blob: dddbc29e20098e0a9f0377f64b0945c87691d315 [file] [log] [blame]
Greg Kroah-Hartmanb2441312017-11-01 15:07:57 +01001/* SPDX-License-Identifier: GPL-2.0 */
Paul Menageddbcc7e2007-10-18 23:39:30 -07002#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>
Paul Menageddbcc7e2007-10-18 23:39:30 -070013#include <linux/cpumask.h>
14#include <linux/nodemask.h>
Tejun Heoeb6fd502012-11-09 09:12:29 -080015#include <linux/rculist.h>
Balbir Singh846c7bb2007-10-18 23:39:44 -070016#include <linux/cgroupstats.h>
Tejun Heo25a7e682013-04-14 20:15:25 -070017#include <linux/fs.h>
Tejun Heo7da11272013-12-05 12:28:04 -050018#include <linux/seq_file.h>
Tejun Heo2bd59d42014-02-11 11:52:49 -050019#include <linux/kernfs.h>
Tejun Heo49d1dc42015-09-18 11:56:28 -040020#include <linux/jump_label.h>
Aditya Kalia79a9082016-01-29 02:54:06 -060021#include <linux/types.h>
22#include <linux/ns_common.h>
23#include <linux/nsproxy.h>
24#include <linux/user_namespace.h>
Elena Reshetova387ad962017-02-20 12:19:00 +020025#include <linux/refcount.h>
Tejun Heob4a04ab2015-05-13 15:38:40 -040026
27#include <linux/cgroup-defs.h>
Paul Menageddbcc7e2007-10-18 23:39:30 -070028
29#ifdef CONFIG_CGROUPS
30
Tejun Heo6abc8ca2015-08-04 15:20:55 -040031/*
32 * All weight knobs on the default hierarhcy should use the following min,
33 * default and max values. The default value is the logarithmic center of
34 * MIN and MAX and allows 100x to be expressed in both directions.
35 */
36#define CGROUP_WEIGHT_MIN 1
37#define CGROUP_WEIGHT_DFL 100
38#define CGROUP_WEIGHT_MAX 10000
39
Tejun Heobc2fb7e2017-05-15 09:34:01 -040040/* walk only threadgroup leaders */
41#define CSS_TASK_ITER_PROCS (1U << 0)
Tejun Heo450ee0c2017-05-15 09:34:03 -040042/* walk all threaded css_sets in the domain */
43#define CSS_TASK_ITER_THREADED (1U << 1)
Tejun Heobc2fb7e2017-05-15 09:34:01 -040044
Tejun Heoc326aa22015-05-13 16:24:16 -040045/* a css_task_iter should be treated as an opaque object */
46struct css_task_iter {
47 struct cgroup_subsys *ss;
Tejun Heobc2fb7e2017-05-15 09:34:01 -040048 unsigned int flags;
Paul Menageddbcc7e2007-10-18 23:39:30 -070049
Tejun Heoc326aa22015-05-13 16:24:16 -040050 struct list_head *cset_pos;
51 struct list_head *cset_head;
Paul Menageddbcc7e2007-10-18 23:39:30 -070052
Tejun Heo450ee0c2017-05-15 09:34:03 -040053 struct list_head *tcset_pos;
54 struct list_head *tcset_head;
55
Tejun Heoc326aa22015-05-13 16:24:16 -040056 struct list_head *task_pos;
57 struct list_head *tasks_head;
58 struct list_head *mg_tasks_head;
Tejun Heoed27b9f2015-10-15 16:41:52 -040059
60 struct css_set *cur_cset;
Tejun Heo450ee0c2017-05-15 09:34:03 -040061 struct css_set *cur_dcset;
Tejun Heoed27b9f2015-10-15 16:41:52 -040062 struct task_struct *cur_task;
63 struct list_head iters_node; /* css_set->task_iters */
Paul Menage817929e2007-10-18 23:39:36 -070064};
Tejun Heoc326aa22015-05-13 16:24:16 -040065
66extern struct cgroup_root cgrp_dfl_root;
67extern struct css_set init_css_set;
68
69#define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys;
70#include <linux/cgroup_subsys.h>
Paul Menage817929e2007-10-18 23:39:36 -070071#undef SUBSYS
72
Tejun Heo49d1dc42015-09-18 11:56:28 -040073#define SUBSYS(_x) \
74 extern struct static_key_true _x ## _cgrp_subsys_enabled_key; \
75 extern struct static_key_true _x ## _cgrp_subsys_on_dfl_key;
76#include <linux/cgroup_subsys.h>
77#undef SUBSYS
78
79/**
80 * cgroup_subsys_enabled - fast test on whether a subsys is enabled
81 * @ss: subsystem in question
82 */
83#define cgroup_subsys_enabled(ss) \
84 static_branch_likely(&ss ## _enabled_key)
85
86/**
87 * cgroup_subsys_on_dfl - fast test on whether a subsys is on default hierarchy
88 * @ss: subsystem in question
89 */
90#define cgroup_subsys_on_dfl(ss) \
91 static_branch_likely(&ss ## _on_dfl_key)
92
Tejun Heoc326aa22015-05-13 16:24:16 -040093bool css_has_online_children(struct cgroup_subsys_state *css);
94struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss);
95struct cgroup_subsys_state *cgroup_get_e_css(struct cgroup *cgroup,
96 struct cgroup_subsys *ss);
97struct cgroup_subsys_state *css_tryget_online_from_dir(struct dentry *dentry,
98 struct cgroup_subsys *ss);
99
Tejun Heo16af4392015-11-20 15:55:52 -0500100struct cgroup *cgroup_get_from_path(const char *path);
Martin KaFai Lau1f3fe7e2016-06-30 10:28:42 -0700101struct cgroup *cgroup_get_from_fd(int fd);
Tejun Heo16af4392015-11-20 15:55:52 -0500102
Tejun Heoc326aa22015-05-13 16:24:16 -0400103int cgroup_attach_task_all(struct task_struct *from, struct task_struct *);
104int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from);
105
106int cgroup_add_dfl_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
107int cgroup_add_legacy_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
108int cgroup_rm_cftypes(struct cftype *cfts);
Tejun Heo34c06252015-11-05 00:12:24 -0500109void cgroup_file_notify(struct cgroup_file *cfile);
Tejun Heoc326aa22015-05-13 16:24:16 -0400110
Tejun Heo4c737b42016-08-10 11:23:44 -0400111int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen);
Tejun Heoc326aa22015-05-13 16:24:16 -0400112int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry);
113int proc_cgroup_show(struct seq_file *m, struct pid_namespace *ns,
114 struct pid *pid, struct task_struct *tsk);
115
116void cgroup_fork(struct task_struct *p);
Oleg Nesterovb53202e2015-12-03 10:24:08 -0500117extern int cgroup_can_fork(struct task_struct *p);
118extern void cgroup_cancel_fork(struct task_struct *p);
119extern void cgroup_post_fork(struct task_struct *p);
Tejun Heoc326aa22015-05-13 16:24:16 -0400120void cgroup_exit(struct task_struct *p);
Tejun Heo2e91fa72015-10-15 16:41:53 -0400121void cgroup_free(struct task_struct *p);
Tejun Heoc326aa22015-05-13 16:24:16 -0400122
123int cgroup_init_early(void);
124int cgroup_init(void);
125
Tejun Heo5c9d5352014-05-16 13:22:48 -0400126/*
Tejun Heoc326aa22015-05-13 16:24:16 -0400127 * Iteration helpers and macros.
Tejun Heo5c9d5352014-05-16 13:22:48 -0400128 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700129
Tejun Heoc326aa22015-05-13 16:24:16 -0400130struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos,
131 struct cgroup_subsys_state *parent);
132struct cgroup_subsys_state *css_next_descendant_pre(struct cgroup_subsys_state *pos,
133 struct cgroup_subsys_state *css);
134struct cgroup_subsys_state *css_rightmost_descendant(struct cgroup_subsys_state *pos);
135struct cgroup_subsys_state *css_next_descendant_post(struct cgroup_subsys_state *pos,
136 struct cgroup_subsys_state *css);
Tejun Heo72c97e52013-08-08 20:11:22 -0400137
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500138struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset,
139 struct cgroup_subsys_state **dst_cssp);
140struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset,
141 struct cgroup_subsys_state **dst_cssp);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700142
Tejun Heobc2fb7e2017-05-15 09:34:01 -0400143void css_task_iter_start(struct cgroup_subsys_state *css, unsigned int flags,
Tejun Heoc326aa22015-05-13 16:24:16 -0400144 struct css_task_iter *it);
145struct task_struct *css_task_iter_next(struct css_task_iter *it);
146void css_task_iter_end(struct css_task_iter *it);
Tejun Heo0ae78e02013-08-13 11:01:54 -0400147
Tejun Heoc326aa22015-05-13 16:24:16 -0400148/**
149 * css_for_each_child - iterate through children of a css
150 * @pos: the css * to use as the loop cursor
151 * @parent: css whose children to walk
152 *
153 * Walk @parent's children. Must be called under rcu_read_lock().
154 *
155 * If a subsystem synchronizes ->css_online() and the start of iteration, a
156 * css which finished ->css_online() is guaranteed to be visible in the
157 * future iterations and will stay visible until the last reference is put.
158 * A css which hasn't finished ->css_online() or already finished
159 * ->css_offline() may show up during traversal. It's each subsystem's
160 * responsibility to synchronize against on/offlining.
161 *
162 * It is allowed to temporarily drop RCU read lock during iteration. The
163 * caller is responsible for ensuring that @pos remains accessible until
164 * the start of the next iteration by, for example, bumping the css refcnt.
165 */
166#define css_for_each_child(pos, parent) \
167 for ((pos) = css_next_child(NULL, (parent)); (pos); \
168 (pos) = css_next_child((pos), (parent)))
Tejun Heod5c419b2014-05-16 13:22:48 -0400169
Tejun Heoc326aa22015-05-13 16:24:16 -0400170/**
171 * css_for_each_descendant_pre - pre-order walk of a css's descendants
172 * @pos: the css * to use as the loop cursor
173 * @root: css whose descendants to walk
174 *
175 * Walk @root's descendants. @root is included in the iteration and the
176 * first node to be visited. Must be called under rcu_read_lock().
177 *
178 * If a subsystem synchronizes ->css_online() and the start of iteration, a
179 * css which finished ->css_online() is guaranteed to be visible in the
180 * future iterations and will stay visible until the last reference is put.
181 * A css which hasn't finished ->css_online() or already finished
182 * ->css_offline() may show up during traversal. It's each subsystem's
183 * responsibility to synchronize against on/offlining.
184 *
185 * For example, the following guarantees that a descendant can't escape
186 * state updates of its ancestors.
187 *
188 * my_online(@css)
189 * {
190 * Lock @css's parent and @css;
191 * Inherit state from the parent;
192 * Unlock both.
193 * }
194 *
195 * my_update_state(@css)
196 * {
197 * css_for_each_descendant_pre(@pos, @css) {
198 * Lock @pos;
199 * if (@pos == @css)
200 * Update @css's state;
201 * else
202 * Verify @pos is alive and inherit state from its parent;
203 * Unlock @pos;
204 * }
205 * }
206 *
207 * As long as the inheriting step, including checking the parent state, is
208 * enclosed inside @pos locking, double-locking the parent isn't necessary
209 * while inheriting. The state update to the parent is guaranteed to be
210 * visible by walking order and, as long as inheriting operations to the
211 * same @pos are atomic to each other, multiple updates racing each other
212 * still result in the correct state. It's guaranateed that at least one
213 * inheritance happens for any css after the latest update to its parent.
214 *
215 * If checking parent's state requires locking the parent, each inheriting
216 * iteration should lock and unlock both @pos->parent and @pos.
217 *
218 * Alternatively, a subsystem may choose to use a single global lock to
219 * synchronize ->css_online() and ->css_offline() against tree-walking
220 * operations.
221 *
222 * It is allowed to temporarily drop RCU read lock during iteration. The
223 * caller is responsible for ensuring that @pos remains accessible until
224 * the start of the next iteration by, for example, bumping the css refcnt.
225 */
226#define css_for_each_descendant_pre(pos, css) \
227 for ((pos) = css_next_descendant_pre(NULL, (css)); (pos); \
228 (pos) = css_next_descendant_pre((pos), (css)))
Tejun Heo15a4c832014-05-04 15:09:14 -0400229
Tejun Heoc326aa22015-05-13 16:24:16 -0400230/**
231 * css_for_each_descendant_post - post-order walk of a css's descendants
232 * @pos: the css * to use as the loop cursor
233 * @css: css whose descendants to walk
234 *
235 * Similar to css_for_each_descendant_pre() but performs post-order
236 * traversal instead. @root is included in the iteration and the last
237 * node to be visited.
238 *
239 * If a subsystem synchronizes ->css_online() and the start of iteration, a
240 * css which finished ->css_online() is guaranteed to be visible in the
241 * future iterations and will stay visible until the last reference is put.
242 * A css which hasn't finished ->css_online() or already finished
243 * ->css_offline() may show up during traversal. It's each subsystem's
244 * responsibility to synchronize against on/offlining.
245 *
246 * Note that the walk visibility guarantee example described in pre-order
247 * walk doesn't apply the same to post-order walks.
248 */
249#define css_for_each_descendant_post(pos, css) \
250 for ((pos) = css_next_descendant_post(NULL, (css)); (pos); \
251 (pos) = css_next_descendant_post((pos), (css)))
Tejun Heo48ddbe12012-04-01 12:09:56 -0700252
Tejun Heoc326aa22015-05-13 16:24:16 -0400253/**
254 * cgroup_taskset_for_each - iterate cgroup_taskset
255 * @task: the loop cursor
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500256 * @dst_css: the destination css
Tejun Heoc326aa22015-05-13 16:24:16 -0400257 * @tset: taskset to iterate
Tejun Heo4530edd2015-09-11 15:00:19 -0400258 *
259 * @tset may contain multiple tasks and they may belong to multiple
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500260 * processes.
261 *
262 * On the v2 hierarchy, there may be tasks from multiple processes and they
263 * may not share the source or destination csses.
264 *
265 * On traditional hierarchies, when there are multiple tasks in @tset, if a
266 * task of a process is in @tset, all tasks of the process are in @tset.
267 * Also, all are guaranteed to share the same source and destination csses.
Tejun Heo4530edd2015-09-11 15:00:19 -0400268 *
269 * Iteration is not in any specific order.
Tejun Heoc326aa22015-05-13 16:24:16 -0400270 */
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500271#define cgroup_taskset_for_each(task, dst_css, tset) \
272 for ((task) = cgroup_taskset_first((tset), &(dst_css)); \
273 (task); \
274 (task) = cgroup_taskset_next((tset), &(dst_css)))
Tejun Heo0cb51d72014-05-16 13:22:49 -0400275
Tejun Heo4530edd2015-09-11 15:00:19 -0400276/**
277 * cgroup_taskset_for_each_leader - iterate group leaders in a cgroup_taskset
278 * @leader: the loop cursor
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500279 * @dst_css: the destination css
Geliang Tang7b4632f2016-12-24 23:28:35 +0800280 * @tset: taskset to iterate
Tejun Heo4530edd2015-09-11 15:00:19 -0400281 *
282 * Iterate threadgroup leaders of @tset. For single-task migrations, @tset
283 * may not contain any.
284 */
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500285#define cgroup_taskset_for_each_leader(leader, dst_css, tset) \
286 for ((leader) = cgroup_taskset_first((tset), &(dst_css)); \
287 (leader); \
288 (leader) = cgroup_taskset_next((tset), &(dst_css))) \
Tejun Heo4530edd2015-09-11 15:00:19 -0400289 if ((leader) != (leader)->group_leader) \
290 ; \
291 else
292
Tejun Heoc326aa22015-05-13 16:24:16 -0400293/*
294 * Inline functions.
295 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700296
Tejun Heo5de01072013-06-12 21:04:52 -0700297/**
298 * css_get - obtain a reference on the specified css
299 * @css: target css
300 *
301 * The caller must already have a reference.
Paul Menageddbcc7e2007-10-18 23:39:30 -0700302 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700303static inline void css_get(struct cgroup_subsys_state *css)
304{
Tejun Heo3b514d22014-05-16 13:22:47 -0400305 if (!(css->flags & CSS_NO_REF))
306 percpu_ref_get(&css->refcnt);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700307}
Paul Menagee7c5ec92009-01-07 18:08:38 -0800308
Tejun Heo5de01072013-06-12 21:04:52 -0700309/**
Johannes Weinere8ea14c2014-12-10 15:42:42 -0800310 * css_get_many - obtain references on the specified css
311 * @css: target css
312 * @n: number of references to get
313 *
314 * The caller must already have a reference.
315 */
316static inline void css_get_many(struct cgroup_subsys_state *css, unsigned int n)
317{
318 if (!(css->flags & CSS_NO_REF))
319 percpu_ref_get_many(&css->refcnt, n);
320}
321
322/**
Tejun Heo6f4524d2014-05-16 13:22:52 -0400323 * css_tryget - try to obtain a reference on the specified css
324 * @css: target css
325 *
326 * Obtain a reference on @css unless it already has reached zero and is
327 * being released. This function doesn't care whether @css is on or
328 * offline. The caller naturally needs to ensure that @css is accessible
329 * but doesn't have to be holding a reference on it - IOW, RCU protected
330 * access is good enough for this function. Returns %true if a reference
331 * count was successfully obtained; %false otherwise.
332 */
333static inline bool css_tryget(struct cgroup_subsys_state *css)
334{
335 if (!(css->flags & CSS_NO_REF))
336 return percpu_ref_tryget(&css->refcnt);
337 return true;
338}
339
340/**
Tejun Heoec903c02014-05-13 12:11:01 -0400341 * css_tryget_online - try to obtain a reference on the specified css if online
Tejun Heo5de01072013-06-12 21:04:52 -0700342 * @css: target css
343 *
Tejun Heoec903c02014-05-13 12:11:01 -0400344 * Obtain a reference on @css if it's online. The caller naturally needs
345 * to ensure that @css is accessible but doesn't have to be holding a
Tejun Heo5de01072013-06-12 21:04:52 -0700346 * reference on it - IOW, RCU protected access is good enough for this
347 * function. Returns %true if a reference count was successfully obtained;
348 * %false otherwise.
349 */
Tejun Heoec903c02014-05-13 12:11:01 -0400350static inline bool css_tryget_online(struct cgroup_subsys_state *css)
Paul Menagee7c5ec92009-01-07 18:08:38 -0800351{
Tejun Heo3b514d22014-05-16 13:22:47 -0400352 if (!(css->flags & CSS_NO_REF))
353 return percpu_ref_tryget_live(&css->refcnt);
354 return true;
Paul Menagee7c5ec92009-01-07 18:08:38 -0800355}
356
Tejun Heo5de01072013-06-12 21:04:52 -0700357/**
Tejun Heo41c25702017-05-24 12:03:48 -0400358 * css_is_dying - test whether the specified css is dying
359 * @css: target css
360 *
361 * Test whether @css is in the process of offlining or already offline. In
362 * most cases, ->css_online() and ->css_offline() callbacks should be
363 * enough; however, the actual offline operations are RCU delayed and this
364 * test returns %true also when @css is scheduled to be offlined.
365 *
366 * This is useful, for example, when the use case requires synchronous
367 * behavior with respect to cgroup removal. cgroup removal schedules css
368 * offlining but the css can seem alive while the operation is being
369 * delayed. If the delay affects user visible semantics, this test can be
370 * used to resolve the situation.
371 */
372static inline bool css_is_dying(struct cgroup_subsys_state *css)
373{
374 return !(css->flags & CSS_NO_REF) && percpu_ref_is_dying(&css->refcnt);
375}
376
377/**
Tejun Heo5de01072013-06-12 21:04:52 -0700378 * css_put - put a css reference
379 * @css: target css
380 *
Tejun Heoec903c02014-05-13 12:11:01 -0400381 * Put a reference obtained via css_get() and css_tryget_online().
Tejun Heo5de01072013-06-12 21:04:52 -0700382 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700383static inline void css_put(struct cgroup_subsys_state *css)
384{
Tejun Heo3b514d22014-05-16 13:22:47 -0400385 if (!(css->flags & CSS_NO_REF))
386 percpu_ref_put(&css->refcnt);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700387}
388
Johannes Weinere8ea14c2014-12-10 15:42:42 -0800389/**
390 * css_put_many - put css references
391 * @css: target css
392 * @n: number of references to put
393 *
394 * Put references obtained via css_get() and css_tryget_online().
395 */
396static inline void css_put_many(struct cgroup_subsys_state *css, unsigned int n)
397{
398 if (!(css->flags & CSS_NO_REF))
399 percpu_ref_put_many(&css->refcnt, n);
400}
401
Tejun Heo3e489302017-08-11 05:49:01 -0700402static inline void cgroup_get(struct cgroup *cgrp)
403{
404 css_get(&cgrp->self);
405}
406
407static inline bool cgroup_tryget(struct cgroup *cgrp)
408{
409 return css_tryget(&cgrp->self);
410}
411
Tejun Heo16af4392015-11-20 15:55:52 -0500412static inline void cgroup_put(struct cgroup *cgrp)
413{
414 css_put(&cgrp->self);
415}
416
Tejun Heoc326aa22015-05-13 16:24:16 -0400417/**
418 * task_css_set_check - obtain a task's css_set with extra access conditions
419 * @task: the task to obtain css_set for
420 * @__c: extra condition expression to be passed to rcu_dereference_check()
Paul Menageddbcc7e2007-10-18 23:39:30 -0700421 *
Tejun Heoc326aa22015-05-13 16:24:16 -0400422 * A task's css_set is RCU protected, initialized and exited while holding
423 * task_lock(), and can only be modified while holding both cgroup_mutex
424 * and task_lock() while the task is alive. This macro verifies that the
425 * caller is inside proper critical section and returns @task's css_set.
426 *
427 * The caller can also specify additional allowed conditions via @__c, such
428 * as locks used during the cgroup_subsys::attach() methods.
Paul Menageddbcc7e2007-10-18 23:39:30 -0700429 */
Tejun Heoc326aa22015-05-13 16:24:16 -0400430#ifdef CONFIG_PROVE_RCU
431extern struct mutex cgroup_mutex;
Tejun Heof0d9a5f2015-10-15 16:41:53 -0400432extern spinlock_t css_set_lock;
Tejun Heoc326aa22015-05-13 16:24:16 -0400433#define task_css_set_check(task, __c) \
434 rcu_dereference_check((task)->cgroups, \
435 lockdep_is_held(&cgroup_mutex) || \
Tejun Heof0d9a5f2015-10-15 16:41:53 -0400436 lockdep_is_held(&css_set_lock) || \
Tejun Heoc326aa22015-05-13 16:24:16 -0400437 ((task)->flags & PF_EXITING) || (__c))
438#else
439#define task_css_set_check(task, __c) \
440 rcu_dereference((task)->cgroups)
Tejun Heo2bd59d42014-02-11 11:52:49 -0500441#endif
Paul Menageddbcc7e2007-10-18 23:39:30 -0700442
Tejun Heoc326aa22015-05-13 16:24:16 -0400443/**
444 * task_css_check - obtain css for (task, subsys) w/ extra access conds
445 * @task: the target task
446 * @subsys_id: the target subsystem ID
447 * @__c: extra condition expression to be passed to rcu_dereference_check()
448 *
449 * Return the cgroup_subsys_state for the (@task, @subsys_id) pair. The
450 * synchronization rules are the same as task_css_set_check().
451 */
452#define task_css_check(task, subsys_id, __c) \
453 task_css_set_check((task), (__c))->subsys[(subsys_id)]
454
455/**
456 * task_css_set - obtain a task's css_set
457 * @task: the task to obtain css_set for
458 *
459 * See task_css_set_check().
460 */
461static inline struct css_set *task_css_set(struct task_struct *task)
462{
463 return task_css_set_check(task, false);
464}
465
466/**
467 * task_css - obtain css for (task, subsys)
468 * @task: the target task
469 * @subsys_id: the target subsystem ID
470 *
471 * See task_css_check().
472 */
473static inline struct cgroup_subsys_state *task_css(struct task_struct *task,
474 int subsys_id)
475{
476 return task_css_check(task, subsys_id, false);
477}
478
479/**
Linus Torvaldsbbe179f2015-06-26 19:50:04 -0700480 * task_get_css - find and get the css for (task, subsys)
481 * @task: the target task
482 * @subsys_id: the target subsystem ID
483 *
484 * Find the css for the (@task, @subsys_id) combination, increment a
485 * reference on and return it. This function is guaranteed to return a
486 * valid css.
487 */
488static inline struct cgroup_subsys_state *
489task_get_css(struct task_struct *task, int subsys_id)
490{
491 struct cgroup_subsys_state *css;
492
493 rcu_read_lock();
494 while (true) {
495 css = task_css(task, subsys_id);
496 if (likely(css_tryget_online(css)))
497 break;
498 cpu_relax();
499 }
500 rcu_read_unlock();
501 return css;
502}
503
504/**
Tejun Heoc326aa22015-05-13 16:24:16 -0400505 * task_css_is_root - test whether a task belongs to the root css
506 * @task: the target task
507 * @subsys_id: the target subsystem ID
508 *
509 * Test whether @task belongs to the root css on the specified subsystem.
510 * May be invoked in any context.
511 */
512static inline bool task_css_is_root(struct task_struct *task, int subsys_id)
513{
514 return task_css_check(task, subsys_id, true) ==
515 init_css_set.subsys[subsys_id];
516}
517
518static inline struct cgroup *task_cgroup(struct task_struct *task,
519 int subsys_id)
520{
521 return task_css(task, subsys_id)->cgroup;
522}
Tejun Heoa2dd4242014-03-19 10:23:55 -0400523
Tejun Heo3e489302017-08-11 05:49:01 -0700524static inline struct cgroup *task_dfl_cgroup(struct task_struct *task)
525{
526 return task_css_set(task)->dfl_cgrp;
527}
528
529static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
530{
531 struct cgroup_subsys_state *parent_css = cgrp->self.parent;
532
533 if (parent_css)
534 return container_of(parent_css, struct cgroup, self);
535 return NULL;
536}
537
Tejun Heob11cfb52015-11-20 15:55:52 -0500538/**
539 * cgroup_is_descendant - test ancestry
540 * @cgrp: the cgroup to be tested
541 * @ancestor: possible ancestor of @cgrp
542 *
543 * Test whether @cgrp is a descendant of @ancestor. It also returns %true
544 * if @cgrp == @ancestor. This function is safe to call as long as @cgrp
545 * and @ancestor are accessible.
546 */
547static inline bool cgroup_is_descendant(struct cgroup *cgrp,
548 struct cgroup *ancestor)
549{
550 if (cgrp->root != ancestor->root || cgrp->level < ancestor->level)
551 return false;
552 return cgrp->ancestor_ids[ancestor->level] == ancestor->id;
553}
554
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700555/**
556 * task_under_cgroup_hierarchy - test task's membership of cgroup ancestry
557 * @task: the task to be tested
558 * @ancestor: possible ancestor of @task's cgroup
559 *
560 * Tests whether @task's default cgroup hierarchy is a descendant of @ancestor.
561 * It follows all the same rules as cgroup_is_descendant, and only applies
562 * to the default hierarchy.
563 */
564static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
565 struct cgroup *ancestor)
566{
567 struct css_set *cset = task_css_set(task);
568
569 return cgroup_is_descendant(cset->dfl_cgrp, ancestor);
570}
571
Tejun Heo07bc3562014-02-13 06:58:39 -0500572/* no synchronization, the result can only be used as a hint */
Tejun Heo27bd4db2015-10-15 16:41:50 -0400573static inline bool cgroup_is_populated(struct cgroup *cgrp)
Tejun Heo07bc3562014-02-13 06:58:39 -0500574{
Tejun Heo454000a2017-05-15 09:34:02 -0400575 return cgrp->nr_populated_csets + cgrp->nr_populated_domain_children +
576 cgrp->nr_populated_threaded_children;
Tejun Heo07bc3562014-02-13 06:58:39 -0500577}
578
Zefan Lif29374b2014-09-19 16:29:31 +0800579/* returns ino associated with a cgroup */
Tejun Heob1664922014-02-11 11:52:49 -0500580static inline ino_t cgroup_ino(struct cgroup *cgrp)
581{
Shaohua Lic53cd492017-07-12 11:49:50 -0700582 return cgrp->kn->id.ino;
Tejun Heob1664922014-02-11 11:52:49 -0500583}
584
Tejun Heob4168642014-05-13 12:16:21 -0400585/* cft/css accessors for cftype->write() operation */
586static inline struct cftype *of_cft(struct kernfs_open_file *of)
Tejun Heo7da11272013-12-05 12:28:04 -0500587{
Tejun Heo2bd59d42014-02-11 11:52:49 -0500588 return of->kn->priv;
Tejun Heo7da11272013-12-05 12:28:04 -0500589}
590
Tejun Heob4168642014-05-13 12:16:21 -0400591struct cgroup_subsys_state *of_css(struct kernfs_open_file *of);
592
593/* cft/css accessors for cftype->seq_*() operations */
594static inline struct cftype *seq_cft(struct seq_file *seq)
595{
596 return of_cft(seq->private);
597}
598
599static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq)
600{
601 return of_css(seq->private);
602}
Tejun Heo59f52962014-02-11 11:52:49 -0500603
Tejun Heoe61734c2014-02-12 09:29:50 -0500604/*
605 * Name / path handling functions. All are thin wrappers around the kernfs
606 * counterparts and can be called under any context.
607 */
608
609static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen)
610{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400611 return kernfs_name(cgrp->kn, buf, buflen);
Tejun Heoe61734c2014-02-12 09:29:50 -0500612}
613
Tejun Heo4c737b42016-08-10 11:23:44 -0400614static inline int cgroup_path(struct cgroup *cgrp, char *buf, size_t buflen)
Tejun Heoe61734c2014-02-12 09:29:50 -0500615{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400616 return kernfs_path(cgrp->kn, buf, buflen);
Tejun Heoe61734c2014-02-12 09:29:50 -0500617}
618
619static inline void pr_cont_cgroup_name(struct cgroup *cgrp)
620{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400621 pr_cont_kernfs_name(cgrp->kn);
Tejun Heoe61734c2014-02-12 09:29:50 -0500622}
623
624static inline void pr_cont_cgroup_path(struct cgroup *cgrp)
625{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400626 pr_cont_kernfs_path(cgrp->kn);
Tejun Heoe61734c2014-02-12 09:29:50 -0500627}
628
Tejun Heo77f88792017-03-16 16:54:24 -0400629static inline void cgroup_init_kthreadd(void)
630{
631 /*
632 * kthreadd is inherited by all kthreads, keep it in the root so
633 * that the new kthreads are guaranteed to stay in the root until
634 * initialization is finished.
635 */
636 current->no_cgroup_migration = 1;
637}
638
639static inline void cgroup_kthread_ready(void)
640{
641 /*
642 * This kthread finished initialization. The creator should have
643 * set PF_NO_SETAFFINITY if this kthread should stay in the root.
644 */
645 current->no_cgroup_migration = 0;
646}
647
Shaohua Li121508d2017-07-12 11:49:52 -0700648static inline union kernfs_node_id *cgroup_get_kernfs_id(struct cgroup *cgrp)
649{
650 return &cgrp->kn->id;
651}
Shaohua Li69fd5c32017-07-12 11:49:55 -0700652
653void cgroup_path_from_kernfs_id(const union kernfs_node_id *id,
654 char *buf, size_t buflen);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700655#else /* !CONFIG_CGROUPS */
656
Tejun Heof3ba5382015-01-06 12:02:46 -0500657struct cgroup_subsys_state;
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700658struct cgroup;
Tejun Heof3ba5382015-01-06 12:02:46 -0500659
Tejun Heoc326aa22015-05-13 16:24:16 -0400660static inline void css_put(struct cgroup_subsys_state *css) {}
661static inline int cgroup_attach_task_all(struct task_struct *from,
662 struct task_struct *t) { return 0; }
663static inline int cgroupstats_build(struct cgroupstats *stats,
664 struct dentry *dentry) { return -EINVAL; }
665
Paul Menageb4f48b62007-10-18 23:39:33 -0700666static inline void cgroup_fork(struct task_struct *p) {}
Oleg Nesterovb53202e2015-12-03 10:24:08 -0500667static inline int cgroup_can_fork(struct task_struct *p) { return 0; }
668static inline void cgroup_cancel_fork(struct task_struct *p) {}
669static inline void cgroup_post_fork(struct task_struct *p) {}
Li Zefan1ec41832014-03-28 15:22:19 +0800670static inline void cgroup_exit(struct task_struct *p) {}
Tejun Heo2e91fa72015-10-15 16:41:53 -0400671static inline void cgroup_free(struct task_struct *p) {}
Paul Menageddbcc7e2007-10-18 23:39:30 -0700672
Tejun Heoc326aa22015-05-13 16:24:16 -0400673static inline int cgroup_init_early(void) { return 0; }
674static inline int cgroup_init(void) { return 0; }
Tejun Heo77f88792017-03-16 16:54:24 -0400675static inline void cgroup_init_kthreadd(void) {}
676static inline void cgroup_kthread_ready(void) {}
Shaohua Li121508d2017-07-12 11:49:52 -0700677static inline union kernfs_node_id *cgroup_get_kernfs_id(struct cgroup *cgrp)
678{
679 return NULL;
680}
Sridhar Samudralad7926ee2010-05-30 22:24:39 +0200681
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700682static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
683 struct cgroup *ancestor)
684{
685 return true;
686}
Shaohua Li69fd5c32017-07-12 11:49:55 -0700687
688static inline void cgroup_path_from_kernfs_id(const union kernfs_node_id *id,
689 char *buf, size_t buflen) {}
Paul Menageddbcc7e2007-10-18 23:39:30 -0700690#endif /* !CONFIG_CGROUPS */
691
Tejun Heobd1060a2015-12-07 17:38:53 -0500692/*
693 * sock->sk_cgrp_data handling. For more info, see sock_cgroup_data
694 * definition in cgroup-defs.h.
695 */
696#ifdef CONFIG_SOCK_CGROUP_DATA
697
698#if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
699extern spinlock_t cgroup_sk_update_lock;
700#endif
701
702void cgroup_sk_alloc_disable(void);
703void cgroup_sk_alloc(struct sock_cgroup_data *skcd);
704void cgroup_sk_free(struct sock_cgroup_data *skcd);
705
706static inline struct cgroup *sock_cgroup_ptr(struct sock_cgroup_data *skcd)
707{
708#if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
709 unsigned long v;
710
711 /*
712 * @skcd->val is 64bit but the following is safe on 32bit too as we
713 * just need the lower ulong to be written and read atomically.
714 */
715 v = READ_ONCE(skcd->val);
716
717 if (v & 1)
718 return &cgrp_dfl_root.cgrp;
719
720 return (struct cgroup *)(unsigned long)v ?: &cgrp_dfl_root.cgrp;
721#else
722 return (struct cgroup *)(unsigned long)skcd->val;
723#endif
724}
725
726#else /* CONFIG_CGROUP_DATA */
727
728static inline void cgroup_sk_alloc(struct sock_cgroup_data *skcd) {}
729static inline void cgroup_sk_free(struct sock_cgroup_data *skcd) {}
730
731#endif /* CONFIG_CGROUP_DATA */
732
Aditya Kalia79a9082016-01-29 02:54:06 -0600733struct cgroup_namespace {
Elena Reshetova387ad962017-02-20 12:19:00 +0200734 refcount_t count;
Aditya Kalia79a9082016-01-29 02:54:06 -0600735 struct ns_common ns;
736 struct user_namespace *user_ns;
Eric W. Biedermand08311d2016-08-08 14:25:30 -0500737 struct ucounts *ucounts;
Aditya Kalia79a9082016-01-29 02:54:06 -0600738 struct css_set *root_cset;
739};
740
741extern struct cgroup_namespace init_cgroup_ns;
742
743#ifdef CONFIG_CGROUPS
744
745void free_cgroup_ns(struct cgroup_namespace *ns);
746
747struct cgroup_namespace *copy_cgroup_ns(unsigned long flags,
748 struct user_namespace *user_ns,
749 struct cgroup_namespace *old_ns);
750
Tejun Heo4c737b42016-08-10 11:23:44 -0400751int cgroup_path_ns(struct cgroup *cgrp, char *buf, size_t buflen,
752 struct cgroup_namespace *ns);
Aditya Kalia79a9082016-01-29 02:54:06 -0600753
754#else /* !CONFIG_CGROUPS */
755
756static inline void free_cgroup_ns(struct cgroup_namespace *ns) { }
757static inline struct cgroup_namespace *
758copy_cgroup_ns(unsigned long flags, struct user_namespace *user_ns,
759 struct cgroup_namespace *old_ns)
760{
761 return old_ns;
762}
763
764#endif /* !CONFIG_CGROUPS */
765
766static inline void get_cgroup_ns(struct cgroup_namespace *ns)
767{
768 if (ns)
Elena Reshetova387ad962017-02-20 12:19:00 +0200769 refcount_inc(&ns->count);
Aditya Kalia79a9082016-01-29 02:54:06 -0600770}
771
772static inline void put_cgroup_ns(struct cgroup_namespace *ns)
773{
Elena Reshetova387ad962017-02-20 12:19:00 +0200774 if (ns && refcount_dec_and_test(&ns->count))
Aditya Kalia79a9082016-01-29 02:54:06 -0600775 free_cgroup_ns(ns);
776}
777
Paul Menageddbcc7e2007-10-18 23:39:30 -0700778#endif /* _LINUX_CGROUP_H */