xref: /OK3568_Linux_fs/kernel/include/linux/cgroup.h (revision 4882a59341e53eb6f0b4789bf948001014eff981)
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 #include <linux/android_kabi.h>
28 
29 #include <linux/cgroup-defs.h>
30 
31 struct kernel_clone_args;
32 
33 #ifdef CONFIG_CGROUPS
34 
35 /*
36  * All weight knobs on the default hierarhcy should use the following min,
37  * default and max values.  The default value is the logarithmic center of
38  * MIN and MAX and allows 100x to be expressed in both directions.
39  */
40 #define CGROUP_WEIGHT_MIN		1
41 #define CGROUP_WEIGHT_DFL		100
42 #define CGROUP_WEIGHT_MAX		10000
43 
44 /* walk only threadgroup leaders */
45 #define CSS_TASK_ITER_PROCS		(1U << 0)
46 /* walk all threaded css_sets in the domain */
47 #define CSS_TASK_ITER_THREADED		(1U << 1)
48 
49 /* internal flags */
50 #define CSS_TASK_ITER_SKIPPED		(1U << 16)
51 
52 /* a css_task_iter should be treated as an opaque object */
53 struct css_task_iter {
54 	struct cgroup_subsys		*ss;
55 	unsigned int			flags;
56 
57 	struct list_head		*cset_pos;
58 	struct list_head		*cset_head;
59 
60 	struct list_head		*tcset_pos;
61 	struct list_head		*tcset_head;
62 
63 	struct list_head		*task_pos;
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 	ANDROID_KABI_RESERVE(1);
72 };
73 
74 extern struct file_system_type cgroup_fs_type;
75 extern struct cgroup_root cgrp_dfl_root;
76 extern struct ext_css_set init_ext_css_set;
77 #define init_css_set init_ext_css_set.cset
78 
79 #define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys;
80 #include <linux/cgroup_subsys.h>
81 #undef SUBSYS
82 
83 #define SUBSYS(_x)								\
84 	extern struct static_key_true _x ## _cgrp_subsys_enabled_key;		\
85 	extern struct static_key_true _x ## _cgrp_subsys_on_dfl_key;
86 #include <linux/cgroup_subsys.h>
87 #undef SUBSYS
88 
89 /**
90  * cgroup_subsys_enabled - fast test on whether a subsys is enabled
91  * @ss: subsystem in question
92  */
93 #define cgroup_subsys_enabled(ss)						\
94 	static_branch_likely(&ss ## _enabled_key)
95 
96 /**
97  * cgroup_subsys_on_dfl - fast test on whether a subsys is on default hierarchy
98  * @ss: subsystem in question
99  */
100 #define cgroup_subsys_on_dfl(ss)						\
101 	static_branch_likely(&ss ## _on_dfl_key)
102 
103 bool css_has_online_children(struct cgroup_subsys_state *css);
104 struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss);
105 struct cgroup_subsys_state *cgroup_e_css(struct cgroup *cgroup,
106 					 struct cgroup_subsys *ss);
107 struct cgroup_subsys_state *cgroup_get_e_css(struct cgroup *cgroup,
108 					     struct cgroup_subsys *ss);
109 struct cgroup_subsys_state *css_tryget_online_from_dir(struct dentry *dentry,
110 						       struct cgroup_subsys *ss);
111 
112 struct cgroup *cgroup_get_from_path(const char *path);
113 struct cgroup *cgroup_get_from_fd(int fd);
114 
115 int cgroup_attach_task_all(struct task_struct *from, struct task_struct *);
116 int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from);
117 
118 int cgroup_add_dfl_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
119 int cgroup_add_legacy_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
120 int cgroup_rm_cftypes(struct cftype *cfts);
121 void cgroup_file_notify(struct cgroup_file *cfile);
122 
123 int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen);
124 int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry);
125 int proc_cgroup_show(struct seq_file *m, struct pid_namespace *ns,
126 		     struct pid *pid, struct task_struct *tsk);
127 
128 void cgroup_fork(struct task_struct *p);
129 extern int cgroup_can_fork(struct task_struct *p,
130 			   struct kernel_clone_args *kargs);
131 extern void cgroup_cancel_fork(struct task_struct *p,
132 			       struct kernel_clone_args *kargs);
133 extern void cgroup_post_fork(struct task_struct *p,
134 			     struct kernel_clone_args *kargs);
135 void cgroup_exit(struct task_struct *p);
136 void cgroup_release(struct task_struct *p);
137 void cgroup_free(struct task_struct *p);
138 
139 int cgroup_init_early(void);
140 int cgroup_init(void);
141 
142 int cgroup_parse_float(const char *input, unsigned dec_shift, s64 *v);
143 
144 /*
145  * Iteration helpers and macros.
146  */
147 
148 struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos,
149 					   struct cgroup_subsys_state *parent);
150 struct cgroup_subsys_state *css_next_descendant_pre(struct cgroup_subsys_state *pos,
151 						    struct cgroup_subsys_state *css);
152 struct cgroup_subsys_state *css_rightmost_descendant(struct cgroup_subsys_state *pos);
153 struct cgroup_subsys_state *css_next_descendant_post(struct cgroup_subsys_state *pos,
154 						     struct cgroup_subsys_state *css);
155 
156 struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset,
157 					 struct cgroup_subsys_state **dst_cssp);
158 struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset,
159 					struct cgroup_subsys_state **dst_cssp);
160 
161 void css_task_iter_start(struct cgroup_subsys_state *css, unsigned int flags,
162 			 struct css_task_iter *it);
163 struct task_struct *css_task_iter_next(struct css_task_iter *it);
164 void css_task_iter_end(struct css_task_iter *it);
165 
166 /**
167  * css_for_each_child - iterate through children of a css
168  * @pos: the css * to use as the loop cursor
169  * @parent: css whose children to walk
170  *
171  * Walk @parent's children.  Must be called under rcu_read_lock().
172  *
173  * If a subsystem synchronizes ->css_online() and the start of iteration, a
174  * css which finished ->css_online() is guaranteed to be visible in the
175  * future iterations and will stay visible until the last reference is put.
176  * A css which hasn't finished ->css_online() or already finished
177  * ->css_offline() may show up during traversal.  It's each subsystem's
178  * responsibility to synchronize against on/offlining.
179  *
180  * It is allowed to temporarily drop RCU read lock during iteration.  The
181  * caller is responsible for ensuring that @pos remains accessible until
182  * the start of the next iteration by, for example, bumping the css refcnt.
183  */
184 #define css_for_each_child(pos, parent)					\
185 	for ((pos) = css_next_child(NULL, (parent)); (pos);		\
186 	     (pos) = css_next_child((pos), (parent)))
187 
188 /**
189  * css_for_each_descendant_pre - pre-order walk of a css's descendants
190  * @pos: the css * to use as the loop cursor
191  * @root: css whose descendants to walk
192  *
193  * Walk @root's descendants.  @root is included in the iteration and the
194  * first node to be visited.  Must be called under rcu_read_lock().
195  *
196  * If a subsystem synchronizes ->css_online() and the start of iteration, a
197  * css which finished ->css_online() is guaranteed to be visible in the
198  * future iterations and will stay visible until the last reference is put.
199  * A css which hasn't finished ->css_online() or already finished
200  * ->css_offline() may show up during traversal.  It's each subsystem's
201  * responsibility to synchronize against on/offlining.
202  *
203  * For example, the following guarantees that a descendant can't escape
204  * state updates of its ancestors.
205  *
206  * my_online(@css)
207  * {
208  *	Lock @css's parent and @css;
209  *	Inherit state from the parent;
210  *	Unlock both.
211  * }
212  *
213  * my_update_state(@css)
214  * {
215  *	css_for_each_descendant_pre(@pos, @css) {
216  *		Lock @pos;
217  *		if (@pos == @css)
218  *			Update @css's state;
219  *		else
220  *			Verify @pos is alive and inherit state from its parent;
221  *		Unlock @pos;
222  *	}
223  * }
224  *
225  * As long as the inheriting step, including checking the parent state, is
226  * enclosed inside @pos locking, double-locking the parent isn't necessary
227  * while inheriting.  The state update to the parent is guaranteed to be
228  * visible by walking order and, as long as inheriting operations to the
229  * same @pos are atomic to each other, multiple updates racing each other
230  * still result in the correct state.  It's guaranateed that at least one
231  * inheritance happens for any css after the latest update to its parent.
232  *
233  * If checking parent's state requires locking the parent, each inheriting
234  * iteration should lock and unlock both @pos->parent and @pos.
235  *
236  * Alternatively, a subsystem may choose to use a single global lock to
237  * synchronize ->css_online() and ->css_offline() against tree-walking
238  * operations.
239  *
240  * It is allowed to temporarily drop RCU read lock during iteration.  The
241  * caller is responsible for ensuring that @pos remains accessible until
242  * the start of the next iteration by, for example, bumping the css refcnt.
243  */
244 #define css_for_each_descendant_pre(pos, css)				\
245 	for ((pos) = css_next_descendant_pre(NULL, (css)); (pos);	\
246 	     (pos) = css_next_descendant_pre((pos), (css)))
247 
248 /**
249  * css_for_each_descendant_post - post-order walk of a css's descendants
250  * @pos: the css * to use as the loop cursor
251  * @css: css whose descendants to walk
252  *
253  * Similar to css_for_each_descendant_pre() but performs post-order
254  * traversal instead.  @root is included in the iteration and the last
255  * node to be visited.
256  *
257  * If a subsystem synchronizes ->css_online() and the start of iteration, a
258  * css which finished ->css_online() is guaranteed to be visible in the
259  * future iterations and will stay visible until the last reference is put.
260  * A css which hasn't finished ->css_online() or already finished
261  * ->css_offline() may show up during traversal.  It's each subsystem's
262  * responsibility to synchronize against on/offlining.
263  *
264  * Note that the walk visibility guarantee example described in pre-order
265  * walk doesn't apply the same to post-order walks.
266  */
267 #define css_for_each_descendant_post(pos, css)				\
268 	for ((pos) = css_next_descendant_post(NULL, (css)); (pos);	\
269 	     (pos) = css_next_descendant_post((pos), (css)))
270 
271 /**
272  * cgroup_taskset_for_each - iterate cgroup_taskset
273  * @task: the loop cursor
274  * @dst_css: the destination css
275  * @tset: taskset to iterate
276  *
277  * @tset may contain multiple tasks and they may belong to multiple
278  * processes.
279  *
280  * On the v2 hierarchy, there may be tasks from multiple processes and they
281  * may not share the source or destination csses.
282  *
283  * On traditional hierarchies, when there are multiple tasks in @tset, if a
284  * task of a process is in @tset, all tasks of the process are in @tset.
285  * Also, all are guaranteed to share the same source and destination csses.
286  *
287  * Iteration is not in any specific order.
288  */
289 #define cgroup_taskset_for_each(task, dst_css, tset)			\
290 	for ((task) = cgroup_taskset_first((tset), &(dst_css));		\
291 	     (task);							\
292 	     (task) = cgroup_taskset_next((tset), &(dst_css)))
293 
294 /**
295  * cgroup_taskset_for_each_leader - iterate group leaders in a cgroup_taskset
296  * @leader: the loop cursor
297  * @dst_css: the destination css
298  * @tset: taskset to iterate
299  *
300  * Iterate threadgroup leaders of @tset.  For single-task migrations, @tset
301  * may not contain any.
302  */
303 #define cgroup_taskset_for_each_leader(leader, dst_css, tset)		\
304 	for ((leader) = cgroup_taskset_first((tset), &(dst_css));	\
305 	     (leader);							\
306 	     (leader) = cgroup_taskset_next((tset), &(dst_css)))	\
307 		if ((leader) != (leader)->group_leader)			\
308 			;						\
309 		else
310 
311 /*
312  * Inline functions.
313  */
314 
cgroup_id(struct cgroup * cgrp)315 static inline u64 cgroup_id(struct cgroup *cgrp)
316 {
317 	return cgrp->kn->id;
318 }
319 
320 /**
321  * css_get - obtain a reference on the specified css
322  * @css: target css
323  *
324  * The caller must already have a reference.
325  */
css_get(struct cgroup_subsys_state * css)326 static inline void css_get(struct cgroup_subsys_state *css)
327 {
328 	if (!(css->flags & CSS_NO_REF))
329 		percpu_ref_get(&css->refcnt);
330 }
331 
332 /**
333  * css_get_many - obtain references on the specified css
334  * @css: target css
335  * @n: number of references to get
336  *
337  * The caller must already have a reference.
338  */
css_get_many(struct cgroup_subsys_state * css,unsigned int n)339 static inline void css_get_many(struct cgroup_subsys_state *css, unsigned int n)
340 {
341 	if (!(css->flags & CSS_NO_REF))
342 		percpu_ref_get_many(&css->refcnt, n);
343 }
344 
345 /**
346  * css_tryget - try to obtain a reference on the specified css
347  * @css: target css
348  *
349  * Obtain a reference on @css unless it already has reached zero and is
350  * being released.  This function doesn't care whether @css is on or
351  * offline.  The caller naturally needs to ensure that @css is accessible
352  * but doesn't have to be holding a reference on it - IOW, RCU protected
353  * access is good enough for this function.  Returns %true if a reference
354  * count was successfully obtained; %false otherwise.
355  */
css_tryget(struct cgroup_subsys_state * css)356 static inline bool css_tryget(struct cgroup_subsys_state *css)
357 {
358 	if (!(css->flags & CSS_NO_REF))
359 		return percpu_ref_tryget(&css->refcnt);
360 	return true;
361 }
362 
363 /**
364  * css_tryget_online - try to obtain a reference on the specified css if online
365  * @css: target css
366  *
367  * Obtain a reference on @css if it's online.  The caller naturally needs
368  * to ensure that @css is accessible but doesn't have to be holding a
369  * reference on it - IOW, RCU protected access is good enough for this
370  * function.  Returns %true if a reference count was successfully obtained;
371  * %false otherwise.
372  */
css_tryget_online(struct cgroup_subsys_state * css)373 static inline bool css_tryget_online(struct cgroup_subsys_state *css)
374 {
375 	if (!(css->flags & CSS_NO_REF))
376 		return percpu_ref_tryget_live(&css->refcnt);
377 	return true;
378 }
379 
380 /**
381  * css_is_dying - test whether the specified css is dying
382  * @css: target css
383  *
384  * Test whether @css is in the process of offlining or already offline.  In
385  * most cases, ->css_online() and ->css_offline() callbacks should be
386  * enough; however, the actual offline operations are RCU delayed and this
387  * test returns %true also when @css is scheduled to be offlined.
388  *
389  * This is useful, for example, when the use case requires synchronous
390  * behavior with respect to cgroup removal.  cgroup removal schedules css
391  * offlining but the css can seem alive while the operation is being
392  * delayed.  If the delay affects user visible semantics, this test can be
393  * used to resolve the situation.
394  */
css_is_dying(struct cgroup_subsys_state * css)395 static inline bool css_is_dying(struct cgroup_subsys_state *css)
396 {
397 	return !(css->flags & CSS_NO_REF) && percpu_ref_is_dying(&css->refcnt);
398 }
399 
400 /**
401  * css_put - put a css reference
402  * @css: target css
403  *
404  * Put a reference obtained via css_get() and css_tryget_online().
405  */
css_put(struct cgroup_subsys_state * css)406 static inline void css_put(struct cgroup_subsys_state *css)
407 {
408 	if (!(css->flags & CSS_NO_REF))
409 		percpu_ref_put(&css->refcnt);
410 }
411 
412 /**
413  * css_put_many - put css references
414  * @css: target css
415  * @n: number of references to put
416  *
417  * Put references obtained via css_get() and css_tryget_online().
418  */
css_put_many(struct cgroup_subsys_state * css,unsigned int n)419 static inline void css_put_many(struct cgroup_subsys_state *css, unsigned int n)
420 {
421 	if (!(css->flags & CSS_NO_REF))
422 		percpu_ref_put_many(&css->refcnt, n);
423 }
424 
cgroup_get(struct cgroup * cgrp)425 static inline void cgroup_get(struct cgroup *cgrp)
426 {
427 	css_get(&cgrp->self);
428 }
429 
cgroup_tryget(struct cgroup * cgrp)430 static inline bool cgroup_tryget(struct cgroup *cgrp)
431 {
432 	return css_tryget(&cgrp->self);
433 }
434 
cgroup_put(struct cgroup * cgrp)435 static inline void cgroup_put(struct cgroup *cgrp)
436 {
437 	css_put(&cgrp->self);
438 }
439 
440 /**
441  * task_css_set_check - obtain a task's css_set with extra access conditions
442  * @task: the task to obtain css_set for
443  * @__c: extra condition expression to be passed to rcu_dereference_check()
444  *
445  * A task's css_set is RCU protected, initialized and exited while holding
446  * task_lock(), and can only be modified while holding both cgroup_mutex
447  * and task_lock() while the task is alive.  This macro verifies that the
448  * caller is inside proper critical section and returns @task's css_set.
449  *
450  * The caller can also specify additional allowed conditions via @__c, such
451  * as locks used during the cgroup_subsys::attach() methods.
452  */
453 #ifdef CONFIG_PROVE_RCU
454 extern struct mutex cgroup_mutex;
455 extern spinlock_t css_set_lock;
456 #define task_css_set_check(task, __c)					\
457 	rcu_dereference_check((task)->cgroups,				\
458 		lockdep_is_held(&cgroup_mutex) ||			\
459 		lockdep_is_held(&css_set_lock) ||			\
460 		((task)->flags & PF_EXITING) || (__c))
461 #else
462 #define task_css_set_check(task, __c)					\
463 	rcu_dereference((task)->cgroups)
464 #endif
465 
466 /**
467  * task_css_check - obtain css for (task, subsys) w/ extra access conds
468  * @task: the target task
469  * @subsys_id: the target subsystem ID
470  * @__c: extra condition expression to be passed to rcu_dereference_check()
471  *
472  * Return the cgroup_subsys_state for the (@task, @subsys_id) pair.  The
473  * synchronization rules are the same as task_css_set_check().
474  */
475 #define task_css_check(task, subsys_id, __c)				\
476 	task_css_set_check((task), (__c))->subsys[(subsys_id)]
477 
478 /**
479  * task_css_set - obtain a task's css_set
480  * @task: the task to obtain css_set for
481  *
482  * See task_css_set_check().
483  */
task_css_set(struct task_struct * task)484 static inline struct css_set *task_css_set(struct task_struct *task)
485 {
486 	return task_css_set_check(task, false);
487 }
488 
489 /**
490  * task_css - obtain css for (task, subsys)
491  * @task: the target task
492  * @subsys_id: the target subsystem ID
493  *
494  * See task_css_check().
495  */
task_css(struct task_struct * task,int subsys_id)496 static inline struct cgroup_subsys_state *task_css(struct task_struct *task,
497 						   int subsys_id)
498 {
499 	return task_css_check(task, subsys_id, false);
500 }
501 
502 /**
503  * task_get_css - find and get the css for (task, subsys)
504  * @task: the target task
505  * @subsys_id: the target subsystem ID
506  *
507  * Find the css for the (@task, @subsys_id) combination, increment a
508  * reference on and return it.  This function is guaranteed to return a
509  * valid css.  The returned css may already have been offlined.
510  */
511 static inline struct cgroup_subsys_state *
task_get_css(struct task_struct * task,int subsys_id)512 task_get_css(struct task_struct *task, int subsys_id)
513 {
514 	struct cgroup_subsys_state *css;
515 
516 	rcu_read_lock();
517 	while (true) {
518 		css = task_css(task, subsys_id);
519 		/*
520 		 * Can't use css_tryget_online() here.  A task which has
521 		 * PF_EXITING set may stay associated with an offline css.
522 		 * If such task calls this function, css_tryget_online()
523 		 * will keep failing.
524 		 */
525 		if (likely(css_tryget(css)))
526 			break;
527 		cpu_relax();
528 	}
529 	rcu_read_unlock();
530 	return css;
531 }
532 
533 /**
534  * task_css_is_root - test whether a task belongs to the root css
535  * @task: the target task
536  * @subsys_id: the target subsystem ID
537  *
538  * Test whether @task belongs to the root css on the specified subsystem.
539  * May be invoked in any context.
540  */
task_css_is_root(struct task_struct * task,int subsys_id)541 static inline bool task_css_is_root(struct task_struct *task, int subsys_id)
542 {
543 	return task_css_check(task, subsys_id, true) ==
544 		init_css_set.subsys[subsys_id];
545 }
546 
task_cgroup(struct task_struct * task,int subsys_id)547 static inline struct cgroup *task_cgroup(struct task_struct *task,
548 					 int subsys_id)
549 {
550 	return task_css(task, subsys_id)->cgroup;
551 }
552 
task_dfl_cgroup(struct task_struct * task)553 static inline struct cgroup *task_dfl_cgroup(struct task_struct *task)
554 {
555 	return task_css_set(task)->dfl_cgrp;
556 }
557 
cgroup_parent(struct cgroup * cgrp)558 static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
559 {
560 	struct cgroup_subsys_state *parent_css = cgrp->self.parent;
561 
562 	if (parent_css)
563 		return container_of(parent_css, struct cgroup, self);
564 	return NULL;
565 }
566 
567 /**
568  * cgroup_is_descendant - test ancestry
569  * @cgrp: the cgroup to be tested
570  * @ancestor: possible ancestor of @cgrp
571  *
572  * Test whether @cgrp is a descendant of @ancestor.  It also returns %true
573  * if @cgrp == @ancestor.  This function is safe to call as long as @cgrp
574  * and @ancestor are accessible.
575  */
cgroup_is_descendant(struct cgroup * cgrp,struct cgroup * ancestor)576 static inline bool cgroup_is_descendant(struct cgroup *cgrp,
577 					struct cgroup *ancestor)
578 {
579 	if (cgrp->root != ancestor->root || cgrp->level < ancestor->level)
580 		return false;
581 	return cgrp->ancestor_ids[ancestor->level] == cgroup_id(ancestor);
582 }
583 
584 /**
585  * cgroup_ancestor - find ancestor of cgroup
586  * @cgrp: cgroup to find ancestor of
587  * @ancestor_level: level of ancestor to find starting from root
588  *
589  * Find ancestor of cgroup at specified level starting from root if it exists
590  * and return pointer to it. Return NULL if @cgrp doesn't have ancestor at
591  * @ancestor_level.
592  *
593  * This function is safe to call as long as @cgrp is accessible.
594  */
cgroup_ancestor(struct cgroup * cgrp,int ancestor_level)595 static inline struct cgroup *cgroup_ancestor(struct cgroup *cgrp,
596 					     int ancestor_level)
597 {
598 	if (cgrp->level < ancestor_level)
599 		return NULL;
600 	while (cgrp && cgrp->level > ancestor_level)
601 		cgrp = cgroup_parent(cgrp);
602 	return cgrp;
603 }
604 
605 /**
606  * task_under_cgroup_hierarchy - test task's membership of cgroup ancestry
607  * @task: the task to be tested
608  * @ancestor: possible ancestor of @task's cgroup
609  *
610  * Tests whether @task's default cgroup hierarchy is a descendant of @ancestor.
611  * It follows all the same rules as cgroup_is_descendant, and only applies
612  * to the default hierarchy.
613  */
task_under_cgroup_hierarchy(struct task_struct * task,struct cgroup * ancestor)614 static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
615 					       struct cgroup *ancestor)
616 {
617 	struct css_set *cset = task_css_set(task);
618 
619 	return cgroup_is_descendant(cset->dfl_cgrp, ancestor);
620 }
621 
622 /* no synchronization, the result can only be used as a hint */
cgroup_is_populated(struct cgroup * cgrp)623 static inline bool cgroup_is_populated(struct cgroup *cgrp)
624 {
625 	return cgrp->nr_populated_csets + cgrp->nr_populated_domain_children +
626 		cgrp->nr_populated_threaded_children;
627 }
628 
629 /* returns ino associated with a cgroup */
cgroup_ino(struct cgroup * cgrp)630 static inline ino_t cgroup_ino(struct cgroup *cgrp)
631 {
632 	return kernfs_ino(cgrp->kn);
633 }
634 
635 /* cft/css accessors for cftype->write() operation */
of_cft(struct kernfs_open_file * of)636 static inline struct cftype *of_cft(struct kernfs_open_file *of)
637 {
638 	return of->kn->priv;
639 }
640 
641 struct cgroup_subsys_state *of_css(struct kernfs_open_file *of);
642 
643 /* cft/css accessors for cftype->seq_*() operations */
seq_cft(struct seq_file * seq)644 static inline struct cftype *seq_cft(struct seq_file *seq)
645 {
646 	return of_cft(seq->private);
647 }
648 
seq_css(struct seq_file * seq)649 static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq)
650 {
651 	return of_css(seq->private);
652 }
653 
654 /*
655  * Name / path handling functions.  All are thin wrappers around the kernfs
656  * counterparts and can be called under any context.
657  */
658 
cgroup_name(struct cgroup * cgrp,char * buf,size_t buflen)659 static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen)
660 {
661 	return kernfs_name(cgrp->kn, buf, buflen);
662 }
663 
cgroup_path(struct cgroup * cgrp,char * buf,size_t buflen)664 static inline int cgroup_path(struct cgroup *cgrp, char *buf, size_t buflen)
665 {
666 	return kernfs_path(cgrp->kn, buf, buflen);
667 }
668 
pr_cont_cgroup_name(struct cgroup * cgrp)669 static inline void pr_cont_cgroup_name(struct cgroup *cgrp)
670 {
671 	pr_cont_kernfs_name(cgrp->kn);
672 }
673 
pr_cont_cgroup_path(struct cgroup * cgrp)674 static inline void pr_cont_cgroup_path(struct cgroup *cgrp)
675 {
676 	pr_cont_kernfs_path(cgrp->kn);
677 }
678 
cgroup_psi(struct cgroup * cgrp)679 static inline struct psi_group *cgroup_psi(struct cgroup *cgrp)
680 {
681 	return &cgrp->psi;
682 }
683 
684 bool cgroup_psi_enabled(void);
685 
cgroup_init_kthreadd(void)686 static inline void cgroup_init_kthreadd(void)
687 {
688 	/*
689 	 * kthreadd is inherited by all kthreads, keep it in the root so
690 	 * that the new kthreads are guaranteed to stay in the root until
691 	 * initialization is finished.
692 	 */
693 	current->no_cgroup_migration = 1;
694 }
695 
cgroup_kthread_ready(void)696 static inline void cgroup_kthread_ready(void)
697 {
698 	/*
699 	 * This kthread finished initialization.  The creator should have
700 	 * set PF_NO_SETAFFINITY if this kthread should stay in the root.
701 	 */
702 	current->no_cgroup_migration = 0;
703 }
704 
705 void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen);
706 #else /* !CONFIG_CGROUPS */
707 
708 struct cgroup_subsys_state;
709 struct cgroup;
710 
cgroup_id(struct cgroup * cgrp)711 static inline u64 cgroup_id(struct cgroup *cgrp) { return 1; }
css_get(struct cgroup_subsys_state * css)712 static inline void css_get(struct cgroup_subsys_state *css) {}
css_put(struct cgroup_subsys_state * css)713 static inline void css_put(struct cgroup_subsys_state *css) {}
cgroup_attach_task_all(struct task_struct * from,struct task_struct * t)714 static inline int cgroup_attach_task_all(struct task_struct *from,
715 					 struct task_struct *t) { return 0; }
cgroupstats_build(struct cgroupstats * stats,struct dentry * dentry)716 static inline int cgroupstats_build(struct cgroupstats *stats,
717 				    struct dentry *dentry) { return -EINVAL; }
718 
cgroup_fork(struct task_struct * p)719 static inline void cgroup_fork(struct task_struct *p) {}
cgroup_can_fork(struct task_struct * p,struct kernel_clone_args * kargs)720 static inline int cgroup_can_fork(struct task_struct *p,
721 				  struct kernel_clone_args *kargs) { return 0; }
cgroup_cancel_fork(struct task_struct * p,struct kernel_clone_args * kargs)722 static inline void cgroup_cancel_fork(struct task_struct *p,
723 				      struct kernel_clone_args *kargs) {}
cgroup_post_fork(struct task_struct * p,struct kernel_clone_args * kargs)724 static inline void cgroup_post_fork(struct task_struct *p,
725 				    struct kernel_clone_args *kargs) {}
cgroup_exit(struct task_struct * p)726 static inline void cgroup_exit(struct task_struct *p) {}
cgroup_release(struct task_struct * p)727 static inline void cgroup_release(struct task_struct *p) {}
cgroup_free(struct task_struct * p)728 static inline void cgroup_free(struct task_struct *p) {}
729 
cgroup_init_early(void)730 static inline int cgroup_init_early(void) { return 0; }
cgroup_init(void)731 static inline int cgroup_init(void) { return 0; }
cgroup_init_kthreadd(void)732 static inline void cgroup_init_kthreadd(void) {}
cgroup_kthread_ready(void)733 static inline void cgroup_kthread_ready(void) {}
734 
cgroup_parent(struct cgroup * cgrp)735 static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
736 {
737 	return NULL;
738 }
739 
cgroup_psi(struct cgroup * cgrp)740 static inline struct psi_group *cgroup_psi(struct cgroup *cgrp)
741 {
742 	return NULL;
743 }
744 
cgroup_psi_enabled(void)745 static inline bool cgroup_psi_enabled(void)
746 {
747 	return false;
748 }
749 
task_under_cgroup_hierarchy(struct task_struct * task,struct cgroup * ancestor)750 static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
751 					       struct cgroup *ancestor)
752 {
753 	return true;
754 }
755 
cgroup_path_from_kernfs_id(u64 id,char * buf,size_t buflen)756 static inline void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen)
757 {}
758 #endif /* !CONFIG_CGROUPS */
759 
760 #ifdef CONFIG_CGROUPS
761 /*
762  * cgroup scalable recursive statistics.
763  */
764 void cgroup_rstat_updated(struct cgroup *cgrp, int cpu);
765 void cgroup_rstat_flush(struct cgroup *cgrp);
766 void cgroup_rstat_flush_irqsafe(struct cgroup *cgrp);
767 void cgroup_rstat_flush_hold(struct cgroup *cgrp);
768 void cgroup_rstat_flush_release(void);
769 
770 /*
771  * Basic resource stats.
772  */
773 #ifdef CONFIG_CGROUP_CPUACCT
774 void cpuacct_charge(struct task_struct *tsk, u64 cputime);
775 void cpuacct_account_field(struct task_struct *tsk, int index, u64 val);
776 #else
cpuacct_charge(struct task_struct * tsk,u64 cputime)777 static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {}
cpuacct_account_field(struct task_struct * tsk,int index,u64 val)778 static inline void cpuacct_account_field(struct task_struct *tsk, int index,
779 					 u64 val) {}
780 #endif
781 
782 void __cgroup_account_cputime(struct cgroup *cgrp, u64 delta_exec);
783 void __cgroup_account_cputime_field(struct cgroup *cgrp,
784 				    enum cpu_usage_stat index, u64 delta_exec);
785 
cgroup_account_cputime(struct task_struct * task,u64 delta_exec)786 static inline void cgroup_account_cputime(struct task_struct *task,
787 					  u64 delta_exec)
788 {
789 	struct cgroup *cgrp;
790 
791 	cpuacct_charge(task, delta_exec);
792 
793 	rcu_read_lock();
794 	cgrp = task_dfl_cgroup(task);
795 	if (cgroup_parent(cgrp))
796 		__cgroup_account_cputime(cgrp, delta_exec);
797 	rcu_read_unlock();
798 }
799 
cgroup_account_cputime_field(struct task_struct * task,enum cpu_usage_stat index,u64 delta_exec)800 static inline void cgroup_account_cputime_field(struct task_struct *task,
801 						enum cpu_usage_stat index,
802 						u64 delta_exec)
803 {
804 	struct cgroup *cgrp;
805 
806 	cpuacct_account_field(task, index, delta_exec);
807 
808 	rcu_read_lock();
809 	cgrp = task_dfl_cgroup(task);
810 	if (cgroup_parent(cgrp))
811 		__cgroup_account_cputime_field(cgrp, index, delta_exec);
812 	rcu_read_unlock();
813 }
814 
815 #else	/* CONFIG_CGROUPS */
816 
cgroup_account_cputime(struct task_struct * task,u64 delta_exec)817 static inline void cgroup_account_cputime(struct task_struct *task,
818 					  u64 delta_exec) {}
cgroup_account_cputime_field(struct task_struct * task,enum cpu_usage_stat index,u64 delta_exec)819 static inline void cgroup_account_cputime_field(struct task_struct *task,
820 						enum cpu_usage_stat index,
821 						u64 delta_exec) {}
822 
823 #endif	/* CONFIG_CGROUPS */
824 
825 /*
826  * sock->sk_cgrp_data handling.  For more info, see sock_cgroup_data
827  * definition in cgroup-defs.h.
828  */
829 #ifdef CONFIG_SOCK_CGROUP_DATA
830 
831 #if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
832 extern spinlock_t cgroup_sk_update_lock;
833 #endif
834 
835 void cgroup_sk_alloc_disable(void);
836 void cgroup_sk_alloc(struct sock_cgroup_data *skcd);
837 void cgroup_sk_clone(struct sock_cgroup_data *skcd);
838 void cgroup_sk_free(struct sock_cgroup_data *skcd);
839 
sock_cgroup_ptr(struct sock_cgroup_data * skcd)840 static inline struct cgroup *sock_cgroup_ptr(struct sock_cgroup_data *skcd)
841 {
842 #if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
843 	unsigned long v;
844 
845 	/*
846 	 * @skcd->val is 64bit but the following is safe on 32bit too as we
847 	 * just need the lower ulong to be written and read atomically.
848 	 */
849 	v = READ_ONCE(skcd->val);
850 
851 	if (v & 3)
852 		return &cgrp_dfl_root.cgrp;
853 
854 	return (struct cgroup *)(unsigned long)v ?: &cgrp_dfl_root.cgrp;
855 #else
856 	return (struct cgroup *)(unsigned long)skcd->val;
857 #endif
858 }
859 
860 #else	/* CONFIG_CGROUP_DATA */
861 
cgroup_sk_alloc(struct sock_cgroup_data * skcd)862 static inline void cgroup_sk_alloc(struct sock_cgroup_data *skcd) {}
cgroup_sk_clone(struct sock_cgroup_data * skcd)863 static inline void cgroup_sk_clone(struct sock_cgroup_data *skcd) {}
cgroup_sk_free(struct sock_cgroup_data * skcd)864 static inline void cgroup_sk_free(struct sock_cgroup_data *skcd) {}
865 
866 #endif	/* CONFIG_CGROUP_DATA */
867 
868 struct cgroup_namespace {
869 	refcount_t		count;
870 	struct ns_common	ns;
871 	struct user_namespace	*user_ns;
872 	struct ucounts		*ucounts;
873 	struct css_set          *root_cset;
874 };
875 
876 extern struct cgroup_namespace init_cgroup_ns;
877 
878 #ifdef CONFIG_CGROUPS
879 
880 void free_cgroup_ns(struct cgroup_namespace *ns);
881 
882 struct cgroup_namespace *copy_cgroup_ns(unsigned long flags,
883 					struct user_namespace *user_ns,
884 					struct cgroup_namespace *old_ns);
885 
886 int cgroup_path_ns(struct cgroup *cgrp, char *buf, size_t buflen,
887 		   struct cgroup_namespace *ns);
888 
889 #else /* !CONFIG_CGROUPS */
890 
free_cgroup_ns(struct cgroup_namespace * ns)891 static inline void free_cgroup_ns(struct cgroup_namespace *ns) { }
892 static inline struct cgroup_namespace *
copy_cgroup_ns(unsigned long flags,struct user_namespace * user_ns,struct cgroup_namespace * old_ns)893 copy_cgroup_ns(unsigned long flags, struct user_namespace *user_ns,
894 	       struct cgroup_namespace *old_ns)
895 {
896 	return old_ns;
897 }
898 
899 #endif /* !CONFIG_CGROUPS */
900 
get_cgroup_ns(struct cgroup_namespace * ns)901 static inline void get_cgroup_ns(struct cgroup_namespace *ns)
902 {
903 	if (ns)
904 		refcount_inc(&ns->count);
905 }
906 
put_cgroup_ns(struct cgroup_namespace * ns)907 static inline void put_cgroup_ns(struct cgroup_namespace *ns)
908 {
909 	if (ns && refcount_dec_and_test(&ns->count))
910 		free_cgroup_ns(ns);
911 }
912 
913 #ifdef CONFIG_CGROUPS
914 
915 void cgroup_enter_frozen(void);
916 void cgroup_leave_frozen(bool always_leave);
917 void cgroup_update_frozen(struct cgroup *cgrp);
918 void cgroup_freeze(struct cgroup *cgrp, bool freeze);
919 void cgroup_freezer_migrate_task(struct task_struct *task, struct cgroup *src,
920 				 struct cgroup *dst);
921 
cgroup_task_freeze(struct task_struct * task)922 static inline bool cgroup_task_freeze(struct task_struct *task)
923 {
924 	bool ret;
925 
926 	if (task->flags & PF_KTHREAD)
927 		return false;
928 
929 	rcu_read_lock();
930 	ret = test_bit(CGRP_FREEZE, &task_dfl_cgroup(task)->flags);
931 	rcu_read_unlock();
932 
933 	return ret;
934 }
935 
cgroup_task_frozen(struct task_struct * task)936 static inline bool cgroup_task_frozen(struct task_struct *task)
937 {
938 	return task->frozen;
939 }
940 
941 #else /* !CONFIG_CGROUPS */
942 
cgroup_enter_frozen(void)943 static inline void cgroup_enter_frozen(void) { }
cgroup_leave_frozen(bool always_leave)944 static inline void cgroup_leave_frozen(bool always_leave) { }
cgroup_task_freeze(struct task_struct * task)945 static inline bool cgroup_task_freeze(struct task_struct *task)
946 {
947 	return false;
948 }
cgroup_task_frozen(struct task_struct * task)949 static inline bool cgroup_task_frozen(struct task_struct *task)
950 {
951 	return false;
952 }
953 
954 #endif /* !CONFIG_CGROUPS */
955 
956 #ifdef CONFIG_CGROUP_BPF
cgroup_bpf_get(struct cgroup * cgrp)957 static inline void cgroup_bpf_get(struct cgroup *cgrp)
958 {
959 	percpu_ref_get(&cgrp->bpf.refcnt);
960 }
961 
cgroup_bpf_put(struct cgroup * cgrp)962 static inline void cgroup_bpf_put(struct cgroup *cgrp)
963 {
964 	percpu_ref_put(&cgrp->bpf.refcnt);
965 }
966 
967 #else /* CONFIG_CGROUP_BPF */
968 
cgroup_bpf_get(struct cgroup * cgrp)969 static inline void cgroup_bpf_get(struct cgroup *cgrp) {}
cgroup_bpf_put(struct cgroup * cgrp)970 static inline void cgroup_bpf_put(struct cgroup *cgrp) {}
971 
972 #endif /* CONFIG_CGROUP_BPF */
973 
974 #endif /* _LINUX_CGROUP_H */
975