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> |
f8f22e53 | 24 | #include <linux/wait.h> |
ddbcc7e8 PM |
25 | |
26 | #ifdef CONFIG_CGROUPS | |
27 | ||
3dd06ffa | 28 | struct cgroup_root; |
ddbcc7e8 PM |
29 | struct cgroup_subsys; |
30 | struct inode; | |
84eea842 | 31 | struct cgroup; |
ddbcc7e8 PM |
32 | |
33 | extern int cgroup_init_early(void); | |
34 | extern int cgroup_init(void); | |
b4f48b63 | 35 | extern void cgroup_fork(struct task_struct *p); |
817929ec | 36 | extern void cgroup_post_fork(struct task_struct *p); |
1ec41830 | 37 | extern void cgroup_exit(struct task_struct *p); |
846c7bb0 BS |
38 | extern int cgroupstats_build(struct cgroupstats *stats, |
39 | struct dentry *dentry); | |
ddbcc7e8 | 40 | |
8d8b97ba | 41 | extern int proc_cgroup_show(struct seq_file *, void *); |
a424316c | 42 | |
3ed80a62 | 43 | /* define the enumeration of all cgroup subsystems */ |
073219e9 | 44 | #define SUBSYS(_x) _x ## _cgrp_id, |
817929ec PM |
45 | enum cgroup_subsys_id { |
46 | #include <linux/cgroup_subsys.h> | |
a6f00298 | 47 | CGROUP_SUBSYS_COUNT, |
817929ec PM |
48 | }; |
49 | #undef SUBSYS | |
50 | ||
ddbcc7e8 PM |
51 | /* Per-subsystem/per-cgroup state maintained by the system. */ |
52 | struct cgroup_subsys_state { | |
72c97e54 | 53 | /* the cgroup that this css is attached to */ |
ddbcc7e8 PM |
54 | struct cgroup *cgroup; |
55 | ||
72c97e54 TH |
56 | /* the cgroup subsystem that this css is attached to */ |
57 | struct cgroup_subsys *ss; | |
58 | ||
d3daf28d TH |
59 | /* reference count - access via css_[try]get() and css_put() */ |
60 | struct percpu_ref refcnt; | |
ddbcc7e8 | 61 | |
0ae78e0b TH |
62 | /* the parent css */ |
63 | struct cgroup_subsys_state *parent; | |
64 | ||
69dfa00c | 65 | unsigned int flags; |
48ddbe19 | 66 | |
0c21ead1 TH |
67 | /* percpu_ref killing and RCU release */ |
68 | struct rcu_head rcu_head; | |
35ef10da | 69 | struct work_struct destroy_work; |
ddbcc7e8 PM |
70 | }; |
71 | ||
72 | /* bits in struct cgroup_subsys_state flags field */ | |
73 | enum { | |
38b53aba | 74 | CSS_ROOT = (1 << 0), /* this CSS is the root of the subsystem */ |
92fb9748 | 75 | CSS_ONLINE = (1 << 1), /* between ->css_online() and ->css_offline() */ |
ddbcc7e8 PM |
76 | }; |
77 | ||
5de0107e TH |
78 | /** |
79 | * css_get - obtain a reference on the specified css | |
80 | * @css: target css | |
81 | * | |
82 | * The caller must already have a reference. | |
ddbcc7e8 | 83 | */ |
ddbcc7e8 PM |
84 | static inline void css_get(struct cgroup_subsys_state *css) |
85 | { | |
86 | /* We don't need to reference count the root state */ | |
38b53aba | 87 | if (!(css->flags & CSS_ROOT)) |
d3daf28d | 88 | percpu_ref_get(&css->refcnt); |
ddbcc7e8 | 89 | } |
e7c5ec91 | 90 | |
5de0107e TH |
91 | /** |
92 | * css_tryget - try to obtain a reference on the specified css | |
93 | * @css: target css | |
94 | * | |
95 | * Obtain a reference on @css if it's alive. The caller naturally needs to | |
96 | * ensure that @css is accessible but doesn't have to be holding a | |
97 | * reference on it - IOW, RCU protected access is good enough for this | |
98 | * function. Returns %true if a reference count was successfully obtained; | |
99 | * %false otherwise. | |
100 | */ | |
e7c5ec91 PM |
101 | static inline bool css_tryget(struct cgroup_subsys_state *css) |
102 | { | |
38b53aba | 103 | if (css->flags & CSS_ROOT) |
e7c5ec91 | 104 | return true; |
d3daf28d | 105 | return percpu_ref_tryget(&css->refcnt); |
e7c5ec91 PM |
106 | } |
107 | ||
5de0107e TH |
108 | /** |
109 | * css_put - put a css reference | |
110 | * @css: target css | |
111 | * | |
112 | * Put a reference obtained via css_get() and css_tryget(). | |
113 | */ | |
ddbcc7e8 PM |
114 | static inline void css_put(struct cgroup_subsys_state *css) |
115 | { | |
38b53aba | 116 | if (!(css->flags & CSS_ROOT)) |
d3daf28d | 117 | percpu_ref_put(&css->refcnt); |
ddbcc7e8 PM |
118 | } |
119 | ||
3116f0e3 PM |
120 | /* bits in struct cgroup flags field */ |
121 | enum { | |
122 | /* Control Group is dead */ | |
54766d4a | 123 | CGRP_DEAD, |
d20a390a PM |
124 | /* |
125 | * Control Group has previously had a child cgroup or a task, | |
126 | * but no longer (only if CGRP_NOTIFY_ON_RELEASE is set) | |
127 | */ | |
3116f0e3 PM |
128 | CGRP_RELEASABLE, |
129 | /* Control Group requires release notifications to userspace */ | |
130 | CGRP_NOTIFY_ON_RELEASE, | |
97978e6d | 131 | /* |
2260e7fc TH |
132 | * Clone the parent's configuration when creating a new child |
133 | * cpuset cgroup. For historical reasons, this option can be | |
134 | * specified at mount time and thus is implemented here. | |
97978e6d | 135 | */ |
2260e7fc | 136 | CGRP_CPUSET_CLONE_CHILDREN, |
873fe09e TH |
137 | /* see the comment above CGRP_ROOT_SANE_BEHAVIOR for details */ |
138 | CGRP_SANE_BEHAVIOR, | |
3116f0e3 PM |
139 | }; |
140 | ||
ddbcc7e8 PM |
141 | struct cgroup { |
142 | unsigned long flags; /* "unsigned long" so bitops work */ | |
143 | ||
b414dc09 LZ |
144 | /* |
145 | * idr allocated in-hierarchy ID. | |
146 | * | |
7d699ddb TH |
147 | * ID 0 is not used, the ID of the root cgroup is always 1, and a |
148 | * new cgroup will be assigned with a smallest available ID. | |
0ab02ca8 LZ |
149 | * |
150 | * Allocating/Removing ID must be protected by cgroup_mutex. | |
b414dc09 LZ |
151 | */ |
152 | int id; | |
0a950f65 | 153 | |
f20104de TH |
154 | /* the number of attached css's */ |
155 | int nr_css; | |
156 | ||
842b597e TH |
157 | /* |
158 | * If this cgroup contains any tasks, it contributes one to | |
159 | * populated_cnt. All children with non-zero popuplated_cnt of | |
160 | * their own contribute one. The count is zero iff there's no task | |
161 | * in this cgroup or its subtree. | |
162 | */ | |
163 | int populated_cnt; | |
164 | ||
2bd59d48 TH |
165 | atomic_t refcnt; |
166 | ||
ddbcc7e8 PM |
167 | /* |
168 | * We link our 'sibling' struct into our parent's 'children'. | |
169 | * Our children link their 'sibling' into our 'children'. | |
170 | */ | |
171 | struct list_head sibling; /* my parent's children */ | |
172 | struct list_head children; /* my children */ | |
173 | ||
d20a390a | 174 | struct cgroup *parent; /* my parent */ |
2bd59d48 | 175 | struct kernfs_node *kn; /* cgroup kernfs entry */ |
f8f22e53 | 176 | struct kernfs_node *control_kn; /* kn for "cgroup.subtree_control" */ |
842b597e | 177 | struct kernfs_node *populated_kn; /* kn for "cgroup.subtree_populated" */ |
ddbcc7e8 | 178 | |
53fa5261 TH |
179 | /* |
180 | * Monotonically increasing unique serial number which defines a | |
181 | * uniform order among all cgroups. It's guaranteed that all | |
182 | * ->children lists are in the ascending order of ->serial_nr. | |
183 | * It's used to allow interrupting and resuming iterations. | |
184 | */ | |
185 | u64 serial_nr; | |
186 | ||
f392e51c | 187 | /* the bitmask of subsystems enabled on the child cgroups */ |
69dfa00c | 188 | unsigned int child_subsys_mask; |
94419627 | 189 | |
ddbcc7e8 | 190 | /* Private pointers for each registered subsystem */ |
73e80ed8 | 191 | struct cgroup_subsys_state __rcu *subsys[CGROUP_SUBSYS_COUNT]; |
ddbcc7e8 | 192 | |
3dd06ffa | 193 | struct cgroup_root *root; |
817929ec PM |
194 | |
195 | /* | |
69d0206c TH |
196 | * List of cgrp_cset_links pointing at css_sets with tasks in this |
197 | * cgroup. Protected by css_set_lock. | |
817929ec | 198 | */ |
69d0206c | 199 | struct list_head cset_links; |
81a6a5cd | 200 | |
2d8f243a TH |
201 | /* |
202 | * On the default hierarchy, a css_set for a cgroup with some | |
203 | * susbsys disabled will point to css's which are associated with | |
204 | * the closest ancestor which has the subsys enabled. The | |
205 | * following lists all css_sets which point to this cgroup's css | |
206 | * for the given subsystem. | |
207 | */ | |
208 | struct list_head e_csets[CGROUP_SUBSYS_COUNT]; | |
209 | ||
81a6a5cd PM |
210 | /* |
211 | * Linked list running through all cgroups that can | |
212 | * potentially be reaped by the release agent. Protected by | |
213 | * release_list_lock | |
214 | */ | |
215 | struct list_head release_list; | |
cc31edce | 216 | |
72a8cb30 BB |
217 | /* |
218 | * list of pidlists, up to two for each namespace (one for procs, one | |
219 | * for tasks); created on demand. | |
220 | */ | |
221 | struct list_head pidlists; | |
222 | struct mutex pidlist_mutex; | |
a47295e6 | 223 | |
67f4c36f TH |
224 | /* dummy css with NULL ->ss, points back to this cgroup */ |
225 | struct cgroup_subsys_state dummy_css; | |
226 | ||
d3daf28d | 227 | /* For css percpu_ref killing and RCU-protected deletion */ |
a47295e6 | 228 | struct rcu_head rcu_head; |
ea15f8cc | 229 | struct work_struct destroy_work; |
f8f22e53 TH |
230 | |
231 | /* used to wait for offlining of csses */ | |
232 | wait_queue_head_t offline_waitq; | |
817929ec PM |
233 | }; |
234 | ||
25a7e684 TH |
235 | #define MAX_CGROUP_ROOT_NAMELEN 64 |
236 | ||
3dd06ffa | 237 | /* cgroup_root->flags */ |
25a7e684 | 238 | enum { |
873fe09e TH |
239 | /* |
240 | * Unfortunately, cgroup core and various controllers are riddled | |
241 | * with idiosyncrasies and pointless options. The following flag, | |
242 | * when set, will force sane behavior - some options are forced on, | |
243 | * others are disallowed, and some controllers will change their | |
244 | * hierarchical or other behaviors. | |
245 | * | |
246 | * The set of behaviors affected by this flag are still being | |
247 | * determined and developed and the mount option for this flag is | |
248 | * prefixed with __DEVEL__. The prefix will be dropped once we | |
249 | * reach the point where all behaviors are compatible with the | |
250 | * planned unified hierarchy, which will automatically turn on this | |
251 | * flag. | |
252 | * | |
253 | * The followings are the behaviors currently affected this flag. | |
254 | * | |
d3ba07c3 TH |
255 | * - Mount options "noprefix", "xattr", "clone_children", |
256 | * "release_agent" and "name" are disallowed. | |
873fe09e TH |
257 | * |
258 | * - When mounting an existing superblock, mount options should | |
259 | * match. | |
260 | * | |
261 | * - Remount is disallowed. | |
262 | * | |
0b0585c3 LT |
263 | * - rename(2) is disallowed. |
264 | * | |
f63674fd TH |
265 | * - "tasks" is removed. Everything should be at process |
266 | * granularity. Use "cgroup.procs" instead. | |
f00baae7 | 267 | * |
afb2bc14 TH |
268 | * - "cgroup.procs" is not sorted. pids will be unique unless they |
269 | * got recycled inbetween reads. | |
270 | * | |
f63674fd TH |
271 | * - "release_agent" and "notify_on_release" are removed. |
272 | * Replacement notification mechanism will be implemented. | |
873fe09e | 273 | * |
d3ba07c3 | 274 | * - "cgroup.clone_children" is removed. |
86bf4b68 | 275 | * |
842b597e TH |
276 | * - "cgroup.subtree_populated" is available. Its value is 0 if |
277 | * the cgroup and its descendants contain no task; otherwise, 1. | |
278 | * The file also generates kernfs notification which can be | |
279 | * monitored through poll and [di]notify when the value of the | |
280 | * file changes. | |
281 | * | |
a2dd4247 TH |
282 | * - If mount is requested with sane_behavior but without any |
283 | * subsystem, the default unified hierarchy is mounted. | |
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, | |
25a7e684 TH |
304 | }; |
305 | ||
306 | /* | |
3dd06ffa | 307 | * A cgroup_root represents the root of a cgroup hierarchy, and may be |
2bd59d48 | 308 | * associated with a kernfs_root to form an active hierarchy. This is |
25a7e684 TH |
309 | * internal to cgroup core. Don't access directly from controllers. |
310 | */ | |
3dd06ffa | 311 | struct cgroup_root { |
2bd59d48 | 312 | struct kernfs_root *kf_root; |
25a7e684 | 313 | |
f392e51c | 314 | /* The bitmask of subsystems attached to this hierarchy */ |
69dfa00c | 315 | unsigned int subsys_mask; |
f392e51c | 316 | |
25a7e684 TH |
317 | /* Unique id for this hierarchy. */ |
318 | int hierarchy_id; | |
319 | ||
776f02fa | 320 | /* The root cgroup. Root is destroyed on its release. */ |
3dd06ffa | 321 | struct cgroup cgrp; |
25a7e684 | 322 | |
3c9c825b TH |
323 | /* Number of cgroups in the hierarchy, used only for /proc/cgroups */ |
324 | atomic_t nr_cgrps; | |
25a7e684 TH |
325 | |
326 | /* A list running through the active hierarchies */ | |
327 | struct list_head root_list; | |
328 | ||
25a7e684 | 329 | /* Hierarchy-specific flags */ |
69dfa00c | 330 | unsigned int flags; |
25a7e684 TH |
331 | |
332 | /* IDs for cgroups in this hierarchy */ | |
4e96ee8e | 333 | struct idr cgroup_idr; |
25a7e684 TH |
334 | |
335 | /* The path to use for release notifications. */ | |
336 | char release_agent_path[PATH_MAX]; | |
337 | ||
338 | /* The name for this hierarchy - may be empty */ | |
339 | char name[MAX_CGROUP_ROOT_NAMELEN]; | |
340 | }; | |
341 | ||
d20a390a PM |
342 | /* |
343 | * A css_set is a structure holding pointers to a set of | |
817929ec PM |
344 | * cgroup_subsys_state objects. This saves space in the task struct |
345 | * object and speeds up fork()/exit(), since a single inc/dec and a | |
d20a390a PM |
346 | * list_add()/del() can bump the reference count on the entire cgroup |
347 | * set for a task. | |
817929ec PM |
348 | */ |
349 | ||
350 | struct css_set { | |
351 | ||
352 | /* Reference count */ | |
146aa1bd | 353 | atomic_t refcount; |
817929ec | 354 | |
472b1053 LZ |
355 | /* |
356 | * List running through all cgroup groups in the same hash | |
357 | * slot. Protected by css_set_lock | |
358 | */ | |
359 | struct hlist_node hlist; | |
360 | ||
817929ec | 361 | /* |
c7561128 TH |
362 | * Lists running through all tasks using this cgroup group. |
363 | * mg_tasks lists tasks which belong to this cset but are in the | |
364 | * process of being migrated out or in. Protected by | |
365 | * css_set_rwsem, but, during migration, once tasks are moved to | |
366 | * mg_tasks, it can be read safely while holding cgroup_mutex. | |
817929ec PM |
367 | */ |
368 | struct list_head tasks; | |
c7561128 | 369 | struct list_head mg_tasks; |
817929ec PM |
370 | |
371 | /* | |
69d0206c TH |
372 | * List of cgrp_cset_links pointing at cgroups referenced from this |
373 | * css_set. Protected by css_set_lock. | |
817929ec | 374 | */ |
69d0206c | 375 | struct list_head cgrp_links; |
817929ec | 376 | |
6803c006 TH |
377 | /* the default cgroup associated with this css_set */ |
378 | struct cgroup *dfl_cgrp; | |
379 | ||
817929ec | 380 | /* |
3ed80a62 TH |
381 | * Set of subsystem states, one for each subsystem. This array is |
382 | * immutable after creation apart from the init_css_set during | |
383 | * subsystem registration (at boot time). | |
817929ec PM |
384 | */ |
385 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
c378369d | 386 | |
b3dc094e TH |
387 | /* |
388 | * List of csets participating in the on-going migration either as | |
389 | * source or destination. Protected by cgroup_mutex. | |
390 | */ | |
1958d2d5 | 391 | struct list_head mg_preload_node; |
b3dc094e TH |
392 | struct list_head mg_node; |
393 | ||
394 | /* | |
395 | * If this cset is acting as the source of migration the following | |
396 | * two fields are set. mg_src_cgrp is the source cgroup of the | |
397 | * on-going migration and mg_dst_cset is the destination cset the | |
398 | * target tasks on this cset should be migrated to. Protected by | |
399 | * cgroup_mutex. | |
400 | */ | |
401 | struct cgroup *mg_src_cgrp; | |
402 | struct css_set *mg_dst_cset; | |
403 | ||
2d8f243a TH |
404 | /* |
405 | * On the default hierarhcy, ->subsys[ssid] may point to a css | |
406 | * attached to an ancestor instead of the cgroup this css_set is | |
407 | * associated with. The following node is anchored at | |
408 | * ->subsys[ssid]->cgroup->e_csets[ssid] and provides a way to | |
409 | * iterate through all css's attached to a given cgroup. | |
410 | */ | |
411 | struct list_head e_cset_node[CGROUP_SUBSYS_COUNT]; | |
412 | ||
c378369d BB |
413 | /* For RCU-protected deletion */ |
414 | struct rcu_head rcu_head; | |
ddbcc7e8 PM |
415 | }; |
416 | ||
d20a390a PM |
417 | /* |
418 | * struct cftype: handler definitions for cgroup control files | |
ddbcc7e8 PM |
419 | * |
420 | * When reading/writing to a file: | |
a043e3b2 | 421 | * - the cgroup to use is file->f_dentry->d_parent->d_fsdata |
ddbcc7e8 PM |
422 | * - the 'cftype' of the file is file->f_dentry->d_fsdata |
423 | */ | |
424 | ||
8e3f6541 | 425 | /* cftype->flags */ |
02c402d9 | 426 | enum { |
6f4b7e63 LZ |
427 | CFTYPE_ONLY_ON_ROOT = (1 << 0), /* only create on root cgrp */ |
428 | CFTYPE_NOT_ON_ROOT = (1 << 1), /* don't create on root cgrp */ | |
02c402d9 | 429 | CFTYPE_INSANE = (1 << 2), /* don't create if sane_behavior */ |
9fa4db33 | 430 | CFTYPE_NO_PREFIX = (1 << 3), /* (DON'T USE FOR NEW FILES) no subsys prefix */ |
8cbbf2c9 | 431 | CFTYPE_ONLY_ON_DFL = (1 << 4), /* only on default hierarchy */ |
02c402d9 | 432 | }; |
8e3f6541 TH |
433 | |
434 | #define MAX_CFTYPE_NAME 64 | |
435 | ||
ddbcc7e8 | 436 | struct cftype { |
d20a390a PM |
437 | /* |
438 | * By convention, the name should begin with the name of the | |
8e3f6541 TH |
439 | * subsystem, followed by a period. Zero length string indicates |
440 | * end of cftype array. | |
d20a390a | 441 | */ |
ddbcc7e8 PM |
442 | char name[MAX_CFTYPE_NAME]; |
443 | int private; | |
099fca32 LZ |
444 | /* |
445 | * If not 0, file mode is set to this value, otherwise it will | |
446 | * be figured out automatically | |
447 | */ | |
a5e7ed32 | 448 | umode_t mode; |
db3b1497 PM |
449 | |
450 | /* | |
5f469907 TH |
451 | * The maximum length of string, excluding trailing nul, that can |
452 | * be passed to write_string. If < PAGE_SIZE-1, PAGE_SIZE-1 is | |
453 | * assumed. | |
db3b1497 PM |
454 | */ |
455 | size_t max_write_len; | |
456 | ||
8e3f6541 TH |
457 | /* CFTYPE_* flags */ |
458 | unsigned int flags; | |
459 | ||
2bb566cb | 460 | /* |
0adb0704 TH |
461 | * Fields used for internal bookkeeping. Initialized automatically |
462 | * during registration. | |
2bb566cb | 463 | */ |
0adb0704 TH |
464 | struct cgroup_subsys *ss; /* NULL for cgroup core files */ |
465 | struct list_head node; /* anchored at ss->cfts */ | |
2bd59d48 TH |
466 | struct kernfs_ops *kf_ops; |
467 | ||
ddbcc7e8 | 468 | /* |
f4c753b7 | 469 | * read_u64() is a shortcut for the common case of returning a |
ddbcc7e8 PM |
470 | * single integer. Use it in place of read() |
471 | */ | |
182446d0 | 472 | u64 (*read_u64)(struct cgroup_subsys_state *css, struct cftype *cft); |
e73d2c61 PM |
473 | /* |
474 | * read_s64() is a signed version of read_u64() | |
475 | */ | |
182446d0 | 476 | s64 (*read_s64)(struct cgroup_subsys_state *css, struct cftype *cft); |
2da8ca82 TH |
477 | |
478 | /* generic seq_file read interface */ | |
479 | int (*seq_show)(struct seq_file *sf, void *v); | |
91796569 | 480 | |
6612f05b TH |
481 | /* optional ops, implement all or none */ |
482 | void *(*seq_start)(struct seq_file *sf, loff_t *ppos); | |
483 | void *(*seq_next)(struct seq_file *sf, void *v, loff_t *ppos); | |
484 | void (*seq_stop)(struct seq_file *sf, void *v); | |
485 | ||
355e0c48 | 486 | /* |
f4c753b7 | 487 | * write_u64() is a shortcut for the common case of accepting |
355e0c48 PM |
488 | * a single integer (as parsed by simple_strtoull) from |
489 | * userspace. Use in place of write(); return 0 or error. | |
490 | */ | |
182446d0 TH |
491 | int (*write_u64)(struct cgroup_subsys_state *css, struct cftype *cft, |
492 | u64 val); | |
e73d2c61 PM |
493 | /* |
494 | * write_s64() is a signed version of write_u64() | |
495 | */ | |
182446d0 TH |
496 | int (*write_s64)(struct cgroup_subsys_state *css, struct cftype *cft, |
497 | s64 val); | |
355e0c48 | 498 | |
db3b1497 PM |
499 | /* |
500 | * write_string() is passed a nul-terminated kernelspace | |
501 | * buffer of maximum length determined by max_write_len. | |
502 | * Returns 0 or -ve error code. | |
503 | */ | |
182446d0 | 504 | int (*write_string)(struct cgroup_subsys_state *css, struct cftype *cft, |
4d3bb511 | 505 | char *buffer); |
d447ea2f PE |
506 | /* |
507 | * trigger() callback can be used to get some kick from the | |
508 | * userspace, when the actual string written is not important | |
509 | * at all. The private field can be used to determine the | |
510 | * kick type for multiplexing. | |
511 | */ | |
182446d0 | 512 | int (*trigger)(struct cgroup_subsys_state *css, unsigned int event); |
2bd59d48 TH |
513 | |
514 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
515 | struct lock_class_key lockdep_key; | |
516 | #endif | |
ddbcc7e8 PM |
517 | }; |
518 | ||
a2dd4247 TH |
519 | extern struct cgroup_root cgrp_dfl_root; |
520 | ||
521 | static inline bool cgroup_on_dfl(const struct cgroup *cgrp) | |
522 | { | |
523 | return cgrp->root == &cgrp_dfl_root; | |
524 | } | |
525 | ||
873fe09e TH |
526 | /* |
527 | * See the comment above CGRP_ROOT_SANE_BEHAVIOR for details. This | |
528 | * function can be called as long as @cgrp is accessible. | |
529 | */ | |
530 | static inline bool cgroup_sane_behavior(const struct cgroup *cgrp) | |
531 | { | |
532 | return cgrp->root->flags & CGRP_ROOT_SANE_BEHAVIOR; | |
533 | } | |
534 | ||
07bc356e TH |
535 | /* no synchronization, the result can only be used as a hint */ |
536 | static inline bool cgroup_has_tasks(struct cgroup *cgrp) | |
537 | { | |
538 | return !list_empty(&cgrp->cset_links); | |
539 | } | |
540 | ||
b1664924 TH |
541 | /* returns ino associated with a cgroup, 0 indicates unmounted root */ |
542 | static inline ino_t cgroup_ino(struct cgroup *cgrp) | |
543 | { | |
2bd59d48 TH |
544 | if (cgrp->kn) |
545 | return cgrp->kn->ino; | |
b1664924 TH |
546 | else |
547 | return 0; | |
548 | } | |
549 | ||
7da11279 TH |
550 | static inline struct cftype *seq_cft(struct seq_file *seq) |
551 | { | |
2bd59d48 TH |
552 | struct kernfs_open_file *of = seq->private; |
553 | ||
554 | return of->kn->priv; | |
7da11279 TH |
555 | } |
556 | ||
59f5296b TH |
557 | struct cgroup_subsys_state *seq_css(struct seq_file *seq); |
558 | ||
e61734c5 TH |
559 | /* |
560 | * Name / path handling functions. All are thin wrappers around the kernfs | |
561 | * counterparts and can be called under any context. | |
562 | */ | |
563 | ||
564 | static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen) | |
565 | { | |
fdce6bf8 | 566 | return kernfs_name(cgrp->kn, buf, buflen); |
e61734c5 TH |
567 | } |
568 | ||
569 | static inline char * __must_check cgroup_path(struct cgroup *cgrp, char *buf, | |
570 | size_t buflen) | |
571 | { | |
fdce6bf8 | 572 | return kernfs_path(cgrp->kn, buf, buflen); |
e61734c5 TH |
573 | } |
574 | ||
575 | static inline void pr_cont_cgroup_name(struct cgroup *cgrp) | |
576 | { | |
fdce6bf8 | 577 | pr_cont_kernfs_name(cgrp->kn); |
e61734c5 TH |
578 | } |
579 | ||
580 | static inline void pr_cont_cgroup_path(struct cgroup *cgrp) | |
581 | { | |
fdce6bf8 | 582 | pr_cont_kernfs_path(cgrp->kn); |
e61734c5 TH |
583 | } |
584 | ||
585 | char *task_cgroup_path(struct task_struct *task, char *buf, size_t buflen); | |
586 | ||
03b1cde6 | 587 | int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts); |
2bb566cb | 588 | int cgroup_rm_cftypes(struct cftype *cfts); |
8e3f6541 | 589 | |
78574cf9 | 590 | bool cgroup_is_descendant(struct cgroup *cgrp, struct cgroup *ancestor); |
ddbcc7e8 | 591 | |
2f7ee569 TH |
592 | /* |
593 | * Control Group taskset, used to pass around set of tasks to cgroup_subsys | |
594 | * methods. | |
595 | */ | |
596 | struct cgroup_taskset; | |
597 | struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset); | |
598 | struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset); | |
2f7ee569 TH |
599 | |
600 | /** | |
601 | * cgroup_taskset_for_each - iterate cgroup_taskset | |
602 | * @task: the loop cursor | |
2f7ee569 TH |
603 | * @tset: taskset to iterate |
604 | */ | |
924f0d9a | 605 | #define cgroup_taskset_for_each(task, tset) \ |
2f7ee569 | 606 | for ((task) = cgroup_taskset_first((tset)); (task); \ |
924f0d9a | 607 | (task) = cgroup_taskset_next((tset))) |
2f7ee569 | 608 | |
21acb9ca TLSC |
609 | /* |
610 | * Control Group subsystem type. | |
611 | * See Documentation/cgroups/cgroups.txt for details | |
612 | */ | |
ddbcc7e8 PM |
613 | |
614 | struct cgroup_subsys { | |
eb95419b TH |
615 | struct cgroup_subsys_state *(*css_alloc)(struct cgroup_subsys_state *parent_css); |
616 | int (*css_online)(struct cgroup_subsys_state *css); | |
617 | void (*css_offline)(struct cgroup_subsys_state *css); | |
618 | void (*css_free)(struct cgroup_subsys_state *css); | |
619 | ||
620 | int (*can_attach)(struct cgroup_subsys_state *css, | |
621 | struct cgroup_taskset *tset); | |
622 | void (*cancel_attach)(struct cgroup_subsys_state *css, | |
623 | struct cgroup_taskset *tset); | |
624 | void (*attach)(struct cgroup_subsys_state *css, | |
625 | struct cgroup_taskset *tset); | |
761b3ef5 | 626 | void (*fork)(struct task_struct *task); |
eb95419b TH |
627 | void (*exit)(struct cgroup_subsys_state *css, |
628 | struct cgroup_subsys_state *old_css, | |
761b3ef5 | 629 | struct task_struct *task); |
eb95419b | 630 | void (*bind)(struct cgroup_subsys_state *root_css); |
e5991371 | 631 | |
8bab8dde | 632 | int disabled; |
ddbcc7e8 | 633 | int early_init; |
48ddbe19 | 634 | |
8c7f6edb TH |
635 | /* |
636 | * If %false, this subsystem is properly hierarchical - | |
637 | * configuration, resource accounting and restriction on a parent | |
638 | * cgroup cover those of its children. If %true, hierarchy support | |
639 | * is broken in some ways - some subsystems ignore hierarchy | |
640 | * completely while others are only implemented half-way. | |
641 | * | |
642 | * It's now disallowed to create nested cgroups if the subsystem is | |
643 | * broken and cgroup core will emit a warning message on such | |
644 | * cases. Eventually, all subsystems will be made properly | |
645 | * hierarchical and this will go away. | |
646 | */ | |
647 | bool broken_hierarchy; | |
648 | bool warned_broken_hierarchy; | |
649 | ||
073219e9 | 650 | /* the following two fields are initialized automtically during boot */ |
aec25020 | 651 | int id; |
ddbcc7e8 PM |
652 | #define MAX_CGROUP_TYPE_NAMELEN 32 |
653 | const char *name; | |
654 | ||
b85d2040 | 655 | /* link to parent, protected by cgroup_lock() */ |
3dd06ffa | 656 | struct cgroup_root *root; |
e6a1105b | 657 | |
0adb0704 TH |
658 | /* |
659 | * List of cftypes. Each entry is the first entry of an array | |
660 | * terminated by zero length name. | |
661 | */ | |
662 | struct list_head cfts; | |
8e3f6541 | 663 | |
de00ffa5 | 664 | /* base cftypes, automatically registered with subsys itself */ |
8e3f6541 | 665 | struct cftype *base_cftypes; |
ddbcc7e8 PM |
666 | }; |
667 | ||
073219e9 | 668 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys; |
ddbcc7e8 PM |
669 | #include <linux/cgroup_subsys.h> |
670 | #undef SUBSYS | |
671 | ||
63876986 TH |
672 | /** |
673 | * css_parent - find the parent css | |
674 | * @css: the target cgroup_subsys_state | |
675 | * | |
676 | * Return the parent css of @css. This function is guaranteed to return | |
677 | * non-NULL parent as long as @css isn't the root. | |
678 | */ | |
679 | static inline | |
680 | struct cgroup_subsys_state *css_parent(struct cgroup_subsys_state *css) | |
681 | { | |
0ae78e0b | 682 | return css->parent; |
63876986 TH |
683 | } |
684 | ||
14611e51 TH |
685 | /** |
686 | * task_css_set_check - obtain a task's css_set with extra access conditions | |
687 | * @task: the task to obtain css_set for | |
688 | * @__c: extra condition expression to be passed to rcu_dereference_check() | |
689 | * | |
690 | * A task's css_set is RCU protected, initialized and exited while holding | |
691 | * task_lock(), and can only be modified while holding both cgroup_mutex | |
692 | * and task_lock() while the task is alive. This macro verifies that the | |
693 | * caller is inside proper critical section and returns @task's css_set. | |
694 | * | |
695 | * The caller can also specify additional allowed conditions via @__c, such | |
696 | * as locks used during the cgroup_subsys::attach() methods. | |
dc61b1d6 | 697 | */ |
2219449a TH |
698 | #ifdef CONFIG_PROVE_RCU |
699 | extern struct mutex cgroup_mutex; | |
0e1d768f | 700 | extern struct rw_semaphore css_set_rwsem; |
14611e51 TH |
701 | #define task_css_set_check(task, __c) \ |
702 | rcu_dereference_check((task)->cgroups, \ | |
0e1d768f TH |
703 | lockdep_is_held(&cgroup_mutex) || \ |
704 | lockdep_is_held(&css_set_rwsem) || \ | |
705 | ((task)->flags & PF_EXITING) || (__c)) | |
2219449a | 706 | #else |
14611e51 TH |
707 | #define task_css_set_check(task, __c) \ |
708 | rcu_dereference((task)->cgroups) | |
2219449a | 709 | #endif |
dc61b1d6 | 710 | |
14611e51 | 711 | /** |
8af01f56 | 712 | * task_css_check - obtain css for (task, subsys) w/ extra access conds |
14611e51 TH |
713 | * @task: the target task |
714 | * @subsys_id: the target subsystem ID | |
715 | * @__c: extra condition expression to be passed to rcu_dereference_check() | |
716 | * | |
717 | * Return the cgroup_subsys_state for the (@task, @subsys_id) pair. The | |
718 | * synchronization rules are the same as task_css_set_check(). | |
719 | */ | |
8af01f56 | 720 | #define task_css_check(task, subsys_id, __c) \ |
14611e51 TH |
721 | task_css_set_check((task), (__c))->subsys[(subsys_id)] |
722 | ||
723 | /** | |
724 | * task_css_set - obtain a task's css_set | |
725 | * @task: the task to obtain css_set for | |
726 | * | |
727 | * See task_css_set_check(). | |
728 | */ | |
729 | static inline struct css_set *task_css_set(struct task_struct *task) | |
730 | { | |
731 | return task_css_set_check(task, false); | |
732 | } | |
733 | ||
734 | /** | |
8af01f56 | 735 | * task_css - obtain css for (task, subsys) |
14611e51 TH |
736 | * @task: the target task |
737 | * @subsys_id: the target subsystem ID | |
738 | * | |
8af01f56 | 739 | * See task_css_check(). |
14611e51 | 740 | */ |
8af01f56 TH |
741 | static inline struct cgroup_subsys_state *task_css(struct task_struct *task, |
742 | int subsys_id) | |
ddbcc7e8 | 743 | { |
8af01f56 | 744 | return task_css_check(task, subsys_id, false); |
ddbcc7e8 PM |
745 | } |
746 | ||
8af01f56 TH |
747 | static inline struct cgroup *task_cgroup(struct task_struct *task, |
748 | int subsys_id) | |
ddbcc7e8 | 749 | { |
8af01f56 | 750 | return task_css(task, subsys_id)->cgroup; |
ddbcc7e8 PM |
751 | } |
752 | ||
492eb21b TH |
753 | struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos, |
754 | struct cgroup_subsys_state *parent); | |
53fa5261 | 755 | |
1cb650b9 LZ |
756 | struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss); |
757 | ||
574bd9f7 | 758 | /** |
492eb21b TH |
759 | * css_for_each_child - iterate through children of a css |
760 | * @pos: the css * to use as the loop cursor | |
761 | * @parent: css whose children to walk | |
574bd9f7 | 762 | * |
492eb21b TH |
763 | * Walk @parent's children. Must be called under rcu_read_lock(). A child |
764 | * css which hasn't finished ->css_online() or already has finished | |
92fb9748 | 765 | * ->css_offline() may show up during traversal and it's each subsystem's |
574bd9f7 TH |
766 | * responsibility to verify that each @pos is alive. |
767 | * | |
92fb9748 | 768 | * If a subsystem synchronizes against the parent in its ->css_online() and |
492eb21b | 769 | * before starting iterating, a css which finished ->css_online() is |
92fb9748 | 770 | * guaranteed to be visible in the future iterations. |
75501a6d TH |
771 | * |
772 | * It is allowed to temporarily drop RCU read lock during iteration. The | |
773 | * caller is responsible for ensuring that @pos remains accessible until | |
774 | * the start of the next iteration by, for example, bumping the css refcnt. | |
574bd9f7 | 775 | */ |
492eb21b TH |
776 | #define css_for_each_child(pos, parent) \ |
777 | for ((pos) = css_next_child(NULL, (parent)); (pos); \ | |
778 | (pos) = css_next_child((pos), (parent))) | |
574bd9f7 | 779 | |
492eb21b TH |
780 | struct cgroup_subsys_state * |
781 | css_next_descendant_pre(struct cgroup_subsys_state *pos, | |
782 | struct cgroup_subsys_state *css); | |
783 | ||
784 | struct cgroup_subsys_state * | |
785 | css_rightmost_descendant(struct cgroup_subsys_state *pos); | |
574bd9f7 TH |
786 | |
787 | /** | |
492eb21b TH |
788 | * css_for_each_descendant_pre - pre-order walk of a css's descendants |
789 | * @pos: the css * to use as the loop cursor | |
790 | * @root: css whose descendants to walk | |
574bd9f7 | 791 | * |
bd8815a6 TH |
792 | * Walk @root's descendants. @root is included in the iteration and the |
793 | * first node to be visited. Must be called under rcu_read_lock(). A | |
492eb21b | 794 | * descendant css which hasn't finished ->css_online() or already has |
92fb9748 | 795 | * finished ->css_offline() may show up during traversal and it's each |
574bd9f7 TH |
796 | * subsystem's responsibility to verify that each @pos is alive. |
797 | * | |
92fb9748 TH |
798 | * If a subsystem synchronizes against the parent in its ->css_online() and |
799 | * before starting iterating, and synchronizes against @pos on each | |
492eb21b | 800 | * iteration, any descendant css which finished ->css_online() is |
574bd9f7 TH |
801 | * guaranteed to be visible in the future iterations. |
802 | * | |
803 | * In other words, the following guarantees that a descendant can't escape | |
804 | * state updates of its ancestors. | |
805 | * | |
492eb21b | 806 | * my_online(@css) |
574bd9f7 | 807 | * { |
492eb21b TH |
808 | * Lock @css's parent and @css; |
809 | * Inherit state from the parent; | |
574bd9f7 TH |
810 | * Unlock both. |
811 | * } | |
812 | * | |
492eb21b | 813 | * my_update_state(@css) |
574bd9f7 | 814 | * { |
492eb21b | 815 | * css_for_each_descendant_pre(@pos, @css) { |
574bd9f7 | 816 | * Lock @pos; |
bd8815a6 TH |
817 | * if (@pos == @css) |
818 | * Update @css's state; | |
819 | * else | |
820 | * Verify @pos is alive and inherit state from its parent; | |
574bd9f7 TH |
821 | * Unlock @pos; |
822 | * } | |
823 | * } | |
824 | * | |
825 | * As long as the inheriting step, including checking the parent state, is | |
826 | * enclosed inside @pos locking, double-locking the parent isn't necessary | |
827 | * while inheriting. The state update to the parent is guaranteed to be | |
828 | * visible by walking order and, as long as inheriting operations to the | |
829 | * same @pos are atomic to each other, multiple updates racing each other | |
830 | * still result in the correct state. It's guaranateed that at least one | |
492eb21b | 831 | * inheritance happens for any css after the latest update to its parent. |
574bd9f7 TH |
832 | * |
833 | * If checking parent's state requires locking the parent, each inheriting | |
834 | * iteration should lock and unlock both @pos->parent and @pos. | |
835 | * | |
836 | * Alternatively, a subsystem may choose to use a single global lock to | |
92fb9748 | 837 | * synchronize ->css_online() and ->css_offline() against tree-walking |
574bd9f7 | 838 | * operations. |
75501a6d TH |
839 | * |
840 | * It is allowed to temporarily drop RCU read lock during iteration. The | |
841 | * caller is responsible for ensuring that @pos remains accessible until | |
842 | * the start of the next iteration by, for example, bumping the css refcnt. | |
574bd9f7 | 843 | */ |
492eb21b TH |
844 | #define css_for_each_descendant_pre(pos, css) \ |
845 | for ((pos) = css_next_descendant_pre(NULL, (css)); (pos); \ | |
846 | (pos) = css_next_descendant_pre((pos), (css))) | |
574bd9f7 | 847 | |
492eb21b TH |
848 | struct cgroup_subsys_state * |
849 | css_next_descendant_post(struct cgroup_subsys_state *pos, | |
850 | struct cgroup_subsys_state *css); | |
574bd9f7 TH |
851 | |
852 | /** | |
492eb21b TH |
853 | * css_for_each_descendant_post - post-order walk of a css's descendants |
854 | * @pos: the css * to use as the loop cursor | |
855 | * @css: css whose descendants to walk | |
574bd9f7 | 856 | * |
492eb21b | 857 | * Similar to css_for_each_descendant_pre() but performs post-order |
bd8815a6 TH |
858 | * traversal instead. @root is included in the iteration and the last |
859 | * node to be visited. Note that the walk visibility guarantee described | |
860 | * in pre-order walk doesn't apply the same to post-order walks. | |
574bd9f7 | 861 | */ |
492eb21b TH |
862 | #define css_for_each_descendant_post(pos, css) \ |
863 | for ((pos) = css_next_descendant_post(NULL, (css)); (pos); \ | |
864 | (pos) = css_next_descendant_post((pos), (css))) | |
574bd9f7 | 865 | |
72ec7029 TH |
866 | /* A css_task_iter should be treated as an opaque object */ |
867 | struct css_task_iter { | |
3ebb2b6e TH |
868 | struct cgroup_subsys *ss; |
869 | ||
0f0a2b4f TH |
870 | struct list_head *cset_pos; |
871 | struct list_head *cset_head; | |
872 | ||
873 | struct list_head *task_pos; | |
874 | struct list_head *tasks_head; | |
875 | struct list_head *mg_tasks_head; | |
817929ec PM |
876 | }; |
877 | ||
72ec7029 TH |
878 | void css_task_iter_start(struct cgroup_subsys_state *css, |
879 | struct css_task_iter *it); | |
880 | struct task_struct *css_task_iter_next(struct css_task_iter *it); | |
881 | void css_task_iter_end(struct css_task_iter *it); | |
e535837b | 882 | |
31583bb0 | 883 | int cgroup_attach_task_all(struct task_struct *from, struct task_struct *); |
8cc99345 | 884 | int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from); |
31583bb0 | 885 | |
5a17f543 TH |
886 | struct cgroup_subsys_state *css_tryget_from_dir(struct dentry *dentry, |
887 | struct cgroup_subsys *ss); | |
38460b48 | 888 | |
ddbcc7e8 PM |
889 | #else /* !CONFIG_CGROUPS */ |
890 | ||
891 | static inline int cgroup_init_early(void) { return 0; } | |
892 | static inline int cgroup_init(void) { return 0; } | |
b4f48b63 | 893 | static inline void cgroup_fork(struct task_struct *p) {} |
817929ec | 894 | static inline void cgroup_post_fork(struct task_struct *p) {} |
1ec41830 | 895 | static inline void cgroup_exit(struct task_struct *p) {} |
ddbcc7e8 | 896 | |
846c7bb0 BS |
897 | static inline int cgroupstats_build(struct cgroupstats *stats, |
898 | struct dentry *dentry) | |
899 | { | |
900 | return -EINVAL; | |
901 | } | |
ddbcc7e8 | 902 | |
d7926ee3 | 903 | /* No cgroups - nothing to do */ |
31583bb0 MT |
904 | static inline int cgroup_attach_task_all(struct task_struct *from, |
905 | struct task_struct *t) | |
906 | { | |
907 | return 0; | |
908 | } | |
d7926ee3 | 909 | |
ddbcc7e8 PM |
910 | #endif /* !CONFIG_CGROUPS */ |
911 | ||
912 | #endif /* _LINUX_CGROUP_H */ |