Commit | Line | Data |
---|---|---|
ddbcc7e8 PM |
1 | #ifndef _LINUX_CGROUP_H |
2 | #define _LINUX_CGROUP_H | |
3 | /* | |
4 | * cgroup interface | |
5 | * | |
6 | * Copyright (C) 2003 BULL SA | |
7 | * Copyright (C) 2004-2006 Silicon Graphics, Inc. | |
8 | * | |
9 | */ | |
10 | ||
11 | #include <linux/sched.h> | |
ddbcc7e8 PM |
12 | #include <linux/cpumask.h> |
13 | #include <linux/nodemask.h> | |
14 | #include <linux/rcupdate.h> | |
eb6fd504 | 15 | #include <linux/rculist.h> |
846c7bb0 | 16 | #include <linux/cgroupstats.h> |
cc31edce | 17 | #include <linux/rwsem.h> |
38460b48 | 18 | #include <linux/idr.h> |
48ddbe19 | 19 | #include <linux/workqueue.h> |
25a7e684 | 20 | #include <linux/fs.h> |
d3daf28d | 21 | #include <linux/percpu-refcount.h> |
7da11279 | 22 | #include <linux/seq_file.h> |
2bd59d48 | 23 | #include <linux/kernfs.h> |
ddbcc7e8 PM |
24 | |
25 | #ifdef CONFIG_CGROUPS | |
26 | ||
3dd06ffa | 27 | struct cgroup_root; |
ddbcc7e8 PM |
28 | struct cgroup_subsys; |
29 | struct inode; | |
84eea842 | 30 | struct cgroup; |
ddbcc7e8 PM |
31 | |
32 | extern int cgroup_init_early(void); | |
33 | extern int cgroup_init(void); | |
b4f48b63 | 34 | extern void cgroup_fork(struct task_struct *p); |
817929ec | 35 | extern void cgroup_post_fork(struct task_struct *p); |
1ec41830 | 36 | extern void cgroup_exit(struct task_struct *p); |
846c7bb0 BS |
37 | extern int cgroupstats_build(struct cgroupstats *stats, |
38 | struct dentry *dentry); | |
ddbcc7e8 | 39 | |
8d8b97ba | 40 | extern int proc_cgroup_show(struct seq_file *, void *); |
a424316c | 41 | |
3ed80a62 | 42 | /* define the enumeration of all cgroup subsystems */ |
073219e9 | 43 | #define SUBSYS(_x) _x ## _cgrp_id, |
817929ec PM |
44 | enum cgroup_subsys_id { |
45 | #include <linux/cgroup_subsys.h> | |
a6f00298 | 46 | CGROUP_SUBSYS_COUNT, |
817929ec PM |
47 | }; |
48 | #undef SUBSYS | |
49 | ||
ddbcc7e8 PM |
50 | /* Per-subsystem/per-cgroup state maintained by the system. */ |
51 | struct cgroup_subsys_state { | |
72c97e54 | 52 | /* the cgroup that this css is attached to */ |
ddbcc7e8 PM |
53 | struct cgroup *cgroup; |
54 | ||
72c97e54 TH |
55 | /* the cgroup subsystem that this css is attached to */ |
56 | struct cgroup_subsys *ss; | |
57 | ||
d3daf28d TH |
58 | /* reference count - access via css_[try]get() and css_put() */ |
59 | struct percpu_ref refcnt; | |
ddbcc7e8 | 60 | |
0ae78e0b TH |
61 | /* the parent css */ |
62 | struct cgroup_subsys_state *parent; | |
63 | ||
ddbcc7e8 | 64 | unsigned long flags; |
48ddbe19 | 65 | |
0c21ead1 TH |
66 | /* percpu_ref killing and RCU release */ |
67 | struct rcu_head rcu_head; | |
35ef10da | 68 | struct work_struct destroy_work; |
ddbcc7e8 PM |
69 | }; |
70 | ||
71 | /* bits in struct cgroup_subsys_state flags field */ | |
72 | enum { | |
38b53aba | 73 | CSS_ROOT = (1 << 0), /* this CSS is the root of the subsystem */ |
92fb9748 | 74 | CSS_ONLINE = (1 << 1), /* between ->css_online() and ->css_offline() */ |
ddbcc7e8 PM |
75 | }; |
76 | ||
5de0107e TH |
77 | /** |
78 | * css_get - obtain a reference on the specified css | |
79 | * @css: target css | |
80 | * | |
81 | * The caller must already have a reference. | |
ddbcc7e8 | 82 | */ |
ddbcc7e8 PM |
83 | static inline void css_get(struct cgroup_subsys_state *css) |
84 | { | |
85 | /* We don't need to reference count the root state */ | |
38b53aba | 86 | if (!(css->flags & CSS_ROOT)) |
d3daf28d | 87 | percpu_ref_get(&css->refcnt); |
ddbcc7e8 | 88 | } |
e7c5ec91 | 89 | |
5de0107e TH |
90 | /** |
91 | * css_tryget - try to obtain a reference on the specified css | |
92 | * @css: target css | |
93 | * | |
94 | * Obtain a reference on @css if it's alive. The caller naturally needs to | |
95 | * ensure that @css is accessible but doesn't have to be holding a | |
96 | * reference on it - IOW, RCU protected access is good enough for this | |
97 | * function. Returns %true if a reference count was successfully obtained; | |
98 | * %false otherwise. | |
99 | */ | |
e7c5ec91 PM |
100 | static inline bool css_tryget(struct cgroup_subsys_state *css) |
101 | { | |
38b53aba | 102 | if (css->flags & CSS_ROOT) |
e7c5ec91 | 103 | return true; |
d3daf28d | 104 | return percpu_ref_tryget(&css->refcnt); |
e7c5ec91 PM |
105 | } |
106 | ||
5de0107e TH |
107 | /** |
108 | * css_put - put a css reference | |
109 | * @css: target css | |
110 | * | |
111 | * Put a reference obtained via css_get() and css_tryget(). | |
112 | */ | |
ddbcc7e8 PM |
113 | static inline void css_put(struct cgroup_subsys_state *css) |
114 | { | |
38b53aba | 115 | if (!(css->flags & CSS_ROOT)) |
d3daf28d | 116 | percpu_ref_put(&css->refcnt); |
ddbcc7e8 PM |
117 | } |
118 | ||
3116f0e3 PM |
119 | /* bits in struct cgroup flags field */ |
120 | enum { | |
121 | /* Control Group is dead */ | |
54766d4a | 122 | CGRP_DEAD, |
d20a390a PM |
123 | /* |
124 | * Control Group has previously had a child cgroup or a task, | |
125 | * but no longer (only if CGRP_NOTIFY_ON_RELEASE is set) | |
126 | */ | |
3116f0e3 PM |
127 | CGRP_RELEASABLE, |
128 | /* Control Group requires release notifications to userspace */ | |
129 | CGRP_NOTIFY_ON_RELEASE, | |
97978e6d | 130 | /* |
2260e7fc TH |
131 | * Clone the parent's configuration when creating a new child |
132 | * cpuset cgroup. For historical reasons, this option can be | |
133 | * specified at mount time and thus is implemented here. | |
97978e6d | 134 | */ |
2260e7fc | 135 | CGRP_CPUSET_CLONE_CHILDREN, |
873fe09e TH |
136 | /* see the comment above CGRP_ROOT_SANE_BEHAVIOR for details */ |
137 | CGRP_SANE_BEHAVIOR, | |
3116f0e3 PM |
138 | }; |
139 | ||
ddbcc7e8 PM |
140 | struct cgroup { |
141 | unsigned long flags; /* "unsigned long" so bitops work */ | |
142 | ||
b414dc09 LZ |
143 | /* |
144 | * idr allocated in-hierarchy ID. | |
145 | * | |
146 | * The ID of the root cgroup is always 0, and a new cgroup | |
147 | * will be assigned with a smallest available ID. | |
0ab02ca8 LZ |
148 | * |
149 | * Allocating/Removing ID must be protected by cgroup_mutex. | |
b414dc09 LZ |
150 | */ |
151 | int id; | |
0a950f65 | 152 | |
f20104de TH |
153 | /* the number of attached css's */ |
154 | int nr_css; | |
155 | ||
2bd59d48 TH |
156 | atomic_t refcnt; |
157 | ||
ddbcc7e8 PM |
158 | /* |
159 | * We link our 'sibling' struct into our parent's 'children'. | |
160 | * Our children link their 'sibling' into our 'children'. | |
161 | */ | |
162 | struct list_head sibling; /* my parent's children */ | |
163 | struct list_head children; /* my children */ | |
164 | ||
d20a390a | 165 | struct cgroup *parent; /* my parent */ |
2bd59d48 | 166 | struct kernfs_node *kn; /* cgroup kernfs entry */ |
ddbcc7e8 | 167 | |
53fa5261 TH |
168 | /* |
169 | * Monotonically increasing unique serial number which defines a | |
170 | * uniform order among all cgroups. It's guaranteed that all | |
171 | * ->children lists are in the ascending order of ->serial_nr. | |
172 | * It's used to allow interrupting and resuming iterations. | |
173 | */ | |
174 | u64 serial_nr; | |
175 | ||
f392e51c TH |
176 | /* the bitmask of subsystems enabled on the child cgroups */ |
177 | unsigned long child_subsys_mask; | |
94419627 | 178 | |
ddbcc7e8 | 179 | /* Private pointers for each registered subsystem */ |
73e80ed8 | 180 | struct cgroup_subsys_state __rcu *subsys[CGROUP_SUBSYS_COUNT]; |
ddbcc7e8 | 181 | |
3dd06ffa | 182 | struct cgroup_root *root; |
817929ec PM |
183 | |
184 | /* | |
69d0206c TH |
185 | * List of cgrp_cset_links pointing at css_sets with tasks in this |
186 | * cgroup. Protected by css_set_lock. | |
817929ec | 187 | */ |
69d0206c | 188 | struct list_head cset_links; |
81a6a5cd PM |
189 | |
190 | /* | |
191 | * Linked list running through all cgroups that can | |
192 | * potentially be reaped by the release agent. Protected by | |
193 | * release_list_lock | |
194 | */ | |
195 | struct list_head release_list; | |
cc31edce | 196 | |
72a8cb30 BB |
197 | /* |
198 | * list of pidlists, up to two for each namespace (one for procs, one | |
199 | * for tasks); created on demand. | |
200 | */ | |
201 | struct list_head pidlists; | |
202 | struct mutex pidlist_mutex; | |
a47295e6 | 203 | |
67f4c36f TH |
204 | /* dummy css with NULL ->ss, points back to this cgroup */ |
205 | struct cgroup_subsys_state dummy_css; | |
206 | ||
d3daf28d | 207 | /* For css percpu_ref killing and RCU-protected deletion */ |
a47295e6 | 208 | struct rcu_head rcu_head; |
ea15f8cc | 209 | struct work_struct destroy_work; |
817929ec PM |
210 | }; |
211 | ||
25a7e684 TH |
212 | #define MAX_CGROUP_ROOT_NAMELEN 64 |
213 | ||
3dd06ffa | 214 | /* cgroup_root->flags */ |
25a7e684 | 215 | enum { |
873fe09e TH |
216 | /* |
217 | * Unfortunately, cgroup core and various controllers are riddled | |
218 | * with idiosyncrasies and pointless options. The following flag, | |
219 | * when set, will force sane behavior - some options are forced on, | |
220 | * others are disallowed, and some controllers will change their | |
221 | * hierarchical or other behaviors. | |
222 | * | |
223 | * The set of behaviors affected by this flag are still being | |
224 | * determined and developed and the mount option for this flag is | |
225 | * prefixed with __DEVEL__. The prefix will be dropped once we | |
226 | * reach the point where all behaviors are compatible with the | |
227 | * planned unified hierarchy, which will automatically turn on this | |
228 | * flag. | |
229 | * | |
230 | * The followings are the behaviors currently affected this flag. | |
231 | * | |
d3ba07c3 TH |
232 | * - Mount options "noprefix", "xattr", "clone_children", |
233 | * "release_agent" and "name" are disallowed. | |
873fe09e TH |
234 | * |
235 | * - When mounting an existing superblock, mount options should | |
236 | * match. | |
237 | * | |
238 | * - Remount is disallowed. | |
239 | * | |
0b0585c3 LT |
240 | * - rename(2) is disallowed. |
241 | * | |
f63674fd TH |
242 | * - "tasks" is removed. Everything should be at process |
243 | * granularity. Use "cgroup.procs" instead. | |
f00baae7 | 244 | * |
afb2bc14 TH |
245 | * - "cgroup.procs" is not sorted. pids will be unique unless they |
246 | * got recycled inbetween reads. | |
247 | * | |
f63674fd TH |
248 | * - "release_agent" and "notify_on_release" are removed. |
249 | * Replacement notification mechanism will be implemented. | |
873fe09e | 250 | * |
d3ba07c3 | 251 | * - "cgroup.clone_children" is removed. |
86bf4b68 | 252 | * |
a2dd4247 TH |
253 | * - If mount is requested with sane_behavior but without any |
254 | * subsystem, the default unified hierarchy is mounted. | |
255 | * | |
5c5cc623 LZ |
256 | * - cpuset: tasks will be kept in empty cpusets when hotplug happens |
257 | * and take masks of ancestors with non-empty cpus/mems, instead of | |
258 | * being moved to an ancestor. | |
259 | * | |
88fa523b LZ |
260 | * - cpuset: a task can be moved into an empty cpuset, and again it |
261 | * takes masks of ancestors. | |
6db8e85c | 262 | * |
f63674fd TH |
263 | * - memcg: use_hierarchy is on by default and the cgroup file for |
264 | * the flag is not created. | |
f00baae7 | 265 | * |
9138125b | 266 | * - blkcg: blk-throttle becomes properly hierarchical. |
873fe09e TH |
267 | */ |
268 | CGRP_ROOT_SANE_BEHAVIOR = (1 << 0), | |
269 | ||
25a7e684 TH |
270 | CGRP_ROOT_NOPREFIX = (1 << 1), /* mounted subsystems have no named prefix */ |
271 | CGRP_ROOT_XATTR = (1 << 2), /* supports extended attributes */ | |
0ce6cba3 TH |
272 | |
273 | /* mount options live below bit 16 */ | |
274 | CGRP_ROOT_OPTION_MASK = (1 << 16) - 1, | |
25a7e684 TH |
275 | }; |
276 | ||
277 | /* | |
3dd06ffa | 278 | * A cgroup_root represents the root of a cgroup hierarchy, and may be |
2bd59d48 | 279 | * associated with a kernfs_root to form an active hierarchy. This is |
25a7e684 TH |
280 | * internal to cgroup core. Don't access directly from controllers. |
281 | */ | |
3dd06ffa | 282 | struct cgroup_root { |
2bd59d48 | 283 | struct kernfs_root *kf_root; |
25a7e684 | 284 | |
f392e51c TH |
285 | /* The bitmask of subsystems attached to this hierarchy */ |
286 | unsigned long subsys_mask; | |
287 | ||
25a7e684 TH |
288 | /* Unique id for this hierarchy. */ |
289 | int hierarchy_id; | |
290 | ||
776f02fa | 291 | /* The root cgroup. Root is destroyed on its release. */ |
3dd06ffa | 292 | struct cgroup cgrp; |
25a7e684 | 293 | |
3c9c825b TH |
294 | /* Number of cgroups in the hierarchy, used only for /proc/cgroups */ |
295 | atomic_t nr_cgrps; | |
25a7e684 TH |
296 | |
297 | /* A list running through the active hierarchies */ | |
298 | struct list_head root_list; | |
299 | ||
25a7e684 TH |
300 | /* Hierarchy-specific flags */ |
301 | unsigned long flags; | |
302 | ||
303 | /* IDs for cgroups in this hierarchy */ | |
4e96ee8e | 304 | struct idr cgroup_idr; |
25a7e684 TH |
305 | |
306 | /* The path to use for release notifications. */ | |
307 | char release_agent_path[PATH_MAX]; | |
308 | ||
309 | /* The name for this hierarchy - may be empty */ | |
310 | char name[MAX_CGROUP_ROOT_NAMELEN]; | |
311 | }; | |
312 | ||
d20a390a PM |
313 | /* |
314 | * A css_set is a structure holding pointers to a set of | |
817929ec PM |
315 | * cgroup_subsys_state objects. This saves space in the task struct |
316 | * object and speeds up fork()/exit(), since a single inc/dec and a | |
d20a390a PM |
317 | * list_add()/del() can bump the reference count on the entire cgroup |
318 | * set for a task. | |
817929ec PM |
319 | */ |
320 | ||
321 | struct css_set { | |
322 | ||
323 | /* Reference count */ | |
146aa1bd | 324 | atomic_t refcount; |
817929ec | 325 | |
472b1053 LZ |
326 | /* |
327 | * List running through all cgroup groups in the same hash | |
328 | * slot. Protected by css_set_lock | |
329 | */ | |
330 | struct hlist_node hlist; | |
331 | ||
817929ec | 332 | /* |
c7561128 TH |
333 | * Lists running through all tasks using this cgroup group. |
334 | * mg_tasks lists tasks which belong to this cset but are in the | |
335 | * process of being migrated out or in. Protected by | |
336 | * css_set_rwsem, but, during migration, once tasks are moved to | |
337 | * mg_tasks, it can be read safely while holding cgroup_mutex. | |
817929ec PM |
338 | */ |
339 | struct list_head tasks; | |
c7561128 | 340 | struct list_head mg_tasks; |
817929ec PM |
341 | |
342 | /* | |
69d0206c TH |
343 | * List of cgrp_cset_links pointing at cgroups referenced from this |
344 | * css_set. Protected by css_set_lock. | |
817929ec | 345 | */ |
69d0206c | 346 | struct list_head cgrp_links; |
817929ec PM |
347 | |
348 | /* | |
3ed80a62 TH |
349 | * Set of subsystem states, one for each subsystem. This array is |
350 | * immutable after creation apart from the init_css_set during | |
351 | * subsystem registration (at boot time). | |
817929ec PM |
352 | */ |
353 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
c378369d | 354 | |
b3dc094e TH |
355 | /* |
356 | * List of csets participating in the on-going migration either as | |
357 | * source or destination. Protected by cgroup_mutex. | |
358 | */ | |
1958d2d5 | 359 | struct list_head mg_preload_node; |
b3dc094e TH |
360 | struct list_head mg_node; |
361 | ||
362 | /* | |
363 | * If this cset is acting as the source of migration the following | |
364 | * two fields are set. mg_src_cgrp is the source cgroup of the | |
365 | * on-going migration and mg_dst_cset is the destination cset the | |
366 | * target tasks on this cset should be migrated to. Protected by | |
367 | * cgroup_mutex. | |
368 | */ | |
369 | struct cgroup *mg_src_cgrp; | |
370 | struct css_set *mg_dst_cset; | |
371 | ||
c378369d BB |
372 | /* For RCU-protected deletion */ |
373 | struct rcu_head rcu_head; | |
ddbcc7e8 PM |
374 | }; |
375 | ||
d20a390a PM |
376 | /* |
377 | * struct cftype: handler definitions for cgroup control files | |
ddbcc7e8 PM |
378 | * |
379 | * When reading/writing to a file: | |
a043e3b2 | 380 | * - the cgroup to use is file->f_dentry->d_parent->d_fsdata |
ddbcc7e8 PM |
381 | * - the 'cftype' of the file is file->f_dentry->d_fsdata |
382 | */ | |
383 | ||
8e3f6541 | 384 | /* cftype->flags */ |
02c402d9 | 385 | enum { |
6f4b7e63 LZ |
386 | CFTYPE_ONLY_ON_ROOT = (1 << 0), /* only create on root cgrp */ |
387 | CFTYPE_NOT_ON_ROOT = (1 << 1), /* don't create on root cgrp */ | |
02c402d9 | 388 | CFTYPE_INSANE = (1 << 2), /* don't create if sane_behavior */ |
9fa4db33 | 389 | CFTYPE_NO_PREFIX = (1 << 3), /* (DON'T USE FOR NEW FILES) no subsys prefix */ |
8cbbf2c9 | 390 | CFTYPE_ONLY_ON_DFL = (1 << 4), /* only on default hierarchy */ |
02c402d9 | 391 | }; |
8e3f6541 TH |
392 | |
393 | #define MAX_CFTYPE_NAME 64 | |
394 | ||
ddbcc7e8 | 395 | struct cftype { |
d20a390a PM |
396 | /* |
397 | * By convention, the name should begin with the name of the | |
8e3f6541 TH |
398 | * subsystem, followed by a period. Zero length string indicates |
399 | * end of cftype array. | |
d20a390a | 400 | */ |
ddbcc7e8 PM |
401 | char name[MAX_CFTYPE_NAME]; |
402 | int private; | |
099fca32 LZ |
403 | /* |
404 | * If not 0, file mode is set to this value, otherwise it will | |
405 | * be figured out automatically | |
406 | */ | |
a5e7ed32 | 407 | umode_t mode; |
db3b1497 PM |
408 | |
409 | /* | |
5f469907 TH |
410 | * The maximum length of string, excluding trailing nul, that can |
411 | * be passed to write_string. If < PAGE_SIZE-1, PAGE_SIZE-1 is | |
412 | * assumed. | |
db3b1497 PM |
413 | */ |
414 | size_t max_write_len; | |
415 | ||
8e3f6541 TH |
416 | /* CFTYPE_* flags */ |
417 | unsigned int flags; | |
418 | ||
2bb566cb | 419 | /* |
0adb0704 TH |
420 | * Fields used for internal bookkeeping. Initialized automatically |
421 | * during registration. | |
2bb566cb | 422 | */ |
0adb0704 TH |
423 | struct cgroup_subsys *ss; /* NULL for cgroup core files */ |
424 | struct list_head node; /* anchored at ss->cfts */ | |
2bd59d48 TH |
425 | struct kernfs_ops *kf_ops; |
426 | ||
ddbcc7e8 | 427 | /* |
f4c753b7 | 428 | * read_u64() is a shortcut for the common case of returning a |
ddbcc7e8 PM |
429 | * single integer. Use it in place of read() |
430 | */ | |
182446d0 | 431 | u64 (*read_u64)(struct cgroup_subsys_state *css, struct cftype *cft); |
e73d2c61 PM |
432 | /* |
433 | * read_s64() is a signed version of read_u64() | |
434 | */ | |
182446d0 | 435 | s64 (*read_s64)(struct cgroup_subsys_state *css, struct cftype *cft); |
2da8ca82 TH |
436 | |
437 | /* generic seq_file read interface */ | |
438 | int (*seq_show)(struct seq_file *sf, void *v); | |
91796569 | 439 | |
6612f05b TH |
440 | /* optional ops, implement all or none */ |
441 | void *(*seq_start)(struct seq_file *sf, loff_t *ppos); | |
442 | void *(*seq_next)(struct seq_file *sf, void *v, loff_t *ppos); | |
443 | void (*seq_stop)(struct seq_file *sf, void *v); | |
444 | ||
355e0c48 | 445 | /* |
f4c753b7 | 446 | * write_u64() is a shortcut for the common case of accepting |
355e0c48 PM |
447 | * a single integer (as parsed by simple_strtoull) from |
448 | * userspace. Use in place of write(); return 0 or error. | |
449 | */ | |
182446d0 TH |
450 | int (*write_u64)(struct cgroup_subsys_state *css, struct cftype *cft, |
451 | u64 val); | |
e73d2c61 PM |
452 | /* |
453 | * write_s64() is a signed version of write_u64() | |
454 | */ | |
182446d0 TH |
455 | int (*write_s64)(struct cgroup_subsys_state *css, struct cftype *cft, |
456 | s64 val); | |
355e0c48 | 457 | |
db3b1497 PM |
458 | /* |
459 | * write_string() is passed a nul-terminated kernelspace | |
460 | * buffer of maximum length determined by max_write_len. | |
461 | * Returns 0 or -ve error code. | |
462 | */ | |
182446d0 | 463 | int (*write_string)(struct cgroup_subsys_state *css, struct cftype *cft, |
4d3bb511 | 464 | char *buffer); |
d447ea2f PE |
465 | /* |
466 | * trigger() callback can be used to get some kick from the | |
467 | * userspace, when the actual string written is not important | |
468 | * at all. The private field can be used to determine the | |
469 | * kick type for multiplexing. | |
470 | */ | |
182446d0 | 471 | int (*trigger)(struct cgroup_subsys_state *css, unsigned int event); |
2bd59d48 TH |
472 | |
473 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
474 | struct lock_class_key lockdep_key; | |
475 | #endif | |
ddbcc7e8 PM |
476 | }; |
477 | ||
a2dd4247 TH |
478 | extern struct cgroup_root cgrp_dfl_root; |
479 | ||
480 | static inline bool cgroup_on_dfl(const struct cgroup *cgrp) | |
481 | { | |
482 | return cgrp->root == &cgrp_dfl_root; | |
483 | } | |
484 | ||
873fe09e TH |
485 | /* |
486 | * See the comment above CGRP_ROOT_SANE_BEHAVIOR for details. This | |
487 | * function can be called as long as @cgrp is accessible. | |
488 | */ | |
489 | static inline bool cgroup_sane_behavior(const struct cgroup *cgrp) | |
490 | { | |
491 | return cgrp->root->flags & CGRP_ROOT_SANE_BEHAVIOR; | |
492 | } | |
493 | ||
07bc356e TH |
494 | /* no synchronization, the result can only be used as a hint */ |
495 | static inline bool cgroup_has_tasks(struct cgroup *cgrp) | |
496 | { | |
497 | return !list_empty(&cgrp->cset_links); | |
498 | } | |
499 | ||
b1664924 TH |
500 | /* returns ino associated with a cgroup, 0 indicates unmounted root */ |
501 | static inline ino_t cgroup_ino(struct cgroup *cgrp) | |
502 | { | |
2bd59d48 TH |
503 | if (cgrp->kn) |
504 | return cgrp->kn->ino; | |
b1664924 TH |
505 | else |
506 | return 0; | |
507 | } | |
508 | ||
7da11279 TH |
509 | static inline struct cftype *seq_cft(struct seq_file *seq) |
510 | { | |
2bd59d48 TH |
511 | struct kernfs_open_file *of = seq->private; |
512 | ||
513 | return of->kn->priv; | |
7da11279 TH |
514 | } |
515 | ||
59f5296b TH |
516 | struct cgroup_subsys_state *seq_css(struct seq_file *seq); |
517 | ||
e61734c5 TH |
518 | /* |
519 | * Name / path handling functions. All are thin wrappers around the kernfs | |
520 | * counterparts and can be called under any context. | |
521 | */ | |
522 | ||
523 | static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen) | |
524 | { | |
fdce6bf8 | 525 | return kernfs_name(cgrp->kn, buf, buflen); |
e61734c5 TH |
526 | } |
527 | ||
528 | static inline char * __must_check cgroup_path(struct cgroup *cgrp, char *buf, | |
529 | size_t buflen) | |
530 | { | |
fdce6bf8 | 531 | return kernfs_path(cgrp->kn, buf, buflen); |
e61734c5 TH |
532 | } |
533 | ||
534 | static inline void pr_cont_cgroup_name(struct cgroup *cgrp) | |
535 | { | |
fdce6bf8 | 536 | pr_cont_kernfs_name(cgrp->kn); |
e61734c5 TH |
537 | } |
538 | ||
539 | static inline void pr_cont_cgroup_path(struct cgroup *cgrp) | |
540 | { | |
fdce6bf8 | 541 | pr_cont_kernfs_path(cgrp->kn); |
e61734c5 TH |
542 | } |
543 | ||
544 | char *task_cgroup_path(struct task_struct *task, char *buf, size_t buflen); | |
545 | ||
03b1cde6 | 546 | int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); |
2bb566cb | 547 | int cgroup_rm_cftypes(struct cftype *cfts); |
8e3f6541 | 548 | |
78574cf9 | 549 | bool cgroup_is_descendant(struct cgroup *cgrp, struct cgroup *ancestor); |
ddbcc7e8 | 550 | |
2f7ee569 TH |
551 | /* |
552 | * Control Group taskset, used to pass around set of tasks to cgroup_subsys | |
553 | * methods. | |
554 | */ | |
555 | struct cgroup_taskset; | |
556 | struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset); | |
557 | struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset); | |
2f7ee569 TH |
558 | |
559 | /** | |
560 | * cgroup_taskset_for_each - iterate cgroup_taskset | |
561 | * @task: the loop cursor | |
2f7ee569 TH |
562 | * @tset: taskset to iterate |
563 | */ | |
924f0d9a | 564 | #define cgroup_taskset_for_each(task, tset) \ |
2f7ee569 | 565 | for ((task) = cgroup_taskset_first((tset)); (task); \ |
924f0d9a | 566 | (task) = cgroup_taskset_next((tset))) |
2f7ee569 | 567 | |
21acb9ca TLSC |
568 | /* |
569 | * Control Group subsystem type. | |
570 | * See Documentation/cgroups/cgroups.txt for details | |
571 | */ | |
ddbcc7e8 PM |
572 | |
573 | struct cgroup_subsys { | |
eb95419b TH |
574 | struct cgroup_subsys_state *(*css_alloc)(struct cgroup_subsys_state *parent_css); |
575 | int (*css_online)(struct cgroup_subsys_state *css); | |
576 | void (*css_offline)(struct cgroup_subsys_state *css); | |
577 | void (*css_free)(struct cgroup_subsys_state *css); | |
578 | ||
579 | int (*can_attach)(struct cgroup_subsys_state *css, | |
580 | struct cgroup_taskset *tset); | |
581 | void (*cancel_attach)(struct cgroup_subsys_state *css, | |
582 | struct cgroup_taskset *tset); | |
583 | void (*attach)(struct cgroup_subsys_state *css, | |
584 | struct cgroup_taskset *tset); | |
761b3ef5 | 585 | void (*fork)(struct task_struct *task); |
eb95419b TH |
586 | void (*exit)(struct cgroup_subsys_state *css, |
587 | struct cgroup_subsys_state *old_css, | |
761b3ef5 | 588 | struct task_struct *task); |
eb95419b | 589 | void (*bind)(struct cgroup_subsys_state *root_css); |
e5991371 | 590 | |
8bab8dde | 591 | int disabled; |
ddbcc7e8 | 592 | int early_init; |
48ddbe19 | 593 | |
8c7f6edb TH |
594 | /* |
595 | * If %false, this subsystem is properly hierarchical - | |
596 | * configuration, resource accounting and restriction on a parent | |
597 | * cgroup cover those of its children. If %true, hierarchy support | |
598 | * is broken in some ways - some subsystems ignore hierarchy | |
599 | * completely while others are only implemented half-way. | |
600 | * | |
601 | * It's now disallowed to create nested cgroups if the subsystem is | |
602 | * broken and cgroup core will emit a warning message on such | |
603 | * cases. Eventually, all subsystems will be made properly | |
604 | * hierarchical and this will go away. | |
605 | */ | |
606 | bool broken_hierarchy; | |
607 | bool warned_broken_hierarchy; | |
608 | ||
073219e9 | 609 | /* the following two fields are initialized automtically during boot */ |
aec25020 | 610 | int id; |
ddbcc7e8 PM |
611 | #define MAX_CGROUP_TYPE_NAMELEN 32 |
612 | const char *name; | |
613 | ||
b85d2040 | 614 | /* link to parent, protected by cgroup_lock() */ |
3dd06ffa | 615 | struct cgroup_root *root; |
e6a1105b | 616 | |
0adb0704 TH |
617 | /* |
618 | * List of cftypes. Each entry is the first entry of an array | |
619 | * terminated by zero length name. | |
620 | */ | |
621 | struct list_head cfts; | |
8e3f6541 | 622 | |
de00ffa5 | 623 | /* base cftypes, automatically registered with subsys itself */ |
8e3f6541 | 624 | struct cftype *base_cftypes; |
ddbcc7e8 PM |
625 | }; |
626 | ||
073219e9 | 627 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys; |
ddbcc7e8 PM |
628 | #include <linux/cgroup_subsys.h> |
629 | #undef SUBSYS | |
630 | ||
63876986 TH |
631 | /** |
632 | * css_parent - find the parent css | |
633 | * @css: the target cgroup_subsys_state | |
634 | * | |
635 | * Return the parent css of @css. This function is guaranteed to return | |
636 | * non-NULL parent as long as @css isn't the root. | |
637 | */ | |
638 | static inline | |
639 | struct cgroup_subsys_state *css_parent(struct cgroup_subsys_state *css) | |
640 | { | |
0ae78e0b | 641 | return css->parent; |
63876986 TH |
642 | } |
643 | ||
14611e51 TH |
644 | /** |
645 | * task_css_set_check - obtain a task's css_set with extra access conditions | |
646 | * @task: the task to obtain css_set for | |
647 | * @__c: extra condition expression to be passed to rcu_dereference_check() | |
648 | * | |
649 | * A task's css_set is RCU protected, initialized and exited while holding | |
650 | * task_lock(), and can only be modified while holding both cgroup_mutex | |
651 | * and task_lock() while the task is alive. This macro verifies that the | |
652 | * caller is inside proper critical section and returns @task's css_set. | |
653 | * | |
654 | * The caller can also specify additional allowed conditions via @__c, such | |
655 | * as locks used during the cgroup_subsys::attach() methods. | |
dc61b1d6 | 656 | */ |
2219449a TH |
657 | #ifdef CONFIG_PROVE_RCU |
658 | extern struct mutex cgroup_mutex; | |
0e1d768f | 659 | extern struct rw_semaphore css_set_rwsem; |
14611e51 TH |
660 | #define task_css_set_check(task, __c) \ |
661 | rcu_dereference_check((task)->cgroups, \ | |
0e1d768f TH |
662 | lockdep_is_held(&cgroup_mutex) || \ |
663 | lockdep_is_held(&css_set_rwsem) || \ | |
664 | ((task)->flags & PF_EXITING) || (__c)) | |
2219449a | 665 | #else |
14611e51 TH |
666 | #define task_css_set_check(task, __c) \ |
667 | rcu_dereference((task)->cgroups) | |
2219449a | 668 | #endif |
dc61b1d6 | 669 | |
14611e51 | 670 | /** |
8af01f56 | 671 | * task_css_check - obtain css for (task, subsys) w/ extra access conds |
14611e51 TH |
672 | * @task: the target task |
673 | * @subsys_id: the target subsystem ID | |
674 | * @__c: extra condition expression to be passed to rcu_dereference_check() | |
675 | * | |
676 | * Return the cgroup_subsys_state for the (@task, @subsys_id) pair. The | |
677 | * synchronization rules are the same as task_css_set_check(). | |
678 | */ | |
8af01f56 | 679 | #define task_css_check(task, subsys_id, __c) \ |
14611e51 TH |
680 | task_css_set_check((task), (__c))->subsys[(subsys_id)] |
681 | ||
682 | /** | |
683 | * task_css_set - obtain a task's css_set | |
684 | * @task: the task to obtain css_set for | |
685 | * | |
686 | * See task_css_set_check(). | |
687 | */ | |
688 | static inline struct css_set *task_css_set(struct task_struct *task) | |
689 | { | |
690 | return task_css_set_check(task, false); | |
691 | } | |
692 | ||
693 | /** | |
8af01f56 | 694 | * task_css - obtain css for (task, subsys) |
14611e51 TH |
695 | * @task: the target task |
696 | * @subsys_id: the target subsystem ID | |
697 | * | |
8af01f56 | 698 | * See task_css_check(). |
14611e51 | 699 | */ |
8af01f56 TH |
700 | static inline struct cgroup_subsys_state *task_css(struct task_struct *task, |
701 | int subsys_id) | |
ddbcc7e8 | 702 | { |
8af01f56 | 703 | return task_css_check(task, subsys_id, false); |
ddbcc7e8 PM |
704 | } |
705 | ||
8af01f56 TH |
706 | static inline struct cgroup *task_cgroup(struct task_struct *task, |
707 | int subsys_id) | |
ddbcc7e8 | 708 | { |
8af01f56 | 709 | return task_css(task, subsys_id)->cgroup; |
ddbcc7e8 PM |
710 | } |
711 | ||
492eb21b TH |
712 | struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos, |
713 | struct cgroup_subsys_state *parent); | |
53fa5261 | 714 | |
1cb650b9 LZ |
715 | struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss); |
716 | ||
574bd9f7 | 717 | /** |
492eb21b TH |
718 | * css_for_each_child - iterate through children of a css |
719 | * @pos: the css * to use as the loop cursor | |
720 | * @parent: css whose children to walk | |
574bd9f7 | 721 | * |
492eb21b TH |
722 | * Walk @parent's children. Must be called under rcu_read_lock(). A child |
723 | * css which hasn't finished ->css_online() or already has finished | |
92fb9748 | 724 | * ->css_offline() may show up during traversal and it's each subsystem's |
574bd9f7 TH |
725 | * responsibility to verify that each @pos is alive. |
726 | * | |
92fb9748 | 727 | * If a subsystem synchronizes against the parent in its ->css_online() and |
492eb21b | 728 | * before starting iterating, a css which finished ->css_online() is |
92fb9748 | 729 | * guaranteed to be visible in the future iterations. |
75501a6d TH |
730 | * |
731 | * It is allowed to temporarily drop RCU read lock during iteration. The | |
732 | * caller is responsible for ensuring that @pos remains accessible until | |
733 | * the start of the next iteration by, for example, bumping the css refcnt. | |
574bd9f7 | 734 | */ |
492eb21b TH |
735 | #define css_for_each_child(pos, parent) \ |
736 | for ((pos) = css_next_child(NULL, (parent)); (pos); \ | |
737 | (pos) = css_next_child((pos), (parent))) | |
574bd9f7 | 738 | |
492eb21b TH |
739 | struct cgroup_subsys_state * |
740 | css_next_descendant_pre(struct cgroup_subsys_state *pos, | |
741 | struct cgroup_subsys_state *css); | |
742 | ||
743 | struct cgroup_subsys_state * | |
744 | css_rightmost_descendant(struct cgroup_subsys_state *pos); | |
574bd9f7 TH |
745 | |
746 | /** | |
492eb21b TH |
747 | * css_for_each_descendant_pre - pre-order walk of a css's descendants |
748 | * @pos: the css * to use as the loop cursor | |
749 | * @root: css whose descendants to walk | |
574bd9f7 | 750 | * |
bd8815a6 TH |
751 | * Walk @root's descendants. @root is included in the iteration and the |
752 | * first node to be visited. Must be called under rcu_read_lock(). A | |
492eb21b | 753 | * descendant css which hasn't finished ->css_online() or already has |
92fb9748 | 754 | * finished ->css_offline() may show up during traversal and it's each |
574bd9f7 TH |
755 | * subsystem's responsibility to verify that each @pos is alive. |
756 | * | |
92fb9748 TH |
757 | * If a subsystem synchronizes against the parent in its ->css_online() and |
758 | * before starting iterating, and synchronizes against @pos on each | |
492eb21b | 759 | * iteration, any descendant css which finished ->css_online() is |
574bd9f7 TH |
760 | * guaranteed to be visible in the future iterations. |
761 | * | |
762 | * In other words, the following guarantees that a descendant can't escape | |
763 | * state updates of its ancestors. | |
764 | * | |
492eb21b | 765 | * my_online(@css) |
574bd9f7 | 766 | * { |
492eb21b TH |
767 | * Lock @css's parent and @css; |
768 | * Inherit state from the parent; | |
574bd9f7 TH |
769 | * Unlock both. |
770 | * } | |
771 | * | |
492eb21b | 772 | * my_update_state(@css) |
574bd9f7 | 773 | * { |
492eb21b | 774 | * css_for_each_descendant_pre(@pos, @css) { |
574bd9f7 | 775 | * Lock @pos; |
bd8815a6 TH |
776 | * if (@pos == @css) |
777 | * Update @css's state; | |
778 | * else | |
779 | * Verify @pos is alive and inherit state from its parent; | |
574bd9f7 TH |
780 | * Unlock @pos; |
781 | * } | |
782 | * } | |
783 | * | |
784 | * As long as the inheriting step, including checking the parent state, is | |
785 | * enclosed inside @pos locking, double-locking the parent isn't necessary | |
786 | * while inheriting. The state update to the parent is guaranteed to be | |
787 | * visible by walking order and, as long as inheriting operations to the | |
788 | * same @pos are atomic to each other, multiple updates racing each other | |
789 | * still result in the correct state. It's guaranateed that at least one | |
492eb21b | 790 | * inheritance happens for any css after the latest update to its parent. |
574bd9f7 TH |
791 | * |
792 | * If checking parent's state requires locking the parent, each inheriting | |
793 | * iteration should lock and unlock both @pos->parent and @pos. | |
794 | * | |
795 | * Alternatively, a subsystem may choose to use a single global lock to | |
92fb9748 | 796 | * synchronize ->css_online() and ->css_offline() against tree-walking |
574bd9f7 | 797 | * operations. |
75501a6d TH |
798 | * |
799 | * It is allowed to temporarily drop RCU read lock during iteration. The | |
800 | * caller is responsible for ensuring that @pos remains accessible until | |
801 | * the start of the next iteration by, for example, bumping the css refcnt. | |
574bd9f7 | 802 | */ |
492eb21b TH |
803 | #define css_for_each_descendant_pre(pos, css) \ |
804 | for ((pos) = css_next_descendant_pre(NULL, (css)); (pos); \ | |
805 | (pos) = css_next_descendant_pre((pos), (css))) | |
574bd9f7 | 806 | |
492eb21b TH |
807 | struct cgroup_subsys_state * |
808 | css_next_descendant_post(struct cgroup_subsys_state *pos, | |
809 | struct cgroup_subsys_state *css); | |
574bd9f7 TH |
810 | |
811 | /** | |
492eb21b TH |
812 | * css_for_each_descendant_post - post-order walk of a css's descendants |
813 | * @pos: the css * to use as the loop cursor | |
814 | * @css: css whose descendants to walk | |
574bd9f7 | 815 | * |
492eb21b | 816 | * Similar to css_for_each_descendant_pre() but performs post-order |
bd8815a6 TH |
817 | * traversal instead. @root is included in the iteration and the last |
818 | * node to be visited. Note that the walk visibility guarantee described | |
819 | * in pre-order walk doesn't apply the same to post-order walks. | |
574bd9f7 | 820 | */ |
492eb21b TH |
821 | #define css_for_each_descendant_post(pos, css) \ |
822 | for ((pos) = css_next_descendant_post(NULL, (css)); (pos); \ | |
823 | (pos) = css_next_descendant_post((pos), (css))) | |
574bd9f7 | 824 | |
72ec7029 TH |
825 | /* A css_task_iter should be treated as an opaque object */ |
826 | struct css_task_iter { | |
827 | struct cgroup_subsys_state *origin_css; | |
0942eeee TH |
828 | struct list_head *cset_link; |
829 | struct list_head *task; | |
817929ec PM |
830 | }; |
831 | ||
72ec7029 TH |
832 | void css_task_iter_start(struct cgroup_subsys_state *css, |
833 | struct css_task_iter *it); | |
834 | struct task_struct *css_task_iter_next(struct css_task_iter *it); | |
835 | void css_task_iter_end(struct css_task_iter *it); | |
e535837b | 836 | |
31583bb0 | 837 | int cgroup_attach_task_all(struct task_struct *from, struct task_struct *); |
8cc99345 | 838 | int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from); |
31583bb0 | 839 | |
5a17f543 TH |
840 | struct cgroup_subsys_state *css_tryget_from_dir(struct dentry *dentry, |
841 | struct cgroup_subsys *ss); | |
38460b48 | 842 | |
ddbcc7e8 PM |
843 | #else /* !CONFIG_CGROUPS */ |
844 | ||
845 | static inline int cgroup_init_early(void) { return 0; } | |
846 | static inline int cgroup_init(void) { return 0; } | |
b4f48b63 | 847 | static inline void cgroup_fork(struct task_struct *p) {} |
817929ec | 848 | static inline void cgroup_post_fork(struct task_struct *p) {} |
1ec41830 | 849 | static inline void cgroup_exit(struct task_struct *p) {} |
ddbcc7e8 | 850 | |
846c7bb0 BS |
851 | static inline int cgroupstats_build(struct cgroupstats *stats, |
852 | struct dentry *dentry) | |
853 | { | |
854 | return -EINVAL; | |
855 | } | |
ddbcc7e8 | 856 | |
d7926ee3 | 857 | /* No cgroups - nothing to do */ |
31583bb0 MT |
858 | static inline int cgroup_attach_task_all(struct task_struct *from, |
859 | struct task_struct *t) | |
860 | { | |
861 | return 0; | |
862 | } | |
d7926ee3 | 863 | |
ddbcc7e8 PM |
864 | #endif /* !CONFIG_CGROUPS */ |
865 | ||
866 | #endif /* _LINUX_CGROUP_H */ |