *cgrp);
767 void cgroup_rstat_flush_irqsafe(struct cgroup *cgrp);
768 void cgroup_rstat_flush_hold(struct cgroup *cgrp);
769 void cgroup_rstat_flush_release(void);
770 
771 /*
772  * Basic resource stats.
773  */
774 #ifdef CONFIG_CGROUP_CPUACCT
775 void cpuacct_charge(struct task_struct *tsk, u64 cputime);
776 void cpuacct_account_field(struct task_struct *tsk, int index, u64 val);
777 #else
cpuacct_charge(struct task_struct * tsk,u64 cputime)778 static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {}
cpuacct_account_field(struct task_struct * tsk,int index,u64 val)779 static inline void cpuacct_account_field(struct task_struct *tsk, int index,
780 					 u64 val) {}
781 #endif
782 
783 void __cgroup_account_cputime(struct cgroup *cgrp, u64 delta_exec);
784 void __cgroup_account_cputime_field(struct cgroup *cgrp,
785 				    enum cpu_usage_stat index, u64 delta_exec);
786 
cgroup_account_cputime(struct task_struct * task,u64 delta_exec)787 static inline void cgroup_account_cputime(struct task_struct *task,
788 					  u64 delta_exec)
789 {
790 	struct cgroup *cgrp;
791 
792 	cpuacct_charge(task, delta_exec);
793 
794 	cgrp = task_dfl_cgroup(task);
795 	if (cgroup_parent(cgrp))
796 		__cgroup_account_cputime(cgrp, delta_exec);
797 }
798 
cgroup_account_cputime_field(struct task_struct * task,enum cpu_usage_stat index,u64 delta_exec)799 static inline void cgroup_account_cputime_field(struct task_struct *task,
800 						enum cpu_usage_stat index,
801 						u64 delta_exec)
802 {
803 	struct cgroup *cgrp;
804 
805 	cpuacct_account_field(task, index, delta_exec);
806 
807 	cgrp = task_dfl_cgroup(task);
808 	if (cgroup_parent(cgrp))
809 		__cgroup_account_cputime_field(cgrp, index, delta_exec);
810 }
811 
812 #else	/* CONFIG_CGROUPS */
813 
cgroup_account_cputime(struct task_struct * task,u64 delta_exec)814 static inline void cgroup_account_cputime(struct task_struct *task,
815 					  u64 delta_exec) {}
cgroup_account_cputime_field(struct task_struct * task,enum cpu_usage_stat index,u64 delta_exec)816 static inline void cgroup_account_cputime_field(struct task_struct *task,
817 						enum cpu_usage_stat index,
818 						u64 delta_exec) {}
819 
820 #endif	/* CONFIG_CGROUPS */
821 
822 /*
823  * sock->sk_cgrp_data handling.  For more info, see sock_cgroup_data
824  * definition in cgroup-defs.h.
825  */
826 #ifdef CONFIG_SOCK_CGROUP_DATA
827 
828 void cgroup_sk_alloc(struct sock_cgroup_data *skcd);
829 void cgroup_sk_clone(struct sock_cgroup_data *skcd);
830 void cgroup_sk_free(struct sock_cgroup_data *skcd);
831 
sock_cgroup_ptr(struct sock_cgroup_data * skcd)832 static inline struct cgroup *sock_cgroup_ptr(struct sock_cgroup_data *skcd)
833 {
834 	return skcd->cgroup;
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 	struct ns_common	ns;
847 	struct user_namespace	*user_ns;
848 	struct ucounts		*ucounts;
849 	struct css_set          *root_cset;
850 };
851 
852 extern struct cgroup_namespace init_cgroup_ns;
853 
854 #ifdef CONFIG_CGROUPS
855 
856 void free_cgroup_ns(struct cgroup_namespace *ns);
857 
858 struct cgroup_namespace *copy_cgroup_ns(unsigned long flags,
859 					struct user_namespace *user_ns,
860 					struct cgroup_namespace *old_ns);
861 
862 int cgroup_path_ns(struct cgroup *cgrp, char *buf, size_t buflen,
863 		   struct cgroup_namespace *ns);
864 
865 #else /* !CONFIG_CGROUPS */
866 
free_cgroup_ns(struct cgroup_namespace * ns)867 static inline void free_cgroup_ns(struct cgroup_namespace *ns) { }
868 static inline struct cgroup_namespace *
copy_cgroup_ns(unsigned long flags,struct user_namespace * user_ns,struct cgroup_namespace * old_ns)869 copy_cgroup_ns(unsigned long flags, struct user_namespace *user_ns,
870 	       struct cgroup_namespace *old_ns)
871 {
872 	return old_ns;
873 }
874 
875 #endif /* !CONFIG_CGROUPS */
876 
get_cgroup_ns(struct cgroup_namespace * ns)877 static inline void get_cgroup_ns(struct cgroup_namespace *ns)
878 {
879 	if (ns)
880 		refcount_inc(&ns->ns.count);
881 }
882 
put_cgroup_ns(struct cgroup_namespace * ns)883 static inline void put_cgroup_ns(struct cgroup_namespace *ns)
884 {
885 	if (ns && refcount_dec_and_test(&ns->ns.count))
886 		free_cgroup_ns(ns);
887 }
888 
889 #ifdef CONFIG_CGROUPS
890 
891 void cgroup_enter_frozen(void);
892 void cgroup_leave_frozen(bool always_leave);
893 void cgroup_update_frozen(struct cgroup *cgrp);
894 void cgroup_freeze(struct cgroup *cgrp, bool freeze);
895 void cgroup_freezer_migrate_task(struct task_struct *task, struct cgroup *src,
896 				 struct cgroup *dst);
897 
cgroup_task_frozen(struct task_struct * task)898 static inline bool cgroup_task_frozen(struct task_struct *task)
899 {
900 	return task->frozen;
901 }
902 
903 #else /* !CONFIG_CGROUPS */
904 
cgroup_enter_frozen(void)905 static inline void cgroup_enter_frozen(void) { }
cgroup_leave_frozen(bool always_leave)906 static inline void cgroup_leave_frozen(bool always_leave) { }
cgroup_task_frozen(struct task_struct * task)907 static inline bool cgroup_task_frozen(struct task_struct *task)
908 {
909 	return false;
910 }
911 
912 #endif /* !CONFIG_CGROUPS */
913 
914 #ifdef CONFIG_CGROUP_BPF
cgroup_bpf_get(struct cgroup * cgrp)915 static inline void cgroup_bpf_get(struct cgroup *cgrp)
916 {
917 	percpu_ref_get(&cgrp->bpf.refcnt);
918 }
919 
cgroup_bpf_put(struct cgroup * cgrp)920 static inline void cgroup_bpf_put(struct cgroup *cgrp)
921 {
922 	percpu_ref_put(&cgrp->bpf.refcnt);
923 }
924 
925 #else /* CONFIG_CGROUP_BPF */
926 
cgroup_bpf_get(struct cgroup * cgrp)927 static inline void cgroup_bpf_get(struct cgroup *cgrp) {}
cgroup_bpf_put(struct cgroup * cgrp)928 static inline void cgroup_bpf_put(struct cgroup *cgrp) {}
929 
930 #endif /* CONFIG_CGROUP_BPF */
931 
932 #endif /* _LINUX_CGROUP_H */
933