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/cpumask.h>
13 #include <linux/nodemask.h>
14 #include <linux/rcupdate.h>
15 #include <linux/rculist.h>
16 #include <linux/cgroupstats.h>
17 #include <linux/prio_heap.h>
18 #include <linux/rwsem.h>
19 #include <linux/idr.h>
20 #include <linux/workqueue.h>
21 #include <linux/xattr.h>
33 extern int cgroup_init_early(void);
34 extern int cgroup_init(void);
35 extern void cgroup_fork(struct task_struct
*p
);
36 extern void cgroup_post_fork(struct task_struct
*p
);
37 extern void cgroup_exit(struct task_struct
*p
, int run_callbacks
);
38 extern int cgroupstats_build(struct cgroupstats
*stats
,
39 struct dentry
*dentry
);
40 extern int cgroup_load_subsys(struct cgroup_subsys
*ss
);
41 extern void cgroup_unload_subsys(struct cgroup_subsys
*ss
);
43 extern int proc_cgroup_show(struct seq_file
*, void *);
46 * Define the enumeration of all cgroup subsystems.
48 * We define ids for builtin subsystems and then modular ones.
50 #define SUBSYS(_x) _x ## _subsys_id,
51 enum cgroup_subsys_id
{
52 #define IS_SUBSYS_ENABLED(option) IS_BUILTIN(option)
53 #include <linux/cgroup_subsys.h>
54 #undef IS_SUBSYS_ENABLED
55 CGROUP_BUILTIN_SUBSYS_COUNT
,
57 __CGROUP_SUBSYS_TEMP_PLACEHOLDER
= CGROUP_BUILTIN_SUBSYS_COUNT
- 1,
59 #define IS_SUBSYS_ENABLED(option) IS_MODULE(option)
60 #include <linux/cgroup_subsys.h>
61 #undef IS_SUBSYS_ENABLED
66 /* Per-subsystem/per-cgroup state maintained by the system. */
67 struct cgroup_subsys_state
{
69 * The cgroup that this subsystem is attached to. Useful
70 * for subsystems that want to know about the cgroup
73 struct cgroup
*cgroup
;
76 * State maintained by the cgroup system to allow subsystems
77 * to be "busy". Should be accessed via css_get(),
78 * css_tryget() and css_put().
84 /* ID for this css, if possible */
85 struct css_id __rcu
*id
;
87 /* Used to put @cgroup->dentry on the last css_put() */
88 struct work_struct dput_work
;
91 /* bits in struct cgroup_subsys_state flags field */
93 CSS_ROOT
= (1 << 0), /* this CSS is the root of the subsystem */
94 CSS_ONLINE
= (1 << 1), /* between ->css_online() and ->css_offline() */
98 * css_get - obtain a reference on the specified css
101 * The caller must already have a reference.
103 static inline void css_get(struct cgroup_subsys_state
*css
)
105 /* We don't need to reference count the root state */
106 if (!(css
->flags
& CSS_ROOT
))
107 atomic_inc(&css
->refcnt
);
110 extern bool __css_tryget(struct cgroup_subsys_state
*css
);
113 * css_tryget - try to obtain a reference on the specified css
116 * Obtain a reference on @css if it's alive. The caller naturally needs to
117 * ensure that @css is accessible but doesn't have to be holding a
118 * reference on it - IOW, RCU protected access is good enough for this
119 * function. Returns %true if a reference count was successfully obtained;
122 static inline bool css_tryget(struct cgroup_subsys_state
*css
)
124 if (css
->flags
& CSS_ROOT
)
126 return __css_tryget(css
);
129 extern void __css_put(struct cgroup_subsys_state
*css
);
132 * css_put - put a css reference
135 * Put a reference obtained via css_get() and css_tryget().
137 static inline void css_put(struct cgroup_subsys_state
*css
)
139 if (!(css
->flags
& CSS_ROOT
))
143 /* bits in struct cgroup flags field */
145 /* Control Group is dead */
148 * Control Group has previously had a child cgroup or a task,
149 * but no longer (only if CGRP_NOTIFY_ON_RELEASE is set)
152 /* Control Group requires release notifications to userspace */
153 CGRP_NOTIFY_ON_RELEASE
,
155 * Clone the parent's configuration when creating a new child
156 * cpuset cgroup. For historical reasons, this option can be
157 * specified at mount time and thus is implemented here.
159 CGRP_CPUSET_CLONE_CHILDREN
,
160 /* see the comment above CGRP_ROOT_SANE_BEHAVIOR for details */
165 struct rcu_head rcu_head
;
170 unsigned long flags
; /* "unsigned long" so bitops work */
173 * count users of this cgroup. >0 means busy, but doesn't
174 * necessarily indicate the number of tasks in the cgroup
178 int id
; /* ida allocated in-hierarchy ID */
181 * We link our 'sibling' struct into our parent's 'children'.
182 * Our children link their 'sibling' into our 'children'.
184 struct list_head sibling
; /* my parent's children */
185 struct list_head children
; /* my children */
186 struct list_head files
; /* my files */
188 struct cgroup
*parent
; /* my parent */
189 struct dentry
*dentry
; /* cgroup fs entry, RCU protected */
192 * Monotonically increasing unique serial number which defines a
193 * uniform order among all cgroups. It's guaranteed that all
194 * ->children lists are in the ascending order of ->serial_nr.
195 * It's used to allow interrupting and resuming iterations.
200 * This is a copy of dentry->d_name, and it's needed because
201 * we can't use dentry->d_name in cgroup_path().
203 * You must acquire rcu_read_lock() to access cgrp->name, and
204 * the only place that can change it is rename(), which is
205 * protected by parent dir's i_mutex.
207 * Normally you should use cgroup_name() wrapper rather than
208 * access it directly.
210 struct cgroup_name __rcu
*name
;
212 /* Private pointers for each registered subsystem */
213 struct cgroup_subsys_state
*subsys
[CGROUP_SUBSYS_COUNT
];
215 struct cgroupfs_root
*root
;
218 * List of cgrp_cset_links pointing at css_sets with tasks in this
219 * cgroup. Protected by css_set_lock.
221 struct list_head cset_links
;
223 struct list_head allcg_node
; /* cgroupfs_root->allcg_list */
224 struct list_head cft_q_node
; /* used during cftype add/rm */
227 * Linked list running through all cgroups that can
228 * potentially be reaped by the release agent. Protected by
231 struct list_head release_list
;
234 * list of pidlists, up to two for each namespace (one for procs, one
235 * for tasks); created on demand.
237 struct list_head pidlists
;
238 struct mutex pidlist_mutex
;
240 /* For RCU-protected deletion */
241 struct rcu_head rcu_head
;
242 struct work_struct free_work
;
244 /* List of events which userspace want to receive */
245 struct list_head event_list
;
246 spinlock_t event_list_lock
;
248 /* directory xattrs */
249 struct simple_xattrs xattrs
;
252 #define MAX_CGROUP_ROOT_NAMELEN 64
254 /* cgroupfs_root->flags */
257 * Unfortunately, cgroup core and various controllers are riddled
258 * with idiosyncrasies and pointless options. The following flag,
259 * when set, will force sane behavior - some options are forced on,
260 * others are disallowed, and some controllers will change their
261 * hierarchical or other behaviors.
263 * The set of behaviors affected by this flag are still being
264 * determined and developed and the mount option for this flag is
265 * prefixed with __DEVEL__. The prefix will be dropped once we
266 * reach the point where all behaviors are compatible with the
267 * planned unified hierarchy, which will automatically turn on this
270 * The followings are the behaviors currently affected this flag.
272 * - Mount options "noprefix" and "clone_children" are disallowed.
273 * Also, cgroupfs file cgroup.clone_children is not created.
275 * - When mounting an existing superblock, mount options should
278 * - Remount is disallowed.
280 * - memcg: use_hierarchy is on by default and the cgroup file for
281 * the flag is not created.
283 * The followings are planned changes.
285 * - release_agent will be disallowed once replacement notification
286 * mechanism is implemented.
288 CGRP_ROOT_SANE_BEHAVIOR
= (1 << 0),
290 CGRP_ROOT_NOPREFIX
= (1 << 1), /* mounted subsystems have no named prefix */
291 CGRP_ROOT_XATTR
= (1 << 2), /* supports extended attributes */
295 * A cgroupfs_root represents the root of a cgroup hierarchy, and may be
296 * associated with a superblock to form an active hierarchy. This is
297 * internal to cgroup core. Don't access directly from controllers.
299 struct cgroupfs_root
{
300 struct super_block
*sb
;
303 * The bitmask of subsystems intended to be attached to this
306 unsigned long subsys_mask
;
308 /* Unique id for this hierarchy. */
311 /* The bitmask of subsystems currently attached to this hierarchy */
312 unsigned long actual_subsys_mask
;
314 /* A list running through the attached subsystems */
315 struct list_head subsys_list
;
317 /* The root cgroup for this hierarchy */
318 struct cgroup top_cgroup
;
320 /* Tracks how many cgroups are currently defined in hierarchy.*/
321 int number_of_cgroups
;
323 /* A list running through the active hierarchies */
324 struct list_head root_list
;
326 /* All cgroups on this root, cgroup_mutex protected */
327 struct list_head allcg_list
;
329 /* Hierarchy-specific flags */
332 /* IDs for cgroups in this hierarchy */
333 struct ida cgroup_ida
;
335 /* The path to use for release notifications. */
336 char release_agent_path
[PATH_MAX
];
338 /* The name for this hierarchy - may be empty */
339 char name
[MAX_CGROUP_ROOT_NAMELEN
];
343 * A css_set is a structure holding pointers to a set of
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
346 * list_add()/del() can bump the reference count on the entire cgroup
352 /* Reference count */
356 * List running through all cgroup groups in the same hash
357 * slot. Protected by css_set_lock
359 struct hlist_node hlist
;
362 * List running through all tasks using this cgroup
363 * group. Protected by css_set_lock
365 struct list_head tasks
;
368 * List of cgrp_cset_links pointing at cgroups referenced from this
369 * css_set. Protected by css_set_lock.
371 struct list_head cgrp_links
;
374 * Set of subsystem states, one for each subsystem. This array
375 * is immutable after creation apart from the init_css_set
376 * during subsystem registration (at boot time) and modular subsystem
379 struct cgroup_subsys_state
*subsys
[CGROUP_SUBSYS_COUNT
];
381 /* For RCU-protected deletion */
382 struct rcu_head rcu_head
;
386 * cgroup_map_cb is an abstract callback API for reporting map-valued
390 struct cgroup_map_cb
{
391 int (*fill
)(struct cgroup_map_cb
*cb
, const char *key
, u64 value
);
396 * struct cftype: handler definitions for cgroup control files
398 * When reading/writing to a file:
399 * - the cgroup to use is file->f_dentry->d_parent->d_fsdata
400 * - the 'cftype' of the file is file->f_dentry->d_fsdata
404 #define CFTYPE_ONLY_ON_ROOT (1U << 0) /* only create on root cg */
405 #define CFTYPE_NOT_ON_ROOT (1U << 1) /* don't create on root cg */
406 #define CFTYPE_INSANE (1U << 2) /* don't create if sane_behavior */
408 #define MAX_CFTYPE_NAME 64
412 * By convention, the name should begin with the name of the
413 * subsystem, followed by a period. Zero length string indicates
414 * end of cftype array.
416 char name
[MAX_CFTYPE_NAME
];
419 * If not 0, file mode is set to this value, otherwise it will
420 * be figured out automatically
425 * If non-zero, defines the maximum length of string that can
426 * be passed to write_string; defaults to 64
428 size_t max_write_len
;
433 int (*open
)(struct inode
*inode
, struct file
*file
);
434 ssize_t (*read
)(struct cgroup
*cgrp
, struct cftype
*cft
,
436 char __user
*buf
, size_t nbytes
, loff_t
*ppos
);
438 * read_u64() is a shortcut for the common case of returning a
439 * single integer. Use it in place of read()
441 u64 (*read_u64
)(struct cgroup
*cgrp
, struct cftype
*cft
);
443 * read_s64() is a signed version of read_u64()
445 s64 (*read_s64
)(struct cgroup
*cgrp
, struct cftype
*cft
);
447 * read_map() is used for defining a map of key/value
448 * pairs. It should call cb->fill(cb, key, value) for each
449 * entry. The key/value pairs (and their ordering) should not
450 * change between reboots.
452 int (*read_map
)(struct cgroup
*cont
, struct cftype
*cft
,
453 struct cgroup_map_cb
*cb
);
455 * read_seq_string() is used for outputting a simple sequence
458 int (*read_seq_string
)(struct cgroup
*cont
, struct cftype
*cft
,
461 ssize_t (*write
)(struct cgroup
*cgrp
, struct cftype
*cft
,
463 const char __user
*buf
, size_t nbytes
, loff_t
*ppos
);
466 * write_u64() is a shortcut for the common case of accepting
467 * a single integer (as parsed by simple_strtoull) from
468 * userspace. Use in place of write(); return 0 or error.
470 int (*write_u64
)(struct cgroup
*cgrp
, struct cftype
*cft
, u64 val
);
472 * write_s64() is a signed version of write_u64()
474 int (*write_s64
)(struct cgroup
*cgrp
, struct cftype
*cft
, s64 val
);
477 * write_string() is passed a nul-terminated kernelspace
478 * buffer of maximum length determined by max_write_len.
479 * Returns 0 or -ve error code.
481 int (*write_string
)(struct cgroup
*cgrp
, struct cftype
*cft
,
484 * trigger() callback can be used to get some kick from the
485 * userspace, when the actual string written is not important
486 * at all. The private field can be used to determine the
487 * kick type for multiplexing.
489 int (*trigger
)(struct cgroup
*cgrp
, unsigned int event
);
491 int (*release
)(struct inode
*inode
, struct file
*file
);
494 * register_event() callback will be used to add new userspace
495 * waiter for changes related to the cftype. Implement it if
496 * you want to provide this functionality. Use eventfd_signal()
497 * on eventfd to send notification to userspace.
499 int (*register_event
)(struct cgroup
*cgrp
, struct cftype
*cft
,
500 struct eventfd_ctx
*eventfd
, const char *args
);
502 * unregister_event() callback will be called when userspace
503 * closes the eventfd or on cgroup removing.
504 * This callback must be implemented, if you want provide
505 * notification functionality.
507 void (*unregister_event
)(struct cgroup
*cgrp
, struct cftype
*cft
,
508 struct eventfd_ctx
*eventfd
);
512 * cftype_sets describe cftypes belonging to a subsystem and are chained at
513 * cgroup_subsys->cftsets. Each cftset points to an array of cftypes
514 * terminated by zero length name.
517 struct list_head node
; /* chained at subsys->cftsets */
521 struct cgroup_scanner
{
523 int (*test_task
)(struct task_struct
*p
, struct cgroup_scanner
*scan
);
524 void (*process_task
)(struct task_struct
*p
,
525 struct cgroup_scanner
*scan
);
526 struct ptr_heap
*heap
;
531 * See the comment above CGRP_ROOT_SANE_BEHAVIOR for details. This
532 * function can be called as long as @cgrp is accessible.
534 static inline bool cgroup_sane_behavior(const struct cgroup
*cgrp
)
536 return cgrp
->root
->flags
& CGRP_ROOT_SANE_BEHAVIOR
;
539 /* Caller should hold rcu_read_lock() */
540 static inline const char *cgroup_name(const struct cgroup
*cgrp
)
542 return rcu_dereference(cgrp
->name
)->name
;
545 int cgroup_add_cftypes(struct cgroup_subsys
*ss
, struct cftype
*cfts
);
546 int cgroup_rm_cftypes(struct cgroup_subsys
*ss
, struct cftype
*cfts
);
548 bool cgroup_is_descendant(struct cgroup
*cgrp
, struct cgroup
*ancestor
);
550 int cgroup_path(const struct cgroup
*cgrp
, char *buf
, int buflen
);
551 int task_cgroup_path_from_hierarchy(struct task_struct
*task
, int hierarchy_id
,
552 char *buf
, size_t buflen
);
554 int cgroup_task_count(const struct cgroup
*cgrp
);
557 * Control Group taskset, used to pass around set of tasks to cgroup_subsys
560 struct cgroup_taskset
;
561 struct task_struct
*cgroup_taskset_first(struct cgroup_taskset
*tset
);
562 struct task_struct
*cgroup_taskset_next(struct cgroup_taskset
*tset
);
563 struct cgroup
*cgroup_taskset_cur_cgroup(struct cgroup_taskset
*tset
);
564 int cgroup_taskset_size(struct cgroup_taskset
*tset
);
567 * cgroup_taskset_for_each - iterate cgroup_taskset
568 * @task: the loop cursor
569 * @skip_cgrp: skip if task's cgroup matches this, %NULL to iterate through all
570 * @tset: taskset to iterate
572 #define cgroup_taskset_for_each(task, skip_cgrp, tset) \
573 for ((task) = cgroup_taskset_first((tset)); (task); \
574 (task) = cgroup_taskset_next((tset))) \
575 if (!(skip_cgrp) || \
576 cgroup_taskset_cur_cgroup((tset)) != (skip_cgrp))
579 * Control Group subsystem type.
580 * See Documentation/cgroups/cgroups.txt for details
583 struct cgroup_subsys
{
584 struct cgroup_subsys_state
*(*css_alloc
)(struct cgroup
*cgrp
);
585 int (*css_online
)(struct cgroup
*cgrp
);
586 void (*css_offline
)(struct cgroup
*cgrp
);
587 void (*css_free
)(struct cgroup
*cgrp
);
589 int (*can_attach
)(struct cgroup
*cgrp
, struct cgroup_taskset
*tset
);
590 void (*cancel_attach
)(struct cgroup
*cgrp
, struct cgroup_taskset
*tset
);
591 void (*attach
)(struct cgroup
*cgrp
, struct cgroup_taskset
*tset
);
592 void (*fork
)(struct task_struct
*task
);
593 void (*exit
)(struct cgroup
*cgrp
, struct cgroup
*old_cgrp
,
594 struct task_struct
*task
);
595 void (*bind
)(struct cgroup
*root
);
601 * True if this subsys uses ID. ID is not available before cgroup_init()
602 * (not available in early_init time.)
607 * If %false, this subsystem is properly hierarchical -
608 * configuration, resource accounting and restriction on a parent
609 * cgroup cover those of its children. If %true, hierarchy support
610 * is broken in some ways - some subsystems ignore hierarchy
611 * completely while others are only implemented half-way.
613 * It's now disallowed to create nested cgroups if the subsystem is
614 * broken and cgroup core will emit a warning message on such
615 * cases. Eventually, all subsystems will be made properly
616 * hierarchical and this will go away.
618 bool broken_hierarchy
;
619 bool warned_broken_hierarchy
;
621 #define MAX_CGROUP_TYPE_NAMELEN 32
625 * Link to parent, and list entry in parent's children.
626 * Protected by cgroup_lock()
628 struct cgroupfs_root
*root
;
629 struct list_head sibling
;
630 /* used when use_id == true */
634 /* list of cftype_sets */
635 struct list_head cftsets
;
637 /* base cftypes, automatically [de]registered with subsys itself */
638 struct cftype
*base_cftypes
;
639 struct cftype_set base_cftset
;
641 /* should be defined only by modular subsystems */
642 struct module
*module
;
645 #define SUBSYS(_x) extern struct cgroup_subsys _x ## _subsys;
646 #define IS_SUBSYS_ENABLED(option) IS_BUILTIN(option)
647 #include <linux/cgroup_subsys.h>
648 #undef IS_SUBSYS_ENABLED
651 static inline struct cgroup_subsys_state
*cgroup_subsys_state(
652 struct cgroup
*cgrp
, int subsys_id
)
654 return cgrp
->subsys
[subsys_id
];
658 * function to get the cgroup_subsys_state which allows for extra
659 * rcu_dereference_check() conditions, such as locks used during the
660 * cgroup_subsys::attach() methods.
662 #ifdef CONFIG_PROVE_RCU
663 extern struct mutex cgroup_mutex
;
664 #define task_subsys_state_check(task, subsys_id, __c) \
665 rcu_dereference_check((task)->cgroups->subsys[(subsys_id)], \
666 lockdep_is_held(&(task)->alloc_lock) || \
667 lockdep_is_held(&cgroup_mutex) || (__c))
669 #define task_subsys_state_check(task, subsys_id, __c) \
670 rcu_dereference((task)->cgroups->subsys[(subsys_id)])
673 static inline struct cgroup_subsys_state
*
674 task_subsys_state(struct task_struct
*task
, int subsys_id
)
676 return task_subsys_state_check(task
, subsys_id
, false);
679 static inline struct cgroup
* task_cgroup(struct task_struct
*task
,
682 return task_subsys_state(task
, subsys_id
)->cgroup
;
685 struct cgroup
*cgroup_next_sibling(struct cgroup
*pos
);
688 * cgroup_for_each_child - iterate through children of a cgroup
689 * @pos: the cgroup * to use as the loop cursor
690 * @cgrp: cgroup whose children to walk
692 * Walk @cgrp's children. Must be called under rcu_read_lock(). A child
693 * cgroup which hasn't finished ->css_online() or already has finished
694 * ->css_offline() may show up during traversal and it's each subsystem's
695 * responsibility to verify that each @pos is alive.
697 * If a subsystem synchronizes against the parent in its ->css_online() and
698 * before starting iterating, a cgroup which finished ->css_online() is
699 * guaranteed to be visible in the future iterations.
701 * It is allowed to temporarily drop RCU read lock during iteration. The
702 * caller is responsible for ensuring that @pos remains accessible until
703 * the start of the next iteration by, for example, bumping the css refcnt.
705 #define cgroup_for_each_child(pos, cgrp) \
706 for ((pos) = list_first_or_null_rcu(&(cgrp)->children, \
707 struct cgroup, sibling); \
708 (pos); (pos) = cgroup_next_sibling((pos)))
710 struct cgroup
*cgroup_next_descendant_pre(struct cgroup
*pos
,
711 struct cgroup
*cgroup
);
712 struct cgroup
*cgroup_rightmost_descendant(struct cgroup
*pos
);
715 * cgroup_for_each_descendant_pre - pre-order walk of a cgroup's descendants
716 * @pos: the cgroup * to use as the loop cursor
717 * @cgroup: cgroup whose descendants to walk
719 * Walk @cgroup's descendants. Must be called under rcu_read_lock(). A
720 * descendant cgroup which hasn't finished ->css_online() or already has
721 * finished ->css_offline() may show up during traversal and it's each
722 * subsystem's responsibility to verify that each @pos is alive.
724 * If a subsystem synchronizes against the parent in its ->css_online() and
725 * before starting iterating, and synchronizes against @pos on each
726 * iteration, any descendant cgroup which finished ->css_online() is
727 * guaranteed to be visible in the future iterations.
729 * In other words, the following guarantees that a descendant can't escape
730 * state updates of its ancestors.
734 * Lock @cgrp->parent and @cgrp;
735 * Inherit state from @cgrp->parent;
739 * my_update_state(@cgrp)
742 * Update @cgrp's state;
745 * cgroup_for_each_descendant_pre(@pos, @cgrp) {
747 * Verify @pos is alive and inherit state from @pos->parent;
752 * As long as the inheriting step, including checking the parent state, is
753 * enclosed inside @pos locking, double-locking the parent isn't necessary
754 * while inheriting. The state update to the parent is guaranteed to be
755 * visible by walking order and, as long as inheriting operations to the
756 * same @pos are atomic to each other, multiple updates racing each other
757 * still result in the correct state. It's guaranateed that at least one
758 * inheritance happens for any cgroup after the latest update to its
761 * If checking parent's state requires locking the parent, each inheriting
762 * iteration should lock and unlock both @pos->parent and @pos.
764 * Alternatively, a subsystem may choose to use a single global lock to
765 * synchronize ->css_online() and ->css_offline() against tree-walking
768 * It is allowed to temporarily drop RCU read lock during iteration. The
769 * caller is responsible for ensuring that @pos remains accessible until
770 * the start of the next iteration by, for example, bumping the css refcnt.
772 #define cgroup_for_each_descendant_pre(pos, cgroup) \
773 for (pos = cgroup_next_descendant_pre(NULL, (cgroup)); (pos); \
774 pos = cgroup_next_descendant_pre((pos), (cgroup)))
776 struct cgroup
*cgroup_next_descendant_post(struct cgroup
*pos
,
777 struct cgroup
*cgroup
);
780 * cgroup_for_each_descendant_post - post-order walk of a cgroup's descendants
781 * @pos: the cgroup * to use as the loop cursor
782 * @cgroup: cgroup whose descendants to walk
784 * Similar to cgroup_for_each_descendant_pre() but performs post-order
785 * traversal instead. Note that the walk visibility guarantee described in
786 * pre-order walk doesn't apply the same to post-order walks.
788 #define cgroup_for_each_descendant_post(pos, cgroup) \
789 for (pos = cgroup_next_descendant_post(NULL, (cgroup)); (pos); \
790 pos = cgroup_next_descendant_post((pos), (cgroup)))
792 /* A cgroup_iter should be treated as an opaque object */
794 struct list_head
*cset_link
;
795 struct list_head
*task
;
799 * To iterate across the tasks in a cgroup:
801 * 1) call cgroup_iter_start to initialize an iterator
803 * 2) call cgroup_iter_next() to retrieve member tasks until it
804 * returns NULL or until you want to end the iteration
806 * 3) call cgroup_iter_end() to destroy the iterator.
808 * Or, call cgroup_scan_tasks() to iterate through every task in a
809 * cgroup - cgroup_scan_tasks() holds the css_set_lock when calling
810 * the test_task() callback, but not while calling the process_task()
813 void cgroup_iter_start(struct cgroup
*cgrp
, struct cgroup_iter
*it
);
814 struct task_struct
*cgroup_iter_next(struct cgroup
*cgrp
,
815 struct cgroup_iter
*it
);
816 void cgroup_iter_end(struct cgroup
*cgrp
, struct cgroup_iter
*it
);
817 int cgroup_scan_tasks(struct cgroup_scanner
*scan
);
818 int cgroup_attach_task_all(struct task_struct
*from
, struct task_struct
*);
819 int cgroup_transfer_tasks(struct cgroup
*to
, struct cgroup
*from
);
822 * CSS ID is ID for cgroup_subsys_state structs under subsys. This only works
823 * if cgroup_subsys.use_id == true. It can be used for looking up and scanning.
824 * CSS ID is assigned at cgroup allocation (create) automatically
825 * and removed when subsys calls free_css_id() function. This is because
826 * the lifetime of cgroup_subsys_state is subsys's matter.
828 * Looking up and scanning function should be called under rcu_read_lock().
829 * Taking cgroup_mutex is not necessary for following calls.
830 * But the css returned by this routine can be "not populated yet" or "being
831 * destroyed". The caller should check css and cgroup's status.
835 * Typically Called at ->destroy(), or somewhere the subsys frees
836 * cgroup_subsys_state.
838 void free_css_id(struct cgroup_subsys
*ss
, struct cgroup_subsys_state
*css
);
840 /* Find a cgroup_subsys_state which has given ID */
842 struct cgroup_subsys_state
*css_lookup(struct cgroup_subsys
*ss
, int id
);
844 /* Returns true if root is ancestor of cg */
845 bool css_is_ancestor(struct cgroup_subsys_state
*cg
,
846 const struct cgroup_subsys_state
*root
);
848 /* Get id and depth of css */
849 unsigned short css_id(struct cgroup_subsys_state
*css
);
850 struct cgroup_subsys_state
*cgroup_css_from_dir(struct file
*f
, int id
);
852 #else /* !CONFIG_CGROUPS */
854 static inline int cgroup_init_early(void) { return 0; }
855 static inline int cgroup_init(void) { return 0; }
856 static inline void cgroup_fork(struct task_struct
*p
) {}
857 static inline void cgroup_post_fork(struct task_struct
*p
) {}
858 static inline void cgroup_exit(struct task_struct
*p
, int callbacks
) {}
860 static inline int cgroupstats_build(struct cgroupstats
*stats
,
861 struct dentry
*dentry
)
866 /* No cgroups - nothing to do */
867 static inline int cgroup_attach_task_all(struct task_struct
*from
,
868 struct task_struct
*t
)
873 #endif /* !CONFIG_CGROUPS */
875 #endif /* _LINUX_CGROUP_H */