2 * linux/cgroup-defs.h - basic definitions for cgroup
4 * This file provides basic type and interface. Include this file directly
5 * only if necessary to avoid cyclic dependencies.
7 #ifndef _LINUX_CGROUP_DEFS_H
8 #define _LINUX_CGROUP_DEFS_H
10 #include <linux/limits.h>
11 #include <linux/list.h>
12 #include <linux/idr.h>
13 #include <linux/wait.h>
14 #include <linux/mutex.h>
15 #include <linux/rcupdate.h>
16 #include <linux/percpu-refcount.h>
17 #include <linux/workqueue.h>
24 struct cgroup_taskset
;
27 struct kernfs_open_file
;
29 #define MAX_CGROUP_TYPE_NAMELEN 32
30 #define MAX_CGROUP_ROOT_NAMELEN 64
31 #define MAX_CFTYPE_NAME 64
33 /* define the enumeration of all cgroup subsystems */
34 #define SUBSYS(_x) _x ## _cgrp_id,
35 enum cgroup_subsys_id
{
36 #include <linux/cgroup_subsys.h>
41 /* bits in struct cgroup_subsys_state flags field */
43 CSS_NO_REF
= (1 << 0), /* no reference counting for this css */
44 CSS_ONLINE
= (1 << 1), /* between ->css_online() and ->css_offline() */
45 CSS_RELEASED
= (1 << 2), /* refcnt reached zero, released */
48 /* bits in struct cgroup flags field */
50 /* Control Group requires release notifications to userspace */
51 CGRP_NOTIFY_ON_RELEASE
,
53 * Clone the parent's configuration when creating a new child
54 * cpuset cgroup. For historical reasons, this option can be
55 * specified at mount time and thus is implemented here.
57 CGRP_CPUSET_CLONE_CHILDREN
,
60 /* cgroup_root->flags */
62 CGRP_ROOT_SANE_BEHAVIOR
= (1 << 0), /* __DEVEL__sane_behavior specified */
63 CGRP_ROOT_NOPREFIX
= (1 << 1), /* mounted subsystems have no named prefix */
64 CGRP_ROOT_XATTR
= (1 << 2), /* supports extended attributes */
69 CFTYPE_ONLY_ON_ROOT
= (1 << 0), /* only create on root cgrp */
70 CFTYPE_NOT_ON_ROOT
= (1 << 1), /* don't create on root cgrp */
71 CFTYPE_NO_PREFIX
= (1 << 3), /* (DON'T USE FOR NEW FILES) no subsys prefix */
73 /* internal flags, do not use outside cgroup core proper */
74 __CFTYPE_ONLY_ON_DFL
= (1 << 16), /* only on default hierarchy */
75 __CFTYPE_NOT_ON_DFL
= (1 << 17), /* not on default hierarchy */
79 * Per-subsystem/per-cgroup state maintained by the system. This is the
80 * fundamental structural building block that controllers deal with.
82 * Fields marked with "PI:" are public and immutable and may be accessed
83 * directly without synchronization.
85 struct cgroup_subsys_state
{
86 /* PI: the cgroup that this css is attached to */
87 struct cgroup
*cgroup
;
89 /* PI: the cgroup subsystem that this css is attached to */
90 struct cgroup_subsys
*ss
;
92 /* reference count - access via css_[try]get() and css_put() */
93 struct percpu_ref refcnt
;
95 /* PI: the parent css */
96 struct cgroup_subsys_state
*parent
;
98 /* siblings list anchored at the parent's ->children */
99 struct list_head sibling
;
100 struct list_head children
;
103 * PI: Subsys-unique ID. 0 is unused and root is always 1. The
104 * matching css can be looked up using css_from_id().
111 * Monotonically increasing unique serial number which defines a
112 * uniform order among all csses. It's guaranteed that all
113 * ->children lists are in the ascending order of ->serial_nr and
114 * used to allow interrupting and resuming iterations.
118 /* percpu_ref killing and RCU release */
119 struct rcu_head rcu_head
;
120 struct work_struct destroy_work
;
124 * A css_set is a structure holding pointers to a set of
125 * cgroup_subsys_state objects. This saves space in the task struct
126 * object and speeds up fork()/exit(), since a single inc/dec and a
127 * list_add()/del() can bump the reference count on the entire cgroup
131 /* Reference count */
135 * List running through all cgroup groups in the same hash
136 * slot. Protected by css_set_lock
138 struct hlist_node hlist
;
141 * Lists running through all tasks using this cgroup group.
142 * mg_tasks lists tasks which belong to this cset but are in the
143 * process of being migrated out or in. Protected by
144 * css_set_rwsem, but, during migration, once tasks are moved to
145 * mg_tasks, it can be read safely while holding cgroup_mutex.
147 struct list_head tasks
;
148 struct list_head mg_tasks
;
151 * List of cgrp_cset_links pointing at cgroups referenced from this
152 * css_set. Protected by css_set_lock.
154 struct list_head cgrp_links
;
156 /* the default cgroup associated with this css_set */
157 struct cgroup
*dfl_cgrp
;
160 * Set of subsystem states, one for each subsystem. This array is
161 * immutable after creation apart from the init_css_set during
162 * subsystem registration (at boot time).
164 struct cgroup_subsys_state
*subsys
[CGROUP_SUBSYS_COUNT
];
167 * List of csets participating in the on-going migration either as
168 * source or destination. Protected by cgroup_mutex.
170 struct list_head mg_preload_node
;
171 struct list_head mg_node
;
174 * If this cset is acting as the source of migration the following
175 * two fields are set. mg_src_cgrp is the source cgroup of the
176 * on-going migration and mg_dst_cset is the destination cset the
177 * target tasks on this cset should be migrated to. Protected by
180 struct cgroup
*mg_src_cgrp
;
181 struct css_set
*mg_dst_cset
;
184 * On the default hierarhcy, ->subsys[ssid] may point to a css
185 * attached to an ancestor instead of the cgroup this css_set is
186 * associated with. The following node is anchored at
187 * ->subsys[ssid]->cgroup->e_csets[ssid] and provides a way to
188 * iterate through all css's attached to a given cgroup.
190 struct list_head e_cset_node
[CGROUP_SUBSYS_COUNT
];
192 /* For RCU-protected deletion */
193 struct rcu_head rcu_head
;
197 /* self css with NULL ->ss, points back to this cgroup */
198 struct cgroup_subsys_state self
;
200 unsigned long flags
; /* "unsigned long" so bitops work */
203 * idr allocated in-hierarchy ID.
205 * ID 0 is not used, the ID of the root cgroup is always 1, and a
206 * new cgroup will be assigned with a smallest available ID.
208 * Allocating/Removing ID must be protected by cgroup_mutex.
213 * If this cgroup contains any tasks, it contributes one to
214 * populated_cnt. All children with non-zero popuplated_cnt of
215 * their own contribute one. The count is zero iff there's no task
216 * in this cgroup or its subtree.
220 struct kernfs_node
*kn
; /* cgroup kernfs entry */
221 struct kernfs_node
*populated_kn
; /* kn for "cgroup.subtree_populated" */
224 * The bitmask of subsystems enabled on the child cgroups.
225 * ->subtree_control is the one configured through
226 * "cgroup.subtree_control" while ->child_subsys_mask is the
227 * effective one which may have more subsystems enabled.
228 * Controller knobs are made available iff it's enabled in
231 unsigned int subtree_control
;
232 unsigned int child_subsys_mask
;
234 /* Private pointers for each registered subsystem */
235 struct cgroup_subsys_state __rcu
*subsys
[CGROUP_SUBSYS_COUNT
];
237 struct cgroup_root
*root
;
240 * List of cgrp_cset_links pointing at css_sets with tasks in this
241 * cgroup. Protected by css_set_lock.
243 struct list_head cset_links
;
246 * On the default hierarchy, a css_set for a cgroup with some
247 * susbsys disabled will point to css's which are associated with
248 * the closest ancestor which has the subsys enabled. The
249 * following lists all css_sets which point to this cgroup's css
250 * for the given subsystem.
252 struct list_head e_csets
[CGROUP_SUBSYS_COUNT
];
255 * list of pidlists, up to two for each namespace (one for procs, one
256 * for tasks); created on demand.
258 struct list_head pidlists
;
259 struct mutex pidlist_mutex
;
261 /* used to wait for offlining of csses */
262 wait_queue_head_t offline_waitq
;
264 /* used to schedule release agent */
265 struct work_struct release_agent_work
;
269 * A cgroup_root represents the root of a cgroup hierarchy, and may be
270 * associated with a kernfs_root to form an active hierarchy. This is
271 * internal to cgroup core. Don't access directly from controllers.
274 struct kernfs_root
*kf_root
;
276 /* The bitmask of subsystems attached to this hierarchy */
277 unsigned int subsys_mask
;
279 /* Unique id for this hierarchy. */
282 /* The root cgroup. Root is destroyed on its release. */
285 /* Number of cgroups in the hierarchy, used only for /proc/cgroups */
288 /* A list running through the active hierarchies */
289 struct list_head root_list
;
291 /* Hierarchy-specific flags */
294 /* IDs for cgroups in this hierarchy */
295 struct idr cgroup_idr
;
297 /* The path to use for release notifications. */
298 char release_agent_path
[PATH_MAX
];
300 /* The name for this hierarchy - may be empty */
301 char name
[MAX_CGROUP_ROOT_NAMELEN
];
305 * struct cftype: handler definitions for cgroup control files
307 * When reading/writing to a file:
308 * - the cgroup to use is file->f_path.dentry->d_parent->d_fsdata
309 * - the 'cftype' of the file is file->f_path.dentry->d_fsdata
313 * By convention, the name should begin with the name of the
314 * subsystem, followed by a period. Zero length string indicates
315 * end of cftype array.
317 char name
[MAX_CFTYPE_NAME
];
320 * If not 0, file mode is set to this value, otherwise it will
321 * be figured out automatically
326 * The maximum length of string, excluding trailing nul, that can
327 * be passed to write. If < PAGE_SIZE-1, PAGE_SIZE-1 is assumed.
329 size_t max_write_len
;
335 * Fields used for internal bookkeeping. Initialized automatically
336 * during registration.
338 struct cgroup_subsys
*ss
; /* NULL for cgroup core files */
339 struct list_head node
; /* anchored at ss->cfts */
340 struct kernfs_ops
*kf_ops
;
343 * read_u64() is a shortcut for the common case of returning a
344 * single integer. Use it in place of read()
346 u64 (*read_u64
)(struct cgroup_subsys_state
*css
, struct cftype
*cft
);
348 * read_s64() is a signed version of read_u64()
350 s64 (*read_s64
)(struct cgroup_subsys_state
*css
, struct cftype
*cft
);
352 /* generic seq_file read interface */
353 int (*seq_show
)(struct seq_file
*sf
, void *v
);
355 /* optional ops, implement all or none */
356 void *(*seq_start
)(struct seq_file
*sf
, loff_t
*ppos
);
357 void *(*seq_next
)(struct seq_file
*sf
, void *v
, loff_t
*ppos
);
358 void (*seq_stop
)(struct seq_file
*sf
, void *v
);
361 * write_u64() is a shortcut for the common case of accepting
362 * a single integer (as parsed by simple_strtoull) from
363 * userspace. Use in place of write(); return 0 or error.
365 int (*write_u64
)(struct cgroup_subsys_state
*css
, struct cftype
*cft
,
368 * write_s64() is a signed version of write_u64()
370 int (*write_s64
)(struct cgroup_subsys_state
*css
, struct cftype
*cft
,
374 * write() is the generic write callback which maps directly to
375 * kernfs write operation and overrides all other operations.
376 * Maximum write size is determined by ->max_write_len. Use
377 * of_css/cft() to access the associated css and cft.
379 ssize_t (*write
)(struct kernfs_open_file
*of
,
380 char *buf
, size_t nbytes
, loff_t off
);
382 #ifdef CONFIG_DEBUG_LOCK_ALLOC
383 struct lock_class_key lockdep_key
;
388 * Control Group subsystem type.
389 * See Documentation/cgroups/cgroups.txt for details
391 struct cgroup_subsys
{
392 struct cgroup_subsys_state
*(*css_alloc
)(struct cgroup_subsys_state
*parent_css
);
393 int (*css_online
)(struct cgroup_subsys_state
*css
);
394 void (*css_offline
)(struct cgroup_subsys_state
*css
);
395 void (*css_released
)(struct cgroup_subsys_state
*css
);
396 void (*css_free
)(struct cgroup_subsys_state
*css
);
397 void (*css_reset
)(struct cgroup_subsys_state
*css
);
398 void (*css_e_css_changed
)(struct cgroup_subsys_state
*css
);
400 int (*can_attach
)(struct cgroup_subsys_state
*css
,
401 struct cgroup_taskset
*tset
);
402 void (*cancel_attach
)(struct cgroup_subsys_state
*css
,
403 struct cgroup_taskset
*tset
);
404 void (*attach
)(struct cgroup_subsys_state
*css
,
405 struct cgroup_taskset
*tset
);
406 void (*fork
)(struct task_struct
*task
);
407 void (*exit
)(struct cgroup_subsys_state
*css
,
408 struct cgroup_subsys_state
*old_css
,
409 struct task_struct
*task
);
410 void (*bind
)(struct cgroup_subsys_state
*root_css
);
416 * If %false, this subsystem is properly hierarchical -
417 * configuration, resource accounting and restriction on a parent
418 * cgroup cover those of its children. If %true, hierarchy support
419 * is broken in some ways - some subsystems ignore hierarchy
420 * completely while others are only implemented half-way.
422 * It's now disallowed to create nested cgroups if the subsystem is
423 * broken and cgroup core will emit a warning message on such
424 * cases. Eventually, all subsystems will be made properly
425 * hierarchical and this will go away.
427 bool broken_hierarchy
;
428 bool warned_broken_hierarchy
;
430 /* the following two fields are initialized automtically during boot */
434 /* link to parent, protected by cgroup_lock() */
435 struct cgroup_root
*root
;
437 /* idr for css->id */
441 * List of cftypes. Each entry is the first entry of an array
442 * terminated by zero length name.
444 struct list_head cfts
;
447 * Base cftypes which are automatically registered. The two can
448 * point to the same array.
450 struct cftype
*dfl_cftypes
; /* for the default hierarchy */
451 struct cftype
*legacy_cftypes
; /* for the legacy hierarchies */
454 * A subsystem may depend on other subsystems. When such subsystem
455 * is enabled on a cgroup, the depended-upon subsystems are enabled
456 * together if available. Subsystems enabled due to dependency are
457 * not visible to userland until explicitly enabled. The following
458 * specifies the mask of subsystems that this one depends on.
460 unsigned int depends_on
;
463 #endif /* CONFIG_CGROUPS */
464 #endif /* _LINUX_CGROUP_DEFS_H */