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 */
172 int id
; /* ida allocated in-hierarchy ID */
175 * We link our 'sibling' struct into our parent's 'children'.
176 * Our children link their 'sibling' into our 'children'.
178 struct list_head sibling
; /* my parent's children */
179 struct list_head children
; /* my children */
180 struct list_head files
; /* my files */
182 struct cgroup
*parent
; /* my parent */
183 struct dentry
*dentry
; /* cgroup fs entry, RCU protected */
186 * Monotonically increasing unique serial number which defines a
187 * uniform order among all cgroups. It's guaranteed that all
188 * ->children lists are in the ascending order of ->serial_nr.
189 * It's used to allow interrupting and resuming iterations.
194 * This is a copy of dentry->d_name, and it's needed because
195 * we can't use dentry->d_name in cgroup_path().
197 * You must acquire rcu_read_lock() to access cgrp->name, and
198 * the only place that can change it is rename(), which is
199 * protected by parent dir's i_mutex.
201 * Normally you should use cgroup_name() wrapper rather than
202 * access it directly.
204 struct cgroup_name __rcu
*name
;
206 /* Private pointers for each registered subsystem */
207 struct cgroup_subsys_state
*subsys
[CGROUP_SUBSYS_COUNT
];
209 struct cgroupfs_root
*root
;
212 * List of cgrp_cset_links pointing at css_sets with tasks in this
213 * cgroup. Protected by css_set_lock.
215 struct list_head cset_links
;
217 struct list_head allcg_node
; /* cgroupfs_root->allcg_list */
218 struct list_head cft_q_node
; /* used during cftype add/rm */
221 * Linked list running through all cgroups that can
222 * potentially be reaped by the release agent. Protected by
225 struct list_head release_list
;
228 * list of pidlists, up to two for each namespace (one for procs, one
229 * for tasks); created on demand.
231 struct list_head pidlists
;
232 struct mutex pidlist_mutex
;
234 /* For RCU-protected deletion */
235 struct rcu_head rcu_head
;
236 struct work_struct destroy_work
;
238 /* List of events which userspace want to receive */
239 struct list_head event_list
;
240 spinlock_t event_list_lock
;
242 /* directory xattrs */
243 struct simple_xattrs xattrs
;
246 #define MAX_CGROUP_ROOT_NAMELEN 64
248 /* cgroupfs_root->flags */
251 * Unfortunately, cgroup core and various controllers are riddled
252 * with idiosyncrasies and pointless options. The following flag,
253 * when set, will force sane behavior - some options are forced on,
254 * others are disallowed, and some controllers will change their
255 * hierarchical or other behaviors.
257 * The set of behaviors affected by this flag are still being
258 * determined and developed and the mount option for this flag is
259 * prefixed with __DEVEL__. The prefix will be dropped once we
260 * reach the point where all behaviors are compatible with the
261 * planned unified hierarchy, which will automatically turn on this
264 * The followings are the behaviors currently affected this flag.
266 * - Mount options "noprefix" and "clone_children" are disallowed.
267 * Also, cgroupfs file cgroup.clone_children is not created.
269 * - When mounting an existing superblock, mount options should
272 * - Remount is disallowed.
274 * - memcg: use_hierarchy is on by default and the cgroup file for
275 * the flag is not created.
277 * The followings are planned changes.
279 * - release_agent will be disallowed once replacement notification
280 * mechanism is implemented.
282 CGRP_ROOT_SANE_BEHAVIOR
= (1 << 0),
284 CGRP_ROOT_NOPREFIX
= (1 << 1), /* mounted subsystems have no named prefix */
285 CGRP_ROOT_XATTR
= (1 << 2), /* supports extended attributes */
289 * A cgroupfs_root represents the root of a cgroup hierarchy, and may be
290 * associated with a superblock to form an active hierarchy. This is
291 * internal to cgroup core. Don't access directly from controllers.
293 struct cgroupfs_root
{
294 struct super_block
*sb
;
297 * The bitmask of subsystems intended to be attached to this
300 unsigned long subsys_mask
;
302 /* Unique id for this hierarchy. */
305 /* The bitmask of subsystems currently attached to this hierarchy */
306 unsigned long actual_subsys_mask
;
308 /* A list running through the attached subsystems */
309 struct list_head subsys_list
;
311 /* The root cgroup for this hierarchy */
312 struct cgroup top_cgroup
;
314 /* Tracks how many cgroups are currently defined in hierarchy.*/
315 int number_of_cgroups
;
317 /* A list running through the active hierarchies */
318 struct list_head root_list
;
320 /* All cgroups on this root, cgroup_mutex protected */
321 struct list_head allcg_list
;
323 /* Hierarchy-specific flags */
326 /* IDs for cgroups in this hierarchy */
327 struct ida cgroup_ida
;
329 /* The path to use for release notifications. */
330 char release_agent_path
[PATH_MAX
];
332 /* The name for this hierarchy - may be empty */
333 char name
[MAX_CGROUP_ROOT_NAMELEN
];
337 * A css_set is a structure holding pointers to a set of
338 * cgroup_subsys_state objects. This saves space in the task struct
339 * object and speeds up fork()/exit(), since a single inc/dec and a
340 * list_add()/del() can bump the reference count on the entire cgroup
346 /* Reference count */
350 * List running through all cgroup groups in the same hash
351 * slot. Protected by css_set_lock
353 struct hlist_node hlist
;
356 * List running through all tasks using this cgroup
357 * group. Protected by css_set_lock
359 struct list_head tasks
;
362 * List of cgrp_cset_links pointing at cgroups referenced from this
363 * css_set. Protected by css_set_lock.
365 struct list_head cgrp_links
;
368 * Set of subsystem states, one for each subsystem. This array
369 * is immutable after creation apart from the init_css_set
370 * during subsystem registration (at boot time) and modular subsystem
373 struct cgroup_subsys_state
*subsys
[CGROUP_SUBSYS_COUNT
];
375 /* For RCU-protected deletion */
376 struct rcu_head rcu_head
;
380 * cgroup_map_cb is an abstract callback API for reporting map-valued
384 struct cgroup_map_cb
{
385 int (*fill
)(struct cgroup_map_cb
*cb
, const char *key
, u64 value
);
390 * struct cftype: handler definitions for cgroup control files
392 * When reading/writing to a file:
393 * - the cgroup to use is file->f_dentry->d_parent->d_fsdata
394 * - the 'cftype' of the file is file->f_dentry->d_fsdata
398 #define CFTYPE_ONLY_ON_ROOT (1U << 0) /* only create on root cg */
399 #define CFTYPE_NOT_ON_ROOT (1U << 1) /* don't create on root cg */
400 #define CFTYPE_INSANE (1U << 2) /* don't create if sane_behavior */
402 #define MAX_CFTYPE_NAME 64
406 * By convention, the name should begin with the name of the
407 * subsystem, followed by a period. Zero length string indicates
408 * end of cftype array.
410 char name
[MAX_CFTYPE_NAME
];
413 * If not 0, file mode is set to this value, otherwise it will
414 * be figured out automatically
419 * If non-zero, defines the maximum length of string that can
420 * be passed to write_string; defaults to 64
422 size_t max_write_len
;
427 int (*open
)(struct inode
*inode
, struct file
*file
);
428 ssize_t (*read
)(struct cgroup
*cgrp
, struct cftype
*cft
,
430 char __user
*buf
, size_t nbytes
, loff_t
*ppos
);
432 * read_u64() is a shortcut for the common case of returning a
433 * single integer. Use it in place of read()
435 u64 (*read_u64
)(struct cgroup
*cgrp
, struct cftype
*cft
);
437 * read_s64() is a signed version of read_u64()
439 s64 (*read_s64
)(struct cgroup
*cgrp
, struct cftype
*cft
);
441 * read_map() is used for defining a map of key/value
442 * pairs. It should call cb->fill(cb, key, value) for each
443 * entry. The key/value pairs (and their ordering) should not
444 * change between reboots.
446 int (*read_map
)(struct cgroup
*cont
, struct cftype
*cft
,
447 struct cgroup_map_cb
*cb
);
449 * read_seq_string() is used for outputting a simple sequence
452 int (*read_seq_string
)(struct cgroup
*cont
, struct cftype
*cft
,
455 ssize_t (*write
)(struct cgroup
*cgrp
, struct cftype
*cft
,
457 const char __user
*buf
, size_t nbytes
, loff_t
*ppos
);
460 * write_u64() is a shortcut for the common case of accepting
461 * a single integer (as parsed by simple_strtoull) from
462 * userspace. Use in place of write(); return 0 or error.
464 int (*write_u64
)(struct cgroup
*cgrp
, struct cftype
*cft
, u64 val
);
466 * write_s64() is a signed version of write_u64()
468 int (*write_s64
)(struct cgroup
*cgrp
, struct cftype
*cft
, s64 val
);
471 * write_string() is passed a nul-terminated kernelspace
472 * buffer of maximum length determined by max_write_len.
473 * Returns 0 or -ve error code.
475 int (*write_string
)(struct cgroup
*cgrp
, struct cftype
*cft
,
478 * trigger() callback can be used to get some kick from the
479 * userspace, when the actual string written is not important
480 * at all. The private field can be used to determine the
481 * kick type for multiplexing.
483 int (*trigger
)(struct cgroup
*cgrp
, unsigned int event
);
485 int (*release
)(struct inode
*inode
, struct file
*file
);
488 * register_event() callback will be used to add new userspace
489 * waiter for changes related to the cftype. Implement it if
490 * you want to provide this functionality. Use eventfd_signal()
491 * on eventfd to send notification to userspace.
493 int (*register_event
)(struct cgroup
*cgrp
, struct cftype
*cft
,
494 struct eventfd_ctx
*eventfd
, const char *args
);
496 * unregister_event() callback will be called when userspace
497 * closes the eventfd or on cgroup removing.
498 * This callback must be implemented, if you want provide
499 * notification functionality.
501 void (*unregister_event
)(struct cgroup
*cgrp
, struct cftype
*cft
,
502 struct eventfd_ctx
*eventfd
);
506 * cftype_sets describe cftypes belonging to a subsystem and are chained at
507 * cgroup_subsys->cftsets. Each cftset points to an array of cftypes
508 * terminated by zero length name.
511 struct list_head node
; /* chained at subsys->cftsets */
515 struct cgroup_scanner
{
517 int (*test_task
)(struct task_struct
*p
, struct cgroup_scanner
*scan
);
518 void (*process_task
)(struct task_struct
*p
,
519 struct cgroup_scanner
*scan
);
520 struct ptr_heap
*heap
;
525 * See the comment above CGRP_ROOT_SANE_BEHAVIOR for details. This
526 * function can be called as long as @cgrp is accessible.
528 static inline bool cgroup_sane_behavior(const struct cgroup
*cgrp
)
530 return cgrp
->root
->flags
& CGRP_ROOT_SANE_BEHAVIOR
;
533 /* Caller should hold rcu_read_lock() */
534 static inline const char *cgroup_name(const struct cgroup
*cgrp
)
536 return rcu_dereference(cgrp
->name
)->name
;
539 int cgroup_add_cftypes(struct cgroup_subsys
*ss
, struct cftype
*cfts
);
540 int cgroup_rm_cftypes(struct cgroup_subsys
*ss
, struct cftype
*cfts
);
542 bool cgroup_is_descendant(struct cgroup
*cgrp
, struct cgroup
*ancestor
);
544 int cgroup_path(const struct cgroup
*cgrp
, char *buf
, int buflen
);
545 int task_cgroup_path_from_hierarchy(struct task_struct
*task
, int hierarchy_id
,
546 char *buf
, size_t buflen
);
548 int cgroup_task_count(const struct cgroup
*cgrp
);
551 * Control Group taskset, used to pass around set of tasks to cgroup_subsys
554 struct cgroup_taskset
;
555 struct task_struct
*cgroup_taskset_first(struct cgroup_taskset
*tset
);
556 struct task_struct
*cgroup_taskset_next(struct cgroup_taskset
*tset
);
557 struct cgroup
*cgroup_taskset_cur_cgroup(struct cgroup_taskset
*tset
);
558 int cgroup_taskset_size(struct cgroup_taskset
*tset
);
561 * cgroup_taskset_for_each - iterate cgroup_taskset
562 * @task: the loop cursor
563 * @skip_cgrp: skip if task's cgroup matches this, %NULL to iterate through all
564 * @tset: taskset to iterate
566 #define cgroup_taskset_for_each(task, skip_cgrp, tset) \
567 for ((task) = cgroup_taskset_first((tset)); (task); \
568 (task) = cgroup_taskset_next((tset))) \
569 if (!(skip_cgrp) || \
570 cgroup_taskset_cur_cgroup((tset)) != (skip_cgrp))
573 * Control Group subsystem type.
574 * See Documentation/cgroups/cgroups.txt for details
577 struct cgroup_subsys
{
578 struct cgroup_subsys_state
*(*css_alloc
)(struct cgroup
*cgrp
);
579 int (*css_online
)(struct cgroup
*cgrp
);
580 void (*css_offline
)(struct cgroup
*cgrp
);
581 void (*css_free
)(struct cgroup
*cgrp
);
583 int (*can_attach
)(struct cgroup
*cgrp
, struct cgroup_taskset
*tset
);
584 void (*cancel_attach
)(struct cgroup
*cgrp
, struct cgroup_taskset
*tset
);
585 void (*attach
)(struct cgroup
*cgrp
, struct cgroup_taskset
*tset
);
586 void (*fork
)(struct task_struct
*task
);
587 void (*exit
)(struct cgroup
*cgrp
, struct cgroup
*old_cgrp
,
588 struct task_struct
*task
);
589 void (*bind
)(struct cgroup
*root
);
595 * True if this subsys uses ID. ID is not available before cgroup_init()
596 * (not available in early_init time.)
601 * If %false, this subsystem is properly hierarchical -
602 * configuration, resource accounting and restriction on a parent
603 * cgroup cover those of its children. If %true, hierarchy support
604 * is broken in some ways - some subsystems ignore hierarchy
605 * completely while others are only implemented half-way.
607 * It's now disallowed to create nested cgroups if the subsystem is
608 * broken and cgroup core will emit a warning message on such
609 * cases. Eventually, all subsystems will be made properly
610 * hierarchical and this will go away.
612 bool broken_hierarchy
;
613 bool warned_broken_hierarchy
;
615 #define MAX_CGROUP_TYPE_NAMELEN 32
619 * Link to parent, and list entry in parent's children.
620 * Protected by cgroup_lock()
622 struct cgroupfs_root
*root
;
623 struct list_head sibling
;
624 /* used when use_id == true */
628 /* list of cftype_sets */
629 struct list_head cftsets
;
631 /* base cftypes, automatically [de]registered with subsys itself */
632 struct cftype
*base_cftypes
;
633 struct cftype_set base_cftset
;
635 /* should be defined only by modular subsystems */
636 struct module
*module
;
639 #define SUBSYS(_x) extern struct cgroup_subsys _x ## _subsys;
640 #define IS_SUBSYS_ENABLED(option) IS_BUILTIN(option)
641 #include <linux/cgroup_subsys.h>
642 #undef IS_SUBSYS_ENABLED
645 static inline struct cgroup_subsys_state
*cgroup_subsys_state(
646 struct cgroup
*cgrp
, int subsys_id
)
648 return cgrp
->subsys
[subsys_id
];
652 * function to get the cgroup_subsys_state which allows for extra
653 * rcu_dereference_check() conditions, such as locks used during the
654 * cgroup_subsys::attach() methods.
656 #ifdef CONFIG_PROVE_RCU
657 extern struct mutex cgroup_mutex
;
658 #define task_subsys_state_check(task, subsys_id, __c) \
659 rcu_dereference_check((task)->cgroups->subsys[(subsys_id)], \
660 lockdep_is_held(&(task)->alloc_lock) || \
661 lockdep_is_held(&cgroup_mutex) || (__c))
663 #define task_subsys_state_check(task, subsys_id, __c) \
664 rcu_dereference((task)->cgroups->subsys[(subsys_id)])
667 static inline struct cgroup_subsys_state
*
668 task_subsys_state(struct task_struct
*task
, int subsys_id
)
670 return task_subsys_state_check(task
, subsys_id
, false);
673 static inline struct cgroup
* task_cgroup(struct task_struct
*task
,
676 return task_subsys_state(task
, subsys_id
)->cgroup
;
679 struct cgroup
*cgroup_next_sibling(struct cgroup
*pos
);
682 * cgroup_for_each_child - iterate through children of a cgroup
683 * @pos: the cgroup * to use as the loop cursor
684 * @cgrp: cgroup whose children to walk
686 * Walk @cgrp's children. Must be called under rcu_read_lock(). A child
687 * cgroup which hasn't finished ->css_online() or already has finished
688 * ->css_offline() may show up during traversal and it's each subsystem's
689 * responsibility to verify that each @pos is alive.
691 * If a subsystem synchronizes against the parent in its ->css_online() and
692 * before starting iterating, a cgroup which finished ->css_online() is
693 * guaranteed to be visible in the future iterations.
695 * It is allowed to temporarily drop RCU read lock during iteration. The
696 * caller is responsible for ensuring that @pos remains accessible until
697 * the start of the next iteration by, for example, bumping the css refcnt.
699 #define cgroup_for_each_child(pos, cgrp) \
700 for ((pos) = list_first_or_null_rcu(&(cgrp)->children, \
701 struct cgroup, sibling); \
702 (pos); (pos) = cgroup_next_sibling((pos)))
704 struct cgroup
*cgroup_next_descendant_pre(struct cgroup
*pos
,
705 struct cgroup
*cgroup
);
706 struct cgroup
*cgroup_rightmost_descendant(struct cgroup
*pos
);
709 * cgroup_for_each_descendant_pre - pre-order walk of a cgroup's descendants
710 * @pos: the cgroup * to use as the loop cursor
711 * @cgroup: cgroup whose descendants to walk
713 * Walk @cgroup's descendants. Must be called under rcu_read_lock(). A
714 * descendant cgroup which hasn't finished ->css_online() or already has
715 * finished ->css_offline() may show up during traversal and it's each
716 * subsystem's responsibility to verify that each @pos is alive.
718 * If a subsystem synchronizes against the parent in its ->css_online() and
719 * before starting iterating, and synchronizes against @pos on each
720 * iteration, any descendant cgroup which finished ->css_online() is
721 * guaranteed to be visible in the future iterations.
723 * In other words, the following guarantees that a descendant can't escape
724 * state updates of its ancestors.
728 * Lock @cgrp->parent and @cgrp;
729 * Inherit state from @cgrp->parent;
733 * my_update_state(@cgrp)
736 * Update @cgrp's state;
739 * cgroup_for_each_descendant_pre(@pos, @cgrp) {
741 * Verify @pos is alive and inherit state from @pos->parent;
746 * As long as the inheriting step, including checking the parent state, is
747 * enclosed inside @pos locking, double-locking the parent isn't necessary
748 * while inheriting. The state update to the parent is guaranteed to be
749 * visible by walking order and, as long as inheriting operations to the
750 * same @pos are atomic to each other, multiple updates racing each other
751 * still result in the correct state. It's guaranateed that at least one
752 * inheritance happens for any cgroup after the latest update to its
755 * If checking parent's state requires locking the parent, each inheriting
756 * iteration should lock and unlock both @pos->parent and @pos.
758 * Alternatively, a subsystem may choose to use a single global lock to
759 * synchronize ->css_online() and ->css_offline() against tree-walking
762 * It is allowed to temporarily drop RCU read lock during iteration. The
763 * caller is responsible for ensuring that @pos remains accessible until
764 * the start of the next iteration by, for example, bumping the css refcnt.
766 #define cgroup_for_each_descendant_pre(pos, cgroup) \
767 for (pos = cgroup_next_descendant_pre(NULL, (cgroup)); (pos); \
768 pos = cgroup_next_descendant_pre((pos), (cgroup)))
770 struct cgroup
*cgroup_next_descendant_post(struct cgroup
*pos
,
771 struct cgroup
*cgroup
);
774 * cgroup_for_each_descendant_post - post-order walk of a cgroup's descendants
775 * @pos: the cgroup * to use as the loop cursor
776 * @cgroup: cgroup whose descendants to walk
778 * Similar to cgroup_for_each_descendant_pre() but performs post-order
779 * traversal instead. Note that the walk visibility guarantee described in
780 * pre-order walk doesn't apply the same to post-order walks.
782 #define cgroup_for_each_descendant_post(pos, cgroup) \
783 for (pos = cgroup_next_descendant_post(NULL, (cgroup)); (pos); \
784 pos = cgroup_next_descendant_post((pos), (cgroup)))
786 /* A cgroup_iter should be treated as an opaque object */
788 struct list_head
*cset_link
;
789 struct list_head
*task
;
793 * To iterate across the tasks in a cgroup:
795 * 1) call cgroup_iter_start to initialize an iterator
797 * 2) call cgroup_iter_next() to retrieve member tasks until it
798 * returns NULL or until you want to end the iteration
800 * 3) call cgroup_iter_end() to destroy the iterator.
802 * Or, call cgroup_scan_tasks() to iterate through every task in a
803 * cgroup - cgroup_scan_tasks() holds the css_set_lock when calling
804 * the test_task() callback, but not while calling the process_task()
807 void cgroup_iter_start(struct cgroup
*cgrp
, struct cgroup_iter
*it
);
808 struct task_struct
*cgroup_iter_next(struct cgroup
*cgrp
,
809 struct cgroup_iter
*it
);
810 void cgroup_iter_end(struct cgroup
*cgrp
, struct cgroup_iter
*it
);
811 int cgroup_scan_tasks(struct cgroup_scanner
*scan
);
812 int cgroup_attach_task_all(struct task_struct
*from
, struct task_struct
*);
813 int cgroup_transfer_tasks(struct cgroup
*to
, struct cgroup
*from
);
816 * CSS ID is ID for cgroup_subsys_state structs under subsys. This only works
817 * if cgroup_subsys.use_id == true. It can be used for looking up and scanning.
818 * CSS ID is assigned at cgroup allocation (create) automatically
819 * and removed when subsys calls free_css_id() function. This is because
820 * the lifetime of cgroup_subsys_state is subsys's matter.
822 * Looking up and scanning function should be called under rcu_read_lock().
823 * Taking cgroup_mutex is not necessary for following calls.
824 * But the css returned by this routine can be "not populated yet" or "being
825 * destroyed". The caller should check css and cgroup's status.
829 * Typically Called at ->destroy(), or somewhere the subsys frees
830 * cgroup_subsys_state.
832 void free_css_id(struct cgroup_subsys
*ss
, struct cgroup_subsys_state
*css
);
834 /* Find a cgroup_subsys_state which has given ID */
836 struct cgroup_subsys_state
*css_lookup(struct cgroup_subsys
*ss
, int id
);
838 /* Returns true if root is ancestor of cg */
839 bool css_is_ancestor(struct cgroup_subsys_state
*cg
,
840 const struct cgroup_subsys_state
*root
);
842 /* Get id and depth of css */
843 unsigned short css_id(struct cgroup_subsys_state
*css
);
844 struct cgroup_subsys_state
*cgroup_css_from_dir(struct file
*f
, int id
);
846 #else /* !CONFIG_CGROUPS */
848 static inline int cgroup_init_early(void) { return 0; }
849 static inline int cgroup_init(void) { return 0; }
850 static inline void cgroup_fork(struct task_struct
*p
) {}
851 static inline void cgroup_post_fork(struct task_struct
*p
) {}
852 static inline void cgroup_exit(struct task_struct
*p
, int callbacks
) {}
854 static inline int cgroupstats_build(struct cgroupstats
*stats
,
855 struct dentry
*dentry
)
860 /* No cgroups - nothing to do */
861 static inline int cgroup_attach_task_all(struct task_struct
*from
,
862 struct task_struct
*t
)
867 #endif /* !CONFIG_CGROUPS */
869 #endif /* _LINUX_CGROUP_H */