blob: d023ac5e377f4bd21b0a6fdc5ae33a100990e32f [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)
Tejun Heo450ee0c2017-05-15 09:34:03 -040041/* walk all threaded css_sets in the domain */
42#define CSS_TASK_ITER_THREADED (1U << 1)
Tejun Heobc2fb7e2017-05-15 09:34:01 -040043
Tejun Heoc326aa22015-05-13 16:24:16 -040044/* a css_task_iter should be treated as an opaque object */
45struct css_task_iter {
46 struct cgroup_subsys *ss;
Tejun Heobc2fb7e2017-05-15 09:34:01 -040047 unsigned int flags;
Paul Menageddbcc7e2007-10-18 23:39:30 -070048
Tejun Heoc326aa22015-05-13 16:24:16 -040049 struct list_head *cset_pos;
50 struct list_head *cset_head;
Paul Menageddbcc7e2007-10-18 23:39:30 -070051
Tejun Heo450ee0c2017-05-15 09:34:03 -040052 struct list_head *tcset_pos;
53 struct list_head *tcset_head;
54
Tejun Heoc326aa22015-05-13 16:24:16 -040055 struct list_head *task_pos;
56 struct list_head *tasks_head;
57 struct list_head *mg_tasks_head;
Tejun Heoed27b9f2015-10-15 16:41:52 -040058
59 struct css_set *cur_cset;
Tejun Heo450ee0c2017-05-15 09:34:03 -040060 struct css_set *cur_dcset;
Tejun Heoed27b9f2015-10-15 16:41:52 -040061 struct task_struct *cur_task;
62 struct list_head iters_node; /* css_set->task_iters */
Paul Menage817929e2007-10-18 23:39:36 -070063};
Tejun Heoc326aa22015-05-13 16:24:16 -040064
65extern struct cgroup_root cgrp_dfl_root;
66extern struct css_set init_css_set;
67
68#define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys;
69#include <linux/cgroup_subsys.h>
Paul Menage817929e2007-10-18 23:39:36 -070070#undef SUBSYS
71
Tejun Heo49d1dc42015-09-18 11:56:28 -040072#define SUBSYS(_x) \
73 extern struct static_key_true _x ## _cgrp_subsys_enabled_key; \
74 extern struct static_key_true _x ## _cgrp_subsys_on_dfl_key;
75#include <linux/cgroup_subsys.h>
76#undef SUBSYS
77
78/**
79 * cgroup_subsys_enabled - fast test on whether a subsys is enabled
80 * @ss: subsystem in question
81 */
82#define cgroup_subsys_enabled(ss) \
83 static_branch_likely(&ss ## _enabled_key)
84
85/**
86 * cgroup_subsys_on_dfl - fast test on whether a subsys is on default hierarchy
87 * @ss: subsystem in question
88 */
89#define cgroup_subsys_on_dfl(ss) \
90 static_branch_likely(&ss ## _on_dfl_key)
91
Tejun Heoc326aa22015-05-13 16:24:16 -040092bool css_has_online_children(struct cgroup_subsys_state *css);
93struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss);
94struct cgroup_subsys_state *cgroup_get_e_css(struct cgroup *cgroup,
95 struct cgroup_subsys *ss);
96struct cgroup_subsys_state *css_tryget_online_from_dir(struct dentry *dentry,
97 struct cgroup_subsys *ss);
98
Tejun Heo16af4392015-11-20 15:55:52 -050099struct cgroup *cgroup_get_from_path(const char *path);
Martin KaFai Lau1f3fe7e2016-06-30 10:28:42 -0700100struct cgroup *cgroup_get_from_fd(int fd);
Tejun Heo16af4392015-11-20 15:55:52 -0500101
Tejun Heoc326aa22015-05-13 16:24:16 -0400102int cgroup_attach_task_all(struct task_struct *from, struct task_struct *);
103int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from);
104
105int cgroup_add_dfl_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
106int cgroup_add_legacy_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
107int cgroup_rm_cftypes(struct cftype *cfts);
Tejun Heo34c06252015-11-05 00:12:24 -0500108void cgroup_file_notify(struct cgroup_file *cfile);
Tejun Heoc326aa22015-05-13 16:24:16 -0400109
Tejun Heo4c737b42016-08-10 11:23:44 -0400110int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen);
Tejun Heoc326aa22015-05-13 16:24:16 -0400111int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry);
112int proc_cgroup_show(struct seq_file *m, struct pid_namespace *ns,
113 struct pid *pid, struct task_struct *tsk);
114
115void cgroup_fork(struct task_struct *p);
Oleg Nesterovb53202e2015-12-03 10:24:08 -0500116extern int cgroup_can_fork(struct task_struct *p);
117extern void cgroup_cancel_fork(struct task_struct *p);
118extern void cgroup_post_fork(struct task_struct *p);
Tejun Heoc326aa22015-05-13 16:24:16 -0400119void cgroup_exit(struct task_struct *p);
Tejun Heo2e91fa72015-10-15 16:41:53 -0400120void cgroup_free(struct task_struct *p);
Tejun Heoc326aa22015-05-13 16:24:16 -0400121
122int cgroup_init_early(void);
123int cgroup_init(void);
124
Tejun Heo5c9d5352014-05-16 13:22:48 -0400125/*
Tejun Heoc326aa22015-05-13 16:24:16 -0400126 * Iteration helpers and macros.
Tejun Heo5c9d5352014-05-16 13:22:48 -0400127 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700128
Tejun Heoc326aa22015-05-13 16:24:16 -0400129struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos,
130 struct cgroup_subsys_state *parent);
131struct cgroup_subsys_state *css_next_descendant_pre(struct cgroup_subsys_state *pos,
132 struct cgroup_subsys_state *css);
133struct cgroup_subsys_state *css_rightmost_descendant(struct cgroup_subsys_state *pos);
134struct cgroup_subsys_state *css_next_descendant_post(struct cgroup_subsys_state *pos,
135 struct cgroup_subsys_state *css);
Tejun Heo72c97e52013-08-08 20:11:22 -0400136
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500137struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset,
138 struct cgroup_subsys_state **dst_cssp);
139struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset,
140 struct cgroup_subsys_state **dst_cssp);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700141
Tejun Heobc2fb7e2017-05-15 09:34:01 -0400142void css_task_iter_start(struct cgroup_subsys_state *css, unsigned int flags,
Tejun Heoc326aa22015-05-13 16:24:16 -0400143 struct css_task_iter *it);
144struct task_struct *css_task_iter_next(struct css_task_iter *it);
145void css_task_iter_end(struct css_task_iter *it);
Tejun Heo0ae78e02013-08-13 11:01:54 -0400146
Tejun Heoc326aa22015-05-13 16:24:16 -0400147/**
148 * css_for_each_child - iterate through children of a css
149 * @pos: the css * to use as the loop cursor
150 * @parent: css whose children to walk
151 *
152 * Walk @parent's children. Must be called under rcu_read_lock().
153 *
154 * If a subsystem synchronizes ->css_online() and the start of iteration, a
155 * css which finished ->css_online() is guaranteed to be visible in the
156 * future iterations and will stay visible until the last reference is put.
157 * A css which hasn't finished ->css_online() or already finished
158 * ->css_offline() may show up during traversal. It's each subsystem's
159 * responsibility to synchronize against on/offlining.
160 *
161 * It is allowed to temporarily drop RCU read lock during iteration. The
162 * caller is responsible for ensuring that @pos remains accessible until
163 * the start of the next iteration by, for example, bumping the css refcnt.
164 */
165#define css_for_each_child(pos, parent) \
166 for ((pos) = css_next_child(NULL, (parent)); (pos); \
167 (pos) = css_next_child((pos), (parent)))
Tejun Heod5c419b2014-05-16 13:22:48 -0400168
Tejun Heoc326aa22015-05-13 16:24:16 -0400169/**
170 * css_for_each_descendant_pre - pre-order walk of a css's descendants
171 * @pos: the css * to use as the loop cursor
172 * @root: css whose descendants to walk
173 *
174 * Walk @root's descendants. @root is included in the iteration and the
175 * first node to be visited. Must be called under rcu_read_lock().
176 *
177 * If a subsystem synchronizes ->css_online() and the start of iteration, a
178 * css which finished ->css_online() is guaranteed to be visible in the
179 * future iterations and will stay visible until the last reference is put.
180 * A css which hasn't finished ->css_online() or already finished
181 * ->css_offline() may show up during traversal. It's each subsystem's
182 * responsibility to synchronize against on/offlining.
183 *
184 * For example, the following guarantees that a descendant can't escape
185 * state updates of its ancestors.
186 *
187 * my_online(@css)
188 * {
189 * Lock @css's parent and @css;
190 * Inherit state from the parent;
191 * Unlock both.
192 * }
193 *
194 * my_update_state(@css)
195 * {
196 * css_for_each_descendant_pre(@pos, @css) {
197 * Lock @pos;
198 * if (@pos == @css)
199 * Update @css's state;
200 * else
201 * Verify @pos is alive and inherit state from its parent;
202 * Unlock @pos;
203 * }
204 * }
205 *
206 * As long as the inheriting step, including checking the parent state, is
207 * enclosed inside @pos locking, double-locking the parent isn't necessary
208 * while inheriting. The state update to the parent is guaranteed to be
209 * visible by walking order and, as long as inheriting operations to the
210 * same @pos are atomic to each other, multiple updates racing each other
211 * still result in the correct state. It's guaranateed that at least one
212 * inheritance happens for any css after the latest update to its parent.
213 *
214 * If checking parent's state requires locking the parent, each inheriting
215 * iteration should lock and unlock both @pos->parent and @pos.
216 *
217 * Alternatively, a subsystem may choose to use a single global lock to
218 * synchronize ->css_online() and ->css_offline() against tree-walking
219 * operations.
220 *
221 * It is allowed to temporarily drop RCU read lock during iteration. The
222 * caller is responsible for ensuring that @pos remains accessible until
223 * the start of the next iteration by, for example, bumping the css refcnt.
224 */
225#define css_for_each_descendant_pre(pos, css) \
226 for ((pos) = css_next_descendant_pre(NULL, (css)); (pos); \
227 (pos) = css_next_descendant_pre((pos), (css)))
Tejun Heo15a4c832014-05-04 15:09:14 -0400228
Tejun Heoc326aa22015-05-13 16:24:16 -0400229/**
230 * css_for_each_descendant_post - post-order walk of a css's descendants
231 * @pos: the css * to use as the loop cursor
232 * @css: css whose descendants to walk
233 *
234 * Similar to css_for_each_descendant_pre() but performs post-order
235 * traversal instead. @root is included in the iteration and the last
236 * node to be visited.
237 *
238 * If a subsystem synchronizes ->css_online() and the start of iteration, a
239 * css which finished ->css_online() is guaranteed to be visible in the
240 * future iterations and will stay visible until the last reference is put.
241 * A css which hasn't finished ->css_online() or already finished
242 * ->css_offline() may show up during traversal. It's each subsystem's
243 * responsibility to synchronize against on/offlining.
244 *
245 * Note that the walk visibility guarantee example described in pre-order
246 * walk doesn't apply the same to post-order walks.
247 */
248#define css_for_each_descendant_post(pos, css) \
249 for ((pos) = css_next_descendant_post(NULL, (css)); (pos); \
250 (pos) = css_next_descendant_post((pos), (css)))
Tejun Heo48ddbe12012-04-01 12:09:56 -0700251
Tejun Heoc326aa22015-05-13 16:24:16 -0400252/**
253 * cgroup_taskset_for_each - iterate cgroup_taskset
254 * @task: the loop cursor
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500255 * @dst_css: the destination css
Tejun Heoc326aa22015-05-13 16:24:16 -0400256 * @tset: taskset to iterate
Tejun Heo4530edd2015-09-11 15:00:19 -0400257 *
258 * @tset may contain multiple tasks and they may belong to multiple
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500259 * processes.
260 *
261 * On the v2 hierarchy, there may be tasks from multiple processes and they
262 * may not share the source or destination csses.
263 *
264 * On traditional hierarchies, when there are multiple tasks in @tset, if a
265 * task of a process is in @tset, all tasks of the process are in @tset.
266 * Also, all are guaranteed to share the same source and destination csses.
Tejun Heo4530edd2015-09-11 15:00:19 -0400267 *
268 * Iteration is not in any specific order.
Tejun Heoc326aa22015-05-13 16:24:16 -0400269 */
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500270#define cgroup_taskset_for_each(task, dst_css, tset) \
271 for ((task) = cgroup_taskset_first((tset), &(dst_css)); \
272 (task); \
273 (task) = cgroup_taskset_next((tset), &(dst_css)))
Tejun Heo0cb51d72014-05-16 13:22:49 -0400274
Tejun Heo4530edd2015-09-11 15:00:19 -0400275/**
276 * cgroup_taskset_for_each_leader - iterate group leaders in a cgroup_taskset
277 * @leader: the loop cursor
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500278 * @dst_css: the destination css
Geliang Tang7b4632f2016-12-24 23:28:35 +0800279 * @tset: taskset to iterate
Tejun Heo4530edd2015-09-11 15:00:19 -0400280 *
281 * Iterate threadgroup leaders of @tset. For single-task migrations, @tset
282 * may not contain any.
283 */
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500284#define cgroup_taskset_for_each_leader(leader, dst_css, tset) \
285 for ((leader) = cgroup_taskset_first((tset), &(dst_css)); \
286 (leader); \
287 (leader) = cgroup_taskset_next((tset), &(dst_css))) \
Tejun Heo4530edd2015-09-11 15:00:19 -0400288 if ((leader) != (leader)->group_leader) \
289 ; \
290 else
291
Tejun Heoc326aa22015-05-13 16:24:16 -0400292/*
293 * Inline functions.
294 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700295
Tejun Heo5de01072013-06-12 21:04:52 -0700296/**
297 * css_get - obtain a reference on the specified css
298 * @css: target css
299 *
300 * The caller must already have a reference.
Paul Menageddbcc7e2007-10-18 23:39:30 -0700301 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700302static inline void css_get(struct cgroup_subsys_state *css)
303{
Tejun Heo3b514d22014-05-16 13:22:47 -0400304 if (!(css->flags & CSS_NO_REF))
305 percpu_ref_get(&css->refcnt);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700306}
Paul Menagee7c5ec92009-01-07 18:08:38 -0800307
Tejun Heo5de01072013-06-12 21:04:52 -0700308/**
Johannes Weinere8ea14c2014-12-10 15:42:42 -0800309 * css_get_many - obtain references on the specified css
310 * @css: target css
311 * @n: number of references to get
312 *
313 * The caller must already have a reference.
314 */
315static inline void css_get_many(struct cgroup_subsys_state *css, unsigned int n)
316{
317 if (!(css->flags & CSS_NO_REF))
318 percpu_ref_get_many(&css->refcnt, n);
319}
320
321/**
Tejun Heo6f4524d2014-05-16 13:22:52 -0400322 * css_tryget - try to obtain a reference on the specified css
323 * @css: target css
324 *
325 * Obtain a reference on @css unless it already has reached zero and is
326 * being released. This function doesn't care whether @css is on or
327 * offline. The caller naturally needs to ensure that @css is accessible
328 * but doesn't have to be holding a reference on it - IOW, RCU protected
329 * access is good enough for this function. Returns %true if a reference
330 * count was successfully obtained; %false otherwise.
331 */
332static inline bool css_tryget(struct cgroup_subsys_state *css)
333{
334 if (!(css->flags & CSS_NO_REF))
335 return percpu_ref_tryget(&css->refcnt);
336 return true;
337}
338
339/**
Tejun Heoec903c02014-05-13 12:11:01 -0400340 * css_tryget_online - try to obtain a reference on the specified css if online
Tejun Heo5de01072013-06-12 21:04:52 -0700341 * @css: target css
342 *
Tejun Heoec903c02014-05-13 12:11:01 -0400343 * Obtain a reference on @css if it's online. The caller naturally needs
344 * to ensure that @css is accessible but doesn't have to be holding a
Tejun Heo5de01072013-06-12 21:04:52 -0700345 * reference on it - IOW, RCU protected access is good enough for this
346 * function. Returns %true if a reference count was successfully obtained;
347 * %false otherwise.
348 */
Tejun Heoec903c02014-05-13 12:11:01 -0400349static inline bool css_tryget_online(struct cgroup_subsys_state *css)
Paul Menagee7c5ec92009-01-07 18:08:38 -0800350{
Tejun Heo3b514d22014-05-16 13:22:47 -0400351 if (!(css->flags & CSS_NO_REF))
352 return percpu_ref_tryget_live(&css->refcnt);
353 return true;
Paul Menagee7c5ec92009-01-07 18:08:38 -0800354}
355
Tejun Heo5de01072013-06-12 21:04:52 -0700356/**
Tejun Heo41c25702017-05-24 12:03:48 -0400357 * css_is_dying - test whether the specified css is dying
358 * @css: target css
359 *
360 * Test whether @css is in the process of offlining or already offline. In
361 * most cases, ->css_online() and ->css_offline() callbacks should be
362 * enough; however, the actual offline operations are RCU delayed and this
363 * test returns %true also when @css is scheduled to be offlined.
364 *
365 * This is useful, for example, when the use case requires synchronous
366 * behavior with respect to cgroup removal. cgroup removal schedules css
367 * offlining but the css can seem alive while the operation is being
368 * delayed. If the delay affects user visible semantics, this test can be
369 * used to resolve the situation.
370 */
371static inline bool css_is_dying(struct cgroup_subsys_state *css)
372{
373 return !(css->flags & CSS_NO_REF) && percpu_ref_is_dying(&css->refcnt);
374}
375
376/**
Tejun Heo5de01072013-06-12 21:04:52 -0700377 * css_put - put a css reference
378 * @css: target css
379 *
Tejun Heoec903c02014-05-13 12:11:01 -0400380 * Put a reference obtained via css_get() and css_tryget_online().
Tejun Heo5de01072013-06-12 21:04:52 -0700381 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700382static inline void css_put(struct cgroup_subsys_state *css)
383{
Tejun Heo3b514d22014-05-16 13:22:47 -0400384 if (!(css->flags & CSS_NO_REF))
385 percpu_ref_put(&css->refcnt);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700386}
387
Johannes Weinere8ea14c2014-12-10 15:42:42 -0800388/**
389 * css_put_many - put css references
390 * @css: target css
391 * @n: number of references to put
392 *
393 * Put references obtained via css_get() and css_tryget_online().
394 */
395static inline void css_put_many(struct cgroup_subsys_state *css, unsigned int n)
396{
397 if (!(css->flags & CSS_NO_REF))
398 percpu_ref_put_many(&css->refcnt, n);
399}
400
Tejun Heo3e489302017-08-11 05:49:01 -0700401static inline void cgroup_get(struct cgroup *cgrp)
402{
403 css_get(&cgrp->self);
404}
405
406static inline bool cgroup_tryget(struct cgroup *cgrp)
407{
408 return css_tryget(&cgrp->self);
409}
410
Tejun Heo16af4392015-11-20 15:55:52 -0500411static inline void cgroup_put(struct cgroup *cgrp)
412{
413 css_put(&cgrp->self);
414}
415
Tejun Heoc326aa22015-05-13 16:24:16 -0400416/**
417 * task_css_set_check - obtain a task's css_set with extra access conditions
418 * @task: the task to obtain css_set for
419 * @__c: extra condition expression to be passed to rcu_dereference_check()
Paul Menageddbcc7e2007-10-18 23:39:30 -0700420 *
Tejun Heoc326aa22015-05-13 16:24:16 -0400421 * A task's css_set is RCU protected, initialized and exited while holding
422 * task_lock(), and can only be modified while holding both cgroup_mutex
423 * and task_lock() while the task is alive. This macro verifies that the
424 * caller is inside proper critical section and returns @task's css_set.
425 *
426 * The caller can also specify additional allowed conditions via @__c, such
427 * as locks used during the cgroup_subsys::attach() methods.
Paul Menageddbcc7e2007-10-18 23:39:30 -0700428 */
Tejun Heoc326aa22015-05-13 16:24:16 -0400429#ifdef CONFIG_PROVE_RCU
430extern struct mutex cgroup_mutex;
Tejun Heof0d9a5f2015-10-15 16:41:53 -0400431extern spinlock_t css_set_lock;
Tejun Heoc326aa22015-05-13 16:24:16 -0400432#define task_css_set_check(task, __c) \
433 rcu_dereference_check((task)->cgroups, \
434 lockdep_is_held(&cgroup_mutex) || \
Tejun Heof0d9a5f2015-10-15 16:41:53 -0400435 lockdep_is_held(&css_set_lock) || \
Tejun Heoc326aa22015-05-13 16:24:16 -0400436 ((task)->flags & PF_EXITING) || (__c))
437#else
438#define task_css_set_check(task, __c) \
439 rcu_dereference((task)->cgroups)
Tejun Heo2bd59d42014-02-11 11:52:49 -0500440#endif
Paul Menageddbcc7e2007-10-18 23:39:30 -0700441
Tejun Heoc326aa22015-05-13 16:24:16 -0400442/**
443 * task_css_check - obtain css for (task, subsys) w/ extra access conds
444 * @task: the target task
445 * @subsys_id: the target subsystem ID
446 * @__c: extra condition expression to be passed to rcu_dereference_check()
447 *
448 * Return the cgroup_subsys_state for the (@task, @subsys_id) pair. The
449 * synchronization rules are the same as task_css_set_check().
450 */
451#define task_css_check(task, subsys_id, __c) \
452 task_css_set_check((task), (__c))->subsys[(subsys_id)]
453
454/**
455 * task_css_set - obtain a task's css_set
456 * @task: the task to obtain css_set for
457 *
458 * See task_css_set_check().
459 */
460static inline struct css_set *task_css_set(struct task_struct *task)
461{
462 return task_css_set_check(task, false);
463}
464
465/**
466 * task_css - obtain css for (task, subsys)
467 * @task: the target task
468 * @subsys_id: the target subsystem ID
469 *
470 * See task_css_check().
471 */
472static inline struct cgroup_subsys_state *task_css(struct task_struct *task,
473 int subsys_id)
474{
475 return task_css_check(task, subsys_id, false);
476}
477
478/**
Linus Torvaldsbbe179f2015-06-26 19:50:04 -0700479 * task_get_css - find and get the css for (task, subsys)
480 * @task: the target task
481 * @subsys_id: the target subsystem ID
482 *
483 * Find the css for the (@task, @subsys_id) combination, increment a
484 * reference on and return it. This function is guaranteed to return a
485 * valid css.
486 */
487static inline struct cgroup_subsys_state *
488task_get_css(struct task_struct *task, int subsys_id)
489{
490 struct cgroup_subsys_state *css;
491
492 rcu_read_lock();
493 while (true) {
494 css = task_css(task, subsys_id);
495 if (likely(css_tryget_online(css)))
496 break;
497 cpu_relax();
498 }
499 rcu_read_unlock();
500 return css;
501}
502
503/**
Tejun Heoc326aa22015-05-13 16:24:16 -0400504 * task_css_is_root - test whether a task belongs to the root css
505 * @task: the target task
506 * @subsys_id: the target subsystem ID
507 *
508 * Test whether @task belongs to the root css on the specified subsystem.
509 * May be invoked in any context.
510 */
511static inline bool task_css_is_root(struct task_struct *task, int subsys_id)
512{
513 return task_css_check(task, subsys_id, true) ==
514 init_css_set.subsys[subsys_id];
515}
516
517static inline struct cgroup *task_cgroup(struct task_struct *task,
518 int subsys_id)
519{
520 return task_css(task, subsys_id)->cgroup;
521}
Tejun Heoa2dd4242014-03-19 10:23:55 -0400522
Tejun Heo3e489302017-08-11 05:49:01 -0700523static inline struct cgroup *task_dfl_cgroup(struct task_struct *task)
524{
525 return task_css_set(task)->dfl_cgrp;
526}
527
528static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
529{
530 struct cgroup_subsys_state *parent_css = cgrp->self.parent;
531
532 if (parent_css)
533 return container_of(parent_css, struct cgroup, self);
534 return NULL;
535}
536
Tejun Heob11cfb52015-11-20 15:55:52 -0500537/**
538 * cgroup_is_descendant - test ancestry
539 * @cgrp: the cgroup to be tested
540 * @ancestor: possible ancestor of @cgrp
541 *
542 * Test whether @cgrp is a descendant of @ancestor. It also returns %true
543 * if @cgrp == @ancestor. This function is safe to call as long as @cgrp
544 * and @ancestor are accessible.
545 */
546static inline bool cgroup_is_descendant(struct cgroup *cgrp,
547 struct cgroup *ancestor)
548{
549 if (cgrp->root != ancestor->root || cgrp->level < ancestor->level)
550 return false;
551 return cgrp->ancestor_ids[ancestor->level] == ancestor->id;
552}
553
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700554/**
555 * task_under_cgroup_hierarchy - test task's membership of cgroup ancestry
556 * @task: the task to be tested
557 * @ancestor: possible ancestor of @task's cgroup
558 *
559 * Tests whether @task's default cgroup hierarchy is a descendant of @ancestor.
560 * It follows all the same rules as cgroup_is_descendant, and only applies
561 * to the default hierarchy.
562 */
563static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
564 struct cgroup *ancestor)
565{
566 struct css_set *cset = task_css_set(task);
567
568 return cgroup_is_descendant(cset->dfl_cgrp, ancestor);
569}
570
Tejun Heo07bc3562014-02-13 06:58:39 -0500571/* no synchronization, the result can only be used as a hint */
Tejun Heo27bd4db2015-10-15 16:41:50 -0400572static inline bool cgroup_is_populated(struct cgroup *cgrp)
Tejun Heo07bc3562014-02-13 06:58:39 -0500573{
Tejun Heo454000a2017-05-15 09:34:02 -0400574 return cgrp->nr_populated_csets + cgrp->nr_populated_domain_children +
575 cgrp->nr_populated_threaded_children;
Tejun Heo07bc3562014-02-13 06:58:39 -0500576}
577
Zefan Lif29374b2014-09-19 16:29:31 +0800578/* returns ino associated with a cgroup */
Tejun Heob1664922014-02-11 11:52:49 -0500579static inline ino_t cgroup_ino(struct cgroup *cgrp)
580{
Shaohua Lic53cd492017-07-12 11:49:50 -0700581 return cgrp->kn->id.ino;
Tejun Heob1664922014-02-11 11:52:49 -0500582}
583
Tejun Heob4168642014-05-13 12:16:21 -0400584/* cft/css accessors for cftype->write() operation */
585static inline struct cftype *of_cft(struct kernfs_open_file *of)
Tejun Heo7da11272013-12-05 12:28:04 -0500586{
Tejun Heo2bd59d42014-02-11 11:52:49 -0500587 return of->kn->priv;
Tejun Heo7da11272013-12-05 12:28:04 -0500588}
589
Tejun Heob4168642014-05-13 12:16:21 -0400590struct cgroup_subsys_state *of_css(struct kernfs_open_file *of);
591
592/* cft/css accessors for cftype->seq_*() operations */
593static inline struct cftype *seq_cft(struct seq_file *seq)
594{
595 return of_cft(seq->private);
596}
597
598static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq)
599{
600 return of_css(seq->private);
601}
Tejun Heo59f52962014-02-11 11:52:49 -0500602
Tejun Heoe61734c2014-02-12 09:29:50 -0500603/*
604 * Name / path handling functions. All are thin wrappers around the kernfs
605 * counterparts and can be called under any context.
606 */
607
608static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen)
609{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400610 return kernfs_name(cgrp->kn, buf, buflen);
Tejun Heoe61734c2014-02-12 09:29:50 -0500611}
612
Tejun Heo4c737b42016-08-10 11:23:44 -0400613static inline int cgroup_path(struct cgroup *cgrp, char *buf, size_t buflen)
Tejun Heoe61734c2014-02-12 09:29:50 -0500614{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400615 return kernfs_path(cgrp->kn, buf, buflen);
Tejun Heoe61734c2014-02-12 09:29:50 -0500616}
617
618static inline void pr_cont_cgroup_name(struct cgroup *cgrp)
619{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400620 pr_cont_kernfs_name(cgrp->kn);
Tejun Heoe61734c2014-02-12 09:29:50 -0500621}
622
623static inline void pr_cont_cgroup_path(struct cgroup *cgrp)
624{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400625 pr_cont_kernfs_path(cgrp->kn);
Tejun Heoe61734c2014-02-12 09:29:50 -0500626}
627
Tejun Heo77f88792017-03-16 16:54:24 -0400628static inline void cgroup_init_kthreadd(void)
629{
630 /*
631 * kthreadd is inherited by all kthreads, keep it in the root so
632 * that the new kthreads are guaranteed to stay in the root until
633 * initialization is finished.
634 */
635 current->no_cgroup_migration = 1;
636}
637
638static inline void cgroup_kthread_ready(void)
639{
640 /*
641 * This kthread finished initialization. The creator should have
642 * set PF_NO_SETAFFINITY if this kthread should stay in the root.
643 */
644 current->no_cgroup_migration = 0;
645}
646
Shaohua Li121508d2017-07-12 11:49:52 -0700647static inline union kernfs_node_id *cgroup_get_kernfs_id(struct cgroup *cgrp)
648{
649 return &cgrp->kn->id;
650}
Shaohua Li69fd5c32017-07-12 11:49:55 -0700651
652void cgroup_path_from_kernfs_id(const union kernfs_node_id *id,
653 char *buf, size_t buflen);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700654#else /* !CONFIG_CGROUPS */
655
Tejun Heof3ba5382015-01-06 12:02:46 -0500656struct cgroup_subsys_state;
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700657struct cgroup;
Tejun Heof3ba5382015-01-06 12:02:46 -0500658
Tejun Heoc326aa22015-05-13 16:24:16 -0400659static inline void css_put(struct cgroup_subsys_state *css) {}
660static inline int cgroup_attach_task_all(struct task_struct *from,
661 struct task_struct *t) { return 0; }
662static inline int cgroupstats_build(struct cgroupstats *stats,
663 struct dentry *dentry) { return -EINVAL; }
664
Paul Menageb4f48b62007-10-18 23:39:33 -0700665static inline void cgroup_fork(struct task_struct *p) {}
Oleg Nesterovb53202e2015-12-03 10:24:08 -0500666static inline int cgroup_can_fork(struct task_struct *p) { return 0; }
667static inline void cgroup_cancel_fork(struct task_struct *p) {}
668static inline void cgroup_post_fork(struct task_struct *p) {}
Li Zefan1ec41832014-03-28 15:22:19 +0800669static inline void cgroup_exit(struct task_struct *p) {}
Tejun Heo2e91fa72015-10-15 16:41:53 -0400670static inline void cgroup_free(struct task_struct *p) {}
Paul Menageddbcc7e2007-10-18 23:39:30 -0700671
Tejun Heoc326aa22015-05-13 16:24:16 -0400672static inline int cgroup_init_early(void) { return 0; }
673static inline int cgroup_init(void) { return 0; }
Tejun Heo77f88792017-03-16 16:54:24 -0400674static inline void cgroup_init_kthreadd(void) {}
675static inline void cgroup_kthread_ready(void) {}
Shaohua Li121508d2017-07-12 11:49:52 -0700676static inline union kernfs_node_id *cgroup_get_kernfs_id(struct cgroup *cgrp)
677{
678 return NULL;
679}
Sridhar Samudralad7926ee2010-05-30 22:24:39 +0200680
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700681static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
682 struct cgroup *ancestor)
683{
684 return true;
685}
Shaohua Li69fd5c32017-07-12 11:49:55 -0700686
687static inline void cgroup_path_from_kernfs_id(const union kernfs_node_id *id,
688 char *buf, size_t buflen) {}
Paul Menageddbcc7e2007-10-18 23:39:30 -0700689#endif /* !CONFIG_CGROUPS */
690
Tejun Heobd1060a2015-12-07 17:38:53 -0500691/*
692 * sock->sk_cgrp_data handling. For more info, see sock_cgroup_data
693 * definition in cgroup-defs.h.
694 */
695#ifdef CONFIG_SOCK_CGROUP_DATA
696
697#if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
698extern spinlock_t cgroup_sk_update_lock;
699#endif
700
701void cgroup_sk_alloc_disable(void);
702void cgroup_sk_alloc(struct sock_cgroup_data *skcd);
703void cgroup_sk_free(struct sock_cgroup_data *skcd);
704
705static inline struct cgroup *sock_cgroup_ptr(struct sock_cgroup_data *skcd)
706{
707#if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
708 unsigned long v;
709
710 /*
711 * @skcd->val is 64bit but the following is safe on 32bit too as we
712 * just need the lower ulong to be written and read atomically.
713 */
714 v = READ_ONCE(skcd->val);
715
716 if (v & 1)
717 return &cgrp_dfl_root.cgrp;
718
719 return (struct cgroup *)(unsigned long)v ?: &cgrp_dfl_root.cgrp;
720#else
721 return (struct cgroup *)(unsigned long)skcd->val;
722#endif
723}
724
725#else /* CONFIG_CGROUP_DATA */
726
727static inline void cgroup_sk_alloc(struct sock_cgroup_data *skcd) {}
728static inline void cgroup_sk_free(struct sock_cgroup_data *skcd) {}
729
730#endif /* CONFIG_CGROUP_DATA */
731
Aditya Kalia79a9082016-01-29 02:54:06 -0600732struct cgroup_namespace {
Elena Reshetova387ad962017-02-20 12:19:00 +0200733 refcount_t count;
Aditya Kalia79a9082016-01-29 02:54:06 -0600734 struct ns_common ns;
735 struct user_namespace *user_ns;
Eric W. Biedermand08311d2016-08-08 14:25:30 -0500736 struct ucounts *ucounts;
Aditya Kalia79a9082016-01-29 02:54:06 -0600737 struct css_set *root_cset;
738};
739
740extern struct cgroup_namespace init_cgroup_ns;
741
742#ifdef CONFIG_CGROUPS
743
744void free_cgroup_ns(struct cgroup_namespace *ns);
745
746struct cgroup_namespace *copy_cgroup_ns(unsigned long flags,
747 struct user_namespace *user_ns,
748 struct cgroup_namespace *old_ns);
749
Tejun Heo4c737b42016-08-10 11:23:44 -0400750int cgroup_path_ns(struct cgroup *cgrp, char *buf, size_t buflen,
751 struct cgroup_namespace *ns);
Aditya Kalia79a9082016-01-29 02:54:06 -0600752
753#else /* !CONFIG_CGROUPS */
754
755static inline void free_cgroup_ns(struct cgroup_namespace *ns) { }
756static inline struct cgroup_namespace *
757copy_cgroup_ns(unsigned long flags, struct user_namespace *user_ns,
758 struct cgroup_namespace *old_ns)
759{
760 return old_ns;
761}
762
763#endif /* !CONFIG_CGROUPS */
764
765static inline void get_cgroup_ns(struct cgroup_namespace *ns)
766{
767 if (ns)
Elena Reshetova387ad962017-02-20 12:19:00 +0200768 refcount_inc(&ns->count);
Aditya Kalia79a9082016-01-29 02:54:06 -0600769}
770
771static inline void put_cgroup_ns(struct cgroup_namespace *ns)
772{
Elena Reshetova387ad962017-02-20 12:19:00 +0200773 if (ns && refcount_dec_and_test(&ns->count))
Aditya Kalia79a9082016-01-29 02:54:06 -0600774 free_cgroup_ns(ns);
775}
776
Paul Menageddbcc7e2007-10-18 23:39:30 -0700777#endif /* _LINUX_CGROUP_H */