Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
ddbcc7e8 PM |
2 | #ifndef _LINUX_CGROUP_H |
3 | #define _LINUX_CGROUP_H | |
4 | /* | |
5 | * cgroup interface | |
6 | * | |
7 | * Copyright (C) 2003 BULL SA | |
8 | * Copyright (C) 2004-2006 Silicon Graphics, Inc. | |
9 | * | |
10 | */ | |
11 | ||
12 | #include <linux/sched.h> | |
ddbcc7e8 PM |
13 | #include <linux/cpumask.h> |
14 | #include <linux/nodemask.h> | |
eb6fd504 | 15 | #include <linux/rculist.h> |
846c7bb0 | 16 | #include <linux/cgroupstats.h> |
25a7e684 | 17 | #include <linux/fs.h> |
7da11279 | 18 | #include <linux/seq_file.h> |
2bd59d48 | 19 | #include <linux/kernfs.h> |
49d1dc4b | 20 | #include <linux/jump_label.h> |
a79a908f AK |
21 | #include <linux/types.h> |
22 | #include <linux/ns_common.h> | |
23 | #include <linux/nsproxy.h> | |
24 | #include <linux/user_namespace.h> | |
387ad967 | 25 | #include <linux/refcount.h> |
d2cc5ed6 | 26 | #include <linux/kernel_stat.h> |
ddbcc7e8 | 27 | |
b4a04ab7 | 28 | #include <linux/cgroup-defs.h> |
ddbcc7e8 | 29 | |
ef2c41cf CB |
30 | struct kernel_clone_args; |
31 | ||
ddbcc7e8 PM |
32 | #ifdef CONFIG_CGROUPS |
33 | ||
6abc8ca1 | 34 | /* |
08b2b6fd | 35 | * All weight knobs on the default hierarchy should use the following min, |
6abc8ca1 TH |
36 | * default and max values. The default value is the logarithmic center of |
37 | * MIN and MAX and allows 100x to be expressed in both directions. | |
38 | */ | |
39 | #define CGROUP_WEIGHT_MIN 1 | |
40 | #define CGROUP_WEIGHT_DFL 100 | |
41 | #define CGROUP_WEIGHT_MAX 10000 | |
42 | ||
bc2fb7ed TH |
43 | /* walk only threadgroup leaders */ |
44 | #define CSS_TASK_ITER_PROCS (1U << 0) | |
450ee0c1 TH |
45 | /* walk all threaded css_sets in the domain */ |
46 | #define CSS_TASK_ITER_THREADED (1U << 1) | |
bc2fb7ed | 47 | |
b636fd38 TH |
48 | /* internal flags */ |
49 | #define CSS_TASK_ITER_SKIPPED (1U << 16) | |
50 | ||
c326aa2b TH |
51 | /* a css_task_iter should be treated as an opaque object */ |
52 | struct css_task_iter { | |
53 | struct cgroup_subsys *ss; | |
bc2fb7ed | 54 | unsigned int flags; |
c326aa2b TH |
55 | |
56 | struct list_head *cset_pos; | |
57 | struct list_head *cset_head; | |
ddbcc7e8 | 58 | |
450ee0c1 TH |
59 | struct list_head *tcset_pos; |
60 | struct list_head *tcset_head; | |
61 | ||
c326aa2b | 62 | struct list_head *task_pos; |
ed27b9f7 | 63 | |
9c974c77 | 64 | struct list_head *cur_tasks_head; |
ed27b9f7 | 65 | struct css_set *cur_cset; |
450ee0c1 | 66 | struct css_set *cur_dcset; |
ed27b9f7 TH |
67 | struct task_struct *cur_task; |
68 | struct list_head iters_node; /* css_set->task_iters */ | |
c326aa2b | 69 | }; |
ddbcc7e8 | 70 | |
fbf83212 | 71 | extern struct file_system_type cgroup_fs_type; |
c326aa2b TH |
72 | extern struct cgroup_root cgrp_dfl_root; |
73 | extern struct css_set init_css_set; | |
a424316c | 74 | |
c326aa2b | 75 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys; |
817929ec | 76 | #include <linux/cgroup_subsys.h> |
817929ec PM |
77 | #undef SUBSYS |
78 | ||
49d1dc4b TH |
79 | #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 | ||
c326aa2b TH |
99 | bool css_has_online_children(struct cgroup_subsys_state *css); |
100 | struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss); | |
fc5a828b DZ |
101 | struct cgroup_subsys_state *cgroup_e_css(struct cgroup *cgroup, |
102 | struct cgroup_subsys *ss); | |
c326aa2b TH |
103 | struct cgroup_subsys_state *cgroup_get_e_css(struct cgroup *cgroup, |
104 | struct cgroup_subsys *ss); | |
105 | struct cgroup_subsys_state *css_tryget_online_from_dir(struct dentry *dentry, | |
106 | struct cgroup_subsys *ss); | |
107 | ||
16af4396 | 108 | struct cgroup *cgroup_get_from_path(const char *path); |
1f3fe7eb | 109 | struct cgroup *cgroup_get_from_fd(int fd); |
a6d1ce59 | 110 | struct cgroup *cgroup_v1v2_get_from_fd(int fd); |
16af4396 | 111 | |
c326aa2b TH |
112 | int cgroup_attach_task_all(struct task_struct *from, struct task_struct *); |
113 | int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from); | |
114 | ||
115 | int cgroup_add_dfl_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); | |
116 | int cgroup_add_legacy_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); | |
117 | int cgroup_rm_cftypes(struct cftype *cfts); | |
34c06254 | 118 | void cgroup_file_notify(struct cgroup_file *cfile); |
e2691f6b | 119 | void cgroup_file_show(struct cgroup_file *cfile, bool show); |
c326aa2b | 120 | |
4c737b41 | 121 | int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen); |
c326aa2b TH |
122 | int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry); |
123 | int proc_cgroup_show(struct seq_file *m, struct pid_namespace *ns, | |
124 | struct pid *pid, struct task_struct *tsk); | |
125 | ||
126 | void cgroup_fork(struct task_struct *p); | |
ef2c41cf CB |
127 | extern int cgroup_can_fork(struct task_struct *p, |
128 | struct kernel_clone_args *kargs); | |
129 | extern void cgroup_cancel_fork(struct task_struct *p, | |
130 | struct kernel_clone_args *kargs); | |
131 | extern void cgroup_post_fork(struct task_struct *p, | |
132 | struct kernel_clone_args *kargs); | |
c326aa2b | 133 | void cgroup_exit(struct task_struct *p); |
51bee5ab | 134 | void cgroup_release(struct task_struct *p); |
2e91fa7f | 135 | void cgroup_free(struct task_struct *p); |
c326aa2b TH |
136 | |
137 | int cgroup_init_early(void); | |
138 | int cgroup_init(void); | |
139 | ||
a5e112e6 TH |
140 | int cgroup_parse_float(const char *input, unsigned dec_shift, s64 *v); |
141 | ||
5c9d535b | 142 | /* |
c326aa2b TH |
143 | * Iteration helpers and macros. |
144 | */ | |
145 | ||
146 | struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos, | |
147 | struct cgroup_subsys_state *parent); | |
148 | struct cgroup_subsys_state *css_next_descendant_pre(struct cgroup_subsys_state *pos, | |
149 | struct cgroup_subsys_state *css); | |
150 | struct cgroup_subsys_state *css_rightmost_descendant(struct cgroup_subsys_state *pos); | |
151 | struct cgroup_subsys_state *css_next_descendant_post(struct cgroup_subsys_state *pos, | |
152 | struct cgroup_subsys_state *css); | |
153 | ||
1f7dd3e5 TH |
154 | struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset, |
155 | struct cgroup_subsys_state **dst_cssp); | |
156 | struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset, | |
157 | struct cgroup_subsys_state **dst_cssp); | |
c326aa2b | 158 | |
bc2fb7ed | 159 | void css_task_iter_start(struct cgroup_subsys_state *css, unsigned int flags, |
c326aa2b TH |
160 | struct css_task_iter *it); |
161 | struct task_struct *css_task_iter_next(struct css_task_iter *it); | |
162 | void css_task_iter_end(struct css_task_iter *it); | |
163 | ||
164 | /** | |
165 | * css_for_each_child - iterate through children of a css | |
166 | * @pos: the css * to use as the loop cursor | |
167 | * @parent: css whose children to walk | |
168 | * | |
169 | * Walk @parent's children. Must be called under rcu_read_lock(). | |
170 | * | |
171 | * If a subsystem synchronizes ->css_online() and the start of iteration, a | |
172 | * css which finished ->css_online() is guaranteed to be visible in the | |
173 | * future iterations and will stay visible until the last reference is put. | |
174 | * A css which hasn't finished ->css_online() or already finished | |
175 | * ->css_offline() may show up during traversal. It's each subsystem's | |
176 | * responsibility to synchronize against on/offlining. | |
5c9d535b | 177 | * |
c326aa2b TH |
178 | * It is allowed to temporarily drop RCU read lock during iteration. The |
179 | * caller is responsible for ensuring that @pos remains accessible until | |
180 | * the start of the next iteration by, for example, bumping the css refcnt. | |
5c9d535b | 181 | */ |
c326aa2b TH |
182 | #define css_for_each_child(pos, parent) \ |
183 | for ((pos) = css_next_child(NULL, (parent)); (pos); \ | |
184 | (pos) = css_next_child((pos), (parent))) | |
ddbcc7e8 | 185 | |
c326aa2b TH |
186 | /** |
187 | * css_for_each_descendant_pre - pre-order walk of a css's descendants | |
188 | * @pos: the css * to use as the loop cursor | |
189 | * @root: css whose descendants to walk | |
190 | * | |
191 | * Walk @root's descendants. @root is included in the iteration and the | |
192 | * first node to be visited. Must be called under rcu_read_lock(). | |
193 | * | |
194 | * If a subsystem synchronizes ->css_online() and the start of iteration, a | |
195 | * css which finished ->css_online() is guaranteed to be visible in the | |
196 | * future iterations and will stay visible until the last reference is put. | |
197 | * A css which hasn't finished ->css_online() or already finished | |
198 | * ->css_offline() may show up during traversal. It's each subsystem's | |
199 | * responsibility to synchronize against on/offlining. | |
200 | * | |
201 | * For example, the following guarantees that a descendant can't escape | |
202 | * state updates of its ancestors. | |
203 | * | |
204 | * my_online(@css) | |
205 | * { | |
206 | * Lock @css's parent and @css; | |
207 | * Inherit state from the parent; | |
208 | * Unlock both. | |
209 | * } | |
210 | * | |
211 | * my_update_state(@css) | |
212 | * { | |
213 | * css_for_each_descendant_pre(@pos, @css) { | |
214 | * Lock @pos; | |
215 | * if (@pos == @css) | |
216 | * Update @css's state; | |
217 | * else | |
218 | * Verify @pos is alive and inherit state from its parent; | |
219 | * Unlock @pos; | |
220 | * } | |
221 | * } | |
222 | * | |
223 | * As long as the inheriting step, including checking the parent state, is | |
224 | * enclosed inside @pos locking, double-locking the parent isn't necessary | |
225 | * while inheriting. The state update to the parent is guaranteed to be | |
226 | * visible by walking order and, as long as inheriting operations to the | |
227 | * same @pos are atomic to each other, multiple updates racing each other | |
228 | * still result in the correct state. It's guaranateed that at least one | |
229 | * inheritance happens for any css after the latest update to its parent. | |
230 | * | |
231 | * If checking parent's state requires locking the parent, each inheriting | |
232 | * iteration should lock and unlock both @pos->parent and @pos. | |
233 | * | |
234 | * Alternatively, a subsystem may choose to use a single global lock to | |
235 | * synchronize ->css_online() and ->css_offline() against tree-walking | |
236 | * operations. | |
237 | * | |
238 | * It is allowed to temporarily drop RCU read lock during iteration. The | |
239 | * caller is responsible for ensuring that @pos remains accessible until | |
240 | * the start of the next iteration by, for example, bumping the css refcnt. | |
241 | */ | |
242 | #define css_for_each_descendant_pre(pos, css) \ | |
243 | for ((pos) = css_next_descendant_pre(NULL, (css)); (pos); \ | |
244 | (pos) = css_next_descendant_pre((pos), (css))) | |
245 | ||
246 | /** | |
247 | * css_for_each_descendant_post - post-order walk of a css's descendants | |
248 | * @pos: the css * to use as the loop cursor | |
249 | * @css: css whose descendants to walk | |
250 | * | |
251 | * Similar to css_for_each_descendant_pre() but performs post-order | |
252 | * traversal instead. @root is included in the iteration and the last | |
253 | * node to be visited. | |
254 | * | |
255 | * If a subsystem synchronizes ->css_online() and the start of iteration, a | |
256 | * css which finished ->css_online() is guaranteed to be visible in the | |
257 | * future iterations and will stay visible until the last reference is put. | |
258 | * A css which hasn't finished ->css_online() or already finished | |
259 | * ->css_offline() may show up during traversal. It's each subsystem's | |
260 | * responsibility to synchronize against on/offlining. | |
261 | * | |
262 | * Note that the walk visibility guarantee example described in pre-order | |
263 | * walk doesn't apply the same to post-order walks. | |
264 | */ | |
265 | #define css_for_each_descendant_post(pos, css) \ | |
266 | for ((pos) = css_next_descendant_post(NULL, (css)); (pos); \ | |
267 | (pos) = css_next_descendant_post((pos), (css))) | |
268 | ||
269 | /** | |
270 | * cgroup_taskset_for_each - iterate cgroup_taskset | |
271 | * @task: the loop cursor | |
1f7dd3e5 | 272 | * @dst_css: the destination css |
c326aa2b | 273 | * @tset: taskset to iterate |
4530eddb TH |
274 | * |
275 | * @tset may contain multiple tasks and they may belong to multiple | |
1f7dd3e5 TH |
276 | * processes. |
277 | * | |
278 | * On the v2 hierarchy, there may be tasks from multiple processes and they | |
279 | * may not share the source or destination csses. | |
280 | * | |
281 | * On traditional hierarchies, when there are multiple tasks in @tset, if a | |
282 | * task of a process is in @tset, all tasks of the process are in @tset. | |
283 | * Also, all are guaranteed to share the same source and destination csses. | |
4530eddb TH |
284 | * |
285 | * Iteration is not in any specific order. | |
c326aa2b | 286 | */ |
1f7dd3e5 TH |
287 | #define cgroup_taskset_for_each(task, dst_css, tset) \ |
288 | for ((task) = cgroup_taskset_first((tset), &(dst_css)); \ | |
289 | (task); \ | |
290 | (task) = cgroup_taskset_next((tset), &(dst_css))) | |
ddbcc7e8 | 291 | |
4530eddb TH |
292 | /** |
293 | * cgroup_taskset_for_each_leader - iterate group leaders in a cgroup_taskset | |
294 | * @leader: the loop cursor | |
1f7dd3e5 | 295 | * @dst_css: the destination css |
7b4632f0 | 296 | * @tset: taskset to iterate |
4530eddb TH |
297 | * |
298 | * Iterate threadgroup leaders of @tset. For single-task migrations, @tset | |
299 | * may not contain any. | |
300 | */ | |
1f7dd3e5 TH |
301 | #define cgroup_taskset_for_each_leader(leader, dst_css, tset) \ |
302 | for ((leader) = cgroup_taskset_first((tset), &(dst_css)); \ | |
303 | (leader); \ | |
304 | (leader) = cgroup_taskset_next((tset), &(dst_css))) \ | |
4530eddb TH |
305 | if ((leader) != (leader)->group_leader) \ |
306 | ; \ | |
307 | else | |
308 | ||
c326aa2b TH |
309 | /* |
310 | * Inline functions. | |
311 | */ | |
ddbcc7e8 | 312 | |
6ab42860 TH |
313 | #ifdef CONFIG_DEBUG_CGROUP_REF |
314 | void css_get(struct cgroup_subsys_state *css); | |
315 | void css_get_many(struct cgroup_subsys_state *css, unsigned int n); | |
316 | bool css_tryget(struct cgroup_subsys_state *css); | |
317 | bool css_tryget_online(struct cgroup_subsys_state *css); | |
318 | void css_put(struct cgroup_subsys_state *css); | |
319 | void css_put_many(struct cgroup_subsys_state *css, unsigned int n); | |
320 | #else | |
321 | #define CGROUP_REF_FN_ATTRS static inline | |
79a7f41f | 322 | #define CGROUP_REF_EXPORT(fn) |
6ab42860 TH |
323 | #include <linux/cgroup_refcnt.h> |
324 | #endif | |
325 | ||
a3b89864 | 326 | static inline u64 cgroup_id(const struct cgroup *cgrp) |
74321038 TH |
327 | { |
328 | return cgrp->kn->id; | |
329 | } | |
330 | ||
41c25707 TH |
331 | /** |
332 | * css_is_dying - test whether the specified css is dying | |
333 | * @css: target css | |
334 | * | |
335 | * Test whether @css is in the process of offlining or already offline. In | |
336 | * most cases, ->css_online() and ->css_offline() callbacks should be | |
337 | * enough; however, the actual offline operations are RCU delayed and this | |
338 | * test returns %true also when @css is scheduled to be offlined. | |
339 | * | |
340 | * This is useful, for example, when the use case requires synchronous | |
341 | * behavior with respect to cgroup removal. cgroup removal schedules css | |
342 | * offlining but the css can seem alive while the operation is being | |
343 | * delayed. If the delay affects user visible semantics, this test can be | |
344 | * used to resolve the situation. | |
345 | */ | |
346 | static inline bool css_is_dying(struct cgroup_subsys_state *css) | |
347 | { | |
348 | return !(css->flags & CSS_NO_REF) && percpu_ref_is_dying(&css->refcnt); | |
349 | } | |
350 | ||
3e48930c TH |
351 | static inline void cgroup_get(struct cgroup *cgrp) |
352 | { | |
353 | css_get(&cgrp->self); | |
354 | } | |
355 | ||
356 | static inline bool cgroup_tryget(struct cgroup *cgrp) | |
357 | { | |
358 | return css_tryget(&cgrp->self); | |
359 | } | |
360 | ||
16af4396 TH |
361 | static inline void cgroup_put(struct cgroup *cgrp) |
362 | { | |
363 | css_put(&cgrp->self); | |
364 | } | |
365 | ||
354ed597 YZ |
366 | extern struct mutex cgroup_mutex; |
367 | ||
368 | static inline void cgroup_lock(void) | |
369 | { | |
370 | mutex_lock(&cgroup_mutex); | |
371 | } | |
372 | ||
373 | static inline void cgroup_unlock(void) | |
374 | { | |
375 | mutex_unlock(&cgroup_mutex); | |
376 | } | |
377 | ||
14611e51 TH |
378 | /** |
379 | * task_css_set_check - obtain a task's css_set with extra access conditions | |
380 | * @task: the task to obtain css_set for | |
381 | * @__c: extra condition expression to be passed to rcu_dereference_check() | |
382 | * | |
383 | * A task's css_set is RCU protected, initialized and exited while holding | |
384 | * task_lock(), and can only be modified while holding both cgroup_mutex | |
385 | * and task_lock() while the task is alive. This macro verifies that the | |
386 | * caller is inside proper critical section and returns @task's css_set. | |
387 | * | |
388 | * The caller can also specify additional allowed conditions via @__c, such | |
389 | * as locks used during the cgroup_subsys::attach() methods. | |
dc61b1d6 | 390 | */ |
2219449a | 391 | #ifdef CONFIG_PROVE_RCU |
f0d9a5f1 | 392 | extern spinlock_t css_set_lock; |
14611e51 TH |
393 | #define task_css_set_check(task, __c) \ |
394 | rcu_dereference_check((task)->cgroups, \ | |
f2aa197e | 395 | rcu_read_lock_sched_held() || \ |
0e1d768f | 396 | lockdep_is_held(&cgroup_mutex) || \ |
f0d9a5f1 | 397 | lockdep_is_held(&css_set_lock) || \ |
0e1d768f | 398 | ((task)->flags & PF_EXITING) || (__c)) |
2219449a | 399 | #else |
14611e51 TH |
400 | #define task_css_set_check(task, __c) \ |
401 | rcu_dereference((task)->cgroups) | |
2219449a | 402 | #endif |
dc61b1d6 | 403 | |
14611e51 | 404 | /** |
8af01f56 | 405 | * task_css_check - obtain css for (task, subsys) w/ extra access conds |
14611e51 TH |
406 | * @task: the target task |
407 | * @subsys_id: the target subsystem ID | |
408 | * @__c: extra condition expression to be passed to rcu_dereference_check() | |
409 | * | |
410 | * Return the cgroup_subsys_state for the (@task, @subsys_id) pair. The | |
411 | * synchronization rules are the same as task_css_set_check(). | |
412 | */ | |
8af01f56 | 413 | #define task_css_check(task, subsys_id, __c) \ |
14611e51 TH |
414 | task_css_set_check((task), (__c))->subsys[(subsys_id)] |
415 | ||
416 | /** | |
417 | * task_css_set - obtain a task's css_set | |
418 | * @task: the task to obtain css_set for | |
419 | * | |
420 | * See task_css_set_check(). | |
421 | */ | |
422 | static inline struct css_set *task_css_set(struct task_struct *task) | |
423 | { | |
424 | return task_css_set_check(task, false); | |
425 | } | |
426 | ||
427 | /** | |
8af01f56 | 428 | * task_css - obtain css for (task, subsys) |
14611e51 TH |
429 | * @task: the target task |
430 | * @subsys_id: the target subsystem ID | |
431 | * | |
8af01f56 | 432 | * See task_css_check(). |
14611e51 | 433 | */ |
8af01f56 TH |
434 | static inline struct cgroup_subsys_state *task_css(struct task_struct *task, |
435 | int subsys_id) | |
ddbcc7e8 | 436 | { |
8af01f56 | 437 | return task_css_check(task, subsys_id, false); |
ddbcc7e8 PM |
438 | } |
439 | ||
ec438699 TH |
440 | /** |
441 | * task_get_css - find and get the css for (task, subsys) | |
442 | * @task: the target task | |
443 | * @subsys_id: the target subsystem ID | |
444 | * | |
445 | * Find the css for the (@task, @subsys_id) combination, increment a | |
446 | * reference on and return it. This function is guaranteed to return a | |
18fa84a2 | 447 | * valid css. The returned css may already have been offlined. |
ec438699 TH |
448 | */ |
449 | static inline struct cgroup_subsys_state * | |
450 | task_get_css(struct task_struct *task, int subsys_id) | |
451 | { | |
452 | struct cgroup_subsys_state *css; | |
453 | ||
454 | rcu_read_lock(); | |
455 | while (true) { | |
456 | css = task_css(task, subsys_id); | |
18fa84a2 TH |
457 | /* |
458 | * Can't use css_tryget_online() here. A task which has | |
459 | * PF_EXITING set may stay associated with an offline css. | |
460 | * If such task calls this function, css_tryget_online() | |
461 | * will keep failing. | |
462 | */ | |
463 | if (likely(css_tryget(css))) | |
ec438699 TH |
464 | break; |
465 | cpu_relax(); | |
466 | } | |
467 | rcu_read_unlock(); | |
468 | return css; | |
469 | } | |
470 | ||
5024ae29 TH |
471 | /** |
472 | * task_css_is_root - test whether a task belongs to the root css | |
473 | * @task: the target task | |
474 | * @subsys_id: the target subsystem ID | |
475 | * | |
476 | * Test whether @task belongs to the root css on the specified subsystem. | |
477 | * May be invoked in any context. | |
478 | */ | |
479 | static inline bool task_css_is_root(struct task_struct *task, int subsys_id) | |
480 | { | |
481 | return task_css_check(task, subsys_id, true) == | |
482 | init_css_set.subsys[subsys_id]; | |
483 | } | |
484 | ||
8af01f56 TH |
485 | static inline struct cgroup *task_cgroup(struct task_struct *task, |
486 | int subsys_id) | |
ddbcc7e8 | 487 | { |
8af01f56 | 488 | return task_css(task, subsys_id)->cgroup; |
ddbcc7e8 PM |
489 | } |
490 | ||
3e48930c TH |
491 | static inline struct cgroup *task_dfl_cgroup(struct task_struct *task) |
492 | { | |
493 | return task_css_set(task)->dfl_cgrp; | |
494 | } | |
495 | ||
496 | static inline struct cgroup *cgroup_parent(struct cgroup *cgrp) | |
497 | { | |
498 | struct cgroup_subsys_state *parent_css = cgrp->self.parent; | |
499 | ||
500 | if (parent_css) | |
501 | return container_of(parent_css, struct cgroup, self); | |
502 | return NULL; | |
503 | } | |
504 | ||
b11cfb58 TH |
505 | /** |
506 | * cgroup_is_descendant - test ancestry | |
507 | * @cgrp: the cgroup to be tested | |
508 | * @ancestor: possible ancestor of @cgrp | |
509 | * | |
510 | * Test whether @cgrp is a descendant of @ancestor. It also returns %true | |
511 | * if @cgrp == @ancestor. This function is safe to call as long as @cgrp | |
512 | * and @ancestor are accessible. | |
513 | */ | |
514 | static inline bool cgroup_is_descendant(struct cgroup *cgrp, | |
515 | struct cgroup *ancestor) | |
516 | { | |
517 | if (cgrp->root != ancestor->root || cgrp->level < ancestor->level) | |
518 | return false; | |
7f203bc8 | 519 | return cgrp->ancestors[ancestor->level] == ancestor; |
b11cfb58 TH |
520 | } |
521 | ||
77236281 AI |
522 | /** |
523 | * cgroup_ancestor - find ancestor of cgroup | |
524 | * @cgrp: cgroup to find ancestor of | |
525 | * @ancestor_level: level of ancestor to find starting from root | |
526 | * | |
527 | * Find ancestor of cgroup at specified level starting from root if it exists | |
528 | * and return pointer to it. Return NULL if @cgrp doesn't have ancestor at | |
529 | * @ancestor_level. | |
530 | * | |
531 | * This function is safe to call as long as @cgrp is accessible. | |
532 | */ | |
533 | static inline struct cgroup *cgroup_ancestor(struct cgroup *cgrp, | |
534 | int ancestor_level) | |
535 | { | |
7f203bc8 | 536 | if (ancestor_level < 0 || ancestor_level > cgrp->level) |
77236281 | 537 | return NULL; |
7f203bc8 | 538 | return cgrp->ancestors[ancestor_level]; |
77236281 AI |
539 | } |
540 | ||
aed704b7 SD |
541 | /** |
542 | * task_under_cgroup_hierarchy - test task's membership of cgroup ancestry | |
543 | * @task: the task to be tested | |
544 | * @ancestor: possible ancestor of @task's cgroup | |
545 | * | |
546 | * Tests whether @task's default cgroup hierarchy is a descendant of @ancestor. | |
547 | * It follows all the same rules as cgroup_is_descendant, and only applies | |
548 | * to the default hierarchy. | |
549 | */ | |
550 | static inline bool task_under_cgroup_hierarchy(struct task_struct *task, | |
551 | struct cgroup *ancestor) | |
552 | { | |
553 | struct css_set *cset = task_css_set(task); | |
554 | ||
555 | return cgroup_is_descendant(cset->dfl_cgrp, ancestor); | |
556 | } | |
557 | ||
07bc356e | 558 | /* no synchronization, the result can only be used as a hint */ |
27bd4dbb | 559 | static inline bool cgroup_is_populated(struct cgroup *cgrp) |
07bc356e | 560 | { |
454000ad TH |
561 | return cgrp->nr_populated_csets + cgrp->nr_populated_domain_children + |
562 | cgrp->nr_populated_threaded_children; | |
07bc356e | 563 | } |
f3d46500 | 564 | |
f29374b1 | 565 | /* returns ino associated with a cgroup */ |
b1664924 TH |
566 | static inline ino_t cgroup_ino(struct cgroup *cgrp) |
567 | { | |
67c0496e | 568 | return kernfs_ino(cgrp->kn); |
b1664924 | 569 | } |
3ebb2b6e | 570 | |
b4168640 TH |
571 | /* cft/css accessors for cftype->write() operation */ |
572 | static inline struct cftype *of_cft(struct kernfs_open_file *of) | |
7da11279 | 573 | { |
2bd59d48 | 574 | return of->kn->priv; |
7da11279 | 575 | } |
0f0a2b4f | 576 | |
b4168640 | 577 | struct cgroup_subsys_state *of_css(struct kernfs_open_file *of); |
817929ec | 578 | |
b4168640 TH |
579 | /* cft/css accessors for cftype->seq_*() operations */ |
580 | static inline struct cftype *seq_cft(struct seq_file *seq) | |
581 | { | |
582 | return of_cft(seq->private); | |
583 | } | |
e535837b | 584 | |
b4168640 TH |
585 | static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq) |
586 | { | |
587 | return of_css(seq->private); | |
588 | } | |
31583bb0 | 589 | |
e61734c5 TH |
590 | /* |
591 | * Name / path handling functions. All are thin wrappers around the kernfs | |
592 | * counterparts and can be called under any context. | |
593 | */ | |
38460b48 | 594 | |
e61734c5 TH |
595 | static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen) |
596 | { | |
fdce6bf8 | 597 | return kernfs_name(cgrp->kn, buf, buflen); |
e61734c5 | 598 | } |
ddbcc7e8 | 599 | |
4c737b41 | 600 | static inline int cgroup_path(struct cgroup *cgrp, char *buf, size_t buflen) |
e61734c5 | 601 | { |
fdce6bf8 | 602 | return kernfs_path(cgrp->kn, buf, buflen); |
e61734c5 | 603 | } |
f3ba5380 | 604 | |
e61734c5 TH |
605 | static inline void pr_cont_cgroup_name(struct cgroup *cgrp) |
606 | { | |
fdce6bf8 | 607 | pr_cont_kernfs_name(cgrp->kn); |
e61734c5 | 608 | } |
ddbcc7e8 | 609 | |
e61734c5 | 610 | static inline void pr_cont_cgroup_path(struct cgroup *cgrp) |
846c7bb0 | 611 | { |
fdce6bf8 | 612 | pr_cont_kernfs_path(cgrp->kn); |
846c7bb0 | 613 | } |
ddbcc7e8 | 614 | |
3958e2d0 SB |
615 | bool cgroup_psi_enabled(void); |
616 | ||
77f88796 TH |
617 | static inline void cgroup_init_kthreadd(void) |
618 | { | |
619 | /* | |
620 | * kthreadd is inherited by all kthreads, keep it in the root so | |
621 | * that the new kthreads are guaranteed to stay in the root until | |
622 | * initialization is finished. | |
623 | */ | |
624 | current->no_cgroup_migration = 1; | |
625 | } | |
626 | ||
627 | static inline void cgroup_kthread_ready(void) | |
628 | { | |
629 | /* | |
630 | * This kthread finished initialization. The creator should have | |
631 | * set PF_NO_SETAFFINITY if this kthread should stay in the root. | |
632 | */ | |
633 | current->no_cgroup_migration = 0; | |
634 | } | |
635 | ||
67c0496e | 636 | void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen); |
6b658c48 | 637 | struct cgroup *cgroup_get_from_id(u64 id); |
ddbcc7e8 PM |
638 | #else /* !CONFIG_CGROUPS */ |
639 | ||
f3ba5380 | 640 | struct cgroup_subsys_state; |
aed704b7 | 641 | struct cgroup; |
f3ba5380 | 642 | |
a3b89864 | 643 | static inline u64 cgroup_id(const struct cgroup *cgrp) { return 1; } |
9b0eb69b | 644 | static inline void css_get(struct cgroup_subsys_state *css) {} |
c326aa2b | 645 | static inline void css_put(struct cgroup_subsys_state *css) {} |
354ed597 YZ |
646 | static inline void cgroup_lock(void) {} |
647 | static inline void cgroup_unlock(void) {} | |
31583bb0 | 648 | static inline int cgroup_attach_task_all(struct task_struct *from, |
c326aa2b TH |
649 | struct task_struct *t) { return 0; } |
650 | static inline int cgroupstats_build(struct cgroupstats *stats, | |
651 | struct dentry *dentry) { return -EINVAL; } | |
652 | ||
b4f48b63 | 653 | static inline void cgroup_fork(struct task_struct *p) {} |
ef2c41cf CB |
654 | static inline int cgroup_can_fork(struct task_struct *p, |
655 | struct kernel_clone_args *kargs) { return 0; } | |
656 | static inline void cgroup_cancel_fork(struct task_struct *p, | |
657 | struct kernel_clone_args *kargs) {} | |
658 | static inline void cgroup_post_fork(struct task_struct *p, | |
659 | struct kernel_clone_args *kargs) {} | |
1ec41830 | 660 | static inline void cgroup_exit(struct task_struct *p) {} |
51bee5ab | 661 | static inline void cgroup_release(struct task_struct *p) {} |
2e91fa7f | 662 | static inline void cgroup_free(struct task_struct *p) {} |
ddbcc7e8 | 663 | |
c326aa2b TH |
664 | static inline int cgroup_init_early(void) { return 0; } |
665 | static inline int cgroup_init(void) { return 0; } | |
77f88796 TH |
666 | static inline void cgroup_init_kthreadd(void) {} |
667 | static inline void cgroup_kthread_ready(void) {} | |
d7926ee3 | 668 | |
2ce7135a JW |
669 | static inline struct cgroup *cgroup_parent(struct cgroup *cgrp) |
670 | { | |
671 | return NULL; | |
672 | } | |
673 | ||
3958e2d0 SB |
674 | static inline bool cgroup_psi_enabled(void) |
675 | { | |
676 | return false; | |
677 | } | |
678 | ||
aed704b7 SD |
679 | static inline bool task_under_cgroup_hierarchy(struct task_struct *task, |
680 | struct cgroup *ancestor) | |
681 | { | |
682 | return true; | |
683 | } | |
69fd5c39 | 684 | |
67c0496e TH |
685 | static inline void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen) |
686 | {} | |
ddbcc7e8 PM |
687 | #endif /* !CONFIG_CGROUPS */ |
688 | ||
6162cef0 | 689 | #ifdef CONFIG_CGROUPS |
d2cc5ed6 | 690 | /* |
6162cef0 | 691 | * cgroup scalable recursive statistics. |
d2cc5ed6 | 692 | */ |
6162cef0 TH |
693 | void cgroup_rstat_updated(struct cgroup *cgrp, int cpu); |
694 | void cgroup_rstat_flush(struct cgroup *cgrp); | |
8bff9a04 | 695 | void cgroup_rstat_flush_atomic(struct cgroup *cgrp); |
6162cef0 TH |
696 | void cgroup_rstat_flush_hold(struct cgroup *cgrp); |
697 | void cgroup_rstat_flush_release(void); | |
d2cc5ed6 | 698 | |
6162cef0 TH |
699 | /* |
700 | * Basic resource stats. | |
701 | */ | |
d2cc5ed6 TH |
702 | #ifdef CONFIG_CGROUP_CPUACCT |
703 | void cpuacct_charge(struct task_struct *tsk, u64 cputime); | |
704 | void cpuacct_account_field(struct task_struct *tsk, int index, u64 val); | |
705 | #else | |
706 | static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {} | |
707 | static inline void cpuacct_account_field(struct task_struct *tsk, int index, | |
708 | u64 val) {} | |
709 | #endif | |
710 | ||
041cd640 TH |
711 | void __cgroup_account_cputime(struct cgroup *cgrp, u64 delta_exec); |
712 | void __cgroup_account_cputime_field(struct cgroup *cgrp, | |
713 | enum cpu_usage_stat index, u64 delta_exec); | |
714 | ||
d2cc5ed6 TH |
715 | static inline void cgroup_account_cputime(struct task_struct *task, |
716 | u64 delta_exec) | |
717 | { | |
041cd640 TH |
718 | struct cgroup *cgrp; |
719 | ||
d2cc5ed6 | 720 | cpuacct_charge(task, delta_exec); |
041cd640 | 721 | |
041cd640 TH |
722 | cgrp = task_dfl_cgroup(task); |
723 | if (cgroup_parent(cgrp)) | |
724 | __cgroup_account_cputime(cgrp, delta_exec); | |
d2cc5ed6 TH |
725 | } |
726 | ||
727 | static inline void cgroup_account_cputime_field(struct task_struct *task, | |
728 | enum cpu_usage_stat index, | |
729 | u64 delta_exec) | |
730 | { | |
041cd640 TH |
731 | struct cgroup *cgrp; |
732 | ||
d2cc5ed6 | 733 | cpuacct_account_field(task, index, delta_exec); |
041cd640 | 734 | |
041cd640 TH |
735 | cgrp = task_dfl_cgroup(task); |
736 | if (cgroup_parent(cgrp)) | |
737 | __cgroup_account_cputime_field(cgrp, index, delta_exec); | |
d2cc5ed6 TH |
738 | } |
739 | ||
740 | #else /* CONFIG_CGROUPS */ | |
741 | ||
742 | static inline void cgroup_account_cputime(struct task_struct *task, | |
743 | u64 delta_exec) {} | |
744 | static inline void cgroup_account_cputime_field(struct task_struct *task, | |
745 | enum cpu_usage_stat index, | |
746 | u64 delta_exec) {} | |
747 | ||
748 | #endif /* CONFIG_CGROUPS */ | |
749 | ||
bd1060a1 TH |
750 | /* |
751 | * sock->sk_cgrp_data handling. For more info, see sock_cgroup_data | |
752 | * definition in cgroup-defs.h. | |
753 | */ | |
754 | #ifdef CONFIG_SOCK_CGROUP_DATA | |
755 | ||
bd1060a1 | 756 | void cgroup_sk_alloc(struct sock_cgroup_data *skcd); |
ad0f75e5 | 757 | void cgroup_sk_clone(struct sock_cgroup_data *skcd); |
bd1060a1 TH |
758 | void cgroup_sk_free(struct sock_cgroup_data *skcd); |
759 | ||
760 | static inline struct cgroup *sock_cgroup_ptr(struct sock_cgroup_data *skcd) | |
761 | { | |
8520e224 | 762 | return skcd->cgroup; |
bd1060a1 TH |
763 | } |
764 | ||
765 | #else /* CONFIG_CGROUP_DATA */ | |
766 | ||
767 | static inline void cgroup_sk_alloc(struct sock_cgroup_data *skcd) {} | |
ad0f75e5 | 768 | static inline void cgroup_sk_clone(struct sock_cgroup_data *skcd) {} |
bd1060a1 TH |
769 | static inline void cgroup_sk_free(struct sock_cgroup_data *skcd) {} |
770 | ||
771 | #endif /* CONFIG_CGROUP_DATA */ | |
772 | ||
a79a908f | 773 | struct cgroup_namespace { |
a79a908f AK |
774 | struct ns_common ns; |
775 | struct user_namespace *user_ns; | |
d08311dd | 776 | struct ucounts *ucounts; |
a79a908f AK |
777 | struct css_set *root_cset; |
778 | }; | |
779 | ||
780 | extern struct cgroup_namespace init_cgroup_ns; | |
781 | ||
782 | #ifdef CONFIG_CGROUPS | |
783 | ||
784 | void free_cgroup_ns(struct cgroup_namespace *ns); | |
785 | ||
786 | struct cgroup_namespace *copy_cgroup_ns(unsigned long flags, | |
787 | struct user_namespace *user_ns, | |
788 | struct cgroup_namespace *old_ns); | |
789 | ||
4c737b41 TH |
790 | int cgroup_path_ns(struct cgroup *cgrp, char *buf, size_t buflen, |
791 | struct cgroup_namespace *ns); | |
a79a908f AK |
792 | |
793 | #else /* !CONFIG_CGROUPS */ | |
794 | ||
795 | static inline void free_cgroup_ns(struct cgroup_namespace *ns) { } | |
796 | static inline struct cgroup_namespace * | |
797 | copy_cgroup_ns(unsigned long flags, struct user_namespace *user_ns, | |
798 | struct cgroup_namespace *old_ns) | |
799 | { | |
800 | return old_ns; | |
801 | } | |
802 | ||
803 | #endif /* !CONFIG_CGROUPS */ | |
804 | ||
805 | static inline void get_cgroup_ns(struct cgroup_namespace *ns) | |
806 | { | |
807 | if (ns) | |
f387882d | 808 | refcount_inc(&ns->ns.count); |
a79a908f AK |
809 | } |
810 | ||
811 | static inline void put_cgroup_ns(struct cgroup_namespace *ns) | |
812 | { | |
f387882d | 813 | if (ns && refcount_dec_and_test(&ns->ns.count)) |
a79a908f AK |
814 | free_cgroup_ns(ns); |
815 | } | |
816 | ||
76f969e8 RG |
817 | #ifdef CONFIG_CGROUPS |
818 | ||
819 | void cgroup_enter_frozen(void); | |
820 | void cgroup_leave_frozen(bool always_leave); | |
821 | void cgroup_update_frozen(struct cgroup *cgrp); | |
822 | void cgroup_freeze(struct cgroup *cgrp, bool freeze); | |
823 | void cgroup_freezer_migrate_task(struct task_struct *task, struct cgroup *src, | |
824 | struct cgroup *dst); | |
96b9c592 | 825 | |
76f969e8 RG |
826 | static inline bool cgroup_task_frozen(struct task_struct *task) |
827 | { | |
828 | return task->frozen; | |
829 | } | |
830 | ||
831 | #else /* !CONFIG_CGROUPS */ | |
832 | ||
833 | static inline void cgroup_enter_frozen(void) { } | |
834 | static inline void cgroup_leave_frozen(bool always_leave) { } | |
76f969e8 RG |
835 | static inline bool cgroup_task_frozen(struct task_struct *task) |
836 | { | |
837 | return false; | |
838 | } | |
839 | ||
840 | #endif /* !CONFIG_CGROUPS */ | |
841 | ||
4bfc0bb2 RG |
842 | #ifdef CONFIG_CGROUP_BPF |
843 | static inline void cgroup_bpf_get(struct cgroup *cgrp) | |
844 | { | |
845 | percpu_ref_get(&cgrp->bpf.refcnt); | |
846 | } | |
847 | ||
848 | static inline void cgroup_bpf_put(struct cgroup *cgrp) | |
849 | { | |
850 | percpu_ref_put(&cgrp->bpf.refcnt); | |
851 | } | |
852 | ||
853 | #else /* CONFIG_CGROUP_BPF */ | |
854 | ||
855 | static inline void cgroup_bpf_get(struct cgroup *cgrp) {} | |
856 | static inline void cgroup_bpf_put(struct cgroup *cgrp) {} | |
857 | ||
858 | #endif /* CONFIG_CGROUP_BPF */ | |
859 | ||
ddbcc7e8 | 860 | #endif /* _LINUX_CGROUP_H */ |