1 #ifndef _LINUX_CGROUP_H
2 #define _LINUX_CGROUP_H
6 * Copyright (C) 2003 BULL SA
7 * Copyright (C) 2004-2006 Silicon Graphics, Inc.
11 #include <linux/sched.h>
12 #include <linux/kref.h>
13 #include <linux/cpumask.h>
14 #include <linux/nodemask.h>
15 #include <linux/rcupdate.h>
16 #include <linux/cgroupstats.h>
17 #include <linux/prio_heap.h>
25 extern int cgroup_init_early(void);
26 extern int cgroup_init(void);
27 extern void cgroup_init_smp(void);
28 extern void cgroup_lock(void);
29 extern void cgroup_unlock(void);
30 extern void cgroup_fork(struct task_struct
*p
);
31 extern void cgroup_fork_callbacks(struct task_struct
*p
);
32 extern void cgroup_post_fork(struct task_struct
*p
);
33 extern void cgroup_exit(struct task_struct
*p
, int run_callbacks
);
34 extern int cgroupstats_build(struct cgroupstats
*stats
,
35 struct dentry
*dentry
);
37 extern struct file_operations proc_cgroup_operations
;
39 /* Define the enumeration of all cgroup subsystems */
40 #define SUBSYS(_x) _x ## _subsys_id,
41 enum cgroup_subsys_id
{
42 #include <linux/cgroup_subsys.h>
47 /* Per-subsystem/per-cgroup state maintained by the system. */
48 struct cgroup_subsys_state
{
49 /* The cgroup that this subsystem is attached to. Useful
50 * for subsystems that want to know about the cgroup
51 * hierarchy structure */
52 struct cgroup
*cgroup
;
54 /* State maintained by the cgroup system to allow
55 * subsystems to be "busy". Should be accessed via css_get()
63 /* bits in struct cgroup_subsys_state flags field */
65 CSS_ROOT
, /* This CSS is the root of the subsystem */
69 * Call css_get() to hold a reference on the cgroup;
73 static inline void css_get(struct cgroup_subsys_state
*css
)
75 /* We don't need to reference count the root state */
76 if (!test_bit(CSS_ROOT
, &css
->flags
))
77 atomic_inc(&css
->refcnt
);
80 * css_put() should be called to release a reference taken by
84 extern void __css_put(struct cgroup_subsys_state
*css
);
85 static inline void css_put(struct cgroup_subsys_state
*css
)
87 if (!test_bit(CSS_ROOT
, &css
->flags
))
91 /* bits in struct cgroup flags field */
93 /* Control Group is dead */
95 /* Control Group has previously had a child cgroup or a task,
96 * but no longer (only if CGRP_NOTIFY_ON_RELEASE is set) */
98 /* Control Group requires release notifications to userspace */
99 CGRP_NOTIFY_ON_RELEASE
,
103 unsigned long flags
; /* "unsigned long" so bitops work */
105 /* count users of this cgroup. >0 means busy, but doesn't
106 * necessarily indicate the number of tasks in the
111 * We link our 'sibling' struct into our parent's 'children'.
112 * Our children link their 'sibling' into our 'children'.
114 struct list_head sibling
; /* my parent's children */
115 struct list_head children
; /* my children */
117 struct cgroup
*parent
; /* my parent */
118 struct dentry
*dentry
; /* cgroup fs entry */
120 /* Private pointers for each registered subsystem */
121 struct cgroup_subsys_state
*subsys
[CGROUP_SUBSYS_COUNT
];
123 struct cgroupfs_root
*root
;
124 struct cgroup
*top_cgroup
;
127 * List of cg_cgroup_links pointing at css_sets with
128 * tasks in this cgroup. Protected by css_set_lock
130 struct list_head css_sets
;
133 * Linked list running through all cgroups that can
134 * potentially be reaped by the release agent. Protected by
137 struct list_head release_list
;
140 /* A css_set is a structure holding pointers to a set of
141 * cgroup_subsys_state objects. This saves space in the task struct
142 * object and speeds up fork()/exit(), since a single inc/dec and a
143 * list_add()/del() can bump the reference count on the entire
144 * cgroup set for a task.
149 /* Reference count */
153 * List running through all cgroup groups. Protected by
156 struct list_head list
;
159 * List running through all tasks using this cgroup
160 * group. Protected by css_set_lock
162 struct list_head tasks
;
165 * List of cg_cgroup_link objects on link chains from
166 * cgroups referenced from this css_set. Protected by
169 struct list_head cg_links
;
172 * Set of subsystem states, one for each subsystem. This array
173 * is immutable after creation apart from the init_css_set
174 * during subsystem registration (at boot time).
176 struct cgroup_subsys_state
*subsys
[CGROUP_SUBSYS_COUNT
];
181 * cgroup_map_cb is an abstract callback API for reporting map-valued
185 struct cgroup_map_cb
{
186 int (*fill
)(struct cgroup_map_cb
*cb
, const char *key
, u64 value
);
192 * The files in the cgroup filesystem mostly have a very simple read/write
193 * handling, some common function will take care of it. Nevertheless some cases
194 * (read tasks) are special and therefore I define this structure for every
198 * When reading/writing to a file:
199 * - the cgroup to use is file->f_dentry->d_parent->d_fsdata
200 * - the 'cftype' of the file is file->f_dentry->d_fsdata
203 #define MAX_CFTYPE_NAME 64
205 /* By convention, the name should begin with the name of the
206 * subsystem, followed by a period */
207 char name
[MAX_CFTYPE_NAME
];
209 int (*open
) (struct inode
*inode
, struct file
*file
);
210 ssize_t (*read
) (struct cgroup
*cgrp
, struct cftype
*cft
,
212 char __user
*buf
, size_t nbytes
, loff_t
*ppos
);
214 * read_u64() is a shortcut for the common case of returning a
215 * single integer. Use it in place of read()
217 u64 (*read_u64
) (struct cgroup
*cgrp
, struct cftype
*cft
);
219 * read_s64() is a signed version of read_u64()
221 s64 (*read_s64
) (struct cgroup
*cgrp
, struct cftype
*cft
);
223 * read_map() is used for defining a map of key/value
224 * pairs. It should call cb->fill(cb, key, value) for each
225 * entry. The key/value pairs (and their ordering) should not
226 * change between reboots.
228 int (*read_map
) (struct cgroup
*cont
, struct cftype
*cft
,
229 struct cgroup_map_cb
*cb
);
231 ssize_t (*write
) (struct cgroup
*cgrp
, struct cftype
*cft
,
233 const char __user
*buf
, size_t nbytes
, loff_t
*ppos
);
236 * write_u64() is a shortcut for the common case of accepting
237 * a single integer (as parsed by simple_strtoull) from
238 * userspace. Use in place of write(); return 0 or error.
240 int (*write_u64
) (struct cgroup
*cgrp
, struct cftype
*cft
, u64 val
);
242 * write_s64() is a signed version of write_u64()
244 int (*write_s64
) (struct cgroup
*cgrp
, struct cftype
*cft
, s64 val
);
246 int (*release
) (struct inode
*inode
, struct file
*file
);
249 struct cgroup_scanner
{
251 int (*test_task
)(struct task_struct
*p
, struct cgroup_scanner
*scan
);
252 void (*process_task
)(struct task_struct
*p
,
253 struct cgroup_scanner
*scan
);
254 struct ptr_heap
*heap
;
257 /* Add a new file to the given cgroup directory. Should only be
258 * called by subsystems from within a populate() method */
259 int cgroup_add_file(struct cgroup
*cgrp
, struct cgroup_subsys
*subsys
,
260 const struct cftype
*cft
);
262 /* Add a set of new files to the given cgroup directory. Should
263 * only be called by subsystems from within a populate() method */
264 int cgroup_add_files(struct cgroup
*cgrp
,
265 struct cgroup_subsys
*subsys
,
266 const struct cftype cft
[],
269 int cgroup_is_removed(const struct cgroup
*cgrp
);
271 int cgroup_path(const struct cgroup
*cgrp
, char *buf
, int buflen
);
273 int cgroup_task_count(const struct cgroup
*cgrp
);
275 /* Return true if the cgroup is a descendant of the current cgroup */
276 int cgroup_is_descendant(const struct cgroup
*cgrp
);
278 /* Control Group subsystem type. See Documentation/cgroups.txt for details */
280 struct cgroup_subsys
{
281 struct cgroup_subsys_state
*(*create
)(struct cgroup_subsys
*ss
,
282 struct cgroup
*cgrp
);
283 void (*pre_destroy
)(struct cgroup_subsys
*ss
, struct cgroup
*cgrp
);
284 void (*destroy
)(struct cgroup_subsys
*ss
, struct cgroup
*cgrp
);
285 int (*can_attach
)(struct cgroup_subsys
*ss
,
286 struct cgroup
*cgrp
, struct task_struct
*tsk
);
287 void (*attach
)(struct cgroup_subsys
*ss
, struct cgroup
*cgrp
,
288 struct cgroup
*old_cgrp
, struct task_struct
*tsk
);
289 void (*fork
)(struct cgroup_subsys
*ss
, struct task_struct
*task
);
290 void (*exit
)(struct cgroup_subsys
*ss
, struct task_struct
*task
);
291 int (*populate
)(struct cgroup_subsys
*ss
,
292 struct cgroup
*cgrp
);
293 void (*post_clone
)(struct cgroup_subsys
*ss
, struct cgroup
*cgrp
);
294 void (*bind
)(struct cgroup_subsys
*ss
, struct cgroup
*root
);
299 #define MAX_CGROUP_TYPE_NAMELEN 32
302 /* Protected by RCU */
303 struct cgroupfs_root
*root
;
305 struct list_head sibling
;
310 #define SUBSYS(_x) extern struct cgroup_subsys _x ## _subsys;
311 #include <linux/cgroup_subsys.h>
314 static inline struct cgroup_subsys_state
*cgroup_subsys_state(
315 struct cgroup
*cgrp
, int subsys_id
)
317 return cgrp
->subsys
[subsys_id
];
320 static inline struct cgroup_subsys_state
*task_subsys_state(
321 struct task_struct
*task
, int subsys_id
)
323 return rcu_dereference(task
->cgroups
->subsys
[subsys_id
]);
326 static inline struct cgroup
* task_cgroup(struct task_struct
*task
,
329 return task_subsys_state(task
, subsys_id
)->cgroup
;
332 int cgroup_clone(struct task_struct
*tsk
, struct cgroup_subsys
*ss
);
334 /* A cgroup_iter should be treated as an opaque object */
336 struct list_head
*cg_link
;
337 struct list_head
*task
;
340 /* To iterate across the tasks in a cgroup:
342 * 1) call cgroup_iter_start to intialize an iterator
344 * 2) call cgroup_iter_next() to retrieve member tasks until it
345 * returns NULL or until you want to end the iteration
347 * 3) call cgroup_iter_end() to destroy the iterator.
349 * Or, call cgroup_scan_tasks() to iterate through every task in a cpuset.
350 * - cgroup_scan_tasks() holds the css_set_lock when calling the test_task()
351 * callback, but not while calling the process_task() callback.
353 void cgroup_iter_start(struct cgroup
*cgrp
, struct cgroup_iter
*it
);
354 struct task_struct
*cgroup_iter_next(struct cgroup
*cgrp
,
355 struct cgroup_iter
*it
);
356 void cgroup_iter_end(struct cgroup
*cgrp
, struct cgroup_iter
*it
);
357 int cgroup_scan_tasks(struct cgroup_scanner
*scan
);
358 int cgroup_attach_task(struct cgroup
*, struct task_struct
*);
360 #else /* !CONFIG_CGROUPS */
362 static inline int cgroup_init_early(void) { return 0; }
363 static inline int cgroup_init(void) { return 0; }
364 static inline void cgroup_init_smp(void) {}
365 static inline void cgroup_fork(struct task_struct
*p
) {}
366 static inline void cgroup_fork_callbacks(struct task_struct
*p
) {}
367 static inline void cgroup_post_fork(struct task_struct
*p
) {}
368 static inline void cgroup_exit(struct task_struct
*p
, int callbacks
) {}
370 static inline void cgroup_lock(void) {}
371 static inline void cgroup_unlock(void) {}
372 static inline int cgroupstats_build(struct cgroupstats
*stats
,
373 struct dentry
*dentry
)
378 #endif /* !CONFIG_CGROUPS */
380 #endif /* _LINUX_CGROUP_H */