]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - include/linux/cgroup.h
Merge tag 'dma-mapping-5.15-1' of git://git.infradead.org/users/hch/dma-mapping
[mirror_ubuntu-jammy-kernel.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 29
ef2c41cf
CB
30struct kernel_clone_args;
31
ddbcc7e8
PM
32#ifdef CONFIG_CGROUPS
33
6abc8ca1 34/*
08b2b6fd 35 * All weight knobs on the default hierarchy should use the following min,
6abc8ca1
TH
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
bc2fb7ed
TH
43/* walk only threadgroup leaders */
44#define CSS_TASK_ITER_PROCS (1U << 0)
450ee0c1
TH
45/* walk all threaded css_sets in the domain */
46#define CSS_TASK_ITER_THREADED (1U << 1)
bc2fb7ed 47
b636fd38
TH
48/* internal flags */
49#define CSS_TASK_ITER_SKIPPED (1U << 16)
50
c326aa2b
TH
51/* a css_task_iter should be treated as an opaque object */
52struct css_task_iter {
53 struct cgroup_subsys *ss;
bc2fb7ed 54 unsigned int flags;
c326aa2b
TH
55
56 struct list_head *cset_pos;
57 struct list_head *cset_head;
ddbcc7e8 58
450ee0c1
TH
59 struct list_head *tcset_pos;
60 struct list_head *tcset_head;
61
c326aa2b 62 struct list_head *task_pos;
ed27b9f7 63
9c974c77 64 struct list_head *cur_tasks_head;
ed27b9f7 65 struct css_set *cur_cset;
450ee0c1 66 struct css_set *cur_dcset;
ed27b9f7
TH
67 struct task_struct *cur_task;
68 struct list_head iters_node; /* css_set->task_iters */
c326aa2b 69};
ddbcc7e8 70
c326aa2b
TH
71extern struct cgroup_root cgrp_dfl_root;
72extern struct css_set init_css_set;
a424316c 73
c326aa2b 74#define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys;
817929ec 75#include <linux/cgroup_subsys.h>
817929ec
PM
76#undef SUBSYS
77
49d1dc4b
TH
78#define SUBSYS(_x) \
79 extern struct static_key_true _x ## _cgrp_subsys_enabled_key; \
80 extern struct static_key_true _x ## _cgrp_subsys_on_dfl_key;
81#include <linux/cgroup_subsys.h>
82#undef SUBSYS
83
84/**
85 * cgroup_subsys_enabled - fast test on whether a subsys is enabled
86 * @ss: subsystem in question
87 */
88#define cgroup_subsys_enabled(ss) \
89 static_branch_likely(&ss ## _enabled_key)
90
91/**
92 * cgroup_subsys_on_dfl - fast test on whether a subsys is on default hierarchy
93 * @ss: subsystem in question
94 */
95#define cgroup_subsys_on_dfl(ss) \
96 static_branch_likely(&ss ## _on_dfl_key)
97
c326aa2b
TH
98bool css_has_online_children(struct cgroup_subsys_state *css);
99struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss);
fc5a828b
DZ
100struct cgroup_subsys_state *cgroup_e_css(struct cgroup *cgroup,
101 struct cgroup_subsys *ss);
c326aa2b
TH
102struct cgroup_subsys_state *cgroup_get_e_css(struct cgroup *cgroup,
103 struct cgroup_subsys *ss);
104struct cgroup_subsys_state *css_tryget_online_from_dir(struct dentry *dentry,
105 struct cgroup_subsys *ss);
106
16af4396 107struct cgroup *cgroup_get_from_path(const char *path);
1f3fe7eb 108struct cgroup *cgroup_get_from_fd(int fd);
16af4396 109
c326aa2b
TH
110int cgroup_attach_task_all(struct task_struct *from, struct task_struct *);
111int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from);
112
113int cgroup_add_dfl_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
114int cgroup_add_legacy_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
115int cgroup_rm_cftypes(struct cftype *cfts);
34c06254 116void cgroup_file_notify(struct cgroup_file *cfile);
c326aa2b 117
4c737b41 118int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen);
c326aa2b
TH
119int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry);
120int proc_cgroup_show(struct seq_file *m, struct pid_namespace *ns,
121 struct pid *pid, struct task_struct *tsk);
122
123void cgroup_fork(struct task_struct *p);
ef2c41cf
CB
124extern int cgroup_can_fork(struct task_struct *p,
125 struct kernel_clone_args *kargs);
126extern void cgroup_cancel_fork(struct task_struct *p,
127 struct kernel_clone_args *kargs);
128extern void cgroup_post_fork(struct task_struct *p,
129 struct kernel_clone_args *kargs);
c326aa2b 130void cgroup_exit(struct task_struct *p);
51bee5ab 131void cgroup_release(struct task_struct *p);
2e91fa7f 132void cgroup_free(struct task_struct *p);
c326aa2b
TH
133
134int cgroup_init_early(void);
135int cgroup_init(void);
136
a5e112e6
TH
137int cgroup_parse_float(const char *input, unsigned dec_shift, s64 *v);
138
5c9d535b 139/*
c326aa2b
TH
140 * Iteration helpers and macros.
141 */
142
143struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos,
144 struct cgroup_subsys_state *parent);
145struct cgroup_subsys_state *css_next_descendant_pre(struct cgroup_subsys_state *pos,
146 struct cgroup_subsys_state *css);
147struct cgroup_subsys_state *css_rightmost_descendant(struct cgroup_subsys_state *pos);
148struct cgroup_subsys_state *css_next_descendant_post(struct cgroup_subsys_state *pos,
149 struct cgroup_subsys_state *css);
150
1f7dd3e5
TH
151struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset,
152 struct cgroup_subsys_state **dst_cssp);
153struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset,
154 struct cgroup_subsys_state **dst_cssp);
c326aa2b 155
bc2fb7ed 156void css_task_iter_start(struct cgroup_subsys_state *css, unsigned int flags,
c326aa2b
TH
157 struct css_task_iter *it);
158struct task_struct *css_task_iter_next(struct css_task_iter *it);
159void css_task_iter_end(struct css_task_iter *it);
160
161/**
162 * css_for_each_child - iterate through children of a css
163 * @pos: the css * to use as the loop cursor
164 * @parent: css whose children to walk
165 *
166 * Walk @parent's children. Must be called under rcu_read_lock().
167 *
168 * If a subsystem synchronizes ->css_online() and the start of iteration, a
169 * css which finished ->css_online() is guaranteed to be visible in the
170 * future iterations and will stay visible until the last reference is put.
171 * A css which hasn't finished ->css_online() or already finished
172 * ->css_offline() may show up during traversal. It's each subsystem's
173 * responsibility to synchronize against on/offlining.
5c9d535b 174 *
c326aa2b
TH
175 * It is allowed to temporarily drop RCU read lock during iteration. The
176 * caller is responsible for ensuring that @pos remains accessible until
177 * the start of the next iteration by, for example, bumping the css refcnt.
5c9d535b 178 */
c326aa2b
TH
179#define css_for_each_child(pos, parent) \
180 for ((pos) = css_next_child(NULL, (parent)); (pos); \
181 (pos) = css_next_child((pos), (parent)))
ddbcc7e8 182
c326aa2b
TH
183/**
184 * css_for_each_descendant_pre - pre-order walk of a css's descendants
185 * @pos: the css * to use as the loop cursor
186 * @root: css whose descendants to walk
187 *
188 * Walk @root's descendants. @root is included in the iteration and the
189 * first node to be visited. Must be called under rcu_read_lock().
190 *
191 * If a subsystem synchronizes ->css_online() and the start of iteration, a
192 * css which finished ->css_online() is guaranteed to be visible in the
193 * future iterations and will stay visible until the last reference is put.
194 * A css which hasn't finished ->css_online() or already finished
195 * ->css_offline() may show up during traversal. It's each subsystem's
196 * responsibility to synchronize against on/offlining.
197 *
198 * For example, the following guarantees that a descendant can't escape
199 * state updates of its ancestors.
200 *
201 * my_online(@css)
202 * {
203 * Lock @css's parent and @css;
204 * Inherit state from the parent;
205 * Unlock both.
206 * }
207 *
208 * my_update_state(@css)
209 * {
210 * css_for_each_descendant_pre(@pos, @css) {
211 * Lock @pos;
212 * if (@pos == @css)
213 * Update @css's state;
214 * else
215 * Verify @pos is alive and inherit state from its parent;
216 * Unlock @pos;
217 * }
218 * }
219 *
220 * As long as the inheriting step, including checking the parent state, is
221 * enclosed inside @pos locking, double-locking the parent isn't necessary
222 * while inheriting. The state update to the parent is guaranteed to be
223 * visible by walking order and, as long as inheriting operations to the
224 * same @pos are atomic to each other, multiple updates racing each other
225 * still result in the correct state. It's guaranateed that at least one
226 * inheritance happens for any css after the latest update to its parent.
227 *
228 * If checking parent's state requires locking the parent, each inheriting
229 * iteration should lock and unlock both @pos->parent and @pos.
230 *
231 * Alternatively, a subsystem may choose to use a single global lock to
232 * synchronize ->css_online() and ->css_offline() against tree-walking
233 * operations.
234 *
235 * It is allowed to temporarily drop RCU read lock during iteration. The
236 * caller is responsible for ensuring that @pos remains accessible until
237 * the start of the next iteration by, for example, bumping the css refcnt.
238 */
239#define css_for_each_descendant_pre(pos, css) \
240 for ((pos) = css_next_descendant_pre(NULL, (css)); (pos); \
241 (pos) = css_next_descendant_pre((pos), (css)))
242
243/**
244 * css_for_each_descendant_post - post-order walk of a css's descendants
245 * @pos: the css * to use as the loop cursor
246 * @css: css whose descendants to walk
247 *
248 * Similar to css_for_each_descendant_pre() but performs post-order
249 * traversal instead. @root is included in the iteration and the last
250 * node to be visited.
251 *
252 * If a subsystem synchronizes ->css_online() and the start of iteration, a
253 * css which finished ->css_online() is guaranteed to be visible in the
254 * future iterations and will stay visible until the last reference is put.
255 * A css which hasn't finished ->css_online() or already finished
256 * ->css_offline() may show up during traversal. It's each subsystem's
257 * responsibility to synchronize against on/offlining.
258 *
259 * Note that the walk visibility guarantee example described in pre-order
260 * walk doesn't apply the same to post-order walks.
261 */
262#define css_for_each_descendant_post(pos, css) \
263 for ((pos) = css_next_descendant_post(NULL, (css)); (pos); \
264 (pos) = css_next_descendant_post((pos), (css)))
265
266/**
267 * cgroup_taskset_for_each - iterate cgroup_taskset
268 * @task: the loop cursor
1f7dd3e5 269 * @dst_css: the destination css
c326aa2b 270 * @tset: taskset to iterate
4530eddb
TH
271 *
272 * @tset may contain multiple tasks and they may belong to multiple
1f7dd3e5
TH
273 * processes.
274 *
275 * On the v2 hierarchy, there may be tasks from multiple processes and they
276 * may not share the source or destination csses.
277 *
278 * On traditional hierarchies, when there are multiple tasks in @tset, if a
279 * task of a process is in @tset, all tasks of the process are in @tset.
280 * Also, all are guaranteed to share the same source and destination csses.
4530eddb
TH
281 *
282 * Iteration is not in any specific order.
c326aa2b 283 */
1f7dd3e5
TH
284#define cgroup_taskset_for_each(task, dst_css, tset) \
285 for ((task) = cgroup_taskset_first((tset), &(dst_css)); \
286 (task); \
287 (task) = cgroup_taskset_next((tset), &(dst_css)))
ddbcc7e8 288
4530eddb
TH
289/**
290 * cgroup_taskset_for_each_leader - iterate group leaders in a cgroup_taskset
291 * @leader: the loop cursor
1f7dd3e5 292 * @dst_css: the destination css
7b4632f0 293 * @tset: taskset to iterate
4530eddb
TH
294 *
295 * Iterate threadgroup leaders of @tset. For single-task migrations, @tset
296 * may not contain any.
297 */
1f7dd3e5
TH
298#define cgroup_taskset_for_each_leader(leader, dst_css, tset) \
299 for ((leader) = cgroup_taskset_first((tset), &(dst_css)); \
300 (leader); \
301 (leader) = cgroup_taskset_next((tset), &(dst_css))) \
4530eddb
TH
302 if ((leader) != (leader)->group_leader) \
303 ; \
304 else
305
c326aa2b
TH
306/*
307 * Inline functions.
308 */
ddbcc7e8 309
a3b89864 310static inline u64 cgroup_id(const struct cgroup *cgrp)
74321038
TH
311{
312 return cgrp->kn->id;
313}
314
5de0107e
TH
315/**
316 * css_get - obtain a reference on the specified css
317 * @css: target css
318 *
319 * The caller must already have a reference.
ddbcc7e8 320 */
ddbcc7e8
PM
321static inline void css_get(struct cgroup_subsys_state *css)
322{
3b514d24
TH
323 if (!(css->flags & CSS_NO_REF))
324 percpu_ref_get(&css->refcnt);
ddbcc7e8 325}
e7c5ec91 326
e8ea14cc
JW
327/**
328 * css_get_many - obtain references on the specified css
329 * @css: target css
330 * @n: number of references to get
331 *
332 * The caller must already have a reference.
333 */
334static inline void css_get_many(struct cgroup_subsys_state *css, unsigned int n)
335{
336 if (!(css->flags & CSS_NO_REF))
337 percpu_ref_get_many(&css->refcnt, n);
338}
339
6f4524d3
TH
340/**
341 * css_tryget - try to obtain a reference on the specified css
342 * @css: target css
343 *
344 * Obtain a reference on @css unless it already has reached zero and is
345 * being released. This function doesn't care whether @css is on or
346 * offline. The caller naturally needs to ensure that @css is accessible
347 * but doesn't have to be holding a reference on it - IOW, RCU protected
348 * access is good enough for this function. Returns %true if a reference
349 * count was successfully obtained; %false otherwise.
350 */
351static inline bool css_tryget(struct cgroup_subsys_state *css)
352{
353 if (!(css->flags & CSS_NO_REF))
354 return percpu_ref_tryget(&css->refcnt);
355 return true;
356}
357
5de0107e 358/**
ec903c0c 359 * css_tryget_online - try to obtain a reference on the specified css if online
5de0107e
TH
360 * @css: target css
361 *
ec903c0c
TH
362 * Obtain a reference on @css if it's online. The caller naturally needs
363 * to ensure that @css is accessible but doesn't have to be holding a
5de0107e
TH
364 * reference on it - IOW, RCU protected access is good enough for this
365 * function. Returns %true if a reference count was successfully obtained;
366 * %false otherwise.
367 */
ec903c0c 368static inline bool css_tryget_online(struct cgroup_subsys_state *css)
e7c5ec91 369{
3b514d24
TH
370 if (!(css->flags & CSS_NO_REF))
371 return percpu_ref_tryget_live(&css->refcnt);
372 return true;
e7c5ec91 373}
2f7ee569 374
41c25707
TH
375/**
376 * css_is_dying - test whether the specified css is dying
377 * @css: target css
378 *
379 * Test whether @css is in the process of offlining or already offline. In
380 * most cases, ->css_online() and ->css_offline() callbacks should be
381 * enough; however, the actual offline operations are RCU delayed and this
382 * test returns %true also when @css is scheduled to be offlined.
383 *
384 * This is useful, for example, when the use case requires synchronous
385 * behavior with respect to cgroup removal. cgroup removal schedules css
386 * offlining but the css can seem alive while the operation is being
387 * delayed. If the delay affects user visible semantics, this test can be
388 * used to resolve the situation.
389 */
390static inline bool css_is_dying(struct cgroup_subsys_state *css)
391{
392 return !(css->flags & CSS_NO_REF) && percpu_ref_is_dying(&css->refcnt);
393}
394
2f7ee569 395/**
5de0107e
TH
396 * css_put - put a css reference
397 * @css: target css
398 *
ec903c0c 399 * Put a reference obtained via css_get() and css_tryget_online().
2f7ee569 400 */
ddbcc7e8
PM
401static inline void css_put(struct cgroup_subsys_state *css)
402{
3b514d24
TH
403 if (!(css->flags & CSS_NO_REF))
404 percpu_ref_put(&css->refcnt);
ddbcc7e8 405}
2f7ee569 406
e8ea14cc
JW
407/**
408 * css_put_many - put css references
409 * @css: target css
410 * @n: number of references to put
411 *
412 * Put references obtained via css_get() and css_tryget_online().
21acb9ca 413 */
e8ea14cc
JW
414static inline void css_put_many(struct cgroup_subsys_state *css, unsigned int n)
415{
416 if (!(css->flags & CSS_NO_REF))
417 percpu_ref_put_many(&css->refcnt, n);
418}
ddbcc7e8 419
3e48930c
TH
420static inline void cgroup_get(struct cgroup *cgrp)
421{
422 css_get(&cgrp->self);
423}
424
425static inline bool cgroup_tryget(struct cgroup *cgrp)
426{
427 return css_tryget(&cgrp->self);
428}
429
16af4396
TH
430static inline void cgroup_put(struct cgroup *cgrp)
431{
432 css_put(&cgrp->self);
433}
434
14611e51
TH
435/**
436 * task_css_set_check - obtain a task's css_set with extra access conditions
437 * @task: the task to obtain css_set for
438 * @__c: extra condition expression to be passed to rcu_dereference_check()
439 *
440 * A task's css_set is RCU protected, initialized and exited while holding
441 * task_lock(), and can only be modified while holding both cgroup_mutex
442 * and task_lock() while the task is alive. This macro verifies that the
443 * caller is inside proper critical section and returns @task's css_set.
444 *
445 * The caller can also specify additional allowed conditions via @__c, such
446 * as locks used during the cgroup_subsys::attach() methods.
dc61b1d6 447 */
2219449a
TH
448#ifdef CONFIG_PROVE_RCU
449extern struct mutex cgroup_mutex;
f0d9a5f1 450extern spinlock_t css_set_lock;
14611e51
TH
451#define task_css_set_check(task, __c) \
452 rcu_dereference_check((task)->cgroups, \
0e1d768f 453 lockdep_is_held(&cgroup_mutex) || \
f0d9a5f1 454 lockdep_is_held(&css_set_lock) || \
0e1d768f 455 ((task)->flags & PF_EXITING) || (__c))
2219449a 456#else
14611e51
TH
457#define task_css_set_check(task, __c) \
458 rcu_dereference((task)->cgroups)
2219449a 459#endif
dc61b1d6 460
14611e51 461/**
8af01f56 462 * task_css_check - obtain css for (task, subsys) w/ extra access conds
14611e51
TH
463 * @task: the target task
464 * @subsys_id: the target subsystem ID
465 * @__c: extra condition expression to be passed to rcu_dereference_check()
466 *
467 * Return the cgroup_subsys_state for the (@task, @subsys_id) pair. The
468 * synchronization rules are the same as task_css_set_check().
469 */
8af01f56 470#define task_css_check(task, subsys_id, __c) \
14611e51
TH
471 task_css_set_check((task), (__c))->subsys[(subsys_id)]
472
473/**
474 * task_css_set - obtain a task's css_set
475 * @task: the task to obtain css_set for
476 *
477 * See task_css_set_check().
478 */
479static inline struct css_set *task_css_set(struct task_struct *task)
480{
481 return task_css_set_check(task, false);
482}
483
484/**
8af01f56 485 * task_css - obtain css for (task, subsys)
14611e51
TH
486 * @task: the target task
487 * @subsys_id: the target subsystem ID
488 *
8af01f56 489 * See task_css_check().
14611e51 490 */
8af01f56
TH
491static inline struct cgroup_subsys_state *task_css(struct task_struct *task,
492 int subsys_id)
ddbcc7e8 493{
8af01f56 494 return task_css_check(task, subsys_id, false);
ddbcc7e8
PM
495}
496
ec438699
TH
497/**
498 * task_get_css - find and get the css for (task, subsys)
499 * @task: the target task
500 * @subsys_id: the target subsystem ID
501 *
502 * Find the css for the (@task, @subsys_id) combination, increment a
503 * reference on and return it. This function is guaranteed to return a
18fa84a2 504 * valid css. The returned css may already have been offlined.
ec438699
TH
505 */
506static inline struct cgroup_subsys_state *
507task_get_css(struct task_struct *task, int subsys_id)
508{
509 struct cgroup_subsys_state *css;
510
511 rcu_read_lock();
512 while (true) {
513 css = task_css(task, subsys_id);
18fa84a2
TH
514 /*
515 * Can't use css_tryget_online() here. A task which has
516 * PF_EXITING set may stay associated with an offline css.
517 * If such task calls this function, css_tryget_online()
518 * will keep failing.
519 */
520 if (likely(css_tryget(css)))
ec438699
TH
521 break;
522 cpu_relax();
523 }
524 rcu_read_unlock();
525 return css;
526}
527
5024ae29
TH
528/**
529 * task_css_is_root - test whether a task belongs to the root css
530 * @task: the target task
531 * @subsys_id: the target subsystem ID
532 *
533 * Test whether @task belongs to the root css on the specified subsystem.
534 * May be invoked in any context.
535 */
536static inline bool task_css_is_root(struct task_struct *task, int subsys_id)
537{
538 return task_css_check(task, subsys_id, true) ==
539 init_css_set.subsys[subsys_id];
540}
541
8af01f56
TH
542static inline struct cgroup *task_cgroup(struct task_struct *task,
543 int subsys_id)
ddbcc7e8 544{
8af01f56 545 return task_css(task, subsys_id)->cgroup;
ddbcc7e8
PM
546}
547
3e48930c
TH
548static inline struct cgroup *task_dfl_cgroup(struct task_struct *task)
549{
550 return task_css_set(task)->dfl_cgrp;
551}
552
553static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
554{
555 struct cgroup_subsys_state *parent_css = cgrp->self.parent;
556
557 if (parent_css)
558 return container_of(parent_css, struct cgroup, self);
559 return NULL;
560}
561
b11cfb58
TH
562/**
563 * cgroup_is_descendant - test ancestry
564 * @cgrp: the cgroup to be tested
565 * @ancestor: possible ancestor of @cgrp
566 *
567 * Test whether @cgrp is a descendant of @ancestor. It also returns %true
568 * if @cgrp == @ancestor. This function is safe to call as long as @cgrp
569 * and @ancestor are accessible.
570 */
571static inline bool cgroup_is_descendant(struct cgroup *cgrp,
572 struct cgroup *ancestor)
573{
574 if (cgrp->root != ancestor->root || cgrp->level < ancestor->level)
575 return false;
74321038 576 return cgrp->ancestor_ids[ancestor->level] == cgroup_id(ancestor);
b11cfb58
TH
577}
578
77236281
AI
579/**
580 * cgroup_ancestor - find ancestor of cgroup
581 * @cgrp: cgroup to find ancestor of
582 * @ancestor_level: level of ancestor to find starting from root
583 *
584 * Find ancestor of cgroup at specified level starting from root if it exists
585 * and return pointer to it. Return NULL if @cgrp doesn't have ancestor at
586 * @ancestor_level.
587 *
588 * This function is safe to call as long as @cgrp is accessible.
589 */
590static inline struct cgroup *cgroup_ancestor(struct cgroup *cgrp,
591 int ancestor_level)
592{
77236281
AI
593 if (cgrp->level < ancestor_level)
594 return NULL;
808c43b7
AI
595 while (cgrp && cgrp->level > ancestor_level)
596 cgrp = cgroup_parent(cgrp);
597 return cgrp;
77236281
AI
598}
599
aed704b7
SD
600/**
601 * task_under_cgroup_hierarchy - test task's membership of cgroup ancestry
602 * @task: the task to be tested
603 * @ancestor: possible ancestor of @task's cgroup
604 *
605 * Tests whether @task's default cgroup hierarchy is a descendant of @ancestor.
606 * It follows all the same rules as cgroup_is_descendant, and only applies
607 * to the default hierarchy.
608 */
609static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
610 struct cgroup *ancestor)
611{
612 struct css_set *cset = task_css_set(task);
613
614 return cgroup_is_descendant(cset->dfl_cgrp, ancestor);
615}
616
07bc356e 617/* no synchronization, the result can only be used as a hint */
27bd4dbb 618static inline bool cgroup_is_populated(struct cgroup *cgrp)
07bc356e 619{
454000ad
TH
620 return cgrp->nr_populated_csets + cgrp->nr_populated_domain_children +
621 cgrp->nr_populated_threaded_children;
07bc356e 622}
f3d46500 623
f29374b1 624/* returns ino associated with a cgroup */
b1664924
TH
625static inline ino_t cgroup_ino(struct cgroup *cgrp)
626{
67c0496e 627 return kernfs_ino(cgrp->kn);
b1664924 628}
3ebb2b6e 629
b4168640
TH
630/* cft/css accessors for cftype->write() operation */
631static inline struct cftype *of_cft(struct kernfs_open_file *of)
7da11279 632{
2bd59d48 633 return of->kn->priv;
7da11279 634}
0f0a2b4f 635
b4168640 636struct cgroup_subsys_state *of_css(struct kernfs_open_file *of);
817929ec 637
b4168640
TH
638/* cft/css accessors for cftype->seq_*() operations */
639static inline struct cftype *seq_cft(struct seq_file *seq)
640{
641 return of_cft(seq->private);
642}
e535837b 643
b4168640
TH
644static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq)
645{
646 return of_css(seq->private);
647}
31583bb0 648
e61734c5
TH
649/*
650 * Name / path handling functions. All are thin wrappers around the kernfs
651 * counterparts and can be called under any context.
652 */
38460b48 653
e61734c5
TH
654static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen)
655{
fdce6bf8 656 return kernfs_name(cgrp->kn, buf, buflen);
e61734c5 657}
ddbcc7e8 658
4c737b41 659static inline int cgroup_path(struct cgroup *cgrp, char *buf, size_t buflen)
e61734c5 660{
fdce6bf8 661 return kernfs_path(cgrp->kn, buf, buflen);
e61734c5 662}
f3ba5380 663
e61734c5
TH
664static inline void pr_cont_cgroup_name(struct cgroup *cgrp)
665{
fdce6bf8 666 pr_cont_kernfs_name(cgrp->kn);
e61734c5 667}
ddbcc7e8 668
e61734c5 669static inline void pr_cont_cgroup_path(struct cgroup *cgrp)
846c7bb0 670{
fdce6bf8 671 pr_cont_kernfs_path(cgrp->kn);
846c7bb0 672}
ddbcc7e8 673
2ce7135a
JW
674static inline struct psi_group *cgroup_psi(struct cgroup *cgrp)
675{
676 return &cgrp->psi;
677}
678
3958e2d0
SB
679bool cgroup_psi_enabled(void);
680
77f88796
TH
681static inline void cgroup_init_kthreadd(void)
682{
683 /*
684 * kthreadd is inherited by all kthreads, keep it in the root so
685 * that the new kthreads are guaranteed to stay in the root until
686 * initialization is finished.
687 */
688 current->no_cgroup_migration = 1;
689}
690
691static inline void cgroup_kthread_ready(void)
692{
693 /*
694 * This kthread finished initialization. The creator should have
695 * set PF_NO_SETAFFINITY if this kthread should stay in the root.
696 */
697 current->no_cgroup_migration = 0;
698}
699
67c0496e 700void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen);
6b658c48 701struct cgroup *cgroup_get_from_id(u64 id);
ddbcc7e8
PM
702#else /* !CONFIG_CGROUPS */
703
f3ba5380 704struct cgroup_subsys_state;
aed704b7 705struct cgroup;
f3ba5380 706
a3b89864 707static inline u64 cgroup_id(const struct cgroup *cgrp) { return 1; }
9b0eb69b 708static inline void css_get(struct cgroup_subsys_state *css) {}
c326aa2b 709static inline void css_put(struct cgroup_subsys_state *css) {}
31583bb0 710static inline int cgroup_attach_task_all(struct task_struct *from,
c326aa2b
TH
711 struct task_struct *t) { return 0; }
712static inline int cgroupstats_build(struct cgroupstats *stats,
713 struct dentry *dentry) { return -EINVAL; }
714
b4f48b63 715static inline void cgroup_fork(struct task_struct *p) {}
ef2c41cf
CB
716static inline int cgroup_can_fork(struct task_struct *p,
717 struct kernel_clone_args *kargs) { return 0; }
718static inline void cgroup_cancel_fork(struct task_struct *p,
719 struct kernel_clone_args *kargs) {}
720static inline void cgroup_post_fork(struct task_struct *p,
721 struct kernel_clone_args *kargs) {}
1ec41830 722static inline void cgroup_exit(struct task_struct *p) {}
51bee5ab 723static inline void cgroup_release(struct task_struct *p) {}
2e91fa7f 724static inline void cgroup_free(struct task_struct *p) {}
ddbcc7e8 725
c326aa2b
TH
726static inline int cgroup_init_early(void) { return 0; }
727static inline int cgroup_init(void) { return 0; }
77f88796
TH
728static inline void cgroup_init_kthreadd(void) {}
729static inline void cgroup_kthread_ready(void) {}
d7926ee3 730
2ce7135a
JW
731static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
732{
733 return NULL;
734}
735
736static inline struct psi_group *cgroup_psi(struct cgroup *cgrp)
737{
738 return NULL;
739}
740
3958e2d0
SB
741static inline bool cgroup_psi_enabled(void)
742{
743 return false;
744}
745
aed704b7
SD
746static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
747 struct cgroup *ancestor)
748{
749 return true;
750}
69fd5c39 751
67c0496e
TH
752static inline void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen)
753{}
6b658c48
MK
754
755static inline struct cgroup *cgroup_get_from_id(u64 id)
756{
757 return NULL;
758}
ddbcc7e8
PM
759#endif /* !CONFIG_CGROUPS */
760
6162cef0 761#ifdef CONFIG_CGROUPS
d2cc5ed6 762/*
6162cef0 763 * cgroup scalable recursive statistics.
d2cc5ed6 764 */
6162cef0
TH
765void cgroup_rstat_updated(struct cgroup *cgrp, int cpu);
766void cgroup_rstat_flush(struct cgroup *cgrp);
0fa294fb 767void cgroup_rstat_flush_irqsafe(struct cgroup *cgrp);
6162cef0
TH
768void cgroup_rstat_flush_hold(struct cgroup *cgrp);
769void cgroup_rstat_flush_release(void);
d2cc5ed6 770
6162cef0
TH
771/*
772 * Basic resource stats.
773 */
d2cc5ed6
TH
774#ifdef CONFIG_CGROUP_CPUACCT
775void cpuacct_charge(struct task_struct *tsk, u64 cputime);
776void cpuacct_account_field(struct task_struct *tsk, int index, u64 val);
777#else
778static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {}
779static inline void cpuacct_account_field(struct task_struct *tsk, int index,
780 u64 val) {}
781#endif
782
041cd640
TH
783void __cgroup_account_cputime(struct cgroup *cgrp, u64 delta_exec);
784void __cgroup_account_cputime_field(struct cgroup *cgrp,
785 enum cpu_usage_stat index, u64 delta_exec);
786
d2cc5ed6
TH
787static inline void cgroup_account_cputime(struct task_struct *task,
788 u64 delta_exec)
789{
041cd640
TH
790 struct cgroup *cgrp;
791
d2cc5ed6 792 cpuacct_charge(task, delta_exec);
041cd640
TH
793
794 rcu_read_lock();
795 cgrp = task_dfl_cgroup(task);
796 if (cgroup_parent(cgrp))
797 __cgroup_account_cputime(cgrp, delta_exec);
798 rcu_read_unlock();
d2cc5ed6
TH
799}
800
801static inline void cgroup_account_cputime_field(struct task_struct *task,
802 enum cpu_usage_stat index,
803 u64 delta_exec)
804{
041cd640
TH
805 struct cgroup *cgrp;
806
d2cc5ed6 807 cpuacct_account_field(task, index, delta_exec);
041cd640
TH
808
809 rcu_read_lock();
810 cgrp = task_dfl_cgroup(task);
811 if (cgroup_parent(cgrp))
812 __cgroup_account_cputime_field(cgrp, index, delta_exec);
813 rcu_read_unlock();
d2cc5ed6
TH
814}
815
816#else /* CONFIG_CGROUPS */
817
818static inline void cgroup_account_cputime(struct task_struct *task,
819 u64 delta_exec) {}
820static inline void cgroup_account_cputime_field(struct task_struct *task,
821 enum cpu_usage_stat index,
822 u64 delta_exec) {}
823
824#endif /* CONFIG_CGROUPS */
825
bd1060a1
TH
826/*
827 * sock->sk_cgrp_data handling. For more info, see sock_cgroup_data
828 * definition in cgroup-defs.h.
829 */
830#ifdef CONFIG_SOCK_CGROUP_DATA
831
bd1060a1 832void cgroup_sk_alloc(struct sock_cgroup_data *skcd);
ad0f75e5 833void cgroup_sk_clone(struct sock_cgroup_data *skcd);
bd1060a1
TH
834void cgroup_sk_free(struct sock_cgroup_data *skcd);
835
836static inline struct cgroup *sock_cgroup_ptr(struct sock_cgroup_data *skcd)
837{
8520e224 838 return skcd->cgroup;
bd1060a1
TH
839}
840
841#else /* CONFIG_CGROUP_DATA */
842
843static inline void cgroup_sk_alloc(struct sock_cgroup_data *skcd) {}
ad0f75e5 844static inline void cgroup_sk_clone(struct sock_cgroup_data *skcd) {}
bd1060a1
TH
845static inline void cgroup_sk_free(struct sock_cgroup_data *skcd) {}
846
847#endif /* CONFIG_CGROUP_DATA */
848
a79a908f 849struct cgroup_namespace {
a79a908f
AK
850 struct ns_common ns;
851 struct user_namespace *user_ns;
d08311dd 852 struct ucounts *ucounts;
a79a908f
AK
853 struct css_set *root_cset;
854};
855
856extern struct cgroup_namespace init_cgroup_ns;
857
858#ifdef CONFIG_CGROUPS
859
860void free_cgroup_ns(struct cgroup_namespace *ns);
861
862struct cgroup_namespace *copy_cgroup_ns(unsigned long flags,
863 struct user_namespace *user_ns,
864 struct cgroup_namespace *old_ns);
865
4c737b41
TH
866int cgroup_path_ns(struct cgroup *cgrp, char *buf, size_t buflen,
867 struct cgroup_namespace *ns);
a79a908f
AK
868
869#else /* !CONFIG_CGROUPS */
870
871static inline void free_cgroup_ns(struct cgroup_namespace *ns) { }
872static inline struct cgroup_namespace *
873copy_cgroup_ns(unsigned long flags, struct user_namespace *user_ns,
874 struct cgroup_namespace *old_ns)
875{
876 return old_ns;
877}
878
879#endif /* !CONFIG_CGROUPS */
880
881static inline void get_cgroup_ns(struct cgroup_namespace *ns)
882{
883 if (ns)
f387882d 884 refcount_inc(&ns->ns.count);
a79a908f
AK
885}
886
887static inline void put_cgroup_ns(struct cgroup_namespace *ns)
888{
f387882d 889 if (ns && refcount_dec_and_test(&ns->ns.count))
a79a908f
AK
890 free_cgroup_ns(ns);
891}
892
76f969e8
RG
893#ifdef CONFIG_CGROUPS
894
895void cgroup_enter_frozen(void);
896void cgroup_leave_frozen(bool always_leave);
897void cgroup_update_frozen(struct cgroup *cgrp);
898void cgroup_freeze(struct cgroup *cgrp, bool freeze);
899void cgroup_freezer_migrate_task(struct task_struct *task, struct cgroup *src,
900 struct cgroup *dst);
96b9c592 901
76f969e8
RG
902static inline bool cgroup_task_frozen(struct task_struct *task)
903{
904 return task->frozen;
905}
906
907#else /* !CONFIG_CGROUPS */
908
909static inline void cgroup_enter_frozen(void) { }
910static inline void cgroup_leave_frozen(bool always_leave) { }
76f969e8
RG
911static inline bool cgroup_task_frozen(struct task_struct *task)
912{
913 return false;
914}
915
916#endif /* !CONFIG_CGROUPS */
917
4bfc0bb2
RG
918#ifdef CONFIG_CGROUP_BPF
919static inline void cgroup_bpf_get(struct cgroup *cgrp)
920{
921 percpu_ref_get(&cgrp->bpf.refcnt);
922}
923
924static inline void cgroup_bpf_put(struct cgroup *cgrp)
925{
926 percpu_ref_put(&cgrp->bpf.refcnt);
927}
928
929#else /* CONFIG_CGROUP_BPF */
930
931static inline void cgroup_bpf_get(struct cgroup *cgrp) {}
932static inline void cgroup_bpf_put(struct cgroup *cgrp) {}
933
934#endif /* CONFIG_CGROUP_BPF */
935
ddbcc7e8 936#endif /* _LINUX_CGROUP_H */