blob: 0008dee665148850f2b46af2918b23635cedc7c0 [file] [log] [blame]
Paul Menageddbcc7e2007-10-18 23:39:30 -07001#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>
Paul Menageddbcc7e2007-10-18 23:39:30 -070012#include <linux/cpumask.h>
13#include <linux/nodemask.h>
14#include <linux/rcupdate.h>
Balbir Singh846c7bb2007-10-18 23:39:44 -070015#include <linux/cgroupstats.h>
Cliff Wickman31a7df02008-02-07 00:14:42 -080016#include <linux/prio_heap.h>
Paul Menagecc31edc2008-10-18 20:28:04 -070017#include <linux/rwsem.h>
KAMEZAWA Hiroyuki38460b42009-04-02 16:57:25 -070018#include <linux/idr.h>
Paul Menageddbcc7e2007-10-18 23:39:30 -070019
20#ifdef CONFIG_CGROUPS
21
22struct cgroupfs_root;
23struct cgroup_subsys;
24struct inode;
Paul Menage84eea842008-07-25 01:47:00 -070025struct cgroup;
KAMEZAWA Hiroyuki38460b42009-04-02 16:57:25 -070026struct css_id;
Paul Menageddbcc7e2007-10-18 23:39:30 -070027
28extern int cgroup_init_early(void);
29extern int cgroup_init(void);
Paul Menageddbcc7e2007-10-18 23:39:30 -070030extern void cgroup_lock(void);
Paul Menage84eea842008-07-25 01:47:00 -070031extern bool cgroup_lock_live_group(struct cgroup *cgrp);
Paul Menageddbcc7e2007-10-18 23:39:30 -070032extern void cgroup_unlock(void);
Paul Menageb4f48b62007-10-18 23:39:33 -070033extern void cgroup_fork(struct task_struct *p);
34extern void cgroup_fork_callbacks(struct task_struct *p);
Paul Menage817929e2007-10-18 23:39:36 -070035extern void cgroup_post_fork(struct task_struct *p);
Paul Menageb4f48b62007-10-18 23:39:33 -070036extern void cgroup_exit(struct task_struct *p, int run_callbacks);
Balbir Singh846c7bb2007-10-18 23:39:44 -070037extern int cgroupstats_build(struct cgroupstats *stats,
38 struct dentry *dentry);
Paul Menageddbcc7e2007-10-18 23:39:30 -070039
Alexey Dobriyan828c0952009-10-01 15:43:56 -070040extern const struct file_operations proc_cgroup_operations;
Paul Menagea4243162007-10-18 23:39:35 -070041
Paul Menage817929e2007-10-18 23:39:36 -070042/* Define the enumeration of all cgroup subsystems */
43#define SUBSYS(_x) _x ## _subsys_id,
44enum cgroup_subsys_id {
45#include <linux/cgroup_subsys.h>
46 CGROUP_SUBSYS_COUNT
47};
48#undef SUBSYS
49
Paul Menageddbcc7e2007-10-18 23:39:30 -070050/* Per-subsystem/per-cgroup state maintained by the system. */
51struct cgroup_subsys_state {
Paul Menaged20a3902009-04-02 16:57:22 -070052 /*
53 * The cgroup that this subsystem is attached to. Useful
Paul Menageddbcc7e2007-10-18 23:39:30 -070054 * for subsystems that want to know about the cgroup
Paul Menaged20a3902009-04-02 16:57:22 -070055 * hierarchy structure
56 */
Paul Menageddbcc7e2007-10-18 23:39:30 -070057 struct cgroup *cgroup;
58
Paul Menaged20a3902009-04-02 16:57:22 -070059 /*
60 * State maintained by the cgroup system to allow subsystems
Paul Menagee7c5ec92009-01-07 18:08:38 -080061 * to be "busy". Should be accessed via css_get(),
Paul Menaged20a3902009-04-02 16:57:22 -070062 * css_tryget() and and css_put().
63 */
Paul Menageddbcc7e2007-10-18 23:39:30 -070064
65 atomic_t refcnt;
66
67 unsigned long flags;
KAMEZAWA Hiroyuki38460b42009-04-02 16:57:25 -070068 /* ID for this css, if possible */
69 struct css_id *id;
Paul Menageddbcc7e2007-10-18 23:39:30 -070070};
71
72/* bits in struct cgroup_subsys_state flags field */
73enum {
74 CSS_ROOT, /* This CSS is the root of the subsystem */
Paul Menagee7c5ec92009-01-07 18:08:38 -080075 CSS_REMOVED, /* This CSS is dead */
Paul Menageddbcc7e2007-10-18 23:39:30 -070076};
77
78/*
Paul Menagee7c5ec92009-01-07 18:08:38 -080079 * Call css_get() to hold a reference on the css; it can be used
80 * for a reference obtained via:
81 * - an existing ref-counted reference to the css
82 * - task->cgroups for a locked task
Paul Menageddbcc7e2007-10-18 23:39:30 -070083 */
84
85static inline void css_get(struct cgroup_subsys_state *css)
86{
87 /* We don't need to reference count the root state */
88 if (!test_bit(CSS_ROOT, &css->flags))
89 atomic_inc(&css->refcnt);
90}
Paul Menagee7c5ec92009-01-07 18:08:38 -080091
92static inline bool css_is_removed(struct cgroup_subsys_state *css)
93{
94 return test_bit(CSS_REMOVED, &css->flags);
95}
96
97/*
98 * Call css_tryget() to take a reference on a css if your existing
99 * (known-valid) reference isn't already ref-counted. Returns false if
100 * the css has been destroyed.
101 */
102
103static inline bool css_tryget(struct cgroup_subsys_state *css)
104{
105 if (test_bit(CSS_ROOT, &css->flags))
106 return true;
107 while (!atomic_inc_not_zero(&css->refcnt)) {
108 if (test_bit(CSS_REMOVED, &css->flags))
109 return false;
Paul Menage804b3c22009-01-29 14:25:21 -0800110 cpu_relax();
Paul Menagee7c5ec92009-01-07 18:08:38 -0800111 }
112 return true;
113}
114
Paul Menageddbcc7e2007-10-18 23:39:30 -0700115/*
116 * css_put() should be called to release a reference taken by
Paul Menagee7c5ec92009-01-07 18:08:38 -0800117 * css_get() or css_tryget()
Paul Menageddbcc7e2007-10-18 23:39:30 -0700118 */
119
Paul Menage81a6a5c2007-10-18 23:39:38 -0700120extern void __css_put(struct cgroup_subsys_state *css);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700121static inline void css_put(struct cgroup_subsys_state *css)
122{
123 if (!test_bit(CSS_ROOT, &css->flags))
Paul Menage81a6a5c2007-10-18 23:39:38 -0700124 __css_put(css);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700125}
126
Paul Menage3116f0e2008-04-29 01:00:04 -0700127/* bits in struct cgroup flags field */
128enum {
129 /* Control Group is dead */
130 CGRP_REMOVED,
Paul Menaged20a3902009-04-02 16:57:22 -0700131 /*
132 * Control Group has previously had a child cgroup or a task,
133 * but no longer (only if CGRP_NOTIFY_ON_RELEASE is set)
134 */
Paul Menage3116f0e2008-04-29 01:00:04 -0700135 CGRP_RELEASABLE,
136 /* Control Group requires release notifications to userspace */
137 CGRP_NOTIFY_ON_RELEASE,
KAMEZAWA Hiroyukiec64f512009-04-02 16:57:26 -0700138 /*
139 * A thread in rmdir() is wating for this cgroup.
140 */
141 CGRP_WAIT_ON_RMDIR,
Paul Menage3116f0e2008-04-29 01:00:04 -0700142};
143
Ben Blum72a8cb32009-09-23 15:56:27 -0700144/* which pidlist file are we talking about? */
145enum cgroup_filetype {
146 CGROUP_FILE_PROCS,
147 CGROUP_FILE_TASKS,
148};
149
150/*
151 * A pidlist is a list of pids that virtually represents the contents of one
152 * of the cgroup files ("procs" or "tasks"). We keep a list of such pidlists,
153 * a pair (one each for procs, tasks) for each pid namespace that's relevant
154 * to the cgroup.
155 */
Ben Blum102a7752009-09-23 15:56:26 -0700156struct cgroup_pidlist {
Ben Blum72a8cb32009-09-23 15:56:27 -0700157 /*
158 * used to find which pidlist is wanted. doesn't change as long as
159 * this particular list stays in the list.
160 */
161 struct { enum cgroup_filetype type; struct pid_namespace *ns; } key;
Ben Blum102a7752009-09-23 15:56:26 -0700162 /* array of xids */
163 pid_t *list;
164 /* how many elements the above list has */
165 int length;
166 /* how many files are using the current array */
167 int use_count;
Ben Blum72a8cb32009-09-23 15:56:27 -0700168 /* each of these stored in a list by its cgroup */
169 struct list_head links;
170 /* pointer to the cgroup we belong to, for list removal purposes */
171 struct cgroup *owner;
172 /* protects the other fields */
173 struct rw_semaphore mutex;
Ben Blum102a7752009-09-23 15:56:26 -0700174};
175
Paul Menageddbcc7e2007-10-18 23:39:30 -0700176struct cgroup {
177 unsigned long flags; /* "unsigned long" so bitops work */
178
Paul Menaged20a3902009-04-02 16:57:22 -0700179 /*
180 * count users of this cgroup. >0 means busy, but doesn't
181 * necessarily indicate the number of tasks in the cgroup
182 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700183 atomic_t count;
184
185 /*
186 * We link our 'sibling' struct into our parent's 'children'.
187 * Our children link their 'sibling' into our 'children'.
188 */
189 struct list_head sibling; /* my parent's children */
190 struct list_head children; /* my children */
191
Paul Menaged20a3902009-04-02 16:57:22 -0700192 struct cgroup *parent; /* my parent */
Paul Menagea47295e2009-01-07 18:07:44 -0800193 struct dentry *dentry; /* cgroup fs entry, RCU protected */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700194
195 /* Private pointers for each registered subsystem */
196 struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT];
197
198 struct cgroupfs_root *root;
199 struct cgroup *top_cgroup;
Paul Menage817929e2007-10-18 23:39:36 -0700200
201 /*
202 * List of cg_cgroup_links pointing at css_sets with
203 * tasks in this cgroup. Protected by css_set_lock
204 */
205 struct list_head css_sets;
Paul Menage81a6a5c2007-10-18 23:39:38 -0700206
207 /*
208 * Linked list running through all cgroups that can
209 * potentially be reaped by the release agent. Protected by
210 * release_list_lock
211 */
212 struct list_head release_list;
Paul Menagecc31edc2008-10-18 20:28:04 -0700213
Ben Blum72a8cb32009-09-23 15:56:27 -0700214 /*
215 * list of pidlists, up to two for each namespace (one for procs, one
216 * for tasks); created on demand.
217 */
218 struct list_head pidlists;
219 struct mutex pidlist_mutex;
Paul Menagea47295e2009-01-07 18:07:44 -0800220
221 /* For RCU-protected deletion */
222 struct rcu_head rcu_head;
Paul Menage817929e2007-10-18 23:39:36 -0700223};
224
Paul Menaged20a3902009-04-02 16:57:22 -0700225/*
226 * A css_set is a structure holding pointers to a set of
Paul Menage817929e2007-10-18 23:39:36 -0700227 * cgroup_subsys_state objects. This saves space in the task struct
228 * object and speeds up fork()/exit(), since a single inc/dec and a
Paul Menaged20a3902009-04-02 16:57:22 -0700229 * list_add()/del() can bump the reference count on the entire cgroup
230 * set for a task.
Paul Menage817929e2007-10-18 23:39:36 -0700231 */
232
233struct css_set {
234
235 /* Reference count */
Lai Jiangshan146aa1b2008-10-18 20:28:03 -0700236 atomic_t refcount;
Paul Menage817929e2007-10-18 23:39:36 -0700237
238 /*
Li Zefan472b1052008-04-29 01:00:11 -0700239 * List running through all cgroup groups in the same hash
240 * slot. Protected by css_set_lock
241 */
242 struct hlist_node hlist;
243
244 /*
Paul Menage817929e2007-10-18 23:39:36 -0700245 * List running through all tasks using this cgroup
246 * group. Protected by css_set_lock
247 */
248 struct list_head tasks;
249
250 /*
251 * List of cg_cgroup_link objects on link chains from
252 * cgroups referenced from this css_set. Protected by
253 * css_set_lock
254 */
255 struct list_head cg_links;
256
257 /*
258 * Set of subsystem states, one for each subsystem. This array
259 * is immutable after creation apart from the init_css_set
260 * during subsystem registration (at boot time).
261 */
262 struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT];
Ben Blumc3783692009-09-23 15:56:29 -0700263
264 /* For RCU-protected deletion */
265 struct rcu_head rcu_head;
Paul Menageddbcc7e2007-10-18 23:39:30 -0700266};
267
Paul Menage91796562008-04-29 01:00:01 -0700268/*
269 * cgroup_map_cb is an abstract callback API for reporting map-valued
270 * control files
271 */
272
273struct cgroup_map_cb {
274 int (*fill)(struct cgroup_map_cb *cb, const char *key, u64 value);
275 void *state;
276};
277
Paul Menaged20a3902009-04-02 16:57:22 -0700278/*
279 * struct cftype: handler definitions for cgroup control files
Paul Menageddbcc7e2007-10-18 23:39:30 -0700280 *
281 * When reading/writing to a file:
Li Zefana043e3b2008-02-23 15:24:09 -0800282 * - the cgroup to use is file->f_dentry->d_parent->d_fsdata
Paul Menageddbcc7e2007-10-18 23:39:30 -0700283 * - the 'cftype' of the file is file->f_dentry->d_fsdata
284 */
285
286#define MAX_CFTYPE_NAME 64
287struct cftype {
Paul Menaged20a3902009-04-02 16:57:22 -0700288 /*
289 * By convention, the name should begin with the name of the
290 * subsystem, followed by a period
291 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700292 char name[MAX_CFTYPE_NAME];
293 int private;
Li Zefan099fca32009-04-02 16:57:29 -0700294 /*
295 * If not 0, file mode is set to this value, otherwise it will
296 * be figured out automatically
297 */
298 mode_t mode;
Paul Menagedb3b1492008-07-25 01:46:58 -0700299
300 /*
301 * If non-zero, defines the maximum length of string that can
302 * be passed to write_string; defaults to 64
303 */
304 size_t max_write_len;
305
Paul Menagece16b492008-07-25 01:46:57 -0700306 int (*open)(struct inode *inode, struct file *file);
307 ssize_t (*read)(struct cgroup *cgrp, struct cftype *cft,
308 struct file *file,
309 char __user *buf, size_t nbytes, loff_t *ppos);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700310 /*
Paul Menagef4c753b2008-04-29 00:59:56 -0700311 * read_u64() is a shortcut for the common case of returning a
Paul Menageddbcc7e2007-10-18 23:39:30 -0700312 * single integer. Use it in place of read()
313 */
Paul Menagece16b492008-07-25 01:46:57 -0700314 u64 (*read_u64)(struct cgroup *cgrp, struct cftype *cft);
Paul Menage91796562008-04-29 01:00:01 -0700315 /*
Paul Menagee73d2c62008-04-29 01:00:06 -0700316 * read_s64() is a signed version of read_u64()
317 */
Paul Menagece16b492008-07-25 01:46:57 -0700318 s64 (*read_s64)(struct cgroup *cgrp, struct cftype *cft);
Paul Menagee73d2c62008-04-29 01:00:06 -0700319 /*
Paul Menage91796562008-04-29 01:00:01 -0700320 * read_map() is used for defining a map of key/value
321 * pairs. It should call cb->fill(cb, key, value) for each
322 * entry. The key/value pairs (and their ordering) should not
323 * change between reboots.
324 */
Paul Menagece16b492008-07-25 01:46:57 -0700325 int (*read_map)(struct cgroup *cont, struct cftype *cft,
326 struct cgroup_map_cb *cb);
Serge E. Hallyn29486df2008-04-29 01:00:14 -0700327 /*
328 * read_seq_string() is used for outputting a simple sequence
329 * using seqfile.
330 */
Paul Menagece16b492008-07-25 01:46:57 -0700331 int (*read_seq_string)(struct cgroup *cont, struct cftype *cft,
332 struct seq_file *m);
Paul Menage91796562008-04-29 01:00:01 -0700333
Paul Menagece16b492008-07-25 01:46:57 -0700334 ssize_t (*write)(struct cgroup *cgrp, struct cftype *cft,
335 struct file *file,
336 const char __user *buf, size_t nbytes, loff_t *ppos);
Paul Menage355e0c42007-10-18 23:39:33 -0700337
338 /*
Paul Menagef4c753b2008-04-29 00:59:56 -0700339 * write_u64() is a shortcut for the common case of accepting
Paul Menage355e0c42007-10-18 23:39:33 -0700340 * a single integer (as parsed by simple_strtoull) from
341 * userspace. Use in place of write(); return 0 or error.
342 */
Paul Menagece16b492008-07-25 01:46:57 -0700343 int (*write_u64)(struct cgroup *cgrp, struct cftype *cft, u64 val);
Paul Menagee73d2c62008-04-29 01:00:06 -0700344 /*
345 * write_s64() is a signed version of write_u64()
346 */
Paul Menagece16b492008-07-25 01:46:57 -0700347 int (*write_s64)(struct cgroup *cgrp, struct cftype *cft, s64 val);
Paul Menage355e0c42007-10-18 23:39:33 -0700348
Pavel Emelyanovd447ea22008-04-29 01:00:08 -0700349 /*
Paul Menagedb3b1492008-07-25 01:46:58 -0700350 * write_string() is passed a nul-terminated kernelspace
351 * buffer of maximum length determined by max_write_len.
352 * Returns 0 or -ve error code.
353 */
354 int (*write_string)(struct cgroup *cgrp, struct cftype *cft,
355 const char *buffer);
356 /*
Pavel Emelyanovd447ea22008-04-29 01:00:08 -0700357 * trigger() callback can be used to get some kick from the
358 * userspace, when the actual string written is not important
359 * at all. The private field can be used to determine the
360 * kick type for multiplexing.
361 */
362 int (*trigger)(struct cgroup *cgrp, unsigned int event);
363
Paul Menagece16b492008-07-25 01:46:57 -0700364 int (*release)(struct inode *inode, struct file *file);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700365};
366
Cliff Wickman31a7df02008-02-07 00:14:42 -0800367struct cgroup_scanner {
368 struct cgroup *cg;
369 int (*test_task)(struct task_struct *p, struct cgroup_scanner *scan);
370 void (*process_task)(struct task_struct *p,
371 struct cgroup_scanner *scan);
372 struct ptr_heap *heap;
Li Zefanbd1a8ab2009-04-02 16:57:50 -0700373 void *data;
Cliff Wickman31a7df02008-02-07 00:14:42 -0800374};
375
Paul Menaged20a3902009-04-02 16:57:22 -0700376/*
377 * Add a new file to the given cgroup directory. Should only be
378 * called by subsystems from within a populate() method
379 */
Li Zefanffd2d882008-02-23 15:24:09 -0800380int cgroup_add_file(struct cgroup *cgrp, struct cgroup_subsys *subsys,
Paul Menageddbcc7e2007-10-18 23:39:30 -0700381 const struct cftype *cft);
382
Paul Menaged20a3902009-04-02 16:57:22 -0700383/*
384 * Add a set of new files to the given cgroup directory. Should
385 * only be called by subsystems from within a populate() method
386 */
Li Zefanffd2d882008-02-23 15:24:09 -0800387int cgroup_add_files(struct cgroup *cgrp,
Paul Menageddbcc7e2007-10-18 23:39:30 -0700388 struct cgroup_subsys *subsys,
389 const struct cftype cft[],
390 int count);
391
Li Zefanffd2d882008-02-23 15:24:09 -0800392int cgroup_is_removed(const struct cgroup *cgrp);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700393
Li Zefanffd2d882008-02-23 15:24:09 -0800394int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700395
Li Zefanffd2d882008-02-23 15:24:09 -0800396int cgroup_task_count(const struct cgroup *cgrp);
Paul Menagebbcb81d2007-10-18 23:39:32 -0700397
Grzegorz Nosek313e9242009-04-02 16:57:23 -0700398/* Return true if cgrp is a descendant of the task's cgroup */
399int cgroup_is_descendant(const struct cgroup *cgrp, struct task_struct *task);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700400
Thadeu Lima de Souza Cascardo21acb9c2009-02-04 10:12:08 +0100401/*
KAMEZAWA Hiroyuki88703262009-07-29 15:04:06 -0700402 * When the subsys has to access css and may add permanent refcnt to css,
403 * it should take care of racy conditions with rmdir(). Following set of
404 * functions, is for stop/restart rmdir if necessary.
405 * Because these will call css_get/put, "css" should be alive css.
406 *
407 * cgroup_exclude_rmdir();
408 * ...do some jobs which may access arbitrary empty cgroup
409 * cgroup_release_and_wakeup_rmdir();
410 *
411 * When someone removes a cgroup while cgroup_exclude_rmdir() holds it,
412 * it sleeps and cgroup_release_and_wakeup_rmdir() will wake him up.
413 */
414
415void cgroup_exclude_rmdir(struct cgroup_subsys_state *css);
416void cgroup_release_and_wakeup_rmdir(struct cgroup_subsys_state *css);
417
418/*
Thadeu Lima de Souza Cascardo21acb9c2009-02-04 10:12:08 +0100419 * Control Group subsystem type.
420 * See Documentation/cgroups/cgroups.txt for details
421 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700422
423struct cgroup_subsys {
424 struct cgroup_subsys_state *(*create)(struct cgroup_subsys *ss,
Li Zefanffd2d882008-02-23 15:24:09 -0800425 struct cgroup *cgrp);
KAMEZAWA Hiroyukiec64f512009-04-02 16:57:26 -0700426 int (*pre_destroy)(struct cgroup_subsys *ss, struct cgroup *cgrp);
Li Zefanffd2d882008-02-23 15:24:09 -0800427 void (*destroy)(struct cgroup_subsys *ss, struct cgroup *cgrp);
Ben Blumbe367d02009-09-23 15:56:31 -0700428 int (*can_attach)(struct cgroup_subsys *ss, struct cgroup *cgrp,
429 struct task_struct *tsk, bool threadgroup);
Li Zefanffd2d882008-02-23 15:24:09 -0800430 void (*attach)(struct cgroup_subsys *ss, struct cgroup *cgrp,
Ben Blumbe367d02009-09-23 15:56:31 -0700431 struct cgroup *old_cgrp, struct task_struct *tsk,
432 bool threadgroup);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700433 void (*fork)(struct cgroup_subsys *ss, struct task_struct *task);
434 void (*exit)(struct cgroup_subsys *ss, struct task_struct *task);
435 int (*populate)(struct cgroup_subsys *ss,
Li Zefanffd2d882008-02-23 15:24:09 -0800436 struct cgroup *cgrp);
437 void (*post_clone)(struct cgroup_subsys *ss, struct cgroup *cgrp);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700438 void (*bind)(struct cgroup_subsys *ss, struct cgroup *root);
Hugh Dickinse5991372009-01-06 14:39:22 -0800439
Paul Menageddbcc7e2007-10-18 23:39:30 -0700440 int subsys_id;
441 int active;
Paul Menage8bab8dd2008-04-04 14:29:57 -0700442 int disabled;
Paul Menageddbcc7e2007-10-18 23:39:30 -0700443 int early_init;
KAMEZAWA Hiroyuki38460b42009-04-02 16:57:25 -0700444 /*
445 * True if this subsys uses ID. ID is not available before cgroup_init()
446 * (not available in early_init time.)
447 */
448 bool use_id;
Paul Menageddbcc7e2007-10-18 23:39:30 -0700449#define MAX_CGROUP_TYPE_NAMELEN 32
450 const char *name;
451
Paul Menage999cd8a2009-01-07 18:08:36 -0800452 /*
453 * Protects sibling/children links of cgroups in this
454 * hierarchy, plus protects which hierarchy (or none) the
455 * subsystem is a part of (i.e. root/sibling). To avoid
456 * potential deadlocks, the following operations should not be
457 * undertaken while holding any hierarchy_mutex:
458 *
459 * - allocating memory
460 * - initiating hotplug events
461 */
462 struct mutex hierarchy_mutex;
Li Zefancfebe562009-02-11 13:04:36 -0800463 struct lock_class_key subsys_key;
Paul Menageddbcc7e2007-10-18 23:39:30 -0700464
Paul Menage999cd8a2009-01-07 18:08:36 -0800465 /*
466 * Link to parent, and list entry in parent's children.
467 * Protected by this->hierarchy_mutex and cgroup_lock()
468 */
469 struct cgroupfs_root *root;
Paul Menageddbcc7e2007-10-18 23:39:30 -0700470 struct list_head sibling;
KAMEZAWA Hiroyuki38460b42009-04-02 16:57:25 -0700471 /* used when use_id == true */
472 struct idr idr;
473 spinlock_t id_lock;
Paul Menageddbcc7e2007-10-18 23:39:30 -0700474};
475
476#define SUBSYS(_x) extern struct cgroup_subsys _x ## _subsys;
477#include <linux/cgroup_subsys.h>
478#undef SUBSYS
479
480static inline struct cgroup_subsys_state *cgroup_subsys_state(
Li Zefanffd2d882008-02-23 15:24:09 -0800481 struct cgroup *cgrp, int subsys_id)
Paul Menageddbcc7e2007-10-18 23:39:30 -0700482{
Li Zefanffd2d882008-02-23 15:24:09 -0800483 return cgrp->subsys[subsys_id];
Paul Menageddbcc7e2007-10-18 23:39:30 -0700484}
485
486static inline struct cgroup_subsys_state *task_subsys_state(
487 struct task_struct *task, int subsys_id)
488{
Paul Menage817929e2007-10-18 23:39:36 -0700489 return rcu_dereference(task->cgroups->subsys[subsys_id]);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700490}
491
492static inline struct cgroup* task_cgroup(struct task_struct *task,
493 int subsys_id)
494{
495 return task_subsys_state(task, subsys_id)->cgroup;
496}
497
Serge E. Hallyne885dcd2008-07-25 01:47:06 -0700498int cgroup_clone(struct task_struct *tsk, struct cgroup_subsys *ss,
499 char *nodename);
Paul Menage697f4162007-10-18 23:39:34 -0700500
Paul Menage817929e2007-10-18 23:39:36 -0700501/* A cgroup_iter should be treated as an opaque object */
502struct cgroup_iter {
503 struct list_head *cg_link;
504 struct list_head *task;
505};
506
Paul Menaged20a3902009-04-02 16:57:22 -0700507/*
508 * To iterate across the tasks in a cgroup:
Paul Menage817929e2007-10-18 23:39:36 -0700509 *
510 * 1) call cgroup_iter_start to intialize an iterator
511 *
512 * 2) call cgroup_iter_next() to retrieve member tasks until it
513 * returns NULL or until you want to end the iteration
514 *
515 * 3) call cgroup_iter_end() to destroy the iterator.
Cliff Wickman31a7df02008-02-07 00:14:42 -0800516 *
Paul Menaged20a3902009-04-02 16:57:22 -0700517 * Or, call cgroup_scan_tasks() to iterate through every task in a
518 * cgroup - cgroup_scan_tasks() holds the css_set_lock when calling
519 * the test_task() callback, but not while calling the process_task()
520 * callback.
Paul Menage817929e2007-10-18 23:39:36 -0700521 */
Li Zefanffd2d882008-02-23 15:24:09 -0800522void cgroup_iter_start(struct cgroup *cgrp, struct cgroup_iter *it);
523struct task_struct *cgroup_iter_next(struct cgroup *cgrp,
Paul Menage817929e2007-10-18 23:39:36 -0700524 struct cgroup_iter *it);
Li Zefanffd2d882008-02-23 15:24:09 -0800525void cgroup_iter_end(struct cgroup *cgrp, struct cgroup_iter *it);
Cliff Wickman31a7df02008-02-07 00:14:42 -0800526int cgroup_scan_tasks(struct cgroup_scanner *scan);
Cliff Wickman956db3c2008-02-07 00:14:43 -0800527int cgroup_attach_task(struct cgroup *, struct task_struct *);
Paul Menage817929e2007-10-18 23:39:36 -0700528
KAMEZAWA Hiroyuki38460b42009-04-02 16:57:25 -0700529/*
530 * CSS ID is ID for cgroup_subsys_state structs under subsys. This only works
531 * if cgroup_subsys.use_id == true. It can be used for looking up and scanning.
532 * CSS ID is assigned at cgroup allocation (create) automatically
533 * and removed when subsys calls free_css_id() function. This is because
534 * the lifetime of cgroup_subsys_state is subsys's matter.
535 *
536 * Looking up and scanning function should be called under rcu_read_lock().
537 * Taking cgroup_mutex()/hierarchy_mutex() is not necessary for following calls.
538 * But the css returned by this routine can be "not populated yet" or "being
539 * destroyed". The caller should check css and cgroup's status.
540 */
541
542/*
543 * Typically Called at ->destroy(), or somewhere the subsys frees
544 * cgroup_subsys_state.
545 */
546void free_css_id(struct cgroup_subsys *ss, struct cgroup_subsys_state *css);
547
548/* Find a cgroup_subsys_state which has given ID */
549
550struct cgroup_subsys_state *css_lookup(struct cgroup_subsys *ss, int id);
551
552/*
553 * Get a cgroup whose id is greater than or equal to id under tree of root.
554 * Returning a cgroup_subsys_state or NULL.
555 */
556struct cgroup_subsys_state *css_get_next(struct cgroup_subsys *ss, int id,
557 struct cgroup_subsys_state *root, int *foundid);
558
559/* Returns true if root is ancestor of cg */
560bool css_is_ancestor(struct cgroup_subsys_state *cg,
KAMEZAWA Hiroyuki0b7f5692009-04-02 16:57:38 -0700561 const struct cgroup_subsys_state *root);
KAMEZAWA Hiroyuki38460b42009-04-02 16:57:25 -0700562
563/* Get id and depth of css */
564unsigned short css_id(struct cgroup_subsys_state *css);
565unsigned short css_depth(struct cgroup_subsys_state *css);
566
Paul Menageddbcc7e2007-10-18 23:39:30 -0700567#else /* !CONFIG_CGROUPS */
568
569static inline int cgroup_init_early(void) { return 0; }
570static inline int cgroup_init(void) { return 0; }
Paul Menageb4f48b62007-10-18 23:39:33 -0700571static inline void cgroup_fork(struct task_struct *p) {}
572static inline void cgroup_fork_callbacks(struct task_struct *p) {}
Paul Menage817929e2007-10-18 23:39:36 -0700573static inline void cgroup_post_fork(struct task_struct *p) {}
Paul Menageb4f48b62007-10-18 23:39:33 -0700574static inline void cgroup_exit(struct task_struct *p, int callbacks) {}
Paul Menageddbcc7e2007-10-18 23:39:30 -0700575
576static inline void cgroup_lock(void) {}
577static inline void cgroup_unlock(void) {}
Balbir Singh846c7bb2007-10-18 23:39:44 -0700578static inline int cgroupstats_build(struct cgroupstats *stats,
579 struct dentry *dentry)
580{
581 return -EINVAL;
582}
Paul Menageddbcc7e2007-10-18 23:39:30 -0700583
584#endif /* !CONFIG_CGROUPS */
585
Paul Menageddbcc7e2007-10-18 23:39:30 -0700586#endif /* _LINUX_CGROUP_H */