]>
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> | |
12 | #include <linux/kref.h> | |
13 | #include <linux/cpumask.h> | |
14 | #include <linux/nodemask.h> | |
15 | #include <linux/rcupdate.h> | |
846c7bb0 | 16 | #include <linux/cgroupstats.h> |
31a7df01 | 17 | #include <linux/prio_heap.h> |
ddbcc7e8 PM |
18 | |
19 | #ifdef CONFIG_CGROUPS | |
20 | ||
21 | struct cgroupfs_root; | |
22 | struct cgroup_subsys; | |
23 | struct inode; | |
24 | ||
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); | |
b4f48b63 PM |
30 | extern void cgroup_fork(struct task_struct *p); |
31 | extern void cgroup_fork_callbacks(struct task_struct *p); | |
817929ec | 32 | extern void cgroup_post_fork(struct task_struct *p); |
b4f48b63 | 33 | extern void cgroup_exit(struct task_struct *p, int run_callbacks); |
846c7bb0 BS |
34 | extern int cgroupstats_build(struct cgroupstats *stats, |
35 | struct dentry *dentry); | |
ddbcc7e8 | 36 | |
a424316c PM |
37 | extern struct file_operations proc_cgroup_operations; |
38 | ||
817929ec PM |
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> | |
43 | CGROUP_SUBSYS_COUNT | |
44 | }; | |
45 | #undef SUBSYS | |
46 | ||
ddbcc7e8 PM |
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; | |
53 | ||
54 | /* State maintained by the cgroup system to allow | |
55 | * subsystems to be "busy". Should be accessed via css_get() | |
56 | * and css_put() */ | |
57 | ||
58 | atomic_t refcnt; | |
59 | ||
60 | unsigned long flags; | |
61 | }; | |
62 | ||
63 | /* bits in struct cgroup_subsys_state flags field */ | |
64 | enum { | |
65 | CSS_ROOT, /* This CSS is the root of the subsystem */ | |
66 | }; | |
67 | ||
68 | /* | |
69 | * Call css_get() to hold a reference on the cgroup; | |
70 | * | |
71 | */ | |
72 | ||
73 | static inline void css_get(struct cgroup_subsys_state *css) | |
74 | { | |
75 | /* We don't need to reference count the root state */ | |
76 | if (!test_bit(CSS_ROOT, &css->flags)) | |
77 | atomic_inc(&css->refcnt); | |
78 | } | |
79 | /* | |
80 | * css_put() should be called to release a reference taken by | |
81 | * css_get() | |
82 | */ | |
83 | ||
81a6a5cd | 84 | extern void __css_put(struct cgroup_subsys_state *css); |
ddbcc7e8 PM |
85 | static inline void css_put(struct cgroup_subsys_state *css) |
86 | { | |
87 | if (!test_bit(CSS_ROOT, &css->flags)) | |
81a6a5cd | 88 | __css_put(css); |
ddbcc7e8 PM |
89 | } |
90 | ||
91 | struct cgroup { | |
92 | unsigned long flags; /* "unsigned long" so bitops work */ | |
93 | ||
94 | /* count users of this cgroup. >0 means busy, but doesn't | |
95 | * necessarily indicate the number of tasks in the | |
96 | * cgroup */ | |
97 | atomic_t count; | |
98 | ||
99 | /* | |
100 | * We link our 'sibling' struct into our parent's 'children'. | |
101 | * Our children link their 'sibling' into our 'children'. | |
102 | */ | |
103 | struct list_head sibling; /* my parent's children */ | |
104 | struct list_head children; /* my children */ | |
105 | ||
106 | struct cgroup *parent; /* my parent */ | |
107 | struct dentry *dentry; /* cgroup fs entry */ | |
108 | ||
109 | /* Private pointers for each registered subsystem */ | |
110 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
111 | ||
112 | struct cgroupfs_root *root; | |
113 | struct cgroup *top_cgroup; | |
817929ec PM |
114 | |
115 | /* | |
116 | * List of cg_cgroup_links pointing at css_sets with | |
117 | * tasks in this cgroup. Protected by css_set_lock | |
118 | */ | |
119 | struct list_head css_sets; | |
81a6a5cd PM |
120 | |
121 | /* | |
122 | * Linked list running through all cgroups that can | |
123 | * potentially be reaped by the release agent. Protected by | |
124 | * release_list_lock | |
125 | */ | |
126 | struct list_head release_list; | |
817929ec PM |
127 | }; |
128 | ||
129 | /* A css_set is a structure holding pointers to a set of | |
130 | * cgroup_subsys_state objects. This saves space in the task struct | |
131 | * object and speeds up fork()/exit(), since a single inc/dec and a | |
132 | * list_add()/del() can bump the reference count on the entire | |
133 | * cgroup set for a task. | |
134 | */ | |
135 | ||
136 | struct css_set { | |
137 | ||
138 | /* Reference count */ | |
139 | struct kref ref; | |
140 | ||
141 | /* | |
142 | * List running through all cgroup groups. Protected by | |
143 | * css_set_lock | |
144 | */ | |
145 | struct list_head list; | |
146 | ||
147 | /* | |
148 | * List running through all tasks using this cgroup | |
149 | * group. Protected by css_set_lock | |
150 | */ | |
151 | struct list_head tasks; | |
152 | ||
153 | /* | |
154 | * List of cg_cgroup_link objects on link chains from | |
155 | * cgroups referenced from this css_set. Protected by | |
156 | * css_set_lock | |
157 | */ | |
158 | struct list_head cg_links; | |
159 | ||
160 | /* | |
161 | * Set of subsystem states, one for each subsystem. This array | |
162 | * is immutable after creation apart from the init_css_set | |
163 | * during subsystem registration (at boot time). | |
164 | */ | |
165 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
166 | ||
ddbcc7e8 PM |
167 | }; |
168 | ||
169 | /* struct cftype: | |
170 | * | |
171 | * The files in the cgroup filesystem mostly have a very simple read/write | |
172 | * handling, some common function will take care of it. Nevertheless some cases | |
173 | * (read tasks) are special and therefore I define this structure for every | |
174 | * kind of file. | |
175 | * | |
176 | * | |
177 | * When reading/writing to a file: | |
a043e3b2 | 178 | * - the cgroup to use is file->f_dentry->d_parent->d_fsdata |
ddbcc7e8 PM |
179 | * - the 'cftype' of the file is file->f_dentry->d_fsdata |
180 | */ | |
181 | ||
182 | #define MAX_CFTYPE_NAME 64 | |
183 | struct cftype { | |
184 | /* By convention, the name should begin with the name of the | |
185 | * subsystem, followed by a period */ | |
186 | char name[MAX_CFTYPE_NAME]; | |
187 | int private; | |
188 | int (*open) (struct inode *inode, struct file *file); | |
ffd2d883 | 189 | ssize_t (*read) (struct cgroup *cgrp, struct cftype *cft, |
ddbcc7e8 PM |
190 | struct file *file, |
191 | char __user *buf, size_t nbytes, loff_t *ppos); | |
192 | /* | |
193 | * read_uint() is a shortcut for the common case of returning a | |
194 | * single integer. Use it in place of read() | |
195 | */ | |
ffd2d883 LZ |
196 | u64 (*read_uint) (struct cgroup *cgrp, struct cftype *cft); |
197 | ssize_t (*write) (struct cgroup *cgrp, struct cftype *cft, | |
ddbcc7e8 PM |
198 | struct file *file, |
199 | const char __user *buf, size_t nbytes, loff_t *ppos); | |
355e0c48 PM |
200 | |
201 | /* | |
202 | * write_uint() is a shortcut for the common case of accepting | |
203 | * a single integer (as parsed by simple_strtoull) from | |
204 | * userspace. Use in place of write(); return 0 or error. | |
205 | */ | |
ffd2d883 | 206 | int (*write_uint) (struct cgroup *cgrp, struct cftype *cft, u64 val); |
355e0c48 | 207 | |
ddbcc7e8 PM |
208 | int (*release) (struct inode *inode, struct file *file); |
209 | }; | |
210 | ||
31a7df01 CW |
211 | struct cgroup_scanner { |
212 | struct cgroup *cg; | |
213 | int (*test_task)(struct task_struct *p, struct cgroup_scanner *scan); | |
214 | void (*process_task)(struct task_struct *p, | |
215 | struct cgroup_scanner *scan); | |
216 | struct ptr_heap *heap; | |
217 | }; | |
218 | ||
ddbcc7e8 PM |
219 | /* Add a new file to the given cgroup directory. Should only be |
220 | * called by subsystems from within a populate() method */ | |
ffd2d883 | 221 | int cgroup_add_file(struct cgroup *cgrp, struct cgroup_subsys *subsys, |
ddbcc7e8 PM |
222 | const struct cftype *cft); |
223 | ||
224 | /* Add a set of new files to the given cgroup directory. Should | |
225 | * only be called by subsystems from within a populate() method */ | |
ffd2d883 | 226 | int cgroup_add_files(struct cgroup *cgrp, |
ddbcc7e8 PM |
227 | struct cgroup_subsys *subsys, |
228 | const struct cftype cft[], | |
229 | int count); | |
230 | ||
ffd2d883 | 231 | int cgroup_is_removed(const struct cgroup *cgrp); |
ddbcc7e8 | 232 | |
ffd2d883 | 233 | int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen); |
ddbcc7e8 | 234 | |
ffd2d883 | 235 | int cgroup_task_count(const struct cgroup *cgrp); |
bbcb81d0 | 236 | |
ddbcc7e8 | 237 | /* Return true if the cgroup is a descendant of the current cgroup */ |
ffd2d883 | 238 | int cgroup_is_descendant(const struct cgroup *cgrp); |
ddbcc7e8 PM |
239 | |
240 | /* Control Group subsystem type. See Documentation/cgroups.txt for details */ | |
241 | ||
242 | struct cgroup_subsys { | |
243 | struct cgroup_subsys_state *(*create)(struct cgroup_subsys *ss, | |
ffd2d883 LZ |
244 | struct cgroup *cgrp); |
245 | void (*pre_destroy)(struct cgroup_subsys *ss, struct cgroup *cgrp); | |
246 | void (*destroy)(struct cgroup_subsys *ss, struct cgroup *cgrp); | |
ddbcc7e8 | 247 | int (*can_attach)(struct cgroup_subsys *ss, |
ffd2d883 LZ |
248 | struct cgroup *cgrp, struct task_struct *tsk); |
249 | void (*attach)(struct cgroup_subsys *ss, struct cgroup *cgrp, | |
250 | struct cgroup *old_cgrp, struct task_struct *tsk); | |
ddbcc7e8 PM |
251 | void (*fork)(struct cgroup_subsys *ss, struct task_struct *task); |
252 | void (*exit)(struct cgroup_subsys *ss, struct task_struct *task); | |
253 | int (*populate)(struct cgroup_subsys *ss, | |
ffd2d883 LZ |
254 | struct cgroup *cgrp); |
255 | void (*post_clone)(struct cgroup_subsys *ss, struct cgroup *cgrp); | |
ddbcc7e8 PM |
256 | void (*bind)(struct cgroup_subsys *ss, struct cgroup *root); |
257 | int subsys_id; | |
258 | int active; | |
259 | int early_init; | |
260 | #define MAX_CGROUP_TYPE_NAMELEN 32 | |
261 | const char *name; | |
262 | ||
263 | /* Protected by RCU */ | |
264 | struct cgroupfs_root *root; | |
265 | ||
266 | struct list_head sibling; | |
267 | ||
268 | void *private; | |
269 | }; | |
270 | ||
271 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _subsys; | |
272 | #include <linux/cgroup_subsys.h> | |
273 | #undef SUBSYS | |
274 | ||
275 | static inline struct cgroup_subsys_state *cgroup_subsys_state( | |
ffd2d883 | 276 | struct cgroup *cgrp, int subsys_id) |
ddbcc7e8 | 277 | { |
ffd2d883 | 278 | return cgrp->subsys[subsys_id]; |
ddbcc7e8 PM |
279 | } |
280 | ||
281 | static inline struct cgroup_subsys_state *task_subsys_state( | |
282 | struct task_struct *task, int subsys_id) | |
283 | { | |
817929ec | 284 | return rcu_dereference(task->cgroups->subsys[subsys_id]); |
ddbcc7e8 PM |
285 | } |
286 | ||
287 | static inline struct cgroup* task_cgroup(struct task_struct *task, | |
288 | int subsys_id) | |
289 | { | |
290 | return task_subsys_state(task, subsys_id)->cgroup; | |
291 | } | |
292 | ||
697f4161 PM |
293 | int cgroup_clone(struct task_struct *tsk, struct cgroup_subsys *ss); |
294 | ||
817929ec PM |
295 | /* A cgroup_iter should be treated as an opaque object */ |
296 | struct cgroup_iter { | |
297 | struct list_head *cg_link; | |
298 | struct list_head *task; | |
299 | }; | |
300 | ||
301 | /* To iterate across the tasks in a cgroup: | |
302 | * | |
303 | * 1) call cgroup_iter_start to intialize an iterator | |
304 | * | |
305 | * 2) call cgroup_iter_next() to retrieve member tasks until it | |
306 | * returns NULL or until you want to end the iteration | |
307 | * | |
308 | * 3) call cgroup_iter_end() to destroy the iterator. | |
31a7df01 CW |
309 | * |
310 | * Or, call cgroup_scan_tasks() to iterate through every task in a cpuset. | |
311 | * - cgroup_scan_tasks() holds the css_set_lock when calling the test_task() | |
312 | * callback, but not while calling the process_task() callback. | |
817929ec | 313 | */ |
ffd2d883 LZ |
314 | void cgroup_iter_start(struct cgroup *cgrp, struct cgroup_iter *it); |
315 | struct task_struct *cgroup_iter_next(struct cgroup *cgrp, | |
817929ec | 316 | struct cgroup_iter *it); |
ffd2d883 | 317 | void cgroup_iter_end(struct cgroup *cgrp, struct cgroup_iter *it); |
31a7df01 | 318 | int cgroup_scan_tasks(struct cgroup_scanner *scan); |
956db3ca | 319 | int cgroup_attach_task(struct cgroup *, struct task_struct *); |
817929ec | 320 | |
ddbcc7e8 PM |
321 | #else /* !CONFIG_CGROUPS */ |
322 | ||
323 | static inline int cgroup_init_early(void) { return 0; } | |
324 | static inline int cgroup_init(void) { return 0; } | |
325 | static inline void cgroup_init_smp(void) {} | |
b4f48b63 PM |
326 | static inline void cgroup_fork(struct task_struct *p) {} |
327 | static inline void cgroup_fork_callbacks(struct task_struct *p) {} | |
817929ec | 328 | static inline void cgroup_post_fork(struct task_struct *p) {} |
b4f48b63 | 329 | static inline void cgroup_exit(struct task_struct *p, int callbacks) {} |
ddbcc7e8 PM |
330 | |
331 | static inline void cgroup_lock(void) {} | |
332 | static inline void cgroup_unlock(void) {} | |
846c7bb0 BS |
333 | static inline int cgroupstats_build(struct cgroupstats *stats, |
334 | struct dentry *dentry) | |
335 | { | |
336 | return -EINVAL; | |
337 | } | |
ddbcc7e8 PM |
338 | |
339 | #endif /* !CONFIG_CGROUPS */ | |
340 | ||
341 | #endif /* _LINUX_CGROUP_H */ |