blob: b7dd23040cd58ad77229bd4991d45b64c7fb034a [file] [log] [blame]
Paul Menageddbcc7e2007-10-18 23:39:30 -07001#ifndef _LINUX_CGROUP_H
2#define _LINUX_CGROUP_H
3/*
4 * cgroup interface
5 *
6 * Copyright (C) 2003 BULL SA
7 * Copyright (C) 2004-2006 Silicon Graphics, Inc.
8 *
9 */
10
11#include <linux/sched.h>
Paul Menageddbcc7e2007-10-18 23:39:30 -070012#include <linux/cpumask.h>
13#include <linux/nodemask.h>
Tejun Heoeb6fd502012-11-09 09:12:29 -080014#include <linux/rculist.h>
Balbir Singh846c7bb2007-10-18 23:39:44 -070015#include <linux/cgroupstats.h>
Tejun Heo25a7e682013-04-14 20:15:25 -070016#include <linux/fs.h>
Tejun Heo7da11272013-12-05 12:28:04 -050017#include <linux/seq_file.h>
Tejun Heo2bd59d42014-02-11 11:52:49 -050018#include <linux/kernfs.h>
Tejun Heo49d1dc42015-09-18 11:56:28 -040019#include <linux/jump_label.h>
Aditya Kalia79a9082016-01-29 02:54:06 -060020#include <linux/types.h>
21#include <linux/ns_common.h>
22#include <linux/nsproxy.h>
23#include <linux/user_namespace.h>
Elena Reshetova387ad962017-02-20 12:19:00 +020024#include <linux/refcount.h>
Tejun Heob4a04ab2015-05-13 15:38:40 -040025
26#include <linux/cgroup-defs.h>
Paul Menageddbcc7e2007-10-18 23:39:30 -070027
28#ifdef CONFIG_CGROUPS
29
Tejun Heo6abc8ca2015-08-04 15:20:55 -040030/*
31 * All weight knobs on the default hierarhcy should use the following min,
32 * default and max values. The default value is the logarithmic center of
33 * MIN and MAX and allows 100x to be expressed in both directions.
34 */
35#define CGROUP_WEIGHT_MIN 1
36#define CGROUP_WEIGHT_DFL 100
37#define CGROUP_WEIGHT_MAX 10000
38
Tejun Heobc2fb7e2017-05-15 09:34:01 -040039/* walk only threadgroup leaders */
40#define CSS_TASK_ITER_PROCS (1U << 0)
41
Tejun Heoc326aa22015-05-13 16:24:16 -040042/* a css_task_iter should be treated as an opaque object */
43struct css_task_iter {
44 struct cgroup_subsys *ss;
Tejun Heobc2fb7e2017-05-15 09:34:01 -040045 unsigned int flags;
Paul Menageddbcc7e2007-10-18 23:39:30 -070046
Tejun Heoc326aa22015-05-13 16:24:16 -040047 struct list_head *cset_pos;
48 struct list_head *cset_head;
Paul Menageddbcc7e2007-10-18 23:39:30 -070049
Tejun Heoc326aa22015-05-13 16:24:16 -040050 struct list_head *task_pos;
51 struct list_head *tasks_head;
52 struct list_head *mg_tasks_head;
Tejun Heoed27b9f2015-10-15 16:41:52 -040053
54 struct css_set *cur_cset;
55 struct task_struct *cur_task;
56 struct list_head iters_node; /* css_set->task_iters */
Paul Menage817929e2007-10-18 23:39:36 -070057};
Tejun Heoc326aa22015-05-13 16:24:16 -040058
59extern struct cgroup_root cgrp_dfl_root;
60extern struct css_set init_css_set;
61
62#define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys;
63#include <linux/cgroup_subsys.h>
Paul Menage817929e2007-10-18 23:39:36 -070064#undef SUBSYS
65
Tejun Heo49d1dc42015-09-18 11:56:28 -040066#define SUBSYS(_x) \
67 extern struct static_key_true _x ## _cgrp_subsys_enabled_key; \
68 extern struct static_key_true _x ## _cgrp_subsys_on_dfl_key;
69#include <linux/cgroup_subsys.h>
70#undef SUBSYS
71
72/**
73 * cgroup_subsys_enabled - fast test on whether a subsys is enabled
74 * @ss: subsystem in question
75 */
76#define cgroup_subsys_enabled(ss) \
77 static_branch_likely(&ss ## _enabled_key)
78
79/**
80 * cgroup_subsys_on_dfl - fast test on whether a subsys is on default hierarchy
81 * @ss: subsystem in question
82 */
83#define cgroup_subsys_on_dfl(ss) \
84 static_branch_likely(&ss ## _on_dfl_key)
85
Tejun Heoc326aa22015-05-13 16:24:16 -040086bool css_has_online_children(struct cgroup_subsys_state *css);
87struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss);
88struct cgroup_subsys_state *cgroup_get_e_css(struct cgroup *cgroup,
89 struct cgroup_subsys *ss);
90struct cgroup_subsys_state *css_tryget_online_from_dir(struct dentry *dentry,
91 struct cgroup_subsys *ss);
92
Tejun Heo16af4392015-11-20 15:55:52 -050093struct cgroup *cgroup_get_from_path(const char *path);
Martin KaFai Lau1f3fe7e2016-06-30 10:28:42 -070094struct cgroup *cgroup_get_from_fd(int fd);
Tejun Heo16af4392015-11-20 15:55:52 -050095
Tejun Heoc326aa22015-05-13 16:24:16 -040096int cgroup_attach_task_all(struct task_struct *from, struct task_struct *);
97int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from);
98
99int cgroup_add_dfl_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
100int cgroup_add_legacy_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
101int cgroup_rm_cftypes(struct cftype *cfts);
Tejun Heo34c06252015-11-05 00:12:24 -0500102void cgroup_file_notify(struct cgroup_file *cfile);
Tejun Heoc326aa22015-05-13 16:24:16 -0400103
Tejun Heo4c737b42016-08-10 11:23:44 -0400104int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen);
Tejun Heoc326aa22015-05-13 16:24:16 -0400105int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry);
106int proc_cgroup_show(struct seq_file *m, struct pid_namespace *ns,
107 struct pid *pid, struct task_struct *tsk);
108
109void cgroup_fork(struct task_struct *p);
Oleg Nesterovb53202e2015-12-03 10:24:08 -0500110extern int cgroup_can_fork(struct task_struct *p);
111extern void cgroup_cancel_fork(struct task_struct *p);
112extern void cgroup_post_fork(struct task_struct *p);
Tejun Heoc326aa22015-05-13 16:24:16 -0400113void cgroup_exit(struct task_struct *p);
Tejun Heo2e91fa72015-10-15 16:41:53 -0400114void cgroup_free(struct task_struct *p);
Tejun Heoc326aa22015-05-13 16:24:16 -0400115
116int cgroup_init_early(void);
117int cgroup_init(void);
118
Tejun Heo5c9d5352014-05-16 13:22:48 -0400119/*
Tejun Heoc326aa22015-05-13 16:24:16 -0400120 * Iteration helpers and macros.
Tejun Heo5c9d5352014-05-16 13:22:48 -0400121 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700122
Tejun Heoc326aa22015-05-13 16:24:16 -0400123struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos,
124 struct cgroup_subsys_state *parent);
125struct cgroup_subsys_state *css_next_descendant_pre(struct cgroup_subsys_state *pos,
126 struct cgroup_subsys_state *css);
127struct cgroup_subsys_state *css_rightmost_descendant(struct cgroup_subsys_state *pos);
128struct cgroup_subsys_state *css_next_descendant_post(struct cgroup_subsys_state *pos,
129 struct cgroup_subsys_state *css);
Tejun Heo72c97e52013-08-08 20:11:22 -0400130
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500131struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset,
132 struct cgroup_subsys_state **dst_cssp);
133struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset,
134 struct cgroup_subsys_state **dst_cssp);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700135
Tejun Heobc2fb7e2017-05-15 09:34:01 -0400136void css_task_iter_start(struct cgroup_subsys_state *css, unsigned int flags,
Tejun Heoc326aa22015-05-13 16:24:16 -0400137 struct css_task_iter *it);
138struct task_struct *css_task_iter_next(struct css_task_iter *it);
139void css_task_iter_end(struct css_task_iter *it);
Tejun Heo0ae78e02013-08-13 11:01:54 -0400140
Tejun Heoc326aa22015-05-13 16:24:16 -0400141/**
142 * css_for_each_child - iterate through children of a css
143 * @pos: the css * to use as the loop cursor
144 * @parent: css whose children to walk
145 *
146 * Walk @parent's children. Must be called under rcu_read_lock().
147 *
148 * If a subsystem synchronizes ->css_online() and the start of iteration, a
149 * css which finished ->css_online() is guaranteed to be visible in the
150 * future iterations and will stay visible until the last reference is put.
151 * A css which hasn't finished ->css_online() or already finished
152 * ->css_offline() may show up during traversal. It's each subsystem's
153 * responsibility to synchronize against on/offlining.
154 *
155 * It is allowed to temporarily drop RCU read lock during iteration. The
156 * caller is responsible for ensuring that @pos remains accessible until
157 * the start of the next iteration by, for example, bumping the css refcnt.
158 */
159#define css_for_each_child(pos, parent) \
160 for ((pos) = css_next_child(NULL, (parent)); (pos); \
161 (pos) = css_next_child((pos), (parent)))
Tejun Heod5c419b2014-05-16 13:22:48 -0400162
Tejun Heoc326aa22015-05-13 16:24:16 -0400163/**
164 * css_for_each_descendant_pre - pre-order walk of a css's descendants
165 * @pos: the css * to use as the loop cursor
166 * @root: css whose descendants to walk
167 *
168 * Walk @root's descendants. @root is included in the iteration and the
169 * first node to be visited. Must be called under rcu_read_lock().
170 *
171 * If a subsystem synchronizes ->css_online() and the start of iteration, a
172 * css which finished ->css_online() is guaranteed to be visible in the
173 * future iterations and will stay visible until the last reference is put.
174 * A css which hasn't finished ->css_online() or already finished
175 * ->css_offline() may show up during traversal. It's each subsystem's
176 * responsibility to synchronize against on/offlining.
177 *
178 * For example, the following guarantees that a descendant can't escape
179 * state updates of its ancestors.
180 *
181 * my_online(@css)
182 * {
183 * Lock @css's parent and @css;
184 * Inherit state from the parent;
185 * Unlock both.
186 * }
187 *
188 * my_update_state(@css)
189 * {
190 * css_for_each_descendant_pre(@pos, @css) {
191 * Lock @pos;
192 * if (@pos == @css)
193 * Update @css's state;
194 * else
195 * Verify @pos is alive and inherit state from its parent;
196 * Unlock @pos;
197 * }
198 * }
199 *
200 * As long as the inheriting step, including checking the parent state, is
201 * enclosed inside @pos locking, double-locking the parent isn't necessary
202 * while inheriting. The state update to the parent is guaranteed to be
203 * visible by walking order and, as long as inheriting operations to the
204 * same @pos are atomic to each other, multiple updates racing each other
205 * still result in the correct state. It's guaranateed that at least one
206 * inheritance happens for any css after the latest update to its parent.
207 *
208 * If checking parent's state requires locking the parent, each inheriting
209 * iteration should lock and unlock both @pos->parent and @pos.
210 *
211 * Alternatively, a subsystem may choose to use a single global lock to
212 * synchronize ->css_online() and ->css_offline() against tree-walking
213 * operations.
214 *
215 * It is allowed to temporarily drop RCU read lock during iteration. The
216 * caller is responsible for ensuring that @pos remains accessible until
217 * the start of the next iteration by, for example, bumping the css refcnt.
218 */
219#define css_for_each_descendant_pre(pos, css) \
220 for ((pos) = css_next_descendant_pre(NULL, (css)); (pos); \
221 (pos) = css_next_descendant_pre((pos), (css)))
Tejun Heo15a4c832014-05-04 15:09:14 -0400222
Tejun Heoc326aa22015-05-13 16:24:16 -0400223/**
224 * css_for_each_descendant_post - post-order walk of a css's descendants
225 * @pos: the css * to use as the loop cursor
226 * @css: css whose descendants to walk
227 *
228 * Similar to css_for_each_descendant_pre() but performs post-order
229 * traversal instead. @root is included in the iteration and the last
230 * node to be visited.
231 *
232 * If a subsystem synchronizes ->css_online() and the start of iteration, a
233 * css which finished ->css_online() is guaranteed to be visible in the
234 * future iterations and will stay visible until the last reference is put.
235 * A css which hasn't finished ->css_online() or already finished
236 * ->css_offline() may show up during traversal. It's each subsystem's
237 * responsibility to synchronize against on/offlining.
238 *
239 * Note that the walk visibility guarantee example described in pre-order
240 * walk doesn't apply the same to post-order walks.
241 */
242#define css_for_each_descendant_post(pos, css) \
243 for ((pos) = css_next_descendant_post(NULL, (css)); (pos); \
244 (pos) = css_next_descendant_post((pos), (css)))
Tejun Heo48ddbe12012-04-01 12:09:56 -0700245
Tejun Heoc326aa22015-05-13 16:24:16 -0400246/**
247 * cgroup_taskset_for_each - iterate cgroup_taskset
248 * @task: the loop cursor
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500249 * @dst_css: the destination css
Tejun Heoc326aa22015-05-13 16:24:16 -0400250 * @tset: taskset to iterate
Tejun Heo4530edd2015-09-11 15:00:19 -0400251 *
252 * @tset may contain multiple tasks and they may belong to multiple
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500253 * processes.
254 *
255 * On the v2 hierarchy, there may be tasks from multiple processes and they
256 * may not share the source or destination csses.
257 *
258 * On traditional hierarchies, when there are multiple tasks in @tset, if a
259 * task of a process is in @tset, all tasks of the process are in @tset.
260 * Also, all are guaranteed to share the same source and destination csses.
Tejun Heo4530edd2015-09-11 15:00:19 -0400261 *
262 * Iteration is not in any specific order.
Tejun Heoc326aa22015-05-13 16:24:16 -0400263 */
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500264#define cgroup_taskset_for_each(task, dst_css, tset) \
265 for ((task) = cgroup_taskset_first((tset), &(dst_css)); \
266 (task); \
267 (task) = cgroup_taskset_next((tset), &(dst_css)))
Tejun Heo0cb51d72014-05-16 13:22:49 -0400268
Tejun Heo4530edd2015-09-11 15:00:19 -0400269/**
270 * cgroup_taskset_for_each_leader - iterate group leaders in a cgroup_taskset
271 * @leader: the loop cursor
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500272 * @dst_css: the destination css
Geliang Tang7b4632f2016-12-24 23:28:35 +0800273 * @tset: taskset to iterate
Tejun Heo4530edd2015-09-11 15:00:19 -0400274 *
275 * Iterate threadgroup leaders of @tset. For single-task migrations, @tset
276 * may not contain any.
277 */
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500278#define cgroup_taskset_for_each_leader(leader, dst_css, tset) \
279 for ((leader) = cgroup_taskset_first((tset), &(dst_css)); \
280 (leader); \
281 (leader) = cgroup_taskset_next((tset), &(dst_css))) \
Tejun Heo4530edd2015-09-11 15:00:19 -0400282 if ((leader) != (leader)->group_leader) \
283 ; \
284 else
285
Tejun Heoc326aa22015-05-13 16:24:16 -0400286/*
287 * Inline functions.
288 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700289
Tejun Heo5de01072013-06-12 21:04:52 -0700290/**
291 * css_get - obtain a reference on the specified css
292 * @css: target css
293 *
294 * The caller must already have a reference.
Paul Menageddbcc7e2007-10-18 23:39:30 -0700295 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700296static inline void css_get(struct cgroup_subsys_state *css)
297{
Tejun Heo3b514d22014-05-16 13:22:47 -0400298 if (!(css->flags & CSS_NO_REF))
299 percpu_ref_get(&css->refcnt);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700300}
Paul Menagee7c5ec92009-01-07 18:08:38 -0800301
Tejun Heo5de01072013-06-12 21:04:52 -0700302/**
Johannes Weinere8ea14c2014-12-10 15:42:42 -0800303 * css_get_many - obtain references on the specified css
304 * @css: target css
305 * @n: number of references to get
306 *
307 * The caller must already have a reference.
308 */
309static inline void css_get_many(struct cgroup_subsys_state *css, unsigned int n)
310{
311 if (!(css->flags & CSS_NO_REF))
312 percpu_ref_get_many(&css->refcnt, n);
313}
314
315/**
Tejun Heo6f4524d2014-05-16 13:22:52 -0400316 * css_tryget - try to obtain a reference on the specified css
317 * @css: target css
318 *
319 * Obtain a reference on @css unless it already has reached zero and is
320 * being released. This function doesn't care whether @css is on or
321 * offline. The caller naturally needs to ensure that @css is accessible
322 * but doesn't have to be holding a reference on it - IOW, RCU protected
323 * access is good enough for this function. Returns %true if a reference
324 * count was successfully obtained; %false otherwise.
325 */
326static inline bool css_tryget(struct cgroup_subsys_state *css)
327{
328 if (!(css->flags & CSS_NO_REF))
329 return percpu_ref_tryget(&css->refcnt);
330 return true;
331}
332
333/**
Tejun Heoec903c02014-05-13 12:11:01 -0400334 * css_tryget_online - try to obtain a reference on the specified css if online
Tejun Heo5de01072013-06-12 21:04:52 -0700335 * @css: target css
336 *
Tejun Heoec903c02014-05-13 12:11:01 -0400337 * Obtain a reference on @css if it's online. The caller naturally needs
338 * to ensure that @css is accessible but doesn't have to be holding a
Tejun Heo5de01072013-06-12 21:04:52 -0700339 * reference on it - IOW, RCU protected access is good enough for this
340 * function. Returns %true if a reference count was successfully obtained;
341 * %false otherwise.
342 */
Tejun Heoec903c02014-05-13 12:11:01 -0400343static inline bool css_tryget_online(struct cgroup_subsys_state *css)
Paul Menagee7c5ec92009-01-07 18:08:38 -0800344{
Tejun Heo3b514d22014-05-16 13:22:47 -0400345 if (!(css->flags & CSS_NO_REF))
346 return percpu_ref_tryget_live(&css->refcnt);
347 return true;
Paul Menagee7c5ec92009-01-07 18:08:38 -0800348}
349
Tejun Heo5de01072013-06-12 21:04:52 -0700350/**
Tejun Heo41c25702017-05-24 12:03:48 -0400351 * css_is_dying - test whether the specified css is dying
352 * @css: target css
353 *
354 * Test whether @css is in the process of offlining or already offline. In
355 * most cases, ->css_online() and ->css_offline() callbacks should be
356 * enough; however, the actual offline operations are RCU delayed and this
357 * test returns %true also when @css is scheduled to be offlined.
358 *
359 * This is useful, for example, when the use case requires synchronous
360 * behavior with respect to cgroup removal. cgroup removal schedules css
361 * offlining but the css can seem alive while the operation is being
362 * delayed. If the delay affects user visible semantics, this test can be
363 * used to resolve the situation.
364 */
365static inline bool css_is_dying(struct cgroup_subsys_state *css)
366{
367 return !(css->flags & CSS_NO_REF) && percpu_ref_is_dying(&css->refcnt);
368}
369
370/**
Tejun Heo5de01072013-06-12 21:04:52 -0700371 * css_put - put a css reference
372 * @css: target css
373 *
Tejun Heoec903c02014-05-13 12:11:01 -0400374 * Put a reference obtained via css_get() and css_tryget_online().
Tejun Heo5de01072013-06-12 21:04:52 -0700375 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700376static inline void css_put(struct cgroup_subsys_state *css)
377{
Tejun Heo3b514d22014-05-16 13:22:47 -0400378 if (!(css->flags & CSS_NO_REF))
379 percpu_ref_put(&css->refcnt);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700380}
381
Johannes Weinere8ea14c2014-12-10 15:42:42 -0800382/**
383 * css_put_many - put css references
384 * @css: target css
385 * @n: number of references to put
386 *
387 * Put references obtained via css_get() and css_tryget_online().
388 */
389static inline void css_put_many(struct cgroup_subsys_state *css, unsigned int n)
390{
391 if (!(css->flags & CSS_NO_REF))
392 percpu_ref_put_many(&css->refcnt, n);
393}
394
Tejun Heo16af4392015-11-20 15:55:52 -0500395static inline void cgroup_put(struct cgroup *cgrp)
396{
397 css_put(&cgrp->self);
398}
399
Tejun Heoc326aa22015-05-13 16:24:16 -0400400/**
401 * task_css_set_check - obtain a task's css_set with extra access conditions
402 * @task: the task to obtain css_set for
403 * @__c: extra condition expression to be passed to rcu_dereference_check()
Paul Menageddbcc7e2007-10-18 23:39:30 -0700404 *
Tejun Heoc326aa22015-05-13 16:24:16 -0400405 * A task's css_set is RCU protected, initialized and exited while holding
406 * task_lock(), and can only be modified while holding both cgroup_mutex
407 * and task_lock() while the task is alive. This macro verifies that the
408 * caller is inside proper critical section and returns @task's css_set.
409 *
410 * The caller can also specify additional allowed conditions via @__c, such
411 * as locks used during the cgroup_subsys::attach() methods.
Paul Menageddbcc7e2007-10-18 23:39:30 -0700412 */
Tejun Heoc326aa22015-05-13 16:24:16 -0400413#ifdef CONFIG_PROVE_RCU
414extern struct mutex cgroup_mutex;
Tejun Heof0d9a5f2015-10-15 16:41:53 -0400415extern spinlock_t css_set_lock;
Tejun Heoc326aa22015-05-13 16:24:16 -0400416#define task_css_set_check(task, __c) \
417 rcu_dereference_check((task)->cgroups, \
418 lockdep_is_held(&cgroup_mutex) || \
Tejun Heof0d9a5f2015-10-15 16:41:53 -0400419 lockdep_is_held(&css_set_lock) || \
Tejun Heoc326aa22015-05-13 16:24:16 -0400420 ((task)->flags & PF_EXITING) || (__c))
421#else
422#define task_css_set_check(task, __c) \
423 rcu_dereference((task)->cgroups)
Tejun Heo2bd59d42014-02-11 11:52:49 -0500424#endif
Paul Menageddbcc7e2007-10-18 23:39:30 -0700425
Tejun Heoc326aa22015-05-13 16:24:16 -0400426/**
427 * task_css_check - obtain css for (task, subsys) w/ extra access conds
428 * @task: the target task
429 * @subsys_id: the target subsystem ID
430 * @__c: extra condition expression to be passed to rcu_dereference_check()
431 *
432 * Return the cgroup_subsys_state for the (@task, @subsys_id) pair. The
433 * synchronization rules are the same as task_css_set_check().
434 */
435#define task_css_check(task, subsys_id, __c) \
436 task_css_set_check((task), (__c))->subsys[(subsys_id)]
437
438/**
439 * task_css_set - obtain a task's css_set
440 * @task: the task to obtain css_set for
441 *
442 * See task_css_set_check().
443 */
444static inline struct css_set *task_css_set(struct task_struct *task)
445{
446 return task_css_set_check(task, false);
447}
448
449/**
450 * task_css - obtain css for (task, subsys)
451 * @task: the target task
452 * @subsys_id: the target subsystem ID
453 *
454 * See task_css_check().
455 */
456static inline struct cgroup_subsys_state *task_css(struct task_struct *task,
457 int subsys_id)
458{
459 return task_css_check(task, subsys_id, false);
460}
461
462/**
Linus Torvaldsbbe179f2015-06-26 19:50:04 -0700463 * task_get_css - find and get the css for (task, subsys)
464 * @task: the target task
465 * @subsys_id: the target subsystem ID
466 *
467 * Find the css for the (@task, @subsys_id) combination, increment a
468 * reference on and return it. This function is guaranteed to return a
469 * valid css.
470 */
471static inline struct cgroup_subsys_state *
472task_get_css(struct task_struct *task, int subsys_id)
473{
474 struct cgroup_subsys_state *css;
475
476 rcu_read_lock();
477 while (true) {
478 css = task_css(task, subsys_id);
479 if (likely(css_tryget_online(css)))
480 break;
481 cpu_relax();
482 }
483 rcu_read_unlock();
484 return css;
485}
486
487/**
Tejun Heoc326aa22015-05-13 16:24:16 -0400488 * task_css_is_root - test whether a task belongs to the root css
489 * @task: the target task
490 * @subsys_id: the target subsystem ID
491 *
492 * Test whether @task belongs to the root css on the specified subsystem.
493 * May be invoked in any context.
494 */
495static inline bool task_css_is_root(struct task_struct *task, int subsys_id)
496{
497 return task_css_check(task, subsys_id, true) ==
498 init_css_set.subsys[subsys_id];
499}
500
501static inline struct cgroup *task_cgroup(struct task_struct *task,
502 int subsys_id)
503{
504 return task_css(task, subsys_id)->cgroup;
505}
Tejun Heoa2dd4242014-03-19 10:23:55 -0400506
Tejun Heob11cfb52015-11-20 15:55:52 -0500507/**
508 * cgroup_is_descendant - test ancestry
509 * @cgrp: the cgroup to be tested
510 * @ancestor: possible ancestor of @cgrp
511 *
512 * Test whether @cgrp is a descendant of @ancestor. It also returns %true
513 * if @cgrp == @ancestor. This function is safe to call as long as @cgrp
514 * and @ancestor are accessible.
515 */
516static inline bool cgroup_is_descendant(struct cgroup *cgrp,
517 struct cgroup *ancestor)
518{
519 if (cgrp->root != ancestor->root || cgrp->level < ancestor->level)
520 return false;
521 return cgrp->ancestor_ids[ancestor->level] == ancestor->id;
522}
523
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700524/**
525 * task_under_cgroup_hierarchy - test task's membership of cgroup ancestry
526 * @task: the task to be tested
527 * @ancestor: possible ancestor of @task's cgroup
528 *
529 * Tests whether @task's default cgroup hierarchy is a descendant of @ancestor.
530 * It follows all the same rules as cgroup_is_descendant, and only applies
531 * to the default hierarchy.
532 */
533static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
534 struct cgroup *ancestor)
535{
536 struct css_set *cset = task_css_set(task);
537
538 return cgroup_is_descendant(cset->dfl_cgrp, ancestor);
539}
540
Tejun Heo07bc3562014-02-13 06:58:39 -0500541/* no synchronization, the result can only be used as a hint */
Tejun Heo27bd4db2015-10-15 16:41:50 -0400542static inline bool cgroup_is_populated(struct cgroup *cgrp)
Tejun Heo07bc3562014-02-13 06:58:39 -0500543{
Tejun Heo454000a2017-05-15 09:34:02 -0400544 return cgrp->nr_populated_csets + cgrp->nr_populated_domain_children +
545 cgrp->nr_populated_threaded_children;
Tejun Heo07bc3562014-02-13 06:58:39 -0500546}
547
Zefan Lif29374b2014-09-19 16:29:31 +0800548/* returns ino associated with a cgroup */
Tejun Heob1664922014-02-11 11:52:49 -0500549static inline ino_t cgroup_ino(struct cgroup *cgrp)
550{
Zefan Lif29374b2014-09-19 16:29:31 +0800551 return cgrp->kn->ino;
Tejun Heob1664922014-02-11 11:52:49 -0500552}
553
Tejun Heob4168642014-05-13 12:16:21 -0400554/* cft/css accessors for cftype->write() operation */
555static inline struct cftype *of_cft(struct kernfs_open_file *of)
Tejun Heo7da11272013-12-05 12:28:04 -0500556{
Tejun Heo2bd59d42014-02-11 11:52:49 -0500557 return of->kn->priv;
Tejun Heo7da11272013-12-05 12:28:04 -0500558}
559
Tejun Heob4168642014-05-13 12:16:21 -0400560struct cgroup_subsys_state *of_css(struct kernfs_open_file *of);
561
562/* cft/css accessors for cftype->seq_*() operations */
563static inline struct cftype *seq_cft(struct seq_file *seq)
564{
565 return of_cft(seq->private);
566}
567
568static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq)
569{
570 return of_css(seq->private);
571}
Tejun Heo59f52962014-02-11 11:52:49 -0500572
Tejun Heoe61734c2014-02-12 09:29:50 -0500573/*
574 * Name / path handling functions. All are thin wrappers around the kernfs
575 * counterparts and can be called under any context.
576 */
577
578static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen)
579{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400580 return kernfs_name(cgrp->kn, buf, buflen);
Tejun Heoe61734c2014-02-12 09:29:50 -0500581}
582
Tejun Heo4c737b42016-08-10 11:23:44 -0400583static inline int cgroup_path(struct cgroup *cgrp, char *buf, size_t buflen)
Tejun Heoe61734c2014-02-12 09:29:50 -0500584{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400585 return kernfs_path(cgrp->kn, buf, buflen);
Tejun Heoe61734c2014-02-12 09:29:50 -0500586}
587
588static inline void pr_cont_cgroup_name(struct cgroup *cgrp)
589{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400590 pr_cont_kernfs_name(cgrp->kn);
Tejun Heoe61734c2014-02-12 09:29:50 -0500591}
592
593static inline void pr_cont_cgroup_path(struct cgroup *cgrp)
594{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400595 pr_cont_kernfs_path(cgrp->kn);
Tejun Heoe61734c2014-02-12 09:29:50 -0500596}
597
Tejun Heo77f88792017-03-16 16:54:24 -0400598static inline void cgroup_init_kthreadd(void)
599{
600 /*
601 * kthreadd is inherited by all kthreads, keep it in the root so
602 * that the new kthreads are guaranteed to stay in the root until
603 * initialization is finished.
604 */
605 current->no_cgroup_migration = 1;
606}
607
608static inline void cgroup_kthread_ready(void)
609{
610 /*
611 * This kthread finished initialization. The creator should have
612 * set PF_NO_SETAFFINITY if this kthread should stay in the root.
613 */
614 current->no_cgroup_migration = 0;
615}
616
Paul Menageddbcc7e2007-10-18 23:39:30 -0700617#else /* !CONFIG_CGROUPS */
618
Tejun Heof3ba5382015-01-06 12:02:46 -0500619struct cgroup_subsys_state;
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700620struct cgroup;
Tejun Heof3ba5382015-01-06 12:02:46 -0500621
Tejun Heoc326aa22015-05-13 16:24:16 -0400622static inline void css_put(struct cgroup_subsys_state *css) {}
623static inline int cgroup_attach_task_all(struct task_struct *from,
624 struct task_struct *t) { return 0; }
625static inline int cgroupstats_build(struct cgroupstats *stats,
626 struct dentry *dentry) { return -EINVAL; }
627
Paul Menageb4f48b62007-10-18 23:39:33 -0700628static inline void cgroup_fork(struct task_struct *p) {}
Oleg Nesterovb53202e2015-12-03 10:24:08 -0500629static inline int cgroup_can_fork(struct task_struct *p) { return 0; }
630static inline void cgroup_cancel_fork(struct task_struct *p) {}
631static inline void cgroup_post_fork(struct task_struct *p) {}
Li Zefan1ec41832014-03-28 15:22:19 +0800632static inline void cgroup_exit(struct task_struct *p) {}
Tejun Heo2e91fa72015-10-15 16:41:53 -0400633static inline void cgroup_free(struct task_struct *p) {}
Paul Menageddbcc7e2007-10-18 23:39:30 -0700634
Tejun Heoc326aa22015-05-13 16:24:16 -0400635static inline int cgroup_init_early(void) { return 0; }
636static inline int cgroup_init(void) { return 0; }
Tejun Heo77f88792017-03-16 16:54:24 -0400637static inline void cgroup_init_kthreadd(void) {}
638static inline void cgroup_kthread_ready(void) {}
Sridhar Samudralad7926ee2010-05-30 22:24:39 +0200639
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700640static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
641 struct cgroup *ancestor)
642{
643 return true;
644}
Paul Menageddbcc7e2007-10-18 23:39:30 -0700645#endif /* !CONFIG_CGROUPS */
646
Tejun Heobd1060a2015-12-07 17:38:53 -0500647/*
648 * sock->sk_cgrp_data handling. For more info, see sock_cgroup_data
649 * definition in cgroup-defs.h.
650 */
651#ifdef CONFIG_SOCK_CGROUP_DATA
652
653#if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
654extern spinlock_t cgroup_sk_update_lock;
655#endif
656
657void cgroup_sk_alloc_disable(void);
658void cgroup_sk_alloc(struct sock_cgroup_data *skcd);
659void cgroup_sk_free(struct sock_cgroup_data *skcd);
660
661static inline struct cgroup *sock_cgroup_ptr(struct sock_cgroup_data *skcd)
662{
663#if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
664 unsigned long v;
665
666 /*
667 * @skcd->val is 64bit but the following is safe on 32bit too as we
668 * just need the lower ulong to be written and read atomically.
669 */
670 v = READ_ONCE(skcd->val);
671
672 if (v & 1)
673 return &cgrp_dfl_root.cgrp;
674
675 return (struct cgroup *)(unsigned long)v ?: &cgrp_dfl_root.cgrp;
676#else
677 return (struct cgroup *)(unsigned long)skcd->val;
678#endif
679}
680
681#else /* CONFIG_CGROUP_DATA */
682
683static inline void cgroup_sk_alloc(struct sock_cgroup_data *skcd) {}
684static inline void cgroup_sk_free(struct sock_cgroup_data *skcd) {}
685
686#endif /* CONFIG_CGROUP_DATA */
687
Aditya Kalia79a9082016-01-29 02:54:06 -0600688struct cgroup_namespace {
Elena Reshetova387ad962017-02-20 12:19:00 +0200689 refcount_t count;
Aditya Kalia79a9082016-01-29 02:54:06 -0600690 struct ns_common ns;
691 struct user_namespace *user_ns;
Eric W. Biedermand08311d2016-08-08 14:25:30 -0500692 struct ucounts *ucounts;
Aditya Kalia79a9082016-01-29 02:54:06 -0600693 struct css_set *root_cset;
694};
695
696extern struct cgroup_namespace init_cgroup_ns;
697
698#ifdef CONFIG_CGROUPS
699
700void free_cgroup_ns(struct cgroup_namespace *ns);
701
702struct cgroup_namespace *copy_cgroup_ns(unsigned long flags,
703 struct user_namespace *user_ns,
704 struct cgroup_namespace *old_ns);
705
Tejun Heo4c737b42016-08-10 11:23:44 -0400706int cgroup_path_ns(struct cgroup *cgrp, char *buf, size_t buflen,
707 struct cgroup_namespace *ns);
Aditya Kalia79a9082016-01-29 02:54:06 -0600708
709#else /* !CONFIG_CGROUPS */
710
711static inline void free_cgroup_ns(struct cgroup_namespace *ns) { }
712static inline struct cgroup_namespace *
713copy_cgroup_ns(unsigned long flags, struct user_namespace *user_ns,
714 struct cgroup_namespace *old_ns)
715{
716 return old_ns;
717}
718
719#endif /* !CONFIG_CGROUPS */
720
721static inline void get_cgroup_ns(struct cgroup_namespace *ns)
722{
723 if (ns)
Elena Reshetova387ad962017-02-20 12:19:00 +0200724 refcount_inc(&ns->count);
Aditya Kalia79a9082016-01-29 02:54:06 -0600725}
726
727static inline void put_cgroup_ns(struct cgroup_namespace *ns)
728{
Elena Reshetova387ad962017-02-20 12:19:00 +0200729 if (ns && refcount_dec_and_test(&ns->count))
Aditya Kalia79a9082016-01-29 02:54:06 -0600730 free_cgroup_ns(ns);
731}
732
Paul Menageddbcc7e2007-10-18 23:39:30 -0700733#endif /* _LINUX_CGROUP_H */