cgroup.c 145.8 KB
Newer Older
1 2 3 4 5 6
/*
 *  Generic process-grouping system.
 *
 *  Based originally on the cpuset system, extracted by Paul Menage
 *  Copyright (C) 2006 Google, Inc
 *
7 8 9 10
 *  Notifications support
 *  Copyright (C) 2009 Nokia Corporation
 *  Author: Kirill A. Shutemov
 *
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
 *  Copyright notices from the original cpuset code:
 *  --------------------------------------------------
 *  Copyright (C) 2003 BULL SA.
 *  Copyright (C) 2004-2006 Silicon Graphics, Inc.
 *
 *  Portions derived from Patrick Mochel's sysfs code.
 *  sysfs is Copyright (c) 2001-3 Patrick Mochel
 *
 *  2003-10-10 Written by Simon Derr.
 *  2003-10-22 Updates by Stephen Hemminger.
 *  2004 May-July Rework by Paul Jackson.
 *  ---------------------------------------------------
 *
 *  This file is subject to the terms and conditions of the GNU General Public
 *  License.  See the file COPYING in the main directory of the Linux
 *  distribution for more details.
 */

#include <linux/cgroup.h>
30
#include <linux/cred.h>
31
#include <linux/ctype.h>
32 33
#include <linux/errno.h>
#include <linux/fs.h>
34
#include <linux/init_task.h>
35 36 37 38 39 40
#include <linux/kernel.h>
#include <linux/list.h>
#include <linux/mm.h>
#include <linux/mutex.h>
#include <linux/mount.h>
#include <linux/pagemap.h>
41
#include <linux/proc_fs.h>
42 43
#include <linux/rcupdate.h>
#include <linux/sched.h>
44
#include <linux/backing-dev.h>
45 46 47 48 49
#include <linux/seq_file.h>
#include <linux/slab.h>
#include <linux/magic.h>
#include <linux/spinlock.h>
#include <linux/string.h>
50
#include <linux/sort.h>
51
#include <linux/kmod.h>
52
#include <linux/module.h>
B
Balbir Singh 已提交
53 54
#include <linux/delayacct.h>
#include <linux/cgroupstats.h>
55
#include <linux/hash.h>
56
#include <linux/namei.h>
L
Li Zefan 已提交
57
#include <linux/pid_namespace.h>
58
#include <linux/idr.h>
59
#include <linux/vmalloc.h> /* TODO: replace with more sophisticated array */
60 61
#include <linux/eventfd.h>
#include <linux/poll.h>
62
#include <linux/flex_array.h> /* used in cgroup_attach_proc */
63
#include <linux/kthread.h>
B
Balbir Singh 已提交
64

A
Arun Sharma 已提交
65
#include <linux/atomic.h>
66

67 68 69
/* css deactivation bias, makes css->refcnt negative to deny new trygets */
#define CSS_DEACT_BIAS		INT_MIN

T
Tejun Heo 已提交
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
/*
 * cgroup_mutex is the master lock.  Any modification to cgroup or its
 * hierarchy must be performed while holding it.
 *
 * cgroup_root_mutex nests inside cgroup_mutex and should be held to modify
 * cgroupfs_root of any cgroup hierarchy - subsys list, flags,
 * release_agent_path and so on.  Modifying requires both cgroup_mutex and
 * cgroup_root_mutex.  Readers can acquire either of the two.  This is to
 * break the following locking order cycle.
 *
 *  A. cgroup_mutex -> cred_guard_mutex -> s_type->i_mutex_key -> namespace_sem
 *  B. namespace_sem -> cgroup_mutex
 *
 * B happens only through cgroup_show_options() and using cgroup_root_mutex
 * breaks it.
 */
86
static DEFINE_MUTEX(cgroup_mutex);
T
Tejun Heo 已提交
87
static DEFINE_MUTEX(cgroup_root_mutex);
88

B
Ben Blum 已提交
89 90 91 92 93 94
/*
 * Generate an array of cgroup subsystem pointers. At boot time, this is
 * populated up to CGROUP_BUILTIN_SUBSYS_COUNT, and modular subsystems are
 * registered after that. The mutable section of this array is protected by
 * cgroup_mutex.
 */
95
#define SUBSYS(_x) &_x ## _subsys,
B
Ben Blum 已提交
96
static struct cgroup_subsys *subsys[CGROUP_SUBSYS_COUNT] = {
97 98 99
#include <linux/cgroup_subsys.h>
};

100 101
#define MAX_CGROUP_ROOT_NAMELEN 64

102 103 104 105 106 107 108 109 110 111 112 113 114 115
/*
 * A cgroupfs_root represents the root of a cgroup hierarchy,
 * and may be associated with a superblock to form an active
 * hierarchy
 */
struct cgroupfs_root {
	struct super_block *sb;

	/*
	 * The bitmask of subsystems intended to be attached to this
	 * hierarchy
	 */
	unsigned long subsys_bits;

116 117 118
	/* Unique id for this hierarchy. */
	int hierarchy_id;

119 120 121 122 123 124 125 126 127 128 129 130
	/* The bitmask of subsystems currently attached to this hierarchy */
	unsigned long actual_subsys_bits;

	/* A list running through the attached subsystems */
	struct list_head subsys_list;

	/* The root cgroup for this hierarchy */
	struct cgroup top_cgroup;

	/* Tracks how many cgroups are currently defined in hierarchy.*/
	int number_of_cgroups;

131
	/* A list running through the active hierarchies */
132 133
	struct list_head root_list;

134 135 136
	/* All cgroups on this root, cgroup_mutex protected */
	struct list_head allcg_list;

137 138
	/* Hierarchy-specific flags */
	unsigned long flags;
139

140
	/* The path to use for release notifications. */
141
	char release_agent_path[PATH_MAX];
142 143 144

	/* The name for this hierarchy - may be empty */
	char name[MAX_CGROUP_ROOT_NAMELEN];
145 146 147 148 149 150 151 152 153
};

/*
 * The "rootnode" hierarchy is the "dummy hierarchy", reserved for the
 * subsystems that are otherwise unattached - it never has more than a
 * single cgroup, and all tasks are part of that cgroup.
 */
static struct cgroupfs_root rootnode;

T
Tejun Heo 已提交
154 155 156 157 158 159 160 161 162
/*
 * cgroupfs file entry, pointed to from leaf dentry->d_fsdata.
 */
struct cfent {
	struct list_head		node;
	struct dentry			*dentry;
	struct cftype			*type;
};

K
KAMEZAWA Hiroyuki 已提交
163 164 165 166 167 168 169 170 171 172 173 174 175
/*
 * CSS ID -- ID per subsys's Cgroup Subsys State(CSS). used only when
 * cgroup_subsys->use_id != 0.
 */
#define CSS_ID_MAX	(65535)
struct css_id {
	/*
	 * The css to which this ID points. This pointer is set to valid value
	 * after cgroup is populated. If cgroup is removed, this will be NULL.
	 * This pointer is expected to be RCU-safe because destroy()
	 * is called after synchronize_rcu(). But for safe use, css_is_removed()
	 * css_tryget() should be used for avoiding race.
	 */
A
Arnd Bergmann 已提交
176
	struct cgroup_subsys_state __rcu *css;
K
KAMEZAWA Hiroyuki 已提交
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194
	/*
	 * ID of this css.
	 */
	unsigned short id;
	/*
	 * Depth in hierarchy which this ID belongs to.
	 */
	unsigned short depth;
	/*
	 * ID is freed by RCU. (and lookup routine is RCU safe.)
	 */
	struct rcu_head rcu_head;
	/*
	 * Hierarchy of CSS ID belongs to.
	 */
	unsigned short stack[0]; /* Array of Length (depth+1) */
};

195
/*
L
Lucas De Marchi 已提交
196
 * cgroup_event represents events which userspace want to receive.
197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
 */
struct cgroup_event {
	/*
	 * Cgroup which the event belongs to.
	 */
	struct cgroup *cgrp;
	/*
	 * Control file which the event associated.
	 */
	struct cftype *cft;
	/*
	 * eventfd to signal userspace about the event.
	 */
	struct eventfd_ctx *eventfd;
	/*
	 * Each of these stored in a list by the cgroup.
	 */
	struct list_head list;
	/*
	 * All fields below needed to unregister event when
	 * userspace closes eventfd.
	 */
	poll_table pt;
	wait_queue_head_t *wqh;
	wait_queue_t wait;
	struct work_struct remove;
};
K
KAMEZAWA Hiroyuki 已提交
224

225 226 227
/* The list of hierarchy roots */

static LIST_HEAD(roots);
228
static int root_count;
229

230 231 232 233
static DEFINE_IDA(hierarchy_ida);
static int next_hierarchy_id;
static DEFINE_SPINLOCK(hierarchy_id_lock);

234 235 236 237
/* dummytop is a shorthand for the dummy hierarchy's top cgroup */
#define dummytop (&rootnode.top_cgroup)

/* This flag indicates whether tasks in the fork and exit paths should
L
Li Zefan 已提交
238 239 240
 * check for fork/exit handlers to call. This avoids us having to do
 * extra work in the fork/exit path if none of the subsystems need to
 * be called.
241
 */
242
static int need_forkexit_callback __read_mostly;
243

244 245 246 247 248 249 250 251 252 253 254 255 256 257
#ifdef CONFIG_PROVE_LOCKING
int cgroup_lock_is_held(void)
{
	return lockdep_is_held(&cgroup_mutex);
}
#else /* #ifdef CONFIG_PROVE_LOCKING */
int cgroup_lock_is_held(void)
{
	return mutex_is_locked(&cgroup_mutex);
}
#endif /* #else #ifdef CONFIG_PROVE_LOCKING */

EXPORT_SYMBOL_GPL(cgroup_lock_is_held);

258 259 260 261 262
static int css_unbias_refcnt(int refcnt)
{
	return refcnt >= 0 ? refcnt : refcnt - CSS_DEACT_BIAS;
}

263 264 265 266 267
/* the current nr of refs, always >= 0 whether @css is deactivated or not */
static int css_refcnt(struct cgroup_subsys_state *css)
{
	int v = atomic_read(&css->refcnt);

268
	return css_unbias_refcnt(v);
269 270
}

271
/* convenient tests for these bits */
272
inline int cgroup_is_removed(const struct cgroup *cgrp)
273
{
274
	return test_bit(CGRP_REMOVED, &cgrp->flags);
275 276 277 278 279 280 281
}

/* bits in struct cgroupfs_root flags field */
enum {
	ROOT_NOPREFIX, /* mounted subsystems have no named prefix */
};

282
static int cgroup_is_releasable(const struct cgroup *cgrp)
283 284
{
	const int bits =
285 286 287
		(1 << CGRP_RELEASABLE) |
		(1 << CGRP_NOTIFY_ON_RELEASE);
	return (cgrp->flags & bits) == bits;
288 289
}

290
static int notify_on_release(const struct cgroup *cgrp)
291
{
292
	return test_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
293 294
}

295 296 297 298 299
static int clone_children(const struct cgroup *cgrp)
{
	return test_bit(CGRP_CLONE_CHILDREN, &cgrp->flags);
}

300 301 302 303 304 305 306
/*
 * for_each_subsys() allows you to iterate on each subsystem attached to
 * an active hierarchy
 */
#define for_each_subsys(_root, _ss) \
list_for_each_entry(_ss, &_root->subsys_list, sibling)

307 308
/* for_each_active_root() allows you to iterate across the active hierarchies */
#define for_each_active_root(_root) \
309 310
list_for_each_entry(_root, &roots, root_list)

311 312 313 314 315
static inline struct cgroup *__d_cgrp(struct dentry *dentry)
{
	return dentry->d_fsdata;
}

T
Tejun Heo 已提交
316
static inline struct cfent *__d_cfe(struct dentry *dentry)
317 318 319 320
{
	return dentry->d_fsdata;
}

T
Tejun Heo 已提交
321 322 323 324 325
static inline struct cftype *__d_cft(struct dentry *dentry)
{
	return __d_cfe(dentry)->type;
}

326 327 328
/* the list of cgroups eligible for automatic release. Protected by
 * release_list_lock */
static LIST_HEAD(release_list);
329
static DEFINE_RAW_SPINLOCK(release_list_lock);
330 331
static void cgroup_release_agent(struct work_struct *work);
static DECLARE_WORK(release_agent_work, cgroup_release_agent);
332
static void check_for_release(struct cgroup *cgrp);
333

334 335 336 337 338 339
/* Link structure for associating css_set objects with cgroups */
struct cg_cgroup_link {
	/*
	 * List running through cg_cgroup_links associated with a
	 * cgroup, anchored on cgroup->css_sets
	 */
340
	struct list_head cgrp_link_list;
341
	struct cgroup *cgrp;
342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359
	/*
	 * List running through cg_cgroup_links pointing at a
	 * single css_set object, anchored on css_set->cg_links
	 */
	struct list_head cg_link_list;
	struct css_set *cg;
};

/* The default css_set - used by init and its children prior to any
 * hierarchies being mounted. It contains a pointer to the root state
 * for each subsystem. Also used to anchor the list of css_sets. Not
 * reference-counted, to improve performance when child cgroups
 * haven't been created.
 */

static struct css_set init_css_set;
static struct cg_cgroup_link init_css_set_link;

360 361
static int cgroup_init_idr(struct cgroup_subsys *ss,
			   struct cgroup_subsys_state *css);
K
KAMEZAWA Hiroyuki 已提交
362

363 364 365 366 367 368
/* css_set_lock protects the list of css_set objects, and the
 * chain of tasks off each css_set.  Nests outside task->alloc_lock
 * due to cgroup_iter_start() */
static DEFINE_RWLOCK(css_set_lock);
static int css_set_count;

369 370 371 372 373
/*
 * hash table for cgroup groups. This improves the performance to find
 * an existing css_set. This hash doesn't (currently) take into
 * account cgroups in empty hierarchies.
 */
374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
#define CSS_SET_HASH_BITS	7
#define CSS_SET_TABLE_SIZE	(1 << CSS_SET_HASH_BITS)
static struct hlist_head css_set_table[CSS_SET_TABLE_SIZE];

static struct hlist_head *css_set_hash(struct cgroup_subsys_state *css[])
{
	int i;
	int index;
	unsigned long tmp = 0UL;

	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++)
		tmp += (unsigned long)css[i];
	tmp = (tmp >> 16) ^ tmp;

	index = hash_long(tmp, CSS_SET_HASH_BITS);

	return &css_set_table[index];
}

393 394 395 396
/* We don't maintain the lists running through each css_set to its
 * task until after the first call to cgroup_iter_start(). This
 * reduces the fork()/exit() overhead for people who have cgroups
 * compiled into their kernel but not actually in use */
397
static int use_task_css_set_links __read_mostly;
398

399
static void __put_css_set(struct css_set *cg, int taskexit)
400
{
K
KOSAKI Motohiro 已提交
401 402
	struct cg_cgroup_link *link;
	struct cg_cgroup_link *saved_link;
403 404 405 406 407 408 409 410 411 412 413 414
	/*
	 * Ensure that the refcount doesn't hit zero while any readers
	 * can see it. Similar to atomic_dec_and_lock(), but for an
	 * rwlock
	 */
	if (atomic_add_unless(&cg->refcount, -1, 1))
		return;
	write_lock(&css_set_lock);
	if (!atomic_dec_and_test(&cg->refcount)) {
		write_unlock(&css_set_lock);
		return;
	}
415

416 417 418 419 420 421 422 423 424
	/* This css_set is dead. unlink it and release cgroup refcounts */
	hlist_del(&cg->hlist);
	css_set_count--;

	list_for_each_entry_safe(link, saved_link, &cg->cg_links,
				 cg_link_list) {
		struct cgroup *cgrp = link->cgrp;
		list_del(&link->cg_link_list);
		list_del(&link->cgrp_link_list);
425 426
		if (atomic_dec_and_test(&cgrp->count) &&
		    notify_on_release(cgrp)) {
427
			if (taskexit)
428 429
				set_bit(CGRP_RELEASABLE, &cgrp->flags);
			check_for_release(cgrp);
430
		}
431 432

		kfree(link);
433
	}
434 435

	write_unlock(&css_set_lock);
436
	kfree_rcu(cg, rcu_head);
437 438
}

439 440 441 442 443
/*
 * refcounted get/put for css_set objects
 */
static inline void get_css_set(struct css_set *cg)
{
444
	atomic_inc(&cg->refcount);
445 446 447 448
}

static inline void put_css_set(struct css_set *cg)
{
449
	__put_css_set(cg, 0);
450 451
}

452 453
static inline void put_css_set_taskexit(struct css_set *cg)
{
454
	__put_css_set(cg, 1);
455 456
}

457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528
/*
 * compare_css_sets - helper function for find_existing_css_set().
 * @cg: candidate css_set being tested
 * @old_cg: existing css_set for a task
 * @new_cgrp: cgroup that's being entered by the task
 * @template: desired set of css pointers in css_set (pre-calculated)
 *
 * Returns true if "cg" matches "old_cg" except for the hierarchy
 * which "new_cgrp" belongs to, for which it should match "new_cgrp".
 */
static bool compare_css_sets(struct css_set *cg,
			     struct css_set *old_cg,
			     struct cgroup *new_cgrp,
			     struct cgroup_subsys_state *template[])
{
	struct list_head *l1, *l2;

	if (memcmp(template, cg->subsys, sizeof(cg->subsys))) {
		/* Not all subsystems matched */
		return false;
	}

	/*
	 * Compare cgroup pointers in order to distinguish between
	 * different cgroups in heirarchies with no subsystems. We
	 * could get by with just this check alone (and skip the
	 * memcmp above) but on most setups the memcmp check will
	 * avoid the need for this more expensive check on almost all
	 * candidates.
	 */

	l1 = &cg->cg_links;
	l2 = &old_cg->cg_links;
	while (1) {
		struct cg_cgroup_link *cgl1, *cgl2;
		struct cgroup *cg1, *cg2;

		l1 = l1->next;
		l2 = l2->next;
		/* See if we reached the end - both lists are equal length. */
		if (l1 == &cg->cg_links) {
			BUG_ON(l2 != &old_cg->cg_links);
			break;
		} else {
			BUG_ON(l2 == &old_cg->cg_links);
		}
		/* Locate the cgroups associated with these links. */
		cgl1 = list_entry(l1, struct cg_cgroup_link, cg_link_list);
		cgl2 = list_entry(l2, struct cg_cgroup_link, cg_link_list);
		cg1 = cgl1->cgrp;
		cg2 = cgl2->cgrp;
		/* Hierarchies should be linked in the same order. */
		BUG_ON(cg1->root != cg2->root);

		/*
		 * If this hierarchy is the hierarchy of the cgroup
		 * that's changing, then we need to check that this
		 * css_set points to the new cgroup; if it's any other
		 * hierarchy, then this css_set should point to the
		 * same cgroup as the old css_set.
		 */
		if (cg1->root == new_cgrp->root) {
			if (cg1 != new_cgrp)
				return false;
		} else {
			if (cg1 != cg2)
				return false;
		}
	}
	return true;
}

529 530 531
/*
 * find_existing_css_set() is a helper for
 * find_css_set(), and checks to see whether an existing
532
 * css_set is suitable.
533 534 535 536
 *
 * oldcg: the cgroup group that we're using before the cgroup
 * transition
 *
537
 * cgrp: the cgroup that we're moving into
538 539 540 541 542 543
 *
 * template: location in which to build the desired set of subsystem
 * state objects for the new cgroup group
 */
static struct css_set *find_existing_css_set(
	struct css_set *oldcg,
544
	struct cgroup *cgrp,
545
	struct cgroup_subsys_state *template[])
546 547
{
	int i;
548
	struct cgroupfs_root *root = cgrp->root;
549 550 551
	struct hlist_head *hhead;
	struct hlist_node *node;
	struct css_set *cg;
552

B
Ben Blum 已提交
553 554 555 556 557
	/*
	 * Build the set of subsystem state objects that we want to see in the
	 * new css_set. while subsystems can change globally, the entries here
	 * won't change, so no need for locking.
	 */
558
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
L
Li Zefan 已提交
559
		if (root->subsys_bits & (1UL << i)) {
560 561 562
			/* Subsystem is in this hierarchy. So we want
			 * the subsystem state from the new
			 * cgroup */
563
			template[i] = cgrp->subsys[i];
564 565 566 567 568 569 570
		} else {
			/* Subsystem is not in this hierarchy, so we
			 * don't want to change the subsystem state */
			template[i] = oldcg->subsys[i];
		}
	}

571 572
	hhead = css_set_hash(template);
	hlist_for_each_entry(cg, node, hhead, hlist) {
573 574 575 576 577
		if (!compare_css_sets(cg, oldcg, cgrp, template))
			continue;

		/* This css_set matches what we need */
		return cg;
578
	}
579 580 581 582 583

	/* No existing cgroup group matched */
	return NULL;
}

584 585 586 587 588 589 590 591 592 593 594
static void free_cg_links(struct list_head *tmp)
{
	struct cg_cgroup_link *link;
	struct cg_cgroup_link *saved_link;

	list_for_each_entry_safe(link, saved_link, tmp, cgrp_link_list) {
		list_del(&link->cgrp_link_list);
		kfree(link);
	}
}

595 596
/*
 * allocate_cg_links() allocates "count" cg_cgroup_link structures
597
 * and chains them on tmp through their cgrp_link_list fields. Returns 0 on
598 599 600 601 602 603 604 605 606 607
 * success or a negative error
 */
static int allocate_cg_links(int count, struct list_head *tmp)
{
	struct cg_cgroup_link *link;
	int i;
	INIT_LIST_HEAD(tmp);
	for (i = 0; i < count; i++) {
		link = kmalloc(sizeof(*link), GFP_KERNEL);
		if (!link) {
608
			free_cg_links(tmp);
609 610
			return -ENOMEM;
		}
611
		list_add(&link->cgrp_link_list, tmp);
612 613 614 615
	}
	return 0;
}

616 617 618 619 620 621 622 623 624 625 626 627 628 629 630
/**
 * link_css_set - a helper function to link a css_set to a cgroup
 * @tmp_cg_links: cg_cgroup_link objects allocated by allocate_cg_links()
 * @cg: the css_set to be linked
 * @cgrp: the destination cgroup
 */
static void link_css_set(struct list_head *tmp_cg_links,
			 struct css_set *cg, struct cgroup *cgrp)
{
	struct cg_cgroup_link *link;

	BUG_ON(list_empty(tmp_cg_links));
	link = list_first_entry(tmp_cg_links, struct cg_cgroup_link,
				cgrp_link_list);
	link->cg = cg;
631
	link->cgrp = cgrp;
632
	atomic_inc(&cgrp->count);
633
	list_move(&link->cgrp_link_list, &cgrp->css_sets);
634 635 636 637 638
	/*
	 * Always add links to the tail of the list so that the list
	 * is sorted by order of hierarchy creation
	 */
	list_add_tail(&link->cg_link_list, &cg->cg_links);
639 640
}

641 642 643 644 645 646 647 648
/*
 * find_css_set() takes an existing cgroup group and a
 * cgroup object, and returns a css_set object that's
 * equivalent to the old group, but with the given cgroup
 * substituted into the appropriate hierarchy. Must be called with
 * cgroup_mutex held
 */
static struct css_set *find_css_set(
649
	struct css_set *oldcg, struct cgroup *cgrp)
650 651 652 653 654 655
{
	struct css_set *res;
	struct cgroup_subsys_state *template[CGROUP_SUBSYS_COUNT];

	struct list_head tmp_cg_links;

656
	struct hlist_head *hhead;
657
	struct cg_cgroup_link *link;
658

659 660
	/* First see if we already have a cgroup group that matches
	 * the desired set */
661
	read_lock(&css_set_lock);
662
	res = find_existing_css_set(oldcg, cgrp, template);
663 664
	if (res)
		get_css_set(res);
665
	read_unlock(&css_set_lock);
666 667 668 669 670 671 672 673 674 675 676 677 678 679

	if (res)
		return res;

	res = kmalloc(sizeof(*res), GFP_KERNEL);
	if (!res)
		return NULL;

	/* Allocate all the cg_cgroup_link objects that we'll need */
	if (allocate_cg_links(root_count, &tmp_cg_links) < 0) {
		kfree(res);
		return NULL;
	}

680
	atomic_set(&res->refcount, 1);
681 682
	INIT_LIST_HEAD(&res->cg_links);
	INIT_LIST_HEAD(&res->tasks);
683
	INIT_HLIST_NODE(&res->hlist);
684 685 686 687 688 689 690

	/* Copy the set of subsystem state objects generated in
	 * find_existing_css_set() */
	memcpy(res->subsys, template, sizeof(res->subsys));

	write_lock(&css_set_lock);
	/* Add reference counts and links from the new css_set. */
691 692 693 694 695 696
	list_for_each_entry(link, &oldcg->cg_links, cg_link_list) {
		struct cgroup *c = link->cgrp;
		if (c->root == cgrp->root)
			c = cgrp;
		link_css_set(&tmp_cg_links, res, c);
	}
697 698 699 700

	BUG_ON(!list_empty(&tmp_cg_links));

	css_set_count++;
701 702 703 704 705

	/* Add this cgroup group to the hash table */
	hhead = css_set_hash(res->subsys);
	hlist_add_head(&res->hlist, hhead);

706 707 708
	write_unlock(&css_set_lock);

	return res;
709 710
}

711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745
/*
 * Return the cgroup for "task" from the given hierarchy. Must be
 * called with cgroup_mutex held.
 */
static struct cgroup *task_cgroup_from_root(struct task_struct *task,
					    struct cgroupfs_root *root)
{
	struct css_set *css;
	struct cgroup *res = NULL;

	BUG_ON(!mutex_is_locked(&cgroup_mutex));
	read_lock(&css_set_lock);
	/*
	 * No need to lock the task - since we hold cgroup_mutex the
	 * task can't change groups, so the only thing that can happen
	 * is that it exits and its css is set back to init_css_set.
	 */
	css = task->cgroups;
	if (css == &init_css_set) {
		res = &root->top_cgroup;
	} else {
		struct cg_cgroup_link *link;
		list_for_each_entry(link, &css->cg_links, cg_link_list) {
			struct cgroup *c = link->cgrp;
			if (c->root == root) {
				res = c;
				break;
			}
		}
	}
	read_unlock(&css_set_lock);
	BUG_ON(!res);
	return res;
}

746 747 748 749 750 751 752 753 754 755
/*
 * There is one global cgroup mutex. We also require taking
 * task_lock() when dereferencing a task's cgroup subsys pointers.
 * See "The task_lock() exception", at the end of this comment.
 *
 * A task must hold cgroup_mutex to modify cgroups.
 *
 * Any task can increment and decrement the count field without lock.
 * So in general, code holding cgroup_mutex can't rely on the count
 * field not changing.  However, if the count goes to zero, then only
756
 * cgroup_attach_task() can increment it again.  Because a count of zero
757 758 759 760 761 762 763 764 765 766 767 768 769
 * means that no tasks are currently attached, therefore there is no
 * way a task attached to that cgroup can fork (the other way to
 * increment the count).  So code holding cgroup_mutex can safely
 * assume that if the count is zero, it will stay zero. Similarly, if
 * a task holds cgroup_mutex on a cgroup with zero count, it
 * knows that the cgroup won't be removed, as cgroup_rmdir()
 * needs that mutex.
 *
 * The fork and exit callbacks cgroup_fork() and cgroup_exit(), don't
 * (usually) take cgroup_mutex.  These are the two most performance
 * critical pieces of code here.  The exception occurs on cgroup_exit(),
 * when a task in a notify_on_release cgroup exits.  Then cgroup_mutex
 * is taken, and if the cgroup count is zero, a usermode call made
L
Li Zefan 已提交
770 771
 * to the release agent with the name of the cgroup (path relative to
 * the root of cgroup file system) as the argument.
772 773 774 775 776 777 778 779 780 781 782
 *
 * A cgroup can only be deleted if both its 'count' of using tasks
 * is zero, and its list of 'children' cgroups is empty.  Since all
 * tasks in the system use _some_ cgroup, and since there is always at
 * least one task in the system (init, pid == 1), therefore, top_cgroup
 * always has either children cgroups and/or using tasks.  So we don't
 * need a special hack to ensure that top_cgroup cannot be deleted.
 *
 *	The task_lock() exception
 *
 * The need for this exception arises from the action of
783
 * cgroup_attach_task(), which overwrites one tasks cgroup pointer with
L
Li Zefan 已提交
784
 * another.  It does so using cgroup_mutex, however there are
785 786 787
 * several performance critical places that need to reference
 * task->cgroup without the expense of grabbing a system global
 * mutex.  Therefore except as noted below, when dereferencing or, as
788
 * in cgroup_attach_task(), modifying a task'ss cgroup pointer we use
789 790 791 792
 * task_lock(), which acts on a spinlock (task->alloc_lock) already in
 * the task_struct routinely used for such matters.
 *
 * P.S.  One more locking exception.  RCU is used to guard the
793
 * update of a tasks cgroup pointer by cgroup_attach_task()
794 795 796 797 798 799 800 801 802 803
 */

/**
 * cgroup_lock - lock out any changes to cgroup structures
 *
 */
void cgroup_lock(void)
{
	mutex_lock(&cgroup_mutex);
}
B
Ben Blum 已提交
804
EXPORT_SYMBOL_GPL(cgroup_lock);
805 806 807 808 809 810 811 812 813 814

/**
 * cgroup_unlock - release lock on cgroup changes
 *
 * Undo the lock taken in a previous cgroup_lock() call.
 */
void cgroup_unlock(void)
{
	mutex_unlock(&cgroup_mutex);
}
B
Ben Blum 已提交
815
EXPORT_SYMBOL_GPL(cgroup_unlock);
816 817 818 819 820 821 822 823

/*
 * A couple of forward declarations required, due to cyclic reference loop:
 * cgroup_mkdir -> cgroup_create -> cgroup_populate_dir ->
 * cgroup_add_file -> cgroup_create_file -> cgroup_dir_inode_operations
 * -> cgroup_mkdir.
 */

824
static int cgroup_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode);
A
Al Viro 已提交
825
static struct dentry *cgroup_lookup(struct inode *, struct dentry *, unsigned int);
826
static int cgroup_rmdir(struct inode *unused_dir, struct dentry *dentry);
827 828
static int cgroup_populate_dir(struct cgroup *cgrp, bool base_files,
			       unsigned long subsys_mask);
829
static const struct inode_operations cgroup_dir_inode_operations;
830
static const struct file_operations proc_cgroupstats_operations;
831 832

static struct backing_dev_info cgroup_backing_dev_info = {
833
	.name		= "cgroup",
834
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK,
835
};
836

K
KAMEZAWA Hiroyuki 已提交
837 838 839
static int alloc_css_id(struct cgroup_subsys *ss,
			struct cgroup *parent, struct cgroup *child);

A
Al Viro 已提交
840
static struct inode *cgroup_new_inode(umode_t mode, struct super_block *sb)
841 842 843 844
{
	struct inode *inode = new_inode(sb);

	if (inode) {
845
		inode->i_ino = get_next_ino();
846
		inode->i_mode = mode;
847 848
		inode->i_uid = current_fsuid();
		inode->i_gid = current_fsgid();
849 850 851 852 853 854
		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
		inode->i_mapping->backing_dev_info = &cgroup_backing_dev_info;
	}
	return inode;
}

855 856 857 858
/*
 * Call subsys's pre_destroy handler.
 * This is called before css refcnt check.
 */
859
static int cgroup_call_pre_destroy(struct cgroup *cgrp)
860 861
{
	struct cgroup_subsys *ss;
862 863
	int ret = 0;

864 865 866 867 868 869 870 871 872
	for_each_subsys(cgrp->root, ss) {
		if (!ss->pre_destroy)
			continue;

		ret = ss->pre_destroy(cgrp);
		if (ret) {
			/* ->pre_destroy() failure is being deprecated */
			WARN_ON_ONCE(!ss->__DEPRECATED_clear_css_refs);
			break;
873
		}
874
	}
875

876
	return ret;
877 878
}

879 880 881 882
static void cgroup_diput(struct dentry *dentry, struct inode *inode)
{
	/* is dentry a directory ? if so, kfree() associated cgroup */
	if (S_ISDIR(inode->i_mode)) {
883
		struct cgroup *cgrp = dentry->d_fsdata;
884
		struct cgroup_subsys *ss;
885
		BUG_ON(!(cgroup_is_removed(cgrp)));
886 887 888 889 890 891 892
		/* It's possible for external users to be holding css
		 * reference counts on a cgroup; css_put() needs to
		 * be able to access the cgroup after decrementing
		 * the reference count in order to know if it needs to
		 * queue the cgroup to be handled by the release
		 * agent */
		synchronize_rcu();
893 894 895 896 897

		mutex_lock(&cgroup_mutex);
		/*
		 * Release the subsystem state objects.
		 */
898
		for_each_subsys(cgrp->root, ss)
899
			ss->destroy(cgrp);
900 901 902 903

		cgrp->root->number_of_cgroups--;
		mutex_unlock(&cgroup_mutex);

904
		/*
905 906
		 * Drop the active superblock reference that we took when we
		 * created the cgroup
907
		 */
908
		deactivate_super(cgrp->root->sb);
909

910 911 912 913 914 915
		/*
		 * if we're getting rid of the cgroup, refcount should ensure
		 * that there are no pidlists left.
		 */
		BUG_ON(!list_empty(&cgrp->pidlists));

916
		kfree_rcu(cgrp, rcu_head);
T
Tejun Heo 已提交
917 918 919 920 921 922 923 924
	} else {
		struct cfent *cfe = __d_cfe(dentry);
		struct cgroup *cgrp = dentry->d_parent->d_fsdata;

		WARN_ONCE(!list_empty(&cfe->node) &&
			  cgrp != &cgrp->root->top_cgroup,
			  "cfe still linked for %s\n", cfe->type->name);
		kfree(cfe);
925 926 927 928
	}
	iput(inode);
}

929 930 931 932 933
static int cgroup_delete(const struct dentry *d)
{
	return 1;
}

934 935 936 937 938 939 940 941 942
static void remove_dir(struct dentry *d)
{
	struct dentry *parent = dget(d->d_parent);

	d_delete(d);
	simple_rmdir(parent->d_inode, d);
	dput(parent);
}

T
Tejun Heo 已提交
943 944 945 946 947 948 949 950 951 952 953 954 955 956 957
static int cgroup_rm_file(struct cgroup *cgrp, const struct cftype *cft)
{
	struct cfent *cfe;

	lockdep_assert_held(&cgrp->dentry->d_inode->i_mutex);
	lockdep_assert_held(&cgroup_mutex);

	list_for_each_entry(cfe, &cgrp->files, node) {
		struct dentry *d = cfe->dentry;

		if (cft && cfe->type != cft)
			continue;

		dget(d);
		d_delete(d);
958
		simple_unlink(cgrp->dentry->d_inode, d);
T
Tejun Heo 已提交
959 960 961 962
		list_del_init(&cfe->node);
		dput(d);

		return 0;
963
	}
T
Tejun Heo 已提交
964 965 966
	return -ENOENT;
}

967 968 969 970 971 972 973 974
/**
 * cgroup_clear_directory - selective removal of base and subsystem files
 * @dir: directory containing the files
 * @base_files: true if the base files should be removed
 * @subsys_mask: mask of the subsystem ids whose files should be removed
 */
static void cgroup_clear_directory(struct dentry *dir, bool base_files,
				   unsigned long subsys_mask)
T
Tejun Heo 已提交
975 976
{
	struct cgroup *cgrp = __d_cgrp(dir);
977
	struct cgroup_subsys *ss;
T
Tejun Heo 已提交
978

979 980 981 982 983 984 985 986 987 988 989
	for_each_subsys(cgrp->root, ss) {
		struct cftype_set *set;
		if (!test_bit(ss->subsys_id, &subsys_mask))
			continue;
		list_for_each_entry(set, &ss->cftsets, node)
			cgroup_rm_file(cgrp, set->cfts);
	}
	if (base_files) {
		while (!list_empty(&cgrp->files))
			cgroup_rm_file(cgrp, NULL);
	}
990 991 992 993 994 995 996
}

/*
 * NOTE : the dentry must have been dget()'ed
 */
static void cgroup_d_remove_dir(struct dentry *dentry)
{
N
Nick Piggin 已提交
997
	struct dentry *parent;
998
	struct cgroupfs_root *root = dentry->d_sb->s_fs_info;
N
Nick Piggin 已提交
999

1000
	cgroup_clear_directory(dentry, true, root->subsys_bits);
1001

N
Nick Piggin 已提交
1002 1003
	parent = dentry->d_parent;
	spin_lock(&parent->d_lock);
1004
	spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
1005
	list_del_init(&dentry->d_u.d_child);
N
Nick Piggin 已提交
1006 1007
	spin_unlock(&dentry->d_lock);
	spin_unlock(&parent->d_lock);
1008 1009 1010
	remove_dir(dentry);
}

1011 1012 1013 1014 1015 1016
/*
 * A queue for waiters to do rmdir() cgroup. A tasks will sleep when
 * cgroup->count == 0 && list_empty(&cgroup->children) && subsys has some
 * reference to css->refcnt. In general, this refcnt is expected to goes down
 * to zero, soon.
 *
1017
 * CGRP_WAIT_ON_RMDIR flag is set under cgroup's inode->i_mutex;
1018
 */
1019
static DECLARE_WAIT_QUEUE_HEAD(cgroup_rmdir_waitq);
1020

1021
static void cgroup_wakeup_rmdir_waiter(struct cgroup *cgrp)
1022
{
1023
	if (unlikely(test_and_clear_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags)))
1024 1025 1026
		wake_up_all(&cgroup_rmdir_waitq);
}

1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037
void cgroup_exclude_rmdir(struct cgroup_subsys_state *css)
{
	css_get(css);
}

void cgroup_release_and_wakeup_rmdir(struct cgroup_subsys_state *css)
{
	cgroup_wakeup_rmdir_waiter(css->cgroup);
	css_put(css);
}

B
Ben Blum 已提交
1038
/*
B
Ben Blum 已提交
1039 1040 1041
 * Call with cgroup_mutex held. Drops reference counts on modules, including
 * any duplicate ones that parse_cgroupfs_options took. If this function
 * returns an error, no reference counts are touched.
B
Ben Blum 已提交
1042
 */
1043 1044 1045 1046
static int rebind_subsystems(struct cgroupfs_root *root,
			      unsigned long final_bits)
{
	unsigned long added_bits, removed_bits;
1047
	struct cgroup *cgrp = &root->top_cgroup;
1048 1049
	int i;

B
Ben Blum 已提交
1050
	BUG_ON(!mutex_is_locked(&cgroup_mutex));
T
Tejun Heo 已提交
1051
	BUG_ON(!mutex_is_locked(&cgroup_root_mutex));
B
Ben Blum 已提交
1052

1053 1054 1055 1056
	removed_bits = root->actual_subsys_bits & ~final_bits;
	added_bits = final_bits & ~root->actual_subsys_bits;
	/* Check that any added subsystems are currently free */
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
L
Li Zefan 已提交
1057
		unsigned long bit = 1UL << i;
1058 1059 1060
		struct cgroup_subsys *ss = subsys[i];
		if (!(bit & added_bits))
			continue;
B
Ben Blum 已提交
1061 1062 1063 1064 1065 1066
		/*
		 * Nobody should tell us to do a subsys that doesn't exist:
		 * parse_cgroupfs_options should catch that case and refcounts
		 * ensure that subsystems won't disappear once selected.
		 */
		BUG_ON(ss == NULL);
1067 1068 1069 1070 1071 1072 1073 1074 1075 1076
		if (ss->root != &rootnode) {
			/* Subsystem isn't free */
			return -EBUSY;
		}
	}

	/* Currently we don't handle adding/removing subsystems when
	 * any child cgroups exist. This is theoretically supportable
	 * but involves complex error handling, so it's being left until
	 * later */
1077
	if (root->number_of_cgroups > 1)
1078 1079 1080 1081 1082 1083 1084 1085
		return -EBUSY;

	/* Process each subsystem */
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
		struct cgroup_subsys *ss = subsys[i];
		unsigned long bit = 1UL << i;
		if (bit & added_bits) {
			/* We're binding this subsystem to this hierarchy */
B
Ben Blum 已提交
1086
			BUG_ON(ss == NULL);
1087
			BUG_ON(cgrp->subsys[i]);
1088 1089
			BUG_ON(!dummytop->subsys[i]);
			BUG_ON(dummytop->subsys[i]->cgroup != dummytop);
1090 1091
			cgrp->subsys[i] = dummytop->subsys[i];
			cgrp->subsys[i]->cgroup = cgrp;
1092
			list_move(&ss->sibling, &root->subsys_list);
1093
			ss->root = root;
1094
			if (ss->bind)
1095
				ss->bind(cgrp);
B
Ben Blum 已提交
1096
			/* refcount was already taken, and we're keeping it */
1097 1098
		} else if (bit & removed_bits) {
			/* We're removing this subsystem */
B
Ben Blum 已提交
1099
			BUG_ON(ss == NULL);
1100 1101
			BUG_ON(cgrp->subsys[i] != dummytop->subsys[i]);
			BUG_ON(cgrp->subsys[i]->cgroup != cgrp);
1102
			if (ss->bind)
1103
				ss->bind(dummytop);
1104
			dummytop->subsys[i]->cgroup = dummytop;
1105
			cgrp->subsys[i] = NULL;
1106
			subsys[i]->root = &rootnode;
1107
			list_move(&ss->sibling, &rootnode.subsys_list);
B
Ben Blum 已提交
1108 1109
			/* subsystem is now free - drop reference on module */
			module_put(ss->module);
1110 1111
		} else if (bit & final_bits) {
			/* Subsystem state should already exist */
B
Ben Blum 已提交
1112
			BUG_ON(ss == NULL);
1113
			BUG_ON(!cgrp->subsys[i]);
B
Ben Blum 已提交
1114 1115 1116 1117 1118 1119 1120 1121
			/*
			 * a refcount was taken, but we already had one, so
			 * drop the extra reference.
			 */
			module_put(ss->module);
#ifdef CONFIG_MODULE_UNLOAD
			BUG_ON(ss->module && !module_refcount(ss->module));
#endif
1122 1123
		} else {
			/* Subsystem state shouldn't exist */
1124
			BUG_ON(cgrp->subsys[i]);
1125 1126 1127 1128 1129 1130 1131 1132
		}
	}
	root->subsys_bits = root->actual_subsys_bits = final_bits;
	synchronize_rcu();

	return 0;
}

1133
static int cgroup_show_options(struct seq_file *seq, struct dentry *dentry)
1134
{
1135
	struct cgroupfs_root *root = dentry->d_sb->s_fs_info;
1136 1137
	struct cgroup_subsys *ss;

T
Tejun Heo 已提交
1138
	mutex_lock(&cgroup_root_mutex);
1139 1140 1141 1142
	for_each_subsys(root, ss)
		seq_printf(seq, ",%s", ss->name);
	if (test_bit(ROOT_NOPREFIX, &root->flags))
		seq_puts(seq, ",noprefix");
1143 1144
	if (strlen(root->release_agent_path))
		seq_printf(seq, ",release_agent=%s", root->release_agent_path);
1145 1146
	if (clone_children(&root->top_cgroup))
		seq_puts(seq, ",clone_children");
1147 1148
	if (strlen(root->name))
		seq_printf(seq, ",name=%s", root->name);
T
Tejun Heo 已提交
1149
	mutex_unlock(&cgroup_root_mutex);
1150 1151 1152 1153 1154 1155
	return 0;
}

struct cgroup_sb_opts {
	unsigned long subsys_bits;
	unsigned long flags;
1156
	char *release_agent;
1157
	bool clone_children;
1158
	char *name;
1159 1160
	/* User explicitly requested empty subsystem */
	bool none;
1161 1162

	struct cgroupfs_root *new_root;
1163

1164 1165
};

B
Ben Blum 已提交
1166 1167
/*
 * Convert a hierarchy specifier into a bitmask of subsystems and flags. Call
B
Ben Blum 已提交
1168 1169 1170
 * with cgroup_mutex held to protect the subsys[] array. This function takes
 * refcounts on subsystems to be used, unless it returns error, in which case
 * no refcounts are taken.
B
Ben Blum 已提交
1171
 */
B
Ben Blum 已提交
1172
static int parse_cgroupfs_options(char *data, struct cgroup_sb_opts *opts)
1173
{
1174 1175
	char *token, *o = data;
	bool all_ss = false, one_ss = false;
1176
	unsigned long mask = (unsigned long)-1;
B
Ben Blum 已提交
1177 1178
	int i;
	bool module_pin_failed = false;
1179

B
Ben Blum 已提交
1180 1181
	BUG_ON(!mutex_is_locked(&cgroup_mutex));

1182 1183 1184
#ifdef CONFIG_CPUSETS
	mask = ~(1UL << cpuset_subsys_id);
#endif
1185

1186
	memset(opts, 0, sizeof(*opts));
1187 1188 1189 1190

	while ((token = strsep(&o, ",")) != NULL) {
		if (!*token)
			return -EINVAL;
1191
		if (!strcmp(token, "none")) {
1192 1193
			/* Explicitly have no subsystems */
			opts->none = true;
1194 1195 1196 1197 1198 1199 1200 1201 1202 1203
			continue;
		}
		if (!strcmp(token, "all")) {
			/* Mutually exclusive option 'all' + subsystem name */
			if (one_ss)
				return -EINVAL;
			all_ss = true;
			continue;
		}
		if (!strcmp(token, "noprefix")) {
1204
			set_bit(ROOT_NOPREFIX, &opts->flags);
1205 1206 1207
			continue;
		}
		if (!strcmp(token, "clone_children")) {
1208
			opts->clone_children = true;
1209 1210 1211
			continue;
		}
		if (!strncmp(token, "release_agent=", 14)) {
1212 1213 1214
			/* Specifying two release agents is forbidden */
			if (opts->release_agent)
				return -EINVAL;
1215
			opts->release_agent =
1216
				kstrndup(token + 14, PATH_MAX - 1, GFP_KERNEL);
1217 1218
			if (!opts->release_agent)
				return -ENOMEM;
1219 1220 1221
			continue;
		}
		if (!strncmp(token, "name=", 5)) {
1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238
			const char *name = token + 5;
			/* Can't specify an empty name */
			if (!strlen(name))
				return -EINVAL;
			/* Must match [\w.-]+ */
			for (i = 0; i < strlen(name); i++) {
				char c = name[i];
				if (isalnum(c))
					continue;
				if ((c == '.') || (c == '-') || (c == '_'))
					continue;
				return -EINVAL;
			}
			/* Specifying two names is forbidden */
			if (opts->name)
				return -EINVAL;
			opts->name = kstrndup(name,
1239
					      MAX_CGROUP_ROOT_NAMELEN - 1,
1240 1241 1242
					      GFP_KERNEL);
			if (!opts->name)
				return -ENOMEM;
1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269

			continue;
		}

		for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
			struct cgroup_subsys *ss = subsys[i];
			if (ss == NULL)
				continue;
			if (strcmp(token, ss->name))
				continue;
			if (ss->disabled)
				continue;

			/* Mutually exclusive option 'all' + subsystem name */
			if (all_ss)
				return -EINVAL;
			set_bit(i, &opts->subsys_bits);
			one_ss = true;

			break;
		}
		if (i == CGROUP_SUBSYS_COUNT)
			return -ENOENT;
	}

	/*
	 * If the 'all' option was specified select all the subsystems,
1270 1271
	 * otherwise if 'none', 'name=' and a subsystem name options
	 * were not specified, let's default to 'all'
1272
	 */
1273
	if (all_ss || (!one_ss && !opts->none && !opts->name)) {
1274 1275 1276 1277 1278 1279 1280
		for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
			struct cgroup_subsys *ss = subsys[i];
			if (ss == NULL)
				continue;
			if (ss->disabled)
				continue;
			set_bit(i, &opts->subsys_bits);
1281 1282 1283
		}
	}

1284 1285
	/* Consistency checks */

1286 1287 1288 1289 1290 1291 1292 1293 1294
	/*
	 * Option noprefix was introduced just for backward compatibility
	 * with the old cpuset, so we allow noprefix only if mounting just
	 * the cpuset subsystem.
	 */
	if (test_bit(ROOT_NOPREFIX, &opts->flags) &&
	    (opts->subsys_bits & mask))
		return -EINVAL;

1295 1296 1297 1298 1299 1300 1301 1302 1303

	/* Can't specify "none" and some subsystems */
	if (opts->subsys_bits && opts->none)
		return -EINVAL;

	/*
	 * We either have to specify by name or by subsystems. (So all
	 * empty hierarchies must have a name).
	 */
1304
	if (!opts->subsys_bits && !opts->name)
1305 1306
		return -EINVAL;

B
Ben Blum 已提交
1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339
	/*
	 * Grab references on all the modules we'll need, so the subsystems
	 * don't dance around before rebind_subsystems attaches them. This may
	 * take duplicate reference counts on a subsystem that's already used,
	 * but rebind_subsystems handles this case.
	 */
	for (i = CGROUP_BUILTIN_SUBSYS_COUNT; i < CGROUP_SUBSYS_COUNT; i++) {
		unsigned long bit = 1UL << i;

		if (!(bit & opts->subsys_bits))
			continue;
		if (!try_module_get(subsys[i]->module)) {
			module_pin_failed = true;
			break;
		}
	}
	if (module_pin_failed) {
		/*
		 * oops, one of the modules was going away. this means that we
		 * raced with a module_delete call, and to the user this is
		 * essentially a "subsystem doesn't exist" case.
		 */
		for (i--; i >= CGROUP_BUILTIN_SUBSYS_COUNT; i--) {
			/* drop refcounts only on the ones we took */
			unsigned long bit = 1UL << i;

			if (!(bit & opts->subsys_bits))
				continue;
			module_put(subsys[i]->module);
		}
		return -ENOENT;
	}

1340 1341 1342
	return 0;
}

B
Ben Blum 已提交
1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354
static void drop_parsed_module_refcounts(unsigned long subsys_bits)
{
	int i;
	for (i = CGROUP_BUILTIN_SUBSYS_COUNT; i < CGROUP_SUBSYS_COUNT; i++) {
		unsigned long bit = 1UL << i;

		if (!(bit & subsys_bits))
			continue;
		module_put(subsys[i]->module);
	}
}

1355 1356 1357 1358
static int cgroup_remount(struct super_block *sb, int *flags, char *data)
{
	int ret = 0;
	struct cgroupfs_root *root = sb->s_fs_info;
1359
	struct cgroup *cgrp = &root->top_cgroup;
1360
	struct cgroup_sb_opts opts;
1361
	unsigned long added_bits, removed_bits;
1362

1363
	mutex_lock(&cgrp->dentry->d_inode->i_mutex);
1364
	mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1365
	mutex_lock(&cgroup_root_mutex);
1366 1367 1368 1369 1370 1371

	/* See what subsystems are wanted */
	ret = parse_cgroupfs_options(data, &opts);
	if (ret)
		goto out_unlock;

1372 1373 1374 1375 1376
	/* See feature-removal-schedule.txt */
	if (opts.subsys_bits != root->actual_subsys_bits || opts.release_agent)
		pr_warning("cgroup: option changes via remount are deprecated (pid=%d comm=%s)\n",
			   task_tgid_nr(current), current->comm);

1377 1378 1379
	added_bits = opts.subsys_bits & ~root->subsys_bits;
	removed_bits = root->subsys_bits & ~opts.subsys_bits;

B
Ben Blum 已提交
1380 1381 1382
	/* Don't allow flags or name to change at remount */
	if (opts.flags != root->flags ||
	    (opts.name && strcmp(opts.name, root->name))) {
1383
		ret = -EINVAL;
B
Ben Blum 已提交
1384
		drop_parsed_module_refcounts(opts.subsys_bits);
1385 1386 1387
		goto out_unlock;
	}

1388
	ret = rebind_subsystems(root, opts.subsys_bits);
B
Ben Blum 已提交
1389 1390
	if (ret) {
		drop_parsed_module_refcounts(opts.subsys_bits);
1391
		goto out_unlock;
B
Ben Blum 已提交
1392
	}
1393

1394
	/* clear out any existing files and repopulate subsystem files */
1395 1396 1397
	cgroup_clear_directory(cgrp->dentry, false, removed_bits);
	/* re-populate subsystem files */
	cgroup_populate_dir(cgrp, false, added_bits);
1398

1399 1400
	if (opts.release_agent)
		strcpy(root->release_agent_path, opts.release_agent);
1401
 out_unlock:
1402
	kfree(opts.release_agent);
1403
	kfree(opts.name);
T
Tejun Heo 已提交
1404
	mutex_unlock(&cgroup_root_mutex);
1405
	mutex_unlock(&cgroup_mutex);
1406
	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
1407 1408 1409
	return ret;
}

1410
static const struct super_operations cgroup_ops = {
1411 1412 1413 1414 1415 1416
	.statfs = simple_statfs,
	.drop_inode = generic_delete_inode,
	.show_options = cgroup_show_options,
	.remount_fs = cgroup_remount,
};

1417 1418 1419 1420
static void init_cgroup_housekeeping(struct cgroup *cgrp)
{
	INIT_LIST_HEAD(&cgrp->sibling);
	INIT_LIST_HEAD(&cgrp->children);
T
Tejun Heo 已提交
1421
	INIT_LIST_HEAD(&cgrp->files);
1422 1423
	INIT_LIST_HEAD(&cgrp->css_sets);
	INIT_LIST_HEAD(&cgrp->release_list);
1424 1425
	INIT_LIST_HEAD(&cgrp->pidlists);
	mutex_init(&cgrp->pidlist_mutex);
1426 1427
	INIT_LIST_HEAD(&cgrp->event_list);
	spin_lock_init(&cgrp->event_list_lock);
1428
}
1429

1430 1431
static void init_cgroup_root(struct cgroupfs_root *root)
{
1432
	struct cgroup *cgrp = &root->top_cgroup;
1433

1434 1435
	INIT_LIST_HEAD(&root->subsys_list);
	INIT_LIST_HEAD(&root->root_list);
1436
	INIT_LIST_HEAD(&root->allcg_list);
1437
	root->number_of_cgroups = 1;
1438 1439
	cgrp->root = root;
	cgrp->top_cgroup = cgrp;
1440
	list_add_tail(&cgrp->allcg_node, &root->allcg_list);
1441
	init_cgroup_housekeeping(cgrp);
1442 1443
}

1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468
static bool init_root_id(struct cgroupfs_root *root)
{
	int ret = 0;

	do {
		if (!ida_pre_get(&hierarchy_ida, GFP_KERNEL))
			return false;
		spin_lock(&hierarchy_id_lock);
		/* Try to allocate the next unused ID */
		ret = ida_get_new_above(&hierarchy_ida, next_hierarchy_id,
					&root->hierarchy_id);
		if (ret == -ENOSPC)
			/* Try again starting from 0 */
			ret = ida_get_new(&hierarchy_ida, &root->hierarchy_id);
		if (!ret) {
			next_hierarchy_id = root->hierarchy_id + 1;
		} else if (ret != -EAGAIN) {
			/* Can only get here if the 31-bit IDR is full ... */
			BUG_ON(ret);
		}
		spin_unlock(&hierarchy_id_lock);
	} while (ret);
	return true;
}

1469 1470
static int cgroup_test_super(struct super_block *sb, void *data)
{
1471
	struct cgroup_sb_opts *opts = data;
1472 1473
	struct cgroupfs_root *root = sb->s_fs_info;

1474 1475 1476
	/* If we asked for a name then it must match */
	if (opts->name && strcmp(opts->name, root->name))
		return 0;
1477

1478 1479 1480 1481 1482 1483
	/*
	 * If we asked for subsystems (or explicitly for no
	 * subsystems) then they must match
	 */
	if ((opts->subsys_bits || opts->none)
	    && (opts->subsys_bits != root->subsys_bits))
1484 1485 1486 1487 1488
		return 0;

	return 1;
}

1489 1490 1491 1492
static struct cgroupfs_root *cgroup_root_from_opts(struct cgroup_sb_opts *opts)
{
	struct cgroupfs_root *root;

1493
	if (!opts->subsys_bits && !opts->none)
1494 1495 1496 1497 1498 1499
		return NULL;

	root = kzalloc(sizeof(*root), GFP_KERNEL);
	if (!root)
		return ERR_PTR(-ENOMEM);

1500 1501 1502 1503
	if (!init_root_id(root)) {
		kfree(root);
		return ERR_PTR(-ENOMEM);
	}
1504
	init_cgroup_root(root);
1505

1506 1507 1508 1509 1510 1511
	root->subsys_bits = opts->subsys_bits;
	root->flags = opts->flags;
	if (opts->release_agent)
		strcpy(root->release_agent_path, opts->release_agent);
	if (opts->name)
		strcpy(root->name, opts->name);
1512 1513
	if (opts->clone_children)
		set_bit(CGRP_CLONE_CHILDREN, &root->top_cgroup.flags);
1514 1515 1516
	return root;
}

1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528
static void cgroup_drop_root(struct cgroupfs_root *root)
{
	if (!root)
		return;

	BUG_ON(!root->hierarchy_id);
	spin_lock(&hierarchy_id_lock);
	ida_remove(&hierarchy_ida, root->hierarchy_id);
	spin_unlock(&hierarchy_id_lock);
	kfree(root);
}

1529 1530 1531
static int cgroup_set_super(struct super_block *sb, void *data)
{
	int ret;
1532 1533 1534 1535 1536 1537
	struct cgroup_sb_opts *opts = data;

	/* If we don't have a new root, we can't set up a new sb */
	if (!opts->new_root)
		return -EINVAL;

1538
	BUG_ON(!opts->subsys_bits && !opts->none);
1539 1540 1541 1542 1543

	ret = set_anon_super(sb, NULL);
	if (ret)
		return ret;

1544 1545
	sb->s_fs_info = opts->new_root;
	opts->new_root->sb = sb;
1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556

	sb->s_blocksize = PAGE_CACHE_SIZE;
	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
	sb->s_magic = CGROUP_SUPER_MAGIC;
	sb->s_op = &cgroup_ops;

	return 0;
}

static int cgroup_get_rootdir(struct super_block *sb)
{
A
Al Viro 已提交
1557 1558
	static const struct dentry_operations cgroup_dops = {
		.d_iput = cgroup_diput,
1559
		.d_delete = cgroup_delete,
A
Al Viro 已提交
1560 1561
	};

1562 1563 1564 1565 1566 1567 1568 1569 1570 1571
	struct inode *inode =
		cgroup_new_inode(S_IFDIR | S_IRUGO | S_IXUGO | S_IWUSR, sb);

	if (!inode)
		return -ENOMEM;

	inode->i_fop = &simple_dir_operations;
	inode->i_op = &cgroup_dir_inode_operations;
	/* directories start off with i_nlink == 2 (for "." entry) */
	inc_nlink(inode);
1572 1573
	sb->s_root = d_make_root(inode);
	if (!sb->s_root)
1574
		return -ENOMEM;
A
Al Viro 已提交
1575 1576
	/* for everything else we want ->d_op set */
	sb->s_d_op = &cgroup_dops;
1577 1578 1579
	return 0;
}

A
Al Viro 已提交
1580
static struct dentry *cgroup_mount(struct file_system_type *fs_type,
1581
			 int flags, const char *unused_dev_name,
A
Al Viro 已提交
1582
			 void *data)
1583 1584
{
	struct cgroup_sb_opts opts;
1585
	struct cgroupfs_root *root;
1586 1587
	int ret = 0;
	struct super_block *sb;
1588
	struct cgroupfs_root *new_root;
T
Tejun Heo 已提交
1589
	struct inode *inode;
1590 1591

	/* First find the desired set of subsystems */
B
Ben Blum 已提交
1592
	mutex_lock(&cgroup_mutex);
1593
	ret = parse_cgroupfs_options(data, &opts);
B
Ben Blum 已提交
1594
	mutex_unlock(&cgroup_mutex);
1595 1596
	if (ret)
		goto out_err;
1597

1598 1599 1600 1601 1602 1603 1604
	/*
	 * Allocate a new cgroup root. We may not need it if we're
	 * reusing an existing hierarchy.
	 */
	new_root = cgroup_root_from_opts(&opts);
	if (IS_ERR(new_root)) {
		ret = PTR_ERR(new_root);
B
Ben Blum 已提交
1605
		goto drop_modules;
1606
	}
1607
	opts.new_root = new_root;
1608

1609
	/* Locate an existing or new sb for this hierarchy */
D
David Howells 已提交
1610
	sb = sget(fs_type, cgroup_test_super, cgroup_set_super, 0, &opts);
1611
	if (IS_ERR(sb)) {
1612
		ret = PTR_ERR(sb);
1613
		cgroup_drop_root(opts.new_root);
B
Ben Blum 已提交
1614
		goto drop_modules;
1615 1616
	}

1617 1618 1619 1620 1621
	root = sb->s_fs_info;
	BUG_ON(!root);
	if (root == opts.new_root) {
		/* We used the new root structure, so this is a new hierarchy */
		struct list_head tmp_cg_links;
1622
		struct cgroup *root_cgrp = &root->top_cgroup;
1623
		struct cgroupfs_root *existing_root;
1624
		const struct cred *cred;
1625
		int i;
1626 1627 1628 1629 1630 1631

		BUG_ON(sb->s_root != NULL);

		ret = cgroup_get_rootdir(sb);
		if (ret)
			goto drop_new_super;
1632
		inode = sb->s_root->d_inode;
1633

1634
		mutex_lock(&inode->i_mutex);
1635
		mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1636
		mutex_lock(&cgroup_root_mutex);
1637

T
Tejun Heo 已提交
1638 1639 1640 1641 1642 1643
		/* Check for name clashes with existing mounts */
		ret = -EBUSY;
		if (strlen(root->name))
			for_each_active_root(existing_root)
				if (!strcmp(existing_root->name, root->name))
					goto unlock_drop;
1644

1645 1646 1647 1648 1649 1650 1651 1652
		/*
		 * We're accessing css_set_count without locking
		 * css_set_lock here, but that's OK - it can only be
		 * increased by someone holding cgroup_lock, and
		 * that's us. The worst that can happen is that we
		 * have some link structures left over
		 */
		ret = allocate_cg_links(css_set_count, &tmp_cg_links);
T
Tejun Heo 已提交
1653 1654
		if (ret)
			goto unlock_drop;
1655

1656 1657
		ret = rebind_subsystems(root, root->subsys_bits);
		if (ret == -EBUSY) {
1658
			free_cg_links(&tmp_cg_links);
T
Tejun Heo 已提交
1659
			goto unlock_drop;
1660
		}
B
Ben Blum 已提交
1661 1662 1663 1664 1665
		/*
		 * There must be no failure case after here, since rebinding
		 * takes care of subsystems' refcounts, which are explicitly
		 * dropped in the failure exit path.
		 */
1666 1667 1668 1669 1670

		/* EBUSY should be the only error here */
		BUG_ON(ret);

		list_add(&root->root_list, &roots);
1671
		root_count++;
1672

1673
		sb->s_root->d_fsdata = root_cgrp;
1674 1675
		root->top_cgroup.dentry = sb->s_root;

1676 1677 1678
		/* Link the top cgroup in this hierarchy into all
		 * the css_set objects */
		write_lock(&css_set_lock);
1679 1680 1681
		for (i = 0; i < CSS_SET_TABLE_SIZE; i++) {
			struct hlist_head *hhead = &css_set_table[i];
			struct hlist_node *node;
1682
			struct css_set *cg;
1683

1684 1685
			hlist_for_each_entry(cg, node, hhead, hlist)
				link_css_set(&tmp_cg_links, cg, root_cgrp);
1686
		}
1687 1688 1689 1690
		write_unlock(&css_set_lock);

		free_cg_links(&tmp_cg_links);

1691 1692
		BUG_ON(!list_empty(&root_cgrp->sibling));
		BUG_ON(!list_empty(&root_cgrp->children));
1693 1694
		BUG_ON(root->number_of_cgroups != 1);

1695
		cred = override_creds(&init_cred);
1696
		cgroup_populate_dir(root_cgrp, true, root->subsys_bits);
1697
		revert_creds(cred);
T
Tejun Heo 已提交
1698
		mutex_unlock(&cgroup_root_mutex);
1699
		mutex_unlock(&cgroup_mutex);
1700
		mutex_unlock(&inode->i_mutex);
1701 1702 1703 1704 1705
	} else {
		/*
		 * We re-used an existing hierarchy - the new root (if
		 * any) is not needed
		 */
1706
		cgroup_drop_root(opts.new_root);
B
Ben Blum 已提交
1707 1708
		/* no subsys rebinding, so refcounts don't change */
		drop_parsed_module_refcounts(opts.subsys_bits);
1709 1710
	}

1711 1712
	kfree(opts.release_agent);
	kfree(opts.name);
A
Al Viro 已提交
1713
	return dget(sb->s_root);
1714

T
Tejun Heo 已提交
1715 1716 1717 1718
 unlock_drop:
	mutex_unlock(&cgroup_root_mutex);
	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&inode->i_mutex);
1719
 drop_new_super:
1720
	deactivate_locked_super(sb);
B
Ben Blum 已提交
1721 1722
 drop_modules:
	drop_parsed_module_refcounts(opts.subsys_bits);
1723 1724 1725
 out_err:
	kfree(opts.release_agent);
	kfree(opts.name);
A
Al Viro 已提交
1726
	return ERR_PTR(ret);
1727 1728 1729 1730
}

static void cgroup_kill_sb(struct super_block *sb) {
	struct cgroupfs_root *root = sb->s_fs_info;
1731
	struct cgroup *cgrp = &root->top_cgroup;
1732
	int ret;
K
KOSAKI Motohiro 已提交
1733 1734
	struct cg_cgroup_link *link;
	struct cg_cgroup_link *saved_link;
1735 1736 1737 1738

	BUG_ON(!root);

	BUG_ON(root->number_of_cgroups != 1);
1739 1740
	BUG_ON(!list_empty(&cgrp->children));
	BUG_ON(!list_empty(&cgrp->sibling));
1741 1742

	mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1743
	mutex_lock(&cgroup_root_mutex);
1744 1745 1746 1747 1748 1749

	/* Rebind all subsystems back to the default hierarchy */
	ret = rebind_subsystems(root, 0);
	/* Shouldn't be able to fail ... */
	BUG_ON(ret);

1750 1751 1752 1753 1754
	/*
	 * Release all the links from css_sets to this hierarchy's
	 * root cgroup
	 */
	write_lock(&css_set_lock);
K
KOSAKI Motohiro 已提交
1755 1756 1757

	list_for_each_entry_safe(link, saved_link, &cgrp->css_sets,
				 cgrp_link_list) {
1758
		list_del(&link->cg_link_list);
1759
		list_del(&link->cgrp_link_list);
1760 1761 1762 1763
		kfree(link);
	}
	write_unlock(&css_set_lock);

1764 1765 1766 1767
	if (!list_empty(&root->root_list)) {
		list_del(&root->root_list);
		root_count--;
	}
1768

T
Tejun Heo 已提交
1769
	mutex_unlock(&cgroup_root_mutex);
1770 1771 1772
	mutex_unlock(&cgroup_mutex);

	kill_litter_super(sb);
1773
	cgroup_drop_root(root);
1774 1775 1776 1777
}

static struct file_system_type cgroup_fs_type = {
	.name = "cgroup",
A
Al Viro 已提交
1778
	.mount = cgroup_mount,
1779 1780 1781
	.kill_sb = cgroup_kill_sb,
};

1782 1783
static struct kobject *cgroup_kobj;

L
Li Zefan 已提交
1784 1785 1786 1787 1788 1789
/**
 * cgroup_path - generate the path of a cgroup
 * @cgrp: the cgroup in question
 * @buf: the buffer to write the path into
 * @buflen: the length of the buffer
 *
1790 1791 1792
 * Called with cgroup_mutex held or else with an RCU-protected cgroup
 * reference.  Writes path of cgroup into buf.  Returns 0 on success,
 * -errno on error.
1793
 */
1794
int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen)
1795 1796
{
	char *start;
1797 1798
	struct dentry *dentry = rcu_dereference_check(cgrp->dentry,
						      cgroup_lock_is_held());
1799

1800
	if (!dentry || cgrp == dummytop) {
1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812
		/*
		 * Inactive subsystems have no dentry for their root
		 * cgroup
		 */
		strcpy(buf, "/");
		return 0;
	}

	start = buf + buflen;

	*--start = '\0';
	for (;;) {
1813
		int len = dentry->d_name.len;
1814

1815 1816
		if ((start -= len) < buf)
			return -ENAMETOOLONG;
1817
		memcpy(start, dentry->d_name.name, len);
1818 1819
		cgrp = cgrp->parent;
		if (!cgrp)
1820
			break;
1821 1822 1823

		dentry = rcu_dereference_check(cgrp->dentry,
					       cgroup_lock_is_held());
1824
		if (!cgrp->parent)
1825 1826 1827 1828 1829 1830 1831 1832
			continue;
		if (--start < buf)
			return -ENAMETOOLONG;
		*start = '/';
	}
	memmove(buf, start, buf + buflen - start);
	return 0;
}
B
Ben Blum 已提交
1833
EXPORT_SYMBOL_GPL(cgroup_path);
1834

1835 1836 1837
/*
 * Control Group taskset
 */
1838 1839 1840
struct task_and_cgroup {
	struct task_struct	*task;
	struct cgroup		*cgrp;
1841
	struct css_set		*cg;
1842 1843
};

1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914
struct cgroup_taskset {
	struct task_and_cgroup	single;
	struct flex_array	*tc_array;
	int			tc_array_len;
	int			idx;
	struct cgroup		*cur_cgrp;
};

/**
 * cgroup_taskset_first - reset taskset and return the first task
 * @tset: taskset of interest
 *
 * @tset iteration is initialized and the first task is returned.
 */
struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset)
{
	if (tset->tc_array) {
		tset->idx = 0;
		return cgroup_taskset_next(tset);
	} else {
		tset->cur_cgrp = tset->single.cgrp;
		return tset->single.task;
	}
}
EXPORT_SYMBOL_GPL(cgroup_taskset_first);

/**
 * cgroup_taskset_next - iterate to the next task in taskset
 * @tset: taskset of interest
 *
 * Return the next task in @tset.  Iteration must have been initialized
 * with cgroup_taskset_first().
 */
struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset)
{
	struct task_and_cgroup *tc;

	if (!tset->tc_array || tset->idx >= tset->tc_array_len)
		return NULL;

	tc = flex_array_get(tset->tc_array, tset->idx++);
	tset->cur_cgrp = tc->cgrp;
	return tc->task;
}
EXPORT_SYMBOL_GPL(cgroup_taskset_next);

/**
 * cgroup_taskset_cur_cgroup - return the matching cgroup for the current task
 * @tset: taskset of interest
 *
 * Return the cgroup for the current (last returned) task of @tset.  This
 * function must be preceded by either cgroup_taskset_first() or
 * cgroup_taskset_next().
 */
struct cgroup *cgroup_taskset_cur_cgroup(struct cgroup_taskset *tset)
{
	return tset->cur_cgrp;
}
EXPORT_SYMBOL_GPL(cgroup_taskset_cur_cgroup);

/**
 * cgroup_taskset_size - return the number of tasks in taskset
 * @tset: taskset of interest
 */
int cgroup_taskset_size(struct cgroup_taskset *tset)
{
	return tset->tc_array ? tset->tc_array_len : 1;
}
EXPORT_SYMBOL_GPL(cgroup_taskset_size);


B
Ben Blum 已提交
1915 1916 1917 1918 1919
/*
 * cgroup_task_migrate - move a task from one cgroup to another.
 *
 * 'guarantee' is set if the caller promises that a new css_set for the task
 * will already exist. If not set, this function might sleep, and can fail with
1920
 * -ENOMEM. Must be called with cgroup_mutex and threadgroup locked.
B
Ben Blum 已提交
1921
 */
1922 1923
static void cgroup_task_migrate(struct cgroup *cgrp, struct cgroup *oldcgrp,
				struct task_struct *tsk, struct css_set *newcg)
B
Ben Blum 已提交
1924 1925 1926 1927
{
	struct css_set *oldcg;

	/*
1928 1929 1930
	 * We are synchronized through threadgroup_lock() against PF_EXITING
	 * setting such that we can't race against cgroup_exit() changing the
	 * css_set to init_css_set and dropping the old one.
B
Ben Blum 已提交
1931
	 */
1932
	WARN_ON_ONCE(tsk->flags & PF_EXITING);
B
Ben Blum 已提交
1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954
	oldcg = tsk->cgroups;

	task_lock(tsk);
	rcu_assign_pointer(tsk->cgroups, newcg);
	task_unlock(tsk);

	/* Update the css_set linked lists if we're using them */
	write_lock(&css_set_lock);
	if (!list_empty(&tsk->cg_list))
		list_move(&tsk->cg_list, &newcg->tasks);
	write_unlock(&css_set_lock);

	/*
	 * We just gained a reference on oldcg by taking it from the task. As
	 * trading it for newcg is protected by cgroup_mutex, we're safe to drop
	 * it here; it will be freed under RCU.
	 */
	put_css_set(oldcg);

	set_bit(CGRP_RELEASABLE, &oldcgrp->flags);
}

L
Li Zefan 已提交
1955 1956 1957 1958
/**
 * cgroup_attach_task - attach task 'tsk' to cgroup 'cgrp'
 * @cgrp: the cgroup the task is attaching to
 * @tsk: the task to be attached
1959
 *
1960 1961
 * Call with cgroup_mutex and threadgroup locked. May take task_lock of
 * @tsk during call.
1962
 */
1963
int cgroup_attach_task(struct cgroup *cgrp, struct task_struct *tsk)
1964
{
1965
	int retval = 0;
1966
	struct cgroup_subsys *ss, *failed_ss = NULL;
1967 1968
	struct cgroup *oldcgrp;
	struct cgroupfs_root *root = cgrp->root;
1969
	struct cgroup_taskset tset = { };
1970
	struct css_set *newcg;
1971

1972 1973 1974
	/* @tsk either already exited or can't exit until the end */
	if (tsk->flags & PF_EXITING)
		return -ESRCH;
1975 1976

	/* Nothing to do if the task is already in that cgroup */
1977
	oldcgrp = task_cgroup_from_root(tsk, root);
1978
	if (cgrp == oldcgrp)
1979 1980
		return 0;

1981 1982 1983
	tset.single.task = tsk;
	tset.single.cgrp = oldcgrp;

1984 1985
	for_each_subsys(root, ss) {
		if (ss->can_attach) {
1986
			retval = ss->can_attach(cgrp, &tset);
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996
			if (retval) {
				/*
				 * Remember on which subsystem the can_attach()
				 * failed, so that we only call cancel_attach()
				 * against the subsystems whose can_attach()
				 * succeeded. (See below)
				 */
				failed_ss = ss;
				goto out;
			}
1997 1998 1999
		}
	}

2000 2001 2002
	newcg = find_css_set(tsk->cgroups, cgrp);
	if (!newcg) {
		retval = -ENOMEM;
2003
		goto out;
2004 2005 2006
	}

	cgroup_task_migrate(cgrp, oldcgrp, tsk, newcg);
2007

2008
	for_each_subsys(root, ss) {
P
Paul Jackson 已提交
2009
		if (ss->attach)
2010
			ss->attach(cgrp, &tset);
2011
	}
B
Ben Blum 已提交
2012

2013
	synchronize_rcu();
2014 2015 2016 2017 2018

	/*
	 * wake up rmdir() waiter. the rmdir should fail since the cgroup
	 * is no longer empty.
	 */
2019
	cgroup_wakeup_rmdir_waiter(cgrp);
2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031
out:
	if (retval) {
		for_each_subsys(root, ss) {
			if (ss == failed_ss)
				/*
				 * This subsystem was the one that failed the
				 * can_attach() check earlier, so we don't need
				 * to call cancel_attach() against it or any
				 * remaining subsystems.
				 */
				break;
			if (ss->cancel_attach)
2032
				ss->cancel_attach(cgrp, &tset);
2033 2034 2035
		}
	}
	return retval;
2036 2037
}

2038
/**
M
Michael S. Tsirkin 已提交
2039 2040
 * cgroup_attach_task_all - attach task 'tsk' to all cgroups of task 'from'
 * @from: attach to all cgroups of a given task
2041 2042
 * @tsk: the task to be attached
 */
M
Michael S. Tsirkin 已提交
2043
int cgroup_attach_task_all(struct task_struct *from, struct task_struct *tsk)
2044 2045 2046 2047 2048 2049
{
	struct cgroupfs_root *root;
	int retval = 0;

	cgroup_lock();
	for_each_active_root(root) {
M
Michael S. Tsirkin 已提交
2050 2051 2052
		struct cgroup *from_cg = task_cgroup_from_root(from, root);

		retval = cgroup_attach_task(from_cg, tsk);
2053 2054 2055 2056 2057 2058 2059
		if (retval)
			break;
	}
	cgroup_unlock();

	return retval;
}
M
Michael S. Tsirkin 已提交
2060
EXPORT_SYMBOL_GPL(cgroup_attach_task_all);
2061

B
Ben Blum 已提交
2062 2063 2064 2065 2066
/**
 * cgroup_attach_proc - attach all threads in a threadgroup to a cgroup
 * @cgrp: the cgroup to attach to
 * @leader: the threadgroup leader task_struct of the group to be attached
 *
2067 2068
 * Call holding cgroup_mutex and the group_rwsem of the leader. Will take
 * task_lock of each thread in leader's threadgroup individually in turn.
B
Ben Blum 已提交
2069
 */
2070
static int cgroup_attach_proc(struct cgroup *cgrp, struct task_struct *leader)
B
Ben Blum 已提交
2071 2072 2073 2074 2075 2076 2077
{
	int retval, i, group_size;
	struct cgroup_subsys *ss, *failed_ss = NULL;
	/* guaranteed to be initialized later, but the compiler needs this */
	struct cgroupfs_root *root = cgrp->root;
	/* threadgroup list cursor and array */
	struct task_struct *tsk;
2078
	struct task_and_cgroup *tc;
2079
	struct flex_array *group;
2080
	struct cgroup_taskset tset = { };
B
Ben Blum 已提交
2081 2082 2083 2084 2085

	/*
	 * step 0: in order to do expensive, possibly blocking operations for
	 * every thread, we cannot iterate the thread group list, since it needs
	 * rcu or tasklist locked. instead, build an array of all threads in the
2086 2087
	 * group - group_rwsem prevents new threads from appearing, and if
	 * threads exit, this will just be an over-estimate.
B
Ben Blum 已提交
2088 2089
	 */
	group_size = get_nr_threads(leader);
2090
	/* flex_array supports very large thread-groups better than kmalloc. */
2091
	group = flex_array_alloc(sizeof(*tc), group_size, GFP_KERNEL);
B
Ben Blum 已提交
2092 2093
	if (!group)
		return -ENOMEM;
2094 2095 2096 2097
	/* pre-allocate to guarantee space while iterating in rcu read-side. */
	retval = flex_array_prealloc(group, 0, group_size - 1, GFP_KERNEL);
	if (retval)
		goto out_free_group_list;
B
Ben Blum 已提交
2098 2099 2100

	tsk = leader;
	i = 0;
2101 2102 2103 2104 2105 2106
	/*
	 * Prevent freeing of tasks while we take a snapshot. Tasks that are
	 * already PF_EXITING could be freed from underneath us unless we
	 * take an rcu_read_lock.
	 */
	rcu_read_lock();
B
Ben Blum 已提交
2107
	do {
2108 2109
		struct task_and_cgroup ent;

2110 2111 2112 2113
		/* @tsk either already exited or can't exit until the end */
		if (tsk->flags & PF_EXITING)
			continue;

B
Ben Blum 已提交
2114 2115
		/* as per above, nr_threads may decrease, but not increase. */
		BUG_ON(i >= group_size);
2116 2117
		ent.task = tsk;
		ent.cgrp = task_cgroup_from_root(tsk, root);
2118 2119 2120
		/* nothing to do if this task is already in the cgroup */
		if (ent.cgrp == cgrp)
			continue;
2121 2122 2123 2124
		/*
		 * saying GFP_ATOMIC has no effect here because we did prealloc
		 * earlier, but it's good form to communicate our expectations.
		 */
2125
		retval = flex_array_put(group, i, &ent, GFP_ATOMIC);
2126
		BUG_ON(retval != 0);
B
Ben Blum 已提交
2127 2128
		i++;
	} while_each_thread(leader, tsk);
2129
	rcu_read_unlock();
B
Ben Blum 已提交
2130 2131
	/* remember the number of threads in the array for later. */
	group_size = i;
2132 2133
	tset.tc_array = group;
	tset.tc_array_len = group_size;
B
Ben Blum 已提交
2134

2135 2136
	/* methods shouldn't be called if no task is actually migrating */
	retval = 0;
2137
	if (!group_size)
2138
		goto out_free_group_list;
2139

B
Ben Blum 已提交
2140 2141 2142 2143 2144
	/*
	 * step 1: check that we can legitimately attach to the cgroup.
	 */
	for_each_subsys(root, ss) {
		if (ss->can_attach) {
2145
			retval = ss->can_attach(cgrp, &tset);
B
Ben Blum 已提交
2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157
			if (retval) {
				failed_ss = ss;
				goto out_cancel_attach;
			}
		}
	}

	/*
	 * step 2: make sure css_sets exist for all threads to be migrated.
	 * we use find_css_set, which allocates a new one if necessary.
	 */
	for (i = 0; i < group_size; i++) {
2158
		tc = flex_array_get(group, i);
2159 2160 2161 2162
		tc->cg = find_css_set(tc->task->cgroups, cgrp);
		if (!tc->cg) {
			retval = -ENOMEM;
			goto out_put_css_set_refs;
B
Ben Blum 已提交
2163 2164 2165 2166
		}
	}

	/*
2167 2168 2169
	 * step 3: now that we're guaranteed success wrt the css_sets,
	 * proceed to move all tasks to the new cgroup.  There are no
	 * failure cases after here, so this is the commit point.
B
Ben Blum 已提交
2170 2171
	 */
	for (i = 0; i < group_size; i++) {
2172
		tc = flex_array_get(group, i);
2173
		cgroup_task_migrate(cgrp, tc->cgrp, tc->task, tc->cg);
B
Ben Blum 已提交
2174 2175 2176 2177
	}
	/* nothing is sensitive to fork() after this point. */

	/*
2178
	 * step 4: do subsystem attach callbacks.
B
Ben Blum 已提交
2179 2180 2181
	 */
	for_each_subsys(root, ss) {
		if (ss->attach)
2182
			ss->attach(cgrp, &tset);
B
Ben Blum 已提交
2183 2184 2185 2186 2187 2188 2189 2190
	}

	/*
	 * step 5: success! and cleanup
	 */
	synchronize_rcu();
	cgroup_wakeup_rmdir_waiter(cgrp);
	retval = 0;
2191 2192 2193 2194 2195 2196 2197 2198
out_put_css_set_refs:
	if (retval) {
		for (i = 0; i < group_size; i++) {
			tc = flex_array_get(group, i);
			if (!tc->cg)
				break;
			put_css_set(tc->cg);
		}
B
Ben Blum 已提交
2199 2200 2201 2202
	}
out_cancel_attach:
	if (retval) {
		for_each_subsys(root, ss) {
2203
			if (ss == failed_ss)
B
Ben Blum 已提交
2204 2205
				break;
			if (ss->cancel_attach)
2206
				ss->cancel_attach(cgrp, &tset);
B
Ben Blum 已提交
2207 2208 2209
		}
	}
out_free_group_list:
2210
	flex_array_free(group);
B
Ben Blum 已提交
2211 2212 2213 2214 2215
	return retval;
}

/*
 * Find the task_struct of the task to attach by vpid and pass it along to the
2216 2217
 * function to attach either it or all tasks in its threadgroup. Will lock
 * cgroup_mutex and threadgroup; may take task_lock of task.
2218
 */
B
Ben Blum 已提交
2219
static int attach_task_by_pid(struct cgroup *cgrp, u64 pid, bool threadgroup)
2220 2221
{
	struct task_struct *tsk;
2222
	const struct cred *cred = current_cred(), *tcred;
2223 2224
	int ret;

B
Ben Blum 已提交
2225 2226 2227
	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;

2228 2229
retry_find_task:
	rcu_read_lock();
2230
	if (pid) {
2231
		tsk = find_task_by_vpid(pid);
B
Ben Blum 已提交
2232 2233
		if (!tsk) {
			rcu_read_unlock();
2234 2235
			ret= -ESRCH;
			goto out_unlock_cgroup;
2236
		}
B
Ben Blum 已提交
2237 2238 2239 2240
		/*
		 * even if we're attaching all tasks in the thread group, we
		 * only need to check permissions on one of them.
		 */
2241
		tcred = __task_cred(tsk);
2242 2243 2244
		if (!uid_eq(cred->euid, GLOBAL_ROOT_UID) &&
		    !uid_eq(cred->euid, tcred->uid) &&
		    !uid_eq(cred->euid, tcred->suid)) {
2245
			rcu_read_unlock();
2246 2247
			ret = -EACCES;
			goto out_unlock_cgroup;
2248
		}
2249 2250
	} else
		tsk = current;
2251 2252

	if (threadgroup)
2253
		tsk = tsk->group_leader;
2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265

	/*
	 * Workqueue threads may acquire PF_THREAD_BOUND and become
	 * trapped in a cpuset, or RT worker may be born in a cgroup
	 * with no rt_runtime allocated.  Just say no.
	 */
	if (tsk == kthreadd_task || (tsk->flags & PF_THREAD_BOUND)) {
		ret = -EINVAL;
		rcu_read_unlock();
		goto out_unlock_cgroup;
	}

2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282
	get_task_struct(tsk);
	rcu_read_unlock();

	threadgroup_lock(tsk);
	if (threadgroup) {
		if (!thread_group_leader(tsk)) {
			/*
			 * a race with de_thread from another thread's exec()
			 * may strip us of our leadership, if this happens,
			 * there is no choice but to throw this task away and
			 * try again; this is
			 * "double-double-toil-and-trouble-check locking".
			 */
			threadgroup_unlock(tsk);
			put_task_struct(tsk);
			goto retry_find_task;
		}
B
Ben Blum 已提交
2283
		ret = cgroup_attach_proc(cgrp, tsk);
2284
	} else
B
Ben Blum 已提交
2285
		ret = cgroup_attach_task(cgrp, tsk);
2286 2287
	threadgroup_unlock(tsk);

2288
	put_task_struct(tsk);
2289
out_unlock_cgroup:
B
Ben Blum 已提交
2290
	cgroup_unlock();
2291 2292 2293
	return ret;
}

2294
static int cgroup_tasks_write(struct cgroup *cgrp, struct cftype *cft, u64 pid)
B
Ben Blum 已提交
2295 2296 2297 2298 2299
{
	return attach_task_by_pid(cgrp, pid, false);
}

static int cgroup_procs_write(struct cgroup *cgrp, struct cftype *cft, u64 tgid)
2300
{
2301
	return attach_task_by_pid(cgrp, tgid, true);
2302 2303
}

2304 2305 2306 2307
/**
 * cgroup_lock_live_group - take cgroup_mutex and check that cgrp is alive.
 * @cgrp: the cgroup to be checked for liveness
 *
2308 2309
 * On success, returns true; the lock should be later released with
 * cgroup_unlock(). On failure returns false with no lock held.
2310
 */
2311
bool cgroup_lock_live_group(struct cgroup *cgrp)
2312 2313 2314 2315 2316 2317 2318 2319
{
	mutex_lock(&cgroup_mutex);
	if (cgroup_is_removed(cgrp)) {
		mutex_unlock(&cgroup_mutex);
		return false;
	}
	return true;
}
B
Ben Blum 已提交
2320
EXPORT_SYMBOL_GPL(cgroup_lock_live_group);
2321 2322 2323 2324 2325

static int cgroup_release_agent_write(struct cgroup *cgrp, struct cftype *cft,
				      const char *buffer)
{
	BUILD_BUG_ON(sizeof(cgrp->root->release_agent_path) < PATH_MAX);
2326 2327
	if (strlen(buffer) >= PATH_MAX)
		return -EINVAL;
2328 2329
	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;
T
Tejun Heo 已提交
2330
	mutex_lock(&cgroup_root_mutex);
2331
	strcpy(cgrp->root->release_agent_path, buffer);
T
Tejun Heo 已提交
2332
	mutex_unlock(&cgroup_root_mutex);
2333
	cgroup_unlock();
2334 2335 2336 2337 2338 2339 2340 2341 2342 2343
	return 0;
}

static int cgroup_release_agent_show(struct cgroup *cgrp, struct cftype *cft,
				     struct seq_file *seq)
{
	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;
	seq_puts(seq, cgrp->root->release_agent_path);
	seq_putc(seq, '\n');
2344
	cgroup_unlock();
2345 2346 2347
	return 0;
}

2348 2349 2350
/* A buffer size big enough for numbers or short strings */
#define CGROUP_LOCAL_BUFFER_SIZE 64

2351
static ssize_t cgroup_write_X64(struct cgroup *cgrp, struct cftype *cft,
2352 2353 2354
				struct file *file,
				const char __user *userbuf,
				size_t nbytes, loff_t *unused_ppos)
2355
{
2356
	char buffer[CGROUP_LOCAL_BUFFER_SIZE];
2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367
	int retval = 0;
	char *end;

	if (!nbytes)
		return -EINVAL;
	if (nbytes >= sizeof(buffer))
		return -E2BIG;
	if (copy_from_user(buffer, userbuf, nbytes))
		return -EFAULT;

	buffer[nbytes] = 0;     /* nul-terminate */
2368
	if (cft->write_u64) {
K
KOSAKI Motohiro 已提交
2369
		u64 val = simple_strtoull(strstrip(buffer), &end, 0);
2370 2371 2372 2373
		if (*end)
			return -EINVAL;
		retval = cft->write_u64(cgrp, cft, val);
	} else {
K
KOSAKI Motohiro 已提交
2374
		s64 val = simple_strtoll(strstrip(buffer), &end, 0);
2375 2376 2377 2378
		if (*end)
			return -EINVAL;
		retval = cft->write_s64(cgrp, cft, val);
	}
2379 2380 2381 2382 2383
	if (!retval)
		retval = nbytes;
	return retval;
}

2384 2385 2386 2387 2388
static ssize_t cgroup_write_string(struct cgroup *cgrp, struct cftype *cft,
				   struct file *file,
				   const char __user *userbuf,
				   size_t nbytes, loff_t *unused_ppos)
{
2389
	char local_buffer[CGROUP_LOCAL_BUFFER_SIZE];
2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403
	int retval = 0;
	size_t max_bytes = cft->max_write_len;
	char *buffer = local_buffer;

	if (!max_bytes)
		max_bytes = sizeof(local_buffer) - 1;
	if (nbytes >= max_bytes)
		return -E2BIG;
	/* Allocate a dynamic buffer if we need one */
	if (nbytes >= sizeof(local_buffer)) {
		buffer = kmalloc(nbytes + 1, GFP_KERNEL);
		if (buffer == NULL)
			return -ENOMEM;
	}
L
Li Zefan 已提交
2404 2405 2406 2407
	if (nbytes && copy_from_user(buffer, userbuf, nbytes)) {
		retval = -EFAULT;
		goto out;
	}
2408 2409

	buffer[nbytes] = 0;     /* nul-terminate */
K
KOSAKI Motohiro 已提交
2410
	retval = cft->write_string(cgrp, cft, strstrip(buffer));
2411 2412
	if (!retval)
		retval = nbytes;
L
Li Zefan 已提交
2413
out:
2414 2415 2416 2417 2418
	if (buffer != local_buffer)
		kfree(buffer);
	return retval;
}

2419 2420 2421 2422
static ssize_t cgroup_file_write(struct file *file, const char __user *buf,
						size_t nbytes, loff_t *ppos)
{
	struct cftype *cft = __d_cft(file->f_dentry);
2423
	struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
2424

2425
	if (cgroup_is_removed(cgrp))
2426
		return -ENODEV;
2427
	if (cft->write)
2428
		return cft->write(cgrp, cft, file, buf, nbytes, ppos);
2429 2430
	if (cft->write_u64 || cft->write_s64)
		return cgroup_write_X64(cgrp, cft, file, buf, nbytes, ppos);
2431 2432
	if (cft->write_string)
		return cgroup_write_string(cgrp, cft, file, buf, nbytes, ppos);
2433 2434 2435 2436
	if (cft->trigger) {
		int ret = cft->trigger(cgrp, (unsigned int)cft->private);
		return ret ? ret : nbytes;
	}
2437
	return -EINVAL;
2438 2439
}

2440 2441 2442 2443
static ssize_t cgroup_read_u64(struct cgroup *cgrp, struct cftype *cft,
			       struct file *file,
			       char __user *buf, size_t nbytes,
			       loff_t *ppos)
2444
{
2445
	char tmp[CGROUP_LOCAL_BUFFER_SIZE];
2446
	u64 val = cft->read_u64(cgrp, cft);
2447 2448 2449 2450 2451
	int len = sprintf(tmp, "%llu\n", (unsigned long long) val);

	return simple_read_from_buffer(buf, nbytes, ppos, tmp, len);
}

2452 2453 2454 2455 2456
static ssize_t cgroup_read_s64(struct cgroup *cgrp, struct cftype *cft,
			       struct file *file,
			       char __user *buf, size_t nbytes,
			       loff_t *ppos)
{
2457
	char tmp[CGROUP_LOCAL_BUFFER_SIZE];
2458 2459 2460 2461 2462 2463
	s64 val = cft->read_s64(cgrp, cft);
	int len = sprintf(tmp, "%lld\n", (long long) val);

	return simple_read_from_buffer(buf, nbytes, ppos, tmp, len);
}

2464 2465 2466 2467
static ssize_t cgroup_file_read(struct file *file, char __user *buf,
				   size_t nbytes, loff_t *ppos)
{
	struct cftype *cft = __d_cft(file->f_dentry);
2468
	struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
2469

2470
	if (cgroup_is_removed(cgrp))
2471 2472 2473
		return -ENODEV;

	if (cft->read)
2474
		return cft->read(cgrp, cft, file, buf, nbytes, ppos);
2475 2476
	if (cft->read_u64)
		return cgroup_read_u64(cgrp, cft, file, buf, nbytes, ppos);
2477 2478
	if (cft->read_s64)
		return cgroup_read_s64(cgrp, cft, file, buf, nbytes, ppos);
2479 2480 2481
	return -EINVAL;
}

2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501
/*
 * seqfile ops/methods for returning structured data. Currently just
 * supports string->u64 maps, but can be extended in future.
 */

struct cgroup_seqfile_state {
	struct cftype *cft;
	struct cgroup *cgroup;
};

static int cgroup_map_add(struct cgroup_map_cb *cb, const char *key, u64 value)
{
	struct seq_file *sf = cb->state;
	return seq_printf(sf, "%s %llu\n", key, (unsigned long long)value);
}

static int cgroup_seqfile_show(struct seq_file *m, void *arg)
{
	struct cgroup_seqfile_state *state = m->private;
	struct cftype *cft = state->cft;
2502 2503 2504 2505 2506 2507 2508 2509
	if (cft->read_map) {
		struct cgroup_map_cb cb = {
			.fill = cgroup_map_add,
			.state = m,
		};
		return cft->read_map(state->cgroup, cft, &cb);
	}
	return cft->read_seq_string(state->cgroup, cft, m);
2510 2511
}

2512
static int cgroup_seqfile_release(struct inode *inode, struct file *file)
2513 2514 2515 2516 2517 2518
{
	struct seq_file *seq = file->private_data;
	kfree(seq->private);
	return single_release(inode, file);
}

2519
static const struct file_operations cgroup_seqfile_operations = {
2520
	.read = seq_read,
2521
	.write = cgroup_file_write,
2522 2523 2524 2525
	.llseek = seq_lseek,
	.release = cgroup_seqfile_release,
};

2526 2527 2528 2529 2530 2531 2532 2533 2534
static int cgroup_file_open(struct inode *inode, struct file *file)
{
	int err;
	struct cftype *cft;

	err = generic_file_open(inode, file);
	if (err)
		return err;
	cft = __d_cft(file->f_dentry);
2535

2536
	if (cft->read_map || cft->read_seq_string) {
2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547
		struct cgroup_seqfile_state *state =
			kzalloc(sizeof(*state), GFP_USER);
		if (!state)
			return -ENOMEM;
		state->cft = cft;
		state->cgroup = __d_cgrp(file->f_dentry->d_parent);
		file->f_op = &cgroup_seqfile_operations;
		err = single_open(file, cgroup_seqfile_show, state);
		if (err < 0)
			kfree(state);
	} else if (cft->open)
2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577
		err = cft->open(inode, file);
	else
		err = 0;

	return err;
}

static int cgroup_file_release(struct inode *inode, struct file *file)
{
	struct cftype *cft = __d_cft(file->f_dentry);
	if (cft->release)
		return cft->release(inode, file);
	return 0;
}

/*
 * cgroup_rename - Only allow simple rename of directories in place.
 */
static int cgroup_rename(struct inode *old_dir, struct dentry *old_dentry,
			    struct inode *new_dir, struct dentry *new_dentry)
{
	if (!S_ISDIR(old_dentry->d_inode->i_mode))
		return -ENOTDIR;
	if (new_dentry->d_inode)
		return -EEXIST;
	if (old_dir != new_dir)
		return -EIO;
	return simple_rename(old_dir, old_dentry, new_dir, new_dentry);
}

2578
static const struct file_operations cgroup_file_operations = {
2579 2580 2581 2582 2583 2584 2585
	.read = cgroup_file_read,
	.write = cgroup_file_write,
	.llseek = generic_file_llseek,
	.open = cgroup_file_open,
	.release = cgroup_file_release,
};

2586
static const struct inode_operations cgroup_dir_inode_operations = {
2587
	.lookup = cgroup_lookup,
2588 2589 2590 2591 2592
	.mkdir = cgroup_mkdir,
	.rmdir = cgroup_rmdir,
	.rename = cgroup_rename,
};

A
Al Viro 已提交
2593
static struct dentry *cgroup_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2594 2595 2596 2597 2598 2599 2600
{
	if (dentry->d_name.len > NAME_MAX)
		return ERR_PTR(-ENAMETOOLONG);
	d_add(dentry, NULL);
	return NULL;
}

2601 2602 2603 2604 2605 2606 2607 2608 2609 2610
/*
 * Check if a file is a control file
 */
static inline struct cftype *__file_cft(struct file *file)
{
	if (file->f_dentry->d_inode->i_fop != &cgroup_file_operations)
		return ERR_PTR(-EINVAL);
	return __d_cft(file->f_dentry);
}

A
Al Viro 已提交
2611
static int cgroup_create_file(struct dentry *dentry, umode_t mode,
2612 2613
				struct super_block *sb)
{
2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633
	struct inode *inode;

	if (!dentry)
		return -ENOENT;
	if (dentry->d_inode)
		return -EEXIST;

	inode = cgroup_new_inode(mode, sb);
	if (!inode)
		return -ENOMEM;

	if (S_ISDIR(mode)) {
		inode->i_op = &cgroup_dir_inode_operations;
		inode->i_fop = &simple_dir_operations;

		/* start off with i_nlink == 2 (for "." entry) */
		inc_nlink(inode);

		/* start with the directory inode held, so that we can
		 * populate it without racing with another mkdir */
2634
		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2635 2636 2637 2638 2639 2640 2641 2642 2643 2644
	} else if (S_ISREG(mode)) {
		inode->i_size = 0;
		inode->i_fop = &cgroup_file_operations;
	}
	d_instantiate(dentry, inode);
	dget(dentry);	/* Extra count - pin the dentry in core */
	return 0;
}

/*
L
Li Zefan 已提交
2645 2646 2647 2648 2649
 * cgroup_create_dir - create a directory for an object.
 * @cgrp: the cgroup we create the directory for. It must have a valid
 *        ->parent field. And we are going to fill its ->dentry field.
 * @dentry: dentry of the new cgroup
 * @mode: mode to set on new directory.
2650
 */
2651
static int cgroup_create_dir(struct cgroup *cgrp, struct dentry *dentry,
A
Al Viro 已提交
2652
				umode_t mode)
2653 2654 2655 2656
{
	struct dentry *parent;
	int error = 0;

2657 2658
	parent = cgrp->parent->dentry;
	error = cgroup_create_file(dentry, S_IFDIR | mode, cgrp->root->sb);
2659
	if (!error) {
2660
		dentry->d_fsdata = cgrp;
2661
		inc_nlink(parent->d_inode);
2662
		rcu_assign_pointer(cgrp->dentry, dentry);
2663 2664 2665 2666 2667 2668 2669
		dget(dentry);
	}
	dput(dentry);

	return error;
}

L
Li Zefan 已提交
2670 2671 2672 2673 2674 2675 2676 2677 2678
/**
 * cgroup_file_mode - deduce file mode of a control file
 * @cft: the control file in question
 *
 * returns cft->mode if ->mode is not 0
 * returns S_IRUGO|S_IWUSR if it has both a read and a write handler
 * returns S_IRUGO if it has only a read handler
 * returns S_IWUSR if it has only a write hander
 */
A
Al Viro 已提交
2679
static umode_t cgroup_file_mode(const struct cftype *cft)
L
Li Zefan 已提交
2680
{
A
Al Viro 已提交
2681
	umode_t mode = 0;
L
Li Zefan 已提交
2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696

	if (cft->mode)
		return cft->mode;

	if (cft->read || cft->read_u64 || cft->read_s64 ||
	    cft->read_map || cft->read_seq_string)
		mode |= S_IRUGO;

	if (cft->write || cft->write_u64 || cft->write_s64 ||
	    cft->write_string || cft->trigger)
		mode |= S_IWUSR;

	return mode;
}

T
Tejun Heo 已提交
2697 2698
static int cgroup_add_file(struct cgroup *cgrp, struct cgroup_subsys *subsys,
			   const struct cftype *cft)
2699
{
2700
	struct dentry *dir = cgrp->dentry;
T
Tejun Heo 已提交
2701
	struct cgroup *parent = __d_cgrp(dir);
2702
	struct dentry *dentry;
T
Tejun Heo 已提交
2703
	struct cfent *cfe;
2704
	int error;
A
Al Viro 已提交
2705
	umode_t mode;
2706
	char name[MAX_CGROUP_TYPE_NAMELEN + MAX_CFTYPE_NAME + 2] = { 0 };
2707 2708 2709 2710 2711 2712 2713

	/* does @cft->flags tell us to skip creation on @cgrp? */
	if ((cft->flags & CFTYPE_NOT_ON_ROOT) && !cgrp->parent)
		return 0;
	if ((cft->flags & CFTYPE_ONLY_ON_ROOT) && cgrp->parent)
		return 0;

2714
	if (subsys && !test_bit(ROOT_NOPREFIX, &cgrp->root->flags)) {
2715 2716 2717 2718
		strcpy(name, subsys->name);
		strcat(name, ".");
	}
	strcat(name, cft->name);
T
Tejun Heo 已提交
2719

2720
	BUG_ON(!mutex_is_locked(&dir->d_inode->i_mutex));
T
Tejun Heo 已提交
2721 2722 2723 2724 2725

	cfe = kzalloc(sizeof(*cfe), GFP_KERNEL);
	if (!cfe)
		return -ENOMEM;

2726
	dentry = lookup_one_len(name, dir, strlen(name));
T
Tejun Heo 已提交
2727
	if (IS_ERR(dentry)) {
2728
		error = PTR_ERR(dentry);
T
Tejun Heo 已提交
2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743
		goto out;
	}

	mode = cgroup_file_mode(cft);
	error = cgroup_create_file(dentry, mode | S_IFREG, cgrp->root->sb);
	if (!error) {
		cfe->type = (void *)cft;
		cfe->dentry = dentry;
		dentry->d_fsdata = cfe;
		list_add_tail(&cfe->node, &parent->files);
		cfe = NULL;
	}
	dput(dentry);
out:
	kfree(cfe);
2744 2745 2746
	return error;
}

2747 2748
static int cgroup_addrm_files(struct cgroup *cgrp, struct cgroup_subsys *subsys,
			      const struct cftype cfts[], bool is_add)
2749
{
T
Tejun Heo 已提交
2750 2751 2752 2753
	const struct cftype *cft;
	int err, ret = 0;

	for (cft = cfts; cft->name[0] != '\0'; cft++) {
2754 2755 2756 2757
		if (is_add)
			err = cgroup_add_file(cgrp, subsys, cft);
		else
			err = cgroup_rm_file(cgrp, cft);
T
Tejun Heo 已提交
2758
		if (err) {
2759 2760
			pr_warning("cgroup_addrm_files: failed to %s %s, err=%d\n",
				   is_add ? "add" : "remove", cft->name, err);
T
Tejun Heo 已提交
2761 2762
			ret = err;
		}
2763
	}
T
Tejun Heo 已提交
2764
	return ret;
2765 2766
}

2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783
static DEFINE_MUTEX(cgroup_cft_mutex);

static void cgroup_cfts_prepare(void)
	__acquires(&cgroup_cft_mutex) __acquires(&cgroup_mutex)
{
	/*
	 * Thanks to the entanglement with vfs inode locking, we can't walk
	 * the existing cgroups under cgroup_mutex and create files.
	 * Instead, we increment reference on all cgroups and build list of
	 * them using @cgrp->cft_q_node.  Grab cgroup_cft_mutex to ensure
	 * exclusive access to the field.
	 */
	mutex_lock(&cgroup_cft_mutex);
	mutex_lock(&cgroup_mutex);
}

static void cgroup_cfts_commit(struct cgroup_subsys *ss,
2784
			       const struct cftype *cfts, bool is_add)
2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809
	__releases(&cgroup_mutex) __releases(&cgroup_cft_mutex)
{
	LIST_HEAD(pending);
	struct cgroup *cgrp, *n;

	/* %NULL @cfts indicates abort and don't bother if @ss isn't attached */
	if (cfts && ss->root != &rootnode) {
		list_for_each_entry(cgrp, &ss->root->allcg_list, allcg_node) {
			dget(cgrp->dentry);
			list_add_tail(&cgrp->cft_q_node, &pending);
		}
	}

	mutex_unlock(&cgroup_mutex);

	/*
	 * All new cgroups will see @cfts update on @ss->cftsets.  Add/rm
	 * files for all cgroups which were created before.
	 */
	list_for_each_entry_safe(cgrp, n, &pending, cft_q_node) {
		struct inode *inode = cgrp->dentry->d_inode;

		mutex_lock(&inode->i_mutex);
		mutex_lock(&cgroup_mutex);
		if (!cgroup_is_removed(cgrp))
2810
			cgroup_addrm_files(cgrp, ss, cfts, is_add);
2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845
		mutex_unlock(&cgroup_mutex);
		mutex_unlock(&inode->i_mutex);

		list_del_init(&cgrp->cft_q_node);
		dput(cgrp->dentry);
	}

	mutex_unlock(&cgroup_cft_mutex);
}

/**
 * cgroup_add_cftypes - add an array of cftypes to a subsystem
 * @ss: target cgroup subsystem
 * @cfts: zero-length name terminated array of cftypes
 *
 * Register @cfts to @ss.  Files described by @cfts are created for all
 * existing cgroups to which @ss is attached and all future cgroups will
 * have them too.  This function can be called anytime whether @ss is
 * attached or not.
 *
 * Returns 0 on successful registration, -errno on failure.  Note that this
 * function currently returns 0 as long as @cfts registration is successful
 * even if some file creation attempts on existing cgroups fail.
 */
int cgroup_add_cftypes(struct cgroup_subsys *ss, const struct cftype *cfts)
{
	struct cftype_set *set;

	set = kzalloc(sizeof(*set), GFP_KERNEL);
	if (!set)
		return -ENOMEM;

	cgroup_cfts_prepare();
	set->cfts = cfts;
	list_add_tail(&set->node, &ss->cftsets);
2846
	cgroup_cfts_commit(ss, cfts, true);
2847 2848 2849 2850 2851

	return 0;
}
EXPORT_SYMBOL_GPL(cgroup_add_cftypes);

2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882
/**
 * cgroup_rm_cftypes - remove an array of cftypes from a subsystem
 * @ss: target cgroup subsystem
 * @cfts: zero-length name terminated array of cftypes
 *
 * Unregister @cfts from @ss.  Files described by @cfts are removed from
 * all existing cgroups to which @ss is attached and all future cgroups
 * won't have them either.  This function can be called anytime whether @ss
 * is attached or not.
 *
 * Returns 0 on successful unregistration, -ENOENT if @cfts is not
 * registered with @ss.
 */
int cgroup_rm_cftypes(struct cgroup_subsys *ss, const struct cftype *cfts)
{
	struct cftype_set *set;

	cgroup_cfts_prepare();

	list_for_each_entry(set, &ss->cftsets, node) {
		if (set->cfts == cfts) {
			list_del_init(&set->node);
			cgroup_cfts_commit(ss, cfts, false);
			return 0;
		}
	}

	cgroup_cfts_commit(ss, NULL, false);
	return -ENOENT;
}

L
Li Zefan 已提交
2883 2884 2885 2886 2887 2888
/**
 * cgroup_task_count - count the number of tasks in a cgroup.
 * @cgrp: the cgroup in question
 *
 * Return the number of tasks in the cgroup.
 */
2889
int cgroup_task_count(const struct cgroup *cgrp)
2890 2891
{
	int count = 0;
K
KOSAKI Motohiro 已提交
2892
	struct cg_cgroup_link *link;
2893 2894

	read_lock(&css_set_lock);
K
KOSAKI Motohiro 已提交
2895
	list_for_each_entry(link, &cgrp->css_sets, cgrp_link_list) {
2896
		count += atomic_read(&link->cg->refcount);
2897 2898
	}
	read_unlock(&css_set_lock);
2899 2900 2901
	return count;
}

2902 2903 2904 2905
/*
 * Advance a list_head iterator.  The iterator should be positioned at
 * the start of a css_set
 */
2906
static void cgroup_advance_iter(struct cgroup *cgrp,
2907
				struct cgroup_iter *it)
2908 2909 2910 2911 2912 2913 2914 2915
{
	struct list_head *l = it->cg_link;
	struct cg_cgroup_link *link;
	struct css_set *cg;

	/* Advance to the next non-empty css_set */
	do {
		l = l->next;
2916
		if (l == &cgrp->css_sets) {
2917 2918 2919
			it->cg_link = NULL;
			return;
		}
2920
		link = list_entry(l, struct cg_cgroup_link, cgrp_link_list);
2921 2922 2923 2924 2925 2926
		cg = link->cg;
	} while (list_empty(&cg->tasks));
	it->cg_link = l;
	it->task = cg->tasks.next;
}

2927 2928 2929 2930 2931 2932
/*
 * To reduce the fork() overhead for systems that are not actually
 * using their cgroups capability, we don't maintain the lists running
 * through each css_set to its tasks until we see the list actually
 * used - in other words after the first call to cgroup_iter_start().
 */
2933
static void cgroup_enable_task_cg_lists(void)
2934 2935 2936 2937
{
	struct task_struct *p, *g;
	write_lock(&css_set_lock);
	use_task_css_set_links = 1;
2938 2939 2940 2941 2942 2943 2944 2945
	/*
	 * We need tasklist_lock because RCU is not safe against
	 * while_each_thread(). Besides, a forking task that has passed
	 * cgroup_post_fork() without seeing use_task_css_set_links = 1
	 * is not guaranteed to have its child immediately visible in the
	 * tasklist if we walk through it with RCU.
	 */
	read_lock(&tasklist_lock);
2946 2947
	do_each_thread(g, p) {
		task_lock(p);
2948 2949 2950 2951 2952 2953
		/*
		 * We should check if the process is exiting, otherwise
		 * it will race with cgroup_exit() in that the list
		 * entry won't be deleted though the process has exited.
		 */
		if (!(p->flags & PF_EXITING) && list_empty(&p->cg_list))
2954 2955 2956
			list_add(&p->cg_list, &p->cgroups->tasks);
		task_unlock(p);
	} while_each_thread(g, p);
2957
	read_unlock(&tasklist_lock);
2958 2959 2960
	write_unlock(&css_set_lock);
}

2961
void cgroup_iter_start(struct cgroup *cgrp, struct cgroup_iter *it)
2962
	__acquires(css_set_lock)
2963 2964 2965 2966 2967 2968
{
	/*
	 * The first time anyone tries to iterate across a cgroup,
	 * we need to enable the list linking each css_set to its
	 * tasks, and fix up all existing tasks.
	 */
2969 2970 2971
	if (!use_task_css_set_links)
		cgroup_enable_task_cg_lists();

2972
	read_lock(&css_set_lock);
2973 2974
	it->cg_link = &cgrp->css_sets;
	cgroup_advance_iter(cgrp, it);
2975 2976
}

2977
struct task_struct *cgroup_iter_next(struct cgroup *cgrp,
2978 2979 2980 2981
					struct cgroup_iter *it)
{
	struct task_struct *res;
	struct list_head *l = it->task;
2982
	struct cg_cgroup_link *link;
2983 2984 2985 2986 2987 2988 2989

	/* If the iterator cg is NULL, we have no tasks */
	if (!it->cg_link)
		return NULL;
	res = list_entry(l, struct task_struct, cg_list);
	/* Advance iterator to find next entry */
	l = l->next;
2990 2991
	link = list_entry(it->cg_link, struct cg_cgroup_link, cgrp_link_list);
	if (l == &link->cg->tasks) {
2992 2993
		/* We reached the end of this task list - move on to
		 * the next cg_cgroup_link */
2994
		cgroup_advance_iter(cgrp, it);
2995 2996 2997 2998 2999 3000
	} else {
		it->task = l;
	}
	return res;
}

3001
void cgroup_iter_end(struct cgroup *cgrp, struct cgroup_iter *it)
3002
	__releases(css_set_lock)
3003 3004 3005 3006
{
	read_unlock(&css_set_lock);
}

3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143
static inline int started_after_time(struct task_struct *t1,
				     struct timespec *time,
				     struct task_struct *t2)
{
	int start_diff = timespec_compare(&t1->start_time, time);
	if (start_diff > 0) {
		return 1;
	} else if (start_diff < 0) {
		return 0;
	} else {
		/*
		 * Arbitrarily, if two processes started at the same
		 * time, we'll say that the lower pointer value
		 * started first. Note that t2 may have exited by now
		 * so this may not be a valid pointer any longer, but
		 * that's fine - it still serves to distinguish
		 * between two tasks started (effectively) simultaneously.
		 */
		return t1 > t2;
	}
}

/*
 * This function is a callback from heap_insert() and is used to order
 * the heap.
 * In this case we order the heap in descending task start time.
 */
static inline int started_after(void *p1, void *p2)
{
	struct task_struct *t1 = p1;
	struct task_struct *t2 = p2;
	return started_after_time(t1, &t2->start_time, t2);
}

/**
 * cgroup_scan_tasks - iterate though all the tasks in a cgroup
 * @scan: struct cgroup_scanner containing arguments for the scan
 *
 * Arguments include pointers to callback functions test_task() and
 * process_task().
 * Iterate through all the tasks in a cgroup, calling test_task() for each,
 * and if it returns true, call process_task() for it also.
 * The test_task pointer may be NULL, meaning always true (select all tasks).
 * Effectively duplicates cgroup_iter_{start,next,end}()
 * but does not lock css_set_lock for the call to process_task().
 * The struct cgroup_scanner may be embedded in any structure of the caller's
 * creation.
 * It is guaranteed that process_task() will act on every task that
 * is a member of the cgroup for the duration of this call. This
 * function may or may not call process_task() for tasks that exit
 * or move to a different cgroup during the call, or are forked or
 * move into the cgroup during the call.
 *
 * Note that test_task() may be called with locks held, and may in some
 * situations be called multiple times for the same task, so it should
 * be cheap.
 * If the heap pointer in the struct cgroup_scanner is non-NULL, a heap has been
 * pre-allocated and will be used for heap operations (and its "gt" member will
 * be overwritten), else a temporary heap will be used (allocation of which
 * may cause this function to fail).
 */
int cgroup_scan_tasks(struct cgroup_scanner *scan)
{
	int retval, i;
	struct cgroup_iter it;
	struct task_struct *p, *dropped;
	/* Never dereference latest_task, since it's not refcounted */
	struct task_struct *latest_task = NULL;
	struct ptr_heap tmp_heap;
	struct ptr_heap *heap;
	struct timespec latest_time = { 0, 0 };

	if (scan->heap) {
		/* The caller supplied our heap and pre-allocated its memory */
		heap = scan->heap;
		heap->gt = &started_after;
	} else {
		/* We need to allocate our own heap memory */
		heap = &tmp_heap;
		retval = heap_init(heap, PAGE_SIZE, GFP_KERNEL, &started_after);
		if (retval)
			/* cannot allocate the heap */
			return retval;
	}

 again:
	/*
	 * Scan tasks in the cgroup, using the scanner's "test_task" callback
	 * to determine which are of interest, and using the scanner's
	 * "process_task" callback to process any of them that need an update.
	 * Since we don't want to hold any locks during the task updates,
	 * gather tasks to be processed in a heap structure.
	 * The heap is sorted by descending task start time.
	 * If the statically-sized heap fills up, we overflow tasks that
	 * started later, and in future iterations only consider tasks that
	 * started after the latest task in the previous pass. This
	 * guarantees forward progress and that we don't miss any tasks.
	 */
	heap->size = 0;
	cgroup_iter_start(scan->cg, &it);
	while ((p = cgroup_iter_next(scan->cg, &it))) {
		/*
		 * Only affect tasks that qualify per the caller's callback,
		 * if he provided one
		 */
		if (scan->test_task && !scan->test_task(p, scan))
			continue;
		/*
		 * Only process tasks that started after the last task
		 * we processed
		 */
		if (!started_after_time(p, &latest_time, latest_task))
			continue;
		dropped = heap_insert(heap, p);
		if (dropped == NULL) {
			/*
			 * The new task was inserted; the heap wasn't
			 * previously full
			 */
			get_task_struct(p);
		} else if (dropped != p) {
			/*
			 * The new task was inserted, and pushed out a
			 * different task
			 */
			get_task_struct(p);
			put_task_struct(dropped);
		}
		/*
		 * Else the new task was newer than anything already in
		 * the heap and wasn't inserted
		 */
	}
	cgroup_iter_end(scan->cg, &it);

	if (heap->size) {
		for (i = 0; i < heap->size; i++) {
3144
			struct task_struct *q = heap->ptrs[i];
3145
			if (i == 0) {
3146 3147
				latest_time = q->start_time;
				latest_task = q;
3148 3149
			}
			/* Process the task per the caller's callback */
3150 3151
			scan->process_task(q, scan);
			put_task_struct(q);
3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166
		}
		/*
		 * If we had to process any tasks at all, scan again
		 * in case some of them were in the middle of forking
		 * children that didn't get processed.
		 * Not the most efficient way to do it, but it avoids
		 * having to take callback_mutex in the fork path
		 */
		goto again;
	}
	if (heap == &tmp_heap)
		heap_free(&tmp_heap);
	return 0;
}

3167
/*
3168
 * Stuff for reading the 'tasks'/'procs' files.
3169 3170 3171 3172 3173 3174 3175 3176
 *
 * Reading this file can return large amounts of data if a cgroup has
 * *lots* of attached tasks. So it may need several calls to read(),
 * but we cannot guarantee that the information we produce is correct
 * unless we produce it entirely atomically.
 *
 */

3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208
/* which pidlist file are we talking about? */
enum cgroup_filetype {
	CGROUP_FILE_PROCS,
	CGROUP_FILE_TASKS,
};

/*
 * A pidlist is a list of pids that virtually represents the contents of one
 * of the cgroup files ("procs" or "tasks"). We keep a list of such pidlists,
 * a pair (one each for procs, tasks) for each pid namespace that's relevant
 * to the cgroup.
 */
struct cgroup_pidlist {
	/*
	 * used to find which pidlist is wanted. doesn't change as long as
	 * this particular list stays in the list.
	*/
	struct { enum cgroup_filetype type; struct pid_namespace *ns; } key;
	/* array of xids */
	pid_t *list;
	/* how many elements the above list has */
	int length;
	/* how many files are using the current array */
	int use_count;
	/* each of these stored in a list by its cgroup */
	struct list_head links;
	/* pointer to the cgroup we belong to, for list removal purposes */
	struct cgroup *owner;
	/* protects the other fields */
	struct rw_semaphore mutex;
};

3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244
/*
 * The following two functions "fix" the issue where there are more pids
 * than kmalloc will give memory for; in such cases, we use vmalloc/vfree.
 * TODO: replace with a kernel-wide solution to this problem
 */
#define PIDLIST_TOO_LARGE(c) ((c) * sizeof(pid_t) > (PAGE_SIZE * 2))
static void *pidlist_allocate(int count)
{
	if (PIDLIST_TOO_LARGE(count))
		return vmalloc(count * sizeof(pid_t));
	else
		return kmalloc(count * sizeof(pid_t), GFP_KERNEL);
}
static void pidlist_free(void *p)
{
	if (is_vmalloc_addr(p))
		vfree(p);
	else
		kfree(p);
}
static void *pidlist_resize(void *p, int newcount)
{
	void *newlist;
	/* note: if new alloc fails, old p will still be valid either way */
	if (is_vmalloc_addr(p)) {
		newlist = vmalloc(newcount * sizeof(pid_t));
		if (!newlist)
			return NULL;
		memcpy(newlist, p, newcount * sizeof(pid_t));
		vfree(p);
	} else {
		newlist = krealloc(p, newcount * sizeof(pid_t), GFP_KERNEL);
	}
	return newlist;
}

3245
/*
3246 3247 3248 3249
 * pidlist_uniq - given a kmalloc()ed list, strip out all duplicate entries
 * If the new stripped list is sufficiently smaller and there's enough memory
 * to allocate a new buffer, will let go of the unneeded memory. Returns the
 * number of unique elements.
3250
 */
3251 3252 3253
/* is the size difference enough that we should re-allocate the array? */
#define PIDLIST_REALLOC_DIFFERENCE(old, new) ((old) - PAGE_SIZE >= (new))
static int pidlist_uniq(pid_t **p, int length)
3254
{
3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283
	int src, dest = 1;
	pid_t *list = *p;
	pid_t *newlist;

	/*
	 * we presume the 0th element is unique, so i starts at 1. trivial
	 * edge cases first; no work needs to be done for either
	 */
	if (length == 0 || length == 1)
		return length;
	/* src and dest walk down the list; dest counts unique elements */
	for (src = 1; src < length; src++) {
		/* find next unique element */
		while (list[src] == list[src-1]) {
			src++;
			if (src == length)
				goto after;
		}
		/* dest always points to where the next unique element goes */
		list[dest] = list[src];
		dest++;
	}
after:
	/*
	 * if the length difference is large enough, we want to allocate a
	 * smaller buffer to save memory. if this fails due to out of memory,
	 * we'll just stay with what we've got.
	 */
	if (PIDLIST_REALLOC_DIFFERENCE(length, dest)) {
3284
		newlist = pidlist_resize(list, dest);
3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295
		if (newlist)
			*p = newlist;
	}
	return dest;
}

static int cmppid(const void *a, const void *b)
{
	return *(pid_t *)a - *(pid_t *)b;
}

3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306
/*
 * find the appropriate pidlist for our purpose (given procs vs tasks)
 * returns with the lock on that pidlist already held, and takes care
 * of the use count, or returns NULL with no locks held if we're out of
 * memory.
 */
static struct cgroup_pidlist *cgroup_pidlist_find(struct cgroup *cgrp,
						  enum cgroup_filetype type)
{
	struct cgroup_pidlist *l;
	/* don't need task_nsproxy() if we're looking at ourself */
3307 3308
	struct pid_namespace *ns = current->nsproxy->pid_ns;

3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332
	/*
	 * We can't drop the pidlist_mutex before taking the l->mutex in case
	 * the last ref-holder is trying to remove l from the list at the same
	 * time. Holding the pidlist_mutex precludes somebody taking whichever
	 * list we find out from under us - compare release_pid_array().
	 */
	mutex_lock(&cgrp->pidlist_mutex);
	list_for_each_entry(l, &cgrp->pidlists, links) {
		if (l->key.type == type && l->key.ns == ns) {
			/* make sure l doesn't vanish out from under us */
			down_write(&l->mutex);
			mutex_unlock(&cgrp->pidlist_mutex);
			return l;
		}
	}
	/* entry not found; create a new one */
	l = kmalloc(sizeof(struct cgroup_pidlist), GFP_KERNEL);
	if (!l) {
		mutex_unlock(&cgrp->pidlist_mutex);
		return l;
	}
	init_rwsem(&l->mutex);
	down_write(&l->mutex);
	l->key.type = type;
3333
	l->key.ns = get_pid_ns(ns);
3334 3335 3336 3337 3338 3339 3340 3341
	l->use_count = 0; /* don't increment here */
	l->list = NULL;
	l->owner = cgrp;
	list_add(&l->links, &cgrp->pidlists);
	mutex_unlock(&cgrp->pidlist_mutex);
	return l;
}

3342 3343 3344
/*
 * Load a cgroup's pidarray with either procs' tgids or tasks' pids
 */
3345 3346
static int pidlist_array_load(struct cgroup *cgrp, enum cgroup_filetype type,
			      struct cgroup_pidlist **lp)
3347 3348 3349 3350
{
	pid_t *array;
	int length;
	int pid, n = 0; /* used for populating the array */
3351 3352
	struct cgroup_iter it;
	struct task_struct *tsk;
3353 3354 3355 3356 3357 3358 3359 3360 3361
	struct cgroup_pidlist *l;

	/*
	 * If cgroup gets more users after we read count, we won't have
	 * enough space - tough.  This race is indistinguishable to the
	 * caller from the case that the additional cgroup users didn't
	 * show up until sometime later on.
	 */
	length = cgroup_task_count(cgrp);
3362
	array = pidlist_allocate(length);
3363 3364 3365
	if (!array)
		return -ENOMEM;
	/* now, populate the array */
3366 3367
	cgroup_iter_start(cgrp, &it);
	while ((tsk = cgroup_iter_next(cgrp, &it))) {
3368
		if (unlikely(n == length))
3369
			break;
3370
		/* get tgid or pid for procs or tasks file respectively */
3371 3372 3373 3374
		if (type == CGROUP_FILE_PROCS)
			pid = task_tgid_vnr(tsk);
		else
			pid = task_pid_vnr(tsk);
3375 3376
		if (pid > 0) /* make sure to only use valid results */
			array[n++] = pid;
3377
	}
3378
	cgroup_iter_end(cgrp, &it);
3379 3380 3381
	length = n;
	/* now sort & (if procs) strip out duplicates */
	sort(array, length, sizeof(pid_t), cmppid, NULL);
3382
	if (type == CGROUP_FILE_PROCS)
3383
		length = pidlist_uniq(&array, length);
3384 3385
	l = cgroup_pidlist_find(cgrp, type);
	if (!l) {
3386
		pidlist_free(array);
3387
		return -ENOMEM;
3388
	}
3389
	/* store array, freeing old if necessary - lock already held */
3390
	pidlist_free(l->list);
3391 3392 3393 3394
	l->list = array;
	l->length = length;
	l->use_count++;
	up_write(&l->mutex);
3395
	*lp = l;
3396
	return 0;
3397 3398
}

B
Balbir Singh 已提交
3399
/**
L
Li Zefan 已提交
3400
 * cgroupstats_build - build and fill cgroupstats
B
Balbir Singh 已提交
3401 3402 3403
 * @stats: cgroupstats to fill information into
 * @dentry: A dentry entry belonging to the cgroup for which stats have
 * been requested.
L
Li Zefan 已提交
3404 3405 3406
 *
 * Build and fill cgroupstats so that taskstats can export it to user
 * space.
B
Balbir Singh 已提交
3407 3408 3409 3410
 */
int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry)
{
	int ret = -EINVAL;
3411
	struct cgroup *cgrp;
B
Balbir Singh 已提交
3412 3413
	struct cgroup_iter it;
	struct task_struct *tsk;
3414

B
Balbir Singh 已提交
3415
	/*
3416 3417
	 * Validate dentry by checking the superblock operations,
	 * and make sure it's a directory.
B
Balbir Singh 已提交
3418
	 */
3419 3420
	if (dentry->d_sb->s_op != &cgroup_ops ||
	    !S_ISDIR(dentry->d_inode->i_mode))
B
Balbir Singh 已提交
3421 3422 3423
		 goto err;

	ret = 0;
3424
	cgrp = dentry->d_fsdata;
B
Balbir Singh 已提交
3425

3426 3427
	cgroup_iter_start(cgrp, &it);
	while ((tsk = cgroup_iter_next(cgrp, &it))) {
B
Balbir Singh 已提交
3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446
		switch (tsk->state) {
		case TASK_RUNNING:
			stats->nr_running++;
			break;
		case TASK_INTERRUPTIBLE:
			stats->nr_sleeping++;
			break;
		case TASK_UNINTERRUPTIBLE:
			stats->nr_uninterruptible++;
			break;
		case TASK_STOPPED:
			stats->nr_stopped++;
			break;
		default:
			if (delayacct_is_task_waiting_on_io(tsk))
				stats->nr_io_wait++;
			break;
		}
	}
3447
	cgroup_iter_end(cgrp, &it);
B
Balbir Singh 已提交
3448 3449 3450 3451 3452

err:
	return ret;
}

3453

3454
/*
3455
 * seq_file methods for the tasks/procs files. The seq_file position is the
3456
 * next pid to display; the seq_file iterator is a pointer to the pid
3457
 * in the cgroup->l->list array.
3458
 */
3459

3460
static void *cgroup_pidlist_start(struct seq_file *s, loff_t *pos)
3461
{
3462 3463 3464 3465 3466 3467
	/*
	 * Initially we receive a position value that corresponds to
	 * one more than the last pid shown (or 0 on the first call or
	 * after a seek to the start). Use a binary-search to find the
	 * next pid to display, if any
	 */
3468
	struct cgroup_pidlist *l = s->private;
3469 3470 3471
	int index = 0, pid = *pos;
	int *iter;

3472
	down_read(&l->mutex);
3473
	if (pid) {
3474
		int end = l->length;
S
Stephen Rothwell 已提交
3475

3476 3477
		while (index < end) {
			int mid = (index + end) / 2;
3478
			if (l->list[mid] == pid) {
3479 3480
				index = mid;
				break;
3481
			} else if (l->list[mid] <= pid)
3482 3483 3484 3485 3486 3487
				index = mid + 1;
			else
				end = mid;
		}
	}
	/* If we're off the end of the array, we're done */
3488
	if (index >= l->length)
3489 3490
		return NULL;
	/* Update the abstract position to be the actual pid that we found */
3491
	iter = l->list + index;
3492 3493 3494 3495
	*pos = *iter;
	return iter;
}

3496
static void cgroup_pidlist_stop(struct seq_file *s, void *v)
3497
{
3498 3499
	struct cgroup_pidlist *l = s->private;
	up_read(&l->mutex);
3500 3501
}

3502
static void *cgroup_pidlist_next(struct seq_file *s, void *v, loff_t *pos)
3503
{
3504 3505 3506
	struct cgroup_pidlist *l = s->private;
	pid_t *p = v;
	pid_t *end = l->list + l->length;
3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519
	/*
	 * Advance to the next pid in the array. If this goes off the
	 * end, we're done
	 */
	p++;
	if (p >= end) {
		return NULL;
	} else {
		*pos = *p;
		return p;
	}
}

3520
static int cgroup_pidlist_show(struct seq_file *s, void *v)
3521 3522 3523
{
	return seq_printf(s, "%d\n", *(int *)v);
}
3524

3525 3526 3527 3528 3529 3530 3531 3532 3533
/*
 * seq_operations functions for iterating on pidlists through seq_file -
 * independent of whether it's tasks or procs
 */
static const struct seq_operations cgroup_pidlist_seq_operations = {
	.start = cgroup_pidlist_start,
	.stop = cgroup_pidlist_stop,
	.next = cgroup_pidlist_next,
	.show = cgroup_pidlist_show,
3534 3535
};

3536
static void cgroup_release_pid_array(struct cgroup_pidlist *l)
3537
{
3538 3539 3540 3541 3542 3543 3544
	/*
	 * the case where we're the last user of this particular pidlist will
	 * have us remove it from the cgroup's list, which entails taking the
	 * mutex. since in pidlist_find the pidlist->lock depends on cgroup->
	 * pidlist_mutex, we have to take pidlist_mutex first.
	 */
	mutex_lock(&l->owner->pidlist_mutex);
3545 3546 3547
	down_write(&l->mutex);
	BUG_ON(!l->use_count);
	if (!--l->use_count) {
3548 3549 3550
		/* we're the last user if refcount is 0; remove and free */
		list_del(&l->links);
		mutex_unlock(&l->owner->pidlist_mutex);
3551
		pidlist_free(l->list);
3552 3553 3554 3555
		put_pid_ns(l->key.ns);
		up_write(&l->mutex);
		kfree(l);
		return;
3556
	}
3557
	mutex_unlock(&l->owner->pidlist_mutex);
3558
	up_write(&l->mutex);
3559 3560
}

3561
static int cgroup_pidlist_release(struct inode *inode, struct file *file)
3562
{
3563
	struct cgroup_pidlist *l;
3564 3565
	if (!(file->f_mode & FMODE_READ))
		return 0;
3566 3567 3568 3569 3570 3571
	/*
	 * the seq_file will only be initialized if the file was opened for
	 * reading; hence we check if it's not null only in that case.
	 */
	l = ((struct seq_file *)file->private_data)->private;
	cgroup_release_pid_array(l);
3572 3573 3574
	return seq_release(inode, file);
}

3575
static const struct file_operations cgroup_pidlist_operations = {
3576 3577 3578
	.read = seq_read,
	.llseek = seq_lseek,
	.write = cgroup_file_write,
3579
	.release = cgroup_pidlist_release,
3580 3581
};

3582
/*
3583 3584 3585
 * The following functions handle opens on a file that displays a pidlist
 * (tasks or procs). Prepare an array of the process/thread IDs of whoever's
 * in the cgroup.
3586
 */
3587
/* helper function for the two below it */
3588
static int cgroup_pidlist_open(struct file *file, enum cgroup_filetype type)
3589
{
3590
	struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
3591
	struct cgroup_pidlist *l;
3592
	int retval;
3593

3594
	/* Nothing to do for write-only files */
3595 3596 3597
	if (!(file->f_mode & FMODE_READ))
		return 0;

3598
	/* have the array populated */
3599
	retval = pidlist_array_load(cgrp, type, &l);
3600 3601 3602 3603
	if (retval)
		return retval;
	/* configure file information */
	file->f_op = &cgroup_pidlist_operations;
3604

3605
	retval = seq_open(file, &cgroup_pidlist_seq_operations);
3606
	if (retval) {
3607
		cgroup_release_pid_array(l);
3608
		return retval;
3609
	}
3610
	((struct seq_file *)file->private_data)->private = l;
3611 3612
	return 0;
}
3613 3614
static int cgroup_tasks_open(struct inode *unused, struct file *file)
{
3615
	return cgroup_pidlist_open(file, CGROUP_FILE_TASKS);
3616 3617 3618
}
static int cgroup_procs_open(struct inode *unused, struct file *file)
{
3619
	return cgroup_pidlist_open(file, CGROUP_FILE_PROCS);
3620
}
3621

3622
static u64 cgroup_read_notify_on_release(struct cgroup *cgrp,
3623 3624
					    struct cftype *cft)
{
3625
	return notify_on_release(cgrp);
3626 3627
}

3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639
static int cgroup_write_notify_on_release(struct cgroup *cgrp,
					  struct cftype *cft,
					  u64 val)
{
	clear_bit(CGRP_RELEASABLE, &cgrp->flags);
	if (val)
		set_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
	else
		clear_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
	return 0;
}

3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654
/*
 * Unregister event and free resources.
 *
 * Gets called from workqueue.
 */
static void cgroup_event_remove(struct work_struct *work)
{
	struct cgroup_event *event = container_of(work, struct cgroup_event,
			remove);
	struct cgroup *cgrp = event->cgrp;

	event->cft->unregister_event(cgrp, event->cft, event->eventfd);

	eventfd_ctx_put(event->eventfd);
	kfree(event);
3655
	dput(cgrp->dentry);
3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671
}

/*
 * Gets called on POLLHUP on eventfd when user closes it.
 *
 * Called with wqh->lock held and interrupts disabled.
 */
static int cgroup_event_wake(wait_queue_t *wait, unsigned mode,
		int sync, void *key)
{
	struct cgroup_event *event = container_of(wait,
			struct cgroup_event, wait);
	struct cgroup *cgrp = event->cgrp;
	unsigned long flags = (unsigned long)key;

	if (flags & POLLHUP) {
C
Changli Gao 已提交
3672
		__remove_wait_queue(event->wqh, &event->wait);
3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749
		spin_lock(&cgrp->event_list_lock);
		list_del(&event->list);
		spin_unlock(&cgrp->event_list_lock);
		/*
		 * We are in atomic context, but cgroup_event_remove() may
		 * sleep, so we have to call it in workqueue.
		 */
		schedule_work(&event->remove);
	}

	return 0;
}

static void cgroup_event_ptable_queue_proc(struct file *file,
		wait_queue_head_t *wqh, poll_table *pt)
{
	struct cgroup_event *event = container_of(pt,
			struct cgroup_event, pt);

	event->wqh = wqh;
	add_wait_queue(wqh, &event->wait);
}

/*
 * Parse input and register new cgroup event handler.
 *
 * Input must be in format '<event_fd> <control_fd> <args>'.
 * Interpretation of args is defined by control file implementation.
 */
static int cgroup_write_event_control(struct cgroup *cgrp, struct cftype *cft,
				      const char *buffer)
{
	struct cgroup_event *event = NULL;
	unsigned int efd, cfd;
	struct file *efile = NULL;
	struct file *cfile = NULL;
	char *endp;
	int ret;

	efd = simple_strtoul(buffer, &endp, 10);
	if (*endp != ' ')
		return -EINVAL;
	buffer = endp + 1;

	cfd = simple_strtoul(buffer, &endp, 10);
	if ((*endp != ' ') && (*endp != '\0'))
		return -EINVAL;
	buffer = endp + 1;

	event = kzalloc(sizeof(*event), GFP_KERNEL);
	if (!event)
		return -ENOMEM;
	event->cgrp = cgrp;
	INIT_LIST_HEAD(&event->list);
	init_poll_funcptr(&event->pt, cgroup_event_ptable_queue_proc);
	init_waitqueue_func_entry(&event->wait, cgroup_event_wake);
	INIT_WORK(&event->remove, cgroup_event_remove);

	efile = eventfd_fget(efd);
	if (IS_ERR(efile)) {
		ret = PTR_ERR(efile);
		goto fail;
	}

	event->eventfd = eventfd_ctx_fileget(efile);
	if (IS_ERR(event->eventfd)) {
		ret = PTR_ERR(event->eventfd);
		goto fail;
	}

	cfile = fget(cfd);
	if (!cfile) {
		ret = -EBADF;
		goto fail;
	}

	/* the process need read permission on control file */
A
Al Viro 已提交
3750 3751
	/* AV: shouldn't we check that it's been opened for read instead? */
	ret = inode_permission(cfile->f_path.dentry->d_inode, MAY_READ);
3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776
	if (ret < 0)
		goto fail;

	event->cft = __file_cft(cfile);
	if (IS_ERR(event->cft)) {
		ret = PTR_ERR(event->cft);
		goto fail;
	}

	if (!event->cft->register_event || !event->cft->unregister_event) {
		ret = -EINVAL;
		goto fail;
	}

	ret = event->cft->register_event(cgrp, event->cft,
			event->eventfd, buffer);
	if (ret)
		goto fail;

	if (efile->f_op->poll(efile, &event->pt) & POLLHUP) {
		event->cft->unregister_event(cgrp, event->cft, event->eventfd);
		ret = 0;
		goto fail;
	}

3777 3778 3779 3780 3781 3782 3783
	/*
	 * Events should be removed after rmdir of cgroup directory, but before
	 * destroying subsystem state objects. Let's take reference to cgroup
	 * directory dentry to do that.
	 */
	dget(cgrp->dentry);

3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807
	spin_lock(&cgrp->event_list_lock);
	list_add(&event->list, &cgrp->event_list);
	spin_unlock(&cgrp->event_list_lock);

	fput(cfile);
	fput(efile);

	return 0;

fail:
	if (cfile)
		fput(cfile);

	if (event && event->eventfd && !IS_ERR(event->eventfd))
		eventfd_ctx_put(event->eventfd);

	if (!IS_ERR_OR_NULL(efile))
		fput(efile);

	kfree(event);

	return ret;
}

3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824
static u64 cgroup_clone_children_read(struct cgroup *cgrp,
				    struct cftype *cft)
{
	return clone_children(cgrp);
}

static int cgroup_clone_children_write(struct cgroup *cgrp,
				     struct cftype *cft,
				     u64 val)
{
	if (val)
		set_bit(CGRP_CLONE_CHILDREN, &cgrp->flags);
	else
		clear_bit(CGRP_CLONE_CHILDREN, &cgrp->flags);
	return 0;
}

3825 3826 3827
/*
 * for the common functions, 'private' gives the type of file
 */
3828 3829
/* for hysterical raisins, we can't put this on the older files */
#define CGROUP_FILE_GENERIC_PREFIX "cgroup."
3830 3831 3832 3833
static struct cftype files[] = {
	{
		.name = "tasks",
		.open = cgroup_tasks_open,
3834
		.write_u64 = cgroup_tasks_write,
3835
		.release = cgroup_pidlist_release,
L
Li Zefan 已提交
3836
		.mode = S_IRUGO | S_IWUSR,
3837
	},
3838 3839 3840
	{
		.name = CGROUP_FILE_GENERIC_PREFIX "procs",
		.open = cgroup_procs_open,
B
Ben Blum 已提交
3841
		.write_u64 = cgroup_procs_write,
3842
		.release = cgroup_pidlist_release,
B
Ben Blum 已提交
3843
		.mode = S_IRUGO | S_IWUSR,
3844
	},
3845 3846
	{
		.name = "notify_on_release",
3847
		.read_u64 = cgroup_read_notify_on_release,
3848
		.write_u64 = cgroup_write_notify_on_release,
3849
	},
3850 3851 3852 3853 3854
	{
		.name = CGROUP_FILE_GENERIC_PREFIX "event_control",
		.write_string = cgroup_write_event_control,
		.mode = S_IWUGO,
	},
3855 3856 3857 3858 3859
	{
		.name = "cgroup.clone_children",
		.read_u64 = cgroup_clone_children_read,
		.write_u64 = cgroup_clone_children_write,
	},
3860 3861 3862 3863 3864 3865 3866
	{
		.name = "release_agent",
		.flags = CFTYPE_ONLY_ON_ROOT,
		.read_seq_string = cgroup_release_agent_show,
		.write_string = cgroup_release_agent_write,
		.max_write_len = PATH_MAX,
	},
T
Tejun Heo 已提交
3867
	{ }	/* terminate */
3868 3869
};

3870 3871 3872 3873 3874 3875 3876 3877
/**
 * cgroup_populate_dir - selectively creation of files in a directory
 * @cgrp: target cgroup
 * @base_files: true if the base files should be added
 * @subsys_mask: mask of the subsystem ids whose files should be added
 */
static int cgroup_populate_dir(struct cgroup *cgrp, bool base_files,
			       unsigned long subsys_mask)
3878 3879 3880 3881
{
	int err;
	struct cgroup_subsys *ss;

3882 3883 3884 3885 3886
	if (base_files) {
		err = cgroup_addrm_files(cgrp, NULL, files, true);
		if (err < 0)
			return err;
	}
3887

3888
	/* process cftsets of each subsystem */
3889
	for_each_subsys(cgrp->root, ss) {
3890
		struct cftype_set *set;
3891 3892
		if (!test_bit(ss->subsys_id, &subsys_mask))
			continue;
3893

T
Tejun Heo 已提交
3894
		list_for_each_entry(set, &ss->cftsets, node)
3895
			cgroup_addrm_files(cgrp, ss, set->cfts, true);
3896
	}
3897

K
KAMEZAWA Hiroyuki 已提交
3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908
	/* This cgroup is ready now */
	for_each_subsys(cgrp->root, ss) {
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];
		/*
		 * Update id->css pointer and make this css visible from
		 * CSS ID functions. This pointer will be dereferened
		 * from RCU-read-side without locks.
		 */
		if (css->id)
			rcu_assign_pointer(css->id->css, css);
	}
3909 3910 3911 3912

	return 0;
}

3913 3914 3915 3916
static void css_dput_fn(struct work_struct *work)
{
	struct cgroup_subsys_state *css =
		container_of(work, struct cgroup_subsys_state, dput_work);
3917 3918
	struct dentry *dentry = css->cgroup->dentry;
	struct super_block *sb = dentry->d_sb;
3919

3920 3921 3922
	atomic_inc(&sb->s_active);
	dput(dentry);
	deactivate_super(sb);
3923 3924
}

3925 3926
static void init_cgroup_css(struct cgroup_subsys_state *css,
			       struct cgroup_subsys *ss,
3927
			       struct cgroup *cgrp)
3928
{
3929
	css->cgroup = cgrp;
P
Paul Menage 已提交
3930
	atomic_set(&css->refcnt, 1);
3931
	css->flags = 0;
K
KAMEZAWA Hiroyuki 已提交
3932
	css->id = NULL;
3933
	if (cgrp == dummytop)
3934
		set_bit(CSS_ROOT, &css->flags);
3935 3936
	BUG_ON(cgrp->subsys[ss->subsys_id]);
	cgrp->subsys[ss->subsys_id] = css;
3937 3938 3939 3940 3941 3942 3943 3944 3945 3946

	/*
	 * If !clear_css_refs, css holds an extra ref to @cgrp->dentry
	 * which is put on the last css_put().  dput() requires process
	 * context, which css_put() may be called without.  @css->dput_work
	 * will be used to invoke dput() asynchronously from css_put().
	 */
	INIT_WORK(&css->dput_work, css_dput_fn);
	if (ss->__DEPRECATED_clear_css_refs)
		set_bit(CSS_CLEAR_CSS_REFS, &css->flags);
3947 3948 3949
}

/*
L
Li Zefan 已提交
3950 3951 3952 3953
 * cgroup_create - create a cgroup
 * @parent: cgroup that will be parent of the new cgroup
 * @dentry: dentry of the new cgroup
 * @mode: mode to set on new inode
3954
 *
L
Li Zefan 已提交
3955
 * Must be called with the mutex on the parent inode held
3956 3957
 */
static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
A
Al Viro 已提交
3958
			     umode_t mode)
3959
{
3960
	struct cgroup *cgrp;
3961 3962 3963 3964 3965
	struct cgroupfs_root *root = parent->root;
	int err = 0;
	struct cgroup_subsys *ss;
	struct super_block *sb = root->sb;

3966 3967
	cgrp = kzalloc(sizeof(*cgrp), GFP_KERNEL);
	if (!cgrp)
3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978
		return -ENOMEM;

	/* Grab a reference on the superblock so the hierarchy doesn't
	 * get deleted on unmount if there are child cgroups.  This
	 * can be done outside cgroup_mutex, since the sb can't
	 * disappear while someone has an open control file on the
	 * fs */
	atomic_inc(&sb->s_active);

	mutex_lock(&cgroup_mutex);

3979
	init_cgroup_housekeeping(cgrp);
3980

3981 3982 3983
	cgrp->parent = parent;
	cgrp->root = parent->root;
	cgrp->top_cgroup = parent->top_cgroup;
3984

3985 3986 3987
	if (notify_on_release(parent))
		set_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);

3988 3989 3990
	if (clone_children(parent))
		set_bit(CGRP_CLONE_CHILDREN, &cgrp->flags);

3991
	for_each_subsys(root, ss) {
3992
		struct cgroup_subsys_state *css = ss->create(cgrp);
3993

3994 3995 3996 3997
		if (IS_ERR(css)) {
			err = PTR_ERR(css);
			goto err_destroy;
		}
3998
		init_cgroup_css(css, ss, cgrp);
3999 4000 4001
		if (ss->use_id) {
			err = alloc_css_id(ss, parent, cgrp);
			if (err)
K
KAMEZAWA Hiroyuki 已提交
4002
				goto err_destroy;
4003
		}
K
KAMEZAWA Hiroyuki 已提交
4004
		/* At error, ->destroy() callback has to free assigned ID. */
4005
		if (clone_children(parent) && ss->post_clone)
4006
			ss->post_clone(cgrp);
4007 4008
	}

4009
	list_add(&cgrp->sibling, &cgrp->parent->children);
4010 4011
	root->number_of_cgroups++;

4012
	err = cgroup_create_dir(cgrp, dentry, mode);
4013 4014 4015
	if (err < 0)
		goto err_remove;

4016 4017 4018 4019 4020
	/* If !clear_css_refs, each css holds a ref to the cgroup's dentry */
	for_each_subsys(root, ss)
		if (!ss->__DEPRECATED_clear_css_refs)
			dget(dentry);

4021
	/* The cgroup directory was pre-locked for us */
4022
	BUG_ON(!mutex_is_locked(&cgrp->dentry->d_inode->i_mutex));
4023

4024 4025
	list_add_tail(&cgrp->allcg_node, &root->allcg_list);

4026
	err = cgroup_populate_dir(cgrp, true, root->subsys_bits);
4027 4028 4029
	/* If err < 0, we have a half-filled directory - oh well ;) */

	mutex_unlock(&cgroup_mutex);
4030
	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
4031 4032 4033 4034 4035

	return 0;

 err_remove:

4036
	list_del(&cgrp->sibling);
4037 4038 4039 4040 4041
	root->number_of_cgroups--;

 err_destroy:

	for_each_subsys(root, ss) {
4042
		if (cgrp->subsys[ss->subsys_id])
4043
			ss->destroy(cgrp);
4044 4045 4046 4047 4048 4049 4050
	}

	mutex_unlock(&cgroup_mutex);

	/* Release the reference count that we took on the superblock */
	deactivate_super(sb);

4051
	kfree(cgrp);
4052 4053 4054
	return err;
}

4055
static int cgroup_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
4056 4057 4058 4059 4060 4061 4062
{
	struct cgroup *c_parent = dentry->d_parent->d_fsdata;

	/* the vfs holds inode->i_mutex already */
	return cgroup_create(c_parent, dentry, mode | S_IFDIR);
}

4063 4064 4065 4066 4067 4068 4069 4070 4071
/*
 * Check the reference count on each subsystem. Since we already
 * established that there are no tasks in the cgroup, if the css refcount
 * is also 1, then there should be no outstanding references, so the
 * subsystem is safe to destroy. We scan across all subsystems rather than
 * using the per-hierarchy linked list of mounted subsystems since we can
 * be called via check_for_release() with no synchronization other than
 * RCU, and the subsystem linked list isn't RCU-safe.
 */
4072
static int cgroup_has_css_refs(struct cgroup *cgrp)
4073 4074
{
	int i;
4075

B
Ben Blum 已提交
4076 4077 4078 4079 4080
	/*
	 * We won't need to lock the subsys array, because the subsystems
	 * we're concerned about aren't going anywhere since our cgroup root
	 * has a reference on them.
	 */
4081 4082 4083
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
		struct cgroup_subsys *ss = subsys[i];
		struct cgroup_subsys_state *css;
4084

B
Ben Blum 已提交
4085 4086
		/* Skip subsystems not present or not in this hierarchy */
		if (ss == NULL || ss->root != cgrp->root)
4087
			continue;
4088

4089
		css = cgrp->subsys[ss->subsys_id];
4090 4091
		/*
		 * When called from check_for_release() it's possible
4092 4093 4094 4095
		 * that by this point the cgroup has been removed
		 * and the css deleted. But a false-positive doesn't
		 * matter, since it can only happen if the cgroup
		 * has been deleted and hence no longer needs the
4096 4097 4098
		 * release agent to be called anyway.
		 */
		if (css && css_refcnt(css) > 1)
4099 4100 4101 4102 4103
			return 1;
	}
	return 0;
}

P
Paul Menage 已提交
4104 4105 4106 4107
/*
 * Atomically mark all (or else none) of the cgroup's CSS objects as
 * CSS_REMOVED. Return true on success, or false if the cgroup has
 * busy subsystems. Call with cgroup_mutex held
4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124
 *
 * Depending on whether a subsys has __DEPRECATED_clear_css_refs set or
 * not, cgroup removal behaves differently.
 *
 * If clear is set, css refcnt for the subsystem should be zero before
 * cgroup removal can be committed.  This is implemented by
 * CGRP_WAIT_ON_RMDIR and retry logic around ->pre_destroy(), which may be
 * called multiple times until all css refcnts reach zero and is allowed to
 * veto removal on any invocation.  This behavior is deprecated and will be
 * removed as soon as the existing user (memcg) is updated.
 *
 * If clear is not set, each css holds an extra reference to the cgroup's
 * dentry and cgroup removal proceeds regardless of css refs.
 * ->pre_destroy() will be called at least once and is not allowed to fail.
 * On the last put of each css, whenever that may be, the extra dentry ref
 * is put so that dentry destruction happens only after all css's are
 * released.
P
Paul Menage 已提交
4125 4126 4127 4128 4129 4130
 */
static int cgroup_clear_css_refs(struct cgroup *cgrp)
{
	struct cgroup_subsys *ss;
	unsigned long flags;
	bool failed = false;
4131

P
Paul Menage 已提交
4132
	local_irq_save(flags);
4133 4134 4135

	/*
	 * Block new css_tryget() by deactivating refcnt.  If all refcnts
4136 4137
	 * for subsystems w/ clear_css_refs set were 1 at the moment of
	 * deactivation, we succeeded.
4138
	 */
P
Paul Menage 已提交
4139 4140
	for_each_subsys(cgrp->root, ss) {
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];
4141 4142 4143

		WARN_ON(atomic_read(&css->refcnt) < 0);
		atomic_add(CSS_DEACT_BIAS, &css->refcnt);
4144 4145 4146

		if (ss->__DEPRECATED_clear_css_refs)
			failed |= css_refcnt(css) != 1;
P
Paul Menage 已提交
4147
	}
4148 4149 4150 4151 4152 4153

	/*
	 * If succeeded, set REMOVED and put all the base refs; otherwise,
	 * restore refcnts to positive values.  Either way, all in-progress
	 * css_tryget() will be released.
	 */
P
Paul Menage 已提交
4154 4155
	for_each_subsys(cgrp->root, ss) {
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];
4156 4157

		if (!failed) {
P
Paul Menage 已提交
4158
			set_bit(CSS_REMOVED, &css->flags);
4159 4160 4161
			css_put(css);
		} else {
			atomic_sub(CSS_DEACT_BIAS, &css->refcnt);
P
Paul Menage 已提交
4162 4163
		}
	}
4164

P
Paul Menage 已提交
4165 4166 4167 4168
	local_irq_restore(flags);
	return !failed;
}

4169 4170
static int cgroup_rmdir(struct inode *unused_dir, struct dentry *dentry)
{
4171
	struct cgroup *cgrp = dentry->d_fsdata;
4172 4173
	struct dentry *d;
	struct cgroup *parent;
4174
	DEFINE_WAIT(wait);
4175
	struct cgroup_event *event, *tmp;
4176
	int ret;
4177 4178

	/* the vfs holds both inode->i_mutex already */
4179
again:
4180
	mutex_lock(&cgroup_mutex);
4181
	if (atomic_read(&cgrp->count) != 0) {
4182 4183 4184
		mutex_unlock(&cgroup_mutex);
		return -EBUSY;
	}
4185
	if (!list_empty(&cgrp->children)) {
4186 4187 4188
		mutex_unlock(&cgroup_mutex);
		return -EBUSY;
	}
4189
	mutex_unlock(&cgroup_mutex);
L
Li Zefan 已提交
4190

4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201
	/*
	 * In general, subsystem has no css->refcnt after pre_destroy(). But
	 * in racy cases, subsystem may have to get css->refcnt after
	 * pre_destroy() and it makes rmdir return with -EBUSY. This sometimes
	 * make rmdir return -EBUSY too often. To avoid that, we use waitqueue
	 * for cgroup's rmdir. CGRP_WAIT_ON_RMDIR is for synchronizing rmdir
	 * and subsystem's reference count handling. Please see css_get/put
	 * and css_tryget() and cgroup_wakeup_rmdir_waiter() implementation.
	 */
	set_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags);

4202
	/*
L
Li Zefan 已提交
4203 4204
	 * Call pre_destroy handlers of subsys. Notify subsystems
	 * that rmdir() request comes.
4205
	 */
4206
	ret = cgroup_call_pre_destroy(cgrp);
4207 4208
	if (ret) {
		clear_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags);
4209
		return ret;
4210
	}
4211

4212 4213
	mutex_lock(&cgroup_mutex);
	parent = cgrp->parent;
4214
	if (atomic_read(&cgrp->count) || !list_empty(&cgrp->children)) {
4215
		clear_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags);
4216 4217 4218
		mutex_unlock(&cgroup_mutex);
		return -EBUSY;
	}
4219 4220 4221
	prepare_to_wait(&cgroup_rmdir_waitq, &wait, TASK_INTERRUPTIBLE);
	if (!cgroup_clear_css_refs(cgrp)) {
		mutex_unlock(&cgroup_mutex);
4222 4223 4224 4225 4226 4227
		/*
		 * Because someone may call cgroup_wakeup_rmdir_waiter() before
		 * prepare_to_wait(), we need to check this flag.
		 */
		if (test_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags))
			schedule();
4228 4229 4230 4231 4232 4233 4234 4235 4236
		finish_wait(&cgroup_rmdir_waitq, &wait);
		clear_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags);
		if (signal_pending(current))
			return -EINTR;
		goto again;
	}
	/* NO css_tryget() can success after here. */
	finish_wait(&cgroup_rmdir_waitq, &wait);
	clear_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags);
4237

4238
	raw_spin_lock(&release_list_lock);
4239 4240
	set_bit(CGRP_REMOVED, &cgrp->flags);
	if (!list_empty(&cgrp->release_list))
4241
		list_del_init(&cgrp->release_list);
4242
	raw_spin_unlock(&release_list_lock);
4243 4244

	/* delete this cgroup from parent->children */
4245
	list_del_init(&cgrp->sibling);
4246

4247 4248
	list_del_init(&cgrp->allcg_node);

4249
	d = dget(cgrp->dentry);
4250 4251 4252 4253

	cgroup_d_remove_dir(d);
	dput(d);

4254
	set_bit(CGRP_RELEASABLE, &parent->flags);
4255 4256
	check_for_release(parent);

4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270
	/*
	 * Unregister events and notify userspace.
	 * Notify userspace about cgroup removing only after rmdir of cgroup
	 * directory to avoid race between userspace and kernelspace
	 */
	spin_lock(&cgrp->event_list_lock);
	list_for_each_entry_safe(event, tmp, &cgrp->event_list, list) {
		list_del(&event->list);
		remove_wait_queue(event->wqh, &event->wait);
		eventfd_signal(event->eventfd, 1);
		schedule_work(&event->remove);
	}
	spin_unlock(&cgrp->event_list_lock);

4271 4272 4273 4274
	mutex_unlock(&cgroup_mutex);
	return 0;
}

4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288
static void __init_or_module cgroup_init_cftsets(struct cgroup_subsys *ss)
{
	INIT_LIST_HEAD(&ss->cftsets);

	/*
	 * base_cftset is embedded in subsys itself, no need to worry about
	 * deregistration.
	 */
	if (ss->base_cftypes) {
		ss->base_cftset.cfts = ss->base_cftypes;
		list_add_tail(&ss->base_cftset.node, &ss->cftsets);
	}
}

4289
static void __init cgroup_init_subsys(struct cgroup_subsys *ss)
4290 4291
{
	struct cgroup_subsys_state *css;
D
Diego Calleja 已提交
4292 4293

	printk(KERN_INFO "Initializing cgroup subsys %s\n", ss->name);
4294

4295 4296 4297
	/* init base cftset */
	cgroup_init_cftsets(ss);

4298
	/* Create the top cgroup state for this subsystem */
4299
	list_add(&ss->sibling, &rootnode.subsys_list);
4300
	ss->root = &rootnode;
4301
	css = ss->create(dummytop);
4302 4303 4304 4305
	/* We don't handle early failures gracefully */
	BUG_ON(IS_ERR(css));
	init_cgroup_css(css, ss, dummytop);

L
Li Zefan 已提交
4306
	/* Update the init_css_set to contain a subsys
4307
	 * pointer to this state - since the subsystem is
L
Li Zefan 已提交
4308 4309 4310
	 * newly registered, all tasks and hence the
	 * init_css_set is in the subsystem's top cgroup. */
	init_css_set.subsys[ss->subsys_id] = dummytop->subsys[ss->subsys_id];
4311 4312 4313

	need_forkexit_callback |= ss->fork || ss->exit;

L
Li Zefan 已提交
4314 4315 4316 4317 4318
	/* At system boot, before all subsystems have been
	 * registered, no tasks have been forked, so we don't
	 * need to invoke fork callbacks here. */
	BUG_ON(!list_empty(&init_task.tasks));

4319
	ss->active = 1;
4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330

	/* this function shouldn't be used with modular subsystems, since they
	 * need to register a subsys_id, among other things */
	BUG_ON(ss->module);
}

/**
 * cgroup_load_subsys: load and register a modular subsystem at runtime
 * @ss: the subsystem to load
 *
 * This function should be called in a modular subsystem's initcall. If the
T
Thomas Weber 已提交
4331
 * subsystem is built as a module, it will be assigned a new subsys_id and set
4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364
 * up for use. If the subsystem is built-in anyway, work is delegated to the
 * simpler cgroup_init_subsys.
 */
int __init_or_module cgroup_load_subsys(struct cgroup_subsys *ss)
{
	int i;
	struct cgroup_subsys_state *css;

	/* check name and function validity */
	if (ss->name == NULL || strlen(ss->name) > MAX_CGROUP_TYPE_NAMELEN ||
	    ss->create == NULL || ss->destroy == NULL)
		return -EINVAL;

	/*
	 * we don't support callbacks in modular subsystems. this check is
	 * before the ss->module check for consistency; a subsystem that could
	 * be a module should still have no callbacks even if the user isn't
	 * compiling it as one.
	 */
	if (ss->fork || ss->exit)
		return -EINVAL;

	/*
	 * an optionally modular subsystem is built-in: we want to do nothing,
	 * since cgroup_init_subsys will have already taken care of it.
	 */
	if (ss->module == NULL) {
		/* a few sanity checks */
		BUG_ON(ss->subsys_id >= CGROUP_BUILTIN_SUBSYS_COUNT);
		BUG_ON(subsys[ss->subsys_id] != ss);
		return 0;
	}

4365 4366 4367
	/* init base cftset */
	cgroup_init_cftsets(ss);

4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390
	/*
	 * need to register a subsys id before anything else - for example,
	 * init_cgroup_css needs it.
	 */
	mutex_lock(&cgroup_mutex);
	/* find the first empty slot in the array */
	for (i = CGROUP_BUILTIN_SUBSYS_COUNT; i < CGROUP_SUBSYS_COUNT; i++) {
		if (subsys[i] == NULL)
			break;
	}
	if (i == CGROUP_SUBSYS_COUNT) {
		/* maximum number of subsystems already registered! */
		mutex_unlock(&cgroup_mutex);
		return -EBUSY;
	}
	/* assign ourselves the subsys_id */
	ss->subsys_id = i;
	subsys[i] = ss;

	/*
	 * no ss->create seems to need anything important in the ss struct, so
	 * this can happen first (i.e. before the rootnode attachment).
	 */
4391
	css = ss->create(dummytop);
4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408
	if (IS_ERR(css)) {
		/* failure case - need to deassign the subsys[] slot. */
		subsys[i] = NULL;
		mutex_unlock(&cgroup_mutex);
		return PTR_ERR(css);
	}

	list_add(&ss->sibling, &rootnode.subsys_list);
	ss->root = &rootnode;

	/* our new subsystem will be attached to the dummy hierarchy. */
	init_cgroup_css(css, ss, dummytop);
	/* init_idr must be after init_cgroup_css because it sets css->id. */
	if (ss->use_id) {
		int ret = cgroup_init_idr(ss, css);
		if (ret) {
			dummytop->subsys[ss->subsys_id] = NULL;
4409
			ss->destroy(dummytop);
4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449
			subsys[i] = NULL;
			mutex_unlock(&cgroup_mutex);
			return ret;
		}
	}

	/*
	 * Now we need to entangle the css into the existing css_sets. unlike
	 * in cgroup_init_subsys, there are now multiple css_sets, so each one
	 * will need a new pointer to it; done by iterating the css_set_table.
	 * furthermore, modifying the existing css_sets will corrupt the hash
	 * table state, so each changed css_set will need its hash recomputed.
	 * this is all done under the css_set_lock.
	 */
	write_lock(&css_set_lock);
	for (i = 0; i < CSS_SET_TABLE_SIZE; i++) {
		struct css_set *cg;
		struct hlist_node *node, *tmp;
		struct hlist_head *bucket = &css_set_table[i], *new_bucket;

		hlist_for_each_entry_safe(cg, node, tmp, bucket, hlist) {
			/* skip entries that we already rehashed */
			if (cg->subsys[ss->subsys_id])
				continue;
			/* remove existing entry */
			hlist_del(&cg->hlist);
			/* set new value */
			cg->subsys[ss->subsys_id] = css;
			/* recompute hash and restore entry */
			new_bucket = css_set_hash(cg->subsys);
			hlist_add_head(&cg->hlist, new_bucket);
		}
	}
	write_unlock(&css_set_lock);

	ss->active = 1;

	/* success! */
	mutex_unlock(&cgroup_mutex);
	return 0;
4450
}
4451
EXPORT_SYMBOL_GPL(cgroup_load_subsys);
4452

B
Ben Blum 已提交
4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480
/**
 * cgroup_unload_subsys: unload a modular subsystem
 * @ss: the subsystem to unload
 *
 * This function should be called in a modular subsystem's exitcall. When this
 * function is invoked, the refcount on the subsystem's module will be 0, so
 * the subsystem will not be attached to any hierarchy.
 */
void cgroup_unload_subsys(struct cgroup_subsys *ss)
{
	struct cg_cgroup_link *link;
	struct hlist_head *hhead;

	BUG_ON(ss->module == NULL);

	/*
	 * we shouldn't be called if the subsystem is in use, and the use of
	 * try_module_get in parse_cgroupfs_options should ensure that it
	 * doesn't start being used while we're killing it off.
	 */
	BUG_ON(ss->root != &rootnode);

	mutex_lock(&cgroup_mutex);
	/* deassign the subsys_id */
	BUG_ON(ss->subsys_id < CGROUP_BUILTIN_SUBSYS_COUNT);
	subsys[ss->subsys_id] = NULL;

	/* remove subsystem from rootnode's list of subsystems */
4481
	list_del_init(&ss->sibling);
B
Ben Blum 已提交
4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504

	/*
	 * disentangle the css from all css_sets attached to the dummytop. as
	 * in loading, we need to pay our respects to the hashtable gods.
	 */
	write_lock(&css_set_lock);
	list_for_each_entry(link, &dummytop->css_sets, cgrp_link_list) {
		struct css_set *cg = link->cg;

		hlist_del(&cg->hlist);
		BUG_ON(!cg->subsys[ss->subsys_id]);
		cg->subsys[ss->subsys_id] = NULL;
		hhead = css_set_hash(cg->subsys);
		hlist_add_head(&cg->hlist, hhead);
	}
	write_unlock(&css_set_lock);

	/*
	 * remove subsystem's css from the dummytop and free it - need to free
	 * before marking as null because ss->destroy needs the cgrp->subsys
	 * pointer to find their state. note that this also takes care of
	 * freeing the css_id.
	 */
4505
	ss->destroy(dummytop);
B
Ben Blum 已提交
4506 4507 4508 4509 4510 4511
	dummytop->subsys[ss->subsys_id] = NULL;

	mutex_unlock(&cgroup_mutex);
}
EXPORT_SYMBOL_GPL(cgroup_unload_subsys);

4512
/**
L
Li Zefan 已提交
4513 4514 4515 4516
 * cgroup_init_early - cgroup initialization at system boot
 *
 * Initialize cgroups at system boot, and initialize any
 * subsystems that request early init.
4517 4518 4519 4520
 */
int __init cgroup_init_early(void)
{
	int i;
4521
	atomic_set(&init_css_set.refcount, 1);
4522 4523
	INIT_LIST_HEAD(&init_css_set.cg_links);
	INIT_LIST_HEAD(&init_css_set.tasks);
4524
	INIT_HLIST_NODE(&init_css_set.hlist);
4525
	css_set_count = 1;
4526
	init_cgroup_root(&rootnode);
4527 4528 4529 4530
	root_count = 1;
	init_task.cgroups = &init_css_set;

	init_css_set_link.cg = &init_css_set;
4531
	init_css_set_link.cgrp = dummytop;
4532
	list_add(&init_css_set_link.cgrp_link_list,
4533 4534 4535
		 &rootnode.top_cgroup.css_sets);
	list_add(&init_css_set_link.cg_link_list,
		 &init_css_set.cg_links);
4536

4537 4538 4539
	for (i = 0; i < CSS_SET_TABLE_SIZE; i++)
		INIT_HLIST_HEAD(&css_set_table[i]);

B
Ben Blum 已提交
4540 4541
	/* at bootup time, we don't worry about modular subsystems */
	for (i = 0; i < CGROUP_BUILTIN_SUBSYS_COUNT; i++) {
4542 4543 4544 4545 4546 4547 4548
		struct cgroup_subsys *ss = subsys[i];

		BUG_ON(!ss->name);
		BUG_ON(strlen(ss->name) > MAX_CGROUP_TYPE_NAMELEN);
		BUG_ON(!ss->create);
		BUG_ON(!ss->destroy);
		if (ss->subsys_id != i) {
D
Diego Calleja 已提交
4549
			printk(KERN_ERR "cgroup: Subsys %s id == %d\n",
4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560
			       ss->name, ss->subsys_id);
			BUG();
		}

		if (ss->early_init)
			cgroup_init_subsys(ss);
	}
	return 0;
}

/**
L
Li Zefan 已提交
4561 4562 4563 4564
 * cgroup_init - cgroup initialization
 *
 * Register cgroup filesystem and /proc file, and initialize
 * any subsystems that didn't request early init.
4565 4566 4567 4568 4569
 */
int __init cgroup_init(void)
{
	int err;
	int i;
4570
	struct hlist_head *hhead;
4571 4572 4573 4574

	err = bdi_init(&cgroup_backing_dev_info);
	if (err)
		return err;
4575

B
Ben Blum 已提交
4576 4577
	/* at bootup time, we don't worry about modular subsystems */
	for (i = 0; i < CGROUP_BUILTIN_SUBSYS_COUNT; i++) {
4578 4579 4580
		struct cgroup_subsys *ss = subsys[i];
		if (!ss->early_init)
			cgroup_init_subsys(ss);
K
KAMEZAWA Hiroyuki 已提交
4581
		if (ss->use_id)
4582
			cgroup_init_idr(ss, init_css_set.subsys[ss->subsys_id]);
4583 4584
	}

4585 4586 4587
	/* Add init_css_set to the hash table */
	hhead = css_set_hash(init_css_set.subsys);
	hlist_add_head(&init_css_set.hlist, hhead);
4588
	BUG_ON(!init_root_id(&rootnode));
4589 4590 4591 4592 4593 4594 4595

	cgroup_kobj = kobject_create_and_add("cgroup", fs_kobj);
	if (!cgroup_kobj) {
		err = -ENOMEM;
		goto out;
	}

4596
	err = register_filesystem(&cgroup_fs_type);
4597 4598
	if (err < 0) {
		kobject_put(cgroup_kobj);
4599
		goto out;
4600
	}
4601

L
Li Zefan 已提交
4602
	proc_create("cgroups", 0, NULL, &proc_cgroupstats_operations);
4603

4604
out:
4605 4606 4607
	if (err)
		bdi_destroy(&cgroup_backing_dev_info);

4608 4609
	return err;
}
4610

4611 4612 4613 4614 4615 4616
/*
 * proc_cgroup_show()
 *  - Print task's cgroup paths into seq_file, one line for each hierarchy
 *  - Used for /proc/<pid>/cgroup.
 *  - No need to task_lock(tsk) on this tsk->cgroup reference, as it
 *    doesn't really matter if tsk->cgroup changes after we read it,
4617
 *    and we take cgroup_mutex, keeping cgroup_attach_task() from changing it
4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646
 *    anyway.  No need to check that tsk->cgroup != NULL, thanks to
 *    the_top_cgroup_hack in cgroup_exit(), which sets an exiting tasks
 *    cgroup to top_cgroup.
 */

/* TODO: Use a proper seq_file iterator */
static int proc_cgroup_show(struct seq_file *m, void *v)
{
	struct pid *pid;
	struct task_struct *tsk;
	char *buf;
	int retval;
	struct cgroupfs_root *root;

	retval = -ENOMEM;
	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
	if (!buf)
		goto out;

	retval = -ESRCH;
	pid = m->private;
	tsk = get_pid_task(pid, PIDTYPE_PID);
	if (!tsk)
		goto out_free;

	retval = 0;

	mutex_lock(&cgroup_mutex);

4647
	for_each_active_root(root) {
4648
		struct cgroup_subsys *ss;
4649
		struct cgroup *cgrp;
4650 4651
		int count = 0;

4652
		seq_printf(m, "%d:", root->hierarchy_id);
4653 4654
		for_each_subsys(root, ss)
			seq_printf(m, "%s%s", count++ ? "," : "", ss->name);
4655 4656 4657
		if (strlen(root->name))
			seq_printf(m, "%sname=%s", count ? "," : "",
				   root->name);
4658
		seq_putc(m, ':');
4659
		cgrp = task_cgroup_from_root(tsk, root);
4660
		retval = cgroup_path(cgrp, buf, PAGE_SIZE);
4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681
		if (retval < 0)
			goto out_unlock;
		seq_puts(m, buf);
		seq_putc(m, '\n');
	}

out_unlock:
	mutex_unlock(&cgroup_mutex);
	put_task_struct(tsk);
out_free:
	kfree(buf);
out:
	return retval;
}

static int cgroup_open(struct inode *inode, struct file *file)
{
	struct pid *pid = PROC_I(inode)->pid;
	return single_open(file, proc_cgroup_show, pid);
}

4682
const struct file_operations proc_cgroup_operations = {
4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693
	.open		= cgroup_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
};

/* Display information about each subsystem and each hierarchy */
static int proc_cgroupstats_show(struct seq_file *m, void *v)
{
	int i;

4694
	seq_puts(m, "#subsys_name\thierarchy\tnum_cgroups\tenabled\n");
B
Ben Blum 已提交
4695 4696 4697 4698 4699
	/*
	 * ideally we don't want subsystems moving around while we do this.
	 * cgroup_mutex is also necessary to guarantee an atomic snapshot of
	 * subsys/hierarchy state.
	 */
4700 4701 4702
	mutex_lock(&cgroup_mutex);
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
		struct cgroup_subsys *ss = subsys[i];
B
Ben Blum 已提交
4703 4704
		if (ss == NULL)
			continue;
4705 4706
		seq_printf(m, "%s\t%d\t%d\t%d\n",
			   ss->name, ss->root->hierarchy_id,
4707
			   ss->root->number_of_cgroups, !ss->disabled);
4708 4709 4710 4711 4712 4713 4714
	}
	mutex_unlock(&cgroup_mutex);
	return 0;
}

static int cgroupstats_open(struct inode *inode, struct file *file)
{
A
Al Viro 已提交
4715
	return single_open(file, proc_cgroupstats_show, NULL);
4716 4717
}

4718
static const struct file_operations proc_cgroupstats_operations = {
4719 4720 4721 4722 4723 4724
	.open = cgroupstats_open,
	.read = seq_read,
	.llseek = seq_lseek,
	.release = single_release,
};

4725 4726
/**
 * cgroup_fork - attach newly forked task to its parents cgroup.
L
Li Zefan 已提交
4727
 * @child: pointer to task_struct of forking parent process.
4728 4729 4730 4731 4732
 *
 * Description: A task inherits its parent's cgroup at fork().
 *
 * A pointer to the shared css_set was automatically copied in
 * fork.c by dup_task_struct().  However, we ignore that copy, since
4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743
 * it was not made under the protection of RCU, cgroup_mutex or
 * threadgroup_change_begin(), so it might no longer be a valid
 * cgroup pointer.  cgroup_attach_task() might have already changed
 * current->cgroups, allowing the previously referenced cgroup
 * group to be removed and freed.
 *
 * Outside the pointer validity we also need to process the css_set
 * inheritance between threadgoup_change_begin() and
 * threadgoup_change_end(), this way there is no leak in any process
 * wide migration performed by cgroup_attach_proc() that could otherwise
 * miss a thread because it is too early or too late in the fork stage.
4744 4745 4746 4747 4748 4749
 *
 * At the point that cgroup_fork() is called, 'current' is the parent
 * task, and the passed argument 'child' points to the child task.
 */
void cgroup_fork(struct task_struct *child)
{
4750 4751 4752 4753 4754 4755
	/*
	 * We don't need to task_lock() current because current->cgroups
	 * can't be changed concurrently here. The parent obviously hasn't
	 * exited and called cgroup_exit(), and we are synchronized against
	 * cgroup migration through threadgroup_change_begin().
	 */
4756 4757 4758
	child->cgroups = current->cgroups;
	get_css_set(child->cgroups);
	INIT_LIST_HEAD(&child->cg_list);
4759 4760 4761
}

/**
L
Li Zefan 已提交
4762 4763 4764 4765 4766 4767
 * cgroup_fork_callbacks - run fork callbacks
 * @child: the new task
 *
 * Called on a new task very soon before adding it to the
 * tasklist. No need to take any locks since no-one can
 * be operating on this task.
4768 4769 4770 4771 4772
 */
void cgroup_fork_callbacks(struct task_struct *child)
{
	if (need_forkexit_callback) {
		int i;
B
Ben Blum 已提交
4773 4774 4775 4776 4777 4778
		/*
		 * forkexit callbacks are only supported for builtin
		 * subsystems, and the builtin section of the subsys array is
		 * immutable, so we don't need to lock the subsys array here.
		 */
		for (i = 0; i < CGROUP_BUILTIN_SUBSYS_COUNT; i++) {
4779 4780
			struct cgroup_subsys *ss = subsys[i];
			if (ss->fork)
4781
				ss->fork(child);
4782 4783 4784 4785
		}
	}
}

4786
/**
L
Li Zefan 已提交
4787 4788 4789 4790 4791 4792 4793 4794
 * cgroup_post_fork - called on a new task after adding it to the task list
 * @child: the task in question
 *
 * Adds the task to the list running through its css_set if necessary.
 * Has to be after the task is visible on the task list in case we race
 * with the first call to cgroup_iter_start() - to guarantee that the
 * new task ends up on its list.
 */
4795 4796
void cgroup_post_fork(struct task_struct *child)
{
4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807
	/*
	 * use_task_css_set_links is set to 1 before we walk the tasklist
	 * under the tasklist_lock and we read it here after we added the child
	 * to the tasklist under the tasklist_lock as well. If the child wasn't
	 * yet in the tasklist when we walked through it from
	 * cgroup_enable_task_cg_lists(), then use_task_css_set_links value
	 * should be visible now due to the paired locking and barriers implied
	 * by LOCK/UNLOCK: it is written before the tasklist_lock unlock
	 * in cgroup_enable_task_cg_lists() and read here after the tasklist_lock
	 * lock on fork.
	 */
4808 4809
	if (use_task_css_set_links) {
		write_lock(&css_set_lock);
4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820
		if (list_empty(&child->cg_list)) {
			/*
			 * It's safe to use child->cgroups without task_lock()
			 * here because we are protected through
			 * threadgroup_change_begin() against concurrent
			 * css_set change in cgroup_task_migrate(). Also
			 * the task can't exit at that point until
			 * wake_up_new_task() is called, so we are protected
			 * against cgroup_exit() setting child->cgroup to
			 * init_css_set.
			 */
4821
			list_add(&child->cg_list, &child->cgroups->tasks);
4822
		}
4823 4824 4825
		write_unlock(&css_set_lock);
	}
}
4826 4827 4828
/**
 * cgroup_exit - detach cgroup from exiting task
 * @tsk: pointer to task_struct of exiting process
L
Li Zefan 已提交
4829
 * @run_callback: run exit callbacks?
4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857
 *
 * Description: Detach cgroup from @tsk and release it.
 *
 * Note that cgroups marked notify_on_release force every task in
 * them to take the global cgroup_mutex mutex when exiting.
 * This could impact scaling on very large systems.  Be reluctant to
 * use notify_on_release cgroups where very high task exit scaling
 * is required on large systems.
 *
 * the_top_cgroup_hack:
 *
 *    Set the exiting tasks cgroup to the root cgroup (top_cgroup).
 *
 *    We call cgroup_exit() while the task is still competent to
 *    handle notify_on_release(), then leave the task attached to the
 *    root cgroup in each hierarchy for the remainder of its exit.
 *
 *    To do this properly, we would increment the reference count on
 *    top_cgroup, and near the very end of the kernel/exit.c do_exit()
 *    code we would add a second cgroup function call, to drop that
 *    reference.  This would just create an unnecessary hot spot on
 *    the top_cgroup reference count, to no avail.
 *
 *    Normally, holding a reference to a cgroup without bumping its
 *    count is unsafe.   The cgroup could go away, or someone could
 *    attach us to a different cgroup, decrementing the count on
 *    the first cgroup that we never incremented.  But in this case,
 *    top_cgroup isn't going away, and either task has PF_EXITING set,
4858 4859
 *    which wards off any cgroup_attach_task() attempts, or task is a failed
 *    fork, never visible to cgroup_attach_task.
4860 4861 4862
 */
void cgroup_exit(struct task_struct *tsk, int run_callbacks)
{
4863
	struct css_set *cg;
4864
	int i;
4865 4866 4867 4868 4869 4870 4871 4872 4873

	/*
	 * Unlink from the css_set task list if necessary.
	 * Optimistically check cg_list before taking
	 * css_set_lock
	 */
	if (!list_empty(&tsk->cg_list)) {
		write_lock(&css_set_lock);
		if (!list_empty(&tsk->cg_list))
4874
			list_del_init(&tsk->cg_list);
4875 4876 4877
		write_unlock(&css_set_lock);
	}

4878 4879
	/* Reassign the task to the init_css_set. */
	task_lock(tsk);
4880 4881
	cg = tsk->cgroups;
	tsk->cgroups = &init_css_set;
4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893

	if (run_callbacks && need_forkexit_callback) {
		/*
		 * modular subsystems can't use callbacks, so no need to lock
		 * the subsys array
		 */
		for (i = 0; i < CGROUP_BUILTIN_SUBSYS_COUNT; i++) {
			struct cgroup_subsys *ss = subsys[i];
			if (ss->exit) {
				struct cgroup *old_cgrp =
					rcu_dereference_raw(cg->subsys[i])->cgroup;
				struct cgroup *cgrp = task_cgroup(tsk, i);
4894
				ss->exit(cgrp, old_cgrp, tsk);
4895 4896 4897
			}
		}
	}
4898
	task_unlock(tsk);
4899

4900
	if (cg)
4901
		put_css_set_taskexit(cg);
4902
}
4903

L
Li Zefan 已提交
4904
/**
4905
 * cgroup_is_descendant - see if @cgrp is a descendant of @task's cgrp
L
Li Zefan 已提交
4906
 * @cgrp: the cgroup in question
4907
 * @task: the task in question
L
Li Zefan 已提交
4908
 *
4909 4910
 * See if @cgrp is a descendant of @task's cgroup in the appropriate
 * hierarchy.
4911 4912 4913 4914 4915 4916
 *
 * If we are sending in dummytop, then presumably we are creating
 * the top cgroup in the subsystem.
 *
 * Called only by the ns (nsproxy) cgroup.
 */
4917
int cgroup_is_descendant(const struct cgroup *cgrp, struct task_struct *task)
4918 4919 4920 4921
{
	int ret;
	struct cgroup *target;

4922
	if (cgrp == dummytop)
4923 4924
		return 1;

4925
	target = task_cgroup_from_root(task, cgrp->root);
4926 4927 4928
	while (cgrp != target && cgrp!= cgrp->top_cgroup)
		cgrp = cgrp->parent;
	ret = (cgrp == target);
4929 4930
	return ret;
}
4931

4932
static void check_for_release(struct cgroup *cgrp)
4933 4934 4935
{
	/* All of these checks rely on RCU to keep the cgroup
	 * structure alive */
4936 4937
	if (cgroup_is_releasable(cgrp) && !atomic_read(&cgrp->count)
	    && list_empty(&cgrp->children) && !cgroup_has_css_refs(cgrp)) {
4938 4939 4940 4941
		/* Control Group is currently removeable. If it's not
		 * already queued for a userspace notification, queue
		 * it now */
		int need_schedule_work = 0;
4942
		raw_spin_lock(&release_list_lock);
4943 4944 4945
		if (!cgroup_is_removed(cgrp) &&
		    list_empty(&cgrp->release_list)) {
			list_add(&cgrp->release_list, &release_list);
4946 4947
			need_schedule_work = 1;
		}
4948
		raw_spin_unlock(&release_list_lock);
4949 4950 4951 4952 4953
		if (need_schedule_work)
			schedule_work(&release_agent_work);
	}
}

4954
/* Caller must verify that the css is not for root cgroup */
4955 4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970
bool __css_tryget(struct cgroup_subsys_state *css)
{
	do {
		int v = css_refcnt(css);

		if (atomic_cmpxchg(&css->refcnt, v, v + 1) == v)
			return true;
		cpu_relax();
	} while (!test_bit(CSS_REMOVED, &css->flags));

	return false;
}
EXPORT_SYMBOL_GPL(__css_tryget);

/* Caller must verify that the css is not for root cgroup */
void __css_put(struct cgroup_subsys_state *css)
4971
{
4972
	struct cgroup *cgrp = css->cgroup;
4973
	int v;
4974

4975
	rcu_read_lock();
4976 4977 4978
	v = css_unbias_refcnt(atomic_dec_return(&css->refcnt));

	switch (v) {
4979
	case 1:
4980 4981 4982 4983
		if (notify_on_release(cgrp)) {
			set_bit(CGRP_RELEASABLE, &cgrp->flags);
			check_for_release(cgrp);
		}
4984
		cgroup_wakeup_rmdir_waiter(cgrp);
4985 4986 4987 4988 4989
		break;
	case 0:
		if (!test_bit(CSS_CLEAR_CSS_REFS, &css->flags))
			schedule_work(&css->dput_work);
		break;
4990 4991 4992
	}
	rcu_read_unlock();
}
B
Ben Blum 已提交
4993
EXPORT_SYMBOL_GPL(__css_put);
4994 4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015 5016 5017 5018 5019 5020 5021

/*
 * Notify userspace when a cgroup is released, by running the
 * configured release agent with the name of the cgroup (path
 * relative to the root of cgroup file system) as the argument.
 *
 * Most likely, this user command will try to rmdir this cgroup.
 *
 * This races with the possibility that some other task will be
 * attached to this cgroup before it is removed, or that some other
 * user task will 'mkdir' a child cgroup of this cgroup.  That's ok.
 * The presumed 'rmdir' will fail quietly if this cgroup is no longer
 * unused, and this cgroup will be reprieved from its death sentence,
 * to continue to serve a useful existence.  Next time it's released,
 * we will get notified again, if it still has 'notify_on_release' set.
 *
 * The final arg to call_usermodehelper() is UMH_WAIT_EXEC, which
 * means only wait until the task is successfully execve()'d.  The
 * separate release agent task is forked by call_usermodehelper(),
 * then control in this thread returns here, without waiting for the
 * release agent task.  We don't bother to wait because the caller of
 * this routine has no use for the exit status of the release agent
 * task, so no sense holding our caller up for that.
 */
static void cgroup_release_agent(struct work_struct *work)
{
	BUG_ON(work != &release_agent_work);
	mutex_lock(&cgroup_mutex);
5022
	raw_spin_lock(&release_list_lock);
5023 5024 5025
	while (!list_empty(&release_list)) {
		char *argv[3], *envp[3];
		int i;
5026
		char *pathbuf = NULL, *agentbuf = NULL;
5027
		struct cgroup *cgrp = list_entry(release_list.next,
5028 5029
						    struct cgroup,
						    release_list);
5030
		list_del_init(&cgrp->release_list);
5031
		raw_spin_unlock(&release_list_lock);
5032
		pathbuf = kmalloc(PAGE_SIZE, GFP_KERNEL);
5033 5034 5035 5036 5037 5038 5039
		if (!pathbuf)
			goto continue_free;
		if (cgroup_path(cgrp, pathbuf, PAGE_SIZE) < 0)
			goto continue_free;
		agentbuf = kstrdup(cgrp->root->release_agent_path, GFP_KERNEL);
		if (!agentbuf)
			goto continue_free;
5040 5041

		i = 0;
5042 5043
		argv[i++] = agentbuf;
		argv[i++] = pathbuf;
5044 5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057
		argv[i] = NULL;

		i = 0;
		/* minimal command environment */
		envp[i++] = "HOME=/";
		envp[i++] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin";
		envp[i] = NULL;

		/* Drop the lock while we invoke the usermode helper,
		 * since the exec could involve hitting disk and hence
		 * be a slow process */
		mutex_unlock(&cgroup_mutex);
		call_usermodehelper(argv[0], argv, envp, UMH_WAIT_EXEC);
		mutex_lock(&cgroup_mutex);
5058 5059 5060
 continue_free:
		kfree(pathbuf);
		kfree(agentbuf);
5061
		raw_spin_lock(&release_list_lock);
5062
	}
5063
	raw_spin_unlock(&release_list_lock);
5064 5065
	mutex_unlock(&cgroup_mutex);
}
5066 5067 5068 5069 5070 5071 5072 5073 5074

static int __init cgroup_disable(char *str)
{
	int i;
	char *token;

	while ((token = strsep(&str, ",")) != NULL) {
		if (!*token)
			continue;
B
Ben Blum 已提交
5075 5076 5077 5078 5079
		/*
		 * cgroup_disable, being at boot time, can't know about module
		 * subsystems, so we don't worry about them.
		 */
		for (i = 0; i < CGROUP_BUILTIN_SUBSYS_COUNT; i++) {
5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092
			struct cgroup_subsys *ss = subsys[i];

			if (!strcmp(token, ss->name)) {
				ss->disabled = 1;
				printk(KERN_INFO "Disabling %s control group"
					" subsystem\n", ss->name);
				break;
			}
		}
	}
	return 1;
}
__setup("cgroup_disable=", cgroup_disable);
K
KAMEZAWA Hiroyuki 已提交
5093 5094 5095 5096 5097 5098 5099 5100 5101 5102

/*
 * Functons for CSS ID.
 */

/*
 *To get ID other than 0, this should be called when !cgroup_is_removed().
 */
unsigned short css_id(struct cgroup_subsys_state *css)
{
5103 5104 5105 5106 5107 5108 5109
	struct css_id *cssid;

	/*
	 * This css_id() can return correct value when somone has refcnt
	 * on this or this is under rcu_read_lock(). Once css->id is allocated,
	 * it's unchanged until freed.
	 */
5110
	cssid = rcu_dereference_check(css->id, css_refcnt(css));
K
KAMEZAWA Hiroyuki 已提交
5111 5112 5113 5114 5115

	if (cssid)
		return cssid->id;
	return 0;
}
B
Ben Blum 已提交
5116
EXPORT_SYMBOL_GPL(css_id);
K
KAMEZAWA Hiroyuki 已提交
5117 5118 5119

unsigned short css_depth(struct cgroup_subsys_state *css)
{
5120 5121
	struct css_id *cssid;

5122
	cssid = rcu_dereference_check(css->id, css_refcnt(css));
K
KAMEZAWA Hiroyuki 已提交
5123 5124 5125 5126 5127

	if (cssid)
		return cssid->depth;
	return 0;
}
B
Ben Blum 已提交
5128
EXPORT_SYMBOL_GPL(css_depth);
K
KAMEZAWA Hiroyuki 已提交
5129

5130 5131 5132 5133 5134 5135
/**
 *  css_is_ancestor - test "root" css is an ancestor of "child"
 * @child: the css to be tested.
 * @root: the css supporsed to be an ancestor of the child.
 *
 * Returns true if "root" is an ancestor of "child" in its hierarchy. Because
5136
 * this function reads css->id, the caller must hold rcu_read_lock().
5137 5138 5139 5140 5141 5142
 * But, considering usual usage, the csses should be valid objects after test.
 * Assuming that the caller will do some action to the child if this returns
 * returns true, the caller must take "child";s reference count.
 * If "child" is valid object and this returns true, "root" is valid, too.
 */

K
KAMEZAWA Hiroyuki 已提交
5143
bool css_is_ancestor(struct cgroup_subsys_state *child,
5144
		    const struct cgroup_subsys_state *root)
K
KAMEZAWA Hiroyuki 已提交
5145
{
5146 5147
	struct css_id *child_id;
	struct css_id *root_id;
K
KAMEZAWA Hiroyuki 已提交
5148

5149
	child_id  = rcu_dereference(child->id);
5150 5151
	if (!child_id)
		return false;
5152
	root_id = rcu_dereference(root->id);
5153 5154 5155 5156 5157 5158 5159
	if (!root_id)
		return false;
	if (child_id->depth < root_id->depth)
		return false;
	if (child_id->stack[root_id->depth] != root_id->id)
		return false;
	return true;
K
KAMEZAWA Hiroyuki 已提交
5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172
}

void free_css_id(struct cgroup_subsys *ss, struct cgroup_subsys_state *css)
{
	struct css_id *id = css->id;
	/* When this is called before css_id initialization, id can be NULL */
	if (!id)
		return;

	BUG_ON(!ss->use_id);

	rcu_assign_pointer(id->css, NULL);
	rcu_assign_pointer(css->id, NULL);
5173
	spin_lock(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5174
	idr_remove(&ss->idr, id->id);
5175
	spin_unlock(&ss->id_lock);
5176
	kfree_rcu(id, rcu_head);
K
KAMEZAWA Hiroyuki 已提交
5177
}
B
Ben Blum 已提交
5178
EXPORT_SYMBOL_GPL(free_css_id);
K
KAMEZAWA Hiroyuki 已提交
5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200

/*
 * This is called by init or create(). Then, calls to this function are
 * always serialized (By cgroup_mutex() at create()).
 */

static struct css_id *get_new_cssid(struct cgroup_subsys *ss, int depth)
{
	struct css_id *newid;
	int myid, error, size;

	BUG_ON(!ss->use_id);

	size = sizeof(*newid) + sizeof(unsigned short) * (depth + 1);
	newid = kzalloc(size, GFP_KERNEL);
	if (!newid)
		return ERR_PTR(-ENOMEM);
	/* get id */
	if (unlikely(!idr_pre_get(&ss->idr, GFP_KERNEL))) {
		error = -ENOMEM;
		goto err_out;
	}
5201
	spin_lock(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5202 5203
	/* Don't use 0. allocates an ID of 1-65535 */
	error = idr_get_new_above(&ss->idr, newid, 1, &myid);
5204
	spin_unlock(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218

	/* Returns error when there are no free spaces for new ID.*/
	if (error) {
		error = -ENOSPC;
		goto err_out;
	}
	if (myid > CSS_ID_MAX)
		goto remove_idr;

	newid->id = myid;
	newid->depth = depth;
	return newid;
remove_idr:
	error = -ENOSPC;
5219
	spin_lock(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5220
	idr_remove(&ss->idr, myid);
5221
	spin_unlock(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5222 5223 5224 5225 5226 5227
err_out:
	kfree(newid);
	return ERR_PTR(error);

}

5228 5229
static int __init_or_module cgroup_init_idr(struct cgroup_subsys *ss,
					    struct cgroup_subsys_state *rootcss)
K
KAMEZAWA Hiroyuki 已提交
5230 5231 5232
{
	struct css_id *newid;

5233
	spin_lock_init(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250
	idr_init(&ss->idr);

	newid = get_new_cssid(ss, 0);
	if (IS_ERR(newid))
		return PTR_ERR(newid);

	newid->stack[0] = newid->id;
	newid->css = rootcss;
	rootcss->id = newid;
	return 0;
}

static int alloc_css_id(struct cgroup_subsys *ss, struct cgroup *parent,
			struct cgroup *child)
{
	int subsys_id, i, depth = 0;
	struct cgroup_subsys_state *parent_css, *child_css;
5251
	struct css_id *child_id, *parent_id;
K
KAMEZAWA Hiroyuki 已提交
5252 5253 5254 5255 5256

	subsys_id = ss->subsys_id;
	parent_css = parent->subsys[subsys_id];
	child_css = child->subsys[subsys_id];
	parent_id = parent_css->id;
5257
	depth = parent_id->depth + 1;
K
KAMEZAWA Hiroyuki 已提交
5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294

	child_id = get_new_cssid(ss, depth);
	if (IS_ERR(child_id))
		return PTR_ERR(child_id);

	for (i = 0; i < depth; i++)
		child_id->stack[i] = parent_id->stack[i];
	child_id->stack[depth] = child_id->id;
	/*
	 * child_id->css pointer will be set after this cgroup is available
	 * see cgroup_populate_dir()
	 */
	rcu_assign_pointer(child_css->id, child_id);

	return 0;
}

/**
 * css_lookup - lookup css by id
 * @ss: cgroup subsys to be looked into.
 * @id: the id
 *
 * Returns pointer to cgroup_subsys_state if there is valid one with id.
 * NULL if not. Should be called under rcu_read_lock()
 */
struct cgroup_subsys_state *css_lookup(struct cgroup_subsys *ss, int id)
{
	struct css_id *cssid = NULL;

	BUG_ON(!ss->use_id);
	cssid = idr_find(&ss->idr, id);

	if (unlikely(!cssid))
		return NULL;

	return rcu_dereference(cssid->css);
}
B
Ben Blum 已提交
5295
EXPORT_SYMBOL_GPL(css_lookup);
K
KAMEZAWA Hiroyuki 已提交
5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313 5314 5315 5316 5317 5318 5319 5320

/**
 * css_get_next - lookup next cgroup under specified hierarchy.
 * @ss: pointer to subsystem
 * @id: current position of iteration.
 * @root: pointer to css. search tree under this.
 * @foundid: position of found object.
 *
 * Search next css under the specified hierarchy of rootid. Calling under
 * rcu_read_lock() is necessary. Returns NULL if it reaches the end.
 */
struct cgroup_subsys_state *
css_get_next(struct cgroup_subsys *ss, int id,
	     struct cgroup_subsys_state *root, int *foundid)
{
	struct cgroup_subsys_state *ret = NULL;
	struct css_id *tmp;
	int tmpid;
	int rootid = css_id(root);
	int depth = css_depth(root);

	if (!rootid)
		return NULL;

	BUG_ON(!ss->use_id);
5321 5322
	WARN_ON_ONCE(!rcu_read_lock_held());

K
KAMEZAWA Hiroyuki 已提交
5323 5324 5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345
	/* fill start point for scan */
	tmpid = id;
	while (1) {
		/*
		 * scan next entry from bitmap(tree), tmpid is updated after
		 * idr_get_next().
		 */
		tmp = idr_get_next(&ss->idr, &tmpid);
		if (!tmp)
			break;
		if (tmp->depth >= depth && tmp->stack[depth] == rootid) {
			ret = rcu_dereference(tmp->css);
			if (ret) {
				*foundid = tmpid;
				break;
			}
		}
		/* continue to scan from next id */
		tmpid = tmpid + 1;
	}
	return ret;
}

S
Stephane Eranian 已提交
5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368
/*
 * get corresponding css from file open on cgroupfs directory
 */
struct cgroup_subsys_state *cgroup_css_from_dir(struct file *f, int id)
{
	struct cgroup *cgrp;
	struct inode *inode;
	struct cgroup_subsys_state *css;

	inode = f->f_dentry->d_inode;
	/* check in cgroup filesystem dir */
	if (inode->i_op != &cgroup_dir_inode_operations)
		return ERR_PTR(-EBADF);

	if (id < 0 || id >= CGROUP_SUBSYS_COUNT)
		return ERR_PTR(-EINVAL);

	/* get cgroup */
	cgrp = __d_cgrp(f->f_dentry);
	css = cgrp->subsys[id];
	return css ? css : ERR_PTR(-ENOENT);
}

5369
#ifdef CONFIG_CGROUP_DEBUG
5370
static struct cgroup_subsys_state *debug_create(struct cgroup *cont)
5371 5372 5373 5374 5375 5376 5377 5378 5379
{
	struct cgroup_subsys_state *css = kzalloc(sizeof(*css), GFP_KERNEL);

	if (!css)
		return ERR_PTR(-ENOMEM);

	return css;
}

5380
static void debug_destroy(struct cgroup *cont)
5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410
{
	kfree(cont->subsys[debug_subsys_id]);
}

static u64 cgroup_refcount_read(struct cgroup *cont, struct cftype *cft)
{
	return atomic_read(&cont->count);
}

static u64 debug_taskcount_read(struct cgroup *cont, struct cftype *cft)
{
	return cgroup_task_count(cont);
}

static u64 current_css_set_read(struct cgroup *cont, struct cftype *cft)
{
	return (u64)(unsigned long)current->cgroups;
}

static u64 current_css_set_refcount_read(struct cgroup *cont,
					   struct cftype *cft)
{
	u64 count;

	rcu_read_lock();
	count = atomic_read(&current->cgroups->refcount);
	rcu_read_unlock();
	return count;
}

5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423 5424 5425 5426 5427 5428
static int current_css_set_cg_links_read(struct cgroup *cont,
					 struct cftype *cft,
					 struct seq_file *seq)
{
	struct cg_cgroup_link *link;
	struct css_set *cg;

	read_lock(&css_set_lock);
	rcu_read_lock();
	cg = rcu_dereference(current->cgroups);
	list_for_each_entry(link, &cg->cg_links, cg_link_list) {
		struct cgroup *c = link->cgrp;
		const char *name;

		if (c->dentry)
			name = c->dentry->d_name.name;
		else
			name = "?";
5429 5430
		seq_printf(seq, "Root %d group %s\n",
			   c->root->hierarchy_id, name);
5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463
	}
	rcu_read_unlock();
	read_unlock(&css_set_lock);
	return 0;
}

#define MAX_TASKS_SHOWN_PER_CSS 25
static int cgroup_css_links_read(struct cgroup *cont,
				 struct cftype *cft,
				 struct seq_file *seq)
{
	struct cg_cgroup_link *link;

	read_lock(&css_set_lock);
	list_for_each_entry(link, &cont->css_sets, cgrp_link_list) {
		struct css_set *cg = link->cg;
		struct task_struct *task;
		int count = 0;
		seq_printf(seq, "css_set %p\n", cg);
		list_for_each_entry(task, &cg->tasks, cg_list) {
			if (count++ > MAX_TASKS_SHOWN_PER_CSS) {
				seq_puts(seq, "  ...\n");
				break;
			} else {
				seq_printf(seq, "  task %d\n",
					   task_pid_vnr(task));
			}
		}
	}
	read_unlock(&css_set_lock);
	return 0;
}

5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488
static u64 releasable_read(struct cgroup *cgrp, struct cftype *cft)
{
	return test_bit(CGRP_RELEASABLE, &cgrp->flags);
}

static struct cftype debug_files[] =  {
	{
		.name = "cgroup_refcount",
		.read_u64 = cgroup_refcount_read,
	},
	{
		.name = "taskcount",
		.read_u64 = debug_taskcount_read,
	},

	{
		.name = "current_css_set",
		.read_u64 = current_css_set_read,
	},

	{
		.name = "current_css_set_refcount",
		.read_u64 = current_css_set_refcount_read,
	},

5489 5490 5491 5492 5493 5494 5495 5496 5497 5498
	{
		.name = "current_css_set_cg_links",
		.read_seq_string = current_css_set_cg_links_read,
	},

	{
		.name = "cgroup_css_links",
		.read_seq_string = cgroup_css_links_read,
	},

5499 5500 5501 5502 5503
	{
		.name = "releasable",
		.read_u64 = releasable_read,
	},

5504 5505
	{ }	/* terminate */
};
5506 5507 5508 5509 5510 5511

struct cgroup_subsys debug_subsys = {
	.name = "debug",
	.create = debug_create,
	.destroy = debug_destroy,
	.subsys_id = debug_subsys_id,
5512
	.base_cftypes = debug_files,
5513 5514
};
#endif /* CONFIG_CGROUP_DEBUG */