]>
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> |
31a7df01 | 17 | #include <linux/prio_heap.h> |
cc31edce | 18 | #include <linux/rwsem.h> |
38460b48 | 19 | #include <linux/idr.h> |
48ddbe19 | 20 | #include <linux/workqueue.h> |
03b1cde6 | 21 | #include <linux/xattr.h> |
25a7e684 | 22 | #include <linux/fs.h> |
d3daf28d | 23 | #include <linux/percpu-refcount.h> |
7da11279 | 24 | #include <linux/seq_file.h> |
ddbcc7e8 PM |
25 | |
26 | #ifdef CONFIG_CGROUPS | |
27 | ||
28 | struct cgroupfs_root; | |
29 | struct cgroup_subsys; | |
30 | struct inode; | |
84eea842 | 31 | struct cgroup; |
38460b48 | 32 | struct css_id; |
ddbcc7e8 PM |
33 | |
34 | extern int cgroup_init_early(void); | |
35 | extern int cgroup_init(void); | |
b4f48b63 | 36 | extern void cgroup_fork(struct task_struct *p); |
817929ec | 37 | extern void cgroup_post_fork(struct task_struct *p); |
b4f48b63 | 38 | extern void cgroup_exit(struct task_struct *p, int run_callbacks); |
846c7bb0 BS |
39 | extern int cgroupstats_build(struct cgroupstats *stats, |
40 | struct dentry *dentry); | |
e6a1105b | 41 | extern int cgroup_load_subsys(struct cgroup_subsys *ss); |
cf5d5941 | 42 | extern void cgroup_unload_subsys(struct cgroup_subsys *ss); |
ddbcc7e8 | 43 | |
8d8b97ba | 44 | extern int proc_cgroup_show(struct seq_file *, void *); |
a424316c | 45 | |
7d8e0bf5 LZ |
46 | /* |
47 | * Define the enumeration of all cgroup subsystems. | |
48 | * | |
49 | * We define ids for builtin subsystems and then modular ones. | |
50 | */ | |
817929ec PM |
51 | #define SUBSYS(_x) _x ## _subsys_id, |
52 | enum cgroup_subsys_id { | |
7d8e0bf5 | 53 | #define IS_SUBSYS_ENABLED(option) IS_BUILTIN(option) |
817929ec | 54 | #include <linux/cgroup_subsys.h> |
7d8e0bf5 LZ |
55 | #undef IS_SUBSYS_ENABLED |
56 | CGROUP_BUILTIN_SUBSYS_COUNT, | |
57 | ||
58 | __CGROUP_SUBSYS_TEMP_PLACEHOLDER = CGROUP_BUILTIN_SUBSYS_COUNT - 1, | |
59 | ||
60 | #define IS_SUBSYS_ENABLED(option) IS_MODULE(option) | |
817929ec | 61 | #include <linux/cgroup_subsys.h> |
7d8e0bf5 | 62 | #undef IS_SUBSYS_ENABLED |
a6f00298 | 63 | CGROUP_SUBSYS_COUNT, |
817929ec PM |
64 | }; |
65 | #undef SUBSYS | |
66 | ||
ddbcc7e8 PM |
67 | /* Per-subsystem/per-cgroup state maintained by the system. */ |
68 | struct cgroup_subsys_state { | |
72c97e54 | 69 | /* the cgroup that this css is attached to */ |
ddbcc7e8 PM |
70 | struct cgroup *cgroup; |
71 | ||
72c97e54 TH |
72 | /* the cgroup subsystem that this css is attached to */ |
73 | struct cgroup_subsys *ss; | |
74 | ||
d3daf28d TH |
75 | /* reference count - access via css_[try]get() and css_put() */ |
76 | struct percpu_ref refcnt; | |
ddbcc7e8 | 77 | |
0ae78e0b TH |
78 | /* the parent css */ |
79 | struct cgroup_subsys_state *parent; | |
80 | ||
ddbcc7e8 | 81 | unsigned long flags; |
38460b48 | 82 | /* ID for this css, if possible */ |
2c392b8c | 83 | struct css_id __rcu *id; |
48ddbe19 | 84 | |
0c21ead1 TH |
85 | /* percpu_ref killing and RCU release */ |
86 | struct rcu_head rcu_head; | |
35ef10da | 87 | struct work_struct destroy_work; |
ddbcc7e8 PM |
88 | }; |
89 | ||
90 | /* bits in struct cgroup_subsys_state flags field */ | |
91 | enum { | |
38b53aba | 92 | CSS_ROOT = (1 << 0), /* this CSS is the root of the subsystem */ |
92fb9748 | 93 | CSS_ONLINE = (1 << 1), /* between ->css_online() and ->css_offline() */ |
ddbcc7e8 PM |
94 | }; |
95 | ||
5de0107e TH |
96 | /** |
97 | * css_get - obtain a reference on the specified css | |
98 | * @css: target css | |
99 | * | |
100 | * The caller must already have a reference. | |
ddbcc7e8 | 101 | */ |
ddbcc7e8 PM |
102 | static inline void css_get(struct cgroup_subsys_state *css) |
103 | { | |
104 | /* We don't need to reference count the root state */ | |
38b53aba | 105 | if (!(css->flags & CSS_ROOT)) |
d3daf28d | 106 | percpu_ref_get(&css->refcnt); |
ddbcc7e8 | 107 | } |
e7c5ec91 | 108 | |
5de0107e TH |
109 | /** |
110 | * css_tryget - try to obtain a reference on the specified css | |
111 | * @css: target css | |
112 | * | |
113 | * Obtain a reference on @css if it's alive. The caller naturally needs to | |
114 | * ensure that @css is accessible but doesn't have to be holding a | |
115 | * reference on it - IOW, RCU protected access is good enough for this | |
116 | * function. Returns %true if a reference count was successfully obtained; | |
117 | * %false otherwise. | |
118 | */ | |
e7c5ec91 PM |
119 | static inline bool css_tryget(struct cgroup_subsys_state *css) |
120 | { | |
38b53aba | 121 | if (css->flags & CSS_ROOT) |
e7c5ec91 | 122 | return true; |
d3daf28d | 123 | return percpu_ref_tryget(&css->refcnt); |
e7c5ec91 PM |
124 | } |
125 | ||
5de0107e TH |
126 | /** |
127 | * css_put - put a css reference | |
128 | * @css: target css | |
129 | * | |
130 | * Put a reference obtained via css_get() and css_tryget(). | |
131 | */ | |
ddbcc7e8 PM |
132 | static inline void css_put(struct cgroup_subsys_state *css) |
133 | { | |
38b53aba | 134 | if (!(css->flags & CSS_ROOT)) |
d3daf28d | 135 | percpu_ref_put(&css->refcnt); |
ddbcc7e8 PM |
136 | } |
137 | ||
3116f0e3 PM |
138 | /* bits in struct cgroup flags field */ |
139 | enum { | |
140 | /* Control Group is dead */ | |
54766d4a | 141 | CGRP_DEAD, |
d20a390a PM |
142 | /* |
143 | * Control Group has previously had a child cgroup or a task, | |
144 | * but no longer (only if CGRP_NOTIFY_ON_RELEASE is set) | |
145 | */ | |
3116f0e3 PM |
146 | CGRP_RELEASABLE, |
147 | /* Control Group requires release notifications to userspace */ | |
148 | CGRP_NOTIFY_ON_RELEASE, | |
97978e6d | 149 | /* |
2260e7fc TH |
150 | * Clone the parent's configuration when creating a new child |
151 | * cpuset cgroup. For historical reasons, this option can be | |
152 | * specified at mount time and thus is implemented here. | |
97978e6d | 153 | */ |
2260e7fc | 154 | CGRP_CPUSET_CLONE_CHILDREN, |
873fe09e TH |
155 | /* see the comment above CGRP_ROOT_SANE_BEHAVIOR for details */ |
156 | CGRP_SANE_BEHAVIOR, | |
3116f0e3 PM |
157 | }; |
158 | ||
65dff759 LZ |
159 | struct cgroup_name { |
160 | struct rcu_head rcu_head; | |
161 | char name[]; | |
3116f0e3 PM |
162 | }; |
163 | ||
ddbcc7e8 PM |
164 | struct cgroup { |
165 | unsigned long flags; /* "unsigned long" so bitops work */ | |
166 | ||
b414dc09 LZ |
167 | /* |
168 | * idr allocated in-hierarchy ID. | |
169 | * | |
170 | * The ID of the root cgroup is always 0, and a new cgroup | |
171 | * will be assigned with a smallest available ID. | |
172 | */ | |
173 | int id; | |
0a950f65 | 174 | |
f20104de TH |
175 | /* the number of attached css's */ |
176 | int nr_css; | |
177 | ||
ddbcc7e8 PM |
178 | /* |
179 | * We link our 'sibling' struct into our parent's 'children'. | |
180 | * Our children link their 'sibling' into our 'children'. | |
181 | */ | |
182 | struct list_head sibling; /* my parent's children */ | |
183 | struct list_head children; /* my children */ | |
05ef1d7c | 184 | struct list_head files; /* my files */ |
ddbcc7e8 | 185 | |
d20a390a | 186 | struct cgroup *parent; /* my parent */ |
febfcef6 | 187 | struct dentry *dentry; /* cgroup fs entry, RCU protected */ |
ddbcc7e8 | 188 | |
53fa5261 TH |
189 | /* |
190 | * Monotonically increasing unique serial number which defines a | |
191 | * uniform order among all cgroups. It's guaranteed that all | |
192 | * ->children lists are in the ascending order of ->serial_nr. | |
193 | * It's used to allow interrupting and resuming iterations. | |
194 | */ | |
195 | u64 serial_nr; | |
196 | ||
65dff759 LZ |
197 | /* |
198 | * This is a copy of dentry->d_name, and it's needed because | |
199 | * we can't use dentry->d_name in cgroup_path(). | |
200 | * | |
201 | * You must acquire rcu_read_lock() to access cgrp->name, and | |
202 | * the only place that can change it is rename(), which is | |
203 | * protected by parent dir's i_mutex. | |
204 | * | |
205 | * Normally you should use cgroup_name() wrapper rather than | |
206 | * access it directly. | |
207 | */ | |
208 | struct cgroup_name __rcu *name; | |
209 | ||
ddbcc7e8 | 210 | /* Private pointers for each registered subsystem */ |
73e80ed8 | 211 | struct cgroup_subsys_state __rcu *subsys[CGROUP_SUBSYS_COUNT]; |
ddbcc7e8 PM |
212 | |
213 | struct cgroupfs_root *root; | |
817929ec PM |
214 | |
215 | /* | |
69d0206c TH |
216 | * List of cgrp_cset_links pointing at css_sets with tasks in this |
217 | * cgroup. Protected by css_set_lock. | |
817929ec | 218 | */ |
69d0206c | 219 | struct list_head cset_links; |
81a6a5cd PM |
220 | |
221 | /* | |
222 | * Linked list running through all cgroups that can | |
223 | * potentially be reaped by the release agent. Protected by | |
224 | * release_list_lock | |
225 | */ | |
226 | struct list_head release_list; | |
cc31edce | 227 | |
72a8cb30 BB |
228 | /* |
229 | * list of pidlists, up to two for each namespace (one for procs, one | |
230 | * for tasks); created on demand. | |
231 | */ | |
232 | struct list_head pidlists; | |
233 | struct mutex pidlist_mutex; | |
a47295e6 | 234 | |
67f4c36f TH |
235 | /* dummy css with NULL ->ss, points back to this cgroup */ |
236 | struct cgroup_subsys_state dummy_css; | |
237 | ||
d3daf28d | 238 | /* For css percpu_ref killing and RCU-protected deletion */ |
a47295e6 | 239 | struct rcu_head rcu_head; |
ea15f8cc | 240 | struct work_struct destroy_work; |
0dea1168 | 241 | |
03b1cde6 AR |
242 | /* directory xattrs */ |
243 | struct simple_xattrs xattrs; | |
817929ec PM |
244 | }; |
245 | ||
25a7e684 TH |
246 | #define MAX_CGROUP_ROOT_NAMELEN 64 |
247 | ||
248 | /* cgroupfs_root->flags */ | |
249 | enum { | |
873fe09e TH |
250 | /* |
251 | * Unfortunately, cgroup core and various controllers are riddled | |
252 | * with idiosyncrasies and pointless options. The following flag, | |
253 | * when set, will force sane behavior - some options are forced on, | |
254 | * others are disallowed, and some controllers will change their | |
255 | * hierarchical or other behaviors. | |
256 | * | |
257 | * The set of behaviors affected by this flag are still being | |
258 | * determined and developed and the mount option for this flag is | |
259 | * prefixed with __DEVEL__. The prefix will be dropped once we | |
260 | * reach the point where all behaviors are compatible with the | |
261 | * planned unified hierarchy, which will automatically turn on this | |
262 | * flag. | |
263 | * | |
264 | * The followings are the behaviors currently affected this flag. | |
265 | * | |
266 | * - Mount options "noprefix" and "clone_children" are disallowed. | |
267 | * Also, cgroupfs file cgroup.clone_children is not created. | |
268 | * | |
269 | * - When mounting an existing superblock, mount options should | |
270 | * match. | |
271 | * | |
272 | * - Remount is disallowed. | |
273 | * | |
0b0585c3 LT |
274 | * - rename(2) is disallowed. |
275 | * | |
f63674fd TH |
276 | * - "tasks" is removed. Everything should be at process |
277 | * granularity. Use "cgroup.procs" instead. | |
f00baae7 | 278 | * |
afb2bc14 TH |
279 | * - "cgroup.procs" is not sorted. pids will be unique unless they |
280 | * got recycled inbetween reads. | |
281 | * | |
f63674fd TH |
282 | * - "release_agent" and "notify_on_release" are removed. |
283 | * Replacement notification mechanism will be implemented. | |
873fe09e | 284 | * |
5c5cc623 LZ |
285 | * - cpuset: tasks will be kept in empty cpusets when hotplug happens |
286 | * and take masks of ancestors with non-empty cpus/mems, instead of | |
287 | * being moved to an ancestor. | |
288 | * | |
88fa523b LZ |
289 | * - cpuset: a task can be moved into an empty cpuset, and again it |
290 | * takes masks of ancestors. | |
6db8e85c | 291 | * |
f63674fd TH |
292 | * - memcg: use_hierarchy is on by default and the cgroup file for |
293 | * the flag is not created. | |
f00baae7 | 294 | * |
9138125b | 295 | * - blkcg: blk-throttle becomes properly hierarchical. |
873fe09e TH |
296 | */ |
297 | CGRP_ROOT_SANE_BEHAVIOR = (1 << 0), | |
298 | ||
25a7e684 TH |
299 | CGRP_ROOT_NOPREFIX = (1 << 1), /* mounted subsystems have no named prefix */ |
300 | CGRP_ROOT_XATTR = (1 << 2), /* supports extended attributes */ | |
0ce6cba3 TH |
301 | |
302 | /* mount options live below bit 16 */ | |
303 | CGRP_ROOT_OPTION_MASK = (1 << 16) - 1, | |
304 | ||
305 | CGRP_ROOT_SUBSYS_BOUND = (1 << 16), /* subsystems finished binding */ | |
25a7e684 TH |
306 | }; |
307 | ||
308 | /* | |
309 | * A cgroupfs_root represents the root of a cgroup hierarchy, and may be | |
310 | * associated with a superblock to form an active hierarchy. This is | |
311 | * internal to cgroup core. Don't access directly from controllers. | |
312 | */ | |
313 | struct cgroupfs_root { | |
314 | struct super_block *sb; | |
315 | ||
a8a648c4 | 316 | /* The bitmask of subsystems attached to this hierarchy */ |
25a7e684 TH |
317 | unsigned long subsys_mask; |
318 | ||
319 | /* Unique id for this hierarchy. */ | |
320 | int hierarchy_id; | |
321 | ||
25a7e684 TH |
322 | /* A list running through the attached subsystems */ |
323 | struct list_head subsys_list; | |
324 | ||
325 | /* The root cgroup for this hierarchy */ | |
326 | struct cgroup top_cgroup; | |
327 | ||
328 | /* Tracks how many cgroups are currently defined in hierarchy.*/ | |
329 | int number_of_cgroups; | |
330 | ||
331 | /* A list running through the active hierarchies */ | |
332 | struct list_head root_list; | |
333 | ||
25a7e684 TH |
334 | /* Hierarchy-specific flags */ |
335 | unsigned long flags; | |
336 | ||
337 | /* IDs for cgroups in this hierarchy */ | |
4e96ee8e | 338 | struct idr cgroup_idr; |
25a7e684 TH |
339 | |
340 | /* The path to use for release notifications. */ | |
341 | char release_agent_path[PATH_MAX]; | |
342 | ||
343 | /* The name for this hierarchy - may be empty */ | |
344 | char name[MAX_CGROUP_ROOT_NAMELEN]; | |
345 | }; | |
346 | ||
d20a390a PM |
347 | /* |
348 | * A css_set is a structure holding pointers to a set of | |
817929ec PM |
349 | * cgroup_subsys_state objects. This saves space in the task struct |
350 | * object and speeds up fork()/exit(), since a single inc/dec and a | |
d20a390a PM |
351 | * list_add()/del() can bump the reference count on the entire cgroup |
352 | * set for a task. | |
817929ec PM |
353 | */ |
354 | ||
355 | struct css_set { | |
356 | ||
357 | /* Reference count */ | |
146aa1bd | 358 | atomic_t refcount; |
817929ec | 359 | |
472b1053 LZ |
360 | /* |
361 | * List running through all cgroup groups in the same hash | |
362 | * slot. Protected by css_set_lock | |
363 | */ | |
364 | struct hlist_node hlist; | |
365 | ||
817929ec PM |
366 | /* |
367 | * List running through all tasks using this cgroup | |
368 | * group. Protected by css_set_lock | |
369 | */ | |
370 | struct list_head tasks; | |
371 | ||
372 | /* | |
69d0206c TH |
373 | * List of cgrp_cset_links pointing at cgroups referenced from this |
374 | * css_set. Protected by css_set_lock. | |
817929ec | 375 | */ |
69d0206c | 376 | struct list_head cgrp_links; |
817929ec PM |
377 | |
378 | /* | |
379 | * Set of subsystem states, one for each subsystem. This array | |
380 | * is immutable after creation apart from the init_css_set | |
cf5d5941 BB |
381 | * during subsystem registration (at boot time) and modular subsystem |
382 | * loading/unloading. | |
817929ec PM |
383 | */ |
384 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
c378369d BB |
385 | |
386 | /* For RCU-protected deletion */ | |
387 | struct rcu_head rcu_head; | |
ddbcc7e8 PM |
388 | }; |
389 | ||
d20a390a PM |
390 | /* |
391 | * struct cftype: handler definitions for cgroup control files | |
ddbcc7e8 PM |
392 | * |
393 | * When reading/writing to a file: | |
a043e3b2 | 394 | * - the cgroup to use is file->f_dentry->d_parent->d_fsdata |
ddbcc7e8 PM |
395 | * - the 'cftype' of the file is file->f_dentry->d_fsdata |
396 | */ | |
397 | ||
8e3f6541 | 398 | /* cftype->flags */ |
02c402d9 | 399 | enum { |
6f4b7e63 LZ |
400 | CFTYPE_ONLY_ON_ROOT = (1 << 0), /* only create on root cgrp */ |
401 | CFTYPE_NOT_ON_ROOT = (1 << 1), /* don't create on root cgrp */ | |
02c402d9 | 402 | CFTYPE_INSANE = (1 << 2), /* don't create if sane_behavior */ |
9fa4db33 | 403 | CFTYPE_NO_PREFIX = (1 << 3), /* (DON'T USE FOR NEW FILES) no subsys prefix */ |
02c402d9 | 404 | }; |
8e3f6541 TH |
405 | |
406 | #define MAX_CFTYPE_NAME 64 | |
407 | ||
ddbcc7e8 | 408 | struct cftype { |
d20a390a PM |
409 | /* |
410 | * By convention, the name should begin with the name of the | |
8e3f6541 TH |
411 | * subsystem, followed by a period. Zero length string indicates |
412 | * end of cftype array. | |
d20a390a | 413 | */ |
ddbcc7e8 PM |
414 | char name[MAX_CFTYPE_NAME]; |
415 | int private; | |
099fca32 LZ |
416 | /* |
417 | * If not 0, file mode is set to this value, otherwise it will | |
418 | * be figured out automatically | |
419 | */ | |
a5e7ed32 | 420 | umode_t mode; |
db3b1497 PM |
421 | |
422 | /* | |
423 | * If non-zero, defines the maximum length of string that can | |
424 | * be passed to write_string; defaults to 64 | |
425 | */ | |
426 | size_t max_write_len; | |
427 | ||
8e3f6541 TH |
428 | /* CFTYPE_* flags */ |
429 | unsigned int flags; | |
430 | ||
2bb566cb TH |
431 | /* |
432 | * The subsys this file belongs to. Initialized automatically | |
433 | * during registration. NULL for cgroup core files. | |
434 | */ | |
435 | struct cgroup_subsys *ss; | |
436 | ||
ddbcc7e8 | 437 | /* |
f4c753b7 | 438 | * read_u64() is a shortcut for the common case of returning a |
ddbcc7e8 PM |
439 | * single integer. Use it in place of read() |
440 | */ | |
182446d0 | 441 | u64 (*read_u64)(struct cgroup_subsys_state *css, struct cftype *cft); |
e73d2c61 PM |
442 | /* |
443 | * read_s64() is a signed version of read_u64() | |
444 | */ | |
182446d0 | 445 | s64 (*read_s64)(struct cgroup_subsys_state *css, struct cftype *cft); |
2da8ca82 TH |
446 | |
447 | /* generic seq_file read interface */ | |
448 | int (*seq_show)(struct seq_file *sf, void *v); | |
91796569 | 449 | |
6612f05b TH |
450 | /* optional ops, implement all or none */ |
451 | void *(*seq_start)(struct seq_file *sf, loff_t *ppos); | |
452 | void *(*seq_next)(struct seq_file *sf, void *v, loff_t *ppos); | |
453 | void (*seq_stop)(struct seq_file *sf, void *v); | |
454 | ||
355e0c48 | 455 | /* |
f4c753b7 | 456 | * write_u64() is a shortcut for the common case of accepting |
355e0c48 PM |
457 | * a single integer (as parsed by simple_strtoull) from |
458 | * userspace. Use in place of write(); return 0 or error. | |
459 | */ | |
182446d0 TH |
460 | int (*write_u64)(struct cgroup_subsys_state *css, struct cftype *cft, |
461 | u64 val); | |
e73d2c61 PM |
462 | /* |
463 | * write_s64() is a signed version of write_u64() | |
464 | */ | |
182446d0 TH |
465 | int (*write_s64)(struct cgroup_subsys_state *css, struct cftype *cft, |
466 | s64 val); | |
355e0c48 | 467 | |
db3b1497 PM |
468 | /* |
469 | * write_string() is passed a nul-terminated kernelspace | |
470 | * buffer of maximum length determined by max_write_len. | |
471 | * Returns 0 or -ve error code. | |
472 | */ | |
182446d0 | 473 | int (*write_string)(struct cgroup_subsys_state *css, struct cftype *cft, |
db3b1497 | 474 | const char *buffer); |
d447ea2f PE |
475 | /* |
476 | * trigger() callback can be used to get some kick from the | |
477 | * userspace, when the actual string written is not important | |
478 | * at all. The private field can be used to determine the | |
479 | * kick type for multiplexing. | |
480 | */ | |
182446d0 | 481 | int (*trigger)(struct cgroup_subsys_state *css, unsigned int event); |
ddbcc7e8 PM |
482 | }; |
483 | ||
8e3f6541 TH |
484 | /* |
485 | * cftype_sets describe cftypes belonging to a subsystem and are chained at | |
486 | * cgroup_subsys->cftsets. Each cftset points to an array of cftypes | |
487 | * terminated by zero length name. | |
488 | */ | |
489 | struct cftype_set { | |
490 | struct list_head node; /* chained at subsys->cftsets */ | |
03b1cde6 | 491 | struct cftype *cfts; |
8e3f6541 TH |
492 | }; |
493 | ||
7da11279 TH |
494 | /* |
495 | * cgroupfs file entry, pointed to from leaf dentry->d_fsdata. Don't | |
496 | * access directly. | |
497 | */ | |
498 | struct cfent { | |
499 | struct list_head node; | |
500 | struct dentry *dentry; | |
501 | struct cftype *type; | |
502 | struct cgroup_subsys_state *css; | |
503 | ||
504 | /* file xattrs */ | |
505 | struct simple_xattrs xattrs; | |
506 | }; | |
507 | ||
508 | /* seq_file->private points to the following, only ->priv is public */ | |
509 | struct cgroup_open_file { | |
510 | struct cfent *cfe; | |
511 | void *priv; | |
512 | }; | |
513 | ||
873fe09e TH |
514 | /* |
515 | * See the comment above CGRP_ROOT_SANE_BEHAVIOR for details. This | |
516 | * function can be called as long as @cgrp is accessible. | |
517 | */ | |
518 | static inline bool cgroup_sane_behavior(const struct cgroup *cgrp) | |
519 | { | |
520 | return cgrp->root->flags & CGRP_ROOT_SANE_BEHAVIOR; | |
521 | } | |
522 | ||
65dff759 LZ |
523 | /* Caller should hold rcu_read_lock() */ |
524 | static inline const char *cgroup_name(const struct cgroup *cgrp) | |
525 | { | |
526 | return rcu_dereference(cgrp->name)->name; | |
527 | } | |
528 | ||
7da11279 TH |
529 | static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq) |
530 | { | |
531 | struct cgroup_open_file *of = seq->private; | |
532 | return of->cfe->css; | |
533 | } | |
534 | ||
535 | static inline struct cftype *seq_cft(struct seq_file *seq) | |
536 | { | |
537 | struct cgroup_open_file *of = seq->private; | |
538 | return of->cfe->type; | |
539 | } | |
540 | ||
03b1cde6 | 541 | int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); |
2bb566cb | 542 | int cgroup_rm_cftypes(struct cftype *cfts); |
8e3f6541 | 543 | |
78574cf9 | 544 | bool cgroup_is_descendant(struct cgroup *cgrp, struct cgroup *ancestor); |
ddbcc7e8 | 545 | |
ffd2d883 | 546 | int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen); |
913ffdb5 | 547 | int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen); |
ddbcc7e8 | 548 | |
ffd2d883 | 549 | int cgroup_task_count(const struct cgroup *cgrp); |
bbcb81d0 | 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); | |
d99c8727 TH |
558 | struct cgroup_subsys_state *cgroup_taskset_cur_css(struct cgroup_taskset *tset, |
559 | int subsys_id); | |
2f7ee569 TH |
560 | int cgroup_taskset_size(struct cgroup_taskset *tset); |
561 | ||
562 | /** | |
563 | * cgroup_taskset_for_each - iterate cgroup_taskset | |
564 | * @task: the loop cursor | |
d99c8727 | 565 | * @skip_css: skip if task's css matches this, %NULL to iterate through all |
2f7ee569 TH |
566 | * @tset: taskset to iterate |
567 | */ | |
d99c8727 | 568 | #define cgroup_taskset_for_each(task, skip_css, tset) \ |
2f7ee569 TH |
569 | for ((task) = cgroup_taskset_first((tset)); (task); \ |
570 | (task) = cgroup_taskset_next((tset))) \ | |
d99c8727 TH |
571 | if (!(skip_css) || \ |
572 | cgroup_taskset_cur_css((tset), \ | |
573 | (skip_css)->ss->subsys_id) != (skip_css)) | |
2f7ee569 | 574 | |
21acb9ca TLSC |
575 | /* |
576 | * Control Group subsystem type. | |
577 | * See Documentation/cgroups/cgroups.txt for details | |
578 | */ | |
ddbcc7e8 PM |
579 | |
580 | struct cgroup_subsys { | |
eb95419b TH |
581 | struct cgroup_subsys_state *(*css_alloc)(struct cgroup_subsys_state *parent_css); |
582 | int (*css_online)(struct cgroup_subsys_state *css); | |
583 | void (*css_offline)(struct cgroup_subsys_state *css); | |
584 | void (*css_free)(struct cgroup_subsys_state *css); | |
585 | ||
586 | int (*can_attach)(struct cgroup_subsys_state *css, | |
587 | struct cgroup_taskset *tset); | |
588 | void (*cancel_attach)(struct cgroup_subsys_state *css, | |
589 | struct cgroup_taskset *tset); | |
590 | void (*attach)(struct cgroup_subsys_state *css, | |
591 | struct cgroup_taskset *tset); | |
761b3ef5 | 592 | void (*fork)(struct task_struct *task); |
eb95419b TH |
593 | void (*exit)(struct cgroup_subsys_state *css, |
594 | struct cgroup_subsys_state *old_css, | |
761b3ef5 | 595 | struct task_struct *task); |
eb95419b | 596 | void (*bind)(struct cgroup_subsys_state *root_css); |
e5991371 | 597 | |
ddbcc7e8 | 598 | int subsys_id; |
8bab8dde | 599 | int disabled; |
ddbcc7e8 | 600 | int early_init; |
48ddbe19 | 601 | |
8c7f6edb TH |
602 | /* |
603 | * If %false, this subsystem is properly hierarchical - | |
604 | * configuration, resource accounting and restriction on a parent | |
605 | * cgroup cover those of its children. If %true, hierarchy support | |
606 | * is broken in some ways - some subsystems ignore hierarchy | |
607 | * completely while others are only implemented half-way. | |
608 | * | |
609 | * It's now disallowed to create nested cgroups if the subsystem is | |
610 | * broken and cgroup core will emit a warning message on such | |
611 | * cases. Eventually, all subsystems will be made properly | |
612 | * hierarchical and this will go away. | |
613 | */ | |
614 | bool broken_hierarchy; | |
615 | bool warned_broken_hierarchy; | |
616 | ||
ddbcc7e8 PM |
617 | #define MAX_CGROUP_TYPE_NAMELEN 32 |
618 | const char *name; | |
619 | ||
999cd8a4 PM |
620 | /* |
621 | * Link to parent, and list entry in parent's children. | |
6be96a5c | 622 | * Protected by cgroup_lock() |
999cd8a4 PM |
623 | */ |
624 | struct cgroupfs_root *root; | |
ddbcc7e8 | 625 | struct list_head sibling; |
e6a1105b | 626 | |
8e3f6541 TH |
627 | /* list of cftype_sets */ |
628 | struct list_head cftsets; | |
629 | ||
630 | /* base cftypes, automatically [de]registered with subsys itself */ | |
631 | struct cftype *base_cftypes; | |
632 | struct cftype_set base_cftset; | |
633 | ||
e6a1105b BB |
634 | /* should be defined only by modular subsystems */ |
635 | struct module *module; | |
ddbcc7e8 PM |
636 | }; |
637 | ||
638 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _subsys; | |
5fc0b025 | 639 | #define IS_SUBSYS_ENABLED(option) IS_BUILTIN(option) |
ddbcc7e8 | 640 | #include <linux/cgroup_subsys.h> |
5fc0b025 | 641 | #undef IS_SUBSYS_ENABLED |
ddbcc7e8 PM |
642 | #undef SUBSYS |
643 | ||
63876986 TH |
644 | /** |
645 | * css_parent - find the parent css | |
646 | * @css: the target cgroup_subsys_state | |
647 | * | |
648 | * Return the parent css of @css. This function is guaranteed to return | |
649 | * non-NULL parent as long as @css isn't the root. | |
650 | */ | |
651 | static inline | |
652 | struct cgroup_subsys_state *css_parent(struct cgroup_subsys_state *css) | |
653 | { | |
0ae78e0b | 654 | return css->parent; |
63876986 TH |
655 | } |
656 | ||
14611e51 TH |
657 | /** |
658 | * task_css_set_check - obtain a task's css_set with extra access conditions | |
659 | * @task: the task to obtain css_set for | |
660 | * @__c: extra condition expression to be passed to rcu_dereference_check() | |
661 | * | |
662 | * A task's css_set is RCU protected, initialized and exited while holding | |
663 | * task_lock(), and can only be modified while holding both cgroup_mutex | |
664 | * and task_lock() while the task is alive. This macro verifies that the | |
665 | * caller is inside proper critical section and returns @task's css_set. | |
666 | * | |
667 | * The caller can also specify additional allowed conditions via @__c, such | |
668 | * as locks used during the cgroup_subsys::attach() methods. | |
dc61b1d6 | 669 | */ |
2219449a TH |
670 | #ifdef CONFIG_PROVE_RCU |
671 | extern struct mutex cgroup_mutex; | |
14611e51 TH |
672 | #define task_css_set_check(task, __c) \ |
673 | rcu_dereference_check((task)->cgroups, \ | |
674 | lockdep_is_held(&(task)->alloc_lock) || \ | |
675 | lockdep_is_held(&cgroup_mutex) || (__c)) | |
2219449a | 676 | #else |
14611e51 TH |
677 | #define task_css_set_check(task, __c) \ |
678 | rcu_dereference((task)->cgroups) | |
2219449a | 679 | #endif |
dc61b1d6 | 680 | |
14611e51 | 681 | /** |
8af01f56 | 682 | * task_css_check - obtain css for (task, subsys) w/ extra access conds |
14611e51 TH |
683 | * @task: the target task |
684 | * @subsys_id: the target subsystem ID | |
685 | * @__c: extra condition expression to be passed to rcu_dereference_check() | |
686 | * | |
687 | * Return the cgroup_subsys_state for the (@task, @subsys_id) pair. The | |
688 | * synchronization rules are the same as task_css_set_check(). | |
689 | */ | |
8af01f56 | 690 | #define task_css_check(task, subsys_id, __c) \ |
14611e51 TH |
691 | task_css_set_check((task), (__c))->subsys[(subsys_id)] |
692 | ||
693 | /** | |
694 | * task_css_set - obtain a task's css_set | |
695 | * @task: the task to obtain css_set for | |
696 | * | |
697 | * See task_css_set_check(). | |
698 | */ | |
699 | static inline struct css_set *task_css_set(struct task_struct *task) | |
700 | { | |
701 | return task_css_set_check(task, false); | |
702 | } | |
703 | ||
704 | /** | |
8af01f56 | 705 | * task_css - obtain css for (task, subsys) |
14611e51 TH |
706 | * @task: the target task |
707 | * @subsys_id: the target subsystem ID | |
708 | * | |
8af01f56 | 709 | * See task_css_check(). |
14611e51 | 710 | */ |
8af01f56 TH |
711 | static inline struct cgroup_subsys_state *task_css(struct task_struct *task, |
712 | int subsys_id) | |
ddbcc7e8 | 713 | { |
8af01f56 | 714 | return task_css_check(task, subsys_id, false); |
ddbcc7e8 PM |
715 | } |
716 | ||
8af01f56 TH |
717 | static inline struct cgroup *task_cgroup(struct task_struct *task, |
718 | int subsys_id) | |
ddbcc7e8 | 719 | { |
8af01f56 | 720 | return task_css(task, subsys_id)->cgroup; |
ddbcc7e8 PM |
721 | } |
722 | ||
492eb21b TH |
723 | struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos, |
724 | struct cgroup_subsys_state *parent); | |
53fa5261 | 725 | |
1cb650b9 LZ |
726 | struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss); |
727 | ||
574bd9f7 | 728 | /** |
492eb21b TH |
729 | * css_for_each_child - iterate through children of a css |
730 | * @pos: the css * to use as the loop cursor | |
731 | * @parent: css whose children to walk | |
574bd9f7 | 732 | * |
492eb21b TH |
733 | * Walk @parent's children. Must be called under rcu_read_lock(). A child |
734 | * css which hasn't finished ->css_online() or already has finished | |
92fb9748 | 735 | * ->css_offline() may show up during traversal and it's each subsystem's |
574bd9f7 TH |
736 | * responsibility to verify that each @pos is alive. |
737 | * | |
92fb9748 | 738 | * If a subsystem synchronizes against the parent in its ->css_online() and |
492eb21b | 739 | * before starting iterating, a css which finished ->css_online() is |
92fb9748 | 740 | * guaranteed to be visible in the future iterations. |
75501a6d TH |
741 | * |
742 | * It is allowed to temporarily drop RCU read lock during iteration. The | |
743 | * caller is responsible for ensuring that @pos remains accessible until | |
744 | * the start of the next iteration by, for example, bumping the css refcnt. | |
574bd9f7 | 745 | */ |
492eb21b TH |
746 | #define css_for_each_child(pos, parent) \ |
747 | for ((pos) = css_next_child(NULL, (parent)); (pos); \ | |
748 | (pos) = css_next_child((pos), (parent))) | |
574bd9f7 | 749 | |
492eb21b TH |
750 | struct cgroup_subsys_state * |
751 | css_next_descendant_pre(struct cgroup_subsys_state *pos, | |
752 | struct cgroup_subsys_state *css); | |
753 | ||
754 | struct cgroup_subsys_state * | |
755 | css_rightmost_descendant(struct cgroup_subsys_state *pos); | |
574bd9f7 TH |
756 | |
757 | /** | |
492eb21b TH |
758 | * css_for_each_descendant_pre - pre-order walk of a css's descendants |
759 | * @pos: the css * to use as the loop cursor | |
760 | * @root: css whose descendants to walk | |
574bd9f7 | 761 | * |
bd8815a6 TH |
762 | * Walk @root's descendants. @root is included in the iteration and the |
763 | * first node to be visited. Must be called under rcu_read_lock(). A | |
492eb21b | 764 | * descendant css which hasn't finished ->css_online() or already has |
92fb9748 | 765 | * finished ->css_offline() may show up during traversal and it's each |
574bd9f7 TH |
766 | * subsystem's responsibility to verify that each @pos is alive. |
767 | * | |
92fb9748 TH |
768 | * If a subsystem synchronizes against the parent in its ->css_online() and |
769 | * before starting iterating, and synchronizes against @pos on each | |
492eb21b | 770 | * iteration, any descendant css which finished ->css_online() is |
574bd9f7 TH |
771 | * guaranteed to be visible in the future iterations. |
772 | * | |
773 | * In other words, the following guarantees that a descendant can't escape | |
774 | * state updates of its ancestors. | |
775 | * | |
492eb21b | 776 | * my_online(@css) |
574bd9f7 | 777 | * { |
492eb21b TH |
778 | * Lock @css's parent and @css; |
779 | * Inherit state from the parent; | |
574bd9f7 TH |
780 | * Unlock both. |
781 | * } | |
782 | * | |
492eb21b | 783 | * my_update_state(@css) |
574bd9f7 | 784 | * { |
492eb21b | 785 | * css_for_each_descendant_pre(@pos, @css) { |
574bd9f7 | 786 | * Lock @pos; |
bd8815a6 TH |
787 | * if (@pos == @css) |
788 | * Update @css's state; | |
789 | * else | |
790 | * Verify @pos is alive and inherit state from its parent; | |
574bd9f7 TH |
791 | * Unlock @pos; |
792 | * } | |
793 | * } | |
794 | * | |
795 | * As long as the inheriting step, including checking the parent state, is | |
796 | * enclosed inside @pos locking, double-locking the parent isn't necessary | |
797 | * while inheriting. The state update to the parent is guaranteed to be | |
798 | * visible by walking order and, as long as inheriting operations to the | |
799 | * same @pos are atomic to each other, multiple updates racing each other | |
800 | * still result in the correct state. It's guaranateed that at least one | |
492eb21b | 801 | * inheritance happens for any css after the latest update to its parent. |
574bd9f7 TH |
802 | * |
803 | * If checking parent's state requires locking the parent, each inheriting | |
804 | * iteration should lock and unlock both @pos->parent and @pos. | |
805 | * | |
806 | * Alternatively, a subsystem may choose to use a single global lock to | |
92fb9748 | 807 | * synchronize ->css_online() and ->css_offline() against tree-walking |
574bd9f7 | 808 | * operations. |
75501a6d TH |
809 | * |
810 | * It is allowed to temporarily drop RCU read lock during iteration. The | |
811 | * caller is responsible for ensuring that @pos remains accessible until | |
812 | * the start of the next iteration by, for example, bumping the css refcnt. | |
574bd9f7 | 813 | */ |
492eb21b TH |
814 | #define css_for_each_descendant_pre(pos, css) \ |
815 | for ((pos) = css_next_descendant_pre(NULL, (css)); (pos); \ | |
816 | (pos) = css_next_descendant_pre((pos), (css))) | |
574bd9f7 | 817 | |
492eb21b TH |
818 | struct cgroup_subsys_state * |
819 | css_next_descendant_post(struct cgroup_subsys_state *pos, | |
820 | struct cgroup_subsys_state *css); | |
574bd9f7 TH |
821 | |
822 | /** | |
492eb21b TH |
823 | * css_for_each_descendant_post - post-order walk of a css's descendants |
824 | * @pos: the css * to use as the loop cursor | |
825 | * @css: css whose descendants to walk | |
574bd9f7 | 826 | * |
492eb21b | 827 | * Similar to css_for_each_descendant_pre() but performs post-order |
bd8815a6 TH |
828 | * traversal instead. @root is included in the iteration and the last |
829 | * node to be visited. Note that the walk visibility guarantee described | |
830 | * in pre-order walk doesn't apply the same to post-order walks. | |
574bd9f7 | 831 | */ |
492eb21b TH |
832 | #define css_for_each_descendant_post(pos, css) \ |
833 | for ((pos) = css_next_descendant_post(NULL, (css)); (pos); \ | |
834 | (pos) = css_next_descendant_post((pos), (css))) | |
574bd9f7 | 835 | |
72ec7029 TH |
836 | /* A css_task_iter should be treated as an opaque object */ |
837 | struct css_task_iter { | |
838 | struct cgroup_subsys_state *origin_css; | |
0942eeee TH |
839 | struct list_head *cset_link; |
840 | struct list_head *task; | |
817929ec PM |
841 | }; |
842 | ||
72ec7029 TH |
843 | void css_task_iter_start(struct cgroup_subsys_state *css, |
844 | struct css_task_iter *it); | |
845 | struct task_struct *css_task_iter_next(struct css_task_iter *it); | |
846 | void css_task_iter_end(struct css_task_iter *it); | |
e535837b | 847 | |
72ec7029 TH |
848 | int css_scan_tasks(struct cgroup_subsys_state *css, |
849 | bool (*test)(struct task_struct *, void *), | |
850 | void (*process)(struct task_struct *, void *), | |
851 | void *data, struct ptr_heap *heap); | |
e535837b | 852 | |
31583bb0 | 853 | int cgroup_attach_task_all(struct task_struct *from, struct task_struct *); |
8cc99345 | 854 | int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from); |
31583bb0 | 855 | |
35cf0836 TH |
856 | struct cgroup_subsys_state *css_from_dir(struct dentry *dentry, |
857 | struct cgroup_subsys *ss); | |
38460b48 | 858 | |
ddbcc7e8 PM |
859 | #else /* !CONFIG_CGROUPS */ |
860 | ||
861 | static inline int cgroup_init_early(void) { return 0; } | |
862 | static inline int cgroup_init(void) { return 0; } | |
b4f48b63 | 863 | static inline void cgroup_fork(struct task_struct *p) {} |
817929ec | 864 | static inline void cgroup_post_fork(struct task_struct *p) {} |
b4f48b63 | 865 | static inline void cgroup_exit(struct task_struct *p, int callbacks) {} |
ddbcc7e8 | 866 | |
846c7bb0 BS |
867 | static inline int cgroupstats_build(struct cgroupstats *stats, |
868 | struct dentry *dentry) | |
869 | { | |
870 | return -EINVAL; | |
871 | } | |
ddbcc7e8 | 872 | |
d7926ee3 | 873 | /* No cgroups - nothing to do */ |
31583bb0 MT |
874 | static inline int cgroup_attach_task_all(struct task_struct *from, |
875 | struct task_struct *t) | |
876 | { | |
877 | return 0; | |
878 | } | |
d7926ee3 | 879 | |
ddbcc7e8 PM |
880 | #endif /* !CONFIG_CGROUPS */ |
881 | ||
882 | #endif /* _LINUX_CGROUP_H */ |