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 TH |
34 | /* |
35 | * All weight knobs on the default hierarhcy should use the following min, | |
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 | |
c326aa2b TH |
71 | extern struct cgroup_root cgrp_dfl_root; |
72 | extern struct css_set init_css_set; | |
a424316c | 73 | |
c326aa2b | 74 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys; |
817929ec | 75 | #include <linux/cgroup_subsys.h> |
817929ec PM |
76 | #undef SUBSYS |
77 | ||
49d1dc4b TH |
78 | #define SUBSYS(_x) \ |
79 | extern struct static_key_true _x ## _cgrp_subsys_enabled_key; \ | |
80 | extern struct static_key_true _x ## _cgrp_subsys_on_dfl_key; | |
81 | #include <linux/cgroup_subsys.h> | |
82 | #undef SUBSYS | |
83 | ||
84 | /** | |
85 | * cgroup_subsys_enabled - fast test on whether a subsys is enabled | |
86 | * @ss: subsystem in question | |
87 | */ | |
88 | #define cgroup_subsys_enabled(ss) \ | |
89 | static_branch_likely(&ss ## _enabled_key) | |
90 | ||
91 | /** | |
92 | * cgroup_subsys_on_dfl - fast test on whether a subsys is on default hierarchy | |
93 | * @ss: subsystem in question | |
94 | */ | |
95 | #define cgroup_subsys_on_dfl(ss) \ | |
96 | static_branch_likely(&ss ## _on_dfl_key) | |
97 | ||
c326aa2b TH |
98 | bool css_has_online_children(struct cgroup_subsys_state *css); |
99 | struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss); | |
fc5a828b DZ |
100 | struct cgroup_subsys_state *cgroup_e_css(struct cgroup *cgroup, |
101 | struct cgroup_subsys *ss); | |
c326aa2b TH |
102 | struct cgroup_subsys_state *cgroup_get_e_css(struct cgroup *cgroup, |
103 | struct cgroup_subsys *ss); | |
104 | struct cgroup_subsys_state *css_tryget_online_from_dir(struct dentry *dentry, | |
105 | struct cgroup_subsys *ss); | |
106 | ||
16af4396 | 107 | struct cgroup *cgroup_get_from_path(const char *path); |
1f3fe7eb | 108 | struct cgroup *cgroup_get_from_fd(int fd); |
16af4396 | 109 | |
c326aa2b TH |
110 | int cgroup_attach_task_all(struct task_struct *from, struct task_struct *); |
111 | int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from); | |
112 | ||
113 | int cgroup_add_dfl_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); | |
114 | int cgroup_add_legacy_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); | |
115 | int cgroup_rm_cftypes(struct cftype *cfts); | |
34c06254 | 116 | void cgroup_file_notify(struct cgroup_file *cfile); |
c326aa2b | 117 | |
4c737b41 | 118 | int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen); |
c326aa2b TH |
119 | int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry); |
120 | int proc_cgroup_show(struct seq_file *m, struct pid_namespace *ns, | |
121 | struct pid *pid, struct task_struct *tsk); | |
122 | ||
123 | void cgroup_fork(struct task_struct *p); | |
ef2c41cf CB |
124 | extern int cgroup_can_fork(struct task_struct *p, |
125 | struct kernel_clone_args *kargs); | |
126 | extern void cgroup_cancel_fork(struct task_struct *p, | |
127 | struct kernel_clone_args *kargs); | |
128 | extern void cgroup_post_fork(struct task_struct *p, | |
129 | struct kernel_clone_args *kargs); | |
c326aa2b | 130 | void cgroup_exit(struct task_struct *p); |
51bee5ab | 131 | void cgroup_release(struct task_struct *p); |
2e91fa7f | 132 | void cgroup_free(struct task_struct *p); |
c326aa2b TH |
133 | |
134 | int cgroup_init_early(void); | |
135 | int cgroup_init(void); | |
136 | ||
a5e112e6 TH |
137 | int cgroup_parse_float(const char *input, unsigned dec_shift, s64 *v); |
138 | ||
5c9d535b | 139 | /* |
c326aa2b TH |
140 | * Iteration helpers and macros. |
141 | */ | |
142 | ||
143 | struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos, | |
144 | struct cgroup_subsys_state *parent); | |
145 | struct cgroup_subsys_state *css_next_descendant_pre(struct cgroup_subsys_state *pos, | |
146 | struct cgroup_subsys_state *css); | |
147 | struct cgroup_subsys_state *css_rightmost_descendant(struct cgroup_subsys_state *pos); | |
148 | struct cgroup_subsys_state *css_next_descendant_post(struct cgroup_subsys_state *pos, | |
149 | struct cgroup_subsys_state *css); | |
150 | ||
1f7dd3e5 TH |
151 | struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset, |
152 | struct cgroup_subsys_state **dst_cssp); | |
153 | struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset, | |
154 | struct cgroup_subsys_state **dst_cssp); | |
c326aa2b | 155 | |
bc2fb7ed | 156 | void css_task_iter_start(struct cgroup_subsys_state *css, unsigned int flags, |
c326aa2b TH |
157 | struct css_task_iter *it); |
158 | struct task_struct *css_task_iter_next(struct css_task_iter *it); | |
159 | void css_task_iter_end(struct css_task_iter *it); | |
160 | ||
161 | /** | |
162 | * css_for_each_child - iterate through children of a css | |
163 | * @pos: the css * to use as the loop cursor | |
164 | * @parent: css whose children to walk | |
165 | * | |
166 | * Walk @parent's children. Must be called under rcu_read_lock(). | |
167 | * | |
168 | * If a subsystem synchronizes ->css_online() and the start of iteration, a | |
169 | * css which finished ->css_online() is guaranteed to be visible in the | |
170 | * future iterations and will stay visible until the last reference is put. | |
171 | * A css which hasn't finished ->css_online() or already finished | |
172 | * ->css_offline() may show up during traversal. It's each subsystem's | |
173 | * responsibility to synchronize against on/offlining. | |
5c9d535b | 174 | * |
c326aa2b TH |
175 | * It is allowed to temporarily drop RCU read lock during iteration. The |
176 | * caller is responsible for ensuring that @pos remains accessible until | |
177 | * the start of the next iteration by, for example, bumping the css refcnt. | |
5c9d535b | 178 | */ |
c326aa2b TH |
179 | #define css_for_each_child(pos, parent) \ |
180 | for ((pos) = css_next_child(NULL, (parent)); (pos); \ | |
181 | (pos) = css_next_child((pos), (parent))) | |
ddbcc7e8 | 182 | |
c326aa2b TH |
183 | /** |
184 | * css_for_each_descendant_pre - pre-order walk of a css's descendants | |
185 | * @pos: the css * to use as the loop cursor | |
186 | * @root: css whose descendants to walk | |
187 | * | |
188 | * Walk @root's descendants. @root is included in the iteration and the | |
189 | * first node to be visited. Must be called under rcu_read_lock(). | |
190 | * | |
191 | * If a subsystem synchronizes ->css_online() and the start of iteration, a | |
192 | * css which finished ->css_online() is guaranteed to be visible in the | |
193 | * future iterations and will stay visible until the last reference is put. | |
194 | * A css which hasn't finished ->css_online() or already finished | |
195 | * ->css_offline() may show up during traversal. It's each subsystem's | |
196 | * responsibility to synchronize against on/offlining. | |
197 | * | |
198 | * For example, the following guarantees that a descendant can't escape | |
199 | * state updates of its ancestors. | |
200 | * | |
201 | * my_online(@css) | |
202 | * { | |
203 | * Lock @css's parent and @css; | |
204 | * Inherit state from the parent; | |
205 | * Unlock both. | |
206 | * } | |
207 | * | |
208 | * my_update_state(@css) | |
209 | * { | |
210 | * css_for_each_descendant_pre(@pos, @css) { | |
211 | * Lock @pos; | |
212 | * if (@pos == @css) | |
213 | * Update @css's state; | |
214 | * else | |
215 | * Verify @pos is alive and inherit state from its parent; | |
216 | * Unlock @pos; | |
217 | * } | |
218 | * } | |
219 | * | |
220 | * As long as the inheriting step, including checking the parent state, is | |
221 | * enclosed inside @pos locking, double-locking the parent isn't necessary | |
222 | * while inheriting. The state update to the parent is guaranteed to be | |
223 | * visible by walking order and, as long as inheriting operations to the | |
224 | * same @pos are atomic to each other, multiple updates racing each other | |
225 | * still result in the correct state. It's guaranateed that at least one | |
226 | * inheritance happens for any css after the latest update to its parent. | |
227 | * | |
228 | * If checking parent's state requires locking the parent, each inheriting | |
229 | * iteration should lock and unlock both @pos->parent and @pos. | |
230 | * | |
231 | * Alternatively, a subsystem may choose to use a single global lock to | |
232 | * synchronize ->css_online() and ->css_offline() against tree-walking | |
233 | * operations. | |
234 | * | |
235 | * It is allowed to temporarily drop RCU read lock during iteration. The | |
236 | * caller is responsible for ensuring that @pos remains accessible until | |
237 | * the start of the next iteration by, for example, bumping the css refcnt. | |
238 | */ | |
239 | #define css_for_each_descendant_pre(pos, css) \ | |
240 | for ((pos) = css_next_descendant_pre(NULL, (css)); (pos); \ | |
241 | (pos) = css_next_descendant_pre((pos), (css))) | |
242 | ||
243 | /** | |
244 | * css_for_each_descendant_post - post-order walk of a css's descendants | |
245 | * @pos: the css * to use as the loop cursor | |
246 | * @css: css whose descendants to walk | |
247 | * | |
248 | * Similar to css_for_each_descendant_pre() but performs post-order | |
249 | * traversal instead. @root is included in the iteration and the last | |
250 | * node to be visited. | |
251 | * | |
252 | * If a subsystem synchronizes ->css_online() and the start of iteration, a | |
253 | * css which finished ->css_online() is guaranteed to be visible in the | |
254 | * future iterations and will stay visible until the last reference is put. | |
255 | * A css which hasn't finished ->css_online() or already finished | |
256 | * ->css_offline() may show up during traversal. It's each subsystem's | |
257 | * responsibility to synchronize against on/offlining. | |
258 | * | |
259 | * Note that the walk visibility guarantee example described in pre-order | |
260 | * walk doesn't apply the same to post-order walks. | |
261 | */ | |
262 | #define css_for_each_descendant_post(pos, css) \ | |
263 | for ((pos) = css_next_descendant_post(NULL, (css)); (pos); \ | |
264 | (pos) = css_next_descendant_post((pos), (css))) | |
265 | ||
266 | /** | |
267 | * cgroup_taskset_for_each - iterate cgroup_taskset | |
268 | * @task: the loop cursor | |
1f7dd3e5 | 269 | * @dst_css: the destination css |
c326aa2b | 270 | * @tset: taskset to iterate |
4530eddb TH |
271 | * |
272 | * @tset may contain multiple tasks and they may belong to multiple | |
1f7dd3e5 TH |
273 | * processes. |
274 | * | |
275 | * On the v2 hierarchy, there may be tasks from multiple processes and they | |
276 | * may not share the source or destination csses. | |
277 | * | |
278 | * On traditional hierarchies, when there are multiple tasks in @tset, if a | |
279 | * task of a process is in @tset, all tasks of the process are in @tset. | |
280 | * Also, all are guaranteed to share the same source and destination csses. | |
4530eddb TH |
281 | * |
282 | * Iteration is not in any specific order. | |
c326aa2b | 283 | */ |
1f7dd3e5 TH |
284 | #define cgroup_taskset_for_each(task, dst_css, tset) \ |
285 | for ((task) = cgroup_taskset_first((tset), &(dst_css)); \ | |
286 | (task); \ | |
287 | (task) = cgroup_taskset_next((tset), &(dst_css))) | |
ddbcc7e8 | 288 | |
4530eddb TH |
289 | /** |
290 | * cgroup_taskset_for_each_leader - iterate group leaders in a cgroup_taskset | |
291 | * @leader: the loop cursor | |
1f7dd3e5 | 292 | * @dst_css: the destination css |
7b4632f0 | 293 | * @tset: taskset to iterate |
4530eddb TH |
294 | * |
295 | * Iterate threadgroup leaders of @tset. For single-task migrations, @tset | |
296 | * may not contain any. | |
297 | */ | |
1f7dd3e5 TH |
298 | #define cgroup_taskset_for_each_leader(leader, dst_css, tset) \ |
299 | for ((leader) = cgroup_taskset_first((tset), &(dst_css)); \ | |
300 | (leader); \ | |
301 | (leader) = cgroup_taskset_next((tset), &(dst_css))) \ | |
4530eddb TH |
302 | if ((leader) != (leader)->group_leader) \ |
303 | ; \ | |
304 | else | |
305 | ||
c326aa2b TH |
306 | /* |
307 | * Inline functions. | |
308 | */ | |
ddbcc7e8 | 309 | |
74321038 TH |
310 | static inline u64 cgroup_id(struct cgroup *cgrp) |
311 | { | |
312 | return cgrp->kn->id; | |
313 | } | |
314 | ||
5de0107e TH |
315 | /** |
316 | * css_get - obtain a reference on the specified css | |
317 | * @css: target css | |
318 | * | |
319 | * The caller must already have a reference. | |
ddbcc7e8 | 320 | */ |
ddbcc7e8 PM |
321 | static inline void css_get(struct cgroup_subsys_state *css) |
322 | { | |
3b514d24 TH |
323 | if (!(css->flags & CSS_NO_REF)) |
324 | percpu_ref_get(&css->refcnt); | |
ddbcc7e8 | 325 | } |
e7c5ec91 | 326 | |
e8ea14cc JW |
327 | /** |
328 | * css_get_many - obtain references on the specified css | |
329 | * @css: target css | |
330 | * @n: number of references to get | |
331 | * | |
332 | * The caller must already have a reference. | |
333 | */ | |
334 | static inline void css_get_many(struct cgroup_subsys_state *css, unsigned int n) | |
335 | { | |
336 | if (!(css->flags & CSS_NO_REF)) | |
337 | percpu_ref_get_many(&css->refcnt, n); | |
338 | } | |
339 | ||
6f4524d3 TH |
340 | /** |
341 | * css_tryget - try to obtain a reference on the specified css | |
342 | * @css: target css | |
343 | * | |
344 | * Obtain a reference on @css unless it already has reached zero and is | |
345 | * being released. This function doesn't care whether @css is on or | |
346 | * offline. The caller naturally needs to ensure that @css is accessible | |
347 | * but doesn't have to be holding a reference on it - IOW, RCU protected | |
348 | * access is good enough for this function. Returns %true if a reference | |
349 | * count was successfully obtained; %false otherwise. | |
350 | */ | |
351 | static inline bool css_tryget(struct cgroup_subsys_state *css) | |
352 | { | |
353 | if (!(css->flags & CSS_NO_REF)) | |
354 | return percpu_ref_tryget(&css->refcnt); | |
355 | return true; | |
356 | } | |
357 | ||
5de0107e | 358 | /** |
ec903c0c | 359 | * css_tryget_online - try to obtain a reference on the specified css if online |
5de0107e TH |
360 | * @css: target css |
361 | * | |
ec903c0c TH |
362 | * Obtain a reference on @css if it's online. The caller naturally needs |
363 | * to ensure that @css is accessible but doesn't have to be holding a | |
5de0107e TH |
364 | * reference on it - IOW, RCU protected access is good enough for this |
365 | * function. Returns %true if a reference count was successfully obtained; | |
366 | * %false otherwise. | |
367 | */ | |
ec903c0c | 368 | static inline bool css_tryget_online(struct cgroup_subsys_state *css) |
e7c5ec91 | 369 | { |
3b514d24 TH |
370 | if (!(css->flags & CSS_NO_REF)) |
371 | return percpu_ref_tryget_live(&css->refcnt); | |
372 | return true; | |
e7c5ec91 | 373 | } |
2f7ee569 | 374 | |
41c25707 TH |
375 | /** |
376 | * css_is_dying - test whether the specified css is dying | |
377 | * @css: target css | |
378 | * | |
379 | * Test whether @css is in the process of offlining or already offline. In | |
380 | * most cases, ->css_online() and ->css_offline() callbacks should be | |
381 | * enough; however, the actual offline operations are RCU delayed and this | |
382 | * test returns %true also when @css is scheduled to be offlined. | |
383 | * | |
384 | * This is useful, for example, when the use case requires synchronous | |
385 | * behavior with respect to cgroup removal. cgroup removal schedules css | |
386 | * offlining but the css can seem alive while the operation is being | |
387 | * delayed. If the delay affects user visible semantics, this test can be | |
388 | * used to resolve the situation. | |
389 | */ | |
390 | static inline bool css_is_dying(struct cgroup_subsys_state *css) | |
391 | { | |
392 | return !(css->flags & CSS_NO_REF) && percpu_ref_is_dying(&css->refcnt); | |
393 | } | |
394 | ||
2f7ee569 | 395 | /** |
5de0107e TH |
396 | * css_put - put a css reference |
397 | * @css: target css | |
398 | * | |
ec903c0c | 399 | * Put a reference obtained via css_get() and css_tryget_online(). |
2f7ee569 | 400 | */ |
ddbcc7e8 PM |
401 | static inline void css_put(struct cgroup_subsys_state *css) |
402 | { | |
3b514d24 TH |
403 | if (!(css->flags & CSS_NO_REF)) |
404 | percpu_ref_put(&css->refcnt); | |
ddbcc7e8 | 405 | } |
2f7ee569 | 406 | |
e8ea14cc JW |
407 | /** |
408 | * css_put_many - put css references | |
409 | * @css: target css | |
410 | * @n: number of references to put | |
411 | * | |
412 | * Put references obtained via css_get() and css_tryget_online(). | |
21acb9ca | 413 | */ |
e8ea14cc JW |
414 | static inline void css_put_many(struct cgroup_subsys_state *css, unsigned int n) |
415 | { | |
416 | if (!(css->flags & CSS_NO_REF)) | |
417 | percpu_ref_put_many(&css->refcnt, n); | |
418 | } | |
ddbcc7e8 | 419 | |
3e48930c TH |
420 | static inline void cgroup_get(struct cgroup *cgrp) |
421 | { | |
422 | css_get(&cgrp->self); | |
423 | } | |
424 | ||
425 | static inline bool cgroup_tryget(struct cgroup *cgrp) | |
426 | { | |
427 | return css_tryget(&cgrp->self); | |
428 | } | |
429 | ||
16af4396 TH |
430 | static inline void cgroup_put(struct cgroup *cgrp) |
431 | { | |
432 | css_put(&cgrp->self); | |
433 | } | |
434 | ||
14611e51 TH |
435 | /** |
436 | * task_css_set_check - obtain a task's css_set with extra access conditions | |
437 | * @task: the task to obtain css_set for | |
438 | * @__c: extra condition expression to be passed to rcu_dereference_check() | |
439 | * | |
440 | * A task's css_set is RCU protected, initialized and exited while holding | |
441 | * task_lock(), and can only be modified while holding both cgroup_mutex | |
442 | * and task_lock() while the task is alive. This macro verifies that the | |
443 | * caller is inside proper critical section and returns @task's css_set. | |
444 | * | |
445 | * The caller can also specify additional allowed conditions via @__c, such | |
446 | * as locks used during the cgroup_subsys::attach() methods. | |
dc61b1d6 | 447 | */ |
2219449a TH |
448 | #ifdef CONFIG_PROVE_RCU |
449 | extern struct mutex cgroup_mutex; | |
f0d9a5f1 | 450 | extern spinlock_t css_set_lock; |
14611e51 TH |
451 | #define task_css_set_check(task, __c) \ |
452 | rcu_dereference_check((task)->cgroups, \ | |
0e1d768f | 453 | lockdep_is_held(&cgroup_mutex) || \ |
f0d9a5f1 | 454 | lockdep_is_held(&css_set_lock) || \ |
0e1d768f | 455 | ((task)->flags & PF_EXITING) || (__c)) |
2219449a | 456 | #else |
14611e51 TH |
457 | #define task_css_set_check(task, __c) \ |
458 | rcu_dereference((task)->cgroups) | |
2219449a | 459 | #endif |
dc61b1d6 | 460 | |
14611e51 | 461 | /** |
8af01f56 | 462 | * task_css_check - obtain css for (task, subsys) w/ extra access conds |
14611e51 TH |
463 | * @task: the target task |
464 | * @subsys_id: the target subsystem ID | |
465 | * @__c: extra condition expression to be passed to rcu_dereference_check() | |
466 | * | |
467 | * Return the cgroup_subsys_state for the (@task, @subsys_id) pair. The | |
468 | * synchronization rules are the same as task_css_set_check(). | |
469 | */ | |
8af01f56 | 470 | #define task_css_check(task, subsys_id, __c) \ |
14611e51 TH |
471 | task_css_set_check((task), (__c))->subsys[(subsys_id)] |
472 | ||
473 | /** | |
474 | * task_css_set - obtain a task's css_set | |
475 | * @task: the task to obtain css_set for | |
476 | * | |
477 | * See task_css_set_check(). | |
478 | */ | |
479 | static inline struct css_set *task_css_set(struct task_struct *task) | |
480 | { | |
481 | return task_css_set_check(task, false); | |
482 | } | |
483 | ||
484 | /** | |
8af01f56 | 485 | * task_css - obtain css for (task, subsys) |
14611e51 TH |
486 | * @task: the target task |
487 | * @subsys_id: the target subsystem ID | |
488 | * | |
8af01f56 | 489 | * See task_css_check(). |
14611e51 | 490 | */ |
8af01f56 TH |
491 | static inline struct cgroup_subsys_state *task_css(struct task_struct *task, |
492 | int subsys_id) | |
ddbcc7e8 | 493 | { |
8af01f56 | 494 | return task_css_check(task, subsys_id, false); |
ddbcc7e8 PM |
495 | } |
496 | ||
ec438699 TH |
497 | /** |
498 | * task_get_css - find and get the css for (task, subsys) | |
499 | * @task: the target task | |
500 | * @subsys_id: the target subsystem ID | |
501 | * | |
502 | * Find the css for the (@task, @subsys_id) combination, increment a | |
503 | * reference on and return it. This function is guaranteed to return a | |
18fa84a2 | 504 | * valid css. The returned css may already have been offlined. |
ec438699 TH |
505 | */ |
506 | static inline struct cgroup_subsys_state * | |
507 | task_get_css(struct task_struct *task, int subsys_id) | |
508 | { | |
509 | struct cgroup_subsys_state *css; | |
510 | ||
511 | rcu_read_lock(); | |
512 | while (true) { | |
513 | css = task_css(task, subsys_id); | |
18fa84a2 TH |
514 | /* |
515 | * Can't use css_tryget_online() here. A task which has | |
516 | * PF_EXITING set may stay associated with an offline css. | |
517 | * If such task calls this function, css_tryget_online() | |
518 | * will keep failing. | |
519 | */ | |
520 | if (likely(css_tryget(css))) | |
ec438699 TH |
521 | break; |
522 | cpu_relax(); | |
523 | } | |
524 | rcu_read_unlock(); | |
525 | return css; | |
526 | } | |
527 | ||
5024ae29 TH |
528 | /** |
529 | * task_css_is_root - test whether a task belongs to the root css | |
530 | * @task: the target task | |
531 | * @subsys_id: the target subsystem ID | |
532 | * | |
533 | * Test whether @task belongs to the root css on the specified subsystem. | |
534 | * May be invoked in any context. | |
535 | */ | |
536 | static inline bool task_css_is_root(struct task_struct *task, int subsys_id) | |
537 | { | |
538 | return task_css_check(task, subsys_id, true) == | |
539 | init_css_set.subsys[subsys_id]; | |
540 | } | |
541 | ||
8af01f56 TH |
542 | static inline struct cgroup *task_cgroup(struct task_struct *task, |
543 | int subsys_id) | |
ddbcc7e8 | 544 | { |
8af01f56 | 545 | return task_css(task, subsys_id)->cgroup; |
ddbcc7e8 PM |
546 | } |
547 | ||
3e48930c TH |
548 | static inline struct cgroup *task_dfl_cgroup(struct task_struct *task) |
549 | { | |
550 | return task_css_set(task)->dfl_cgrp; | |
551 | } | |
552 | ||
553 | static inline struct cgroup *cgroup_parent(struct cgroup *cgrp) | |
554 | { | |
555 | struct cgroup_subsys_state *parent_css = cgrp->self.parent; | |
556 | ||
557 | if (parent_css) | |
558 | return container_of(parent_css, struct cgroup, self); | |
559 | return NULL; | |
560 | } | |
561 | ||
b11cfb58 TH |
562 | /** |
563 | * cgroup_is_descendant - test ancestry | |
564 | * @cgrp: the cgroup to be tested | |
565 | * @ancestor: possible ancestor of @cgrp | |
566 | * | |
567 | * Test whether @cgrp is a descendant of @ancestor. It also returns %true | |
568 | * if @cgrp == @ancestor. This function is safe to call as long as @cgrp | |
569 | * and @ancestor are accessible. | |
570 | */ | |
571 | static inline bool cgroup_is_descendant(struct cgroup *cgrp, | |
572 | struct cgroup *ancestor) | |
573 | { | |
574 | if (cgrp->root != ancestor->root || cgrp->level < ancestor->level) | |
575 | return false; | |
74321038 | 576 | return cgrp->ancestor_ids[ancestor->level] == cgroup_id(ancestor); |
b11cfb58 TH |
577 | } |
578 | ||
77236281 AI |
579 | /** |
580 | * cgroup_ancestor - find ancestor of cgroup | |
581 | * @cgrp: cgroup to find ancestor of | |
582 | * @ancestor_level: level of ancestor to find starting from root | |
583 | * | |
584 | * Find ancestor of cgroup at specified level starting from root if it exists | |
585 | * and return pointer to it. Return NULL if @cgrp doesn't have ancestor at | |
586 | * @ancestor_level. | |
587 | * | |
588 | * This function is safe to call as long as @cgrp is accessible. | |
589 | */ | |
590 | static inline struct cgroup *cgroup_ancestor(struct cgroup *cgrp, | |
591 | int ancestor_level) | |
592 | { | |
77236281 AI |
593 | if (cgrp->level < ancestor_level) |
594 | return NULL; | |
808c43b7 AI |
595 | while (cgrp && cgrp->level > ancestor_level) |
596 | cgrp = cgroup_parent(cgrp); | |
597 | return cgrp; | |
77236281 AI |
598 | } |
599 | ||
aed704b7 SD |
600 | /** |
601 | * task_under_cgroup_hierarchy - test task's membership of cgroup ancestry | |
602 | * @task: the task to be tested | |
603 | * @ancestor: possible ancestor of @task's cgroup | |
604 | * | |
605 | * Tests whether @task's default cgroup hierarchy is a descendant of @ancestor. | |
606 | * It follows all the same rules as cgroup_is_descendant, and only applies | |
607 | * to the default hierarchy. | |
608 | */ | |
609 | static inline bool task_under_cgroup_hierarchy(struct task_struct *task, | |
610 | struct cgroup *ancestor) | |
611 | { | |
612 | struct css_set *cset = task_css_set(task); | |
613 | ||
614 | return cgroup_is_descendant(cset->dfl_cgrp, ancestor); | |
615 | } | |
616 | ||
07bc356e | 617 | /* no synchronization, the result can only be used as a hint */ |
27bd4dbb | 618 | static inline bool cgroup_is_populated(struct cgroup *cgrp) |
07bc356e | 619 | { |
454000ad TH |
620 | return cgrp->nr_populated_csets + cgrp->nr_populated_domain_children + |
621 | cgrp->nr_populated_threaded_children; | |
07bc356e | 622 | } |
f3d46500 | 623 | |
f29374b1 | 624 | /* returns ino associated with a cgroup */ |
b1664924 TH |
625 | static inline ino_t cgroup_ino(struct cgroup *cgrp) |
626 | { | |
67c0496e | 627 | return kernfs_ino(cgrp->kn); |
b1664924 | 628 | } |
3ebb2b6e | 629 | |
b4168640 TH |
630 | /* cft/css accessors for cftype->write() operation */ |
631 | static inline struct cftype *of_cft(struct kernfs_open_file *of) | |
7da11279 | 632 | { |
2bd59d48 | 633 | return of->kn->priv; |
7da11279 | 634 | } |
0f0a2b4f | 635 | |
b4168640 | 636 | struct cgroup_subsys_state *of_css(struct kernfs_open_file *of); |
817929ec | 637 | |
b4168640 TH |
638 | /* cft/css accessors for cftype->seq_*() operations */ |
639 | static inline struct cftype *seq_cft(struct seq_file *seq) | |
640 | { | |
641 | return of_cft(seq->private); | |
642 | } | |
e535837b | 643 | |
b4168640 TH |
644 | static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq) |
645 | { | |
646 | return of_css(seq->private); | |
647 | } | |
31583bb0 | 648 | |
e61734c5 TH |
649 | /* |
650 | * Name / path handling functions. All are thin wrappers around the kernfs | |
651 | * counterparts and can be called under any context. | |
652 | */ | |
38460b48 | 653 | |
e61734c5 TH |
654 | static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen) |
655 | { | |
fdce6bf8 | 656 | return kernfs_name(cgrp->kn, buf, buflen); |
e61734c5 | 657 | } |
ddbcc7e8 | 658 | |
4c737b41 | 659 | static inline int cgroup_path(struct cgroup *cgrp, char *buf, size_t buflen) |
e61734c5 | 660 | { |
fdce6bf8 | 661 | return kernfs_path(cgrp->kn, buf, buflen); |
e61734c5 | 662 | } |
f3ba5380 | 663 | |
e61734c5 TH |
664 | static inline void pr_cont_cgroup_name(struct cgroup *cgrp) |
665 | { | |
fdce6bf8 | 666 | pr_cont_kernfs_name(cgrp->kn); |
e61734c5 | 667 | } |
ddbcc7e8 | 668 | |
e61734c5 | 669 | static inline void pr_cont_cgroup_path(struct cgroup *cgrp) |
846c7bb0 | 670 | { |
fdce6bf8 | 671 | pr_cont_kernfs_path(cgrp->kn); |
846c7bb0 | 672 | } |
ddbcc7e8 | 673 | |
2ce7135a JW |
674 | static inline struct psi_group *cgroup_psi(struct cgroup *cgrp) |
675 | { | |
676 | return &cgrp->psi; | |
677 | } | |
678 | ||
77f88796 TH |
679 | static inline void cgroup_init_kthreadd(void) |
680 | { | |
681 | /* | |
682 | * kthreadd is inherited by all kthreads, keep it in the root so | |
683 | * that the new kthreads are guaranteed to stay in the root until | |
684 | * initialization is finished. | |
685 | */ | |
686 | current->no_cgroup_migration = 1; | |
687 | } | |
688 | ||
689 | static inline void cgroup_kthread_ready(void) | |
690 | { | |
691 | /* | |
692 | * This kthread finished initialization. The creator should have | |
693 | * set PF_NO_SETAFFINITY if this kthread should stay in the root. | |
694 | */ | |
695 | current->no_cgroup_migration = 0; | |
696 | } | |
697 | ||
67c0496e | 698 | void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen); |
ddbcc7e8 PM |
699 | #else /* !CONFIG_CGROUPS */ |
700 | ||
f3ba5380 | 701 | struct cgroup_subsys_state; |
aed704b7 | 702 | struct cgroup; |
f3ba5380 | 703 | |
74321038 | 704 | static inline u64 cgroup_id(struct cgroup *cgrp) { return 1; } |
9b0eb69b | 705 | static inline void css_get(struct cgroup_subsys_state *css) {} |
c326aa2b | 706 | static inline void css_put(struct cgroup_subsys_state *css) {} |
31583bb0 | 707 | static inline int cgroup_attach_task_all(struct task_struct *from, |
c326aa2b TH |
708 | struct task_struct *t) { return 0; } |
709 | static inline int cgroupstats_build(struct cgroupstats *stats, | |
710 | struct dentry *dentry) { return -EINVAL; } | |
711 | ||
b4f48b63 | 712 | static inline void cgroup_fork(struct task_struct *p) {} |
ef2c41cf CB |
713 | static inline int cgroup_can_fork(struct task_struct *p, |
714 | struct kernel_clone_args *kargs) { return 0; } | |
715 | static inline void cgroup_cancel_fork(struct task_struct *p, | |
716 | struct kernel_clone_args *kargs) {} | |
717 | static inline void cgroup_post_fork(struct task_struct *p, | |
718 | struct kernel_clone_args *kargs) {} | |
1ec41830 | 719 | static inline void cgroup_exit(struct task_struct *p) {} |
51bee5ab | 720 | static inline void cgroup_release(struct task_struct *p) {} |
2e91fa7f | 721 | static inline void cgroup_free(struct task_struct *p) {} |
ddbcc7e8 | 722 | |
c326aa2b TH |
723 | static inline int cgroup_init_early(void) { return 0; } |
724 | static inline int cgroup_init(void) { return 0; } | |
77f88796 TH |
725 | static inline void cgroup_init_kthreadd(void) {} |
726 | static inline void cgroup_kthread_ready(void) {} | |
d7926ee3 | 727 | |
2ce7135a JW |
728 | static inline struct cgroup *cgroup_parent(struct cgroup *cgrp) |
729 | { | |
730 | return NULL; | |
731 | } | |
732 | ||
733 | static inline struct psi_group *cgroup_psi(struct cgroup *cgrp) | |
734 | { | |
735 | return NULL; | |
736 | } | |
737 | ||
aed704b7 SD |
738 | static inline bool task_under_cgroup_hierarchy(struct task_struct *task, |
739 | struct cgroup *ancestor) | |
740 | { | |
741 | return true; | |
742 | } | |
69fd5c39 | 743 | |
67c0496e TH |
744 | static inline void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen) |
745 | {} | |
ddbcc7e8 PM |
746 | #endif /* !CONFIG_CGROUPS */ |
747 | ||
6162cef0 | 748 | #ifdef CONFIG_CGROUPS |
d2cc5ed6 | 749 | /* |
6162cef0 | 750 | * cgroup scalable recursive statistics. |
d2cc5ed6 | 751 | */ |
6162cef0 TH |
752 | void cgroup_rstat_updated(struct cgroup *cgrp, int cpu); |
753 | void cgroup_rstat_flush(struct cgroup *cgrp); | |
0fa294fb | 754 | void cgroup_rstat_flush_irqsafe(struct cgroup *cgrp); |
6162cef0 TH |
755 | void cgroup_rstat_flush_hold(struct cgroup *cgrp); |
756 | void cgroup_rstat_flush_release(void); | |
d2cc5ed6 | 757 | |
6162cef0 TH |
758 | /* |
759 | * Basic resource stats. | |
760 | */ | |
d2cc5ed6 TH |
761 | #ifdef CONFIG_CGROUP_CPUACCT |
762 | void cpuacct_charge(struct task_struct *tsk, u64 cputime); | |
763 | void cpuacct_account_field(struct task_struct *tsk, int index, u64 val); | |
764 | #else | |
765 | static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {} | |
766 | static inline void cpuacct_account_field(struct task_struct *tsk, int index, | |
767 | u64 val) {} | |
768 | #endif | |
769 | ||
041cd640 TH |
770 | void __cgroup_account_cputime(struct cgroup *cgrp, u64 delta_exec); |
771 | void __cgroup_account_cputime_field(struct cgroup *cgrp, | |
772 | enum cpu_usage_stat index, u64 delta_exec); | |
773 | ||
d2cc5ed6 TH |
774 | static inline void cgroup_account_cputime(struct task_struct *task, |
775 | u64 delta_exec) | |
776 | { | |
041cd640 TH |
777 | struct cgroup *cgrp; |
778 | ||
d2cc5ed6 | 779 | cpuacct_charge(task, delta_exec); |
041cd640 TH |
780 | |
781 | rcu_read_lock(); | |
782 | cgrp = task_dfl_cgroup(task); | |
783 | if (cgroup_parent(cgrp)) | |
784 | __cgroup_account_cputime(cgrp, delta_exec); | |
785 | rcu_read_unlock(); | |
d2cc5ed6 TH |
786 | } |
787 | ||
788 | static inline void cgroup_account_cputime_field(struct task_struct *task, | |
789 | enum cpu_usage_stat index, | |
790 | u64 delta_exec) | |
791 | { | |
041cd640 TH |
792 | struct cgroup *cgrp; |
793 | ||
d2cc5ed6 | 794 | cpuacct_account_field(task, index, delta_exec); |
041cd640 TH |
795 | |
796 | rcu_read_lock(); | |
797 | cgrp = task_dfl_cgroup(task); | |
798 | if (cgroup_parent(cgrp)) | |
799 | __cgroup_account_cputime_field(cgrp, index, delta_exec); | |
800 | rcu_read_unlock(); | |
d2cc5ed6 TH |
801 | } |
802 | ||
803 | #else /* CONFIG_CGROUPS */ | |
804 | ||
805 | static inline void cgroup_account_cputime(struct task_struct *task, | |
806 | u64 delta_exec) {} | |
807 | static inline void cgroup_account_cputime_field(struct task_struct *task, | |
808 | enum cpu_usage_stat index, | |
809 | u64 delta_exec) {} | |
810 | ||
811 | #endif /* CONFIG_CGROUPS */ | |
812 | ||
bd1060a1 TH |
813 | /* |
814 | * sock->sk_cgrp_data handling. For more info, see sock_cgroup_data | |
815 | * definition in cgroup-defs.h. | |
816 | */ | |
817 | #ifdef CONFIG_SOCK_CGROUP_DATA | |
818 | ||
819 | #if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID) | |
820 | extern spinlock_t cgroup_sk_update_lock; | |
821 | #endif | |
822 | ||
823 | void cgroup_sk_alloc_disable(void); | |
824 | void cgroup_sk_alloc(struct sock_cgroup_data *skcd); | |
ad0f75e5 | 825 | void cgroup_sk_clone(struct sock_cgroup_data *skcd); |
bd1060a1 TH |
826 | void cgroup_sk_free(struct sock_cgroup_data *skcd); |
827 | ||
828 | static inline struct cgroup *sock_cgroup_ptr(struct sock_cgroup_data *skcd) | |
829 | { | |
830 | #if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID) | |
831 | unsigned long v; | |
832 | ||
833 | /* | |
834 | * @skcd->val is 64bit but the following is safe on 32bit too as we | |
835 | * just need the lower ulong to be written and read atomically. | |
836 | */ | |
837 | v = READ_ONCE(skcd->val); | |
838 | ||
ad0f75e5 | 839 | if (v & 3) |
bd1060a1 TH |
840 | return &cgrp_dfl_root.cgrp; |
841 | ||
842 | return (struct cgroup *)(unsigned long)v ?: &cgrp_dfl_root.cgrp; | |
843 | #else | |
844 | return (struct cgroup *)(unsigned long)skcd->val; | |
845 | #endif | |
846 | } | |
847 | ||
848 | #else /* CONFIG_CGROUP_DATA */ | |
849 | ||
850 | static inline void cgroup_sk_alloc(struct sock_cgroup_data *skcd) {} | |
ad0f75e5 | 851 | static inline void cgroup_sk_clone(struct sock_cgroup_data *skcd) {} |
bd1060a1 TH |
852 | static inline void cgroup_sk_free(struct sock_cgroup_data *skcd) {} |
853 | ||
854 | #endif /* CONFIG_CGROUP_DATA */ | |
855 | ||
a79a908f | 856 | struct cgroup_namespace { |
a79a908f AK |
857 | struct ns_common ns; |
858 | struct user_namespace *user_ns; | |
d08311dd | 859 | struct ucounts *ucounts; |
a79a908f AK |
860 | struct css_set *root_cset; |
861 | }; | |
862 | ||
863 | extern struct cgroup_namespace init_cgroup_ns; | |
864 | ||
865 | #ifdef CONFIG_CGROUPS | |
866 | ||
867 | void free_cgroup_ns(struct cgroup_namespace *ns); | |
868 | ||
869 | struct cgroup_namespace *copy_cgroup_ns(unsigned long flags, | |
870 | struct user_namespace *user_ns, | |
871 | struct cgroup_namespace *old_ns); | |
872 | ||
4c737b41 TH |
873 | int cgroup_path_ns(struct cgroup *cgrp, char *buf, size_t buflen, |
874 | struct cgroup_namespace *ns); | |
a79a908f AK |
875 | |
876 | #else /* !CONFIG_CGROUPS */ | |
877 | ||
878 | static inline void free_cgroup_ns(struct cgroup_namespace *ns) { } | |
879 | static inline struct cgroup_namespace * | |
880 | copy_cgroup_ns(unsigned long flags, struct user_namespace *user_ns, | |
881 | struct cgroup_namespace *old_ns) | |
882 | { | |
883 | return old_ns; | |
884 | } | |
885 | ||
886 | #endif /* !CONFIG_CGROUPS */ | |
887 | ||
888 | static inline void get_cgroup_ns(struct cgroup_namespace *ns) | |
889 | { | |
890 | if (ns) | |
f387882d | 891 | refcount_inc(&ns->ns.count); |
a79a908f AK |
892 | } |
893 | ||
894 | static inline void put_cgroup_ns(struct cgroup_namespace *ns) | |
895 | { | |
f387882d | 896 | if (ns && refcount_dec_and_test(&ns->ns.count)) |
a79a908f AK |
897 | free_cgroup_ns(ns); |
898 | } | |
899 | ||
76f969e8 RG |
900 | #ifdef CONFIG_CGROUPS |
901 | ||
902 | void cgroup_enter_frozen(void); | |
903 | void cgroup_leave_frozen(bool always_leave); | |
904 | void cgroup_update_frozen(struct cgroup *cgrp); | |
905 | void cgroup_freeze(struct cgroup *cgrp, bool freeze); | |
906 | void cgroup_freezer_migrate_task(struct task_struct *task, struct cgroup *src, | |
907 | struct cgroup *dst); | |
96b9c592 | 908 | |
76f969e8 RG |
909 | static inline bool cgroup_task_freeze(struct task_struct *task) |
910 | { | |
911 | bool ret; | |
912 | ||
913 | if (task->flags & PF_KTHREAD) | |
914 | return false; | |
915 | ||
916 | rcu_read_lock(); | |
917 | ret = test_bit(CGRP_FREEZE, &task_dfl_cgroup(task)->flags); | |
918 | rcu_read_unlock(); | |
919 | ||
920 | return ret; | |
921 | } | |
922 | ||
923 | static inline bool cgroup_task_frozen(struct task_struct *task) | |
924 | { | |
925 | return task->frozen; | |
926 | } | |
927 | ||
928 | #else /* !CONFIG_CGROUPS */ | |
929 | ||
930 | static inline void cgroup_enter_frozen(void) { } | |
931 | static inline void cgroup_leave_frozen(bool always_leave) { } | |
932 | static inline bool cgroup_task_freeze(struct task_struct *task) | |
933 | { | |
934 | return false; | |
935 | } | |
936 | static inline bool cgroup_task_frozen(struct task_struct *task) | |
937 | { | |
938 | return false; | |
939 | } | |
940 | ||
941 | #endif /* !CONFIG_CGROUPS */ | |
942 | ||
4bfc0bb2 RG |
943 | #ifdef CONFIG_CGROUP_BPF |
944 | static inline void cgroup_bpf_get(struct cgroup *cgrp) | |
945 | { | |
946 | percpu_ref_get(&cgrp->bpf.refcnt); | |
947 | } | |
948 | ||
949 | static inline void cgroup_bpf_put(struct cgroup *cgrp) | |
950 | { | |
951 | percpu_ref_put(&cgrp->bpf.refcnt); | |
952 | } | |
953 | ||
954 | #else /* CONFIG_CGROUP_BPF */ | |
955 | ||
956 | static inline void cgroup_bpf_get(struct cgroup *cgrp) {} | |
957 | static inline void cgroup_bpf_put(struct cgroup *cgrp) {} | |
958 | ||
959 | #endif /* CONFIG_CGROUP_BPF */ | |
960 | ||
ddbcc7e8 | 961 | #endif /* _LINUX_CGROUP_H */ |