cgroup.h 25.4 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14
#ifndef _LINUX_CGROUP_H
#define _LINUX_CGROUP_H
/*
 *  cgroup interface
 *
 *  Copyright (C) 2003 BULL SA
 *  Copyright (C) 2004-2006 Silicon Graphics, Inc.
 *
 */

#include <linux/sched.h>
#include <linux/cpumask.h>
#include <linux/nodemask.h>
#include <linux/rcupdate.h>
15
#include <linux/rculist.h>
B
Balbir Singh 已提交
16
#include <linux/cgroupstats.h>
17
#include <linux/rwsem.h>
K
KAMEZAWA Hiroyuki 已提交
18
#include <linux/idr.h>
19
#include <linux/workqueue.h>
20
#include <linux/fs.h>
21
#include <linux/percpu-refcount.h>
22
#include <linux/seq_file.h>
T
Tejun Heo 已提交
23
#include <linux/kernfs.h>
24 25 26 27 28 29

#ifdef CONFIG_CGROUPS

struct cgroupfs_root;
struct cgroup_subsys;
struct inode;
30
struct cgroup;
31 32 33

extern int cgroup_init_early(void);
extern int cgroup_init(void);
34
extern void cgroup_fork(struct task_struct *p);
35
extern void cgroup_post_fork(struct task_struct *p);
36
extern void cgroup_exit(struct task_struct *p, int run_callbacks);
B
Balbir Singh 已提交
37 38
extern int cgroupstats_build(struct cgroupstats *stats,
				struct dentry *dentry);
39

40
extern int proc_cgroup_show(struct seq_file *, void *);
41

T
Tejun Heo 已提交
42
/* define the enumeration of all cgroup subsystems */
43
#define SUBSYS(_x) _x ## _cgrp_id,
44 45
enum cgroup_subsys_id {
#include <linux/cgroup_subsys.h>
46
	CGROUP_SUBSYS_COUNT,
47 48 49
};
#undef SUBSYS

50 51
/* Per-subsystem/per-cgroup state maintained by the system. */
struct cgroup_subsys_state {
52
	/* the cgroup that this css is attached to */
53 54
	struct cgroup *cgroup;

55 56 57
	/* the cgroup subsystem that this css is attached to */
	struct cgroup_subsys *ss;

58 59
	/* reference count - access via css_[try]get() and css_put() */
	struct percpu_ref refcnt;
60

61 62 63
	/* the parent css */
	struct cgroup_subsys_state *parent;

64
	unsigned long flags;
65

66 67
	/* percpu_ref killing and RCU release */
	struct rcu_head rcu_head;
68
	struct work_struct destroy_work;
69 70 71 72
};

/* bits in struct cgroup_subsys_state flags field */
enum {
73
	CSS_ROOT	= (1 << 0), /* this CSS is the root of the subsystem */
74
	CSS_ONLINE	= (1 << 1), /* between ->css_online() and ->css_offline() */
75 76
};

77 78 79 80 81
/**
 * css_get - obtain a reference on the specified css
 * @css: target css
 *
 * The caller must already have a reference.
82 83 84 85
 */
static inline void css_get(struct cgroup_subsys_state *css)
{
	/* We don't need to reference count the root state */
86
	if (!(css->flags & CSS_ROOT))
87
		percpu_ref_get(&css->refcnt);
88
}
P
Paul Menage 已提交
89

90 91 92 93 94 95 96 97 98 99
/**
 * css_tryget - try to obtain a reference on the specified css
 * @css: target css
 *
 * Obtain a reference on @css if it's alive.  The caller naturally needs to
 * ensure that @css is accessible but doesn't have to be holding a
 * reference on it - IOW, RCU protected access is good enough for this
 * function.  Returns %true if a reference count was successfully obtained;
 * %false otherwise.
 */
P
Paul Menage 已提交
100 101
static inline bool css_tryget(struct cgroup_subsys_state *css)
{
102
	if (css->flags & CSS_ROOT)
P
Paul Menage 已提交
103
		return true;
104
	return percpu_ref_tryget(&css->refcnt);
P
Paul Menage 已提交
105 106
}

107 108 109 110 111 112
/**
 * css_put - put a css reference
 * @css: target css
 *
 * Put a reference obtained via css_get() and css_tryget().
 */
113 114
static inline void css_put(struct cgroup_subsys_state *css)
{
115
	if (!(css->flags & CSS_ROOT))
116
		percpu_ref_put(&css->refcnt);
117 118
}

119 120 121
/* bits in struct cgroup flags field */
enum {
	/* Control Group is dead */
122
	CGRP_DEAD,
P
Paul Menage 已提交
123 124 125 126
	/*
	 * Control Group has previously had a child cgroup or a task,
	 * but no longer (only if CGRP_NOTIFY_ON_RELEASE is set)
	 */
127 128 129
	CGRP_RELEASABLE,
	/* Control Group requires release notifications to userspace */
	CGRP_NOTIFY_ON_RELEASE,
130
	/*
131 132 133
	 * Clone the parent's configuration when creating a new child
	 * cpuset cgroup.  For historical reasons, this option can be
	 * specified at mount time and thus is implemented here.
134
	 */
135
	CGRP_CPUSET_CLONE_CHILDREN,
136 137
	/* see the comment above CGRP_ROOT_SANE_BEHAVIOR for details */
	CGRP_SANE_BEHAVIOR,
138 139
};

140 141 142
struct cgroup {
	unsigned long flags;		/* "unsigned long" so bitops work */

143 144 145 146 147
	/*
	 * idr allocated in-hierarchy ID.
	 *
	 * The ID of the root cgroup is always 0, and a new cgroup
	 * will be assigned with a smallest available ID.
148 149
	 *
	 * Allocating/Removing ID must be protected by cgroup_mutex.
150 151
	 */
	int id;
T
Tejun Heo 已提交
152

153 154 155
	/* the number of attached css's */
	int nr_css;

T
Tejun Heo 已提交
156 157
	atomic_t refcnt;

158 159 160 161 162 163 164
	/*
	 * We link our 'sibling' struct into our parent's 'children'.
	 * Our children link their 'sibling' into our 'children'.
	 */
	struct list_head sibling;	/* my parent's children */
	struct list_head children;	/* my children */

P
Paul Menage 已提交
165
	struct cgroup *parent;		/* my parent */
T
Tejun Heo 已提交
166
	struct kernfs_node *kn;		/* cgroup kernfs entry */
167

168 169 170 171 172 173 174 175
	/*
	 * Monotonically increasing unique serial number which defines a
	 * uniform order among all cgroups.  It's guaranteed that all
	 * ->children lists are in the ascending order of ->serial_nr.
	 * It's used to allow interrupting and resuming iterations.
	 */
	u64 serial_nr;

176
	/* Private pointers for each registered subsystem */
177
	struct cgroup_subsys_state __rcu *subsys[CGROUP_SUBSYS_COUNT];
178 179

	struct cgroupfs_root *root;
180 181

	/*
182 183
	 * List of cgrp_cset_links pointing at css_sets with tasks in this
	 * cgroup.  Protected by css_set_lock.
184
	 */
185
	struct list_head cset_links;
186 187 188 189 190 191 192

	/*
	 * Linked list running through all cgroups that can
	 * potentially be reaped by the release agent. Protected by
	 * release_list_lock
	 */
	struct list_head release_list;
193

194 195 196 197 198 199
	/*
	 * list of pidlists, up to two for each namespace (one for procs, one
	 * for tasks); created on demand.
	 */
	struct list_head pidlists;
	struct mutex pidlist_mutex;
200

T
Tejun Heo 已提交
201 202 203
	/* dummy css with NULL ->ss, points back to this cgroup */
	struct cgroup_subsys_state dummy_css;

204
	/* For css percpu_ref killing and RCU-protected deletion */
205
	struct rcu_head rcu_head;
206
	struct work_struct destroy_work;
207 208
};

209 210 211 212
#define MAX_CGROUP_ROOT_NAMELEN 64

/* cgroupfs_root->flags */
enum {
213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228
	/*
	 * Unfortunately, cgroup core and various controllers are riddled
	 * with idiosyncrasies and pointless options.  The following flag,
	 * when set, will force sane behavior - some options are forced on,
	 * others are disallowed, and some controllers will change their
	 * hierarchical or other behaviors.
	 *
	 * The set of behaviors affected by this flag are still being
	 * determined and developed and the mount option for this flag is
	 * prefixed with __DEVEL__.  The prefix will be dropped once we
	 * reach the point where all behaviors are compatible with the
	 * planned unified hierarchy, which will automatically turn on this
	 * flag.
	 *
	 * The followings are the behaviors currently affected this flag.
	 *
229 230
	 * - Mount options "noprefix", "xattr", "clone_children",
	 *   "release_agent" and "name" are disallowed.
231 232 233 234 235 236
	 *
	 * - When mounting an existing superblock, mount options should
	 *   match.
	 *
	 * - Remount is disallowed.
	 *
237 238
	 * - rename(2) is disallowed.
	 *
239 240
	 * - "tasks" is removed.  Everything should be at process
	 *   granularity.  Use "cgroup.procs" instead.
241
	 *
242 243 244
	 * - "cgroup.procs" is not sorted.  pids will be unique unless they
	 *   got recycled inbetween reads.
	 *
245 246
	 * - "release_agent" and "notify_on_release" are removed.
	 *   Replacement notification mechanism will be implemented.
247
	 *
248
	 * - "cgroup.clone_children" is removed.
249
	 *
250 251 252 253
	 * - cpuset: tasks will be kept in empty cpusets when hotplug happens
	 *   and take masks of ancestors with non-empty cpus/mems, instead of
	 *   being moved to an ancestor.
	 *
254 255
	 * - cpuset: a task can be moved into an empty cpuset, and again it
	 *   takes masks of ancestors.
256
	 *
257 258
	 * - memcg: use_hierarchy is on by default and the cgroup file for
	 *   the flag is not created.
259
	 *
260
	 * - blkcg: blk-throttle becomes properly hierarchical.
261 262 263
	 */
	CGRP_ROOT_SANE_BEHAVIOR	= (1 << 0),

264 265
	CGRP_ROOT_NOPREFIX	= (1 << 1), /* mounted subsystems have no named prefix */
	CGRP_ROOT_XATTR		= (1 << 2), /* supports extended attributes */
266 267 268

	/* mount options live below bit 16 */
	CGRP_ROOT_OPTION_MASK	= (1 << 16) - 1,
269 270 271 272
};

/*
 * A cgroupfs_root represents the root of a cgroup hierarchy, and may be
T
Tejun Heo 已提交
273
 * associated with a kernfs_root to form an active hierarchy.  This is
274 275 276
 * internal to cgroup core.  Don't access directly from controllers.
 */
struct cgroupfs_root {
T
Tejun Heo 已提交
277
	struct kernfs_root *kf_root;
278

279
	/* The bitmask of subsystems attached to this hierarchy */
280 281 282 283 284
	unsigned long subsys_mask;

	/* Unique id for this hierarchy. */
	int hierarchy_id;

T
Tejun Heo 已提交
285
	/* The root cgroup.  Root is destroyed on its release. */
286 287
	struct cgroup top_cgroup;

288 289
	/* Number of cgroups in the hierarchy, used only for /proc/cgroups */
	atomic_t nr_cgrps;
290 291 292 293 294 295 296 297

	/* A list running through the active hierarchies */
	struct list_head root_list;

	/* Hierarchy-specific flags */
	unsigned long flags;

	/* IDs for cgroups in this hierarchy */
298
	struct idr cgroup_idr;
299 300 301 302 303 304 305 306

	/* The path to use for release notifications. */
	char release_agent_path[PATH_MAX];

	/* The name for this hierarchy - may be empty */
	char name[MAX_CGROUP_ROOT_NAMELEN];
};

P
Paul Menage 已提交
307 308
/*
 * A css_set is a structure holding pointers to a set of
309 310
 * cgroup_subsys_state objects. This saves space in the task struct
 * object and speeds up fork()/exit(), since a single inc/dec and a
P
Paul Menage 已提交
311 312
 * list_add()/del() can bump the reference count on the entire cgroup
 * set for a task.
313 314 315 316 317
 */

struct css_set {

	/* Reference count */
318
	atomic_t refcount;
319

320 321 322 323 324 325
	/*
	 * List running through all cgroup groups in the same hash
	 * slot. Protected by css_set_lock
	 */
	struct hlist_node hlist;

326
	/*
T
Tejun Heo 已提交
327 328 329 330 331
	 * Lists running through all tasks using this cgroup group.
	 * mg_tasks lists tasks which belong to this cset but are in the
	 * process of being migrated out or in.  Protected by
	 * css_set_rwsem, but, during migration, once tasks are moved to
	 * mg_tasks, it can be read safely while holding cgroup_mutex.
332 333
	 */
	struct list_head tasks;
T
Tejun Heo 已提交
334
	struct list_head mg_tasks;
335 336

	/*
337 338
	 * List of cgrp_cset_links pointing at cgroups referenced from this
	 * css_set.  Protected by css_set_lock.
339
	 */
340
	struct list_head cgrp_links;
341 342

	/*
T
Tejun Heo 已提交
343 344 345
	 * Set of subsystem states, one for each subsystem. This array is
	 * immutable after creation apart from the init_css_set during
	 * subsystem registration (at boot time).
346 347
	 */
	struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT];
348 349 350

	/* For RCU-protected deletion */
	struct rcu_head rcu_head;
351 352
};

P
Paul Menage 已提交
353 354
/*
 * struct cftype: handler definitions for cgroup control files
355 356
 *
 * When reading/writing to a file:
L
Li Zefan 已提交
357
 *	- the cgroup to use is file->f_dentry->d_parent->d_fsdata
358 359 360
 *	- the 'cftype' of the file is file->f_dentry->d_fsdata
 */

361
/* cftype->flags */
362
enum {
L
Li Zefan 已提交
363 364
	CFTYPE_ONLY_ON_ROOT	= (1 << 0),	/* only create on root cgrp */
	CFTYPE_NOT_ON_ROOT	= (1 << 1),	/* don't create on root cgrp */
365
	CFTYPE_INSANE		= (1 << 2),	/* don't create if sane_behavior */
T
Tejun Heo 已提交
366
	CFTYPE_NO_PREFIX	= (1 << 3),	/* (DON'T USE FOR NEW FILES) no subsys prefix */
367
};
368 369 370

#define MAX_CFTYPE_NAME		64

371
struct cftype {
P
Paul Menage 已提交
372 373
	/*
	 * By convention, the name should begin with the name of the
374 375
	 * subsystem, followed by a period.  Zero length string indicates
	 * end of cftype array.
P
Paul Menage 已提交
376
	 */
377 378
	char name[MAX_CFTYPE_NAME];
	int private;
L
Li Zefan 已提交
379 380 381 382
	/*
	 * If not 0, file mode is set to this value, otherwise it will
	 * be figured out automatically
	 */
A
Al Viro 已提交
383
	umode_t mode;
384 385

	/*
386 387 388
	 * The maximum length of string, excluding trailing nul, that can
	 * be passed to write_string.  If < PAGE_SIZE-1, PAGE_SIZE-1 is
	 * assumed.
389 390 391
	 */
	size_t max_write_len;

392 393 394
	/* CFTYPE_* flags */
	unsigned int flags;

395
	/*
T
Tejun Heo 已提交
396 397
	 * Fields used for internal bookkeeping.  Initialized automatically
	 * during registration.
398
	 */
T
Tejun Heo 已提交
399 400
	struct cgroup_subsys *ss;	/* NULL for cgroup core files */
	struct list_head node;		/* anchored at ss->cfts */
T
Tejun Heo 已提交
401 402
	struct kernfs_ops *kf_ops;

403
	/*
404
	 * read_u64() is a shortcut for the common case of returning a
405 406
	 * single integer. Use it in place of read()
	 */
407
	u64 (*read_u64)(struct cgroup_subsys_state *css, struct cftype *cft);
408 409 410
	/*
	 * read_s64() is a signed version of read_u64()
	 */
411
	s64 (*read_s64)(struct cgroup_subsys_state *css, struct cftype *cft);
412 413 414

	/* generic seq_file read interface */
	int (*seq_show)(struct seq_file *sf, void *v);
415

416 417 418 419 420
	/* optional ops, implement all or none */
	void *(*seq_start)(struct seq_file *sf, loff_t *ppos);
	void *(*seq_next)(struct seq_file *sf, void *v, loff_t *ppos);
	void (*seq_stop)(struct seq_file *sf, void *v);

421
	/*
422
	 * write_u64() is a shortcut for the common case of accepting
423 424 425
	 * a single integer (as parsed by simple_strtoull) from
	 * userspace. Use in place of write(); return 0 or error.
	 */
426 427
	int (*write_u64)(struct cgroup_subsys_state *css, struct cftype *cft,
			 u64 val);
428 429 430
	/*
	 * write_s64() is a signed version of write_u64()
	 */
431 432
	int (*write_s64)(struct cgroup_subsys_state *css, struct cftype *cft,
			 s64 val);
433

434 435 436 437 438
	/*
	 * write_string() is passed a nul-terminated kernelspace
	 * buffer of maximum length determined by max_write_len.
	 * Returns 0 or -ve error code.
	 */
439
	int (*write_string)(struct cgroup_subsys_state *css, struct cftype *cft,
440
			    const char *buffer);
441 442 443 444 445 446
	/*
	 * trigger() callback can be used to get some kick from the
	 * userspace, when the actual string written is not important
	 * at all. The private field can be used to determine the
	 * kick type for multiplexing.
	 */
447
	int (*trigger)(struct cgroup_subsys_state *css, unsigned int event);
T
Tejun Heo 已提交
448 449 450 451

#ifdef CONFIG_DEBUG_LOCK_ALLOC
	struct lock_class_key	lockdep_key;
#endif
452 453
};

454 455 456 457 458 459 460 461 462
/*
 * See the comment above CGRP_ROOT_SANE_BEHAVIOR for details.  This
 * function can be called as long as @cgrp is accessible.
 */
static inline bool cgroup_sane_behavior(const struct cgroup *cgrp)
{
	return cgrp->root->flags & CGRP_ROOT_SANE_BEHAVIOR;
}

463 464 465 466 467 468
/* no synchronization, the result can only be used as a hint */
static inline bool cgroup_has_tasks(struct cgroup *cgrp)
{
	return !list_empty(&cgrp->cset_links);
}

T
Tejun Heo 已提交
469 470 471
/* returns ino associated with a cgroup, 0 indicates unmounted root */
static inline ino_t cgroup_ino(struct cgroup *cgrp)
{
T
Tejun Heo 已提交
472 473
	if (cgrp->kn)
		return cgrp->kn->ino;
T
Tejun Heo 已提交
474 475 476 477
	else
		return 0;
}

478 479
static inline struct cftype *seq_cft(struct seq_file *seq)
{
T
Tejun Heo 已提交
480 481 482
	struct kernfs_open_file *of = seq->private;

	return of->kn->priv;
483 484
}

485 486
struct cgroup_subsys_state *seq_css(struct seq_file *seq);

T
Tejun Heo 已提交
487 488 489 490 491 492 493
/*
 * Name / path handling functions.  All are thin wrappers around the kernfs
 * counterparts and can be called under any context.
 */

static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen)
{
494 495 496 497 498
	/* dummy_top doesn't have a kn associated */
	if (cgrp->kn)
		return kernfs_name(cgrp->kn, buf, buflen);
	else
		return strlcpy(buf, "/", buflen);
T
Tejun Heo 已提交
499 500 501 502 503
}

static inline char * __must_check cgroup_path(struct cgroup *cgrp, char *buf,
					      size_t buflen)
{
504 505 506 507 508
	/* dummy_top doesn't have a kn associated */
	if (cgrp->kn)
		return kernfs_path(cgrp->kn, buf, buflen);
	strlcpy(buf, "/", buflen);
	return (buflen <= 2) ? NULL : buf;
T
Tejun Heo 已提交
509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530
}

static inline void pr_cont_cgroup_name(struct cgroup *cgrp)
{
	/* dummy_top doesn't have a kn associated */
	if (cgrp->kn)
		pr_cont_kernfs_name(cgrp->kn);
	else
		pr_cont("/");
}

static inline void pr_cont_cgroup_path(struct cgroup *cgrp)
{
	/* dummy_top doesn't have a kn associated */
	if (cgrp->kn)
		pr_cont_kernfs_path(cgrp->kn);
	else
		pr_cont("/");
}

char *task_cgroup_path(struct task_struct *task, char *buf, size_t buflen);

A
Aristeu Rozanski 已提交
531
int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
532
int cgroup_rm_cftypes(struct cftype *cfts);
533

534
bool cgroup_is_descendant(struct cgroup *cgrp, struct cgroup *ancestor);
535

536 537 538 539 540 541 542 543 544 545 546 547 548
/*
 * Control Group taskset, used to pass around set of tasks to cgroup_subsys
 * methods.
 */
struct cgroup_taskset;
struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset);
struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset);

/**
 * cgroup_taskset_for_each - iterate cgroup_taskset
 * @task: the loop cursor
 * @tset: taskset to iterate
 */
549
#define cgroup_taskset_for_each(task, tset)				\
550
	for ((task) = cgroup_taskset_first((tset)); (task);		\
551
	     (task) = cgroup_taskset_next((tset)))
552

553 554 555 556
/*
 * Control Group subsystem type.
 * See Documentation/cgroups/cgroups.txt for details
 */
557 558

struct cgroup_subsys {
559 560 561 562 563 564 565 566 567 568 569
	struct cgroup_subsys_state *(*css_alloc)(struct cgroup_subsys_state *parent_css);
	int (*css_online)(struct cgroup_subsys_state *css);
	void (*css_offline)(struct cgroup_subsys_state *css);
	void (*css_free)(struct cgroup_subsys_state *css);

	int (*can_attach)(struct cgroup_subsys_state *css,
			  struct cgroup_taskset *tset);
	void (*cancel_attach)(struct cgroup_subsys_state *css,
			      struct cgroup_taskset *tset);
	void (*attach)(struct cgroup_subsys_state *css,
		       struct cgroup_taskset *tset);
570
	void (*fork)(struct task_struct *task);
571 572
	void (*exit)(struct cgroup_subsys_state *css,
		     struct cgroup_subsys_state *old_css,
573
		     struct task_struct *task);
574
	void (*bind)(struct cgroup_subsys_state *root_css);
575

576
	int disabled;
577
	int early_init;
578

579 580 581 582 583 584 585 586 587 588 589 590 591 592 593
	/*
	 * If %false, this subsystem is properly hierarchical -
	 * configuration, resource accounting and restriction on a parent
	 * cgroup cover those of its children.  If %true, hierarchy support
	 * is broken in some ways - some subsystems ignore hierarchy
	 * completely while others are only implemented half-way.
	 *
	 * It's now disallowed to create nested cgroups if the subsystem is
	 * broken and cgroup core will emit a warning message on such
	 * cases.  Eventually, all subsystems will be made properly
	 * hierarchical and this will go away.
	 */
	bool broken_hierarchy;
	bool warned_broken_hierarchy;

594
	/* the following two fields are initialized automtically during boot */
595
	int id;
596 597 598
#define MAX_CGROUP_TYPE_NAMELEN 32
	const char *name;

T
Tejun Heo 已提交
599
	/* link to parent, protected by cgroup_lock() */
600
	struct cgroupfs_root *root;
601

T
Tejun Heo 已提交
602 603 604 605 606
	/*
	 * List of cftypes.  Each entry is the first entry of an array
	 * terminated by zero length name.
	 */
	struct list_head cfts;
607

608
	/* base cftypes, automatically registered with subsys itself */
609
	struct cftype *base_cftypes;
610 611
};

612
#define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys;
613 614 615
#include <linux/cgroup_subsys.h>
#undef SUBSYS

T
Tejun Heo 已提交
616 617 618 619 620 621 622 623 624 625
/**
 * css_parent - find the parent css
 * @css: the target cgroup_subsys_state
 *
 * Return the parent css of @css.  This function is guaranteed to return
 * non-NULL parent as long as @css isn't the root.
 */
static inline
struct cgroup_subsys_state *css_parent(struct cgroup_subsys_state *css)
{
626
	return css->parent;
T
Tejun Heo 已提交
627 628
}

629 630 631 632 633 634 635 636 637 638 639 640
/**
 * task_css_set_check - obtain a task's css_set with extra access conditions
 * @task: the task to obtain css_set for
 * @__c: extra condition expression to be passed to rcu_dereference_check()
 *
 * A task's css_set is RCU protected, initialized and exited while holding
 * task_lock(), and can only be modified while holding both cgroup_mutex
 * and task_lock() while the task is alive.  This macro verifies that the
 * caller is inside proper critical section and returns @task's css_set.
 *
 * The caller can also specify additional allowed conditions via @__c, such
 * as locks used during the cgroup_subsys::attach() methods.
641
 */
T
Tejun Heo 已提交
642 643
#ifdef CONFIG_PROVE_RCU
extern struct mutex cgroup_mutex;
644 645 646 647
#define task_css_set_check(task, __c)					\
	rcu_dereference_check((task)->cgroups,				\
		lockdep_is_held(&(task)->alloc_lock) ||			\
		lockdep_is_held(&cgroup_mutex) || (__c))
T
Tejun Heo 已提交
648
#else
649 650
#define task_css_set_check(task, __c)					\
	rcu_dereference((task)->cgroups)
T
Tejun Heo 已提交
651
#endif
652

653
/**
654
 * task_css_check - obtain css for (task, subsys) w/ extra access conds
655 656 657 658 659 660 661
 * @task: the target task
 * @subsys_id: the target subsystem ID
 * @__c: extra condition expression to be passed to rcu_dereference_check()
 *
 * Return the cgroup_subsys_state for the (@task, @subsys_id) pair.  The
 * synchronization rules are the same as task_css_set_check().
 */
662
#define task_css_check(task, subsys_id, __c)				\
663 664 665 666 667 668 669 670 671 672 673 674 675 676
	task_css_set_check((task), (__c))->subsys[(subsys_id)]

/**
 * task_css_set - obtain a task's css_set
 * @task: the task to obtain css_set for
 *
 * See task_css_set_check().
 */
static inline struct css_set *task_css_set(struct task_struct *task)
{
	return task_css_set_check(task, false);
}

/**
677
 * task_css - obtain css for (task, subsys)
678 679 680
 * @task: the target task
 * @subsys_id: the target subsystem ID
 *
681
 * See task_css_check().
682
 */
683 684
static inline struct cgroup_subsys_state *task_css(struct task_struct *task,
						   int subsys_id)
685
{
686
	return task_css_check(task, subsys_id, false);
687 688
}

689 690
static inline struct cgroup *task_cgroup(struct task_struct *task,
					 int subsys_id)
691
{
692
	return task_css(task, subsys_id)->cgroup;
693 694
}

695 696
struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos,
					   struct cgroup_subsys_state *parent);
697

698 699
struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss);

700
/**
701 702 703
 * css_for_each_child - iterate through children of a css
 * @pos: the css * to use as the loop cursor
 * @parent: css whose children to walk
704
 *
705 706
 * Walk @parent's children.  Must be called under rcu_read_lock().  A child
 * css which hasn't finished ->css_online() or already has finished
707
 * ->css_offline() may show up during traversal and it's each subsystem's
708 709
 * responsibility to verify that each @pos is alive.
 *
710
 * If a subsystem synchronizes against the parent in its ->css_online() and
711
 * before starting iterating, a css which finished ->css_online() is
712
 * guaranteed to be visible in the future iterations.
713 714 715 716
 *
 * It is allowed to temporarily drop RCU read lock during iteration.  The
 * caller is responsible for ensuring that @pos remains accessible until
 * the start of the next iteration by, for example, bumping the css refcnt.
717
 */
718 719 720
#define css_for_each_child(pos, parent)					\
	for ((pos) = css_next_child(NULL, (parent)); (pos);		\
	     (pos) = css_next_child((pos), (parent)))
721

722 723 724 725 726 727
struct cgroup_subsys_state *
css_next_descendant_pre(struct cgroup_subsys_state *pos,
			struct cgroup_subsys_state *css);

struct cgroup_subsys_state *
css_rightmost_descendant(struct cgroup_subsys_state *pos);
728 729

/**
730 731 732
 * css_for_each_descendant_pre - pre-order walk of a css's descendants
 * @pos: the css * to use as the loop cursor
 * @root: css whose descendants to walk
733
 *
734 735
 * Walk @root's descendants.  @root is included in the iteration and the
 * first node to be visited.  Must be called under rcu_read_lock().  A
736
 * descendant css which hasn't finished ->css_online() or already has
737
 * finished ->css_offline() may show up during traversal and it's each
738 739
 * subsystem's responsibility to verify that each @pos is alive.
 *
740 741
 * If a subsystem synchronizes against the parent in its ->css_online() and
 * before starting iterating, and synchronizes against @pos on each
742
 * iteration, any descendant css which finished ->css_online() is
743 744 745 746 747
 * guaranteed to be visible in the future iterations.
 *
 * In other words, the following guarantees that a descendant can't escape
 * state updates of its ancestors.
 *
748
 * my_online(@css)
749
 * {
750 751
 *	Lock @css's parent and @css;
 *	Inherit state from the parent;
752 753 754
 *	Unlock both.
 * }
 *
755
 * my_update_state(@css)
756
 * {
757
 *	css_for_each_descendant_pre(@pos, @css) {
758
 *		Lock @pos;
759 760 761 762
 *		if (@pos == @css)
 *			Update @css's state;
 *		else
 *			Verify @pos is alive and inherit state from its parent;
763 764 765 766 767 768 769 770 771 772
 *		Unlock @pos;
 *	}
 * }
 *
 * As long as the inheriting step, including checking the parent state, is
 * enclosed inside @pos locking, double-locking the parent isn't necessary
 * while inheriting.  The state update to the parent is guaranteed to be
 * visible by walking order and, as long as inheriting operations to the
 * same @pos are atomic to each other, multiple updates racing each other
 * still result in the correct state.  It's guaranateed that at least one
773
 * inheritance happens for any css after the latest update to its parent.
774 775 776 777 778
 *
 * If checking parent's state requires locking the parent, each inheriting
 * iteration should lock and unlock both @pos->parent and @pos.
 *
 * Alternatively, a subsystem may choose to use a single global lock to
779
 * synchronize ->css_online() and ->css_offline() against tree-walking
780
 * operations.
781 782 783 784
 *
 * It is allowed to temporarily drop RCU read lock during iteration.  The
 * caller is responsible for ensuring that @pos remains accessible until
 * the start of the next iteration by, for example, bumping the css refcnt.
785
 */
786 787 788
#define css_for_each_descendant_pre(pos, css)				\
	for ((pos) = css_next_descendant_pre(NULL, (css)); (pos);	\
	     (pos) = css_next_descendant_pre((pos), (css)))
789

790 791 792
struct cgroup_subsys_state *
css_next_descendant_post(struct cgroup_subsys_state *pos,
			 struct cgroup_subsys_state *css);
793 794

/**
795 796 797
 * css_for_each_descendant_post - post-order walk of a css's descendants
 * @pos: the css * to use as the loop cursor
 * @css: css whose descendants to walk
798
 *
799
 * Similar to css_for_each_descendant_pre() but performs post-order
800 801 802
 * traversal instead.  @root is included in the iteration and the last
 * node to be visited.  Note that the walk visibility guarantee described
 * in pre-order walk doesn't apply the same to post-order walks.
803
 */
804 805 806
#define css_for_each_descendant_post(pos, css)				\
	for ((pos) = css_next_descendant_post(NULL, (css)); (pos);	\
	     (pos) = css_next_descendant_post((pos), (css)))
807

808 809 810
/* A css_task_iter should be treated as an opaque object */
struct css_task_iter {
	struct cgroup_subsys_state	*origin_css;
811 812
	struct list_head		*cset_link;
	struct list_head		*task;
813 814
};

815 816 817 818
void css_task_iter_start(struct cgroup_subsys_state *css,
			 struct css_task_iter *it);
struct task_struct *css_task_iter_next(struct css_task_iter *it);
void css_task_iter_end(struct css_task_iter *it);
T
Tejun Heo 已提交
819

M
Michael S. Tsirkin 已提交
820
int cgroup_attach_task_all(struct task_struct *from, struct task_struct *);
821
int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from);
M
Michael S. Tsirkin 已提交
822

823 824
struct cgroup_subsys_state *css_tryget_from_dir(struct dentry *dentry,
						struct cgroup_subsys *ss);
K
KAMEZAWA Hiroyuki 已提交
825

826 827 828 829
#else /* !CONFIG_CGROUPS */

static inline int cgroup_init_early(void) { return 0; }
static inline int cgroup_init(void) { return 0; }
830
static inline void cgroup_fork(struct task_struct *p) {}
831
static inline void cgroup_post_fork(struct task_struct *p) {}
832
static inline void cgroup_exit(struct task_struct *p, int callbacks) {}
833

B
Balbir Singh 已提交
834 835 836 837 838
static inline int cgroupstats_build(struct cgroupstats *stats,
					struct dentry *dentry)
{
	return -EINVAL;
}
839

840
/* No cgroups - nothing to do */
M
Michael S. Tsirkin 已提交
841 842 843 844 845
static inline int cgroup_attach_task_all(struct task_struct *from,
					 struct task_struct *t)
{
	return 0;
}
846

847 848 849
#endif /* !CONFIG_CGROUPS */

#endif /* _LINUX_CGROUP_H */