blob: 02da4e1def61ed7ca9f9d6a2b6ae181a5220e461 [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 Heod2cc5ed2017-09-25 08:12:04 -070026#include <linux/kernel_stat.h>
Tejun Heob4a04ab2015-05-13 15:38:40 -040027
28#include <linux/cgroup-defs.h>
Paul Menageddbcc7e2007-10-18 23:39:30 -070029
30#ifdef CONFIG_CGROUPS
31
Tejun Heo6abc8ca2015-08-04 15:20:55 -040032/*
33 * All weight knobs on the default hierarhcy should use the following min,
34 * default and max values. The default value is the logarithmic center of
35 * MIN and MAX and allows 100x to be expressed in both directions.
36 */
37#define CGROUP_WEIGHT_MIN 1
38#define CGROUP_WEIGHT_DFL 100
39#define CGROUP_WEIGHT_MAX 10000
40
Tejun Heobc2fb7e2017-05-15 09:34:01 -040041/* walk only threadgroup leaders */
42#define CSS_TASK_ITER_PROCS (1U << 0)
Tejun Heo450ee0c2017-05-15 09:34:03 -040043/* walk all threaded css_sets in the domain */
44#define CSS_TASK_ITER_THREADED (1U << 1)
Tejun Heobc2fb7e2017-05-15 09:34:01 -040045
Tejun Heo370b9e62019-05-31 10:38:58 -070046/* internal flags */
47#define CSS_TASK_ITER_SKIPPED (1U << 16)
48
Tejun Heoc326aa22015-05-13 16:24:16 -040049/* a css_task_iter should be treated as an opaque object */
50struct css_task_iter {
51 struct cgroup_subsys *ss;
Tejun Heobc2fb7e2017-05-15 09:34:01 -040052 unsigned int flags;
Paul Menageddbcc7e2007-10-18 23:39:30 -070053
Tejun Heoc326aa22015-05-13 16:24:16 -040054 struct list_head *cset_pos;
55 struct list_head *cset_head;
Paul Menageddbcc7e2007-10-18 23:39:30 -070056
Tejun Heo450ee0c2017-05-15 09:34:03 -040057 struct list_head *tcset_pos;
58 struct list_head *tcset_head;
59
Tejun Heoc326aa22015-05-13 16:24:16 -040060 struct list_head *task_pos;
61 struct list_head *tasks_head;
62 struct list_head *mg_tasks_head;
Tejun Heo4340d172019-05-31 10:38:58 -070063 struct list_head *dying_tasks_head;
Tejun Heoed27b9f2015-10-15 16:41:52 -040064
Michal Koutnýab3e3b22020-01-24 12:40:15 +010065 struct list_head *cur_tasks_head;
Tejun Heoed27b9f2015-10-15 16:41:52 -040066 struct css_set *cur_cset;
Tejun Heo450ee0c2017-05-15 09:34:03 -040067 struct css_set *cur_dcset;
Tejun Heoed27b9f2015-10-15 16:41:52 -040068 struct task_struct *cur_task;
69 struct list_head iters_node; /* css_set->task_iters */
Paul Menage817929e2007-10-18 23:39:36 -070070};
Tejun Heoc326aa22015-05-13 16:24:16 -040071
72extern struct cgroup_root cgrp_dfl_root;
73extern struct css_set init_css_set;
74
75#define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys;
76#include <linux/cgroup_subsys.h>
Paul Menage817929e2007-10-18 23:39:36 -070077#undef SUBSYS
78
Tejun Heo49d1dc42015-09-18 11:56:28 -040079#define SUBSYS(_x) \
80 extern struct static_key_true _x ## _cgrp_subsys_enabled_key; \
81 extern struct static_key_true _x ## _cgrp_subsys_on_dfl_key;
82#include <linux/cgroup_subsys.h>
83#undef SUBSYS
84
85/**
86 * cgroup_subsys_enabled - fast test on whether a subsys is enabled
87 * @ss: subsystem in question
88 */
89#define cgroup_subsys_enabled(ss) \
90 static_branch_likely(&ss ## _enabled_key)
91
92/**
93 * cgroup_subsys_on_dfl - fast test on whether a subsys is on default hierarchy
94 * @ss: subsystem in question
95 */
96#define cgroup_subsys_on_dfl(ss) \
97 static_branch_likely(&ss ## _on_dfl_key)
98
Tejun Heoc326aa22015-05-13 16:24:16 -040099bool css_has_online_children(struct cgroup_subsys_state *css);
100struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss);
101struct cgroup_subsys_state *cgroup_get_e_css(struct cgroup *cgroup,
102 struct cgroup_subsys *ss);
103struct cgroup_subsys_state *css_tryget_online_from_dir(struct dentry *dentry,
104 struct cgroup_subsys *ss);
105
Tejun Heo16af4392015-11-20 15:55:52 -0500106struct cgroup *cgroup_get_from_path(const char *path);
Martin KaFai Lau1f3fe7e2016-06-30 10:28:42 -0700107struct cgroup *cgroup_get_from_fd(int fd);
Tejun Heo16af4392015-11-20 15:55:52 -0500108
Tejun Heoc326aa22015-05-13 16:24:16 -0400109int cgroup_attach_task_all(struct task_struct *from, struct task_struct *);
110int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from);
111
112int cgroup_add_dfl_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
113int cgroup_add_legacy_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
114int cgroup_rm_cftypes(struct cftype *cfts);
Tejun Heo34c06252015-11-05 00:12:24 -0500115void cgroup_file_notify(struct cgroup_file *cfile);
Tejun Heoc326aa22015-05-13 16:24:16 -0400116
Tejun Heo4c737b42016-08-10 11:23:44 -0400117int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen);
Tejun Heoc326aa22015-05-13 16:24:16 -0400118int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry);
119int proc_cgroup_show(struct seq_file *m, struct pid_namespace *ns,
120 struct pid *pid, struct task_struct *tsk);
121
122void cgroup_fork(struct task_struct *p);
Oleg Nesterovb53202e2015-12-03 10:24:08 -0500123extern int cgroup_can_fork(struct task_struct *p);
124extern void cgroup_cancel_fork(struct task_struct *p);
125extern void cgroup_post_fork(struct task_struct *p);
Tejun Heoc326aa22015-05-13 16:24:16 -0400126void cgroup_exit(struct task_struct *p);
Oleg Nesterovd0bc74c2019-01-28 17:00:13 +0100127void cgroup_release(struct task_struct *p);
Tejun Heo2e91fa72015-10-15 16:41:53 -0400128void cgroup_free(struct task_struct *p);
Tejun Heoc326aa22015-05-13 16:24:16 -0400129
130int cgroup_init_early(void);
131int cgroup_init(void);
132
Tejun Heo5c9d5352014-05-16 13:22:48 -0400133/*
Tejun Heoc326aa22015-05-13 16:24:16 -0400134 * Iteration helpers and macros.
Tejun Heo5c9d5352014-05-16 13:22:48 -0400135 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700136
Tejun Heoc326aa22015-05-13 16:24:16 -0400137struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos,
138 struct cgroup_subsys_state *parent);
139struct cgroup_subsys_state *css_next_descendant_pre(struct cgroup_subsys_state *pos,
140 struct cgroup_subsys_state *css);
141struct cgroup_subsys_state *css_rightmost_descendant(struct cgroup_subsys_state *pos);
142struct cgroup_subsys_state *css_next_descendant_post(struct cgroup_subsys_state *pos,
143 struct cgroup_subsys_state *css);
Tejun Heo72c97e52013-08-08 20:11:22 -0400144
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500145struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset,
146 struct cgroup_subsys_state **dst_cssp);
147struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset,
148 struct cgroup_subsys_state **dst_cssp);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700149
Tejun Heobc2fb7e2017-05-15 09:34:01 -0400150void css_task_iter_start(struct cgroup_subsys_state *css, unsigned int flags,
Tejun Heoc326aa22015-05-13 16:24:16 -0400151 struct css_task_iter *it);
152struct task_struct *css_task_iter_next(struct css_task_iter *it);
153void css_task_iter_end(struct css_task_iter *it);
Tejun Heo0ae78e02013-08-13 11:01:54 -0400154
Tejun Heoc326aa22015-05-13 16:24:16 -0400155/**
156 * css_for_each_child - iterate through children of a css
157 * @pos: the css * to use as the loop cursor
158 * @parent: css whose children to walk
159 *
160 * Walk @parent's children. Must be called under rcu_read_lock().
161 *
162 * If a subsystem synchronizes ->css_online() and the start of iteration, a
163 * css which finished ->css_online() is guaranteed to be visible in the
164 * future iterations and will stay visible until the last reference is put.
165 * A css which hasn't finished ->css_online() or already finished
166 * ->css_offline() may show up during traversal. It's each subsystem's
167 * responsibility to synchronize against on/offlining.
168 *
169 * It is allowed to temporarily drop RCU read lock during iteration. The
170 * caller is responsible for ensuring that @pos remains accessible until
171 * the start of the next iteration by, for example, bumping the css refcnt.
172 */
173#define css_for_each_child(pos, parent) \
174 for ((pos) = css_next_child(NULL, (parent)); (pos); \
175 (pos) = css_next_child((pos), (parent)))
Tejun Heod5c419b2014-05-16 13:22:48 -0400176
Tejun Heoc326aa22015-05-13 16:24:16 -0400177/**
178 * css_for_each_descendant_pre - pre-order walk of a css's descendants
179 * @pos: the css * to use as the loop cursor
180 * @root: css whose descendants to walk
181 *
182 * Walk @root's descendants. @root is included in the iteration and the
183 * first node to be visited. Must be called under rcu_read_lock().
184 *
185 * If a subsystem synchronizes ->css_online() and the start of iteration, a
186 * css which finished ->css_online() is guaranteed to be visible in the
187 * future iterations and will stay visible until the last reference is put.
188 * A css which hasn't finished ->css_online() or already finished
189 * ->css_offline() may show up during traversal. It's each subsystem's
190 * responsibility to synchronize against on/offlining.
191 *
192 * For example, the following guarantees that a descendant can't escape
193 * state updates of its ancestors.
194 *
195 * my_online(@css)
196 * {
197 * Lock @css's parent and @css;
198 * Inherit state from the parent;
199 * Unlock both.
200 * }
201 *
202 * my_update_state(@css)
203 * {
204 * css_for_each_descendant_pre(@pos, @css) {
205 * Lock @pos;
206 * if (@pos == @css)
207 * Update @css's state;
208 * else
209 * Verify @pos is alive and inherit state from its parent;
210 * Unlock @pos;
211 * }
212 * }
213 *
214 * As long as the inheriting step, including checking the parent state, is
215 * enclosed inside @pos locking, double-locking the parent isn't necessary
216 * while inheriting. The state update to the parent is guaranteed to be
217 * visible by walking order and, as long as inheriting operations to the
218 * same @pos are atomic to each other, multiple updates racing each other
219 * still result in the correct state. It's guaranateed that at least one
220 * inheritance happens for any css after the latest update to its parent.
221 *
222 * If checking parent's state requires locking the parent, each inheriting
223 * iteration should lock and unlock both @pos->parent and @pos.
224 *
225 * Alternatively, a subsystem may choose to use a single global lock to
226 * synchronize ->css_online() and ->css_offline() against tree-walking
227 * operations.
228 *
229 * It is allowed to temporarily drop RCU read lock during iteration. The
230 * caller is responsible for ensuring that @pos remains accessible until
231 * the start of the next iteration by, for example, bumping the css refcnt.
232 */
233#define css_for_each_descendant_pre(pos, css) \
234 for ((pos) = css_next_descendant_pre(NULL, (css)); (pos); \
235 (pos) = css_next_descendant_pre((pos), (css)))
Tejun Heo15a4c832014-05-04 15:09:14 -0400236
Tejun Heoc326aa22015-05-13 16:24:16 -0400237/**
238 * css_for_each_descendant_post - post-order walk of a css's descendants
239 * @pos: the css * to use as the loop cursor
240 * @css: css whose descendants to walk
241 *
242 * Similar to css_for_each_descendant_pre() but performs post-order
243 * traversal instead. @root is included in the iteration and the last
244 * node to be visited.
245 *
246 * If a subsystem synchronizes ->css_online() and the start of iteration, a
247 * css which finished ->css_online() is guaranteed to be visible in the
248 * future iterations and will stay visible until the last reference is put.
249 * A css which hasn't finished ->css_online() or already finished
250 * ->css_offline() may show up during traversal. It's each subsystem's
251 * responsibility to synchronize against on/offlining.
252 *
253 * Note that the walk visibility guarantee example described in pre-order
254 * walk doesn't apply the same to post-order walks.
255 */
256#define css_for_each_descendant_post(pos, css) \
257 for ((pos) = css_next_descendant_post(NULL, (css)); (pos); \
258 (pos) = css_next_descendant_post((pos), (css)))
Tejun Heo48ddbe12012-04-01 12:09:56 -0700259
Tejun Heoc326aa22015-05-13 16:24:16 -0400260/**
261 * cgroup_taskset_for_each - iterate cgroup_taskset
262 * @task: the loop cursor
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500263 * @dst_css: the destination css
Tejun Heoc326aa22015-05-13 16:24:16 -0400264 * @tset: taskset to iterate
Tejun Heo4530edd2015-09-11 15:00:19 -0400265 *
266 * @tset may contain multiple tasks and they may belong to multiple
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500267 * processes.
268 *
269 * On the v2 hierarchy, there may be tasks from multiple processes and they
270 * may not share the source or destination csses.
271 *
272 * On traditional hierarchies, when there are multiple tasks in @tset, if a
273 * task of a process is in @tset, all tasks of the process are in @tset.
274 * Also, all are guaranteed to share the same source and destination csses.
Tejun Heo4530edd2015-09-11 15:00:19 -0400275 *
276 * Iteration is not in any specific order.
Tejun Heoc326aa22015-05-13 16:24:16 -0400277 */
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500278#define cgroup_taskset_for_each(task, dst_css, tset) \
279 for ((task) = cgroup_taskset_first((tset), &(dst_css)); \
280 (task); \
281 (task) = cgroup_taskset_next((tset), &(dst_css)))
Tejun Heo0cb51d72014-05-16 13:22:49 -0400282
Tejun Heo4530edd2015-09-11 15:00:19 -0400283/**
284 * cgroup_taskset_for_each_leader - iterate group leaders in a cgroup_taskset
285 * @leader: the loop cursor
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500286 * @dst_css: the destination css
Geliang Tang7b4632f2016-12-24 23:28:35 +0800287 * @tset: taskset to iterate
Tejun Heo4530edd2015-09-11 15:00:19 -0400288 *
289 * Iterate threadgroup leaders of @tset. For single-task migrations, @tset
290 * may not contain any.
291 */
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500292#define cgroup_taskset_for_each_leader(leader, dst_css, tset) \
293 for ((leader) = cgroup_taskset_first((tset), &(dst_css)); \
294 (leader); \
295 (leader) = cgroup_taskset_next((tset), &(dst_css))) \
Tejun Heo4530edd2015-09-11 15:00:19 -0400296 if ((leader) != (leader)->group_leader) \
297 ; \
298 else
299
Tejun Heoc326aa22015-05-13 16:24:16 -0400300/*
301 * Inline functions.
302 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700303
Tejun Heo5de01072013-06-12 21:04:52 -0700304/**
305 * css_get - obtain a reference on the specified css
306 * @css: target css
307 *
308 * The caller must already have a reference.
Paul Menageddbcc7e2007-10-18 23:39:30 -0700309 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700310static inline void css_get(struct cgroup_subsys_state *css)
311{
Tejun Heo3b514d22014-05-16 13:22:47 -0400312 if (!(css->flags & CSS_NO_REF))
313 percpu_ref_get(&css->refcnt);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700314}
Paul Menagee7c5ec92009-01-07 18:08:38 -0800315
Tejun Heo5de01072013-06-12 21:04:52 -0700316/**
Johannes Weinere8ea14c2014-12-10 15:42:42 -0800317 * css_get_many - obtain references on the specified css
318 * @css: target css
319 * @n: number of references to get
320 *
321 * The caller must already have a reference.
322 */
323static inline void css_get_many(struct cgroup_subsys_state *css, unsigned int n)
324{
325 if (!(css->flags & CSS_NO_REF))
326 percpu_ref_get_many(&css->refcnt, n);
327}
328
329/**
Tejun Heo6f4524d2014-05-16 13:22:52 -0400330 * css_tryget - try to obtain a reference on the specified css
331 * @css: target css
332 *
333 * Obtain a reference on @css unless it already has reached zero and is
334 * being released. This function doesn't care whether @css is on or
335 * offline. The caller naturally needs to ensure that @css is accessible
336 * but doesn't have to be holding a reference on it - IOW, RCU protected
337 * access is good enough for this function. Returns %true if a reference
338 * count was successfully obtained; %false otherwise.
339 */
340static inline bool css_tryget(struct cgroup_subsys_state *css)
341{
342 if (!(css->flags & CSS_NO_REF))
343 return percpu_ref_tryget(&css->refcnt);
344 return true;
345}
346
347/**
Tejun Heoec903c02014-05-13 12:11:01 -0400348 * css_tryget_online - try to obtain a reference on the specified css if online
Tejun Heo5de01072013-06-12 21:04:52 -0700349 * @css: target css
350 *
Tejun Heoec903c02014-05-13 12:11:01 -0400351 * Obtain a reference on @css if it's online. The caller naturally needs
352 * to ensure that @css is accessible but doesn't have to be holding a
Tejun Heo5de01072013-06-12 21:04:52 -0700353 * reference on it - IOW, RCU protected access is good enough for this
354 * function. Returns %true if a reference count was successfully obtained;
355 * %false otherwise.
356 */
Tejun Heoec903c02014-05-13 12:11:01 -0400357static inline bool css_tryget_online(struct cgroup_subsys_state *css)
Paul Menagee7c5ec92009-01-07 18:08:38 -0800358{
Tejun Heo3b514d22014-05-16 13:22:47 -0400359 if (!(css->flags & CSS_NO_REF))
360 return percpu_ref_tryget_live(&css->refcnt);
361 return true;
Paul Menagee7c5ec92009-01-07 18:08:38 -0800362}
363
Tejun Heo5de01072013-06-12 21:04:52 -0700364/**
Tejun Heo41c25702017-05-24 12:03:48 -0400365 * css_is_dying - test whether the specified css is dying
366 * @css: target css
367 *
368 * Test whether @css is in the process of offlining or already offline. In
369 * most cases, ->css_online() and ->css_offline() callbacks should be
370 * enough; however, the actual offline operations are RCU delayed and this
371 * test returns %true also when @css is scheduled to be offlined.
372 *
373 * This is useful, for example, when the use case requires synchronous
374 * behavior with respect to cgroup removal. cgroup removal schedules css
375 * offlining but the css can seem alive while the operation is being
376 * delayed. If the delay affects user visible semantics, this test can be
377 * used to resolve the situation.
378 */
379static inline bool css_is_dying(struct cgroup_subsys_state *css)
380{
381 return !(css->flags & CSS_NO_REF) && percpu_ref_is_dying(&css->refcnt);
382}
383
384/**
Tejun Heo5de01072013-06-12 21:04:52 -0700385 * css_put - put a css reference
386 * @css: target css
387 *
Tejun Heoec903c02014-05-13 12:11:01 -0400388 * Put a reference obtained via css_get() and css_tryget_online().
Tejun Heo5de01072013-06-12 21:04:52 -0700389 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700390static inline void css_put(struct cgroup_subsys_state *css)
391{
Tejun Heo3b514d22014-05-16 13:22:47 -0400392 if (!(css->flags & CSS_NO_REF))
393 percpu_ref_put(&css->refcnt);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700394}
395
Johannes Weinere8ea14c2014-12-10 15:42:42 -0800396/**
397 * css_put_many - put css references
398 * @css: target css
399 * @n: number of references to put
400 *
401 * Put references obtained via css_get() and css_tryget_online().
402 */
403static inline void css_put_many(struct cgroup_subsys_state *css, unsigned int n)
404{
405 if (!(css->flags & CSS_NO_REF))
406 percpu_ref_put_many(&css->refcnt, n);
407}
408
Tejun Heo3e489302017-08-11 05:49:01 -0700409static inline void cgroup_get(struct cgroup *cgrp)
410{
411 css_get(&cgrp->self);
412}
413
414static inline bool cgroup_tryget(struct cgroup *cgrp)
415{
416 return css_tryget(&cgrp->self);
417}
418
Tejun Heo16af4392015-11-20 15:55:52 -0500419static inline void cgroup_put(struct cgroup *cgrp)
420{
421 css_put(&cgrp->self);
422}
423
Tejun Heoc326aa22015-05-13 16:24:16 -0400424/**
425 * task_css_set_check - obtain a task's css_set with extra access conditions
426 * @task: the task to obtain css_set for
427 * @__c: extra condition expression to be passed to rcu_dereference_check()
Paul Menageddbcc7e2007-10-18 23:39:30 -0700428 *
Tejun Heoc326aa22015-05-13 16:24:16 -0400429 * A task's css_set is RCU protected, initialized and exited while holding
430 * task_lock(), and can only be modified while holding both cgroup_mutex
431 * and task_lock() while the task is alive. This macro verifies that the
432 * caller is inside proper critical section and returns @task's css_set.
433 *
434 * The caller can also specify additional allowed conditions via @__c, such
435 * as locks used during the cgroup_subsys::attach() methods.
Paul Menageddbcc7e2007-10-18 23:39:30 -0700436 */
Tejun Heoc326aa22015-05-13 16:24:16 -0400437#ifdef CONFIG_PROVE_RCU
438extern struct mutex cgroup_mutex;
Tejun Heof0d9a5f2015-10-15 16:41:53 -0400439extern spinlock_t css_set_lock;
Tejun Heoc326aa22015-05-13 16:24:16 -0400440#define task_css_set_check(task, __c) \
441 rcu_dereference_check((task)->cgroups, \
442 lockdep_is_held(&cgroup_mutex) || \
Tejun Heof0d9a5f2015-10-15 16:41:53 -0400443 lockdep_is_held(&css_set_lock) || \
Tejun Heoc326aa22015-05-13 16:24:16 -0400444 ((task)->flags & PF_EXITING) || (__c))
445#else
446#define task_css_set_check(task, __c) \
447 rcu_dereference((task)->cgroups)
Tejun Heo2bd59d42014-02-11 11:52:49 -0500448#endif
Paul Menageddbcc7e2007-10-18 23:39:30 -0700449
Tejun Heoc326aa22015-05-13 16:24:16 -0400450/**
451 * task_css_check - obtain css for (task, subsys) w/ extra access conds
452 * @task: the target task
453 * @subsys_id: the target subsystem ID
454 * @__c: extra condition expression to be passed to rcu_dereference_check()
455 *
456 * Return the cgroup_subsys_state for the (@task, @subsys_id) pair. The
457 * synchronization rules are the same as task_css_set_check().
458 */
459#define task_css_check(task, subsys_id, __c) \
460 task_css_set_check((task), (__c))->subsys[(subsys_id)]
461
462/**
463 * task_css_set - obtain a task's css_set
464 * @task: the task to obtain css_set for
465 *
466 * See task_css_set_check().
467 */
468static inline struct css_set *task_css_set(struct task_struct *task)
469{
470 return task_css_set_check(task, false);
471}
472
473/**
474 * task_css - obtain css for (task, subsys)
475 * @task: the target task
476 * @subsys_id: the target subsystem ID
477 *
478 * See task_css_check().
479 */
480static inline struct cgroup_subsys_state *task_css(struct task_struct *task,
481 int subsys_id)
482{
483 return task_css_check(task, subsys_id, false);
484}
485
486/**
Linus Torvaldsbbe179f2015-06-26 19:50:04 -0700487 * task_get_css - find and get the css for (task, subsys)
488 * @task: the target task
489 * @subsys_id: the target subsystem ID
490 *
491 * Find the css for the (@task, @subsys_id) combination, increment a
492 * reference on and return it. This function is guaranteed to return a
Tejun Heoc3b85bd2019-05-29 13:46:25 -0700493 * valid css. The returned css may already have been offlined.
Linus Torvaldsbbe179f2015-06-26 19:50:04 -0700494 */
495static inline struct cgroup_subsys_state *
496task_get_css(struct task_struct *task, int subsys_id)
497{
498 struct cgroup_subsys_state *css;
499
500 rcu_read_lock();
501 while (true) {
502 css = task_css(task, subsys_id);
Tejun Heoc3b85bd2019-05-29 13:46:25 -0700503 /*
504 * Can't use css_tryget_online() here. A task which has
505 * PF_EXITING set may stay associated with an offline css.
506 * If such task calls this function, css_tryget_online()
507 * will keep failing.
508 */
509 if (likely(css_tryget(css)))
Linus Torvaldsbbe179f2015-06-26 19:50:04 -0700510 break;
511 cpu_relax();
512 }
513 rcu_read_unlock();
514 return css;
515}
516
517/**
Tejun Heoc326aa22015-05-13 16:24:16 -0400518 * task_css_is_root - test whether a task belongs to the root css
519 * @task: the target task
520 * @subsys_id: the target subsystem ID
521 *
522 * Test whether @task belongs to the root css on the specified subsystem.
523 * May be invoked in any context.
524 */
525static inline bool task_css_is_root(struct task_struct *task, int subsys_id)
526{
527 return task_css_check(task, subsys_id, true) ==
528 init_css_set.subsys[subsys_id];
529}
530
531static inline struct cgroup *task_cgroup(struct task_struct *task,
532 int subsys_id)
533{
534 return task_css(task, subsys_id)->cgroup;
535}
Tejun Heoa2dd4242014-03-19 10:23:55 -0400536
Tejun Heo3e489302017-08-11 05:49:01 -0700537static inline struct cgroup *task_dfl_cgroup(struct task_struct *task)
538{
539 return task_css_set(task)->dfl_cgrp;
540}
541
542static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
543{
544 struct cgroup_subsys_state *parent_css = cgrp->self.parent;
545
546 if (parent_css)
547 return container_of(parent_css, struct cgroup, self);
548 return NULL;
549}
550
Tejun Heob11cfb52015-11-20 15:55:52 -0500551/**
552 * cgroup_is_descendant - test ancestry
553 * @cgrp: the cgroup to be tested
554 * @ancestor: possible ancestor of @cgrp
555 *
556 * Test whether @cgrp is a descendant of @ancestor. It also returns %true
557 * if @cgrp == @ancestor. This function is safe to call as long as @cgrp
558 * and @ancestor are accessible.
559 */
560static inline bool cgroup_is_descendant(struct cgroup *cgrp,
561 struct cgroup *ancestor)
562{
563 if (cgrp->root != ancestor->root || cgrp->level < ancestor->level)
564 return false;
565 return cgrp->ancestor_ids[ancestor->level] == ancestor->id;
566}
567
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700568/**
Andrey Ignatov77236282018-08-12 10:49:27 -0700569 * cgroup_ancestor - find ancestor of cgroup
570 * @cgrp: cgroup to find ancestor of
571 * @ancestor_level: level of ancestor to find starting from root
572 *
573 * Find ancestor of cgroup at specified level starting from root if it exists
574 * and return pointer to it. Return NULL if @cgrp doesn't have ancestor at
575 * @ancestor_level.
576 *
577 * This function is safe to call as long as @cgrp is accessible.
578 */
579static inline struct cgroup *cgroup_ancestor(struct cgroup *cgrp,
580 int ancestor_level)
581{
582 struct cgroup *ptr;
583
584 if (cgrp->level < ancestor_level)
585 return NULL;
586
587 for (ptr = cgrp;
588 ptr && ptr->level > ancestor_level;
589 ptr = cgroup_parent(ptr))
590 ;
591
592 if (ptr && ptr->level == ancestor_level)
593 return ptr;
594
595 return NULL;
596}
597
598/**
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700599 * task_under_cgroup_hierarchy - test task's membership of cgroup ancestry
600 * @task: the task to be tested
601 * @ancestor: possible ancestor of @task's cgroup
602 *
603 * Tests whether @task's default cgroup hierarchy is a descendant of @ancestor.
604 * It follows all the same rules as cgroup_is_descendant, and only applies
605 * to the default hierarchy.
606 */
607static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
608 struct cgroup *ancestor)
609{
610 struct css_set *cset = task_css_set(task);
611
612 return cgroup_is_descendant(cset->dfl_cgrp, ancestor);
613}
614
Tejun Heo07bc3562014-02-13 06:58:39 -0500615/* no synchronization, the result can only be used as a hint */
Tejun Heo27bd4db2015-10-15 16:41:50 -0400616static inline bool cgroup_is_populated(struct cgroup *cgrp)
Tejun Heo07bc3562014-02-13 06:58:39 -0500617{
Tejun Heo454000a2017-05-15 09:34:02 -0400618 return cgrp->nr_populated_csets + cgrp->nr_populated_domain_children +
619 cgrp->nr_populated_threaded_children;
Tejun Heo07bc3562014-02-13 06:58:39 -0500620}
621
Zefan Lif29374b2014-09-19 16:29:31 +0800622/* returns ino associated with a cgroup */
Tejun Heob1664922014-02-11 11:52:49 -0500623static inline ino_t cgroup_ino(struct cgroup *cgrp)
624{
Shaohua Lic53cd492017-07-12 11:49:50 -0700625 return cgrp->kn->id.ino;
Tejun Heob1664922014-02-11 11:52:49 -0500626}
627
Tejun Heob4168642014-05-13 12:16:21 -0400628/* cft/css accessors for cftype->write() operation */
629static inline struct cftype *of_cft(struct kernfs_open_file *of)
Tejun Heo7da11272013-12-05 12:28:04 -0500630{
Tejun Heo2bd59d42014-02-11 11:52:49 -0500631 return of->kn->priv;
Tejun Heo7da11272013-12-05 12:28:04 -0500632}
633
Tejun Heob4168642014-05-13 12:16:21 -0400634struct cgroup_subsys_state *of_css(struct kernfs_open_file *of);
635
636/* cft/css accessors for cftype->seq_*() operations */
637static inline struct cftype *seq_cft(struct seq_file *seq)
638{
639 return of_cft(seq->private);
640}
641
642static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq)
643{
644 return of_css(seq->private);
645}
Tejun Heo59f52962014-02-11 11:52:49 -0500646
Tejun Heoe61734c2014-02-12 09:29:50 -0500647/*
648 * Name / path handling functions. All are thin wrappers around the kernfs
649 * counterparts and can be called under any context.
650 */
651
652static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen)
653{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400654 return kernfs_name(cgrp->kn, buf, buflen);
Tejun Heoe61734c2014-02-12 09:29:50 -0500655}
656
Tejun Heo4c737b42016-08-10 11:23:44 -0400657static inline int cgroup_path(struct cgroup *cgrp, char *buf, size_t buflen)
Tejun Heoe61734c2014-02-12 09:29:50 -0500658{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400659 return kernfs_path(cgrp->kn, buf, buflen);
Tejun Heoe61734c2014-02-12 09:29:50 -0500660}
661
662static inline void pr_cont_cgroup_name(struct cgroup *cgrp)
663{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400664 pr_cont_kernfs_name(cgrp->kn);
Tejun Heoe61734c2014-02-12 09:29:50 -0500665}
666
667static inline void pr_cont_cgroup_path(struct cgroup *cgrp)
668{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400669 pr_cont_kernfs_path(cgrp->kn);
Tejun Heoe61734c2014-02-12 09:29:50 -0500670}
671
Tejun Heo77f88792017-03-16 16:54:24 -0400672static inline void cgroup_init_kthreadd(void)
673{
674 /*
675 * kthreadd is inherited by all kthreads, keep it in the root so
676 * that the new kthreads are guaranteed to stay in the root until
677 * initialization is finished.
678 */
679 current->no_cgroup_migration = 1;
680}
681
682static inline void cgroup_kthread_ready(void)
683{
684 /*
685 * This kthread finished initialization. The creator should have
686 * set PF_NO_SETAFFINITY if this kthread should stay in the root.
687 */
688 current->no_cgroup_migration = 0;
689}
690
Shaohua Li121508d2017-07-12 11:49:52 -0700691static inline union kernfs_node_id *cgroup_get_kernfs_id(struct cgroup *cgrp)
692{
693 return &cgrp->kn->id;
694}
Shaohua Li69fd5c32017-07-12 11:49:55 -0700695
696void cgroup_path_from_kernfs_id(const union kernfs_node_id *id,
697 char *buf, size_t buflen);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700698#else /* !CONFIG_CGROUPS */
699
Tejun Heof3ba5382015-01-06 12:02:46 -0500700struct cgroup_subsys_state;
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700701struct cgroup;
Tejun Heof3ba5382015-01-06 12:02:46 -0500702
Tejun Heoc326aa22015-05-13 16:24:16 -0400703static inline void css_put(struct cgroup_subsys_state *css) {}
704static inline int cgroup_attach_task_all(struct task_struct *from,
705 struct task_struct *t) { return 0; }
706static inline int cgroupstats_build(struct cgroupstats *stats,
707 struct dentry *dentry) { return -EINVAL; }
708
Paul Menageb4f48b62007-10-18 23:39:33 -0700709static inline void cgroup_fork(struct task_struct *p) {}
Oleg Nesterovb53202e2015-12-03 10:24:08 -0500710static inline int cgroup_can_fork(struct task_struct *p) { return 0; }
711static inline void cgroup_cancel_fork(struct task_struct *p) {}
712static inline void cgroup_post_fork(struct task_struct *p) {}
Li Zefan1ec41832014-03-28 15:22:19 +0800713static inline void cgroup_exit(struct task_struct *p) {}
Oleg Nesterovd0bc74c2019-01-28 17:00:13 +0100714static inline void cgroup_release(struct task_struct *p) {}
Tejun Heo2e91fa72015-10-15 16:41:53 -0400715static inline void cgroup_free(struct task_struct *p) {}
Paul Menageddbcc7e2007-10-18 23:39:30 -0700716
Tejun Heoc326aa22015-05-13 16:24:16 -0400717static inline int cgroup_init_early(void) { return 0; }
718static inline int cgroup_init(void) { return 0; }
Tejun Heo77f88792017-03-16 16:54:24 -0400719static inline void cgroup_init_kthreadd(void) {}
720static inline void cgroup_kthread_ready(void) {}
Shaohua Li121508d2017-07-12 11:49:52 -0700721static inline union kernfs_node_id *cgroup_get_kernfs_id(struct cgroup *cgrp)
722{
723 return NULL;
724}
Sridhar Samudralad7926ee2010-05-30 22:24:39 +0200725
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700726static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
727 struct cgroup *ancestor)
728{
729 return true;
730}
Shaohua Li69fd5c32017-07-12 11:49:55 -0700731
732static inline void cgroup_path_from_kernfs_id(const union kernfs_node_id *id,
733 char *buf, size_t buflen) {}
Paul Menageddbcc7e2007-10-18 23:39:30 -0700734#endif /* !CONFIG_CGROUPS */
735
Tejun Heo6162cef2018-04-26 14:29:05 -0700736#ifdef CONFIG_CGROUPS
737/*
738 * cgroup scalable recursive statistics.
739 */
740void cgroup_rstat_updated(struct cgroup *cgrp, int cpu);
741void cgroup_rstat_flush(struct cgroup *cgrp);
Tejun Heo0fa294f2018-04-26 14:29:05 -0700742void cgroup_rstat_flush_irqsafe(struct cgroup *cgrp);
Tejun Heo6162cef2018-04-26 14:29:05 -0700743void cgroup_rstat_flush_hold(struct cgroup *cgrp);
744void cgroup_rstat_flush_release(void);
745
Tejun Heobd1060a2015-12-07 17:38:53 -0500746/*
Tejun Heod2cc5ed2017-09-25 08:12:04 -0700747 * Basic resource stats.
748 */
Tejun Heod2cc5ed2017-09-25 08:12:04 -0700749#ifdef CONFIG_CGROUP_CPUACCT
750void cpuacct_charge(struct task_struct *tsk, u64 cputime);
751void cpuacct_account_field(struct task_struct *tsk, int index, u64 val);
752#else
753static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {}
754static inline void cpuacct_account_field(struct task_struct *tsk, int index,
755 u64 val) {}
756#endif
757
Tejun Heo041cd642017-09-25 08:12:05 -0700758void __cgroup_account_cputime(struct cgroup *cgrp, u64 delta_exec);
759void __cgroup_account_cputime_field(struct cgroup *cgrp,
760 enum cpu_usage_stat index, u64 delta_exec);
761
Tejun Heod2cc5ed2017-09-25 08:12:04 -0700762static inline void cgroup_account_cputime(struct task_struct *task,
763 u64 delta_exec)
764{
Tejun Heo041cd642017-09-25 08:12:05 -0700765 struct cgroup *cgrp;
766
Tejun Heod2cc5ed2017-09-25 08:12:04 -0700767 cpuacct_charge(task, delta_exec);
Tejun Heo041cd642017-09-25 08:12:05 -0700768
769 rcu_read_lock();
770 cgrp = task_dfl_cgroup(task);
771 if (cgroup_parent(cgrp))
772 __cgroup_account_cputime(cgrp, delta_exec);
773 rcu_read_unlock();
Tejun Heod2cc5ed2017-09-25 08:12:04 -0700774}
775
776static inline void cgroup_account_cputime_field(struct task_struct *task,
777 enum cpu_usage_stat index,
778 u64 delta_exec)
779{
Tejun Heo041cd642017-09-25 08:12:05 -0700780 struct cgroup *cgrp;
781
Tejun Heod2cc5ed2017-09-25 08:12:04 -0700782 cpuacct_account_field(task, index, delta_exec);
Tejun Heo041cd642017-09-25 08:12:05 -0700783
784 rcu_read_lock();
785 cgrp = task_dfl_cgroup(task);
786 if (cgroup_parent(cgrp))
787 __cgroup_account_cputime_field(cgrp, index, delta_exec);
788 rcu_read_unlock();
Tejun Heod2cc5ed2017-09-25 08:12:04 -0700789}
790
791#else /* CONFIG_CGROUPS */
792
793static inline void cgroup_account_cputime(struct task_struct *task,
794 u64 delta_exec) {}
795static inline void cgroup_account_cputime_field(struct task_struct *task,
796 enum cpu_usage_stat index,
797 u64 delta_exec) {}
798
799#endif /* CONFIG_CGROUPS */
800
801/*
Tejun Heobd1060a2015-12-07 17:38:53 -0500802 * sock->sk_cgrp_data handling. For more info, see sock_cgroup_data
803 * definition in cgroup-defs.h.
804 */
805#ifdef CONFIG_SOCK_CGROUP_DATA
806
807#if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
808extern spinlock_t cgroup_sk_update_lock;
809#endif
810
811void cgroup_sk_alloc_disable(void);
812void cgroup_sk_alloc(struct sock_cgroup_data *skcd);
Cong Wang0505cc42020-07-02 11:52:56 -0700813void cgroup_sk_clone(struct sock_cgroup_data *skcd);
Tejun Heobd1060a2015-12-07 17:38:53 -0500814void cgroup_sk_free(struct sock_cgroup_data *skcd);
815
816static inline struct cgroup *sock_cgroup_ptr(struct sock_cgroup_data *skcd)
817{
818#if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
819 unsigned long v;
820
821 /*
822 * @skcd->val is 64bit but the following is safe on 32bit too as we
823 * just need the lower ulong to be written and read atomically.
824 */
825 v = READ_ONCE(skcd->val);
826
Cong Wang0505cc42020-07-02 11:52:56 -0700827 if (v & 3)
Tejun Heobd1060a2015-12-07 17:38:53 -0500828 return &cgrp_dfl_root.cgrp;
829
830 return (struct cgroup *)(unsigned long)v ?: &cgrp_dfl_root.cgrp;
831#else
832 return (struct cgroup *)(unsigned long)skcd->val;
833#endif
834}
835
836#else /* CONFIG_CGROUP_DATA */
837
838static inline void cgroup_sk_alloc(struct sock_cgroup_data *skcd) {}
Cong Wang0505cc42020-07-02 11:52:56 -0700839static inline void cgroup_sk_clone(struct sock_cgroup_data *skcd) {}
Tejun Heobd1060a2015-12-07 17:38:53 -0500840static inline void cgroup_sk_free(struct sock_cgroup_data *skcd) {}
841
842#endif /* CONFIG_CGROUP_DATA */
843
Aditya Kalia79a9082016-01-29 02:54:06 -0600844struct cgroup_namespace {
Elena Reshetova387ad962017-02-20 12:19:00 +0200845 refcount_t count;
Aditya Kalia79a9082016-01-29 02:54:06 -0600846 struct ns_common ns;
847 struct user_namespace *user_ns;
Eric W. Biedermand08311d2016-08-08 14:25:30 -0500848 struct ucounts *ucounts;
Aditya Kalia79a9082016-01-29 02:54:06 -0600849 struct css_set *root_cset;
850};
851
852extern struct cgroup_namespace init_cgroup_ns;
853
854#ifdef CONFIG_CGROUPS
855
856void free_cgroup_ns(struct cgroup_namespace *ns);
857
858struct cgroup_namespace *copy_cgroup_ns(unsigned long flags,
859 struct user_namespace *user_ns,
860 struct cgroup_namespace *old_ns);
861
Tejun Heo4c737b42016-08-10 11:23:44 -0400862int cgroup_path_ns(struct cgroup *cgrp, char *buf, size_t buflen,
863 struct cgroup_namespace *ns);
Aditya Kalia79a9082016-01-29 02:54:06 -0600864
865#else /* !CONFIG_CGROUPS */
866
867static inline void free_cgroup_ns(struct cgroup_namespace *ns) { }
868static inline struct cgroup_namespace *
869copy_cgroup_ns(unsigned long flags, struct user_namespace *user_ns,
870 struct cgroup_namespace *old_ns)
871{
872 return old_ns;
873}
874
875#endif /* !CONFIG_CGROUPS */
876
877static inline void get_cgroup_ns(struct cgroup_namespace *ns)
878{
879 if (ns)
Elena Reshetova387ad962017-02-20 12:19:00 +0200880 refcount_inc(&ns->count);
Aditya Kalia79a9082016-01-29 02:54:06 -0600881}
882
883static inline void put_cgroup_ns(struct cgroup_namespace *ns)
884{
Elena Reshetova387ad962017-02-20 12:19:00 +0200885 if (ns && refcount_dec_and_test(&ns->count))
Aditya Kalia79a9082016-01-29 02:54:06 -0600886 free_cgroup_ns(ns);
887}
888
Paul Menageddbcc7e2007-10-18 23:39:30 -0700889#endif /* _LINUX_CGROUP_H */