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 | 13 | #include <linux/nodemask.h> |
c6f53ed8 | 14 | #include <linux/list.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 | 21 | #include <linux/types.h> |
9e8c67a9 | 22 | #include <linux/notifier.h> |
a79a908f AK |
23 | #include <linux/ns_common.h> |
24 | #include <linux/nsproxy.h> | |
25 | #include <linux/user_namespace.h> | |
387ad967 | 26 | #include <linux/refcount.h> |
d2cc5ed6 | 27 | #include <linux/kernel_stat.h> |
ddbcc7e8 | 28 | |
b4a04ab7 | 29 | #include <linux/cgroup-defs.h> |
ddbcc7e8 | 30 | |
ef2c41cf CB |
31 | struct kernel_clone_args; |
32 | ||
6abc8ca1 | 33 | /* |
08b2b6fd | 34 | * All weight knobs on the default hierarchy should use the following min, |
6abc8ca1 TH |
35 | * default and max values. The default value is the logarithmic center of |
36 | * MIN and MAX and allows 100x to be expressed in both directions. | |
37 | */ | |
38 | #define CGROUP_WEIGHT_MIN 1 | |
39 | #define CGROUP_WEIGHT_DFL 100 | |
40 | #define CGROUP_WEIGHT_MAX 10000 | |
41 | ||
4f9c7ca8 TH |
42 | #ifdef CONFIG_CGROUPS |
43 | ||
9e8c67a9 | 44 | enum css_task_iter_flags { |
6da88306 CZ |
45 | CSS_TASK_ITER_PROCS = (1U << 0), /* walk only threadgroup leaders */ |
46 | CSS_TASK_ITER_THREADED = (1U << 1), /* walk all threaded css_sets in the domain */ | |
47 | CSS_TASK_ITER_SKIPPED = (1U << 16), /* internal flags */ | |
48 | }; | |
b636fd38 | 49 | |
c326aa2b TH |
50 | /* a css_task_iter should be treated as an opaque object */ |
51 | struct css_task_iter { | |
52 | struct cgroup_subsys *ss; | |
bc2fb7ed | 53 | unsigned int flags; |
c326aa2b TH |
54 | |
55 | struct list_head *cset_pos; | |
56 | struct list_head *cset_head; | |
ddbcc7e8 | 57 | |
450ee0c1 TH |
58 | struct list_head *tcset_pos; |
59 | struct list_head *tcset_head; | |
60 | ||
c326aa2b | 61 | struct list_head *task_pos; |
ed27b9f7 | 62 | |
9c974c77 | 63 | struct list_head *cur_tasks_head; |
ed27b9f7 | 64 | struct css_set *cur_cset; |
450ee0c1 | 65 | struct css_set *cur_dcset; |
ed27b9f7 TH |
66 | struct task_struct *cur_task; |
67 | struct list_head iters_node; /* css_set->task_iters */ | |
c326aa2b | 68 | }; |
ddbcc7e8 | 69 | |
9e8c67a9 TH |
70 | enum cgroup_lifetime_events { |
71 | CGROUP_LIFETIME_ONLINE, | |
72 | CGROUP_LIFETIME_OFFLINE, | |
73 | }; | |
74 | ||
fbf83212 | 75 | extern struct file_system_type cgroup_fs_type; |
c326aa2b TH |
76 | extern struct cgroup_root cgrp_dfl_root; |
77 | extern struct css_set init_css_set; | |
aecd408b | 78 | extern spinlock_t css_set_lock; |
9e8c67a9 | 79 | extern struct blocking_notifier_head cgroup_lifetime_notifier; |
a424316c | 80 | |
c326aa2b | 81 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys; |
817929ec | 82 | #include <linux/cgroup_subsys.h> |
817929ec PM |
83 | #undef SUBSYS |
84 | ||
49d1dc4b TH |
85 | #define SUBSYS(_x) \ |
86 | extern struct static_key_true _x ## _cgrp_subsys_enabled_key; \ | |
87 | extern struct static_key_true _x ## _cgrp_subsys_on_dfl_key; | |
88 | #include <linux/cgroup_subsys.h> | |
89 | #undef SUBSYS | |
90 | ||
91 | /** | |
92 | * cgroup_subsys_enabled - fast test on whether a subsys is enabled | |
93 | * @ss: subsystem in question | |
94 | */ | |
95 | #define cgroup_subsys_enabled(ss) \ | |
96 | static_branch_likely(&ss ## _enabled_key) | |
97 | ||
98 | /** | |
99 | * cgroup_subsys_on_dfl - fast test on whether a subsys is on default hierarchy | |
100 | * @ss: subsystem in question | |
101 | */ | |
102 | #define cgroup_subsys_on_dfl(ss) \ | |
103 | static_branch_likely(&ss ## _on_dfl_key) | |
104 | ||
c326aa2b TH |
105 | bool css_has_online_children(struct cgroup_subsys_state *css); |
106 | struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss); | |
fc5a828b DZ |
107 | struct cgroup_subsys_state *cgroup_e_css(struct cgroup *cgroup, |
108 | struct cgroup_subsys *ss); | |
c326aa2b TH |
109 | struct cgroup_subsys_state *cgroup_get_e_css(struct cgroup *cgroup, |
110 | struct cgroup_subsys *ss); | |
111 | struct cgroup_subsys_state *css_tryget_online_from_dir(struct dentry *dentry, | |
112 | struct cgroup_subsys *ss); | |
113 | ||
16af4396 | 114 | struct cgroup *cgroup_get_from_path(const char *path); |
1f3fe7eb | 115 | struct cgroup *cgroup_get_from_fd(int fd); |
a6d1ce59 | 116 | struct cgroup *cgroup_v1v2_get_from_fd(int fd); |
16af4396 | 117 | |
c326aa2b TH |
118 | int cgroup_attach_task_all(struct task_struct *from, struct task_struct *); |
119 | int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from); | |
120 | ||
121 | int cgroup_add_dfl_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); | |
122 | int cgroup_add_legacy_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); | |
4a893bdc | 123 | int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); |
c326aa2b | 124 | int cgroup_rm_cftypes(struct cftype *cfts); |
34c06254 | 125 | void cgroup_file_notify(struct cgroup_file *cfile); |
e2691f6b | 126 | void cgroup_file_show(struct cgroup_file *cfile, bool show); |
c326aa2b | 127 | |
c326aa2b TH |
128 | int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry); |
129 | int proc_cgroup_show(struct seq_file *m, struct pid_namespace *ns, | |
130 | struct pid *pid, struct task_struct *tsk); | |
131 | ||
132 | void cgroup_fork(struct task_struct *p); | |
ef2c41cf CB |
133 | extern int cgroup_can_fork(struct task_struct *p, |
134 | struct kernel_clone_args *kargs); | |
135 | extern void cgroup_cancel_fork(struct task_struct *p, | |
136 | struct kernel_clone_args *kargs); | |
137 | extern void cgroup_post_fork(struct task_struct *p, | |
138 | struct kernel_clone_args *kargs); | |
c326aa2b | 139 | void cgroup_exit(struct task_struct *p); |
51bee5ab | 140 | void cgroup_release(struct task_struct *p); |
2e91fa7f | 141 | void cgroup_free(struct task_struct *p); |
c326aa2b TH |
142 | |
143 | int cgroup_init_early(void); | |
144 | int cgroup_init(void); | |
145 | ||
a5e112e6 TH |
146 | int cgroup_parse_float(const char *input, unsigned dec_shift, s64 *v); |
147 | ||
5c9d535b | 148 | /* |
c326aa2b TH |
149 | * Iteration helpers and macros. |
150 | */ | |
151 | ||
152 | struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos, | |
153 | struct cgroup_subsys_state *parent); | |
154 | struct cgroup_subsys_state *css_next_descendant_pre(struct cgroup_subsys_state *pos, | |
155 | struct cgroup_subsys_state *css); | |
156 | struct cgroup_subsys_state *css_rightmost_descendant(struct cgroup_subsys_state *pos); | |
157 | struct cgroup_subsys_state *css_next_descendant_post(struct cgroup_subsys_state *pos, | |
158 | struct cgroup_subsys_state *css); | |
159 | ||
1f7dd3e5 TH |
160 | struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset, |
161 | struct cgroup_subsys_state **dst_cssp); | |
162 | struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset, | |
163 | struct cgroup_subsys_state **dst_cssp); | |
c326aa2b | 164 | |
bc2fb7ed | 165 | void css_task_iter_start(struct cgroup_subsys_state *css, unsigned int flags, |
c326aa2b TH |
166 | struct css_task_iter *it); |
167 | struct task_struct *css_task_iter_next(struct css_task_iter *it); | |
168 | void css_task_iter_end(struct css_task_iter *it); | |
169 | ||
170 | /** | |
171 | * css_for_each_child - iterate through children of a css | |
172 | * @pos: the css * to use as the loop cursor | |
173 | * @parent: css whose children to walk | |
174 | * | |
175 | * Walk @parent's children. Must be called under rcu_read_lock(). | |
176 | * | |
177 | * If a subsystem synchronizes ->css_online() and the start of iteration, a | |
178 | * css which finished ->css_online() is guaranteed to be visible in the | |
179 | * future iterations and will stay visible until the last reference is put. | |
180 | * A css which hasn't finished ->css_online() or already finished | |
181 | * ->css_offline() may show up during traversal. It's each subsystem's | |
182 | * responsibility to synchronize against on/offlining. | |
5c9d535b | 183 | * |
c326aa2b TH |
184 | * It is allowed to temporarily drop RCU read lock during iteration. The |
185 | * caller is responsible for ensuring that @pos remains accessible until | |
186 | * the start of the next iteration by, for example, bumping the css refcnt. | |
5c9d535b | 187 | */ |
c326aa2b TH |
188 | #define css_for_each_child(pos, parent) \ |
189 | for ((pos) = css_next_child(NULL, (parent)); (pos); \ | |
190 | (pos) = css_next_child((pos), (parent))) | |
ddbcc7e8 | 191 | |
c326aa2b TH |
192 | /** |
193 | * css_for_each_descendant_pre - pre-order walk of a css's descendants | |
194 | * @pos: the css * to use as the loop cursor | |
195 | * @root: css whose descendants to walk | |
196 | * | |
197 | * Walk @root's descendants. @root is included in the iteration and the | |
198 | * first node to be visited. Must be called under rcu_read_lock(). | |
199 | * | |
200 | * If a subsystem synchronizes ->css_online() and the start of iteration, a | |
201 | * css which finished ->css_online() is guaranteed to be visible in the | |
202 | * future iterations and will stay visible until the last reference is put. | |
203 | * A css which hasn't finished ->css_online() or already finished | |
204 | * ->css_offline() may show up during traversal. It's each subsystem's | |
205 | * responsibility to synchronize against on/offlining. | |
206 | * | |
207 | * For example, the following guarantees that a descendant can't escape | |
208 | * state updates of its ancestors. | |
209 | * | |
210 | * my_online(@css) | |
211 | * { | |
212 | * Lock @css's parent and @css; | |
213 | * Inherit state from the parent; | |
214 | * Unlock both. | |
215 | * } | |
216 | * | |
217 | * my_update_state(@css) | |
218 | * { | |
219 | * css_for_each_descendant_pre(@pos, @css) { | |
220 | * Lock @pos; | |
221 | * if (@pos == @css) | |
222 | * Update @css's state; | |
223 | * else | |
224 | * Verify @pos is alive and inherit state from its parent; | |
225 | * Unlock @pos; | |
226 | * } | |
227 | * } | |
228 | * | |
229 | * As long as the inheriting step, including checking the parent state, is | |
230 | * enclosed inside @pos locking, double-locking the parent isn't necessary | |
231 | * while inheriting. The state update to the parent is guaranteed to be | |
232 | * visible by walking order and, as long as inheriting operations to the | |
233 | * same @pos are atomic to each other, multiple updates racing each other | |
234 | * still result in the correct state. It's guaranateed that at least one | |
235 | * inheritance happens for any css after the latest update to its parent. | |
236 | * | |
237 | * If checking parent's state requires locking the parent, each inheriting | |
238 | * iteration should lock and unlock both @pos->parent and @pos. | |
239 | * | |
240 | * Alternatively, a subsystem may choose to use a single global lock to | |
241 | * synchronize ->css_online() and ->css_offline() against tree-walking | |
242 | * operations. | |
243 | * | |
244 | * It is allowed to temporarily drop RCU read lock during iteration. The | |
245 | * caller is responsible for ensuring that @pos remains accessible until | |
246 | * the start of the next iteration by, for example, bumping the css refcnt. | |
247 | */ | |
248 | #define css_for_each_descendant_pre(pos, css) \ | |
249 | for ((pos) = css_next_descendant_pre(NULL, (css)); (pos); \ | |
250 | (pos) = css_next_descendant_pre((pos), (css))) | |
251 | ||
252 | /** | |
253 | * css_for_each_descendant_post - post-order walk of a css's descendants | |
254 | * @pos: the css * to use as the loop cursor | |
255 | * @css: css whose descendants to walk | |
256 | * | |
257 | * Similar to css_for_each_descendant_pre() but performs post-order | |
258 | * traversal instead. @root is included in the iteration and the last | |
259 | * node to be visited. | |
260 | * | |
261 | * If a subsystem synchronizes ->css_online() and the start of iteration, a | |
262 | * css which finished ->css_online() is guaranteed to be visible in the | |
263 | * future iterations and will stay visible until the last reference is put. | |
264 | * A css which hasn't finished ->css_online() or already finished | |
265 | * ->css_offline() may show up during traversal. It's each subsystem's | |
266 | * responsibility to synchronize against on/offlining. | |
267 | * | |
268 | * Note that the walk visibility guarantee example described in pre-order | |
269 | * walk doesn't apply the same to post-order walks. | |
270 | */ | |
271 | #define css_for_each_descendant_post(pos, css) \ | |
272 | for ((pos) = css_next_descendant_post(NULL, (css)); (pos); \ | |
273 | (pos) = css_next_descendant_post((pos), (css))) | |
274 | ||
275 | /** | |
276 | * cgroup_taskset_for_each - iterate cgroup_taskset | |
277 | * @task: the loop cursor | |
1f7dd3e5 | 278 | * @dst_css: the destination css |
c326aa2b | 279 | * @tset: taskset to iterate |
4530eddb TH |
280 | * |
281 | * @tset may contain multiple tasks and they may belong to multiple | |
1f7dd3e5 TH |
282 | * processes. |
283 | * | |
284 | * On the v2 hierarchy, there may be tasks from multiple processes and they | |
285 | * may not share the source or destination csses. | |
286 | * | |
287 | * On traditional hierarchies, when there are multiple tasks in @tset, if a | |
288 | * task of a process is in @tset, all tasks of the process are in @tset. | |
289 | * Also, all are guaranteed to share the same source and destination csses. | |
4530eddb TH |
290 | * |
291 | * Iteration is not in any specific order. | |
c326aa2b | 292 | */ |
1f7dd3e5 TH |
293 | #define cgroup_taskset_for_each(task, dst_css, tset) \ |
294 | for ((task) = cgroup_taskset_first((tset), &(dst_css)); \ | |
295 | (task); \ | |
296 | (task) = cgroup_taskset_next((tset), &(dst_css))) | |
ddbcc7e8 | 297 | |
4530eddb TH |
298 | /** |
299 | * cgroup_taskset_for_each_leader - iterate group leaders in a cgroup_taskset | |
300 | * @leader: the loop cursor | |
1f7dd3e5 | 301 | * @dst_css: the destination css |
7b4632f0 | 302 | * @tset: taskset to iterate |
4530eddb TH |
303 | * |
304 | * Iterate threadgroup leaders of @tset. For single-task migrations, @tset | |
305 | * may not contain any. | |
306 | */ | |
1f7dd3e5 TH |
307 | #define cgroup_taskset_for_each_leader(leader, dst_css, tset) \ |
308 | for ((leader) = cgroup_taskset_first((tset), &(dst_css)); \ | |
309 | (leader); \ | |
310 | (leader) = cgroup_taskset_next((tset), &(dst_css))) \ | |
4530eddb TH |
311 | if ((leader) != (leader)->group_leader) \ |
312 | ; \ | |
313 | else | |
314 | ||
c326aa2b TH |
315 | /* |
316 | * Inline functions. | |
317 | */ | |
ddbcc7e8 | 318 | |
6ab42860 TH |
319 | #ifdef CONFIG_DEBUG_CGROUP_REF |
320 | void css_get(struct cgroup_subsys_state *css); | |
321 | void css_get_many(struct cgroup_subsys_state *css, unsigned int n); | |
322 | bool css_tryget(struct cgroup_subsys_state *css); | |
323 | bool css_tryget_online(struct cgroup_subsys_state *css); | |
324 | void css_put(struct cgroup_subsys_state *css); | |
325 | void css_put_many(struct cgroup_subsys_state *css, unsigned int n); | |
326 | #else | |
327 | #define CGROUP_REF_FN_ATTRS static inline | |
79a7f41f | 328 | #define CGROUP_REF_EXPORT(fn) |
6ab42860 TH |
329 | #include <linux/cgroup_refcnt.h> |
330 | #endif | |
331 | ||
a3b89864 | 332 | static inline u64 cgroup_id(const struct cgroup *cgrp) |
74321038 TH |
333 | { |
334 | return cgrp->kn->id; | |
335 | } | |
336 | ||
41c25707 TH |
337 | /** |
338 | * css_is_dying - test whether the specified css is dying | |
339 | * @css: target css | |
340 | * | |
341 | * Test whether @css is in the process of offlining or already offline. In | |
342 | * most cases, ->css_online() and ->css_offline() callbacks should be | |
343 | * enough; however, the actual offline operations are RCU delayed and this | |
344 | * test returns %true also when @css is scheduled to be offlined. | |
345 | * | |
346 | * This is useful, for example, when the use case requires synchronous | |
347 | * behavior with respect to cgroup removal. cgroup removal schedules css | |
348 | * offlining but the css can seem alive while the operation is being | |
349 | * delayed. If the delay affects user visible semantics, this test can be | |
350 | * used to resolve the situation. | |
351 | */ | |
352 | static inline bool css_is_dying(struct cgroup_subsys_state *css) | |
353 | { | |
a22b3d54 | 354 | return css->flags & CSS_DYING; |
41c25707 TH |
355 | } |
356 | ||
541a4219 | 357 | static inline bool css_is_self(struct cgroup_subsys_state *css) |
845a7245 | 358 | { |
541a4219 JK |
359 | if (css == &css->cgroup->self) { |
360 | /* cgroup::self should not have subsystem association */ | |
361 | WARN_ON(css->ss != NULL); | |
362 | return true; | |
363 | } | |
364 | ||
365 | return false; | |
845a7245 JK |
366 | } |
367 | ||
3e48930c TH |
368 | static inline void cgroup_get(struct cgroup *cgrp) |
369 | { | |
370 | css_get(&cgrp->self); | |
371 | } | |
372 | ||
373 | static inline bool cgroup_tryget(struct cgroup *cgrp) | |
374 | { | |
375 | return css_tryget(&cgrp->self); | |
376 | } | |
377 | ||
16af4396 TH |
378 | static inline void cgroup_put(struct cgroup *cgrp) |
379 | { | |
380 | css_put(&cgrp->self); | |
381 | } | |
382 | ||
354ed597 YZ |
383 | extern struct mutex cgroup_mutex; |
384 | ||
385 | static inline void cgroup_lock(void) | |
386 | { | |
387 | mutex_lock(&cgroup_mutex); | |
388 | } | |
389 | ||
390 | static inline void cgroup_unlock(void) | |
391 | { | |
392 | mutex_unlock(&cgroup_mutex); | |
393 | } | |
394 | ||
14611e51 TH |
395 | /** |
396 | * task_css_set_check - obtain a task's css_set with extra access conditions | |
397 | * @task: the task to obtain css_set for | |
398 | * @__c: extra condition expression to be passed to rcu_dereference_check() | |
399 | * | |
400 | * A task's css_set is RCU protected, initialized and exited while holding | |
401 | * task_lock(), and can only be modified while holding both cgroup_mutex | |
402 | * and task_lock() while the task is alive. This macro verifies that the | |
403 | * caller is inside proper critical section and returns @task's css_set. | |
404 | * | |
405 | * The caller can also specify additional allowed conditions via @__c, such | |
406 | * as locks used during the cgroup_subsys::attach() methods. | |
dc61b1d6 | 407 | */ |
2219449a | 408 | #ifdef CONFIG_PROVE_RCU |
14611e51 TH |
409 | #define task_css_set_check(task, __c) \ |
410 | rcu_dereference_check((task)->cgroups, \ | |
f2aa197e | 411 | rcu_read_lock_sched_held() || \ |
0e1d768f | 412 | lockdep_is_held(&cgroup_mutex) || \ |
f0d9a5f1 | 413 | lockdep_is_held(&css_set_lock) || \ |
0e1d768f | 414 | ((task)->flags & PF_EXITING) || (__c)) |
2219449a | 415 | #else |
14611e51 TH |
416 | #define task_css_set_check(task, __c) \ |
417 | rcu_dereference((task)->cgroups) | |
2219449a | 418 | #endif |
dc61b1d6 | 419 | |
14611e51 | 420 | /** |
8af01f56 | 421 | * task_css_check - obtain css for (task, subsys) w/ extra access conds |
14611e51 TH |
422 | * @task: the target task |
423 | * @subsys_id: the target subsystem ID | |
424 | * @__c: extra condition expression to be passed to rcu_dereference_check() | |
425 | * | |
426 | * Return the cgroup_subsys_state for the (@task, @subsys_id) pair. The | |
427 | * synchronization rules are the same as task_css_set_check(). | |
428 | */ | |
8af01f56 | 429 | #define task_css_check(task, subsys_id, __c) \ |
14611e51 TH |
430 | task_css_set_check((task), (__c))->subsys[(subsys_id)] |
431 | ||
432 | /** | |
433 | * task_css_set - obtain a task's css_set | |
434 | * @task: the task to obtain css_set for | |
435 | * | |
436 | * See task_css_set_check(). | |
437 | */ | |
438 | static inline struct css_set *task_css_set(struct task_struct *task) | |
439 | { | |
440 | return task_css_set_check(task, false); | |
441 | } | |
442 | ||
443 | /** | |
8af01f56 | 444 | * task_css - obtain css for (task, subsys) |
14611e51 TH |
445 | * @task: the target task |
446 | * @subsys_id: the target subsystem ID | |
447 | * | |
8af01f56 | 448 | * See task_css_check(). |
14611e51 | 449 | */ |
8af01f56 TH |
450 | static inline struct cgroup_subsys_state *task_css(struct task_struct *task, |
451 | int subsys_id) | |
ddbcc7e8 | 452 | { |
8af01f56 | 453 | return task_css_check(task, subsys_id, false); |
ddbcc7e8 PM |
454 | } |
455 | ||
ec438699 TH |
456 | /** |
457 | * task_get_css - find and get the css for (task, subsys) | |
458 | * @task: the target task | |
459 | * @subsys_id: the target subsystem ID | |
460 | * | |
461 | * Find the css for the (@task, @subsys_id) combination, increment a | |
462 | * reference on and return it. This function is guaranteed to return a | |
18fa84a2 | 463 | * valid css. The returned css may already have been offlined. |
ec438699 TH |
464 | */ |
465 | static inline struct cgroup_subsys_state * | |
466 | task_get_css(struct task_struct *task, int subsys_id) | |
467 | { | |
468 | struct cgroup_subsys_state *css; | |
469 | ||
470 | rcu_read_lock(); | |
471 | while (true) { | |
472 | css = task_css(task, subsys_id); | |
18fa84a2 TH |
473 | /* |
474 | * Can't use css_tryget_online() here. A task which has | |
475 | * PF_EXITING set may stay associated with an offline css. | |
476 | * If such task calls this function, css_tryget_online() | |
477 | * will keep failing. | |
478 | */ | |
479 | if (likely(css_tryget(css))) | |
ec438699 TH |
480 | break; |
481 | cpu_relax(); | |
482 | } | |
483 | rcu_read_unlock(); | |
484 | return css; | |
485 | } | |
486 | ||
5024ae29 TH |
487 | /** |
488 | * task_css_is_root - test whether a task belongs to the root css | |
489 | * @task: the target task | |
490 | * @subsys_id: the target subsystem ID | |
491 | * | |
492 | * Test whether @task belongs to the root css on the specified subsystem. | |
493 | * May be invoked in any context. | |
494 | */ | |
495 | static inline bool task_css_is_root(struct task_struct *task, int subsys_id) | |
496 | { | |
497 | return task_css_check(task, subsys_id, true) == | |
498 | init_css_set.subsys[subsys_id]; | |
499 | } | |
500 | ||
8af01f56 TH |
501 | static inline struct cgroup *task_cgroup(struct task_struct *task, |
502 | int subsys_id) | |
ddbcc7e8 | 503 | { |
8af01f56 | 504 | return task_css(task, subsys_id)->cgroup; |
ddbcc7e8 PM |
505 | } |
506 | ||
3e48930c TH |
507 | static inline struct cgroup *task_dfl_cgroup(struct task_struct *task) |
508 | { | |
509 | return task_css_set(task)->dfl_cgrp; | |
510 | } | |
511 | ||
512 | static inline struct cgroup *cgroup_parent(struct cgroup *cgrp) | |
513 | { | |
514 | struct cgroup_subsys_state *parent_css = cgrp->self.parent; | |
515 | ||
516 | if (parent_css) | |
517 | return container_of(parent_css, struct cgroup, self); | |
518 | return NULL; | |
519 | } | |
520 | ||
b11cfb58 TH |
521 | /** |
522 | * cgroup_is_descendant - test ancestry | |
523 | * @cgrp: the cgroup to be tested | |
524 | * @ancestor: possible ancestor of @cgrp | |
525 | * | |
526 | * Test whether @cgrp is a descendant of @ancestor. It also returns %true | |
527 | * if @cgrp == @ancestor. This function is safe to call as long as @cgrp | |
528 | * and @ancestor are accessible. | |
529 | */ | |
530 | static inline bool cgroup_is_descendant(struct cgroup *cgrp, | |
531 | struct cgroup *ancestor) | |
532 | { | |
533 | if (cgrp->root != ancestor->root || cgrp->level < ancestor->level) | |
534 | return false; | |
7f203bc8 | 535 | return cgrp->ancestors[ancestor->level] == ancestor; |
b11cfb58 TH |
536 | } |
537 | ||
77236281 AI |
538 | /** |
539 | * cgroup_ancestor - find ancestor of cgroup | |
540 | * @cgrp: cgroup to find ancestor of | |
541 | * @ancestor_level: level of ancestor to find starting from root | |
542 | * | |
543 | * Find ancestor of cgroup at specified level starting from root if it exists | |
544 | * and return pointer to it. Return NULL if @cgrp doesn't have ancestor at | |
545 | * @ancestor_level. | |
546 | * | |
547 | * This function is safe to call as long as @cgrp is accessible. | |
548 | */ | |
549 | static inline struct cgroup *cgroup_ancestor(struct cgroup *cgrp, | |
550 | int ancestor_level) | |
551 | { | |
7f203bc8 | 552 | if (ancestor_level < 0 || ancestor_level > cgrp->level) |
77236281 | 553 | return NULL; |
7f203bc8 | 554 | return cgrp->ancestors[ancestor_level]; |
77236281 AI |
555 | } |
556 | ||
aed704b7 SD |
557 | /** |
558 | * task_under_cgroup_hierarchy - test task's membership of cgroup ancestry | |
559 | * @task: the task to be tested | |
560 | * @ancestor: possible ancestor of @task's cgroup | |
561 | * | |
562 | * Tests whether @task's default cgroup hierarchy is a descendant of @ancestor. | |
563 | * It follows all the same rules as cgroup_is_descendant, and only applies | |
564 | * to the default hierarchy. | |
565 | */ | |
566 | static inline bool task_under_cgroup_hierarchy(struct task_struct *task, | |
567 | struct cgroup *ancestor) | |
568 | { | |
569 | struct css_set *cset = task_css_set(task); | |
570 | ||
571 | return cgroup_is_descendant(cset->dfl_cgrp, ancestor); | |
572 | } | |
573 | ||
07bc356e | 574 | /* no synchronization, the result can only be used as a hint */ |
27bd4dbb | 575 | static inline bool cgroup_is_populated(struct cgroup *cgrp) |
07bc356e | 576 | { |
454000ad TH |
577 | return cgrp->nr_populated_csets + cgrp->nr_populated_domain_children + |
578 | cgrp->nr_populated_threaded_children; | |
07bc356e | 579 | } |
f3d46500 | 580 | |
f29374b1 | 581 | /* returns ino associated with a cgroup */ |
b1664924 TH |
582 | static inline ino_t cgroup_ino(struct cgroup *cgrp) |
583 | { | |
67c0496e | 584 | return kernfs_ino(cgrp->kn); |
b1664924 | 585 | } |
3ebb2b6e | 586 | |
b4168640 TH |
587 | /* cft/css accessors for cftype->write() operation */ |
588 | static inline struct cftype *of_cft(struct kernfs_open_file *of) | |
7da11279 | 589 | { |
2bd59d48 | 590 | return of->kn->priv; |
7da11279 | 591 | } |
0f0a2b4f | 592 | |
b4168640 | 593 | struct cgroup_subsys_state *of_css(struct kernfs_open_file *of); |
817929ec | 594 | |
b4168640 TH |
595 | /* cft/css accessors for cftype->seq_*() operations */ |
596 | static inline struct cftype *seq_cft(struct seq_file *seq) | |
597 | { | |
598 | return of_cft(seq->private); | |
599 | } | |
e535837b | 600 | |
b4168640 TH |
601 | static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq) |
602 | { | |
603 | return of_css(seq->private); | |
604 | } | |
31583bb0 | 605 | |
e61734c5 TH |
606 | /* |
607 | * Name / path handling functions. All are thin wrappers around the kernfs | |
608 | * counterparts and can be called under any context. | |
609 | */ | |
38460b48 | 610 | |
e61734c5 TH |
611 | static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen) |
612 | { | |
fdce6bf8 | 613 | return kernfs_name(cgrp->kn, buf, buflen); |
e61734c5 | 614 | } |
ddbcc7e8 | 615 | |
4c737b41 | 616 | static inline int cgroup_path(struct cgroup *cgrp, char *buf, size_t buflen) |
e61734c5 | 617 | { |
fdce6bf8 | 618 | return kernfs_path(cgrp->kn, buf, buflen); |
e61734c5 | 619 | } |
f3ba5380 | 620 | |
e61734c5 TH |
621 | static inline void pr_cont_cgroup_name(struct cgroup *cgrp) |
622 | { | |
fdce6bf8 | 623 | pr_cont_kernfs_name(cgrp->kn); |
e61734c5 | 624 | } |
ddbcc7e8 | 625 | |
e61734c5 | 626 | static inline void pr_cont_cgroup_path(struct cgroup *cgrp) |
846c7bb0 | 627 | { |
fdce6bf8 | 628 | pr_cont_kernfs_path(cgrp->kn); |
846c7bb0 | 629 | } |
ddbcc7e8 | 630 | |
3958e2d0 SB |
631 | bool cgroup_psi_enabled(void); |
632 | ||
77f88796 TH |
633 | static inline void cgroup_init_kthreadd(void) |
634 | { | |
635 | /* | |
636 | * kthreadd is inherited by all kthreads, keep it in the root so | |
637 | * that the new kthreads are guaranteed to stay in the root until | |
638 | * initialization is finished. | |
639 | */ | |
640 | current->no_cgroup_migration = 1; | |
641 | } | |
642 | ||
643 | static inline void cgroup_kthread_ready(void) | |
644 | { | |
645 | /* | |
646 | * This kthread finished initialization. The creator should have | |
647 | * set PF_NO_SETAFFINITY if this kthread should stay in the root. | |
648 | */ | |
649 | current->no_cgroup_migration = 0; | |
650 | } | |
651 | ||
67c0496e | 652 | void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen); |
6b658c48 | 653 | struct cgroup *cgroup_get_from_id(u64 id); |
ddbcc7e8 PM |
654 | #else /* !CONFIG_CGROUPS */ |
655 | ||
f3ba5380 | 656 | struct cgroup_subsys_state; |
aed704b7 | 657 | struct cgroup; |
f3ba5380 | 658 | |
a3b89864 | 659 | static inline u64 cgroup_id(const struct cgroup *cgrp) { return 1; } |
9b0eb69b | 660 | static inline void css_get(struct cgroup_subsys_state *css) {} |
c326aa2b | 661 | static inline void css_put(struct cgroup_subsys_state *css) {} |
354ed597 YZ |
662 | static inline void cgroup_lock(void) {} |
663 | static inline void cgroup_unlock(void) {} | |
31583bb0 | 664 | static inline int cgroup_attach_task_all(struct task_struct *from, |
c326aa2b TH |
665 | struct task_struct *t) { return 0; } |
666 | static inline int cgroupstats_build(struct cgroupstats *stats, | |
667 | struct dentry *dentry) { return -EINVAL; } | |
668 | ||
b4f48b63 | 669 | static inline void cgroup_fork(struct task_struct *p) {} |
ef2c41cf CB |
670 | static inline int cgroup_can_fork(struct task_struct *p, |
671 | struct kernel_clone_args *kargs) { return 0; } | |
672 | static inline void cgroup_cancel_fork(struct task_struct *p, | |
673 | struct kernel_clone_args *kargs) {} | |
674 | static inline void cgroup_post_fork(struct task_struct *p, | |
675 | struct kernel_clone_args *kargs) {} | |
1ec41830 | 676 | static inline void cgroup_exit(struct task_struct *p) {} |
51bee5ab | 677 | static inline void cgroup_release(struct task_struct *p) {} |
2e91fa7f | 678 | static inline void cgroup_free(struct task_struct *p) {} |
ddbcc7e8 | 679 | |
c326aa2b TH |
680 | static inline int cgroup_init_early(void) { return 0; } |
681 | static inline int cgroup_init(void) { return 0; } | |
77f88796 TH |
682 | static inline void cgroup_init_kthreadd(void) {} |
683 | static inline void cgroup_kthread_ready(void) {} | |
d7926ee3 | 684 | |
2ce7135a JW |
685 | static inline struct cgroup *cgroup_parent(struct cgroup *cgrp) |
686 | { | |
687 | return NULL; | |
688 | } | |
689 | ||
3958e2d0 SB |
690 | static inline bool cgroup_psi_enabled(void) |
691 | { | |
692 | return false; | |
693 | } | |
694 | ||
aed704b7 SD |
695 | static inline bool task_under_cgroup_hierarchy(struct task_struct *task, |
696 | struct cgroup *ancestor) | |
697 | { | |
698 | return true; | |
699 | } | |
69fd5c39 | 700 | |
67c0496e TH |
701 | static inline void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen) |
702 | {} | |
ddbcc7e8 PM |
703 | #endif /* !CONFIG_CGROUPS */ |
704 | ||
6162cef0 | 705 | #ifdef CONFIG_CGROUPS |
d2cc5ed6 | 706 | /* |
6162cef0 | 707 | * cgroup scalable recursive statistics. |
d2cc5ed6 | 708 | */ |
a9791555 JK |
709 | void css_rstat_updated(struct cgroup_subsys_state *css, int cpu); |
710 | void css_rstat_flush(struct cgroup_subsys_state *css); | |
d2cc5ed6 | 711 | |
6162cef0 TH |
712 | /* |
713 | * Basic resource stats. | |
714 | */ | |
d2cc5ed6 TH |
715 | #ifdef CONFIG_CGROUP_CPUACCT |
716 | void cpuacct_charge(struct task_struct *tsk, u64 cputime); | |
717 | void cpuacct_account_field(struct task_struct *tsk, int index, u64 val); | |
718 | #else | |
719 | static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {} | |
720 | static inline void cpuacct_account_field(struct task_struct *tsk, int index, | |
721 | u64 val) {} | |
722 | #endif | |
723 | ||
041cd640 TH |
724 | void __cgroup_account_cputime(struct cgroup *cgrp, u64 delta_exec); |
725 | void __cgroup_account_cputime_field(struct cgroup *cgrp, | |
726 | enum cpu_usage_stat index, u64 delta_exec); | |
727 | ||
d2cc5ed6 TH |
728 | static inline void cgroup_account_cputime(struct task_struct *task, |
729 | u64 delta_exec) | |
730 | { | |
041cd640 TH |
731 | struct cgroup *cgrp; |
732 | ||
d2cc5ed6 | 733 | cpuacct_charge(task, delta_exec); |
041cd640 | 734 | |
041cd640 TH |
735 | cgrp = task_dfl_cgroup(task); |
736 | if (cgroup_parent(cgrp)) | |
737 | __cgroup_account_cputime(cgrp, delta_exec); | |
d2cc5ed6 TH |
738 | } |
739 | ||
740 | static inline void cgroup_account_cputime_field(struct task_struct *task, | |
741 | enum cpu_usage_stat index, | |
742 | u64 delta_exec) | |
743 | { | |
041cd640 TH |
744 | struct cgroup *cgrp; |
745 | ||
d2cc5ed6 | 746 | cpuacct_account_field(task, index, delta_exec); |
041cd640 | 747 | |
041cd640 TH |
748 | cgrp = task_dfl_cgroup(task); |
749 | if (cgroup_parent(cgrp)) | |
750 | __cgroup_account_cputime_field(cgrp, index, delta_exec); | |
d2cc5ed6 TH |
751 | } |
752 | ||
753 | #else /* CONFIG_CGROUPS */ | |
754 | ||
755 | static inline void cgroup_account_cputime(struct task_struct *task, | |
756 | u64 delta_exec) {} | |
757 | static inline void cgroup_account_cputime_field(struct task_struct *task, | |
758 | enum cpu_usage_stat index, | |
759 | u64 delta_exec) {} | |
760 | ||
761 | #endif /* CONFIG_CGROUPS */ | |
762 | ||
bd1060a1 TH |
763 | /* |
764 | * sock->sk_cgrp_data handling. For more info, see sock_cgroup_data | |
765 | * definition in cgroup-defs.h. | |
766 | */ | |
767 | #ifdef CONFIG_SOCK_CGROUP_DATA | |
768 | ||
bd1060a1 | 769 | void cgroup_sk_alloc(struct sock_cgroup_data *skcd); |
ad0f75e5 | 770 | void cgroup_sk_clone(struct sock_cgroup_data *skcd); |
bd1060a1 TH |
771 | void cgroup_sk_free(struct sock_cgroup_data *skcd); |
772 | ||
773 | static inline struct cgroup *sock_cgroup_ptr(struct sock_cgroup_data *skcd) | |
774 | { | |
8520e224 | 775 | return skcd->cgroup; |
bd1060a1 TH |
776 | } |
777 | ||
778 | #else /* CONFIG_CGROUP_DATA */ | |
779 | ||
780 | static inline void cgroup_sk_alloc(struct sock_cgroup_data *skcd) {} | |
ad0f75e5 | 781 | static inline void cgroup_sk_clone(struct sock_cgroup_data *skcd) {} |
bd1060a1 TH |
782 | static inline void cgroup_sk_free(struct sock_cgroup_data *skcd) {} |
783 | ||
784 | #endif /* CONFIG_CGROUP_DATA */ | |
785 | ||
a79a908f | 786 | struct cgroup_namespace { |
a79a908f AK |
787 | struct ns_common ns; |
788 | struct user_namespace *user_ns; | |
d08311dd | 789 | struct ucounts *ucounts; |
a79a908f AK |
790 | struct css_set *root_cset; |
791 | }; | |
792 | ||
793 | extern struct cgroup_namespace init_cgroup_ns; | |
794 | ||
795 | #ifdef CONFIG_CGROUPS | |
796 | ||
797 | void free_cgroup_ns(struct cgroup_namespace *ns); | |
798 | ||
799 | struct cgroup_namespace *copy_cgroup_ns(unsigned long flags, | |
800 | struct user_namespace *user_ns, | |
801 | struct cgroup_namespace *old_ns); | |
802 | ||
4c737b41 TH |
803 | int cgroup_path_ns(struct cgroup *cgrp, char *buf, size_t buflen, |
804 | struct cgroup_namespace *ns); | |
a79a908f | 805 | |
79fb8d8d JS |
806 | static inline void get_cgroup_ns(struct cgroup_namespace *ns) |
807 | { | |
808 | refcount_inc(&ns->ns.count); | |
809 | } | |
810 | ||
811 | static inline void put_cgroup_ns(struct cgroup_namespace *ns) | |
812 | { | |
813 | if (refcount_dec_and_test(&ns->ns.count)) | |
814 | free_cgroup_ns(ns); | |
815 | } | |
816 | ||
a79a908f AK |
817 | #else /* !CONFIG_CGROUPS */ |
818 | ||
819 | static inline void free_cgroup_ns(struct cgroup_namespace *ns) { } | |
820 | static inline struct cgroup_namespace * | |
821 | copy_cgroup_ns(unsigned long flags, struct user_namespace *user_ns, | |
822 | struct cgroup_namespace *old_ns) | |
823 | { | |
824 | return old_ns; | |
825 | } | |
826 | ||
79fb8d8d JS |
827 | static inline void get_cgroup_ns(struct cgroup_namespace *ns) { } |
828 | static inline void put_cgroup_ns(struct cgroup_namespace *ns) { } | |
a79a908f | 829 | |
79fb8d8d | 830 | #endif /* !CONFIG_CGROUPS */ |
a79a908f | 831 | |
76f969e8 RG |
832 | #ifdef CONFIG_CGROUPS |
833 | ||
834 | void cgroup_enter_frozen(void); | |
835 | void cgroup_leave_frozen(bool always_leave); | |
836 | void cgroup_update_frozen(struct cgroup *cgrp); | |
837 | void cgroup_freeze(struct cgroup *cgrp, bool freeze); | |
838 | void cgroup_freezer_migrate_task(struct task_struct *task, struct cgroup *src, | |
839 | struct cgroup *dst); | |
96b9c592 | 840 | |
76f969e8 RG |
841 | static inline bool cgroup_task_frozen(struct task_struct *task) |
842 | { | |
843 | return task->frozen; | |
844 | } | |
845 | ||
846 | #else /* !CONFIG_CGROUPS */ | |
847 | ||
848 | static inline void cgroup_enter_frozen(void) { } | |
849 | static inline void cgroup_leave_frozen(bool always_leave) { } | |
76f969e8 RG |
850 | static inline bool cgroup_task_frozen(struct task_struct *task) |
851 | { | |
852 | return false; | |
853 | } | |
854 | ||
855 | #endif /* !CONFIG_CGROUPS */ | |
856 | ||
4bfc0bb2 RG |
857 | #ifdef CONFIG_CGROUP_BPF |
858 | static inline void cgroup_bpf_get(struct cgroup *cgrp) | |
859 | { | |
860 | percpu_ref_get(&cgrp->bpf.refcnt); | |
861 | } | |
862 | ||
863 | static inline void cgroup_bpf_put(struct cgroup *cgrp) | |
864 | { | |
865 | percpu_ref_put(&cgrp->bpf.refcnt); | |
866 | } | |
867 | ||
868 | #else /* CONFIG_CGROUP_BPF */ | |
869 | ||
870 | static inline void cgroup_bpf_get(struct cgroup *cgrp) {} | |
871 | static inline void cgroup_bpf_put(struct cgroup *cgrp) {} | |
872 | ||
873 | #endif /* CONFIG_CGROUP_BPF */ | |
874 | ||
aecd408b YS |
875 | struct cgroup *task_get_cgroup1(struct task_struct *tsk, int hierarchy_id); |
876 | ||
c6f53ed8 DF |
877 | struct cgroup_of_peak *of_peak(struct kernfs_open_file *of); |
878 | ||
ddbcc7e8 | 879 | #endif /* _LINUX_CGROUP_H */ |