]> git.ipfire.org Git - thirdparty/kernel/stable.git/blame - include/linux/cgroup.h
cgroup: Use css_tryget() instead of css_tryget_online() in task_get_css()
[thirdparty/kernel/stable.git] / include / linux / cgroup.h
CommitLineData
b2441318 1/* SPDX-License-Identifier: GPL-2.0 */
ddbcc7e8
PM
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>
ddbcc7e8
PM
13#include <linux/cpumask.h>
14#include <linux/nodemask.h>
eb6fd504 15#include <linux/rculist.h>
846c7bb0 16#include <linux/cgroupstats.h>
25a7e684 17#include <linux/fs.h>
7da11279 18#include <linux/seq_file.h>
2bd59d48 19#include <linux/kernfs.h>
49d1dc4b 20#include <linux/jump_label.h>
a79a908f
AK
21#include <linux/types.h>
22#include <linux/ns_common.h>
23#include <linux/nsproxy.h>
24#include <linux/user_namespace.h>
387ad967 25#include <linux/refcount.h>
d2cc5ed6 26#include <linux/kernel_stat.h>
ddbcc7e8 27
b4a04ab7 28#include <linux/cgroup-defs.h>
ddbcc7e8
PM
29
30#ifdef CONFIG_CGROUPS
31
6abc8ca1
TH
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
bc2fb7ed
TH
41/* walk only threadgroup leaders */
42#define CSS_TASK_ITER_PROCS (1U << 0)
450ee0c1
TH
43/* walk all threaded css_sets in the domain */
44#define CSS_TASK_ITER_THREADED (1U << 1)
bc2fb7ed 45
c326aa2b
TH
46/* a css_task_iter should be treated as an opaque object */
47struct css_task_iter {
48 struct cgroup_subsys *ss;
bc2fb7ed 49 unsigned int flags;
c326aa2b
TH
50
51 struct list_head *cset_pos;
52 struct list_head *cset_head;
ddbcc7e8 53
450ee0c1
TH
54 struct list_head *tcset_pos;
55 struct list_head *tcset_head;
56
c326aa2b
TH
57 struct list_head *task_pos;
58 struct list_head *tasks_head;
59 struct list_head *mg_tasks_head;
ed27b9f7
TH
60
61 struct css_set *cur_cset;
450ee0c1 62 struct css_set *cur_dcset;
ed27b9f7
TH
63 struct task_struct *cur_task;
64 struct list_head iters_node; /* css_set->task_iters */
c326aa2b 65};
ddbcc7e8 66
c326aa2b
TH
67extern struct cgroup_root cgrp_dfl_root;
68extern struct css_set init_css_set;
a424316c 69
c326aa2b 70#define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys;
817929ec 71#include <linux/cgroup_subsys.h>
817929ec
PM
72#undef SUBSYS
73
49d1dc4b
TH
74#define SUBSYS(_x) \
75 extern struct static_key_true _x ## _cgrp_subsys_enabled_key; \
76 extern struct static_key_true _x ## _cgrp_subsys_on_dfl_key;
77#include <linux/cgroup_subsys.h>
78#undef SUBSYS
79
80/**
81 * cgroup_subsys_enabled - fast test on whether a subsys is enabled
82 * @ss: subsystem in question
83 */
84#define cgroup_subsys_enabled(ss) \
85 static_branch_likely(&ss ## _enabled_key)
86
87/**
88 * cgroup_subsys_on_dfl - fast test on whether a subsys is on default hierarchy
89 * @ss: subsystem in question
90 */
91#define cgroup_subsys_on_dfl(ss) \
92 static_branch_likely(&ss ## _on_dfl_key)
93
c326aa2b
TH
94bool css_has_online_children(struct cgroup_subsys_state *css);
95struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss);
fc5a828b
DZ
96struct cgroup_subsys_state *cgroup_e_css(struct cgroup *cgroup,
97 struct cgroup_subsys *ss);
c326aa2b
TH
98struct cgroup_subsys_state *cgroup_get_e_css(struct cgroup *cgroup,
99 struct cgroup_subsys *ss);
100struct cgroup_subsys_state *css_tryget_online_from_dir(struct dentry *dentry,
101 struct cgroup_subsys *ss);
102
16af4396 103struct cgroup *cgroup_get_from_path(const char *path);
1f3fe7eb 104struct cgroup *cgroup_get_from_fd(int fd);
16af4396 105
c326aa2b
TH
106int cgroup_attach_task_all(struct task_struct *from, struct task_struct *);
107int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from);
108
109int cgroup_add_dfl_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
110int cgroup_add_legacy_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
111int cgroup_rm_cftypes(struct cftype *cfts);
34c06254 112void cgroup_file_notify(struct cgroup_file *cfile);
c326aa2b 113
4c737b41 114int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen);
c326aa2b
TH
115int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry);
116int proc_cgroup_show(struct seq_file *m, struct pid_namespace *ns,
117 struct pid *pid, struct task_struct *tsk);
118
119void cgroup_fork(struct task_struct *p);
b53202e6
ON
120extern int cgroup_can_fork(struct task_struct *p);
121extern void cgroup_cancel_fork(struct task_struct *p);
122extern void cgroup_post_fork(struct task_struct *p);
c326aa2b 123void cgroup_exit(struct task_struct *p);
51bee5ab 124void cgroup_release(struct task_struct *p);
2e91fa7f 125void cgroup_free(struct task_struct *p);
c326aa2b
TH
126
127int cgroup_init_early(void);
128int cgroup_init(void);
129
5c9d535b 130/*
c326aa2b
TH
131 * Iteration helpers and macros.
132 */
133
134struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos,
135 struct cgroup_subsys_state *parent);
136struct cgroup_subsys_state *css_next_descendant_pre(struct cgroup_subsys_state *pos,
137 struct cgroup_subsys_state *css);
138struct cgroup_subsys_state *css_rightmost_descendant(struct cgroup_subsys_state *pos);
139struct cgroup_subsys_state *css_next_descendant_post(struct cgroup_subsys_state *pos,
140 struct cgroup_subsys_state *css);
141
1f7dd3e5
TH
142struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset,
143 struct cgroup_subsys_state **dst_cssp);
144struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset,
145 struct cgroup_subsys_state **dst_cssp);
c326aa2b 146
bc2fb7ed 147void css_task_iter_start(struct cgroup_subsys_state *css, unsigned int flags,
c326aa2b
TH
148 struct css_task_iter *it);
149struct task_struct *css_task_iter_next(struct css_task_iter *it);
150void css_task_iter_end(struct css_task_iter *it);
151
152/**
153 * css_for_each_child - iterate through children of a css
154 * @pos: the css * to use as the loop cursor
155 * @parent: css whose children to walk
156 *
157 * Walk @parent's children. Must be called under rcu_read_lock().
158 *
159 * If a subsystem synchronizes ->css_online() and the start of iteration, a
160 * css which finished ->css_online() is guaranteed to be visible in the
161 * future iterations and will stay visible until the last reference is put.
162 * A css which hasn't finished ->css_online() or already finished
163 * ->css_offline() may show up during traversal. It's each subsystem's
164 * responsibility to synchronize against on/offlining.
5c9d535b 165 *
c326aa2b
TH
166 * It is allowed to temporarily drop RCU read lock during iteration. The
167 * caller is responsible for ensuring that @pos remains accessible until
168 * the start of the next iteration by, for example, bumping the css refcnt.
5c9d535b 169 */
c326aa2b
TH
170#define css_for_each_child(pos, parent) \
171 for ((pos) = css_next_child(NULL, (parent)); (pos); \
172 (pos) = css_next_child((pos), (parent)))
ddbcc7e8 173
c326aa2b
TH
174/**
175 * css_for_each_descendant_pre - pre-order walk of a css's descendants
176 * @pos: the css * to use as the loop cursor
177 * @root: css whose descendants to walk
178 *
179 * Walk @root's descendants. @root is included in the iteration and the
180 * first node to be visited. Must be called under rcu_read_lock().
181 *
182 * If a subsystem synchronizes ->css_online() and the start of iteration, a
183 * css which finished ->css_online() is guaranteed to be visible in the
184 * future iterations and will stay visible until the last reference is put.
185 * A css which hasn't finished ->css_online() or already finished
186 * ->css_offline() may show up during traversal. It's each subsystem's
187 * responsibility to synchronize against on/offlining.
188 *
189 * For example, the following guarantees that a descendant can't escape
190 * state updates of its ancestors.
191 *
192 * my_online(@css)
193 * {
194 * Lock @css's parent and @css;
195 * Inherit state from the parent;
196 * Unlock both.
197 * }
198 *
199 * my_update_state(@css)
200 * {
201 * css_for_each_descendant_pre(@pos, @css) {
202 * Lock @pos;
203 * if (@pos == @css)
204 * Update @css's state;
205 * else
206 * Verify @pos is alive and inherit state from its parent;
207 * Unlock @pos;
208 * }
209 * }
210 *
211 * As long as the inheriting step, including checking the parent state, is
212 * enclosed inside @pos locking, double-locking the parent isn't necessary
213 * while inheriting. The state update to the parent is guaranteed to be
214 * visible by walking order and, as long as inheriting operations to the
215 * same @pos are atomic to each other, multiple updates racing each other
216 * still result in the correct state. It's guaranateed that at least one
217 * inheritance happens for any css after the latest update to its parent.
218 *
219 * If checking parent's state requires locking the parent, each inheriting
220 * iteration should lock and unlock both @pos->parent and @pos.
221 *
222 * Alternatively, a subsystem may choose to use a single global lock to
223 * synchronize ->css_online() and ->css_offline() against tree-walking
224 * operations.
225 *
226 * It is allowed to temporarily drop RCU read lock during iteration. The
227 * caller is responsible for ensuring that @pos remains accessible until
228 * the start of the next iteration by, for example, bumping the css refcnt.
229 */
230#define css_for_each_descendant_pre(pos, css) \
231 for ((pos) = css_next_descendant_pre(NULL, (css)); (pos); \
232 (pos) = css_next_descendant_pre((pos), (css)))
233
234/**
235 * css_for_each_descendant_post - post-order walk of a css's descendants
236 * @pos: the css * to use as the loop cursor
237 * @css: css whose descendants to walk
238 *
239 * Similar to css_for_each_descendant_pre() but performs post-order
240 * traversal instead. @root is included in the iteration and the last
241 * node to be visited.
242 *
243 * If a subsystem synchronizes ->css_online() and the start of iteration, a
244 * css which finished ->css_online() is guaranteed to be visible in the
245 * future iterations and will stay visible until the last reference is put.
246 * A css which hasn't finished ->css_online() or already finished
247 * ->css_offline() may show up during traversal. It's each subsystem's
248 * responsibility to synchronize against on/offlining.
249 *
250 * Note that the walk visibility guarantee example described in pre-order
251 * walk doesn't apply the same to post-order walks.
252 */
253#define css_for_each_descendant_post(pos, css) \
254 for ((pos) = css_next_descendant_post(NULL, (css)); (pos); \
255 (pos) = css_next_descendant_post((pos), (css)))
256
257/**
258 * cgroup_taskset_for_each - iterate cgroup_taskset
259 * @task: the loop cursor
1f7dd3e5 260 * @dst_css: the destination css
c326aa2b 261 * @tset: taskset to iterate
4530eddb
TH
262 *
263 * @tset may contain multiple tasks and they may belong to multiple
1f7dd3e5
TH
264 * processes.
265 *
266 * On the v2 hierarchy, there may be tasks from multiple processes and they
267 * may not share the source or destination csses.
268 *
269 * On traditional hierarchies, when there are multiple tasks in @tset, if a
270 * task of a process is in @tset, all tasks of the process are in @tset.
271 * Also, all are guaranteed to share the same source and destination csses.
4530eddb
TH
272 *
273 * Iteration is not in any specific order.
c326aa2b 274 */
1f7dd3e5
TH
275#define cgroup_taskset_for_each(task, dst_css, tset) \
276 for ((task) = cgroup_taskset_first((tset), &(dst_css)); \
277 (task); \
278 (task) = cgroup_taskset_next((tset), &(dst_css)))
ddbcc7e8 279
4530eddb
TH
280/**
281 * cgroup_taskset_for_each_leader - iterate group leaders in a cgroup_taskset
282 * @leader: the loop cursor
1f7dd3e5 283 * @dst_css: the destination css
7b4632f0 284 * @tset: taskset to iterate
4530eddb
TH
285 *
286 * Iterate threadgroup leaders of @tset. For single-task migrations, @tset
287 * may not contain any.
288 */
1f7dd3e5
TH
289#define cgroup_taskset_for_each_leader(leader, dst_css, tset) \
290 for ((leader) = cgroup_taskset_first((tset), &(dst_css)); \
291 (leader); \
292 (leader) = cgroup_taskset_next((tset), &(dst_css))) \
4530eddb
TH
293 if ((leader) != (leader)->group_leader) \
294 ; \
295 else
296
c326aa2b
TH
297/*
298 * Inline functions.
299 */
ddbcc7e8 300
5de0107e
TH
301/**
302 * css_get - obtain a reference on the specified css
303 * @css: target css
304 *
305 * The caller must already have a reference.
ddbcc7e8 306 */
ddbcc7e8
PM
307static inline void css_get(struct cgroup_subsys_state *css)
308{
3b514d24
TH
309 if (!(css->flags & CSS_NO_REF))
310 percpu_ref_get(&css->refcnt);
ddbcc7e8 311}
e7c5ec91 312
e8ea14cc
JW
313/**
314 * css_get_many - obtain references on the specified css
315 * @css: target css
316 * @n: number of references to get
317 *
318 * The caller must already have a reference.
319 */
320static inline void css_get_many(struct cgroup_subsys_state *css, unsigned int n)
321{
322 if (!(css->flags & CSS_NO_REF))
323 percpu_ref_get_many(&css->refcnt, n);
324}
325
6f4524d3
TH
326/**
327 * css_tryget - try to obtain a reference on the specified css
328 * @css: target css
329 *
330 * Obtain a reference on @css unless it already has reached zero and is
331 * being released. This function doesn't care whether @css is on or
332 * offline. The caller naturally needs to ensure that @css is accessible
333 * but doesn't have to be holding a reference on it - IOW, RCU protected
334 * access is good enough for this function. Returns %true if a reference
335 * count was successfully obtained; %false otherwise.
336 */
337static inline bool css_tryget(struct cgroup_subsys_state *css)
338{
339 if (!(css->flags & CSS_NO_REF))
340 return percpu_ref_tryget(&css->refcnt);
341 return true;
342}
343
5de0107e 344/**
ec903c0c 345 * css_tryget_online - try to obtain a reference on the specified css if online
5de0107e
TH
346 * @css: target css
347 *
ec903c0c
TH
348 * Obtain a reference on @css if it's online. The caller naturally needs
349 * to ensure that @css is accessible but doesn't have to be holding a
5de0107e
TH
350 * reference on it - IOW, RCU protected access is good enough for this
351 * function. Returns %true if a reference count was successfully obtained;
352 * %false otherwise.
353 */
ec903c0c 354static inline bool css_tryget_online(struct cgroup_subsys_state *css)
e7c5ec91 355{
3b514d24
TH
356 if (!(css->flags & CSS_NO_REF))
357 return percpu_ref_tryget_live(&css->refcnt);
358 return true;
e7c5ec91 359}
2f7ee569 360
41c25707
TH
361/**
362 * css_is_dying - test whether the specified css is dying
363 * @css: target css
364 *
365 * Test whether @css is in the process of offlining or already offline. In
366 * most cases, ->css_online() and ->css_offline() callbacks should be
367 * enough; however, the actual offline operations are RCU delayed and this
368 * test returns %true also when @css is scheduled to be offlined.
369 *
370 * This is useful, for example, when the use case requires synchronous
371 * behavior with respect to cgroup removal. cgroup removal schedules css
372 * offlining but the css can seem alive while the operation is being
373 * delayed. If the delay affects user visible semantics, this test can be
374 * used to resolve the situation.
375 */
376static inline bool css_is_dying(struct cgroup_subsys_state *css)
377{
378 return !(css->flags & CSS_NO_REF) && percpu_ref_is_dying(&css->refcnt);
379}
380
2f7ee569 381/**
5de0107e
TH
382 * css_put - put a css reference
383 * @css: target css
384 *
ec903c0c 385 * Put a reference obtained via css_get() and css_tryget_online().
2f7ee569 386 */
ddbcc7e8
PM
387static inline void css_put(struct cgroup_subsys_state *css)
388{
3b514d24
TH
389 if (!(css->flags & CSS_NO_REF))
390 percpu_ref_put(&css->refcnt);
ddbcc7e8 391}
2f7ee569 392
e8ea14cc
JW
393/**
394 * css_put_many - put css references
395 * @css: target css
396 * @n: number of references to put
397 *
398 * Put references obtained via css_get() and css_tryget_online().
21acb9ca 399 */
e8ea14cc
JW
400static inline void css_put_many(struct cgroup_subsys_state *css, unsigned int n)
401{
402 if (!(css->flags & CSS_NO_REF))
403 percpu_ref_put_many(&css->refcnt, n);
404}
ddbcc7e8 405
3e48930c
TH
406static inline void cgroup_get(struct cgroup *cgrp)
407{
408 css_get(&cgrp->self);
409}
410
411static inline bool cgroup_tryget(struct cgroup *cgrp)
412{
413 return css_tryget(&cgrp->self);
414}
415
16af4396
TH
416static inline void cgroup_put(struct cgroup *cgrp)
417{
418 css_put(&cgrp->self);
419}
420
14611e51
TH
421/**
422 * task_css_set_check - obtain a task's css_set with extra access conditions
423 * @task: the task to obtain css_set for
424 * @__c: extra condition expression to be passed to rcu_dereference_check()
425 *
426 * A task's css_set is RCU protected, initialized and exited while holding
427 * task_lock(), and can only be modified while holding both cgroup_mutex
428 * and task_lock() while the task is alive. This macro verifies that the
429 * caller is inside proper critical section and returns @task's css_set.
430 *
431 * The caller can also specify additional allowed conditions via @__c, such
432 * as locks used during the cgroup_subsys::attach() methods.
dc61b1d6 433 */
2219449a
TH
434#ifdef CONFIG_PROVE_RCU
435extern struct mutex cgroup_mutex;
f0d9a5f1 436extern spinlock_t css_set_lock;
14611e51
TH
437#define task_css_set_check(task, __c) \
438 rcu_dereference_check((task)->cgroups, \
0e1d768f 439 lockdep_is_held(&cgroup_mutex) || \
f0d9a5f1 440 lockdep_is_held(&css_set_lock) || \
0e1d768f 441 ((task)->flags & PF_EXITING) || (__c))
2219449a 442#else
14611e51
TH
443#define task_css_set_check(task, __c) \
444 rcu_dereference((task)->cgroups)
2219449a 445#endif
dc61b1d6 446
14611e51 447/**
8af01f56 448 * task_css_check - obtain css for (task, subsys) w/ extra access conds
14611e51
TH
449 * @task: the target task
450 * @subsys_id: the target subsystem ID
451 * @__c: extra condition expression to be passed to rcu_dereference_check()
452 *
453 * Return the cgroup_subsys_state for the (@task, @subsys_id) pair. The
454 * synchronization rules are the same as task_css_set_check().
455 */
8af01f56 456#define task_css_check(task, subsys_id, __c) \
14611e51
TH
457 task_css_set_check((task), (__c))->subsys[(subsys_id)]
458
459/**
460 * task_css_set - obtain a task's css_set
461 * @task: the task to obtain css_set for
462 *
463 * See task_css_set_check().
464 */
465static inline struct css_set *task_css_set(struct task_struct *task)
466{
467 return task_css_set_check(task, false);
468}
469
470/**
8af01f56 471 * task_css - obtain css for (task, subsys)
14611e51
TH
472 * @task: the target task
473 * @subsys_id: the target subsystem ID
474 *
8af01f56 475 * See task_css_check().
14611e51 476 */
8af01f56
TH
477static inline struct cgroup_subsys_state *task_css(struct task_struct *task,
478 int subsys_id)
ddbcc7e8 479{
8af01f56 480 return task_css_check(task, subsys_id, false);
ddbcc7e8
PM
481}
482
ec438699
TH
483/**
484 * task_get_css - find and get the css for (task, subsys)
485 * @task: the target task
486 * @subsys_id: the target subsystem ID
487 *
488 * Find the css for the (@task, @subsys_id) combination, increment a
489 * reference on and return it. This function is guaranteed to return a
097b73d4 490 * valid css. The returned css may already have been offlined.
ec438699
TH
491 */
492static inline struct cgroup_subsys_state *
493task_get_css(struct task_struct *task, int subsys_id)
494{
495 struct cgroup_subsys_state *css;
496
497 rcu_read_lock();
498 while (true) {
499 css = task_css(task, subsys_id);
097b73d4
TH
500 /*
501 * Can't use css_tryget_online() here. A task which has
502 * PF_EXITING set may stay associated with an offline css.
503 * If such task calls this function, css_tryget_online()
504 * will keep failing.
505 */
506 if (likely(css_tryget(css)))
ec438699
TH
507 break;
508 cpu_relax();
509 }
510 rcu_read_unlock();
511 return css;
512}
513
5024ae29
TH
514/**
515 * task_css_is_root - test whether a task belongs to the root css
516 * @task: the target task
517 * @subsys_id: the target subsystem ID
518 *
519 * Test whether @task belongs to the root css on the specified subsystem.
520 * May be invoked in any context.
521 */
522static inline bool task_css_is_root(struct task_struct *task, int subsys_id)
523{
524 return task_css_check(task, subsys_id, true) ==
525 init_css_set.subsys[subsys_id];
526}
527
8af01f56
TH
528static inline struct cgroup *task_cgroup(struct task_struct *task,
529 int subsys_id)
ddbcc7e8 530{
8af01f56 531 return task_css(task, subsys_id)->cgroup;
ddbcc7e8
PM
532}
533
3e48930c
TH
534static inline struct cgroup *task_dfl_cgroup(struct task_struct *task)
535{
536 return task_css_set(task)->dfl_cgrp;
537}
538
539static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
540{
541 struct cgroup_subsys_state *parent_css = cgrp->self.parent;
542
543 if (parent_css)
544 return container_of(parent_css, struct cgroup, self);
545 return NULL;
546}
547
b11cfb58
TH
548/**
549 * cgroup_is_descendant - test ancestry
550 * @cgrp: the cgroup to be tested
551 * @ancestor: possible ancestor of @cgrp
552 *
553 * Test whether @cgrp is a descendant of @ancestor. It also returns %true
554 * if @cgrp == @ancestor. This function is safe to call as long as @cgrp
555 * and @ancestor are accessible.
556 */
557static inline bool cgroup_is_descendant(struct cgroup *cgrp,
558 struct cgroup *ancestor)
559{
560 if (cgrp->root != ancestor->root || cgrp->level < ancestor->level)
561 return false;
562 return cgrp->ancestor_ids[ancestor->level] == ancestor->id;
563}
564
77236281
AI
565/**
566 * cgroup_ancestor - find ancestor of cgroup
567 * @cgrp: cgroup to find ancestor of
568 * @ancestor_level: level of ancestor to find starting from root
569 *
570 * Find ancestor of cgroup at specified level starting from root if it exists
571 * and return pointer to it. Return NULL if @cgrp doesn't have ancestor at
572 * @ancestor_level.
573 *
574 * This function is safe to call as long as @cgrp is accessible.
575 */
576static inline struct cgroup *cgroup_ancestor(struct cgroup *cgrp,
577 int ancestor_level)
578{
77236281
AI
579 if (cgrp->level < ancestor_level)
580 return NULL;
808c43b7
AI
581 while (cgrp && cgrp->level > ancestor_level)
582 cgrp = cgroup_parent(cgrp);
583 return cgrp;
77236281
AI
584}
585
aed704b7
SD
586/**
587 * task_under_cgroup_hierarchy - test task's membership of cgroup ancestry
588 * @task: the task to be tested
589 * @ancestor: possible ancestor of @task's cgroup
590 *
591 * Tests whether @task's default cgroup hierarchy is a descendant of @ancestor.
592 * It follows all the same rules as cgroup_is_descendant, and only applies
593 * to the default hierarchy.
594 */
595static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
596 struct cgroup *ancestor)
597{
598 struct css_set *cset = task_css_set(task);
599
600 return cgroup_is_descendant(cset->dfl_cgrp, ancestor);
601}
602
07bc356e 603/* no synchronization, the result can only be used as a hint */
27bd4dbb 604static inline bool cgroup_is_populated(struct cgroup *cgrp)
07bc356e 605{
454000ad
TH
606 return cgrp->nr_populated_csets + cgrp->nr_populated_domain_children +
607 cgrp->nr_populated_threaded_children;
07bc356e 608}
f3d46500 609
f29374b1 610/* returns ino associated with a cgroup */
b1664924
TH
611static inline ino_t cgroup_ino(struct cgroup *cgrp)
612{
c53cd490 613 return cgrp->kn->id.ino;
b1664924 614}
3ebb2b6e 615
b4168640
TH
616/* cft/css accessors for cftype->write() operation */
617static inline struct cftype *of_cft(struct kernfs_open_file *of)
7da11279 618{
2bd59d48 619 return of->kn->priv;
7da11279 620}
0f0a2b4f 621
b4168640 622struct cgroup_subsys_state *of_css(struct kernfs_open_file *of);
817929ec 623
b4168640
TH
624/* cft/css accessors for cftype->seq_*() operations */
625static inline struct cftype *seq_cft(struct seq_file *seq)
626{
627 return of_cft(seq->private);
628}
e535837b 629
b4168640
TH
630static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq)
631{
632 return of_css(seq->private);
633}
31583bb0 634
e61734c5
TH
635/*
636 * Name / path handling functions. All are thin wrappers around the kernfs
637 * counterparts and can be called under any context.
638 */
38460b48 639
e61734c5
TH
640static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen)
641{
fdce6bf8 642 return kernfs_name(cgrp->kn, buf, buflen);
e61734c5 643}
ddbcc7e8 644
4c737b41 645static inline int cgroup_path(struct cgroup *cgrp, char *buf, size_t buflen)
e61734c5 646{
fdce6bf8 647 return kernfs_path(cgrp->kn, buf, buflen);
e61734c5 648}
f3ba5380 649
e61734c5
TH
650static inline void pr_cont_cgroup_name(struct cgroup *cgrp)
651{
fdce6bf8 652 pr_cont_kernfs_name(cgrp->kn);
e61734c5 653}
ddbcc7e8 654
e61734c5 655static inline void pr_cont_cgroup_path(struct cgroup *cgrp)
846c7bb0 656{
fdce6bf8 657 pr_cont_kernfs_path(cgrp->kn);
846c7bb0 658}
ddbcc7e8 659
2ce7135a
JW
660static inline struct psi_group *cgroup_psi(struct cgroup *cgrp)
661{
662 return &cgrp->psi;
663}
664
77f88796
TH
665static inline void cgroup_init_kthreadd(void)
666{
667 /*
668 * kthreadd is inherited by all kthreads, keep it in the root so
669 * that the new kthreads are guaranteed to stay in the root until
670 * initialization is finished.
671 */
672 current->no_cgroup_migration = 1;
673}
674
675static inline void cgroup_kthread_ready(void)
676{
677 /*
678 * This kthread finished initialization. The creator should have
679 * set PF_NO_SETAFFINITY if this kthread should stay in the root.
680 */
681 current->no_cgroup_migration = 0;
682}
683
121508df
SL
684static inline union kernfs_node_id *cgroup_get_kernfs_id(struct cgroup *cgrp)
685{
686 return &cgrp->kn->id;
687}
69fd5c39
SL
688
689void cgroup_path_from_kernfs_id(const union kernfs_node_id *id,
690 char *buf, size_t buflen);
ddbcc7e8
PM
691#else /* !CONFIG_CGROUPS */
692
f3ba5380 693struct cgroup_subsys_state;
aed704b7 694struct cgroup;
f3ba5380 695
c326aa2b 696static inline void css_put(struct cgroup_subsys_state *css) {}
31583bb0 697static inline int cgroup_attach_task_all(struct task_struct *from,
c326aa2b
TH
698 struct task_struct *t) { return 0; }
699static inline int cgroupstats_build(struct cgroupstats *stats,
700 struct dentry *dentry) { return -EINVAL; }
701
b4f48b63 702static inline void cgroup_fork(struct task_struct *p) {}
b53202e6
ON
703static inline int cgroup_can_fork(struct task_struct *p) { return 0; }
704static inline void cgroup_cancel_fork(struct task_struct *p) {}
705static inline void cgroup_post_fork(struct task_struct *p) {}
1ec41830 706static inline void cgroup_exit(struct task_struct *p) {}
51bee5ab 707static inline void cgroup_release(struct task_struct *p) {}
2e91fa7f 708static inline void cgroup_free(struct task_struct *p) {}
ddbcc7e8 709
c326aa2b
TH
710static inline int cgroup_init_early(void) { return 0; }
711static inline int cgroup_init(void) { return 0; }
77f88796
TH
712static inline void cgroup_init_kthreadd(void) {}
713static inline void cgroup_kthread_ready(void) {}
121508df
SL
714static inline union kernfs_node_id *cgroup_get_kernfs_id(struct cgroup *cgrp)
715{
716 return NULL;
717}
d7926ee3 718
2ce7135a
JW
719static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
720{
721 return NULL;
722}
723
724static inline struct psi_group *cgroup_psi(struct cgroup *cgrp)
725{
726 return NULL;
727}
728
aed704b7
SD
729static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
730 struct cgroup *ancestor)
731{
732 return true;
733}
69fd5c39
SL
734
735static inline void cgroup_path_from_kernfs_id(const union kernfs_node_id *id,
736 char *buf, size_t buflen) {}
ddbcc7e8
PM
737#endif /* !CONFIG_CGROUPS */
738
6162cef0 739#ifdef CONFIG_CGROUPS
d2cc5ed6 740/*
6162cef0 741 * cgroup scalable recursive statistics.
d2cc5ed6 742 */
6162cef0
TH
743void cgroup_rstat_updated(struct cgroup *cgrp, int cpu);
744void cgroup_rstat_flush(struct cgroup *cgrp);
0fa294fb 745void cgroup_rstat_flush_irqsafe(struct cgroup *cgrp);
6162cef0
TH
746void cgroup_rstat_flush_hold(struct cgroup *cgrp);
747void cgroup_rstat_flush_release(void);
d2cc5ed6 748
6162cef0
TH
749/*
750 * Basic resource stats.
751 */
d2cc5ed6
TH
752#ifdef CONFIG_CGROUP_CPUACCT
753void cpuacct_charge(struct task_struct *tsk, u64 cputime);
754void cpuacct_account_field(struct task_struct *tsk, int index, u64 val);
755#else
756static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {}
757static inline void cpuacct_account_field(struct task_struct *tsk, int index,
758 u64 val) {}
759#endif
760
041cd640
TH
761void __cgroup_account_cputime(struct cgroup *cgrp, u64 delta_exec);
762void __cgroup_account_cputime_field(struct cgroup *cgrp,
763 enum cpu_usage_stat index, u64 delta_exec);
764
d2cc5ed6
TH
765static inline void cgroup_account_cputime(struct task_struct *task,
766 u64 delta_exec)
767{
041cd640
TH
768 struct cgroup *cgrp;
769
d2cc5ed6 770 cpuacct_charge(task, delta_exec);
041cd640
TH
771
772 rcu_read_lock();
773 cgrp = task_dfl_cgroup(task);
774 if (cgroup_parent(cgrp))
775 __cgroup_account_cputime(cgrp, delta_exec);
776 rcu_read_unlock();
d2cc5ed6
TH
777}
778
779static inline void cgroup_account_cputime_field(struct task_struct *task,
780 enum cpu_usage_stat index,
781 u64 delta_exec)
782{
041cd640
TH
783 struct cgroup *cgrp;
784
d2cc5ed6 785 cpuacct_account_field(task, index, delta_exec);
041cd640
TH
786
787 rcu_read_lock();
788 cgrp = task_dfl_cgroup(task);
789 if (cgroup_parent(cgrp))
790 __cgroup_account_cputime_field(cgrp, index, delta_exec);
791 rcu_read_unlock();
d2cc5ed6
TH
792}
793
794#else /* CONFIG_CGROUPS */
795
796static inline void cgroup_account_cputime(struct task_struct *task,
797 u64 delta_exec) {}
798static inline void cgroup_account_cputime_field(struct task_struct *task,
799 enum cpu_usage_stat index,
800 u64 delta_exec) {}
801
802#endif /* CONFIG_CGROUPS */
803
bd1060a1
TH
804/*
805 * sock->sk_cgrp_data handling. For more info, see sock_cgroup_data
806 * definition in cgroup-defs.h.
807 */
808#ifdef CONFIG_SOCK_CGROUP_DATA
809
810#if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
811extern spinlock_t cgroup_sk_update_lock;
812#endif
813
814void cgroup_sk_alloc_disable(void);
815void cgroup_sk_alloc(struct sock_cgroup_data *skcd);
816void cgroup_sk_free(struct sock_cgroup_data *skcd);
817
818static inline struct cgroup *sock_cgroup_ptr(struct sock_cgroup_data *skcd)
819{
820#if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
821 unsigned long v;
822
823 /*
824 * @skcd->val is 64bit but the following is safe on 32bit too as we
825 * just need the lower ulong to be written and read atomically.
826 */
827 v = READ_ONCE(skcd->val);
828
829 if (v & 1)
830 return &cgrp_dfl_root.cgrp;
831
832 return (struct cgroup *)(unsigned long)v ?: &cgrp_dfl_root.cgrp;
833#else
834 return (struct cgroup *)(unsigned long)skcd->val;
835#endif
836}
837
838#else /* CONFIG_CGROUP_DATA */
839
840static inline void cgroup_sk_alloc(struct sock_cgroup_data *skcd) {}
841static inline void cgroup_sk_free(struct sock_cgroup_data *skcd) {}
842
843#endif /* CONFIG_CGROUP_DATA */
844
a79a908f 845struct cgroup_namespace {
387ad967 846 refcount_t count;
a79a908f
AK
847 struct ns_common ns;
848 struct user_namespace *user_ns;
d08311dd 849 struct ucounts *ucounts;
a79a908f
AK
850 struct css_set *root_cset;
851};
852
853extern struct cgroup_namespace init_cgroup_ns;
854
855#ifdef CONFIG_CGROUPS
856
857void free_cgroup_ns(struct cgroup_namespace *ns);
858
859struct cgroup_namespace *copy_cgroup_ns(unsigned long flags,
860 struct user_namespace *user_ns,
861 struct cgroup_namespace *old_ns);
862
4c737b41
TH
863int cgroup_path_ns(struct cgroup *cgrp, char *buf, size_t buflen,
864 struct cgroup_namespace *ns);
a79a908f
AK
865
866#else /* !CONFIG_CGROUPS */
867
868static inline void free_cgroup_ns(struct cgroup_namespace *ns) { }
869static inline struct cgroup_namespace *
870copy_cgroup_ns(unsigned long flags, struct user_namespace *user_ns,
871 struct cgroup_namespace *old_ns)
872{
873 return old_ns;
874}
875
876#endif /* !CONFIG_CGROUPS */
877
878static inline void get_cgroup_ns(struct cgroup_namespace *ns)
879{
880 if (ns)
387ad967 881 refcount_inc(&ns->count);
a79a908f
AK
882}
883
884static inline void put_cgroup_ns(struct cgroup_namespace *ns)
885{
387ad967 886 if (ns && refcount_dec_and_test(&ns->count))
a79a908f
AK
887 free_cgroup_ns(ns);
888}
889
ddbcc7e8 890#endif /* _LINUX_CGROUP_H */